Adding upstream version 4.19.249.upstream/4.19.249upstream

Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>

22 files changed, 23136 insertions, 0 deletions

diff --git a/drivers/infiniband/hw/mlx5/Kconfig b/drivers/infiniband/hw/mlx5/Kconfig
new file mode 100644
index 000000000..0440966bc
--- /dev/null
+++ b/drivers/infiniband/hw/mlx5/Kconfig
@@ -0,0 +1,9 @@
+config MLX5_INFINIBAND
+	tristate "Mellanox 5th generation network adapters (ConnectX series) support"
+	depends on NETDEVICES && ETHERNET && PCI && MLX5_CORE
+	depends on INFINIBAND_USER_ACCESS || INFINIBAND_USER_ACCESS=n
+	---help---
+	  This driver provides low-level InfiniBand support for
+	  Mellanox Connect-IB PCI Express host channel adapters (HCAs).
+	  This is required to use InfiniBand protocols such as
+	  IP-over-IB or SRP with these devices.
diff --git a/drivers/infiniband/hw/mlx5/Makefile b/drivers/infiniband/hw/mlx5/Makefile
new file mode 100644
index 000000000..b8e4b15e2
--- /dev/null
+++ b/drivers/infiniband/hw/mlx5/Makefile
@@ -0,0 +1,7 @@
+obj-$(CONFIG_MLX5_INFINIBAND)	+= mlx5_ib.o
+
+mlx5_ib-y :=	main.o cq.o doorbell.o qp.o mem.o srq.o mr.o ah.o mad.o gsi.o ib_virt.o cmd.o cong.o
+mlx5_ib-$(CONFIG_INFINIBAND_ON_DEMAND_PAGING) += odp.o
+mlx5_ib-$(CONFIG_MLX5_ESWITCH) += ib_rep.o
+mlx5_ib-$(CONFIG_INFINIBAND_USER_ACCESS) += devx.o
+mlx5_ib-$(CONFIG_INFINIBAND_USER_ACCESS) += flow.o
diff --git a/drivers/infiniband/hw/mlx5/ah.c b/drivers/infiniband/hw/mlx5/ah.c
new file mode 100644
index 000000000..ffd03bf1a
--- /dev/null
+++ b/drivers/infiniband/hw/mlx5/ah.c
@@ -0,0 +1,138 @@
+/*
+ * Copyright (c) 2013-2015, Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "mlx5_ib.h"
+
+static struct ib_ah *create_ib_ah(struct mlx5_ib_dev *dev,
+				  struct mlx5_ib_ah *ah,
+				  struct rdma_ah_attr *ah_attr)
+{
+	enum ib_gid_type gid_type;
+
+	if (rdma_ah_get_ah_flags(ah_attr) & IB_AH_GRH) {
+		const struct ib_global_route *grh = rdma_ah_read_grh(ah_attr);
+
+		memcpy(ah->av.rgid, &grh->dgid, 16);
+		ah->av.grh_gid_fl = cpu_to_be32(grh->flow_label |
+						(1 << 30) |
+						grh->sgid_index << 20);
+		ah->av.hop_limit = grh->hop_limit;
+		ah->av.tclass = grh->traffic_class;
+	}
+
+	ah->av.stat_rate_sl = (rdma_ah_get_static_rate(ah_attr) << 4);
+
+	if (ah_attr->type == RDMA_AH_ATTR_TYPE_ROCE) {
+		gid_type = ah_attr->grh.sgid_attr->gid_type;
+
+		memcpy(ah->av.rmac, ah_attr->roce.dmac,
+		       sizeof(ah_attr->roce.dmac));
+		ah->av.udp_sport =
+			mlx5_get_roce_udp_sport(dev, ah_attr->grh.sgid_attr);
+		ah->av.stat_rate_sl |= (rdma_ah_get_sl(ah_attr) & 0x7) << 1;
+		if (gid_type == IB_GID_TYPE_ROCE_UDP_ENCAP)
+#define MLX5_ECN_ENABLED BIT(1)
+			ah->av.tclass |= MLX5_ECN_ENABLED;
+	} else {
+		ah->av.rlid = cpu_to_be16(rdma_ah_get_dlid(ah_attr));
+		ah->av.fl_mlid = rdma_ah_get_path_bits(ah_attr) & 0x7f;
+		ah->av.stat_rate_sl |= (rdma_ah_get_sl(ah_attr) & 0xf);
+	}
+
+	return &ah->ibah;
+}
+
+struct ib_ah *mlx5_ib_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr,
+				struct ib_udata *udata)
+
+{
+	struct mlx5_ib_ah *ah;
+	struct mlx5_ib_dev *dev = to_mdev(pd->device);
+	enum rdma_ah_attr_type ah_type = ah_attr->type;
+
+	if ((ah_type == RDMA_AH_ATTR_TYPE_ROCE) &&
+	    !(rdma_ah_get_ah_flags(ah_attr) & IB_AH_GRH))
+		return ERR_PTR(-EINVAL);
+
+	if (ah_type == RDMA_AH_ATTR_TYPE_ROCE && udata) {
+		int err;
+		struct mlx5_ib_create_ah_resp resp = {};
+		u32 min_resp_len = offsetof(typeof(resp), dmac) +
+				   sizeof(resp.dmac);
+
+		if (udata->outlen < min_resp_len)
+			return ERR_PTR(-EINVAL);
+
+		resp.response_length = min_resp_len;
+
+		memcpy(resp.dmac, ah_attr->roce.dmac, ETH_ALEN);
+		err = ib_copy_to_udata(udata, &resp, resp.response_length);
+		if (err)
+			return ERR_PTR(err);
+	}
+
+	ah = kzalloc(sizeof(*ah), GFP_ATOMIC);
+	if (!ah)
+		return ERR_PTR(-ENOMEM);
+
+	return create_ib_ah(dev, ah, ah_attr); /* never fails */
+}
+
+int mlx5_ib_query_ah(struct ib_ah *ibah, struct rdma_ah_attr *ah_attr)
+{
+	struct mlx5_ib_ah *ah = to_mah(ibah);
+	u32 tmp;
+
+	memset(ah_attr, 0, sizeof(*ah_attr));
+	ah_attr->type = ibah->type;
+
+	tmp = be32_to_cpu(ah->av.grh_gid_fl);
+	if (tmp & (1 << 30)) {
+		rdma_ah_set_grh(ah_attr, NULL,
+				tmp & 0xfffff,
+				(tmp >> 20) & 0xff,
+				ah->av.hop_limit,
+				ah->av.tclass);
+		rdma_ah_set_dgid_raw(ah_attr, ah->av.rgid);
+	}
+	rdma_ah_set_dlid(ah_attr, be16_to_cpu(ah->av.rlid));
+	rdma_ah_set_static_rate(ah_attr, ah->av.stat_rate_sl >> 4);
+	rdma_ah_set_sl(ah_attr, ah->av.stat_rate_sl & 0xf);
+
+	return 0;
+}
+
+int mlx5_ib_destroy_ah(struct ib_ah *ah)
+{
+	kfree(to_mah(ah));
+	return 0;
+}
diff --git a/drivers/infiniband/hw/mlx5/cmd.c b/drivers/infiniband/hw/mlx5/cmd.c
new file mode 100644
index 000000000..c84fef9a8
--- /dev/null
+++ b/drivers/infiniband/hw/mlx5/cmd.c
@@ -0,0 +1,199 @@
+/*
+ * Copyright (c) 2017, Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "cmd.h"
+
+int mlx5_cmd_dump_fill_mkey(struct mlx5_core_dev *dev, u32 *mkey)
+{
+	u32 out[MLX5_ST_SZ_DW(query_special_contexts_out)] = {0};
+	u32 in[MLX5_ST_SZ_DW(query_special_contexts_in)]   = {0};
+	int err;
+
+	MLX5_SET(query_special_contexts_in, in, opcode,
+		 MLX5_CMD_OP_QUERY_SPECIAL_CONTEXTS);
+	err = mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out));
+	if (!err)
+		*mkey = MLX5_GET(query_special_contexts_out, out,
+				 dump_fill_mkey);
+	return err;
+}
+
+int mlx5_cmd_null_mkey(struct mlx5_core_dev *dev, u32 *null_mkey)
+{
+	u32 out[MLX5_ST_SZ_DW(query_special_contexts_out)] = {};
+	u32 in[MLX5_ST_SZ_DW(query_special_contexts_in)]   = {};
+	int err;
+
+	MLX5_SET(query_special_contexts_in, in, opcode,
+		 MLX5_CMD_OP_QUERY_SPECIAL_CONTEXTS);
+	err = mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out));
+	if (!err)
+		*null_mkey = MLX5_GET(query_special_contexts_out, out,
+				      null_mkey);
+	return err;
+}
+
+int mlx5_cmd_query_cong_params(struct mlx5_core_dev *dev, int cong_point,
+			       void *out, int out_size)
+{
+	u32 in[MLX5_ST_SZ_DW(query_cong_params_in)] = { };
+
+	MLX5_SET(query_cong_params_in, in, opcode,
+		 MLX5_CMD_OP_QUERY_CONG_PARAMS);
+	MLX5_SET(query_cong_params_in, in, cong_protocol, cong_point);
+
+	return mlx5_cmd_exec(dev, in, sizeof(in), out, out_size);
+}
+
+int mlx5_cmd_modify_cong_params(struct mlx5_core_dev *dev,
+				void *in, int in_size)
+{
+	u32 out[MLX5_ST_SZ_DW(modify_cong_params_out)] = { };
+
+	return mlx5_cmd_exec(dev, in, in_size, out, sizeof(out));
+}
+
+int mlx5_cmd_alloc_memic(struct mlx5_memic *memic, phys_addr_t *addr,
+			  u64 length, u32 alignment)
+{
+	struct mlx5_core_dev *dev = memic->dev;
+	u64 num_memic_hw_pages = MLX5_CAP_DEV_MEM(dev, memic_bar_size)
+					>> PAGE_SHIFT;
+	u64 hw_start_addr = MLX5_CAP64_DEV_MEM(dev, memic_bar_start_addr);
+	u32 max_alignment = MLX5_CAP_DEV_MEM(dev, log_max_memic_addr_alignment);
+	u32 num_pages = DIV_ROUND_UP(length, PAGE_SIZE);
+	u32 out[MLX5_ST_SZ_DW(alloc_memic_out)] = {};
+	u32 in[MLX5_ST_SZ_DW(alloc_memic_in)] = {};
+	u32 mlx5_alignment;
+	u64 page_idx = 0;
+	int ret = 0;
+
+	if (!length || (length & MLX5_MEMIC_ALLOC_SIZE_MASK))
+		return -EINVAL;
+
+	/* mlx5 device sets alignment as 64*2^driver_value
+	 * so normalizing is needed.
+	 */
+	mlx5_alignment = (alignment < MLX5_MEMIC_BASE_ALIGN) ? 0 :
+			 alignment - MLX5_MEMIC_BASE_ALIGN;
+	if (mlx5_alignment > max_alignment)
+		return -EINVAL;
+
+	MLX5_SET(alloc_memic_in, in, opcode, MLX5_CMD_OP_ALLOC_MEMIC);
+	MLX5_SET(alloc_memic_in, in, range_size, num_pages * PAGE_SIZE);
+	MLX5_SET(alloc_memic_in, in, memic_size, length);
+	MLX5_SET(alloc_memic_in, in, log_memic_addr_alignment,
+		 mlx5_alignment);
+
+	while (page_idx < num_memic_hw_pages) {
+		spin_lock(&memic->memic_lock);
+		page_idx = bitmap_find_next_zero_area(memic->memic_alloc_pages,
+						      num_memic_hw_pages,
+						      page_idx,
+						      num_pages, 0);
+
+		if (page_idx < num_memic_hw_pages)
+			bitmap_set(memic->memic_alloc_pages,
+				   page_idx, num_pages);
+
+		spin_unlock(&memic->memic_lock);
+
+		if (page_idx >= num_memic_hw_pages)
+			break;
+
+		MLX5_SET64(alloc_memic_in, in, range_start_addr,
+			   hw_start_addr + (page_idx * PAGE_SIZE));
+
+		ret = mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out));
+		if (ret) {
+			spin_lock(&memic->memic_lock);
+			bitmap_clear(memic->memic_alloc_pages,
+				     page_idx, num_pages);
+			spin_unlock(&memic->memic_lock);
+
+			if (ret == -EAGAIN) {
+				page_idx++;
+				continue;
+			}
+
+			return ret;
+		}
+
+		*addr = pci_resource_start(dev->pdev, 0) +
+			MLX5_GET64(alloc_memic_out, out, memic_start_addr);
+
+		return 0;
+	}
+
+	return -ENOMEM;
+}
+
+int mlx5_cmd_dealloc_memic(struct mlx5_memic *memic, u64 addr, u64 length)
+{
+	struct mlx5_core_dev *dev = memic->dev;
+	u64 hw_start_addr = MLX5_CAP64_DEV_MEM(dev, memic_bar_start_addr);
+	u32 num_pages = DIV_ROUND_UP(length, PAGE_SIZE);
+	u32 out[MLX5_ST_SZ_DW(dealloc_memic_out)] = {0};
+	u32 in[MLX5_ST_SZ_DW(dealloc_memic_in)] = {0};
+	u64 start_page_idx;
+	int err;
+
+	addr -= pci_resource_start(dev->pdev, 0);
+	start_page_idx = (addr - hw_start_addr) >> PAGE_SHIFT;
+
+	MLX5_SET(dealloc_memic_in, in, opcode, MLX5_CMD_OP_DEALLOC_MEMIC);
+	MLX5_SET64(dealloc_memic_in, in, memic_start_addr, addr);
+	MLX5_SET(dealloc_memic_in, in, memic_size, length);
+
+	err =  mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out));
+
+	if (!err) {
+		spin_lock(&memic->memic_lock);
+		bitmap_clear(memic->memic_alloc_pages,
+			     start_page_idx, num_pages);
+		spin_unlock(&memic->memic_lock);
+	}
+
+	return err;
+}
+
+int mlx5_cmd_query_ext_ppcnt_counters(struct mlx5_core_dev *dev, void *out)
+{
+	u32 in[MLX5_ST_SZ_DW(ppcnt_reg)] = {};
+	int sz = MLX5_ST_SZ_BYTES(ppcnt_reg);
+
+	MLX5_SET(ppcnt_reg, in, local_port, 1);
+
+	MLX5_SET(ppcnt_reg, in, grp, MLX5_ETHERNET_EXTENDED_COUNTERS_GROUP);
+	return  mlx5_core_access_reg(dev, in, sz, out, sz, MLX5_REG_PPCNT,
+				     0, 0);
+}
diff --git a/drivers/infiniband/hw/mlx5/cmd.h b/drivers/infiniband/hw/mlx5/cmd.h
new file mode 100644
index 000000000..88cbb1c41
--- /dev/null
+++ b/drivers/infiniband/hw/mlx5/cmd.h
@@ -0,0 +1,50 @@
+/*
+ * Copyright (c) 2017, Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef MLX5_IB_CMD_H
+#define MLX5_IB_CMD_H
+
+#include "mlx5_ib.h"
+#include <linux/kernel.h>
+#include <linux/mlx5/driver.h>
+
+int mlx5_cmd_dump_fill_mkey(struct mlx5_core_dev *dev, u32 *mkey);
+int mlx5_cmd_null_mkey(struct mlx5_core_dev *dev, u32 *null_mkey);
+int mlx5_cmd_query_cong_params(struct mlx5_core_dev *dev, int cong_point,
+			       void *out, int out_size);
+int mlx5_cmd_query_ext_ppcnt_counters(struct mlx5_core_dev *dev, void *out);
+int mlx5_cmd_modify_cong_params(struct mlx5_core_dev *mdev,
+				void *in, int in_size);
+int mlx5_cmd_alloc_memic(struct mlx5_memic *memic, phys_addr_t *addr,
+			 u64 length, u32 alignment);
+int mlx5_cmd_dealloc_memic(struct mlx5_memic *memic, u64 addr, u64 length);
+#endif /* MLX5_IB_CMD_H */
diff --git a/drivers/infiniband/hw/mlx5/cong.c b/drivers/infiniband/hw/mlx5/cong.c
new file mode 100644
index 000000000..7e4e358a4
--- /dev/null
+++ b/drivers/infiniband/hw/mlx5/cong.c
@@ -0,0 +1,449 @@
+/*
+ * Copyright (c) 2013-2017, Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/debugfs.h>
+
+#include "mlx5_ib.h"
+#include "cmd.h"
+
+enum mlx5_ib_cong_node_type {
+	MLX5_IB_RROCE_ECN_RP = 1,
+	MLX5_IB_RROCE_ECN_NP = 2,
+};
+
+static const char * const mlx5_ib_dbg_cc_name[] = {
+	"rp_clamp_tgt_rate",
+	"rp_clamp_tgt_rate_ati",
+	"rp_time_reset",
+	"rp_byte_reset",
+	"rp_threshold",
+	"rp_ai_rate",
+	"rp_hai_rate",
+	"rp_min_dec_fac",
+	"rp_min_rate",
+	"rp_rate_to_set_on_first_cnp",
+	"rp_dce_tcp_g",
+	"rp_dce_tcp_rtt",
+	"rp_rate_reduce_monitor_period",
+	"rp_initial_alpha_value",
+	"rp_gd",
+	"np_cnp_dscp",
+	"np_cnp_prio_mode",
+	"np_cnp_prio",
+};
+
+#define MLX5_IB_RP_CLAMP_TGT_RATE_ATTR			BIT(1)
+#define MLX5_IB_RP_CLAMP_TGT_RATE_ATI_ATTR		BIT(2)
+#define MLX5_IB_RP_TIME_RESET_ATTR			BIT(3)
+#define MLX5_IB_RP_BYTE_RESET_ATTR			BIT(4)
+#define MLX5_IB_RP_THRESHOLD_ATTR			BIT(5)
+#define MLX5_IB_RP_AI_RATE_ATTR				BIT(7)
+#define MLX5_IB_RP_HAI_RATE_ATTR			BIT(8)
+#define MLX5_IB_RP_MIN_DEC_FAC_ATTR			BIT(9)
+#define MLX5_IB_RP_MIN_RATE_ATTR			BIT(10)
+#define MLX5_IB_RP_RATE_TO_SET_ON_FIRST_CNP_ATTR	BIT(11)
+#define MLX5_IB_RP_DCE_TCP_G_ATTR			BIT(12)
+#define MLX5_IB_RP_DCE_TCP_RTT_ATTR			BIT(13)
+#define MLX5_IB_RP_RATE_REDUCE_MONITOR_PERIOD_ATTR	BIT(14)
+#define MLX5_IB_RP_INITIAL_ALPHA_VALUE_ATTR		BIT(15)
+#define MLX5_IB_RP_GD_ATTR				BIT(16)
+
+#define MLX5_IB_NP_CNP_DSCP_ATTR			BIT(3)
+#define MLX5_IB_NP_CNP_PRIO_MODE_ATTR			BIT(4)
+
+static enum mlx5_ib_cong_node_type
+mlx5_ib_param_to_node(enum mlx5_ib_dbg_cc_types param_offset)
+{
+	if (param_offset >= MLX5_IB_DBG_CC_RP_CLAMP_TGT_RATE &&
+	    param_offset <= MLX5_IB_DBG_CC_RP_GD)
+		return MLX5_IB_RROCE_ECN_RP;
+	else
+		return MLX5_IB_RROCE_ECN_NP;
+}
+
+static u32 mlx5_get_cc_param_val(void *field, int offset)
+{
+	switch (offset) {
+	case MLX5_IB_DBG_CC_RP_CLAMP_TGT_RATE:
+		return MLX5_GET(cong_control_r_roce_ecn_rp, field,
+				clamp_tgt_rate);
+	case MLX5_IB_DBG_CC_RP_CLAMP_TGT_RATE_ATI:
+		return MLX5_GET(cong_control_r_roce_ecn_rp, field,
+				clamp_tgt_rate_after_time_inc);
+	case MLX5_IB_DBG_CC_RP_TIME_RESET:
+		return MLX5_GET(cong_control_r_roce_ecn_rp, field,
+				rpg_time_reset);
+	case MLX5_IB_DBG_CC_RP_BYTE_RESET:
+		return MLX5_GET(cong_control_r_roce_ecn_rp, field,
+				rpg_byte_reset);
+	case MLX5_IB_DBG_CC_RP_THRESHOLD:
+		return MLX5_GET(cong_control_r_roce_ecn_rp, field,
+				rpg_threshold);
+	case MLX5_IB_DBG_CC_RP_AI_RATE:
+		return MLX5_GET(cong_control_r_roce_ecn_rp, field,
+				rpg_ai_rate);
+	case MLX5_IB_DBG_CC_RP_HAI_RATE:
+		return MLX5_GET(cong_control_r_roce_ecn_rp, field,
+				rpg_hai_rate);
+	case MLX5_IB_DBG_CC_RP_MIN_DEC_FAC:
+		return MLX5_GET(cong_control_r_roce_ecn_rp, field,
+				rpg_min_dec_fac);
+	case MLX5_IB_DBG_CC_RP_MIN_RATE:
+		return MLX5_GET(cong_control_r_roce_ecn_rp, field,
+				rpg_min_rate);
+	case MLX5_IB_DBG_CC_RP_RATE_TO_SET_ON_FIRST_CNP:
+		return MLX5_GET(cong_control_r_roce_ecn_rp, field,
+				rate_to_set_on_first_cnp);
+	case MLX5_IB_DBG_CC_RP_DCE_TCP_G:
+		return MLX5_GET(cong_control_r_roce_ecn_rp, field,
+				dce_tcp_g);
+	case MLX5_IB_DBG_CC_RP_DCE_TCP_RTT:
+		return MLX5_GET(cong_control_r_roce_ecn_rp, field,
+				dce_tcp_rtt);
+	case MLX5_IB_DBG_CC_RP_RATE_REDUCE_MONITOR_PERIOD:
+		return MLX5_GET(cong_control_r_roce_ecn_rp, field,
+				rate_reduce_monitor_period);
+	case MLX5_IB_DBG_CC_RP_INITIAL_ALPHA_VALUE:
+		return MLX5_GET(cong_control_r_roce_ecn_rp, field,
+				initial_alpha_value);
+	case MLX5_IB_DBG_CC_RP_GD:
+		return MLX5_GET(cong_control_r_roce_ecn_rp, field,
+				rpg_gd);
+	case MLX5_IB_DBG_CC_NP_CNP_DSCP:
+		return MLX5_GET(cong_control_r_roce_ecn_np, field,
+				cnp_dscp);
+	case MLX5_IB_DBG_CC_NP_CNP_PRIO_MODE:
+		return MLX5_GET(cong_control_r_roce_ecn_np, field,
+				cnp_prio_mode);
+	case MLX5_IB_DBG_CC_NP_CNP_PRIO:
+		return MLX5_GET(cong_control_r_roce_ecn_np, field,
+				cnp_802p_prio);
+	default:
+		return 0;
+	}
+}
+
+static void mlx5_ib_set_cc_param_mask_val(void *field, int offset,
+					  u32 var, u32 *attr_mask)
+{
+	switch (offset) {
+	case MLX5_IB_DBG_CC_RP_CLAMP_TGT_RATE:
+		*attr_mask |= MLX5_IB_RP_CLAMP_TGT_RATE_ATTR;
+		MLX5_SET(cong_control_r_roce_ecn_rp, field,
+			 clamp_tgt_rate, var);
+		break;
+	case MLX5_IB_DBG_CC_RP_CLAMP_TGT_RATE_ATI:
+		*attr_mask |= MLX5_IB_RP_CLAMP_TGT_RATE_ATI_ATTR;
+		MLX5_SET(cong_control_r_roce_ecn_rp, field,
+			 clamp_tgt_rate_after_time_inc, var);
+		break;
+	case MLX5_IB_DBG_CC_RP_TIME_RESET:
+		*attr_mask |= MLX5_IB_RP_TIME_RESET_ATTR;
+		MLX5_SET(cong_control_r_roce_ecn_rp, field,
+			 rpg_time_reset, var);
+		break;
+	case MLX5_IB_DBG_CC_RP_BYTE_RESET:
+		*attr_mask |= MLX5_IB_RP_BYTE_RESET_ATTR;
+		MLX5_SET(cong_control_r_roce_ecn_rp, field,
+			 rpg_byte_reset, var);
+		break;
+	case MLX5_IB_DBG_CC_RP_THRESHOLD:
+		*attr_mask |= MLX5_IB_RP_THRESHOLD_ATTR;
+		MLX5_SET(cong_control_r_roce_ecn_rp, field,
+			 rpg_threshold, var);
+		break;
+	case MLX5_IB_DBG_CC_RP_AI_RATE:
+		*attr_mask |= MLX5_IB_RP_AI_RATE_ATTR;
+		MLX5_SET(cong_control_r_roce_ecn_rp, field,
+			 rpg_ai_rate, var);
+		break;
+	case MLX5_IB_DBG_CC_RP_HAI_RATE:
+		*attr_mask |= MLX5_IB_RP_HAI_RATE_ATTR;
+		MLX5_SET(cong_control_r_roce_ecn_rp, field,
+			 rpg_hai_rate, var);
+		break;
+	case MLX5_IB_DBG_CC_RP_MIN_DEC_FAC:
+		*attr_mask |= MLX5_IB_RP_MIN_DEC_FAC_ATTR;
+		MLX5_SET(cong_control_r_roce_ecn_rp, field,
+			 rpg_min_dec_fac, var);
+		break;
+	case MLX5_IB_DBG_CC_RP_MIN_RATE:
+		*attr_mask |= MLX5_IB_RP_MIN_RATE_ATTR;
+		MLX5_SET(cong_control_r_roce_ecn_rp, field,
+			 rpg_min_rate, var);
+		break;
+	case MLX5_IB_DBG_CC_RP_RATE_TO_SET_ON_FIRST_CNP:
+		*attr_mask |= MLX5_IB_RP_RATE_TO_SET_ON_FIRST_CNP_ATTR;
+		MLX5_SET(cong_control_r_roce_ecn_rp, field,
+			 rate_to_set_on_first_cnp, var);
+		break;
+	case MLX5_IB_DBG_CC_RP_DCE_TCP_G:
+		*attr_mask |= MLX5_IB_RP_DCE_TCP_G_ATTR;
+		MLX5_SET(cong_control_r_roce_ecn_rp, field,
+			 dce_tcp_g, var);
+		break;
+	case MLX5_IB_DBG_CC_RP_DCE_TCP_RTT:
+		*attr_mask |= MLX5_IB_RP_DCE_TCP_RTT_ATTR;
+		MLX5_SET(cong_control_r_roce_ecn_rp, field,
+			 dce_tcp_rtt, var);
+		break;
+	case MLX5_IB_DBG_CC_RP_RATE_REDUCE_MONITOR_PERIOD:
+		*attr_mask |= MLX5_IB_RP_RATE_REDUCE_MONITOR_PERIOD_ATTR;
+		MLX5_SET(cong_control_r_roce_ecn_rp, field,
+			 rate_reduce_monitor_period, var);
+		break;
+	case MLX5_IB_DBG_CC_RP_INITIAL_ALPHA_VALUE:
+		*attr_mask |= MLX5_IB_RP_INITIAL_ALPHA_VALUE_ATTR;
+		MLX5_SET(cong_control_r_roce_ecn_rp, field,
+			 initial_alpha_value, var);
+		break;
+	case MLX5_IB_DBG_CC_RP_GD:
+		*attr_mask |= MLX5_IB_RP_GD_ATTR;
+		MLX5_SET(cong_control_r_roce_ecn_rp, field,
+			 rpg_gd, var);
+		break;
+	case MLX5_IB_DBG_CC_NP_CNP_DSCP:
+		*attr_mask |= MLX5_IB_NP_CNP_DSCP_ATTR;
+		MLX5_SET(cong_control_r_roce_ecn_np, field, cnp_dscp, var);
+		break;
+	case MLX5_IB_DBG_CC_NP_CNP_PRIO_MODE:
+		*attr_mask |= MLX5_IB_NP_CNP_PRIO_MODE_ATTR;
+		MLX5_SET(cong_control_r_roce_ecn_np, field, cnp_prio_mode, var);
+		break;
+	case MLX5_IB_DBG_CC_NP_CNP_PRIO:
+		*attr_mask |= MLX5_IB_NP_CNP_PRIO_MODE_ATTR;
+		MLX5_SET(cong_control_r_roce_ecn_np, field, cnp_prio_mode, 0);
+		MLX5_SET(cong_control_r_roce_ecn_np, field, cnp_802p_prio, var);
+		break;
+	}
+}
+
+static int mlx5_ib_get_cc_params(struct mlx5_ib_dev *dev, u8 port_num,
+				 int offset, u32 *var)
+{
+	int outlen = MLX5_ST_SZ_BYTES(query_cong_params_out);
+	void *out;
+	void *field;
+	int err;
+	enum mlx5_ib_cong_node_type node;
+	struct mlx5_core_dev *mdev;
+
+	/* Takes a 1-based port number */
+	mdev = mlx5_ib_get_native_port_mdev(dev, port_num + 1, NULL);
+	if (!mdev)
+		return -ENODEV;
+
+	out = kvzalloc(outlen, GFP_KERNEL);
+	if (!out) {
+		err = -ENOMEM;
+		goto alloc_err;
+	}
+
+	node = mlx5_ib_param_to_node(offset);
+
+	err = mlx5_cmd_query_cong_params(mdev, node, out, outlen);
+	if (err)
+		goto free;
+
+	field = MLX5_ADDR_OF(query_cong_params_out, out, congestion_parameters);
+	*var = mlx5_get_cc_param_val(field, offset);
+
+free:
+	kvfree(out);
+alloc_err:
+	mlx5_ib_put_native_port_mdev(dev, port_num + 1);
+	return err;
+}
+
+static int mlx5_ib_set_cc_params(struct mlx5_ib_dev *dev, u8 port_num,
+				 int offset, u32 var)
+{
+	int inlen = MLX5_ST_SZ_BYTES(modify_cong_params_in);
+	void *in;
+	void *field;
+	enum mlx5_ib_cong_node_type node;
+	struct mlx5_core_dev *mdev;
+	u32 attr_mask = 0;
+	int err;
+
+	/* Takes a 1-based port number */
+	mdev = mlx5_ib_get_native_port_mdev(dev, port_num + 1, NULL);
+	if (!mdev)
+		return -ENODEV;
+
+	in = kvzalloc(inlen, GFP_KERNEL);
+	if (!in) {
+		err = -ENOMEM;
+		goto alloc_err;
+	}
+
+	MLX5_SET(modify_cong_params_in, in, opcode,
+		 MLX5_CMD_OP_MODIFY_CONG_PARAMS);
+
+	node = mlx5_ib_param_to_node(offset);
+	MLX5_SET(modify_cong_params_in, in, cong_protocol, node);
+
+	field = MLX5_ADDR_OF(modify_cong_params_in, in, congestion_parameters);
+	mlx5_ib_set_cc_param_mask_val(field, offset, var, &attr_mask);
+
+	field = MLX5_ADDR_OF(modify_cong_params_in, in, field_select);
+	MLX5_SET(field_select_r_roce_rp, field, field_select_r_roce_rp,
+		 attr_mask);
+
+	err = mlx5_cmd_modify_cong_params(mdev, in, inlen);
+	kvfree(in);
+alloc_err:
+	mlx5_ib_put_native_port_mdev(dev, port_num + 1);
+	return err;
+}
+
+static ssize_t set_param(struct file *filp, const char __user *buf,
+			 size_t count, loff_t *pos)
+{
+	struct mlx5_ib_dbg_param *param = filp->private_data;
+	int offset = param->offset;
+	char lbuf[11] = { };
+	u32 var;
+	int ret;
+
+	if (count > sizeof(lbuf))
+		return -EINVAL;
+
+	if (copy_from_user(lbuf, buf, count))
+		return -EFAULT;
+
+	lbuf[sizeof(lbuf) - 1] = '\0';
+
+	if (kstrtou32(lbuf, 0, &var))
+		return -EINVAL;
+
+	ret = mlx5_ib_set_cc_params(param->dev, param->port_num, offset, var);
+	return ret ? ret : count;
+}
+
+static ssize_t get_param(struct file *filp, char __user *buf, size_t count,
+			 loff_t *pos)
+{
+	struct mlx5_ib_dbg_param *param = filp->private_data;
+	int offset = param->offset;
+	u32 var = 0;
+	int ret;
+	char lbuf[11];
+
+	ret = mlx5_ib_get_cc_params(param->dev, param->port_num, offset, &var);
+	if (ret)
+		return ret;
+
+	ret = snprintf(lbuf, sizeof(lbuf), "%d\n", var);
+	if (ret < 0)
+		return ret;
+
+	return simple_read_from_buffer(buf, count, pos, lbuf, ret);
+}
+
+static const struct file_operations dbg_cc_fops = {
+	.owner	= THIS_MODULE,
+	.open	= simple_open,
+	.write	= set_param,
+	.read	= get_param,
+};
+
+void mlx5_ib_cleanup_cong_debugfs(struct mlx5_ib_dev *dev, u8 port_num)
+{
+	if (!mlx5_debugfs_root ||
+	    !dev->port[port_num].dbg_cc_params ||
+	    !dev->port[port_num].dbg_cc_params->root)
+		return;
+
+	debugfs_remove_recursive(dev->port[port_num].dbg_cc_params->root);
+	kfree(dev->port[port_num].dbg_cc_params);
+	dev->port[port_num].dbg_cc_params = NULL;
+}
+
+int mlx5_ib_init_cong_debugfs(struct mlx5_ib_dev *dev, u8 port_num)
+{
+	struct mlx5_ib_dbg_cc_params *dbg_cc_params;
+	struct mlx5_core_dev *mdev;
+	int i;
+
+	if (!mlx5_debugfs_root)
+		goto out;
+
+	/* Takes a 1-based port number */
+	mdev = mlx5_ib_get_native_port_mdev(dev, port_num + 1, NULL);
+	if (!mdev)
+		goto out;
+
+	if (!MLX5_CAP_GEN(mdev, cc_query_allowed) ||
+	    !MLX5_CAP_GEN(mdev, cc_modify_allowed))
+		goto put_mdev;
+
+	dbg_cc_params = kzalloc(sizeof(*dbg_cc_params), GFP_KERNEL);
+	if (!dbg_cc_params)
+		goto err;
+
+	dev->port[port_num].dbg_cc_params = dbg_cc_params;
+
+	dbg_cc_params->root = debugfs_create_dir("cc_params",
+						 mdev->priv.dbg_root);
+	if (!dbg_cc_params->root)
+		goto err;
+
+	for (i = 0; i < MLX5_IB_DBG_CC_MAX; i++) {
+		dbg_cc_params->params[i].offset = i;
+		dbg_cc_params->params[i].dev = dev;
+		dbg_cc_params->params[i].port_num = port_num;
+		dbg_cc_params->params[i].dentry =
+			debugfs_create_file(mlx5_ib_dbg_cc_name[i],
+					    0600, dbg_cc_params->root,
+					    &dbg_cc_params->params[i],
+					    &dbg_cc_fops);
+		if (!dbg_cc_params->params[i].dentry)
+			goto err;
+	}
+
+put_mdev:
+	mlx5_ib_put_native_port_mdev(dev, port_num + 1);
+out:
+	return 0;
+
+err:
+	mlx5_ib_warn(dev, "cong debugfs failure\n");
+	mlx5_ib_cleanup_cong_debugfs(dev, port_num);
+	mlx5_ib_put_native_port_mdev(dev, port_num + 1);
+
+	/*
+	 * We don't want to fail driver if debugfs failed to initialize,
+	 * so we are not forwarding error to the user.
+	 */
+	return 0;
+}
diff --git a/drivers/infiniband/hw/mlx5/cq.c b/drivers/infiniband/hw/mlx5/cq.c
new file mode 100644
index 000000000..872985e4e
--- /dev/null
+++ b/drivers/infiniband/hw/mlx5/cq.c
@@ -0,0 +1,1494 @@
+/*
+ * Copyright (c) 2013-2015, Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/kref.h>
+#include <rdma/ib_umem.h>
+#include <rdma/ib_user_verbs.h>
+#include <rdma/ib_cache.h>
+#include "mlx5_ib.h"
+
+static void mlx5_ib_cq_comp(struct mlx5_core_cq *cq)
+{
+	struct ib_cq *ibcq = &to_mibcq(cq)->ibcq;
+
+	ibcq->comp_handler(ibcq, ibcq->cq_context);
+}
+
+static void mlx5_ib_cq_event(struct mlx5_core_cq *mcq, enum mlx5_event type)
+{
+	struct mlx5_ib_cq *cq = container_of(mcq, struct mlx5_ib_cq, mcq);
+	struct mlx5_ib_dev *dev = to_mdev(cq->ibcq.device);
+	struct ib_cq *ibcq = &cq->ibcq;
+	struct ib_event event;
+
+	if (type != MLX5_EVENT_TYPE_CQ_ERROR) {
+		mlx5_ib_warn(dev, "Unexpected event type %d on CQ %06x\n",
+			     type, mcq->cqn);
+		return;
+	}
+
+	if (ibcq->event_handler) {
+		event.device     = &dev->ib_dev;
+		event.event      = IB_EVENT_CQ_ERR;
+		event.element.cq = ibcq;
+		ibcq->event_handler(&event, ibcq->cq_context);
+	}
+}
+
+static void *get_cqe(struct mlx5_ib_cq *cq, int n)
+{
+	return mlx5_frag_buf_get_wqe(&cq->buf.fbc, n);
+}
+
+static u8 sw_ownership_bit(int n, int nent)
+{
+	return (n & nent) ? 1 : 0;
+}
+
+static void *get_sw_cqe(struct mlx5_ib_cq *cq, int n)
+{
+	void *cqe = get_cqe(cq, n & cq->ibcq.cqe);
+	struct mlx5_cqe64 *cqe64;
+
+	cqe64 = (cq->mcq.cqe_sz == 64) ? cqe : cqe + 64;
+
+	if (likely((cqe64->op_own) >> 4 != MLX5_CQE_INVALID) &&
+	    !((cqe64->op_own & MLX5_CQE_OWNER_MASK) ^ !!(n & (cq->ibcq.cqe + 1)))) {
+		return cqe;
+	} else {
+		return NULL;
+	}
+}
+
+static void *next_cqe_sw(struct mlx5_ib_cq *cq)
+{
+	return get_sw_cqe(cq, cq->mcq.cons_index);
+}
+
+static enum ib_wc_opcode get_umr_comp(struct mlx5_ib_wq *wq, int idx)
+{
+	switch (wq->wr_data[idx]) {
+	case MLX5_IB_WR_UMR:
+		return 0;
+
+	case IB_WR_LOCAL_INV:
+		return IB_WC_LOCAL_INV;
+
+	case IB_WR_REG_MR:
+		return IB_WC_REG_MR;
+
+	default:
+		pr_warn("unknown completion status\n");
+		return 0;
+	}
+}
+
+static void handle_good_req(struct ib_wc *wc, struct mlx5_cqe64 *cqe,
+			    struct mlx5_ib_wq *wq, int idx)
+{
+	wc->wc_flags = 0;
+	switch (be32_to_cpu(cqe->sop_drop_qpn) >> 24) {
+	case MLX5_OPCODE_RDMA_WRITE_IMM:
+		wc->wc_flags |= IB_WC_WITH_IMM;
+		/* fall through */
+	case MLX5_OPCODE_RDMA_WRITE:
+		wc->opcode    = IB_WC_RDMA_WRITE;
+		break;
+	case MLX5_OPCODE_SEND_IMM:
+		wc->wc_flags |= IB_WC_WITH_IMM;
+		/* fall through */
+	case MLX5_OPCODE_SEND:
+	case MLX5_OPCODE_SEND_INVAL:
+		wc->opcode    = IB_WC_SEND;
+		break;
+	case MLX5_OPCODE_RDMA_READ:
+		wc->opcode    = IB_WC_RDMA_READ;
+		wc->byte_len  = be32_to_cpu(cqe->byte_cnt);
+		break;
+	case MLX5_OPCODE_ATOMIC_CS:
+		wc->opcode    = IB_WC_COMP_SWAP;
+		wc->byte_len  = 8;
+		break;
+	case MLX5_OPCODE_ATOMIC_FA:
+		wc->opcode    = IB_WC_FETCH_ADD;
+		wc->byte_len  = 8;
+		break;
+	case MLX5_OPCODE_ATOMIC_MASKED_CS:
+		wc->opcode    = IB_WC_MASKED_COMP_SWAP;
+		wc->byte_len  = 8;
+		break;
+	case MLX5_OPCODE_ATOMIC_MASKED_FA:
+		wc->opcode    = IB_WC_MASKED_FETCH_ADD;
+		wc->byte_len  = 8;
+		break;
+	case MLX5_OPCODE_UMR:
+		wc->opcode = get_umr_comp(wq, idx);
+		break;
+	}
+}
+
+enum {
+	MLX5_GRH_IN_BUFFER = 1,
+	MLX5_GRH_IN_CQE	   = 2,
+};
+
+static void handle_responder(struct ib_wc *wc, struct mlx5_cqe64 *cqe,
+			     struct mlx5_ib_qp *qp)
+{
+	enum rdma_link_layer ll = rdma_port_get_link_layer(qp->ibqp.device, 1);
+	struct mlx5_ib_dev *dev = to_mdev(qp->ibqp.device);
+	struct mlx5_ib_srq *srq;
+	struct mlx5_ib_wq *wq;
+	u16 wqe_ctr;
+	u8  roce_packet_type;
+	bool vlan_present;
+	u8 g;
+
+	if (qp->ibqp.srq || qp->ibqp.xrcd) {
+		struct mlx5_core_srq *msrq = NULL;
+
+		if (qp->ibqp.xrcd) {
+			msrq = mlx5_core_get_srq(dev->mdev,
+						 be32_to_cpu(cqe->srqn));
+			srq = to_mibsrq(msrq);
+		} else {
+			srq = to_msrq(qp->ibqp.srq);
+		}
+		if (srq) {
+			wqe_ctr = be16_to_cpu(cqe->wqe_counter);
+			wc->wr_id = srq->wrid[wqe_ctr];
+			mlx5_ib_free_srq_wqe(srq, wqe_ctr);
+			if (msrq && atomic_dec_and_test(&msrq->refcount))
+				complete(&msrq->free);
+		}
+	} else {
+		wq	  = &qp->rq;
+		wc->wr_id = wq->wrid[wq->tail & (wq->wqe_cnt - 1)];
+		++wq->tail;
+	}
+	wc->byte_len = be32_to_cpu(cqe->byte_cnt);
+
+	switch (cqe->op_own >> 4) {
+	case MLX5_CQE_RESP_WR_IMM:
+		wc->opcode	= IB_WC_RECV_RDMA_WITH_IMM;
+		wc->wc_flags	= IB_WC_WITH_IMM;
+		wc->ex.imm_data = cqe->imm_inval_pkey;
+		break;
+	case MLX5_CQE_RESP_SEND:
+		wc->opcode   = IB_WC_RECV;
+		wc->wc_flags = IB_WC_IP_CSUM_OK;
+		if (unlikely(!((cqe->hds_ip_ext & CQE_L3_OK) &&
+			       (cqe->hds_ip_ext & CQE_L4_OK))))
+			wc->wc_flags = 0;
+		break;
+	case MLX5_CQE_RESP_SEND_IMM:
+		wc->opcode	= IB_WC_RECV;
+		wc->wc_flags	= IB_WC_WITH_IMM;
+		wc->ex.imm_data = cqe->imm_inval_pkey;
+		break;
+	case MLX5_CQE_RESP_SEND_INV:
+		wc->opcode	= IB_WC_RECV;
+		wc->wc_flags	= IB_WC_WITH_INVALIDATE;
+		wc->ex.invalidate_rkey = be32_to_cpu(cqe->imm_inval_pkey);
+		break;
+	}
+	wc->src_qp	   = be32_to_cpu(cqe->flags_rqpn) & 0xffffff;
+	wc->dlid_path_bits = cqe->ml_path;
+	g = (be32_to_cpu(cqe->flags_rqpn) >> 28) & 3;
+	wc->wc_flags |= g ? IB_WC_GRH : 0;
+	if (unlikely(is_qp1(qp->ibqp.qp_type))) {
+		u16 pkey = be32_to_cpu(cqe->imm_inval_pkey) & 0xffff;
+
+		ib_find_cached_pkey(&dev->ib_dev, qp->port, pkey,
+				    &wc->pkey_index);
+	} else {
+		wc->pkey_index = 0;
+	}
+
+	if (ll != IB_LINK_LAYER_ETHERNET) {
+		wc->slid = be16_to_cpu(cqe->slid);
+		wc->sl = (be32_to_cpu(cqe->flags_rqpn) >> 24) & 0xf;
+		return;
+	}
+
+	wc->slid = 0;
+	vlan_present = cqe->l4_l3_hdr_type & 0x1;
+	roce_packet_type   = (be32_to_cpu(cqe->flags_rqpn) >> 24) & 0x3;
+	if (vlan_present) {
+		wc->vlan_id = (be16_to_cpu(cqe->vlan_info)) & 0xfff;
+		wc->sl = (be16_to_cpu(cqe->vlan_info) >> 13) & 0x7;
+		wc->wc_flags |= IB_WC_WITH_VLAN;
+	} else {
+		wc->sl = 0;
+	}
+
+	switch (roce_packet_type) {
+	case MLX5_CQE_ROCE_L3_HEADER_TYPE_GRH:
+		wc->network_hdr_type = RDMA_NETWORK_IB;
+		break;
+	case MLX5_CQE_ROCE_L3_HEADER_TYPE_IPV6:
+		wc->network_hdr_type = RDMA_NETWORK_IPV6;
+		break;
+	case MLX5_CQE_ROCE_L3_HEADER_TYPE_IPV4:
+		wc->network_hdr_type = RDMA_NETWORK_IPV4;
+		break;
+	}
+	wc->wc_flags |= IB_WC_WITH_NETWORK_HDR_TYPE;
+}
+
+static void dump_cqe(struct mlx5_ib_dev *dev, struct mlx5_err_cqe *cqe)
+{
+	mlx5_ib_warn(dev, "dump error cqe\n");
+	mlx5_dump_err_cqe(dev->mdev, cqe);
+}
+
+static void mlx5_handle_error_cqe(struct mlx5_ib_dev *dev,
+				  struct mlx5_err_cqe *cqe,
+				  struct ib_wc *wc)
+{
+	int dump = 1;
+
+	switch (cqe->syndrome) {
+	case MLX5_CQE_SYNDROME_LOCAL_LENGTH_ERR:
+		wc->status = IB_WC_LOC_LEN_ERR;
+		break;
+	case MLX5_CQE_SYNDROME_LOCAL_QP_OP_ERR:
+		wc->status = IB_WC_LOC_QP_OP_ERR;
+		break;
+	case MLX5_CQE_SYNDROME_LOCAL_PROT_ERR:
+		wc->status = IB_WC_LOC_PROT_ERR;
+		break;
+	case MLX5_CQE_SYNDROME_WR_FLUSH_ERR:
+		dump = 0;
+		wc->status = IB_WC_WR_FLUSH_ERR;
+		break;
+	case MLX5_CQE_SYNDROME_MW_BIND_ERR:
+		wc->status = IB_WC_MW_BIND_ERR;
+		break;
+	case MLX5_CQE_SYNDROME_BAD_RESP_ERR:
+		wc->status = IB_WC_BAD_RESP_ERR;
+		break;
+	case MLX5_CQE_SYNDROME_LOCAL_ACCESS_ERR:
+		wc->status = IB_WC_LOC_ACCESS_ERR;
+		break;
+	case MLX5_CQE_SYNDROME_REMOTE_INVAL_REQ_ERR:
+		wc->status = IB_WC_REM_INV_REQ_ERR;
+		break;
+	case MLX5_CQE_SYNDROME_REMOTE_ACCESS_ERR:
+		wc->status = IB_WC_REM_ACCESS_ERR;
+		break;
+	case MLX5_CQE_SYNDROME_REMOTE_OP_ERR:
+		wc->status = IB_WC_REM_OP_ERR;
+		break;
+	case MLX5_CQE_SYNDROME_TRANSPORT_RETRY_EXC_ERR:
+		wc->status = IB_WC_RETRY_EXC_ERR;
+		dump = 0;
+		break;
+	case MLX5_CQE_SYNDROME_RNR_RETRY_EXC_ERR:
+		wc->status = IB_WC_RNR_RETRY_EXC_ERR;
+		dump = 0;
+		break;
+	case MLX5_CQE_SYNDROME_REMOTE_ABORTED_ERR:
+		wc->status = IB_WC_REM_ABORT_ERR;
+		break;
+	default:
+		wc->status = IB_WC_GENERAL_ERR;
+		break;
+	}
+
+	wc->vendor_err = cqe->vendor_err_synd;
+	if (dump)
+		dump_cqe(dev, cqe);
+}
+
+static int is_atomic_response(struct mlx5_ib_qp *qp, uint16_t idx)
+{
+	/* TBD: waiting decision
+	*/
+	return 0;
+}
+
+static void *mlx5_get_atomic_laddr(struct mlx5_ib_qp *qp, uint16_t idx)
+{
+	struct mlx5_wqe_data_seg *dpseg;
+	void *addr;
+
+	dpseg = mlx5_get_send_wqe(qp, idx) + sizeof(struct mlx5_wqe_ctrl_seg) +
+		sizeof(struct mlx5_wqe_raddr_seg) +
+		sizeof(struct mlx5_wqe_atomic_seg);
+	addr = (void *)(unsigned long)be64_to_cpu(dpseg->addr);
+	return addr;
+}
+
+static void handle_atomic(struct mlx5_ib_qp *qp, struct mlx5_cqe64 *cqe64,
+			  uint16_t idx)
+{
+	void *addr;
+	int byte_count;
+	int i;
+
+	if (!is_atomic_response(qp, idx))
+		return;
+
+	byte_count = be32_to_cpu(cqe64->byte_cnt);
+	addr = mlx5_get_atomic_laddr(qp, idx);
+
+	if (byte_count == 4) {
+		*(uint32_t *)addr = be32_to_cpu(*((__be32 *)addr));
+	} else {
+		for (i = 0; i < byte_count; i += 8) {
+			*(uint64_t *)addr = be64_to_cpu(*((__be64 *)addr));
+			addr += 8;
+		}
+	}
+
+	return;
+}
+
+static void handle_atomics(struct mlx5_ib_qp *qp, struct mlx5_cqe64 *cqe64,
+			   u16 tail, u16 head)
+{
+	u16 idx;
+
+	do {
+		idx = tail & (qp->sq.wqe_cnt - 1);
+		handle_atomic(qp, cqe64, idx);
+		if (idx == head)
+			break;
+
+		tail = qp->sq.w_list[idx].next;
+	} while (1);
+	tail = qp->sq.w_list[idx].next;
+	qp->sq.last_poll = tail;
+}
+
+static void free_cq_buf(struct mlx5_ib_dev *dev, struct mlx5_ib_cq_buf *buf)
+{
+	mlx5_frag_buf_free(dev->mdev, &buf->fbc.frag_buf);
+}
+
+static void get_sig_err_item(struct mlx5_sig_err_cqe *cqe,
+			     struct ib_sig_err *item)
+{
+	u16 syndrome = be16_to_cpu(cqe->syndrome);
+
+#define GUARD_ERR   (1 << 13)
+#define APPTAG_ERR  (1 << 12)
+#define REFTAG_ERR  (1 << 11)
+
+	if (syndrome & GUARD_ERR) {
+		item->err_type = IB_SIG_BAD_GUARD;
+		item->expected = be32_to_cpu(cqe->expected_trans_sig) >> 16;
+		item->actual = be32_to_cpu(cqe->actual_trans_sig) >> 16;
+	} else
+	if (syndrome & REFTAG_ERR) {
+		item->err_type = IB_SIG_BAD_REFTAG;
+		item->expected = be32_to_cpu(cqe->expected_reftag);
+		item->actual = be32_to_cpu(cqe->actual_reftag);
+	} else
+	if (syndrome & APPTAG_ERR) {
+		item->err_type = IB_SIG_BAD_APPTAG;
+		item->expected = be32_to_cpu(cqe->expected_trans_sig) & 0xffff;
+		item->actual = be32_to_cpu(cqe->actual_trans_sig) & 0xffff;
+	} else {
+		pr_err("Got signature completion error with bad syndrome %04x\n",
+		       syndrome);
+	}
+
+	item->sig_err_offset = be64_to_cpu(cqe->err_offset);
+	item->key = be32_to_cpu(cqe->mkey);
+}
+
+static void sw_send_comp(struct mlx5_ib_qp *qp, int num_entries,
+			 struct ib_wc *wc, int *npolled)
+{
+	struct mlx5_ib_wq *wq;
+	unsigned int cur;
+	unsigned int idx;
+	int np;
+	int i;
+
+	wq = &qp->sq;
+	cur = wq->head - wq->tail;
+	np = *npolled;
+
+	if (cur == 0)
+		return;
+
+	for (i = 0;  i < cur && np < num_entries; i++) {
+		idx = wq->last_poll & (wq->wqe_cnt - 1);
+		wc->wr_id = wq->wrid[idx];
+		wc->status = IB_WC_WR_FLUSH_ERR;
+		wc->vendor_err = MLX5_CQE_SYNDROME_WR_FLUSH_ERR;
+		wq->tail++;
+		np++;
+		wc->qp = &qp->ibqp;
+		wc++;
+		wq->last_poll = wq->w_list[idx].next;
+	}
+	*npolled = np;
+}
+
+static void sw_recv_comp(struct mlx5_ib_qp *qp, int num_entries,
+			 struct ib_wc *wc, int *npolled)
+{
+	struct mlx5_ib_wq *wq;
+	unsigned int cur;
+	int np;
+	int i;
+
+	wq = &qp->rq;
+	cur = wq->head - wq->tail;
+	np = *npolled;
+
+	if (cur == 0)
+		return;
+
+	for (i = 0;  i < cur && np < num_entries; i++) {
+		wc->wr_id = wq->wrid[wq->tail & (wq->wqe_cnt - 1)];
+		wc->status = IB_WC_WR_FLUSH_ERR;
+		wc->vendor_err = MLX5_CQE_SYNDROME_WR_FLUSH_ERR;
+		wq->tail++;
+		np++;
+		wc->qp = &qp->ibqp;
+		wc++;
+	}
+	*npolled = np;
+}
+
+static void mlx5_ib_poll_sw_comp(struct mlx5_ib_cq *cq, int num_entries,
+				 struct ib_wc *wc, int *npolled)
+{
+	struct mlx5_ib_qp *qp;
+
+	*npolled = 0;
+	/* Find uncompleted WQEs belonging to that cq and return mmics ones */
+	list_for_each_entry(qp, &cq->list_send_qp, cq_send_list) {
+		sw_send_comp(qp, num_entries, wc + *npolled, npolled);
+		if (*npolled >= num_entries)
+			return;
+	}
+
+	list_for_each_entry(qp, &cq->list_recv_qp, cq_recv_list) {
+		sw_recv_comp(qp, num_entries, wc + *npolled, npolled);
+		if (*npolled >= num_entries)
+			return;
+	}
+}
+
+static int mlx5_poll_one(struct mlx5_ib_cq *cq,
+			 struct mlx5_ib_qp **cur_qp,
+			 struct ib_wc *wc)
+{
+	struct mlx5_ib_dev *dev = to_mdev(cq->ibcq.device);
+	struct mlx5_err_cqe *err_cqe;
+	struct mlx5_cqe64 *cqe64;
+	struct mlx5_core_qp *mqp;
+	struct mlx5_ib_wq *wq;
+	struct mlx5_sig_err_cqe *sig_err_cqe;
+	struct mlx5_core_mkey *mmkey;
+	struct mlx5_ib_mr *mr;
+	uint8_t opcode;
+	uint32_t qpn;
+	u16 wqe_ctr;
+	void *cqe;
+	int idx;
+
+repoll:
+	cqe = next_cqe_sw(cq);
+	if (!cqe)
+		return -EAGAIN;
+
+	cqe64 = (cq->mcq.cqe_sz == 64) ? cqe : cqe + 64;
+
+	++cq->mcq.cons_index;
+
+	/* Make sure we read CQ entry contents after we've checked the
+	 * ownership bit.
+	 */
+	rmb();
+
+	opcode = cqe64->op_own >> 4;
+	if (unlikely(opcode == MLX5_CQE_RESIZE_CQ)) {
+		if (likely(cq->resize_buf)) {
+			free_cq_buf(dev, &cq->buf);
+			cq->buf = *cq->resize_buf;
+			kfree(cq->resize_buf);
+			cq->resize_buf = NULL;
+			goto repoll;
+		} else {
+			mlx5_ib_warn(dev, "unexpected resize cqe\n");
+		}
+	}
+
+	qpn = ntohl(cqe64->sop_drop_qpn) & 0xffffff;
+	if (!*cur_qp || (qpn != (*cur_qp)->ibqp.qp_num)) {
+		/* We do not have to take the QP table lock here,
+		 * because CQs will be locked while QPs are removed
+		 * from the table.
+		 */
+		mqp = __mlx5_qp_lookup(dev->mdev, qpn);
+		*cur_qp = to_mibqp(mqp);
+	}
+
+	wc->qp  = &(*cur_qp)->ibqp;
+	switch (opcode) {
+	case MLX5_CQE_REQ:
+		wq = &(*cur_qp)->sq;
+		wqe_ctr = be16_to_cpu(cqe64->wqe_counter);
+		idx = wqe_ctr & (wq->wqe_cnt - 1);
+		handle_good_req(wc, cqe64, wq, idx);
+		handle_atomics(*cur_qp, cqe64, wq->last_poll, idx);
+		wc->wr_id = wq->wrid[idx];
+		wq->tail = wq->wqe_head[idx] + 1;
+		wc->status = IB_WC_SUCCESS;
+		break;
+	case MLX5_CQE_RESP_WR_IMM:
+	case MLX5_CQE_RESP_SEND:
+	case MLX5_CQE_RESP_SEND_IMM:
+	case MLX5_CQE_RESP_SEND_INV:
+		handle_responder(wc, cqe64, *cur_qp);
+		wc->status = IB_WC_SUCCESS;
+		break;
+	case MLX5_CQE_RESIZE_CQ:
+		break;
+	case MLX5_CQE_REQ_ERR:
+	case MLX5_CQE_RESP_ERR:
+		err_cqe = (struct mlx5_err_cqe *)cqe64;
+		mlx5_handle_error_cqe(dev, err_cqe, wc);
+		mlx5_ib_dbg(dev, "%s error cqe on cqn 0x%x:\n",
+			    opcode == MLX5_CQE_REQ_ERR ?
+			    "Requestor" : "Responder", cq->mcq.cqn);
+		mlx5_ib_dbg(dev, "syndrome 0x%x, vendor syndrome 0x%x\n",
+			    err_cqe->syndrome, err_cqe->vendor_err_synd);
+		if (opcode == MLX5_CQE_REQ_ERR) {
+			wq = &(*cur_qp)->sq;
+			wqe_ctr = be16_to_cpu(cqe64->wqe_counter);
+			idx = wqe_ctr & (wq->wqe_cnt - 1);
+			wc->wr_id = wq->wrid[idx];
+			wq->tail = wq->wqe_head[idx] + 1;
+		} else {
+			struct mlx5_ib_srq *srq;
+
+			if ((*cur_qp)->ibqp.srq) {
+				srq = to_msrq((*cur_qp)->ibqp.srq);
+				wqe_ctr = be16_to_cpu(cqe64->wqe_counter);
+				wc->wr_id = srq->wrid[wqe_ctr];
+				mlx5_ib_free_srq_wqe(srq, wqe_ctr);
+			} else {
+				wq = &(*cur_qp)->rq;
+				wc->wr_id = wq->wrid[wq->tail & (wq->wqe_cnt - 1)];
+				++wq->tail;
+			}
+		}
+		break;
+	case MLX5_CQE_SIG_ERR:
+		sig_err_cqe = (struct mlx5_sig_err_cqe *)cqe64;
+
+		read_lock(&dev->mdev->priv.mkey_table.lock);
+		mmkey = __mlx5_mr_lookup(dev->mdev,
+					 mlx5_base_mkey(be32_to_cpu(sig_err_cqe->mkey)));
+		mr = to_mibmr(mmkey);
+		get_sig_err_item(sig_err_cqe, &mr->sig->err_item);
+		mr->sig->sig_err_exists = true;
+		mr->sig->sigerr_count++;
+
+		mlx5_ib_warn(dev, "CQN: 0x%x Got SIGERR on key: 0x%x err_type %x err_offset %llx expected %x actual %x\n",
+			     cq->mcq.cqn, mr->sig->err_item.key,
+			     mr->sig->err_item.err_type,
+			     mr->sig->err_item.sig_err_offset,
+			     mr->sig->err_item.expected,
+			     mr->sig->err_item.actual);
+
+		read_unlock(&dev->mdev->priv.mkey_table.lock);
+		goto repoll;
+	}
+
+	return 0;
+}
+
+static int poll_soft_wc(struct mlx5_ib_cq *cq, int num_entries,
+			struct ib_wc *wc, bool is_fatal_err)
+{
+	struct mlx5_ib_dev *dev = to_mdev(cq->ibcq.device);
+	struct mlx5_ib_wc *soft_wc, *next;
+	int npolled = 0;
+
+	list_for_each_entry_safe(soft_wc, next, &cq->wc_list, list) {
+		if (npolled >= num_entries)
+			break;
+
+		mlx5_ib_dbg(dev, "polled software generated completion on CQ 0x%x\n",
+			    cq->mcq.cqn);
+
+		if (unlikely(is_fatal_err)) {
+			soft_wc->wc.status = IB_WC_WR_FLUSH_ERR;
+			soft_wc->wc.vendor_err = MLX5_CQE_SYNDROME_WR_FLUSH_ERR;
+		}
+		wc[npolled++] = soft_wc->wc;
+		list_del(&soft_wc->list);
+		kfree(soft_wc);
+	}
+
+	return npolled;
+}
+
+int mlx5_ib_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *wc)
+{
+	struct mlx5_ib_cq *cq = to_mcq(ibcq);
+	struct mlx5_ib_qp *cur_qp = NULL;
+	struct mlx5_ib_dev *dev = to_mdev(cq->ibcq.device);
+	struct mlx5_core_dev *mdev = dev->mdev;
+	unsigned long flags;
+	int soft_polled = 0;
+	int npolled;
+
+	spin_lock_irqsave(&cq->lock, flags);
+	if (mdev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR) {
+		/* make sure no soft wqe's are waiting */
+		if (unlikely(!list_empty(&cq->wc_list)))
+			soft_polled = poll_soft_wc(cq, num_entries, wc, true);
+
+		mlx5_ib_poll_sw_comp(cq, num_entries - soft_polled,
+				     wc + soft_polled, &npolled);
+		goto out;
+	}
+
+	if (unlikely(!list_empty(&cq->wc_list)))
+		soft_polled = poll_soft_wc(cq, num_entries, wc, false);
+
+	for (npolled = 0; npolled < num_entries - soft_polled; npolled++) {
+		if (mlx5_poll_one(cq, &cur_qp, wc + soft_polled + npolled))
+			break;
+	}
+
+	if (npolled)
+		mlx5_cq_set_ci(&cq->mcq);
+out:
+	spin_unlock_irqrestore(&cq->lock, flags);
+
+	return soft_polled + npolled;
+}
+
+int mlx5_ib_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags flags)
+{
+	struct mlx5_core_dev *mdev = to_mdev(ibcq->device)->mdev;
+	struct mlx5_ib_cq *cq = to_mcq(ibcq);
+	void __iomem *uar_page = mdev->priv.uar->map;
+	unsigned long irq_flags;
+	int ret = 0;
+
+	spin_lock_irqsave(&cq->lock, irq_flags);
+	if (cq->notify_flags != IB_CQ_NEXT_COMP)
+		cq->notify_flags = flags & IB_CQ_SOLICITED_MASK;
+
+	if ((flags & IB_CQ_REPORT_MISSED_EVENTS) && !list_empty(&cq->wc_list))
+		ret = 1;
+	spin_unlock_irqrestore(&cq->lock, irq_flags);
+
+	mlx5_cq_arm(&cq->mcq,
+		    (flags & IB_CQ_SOLICITED_MASK) == IB_CQ_SOLICITED ?
+		    MLX5_CQ_DB_REQ_NOT_SOL : MLX5_CQ_DB_REQ_NOT,
+		    uar_page, to_mcq(ibcq)->mcq.cons_index);
+
+	return ret;
+}
+
+static int alloc_cq_frag_buf(struct mlx5_ib_dev *dev,
+			     struct mlx5_ib_cq_buf *buf,
+			     int nent,
+			     int cqe_size)
+{
+	struct mlx5_frag_buf_ctrl *c = &buf->fbc;
+	struct mlx5_frag_buf *frag_buf = &c->frag_buf;
+	u32 cqc_buff[MLX5_ST_SZ_DW(cqc)] = {0};
+	int err;
+
+	MLX5_SET(cqc, cqc_buff, log_cq_size, ilog2(cqe_size));
+	MLX5_SET(cqc, cqc_buff, cqe_sz, (cqe_size == 128) ? 1 : 0);
+
+	mlx5_core_init_cq_frag_buf(&buf->fbc, cqc_buff);
+
+	err = mlx5_frag_buf_alloc_node(dev->mdev,
+				       nent * cqe_size,
+				       frag_buf,
+				       dev->mdev->priv.numa_node);
+	if (err)
+		return err;
+
+	buf->cqe_size = cqe_size;
+	buf->nent = nent;
+
+	return 0;
+}
+
+enum {
+	MLX5_CQE_RES_FORMAT_HASH = 0,
+	MLX5_CQE_RES_FORMAT_CSUM = 1,
+	MLX5_CQE_RES_FORMAT_CSUM_STRIDX = 3,
+};
+
+static int mini_cqe_res_format_to_hw(struct mlx5_ib_dev *dev, u8 format)
+{
+	switch (format) {
+	case MLX5_IB_CQE_RES_FORMAT_HASH:
+		return MLX5_CQE_RES_FORMAT_HASH;
+	case MLX5_IB_CQE_RES_FORMAT_CSUM:
+		return MLX5_CQE_RES_FORMAT_CSUM;
+	case MLX5_IB_CQE_RES_FORMAT_CSUM_STRIDX:
+		if (MLX5_CAP_GEN(dev->mdev, mini_cqe_resp_stride_index))
+			return MLX5_CQE_RES_FORMAT_CSUM_STRIDX;
+		return -EOPNOTSUPP;
+	default:
+		return -EINVAL;
+	}
+}
+
+static int create_cq_user(struct mlx5_ib_dev *dev, struct ib_udata *udata,
+			  struct ib_ucontext *context, struct mlx5_ib_cq *cq,
+			  int entries, u32 **cqb,
+			  int *cqe_size, int *index, int *inlen)
+{
+	struct mlx5_ib_create_cq ucmd = {};
+	size_t ucmdlen;
+	int page_shift;
+	__be64 *pas;
+	int npages;
+	int ncont;
+	void *cqc;
+	int err;
+
+	ucmdlen = udata->inlen < sizeof(ucmd) ?
+		  (sizeof(ucmd) - sizeof(ucmd.flags)) : sizeof(ucmd);
+
+	if (ib_copy_from_udata(&ucmd, udata, ucmdlen))
+		return -EFAULT;
+
+	if (ucmdlen == sizeof(ucmd) &&
+	    (ucmd.flags & ~(MLX5_IB_CREATE_CQ_FLAGS_CQE_128B_PAD)))
+		return -EINVAL;
+
+	if (ucmd.cqe_size != 64 && ucmd.cqe_size != 128)
+		return -EINVAL;
+
+	*cqe_size = ucmd.cqe_size;
+
+	cq->buf.umem = ib_umem_get(context, ucmd.buf_addr,
+				   entries * ucmd.cqe_size,
+				   IB_ACCESS_LOCAL_WRITE, 1);
+	if (IS_ERR(cq->buf.umem)) {
+		err = PTR_ERR(cq->buf.umem);
+		return err;
+	}
+
+	err = mlx5_ib_db_map_user(to_mucontext(context), ucmd.db_addr,
+				  &cq->db);
+	if (err)
+		goto err_umem;
+
+	mlx5_ib_cont_pages(cq->buf.umem, ucmd.buf_addr, 0, &npages, &page_shift,
+			   &ncont, NULL);
+	mlx5_ib_dbg(dev, "addr 0x%llx, size %u, npages %d, page_shift %d, ncont %d\n",
+		    ucmd.buf_addr, entries * ucmd.cqe_size, npages, page_shift, ncont);
+
+	*inlen = MLX5_ST_SZ_BYTES(create_cq_in) +
+		 MLX5_FLD_SZ_BYTES(create_cq_in, pas[0]) * ncont;
+	*cqb = kvzalloc(*inlen, GFP_KERNEL);
+	if (!*cqb) {
+		err = -ENOMEM;
+		goto err_db;
+	}
+
+	pas = (__be64 *)MLX5_ADDR_OF(create_cq_in, *cqb, pas);
+	mlx5_ib_populate_pas(dev, cq->buf.umem, page_shift, pas, 0);
+
+	cqc = MLX5_ADDR_OF(create_cq_in, *cqb, cq_context);
+	MLX5_SET(cqc, cqc, log_page_size,
+		 page_shift - MLX5_ADAPTER_PAGE_SHIFT);
+
+	*index = to_mucontext(context)->bfregi.sys_pages[0];
+
+	if (ucmd.cqe_comp_en == 1) {
+		int mini_cqe_format;
+
+		if (!((*cqe_size == 128 &&
+		       MLX5_CAP_GEN(dev->mdev, cqe_compression_128)) ||
+		      (*cqe_size == 64  &&
+		       MLX5_CAP_GEN(dev->mdev, cqe_compression)))) {
+			err = -EOPNOTSUPP;
+			mlx5_ib_warn(dev, "CQE compression is not supported for size %d!\n",
+				     *cqe_size);
+			goto err_cqb;
+		}
+
+		mini_cqe_format =
+			mini_cqe_res_format_to_hw(dev,
+						  ucmd.cqe_comp_res_format);
+		if (mini_cqe_format < 0) {
+			err = mini_cqe_format;
+			mlx5_ib_dbg(dev, "CQE compression res format %d error: %d\n",
+				    ucmd.cqe_comp_res_format, err);
+			goto err_cqb;
+		}
+
+		MLX5_SET(cqc, cqc, cqe_comp_en, 1);
+		MLX5_SET(cqc, cqc, mini_cqe_res_format, mini_cqe_format);
+	}
+
+	if (ucmd.flags & MLX5_IB_CREATE_CQ_FLAGS_CQE_128B_PAD) {
+		if (*cqe_size != 128 ||
+		    !MLX5_CAP_GEN(dev->mdev, cqe_128_always)) {
+			err = -EOPNOTSUPP;
+			mlx5_ib_warn(dev,
+				     "CQE padding is not supported for CQE size of %dB!\n",
+				     *cqe_size);
+			goto err_cqb;
+		}
+
+		cq->private_flags |= MLX5_IB_CQ_PR_FLAGS_CQE_128_PAD;
+	}
+
+	return 0;
+
+err_cqb:
+	kvfree(*cqb);
+
+err_db:
+	mlx5_ib_db_unmap_user(to_mucontext(context), &cq->db);
+
+err_umem:
+	ib_umem_release(cq->buf.umem);
+	return err;
+}
+
+static void destroy_cq_user(struct mlx5_ib_cq *cq, struct ib_ucontext *context)
+{
+	mlx5_ib_db_unmap_user(to_mucontext(context), &cq->db);
+	ib_umem_release(cq->buf.umem);
+}
+
+static void init_cq_frag_buf(struct mlx5_ib_cq_buf *buf)
+{
+	int i;
+	void *cqe;
+	struct mlx5_cqe64 *cqe64;
+
+	for (i = 0; i < buf->nent; i++) {
+		cqe = mlx5_frag_buf_get_wqe(&buf->fbc, i);
+		cqe64 = buf->cqe_size == 64 ? cqe : cqe + 64;
+		cqe64->op_own = MLX5_CQE_INVALID << 4;
+	}
+}
+
+static int create_cq_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_cq *cq,
+			    int entries, int cqe_size,
+			    u32 **cqb, int *index, int *inlen)
+{
+	__be64 *pas;
+	void *cqc;
+	int err;
+
+	err = mlx5_db_alloc(dev->mdev, &cq->db);
+	if (err)
+		return err;
+
+	cq->mcq.set_ci_db  = cq->db.db;
+	cq->mcq.arm_db     = cq->db.db + 1;
+	cq->mcq.cqe_sz = cqe_size;
+
+	err = alloc_cq_frag_buf(dev, &cq->buf, entries, cqe_size);
+	if (err)
+		goto err_db;
+
+	init_cq_frag_buf(&cq->buf);
+
+	*inlen = MLX5_ST_SZ_BYTES(create_cq_in) +
+		 MLX5_FLD_SZ_BYTES(create_cq_in, pas[0]) *
+		 cq->buf.fbc.frag_buf.npages;
+	*cqb = kvzalloc(*inlen, GFP_KERNEL);
+	if (!*cqb) {
+		err = -ENOMEM;
+		goto err_buf;
+	}
+
+	pas = (__be64 *)MLX5_ADDR_OF(create_cq_in, *cqb, pas);
+	mlx5_fill_page_frag_array(&cq->buf.fbc.frag_buf, pas);
+
+	cqc = MLX5_ADDR_OF(create_cq_in, *cqb, cq_context);
+	MLX5_SET(cqc, cqc, log_page_size,
+		 cq->buf.fbc.frag_buf.page_shift -
+		 MLX5_ADAPTER_PAGE_SHIFT);
+
+	*index = dev->mdev->priv.uar->index;
+
+	return 0;
+
+err_buf:
+	free_cq_buf(dev, &cq->buf);
+
+err_db:
+	mlx5_db_free(dev->mdev, &cq->db);
+	return err;
+}
+
+static void destroy_cq_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_cq *cq)
+{
+	free_cq_buf(dev, &cq->buf);
+	mlx5_db_free(dev->mdev, &cq->db);
+}
+
+static void notify_soft_wc_handler(struct work_struct *work)
+{
+	struct mlx5_ib_cq *cq = container_of(work, struct mlx5_ib_cq,
+					     notify_work);
+
+	cq->ibcq.comp_handler(&cq->ibcq, cq->ibcq.cq_context);
+}
+
+struct ib_cq *mlx5_ib_create_cq(struct ib_device *ibdev,
+				const struct ib_cq_init_attr *attr,
+				struct ib_ucontext *context,
+				struct ib_udata *udata)
+{
+	int entries = attr->cqe;
+	int vector = attr->comp_vector;
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+	struct mlx5_ib_cq *cq;
+	int uninitialized_var(index);
+	int uninitialized_var(inlen);
+	u32 *cqb = NULL;
+	void *cqc;
+	int cqe_size;
+	unsigned int irqn;
+	int eqn;
+	int err;
+
+	if (entries < 0 ||
+	    (entries > (1 << MLX5_CAP_GEN(dev->mdev, log_max_cq_sz))))
+		return ERR_PTR(-EINVAL);
+
+	if (check_cq_create_flags(attr->flags))
+		return ERR_PTR(-EOPNOTSUPP);
+
+	entries = roundup_pow_of_two(entries + 1);
+	if (entries > (1 << MLX5_CAP_GEN(dev->mdev, log_max_cq_sz)))
+		return ERR_PTR(-EINVAL);
+
+	cq = kzalloc(sizeof(*cq), GFP_KERNEL);
+	if (!cq)
+		return ERR_PTR(-ENOMEM);
+
+	cq->ibcq.cqe = entries - 1;
+	mutex_init(&cq->resize_mutex);
+	spin_lock_init(&cq->lock);
+	cq->resize_buf = NULL;
+	cq->resize_umem = NULL;
+	cq->create_flags = attr->flags;
+	INIT_LIST_HEAD(&cq->list_send_qp);
+	INIT_LIST_HEAD(&cq->list_recv_qp);
+
+	if (context) {
+		err = create_cq_user(dev, udata, context, cq, entries,
+				     &cqb, &cqe_size, &index, &inlen);
+		if (err)
+			goto err_create;
+	} else {
+		cqe_size = cache_line_size() == 128 ? 128 : 64;
+		err = create_cq_kernel(dev, cq, entries, cqe_size, &cqb,
+				       &index, &inlen);
+		if (err)
+			goto err_create;
+
+		INIT_WORK(&cq->notify_work, notify_soft_wc_handler);
+	}
+
+	err = mlx5_vector2eqn(dev->mdev, vector, &eqn, &irqn);
+	if (err)
+		goto err_cqb;
+
+	cq->cqe_size = cqe_size;
+
+	cqc = MLX5_ADDR_OF(create_cq_in, cqb, cq_context);
+	MLX5_SET(cqc, cqc, cqe_sz,
+		 cqe_sz_to_mlx_sz(cqe_size,
+				  cq->private_flags &
+				  MLX5_IB_CQ_PR_FLAGS_CQE_128_PAD));
+	MLX5_SET(cqc, cqc, log_cq_size, ilog2(entries));
+	MLX5_SET(cqc, cqc, uar_page, index);
+	MLX5_SET(cqc, cqc, c_eqn, eqn);
+	MLX5_SET64(cqc, cqc, dbr_addr, cq->db.dma);
+	if (cq->create_flags & IB_UVERBS_CQ_FLAGS_IGNORE_OVERRUN)
+		MLX5_SET(cqc, cqc, oi, 1);
+
+	err = mlx5_core_create_cq(dev->mdev, &cq->mcq, cqb, inlen);
+	if (err)
+		goto err_cqb;
+
+	mlx5_ib_dbg(dev, "cqn 0x%x\n", cq->mcq.cqn);
+	cq->mcq.irqn = irqn;
+	if (context)
+		cq->mcq.tasklet_ctx.comp = mlx5_ib_cq_comp;
+	else
+		cq->mcq.comp  = mlx5_ib_cq_comp;
+	cq->mcq.event = mlx5_ib_cq_event;
+
+	INIT_LIST_HEAD(&cq->wc_list);
+
+	if (context)
+		if (ib_copy_to_udata(udata, &cq->mcq.cqn, sizeof(__u32))) {
+			err = -EFAULT;
+			goto err_cmd;
+		}
+
+
+	kvfree(cqb);
+	return &cq->ibcq;
+
+err_cmd:
+	mlx5_core_destroy_cq(dev->mdev, &cq->mcq);
+
+err_cqb:
+	kvfree(cqb);
+	if (context)
+		destroy_cq_user(cq, context);
+	else
+		destroy_cq_kernel(dev, cq);
+
+err_create:
+	kfree(cq);
+
+	return ERR_PTR(err);
+}
+
+
+int mlx5_ib_destroy_cq(struct ib_cq *cq)
+{
+	struct mlx5_ib_dev *dev = to_mdev(cq->device);
+	struct mlx5_ib_cq *mcq = to_mcq(cq);
+	struct ib_ucontext *context = NULL;
+
+	if (cq->uobject)
+		context = cq->uobject->context;
+
+	mlx5_core_destroy_cq(dev->mdev, &mcq->mcq);
+	if (context)
+		destroy_cq_user(mcq, context);
+	else
+		destroy_cq_kernel(dev, mcq);
+
+	kfree(mcq);
+
+	return 0;
+}
+
+static int is_equal_rsn(struct mlx5_cqe64 *cqe64, u32 rsn)
+{
+	return rsn == (ntohl(cqe64->sop_drop_qpn) & 0xffffff);
+}
+
+void __mlx5_ib_cq_clean(struct mlx5_ib_cq *cq, u32 rsn, struct mlx5_ib_srq *srq)
+{
+	struct mlx5_cqe64 *cqe64, *dest64;
+	void *cqe, *dest;
+	u32 prod_index;
+	int nfreed = 0;
+	u8 owner_bit;
+
+	if (!cq)
+		return;
+
+	/* First we need to find the current producer index, so we
+	 * know where to start cleaning from.  It doesn't matter if HW
+	 * adds new entries after this loop -- the QP we're worried
+	 * about is already in RESET, so the new entries won't come
+	 * from our QP and therefore don't need to be checked.
+	 */
+	for (prod_index = cq->mcq.cons_index; get_sw_cqe(cq, prod_index); prod_index++)
+		if (prod_index == cq->mcq.cons_index + cq->ibcq.cqe)
+			break;
+
+	/* Now sweep backwards through the CQ, removing CQ entries
+	 * that match our QP by copying older entries on top of them.
+	 */
+	while ((int) --prod_index - (int) cq->mcq.cons_index >= 0) {
+		cqe = get_cqe(cq, prod_index & cq->ibcq.cqe);
+		cqe64 = (cq->mcq.cqe_sz == 64) ? cqe : cqe + 64;
+		if (is_equal_rsn(cqe64, rsn)) {
+			if (srq && (ntohl(cqe64->srqn) & 0xffffff))
+				mlx5_ib_free_srq_wqe(srq, be16_to_cpu(cqe64->wqe_counter));
+			++nfreed;
+		} else if (nfreed) {
+			dest = get_cqe(cq, (prod_index + nfreed) & cq->ibcq.cqe);
+			dest64 = (cq->mcq.cqe_sz == 64) ? dest : dest + 64;
+			owner_bit = dest64->op_own & MLX5_CQE_OWNER_MASK;
+			memcpy(dest, cqe, cq->mcq.cqe_sz);
+			dest64->op_own = owner_bit |
+				(dest64->op_own & ~MLX5_CQE_OWNER_MASK);
+		}
+	}
+
+	if (nfreed) {
+		cq->mcq.cons_index += nfreed;
+		/* Make sure update of buffer contents is done before
+		 * updating consumer index.
+		 */
+		wmb();
+		mlx5_cq_set_ci(&cq->mcq);
+	}
+}
+
+void mlx5_ib_cq_clean(struct mlx5_ib_cq *cq, u32 qpn, struct mlx5_ib_srq *srq)
+{
+	if (!cq)
+		return;
+
+	spin_lock_irq(&cq->lock);
+	__mlx5_ib_cq_clean(cq, qpn, srq);
+	spin_unlock_irq(&cq->lock);
+}
+
+int mlx5_ib_modify_cq(struct ib_cq *cq, u16 cq_count, u16 cq_period)
+{
+	struct mlx5_ib_dev *dev = to_mdev(cq->device);
+	struct mlx5_ib_cq *mcq = to_mcq(cq);
+	int err;
+
+	if (!MLX5_CAP_GEN(dev->mdev, cq_moderation))
+		return -EOPNOTSUPP;
+
+	if (cq_period > MLX5_MAX_CQ_PERIOD)
+		return -EINVAL;
+
+	err = mlx5_core_modify_cq_moderation(dev->mdev, &mcq->mcq,
+					     cq_period, cq_count);
+	if (err)
+		mlx5_ib_warn(dev, "modify cq 0x%x failed\n", mcq->mcq.cqn);
+
+	return err;
+}
+
+static int resize_user(struct mlx5_ib_dev *dev, struct mlx5_ib_cq *cq,
+		       int entries, struct ib_udata *udata, int *npas,
+		       int *page_shift, int *cqe_size)
+{
+	struct mlx5_ib_resize_cq ucmd;
+	struct ib_umem *umem;
+	int err;
+	int npages;
+	struct ib_ucontext *context = cq->buf.umem->context;
+
+	err = ib_copy_from_udata(&ucmd, udata, sizeof(ucmd));
+	if (err)
+		return err;
+
+	if (ucmd.reserved0 || ucmd.reserved1)
+		return -EINVAL;
+
+	/* check multiplication overflow */
+	if (ucmd.cqe_size && SIZE_MAX / ucmd.cqe_size <= entries - 1)
+		return -EINVAL;
+
+	umem = ib_umem_get(context, ucmd.buf_addr,
+			   (size_t)ucmd.cqe_size * entries,
+			   IB_ACCESS_LOCAL_WRITE, 1);
+	if (IS_ERR(umem)) {
+		err = PTR_ERR(umem);
+		return err;
+	}
+
+	mlx5_ib_cont_pages(umem, ucmd.buf_addr, 0, &npages, page_shift,
+			   npas, NULL);
+
+	cq->resize_umem = umem;
+	*cqe_size = ucmd.cqe_size;
+
+	return 0;
+}
+
+static void un_resize_user(struct mlx5_ib_cq *cq)
+{
+	ib_umem_release(cq->resize_umem);
+}
+
+static int resize_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_cq *cq,
+			 int entries, int cqe_size)
+{
+	int err;
+
+	cq->resize_buf = kzalloc(sizeof(*cq->resize_buf), GFP_KERNEL);
+	if (!cq->resize_buf)
+		return -ENOMEM;
+
+	err = alloc_cq_frag_buf(dev, cq->resize_buf, entries, cqe_size);
+	if (err)
+		goto ex;
+
+	init_cq_frag_buf(cq->resize_buf);
+
+	return 0;
+
+ex:
+	kfree(cq->resize_buf);
+	return err;
+}
+
+static void un_resize_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_cq *cq)
+{
+	free_cq_buf(dev, cq->resize_buf);
+	cq->resize_buf = NULL;
+}
+
+static int copy_resize_cqes(struct mlx5_ib_cq *cq)
+{
+	struct mlx5_ib_dev *dev = to_mdev(cq->ibcq.device);
+	struct mlx5_cqe64 *scqe64;
+	struct mlx5_cqe64 *dcqe64;
+	void *start_cqe;
+	void *scqe;
+	void *dcqe;
+	int ssize;
+	int dsize;
+	int i;
+	u8 sw_own;
+
+	ssize = cq->buf.cqe_size;
+	dsize = cq->resize_buf->cqe_size;
+	if (ssize != dsize) {
+		mlx5_ib_warn(dev, "resize from different cqe size is not supported\n");
+		return -EINVAL;
+	}
+
+	i = cq->mcq.cons_index;
+	scqe = get_sw_cqe(cq, i);
+	scqe64 = ssize == 64 ? scqe : scqe + 64;
+	start_cqe = scqe;
+	if (!scqe) {
+		mlx5_ib_warn(dev, "expected cqe in sw ownership\n");
+		return -EINVAL;
+	}
+
+	while ((scqe64->op_own >> 4) != MLX5_CQE_RESIZE_CQ) {
+		dcqe = mlx5_frag_buf_get_wqe(&cq->resize_buf->fbc,
+					     (i + 1) & cq->resize_buf->nent);
+		dcqe64 = dsize == 64 ? dcqe : dcqe + 64;
+		sw_own = sw_ownership_bit(i + 1, cq->resize_buf->nent);
+		memcpy(dcqe, scqe, dsize);
+		dcqe64->op_own = (dcqe64->op_own & ~MLX5_CQE_OWNER_MASK) | sw_own;
+
+		++i;
+		scqe = get_sw_cqe(cq, i);
+		scqe64 = ssize == 64 ? scqe : scqe + 64;
+		if (!scqe) {
+			mlx5_ib_warn(dev, "expected cqe in sw ownership\n");
+			return -EINVAL;
+		}
+
+		if (scqe == start_cqe) {
+			pr_warn("resize CQ failed to get resize CQE, CQN 0x%x\n",
+				cq->mcq.cqn);
+			return -ENOMEM;
+		}
+	}
+	++cq->mcq.cons_index;
+	return 0;
+}
+
+int mlx5_ib_resize_cq(struct ib_cq *ibcq, int entries, struct ib_udata *udata)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibcq->device);
+	struct mlx5_ib_cq *cq = to_mcq(ibcq);
+	void *cqc;
+	u32 *in;
+	int err;
+	int npas;
+	__be64 *pas;
+	int page_shift;
+	int inlen;
+	int uninitialized_var(cqe_size);
+	unsigned long flags;
+
+	if (!MLX5_CAP_GEN(dev->mdev, cq_resize)) {
+		pr_info("Firmware does not support resize CQ\n");
+		return -ENOSYS;
+	}
+
+	if (entries < 1 ||
+	    entries > (1 << MLX5_CAP_GEN(dev->mdev, log_max_cq_sz))) {
+		mlx5_ib_warn(dev, "wrong entries number %d, max %d\n",
+			     entries,
+			     1 << MLX5_CAP_GEN(dev->mdev, log_max_cq_sz));
+		return -EINVAL;
+	}
+
+	entries = roundup_pow_of_two(entries + 1);
+	if (entries > (1 << MLX5_CAP_GEN(dev->mdev, log_max_cq_sz)) + 1)
+		return -EINVAL;
+
+	if (entries == ibcq->cqe + 1)
+		return 0;
+
+	mutex_lock(&cq->resize_mutex);
+	if (udata) {
+		err = resize_user(dev, cq, entries, udata, &npas, &page_shift,
+				  &cqe_size);
+	} else {
+		cqe_size = 64;
+		err = resize_kernel(dev, cq, entries, cqe_size);
+		if (!err) {
+			struct mlx5_frag_buf_ctrl *c;
+
+			c = &cq->resize_buf->fbc;
+			npas = c->frag_buf.npages;
+			page_shift = c->frag_buf.page_shift;
+		}
+	}
+
+	if (err)
+		goto ex;
+
+	inlen = MLX5_ST_SZ_BYTES(modify_cq_in) +
+		MLX5_FLD_SZ_BYTES(modify_cq_in, pas[0]) * npas;
+
+	in = kvzalloc(inlen, GFP_KERNEL);
+	if (!in) {
+		err = -ENOMEM;
+		goto ex_resize;
+	}
+
+	pas = (__be64 *)MLX5_ADDR_OF(modify_cq_in, in, pas);
+	if (udata)
+		mlx5_ib_populate_pas(dev, cq->resize_umem, page_shift,
+				     pas, 0);
+	else
+		mlx5_fill_page_frag_array(&cq->resize_buf->fbc.frag_buf,
+					  pas);
+
+	MLX5_SET(modify_cq_in, in,
+		 modify_field_select_resize_field_select.resize_field_select.resize_field_select,
+		 MLX5_MODIFY_CQ_MASK_LOG_SIZE  |
+		 MLX5_MODIFY_CQ_MASK_PG_OFFSET |
+		 MLX5_MODIFY_CQ_MASK_PG_SIZE);
+
+	cqc = MLX5_ADDR_OF(modify_cq_in, in, cq_context);
+
+	MLX5_SET(cqc, cqc, log_page_size,
+		 page_shift - MLX5_ADAPTER_PAGE_SHIFT);
+	MLX5_SET(cqc, cqc, cqe_sz,
+		 cqe_sz_to_mlx_sz(cqe_size,
+				  cq->private_flags &
+				  MLX5_IB_CQ_PR_FLAGS_CQE_128_PAD));
+	MLX5_SET(cqc, cqc, log_cq_size, ilog2(entries));
+
+	MLX5_SET(modify_cq_in, in, op_mod, MLX5_CQ_OPMOD_RESIZE);
+	MLX5_SET(modify_cq_in, in, cqn, cq->mcq.cqn);
+
+	err = mlx5_core_modify_cq(dev->mdev, &cq->mcq, in, inlen);
+	if (err)
+		goto ex_alloc;
+
+	if (udata) {
+		cq->ibcq.cqe = entries - 1;
+		ib_umem_release(cq->buf.umem);
+		cq->buf.umem = cq->resize_umem;
+		cq->resize_umem = NULL;
+	} else {
+		struct mlx5_ib_cq_buf tbuf;
+		int resized = 0;
+
+		spin_lock_irqsave(&cq->lock, flags);
+		if (cq->resize_buf) {
+			err = copy_resize_cqes(cq);
+			if (!err) {
+				tbuf = cq->buf;
+				cq->buf = *cq->resize_buf;
+				kfree(cq->resize_buf);
+				cq->resize_buf = NULL;
+				resized = 1;
+			}
+		}
+		cq->ibcq.cqe = entries - 1;
+		spin_unlock_irqrestore(&cq->lock, flags);
+		if (resized)
+			free_cq_buf(dev, &tbuf);
+	}
+	mutex_unlock(&cq->resize_mutex);
+
+	kvfree(in);
+	return 0;
+
+ex_alloc:
+	kvfree(in);
+
+ex_resize:
+	if (udata)
+		un_resize_user(cq);
+	else
+		un_resize_kernel(dev, cq);
+ex:
+	mutex_unlock(&cq->resize_mutex);
+	return err;
+}
+
+int mlx5_ib_get_cqe_size(struct mlx5_ib_dev *dev, struct ib_cq *ibcq)
+{
+	struct mlx5_ib_cq *cq;
+
+	if (!ibcq)
+		return 128;
+
+	cq = to_mcq(ibcq);
+	return cq->cqe_size;
+}
+
+/* Called from atomic context */
+int mlx5_ib_generate_wc(struct ib_cq *ibcq, struct ib_wc *wc)
+{
+	struct mlx5_ib_wc *soft_wc;
+	struct mlx5_ib_cq *cq = to_mcq(ibcq);
+	unsigned long flags;
+
+	soft_wc = kmalloc(sizeof(*soft_wc), GFP_ATOMIC);
+	if (!soft_wc)
+		return -ENOMEM;
+
+	soft_wc->wc = *wc;
+	spin_lock_irqsave(&cq->lock, flags);
+	list_add_tail(&soft_wc->list, &cq->wc_list);
+	if (cq->notify_flags == IB_CQ_NEXT_COMP ||
+	    wc->status != IB_WC_SUCCESS) {
+		cq->notify_flags = 0;
+		schedule_work(&cq->notify_work);
+	}
+	spin_unlock_irqrestore(&cq->lock, flags);
+
+	return 0;
+}
diff --git a/drivers/infiniband/hw/mlx5/devx.c b/drivers/infiniband/hw/mlx5/devx.c
new file mode 100644
index 000000000..c89aec834
--- /dev/null
+++ b/drivers/infiniband/hw/mlx5/devx.c
@@ -0,0 +1,1127 @@
+// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
+/*
+ * Copyright (c) 2018, Mellanox Technologies inc.  All rights reserved.
+ */
+
+#include <rdma/ib_user_verbs.h>
+#include <rdma/ib_verbs.h>
+#include <rdma/uverbs_types.h>
+#include <rdma/uverbs_ioctl.h>
+#include <rdma/mlx5_user_ioctl_cmds.h>
+#include <rdma/ib_umem.h>
+#include <linux/mlx5/driver.h>
+#include <linux/mlx5/fs.h>
+#include "mlx5_ib.h"
+
+#define UVERBS_MODULE_NAME mlx5_ib
+#include <rdma/uverbs_named_ioctl.h>
+
+#define MLX5_MAX_DESTROY_INBOX_SIZE_DW MLX5_ST_SZ_DW(delete_fte_in)
+struct devx_obj {
+	struct mlx5_core_dev	*mdev;
+	u32			obj_id;
+	u32			dinlen; /* destroy inbox length */
+	u32			dinbox[MLX5_MAX_DESTROY_INBOX_SIZE_DW];
+};
+
+struct devx_umem {
+	struct mlx5_core_dev		*mdev;
+	struct ib_umem			*umem;
+	u32				page_offset;
+	int				page_shift;
+	int				ncont;
+	u32				dinlen;
+	u32				dinbox[MLX5_ST_SZ_DW(general_obj_in_cmd_hdr)];
+};
+
+struct devx_umem_reg_cmd {
+	void				*in;
+	u32				inlen;
+	u32				out[MLX5_ST_SZ_DW(general_obj_out_cmd_hdr)];
+};
+
+static struct mlx5_ib_ucontext *devx_ufile2uctx(struct ib_uverbs_file *file)
+{
+	return to_mucontext(ib_uverbs_get_ucontext(file));
+}
+
+int mlx5_ib_devx_create(struct mlx5_ib_dev *dev, struct mlx5_ib_ucontext *context)
+{
+	u32 in[MLX5_ST_SZ_DW(create_uctx_in)] = {0};
+	u32 out[MLX5_ST_SZ_DW(general_obj_out_cmd_hdr)] = {0};
+	u64 general_obj_types;
+	void *hdr;
+	int err;
+
+	hdr = MLX5_ADDR_OF(create_uctx_in, in, hdr);
+
+	general_obj_types = MLX5_CAP_GEN_64(dev->mdev, general_obj_types);
+	if (!(general_obj_types & MLX5_GENERAL_OBJ_TYPES_CAP_UCTX) ||
+	    !(general_obj_types & MLX5_GENERAL_OBJ_TYPES_CAP_UMEM))
+		return -EINVAL;
+
+	if (!capable(CAP_NET_RAW))
+		return -EPERM;
+
+	MLX5_SET(general_obj_in_cmd_hdr, hdr, opcode, MLX5_CMD_OP_CREATE_GENERAL_OBJECT);
+	MLX5_SET(general_obj_in_cmd_hdr, hdr, obj_type, MLX5_OBJ_TYPE_UCTX);
+
+	err = mlx5_cmd_exec(dev->mdev, in, sizeof(in), out, sizeof(out));
+	if (err)
+		return err;
+
+	context->devx_uid = MLX5_GET(general_obj_out_cmd_hdr, out, obj_id);
+	return 0;
+}
+
+void mlx5_ib_devx_destroy(struct mlx5_ib_dev *dev,
+			  struct mlx5_ib_ucontext *context)
+{
+	u32 in[MLX5_ST_SZ_DW(general_obj_in_cmd_hdr)] = {0};
+	u32 out[MLX5_ST_SZ_DW(general_obj_out_cmd_hdr)] = {0};
+
+	MLX5_SET(general_obj_in_cmd_hdr, in, opcode, MLX5_CMD_OP_DESTROY_GENERAL_OBJECT);
+	MLX5_SET(general_obj_in_cmd_hdr, in, obj_type, MLX5_OBJ_TYPE_UCTX);
+	MLX5_SET(general_obj_in_cmd_hdr, in, obj_id, context->devx_uid);
+
+	mlx5_cmd_exec(dev->mdev, in, sizeof(in), out, sizeof(out));
+}
+
+bool mlx5_ib_devx_is_flow_dest(void *obj, int *dest_id, int *dest_type)
+{
+	struct devx_obj *devx_obj = obj;
+	u16 opcode = MLX5_GET(general_obj_in_cmd_hdr, devx_obj->dinbox, opcode);
+
+	switch (opcode) {
+	case MLX5_CMD_OP_DESTROY_TIR:
+		*dest_type = MLX5_FLOW_DESTINATION_TYPE_TIR;
+		*dest_id = MLX5_GET(general_obj_in_cmd_hdr, devx_obj->dinbox,
+				    obj_id);
+		return true;
+
+	case MLX5_CMD_OP_DESTROY_FLOW_TABLE:
+		*dest_type = MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE;
+		*dest_id = MLX5_GET(destroy_flow_table_in, devx_obj->dinbox,
+				    table_id);
+		return true;
+	default:
+		return false;
+	}
+}
+
+static int devx_is_valid_obj_id(struct devx_obj *obj, const void *in)
+{
+	u16 opcode = MLX5_GET(general_obj_in_cmd_hdr, in, opcode);
+	u32 obj_id;
+
+	switch (opcode) {
+	case MLX5_CMD_OP_MODIFY_GENERAL_OBJECT:
+	case MLX5_CMD_OP_QUERY_GENERAL_OBJECT:
+		obj_id = MLX5_GET(general_obj_in_cmd_hdr, in, obj_id);
+		break;
+	case MLX5_CMD_OP_QUERY_MKEY:
+		obj_id = MLX5_GET(query_mkey_in, in, mkey_index);
+		break;
+	case MLX5_CMD_OP_QUERY_CQ:
+		obj_id = MLX5_GET(query_cq_in, in, cqn);
+		break;
+	case MLX5_CMD_OP_MODIFY_CQ:
+		obj_id = MLX5_GET(modify_cq_in, in, cqn);
+		break;
+	case MLX5_CMD_OP_QUERY_SQ:
+		obj_id = MLX5_GET(query_sq_in, in, sqn);
+		break;
+	case MLX5_CMD_OP_MODIFY_SQ:
+		obj_id = MLX5_GET(modify_sq_in, in, sqn);
+		break;
+	case MLX5_CMD_OP_QUERY_RQ:
+		obj_id = MLX5_GET(query_rq_in, in, rqn);
+		break;
+	case MLX5_CMD_OP_MODIFY_RQ:
+		obj_id = MLX5_GET(modify_rq_in, in, rqn);
+		break;
+	case MLX5_CMD_OP_QUERY_RMP:
+		obj_id = MLX5_GET(query_rmp_in, in, rmpn);
+		break;
+	case MLX5_CMD_OP_MODIFY_RMP:
+		obj_id = MLX5_GET(modify_rmp_in, in, rmpn);
+		break;
+	case MLX5_CMD_OP_QUERY_RQT:
+		obj_id = MLX5_GET(query_rqt_in, in, rqtn);
+		break;
+	case MLX5_CMD_OP_MODIFY_RQT:
+		obj_id = MLX5_GET(modify_rqt_in, in, rqtn);
+		break;
+	case MLX5_CMD_OP_QUERY_TIR:
+		obj_id = MLX5_GET(query_tir_in, in, tirn);
+		break;
+	case MLX5_CMD_OP_MODIFY_TIR:
+		obj_id = MLX5_GET(modify_tir_in, in, tirn);
+		break;
+	case MLX5_CMD_OP_QUERY_TIS:
+		obj_id = MLX5_GET(query_tis_in, in, tisn);
+		break;
+	case MLX5_CMD_OP_MODIFY_TIS:
+		obj_id = MLX5_GET(modify_tis_in, in, tisn);
+		break;
+	case MLX5_CMD_OP_QUERY_FLOW_TABLE:
+		obj_id = MLX5_GET(query_flow_table_in, in, table_id);
+		break;
+	case MLX5_CMD_OP_MODIFY_FLOW_TABLE:
+		obj_id = MLX5_GET(modify_flow_table_in, in, table_id);
+		break;
+	case MLX5_CMD_OP_QUERY_FLOW_GROUP:
+		obj_id = MLX5_GET(query_flow_group_in, in, group_id);
+		break;
+	case MLX5_CMD_OP_QUERY_FLOW_TABLE_ENTRY:
+		obj_id = MLX5_GET(query_fte_in, in, flow_index);
+		break;
+	case MLX5_CMD_OP_SET_FLOW_TABLE_ENTRY:
+		obj_id = MLX5_GET(set_fte_in, in, flow_index);
+		break;
+	case MLX5_CMD_OP_QUERY_Q_COUNTER:
+		obj_id = MLX5_GET(query_q_counter_in, in, counter_set_id);
+		break;
+	case MLX5_CMD_OP_QUERY_FLOW_COUNTER:
+		obj_id = MLX5_GET(query_flow_counter_in, in, flow_counter_id);
+		break;
+	case MLX5_CMD_OP_QUERY_MODIFY_HEADER_CONTEXT:
+		obj_id = MLX5_GET(general_obj_in_cmd_hdr, in, obj_id);
+		break;
+	case MLX5_CMD_OP_QUERY_SCHEDULING_ELEMENT:
+		obj_id = MLX5_GET(query_scheduling_element_in, in,
+				  scheduling_element_id);
+		break;
+	case MLX5_CMD_OP_MODIFY_SCHEDULING_ELEMENT:
+		obj_id = MLX5_GET(modify_scheduling_element_in, in,
+				  scheduling_element_id);
+		break;
+	case MLX5_CMD_OP_ADD_VXLAN_UDP_DPORT:
+		obj_id = MLX5_GET(add_vxlan_udp_dport_in, in, vxlan_udp_port);
+		break;
+	case MLX5_CMD_OP_QUERY_L2_TABLE_ENTRY:
+		obj_id = MLX5_GET(query_l2_table_entry_in, in, table_index);
+		break;
+	case MLX5_CMD_OP_SET_L2_TABLE_ENTRY:
+		obj_id = MLX5_GET(set_l2_table_entry_in, in, table_index);
+		break;
+	case MLX5_CMD_OP_QUERY_QP:
+		obj_id = MLX5_GET(query_qp_in, in, qpn);
+		break;
+	case MLX5_CMD_OP_RST2INIT_QP:
+		obj_id = MLX5_GET(rst2init_qp_in, in, qpn);
+		break;
+	case MLX5_CMD_OP_INIT2RTR_QP:
+		obj_id = MLX5_GET(init2rtr_qp_in, in, qpn);
+		break;
+	case MLX5_CMD_OP_RTR2RTS_QP:
+		obj_id = MLX5_GET(rtr2rts_qp_in, in, qpn);
+		break;
+	case MLX5_CMD_OP_RTS2RTS_QP:
+		obj_id = MLX5_GET(rts2rts_qp_in, in, qpn);
+		break;
+	case MLX5_CMD_OP_SQERR2RTS_QP:
+		obj_id = MLX5_GET(sqerr2rts_qp_in, in, qpn);
+		break;
+	case MLX5_CMD_OP_2ERR_QP:
+		obj_id = MLX5_GET(qp_2err_in, in, qpn);
+		break;
+	case MLX5_CMD_OP_2RST_QP:
+		obj_id = MLX5_GET(qp_2rst_in, in, qpn);
+		break;
+	case MLX5_CMD_OP_QUERY_DCT:
+		obj_id = MLX5_GET(query_dct_in, in, dctn);
+		break;
+	case MLX5_CMD_OP_QUERY_XRQ:
+		obj_id = MLX5_GET(query_xrq_in, in, xrqn);
+		break;
+	case MLX5_CMD_OP_QUERY_XRC_SRQ:
+		obj_id = MLX5_GET(query_xrc_srq_in, in, xrc_srqn);
+		break;
+	case MLX5_CMD_OP_ARM_XRC_SRQ:
+		obj_id = MLX5_GET(arm_xrc_srq_in, in, xrc_srqn);
+		break;
+	case MLX5_CMD_OP_QUERY_SRQ:
+		obj_id = MLX5_GET(query_srq_in, in, srqn);
+		break;
+	case MLX5_CMD_OP_ARM_RQ:
+		obj_id = 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 = MLX5_GET(drain_dct_in, in, dctn);
+		break;
+	case MLX5_CMD_OP_ARM_XRQ:
+		obj_id = MLX5_GET(arm_xrq_in, in, xrqn);
+		break;
+	default:
+		return false;
+	}
+
+	if (obj_id == obj->obj_id)
+		return true;
+
+	return false;
+}
+
+static bool devx_is_obj_create_cmd(const void *in)
+{
+	u16 opcode = MLX5_GET(general_obj_in_cmd_hdr, in, opcode);
+
+	switch (opcode) {
+	case MLX5_CMD_OP_CREATE_GENERAL_OBJECT:
+	case MLX5_CMD_OP_CREATE_MKEY:
+	case MLX5_CMD_OP_CREATE_CQ:
+	case MLX5_CMD_OP_ALLOC_PD:
+	case MLX5_CMD_OP_ALLOC_TRANSPORT_DOMAIN:
+	case MLX5_CMD_OP_CREATE_RMP:
+	case MLX5_CMD_OP_CREATE_SQ:
+	case MLX5_CMD_OP_CREATE_RQ:
+	case MLX5_CMD_OP_CREATE_RQT:
+	case MLX5_CMD_OP_CREATE_TIR:
+	case MLX5_CMD_OP_CREATE_TIS:
+	case MLX5_CMD_OP_ALLOC_Q_COUNTER:
+	case MLX5_CMD_OP_CREATE_FLOW_TABLE:
+	case MLX5_CMD_OP_CREATE_FLOW_GROUP:
+	case MLX5_CMD_OP_ALLOC_FLOW_COUNTER:
+	case MLX5_CMD_OP_ALLOC_ENCAP_HEADER:
+	case MLX5_CMD_OP_ALLOC_MODIFY_HEADER_CONTEXT:
+	case MLX5_CMD_OP_CREATE_SCHEDULING_ELEMENT:
+	case MLX5_CMD_OP_ADD_VXLAN_UDP_DPORT:
+	case MLX5_CMD_OP_SET_L2_TABLE_ENTRY:
+	case MLX5_CMD_OP_CREATE_QP:
+	case MLX5_CMD_OP_CREATE_SRQ:
+	case MLX5_CMD_OP_CREATE_XRC_SRQ:
+	case MLX5_CMD_OP_CREATE_DCT:
+	case MLX5_CMD_OP_CREATE_XRQ:
+	case MLX5_CMD_OP_ATTACH_TO_MCG:
+	case MLX5_CMD_OP_ALLOC_XRCD:
+		return true;
+	case MLX5_CMD_OP_SET_FLOW_TABLE_ENTRY:
+	{
+		u16 op_mod = MLX5_GET(set_fte_in, in, op_mod);
+		if (op_mod == 0)
+			return true;
+		return false;
+	}
+	default:
+		return false;
+	}
+}
+
+static bool devx_is_obj_modify_cmd(const void *in)
+{
+	u16 opcode = MLX5_GET(general_obj_in_cmd_hdr, in, opcode);
+
+	switch (opcode) {
+	case MLX5_CMD_OP_MODIFY_GENERAL_OBJECT:
+	case MLX5_CMD_OP_MODIFY_CQ:
+	case MLX5_CMD_OP_MODIFY_RMP:
+	case MLX5_CMD_OP_MODIFY_SQ:
+	case MLX5_CMD_OP_MODIFY_RQ:
+	case MLX5_CMD_OP_MODIFY_RQT:
+	case MLX5_CMD_OP_MODIFY_TIR:
+	case MLX5_CMD_OP_MODIFY_TIS:
+	case MLX5_CMD_OP_MODIFY_FLOW_TABLE:
+	case MLX5_CMD_OP_MODIFY_SCHEDULING_ELEMENT:
+	case MLX5_CMD_OP_ADD_VXLAN_UDP_DPORT:
+	case MLX5_CMD_OP_SET_L2_TABLE_ENTRY:
+	case MLX5_CMD_OP_RST2INIT_QP:
+	case MLX5_CMD_OP_INIT2RTR_QP:
+	case MLX5_CMD_OP_INIT2INIT_QP:
+	case MLX5_CMD_OP_RTR2RTS_QP:
+	case MLX5_CMD_OP_RTS2RTS_QP:
+	case MLX5_CMD_OP_SQERR2RTS_QP:
+	case MLX5_CMD_OP_2ERR_QP:
+	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:
+		return true;
+	case MLX5_CMD_OP_SET_FLOW_TABLE_ENTRY:
+	{
+		u16 op_mod = MLX5_GET(set_fte_in, in, op_mod);
+
+		if (op_mod == 1)
+			return true;
+		return false;
+	}
+	default:
+		return false;
+	}
+}
+
+static bool devx_is_obj_query_cmd(const void *in)
+{
+	u16 opcode = MLX5_GET(general_obj_in_cmd_hdr, in, opcode);
+
+	switch (opcode) {
+	case MLX5_CMD_OP_QUERY_GENERAL_OBJECT:
+	case MLX5_CMD_OP_QUERY_MKEY:
+	case MLX5_CMD_OP_QUERY_CQ:
+	case MLX5_CMD_OP_QUERY_RMP:
+	case MLX5_CMD_OP_QUERY_SQ:
+	case MLX5_CMD_OP_QUERY_RQ:
+	case MLX5_CMD_OP_QUERY_RQT:
+	case MLX5_CMD_OP_QUERY_TIR:
+	case MLX5_CMD_OP_QUERY_TIS:
+	case MLX5_CMD_OP_QUERY_Q_COUNTER:
+	case MLX5_CMD_OP_QUERY_FLOW_TABLE:
+	case MLX5_CMD_OP_QUERY_FLOW_GROUP:
+	case MLX5_CMD_OP_QUERY_FLOW_TABLE_ENTRY:
+	case MLX5_CMD_OP_QUERY_FLOW_COUNTER:
+	case MLX5_CMD_OP_QUERY_MODIFY_HEADER_CONTEXT:
+	case MLX5_CMD_OP_QUERY_SCHEDULING_ELEMENT:
+	case MLX5_CMD_OP_QUERY_L2_TABLE_ENTRY:
+	case MLX5_CMD_OP_QUERY_QP:
+	case MLX5_CMD_OP_QUERY_SRQ:
+	case MLX5_CMD_OP_QUERY_XRC_SRQ:
+	case MLX5_CMD_OP_QUERY_DCT:
+	case MLX5_CMD_OP_QUERY_XRQ:
+		return true;
+	default:
+		return false;
+	}
+}
+
+static bool devx_is_general_cmd(void *in)
+{
+	u16 opcode = MLX5_GET(general_obj_in_cmd_hdr, in, opcode);
+
+	switch (opcode) {
+	case MLX5_CMD_OP_QUERY_HCA_CAP:
+	case MLX5_CMD_OP_QUERY_VPORT_STATE:
+	case MLX5_CMD_OP_QUERY_ADAPTER:
+	case MLX5_CMD_OP_QUERY_ISSI:
+	case MLX5_CMD_OP_QUERY_NIC_VPORT_CONTEXT:
+	case MLX5_CMD_OP_QUERY_ROCE_ADDRESS:
+	case MLX5_CMD_OP_QUERY_VNIC_ENV:
+	case MLX5_CMD_OP_QUERY_VPORT_COUNTER:
+	case MLX5_CMD_OP_GET_DROPPED_PACKET_LOG:
+	case MLX5_CMD_OP_NOP:
+	case MLX5_CMD_OP_QUERY_CONG_STATUS:
+	case MLX5_CMD_OP_QUERY_CONG_PARAMS:
+	case MLX5_CMD_OP_QUERY_CONG_STATISTICS:
+		return true;
+	default:
+		return false;
+	}
+}
+
+static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_QUERY_EQN)(
+	struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs)
+{
+	struct mlx5_ib_ucontext *c;
+	struct mlx5_ib_dev *dev;
+	int user_vector;
+	int dev_eqn;
+	unsigned int irqn;
+	int err;
+
+	if (uverbs_copy_from(&user_vector, attrs,
+			     MLX5_IB_ATTR_DEVX_QUERY_EQN_USER_VEC))
+		return -EFAULT;
+
+	c = devx_ufile2uctx(file);
+	if (IS_ERR(c))
+		return PTR_ERR(c);
+	dev = to_mdev(c->ibucontext.device);
+
+	err = mlx5_vector2eqn(dev->mdev, user_vector, &dev_eqn, &irqn);
+	if (err < 0)
+		return err;
+
+	if (uverbs_copy_to(attrs, MLX5_IB_ATTR_DEVX_QUERY_EQN_DEV_EQN,
+			   &dev_eqn, sizeof(dev_eqn)))
+		return -EFAULT;
+
+	return 0;
+}
+
+/*
+ *Security note:
+ * The hardware protection mechanism works like this: Each device object that
+ * is subject to UAR doorbells (QP/SQ/CQ) gets a UAR ID (called uar_page in
+ * the device specification manual) upon its creation. Then upon doorbell,
+ * hardware fetches the object context for which the doorbell was rang, and
+ * validates that the UAR through which the DB was rang matches the UAR ID
+ * of the object.
+ * If no match the doorbell is silently ignored by the hardware. Of course,
+ * the user cannot ring a doorbell on a UAR that was not mapped to it.
+ * Now in devx, as the devx kernel does not manipulate the QP/SQ/CQ command
+ * mailboxes (except tagging them with UID), we expose to the user its UAR
+ * ID, so it can embed it in these objects in the expected specification
+ * format. So the only thing the user can do is hurt itself by creating a
+ * QP/SQ/CQ with a UAR ID other than his, and then in this case other users
+ * may ring a doorbell on its objects.
+ * The consequence of that will be that another user can schedule a QP/SQ
+ * of the buggy user for execution (just insert it to the hardware schedule
+ * queue or arm its CQ for event generation), no further harm is expected.
+ */
+static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_QUERY_UAR)(
+	struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs)
+{
+	struct mlx5_ib_ucontext *c;
+	struct mlx5_ib_dev *dev;
+	u32 user_idx;
+	s32 dev_idx;
+
+	c = devx_ufile2uctx(file);
+	if (IS_ERR(c))
+		return PTR_ERR(c);
+	dev = to_mdev(c->ibucontext.device);
+
+	if (uverbs_copy_from(&user_idx, attrs,
+			     MLX5_IB_ATTR_DEVX_QUERY_UAR_USER_IDX))
+		return -EFAULT;
+
+	dev_idx = bfregn_to_uar_index(dev, &c->bfregi, user_idx, true);
+	if (dev_idx < 0)
+		return dev_idx;
+
+	if (uverbs_copy_to(attrs, MLX5_IB_ATTR_DEVX_QUERY_UAR_DEV_IDX,
+			   &dev_idx, sizeof(dev_idx)))
+		return -EFAULT;
+
+	return 0;
+}
+
+static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OTHER)(
+	struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs)
+{
+	struct mlx5_ib_ucontext *c;
+	struct mlx5_ib_dev *dev;
+	void *cmd_in = uverbs_attr_get_alloced_ptr(
+		attrs, MLX5_IB_ATTR_DEVX_OTHER_CMD_IN);
+	int cmd_out_len = uverbs_attr_get_len(attrs,
+					MLX5_IB_ATTR_DEVX_OTHER_CMD_OUT);
+	void *cmd_out;
+	int err;
+
+	c = devx_ufile2uctx(file);
+	if (IS_ERR(c))
+		return PTR_ERR(c);
+	dev = to_mdev(c->ibucontext.device);
+
+	if (!c->devx_uid)
+		return -EPERM;
+
+	/* Only white list of some general HCA commands are allowed for this method. */
+	if (!devx_is_general_cmd(cmd_in))
+		return -EINVAL;
+
+	cmd_out = uverbs_zalloc(attrs, cmd_out_len);
+	if (IS_ERR(cmd_out))
+		return PTR_ERR(cmd_out);
+
+	MLX5_SET(general_obj_in_cmd_hdr, cmd_in, uid, c->devx_uid);
+	err = mlx5_cmd_exec(dev->mdev, cmd_in,
+			    uverbs_attr_get_len(attrs, MLX5_IB_ATTR_DEVX_OTHER_CMD_IN),
+			    cmd_out, cmd_out_len);
+	if (err)
+		return err;
+
+	return uverbs_copy_to(attrs, MLX5_IB_ATTR_DEVX_OTHER_CMD_OUT, cmd_out,
+			      cmd_out_len);
+}
+
+static void devx_obj_build_destroy_cmd(void *in, void *out, void *din,
+				       u32 *dinlen,
+				       u32 *obj_id)
+{
+	u16 obj_type = MLX5_GET(general_obj_in_cmd_hdr, in, obj_type);
+	u16 uid = MLX5_GET(general_obj_in_cmd_hdr, in, uid);
+
+	*obj_id = MLX5_GET(general_obj_out_cmd_hdr, out, obj_id);
+	*dinlen = MLX5_ST_SZ_BYTES(general_obj_in_cmd_hdr);
+
+	MLX5_SET(general_obj_in_cmd_hdr, din, obj_id, *obj_id);
+	MLX5_SET(general_obj_in_cmd_hdr, din, uid, uid);
+
+	switch (MLX5_GET(general_obj_in_cmd_hdr, in, opcode)) {
+	case MLX5_CMD_OP_CREATE_GENERAL_OBJECT:
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode, MLX5_CMD_OP_DESTROY_GENERAL_OBJECT);
+		MLX5_SET(general_obj_in_cmd_hdr, din, obj_type, obj_type);
+		break;
+
+	case MLX5_CMD_OP_CREATE_MKEY:
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode, MLX5_CMD_OP_DESTROY_MKEY);
+		break;
+	case MLX5_CMD_OP_CREATE_CQ:
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode, MLX5_CMD_OP_DESTROY_CQ);
+		break;
+	case MLX5_CMD_OP_ALLOC_PD:
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode, MLX5_CMD_OP_DEALLOC_PD);
+		break;
+	case MLX5_CMD_OP_ALLOC_TRANSPORT_DOMAIN:
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode,
+			 MLX5_CMD_OP_DEALLOC_TRANSPORT_DOMAIN);
+		break;
+	case MLX5_CMD_OP_CREATE_RMP:
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode, MLX5_CMD_OP_DESTROY_RMP);
+		break;
+	case MLX5_CMD_OP_CREATE_SQ:
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode, MLX5_CMD_OP_DESTROY_SQ);
+		break;
+	case MLX5_CMD_OP_CREATE_RQ:
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode, MLX5_CMD_OP_DESTROY_RQ);
+		break;
+	case MLX5_CMD_OP_CREATE_RQT:
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode, MLX5_CMD_OP_DESTROY_RQT);
+		break;
+	case MLX5_CMD_OP_CREATE_TIR:
+		*obj_id = MLX5_GET(create_tir_out, out, tirn);
+		MLX5_SET(destroy_tir_in, din, opcode, MLX5_CMD_OP_DESTROY_TIR);
+		MLX5_SET(destroy_tir_in, din, tirn, *obj_id);
+		break;
+	case MLX5_CMD_OP_CREATE_TIS:
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode, MLX5_CMD_OP_DESTROY_TIS);
+		break;
+	case MLX5_CMD_OP_ALLOC_Q_COUNTER:
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode,
+			 MLX5_CMD_OP_DEALLOC_Q_COUNTER);
+		break;
+	case MLX5_CMD_OP_CREATE_FLOW_TABLE:
+		*dinlen = MLX5_ST_SZ_BYTES(destroy_flow_table_in);
+		*obj_id = MLX5_GET(create_flow_table_out, out, table_id);
+		MLX5_SET(destroy_flow_table_in, din, other_vport,
+			 MLX5_GET(create_flow_table_in,  in, other_vport));
+		MLX5_SET(destroy_flow_table_in, din, vport_number,
+			 MLX5_GET(create_flow_table_in,  in, vport_number));
+		MLX5_SET(destroy_flow_table_in, din, table_type,
+			 MLX5_GET(create_flow_table_in,  in, table_type));
+		MLX5_SET(destroy_flow_table_in, din, table_id, *obj_id);
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode,
+			 MLX5_CMD_OP_DESTROY_FLOW_TABLE);
+		break;
+	case MLX5_CMD_OP_CREATE_FLOW_GROUP:
+		*dinlen = MLX5_ST_SZ_BYTES(destroy_flow_group_in);
+		*obj_id = MLX5_GET(create_flow_group_out, out, group_id);
+		MLX5_SET(destroy_flow_group_in, din, other_vport,
+			 MLX5_GET(create_flow_group_in, in, other_vport));
+		MLX5_SET(destroy_flow_group_in, din, vport_number,
+			 MLX5_GET(create_flow_group_in, in, vport_number));
+		MLX5_SET(destroy_flow_group_in, din, table_type,
+			 MLX5_GET(create_flow_group_in, in, table_type));
+		MLX5_SET(destroy_flow_group_in, din, table_id,
+			 MLX5_GET(create_flow_group_in, in, table_id));
+		MLX5_SET(destroy_flow_group_in, din, group_id, *obj_id);
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode,
+			 MLX5_CMD_OP_DESTROY_FLOW_GROUP);
+		break;
+	case MLX5_CMD_OP_SET_FLOW_TABLE_ENTRY:
+		*dinlen = MLX5_ST_SZ_BYTES(delete_fte_in);
+		*obj_id = MLX5_GET(set_fte_in, in, flow_index);
+		MLX5_SET(delete_fte_in, din, other_vport,
+			 MLX5_GET(set_fte_in,  in, other_vport));
+		MLX5_SET(delete_fte_in, din, vport_number,
+			 MLX5_GET(set_fte_in, in, vport_number));
+		MLX5_SET(delete_fte_in, din, table_type,
+			 MLX5_GET(set_fte_in, in, table_type));
+		MLX5_SET(delete_fte_in, din, table_id,
+			 MLX5_GET(set_fte_in, in, table_id));
+		MLX5_SET(delete_fte_in, din, flow_index, *obj_id);
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode,
+			 MLX5_CMD_OP_DELETE_FLOW_TABLE_ENTRY);
+		break;
+	case MLX5_CMD_OP_ALLOC_FLOW_COUNTER:
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode,
+			 MLX5_CMD_OP_DEALLOC_FLOW_COUNTER);
+		break;
+	case MLX5_CMD_OP_ALLOC_ENCAP_HEADER:
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode,
+			 MLX5_CMD_OP_DEALLOC_ENCAP_HEADER);
+		break;
+	case MLX5_CMD_OP_ALLOC_MODIFY_HEADER_CONTEXT:
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode,
+			 MLX5_CMD_OP_DEALLOC_MODIFY_HEADER_CONTEXT);
+		break;
+	case MLX5_CMD_OP_CREATE_SCHEDULING_ELEMENT:
+		*dinlen = MLX5_ST_SZ_BYTES(destroy_scheduling_element_in);
+		*obj_id = MLX5_GET(create_scheduling_element_out, out,
+				   scheduling_element_id);
+		MLX5_SET(destroy_scheduling_element_in, din,
+			 scheduling_hierarchy,
+			 MLX5_GET(create_scheduling_element_in, in,
+				  scheduling_hierarchy));
+		MLX5_SET(destroy_scheduling_element_in, din,
+			 scheduling_element_id, *obj_id);
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode,
+			 MLX5_CMD_OP_DESTROY_SCHEDULING_ELEMENT);
+		break;
+	case MLX5_CMD_OP_ADD_VXLAN_UDP_DPORT:
+		*dinlen = MLX5_ST_SZ_BYTES(delete_vxlan_udp_dport_in);
+		*obj_id = MLX5_GET(add_vxlan_udp_dport_in, in, vxlan_udp_port);
+		MLX5_SET(delete_vxlan_udp_dport_in, din, vxlan_udp_port, *obj_id);
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode,
+			 MLX5_CMD_OP_DELETE_VXLAN_UDP_DPORT);
+		break;
+	case MLX5_CMD_OP_SET_L2_TABLE_ENTRY:
+		*dinlen = MLX5_ST_SZ_BYTES(delete_l2_table_entry_in);
+		*obj_id = MLX5_GET(set_l2_table_entry_in, in, table_index);
+		MLX5_SET(delete_l2_table_entry_in, din, table_index, *obj_id);
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode,
+			 MLX5_CMD_OP_DELETE_L2_TABLE_ENTRY);
+		break;
+	case MLX5_CMD_OP_CREATE_QP:
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode, MLX5_CMD_OP_DESTROY_QP);
+		break;
+	case MLX5_CMD_OP_CREATE_SRQ:
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode, MLX5_CMD_OP_DESTROY_SRQ);
+		break;
+	case MLX5_CMD_OP_CREATE_XRC_SRQ:
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode,
+			 MLX5_CMD_OP_DESTROY_XRC_SRQ);
+		break;
+	case MLX5_CMD_OP_CREATE_DCT:
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode, MLX5_CMD_OP_DESTROY_DCT);
+		break;
+	case MLX5_CMD_OP_CREATE_XRQ:
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode, MLX5_CMD_OP_DESTROY_XRQ);
+		break;
+	case MLX5_CMD_OP_ATTACH_TO_MCG:
+		*dinlen = MLX5_ST_SZ_BYTES(detach_from_mcg_in);
+		MLX5_SET(detach_from_mcg_in, din, qpn,
+			 MLX5_GET(attach_to_mcg_in, in, qpn));
+		memcpy(MLX5_ADDR_OF(detach_from_mcg_in, din, multicast_gid),
+		       MLX5_ADDR_OF(attach_to_mcg_in, in, multicast_gid),
+		       MLX5_FLD_SZ_BYTES(attach_to_mcg_in, multicast_gid));
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode, MLX5_CMD_OP_DETACH_FROM_MCG);
+		break;
+	case MLX5_CMD_OP_ALLOC_XRCD:
+		MLX5_SET(general_obj_in_cmd_hdr, din, opcode, MLX5_CMD_OP_DEALLOC_XRCD);
+		break;
+	default:
+		/* The entry must match to one of the devx_is_obj_create_cmd */
+		WARN_ON(true);
+		break;
+	}
+}
+
+static int devx_obj_cleanup(struct ib_uobject *uobject,
+			    enum rdma_remove_reason why)
+{
+	u32 out[MLX5_ST_SZ_DW(general_obj_out_cmd_hdr)];
+	struct devx_obj *obj = uobject->object;
+	int ret;
+
+	ret = mlx5_cmd_exec(obj->mdev, obj->dinbox, obj->dinlen, out, sizeof(out));
+	if (ib_is_destroy_retryable(ret, why, uobject))
+		return ret;
+
+	kfree(obj);
+	return ret;
+}
+
+static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_CREATE)(
+	struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs)
+{
+	void *cmd_in = uverbs_attr_get_alloced_ptr(attrs, MLX5_IB_ATTR_DEVX_OBJ_CREATE_CMD_IN);
+	int cmd_out_len =  uverbs_attr_get_len(attrs,
+					MLX5_IB_ATTR_DEVX_OBJ_CREATE_CMD_OUT);
+	void *cmd_out;
+	struct ib_uobject *uobj = uverbs_attr_get_uobject(
+		attrs, MLX5_IB_ATTR_DEVX_OBJ_CREATE_HANDLE);
+	struct mlx5_ib_ucontext *c = to_mucontext(uobj->context);
+	struct mlx5_ib_dev *dev = to_mdev(c->ibucontext.device);
+	u32 out[MLX5_ST_SZ_DW(general_obj_out_cmd_hdr)];
+	struct devx_obj *obj;
+	int err;
+
+	if (!c->devx_uid)
+		return -EPERM;
+
+	if (!devx_is_obj_create_cmd(cmd_in))
+		return -EINVAL;
+
+	cmd_out = uverbs_zalloc(attrs, cmd_out_len);
+	if (IS_ERR(cmd_out))
+		return PTR_ERR(cmd_out);
+
+	obj = kzalloc(sizeof(struct devx_obj), GFP_KERNEL);
+	if (!obj)
+		return -ENOMEM;
+
+	MLX5_SET(general_obj_in_cmd_hdr, cmd_in, uid, c->devx_uid);
+	err = mlx5_cmd_exec(dev->mdev, cmd_in,
+			    uverbs_attr_get_len(attrs, MLX5_IB_ATTR_DEVX_OBJ_CREATE_CMD_IN),
+			    cmd_out, cmd_out_len);
+	if (err)
+		goto obj_free;
+
+	uobj->object = obj;
+	obj->mdev = dev->mdev;
+	devx_obj_build_destroy_cmd(cmd_in, cmd_out, obj->dinbox, &obj->dinlen, &obj->obj_id);
+	WARN_ON(obj->dinlen > MLX5_MAX_DESTROY_INBOX_SIZE_DW * sizeof(u32));
+
+	err = uverbs_copy_to(attrs, MLX5_IB_ATTR_DEVX_OBJ_CREATE_CMD_OUT, cmd_out, cmd_out_len);
+	if (err)
+		goto obj_destroy;
+
+	return 0;
+
+obj_destroy:
+	mlx5_cmd_exec(obj->mdev, obj->dinbox, obj->dinlen, out, sizeof(out));
+obj_free:
+	kfree(obj);
+	return err;
+}
+
+static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_MODIFY)(
+	struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs)
+{
+	void *cmd_in = uverbs_attr_get_alloced_ptr(attrs, MLX5_IB_ATTR_DEVX_OBJ_MODIFY_CMD_IN);
+	int cmd_out_len = uverbs_attr_get_len(attrs,
+					MLX5_IB_ATTR_DEVX_OBJ_MODIFY_CMD_OUT);
+	struct ib_uobject *uobj = uverbs_attr_get_uobject(attrs,
+							  MLX5_IB_ATTR_DEVX_OBJ_MODIFY_HANDLE);
+	struct mlx5_ib_ucontext *c = to_mucontext(uobj->context);
+	struct devx_obj *obj = uobj->object;
+	void *cmd_out;
+	int err;
+
+	if (!c->devx_uid)
+		return -EPERM;
+
+	if (!devx_is_obj_modify_cmd(cmd_in))
+		return -EINVAL;
+
+	if (!devx_is_valid_obj_id(obj, cmd_in))
+		return -EINVAL;
+
+	cmd_out = uverbs_zalloc(attrs, cmd_out_len);
+	if (IS_ERR(cmd_out))
+		return PTR_ERR(cmd_out);
+
+	MLX5_SET(general_obj_in_cmd_hdr, cmd_in, uid, c->devx_uid);
+	err = mlx5_cmd_exec(obj->mdev, cmd_in,
+			    uverbs_attr_get_len(attrs, MLX5_IB_ATTR_DEVX_OBJ_MODIFY_CMD_IN),
+			    cmd_out, cmd_out_len);
+	if (err)
+		return err;
+
+	return uverbs_copy_to(attrs, MLX5_IB_ATTR_DEVX_OBJ_MODIFY_CMD_OUT,
+			      cmd_out, cmd_out_len);
+}
+
+static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_QUERY)(
+	struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs)
+{
+	void *cmd_in = uverbs_attr_get_alloced_ptr(attrs, MLX5_IB_ATTR_DEVX_OBJ_QUERY_CMD_IN);
+	int cmd_out_len = uverbs_attr_get_len(attrs,
+					      MLX5_IB_ATTR_DEVX_OBJ_QUERY_CMD_OUT);
+	struct ib_uobject *uobj = uverbs_attr_get_uobject(attrs,
+							  MLX5_IB_ATTR_DEVX_OBJ_QUERY_HANDLE);
+	struct mlx5_ib_ucontext *c = to_mucontext(uobj->context);
+	struct devx_obj *obj = uobj->object;
+	void *cmd_out;
+	int err;
+
+	if (!c->devx_uid)
+		return -EPERM;
+
+	if (!devx_is_obj_query_cmd(cmd_in))
+		return -EINVAL;
+
+	if (!devx_is_valid_obj_id(obj, cmd_in))
+		return -EINVAL;
+
+	cmd_out = uverbs_zalloc(attrs, cmd_out_len);
+	if (IS_ERR(cmd_out))
+		return PTR_ERR(cmd_out);
+
+	MLX5_SET(general_obj_in_cmd_hdr, cmd_in, uid, c->devx_uid);
+	err = mlx5_cmd_exec(obj->mdev, cmd_in,
+			    uverbs_attr_get_len(attrs, MLX5_IB_ATTR_DEVX_OBJ_QUERY_CMD_IN),
+			    cmd_out, cmd_out_len);
+	if (err)
+		return err;
+
+	return uverbs_copy_to(attrs, MLX5_IB_ATTR_DEVX_OBJ_QUERY_CMD_OUT,
+			      cmd_out, cmd_out_len);
+}
+
+static int devx_umem_get(struct mlx5_ib_dev *dev, struct ib_ucontext *ucontext,
+			 struct uverbs_attr_bundle *attrs,
+			 struct devx_umem *obj)
+{
+	u64 addr;
+	size_t size;
+	u32 access;
+	int npages;
+	int err;
+	u32 page_mask;
+
+	if (uverbs_copy_from(&addr, attrs, MLX5_IB_ATTR_DEVX_UMEM_REG_ADDR) ||
+	    uverbs_copy_from(&size, attrs, MLX5_IB_ATTR_DEVX_UMEM_REG_LEN))
+		return -EFAULT;
+
+	err = uverbs_get_flags32(&access, attrs,
+				 MLX5_IB_ATTR_DEVX_UMEM_REG_ACCESS,
+				 IB_ACCESS_LOCAL_WRITE |
+				 IB_ACCESS_REMOTE_WRITE |
+				 IB_ACCESS_REMOTE_READ);
+	if (err)
+		return err;
+
+	err = ib_check_mr_access(access);
+	if (err)
+		return err;
+
+	obj->umem = ib_umem_get(ucontext, addr, size, access, 0);
+	if (IS_ERR(obj->umem))
+		return PTR_ERR(obj->umem);
+
+	mlx5_ib_cont_pages(obj->umem, obj->umem->address,
+			   MLX5_MKEY_PAGE_SHIFT_MASK, &npages,
+			   &obj->page_shift, &obj->ncont, NULL);
+
+	if (!npages) {
+		ib_umem_release(obj->umem);
+		return -EINVAL;
+	}
+
+	page_mask = (1 << obj->page_shift) - 1;
+	obj->page_offset = obj->umem->address & page_mask;
+
+	return 0;
+}
+
+static int devx_umem_reg_cmd_alloc(struct uverbs_attr_bundle *attrs,
+				   struct devx_umem *obj,
+				   struct devx_umem_reg_cmd *cmd)
+{
+	cmd->inlen = MLX5_ST_SZ_BYTES(create_umem_in) +
+		    (MLX5_ST_SZ_BYTES(mtt) * obj->ncont);
+	cmd->in = uverbs_zalloc(attrs, cmd->inlen);
+	return PTR_ERR_OR_ZERO(cmd->in);
+}
+
+static void devx_umem_reg_cmd_build(struct mlx5_ib_dev *dev,
+				    struct devx_umem *obj,
+				    struct devx_umem_reg_cmd *cmd)
+{
+	void *umem;
+	__be64 *mtt;
+
+	umem = MLX5_ADDR_OF(create_umem_in, cmd->in, umem);
+	mtt = (__be64 *)MLX5_ADDR_OF(umem, umem, mtt);
+
+	MLX5_SET(general_obj_in_cmd_hdr, cmd->in, opcode, MLX5_CMD_OP_CREATE_GENERAL_OBJECT);
+	MLX5_SET(general_obj_in_cmd_hdr, cmd->in, obj_type, MLX5_OBJ_TYPE_UMEM);
+	MLX5_SET64(umem, umem, num_of_mtt, obj->ncont);
+	MLX5_SET(umem, umem, log_page_size, obj->page_shift -
+					    MLX5_ADAPTER_PAGE_SHIFT);
+	MLX5_SET(umem, umem, page_offset, obj->page_offset);
+	mlx5_ib_populate_pas(dev, obj->umem, obj->page_shift, mtt,
+			     (obj->umem->writable ? MLX5_IB_MTT_WRITE : 0) |
+			     MLX5_IB_MTT_READ);
+}
+
+static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_UMEM_REG)(
+	struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs)
+{
+	struct devx_umem_reg_cmd cmd;
+	struct devx_umem *obj;
+	struct ib_uobject *uobj = uverbs_attr_get_uobject(
+		attrs, MLX5_IB_ATTR_DEVX_UMEM_REG_HANDLE);
+	u32 obj_id;
+	struct mlx5_ib_ucontext *c = to_mucontext(uobj->context);
+	struct mlx5_ib_dev *dev = to_mdev(c->ibucontext.device);
+	int err;
+
+	if (!c->devx_uid)
+		return -EPERM;
+
+	obj = kzalloc(sizeof(struct devx_umem), GFP_KERNEL);
+	if (!obj)
+		return -ENOMEM;
+
+	err = devx_umem_get(dev, &c->ibucontext, attrs, obj);
+	if (err)
+		goto err_obj_free;
+
+	err = devx_umem_reg_cmd_alloc(attrs, obj, &cmd);
+	if (err)
+		goto err_umem_release;
+
+	devx_umem_reg_cmd_build(dev, obj, &cmd);
+
+	MLX5_SET(general_obj_in_cmd_hdr, cmd.in, uid, c->devx_uid);
+	err = mlx5_cmd_exec(dev->mdev, cmd.in, cmd.inlen, cmd.out,
+			    sizeof(cmd.out));
+	if (err)
+		goto err_umem_release;
+
+	obj->mdev = dev->mdev;
+	uobj->object = obj;
+	devx_obj_build_destroy_cmd(cmd.in, cmd.out, obj->dinbox, &obj->dinlen, &obj_id);
+	err = uverbs_copy_to(attrs, MLX5_IB_ATTR_DEVX_UMEM_REG_OUT_ID, &obj_id, sizeof(obj_id));
+	if (err)
+		goto err_umem_destroy;
+
+	return 0;
+
+err_umem_destroy:
+	mlx5_cmd_exec(obj->mdev, obj->dinbox, obj->dinlen, cmd.out, sizeof(cmd.out));
+err_umem_release:
+	ib_umem_release(obj->umem);
+err_obj_free:
+	kfree(obj);
+	return err;
+}
+
+static int devx_umem_cleanup(struct ib_uobject *uobject,
+			     enum rdma_remove_reason why)
+{
+	struct devx_umem *obj = uobject->object;
+	u32 out[MLX5_ST_SZ_DW(general_obj_out_cmd_hdr)];
+	int err;
+
+	err = mlx5_cmd_exec(obj->mdev, obj->dinbox, obj->dinlen, out, sizeof(out));
+	if (ib_is_destroy_retryable(err, why, uobject))
+		return err;
+
+	ib_umem_release(obj->umem);
+	kfree(obj);
+	return 0;
+}
+
+DECLARE_UVERBS_NAMED_METHOD(
+	MLX5_IB_METHOD_DEVX_UMEM_REG,
+	UVERBS_ATTR_IDR(MLX5_IB_ATTR_DEVX_UMEM_REG_HANDLE,
+			MLX5_IB_OBJECT_DEVX_UMEM,
+			UVERBS_ACCESS_NEW,
+			UA_MANDATORY),
+	UVERBS_ATTR_PTR_IN(MLX5_IB_ATTR_DEVX_UMEM_REG_ADDR,
+			   UVERBS_ATTR_TYPE(u64),
+			   UA_MANDATORY),
+	UVERBS_ATTR_PTR_IN(MLX5_IB_ATTR_DEVX_UMEM_REG_LEN,
+			   UVERBS_ATTR_TYPE(u64),
+			   UA_MANDATORY),
+	UVERBS_ATTR_FLAGS_IN(MLX5_IB_ATTR_DEVX_UMEM_REG_ACCESS,
+			     enum ib_access_flags),
+	UVERBS_ATTR_PTR_OUT(MLX5_IB_ATTR_DEVX_UMEM_REG_OUT_ID,
+			    UVERBS_ATTR_TYPE(u32),
+			    UA_MANDATORY));
+
+DECLARE_UVERBS_NAMED_METHOD_DESTROY(
+	MLX5_IB_METHOD_DEVX_UMEM_DEREG,
+	UVERBS_ATTR_IDR(MLX5_IB_ATTR_DEVX_UMEM_DEREG_HANDLE,
+			MLX5_IB_OBJECT_DEVX_UMEM,
+			UVERBS_ACCESS_DESTROY,
+			UA_MANDATORY));
+
+DECLARE_UVERBS_NAMED_METHOD(
+	MLX5_IB_METHOD_DEVX_QUERY_EQN,
+	UVERBS_ATTR_PTR_IN(MLX5_IB_ATTR_DEVX_QUERY_EQN_USER_VEC,
+			   UVERBS_ATTR_TYPE(u32),
+			   UA_MANDATORY),
+	UVERBS_ATTR_PTR_OUT(MLX5_IB_ATTR_DEVX_QUERY_EQN_DEV_EQN,
+			    UVERBS_ATTR_TYPE(u32),
+			    UA_MANDATORY));
+
+DECLARE_UVERBS_NAMED_METHOD(
+	MLX5_IB_METHOD_DEVX_QUERY_UAR,
+	UVERBS_ATTR_PTR_IN(MLX5_IB_ATTR_DEVX_QUERY_UAR_USER_IDX,
+			   UVERBS_ATTR_TYPE(u32),
+			   UA_MANDATORY),
+	UVERBS_ATTR_PTR_OUT(MLX5_IB_ATTR_DEVX_QUERY_UAR_DEV_IDX,
+			    UVERBS_ATTR_TYPE(u32),
+			    UA_MANDATORY));
+
+DECLARE_UVERBS_NAMED_METHOD(
+	MLX5_IB_METHOD_DEVX_OTHER,
+	UVERBS_ATTR_PTR_IN(
+		MLX5_IB_ATTR_DEVX_OTHER_CMD_IN,
+		UVERBS_ATTR_MIN_SIZE(MLX5_ST_SZ_BYTES(general_obj_in_cmd_hdr)),
+		UA_MANDATORY,
+		UA_ALLOC_AND_COPY),
+	UVERBS_ATTR_PTR_OUT(
+		MLX5_IB_ATTR_DEVX_OTHER_CMD_OUT,
+		UVERBS_ATTR_MIN_SIZE(MLX5_ST_SZ_BYTES(general_obj_out_cmd_hdr)),
+		UA_MANDATORY));
+
+DECLARE_UVERBS_NAMED_METHOD(
+	MLX5_IB_METHOD_DEVX_OBJ_CREATE,
+	UVERBS_ATTR_IDR(MLX5_IB_ATTR_DEVX_OBJ_CREATE_HANDLE,
+			MLX5_IB_OBJECT_DEVX_OBJ,
+			UVERBS_ACCESS_NEW,
+			UA_MANDATORY),
+	UVERBS_ATTR_PTR_IN(
+		MLX5_IB_ATTR_DEVX_OBJ_CREATE_CMD_IN,
+		UVERBS_ATTR_MIN_SIZE(MLX5_ST_SZ_BYTES(general_obj_in_cmd_hdr)),
+		UA_MANDATORY,
+		UA_ALLOC_AND_COPY),
+	UVERBS_ATTR_PTR_OUT(
+		MLX5_IB_ATTR_DEVX_OBJ_CREATE_CMD_OUT,
+		UVERBS_ATTR_MIN_SIZE(MLX5_ST_SZ_BYTES(general_obj_out_cmd_hdr)),
+		UA_MANDATORY));
+
+DECLARE_UVERBS_NAMED_METHOD_DESTROY(
+	MLX5_IB_METHOD_DEVX_OBJ_DESTROY,
+	UVERBS_ATTR_IDR(MLX5_IB_ATTR_DEVX_OBJ_DESTROY_HANDLE,
+			MLX5_IB_OBJECT_DEVX_OBJ,
+			UVERBS_ACCESS_DESTROY,
+			UA_MANDATORY));
+
+DECLARE_UVERBS_NAMED_METHOD(
+	MLX5_IB_METHOD_DEVX_OBJ_MODIFY,
+	UVERBS_ATTR_IDR(MLX5_IB_ATTR_DEVX_OBJ_MODIFY_HANDLE,
+			MLX5_IB_OBJECT_DEVX_OBJ,
+			UVERBS_ACCESS_WRITE,
+			UA_MANDATORY),
+	UVERBS_ATTR_PTR_IN(
+		MLX5_IB_ATTR_DEVX_OBJ_MODIFY_CMD_IN,
+		UVERBS_ATTR_MIN_SIZE(MLX5_ST_SZ_BYTES(general_obj_in_cmd_hdr)),
+		UA_MANDATORY,
+		UA_ALLOC_AND_COPY),
+	UVERBS_ATTR_PTR_OUT(
+		MLX5_IB_ATTR_DEVX_OBJ_MODIFY_CMD_OUT,
+		UVERBS_ATTR_MIN_SIZE(MLX5_ST_SZ_BYTES(general_obj_out_cmd_hdr)),
+		UA_MANDATORY));
+
+DECLARE_UVERBS_NAMED_METHOD(
+	MLX5_IB_METHOD_DEVX_OBJ_QUERY,
+	UVERBS_ATTR_IDR(MLX5_IB_ATTR_DEVX_OBJ_QUERY_HANDLE,
+			MLX5_IB_OBJECT_DEVX_OBJ,
+			UVERBS_ACCESS_READ,
+			UA_MANDATORY),
+	UVERBS_ATTR_PTR_IN(
+		MLX5_IB_ATTR_DEVX_OBJ_QUERY_CMD_IN,
+		UVERBS_ATTR_MIN_SIZE(MLX5_ST_SZ_BYTES(general_obj_in_cmd_hdr)),
+		UA_MANDATORY,
+		UA_ALLOC_AND_COPY),
+	UVERBS_ATTR_PTR_OUT(
+		MLX5_IB_ATTR_DEVX_OBJ_QUERY_CMD_OUT,
+		UVERBS_ATTR_MIN_SIZE(MLX5_ST_SZ_BYTES(general_obj_out_cmd_hdr)),
+		UA_MANDATORY));
+
+DECLARE_UVERBS_GLOBAL_METHODS(MLX5_IB_OBJECT_DEVX,
+			      &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_OTHER),
+			      &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_QUERY_UAR),
+			      &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_QUERY_EQN));
+
+DECLARE_UVERBS_NAMED_OBJECT(MLX5_IB_OBJECT_DEVX_OBJ,
+			    UVERBS_TYPE_ALLOC_IDR(devx_obj_cleanup),
+			    &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_OBJ_CREATE),
+			    &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_OBJ_DESTROY),
+			    &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_OBJ_MODIFY),
+			    &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_OBJ_QUERY));
+
+DECLARE_UVERBS_NAMED_OBJECT(MLX5_IB_OBJECT_DEVX_UMEM,
+			    UVERBS_TYPE_ALLOC_IDR(devx_umem_cleanup),
+			    &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_UMEM_REG),
+			    &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_UMEM_DEREG));
+
+DECLARE_UVERBS_OBJECT_TREE(devx_objects,
+			   &UVERBS_OBJECT(MLX5_IB_OBJECT_DEVX),
+			   &UVERBS_OBJECT(MLX5_IB_OBJECT_DEVX_OBJ),
+			   &UVERBS_OBJECT(MLX5_IB_OBJECT_DEVX_UMEM));
+
+const struct uverbs_object_tree_def *mlx5_ib_get_devx_tree(void)
+{
+	return &devx_objects;
+}
diff --git a/drivers/infiniband/hw/mlx5/doorbell.c b/drivers/infiniband/hw/mlx5/doorbell.c
new file mode 100644
index 000000000..a0e4e6ddb
--- /dev/null
+++ b/drivers/infiniband/hw/mlx5/doorbell.c
@@ -0,0 +1,98 @@
+/*
+ * Copyright (c) 2013-2015, Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/kref.h>
+#include <linux/slab.h>
+#include <rdma/ib_umem.h>
+
+#include "mlx5_ib.h"
+
+struct mlx5_ib_user_db_page {
+	struct list_head	list;
+	struct ib_umem	       *umem;
+	unsigned long		user_virt;
+	int			refcnt;
+};
+
+int mlx5_ib_db_map_user(struct mlx5_ib_ucontext *context, unsigned long virt,
+			struct mlx5_db *db)
+{
+	struct mlx5_ib_user_db_page *page;
+	int err = 0;
+
+	mutex_lock(&context->db_page_mutex);
+
+	list_for_each_entry(page, &context->db_page_list, list)
+		if (page->user_virt == (virt & PAGE_MASK))
+			goto found;
+
+	page = kmalloc(sizeof(*page), GFP_KERNEL);
+	if (!page) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	page->user_virt = (virt & PAGE_MASK);
+	page->refcnt    = 0;
+	page->umem      = ib_umem_get(&context->ibucontext, virt & PAGE_MASK,
+				      PAGE_SIZE, 0, 0);
+	if (IS_ERR(page->umem)) {
+		err = PTR_ERR(page->umem);
+		kfree(page);
+		goto out;
+	}
+
+	list_add(&page->list, &context->db_page_list);
+
+found:
+	db->dma = sg_dma_address(page->umem->sg_head.sgl) + (virt & ~PAGE_MASK);
+	db->u.user_page = page;
+	++page->refcnt;
+
+out:
+	mutex_unlock(&context->db_page_mutex);
+
+	return err;
+}
+
+void mlx5_ib_db_unmap_user(struct mlx5_ib_ucontext *context, struct mlx5_db *db)
+{
+	mutex_lock(&context->db_page_mutex);
+
+	if (!--db->u.user_page->refcnt) {
+		list_del(&db->u.user_page->list);
+		ib_umem_release(db->u.user_page->umem);
+		kfree(db->u.user_page);
+	}
+
+	mutex_unlock(&context->db_page_mutex);
+}
diff --git a/drivers/infiniband/hw/mlx5/flow.c b/drivers/infiniband/hw/mlx5/flow.c
new file mode 100644
index 000000000..1a29f47f8
--- /dev/null
+++ b/drivers/infiniband/hw/mlx5/flow.c
@@ -0,0 +1,252 @@
+// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
+/*
+ * Copyright (c) 2018, Mellanox Technologies inc.  All rights reserved.
+ */
+
+#include <rdma/ib_user_verbs.h>
+#include <rdma/ib_verbs.h>
+#include <rdma/uverbs_types.h>
+#include <rdma/uverbs_ioctl.h>
+#include <rdma/mlx5_user_ioctl_cmds.h>
+#include <rdma/ib_umem.h>
+#include <linux/mlx5/driver.h>
+#include <linux/mlx5/fs.h>
+#include "mlx5_ib.h"
+
+#define UVERBS_MODULE_NAME mlx5_ib
+#include <rdma/uverbs_named_ioctl.h>
+
+static const struct uverbs_attr_spec mlx5_ib_flow_type[] = {
+	[MLX5_IB_FLOW_TYPE_NORMAL] = {
+		.type = UVERBS_ATTR_TYPE_PTR_IN,
+		.u.ptr = {
+			.len = sizeof(u16), /* data is priority */
+			.min_len = sizeof(u16),
+		}
+	},
+	[MLX5_IB_FLOW_TYPE_SNIFFER] = {
+		.type = UVERBS_ATTR_TYPE_PTR_IN,
+		UVERBS_ATTR_NO_DATA(),
+	},
+	[MLX5_IB_FLOW_TYPE_ALL_DEFAULT] = {
+		.type = UVERBS_ATTR_TYPE_PTR_IN,
+		UVERBS_ATTR_NO_DATA(),
+	},
+	[MLX5_IB_FLOW_TYPE_MC_DEFAULT] = {
+		.type = UVERBS_ATTR_TYPE_PTR_IN,
+		UVERBS_ATTR_NO_DATA(),
+	},
+};
+
+static int UVERBS_HANDLER(MLX5_IB_METHOD_CREATE_FLOW)(
+	struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs)
+{
+	struct mlx5_ib_flow_handler *flow_handler;
+	struct mlx5_ib_flow_matcher *fs_matcher;
+	void *devx_obj;
+	int dest_id, dest_type;
+	void *cmd_in;
+	int inlen;
+	bool dest_devx, dest_qp;
+	struct ib_qp *qp = NULL;
+	struct ib_uobject *uobj =
+		uverbs_attr_get_uobject(attrs, MLX5_IB_ATTR_CREATE_FLOW_HANDLE);
+	struct mlx5_ib_dev *dev = to_mdev(uobj->context->device);
+
+	if (!capable(CAP_NET_RAW))
+		return -EPERM;
+
+	dest_devx =
+		uverbs_attr_is_valid(attrs, MLX5_IB_ATTR_CREATE_FLOW_DEST_DEVX);
+	dest_qp = uverbs_attr_is_valid(attrs,
+				       MLX5_IB_ATTR_CREATE_FLOW_DEST_QP);
+
+	if ((dest_devx && dest_qp) || (!dest_devx && !dest_qp))
+		return -EINVAL;
+
+	if (dest_devx) {
+		devx_obj = uverbs_attr_get_obj(
+			attrs, MLX5_IB_ATTR_CREATE_FLOW_DEST_DEVX);
+		if (IS_ERR(devx_obj))
+			return PTR_ERR(devx_obj);
+
+		/* Verify that the given DEVX object is a flow
+		 * steering destination.
+		 */
+		if (!mlx5_ib_devx_is_flow_dest(devx_obj, &dest_id, &dest_type))
+			return -EINVAL;
+	} else {
+		struct mlx5_ib_qp *mqp;
+
+		qp = uverbs_attr_get_obj(attrs,
+					 MLX5_IB_ATTR_CREATE_FLOW_DEST_QP);
+		if (IS_ERR(qp))
+			return PTR_ERR(qp);
+
+		if (qp->qp_type != IB_QPT_RAW_PACKET)
+			return -EINVAL;
+
+		mqp = to_mqp(qp);
+		if (mqp->flags & MLX5_IB_QP_RSS)
+			dest_id = mqp->rss_qp.tirn;
+		else
+			dest_id = mqp->raw_packet_qp.rq.tirn;
+		dest_type = MLX5_FLOW_DESTINATION_TYPE_TIR;
+	}
+
+	if (dev->rep)
+		return -ENOTSUPP;
+
+	cmd_in = uverbs_attr_get_alloced_ptr(
+		attrs, MLX5_IB_ATTR_CREATE_FLOW_MATCH_VALUE);
+	inlen = uverbs_attr_get_len(attrs,
+				    MLX5_IB_ATTR_CREATE_FLOW_MATCH_VALUE);
+	fs_matcher = uverbs_attr_get_obj(attrs,
+					 MLX5_IB_ATTR_CREATE_FLOW_MATCHER);
+	flow_handler = mlx5_ib_raw_fs_rule_add(dev, fs_matcher, cmd_in, inlen,
+					       dest_id, dest_type);
+	if (IS_ERR(flow_handler))
+		return PTR_ERR(flow_handler);
+
+	ib_set_flow(uobj, &flow_handler->ibflow, qp, &dev->ib_dev);
+
+	return 0;
+}
+
+static int flow_matcher_cleanup(struct ib_uobject *uobject,
+				enum rdma_remove_reason why)
+{
+	struct mlx5_ib_flow_matcher *obj = uobject->object;
+	int ret;
+
+	ret = ib_destroy_usecnt(&obj->usecnt, why, uobject);
+	if (ret)
+		return ret;
+
+	kfree(obj);
+	return 0;
+}
+
+static int UVERBS_HANDLER(MLX5_IB_METHOD_FLOW_MATCHER_CREATE)(
+	struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs)
+{
+	struct ib_uobject *uobj = uverbs_attr_get_uobject(
+		attrs, MLX5_IB_ATTR_FLOW_MATCHER_CREATE_HANDLE);
+	struct mlx5_ib_dev *dev = to_mdev(uobj->context->device);
+	struct mlx5_ib_flow_matcher *obj;
+	int err;
+
+	obj = kzalloc(sizeof(struct mlx5_ib_flow_matcher), GFP_KERNEL);
+	if (!obj)
+		return -ENOMEM;
+
+	obj->mask_len = uverbs_attr_get_len(
+		attrs, MLX5_IB_ATTR_FLOW_MATCHER_MATCH_MASK);
+	err = uverbs_copy_from(&obj->matcher_mask,
+			       attrs,
+			       MLX5_IB_ATTR_FLOW_MATCHER_MATCH_MASK);
+	if (err)
+		goto end;
+
+	obj->flow_type = uverbs_attr_get_enum_id(
+		attrs, MLX5_IB_ATTR_FLOW_MATCHER_FLOW_TYPE);
+
+	if (obj->flow_type == MLX5_IB_FLOW_TYPE_NORMAL) {
+		err = uverbs_copy_from(&obj->priority,
+				       attrs,
+				       MLX5_IB_ATTR_FLOW_MATCHER_FLOW_TYPE);
+		if (err)
+			goto end;
+	}
+
+	err = uverbs_copy_from(&obj->match_criteria_enable,
+			       attrs,
+			       MLX5_IB_ATTR_FLOW_MATCHER_MATCH_CRITERIA);
+	if (err)
+		goto end;
+
+	uobj->object = obj;
+	obj->mdev = dev->mdev;
+	atomic_set(&obj->usecnt, 0);
+	return 0;
+
+end:
+	kfree(obj);
+	return err;
+}
+
+DECLARE_UVERBS_NAMED_METHOD(
+	MLX5_IB_METHOD_CREATE_FLOW,
+	UVERBS_ATTR_IDR(MLX5_IB_ATTR_CREATE_FLOW_HANDLE,
+			UVERBS_OBJECT_FLOW,
+			UVERBS_ACCESS_NEW,
+			UA_MANDATORY),
+	UVERBS_ATTR_PTR_IN(
+		MLX5_IB_ATTR_CREATE_FLOW_MATCH_VALUE,
+		UVERBS_ATTR_SIZE(1, sizeof(struct mlx5_ib_match_params)),
+		UA_MANDATORY,
+		UA_ALLOC_AND_COPY),
+	UVERBS_ATTR_IDR(MLX5_IB_ATTR_CREATE_FLOW_MATCHER,
+			MLX5_IB_OBJECT_FLOW_MATCHER,
+			UVERBS_ACCESS_READ,
+			UA_MANDATORY),
+	UVERBS_ATTR_IDR(MLX5_IB_ATTR_CREATE_FLOW_DEST_QP,
+			UVERBS_OBJECT_QP,
+			UVERBS_ACCESS_READ),
+	UVERBS_ATTR_IDR(MLX5_IB_ATTR_CREATE_FLOW_DEST_DEVX,
+			MLX5_IB_OBJECT_DEVX_OBJ,
+			UVERBS_ACCESS_READ));
+
+DECLARE_UVERBS_NAMED_METHOD_DESTROY(
+	MLX5_IB_METHOD_DESTROY_FLOW,
+	UVERBS_ATTR_IDR(MLX5_IB_ATTR_CREATE_FLOW_HANDLE,
+			UVERBS_OBJECT_FLOW,
+			UVERBS_ACCESS_DESTROY,
+			UA_MANDATORY));
+
+ADD_UVERBS_METHODS(mlx5_ib_fs,
+		   UVERBS_OBJECT_FLOW,
+		   &UVERBS_METHOD(MLX5_IB_METHOD_CREATE_FLOW),
+		   &UVERBS_METHOD(MLX5_IB_METHOD_DESTROY_FLOW));
+
+DECLARE_UVERBS_NAMED_METHOD(
+	MLX5_IB_METHOD_FLOW_MATCHER_CREATE,
+	UVERBS_ATTR_IDR(MLX5_IB_ATTR_FLOW_MATCHER_CREATE_HANDLE,
+			MLX5_IB_OBJECT_FLOW_MATCHER,
+			UVERBS_ACCESS_NEW,
+			UA_MANDATORY),
+	UVERBS_ATTR_PTR_IN(
+		MLX5_IB_ATTR_FLOW_MATCHER_MATCH_MASK,
+		UVERBS_ATTR_SIZE(1, sizeof(struct mlx5_ib_match_params)),
+		UA_MANDATORY),
+	UVERBS_ATTR_ENUM_IN(MLX5_IB_ATTR_FLOW_MATCHER_FLOW_TYPE,
+			    mlx5_ib_flow_type,
+			    UA_MANDATORY),
+	UVERBS_ATTR_PTR_IN(MLX5_IB_ATTR_FLOW_MATCHER_MATCH_CRITERIA,
+			   UVERBS_ATTR_TYPE(u8),
+			   UA_MANDATORY));
+
+DECLARE_UVERBS_NAMED_METHOD_DESTROY(
+	MLX5_IB_METHOD_FLOW_MATCHER_DESTROY,
+	UVERBS_ATTR_IDR(MLX5_IB_ATTR_FLOW_MATCHER_DESTROY_HANDLE,
+			MLX5_IB_OBJECT_FLOW_MATCHER,
+			UVERBS_ACCESS_DESTROY,
+			UA_MANDATORY));
+
+DECLARE_UVERBS_NAMED_OBJECT(MLX5_IB_OBJECT_FLOW_MATCHER,
+			    UVERBS_TYPE_ALLOC_IDR(flow_matcher_cleanup),
+			    &UVERBS_METHOD(MLX5_IB_METHOD_FLOW_MATCHER_CREATE),
+			    &UVERBS_METHOD(MLX5_IB_METHOD_FLOW_MATCHER_DESTROY));
+
+DECLARE_UVERBS_OBJECT_TREE(flow_objects,
+			   &UVERBS_OBJECT(MLX5_IB_OBJECT_FLOW_MATCHER));
+
+int mlx5_ib_get_flow_trees(const struct uverbs_object_tree_def **root)
+{
+	int i = 0;
+
+	root[i++] = &flow_objects;
+	root[i++] = &mlx5_ib_fs;
+
+	return i;
+}
diff --git a/drivers/infiniband/hw/mlx5/gsi.c b/drivers/infiniband/hw/mlx5/gsi.c
new file mode 100644
index 000000000..5c73c0a79
--- /dev/null
+++ b/drivers/infiniband/hw/mlx5/gsi.c
@@ -0,0 +1,540 @@
+/*
+ * Copyright (c) 2016, Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "mlx5_ib.h"
+
+struct mlx5_ib_gsi_wr {
+	struct ib_cqe cqe;
+	struct ib_wc wc;
+	int send_flags;
+	bool completed:1;
+};
+
+struct mlx5_ib_gsi_qp {
+	struct ib_qp ibqp;
+	struct ib_qp *rx_qp;
+	u8 port_num;
+	struct ib_qp_cap cap;
+	enum ib_sig_type sq_sig_type;
+	/* Serialize qp state modifications */
+	struct mutex mutex;
+	struct ib_cq *cq;
+	struct mlx5_ib_gsi_wr *outstanding_wrs;
+	u32 outstanding_pi, outstanding_ci;
+	int num_qps;
+	/* Protects access to the tx_qps. Post send operations synchronize
+	 * with tx_qp creation in setup_qp(). Also protects the
+	 * outstanding_wrs array and indices.
+	 */
+	spinlock_t lock;
+	struct ib_qp **tx_qps;
+};
+
+static struct mlx5_ib_gsi_qp *gsi_qp(struct ib_qp *qp)
+{
+	return container_of(qp, struct mlx5_ib_gsi_qp, ibqp);
+}
+
+static bool mlx5_ib_deth_sqpn_cap(struct mlx5_ib_dev *dev)
+{
+	return MLX5_CAP_GEN(dev->mdev, set_deth_sqpn);
+}
+
+/* Call with gsi->lock locked */
+static void generate_completions(struct mlx5_ib_gsi_qp *gsi)
+{
+	struct ib_cq *gsi_cq = gsi->ibqp.send_cq;
+	struct mlx5_ib_gsi_wr *wr;
+	u32 index;
+
+	for (index = gsi->outstanding_ci; index != gsi->outstanding_pi;
+	     index++) {
+		wr = &gsi->outstanding_wrs[index % gsi->cap.max_send_wr];
+
+		if (!wr->completed)
+			break;
+
+		if (gsi->sq_sig_type == IB_SIGNAL_ALL_WR ||
+		    wr->send_flags & IB_SEND_SIGNALED)
+			WARN_ON_ONCE(mlx5_ib_generate_wc(gsi_cq, &wr->wc));
+
+		wr->completed = false;
+	}
+
+	gsi->outstanding_ci = index;
+}
+
+static void handle_single_completion(struct ib_cq *cq, struct ib_wc *wc)
+{
+	struct mlx5_ib_gsi_qp *gsi = cq->cq_context;
+	struct mlx5_ib_gsi_wr *wr =
+		container_of(wc->wr_cqe, struct mlx5_ib_gsi_wr, cqe);
+	u64 wr_id;
+	unsigned long flags;
+
+	spin_lock_irqsave(&gsi->lock, flags);
+	wr->completed = true;
+	wr_id = wr->wc.wr_id;
+	wr->wc = *wc;
+	wr->wc.wr_id = wr_id;
+	wr->wc.qp = &gsi->ibqp;
+
+	generate_completions(gsi);
+	spin_unlock_irqrestore(&gsi->lock, flags);
+}
+
+struct ib_qp *mlx5_ib_gsi_create_qp(struct ib_pd *pd,
+				    struct ib_qp_init_attr *init_attr)
+{
+	struct mlx5_ib_dev *dev = to_mdev(pd->device);
+	struct mlx5_ib_gsi_qp *gsi;
+	struct ib_qp_init_attr hw_init_attr = *init_attr;
+	const u8 port_num = init_attr->port_num;
+	const int num_pkeys = pd->device->attrs.max_pkeys;
+	const int num_qps = mlx5_ib_deth_sqpn_cap(dev) ? num_pkeys : 0;
+	int ret;
+
+	mlx5_ib_dbg(dev, "creating GSI QP\n");
+
+	if (port_num > ARRAY_SIZE(dev->devr.ports) || port_num < 1) {
+		mlx5_ib_warn(dev,
+			     "invalid port number %d during GSI QP creation\n",
+			     port_num);
+		return ERR_PTR(-EINVAL);
+	}
+
+	gsi = kzalloc(sizeof(*gsi), GFP_KERNEL);
+	if (!gsi)
+		return ERR_PTR(-ENOMEM);
+
+	gsi->tx_qps = kcalloc(num_qps, sizeof(*gsi->tx_qps), GFP_KERNEL);
+	if (!gsi->tx_qps) {
+		ret = -ENOMEM;
+		goto err_free;
+	}
+
+	gsi->outstanding_wrs = kcalloc(init_attr->cap.max_send_wr,
+				       sizeof(*gsi->outstanding_wrs),
+				       GFP_KERNEL);
+	if (!gsi->outstanding_wrs) {
+		ret = -ENOMEM;
+		goto err_free_tx;
+	}
+
+	mutex_init(&gsi->mutex);
+
+	mutex_lock(&dev->devr.mutex);
+
+	if (dev->devr.ports[port_num - 1].gsi) {
+		mlx5_ib_warn(dev, "GSI QP already exists on port %d\n",
+			     port_num);
+		ret = -EBUSY;
+		goto err_free_wrs;
+	}
+	gsi->num_qps = num_qps;
+	spin_lock_init(&gsi->lock);
+
+	gsi->cap = init_attr->cap;
+	gsi->sq_sig_type = init_attr->sq_sig_type;
+	gsi->ibqp.qp_num = 1;
+	gsi->port_num = port_num;
+
+	gsi->cq = ib_alloc_cq(pd->device, gsi, init_attr->cap.max_send_wr, 0,
+			      IB_POLL_SOFTIRQ);
+	if (IS_ERR(gsi->cq)) {
+		mlx5_ib_warn(dev, "unable to create send CQ for GSI QP. error %ld\n",
+			     PTR_ERR(gsi->cq));
+		ret = PTR_ERR(gsi->cq);
+		goto err_free_wrs;
+	}
+
+	hw_init_attr.qp_type = MLX5_IB_QPT_HW_GSI;
+	hw_init_attr.send_cq = gsi->cq;
+	if (num_qps) {
+		hw_init_attr.cap.max_send_wr = 0;
+		hw_init_attr.cap.max_send_sge = 0;
+		hw_init_attr.cap.max_inline_data = 0;
+	}
+	gsi->rx_qp = ib_create_qp(pd, &hw_init_attr);
+	if (IS_ERR(gsi->rx_qp)) {
+		mlx5_ib_warn(dev, "unable to create hardware GSI QP. error %ld\n",
+			     PTR_ERR(gsi->rx_qp));
+		ret = PTR_ERR(gsi->rx_qp);
+		goto err_destroy_cq;
+	}
+
+	dev->devr.ports[init_attr->port_num - 1].gsi = gsi;
+
+	mutex_unlock(&dev->devr.mutex);
+
+	return &gsi->ibqp;
+
+err_destroy_cq:
+	ib_free_cq(gsi->cq);
+err_free_wrs:
+	mutex_unlock(&dev->devr.mutex);
+	kfree(gsi->outstanding_wrs);
+err_free_tx:
+	kfree(gsi->tx_qps);
+err_free:
+	kfree(gsi);
+	return ERR_PTR(ret);
+}
+
+int mlx5_ib_gsi_destroy_qp(struct ib_qp *qp)
+{
+	struct mlx5_ib_dev *dev = to_mdev(qp->device);
+	struct mlx5_ib_gsi_qp *gsi = gsi_qp(qp);
+	const int port_num = gsi->port_num;
+	int qp_index;
+	int ret;
+
+	mlx5_ib_dbg(dev, "destroying GSI QP\n");
+
+	mutex_lock(&dev->devr.mutex);
+	ret = ib_destroy_qp(gsi->rx_qp);
+	if (ret) {
+		mlx5_ib_warn(dev, "unable to destroy hardware GSI QP. error %d\n",
+			     ret);
+		mutex_unlock(&dev->devr.mutex);
+		return ret;
+	}
+	dev->devr.ports[port_num - 1].gsi = NULL;
+	mutex_unlock(&dev->devr.mutex);
+	gsi->rx_qp = NULL;
+
+	for (qp_index = 0; qp_index < gsi->num_qps; ++qp_index) {
+		if (!gsi->tx_qps[qp_index])
+			continue;
+		WARN_ON_ONCE(ib_destroy_qp(gsi->tx_qps[qp_index]));
+		gsi->tx_qps[qp_index] = NULL;
+	}
+
+	ib_free_cq(gsi->cq);
+
+	kfree(gsi->outstanding_wrs);
+	kfree(gsi->tx_qps);
+	kfree(gsi);
+
+	return 0;
+}
+
+static struct ib_qp *create_gsi_ud_qp(struct mlx5_ib_gsi_qp *gsi)
+{
+	struct ib_pd *pd = gsi->rx_qp->pd;
+	struct ib_qp_init_attr init_attr = {
+		.event_handler = gsi->rx_qp->event_handler,
+		.qp_context = gsi->rx_qp->qp_context,
+		.send_cq = gsi->cq,
+		.recv_cq = gsi->rx_qp->recv_cq,
+		.cap = {
+			.max_send_wr = gsi->cap.max_send_wr,
+			.max_send_sge = gsi->cap.max_send_sge,
+			.max_inline_data = gsi->cap.max_inline_data,
+		},
+		.sq_sig_type = gsi->sq_sig_type,
+		.qp_type = IB_QPT_UD,
+		.create_flags = mlx5_ib_create_qp_sqpn_qp1(),
+	};
+
+	return ib_create_qp(pd, &init_attr);
+}
+
+static int modify_to_rts(struct mlx5_ib_gsi_qp *gsi, struct ib_qp *qp,
+			 u16 qp_index)
+{
+	struct mlx5_ib_dev *dev = to_mdev(qp->device);
+	struct ib_qp_attr attr;
+	int mask;
+	int ret;
+
+	mask = IB_QP_STATE | IB_QP_PKEY_INDEX | IB_QP_QKEY | IB_QP_PORT;
+	attr.qp_state = IB_QPS_INIT;
+	attr.pkey_index = qp_index;
+	attr.qkey = IB_QP1_QKEY;
+	attr.port_num = gsi->port_num;
+	ret = ib_modify_qp(qp, &attr, mask);
+	if (ret) {
+		mlx5_ib_err(dev, "could not change QP%d state to INIT: %d\n",
+			    qp->qp_num, ret);
+		return ret;
+	}
+
+	attr.qp_state = IB_QPS_RTR;
+	ret = ib_modify_qp(qp, &attr, IB_QP_STATE);
+	if (ret) {
+		mlx5_ib_err(dev, "could not change QP%d state to RTR: %d\n",
+			    qp->qp_num, ret);
+		return ret;
+	}
+
+	attr.qp_state = IB_QPS_RTS;
+	attr.sq_psn = 0;
+	ret = ib_modify_qp(qp, &attr, IB_QP_STATE | IB_QP_SQ_PSN);
+	if (ret) {
+		mlx5_ib_err(dev, "could not change QP%d state to RTS: %d\n",
+			    qp->qp_num, ret);
+		return ret;
+	}
+
+	return 0;
+}
+
+static void setup_qp(struct mlx5_ib_gsi_qp *gsi, u16 qp_index)
+{
+	struct ib_device *device = gsi->rx_qp->device;
+	struct mlx5_ib_dev *dev = to_mdev(device);
+	struct ib_qp *qp;
+	unsigned long flags;
+	u16 pkey;
+	int ret;
+
+	ret = ib_query_pkey(device, gsi->port_num, qp_index, &pkey);
+	if (ret) {
+		mlx5_ib_warn(dev, "unable to read P_Key at port %d, index %d\n",
+			     gsi->port_num, qp_index);
+		return;
+	}
+
+	if (!pkey) {
+		mlx5_ib_dbg(dev, "invalid P_Key at port %d, index %d.  Skipping.\n",
+			    gsi->port_num, qp_index);
+		return;
+	}
+
+	spin_lock_irqsave(&gsi->lock, flags);
+	qp = gsi->tx_qps[qp_index];
+	spin_unlock_irqrestore(&gsi->lock, flags);
+	if (qp) {
+		mlx5_ib_dbg(dev, "already existing GSI TX QP at port %d, index %d. Skipping\n",
+			    gsi->port_num, qp_index);
+		return;
+	}
+
+	qp = create_gsi_ud_qp(gsi);
+	if (IS_ERR(qp)) {
+		mlx5_ib_warn(dev, "unable to create hardware UD QP for GSI: %ld\n",
+			     PTR_ERR(qp));
+		return;
+	}
+
+	ret = modify_to_rts(gsi, qp, qp_index);
+	if (ret)
+		goto err_destroy_qp;
+
+	spin_lock_irqsave(&gsi->lock, flags);
+	WARN_ON_ONCE(gsi->tx_qps[qp_index]);
+	gsi->tx_qps[qp_index] = qp;
+	spin_unlock_irqrestore(&gsi->lock, flags);
+
+	return;
+
+err_destroy_qp:
+	WARN_ON_ONCE(qp);
+}
+
+static void setup_qps(struct mlx5_ib_gsi_qp *gsi)
+{
+	u16 qp_index;
+
+	for (qp_index = 0; qp_index < gsi->num_qps; ++qp_index)
+		setup_qp(gsi, qp_index);
+}
+
+int mlx5_ib_gsi_modify_qp(struct ib_qp *qp, struct ib_qp_attr *attr,
+			  int attr_mask)
+{
+	struct mlx5_ib_dev *dev = to_mdev(qp->device);
+	struct mlx5_ib_gsi_qp *gsi = gsi_qp(qp);
+	int ret;
+
+	mlx5_ib_dbg(dev, "modifying GSI QP to state %d\n", attr->qp_state);
+
+	mutex_lock(&gsi->mutex);
+	ret = ib_modify_qp(gsi->rx_qp, attr, attr_mask);
+	if (ret) {
+		mlx5_ib_warn(dev, "unable to modify GSI rx QP: %d\n", ret);
+		goto unlock;
+	}
+
+	if (to_mqp(gsi->rx_qp)->state == IB_QPS_RTS)
+		setup_qps(gsi);
+
+unlock:
+	mutex_unlock(&gsi->mutex);
+
+	return ret;
+}
+
+int mlx5_ib_gsi_query_qp(struct ib_qp *qp, struct ib_qp_attr *qp_attr,
+			 int qp_attr_mask,
+			 struct ib_qp_init_attr *qp_init_attr)
+{
+	struct mlx5_ib_gsi_qp *gsi = gsi_qp(qp);
+	int ret;
+
+	mutex_lock(&gsi->mutex);
+	ret = ib_query_qp(gsi->rx_qp, qp_attr, qp_attr_mask, qp_init_attr);
+	qp_init_attr->cap = gsi->cap;
+	mutex_unlock(&gsi->mutex);
+
+	return ret;
+}
+
+/* Call with gsi->lock locked */
+static int mlx5_ib_add_outstanding_wr(struct mlx5_ib_gsi_qp *gsi,
+				      struct ib_ud_wr *wr, struct ib_wc *wc)
+{
+	struct mlx5_ib_dev *dev = to_mdev(gsi->rx_qp->device);
+	struct mlx5_ib_gsi_wr *gsi_wr;
+
+	if (gsi->outstanding_pi == gsi->outstanding_ci + gsi->cap.max_send_wr) {
+		mlx5_ib_warn(dev, "no available GSI work request.\n");
+		return -ENOMEM;
+	}
+
+	gsi_wr = &gsi->outstanding_wrs[gsi->outstanding_pi %
+				       gsi->cap.max_send_wr];
+	gsi->outstanding_pi++;
+
+	if (!wc) {
+		memset(&gsi_wr->wc, 0, sizeof(gsi_wr->wc));
+		gsi_wr->wc.pkey_index = wr->pkey_index;
+		gsi_wr->wc.wr_id = wr->wr.wr_id;
+	} else {
+		gsi_wr->wc = *wc;
+		gsi_wr->completed = true;
+	}
+
+	gsi_wr->cqe.done = &handle_single_completion;
+	wr->wr.wr_cqe = &gsi_wr->cqe;
+
+	return 0;
+}
+
+/* Call with gsi->lock locked */
+static int mlx5_ib_gsi_silent_drop(struct mlx5_ib_gsi_qp *gsi,
+				    struct ib_ud_wr *wr)
+{
+	struct ib_wc wc = {
+		{ .wr_id = wr->wr.wr_id },
+		.status = IB_WC_SUCCESS,
+		.opcode = IB_WC_SEND,
+		.qp = &gsi->ibqp,
+	};
+	int ret;
+
+	ret = mlx5_ib_add_outstanding_wr(gsi, wr, &wc);
+	if (ret)
+		return ret;
+
+	generate_completions(gsi);
+
+	return 0;
+}
+
+/* Call with gsi->lock locked */
+static struct ib_qp *get_tx_qp(struct mlx5_ib_gsi_qp *gsi, struct ib_ud_wr *wr)
+{
+	struct mlx5_ib_dev *dev = to_mdev(gsi->rx_qp->device);
+	int qp_index = wr->pkey_index;
+
+	if (!mlx5_ib_deth_sqpn_cap(dev))
+		return gsi->rx_qp;
+
+	if (qp_index >= gsi->num_qps)
+		return NULL;
+
+	return gsi->tx_qps[qp_index];
+}
+
+int mlx5_ib_gsi_post_send(struct ib_qp *qp, const struct ib_send_wr *wr,
+			  const struct ib_send_wr **bad_wr)
+{
+	struct mlx5_ib_gsi_qp *gsi = gsi_qp(qp);
+	struct ib_qp *tx_qp;
+	unsigned long flags;
+	int ret;
+
+	for (; wr; wr = wr->next) {
+		struct ib_ud_wr cur_wr = *ud_wr(wr);
+
+		cur_wr.wr.next = NULL;
+
+		spin_lock_irqsave(&gsi->lock, flags);
+		tx_qp = get_tx_qp(gsi, &cur_wr);
+		if (!tx_qp) {
+			ret = mlx5_ib_gsi_silent_drop(gsi, &cur_wr);
+			if (ret)
+				goto err;
+			spin_unlock_irqrestore(&gsi->lock, flags);
+			continue;
+		}
+
+		ret = mlx5_ib_add_outstanding_wr(gsi, &cur_wr, NULL);
+		if (ret)
+			goto err;
+
+		ret = ib_post_send(tx_qp, &cur_wr.wr, bad_wr);
+		if (ret) {
+			/* Undo the effect of adding the outstanding wr */
+			gsi->outstanding_pi--;
+			goto err;
+		}
+		spin_unlock_irqrestore(&gsi->lock, flags);
+	}
+
+	return 0;
+
+err:
+	spin_unlock_irqrestore(&gsi->lock, flags);
+	*bad_wr = wr;
+	return ret;
+}
+
+int mlx5_ib_gsi_post_recv(struct ib_qp *qp, const struct ib_recv_wr *wr,
+			  const struct ib_recv_wr **bad_wr)
+{
+	struct mlx5_ib_gsi_qp *gsi = gsi_qp(qp);
+
+	return ib_post_recv(gsi->rx_qp, wr, bad_wr);
+}
+
+void mlx5_ib_gsi_pkey_change(struct mlx5_ib_gsi_qp *gsi)
+{
+	if (!gsi)
+		return;
+
+	mutex_lock(&gsi->mutex);
+	setup_qps(gsi);
+	mutex_unlock(&gsi->mutex);
+}
diff --git a/drivers/infiniband/hw/mlx5/ib_rep.c b/drivers/infiniband/hw/mlx5/ib_rep.c
new file mode 100644
index 000000000..b841589c2
--- /dev/null
+++ b/drivers/infiniband/hw/mlx5/ib_rep.c
@@ -0,0 +1,194 @@
+// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
+/*
+ * Copyright (c) 2018 Mellanox Technologies. All rights reserved.
+ */
+
+#include "ib_rep.h"
+
+static const struct mlx5_ib_profile rep_profile = {
+	STAGE_CREATE(MLX5_IB_STAGE_INIT,
+		     mlx5_ib_stage_init_init,
+		     mlx5_ib_stage_init_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_FLOW_DB,
+		     mlx5_ib_stage_rep_flow_db_init,
+		     NULL),
+	STAGE_CREATE(MLX5_IB_STAGE_CAPS,
+		     mlx5_ib_stage_caps_init,
+		     NULL),
+	STAGE_CREATE(MLX5_IB_STAGE_NON_DEFAULT_CB,
+		     mlx5_ib_stage_rep_non_default_cb,
+		     NULL),
+	STAGE_CREATE(MLX5_IB_STAGE_ROCE,
+		     mlx5_ib_stage_rep_roce_init,
+		     mlx5_ib_stage_rep_roce_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_DEVICE_RESOURCES,
+		     mlx5_ib_stage_dev_res_init,
+		     mlx5_ib_stage_dev_res_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_COUNTERS,
+		     mlx5_ib_stage_counters_init,
+		     mlx5_ib_stage_counters_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_BFREG,
+		     mlx5_ib_stage_bfrag_init,
+		     mlx5_ib_stage_bfrag_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_PRE_IB_REG_UMR,
+		     NULL,
+		     mlx5_ib_stage_pre_ib_reg_umr_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_IB_REG,
+		     mlx5_ib_stage_ib_reg_init,
+		     mlx5_ib_stage_ib_reg_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_POST_IB_REG_UMR,
+		     mlx5_ib_stage_post_ib_reg_umr_init,
+		     NULL),
+	STAGE_CREATE(MLX5_IB_STAGE_CLASS_ATTR,
+		     mlx5_ib_stage_class_attr_init,
+		     NULL),
+};
+
+static int
+mlx5_ib_nic_rep_load(struct mlx5_core_dev *dev, struct mlx5_eswitch_rep *rep)
+{
+	return 0;
+}
+
+static void
+mlx5_ib_nic_rep_unload(struct mlx5_eswitch_rep *rep)
+{
+	rep->rep_if[REP_IB].priv = NULL;
+}
+
+static int
+mlx5_ib_vport_rep_load(struct mlx5_core_dev *dev, struct mlx5_eswitch_rep *rep)
+{
+	struct mlx5_ib_dev *ibdev;
+
+	ibdev = (struct mlx5_ib_dev *)ib_alloc_device(sizeof(*ibdev));
+	if (!ibdev)
+		return -ENOMEM;
+
+	ibdev->rep = rep;
+	ibdev->mdev = dev;
+	ibdev->num_ports = max(MLX5_CAP_GEN(dev, num_ports),
+			       MLX5_CAP_GEN(dev, num_vhca_ports));
+	if (!__mlx5_ib_add(ibdev, &rep_profile)) {
+		ib_dealloc_device(&ibdev->ib_dev);
+		return -EINVAL;
+	}
+
+	rep->rep_if[REP_IB].priv = ibdev;
+
+	return 0;
+}
+
+static void
+mlx5_ib_vport_rep_unload(struct mlx5_eswitch_rep *rep)
+{
+	struct mlx5_ib_dev *dev;
+
+	if (!rep->rep_if[REP_IB].priv)
+		return;
+
+	dev = mlx5_ib_rep_to_dev(rep);
+	__mlx5_ib_remove(dev, dev->profile, MLX5_IB_STAGE_MAX);
+	rep->rep_if[REP_IB].priv = NULL;
+}
+
+static void *mlx5_ib_vport_get_proto_dev(struct mlx5_eswitch_rep *rep)
+{
+	return mlx5_ib_rep_to_dev(rep);
+}
+
+static void mlx5_ib_rep_register_vf_vports(struct mlx5_ib_dev *dev)
+{
+	struct mlx5_eswitch *esw   = dev->mdev->priv.eswitch;
+	int total_vfs = MLX5_TOTAL_VPORTS(dev->mdev);
+	int vport;
+
+	for (vport = 1; vport < total_vfs; vport++) {
+		struct mlx5_eswitch_rep_if rep_if = {};
+
+		rep_if.load = mlx5_ib_vport_rep_load;
+		rep_if.unload = mlx5_ib_vport_rep_unload;
+		rep_if.get_proto_dev = mlx5_ib_vport_get_proto_dev;
+		mlx5_eswitch_register_vport_rep(esw, vport, &rep_if, REP_IB);
+	}
+}
+
+static void mlx5_ib_rep_unregister_vf_vports(struct mlx5_ib_dev *dev)
+{
+	struct mlx5_eswitch *esw   = dev->mdev->priv.eswitch;
+	int total_vfs = MLX5_TOTAL_VPORTS(dev->mdev);
+	int vport;
+
+	for (vport = 1; vport < total_vfs; vport++)
+		mlx5_eswitch_unregister_vport_rep(esw, vport, REP_IB);
+}
+
+void mlx5_ib_register_vport_reps(struct mlx5_ib_dev *dev)
+{
+	struct mlx5_eswitch *esw = dev->mdev->priv.eswitch;
+	struct mlx5_eswitch_rep_if rep_if = {};
+
+	rep_if.load = mlx5_ib_nic_rep_load;
+	rep_if.unload = mlx5_ib_nic_rep_unload;
+	rep_if.get_proto_dev = mlx5_ib_vport_get_proto_dev;
+	rep_if.priv = dev;
+
+	mlx5_eswitch_register_vport_rep(esw, 0, &rep_if, REP_IB);
+
+	mlx5_ib_rep_register_vf_vports(dev);
+}
+
+void mlx5_ib_unregister_vport_reps(struct mlx5_ib_dev *dev)
+{
+	struct mlx5_eswitch *esw   = dev->mdev->priv.eswitch;
+
+	mlx5_ib_rep_unregister_vf_vports(dev); /* VFs vports */
+	mlx5_eswitch_unregister_vport_rep(esw, 0, REP_IB); /* UPLINK PF*/
+}
+
+u8 mlx5_ib_eswitch_mode(struct mlx5_eswitch *esw)
+{
+	return mlx5_eswitch_mode(esw);
+}
+
+struct mlx5_ib_dev *mlx5_ib_get_rep_ibdev(struct mlx5_eswitch *esw,
+					  int vport_index)
+{
+	return mlx5_eswitch_get_proto_dev(esw, vport_index, REP_IB);
+}
+
+struct net_device *mlx5_ib_get_rep_netdev(struct mlx5_eswitch *esw,
+					  int vport_index)
+{
+	return mlx5_eswitch_get_proto_dev(esw, vport_index, REP_ETH);
+}
+
+struct mlx5_ib_dev *mlx5_ib_get_uplink_ibdev(struct mlx5_eswitch *esw)
+{
+	return mlx5_eswitch_uplink_get_proto_dev(esw, REP_IB);
+}
+
+struct mlx5_eswitch_rep *mlx5_ib_vport_rep(struct mlx5_eswitch *esw, int vport)
+{
+	return mlx5_eswitch_vport_rep(esw, vport);
+}
+
+int create_flow_rule_vport_sq(struct mlx5_ib_dev *dev,
+			      struct mlx5_ib_sq *sq)
+{
+	struct mlx5_flow_handle *flow_rule;
+	struct mlx5_eswitch *esw = dev->mdev->priv.eswitch;
+
+	if (!dev->rep)
+		return 0;
+
+	flow_rule =
+		mlx5_eswitch_add_send_to_vport_rule(esw,
+						    dev->rep->vport,
+						    sq->base.mqp.qpn);
+	if (IS_ERR(flow_rule))
+		return PTR_ERR(flow_rule);
+	sq->flow_rule = flow_rule;
+
+	return 0;
+}
diff --git a/drivers/infiniband/hw/mlx5/ib_rep.h b/drivers/infiniband/hw/mlx5/ib_rep.h
new file mode 100644
index 000000000..2ba73636a
--- /dev/null
+++ b/drivers/infiniband/hw/mlx5/ib_rep.h
@@ -0,0 +1,72 @@
+/* SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB */
+/*
+ * Copyright (c) 2018 Mellanox Technologies. All rights reserved.
+ */
+
+#ifndef __MLX5_IB_REP_H__
+#define __MLX5_IB_REP_H__
+
+#include <linux/mlx5/eswitch.h>
+#include "mlx5_ib.h"
+
+#ifdef CONFIG_MLX5_ESWITCH
+u8 mlx5_ib_eswitch_mode(struct mlx5_eswitch *esw);
+struct mlx5_ib_dev *mlx5_ib_get_rep_ibdev(struct mlx5_eswitch *esw,
+					  int vport_index);
+struct mlx5_ib_dev *mlx5_ib_get_uplink_ibdev(struct mlx5_eswitch *esw);
+struct mlx5_eswitch_rep *mlx5_ib_vport_rep(struct mlx5_eswitch *esw,
+					   int vport_index);
+void mlx5_ib_register_vport_reps(struct mlx5_ib_dev *dev);
+void mlx5_ib_unregister_vport_reps(struct mlx5_ib_dev *dev);
+int create_flow_rule_vport_sq(struct mlx5_ib_dev *dev,
+			      struct mlx5_ib_sq *sq);
+struct net_device *mlx5_ib_get_rep_netdev(struct mlx5_eswitch *esw,
+					  int vport_index);
+#else /* CONFIG_MLX5_ESWITCH */
+static inline u8 mlx5_ib_eswitch_mode(struct mlx5_eswitch *esw)
+{
+	return SRIOV_NONE;
+}
+
+static inline
+struct mlx5_ib_dev *mlx5_ib_get_rep_ibdev(struct mlx5_eswitch *esw,
+					  int vport_index)
+{
+	return NULL;
+}
+
+static inline
+struct mlx5_ib_dev *mlx5_ib_get_uplink_ibdev(struct mlx5_eswitch *esw)
+{
+	return NULL;
+}
+
+static inline
+struct mlx5_eswitch_rep *mlx5_ib_vport_rep(struct mlx5_eswitch *esw,
+					   int vport_index)
+{
+	return NULL;
+}
+
+static inline void mlx5_ib_register_vport_reps(struct mlx5_ib_dev *dev) {}
+static inline void mlx5_ib_unregister_vport_reps(struct mlx5_ib_dev *dev) {}
+static inline int create_flow_rule_vport_sq(struct mlx5_ib_dev *dev,
+					    struct mlx5_ib_sq *sq)
+{
+	return 0;
+}
+
+static inline
+struct net_device *mlx5_ib_get_rep_netdev(struct mlx5_eswitch *esw,
+					  int vport_index)
+{
+	return NULL;
+}
+#endif
+
+static inline
+struct mlx5_ib_dev *mlx5_ib_rep_to_dev(struct mlx5_eswitch_rep *rep)
+{
+	return (struct mlx5_ib_dev *)rep->rep_if[REP_IB].priv;
+}
+#endif /* __MLX5_IB_REP_H__ */
diff --git a/drivers/infiniband/hw/mlx5/ib_virt.c b/drivers/infiniband/hw/mlx5/ib_virt.c
new file mode 100644
index 000000000..649a3364f
--- /dev/null
+++ b/drivers/infiniband/hw/mlx5/ib_virt.c
@@ -0,0 +1,203 @@
+/*
+ * Copyright (c) 2016, Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/module.h>
+#include <linux/mlx5/vport.h>
+#include "mlx5_ib.h"
+
+static inline u32 mlx_to_net_policy(enum port_state_policy mlx_policy)
+{
+	switch (mlx_policy) {
+	case MLX5_POLICY_DOWN:
+		return IFLA_VF_LINK_STATE_DISABLE;
+	case MLX5_POLICY_UP:
+		return IFLA_VF_LINK_STATE_ENABLE;
+	case MLX5_POLICY_FOLLOW:
+		return IFLA_VF_LINK_STATE_AUTO;
+	default:
+		return __IFLA_VF_LINK_STATE_MAX;
+	}
+}
+
+int mlx5_ib_get_vf_config(struct ib_device *device, int vf, u8 port,
+			  struct ifla_vf_info *info)
+{
+	struct mlx5_ib_dev *dev = to_mdev(device);
+	struct mlx5_core_dev *mdev = dev->mdev;
+	struct mlx5_hca_vport_context *rep;
+	int err;
+
+	rep = kzalloc(sizeof(*rep), GFP_KERNEL);
+	if (!rep)
+		return -ENOMEM;
+
+	err = mlx5_query_hca_vport_context(mdev, 1, 1,  vf + 1, rep);
+	if (err) {
+		mlx5_ib_warn(dev, "failed to query port policy for vf %d (%d)\n",
+			     vf, err);
+		goto free;
+	}
+	memset(info, 0, sizeof(*info));
+	info->linkstate = mlx_to_net_policy(rep->policy);
+	if (info->linkstate == __IFLA_VF_LINK_STATE_MAX)
+		err = -EINVAL;
+
+free:
+	kfree(rep);
+	return err;
+}
+
+static inline enum port_state_policy net_to_mlx_policy(int policy)
+{
+	switch (policy) {
+	case IFLA_VF_LINK_STATE_DISABLE:
+		return MLX5_POLICY_DOWN;
+	case IFLA_VF_LINK_STATE_ENABLE:
+		return MLX5_POLICY_UP;
+	case IFLA_VF_LINK_STATE_AUTO:
+		return MLX5_POLICY_FOLLOW;
+	default:
+		return MLX5_POLICY_INVALID;
+	}
+}
+
+int mlx5_ib_set_vf_link_state(struct ib_device *device, int vf,
+			      u8 port, int state)
+{
+	struct mlx5_ib_dev *dev = to_mdev(device);
+	struct mlx5_core_dev *mdev = dev->mdev;
+	struct mlx5_hca_vport_context *in;
+	struct mlx5_vf_context *vfs_ctx = mdev->priv.sriov.vfs_ctx;
+	int err;
+
+	in = kzalloc(sizeof(*in), GFP_KERNEL);
+	if (!in)
+		return -ENOMEM;
+
+	in->policy = net_to_mlx_policy(state);
+	if (in->policy == MLX5_POLICY_INVALID) {
+		err = -EINVAL;
+		goto out;
+	}
+	in->field_select = MLX5_HCA_VPORT_SEL_STATE_POLICY;
+	err = mlx5_core_modify_hca_vport_context(mdev, 1, 1, vf + 1, in);
+	if (!err)
+		vfs_ctx[vf].policy = in->policy;
+
+out:
+	kfree(in);
+	return err;
+}
+
+int mlx5_ib_get_vf_stats(struct ib_device *device, int vf,
+			 u8 port, struct ifla_vf_stats *stats)
+{
+	int out_sz = MLX5_ST_SZ_BYTES(query_vport_counter_out);
+	struct mlx5_core_dev *mdev;
+	struct mlx5_ib_dev *dev;
+	void *out;
+	int err;
+
+	dev = to_mdev(device);
+	mdev = dev->mdev;
+
+	out = kzalloc(out_sz, GFP_KERNEL);
+	if (!out)
+		return -ENOMEM;
+
+	err = mlx5_core_query_vport_counter(mdev, true, vf, port, out, out_sz);
+	if (err)
+		goto ex;
+
+	stats->rx_packets = MLX5_GET64_PR(query_vport_counter_out, out, received_ib_unicast.packets);
+	stats->tx_packets = MLX5_GET64_PR(query_vport_counter_out, out, transmitted_ib_unicast.packets);
+	stats->rx_bytes = MLX5_GET64_PR(query_vport_counter_out, out, received_ib_unicast.octets);
+	stats->tx_bytes = MLX5_GET64_PR(query_vport_counter_out, out, transmitted_ib_unicast.octets);
+	stats->multicast = MLX5_GET64_PR(query_vport_counter_out, out, received_ib_multicast.packets);
+
+ex:
+	kfree(out);
+	return err;
+}
+
+static int set_vf_node_guid(struct ib_device *device, int vf, u8 port, u64 guid)
+{
+	struct mlx5_ib_dev *dev = to_mdev(device);
+	struct mlx5_core_dev *mdev = dev->mdev;
+	struct mlx5_hca_vport_context *in;
+	struct mlx5_vf_context *vfs_ctx = mdev->priv.sriov.vfs_ctx;
+	int err;
+
+	in = kzalloc(sizeof(*in), GFP_KERNEL);
+	if (!in)
+		return -ENOMEM;
+
+	in->field_select = MLX5_HCA_VPORT_SEL_NODE_GUID;
+	in->node_guid = guid;
+	err = mlx5_core_modify_hca_vport_context(mdev, 1, 1, vf + 1, in);
+	if (!err)
+		vfs_ctx[vf].node_guid = guid;
+	kfree(in);
+	return err;
+}
+
+static int set_vf_port_guid(struct ib_device *device, int vf, u8 port, u64 guid)
+{
+	struct mlx5_ib_dev *dev = to_mdev(device);
+	struct mlx5_core_dev *mdev = dev->mdev;
+	struct mlx5_hca_vport_context *in;
+	struct mlx5_vf_context *vfs_ctx = mdev->priv.sriov.vfs_ctx;
+	int err;
+
+	in = kzalloc(sizeof(*in), GFP_KERNEL);
+	if (!in)
+		return -ENOMEM;
+
+	in->field_select = MLX5_HCA_VPORT_SEL_PORT_GUID;
+	in->port_guid = guid;
+	err = mlx5_core_modify_hca_vport_context(mdev, 1, 1, vf + 1, in);
+	if (!err)
+		vfs_ctx[vf].port_guid = guid;
+	kfree(in);
+	return err;
+}
+
+int mlx5_ib_set_vf_guid(struct ib_device *device, int vf, u8 port,
+			u64 guid, int type)
+{
+	if (type == IFLA_VF_IB_NODE_GUID)
+		return set_vf_node_guid(device, vf, port, guid);
+	else if (type == IFLA_VF_IB_PORT_GUID)
+		return set_vf_port_guid(device, vf, port, guid);
+
+	return -EINVAL;
+}
diff --git a/drivers/infiniband/hw/mlx5/mad.c b/drivers/infiniband/hw/mlx5/mad.c
new file mode 100644
index 000000000..cdf6e26eb
--- /dev/null
+++ b/drivers/infiniband/hw/mlx5/mad.c
@@ -0,0 +1,621 @@
+/*
+ * Copyright (c) 2013-2015, Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/mlx5/cmd.h>
+#include <linux/mlx5/vport.h>
+#include <rdma/ib_mad.h>
+#include <rdma/ib_smi.h>
+#include <rdma/ib_pma.h>
+#include "mlx5_ib.h"
+
+enum {
+	MLX5_IB_VENDOR_CLASS1 = 0x9,
+	MLX5_IB_VENDOR_CLASS2 = 0xa
+};
+
+static bool can_do_mad_ifc(struct mlx5_ib_dev *dev, u8 port_num,
+			   struct ib_mad *in_mad)
+{
+	if (in_mad->mad_hdr.mgmt_class != IB_MGMT_CLASS_SUBN_LID_ROUTED &&
+	    in_mad->mad_hdr.mgmt_class != IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE)
+		return true;
+	return dev->mdev->port_caps[port_num - 1].has_smi;
+}
+
+int mlx5_MAD_IFC(struct mlx5_ib_dev *dev, int ignore_mkey, int ignore_bkey,
+		 u8 port, const struct ib_wc *in_wc, const struct ib_grh *in_grh,
+		 const void *in_mad, void *response_mad)
+{
+	u8 op_modifier = 0;
+
+	if (!can_do_mad_ifc(dev, port, (struct ib_mad *)in_mad))
+		return -EPERM;
+
+	/* Key check traps can't be generated unless we have in_wc to
+	 * tell us where to send the trap.
+	 */
+	if (ignore_mkey || !in_wc)
+		op_modifier |= 0x1;
+	if (ignore_bkey || !in_wc)
+		op_modifier |= 0x2;
+
+	return mlx5_core_mad_ifc(dev->mdev, in_mad, response_mad, op_modifier, port);
+}
+
+static int process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num,
+		       const struct ib_wc *in_wc, const struct ib_grh *in_grh,
+		       const struct ib_mad *in_mad, struct ib_mad *out_mad)
+{
+	u16 slid;
+	int err;
+
+	slid = in_wc ? ib_lid_cpu16(in_wc->slid) : be16_to_cpu(IB_LID_PERMISSIVE);
+
+	if (in_mad->mad_hdr.method == IB_MGMT_METHOD_TRAP && slid == 0)
+		return IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_CONSUMED;
+
+	if (in_mad->mad_hdr.mgmt_class == IB_MGMT_CLASS_SUBN_LID_ROUTED ||
+	    in_mad->mad_hdr.mgmt_class == IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE) {
+		if (in_mad->mad_hdr.method   != IB_MGMT_METHOD_GET &&
+		    in_mad->mad_hdr.method   != IB_MGMT_METHOD_SET &&
+		    in_mad->mad_hdr.method   != IB_MGMT_METHOD_TRAP_REPRESS)
+			return IB_MAD_RESULT_SUCCESS;
+
+		/* Don't process SMInfo queries -- the SMA can't handle them.
+		 */
+		if (in_mad->mad_hdr.attr_id == IB_SMP_ATTR_SM_INFO)
+			return IB_MAD_RESULT_SUCCESS;
+	} else if (in_mad->mad_hdr.mgmt_class == IB_MGMT_CLASS_PERF_MGMT ||
+		   in_mad->mad_hdr.mgmt_class == MLX5_IB_VENDOR_CLASS1   ||
+		   in_mad->mad_hdr.mgmt_class == MLX5_IB_VENDOR_CLASS2   ||
+		   in_mad->mad_hdr.mgmt_class == IB_MGMT_CLASS_CONG_MGMT) {
+		if (in_mad->mad_hdr.method  != IB_MGMT_METHOD_GET &&
+		    in_mad->mad_hdr.method  != IB_MGMT_METHOD_SET)
+			return IB_MAD_RESULT_SUCCESS;
+	} else {
+		return IB_MAD_RESULT_SUCCESS;
+	}
+
+	err = mlx5_MAD_IFC(to_mdev(ibdev),
+			   mad_flags & IB_MAD_IGNORE_MKEY,
+			   mad_flags & IB_MAD_IGNORE_BKEY,
+			   port_num, in_wc, in_grh, in_mad, out_mad);
+	if (err)
+		return IB_MAD_RESULT_FAILURE;
+
+	/* set return bit in status of directed route responses */
+	if (in_mad->mad_hdr.mgmt_class == IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE)
+		out_mad->mad_hdr.status |= cpu_to_be16(1 << 15);
+
+	if (in_mad->mad_hdr.method == IB_MGMT_METHOD_TRAP_REPRESS)
+		/* no response for trap repress */
+		return IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_CONSUMED;
+
+	return IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_REPLY;
+}
+
+static void pma_cnt_ext_assign(struct ib_pma_portcounters_ext *pma_cnt_ext,
+			       void *out)
+{
+#define MLX5_SUM_CNT(p, cntr1, cntr2)	\
+	(MLX5_GET64(query_vport_counter_out, p, cntr1) + \
+	MLX5_GET64(query_vport_counter_out, p, cntr2))
+
+	pma_cnt_ext->port_xmit_data =
+		cpu_to_be64(MLX5_SUM_CNT(out, transmitted_ib_unicast.octets,
+					 transmitted_ib_multicast.octets) >> 2);
+	pma_cnt_ext->port_rcv_data =
+		cpu_to_be64(MLX5_SUM_CNT(out, received_ib_unicast.octets,
+					 received_ib_multicast.octets) >> 2);
+	pma_cnt_ext->port_xmit_packets =
+		cpu_to_be64(MLX5_SUM_CNT(out, transmitted_ib_unicast.packets,
+					 transmitted_ib_multicast.packets));
+	pma_cnt_ext->port_rcv_packets =
+		cpu_to_be64(MLX5_SUM_CNT(out, received_ib_unicast.packets,
+					 received_ib_multicast.packets));
+	pma_cnt_ext->port_unicast_xmit_packets =
+		MLX5_GET64_BE(query_vport_counter_out,
+			      out, transmitted_ib_unicast.packets);
+	pma_cnt_ext->port_unicast_rcv_packets =
+		MLX5_GET64_BE(query_vport_counter_out,
+			      out, received_ib_unicast.packets);
+	pma_cnt_ext->port_multicast_xmit_packets =
+		MLX5_GET64_BE(query_vport_counter_out,
+			      out, transmitted_ib_multicast.packets);
+	pma_cnt_ext->port_multicast_rcv_packets =
+		MLX5_GET64_BE(query_vport_counter_out,
+			      out, received_ib_multicast.packets);
+}
+
+static void pma_cnt_assign(struct ib_pma_portcounters *pma_cnt,
+			   void *out)
+{
+	/* Traffic counters will be reported in
+	 * their 64bit form via ib_pma_portcounters_ext by default.
+	 */
+	void *out_pma = MLX5_ADDR_OF(ppcnt_reg, out,
+				     counter_set);
+
+#define MLX5_ASSIGN_PMA_CNTR(counter_var, counter_name)	{		\
+	counter_var = MLX5_GET_BE(typeof(counter_var),			\
+				  ib_port_cntrs_grp_data_layout,	\
+				  out_pma, counter_name);		\
+	}
+
+	MLX5_ASSIGN_PMA_CNTR(pma_cnt->symbol_error_counter,
+			     symbol_error_counter);
+	MLX5_ASSIGN_PMA_CNTR(pma_cnt->link_error_recovery_counter,
+			     link_error_recovery_counter);
+	MLX5_ASSIGN_PMA_CNTR(pma_cnt->link_downed_counter,
+			     link_downed_counter);
+	MLX5_ASSIGN_PMA_CNTR(pma_cnt->port_rcv_errors,
+			     port_rcv_errors);
+	MLX5_ASSIGN_PMA_CNTR(pma_cnt->port_rcv_remphys_errors,
+			     port_rcv_remote_physical_errors);
+	MLX5_ASSIGN_PMA_CNTR(pma_cnt->port_rcv_switch_relay_errors,
+			     port_rcv_switch_relay_errors);
+	MLX5_ASSIGN_PMA_CNTR(pma_cnt->port_xmit_discards,
+			     port_xmit_discards);
+	MLX5_ASSIGN_PMA_CNTR(pma_cnt->port_xmit_constraint_errors,
+			     port_xmit_constraint_errors);
+	MLX5_ASSIGN_PMA_CNTR(pma_cnt->port_xmit_wait,
+			     port_xmit_wait);
+	MLX5_ASSIGN_PMA_CNTR(pma_cnt->port_rcv_constraint_errors,
+			     port_rcv_constraint_errors);
+	MLX5_ASSIGN_PMA_CNTR(pma_cnt->link_overrun_errors,
+			     link_overrun_errors);
+	MLX5_ASSIGN_PMA_CNTR(pma_cnt->vl15_dropped,
+			     vl_15_dropped);
+}
+
+static int process_pma_cmd(struct mlx5_ib_dev *dev, u8 port_num,
+			   const struct ib_mad *in_mad, struct ib_mad *out_mad)
+{
+	struct mlx5_core_dev *mdev;
+	bool native_port = true;
+	u8 mdev_port_num;
+	void *out_cnt;
+	int err;
+
+	mdev = mlx5_ib_get_native_port_mdev(dev, port_num, &mdev_port_num);
+	if (!mdev) {
+		/* Fail to get the native port, likely due to 2nd port is still
+		 * unaffiliated. In such case default to 1st port and attached
+		 * PF device.
+		 */
+		native_port = false;
+		mdev = dev->mdev;
+		mdev_port_num = 1;
+	}
+	/* Declaring support of extended counters */
+	if (in_mad->mad_hdr.attr_id == IB_PMA_CLASS_PORT_INFO) {
+		struct ib_class_port_info cpi = {};
+
+		cpi.capability_mask = IB_PMA_CLASS_CAP_EXT_WIDTH;
+		memcpy((out_mad->data + 40), &cpi, sizeof(cpi));
+		err = IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_REPLY;
+		goto done;
+	}
+
+	if (in_mad->mad_hdr.attr_id == IB_PMA_PORT_COUNTERS_EXT) {
+		struct ib_pma_portcounters_ext *pma_cnt_ext =
+			(struct ib_pma_portcounters_ext *)(out_mad->data + 40);
+		int sz = MLX5_ST_SZ_BYTES(query_vport_counter_out);
+
+		out_cnt = kvzalloc(sz, GFP_KERNEL);
+		if (!out_cnt) {
+			err = IB_MAD_RESULT_FAILURE;
+			goto done;
+		}
+
+		err = mlx5_core_query_vport_counter(mdev, 0, 0,
+						    mdev_port_num, out_cnt, sz);
+		if (!err)
+			pma_cnt_ext_assign(pma_cnt_ext, out_cnt);
+	} else {
+		struct ib_pma_portcounters *pma_cnt =
+			(struct ib_pma_portcounters *)(out_mad->data + 40);
+		int sz = MLX5_ST_SZ_BYTES(ppcnt_reg);
+
+		out_cnt = kvzalloc(sz, GFP_KERNEL);
+		if (!out_cnt) {
+			err = IB_MAD_RESULT_FAILURE;
+			goto done;
+		}
+
+		err = mlx5_core_query_ib_ppcnt(mdev, mdev_port_num,
+					       out_cnt, sz);
+		if (!err)
+			pma_cnt_assign(pma_cnt, out_cnt);
+	}
+	kvfree(out_cnt);
+	err = err ? IB_MAD_RESULT_FAILURE :
+		    IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_REPLY;
+done:
+	if (native_port)
+		mlx5_ib_put_native_port_mdev(dev, port_num);
+	return err;
+}
+
+int mlx5_ib_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num,
+			const struct ib_wc *in_wc, const struct ib_grh *in_grh,
+			const struct ib_mad_hdr *in, size_t in_mad_size,
+			struct ib_mad_hdr *out, size_t *out_mad_size,
+			u16 *out_mad_pkey_index)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+	const struct ib_mad *in_mad = (const struct ib_mad *)in;
+	struct ib_mad *out_mad = (struct ib_mad *)out;
+	int ret;
+
+	if (WARN_ON_ONCE(in_mad_size != sizeof(*in_mad) ||
+			 *out_mad_size != sizeof(*out_mad)))
+		return IB_MAD_RESULT_FAILURE;
+
+	memset(out_mad->data, 0, sizeof(out_mad->data));
+
+	if (MLX5_CAP_GEN(dev->mdev, vport_counters) &&
+	    in_mad->mad_hdr.mgmt_class == IB_MGMT_CLASS_PERF_MGMT &&
+	    in_mad->mad_hdr.method == IB_MGMT_METHOD_GET) {
+		ret = process_pma_cmd(dev, port_num, in_mad, out_mad);
+	} else {
+		ret =  process_mad(ibdev, mad_flags, port_num, in_wc, in_grh,
+				   in_mad, out_mad);
+	}
+	return ret;
+}
+
+int mlx5_query_ext_port_caps(struct mlx5_ib_dev *dev, u8 port)
+{
+	struct ib_smp *in_mad  = NULL;
+	struct ib_smp *out_mad = NULL;
+	int err = -ENOMEM;
+	u16 packet_error;
+
+	in_mad  = kzalloc(sizeof(*in_mad), GFP_KERNEL);
+	out_mad = kmalloc(sizeof(*out_mad), GFP_KERNEL);
+	if (!in_mad || !out_mad)
+		goto out;
+
+	init_query_mad(in_mad);
+	in_mad->attr_id = MLX5_ATTR_EXTENDED_PORT_INFO;
+	in_mad->attr_mod = cpu_to_be32(port);
+
+	err = mlx5_MAD_IFC(dev, 1, 1, 1, NULL, NULL, in_mad, out_mad);
+
+	packet_error = be16_to_cpu(out_mad->status);
+
+	dev->mdev->port_caps[port - 1].ext_port_cap = (!err && !packet_error) ?
+		MLX_EXT_PORT_CAP_FLAG_EXTENDED_PORT_INFO : 0;
+
+out:
+	kfree(in_mad);
+	kfree(out_mad);
+	return err;
+}
+
+int mlx5_query_mad_ifc_smp_attr_node_info(struct ib_device *ibdev,
+					  struct ib_smp *out_mad)
+{
+	struct ib_smp *in_mad = NULL;
+	int err = -ENOMEM;
+
+	in_mad = kzalloc(sizeof(*in_mad), GFP_KERNEL);
+	if (!in_mad)
+		return -ENOMEM;
+
+	init_query_mad(in_mad);
+	in_mad->attr_id = IB_SMP_ATTR_NODE_INFO;
+
+	err = mlx5_MAD_IFC(to_mdev(ibdev), 1, 1, 1, NULL, NULL, in_mad,
+			   out_mad);
+
+	kfree(in_mad);
+	return err;
+}
+
+int mlx5_query_mad_ifc_system_image_guid(struct ib_device *ibdev,
+					 __be64 *sys_image_guid)
+{
+	struct ib_smp *out_mad = NULL;
+	int err = -ENOMEM;
+
+	out_mad = kmalloc(sizeof(*out_mad), GFP_KERNEL);
+	if (!out_mad)
+		return -ENOMEM;
+
+	err = mlx5_query_mad_ifc_smp_attr_node_info(ibdev, out_mad);
+	if (err)
+		goto out;
+
+	memcpy(sys_image_guid, out_mad->data + 4, 8);
+
+out:
+	kfree(out_mad);
+
+	return err;
+}
+
+int mlx5_query_mad_ifc_max_pkeys(struct ib_device *ibdev,
+				 u16 *max_pkeys)
+{
+	struct ib_smp *out_mad = NULL;
+	int err = -ENOMEM;
+
+	out_mad = kmalloc(sizeof(*out_mad), GFP_KERNEL);
+	if (!out_mad)
+		return -ENOMEM;
+
+	err = mlx5_query_mad_ifc_smp_attr_node_info(ibdev, out_mad);
+	if (err)
+		goto out;
+
+	*max_pkeys = be16_to_cpup((__be16 *)(out_mad->data + 28));
+
+out:
+	kfree(out_mad);
+
+	return err;
+}
+
+int mlx5_query_mad_ifc_vendor_id(struct ib_device *ibdev,
+				 u32 *vendor_id)
+{
+	struct ib_smp *out_mad = NULL;
+	int err = -ENOMEM;
+
+	out_mad = kmalloc(sizeof(*out_mad), GFP_KERNEL);
+	if (!out_mad)
+		return -ENOMEM;
+
+	err = mlx5_query_mad_ifc_smp_attr_node_info(ibdev, out_mad);
+	if (err)
+		goto out;
+
+	*vendor_id = be32_to_cpup((__be32 *)(out_mad->data + 36)) & 0xffff;
+
+out:
+	kfree(out_mad);
+
+	return err;
+}
+
+int mlx5_query_mad_ifc_node_desc(struct mlx5_ib_dev *dev, char *node_desc)
+{
+	struct ib_smp *in_mad  = NULL;
+	struct ib_smp *out_mad = NULL;
+	int err = -ENOMEM;
+
+	in_mad  = kzalloc(sizeof(*in_mad), GFP_KERNEL);
+	out_mad = kmalloc(sizeof(*out_mad), GFP_KERNEL);
+	if (!in_mad || !out_mad)
+		goto out;
+
+	init_query_mad(in_mad);
+	in_mad->attr_id = IB_SMP_ATTR_NODE_DESC;
+
+	err = mlx5_MAD_IFC(dev, 1, 1, 1, NULL, NULL, in_mad, out_mad);
+	if (err)
+		goto out;
+
+	memcpy(node_desc, out_mad->data, IB_DEVICE_NODE_DESC_MAX);
+out:
+	kfree(in_mad);
+	kfree(out_mad);
+	return err;
+}
+
+int mlx5_query_mad_ifc_node_guid(struct mlx5_ib_dev *dev, __be64 *node_guid)
+{
+	struct ib_smp *in_mad  = NULL;
+	struct ib_smp *out_mad = NULL;
+	int err = -ENOMEM;
+
+	in_mad  = kzalloc(sizeof(*in_mad), GFP_KERNEL);
+	out_mad = kmalloc(sizeof(*out_mad), GFP_KERNEL);
+	if (!in_mad || !out_mad)
+		goto out;
+
+	init_query_mad(in_mad);
+	in_mad->attr_id = IB_SMP_ATTR_NODE_INFO;
+
+	err = mlx5_MAD_IFC(dev, 1, 1, 1, NULL, NULL, in_mad, out_mad);
+	if (err)
+		goto out;
+
+	memcpy(node_guid, out_mad->data + 12, 8);
+out:
+	kfree(in_mad);
+	kfree(out_mad);
+	return err;
+}
+
+int mlx5_query_mad_ifc_pkey(struct ib_device *ibdev, u8 port, u16 index,
+			    u16 *pkey)
+{
+	struct ib_smp *in_mad  = NULL;
+	struct ib_smp *out_mad = NULL;
+	int err = -ENOMEM;
+
+	in_mad  = kzalloc(sizeof(*in_mad), GFP_KERNEL);
+	out_mad = kmalloc(sizeof(*out_mad), GFP_KERNEL);
+	if (!in_mad || !out_mad)
+		goto out;
+
+	init_query_mad(in_mad);
+	in_mad->attr_id  = IB_SMP_ATTR_PKEY_TABLE;
+	in_mad->attr_mod = cpu_to_be32(index / 32);
+
+	err = mlx5_MAD_IFC(to_mdev(ibdev), 1, 1, port, NULL, NULL, in_mad,
+			   out_mad);
+	if (err)
+		goto out;
+
+	*pkey = be16_to_cpu(((__be16 *)out_mad->data)[index % 32]);
+
+out:
+	kfree(in_mad);
+	kfree(out_mad);
+	return err;
+}
+
+int mlx5_query_mad_ifc_gids(struct ib_device *ibdev, u8 port, int index,
+			    union ib_gid *gid)
+{
+	struct ib_smp *in_mad  = NULL;
+	struct ib_smp *out_mad = NULL;
+	int err = -ENOMEM;
+
+	in_mad  = kzalloc(sizeof(*in_mad), GFP_KERNEL);
+	out_mad = kmalloc(sizeof(*out_mad), GFP_KERNEL);
+	if (!in_mad || !out_mad)
+		goto out;
+
+	init_query_mad(in_mad);
+	in_mad->attr_id  = IB_SMP_ATTR_PORT_INFO;
+	in_mad->attr_mod = cpu_to_be32(port);
+
+	err = mlx5_MAD_IFC(to_mdev(ibdev), 1, 1, port, NULL, NULL, in_mad,
+			   out_mad);
+	if (err)
+		goto out;
+
+	memcpy(gid->raw, out_mad->data + 8, 8);
+
+	init_query_mad(in_mad);
+	in_mad->attr_id  = IB_SMP_ATTR_GUID_INFO;
+	in_mad->attr_mod = cpu_to_be32(index / 8);
+
+	err = mlx5_MAD_IFC(to_mdev(ibdev), 1, 1, port, NULL, NULL, in_mad,
+			   out_mad);
+	if (err)
+		goto out;
+
+	memcpy(gid->raw + 8, out_mad->data + (index % 8) * 8, 8);
+
+out:
+	kfree(in_mad);
+	kfree(out_mad);
+	return err;
+}
+
+int mlx5_query_mad_ifc_port(struct ib_device *ibdev, u8 port,
+			    struct ib_port_attr *props)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+	struct mlx5_core_dev *mdev = dev->mdev;
+	struct ib_smp *in_mad  = NULL;
+	struct ib_smp *out_mad = NULL;
+	int ext_active_speed;
+	int err = -ENOMEM;
+
+	if (port < 1 || port > dev->num_ports) {
+		mlx5_ib_warn(dev, "invalid port number %d\n", port);
+		return -EINVAL;
+	}
+
+	in_mad  = kzalloc(sizeof(*in_mad), GFP_KERNEL);
+	out_mad = kmalloc(sizeof(*out_mad), GFP_KERNEL);
+	if (!in_mad || !out_mad)
+		goto out;
+
+	/* props being zeroed by the caller, avoid zeroing it here */
+
+	init_query_mad(in_mad);
+	in_mad->attr_id  = IB_SMP_ATTR_PORT_INFO;
+	in_mad->attr_mod = cpu_to_be32(port);
+
+	err = mlx5_MAD_IFC(dev, 1, 1, port, NULL, NULL, in_mad, out_mad);
+	if (err) {
+		mlx5_ib_warn(dev, "err %d\n", err);
+		goto out;
+	}
+
+	props->lid		= be16_to_cpup((__be16 *)(out_mad->data + 16));
+	props->lmc		= out_mad->data[34] & 0x7;
+	props->sm_lid		= be16_to_cpup((__be16 *)(out_mad->data + 18));
+	props->sm_sl		= out_mad->data[36] & 0xf;
+	props->state		= out_mad->data[32] & 0xf;
+	props->phys_state	= out_mad->data[33] >> 4;
+	props->port_cap_flags	= be32_to_cpup((__be32 *)(out_mad->data + 20));
+	props->gid_tbl_len	= out_mad->data[50];
+	props->max_msg_sz	= 1 << MLX5_CAP_GEN(mdev, log_max_msg);
+	props->pkey_tbl_len	= mdev->port_caps[port - 1].pkey_table_len;
+	props->bad_pkey_cntr	= be16_to_cpup((__be16 *)(out_mad->data + 46));
+	props->qkey_viol_cntr	= be16_to_cpup((__be16 *)(out_mad->data + 48));
+	props->active_width	= out_mad->data[31] & 0xf;
+	props->active_speed	= out_mad->data[35] >> 4;
+	props->max_mtu		= out_mad->data[41] & 0xf;
+	props->active_mtu	= out_mad->data[36] >> 4;
+	props->subnet_timeout	= out_mad->data[51] & 0x1f;
+	props->max_vl_num	= out_mad->data[37] >> 4;
+	props->init_type_reply	= out_mad->data[41] >> 4;
+
+	/* Check if extended speeds (EDR/FDR/...) are supported */
+	if (props->port_cap_flags & IB_PORT_EXTENDED_SPEEDS_SUP) {
+		ext_active_speed = out_mad->data[62] >> 4;
+
+		switch (ext_active_speed) {
+		case 1:
+			props->active_speed = 16; /* FDR */
+			break;
+		case 2:
+			props->active_speed = 32; /* EDR */
+			break;
+		}
+	}
+
+	/* If reported active speed is QDR, check if is FDR-10 */
+	if (props->active_speed == 4) {
+		if (mdev->port_caps[port - 1].ext_port_cap &
+		    MLX_EXT_PORT_CAP_FLAG_EXTENDED_PORT_INFO) {
+			init_query_mad(in_mad);
+			in_mad->attr_id = MLX5_ATTR_EXTENDED_PORT_INFO;
+			in_mad->attr_mod = cpu_to_be32(port);
+
+			err = mlx5_MAD_IFC(dev, 1, 1, port,
+					   NULL, NULL, in_mad, out_mad);
+			if (err)
+				goto out;
+
+			/* Checking LinkSpeedActive for FDR-10 */
+			if (out_mad->data[15] & 0x1)
+				props->active_speed = 8;
+		}
+	}
+
+out:
+	kfree(in_mad);
+	kfree(out_mad);
+
+	return err;
+}
diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
new file mode 100644
index 000000000..1688c06d5
--- /dev/null
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -0,0 +1,6461 @@
+/*
+ * Copyright (c) 2013-2015, Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/debugfs.h>
+#include <linux/highmem.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/errno.h>
+#include <linux/pci.h>
+#include <linux/dma-mapping.h>
+#include <linux/slab.h>
+#include <linux/bitmap.h>
+#if defined(CONFIG_X86)
+#include <asm/pat.h>
+#endif
+#include <linux/sched.h>
+#include <linux/sched/mm.h>
+#include <linux/sched/task.h>
+#include <linux/delay.h>
+#include <rdma/ib_user_verbs.h>
+#include <rdma/ib_addr.h>
+#include <rdma/ib_cache.h>
+#include <linux/mlx5/port.h>
+#include <linux/mlx5/vport.h>
+#include <linux/mlx5/fs.h>
+#include <linux/list.h>
+#include <rdma/ib_smi.h>
+#include <rdma/ib_umem.h>
+#include <linux/in.h>
+#include <linux/etherdevice.h>
+#include "mlx5_ib.h"
+#include "ib_rep.h"
+#include "cmd.h"
+#include <linux/mlx5/fs_helpers.h>
+#include <linux/mlx5/accel.h>
+#include <rdma/uverbs_std_types.h>
+#include <rdma/mlx5_user_ioctl_verbs.h>
+#include <rdma/mlx5_user_ioctl_cmds.h>
+
+#define UVERBS_MODULE_NAME mlx5_ib
+#include <rdma/uverbs_named_ioctl.h>
+
+#define DRIVER_NAME "mlx5_ib"
+#define DRIVER_VERSION "5.0-0"
+
+MODULE_AUTHOR("Eli Cohen <eli@mellanox.com>");
+MODULE_DESCRIPTION("Mellanox Connect-IB HCA IB driver");
+MODULE_LICENSE("Dual BSD/GPL");
+
+static char mlx5_version[] =
+	DRIVER_NAME ": Mellanox Connect-IB Infiniband driver v"
+	DRIVER_VERSION "\n";
+
+struct mlx5_ib_event_work {
+	struct work_struct	work;
+	struct mlx5_core_dev	*dev;
+	void			*context;
+	enum mlx5_dev_event	event;
+	unsigned long		param;
+};
+
+enum {
+	MLX5_ATOMIC_SIZE_QP_8BYTES = 1 << 3,
+};
+
+static struct workqueue_struct *mlx5_ib_event_wq;
+static LIST_HEAD(mlx5_ib_unaffiliated_port_list);
+static LIST_HEAD(mlx5_ib_dev_list);
+/*
+ * This mutex should be held when accessing either of the above lists
+ */
+static DEFINE_MUTEX(mlx5_ib_multiport_mutex);
+
+/* We can't use an array for xlt_emergency_page because dma_map_single
+ * doesn't work on kernel modules memory
+ */
+static unsigned long xlt_emergency_page;
+static struct mutex xlt_emergency_page_mutex;
+
+struct mlx5_ib_dev *mlx5_ib_get_ibdev_from_mpi(struct mlx5_ib_multiport_info *mpi)
+{
+	struct mlx5_ib_dev *dev;
+
+	mutex_lock(&mlx5_ib_multiport_mutex);
+	dev = mpi->ibdev;
+	mutex_unlock(&mlx5_ib_multiport_mutex);
+	return dev;
+}
+
+static enum rdma_link_layer
+mlx5_port_type_cap_to_rdma_ll(int port_type_cap)
+{
+	switch (port_type_cap) {
+	case MLX5_CAP_PORT_TYPE_IB:
+		return IB_LINK_LAYER_INFINIBAND;
+	case MLX5_CAP_PORT_TYPE_ETH:
+		return IB_LINK_LAYER_ETHERNET;
+	default:
+		return IB_LINK_LAYER_UNSPECIFIED;
+	}
+}
+
+static enum rdma_link_layer
+mlx5_ib_port_link_layer(struct ib_device *device, u8 port_num)
+{
+	struct mlx5_ib_dev *dev = to_mdev(device);
+	int port_type_cap = MLX5_CAP_GEN(dev->mdev, port_type);
+
+	return mlx5_port_type_cap_to_rdma_ll(port_type_cap);
+}
+
+static int get_port_state(struct ib_device *ibdev,
+			  u8 port_num,
+			  enum ib_port_state *state)
+{
+	struct ib_port_attr attr;
+	int ret;
+
+	memset(&attr, 0, sizeof(attr));
+	ret = ibdev->query_port(ibdev, port_num, &attr);
+	if (!ret)
+		*state = attr.state;
+	return ret;
+}
+
+static int mlx5_netdev_event(struct notifier_block *this,
+			     unsigned long event, void *ptr)
+{
+	struct mlx5_roce *roce = container_of(this, struct mlx5_roce, nb);
+	struct net_device *ndev = netdev_notifier_info_to_dev(ptr);
+	u8 port_num = roce->native_port_num;
+	struct mlx5_core_dev *mdev;
+	struct mlx5_ib_dev *ibdev;
+
+	ibdev = roce->dev;
+	mdev = mlx5_ib_get_native_port_mdev(ibdev, port_num, NULL);
+	if (!mdev)
+		return NOTIFY_DONE;
+
+	switch (event) {
+	case NETDEV_REGISTER:
+	case NETDEV_UNREGISTER:
+		write_lock(&roce->netdev_lock);
+		if (ibdev->rep) {
+			struct mlx5_eswitch *esw = ibdev->mdev->priv.eswitch;
+			struct net_device *rep_ndev;
+
+			rep_ndev = mlx5_ib_get_rep_netdev(esw,
+							  ibdev->rep->vport);
+			if (rep_ndev == ndev)
+				roce->netdev = (event == NETDEV_UNREGISTER) ?
+					NULL : ndev;
+		} else if (ndev->dev.parent == &mdev->pdev->dev) {
+			roce->netdev = (event == NETDEV_UNREGISTER) ?
+				NULL : ndev;
+		}
+		write_unlock(&roce->netdev_lock);
+		break;
+
+	case NETDEV_CHANGE:
+	case NETDEV_UP:
+	case NETDEV_DOWN: {
+		struct net_device *lag_ndev = mlx5_lag_get_roce_netdev(mdev);
+		struct net_device *upper = NULL;
+
+		if (lag_ndev) {
+			upper = netdev_master_upper_dev_get(lag_ndev);
+			dev_put(lag_ndev);
+		}
+
+		if ((upper == ndev || (!upper && ndev == roce->netdev))
+		    && ibdev->ib_active) {
+			struct ib_event ibev = { };
+			enum ib_port_state port_state;
+
+			if (get_port_state(&ibdev->ib_dev, port_num,
+					   &port_state))
+				goto done;
+
+			if (roce->last_port_state == port_state)
+				goto done;
+
+			roce->last_port_state = port_state;
+			ibev.device = &ibdev->ib_dev;
+			if (port_state == IB_PORT_DOWN)
+				ibev.event = IB_EVENT_PORT_ERR;
+			else if (port_state == IB_PORT_ACTIVE)
+				ibev.event = IB_EVENT_PORT_ACTIVE;
+			else
+				goto done;
+
+			ibev.element.port_num = port_num;
+			ib_dispatch_event(&ibev);
+		}
+		break;
+	}
+
+	default:
+		break;
+	}
+done:
+	mlx5_ib_put_native_port_mdev(ibdev, port_num);
+	return NOTIFY_DONE;
+}
+
+static struct net_device *mlx5_ib_get_netdev(struct ib_device *device,
+					     u8 port_num)
+{
+	struct mlx5_ib_dev *ibdev = to_mdev(device);
+	struct net_device *ndev;
+	struct mlx5_core_dev *mdev;
+
+	mdev = mlx5_ib_get_native_port_mdev(ibdev, port_num, NULL);
+	if (!mdev)
+		return NULL;
+
+	ndev = mlx5_lag_get_roce_netdev(mdev);
+	if (ndev)
+		goto out;
+
+	/* Ensure ndev does not disappear before we invoke dev_hold()
+	 */
+	read_lock(&ibdev->roce[port_num - 1].netdev_lock);
+	ndev = ibdev->roce[port_num - 1].netdev;
+	if (ndev)
+		dev_hold(ndev);
+	read_unlock(&ibdev->roce[port_num - 1].netdev_lock);
+
+out:
+	mlx5_ib_put_native_port_mdev(ibdev, port_num);
+	return ndev;
+}
+
+struct mlx5_core_dev *mlx5_ib_get_native_port_mdev(struct mlx5_ib_dev *ibdev,
+						   u8 ib_port_num,
+						   u8 *native_port_num)
+{
+	enum rdma_link_layer ll = mlx5_ib_port_link_layer(&ibdev->ib_dev,
+							  ib_port_num);
+	struct mlx5_core_dev *mdev = NULL;
+	struct mlx5_ib_multiport_info *mpi;
+	struct mlx5_ib_port *port;
+
+	if (!mlx5_core_mp_enabled(ibdev->mdev) ||
+	    ll != IB_LINK_LAYER_ETHERNET) {
+		if (native_port_num)
+			*native_port_num = ib_port_num;
+		return ibdev->mdev;
+	}
+
+	if (native_port_num)
+		*native_port_num = 1;
+
+	port = &ibdev->port[ib_port_num - 1];
+	if (!port)
+		return NULL;
+
+	spin_lock(&port->mp.mpi_lock);
+	mpi = ibdev->port[ib_port_num - 1].mp.mpi;
+	if (mpi && !mpi->unaffiliate) {
+		mdev = mpi->mdev;
+		/* If it's the master no need to refcount, it'll exist
+		 * as long as the ib_dev exists.
+		 */
+		if (!mpi->is_master)
+			mpi->mdev_refcnt++;
+	}
+	spin_unlock(&port->mp.mpi_lock);
+
+	return mdev;
+}
+
+void mlx5_ib_put_native_port_mdev(struct mlx5_ib_dev *ibdev, u8 port_num)
+{
+	enum rdma_link_layer ll = mlx5_ib_port_link_layer(&ibdev->ib_dev,
+							  port_num);
+	struct mlx5_ib_multiport_info *mpi;
+	struct mlx5_ib_port *port;
+
+	if (!mlx5_core_mp_enabled(ibdev->mdev) || ll != IB_LINK_LAYER_ETHERNET)
+		return;
+
+	port = &ibdev->port[port_num - 1];
+
+	spin_lock(&port->mp.mpi_lock);
+	mpi = ibdev->port[port_num - 1].mp.mpi;
+	if (mpi->is_master)
+		goto out;
+
+	mpi->mdev_refcnt--;
+	if (mpi->unaffiliate)
+		complete(&mpi->unref_comp);
+out:
+	spin_unlock(&port->mp.mpi_lock);
+}
+
+static int translate_eth_proto_oper(u32 eth_proto_oper, u8 *active_speed,
+				    u8 *active_width)
+{
+	switch (eth_proto_oper) {
+	case MLX5E_PROT_MASK(MLX5E_1000BASE_CX_SGMII):
+	case MLX5E_PROT_MASK(MLX5E_1000BASE_KX):
+	case MLX5E_PROT_MASK(MLX5E_100BASE_TX):
+	case MLX5E_PROT_MASK(MLX5E_1000BASE_T):
+		*active_width = IB_WIDTH_1X;
+		*active_speed = IB_SPEED_SDR;
+		break;
+	case MLX5E_PROT_MASK(MLX5E_10GBASE_T):
+	case MLX5E_PROT_MASK(MLX5E_10GBASE_CX4):
+	case MLX5E_PROT_MASK(MLX5E_10GBASE_KX4):
+	case MLX5E_PROT_MASK(MLX5E_10GBASE_KR):
+	case MLX5E_PROT_MASK(MLX5E_10GBASE_CR):
+	case MLX5E_PROT_MASK(MLX5E_10GBASE_SR):
+	case MLX5E_PROT_MASK(MLX5E_10GBASE_ER):
+		*active_width = IB_WIDTH_1X;
+		*active_speed = IB_SPEED_QDR;
+		break;
+	case MLX5E_PROT_MASK(MLX5E_25GBASE_CR):
+	case MLX5E_PROT_MASK(MLX5E_25GBASE_KR):
+	case MLX5E_PROT_MASK(MLX5E_25GBASE_SR):
+		*active_width = IB_WIDTH_1X;
+		*active_speed = IB_SPEED_EDR;
+		break;
+	case MLX5E_PROT_MASK(MLX5E_40GBASE_CR4):
+	case MLX5E_PROT_MASK(MLX5E_40GBASE_KR4):
+	case MLX5E_PROT_MASK(MLX5E_40GBASE_SR4):
+	case MLX5E_PROT_MASK(MLX5E_40GBASE_LR4):
+		*active_width = IB_WIDTH_4X;
+		*active_speed = IB_SPEED_QDR;
+		break;
+	case MLX5E_PROT_MASK(MLX5E_50GBASE_CR2):
+	case MLX5E_PROT_MASK(MLX5E_50GBASE_KR2):
+	case MLX5E_PROT_MASK(MLX5E_50GBASE_SR2):
+		*active_width = IB_WIDTH_1X;
+		*active_speed = IB_SPEED_HDR;
+		break;
+	case MLX5E_PROT_MASK(MLX5E_56GBASE_R4):
+		*active_width = IB_WIDTH_4X;
+		*active_speed = IB_SPEED_FDR;
+		break;
+	case MLX5E_PROT_MASK(MLX5E_100GBASE_CR4):
+	case MLX5E_PROT_MASK(MLX5E_100GBASE_SR4):
+	case MLX5E_PROT_MASK(MLX5E_100GBASE_KR4):
+	case MLX5E_PROT_MASK(MLX5E_100GBASE_LR4):
+		*active_width = IB_WIDTH_4X;
+		*active_speed = IB_SPEED_EDR;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int mlx5_query_port_roce(struct ib_device *device, u8 port_num,
+				struct ib_port_attr *props)
+{
+	struct mlx5_ib_dev *dev = to_mdev(device);
+	struct mlx5_core_dev *mdev;
+	struct net_device *ndev, *upper;
+	enum ib_mtu ndev_ib_mtu;
+	bool put_mdev = true;
+	u16 qkey_viol_cntr;
+	u32 eth_prot_oper;
+	u8 mdev_port_num;
+	int err;
+
+	mdev = mlx5_ib_get_native_port_mdev(dev, port_num, &mdev_port_num);
+	if (!mdev) {
+		/* This means the port isn't affiliated yet. Get the
+		 * info for the master port instead.
+		 */
+		put_mdev = false;
+		mdev = dev->mdev;
+		mdev_port_num = 1;
+		port_num = 1;
+	}
+
+	/* Possible bad flows are checked before filling out props so in case
+	 * of an error it will still be zeroed out.
+	 */
+	err = mlx5_query_port_eth_proto_oper(mdev, &eth_prot_oper,
+					     mdev_port_num);
+	if (err)
+		goto out;
+
+	props->active_width     = IB_WIDTH_4X;
+	props->active_speed     = IB_SPEED_QDR;
+
+	translate_eth_proto_oper(eth_prot_oper, &props->active_speed,
+				 &props->active_width);
+
+	props->port_cap_flags |= IB_PORT_CM_SUP;
+	props->ip_gids = true;
+
+	props->gid_tbl_len      = MLX5_CAP_ROCE(dev->mdev,
+						roce_address_table_size);
+	props->max_mtu          = IB_MTU_4096;
+	props->max_msg_sz       = 1 << MLX5_CAP_GEN(dev->mdev, log_max_msg);
+	props->pkey_tbl_len     = 1;
+	props->state            = IB_PORT_DOWN;
+	props->phys_state       = 3;
+
+	mlx5_query_nic_vport_qkey_viol_cntr(mdev, &qkey_viol_cntr);
+	props->qkey_viol_cntr = qkey_viol_cntr;
+
+	/* If this is a stub query for an unaffiliated port stop here */
+	if (!put_mdev)
+		goto out;
+
+	ndev = mlx5_ib_get_netdev(device, port_num);
+	if (!ndev)
+		goto out;
+
+	if (mlx5_lag_is_active(dev->mdev)) {
+		rcu_read_lock();
+		upper = netdev_master_upper_dev_get_rcu(ndev);
+		if (upper) {
+			dev_put(ndev);
+			ndev = upper;
+			dev_hold(ndev);
+		}
+		rcu_read_unlock();
+	}
+
+	if (netif_running(ndev) && netif_carrier_ok(ndev)) {
+		props->state      = IB_PORT_ACTIVE;
+		props->phys_state = 5;
+	}
+
+	ndev_ib_mtu = iboe_get_mtu(ndev->mtu);
+
+	dev_put(ndev);
+
+	props->active_mtu	= min(props->max_mtu, ndev_ib_mtu);
+out:
+	if (put_mdev)
+		mlx5_ib_put_native_port_mdev(dev, port_num);
+	return err;
+}
+
+static int set_roce_addr(struct mlx5_ib_dev *dev, u8 port_num,
+			 unsigned int index, const union ib_gid *gid,
+			 const struct ib_gid_attr *attr)
+{
+	enum ib_gid_type gid_type = IB_GID_TYPE_IB;
+	u8 roce_version = 0;
+	u8 roce_l3_type = 0;
+	bool vlan = false;
+	u8 mac[ETH_ALEN];
+	u16 vlan_id = 0;
+
+	if (gid) {
+		gid_type = attr->gid_type;
+		ether_addr_copy(mac, attr->ndev->dev_addr);
+
+		if (is_vlan_dev(attr->ndev)) {
+			vlan = true;
+			vlan_id = vlan_dev_vlan_id(attr->ndev);
+		}
+	}
+
+	switch (gid_type) {
+	case IB_GID_TYPE_IB:
+		roce_version = MLX5_ROCE_VERSION_1;
+		break;
+	case IB_GID_TYPE_ROCE_UDP_ENCAP:
+		roce_version = MLX5_ROCE_VERSION_2;
+		if (ipv6_addr_v4mapped((void *)gid))
+			roce_l3_type = MLX5_ROCE_L3_TYPE_IPV4;
+		else
+			roce_l3_type = MLX5_ROCE_L3_TYPE_IPV6;
+		break;
+
+	default:
+		mlx5_ib_warn(dev, "Unexpected GID type %u\n", gid_type);
+	}
+
+	return mlx5_core_roce_gid_set(dev->mdev, index, roce_version,
+				      roce_l3_type, gid->raw, mac, vlan,
+				      vlan_id, port_num);
+}
+
+static int mlx5_ib_add_gid(const struct ib_gid_attr *attr,
+			   __always_unused void **context)
+{
+	return set_roce_addr(to_mdev(attr->device), attr->port_num,
+			     attr->index, &attr->gid, attr);
+}
+
+static int mlx5_ib_del_gid(const struct ib_gid_attr *attr,
+			   __always_unused void **context)
+{
+	return set_roce_addr(to_mdev(attr->device), attr->port_num,
+			     attr->index, NULL, NULL);
+}
+
+__be16 mlx5_get_roce_udp_sport(struct mlx5_ib_dev *dev,
+			       const struct ib_gid_attr *attr)
+{
+	if (attr->gid_type != IB_GID_TYPE_ROCE_UDP_ENCAP)
+		return 0;
+
+	return cpu_to_be16(MLX5_CAP_ROCE(dev->mdev, r_roce_min_src_udp_port));
+}
+
+static int mlx5_use_mad_ifc(struct mlx5_ib_dev *dev)
+{
+	if (MLX5_CAP_GEN(dev->mdev, port_type) == MLX5_CAP_PORT_TYPE_IB)
+		return !MLX5_CAP_GEN(dev->mdev, ib_virt);
+	return 0;
+}
+
+enum {
+	MLX5_VPORT_ACCESS_METHOD_MAD,
+	MLX5_VPORT_ACCESS_METHOD_HCA,
+	MLX5_VPORT_ACCESS_METHOD_NIC,
+};
+
+static int mlx5_get_vport_access_method(struct ib_device *ibdev)
+{
+	if (mlx5_use_mad_ifc(to_mdev(ibdev)))
+		return MLX5_VPORT_ACCESS_METHOD_MAD;
+
+	if (mlx5_ib_port_link_layer(ibdev, 1) ==
+	    IB_LINK_LAYER_ETHERNET)
+		return MLX5_VPORT_ACCESS_METHOD_NIC;
+
+	return MLX5_VPORT_ACCESS_METHOD_HCA;
+}
+
+static void get_atomic_caps(struct mlx5_ib_dev *dev,
+			    u8 atomic_size_qp,
+			    struct ib_device_attr *props)
+{
+	u8 tmp;
+	u8 atomic_operations = MLX5_CAP_ATOMIC(dev->mdev, atomic_operations);
+	u8 atomic_req_8B_endianness_mode =
+		MLX5_CAP_ATOMIC(dev->mdev, atomic_req_8B_endianness_mode);
+
+	/* Check if HW supports 8 bytes standard atomic operations and capable
+	 * of host endianness respond
+	 */
+	tmp = MLX5_ATOMIC_OPS_CMP_SWAP | MLX5_ATOMIC_OPS_FETCH_ADD;
+	if (((atomic_operations & tmp) == tmp) &&
+	    (atomic_size_qp & MLX5_ATOMIC_SIZE_QP_8BYTES) &&
+	    (atomic_req_8B_endianness_mode)) {
+		props->atomic_cap = IB_ATOMIC_HCA;
+	} else {
+		props->atomic_cap = IB_ATOMIC_NONE;
+	}
+}
+
+static void get_atomic_caps_qp(struct mlx5_ib_dev *dev,
+			       struct ib_device_attr *props)
+{
+	u8 atomic_size_qp = MLX5_CAP_ATOMIC(dev->mdev, atomic_size_qp);
+
+	get_atomic_caps(dev, atomic_size_qp, props);
+}
+
+static void get_atomic_caps_dc(struct mlx5_ib_dev *dev,
+			       struct ib_device_attr *props)
+{
+	u8 atomic_size_qp = MLX5_CAP_ATOMIC(dev->mdev, atomic_size_dc);
+
+	get_atomic_caps(dev, atomic_size_qp, props);
+}
+
+bool mlx5_ib_dc_atomic_is_supported(struct mlx5_ib_dev *dev)
+{
+	struct ib_device_attr props = {};
+
+	get_atomic_caps_dc(dev, &props);
+	return (props.atomic_cap == IB_ATOMIC_HCA) ? true : false;
+}
+static int mlx5_query_system_image_guid(struct ib_device *ibdev,
+					__be64 *sys_image_guid)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+	struct mlx5_core_dev *mdev = dev->mdev;
+	u64 tmp;
+	int err;
+
+	switch (mlx5_get_vport_access_method(ibdev)) {
+	case MLX5_VPORT_ACCESS_METHOD_MAD:
+		return mlx5_query_mad_ifc_system_image_guid(ibdev,
+							    sys_image_guid);
+
+	case MLX5_VPORT_ACCESS_METHOD_HCA:
+		err = mlx5_query_hca_vport_system_image_guid(mdev, &tmp);
+		break;
+
+	case MLX5_VPORT_ACCESS_METHOD_NIC:
+		err = mlx5_query_nic_vport_system_image_guid(mdev, &tmp);
+		break;
+
+	default:
+		return -EINVAL;
+	}
+
+	if (!err)
+		*sys_image_guid = cpu_to_be64(tmp);
+
+	return err;
+
+}
+
+static int mlx5_query_max_pkeys(struct ib_device *ibdev,
+				u16 *max_pkeys)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+	struct mlx5_core_dev *mdev = dev->mdev;
+
+	switch (mlx5_get_vport_access_method(ibdev)) {
+	case MLX5_VPORT_ACCESS_METHOD_MAD:
+		return mlx5_query_mad_ifc_max_pkeys(ibdev, max_pkeys);
+
+	case MLX5_VPORT_ACCESS_METHOD_HCA:
+	case MLX5_VPORT_ACCESS_METHOD_NIC:
+		*max_pkeys = mlx5_to_sw_pkey_sz(MLX5_CAP_GEN(mdev,
+						pkey_table_size));
+		return 0;
+
+	default:
+		return -EINVAL;
+	}
+}
+
+static int mlx5_query_vendor_id(struct ib_device *ibdev,
+				u32 *vendor_id)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+
+	switch (mlx5_get_vport_access_method(ibdev)) {
+	case MLX5_VPORT_ACCESS_METHOD_MAD:
+		return mlx5_query_mad_ifc_vendor_id(ibdev, vendor_id);
+
+	case MLX5_VPORT_ACCESS_METHOD_HCA:
+	case MLX5_VPORT_ACCESS_METHOD_NIC:
+		return mlx5_core_query_vendor_id(dev->mdev, vendor_id);
+
+	default:
+		return -EINVAL;
+	}
+}
+
+static int mlx5_query_node_guid(struct mlx5_ib_dev *dev,
+				__be64 *node_guid)
+{
+	u64 tmp;
+	int err;
+
+	switch (mlx5_get_vport_access_method(&dev->ib_dev)) {
+	case MLX5_VPORT_ACCESS_METHOD_MAD:
+		return mlx5_query_mad_ifc_node_guid(dev, node_guid);
+
+	case MLX5_VPORT_ACCESS_METHOD_HCA:
+		err = mlx5_query_hca_vport_node_guid(dev->mdev, &tmp);
+		break;
+
+	case MLX5_VPORT_ACCESS_METHOD_NIC:
+		err = mlx5_query_nic_vport_node_guid(dev->mdev, &tmp);
+		break;
+
+	default:
+		return -EINVAL;
+	}
+
+	if (!err)
+		*node_guid = cpu_to_be64(tmp);
+
+	return err;
+}
+
+struct mlx5_reg_node_desc {
+	u8	desc[IB_DEVICE_NODE_DESC_MAX];
+};
+
+static int mlx5_query_node_desc(struct mlx5_ib_dev *dev, char *node_desc)
+{
+	struct mlx5_reg_node_desc in;
+
+	if (mlx5_use_mad_ifc(dev))
+		return mlx5_query_mad_ifc_node_desc(dev, node_desc);
+
+	memset(&in, 0, sizeof(in));
+
+	return mlx5_core_access_reg(dev->mdev, &in, sizeof(in), node_desc,
+				    sizeof(struct mlx5_reg_node_desc),
+				    MLX5_REG_NODE_DESC, 0, 0);
+}
+
+static int mlx5_ib_query_device(struct ib_device *ibdev,
+				struct ib_device_attr *props,
+				struct ib_udata *uhw)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+	struct mlx5_core_dev *mdev = dev->mdev;
+	int err = -ENOMEM;
+	int max_sq_desc;
+	int max_rq_sg;
+	int max_sq_sg;
+	u64 min_page_size = 1ull << MLX5_CAP_GEN(mdev, log_pg_sz);
+	bool raw_support = !mlx5_core_mp_enabled(mdev);
+	struct mlx5_ib_query_device_resp resp = {};
+	size_t resp_len;
+	u64 max_tso;
+
+	resp_len = sizeof(resp.comp_mask) + sizeof(resp.response_length);
+	if (uhw->outlen && uhw->outlen < resp_len)
+		return -EINVAL;
+	else
+		resp.response_length = resp_len;
+
+	if (uhw->inlen && !ib_is_udata_cleared(uhw, 0, uhw->inlen))
+		return -EINVAL;
+
+	memset(props, 0, sizeof(*props));
+	err = mlx5_query_system_image_guid(ibdev,
+					   &props->sys_image_guid);
+	if (err)
+		return err;
+
+	err = mlx5_query_max_pkeys(ibdev, &props->max_pkeys);
+	if (err)
+		return err;
+
+	err = mlx5_query_vendor_id(ibdev, &props->vendor_id);
+	if (err)
+		return err;
+
+	props->fw_ver = ((u64)fw_rev_maj(dev->mdev) << 32) |
+		(fw_rev_min(dev->mdev) << 16) |
+		fw_rev_sub(dev->mdev);
+	props->device_cap_flags    = IB_DEVICE_CHANGE_PHY_PORT |
+		IB_DEVICE_PORT_ACTIVE_EVENT		|
+		IB_DEVICE_SYS_IMAGE_GUID		|
+		IB_DEVICE_RC_RNR_NAK_GEN;
+
+	if (MLX5_CAP_GEN(mdev, pkv))
+		props->device_cap_flags |= IB_DEVICE_BAD_PKEY_CNTR;
+	if (MLX5_CAP_GEN(mdev, qkv))
+		props->device_cap_flags |= IB_DEVICE_BAD_QKEY_CNTR;
+	if (MLX5_CAP_GEN(mdev, apm))
+		props->device_cap_flags |= IB_DEVICE_AUTO_PATH_MIG;
+	if (MLX5_CAP_GEN(mdev, xrc))
+		props->device_cap_flags |= IB_DEVICE_XRC;
+	if (MLX5_CAP_GEN(mdev, imaicl)) {
+		props->device_cap_flags |= IB_DEVICE_MEM_WINDOW |
+					   IB_DEVICE_MEM_WINDOW_TYPE_2B;
+		props->max_mw = 1 << MLX5_CAP_GEN(mdev, log_max_mkey);
+		/* We support 'Gappy' memory registration too */
+		props->device_cap_flags |= IB_DEVICE_SG_GAPS_REG;
+	}
+	props->device_cap_flags |= IB_DEVICE_MEM_MGT_EXTENSIONS;
+	if (MLX5_CAP_GEN(mdev, sho)) {
+		props->device_cap_flags |= IB_DEVICE_SIGNATURE_HANDOVER;
+		/* At this stage no support for signature handover */
+		props->sig_prot_cap = IB_PROT_T10DIF_TYPE_1 |
+				      IB_PROT_T10DIF_TYPE_2 |
+				      IB_PROT_T10DIF_TYPE_3;
+		props->sig_guard_cap = IB_GUARD_T10DIF_CRC |
+				       IB_GUARD_T10DIF_CSUM;
+	}
+	if (MLX5_CAP_GEN(mdev, block_lb_mc))
+		props->device_cap_flags |= IB_DEVICE_BLOCK_MULTICAST_LOOPBACK;
+
+	if (MLX5_CAP_GEN(dev->mdev, eth_net_offloads) && raw_support) {
+		if (MLX5_CAP_ETH(mdev, csum_cap)) {
+			/* Legacy bit to support old userspace libraries */
+			props->device_cap_flags |= IB_DEVICE_RAW_IP_CSUM;
+			props->raw_packet_caps |= IB_RAW_PACKET_CAP_IP_CSUM;
+		}
+
+		if (MLX5_CAP_ETH(dev->mdev, vlan_cap))
+			props->raw_packet_caps |=
+				IB_RAW_PACKET_CAP_CVLAN_STRIPPING;
+
+		if (field_avail(typeof(resp), tso_caps, uhw->outlen)) {
+			max_tso = MLX5_CAP_ETH(mdev, max_lso_cap);
+			if (max_tso) {
+				resp.tso_caps.max_tso = 1 << max_tso;
+				resp.tso_caps.supported_qpts |=
+					1 << IB_QPT_RAW_PACKET;
+				resp.response_length += sizeof(resp.tso_caps);
+			}
+		}
+
+		if (field_avail(typeof(resp), rss_caps, uhw->outlen)) {
+			resp.rss_caps.rx_hash_function =
+						MLX5_RX_HASH_FUNC_TOEPLITZ;
+			resp.rss_caps.rx_hash_fields_mask =
+						MLX5_RX_HASH_SRC_IPV4 |
+						MLX5_RX_HASH_DST_IPV4 |
+						MLX5_RX_HASH_SRC_IPV6 |
+						MLX5_RX_HASH_DST_IPV6 |
+						MLX5_RX_HASH_SRC_PORT_TCP |
+						MLX5_RX_HASH_DST_PORT_TCP |
+						MLX5_RX_HASH_SRC_PORT_UDP |
+						MLX5_RX_HASH_DST_PORT_UDP |
+						MLX5_RX_HASH_INNER;
+			if (mlx5_accel_ipsec_device_caps(dev->mdev) &
+			    MLX5_ACCEL_IPSEC_CAP_DEVICE)
+				resp.rss_caps.rx_hash_fields_mask |=
+					MLX5_RX_HASH_IPSEC_SPI;
+			resp.response_length += sizeof(resp.rss_caps);
+		}
+	} else {
+		if (field_avail(typeof(resp), tso_caps, uhw->outlen))
+			resp.response_length += sizeof(resp.tso_caps);
+		if (field_avail(typeof(resp), rss_caps, uhw->outlen))
+			resp.response_length += sizeof(resp.rss_caps);
+	}
+
+	if (MLX5_CAP_GEN(mdev, ipoib_basic_offloads)) {
+		props->device_cap_flags |= IB_DEVICE_UD_IP_CSUM;
+		props->device_cap_flags |= IB_DEVICE_UD_TSO;
+	}
+
+	if (MLX5_CAP_GEN(dev->mdev, rq_delay_drop) &&
+	    MLX5_CAP_GEN(dev->mdev, general_notification_event) &&
+	    raw_support)
+		props->raw_packet_caps |= IB_RAW_PACKET_CAP_DELAY_DROP;
+
+	if (MLX5_CAP_GEN(mdev, ipoib_enhanced_offloads) &&
+	    MLX5_CAP_IPOIB_ENHANCED(mdev, csum_cap))
+		props->device_cap_flags |= IB_DEVICE_UD_IP_CSUM;
+
+	if (MLX5_CAP_GEN(dev->mdev, eth_net_offloads) &&
+	    MLX5_CAP_ETH(dev->mdev, scatter_fcs) &&
+	    raw_support) {
+		/* Legacy bit to support old userspace libraries */
+		props->device_cap_flags |= IB_DEVICE_RAW_SCATTER_FCS;
+		props->raw_packet_caps |= IB_RAW_PACKET_CAP_SCATTER_FCS;
+	}
+
+	if (MLX5_CAP_DEV_MEM(mdev, memic)) {
+		props->max_dm_size =
+			MLX5_CAP_DEV_MEM(mdev, max_memic_size);
+	}
+
+	if (mlx5_get_flow_namespace(dev->mdev, MLX5_FLOW_NAMESPACE_BYPASS))
+		props->device_cap_flags |= IB_DEVICE_MANAGED_FLOW_STEERING;
+
+	if (MLX5_CAP_GEN(mdev, end_pad))
+		props->device_cap_flags |= IB_DEVICE_PCI_WRITE_END_PADDING;
+
+	props->vendor_part_id	   = mdev->pdev->device;
+	props->hw_ver		   = mdev->pdev->revision;
+
+	props->max_mr_size	   = ~0ull;
+	props->page_size_cap	   = ~(min_page_size - 1);
+	props->max_qp		   = 1 << MLX5_CAP_GEN(mdev, log_max_qp);
+	props->max_qp_wr	   = 1 << MLX5_CAP_GEN(mdev, log_max_qp_sz);
+	max_rq_sg =  MLX5_CAP_GEN(mdev, max_wqe_sz_rq) /
+		     sizeof(struct mlx5_wqe_data_seg);
+	max_sq_desc = min_t(int, MLX5_CAP_GEN(mdev, max_wqe_sz_sq), 512);
+	max_sq_sg = (max_sq_desc - sizeof(struct mlx5_wqe_ctrl_seg) -
+		     sizeof(struct mlx5_wqe_raddr_seg)) /
+		sizeof(struct mlx5_wqe_data_seg);
+	props->max_send_sge = max_sq_sg;
+	props->max_recv_sge = max_rq_sg;
+	props->max_sge_rd	   = MLX5_MAX_SGE_RD;
+	props->max_cq		   = 1 << MLX5_CAP_GEN(mdev, log_max_cq);
+	props->max_cqe = (1 << MLX5_CAP_GEN(mdev, log_max_cq_sz)) - 1;
+	props->max_mr		   = 1 << MLX5_CAP_GEN(mdev, log_max_mkey);
+	props->max_pd		   = 1 << MLX5_CAP_GEN(mdev, log_max_pd);
+	props->max_qp_rd_atom	   = 1 << MLX5_CAP_GEN(mdev, log_max_ra_req_qp);
+	props->max_qp_init_rd_atom = 1 << MLX5_CAP_GEN(mdev, log_max_ra_res_qp);
+	props->max_srq		   = 1 << MLX5_CAP_GEN(mdev, log_max_srq);
+	props->max_srq_wr = (1 << MLX5_CAP_GEN(mdev, log_max_srq_sz)) - 1;
+	props->local_ca_ack_delay  = MLX5_CAP_GEN(mdev, local_ca_ack_delay);
+	props->max_res_rd_atom	   = props->max_qp_rd_atom * props->max_qp;
+	props->max_srq_sge	   = max_rq_sg - 1;
+	props->max_fast_reg_page_list_len =
+		1 << MLX5_CAP_GEN(mdev, log_max_klm_list_size);
+	get_atomic_caps_qp(dev, props);
+	props->masked_atomic_cap   = IB_ATOMIC_NONE;
+	props->max_mcast_grp	   = 1 << MLX5_CAP_GEN(mdev, log_max_mcg);
+	props->max_mcast_qp_attach = MLX5_CAP_GEN(mdev, max_qp_mcg);
+	props->max_total_mcast_qp_attach = props->max_mcast_qp_attach *
+					   props->max_mcast_grp;
+	props->max_map_per_fmr = INT_MAX; /* no limit in ConnectIB */
+	props->max_ah = INT_MAX;
+	props->hca_core_clock = MLX5_CAP_GEN(mdev, device_frequency_khz);
+	props->timestamp_mask = 0x7FFFFFFFFFFFFFFFULL;
+
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
+	if (MLX5_CAP_GEN(mdev, pg))
+		props->device_cap_flags |= IB_DEVICE_ON_DEMAND_PAGING;
+	props->odp_caps = dev->odp_caps;
+#endif
+
+	if (MLX5_CAP_GEN(mdev, cd))
+		props->device_cap_flags |= IB_DEVICE_CROSS_CHANNEL;
+
+	if (!mlx5_core_is_pf(mdev))
+		props->device_cap_flags |= IB_DEVICE_VIRTUAL_FUNCTION;
+
+	if (mlx5_ib_port_link_layer(ibdev, 1) ==
+	    IB_LINK_LAYER_ETHERNET && raw_support) {
+		props->rss_caps.max_rwq_indirection_tables =
+			1 << MLX5_CAP_GEN(dev->mdev, log_max_rqt);
+		props->rss_caps.max_rwq_indirection_table_size =
+			1 << MLX5_CAP_GEN(dev->mdev, log_max_rqt_size);
+		props->rss_caps.supported_qpts = 1 << IB_QPT_RAW_PACKET;
+		props->max_wq_type_rq =
+			1 << MLX5_CAP_GEN(dev->mdev, log_max_rq);
+	}
+
+	if (MLX5_CAP_GEN(mdev, tag_matching)) {
+		props->tm_caps.max_num_tags =
+			(1 << MLX5_CAP_GEN(mdev, log_tag_matching_list_sz)) - 1;
+		props->tm_caps.max_ops =
+			1 << MLX5_CAP_GEN(mdev, log_max_qp_sz);
+		props->tm_caps.max_sge = MLX5_TM_MAX_SGE;
+	}
+
+	if (MLX5_CAP_GEN(mdev, tag_matching) &&
+	    MLX5_CAP_GEN(mdev, rndv_offload_rc)) {
+		props->tm_caps.flags = IB_TM_CAP_RNDV_RC;
+		props->tm_caps.max_rndv_hdr_size = MLX5_TM_MAX_RNDV_MSG_SIZE;
+	}
+
+	if (MLX5_CAP_GEN(dev->mdev, cq_moderation)) {
+		props->cq_caps.max_cq_moderation_count =
+						MLX5_MAX_CQ_COUNT;
+		props->cq_caps.max_cq_moderation_period =
+						MLX5_MAX_CQ_PERIOD;
+	}
+
+	if (field_avail(typeof(resp), cqe_comp_caps, uhw->outlen)) {
+		resp.response_length += sizeof(resp.cqe_comp_caps);
+
+		if (MLX5_CAP_GEN(dev->mdev, cqe_compression)) {
+			resp.cqe_comp_caps.max_num =
+				MLX5_CAP_GEN(dev->mdev,
+					     cqe_compression_max_num);
+
+			resp.cqe_comp_caps.supported_format =
+				MLX5_IB_CQE_RES_FORMAT_HASH |
+				MLX5_IB_CQE_RES_FORMAT_CSUM;
+
+			if (MLX5_CAP_GEN(dev->mdev, mini_cqe_resp_stride_index))
+				resp.cqe_comp_caps.supported_format |=
+					MLX5_IB_CQE_RES_FORMAT_CSUM_STRIDX;
+		}
+	}
+
+	if (field_avail(typeof(resp), packet_pacing_caps, uhw->outlen) &&
+	    raw_support) {
+		if (MLX5_CAP_QOS(mdev, packet_pacing) &&
+		    MLX5_CAP_GEN(mdev, qos)) {
+			resp.packet_pacing_caps.qp_rate_limit_max =
+				MLX5_CAP_QOS(mdev, packet_pacing_max_rate);
+			resp.packet_pacing_caps.qp_rate_limit_min =
+				MLX5_CAP_QOS(mdev, packet_pacing_min_rate);
+			resp.packet_pacing_caps.supported_qpts |=
+				1 << IB_QPT_RAW_PACKET;
+			if (MLX5_CAP_QOS(mdev, packet_pacing_burst_bound) &&
+			    MLX5_CAP_QOS(mdev, packet_pacing_typical_size))
+				resp.packet_pacing_caps.cap_flags |=
+					MLX5_IB_PP_SUPPORT_BURST;
+		}
+		resp.response_length += sizeof(resp.packet_pacing_caps);
+	}
+
+	if (field_avail(typeof(resp), mlx5_ib_support_multi_pkt_send_wqes,
+			uhw->outlen)) {
+		if (MLX5_CAP_ETH(mdev, multi_pkt_send_wqe))
+			resp.mlx5_ib_support_multi_pkt_send_wqes =
+				MLX5_IB_ALLOW_MPW;
+
+		if (MLX5_CAP_ETH(mdev, enhanced_multi_pkt_send_wqe))
+			resp.mlx5_ib_support_multi_pkt_send_wqes |=
+				MLX5_IB_SUPPORT_EMPW;
+
+		resp.response_length +=
+			sizeof(resp.mlx5_ib_support_multi_pkt_send_wqes);
+	}
+
+	if (field_avail(typeof(resp), flags, uhw->outlen)) {
+		resp.response_length += sizeof(resp.flags);
+
+		if (MLX5_CAP_GEN(mdev, cqe_compression_128))
+			resp.flags |=
+				MLX5_IB_QUERY_DEV_RESP_FLAGS_CQE_128B_COMP;
+
+		if (MLX5_CAP_GEN(mdev, cqe_128_always))
+			resp.flags |= MLX5_IB_QUERY_DEV_RESP_FLAGS_CQE_128B_PAD;
+	}
+
+	if (field_avail(typeof(resp), sw_parsing_caps,
+			uhw->outlen)) {
+		resp.response_length += sizeof(resp.sw_parsing_caps);
+		if (MLX5_CAP_ETH(mdev, swp)) {
+			resp.sw_parsing_caps.sw_parsing_offloads |=
+				MLX5_IB_SW_PARSING;
+
+			if (MLX5_CAP_ETH(mdev, swp_csum))
+				resp.sw_parsing_caps.sw_parsing_offloads |=
+					MLX5_IB_SW_PARSING_CSUM;
+
+			if (MLX5_CAP_ETH(mdev, swp_lso))
+				resp.sw_parsing_caps.sw_parsing_offloads |=
+					MLX5_IB_SW_PARSING_LSO;
+
+			if (resp.sw_parsing_caps.sw_parsing_offloads)
+				resp.sw_parsing_caps.supported_qpts =
+					BIT(IB_QPT_RAW_PACKET);
+		}
+	}
+
+	if (field_avail(typeof(resp), striding_rq_caps, uhw->outlen) &&
+	    raw_support) {
+		resp.response_length += sizeof(resp.striding_rq_caps);
+		if (MLX5_CAP_GEN(mdev, striding_rq)) {
+			resp.striding_rq_caps.min_single_stride_log_num_of_bytes =
+				MLX5_MIN_SINGLE_STRIDE_LOG_NUM_BYTES;
+			resp.striding_rq_caps.max_single_stride_log_num_of_bytes =
+				MLX5_MAX_SINGLE_STRIDE_LOG_NUM_BYTES;
+			resp.striding_rq_caps.min_single_wqe_log_num_of_strides =
+				MLX5_MIN_SINGLE_WQE_LOG_NUM_STRIDES;
+			resp.striding_rq_caps.max_single_wqe_log_num_of_strides =
+				MLX5_MAX_SINGLE_WQE_LOG_NUM_STRIDES;
+			resp.striding_rq_caps.supported_qpts =
+				BIT(IB_QPT_RAW_PACKET);
+		}
+	}
+
+	if (field_avail(typeof(resp), tunnel_offloads_caps,
+			uhw->outlen)) {
+		resp.response_length += sizeof(resp.tunnel_offloads_caps);
+		if (MLX5_CAP_ETH(mdev, tunnel_stateless_vxlan))
+			resp.tunnel_offloads_caps |=
+				MLX5_IB_TUNNELED_OFFLOADS_VXLAN;
+		if (MLX5_CAP_ETH(mdev, tunnel_stateless_geneve_rx))
+			resp.tunnel_offloads_caps |=
+				MLX5_IB_TUNNELED_OFFLOADS_GENEVE;
+		if (MLX5_CAP_ETH(mdev, tunnel_stateless_gre))
+			resp.tunnel_offloads_caps |=
+				MLX5_IB_TUNNELED_OFFLOADS_GRE;
+		if (MLX5_CAP_ETH(mdev, tunnel_stateless_mpls_over_gre))
+			resp.tunnel_offloads_caps |=
+				MLX5_IB_TUNNELED_OFFLOADS_MPLS_GRE;
+		if (MLX5_CAP_ETH(mdev, tunnel_stateless_mpls_over_udp))
+			resp.tunnel_offloads_caps |=
+				MLX5_IB_TUNNELED_OFFLOADS_MPLS_UDP;
+	}
+
+	if (uhw->outlen) {
+		err = ib_copy_to_udata(uhw, &resp, resp.response_length);
+
+		if (err)
+			return err;
+	}
+
+	return 0;
+}
+
+enum mlx5_ib_width {
+	MLX5_IB_WIDTH_1X	= 1 << 0,
+	MLX5_IB_WIDTH_2X	= 1 << 1,
+	MLX5_IB_WIDTH_4X	= 1 << 2,
+	MLX5_IB_WIDTH_8X	= 1 << 3,
+	MLX5_IB_WIDTH_12X	= 1 << 4
+};
+
+static void translate_active_width(struct ib_device *ibdev, u8 active_width,
+				  u8 *ib_width)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+
+	if (active_width & MLX5_IB_WIDTH_1X)
+		*ib_width = IB_WIDTH_1X;
+	else if (active_width & MLX5_IB_WIDTH_4X)
+		*ib_width = IB_WIDTH_4X;
+	else if (active_width & MLX5_IB_WIDTH_8X)
+		*ib_width = IB_WIDTH_8X;
+	else if (active_width & MLX5_IB_WIDTH_12X)
+		*ib_width = IB_WIDTH_12X;
+	else {
+		mlx5_ib_dbg(dev, "Invalid active_width %d, setting width to default value: 4x\n",
+			    (int)active_width);
+		*ib_width = IB_WIDTH_4X;
+	}
+
+	return;
+}
+
+static int mlx5_mtu_to_ib_mtu(int mtu)
+{
+	switch (mtu) {
+	case 256: return 1;
+	case 512: return 2;
+	case 1024: return 3;
+	case 2048: return 4;
+	case 4096: return 5;
+	default:
+		pr_warn("invalid mtu\n");
+		return -1;
+	}
+}
+
+enum ib_max_vl_num {
+	__IB_MAX_VL_0		= 1,
+	__IB_MAX_VL_0_1		= 2,
+	__IB_MAX_VL_0_3		= 3,
+	__IB_MAX_VL_0_7		= 4,
+	__IB_MAX_VL_0_14	= 5,
+};
+
+enum mlx5_vl_hw_cap {
+	MLX5_VL_HW_0	= 1,
+	MLX5_VL_HW_0_1	= 2,
+	MLX5_VL_HW_0_2	= 3,
+	MLX5_VL_HW_0_3	= 4,
+	MLX5_VL_HW_0_4	= 5,
+	MLX5_VL_HW_0_5	= 6,
+	MLX5_VL_HW_0_6	= 7,
+	MLX5_VL_HW_0_7	= 8,
+	MLX5_VL_HW_0_14	= 15
+};
+
+static int translate_max_vl_num(struct ib_device *ibdev, u8 vl_hw_cap,
+				u8 *max_vl_num)
+{
+	switch (vl_hw_cap) {
+	case MLX5_VL_HW_0:
+		*max_vl_num = __IB_MAX_VL_0;
+		break;
+	case MLX5_VL_HW_0_1:
+		*max_vl_num = __IB_MAX_VL_0_1;
+		break;
+	case MLX5_VL_HW_0_3:
+		*max_vl_num = __IB_MAX_VL_0_3;
+		break;
+	case MLX5_VL_HW_0_7:
+		*max_vl_num = __IB_MAX_VL_0_7;
+		break;
+	case MLX5_VL_HW_0_14:
+		*max_vl_num = __IB_MAX_VL_0_14;
+		break;
+
+	default:
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int mlx5_query_hca_port(struct ib_device *ibdev, u8 port,
+			       struct ib_port_attr *props)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+	struct mlx5_core_dev *mdev = dev->mdev;
+	struct mlx5_hca_vport_context *rep;
+	u16 max_mtu;
+	u16 oper_mtu;
+	int err;
+	u8 ib_link_width_oper;
+	u8 vl_hw_cap;
+
+	rep = kzalloc(sizeof(*rep), GFP_KERNEL);
+	if (!rep) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	/* props being zeroed by the caller, avoid zeroing it here */
+
+	err = mlx5_query_hca_vport_context(mdev, 0, port, 0, rep);
+	if (err)
+		goto out;
+
+	props->lid		= rep->lid;
+	props->lmc		= rep->lmc;
+	props->sm_lid		= rep->sm_lid;
+	props->sm_sl		= rep->sm_sl;
+	props->state		= rep->vport_state;
+	props->phys_state	= rep->port_physical_state;
+	props->port_cap_flags	= rep->cap_mask1;
+	props->gid_tbl_len	= mlx5_get_gid_table_len(MLX5_CAP_GEN(mdev, gid_table_size));
+	props->max_msg_sz	= 1 << MLX5_CAP_GEN(mdev, log_max_msg);
+	props->pkey_tbl_len	= mlx5_to_sw_pkey_sz(MLX5_CAP_GEN(mdev, pkey_table_size));
+	props->bad_pkey_cntr	= rep->pkey_violation_counter;
+	props->qkey_viol_cntr	= rep->qkey_violation_counter;
+	props->subnet_timeout	= rep->subnet_timeout;
+	props->init_type_reply	= rep->init_type_reply;
+
+	err = mlx5_query_port_link_width_oper(mdev, &ib_link_width_oper, port);
+	if (err)
+		goto out;
+
+	translate_active_width(ibdev, ib_link_width_oper, &props->active_width);
+
+	err = mlx5_query_port_ib_proto_oper(mdev, &props->active_speed, port);
+	if (err)
+		goto out;
+
+	mlx5_query_port_max_mtu(mdev, &max_mtu, port);
+
+	props->max_mtu = mlx5_mtu_to_ib_mtu(max_mtu);
+
+	mlx5_query_port_oper_mtu(mdev, &oper_mtu, port);
+
+	props->active_mtu = mlx5_mtu_to_ib_mtu(oper_mtu);
+
+	err = mlx5_query_port_vl_hw_cap(mdev, &vl_hw_cap, port);
+	if (err)
+		goto out;
+
+	err = translate_max_vl_num(ibdev, vl_hw_cap,
+				   &props->max_vl_num);
+out:
+	kfree(rep);
+	return err;
+}
+
+int mlx5_ib_query_port(struct ib_device *ibdev, u8 port,
+		       struct ib_port_attr *props)
+{
+	unsigned int count;
+	int ret;
+
+	switch (mlx5_get_vport_access_method(ibdev)) {
+	case MLX5_VPORT_ACCESS_METHOD_MAD:
+		ret = mlx5_query_mad_ifc_port(ibdev, port, props);
+		break;
+
+	case MLX5_VPORT_ACCESS_METHOD_HCA:
+		ret = mlx5_query_hca_port(ibdev, port, props);
+		break;
+
+	case MLX5_VPORT_ACCESS_METHOD_NIC:
+		ret = mlx5_query_port_roce(ibdev, port, props);
+		break;
+
+	default:
+		ret = -EINVAL;
+	}
+
+	if (!ret && props) {
+		struct mlx5_ib_dev *dev = to_mdev(ibdev);
+		struct mlx5_core_dev *mdev;
+		bool put_mdev = true;
+
+		mdev = mlx5_ib_get_native_port_mdev(dev, port, NULL);
+		if (!mdev) {
+			/* If the port isn't affiliated yet query the master.
+			 * The master and slave will have the same values.
+			 */
+			mdev = dev->mdev;
+			port = 1;
+			put_mdev = false;
+		}
+		count = mlx5_core_reserved_gids_count(mdev);
+		if (put_mdev)
+			mlx5_ib_put_native_port_mdev(dev, port);
+		props->gid_tbl_len -= count;
+	}
+	return ret;
+}
+
+static int mlx5_ib_rep_query_port(struct ib_device *ibdev, u8 port,
+				  struct ib_port_attr *props)
+{
+	int ret;
+
+	/* Only link layer == ethernet is valid for representors */
+	ret = mlx5_query_port_roce(ibdev, port, props);
+	if (ret || !props)
+		return ret;
+
+	/* We don't support GIDS */
+	props->gid_tbl_len = 0;
+
+	return ret;
+}
+
+static int mlx5_ib_query_gid(struct ib_device *ibdev, u8 port, int index,
+			     union ib_gid *gid)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+	struct mlx5_core_dev *mdev = dev->mdev;
+
+	switch (mlx5_get_vport_access_method(ibdev)) {
+	case MLX5_VPORT_ACCESS_METHOD_MAD:
+		return mlx5_query_mad_ifc_gids(ibdev, port, index, gid);
+
+	case MLX5_VPORT_ACCESS_METHOD_HCA:
+		return mlx5_query_hca_vport_gid(mdev, 0, port, 0, index, gid);
+
+	default:
+		return -EINVAL;
+	}
+
+}
+
+static int mlx5_query_hca_nic_pkey(struct ib_device *ibdev, u8 port,
+				   u16 index, u16 *pkey)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+	struct mlx5_core_dev *mdev;
+	bool put_mdev = true;
+	u8 mdev_port_num;
+	int err;
+
+	mdev = mlx5_ib_get_native_port_mdev(dev, port, &mdev_port_num);
+	if (!mdev) {
+		/* The port isn't affiliated yet, get the PKey from the master
+		 * port. For RoCE the PKey tables will be the same.
+		 */
+		put_mdev = false;
+		mdev = dev->mdev;
+		mdev_port_num = 1;
+	}
+
+	err = mlx5_query_hca_vport_pkey(mdev, 0, mdev_port_num, 0,
+					index, pkey);
+	if (put_mdev)
+		mlx5_ib_put_native_port_mdev(dev, port);
+
+	return err;
+}
+
+static int mlx5_ib_query_pkey(struct ib_device *ibdev, u8 port, u16 index,
+			      u16 *pkey)
+{
+	switch (mlx5_get_vport_access_method(ibdev)) {
+	case MLX5_VPORT_ACCESS_METHOD_MAD:
+		return mlx5_query_mad_ifc_pkey(ibdev, port, index, pkey);
+
+	case MLX5_VPORT_ACCESS_METHOD_HCA:
+	case MLX5_VPORT_ACCESS_METHOD_NIC:
+		return mlx5_query_hca_nic_pkey(ibdev, port, index, pkey);
+	default:
+		return -EINVAL;
+	}
+}
+
+static int mlx5_ib_modify_device(struct ib_device *ibdev, int mask,
+				 struct ib_device_modify *props)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+	struct mlx5_reg_node_desc in;
+	struct mlx5_reg_node_desc out;
+	int err;
+
+	if (mask & ~IB_DEVICE_MODIFY_NODE_DESC)
+		return -EOPNOTSUPP;
+
+	if (!(mask & IB_DEVICE_MODIFY_NODE_DESC))
+		return 0;
+
+	/*
+	 * If possible, pass node desc to FW, so it can generate
+	 * a 144 trap.  If cmd fails, just ignore.
+	 */
+	memcpy(&in, props->node_desc, IB_DEVICE_NODE_DESC_MAX);
+	err = mlx5_core_access_reg(dev->mdev, &in, sizeof(in), &out,
+				   sizeof(out), MLX5_REG_NODE_DESC, 0, 1);
+	if (err)
+		return err;
+
+	memcpy(ibdev->node_desc, props->node_desc, IB_DEVICE_NODE_DESC_MAX);
+
+	return err;
+}
+
+static int set_port_caps_atomic(struct mlx5_ib_dev *dev, u8 port_num, u32 mask,
+				u32 value)
+{
+	struct mlx5_hca_vport_context ctx = {};
+	struct mlx5_core_dev *mdev;
+	u8 mdev_port_num;
+	int err;
+
+	mdev = mlx5_ib_get_native_port_mdev(dev, port_num, &mdev_port_num);
+	if (!mdev)
+		return -ENODEV;
+
+	err = mlx5_query_hca_vport_context(mdev, 0, mdev_port_num, 0, &ctx);
+	if (err)
+		goto out;
+
+	if (~ctx.cap_mask1_perm & mask) {
+		mlx5_ib_warn(dev, "trying to change bitmask 0x%X but change supported 0x%X\n",
+			     mask, ctx.cap_mask1_perm);
+		err = -EINVAL;
+		goto out;
+	}
+
+	ctx.cap_mask1 = value;
+	ctx.cap_mask1_perm = mask;
+	err = mlx5_core_modify_hca_vport_context(mdev, 0, mdev_port_num,
+						 0, &ctx);
+
+out:
+	mlx5_ib_put_native_port_mdev(dev, port_num);
+
+	return err;
+}
+
+static int mlx5_ib_modify_port(struct ib_device *ibdev, u8 port, int mask,
+			       struct ib_port_modify *props)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+	struct ib_port_attr attr;
+	u32 tmp;
+	int err;
+	u32 change_mask;
+	u32 value;
+	bool is_ib = (mlx5_ib_port_link_layer(ibdev, port) ==
+		      IB_LINK_LAYER_INFINIBAND);
+
+	/* CM layer calls ib_modify_port() regardless of the link layer. For
+	 * Ethernet ports, qkey violation and Port capabilities are meaningless.
+	 */
+	if (!is_ib)
+		return 0;
+
+	if (MLX5_CAP_GEN(dev->mdev, ib_virt) && is_ib) {
+		change_mask = props->clr_port_cap_mask | props->set_port_cap_mask;
+		value = ~props->clr_port_cap_mask | props->set_port_cap_mask;
+		return set_port_caps_atomic(dev, port, change_mask, value);
+	}
+
+	mutex_lock(&dev->cap_mask_mutex);
+
+	err = ib_query_port(ibdev, port, &attr);
+	if (err)
+		goto out;
+
+	tmp = (attr.port_cap_flags | props->set_port_cap_mask) &
+		~props->clr_port_cap_mask;
+
+	err = mlx5_set_port_caps(dev->mdev, port, tmp);
+
+out:
+	mutex_unlock(&dev->cap_mask_mutex);
+	return err;
+}
+
+static void print_lib_caps(struct mlx5_ib_dev *dev, u64 caps)
+{
+	mlx5_ib_dbg(dev, "MLX5_LIB_CAP_4K_UAR = %s\n",
+		    caps & MLX5_LIB_CAP_4K_UAR ? "y" : "n");
+}
+
+static u16 calc_dynamic_bfregs(int uars_per_sys_page)
+{
+	/* Large page with non 4k uar support might limit the dynamic size */
+	if (uars_per_sys_page == 1  && PAGE_SIZE > 4096)
+		return MLX5_MIN_DYN_BFREGS;
+
+	return MLX5_MAX_DYN_BFREGS;
+}
+
+static int calc_total_bfregs(struct mlx5_ib_dev *dev, bool lib_uar_4k,
+			     struct mlx5_ib_alloc_ucontext_req_v2 *req,
+			     struct mlx5_bfreg_info *bfregi)
+{
+	int uars_per_sys_page;
+	int bfregs_per_sys_page;
+	int ref_bfregs = req->total_num_bfregs;
+
+	if (req->total_num_bfregs == 0)
+		return -EINVAL;
+
+	BUILD_BUG_ON(MLX5_MAX_BFREGS % MLX5_NON_FP_BFREGS_IN_PAGE);
+	BUILD_BUG_ON(MLX5_MAX_BFREGS < MLX5_NON_FP_BFREGS_IN_PAGE);
+
+	if (req->total_num_bfregs > MLX5_MAX_BFREGS)
+		return -ENOMEM;
+
+	uars_per_sys_page = get_uars_per_sys_page(dev, lib_uar_4k);
+	bfregs_per_sys_page = uars_per_sys_page * MLX5_NON_FP_BFREGS_PER_UAR;
+	/* This holds the required static allocation asked by the user */
+	req->total_num_bfregs = ALIGN(req->total_num_bfregs, bfregs_per_sys_page);
+	if (req->num_low_latency_bfregs > req->total_num_bfregs - 1)
+		return -EINVAL;
+
+	bfregi->num_static_sys_pages = req->total_num_bfregs / bfregs_per_sys_page;
+	bfregi->num_dyn_bfregs = ALIGN(calc_dynamic_bfregs(uars_per_sys_page), bfregs_per_sys_page);
+	bfregi->total_num_bfregs = req->total_num_bfregs + bfregi->num_dyn_bfregs;
+	bfregi->num_sys_pages = bfregi->total_num_bfregs / bfregs_per_sys_page;
+
+	mlx5_ib_dbg(dev, "uar_4k: fw support %s, lib support %s, user requested %d bfregs, allocated %d, total bfregs %d, using %d sys pages\n",
+		    MLX5_CAP_GEN(dev->mdev, uar_4k) ? "yes" : "no",
+		    lib_uar_4k ? "yes" : "no", ref_bfregs,
+		    req->total_num_bfregs, bfregi->total_num_bfregs,
+		    bfregi->num_sys_pages);
+
+	return 0;
+}
+
+static int allocate_uars(struct mlx5_ib_dev *dev, struct mlx5_ib_ucontext *context)
+{
+	struct mlx5_bfreg_info *bfregi;
+	int err;
+	int i;
+
+	bfregi = &context->bfregi;
+	for (i = 0; i < bfregi->num_static_sys_pages; i++) {
+		err = mlx5_cmd_alloc_uar(dev->mdev, &bfregi->sys_pages[i]);
+		if (err)
+			goto error;
+
+		mlx5_ib_dbg(dev, "allocated uar %d\n", bfregi->sys_pages[i]);
+	}
+
+	for (i = bfregi->num_static_sys_pages; i < bfregi->num_sys_pages; i++)
+		bfregi->sys_pages[i] = MLX5_IB_INVALID_UAR_INDEX;
+
+	return 0;
+
+error:
+	for (--i; i >= 0; i--)
+		if (mlx5_cmd_free_uar(dev->mdev, bfregi->sys_pages[i]))
+			mlx5_ib_warn(dev, "failed to free uar %d\n", i);
+
+	return err;
+}
+
+static void deallocate_uars(struct mlx5_ib_dev *dev,
+			    struct mlx5_ib_ucontext *context)
+{
+	struct mlx5_bfreg_info *bfregi;
+	int i;
+
+	bfregi = &context->bfregi;
+	for (i = 0; i < bfregi->num_sys_pages; i++)
+		if (i < bfregi->num_static_sys_pages ||
+		    bfregi->sys_pages[i] != MLX5_IB_INVALID_UAR_INDEX)
+			mlx5_cmd_free_uar(dev->mdev, bfregi->sys_pages[i]);
+}
+
+static int mlx5_ib_alloc_transport_domain(struct mlx5_ib_dev *dev, u32 *tdn)
+{
+	int err;
+
+	if (!MLX5_CAP_GEN(dev->mdev, log_max_transport_domain))
+		return 0;
+
+	err = mlx5_core_alloc_transport_domain(dev->mdev, tdn);
+	if (err)
+		return err;
+
+	if ((MLX5_CAP_GEN(dev->mdev, port_type) != MLX5_CAP_PORT_TYPE_ETH) ||
+	    (!MLX5_CAP_GEN(dev->mdev, disable_local_lb_uc) &&
+	     !MLX5_CAP_GEN(dev->mdev, disable_local_lb_mc)))
+		return err;
+
+	mutex_lock(&dev->lb_mutex);
+	dev->user_td++;
+
+	if (dev->user_td == 2)
+		err = mlx5_nic_vport_update_local_lb(dev->mdev, true);
+
+	mutex_unlock(&dev->lb_mutex);
+	return err;
+}
+
+static void mlx5_ib_dealloc_transport_domain(struct mlx5_ib_dev *dev, u32 tdn)
+{
+	if (!MLX5_CAP_GEN(dev->mdev, log_max_transport_domain))
+		return;
+
+	mlx5_core_dealloc_transport_domain(dev->mdev, tdn);
+
+	if ((MLX5_CAP_GEN(dev->mdev, port_type) != MLX5_CAP_PORT_TYPE_ETH) ||
+	    (!MLX5_CAP_GEN(dev->mdev, disable_local_lb_uc) &&
+	     !MLX5_CAP_GEN(dev->mdev, disable_local_lb_mc)))
+		return;
+
+	mutex_lock(&dev->lb_mutex);
+	dev->user_td--;
+
+	if (dev->user_td < 2)
+		mlx5_nic_vport_update_local_lb(dev->mdev, false);
+
+	mutex_unlock(&dev->lb_mutex);
+}
+
+static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev,
+						  struct ib_udata *udata)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+	struct mlx5_ib_alloc_ucontext_req_v2 req = {};
+	struct mlx5_ib_alloc_ucontext_resp resp = {};
+	struct mlx5_core_dev *mdev = dev->mdev;
+	struct mlx5_ib_ucontext *context;
+	struct mlx5_bfreg_info *bfregi;
+	int ver;
+	int err;
+	size_t min_req_v2 = offsetof(struct mlx5_ib_alloc_ucontext_req_v2,
+				     max_cqe_version);
+	u32 dump_fill_mkey;
+	bool lib_uar_4k;
+
+	if (!dev->ib_active)
+		return ERR_PTR(-EAGAIN);
+
+	if (udata->inlen == sizeof(struct mlx5_ib_alloc_ucontext_req))
+		ver = 0;
+	else if (udata->inlen >= min_req_v2)
+		ver = 2;
+	else
+		return ERR_PTR(-EINVAL);
+
+	err = ib_copy_from_udata(&req, udata, min(udata->inlen, sizeof(req)));
+	if (err)
+		return ERR_PTR(err);
+
+	if (req.flags & ~MLX5_IB_ALLOC_UCTX_DEVX)
+		return ERR_PTR(-EOPNOTSUPP);
+
+	if (req.comp_mask || req.reserved0 || req.reserved1 || req.reserved2)
+		return ERR_PTR(-EOPNOTSUPP);
+
+	req.total_num_bfregs = ALIGN(req.total_num_bfregs,
+				    MLX5_NON_FP_BFREGS_PER_UAR);
+	if (req.num_low_latency_bfregs > req.total_num_bfregs - 1)
+		return ERR_PTR(-EINVAL);
+
+	resp.qp_tab_size = 1 << MLX5_CAP_GEN(dev->mdev, log_max_qp);
+	if (mlx5_core_is_pf(dev->mdev) && MLX5_CAP_GEN(dev->mdev, bf))
+		resp.bf_reg_size = 1 << MLX5_CAP_GEN(dev->mdev, log_bf_reg_size);
+	resp.cache_line_size = cache_line_size();
+	resp.max_sq_desc_sz = MLX5_CAP_GEN(dev->mdev, max_wqe_sz_sq);
+	resp.max_rq_desc_sz = MLX5_CAP_GEN(dev->mdev, max_wqe_sz_rq);
+	resp.max_send_wqebb = 1 << MLX5_CAP_GEN(dev->mdev, log_max_qp_sz);
+	resp.max_recv_wr = 1 << MLX5_CAP_GEN(dev->mdev, log_max_qp_sz);
+	resp.max_srq_recv_wr = 1 << MLX5_CAP_GEN(dev->mdev, log_max_srq_sz);
+	resp.cqe_version = min_t(__u8,
+				 (__u8)MLX5_CAP_GEN(dev->mdev, cqe_version),
+				 req.max_cqe_version);
+	resp.log_uar_size = MLX5_CAP_GEN(dev->mdev, uar_4k) ?
+				MLX5_ADAPTER_PAGE_SHIFT : PAGE_SHIFT;
+	resp.num_uars_per_page = MLX5_CAP_GEN(dev->mdev, uar_4k) ?
+					MLX5_CAP_GEN(dev->mdev, num_of_uars_per_page) : 1;
+	resp.response_length = min(offsetof(typeof(resp), response_length) +
+				   sizeof(resp.response_length), udata->outlen);
+
+	if (mlx5_accel_ipsec_device_caps(dev->mdev) & MLX5_ACCEL_IPSEC_CAP_DEVICE) {
+		if (mlx5_get_flow_namespace(dev->mdev, MLX5_FLOW_NAMESPACE_EGRESS))
+			resp.flow_action_flags |= MLX5_USER_ALLOC_UCONTEXT_FLOW_ACTION_FLAGS_ESP_AES_GCM;
+		if (mlx5_accel_ipsec_device_caps(dev->mdev) & MLX5_ACCEL_IPSEC_CAP_REQUIRED_METADATA)
+			resp.flow_action_flags |= MLX5_USER_ALLOC_UCONTEXT_FLOW_ACTION_FLAGS_ESP_AES_GCM_REQ_METADATA;
+		if (MLX5_CAP_FLOWTABLE(dev->mdev, flow_table_properties_nic_receive.ft_field_support.outer_esp_spi))
+			resp.flow_action_flags |= MLX5_USER_ALLOC_UCONTEXT_FLOW_ACTION_FLAGS_ESP_AES_GCM_SPI_STEERING;
+		if (mlx5_accel_ipsec_device_caps(dev->mdev) & MLX5_ACCEL_IPSEC_CAP_TX_IV_IS_ESN)
+			resp.flow_action_flags |= MLX5_USER_ALLOC_UCONTEXT_FLOW_ACTION_FLAGS_ESP_AES_GCM_TX_IV_IS_ESN;
+		/* MLX5_USER_ALLOC_UCONTEXT_FLOW_ACTION_FLAGS_ESP_AES_GCM_FULL_OFFLOAD is currently always 0 */
+	}
+
+	context = kzalloc(sizeof(*context), GFP_KERNEL);
+	if (!context)
+		return ERR_PTR(-ENOMEM);
+
+	lib_uar_4k = req.lib_caps & MLX5_LIB_CAP_4K_UAR;
+	bfregi = &context->bfregi;
+
+	/* updates req->total_num_bfregs */
+	err = calc_total_bfregs(dev, lib_uar_4k, &req, bfregi);
+	if (err)
+		goto out_ctx;
+
+	mutex_init(&bfregi->lock);
+	bfregi->lib_uar_4k = lib_uar_4k;
+	bfregi->count = kcalloc(bfregi->total_num_bfregs, sizeof(*bfregi->count),
+				GFP_KERNEL);
+	if (!bfregi->count) {
+		err = -ENOMEM;
+		goto out_ctx;
+	}
+
+	bfregi->sys_pages = kcalloc(bfregi->num_sys_pages,
+				    sizeof(*bfregi->sys_pages),
+				    GFP_KERNEL);
+	if (!bfregi->sys_pages) {
+		err = -ENOMEM;
+		goto out_count;
+	}
+
+	err = allocate_uars(dev, context);
+	if (err)
+		goto out_sys_pages;
+
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
+	context->ibucontext.invalidate_range = &mlx5_ib_invalidate_range;
+#endif
+
+	err = mlx5_ib_alloc_transport_domain(dev, &context->tdn);
+	if (err)
+		goto out_uars;
+
+	if (req.flags & MLX5_IB_ALLOC_UCTX_DEVX) {
+		/* Block DEVX on Infiniband as of SELinux */
+		if (mlx5_ib_port_link_layer(ibdev, 1) != IB_LINK_LAYER_ETHERNET) {
+			err = -EPERM;
+			goto out_td;
+		}
+
+		err = mlx5_ib_devx_create(dev, context);
+		if (err)
+			goto out_td;
+	}
+
+	if (MLX5_CAP_GEN(dev->mdev, dump_fill_mkey)) {
+		err = mlx5_cmd_dump_fill_mkey(dev->mdev, &dump_fill_mkey);
+		if (err)
+			goto out_mdev;
+	}
+
+	INIT_LIST_HEAD(&context->vma_private_list);
+	mutex_init(&context->vma_private_list_mutex);
+	INIT_LIST_HEAD(&context->db_page_list);
+	mutex_init(&context->db_page_mutex);
+
+	resp.tot_bfregs = req.total_num_bfregs;
+	resp.num_ports = dev->num_ports;
+
+	if (field_avail(typeof(resp), cqe_version, udata->outlen))
+		resp.response_length += sizeof(resp.cqe_version);
+
+	if (field_avail(typeof(resp), cmds_supp_uhw, udata->outlen)) {
+		resp.cmds_supp_uhw |= MLX5_USER_CMDS_SUPP_UHW_QUERY_DEVICE |
+				      MLX5_USER_CMDS_SUPP_UHW_CREATE_AH;
+		resp.response_length += sizeof(resp.cmds_supp_uhw);
+	}
+
+	if (field_avail(typeof(resp), eth_min_inline, udata->outlen)) {
+		if (mlx5_ib_port_link_layer(ibdev, 1) == IB_LINK_LAYER_ETHERNET) {
+			mlx5_query_min_inline(dev->mdev, &resp.eth_min_inline);
+			resp.eth_min_inline++;
+		}
+		resp.response_length += sizeof(resp.eth_min_inline);
+	}
+
+	if (field_avail(typeof(resp), clock_info_versions, udata->outlen)) {
+		if (mdev->clock_info)
+			resp.clock_info_versions = BIT(MLX5_IB_CLOCK_INFO_V1);
+		resp.response_length += sizeof(resp.clock_info_versions);
+	}
+
+	/*
+	 * We don't want to expose information from the PCI bar that is located
+	 * after 4096 bytes, so if the arch only supports larger pages, let's
+	 * pretend we don't support reading the HCA's core clock. This is also
+	 * forced by mmap function.
+	 */
+	if (field_avail(typeof(resp), hca_core_clock_offset, udata->outlen)) {
+		if (PAGE_SIZE <= 4096) {
+			resp.comp_mask |=
+				MLX5_IB_ALLOC_UCONTEXT_RESP_MASK_CORE_CLOCK_OFFSET;
+			resp.hca_core_clock_offset =
+				offsetof(struct mlx5_init_seg, internal_timer_h) % PAGE_SIZE;
+		}
+		resp.response_length += sizeof(resp.hca_core_clock_offset);
+	}
+
+	if (field_avail(typeof(resp), log_uar_size, udata->outlen))
+		resp.response_length += sizeof(resp.log_uar_size);
+
+	if (field_avail(typeof(resp), num_uars_per_page, udata->outlen))
+		resp.response_length += sizeof(resp.num_uars_per_page);
+
+	if (field_avail(typeof(resp), num_dyn_bfregs, udata->outlen)) {
+		resp.num_dyn_bfregs = bfregi->num_dyn_bfregs;
+		resp.response_length += sizeof(resp.num_dyn_bfregs);
+	}
+
+	if (field_avail(typeof(resp), dump_fill_mkey, udata->outlen)) {
+		if (MLX5_CAP_GEN(dev->mdev, dump_fill_mkey)) {
+			resp.dump_fill_mkey = dump_fill_mkey;
+			resp.comp_mask |=
+				MLX5_IB_ALLOC_UCONTEXT_RESP_MASK_DUMP_FILL_MKEY;
+		}
+		resp.response_length += sizeof(resp.dump_fill_mkey);
+	}
+
+	err = ib_copy_to_udata(udata, &resp, resp.response_length);
+	if (err)
+		goto out_mdev;
+
+	bfregi->ver = ver;
+	bfregi->num_low_latency_bfregs = req.num_low_latency_bfregs;
+	context->cqe_version = resp.cqe_version;
+	context->lib_caps = req.lib_caps;
+	print_lib_caps(dev, context->lib_caps);
+
+	if (mlx5_lag_is_active(dev->mdev)) {
+		u8 port = mlx5_core_native_port_num(dev->mdev);
+
+		atomic_set(&context->tx_port_affinity,
+			   atomic_add_return(
+				   1, &dev->roce[port].tx_port_affinity));
+	}
+
+	return &context->ibucontext;
+
+out_mdev:
+	if (req.flags & MLX5_IB_ALLOC_UCTX_DEVX)
+		mlx5_ib_devx_destroy(dev, context);
+out_td:
+	mlx5_ib_dealloc_transport_domain(dev, context->tdn);
+
+out_uars:
+	deallocate_uars(dev, context);
+
+out_sys_pages:
+	kfree(bfregi->sys_pages);
+
+out_count:
+	kfree(bfregi->count);
+
+out_ctx:
+	kfree(context);
+
+	return ERR_PTR(err);
+}
+
+static int mlx5_ib_dealloc_ucontext(struct ib_ucontext *ibcontext)
+{
+	struct mlx5_ib_ucontext *context = to_mucontext(ibcontext);
+	struct mlx5_ib_dev *dev = to_mdev(ibcontext->device);
+	struct mlx5_bfreg_info *bfregi;
+
+	if (context->devx_uid)
+		mlx5_ib_devx_destroy(dev, context);
+
+	bfregi = &context->bfregi;
+	mlx5_ib_dealloc_transport_domain(dev, context->tdn);
+
+	deallocate_uars(dev, context);
+	kfree(bfregi->sys_pages);
+	kfree(bfregi->count);
+	kfree(context);
+
+	return 0;
+}
+
+static phys_addr_t uar_index2pfn(struct mlx5_ib_dev *dev,
+				 int uar_idx)
+{
+	int fw_uars_per_page;
+
+	fw_uars_per_page = MLX5_CAP_GEN(dev->mdev, uar_4k) ? MLX5_UARS_IN_PAGE : 1;
+
+	return (pci_resource_start(dev->mdev->pdev, 0) >> PAGE_SHIFT) + uar_idx / fw_uars_per_page;
+}
+
+static int get_command(unsigned long offset)
+{
+	return (offset >> MLX5_IB_MMAP_CMD_SHIFT) & MLX5_IB_MMAP_CMD_MASK;
+}
+
+static int get_arg(unsigned long offset)
+{
+	return offset & ((1 << MLX5_IB_MMAP_CMD_SHIFT) - 1);
+}
+
+static int get_index(unsigned long offset)
+{
+	return get_arg(offset);
+}
+
+/* Index resides in an extra byte to enable larger values than 255 */
+static int get_extended_index(unsigned long offset)
+{
+	return get_arg(offset) | ((offset >> 16) & 0xff) << 8;
+}
+
+static void  mlx5_ib_vma_open(struct vm_area_struct *area)
+{
+	/* vma_open is called when a new VMA is created on top of our VMA.  This
+	 * is done through either mremap flow or split_vma (usually due to
+	 * mlock, madvise, munmap, etc.) We do not support a clone of the VMA,
+	 * as this VMA is strongly hardware related.  Therefore we set the
+	 * vm_ops of the newly created/cloned VMA to NULL, to prevent it from
+	 * calling us again and trying to do incorrect actions.  We assume that
+	 * the original VMA size is exactly a single page, and therefore all
+	 * "splitting" operation will not happen to it.
+	 */
+	area->vm_ops = NULL;
+}
+
+static void  mlx5_ib_vma_close(struct vm_area_struct *area)
+{
+	struct mlx5_ib_vma_private_data *mlx5_ib_vma_priv_data;
+
+	/* It's guaranteed that all VMAs opened on a FD are closed before the
+	 * file itself is closed, therefore no sync is needed with the regular
+	 * closing flow. (e.g. mlx5 ib_dealloc_ucontext)
+	 * However need a sync with accessing the vma as part of
+	 * mlx5_ib_disassociate_ucontext.
+	 * The close operation is usually called under mm->mmap_sem except when
+	 * process is exiting.
+	 * The exiting case is handled explicitly as part of
+	 * mlx5_ib_disassociate_ucontext.
+	 */
+	mlx5_ib_vma_priv_data = (struct mlx5_ib_vma_private_data *)area->vm_private_data;
+
+	/* setting the vma context pointer to null in the mlx5_ib driver's
+	 * private data, to protect a race condition in
+	 * mlx5_ib_disassociate_ucontext().
+	 */
+	mlx5_ib_vma_priv_data->vma = NULL;
+	mutex_lock(mlx5_ib_vma_priv_data->vma_private_list_mutex);
+	list_del(&mlx5_ib_vma_priv_data->list);
+	mutex_unlock(mlx5_ib_vma_priv_data->vma_private_list_mutex);
+	kfree(mlx5_ib_vma_priv_data);
+}
+
+static const struct vm_operations_struct mlx5_ib_vm_ops = {
+	.open = mlx5_ib_vma_open,
+	.close = mlx5_ib_vma_close
+};
+
+static int mlx5_ib_set_vma_data(struct vm_area_struct *vma,
+				struct mlx5_ib_ucontext *ctx)
+{
+	struct mlx5_ib_vma_private_data *vma_prv;
+	struct list_head *vma_head = &ctx->vma_private_list;
+
+	vma_prv = kzalloc(sizeof(*vma_prv), GFP_KERNEL);
+	if (!vma_prv)
+		return -ENOMEM;
+
+	vma_prv->vma = vma;
+	vma_prv->vma_private_list_mutex = &ctx->vma_private_list_mutex;
+	vma->vm_private_data = vma_prv;
+	vma->vm_ops =  &mlx5_ib_vm_ops;
+
+	mutex_lock(&ctx->vma_private_list_mutex);
+	list_add(&vma_prv->list, vma_head);
+	mutex_unlock(&ctx->vma_private_list_mutex);
+
+	return 0;
+}
+
+static void mlx5_ib_disassociate_ucontext(struct ib_ucontext *ibcontext)
+{
+	struct vm_area_struct *vma;
+	struct mlx5_ib_vma_private_data *vma_private, *n;
+	struct mlx5_ib_ucontext *context = to_mucontext(ibcontext);
+
+	mutex_lock(&context->vma_private_list_mutex);
+	list_for_each_entry_safe(vma_private, n, &context->vma_private_list,
+				 list) {
+		vma = vma_private->vma;
+		zap_vma_ptes(vma, vma->vm_start, PAGE_SIZE);
+		/* context going to be destroyed, should
+		 * not access ops any more.
+		 */
+		vma->vm_flags &= ~(VM_SHARED | VM_MAYSHARE);
+		vma->vm_ops = NULL;
+		list_del(&vma_private->list);
+		kfree(vma_private);
+	}
+	mutex_unlock(&context->vma_private_list_mutex);
+}
+
+static inline char *mmap_cmd2str(enum mlx5_ib_mmap_cmd cmd)
+{
+	switch (cmd) {
+	case MLX5_IB_MMAP_WC_PAGE:
+		return "WC";
+	case MLX5_IB_MMAP_REGULAR_PAGE:
+		return "best effort WC";
+	case MLX5_IB_MMAP_NC_PAGE:
+		return "NC";
+	case MLX5_IB_MMAP_DEVICE_MEM:
+		return "Device Memory";
+	default:
+		return NULL;
+	}
+}
+
+static int mlx5_ib_mmap_clock_info_page(struct mlx5_ib_dev *dev,
+					struct vm_area_struct *vma,
+					struct mlx5_ib_ucontext *context)
+{
+	phys_addr_t pfn;
+	int err;
+
+	if (vma->vm_end - vma->vm_start != PAGE_SIZE)
+		return -EINVAL;
+
+	if (get_index(vma->vm_pgoff) != MLX5_IB_CLOCK_INFO_V1)
+		return -EOPNOTSUPP;
+
+	if (vma->vm_flags & VM_WRITE)
+		return -EPERM;
+	vma->vm_flags &= ~VM_MAYWRITE;
+
+	if (!dev->mdev->clock_info_page)
+		return -EOPNOTSUPP;
+
+	pfn = page_to_pfn(dev->mdev->clock_info_page);
+	err = remap_pfn_range(vma, vma->vm_start, pfn, PAGE_SIZE,
+			      vma->vm_page_prot);
+	if (err)
+		return err;
+
+	return mlx5_ib_set_vma_data(vma, context);
+}
+
+static int uar_mmap(struct mlx5_ib_dev *dev, enum mlx5_ib_mmap_cmd cmd,
+		    struct vm_area_struct *vma,
+		    struct mlx5_ib_ucontext *context)
+{
+	struct mlx5_bfreg_info *bfregi = &context->bfregi;
+	int err;
+	unsigned long idx;
+	phys_addr_t pfn;
+	pgprot_t prot;
+	u32 bfreg_dyn_idx = 0;
+	u32 uar_index;
+	int dyn_uar = (cmd == MLX5_IB_MMAP_ALLOC_WC);
+	int max_valid_idx = dyn_uar ? bfregi->num_sys_pages :
+				bfregi->num_static_sys_pages;
+
+	if (vma->vm_end - vma->vm_start != PAGE_SIZE)
+		return -EINVAL;
+
+	if (dyn_uar)
+		idx = get_extended_index(vma->vm_pgoff) + bfregi->num_static_sys_pages;
+	else
+		idx = get_index(vma->vm_pgoff);
+
+	if (idx >= max_valid_idx) {
+		mlx5_ib_warn(dev, "invalid uar index %lu, max=%d\n",
+			     idx, max_valid_idx);
+		return -EINVAL;
+	}
+
+	switch (cmd) {
+	case MLX5_IB_MMAP_WC_PAGE:
+	case MLX5_IB_MMAP_ALLOC_WC:
+/* Some architectures don't support WC memory */
+#if defined(CONFIG_X86)
+		if (!pat_enabled())
+			return -EPERM;
+#elif !(defined(CONFIG_PPC) || (defined(CONFIG_ARM) && defined(CONFIG_MMU)))
+			return -EPERM;
+#endif
+	/* fall through */
+	case MLX5_IB_MMAP_REGULAR_PAGE:
+		/* For MLX5_IB_MMAP_REGULAR_PAGE do the best effort to get WC */
+		prot = pgprot_writecombine(vma->vm_page_prot);
+		break;
+	case MLX5_IB_MMAP_NC_PAGE:
+		prot = pgprot_noncached(vma->vm_page_prot);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	if (dyn_uar) {
+		int uars_per_page;
+
+		uars_per_page = get_uars_per_sys_page(dev, bfregi->lib_uar_4k);
+		bfreg_dyn_idx = idx * (uars_per_page * MLX5_NON_FP_BFREGS_PER_UAR);
+		if (bfreg_dyn_idx >= bfregi->total_num_bfregs) {
+			mlx5_ib_warn(dev, "invalid bfreg_dyn_idx %u, max=%u\n",
+				     bfreg_dyn_idx, bfregi->total_num_bfregs);
+			return -EINVAL;
+		}
+
+		mutex_lock(&bfregi->lock);
+		/* Fail if uar already allocated, first bfreg index of each
+		 * page holds its count.
+		 */
+		if (bfregi->count[bfreg_dyn_idx]) {
+			mlx5_ib_warn(dev, "wrong offset, idx %lu is busy, bfregn=%u\n", idx, bfreg_dyn_idx);
+			mutex_unlock(&bfregi->lock);
+			return -EINVAL;
+		}
+
+		bfregi->count[bfreg_dyn_idx]++;
+		mutex_unlock(&bfregi->lock);
+
+		err = mlx5_cmd_alloc_uar(dev->mdev, &uar_index);
+		if (err) {
+			mlx5_ib_warn(dev, "UAR alloc failed\n");
+			goto free_bfreg;
+		}
+	} else {
+		uar_index = bfregi->sys_pages[idx];
+	}
+
+	pfn = uar_index2pfn(dev, uar_index);
+	mlx5_ib_dbg(dev, "uar idx 0x%lx, pfn %pa\n", idx, &pfn);
+
+	vma->vm_page_prot = prot;
+	err = io_remap_pfn_range(vma, vma->vm_start, pfn,
+				 PAGE_SIZE, vma->vm_page_prot);
+	if (err) {
+		mlx5_ib_err(dev,
+			    "io_remap_pfn_range failed with error=%d, mmap_cmd=%s\n",
+			    err, mmap_cmd2str(cmd));
+		err = -EAGAIN;
+		goto err;
+	}
+
+	err = mlx5_ib_set_vma_data(vma, context);
+	if (err)
+		goto err;
+
+	if (dyn_uar)
+		bfregi->sys_pages[idx] = uar_index;
+	return 0;
+
+err:
+	if (!dyn_uar)
+		return err;
+
+	mlx5_cmd_free_uar(dev->mdev, idx);
+
+free_bfreg:
+	mlx5_ib_free_bfreg(dev, bfregi, bfreg_dyn_idx);
+
+	return err;
+}
+
+static int dm_mmap(struct ib_ucontext *context, struct vm_area_struct *vma)
+{
+	struct mlx5_ib_ucontext *mctx = to_mucontext(context);
+	struct mlx5_ib_dev *dev = to_mdev(context->device);
+	u16 page_idx = get_extended_index(vma->vm_pgoff);
+	size_t map_size = vma->vm_end - vma->vm_start;
+	u32 npages = map_size >> PAGE_SHIFT;
+	phys_addr_t pfn;
+	pgprot_t prot;
+
+	if (find_next_zero_bit(mctx->dm_pages, page_idx + npages, page_idx) !=
+	    page_idx + npages)
+		return -EINVAL;
+
+	pfn = ((pci_resource_start(dev->mdev->pdev, 0) +
+	      MLX5_CAP64_DEV_MEM(dev->mdev, memic_bar_start_addr)) >>
+	      PAGE_SHIFT) +
+	      page_idx;
+	prot = pgprot_writecombine(vma->vm_page_prot);
+	vma->vm_page_prot = prot;
+
+	if (io_remap_pfn_range(vma, vma->vm_start, pfn, map_size,
+			       vma->vm_page_prot))
+		return -EAGAIN;
+
+	return mlx5_ib_set_vma_data(vma, mctx);
+}
+
+static int mlx5_ib_mmap(struct ib_ucontext *ibcontext, struct vm_area_struct *vma)
+{
+	struct mlx5_ib_ucontext *context = to_mucontext(ibcontext);
+	struct mlx5_ib_dev *dev = to_mdev(ibcontext->device);
+	unsigned long command;
+	phys_addr_t pfn;
+
+	command = get_command(vma->vm_pgoff);
+	switch (command) {
+	case MLX5_IB_MMAP_WC_PAGE:
+	case MLX5_IB_MMAP_NC_PAGE:
+	case MLX5_IB_MMAP_REGULAR_PAGE:
+	case MLX5_IB_MMAP_ALLOC_WC:
+		return uar_mmap(dev, command, vma, context);
+
+	case MLX5_IB_MMAP_GET_CONTIGUOUS_PAGES:
+		return -ENOSYS;
+
+	case MLX5_IB_MMAP_CORE_CLOCK:
+		if (vma->vm_end - vma->vm_start != PAGE_SIZE)
+			return -EINVAL;
+
+		if (vma->vm_flags & VM_WRITE)
+			return -EPERM;
+		vma->vm_flags &= ~VM_MAYWRITE;
+
+		/* Don't expose to user-space information it shouldn't have */
+		if (PAGE_SIZE > 4096)
+			return -EOPNOTSUPP;
+
+		vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
+		pfn = (dev->mdev->iseg_base +
+		       offsetof(struct mlx5_init_seg, internal_timer_h)) >>
+			PAGE_SHIFT;
+		if (io_remap_pfn_range(vma, vma->vm_start, pfn,
+				       PAGE_SIZE, vma->vm_page_prot))
+			return -EAGAIN;
+		break;
+	case MLX5_IB_MMAP_CLOCK_INFO:
+		return mlx5_ib_mmap_clock_info_page(dev, vma, context);
+
+	case MLX5_IB_MMAP_DEVICE_MEM:
+		return dm_mmap(ibcontext, vma);
+
+	default:
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+struct ib_dm *mlx5_ib_alloc_dm(struct ib_device *ibdev,
+			       struct ib_ucontext *context,
+			       struct ib_dm_alloc_attr *attr,
+			       struct uverbs_attr_bundle *attrs)
+{
+	u64 act_size = roundup(attr->length, MLX5_MEMIC_BASE_SIZE);
+	struct mlx5_memic *memic = &to_mdev(ibdev)->memic;
+	phys_addr_t memic_addr;
+	struct mlx5_ib_dm *dm;
+	u64 start_offset;
+	u32 page_idx;
+	int err;
+
+	dm = kzalloc(sizeof(*dm), GFP_KERNEL);
+	if (!dm)
+		return ERR_PTR(-ENOMEM);
+
+	mlx5_ib_dbg(to_mdev(ibdev), "alloc_memic req: user_length=0x%llx act_length=0x%llx log_alignment=%d\n",
+		    attr->length, act_size, attr->alignment);
+
+	err = mlx5_cmd_alloc_memic(memic, &memic_addr,
+				   act_size, attr->alignment);
+	if (err)
+		goto err_free;
+
+	start_offset = memic_addr & ~PAGE_MASK;
+	page_idx = (memic_addr - pci_resource_start(memic->dev->pdev, 0) -
+		    MLX5_CAP64_DEV_MEM(memic->dev, memic_bar_start_addr)) >>
+		    PAGE_SHIFT;
+
+	err = uverbs_copy_to(attrs,
+			     MLX5_IB_ATTR_ALLOC_DM_RESP_START_OFFSET,
+			     &start_offset, sizeof(start_offset));
+	if (err)
+		goto err_dealloc;
+
+	err = uverbs_copy_to(attrs,
+			     MLX5_IB_ATTR_ALLOC_DM_RESP_PAGE_INDEX,
+			     &page_idx, sizeof(page_idx));
+	if (err)
+		goto err_dealloc;
+
+	bitmap_set(to_mucontext(context)->dm_pages, page_idx,
+		   DIV_ROUND_UP(act_size, PAGE_SIZE));
+
+	dm->dev_addr = memic_addr;
+
+	return &dm->ibdm;
+
+err_dealloc:
+	mlx5_cmd_dealloc_memic(memic, memic_addr,
+			       act_size);
+err_free:
+	kfree(dm);
+	return ERR_PTR(err);
+}
+
+int mlx5_ib_dealloc_dm(struct ib_dm *ibdm)
+{
+	struct mlx5_memic *memic = &to_mdev(ibdm->device)->memic;
+	struct mlx5_ib_dm *dm = to_mdm(ibdm);
+	u64 act_size = roundup(dm->ibdm.length, MLX5_MEMIC_BASE_SIZE);
+	u32 page_idx;
+	int ret;
+
+	ret = mlx5_cmd_dealloc_memic(memic, dm->dev_addr, act_size);
+	if (ret)
+		return ret;
+
+	page_idx = (dm->dev_addr - pci_resource_start(memic->dev->pdev, 0) -
+		    MLX5_CAP64_DEV_MEM(memic->dev, memic_bar_start_addr)) >>
+		    PAGE_SHIFT;
+	bitmap_clear(to_mucontext(ibdm->uobject->context)->dm_pages,
+		     page_idx,
+		     DIV_ROUND_UP(act_size, PAGE_SIZE));
+
+	kfree(dm);
+
+	return 0;
+}
+
+static struct ib_pd *mlx5_ib_alloc_pd(struct ib_device *ibdev,
+				      struct ib_ucontext *context,
+				      struct ib_udata *udata)
+{
+	struct mlx5_ib_alloc_pd_resp resp;
+	struct mlx5_ib_pd *pd;
+	int err;
+
+	pd = kmalloc(sizeof(*pd), GFP_KERNEL);
+	if (!pd)
+		return ERR_PTR(-ENOMEM);
+
+	err = mlx5_core_alloc_pd(to_mdev(ibdev)->mdev, &pd->pdn);
+	if (err) {
+		kfree(pd);
+		return ERR_PTR(err);
+	}
+
+	if (context) {
+		resp.pdn = pd->pdn;
+		if (ib_copy_to_udata(udata, &resp, sizeof(resp))) {
+			mlx5_core_dealloc_pd(to_mdev(ibdev)->mdev, pd->pdn);
+			kfree(pd);
+			return ERR_PTR(-EFAULT);
+		}
+	}
+
+	return &pd->ibpd;
+}
+
+static int mlx5_ib_dealloc_pd(struct ib_pd *pd)
+{
+	struct mlx5_ib_dev *mdev = to_mdev(pd->device);
+	struct mlx5_ib_pd *mpd = to_mpd(pd);
+
+	mlx5_core_dealloc_pd(mdev->mdev, mpd->pdn);
+	kfree(mpd);
+
+	return 0;
+}
+
+enum {
+	MATCH_CRITERIA_ENABLE_OUTER_BIT,
+	MATCH_CRITERIA_ENABLE_MISC_BIT,
+	MATCH_CRITERIA_ENABLE_INNER_BIT,
+	MATCH_CRITERIA_ENABLE_MISC2_BIT
+};
+
+#define HEADER_IS_ZERO(match_criteria, headers)			           \
+	!(memchr_inv(MLX5_ADDR_OF(fte_match_param, match_criteria, headers), \
+		    0, MLX5_FLD_SZ_BYTES(fte_match_param, headers)))       \
+
+static u8 get_match_criteria_enable(u32 *match_criteria)
+{
+	u8 match_criteria_enable;
+
+	match_criteria_enable =
+		(!HEADER_IS_ZERO(match_criteria, outer_headers)) <<
+		MATCH_CRITERIA_ENABLE_OUTER_BIT;
+	match_criteria_enable |=
+		(!HEADER_IS_ZERO(match_criteria, misc_parameters)) <<
+		MATCH_CRITERIA_ENABLE_MISC_BIT;
+	match_criteria_enable |=
+		(!HEADER_IS_ZERO(match_criteria, inner_headers)) <<
+		MATCH_CRITERIA_ENABLE_INNER_BIT;
+	match_criteria_enable |=
+		(!HEADER_IS_ZERO(match_criteria, misc_parameters_2)) <<
+		MATCH_CRITERIA_ENABLE_MISC2_BIT;
+
+	return match_criteria_enable;
+}
+
+static int set_proto(void *outer_c, void *outer_v, u8 mask, u8 val)
+{
+	u8 entry_mask;
+	u8 entry_val;
+	int err = 0;
+
+	if (!mask)
+		goto out;
+
+	entry_mask = MLX5_GET(fte_match_set_lyr_2_4, outer_c,
+			      ip_protocol);
+	entry_val = MLX5_GET(fte_match_set_lyr_2_4, outer_v,
+			     ip_protocol);
+	if (!entry_mask) {
+		MLX5_SET(fte_match_set_lyr_2_4, outer_c, ip_protocol, mask);
+		MLX5_SET(fte_match_set_lyr_2_4, outer_v, ip_protocol, val);
+		goto out;
+	}
+	/* Don't override existing ip protocol */
+	if (mask != entry_mask || val != entry_val)
+		err = -EINVAL;
+out:
+	return err;
+}
+
+static void set_flow_label(void *misc_c, void *misc_v, u32 mask, u32 val,
+			   bool inner)
+{
+	if (inner) {
+		MLX5_SET(fte_match_set_misc,
+			 misc_c, inner_ipv6_flow_label, mask);
+		MLX5_SET(fte_match_set_misc,
+			 misc_v, inner_ipv6_flow_label, val);
+	} else {
+		MLX5_SET(fte_match_set_misc,
+			 misc_c, outer_ipv6_flow_label, mask);
+		MLX5_SET(fte_match_set_misc,
+			 misc_v, outer_ipv6_flow_label, val);
+	}
+}
+
+static void set_tos(void *outer_c, void *outer_v, u8 mask, u8 val)
+{
+	MLX5_SET(fte_match_set_lyr_2_4, outer_c, ip_ecn, mask);
+	MLX5_SET(fte_match_set_lyr_2_4, outer_v, ip_ecn, val);
+	MLX5_SET(fte_match_set_lyr_2_4, outer_c, ip_dscp, mask >> 2);
+	MLX5_SET(fte_match_set_lyr_2_4, outer_v, ip_dscp, val >> 2);
+}
+
+static int check_mpls_supp_fields(u32 field_support, const __be32 *set_mask)
+{
+	if (MLX5_GET(fte_match_mpls, set_mask, mpls_label) &&
+	    !(field_support & MLX5_FIELD_SUPPORT_MPLS_LABEL))
+		return -EOPNOTSUPP;
+
+	if (MLX5_GET(fte_match_mpls, set_mask, mpls_exp) &&
+	    !(field_support & MLX5_FIELD_SUPPORT_MPLS_EXP))
+		return -EOPNOTSUPP;
+
+	if (MLX5_GET(fte_match_mpls, set_mask, mpls_s_bos) &&
+	    !(field_support & MLX5_FIELD_SUPPORT_MPLS_S_BOS))
+		return -EOPNOTSUPP;
+
+	if (MLX5_GET(fte_match_mpls, set_mask, mpls_ttl) &&
+	    !(field_support & MLX5_FIELD_SUPPORT_MPLS_TTL))
+		return -EOPNOTSUPP;
+
+	return 0;
+}
+
+#define LAST_ETH_FIELD vlan_tag
+#define LAST_IB_FIELD sl
+#define LAST_IPV4_FIELD tos
+#define LAST_IPV6_FIELD traffic_class
+#define LAST_TCP_UDP_FIELD src_port
+#define LAST_TUNNEL_FIELD tunnel_id
+#define LAST_FLOW_TAG_FIELD tag_id
+#define LAST_DROP_FIELD size
+#define LAST_COUNTERS_FIELD counters
+
+/* Field is the last supported field */
+#define FIELDS_NOT_SUPPORTED(filter, field)\
+	memchr_inv((void *)&filter.field  +\
+		   sizeof(filter.field), 0,\
+		   sizeof(filter) -\
+		   offsetof(typeof(filter), field) -\
+		   sizeof(filter.field))
+
+static int parse_flow_flow_action(const union ib_flow_spec *ib_spec,
+				  const struct ib_flow_attr *flow_attr,
+				  struct mlx5_flow_act *action)
+{
+	struct mlx5_ib_flow_action *maction = to_mflow_act(ib_spec->action.act);
+
+	switch (maction->ib_action.type) {
+	case IB_FLOW_ACTION_ESP:
+		/* Currently only AES_GCM keymat is supported by the driver */
+		action->esp_id = (uintptr_t)maction->esp_aes_gcm.ctx;
+		action->action |= flow_attr->flags & IB_FLOW_ATTR_FLAGS_EGRESS ?
+			MLX5_FLOW_CONTEXT_ACTION_ENCRYPT :
+			MLX5_FLOW_CONTEXT_ACTION_DECRYPT;
+		return 0;
+	default:
+		return -EOPNOTSUPP;
+	}
+}
+
+static int parse_flow_attr(struct mlx5_core_dev *mdev, u32 *match_c,
+			   u32 *match_v, const union ib_flow_spec *ib_spec,
+			   const struct ib_flow_attr *flow_attr,
+			   struct mlx5_flow_act *action, u32 prev_type)
+{
+	void *misc_params_c = MLX5_ADDR_OF(fte_match_param, match_c,
+					   misc_parameters);
+	void *misc_params_v = MLX5_ADDR_OF(fte_match_param, match_v,
+					   misc_parameters);
+	void *misc_params2_c = MLX5_ADDR_OF(fte_match_param, match_c,
+					    misc_parameters_2);
+	void *misc_params2_v = MLX5_ADDR_OF(fte_match_param, match_v,
+					    misc_parameters_2);
+	void *headers_c;
+	void *headers_v;
+	int match_ipv;
+	int ret;
+
+	if (ib_spec->type & IB_FLOW_SPEC_INNER) {
+		headers_c = MLX5_ADDR_OF(fte_match_param, match_c,
+					 inner_headers);
+		headers_v = MLX5_ADDR_OF(fte_match_param, match_v,
+					 inner_headers);
+		match_ipv = MLX5_CAP_FLOWTABLE_NIC_RX(mdev,
+					ft_field_support.inner_ip_version);
+	} else {
+		headers_c = MLX5_ADDR_OF(fte_match_param, match_c,
+					 outer_headers);
+		headers_v = MLX5_ADDR_OF(fte_match_param, match_v,
+					 outer_headers);
+		match_ipv = MLX5_CAP_FLOWTABLE_NIC_RX(mdev,
+					ft_field_support.outer_ip_version);
+	}
+
+	switch (ib_spec->type & ~IB_FLOW_SPEC_INNER) {
+	case IB_FLOW_SPEC_ETH:
+		if (FIELDS_NOT_SUPPORTED(ib_spec->eth.mask, LAST_ETH_FIELD))
+			return -EOPNOTSUPP;
+
+		ether_addr_copy(MLX5_ADDR_OF(fte_match_set_lyr_2_4, headers_c,
+					     dmac_47_16),
+				ib_spec->eth.mask.dst_mac);
+		ether_addr_copy(MLX5_ADDR_OF(fte_match_set_lyr_2_4, headers_v,
+					     dmac_47_16),
+				ib_spec->eth.val.dst_mac);
+
+		ether_addr_copy(MLX5_ADDR_OF(fte_match_set_lyr_2_4, headers_c,
+					     smac_47_16),
+				ib_spec->eth.mask.src_mac);
+		ether_addr_copy(MLX5_ADDR_OF(fte_match_set_lyr_2_4, headers_v,
+					     smac_47_16),
+				ib_spec->eth.val.src_mac);
+
+		if (ib_spec->eth.mask.vlan_tag) {
+			MLX5_SET(fte_match_set_lyr_2_4, headers_c,
+				 cvlan_tag, 1);
+			MLX5_SET(fte_match_set_lyr_2_4, headers_v,
+				 cvlan_tag, 1);
+
+			MLX5_SET(fte_match_set_lyr_2_4, headers_c,
+				 first_vid, ntohs(ib_spec->eth.mask.vlan_tag));
+			MLX5_SET(fte_match_set_lyr_2_4, headers_v,
+				 first_vid, ntohs(ib_spec->eth.val.vlan_tag));
+
+			MLX5_SET(fte_match_set_lyr_2_4, headers_c,
+				 first_cfi,
+				 ntohs(ib_spec->eth.mask.vlan_tag) >> 12);
+			MLX5_SET(fte_match_set_lyr_2_4, headers_v,
+				 first_cfi,
+				 ntohs(ib_spec->eth.val.vlan_tag) >> 12);
+
+			MLX5_SET(fte_match_set_lyr_2_4, headers_c,
+				 first_prio,
+				 ntohs(ib_spec->eth.mask.vlan_tag) >> 13);
+			MLX5_SET(fte_match_set_lyr_2_4, headers_v,
+				 first_prio,
+				 ntohs(ib_spec->eth.val.vlan_tag) >> 13);
+		}
+		MLX5_SET(fte_match_set_lyr_2_4, headers_c,
+			 ethertype, ntohs(ib_spec->eth.mask.ether_type));
+		MLX5_SET(fte_match_set_lyr_2_4, headers_v,
+			 ethertype, ntohs(ib_spec->eth.val.ether_type));
+		break;
+	case IB_FLOW_SPEC_IPV4:
+		if (FIELDS_NOT_SUPPORTED(ib_spec->ipv4.mask, LAST_IPV4_FIELD))
+			return -EOPNOTSUPP;
+
+		if (match_ipv) {
+			MLX5_SET(fte_match_set_lyr_2_4, headers_c,
+				 ip_version, 0xf);
+			MLX5_SET(fte_match_set_lyr_2_4, headers_v,
+				 ip_version, MLX5_FS_IPV4_VERSION);
+		} else {
+			MLX5_SET(fte_match_set_lyr_2_4, headers_c,
+				 ethertype, 0xffff);
+			MLX5_SET(fte_match_set_lyr_2_4, headers_v,
+				 ethertype, ETH_P_IP);
+		}
+
+		memcpy(MLX5_ADDR_OF(fte_match_set_lyr_2_4, headers_c,
+				    src_ipv4_src_ipv6.ipv4_layout.ipv4),
+		       &ib_spec->ipv4.mask.src_ip,
+		       sizeof(ib_spec->ipv4.mask.src_ip));
+		memcpy(MLX5_ADDR_OF(fte_match_set_lyr_2_4, headers_v,
+				    src_ipv4_src_ipv6.ipv4_layout.ipv4),
+		       &ib_spec->ipv4.val.src_ip,
+		       sizeof(ib_spec->ipv4.val.src_ip));
+		memcpy(MLX5_ADDR_OF(fte_match_set_lyr_2_4, headers_c,
+				    dst_ipv4_dst_ipv6.ipv4_layout.ipv4),
+		       &ib_spec->ipv4.mask.dst_ip,
+		       sizeof(ib_spec->ipv4.mask.dst_ip));
+		memcpy(MLX5_ADDR_OF(fte_match_set_lyr_2_4, headers_v,
+				    dst_ipv4_dst_ipv6.ipv4_layout.ipv4),
+		       &ib_spec->ipv4.val.dst_ip,
+		       sizeof(ib_spec->ipv4.val.dst_ip));
+
+		set_tos(headers_c, headers_v,
+			ib_spec->ipv4.mask.tos, ib_spec->ipv4.val.tos);
+
+		if (set_proto(headers_c, headers_v,
+			      ib_spec->ipv4.mask.proto,
+			      ib_spec->ipv4.val.proto))
+			return -EINVAL;
+		break;
+	case IB_FLOW_SPEC_IPV6:
+		if (FIELDS_NOT_SUPPORTED(ib_spec->ipv6.mask, LAST_IPV6_FIELD))
+			return -EOPNOTSUPP;
+
+		if (match_ipv) {
+			MLX5_SET(fte_match_set_lyr_2_4, headers_c,
+				 ip_version, 0xf);
+			MLX5_SET(fte_match_set_lyr_2_4, headers_v,
+				 ip_version, MLX5_FS_IPV6_VERSION);
+		} else {
+			MLX5_SET(fte_match_set_lyr_2_4, headers_c,
+				 ethertype, 0xffff);
+			MLX5_SET(fte_match_set_lyr_2_4, headers_v,
+				 ethertype, ETH_P_IPV6);
+		}
+
+		memcpy(MLX5_ADDR_OF(fte_match_set_lyr_2_4, headers_c,
+				    src_ipv4_src_ipv6.ipv6_layout.ipv6),
+		       &ib_spec->ipv6.mask.src_ip,
+		       sizeof(ib_spec->ipv6.mask.src_ip));
+		memcpy(MLX5_ADDR_OF(fte_match_set_lyr_2_4, headers_v,
+				    src_ipv4_src_ipv6.ipv6_layout.ipv6),
+		       &ib_spec->ipv6.val.src_ip,
+		       sizeof(ib_spec->ipv6.val.src_ip));
+		memcpy(MLX5_ADDR_OF(fte_match_set_lyr_2_4, headers_c,
+				    dst_ipv4_dst_ipv6.ipv6_layout.ipv6),
+		       &ib_spec->ipv6.mask.dst_ip,
+		       sizeof(ib_spec->ipv6.mask.dst_ip));
+		memcpy(MLX5_ADDR_OF(fte_match_set_lyr_2_4, headers_v,
+				    dst_ipv4_dst_ipv6.ipv6_layout.ipv6),
+		       &ib_spec->ipv6.val.dst_ip,
+		       sizeof(ib_spec->ipv6.val.dst_ip));
+
+		set_tos(headers_c, headers_v,
+			ib_spec->ipv6.mask.traffic_class,
+			ib_spec->ipv6.val.traffic_class);
+
+		if (set_proto(headers_c, headers_v,
+			      ib_spec->ipv6.mask.next_hdr,
+			      ib_spec->ipv6.val.next_hdr))
+			return -EINVAL;
+
+		set_flow_label(misc_params_c, misc_params_v,
+			       ntohl(ib_spec->ipv6.mask.flow_label),
+			       ntohl(ib_spec->ipv6.val.flow_label),
+			       ib_spec->type & IB_FLOW_SPEC_INNER);
+		break;
+	case IB_FLOW_SPEC_ESP:
+		if (ib_spec->esp.mask.seq)
+			return -EOPNOTSUPP;
+
+		MLX5_SET(fte_match_set_misc, misc_params_c, outer_esp_spi,
+			 ntohl(ib_spec->esp.mask.spi));
+		MLX5_SET(fte_match_set_misc, misc_params_v, outer_esp_spi,
+			 ntohl(ib_spec->esp.val.spi));
+		break;
+	case IB_FLOW_SPEC_TCP:
+		if (FIELDS_NOT_SUPPORTED(ib_spec->tcp_udp.mask,
+					 LAST_TCP_UDP_FIELD))
+			return -EOPNOTSUPP;
+
+		if (set_proto(headers_c, headers_v, 0xff, IPPROTO_TCP))
+			return -EINVAL;
+
+		MLX5_SET(fte_match_set_lyr_2_4, headers_c, tcp_sport,
+			 ntohs(ib_spec->tcp_udp.mask.src_port));
+		MLX5_SET(fte_match_set_lyr_2_4, headers_v, tcp_sport,
+			 ntohs(ib_spec->tcp_udp.val.src_port));
+
+		MLX5_SET(fte_match_set_lyr_2_4, headers_c, tcp_dport,
+			 ntohs(ib_spec->tcp_udp.mask.dst_port));
+		MLX5_SET(fte_match_set_lyr_2_4, headers_v, tcp_dport,
+			 ntohs(ib_spec->tcp_udp.val.dst_port));
+		break;
+	case IB_FLOW_SPEC_UDP:
+		if (FIELDS_NOT_SUPPORTED(ib_spec->tcp_udp.mask,
+					 LAST_TCP_UDP_FIELD))
+			return -EOPNOTSUPP;
+
+		if (set_proto(headers_c, headers_v, 0xff, IPPROTO_UDP))
+			return -EINVAL;
+
+		MLX5_SET(fte_match_set_lyr_2_4, headers_c, udp_sport,
+			 ntohs(ib_spec->tcp_udp.mask.src_port));
+		MLX5_SET(fte_match_set_lyr_2_4, headers_v, udp_sport,
+			 ntohs(ib_spec->tcp_udp.val.src_port));
+
+		MLX5_SET(fte_match_set_lyr_2_4, headers_c, udp_dport,
+			 ntohs(ib_spec->tcp_udp.mask.dst_port));
+		MLX5_SET(fte_match_set_lyr_2_4, headers_v, udp_dport,
+			 ntohs(ib_spec->tcp_udp.val.dst_port));
+		break;
+	case IB_FLOW_SPEC_GRE:
+		if (ib_spec->gre.mask.c_ks_res0_ver)
+			return -EOPNOTSUPP;
+
+		if (set_proto(headers_c, headers_v, 0xff, IPPROTO_GRE))
+			return -EINVAL;
+
+		MLX5_SET(fte_match_set_lyr_2_4, headers_c, ip_protocol,
+			 0xff);
+		MLX5_SET(fte_match_set_lyr_2_4, headers_v, ip_protocol,
+			 IPPROTO_GRE);
+
+		MLX5_SET(fte_match_set_misc, misc_params_c, gre_protocol,
+			 ntohs(ib_spec->gre.mask.protocol));
+		MLX5_SET(fte_match_set_misc, misc_params_v, gre_protocol,
+			 ntohs(ib_spec->gre.val.protocol));
+
+		memcpy(MLX5_ADDR_OF(fte_match_set_misc, misc_params_c,
+				    gre_key_h),
+		       &ib_spec->gre.mask.key,
+		       sizeof(ib_spec->gre.mask.key));
+		memcpy(MLX5_ADDR_OF(fte_match_set_misc, misc_params_v,
+				    gre_key_h),
+		       &ib_spec->gre.val.key,
+		       sizeof(ib_spec->gre.val.key));
+		break;
+	case IB_FLOW_SPEC_MPLS:
+		switch (prev_type) {
+		case IB_FLOW_SPEC_UDP:
+			if (check_mpls_supp_fields(MLX5_CAP_FLOWTABLE_NIC_RX(mdev,
+						   ft_field_support.outer_first_mpls_over_udp),
+						   &ib_spec->mpls.mask.tag))
+				return -EOPNOTSUPP;
+
+			memcpy(MLX5_ADDR_OF(fte_match_set_misc2, misc_params2_v,
+					    outer_first_mpls_over_udp),
+			       &ib_spec->mpls.val.tag,
+			       sizeof(ib_spec->mpls.val.tag));
+			memcpy(MLX5_ADDR_OF(fte_match_set_misc2, misc_params2_c,
+					    outer_first_mpls_over_udp),
+			       &ib_spec->mpls.mask.tag,
+			       sizeof(ib_spec->mpls.mask.tag));
+			break;
+		case IB_FLOW_SPEC_GRE:
+			if (check_mpls_supp_fields(MLX5_CAP_FLOWTABLE_NIC_RX(mdev,
+						   ft_field_support.outer_first_mpls_over_gre),
+						   &ib_spec->mpls.mask.tag))
+				return -EOPNOTSUPP;
+
+			memcpy(MLX5_ADDR_OF(fte_match_set_misc2, misc_params2_v,
+					    outer_first_mpls_over_gre),
+			       &ib_spec->mpls.val.tag,
+			       sizeof(ib_spec->mpls.val.tag));
+			memcpy(MLX5_ADDR_OF(fte_match_set_misc2, misc_params2_c,
+					    outer_first_mpls_over_gre),
+			       &ib_spec->mpls.mask.tag,
+			       sizeof(ib_spec->mpls.mask.tag));
+			break;
+		default:
+			if (ib_spec->type & IB_FLOW_SPEC_INNER) {
+				if (check_mpls_supp_fields(MLX5_CAP_FLOWTABLE_NIC_RX(mdev,
+							   ft_field_support.inner_first_mpls),
+							   &ib_spec->mpls.mask.tag))
+					return -EOPNOTSUPP;
+
+				memcpy(MLX5_ADDR_OF(fte_match_set_misc2, misc_params2_v,
+						    inner_first_mpls),
+				       &ib_spec->mpls.val.tag,
+				       sizeof(ib_spec->mpls.val.tag));
+				memcpy(MLX5_ADDR_OF(fte_match_set_misc2, misc_params2_c,
+						    inner_first_mpls),
+				       &ib_spec->mpls.mask.tag,
+				       sizeof(ib_spec->mpls.mask.tag));
+			} else {
+				if (check_mpls_supp_fields(MLX5_CAP_FLOWTABLE_NIC_RX(mdev,
+							   ft_field_support.outer_first_mpls),
+							   &ib_spec->mpls.mask.tag))
+					return -EOPNOTSUPP;
+
+				memcpy(MLX5_ADDR_OF(fte_match_set_misc2, misc_params2_v,
+						    outer_first_mpls),
+				       &ib_spec->mpls.val.tag,
+				       sizeof(ib_spec->mpls.val.tag));
+				memcpy(MLX5_ADDR_OF(fte_match_set_misc2, misc_params2_c,
+						    outer_first_mpls),
+				       &ib_spec->mpls.mask.tag,
+				       sizeof(ib_spec->mpls.mask.tag));
+			}
+		}
+		break;
+	case IB_FLOW_SPEC_VXLAN_TUNNEL:
+		if (FIELDS_NOT_SUPPORTED(ib_spec->tunnel.mask,
+					 LAST_TUNNEL_FIELD))
+			return -EOPNOTSUPP;
+
+		MLX5_SET(fte_match_set_misc, misc_params_c, vxlan_vni,
+			 ntohl(ib_spec->tunnel.mask.tunnel_id));
+		MLX5_SET(fte_match_set_misc, misc_params_v, vxlan_vni,
+			 ntohl(ib_spec->tunnel.val.tunnel_id));
+		break;
+	case IB_FLOW_SPEC_ACTION_TAG:
+		if (FIELDS_NOT_SUPPORTED(ib_spec->flow_tag,
+					 LAST_FLOW_TAG_FIELD))
+			return -EOPNOTSUPP;
+		if (ib_spec->flow_tag.tag_id >= BIT(24))
+			return -EINVAL;
+
+		action->flow_tag = ib_spec->flow_tag.tag_id;
+		action->has_flow_tag = true;
+		break;
+	case IB_FLOW_SPEC_ACTION_DROP:
+		if (FIELDS_NOT_SUPPORTED(ib_spec->drop,
+					 LAST_DROP_FIELD))
+			return -EOPNOTSUPP;
+		action->action |= MLX5_FLOW_CONTEXT_ACTION_DROP;
+		break;
+	case IB_FLOW_SPEC_ACTION_HANDLE:
+		ret = parse_flow_flow_action(ib_spec, flow_attr, action);
+		if (ret)
+			return ret;
+		break;
+	case IB_FLOW_SPEC_ACTION_COUNT:
+		if (FIELDS_NOT_SUPPORTED(ib_spec->flow_count,
+					 LAST_COUNTERS_FIELD))
+			return -EOPNOTSUPP;
+
+		/* for now support only one counters spec per flow */
+		if (action->action & MLX5_FLOW_CONTEXT_ACTION_COUNT)
+			return -EINVAL;
+
+		action->counters = ib_spec->flow_count.counters;
+		action->action |= MLX5_FLOW_CONTEXT_ACTION_COUNT;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+/* If a flow could catch both multicast and unicast packets,
+ * it won't fall into the multicast flow steering table and this rule
+ * could steal other multicast packets.
+ */
+static bool flow_is_multicast_only(const struct ib_flow_attr *ib_attr)
+{
+	union ib_flow_spec *flow_spec;
+
+	if (ib_attr->type != IB_FLOW_ATTR_NORMAL ||
+	    ib_attr->num_of_specs < 1)
+		return false;
+
+	flow_spec = (union ib_flow_spec *)(ib_attr + 1);
+	if (flow_spec->type == IB_FLOW_SPEC_IPV4) {
+		struct ib_flow_spec_ipv4 *ipv4_spec;
+
+		ipv4_spec = (struct ib_flow_spec_ipv4 *)flow_spec;
+		if (ipv4_is_multicast(ipv4_spec->val.dst_ip))
+			return true;
+
+		return false;
+	}
+
+	if (flow_spec->type == IB_FLOW_SPEC_ETH) {
+		struct ib_flow_spec_eth *eth_spec;
+
+		eth_spec = (struct ib_flow_spec_eth *)flow_spec;
+		return is_multicast_ether_addr(eth_spec->mask.dst_mac) &&
+		       is_multicast_ether_addr(eth_spec->val.dst_mac);
+	}
+
+	return false;
+}
+
+enum valid_spec {
+	VALID_SPEC_INVALID,
+	VALID_SPEC_VALID,
+	VALID_SPEC_NA,
+};
+
+static enum valid_spec
+is_valid_esp_aes_gcm(struct mlx5_core_dev *mdev,
+		     const struct mlx5_flow_spec *spec,
+		     const struct mlx5_flow_act *flow_act,
+		     bool egress)
+{
+	const u32 *match_c = spec->match_criteria;
+	bool is_crypto =
+		(flow_act->action & (MLX5_FLOW_CONTEXT_ACTION_ENCRYPT |
+				     MLX5_FLOW_CONTEXT_ACTION_DECRYPT));
+	bool is_ipsec = mlx5_fs_is_ipsec_flow(match_c);
+	bool is_drop = flow_act->action & MLX5_FLOW_CONTEXT_ACTION_DROP;
+
+	/*
+	 * Currently only crypto is supported in egress, when regular egress
+	 * rules would be supported, always return VALID_SPEC_NA.
+	 */
+	if (!is_crypto)
+		return egress ? VALID_SPEC_INVALID : VALID_SPEC_NA;
+
+	return is_crypto && is_ipsec &&
+		(!egress || (!is_drop && !flow_act->has_flow_tag)) ?
+		VALID_SPEC_VALID : VALID_SPEC_INVALID;
+}
+
+static bool is_valid_spec(struct mlx5_core_dev *mdev,
+			  const struct mlx5_flow_spec *spec,
+			  const struct mlx5_flow_act *flow_act,
+			  bool egress)
+{
+	/* We curretly only support ipsec egress flow */
+	return is_valid_esp_aes_gcm(mdev, spec, flow_act, egress) != VALID_SPEC_INVALID;
+}
+
+static bool is_valid_ethertype(struct mlx5_core_dev *mdev,
+			       const struct ib_flow_attr *flow_attr,
+			       bool check_inner)
+{
+	union ib_flow_spec *ib_spec = (union ib_flow_spec *)(flow_attr + 1);
+	int match_ipv = check_inner ?
+			MLX5_CAP_FLOWTABLE_NIC_RX(mdev,
+					ft_field_support.inner_ip_version) :
+			MLX5_CAP_FLOWTABLE_NIC_RX(mdev,
+					ft_field_support.outer_ip_version);
+	int inner_bit = check_inner ? IB_FLOW_SPEC_INNER : 0;
+	bool ipv4_spec_valid, ipv6_spec_valid;
+	unsigned int ip_spec_type = 0;
+	bool has_ethertype = false;
+	unsigned int spec_index;
+	bool mask_valid = true;
+	u16 eth_type = 0;
+	bool type_valid;
+
+	/* Validate that ethertype is correct */
+	for (spec_index = 0; spec_index < flow_attr->num_of_specs; spec_index++) {
+		if ((ib_spec->type == (IB_FLOW_SPEC_ETH | inner_bit)) &&
+		    ib_spec->eth.mask.ether_type) {
+			mask_valid = (ib_spec->eth.mask.ether_type ==
+				      htons(0xffff));
+			has_ethertype = true;
+			eth_type = ntohs(ib_spec->eth.val.ether_type);
+		} else if ((ib_spec->type == (IB_FLOW_SPEC_IPV4 | inner_bit)) ||
+			   (ib_spec->type == (IB_FLOW_SPEC_IPV6 | inner_bit))) {
+			ip_spec_type = ib_spec->type;
+		}
+		ib_spec = (void *)ib_spec + ib_spec->size;
+	}
+
+	type_valid = (!has_ethertype) || (!ip_spec_type);
+	if (!type_valid && mask_valid) {
+		ipv4_spec_valid = (eth_type == ETH_P_IP) &&
+			(ip_spec_type == (IB_FLOW_SPEC_IPV4 | inner_bit));
+		ipv6_spec_valid = (eth_type == ETH_P_IPV6) &&
+			(ip_spec_type == (IB_FLOW_SPEC_IPV6 | inner_bit));
+
+		type_valid = (ipv4_spec_valid) || (ipv6_spec_valid) ||
+			     (((eth_type == ETH_P_MPLS_UC) ||
+			       (eth_type == ETH_P_MPLS_MC)) && match_ipv);
+	}
+
+	return type_valid;
+}
+
+static bool is_valid_attr(struct mlx5_core_dev *mdev,
+			  const struct ib_flow_attr *flow_attr)
+{
+	return is_valid_ethertype(mdev, flow_attr, false) &&
+	       is_valid_ethertype(mdev, flow_attr, true);
+}
+
+static void put_flow_table(struct mlx5_ib_dev *dev,
+			   struct mlx5_ib_flow_prio *prio, bool ft_added)
+{
+	prio->refcount -= !!ft_added;
+	if (!prio->refcount) {
+		mlx5_destroy_flow_table(prio->flow_table);
+		prio->flow_table = NULL;
+	}
+}
+
+static void counters_clear_description(struct ib_counters *counters)
+{
+	struct mlx5_ib_mcounters *mcounters = to_mcounters(counters);
+
+	mutex_lock(&mcounters->mcntrs_mutex);
+	kfree(mcounters->counters_data);
+	mcounters->counters_data = NULL;
+	mcounters->cntrs_max_index = 0;
+	mutex_unlock(&mcounters->mcntrs_mutex);
+}
+
+static int mlx5_ib_destroy_flow(struct ib_flow *flow_id)
+{
+	struct mlx5_ib_flow_handler *handler = container_of(flow_id,
+							  struct mlx5_ib_flow_handler,
+							  ibflow);
+	struct mlx5_ib_flow_handler *iter, *tmp;
+	struct mlx5_ib_dev *dev = handler->dev;
+
+	mutex_lock(&dev->flow_db->lock);
+
+	list_for_each_entry_safe(iter, tmp, &handler->list, list) {
+		mlx5_del_flow_rules(iter->rule);
+		put_flow_table(dev, iter->prio, true);
+		list_del(&iter->list);
+		kfree(iter);
+	}
+
+	mlx5_del_flow_rules(handler->rule);
+	put_flow_table(dev, handler->prio, true);
+	if (handler->ibcounters &&
+	    atomic_read(&handler->ibcounters->usecnt) == 1)
+		counters_clear_description(handler->ibcounters);
+
+	mutex_unlock(&dev->flow_db->lock);
+	if (handler->flow_matcher)
+		atomic_dec(&handler->flow_matcher->usecnt);
+	kfree(handler);
+
+	return 0;
+}
+
+static int ib_prio_to_core_prio(unsigned int priority, bool dont_trap)
+{
+	priority *= 2;
+	if (!dont_trap)
+		priority++;
+	return priority;
+}
+
+enum flow_table_type {
+	MLX5_IB_FT_RX,
+	MLX5_IB_FT_TX
+};
+
+#define MLX5_FS_MAX_TYPES	 6
+#define MLX5_FS_MAX_ENTRIES	 BIT(16)
+
+static struct mlx5_ib_flow_prio *_get_prio(struct mlx5_flow_namespace *ns,
+					   struct mlx5_ib_flow_prio *prio,
+					   int priority,
+					   int num_entries, int num_groups)
+{
+	struct mlx5_flow_table *ft;
+
+	ft = mlx5_create_auto_grouped_flow_table(ns, priority,
+						 num_entries,
+						 num_groups,
+						 0, 0);
+	if (IS_ERR(ft))
+		return ERR_CAST(ft);
+
+	prio->flow_table = ft;
+	prio->refcount = 0;
+	return prio;
+}
+
+static struct mlx5_ib_flow_prio *get_flow_table(struct mlx5_ib_dev *dev,
+						struct ib_flow_attr *flow_attr,
+						enum flow_table_type ft_type)
+{
+	bool dont_trap = flow_attr->flags & IB_FLOW_ATTR_FLAGS_DONT_TRAP;
+	struct mlx5_flow_namespace *ns = NULL;
+	struct mlx5_ib_flow_prio *prio;
+	struct mlx5_flow_table *ft;
+	int max_table_size;
+	int num_entries;
+	int num_groups;
+	int priority;
+
+	max_table_size = BIT(MLX5_CAP_FLOWTABLE_NIC_RX(dev->mdev,
+						       log_max_ft_size));
+	if (flow_attr->type == IB_FLOW_ATTR_NORMAL) {
+		if (ft_type == MLX5_IB_FT_TX)
+			priority = 0;
+		else if (flow_is_multicast_only(flow_attr) &&
+			 !dont_trap)
+			priority = MLX5_IB_FLOW_MCAST_PRIO;
+		else
+			priority = ib_prio_to_core_prio(flow_attr->priority,
+							dont_trap);
+		ns = mlx5_get_flow_namespace(dev->mdev,
+					     ft_type == MLX5_IB_FT_TX ?
+					     MLX5_FLOW_NAMESPACE_EGRESS :
+					     MLX5_FLOW_NAMESPACE_BYPASS);
+		num_entries = MLX5_FS_MAX_ENTRIES;
+		num_groups = MLX5_FS_MAX_TYPES;
+		prio = &dev->flow_db->prios[priority];
+	} else if (flow_attr->type == IB_FLOW_ATTR_ALL_DEFAULT ||
+		   flow_attr->type == IB_FLOW_ATTR_MC_DEFAULT) {
+		ns = mlx5_get_flow_namespace(dev->mdev,
+					     MLX5_FLOW_NAMESPACE_LEFTOVERS);
+		build_leftovers_ft_param(&priority,
+					 &num_entries,
+					 &num_groups);
+		prio = &dev->flow_db->prios[MLX5_IB_FLOW_LEFTOVERS_PRIO];
+	} else if (flow_attr->type == IB_FLOW_ATTR_SNIFFER) {
+		if (!MLX5_CAP_FLOWTABLE(dev->mdev,
+					allow_sniffer_and_nic_rx_shared_tir))
+			return ERR_PTR(-ENOTSUPP);
+
+		ns = mlx5_get_flow_namespace(dev->mdev, ft_type == MLX5_IB_FT_RX ?
+					     MLX5_FLOW_NAMESPACE_SNIFFER_RX :
+					     MLX5_FLOW_NAMESPACE_SNIFFER_TX);
+
+		prio = &dev->flow_db->sniffer[ft_type];
+		priority = 0;
+		num_entries = 1;
+		num_groups = 1;
+	}
+
+	if (!ns)
+		return ERR_PTR(-ENOTSUPP);
+
+	if (num_entries > max_table_size)
+		return ERR_PTR(-ENOMEM);
+
+	ft = prio->flow_table;
+	if (!ft)
+		return _get_prio(ns, prio, priority, num_entries, num_groups);
+
+	return prio;
+}
+
+static void set_underlay_qp(struct mlx5_ib_dev *dev,
+			    struct mlx5_flow_spec *spec,
+			    u32 underlay_qpn)
+{
+	void *misc_params_c = MLX5_ADDR_OF(fte_match_param,
+					   spec->match_criteria,
+					   misc_parameters);
+	void *misc_params_v = MLX5_ADDR_OF(fte_match_param, spec->match_value,
+					   misc_parameters);
+
+	if (underlay_qpn &&
+	    MLX5_CAP_FLOWTABLE_NIC_RX(dev->mdev,
+				      ft_field_support.bth_dst_qp)) {
+		MLX5_SET(fte_match_set_misc,
+			 misc_params_v, bth_dst_qp, underlay_qpn);
+		MLX5_SET(fte_match_set_misc,
+			 misc_params_c, bth_dst_qp, 0xffffff);
+	}
+}
+
+static int read_flow_counters(struct ib_device *ibdev,
+			      struct mlx5_read_counters_attr *read_attr)
+{
+	struct mlx5_fc *fc = read_attr->hw_cntrs_hndl;
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+
+	return mlx5_fc_query(dev->mdev, fc,
+			     &read_attr->out[IB_COUNTER_PACKETS],
+			     &read_attr->out[IB_COUNTER_BYTES]);
+}
+
+/* flow counters currently expose two counters packets and bytes */
+#define FLOW_COUNTERS_NUM 2
+static int counters_set_description(struct ib_counters *counters,
+				    enum mlx5_ib_counters_type counters_type,
+				    struct mlx5_ib_flow_counters_desc *desc_data,
+				    u32 ncounters)
+{
+	struct mlx5_ib_mcounters *mcounters = to_mcounters(counters);
+	u32 cntrs_max_index = 0;
+	int i;
+
+	if (counters_type != MLX5_IB_COUNTERS_FLOW)
+		return -EINVAL;
+
+	/* init the fields for the object */
+	mcounters->type = counters_type;
+	mcounters->read_counters = read_flow_counters;
+	mcounters->counters_num = FLOW_COUNTERS_NUM;
+	mcounters->ncounters = ncounters;
+	/* each counter entry have both description and index pair */
+	for (i = 0; i < ncounters; i++) {
+		if (desc_data[i].description > IB_COUNTER_BYTES)
+			return -EINVAL;
+
+		if (cntrs_max_index <= desc_data[i].index)
+			cntrs_max_index = desc_data[i].index + 1;
+	}
+
+	mutex_lock(&mcounters->mcntrs_mutex);
+	mcounters->counters_data = desc_data;
+	mcounters->cntrs_max_index = cntrs_max_index;
+	mutex_unlock(&mcounters->mcntrs_mutex);
+
+	return 0;
+}
+
+#define MAX_COUNTERS_NUM (USHRT_MAX / (sizeof(u32) * 2))
+static int flow_counters_set_data(struct ib_counters *ibcounters,
+				  struct mlx5_ib_create_flow *ucmd)
+{
+	struct mlx5_ib_mcounters *mcounters = to_mcounters(ibcounters);
+	struct mlx5_ib_flow_counters_data *cntrs_data = NULL;
+	struct mlx5_ib_flow_counters_desc *desc_data = NULL;
+	bool hw_hndl = false;
+	int ret = 0;
+
+	if (ucmd && ucmd->ncounters_data != 0) {
+		cntrs_data = ucmd->data;
+		if (cntrs_data->ncounters > MAX_COUNTERS_NUM)
+			return -EINVAL;
+
+		desc_data = kcalloc(cntrs_data->ncounters,
+				    sizeof(*desc_data),
+				    GFP_KERNEL);
+		if (!desc_data)
+			return  -ENOMEM;
+
+		if (copy_from_user(desc_data,
+				   u64_to_user_ptr(cntrs_data->counters_data),
+				   sizeof(*desc_data) * cntrs_data->ncounters)) {
+			ret = -EFAULT;
+			goto free;
+		}
+	}
+
+	if (!mcounters->hw_cntrs_hndl) {
+		mcounters->hw_cntrs_hndl = mlx5_fc_create(
+			to_mdev(ibcounters->device)->mdev, false);
+		if (IS_ERR(mcounters->hw_cntrs_hndl)) {
+			ret = PTR_ERR(mcounters->hw_cntrs_hndl);
+			goto free;
+		}
+		hw_hndl = true;
+	}
+
+	if (desc_data) {
+		/* counters already bound to at least one flow */
+		if (mcounters->cntrs_max_index) {
+			ret = -EINVAL;
+			goto free_hndl;
+		}
+
+		ret = counters_set_description(ibcounters,
+					       MLX5_IB_COUNTERS_FLOW,
+					       desc_data,
+					       cntrs_data->ncounters);
+		if (ret)
+			goto free_hndl;
+
+	} else if (!mcounters->cntrs_max_index) {
+		/* counters not bound yet, must have udata passed */
+		ret = -EINVAL;
+		goto free_hndl;
+	}
+
+	return 0;
+
+free_hndl:
+	if (hw_hndl) {
+		mlx5_fc_destroy(to_mdev(ibcounters->device)->mdev,
+				mcounters->hw_cntrs_hndl);
+		mcounters->hw_cntrs_hndl = NULL;
+	}
+free:
+	kfree(desc_data);
+	return ret;
+}
+
+static struct mlx5_ib_flow_handler *_create_flow_rule(struct mlx5_ib_dev *dev,
+						      struct mlx5_ib_flow_prio *ft_prio,
+						      const struct ib_flow_attr *flow_attr,
+						      struct mlx5_flow_destination *dst,
+						      u32 underlay_qpn,
+						      struct mlx5_ib_create_flow *ucmd)
+{
+	struct mlx5_flow_table	*ft = ft_prio->flow_table;
+	struct mlx5_ib_flow_handler *handler;
+	struct mlx5_flow_act flow_act = {.flow_tag = MLX5_FS_DEFAULT_FLOW_TAG};
+	struct mlx5_flow_spec *spec;
+	struct mlx5_flow_destination dest_arr[2] = {};
+	struct mlx5_flow_destination *rule_dst = dest_arr;
+	const void *ib_flow = (const void *)flow_attr + sizeof(*flow_attr);
+	unsigned int spec_index;
+	u32 prev_type = 0;
+	int err = 0;
+	int dest_num = 0;
+	bool is_egress = flow_attr->flags & IB_FLOW_ATTR_FLAGS_EGRESS;
+
+	if (!is_valid_attr(dev->mdev, flow_attr))
+		return ERR_PTR(-EINVAL);
+
+	spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
+	handler = kzalloc(sizeof(*handler), GFP_KERNEL);
+	if (!handler || !spec) {
+		err = -ENOMEM;
+		goto free;
+	}
+
+	INIT_LIST_HEAD(&handler->list);
+
+	for (spec_index = 0; spec_index < flow_attr->num_of_specs; spec_index++) {
+		err = parse_flow_attr(dev->mdev, spec->match_criteria,
+				      spec->match_value,
+				      ib_flow, flow_attr, &flow_act,
+				      prev_type);
+		if (err < 0)
+			goto free;
+
+		prev_type = ((union ib_flow_spec *)ib_flow)->type;
+		ib_flow += ((union ib_flow_spec *)ib_flow)->size;
+	}
+
+	if (dst && !(flow_act.action & MLX5_FLOW_CONTEXT_ACTION_DROP)) {
+		memcpy(&dest_arr[0], dst, sizeof(*dst));
+		dest_num++;
+	}
+
+	if (!flow_is_multicast_only(flow_attr))
+		set_underlay_qp(dev, spec, underlay_qpn);
+
+	if (dev->rep) {
+		void *misc;
+
+		misc = MLX5_ADDR_OF(fte_match_param, spec->match_value,
+				    misc_parameters);
+		MLX5_SET(fte_match_set_misc, misc, source_port,
+			 dev->rep->vport);
+		misc = MLX5_ADDR_OF(fte_match_param, spec->match_criteria,
+				    misc_parameters);
+		MLX5_SET_TO_ONES(fte_match_set_misc, misc, source_port);
+	}
+
+	spec->match_criteria_enable = get_match_criteria_enable(spec->match_criteria);
+
+	if (is_egress &&
+	    !is_valid_spec(dev->mdev, spec, &flow_act, is_egress)) {
+		err = -EINVAL;
+		goto free;
+	}
+
+	if (flow_act.action & MLX5_FLOW_CONTEXT_ACTION_COUNT) {
+		err = flow_counters_set_data(flow_act.counters, ucmd);
+		if (err)
+			goto free;
+
+		handler->ibcounters = flow_act.counters;
+		dest_arr[dest_num].type =
+			MLX5_FLOW_DESTINATION_TYPE_COUNTER;
+		dest_arr[dest_num].counter =
+			to_mcounters(flow_act.counters)->hw_cntrs_hndl;
+		dest_num++;
+	}
+
+	if (flow_act.action & MLX5_FLOW_CONTEXT_ACTION_DROP) {
+		if (!dest_num)
+			rule_dst = NULL;
+	} else {
+		if (is_egress)
+			flow_act.action |= MLX5_FLOW_CONTEXT_ACTION_ALLOW;
+		else
+			flow_act.action |=
+				dest_num ?  MLX5_FLOW_CONTEXT_ACTION_FWD_DEST :
+					MLX5_FLOW_CONTEXT_ACTION_FWD_NEXT_PRIO;
+	}
+
+	if (flow_act.has_flow_tag &&
+	    (flow_attr->type == IB_FLOW_ATTR_ALL_DEFAULT ||
+	     flow_attr->type == IB_FLOW_ATTR_MC_DEFAULT)) {
+		mlx5_ib_warn(dev, "Flow tag %u and attribute type %x isn't allowed in leftovers\n",
+			     flow_act.flow_tag, flow_attr->type);
+		err = -EINVAL;
+		goto free;
+	}
+	handler->rule = mlx5_add_flow_rules(ft, spec,
+					    &flow_act,
+					    rule_dst, dest_num);
+
+	if (IS_ERR(handler->rule)) {
+		err = PTR_ERR(handler->rule);
+		goto free;
+	}
+
+	ft_prio->refcount++;
+	handler->prio = ft_prio;
+	handler->dev = dev;
+
+	ft_prio->flow_table = ft;
+free:
+	if (err && handler) {
+		if (handler->ibcounters &&
+		    atomic_read(&handler->ibcounters->usecnt) == 1)
+			counters_clear_description(handler->ibcounters);
+		kfree(handler);
+	}
+	kvfree(spec);
+	return err ? ERR_PTR(err) : handler;
+}
+
+static struct mlx5_ib_flow_handler *create_flow_rule(struct mlx5_ib_dev *dev,
+						     struct mlx5_ib_flow_prio *ft_prio,
+						     const struct ib_flow_attr *flow_attr,
+						     struct mlx5_flow_destination *dst)
+{
+	return _create_flow_rule(dev, ft_prio, flow_attr, dst, 0, NULL);
+}
+
+static struct mlx5_ib_flow_handler *create_dont_trap_rule(struct mlx5_ib_dev *dev,
+							  struct mlx5_ib_flow_prio *ft_prio,
+							  struct ib_flow_attr *flow_attr,
+							  struct mlx5_flow_destination *dst)
+{
+	struct mlx5_ib_flow_handler *handler_dst = NULL;
+	struct mlx5_ib_flow_handler *handler = NULL;
+
+	handler = create_flow_rule(dev, ft_prio, flow_attr, NULL);
+	if (!IS_ERR(handler)) {
+		handler_dst = create_flow_rule(dev, ft_prio,
+					       flow_attr, dst);
+		if (IS_ERR(handler_dst)) {
+			mlx5_del_flow_rules(handler->rule);
+			ft_prio->refcount--;
+			kfree(handler);
+			handler = handler_dst;
+		} else {
+			list_add(&handler_dst->list, &handler->list);
+		}
+	}
+
+	return handler;
+}
+enum {
+	LEFTOVERS_MC,
+	LEFTOVERS_UC,
+};
+
+static struct mlx5_ib_flow_handler *create_leftovers_rule(struct mlx5_ib_dev *dev,
+							  struct mlx5_ib_flow_prio *ft_prio,
+							  struct ib_flow_attr *flow_attr,
+							  struct mlx5_flow_destination *dst)
+{
+	struct mlx5_ib_flow_handler *handler_ucast = NULL;
+	struct mlx5_ib_flow_handler *handler = NULL;
+
+	static struct {
+		struct ib_flow_attr	flow_attr;
+		struct ib_flow_spec_eth eth_flow;
+	} leftovers_specs[] = {
+		[LEFTOVERS_MC] = {
+			.flow_attr = {
+				.num_of_specs = 1,
+				.size = sizeof(leftovers_specs[0])
+			},
+			.eth_flow = {
+				.type = IB_FLOW_SPEC_ETH,
+				.size = sizeof(struct ib_flow_spec_eth),
+				.mask = {.dst_mac = {0x1} },
+				.val =  {.dst_mac = {0x1} }
+			}
+		},
+		[LEFTOVERS_UC] = {
+			.flow_attr = {
+				.num_of_specs = 1,
+				.size = sizeof(leftovers_specs[0])
+			},
+			.eth_flow = {
+				.type = IB_FLOW_SPEC_ETH,
+				.size = sizeof(struct ib_flow_spec_eth),
+				.mask = {.dst_mac = {0x1} },
+				.val = {.dst_mac = {} }
+			}
+		}
+	};
+
+	handler = create_flow_rule(dev, ft_prio,
+				   &leftovers_specs[LEFTOVERS_MC].flow_attr,
+				   dst);
+	if (!IS_ERR(handler) &&
+	    flow_attr->type == IB_FLOW_ATTR_ALL_DEFAULT) {
+		handler_ucast = create_flow_rule(dev, ft_prio,
+						 &leftovers_specs[LEFTOVERS_UC].flow_attr,
+						 dst);
+		if (IS_ERR(handler_ucast)) {
+			mlx5_del_flow_rules(handler->rule);
+			ft_prio->refcount--;
+			kfree(handler);
+			handler = handler_ucast;
+		} else {
+			list_add(&handler_ucast->list, &handler->list);
+		}
+	}
+
+	return handler;
+}
+
+static struct mlx5_ib_flow_handler *create_sniffer_rule(struct mlx5_ib_dev *dev,
+							struct mlx5_ib_flow_prio *ft_rx,
+							struct mlx5_ib_flow_prio *ft_tx,
+							struct mlx5_flow_destination *dst)
+{
+	struct mlx5_ib_flow_handler *handler_rx;
+	struct mlx5_ib_flow_handler *handler_tx;
+	int err;
+	static const struct ib_flow_attr flow_attr  = {
+		.num_of_specs = 0,
+		.size = sizeof(flow_attr)
+	};
+
+	handler_rx = create_flow_rule(dev, ft_rx, &flow_attr, dst);
+	if (IS_ERR(handler_rx)) {
+		err = PTR_ERR(handler_rx);
+		goto err;
+	}
+
+	handler_tx = create_flow_rule(dev, ft_tx, &flow_attr, dst);
+	if (IS_ERR(handler_tx)) {
+		err = PTR_ERR(handler_tx);
+		goto err_tx;
+	}
+
+	list_add(&handler_tx->list, &handler_rx->list);
+
+	return handler_rx;
+
+err_tx:
+	mlx5_del_flow_rules(handler_rx->rule);
+	ft_rx->refcount--;
+	kfree(handler_rx);
+err:
+	return ERR_PTR(err);
+}
+
+static struct ib_flow *mlx5_ib_create_flow(struct ib_qp *qp,
+					   struct ib_flow_attr *flow_attr,
+					   int domain,
+					   struct ib_udata *udata)
+{
+	struct mlx5_ib_dev *dev = to_mdev(qp->device);
+	struct mlx5_ib_qp *mqp = to_mqp(qp);
+	struct mlx5_ib_flow_handler *handler = NULL;
+	struct mlx5_flow_destination *dst = NULL;
+	struct mlx5_ib_flow_prio *ft_prio_tx = NULL;
+	struct mlx5_ib_flow_prio *ft_prio;
+	bool is_egress = flow_attr->flags & IB_FLOW_ATTR_FLAGS_EGRESS;
+	struct mlx5_ib_create_flow *ucmd = NULL, ucmd_hdr;
+	size_t min_ucmd_sz, required_ucmd_sz;
+	int err;
+	int underlay_qpn;
+
+	if (udata && udata->inlen) {
+		min_ucmd_sz = offsetof(typeof(ucmd_hdr), reserved) +
+				sizeof(ucmd_hdr.reserved);
+		if (udata->inlen < min_ucmd_sz)
+			return ERR_PTR(-EOPNOTSUPP);
+
+		err = ib_copy_from_udata(&ucmd_hdr, udata, min_ucmd_sz);
+		if (err)
+			return ERR_PTR(err);
+
+		/* currently supports only one counters data */
+		if (ucmd_hdr.ncounters_data > 1)
+			return ERR_PTR(-EINVAL);
+
+		required_ucmd_sz = min_ucmd_sz +
+			sizeof(struct mlx5_ib_flow_counters_data) *
+			ucmd_hdr.ncounters_data;
+		if (udata->inlen > required_ucmd_sz &&
+		    !ib_is_udata_cleared(udata, required_ucmd_sz,
+					 udata->inlen - required_ucmd_sz))
+			return ERR_PTR(-EOPNOTSUPP);
+
+		ucmd = kzalloc(required_ucmd_sz, GFP_KERNEL);
+		if (!ucmd)
+			return ERR_PTR(-ENOMEM);
+
+		err = ib_copy_from_udata(ucmd, udata, required_ucmd_sz);
+		if (err)
+			goto free_ucmd;
+	}
+
+	if (flow_attr->priority > MLX5_IB_FLOW_LAST_PRIO) {
+		err = -ENOMEM;
+		goto free_ucmd;
+	}
+
+	if (domain != IB_FLOW_DOMAIN_USER ||
+	    flow_attr->port > dev->num_ports ||
+	    (flow_attr->flags & ~(IB_FLOW_ATTR_FLAGS_DONT_TRAP |
+				  IB_FLOW_ATTR_FLAGS_EGRESS))) {
+		err = -EINVAL;
+		goto free_ucmd;
+	}
+
+	if (is_egress &&
+	    (flow_attr->type == IB_FLOW_ATTR_ALL_DEFAULT ||
+	     flow_attr->type == IB_FLOW_ATTR_MC_DEFAULT)) {
+		err = -EINVAL;
+		goto free_ucmd;
+	}
+
+	dst = kzalloc(sizeof(*dst), GFP_KERNEL);
+	if (!dst) {
+		err = -ENOMEM;
+		goto free_ucmd;
+	}
+
+	mutex_lock(&dev->flow_db->lock);
+
+	ft_prio = get_flow_table(dev, flow_attr,
+				 is_egress ? MLX5_IB_FT_TX : MLX5_IB_FT_RX);
+	if (IS_ERR(ft_prio)) {
+		err = PTR_ERR(ft_prio);
+		goto unlock;
+	}
+	if (flow_attr->type == IB_FLOW_ATTR_SNIFFER) {
+		ft_prio_tx = get_flow_table(dev, flow_attr, MLX5_IB_FT_TX);
+		if (IS_ERR(ft_prio_tx)) {
+			err = PTR_ERR(ft_prio_tx);
+			ft_prio_tx = NULL;
+			goto destroy_ft;
+		}
+	}
+
+	if (is_egress) {
+		dst->type = MLX5_FLOW_DESTINATION_TYPE_PORT;
+	} else {
+		dst->type = MLX5_FLOW_DESTINATION_TYPE_TIR;
+		if (mqp->flags & MLX5_IB_QP_RSS)
+			dst->tir_num = mqp->rss_qp.tirn;
+		else
+			dst->tir_num = mqp->raw_packet_qp.rq.tirn;
+	}
+
+	if (flow_attr->type == IB_FLOW_ATTR_NORMAL) {
+		if (flow_attr->flags & IB_FLOW_ATTR_FLAGS_DONT_TRAP)  {
+			handler = create_dont_trap_rule(dev, ft_prio,
+							flow_attr, dst);
+		} else {
+			underlay_qpn = (mqp->flags & MLX5_IB_QP_UNDERLAY) ?
+					mqp->underlay_qpn : 0;
+			handler = _create_flow_rule(dev, ft_prio, flow_attr,
+						    dst, underlay_qpn, ucmd);
+		}
+	} else if (flow_attr->type == IB_FLOW_ATTR_ALL_DEFAULT ||
+		   flow_attr->type == IB_FLOW_ATTR_MC_DEFAULT) {
+		handler = create_leftovers_rule(dev, ft_prio, flow_attr,
+						dst);
+	} else if (flow_attr->type == IB_FLOW_ATTR_SNIFFER) {
+		handler = create_sniffer_rule(dev, ft_prio, ft_prio_tx, dst);
+	} else {
+		err = -EINVAL;
+		goto destroy_ft;
+	}
+
+	if (IS_ERR(handler)) {
+		err = PTR_ERR(handler);
+		handler = NULL;
+		goto destroy_ft;
+	}
+
+	mutex_unlock(&dev->flow_db->lock);
+	kfree(dst);
+	kfree(ucmd);
+
+	return &handler->ibflow;
+
+destroy_ft:
+	put_flow_table(dev, ft_prio, false);
+	if (ft_prio_tx)
+		put_flow_table(dev, ft_prio_tx, false);
+unlock:
+	mutex_unlock(&dev->flow_db->lock);
+	kfree(dst);
+free_ucmd:
+	kfree(ucmd);
+	return ERR_PTR(err);
+}
+
+static struct mlx5_ib_flow_prio *_get_flow_table(struct mlx5_ib_dev *dev,
+						 int priority, bool mcast)
+{
+	int max_table_size;
+	struct mlx5_flow_namespace *ns = NULL;
+	struct mlx5_ib_flow_prio *prio;
+
+	max_table_size = BIT(MLX5_CAP_FLOWTABLE_NIC_RX(dev->mdev,
+			     log_max_ft_size));
+	if (max_table_size < MLX5_FS_MAX_ENTRIES)
+		return ERR_PTR(-ENOMEM);
+
+	if (mcast)
+		priority = MLX5_IB_FLOW_MCAST_PRIO;
+	else
+		priority = ib_prio_to_core_prio(priority, false);
+
+	ns = mlx5_get_flow_namespace(dev->mdev, MLX5_FLOW_NAMESPACE_BYPASS);
+	if (!ns)
+		return ERR_PTR(-ENOTSUPP);
+
+	prio = &dev->flow_db->prios[priority];
+
+	if (prio->flow_table)
+		return prio;
+
+	return _get_prio(ns, prio, priority, MLX5_FS_MAX_ENTRIES,
+			 MLX5_FS_MAX_TYPES);
+}
+
+static struct mlx5_ib_flow_handler *
+_create_raw_flow_rule(struct mlx5_ib_dev *dev,
+		      struct mlx5_ib_flow_prio *ft_prio,
+		      struct mlx5_flow_destination *dst,
+		      struct mlx5_ib_flow_matcher  *fs_matcher,
+		      void *cmd_in, int inlen)
+{
+	struct mlx5_ib_flow_handler *handler;
+	struct mlx5_flow_act flow_act = {.flow_tag = MLX5_FS_DEFAULT_FLOW_TAG};
+	struct mlx5_flow_spec *spec;
+	struct mlx5_flow_table *ft = ft_prio->flow_table;
+	int err = 0;
+
+	spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
+	handler = kzalloc(sizeof(*handler), GFP_KERNEL);
+	if (!handler || !spec) {
+		err = -ENOMEM;
+		goto free;
+	}
+
+	INIT_LIST_HEAD(&handler->list);
+
+	memcpy(spec->match_value, cmd_in, inlen);
+	memcpy(spec->match_criteria, fs_matcher->matcher_mask.match_params,
+	       fs_matcher->mask_len);
+	spec->match_criteria_enable = fs_matcher->match_criteria_enable;
+
+	flow_act.action |= MLX5_FLOW_CONTEXT_ACTION_FWD_DEST;
+	handler->rule = mlx5_add_flow_rules(ft, spec,
+					    &flow_act, dst, 1);
+
+	if (IS_ERR(handler->rule)) {
+		err = PTR_ERR(handler->rule);
+		goto free;
+	}
+
+	ft_prio->refcount++;
+	handler->prio = ft_prio;
+	handler->dev = dev;
+	ft_prio->flow_table = ft;
+
+free:
+	if (err)
+		kfree(handler);
+	kvfree(spec);
+	return err ? ERR_PTR(err) : handler;
+}
+
+static bool raw_fs_is_multicast(struct mlx5_ib_flow_matcher *fs_matcher,
+				void *match_v)
+{
+	void *match_c;
+	void *match_v_set_lyr_2_4, *match_c_set_lyr_2_4;
+	void *dmac, *dmac_mask;
+	void *ipv4, *ipv4_mask;
+
+	if (!(fs_matcher->match_criteria_enable &
+	      (1 << MATCH_CRITERIA_ENABLE_OUTER_BIT)))
+		return false;
+
+	match_c = fs_matcher->matcher_mask.match_params;
+	match_v_set_lyr_2_4 = MLX5_ADDR_OF(fte_match_param, match_v,
+					   outer_headers);
+	match_c_set_lyr_2_4 = MLX5_ADDR_OF(fte_match_param, match_c,
+					   outer_headers);
+
+	dmac = MLX5_ADDR_OF(fte_match_set_lyr_2_4, match_v_set_lyr_2_4,
+			    dmac_47_16);
+	dmac_mask = MLX5_ADDR_OF(fte_match_set_lyr_2_4, match_c_set_lyr_2_4,
+				 dmac_47_16);
+
+	if (is_multicast_ether_addr(dmac) &&
+	    is_multicast_ether_addr(dmac_mask))
+		return true;
+
+	ipv4 = MLX5_ADDR_OF(fte_match_set_lyr_2_4, match_v_set_lyr_2_4,
+			    dst_ipv4_dst_ipv6.ipv4_layout.ipv4);
+
+	ipv4_mask = MLX5_ADDR_OF(fte_match_set_lyr_2_4, match_c_set_lyr_2_4,
+				 dst_ipv4_dst_ipv6.ipv4_layout.ipv4);
+
+	if (ipv4_is_multicast(*(__be32 *)(ipv4)) &&
+	    ipv4_is_multicast(*(__be32 *)(ipv4_mask)))
+		return true;
+
+	return false;
+}
+
+struct mlx5_ib_flow_handler *
+mlx5_ib_raw_fs_rule_add(struct mlx5_ib_dev *dev,
+			struct mlx5_ib_flow_matcher *fs_matcher,
+			void *cmd_in, int inlen, int dest_id,
+			int dest_type)
+{
+	struct mlx5_flow_destination *dst;
+	struct mlx5_ib_flow_prio *ft_prio;
+	int priority = fs_matcher->priority;
+	struct mlx5_ib_flow_handler *handler;
+	bool mcast;
+	int err;
+
+	if (fs_matcher->flow_type != MLX5_IB_FLOW_TYPE_NORMAL)
+		return ERR_PTR(-EOPNOTSUPP);
+
+	if (fs_matcher->priority > MLX5_IB_FLOW_LAST_PRIO)
+		return ERR_PTR(-ENOMEM);
+
+	dst = kzalloc(sizeof(*dst), GFP_KERNEL);
+	if (!dst)
+		return ERR_PTR(-ENOMEM);
+
+	mcast = raw_fs_is_multicast(fs_matcher, cmd_in);
+	mutex_lock(&dev->flow_db->lock);
+
+	ft_prio = _get_flow_table(dev, priority, mcast);
+	if (IS_ERR(ft_prio)) {
+		err = PTR_ERR(ft_prio);
+		goto unlock;
+	}
+
+	if (dest_type == MLX5_FLOW_DESTINATION_TYPE_TIR) {
+		dst->type = dest_type;
+		dst->tir_num = dest_id;
+	} else {
+		dst->type = MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE_NUM;
+		dst->ft_num = dest_id;
+	}
+
+	handler = _create_raw_flow_rule(dev, ft_prio, dst, fs_matcher, cmd_in,
+					inlen);
+
+	if (IS_ERR(handler)) {
+		err = PTR_ERR(handler);
+		goto destroy_ft;
+	}
+
+	mutex_unlock(&dev->flow_db->lock);
+	atomic_inc(&fs_matcher->usecnt);
+	handler->flow_matcher = fs_matcher;
+
+	kfree(dst);
+
+	return handler;
+
+destroy_ft:
+	put_flow_table(dev, ft_prio, false);
+unlock:
+	mutex_unlock(&dev->flow_db->lock);
+	kfree(dst);
+
+	return ERR_PTR(err);
+}
+
+static u32 mlx5_ib_flow_action_flags_to_accel_xfrm_flags(u32 mlx5_flags)
+{
+	u32 flags = 0;
+
+	if (mlx5_flags & MLX5_IB_UAPI_FLOW_ACTION_FLAGS_REQUIRE_METADATA)
+		flags |= MLX5_ACCEL_XFRM_FLAG_REQUIRE_METADATA;
+
+	return flags;
+}
+
+#define MLX5_FLOW_ACTION_ESP_CREATE_LAST_SUPPORTED	MLX5_IB_UAPI_FLOW_ACTION_FLAGS_REQUIRE_METADATA
+static struct ib_flow_action *
+mlx5_ib_create_flow_action_esp(struct ib_device *device,
+			       const struct ib_flow_action_attrs_esp *attr,
+			       struct uverbs_attr_bundle *attrs)
+{
+	struct mlx5_ib_dev *mdev = to_mdev(device);
+	struct ib_uverbs_flow_action_esp_keymat_aes_gcm *aes_gcm;
+	struct mlx5_accel_esp_xfrm_attrs accel_attrs = {};
+	struct mlx5_ib_flow_action *action;
+	u64 action_flags;
+	u64 flags;
+	int err = 0;
+
+	err = uverbs_get_flags64(
+		&action_flags, attrs, MLX5_IB_ATTR_CREATE_FLOW_ACTION_FLAGS,
+		((MLX5_FLOW_ACTION_ESP_CREATE_LAST_SUPPORTED << 1) - 1));
+	if (err)
+		return ERR_PTR(err);
+
+	flags = mlx5_ib_flow_action_flags_to_accel_xfrm_flags(action_flags);
+
+	/* We current only support a subset of the standard features. Only a
+	 * keymat of type AES_GCM, with icv_len == 16, iv_algo == SEQ and esn
+	 * (with overlap). Full offload mode isn't supported.
+	 */
+	if (!attr->keymat || attr->replay || attr->encap ||
+	    attr->spi || attr->seq || attr->tfc_pad ||
+	    attr->hard_limit_pkts ||
+	    (attr->flags & ~(IB_FLOW_ACTION_ESP_FLAGS_ESN_TRIGGERED |
+			     IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ENCRYPT)))
+		return ERR_PTR(-EOPNOTSUPP);
+
+	if (attr->keymat->protocol !=
+	    IB_UVERBS_FLOW_ACTION_ESP_KEYMAT_AES_GCM)
+		return ERR_PTR(-EOPNOTSUPP);
+
+	aes_gcm = &attr->keymat->keymat.aes_gcm;
+
+	if (aes_gcm->icv_len != 16 ||
+	    aes_gcm->iv_algo != IB_UVERBS_FLOW_ACTION_IV_ALGO_SEQ)
+		return ERR_PTR(-EOPNOTSUPP);
+
+	action = kmalloc(sizeof(*action), GFP_KERNEL);
+	if (!action)
+		return ERR_PTR(-ENOMEM);
+
+	action->esp_aes_gcm.ib_flags = attr->flags;
+	memcpy(&accel_attrs.keymat.aes_gcm.aes_key, &aes_gcm->aes_key,
+	       sizeof(accel_attrs.keymat.aes_gcm.aes_key));
+	accel_attrs.keymat.aes_gcm.key_len = aes_gcm->key_len * 8;
+	memcpy(&accel_attrs.keymat.aes_gcm.salt, &aes_gcm->salt,
+	       sizeof(accel_attrs.keymat.aes_gcm.salt));
+	memcpy(&accel_attrs.keymat.aes_gcm.seq_iv, &aes_gcm->iv,
+	       sizeof(accel_attrs.keymat.aes_gcm.seq_iv));
+	accel_attrs.keymat.aes_gcm.icv_len = aes_gcm->icv_len * 8;
+	accel_attrs.keymat.aes_gcm.iv_algo = MLX5_ACCEL_ESP_AES_GCM_IV_ALGO_SEQ;
+	accel_attrs.keymat_type = MLX5_ACCEL_ESP_KEYMAT_AES_GCM;
+
+	accel_attrs.esn = attr->esn;
+	if (attr->flags & IB_FLOW_ACTION_ESP_FLAGS_ESN_TRIGGERED)
+		accel_attrs.flags |= MLX5_ACCEL_ESP_FLAGS_ESN_TRIGGERED;
+	if (attr->flags & IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ESN_NEW_WINDOW)
+		accel_attrs.flags |= MLX5_ACCEL_ESP_FLAGS_ESN_STATE_OVERLAP;
+
+	if (attr->flags & IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ENCRYPT)
+		accel_attrs.action |= MLX5_ACCEL_ESP_ACTION_ENCRYPT;
+
+	action->esp_aes_gcm.ctx =
+		mlx5_accel_esp_create_xfrm(mdev->mdev, &accel_attrs, flags);
+	if (IS_ERR(action->esp_aes_gcm.ctx)) {
+		err = PTR_ERR(action->esp_aes_gcm.ctx);
+		goto err_parse;
+	}
+
+	action->esp_aes_gcm.ib_flags = attr->flags;
+
+	return &action->ib_action;
+
+err_parse:
+	kfree(action);
+	return ERR_PTR(err);
+}
+
+static int
+mlx5_ib_modify_flow_action_esp(struct ib_flow_action *action,
+			       const struct ib_flow_action_attrs_esp *attr,
+			       struct uverbs_attr_bundle *attrs)
+{
+	struct mlx5_ib_flow_action *maction = to_mflow_act(action);
+	struct mlx5_accel_esp_xfrm_attrs accel_attrs;
+	int err = 0;
+
+	if (attr->keymat || attr->replay || attr->encap ||
+	    attr->spi || attr->seq || attr->tfc_pad ||
+	    attr->hard_limit_pkts ||
+	    (attr->flags & ~(IB_FLOW_ACTION_ESP_FLAGS_ESN_TRIGGERED |
+			     IB_FLOW_ACTION_ESP_FLAGS_MOD_ESP_ATTRS |
+			     IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ESN_NEW_WINDOW)))
+		return -EOPNOTSUPP;
+
+	/* Only the ESN value or the MLX5_ACCEL_ESP_FLAGS_ESN_STATE_OVERLAP can
+	 * be modified.
+	 */
+	if (!(maction->esp_aes_gcm.ib_flags &
+	      IB_FLOW_ACTION_ESP_FLAGS_ESN_TRIGGERED) &&
+	    attr->flags & (IB_FLOW_ACTION_ESP_FLAGS_ESN_TRIGGERED |
+			   IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ESN_NEW_WINDOW))
+		return -EINVAL;
+
+	memcpy(&accel_attrs, &maction->esp_aes_gcm.ctx->attrs,
+	       sizeof(accel_attrs));
+
+	accel_attrs.esn = attr->esn;
+	if (attr->flags & IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ESN_NEW_WINDOW)
+		accel_attrs.flags |= MLX5_ACCEL_ESP_FLAGS_ESN_STATE_OVERLAP;
+	else
+		accel_attrs.flags &= ~MLX5_ACCEL_ESP_FLAGS_ESN_STATE_OVERLAP;
+
+	err = mlx5_accel_esp_modify_xfrm(maction->esp_aes_gcm.ctx,
+					 &accel_attrs);
+	if (err)
+		return err;
+
+	maction->esp_aes_gcm.ib_flags &=
+		~IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ESN_NEW_WINDOW;
+	maction->esp_aes_gcm.ib_flags |=
+		attr->flags & IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ESN_NEW_WINDOW;
+
+	return 0;
+}
+
+static int mlx5_ib_destroy_flow_action(struct ib_flow_action *action)
+{
+	struct mlx5_ib_flow_action *maction = to_mflow_act(action);
+
+	switch (action->type) {
+	case IB_FLOW_ACTION_ESP:
+		/*
+		 * We only support aes_gcm by now, so we implicitly know this is
+		 * the underline crypto.
+		 */
+		mlx5_accel_esp_destroy_xfrm(maction->esp_aes_gcm.ctx);
+		break;
+	default:
+		WARN_ON(true);
+		break;
+	}
+
+	kfree(maction);
+	return 0;
+}
+
+static int mlx5_ib_mcg_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibqp->device);
+	struct mlx5_ib_qp *mqp = to_mqp(ibqp);
+	int err;
+
+	if (mqp->flags & MLX5_IB_QP_UNDERLAY) {
+		mlx5_ib_dbg(dev, "Attaching a multi cast group to underlay QP is not supported\n");
+		return -EOPNOTSUPP;
+	}
+
+	err = mlx5_core_attach_mcg(dev->mdev, gid, ibqp->qp_num);
+	if (err)
+		mlx5_ib_warn(dev, "failed attaching QPN 0x%x, MGID %pI6\n",
+			     ibqp->qp_num, gid->raw);
+
+	return err;
+}
+
+static int mlx5_ib_mcg_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibqp->device);
+	int err;
+
+	err = mlx5_core_detach_mcg(dev->mdev, gid, ibqp->qp_num);
+	if (err)
+		mlx5_ib_warn(dev, "failed detaching QPN 0x%x, MGID %pI6\n",
+			     ibqp->qp_num, gid->raw);
+
+	return err;
+}
+
+static int init_node_data(struct mlx5_ib_dev *dev)
+{
+	int err;
+
+	err = mlx5_query_node_desc(dev, dev->ib_dev.node_desc);
+	if (err)
+		return err;
+
+	dev->mdev->rev_id = dev->mdev->pdev->revision;
+
+	return mlx5_query_node_guid(dev, &dev->ib_dev.node_guid);
+}
+
+static ssize_t show_fw_pages(struct device *device, struct device_attribute *attr,
+			     char *buf)
+{
+	struct mlx5_ib_dev *dev =
+		container_of(device, struct mlx5_ib_dev, ib_dev.dev);
+
+	return sprintf(buf, "%d\n", dev->mdev->priv.fw_pages);
+}
+
+static ssize_t show_reg_pages(struct device *device,
+			      struct device_attribute *attr, char *buf)
+{
+	struct mlx5_ib_dev *dev =
+		container_of(device, struct mlx5_ib_dev, ib_dev.dev);
+
+	return sprintf(buf, "%d\n", atomic_read(&dev->mdev->priv.reg_pages));
+}
+
+static ssize_t show_hca(struct device *device, struct device_attribute *attr,
+			char *buf)
+{
+	struct mlx5_ib_dev *dev =
+		container_of(device, struct mlx5_ib_dev, ib_dev.dev);
+	return sprintf(buf, "MT%d\n", dev->mdev->pdev->device);
+}
+
+static ssize_t show_rev(struct device *device, struct device_attribute *attr,
+			char *buf)
+{
+	struct mlx5_ib_dev *dev =
+		container_of(device, struct mlx5_ib_dev, ib_dev.dev);
+	return sprintf(buf, "%x\n", dev->mdev->rev_id);
+}
+
+static ssize_t show_board(struct device *device, struct device_attribute *attr,
+			  char *buf)
+{
+	struct mlx5_ib_dev *dev =
+		container_of(device, struct mlx5_ib_dev, ib_dev.dev);
+	return sprintf(buf, "%.*s\n", MLX5_BOARD_ID_LEN,
+		       dev->mdev->board_id);
+}
+
+static DEVICE_ATTR(hw_rev,   S_IRUGO, show_rev,    NULL);
+static DEVICE_ATTR(hca_type, S_IRUGO, show_hca,    NULL);
+static DEVICE_ATTR(board_id, S_IRUGO, show_board,  NULL);
+static DEVICE_ATTR(fw_pages, S_IRUGO, show_fw_pages, NULL);
+static DEVICE_ATTR(reg_pages, S_IRUGO, show_reg_pages, NULL);
+
+static struct device_attribute *mlx5_class_attributes[] = {
+	&dev_attr_hw_rev,
+	&dev_attr_hca_type,
+	&dev_attr_board_id,
+	&dev_attr_fw_pages,
+	&dev_attr_reg_pages,
+};
+
+static void pkey_change_handler(struct work_struct *work)
+{
+	struct mlx5_ib_port_resources *ports =
+		container_of(work, struct mlx5_ib_port_resources,
+			     pkey_change_work);
+
+	mutex_lock(&ports->devr->mutex);
+	mlx5_ib_gsi_pkey_change(ports->gsi);
+	mutex_unlock(&ports->devr->mutex);
+}
+
+static void mlx5_ib_handle_internal_error(struct mlx5_ib_dev *ibdev)
+{
+	struct mlx5_ib_qp *mqp;
+	struct mlx5_ib_cq *send_mcq, *recv_mcq;
+	struct mlx5_core_cq *mcq;
+	struct list_head cq_armed_list;
+	unsigned long flags_qp;
+	unsigned long flags_cq;
+	unsigned long flags;
+
+	INIT_LIST_HEAD(&cq_armed_list);
+
+	/* Go over qp list reside on that ibdev, sync with create/destroy qp.*/
+	spin_lock_irqsave(&ibdev->reset_flow_resource_lock, flags);
+	list_for_each_entry(mqp, &ibdev->qp_list, qps_list) {
+		spin_lock_irqsave(&mqp->sq.lock, flags_qp);
+		if (mqp->sq.tail != mqp->sq.head) {
+			send_mcq = to_mcq(mqp->ibqp.send_cq);
+			spin_lock_irqsave(&send_mcq->lock, flags_cq);
+			if (send_mcq->mcq.comp &&
+			    mqp->ibqp.send_cq->comp_handler) {
+				if (!send_mcq->mcq.reset_notify_added) {
+					send_mcq->mcq.reset_notify_added = 1;
+					list_add_tail(&send_mcq->mcq.reset_notify,
+						      &cq_armed_list);
+				}
+			}
+			spin_unlock_irqrestore(&send_mcq->lock, flags_cq);
+		}
+		spin_unlock_irqrestore(&mqp->sq.lock, flags_qp);
+		spin_lock_irqsave(&mqp->rq.lock, flags_qp);
+		/* no handling is needed for SRQ */
+		if (!mqp->ibqp.srq) {
+			if (mqp->rq.tail != mqp->rq.head) {
+				recv_mcq = to_mcq(mqp->ibqp.recv_cq);
+				spin_lock_irqsave(&recv_mcq->lock, flags_cq);
+				if (recv_mcq->mcq.comp &&
+				    mqp->ibqp.recv_cq->comp_handler) {
+					if (!recv_mcq->mcq.reset_notify_added) {
+						recv_mcq->mcq.reset_notify_added = 1;
+						list_add_tail(&recv_mcq->mcq.reset_notify,
+							      &cq_armed_list);
+					}
+				}
+				spin_unlock_irqrestore(&recv_mcq->lock,
+						       flags_cq);
+			}
+		}
+		spin_unlock_irqrestore(&mqp->rq.lock, flags_qp);
+	}
+	/*At that point all inflight post send were put to be executed as of we
+	 * lock/unlock above locks Now need to arm all involved CQs.
+	 */
+	list_for_each_entry(mcq, &cq_armed_list, reset_notify) {
+		mcq->comp(mcq);
+	}
+	spin_unlock_irqrestore(&ibdev->reset_flow_resource_lock, flags);
+}
+
+static void delay_drop_handler(struct work_struct *work)
+{
+	int err;
+	struct mlx5_ib_delay_drop *delay_drop =
+		container_of(work, struct mlx5_ib_delay_drop,
+			     delay_drop_work);
+
+	atomic_inc(&delay_drop->events_cnt);
+
+	mutex_lock(&delay_drop->lock);
+	err = mlx5_core_set_delay_drop(delay_drop->dev->mdev,
+				       delay_drop->timeout);
+	if (err) {
+		mlx5_ib_warn(delay_drop->dev, "Failed to set delay drop, timeout=%u\n",
+			     delay_drop->timeout);
+		delay_drop->activate = false;
+	}
+	mutex_unlock(&delay_drop->lock);
+}
+
+static void mlx5_ib_handle_event(struct work_struct *_work)
+{
+	struct mlx5_ib_event_work *work =
+		container_of(_work, struct mlx5_ib_event_work, work);
+	struct mlx5_ib_dev *ibdev;
+	struct ib_event ibev;
+	bool fatal = false;
+	u8 port = (u8)work->param;
+
+	if (mlx5_core_is_mp_slave(work->dev)) {
+		ibdev = mlx5_ib_get_ibdev_from_mpi(work->context);
+		if (!ibdev)
+			goto out;
+	} else {
+		ibdev = work->context;
+	}
+
+	switch (work->event) {
+	case MLX5_DEV_EVENT_SYS_ERROR:
+		ibev.event = IB_EVENT_DEVICE_FATAL;
+		mlx5_ib_handle_internal_error(ibdev);
+		fatal = true;
+		break;
+
+	case MLX5_DEV_EVENT_PORT_UP:
+	case MLX5_DEV_EVENT_PORT_DOWN:
+	case MLX5_DEV_EVENT_PORT_INITIALIZED:
+		/* In RoCE, port up/down events are handled in
+		 * mlx5_netdev_event().
+		 */
+		if (mlx5_ib_port_link_layer(&ibdev->ib_dev, port) ==
+			IB_LINK_LAYER_ETHERNET)
+			goto out;
+
+		ibev.event = (work->event == MLX5_DEV_EVENT_PORT_UP) ?
+			     IB_EVENT_PORT_ACTIVE : IB_EVENT_PORT_ERR;
+		break;
+
+	case MLX5_DEV_EVENT_LID_CHANGE:
+		ibev.event = IB_EVENT_LID_CHANGE;
+		break;
+
+	case MLX5_DEV_EVENT_PKEY_CHANGE:
+		ibev.event = IB_EVENT_PKEY_CHANGE;
+		schedule_work(&ibdev->devr.ports[port - 1].pkey_change_work);
+		break;
+
+	case MLX5_DEV_EVENT_GUID_CHANGE:
+		ibev.event = IB_EVENT_GID_CHANGE;
+		break;
+
+	case MLX5_DEV_EVENT_CLIENT_REREG:
+		ibev.event = IB_EVENT_CLIENT_REREGISTER;
+		break;
+	case MLX5_DEV_EVENT_DELAY_DROP_TIMEOUT:
+		schedule_work(&ibdev->delay_drop.delay_drop_work);
+		goto out;
+	default:
+		goto out;
+	}
+
+	ibev.device	      = &ibdev->ib_dev;
+	ibev.element.port_num = port;
+
+	if (!rdma_is_port_valid(&ibdev->ib_dev, port)) {
+		mlx5_ib_warn(ibdev, "warning: event on port %d\n", port);
+		goto out;
+	}
+
+	if (ibdev->ib_active)
+		ib_dispatch_event(&ibev);
+
+	if (fatal)
+		ibdev->ib_active = false;
+out:
+	kfree(work);
+}
+
+static void mlx5_ib_event(struct mlx5_core_dev *dev, void *context,
+			  enum mlx5_dev_event event, unsigned long param)
+{
+	struct mlx5_ib_event_work *work;
+
+	work = kmalloc(sizeof(*work), GFP_ATOMIC);
+	if (!work)
+		return;
+
+	INIT_WORK(&work->work, mlx5_ib_handle_event);
+	work->dev = dev;
+	work->param = param;
+	work->context = context;
+	work->event = event;
+
+	queue_work(mlx5_ib_event_wq, &work->work);
+}
+
+static int set_has_smi_cap(struct mlx5_ib_dev *dev)
+{
+	struct mlx5_hca_vport_context vport_ctx;
+	int err;
+	int port;
+
+	for (port = 1; port <= dev->num_ports; port++) {
+		dev->mdev->port_caps[port - 1].has_smi = false;
+		if (MLX5_CAP_GEN(dev->mdev, port_type) ==
+		    MLX5_CAP_PORT_TYPE_IB) {
+			if (MLX5_CAP_GEN(dev->mdev, ib_virt)) {
+				err = mlx5_query_hca_vport_context(dev->mdev, 0,
+								   port, 0,
+								   &vport_ctx);
+				if (err) {
+					mlx5_ib_err(dev, "query_hca_vport_context for port=%d failed %d\n",
+						    port, err);
+					return err;
+				}
+				dev->mdev->port_caps[port - 1].has_smi =
+					vport_ctx.has_smi;
+			} else {
+				dev->mdev->port_caps[port - 1].has_smi = true;
+			}
+		}
+	}
+	return 0;
+}
+
+static void get_ext_port_caps(struct mlx5_ib_dev *dev)
+{
+	int port;
+
+	for (port = 1; port <= dev->num_ports; port++)
+		mlx5_query_ext_port_caps(dev, port);
+}
+
+static int get_port_caps(struct mlx5_ib_dev *dev, u8 port)
+{
+	struct ib_device_attr *dprops = NULL;
+	struct ib_port_attr *pprops = NULL;
+	int err = -ENOMEM;
+	struct ib_udata uhw = {.inlen = 0, .outlen = 0};
+
+	pprops = kmalloc(sizeof(*pprops), GFP_KERNEL);
+	if (!pprops)
+		goto out;
+
+	dprops = kmalloc(sizeof(*dprops), GFP_KERNEL);
+	if (!dprops)
+		goto out;
+
+	err = set_has_smi_cap(dev);
+	if (err)
+		goto out;
+
+	err = mlx5_ib_query_device(&dev->ib_dev, dprops, &uhw);
+	if (err) {
+		mlx5_ib_warn(dev, "query_device failed %d\n", err);
+		goto out;
+	}
+
+	memset(pprops, 0, sizeof(*pprops));
+	err = mlx5_ib_query_port(&dev->ib_dev, port, pprops);
+	if (err) {
+		mlx5_ib_warn(dev, "query_port %d failed %d\n",
+			     port, err);
+		goto out;
+	}
+
+	dev->mdev->port_caps[port - 1].pkey_table_len =
+					dprops->max_pkeys;
+	dev->mdev->port_caps[port - 1].gid_table_len =
+					pprops->gid_tbl_len;
+	mlx5_ib_dbg(dev, "port %d: pkey_table_len %d, gid_table_len %d\n",
+		    port, dprops->max_pkeys, pprops->gid_tbl_len);
+
+out:
+	kfree(pprops);
+	kfree(dprops);
+
+	return err;
+}
+
+static void destroy_umrc_res(struct mlx5_ib_dev *dev)
+{
+	int err;
+
+	err = mlx5_mr_cache_cleanup(dev);
+	if (err)
+		mlx5_ib_warn(dev, "mr cache cleanup failed\n");
+
+	if (dev->umrc.qp)
+		mlx5_ib_destroy_qp(dev->umrc.qp);
+	if (dev->umrc.cq)
+		ib_free_cq(dev->umrc.cq);
+	if (dev->umrc.pd)
+		ib_dealloc_pd(dev->umrc.pd);
+}
+
+enum {
+	MAX_UMR_WR = 128,
+};
+
+static int create_umr_res(struct mlx5_ib_dev *dev)
+{
+	struct ib_qp_init_attr *init_attr = NULL;
+	struct ib_qp_attr *attr = NULL;
+	struct ib_pd *pd;
+	struct ib_cq *cq;
+	struct ib_qp *qp;
+	int ret;
+
+	attr = kzalloc(sizeof(*attr), GFP_KERNEL);
+	init_attr = kzalloc(sizeof(*init_attr), GFP_KERNEL);
+	if (!attr || !init_attr) {
+		ret = -ENOMEM;
+		goto error_0;
+	}
+
+	pd = ib_alloc_pd(&dev->ib_dev, 0);
+	if (IS_ERR(pd)) {
+		mlx5_ib_dbg(dev, "Couldn't create PD for sync UMR QP\n");
+		ret = PTR_ERR(pd);
+		goto error_0;
+	}
+
+	cq = ib_alloc_cq(&dev->ib_dev, NULL, 128, 0, IB_POLL_SOFTIRQ);
+	if (IS_ERR(cq)) {
+		mlx5_ib_dbg(dev, "Couldn't create CQ for sync UMR QP\n");
+		ret = PTR_ERR(cq);
+		goto error_2;
+	}
+
+	init_attr->send_cq = cq;
+	init_attr->recv_cq = cq;
+	init_attr->sq_sig_type = IB_SIGNAL_ALL_WR;
+	init_attr->cap.max_send_wr = MAX_UMR_WR;
+	init_attr->cap.max_send_sge = 1;
+	init_attr->qp_type = MLX5_IB_QPT_REG_UMR;
+	init_attr->port_num = 1;
+	qp = mlx5_ib_create_qp(pd, init_attr, NULL);
+	if (IS_ERR(qp)) {
+		mlx5_ib_dbg(dev, "Couldn't create sync UMR QP\n");
+		ret = PTR_ERR(qp);
+		goto error_3;
+	}
+	qp->device     = &dev->ib_dev;
+	qp->real_qp    = qp;
+	qp->uobject    = NULL;
+	qp->qp_type    = MLX5_IB_QPT_REG_UMR;
+	qp->send_cq    = init_attr->send_cq;
+	qp->recv_cq    = init_attr->recv_cq;
+
+	attr->qp_state = IB_QPS_INIT;
+	attr->port_num = 1;
+	ret = mlx5_ib_modify_qp(qp, attr, IB_QP_STATE | IB_QP_PKEY_INDEX |
+				IB_QP_PORT, NULL);
+	if (ret) {
+		mlx5_ib_dbg(dev, "Couldn't modify UMR QP\n");
+		goto error_4;
+	}
+
+	memset(attr, 0, sizeof(*attr));
+	attr->qp_state = IB_QPS_RTR;
+	attr->path_mtu = IB_MTU_256;
+
+	ret = mlx5_ib_modify_qp(qp, attr, IB_QP_STATE, NULL);
+	if (ret) {
+		mlx5_ib_dbg(dev, "Couldn't modify umr QP to rtr\n");
+		goto error_4;
+	}
+
+	memset(attr, 0, sizeof(*attr));
+	attr->qp_state = IB_QPS_RTS;
+	ret = mlx5_ib_modify_qp(qp, attr, IB_QP_STATE, NULL);
+	if (ret) {
+		mlx5_ib_dbg(dev, "Couldn't modify umr QP to rts\n");
+		goto error_4;
+	}
+
+	dev->umrc.qp = qp;
+	dev->umrc.cq = cq;
+	dev->umrc.pd = pd;
+
+	sema_init(&dev->umrc.sem, MAX_UMR_WR);
+	ret = mlx5_mr_cache_init(dev);
+	if (ret) {
+		mlx5_ib_warn(dev, "mr cache init failed %d\n", ret);
+		goto error_4;
+	}
+
+	kfree(attr);
+	kfree(init_attr);
+
+	return 0;
+
+error_4:
+	mlx5_ib_destroy_qp(qp);
+	dev->umrc.qp = NULL;
+
+error_3:
+	ib_free_cq(cq);
+	dev->umrc.cq = NULL;
+
+error_2:
+	ib_dealloc_pd(pd);
+	dev->umrc.pd = NULL;
+
+error_0:
+	kfree(attr);
+	kfree(init_attr);
+	return ret;
+}
+
+static u8 mlx5_get_umr_fence(u8 umr_fence_cap)
+{
+	switch (umr_fence_cap) {
+	case MLX5_CAP_UMR_FENCE_NONE:
+		return MLX5_FENCE_MODE_NONE;
+	case MLX5_CAP_UMR_FENCE_SMALL:
+		return MLX5_FENCE_MODE_INITIATOR_SMALL;
+	default:
+		return MLX5_FENCE_MODE_STRONG_ORDERING;
+	}
+}
+
+static int create_dev_resources(struct mlx5_ib_resources *devr)
+{
+	struct ib_srq_init_attr attr;
+	struct mlx5_ib_dev *dev;
+	struct ib_cq_init_attr cq_attr = {.cqe = 1};
+	int port;
+	int ret = 0;
+
+	dev = container_of(devr, struct mlx5_ib_dev, devr);
+
+	mutex_init(&devr->mutex);
+
+	devr->p0 = mlx5_ib_alloc_pd(&dev->ib_dev, NULL, NULL);
+	if (IS_ERR(devr->p0)) {
+		ret = PTR_ERR(devr->p0);
+		goto error0;
+	}
+	devr->p0->device  = &dev->ib_dev;
+	devr->p0->uobject = NULL;
+	atomic_set(&devr->p0->usecnt, 0);
+
+	devr->c0 = mlx5_ib_create_cq(&dev->ib_dev, &cq_attr, NULL, NULL);
+	if (IS_ERR(devr->c0)) {
+		ret = PTR_ERR(devr->c0);
+		goto error1;
+	}
+	devr->c0->device        = &dev->ib_dev;
+	devr->c0->uobject       = NULL;
+	devr->c0->comp_handler  = NULL;
+	devr->c0->event_handler = NULL;
+	devr->c0->cq_context    = NULL;
+	atomic_set(&devr->c0->usecnt, 0);
+
+	devr->x0 = mlx5_ib_alloc_xrcd(&dev->ib_dev, NULL, NULL);
+	if (IS_ERR(devr->x0)) {
+		ret = PTR_ERR(devr->x0);
+		goto error2;
+	}
+	devr->x0->device = &dev->ib_dev;
+	devr->x0->inode = NULL;
+	atomic_set(&devr->x0->usecnt, 0);
+	mutex_init(&devr->x0->tgt_qp_mutex);
+	INIT_LIST_HEAD(&devr->x0->tgt_qp_list);
+
+	devr->x1 = mlx5_ib_alloc_xrcd(&dev->ib_dev, NULL, NULL);
+	if (IS_ERR(devr->x1)) {
+		ret = PTR_ERR(devr->x1);
+		goto error3;
+	}
+	devr->x1->device = &dev->ib_dev;
+	devr->x1->inode = NULL;
+	atomic_set(&devr->x1->usecnt, 0);
+	mutex_init(&devr->x1->tgt_qp_mutex);
+	INIT_LIST_HEAD(&devr->x1->tgt_qp_list);
+
+	memset(&attr, 0, sizeof(attr));
+	attr.attr.max_sge = 1;
+	attr.attr.max_wr = 1;
+	attr.srq_type = IB_SRQT_XRC;
+	attr.ext.cq = devr->c0;
+	attr.ext.xrc.xrcd = devr->x0;
+
+	devr->s0 = mlx5_ib_create_srq(devr->p0, &attr, NULL);
+	if (IS_ERR(devr->s0)) {
+		ret = PTR_ERR(devr->s0);
+		goto error4;
+	}
+	devr->s0->device	= &dev->ib_dev;
+	devr->s0->pd		= devr->p0;
+	devr->s0->uobject       = NULL;
+	devr->s0->event_handler = NULL;
+	devr->s0->srq_context   = NULL;
+	devr->s0->srq_type      = IB_SRQT_XRC;
+	devr->s0->ext.xrc.xrcd	= devr->x0;
+	devr->s0->ext.cq	= devr->c0;
+	atomic_inc(&devr->s0->ext.xrc.xrcd->usecnt);
+	atomic_inc(&devr->s0->ext.cq->usecnt);
+	atomic_inc(&devr->p0->usecnt);
+	atomic_set(&devr->s0->usecnt, 0);
+
+	memset(&attr, 0, sizeof(attr));
+	attr.attr.max_sge = 1;
+	attr.attr.max_wr = 1;
+	attr.srq_type = IB_SRQT_BASIC;
+	devr->s1 = mlx5_ib_create_srq(devr->p0, &attr, NULL);
+	if (IS_ERR(devr->s1)) {
+		ret = PTR_ERR(devr->s1);
+		goto error5;
+	}
+	devr->s1->device	= &dev->ib_dev;
+	devr->s1->pd		= devr->p0;
+	devr->s1->uobject       = NULL;
+	devr->s1->event_handler = NULL;
+	devr->s1->srq_context   = NULL;
+	devr->s1->srq_type      = IB_SRQT_BASIC;
+	devr->s1->ext.cq	= devr->c0;
+	atomic_inc(&devr->p0->usecnt);
+	atomic_set(&devr->s1->usecnt, 0);
+
+	for (port = 0; port < ARRAY_SIZE(devr->ports); ++port) {
+		INIT_WORK(&devr->ports[port].pkey_change_work,
+			  pkey_change_handler);
+		devr->ports[port].devr = devr;
+	}
+
+	return 0;
+
+error5:
+	mlx5_ib_destroy_srq(devr->s0);
+error4:
+	mlx5_ib_dealloc_xrcd(devr->x1);
+error3:
+	mlx5_ib_dealloc_xrcd(devr->x0);
+error2:
+	mlx5_ib_destroy_cq(devr->c0);
+error1:
+	mlx5_ib_dealloc_pd(devr->p0);
+error0:
+	return ret;
+}
+
+static void destroy_dev_resources(struct mlx5_ib_resources *devr)
+{
+	struct mlx5_ib_dev *dev =
+		container_of(devr, struct mlx5_ib_dev, devr);
+	int port;
+
+	mlx5_ib_destroy_srq(devr->s1);
+	mlx5_ib_destroy_srq(devr->s0);
+	mlx5_ib_dealloc_xrcd(devr->x0);
+	mlx5_ib_dealloc_xrcd(devr->x1);
+	mlx5_ib_destroy_cq(devr->c0);
+	mlx5_ib_dealloc_pd(devr->p0);
+
+	/* Make sure no change P_Key work items are still executing */
+	for (port = 0; port < dev->num_ports; ++port)
+		cancel_work_sync(&devr->ports[port].pkey_change_work);
+}
+
+static u32 get_core_cap_flags(struct ib_device *ibdev,
+			      struct mlx5_hca_vport_context *rep)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+	enum rdma_link_layer ll = mlx5_ib_port_link_layer(ibdev, 1);
+	u8 l3_type_cap = MLX5_CAP_ROCE(dev->mdev, l3_type);
+	u8 roce_version_cap = MLX5_CAP_ROCE(dev->mdev, roce_version);
+	bool raw_support = !mlx5_core_mp_enabled(dev->mdev);
+	u32 ret = 0;
+
+	if (rep->grh_required)
+		ret |= RDMA_CORE_CAP_IB_GRH_REQUIRED;
+
+	if (ll == IB_LINK_LAYER_INFINIBAND)
+		return ret | RDMA_CORE_PORT_IBA_IB;
+
+	if (raw_support)
+		ret |= RDMA_CORE_PORT_RAW_PACKET;
+
+	if (!(l3_type_cap & MLX5_ROCE_L3_TYPE_IPV4_CAP))
+		return ret;
+
+	if (!(l3_type_cap & MLX5_ROCE_L3_TYPE_IPV6_CAP))
+		return ret;
+
+	if (roce_version_cap & MLX5_ROCE_VERSION_1_CAP)
+		ret |= RDMA_CORE_PORT_IBA_ROCE;
+
+	if (roce_version_cap & MLX5_ROCE_VERSION_2_CAP)
+		ret |= RDMA_CORE_PORT_IBA_ROCE_UDP_ENCAP;
+
+	return ret;
+}
+
+static int mlx5_port_immutable(struct ib_device *ibdev, u8 port_num,
+			       struct ib_port_immutable *immutable)
+{
+	struct ib_port_attr attr;
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+	enum rdma_link_layer ll = mlx5_ib_port_link_layer(ibdev, port_num);
+	struct mlx5_hca_vport_context rep = {0};
+	int err;
+
+	err = ib_query_port(ibdev, port_num, &attr);
+	if (err)
+		return err;
+
+	if (ll == IB_LINK_LAYER_INFINIBAND) {
+		err = mlx5_query_hca_vport_context(dev->mdev, 0, port_num, 0,
+						   &rep);
+		if (err)
+			return err;
+	}
+
+	immutable->pkey_tbl_len = attr.pkey_tbl_len;
+	immutable->gid_tbl_len = attr.gid_tbl_len;
+	immutable->core_cap_flags = get_core_cap_flags(ibdev, &rep);
+	if ((ll == IB_LINK_LAYER_INFINIBAND) || MLX5_CAP_GEN(dev->mdev, roce))
+		immutable->max_mad_size = IB_MGMT_MAD_SIZE;
+
+	return 0;
+}
+
+static int mlx5_port_rep_immutable(struct ib_device *ibdev, u8 port_num,
+				   struct ib_port_immutable *immutable)
+{
+	struct ib_port_attr attr;
+	int err;
+
+	immutable->core_cap_flags = RDMA_CORE_PORT_RAW_PACKET;
+
+	err = ib_query_port(ibdev, port_num, &attr);
+	if (err)
+		return err;
+
+	immutable->pkey_tbl_len = attr.pkey_tbl_len;
+	immutable->gid_tbl_len = attr.gid_tbl_len;
+	immutable->core_cap_flags = RDMA_CORE_PORT_RAW_PACKET;
+
+	return 0;
+}
+
+static void get_dev_fw_str(struct ib_device *ibdev, char *str)
+{
+	struct mlx5_ib_dev *dev =
+		container_of(ibdev, struct mlx5_ib_dev, ib_dev);
+	snprintf(str, IB_FW_VERSION_NAME_MAX, "%d.%d.%04d",
+		 fw_rev_maj(dev->mdev), fw_rev_min(dev->mdev),
+		 fw_rev_sub(dev->mdev));
+}
+
+static int mlx5_eth_lag_init(struct mlx5_ib_dev *dev)
+{
+	struct mlx5_core_dev *mdev = dev->mdev;
+	struct mlx5_flow_namespace *ns = mlx5_get_flow_namespace(mdev,
+								 MLX5_FLOW_NAMESPACE_LAG);
+	struct mlx5_flow_table *ft;
+	int err;
+
+	if (!ns || !mlx5_lag_is_active(mdev))
+		return 0;
+
+	err = mlx5_cmd_create_vport_lag(mdev);
+	if (err)
+		return err;
+
+	ft = mlx5_create_lag_demux_flow_table(ns, 0, 0);
+	if (IS_ERR(ft)) {
+		err = PTR_ERR(ft);
+		goto err_destroy_vport_lag;
+	}
+
+	dev->flow_db->lag_demux_ft = ft;
+	return 0;
+
+err_destroy_vport_lag:
+	mlx5_cmd_destroy_vport_lag(mdev);
+	return err;
+}
+
+static void mlx5_eth_lag_cleanup(struct mlx5_ib_dev *dev)
+{
+	struct mlx5_core_dev *mdev = dev->mdev;
+
+	if (dev->flow_db->lag_demux_ft) {
+		mlx5_destroy_flow_table(dev->flow_db->lag_demux_ft);
+		dev->flow_db->lag_demux_ft = NULL;
+
+		mlx5_cmd_destroy_vport_lag(mdev);
+	}
+}
+
+static int mlx5_add_netdev_notifier(struct mlx5_ib_dev *dev, u8 port_num)
+{
+	int err;
+
+	dev->roce[port_num].nb.notifier_call = mlx5_netdev_event;
+	err = register_netdevice_notifier(&dev->roce[port_num].nb);
+	if (err) {
+		dev->roce[port_num].nb.notifier_call = NULL;
+		return err;
+	}
+
+	return 0;
+}
+
+static void mlx5_remove_netdev_notifier(struct mlx5_ib_dev *dev, u8 port_num)
+{
+	if (dev->roce[port_num].nb.notifier_call) {
+		unregister_netdevice_notifier(&dev->roce[port_num].nb);
+		dev->roce[port_num].nb.notifier_call = NULL;
+	}
+}
+
+static int mlx5_enable_eth(struct mlx5_ib_dev *dev)
+{
+	int err;
+
+	if (MLX5_CAP_GEN(dev->mdev, roce)) {
+		err = mlx5_nic_vport_enable_roce(dev->mdev);
+		if (err)
+			return err;
+	}
+
+	err = mlx5_eth_lag_init(dev);
+	if (err)
+		goto err_disable_roce;
+
+	return 0;
+
+err_disable_roce:
+	if (MLX5_CAP_GEN(dev->mdev, roce))
+		mlx5_nic_vport_disable_roce(dev->mdev);
+
+	return err;
+}
+
+static void mlx5_disable_eth(struct mlx5_ib_dev *dev)
+{
+	mlx5_eth_lag_cleanup(dev);
+	if (MLX5_CAP_GEN(dev->mdev, roce))
+		mlx5_nic_vport_disable_roce(dev->mdev);
+}
+
+struct mlx5_ib_counter {
+	const char *name;
+	size_t offset;
+};
+
+#define INIT_Q_COUNTER(_name)		\
+	{ .name = #_name, .offset = MLX5_BYTE_OFF(query_q_counter_out, _name)}
+
+static const struct mlx5_ib_counter basic_q_cnts[] = {
+	INIT_Q_COUNTER(rx_write_requests),
+	INIT_Q_COUNTER(rx_read_requests),
+	INIT_Q_COUNTER(rx_atomic_requests),
+	INIT_Q_COUNTER(out_of_buffer),
+};
+
+static const struct mlx5_ib_counter out_of_seq_q_cnts[] = {
+	INIT_Q_COUNTER(out_of_sequence),
+};
+
+static const struct mlx5_ib_counter retrans_q_cnts[] = {
+	INIT_Q_COUNTER(duplicate_request),
+	INIT_Q_COUNTER(rnr_nak_retry_err),
+	INIT_Q_COUNTER(packet_seq_err),
+	INIT_Q_COUNTER(implied_nak_seq_err),
+	INIT_Q_COUNTER(local_ack_timeout_err),
+};
+
+#define INIT_CONG_COUNTER(_name)		\
+	{ .name = #_name, .offset =	\
+		MLX5_BYTE_OFF(query_cong_statistics_out, _name ## _high)}
+
+static const struct mlx5_ib_counter cong_cnts[] = {
+	INIT_CONG_COUNTER(rp_cnp_ignored),
+	INIT_CONG_COUNTER(rp_cnp_handled),
+	INIT_CONG_COUNTER(np_ecn_marked_roce_packets),
+	INIT_CONG_COUNTER(np_cnp_sent),
+};
+
+static const struct mlx5_ib_counter extended_err_cnts[] = {
+	INIT_Q_COUNTER(resp_local_length_error),
+	INIT_Q_COUNTER(resp_cqe_error),
+	INIT_Q_COUNTER(req_cqe_error),
+	INIT_Q_COUNTER(req_remote_invalid_request),
+	INIT_Q_COUNTER(req_remote_access_errors),
+	INIT_Q_COUNTER(resp_remote_access_errors),
+	INIT_Q_COUNTER(resp_cqe_flush_error),
+	INIT_Q_COUNTER(req_cqe_flush_error),
+};
+
+#define INIT_EXT_PPCNT_COUNTER(_name)		\
+	{ .name = #_name, .offset =	\
+	MLX5_BYTE_OFF(ppcnt_reg, \
+		      counter_set.eth_extended_cntrs_grp_data_layout._name##_high)}
+
+static const struct mlx5_ib_counter ext_ppcnt_cnts[] = {
+	INIT_EXT_PPCNT_COUNTER(rx_icrc_encapsulated),
+};
+
+static void mlx5_ib_dealloc_counters(struct mlx5_ib_dev *dev)
+{
+	int i;
+
+	for (i = 0; i < dev->num_ports; i++) {
+		if (dev->port[i].cnts.set_id_valid)
+			mlx5_core_dealloc_q_counter(dev->mdev,
+						    dev->port[i].cnts.set_id);
+		kfree(dev->port[i].cnts.names);
+		kfree(dev->port[i].cnts.offsets);
+	}
+}
+
+static int __mlx5_ib_alloc_counters(struct mlx5_ib_dev *dev,
+				    struct mlx5_ib_counters *cnts)
+{
+	u32 num_counters;
+
+	num_counters = ARRAY_SIZE(basic_q_cnts);
+
+	if (MLX5_CAP_GEN(dev->mdev, out_of_seq_cnt))
+		num_counters += ARRAY_SIZE(out_of_seq_q_cnts);
+
+	if (MLX5_CAP_GEN(dev->mdev, retransmission_q_counters))
+		num_counters += ARRAY_SIZE(retrans_q_cnts);
+
+	if (MLX5_CAP_GEN(dev->mdev, enhanced_error_q_counters))
+		num_counters += ARRAY_SIZE(extended_err_cnts);
+
+	cnts->num_q_counters = num_counters;
+
+	if (MLX5_CAP_GEN(dev->mdev, cc_query_allowed)) {
+		cnts->num_cong_counters = ARRAY_SIZE(cong_cnts);
+		num_counters += ARRAY_SIZE(cong_cnts);
+	}
+	if (MLX5_CAP_PCAM_FEATURE(dev->mdev, rx_icrc_encapsulated_counter)) {
+		cnts->num_ext_ppcnt_counters = ARRAY_SIZE(ext_ppcnt_cnts);
+		num_counters += ARRAY_SIZE(ext_ppcnt_cnts);
+	}
+	cnts->names = kcalloc(num_counters, sizeof(cnts->names), GFP_KERNEL);
+	if (!cnts->names)
+		return -ENOMEM;
+
+	cnts->offsets = kcalloc(num_counters,
+				sizeof(cnts->offsets), GFP_KERNEL);
+	if (!cnts->offsets)
+		goto err_names;
+
+	return 0;
+
+err_names:
+	kfree(cnts->names);
+	cnts->names = NULL;
+	return -ENOMEM;
+}
+
+static void mlx5_ib_fill_counters(struct mlx5_ib_dev *dev,
+				  const char **names,
+				  size_t *offsets)
+{
+	int i;
+	int j = 0;
+
+	for (i = 0; i < ARRAY_SIZE(basic_q_cnts); i++, j++) {
+		names[j] = basic_q_cnts[i].name;
+		offsets[j] = basic_q_cnts[i].offset;
+	}
+
+	if (MLX5_CAP_GEN(dev->mdev, out_of_seq_cnt)) {
+		for (i = 0; i < ARRAY_SIZE(out_of_seq_q_cnts); i++, j++) {
+			names[j] = out_of_seq_q_cnts[i].name;
+			offsets[j] = out_of_seq_q_cnts[i].offset;
+		}
+	}
+
+	if (MLX5_CAP_GEN(dev->mdev, retransmission_q_counters)) {
+		for (i = 0; i < ARRAY_SIZE(retrans_q_cnts); i++, j++) {
+			names[j] = retrans_q_cnts[i].name;
+			offsets[j] = retrans_q_cnts[i].offset;
+		}
+	}
+
+	if (MLX5_CAP_GEN(dev->mdev, enhanced_error_q_counters)) {
+		for (i = 0; i < ARRAY_SIZE(extended_err_cnts); i++, j++) {
+			names[j] = extended_err_cnts[i].name;
+			offsets[j] = extended_err_cnts[i].offset;
+		}
+	}
+
+	if (MLX5_CAP_GEN(dev->mdev, cc_query_allowed)) {
+		for (i = 0; i < ARRAY_SIZE(cong_cnts); i++, j++) {
+			names[j] = cong_cnts[i].name;
+			offsets[j] = cong_cnts[i].offset;
+		}
+	}
+
+	if (MLX5_CAP_PCAM_FEATURE(dev->mdev, rx_icrc_encapsulated_counter)) {
+		for (i = 0; i < ARRAY_SIZE(ext_ppcnt_cnts); i++, j++) {
+			names[j] = ext_ppcnt_cnts[i].name;
+			offsets[j] = ext_ppcnt_cnts[i].offset;
+		}
+	}
+}
+
+static int mlx5_ib_alloc_counters(struct mlx5_ib_dev *dev)
+{
+	int err = 0;
+	int i;
+
+	for (i = 0; i < dev->num_ports; i++) {
+		err = __mlx5_ib_alloc_counters(dev, &dev->port[i].cnts);
+		if (err)
+			goto err_alloc;
+
+		mlx5_ib_fill_counters(dev, dev->port[i].cnts.names,
+				      dev->port[i].cnts.offsets);
+
+		err = mlx5_core_alloc_q_counter(dev->mdev,
+						&dev->port[i].cnts.set_id);
+		if (err) {
+			mlx5_ib_warn(dev,
+				     "couldn't allocate queue counter for port %d, err %d\n",
+				     i + 1, err);
+			goto err_alloc;
+		}
+		dev->port[i].cnts.set_id_valid = true;
+	}
+
+	return 0;
+
+err_alloc:
+	mlx5_ib_dealloc_counters(dev);
+	return err;
+}
+
+static struct rdma_hw_stats *mlx5_ib_alloc_hw_stats(struct ib_device *ibdev,
+						    u8 port_num)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+	struct mlx5_ib_port *port = &dev->port[port_num - 1];
+
+	/* We support only per port stats */
+	if (port_num == 0)
+		return NULL;
+
+	return rdma_alloc_hw_stats_struct(port->cnts.names,
+					  port->cnts.num_q_counters +
+					  port->cnts.num_cong_counters +
+					  port->cnts.num_ext_ppcnt_counters,
+					  RDMA_HW_STATS_DEFAULT_LIFESPAN);
+}
+
+static int mlx5_ib_query_q_counters(struct mlx5_core_dev *mdev,
+				    struct mlx5_ib_port *port,
+				    struct rdma_hw_stats *stats)
+{
+	int outlen = MLX5_ST_SZ_BYTES(query_q_counter_out);
+	void *out;
+	__be32 val;
+	int ret, i;
+
+	out = kvzalloc(outlen, GFP_KERNEL);
+	if (!out)
+		return -ENOMEM;
+
+	ret = mlx5_core_query_q_counter(mdev,
+					port->cnts.set_id, 0,
+					out, outlen);
+	if (ret)
+		goto free;
+
+	for (i = 0; i < port->cnts.num_q_counters; i++) {
+		val = *(__be32 *)(out + port->cnts.offsets[i]);
+		stats->value[i] = (u64)be32_to_cpu(val);
+	}
+
+free:
+	kvfree(out);
+	return ret;
+}
+
+static int mlx5_ib_query_ext_ppcnt_counters(struct mlx5_ib_dev *dev,
+					  struct mlx5_ib_port *port,
+					  struct rdma_hw_stats *stats)
+{
+	int offset = port->cnts.num_q_counters + port->cnts.num_cong_counters;
+	int sz = MLX5_ST_SZ_BYTES(ppcnt_reg);
+	int ret, i;
+	void *out;
+
+	out = kvzalloc(sz, GFP_KERNEL);
+	if (!out)
+		return -ENOMEM;
+
+	ret = mlx5_cmd_query_ext_ppcnt_counters(dev->mdev, out);
+	if (ret)
+		goto free;
+
+	for (i = 0; i < port->cnts.num_ext_ppcnt_counters; i++) {
+		stats->value[i + offset] =
+			be64_to_cpup((__be64 *)(out +
+				    port->cnts.offsets[i + offset]));
+	}
+
+free:
+	kvfree(out);
+	return ret;
+}
+
+static int mlx5_ib_get_hw_stats(struct ib_device *ibdev,
+				struct rdma_hw_stats *stats,
+				u8 port_num, int index)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+	struct mlx5_ib_port *port = &dev->port[port_num - 1];
+	struct mlx5_core_dev *mdev;
+	int ret, num_counters;
+	u8 mdev_port_num;
+
+	if (!stats)
+		return -EINVAL;
+
+	num_counters = port->cnts.num_q_counters +
+		       port->cnts.num_cong_counters +
+		       port->cnts.num_ext_ppcnt_counters;
+
+	/* q_counters are per IB device, query the master mdev */
+	ret = mlx5_ib_query_q_counters(dev->mdev, port, stats);
+	if (ret)
+		return ret;
+
+	if (MLX5_CAP_PCAM_FEATURE(dev->mdev, rx_icrc_encapsulated_counter)) {
+		ret =  mlx5_ib_query_ext_ppcnt_counters(dev, port, stats);
+		if (ret)
+			return ret;
+	}
+
+	if (MLX5_CAP_GEN(dev->mdev, cc_query_allowed)) {
+		mdev = mlx5_ib_get_native_port_mdev(dev, port_num,
+						    &mdev_port_num);
+		if (!mdev) {
+			/* If port is not affiliated yet, its in down state
+			 * which doesn't have any counters yet, so it would be
+			 * zero. So no need to read from the HCA.
+			 */
+			goto done;
+		}
+		ret = mlx5_lag_query_cong_counters(dev->mdev,
+						   stats->value +
+						   port->cnts.num_q_counters,
+						   port->cnts.num_cong_counters,
+						   port->cnts.offsets +
+						   port->cnts.num_q_counters);
+
+		mlx5_ib_put_native_port_mdev(dev, port_num);
+		if (ret)
+			return ret;
+	}
+
+done:
+	return num_counters;
+}
+
+static struct net_device*
+mlx5_ib_alloc_rdma_netdev(struct ib_device *hca,
+			  u8 port_num,
+			  enum rdma_netdev_t type,
+			  const char *name,
+			  unsigned char name_assign_type,
+			  void (*setup)(struct net_device *))
+{
+	struct net_device *netdev;
+
+	if (type != RDMA_NETDEV_IPOIB)
+		return ERR_PTR(-EOPNOTSUPP);
+
+	netdev = mlx5_rdma_netdev_alloc(to_mdev(hca)->mdev, hca,
+					name, setup);
+	return netdev;
+}
+
+static void delay_drop_debugfs_cleanup(struct mlx5_ib_dev *dev)
+{
+	if (!dev->delay_drop.dbg)
+		return;
+	debugfs_remove_recursive(dev->delay_drop.dbg->dir_debugfs);
+	kfree(dev->delay_drop.dbg);
+	dev->delay_drop.dbg = NULL;
+}
+
+static void cancel_delay_drop(struct mlx5_ib_dev *dev)
+{
+	if (!(dev->ib_dev.attrs.raw_packet_caps & IB_RAW_PACKET_CAP_DELAY_DROP))
+		return;
+
+	cancel_work_sync(&dev->delay_drop.delay_drop_work);
+	delay_drop_debugfs_cleanup(dev);
+}
+
+static ssize_t delay_drop_timeout_read(struct file *filp, char __user *buf,
+				       size_t count, loff_t *pos)
+{
+	struct mlx5_ib_delay_drop *delay_drop = filp->private_data;
+	char lbuf[20];
+	int len;
+
+	len = snprintf(lbuf, sizeof(lbuf), "%u\n", delay_drop->timeout);
+	return simple_read_from_buffer(buf, count, pos, lbuf, len);
+}
+
+static ssize_t delay_drop_timeout_write(struct file *filp, const char __user *buf,
+					size_t count, loff_t *pos)
+{
+	struct mlx5_ib_delay_drop *delay_drop = filp->private_data;
+	u32 timeout;
+	u32 var;
+
+	if (kstrtouint_from_user(buf, count, 0, &var))
+		return -EFAULT;
+
+	timeout = min_t(u32, roundup(var, 100), MLX5_MAX_DELAY_DROP_TIMEOUT_MS *
+			1000);
+	if (timeout != var)
+		mlx5_ib_dbg(delay_drop->dev, "Round delay drop timeout to %u usec\n",
+			    timeout);
+
+	delay_drop->timeout = timeout;
+
+	return count;
+}
+
+static const struct file_operations fops_delay_drop_timeout = {
+	.owner	= THIS_MODULE,
+	.open	= simple_open,
+	.write	= delay_drop_timeout_write,
+	.read	= delay_drop_timeout_read,
+};
+
+static int delay_drop_debugfs_init(struct mlx5_ib_dev *dev)
+{
+	struct mlx5_ib_dbg_delay_drop *dbg;
+
+	if (!mlx5_debugfs_root)
+		return 0;
+
+	dbg = kzalloc(sizeof(*dbg), GFP_KERNEL);
+	if (!dbg)
+		return -ENOMEM;
+
+	dev->delay_drop.dbg = dbg;
+
+	dbg->dir_debugfs =
+		debugfs_create_dir("delay_drop",
+				   dev->mdev->priv.dbg_root);
+	if (!dbg->dir_debugfs)
+		goto out_debugfs;
+
+	dbg->events_cnt_debugfs =
+		debugfs_create_atomic_t("num_timeout_events", 0400,
+					dbg->dir_debugfs,
+					&dev->delay_drop.events_cnt);
+	if (!dbg->events_cnt_debugfs)
+		goto out_debugfs;
+
+	dbg->rqs_cnt_debugfs =
+		debugfs_create_atomic_t("num_rqs", 0400,
+					dbg->dir_debugfs,
+					&dev->delay_drop.rqs_cnt);
+	if (!dbg->rqs_cnt_debugfs)
+		goto out_debugfs;
+
+	dbg->timeout_debugfs =
+		debugfs_create_file("timeout", 0600,
+				    dbg->dir_debugfs,
+				    &dev->delay_drop,
+				    &fops_delay_drop_timeout);
+	if (!dbg->timeout_debugfs)
+		goto out_debugfs;
+
+	return 0;
+
+out_debugfs:
+	delay_drop_debugfs_cleanup(dev);
+	return -ENOMEM;
+}
+
+static void init_delay_drop(struct mlx5_ib_dev *dev)
+{
+	if (!(dev->ib_dev.attrs.raw_packet_caps & IB_RAW_PACKET_CAP_DELAY_DROP))
+		return;
+
+	mutex_init(&dev->delay_drop.lock);
+	dev->delay_drop.dev = dev;
+	dev->delay_drop.activate = false;
+	dev->delay_drop.timeout = MLX5_MAX_DELAY_DROP_TIMEOUT_MS * 1000;
+	INIT_WORK(&dev->delay_drop.delay_drop_work, delay_drop_handler);
+	atomic_set(&dev->delay_drop.rqs_cnt, 0);
+	atomic_set(&dev->delay_drop.events_cnt, 0);
+
+	if (delay_drop_debugfs_init(dev))
+		mlx5_ib_warn(dev, "Failed to init delay drop debugfs\n");
+}
+
+static const struct cpumask *
+mlx5_ib_get_vector_affinity(struct ib_device *ibdev, int comp_vector)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+
+	return mlx5_get_vector_affinity_hint(dev->mdev, comp_vector);
+}
+
+/* The mlx5_ib_multiport_mutex should be held when calling this function */
+static void mlx5_ib_unbind_slave_port(struct mlx5_ib_dev *ibdev,
+				      struct mlx5_ib_multiport_info *mpi)
+{
+	u8 port_num = mlx5_core_native_port_num(mpi->mdev) - 1;
+	struct mlx5_ib_port *port = &ibdev->port[port_num];
+	int comps;
+	int err;
+	int i;
+
+	mlx5_ib_cleanup_cong_debugfs(ibdev, port_num);
+
+	spin_lock(&port->mp.mpi_lock);
+	if (!mpi->ibdev) {
+		spin_unlock(&port->mp.mpi_lock);
+		return;
+	}
+	mpi->ibdev = NULL;
+
+	spin_unlock(&port->mp.mpi_lock);
+	mlx5_remove_netdev_notifier(ibdev, port_num);
+	spin_lock(&port->mp.mpi_lock);
+
+	comps = mpi->mdev_refcnt;
+	if (comps) {
+		mpi->unaffiliate = true;
+		init_completion(&mpi->unref_comp);
+		spin_unlock(&port->mp.mpi_lock);
+
+		for (i = 0; i < comps; i++)
+			wait_for_completion(&mpi->unref_comp);
+
+		spin_lock(&port->mp.mpi_lock);
+		mpi->unaffiliate = false;
+	}
+
+	port->mp.mpi = NULL;
+
+	spin_unlock(&port->mp.mpi_lock);
+
+	err = mlx5_nic_vport_unaffiliate_multiport(mpi->mdev);
+
+	mlx5_ib_dbg(ibdev, "unaffiliated port %d\n", port_num + 1);
+	/* Log an error, still needed to cleanup the pointers and add
+	 * it back to the list.
+	 */
+	if (err)
+		mlx5_ib_err(ibdev, "Failed to unaffiliate port %u\n",
+			    port_num + 1);
+
+	ibdev->roce[port_num].last_port_state = IB_PORT_DOWN;
+}
+
+/* The mlx5_ib_multiport_mutex should be held when calling this function */
+static bool mlx5_ib_bind_slave_port(struct mlx5_ib_dev *ibdev,
+				    struct mlx5_ib_multiport_info *mpi)
+{
+	u8 port_num = mlx5_core_native_port_num(mpi->mdev) - 1;
+	int err;
+
+	spin_lock(&ibdev->port[port_num].mp.mpi_lock);
+	if (ibdev->port[port_num].mp.mpi) {
+		mlx5_ib_dbg(ibdev, "port %d already affiliated.\n",
+			    port_num + 1);
+		spin_unlock(&ibdev->port[port_num].mp.mpi_lock);
+		return false;
+	}
+
+	ibdev->port[port_num].mp.mpi = mpi;
+	mpi->ibdev = ibdev;
+	spin_unlock(&ibdev->port[port_num].mp.mpi_lock);
+
+	err = mlx5_nic_vport_affiliate_multiport(ibdev->mdev, mpi->mdev);
+	if (err)
+		goto unbind;
+
+	err = get_port_caps(ibdev, mlx5_core_native_port_num(mpi->mdev));
+	if (err)
+		goto unbind;
+
+	err = mlx5_add_netdev_notifier(ibdev, port_num);
+	if (err) {
+		mlx5_ib_err(ibdev, "failed adding netdev notifier for port %u\n",
+			    port_num + 1);
+		goto unbind;
+	}
+
+	err = mlx5_ib_init_cong_debugfs(ibdev, port_num);
+	if (err)
+		goto unbind;
+
+	return true;
+
+unbind:
+	mlx5_ib_unbind_slave_port(ibdev, mpi);
+	return false;
+}
+
+static int mlx5_ib_init_multiport_master(struct mlx5_ib_dev *dev)
+{
+	int port_num = mlx5_core_native_port_num(dev->mdev) - 1;
+	enum rdma_link_layer ll = mlx5_ib_port_link_layer(&dev->ib_dev,
+							  port_num + 1);
+	struct mlx5_ib_multiport_info *mpi;
+	int err;
+	int i;
+
+	if (!mlx5_core_is_mp_master(dev->mdev) || ll != IB_LINK_LAYER_ETHERNET)
+		return 0;
+
+	err = mlx5_query_nic_vport_system_image_guid(dev->mdev,
+						     &dev->sys_image_guid);
+	if (err)
+		return err;
+
+	err = mlx5_nic_vport_enable_roce(dev->mdev);
+	if (err)
+		return err;
+
+	mutex_lock(&mlx5_ib_multiport_mutex);
+	for (i = 0; i < dev->num_ports; i++) {
+		bool bound = false;
+
+		/* build a stub multiport info struct for the native port. */
+		if (i == port_num) {
+			mpi = kzalloc(sizeof(*mpi), GFP_KERNEL);
+			if (!mpi) {
+				mutex_unlock(&mlx5_ib_multiport_mutex);
+				mlx5_nic_vport_disable_roce(dev->mdev);
+				return -ENOMEM;
+			}
+
+			mpi->is_master = true;
+			mpi->mdev = dev->mdev;
+			mpi->sys_image_guid = dev->sys_image_guid;
+			dev->port[i].mp.mpi = mpi;
+			mpi->ibdev = dev;
+			mpi = NULL;
+			continue;
+		}
+
+		list_for_each_entry(mpi, &mlx5_ib_unaffiliated_port_list,
+				    list) {
+			if (dev->sys_image_guid == mpi->sys_image_guid &&
+			    (mlx5_core_native_port_num(mpi->mdev) - 1) == i) {
+				bound = mlx5_ib_bind_slave_port(dev, mpi);
+			}
+
+			if (bound) {
+				dev_dbg(&mpi->mdev->pdev->dev, "removing port from unaffiliated list.\n");
+				mlx5_ib_dbg(dev, "port %d bound\n", i + 1);
+				list_del(&mpi->list);
+				break;
+			}
+		}
+		if (!bound) {
+			get_port_caps(dev, i + 1);
+			mlx5_ib_dbg(dev, "no free port found for port %d\n",
+				    i + 1);
+		}
+	}
+
+	list_add_tail(&dev->ib_dev_list, &mlx5_ib_dev_list);
+	mutex_unlock(&mlx5_ib_multiport_mutex);
+	return err;
+}
+
+static void mlx5_ib_cleanup_multiport_master(struct mlx5_ib_dev *dev)
+{
+	int port_num = mlx5_core_native_port_num(dev->mdev) - 1;
+	enum rdma_link_layer ll = mlx5_ib_port_link_layer(&dev->ib_dev,
+							  port_num + 1);
+	int i;
+
+	if (!mlx5_core_is_mp_master(dev->mdev) || ll != IB_LINK_LAYER_ETHERNET)
+		return;
+
+	mutex_lock(&mlx5_ib_multiport_mutex);
+	for (i = 0; i < dev->num_ports; i++) {
+		if (dev->port[i].mp.mpi) {
+			/* Destroy the native port stub */
+			if (i == port_num) {
+				kfree(dev->port[i].mp.mpi);
+				dev->port[i].mp.mpi = NULL;
+			} else {
+				mlx5_ib_dbg(dev, "unbinding port_num: %d\n", i + 1);
+				list_add_tail(&dev->port[i].mp.mpi->list,
+					      &mlx5_ib_unaffiliated_port_list);
+				mlx5_ib_unbind_slave_port(dev, dev->port[i].mp.mpi);
+			}
+		}
+	}
+
+	mlx5_ib_dbg(dev, "removing from devlist\n");
+	list_del(&dev->ib_dev_list);
+	mutex_unlock(&mlx5_ib_multiport_mutex);
+
+	mlx5_nic_vport_disable_roce(dev->mdev);
+}
+
+ADD_UVERBS_ATTRIBUTES_SIMPLE(
+	mlx5_ib_dm,
+	UVERBS_OBJECT_DM,
+	UVERBS_METHOD_DM_ALLOC,
+	UVERBS_ATTR_PTR_OUT(MLX5_IB_ATTR_ALLOC_DM_RESP_START_OFFSET,
+			    UVERBS_ATTR_TYPE(u64),
+			    UA_MANDATORY),
+	UVERBS_ATTR_PTR_OUT(MLX5_IB_ATTR_ALLOC_DM_RESP_PAGE_INDEX,
+			    UVERBS_ATTR_TYPE(u16),
+			    UA_MANDATORY));
+
+ADD_UVERBS_ATTRIBUTES_SIMPLE(
+	mlx5_ib_flow_action,
+	UVERBS_OBJECT_FLOW_ACTION,
+	UVERBS_METHOD_FLOW_ACTION_ESP_CREATE,
+	UVERBS_ATTR_FLAGS_IN(MLX5_IB_ATTR_CREATE_FLOW_ACTION_FLAGS,
+			     enum mlx5_ib_uapi_flow_action_flags));
+
+static int populate_specs_root(struct mlx5_ib_dev *dev)
+{
+	const struct uverbs_object_tree_def **trees = dev->driver_trees;
+	size_t num_trees = 0;
+
+	if (mlx5_accel_ipsec_device_caps(dev->mdev) &
+	    MLX5_ACCEL_IPSEC_CAP_DEVICE)
+		trees[num_trees++] = &mlx5_ib_flow_action;
+
+	if (MLX5_CAP_DEV_MEM(dev->mdev, memic))
+		trees[num_trees++] = &mlx5_ib_dm;
+
+	if (MLX5_CAP_GEN_64(dev->mdev, general_obj_types) &
+	    MLX5_GENERAL_OBJ_TYPES_CAP_UCTX)
+		trees[num_trees++] = mlx5_ib_get_devx_tree();
+
+	num_trees += mlx5_ib_get_flow_trees(trees + num_trees);
+
+	WARN_ON(num_trees >= ARRAY_SIZE(dev->driver_trees));
+	trees[num_trees] = NULL;
+	dev->ib_dev.driver_specs = trees;
+
+	return 0;
+}
+
+static int mlx5_ib_read_counters(struct ib_counters *counters,
+				 struct ib_counters_read_attr *read_attr,
+				 struct uverbs_attr_bundle *attrs)
+{
+	struct mlx5_ib_mcounters *mcounters = to_mcounters(counters);
+	struct mlx5_read_counters_attr mread_attr = {};
+	struct mlx5_ib_flow_counters_desc *desc;
+	int ret, i;
+
+	mutex_lock(&mcounters->mcntrs_mutex);
+	if (mcounters->cntrs_max_index > read_attr->ncounters) {
+		ret = -EINVAL;
+		goto err_bound;
+	}
+
+	mread_attr.out = kcalloc(mcounters->counters_num, sizeof(u64),
+				 GFP_KERNEL);
+	if (!mread_attr.out) {
+		ret = -ENOMEM;
+		goto err_bound;
+	}
+
+	mread_attr.hw_cntrs_hndl = mcounters->hw_cntrs_hndl;
+	mread_attr.flags = read_attr->flags;
+	ret = mcounters->read_counters(counters->device, &mread_attr);
+	if (ret)
+		goto err_read;
+
+	/* do the pass over the counters data array to assign according to the
+	 * descriptions and indexing pairs
+	 */
+	desc = mcounters->counters_data;
+	for (i = 0; i < mcounters->ncounters; i++)
+		read_attr->counters_buff[desc[i].index] += mread_attr.out[desc[i].description];
+
+err_read:
+	kfree(mread_attr.out);
+err_bound:
+	mutex_unlock(&mcounters->mcntrs_mutex);
+	return ret;
+}
+
+static int mlx5_ib_destroy_counters(struct ib_counters *counters)
+{
+	struct mlx5_ib_mcounters *mcounters = to_mcounters(counters);
+
+	counters_clear_description(counters);
+	if (mcounters->hw_cntrs_hndl)
+		mlx5_fc_destroy(to_mdev(counters->device)->mdev,
+				mcounters->hw_cntrs_hndl);
+
+	kfree(mcounters);
+
+	return 0;
+}
+
+static struct ib_counters *mlx5_ib_create_counters(struct ib_device *device,
+						   struct uverbs_attr_bundle *attrs)
+{
+	struct mlx5_ib_mcounters *mcounters;
+
+	mcounters = kzalloc(sizeof(*mcounters), GFP_KERNEL);
+	if (!mcounters)
+		return ERR_PTR(-ENOMEM);
+
+	mutex_init(&mcounters->mcntrs_mutex);
+
+	return &mcounters->ibcntrs;
+}
+
+void mlx5_ib_stage_init_cleanup(struct mlx5_ib_dev *dev)
+{
+	mlx5_ib_cleanup_multiport_master(dev);
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
+	cleanup_srcu_struct(&dev->mr_srcu);
+#endif
+	kfree(dev->port);
+}
+
+int mlx5_ib_stage_init_init(struct mlx5_ib_dev *dev)
+{
+	struct mlx5_core_dev *mdev = dev->mdev;
+	const char *name;
+	int err;
+	int i;
+
+	dev->port = kcalloc(dev->num_ports, sizeof(*dev->port),
+			    GFP_KERNEL);
+	if (!dev->port)
+		return -ENOMEM;
+
+	for (i = 0; i < dev->num_ports; i++) {
+		spin_lock_init(&dev->port[i].mp.mpi_lock);
+		rwlock_init(&dev->roce[i].netdev_lock);
+	}
+
+	err = mlx5_ib_init_multiport_master(dev);
+	if (err)
+		goto err_free_port;
+
+	if (!mlx5_core_mp_enabled(mdev)) {
+		for (i = 1; i <= dev->num_ports; i++) {
+			err = get_port_caps(dev, i);
+			if (err)
+				break;
+		}
+	} else {
+		err = get_port_caps(dev, mlx5_core_native_port_num(mdev));
+	}
+	if (err)
+		goto err_mp;
+
+	if (mlx5_use_mad_ifc(dev))
+		get_ext_port_caps(dev);
+
+	if (!mlx5_lag_is_active(mdev))
+		name = "mlx5_%d";
+	else
+		name = "mlx5_bond_%d";
+
+	strlcpy(dev->ib_dev.name, name, IB_DEVICE_NAME_MAX);
+	dev->ib_dev.owner		= THIS_MODULE;
+	dev->ib_dev.node_type		= RDMA_NODE_IB_CA;
+	dev->ib_dev.local_dma_lkey	= 0 /* not supported for now */;
+	dev->ib_dev.phys_port_cnt	= dev->num_ports;
+	dev->ib_dev.num_comp_vectors    =
+		dev->mdev->priv.eq_table.num_comp_vectors;
+	dev->ib_dev.dev.parent		= &mdev->pdev->dev;
+
+	mutex_init(&dev->cap_mask_mutex);
+	INIT_LIST_HEAD(&dev->qp_list);
+	spin_lock_init(&dev->reset_flow_resource_lock);
+
+	spin_lock_init(&dev->memic.memic_lock);
+	dev->memic.dev = mdev;
+
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
+	err = init_srcu_struct(&dev->mr_srcu);
+	if (err)
+		goto err_free_port;
+#endif
+
+	return 0;
+err_mp:
+	mlx5_ib_cleanup_multiport_master(dev);
+
+err_free_port:
+	kfree(dev->port);
+
+	return -ENOMEM;
+}
+
+static int mlx5_ib_stage_flow_db_init(struct mlx5_ib_dev *dev)
+{
+	dev->flow_db = kzalloc(sizeof(*dev->flow_db), GFP_KERNEL);
+
+	if (!dev->flow_db)
+		return -ENOMEM;
+
+	mutex_init(&dev->flow_db->lock);
+
+	return 0;
+}
+
+int mlx5_ib_stage_rep_flow_db_init(struct mlx5_ib_dev *dev)
+{
+	struct mlx5_ib_dev *nic_dev;
+
+	nic_dev = mlx5_ib_get_uplink_ibdev(dev->mdev->priv.eswitch);
+
+	if (!nic_dev)
+		return -EINVAL;
+
+	dev->flow_db = nic_dev->flow_db;
+
+	return 0;
+}
+
+static void mlx5_ib_stage_flow_db_cleanup(struct mlx5_ib_dev *dev)
+{
+	kfree(dev->flow_db);
+}
+
+int mlx5_ib_stage_caps_init(struct mlx5_ib_dev *dev)
+{
+	struct mlx5_core_dev *mdev = dev->mdev;
+	int err;
+
+	dev->ib_dev.uverbs_abi_ver	= MLX5_IB_UVERBS_ABI_VERSION;
+	dev->ib_dev.uverbs_cmd_mask	=
+		(1ull << IB_USER_VERBS_CMD_GET_CONTEXT)		|
+		(1ull << IB_USER_VERBS_CMD_QUERY_DEVICE)	|
+		(1ull << IB_USER_VERBS_CMD_QUERY_PORT)		|
+		(1ull << IB_USER_VERBS_CMD_ALLOC_PD)		|
+		(1ull << IB_USER_VERBS_CMD_DEALLOC_PD)		|
+		(1ull << IB_USER_VERBS_CMD_CREATE_AH)		|
+		(1ull << IB_USER_VERBS_CMD_DESTROY_AH)		|
+		(1ull << IB_USER_VERBS_CMD_REG_MR)		|
+		(1ull << IB_USER_VERBS_CMD_REREG_MR)		|
+		(1ull << IB_USER_VERBS_CMD_DEREG_MR)		|
+		(1ull << IB_USER_VERBS_CMD_CREATE_COMP_CHANNEL)	|
+		(1ull << IB_USER_VERBS_CMD_CREATE_CQ)		|
+		(1ull << IB_USER_VERBS_CMD_RESIZE_CQ)		|
+		(1ull << IB_USER_VERBS_CMD_DESTROY_CQ)		|
+		(1ull << IB_USER_VERBS_CMD_CREATE_QP)		|
+		(1ull << IB_USER_VERBS_CMD_MODIFY_QP)		|
+		(1ull << IB_USER_VERBS_CMD_QUERY_QP)		|
+		(1ull << IB_USER_VERBS_CMD_DESTROY_QP)		|
+		(1ull << IB_USER_VERBS_CMD_ATTACH_MCAST)	|
+		(1ull << IB_USER_VERBS_CMD_DETACH_MCAST)	|
+		(1ull << IB_USER_VERBS_CMD_CREATE_SRQ)		|
+		(1ull << IB_USER_VERBS_CMD_MODIFY_SRQ)		|
+		(1ull << IB_USER_VERBS_CMD_QUERY_SRQ)		|
+		(1ull << IB_USER_VERBS_CMD_DESTROY_SRQ)		|
+		(1ull << IB_USER_VERBS_CMD_CREATE_XSRQ)		|
+		(1ull << IB_USER_VERBS_CMD_OPEN_QP);
+	dev->ib_dev.uverbs_ex_cmd_mask =
+		(1ull << IB_USER_VERBS_EX_CMD_QUERY_DEVICE)	|
+		(1ull << IB_USER_VERBS_EX_CMD_CREATE_CQ)	|
+		(1ull << IB_USER_VERBS_EX_CMD_CREATE_QP)	|
+		(1ull << IB_USER_VERBS_EX_CMD_MODIFY_QP)	|
+		(1ull << IB_USER_VERBS_EX_CMD_MODIFY_CQ);
+
+	dev->ib_dev.query_device	= mlx5_ib_query_device;
+	dev->ib_dev.get_link_layer	= mlx5_ib_port_link_layer;
+	dev->ib_dev.query_gid		= mlx5_ib_query_gid;
+	dev->ib_dev.add_gid		= mlx5_ib_add_gid;
+	dev->ib_dev.del_gid		= mlx5_ib_del_gid;
+	dev->ib_dev.query_pkey		= mlx5_ib_query_pkey;
+	dev->ib_dev.modify_device	= mlx5_ib_modify_device;
+	dev->ib_dev.modify_port		= mlx5_ib_modify_port;
+	dev->ib_dev.alloc_ucontext	= mlx5_ib_alloc_ucontext;
+	dev->ib_dev.dealloc_ucontext	= mlx5_ib_dealloc_ucontext;
+	dev->ib_dev.mmap		= mlx5_ib_mmap;
+	dev->ib_dev.alloc_pd		= mlx5_ib_alloc_pd;
+	dev->ib_dev.dealloc_pd		= mlx5_ib_dealloc_pd;
+	dev->ib_dev.create_ah		= mlx5_ib_create_ah;
+	dev->ib_dev.query_ah		= mlx5_ib_query_ah;
+	dev->ib_dev.destroy_ah		= mlx5_ib_destroy_ah;
+	dev->ib_dev.create_srq		= mlx5_ib_create_srq;
+	dev->ib_dev.modify_srq		= mlx5_ib_modify_srq;
+	dev->ib_dev.query_srq		= mlx5_ib_query_srq;
+	dev->ib_dev.destroy_srq		= mlx5_ib_destroy_srq;
+	dev->ib_dev.post_srq_recv	= mlx5_ib_post_srq_recv;
+	dev->ib_dev.create_qp		= mlx5_ib_create_qp;
+	dev->ib_dev.modify_qp		= mlx5_ib_modify_qp;
+	dev->ib_dev.query_qp		= mlx5_ib_query_qp;
+	dev->ib_dev.destroy_qp		= mlx5_ib_destroy_qp;
+	dev->ib_dev.drain_sq		= mlx5_ib_drain_sq;
+	dev->ib_dev.drain_rq		= mlx5_ib_drain_rq;
+	dev->ib_dev.post_send		= mlx5_ib_post_send;
+	dev->ib_dev.post_recv		= mlx5_ib_post_recv;
+	dev->ib_dev.create_cq		= mlx5_ib_create_cq;
+	dev->ib_dev.modify_cq		= mlx5_ib_modify_cq;
+	dev->ib_dev.resize_cq		= mlx5_ib_resize_cq;
+	dev->ib_dev.destroy_cq		= mlx5_ib_destroy_cq;
+	dev->ib_dev.poll_cq		= mlx5_ib_poll_cq;
+	dev->ib_dev.req_notify_cq	= mlx5_ib_arm_cq;
+	dev->ib_dev.get_dma_mr		= mlx5_ib_get_dma_mr;
+	dev->ib_dev.reg_user_mr		= mlx5_ib_reg_user_mr;
+	dev->ib_dev.rereg_user_mr	= mlx5_ib_rereg_user_mr;
+	dev->ib_dev.dereg_mr		= mlx5_ib_dereg_mr;
+	dev->ib_dev.attach_mcast	= mlx5_ib_mcg_attach;
+	dev->ib_dev.detach_mcast	= mlx5_ib_mcg_detach;
+	dev->ib_dev.process_mad		= mlx5_ib_process_mad;
+	dev->ib_dev.alloc_mr		= mlx5_ib_alloc_mr;
+	dev->ib_dev.map_mr_sg		= mlx5_ib_map_mr_sg;
+	dev->ib_dev.check_mr_status	= mlx5_ib_check_mr_status;
+	dev->ib_dev.get_dev_fw_str      = get_dev_fw_str;
+	dev->ib_dev.get_vector_affinity	= mlx5_ib_get_vector_affinity;
+	if (MLX5_CAP_GEN(mdev, ipoib_enhanced_offloads))
+		dev->ib_dev.alloc_rdma_netdev	= mlx5_ib_alloc_rdma_netdev;
+
+	if (mlx5_core_is_pf(mdev)) {
+		dev->ib_dev.get_vf_config	= mlx5_ib_get_vf_config;
+		dev->ib_dev.set_vf_link_state	= mlx5_ib_set_vf_link_state;
+		dev->ib_dev.get_vf_stats	= mlx5_ib_get_vf_stats;
+		dev->ib_dev.set_vf_guid		= mlx5_ib_set_vf_guid;
+	}
+
+	dev->ib_dev.disassociate_ucontext = mlx5_ib_disassociate_ucontext;
+
+	dev->umr_fence = mlx5_get_umr_fence(MLX5_CAP_GEN(mdev, umr_fence));
+
+	if (MLX5_CAP_GEN(mdev, imaicl)) {
+		dev->ib_dev.alloc_mw		= mlx5_ib_alloc_mw;
+		dev->ib_dev.dealloc_mw		= mlx5_ib_dealloc_mw;
+		dev->ib_dev.uverbs_cmd_mask |=
+			(1ull << IB_USER_VERBS_CMD_ALLOC_MW)	|
+			(1ull << IB_USER_VERBS_CMD_DEALLOC_MW);
+	}
+
+	if (MLX5_CAP_GEN(mdev, xrc)) {
+		dev->ib_dev.alloc_xrcd = mlx5_ib_alloc_xrcd;
+		dev->ib_dev.dealloc_xrcd = mlx5_ib_dealloc_xrcd;
+		dev->ib_dev.uverbs_cmd_mask |=
+			(1ull << IB_USER_VERBS_CMD_OPEN_XRCD) |
+			(1ull << IB_USER_VERBS_CMD_CLOSE_XRCD);
+	}
+
+	if (MLX5_CAP_DEV_MEM(mdev, memic)) {
+		dev->ib_dev.alloc_dm = mlx5_ib_alloc_dm;
+		dev->ib_dev.dealloc_dm = mlx5_ib_dealloc_dm;
+		dev->ib_dev.reg_dm_mr = mlx5_ib_reg_dm_mr;
+	}
+
+	dev->ib_dev.create_flow	= mlx5_ib_create_flow;
+	dev->ib_dev.destroy_flow = mlx5_ib_destroy_flow;
+	dev->ib_dev.uverbs_ex_cmd_mask |=
+			(1ull << IB_USER_VERBS_EX_CMD_CREATE_FLOW) |
+			(1ull << IB_USER_VERBS_EX_CMD_DESTROY_FLOW);
+	dev->ib_dev.create_flow_action_esp = mlx5_ib_create_flow_action_esp;
+	dev->ib_dev.destroy_flow_action = mlx5_ib_destroy_flow_action;
+	dev->ib_dev.modify_flow_action_esp = mlx5_ib_modify_flow_action_esp;
+	dev->ib_dev.driver_id = RDMA_DRIVER_MLX5;
+	dev->ib_dev.create_counters = mlx5_ib_create_counters;
+	dev->ib_dev.destroy_counters = mlx5_ib_destroy_counters;
+	dev->ib_dev.read_counters = mlx5_ib_read_counters;
+
+	err = init_node_data(dev);
+	if (err)
+		return err;
+
+	if ((MLX5_CAP_GEN(dev->mdev, port_type) == MLX5_CAP_PORT_TYPE_ETH) &&
+	    (MLX5_CAP_GEN(dev->mdev, disable_local_lb_uc) ||
+	     MLX5_CAP_GEN(dev->mdev, disable_local_lb_mc)))
+		mutex_init(&dev->lb_mutex);
+
+	return 0;
+}
+
+static int mlx5_ib_stage_non_default_cb(struct mlx5_ib_dev *dev)
+{
+	dev->ib_dev.get_port_immutable  = mlx5_port_immutable;
+	dev->ib_dev.query_port		= mlx5_ib_query_port;
+
+	return 0;
+}
+
+int mlx5_ib_stage_rep_non_default_cb(struct mlx5_ib_dev *dev)
+{
+	dev->ib_dev.get_port_immutable  = mlx5_port_rep_immutable;
+	dev->ib_dev.query_port		= mlx5_ib_rep_query_port;
+
+	return 0;
+}
+
+static int mlx5_ib_stage_common_roce_init(struct mlx5_ib_dev *dev)
+{
+	u8 port_num;
+	int i;
+
+	for (i = 0; i < dev->num_ports; i++) {
+		dev->roce[i].dev = dev;
+		dev->roce[i].native_port_num = i + 1;
+		dev->roce[i].last_port_state = IB_PORT_DOWN;
+	}
+
+	dev->ib_dev.get_netdev	= mlx5_ib_get_netdev;
+	dev->ib_dev.create_wq	 = mlx5_ib_create_wq;
+	dev->ib_dev.modify_wq	 = mlx5_ib_modify_wq;
+	dev->ib_dev.destroy_wq	 = mlx5_ib_destroy_wq;
+	dev->ib_dev.create_rwq_ind_table = mlx5_ib_create_rwq_ind_table;
+	dev->ib_dev.destroy_rwq_ind_table = mlx5_ib_destroy_rwq_ind_table;
+
+	dev->ib_dev.uverbs_ex_cmd_mask |=
+			(1ull << IB_USER_VERBS_EX_CMD_CREATE_WQ) |
+			(1ull << IB_USER_VERBS_EX_CMD_MODIFY_WQ) |
+			(1ull << IB_USER_VERBS_EX_CMD_DESTROY_WQ) |
+			(1ull << IB_USER_VERBS_EX_CMD_CREATE_RWQ_IND_TBL) |
+			(1ull << IB_USER_VERBS_EX_CMD_DESTROY_RWQ_IND_TBL);
+
+	port_num = mlx5_core_native_port_num(dev->mdev) - 1;
+
+	return mlx5_add_netdev_notifier(dev, port_num);
+}
+
+static void mlx5_ib_stage_common_roce_cleanup(struct mlx5_ib_dev *dev)
+{
+	u8 port_num = mlx5_core_native_port_num(dev->mdev) - 1;
+
+	mlx5_remove_netdev_notifier(dev, port_num);
+}
+
+int mlx5_ib_stage_rep_roce_init(struct mlx5_ib_dev *dev)
+{
+	struct mlx5_core_dev *mdev = dev->mdev;
+	enum rdma_link_layer ll;
+	int port_type_cap;
+	int err = 0;
+
+	port_type_cap = MLX5_CAP_GEN(mdev, port_type);
+	ll = mlx5_port_type_cap_to_rdma_ll(port_type_cap);
+
+	if (ll == IB_LINK_LAYER_ETHERNET)
+		err = mlx5_ib_stage_common_roce_init(dev);
+
+	return err;
+}
+
+void mlx5_ib_stage_rep_roce_cleanup(struct mlx5_ib_dev *dev)
+{
+	mlx5_ib_stage_common_roce_cleanup(dev);
+}
+
+static int mlx5_ib_stage_roce_init(struct mlx5_ib_dev *dev)
+{
+	struct mlx5_core_dev *mdev = dev->mdev;
+	enum rdma_link_layer ll;
+	int port_type_cap;
+	int err;
+
+	port_type_cap = MLX5_CAP_GEN(mdev, port_type);
+	ll = mlx5_port_type_cap_to_rdma_ll(port_type_cap);
+
+	if (ll == IB_LINK_LAYER_ETHERNET) {
+		err = mlx5_ib_stage_common_roce_init(dev);
+		if (err)
+			return err;
+
+		err = mlx5_enable_eth(dev);
+		if (err)
+			goto cleanup;
+	}
+
+	return 0;
+cleanup:
+	mlx5_ib_stage_common_roce_cleanup(dev);
+
+	return err;
+}
+
+static void mlx5_ib_stage_roce_cleanup(struct mlx5_ib_dev *dev)
+{
+	struct mlx5_core_dev *mdev = dev->mdev;
+	enum rdma_link_layer ll;
+	int port_type_cap;
+
+	port_type_cap = MLX5_CAP_GEN(mdev, port_type);
+	ll = mlx5_port_type_cap_to_rdma_ll(port_type_cap);
+
+	if (ll == IB_LINK_LAYER_ETHERNET) {
+		mlx5_disable_eth(dev);
+		mlx5_ib_stage_common_roce_cleanup(dev);
+	}
+}
+
+int mlx5_ib_stage_dev_res_init(struct mlx5_ib_dev *dev)
+{
+	return create_dev_resources(&dev->devr);
+}
+
+void mlx5_ib_stage_dev_res_cleanup(struct mlx5_ib_dev *dev)
+{
+	destroy_dev_resources(&dev->devr);
+}
+
+static int mlx5_ib_stage_odp_init(struct mlx5_ib_dev *dev)
+{
+	mlx5_ib_internal_fill_odp_caps(dev);
+
+	return mlx5_ib_odp_init_one(dev);
+}
+
+int mlx5_ib_stage_counters_init(struct mlx5_ib_dev *dev)
+{
+	if (MLX5_CAP_GEN(dev->mdev, max_qp_cnt)) {
+		dev->ib_dev.get_hw_stats	= mlx5_ib_get_hw_stats;
+		dev->ib_dev.alloc_hw_stats	= mlx5_ib_alloc_hw_stats;
+
+		return mlx5_ib_alloc_counters(dev);
+	}
+
+	return 0;
+}
+
+void mlx5_ib_stage_counters_cleanup(struct mlx5_ib_dev *dev)
+{
+	if (MLX5_CAP_GEN(dev->mdev, max_qp_cnt))
+		mlx5_ib_dealloc_counters(dev);
+}
+
+static int mlx5_ib_stage_cong_debugfs_init(struct mlx5_ib_dev *dev)
+{
+	return mlx5_ib_init_cong_debugfs(dev,
+					 mlx5_core_native_port_num(dev->mdev) - 1);
+}
+
+static void mlx5_ib_stage_cong_debugfs_cleanup(struct mlx5_ib_dev *dev)
+{
+	mlx5_ib_cleanup_cong_debugfs(dev,
+				     mlx5_core_native_port_num(dev->mdev) - 1);
+}
+
+static int mlx5_ib_stage_uar_init(struct mlx5_ib_dev *dev)
+{
+	dev->mdev->priv.uar = mlx5_get_uars_page(dev->mdev);
+	return PTR_ERR_OR_ZERO(dev->mdev->priv.uar);
+}
+
+static void mlx5_ib_stage_uar_cleanup(struct mlx5_ib_dev *dev)
+{
+	mlx5_put_uars_page(dev->mdev, dev->mdev->priv.uar);
+}
+
+int mlx5_ib_stage_bfrag_init(struct mlx5_ib_dev *dev)
+{
+	int err;
+
+	err = mlx5_alloc_bfreg(dev->mdev, &dev->bfreg, false, false);
+	if (err)
+		return err;
+
+	err = mlx5_alloc_bfreg(dev->mdev, &dev->fp_bfreg, false, true);
+	if (err)
+		mlx5_free_bfreg(dev->mdev, &dev->bfreg);
+
+	return err;
+}
+
+void mlx5_ib_stage_bfrag_cleanup(struct mlx5_ib_dev *dev)
+{
+	mlx5_free_bfreg(dev->mdev, &dev->fp_bfreg);
+	mlx5_free_bfreg(dev->mdev, &dev->bfreg);
+}
+
+static int mlx5_ib_stage_populate_specs(struct mlx5_ib_dev *dev)
+{
+	return populate_specs_root(dev);
+}
+
+int mlx5_ib_stage_ib_reg_init(struct mlx5_ib_dev *dev)
+{
+	return ib_register_device(&dev->ib_dev, NULL);
+}
+
+void mlx5_ib_stage_pre_ib_reg_umr_cleanup(struct mlx5_ib_dev *dev)
+{
+	destroy_umrc_res(dev);
+}
+
+void mlx5_ib_stage_ib_reg_cleanup(struct mlx5_ib_dev *dev)
+{
+	ib_unregister_device(&dev->ib_dev);
+}
+
+int mlx5_ib_stage_post_ib_reg_umr_init(struct mlx5_ib_dev *dev)
+{
+	return create_umr_res(dev);
+}
+
+static int mlx5_ib_stage_delay_drop_init(struct mlx5_ib_dev *dev)
+{
+	init_delay_drop(dev);
+
+	return 0;
+}
+
+static void mlx5_ib_stage_delay_drop_cleanup(struct mlx5_ib_dev *dev)
+{
+	cancel_delay_drop(dev);
+}
+
+int mlx5_ib_stage_class_attr_init(struct mlx5_ib_dev *dev)
+{
+	int err;
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(mlx5_class_attributes); i++) {
+		err = device_create_file(&dev->ib_dev.dev,
+					 mlx5_class_attributes[i]);
+		if (err)
+			return err;
+	}
+
+	return 0;
+}
+
+static int mlx5_ib_stage_rep_reg_init(struct mlx5_ib_dev *dev)
+{
+	mlx5_ib_register_vport_reps(dev);
+
+	return 0;
+}
+
+static void mlx5_ib_stage_rep_reg_cleanup(struct mlx5_ib_dev *dev)
+{
+	mlx5_ib_unregister_vport_reps(dev);
+}
+
+void __mlx5_ib_remove(struct mlx5_ib_dev *dev,
+		      const struct mlx5_ib_profile *profile,
+		      int stage)
+{
+	/* Number of stages to cleanup */
+	while (stage) {
+		stage--;
+		if (profile->stage[stage].cleanup)
+			profile->stage[stage].cleanup(dev);
+	}
+
+	ib_dealloc_device((struct ib_device *)dev);
+}
+
+void *__mlx5_ib_add(struct mlx5_ib_dev *dev,
+		    const struct mlx5_ib_profile *profile)
+{
+	int err;
+	int i;
+
+	printk_once(KERN_INFO "%s", mlx5_version);
+
+	for (i = 0; i < MLX5_IB_STAGE_MAX; i++) {
+		if (profile->stage[i].init) {
+			err = profile->stage[i].init(dev);
+			if (err)
+				goto err_out;
+		}
+	}
+
+	dev->profile = profile;
+	dev->ib_active = true;
+
+	return dev;
+
+err_out:
+	__mlx5_ib_remove(dev, profile, i);
+
+	return NULL;
+}
+
+static const struct mlx5_ib_profile pf_profile = {
+	STAGE_CREATE(MLX5_IB_STAGE_INIT,
+		     mlx5_ib_stage_init_init,
+		     mlx5_ib_stage_init_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_FLOW_DB,
+		     mlx5_ib_stage_flow_db_init,
+		     mlx5_ib_stage_flow_db_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_CAPS,
+		     mlx5_ib_stage_caps_init,
+		     NULL),
+	STAGE_CREATE(MLX5_IB_STAGE_NON_DEFAULT_CB,
+		     mlx5_ib_stage_non_default_cb,
+		     NULL),
+	STAGE_CREATE(MLX5_IB_STAGE_ROCE,
+		     mlx5_ib_stage_roce_init,
+		     mlx5_ib_stage_roce_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_DEVICE_RESOURCES,
+		     mlx5_ib_stage_dev_res_init,
+		     mlx5_ib_stage_dev_res_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_ODP,
+		     mlx5_ib_stage_odp_init,
+		     NULL),
+	STAGE_CREATE(MLX5_IB_STAGE_COUNTERS,
+		     mlx5_ib_stage_counters_init,
+		     mlx5_ib_stage_counters_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_CONG_DEBUGFS,
+		     mlx5_ib_stage_cong_debugfs_init,
+		     mlx5_ib_stage_cong_debugfs_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_UAR,
+		     mlx5_ib_stage_uar_init,
+		     mlx5_ib_stage_uar_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_BFREG,
+		     mlx5_ib_stage_bfrag_init,
+		     mlx5_ib_stage_bfrag_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_PRE_IB_REG_UMR,
+		     NULL,
+		     mlx5_ib_stage_pre_ib_reg_umr_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_SPECS,
+		     mlx5_ib_stage_populate_specs,
+		     NULL),
+	STAGE_CREATE(MLX5_IB_STAGE_IB_REG,
+		     mlx5_ib_stage_ib_reg_init,
+		     mlx5_ib_stage_ib_reg_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_POST_IB_REG_UMR,
+		     mlx5_ib_stage_post_ib_reg_umr_init,
+		     NULL),
+	STAGE_CREATE(MLX5_IB_STAGE_DELAY_DROP,
+		     mlx5_ib_stage_delay_drop_init,
+		     mlx5_ib_stage_delay_drop_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_CLASS_ATTR,
+		     mlx5_ib_stage_class_attr_init,
+		     NULL),
+};
+
+static const struct mlx5_ib_profile nic_rep_profile = {
+	STAGE_CREATE(MLX5_IB_STAGE_INIT,
+		     mlx5_ib_stage_init_init,
+		     mlx5_ib_stage_init_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_FLOW_DB,
+		     mlx5_ib_stage_flow_db_init,
+		     mlx5_ib_stage_flow_db_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_CAPS,
+		     mlx5_ib_stage_caps_init,
+		     NULL),
+	STAGE_CREATE(MLX5_IB_STAGE_NON_DEFAULT_CB,
+		     mlx5_ib_stage_rep_non_default_cb,
+		     NULL),
+	STAGE_CREATE(MLX5_IB_STAGE_ROCE,
+		     mlx5_ib_stage_rep_roce_init,
+		     mlx5_ib_stage_rep_roce_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_DEVICE_RESOURCES,
+		     mlx5_ib_stage_dev_res_init,
+		     mlx5_ib_stage_dev_res_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_COUNTERS,
+		     mlx5_ib_stage_counters_init,
+		     mlx5_ib_stage_counters_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_UAR,
+		     mlx5_ib_stage_uar_init,
+		     mlx5_ib_stage_uar_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_BFREG,
+		     mlx5_ib_stage_bfrag_init,
+		     mlx5_ib_stage_bfrag_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_PRE_IB_REG_UMR,
+		     NULL,
+		     mlx5_ib_stage_pre_ib_reg_umr_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_SPECS,
+		     mlx5_ib_stage_populate_specs,
+		     NULL),
+	STAGE_CREATE(MLX5_IB_STAGE_IB_REG,
+		     mlx5_ib_stage_ib_reg_init,
+		     mlx5_ib_stage_ib_reg_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_POST_IB_REG_UMR,
+		     mlx5_ib_stage_post_ib_reg_umr_init,
+		     NULL),
+	STAGE_CREATE(MLX5_IB_STAGE_CLASS_ATTR,
+		     mlx5_ib_stage_class_attr_init,
+		     NULL),
+	STAGE_CREATE(MLX5_IB_STAGE_REP_REG,
+		     mlx5_ib_stage_rep_reg_init,
+		     mlx5_ib_stage_rep_reg_cleanup),
+};
+
+static void *mlx5_ib_add_slave_port(struct mlx5_core_dev *mdev)
+{
+	struct mlx5_ib_multiport_info *mpi;
+	struct mlx5_ib_dev *dev;
+	bool bound = false;
+	int err;
+
+	mpi = kzalloc(sizeof(*mpi), GFP_KERNEL);
+	if (!mpi)
+		return NULL;
+
+	mpi->mdev = mdev;
+
+	err = mlx5_query_nic_vport_system_image_guid(mdev,
+						     &mpi->sys_image_guid);
+	if (err) {
+		kfree(mpi);
+		return NULL;
+	}
+
+	mutex_lock(&mlx5_ib_multiport_mutex);
+	list_for_each_entry(dev, &mlx5_ib_dev_list, ib_dev_list) {
+		if (dev->sys_image_guid == mpi->sys_image_guid)
+			bound = mlx5_ib_bind_slave_port(dev, mpi);
+
+		if (bound) {
+			rdma_roce_rescan_device(&dev->ib_dev);
+			mpi->ibdev->ib_active = true;
+			break;
+		}
+	}
+
+	if (!bound) {
+		list_add_tail(&mpi->list, &mlx5_ib_unaffiliated_port_list);
+		dev_dbg(&mdev->pdev->dev, "no suitable IB device found to bind to, added to unaffiliated list.\n");
+	}
+	mutex_unlock(&mlx5_ib_multiport_mutex);
+
+	return mpi;
+}
+
+static void *mlx5_ib_add(struct mlx5_core_dev *mdev)
+{
+	enum rdma_link_layer ll;
+	struct mlx5_ib_dev *dev;
+	int port_type_cap;
+
+	printk_once(KERN_INFO "%s", mlx5_version);
+
+	port_type_cap = MLX5_CAP_GEN(mdev, port_type);
+	ll = mlx5_port_type_cap_to_rdma_ll(port_type_cap);
+
+	if (mlx5_core_is_mp_slave(mdev) && ll == IB_LINK_LAYER_ETHERNET)
+		return mlx5_ib_add_slave_port(mdev);
+
+	dev = (struct mlx5_ib_dev *)ib_alloc_device(sizeof(*dev));
+	if (!dev)
+		return NULL;
+
+	dev->mdev = mdev;
+	dev->num_ports = max(MLX5_CAP_GEN(mdev, num_ports),
+			     MLX5_CAP_GEN(mdev, num_vhca_ports));
+
+	if (MLX5_ESWITCH_MANAGER(mdev) &&
+	    mlx5_ib_eswitch_mode(mdev->priv.eswitch) == SRIOV_OFFLOADS) {
+		dev->rep = mlx5_ib_vport_rep(mdev->priv.eswitch, 0);
+
+		return __mlx5_ib_add(dev, &nic_rep_profile);
+	}
+
+	return __mlx5_ib_add(dev, &pf_profile);
+}
+
+static void mlx5_ib_remove(struct mlx5_core_dev *mdev, void *context)
+{
+	struct mlx5_ib_multiport_info *mpi;
+	struct mlx5_ib_dev *dev;
+
+	if (mlx5_core_is_mp_slave(mdev)) {
+		mpi = context;
+		mutex_lock(&mlx5_ib_multiport_mutex);
+		if (mpi->ibdev)
+			mlx5_ib_unbind_slave_port(mpi->ibdev, mpi);
+		list_del(&mpi->list);
+		mutex_unlock(&mlx5_ib_multiport_mutex);
+		kfree(mpi);
+		return;
+	}
+
+	dev = context;
+	__mlx5_ib_remove(dev, dev->profile, MLX5_IB_STAGE_MAX);
+}
+
+static struct mlx5_interface mlx5_ib_interface = {
+	.add            = mlx5_ib_add,
+	.remove         = mlx5_ib_remove,
+	.event          = mlx5_ib_event,
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
+	.pfault		= mlx5_ib_pfault,
+#endif
+	.protocol	= MLX5_INTERFACE_PROTOCOL_IB,
+};
+
+unsigned long mlx5_ib_get_xlt_emergency_page(void)
+{
+	mutex_lock(&xlt_emergency_page_mutex);
+	return xlt_emergency_page;
+}
+
+void mlx5_ib_put_xlt_emergency_page(void)
+{
+	mutex_unlock(&xlt_emergency_page_mutex);
+}
+
+static int __init mlx5_ib_init(void)
+{
+	int err;
+
+	xlt_emergency_page = __get_free_page(GFP_KERNEL);
+	if (!xlt_emergency_page)
+		return -ENOMEM;
+
+	mutex_init(&xlt_emergency_page_mutex);
+
+	mlx5_ib_event_wq = alloc_ordered_workqueue("mlx5_ib_event_wq", 0);
+	if (!mlx5_ib_event_wq) {
+		free_page(xlt_emergency_page);
+		return -ENOMEM;
+	}
+
+	mlx5_ib_odp_init();
+
+	err = mlx5_register_interface(&mlx5_ib_interface);
+
+	return err;
+}
+
+static void __exit mlx5_ib_cleanup(void)
+{
+	mlx5_unregister_interface(&mlx5_ib_interface);
+	destroy_workqueue(mlx5_ib_event_wq);
+	mutex_destroy(&xlt_emergency_page_mutex);
+	free_page(xlt_emergency_page);
+}
+
+module_init(mlx5_ib_init);
+module_exit(mlx5_ib_cleanup);
diff --git a/drivers/infiniband/hw/mlx5/mem.c b/drivers/infiniband/hw/mlx5/mem.c
new file mode 100644
index 000000000..f3dbd75a0
--- /dev/null
+++ b/drivers/infiniband/hw/mlx5/mem.c
@@ -0,0 +1,234 @@
+/*
+ * Copyright (c) 2013-2015, Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/module.h>
+#include <rdma/ib_umem.h>
+#include <rdma/ib_umem_odp.h>
+#include "mlx5_ib.h"
+
+/* @umem: umem object to scan
+ * @addr: ib virtual address requested by the user
+ * @max_page_shift: high limit for page_shift - 0 means no limit
+ * @count: number of PAGE_SIZE pages covered by umem
+ * @shift: page shift for the compound pages found in the region
+ * @ncont: number of compund pages
+ * @order: log2 of the number of compound pages
+ */
+void mlx5_ib_cont_pages(struct ib_umem *umem, u64 addr,
+			unsigned long max_page_shift,
+			int *count, int *shift,
+			int *ncont, int *order)
+{
+	unsigned long tmp;
+	unsigned long m;
+	u64 base = ~0, p = 0;
+	u64 len, pfn;
+	int i = 0;
+	struct scatterlist *sg;
+	int entry;
+	unsigned long page_shift = umem->page_shift;
+
+	if (umem->odp_data) {
+		*ncont = ib_umem_page_count(umem);
+		*count = *ncont << (page_shift - PAGE_SHIFT);
+		*shift = page_shift;
+		if (order)
+			*order = ilog2(roundup_pow_of_two(*ncont));
+
+		return;
+	}
+
+	addr = addr >> page_shift;
+	tmp = (unsigned long)addr;
+	m = find_first_bit(&tmp, BITS_PER_LONG);
+	if (max_page_shift)
+		m = min_t(unsigned long, max_page_shift - page_shift, m);
+
+	for_each_sg(umem->sg_head.sgl, sg, umem->nmap, entry) {
+		len = sg_dma_len(sg) >> page_shift;
+		pfn = sg_dma_address(sg) >> page_shift;
+		if (base + p != pfn) {
+			/* If either the offset or the new
+			 * base are unaligned update m
+			 */
+			tmp = (unsigned long)(pfn | p);
+			if (!IS_ALIGNED(tmp, 1 << m))
+				m = find_first_bit(&tmp, BITS_PER_LONG);
+
+			base = pfn;
+			p = 0;
+		}
+
+		p += len;
+		i += len;
+	}
+
+	if (i) {
+		m = min_t(unsigned long, ilog2(roundup_pow_of_two(i)), m);
+
+		if (order)
+			*order = ilog2(roundup_pow_of_two(i) >> m);
+
+		*ncont = DIV_ROUND_UP(i, (1 << m));
+	} else {
+		m  = 0;
+
+		if (order)
+			*order = 0;
+
+		*ncont = 0;
+	}
+	*shift = page_shift + m;
+	*count = i;
+}
+
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
+static u64 umem_dma_to_mtt(dma_addr_t umem_dma)
+{
+	u64 mtt_entry = umem_dma & ODP_DMA_ADDR_MASK;
+
+	if (umem_dma & ODP_READ_ALLOWED_BIT)
+		mtt_entry |= MLX5_IB_MTT_READ;
+	if (umem_dma & ODP_WRITE_ALLOWED_BIT)
+		mtt_entry |= MLX5_IB_MTT_WRITE;
+
+	return mtt_entry;
+}
+#endif
+
+/*
+ * Populate the given array with bus addresses from the umem.
+ *
+ * dev - mlx5_ib device
+ * umem - umem to use to fill the pages
+ * page_shift - determines the page size used in the resulting array
+ * offset - offset into the umem to start from,
+ *          only implemented for ODP umems
+ * num_pages - total number of pages to fill
+ * pas - bus addresses array to fill
+ * access_flags - access flags to set on all present pages.
+		  use enum mlx5_ib_mtt_access_flags for this.
+ */
+void __mlx5_ib_populate_pas(struct mlx5_ib_dev *dev, struct ib_umem *umem,
+			    int page_shift, size_t offset, size_t num_pages,
+			    __be64 *pas, int access_flags)
+{
+	unsigned long umem_page_shift = umem->page_shift;
+	int shift = page_shift - umem_page_shift;
+	int mask = (1 << shift) - 1;
+	int i, k, idx;
+	u64 cur = 0;
+	u64 base;
+	int len;
+	struct scatterlist *sg;
+	int entry;
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
+	const bool odp = umem->odp_data != NULL;
+
+	if (odp) {
+		WARN_ON(shift != 0);
+		WARN_ON(access_flags != (MLX5_IB_MTT_READ | MLX5_IB_MTT_WRITE));
+
+		for (i = 0; i < num_pages; ++i) {
+			dma_addr_t pa = umem->odp_data->dma_list[offset + i];
+
+			pas[i] = cpu_to_be64(umem_dma_to_mtt(pa));
+		}
+		return;
+	}
+#endif
+
+	i = 0;
+	for_each_sg(umem->sg_head.sgl, sg, umem->nmap, entry) {
+		len = sg_dma_len(sg) >> umem_page_shift;
+		base = sg_dma_address(sg);
+
+		/* Skip elements below offset */
+		if (i + len < offset << shift) {
+			i += len;
+			continue;
+		}
+
+		/* Skip pages below offset */
+		if (i < offset << shift) {
+			k = (offset << shift) - i;
+			i = offset << shift;
+		} else {
+			k = 0;
+		}
+
+		for (; k < len; k++) {
+			if (!(i & mask)) {
+				cur = base + (k << umem_page_shift);
+				cur |= access_flags;
+				idx = (i >> shift) - offset;
+
+				pas[idx] = cpu_to_be64(cur);
+				mlx5_ib_dbg(dev, "pas[%d] 0x%llx\n",
+					    i >> shift, be64_to_cpu(pas[idx]));
+			}
+			i++;
+
+			/* Stop after num_pages reached */
+			if (i >> shift >= offset + num_pages)
+				return;
+		}
+	}
+}
+
+void mlx5_ib_populate_pas(struct mlx5_ib_dev *dev, struct ib_umem *umem,
+			  int page_shift, __be64 *pas, int access_flags)
+{
+	return __mlx5_ib_populate_pas(dev, umem, page_shift, 0,
+				      ib_umem_num_pages(umem), pas,
+				      access_flags);
+}
+int mlx5_ib_get_buf_offset(u64 addr, int page_shift, u32 *offset)
+{
+	u64 page_size;
+	u64 page_mask;
+	u64 off_size;
+	u64 off_mask;
+	u64 buf_off;
+
+	page_size = (u64)1 << page_shift;
+	page_mask = page_size - 1;
+	buf_off = addr & page_mask;
+	off_size = page_size >> 6;
+	off_mask = off_size - 1;
+
+	if (buf_off & off_mask)
+		return -EINVAL;
+
+	*offset = buf_off >> ilog2(off_size);
+	return 0;
+}
diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
new file mode 100644
index 000000000..6a060c845
--- /dev/null
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -0,0 +1,1367 @@
+/*
+ * Copyright (c) 2013-2015, Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef MLX5_IB_H
+#define MLX5_IB_H
+
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <rdma/ib_verbs.h>
+#include <rdma/ib_smi.h>
+#include <linux/mlx5/driver.h>
+#include <linux/mlx5/cq.h>
+#include <linux/mlx5/qp.h>
+#include <linux/mlx5/srq.h>
+#include <linux/types.h>
+#include <linux/mlx5/transobj.h>
+#include <rdma/ib_user_verbs.h>
+#include <rdma/mlx5-abi.h>
+#include <rdma/uverbs_ioctl.h>
+#include <rdma/mlx5_user_ioctl_cmds.h>
+
+#define mlx5_ib_dbg(dev, format, arg...)				\
+pr_debug("%s:%s:%d:(pid %d): " format, (dev)->ib_dev.name, __func__,	\
+	 __LINE__, current->pid, ##arg)
+
+#define mlx5_ib_err(dev, format, arg...)				\
+pr_err("%s:%s:%d:(pid %d): " format, (dev)->ib_dev.name, __func__,	\
+	__LINE__, current->pid, ##arg)
+
+#define mlx5_ib_warn(dev, format, arg...)				\
+pr_warn("%s:%s:%d:(pid %d): " format, (dev)->ib_dev.name, __func__,	\
+	__LINE__, current->pid, ##arg)
+
+#define field_avail(type, fld, sz) (offsetof(type, fld) +		\
+				    sizeof(((type *)0)->fld) <= (sz))
+#define MLX5_IB_DEFAULT_UIDX 0xffffff
+#define MLX5_USER_ASSIGNED_UIDX_MASK __mlx5_mask(qpc, user_index)
+
+#define MLX5_MKEY_PAGE_SHIFT_MASK __mlx5_mask(mkc, log_page_size)
+
+enum {
+	MLX5_IB_MMAP_CMD_SHIFT	= 8,
+	MLX5_IB_MMAP_CMD_MASK	= 0xff,
+};
+
+enum {
+	MLX5_RES_SCAT_DATA32_CQE	= 0x1,
+	MLX5_RES_SCAT_DATA64_CQE	= 0x2,
+	MLX5_REQ_SCAT_DATA32_CQE	= 0x11,
+	MLX5_REQ_SCAT_DATA64_CQE	= 0x22,
+};
+
+enum mlx5_ib_mad_ifc_flags {
+	MLX5_MAD_IFC_IGNORE_MKEY	= 1,
+	MLX5_MAD_IFC_IGNORE_BKEY	= 2,
+	MLX5_MAD_IFC_NET_VIEW		= 4,
+};
+
+enum {
+	MLX5_CROSS_CHANNEL_BFREG         = 0,
+};
+
+enum {
+	MLX5_CQE_VERSION_V0,
+	MLX5_CQE_VERSION_V1,
+};
+
+enum {
+	MLX5_TM_MAX_RNDV_MSG_SIZE	= 64,
+	MLX5_TM_MAX_SGE			= 1,
+};
+
+enum {
+	MLX5_IB_INVALID_UAR_INDEX	= BIT(31),
+	MLX5_IB_INVALID_BFREG		= BIT(31),
+};
+
+enum {
+	MLX5_MAX_MEMIC_PAGES = 0x100,
+	MLX5_MEMIC_ALLOC_SIZE_MASK = 0x3f,
+};
+
+enum {
+	MLX5_MEMIC_BASE_ALIGN	= 6,
+	MLX5_MEMIC_BASE_SIZE	= 1 << MLX5_MEMIC_BASE_ALIGN,
+};
+
+struct mlx5_ib_vma_private_data {
+	struct list_head list;
+	struct vm_area_struct *vma;
+	/* protect vma_private_list add/del */
+	struct mutex *vma_private_list_mutex;
+};
+
+struct mlx5_ib_ucontext {
+	struct ib_ucontext	ibucontext;
+	struct list_head	db_page_list;
+
+	/* protect doorbell record alloc/free
+	 */
+	struct mutex		db_page_mutex;
+	struct mlx5_bfreg_info	bfregi;
+	u8			cqe_version;
+	/* Transport Domain number */
+	u32			tdn;
+	struct list_head	vma_private_list;
+	/* protect vma_private_list add/del */
+	struct mutex		vma_private_list_mutex;
+
+	u64			lib_caps;
+	DECLARE_BITMAP(dm_pages, MLX5_MAX_MEMIC_PAGES);
+	u16			devx_uid;
+	/* For RoCE LAG TX affinity */
+	atomic_t		tx_port_affinity;
+};
+
+static inline struct mlx5_ib_ucontext *to_mucontext(struct ib_ucontext *ibucontext)
+{
+	return container_of(ibucontext, struct mlx5_ib_ucontext, ibucontext);
+}
+
+struct mlx5_ib_pd {
+	struct ib_pd		ibpd;
+	u32			pdn;
+};
+
+#define MLX5_IB_FLOW_MCAST_PRIO		(MLX5_BY_PASS_NUM_PRIOS - 1)
+#define MLX5_IB_FLOW_LAST_PRIO		(MLX5_BY_PASS_NUM_REGULAR_PRIOS - 1)
+#if (MLX5_IB_FLOW_LAST_PRIO <= 0)
+#error "Invalid number of bypass priorities"
+#endif
+#define MLX5_IB_FLOW_LEFTOVERS_PRIO	(MLX5_IB_FLOW_MCAST_PRIO + 1)
+
+#define MLX5_IB_NUM_FLOW_FT		(MLX5_IB_FLOW_LEFTOVERS_PRIO + 1)
+#define MLX5_IB_NUM_SNIFFER_FTS		2
+#define MLX5_IB_NUM_EGRESS_FTS		1
+struct mlx5_ib_flow_prio {
+	struct mlx5_flow_table		*flow_table;
+	unsigned int			refcount;
+};
+
+struct mlx5_ib_flow_handler {
+	struct list_head		list;
+	struct ib_flow			ibflow;
+	struct mlx5_ib_flow_prio	*prio;
+	struct mlx5_flow_handle		*rule;
+	struct ib_counters		*ibcounters;
+	struct mlx5_ib_dev		*dev;
+	struct mlx5_ib_flow_matcher	*flow_matcher;
+};
+
+struct mlx5_ib_flow_matcher {
+	struct mlx5_ib_match_params matcher_mask;
+	int			mask_len;
+	enum mlx5_ib_flow_type	flow_type;
+	u16			priority;
+	struct mlx5_core_dev	*mdev;
+	atomic_t		usecnt;
+	u8			match_criteria_enable;
+};
+
+struct mlx5_ib_flow_db {
+	struct mlx5_ib_flow_prio	prios[MLX5_IB_NUM_FLOW_FT];
+	struct mlx5_ib_flow_prio	sniffer[MLX5_IB_NUM_SNIFFER_FTS];
+	struct mlx5_ib_flow_prio	egress[MLX5_IB_NUM_EGRESS_FTS];
+	struct mlx5_flow_table		*lag_demux_ft;
+	/* Protect flow steering bypass flow tables
+	 * when add/del flow rules.
+	 * only single add/removal of flow steering rule could be done
+	 * simultaneously.
+	 */
+	struct mutex			lock;
+};
+
+/* Use macros here so that don't have to duplicate
+ * enum ib_send_flags and enum ib_qp_type for low-level driver
+ */
+
+#define MLX5_IB_SEND_UMR_ENABLE_MR	       (IB_SEND_RESERVED_START << 0)
+#define MLX5_IB_SEND_UMR_DISABLE_MR	       (IB_SEND_RESERVED_START << 1)
+#define MLX5_IB_SEND_UMR_FAIL_IF_FREE	       (IB_SEND_RESERVED_START << 2)
+#define MLX5_IB_SEND_UMR_UPDATE_XLT	       (IB_SEND_RESERVED_START << 3)
+#define MLX5_IB_SEND_UMR_UPDATE_TRANSLATION    (IB_SEND_RESERVED_START << 4)
+#define MLX5_IB_SEND_UMR_UPDATE_PD_ACCESS       IB_SEND_RESERVED_END
+
+#define MLX5_IB_QPT_REG_UMR	IB_QPT_RESERVED1
+/*
+ * IB_QPT_GSI creates the software wrapper around GSI, and MLX5_IB_QPT_HW_GSI
+ * creates the actual hardware QP.
+ */
+#define MLX5_IB_QPT_HW_GSI	IB_QPT_RESERVED2
+#define MLX5_IB_QPT_DCI		IB_QPT_RESERVED3
+#define MLX5_IB_QPT_DCT		IB_QPT_RESERVED4
+#define MLX5_IB_WR_UMR		IB_WR_RESERVED1
+
+#define MLX5_IB_UMR_OCTOWORD	       16
+#define MLX5_IB_UMR_XLT_ALIGNMENT      64
+
+#define MLX5_IB_UPD_XLT_ZAP	      BIT(0)
+#define MLX5_IB_UPD_XLT_ENABLE	      BIT(1)
+#define MLX5_IB_UPD_XLT_ATOMIC	      BIT(2)
+#define MLX5_IB_UPD_XLT_ADDR	      BIT(3)
+#define MLX5_IB_UPD_XLT_PD	      BIT(4)
+#define MLX5_IB_UPD_XLT_ACCESS	      BIT(5)
+#define MLX5_IB_UPD_XLT_INDIRECT      BIT(6)
+
+/* Private QP creation flags to be passed in ib_qp_init_attr.create_flags.
+ *
+ * These flags are intended for internal use by the mlx5_ib driver, and they
+ * rely on the range reserved for that use in the ib_qp_create_flags enum.
+ */
+
+/* Create a UD QP whose source QP number is 1 */
+static inline enum ib_qp_create_flags mlx5_ib_create_qp_sqpn_qp1(void)
+{
+	return IB_QP_CREATE_RESERVED_START;
+}
+
+struct wr_list {
+	u16	opcode;
+	u16	next;
+};
+
+enum mlx5_ib_rq_flags {
+	MLX5_IB_RQ_CVLAN_STRIPPING	= 1 << 0,
+	MLX5_IB_RQ_PCI_WRITE_END_PADDING	= 1 << 1,
+};
+
+struct mlx5_ib_wq {
+	u64		       *wrid;
+	u32		       *wr_data;
+	struct wr_list	       *w_list;
+	unsigned	       *wqe_head;
+	u16		        unsig_count;
+
+	/* serialize post to the work queue
+	 */
+	spinlock_t		lock;
+	int			wqe_cnt;
+	int			max_post;
+	int			max_gs;
+	int			offset;
+	int			wqe_shift;
+	unsigned		head;
+	unsigned		tail;
+	u16			cur_post;
+	u16			last_poll;
+	void		       *qend;
+};
+
+enum mlx5_ib_wq_flags {
+	MLX5_IB_WQ_FLAGS_DELAY_DROP = 0x1,
+	MLX5_IB_WQ_FLAGS_STRIDING_RQ = 0x2,
+};
+
+#define MLX5_MIN_SINGLE_WQE_LOG_NUM_STRIDES 9
+#define MLX5_MAX_SINGLE_WQE_LOG_NUM_STRIDES 16
+#define MLX5_MIN_SINGLE_STRIDE_LOG_NUM_BYTES 6
+#define MLX5_MAX_SINGLE_STRIDE_LOG_NUM_BYTES 13
+
+struct mlx5_ib_rwq {
+	struct ib_wq		ibwq;
+	struct mlx5_core_qp	core_qp;
+	u32			rq_num_pas;
+	u32			log_rq_stride;
+	u32			log_rq_size;
+	u32			rq_page_offset;
+	u32			log_page_size;
+	u32			log_num_strides;
+	u32			two_byte_shift_en;
+	u32			single_stride_log_num_of_bytes;
+	struct ib_umem		*umem;
+	size_t			buf_size;
+	unsigned int		page_shift;
+	int			create_type;
+	struct mlx5_db		db;
+	u32			user_index;
+	u32			wqe_count;
+	u32			wqe_shift;
+	int			wq_sig;
+	u32			create_flags; /* Use enum mlx5_ib_wq_flags */
+};
+
+enum {
+	MLX5_QP_USER,
+	MLX5_QP_KERNEL,
+	MLX5_QP_EMPTY
+};
+
+enum {
+	MLX5_WQ_USER,
+	MLX5_WQ_KERNEL
+};
+
+struct mlx5_ib_rwq_ind_table {
+	struct ib_rwq_ind_table ib_rwq_ind_tbl;
+	u32			rqtn;
+};
+
+struct mlx5_ib_ubuffer {
+	struct ib_umem	       *umem;
+	int			buf_size;
+	u64			buf_addr;
+};
+
+struct mlx5_ib_qp_base {
+	struct mlx5_ib_qp	*container_mibqp;
+	struct mlx5_core_qp	mqp;
+	struct mlx5_ib_ubuffer	ubuffer;
+};
+
+struct mlx5_ib_qp_trans {
+	struct mlx5_ib_qp_base	base;
+	u16			xrcdn;
+	u8			alt_port;
+	u8			atomic_rd_en;
+	u8			resp_depth;
+};
+
+struct mlx5_ib_rss_qp {
+	u32	tirn;
+};
+
+struct mlx5_ib_rq {
+	struct mlx5_ib_qp_base base;
+	struct mlx5_ib_wq	*rq;
+	struct mlx5_ib_ubuffer	ubuffer;
+	struct mlx5_db		*doorbell;
+	u32			tirn;
+	u8			state;
+	u32			flags;
+};
+
+struct mlx5_ib_sq {
+	struct mlx5_ib_qp_base base;
+	struct mlx5_ib_wq	*sq;
+	struct mlx5_ib_ubuffer  ubuffer;
+	struct mlx5_db		*doorbell;
+	struct mlx5_flow_handle	*flow_rule;
+	u32			tisn;
+	u8			state;
+};
+
+struct mlx5_ib_raw_packet_qp {
+	struct mlx5_ib_sq sq;
+	struct mlx5_ib_rq rq;
+};
+
+struct mlx5_bf {
+	int			buf_size;
+	unsigned long		offset;
+	struct mlx5_sq_bfreg   *bfreg;
+};
+
+struct mlx5_ib_dct {
+	struct mlx5_core_dct    mdct;
+	u32                     *in;
+};
+
+struct mlx5_ib_qp {
+	struct ib_qp		ibqp;
+	union {
+		struct mlx5_ib_qp_trans trans_qp;
+		struct mlx5_ib_raw_packet_qp raw_packet_qp;
+		struct mlx5_ib_rss_qp rss_qp;
+		struct mlx5_ib_dct dct;
+	};
+	struct mlx5_frag_buf	buf;
+
+	struct mlx5_db		db;
+	struct mlx5_ib_wq	rq;
+
+	u8			sq_signal_bits;
+	u8			next_fence;
+	struct mlx5_ib_wq	sq;
+
+	/* serialize qp state modifications
+	 */
+	struct mutex		mutex;
+	u32			flags;
+	u8			port;
+	u8			state;
+	int			wq_sig;
+	int			scat_cqe;
+	int			max_inline_data;
+	struct mlx5_bf	        bf;
+	int			has_rq;
+
+	/* only for user space QPs. For kernel
+	 * we have it from the bf object
+	 */
+	int			bfregn;
+
+	int			create_type;
+
+	/* Store signature errors */
+	bool			signature_en;
+
+	struct list_head	qps_list;
+	struct list_head	cq_recv_list;
+	struct list_head	cq_send_list;
+	struct mlx5_rate_limit	rl;
+	u32                     underlay_qpn;
+	bool			tunnel_offload_en;
+	/* storage for qp sub type when core qp type is IB_QPT_DRIVER */
+	enum ib_qp_type		qp_sub_type;
+};
+
+struct mlx5_ib_cq_buf {
+	struct mlx5_frag_buf_ctrl fbc;
+	struct ib_umem		*umem;
+	int			cqe_size;
+	int			nent;
+};
+
+enum mlx5_ib_qp_flags {
+	MLX5_IB_QP_LSO                          = IB_QP_CREATE_IPOIB_UD_LSO,
+	MLX5_IB_QP_BLOCK_MULTICAST_LOOPBACK     = IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK,
+	MLX5_IB_QP_CROSS_CHANNEL            = IB_QP_CREATE_CROSS_CHANNEL,
+	MLX5_IB_QP_MANAGED_SEND             = IB_QP_CREATE_MANAGED_SEND,
+	MLX5_IB_QP_MANAGED_RECV             = IB_QP_CREATE_MANAGED_RECV,
+	MLX5_IB_QP_SIGNATURE_HANDLING           = 1 << 5,
+	/* QP uses 1 as its source QP number */
+	MLX5_IB_QP_SQPN_QP1			= 1 << 6,
+	MLX5_IB_QP_CAP_SCATTER_FCS		= 1 << 7,
+	MLX5_IB_QP_RSS				= 1 << 8,
+	MLX5_IB_QP_CVLAN_STRIPPING		= 1 << 9,
+	MLX5_IB_QP_UNDERLAY			= 1 << 10,
+	MLX5_IB_QP_PCI_WRITE_END_PADDING	= 1 << 11,
+	MLX5_IB_QP_TUNNEL_OFFLOAD		= 1 << 12,
+};
+
+struct mlx5_umr_wr {
+	struct ib_send_wr		wr;
+	u64				virt_addr;
+	u64				offset;
+	struct ib_pd		       *pd;
+	unsigned int			page_shift;
+	unsigned int			xlt_size;
+	u64				length;
+	int				access_flags;
+	u32				mkey;
+	u8				ignore_free_state:1;
+};
+
+static inline const struct mlx5_umr_wr *umr_wr(const struct ib_send_wr *wr)
+{
+	return container_of(wr, struct mlx5_umr_wr, wr);
+}
+
+struct mlx5_shared_mr_info {
+	int mr_id;
+	struct ib_umem		*umem;
+};
+
+enum mlx5_ib_cq_pr_flags {
+	MLX5_IB_CQ_PR_FLAGS_CQE_128_PAD	= 1 << 0,
+};
+
+struct mlx5_ib_cq {
+	struct ib_cq		ibcq;
+	struct mlx5_core_cq	mcq;
+	struct mlx5_ib_cq_buf	buf;
+	struct mlx5_db		db;
+
+	/* serialize access to the CQ
+	 */
+	spinlock_t		lock;
+
+	/* protect resize cq
+	 */
+	struct mutex		resize_mutex;
+	struct mlx5_ib_cq_buf  *resize_buf;
+	struct ib_umem	       *resize_umem;
+	int			cqe_size;
+	struct list_head	list_send_qp;
+	struct list_head	list_recv_qp;
+	u32			create_flags;
+	struct list_head	wc_list;
+	enum ib_cq_notify_flags notify_flags;
+	struct work_struct	notify_work;
+	u16			private_flags; /* Use mlx5_ib_cq_pr_flags */
+};
+
+struct mlx5_ib_wc {
+	struct ib_wc wc;
+	struct list_head list;
+};
+
+struct mlx5_ib_srq {
+	struct ib_srq		ibsrq;
+	struct mlx5_core_srq	msrq;
+	struct mlx5_frag_buf	buf;
+	struct mlx5_db		db;
+	u64		       *wrid;
+	/* protect SRQ hanlding
+	 */
+	spinlock_t		lock;
+	int			head;
+	int			tail;
+	u16			wqe_ctr;
+	struct ib_umem	       *umem;
+	/* serialize arming a SRQ
+	 */
+	struct mutex		mutex;
+	int			wq_sig;
+};
+
+struct mlx5_ib_xrcd {
+	struct ib_xrcd		ibxrcd;
+	u32			xrcdn;
+};
+
+enum mlx5_ib_mtt_access_flags {
+	MLX5_IB_MTT_READ  = (1 << 0),
+	MLX5_IB_MTT_WRITE = (1 << 1),
+};
+
+struct mlx5_ib_dm {
+	struct ib_dm		ibdm;
+	phys_addr_t		dev_addr;
+};
+
+#define MLX5_IB_MTT_PRESENT (MLX5_IB_MTT_READ | MLX5_IB_MTT_WRITE)
+
+#define MLX5_IB_DM_ALLOWED_ACCESS (IB_ACCESS_LOCAL_WRITE   |\
+				   IB_ACCESS_REMOTE_WRITE  |\
+				   IB_ACCESS_REMOTE_READ   |\
+				   IB_ACCESS_REMOTE_ATOMIC |\
+				   IB_ZERO_BASED)
+
+struct mlx5_ib_mr {
+	struct ib_mr		ibmr;
+	void			*descs;
+	dma_addr_t		desc_map;
+	int			ndescs;
+	int			max_descs;
+	int			desc_size;
+	int			access_mode;
+	struct mlx5_core_mkey	mmkey;
+	struct ib_umem	       *umem;
+	struct mlx5_shared_mr_info	*smr_info;
+	struct list_head	list;
+	int			order;
+	bool			allocated_from_cache;
+	int			npages;
+	struct mlx5_ib_dev     *dev;
+	u32 out[MLX5_ST_SZ_DW(create_mkey_out)];
+	struct mlx5_core_sig_ctx    *sig;
+	int			live;
+	void			*descs_alloc;
+	int			access_flags; /* Needed for rereg MR */
+
+	struct mlx5_ib_mr      *parent;
+	atomic_t		num_leaf_free;
+	wait_queue_head_t       q_leaf_free;
+};
+
+struct mlx5_ib_mw {
+	struct ib_mw		ibmw;
+	struct mlx5_core_mkey	mmkey;
+	int			ndescs;
+};
+
+struct mlx5_ib_umr_context {
+	struct ib_cqe		cqe;
+	enum ib_wc_status	status;
+	struct completion	done;
+};
+
+struct umr_common {
+	struct ib_pd	*pd;
+	struct ib_cq	*cq;
+	struct ib_qp	*qp;
+	/* control access to UMR QP
+	 */
+	struct semaphore	sem;
+};
+
+enum {
+	MLX5_FMR_INVALID,
+	MLX5_FMR_VALID,
+	MLX5_FMR_BUSY,
+};
+
+struct mlx5_cache_ent {
+	struct list_head	head;
+	/* sync access to the cahce entry
+	 */
+	spinlock_t		lock;
+
+
+	struct dentry	       *dir;
+	char                    name[4];
+	u32                     order;
+	u32			xlt;
+	u32			access_mode;
+	u32			page;
+
+	u32			size;
+	u32                     cur;
+	u32                     miss;
+	u32			limit;
+
+	struct dentry          *fsize;
+	struct dentry          *fcur;
+	struct dentry          *fmiss;
+	struct dentry          *flimit;
+
+	struct mlx5_ib_dev     *dev;
+	struct work_struct	work;
+	struct delayed_work	dwork;
+	int			pending;
+	struct completion	compl;
+};
+
+struct mlx5_mr_cache {
+	struct workqueue_struct *wq;
+	struct mlx5_cache_ent	ent[MAX_MR_CACHE_ENTRIES];
+	int			stopped;
+	struct dentry		*root;
+	unsigned long		last_add;
+};
+
+struct mlx5_ib_gsi_qp;
+
+struct mlx5_ib_port_resources {
+	struct mlx5_ib_resources *devr;
+	struct mlx5_ib_gsi_qp *gsi;
+	struct work_struct pkey_change_work;
+};
+
+struct mlx5_ib_resources {
+	struct ib_cq	*c0;
+	struct ib_xrcd	*x0;
+	struct ib_xrcd	*x1;
+	struct ib_pd	*p0;
+	struct ib_srq	*s0;
+	struct ib_srq	*s1;
+	struct mlx5_ib_port_resources ports[2];
+	/* Protects changes to the port resources */
+	struct mutex	mutex;
+};
+
+struct mlx5_ib_counters {
+	const char **names;
+	size_t *offsets;
+	u32 num_q_counters;
+	u32 num_cong_counters;
+	u32 num_ext_ppcnt_counters;
+	u16 set_id;
+	bool set_id_valid;
+};
+
+struct mlx5_ib_multiport_info;
+
+struct mlx5_ib_multiport {
+	struct mlx5_ib_multiport_info *mpi;
+	/* To be held when accessing the multiport info */
+	spinlock_t mpi_lock;
+};
+
+struct mlx5_ib_port {
+	struct mlx5_ib_counters cnts;
+	struct mlx5_ib_multiport mp;
+	struct mlx5_ib_dbg_cc_params	*dbg_cc_params;
+};
+
+struct mlx5_roce {
+	/* Protect mlx5_ib_get_netdev from invoking dev_hold() with a NULL
+	 * netdev pointer
+	 */
+	rwlock_t		netdev_lock;
+	struct net_device	*netdev;
+	struct notifier_block	nb;
+	atomic_t		tx_port_affinity;
+	enum ib_port_state last_port_state;
+	struct mlx5_ib_dev	*dev;
+	u8			native_port_num;
+};
+
+struct mlx5_ib_dbg_param {
+	int			offset;
+	struct mlx5_ib_dev	*dev;
+	struct dentry		*dentry;
+	u8			port_num;
+};
+
+enum mlx5_ib_dbg_cc_types {
+	MLX5_IB_DBG_CC_RP_CLAMP_TGT_RATE,
+	MLX5_IB_DBG_CC_RP_CLAMP_TGT_RATE_ATI,
+	MLX5_IB_DBG_CC_RP_TIME_RESET,
+	MLX5_IB_DBG_CC_RP_BYTE_RESET,
+	MLX5_IB_DBG_CC_RP_THRESHOLD,
+	MLX5_IB_DBG_CC_RP_AI_RATE,
+	MLX5_IB_DBG_CC_RP_HAI_RATE,
+	MLX5_IB_DBG_CC_RP_MIN_DEC_FAC,
+	MLX5_IB_DBG_CC_RP_MIN_RATE,
+	MLX5_IB_DBG_CC_RP_RATE_TO_SET_ON_FIRST_CNP,
+	MLX5_IB_DBG_CC_RP_DCE_TCP_G,
+	MLX5_IB_DBG_CC_RP_DCE_TCP_RTT,
+	MLX5_IB_DBG_CC_RP_RATE_REDUCE_MONITOR_PERIOD,
+	MLX5_IB_DBG_CC_RP_INITIAL_ALPHA_VALUE,
+	MLX5_IB_DBG_CC_RP_GD,
+	MLX5_IB_DBG_CC_NP_CNP_DSCP,
+	MLX5_IB_DBG_CC_NP_CNP_PRIO_MODE,
+	MLX5_IB_DBG_CC_NP_CNP_PRIO,
+	MLX5_IB_DBG_CC_MAX,
+};
+
+struct mlx5_ib_dbg_cc_params {
+	struct dentry			*root;
+	struct mlx5_ib_dbg_param	params[MLX5_IB_DBG_CC_MAX];
+};
+
+enum {
+	MLX5_MAX_DELAY_DROP_TIMEOUT_MS = 100,
+};
+
+struct mlx5_ib_dbg_delay_drop {
+	struct dentry		*dir_debugfs;
+	struct dentry		*rqs_cnt_debugfs;
+	struct dentry		*events_cnt_debugfs;
+	struct dentry		*timeout_debugfs;
+};
+
+struct mlx5_ib_delay_drop {
+	struct mlx5_ib_dev     *dev;
+	struct work_struct	delay_drop_work;
+	/* serialize setting of delay drop */
+	struct mutex		lock;
+	u32			timeout;
+	bool			activate;
+	atomic_t		events_cnt;
+	atomic_t		rqs_cnt;
+	struct mlx5_ib_dbg_delay_drop *dbg;
+};
+
+enum mlx5_ib_stages {
+	MLX5_IB_STAGE_INIT,
+	MLX5_IB_STAGE_FLOW_DB,
+	MLX5_IB_STAGE_CAPS,
+	MLX5_IB_STAGE_NON_DEFAULT_CB,
+	MLX5_IB_STAGE_ROCE,
+	MLX5_IB_STAGE_DEVICE_RESOURCES,
+	MLX5_IB_STAGE_ODP,
+	MLX5_IB_STAGE_COUNTERS,
+	MLX5_IB_STAGE_CONG_DEBUGFS,
+	MLX5_IB_STAGE_UAR,
+	MLX5_IB_STAGE_BFREG,
+	MLX5_IB_STAGE_PRE_IB_REG_UMR,
+	MLX5_IB_STAGE_SPECS,
+	MLX5_IB_STAGE_IB_REG,
+	MLX5_IB_STAGE_POST_IB_REG_UMR,
+	MLX5_IB_STAGE_DELAY_DROP,
+	MLX5_IB_STAGE_CLASS_ATTR,
+	MLX5_IB_STAGE_REP_REG,
+	MLX5_IB_STAGE_MAX,
+};
+
+struct mlx5_ib_stage {
+	int (*init)(struct mlx5_ib_dev *dev);
+	void (*cleanup)(struct mlx5_ib_dev *dev);
+};
+
+#define STAGE_CREATE(_stage, _init, _cleanup) \
+	.stage[_stage] = {.init = _init, .cleanup = _cleanup}
+
+struct mlx5_ib_profile {
+	struct mlx5_ib_stage stage[MLX5_IB_STAGE_MAX];
+};
+
+struct mlx5_ib_multiport_info {
+	struct list_head list;
+	struct mlx5_ib_dev *ibdev;
+	struct mlx5_core_dev *mdev;
+	struct completion unref_comp;
+	u64 sys_image_guid;
+	u32 mdev_refcnt;
+	bool is_master;
+	bool unaffiliate;
+};
+
+struct mlx5_ib_flow_action {
+	struct ib_flow_action		ib_action;
+	union {
+		struct {
+			u64			    ib_flags;
+			struct mlx5_accel_esp_xfrm *ctx;
+		} esp_aes_gcm;
+	};
+};
+
+struct mlx5_memic {
+	struct mlx5_core_dev *dev;
+	spinlock_t		memic_lock;
+	DECLARE_BITMAP(memic_alloc_pages, MLX5_MAX_MEMIC_PAGES);
+};
+
+struct mlx5_read_counters_attr {
+	struct mlx5_fc *hw_cntrs_hndl;
+	u64 *out;
+	u32 flags;
+};
+
+enum mlx5_ib_counters_type {
+	MLX5_IB_COUNTERS_FLOW,
+};
+
+struct mlx5_ib_mcounters {
+	struct ib_counters ibcntrs;
+	enum mlx5_ib_counters_type type;
+	/* number of counters supported for this counters type */
+	u32 counters_num;
+	struct mlx5_fc *hw_cntrs_hndl;
+	/* read function for this counters type */
+	int (*read_counters)(struct ib_device *ibdev,
+			     struct mlx5_read_counters_attr *read_attr);
+	/* max index set as part of create_flow */
+	u32 cntrs_max_index;
+	/* number of counters data entries (<description,index> pair) */
+	u32 ncounters;
+	/* counters data array for descriptions and indexes */
+	struct mlx5_ib_flow_counters_desc *counters_data;
+	/* protects access to mcounters internal data */
+	struct mutex mcntrs_mutex;
+};
+
+static inline struct mlx5_ib_mcounters *
+to_mcounters(struct ib_counters *ibcntrs)
+{
+	return container_of(ibcntrs, struct mlx5_ib_mcounters, ibcntrs);
+}
+
+struct mlx5_ib_dev {
+	struct ib_device		ib_dev;
+	const struct uverbs_object_tree_def *driver_trees[6];
+	struct mlx5_core_dev		*mdev;
+	struct mlx5_roce		roce[MLX5_MAX_PORTS];
+	int				num_ports;
+	/* serialize update of capability mask
+	 */
+	struct mutex			cap_mask_mutex;
+	bool				ib_active;
+	struct umr_common		umrc;
+	/* sync used page count stats
+	 */
+	struct mlx5_ib_resources	devr;
+	struct mlx5_mr_cache		cache;
+	struct timer_list		delay_timer;
+	/* Prevents soft lock on massive reg MRs */
+	struct mutex			slow_path_mutex;
+	int				fill_delay;
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
+	struct ib_odp_caps	odp_caps;
+	u64			odp_max_size;
+	/*
+	 * Sleepable RCU that prevents destruction of MRs while they are still
+	 * being used by a page fault handler.
+	 */
+	struct srcu_struct      mr_srcu;
+	u32			null_mkey;
+#endif
+	struct mlx5_ib_flow_db	*flow_db;
+	/* protect resources needed as part of reset flow */
+	spinlock_t		reset_flow_resource_lock;
+	struct list_head	qp_list;
+	/* Array with num_ports elements */
+	struct mlx5_ib_port	*port;
+	struct mlx5_sq_bfreg	bfreg;
+	struct mlx5_sq_bfreg	fp_bfreg;
+	struct mlx5_ib_delay_drop	delay_drop;
+	const struct mlx5_ib_profile	*profile;
+	struct mlx5_eswitch_rep		*rep;
+
+	/* protect the user_td */
+	struct mutex		lb_mutex;
+	u32			user_td;
+	u8			umr_fence;
+	struct list_head	ib_dev_list;
+	u64			sys_image_guid;
+	struct mlx5_memic	memic;
+};
+
+static inline struct mlx5_ib_cq *to_mibcq(struct mlx5_core_cq *mcq)
+{
+	return container_of(mcq, struct mlx5_ib_cq, mcq);
+}
+
+static inline struct mlx5_ib_xrcd *to_mxrcd(struct ib_xrcd *ibxrcd)
+{
+	return container_of(ibxrcd, struct mlx5_ib_xrcd, ibxrcd);
+}
+
+static inline struct mlx5_ib_dev *to_mdev(struct ib_device *ibdev)
+{
+	return container_of(ibdev, struct mlx5_ib_dev, ib_dev);
+}
+
+static inline struct mlx5_ib_cq *to_mcq(struct ib_cq *ibcq)
+{
+	return container_of(ibcq, struct mlx5_ib_cq, ibcq);
+}
+
+static inline struct mlx5_ib_qp *to_mibqp(struct mlx5_core_qp *mqp)
+{
+	return container_of(mqp, struct mlx5_ib_qp_base, mqp)->container_mibqp;
+}
+
+static inline struct mlx5_ib_rwq *to_mibrwq(struct mlx5_core_qp *core_qp)
+{
+	return container_of(core_qp, struct mlx5_ib_rwq, core_qp);
+}
+
+static inline struct mlx5_ib_mr *to_mibmr(struct mlx5_core_mkey *mmkey)
+{
+	return container_of(mmkey, struct mlx5_ib_mr, mmkey);
+}
+
+static inline struct mlx5_ib_pd *to_mpd(struct ib_pd *ibpd)
+{
+	return container_of(ibpd, struct mlx5_ib_pd, ibpd);
+}
+
+static inline struct mlx5_ib_srq *to_msrq(struct ib_srq *ibsrq)
+{
+	return container_of(ibsrq, struct mlx5_ib_srq, ibsrq);
+}
+
+static inline struct mlx5_ib_qp *to_mqp(struct ib_qp *ibqp)
+{
+	return container_of(ibqp, struct mlx5_ib_qp, ibqp);
+}
+
+static inline struct mlx5_ib_rwq *to_mrwq(struct ib_wq *ibwq)
+{
+	return container_of(ibwq, struct mlx5_ib_rwq, ibwq);
+}
+
+static inline struct mlx5_ib_rwq_ind_table *to_mrwq_ind_table(struct ib_rwq_ind_table *ib_rwq_ind_tbl)
+{
+	return container_of(ib_rwq_ind_tbl, struct mlx5_ib_rwq_ind_table, ib_rwq_ind_tbl);
+}
+
+static inline struct mlx5_ib_srq *to_mibsrq(struct mlx5_core_srq *msrq)
+{
+	return container_of(msrq, struct mlx5_ib_srq, msrq);
+}
+
+static inline struct mlx5_ib_dm *to_mdm(struct ib_dm *ibdm)
+{
+	return container_of(ibdm, struct mlx5_ib_dm, ibdm);
+}
+
+static inline struct mlx5_ib_mr *to_mmr(struct ib_mr *ibmr)
+{
+	return container_of(ibmr, struct mlx5_ib_mr, ibmr);
+}
+
+static inline struct mlx5_ib_mw *to_mmw(struct ib_mw *ibmw)
+{
+	return container_of(ibmw, struct mlx5_ib_mw, ibmw);
+}
+
+static inline struct mlx5_ib_flow_action *
+to_mflow_act(struct ib_flow_action *ibact)
+{
+	return container_of(ibact, struct mlx5_ib_flow_action, ib_action);
+}
+
+int mlx5_ib_db_map_user(struct mlx5_ib_ucontext *context, unsigned long virt,
+			struct mlx5_db *db);
+void mlx5_ib_db_unmap_user(struct mlx5_ib_ucontext *context, struct mlx5_db *db);
+void __mlx5_ib_cq_clean(struct mlx5_ib_cq *cq, u32 qpn, struct mlx5_ib_srq *srq);
+void mlx5_ib_cq_clean(struct mlx5_ib_cq *cq, u32 qpn, struct mlx5_ib_srq *srq);
+void mlx5_ib_free_srq_wqe(struct mlx5_ib_srq *srq, int wqe_index);
+int mlx5_MAD_IFC(struct mlx5_ib_dev *dev, int ignore_mkey, int ignore_bkey,
+		 u8 port, const struct ib_wc *in_wc, const struct ib_grh *in_grh,
+		 const void *in_mad, void *response_mad);
+struct ib_ah *mlx5_ib_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr,
+				struct ib_udata *udata);
+int mlx5_ib_query_ah(struct ib_ah *ibah, struct rdma_ah_attr *ah_attr);
+int mlx5_ib_destroy_ah(struct ib_ah *ah);
+struct ib_srq *mlx5_ib_create_srq(struct ib_pd *pd,
+				  struct ib_srq_init_attr *init_attr,
+				  struct ib_udata *udata);
+int mlx5_ib_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
+		       enum ib_srq_attr_mask attr_mask, struct ib_udata *udata);
+int mlx5_ib_query_srq(struct ib_srq *ibsrq, struct ib_srq_attr *srq_attr);
+int mlx5_ib_destroy_srq(struct ib_srq *srq);
+int mlx5_ib_post_srq_recv(struct ib_srq *ibsrq, const struct ib_recv_wr *wr,
+			  const struct ib_recv_wr **bad_wr);
+struct ib_qp *mlx5_ib_create_qp(struct ib_pd *pd,
+				struct ib_qp_init_attr *init_attr,
+				struct ib_udata *udata);
+int mlx5_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
+		      int attr_mask, struct ib_udata *udata);
+int mlx5_ib_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *qp_attr, int qp_attr_mask,
+		     struct ib_qp_init_attr *qp_init_attr);
+int mlx5_ib_destroy_qp(struct ib_qp *qp);
+void mlx5_ib_drain_sq(struct ib_qp *qp);
+void mlx5_ib_drain_rq(struct ib_qp *qp);
+int mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr,
+		      const struct ib_send_wr **bad_wr);
+int mlx5_ib_post_recv(struct ib_qp *ibqp, const struct ib_recv_wr *wr,
+		      const struct ib_recv_wr **bad_wr);
+void *mlx5_get_send_wqe(struct mlx5_ib_qp *qp, int n);
+int mlx5_ib_read_user_wqe(struct mlx5_ib_qp *qp, int send, int wqe_index,
+			  void *buffer, u32 length,
+			  struct mlx5_ib_qp_base *base);
+struct ib_cq *mlx5_ib_create_cq(struct ib_device *ibdev,
+				const struct ib_cq_init_attr *attr,
+				struct ib_ucontext *context,
+				struct ib_udata *udata);
+int mlx5_ib_destroy_cq(struct ib_cq *cq);
+int mlx5_ib_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *wc);
+int mlx5_ib_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags flags);
+int mlx5_ib_modify_cq(struct ib_cq *cq, u16 cq_count, u16 cq_period);
+int mlx5_ib_resize_cq(struct ib_cq *ibcq, int entries, struct ib_udata *udata);
+struct ib_mr *mlx5_ib_get_dma_mr(struct ib_pd *pd, int acc);
+struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
+				  u64 virt_addr, int access_flags,
+				  struct ib_udata *udata);
+struct ib_mw *mlx5_ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
+			       struct ib_udata *udata);
+int mlx5_ib_dealloc_mw(struct ib_mw *mw);
+int mlx5_ib_update_xlt(struct mlx5_ib_mr *mr, u64 idx, int npages,
+		       int page_shift, int flags);
+struct mlx5_ib_mr *mlx5_ib_alloc_implicit_mr(struct mlx5_ib_pd *pd,
+					     int access_flags);
+void mlx5_ib_free_implicit_mr(struct mlx5_ib_mr *mr);
+int mlx5_ib_rereg_user_mr(struct ib_mr *ib_mr, int flags, u64 start,
+			  u64 length, u64 virt_addr, int access_flags,
+			  struct ib_pd *pd, struct ib_udata *udata);
+int mlx5_ib_dereg_mr(struct ib_mr *ibmr);
+struct ib_mr *mlx5_ib_alloc_mr(struct ib_pd *pd,
+			       enum ib_mr_type mr_type,
+			       u32 max_num_sg);
+int mlx5_ib_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg, int sg_nents,
+		      unsigned int *sg_offset);
+int mlx5_ib_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num,
+			const struct ib_wc *in_wc, const struct ib_grh *in_grh,
+			const struct ib_mad_hdr *in, size_t in_mad_size,
+			struct ib_mad_hdr *out, size_t *out_mad_size,
+			u16 *out_mad_pkey_index);
+struct ib_xrcd *mlx5_ib_alloc_xrcd(struct ib_device *ibdev,
+					  struct ib_ucontext *context,
+					  struct ib_udata *udata);
+int mlx5_ib_dealloc_xrcd(struct ib_xrcd *xrcd);
+int mlx5_ib_get_buf_offset(u64 addr, int page_shift, u32 *offset);
+int mlx5_query_ext_port_caps(struct mlx5_ib_dev *dev, u8 port);
+int mlx5_query_mad_ifc_smp_attr_node_info(struct ib_device *ibdev,
+					  struct ib_smp *out_mad);
+int mlx5_query_mad_ifc_system_image_guid(struct ib_device *ibdev,
+					 __be64 *sys_image_guid);
+int mlx5_query_mad_ifc_max_pkeys(struct ib_device *ibdev,
+				 u16 *max_pkeys);
+int mlx5_query_mad_ifc_vendor_id(struct ib_device *ibdev,
+				 u32 *vendor_id);
+int mlx5_query_mad_ifc_node_desc(struct mlx5_ib_dev *dev, char *node_desc);
+int mlx5_query_mad_ifc_node_guid(struct mlx5_ib_dev *dev, __be64 *node_guid);
+int mlx5_query_mad_ifc_pkey(struct ib_device *ibdev, u8 port, u16 index,
+			    u16 *pkey);
+int mlx5_query_mad_ifc_gids(struct ib_device *ibdev, u8 port, int index,
+			    union ib_gid *gid);
+int mlx5_query_mad_ifc_port(struct ib_device *ibdev, u8 port,
+			    struct ib_port_attr *props);
+int mlx5_ib_query_port(struct ib_device *ibdev, u8 port,
+		       struct ib_port_attr *props);
+int mlx5_ib_init_fmr(struct mlx5_ib_dev *dev);
+void mlx5_ib_cleanup_fmr(struct mlx5_ib_dev *dev);
+void mlx5_ib_cont_pages(struct ib_umem *umem, u64 addr,
+			unsigned long max_page_shift,
+			int *count, int *shift,
+			int *ncont, int *order);
+void __mlx5_ib_populate_pas(struct mlx5_ib_dev *dev, struct ib_umem *umem,
+			    int page_shift, size_t offset, size_t num_pages,
+			    __be64 *pas, int access_flags);
+void mlx5_ib_populate_pas(struct mlx5_ib_dev *dev, struct ib_umem *umem,
+			  int page_shift, __be64 *pas, int access_flags);
+void mlx5_ib_copy_pas(u64 *old, u64 *new, int step, int num);
+int mlx5_ib_get_cqe_size(struct mlx5_ib_dev *dev, struct ib_cq *ibcq);
+int mlx5_mr_cache_init(struct mlx5_ib_dev *dev);
+int mlx5_mr_cache_cleanup(struct mlx5_ib_dev *dev);
+
+struct mlx5_ib_mr *mlx5_mr_cache_alloc(struct mlx5_ib_dev *dev, int entry);
+void mlx5_mr_cache_free(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr);
+int mlx5_ib_check_mr_status(struct ib_mr *ibmr, u32 check_mask,
+			    struct ib_mr_status *mr_status);
+struct ib_wq *mlx5_ib_create_wq(struct ib_pd *pd,
+				struct ib_wq_init_attr *init_attr,
+				struct ib_udata *udata);
+int mlx5_ib_destroy_wq(struct ib_wq *wq);
+int mlx5_ib_modify_wq(struct ib_wq *wq, struct ib_wq_attr *wq_attr,
+		      u32 wq_attr_mask, struct ib_udata *udata);
+struct ib_rwq_ind_table *mlx5_ib_create_rwq_ind_table(struct ib_device *device,
+						      struct ib_rwq_ind_table_init_attr *init_attr,
+						      struct ib_udata *udata);
+int mlx5_ib_destroy_rwq_ind_table(struct ib_rwq_ind_table *wq_ind_table);
+bool mlx5_ib_dc_atomic_is_supported(struct mlx5_ib_dev *dev);
+struct ib_dm *mlx5_ib_alloc_dm(struct ib_device *ibdev,
+			       struct ib_ucontext *context,
+			       struct ib_dm_alloc_attr *attr,
+			       struct uverbs_attr_bundle *attrs);
+int mlx5_ib_dealloc_dm(struct ib_dm *ibdm);
+struct ib_mr *mlx5_ib_reg_dm_mr(struct ib_pd *pd, struct ib_dm *dm,
+				struct ib_dm_mr_attr *attr,
+				struct uverbs_attr_bundle *attrs);
+
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
+void mlx5_ib_internal_fill_odp_caps(struct mlx5_ib_dev *dev);
+void mlx5_ib_pfault(struct mlx5_core_dev *mdev, void *context,
+		    struct mlx5_pagefault *pfault);
+int mlx5_ib_odp_init_one(struct mlx5_ib_dev *ibdev);
+int __init mlx5_ib_odp_init(void);
+void mlx5_ib_odp_cleanup(void);
+void mlx5_ib_invalidate_range(struct ib_umem *umem, unsigned long start,
+			      unsigned long end);
+void mlx5_odp_init_mr_cache_entry(struct mlx5_cache_ent *ent);
+void mlx5_odp_populate_klm(struct mlx5_klm *pklm, size_t offset,
+			   size_t nentries, struct mlx5_ib_mr *mr, int flags);
+#else /* CONFIG_INFINIBAND_ON_DEMAND_PAGING */
+static inline void mlx5_ib_internal_fill_odp_caps(struct mlx5_ib_dev *dev)
+{
+	return;
+}
+
+static inline int mlx5_ib_odp_init_one(struct mlx5_ib_dev *ibdev) { return 0; }
+static inline int mlx5_ib_odp_init(void) { return 0; }
+static inline void mlx5_ib_odp_cleanup(void)				    {}
+static inline void mlx5_odp_init_mr_cache_entry(struct mlx5_cache_ent *ent) {}
+static inline void mlx5_odp_populate_klm(struct mlx5_klm *pklm, size_t offset,
+					 size_t nentries, struct mlx5_ib_mr *mr,
+					 int flags) {}
+
+#endif /* CONFIG_INFINIBAND_ON_DEMAND_PAGING */
+
+/* Needed for rep profile */
+int mlx5_ib_stage_init_init(struct mlx5_ib_dev *dev);
+void mlx5_ib_stage_init_cleanup(struct mlx5_ib_dev *dev);
+int mlx5_ib_stage_rep_flow_db_init(struct mlx5_ib_dev *dev);
+int mlx5_ib_stage_caps_init(struct mlx5_ib_dev *dev);
+int mlx5_ib_stage_rep_non_default_cb(struct mlx5_ib_dev *dev);
+int mlx5_ib_stage_rep_roce_init(struct mlx5_ib_dev *dev);
+void mlx5_ib_stage_rep_roce_cleanup(struct mlx5_ib_dev *dev);
+int mlx5_ib_stage_dev_res_init(struct mlx5_ib_dev *dev);
+void mlx5_ib_stage_dev_res_cleanup(struct mlx5_ib_dev *dev);
+int mlx5_ib_stage_counters_init(struct mlx5_ib_dev *dev);
+void mlx5_ib_stage_counters_cleanup(struct mlx5_ib_dev *dev);
+int mlx5_ib_stage_bfrag_init(struct mlx5_ib_dev *dev);
+void mlx5_ib_stage_bfrag_cleanup(struct mlx5_ib_dev *dev);
+void mlx5_ib_stage_pre_ib_reg_umr_cleanup(struct mlx5_ib_dev *dev);
+int mlx5_ib_stage_ib_reg_init(struct mlx5_ib_dev *dev);
+void mlx5_ib_stage_ib_reg_cleanup(struct mlx5_ib_dev *dev);
+int mlx5_ib_stage_post_ib_reg_umr_init(struct mlx5_ib_dev *dev);
+int mlx5_ib_stage_class_attr_init(struct mlx5_ib_dev *dev);
+void __mlx5_ib_remove(struct mlx5_ib_dev *dev,
+		      const struct mlx5_ib_profile *profile,
+		      int stage);
+void *__mlx5_ib_add(struct mlx5_ib_dev *dev,
+		    const struct mlx5_ib_profile *profile);
+
+int mlx5_ib_get_vf_config(struct ib_device *device, int vf,
+			  u8 port, struct ifla_vf_info *info);
+int mlx5_ib_set_vf_link_state(struct ib_device *device, int vf,
+			      u8 port, int state);
+int mlx5_ib_get_vf_stats(struct ib_device *device, int vf,
+			 u8 port, struct ifla_vf_stats *stats);
+int mlx5_ib_set_vf_guid(struct ib_device *device, int vf, u8 port,
+			u64 guid, int type);
+
+__be16 mlx5_get_roce_udp_sport(struct mlx5_ib_dev *dev,
+			       const struct ib_gid_attr *attr);
+
+void mlx5_ib_cleanup_cong_debugfs(struct mlx5_ib_dev *dev, u8 port_num);
+int mlx5_ib_init_cong_debugfs(struct mlx5_ib_dev *dev, u8 port_num);
+
+/* GSI QP helper functions */
+struct ib_qp *mlx5_ib_gsi_create_qp(struct ib_pd *pd,
+				    struct ib_qp_init_attr *init_attr);
+int mlx5_ib_gsi_destroy_qp(struct ib_qp *qp);
+int mlx5_ib_gsi_modify_qp(struct ib_qp *qp, struct ib_qp_attr *attr,
+			  int attr_mask);
+int mlx5_ib_gsi_query_qp(struct ib_qp *qp, struct ib_qp_attr *qp_attr,
+			 int qp_attr_mask,
+			 struct ib_qp_init_attr *qp_init_attr);
+int mlx5_ib_gsi_post_send(struct ib_qp *qp, const struct ib_send_wr *wr,
+			  const struct ib_send_wr **bad_wr);
+int mlx5_ib_gsi_post_recv(struct ib_qp *qp, const struct ib_recv_wr *wr,
+			  const struct ib_recv_wr **bad_wr);
+void mlx5_ib_gsi_pkey_change(struct mlx5_ib_gsi_qp *gsi);
+
+int mlx5_ib_generate_wc(struct ib_cq *ibcq, struct ib_wc *wc);
+
+void mlx5_ib_free_bfreg(struct mlx5_ib_dev *dev, struct mlx5_bfreg_info *bfregi,
+			int bfregn);
+struct mlx5_ib_dev *mlx5_ib_get_ibdev_from_mpi(struct mlx5_ib_multiport_info *mpi);
+struct mlx5_core_dev *mlx5_ib_get_native_port_mdev(struct mlx5_ib_dev *dev,
+						   u8 ib_port_num,
+						   u8 *native_port_num);
+void mlx5_ib_put_native_port_mdev(struct mlx5_ib_dev *dev,
+				  u8 port_num);
+
+#if IS_ENABLED(CONFIG_INFINIBAND_USER_ACCESS)
+int mlx5_ib_devx_create(struct mlx5_ib_dev *dev,
+			struct mlx5_ib_ucontext *context);
+void mlx5_ib_devx_destroy(struct mlx5_ib_dev *dev,
+			  struct mlx5_ib_ucontext *context);
+const struct uverbs_object_tree_def *mlx5_ib_get_devx_tree(void);
+struct mlx5_ib_flow_handler *mlx5_ib_raw_fs_rule_add(
+	struct mlx5_ib_dev *dev, struct mlx5_ib_flow_matcher *fs_matcher,
+	void *cmd_in, int inlen, int dest_id, int dest_type);
+bool mlx5_ib_devx_is_flow_dest(void *obj, int *dest_id, int *dest_type);
+int mlx5_ib_get_flow_trees(const struct uverbs_object_tree_def **root);
+#else
+static inline int
+mlx5_ib_devx_create(struct mlx5_ib_dev *dev,
+		    struct mlx5_ib_ucontext *context) { return -EOPNOTSUPP; };
+static inline void mlx5_ib_devx_destroy(struct mlx5_ib_dev *dev,
+					struct mlx5_ib_ucontext *context) {}
+static inline const struct uverbs_object_tree_def *
+mlx5_ib_get_devx_tree(void) { return NULL; }
+static inline bool mlx5_ib_devx_is_flow_dest(void *obj, int *dest_id,
+					     int *dest_type)
+{
+	return false;
+}
+static inline int
+mlx5_ib_get_flow_trees(const struct uverbs_object_tree_def **root)
+{
+	return 0;
+}
+#endif
+static inline void init_query_mad(struct ib_smp *mad)
+{
+	mad->base_version  = 1;
+	mad->mgmt_class    = IB_MGMT_CLASS_SUBN_LID_ROUTED;
+	mad->class_version = 1;
+	mad->method	   = IB_MGMT_METHOD_GET;
+}
+
+static inline u8 convert_access(int acc)
+{
+	return (acc & IB_ACCESS_REMOTE_ATOMIC ? MLX5_PERM_ATOMIC       : 0) |
+	       (acc & IB_ACCESS_REMOTE_WRITE  ? MLX5_PERM_REMOTE_WRITE : 0) |
+	       (acc & IB_ACCESS_REMOTE_READ   ? MLX5_PERM_REMOTE_READ  : 0) |
+	       (acc & IB_ACCESS_LOCAL_WRITE   ? MLX5_PERM_LOCAL_WRITE  : 0) |
+	       MLX5_PERM_LOCAL_READ;
+}
+
+static inline int is_qp1(enum ib_qp_type qp_type)
+{
+	return qp_type == MLX5_IB_QPT_HW_GSI;
+}
+
+#define MLX5_MAX_UMR_SHIFT 16
+#define MLX5_MAX_UMR_PAGES (1 << MLX5_MAX_UMR_SHIFT)
+
+static inline u32 check_cq_create_flags(u32 flags)
+{
+	/*
+	 * It returns non-zero value for unsupported CQ
+	 * create flags, otherwise it returns zero.
+	 */
+	return (flags & ~(IB_UVERBS_CQ_FLAGS_IGNORE_OVERRUN |
+			  IB_UVERBS_CQ_FLAGS_TIMESTAMP_COMPLETION));
+}
+
+static inline int verify_assign_uidx(u8 cqe_version, u32 cmd_uidx,
+				     u32 *user_index)
+{
+	if (cqe_version) {
+		if ((cmd_uidx == MLX5_IB_DEFAULT_UIDX) ||
+		    (cmd_uidx & ~MLX5_USER_ASSIGNED_UIDX_MASK))
+			return -EINVAL;
+		*user_index = cmd_uidx;
+	} else {
+		*user_index = MLX5_IB_DEFAULT_UIDX;
+	}
+
+	return 0;
+}
+
+static inline int get_qp_user_index(struct mlx5_ib_ucontext *ucontext,
+				    struct mlx5_ib_create_qp *ucmd,
+				    int inlen,
+				    u32 *user_index)
+{
+	u8 cqe_version = ucontext->cqe_version;
+
+	if (field_avail(struct mlx5_ib_create_qp, uidx, inlen) &&
+	    !cqe_version && (ucmd->uidx == MLX5_IB_DEFAULT_UIDX))
+		return 0;
+
+	if (!!(field_avail(struct mlx5_ib_create_qp, uidx, inlen) !=
+	       !!cqe_version))
+		return -EINVAL;
+
+	return verify_assign_uidx(cqe_version, ucmd->uidx, user_index);
+}
+
+static inline int get_srq_user_index(struct mlx5_ib_ucontext *ucontext,
+				     struct mlx5_ib_create_srq *ucmd,
+				     int inlen,
+				     u32 *user_index)
+{
+	u8 cqe_version = ucontext->cqe_version;
+
+	if (field_avail(struct mlx5_ib_create_srq, uidx, inlen) &&
+	    !cqe_version && (ucmd->uidx == MLX5_IB_DEFAULT_UIDX))
+		return 0;
+
+	if (!!(field_avail(struct mlx5_ib_create_srq, uidx, inlen) !=
+	       !!cqe_version))
+		return -EINVAL;
+
+	return verify_assign_uidx(cqe_version, ucmd->uidx, user_index);
+}
+
+static inline int get_uars_per_sys_page(struct mlx5_ib_dev *dev, bool lib_support)
+{
+	return lib_support && MLX5_CAP_GEN(dev->mdev, uar_4k) ?
+				MLX5_UARS_IN_PAGE : 1;
+}
+
+static inline int get_num_static_uars(struct mlx5_ib_dev *dev,
+				      struct mlx5_bfreg_info *bfregi)
+{
+	return get_uars_per_sys_page(dev, bfregi->lib_uar_4k) * bfregi->num_static_sys_pages;
+}
+
+unsigned long mlx5_ib_get_xlt_emergency_page(void);
+void mlx5_ib_put_xlt_emergency_page(void);
+
+int bfregn_to_uar_index(struct mlx5_ib_dev *dev,
+			struct mlx5_bfreg_info *bfregi, u32 bfregn,
+			bool dyn_bfreg);
+#endif /* MLX5_IB_H */
diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c
new file mode 100644
index 000000000..18fd9aa65
--- /dev/null
+++ b/drivers/infiniband/hw/mlx5/mr.c
@@ -0,0 +1,1984 @@
+/*
+ * Copyright (c) 2013-2015, Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+
+#include <linux/kref.h>
+#include <linux/random.h>
+#include <linux/debugfs.h>
+#include <linux/export.h>
+#include <linux/delay.h>
+#include <rdma/ib_umem.h>
+#include <rdma/ib_umem_odp.h>
+#include <rdma/ib_verbs.h>
+#include "mlx5_ib.h"
+
+enum {
+	MAX_PENDING_REG_MR = 8,
+};
+
+#define MLX5_UMR_ALIGN 2048
+
+static void clean_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr);
+static void dereg_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr);
+static int mr_cache_max_order(struct mlx5_ib_dev *dev);
+static int unreg_umr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr);
+
+static bool umr_can_use_indirect_mkey(struct mlx5_ib_dev *dev)
+{
+	return !MLX5_CAP_GEN(dev->mdev, umr_indirect_mkey_disabled);
+}
+
+static int destroy_mkey(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
+{
+	int err = mlx5_core_destroy_mkey(dev->mdev, &mr->mmkey);
+
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
+	/* Wait until all page fault handlers using the mr complete. */
+	synchronize_srcu(&dev->mr_srcu);
+#endif
+
+	return err;
+}
+
+static int order2idx(struct mlx5_ib_dev *dev, int order)
+{
+	struct mlx5_mr_cache *cache = &dev->cache;
+
+	if (order < cache->ent[0].order)
+		return 0;
+	else
+		return order - cache->ent[0].order;
+}
+
+static bool use_umr_mtt_update(struct mlx5_ib_mr *mr, u64 start, u64 length)
+{
+	return ((u64)1 << mr->order) * MLX5_ADAPTER_PAGE_SIZE >=
+		length + (start & (MLX5_ADAPTER_PAGE_SIZE - 1));
+}
+
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
+static void update_odp_mr(struct mlx5_ib_mr *mr)
+{
+	if (mr->umem->odp_data) {
+		/*
+		 * This barrier prevents the compiler from moving the
+		 * setting of umem->odp_data->private to point to our
+		 * MR, before reg_umr finished, to ensure that the MR
+		 * initialization have finished before starting to
+		 * handle invalidations.
+		 */
+		smp_wmb();
+		mr->umem->odp_data->private = mr;
+		/*
+		 * Make sure we will see the new
+		 * umem->odp_data->private value in the invalidation
+		 * routines, before we can get page faults on the
+		 * MR. Page faults can happen once we put the MR in
+		 * the tree, below this line. Without the barrier,
+		 * there can be a fault handling and an invalidation
+		 * before umem->odp_data->private == mr is visible to
+		 * the invalidation handler.
+		 */
+		smp_wmb();
+	}
+}
+#endif
+
+static void reg_mr_callback(int status, void *context)
+{
+	struct mlx5_ib_mr *mr = context;
+	struct mlx5_ib_dev *dev = mr->dev;
+	struct mlx5_mr_cache *cache = &dev->cache;
+	int c = order2idx(dev, mr->order);
+	struct mlx5_cache_ent *ent = &cache->ent[c];
+	u8 key;
+	unsigned long flags;
+	struct mlx5_mkey_table *table = &dev->mdev->priv.mkey_table;
+	int err;
+
+	spin_lock_irqsave(&ent->lock, flags);
+	ent->pending--;
+	spin_unlock_irqrestore(&ent->lock, flags);
+	if (status) {
+		mlx5_ib_warn(dev, "async reg mr failed. status %d\n", status);
+		kfree(mr);
+		dev->fill_delay = 1;
+		mod_timer(&dev->delay_timer, jiffies + HZ);
+		return;
+	}
+
+	mr->mmkey.type = MLX5_MKEY_MR;
+	spin_lock_irqsave(&dev->mdev->priv.mkey_lock, flags);
+	key = dev->mdev->priv.mkey_key++;
+	spin_unlock_irqrestore(&dev->mdev->priv.mkey_lock, flags);
+	mr->mmkey.key = mlx5_idx_to_mkey(MLX5_GET(create_mkey_out, mr->out, mkey_index)) | key;
+
+	cache->last_add = jiffies;
+
+	spin_lock_irqsave(&ent->lock, flags);
+	list_add_tail(&mr->list, &ent->head);
+	ent->cur++;
+	ent->size++;
+	spin_unlock_irqrestore(&ent->lock, flags);
+
+	write_lock_irqsave(&table->lock, flags);
+	err = radix_tree_insert(&table->tree, mlx5_base_mkey(mr->mmkey.key),
+				&mr->mmkey);
+	if (err)
+		pr_err("Error inserting to mkey tree. 0x%x\n", -err);
+	write_unlock_irqrestore(&table->lock, flags);
+
+	if (!completion_done(&ent->compl))
+		complete(&ent->compl);
+}
+
+static int add_keys(struct mlx5_ib_dev *dev, int c, int num)
+{
+	struct mlx5_mr_cache *cache = &dev->cache;
+	struct mlx5_cache_ent *ent = &cache->ent[c];
+	int inlen = MLX5_ST_SZ_BYTES(create_mkey_in);
+	struct mlx5_ib_mr *mr;
+	void *mkc;
+	u32 *in;
+	int err = 0;
+	int i;
+
+	in = kzalloc(inlen, GFP_KERNEL);
+	if (!in)
+		return -ENOMEM;
+
+	mkc = MLX5_ADDR_OF(create_mkey_in, in, memory_key_mkey_entry);
+	for (i = 0; i < num; i++) {
+		if (ent->pending >= MAX_PENDING_REG_MR) {
+			err = -EAGAIN;
+			break;
+		}
+
+		mr = kzalloc(sizeof(*mr), GFP_KERNEL);
+		if (!mr) {
+			err = -ENOMEM;
+			break;
+		}
+		mr->order = ent->order;
+		mr->allocated_from_cache = 1;
+		mr->dev = dev;
+
+		MLX5_SET(mkc, mkc, free, 1);
+		MLX5_SET(mkc, mkc, umr_en, 1);
+		MLX5_SET(mkc, mkc, access_mode_1_0, ent->access_mode & 0x3);
+		MLX5_SET(mkc, mkc, access_mode_4_2,
+			 (ent->access_mode >> 2) & 0x7);
+
+		MLX5_SET(mkc, mkc, qpn, 0xffffff);
+		MLX5_SET(mkc, mkc, translations_octword_size, ent->xlt);
+		MLX5_SET(mkc, mkc, log_page_size, ent->page);
+
+		spin_lock_irq(&ent->lock);
+		ent->pending++;
+		spin_unlock_irq(&ent->lock);
+		err = mlx5_core_create_mkey_cb(dev->mdev, &mr->mmkey,
+					       in, inlen,
+					       mr->out, sizeof(mr->out),
+					       reg_mr_callback, mr);
+		if (err) {
+			spin_lock_irq(&ent->lock);
+			ent->pending--;
+			spin_unlock_irq(&ent->lock);
+			mlx5_ib_warn(dev, "create mkey failed %d\n", err);
+			kfree(mr);
+			break;
+		}
+	}
+
+	kfree(in);
+	return err;
+}
+
+static void remove_keys(struct mlx5_ib_dev *dev, int c, int num)
+{
+	struct mlx5_mr_cache *cache = &dev->cache;
+	struct mlx5_cache_ent *ent = &cache->ent[c];
+	struct mlx5_ib_mr *tmp_mr;
+	struct mlx5_ib_mr *mr;
+	LIST_HEAD(del_list);
+	int i;
+
+	for (i = 0; i < num; i++) {
+		spin_lock_irq(&ent->lock);
+		if (list_empty(&ent->head)) {
+			spin_unlock_irq(&ent->lock);
+			break;
+		}
+		mr = list_first_entry(&ent->head, struct mlx5_ib_mr, list);
+		list_move(&mr->list, &del_list);
+		ent->cur--;
+		ent->size--;
+		spin_unlock_irq(&ent->lock);
+		mlx5_core_destroy_mkey(dev->mdev, &mr->mmkey);
+	}
+
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
+	synchronize_srcu(&dev->mr_srcu);
+#endif
+
+	list_for_each_entry_safe(mr, tmp_mr, &del_list, list) {
+		list_del(&mr->list);
+		kfree(mr);
+	}
+}
+
+static ssize_t size_write(struct file *filp, const char __user *buf,
+			  size_t count, loff_t *pos)
+{
+	struct mlx5_cache_ent *ent = filp->private_data;
+	struct mlx5_ib_dev *dev = ent->dev;
+	char lbuf[20] = {0};
+	u32 var;
+	int err;
+	int c;
+
+	count = min(count, sizeof(lbuf) - 1);
+	if (copy_from_user(lbuf, buf, count))
+		return -EFAULT;
+
+	c = order2idx(dev, ent->order);
+
+	if (sscanf(lbuf, "%u", &var) != 1)
+		return -EINVAL;
+
+	if (var < ent->limit)
+		return -EINVAL;
+
+	if (var > ent->size) {
+		do {
+			err = add_keys(dev, c, var - ent->size);
+			if (err && err != -EAGAIN)
+				return err;
+
+			usleep_range(3000, 5000);
+		} while (err);
+	} else if (var < ent->size) {
+		remove_keys(dev, c, ent->size - var);
+	}
+
+	return count;
+}
+
+static ssize_t size_read(struct file *filp, char __user *buf, size_t count,
+			 loff_t *pos)
+{
+	struct mlx5_cache_ent *ent = filp->private_data;
+	char lbuf[20];
+	int err;
+
+	err = snprintf(lbuf, sizeof(lbuf), "%d\n", ent->size);
+	if (err < 0)
+		return err;
+
+	return simple_read_from_buffer(buf, count, pos, lbuf, err);
+}
+
+static const struct file_operations size_fops = {
+	.owner	= THIS_MODULE,
+	.open	= simple_open,
+	.write	= size_write,
+	.read	= size_read,
+};
+
+static ssize_t limit_write(struct file *filp, const char __user *buf,
+			   size_t count, loff_t *pos)
+{
+	struct mlx5_cache_ent *ent = filp->private_data;
+	struct mlx5_ib_dev *dev = ent->dev;
+	char lbuf[20] = {0};
+	u32 var;
+	int err;
+	int c;
+
+	count = min(count, sizeof(lbuf) - 1);
+	if (copy_from_user(lbuf, buf, count))
+		return -EFAULT;
+
+	c = order2idx(dev, ent->order);
+
+	if (sscanf(lbuf, "%u", &var) != 1)
+		return -EINVAL;
+
+	if (var > ent->size)
+		return -EINVAL;
+
+	ent->limit = var;
+
+	if (ent->cur < ent->limit) {
+		err = add_keys(dev, c, 2 * ent->limit - ent->cur);
+		if (err)
+			return err;
+	}
+
+	return count;
+}
+
+static ssize_t limit_read(struct file *filp, char __user *buf, size_t count,
+			  loff_t *pos)
+{
+	struct mlx5_cache_ent *ent = filp->private_data;
+	char lbuf[20];
+	int err;
+
+	err = snprintf(lbuf, sizeof(lbuf), "%d\n", ent->limit);
+	if (err < 0)
+		return err;
+
+	return simple_read_from_buffer(buf, count, pos, lbuf, err);
+}
+
+static const struct file_operations limit_fops = {
+	.owner	= THIS_MODULE,
+	.open	= simple_open,
+	.write	= limit_write,
+	.read	= limit_read,
+};
+
+static int someone_adding(struct mlx5_mr_cache *cache)
+{
+	int i;
+
+	for (i = 0; i < MAX_MR_CACHE_ENTRIES; i++) {
+		if (cache->ent[i].cur < cache->ent[i].limit)
+			return 1;
+	}
+
+	return 0;
+}
+
+static void __cache_work_func(struct mlx5_cache_ent *ent)
+{
+	struct mlx5_ib_dev *dev = ent->dev;
+	struct mlx5_mr_cache *cache = &dev->cache;
+	int i = order2idx(dev, ent->order);
+	int err;
+
+	if (cache->stopped)
+		return;
+
+	ent = &dev->cache.ent[i];
+	if (ent->cur < 2 * ent->limit && !dev->fill_delay) {
+		err = add_keys(dev, i, 1);
+		if (ent->cur < 2 * ent->limit) {
+			if (err == -EAGAIN) {
+				mlx5_ib_dbg(dev, "returned eagain, order %d\n",
+					    i + 2);
+				queue_delayed_work(cache->wq, &ent->dwork,
+						   msecs_to_jiffies(3));
+			} else if (err) {
+				mlx5_ib_warn(dev, "command failed order %d, err %d\n",
+					     i + 2, err);
+				queue_delayed_work(cache->wq, &ent->dwork,
+						   msecs_to_jiffies(1000));
+			} else {
+				queue_work(cache->wq, &ent->work);
+			}
+		}
+	} else if (ent->cur > 2 * ent->limit) {
+		/*
+		 * The remove_keys() logic is performed as garbage collection
+		 * task. Such task is intended to be run when no other active
+		 * processes are running.
+		 *
+		 * The need_resched() will return TRUE if there are user tasks
+		 * to be activated in near future.
+		 *
+		 * In such case, we don't execute remove_keys() and postpone
+		 * the garbage collection work to try to run in next cycle,
+		 * in order to free CPU resources to other tasks.
+		 */
+		if (!need_resched() && !someone_adding(cache) &&
+		    time_after(jiffies, cache->last_add + 300 * HZ)) {
+			remove_keys(dev, i, 1);
+			if (ent->cur > ent->limit)
+				queue_work(cache->wq, &ent->work);
+		} else {
+			queue_delayed_work(cache->wq, &ent->dwork, 300 * HZ);
+		}
+	}
+}
+
+static void delayed_cache_work_func(struct work_struct *work)
+{
+	struct mlx5_cache_ent *ent;
+
+	ent = container_of(work, struct mlx5_cache_ent, dwork.work);
+	__cache_work_func(ent);
+}
+
+static void cache_work_func(struct work_struct *work)
+{
+	struct mlx5_cache_ent *ent;
+
+	ent = container_of(work, struct mlx5_cache_ent, work);
+	__cache_work_func(ent);
+}
+
+struct mlx5_ib_mr *mlx5_mr_cache_alloc(struct mlx5_ib_dev *dev, int entry)
+{
+	struct mlx5_mr_cache *cache = &dev->cache;
+	struct mlx5_cache_ent *ent;
+	struct mlx5_ib_mr *mr;
+	int err;
+
+	if (entry < 0 || entry >= MAX_MR_CACHE_ENTRIES) {
+		mlx5_ib_err(dev, "cache entry %d is out of range\n", entry);
+		return ERR_PTR(-EINVAL);
+	}
+
+	ent = &cache->ent[entry];
+	while (1) {
+		spin_lock_irq(&ent->lock);
+		if (list_empty(&ent->head)) {
+			spin_unlock_irq(&ent->lock);
+
+			err = add_keys(dev, entry, 1);
+			if (err && err != -EAGAIN)
+				return ERR_PTR(err);
+
+			wait_for_completion(&ent->compl);
+		} else {
+			mr = list_first_entry(&ent->head, struct mlx5_ib_mr,
+					      list);
+			list_del(&mr->list);
+			ent->cur--;
+			spin_unlock_irq(&ent->lock);
+			if (ent->cur < ent->limit)
+				queue_work(cache->wq, &ent->work);
+			return mr;
+		}
+	}
+}
+
+static struct mlx5_ib_mr *alloc_cached_mr(struct mlx5_ib_dev *dev, int order)
+{
+	struct mlx5_mr_cache *cache = &dev->cache;
+	struct mlx5_ib_mr *mr = NULL;
+	struct mlx5_cache_ent *ent;
+	int last_umr_cache_entry;
+	int c;
+	int i;
+
+	c = order2idx(dev, order);
+	last_umr_cache_entry = order2idx(dev, mr_cache_max_order(dev));
+	if (c < 0 || c > last_umr_cache_entry) {
+		mlx5_ib_warn(dev, "order %d, cache index %d\n", order, c);
+		return NULL;
+	}
+
+	for (i = c; i <= last_umr_cache_entry; i++) {
+		ent = &cache->ent[i];
+
+		mlx5_ib_dbg(dev, "order %d, cache index %d\n", ent->order, i);
+
+		spin_lock_irq(&ent->lock);
+		if (!list_empty(&ent->head)) {
+			mr = list_first_entry(&ent->head, struct mlx5_ib_mr,
+					      list);
+			list_del(&mr->list);
+			ent->cur--;
+			spin_unlock_irq(&ent->lock);
+			if (ent->cur < ent->limit)
+				queue_work(cache->wq, &ent->work);
+			break;
+		}
+		spin_unlock_irq(&ent->lock);
+
+		queue_work(cache->wq, &ent->work);
+	}
+
+	if (!mr)
+		cache->ent[c].miss++;
+
+	return mr;
+}
+
+void mlx5_mr_cache_free(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
+{
+	struct mlx5_mr_cache *cache = &dev->cache;
+	struct mlx5_cache_ent *ent;
+	int shrink = 0;
+	int c;
+
+	if (!mr->allocated_from_cache)
+		return;
+
+	c = order2idx(dev, mr->order);
+	WARN_ON(c < 0 || c >= MAX_MR_CACHE_ENTRIES);
+
+	if (unreg_umr(dev, mr)) {
+		mr->allocated_from_cache = false;
+		destroy_mkey(dev, mr);
+		ent = &cache->ent[c];
+		if (ent->cur < ent->limit)
+			queue_work(cache->wq, &ent->work);
+		return;
+	}
+
+	ent = &cache->ent[c];
+	spin_lock_irq(&ent->lock);
+	list_add_tail(&mr->list, &ent->head);
+	ent->cur++;
+	if (ent->cur > 2 * ent->limit)
+		shrink = 1;
+	spin_unlock_irq(&ent->lock);
+
+	if (shrink)
+		queue_work(cache->wq, &ent->work);
+}
+
+static void clean_keys(struct mlx5_ib_dev *dev, int c)
+{
+	struct mlx5_mr_cache *cache = &dev->cache;
+	struct mlx5_cache_ent *ent = &cache->ent[c];
+	struct mlx5_ib_mr *tmp_mr;
+	struct mlx5_ib_mr *mr;
+	LIST_HEAD(del_list);
+
+	cancel_delayed_work(&ent->dwork);
+	while (1) {
+		spin_lock_irq(&ent->lock);
+		if (list_empty(&ent->head)) {
+			spin_unlock_irq(&ent->lock);
+			break;
+		}
+		mr = list_first_entry(&ent->head, struct mlx5_ib_mr, list);
+		list_move(&mr->list, &del_list);
+		ent->cur--;
+		ent->size--;
+		spin_unlock_irq(&ent->lock);
+		mlx5_core_destroy_mkey(dev->mdev, &mr->mmkey);
+	}
+
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
+	synchronize_srcu(&dev->mr_srcu);
+#endif
+
+	list_for_each_entry_safe(mr, tmp_mr, &del_list, list) {
+		list_del(&mr->list);
+		kfree(mr);
+	}
+}
+
+static void mlx5_mr_cache_debugfs_cleanup(struct mlx5_ib_dev *dev)
+{
+	if (!mlx5_debugfs_root || dev->rep)
+		return;
+
+	debugfs_remove_recursive(dev->cache.root);
+	dev->cache.root = NULL;
+}
+
+static int mlx5_mr_cache_debugfs_init(struct mlx5_ib_dev *dev)
+{
+	struct mlx5_mr_cache *cache = &dev->cache;
+	struct mlx5_cache_ent *ent;
+	int i;
+
+	if (!mlx5_debugfs_root || dev->rep)
+		return 0;
+
+	cache->root = debugfs_create_dir("mr_cache", dev->mdev->priv.dbg_root);
+	if (!cache->root)
+		return -ENOMEM;
+
+	for (i = 0; i < MAX_MR_CACHE_ENTRIES; i++) {
+		ent = &cache->ent[i];
+		sprintf(ent->name, "%d", ent->order);
+		ent->dir = debugfs_create_dir(ent->name,  cache->root);
+		if (!ent->dir)
+			goto err;
+
+		ent->fsize = debugfs_create_file("size", 0600, ent->dir, ent,
+						 &size_fops);
+		if (!ent->fsize)
+			goto err;
+
+		ent->flimit = debugfs_create_file("limit", 0600, ent->dir, ent,
+						  &limit_fops);
+		if (!ent->flimit)
+			goto err;
+
+		ent->fcur = debugfs_create_u32("cur", 0400, ent->dir,
+					       &ent->cur);
+		if (!ent->fcur)
+			goto err;
+
+		ent->fmiss = debugfs_create_u32("miss", 0600, ent->dir,
+						&ent->miss);
+		if (!ent->fmiss)
+			goto err;
+	}
+
+	return 0;
+err:
+	mlx5_mr_cache_debugfs_cleanup(dev);
+
+	return -ENOMEM;
+}
+
+static void delay_time_func(struct timer_list *t)
+{
+	struct mlx5_ib_dev *dev = from_timer(dev, t, delay_timer);
+
+	dev->fill_delay = 0;
+}
+
+int mlx5_mr_cache_init(struct mlx5_ib_dev *dev)
+{
+	struct mlx5_mr_cache *cache = &dev->cache;
+	struct mlx5_cache_ent *ent;
+	int err;
+	int i;
+
+	mutex_init(&dev->slow_path_mutex);
+	cache->wq = alloc_ordered_workqueue("mkey_cache", WQ_MEM_RECLAIM);
+	if (!cache->wq) {
+		mlx5_ib_warn(dev, "failed to create work queue\n");
+		return -ENOMEM;
+	}
+
+	timer_setup(&dev->delay_timer, delay_time_func, 0);
+	for (i = 0; i < MAX_MR_CACHE_ENTRIES; i++) {
+		ent = &cache->ent[i];
+		INIT_LIST_HEAD(&ent->head);
+		spin_lock_init(&ent->lock);
+		ent->order = i + 2;
+		ent->dev = dev;
+		ent->limit = 0;
+
+		init_completion(&ent->compl);
+		INIT_WORK(&ent->work, cache_work_func);
+		INIT_DELAYED_WORK(&ent->dwork, delayed_cache_work_func);
+
+		if (i > MR_CACHE_LAST_STD_ENTRY) {
+			mlx5_odp_init_mr_cache_entry(ent);
+			continue;
+		}
+
+		if (ent->order > mr_cache_max_order(dev))
+			continue;
+
+		ent->page = PAGE_SHIFT;
+		ent->xlt = (1 << ent->order) * sizeof(struct mlx5_mtt) /
+			   MLX5_IB_UMR_OCTOWORD;
+		ent->access_mode = MLX5_MKC_ACCESS_MODE_MTT;
+		if ((dev->mdev->profile->mask & MLX5_PROF_MASK_MR_CACHE) &&
+		    !dev->rep &&
+		    mlx5_core_is_pf(dev->mdev))
+			ent->limit = dev->mdev->profile->mr_cache[i].limit;
+		else
+			ent->limit = 0;
+		queue_work(cache->wq, &ent->work);
+	}
+
+	err = mlx5_mr_cache_debugfs_init(dev);
+	if (err)
+		mlx5_ib_warn(dev, "cache debugfs failure\n");
+
+	/*
+	 * We don't want to fail driver if debugfs failed to initialize,
+	 * so we are not forwarding error to the user.
+	 */
+
+	return 0;
+}
+
+static void wait_for_async_commands(struct mlx5_ib_dev *dev)
+{
+	struct mlx5_mr_cache *cache = &dev->cache;
+	struct mlx5_cache_ent *ent;
+	int total = 0;
+	int i;
+	int j;
+
+	for (i = 0; i < MAX_MR_CACHE_ENTRIES; i++) {
+		ent = &cache->ent[i];
+		for (j = 0 ; j < 1000; j++) {
+			if (!ent->pending)
+				break;
+			msleep(50);
+		}
+	}
+	for (i = 0; i < MAX_MR_CACHE_ENTRIES; i++) {
+		ent = &cache->ent[i];
+		total += ent->pending;
+	}
+
+	if (total)
+		mlx5_ib_warn(dev, "aborted while there are %d pending mr requests\n", total);
+	else
+		mlx5_ib_warn(dev, "done with all pending requests\n");
+}
+
+int mlx5_mr_cache_cleanup(struct mlx5_ib_dev *dev)
+{
+	int i;
+
+	if (!dev->cache.wq)
+		return 0;
+
+	dev->cache.stopped = 1;
+	flush_workqueue(dev->cache.wq);
+
+	mlx5_mr_cache_debugfs_cleanup(dev);
+
+	for (i = 0; i < MAX_MR_CACHE_ENTRIES; i++)
+		clean_keys(dev, i);
+
+	destroy_workqueue(dev->cache.wq);
+	wait_for_async_commands(dev);
+	del_timer_sync(&dev->delay_timer);
+
+	return 0;
+}
+
+struct ib_mr *mlx5_ib_get_dma_mr(struct ib_pd *pd, int acc)
+{
+	struct mlx5_ib_dev *dev = to_mdev(pd->device);
+	int inlen = MLX5_ST_SZ_BYTES(create_mkey_in);
+	struct mlx5_core_dev *mdev = dev->mdev;
+	struct mlx5_ib_mr *mr;
+	void *mkc;
+	u32 *in;
+	int err;
+
+	mr = kzalloc(sizeof(*mr), GFP_KERNEL);
+	if (!mr)
+		return ERR_PTR(-ENOMEM);
+
+	in = kzalloc(inlen, GFP_KERNEL);
+	if (!in) {
+		err = -ENOMEM;
+		goto err_free;
+	}
+
+	mkc = MLX5_ADDR_OF(create_mkey_in, in, memory_key_mkey_entry);
+
+	MLX5_SET(mkc, mkc, access_mode_1_0, MLX5_MKC_ACCESS_MODE_PA);
+	MLX5_SET(mkc, mkc, a, !!(acc & IB_ACCESS_REMOTE_ATOMIC));
+	MLX5_SET(mkc, mkc, rw, !!(acc & IB_ACCESS_REMOTE_WRITE));
+	MLX5_SET(mkc, mkc, rr, !!(acc & IB_ACCESS_REMOTE_READ));
+	MLX5_SET(mkc, mkc, lw, !!(acc & IB_ACCESS_LOCAL_WRITE));
+	MLX5_SET(mkc, mkc, lr, 1);
+
+	MLX5_SET(mkc, mkc, length64, 1);
+	MLX5_SET(mkc, mkc, pd, to_mpd(pd)->pdn);
+	MLX5_SET(mkc, mkc, qpn, 0xffffff);
+	MLX5_SET64(mkc, mkc, start_addr, 0);
+
+	err = mlx5_core_create_mkey(mdev, &mr->mmkey, in, inlen);
+	if (err)
+		goto err_in;
+
+	kfree(in);
+	mr->mmkey.type = MLX5_MKEY_MR;
+	mr->ibmr.lkey = mr->mmkey.key;
+	mr->ibmr.rkey = mr->mmkey.key;
+	mr->umem = NULL;
+
+	return &mr->ibmr;
+
+err_in:
+	kfree(in);
+
+err_free:
+	kfree(mr);
+
+	return ERR_PTR(err);
+}
+
+static int get_octo_len(u64 addr, u64 len, int page_shift)
+{
+	u64 page_size = 1ULL << page_shift;
+	u64 offset;
+	int npages;
+
+	offset = addr & (page_size - 1);
+	npages = ALIGN(len + offset, page_size) >> page_shift;
+	return (npages + 1) / 2;
+}
+
+static int mr_cache_max_order(struct mlx5_ib_dev *dev)
+{
+	if (MLX5_CAP_GEN(dev->mdev, umr_extended_translation_offset))
+		return MR_CACHE_LAST_STD_ENTRY + 2;
+	return MLX5_MAX_UMR_SHIFT;
+}
+
+static int mr_umem_get(struct ib_pd *pd, u64 start, u64 length,
+		       int access_flags, struct ib_umem **umem,
+		       int *npages, int *page_shift, int *ncont,
+		       int *order)
+{
+	struct mlx5_ib_dev *dev = to_mdev(pd->device);
+	struct ib_umem *u;
+	int err;
+
+	*umem = NULL;
+
+	u = ib_umem_get(pd->uobject->context, start, length, access_flags, 0);
+	err = PTR_ERR_OR_ZERO(u);
+	if (err) {
+		mlx5_ib_dbg(dev, "umem get failed (%d)\n", err);
+		return err;
+	}
+
+	mlx5_ib_cont_pages(u, start, MLX5_MKEY_PAGE_SHIFT_MASK, npages,
+			   page_shift, ncont, order);
+	if (!*npages) {
+		mlx5_ib_warn(dev, "avoid zero region\n");
+		ib_umem_release(u);
+		return -EINVAL;
+	}
+
+	*umem = u;
+
+	mlx5_ib_dbg(dev, "npages %d, ncont %d, order %d, page_shift %d\n",
+		    *npages, *ncont, *order, *page_shift);
+
+	return 0;
+}
+
+static void mlx5_ib_umr_done(struct ib_cq *cq, struct ib_wc *wc)
+{
+	struct mlx5_ib_umr_context *context =
+		container_of(wc->wr_cqe, struct mlx5_ib_umr_context, cqe);
+
+	context->status = wc->status;
+	complete(&context->done);
+}
+
+static inline void mlx5_ib_init_umr_context(struct mlx5_ib_umr_context *context)
+{
+	context->cqe.done = mlx5_ib_umr_done;
+	context->status = -1;
+	init_completion(&context->done);
+}
+
+static int mlx5_ib_post_send_wait(struct mlx5_ib_dev *dev,
+				  struct mlx5_umr_wr *umrwr)
+{
+	struct umr_common *umrc = &dev->umrc;
+	const struct ib_send_wr *bad;
+	int err;
+	struct mlx5_ib_umr_context umr_context;
+
+	mlx5_ib_init_umr_context(&umr_context);
+	umrwr->wr.wr_cqe = &umr_context.cqe;
+
+	down(&umrc->sem);
+	err = ib_post_send(umrc->qp, &umrwr->wr, &bad);
+	if (err) {
+		mlx5_ib_warn(dev, "UMR post send failed, err %d\n", err);
+	} else {
+		wait_for_completion(&umr_context.done);
+		if (umr_context.status != IB_WC_SUCCESS) {
+			mlx5_ib_warn(dev, "reg umr failed (%u)\n",
+				     umr_context.status);
+			err = -EFAULT;
+		}
+	}
+	up(&umrc->sem);
+	return err;
+}
+
+static struct mlx5_ib_mr *alloc_mr_from_cache(
+				  struct ib_pd *pd, struct ib_umem *umem,
+				  u64 virt_addr, u64 len, int npages,
+				  int page_shift, int order, int access_flags)
+{
+	struct mlx5_ib_dev *dev = to_mdev(pd->device);
+	struct mlx5_ib_mr *mr;
+	int err = 0;
+	int i;
+
+	for (i = 0; i < 1; i++) {
+		mr = alloc_cached_mr(dev, order);
+		if (mr)
+			break;
+
+		err = add_keys(dev, order2idx(dev, order), 1);
+		if (err && err != -EAGAIN) {
+			mlx5_ib_warn(dev, "add_keys failed, err %d\n", err);
+			break;
+		}
+	}
+
+	if (!mr)
+		return ERR_PTR(-EAGAIN);
+
+	mr->ibmr.pd = pd;
+	mr->umem = umem;
+	mr->access_flags = access_flags;
+	mr->desc_size = sizeof(struct mlx5_mtt);
+	mr->mmkey.iova = virt_addr;
+	mr->mmkey.size = len;
+	mr->mmkey.pd = to_mpd(pd)->pdn;
+
+	return mr;
+}
+
+static inline int populate_xlt(struct mlx5_ib_mr *mr, int idx, int npages,
+			       void *xlt, int page_shift, size_t size,
+			       int flags)
+{
+	struct mlx5_ib_dev *dev = mr->dev;
+	struct ib_umem *umem = mr->umem;
+
+	if (flags & MLX5_IB_UPD_XLT_INDIRECT) {
+		if (!umr_can_use_indirect_mkey(dev))
+			return -EPERM;
+		mlx5_odp_populate_klm(xlt, idx, npages, mr, flags);
+		return npages;
+	}
+
+	npages = min_t(size_t, npages, ib_umem_num_pages(umem) - idx);
+
+	if (!(flags & MLX5_IB_UPD_XLT_ZAP)) {
+		__mlx5_ib_populate_pas(dev, umem, page_shift,
+				       idx, npages, xlt,
+				       MLX5_IB_MTT_PRESENT);
+		/* Clear padding after the pages
+		 * brought from the umem.
+		 */
+		memset(xlt + (npages * sizeof(struct mlx5_mtt)), 0,
+		       size - npages * sizeof(struct mlx5_mtt));
+	}
+
+	return npages;
+}
+
+#define MLX5_MAX_UMR_CHUNK ((1 << (MLX5_MAX_UMR_SHIFT + 4)) - \
+			    MLX5_UMR_MTT_ALIGNMENT)
+#define MLX5_SPARE_UMR_CHUNK 0x10000
+
+int mlx5_ib_update_xlt(struct mlx5_ib_mr *mr, u64 idx, int npages,
+		       int page_shift, int flags)
+{
+	struct mlx5_ib_dev *dev = mr->dev;
+	struct device *ddev = dev->ib_dev.dev.parent;
+	int size;
+	void *xlt;
+	dma_addr_t dma;
+	struct mlx5_umr_wr wr;
+	struct ib_sge sg;
+	int err = 0;
+	int desc_size = (flags & MLX5_IB_UPD_XLT_INDIRECT)
+			       ? sizeof(struct mlx5_klm)
+			       : sizeof(struct mlx5_mtt);
+	const int page_align = MLX5_UMR_MTT_ALIGNMENT / desc_size;
+	const int page_mask = page_align - 1;
+	size_t pages_mapped = 0;
+	size_t pages_to_map = 0;
+	size_t pages_iter = 0;
+	gfp_t gfp;
+	bool use_emergency_page = false;
+
+	if ((flags & MLX5_IB_UPD_XLT_INDIRECT) &&
+	    !umr_can_use_indirect_mkey(dev))
+		return -EPERM;
+
+	/* UMR copies MTTs in units of MLX5_UMR_MTT_ALIGNMENT bytes,
+	 * so we need to align the offset and length accordingly
+	 */
+	if (idx & page_mask) {
+		npages += idx & page_mask;
+		idx &= ~page_mask;
+	}
+
+	gfp = flags & MLX5_IB_UPD_XLT_ATOMIC ? GFP_ATOMIC : GFP_KERNEL;
+	gfp |= __GFP_ZERO | __GFP_NOWARN;
+
+	pages_to_map = ALIGN(npages, page_align);
+	size = desc_size * pages_to_map;
+	size = min_t(int, size, MLX5_MAX_UMR_CHUNK);
+
+	xlt = (void *)__get_free_pages(gfp, get_order(size));
+	if (!xlt && size > MLX5_SPARE_UMR_CHUNK) {
+		mlx5_ib_dbg(dev, "Failed to allocate %d bytes of order %d. fallback to spare UMR allocation od %d bytes\n",
+			    size, get_order(size), MLX5_SPARE_UMR_CHUNK);
+
+		size = MLX5_SPARE_UMR_CHUNK;
+		xlt = (void *)__get_free_pages(gfp, get_order(size));
+	}
+
+	if (!xlt) {
+		mlx5_ib_warn(dev, "Using XLT emergency buffer\n");
+		xlt = (void *)mlx5_ib_get_xlt_emergency_page();
+		size = PAGE_SIZE;
+		memset(xlt, 0, size);
+		use_emergency_page = true;
+	}
+	pages_iter = size / desc_size;
+	dma = dma_map_single(ddev, xlt, size, DMA_TO_DEVICE);
+	if (dma_mapping_error(ddev, dma)) {
+		mlx5_ib_err(dev, "unable to map DMA during XLT update.\n");
+		err = -ENOMEM;
+		goto free_xlt;
+	}
+
+	sg.addr = dma;
+	sg.lkey = dev->umrc.pd->local_dma_lkey;
+
+	memset(&wr, 0, sizeof(wr));
+	wr.wr.send_flags = MLX5_IB_SEND_UMR_UPDATE_XLT;
+	if (!(flags & MLX5_IB_UPD_XLT_ENABLE))
+		wr.wr.send_flags |= MLX5_IB_SEND_UMR_FAIL_IF_FREE;
+	wr.wr.sg_list = &sg;
+	wr.wr.num_sge = 1;
+	wr.wr.opcode = MLX5_IB_WR_UMR;
+
+	wr.pd = mr->ibmr.pd;
+	wr.mkey = mr->mmkey.key;
+	wr.length = mr->mmkey.size;
+	wr.virt_addr = mr->mmkey.iova;
+	wr.access_flags = mr->access_flags;
+	wr.page_shift = page_shift;
+
+	for (pages_mapped = 0;
+	     pages_mapped < pages_to_map && !err;
+	     pages_mapped += pages_iter, idx += pages_iter) {
+		npages = min_t(int, pages_iter, pages_to_map - pages_mapped);
+		dma_sync_single_for_cpu(ddev, dma, size, DMA_TO_DEVICE);
+		npages = populate_xlt(mr, idx, npages, xlt,
+				      page_shift, size, flags);
+
+		dma_sync_single_for_device(ddev, dma, size, DMA_TO_DEVICE);
+
+		sg.length = ALIGN(npages * desc_size,
+				  MLX5_UMR_MTT_ALIGNMENT);
+
+		if (pages_mapped + pages_iter >= pages_to_map) {
+			if (flags & MLX5_IB_UPD_XLT_ENABLE)
+				wr.wr.send_flags |=
+					MLX5_IB_SEND_UMR_ENABLE_MR |
+					MLX5_IB_SEND_UMR_UPDATE_PD_ACCESS |
+					MLX5_IB_SEND_UMR_UPDATE_TRANSLATION;
+			if (flags & MLX5_IB_UPD_XLT_PD ||
+			    flags & MLX5_IB_UPD_XLT_ACCESS)
+				wr.wr.send_flags |=
+					MLX5_IB_SEND_UMR_UPDATE_PD_ACCESS;
+			if (flags & MLX5_IB_UPD_XLT_ADDR)
+				wr.wr.send_flags |=
+					MLX5_IB_SEND_UMR_UPDATE_TRANSLATION;
+		}
+
+		wr.offset = idx * desc_size;
+		wr.xlt_size = sg.length;
+
+		err = mlx5_ib_post_send_wait(dev, &wr);
+	}
+	dma_unmap_single(ddev, dma, size, DMA_TO_DEVICE);
+
+free_xlt:
+	if (use_emergency_page)
+		mlx5_ib_put_xlt_emergency_page();
+	else
+		free_pages((unsigned long)xlt, get_order(size));
+
+	return err;
+}
+
+/*
+ * If ibmr is NULL it will be allocated by reg_create.
+ * Else, the given ibmr will be used.
+ */
+static struct mlx5_ib_mr *reg_create(struct ib_mr *ibmr, struct ib_pd *pd,
+				     u64 virt_addr, u64 length,
+				     struct ib_umem *umem, int npages,
+				     int page_shift, int access_flags,
+				     bool populate)
+{
+	struct mlx5_ib_dev *dev = to_mdev(pd->device);
+	struct mlx5_ib_mr *mr;
+	__be64 *pas;
+	void *mkc;
+	int inlen;
+	u32 *in;
+	int err;
+	bool pg_cap = !!(MLX5_CAP_GEN(dev->mdev, pg));
+
+	mr = ibmr ? to_mmr(ibmr) : kzalloc(sizeof(*mr), GFP_KERNEL);
+	if (!mr)
+		return ERR_PTR(-ENOMEM);
+
+	mr->ibmr.pd = pd;
+	mr->access_flags = access_flags;
+
+	inlen = MLX5_ST_SZ_BYTES(create_mkey_in);
+	if (populate)
+		inlen += sizeof(*pas) * roundup(npages, 2);
+	in = kvzalloc(inlen, GFP_KERNEL);
+	if (!in) {
+		err = -ENOMEM;
+		goto err_1;
+	}
+	pas = (__be64 *)MLX5_ADDR_OF(create_mkey_in, in, klm_pas_mtt);
+	if (populate && !(access_flags & IB_ACCESS_ON_DEMAND))
+		mlx5_ib_populate_pas(dev, umem, page_shift, pas,
+				     pg_cap ? MLX5_IB_MTT_PRESENT : 0);
+
+	/* The pg_access bit allows setting the access flags
+	 * in the page list submitted with the command. */
+	MLX5_SET(create_mkey_in, in, pg_access, !!(pg_cap));
+
+	mkc = MLX5_ADDR_OF(create_mkey_in, in, memory_key_mkey_entry);
+	MLX5_SET(mkc, mkc, free, !populate);
+	MLX5_SET(mkc, mkc, access_mode_1_0, MLX5_MKC_ACCESS_MODE_MTT);
+	MLX5_SET(mkc, mkc, a, !!(access_flags & IB_ACCESS_REMOTE_ATOMIC));
+	MLX5_SET(mkc, mkc, rw, !!(access_flags & IB_ACCESS_REMOTE_WRITE));
+	MLX5_SET(mkc, mkc, rr, !!(access_flags & IB_ACCESS_REMOTE_READ));
+	MLX5_SET(mkc, mkc, lw, !!(access_flags & IB_ACCESS_LOCAL_WRITE));
+	MLX5_SET(mkc, mkc, lr, 1);
+	MLX5_SET(mkc, mkc, umr_en, 1);
+
+	MLX5_SET64(mkc, mkc, start_addr, virt_addr);
+	MLX5_SET64(mkc, mkc, len, length);
+	MLX5_SET(mkc, mkc, pd, to_mpd(pd)->pdn);
+	MLX5_SET(mkc, mkc, bsf_octword_size, 0);
+	MLX5_SET(mkc, mkc, translations_octword_size,
+		 get_octo_len(virt_addr, length, page_shift));
+	MLX5_SET(mkc, mkc, log_page_size, page_shift);
+	MLX5_SET(mkc, mkc, qpn, 0xffffff);
+	if (populate) {
+		MLX5_SET(create_mkey_in, in, translations_octword_actual_size,
+			 get_octo_len(virt_addr, length, page_shift));
+	}
+
+	err = mlx5_core_create_mkey(dev->mdev, &mr->mmkey, in, inlen);
+	if (err) {
+		mlx5_ib_warn(dev, "create mkey failed\n");
+		goto err_2;
+	}
+	mr->mmkey.type = MLX5_MKEY_MR;
+	mr->desc_size = sizeof(struct mlx5_mtt);
+	mr->dev = dev;
+	kvfree(in);
+
+	mlx5_ib_dbg(dev, "mkey = 0x%x\n", mr->mmkey.key);
+
+	return mr;
+
+err_2:
+	kvfree(in);
+
+err_1:
+	if (!ibmr)
+		kfree(mr);
+
+	return ERR_PTR(err);
+}
+
+static void set_mr_fileds(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr,
+			  int npages, u64 length, int access_flags)
+{
+	mr->npages = npages;
+	atomic_add(npages, &dev->mdev->priv.reg_pages);
+	mr->ibmr.lkey = mr->mmkey.key;
+	mr->ibmr.rkey = mr->mmkey.key;
+	mr->ibmr.length = length;
+	mr->access_flags = access_flags;
+}
+
+static struct ib_mr *mlx5_ib_get_memic_mr(struct ib_pd *pd, u64 memic_addr,
+					  u64 length, int acc)
+{
+	struct mlx5_ib_dev *dev = to_mdev(pd->device);
+	int inlen = MLX5_ST_SZ_BYTES(create_mkey_in);
+	struct mlx5_core_dev *mdev = dev->mdev;
+	struct mlx5_ib_mr *mr;
+	void *mkc;
+	u32 *in;
+	int err;
+
+	mr = kzalloc(sizeof(*mr), GFP_KERNEL);
+	if (!mr)
+		return ERR_PTR(-ENOMEM);
+
+	in = kzalloc(inlen, GFP_KERNEL);
+	if (!in) {
+		err = -ENOMEM;
+		goto err_free;
+	}
+
+	mkc = MLX5_ADDR_OF(create_mkey_in, in, memory_key_mkey_entry);
+
+	MLX5_SET(mkc, mkc, access_mode_1_0, MLX5_MKC_ACCESS_MODE_MEMIC & 0x3);
+	MLX5_SET(mkc, mkc, access_mode_4_2,
+		 (MLX5_MKC_ACCESS_MODE_MEMIC >> 2) & 0x7);
+	MLX5_SET(mkc, mkc, a, !!(acc & IB_ACCESS_REMOTE_ATOMIC));
+	MLX5_SET(mkc, mkc, rw, !!(acc & IB_ACCESS_REMOTE_WRITE));
+	MLX5_SET(mkc, mkc, rr, !!(acc & IB_ACCESS_REMOTE_READ));
+	MLX5_SET(mkc, mkc, lw, !!(acc & IB_ACCESS_LOCAL_WRITE));
+	MLX5_SET(mkc, mkc, lr, 1);
+
+	MLX5_SET64(mkc, mkc, len, length);
+	MLX5_SET(mkc, mkc, pd, to_mpd(pd)->pdn);
+	MLX5_SET(mkc, mkc, qpn, 0xffffff);
+	MLX5_SET64(mkc, mkc, start_addr,
+		   memic_addr - pci_resource_start(dev->mdev->pdev, 0));
+
+	err = mlx5_core_create_mkey(mdev, &mr->mmkey, in, inlen);
+	if (err)
+		goto err_in;
+
+	kfree(in);
+
+	mr->umem = NULL;
+	set_mr_fileds(dev, mr, 0, length, acc);
+
+	return &mr->ibmr;
+
+err_in:
+	kfree(in);
+
+err_free:
+	kfree(mr);
+
+	return ERR_PTR(err);
+}
+
+struct ib_mr *mlx5_ib_reg_dm_mr(struct ib_pd *pd, struct ib_dm *dm,
+				struct ib_dm_mr_attr *attr,
+				struct uverbs_attr_bundle *attrs)
+{
+	struct mlx5_ib_dm *mdm = to_mdm(dm);
+	u64 memic_addr;
+
+	if (attr->access_flags & ~MLX5_IB_DM_ALLOWED_ACCESS)
+		return ERR_PTR(-EINVAL);
+
+	memic_addr = mdm->dev_addr + attr->offset;
+
+	return mlx5_ib_get_memic_mr(pd, memic_addr, attr->length,
+				    attr->access_flags);
+}
+
+struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
+				  u64 virt_addr, int access_flags,
+				  struct ib_udata *udata)
+{
+	struct mlx5_ib_dev *dev = to_mdev(pd->device);
+	struct mlx5_ib_mr *mr = NULL;
+	bool use_umr;
+	struct ib_umem *umem;
+	int page_shift;
+	int npages;
+	int ncont;
+	int order;
+	int err;
+
+	if (!IS_ENABLED(CONFIG_INFINIBAND_USER_MEM))
+		return ERR_PTR(-EOPNOTSUPP);
+
+	mlx5_ib_dbg(dev, "start 0x%llx, virt_addr 0x%llx, length 0x%llx, access_flags 0x%x\n",
+		    start, virt_addr, length, access_flags);
+
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
+	if (!start && length == U64_MAX) {
+		if (!(access_flags & IB_ACCESS_ON_DEMAND) ||
+		    !(dev->odp_caps.general_caps & IB_ODP_SUPPORT_IMPLICIT))
+			return ERR_PTR(-EINVAL);
+
+		mr = mlx5_ib_alloc_implicit_mr(to_mpd(pd), access_flags);
+		if (IS_ERR(mr))
+			return ERR_CAST(mr);
+		return &mr->ibmr;
+	}
+#endif
+
+	err = mr_umem_get(pd, start, length, access_flags, &umem, &npages,
+			   &page_shift, &ncont, &order);
+
+	if (err < 0)
+		return ERR_PTR(err);
+
+	use_umr = !MLX5_CAP_GEN(dev->mdev, umr_modify_entity_size_disabled) &&
+		  (!MLX5_CAP_GEN(dev->mdev, umr_modify_atomic_disabled) ||
+		   !MLX5_CAP_GEN(dev->mdev, atomic));
+
+	if (order <= mr_cache_max_order(dev) && use_umr) {
+		mr = alloc_mr_from_cache(pd, umem, virt_addr, length, ncont,
+					 page_shift, order, access_flags);
+		if (PTR_ERR(mr) == -EAGAIN) {
+			mlx5_ib_dbg(dev, "cache empty for order %d\n", order);
+			mr = NULL;
+		}
+	} else if (!MLX5_CAP_GEN(dev->mdev, umr_extended_translation_offset)) {
+		if (access_flags & IB_ACCESS_ON_DEMAND) {
+			err = -EINVAL;
+			pr_err("Got MR registration for ODP MR > 512MB, not supported for Connect-IB\n");
+			goto error;
+		}
+		use_umr = false;
+	}
+
+	if (!mr) {
+		mutex_lock(&dev->slow_path_mutex);
+		mr = reg_create(NULL, pd, virt_addr, length, umem, ncont,
+				page_shift, access_flags, !use_umr);
+		mutex_unlock(&dev->slow_path_mutex);
+	}
+
+	if (IS_ERR(mr)) {
+		err = PTR_ERR(mr);
+		goto error;
+	}
+
+	mlx5_ib_dbg(dev, "mkey 0x%x\n", mr->mmkey.key);
+
+	mr->umem = umem;
+	set_mr_fileds(dev, mr, npages, length, access_flags);
+
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
+	update_odp_mr(mr);
+#endif
+
+	if (use_umr) {
+		int update_xlt_flags = MLX5_IB_UPD_XLT_ENABLE;
+
+		if (access_flags & IB_ACCESS_ON_DEMAND)
+			update_xlt_flags |= MLX5_IB_UPD_XLT_ZAP;
+
+		err = mlx5_ib_update_xlt(mr, 0, ncont, page_shift,
+					 update_xlt_flags);
+
+		if (err) {
+			dereg_mr(dev, mr);
+			return ERR_PTR(err);
+		}
+	}
+
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
+	mr->live = 1;
+#endif
+	return &mr->ibmr;
+error:
+	ib_umem_release(umem);
+	return ERR_PTR(err);
+}
+
+static int unreg_umr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
+{
+	struct mlx5_core_dev *mdev = dev->mdev;
+	struct mlx5_umr_wr umrwr = {};
+
+	if (mdev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR)
+		return 0;
+
+	umrwr.wr.send_flags = MLX5_IB_SEND_UMR_DISABLE_MR |
+			      MLX5_IB_SEND_UMR_UPDATE_PD_ACCESS;
+	umrwr.wr.opcode = MLX5_IB_WR_UMR;
+	umrwr.pd = dev->umrc.pd;
+	umrwr.mkey = mr->mmkey.key;
+	umrwr.ignore_free_state = 1;
+
+	return mlx5_ib_post_send_wait(dev, &umrwr);
+}
+
+static int rereg_umr(struct ib_pd *pd, struct mlx5_ib_mr *mr,
+		     int access_flags, int flags)
+{
+	struct mlx5_ib_dev *dev = to_mdev(pd->device);
+	struct mlx5_umr_wr umrwr = {};
+	int err;
+
+	umrwr.wr.send_flags = MLX5_IB_SEND_UMR_FAIL_IF_FREE;
+
+	umrwr.wr.opcode = MLX5_IB_WR_UMR;
+	umrwr.mkey = mr->mmkey.key;
+
+	if (flags & IB_MR_REREG_PD || flags & IB_MR_REREG_ACCESS) {
+		umrwr.pd = pd;
+		umrwr.access_flags = access_flags;
+		umrwr.wr.send_flags |= MLX5_IB_SEND_UMR_UPDATE_PD_ACCESS;
+	}
+
+	err = mlx5_ib_post_send_wait(dev, &umrwr);
+
+	return err;
+}
+
+int mlx5_ib_rereg_user_mr(struct ib_mr *ib_mr, int flags, u64 start,
+			  u64 length, u64 virt_addr, int new_access_flags,
+			  struct ib_pd *new_pd, struct ib_udata *udata)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ib_mr->device);
+	struct mlx5_ib_mr *mr = to_mmr(ib_mr);
+	struct ib_pd *pd = (flags & IB_MR_REREG_PD) ? new_pd : ib_mr->pd;
+	int access_flags = flags & IB_MR_REREG_ACCESS ?
+			    new_access_flags :
+			    mr->access_flags;
+	int page_shift = 0;
+	int upd_flags = 0;
+	int npages = 0;
+	int ncont = 0;
+	int order = 0;
+	u64 addr, len;
+	int err;
+
+	mlx5_ib_dbg(dev, "start 0x%llx, virt_addr 0x%llx, length 0x%llx, access_flags 0x%x\n",
+		    start, virt_addr, length, access_flags);
+
+	atomic_sub(mr->npages, &dev->mdev->priv.reg_pages);
+
+	if (!mr->umem)
+		return -EINVAL;
+
+	if (flags & IB_MR_REREG_TRANS) {
+		addr = virt_addr;
+		len = length;
+	} else {
+		addr = mr->umem->address;
+		len = mr->umem->length;
+	}
+
+	if (flags != IB_MR_REREG_PD) {
+		/*
+		 * Replace umem. This needs to be done whether or not UMR is
+		 * used.
+		 */
+		flags |= IB_MR_REREG_TRANS;
+		ib_umem_release(mr->umem);
+		mr->umem = NULL;
+		err = mr_umem_get(pd, addr, len, access_flags, &mr->umem,
+				  &npages, &page_shift, &ncont, &order);
+		if (err)
+			goto err;
+	}
+
+	if (flags & IB_MR_REREG_TRANS && !use_umr_mtt_update(mr, addr, len)) {
+		/*
+		 * UMR can't be used - MKey needs to be replaced.
+		 */
+		if (mr->allocated_from_cache)
+			err = unreg_umr(dev, mr);
+		else
+			err = destroy_mkey(dev, mr);
+		if (err)
+			goto err;
+
+		mr = reg_create(ib_mr, pd, addr, len, mr->umem, ncont,
+				page_shift, access_flags, true);
+
+		if (IS_ERR(mr)) {
+			err = PTR_ERR(mr);
+			mr = to_mmr(ib_mr);
+			goto err;
+		}
+
+		mr->allocated_from_cache = 0;
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
+		mr->live = 1;
+#endif
+	} else {
+		/*
+		 * Send a UMR WQE
+		 */
+		mr->ibmr.pd = pd;
+		mr->access_flags = access_flags;
+		mr->mmkey.iova = addr;
+		mr->mmkey.size = len;
+		mr->mmkey.pd = to_mpd(pd)->pdn;
+
+		if (flags & IB_MR_REREG_TRANS) {
+			upd_flags = MLX5_IB_UPD_XLT_ADDR;
+			if (flags & IB_MR_REREG_PD)
+				upd_flags |= MLX5_IB_UPD_XLT_PD;
+			if (flags & IB_MR_REREG_ACCESS)
+				upd_flags |= MLX5_IB_UPD_XLT_ACCESS;
+			err = mlx5_ib_update_xlt(mr, 0, npages, page_shift,
+						 upd_flags);
+		} else {
+			err = rereg_umr(pd, mr, access_flags, flags);
+		}
+
+		if (err)
+			goto err;
+	}
+
+	set_mr_fileds(dev, mr, npages, len, access_flags);
+
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
+	update_odp_mr(mr);
+#endif
+	return 0;
+
+err:
+	if (mr->umem) {
+		ib_umem_release(mr->umem);
+		mr->umem = NULL;
+	}
+	clean_mr(dev, mr);
+	return err;
+}
+
+static int
+mlx5_alloc_priv_descs(struct ib_device *device,
+		      struct mlx5_ib_mr *mr,
+		      int ndescs,
+		      int desc_size)
+{
+	int size = ndescs * desc_size;
+	int add_size;
+	int ret;
+
+	add_size = max_t(int, MLX5_UMR_ALIGN - ARCH_KMALLOC_MINALIGN, 0);
+
+	mr->descs_alloc = kzalloc(size + add_size, GFP_KERNEL);
+	if (!mr->descs_alloc)
+		return -ENOMEM;
+
+	mr->descs = PTR_ALIGN(mr->descs_alloc, MLX5_UMR_ALIGN);
+
+	mr->desc_map = dma_map_single(device->dev.parent, mr->descs,
+				      size, DMA_TO_DEVICE);
+	if (dma_mapping_error(device->dev.parent, mr->desc_map)) {
+		ret = -ENOMEM;
+		goto err;
+	}
+
+	return 0;
+err:
+	kfree(mr->descs_alloc);
+
+	return ret;
+}
+
+static void
+mlx5_free_priv_descs(struct mlx5_ib_mr *mr)
+{
+	if (mr->descs) {
+		struct ib_device *device = mr->ibmr.device;
+		int size = mr->max_descs * mr->desc_size;
+
+		dma_unmap_single(device->dev.parent, mr->desc_map,
+				 size, DMA_TO_DEVICE);
+		kfree(mr->descs_alloc);
+		mr->descs = NULL;
+	}
+}
+
+static void clean_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
+{
+	int allocated_from_cache = mr->allocated_from_cache;
+
+	if (mr->sig) {
+		if (mlx5_core_destroy_psv(dev->mdev,
+					  mr->sig->psv_memory.psv_idx))
+			mlx5_ib_warn(dev, "failed to destroy mem psv %d\n",
+				     mr->sig->psv_memory.psv_idx);
+		if (mlx5_core_destroy_psv(dev->mdev,
+					  mr->sig->psv_wire.psv_idx))
+			mlx5_ib_warn(dev, "failed to destroy wire psv %d\n",
+				     mr->sig->psv_wire.psv_idx);
+		kfree(mr->sig);
+		mr->sig = NULL;
+	}
+
+	if (!allocated_from_cache) {
+		destroy_mkey(dev, mr);
+		mlx5_free_priv_descs(mr);
+	}
+}
+
+static void dereg_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
+{
+	int npages = mr->npages;
+	struct ib_umem *umem = mr->umem;
+
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
+	if (umem && umem->odp_data) {
+		/* Prevent new page faults from succeeding */
+		mr->live = 0;
+		/* Wait for all running page-fault handlers to finish. */
+		synchronize_srcu(&dev->mr_srcu);
+		/* Destroy all page mappings */
+		if (umem->odp_data->page_list)
+			mlx5_ib_invalidate_range(umem, ib_umem_start(umem),
+						 ib_umem_end(umem));
+		else
+			mlx5_ib_free_implicit_mr(mr);
+		/*
+		 * We kill the umem before the MR for ODP,
+		 * so that there will not be any invalidations in
+		 * flight, looking at the *mr struct.
+		 */
+		ib_umem_release(umem);
+		atomic_sub(npages, &dev->mdev->priv.reg_pages);
+
+		/* Avoid double-freeing the umem. */
+		umem = NULL;
+	}
+#endif
+	clean_mr(dev, mr);
+
+	/*
+	 * We should unregister the DMA address from the HCA before
+	 * remove the DMA mapping.
+	 */
+	mlx5_mr_cache_free(dev, mr);
+	if (umem) {
+		ib_umem_release(umem);
+		atomic_sub(npages, &dev->mdev->priv.reg_pages);
+	}
+	if (!mr->allocated_from_cache)
+		kfree(mr);
+}
+
+int mlx5_ib_dereg_mr(struct ib_mr *ibmr)
+{
+	dereg_mr(to_mdev(ibmr->device), to_mmr(ibmr));
+	return 0;
+}
+
+struct ib_mr *mlx5_ib_alloc_mr(struct ib_pd *pd,
+			       enum ib_mr_type mr_type,
+			       u32 max_num_sg)
+{
+	struct mlx5_ib_dev *dev = to_mdev(pd->device);
+	int inlen = MLX5_ST_SZ_BYTES(create_mkey_in);
+	int ndescs = ALIGN(max_num_sg, 4);
+	struct mlx5_ib_mr *mr;
+	void *mkc;
+	u32 *in;
+	int err;
+
+	mr = kzalloc(sizeof(*mr), GFP_KERNEL);
+	if (!mr)
+		return ERR_PTR(-ENOMEM);
+
+	in = kzalloc(inlen, GFP_KERNEL);
+	if (!in) {
+		err = -ENOMEM;
+		goto err_free;
+	}
+
+	mkc = MLX5_ADDR_OF(create_mkey_in, in, memory_key_mkey_entry);
+	MLX5_SET(mkc, mkc, free, 1);
+	MLX5_SET(mkc, mkc, translations_octword_size, ndescs);
+	MLX5_SET(mkc, mkc, qpn, 0xffffff);
+	MLX5_SET(mkc, mkc, pd, to_mpd(pd)->pdn);
+
+	if (mr_type == IB_MR_TYPE_MEM_REG) {
+		mr->access_mode = MLX5_MKC_ACCESS_MODE_MTT;
+		MLX5_SET(mkc, mkc, log_page_size, PAGE_SHIFT);
+		err = mlx5_alloc_priv_descs(pd->device, mr,
+					    ndescs, sizeof(struct mlx5_mtt));
+		if (err)
+			goto err_free_in;
+
+		mr->desc_size = sizeof(struct mlx5_mtt);
+		mr->max_descs = ndescs;
+	} else if (mr_type == IB_MR_TYPE_SG_GAPS) {
+		mr->access_mode = MLX5_MKC_ACCESS_MODE_KLMS;
+
+		err = mlx5_alloc_priv_descs(pd->device, mr,
+					    ndescs, sizeof(struct mlx5_klm));
+		if (err)
+			goto err_free_in;
+		mr->desc_size = sizeof(struct mlx5_klm);
+		mr->max_descs = ndescs;
+	} else if (mr_type == IB_MR_TYPE_SIGNATURE) {
+		u32 psv_index[2];
+
+		MLX5_SET(mkc, mkc, bsf_en, 1);
+		MLX5_SET(mkc, mkc, bsf_octword_size, MLX5_MKEY_BSF_OCTO_SIZE);
+		mr->sig = kzalloc(sizeof(*mr->sig), GFP_KERNEL);
+		if (!mr->sig) {
+			err = -ENOMEM;
+			goto err_free_in;
+		}
+
+		/* create mem & wire PSVs */
+		err = mlx5_core_create_psv(dev->mdev, to_mpd(pd)->pdn,
+					   2, psv_index);
+		if (err)
+			goto err_free_sig;
+
+		mr->access_mode = MLX5_MKC_ACCESS_MODE_KLMS;
+		mr->sig->psv_memory.psv_idx = psv_index[0];
+		mr->sig->psv_wire.psv_idx = psv_index[1];
+
+		mr->sig->sig_status_checked = true;
+		mr->sig->sig_err_exists = false;
+		/* Next UMR, Arm SIGERR */
+		++mr->sig->sigerr_count;
+	} else {
+		mlx5_ib_warn(dev, "Invalid mr type %d\n", mr_type);
+		err = -EINVAL;
+		goto err_free_in;
+	}
+
+	MLX5_SET(mkc, mkc, access_mode_1_0, mr->access_mode & 0x3);
+	MLX5_SET(mkc, mkc, access_mode_4_2, (mr->access_mode >> 2) & 0x7);
+	MLX5_SET(mkc, mkc, umr_en, 1);
+
+	mr->ibmr.device = pd->device;
+	err = mlx5_core_create_mkey(dev->mdev, &mr->mmkey, in, inlen);
+	if (err)
+		goto err_destroy_psv;
+
+	mr->mmkey.type = MLX5_MKEY_MR;
+	mr->ibmr.lkey = mr->mmkey.key;
+	mr->ibmr.rkey = mr->mmkey.key;
+	mr->umem = NULL;
+	kfree(in);
+
+	return &mr->ibmr;
+
+err_destroy_psv:
+	if (mr->sig) {
+		if (mlx5_core_destroy_psv(dev->mdev,
+					  mr->sig->psv_memory.psv_idx))
+			mlx5_ib_warn(dev, "failed to destroy mem psv %d\n",
+				     mr->sig->psv_memory.psv_idx);
+		if (mlx5_core_destroy_psv(dev->mdev,
+					  mr->sig->psv_wire.psv_idx))
+			mlx5_ib_warn(dev, "failed to destroy wire psv %d\n",
+				     mr->sig->psv_wire.psv_idx);
+	}
+	mlx5_free_priv_descs(mr);
+err_free_sig:
+	kfree(mr->sig);
+err_free_in:
+	kfree(in);
+err_free:
+	kfree(mr);
+	return ERR_PTR(err);
+}
+
+struct ib_mw *mlx5_ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
+			       struct ib_udata *udata)
+{
+	struct mlx5_ib_dev *dev = to_mdev(pd->device);
+	int inlen = MLX5_ST_SZ_BYTES(create_mkey_in);
+	struct mlx5_ib_mw *mw = NULL;
+	u32 *in = NULL;
+	void *mkc;
+	int ndescs;
+	int err;
+	struct mlx5_ib_alloc_mw req = {};
+	struct {
+		__u32	comp_mask;
+		__u32	response_length;
+	} resp = {};
+
+	err = ib_copy_from_udata(&req, udata, min(udata->inlen, sizeof(req)));
+	if (err)
+		return ERR_PTR(err);
+
+	if (req.comp_mask || req.reserved1 || req.reserved2)
+		return ERR_PTR(-EOPNOTSUPP);
+
+	if (udata->inlen > sizeof(req) &&
+	    !ib_is_udata_cleared(udata, sizeof(req),
+				 udata->inlen - sizeof(req)))
+		return ERR_PTR(-EOPNOTSUPP);
+
+	ndescs = req.num_klms ? roundup(req.num_klms, 4) : roundup(1, 4);
+
+	mw = kzalloc(sizeof(*mw), GFP_KERNEL);
+	in = kzalloc(inlen, GFP_KERNEL);
+	if (!mw || !in) {
+		err = -ENOMEM;
+		goto free;
+	}
+
+	mkc = MLX5_ADDR_OF(create_mkey_in, in, memory_key_mkey_entry);
+
+	MLX5_SET(mkc, mkc, free, 1);
+	MLX5_SET(mkc, mkc, translations_octword_size, ndescs);
+	MLX5_SET(mkc, mkc, pd, to_mpd(pd)->pdn);
+	MLX5_SET(mkc, mkc, umr_en, 1);
+	MLX5_SET(mkc, mkc, lr, 1);
+	MLX5_SET(mkc, mkc, access_mode_1_0, MLX5_MKC_ACCESS_MODE_KLMS);
+	MLX5_SET(mkc, mkc, en_rinval, !!((type == IB_MW_TYPE_2)));
+	MLX5_SET(mkc, mkc, qpn, 0xffffff);
+
+	err = mlx5_core_create_mkey(dev->mdev, &mw->mmkey, in, inlen);
+	if (err)
+		goto free;
+
+	mw->mmkey.type = MLX5_MKEY_MW;
+	mw->ibmw.rkey = mw->mmkey.key;
+	mw->ndescs = ndescs;
+
+	resp.response_length = min(offsetof(typeof(resp), response_length) +
+				   sizeof(resp.response_length), udata->outlen);
+	if (resp.response_length) {
+		err = ib_copy_to_udata(udata, &resp, resp.response_length);
+		if (err) {
+			mlx5_core_destroy_mkey(dev->mdev, &mw->mmkey);
+			goto free;
+		}
+	}
+
+	kfree(in);
+	return &mw->ibmw;
+
+free:
+	kfree(mw);
+	kfree(in);
+	return ERR_PTR(err);
+}
+
+int mlx5_ib_dealloc_mw(struct ib_mw *mw)
+{
+	struct mlx5_ib_mw *mmw = to_mmw(mw);
+	int err;
+
+	err =  mlx5_core_destroy_mkey((to_mdev(mw->device))->mdev,
+				      &mmw->mmkey);
+	if (!err)
+		kfree(mmw);
+	return err;
+}
+
+int mlx5_ib_check_mr_status(struct ib_mr *ibmr, u32 check_mask,
+			    struct ib_mr_status *mr_status)
+{
+	struct mlx5_ib_mr *mmr = to_mmr(ibmr);
+	int ret = 0;
+
+	if (check_mask & ~IB_MR_CHECK_SIG_STATUS) {
+		pr_err("Invalid status check mask\n");
+		ret = -EINVAL;
+		goto done;
+	}
+
+	mr_status->fail_status = 0;
+	if (check_mask & IB_MR_CHECK_SIG_STATUS) {
+		if (!mmr->sig) {
+			ret = -EINVAL;
+			pr_err("signature status check requested on a non-signature enabled MR\n");
+			goto done;
+		}
+
+		mmr->sig->sig_status_checked = true;
+		if (!mmr->sig->sig_err_exists)
+			goto done;
+
+		if (ibmr->lkey == mmr->sig->err_item.key)
+			memcpy(&mr_status->sig_err, &mmr->sig->err_item,
+			       sizeof(mr_status->sig_err));
+		else {
+			mr_status->sig_err.err_type = IB_SIG_BAD_GUARD;
+			mr_status->sig_err.sig_err_offset = 0;
+			mr_status->sig_err.key = mmr->sig->err_item.key;
+		}
+
+		mmr->sig->sig_err_exists = false;
+		mr_status->fail_status |= IB_MR_CHECK_SIG_STATUS;
+	}
+
+done:
+	return ret;
+}
+
+static int
+mlx5_ib_sg_to_klms(struct mlx5_ib_mr *mr,
+		   struct scatterlist *sgl,
+		   unsigned short sg_nents,
+		   unsigned int *sg_offset_p)
+{
+	struct scatterlist *sg = sgl;
+	struct mlx5_klm *klms = mr->descs;
+	unsigned int sg_offset = sg_offset_p ? *sg_offset_p : 0;
+	u32 lkey = mr->ibmr.pd->local_dma_lkey;
+	int i;
+
+	mr->ibmr.iova = sg_dma_address(sg) + sg_offset;
+	mr->ibmr.length = 0;
+
+	for_each_sg(sgl, sg, sg_nents, i) {
+		if (unlikely(i >= mr->max_descs))
+			break;
+		klms[i].va = cpu_to_be64(sg_dma_address(sg) + sg_offset);
+		klms[i].bcount = cpu_to_be32(sg_dma_len(sg) - sg_offset);
+		klms[i].key = cpu_to_be32(lkey);
+		mr->ibmr.length += sg_dma_len(sg) - sg_offset;
+
+		sg_offset = 0;
+	}
+	mr->ndescs = i;
+
+	if (sg_offset_p)
+		*sg_offset_p = sg_offset;
+
+	return i;
+}
+
+static int mlx5_set_page(struct ib_mr *ibmr, u64 addr)
+{
+	struct mlx5_ib_mr *mr = to_mmr(ibmr);
+	__be64 *descs;
+
+	if (unlikely(mr->ndescs == mr->max_descs))
+		return -ENOMEM;
+
+	descs = mr->descs;
+	descs[mr->ndescs++] = cpu_to_be64(addr | MLX5_EN_RD | MLX5_EN_WR);
+
+	return 0;
+}
+
+int mlx5_ib_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg, int sg_nents,
+		      unsigned int *sg_offset)
+{
+	struct mlx5_ib_mr *mr = to_mmr(ibmr);
+	int n;
+
+	mr->ndescs = 0;
+
+	ib_dma_sync_single_for_cpu(ibmr->device, mr->desc_map,
+				   mr->desc_size * mr->max_descs,
+				   DMA_TO_DEVICE);
+
+	if (mr->access_mode == MLX5_MKC_ACCESS_MODE_KLMS)
+		n = mlx5_ib_sg_to_klms(mr, sg, sg_nents, sg_offset);
+	else
+		n = ib_sg_to_pages(ibmr, sg, sg_nents, sg_offset,
+				mlx5_set_page);
+
+	ib_dma_sync_single_for_device(ibmr->device, mr->desc_map,
+				      mr->desc_size * mr->max_descs,
+				      DMA_TO_DEVICE);
+
+	return n;
+}
diff --git a/drivers/infiniband/hw/mlx5/odp.c b/drivers/infiniband/hw/mlx5/odp.c
new file mode 100644
index 000000000..453e5c4ac
--- /dev/null
+++ b/drivers/infiniband/hw/mlx5/odp.c
@@ -0,0 +1,1230 @@
+/*
+ * Copyright (c) 2013-2015, Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <rdma/ib_umem.h>
+#include <rdma/ib_umem_odp.h>
+#include <linux/kernel.h>
+
+#include "mlx5_ib.h"
+#include "cmd.h"
+
+#define MAX_PREFETCH_LEN (4*1024*1024U)
+
+/* Timeout in ms to wait for an active mmu notifier to complete when handling
+ * a pagefault. */
+#define MMU_NOTIFIER_TIMEOUT 1000
+
+#define MLX5_IMR_MTT_BITS (30 - PAGE_SHIFT)
+#define MLX5_IMR_MTT_SHIFT (MLX5_IMR_MTT_BITS + PAGE_SHIFT)
+#define MLX5_IMR_MTT_ENTRIES BIT_ULL(MLX5_IMR_MTT_BITS)
+#define MLX5_IMR_MTT_SIZE BIT_ULL(MLX5_IMR_MTT_SHIFT)
+#define MLX5_IMR_MTT_MASK (~(MLX5_IMR_MTT_SIZE - 1))
+
+#define MLX5_KSM_PAGE_SHIFT MLX5_IMR_MTT_SHIFT
+
+static u64 mlx5_imr_ksm_entries;
+
+static int check_parent(struct ib_umem_odp *odp,
+			       struct mlx5_ib_mr *parent)
+{
+	struct mlx5_ib_mr *mr = odp->private;
+
+	return mr && mr->parent == parent && !odp->dying;
+}
+
+static struct ib_umem_odp *odp_next(struct ib_umem_odp *odp)
+{
+	struct mlx5_ib_mr *mr = odp->private, *parent = mr->parent;
+	struct ib_ucontext *ctx = odp->umem->context;
+	struct rb_node *rb;
+
+	down_read(&ctx->umem_rwsem);
+	while (1) {
+		rb = rb_next(&odp->interval_tree.rb);
+		if (!rb)
+			goto not_found;
+		odp = rb_entry(rb, struct ib_umem_odp, interval_tree.rb);
+		if (check_parent(odp, parent))
+			goto end;
+	}
+not_found:
+	odp = NULL;
+end:
+	up_read(&ctx->umem_rwsem);
+	return odp;
+}
+
+static struct ib_umem_odp *odp_lookup(struct ib_ucontext *ctx,
+				      u64 start, u64 length,
+				      struct mlx5_ib_mr *parent)
+{
+	struct ib_umem_odp *odp;
+	struct rb_node *rb;
+
+	down_read(&ctx->umem_rwsem);
+	odp = rbt_ib_umem_lookup(&ctx->umem_tree, start, length);
+	if (!odp)
+		goto end;
+
+	while (1) {
+		if (check_parent(odp, parent))
+			goto end;
+		rb = rb_next(&odp->interval_tree.rb);
+		if (!rb)
+			goto not_found;
+		odp = rb_entry(rb, struct ib_umem_odp, interval_tree.rb);
+		if (ib_umem_start(odp->umem) > start + length)
+			goto not_found;
+	}
+not_found:
+	odp = NULL;
+end:
+	up_read(&ctx->umem_rwsem);
+	return odp;
+}
+
+void mlx5_odp_populate_klm(struct mlx5_klm *pklm, size_t offset,
+			   size_t nentries, struct mlx5_ib_mr *mr, int flags)
+{
+	struct ib_pd *pd = mr->ibmr.pd;
+	struct ib_ucontext *ctx = pd->uobject->context;
+	struct mlx5_ib_dev *dev = to_mdev(pd->device);
+	struct ib_umem_odp *odp;
+	unsigned long va;
+	int i;
+
+	if (flags & MLX5_IB_UPD_XLT_ZAP) {
+		for (i = 0; i < nentries; i++, pklm++) {
+			pklm->bcount = cpu_to_be32(MLX5_IMR_MTT_SIZE);
+			pklm->key = cpu_to_be32(dev->null_mkey);
+			pklm->va = 0;
+		}
+		return;
+	}
+
+	odp = odp_lookup(ctx, offset * MLX5_IMR_MTT_SIZE,
+			     nentries * MLX5_IMR_MTT_SIZE, mr);
+
+	for (i = 0; i < nentries; i++, pklm++) {
+		pklm->bcount = cpu_to_be32(MLX5_IMR_MTT_SIZE);
+		va = (offset + i) * MLX5_IMR_MTT_SIZE;
+		if (odp && odp->umem->address == va) {
+			struct mlx5_ib_mr *mtt = odp->private;
+
+			pklm->key = cpu_to_be32(mtt->ibmr.lkey);
+			odp = odp_next(odp);
+		} else {
+			pklm->key = cpu_to_be32(dev->null_mkey);
+		}
+		mlx5_ib_dbg(dev, "[%d] va %lx key %x\n",
+			    i, va, be32_to_cpu(pklm->key));
+	}
+}
+
+static void mr_leaf_free_action(struct work_struct *work)
+{
+	struct ib_umem_odp *odp = container_of(work, struct ib_umem_odp, work);
+	int idx = ib_umem_start(odp->umem) >> MLX5_IMR_MTT_SHIFT;
+	struct mlx5_ib_mr *mr = odp->private, *imr = mr->parent;
+
+	mr->parent = NULL;
+	synchronize_srcu(&mr->dev->mr_srcu);
+
+	ib_umem_release(odp->umem);
+	if (imr->live)
+		mlx5_ib_update_xlt(imr, idx, 1, 0,
+				   MLX5_IB_UPD_XLT_INDIRECT |
+				   MLX5_IB_UPD_XLT_ATOMIC);
+	mlx5_mr_cache_free(mr->dev, mr);
+
+	if (atomic_dec_and_test(&imr->num_leaf_free))
+		wake_up(&imr->q_leaf_free);
+}
+
+void mlx5_ib_invalidate_range(struct ib_umem *umem, unsigned long start,
+			      unsigned long end)
+{
+	struct mlx5_ib_mr *mr;
+	const u64 umr_block_mask = (MLX5_UMR_MTT_ALIGNMENT /
+				    sizeof(struct mlx5_mtt)) - 1;
+	u64 idx = 0, blk_start_idx = 0;
+	int in_block = 0;
+	u64 addr;
+
+	if (!umem || !umem->odp_data) {
+		pr_err("invalidation called on NULL umem or non-ODP umem\n");
+		return;
+	}
+
+	mr = umem->odp_data->private;
+
+	if (!mr || !mr->ibmr.pd)
+		return;
+
+	start = max_t(u64, ib_umem_start(umem), start);
+	end = min_t(u64, ib_umem_end(umem), end);
+
+	/*
+	 * Iteration one - zap the HW's MTTs. The notifiers_count ensures that
+	 * while we are doing the invalidation, no page fault will attempt to
+	 * overwrite the same MTTs.  Concurent invalidations might race us,
+	 * but they will write 0s as well, so no difference in the end result.
+	 */
+
+	for (addr = start; addr < end; addr += BIT(umem->page_shift)) {
+		idx = (addr - ib_umem_start(umem)) >> umem->page_shift;
+		/*
+		 * Strive to write the MTTs in chunks, but avoid overwriting
+		 * non-existing MTTs. The huristic here can be improved to
+		 * estimate the cost of another UMR vs. the cost of bigger
+		 * UMR.
+		 */
+		if (umem->odp_data->dma_list[idx] &
+		    (ODP_READ_ALLOWED_BIT | ODP_WRITE_ALLOWED_BIT)) {
+			if (!in_block) {
+				blk_start_idx = idx;
+				in_block = 1;
+			}
+		} else {
+			u64 umr_offset = idx & umr_block_mask;
+
+			if (in_block && umr_offset == 0) {
+				mlx5_ib_update_xlt(mr, blk_start_idx,
+						   idx - blk_start_idx, 0,
+						   MLX5_IB_UPD_XLT_ZAP |
+						   MLX5_IB_UPD_XLT_ATOMIC);
+				in_block = 0;
+			}
+		}
+	}
+	if (in_block)
+		mlx5_ib_update_xlt(mr, blk_start_idx,
+				   idx - blk_start_idx + 1, 0,
+				   MLX5_IB_UPD_XLT_ZAP |
+				   MLX5_IB_UPD_XLT_ATOMIC);
+	/*
+	 * We are now sure that the device will not access the
+	 * memory. We can safely unmap it, and mark it as dirty if
+	 * needed.
+	 */
+
+	ib_umem_odp_unmap_dma_pages(umem, start, end);
+
+	if (unlikely(!umem->npages && mr->parent &&
+		     !umem->odp_data->dying)) {
+		WRITE_ONCE(umem->odp_data->dying, 1);
+		atomic_inc(&mr->parent->num_leaf_free);
+		schedule_work(&umem->odp_data->work);
+	}
+}
+
+void mlx5_ib_internal_fill_odp_caps(struct mlx5_ib_dev *dev)
+{
+	struct ib_odp_caps *caps = &dev->odp_caps;
+
+	memset(caps, 0, sizeof(*caps));
+
+	if (!MLX5_CAP_GEN(dev->mdev, pg))
+		return;
+
+	caps->general_caps = IB_ODP_SUPPORT;
+
+	if (MLX5_CAP_GEN(dev->mdev, umr_extended_translation_offset))
+		dev->odp_max_size = U64_MAX;
+	else
+		dev->odp_max_size = BIT_ULL(MLX5_MAX_UMR_SHIFT + PAGE_SHIFT);
+
+	if (MLX5_CAP_ODP(dev->mdev, ud_odp_caps.send))
+		caps->per_transport_caps.ud_odp_caps |= IB_ODP_SUPPORT_SEND;
+
+	if (MLX5_CAP_ODP(dev->mdev, rc_odp_caps.send))
+		caps->per_transport_caps.rc_odp_caps |= IB_ODP_SUPPORT_SEND;
+
+	if (MLX5_CAP_ODP(dev->mdev, rc_odp_caps.receive))
+		caps->per_transport_caps.rc_odp_caps |= IB_ODP_SUPPORT_RECV;
+
+	if (MLX5_CAP_ODP(dev->mdev, rc_odp_caps.write))
+		caps->per_transport_caps.rc_odp_caps |= IB_ODP_SUPPORT_WRITE;
+
+	if (MLX5_CAP_ODP(dev->mdev, rc_odp_caps.read))
+		caps->per_transport_caps.rc_odp_caps |= IB_ODP_SUPPORT_READ;
+
+	if (MLX5_CAP_ODP(dev->mdev, rc_odp_caps.atomic))
+		caps->per_transport_caps.rc_odp_caps |= IB_ODP_SUPPORT_ATOMIC;
+
+	if (MLX5_CAP_GEN(dev->mdev, fixed_buffer_size) &&
+	    MLX5_CAP_GEN(dev->mdev, null_mkey) &&
+	    MLX5_CAP_GEN(dev->mdev, umr_extended_translation_offset))
+		caps->general_caps |= IB_ODP_SUPPORT_IMPLICIT;
+
+	return;
+}
+
+static void mlx5_ib_page_fault_resume(struct mlx5_ib_dev *dev,
+				      struct mlx5_pagefault *pfault,
+				      int error)
+{
+	int wq_num = pfault->event_subtype == MLX5_PFAULT_SUBTYPE_WQE ?
+		     pfault->wqe.wq_num : pfault->token;
+	int ret = mlx5_core_page_fault_resume(dev->mdev,
+					      pfault->token,
+					      wq_num,
+					      pfault->type,
+					      error);
+	if (ret)
+		mlx5_ib_err(dev, "Failed to resolve the page fault on WQ 0x%x\n",
+			    wq_num);
+}
+
+static struct mlx5_ib_mr *implicit_mr_alloc(struct ib_pd *pd,
+					    struct ib_umem *umem,
+					    bool ksm, int access_flags)
+{
+	struct mlx5_ib_dev *dev = to_mdev(pd->device);
+	struct mlx5_ib_mr *mr;
+	int err;
+
+	mr = mlx5_mr_cache_alloc(dev, ksm ? MLX5_IMR_KSM_CACHE_ENTRY :
+					    MLX5_IMR_MTT_CACHE_ENTRY);
+
+	if (IS_ERR(mr))
+		return mr;
+
+	mr->ibmr.pd = pd;
+
+	mr->dev = dev;
+	mr->access_flags = access_flags;
+	mr->mmkey.iova = 0;
+	mr->umem = umem;
+
+	if (ksm) {
+		err = mlx5_ib_update_xlt(mr, 0,
+					 mlx5_imr_ksm_entries,
+					 MLX5_KSM_PAGE_SHIFT,
+					 MLX5_IB_UPD_XLT_INDIRECT |
+					 MLX5_IB_UPD_XLT_ZAP |
+					 MLX5_IB_UPD_XLT_ENABLE);
+
+	} else {
+		err = mlx5_ib_update_xlt(mr, 0,
+					 MLX5_IMR_MTT_ENTRIES,
+					 PAGE_SHIFT,
+					 MLX5_IB_UPD_XLT_ZAP |
+					 MLX5_IB_UPD_XLT_ENABLE |
+					 MLX5_IB_UPD_XLT_ATOMIC);
+	}
+
+	if (err)
+		goto fail;
+
+	mr->ibmr.lkey = mr->mmkey.key;
+	mr->ibmr.rkey = mr->mmkey.key;
+
+	mr->live = 1;
+
+	mlx5_ib_dbg(dev, "key %x dev %p mr %p\n",
+		    mr->mmkey.key, dev->mdev, mr);
+
+	return mr;
+
+fail:
+	mlx5_ib_err(dev, "Failed to register MKEY %d\n", err);
+	mlx5_mr_cache_free(dev, mr);
+
+	return ERR_PTR(err);
+}
+
+static struct ib_umem_odp *implicit_mr_get_data(struct mlx5_ib_mr *mr,
+						u64 io_virt, size_t bcnt)
+{
+	struct ib_ucontext *ctx = mr->ibmr.pd->uobject->context;
+	struct mlx5_ib_dev *dev = to_mdev(mr->ibmr.pd->device);
+	struct ib_umem_odp *odp, *result = NULL;
+	u64 addr = io_virt & MLX5_IMR_MTT_MASK;
+	int nentries = 0, start_idx = 0, ret;
+	struct mlx5_ib_mr *mtt;
+	struct ib_umem *umem;
+
+	mutex_lock(&mr->umem->odp_data->umem_mutex);
+	odp = odp_lookup(ctx, addr, 1, mr);
+
+	mlx5_ib_dbg(dev, "io_virt:%llx bcnt:%zx addr:%llx odp:%p\n",
+		    io_virt, bcnt, addr, odp);
+
+next_mr:
+	if (likely(odp)) {
+		if (nentries)
+			nentries++;
+	} else {
+		umem = ib_alloc_odp_umem(ctx, addr, MLX5_IMR_MTT_SIZE);
+		if (IS_ERR(umem)) {
+			mutex_unlock(&mr->umem->odp_data->umem_mutex);
+			return ERR_CAST(umem);
+		}
+
+		mtt = implicit_mr_alloc(mr->ibmr.pd, umem, 0, mr->access_flags);
+		if (IS_ERR(mtt)) {
+			mutex_unlock(&mr->umem->odp_data->umem_mutex);
+			ib_umem_release(umem);
+			return ERR_CAST(mtt);
+		}
+
+		odp = umem->odp_data;
+		odp->private = mtt;
+		mtt->umem = umem;
+		mtt->mmkey.iova = addr;
+		mtt->parent = mr;
+		INIT_WORK(&odp->work, mr_leaf_free_action);
+
+		if (!nentries)
+			start_idx = addr >> MLX5_IMR_MTT_SHIFT;
+		nentries++;
+	}
+
+	/* Return first odp if region not covered by single one */
+	if (likely(!result))
+		result = odp;
+
+	addr += MLX5_IMR_MTT_SIZE;
+	if (unlikely(addr < io_virt + bcnt)) {
+		odp = odp_next(odp);
+		if (odp && odp->umem->address != addr)
+			odp = NULL;
+		goto next_mr;
+	}
+
+	if (unlikely(nentries)) {
+		ret = mlx5_ib_update_xlt(mr, start_idx, nentries, 0,
+					 MLX5_IB_UPD_XLT_INDIRECT |
+					 MLX5_IB_UPD_XLT_ATOMIC);
+		if (ret) {
+			mlx5_ib_err(dev, "Failed to update PAS\n");
+			result = ERR_PTR(ret);
+		}
+	}
+
+	mutex_unlock(&mr->umem->odp_data->umem_mutex);
+	return result;
+}
+
+struct mlx5_ib_mr *mlx5_ib_alloc_implicit_mr(struct mlx5_ib_pd *pd,
+					     int access_flags)
+{
+	struct ib_ucontext *ctx = pd->ibpd.uobject->context;
+	struct mlx5_ib_mr *imr;
+	struct ib_umem *umem;
+
+	umem = ib_umem_get(ctx, 0, 0, IB_ACCESS_ON_DEMAND, 0);
+	if (IS_ERR(umem))
+		return ERR_CAST(umem);
+
+	imr = implicit_mr_alloc(&pd->ibpd, umem, 1, access_flags);
+	if (IS_ERR(imr)) {
+		ib_umem_release(umem);
+		return ERR_CAST(imr);
+	}
+
+	imr->umem = umem;
+	init_waitqueue_head(&imr->q_leaf_free);
+	atomic_set(&imr->num_leaf_free, 0);
+
+	return imr;
+}
+
+static int mr_leaf_free(struct ib_umem *umem, u64 start,
+			u64 end, void *cookie)
+{
+	struct mlx5_ib_mr *mr = umem->odp_data->private, *imr = cookie;
+
+	if (mr->parent != imr)
+		return 0;
+
+	ib_umem_odp_unmap_dma_pages(umem,
+				    ib_umem_start(umem),
+				    ib_umem_end(umem));
+
+	if (umem->odp_data->dying)
+		return 0;
+
+	WRITE_ONCE(umem->odp_data->dying, 1);
+	atomic_inc(&imr->num_leaf_free);
+	schedule_work(&umem->odp_data->work);
+
+	return 0;
+}
+
+void mlx5_ib_free_implicit_mr(struct mlx5_ib_mr *imr)
+{
+	struct ib_ucontext *ctx = imr->ibmr.pd->uobject->context;
+
+	down_read(&ctx->umem_rwsem);
+	rbt_ib_umem_for_each_in_range(&ctx->umem_tree, 0, ULLONG_MAX,
+				      mr_leaf_free, true, imr);
+	up_read(&ctx->umem_rwsem);
+
+	wait_event(imr->q_leaf_free, !atomic_read(&imr->num_leaf_free));
+}
+
+static int pagefault_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr,
+			u64 io_virt, size_t bcnt, u32 *bytes_mapped)
+{
+	u64 access_mask;
+	int npages = 0, page_shift, np;
+	u64 start_idx, page_mask;
+	struct ib_umem_odp *odp;
+	int current_seq;
+	size_t size;
+	int ret;
+
+	if (!mr->umem->odp_data->page_list) {
+		odp = implicit_mr_get_data(mr, io_virt, bcnt);
+
+		if (IS_ERR(odp))
+			return PTR_ERR(odp);
+		mr = odp->private;
+
+	} else {
+		odp = mr->umem->odp_data;
+	}
+
+next_mr:
+	size = min_t(size_t, bcnt, ib_umem_end(odp->umem) - io_virt);
+
+	page_shift = mr->umem->page_shift;
+	page_mask = ~(BIT(page_shift) - 1);
+	start_idx = (io_virt - (mr->mmkey.iova & page_mask)) >> page_shift;
+	access_mask = ODP_READ_ALLOWED_BIT;
+
+	if (mr->umem->writable)
+		access_mask |= ODP_WRITE_ALLOWED_BIT;
+
+	current_seq = READ_ONCE(odp->notifiers_seq);
+	/*
+	 * Ensure the sequence number is valid for some time before we call
+	 * gup.
+	 */
+	smp_rmb();
+
+	ret = ib_umem_odp_map_dma_pages(mr->umem, io_virt, size,
+					access_mask, current_seq);
+
+	if (ret < 0)
+		goto out;
+
+	np = ret;
+
+	mutex_lock(&odp->umem_mutex);
+	if (!ib_umem_mmu_notifier_retry(mr->umem, current_seq)) {
+		/*
+		 * No need to check whether the MTTs really belong to
+		 * this MR, since ib_umem_odp_map_dma_pages already
+		 * checks this.
+		 */
+		ret = mlx5_ib_update_xlt(mr, start_idx, np,
+					 page_shift, MLX5_IB_UPD_XLT_ATOMIC);
+	} else {
+		ret = -EAGAIN;
+	}
+	mutex_unlock(&odp->umem_mutex);
+
+	if (ret < 0) {
+		if (ret != -EAGAIN)
+			mlx5_ib_err(dev, "Failed to update mkey page tables\n");
+		goto out;
+	}
+
+	if (bytes_mapped) {
+		u32 new_mappings = (np << page_shift) -
+			(io_virt - round_down(io_virt, 1 << page_shift));
+		*bytes_mapped += min_t(u32, new_mappings, size);
+	}
+
+	npages += np << (page_shift - PAGE_SHIFT);
+	bcnt -= size;
+
+	if (unlikely(bcnt)) {
+		struct ib_umem_odp *next;
+
+		io_virt += size;
+		next = odp_next(odp);
+		if (unlikely(!next || next->umem->address != io_virt)) {
+			mlx5_ib_dbg(dev, "next implicit leaf removed at 0x%llx. got %p\n",
+				    io_virt, next);
+			return -EAGAIN;
+		}
+		odp = next;
+		mr = odp->private;
+		goto next_mr;
+	}
+
+	return npages;
+
+out:
+	if (ret == -EAGAIN) {
+		if (mr->parent || !odp->dying) {
+			unsigned long timeout =
+				msecs_to_jiffies(MMU_NOTIFIER_TIMEOUT);
+
+			if (!wait_for_completion_timeout(
+					&odp->notifier_completion,
+					timeout)) {
+				mlx5_ib_warn(dev, "timeout waiting for mmu notifier. seq %d against %d\n",
+					     current_seq, odp->notifiers_seq);
+			}
+		} else {
+			/* The MR is being killed, kill the QP as well. */
+			ret = -EFAULT;
+		}
+	}
+
+	return ret;
+}
+
+struct pf_frame {
+	struct pf_frame *next;
+	u32 key;
+	u64 io_virt;
+	size_t bcnt;
+	int depth;
+};
+
+/*
+ * Handle a single data segment in a page-fault WQE or RDMA region.
+ *
+ * Returns number of OS pages retrieved on success. The caller may continue to
+ * the next data segment.
+ * Can return the following error codes:
+ * -EAGAIN to designate a temporary error. The caller will abort handling the
+ *  page fault and resolve it.
+ * -EFAULT when there's an error mapping the requested pages. The caller will
+ *  abort the page fault handling.
+ */
+static int pagefault_single_data_segment(struct mlx5_ib_dev *dev,
+					 u32 key, u64 io_virt, size_t bcnt,
+					 u32 *bytes_committed,
+					 u32 *bytes_mapped)
+{
+	int npages = 0, srcu_key, ret, i, outlen, cur_outlen = 0, depth = 0;
+	struct pf_frame *head = NULL, *frame;
+	struct mlx5_core_mkey *mmkey;
+	struct mlx5_ib_mw *mw;
+	struct mlx5_ib_mr *mr;
+	struct mlx5_klm *pklm;
+	u32 *out = NULL;
+	size_t offset;
+
+	srcu_key = srcu_read_lock(&dev->mr_srcu);
+
+	io_virt += *bytes_committed;
+	bcnt -= *bytes_committed;
+
+next_mr:
+	mmkey = __mlx5_mr_lookup(dev->mdev, mlx5_base_mkey(key));
+	if (!mmkey || mmkey->key != key) {
+		mlx5_ib_dbg(dev, "failed to find mkey %x\n", key);
+		ret = -EFAULT;
+		goto srcu_unlock;
+	}
+
+	switch (mmkey->type) {
+	case MLX5_MKEY_MR:
+		mr = container_of(mmkey, struct mlx5_ib_mr, mmkey);
+		if (!mr->live || !mr->ibmr.pd) {
+			mlx5_ib_dbg(dev, "got dead MR\n");
+			ret = -EFAULT;
+			goto srcu_unlock;
+		}
+
+		ret = pagefault_mr(dev, mr, io_virt, bcnt, bytes_mapped);
+		if (ret < 0)
+			goto srcu_unlock;
+
+		npages += ret;
+		ret = 0;
+		break;
+
+	case MLX5_MKEY_MW:
+		mw = container_of(mmkey, struct mlx5_ib_mw, mmkey);
+
+		if (depth >= MLX5_CAP_GEN(dev->mdev, max_indirection)) {
+			mlx5_ib_dbg(dev, "indirection level exceeded\n");
+			ret = -EFAULT;
+			goto srcu_unlock;
+		}
+
+		outlen = MLX5_ST_SZ_BYTES(query_mkey_out) +
+			sizeof(*pklm) * (mw->ndescs - 2);
+
+		if (outlen > cur_outlen) {
+			kfree(out);
+			out = kzalloc(outlen, GFP_KERNEL);
+			if (!out) {
+				ret = -ENOMEM;
+				goto srcu_unlock;
+			}
+			cur_outlen = outlen;
+		}
+
+		pklm = (struct mlx5_klm *)MLX5_ADDR_OF(query_mkey_out, out,
+						       bsf0_klm0_pas_mtt0_1);
+
+		ret = mlx5_core_query_mkey(dev->mdev, &mw->mmkey, out, outlen);
+		if (ret)
+			goto srcu_unlock;
+
+		offset = io_virt - MLX5_GET64(query_mkey_out, out,
+					      memory_key_mkey_entry.start_addr);
+
+		for (i = 0; bcnt && i < mw->ndescs; i++, pklm++) {
+			if (offset >= be32_to_cpu(pklm->bcount)) {
+				offset -= be32_to_cpu(pklm->bcount);
+				continue;
+			}
+
+			frame = kzalloc(sizeof(*frame), GFP_KERNEL);
+			if (!frame) {
+				ret = -ENOMEM;
+				goto srcu_unlock;
+			}
+
+			frame->key = be32_to_cpu(pklm->key);
+			frame->io_virt = be64_to_cpu(pklm->va) + offset;
+			frame->bcnt = min_t(size_t, bcnt,
+					    be32_to_cpu(pklm->bcount) - offset);
+			frame->depth = depth + 1;
+			frame->next = head;
+			head = frame;
+
+			bcnt -= frame->bcnt;
+			offset = 0;
+		}
+		break;
+
+	default:
+		mlx5_ib_dbg(dev, "wrong mkey type %d\n", mmkey->type);
+		ret = -EFAULT;
+		goto srcu_unlock;
+	}
+
+	if (head) {
+		frame = head;
+		head = frame->next;
+
+		key = frame->key;
+		io_virt = frame->io_virt;
+		bcnt = frame->bcnt;
+		depth = frame->depth;
+		kfree(frame);
+
+		goto next_mr;
+	}
+
+srcu_unlock:
+	while (head) {
+		frame = head;
+		head = frame->next;
+		kfree(frame);
+	}
+	kfree(out);
+
+	srcu_read_unlock(&dev->mr_srcu, srcu_key);
+	*bytes_committed = 0;
+	return ret ? ret : npages;
+}
+
+/**
+ * Parse a series of data segments for page fault handling.
+ *
+ * @qp the QP on which the fault occurred.
+ * @pfault contains page fault information.
+ * @wqe points at the first data segment in the WQE.
+ * @wqe_end points after the end of the WQE.
+ * @bytes_mapped receives the number of bytes that the function was able to
+ *               map. This allows the caller to decide intelligently whether
+ *               enough memory was mapped to resolve the page fault
+ *               successfully (e.g. enough for the next MTU, or the entire
+ *               WQE).
+ * @total_wqe_bytes receives the total data size of this WQE in bytes (minus
+ *                  the committed bytes).
+ *
+ * Returns the number of pages loaded if positive, zero for an empty WQE, or a
+ * negative error code.
+ */
+static int pagefault_data_segments(struct mlx5_ib_dev *dev,
+				   struct mlx5_pagefault *pfault,
+				   struct mlx5_ib_qp *qp, void *wqe,
+				   void *wqe_end, u32 *bytes_mapped,
+				   u32 *total_wqe_bytes, int receive_queue)
+{
+	int ret = 0, npages = 0;
+	u64 io_virt;
+	u32 key;
+	u32 byte_count;
+	size_t bcnt;
+	int inline_segment;
+
+	/* Skip SRQ next-WQE segment. */
+	if (receive_queue && qp->ibqp.srq)
+		wqe += sizeof(struct mlx5_wqe_srq_next_seg);
+
+	if (bytes_mapped)
+		*bytes_mapped = 0;
+	if (total_wqe_bytes)
+		*total_wqe_bytes = 0;
+
+	while (wqe < wqe_end) {
+		struct mlx5_wqe_data_seg *dseg = wqe;
+
+		io_virt = be64_to_cpu(dseg->addr);
+		key = be32_to_cpu(dseg->lkey);
+		byte_count = be32_to_cpu(dseg->byte_count);
+		inline_segment = !!(byte_count &  MLX5_INLINE_SEG);
+		bcnt	       = byte_count & ~MLX5_INLINE_SEG;
+
+		if (inline_segment) {
+			bcnt = bcnt & MLX5_WQE_INLINE_SEG_BYTE_COUNT_MASK;
+			wqe += ALIGN(sizeof(struct mlx5_wqe_inline_seg) + bcnt,
+				     16);
+		} else {
+			wqe += sizeof(*dseg);
+		}
+
+		/* receive WQE end of sg list. */
+		if (receive_queue && bcnt == 0 && key == MLX5_INVALID_LKEY &&
+		    io_virt == 0)
+			break;
+
+		if (!inline_segment && total_wqe_bytes) {
+			*total_wqe_bytes += bcnt - min_t(size_t, bcnt,
+					pfault->bytes_committed);
+		}
+
+		/* A zero length data segment designates a length of 2GB. */
+		if (bcnt == 0)
+			bcnt = 1U << 31;
+
+		if (inline_segment || bcnt <= pfault->bytes_committed) {
+			pfault->bytes_committed -=
+				min_t(size_t, bcnt,
+				      pfault->bytes_committed);
+			continue;
+		}
+
+		ret = pagefault_single_data_segment(dev, key, io_virt, bcnt,
+						    &pfault->bytes_committed,
+						    bytes_mapped);
+		if (ret < 0)
+			break;
+		npages += ret;
+	}
+
+	return ret < 0 ? ret : npages;
+}
+
+static const u32 mlx5_ib_odp_opcode_cap[] = {
+	[MLX5_OPCODE_SEND]	       = IB_ODP_SUPPORT_SEND,
+	[MLX5_OPCODE_SEND_IMM]	       = IB_ODP_SUPPORT_SEND,
+	[MLX5_OPCODE_SEND_INVAL]       = IB_ODP_SUPPORT_SEND,
+	[MLX5_OPCODE_RDMA_WRITE]       = IB_ODP_SUPPORT_WRITE,
+	[MLX5_OPCODE_RDMA_WRITE_IMM]   = IB_ODP_SUPPORT_WRITE,
+	[MLX5_OPCODE_RDMA_READ]	       = IB_ODP_SUPPORT_READ,
+	[MLX5_OPCODE_ATOMIC_CS]	       = IB_ODP_SUPPORT_ATOMIC,
+	[MLX5_OPCODE_ATOMIC_FA]	       = IB_ODP_SUPPORT_ATOMIC,
+};
+
+/*
+ * Parse initiator WQE. Advances the wqe pointer to point at the
+ * scatter-gather list, and set wqe_end to the end of the WQE.
+ */
+static int mlx5_ib_mr_initiator_pfault_handler(
+	struct mlx5_ib_dev *dev, struct mlx5_pagefault *pfault,
+	struct mlx5_ib_qp *qp, void **wqe, void **wqe_end, int wqe_length)
+{
+	struct mlx5_wqe_ctrl_seg *ctrl = *wqe;
+	u16 wqe_index = pfault->wqe.wqe_index;
+	u32 transport_caps;
+	struct mlx5_base_av *av;
+	unsigned ds, opcode;
+#if defined(DEBUG)
+	u32 ctrl_wqe_index, ctrl_qpn;
+#endif
+	u32 qpn = qp->trans_qp.base.mqp.qpn;
+
+	ds = be32_to_cpu(ctrl->qpn_ds) & MLX5_WQE_CTRL_DS_MASK;
+	if (ds * MLX5_WQE_DS_UNITS > wqe_length) {
+		mlx5_ib_err(dev, "Unable to read the complete WQE. ds = 0x%x, ret = 0x%x\n",
+			    ds, wqe_length);
+		return -EFAULT;
+	}
+
+	if (ds == 0) {
+		mlx5_ib_err(dev, "Got WQE with zero DS. wqe_index=%x, qpn=%x\n",
+			    wqe_index, qpn);
+		return -EFAULT;
+	}
+
+#if defined(DEBUG)
+	ctrl_wqe_index = (be32_to_cpu(ctrl->opmod_idx_opcode) &
+			MLX5_WQE_CTRL_WQE_INDEX_MASK) >>
+			MLX5_WQE_CTRL_WQE_INDEX_SHIFT;
+	if (wqe_index != ctrl_wqe_index) {
+		mlx5_ib_err(dev, "Got WQE with invalid wqe_index. wqe_index=0x%x, qpn=0x%x ctrl->wqe_index=0x%x\n",
+			    wqe_index, qpn,
+			    ctrl_wqe_index);
+		return -EFAULT;
+	}
+
+	ctrl_qpn = (be32_to_cpu(ctrl->qpn_ds) & MLX5_WQE_CTRL_QPN_MASK) >>
+		MLX5_WQE_CTRL_QPN_SHIFT;
+	if (qpn != ctrl_qpn) {
+		mlx5_ib_err(dev, "Got WQE with incorrect QP number. wqe_index=0x%x, qpn=0x%x ctrl->qpn=0x%x\n",
+			    wqe_index, qpn,
+			    ctrl_qpn);
+		return -EFAULT;
+	}
+#endif /* DEBUG */
+
+	*wqe_end = *wqe + ds * MLX5_WQE_DS_UNITS;
+	*wqe += sizeof(*ctrl);
+
+	opcode = be32_to_cpu(ctrl->opmod_idx_opcode) &
+		 MLX5_WQE_CTRL_OPCODE_MASK;
+
+	switch (qp->ibqp.qp_type) {
+	case IB_QPT_RC:
+		transport_caps = dev->odp_caps.per_transport_caps.rc_odp_caps;
+		break;
+	case IB_QPT_UD:
+		transport_caps = dev->odp_caps.per_transport_caps.ud_odp_caps;
+		break;
+	default:
+		mlx5_ib_err(dev, "ODP fault on QP of an unsupported transport 0x%x\n",
+			    qp->ibqp.qp_type);
+		return -EFAULT;
+	}
+
+	if (unlikely(opcode >= ARRAY_SIZE(mlx5_ib_odp_opcode_cap) ||
+		     !(transport_caps & mlx5_ib_odp_opcode_cap[opcode]))) {
+		mlx5_ib_err(dev, "ODP fault on QP of an unsupported opcode 0x%x\n",
+			    opcode);
+		return -EFAULT;
+	}
+
+	if (qp->ibqp.qp_type != IB_QPT_RC) {
+		av = *wqe;
+		if (av->dqp_dct & cpu_to_be32(MLX5_EXTENDED_UD_AV))
+			*wqe += sizeof(struct mlx5_av);
+		else
+			*wqe += sizeof(struct mlx5_base_av);
+	}
+
+	switch (opcode) {
+	case MLX5_OPCODE_RDMA_WRITE:
+	case MLX5_OPCODE_RDMA_WRITE_IMM:
+	case MLX5_OPCODE_RDMA_READ:
+		*wqe += sizeof(struct mlx5_wqe_raddr_seg);
+		break;
+	case MLX5_OPCODE_ATOMIC_CS:
+	case MLX5_OPCODE_ATOMIC_FA:
+		*wqe += sizeof(struct mlx5_wqe_raddr_seg);
+		*wqe += sizeof(struct mlx5_wqe_atomic_seg);
+		break;
+	}
+
+	return 0;
+}
+
+/*
+ * Parse responder WQE. Advances the wqe pointer to point at the
+ * scatter-gather list, and set wqe_end to the end of the WQE.
+ */
+static int mlx5_ib_mr_responder_pfault_handler(
+	struct mlx5_ib_dev *dev, struct mlx5_pagefault *pfault,
+	struct mlx5_ib_qp *qp, void **wqe, void **wqe_end, int wqe_length)
+{
+	struct mlx5_ib_wq *wq = &qp->rq;
+	int wqe_size = 1 << wq->wqe_shift;
+
+	if (qp->ibqp.srq) {
+		mlx5_ib_err(dev, "ODP fault on SRQ is not supported\n");
+		return -EFAULT;
+	}
+
+	if (qp->wq_sig) {
+		mlx5_ib_err(dev, "ODP fault with WQE signatures is not supported\n");
+		return -EFAULT;
+	}
+
+	if (wqe_size > wqe_length) {
+		mlx5_ib_err(dev, "Couldn't read all of the receive WQE's content\n");
+		return -EFAULT;
+	}
+
+	switch (qp->ibqp.qp_type) {
+	case IB_QPT_RC:
+		if (!(dev->odp_caps.per_transport_caps.rc_odp_caps &
+		      IB_ODP_SUPPORT_RECV))
+			goto invalid_transport_or_opcode;
+		break;
+	default:
+invalid_transport_or_opcode:
+		mlx5_ib_err(dev, "ODP fault on QP of an unsupported transport. transport: 0x%x\n",
+			    qp->ibqp.qp_type);
+		return -EFAULT;
+	}
+
+	*wqe_end = *wqe + wqe_size;
+
+	return 0;
+}
+
+static struct mlx5_ib_qp *mlx5_ib_odp_find_qp(struct mlx5_ib_dev *dev,
+					      u32 wq_num)
+{
+	struct mlx5_core_qp *mqp = __mlx5_qp_lookup(dev->mdev, wq_num);
+
+	if (!mqp) {
+		mlx5_ib_err(dev, "QPN 0x%6x not found\n", wq_num);
+		return NULL;
+	}
+
+	return to_mibqp(mqp);
+}
+
+static void mlx5_ib_mr_wqe_pfault_handler(struct mlx5_ib_dev *dev,
+					  struct mlx5_pagefault *pfault)
+{
+	int ret;
+	void *wqe, *wqe_end;
+	u32 bytes_mapped, total_wqe_bytes;
+	char *buffer = NULL;
+	int resume_with_error = 1;
+	u16 wqe_index = pfault->wqe.wqe_index;
+	int requestor = pfault->type & MLX5_PFAULT_REQUESTOR;
+	struct mlx5_ib_qp *qp;
+
+	buffer = (char *)__get_free_page(GFP_KERNEL);
+	if (!buffer) {
+		mlx5_ib_err(dev, "Error allocating memory for IO page fault handling.\n");
+		goto resolve_page_fault;
+	}
+
+	qp = mlx5_ib_odp_find_qp(dev, pfault->wqe.wq_num);
+	if (!qp)
+		goto resolve_page_fault;
+
+	ret = mlx5_ib_read_user_wqe(qp, requestor, wqe_index, buffer,
+				    PAGE_SIZE, &qp->trans_qp.base);
+	if (ret < 0) {
+		mlx5_ib_err(dev, "Failed reading a WQE following page fault, error=%d, wqe_index=%x, qpn=%x\n",
+			    ret, wqe_index, pfault->token);
+		goto resolve_page_fault;
+	}
+
+	wqe = buffer;
+	if (requestor)
+		ret = mlx5_ib_mr_initiator_pfault_handler(dev, pfault, qp, &wqe,
+							  &wqe_end, ret);
+	else
+		ret = mlx5_ib_mr_responder_pfault_handler(dev, pfault, qp, &wqe,
+							  &wqe_end, ret);
+	if (ret < 0)
+		goto resolve_page_fault;
+
+	if (wqe >= wqe_end) {
+		mlx5_ib_err(dev, "ODP fault on invalid WQE.\n");
+		goto resolve_page_fault;
+	}
+
+	ret = pagefault_data_segments(dev, pfault, qp, wqe, wqe_end,
+				      &bytes_mapped, &total_wqe_bytes,
+				      !requestor);
+	if (ret == -EAGAIN) {
+		resume_with_error = 0;
+		goto resolve_page_fault;
+	} else if (ret < 0 || total_wqe_bytes > bytes_mapped) {
+		goto resolve_page_fault;
+	}
+
+	resume_with_error = 0;
+resolve_page_fault:
+	mlx5_ib_page_fault_resume(dev, pfault, resume_with_error);
+	mlx5_ib_dbg(dev, "PAGE FAULT completed. QP 0x%x resume_with_error=%d, type: 0x%x\n",
+		    pfault->wqe.wq_num, resume_with_error,
+		    pfault->type);
+	free_page((unsigned long)buffer);
+}
+
+static int pages_in_range(u64 address, u32 length)
+{
+	return (ALIGN(address + length, PAGE_SIZE) -
+		(address & PAGE_MASK)) >> PAGE_SHIFT;
+}
+
+static void mlx5_ib_mr_rdma_pfault_handler(struct mlx5_ib_dev *dev,
+					   struct mlx5_pagefault *pfault)
+{
+	u64 address;
+	u32 length;
+	u32 prefetch_len = pfault->bytes_committed;
+	int prefetch_activated = 0;
+	u32 rkey = pfault->rdma.r_key;
+	int ret;
+
+	/* The RDMA responder handler handles the page fault in two parts.
+	 * First it brings the necessary pages for the current packet
+	 * (and uses the pfault context), and then (after resuming the QP)
+	 * prefetches more pages. The second operation cannot use the pfault
+	 * context and therefore uses the dummy_pfault context allocated on
+	 * the stack */
+	pfault->rdma.rdma_va += pfault->bytes_committed;
+	pfault->rdma.rdma_op_len -= min(pfault->bytes_committed,
+					 pfault->rdma.rdma_op_len);
+	pfault->bytes_committed = 0;
+
+	address = pfault->rdma.rdma_va;
+	length  = pfault->rdma.rdma_op_len;
+
+	/* For some operations, the hardware cannot tell the exact message
+	 * length, and in those cases it reports zero. Use prefetch
+	 * logic. */
+	if (length == 0) {
+		prefetch_activated = 1;
+		length = pfault->rdma.packet_size;
+		prefetch_len = min(MAX_PREFETCH_LEN, prefetch_len);
+	}
+
+	ret = pagefault_single_data_segment(dev, rkey, address, length,
+					    &pfault->bytes_committed, NULL);
+	if (ret == -EAGAIN) {
+		/* We're racing with an invalidation, don't prefetch */
+		prefetch_activated = 0;
+	} else if (ret < 0 || pages_in_range(address, length) > ret) {
+		mlx5_ib_page_fault_resume(dev, pfault, 1);
+		if (ret != -ENOENT)
+			mlx5_ib_dbg(dev, "PAGE FAULT error %d. QP 0x%x, type: 0x%x\n",
+				    ret, pfault->token, pfault->type);
+		return;
+	}
+
+	mlx5_ib_page_fault_resume(dev, pfault, 0);
+	mlx5_ib_dbg(dev, "PAGE FAULT completed. QP 0x%x, type: 0x%x, prefetch_activated: %d\n",
+		    pfault->token, pfault->type,
+		    prefetch_activated);
+
+	/* At this point, there might be a new pagefault already arriving in
+	 * the eq, switch to the dummy pagefault for the rest of the
+	 * processing. We're still OK with the objects being alive as the
+	 * work-queue is being fenced. */
+
+	if (prefetch_activated) {
+		u32 bytes_committed = 0;
+
+		ret = pagefault_single_data_segment(dev, rkey, address,
+						    prefetch_len,
+						    &bytes_committed, NULL);
+		if (ret < 0 && ret != -EAGAIN) {
+			mlx5_ib_dbg(dev, "Prefetch failed. ret: %d, QP 0x%x, address: 0x%.16llx, length = 0x%.16x\n",
+				    ret, pfault->token, address, prefetch_len);
+		}
+	}
+}
+
+void mlx5_ib_pfault(struct mlx5_core_dev *mdev, void *context,
+		    struct mlx5_pagefault *pfault)
+{
+	struct mlx5_ib_dev *dev = context;
+	u8 event_subtype = pfault->event_subtype;
+
+	switch (event_subtype) {
+	case MLX5_PFAULT_SUBTYPE_WQE:
+		mlx5_ib_mr_wqe_pfault_handler(dev, pfault);
+		break;
+	case MLX5_PFAULT_SUBTYPE_RDMA:
+		mlx5_ib_mr_rdma_pfault_handler(dev, pfault);
+		break;
+	default:
+		mlx5_ib_err(dev, "Invalid page fault event subtype: 0x%x\n",
+			    event_subtype);
+		mlx5_ib_page_fault_resume(dev, pfault, 1);
+	}
+}
+
+void mlx5_odp_init_mr_cache_entry(struct mlx5_cache_ent *ent)
+{
+	if (!(ent->dev->odp_caps.general_caps & IB_ODP_SUPPORT_IMPLICIT))
+		return;
+
+	switch (ent->order - 2) {
+	case MLX5_IMR_MTT_CACHE_ENTRY:
+		ent->page = PAGE_SHIFT;
+		ent->xlt = MLX5_IMR_MTT_ENTRIES *
+			   sizeof(struct mlx5_mtt) /
+			   MLX5_IB_UMR_OCTOWORD;
+		ent->access_mode = MLX5_MKC_ACCESS_MODE_MTT;
+		ent->limit = 0;
+		break;
+
+	case MLX5_IMR_KSM_CACHE_ENTRY:
+		ent->page = MLX5_KSM_PAGE_SHIFT;
+		ent->xlt = mlx5_imr_ksm_entries *
+			   sizeof(struct mlx5_klm) /
+			   MLX5_IB_UMR_OCTOWORD;
+		ent->access_mode = MLX5_MKC_ACCESS_MODE_KSM;
+		ent->limit = 0;
+		break;
+	}
+}
+
+int mlx5_ib_odp_init_one(struct mlx5_ib_dev *dev)
+{
+	int ret;
+
+	if (dev->odp_caps.general_caps & IB_ODP_SUPPORT_IMPLICIT) {
+		ret = mlx5_cmd_null_mkey(dev->mdev, &dev->null_mkey);
+		if (ret) {
+			mlx5_ib_err(dev, "Error getting null_mkey %d\n", ret);
+			return ret;
+		}
+	}
+
+	return 0;
+}
+
+int mlx5_ib_odp_init(void)
+{
+	mlx5_imr_ksm_entries = BIT_ULL(get_order(TASK_SIZE) -
+				       MLX5_IMR_MTT_BITS);
+
+	return 0;
+}
+
diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c
new file mode 100644
index 000000000..361b1b859
--- /dev/null
+++ b/drivers/infiniband/hw/mlx5/qp.c
@@ -0,0 +1,5891 @@
+/*
+ * Copyright (c) 2013-2015, Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/module.h>
+#include <rdma/ib_umem.h>
+#include <rdma/ib_cache.h>
+#include <rdma/ib_user_verbs.h>
+#include <linux/mlx5/fs.h>
+#include "mlx5_ib.h"
+#include "ib_rep.h"
+
+/* not supported currently */
+static int wq_signature;
+
+enum {
+	MLX5_IB_ACK_REQ_FREQ	= 8,
+};
+
+enum {
+	MLX5_IB_DEFAULT_SCHED_QUEUE	= 0x83,
+	MLX5_IB_DEFAULT_QP0_SCHED_QUEUE	= 0x3f,
+	MLX5_IB_LINK_TYPE_IB		= 0,
+	MLX5_IB_LINK_TYPE_ETH		= 1
+};
+
+enum {
+	MLX5_IB_SQ_STRIDE	= 6,
+	MLX5_IB_SQ_UMR_INLINE_THRESHOLD = 64,
+};
+
+static const u32 mlx5_ib_opcode[] = {
+	[IB_WR_SEND]				= MLX5_OPCODE_SEND,
+	[IB_WR_LSO]				= MLX5_OPCODE_LSO,
+	[IB_WR_SEND_WITH_IMM]			= MLX5_OPCODE_SEND_IMM,
+	[IB_WR_RDMA_WRITE]			= MLX5_OPCODE_RDMA_WRITE,
+	[IB_WR_RDMA_WRITE_WITH_IMM]		= MLX5_OPCODE_RDMA_WRITE_IMM,
+	[IB_WR_RDMA_READ]			= MLX5_OPCODE_RDMA_READ,
+	[IB_WR_ATOMIC_CMP_AND_SWP]		= MLX5_OPCODE_ATOMIC_CS,
+	[IB_WR_ATOMIC_FETCH_AND_ADD]		= MLX5_OPCODE_ATOMIC_FA,
+	[IB_WR_SEND_WITH_INV]			= MLX5_OPCODE_SEND_INVAL,
+	[IB_WR_LOCAL_INV]			= MLX5_OPCODE_UMR,
+	[IB_WR_REG_MR]				= MLX5_OPCODE_UMR,
+	[IB_WR_MASKED_ATOMIC_CMP_AND_SWP]	= MLX5_OPCODE_ATOMIC_MASKED_CS,
+	[IB_WR_MASKED_ATOMIC_FETCH_AND_ADD]	= MLX5_OPCODE_ATOMIC_MASKED_FA,
+	[MLX5_IB_WR_UMR]			= MLX5_OPCODE_UMR,
+};
+
+struct mlx5_wqe_eth_pad {
+	u8 rsvd0[16];
+};
+
+enum raw_qp_set_mask_map {
+	MLX5_RAW_QP_MOD_SET_RQ_Q_CTR_ID		= 1UL << 0,
+	MLX5_RAW_QP_RATE_LIMIT			= 1UL << 1,
+};
+
+struct mlx5_modify_raw_qp_param {
+	u16 operation;
+
+	u32 set_mask; /* raw_qp_set_mask_map */
+
+	struct mlx5_rate_limit rl;
+
+	u8 rq_q_ctr_id;
+};
+
+static void get_cqs(enum ib_qp_type qp_type,
+		    struct ib_cq *ib_send_cq, struct ib_cq *ib_recv_cq,
+		    struct mlx5_ib_cq **send_cq, struct mlx5_ib_cq **recv_cq);
+
+static int is_qp0(enum ib_qp_type qp_type)
+{
+	return qp_type == IB_QPT_SMI;
+}
+
+static int is_sqp(enum ib_qp_type qp_type)
+{
+	return is_qp0(qp_type) || is_qp1(qp_type);
+}
+
+static void *get_wqe(struct mlx5_ib_qp *qp, int offset)
+{
+	return mlx5_buf_offset(&qp->buf, offset);
+}
+
+static void *get_recv_wqe(struct mlx5_ib_qp *qp, int n)
+{
+	return get_wqe(qp, qp->rq.offset + (n << qp->rq.wqe_shift));
+}
+
+void *mlx5_get_send_wqe(struct mlx5_ib_qp *qp, int n)
+{
+	return get_wqe(qp, qp->sq.offset + (n << MLX5_IB_SQ_STRIDE));
+}
+
+/**
+ * mlx5_ib_read_user_wqe() - Copy a user-space WQE to kernel space.
+ *
+ * @qp: QP to copy from.
+ * @send: copy from the send queue when non-zero, use the receive queue
+ *	  otherwise.
+ * @wqe_index:  index to start copying from. For send work queues, the
+ *		wqe_index is in units of MLX5_SEND_WQE_BB.
+ *		For receive work queue, it is the number of work queue
+ *		element in the queue.
+ * @buffer: destination buffer.
+ * @length: maximum number of bytes to copy.
+ *
+ * Copies at least a single WQE, but may copy more data.
+ *
+ * Return: the number of bytes copied, or an error code.
+ */
+int mlx5_ib_read_user_wqe(struct mlx5_ib_qp *qp, int send, int wqe_index,
+			  void *buffer, u32 length,
+			  struct mlx5_ib_qp_base *base)
+{
+	struct ib_device *ibdev = qp->ibqp.device;
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+	struct mlx5_ib_wq *wq = send ? &qp->sq : &qp->rq;
+	size_t offset;
+	size_t wq_end;
+	struct ib_umem *umem = base->ubuffer.umem;
+	u32 first_copy_length;
+	int wqe_length;
+	int ret;
+
+	if (wq->wqe_cnt == 0) {
+		mlx5_ib_dbg(dev, "mlx5_ib_read_user_wqe for a QP with wqe_cnt == 0. qp_type: 0x%x\n",
+			    qp->ibqp.qp_type);
+		return -EINVAL;
+	}
+
+	offset = wq->offset + ((wqe_index % wq->wqe_cnt) << wq->wqe_shift);
+	wq_end = wq->offset + (wq->wqe_cnt << wq->wqe_shift);
+
+	if (send && length < sizeof(struct mlx5_wqe_ctrl_seg))
+		return -EINVAL;
+
+	if (offset > umem->length ||
+	    (send && offset + sizeof(struct mlx5_wqe_ctrl_seg) > umem->length))
+		return -EINVAL;
+
+	first_copy_length = min_t(u32, offset + length, wq_end) - offset;
+	ret = ib_umem_copy_from(buffer, umem, offset, first_copy_length);
+	if (ret)
+		return ret;
+
+	if (send) {
+		struct mlx5_wqe_ctrl_seg *ctrl = buffer;
+		int ds = be32_to_cpu(ctrl->qpn_ds) & MLX5_WQE_CTRL_DS_MASK;
+
+		wqe_length = ds * MLX5_WQE_DS_UNITS;
+	} else {
+		wqe_length = 1 << wq->wqe_shift;
+	}
+
+	if (wqe_length <= first_copy_length)
+		return first_copy_length;
+
+	ret = ib_umem_copy_from(buffer + first_copy_length, umem, wq->offset,
+				wqe_length - first_copy_length);
+	if (ret)
+		return ret;
+
+	return wqe_length;
+}
+
+static void mlx5_ib_qp_event(struct mlx5_core_qp *qp, int type)
+{
+	struct ib_qp *ibqp = &to_mibqp(qp)->ibqp;
+	struct ib_event event;
+
+	if (type == MLX5_EVENT_TYPE_PATH_MIG) {
+		/* This event is only valid for trans_qps */
+		to_mibqp(qp)->port = to_mibqp(qp)->trans_qp.alt_port;
+	}
+
+	if (ibqp->event_handler) {
+		event.device     = ibqp->device;
+		event.element.qp = ibqp;
+		switch (type) {
+		case MLX5_EVENT_TYPE_PATH_MIG:
+			event.event = IB_EVENT_PATH_MIG;
+			break;
+		case MLX5_EVENT_TYPE_COMM_EST:
+			event.event = IB_EVENT_COMM_EST;
+			break;
+		case MLX5_EVENT_TYPE_SQ_DRAINED:
+			event.event = IB_EVENT_SQ_DRAINED;
+			break;
+		case MLX5_EVENT_TYPE_SRQ_LAST_WQE:
+			event.event = IB_EVENT_QP_LAST_WQE_REACHED;
+			break;
+		case MLX5_EVENT_TYPE_WQ_CATAS_ERROR:
+			event.event = IB_EVENT_QP_FATAL;
+			break;
+		case MLX5_EVENT_TYPE_PATH_MIG_FAILED:
+			event.event = IB_EVENT_PATH_MIG_ERR;
+			break;
+		case MLX5_EVENT_TYPE_WQ_INVAL_REQ_ERROR:
+			event.event = IB_EVENT_QP_REQ_ERR;
+			break;
+		case MLX5_EVENT_TYPE_WQ_ACCESS_ERROR:
+			event.event = IB_EVENT_QP_ACCESS_ERR;
+			break;
+		default:
+			pr_warn("mlx5_ib: Unexpected event type %d on QP %06x\n", type, qp->qpn);
+			return;
+		}
+
+		ibqp->event_handler(&event, ibqp->qp_context);
+	}
+}
+
+static int set_rq_size(struct mlx5_ib_dev *dev, struct ib_qp_cap *cap,
+		       int has_rq, struct mlx5_ib_qp *qp, struct mlx5_ib_create_qp *ucmd)
+{
+	int wqe_size;
+	int wq_size;
+
+	/* Sanity check RQ size before proceeding */
+	if (cap->max_recv_wr > (1 << MLX5_CAP_GEN(dev->mdev, log_max_qp_sz)))
+		return -EINVAL;
+
+	if (!has_rq) {
+		qp->rq.max_gs = 0;
+		qp->rq.wqe_cnt = 0;
+		qp->rq.wqe_shift = 0;
+		cap->max_recv_wr = 0;
+		cap->max_recv_sge = 0;
+	} else {
+		if (ucmd) {
+			qp->rq.wqe_cnt = ucmd->rq_wqe_count;
+			if (ucmd->rq_wqe_shift > BITS_PER_BYTE * sizeof(ucmd->rq_wqe_shift))
+				return -EINVAL;
+			qp->rq.wqe_shift = ucmd->rq_wqe_shift;
+			if ((1 << qp->rq.wqe_shift) / sizeof(struct mlx5_wqe_data_seg) < qp->wq_sig)
+				return -EINVAL;
+			qp->rq.max_gs = (1 << qp->rq.wqe_shift) / sizeof(struct mlx5_wqe_data_seg) - qp->wq_sig;
+			qp->rq.max_post = qp->rq.wqe_cnt;
+		} else {
+			wqe_size = qp->wq_sig ? sizeof(struct mlx5_wqe_signature_seg) : 0;
+			wqe_size += cap->max_recv_sge * sizeof(struct mlx5_wqe_data_seg);
+			wqe_size = roundup_pow_of_two(wqe_size);
+			wq_size = roundup_pow_of_two(cap->max_recv_wr) * wqe_size;
+			wq_size = max_t(int, wq_size, MLX5_SEND_WQE_BB);
+			qp->rq.wqe_cnt = wq_size / wqe_size;
+			if (wqe_size > MLX5_CAP_GEN(dev->mdev, max_wqe_sz_rq)) {
+				mlx5_ib_dbg(dev, "wqe_size %d, max %d\n",
+					    wqe_size,
+					    MLX5_CAP_GEN(dev->mdev,
+							 max_wqe_sz_rq));
+				return -EINVAL;
+			}
+			qp->rq.wqe_shift = ilog2(wqe_size);
+			qp->rq.max_gs = (1 << qp->rq.wqe_shift) / sizeof(struct mlx5_wqe_data_seg) - qp->wq_sig;
+			qp->rq.max_post = qp->rq.wqe_cnt;
+		}
+	}
+
+	return 0;
+}
+
+static int sq_overhead(struct ib_qp_init_attr *attr)
+{
+	int size = 0;
+
+	switch (attr->qp_type) {
+	case IB_QPT_XRC_INI:
+		size += sizeof(struct mlx5_wqe_xrc_seg);
+		/* fall through */
+	case IB_QPT_RC:
+		size += sizeof(struct mlx5_wqe_ctrl_seg) +
+			max(sizeof(struct mlx5_wqe_atomic_seg) +
+			    sizeof(struct mlx5_wqe_raddr_seg),
+			    sizeof(struct mlx5_wqe_umr_ctrl_seg) +
+			    sizeof(struct mlx5_mkey_seg) +
+			    MLX5_IB_SQ_UMR_INLINE_THRESHOLD /
+			    MLX5_IB_UMR_OCTOWORD);
+		break;
+
+	case IB_QPT_XRC_TGT:
+		return 0;
+
+	case IB_QPT_UC:
+		size += sizeof(struct mlx5_wqe_ctrl_seg) +
+			max(sizeof(struct mlx5_wqe_raddr_seg),
+			    sizeof(struct mlx5_wqe_umr_ctrl_seg) +
+			    sizeof(struct mlx5_mkey_seg));
+		break;
+
+	case IB_QPT_UD:
+		if (attr->create_flags & IB_QP_CREATE_IPOIB_UD_LSO)
+			size += sizeof(struct mlx5_wqe_eth_pad) +
+				sizeof(struct mlx5_wqe_eth_seg);
+		/* fall through */
+	case IB_QPT_SMI:
+	case MLX5_IB_QPT_HW_GSI:
+		size += sizeof(struct mlx5_wqe_ctrl_seg) +
+			sizeof(struct mlx5_wqe_datagram_seg);
+		break;
+
+	case MLX5_IB_QPT_REG_UMR:
+		size += sizeof(struct mlx5_wqe_ctrl_seg) +
+			sizeof(struct mlx5_wqe_umr_ctrl_seg) +
+			sizeof(struct mlx5_mkey_seg);
+		break;
+
+	default:
+		return -EINVAL;
+	}
+
+	return size;
+}
+
+static int calc_send_wqe(struct ib_qp_init_attr *attr)
+{
+	int inl_size = 0;
+	int size;
+
+	size = sq_overhead(attr);
+	if (size < 0)
+		return size;
+
+	if (attr->cap.max_inline_data) {
+		inl_size = size + sizeof(struct mlx5_wqe_inline_seg) +
+			attr->cap.max_inline_data;
+	}
+
+	size += attr->cap.max_send_sge * sizeof(struct mlx5_wqe_data_seg);
+	if (attr->create_flags & IB_QP_CREATE_SIGNATURE_EN &&
+	    ALIGN(max_t(int, inl_size, size), MLX5_SEND_WQE_BB) < MLX5_SIG_WQE_SIZE)
+			return MLX5_SIG_WQE_SIZE;
+	else
+		return ALIGN(max_t(int, inl_size, size), MLX5_SEND_WQE_BB);
+}
+
+static int get_send_sge(struct ib_qp_init_attr *attr, int wqe_size)
+{
+	int max_sge;
+
+	if (attr->qp_type == IB_QPT_RC)
+		max_sge = (min_t(int, wqe_size, 512) -
+			   sizeof(struct mlx5_wqe_ctrl_seg) -
+			   sizeof(struct mlx5_wqe_raddr_seg)) /
+			sizeof(struct mlx5_wqe_data_seg);
+	else if (attr->qp_type == IB_QPT_XRC_INI)
+		max_sge = (min_t(int, wqe_size, 512) -
+			   sizeof(struct mlx5_wqe_ctrl_seg) -
+			   sizeof(struct mlx5_wqe_xrc_seg) -
+			   sizeof(struct mlx5_wqe_raddr_seg)) /
+			sizeof(struct mlx5_wqe_data_seg);
+	else
+		max_sge = (wqe_size - sq_overhead(attr)) /
+			sizeof(struct mlx5_wqe_data_seg);
+
+	return min_t(int, max_sge, wqe_size - sq_overhead(attr) /
+		     sizeof(struct mlx5_wqe_data_seg));
+}
+
+static int calc_sq_size(struct mlx5_ib_dev *dev, struct ib_qp_init_attr *attr,
+			struct mlx5_ib_qp *qp)
+{
+	int wqe_size;
+	int wq_size;
+
+	if (!attr->cap.max_send_wr)
+		return 0;
+
+	wqe_size = calc_send_wqe(attr);
+	mlx5_ib_dbg(dev, "wqe_size %d\n", wqe_size);
+	if (wqe_size < 0)
+		return wqe_size;
+
+	if (wqe_size > MLX5_CAP_GEN(dev->mdev, max_wqe_sz_sq)) {
+		mlx5_ib_dbg(dev, "wqe_size(%d) > max_sq_desc_sz(%d)\n",
+			    wqe_size, MLX5_CAP_GEN(dev->mdev, max_wqe_sz_sq));
+		return -EINVAL;
+	}
+
+	qp->max_inline_data = wqe_size - sq_overhead(attr) -
+			      sizeof(struct mlx5_wqe_inline_seg);
+	attr->cap.max_inline_data = qp->max_inline_data;
+
+	if (attr->create_flags & IB_QP_CREATE_SIGNATURE_EN)
+		qp->signature_en = true;
+
+	wq_size = roundup_pow_of_two(attr->cap.max_send_wr * wqe_size);
+	qp->sq.wqe_cnt = wq_size / MLX5_SEND_WQE_BB;
+	if (qp->sq.wqe_cnt > (1 << MLX5_CAP_GEN(dev->mdev, log_max_qp_sz))) {
+		mlx5_ib_dbg(dev, "send queue size (%d * %d / %d -> %d) exceeds limits(%d)\n",
+			    attr->cap.max_send_wr, wqe_size, MLX5_SEND_WQE_BB,
+			    qp->sq.wqe_cnt,
+			    1 << MLX5_CAP_GEN(dev->mdev, log_max_qp_sz));
+		return -ENOMEM;
+	}
+	qp->sq.wqe_shift = ilog2(MLX5_SEND_WQE_BB);
+	qp->sq.max_gs = get_send_sge(attr, wqe_size);
+	if (qp->sq.max_gs < attr->cap.max_send_sge)
+		return -ENOMEM;
+
+	attr->cap.max_send_sge = qp->sq.max_gs;
+	qp->sq.max_post = wq_size / wqe_size;
+	attr->cap.max_send_wr = qp->sq.max_post;
+
+	return wq_size;
+}
+
+static int set_user_buf_size(struct mlx5_ib_dev *dev,
+			    struct mlx5_ib_qp *qp,
+			    struct mlx5_ib_create_qp *ucmd,
+			    struct mlx5_ib_qp_base *base,
+			    struct ib_qp_init_attr *attr)
+{
+	int desc_sz = 1 << qp->sq.wqe_shift;
+
+	if (desc_sz > MLX5_CAP_GEN(dev->mdev, max_wqe_sz_sq)) {
+		mlx5_ib_warn(dev, "desc_sz %d, max_sq_desc_sz %d\n",
+			     desc_sz, MLX5_CAP_GEN(dev->mdev, max_wqe_sz_sq));
+		return -EINVAL;
+	}
+
+	if (ucmd->sq_wqe_count && ((1 << ilog2(ucmd->sq_wqe_count)) != ucmd->sq_wqe_count)) {
+		mlx5_ib_warn(dev, "sq_wqe_count %d, sq_wqe_count %d\n",
+			     ucmd->sq_wqe_count, ucmd->sq_wqe_count);
+		return -EINVAL;
+	}
+
+	qp->sq.wqe_cnt = ucmd->sq_wqe_count;
+
+	if (qp->sq.wqe_cnt > (1 << MLX5_CAP_GEN(dev->mdev, log_max_qp_sz))) {
+		mlx5_ib_warn(dev, "wqe_cnt %d, max_wqes %d\n",
+			     qp->sq.wqe_cnt,
+			     1 << MLX5_CAP_GEN(dev->mdev, log_max_qp_sz));
+		return -EINVAL;
+	}
+
+	if (attr->qp_type == IB_QPT_RAW_PACKET ||
+	    qp->flags & MLX5_IB_QP_UNDERLAY) {
+		base->ubuffer.buf_size = qp->rq.wqe_cnt << qp->rq.wqe_shift;
+		qp->raw_packet_qp.sq.ubuffer.buf_size = qp->sq.wqe_cnt << 6;
+	} else {
+		base->ubuffer.buf_size = (qp->rq.wqe_cnt << qp->rq.wqe_shift) +
+					 (qp->sq.wqe_cnt << 6);
+	}
+
+	return 0;
+}
+
+static int qp_has_rq(struct ib_qp_init_attr *attr)
+{
+	if (attr->qp_type == IB_QPT_XRC_INI ||
+	    attr->qp_type == IB_QPT_XRC_TGT || attr->srq ||
+	    attr->qp_type == MLX5_IB_QPT_REG_UMR ||
+	    !attr->cap.max_recv_wr)
+		return 0;
+
+	return 1;
+}
+
+enum {
+	/* this is the first blue flame register in the array of bfregs assigned
+	 * to a processes. Since we do not use it for blue flame but rather
+	 * regular 64 bit doorbells, we do not need a lock for maintaiing
+	 * "odd/even" order
+	 */
+	NUM_NON_BLUE_FLAME_BFREGS = 1,
+};
+
+static int max_bfregs(struct mlx5_ib_dev *dev, struct mlx5_bfreg_info *bfregi)
+{
+	return get_num_static_uars(dev, bfregi) * MLX5_NON_FP_BFREGS_PER_UAR;
+}
+
+static int num_med_bfreg(struct mlx5_ib_dev *dev,
+			 struct mlx5_bfreg_info *bfregi)
+{
+	int n;
+
+	n = max_bfregs(dev, bfregi) - bfregi->num_low_latency_bfregs -
+	    NUM_NON_BLUE_FLAME_BFREGS;
+
+	return n >= 0 ? n : 0;
+}
+
+static int first_med_bfreg(struct mlx5_ib_dev *dev,
+			   struct mlx5_bfreg_info *bfregi)
+{
+	return num_med_bfreg(dev, bfregi) ? 1 : -ENOMEM;
+}
+
+static int first_hi_bfreg(struct mlx5_ib_dev *dev,
+			  struct mlx5_bfreg_info *bfregi)
+{
+	int med;
+
+	med = num_med_bfreg(dev, bfregi);
+	return ++med;
+}
+
+static int alloc_high_class_bfreg(struct mlx5_ib_dev *dev,
+				  struct mlx5_bfreg_info *bfregi)
+{
+	int i;
+
+	for (i = first_hi_bfreg(dev, bfregi); i < max_bfregs(dev, bfregi); i++) {
+		if (!bfregi->count[i]) {
+			bfregi->count[i]++;
+			return i;
+		}
+	}
+
+	return -ENOMEM;
+}
+
+static int alloc_med_class_bfreg(struct mlx5_ib_dev *dev,
+				 struct mlx5_bfreg_info *bfregi)
+{
+	int minidx = first_med_bfreg(dev, bfregi);
+	int i;
+
+	if (minidx < 0)
+		return minidx;
+
+	for (i = minidx; i < first_hi_bfreg(dev, bfregi); i++) {
+		if (bfregi->count[i] < bfregi->count[minidx])
+			minidx = i;
+		if (!bfregi->count[minidx])
+			break;
+	}
+
+	bfregi->count[minidx]++;
+	return minidx;
+}
+
+static int alloc_bfreg(struct mlx5_ib_dev *dev,
+		       struct mlx5_bfreg_info *bfregi)
+{
+	int bfregn = -ENOMEM;
+
+	mutex_lock(&bfregi->lock);
+	if (bfregi->ver >= 2) {
+		bfregn = alloc_high_class_bfreg(dev, bfregi);
+		if (bfregn < 0)
+			bfregn = alloc_med_class_bfreg(dev, bfregi);
+	}
+
+	if (bfregn < 0) {
+		BUILD_BUG_ON(NUM_NON_BLUE_FLAME_BFREGS != 1);
+		bfregn = 0;
+		bfregi->count[bfregn]++;
+	}
+	mutex_unlock(&bfregi->lock);
+
+	return bfregn;
+}
+
+void mlx5_ib_free_bfreg(struct mlx5_ib_dev *dev, struct mlx5_bfreg_info *bfregi, int bfregn)
+{
+	mutex_lock(&bfregi->lock);
+	bfregi->count[bfregn]--;
+	mutex_unlock(&bfregi->lock);
+}
+
+static enum mlx5_qp_state to_mlx5_state(enum ib_qp_state state)
+{
+	switch (state) {
+	case IB_QPS_RESET:	return MLX5_QP_STATE_RST;
+	case IB_QPS_INIT:	return MLX5_QP_STATE_INIT;
+	case IB_QPS_RTR:	return MLX5_QP_STATE_RTR;
+	case IB_QPS_RTS:	return MLX5_QP_STATE_RTS;
+	case IB_QPS_SQD:	return MLX5_QP_STATE_SQD;
+	case IB_QPS_SQE:	return MLX5_QP_STATE_SQER;
+	case IB_QPS_ERR:	return MLX5_QP_STATE_ERR;
+	default:		return -1;
+	}
+}
+
+static int to_mlx5_st(enum ib_qp_type type)
+{
+	switch (type) {
+	case IB_QPT_RC:			return MLX5_QP_ST_RC;
+	case IB_QPT_UC:			return MLX5_QP_ST_UC;
+	case IB_QPT_UD:			return MLX5_QP_ST_UD;
+	case MLX5_IB_QPT_REG_UMR:	return MLX5_QP_ST_REG_UMR;
+	case IB_QPT_XRC_INI:
+	case IB_QPT_XRC_TGT:		return MLX5_QP_ST_XRC;
+	case IB_QPT_SMI:		return MLX5_QP_ST_QP0;
+	case MLX5_IB_QPT_HW_GSI:	return MLX5_QP_ST_QP1;
+	case MLX5_IB_QPT_DCI:		return MLX5_QP_ST_DCI;
+	case IB_QPT_RAW_IPV6:		return MLX5_QP_ST_RAW_IPV6;
+	case IB_QPT_RAW_PACKET:
+	case IB_QPT_RAW_ETHERTYPE:	return MLX5_QP_ST_RAW_ETHERTYPE;
+	case IB_QPT_MAX:
+	default:		return -EINVAL;
+	}
+}
+
+static void mlx5_ib_lock_cqs(struct mlx5_ib_cq *send_cq,
+			     struct mlx5_ib_cq *recv_cq);
+static void mlx5_ib_unlock_cqs(struct mlx5_ib_cq *send_cq,
+			       struct mlx5_ib_cq *recv_cq);
+
+int bfregn_to_uar_index(struct mlx5_ib_dev *dev,
+			struct mlx5_bfreg_info *bfregi, u32 bfregn,
+			bool dyn_bfreg)
+{
+	unsigned int bfregs_per_sys_page;
+	u32 index_of_sys_page;
+	u32 offset;
+
+	bfregs_per_sys_page = get_uars_per_sys_page(dev, bfregi->lib_uar_4k) *
+				MLX5_NON_FP_BFREGS_PER_UAR;
+	index_of_sys_page = bfregn / bfregs_per_sys_page;
+
+	if (dyn_bfreg) {
+		index_of_sys_page += bfregi->num_static_sys_pages;
+
+		if (index_of_sys_page >= bfregi->num_sys_pages)
+			return -EINVAL;
+
+		if (bfregn > bfregi->num_dyn_bfregs ||
+		    bfregi->sys_pages[index_of_sys_page] == MLX5_IB_INVALID_UAR_INDEX) {
+			mlx5_ib_dbg(dev, "Invalid dynamic uar index\n");
+			return -EINVAL;
+		}
+	}
+
+	offset = bfregn % bfregs_per_sys_page / MLX5_NON_FP_BFREGS_PER_UAR;
+	return bfregi->sys_pages[index_of_sys_page] + offset;
+}
+
+static int mlx5_ib_umem_get(struct mlx5_ib_dev *dev,
+			    struct ib_pd *pd,
+			    unsigned long addr, size_t size,
+			    struct ib_umem **umem,
+			    int *npages, int *page_shift, int *ncont,
+			    u32 *offset)
+{
+	int err;
+
+	*umem = ib_umem_get(pd->uobject->context, addr, size, 0, 0);
+	if (IS_ERR(*umem)) {
+		mlx5_ib_dbg(dev, "umem_get failed\n");
+		return PTR_ERR(*umem);
+	}
+
+	mlx5_ib_cont_pages(*umem, addr, 0, npages, page_shift, ncont, NULL);
+
+	err = mlx5_ib_get_buf_offset(addr, *page_shift, offset);
+	if (err) {
+		mlx5_ib_warn(dev, "bad offset\n");
+		goto err_umem;
+	}
+
+	mlx5_ib_dbg(dev, "addr 0x%lx, size %zu, npages %d, page_shift %d, ncont %d, offset %d\n",
+		    addr, size, *npages, *page_shift, *ncont, *offset);
+
+	return 0;
+
+err_umem:
+	ib_umem_release(*umem);
+	*umem = NULL;
+
+	return err;
+}
+
+static void destroy_user_rq(struct mlx5_ib_dev *dev, struct ib_pd *pd,
+			    struct mlx5_ib_rwq *rwq)
+{
+	struct mlx5_ib_ucontext *context;
+
+	if (rwq->create_flags & MLX5_IB_WQ_FLAGS_DELAY_DROP)
+		atomic_dec(&dev->delay_drop.rqs_cnt);
+
+	context = to_mucontext(pd->uobject->context);
+	mlx5_ib_db_unmap_user(context, &rwq->db);
+	if (rwq->umem)
+		ib_umem_release(rwq->umem);
+}
+
+static int create_user_rq(struct mlx5_ib_dev *dev, struct ib_pd *pd,
+			  struct mlx5_ib_rwq *rwq,
+			  struct mlx5_ib_create_wq *ucmd)
+{
+	struct mlx5_ib_ucontext *context;
+	int page_shift = 0;
+	int npages;
+	u32 offset = 0;
+	int ncont = 0;
+	int err;
+
+	if (!ucmd->buf_addr)
+		return -EINVAL;
+
+	context = to_mucontext(pd->uobject->context);
+	rwq->umem = ib_umem_get(pd->uobject->context, ucmd->buf_addr,
+			       rwq->buf_size, 0, 0);
+	if (IS_ERR(rwq->umem)) {
+		mlx5_ib_dbg(dev, "umem_get failed\n");
+		err = PTR_ERR(rwq->umem);
+		return err;
+	}
+
+	mlx5_ib_cont_pages(rwq->umem, ucmd->buf_addr, 0, &npages, &page_shift,
+			   &ncont, NULL);
+	err = mlx5_ib_get_buf_offset(ucmd->buf_addr, page_shift,
+				     &rwq->rq_page_offset);
+	if (err) {
+		mlx5_ib_warn(dev, "bad offset\n");
+		goto err_umem;
+	}
+
+	rwq->rq_num_pas = ncont;
+	rwq->page_shift = page_shift;
+	rwq->log_page_size =  page_shift - MLX5_ADAPTER_PAGE_SHIFT;
+	rwq->wq_sig = !!(ucmd->flags & MLX5_WQ_FLAG_SIGNATURE);
+
+	mlx5_ib_dbg(dev, "addr 0x%llx, size %zd, npages %d, page_shift %d, ncont %d, offset %d\n",
+		    (unsigned long long)ucmd->buf_addr, rwq->buf_size,
+		    npages, page_shift, ncont, offset);
+
+	err = mlx5_ib_db_map_user(context, ucmd->db_addr, &rwq->db);
+	if (err) {
+		mlx5_ib_dbg(dev, "map failed\n");
+		goto err_umem;
+	}
+
+	rwq->create_type = MLX5_WQ_USER;
+	return 0;
+
+err_umem:
+	ib_umem_release(rwq->umem);
+	return err;
+}
+
+static int adjust_bfregn(struct mlx5_ib_dev *dev,
+			 struct mlx5_bfreg_info *bfregi, int bfregn)
+{
+	return bfregn / MLX5_NON_FP_BFREGS_PER_UAR * MLX5_BFREGS_PER_UAR +
+				bfregn % MLX5_NON_FP_BFREGS_PER_UAR;
+}
+
+static int create_user_qp(struct mlx5_ib_dev *dev, struct ib_pd *pd,
+			  struct mlx5_ib_qp *qp, struct ib_udata *udata,
+			  struct ib_qp_init_attr *attr,
+			  u32 **in,
+			  struct mlx5_ib_create_qp_resp *resp, int *inlen,
+			  struct mlx5_ib_qp_base *base)
+{
+	struct mlx5_ib_ucontext *context;
+	struct mlx5_ib_create_qp ucmd;
+	struct mlx5_ib_ubuffer *ubuffer = &base->ubuffer;
+	int page_shift = 0;
+	int uar_index = 0;
+	int npages;
+	u32 offset = 0;
+	int bfregn;
+	int ncont = 0;
+	__be64 *pas;
+	void *qpc;
+	int err;
+
+	err = ib_copy_from_udata(&ucmd, udata, sizeof(ucmd));
+	if (err) {
+		mlx5_ib_dbg(dev, "copy failed\n");
+		return err;
+	}
+
+	context = to_mucontext(pd->uobject->context);
+	if (ucmd.flags & MLX5_QP_FLAG_BFREG_INDEX) {
+		uar_index = bfregn_to_uar_index(dev, &context->bfregi,
+						ucmd.bfreg_index, true);
+		if (uar_index < 0)
+			return uar_index;
+
+		bfregn = MLX5_IB_INVALID_BFREG;
+	} else if (qp->flags & MLX5_IB_QP_CROSS_CHANNEL) {
+		/*
+		 * TBD: should come from the verbs when we have the API
+		 */
+		/* In CROSS_CHANNEL CQ and QP must use the same UAR */
+		bfregn = MLX5_CROSS_CHANNEL_BFREG;
+	}
+	else {
+		bfregn = alloc_bfreg(dev, &context->bfregi);
+		if (bfregn < 0)
+			return bfregn;
+	}
+
+	mlx5_ib_dbg(dev, "bfregn 0x%x, uar_index 0x%x\n", bfregn, uar_index);
+	if (bfregn != MLX5_IB_INVALID_BFREG)
+		uar_index = bfregn_to_uar_index(dev, &context->bfregi, bfregn,
+						false);
+
+	qp->rq.offset = 0;
+	qp->sq.wqe_shift = ilog2(MLX5_SEND_WQE_BB);
+	qp->sq.offset = qp->rq.wqe_cnt << qp->rq.wqe_shift;
+
+	err = set_user_buf_size(dev, qp, &ucmd, base, attr);
+	if (err)
+		goto err_bfreg;
+
+	if (ucmd.buf_addr && ubuffer->buf_size) {
+		ubuffer->buf_addr = ucmd.buf_addr;
+		err = mlx5_ib_umem_get(dev, pd, ubuffer->buf_addr,
+				       ubuffer->buf_size,
+				       &ubuffer->umem, &npages, &page_shift,
+				       &ncont, &offset);
+		if (err)
+			goto err_bfreg;
+	} else {
+		ubuffer->umem = NULL;
+	}
+
+	*inlen = MLX5_ST_SZ_BYTES(create_qp_in) +
+		 MLX5_FLD_SZ_BYTES(create_qp_in, pas[0]) * ncont;
+	*in = kvzalloc(*inlen, GFP_KERNEL);
+	if (!*in) {
+		err = -ENOMEM;
+		goto err_umem;
+	}
+
+	pas = (__be64 *)MLX5_ADDR_OF(create_qp_in, *in, pas);
+	if (ubuffer->umem)
+		mlx5_ib_populate_pas(dev, ubuffer->umem, page_shift, pas, 0);
+
+	qpc = MLX5_ADDR_OF(create_qp_in, *in, qpc);
+
+	MLX5_SET(qpc, qpc, log_page_size, page_shift - MLX5_ADAPTER_PAGE_SHIFT);
+	MLX5_SET(qpc, qpc, page_offset, offset);
+
+	MLX5_SET(qpc, qpc, uar_page, uar_index);
+	if (bfregn != MLX5_IB_INVALID_BFREG)
+		resp->bfreg_index = adjust_bfregn(dev, &context->bfregi, bfregn);
+	else
+		resp->bfreg_index = MLX5_IB_INVALID_BFREG;
+	qp->bfregn = bfregn;
+
+	err = mlx5_ib_db_map_user(context, ucmd.db_addr, &qp->db);
+	if (err) {
+		mlx5_ib_dbg(dev, "map failed\n");
+		goto err_free;
+	}
+
+	err = ib_copy_to_udata(udata, resp, min(udata->outlen, sizeof(*resp)));
+	if (err) {
+		mlx5_ib_dbg(dev, "copy failed\n");
+		goto err_unmap;
+	}
+	qp->create_type = MLX5_QP_USER;
+
+	return 0;
+
+err_unmap:
+	mlx5_ib_db_unmap_user(context, &qp->db);
+
+err_free:
+	kvfree(*in);
+
+err_umem:
+	if (ubuffer->umem)
+		ib_umem_release(ubuffer->umem);
+
+err_bfreg:
+	if (bfregn != MLX5_IB_INVALID_BFREG)
+		mlx5_ib_free_bfreg(dev, &context->bfregi, bfregn);
+	return err;
+}
+
+static void destroy_qp_user(struct mlx5_ib_dev *dev, struct ib_pd *pd,
+			    struct mlx5_ib_qp *qp, struct mlx5_ib_qp_base *base)
+{
+	struct mlx5_ib_ucontext *context;
+
+	context = to_mucontext(pd->uobject->context);
+	mlx5_ib_db_unmap_user(context, &qp->db);
+	if (base->ubuffer.umem)
+		ib_umem_release(base->ubuffer.umem);
+
+	/*
+	 * Free only the BFREGs which are handled by the kernel.
+	 * BFREGs of UARs allocated dynamically are handled by user.
+	 */
+	if (qp->bfregn != MLX5_IB_INVALID_BFREG)
+		mlx5_ib_free_bfreg(dev, &context->bfregi, qp->bfregn);
+}
+
+static int create_kernel_qp(struct mlx5_ib_dev *dev,
+			    struct ib_qp_init_attr *init_attr,
+			    struct mlx5_ib_qp *qp,
+			    u32 **in, int *inlen,
+			    struct mlx5_ib_qp_base *base)
+{
+	int uar_index;
+	void *qpc;
+	int err;
+
+	if (init_attr->create_flags & ~(IB_QP_CREATE_SIGNATURE_EN |
+					IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK |
+					IB_QP_CREATE_IPOIB_UD_LSO |
+					IB_QP_CREATE_NETIF_QP |
+					mlx5_ib_create_qp_sqpn_qp1()))
+		return -EINVAL;
+
+	if (init_attr->qp_type == MLX5_IB_QPT_REG_UMR)
+		qp->bf.bfreg = &dev->fp_bfreg;
+	else
+		qp->bf.bfreg = &dev->bfreg;
+
+	/* We need to divide by two since each register is comprised of
+	 * two buffers of identical size, namely odd and even
+	 */
+	qp->bf.buf_size = (1 << MLX5_CAP_GEN(dev->mdev, log_bf_reg_size)) / 2;
+	uar_index = qp->bf.bfreg->index;
+
+	err = calc_sq_size(dev, init_attr, qp);
+	if (err < 0) {
+		mlx5_ib_dbg(dev, "err %d\n", err);
+		return err;
+	}
+
+	qp->rq.offset = 0;
+	qp->sq.offset = qp->rq.wqe_cnt << qp->rq.wqe_shift;
+	base->ubuffer.buf_size = err + (qp->rq.wqe_cnt << qp->rq.wqe_shift);
+
+	err = mlx5_buf_alloc(dev->mdev, base->ubuffer.buf_size, &qp->buf);
+	if (err) {
+		mlx5_ib_dbg(dev, "err %d\n", err);
+		return err;
+	}
+
+	qp->sq.qend = mlx5_get_send_wqe(qp, qp->sq.wqe_cnt);
+	*inlen = MLX5_ST_SZ_BYTES(create_qp_in) +
+		 MLX5_FLD_SZ_BYTES(create_qp_in, pas[0]) * qp->buf.npages;
+	*in = kvzalloc(*inlen, GFP_KERNEL);
+	if (!*in) {
+		err = -ENOMEM;
+		goto err_buf;
+	}
+
+	qpc = MLX5_ADDR_OF(create_qp_in, *in, qpc);
+	MLX5_SET(qpc, qpc, uar_page, uar_index);
+	MLX5_SET(qpc, qpc, log_page_size, qp->buf.page_shift - MLX5_ADAPTER_PAGE_SHIFT);
+
+	/* Set "fast registration enabled" for all kernel QPs */
+	MLX5_SET(qpc, qpc, fre, 1);
+	MLX5_SET(qpc, qpc, rlky, 1);
+
+	if (init_attr->create_flags & mlx5_ib_create_qp_sqpn_qp1()) {
+		MLX5_SET(qpc, qpc, deth_sqpn, 1);
+		qp->flags |= MLX5_IB_QP_SQPN_QP1;
+	}
+
+	mlx5_fill_page_array(&qp->buf,
+			     (__be64 *)MLX5_ADDR_OF(create_qp_in, *in, pas));
+
+	err = mlx5_db_alloc(dev->mdev, &qp->db);
+	if (err) {
+		mlx5_ib_dbg(dev, "err %d\n", err);
+		goto err_free;
+	}
+
+	qp->sq.wrid = kvmalloc_array(qp->sq.wqe_cnt,
+				     sizeof(*qp->sq.wrid), GFP_KERNEL);
+	qp->sq.wr_data = kvmalloc_array(qp->sq.wqe_cnt,
+					sizeof(*qp->sq.wr_data), GFP_KERNEL);
+	qp->rq.wrid = kvmalloc_array(qp->rq.wqe_cnt,
+				     sizeof(*qp->rq.wrid), GFP_KERNEL);
+	qp->sq.w_list = kvmalloc_array(qp->sq.wqe_cnt,
+				       sizeof(*qp->sq.w_list), GFP_KERNEL);
+	qp->sq.wqe_head = kvmalloc_array(qp->sq.wqe_cnt,
+					 sizeof(*qp->sq.wqe_head), GFP_KERNEL);
+
+	if (!qp->sq.wrid || !qp->sq.wr_data || !qp->rq.wrid ||
+	    !qp->sq.w_list || !qp->sq.wqe_head) {
+		err = -ENOMEM;
+		goto err_wrid;
+	}
+	qp->create_type = MLX5_QP_KERNEL;
+
+	return 0;
+
+err_wrid:
+	kvfree(qp->sq.wqe_head);
+	kvfree(qp->sq.w_list);
+	kvfree(qp->sq.wrid);
+	kvfree(qp->sq.wr_data);
+	kvfree(qp->rq.wrid);
+	mlx5_db_free(dev->mdev, &qp->db);
+
+err_free:
+	kvfree(*in);
+
+err_buf:
+	mlx5_buf_free(dev->mdev, &qp->buf);
+	return err;
+}
+
+static void destroy_qp_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp)
+{
+	kvfree(qp->sq.wqe_head);
+	kvfree(qp->sq.w_list);
+	kvfree(qp->sq.wrid);
+	kvfree(qp->sq.wr_data);
+	kvfree(qp->rq.wrid);
+	mlx5_db_free(dev->mdev, &qp->db);
+	mlx5_buf_free(dev->mdev, &qp->buf);
+}
+
+static u32 get_rx_type(struct mlx5_ib_qp *qp, struct ib_qp_init_attr *attr)
+{
+	if (attr->srq || (attr->qp_type == IB_QPT_XRC_TGT) ||
+	    (attr->qp_type == MLX5_IB_QPT_DCI) ||
+	    (attr->qp_type == IB_QPT_XRC_INI))
+		return MLX5_SRQ_RQ;
+	else if (!qp->has_rq)
+		return MLX5_ZERO_LEN_RQ;
+	else
+		return MLX5_NON_ZERO_RQ;
+}
+
+static int is_connected(enum ib_qp_type qp_type)
+{
+	if (qp_type == IB_QPT_RC || qp_type == IB_QPT_UC)
+		return 1;
+
+	return 0;
+}
+
+static int create_raw_packet_qp_tis(struct mlx5_ib_dev *dev,
+				    struct mlx5_ib_qp *qp,
+				    struct mlx5_ib_sq *sq, u32 tdn)
+{
+	u32 in[MLX5_ST_SZ_DW(create_tis_in)] = {0};
+	void *tisc = MLX5_ADDR_OF(create_tis_in, in, ctx);
+
+	MLX5_SET(tisc, tisc, transport_domain, tdn);
+	if (qp->flags & MLX5_IB_QP_UNDERLAY)
+		MLX5_SET(tisc, tisc, underlay_qpn, qp->underlay_qpn);
+
+	return mlx5_core_create_tis(dev->mdev, in, sizeof(in), &sq->tisn);
+}
+
+static void destroy_raw_packet_qp_tis(struct mlx5_ib_dev *dev,
+				      struct mlx5_ib_sq *sq)
+{
+	mlx5_core_destroy_tis(dev->mdev, sq->tisn);
+}
+
+static void destroy_flow_rule_vport_sq(struct mlx5_ib_dev *dev,
+				       struct mlx5_ib_sq *sq)
+{
+	if (sq->flow_rule)
+		mlx5_del_flow_rules(sq->flow_rule);
+}
+
+static int create_raw_packet_qp_sq(struct mlx5_ib_dev *dev,
+				   struct mlx5_ib_sq *sq, void *qpin,
+				   struct ib_pd *pd)
+{
+	struct mlx5_ib_ubuffer *ubuffer = &sq->ubuffer;
+	__be64 *pas;
+	void *in;
+	void *sqc;
+	void *qpc = MLX5_ADDR_OF(create_qp_in, qpin, qpc);
+	void *wq;
+	int inlen;
+	int err;
+	int page_shift = 0;
+	int npages;
+	int ncont = 0;
+	u32 offset = 0;
+
+	err = mlx5_ib_umem_get(dev, pd, ubuffer->buf_addr, ubuffer->buf_size,
+			       &sq->ubuffer.umem, &npages, &page_shift,
+			       &ncont, &offset);
+	if (err)
+		return err;
+
+	inlen = MLX5_ST_SZ_BYTES(create_sq_in) + sizeof(u64) * ncont;
+	in = kvzalloc(inlen, GFP_KERNEL);
+	if (!in) {
+		err = -ENOMEM;
+		goto err_umem;
+	}
+
+	sqc = MLX5_ADDR_OF(create_sq_in, in, ctx);
+	MLX5_SET(sqc, sqc, flush_in_error_en, 1);
+	if (MLX5_CAP_ETH(dev->mdev, multi_pkt_send_wqe))
+		MLX5_SET(sqc, sqc, allow_multi_pkt_send_wqe, 1);
+	MLX5_SET(sqc, sqc, state, MLX5_SQC_STATE_RST);
+	MLX5_SET(sqc, sqc, user_index, MLX5_GET(qpc, qpc, user_index));
+	MLX5_SET(sqc, sqc, cqn, MLX5_GET(qpc, qpc, cqn_snd));
+	MLX5_SET(sqc, sqc, tis_lst_sz, 1);
+	MLX5_SET(sqc, sqc, tis_num_0, sq->tisn);
+	if (MLX5_CAP_GEN(dev->mdev, eth_net_offloads) &&
+	    MLX5_CAP_ETH(dev->mdev, swp))
+		MLX5_SET(sqc, sqc, allow_swp, 1);
+
+	wq = MLX5_ADDR_OF(sqc, sqc, wq);
+	MLX5_SET(wq, wq, wq_type, MLX5_WQ_TYPE_CYCLIC);
+	MLX5_SET(wq, wq, pd, MLX5_GET(qpc, qpc, pd));
+	MLX5_SET(wq, wq, uar_page, MLX5_GET(qpc, qpc, uar_page));
+	MLX5_SET64(wq, wq, dbr_addr, MLX5_GET64(qpc, qpc, dbr_addr));
+	MLX5_SET(wq, wq, log_wq_stride, ilog2(MLX5_SEND_WQE_BB));
+	MLX5_SET(wq, wq, log_wq_sz, MLX5_GET(qpc, qpc, log_sq_size));
+	MLX5_SET(wq, wq, log_wq_pg_sz,  page_shift - MLX5_ADAPTER_PAGE_SHIFT);
+	MLX5_SET(wq, wq, page_offset, offset);
+
+	pas = (__be64 *)MLX5_ADDR_OF(wq, wq, pas);
+	mlx5_ib_populate_pas(dev, sq->ubuffer.umem, page_shift, pas, 0);
+
+	err = mlx5_core_create_sq_tracked(dev->mdev, in, inlen, &sq->base.mqp);
+
+	kvfree(in);
+
+	if (err)
+		goto err_umem;
+
+	err = create_flow_rule_vport_sq(dev, sq);
+	if (err)
+		goto err_flow;
+
+	return 0;
+
+err_flow:
+	mlx5_core_destroy_sq_tracked(dev->mdev, &sq->base.mqp);
+
+err_umem:
+	ib_umem_release(sq->ubuffer.umem);
+	sq->ubuffer.umem = NULL;
+
+	return err;
+}
+
+static void destroy_raw_packet_qp_sq(struct mlx5_ib_dev *dev,
+				     struct mlx5_ib_sq *sq)
+{
+	destroy_flow_rule_vport_sq(dev, sq);
+	mlx5_core_destroy_sq_tracked(dev->mdev, &sq->base.mqp);
+	ib_umem_release(sq->ubuffer.umem);
+}
+
+static size_t get_rq_pas_size(void *qpc)
+{
+	u32 log_page_size = MLX5_GET(qpc, qpc, log_page_size) + 12;
+	u32 log_rq_stride = MLX5_GET(qpc, qpc, log_rq_stride);
+	u32 log_rq_size   = MLX5_GET(qpc, qpc, log_rq_size);
+	u32 page_offset   = MLX5_GET(qpc, qpc, page_offset);
+	u32 po_quanta	  = 1 << (log_page_size - 6);
+	u32 rq_sz	  = 1 << (log_rq_size + 4 + log_rq_stride);
+	u32 page_size	  = 1 << log_page_size;
+	u32 rq_sz_po      = rq_sz + (page_offset * po_quanta);
+	u32 rq_num_pas	  = (rq_sz_po + page_size - 1) / page_size;
+
+	return rq_num_pas * sizeof(u64);
+}
+
+static int create_raw_packet_qp_rq(struct mlx5_ib_dev *dev,
+				   struct mlx5_ib_rq *rq, void *qpin,
+				   size_t qpinlen)
+{
+	struct mlx5_ib_qp *mqp = rq->base.container_mibqp;
+	__be64 *pas;
+	__be64 *qp_pas;
+	void *in;
+	void *rqc;
+	void *wq;
+	void *qpc = MLX5_ADDR_OF(create_qp_in, qpin, qpc);
+	size_t rq_pas_size = get_rq_pas_size(qpc);
+	size_t inlen;
+	int err;
+
+	if (qpinlen < rq_pas_size + MLX5_BYTE_OFF(create_qp_in, pas))
+		return -EINVAL;
+
+	inlen = MLX5_ST_SZ_BYTES(create_rq_in) + rq_pas_size;
+	in = kvzalloc(inlen, GFP_KERNEL);
+	if (!in)
+		return -ENOMEM;
+
+	rqc = MLX5_ADDR_OF(create_rq_in, in, ctx);
+	if (!(rq->flags & MLX5_IB_RQ_CVLAN_STRIPPING))
+		MLX5_SET(rqc, rqc, vsd, 1);
+	MLX5_SET(rqc, rqc, mem_rq_type, MLX5_RQC_MEM_RQ_TYPE_MEMORY_RQ_INLINE);
+	MLX5_SET(rqc, rqc, state, MLX5_RQC_STATE_RST);
+	MLX5_SET(rqc, rqc, flush_in_error_en, 1);
+	MLX5_SET(rqc, rqc, user_index, MLX5_GET(qpc, qpc, user_index));
+	MLX5_SET(rqc, rqc, cqn, MLX5_GET(qpc, qpc, cqn_rcv));
+
+	if (mqp->flags & MLX5_IB_QP_CAP_SCATTER_FCS)
+		MLX5_SET(rqc, rqc, scatter_fcs, 1);
+
+	wq = MLX5_ADDR_OF(rqc, rqc, wq);
+	MLX5_SET(wq, wq, wq_type, MLX5_WQ_TYPE_CYCLIC);
+	if (rq->flags & MLX5_IB_RQ_PCI_WRITE_END_PADDING)
+		MLX5_SET(wq, wq, end_padding_mode, MLX5_WQ_END_PAD_MODE_ALIGN);
+	MLX5_SET(wq, wq, page_offset, MLX5_GET(qpc, qpc, page_offset));
+	MLX5_SET(wq, wq, pd, MLX5_GET(qpc, qpc, pd));
+	MLX5_SET64(wq, wq, dbr_addr, MLX5_GET64(qpc, qpc, dbr_addr));
+	MLX5_SET(wq, wq, log_wq_stride, MLX5_GET(qpc, qpc, log_rq_stride) + 4);
+	MLX5_SET(wq, wq, log_wq_pg_sz, MLX5_GET(qpc, qpc, log_page_size));
+	MLX5_SET(wq, wq, log_wq_sz, MLX5_GET(qpc, qpc, log_rq_size));
+
+	pas = (__be64 *)MLX5_ADDR_OF(wq, wq, pas);
+	qp_pas = (__be64 *)MLX5_ADDR_OF(create_qp_in, qpin, pas);
+	memcpy(pas, qp_pas, rq_pas_size);
+
+	err = mlx5_core_create_rq_tracked(dev->mdev, in, inlen, &rq->base.mqp);
+
+	kvfree(in);
+
+	return err;
+}
+
+static void destroy_raw_packet_qp_rq(struct mlx5_ib_dev *dev,
+				     struct mlx5_ib_rq *rq)
+{
+	mlx5_core_destroy_rq_tracked(dev->mdev, &rq->base.mqp);
+}
+
+static bool tunnel_offload_supported(struct mlx5_core_dev *dev)
+{
+	return  (MLX5_CAP_ETH(dev, tunnel_stateless_vxlan) ||
+		 MLX5_CAP_ETH(dev, tunnel_stateless_gre) ||
+		 MLX5_CAP_ETH(dev, tunnel_stateless_geneve_rx));
+}
+
+static int create_raw_packet_qp_tir(struct mlx5_ib_dev *dev,
+				    struct mlx5_ib_rq *rq, u32 tdn,
+				    bool tunnel_offload_en)
+{
+	u32 *in;
+	void *tirc;
+	int inlen;
+	int err;
+
+	inlen = MLX5_ST_SZ_BYTES(create_tir_in);
+	in = kvzalloc(inlen, GFP_KERNEL);
+	if (!in)
+		return -ENOMEM;
+
+	tirc = MLX5_ADDR_OF(create_tir_in, in, ctx);
+	MLX5_SET(tirc, tirc, disp_type, MLX5_TIRC_DISP_TYPE_DIRECT);
+	MLX5_SET(tirc, tirc, inline_rqn, rq->base.mqp.qpn);
+	MLX5_SET(tirc, tirc, transport_domain, tdn);
+	if (tunnel_offload_en)
+		MLX5_SET(tirc, tirc, tunneled_offload_en, 1);
+
+	if (dev->rep)
+		MLX5_SET(tirc, tirc, self_lb_block,
+			 MLX5_TIRC_SELF_LB_BLOCK_BLOCK_UNICAST_);
+
+	err = mlx5_core_create_tir(dev->mdev, in, inlen, &rq->tirn);
+
+	kvfree(in);
+
+	return err;
+}
+
+static void destroy_raw_packet_qp_tir(struct mlx5_ib_dev *dev,
+				      struct mlx5_ib_rq *rq)
+{
+	mlx5_core_destroy_tir(dev->mdev, rq->tirn);
+}
+
+static int create_raw_packet_qp(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
+				u32 *in, size_t inlen,
+				struct ib_pd *pd)
+{
+	struct mlx5_ib_raw_packet_qp *raw_packet_qp = &qp->raw_packet_qp;
+	struct mlx5_ib_sq *sq = &raw_packet_qp->sq;
+	struct mlx5_ib_rq *rq = &raw_packet_qp->rq;
+	struct ib_uobject *uobj = pd->uobject;
+	struct ib_ucontext *ucontext = uobj->context;
+	struct mlx5_ib_ucontext *mucontext = to_mucontext(ucontext);
+	int err;
+	u32 tdn = mucontext->tdn;
+
+	if (qp->sq.wqe_cnt) {
+		err = create_raw_packet_qp_tis(dev, qp, sq, tdn);
+		if (err)
+			return err;
+
+		err = create_raw_packet_qp_sq(dev, sq, in, pd);
+		if (err)
+			goto err_destroy_tis;
+
+		sq->base.container_mibqp = qp;
+		sq->base.mqp.event = mlx5_ib_qp_event;
+	}
+
+	if (qp->rq.wqe_cnt) {
+		rq->base.container_mibqp = qp;
+
+		if (qp->flags & MLX5_IB_QP_CVLAN_STRIPPING)
+			rq->flags |= MLX5_IB_RQ_CVLAN_STRIPPING;
+		if (qp->flags & MLX5_IB_QP_PCI_WRITE_END_PADDING)
+			rq->flags |= MLX5_IB_RQ_PCI_WRITE_END_PADDING;
+		err = create_raw_packet_qp_rq(dev, rq, in, inlen);
+		if (err)
+			goto err_destroy_sq;
+
+
+		err = create_raw_packet_qp_tir(dev, rq, tdn,
+					       qp->tunnel_offload_en);
+		if (err)
+			goto err_destroy_rq;
+	}
+
+	qp->trans_qp.base.mqp.qpn = qp->sq.wqe_cnt ? sq->base.mqp.qpn :
+						     rq->base.mqp.qpn;
+
+	return 0;
+
+err_destroy_rq:
+	destroy_raw_packet_qp_rq(dev, rq);
+err_destroy_sq:
+	if (!qp->sq.wqe_cnt)
+		return err;
+	destroy_raw_packet_qp_sq(dev, sq);
+err_destroy_tis:
+	destroy_raw_packet_qp_tis(dev, sq);
+
+	return err;
+}
+
+static void destroy_raw_packet_qp(struct mlx5_ib_dev *dev,
+				  struct mlx5_ib_qp *qp)
+{
+	struct mlx5_ib_raw_packet_qp *raw_packet_qp = &qp->raw_packet_qp;
+	struct mlx5_ib_sq *sq = &raw_packet_qp->sq;
+	struct mlx5_ib_rq *rq = &raw_packet_qp->rq;
+
+	if (qp->rq.wqe_cnt) {
+		destroy_raw_packet_qp_tir(dev, rq);
+		destroy_raw_packet_qp_rq(dev, rq);
+	}
+
+	if (qp->sq.wqe_cnt) {
+		destroy_raw_packet_qp_sq(dev, sq);
+		destroy_raw_packet_qp_tis(dev, sq);
+	}
+}
+
+static void raw_packet_qp_copy_info(struct mlx5_ib_qp *qp,
+				    struct mlx5_ib_raw_packet_qp *raw_packet_qp)
+{
+	struct mlx5_ib_sq *sq = &raw_packet_qp->sq;
+	struct mlx5_ib_rq *rq = &raw_packet_qp->rq;
+
+	sq->sq = &qp->sq;
+	rq->rq = &qp->rq;
+	sq->doorbell = &qp->db;
+	rq->doorbell = &qp->db;
+}
+
+static void destroy_rss_raw_qp_tir(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp)
+{
+	mlx5_core_destroy_tir(dev->mdev, qp->rss_qp.tirn);
+}
+
+static int create_rss_raw_qp_tir(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
+				 struct ib_pd *pd,
+				 struct ib_qp_init_attr *init_attr,
+				 struct ib_udata *udata)
+{
+	struct ib_uobject *uobj = pd->uobject;
+	struct ib_ucontext *ucontext = uobj->context;
+	struct mlx5_ib_ucontext *mucontext = to_mucontext(ucontext);
+	struct mlx5_ib_create_qp_resp resp = {};
+	int inlen;
+	int err;
+	u32 *in;
+	void *tirc;
+	void *hfso;
+	u32 selected_fields = 0;
+	u32 outer_l4;
+	size_t min_resp_len;
+	u32 tdn = mucontext->tdn;
+	struct mlx5_ib_create_qp_rss ucmd = {};
+	size_t required_cmd_sz;
+
+	if (init_attr->qp_type != IB_QPT_RAW_PACKET)
+		return -EOPNOTSUPP;
+
+	if (init_attr->create_flags || init_attr->send_cq)
+		return -EINVAL;
+
+	min_resp_len = offsetof(typeof(resp), bfreg_index) + sizeof(resp.bfreg_index);
+	if (udata->outlen < min_resp_len)
+		return -EINVAL;
+
+	required_cmd_sz = offsetof(typeof(ucmd), flags) + sizeof(ucmd.flags);
+	if (udata->inlen < required_cmd_sz) {
+		mlx5_ib_dbg(dev, "invalid inlen\n");
+		return -EINVAL;
+	}
+
+	if (udata->inlen > sizeof(ucmd) &&
+	    !ib_is_udata_cleared(udata, sizeof(ucmd),
+				 udata->inlen - sizeof(ucmd))) {
+		mlx5_ib_dbg(dev, "inlen is not supported\n");
+		return -EOPNOTSUPP;
+	}
+
+	if (ib_copy_from_udata(&ucmd, udata, min(sizeof(ucmd), udata->inlen))) {
+		mlx5_ib_dbg(dev, "copy failed\n");
+		return -EFAULT;
+	}
+
+	if (ucmd.comp_mask) {
+		mlx5_ib_dbg(dev, "invalid comp mask\n");
+		return -EOPNOTSUPP;
+	}
+
+	if (ucmd.flags & ~MLX5_QP_FLAG_TUNNEL_OFFLOADS) {
+		mlx5_ib_dbg(dev, "invalid flags\n");
+		return -EOPNOTSUPP;
+	}
+
+	if (ucmd.flags & MLX5_QP_FLAG_TUNNEL_OFFLOADS &&
+	    !tunnel_offload_supported(dev->mdev)) {
+		mlx5_ib_dbg(dev, "tunnel offloads isn't supported\n");
+		return -EOPNOTSUPP;
+	}
+
+	if (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_INNER &&
+	    !(ucmd.flags & MLX5_QP_FLAG_TUNNEL_OFFLOADS)) {
+		mlx5_ib_dbg(dev, "Tunnel offloads must be set for inner RSS\n");
+		return -EOPNOTSUPP;
+	}
+
+	err = ib_copy_to_udata(udata, &resp, min(udata->outlen, sizeof(resp)));
+	if (err) {
+		mlx5_ib_dbg(dev, "copy failed\n");
+		return -EINVAL;
+	}
+
+	inlen = MLX5_ST_SZ_BYTES(create_tir_in);
+	in = kvzalloc(inlen, GFP_KERNEL);
+	if (!in)
+		return -ENOMEM;
+
+	tirc = MLX5_ADDR_OF(create_tir_in, in, ctx);
+	MLX5_SET(tirc, tirc, disp_type,
+		 MLX5_TIRC_DISP_TYPE_INDIRECT);
+	MLX5_SET(tirc, tirc, indirect_table,
+		 init_attr->rwq_ind_tbl->ind_tbl_num);
+	MLX5_SET(tirc, tirc, transport_domain, tdn);
+
+	hfso = MLX5_ADDR_OF(tirc, tirc, rx_hash_field_selector_outer);
+
+	if (ucmd.flags & MLX5_QP_FLAG_TUNNEL_OFFLOADS)
+		MLX5_SET(tirc, tirc, tunneled_offload_en, 1);
+
+	if (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_INNER)
+		hfso = MLX5_ADDR_OF(tirc, tirc, rx_hash_field_selector_inner);
+	else
+		hfso = MLX5_ADDR_OF(tirc, tirc, rx_hash_field_selector_outer);
+
+	switch (ucmd.rx_hash_function) {
+	case MLX5_RX_HASH_FUNC_TOEPLITZ:
+	{
+		void *rss_key = MLX5_ADDR_OF(tirc, tirc, rx_hash_toeplitz_key);
+		size_t len = MLX5_FLD_SZ_BYTES(tirc, rx_hash_toeplitz_key);
+
+		if (len != ucmd.rx_key_len) {
+			err = -EINVAL;
+			goto err;
+		}
+
+		MLX5_SET(tirc, tirc, rx_hash_fn, MLX5_RX_HASH_FN_TOEPLITZ);
+		memcpy(rss_key, ucmd.rx_hash_key, len);
+		break;
+	}
+	default:
+		err = -EOPNOTSUPP;
+		goto err;
+	}
+
+	if (!ucmd.rx_hash_fields_mask) {
+		/* special case when this TIR serves as steering entry without hashing */
+		if (!init_attr->rwq_ind_tbl->log_ind_tbl_size)
+			goto create_tir;
+		err = -EINVAL;
+		goto err;
+	}
+
+	if (((ucmd.rx_hash_fields_mask & MLX5_RX_HASH_SRC_IPV4) ||
+	     (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_DST_IPV4)) &&
+	     ((ucmd.rx_hash_fields_mask & MLX5_RX_HASH_SRC_IPV6) ||
+	     (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_DST_IPV6))) {
+		err = -EINVAL;
+		goto err;
+	}
+
+	/* If none of IPV4 & IPV6 SRC/DST was set - this bit field is ignored */
+	if ((ucmd.rx_hash_fields_mask & MLX5_RX_HASH_SRC_IPV4) ||
+	    (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_DST_IPV4))
+		MLX5_SET(rx_hash_field_select, hfso, l3_prot_type,
+			 MLX5_L3_PROT_TYPE_IPV4);
+	else if ((ucmd.rx_hash_fields_mask & MLX5_RX_HASH_SRC_IPV6) ||
+		 (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_DST_IPV6))
+		MLX5_SET(rx_hash_field_select, hfso, l3_prot_type,
+			 MLX5_L3_PROT_TYPE_IPV6);
+
+	outer_l4 = ((ucmd.rx_hash_fields_mask & MLX5_RX_HASH_SRC_PORT_TCP) ||
+		    (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_DST_PORT_TCP)) << 0 |
+		   ((ucmd.rx_hash_fields_mask & MLX5_RX_HASH_SRC_PORT_UDP) ||
+		    (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_DST_PORT_UDP)) << 1 |
+		   (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_IPSEC_SPI) << 2;
+
+	/* Check that only one l4 protocol is set */
+	if (outer_l4 & (outer_l4 - 1)) {
+		err = -EINVAL;
+		goto err;
+	}
+
+	/* If none of TCP & UDP SRC/DST was set - this bit field is ignored */
+	if ((ucmd.rx_hash_fields_mask & MLX5_RX_HASH_SRC_PORT_TCP) ||
+	    (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_DST_PORT_TCP))
+		MLX5_SET(rx_hash_field_select, hfso, l4_prot_type,
+			 MLX5_L4_PROT_TYPE_TCP);
+	else if ((ucmd.rx_hash_fields_mask & MLX5_RX_HASH_SRC_PORT_UDP) ||
+		 (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_DST_PORT_UDP))
+		MLX5_SET(rx_hash_field_select, hfso, l4_prot_type,
+			 MLX5_L4_PROT_TYPE_UDP);
+
+	if ((ucmd.rx_hash_fields_mask & MLX5_RX_HASH_SRC_IPV4) ||
+	    (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_SRC_IPV6))
+		selected_fields |= MLX5_HASH_FIELD_SEL_SRC_IP;
+
+	if ((ucmd.rx_hash_fields_mask & MLX5_RX_HASH_DST_IPV4) ||
+	    (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_DST_IPV6))
+		selected_fields |= MLX5_HASH_FIELD_SEL_DST_IP;
+
+	if ((ucmd.rx_hash_fields_mask & MLX5_RX_HASH_SRC_PORT_TCP) ||
+	    (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_SRC_PORT_UDP))
+		selected_fields |= MLX5_HASH_FIELD_SEL_L4_SPORT;
+
+	if ((ucmd.rx_hash_fields_mask & MLX5_RX_HASH_DST_PORT_TCP) ||
+	    (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_DST_PORT_UDP))
+		selected_fields |= MLX5_HASH_FIELD_SEL_L4_DPORT;
+
+	if (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_IPSEC_SPI)
+		selected_fields |= MLX5_HASH_FIELD_SEL_IPSEC_SPI;
+
+	MLX5_SET(rx_hash_field_select, hfso, selected_fields, selected_fields);
+
+create_tir:
+	if (dev->rep)
+		MLX5_SET(tirc, tirc, self_lb_block,
+			 MLX5_TIRC_SELF_LB_BLOCK_BLOCK_UNICAST_);
+
+	err = mlx5_core_create_tir(dev->mdev, in, inlen, &qp->rss_qp.tirn);
+
+	if (err)
+		goto err;
+
+	kvfree(in);
+	/* qpn is reserved for that QP */
+	qp->trans_qp.base.mqp.qpn = 0;
+	qp->flags |= MLX5_IB_QP_RSS;
+	return 0;
+
+err:
+	kvfree(in);
+	return err;
+}
+
+static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd,
+			    struct ib_qp_init_attr *init_attr,
+			    struct ib_udata *udata, struct mlx5_ib_qp *qp)
+{
+	struct mlx5_ib_resources *devr = &dev->devr;
+	int inlen = MLX5_ST_SZ_BYTES(create_qp_in);
+	struct mlx5_core_dev *mdev = dev->mdev;
+	struct mlx5_ib_create_qp_resp resp = {};
+	struct mlx5_ib_cq *send_cq;
+	struct mlx5_ib_cq *recv_cq;
+	unsigned long flags;
+	u32 uidx = MLX5_IB_DEFAULT_UIDX;
+	struct mlx5_ib_create_qp ucmd;
+	struct mlx5_ib_qp_base *base;
+	int mlx5_st;
+	void *qpc;
+	u32 *in;
+	int err;
+
+	mutex_init(&qp->mutex);
+	spin_lock_init(&qp->sq.lock);
+	spin_lock_init(&qp->rq.lock);
+
+	mlx5_st = to_mlx5_st(init_attr->qp_type);
+	if (mlx5_st < 0)
+		return -EINVAL;
+
+	if (init_attr->rwq_ind_tbl) {
+		if (!udata)
+			return -ENOSYS;
+
+		err = create_rss_raw_qp_tir(dev, qp, pd, init_attr, udata);
+		return err;
+	}
+
+	if (init_attr->create_flags & IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK) {
+		if (!MLX5_CAP_GEN(mdev, block_lb_mc)) {
+			mlx5_ib_dbg(dev, "block multicast loopback isn't supported\n");
+			return -EINVAL;
+		} else {
+			qp->flags |= MLX5_IB_QP_BLOCK_MULTICAST_LOOPBACK;
+		}
+	}
+
+	if (init_attr->create_flags &
+			(IB_QP_CREATE_CROSS_CHANNEL |
+			 IB_QP_CREATE_MANAGED_SEND |
+			 IB_QP_CREATE_MANAGED_RECV)) {
+		if (!MLX5_CAP_GEN(mdev, cd)) {
+			mlx5_ib_dbg(dev, "cross-channel isn't supported\n");
+			return -EINVAL;
+		}
+		if (init_attr->create_flags & IB_QP_CREATE_CROSS_CHANNEL)
+			qp->flags |= MLX5_IB_QP_CROSS_CHANNEL;
+		if (init_attr->create_flags & IB_QP_CREATE_MANAGED_SEND)
+			qp->flags |= MLX5_IB_QP_MANAGED_SEND;
+		if (init_attr->create_flags & IB_QP_CREATE_MANAGED_RECV)
+			qp->flags |= MLX5_IB_QP_MANAGED_RECV;
+	}
+
+	if (init_attr->qp_type == IB_QPT_UD &&
+	    (init_attr->create_flags & IB_QP_CREATE_IPOIB_UD_LSO))
+		if (!MLX5_CAP_GEN(mdev, ipoib_basic_offloads)) {
+			mlx5_ib_dbg(dev, "ipoib UD lso qp isn't supported\n");
+			return -EOPNOTSUPP;
+		}
+
+	if (init_attr->create_flags & IB_QP_CREATE_SCATTER_FCS) {
+		if (init_attr->qp_type != IB_QPT_RAW_PACKET) {
+			mlx5_ib_dbg(dev, "Scatter FCS is supported only for Raw Packet QPs");
+			return -EOPNOTSUPP;
+		}
+		if (!MLX5_CAP_GEN(dev->mdev, eth_net_offloads) ||
+		    !MLX5_CAP_ETH(dev->mdev, scatter_fcs)) {
+			mlx5_ib_dbg(dev, "Scatter FCS isn't supported\n");
+			return -EOPNOTSUPP;
+		}
+		qp->flags |= MLX5_IB_QP_CAP_SCATTER_FCS;
+	}
+
+	if (init_attr->sq_sig_type == IB_SIGNAL_ALL_WR)
+		qp->sq_signal_bits = MLX5_WQE_CTRL_CQ_UPDATE;
+
+	if (init_attr->create_flags & IB_QP_CREATE_CVLAN_STRIPPING) {
+		if (!(MLX5_CAP_GEN(dev->mdev, eth_net_offloads) &&
+		      MLX5_CAP_ETH(dev->mdev, vlan_cap)) ||
+		    (init_attr->qp_type != IB_QPT_RAW_PACKET))
+			return -EOPNOTSUPP;
+		qp->flags |= MLX5_IB_QP_CVLAN_STRIPPING;
+	}
+
+	if (pd && pd->uobject) {
+		if (ib_copy_from_udata(&ucmd, udata, sizeof(ucmd))) {
+			mlx5_ib_dbg(dev, "copy failed\n");
+			return -EFAULT;
+		}
+
+		err = get_qp_user_index(to_mucontext(pd->uobject->context),
+					&ucmd, udata->inlen, &uidx);
+		if (err)
+			return err;
+
+		qp->wq_sig = !!(ucmd.flags & MLX5_QP_FLAG_SIGNATURE);
+		qp->scat_cqe = !!(ucmd.flags & MLX5_QP_FLAG_SCATTER_CQE);
+		if (ucmd.flags & MLX5_QP_FLAG_TUNNEL_OFFLOADS) {
+			if (init_attr->qp_type != IB_QPT_RAW_PACKET ||
+			    !tunnel_offload_supported(mdev)) {
+				mlx5_ib_dbg(dev, "Tunnel offload isn't supported\n");
+				return -EOPNOTSUPP;
+			}
+			qp->tunnel_offload_en = true;
+		}
+
+		if (init_attr->create_flags & IB_QP_CREATE_SOURCE_QPN) {
+			if (init_attr->qp_type != IB_QPT_UD ||
+			    (MLX5_CAP_GEN(dev->mdev, port_type) !=
+			     MLX5_CAP_PORT_TYPE_IB) ||
+			    !mlx5_get_flow_namespace(dev->mdev, MLX5_FLOW_NAMESPACE_BYPASS)) {
+				mlx5_ib_dbg(dev, "Source QP option isn't supported\n");
+				return -EOPNOTSUPP;
+			}
+
+			qp->flags |= MLX5_IB_QP_UNDERLAY;
+			qp->underlay_qpn = init_attr->source_qpn;
+		}
+	} else {
+		qp->wq_sig = !!wq_signature;
+	}
+
+	base = (init_attr->qp_type == IB_QPT_RAW_PACKET ||
+		qp->flags & MLX5_IB_QP_UNDERLAY) ?
+	       &qp->raw_packet_qp.rq.base :
+	       &qp->trans_qp.base;
+
+	qp->has_rq = qp_has_rq(init_attr);
+	err = set_rq_size(dev, &init_attr->cap, qp->has_rq,
+			  qp, (pd && pd->uobject) ? &ucmd : NULL);
+	if (err) {
+		mlx5_ib_dbg(dev, "err %d\n", err);
+		return err;
+	}
+
+	if (pd) {
+		if (pd->uobject) {
+			__u32 max_wqes =
+				1 << MLX5_CAP_GEN(mdev, log_max_qp_sz);
+			mlx5_ib_dbg(dev, "requested sq_wqe_count (%d)\n", ucmd.sq_wqe_count);
+			if (ucmd.rq_wqe_shift != qp->rq.wqe_shift ||
+			    ucmd.rq_wqe_count != qp->rq.wqe_cnt) {
+				mlx5_ib_dbg(dev, "invalid rq params\n");
+				return -EINVAL;
+			}
+			if (ucmd.sq_wqe_count > max_wqes) {
+				mlx5_ib_dbg(dev, "requested sq_wqe_count (%d) > max allowed (%d)\n",
+					    ucmd.sq_wqe_count, max_wqes);
+				return -EINVAL;
+			}
+			if (init_attr->create_flags &
+			    mlx5_ib_create_qp_sqpn_qp1()) {
+				mlx5_ib_dbg(dev, "user-space is not allowed to create UD QPs spoofing as QP1\n");
+				return -EINVAL;
+			}
+			err = create_user_qp(dev, pd, qp, udata, init_attr, &in,
+					     &resp, &inlen, base);
+			if (err)
+				mlx5_ib_dbg(dev, "err %d\n", err);
+		} else {
+			err = create_kernel_qp(dev, init_attr, qp, &in, &inlen,
+					       base);
+			if (err)
+				mlx5_ib_dbg(dev, "err %d\n", err);
+		}
+
+		if (err)
+			return err;
+	} else {
+		in = kvzalloc(inlen, GFP_KERNEL);
+		if (!in)
+			return -ENOMEM;
+
+		qp->create_type = MLX5_QP_EMPTY;
+	}
+
+	if (is_sqp(init_attr->qp_type))
+		qp->port = init_attr->port_num;
+
+	qpc = MLX5_ADDR_OF(create_qp_in, in, qpc);
+
+	MLX5_SET(qpc, qpc, st, mlx5_st);
+	MLX5_SET(qpc, qpc, pm_state, MLX5_QP_PM_MIGRATED);
+
+	if (init_attr->qp_type != MLX5_IB_QPT_REG_UMR)
+		MLX5_SET(qpc, qpc, pd, to_mpd(pd ? pd : devr->p0)->pdn);
+	else
+		MLX5_SET(qpc, qpc, latency_sensitive, 1);
+
+
+	if (qp->wq_sig)
+		MLX5_SET(qpc, qpc, wq_signature, 1);
+
+	if (qp->flags & MLX5_IB_QP_BLOCK_MULTICAST_LOOPBACK)
+		MLX5_SET(qpc, qpc, block_lb_mc, 1);
+
+	if (qp->flags & MLX5_IB_QP_CROSS_CHANNEL)
+		MLX5_SET(qpc, qpc, cd_master, 1);
+	if (qp->flags & MLX5_IB_QP_MANAGED_SEND)
+		MLX5_SET(qpc, qpc, cd_slave_send, 1);
+	if (qp->flags & MLX5_IB_QP_MANAGED_RECV)
+		MLX5_SET(qpc, qpc, cd_slave_receive, 1);
+
+	if (qp->scat_cqe && is_connected(init_attr->qp_type)) {
+		int rcqe_sz;
+		int scqe_sz;
+
+		rcqe_sz = mlx5_ib_get_cqe_size(dev, init_attr->recv_cq);
+		scqe_sz = mlx5_ib_get_cqe_size(dev, init_attr->send_cq);
+
+		if (rcqe_sz == 128)
+			MLX5_SET(qpc, qpc, cs_res, MLX5_RES_SCAT_DATA64_CQE);
+		else
+			MLX5_SET(qpc, qpc, cs_res, MLX5_RES_SCAT_DATA32_CQE);
+
+		if (init_attr->sq_sig_type == IB_SIGNAL_ALL_WR) {
+			if (scqe_sz == 128)
+				MLX5_SET(qpc, qpc, cs_req, MLX5_REQ_SCAT_DATA64_CQE);
+			else
+				MLX5_SET(qpc, qpc, cs_req, MLX5_REQ_SCAT_DATA32_CQE);
+		}
+	}
+
+	if (qp->rq.wqe_cnt) {
+		MLX5_SET(qpc, qpc, log_rq_stride, qp->rq.wqe_shift - 4);
+		MLX5_SET(qpc, qpc, log_rq_size, ilog2(qp->rq.wqe_cnt));
+	}
+
+	MLX5_SET(qpc, qpc, rq_type, get_rx_type(qp, init_attr));
+
+	if (qp->sq.wqe_cnt) {
+		MLX5_SET(qpc, qpc, log_sq_size, ilog2(qp->sq.wqe_cnt));
+	} else {
+		MLX5_SET(qpc, qpc, no_sq, 1);
+		if (init_attr->srq &&
+		    init_attr->srq->srq_type == IB_SRQT_TM)
+			MLX5_SET(qpc, qpc, offload_type,
+				 MLX5_QPC_OFFLOAD_TYPE_RNDV);
+	}
+
+	/* Set default resources */
+	switch (init_attr->qp_type) {
+	case IB_QPT_XRC_TGT:
+		MLX5_SET(qpc, qpc, cqn_rcv, to_mcq(devr->c0)->mcq.cqn);
+		MLX5_SET(qpc, qpc, cqn_snd, to_mcq(devr->c0)->mcq.cqn);
+		MLX5_SET(qpc, qpc, srqn_rmpn_xrqn, to_msrq(devr->s0)->msrq.srqn);
+		MLX5_SET(qpc, qpc, xrcd, to_mxrcd(init_attr->xrcd)->xrcdn);
+		break;
+	case IB_QPT_XRC_INI:
+		MLX5_SET(qpc, qpc, cqn_rcv, to_mcq(devr->c0)->mcq.cqn);
+		MLX5_SET(qpc, qpc, xrcd, to_mxrcd(devr->x1)->xrcdn);
+		MLX5_SET(qpc, qpc, srqn_rmpn_xrqn, to_msrq(devr->s0)->msrq.srqn);
+		break;
+	default:
+		if (init_attr->srq) {
+			MLX5_SET(qpc, qpc, xrcd, to_mxrcd(devr->x0)->xrcdn);
+			MLX5_SET(qpc, qpc, srqn_rmpn_xrqn, to_msrq(init_attr->srq)->msrq.srqn);
+		} else {
+			MLX5_SET(qpc, qpc, xrcd, to_mxrcd(devr->x1)->xrcdn);
+			MLX5_SET(qpc, qpc, srqn_rmpn_xrqn, to_msrq(devr->s1)->msrq.srqn);
+		}
+	}
+
+	if (init_attr->send_cq)
+		MLX5_SET(qpc, qpc, cqn_snd, to_mcq(init_attr->send_cq)->mcq.cqn);
+
+	if (init_attr->recv_cq)
+		MLX5_SET(qpc, qpc, cqn_rcv, to_mcq(init_attr->recv_cq)->mcq.cqn);
+
+	MLX5_SET64(qpc, qpc, dbr_addr, qp->db.dma);
+
+	/* 0xffffff means we ask to work with cqe version 0 */
+	if (MLX5_CAP_GEN(mdev, cqe_version) == MLX5_CQE_VERSION_V1)
+		MLX5_SET(qpc, qpc, user_index, uidx);
+
+	/* we use IB_QP_CREATE_IPOIB_UD_LSO to indicates ipoib qp */
+	if (init_attr->qp_type == IB_QPT_UD &&
+	    (init_attr->create_flags & IB_QP_CREATE_IPOIB_UD_LSO)) {
+		MLX5_SET(qpc, qpc, ulp_stateless_offload_mode, 1);
+		qp->flags |= MLX5_IB_QP_LSO;
+	}
+
+	if (init_attr->create_flags & IB_QP_CREATE_PCI_WRITE_END_PADDING) {
+		if (!MLX5_CAP_GEN(dev->mdev, end_pad)) {
+			mlx5_ib_dbg(dev, "scatter end padding is not supported\n");
+			err = -EOPNOTSUPP;
+			goto err;
+		} else if (init_attr->qp_type != IB_QPT_RAW_PACKET) {
+			MLX5_SET(qpc, qpc, end_padding_mode,
+				 MLX5_WQ_END_PAD_MODE_ALIGN);
+		} else {
+			qp->flags |= MLX5_IB_QP_PCI_WRITE_END_PADDING;
+		}
+	}
+
+	if (inlen < 0) {
+		err = -EINVAL;
+		goto err;
+	}
+
+	if (init_attr->qp_type == IB_QPT_RAW_PACKET ||
+	    qp->flags & MLX5_IB_QP_UNDERLAY) {
+		qp->raw_packet_qp.sq.ubuffer.buf_addr = ucmd.sq_buf_addr;
+		raw_packet_qp_copy_info(qp, &qp->raw_packet_qp);
+		err = create_raw_packet_qp(dev, qp, in, inlen, pd);
+	} else {
+		err = mlx5_core_create_qp(dev->mdev, &base->mqp, in, inlen);
+	}
+
+	if (err) {
+		mlx5_ib_dbg(dev, "create qp failed\n");
+		goto err_create;
+	}
+
+	kvfree(in);
+
+	base->container_mibqp = qp;
+	base->mqp.event = mlx5_ib_qp_event;
+
+	get_cqs(init_attr->qp_type, init_attr->send_cq, init_attr->recv_cq,
+		&send_cq, &recv_cq);
+	spin_lock_irqsave(&dev->reset_flow_resource_lock, flags);
+	mlx5_ib_lock_cqs(send_cq, recv_cq);
+	/* Maintain device to QPs access, needed for further handling via reset
+	 * flow
+	 */
+	list_add_tail(&qp->qps_list, &dev->qp_list);
+	/* Maintain CQ to QPs access, needed for further handling via reset flow
+	 */
+	if (send_cq)
+		list_add_tail(&qp->cq_send_list, &send_cq->list_send_qp);
+	if (recv_cq)
+		list_add_tail(&qp->cq_recv_list, &recv_cq->list_recv_qp);
+	mlx5_ib_unlock_cqs(send_cq, recv_cq);
+	spin_unlock_irqrestore(&dev->reset_flow_resource_lock, flags);
+
+	return 0;
+
+err_create:
+	if (qp->create_type == MLX5_QP_USER)
+		destroy_qp_user(dev, pd, qp, base);
+	else if (qp->create_type == MLX5_QP_KERNEL)
+		destroy_qp_kernel(dev, qp);
+
+err:
+	kvfree(in);
+	return err;
+}
+
+static void mlx5_ib_lock_cqs(struct mlx5_ib_cq *send_cq, struct mlx5_ib_cq *recv_cq)
+	__acquires(&send_cq->lock) __acquires(&recv_cq->lock)
+{
+	if (send_cq) {
+		if (recv_cq) {
+			if (send_cq->mcq.cqn < recv_cq->mcq.cqn)  {
+				spin_lock(&send_cq->lock);
+				spin_lock_nested(&recv_cq->lock,
+						 SINGLE_DEPTH_NESTING);
+			} else if (send_cq->mcq.cqn == recv_cq->mcq.cqn) {
+				spin_lock(&send_cq->lock);
+				__acquire(&recv_cq->lock);
+			} else {
+				spin_lock(&recv_cq->lock);
+				spin_lock_nested(&send_cq->lock,
+						 SINGLE_DEPTH_NESTING);
+			}
+		} else {
+			spin_lock(&send_cq->lock);
+			__acquire(&recv_cq->lock);
+		}
+	} else if (recv_cq) {
+		spin_lock(&recv_cq->lock);
+		__acquire(&send_cq->lock);
+	} else {
+		__acquire(&send_cq->lock);
+		__acquire(&recv_cq->lock);
+	}
+}
+
+static void mlx5_ib_unlock_cqs(struct mlx5_ib_cq *send_cq, struct mlx5_ib_cq *recv_cq)
+	__releases(&send_cq->lock) __releases(&recv_cq->lock)
+{
+	if (send_cq) {
+		if (recv_cq) {
+			if (send_cq->mcq.cqn < recv_cq->mcq.cqn)  {
+				spin_unlock(&recv_cq->lock);
+				spin_unlock(&send_cq->lock);
+			} else if (send_cq->mcq.cqn == recv_cq->mcq.cqn) {
+				__release(&recv_cq->lock);
+				spin_unlock(&send_cq->lock);
+			} else {
+				spin_unlock(&send_cq->lock);
+				spin_unlock(&recv_cq->lock);
+			}
+		} else {
+			__release(&recv_cq->lock);
+			spin_unlock(&send_cq->lock);
+		}
+	} else if (recv_cq) {
+		__release(&send_cq->lock);
+		spin_unlock(&recv_cq->lock);
+	} else {
+		__release(&recv_cq->lock);
+		__release(&send_cq->lock);
+	}
+}
+
+static struct mlx5_ib_pd *get_pd(struct mlx5_ib_qp *qp)
+{
+	return to_mpd(qp->ibqp.pd);
+}
+
+static void get_cqs(enum ib_qp_type qp_type,
+		    struct ib_cq *ib_send_cq, struct ib_cq *ib_recv_cq,
+		    struct mlx5_ib_cq **send_cq, struct mlx5_ib_cq **recv_cq)
+{
+	switch (qp_type) {
+	case IB_QPT_XRC_TGT:
+		*send_cq = NULL;
+		*recv_cq = NULL;
+		break;
+	case MLX5_IB_QPT_REG_UMR:
+	case IB_QPT_XRC_INI:
+		*send_cq = ib_send_cq ? to_mcq(ib_send_cq) : NULL;
+		*recv_cq = NULL;
+		break;
+
+	case IB_QPT_SMI:
+	case MLX5_IB_QPT_HW_GSI:
+	case IB_QPT_RC:
+	case IB_QPT_UC:
+	case IB_QPT_UD:
+	case IB_QPT_RAW_IPV6:
+	case IB_QPT_RAW_ETHERTYPE:
+	case IB_QPT_RAW_PACKET:
+		*send_cq = ib_send_cq ? to_mcq(ib_send_cq) : NULL;
+		*recv_cq = ib_recv_cq ? to_mcq(ib_recv_cq) : NULL;
+		break;
+
+	case IB_QPT_MAX:
+	default:
+		*send_cq = NULL;
+		*recv_cq = NULL;
+		break;
+	}
+}
+
+static int modify_raw_packet_qp(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
+				const struct mlx5_modify_raw_qp_param *raw_qp_param,
+				u8 lag_tx_affinity);
+
+static void destroy_qp_common(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp)
+{
+	struct mlx5_ib_cq *send_cq, *recv_cq;
+	struct mlx5_ib_qp_base *base;
+	unsigned long flags;
+	int err;
+
+	if (qp->ibqp.rwq_ind_tbl) {
+		destroy_rss_raw_qp_tir(dev, qp);
+		return;
+	}
+
+	base = (qp->ibqp.qp_type == IB_QPT_RAW_PACKET ||
+		qp->flags & MLX5_IB_QP_UNDERLAY) ?
+	       &qp->raw_packet_qp.rq.base :
+	       &qp->trans_qp.base;
+
+	if (qp->state != IB_QPS_RESET) {
+		if (qp->ibqp.qp_type != IB_QPT_RAW_PACKET &&
+		    !(qp->flags & MLX5_IB_QP_UNDERLAY)) {
+			err = mlx5_core_qp_modify(dev->mdev,
+						  MLX5_CMD_OP_2RST_QP, 0,
+						  NULL, &base->mqp);
+		} else {
+			struct mlx5_modify_raw_qp_param raw_qp_param = {
+				.operation = MLX5_CMD_OP_2RST_QP
+			};
+
+			err = modify_raw_packet_qp(dev, qp, &raw_qp_param, 0);
+		}
+		if (err)
+			mlx5_ib_warn(dev, "mlx5_ib: modify QP 0x%06x to RESET failed\n",
+				     base->mqp.qpn);
+	}
+
+	get_cqs(qp->ibqp.qp_type, qp->ibqp.send_cq, qp->ibqp.recv_cq,
+		&send_cq, &recv_cq);
+
+	spin_lock_irqsave(&dev->reset_flow_resource_lock, flags);
+	mlx5_ib_lock_cqs(send_cq, recv_cq);
+	/* del from lists under both locks above to protect reset flow paths */
+	list_del(&qp->qps_list);
+	if (send_cq)
+		list_del(&qp->cq_send_list);
+
+	if (recv_cq)
+		list_del(&qp->cq_recv_list);
+
+	if (qp->create_type == MLX5_QP_KERNEL) {
+		__mlx5_ib_cq_clean(recv_cq, base->mqp.qpn,
+				   qp->ibqp.srq ? to_msrq(qp->ibqp.srq) : NULL);
+		if (send_cq != recv_cq)
+			__mlx5_ib_cq_clean(send_cq, base->mqp.qpn,
+					   NULL);
+	}
+	mlx5_ib_unlock_cqs(send_cq, recv_cq);
+	spin_unlock_irqrestore(&dev->reset_flow_resource_lock, flags);
+
+	if (qp->ibqp.qp_type == IB_QPT_RAW_PACKET ||
+	    qp->flags & MLX5_IB_QP_UNDERLAY) {
+		destroy_raw_packet_qp(dev, qp);
+	} else {
+		err = mlx5_core_destroy_qp(dev->mdev, &base->mqp);
+		if (err)
+			mlx5_ib_warn(dev, "failed to destroy QP 0x%x\n",
+				     base->mqp.qpn);
+	}
+
+	if (qp->create_type == MLX5_QP_KERNEL)
+		destroy_qp_kernel(dev, qp);
+	else if (qp->create_type == MLX5_QP_USER)
+		destroy_qp_user(dev, &get_pd(qp)->ibpd, qp, base);
+}
+
+static const char *ib_qp_type_str(enum ib_qp_type type)
+{
+	switch (type) {
+	case IB_QPT_SMI:
+		return "IB_QPT_SMI";
+	case IB_QPT_GSI:
+		return "IB_QPT_GSI";
+	case IB_QPT_RC:
+		return "IB_QPT_RC";
+	case IB_QPT_UC:
+		return "IB_QPT_UC";
+	case IB_QPT_UD:
+		return "IB_QPT_UD";
+	case IB_QPT_RAW_IPV6:
+		return "IB_QPT_RAW_IPV6";
+	case IB_QPT_RAW_ETHERTYPE:
+		return "IB_QPT_RAW_ETHERTYPE";
+	case IB_QPT_XRC_INI:
+		return "IB_QPT_XRC_INI";
+	case IB_QPT_XRC_TGT:
+		return "IB_QPT_XRC_TGT";
+	case IB_QPT_RAW_PACKET:
+		return "IB_QPT_RAW_PACKET";
+	case MLX5_IB_QPT_REG_UMR:
+		return "MLX5_IB_QPT_REG_UMR";
+	case IB_QPT_DRIVER:
+		return "IB_QPT_DRIVER";
+	case IB_QPT_MAX:
+	default:
+		return "Invalid QP type";
+	}
+}
+
+static struct ib_qp *mlx5_ib_create_dct(struct ib_pd *pd,
+					struct ib_qp_init_attr *attr,
+					struct mlx5_ib_create_qp *ucmd)
+{
+	struct mlx5_ib_qp *qp;
+	int err = 0;
+	u32 uidx = MLX5_IB_DEFAULT_UIDX;
+	void *dctc;
+
+	if (!attr->srq || !attr->recv_cq)
+		return ERR_PTR(-EINVAL);
+
+	err = get_qp_user_index(to_mucontext(pd->uobject->context),
+				ucmd, sizeof(*ucmd), &uidx);
+	if (err)
+		return ERR_PTR(err);
+
+	qp = kzalloc(sizeof(*qp), GFP_KERNEL);
+	if (!qp)
+		return ERR_PTR(-ENOMEM);
+
+	qp->dct.in = kzalloc(MLX5_ST_SZ_BYTES(create_dct_in), GFP_KERNEL);
+	if (!qp->dct.in) {
+		err = -ENOMEM;
+		goto err_free;
+	}
+
+	dctc = MLX5_ADDR_OF(create_dct_in, qp->dct.in, dct_context_entry);
+	qp->qp_sub_type = MLX5_IB_QPT_DCT;
+	MLX5_SET(dctc, dctc, pd, to_mpd(pd)->pdn);
+	MLX5_SET(dctc, dctc, srqn_xrqn, to_msrq(attr->srq)->msrq.srqn);
+	MLX5_SET(dctc, dctc, cqn, to_mcq(attr->recv_cq)->mcq.cqn);
+	MLX5_SET64(dctc, dctc, dc_access_key, ucmd->access_key);
+	MLX5_SET(dctc, dctc, user_index, uidx);
+
+	qp->state = IB_QPS_RESET;
+
+	return &qp->ibqp;
+err_free:
+	kfree(qp);
+	return ERR_PTR(err);
+}
+
+static int set_mlx_qp_type(struct mlx5_ib_dev *dev,
+			   struct ib_qp_init_attr *init_attr,
+			   struct mlx5_ib_create_qp *ucmd,
+			   struct ib_udata *udata)
+{
+	enum { MLX_QP_FLAGS = MLX5_QP_FLAG_TYPE_DCT | MLX5_QP_FLAG_TYPE_DCI };
+	int err;
+
+	if (!udata)
+		return -EINVAL;
+
+	if (udata->inlen < sizeof(*ucmd)) {
+		mlx5_ib_dbg(dev, "create_qp user command is smaller than expected\n");
+		return -EINVAL;
+	}
+	err = ib_copy_from_udata(ucmd, udata, sizeof(*ucmd));
+	if (err)
+		return err;
+
+	if ((ucmd->flags & MLX_QP_FLAGS) == MLX5_QP_FLAG_TYPE_DCI) {
+		init_attr->qp_type = MLX5_IB_QPT_DCI;
+	} else {
+		if ((ucmd->flags & MLX_QP_FLAGS) == MLX5_QP_FLAG_TYPE_DCT) {
+			init_attr->qp_type = MLX5_IB_QPT_DCT;
+		} else {
+			mlx5_ib_dbg(dev, "Invalid QP flags\n");
+			return -EINVAL;
+		}
+	}
+
+	if (!MLX5_CAP_GEN(dev->mdev, dct)) {
+		mlx5_ib_dbg(dev, "DC transport is not supported\n");
+		return -EOPNOTSUPP;
+	}
+
+	return 0;
+}
+
+struct ib_qp *mlx5_ib_create_qp(struct ib_pd *pd,
+				struct ib_qp_init_attr *verbs_init_attr,
+				struct ib_udata *udata)
+{
+	struct mlx5_ib_dev *dev;
+	struct mlx5_ib_qp *qp;
+	u16 xrcdn = 0;
+	int err;
+	struct ib_qp_init_attr mlx_init_attr;
+	struct ib_qp_init_attr *init_attr = verbs_init_attr;
+
+	if (pd) {
+		dev = to_mdev(pd->device);
+
+		if (init_attr->qp_type == IB_QPT_RAW_PACKET) {
+			if (!pd->uobject) {
+				mlx5_ib_dbg(dev, "Raw Packet QP is not supported for kernel consumers\n");
+				return ERR_PTR(-EINVAL);
+			} else if (!to_mucontext(pd->uobject->context)->cqe_version) {
+				mlx5_ib_dbg(dev, "Raw Packet QP is only supported for CQE version > 0\n");
+				return ERR_PTR(-EINVAL);
+			}
+		}
+	} else {
+		/* being cautious here */
+		if (init_attr->qp_type != IB_QPT_XRC_TGT &&
+		    init_attr->qp_type != MLX5_IB_QPT_REG_UMR) {
+			pr_warn("%s: no PD for transport %s\n", __func__,
+				ib_qp_type_str(init_attr->qp_type));
+			return ERR_PTR(-EINVAL);
+		}
+		dev = to_mdev(to_mxrcd(init_attr->xrcd)->ibxrcd.device);
+	}
+
+	if (init_attr->qp_type == IB_QPT_DRIVER) {
+		struct mlx5_ib_create_qp ucmd;
+
+		init_attr = &mlx_init_attr;
+		memcpy(init_attr, verbs_init_attr, sizeof(*verbs_init_attr));
+		err = set_mlx_qp_type(dev, init_attr, &ucmd, udata);
+		if (err)
+			return ERR_PTR(err);
+
+		if (init_attr->qp_type == MLX5_IB_QPT_DCI) {
+			if (init_attr->cap.max_recv_wr ||
+			    init_attr->cap.max_recv_sge) {
+				mlx5_ib_dbg(dev, "DCI QP requires zero size receive queue\n");
+				return ERR_PTR(-EINVAL);
+			}
+		} else {
+			return mlx5_ib_create_dct(pd, init_attr, &ucmd);
+		}
+	}
+
+	switch (init_attr->qp_type) {
+	case IB_QPT_XRC_TGT:
+	case IB_QPT_XRC_INI:
+		if (!MLX5_CAP_GEN(dev->mdev, xrc)) {
+			mlx5_ib_dbg(dev, "XRC not supported\n");
+			return ERR_PTR(-ENOSYS);
+		}
+		init_attr->recv_cq = NULL;
+		if (init_attr->qp_type == IB_QPT_XRC_TGT) {
+			xrcdn = to_mxrcd(init_attr->xrcd)->xrcdn;
+			init_attr->send_cq = NULL;
+		}
+
+		/* fall through */
+	case IB_QPT_RAW_PACKET:
+	case IB_QPT_RC:
+	case IB_QPT_UC:
+	case IB_QPT_UD:
+	case IB_QPT_SMI:
+	case MLX5_IB_QPT_HW_GSI:
+	case MLX5_IB_QPT_REG_UMR:
+	case MLX5_IB_QPT_DCI:
+		qp = kzalloc(sizeof(*qp), GFP_KERNEL);
+		if (!qp)
+			return ERR_PTR(-ENOMEM);
+
+		err = create_qp_common(dev, pd, init_attr, udata, qp);
+		if (err) {
+			mlx5_ib_dbg(dev, "create_qp_common failed\n");
+			kfree(qp);
+			return ERR_PTR(err);
+		}
+
+		if (is_qp0(init_attr->qp_type))
+			qp->ibqp.qp_num = 0;
+		else if (is_qp1(init_attr->qp_type))
+			qp->ibqp.qp_num = 1;
+		else
+			qp->ibqp.qp_num = qp->trans_qp.base.mqp.qpn;
+
+		mlx5_ib_dbg(dev, "ib qpnum 0x%x, mlx qpn 0x%x, rcqn 0x%x, scqn 0x%x\n",
+			    qp->ibqp.qp_num, qp->trans_qp.base.mqp.qpn,
+			    init_attr->recv_cq ? to_mcq(init_attr->recv_cq)->mcq.cqn : -1,
+			    init_attr->send_cq ? to_mcq(init_attr->send_cq)->mcq.cqn : -1);
+
+		qp->trans_qp.xrcdn = xrcdn;
+
+		break;
+
+	case IB_QPT_GSI:
+		return mlx5_ib_gsi_create_qp(pd, init_attr);
+
+	case IB_QPT_RAW_IPV6:
+	case IB_QPT_RAW_ETHERTYPE:
+	case IB_QPT_MAX:
+	default:
+		mlx5_ib_dbg(dev, "unsupported qp type %d\n",
+			    init_attr->qp_type);
+		/* Don't support raw QPs */
+		return ERR_PTR(-EINVAL);
+	}
+
+	if (verbs_init_attr->qp_type == IB_QPT_DRIVER)
+		qp->qp_sub_type = init_attr->qp_type;
+
+	return &qp->ibqp;
+}
+
+static int mlx5_ib_destroy_dct(struct mlx5_ib_qp *mqp)
+{
+	struct mlx5_ib_dev *dev = to_mdev(mqp->ibqp.device);
+
+	if (mqp->state == IB_QPS_RTR) {
+		int err;
+
+		err = mlx5_core_destroy_dct(dev->mdev, &mqp->dct.mdct);
+		if (err) {
+			mlx5_ib_warn(dev, "failed to destroy DCT %d\n", err);
+			return err;
+		}
+	}
+
+	kfree(mqp->dct.in);
+	kfree(mqp);
+	return 0;
+}
+
+int mlx5_ib_destroy_qp(struct ib_qp *qp)
+{
+	struct mlx5_ib_dev *dev = to_mdev(qp->device);
+	struct mlx5_ib_qp *mqp = to_mqp(qp);
+
+	if (unlikely(qp->qp_type == IB_QPT_GSI))
+		return mlx5_ib_gsi_destroy_qp(qp);
+
+	if (mqp->qp_sub_type == MLX5_IB_QPT_DCT)
+		return mlx5_ib_destroy_dct(mqp);
+
+	destroy_qp_common(dev, mqp);
+
+	kfree(mqp);
+
+	return 0;
+}
+
+static __be32 to_mlx5_access_flags(struct mlx5_ib_qp *qp, const struct ib_qp_attr *attr,
+				   int attr_mask)
+{
+	u32 hw_access_flags = 0;
+	u8 dest_rd_atomic;
+	u32 access_flags;
+
+	if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC)
+		dest_rd_atomic = attr->max_dest_rd_atomic;
+	else
+		dest_rd_atomic = qp->trans_qp.resp_depth;
+
+	if (attr_mask & IB_QP_ACCESS_FLAGS)
+		access_flags = attr->qp_access_flags;
+	else
+		access_flags = qp->trans_qp.atomic_rd_en;
+
+	if (!dest_rd_atomic)
+		access_flags &= IB_ACCESS_REMOTE_WRITE;
+
+	if (access_flags & IB_ACCESS_REMOTE_READ)
+		hw_access_flags |= MLX5_QP_BIT_RRE;
+	if (access_flags & IB_ACCESS_REMOTE_ATOMIC)
+		hw_access_flags |= (MLX5_QP_BIT_RAE | MLX5_ATOMIC_MODE_CX);
+	if (access_flags & IB_ACCESS_REMOTE_WRITE)
+		hw_access_flags |= MLX5_QP_BIT_RWE;
+
+	return cpu_to_be32(hw_access_flags);
+}
+
+enum {
+	MLX5_PATH_FLAG_FL	= 1 << 0,
+	MLX5_PATH_FLAG_FREE_AR	= 1 << 1,
+	MLX5_PATH_FLAG_COUNTER	= 1 << 2,
+};
+
+static int ib_rate_to_mlx5(struct mlx5_ib_dev *dev, u8 rate)
+{
+	if (rate == IB_RATE_PORT_CURRENT)
+		return 0;
+
+	if (rate < IB_RATE_2_5_GBPS || rate > IB_RATE_300_GBPS)
+		return -EINVAL;
+
+	while (rate != IB_RATE_PORT_CURRENT &&
+	       !(1 << (rate + MLX5_STAT_RATE_OFFSET) &
+		 MLX5_CAP_GEN(dev->mdev, stat_rate_support)))
+		--rate;
+
+	return rate ? rate + MLX5_STAT_RATE_OFFSET : rate;
+}
+
+static int modify_raw_packet_eth_prio(struct mlx5_core_dev *dev,
+				      struct mlx5_ib_sq *sq, u8 sl)
+{
+	void *in;
+	void *tisc;
+	int inlen;
+	int err;
+
+	inlen = MLX5_ST_SZ_BYTES(modify_tis_in);
+	in = kvzalloc(inlen, GFP_KERNEL);
+	if (!in)
+		return -ENOMEM;
+
+	MLX5_SET(modify_tis_in, in, bitmask.prio, 1);
+
+	tisc = MLX5_ADDR_OF(modify_tis_in, in, ctx);
+	MLX5_SET(tisc, tisc, prio, ((sl & 0x7) << 1));
+
+	err = mlx5_core_modify_tis(dev, sq->tisn, in, inlen);
+
+	kvfree(in);
+
+	return err;
+}
+
+static int modify_raw_packet_tx_affinity(struct mlx5_core_dev *dev,
+					 struct mlx5_ib_sq *sq, u8 tx_affinity)
+{
+	void *in;
+	void *tisc;
+	int inlen;
+	int err;
+
+	inlen = MLX5_ST_SZ_BYTES(modify_tis_in);
+	in = kvzalloc(inlen, GFP_KERNEL);
+	if (!in)
+		return -ENOMEM;
+
+	MLX5_SET(modify_tis_in, in, bitmask.lag_tx_port_affinity, 1);
+
+	tisc = MLX5_ADDR_OF(modify_tis_in, in, ctx);
+	MLX5_SET(tisc, tisc, lag_tx_port_affinity, tx_affinity);
+
+	err = mlx5_core_modify_tis(dev, sq->tisn, in, inlen);
+
+	kvfree(in);
+
+	return err;
+}
+
+static int mlx5_set_path(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
+			 const struct rdma_ah_attr *ah,
+			 struct mlx5_qp_path *path, u8 port, int attr_mask,
+			 u32 path_flags, const struct ib_qp_attr *attr,
+			 bool alt)
+{
+	const struct ib_global_route *grh = rdma_ah_read_grh(ah);
+	int err;
+	enum ib_gid_type gid_type;
+	u8 ah_flags = rdma_ah_get_ah_flags(ah);
+	u8 sl = rdma_ah_get_sl(ah);
+
+	if (attr_mask & IB_QP_PKEY_INDEX)
+		path->pkey_index = cpu_to_be16(alt ? attr->alt_pkey_index :
+						     attr->pkey_index);
+
+	if (ah_flags & IB_AH_GRH) {
+		if (grh->sgid_index >=
+		    dev->mdev->port_caps[port - 1].gid_table_len) {
+			pr_err("sgid_index (%u) too large. max is %d\n",
+			       grh->sgid_index,
+			       dev->mdev->port_caps[port - 1].gid_table_len);
+			return -EINVAL;
+		}
+	}
+
+	if (ah->type == RDMA_AH_ATTR_TYPE_ROCE) {
+		if (!(ah_flags & IB_AH_GRH))
+			return -EINVAL;
+
+		memcpy(path->rmac, ah->roce.dmac, sizeof(ah->roce.dmac));
+		if (qp->ibqp.qp_type == IB_QPT_RC ||
+		    qp->ibqp.qp_type == IB_QPT_UC ||
+		    qp->ibqp.qp_type == IB_QPT_XRC_INI ||
+		    qp->ibqp.qp_type == IB_QPT_XRC_TGT)
+			path->udp_sport =
+				mlx5_get_roce_udp_sport(dev, ah->grh.sgid_attr);
+		path->dci_cfi_prio_sl = (sl & 0x7) << 4;
+		gid_type = ah->grh.sgid_attr->gid_type;
+		if (gid_type == IB_GID_TYPE_ROCE_UDP_ENCAP)
+			path->ecn_dscp = (grh->traffic_class >> 2) & 0x3f;
+	} else {
+		path->fl_free_ar = (path_flags & MLX5_PATH_FLAG_FL) ? 0x80 : 0;
+		path->fl_free_ar |=
+			(path_flags & MLX5_PATH_FLAG_FREE_AR) ? 0x40 : 0;
+		path->rlid = cpu_to_be16(rdma_ah_get_dlid(ah));
+		path->grh_mlid = rdma_ah_get_path_bits(ah) & 0x7f;
+		if (ah_flags & IB_AH_GRH)
+			path->grh_mlid	|= 1 << 7;
+		path->dci_cfi_prio_sl = sl & 0xf;
+	}
+
+	if (ah_flags & IB_AH_GRH) {
+		path->mgid_index = grh->sgid_index;
+		path->hop_limit  = grh->hop_limit;
+		path->tclass_flowlabel =
+			cpu_to_be32((grh->traffic_class << 20) |
+				    (grh->flow_label));
+		memcpy(path->rgid, grh->dgid.raw, 16);
+	}
+
+	err = ib_rate_to_mlx5(dev, rdma_ah_get_static_rate(ah));
+	if (err < 0)
+		return err;
+	path->static_rate = err;
+	path->port = port;
+
+	if (attr_mask & IB_QP_TIMEOUT)
+		path->ackto_lt = (alt ? attr->alt_timeout : attr->timeout) << 3;
+
+	if ((qp->ibqp.qp_type == IB_QPT_RAW_PACKET) && qp->sq.wqe_cnt)
+		return modify_raw_packet_eth_prio(dev->mdev,
+						  &qp->raw_packet_qp.sq,
+						  sl & 0xf);
+
+	return 0;
+}
+
+static enum mlx5_qp_optpar opt_mask[MLX5_QP_NUM_STATE][MLX5_QP_NUM_STATE][MLX5_QP_ST_MAX] = {
+	[MLX5_QP_STATE_INIT] = {
+		[MLX5_QP_STATE_INIT] = {
+			[MLX5_QP_ST_RC] = MLX5_QP_OPTPAR_RRE		|
+					  MLX5_QP_OPTPAR_RAE		|
+					  MLX5_QP_OPTPAR_RWE		|
+					  MLX5_QP_OPTPAR_PKEY_INDEX	|
+					  MLX5_QP_OPTPAR_PRI_PORT,
+			[MLX5_QP_ST_UC] = MLX5_QP_OPTPAR_RWE		|
+					  MLX5_QP_OPTPAR_PKEY_INDEX	|
+					  MLX5_QP_OPTPAR_PRI_PORT,
+			[MLX5_QP_ST_UD] = MLX5_QP_OPTPAR_PKEY_INDEX	|
+					  MLX5_QP_OPTPAR_Q_KEY		|
+					  MLX5_QP_OPTPAR_PRI_PORT,
+			[MLX5_QP_ST_XRC] = MLX5_QP_OPTPAR_RRE		|
+					  MLX5_QP_OPTPAR_RAE		|
+					  MLX5_QP_OPTPAR_RWE		|
+					  MLX5_QP_OPTPAR_PKEY_INDEX	|
+					  MLX5_QP_OPTPAR_PRI_PORT,
+		},
+		[MLX5_QP_STATE_RTR] = {
+			[MLX5_QP_ST_RC] = MLX5_QP_OPTPAR_ALT_ADDR_PATH  |
+					  MLX5_QP_OPTPAR_RRE            |
+					  MLX5_QP_OPTPAR_RAE            |
+					  MLX5_QP_OPTPAR_RWE            |
+					  MLX5_QP_OPTPAR_PKEY_INDEX,
+			[MLX5_QP_ST_UC] = MLX5_QP_OPTPAR_ALT_ADDR_PATH  |
+					  MLX5_QP_OPTPAR_RWE            |
+					  MLX5_QP_OPTPAR_PKEY_INDEX,
+			[MLX5_QP_ST_UD] = MLX5_QP_OPTPAR_PKEY_INDEX     |
+					  MLX5_QP_OPTPAR_Q_KEY,
+			[MLX5_QP_ST_MLX] = MLX5_QP_OPTPAR_PKEY_INDEX	|
+					   MLX5_QP_OPTPAR_Q_KEY,
+			[MLX5_QP_ST_XRC] = MLX5_QP_OPTPAR_ALT_ADDR_PATH |
+					  MLX5_QP_OPTPAR_RRE            |
+					  MLX5_QP_OPTPAR_RAE            |
+					  MLX5_QP_OPTPAR_RWE            |
+					  MLX5_QP_OPTPAR_PKEY_INDEX,
+		},
+	},
+	[MLX5_QP_STATE_RTR] = {
+		[MLX5_QP_STATE_RTS] = {
+			[MLX5_QP_ST_RC] = MLX5_QP_OPTPAR_ALT_ADDR_PATH	|
+					  MLX5_QP_OPTPAR_RRE		|
+					  MLX5_QP_OPTPAR_RAE		|
+					  MLX5_QP_OPTPAR_RWE		|
+					  MLX5_QP_OPTPAR_PM_STATE	|
+					  MLX5_QP_OPTPAR_RNR_TIMEOUT,
+			[MLX5_QP_ST_UC] = MLX5_QP_OPTPAR_ALT_ADDR_PATH	|
+					  MLX5_QP_OPTPAR_RWE		|
+					  MLX5_QP_OPTPAR_PM_STATE,
+			[MLX5_QP_ST_UD] = MLX5_QP_OPTPAR_Q_KEY,
+			[MLX5_QP_ST_XRC] = MLX5_QP_OPTPAR_ALT_ADDR_PATH	|
+					  MLX5_QP_OPTPAR_RRE		|
+					  MLX5_QP_OPTPAR_RAE		|
+					  MLX5_QP_OPTPAR_RWE		|
+					  MLX5_QP_OPTPAR_PM_STATE	|
+					  MLX5_QP_OPTPAR_RNR_TIMEOUT,
+		},
+	},
+	[MLX5_QP_STATE_RTS] = {
+		[MLX5_QP_STATE_RTS] = {
+			[MLX5_QP_ST_RC] = MLX5_QP_OPTPAR_RRE		|
+					  MLX5_QP_OPTPAR_RAE		|
+					  MLX5_QP_OPTPAR_RWE		|
+					  MLX5_QP_OPTPAR_RNR_TIMEOUT	|
+					  MLX5_QP_OPTPAR_PM_STATE	|
+					  MLX5_QP_OPTPAR_ALT_ADDR_PATH,
+			[MLX5_QP_ST_UC] = MLX5_QP_OPTPAR_RWE		|
+					  MLX5_QP_OPTPAR_PM_STATE	|
+					  MLX5_QP_OPTPAR_ALT_ADDR_PATH,
+			[MLX5_QP_ST_UD] = MLX5_QP_OPTPAR_Q_KEY		|
+					  MLX5_QP_OPTPAR_SRQN		|
+					  MLX5_QP_OPTPAR_CQN_RCV,
+			[MLX5_QP_ST_XRC] = MLX5_QP_OPTPAR_RRE		|
+					  MLX5_QP_OPTPAR_RAE		|
+					  MLX5_QP_OPTPAR_RWE		|
+					  MLX5_QP_OPTPAR_RNR_TIMEOUT	|
+					  MLX5_QP_OPTPAR_PM_STATE	|
+					  MLX5_QP_OPTPAR_ALT_ADDR_PATH,
+		},
+	},
+	[MLX5_QP_STATE_SQER] = {
+		[MLX5_QP_STATE_RTS] = {
+			[MLX5_QP_ST_UD]	 = MLX5_QP_OPTPAR_Q_KEY,
+			[MLX5_QP_ST_MLX] = MLX5_QP_OPTPAR_Q_KEY,
+			[MLX5_QP_ST_UC]	 = MLX5_QP_OPTPAR_RWE,
+			[MLX5_QP_ST_RC]	 = MLX5_QP_OPTPAR_RNR_TIMEOUT	|
+					   MLX5_QP_OPTPAR_RWE		|
+					   MLX5_QP_OPTPAR_RAE		|
+					   MLX5_QP_OPTPAR_RRE,
+			[MLX5_QP_ST_XRC]  = MLX5_QP_OPTPAR_RNR_TIMEOUT	|
+					   MLX5_QP_OPTPAR_RWE		|
+					   MLX5_QP_OPTPAR_RAE		|
+					   MLX5_QP_OPTPAR_RRE,
+		},
+	},
+};
+
+static int ib_nr_to_mlx5_nr(int ib_mask)
+{
+	switch (ib_mask) {
+	case IB_QP_STATE:
+		return 0;
+	case IB_QP_CUR_STATE:
+		return 0;
+	case IB_QP_EN_SQD_ASYNC_NOTIFY:
+		return 0;
+	case IB_QP_ACCESS_FLAGS:
+		return MLX5_QP_OPTPAR_RWE | MLX5_QP_OPTPAR_RRE |
+			MLX5_QP_OPTPAR_RAE;
+	case IB_QP_PKEY_INDEX:
+		return MLX5_QP_OPTPAR_PKEY_INDEX;
+	case IB_QP_PORT:
+		return MLX5_QP_OPTPAR_PRI_PORT;
+	case IB_QP_QKEY:
+		return MLX5_QP_OPTPAR_Q_KEY;
+	case IB_QP_AV:
+		return MLX5_QP_OPTPAR_PRIMARY_ADDR_PATH |
+			MLX5_QP_OPTPAR_PRI_PORT;
+	case IB_QP_PATH_MTU:
+		return 0;
+	case IB_QP_TIMEOUT:
+		return MLX5_QP_OPTPAR_ACK_TIMEOUT;
+	case IB_QP_RETRY_CNT:
+		return MLX5_QP_OPTPAR_RETRY_COUNT;
+	case IB_QP_RNR_RETRY:
+		return MLX5_QP_OPTPAR_RNR_RETRY;
+	case IB_QP_RQ_PSN:
+		return 0;
+	case IB_QP_MAX_QP_RD_ATOMIC:
+		return MLX5_QP_OPTPAR_SRA_MAX;
+	case IB_QP_ALT_PATH:
+		return MLX5_QP_OPTPAR_ALT_ADDR_PATH;
+	case IB_QP_MIN_RNR_TIMER:
+		return MLX5_QP_OPTPAR_RNR_TIMEOUT;
+	case IB_QP_SQ_PSN:
+		return 0;
+	case IB_QP_MAX_DEST_RD_ATOMIC:
+		return MLX5_QP_OPTPAR_RRA_MAX | MLX5_QP_OPTPAR_RWE |
+			MLX5_QP_OPTPAR_RRE | MLX5_QP_OPTPAR_RAE;
+	case IB_QP_PATH_MIG_STATE:
+		return MLX5_QP_OPTPAR_PM_STATE;
+	case IB_QP_CAP:
+		return 0;
+	case IB_QP_DEST_QPN:
+		return 0;
+	}
+	return 0;
+}
+
+static int ib_mask_to_mlx5_opt(int ib_mask)
+{
+	int result = 0;
+	int i;
+
+	for (i = 0; i < 8 * sizeof(int); i++) {
+		if ((1 << i) & ib_mask)
+			result |= ib_nr_to_mlx5_nr(1 << i);
+	}
+
+	return result;
+}
+
+static int modify_raw_packet_qp_rq(struct mlx5_ib_dev *dev,
+				   struct mlx5_ib_rq *rq, int new_state,
+				   const struct mlx5_modify_raw_qp_param *raw_qp_param)
+{
+	void *in;
+	void *rqc;
+	int inlen;
+	int err;
+
+	inlen = MLX5_ST_SZ_BYTES(modify_rq_in);
+	in = kvzalloc(inlen, GFP_KERNEL);
+	if (!in)
+		return -ENOMEM;
+
+	MLX5_SET(modify_rq_in, in, rq_state, rq->state);
+
+	rqc = MLX5_ADDR_OF(modify_rq_in, in, ctx);
+	MLX5_SET(rqc, rqc, state, new_state);
+
+	if (raw_qp_param->set_mask & MLX5_RAW_QP_MOD_SET_RQ_Q_CTR_ID) {
+		if (MLX5_CAP_GEN(dev->mdev, modify_rq_counter_set_id)) {
+			MLX5_SET64(modify_rq_in, in, modify_bitmask,
+				   MLX5_MODIFY_RQ_IN_MODIFY_BITMASK_RQ_COUNTER_SET_ID);
+			MLX5_SET(rqc, rqc, counter_set_id, raw_qp_param->rq_q_ctr_id);
+		} else
+			pr_info_once("%s: RAW PACKET QP counters are not supported on current FW\n",
+				     dev->ib_dev.name);
+	}
+
+	err = mlx5_core_modify_rq(dev->mdev, rq->base.mqp.qpn, in, inlen);
+	if (err)
+		goto out;
+
+	rq->state = new_state;
+
+out:
+	kvfree(in);
+	return err;
+}
+
+static int modify_raw_packet_qp_sq(struct mlx5_core_dev *dev,
+				   struct mlx5_ib_sq *sq,
+				   int new_state,
+				   const struct mlx5_modify_raw_qp_param *raw_qp_param)
+{
+	struct mlx5_ib_qp *ibqp = sq->base.container_mibqp;
+	struct mlx5_rate_limit old_rl = ibqp->rl;
+	struct mlx5_rate_limit new_rl = old_rl;
+	bool new_rate_added = false;
+	u16 rl_index = 0;
+	void *in;
+	void *sqc;
+	int inlen;
+	int err;
+
+	inlen = MLX5_ST_SZ_BYTES(modify_sq_in);
+	in = kvzalloc(inlen, GFP_KERNEL);
+	if (!in)
+		return -ENOMEM;
+
+	MLX5_SET(modify_sq_in, in, sq_state, sq->state);
+
+	sqc = MLX5_ADDR_OF(modify_sq_in, in, ctx);
+	MLX5_SET(sqc, sqc, state, new_state);
+
+	if (raw_qp_param->set_mask & MLX5_RAW_QP_RATE_LIMIT) {
+		if (new_state != MLX5_SQC_STATE_RDY)
+			pr_warn("%s: Rate limit can only be changed when SQ is moving to RDY\n",
+				__func__);
+		else
+			new_rl = raw_qp_param->rl;
+	}
+
+	if (!mlx5_rl_are_equal(&old_rl, &new_rl)) {
+		if (new_rl.rate) {
+			err = mlx5_rl_add_rate(dev, &rl_index, &new_rl);
+			if (err) {
+				pr_err("Failed configuring rate limit(err %d): \
+				       rate %u, max_burst_sz %u, typical_pkt_sz %u\n",
+				       err, new_rl.rate, new_rl.max_burst_sz,
+				       new_rl.typical_pkt_sz);
+
+				goto out;
+			}
+			new_rate_added = true;
+		}
+
+		MLX5_SET64(modify_sq_in, in, modify_bitmask, 1);
+		/* index 0 means no limit */
+		MLX5_SET(sqc, sqc, packet_pacing_rate_limit_index, rl_index);
+	}
+
+	err = mlx5_core_modify_sq(dev, sq->base.mqp.qpn, in, inlen);
+	if (err) {
+		/* Remove new rate from table if failed */
+		if (new_rate_added)
+			mlx5_rl_remove_rate(dev, &new_rl);
+		goto out;
+	}
+
+	/* Only remove the old rate after new rate was set */
+	if ((old_rl.rate && !mlx5_rl_are_equal(&old_rl, &new_rl)) ||
+	    (new_state != MLX5_SQC_STATE_RDY)) {
+		mlx5_rl_remove_rate(dev, &old_rl);
+		if (new_state != MLX5_SQC_STATE_RDY)
+			memset(&new_rl, 0, sizeof(new_rl));
+	}
+
+	ibqp->rl = new_rl;
+	sq->state = new_state;
+
+out:
+	kvfree(in);
+	return err;
+}
+
+static int modify_raw_packet_qp(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
+				const struct mlx5_modify_raw_qp_param *raw_qp_param,
+				u8 tx_affinity)
+{
+	struct mlx5_ib_raw_packet_qp *raw_packet_qp = &qp->raw_packet_qp;
+	struct mlx5_ib_rq *rq = &raw_packet_qp->rq;
+	struct mlx5_ib_sq *sq = &raw_packet_qp->sq;
+	int modify_rq = !!qp->rq.wqe_cnt;
+	int modify_sq = !!qp->sq.wqe_cnt;
+	int rq_state;
+	int sq_state;
+	int err;
+
+	switch (raw_qp_param->operation) {
+	case MLX5_CMD_OP_RST2INIT_QP:
+		rq_state = MLX5_RQC_STATE_RDY;
+		sq_state = MLX5_SQC_STATE_RDY;
+		break;
+	case MLX5_CMD_OP_2ERR_QP:
+		rq_state = MLX5_RQC_STATE_ERR;
+		sq_state = MLX5_SQC_STATE_ERR;
+		break;
+	case MLX5_CMD_OP_2RST_QP:
+		rq_state = MLX5_RQC_STATE_RST;
+		sq_state = MLX5_SQC_STATE_RST;
+		break;
+	case MLX5_CMD_OP_RTR2RTS_QP:
+	case MLX5_CMD_OP_RTS2RTS_QP:
+		if (raw_qp_param->set_mask ==
+		    MLX5_RAW_QP_RATE_LIMIT) {
+			modify_rq = 0;
+			sq_state = sq->state;
+		} else {
+			return raw_qp_param->set_mask ? -EINVAL : 0;
+		}
+		break;
+	case MLX5_CMD_OP_INIT2INIT_QP:
+	case MLX5_CMD_OP_INIT2RTR_QP:
+		if (raw_qp_param->set_mask)
+			return -EINVAL;
+		else
+			return 0;
+	default:
+		WARN_ON(1);
+		return -EINVAL;
+	}
+
+	if (modify_rq) {
+		err =  modify_raw_packet_qp_rq(dev, rq, rq_state, raw_qp_param);
+		if (err)
+			return err;
+	}
+
+	if (modify_sq) {
+		if (tx_affinity) {
+			err = modify_raw_packet_tx_affinity(dev->mdev, sq,
+							    tx_affinity);
+			if (err)
+				return err;
+		}
+
+		return modify_raw_packet_qp_sq(dev->mdev, sq, sq_state, raw_qp_param);
+	}
+
+	return 0;
+}
+
+static unsigned int get_tx_affinity(struct mlx5_ib_dev *dev,
+				    struct mlx5_ib_pd *pd,
+				    struct mlx5_ib_qp_base *qp_base,
+				    u8 port_num)
+{
+	struct mlx5_ib_ucontext *ucontext = NULL;
+	unsigned int tx_port_affinity;
+
+	if (pd && pd->ibpd.uobject && pd->ibpd.uobject->context)
+		ucontext = to_mucontext(pd->ibpd.uobject->context);
+
+	if (ucontext) {
+		tx_port_affinity = (unsigned int)atomic_add_return(
+					   1, &ucontext->tx_port_affinity) %
+					   MLX5_MAX_PORTS +
+				   1;
+		mlx5_ib_dbg(dev, "Set tx affinity 0x%x to qpn 0x%x ucontext %p\n",
+				tx_port_affinity, qp_base->mqp.qpn, ucontext);
+	} else {
+		tx_port_affinity =
+			(unsigned int)atomic_add_return(
+				1, &dev->roce[port_num].tx_port_affinity) %
+				MLX5_MAX_PORTS +
+			1;
+		mlx5_ib_dbg(dev, "Set tx affinity 0x%x to qpn 0x%x\n",
+				tx_port_affinity, qp_base->mqp.qpn);
+	}
+
+	return tx_port_affinity;
+}
+
+static int __mlx5_ib_modify_qp(struct ib_qp *ibqp,
+			       const struct ib_qp_attr *attr, int attr_mask,
+			       enum ib_qp_state cur_state, enum ib_qp_state new_state,
+			       const struct mlx5_ib_modify_qp *ucmd)
+{
+	static const u16 optab[MLX5_QP_NUM_STATE][MLX5_QP_NUM_STATE] = {
+		[MLX5_QP_STATE_RST] = {
+			[MLX5_QP_STATE_RST]	= MLX5_CMD_OP_2RST_QP,
+			[MLX5_QP_STATE_ERR]	= MLX5_CMD_OP_2ERR_QP,
+			[MLX5_QP_STATE_INIT]	= MLX5_CMD_OP_RST2INIT_QP,
+		},
+		[MLX5_QP_STATE_INIT]  = {
+			[MLX5_QP_STATE_RST]	= MLX5_CMD_OP_2RST_QP,
+			[MLX5_QP_STATE_ERR]	= MLX5_CMD_OP_2ERR_QP,
+			[MLX5_QP_STATE_INIT]	= MLX5_CMD_OP_INIT2INIT_QP,
+			[MLX5_QP_STATE_RTR]	= MLX5_CMD_OP_INIT2RTR_QP,
+		},
+		[MLX5_QP_STATE_RTR]   = {
+			[MLX5_QP_STATE_RST]	= MLX5_CMD_OP_2RST_QP,
+			[MLX5_QP_STATE_ERR]	= MLX5_CMD_OP_2ERR_QP,
+			[MLX5_QP_STATE_RTS]	= MLX5_CMD_OP_RTR2RTS_QP,
+		},
+		[MLX5_QP_STATE_RTS]   = {
+			[MLX5_QP_STATE_RST]	= MLX5_CMD_OP_2RST_QP,
+			[MLX5_QP_STATE_ERR]	= MLX5_CMD_OP_2ERR_QP,
+			[MLX5_QP_STATE_RTS]	= MLX5_CMD_OP_RTS2RTS_QP,
+		},
+		[MLX5_QP_STATE_SQD] = {
+			[MLX5_QP_STATE_RST]	= MLX5_CMD_OP_2RST_QP,
+			[MLX5_QP_STATE_ERR]	= MLX5_CMD_OP_2ERR_QP,
+		},
+		[MLX5_QP_STATE_SQER] = {
+			[MLX5_QP_STATE_RST]	= MLX5_CMD_OP_2RST_QP,
+			[MLX5_QP_STATE_ERR]	= MLX5_CMD_OP_2ERR_QP,
+			[MLX5_QP_STATE_RTS]	= MLX5_CMD_OP_SQERR2RTS_QP,
+		},
+		[MLX5_QP_STATE_ERR] = {
+			[MLX5_QP_STATE_RST]	= MLX5_CMD_OP_2RST_QP,
+			[MLX5_QP_STATE_ERR]	= MLX5_CMD_OP_2ERR_QP,
+		}
+	};
+
+	struct mlx5_ib_dev *dev = to_mdev(ibqp->device);
+	struct mlx5_ib_qp *qp = to_mqp(ibqp);
+	struct mlx5_ib_qp_base *base = &qp->trans_qp.base;
+	struct mlx5_ib_cq *send_cq, *recv_cq;
+	struct mlx5_qp_context *context;
+	struct mlx5_ib_pd *pd;
+	struct mlx5_ib_port *mibport = NULL;
+	enum mlx5_qp_state mlx5_cur, mlx5_new;
+	enum mlx5_qp_optpar optpar;
+	int mlx5_st;
+	int err;
+	u16 op;
+	u8 tx_affinity = 0;
+
+	mlx5_st = to_mlx5_st(ibqp->qp_type == IB_QPT_DRIVER ?
+			     qp->qp_sub_type : ibqp->qp_type);
+	if (mlx5_st < 0)
+		return -EINVAL;
+
+	context = kzalloc(sizeof(*context), GFP_KERNEL);
+	if (!context)
+		return -ENOMEM;
+
+	pd = get_pd(qp);
+	context->flags = cpu_to_be32(mlx5_st << 16);
+
+	if (!(attr_mask & IB_QP_PATH_MIG_STATE)) {
+		context->flags |= cpu_to_be32(MLX5_QP_PM_MIGRATED << 11);
+	} else {
+		switch (attr->path_mig_state) {
+		case IB_MIG_MIGRATED:
+			context->flags |= cpu_to_be32(MLX5_QP_PM_MIGRATED << 11);
+			break;
+		case IB_MIG_REARM:
+			context->flags |= cpu_to_be32(MLX5_QP_PM_REARM << 11);
+			break;
+		case IB_MIG_ARMED:
+			context->flags |= cpu_to_be32(MLX5_QP_PM_ARMED << 11);
+			break;
+		}
+	}
+
+	if ((cur_state == IB_QPS_RESET) && (new_state == IB_QPS_INIT)) {
+		if ((ibqp->qp_type == IB_QPT_RC) ||
+		    (ibqp->qp_type == IB_QPT_UD &&
+		     !(qp->flags & MLX5_IB_QP_SQPN_QP1)) ||
+		    (ibqp->qp_type == IB_QPT_UC) ||
+		    (ibqp->qp_type == IB_QPT_RAW_PACKET) ||
+		    (ibqp->qp_type == IB_QPT_XRC_INI) ||
+		    (ibqp->qp_type == IB_QPT_XRC_TGT)) {
+			if (mlx5_lag_is_active(dev->mdev)) {
+				u8 p = mlx5_core_native_port_num(dev->mdev);
+				tx_affinity = get_tx_affinity(dev, pd, base, p);
+				context->flags |= cpu_to_be32(tx_affinity << 24);
+			}
+		}
+	}
+
+	if (is_sqp(ibqp->qp_type)) {
+		context->mtu_msgmax = (IB_MTU_256 << 5) | 8;
+	} else if ((ibqp->qp_type == IB_QPT_UD &&
+		    !(qp->flags & MLX5_IB_QP_UNDERLAY)) ||
+		   ibqp->qp_type == MLX5_IB_QPT_REG_UMR) {
+		context->mtu_msgmax = (IB_MTU_4096 << 5) | 12;
+	} else if (attr_mask & IB_QP_PATH_MTU) {
+		if (attr->path_mtu < IB_MTU_256 ||
+		    attr->path_mtu > IB_MTU_4096) {
+			mlx5_ib_warn(dev, "invalid mtu %d\n", attr->path_mtu);
+			err = -EINVAL;
+			goto out;
+		}
+		context->mtu_msgmax = (attr->path_mtu << 5) |
+				      (u8)MLX5_CAP_GEN(dev->mdev, log_max_msg);
+	}
+
+	if (attr_mask & IB_QP_DEST_QPN)
+		context->log_pg_sz_remote_qpn = cpu_to_be32(attr->dest_qp_num);
+
+	if (attr_mask & IB_QP_PKEY_INDEX)
+		context->pri_path.pkey_index = cpu_to_be16(attr->pkey_index);
+
+	/* todo implement counter_index functionality */
+
+	if (is_sqp(ibqp->qp_type))
+		context->pri_path.port = qp->port;
+
+	if (attr_mask & IB_QP_PORT)
+		context->pri_path.port = attr->port_num;
+
+	if (attr_mask & IB_QP_AV) {
+		err = mlx5_set_path(dev, qp, &attr->ah_attr, &context->pri_path,
+				    attr_mask & IB_QP_PORT ? attr->port_num : qp->port,
+				    attr_mask, 0, attr, false);
+		if (err)
+			goto out;
+	}
+
+	if (attr_mask & IB_QP_TIMEOUT)
+		context->pri_path.ackto_lt |= attr->timeout << 3;
+
+	if (attr_mask & IB_QP_ALT_PATH) {
+		err = mlx5_set_path(dev, qp, &attr->alt_ah_attr,
+				    &context->alt_path,
+				    attr->alt_port_num,
+				    attr_mask | IB_QP_PKEY_INDEX | IB_QP_TIMEOUT,
+				    0, attr, true);
+		if (err)
+			goto out;
+	}
+
+	get_cqs(qp->ibqp.qp_type, qp->ibqp.send_cq, qp->ibqp.recv_cq,
+		&send_cq, &recv_cq);
+
+	context->flags_pd = cpu_to_be32(pd ? pd->pdn : to_mpd(dev->devr.p0)->pdn);
+	context->cqn_send = send_cq ? cpu_to_be32(send_cq->mcq.cqn) : 0;
+	context->cqn_recv = recv_cq ? cpu_to_be32(recv_cq->mcq.cqn) : 0;
+	context->params1  = cpu_to_be32(MLX5_IB_ACK_REQ_FREQ << 28);
+
+	if (attr_mask & IB_QP_RNR_RETRY)
+		context->params1 |= cpu_to_be32(attr->rnr_retry << 13);
+
+	if (attr_mask & IB_QP_RETRY_CNT)
+		context->params1 |= cpu_to_be32(attr->retry_cnt << 16);
+
+	if (attr_mask & IB_QP_MAX_QP_RD_ATOMIC) {
+		if (attr->max_rd_atomic)
+			context->params1 |=
+				cpu_to_be32(fls(attr->max_rd_atomic - 1) << 21);
+	}
+
+	if (attr_mask & IB_QP_SQ_PSN)
+		context->next_send_psn = cpu_to_be32(attr->sq_psn);
+
+	if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC) {
+		if (attr->max_dest_rd_atomic)
+			context->params2 |=
+				cpu_to_be32(fls(attr->max_dest_rd_atomic - 1) << 21);
+	}
+
+	if (attr_mask & (IB_QP_ACCESS_FLAGS | IB_QP_MAX_DEST_RD_ATOMIC))
+		context->params2 |= to_mlx5_access_flags(qp, attr, attr_mask);
+
+	if (attr_mask & IB_QP_MIN_RNR_TIMER)
+		context->rnr_nextrecvpsn |= cpu_to_be32(attr->min_rnr_timer << 24);
+
+	if (attr_mask & IB_QP_RQ_PSN)
+		context->rnr_nextrecvpsn |= cpu_to_be32(attr->rq_psn);
+
+	if (attr_mask & IB_QP_QKEY)
+		context->qkey = cpu_to_be32(attr->qkey);
+
+	if (qp->rq.wqe_cnt && cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT)
+		context->db_rec_addr = cpu_to_be64(qp->db.dma);
+
+	if (cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT) {
+		u8 port_num = (attr_mask & IB_QP_PORT ? attr->port_num :
+			       qp->port) - 1;
+
+		/* Underlay port should be used - index 0 function per port */
+		if (qp->flags & MLX5_IB_QP_UNDERLAY)
+			port_num = 0;
+
+		mibport = &dev->port[port_num];
+		context->qp_counter_set_usr_page |=
+			cpu_to_be32((u32)(mibport->cnts.set_id) << 24);
+	}
+
+	if (!ibqp->uobject && cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT)
+		context->sq_crq_size |= cpu_to_be16(1 << 4);
+
+	if (qp->flags & MLX5_IB_QP_SQPN_QP1)
+		context->deth_sqpn = cpu_to_be32(1);
+
+	mlx5_cur = to_mlx5_state(cur_state);
+	mlx5_new = to_mlx5_state(new_state);
+
+	if (mlx5_cur >= MLX5_QP_NUM_STATE || mlx5_new >= MLX5_QP_NUM_STATE ||
+	    !optab[mlx5_cur][mlx5_new]) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	op = optab[mlx5_cur][mlx5_new];
+	optpar = ib_mask_to_mlx5_opt(attr_mask);
+	optpar &= opt_mask[mlx5_cur][mlx5_new][mlx5_st];
+
+	if (qp->ibqp.qp_type == IB_QPT_RAW_PACKET ||
+	    qp->flags & MLX5_IB_QP_UNDERLAY) {
+		struct mlx5_modify_raw_qp_param raw_qp_param = {};
+
+		raw_qp_param.operation = op;
+		if (cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT) {
+			raw_qp_param.rq_q_ctr_id = mibport->cnts.set_id;
+			raw_qp_param.set_mask |= MLX5_RAW_QP_MOD_SET_RQ_Q_CTR_ID;
+		}
+
+		if (attr_mask & IB_QP_RATE_LIMIT) {
+			raw_qp_param.rl.rate = attr->rate_limit;
+
+			if (ucmd->burst_info.max_burst_sz) {
+				if (attr->rate_limit &&
+				    MLX5_CAP_QOS(dev->mdev, packet_pacing_burst_bound)) {
+					raw_qp_param.rl.max_burst_sz =
+						ucmd->burst_info.max_burst_sz;
+				} else {
+					err = -EINVAL;
+					goto out;
+				}
+			}
+
+			if (ucmd->burst_info.typical_pkt_sz) {
+				if (attr->rate_limit &&
+				    MLX5_CAP_QOS(dev->mdev, packet_pacing_typical_size)) {
+					raw_qp_param.rl.typical_pkt_sz =
+						ucmd->burst_info.typical_pkt_sz;
+				} else {
+					err = -EINVAL;
+					goto out;
+				}
+			}
+
+			raw_qp_param.set_mask |= MLX5_RAW_QP_RATE_LIMIT;
+		}
+
+		err = modify_raw_packet_qp(dev, qp, &raw_qp_param, tx_affinity);
+	} else {
+		err = mlx5_core_qp_modify(dev->mdev, op, optpar, context,
+					  &base->mqp);
+	}
+
+	if (err)
+		goto out;
+
+	qp->state = new_state;
+
+	if (attr_mask & IB_QP_ACCESS_FLAGS)
+		qp->trans_qp.atomic_rd_en = attr->qp_access_flags;
+	if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC)
+		qp->trans_qp.resp_depth = attr->max_dest_rd_atomic;
+	if (attr_mask & IB_QP_PORT)
+		qp->port = attr->port_num;
+	if (attr_mask & IB_QP_ALT_PATH)
+		qp->trans_qp.alt_port = attr->alt_port_num;
+
+	/*
+	 * If we moved a kernel QP to RESET, clean up all old CQ
+	 * entries and reinitialize the QP.
+	 */
+	if (new_state == IB_QPS_RESET &&
+	    !ibqp->uobject && ibqp->qp_type != IB_QPT_XRC_TGT) {
+		mlx5_ib_cq_clean(recv_cq, base->mqp.qpn,
+				 ibqp->srq ? to_msrq(ibqp->srq) : NULL);
+		if (send_cq != recv_cq)
+			mlx5_ib_cq_clean(send_cq, base->mqp.qpn, NULL);
+
+		qp->rq.head = 0;
+		qp->rq.tail = 0;
+		qp->sq.head = 0;
+		qp->sq.tail = 0;
+		qp->sq.cur_post = 0;
+		qp->sq.last_poll = 0;
+		qp->db.db[MLX5_RCV_DBR] = 0;
+		qp->db.db[MLX5_SND_DBR] = 0;
+	}
+
+out:
+	kfree(context);
+	return err;
+}
+
+static inline bool is_valid_mask(int mask, int req, int opt)
+{
+	if ((mask & req) != req)
+		return false;
+
+	if (mask & ~(req | opt))
+		return false;
+
+	return true;
+}
+
+/* check valid transition for driver QP types
+ * for now the only QP type that this function supports is DCI
+ */
+static bool modify_dci_qp_is_ok(enum ib_qp_state cur_state, enum ib_qp_state new_state,
+				enum ib_qp_attr_mask attr_mask)
+{
+	int req = IB_QP_STATE;
+	int opt = 0;
+
+	if (new_state == IB_QPS_RESET) {
+		return is_valid_mask(attr_mask, req, opt);
+	} else if (cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT) {
+		req |= IB_QP_PKEY_INDEX | IB_QP_PORT;
+		return is_valid_mask(attr_mask, req, opt);
+	} else if (cur_state == IB_QPS_INIT && new_state == IB_QPS_INIT) {
+		opt = IB_QP_PKEY_INDEX | IB_QP_PORT;
+		return is_valid_mask(attr_mask, req, opt);
+	} else if (cur_state == IB_QPS_INIT && new_state == IB_QPS_RTR) {
+		req |= IB_QP_PATH_MTU;
+		opt = IB_QP_PKEY_INDEX;
+		return is_valid_mask(attr_mask, req, opt);
+	} else if (cur_state == IB_QPS_RTR && new_state == IB_QPS_RTS) {
+		req |= IB_QP_TIMEOUT | IB_QP_RETRY_CNT | IB_QP_RNR_RETRY |
+		       IB_QP_MAX_QP_RD_ATOMIC | IB_QP_SQ_PSN;
+		opt = IB_QP_MIN_RNR_TIMER;
+		return is_valid_mask(attr_mask, req, opt);
+	} else if (cur_state == IB_QPS_RTS && new_state == IB_QPS_RTS) {
+		opt = IB_QP_MIN_RNR_TIMER;
+		return is_valid_mask(attr_mask, req, opt);
+	} else if (cur_state != IB_QPS_RESET && new_state == IB_QPS_ERR) {
+		return is_valid_mask(attr_mask, req, opt);
+	}
+	return false;
+}
+
+/* mlx5_ib_modify_dct: modify a DCT QP
+ * valid transitions are:
+ * RESET to INIT: must set access_flags, pkey_index and port
+ * INIT  to RTR : must set min_rnr_timer, tclass, flow_label,
+ *			   mtu, gid_index and hop_limit
+ * Other transitions and attributes are illegal
+ */
+static int mlx5_ib_modify_dct(struct ib_qp *ibqp, struct ib_qp_attr *attr,
+			      int attr_mask, struct ib_udata *udata)
+{
+	struct mlx5_ib_qp *qp = to_mqp(ibqp);
+	struct mlx5_ib_dev *dev = to_mdev(ibqp->device);
+	enum ib_qp_state cur_state, new_state;
+	int err = 0;
+	int required = IB_QP_STATE;
+	void *dctc;
+
+	if (!(attr_mask & IB_QP_STATE))
+		return -EINVAL;
+
+	cur_state = qp->state;
+	new_state = attr->qp_state;
+
+	dctc = MLX5_ADDR_OF(create_dct_in, qp->dct.in, dct_context_entry);
+	if (cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT) {
+		required |= IB_QP_ACCESS_FLAGS | IB_QP_PKEY_INDEX | IB_QP_PORT;
+		if (!is_valid_mask(attr_mask, required, 0))
+			return -EINVAL;
+
+		if (attr->port_num == 0 ||
+		    attr->port_num > MLX5_CAP_GEN(dev->mdev, num_ports)) {
+			mlx5_ib_dbg(dev, "invalid port number %d. number of ports is %d\n",
+				    attr->port_num, dev->num_ports);
+			return -EINVAL;
+		}
+		if (attr->qp_access_flags & IB_ACCESS_REMOTE_READ)
+			MLX5_SET(dctc, dctc, rre, 1);
+		if (attr->qp_access_flags & IB_ACCESS_REMOTE_WRITE)
+			MLX5_SET(dctc, dctc, rwe, 1);
+		if (attr->qp_access_flags & IB_ACCESS_REMOTE_ATOMIC) {
+			if (!mlx5_ib_dc_atomic_is_supported(dev))
+				return -EOPNOTSUPP;
+			MLX5_SET(dctc, dctc, rae, 1);
+			MLX5_SET(dctc, dctc, atomic_mode, MLX5_ATOMIC_MODE_DCT_CX);
+		}
+		MLX5_SET(dctc, dctc, pkey_index, attr->pkey_index);
+		MLX5_SET(dctc, dctc, port, attr->port_num);
+		MLX5_SET(dctc, dctc, counter_set_id, dev->port[attr->port_num - 1].cnts.set_id);
+
+	} else if (cur_state == IB_QPS_INIT && new_state == IB_QPS_RTR) {
+		struct mlx5_ib_modify_qp_resp resp = {};
+		u32 min_resp_len = offsetof(typeof(resp), dctn) +
+				   sizeof(resp.dctn);
+
+		if (udata->outlen < min_resp_len)
+			return -EINVAL;
+		resp.response_length = min_resp_len;
+
+		required |= IB_QP_MIN_RNR_TIMER | IB_QP_AV | IB_QP_PATH_MTU;
+		if (!is_valid_mask(attr_mask, required, 0))
+			return -EINVAL;
+		MLX5_SET(dctc, dctc, min_rnr_nak, attr->min_rnr_timer);
+		MLX5_SET(dctc, dctc, tclass, attr->ah_attr.grh.traffic_class);
+		MLX5_SET(dctc, dctc, flow_label, attr->ah_attr.grh.flow_label);
+		MLX5_SET(dctc, dctc, mtu, attr->path_mtu);
+		MLX5_SET(dctc, dctc, my_addr_index, attr->ah_attr.grh.sgid_index);
+		MLX5_SET(dctc, dctc, hop_limit, attr->ah_attr.grh.hop_limit);
+		if (attr->ah_attr.type == RDMA_AH_ATTR_TYPE_ROCE)
+			MLX5_SET(dctc, dctc, eth_prio, attr->ah_attr.sl & 0x7);
+
+		err = mlx5_core_create_dct(dev->mdev, &qp->dct.mdct, qp->dct.in,
+					   MLX5_ST_SZ_BYTES(create_dct_in));
+		if (err)
+			return err;
+		resp.dctn = qp->dct.mdct.mqp.qpn;
+		err = ib_copy_to_udata(udata, &resp, resp.response_length);
+		if (err) {
+			mlx5_core_destroy_dct(dev->mdev, &qp->dct.mdct);
+			return err;
+		}
+	} else {
+		mlx5_ib_warn(dev, "Modify DCT: Invalid transition from %d to %d\n", cur_state, new_state);
+		return -EINVAL;
+	}
+	if (err)
+		qp->state = IB_QPS_ERR;
+	else
+		qp->state = new_state;
+	return err;
+}
+
+int mlx5_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
+		      int attr_mask, struct ib_udata *udata)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibqp->device);
+	struct mlx5_ib_qp *qp = to_mqp(ibqp);
+	struct mlx5_ib_modify_qp ucmd = {};
+	enum ib_qp_type qp_type;
+	enum ib_qp_state cur_state, new_state;
+	size_t required_cmd_sz;
+	int err = -EINVAL;
+	int port;
+	enum rdma_link_layer ll = IB_LINK_LAYER_UNSPECIFIED;
+
+	if (ibqp->rwq_ind_tbl)
+		return -ENOSYS;
+
+	if (udata && udata->inlen) {
+		required_cmd_sz = offsetof(typeof(ucmd), reserved) +
+			sizeof(ucmd.reserved);
+		if (udata->inlen < required_cmd_sz)
+			return -EINVAL;
+
+		if (udata->inlen > sizeof(ucmd) &&
+		    !ib_is_udata_cleared(udata, sizeof(ucmd),
+					 udata->inlen - sizeof(ucmd)))
+			return -EOPNOTSUPP;
+
+		if (ib_copy_from_udata(&ucmd, udata,
+				       min(udata->inlen, sizeof(ucmd))))
+			return -EFAULT;
+
+		if (ucmd.comp_mask ||
+		    memchr_inv(&ucmd.reserved, 0, sizeof(ucmd.reserved)) ||
+		    memchr_inv(&ucmd.burst_info.reserved, 0,
+			       sizeof(ucmd.burst_info.reserved)))
+			return -EOPNOTSUPP;
+	}
+
+	if (unlikely(ibqp->qp_type == IB_QPT_GSI))
+		return mlx5_ib_gsi_modify_qp(ibqp, attr, attr_mask);
+
+	if (ibqp->qp_type == IB_QPT_DRIVER)
+		qp_type = qp->qp_sub_type;
+	else
+		qp_type = (unlikely(ibqp->qp_type == MLX5_IB_QPT_HW_GSI)) ?
+			IB_QPT_GSI : ibqp->qp_type;
+
+	if (qp_type == MLX5_IB_QPT_DCT)
+		return mlx5_ib_modify_dct(ibqp, attr, attr_mask, udata);
+
+	mutex_lock(&qp->mutex);
+
+	cur_state = attr_mask & IB_QP_CUR_STATE ? attr->cur_qp_state : qp->state;
+	new_state = attr_mask & IB_QP_STATE ? attr->qp_state : cur_state;
+
+	if (!(cur_state == new_state && cur_state == IB_QPS_RESET)) {
+		port = attr_mask & IB_QP_PORT ? attr->port_num : qp->port;
+		ll = dev->ib_dev.get_link_layer(&dev->ib_dev, port);
+	}
+
+	if (qp->flags & MLX5_IB_QP_UNDERLAY) {
+		if (attr_mask & ~(IB_QP_STATE | IB_QP_CUR_STATE)) {
+			mlx5_ib_dbg(dev, "invalid attr_mask 0x%x when underlay QP is used\n",
+				    attr_mask);
+			goto out;
+		}
+	} else if (qp_type != MLX5_IB_QPT_REG_UMR &&
+		   qp_type != MLX5_IB_QPT_DCI &&
+		   !ib_modify_qp_is_ok(cur_state, new_state, qp_type, attr_mask, ll)) {
+		mlx5_ib_dbg(dev, "invalid QP state transition from %d to %d, qp_type %d, attr_mask 0x%x\n",
+			    cur_state, new_state, ibqp->qp_type, attr_mask);
+		goto out;
+	} else if (qp_type == MLX5_IB_QPT_DCI &&
+		   !modify_dci_qp_is_ok(cur_state, new_state, attr_mask)) {
+		mlx5_ib_dbg(dev, "invalid QP state transition from %d to %d, qp_type %d, attr_mask 0x%x\n",
+			    cur_state, new_state, qp_type, attr_mask);
+		goto out;
+	}
+
+	if ((attr_mask & IB_QP_PORT) &&
+	    (attr->port_num == 0 ||
+	     attr->port_num > dev->num_ports)) {
+		mlx5_ib_dbg(dev, "invalid port number %d. number of ports is %d\n",
+			    attr->port_num, dev->num_ports);
+		goto out;
+	}
+
+	if (attr_mask & IB_QP_PKEY_INDEX) {
+		port = attr_mask & IB_QP_PORT ? attr->port_num : qp->port;
+		if (attr->pkey_index >=
+		    dev->mdev->port_caps[port - 1].pkey_table_len) {
+			mlx5_ib_dbg(dev, "invalid pkey index %d\n",
+				    attr->pkey_index);
+			goto out;
+		}
+	}
+
+	if (attr_mask & IB_QP_MAX_QP_RD_ATOMIC &&
+	    attr->max_rd_atomic >
+	    (1 << MLX5_CAP_GEN(dev->mdev, log_max_ra_res_qp))) {
+		mlx5_ib_dbg(dev, "invalid max_rd_atomic value %d\n",
+			    attr->max_rd_atomic);
+		goto out;
+	}
+
+	if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC &&
+	    attr->max_dest_rd_atomic >
+	    (1 << MLX5_CAP_GEN(dev->mdev, log_max_ra_req_qp))) {
+		mlx5_ib_dbg(dev, "invalid max_dest_rd_atomic value %d\n",
+			    attr->max_dest_rd_atomic);
+		goto out;
+	}
+
+	if (cur_state == new_state && cur_state == IB_QPS_RESET) {
+		err = 0;
+		goto out;
+	}
+
+	err = __mlx5_ib_modify_qp(ibqp, attr, attr_mask, cur_state,
+				  new_state, &ucmd);
+
+out:
+	mutex_unlock(&qp->mutex);
+	return err;
+}
+
+static int mlx5_wq_overflow(struct mlx5_ib_wq *wq, int nreq, struct ib_cq *ib_cq)
+{
+	struct mlx5_ib_cq *cq;
+	unsigned cur;
+
+	cur = wq->head - wq->tail;
+	if (likely(cur + nreq < wq->max_post))
+		return 0;
+
+	cq = to_mcq(ib_cq);
+	spin_lock(&cq->lock);
+	cur = wq->head - wq->tail;
+	spin_unlock(&cq->lock);
+
+	return cur + nreq >= wq->max_post;
+}
+
+static __always_inline void set_raddr_seg(struct mlx5_wqe_raddr_seg *rseg,
+					  u64 remote_addr, u32 rkey)
+{
+	rseg->raddr    = cpu_to_be64(remote_addr);
+	rseg->rkey     = cpu_to_be32(rkey);
+	rseg->reserved = 0;
+}
+
+static void *set_eth_seg(struct mlx5_wqe_eth_seg *eseg,
+			 const struct ib_send_wr *wr, void *qend,
+			 struct mlx5_ib_qp *qp, int *size)
+{
+	void *seg = eseg;
+
+	memset(eseg, 0, sizeof(struct mlx5_wqe_eth_seg));
+
+	if (wr->send_flags & IB_SEND_IP_CSUM)
+		eseg->cs_flags = MLX5_ETH_WQE_L3_CSUM |
+				 MLX5_ETH_WQE_L4_CSUM;
+
+	seg += sizeof(struct mlx5_wqe_eth_seg);
+	*size += sizeof(struct mlx5_wqe_eth_seg) / 16;
+
+	if (wr->opcode == IB_WR_LSO) {
+		struct ib_ud_wr *ud_wr = container_of(wr, struct ib_ud_wr, wr);
+		int size_of_inl_hdr_start = sizeof(eseg->inline_hdr.start);
+		u64 left, leftlen, copysz;
+		void *pdata = ud_wr->header;
+
+		left = ud_wr->hlen;
+		eseg->mss = cpu_to_be16(ud_wr->mss);
+		eseg->inline_hdr.sz = cpu_to_be16(left);
+
+		/*
+		 * check if there is space till the end of queue, if yes,
+		 * copy all in one shot, otherwise copy till the end of queue,
+		 * rollback and than the copy the left
+		 */
+		leftlen = qend - (void *)eseg->inline_hdr.start;
+		copysz = min_t(u64, leftlen, left);
+
+		memcpy(seg - size_of_inl_hdr_start, pdata, copysz);
+
+		if (likely(copysz > size_of_inl_hdr_start)) {
+			seg += ALIGN(copysz - size_of_inl_hdr_start, 16);
+			*size += ALIGN(copysz - size_of_inl_hdr_start, 16) / 16;
+		}
+
+		if (unlikely(copysz < left)) { /* the last wqe in the queue */
+			seg = mlx5_get_send_wqe(qp, 0);
+			left -= copysz;
+			pdata += copysz;
+			memcpy(seg, pdata, left);
+			seg += ALIGN(left, 16);
+			*size += ALIGN(left, 16) / 16;
+		}
+	}
+
+	return seg;
+}
+
+static void set_datagram_seg(struct mlx5_wqe_datagram_seg *dseg,
+			     const struct ib_send_wr *wr)
+{
+	memcpy(&dseg->av, &to_mah(ud_wr(wr)->ah)->av, sizeof(struct mlx5_av));
+	dseg->av.dqp_dct = cpu_to_be32(ud_wr(wr)->remote_qpn | MLX5_EXTENDED_UD_AV);
+	dseg->av.key.qkey.qkey = cpu_to_be32(ud_wr(wr)->remote_qkey);
+}
+
+static void set_data_ptr_seg(struct mlx5_wqe_data_seg *dseg, struct ib_sge *sg)
+{
+	dseg->byte_count = cpu_to_be32(sg->length);
+	dseg->lkey       = cpu_to_be32(sg->lkey);
+	dseg->addr       = cpu_to_be64(sg->addr);
+}
+
+static u64 get_xlt_octo(u64 bytes)
+{
+	return ALIGN(bytes, MLX5_IB_UMR_XLT_ALIGNMENT) /
+	       MLX5_IB_UMR_OCTOWORD;
+}
+
+static __be64 frwr_mkey_mask(void)
+{
+	u64 result;
+
+	result = MLX5_MKEY_MASK_LEN		|
+		MLX5_MKEY_MASK_PAGE_SIZE	|
+		MLX5_MKEY_MASK_START_ADDR	|
+		MLX5_MKEY_MASK_EN_RINVAL	|
+		MLX5_MKEY_MASK_KEY		|
+		MLX5_MKEY_MASK_LR		|
+		MLX5_MKEY_MASK_LW		|
+		MLX5_MKEY_MASK_RR		|
+		MLX5_MKEY_MASK_RW		|
+		MLX5_MKEY_MASK_A		|
+		MLX5_MKEY_MASK_SMALL_FENCE	|
+		MLX5_MKEY_MASK_FREE;
+
+	return cpu_to_be64(result);
+}
+
+static __be64 sig_mkey_mask(void)
+{
+	u64 result;
+
+	result = MLX5_MKEY_MASK_LEN		|
+		MLX5_MKEY_MASK_PAGE_SIZE	|
+		MLX5_MKEY_MASK_START_ADDR	|
+		MLX5_MKEY_MASK_EN_SIGERR	|
+		MLX5_MKEY_MASK_EN_RINVAL	|
+		MLX5_MKEY_MASK_KEY		|
+		MLX5_MKEY_MASK_LR		|
+		MLX5_MKEY_MASK_LW		|
+		MLX5_MKEY_MASK_RR		|
+		MLX5_MKEY_MASK_RW		|
+		MLX5_MKEY_MASK_SMALL_FENCE	|
+		MLX5_MKEY_MASK_FREE		|
+		MLX5_MKEY_MASK_BSF_EN;
+
+	return cpu_to_be64(result);
+}
+
+static void set_reg_umr_seg(struct mlx5_wqe_umr_ctrl_seg *umr,
+			    struct mlx5_ib_mr *mr, bool umr_inline)
+{
+	int size = mr->ndescs * mr->desc_size;
+
+	memset(umr, 0, sizeof(*umr));
+
+	umr->flags = MLX5_UMR_CHECK_NOT_FREE;
+	if (umr_inline)
+		umr->flags |= MLX5_UMR_INLINE;
+	umr->xlt_octowords = cpu_to_be16(get_xlt_octo(size));
+	umr->mkey_mask = frwr_mkey_mask();
+}
+
+static void set_linv_umr_seg(struct mlx5_wqe_umr_ctrl_seg *umr)
+{
+	memset(umr, 0, sizeof(*umr));
+	umr->mkey_mask = cpu_to_be64(MLX5_MKEY_MASK_FREE);
+	umr->flags = MLX5_UMR_INLINE;
+}
+
+static __be64 get_umr_enable_mr_mask(void)
+{
+	u64 result;
+
+	result = MLX5_MKEY_MASK_KEY |
+		 MLX5_MKEY_MASK_FREE;
+
+	return cpu_to_be64(result);
+}
+
+static __be64 get_umr_disable_mr_mask(void)
+{
+	u64 result;
+
+	result = MLX5_MKEY_MASK_FREE;
+
+	return cpu_to_be64(result);
+}
+
+static __be64 get_umr_update_translation_mask(void)
+{
+	u64 result;
+
+	result = MLX5_MKEY_MASK_LEN |
+		 MLX5_MKEY_MASK_PAGE_SIZE |
+		 MLX5_MKEY_MASK_START_ADDR;
+
+	return cpu_to_be64(result);
+}
+
+static __be64 get_umr_update_access_mask(int atomic)
+{
+	u64 result;
+
+	result = MLX5_MKEY_MASK_LR |
+		 MLX5_MKEY_MASK_LW |
+		 MLX5_MKEY_MASK_RR |
+		 MLX5_MKEY_MASK_RW;
+
+	if (atomic)
+		result |= MLX5_MKEY_MASK_A;
+
+	return cpu_to_be64(result);
+}
+
+static __be64 get_umr_update_pd_mask(void)
+{
+	u64 result;
+
+	result = MLX5_MKEY_MASK_PD;
+
+	return cpu_to_be64(result);
+}
+
+static int umr_check_mkey_mask(struct mlx5_ib_dev *dev, u64 mask)
+{
+	if ((mask & MLX5_MKEY_MASK_PAGE_SIZE &&
+	     MLX5_CAP_GEN(dev->mdev, umr_modify_entity_size_disabled)) ||
+	    (mask & MLX5_MKEY_MASK_A &&
+	     MLX5_CAP_GEN(dev->mdev, umr_modify_atomic_disabled)))
+		return -EPERM;
+	return 0;
+}
+
+static int set_reg_umr_segment(struct mlx5_ib_dev *dev,
+			       struct mlx5_wqe_umr_ctrl_seg *umr,
+			       const struct ib_send_wr *wr, int atomic)
+{
+	const struct mlx5_umr_wr *umrwr = umr_wr(wr);
+
+	memset(umr, 0, sizeof(*umr));
+
+	if (!umrwr->ignore_free_state) {
+		if (wr->send_flags & MLX5_IB_SEND_UMR_FAIL_IF_FREE)
+			 /* fail if free */
+			umr->flags = MLX5_UMR_CHECK_FREE;
+		else
+			/* fail if not free */
+			umr->flags = MLX5_UMR_CHECK_NOT_FREE;
+	}
+
+	umr->xlt_octowords = cpu_to_be16(get_xlt_octo(umrwr->xlt_size));
+	if (wr->send_flags & MLX5_IB_SEND_UMR_UPDATE_XLT) {
+		u64 offset = get_xlt_octo(umrwr->offset);
+
+		umr->xlt_offset = cpu_to_be16(offset & 0xffff);
+		umr->xlt_offset_47_16 = cpu_to_be32(offset >> 16);
+		umr->flags |= MLX5_UMR_TRANSLATION_OFFSET_EN;
+	}
+	if (wr->send_flags & MLX5_IB_SEND_UMR_UPDATE_TRANSLATION)
+		umr->mkey_mask |= get_umr_update_translation_mask();
+	if (wr->send_flags & MLX5_IB_SEND_UMR_UPDATE_PD_ACCESS) {
+		umr->mkey_mask |= get_umr_update_access_mask(atomic);
+		umr->mkey_mask |= get_umr_update_pd_mask();
+	}
+	if (wr->send_flags & MLX5_IB_SEND_UMR_ENABLE_MR)
+		umr->mkey_mask |= get_umr_enable_mr_mask();
+	if (wr->send_flags & MLX5_IB_SEND_UMR_DISABLE_MR)
+		umr->mkey_mask |= get_umr_disable_mr_mask();
+
+	if (!wr->num_sge)
+		umr->flags |= MLX5_UMR_INLINE;
+
+	return umr_check_mkey_mask(dev, be64_to_cpu(umr->mkey_mask));
+}
+
+static u8 get_umr_flags(int acc)
+{
+	return (acc & IB_ACCESS_REMOTE_ATOMIC ? MLX5_PERM_ATOMIC       : 0) |
+	       (acc & IB_ACCESS_REMOTE_WRITE  ? MLX5_PERM_REMOTE_WRITE : 0) |
+	       (acc & IB_ACCESS_REMOTE_READ   ? MLX5_PERM_REMOTE_READ  : 0) |
+	       (acc & IB_ACCESS_LOCAL_WRITE   ? MLX5_PERM_LOCAL_WRITE  : 0) |
+		MLX5_PERM_LOCAL_READ | MLX5_PERM_UMR_EN;
+}
+
+static void set_reg_mkey_seg(struct mlx5_mkey_seg *seg,
+			     struct mlx5_ib_mr *mr,
+			     u32 key, int access)
+{
+	int ndescs = ALIGN(mr->ndescs, 8) >> 1;
+
+	memset(seg, 0, sizeof(*seg));
+
+	if (mr->access_mode == MLX5_MKC_ACCESS_MODE_MTT)
+		seg->log2_page_size = ilog2(mr->ibmr.page_size);
+	else if (mr->access_mode == MLX5_MKC_ACCESS_MODE_KLMS)
+		/* KLMs take twice the size of MTTs */
+		ndescs *= 2;
+
+	seg->flags = get_umr_flags(access) | mr->access_mode;
+	seg->qpn_mkey7_0 = cpu_to_be32((key & 0xff) | 0xffffff00);
+	seg->flags_pd = cpu_to_be32(MLX5_MKEY_REMOTE_INVAL);
+	seg->start_addr = cpu_to_be64(mr->ibmr.iova);
+	seg->len = cpu_to_be64(mr->ibmr.length);
+	seg->xlt_oct_size = cpu_to_be32(ndescs);
+}
+
+static void set_linv_mkey_seg(struct mlx5_mkey_seg *seg)
+{
+	memset(seg, 0, sizeof(*seg));
+	seg->status = MLX5_MKEY_STATUS_FREE;
+}
+
+static void set_reg_mkey_segment(struct mlx5_mkey_seg *seg,
+				 const struct ib_send_wr *wr)
+{
+	const struct mlx5_umr_wr *umrwr = umr_wr(wr);
+
+	memset(seg, 0, sizeof(*seg));
+	if (wr->send_flags & MLX5_IB_SEND_UMR_DISABLE_MR)
+		seg->status = MLX5_MKEY_STATUS_FREE;
+
+	seg->flags = convert_access(umrwr->access_flags);
+	if (umrwr->pd)
+		seg->flags_pd = cpu_to_be32(to_mpd(umrwr->pd)->pdn);
+	if (wr->send_flags & MLX5_IB_SEND_UMR_UPDATE_TRANSLATION &&
+	    !umrwr->length)
+		seg->flags_pd |= cpu_to_be32(MLX5_MKEY_LEN64);
+
+	seg->start_addr = cpu_to_be64(umrwr->virt_addr);
+	seg->len = cpu_to_be64(umrwr->length);
+	seg->log2_page_size = umrwr->page_shift;
+	seg->qpn_mkey7_0 = cpu_to_be32(0xffffff00 |
+				       mlx5_mkey_variant(umrwr->mkey));
+}
+
+static void set_reg_data_seg(struct mlx5_wqe_data_seg *dseg,
+			     struct mlx5_ib_mr *mr,
+			     struct mlx5_ib_pd *pd)
+{
+	int bcount = mr->desc_size * mr->ndescs;
+
+	dseg->addr = cpu_to_be64(mr->desc_map);
+	dseg->byte_count = cpu_to_be32(ALIGN(bcount, 64));
+	dseg->lkey = cpu_to_be32(pd->ibpd.local_dma_lkey);
+}
+
+static void set_reg_umr_inline_seg(void *seg, struct mlx5_ib_qp *qp,
+				   struct mlx5_ib_mr *mr, int mr_list_size)
+{
+	void *qend = qp->sq.qend;
+	void *addr = mr->descs;
+	int copy;
+
+	if (unlikely(seg + mr_list_size > qend)) {
+		copy = qend - seg;
+		memcpy(seg, addr, copy);
+		addr += copy;
+		mr_list_size -= copy;
+		seg = mlx5_get_send_wqe(qp, 0);
+	}
+	memcpy(seg, addr, mr_list_size);
+	seg += mr_list_size;
+}
+
+static __be32 send_ieth(const struct ib_send_wr *wr)
+{
+	switch (wr->opcode) {
+	case IB_WR_SEND_WITH_IMM:
+	case IB_WR_RDMA_WRITE_WITH_IMM:
+		return wr->ex.imm_data;
+
+	case IB_WR_SEND_WITH_INV:
+		return cpu_to_be32(wr->ex.invalidate_rkey);
+
+	default:
+		return 0;
+	}
+}
+
+static u8 calc_sig(void *wqe, int size)
+{
+	u8 *p = wqe;
+	u8 res = 0;
+	int i;
+
+	for (i = 0; i < size; i++)
+		res ^= p[i];
+
+	return ~res;
+}
+
+static u8 wq_sig(void *wqe)
+{
+	return calc_sig(wqe, (*((u8 *)wqe + 8) & 0x3f) << 4);
+}
+
+static int set_data_inl_seg(struct mlx5_ib_qp *qp, const struct ib_send_wr *wr,
+			    void *wqe, int *sz)
+{
+	struct mlx5_wqe_inline_seg *seg;
+	void *qend = qp->sq.qend;
+	void *addr;
+	int inl = 0;
+	int copy;
+	int len;
+	int i;
+
+	seg = wqe;
+	wqe += sizeof(*seg);
+	for (i = 0; i < wr->num_sge; i++) {
+		addr = (void *)(unsigned long)(wr->sg_list[i].addr);
+		len  = wr->sg_list[i].length;
+		inl += len;
+
+		if (unlikely(inl > qp->max_inline_data))
+			return -ENOMEM;
+
+		if (unlikely(wqe + len > qend)) {
+			copy = qend - wqe;
+			memcpy(wqe, addr, copy);
+			addr += copy;
+			len -= copy;
+			wqe = mlx5_get_send_wqe(qp, 0);
+		}
+		memcpy(wqe, addr, len);
+		wqe += len;
+	}
+
+	seg->byte_count = cpu_to_be32(inl | MLX5_INLINE_SEG);
+
+	*sz = ALIGN(inl + sizeof(seg->byte_count), 16) / 16;
+
+	return 0;
+}
+
+static u16 prot_field_size(enum ib_signature_type type)
+{
+	switch (type) {
+	case IB_SIG_TYPE_T10_DIF:
+		return MLX5_DIF_SIZE;
+	default:
+		return 0;
+	}
+}
+
+static u8 bs_selector(int block_size)
+{
+	switch (block_size) {
+	case 512:	    return 0x1;
+	case 520:	    return 0x2;
+	case 4096:	    return 0x3;
+	case 4160:	    return 0x4;
+	case 1073741824:    return 0x5;
+	default:	    return 0;
+	}
+}
+
+static void mlx5_fill_inl_bsf(struct ib_sig_domain *domain,
+			      struct mlx5_bsf_inl *inl)
+{
+	/* Valid inline section and allow BSF refresh */
+	inl->vld_refresh = cpu_to_be16(MLX5_BSF_INL_VALID |
+				       MLX5_BSF_REFRESH_DIF);
+	inl->dif_apptag = cpu_to_be16(domain->sig.dif.app_tag);
+	inl->dif_reftag = cpu_to_be32(domain->sig.dif.ref_tag);
+	/* repeating block */
+	inl->rp_inv_seed = MLX5_BSF_REPEAT_BLOCK;
+	inl->sig_type = domain->sig.dif.bg_type == IB_T10DIF_CRC ?
+			MLX5_DIF_CRC : MLX5_DIF_IPCS;
+
+	if (domain->sig.dif.ref_remap)
+		inl->dif_inc_ref_guard_check |= MLX5_BSF_INC_REFTAG;
+
+	if (domain->sig.dif.app_escape) {
+		if (domain->sig.dif.ref_escape)
+			inl->dif_inc_ref_guard_check |= MLX5_BSF_APPREF_ESCAPE;
+		else
+			inl->dif_inc_ref_guard_check |= MLX5_BSF_APPTAG_ESCAPE;
+	}
+
+	inl->dif_app_bitmask_check =
+		cpu_to_be16(domain->sig.dif.apptag_check_mask);
+}
+
+static int mlx5_set_bsf(struct ib_mr *sig_mr,
+			struct ib_sig_attrs *sig_attrs,
+			struct mlx5_bsf *bsf, u32 data_size)
+{
+	struct mlx5_core_sig_ctx *msig = to_mmr(sig_mr)->sig;
+	struct mlx5_bsf_basic *basic = &bsf->basic;
+	struct ib_sig_domain *mem = &sig_attrs->mem;
+	struct ib_sig_domain *wire = &sig_attrs->wire;
+
+	memset(bsf, 0, sizeof(*bsf));
+
+	/* Basic + Extended + Inline */
+	basic->bsf_size_sbs = 1 << 7;
+	/* Input domain check byte mask */
+	basic->check_byte_mask = sig_attrs->check_mask;
+	basic->raw_data_size = cpu_to_be32(data_size);
+
+	/* Memory domain */
+	switch (sig_attrs->mem.sig_type) {
+	case IB_SIG_TYPE_NONE:
+		break;
+	case IB_SIG_TYPE_T10_DIF:
+		basic->mem.bs_selector = bs_selector(mem->sig.dif.pi_interval);
+		basic->m_bfs_psv = cpu_to_be32(msig->psv_memory.psv_idx);
+		mlx5_fill_inl_bsf(mem, &bsf->m_inl);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	/* Wire domain */
+	switch (sig_attrs->wire.sig_type) {
+	case IB_SIG_TYPE_NONE:
+		break;
+	case IB_SIG_TYPE_T10_DIF:
+		if (mem->sig.dif.pi_interval == wire->sig.dif.pi_interval &&
+		    mem->sig_type == wire->sig_type) {
+			/* Same block structure */
+			basic->bsf_size_sbs |= 1 << 4;
+			if (mem->sig.dif.bg_type == wire->sig.dif.bg_type)
+				basic->wire.copy_byte_mask |= MLX5_CPY_GRD_MASK;
+			if (mem->sig.dif.app_tag == wire->sig.dif.app_tag)
+				basic->wire.copy_byte_mask |= MLX5_CPY_APP_MASK;
+			if (mem->sig.dif.ref_tag == wire->sig.dif.ref_tag)
+				basic->wire.copy_byte_mask |= MLX5_CPY_REF_MASK;
+		} else
+			basic->wire.bs_selector = bs_selector(wire->sig.dif.pi_interval);
+
+		basic->w_bfs_psv = cpu_to_be32(msig->psv_wire.psv_idx);
+		mlx5_fill_inl_bsf(wire, &bsf->w_inl);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int set_sig_data_segment(const struct ib_sig_handover_wr *wr,
+				struct mlx5_ib_qp *qp, void **seg, int *size)
+{
+	struct ib_sig_attrs *sig_attrs = wr->sig_attrs;
+	struct ib_mr *sig_mr = wr->sig_mr;
+	struct mlx5_bsf *bsf;
+	u32 data_len = wr->wr.sg_list->length;
+	u32 data_key = wr->wr.sg_list->lkey;
+	u64 data_va = wr->wr.sg_list->addr;
+	int ret;
+	int wqe_size;
+
+	if (!wr->prot ||
+	    (data_key == wr->prot->lkey &&
+	     data_va == wr->prot->addr &&
+	     data_len == wr->prot->length)) {
+		/**
+		 * Source domain doesn't contain signature information
+		 * or data and protection are interleaved in memory.
+		 * So need construct:
+		 *                  ------------------
+		 *                 |     data_klm     |
+		 *                  ------------------
+		 *                 |       BSF        |
+		 *                  ------------------
+		 **/
+		struct mlx5_klm *data_klm = *seg;
+
+		data_klm->bcount = cpu_to_be32(data_len);
+		data_klm->key = cpu_to_be32(data_key);
+		data_klm->va = cpu_to_be64(data_va);
+		wqe_size = ALIGN(sizeof(*data_klm), 64);
+	} else {
+		/**
+		 * Source domain contains signature information
+		 * So need construct a strided block format:
+		 *               ---------------------------
+		 *              |     stride_block_ctrl     |
+		 *               ---------------------------
+		 *              |          data_klm         |
+		 *               ---------------------------
+		 *              |          prot_klm         |
+		 *               ---------------------------
+		 *              |             BSF           |
+		 *               ---------------------------
+		 **/
+		struct mlx5_stride_block_ctrl_seg *sblock_ctrl;
+		struct mlx5_stride_block_entry *data_sentry;
+		struct mlx5_stride_block_entry *prot_sentry;
+		u32 prot_key = wr->prot->lkey;
+		u64 prot_va = wr->prot->addr;
+		u16 block_size = sig_attrs->mem.sig.dif.pi_interval;
+		int prot_size;
+
+		sblock_ctrl = *seg;
+		data_sentry = (void *)sblock_ctrl + sizeof(*sblock_ctrl);
+		prot_sentry = (void *)data_sentry + sizeof(*data_sentry);
+
+		prot_size = prot_field_size(sig_attrs->mem.sig_type);
+		if (!prot_size) {
+			pr_err("Bad block size given: %u\n", block_size);
+			return -EINVAL;
+		}
+		sblock_ctrl->bcount_per_cycle = cpu_to_be32(block_size +
+							    prot_size);
+		sblock_ctrl->op = cpu_to_be32(MLX5_STRIDE_BLOCK_OP);
+		sblock_ctrl->repeat_count = cpu_to_be32(data_len / block_size);
+		sblock_ctrl->num_entries = cpu_to_be16(2);
+
+		data_sentry->bcount = cpu_to_be16(block_size);
+		data_sentry->key = cpu_to_be32(data_key);
+		data_sentry->va = cpu_to_be64(data_va);
+		data_sentry->stride = cpu_to_be16(block_size);
+
+		prot_sentry->bcount = cpu_to_be16(prot_size);
+		prot_sentry->key = cpu_to_be32(prot_key);
+		prot_sentry->va = cpu_to_be64(prot_va);
+		prot_sentry->stride = cpu_to_be16(prot_size);
+
+		wqe_size = ALIGN(sizeof(*sblock_ctrl) + sizeof(*data_sentry) +
+				 sizeof(*prot_sentry), 64);
+	}
+
+	*seg += wqe_size;
+	*size += wqe_size / 16;
+	if (unlikely((*seg == qp->sq.qend)))
+		*seg = mlx5_get_send_wqe(qp, 0);
+
+	bsf = *seg;
+	ret = mlx5_set_bsf(sig_mr, sig_attrs, bsf, data_len);
+	if (ret)
+		return -EINVAL;
+
+	*seg += sizeof(*bsf);
+	*size += sizeof(*bsf) / 16;
+	if (unlikely((*seg == qp->sq.qend)))
+		*seg = mlx5_get_send_wqe(qp, 0);
+
+	return 0;
+}
+
+static void set_sig_mkey_segment(struct mlx5_mkey_seg *seg,
+				 const struct ib_sig_handover_wr *wr, u32 size,
+				 u32 length, u32 pdn)
+{
+	struct ib_mr *sig_mr = wr->sig_mr;
+	u32 sig_key = sig_mr->rkey;
+	u8 sigerr = to_mmr(sig_mr)->sig->sigerr_count & 1;
+
+	memset(seg, 0, sizeof(*seg));
+
+	seg->flags = get_umr_flags(wr->access_flags) |
+				   MLX5_MKC_ACCESS_MODE_KLMS;
+	seg->qpn_mkey7_0 = cpu_to_be32((sig_key & 0xff) | 0xffffff00);
+	seg->flags_pd = cpu_to_be32(MLX5_MKEY_REMOTE_INVAL | sigerr << 26 |
+				    MLX5_MKEY_BSF_EN | pdn);
+	seg->len = cpu_to_be64(length);
+	seg->xlt_oct_size = cpu_to_be32(get_xlt_octo(size));
+	seg->bsfs_octo_size = cpu_to_be32(MLX5_MKEY_BSF_OCTO_SIZE);
+}
+
+static void set_sig_umr_segment(struct mlx5_wqe_umr_ctrl_seg *umr,
+				u32 size)
+{
+	memset(umr, 0, sizeof(*umr));
+
+	umr->flags = MLX5_FLAGS_INLINE | MLX5_FLAGS_CHECK_FREE;
+	umr->xlt_octowords = cpu_to_be16(get_xlt_octo(size));
+	umr->bsf_octowords = cpu_to_be16(MLX5_MKEY_BSF_OCTO_SIZE);
+	umr->mkey_mask = sig_mkey_mask();
+}
+
+
+static int set_sig_umr_wr(const struct ib_send_wr *send_wr,
+			  struct mlx5_ib_qp *qp, void **seg, int *size)
+{
+	const struct ib_sig_handover_wr *wr = sig_handover_wr(send_wr);
+	struct mlx5_ib_mr *sig_mr = to_mmr(wr->sig_mr);
+	u32 pdn = get_pd(qp)->pdn;
+	u32 xlt_size;
+	int region_len, ret;
+
+	if (unlikely(wr->wr.num_sge != 1) ||
+	    unlikely(wr->access_flags & IB_ACCESS_REMOTE_ATOMIC) ||
+	    unlikely(!sig_mr->sig) || unlikely(!qp->signature_en) ||
+	    unlikely(!sig_mr->sig->sig_status_checked))
+		return -EINVAL;
+
+	/* length of the protected region, data + protection */
+	region_len = wr->wr.sg_list->length;
+	if (wr->prot &&
+	    (wr->prot->lkey != wr->wr.sg_list->lkey  ||
+	     wr->prot->addr != wr->wr.sg_list->addr  ||
+	     wr->prot->length != wr->wr.sg_list->length))
+		region_len += wr->prot->length;
+
+	/**
+	 * KLM octoword size - if protection was provided
+	 * then we use strided block format (3 octowords),
+	 * else we use single KLM (1 octoword)
+	 **/
+	xlt_size = wr->prot ? 0x30 : sizeof(struct mlx5_klm);
+
+	set_sig_umr_segment(*seg, xlt_size);
+	*seg += sizeof(struct mlx5_wqe_umr_ctrl_seg);
+	*size += sizeof(struct mlx5_wqe_umr_ctrl_seg) / 16;
+	if (unlikely((*seg == qp->sq.qend)))
+		*seg = mlx5_get_send_wqe(qp, 0);
+
+	set_sig_mkey_segment(*seg, wr, xlt_size, region_len, pdn);
+	*seg += sizeof(struct mlx5_mkey_seg);
+	*size += sizeof(struct mlx5_mkey_seg) / 16;
+	if (unlikely((*seg == qp->sq.qend)))
+		*seg = mlx5_get_send_wqe(qp, 0);
+
+	ret = set_sig_data_segment(wr, qp, seg, size);
+	if (ret)
+		return ret;
+
+	sig_mr->sig->sig_status_checked = false;
+	return 0;
+}
+
+static int set_psv_wr(struct ib_sig_domain *domain,
+		      u32 psv_idx, void **seg, int *size)
+{
+	struct mlx5_seg_set_psv *psv_seg = *seg;
+
+	memset(psv_seg, 0, sizeof(*psv_seg));
+	psv_seg->psv_num = cpu_to_be32(psv_idx);
+	switch (domain->sig_type) {
+	case IB_SIG_TYPE_NONE:
+		break;
+	case IB_SIG_TYPE_T10_DIF:
+		psv_seg->transient_sig = cpu_to_be32(domain->sig.dif.bg << 16 |
+						     domain->sig.dif.app_tag);
+		psv_seg->ref_tag = cpu_to_be32(domain->sig.dif.ref_tag);
+		break;
+	default:
+		pr_err("Bad signature type (%d) is given.\n",
+		       domain->sig_type);
+		return -EINVAL;
+	}
+
+	*seg += sizeof(*psv_seg);
+	*size += sizeof(*psv_seg) / 16;
+
+	return 0;
+}
+
+static int set_reg_wr(struct mlx5_ib_qp *qp,
+		      const struct ib_reg_wr *wr,
+		      void **seg, int *size)
+{
+	struct mlx5_ib_mr *mr = to_mmr(wr->mr);
+	struct mlx5_ib_pd *pd = to_mpd(qp->ibqp.pd);
+	int mr_list_size = mr->ndescs * mr->desc_size;
+	bool umr_inline = mr_list_size <= MLX5_IB_SQ_UMR_INLINE_THRESHOLD;
+
+	if (unlikely(wr->wr.send_flags & IB_SEND_INLINE)) {
+		mlx5_ib_warn(to_mdev(qp->ibqp.device),
+			     "Invalid IB_SEND_INLINE send flag\n");
+		return -EINVAL;
+	}
+
+	set_reg_umr_seg(*seg, mr, umr_inline);
+	*seg += sizeof(struct mlx5_wqe_umr_ctrl_seg);
+	*size += sizeof(struct mlx5_wqe_umr_ctrl_seg) / 16;
+	if (unlikely((*seg == qp->sq.qend)))
+		*seg = mlx5_get_send_wqe(qp, 0);
+
+	set_reg_mkey_seg(*seg, mr, wr->key, wr->access);
+	*seg += sizeof(struct mlx5_mkey_seg);
+	*size += sizeof(struct mlx5_mkey_seg) / 16;
+	if (unlikely((*seg == qp->sq.qend)))
+		*seg = mlx5_get_send_wqe(qp, 0);
+
+	if (umr_inline) {
+		set_reg_umr_inline_seg(*seg, qp, mr, mr_list_size);
+		*size += get_xlt_octo(mr_list_size);
+	} else {
+		set_reg_data_seg(*seg, mr, pd);
+		*seg += sizeof(struct mlx5_wqe_data_seg);
+		*size += (sizeof(struct mlx5_wqe_data_seg) / 16);
+	}
+	return 0;
+}
+
+static void set_linv_wr(struct mlx5_ib_qp *qp, void **seg, int *size)
+{
+	set_linv_umr_seg(*seg);
+	*seg += sizeof(struct mlx5_wqe_umr_ctrl_seg);
+	*size += sizeof(struct mlx5_wqe_umr_ctrl_seg) / 16;
+	if (unlikely((*seg == qp->sq.qend)))
+		*seg = mlx5_get_send_wqe(qp, 0);
+	set_linv_mkey_seg(*seg);
+	*seg += sizeof(struct mlx5_mkey_seg);
+	*size += sizeof(struct mlx5_mkey_seg) / 16;
+	if (unlikely((*seg == qp->sq.qend)))
+		*seg = mlx5_get_send_wqe(qp, 0);
+}
+
+static void dump_wqe(struct mlx5_ib_qp *qp, int idx, int size_16)
+{
+	__be32 *p = NULL;
+	int tidx = idx;
+	int i, j;
+
+	pr_debug("dump wqe at %p\n", mlx5_get_send_wqe(qp, tidx));
+	for (i = 0, j = 0; i < size_16 * 4; i += 4, j += 4) {
+		if ((i & 0xf) == 0) {
+			void *buf = mlx5_get_send_wqe(qp, tidx);
+			tidx = (tidx + 1) & (qp->sq.wqe_cnt - 1);
+			p = buf;
+			j = 0;
+		}
+		pr_debug("%08x %08x %08x %08x\n", be32_to_cpu(p[j]),
+			 be32_to_cpu(p[j + 1]), be32_to_cpu(p[j + 2]),
+			 be32_to_cpu(p[j + 3]));
+	}
+}
+
+static int __begin_wqe(struct mlx5_ib_qp *qp, void **seg,
+		     struct mlx5_wqe_ctrl_seg **ctrl,
+		     const struct ib_send_wr *wr, unsigned *idx,
+		     int *size, int nreq, bool send_signaled, bool solicited)
+{
+	if (unlikely(mlx5_wq_overflow(&qp->sq, nreq, qp->ibqp.send_cq)))
+		return -ENOMEM;
+
+	*idx = qp->sq.cur_post & (qp->sq.wqe_cnt - 1);
+	*seg = mlx5_get_send_wqe(qp, *idx);
+	*ctrl = *seg;
+	*(uint32_t *)(*seg + 8) = 0;
+	(*ctrl)->imm = send_ieth(wr);
+	(*ctrl)->fm_ce_se = qp->sq_signal_bits |
+		(send_signaled ? MLX5_WQE_CTRL_CQ_UPDATE : 0) |
+		(solicited ? MLX5_WQE_CTRL_SOLICITED : 0);
+
+	*seg += sizeof(**ctrl);
+	*size = sizeof(**ctrl) / 16;
+
+	return 0;
+}
+
+static int begin_wqe(struct mlx5_ib_qp *qp, void **seg,
+		     struct mlx5_wqe_ctrl_seg **ctrl,
+		     const struct ib_send_wr *wr, unsigned *idx,
+		     int *size, int nreq)
+{
+	return __begin_wqe(qp, seg, ctrl, wr, idx, size, nreq,
+			   wr->send_flags & IB_SEND_SIGNALED,
+			   wr->send_flags & IB_SEND_SOLICITED);
+}
+
+static void finish_wqe(struct mlx5_ib_qp *qp,
+		       struct mlx5_wqe_ctrl_seg *ctrl,
+		       u8 size, unsigned idx, u64 wr_id,
+		       int nreq, u8 fence, u32 mlx5_opcode)
+{
+	u8 opmod = 0;
+
+	ctrl->opmod_idx_opcode = cpu_to_be32(((u32)(qp->sq.cur_post) << 8) |
+					     mlx5_opcode | ((u32)opmod << 24));
+	ctrl->qpn_ds = cpu_to_be32(size | (qp->trans_qp.base.mqp.qpn << 8));
+	ctrl->fm_ce_se |= fence;
+	if (unlikely(qp->wq_sig))
+		ctrl->signature = wq_sig(ctrl);
+
+	qp->sq.wrid[idx] = wr_id;
+	qp->sq.w_list[idx].opcode = mlx5_opcode;
+	qp->sq.wqe_head[idx] = qp->sq.head + nreq;
+	qp->sq.cur_post += DIV_ROUND_UP(size * 16, MLX5_SEND_WQE_BB);
+	qp->sq.w_list[idx].next = qp->sq.cur_post;
+}
+
+static int _mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr,
+			      const struct ib_send_wr **bad_wr, bool drain)
+{
+	struct mlx5_wqe_ctrl_seg *ctrl = NULL;  /* compiler warning */
+	struct mlx5_ib_dev *dev = to_mdev(ibqp->device);
+	struct mlx5_core_dev *mdev = dev->mdev;
+	struct mlx5_ib_qp *qp;
+	struct mlx5_ib_mr *mr;
+	struct mlx5_wqe_data_seg *dpseg;
+	struct mlx5_wqe_xrc_seg *xrc;
+	struct mlx5_bf *bf;
+	int uninitialized_var(size);
+	void *qend;
+	unsigned long flags;
+	unsigned idx;
+	int err = 0;
+	int num_sge;
+	void *seg;
+	int nreq;
+	int i;
+	u8 next_fence = 0;
+	u8 fence;
+
+	if (unlikely(mdev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR &&
+		     !drain)) {
+		*bad_wr = wr;
+		return -EIO;
+	}
+
+	if (unlikely(ibqp->qp_type == IB_QPT_GSI))
+		return mlx5_ib_gsi_post_send(ibqp, wr, bad_wr);
+
+	qp = to_mqp(ibqp);
+	bf = &qp->bf;
+	qend = qp->sq.qend;
+
+	spin_lock_irqsave(&qp->sq.lock, flags);
+
+	for (nreq = 0; wr; nreq++, wr = wr->next) {
+		if (unlikely(wr->opcode >= ARRAY_SIZE(mlx5_ib_opcode))) {
+			mlx5_ib_warn(dev, "\n");
+			err = -EINVAL;
+			*bad_wr = wr;
+			goto out;
+		}
+
+		num_sge = wr->num_sge;
+		if (unlikely(num_sge > qp->sq.max_gs)) {
+			mlx5_ib_warn(dev, "\n");
+			err = -EINVAL;
+			*bad_wr = wr;
+			goto out;
+		}
+
+		err = begin_wqe(qp, &seg, &ctrl, wr, &idx, &size, nreq);
+		if (err) {
+			mlx5_ib_warn(dev, "\n");
+			err = -ENOMEM;
+			*bad_wr = wr;
+			goto out;
+		}
+
+		if (wr->opcode == IB_WR_REG_MR) {
+			fence = dev->umr_fence;
+			next_fence = MLX5_FENCE_MODE_INITIATOR_SMALL;
+		} else  {
+			if (wr->send_flags & IB_SEND_FENCE) {
+				if (qp->next_fence)
+					fence = MLX5_FENCE_MODE_SMALL_AND_FENCE;
+				else
+					fence = MLX5_FENCE_MODE_FENCE;
+			} else {
+				fence = qp->next_fence;
+			}
+		}
+
+		switch (ibqp->qp_type) {
+		case IB_QPT_XRC_INI:
+			xrc = seg;
+			seg += sizeof(*xrc);
+			size += sizeof(*xrc) / 16;
+			/* fall through */
+		case IB_QPT_RC:
+			switch (wr->opcode) {
+			case IB_WR_RDMA_READ:
+			case IB_WR_RDMA_WRITE:
+			case IB_WR_RDMA_WRITE_WITH_IMM:
+				set_raddr_seg(seg, rdma_wr(wr)->remote_addr,
+					      rdma_wr(wr)->rkey);
+				seg += sizeof(struct mlx5_wqe_raddr_seg);
+				size += sizeof(struct mlx5_wqe_raddr_seg) / 16;
+				break;
+
+			case IB_WR_ATOMIC_CMP_AND_SWP:
+			case IB_WR_ATOMIC_FETCH_AND_ADD:
+			case IB_WR_MASKED_ATOMIC_CMP_AND_SWP:
+				mlx5_ib_warn(dev, "Atomic operations are not supported yet\n");
+				err = -ENOSYS;
+				*bad_wr = wr;
+				goto out;
+
+			case IB_WR_LOCAL_INV:
+				qp->sq.wr_data[idx] = IB_WR_LOCAL_INV;
+				ctrl->imm = cpu_to_be32(wr->ex.invalidate_rkey);
+				set_linv_wr(qp, &seg, &size);
+				num_sge = 0;
+				break;
+
+			case IB_WR_REG_MR:
+				qp->sq.wr_data[idx] = IB_WR_REG_MR;
+				ctrl->imm = cpu_to_be32(reg_wr(wr)->key);
+				err = set_reg_wr(qp, reg_wr(wr), &seg, &size);
+				if (err) {
+					*bad_wr = wr;
+					goto out;
+				}
+				num_sge = 0;
+				break;
+
+			case IB_WR_REG_SIG_MR:
+				qp->sq.wr_data[idx] = IB_WR_REG_SIG_MR;
+				mr = to_mmr(sig_handover_wr(wr)->sig_mr);
+
+				ctrl->imm = cpu_to_be32(mr->ibmr.rkey);
+				err = set_sig_umr_wr(wr, qp, &seg, &size);
+				if (err) {
+					mlx5_ib_warn(dev, "\n");
+					*bad_wr = wr;
+					goto out;
+				}
+
+				finish_wqe(qp, ctrl, size, idx, wr->wr_id, nreq,
+					   fence, MLX5_OPCODE_UMR);
+				/*
+				 * SET_PSV WQEs are not signaled and solicited
+				 * on error
+				 */
+				err = __begin_wqe(qp, &seg, &ctrl, wr, &idx,
+						  &size, nreq, false, true);
+				if (err) {
+					mlx5_ib_warn(dev, "\n");
+					err = -ENOMEM;
+					*bad_wr = wr;
+					goto out;
+				}
+
+				err = set_psv_wr(&sig_handover_wr(wr)->sig_attrs->mem,
+						 mr->sig->psv_memory.psv_idx, &seg,
+						 &size);
+				if (err) {
+					mlx5_ib_warn(dev, "\n");
+					*bad_wr = wr;
+					goto out;
+				}
+
+				finish_wqe(qp, ctrl, size, idx, wr->wr_id, nreq,
+					   fence, MLX5_OPCODE_SET_PSV);
+				err = __begin_wqe(qp, &seg, &ctrl, wr, &idx,
+						  &size, nreq, false, true);
+				if (err) {
+					mlx5_ib_warn(dev, "\n");
+					err = -ENOMEM;
+					*bad_wr = wr;
+					goto out;
+				}
+
+				err = set_psv_wr(&sig_handover_wr(wr)->sig_attrs->wire,
+						 mr->sig->psv_wire.psv_idx, &seg,
+						 &size);
+				if (err) {
+					mlx5_ib_warn(dev, "\n");
+					*bad_wr = wr;
+					goto out;
+				}
+
+				finish_wqe(qp, ctrl, size, idx, wr->wr_id, nreq,
+					   fence, MLX5_OPCODE_SET_PSV);
+				qp->next_fence = MLX5_FENCE_MODE_INITIATOR_SMALL;
+				num_sge = 0;
+				goto skip_psv;
+
+			default:
+				break;
+			}
+			break;
+
+		case IB_QPT_UC:
+			switch (wr->opcode) {
+			case IB_WR_RDMA_WRITE:
+			case IB_WR_RDMA_WRITE_WITH_IMM:
+				set_raddr_seg(seg, rdma_wr(wr)->remote_addr,
+					      rdma_wr(wr)->rkey);
+				seg  += sizeof(struct mlx5_wqe_raddr_seg);
+				size += sizeof(struct mlx5_wqe_raddr_seg) / 16;
+				break;
+
+			default:
+				break;
+			}
+			break;
+
+		case IB_QPT_SMI:
+			if (unlikely(!mdev->port_caps[qp->port - 1].has_smi)) {
+				mlx5_ib_warn(dev, "Send SMP MADs is not allowed\n");
+				err = -EPERM;
+				*bad_wr = wr;
+				goto out;
+			}
+			/* fall through */
+		case MLX5_IB_QPT_HW_GSI:
+			set_datagram_seg(seg, wr);
+			seg += sizeof(struct mlx5_wqe_datagram_seg);
+			size += sizeof(struct mlx5_wqe_datagram_seg) / 16;
+			if (unlikely((seg == qend)))
+				seg = mlx5_get_send_wqe(qp, 0);
+			break;
+		case IB_QPT_UD:
+			set_datagram_seg(seg, wr);
+			seg += sizeof(struct mlx5_wqe_datagram_seg);
+			size += sizeof(struct mlx5_wqe_datagram_seg) / 16;
+
+			if (unlikely((seg == qend)))
+				seg = mlx5_get_send_wqe(qp, 0);
+
+			/* handle qp that supports ud offload */
+			if (qp->flags & IB_QP_CREATE_IPOIB_UD_LSO) {
+				struct mlx5_wqe_eth_pad *pad;
+
+				pad = seg;
+				memset(pad, 0, sizeof(struct mlx5_wqe_eth_pad));
+				seg += sizeof(struct mlx5_wqe_eth_pad);
+				size += sizeof(struct mlx5_wqe_eth_pad) / 16;
+
+				seg = set_eth_seg(seg, wr, qend, qp, &size);
+
+				if (unlikely((seg == qend)))
+					seg = mlx5_get_send_wqe(qp, 0);
+			}
+			break;
+		case MLX5_IB_QPT_REG_UMR:
+			if (wr->opcode != MLX5_IB_WR_UMR) {
+				err = -EINVAL;
+				mlx5_ib_warn(dev, "bad opcode\n");
+				goto out;
+			}
+			qp->sq.wr_data[idx] = MLX5_IB_WR_UMR;
+			ctrl->imm = cpu_to_be32(umr_wr(wr)->mkey);
+			err = set_reg_umr_segment(dev, seg, wr, !!(MLX5_CAP_GEN(mdev, atomic)));
+			if (unlikely(err))
+				goto out;
+			seg += sizeof(struct mlx5_wqe_umr_ctrl_seg);
+			size += sizeof(struct mlx5_wqe_umr_ctrl_seg) / 16;
+			if (unlikely((seg == qend)))
+				seg = mlx5_get_send_wqe(qp, 0);
+			set_reg_mkey_segment(seg, wr);
+			seg += sizeof(struct mlx5_mkey_seg);
+			size += sizeof(struct mlx5_mkey_seg) / 16;
+			if (unlikely((seg == qend)))
+				seg = mlx5_get_send_wqe(qp, 0);
+			break;
+
+		default:
+			break;
+		}
+
+		if (wr->send_flags & IB_SEND_INLINE && num_sge) {
+			int uninitialized_var(sz);
+
+			err = set_data_inl_seg(qp, wr, seg, &sz);
+			if (unlikely(err)) {
+				mlx5_ib_warn(dev, "\n");
+				*bad_wr = wr;
+				goto out;
+			}
+			size += sz;
+		} else {
+			dpseg = seg;
+			for (i = 0; i < num_sge; i++) {
+				if (unlikely(dpseg == qend)) {
+					seg = mlx5_get_send_wqe(qp, 0);
+					dpseg = seg;
+				}
+				if (likely(wr->sg_list[i].length)) {
+					set_data_ptr_seg(dpseg, wr->sg_list + i);
+					size += sizeof(struct mlx5_wqe_data_seg) / 16;
+					dpseg++;
+				}
+			}
+		}
+
+		qp->next_fence = next_fence;
+		finish_wqe(qp, ctrl, size, idx, wr->wr_id, nreq, fence,
+			   mlx5_ib_opcode[wr->opcode]);
+skip_psv:
+		if (0)
+			dump_wqe(qp, idx, size);
+	}
+
+out:
+	if (likely(nreq)) {
+		qp->sq.head += nreq;
+
+		/* 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();
+
+		/* currently we support only regular doorbells */
+		mlx5_write64((__be32 *)ctrl, bf->bfreg->map + bf->offset, NULL);
+		/* Make sure doorbells don't leak out of SQ spinlock
+		 * and reach the HCA out of order.
+		 */
+		mmiowb();
+		bf->offset ^= bf->buf_size;
+	}
+
+	spin_unlock_irqrestore(&qp->sq.lock, flags);
+
+	return err;
+}
+
+int mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr,
+		      const struct ib_send_wr **bad_wr)
+{
+	return _mlx5_ib_post_send(ibqp, wr, bad_wr, false);
+}
+
+static void set_sig_seg(struct mlx5_rwqe_sig *sig, int size)
+{
+	sig->signature = calc_sig(sig, size);
+}
+
+static int _mlx5_ib_post_recv(struct ib_qp *ibqp, const struct ib_recv_wr *wr,
+		      const struct ib_recv_wr **bad_wr, bool drain)
+{
+	struct mlx5_ib_qp *qp = to_mqp(ibqp);
+	struct mlx5_wqe_data_seg *scat;
+	struct mlx5_rwqe_sig *sig;
+	struct mlx5_ib_dev *dev = to_mdev(ibqp->device);
+	struct mlx5_core_dev *mdev = dev->mdev;
+	unsigned long flags;
+	int err = 0;
+	int nreq;
+	int ind;
+	int i;
+
+	if (unlikely(mdev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR &&
+		     !drain)) {
+		*bad_wr = wr;
+		return -EIO;
+	}
+
+	if (unlikely(ibqp->qp_type == IB_QPT_GSI))
+		return mlx5_ib_gsi_post_recv(ibqp, wr, bad_wr);
+
+	spin_lock_irqsave(&qp->rq.lock, flags);
+
+	ind = qp->rq.head & (qp->rq.wqe_cnt - 1);
+
+	for (nreq = 0; wr; nreq++, wr = wr->next) {
+		if (mlx5_wq_overflow(&qp->rq, nreq, qp->ibqp.recv_cq)) {
+			err = -ENOMEM;
+			*bad_wr = wr;
+			goto out;
+		}
+
+		if (unlikely(wr->num_sge > qp->rq.max_gs)) {
+			err = -EINVAL;
+			*bad_wr = wr;
+			goto out;
+		}
+
+		scat = get_recv_wqe(qp, ind);
+		if (qp->wq_sig)
+			scat++;
+
+		for (i = 0; i < wr->num_sge; i++)
+			set_data_ptr_seg(scat + i, wr->sg_list + i);
+
+		if (i < qp->rq.max_gs) {
+			scat[i].byte_count = 0;
+			scat[i].lkey       = cpu_to_be32(MLX5_INVALID_LKEY);
+			scat[i].addr       = 0;
+		}
+
+		if (qp->wq_sig) {
+			sig = (struct mlx5_rwqe_sig *)scat;
+			set_sig_seg(sig, (qp->rq.max_gs + 1) << 2);
+		}
+
+		qp->rq.wrid[ind] = wr->wr_id;
+
+		ind = (ind + 1) & (qp->rq.wqe_cnt - 1);
+	}
+
+out:
+	if (likely(nreq)) {
+		qp->rq.head += nreq;
+
+		/* Make sure that descriptors are written before
+		 * doorbell record.
+		 */
+		wmb();
+
+		*qp->db.db = cpu_to_be32(qp->rq.head & 0xffff);
+	}
+
+	spin_unlock_irqrestore(&qp->rq.lock, flags);
+
+	return err;
+}
+
+int mlx5_ib_post_recv(struct ib_qp *ibqp, const struct ib_recv_wr *wr,
+		      const struct ib_recv_wr **bad_wr)
+{
+	return _mlx5_ib_post_recv(ibqp, wr, bad_wr, false);
+}
+
+static inline enum ib_qp_state to_ib_qp_state(enum mlx5_qp_state mlx5_state)
+{
+	switch (mlx5_state) {
+	case MLX5_QP_STATE_RST:      return IB_QPS_RESET;
+	case MLX5_QP_STATE_INIT:     return IB_QPS_INIT;
+	case MLX5_QP_STATE_RTR:      return IB_QPS_RTR;
+	case MLX5_QP_STATE_RTS:      return IB_QPS_RTS;
+	case MLX5_QP_STATE_SQ_DRAINING:
+	case MLX5_QP_STATE_SQD:      return IB_QPS_SQD;
+	case MLX5_QP_STATE_SQER:     return IB_QPS_SQE;
+	case MLX5_QP_STATE_ERR:      return IB_QPS_ERR;
+	default:		     return -1;
+	}
+}
+
+static inline enum ib_mig_state to_ib_mig_state(int mlx5_mig_state)
+{
+	switch (mlx5_mig_state) {
+	case MLX5_QP_PM_ARMED:		return IB_MIG_ARMED;
+	case MLX5_QP_PM_REARM:		return IB_MIG_REARM;
+	case MLX5_QP_PM_MIGRATED:	return IB_MIG_MIGRATED;
+	default: return -1;
+	}
+}
+
+static int to_ib_qp_access_flags(int mlx5_flags)
+{
+	int ib_flags = 0;
+
+	if (mlx5_flags & MLX5_QP_BIT_RRE)
+		ib_flags |= IB_ACCESS_REMOTE_READ;
+	if (mlx5_flags & MLX5_QP_BIT_RWE)
+		ib_flags |= IB_ACCESS_REMOTE_WRITE;
+	if (mlx5_flags & MLX5_QP_BIT_RAE)
+		ib_flags |= IB_ACCESS_REMOTE_ATOMIC;
+
+	return ib_flags;
+}
+
+static void to_rdma_ah_attr(struct mlx5_ib_dev *ibdev,
+			    struct rdma_ah_attr *ah_attr,
+			    struct mlx5_qp_path *path)
+{
+
+	memset(ah_attr, 0, sizeof(*ah_attr));
+
+	if (!path->port || path->port > ibdev->num_ports)
+		return;
+
+	ah_attr->type = rdma_ah_find_type(&ibdev->ib_dev, path->port);
+
+	rdma_ah_set_port_num(ah_attr, path->port);
+	rdma_ah_set_sl(ah_attr, path->dci_cfi_prio_sl & 0xf);
+
+	rdma_ah_set_dlid(ah_attr, be16_to_cpu(path->rlid));
+	rdma_ah_set_path_bits(ah_attr, path->grh_mlid & 0x7f);
+	rdma_ah_set_static_rate(ah_attr,
+				path->static_rate ? path->static_rate - 5 : 0);
+
+	if (path->grh_mlid & (1 << 7) ||
+	    ah_attr->type == RDMA_AH_ATTR_TYPE_ROCE) {
+		u32 tc_fl = be32_to_cpu(path->tclass_flowlabel);
+
+		rdma_ah_set_grh(ah_attr, NULL,
+				tc_fl & 0xfffff,
+				path->mgid_index,
+				path->hop_limit,
+				(tc_fl >> 20) & 0xff);
+		rdma_ah_set_dgid_raw(ah_attr, path->rgid);
+	}
+}
+
+static int query_raw_packet_qp_sq_state(struct mlx5_ib_dev *dev,
+					struct mlx5_ib_sq *sq,
+					u8 *sq_state)
+{
+	int err;
+
+	err = mlx5_core_query_sq_state(dev->mdev, sq->base.mqp.qpn, sq_state);
+	if (err)
+		goto out;
+	sq->state = *sq_state;
+
+out:
+	return err;
+}
+
+static int query_raw_packet_qp_rq_state(struct mlx5_ib_dev *dev,
+					struct mlx5_ib_rq *rq,
+					u8 *rq_state)
+{
+	void *out;
+	void *rqc;
+	int inlen;
+	int err;
+
+	inlen = MLX5_ST_SZ_BYTES(query_rq_out);
+	out = kvzalloc(inlen, GFP_KERNEL);
+	if (!out)
+		return -ENOMEM;
+
+	err = mlx5_core_query_rq(dev->mdev, rq->base.mqp.qpn, out);
+	if (err)
+		goto out;
+
+	rqc = MLX5_ADDR_OF(query_rq_out, out, rq_context);
+	*rq_state = MLX5_GET(rqc, rqc, state);
+	rq->state = *rq_state;
+
+out:
+	kvfree(out);
+	return err;
+}
+
+static int sqrq_state_to_qp_state(u8 sq_state, u8 rq_state,
+				  struct mlx5_ib_qp *qp, u8 *qp_state)
+{
+	static const u8 sqrq_trans[MLX5_RQ_NUM_STATE][MLX5_SQ_NUM_STATE] = {
+		[MLX5_RQC_STATE_RST] = {
+			[MLX5_SQC_STATE_RST]	= IB_QPS_RESET,
+			[MLX5_SQC_STATE_RDY]	= MLX5_QP_STATE_BAD,
+			[MLX5_SQC_STATE_ERR]	= MLX5_QP_STATE_BAD,
+			[MLX5_SQ_STATE_NA]	= IB_QPS_RESET,
+		},
+		[MLX5_RQC_STATE_RDY] = {
+			[MLX5_SQC_STATE_RST]	= MLX5_QP_STATE_BAD,
+			[MLX5_SQC_STATE_RDY]	= MLX5_QP_STATE,
+			[MLX5_SQC_STATE_ERR]	= IB_QPS_SQE,
+			[MLX5_SQ_STATE_NA]	= MLX5_QP_STATE,
+		},
+		[MLX5_RQC_STATE_ERR] = {
+			[MLX5_SQC_STATE_RST]    = MLX5_QP_STATE_BAD,
+			[MLX5_SQC_STATE_RDY]	= MLX5_QP_STATE_BAD,
+			[MLX5_SQC_STATE_ERR]	= IB_QPS_ERR,
+			[MLX5_SQ_STATE_NA]	= IB_QPS_ERR,
+		},
+		[MLX5_RQ_STATE_NA] = {
+			[MLX5_SQC_STATE_RST]    = IB_QPS_RESET,
+			[MLX5_SQC_STATE_RDY]	= MLX5_QP_STATE,
+			[MLX5_SQC_STATE_ERR]	= MLX5_QP_STATE,
+			[MLX5_SQ_STATE_NA]	= MLX5_QP_STATE_BAD,
+		},
+	};
+
+	*qp_state = sqrq_trans[rq_state][sq_state];
+
+	if (*qp_state == MLX5_QP_STATE_BAD) {
+		WARN(1, "Buggy Raw Packet QP state, SQ 0x%x state: 0x%x, RQ 0x%x state: 0x%x",
+		     qp->raw_packet_qp.sq.base.mqp.qpn, sq_state,
+		     qp->raw_packet_qp.rq.base.mqp.qpn, rq_state);
+		return -EINVAL;
+	}
+
+	if (*qp_state == MLX5_QP_STATE)
+		*qp_state = qp->state;
+
+	return 0;
+}
+
+static int query_raw_packet_qp_state(struct mlx5_ib_dev *dev,
+				     struct mlx5_ib_qp *qp,
+				     u8 *raw_packet_qp_state)
+{
+	struct mlx5_ib_raw_packet_qp *raw_packet_qp = &qp->raw_packet_qp;
+	struct mlx5_ib_sq *sq = &raw_packet_qp->sq;
+	struct mlx5_ib_rq *rq = &raw_packet_qp->rq;
+	int err;
+	u8 sq_state = MLX5_SQ_STATE_NA;
+	u8 rq_state = MLX5_RQ_STATE_NA;
+
+	if (qp->sq.wqe_cnt) {
+		err = query_raw_packet_qp_sq_state(dev, sq, &sq_state);
+		if (err)
+			return err;
+	}
+
+	if (qp->rq.wqe_cnt) {
+		err = query_raw_packet_qp_rq_state(dev, rq, &rq_state);
+		if (err)
+			return err;
+	}
+
+	return sqrq_state_to_qp_state(sq_state, rq_state, qp,
+				      raw_packet_qp_state);
+}
+
+static int query_qp_attr(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
+			 struct ib_qp_attr *qp_attr)
+{
+	int outlen = MLX5_ST_SZ_BYTES(query_qp_out);
+	struct mlx5_qp_context *context;
+	int mlx5_state;
+	u32 *outb;
+	int err = 0;
+
+	outb = kzalloc(outlen, GFP_KERNEL);
+	if (!outb)
+		return -ENOMEM;
+
+	err = mlx5_core_qp_query(dev->mdev, &qp->trans_qp.base.mqp, outb,
+				 outlen);
+	if (err)
+		goto out;
+
+	/* FIXME: use MLX5_GET rather than mlx5_qp_context manual struct */
+	context = (struct mlx5_qp_context *)MLX5_ADDR_OF(query_qp_out, outb, qpc);
+
+	mlx5_state = be32_to_cpu(context->flags) >> 28;
+
+	qp->state		     = to_ib_qp_state(mlx5_state);
+	qp_attr->path_mtu	     = context->mtu_msgmax >> 5;
+	qp_attr->path_mig_state	     =
+		to_ib_mig_state((be32_to_cpu(context->flags) >> 11) & 0x3);
+	qp_attr->qkey		     = be32_to_cpu(context->qkey);
+	qp_attr->rq_psn		     = be32_to_cpu(context->rnr_nextrecvpsn) & 0xffffff;
+	qp_attr->sq_psn		     = be32_to_cpu(context->next_send_psn) & 0xffffff;
+	qp_attr->dest_qp_num	     = be32_to_cpu(context->log_pg_sz_remote_qpn) & 0xffffff;
+	qp_attr->qp_access_flags     =
+		to_ib_qp_access_flags(be32_to_cpu(context->params2));
+
+	if (qp->ibqp.qp_type == IB_QPT_RC || qp->ibqp.qp_type == IB_QPT_UC) {
+		to_rdma_ah_attr(dev, &qp_attr->ah_attr, &context->pri_path);
+		to_rdma_ah_attr(dev, &qp_attr->alt_ah_attr, &context->alt_path);
+		qp_attr->alt_pkey_index =
+			be16_to_cpu(context->alt_path.pkey_index);
+		qp_attr->alt_port_num	=
+			rdma_ah_get_port_num(&qp_attr->alt_ah_attr);
+	}
+
+	qp_attr->pkey_index = be16_to_cpu(context->pri_path.pkey_index);
+	qp_attr->port_num = context->pri_path.port;
+
+	/* qp_attr->en_sqd_async_notify is only applicable in modify qp */
+	qp_attr->sq_draining = mlx5_state == MLX5_QP_STATE_SQ_DRAINING;
+
+	qp_attr->max_rd_atomic = 1 << ((be32_to_cpu(context->params1) >> 21) & 0x7);
+
+	qp_attr->max_dest_rd_atomic =
+		1 << ((be32_to_cpu(context->params2) >> 21) & 0x7);
+	qp_attr->min_rnr_timer	    =
+		(be32_to_cpu(context->rnr_nextrecvpsn) >> 24) & 0x1f;
+	qp_attr->timeout	    = context->pri_path.ackto_lt >> 3;
+	qp_attr->retry_cnt	    = (be32_to_cpu(context->params1) >> 16) & 0x7;
+	qp_attr->rnr_retry	    = (be32_to_cpu(context->params1) >> 13) & 0x7;
+	qp_attr->alt_timeout	    = context->alt_path.ackto_lt >> 3;
+
+out:
+	kfree(outb);
+	return err;
+}
+
+static int mlx5_ib_dct_query_qp(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *mqp,
+				struct ib_qp_attr *qp_attr, int qp_attr_mask,
+				struct ib_qp_init_attr *qp_init_attr)
+{
+	struct mlx5_core_dct	*dct = &mqp->dct.mdct;
+	u32 *out;
+	u32 access_flags = 0;
+	int outlen = MLX5_ST_SZ_BYTES(query_dct_out);
+	void *dctc;
+	int err;
+	int supported_mask = IB_QP_STATE |
+			     IB_QP_ACCESS_FLAGS |
+			     IB_QP_PORT |
+			     IB_QP_MIN_RNR_TIMER |
+			     IB_QP_AV |
+			     IB_QP_PATH_MTU |
+			     IB_QP_PKEY_INDEX;
+
+	if (qp_attr_mask & ~supported_mask)
+		return -EINVAL;
+	if (mqp->state != IB_QPS_RTR)
+		return -EINVAL;
+
+	out = kzalloc(outlen, GFP_KERNEL);
+	if (!out)
+		return -ENOMEM;
+
+	err = mlx5_core_dct_query(dev->mdev, dct, out, outlen);
+	if (err)
+		goto out;
+
+	dctc = MLX5_ADDR_OF(query_dct_out, out, dct_context_entry);
+
+	if (qp_attr_mask & IB_QP_STATE)
+		qp_attr->qp_state = IB_QPS_RTR;
+
+	if (qp_attr_mask & IB_QP_ACCESS_FLAGS) {
+		if (MLX5_GET(dctc, dctc, rre))
+			access_flags |= IB_ACCESS_REMOTE_READ;
+		if (MLX5_GET(dctc, dctc, rwe))
+			access_flags |= IB_ACCESS_REMOTE_WRITE;
+		if (MLX5_GET(dctc, dctc, rae))
+			access_flags |= IB_ACCESS_REMOTE_ATOMIC;
+		qp_attr->qp_access_flags = access_flags;
+	}
+
+	if (qp_attr_mask & IB_QP_PORT)
+		qp_attr->port_num = MLX5_GET(dctc, dctc, port);
+	if (qp_attr_mask & IB_QP_MIN_RNR_TIMER)
+		qp_attr->min_rnr_timer = MLX5_GET(dctc, dctc, min_rnr_nak);
+	if (qp_attr_mask & IB_QP_AV) {
+		qp_attr->ah_attr.grh.traffic_class = MLX5_GET(dctc, dctc, tclass);
+		qp_attr->ah_attr.grh.flow_label = MLX5_GET(dctc, dctc, flow_label);
+		qp_attr->ah_attr.grh.sgid_index = MLX5_GET(dctc, dctc, my_addr_index);
+		qp_attr->ah_attr.grh.hop_limit = MLX5_GET(dctc, dctc, hop_limit);
+	}
+	if (qp_attr_mask & IB_QP_PATH_MTU)
+		qp_attr->path_mtu = MLX5_GET(dctc, dctc, mtu);
+	if (qp_attr_mask & IB_QP_PKEY_INDEX)
+		qp_attr->pkey_index = MLX5_GET(dctc, dctc, pkey_index);
+out:
+	kfree(out);
+	return err;
+}
+
+int mlx5_ib_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *qp_attr,
+		     int qp_attr_mask, struct ib_qp_init_attr *qp_init_attr)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibqp->device);
+	struct mlx5_ib_qp *qp = to_mqp(ibqp);
+	int err = 0;
+	u8 raw_packet_qp_state;
+
+	if (ibqp->rwq_ind_tbl)
+		return -ENOSYS;
+
+	if (unlikely(ibqp->qp_type == IB_QPT_GSI))
+		return mlx5_ib_gsi_query_qp(ibqp, qp_attr, qp_attr_mask,
+					    qp_init_attr);
+
+	/* Not all of output fields are applicable, make sure to zero them */
+	memset(qp_init_attr, 0, sizeof(*qp_init_attr));
+	memset(qp_attr, 0, sizeof(*qp_attr));
+
+	if (unlikely(qp->qp_sub_type == MLX5_IB_QPT_DCT))
+		return mlx5_ib_dct_query_qp(dev, qp, qp_attr,
+					    qp_attr_mask, qp_init_attr);
+
+	mutex_lock(&qp->mutex);
+
+	if (qp->ibqp.qp_type == IB_QPT_RAW_PACKET ||
+	    qp->flags & MLX5_IB_QP_UNDERLAY) {
+		err = query_raw_packet_qp_state(dev, qp, &raw_packet_qp_state);
+		if (err)
+			goto out;
+		qp->state = raw_packet_qp_state;
+		qp_attr->port_num = 1;
+	} else {
+		err = query_qp_attr(dev, qp, qp_attr);
+		if (err)
+			goto out;
+	}
+
+	qp_attr->qp_state	     = qp->state;
+	qp_attr->cur_qp_state	     = qp_attr->qp_state;
+	qp_attr->cap.max_recv_wr     = qp->rq.wqe_cnt;
+	qp_attr->cap.max_recv_sge    = qp->rq.max_gs;
+
+	if (!ibqp->uobject) {
+		qp_attr->cap.max_send_wr  = qp->sq.max_post;
+		qp_attr->cap.max_send_sge = qp->sq.max_gs;
+		qp_init_attr->qp_context = ibqp->qp_context;
+	} else {
+		qp_attr->cap.max_send_wr  = 0;
+		qp_attr->cap.max_send_sge = 0;
+	}
+
+	qp_init_attr->qp_type = ibqp->qp_type;
+	qp_init_attr->recv_cq = ibqp->recv_cq;
+	qp_init_attr->send_cq = ibqp->send_cq;
+	qp_init_attr->srq = ibqp->srq;
+	qp_attr->cap.max_inline_data = qp->max_inline_data;
+
+	qp_init_attr->cap	     = qp_attr->cap;
+
+	qp_init_attr->create_flags = 0;
+	if (qp->flags & MLX5_IB_QP_BLOCK_MULTICAST_LOOPBACK)
+		qp_init_attr->create_flags |= IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK;
+
+	if (qp->flags & MLX5_IB_QP_CROSS_CHANNEL)
+		qp_init_attr->create_flags |= IB_QP_CREATE_CROSS_CHANNEL;
+	if (qp->flags & MLX5_IB_QP_MANAGED_SEND)
+		qp_init_attr->create_flags |= IB_QP_CREATE_MANAGED_SEND;
+	if (qp->flags & MLX5_IB_QP_MANAGED_RECV)
+		qp_init_attr->create_flags |= IB_QP_CREATE_MANAGED_RECV;
+	if (qp->flags & MLX5_IB_QP_SQPN_QP1)
+		qp_init_attr->create_flags |= mlx5_ib_create_qp_sqpn_qp1();
+
+	qp_init_attr->sq_sig_type = qp->sq_signal_bits & MLX5_WQE_CTRL_CQ_UPDATE ?
+		IB_SIGNAL_ALL_WR : IB_SIGNAL_REQ_WR;
+
+out:
+	mutex_unlock(&qp->mutex);
+	return err;
+}
+
+struct ib_xrcd *mlx5_ib_alloc_xrcd(struct ib_device *ibdev,
+					  struct ib_ucontext *context,
+					  struct ib_udata *udata)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+	struct mlx5_ib_xrcd *xrcd;
+	int err;
+
+	if (!MLX5_CAP_GEN(dev->mdev, xrc))
+		return ERR_PTR(-ENOSYS);
+
+	xrcd = kmalloc(sizeof(*xrcd), GFP_KERNEL);
+	if (!xrcd)
+		return ERR_PTR(-ENOMEM);
+
+	err = mlx5_core_xrcd_alloc(dev->mdev, &xrcd->xrcdn);
+	if (err) {
+		kfree(xrcd);
+		return ERR_PTR(-ENOMEM);
+	}
+
+	return &xrcd->ibxrcd;
+}
+
+int mlx5_ib_dealloc_xrcd(struct ib_xrcd *xrcd)
+{
+	struct mlx5_ib_dev *dev = to_mdev(xrcd->device);
+	u32 xrcdn = to_mxrcd(xrcd)->xrcdn;
+	int err;
+
+	err = mlx5_core_xrcd_dealloc(dev->mdev, xrcdn);
+	if (err)
+		mlx5_ib_warn(dev, "failed to dealloc xrcdn 0x%x\n", xrcdn);
+
+	kfree(xrcd);
+	return 0;
+}
+
+static void mlx5_ib_wq_event(struct mlx5_core_qp *core_qp, int type)
+{
+	struct mlx5_ib_rwq *rwq = to_mibrwq(core_qp);
+	struct mlx5_ib_dev *dev = to_mdev(rwq->ibwq.device);
+	struct ib_event event;
+
+	if (rwq->ibwq.event_handler) {
+		event.device     = rwq->ibwq.device;
+		event.element.wq = &rwq->ibwq;
+		switch (type) {
+		case MLX5_EVENT_TYPE_WQ_CATAS_ERROR:
+			event.event = IB_EVENT_WQ_FATAL;
+			break;
+		default:
+			mlx5_ib_warn(dev, "Unexpected event type %d on WQ %06x\n", type, core_qp->qpn);
+			return;
+		}
+
+		rwq->ibwq.event_handler(&event, rwq->ibwq.wq_context);
+	}
+}
+
+static int set_delay_drop(struct mlx5_ib_dev *dev)
+{
+	int err = 0;
+
+	mutex_lock(&dev->delay_drop.lock);
+	if (dev->delay_drop.activate)
+		goto out;
+
+	err = mlx5_core_set_delay_drop(dev->mdev, dev->delay_drop.timeout);
+	if (err)
+		goto out;
+
+	dev->delay_drop.activate = true;
+out:
+	mutex_unlock(&dev->delay_drop.lock);
+
+	if (!err)
+		atomic_inc(&dev->delay_drop.rqs_cnt);
+	return err;
+}
+
+static int  create_rq(struct mlx5_ib_rwq *rwq, struct ib_pd *pd,
+		      struct ib_wq_init_attr *init_attr)
+{
+	struct mlx5_ib_dev *dev;
+	int has_net_offloads;
+	__be64 *rq_pas0;
+	void *in;
+	void *rqc;
+	void *wq;
+	int inlen;
+	int err;
+
+	dev = to_mdev(pd->device);
+
+	inlen = MLX5_ST_SZ_BYTES(create_rq_in) + sizeof(u64) * rwq->rq_num_pas;
+	in = kvzalloc(inlen, GFP_KERNEL);
+	if (!in)
+		return -ENOMEM;
+
+	rqc = MLX5_ADDR_OF(create_rq_in, in, ctx);
+	MLX5_SET(rqc,  rqc, mem_rq_type,
+		 MLX5_RQC_MEM_RQ_TYPE_MEMORY_RQ_INLINE);
+	MLX5_SET(rqc, rqc, user_index, rwq->user_index);
+	MLX5_SET(rqc,  rqc, cqn, to_mcq(init_attr->cq)->mcq.cqn);
+	MLX5_SET(rqc,  rqc, state, MLX5_RQC_STATE_RST);
+	MLX5_SET(rqc,  rqc, flush_in_error_en, 1);
+	wq = MLX5_ADDR_OF(rqc, rqc, wq);
+	MLX5_SET(wq, wq, wq_type,
+		 rwq->create_flags & MLX5_IB_WQ_FLAGS_STRIDING_RQ ?
+		 MLX5_WQ_TYPE_CYCLIC_STRIDING_RQ : MLX5_WQ_TYPE_CYCLIC);
+	if (init_attr->create_flags & IB_WQ_FLAGS_PCI_WRITE_END_PADDING) {
+		if (!MLX5_CAP_GEN(dev->mdev, end_pad)) {
+			mlx5_ib_dbg(dev, "Scatter end padding is not supported\n");
+			err = -EOPNOTSUPP;
+			goto out;
+		} else {
+			MLX5_SET(wq, wq, end_padding_mode, MLX5_WQ_END_PAD_MODE_ALIGN);
+		}
+	}
+	MLX5_SET(wq, wq, log_wq_stride, rwq->log_rq_stride);
+	if (rwq->create_flags & MLX5_IB_WQ_FLAGS_STRIDING_RQ) {
+		MLX5_SET(wq, wq, two_byte_shift_en, rwq->two_byte_shift_en);
+		MLX5_SET(wq, wq, log_wqe_stride_size,
+			 rwq->single_stride_log_num_of_bytes -
+			 MLX5_MIN_SINGLE_STRIDE_LOG_NUM_BYTES);
+		MLX5_SET(wq, wq, log_wqe_num_of_strides, rwq->log_num_strides -
+			 MLX5_MIN_SINGLE_WQE_LOG_NUM_STRIDES);
+	}
+	MLX5_SET(wq, wq, log_wq_sz, rwq->log_rq_size);
+	MLX5_SET(wq, wq, pd, to_mpd(pd)->pdn);
+	MLX5_SET(wq, wq, page_offset, rwq->rq_page_offset);
+	MLX5_SET(wq, wq, log_wq_pg_sz, rwq->log_page_size);
+	MLX5_SET(wq, wq, wq_signature, rwq->wq_sig);
+	MLX5_SET64(wq, wq, dbr_addr, rwq->db.dma);
+	has_net_offloads = MLX5_CAP_GEN(dev->mdev, eth_net_offloads);
+	if (init_attr->create_flags & IB_WQ_FLAGS_CVLAN_STRIPPING) {
+		if (!(has_net_offloads && MLX5_CAP_ETH(dev->mdev, vlan_cap))) {
+			mlx5_ib_dbg(dev, "VLAN offloads are not supported\n");
+			err = -EOPNOTSUPP;
+			goto out;
+		}
+	} else {
+		MLX5_SET(rqc, rqc, vsd, 1);
+	}
+	if (init_attr->create_flags & IB_WQ_FLAGS_SCATTER_FCS) {
+		if (!(has_net_offloads && MLX5_CAP_ETH(dev->mdev, scatter_fcs))) {
+			mlx5_ib_dbg(dev, "Scatter FCS is not supported\n");
+			err = -EOPNOTSUPP;
+			goto out;
+		}
+		MLX5_SET(rqc, rqc, scatter_fcs, 1);
+	}
+	if (init_attr->create_flags & IB_WQ_FLAGS_DELAY_DROP) {
+		if (!(dev->ib_dev.attrs.raw_packet_caps &
+		      IB_RAW_PACKET_CAP_DELAY_DROP)) {
+			mlx5_ib_dbg(dev, "Delay drop is not supported\n");
+			err = -EOPNOTSUPP;
+			goto out;
+		}
+		MLX5_SET(rqc, rqc, delay_drop_en, 1);
+	}
+	rq_pas0 = (__be64 *)MLX5_ADDR_OF(wq, wq, pas);
+	mlx5_ib_populate_pas(dev, rwq->umem, rwq->page_shift, rq_pas0, 0);
+	err = mlx5_core_create_rq_tracked(dev->mdev, in, inlen, &rwq->core_qp);
+	if (!err && init_attr->create_flags & IB_WQ_FLAGS_DELAY_DROP) {
+		err = set_delay_drop(dev);
+		if (err) {
+			mlx5_ib_warn(dev, "Failed to enable delay drop err=%d\n",
+				     err);
+			mlx5_core_destroy_rq_tracked(dev->mdev, &rwq->core_qp);
+		} else {
+			rwq->create_flags |= MLX5_IB_WQ_FLAGS_DELAY_DROP;
+		}
+	}
+out:
+	kvfree(in);
+	return err;
+}
+
+static int set_user_rq_size(struct mlx5_ib_dev *dev,
+			    struct ib_wq_init_attr *wq_init_attr,
+			    struct mlx5_ib_create_wq *ucmd,
+			    struct mlx5_ib_rwq *rwq)
+{
+	/* Sanity check RQ size before proceeding */
+	if (wq_init_attr->max_wr > (1 << MLX5_CAP_GEN(dev->mdev, log_max_wq_sz)))
+		return -EINVAL;
+
+	if (!ucmd->rq_wqe_count)
+		return -EINVAL;
+
+	rwq->wqe_count = ucmd->rq_wqe_count;
+	rwq->wqe_shift = ucmd->rq_wqe_shift;
+	if (check_shl_overflow(rwq->wqe_count, rwq->wqe_shift, &rwq->buf_size))
+		return -EINVAL;
+
+	rwq->log_rq_stride = rwq->wqe_shift;
+	rwq->log_rq_size = ilog2(rwq->wqe_count);
+	return 0;
+}
+
+static int prepare_user_rq(struct ib_pd *pd,
+			   struct ib_wq_init_attr *init_attr,
+			   struct ib_udata *udata,
+			   struct mlx5_ib_rwq *rwq)
+{
+	struct mlx5_ib_dev *dev = to_mdev(pd->device);
+	struct mlx5_ib_create_wq ucmd = {};
+	int err;
+	size_t required_cmd_sz;
+
+	required_cmd_sz = offsetof(typeof(ucmd), single_stride_log_num_of_bytes)
+		+ sizeof(ucmd.single_stride_log_num_of_bytes);
+	if (udata->inlen < required_cmd_sz) {
+		mlx5_ib_dbg(dev, "invalid inlen\n");
+		return -EINVAL;
+	}
+
+	if (udata->inlen > sizeof(ucmd) &&
+	    !ib_is_udata_cleared(udata, sizeof(ucmd),
+				 udata->inlen - sizeof(ucmd))) {
+		mlx5_ib_dbg(dev, "inlen is not supported\n");
+		return -EOPNOTSUPP;
+	}
+
+	if (ib_copy_from_udata(&ucmd, udata, min(sizeof(ucmd), udata->inlen))) {
+		mlx5_ib_dbg(dev, "copy failed\n");
+		return -EFAULT;
+	}
+
+	if (ucmd.comp_mask & (~MLX5_IB_CREATE_WQ_STRIDING_RQ)) {
+		mlx5_ib_dbg(dev, "invalid comp mask\n");
+		return -EOPNOTSUPP;
+	} else if (ucmd.comp_mask & MLX5_IB_CREATE_WQ_STRIDING_RQ) {
+		if (!MLX5_CAP_GEN(dev->mdev, striding_rq)) {
+			mlx5_ib_dbg(dev, "Striding RQ is not supported\n");
+			return -EOPNOTSUPP;
+		}
+		if ((ucmd.single_stride_log_num_of_bytes <
+		    MLX5_MIN_SINGLE_STRIDE_LOG_NUM_BYTES) ||
+		    (ucmd.single_stride_log_num_of_bytes >
+		     MLX5_MAX_SINGLE_STRIDE_LOG_NUM_BYTES)) {
+			mlx5_ib_dbg(dev, "Invalid log stride size (%u. Range is %u - %u)\n",
+				    ucmd.single_stride_log_num_of_bytes,
+				    MLX5_MIN_SINGLE_STRIDE_LOG_NUM_BYTES,
+				    MLX5_MAX_SINGLE_STRIDE_LOG_NUM_BYTES);
+			return -EINVAL;
+		}
+		if ((ucmd.single_wqe_log_num_of_strides >
+		    MLX5_MAX_SINGLE_WQE_LOG_NUM_STRIDES) ||
+		     (ucmd.single_wqe_log_num_of_strides <
+			MLX5_MIN_SINGLE_WQE_LOG_NUM_STRIDES)) {
+			mlx5_ib_dbg(dev, "Invalid log num strides (%u. Range is %u - %u)\n",
+				    ucmd.single_wqe_log_num_of_strides,
+				    MLX5_MIN_SINGLE_WQE_LOG_NUM_STRIDES,
+				    MLX5_MAX_SINGLE_WQE_LOG_NUM_STRIDES);
+			return -EINVAL;
+		}
+		rwq->single_stride_log_num_of_bytes =
+			ucmd.single_stride_log_num_of_bytes;
+		rwq->log_num_strides = ucmd.single_wqe_log_num_of_strides;
+		rwq->two_byte_shift_en = !!ucmd.two_byte_shift_en;
+		rwq->create_flags |= MLX5_IB_WQ_FLAGS_STRIDING_RQ;
+	}
+
+	err = set_user_rq_size(dev, init_attr, &ucmd, rwq);
+	if (err) {
+		mlx5_ib_dbg(dev, "err %d\n", err);
+		return err;
+	}
+
+	err = create_user_rq(dev, pd, rwq, &ucmd);
+	if (err) {
+		mlx5_ib_dbg(dev, "err %d\n", err);
+		if (err)
+			return err;
+	}
+
+	rwq->user_index = ucmd.user_index;
+	return 0;
+}
+
+struct ib_wq *mlx5_ib_create_wq(struct ib_pd *pd,
+				struct ib_wq_init_attr *init_attr,
+				struct ib_udata *udata)
+{
+	struct mlx5_ib_dev *dev;
+	struct mlx5_ib_rwq *rwq;
+	struct mlx5_ib_create_wq_resp resp = {};
+	size_t min_resp_len;
+	int err;
+
+	if (!udata)
+		return ERR_PTR(-ENOSYS);
+
+	min_resp_len = offsetof(typeof(resp), reserved) + sizeof(resp.reserved);
+	if (udata->outlen && udata->outlen < min_resp_len)
+		return ERR_PTR(-EINVAL);
+
+	if (!capable(CAP_SYS_RAWIO) &&
+	    init_attr->create_flags & IB_WQ_FLAGS_DELAY_DROP)
+		return ERR_PTR(-EPERM);
+
+	dev = to_mdev(pd->device);
+	switch (init_attr->wq_type) {
+	case IB_WQT_RQ:
+		rwq = kzalloc(sizeof(*rwq), GFP_KERNEL);
+		if (!rwq)
+			return ERR_PTR(-ENOMEM);
+		err = prepare_user_rq(pd, init_attr, udata, rwq);
+		if (err)
+			goto err;
+		err = create_rq(rwq, pd, init_attr);
+		if (err)
+			goto err_user_rq;
+		break;
+	default:
+		mlx5_ib_dbg(dev, "unsupported wq type %d\n",
+			    init_attr->wq_type);
+		return ERR_PTR(-EINVAL);
+	}
+
+	rwq->ibwq.wq_num = rwq->core_qp.qpn;
+	rwq->ibwq.state = IB_WQS_RESET;
+	if (udata->outlen) {
+		resp.response_length = offsetof(typeof(resp), response_length) +
+				sizeof(resp.response_length);
+		err = ib_copy_to_udata(udata, &resp, resp.response_length);
+		if (err)
+			goto err_copy;
+	}
+
+	rwq->core_qp.event = mlx5_ib_wq_event;
+	rwq->ibwq.event_handler = init_attr->event_handler;
+	return &rwq->ibwq;
+
+err_copy:
+	mlx5_core_destroy_rq_tracked(dev->mdev, &rwq->core_qp);
+err_user_rq:
+	destroy_user_rq(dev, pd, rwq);
+err:
+	kfree(rwq);
+	return ERR_PTR(err);
+}
+
+int mlx5_ib_destroy_wq(struct ib_wq *wq)
+{
+	struct mlx5_ib_dev *dev = to_mdev(wq->device);
+	struct mlx5_ib_rwq *rwq = to_mrwq(wq);
+
+	mlx5_core_destroy_rq_tracked(dev->mdev, &rwq->core_qp);
+	destroy_user_rq(dev, wq->pd, rwq);
+	kfree(rwq);
+
+	return 0;
+}
+
+struct ib_rwq_ind_table *mlx5_ib_create_rwq_ind_table(struct ib_device *device,
+						      struct ib_rwq_ind_table_init_attr *init_attr,
+						      struct ib_udata *udata)
+{
+	struct mlx5_ib_dev *dev = to_mdev(device);
+	struct mlx5_ib_rwq_ind_table *rwq_ind_tbl;
+	int sz = 1 << init_attr->log_ind_tbl_size;
+	struct mlx5_ib_create_rwq_ind_tbl_resp resp = {};
+	size_t min_resp_len;
+	int inlen;
+	int err;
+	int i;
+	u32 *in;
+	void *rqtc;
+
+	if (udata->inlen > 0 &&
+	    !ib_is_udata_cleared(udata, 0,
+				 udata->inlen))
+		return ERR_PTR(-EOPNOTSUPP);
+
+	if (init_attr->log_ind_tbl_size >
+	    MLX5_CAP_GEN(dev->mdev, log_max_rqt_size)) {
+		mlx5_ib_dbg(dev, "log_ind_tbl_size = %d is bigger than supported = %d\n",
+			    init_attr->log_ind_tbl_size,
+			    MLX5_CAP_GEN(dev->mdev, log_max_rqt_size));
+		return ERR_PTR(-EINVAL);
+	}
+
+	min_resp_len = offsetof(typeof(resp), reserved) + sizeof(resp.reserved);
+	if (udata->outlen && udata->outlen < min_resp_len)
+		return ERR_PTR(-EINVAL);
+
+	rwq_ind_tbl = kzalloc(sizeof(*rwq_ind_tbl), GFP_KERNEL);
+	if (!rwq_ind_tbl)
+		return ERR_PTR(-ENOMEM);
+
+	inlen = MLX5_ST_SZ_BYTES(create_rqt_in) + sizeof(u32) * sz;
+	in = kvzalloc(inlen, GFP_KERNEL);
+	if (!in) {
+		err = -ENOMEM;
+		goto err;
+	}
+
+	rqtc = MLX5_ADDR_OF(create_rqt_in, in, rqt_context);
+
+	MLX5_SET(rqtc, rqtc, rqt_actual_size, sz);
+	MLX5_SET(rqtc, rqtc, rqt_max_size, sz);
+
+	for (i = 0; i < sz; i++)
+		MLX5_SET(rqtc, rqtc, rq_num[i], init_attr->ind_tbl[i]->wq_num);
+
+	err = mlx5_core_create_rqt(dev->mdev, in, inlen, &rwq_ind_tbl->rqtn);
+	kvfree(in);
+
+	if (err)
+		goto err;
+
+	rwq_ind_tbl->ib_rwq_ind_tbl.ind_tbl_num = rwq_ind_tbl->rqtn;
+	if (udata->outlen) {
+		resp.response_length = offsetof(typeof(resp), response_length) +
+					sizeof(resp.response_length);
+		err = ib_copy_to_udata(udata, &resp, resp.response_length);
+		if (err)
+			goto err_copy;
+	}
+
+	return &rwq_ind_tbl->ib_rwq_ind_tbl;
+
+err_copy:
+	mlx5_core_destroy_rqt(dev->mdev, rwq_ind_tbl->rqtn);
+err:
+	kfree(rwq_ind_tbl);
+	return ERR_PTR(err);
+}
+
+int mlx5_ib_destroy_rwq_ind_table(struct ib_rwq_ind_table *ib_rwq_ind_tbl)
+{
+	struct mlx5_ib_rwq_ind_table *rwq_ind_tbl = to_mrwq_ind_table(ib_rwq_ind_tbl);
+	struct mlx5_ib_dev *dev = to_mdev(ib_rwq_ind_tbl->device);
+
+	mlx5_core_destroy_rqt(dev->mdev, rwq_ind_tbl->rqtn);
+
+	kfree(rwq_ind_tbl);
+	return 0;
+}
+
+int mlx5_ib_modify_wq(struct ib_wq *wq, struct ib_wq_attr *wq_attr,
+		      u32 wq_attr_mask, struct ib_udata *udata)
+{
+	struct mlx5_ib_dev *dev = to_mdev(wq->device);
+	struct mlx5_ib_rwq *rwq = to_mrwq(wq);
+	struct mlx5_ib_modify_wq ucmd = {};
+	size_t required_cmd_sz;
+	int curr_wq_state;
+	int wq_state;
+	int inlen;
+	int err;
+	void *rqc;
+	void *in;
+
+	required_cmd_sz = offsetof(typeof(ucmd), reserved) + sizeof(ucmd.reserved);
+	if (udata->inlen < required_cmd_sz)
+		return -EINVAL;
+
+	if (udata->inlen > sizeof(ucmd) &&
+	    !ib_is_udata_cleared(udata, sizeof(ucmd),
+				 udata->inlen - sizeof(ucmd)))
+		return -EOPNOTSUPP;
+
+	if (ib_copy_from_udata(&ucmd, udata, min(sizeof(ucmd), udata->inlen)))
+		return -EFAULT;
+
+	if (ucmd.comp_mask || ucmd.reserved)
+		return -EOPNOTSUPP;
+
+	inlen = MLX5_ST_SZ_BYTES(modify_rq_in);
+	in = kvzalloc(inlen, GFP_KERNEL);
+	if (!in)
+		return -ENOMEM;
+
+	rqc = MLX5_ADDR_OF(modify_rq_in, in, ctx);
+
+	curr_wq_state = (wq_attr_mask & IB_WQ_CUR_STATE) ?
+		wq_attr->curr_wq_state : wq->state;
+	wq_state = (wq_attr_mask & IB_WQ_STATE) ?
+		wq_attr->wq_state : curr_wq_state;
+	if (curr_wq_state == IB_WQS_ERR)
+		curr_wq_state = MLX5_RQC_STATE_ERR;
+	if (wq_state == IB_WQS_ERR)
+		wq_state = MLX5_RQC_STATE_ERR;
+	MLX5_SET(modify_rq_in, in, rq_state, curr_wq_state);
+	MLX5_SET(rqc, rqc, state, wq_state);
+
+	if (wq_attr_mask & IB_WQ_FLAGS) {
+		if (wq_attr->flags_mask & IB_WQ_FLAGS_CVLAN_STRIPPING) {
+			if (!(MLX5_CAP_GEN(dev->mdev, eth_net_offloads) &&
+			      MLX5_CAP_ETH(dev->mdev, vlan_cap))) {
+				mlx5_ib_dbg(dev, "VLAN offloads are not "
+					    "supported\n");
+				err = -EOPNOTSUPP;
+				goto out;
+			}
+			MLX5_SET64(modify_rq_in, in, modify_bitmask,
+				   MLX5_MODIFY_RQ_IN_MODIFY_BITMASK_VSD);
+			MLX5_SET(rqc, rqc, vsd,
+				 (wq_attr->flags & IB_WQ_FLAGS_CVLAN_STRIPPING) ? 0 : 1);
+		}
+
+		if (wq_attr->flags_mask & IB_WQ_FLAGS_PCI_WRITE_END_PADDING) {
+			mlx5_ib_dbg(dev, "Modifying scatter end padding is not supported\n");
+			err = -EOPNOTSUPP;
+			goto out;
+		}
+	}
+
+	if (curr_wq_state == IB_WQS_RESET && wq_state == IB_WQS_RDY) {
+		if (MLX5_CAP_GEN(dev->mdev, modify_rq_counter_set_id)) {
+			MLX5_SET64(modify_rq_in, in, modify_bitmask,
+				   MLX5_MODIFY_RQ_IN_MODIFY_BITMASK_RQ_COUNTER_SET_ID);
+			MLX5_SET(rqc, rqc, counter_set_id,
+				 dev->port->cnts.set_id);
+		} else
+			pr_info_once("%s: Receive WQ counters are not supported on current FW\n",
+				     dev->ib_dev.name);
+	}
+
+	err = mlx5_core_modify_rq(dev->mdev, rwq->core_qp.qpn, in, inlen);
+	if (!err)
+		rwq->ibwq.state = (wq_state == MLX5_RQC_STATE_ERR) ? IB_WQS_ERR : wq_state;
+
+out:
+	kvfree(in);
+	return err;
+}
+
+struct mlx5_ib_drain_cqe {
+	struct ib_cqe cqe;
+	struct completion done;
+};
+
+static void mlx5_ib_drain_qp_done(struct ib_cq *cq, struct ib_wc *wc)
+{
+	struct mlx5_ib_drain_cqe *cqe = container_of(wc->wr_cqe,
+						     struct mlx5_ib_drain_cqe,
+						     cqe);
+
+	complete(&cqe->done);
+}
+
+/* This function returns only once the drained WR was completed */
+static void handle_drain_completion(struct ib_cq *cq,
+				    struct mlx5_ib_drain_cqe *sdrain,
+				    struct mlx5_ib_dev *dev)
+{
+	struct mlx5_core_dev *mdev = dev->mdev;
+
+	if (cq->poll_ctx == IB_POLL_DIRECT) {
+		while (wait_for_completion_timeout(&sdrain->done, HZ / 10) <= 0)
+			ib_process_cq_direct(cq, -1);
+		return;
+	}
+
+	if (mdev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR) {
+		struct mlx5_ib_cq *mcq = to_mcq(cq);
+		bool triggered = false;
+		unsigned long flags;
+
+		spin_lock_irqsave(&dev->reset_flow_resource_lock, flags);
+		/* Make sure that the CQ handler won't run if wasn't run yet */
+		if (!mcq->mcq.reset_notify_added)
+			mcq->mcq.reset_notify_added = 1;
+		else
+			triggered = true;
+		spin_unlock_irqrestore(&dev->reset_flow_resource_lock, flags);
+
+		if (triggered) {
+			/* Wait for any scheduled/running task to be ended */
+			switch (cq->poll_ctx) {
+			case IB_POLL_SOFTIRQ:
+				irq_poll_disable(&cq->iop);
+				irq_poll_enable(&cq->iop);
+				break;
+			case IB_POLL_WORKQUEUE:
+				cancel_work_sync(&cq->work);
+				break;
+			default:
+				WARN_ON_ONCE(1);
+			}
+		}
+
+		/* Run the CQ handler - this makes sure that the drain WR will
+		 * be processed if wasn't processed yet.
+		 */
+		mcq->mcq.comp(&mcq->mcq);
+	}
+
+	wait_for_completion(&sdrain->done);
+}
+
+void mlx5_ib_drain_sq(struct ib_qp *qp)
+{
+	struct ib_cq *cq = qp->send_cq;
+	struct ib_qp_attr attr = { .qp_state = IB_QPS_ERR };
+	struct mlx5_ib_drain_cqe sdrain;
+	const struct ib_send_wr *bad_swr;
+	struct ib_rdma_wr swr = {
+		.wr = {
+			.next = NULL,
+			{ .wr_cqe	= &sdrain.cqe, },
+			.opcode	= IB_WR_RDMA_WRITE,
+		},
+	};
+	int ret;
+	struct mlx5_ib_dev *dev = to_mdev(qp->device);
+	struct mlx5_core_dev *mdev = dev->mdev;
+
+	ret = ib_modify_qp(qp, &attr, IB_QP_STATE);
+	if (ret && mdev->state != MLX5_DEVICE_STATE_INTERNAL_ERROR) {
+		WARN_ONCE(ret, "failed to drain send queue: %d\n", ret);
+		return;
+	}
+
+	sdrain.cqe.done = mlx5_ib_drain_qp_done;
+	init_completion(&sdrain.done);
+
+	ret = _mlx5_ib_post_send(qp, &swr.wr, &bad_swr, true);
+	if (ret) {
+		WARN_ONCE(ret, "failed to drain send queue: %d\n", ret);
+		return;
+	}
+
+	handle_drain_completion(cq, &sdrain, dev);
+}
+
+void mlx5_ib_drain_rq(struct ib_qp *qp)
+{
+	struct ib_cq *cq = qp->recv_cq;
+	struct ib_qp_attr attr = { .qp_state = IB_QPS_ERR };
+	struct mlx5_ib_drain_cqe rdrain;
+	struct ib_recv_wr rwr = {};
+	const struct ib_recv_wr *bad_rwr;
+	int ret;
+	struct mlx5_ib_dev *dev = to_mdev(qp->device);
+	struct mlx5_core_dev *mdev = dev->mdev;
+
+	ret = ib_modify_qp(qp, &attr, IB_QP_STATE);
+	if (ret && mdev->state != MLX5_DEVICE_STATE_INTERNAL_ERROR) {
+		WARN_ONCE(ret, "failed to drain recv queue: %d\n", ret);
+		return;
+	}
+
+	rwr.wr_cqe = &rdrain.cqe;
+	rdrain.cqe.done = mlx5_ib_drain_qp_done;
+	init_completion(&rdrain.done);
+
+	ret = _mlx5_ib_post_recv(qp, &rwr, &bad_rwr, true);
+	if (ret) {
+		WARN_ONCE(ret, "failed to drain recv queue: %d\n", ret);
+		return;
+	}
+
+	handle_drain_completion(cq, &rdrain, dev);
+}
diff --git a/drivers/infiniband/hw/mlx5/srq.c b/drivers/infiniband/hw/mlx5/srq.c
new file mode 100644
index 000000000..d359fecf7
--- /dev/null
+++ b/drivers/infiniband/hw/mlx5/srq.c
@@ -0,0 +1,516 @@
+/*
+ * Copyright (c) 2013-2015, Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/module.h>
+#include <linux/mlx5/qp.h>
+#include <linux/mlx5/srq.h>
+#include <linux/slab.h>
+#include <rdma/ib_umem.h>
+#include <rdma/ib_user_verbs.h>
+
+#include "mlx5_ib.h"
+
+/* not supported currently */
+static int srq_signature;
+
+static void *get_wqe(struct mlx5_ib_srq *srq, int n)
+{
+	return mlx5_buf_offset(&srq->buf, n << srq->msrq.wqe_shift);
+}
+
+static void mlx5_ib_srq_event(struct mlx5_core_srq *srq, enum mlx5_event type)
+{
+	struct ib_event event;
+	struct ib_srq *ibsrq = &to_mibsrq(srq)->ibsrq;
+
+	if (ibsrq->event_handler) {
+		event.device      = ibsrq->device;
+		event.element.srq = ibsrq;
+		switch (type) {
+		case MLX5_EVENT_TYPE_SRQ_RQ_LIMIT:
+			event.event = IB_EVENT_SRQ_LIMIT_REACHED;
+			break;
+		case MLX5_EVENT_TYPE_SRQ_CATAS_ERROR:
+			event.event = IB_EVENT_SRQ_ERR;
+			break;
+		default:
+			pr_warn("mlx5_ib: Unexpected event type %d on SRQ %06x\n",
+				type, srq->srqn);
+			return;
+		}
+
+		ibsrq->event_handler(&event, ibsrq->srq_context);
+	}
+}
+
+static int create_srq_user(struct ib_pd *pd, struct mlx5_ib_srq *srq,
+			   struct mlx5_srq_attr *in,
+			   struct ib_udata *udata, int buf_size)
+{
+	struct mlx5_ib_dev *dev = to_mdev(pd->device);
+	struct mlx5_ib_create_srq ucmd = {};
+	size_t ucmdlen;
+	int err;
+	int npages;
+	int page_shift;
+	int ncont;
+	u32 offset;
+	u32 uidx = MLX5_IB_DEFAULT_UIDX;
+
+	ucmdlen = min(udata->inlen, sizeof(ucmd));
+
+	if (ib_copy_from_udata(&ucmd, udata, ucmdlen)) {
+		mlx5_ib_dbg(dev, "failed copy udata\n");
+		return -EFAULT;
+	}
+
+	if (ucmd.reserved0 || ucmd.reserved1)
+		return -EINVAL;
+
+	if (udata->inlen > sizeof(ucmd) &&
+	    !ib_is_udata_cleared(udata, sizeof(ucmd),
+				 udata->inlen - sizeof(ucmd)))
+		return -EINVAL;
+
+	if (in->type != IB_SRQT_BASIC) {
+		err = get_srq_user_index(to_mucontext(pd->uobject->context),
+					 &ucmd, udata->inlen, &uidx);
+		if (err)
+			return err;
+	}
+
+	srq->wq_sig = !!(ucmd.flags & MLX5_SRQ_FLAG_SIGNATURE);
+
+	srq->umem = ib_umem_get(pd->uobject->context, ucmd.buf_addr, buf_size,
+				0, 0);
+	if (IS_ERR(srq->umem)) {
+		mlx5_ib_dbg(dev, "failed umem get, size %d\n", buf_size);
+		err = PTR_ERR(srq->umem);
+		return err;
+	}
+
+	mlx5_ib_cont_pages(srq->umem, ucmd.buf_addr, 0, &npages,
+			   &page_shift, &ncont, NULL);
+	err = mlx5_ib_get_buf_offset(ucmd.buf_addr, page_shift,
+				     &offset);
+	if (err) {
+		mlx5_ib_warn(dev, "bad offset\n");
+		goto err_umem;
+	}
+
+	in->pas = kvcalloc(ncont, sizeof(*in->pas), GFP_KERNEL);
+	if (!in->pas) {
+		err = -ENOMEM;
+		goto err_umem;
+	}
+
+	mlx5_ib_populate_pas(dev, srq->umem, page_shift, in->pas, 0);
+
+	err = mlx5_ib_db_map_user(to_mucontext(pd->uobject->context),
+				  ucmd.db_addr, &srq->db);
+	if (err) {
+		mlx5_ib_dbg(dev, "map doorbell failed\n");
+		goto err_in;
+	}
+
+	in->log_page_size = page_shift - MLX5_ADAPTER_PAGE_SHIFT;
+	in->page_offset = offset;
+	if (MLX5_CAP_GEN(dev->mdev, cqe_version) == MLX5_CQE_VERSION_V1 &&
+	    in->type != IB_SRQT_BASIC)
+		in->user_index = uidx;
+
+	return 0;
+
+err_in:
+	kvfree(in->pas);
+
+err_umem:
+	ib_umem_release(srq->umem);
+
+	return err;
+}
+
+static int create_srq_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_srq *srq,
+			     struct mlx5_srq_attr *in, int buf_size)
+{
+	int err;
+	int i;
+	struct mlx5_wqe_srq_next_seg *next;
+
+	err = mlx5_db_alloc(dev->mdev, &srq->db);
+	if (err) {
+		mlx5_ib_warn(dev, "alloc dbell rec failed\n");
+		return err;
+	}
+
+	if (mlx5_buf_alloc(dev->mdev, buf_size, &srq->buf)) {
+		mlx5_ib_dbg(dev, "buf alloc failed\n");
+		err = -ENOMEM;
+		goto err_db;
+	}
+
+	srq->head    = 0;
+	srq->tail    = srq->msrq.max - 1;
+	srq->wqe_ctr = 0;
+
+	for (i = 0; i < srq->msrq.max; i++) {
+		next = get_wqe(srq, i);
+		next->next_wqe_index =
+			cpu_to_be16((i + 1) & (srq->msrq.max - 1));
+	}
+
+	mlx5_ib_dbg(dev, "srq->buf.page_shift = %d\n", srq->buf.page_shift);
+	in->pas = kvcalloc(srq->buf.npages, sizeof(*in->pas), GFP_KERNEL);
+	if (!in->pas) {
+		err = -ENOMEM;
+		goto err_buf;
+	}
+	mlx5_fill_page_array(&srq->buf, in->pas);
+
+	srq->wrid = kvmalloc_array(srq->msrq.max, sizeof(u64), GFP_KERNEL);
+	if (!srq->wrid) {
+		err = -ENOMEM;
+		goto err_in;
+	}
+	srq->wq_sig = !!srq_signature;
+
+	in->log_page_size = srq->buf.page_shift - MLX5_ADAPTER_PAGE_SHIFT;
+	if (MLX5_CAP_GEN(dev->mdev, cqe_version) == MLX5_CQE_VERSION_V1 &&
+	    in->type != IB_SRQT_BASIC)
+		in->user_index = MLX5_IB_DEFAULT_UIDX;
+
+	return 0;
+
+err_in:
+	kvfree(in->pas);
+
+err_buf:
+	mlx5_buf_free(dev->mdev, &srq->buf);
+
+err_db:
+	mlx5_db_free(dev->mdev, &srq->db);
+	return err;
+}
+
+static void destroy_srq_user(struct ib_pd *pd, struct mlx5_ib_srq *srq)
+{
+	mlx5_ib_db_unmap_user(to_mucontext(pd->uobject->context), &srq->db);
+	ib_umem_release(srq->umem);
+}
+
+
+static void destroy_srq_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_srq *srq)
+{
+	kvfree(srq->wrid);
+	mlx5_buf_free(dev->mdev, &srq->buf);
+	mlx5_db_free(dev->mdev, &srq->db);
+}
+
+struct ib_srq *mlx5_ib_create_srq(struct ib_pd *pd,
+				  struct ib_srq_init_attr *init_attr,
+				  struct ib_udata *udata)
+{
+	struct mlx5_ib_dev *dev = to_mdev(pd->device);
+	struct mlx5_ib_srq *srq;
+	size_t desc_size;
+	size_t buf_size;
+	int err;
+	struct mlx5_srq_attr in = {0};
+	__u32 max_srq_wqes = 1 << MLX5_CAP_GEN(dev->mdev, log_max_srq_sz);
+
+	/* Sanity check SRQ size before proceeding */
+	if (init_attr->attr.max_wr >= max_srq_wqes) {
+		mlx5_ib_dbg(dev, "max_wr %d, cap %d\n",
+			    init_attr->attr.max_wr,
+			    max_srq_wqes);
+		return ERR_PTR(-EINVAL);
+	}
+
+	srq = kmalloc(sizeof(*srq), GFP_KERNEL);
+	if (!srq)
+		return ERR_PTR(-ENOMEM);
+
+	mutex_init(&srq->mutex);
+	spin_lock_init(&srq->lock);
+	srq->msrq.max    = roundup_pow_of_two(init_attr->attr.max_wr + 1);
+	srq->msrq.max_gs = init_attr->attr.max_sge;
+
+	desc_size = sizeof(struct mlx5_wqe_srq_next_seg) +
+		    srq->msrq.max_gs * sizeof(struct mlx5_wqe_data_seg);
+	if (desc_size == 0 || srq->msrq.max_gs > desc_size) {
+		err = -EINVAL;
+		goto err_srq;
+	}
+	desc_size = roundup_pow_of_two(desc_size);
+	desc_size = max_t(size_t, 32, desc_size);
+	if (desc_size < sizeof(struct mlx5_wqe_srq_next_seg)) {
+		err = -EINVAL;
+		goto err_srq;
+	}
+	srq->msrq.max_avail_gather = (desc_size - sizeof(struct mlx5_wqe_srq_next_seg)) /
+		sizeof(struct mlx5_wqe_data_seg);
+	srq->msrq.wqe_shift = ilog2(desc_size);
+	buf_size = srq->msrq.max * desc_size;
+	if (buf_size < desc_size) {
+		err = -EINVAL;
+		goto err_srq;
+	}
+	in.type = init_attr->srq_type;
+
+	if (pd->uobject)
+		err = create_srq_user(pd, srq, &in, udata, buf_size);
+	else
+		err = create_srq_kernel(dev, srq, &in, buf_size);
+
+	if (err) {
+		mlx5_ib_warn(dev, "create srq %s failed, err %d\n",
+			     pd->uobject ? "user" : "kernel", err);
+		goto err_srq;
+	}
+
+	in.log_size = ilog2(srq->msrq.max);
+	in.wqe_shift = srq->msrq.wqe_shift - 4;
+	if (srq->wq_sig)
+		in.flags |= MLX5_SRQ_FLAG_WQ_SIG;
+
+	if (init_attr->srq_type == IB_SRQT_XRC)
+		in.xrcd = to_mxrcd(init_attr->ext.xrc.xrcd)->xrcdn;
+	else
+		in.xrcd = to_mxrcd(dev->devr.x0)->xrcdn;
+
+	if (init_attr->srq_type == IB_SRQT_TM) {
+		in.tm_log_list_size =
+			ilog2(init_attr->ext.tag_matching.max_num_tags) + 1;
+		if (in.tm_log_list_size >
+		    MLX5_CAP_GEN(dev->mdev, log_tag_matching_list_sz)) {
+			mlx5_ib_dbg(dev, "TM SRQ max_num_tags exceeding limit\n");
+			err = -EINVAL;
+			goto err_usr_kern_srq;
+		}
+		in.flags |= MLX5_SRQ_FLAG_RNDV;
+	}
+
+	if (ib_srq_has_cq(init_attr->srq_type))
+		in.cqn = to_mcq(init_attr->ext.cq)->mcq.cqn;
+	else
+		in.cqn = to_mcq(dev->devr.c0)->mcq.cqn;
+
+	in.pd = to_mpd(pd)->pdn;
+	in.db_record = srq->db.dma;
+	err = mlx5_core_create_srq(dev->mdev, &srq->msrq, &in);
+	kvfree(in.pas);
+	if (err) {
+		mlx5_ib_dbg(dev, "create SRQ failed, err %d\n", err);
+		goto err_usr_kern_srq;
+	}
+
+	mlx5_ib_dbg(dev, "create SRQ with srqn 0x%x\n", srq->msrq.srqn);
+
+	srq->msrq.event = mlx5_ib_srq_event;
+	srq->ibsrq.ext.xrc.srq_num = srq->msrq.srqn;
+
+	if (pd->uobject)
+		if (ib_copy_to_udata(udata, &srq->msrq.srqn, sizeof(__u32))) {
+			mlx5_ib_dbg(dev, "copy to user failed\n");
+			err = -EFAULT;
+			goto err_core;
+		}
+
+	init_attr->attr.max_wr = srq->msrq.max - 1;
+
+	return &srq->ibsrq;
+
+err_core:
+	mlx5_core_destroy_srq(dev->mdev, &srq->msrq);
+
+err_usr_kern_srq:
+	if (pd->uobject)
+		destroy_srq_user(pd, srq);
+	else
+		destroy_srq_kernel(dev, srq);
+
+err_srq:
+	kfree(srq);
+
+	return ERR_PTR(err);
+}
+
+int mlx5_ib_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
+		       enum ib_srq_attr_mask attr_mask, struct ib_udata *udata)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibsrq->device);
+	struct mlx5_ib_srq *srq = to_msrq(ibsrq);
+	int ret;
+
+	/* We don't support resizing SRQs yet */
+	if (attr_mask & IB_SRQ_MAX_WR)
+		return -EINVAL;
+
+	if (attr_mask & IB_SRQ_LIMIT) {
+		if (attr->srq_limit >= srq->msrq.max)
+			return -EINVAL;
+
+		mutex_lock(&srq->mutex);
+		ret = mlx5_core_arm_srq(dev->mdev, &srq->msrq, attr->srq_limit, 1);
+		mutex_unlock(&srq->mutex);
+
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
+int mlx5_ib_query_srq(struct ib_srq *ibsrq, struct ib_srq_attr *srq_attr)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibsrq->device);
+	struct mlx5_ib_srq *srq = to_msrq(ibsrq);
+	int ret;
+	struct mlx5_srq_attr *out;
+
+	out = kzalloc(sizeof(*out), GFP_KERNEL);
+	if (!out)
+		return -ENOMEM;
+
+	ret = mlx5_core_query_srq(dev->mdev, &srq->msrq, out);
+	if (ret)
+		goto out_box;
+
+	srq_attr->srq_limit = out->lwm;
+	srq_attr->max_wr    = srq->msrq.max - 1;
+	srq_attr->max_sge   = srq->msrq.max_gs;
+
+out_box:
+	kfree(out);
+	return ret;
+}
+
+int mlx5_ib_destroy_srq(struct ib_srq *srq)
+{
+	struct mlx5_ib_dev *dev = to_mdev(srq->device);
+	struct mlx5_ib_srq *msrq = to_msrq(srq);
+
+	mlx5_core_destroy_srq(dev->mdev, &msrq->msrq);
+
+	if (srq->uobject) {
+		mlx5_ib_db_unmap_user(to_mucontext(srq->uobject->context), &msrq->db);
+		ib_umem_release(msrq->umem);
+	} else {
+		destroy_srq_kernel(dev, msrq);
+	}
+
+	kfree(srq);
+	return 0;
+}
+
+void mlx5_ib_free_srq_wqe(struct mlx5_ib_srq *srq, int wqe_index)
+{
+	struct mlx5_wqe_srq_next_seg *next;
+
+	/* always called with interrupts disabled. */
+	spin_lock(&srq->lock);
+
+	next = get_wqe(srq, srq->tail);
+	next->next_wqe_index = cpu_to_be16(wqe_index);
+	srq->tail = wqe_index;
+
+	spin_unlock(&srq->lock);
+}
+
+int mlx5_ib_post_srq_recv(struct ib_srq *ibsrq, const struct ib_recv_wr *wr,
+			  const struct ib_recv_wr **bad_wr)
+{
+	struct mlx5_ib_srq *srq = to_msrq(ibsrq);
+	struct mlx5_wqe_srq_next_seg *next;
+	struct mlx5_wqe_data_seg *scat;
+	struct mlx5_ib_dev *dev = to_mdev(ibsrq->device);
+	struct mlx5_core_dev *mdev = dev->mdev;
+	unsigned long flags;
+	int err = 0;
+	int nreq;
+	int i;
+
+	spin_lock_irqsave(&srq->lock, flags);
+
+	if (mdev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR) {
+		err = -EIO;
+		*bad_wr = wr;
+		goto out;
+	}
+
+	for (nreq = 0; wr; nreq++, wr = wr->next) {
+		if (unlikely(wr->num_sge > srq->msrq.max_gs)) {
+			err = -EINVAL;
+			*bad_wr = wr;
+			break;
+		}
+
+		if (unlikely(srq->head == srq->tail)) {
+			err = -ENOMEM;
+			*bad_wr = wr;
+			break;
+		}
+
+		srq->wrid[srq->head] = wr->wr_id;
+
+		next      = get_wqe(srq, srq->head);
+		srq->head = be16_to_cpu(next->next_wqe_index);
+		scat      = (struct mlx5_wqe_data_seg *)(next + 1);
+
+		for (i = 0; i < wr->num_sge; i++) {
+			scat[i].byte_count = cpu_to_be32(wr->sg_list[i].length);
+			scat[i].lkey       = cpu_to_be32(wr->sg_list[i].lkey);
+			scat[i].addr       = cpu_to_be64(wr->sg_list[i].addr);
+		}
+
+		if (i < srq->msrq.max_avail_gather) {
+			scat[i].byte_count = 0;
+			scat[i].lkey       = cpu_to_be32(MLX5_INVALID_LKEY);
+			scat[i].addr       = 0;
+		}
+	}
+
+	if (likely(nreq)) {
+		srq->wqe_ctr += nreq;
+
+		/* Make sure that descriptors are written before
+		 * doorbell record.
+		 */
+		wmb();
+
+		*srq->db.db = cpu_to_be32(srq->wqe_ctr);
+	}
+out:
+	spin_unlock_irqrestore(&srq->lock, flags);
+
+	return err;
+}