net/mlx5_core: Update device capabilities handling

Rearrange struct mlx5_caps so it has a "gen" field to represent the current
capabilities configured for the device. Max capabilities can also be queried
from the device. Also update capabilities struct to contain more fields as per
the latest revision if firmware specification.

Signed-off-by: Eli Cohen <eli@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

12 files changed, 331 insertions, 222 deletions

diff --git a/drivers/infiniband/hw/mlx5/cq.c b/drivers/infiniband/hw/mlx5/cq.c
index e4056279166d..10cfce5119a9 100644
--- a/drivers/infiniband/hw/mlx5/cq.c
+++ b/drivers/infiniband/hw/mlx5/cq.c
@@ -752,7 +752,7 @@ struct ib_cq *mlx5_ib_create_cq(struct ib_device *ibdev, int entries,
 		return ERR_PTR(-EINVAL);
 
 	entries = roundup_pow_of_two(entries + 1);
-	if (entries > dev->mdev->caps.max_cqes)
+	if (entries > dev->mdev->caps.gen.max_cqes)
 		return ERR_PTR(-EINVAL);
 
 	cq = kzalloc(sizeof(*cq), GFP_KERNEL);
@@ -919,7 +919,7 @@ int mlx5_ib_modify_cq(struct ib_cq *cq, u16 cq_count, u16 cq_period)
 	int err;
 	u32 fsel;
 
-	if (!(dev->mdev->caps.flags & MLX5_DEV_CAP_FLAG_CQ_MODER))
+	if (!(dev->mdev->caps.gen.flags & MLX5_DEV_CAP_FLAG_CQ_MODER))
 		return -ENOSYS;
 
 	in = kzalloc(sizeof(*in), GFP_KERNEL);
@@ -1074,7 +1074,7 @@ int mlx5_ib_resize_cq(struct ib_cq *ibcq, int entries, struct ib_udata *udata)
 	int uninitialized_var(cqe_size);
 	unsigned long flags;
 
-	if (!(dev->mdev->caps.flags & MLX5_DEV_CAP_FLAG_RESIZE_CQ)) {
+	if (!(dev->mdev->caps.gen.flags & MLX5_DEV_CAP_FLAG_RESIZE_CQ)) {
 		pr_info("Firmware does not support resize CQ\n");
 		return -ENOSYS;
 	}
@@ -1083,7 +1083,7 @@ int mlx5_ib_resize_cq(struct ib_cq *ibcq, int entries, struct ib_udata *udata)
 		return -EINVAL;
 
 	entries = roundup_pow_of_two(entries + 1);
-	if (entries > dev->mdev->caps.max_cqes + 1)
+	if (entries > dev->mdev->caps.gen.max_cqes + 1)
 		return -EINVAL;
 
 	if (entries == ibcq->cqe + 1)
diff --git a/drivers/infiniband/hw/mlx5/mad.c b/drivers/infiniband/hw/mlx5/mad.c
index b514bbb5610f..657af9a1167c 100644
--- a/drivers/infiniband/hw/mlx5/mad.c
+++ b/drivers/infiniband/hw/mlx5/mad.c
@@ -129,7 +129,7 @@ int mlx5_query_ext_port_caps(struct mlx5_ib_dev *dev, u8 port)
 
 	packet_error = be16_to_cpu(out_mad->status);
 
-	dev->mdev->caps.ext_port_cap[port - 1] = (!err && !packet_error) ?
+	dev->mdev->caps.gen.ext_port_cap[port - 1] = (!err && !packet_error) ?
 		MLX_EXT_PORT_CAP_FLAG_EXTENDED_PORT_INFO : 0;
 
 out:
diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index d8907b20522a..f3114d1132fb 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -157,11 +157,13 @@ static int mlx5_ib_query_device(struct ib_device *ibdev,
 	struct mlx5_ib_dev *dev = to_mdev(ibdev);
 	struct ib_smp *in_mad  = NULL;
 	struct ib_smp *out_mad = NULL;
+	struct mlx5_general_caps *gen;
 	int err = -ENOMEM;
 	int max_rq_sg;
 	int max_sq_sg;
 	u64 flags;
 
+	gen = &dev->mdev->caps.gen;
 	in_mad  = kzalloc(sizeof(*in_mad), GFP_KERNEL);
 	out_mad = kmalloc(sizeof(*out_mad), GFP_KERNEL);
 	if (!in_mad || !out_mad)
@@ -183,7 +185,7 @@ static int mlx5_ib_query_device(struct ib_device *ibdev,
 		IB_DEVICE_PORT_ACTIVE_EVENT		|
 		IB_DEVICE_SYS_IMAGE_GUID		|
 		IB_DEVICE_RC_RNR_NAK_GEN;
-	flags = dev->mdev->caps.flags;
+	flags = gen->flags;
 	if (flags & MLX5_DEV_CAP_FLAG_BAD_PKEY_CNTR)
 		props->device_cap_flags |= IB_DEVICE_BAD_PKEY_CNTR;
 	if (flags & MLX5_DEV_CAP_FLAG_BAD_QKEY_CNTR)
@@ -213,30 +215,31 @@ static int mlx5_ib_query_device(struct ib_device *ibdev,
 	memcpy(&props->sys_image_guid, out_mad->data +	4, 8);
 
 	props->max_mr_size	   = ~0ull;
-	props->page_size_cap	   = dev->mdev->caps.min_page_sz;
-	props->max_qp		   = 1 << dev->mdev->caps.log_max_qp;
-	props->max_qp_wr	   = dev->mdev->caps.max_wqes;
-	max_rq_sg = dev->mdev->caps.max_rq_desc_sz / sizeof(struct mlx5_wqe_data_seg);
-	max_sq_sg = (dev->mdev->caps.max_sq_desc_sz - sizeof(struct mlx5_wqe_ctrl_seg)) /
+	props->page_size_cap	   = gen->min_page_sz;
+	props->max_qp		   = 1 << gen->log_max_qp;
+	props->max_qp_wr	   = gen->max_wqes;
+	max_rq_sg = gen->max_rq_desc_sz / sizeof(struct mlx5_wqe_data_seg);
+	max_sq_sg = (gen->max_sq_desc_sz - sizeof(struct mlx5_wqe_ctrl_seg)) /
 		sizeof(struct mlx5_wqe_data_seg);
 	props->max_sge = min(max_rq_sg, max_sq_sg);
-	props->max_cq		   = 1 << dev->mdev->caps.log_max_cq;
-	props->max_cqe		   = dev->mdev->caps.max_cqes - 1;
-	props->max_mr		   = 1 << dev->mdev->caps.log_max_mkey;
-	props->max_pd		   = 1 << dev->mdev->caps.log_max_pd;
-	props->max_qp_rd_atom	   = dev->mdev->caps.max_ra_req_qp;
-	props->max_qp_init_rd_atom = dev->mdev->caps.max_ra_res_qp;
+	props->max_cq		   = 1 << gen->log_max_cq;
+	props->max_cqe		   = gen->max_cqes - 1;
+	props->max_mr		   = 1 << gen->log_max_mkey;
+	props->max_pd		   = 1 << gen->log_max_pd;
+	props->max_qp_rd_atom	   = 1 << gen->log_max_ra_req_qp;
+	props->max_qp_init_rd_atom = 1 << gen->log_max_ra_res_qp;
+	props->max_srq		   = 1 << gen->log_max_srq;
+	props->max_srq_wr	   = gen->max_srq_wqes - 1;
+	props->local_ca_ack_delay  = gen->local_ca_ack_delay;
 	props->max_res_rd_atom	   = props->max_qp_rd_atom * props->max_qp;
-	props->max_srq		   = 1 << dev->mdev->caps.log_max_srq;
-	props->max_srq_wr	   = dev->mdev->caps.max_srq_wqes - 1;
 	props->max_srq_sge	   = max_rq_sg - 1;
 	props->max_fast_reg_page_list_len = (unsigned int)-1;
-	props->local_ca_ack_delay  = dev->mdev->caps.local_ca_ack_delay;
+	props->local_ca_ack_delay  = gen->local_ca_ack_delay;
 	props->atomic_cap	   = IB_ATOMIC_NONE;
 	props->masked_atomic_cap   = IB_ATOMIC_NONE;
 	props->max_pkeys	   = be16_to_cpup((__be16 *)(out_mad->data + 28));
-	props->max_mcast_grp	   = 1 << dev->mdev->caps.log_max_mcg;
-	props->max_mcast_qp_attach = dev->mdev->caps.max_qp_mcg;
+	props->max_mcast_grp	   = 1 << gen->log_max_mcg;
+	props->max_mcast_qp_attach = gen->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 */
@@ -254,10 +257,12 @@ int mlx5_ib_query_port(struct ib_device *ibdev, u8 port,
 	struct mlx5_ib_dev *dev = to_mdev(ibdev);
 	struct ib_smp *in_mad  = NULL;
 	struct ib_smp *out_mad = NULL;
+	struct mlx5_general_caps *gen;
 	int ext_active_speed;
 	int err = -ENOMEM;
 
-	if (port < 1 || port > dev->mdev->caps.num_ports) {
+	gen = &dev->mdev->caps.gen;
+	if (port < 1 || port > gen->num_ports) {
 		mlx5_ib_warn(dev, "invalid port number %d\n", port);
 		return -EINVAL;
 	}
@@ -288,8 +293,8 @@ int mlx5_ib_query_port(struct ib_device *ibdev, u8 port,
 	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 << to_mdev(ibdev)->mdev->caps.log_max_msg;
-	props->pkey_tbl_len	= to_mdev(ibdev)->mdev->caps.port[port - 1].pkey_table_len;
+	props->max_msg_sz	= 1 << gen->log_max_msg;
+	props->pkey_tbl_len	= gen->port[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;
@@ -316,7 +321,7 @@ int mlx5_ib_query_port(struct ib_device *ibdev, u8 port,
 
 	/* If reported active speed is QDR, check if is FDR-10 */
 	if (props->active_speed == 4) {
-		if (dev->mdev->caps.ext_port_cap[port - 1] &
+		if (gen->ext_port_cap[port - 1] &
 		    MLX_EXT_PORT_CAP_FLAG_EXTENDED_PORT_INFO) {
 			init_query_mad(in_mad);
 			in_mad->attr_id = MLX5_ATTR_EXTENDED_PORT_INFO;
@@ -470,6 +475,7 @@ static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev,
 	struct mlx5_ib_alloc_ucontext_req_v2 req;
 	struct mlx5_ib_alloc_ucontext_resp resp;
 	struct mlx5_ib_ucontext *context;
+	struct mlx5_general_caps *gen;
 	struct mlx5_uuar_info *uuari;
 	struct mlx5_uar *uars;
 	int gross_uuars;
@@ -480,6 +486,7 @@ static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev,
 	int i;
 	size_t reqlen;
 
+	gen = &dev->mdev->caps.gen;
 	if (!dev->ib_active)
 		return ERR_PTR(-EAGAIN);
 
@@ -512,14 +519,14 @@ static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev,
 
 	num_uars = req.total_num_uuars / MLX5_NON_FP_BF_REGS_PER_PAGE;
 	gross_uuars = num_uars * MLX5_BF_REGS_PER_PAGE;
-	resp.qp_tab_size      = 1 << dev->mdev->caps.log_max_qp;
-	resp.bf_reg_size      = dev->mdev->caps.bf_reg_size;
+	resp.qp_tab_size      = 1 << gen->log_max_qp;
+	resp.bf_reg_size      = gen->bf_reg_size;
 	resp.cache_line_size  = L1_CACHE_BYTES;
-	resp.max_sq_desc_sz = dev->mdev->caps.max_sq_desc_sz;
-	resp.max_rq_desc_sz = dev->mdev->caps.max_rq_desc_sz;
-	resp.max_send_wqebb = dev->mdev->caps.max_wqes;
-	resp.max_recv_wr = dev->mdev->caps.max_wqes;
-	resp.max_srq_recv_wr = dev->mdev->caps.max_srq_wqes;
+	resp.max_sq_desc_sz = gen->max_sq_desc_sz;
+	resp.max_rq_desc_sz = gen->max_rq_desc_sz;
+	resp.max_send_wqebb = gen->max_wqes;
+	resp.max_recv_wr = gen->max_wqes;
+	resp.max_srq_recv_wr = gen->max_srq_wqes;
 
 	context = kzalloc(sizeof(*context), GFP_KERNEL);
 	if (!context)
@@ -565,7 +572,7 @@ static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev,
 	mutex_init(&context->db_page_mutex);
 
 	resp.tot_uuars = req.total_num_uuars;
-	resp.num_ports = dev->mdev->caps.num_ports;
+	resp.num_ports = gen->num_ports;
 	err = ib_copy_to_udata(udata, &resp,
 			       sizeof(resp) - sizeof(resp.reserved));
 	if (err)
@@ -967,9 +974,11 @@ static void mlx5_ib_event(struct mlx5_core_dev *dev, void *context,
 
 static void get_ext_port_caps(struct mlx5_ib_dev *dev)
 {
+	struct mlx5_general_caps *gen;
 	int port;
 
-	for (port = 1; port <= dev->mdev->caps.num_ports; port++)
+	gen = &dev->mdev->caps.gen;
+	for (port = 1; port <= gen->num_ports; port++)
 		mlx5_query_ext_port_caps(dev, port);
 }
 
@@ -977,9 +986,11 @@ static int get_port_caps(struct mlx5_ib_dev *dev)
 {
 	struct ib_device_attr *dprops = NULL;
 	struct ib_port_attr *pprops = NULL;
+	struct mlx5_general_caps *gen;
 	int err = 0;
 	int port;
 
+	gen = &dev->mdev->caps.gen;
 	pprops = kmalloc(sizeof(*pprops), GFP_KERNEL);
 	if (!pprops)
 		goto out;
@@ -994,14 +1005,14 @@ static int get_port_caps(struct mlx5_ib_dev *dev)
 		goto out;
 	}
 
-	for (port = 1; port <= dev->mdev->caps.num_ports; port++) {
+	for (port = 1; port <= gen->num_ports; port++) {
 		err = mlx5_ib_query_port(&dev->ib_dev, port, pprops);
 		if (err) {
 			mlx5_ib_warn(dev, "query_port %d failed %d\n", port, err);
 			break;
 		}
-		dev->mdev->caps.port[port - 1].pkey_table_len = dprops->max_pkeys;
-		dev->mdev->caps.port[port - 1].gid_table_len = pprops->gid_tbl_len;
+		gen->port[port - 1].pkey_table_len = dprops->max_pkeys;
+		gen->port[port - 1].gid_table_len = pprops->gid_tbl_len;
 		mlx5_ib_dbg(dev, "pkey_table_len %d, gid_table_len %d\n",
 			    dprops->max_pkeys, pprops->gid_tbl_len);
 	}
@@ -1279,8 +1290,8 @@ static void *mlx5_ib_add(struct mlx5_core_dev *mdev)
 	strlcpy(dev->ib_dev.name, "mlx5_%d", 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	= mdev->caps.reserved_lkey;
-	dev->num_ports		= mdev->caps.num_ports;
+	dev->ib_dev.local_dma_lkey	= mdev->caps.gen.reserved_lkey;
+	dev->num_ports		= mdev->caps.gen.num_ports;
 	dev->ib_dev.phys_port_cnt     = dev->num_ports;
 	dev->ib_dev.num_comp_vectors	= dev->num_comp_vectors;
 	dev->ib_dev.dma_device	= &mdev->pdev->dev;
@@ -1355,7 +1366,7 @@ static void *mlx5_ib_add(struct mlx5_core_dev *mdev)
 	dev->ib_dev.free_fast_reg_page_list  = mlx5_ib_free_fast_reg_page_list;
 	dev->ib_dev.check_mr_status	= mlx5_ib_check_mr_status;
 
-	if (mdev->caps.flags & MLX5_DEV_CAP_FLAG_XRC) {
+	if (mdev->caps.gen.flags & MLX5_DEV_CAP_FLAG_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 |=
diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c
index 8c574b63d77b..dbfe498870c1 100644
--- a/drivers/infiniband/hw/mlx5/qp.c
+++ b/drivers/infiniband/hw/mlx5/qp.c
@@ -158,11 +158,13 @@ static void mlx5_ib_qp_event(struct mlx5_core_qp *qp, int type)
 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)
 {
+	struct mlx5_general_caps *gen;
 	int wqe_size;
 	int wq_size;
 
+	gen = &dev->mdev->caps.gen;
 	/* Sanity check RQ size before proceeding */
-	if (cap->max_recv_wr  > dev->mdev->caps.max_wqes)
+	if (cap->max_recv_wr  > gen->max_wqes)
 		return -EINVAL;
 
 	if (!has_rq) {
@@ -182,10 +184,10 @@ static int set_rq_size(struct mlx5_ib_dev *dev, struct ib_qp_cap *cap,
 			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 > dev->mdev->caps.max_rq_desc_sz) {
+			if (wqe_size > gen->max_rq_desc_sz) {
 				mlx5_ib_dbg(dev, "wqe_size %d, max %d\n",
 					    wqe_size,
-					    dev->mdev->caps.max_rq_desc_sz);
+					    gen->max_rq_desc_sz);
 				return -EINVAL;
 			}
 			qp->rq.wqe_shift = ilog2(wqe_size);
@@ -266,9 +268,11 @@ static int calc_send_wqe(struct ib_qp_init_attr *attr)
 static int calc_sq_size(struct mlx5_ib_dev *dev, struct ib_qp_init_attr *attr,
 			struct mlx5_ib_qp *qp)
 {
+	struct mlx5_general_caps *gen;
 	int wqe_size;
 	int wq_size;
 
+	gen = &dev->mdev->caps.gen;
 	if (!attr->cap.max_send_wr)
 		return 0;
 
@@ -277,9 +281,9 @@ static int calc_sq_size(struct mlx5_ib_dev *dev, struct ib_qp_init_attr *attr,
 	if (wqe_size < 0)
 		return wqe_size;
 
-	if (wqe_size > dev->mdev->caps.max_sq_desc_sz) {
+	if (wqe_size > gen->max_sq_desc_sz) {
 		mlx5_ib_dbg(dev, "wqe_size(%d) > max_sq_desc_sz(%d)\n",
-			    wqe_size, dev->mdev->caps.max_sq_desc_sz);
+			    wqe_size, gen->max_sq_desc_sz);
 		return -EINVAL;
 	}
 
@@ -292,9 +296,9 @@ static int calc_sq_size(struct mlx5_ib_dev *dev, struct ib_qp_init_attr *attr,
 
 	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 > dev->mdev->caps.max_wqes) {
+	if (qp->sq.wqe_cnt > gen->max_wqes) {
 		mlx5_ib_dbg(dev, "wqe count(%d) exceeds limits(%d)\n",
-			    qp->sq.wqe_cnt, dev->mdev->caps.max_wqes);
+			    qp->sq.wqe_cnt, gen->max_wqes);
 		return -ENOMEM;
 	}
 	qp->sq.wqe_shift = ilog2(MLX5_SEND_WQE_BB);
@@ -309,11 +313,13 @@ static int set_user_buf_size(struct mlx5_ib_dev *dev,
 			    struct mlx5_ib_qp *qp,
 			    struct mlx5_ib_create_qp *ucmd)
 {
+	struct mlx5_general_caps *gen;
 	int desc_sz = 1 << qp->sq.wqe_shift;
 
-	if (desc_sz > dev->mdev->caps.max_sq_desc_sz) {
+	gen = &dev->mdev->caps.gen;
+	if (desc_sz > gen->max_sq_desc_sz) {
 		mlx5_ib_warn(dev, "desc_sz %d, max_sq_desc_sz %d\n",
-			     desc_sz, dev->mdev->caps.max_sq_desc_sz);
+			     desc_sz, gen->max_sq_desc_sz);
 		return -EINVAL;
 	}
 
@@ -325,9 +331,9 @@ static int set_user_buf_size(struct mlx5_ib_dev *dev,
 
 	qp->sq.wqe_cnt = ucmd->sq_wqe_count;
 
-	if (qp->sq.wqe_cnt > dev->mdev->caps.max_wqes) {
+	if (qp->sq.wqe_cnt > gen->max_wqes) {
 		mlx5_ib_warn(dev, "wqe_cnt %d, max_wqes %d\n",
-			     qp->sq.wqe_cnt, dev->mdev->caps.max_wqes);
+			     qp->sq.wqe_cnt, gen->max_wqes);
 		return -EINVAL;
 	}
 
@@ -803,16 +809,18 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd,
 	struct mlx5_ib_resources *devr = &dev->devr;
 	struct mlx5_ib_create_qp_resp resp;
 	struct mlx5_create_qp_mbox_in *in;
+	struct mlx5_general_caps *gen;
 	struct mlx5_ib_create_qp ucmd;
 	int inlen = sizeof(*in);
 	int err;
 
+	gen = &dev->mdev->caps.gen;
 	mutex_init(&qp->mutex);
 	spin_lock_init(&qp->sq.lock);
 	spin_lock_init(&qp->rq.lock);
 
 	if (init_attr->create_flags & IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK) {
-		if (!(dev->mdev->caps.flags & MLX5_DEV_CAP_FLAG_BLOCK_MCAST)) {
+		if (!(gen->flags & MLX5_DEV_CAP_FLAG_BLOCK_MCAST)) {
 			mlx5_ib_dbg(dev, "block multicast loopback isn't supported\n");
 			return -EINVAL;
 		} else {
@@ -851,9 +859,9 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd,
 				mlx5_ib_dbg(dev, "invalid rq params\n");
 				return -EINVAL;
 			}
-			if (ucmd.sq_wqe_count > dev->mdev->caps.max_wqes) {
+			if (ucmd.sq_wqe_count > gen->max_wqes) {
 				mlx5_ib_dbg(dev, "requested sq_wqe_count (%d) > max allowed (%d)\n",
-					    ucmd.sq_wqe_count, dev->mdev->caps.max_wqes);
+					    ucmd.sq_wqe_count, gen->max_wqes);
 				return -EINVAL;
 			}
 			err = create_user_qp(dev, pd, qp, udata, &in, &resp, &inlen);
@@ -1144,6 +1152,7 @@ struct ib_qp *mlx5_ib_create_qp(struct ib_pd *pd,
 				struct ib_qp_init_attr *init_attr,
 				struct ib_udata *udata)
 {
+	struct mlx5_general_caps *gen;
 	struct mlx5_ib_dev *dev;
 	struct mlx5_ib_qp *qp;
 	u16 xrcdn = 0;
@@ -1161,11 +1170,12 @@ struct ib_qp *mlx5_ib_create_qp(struct ib_pd *pd,
 		}
 		dev = to_mdev(to_mxrcd(init_attr->xrcd)->ibxrcd.device);
 	}
+	gen = &dev->mdev->caps.gen;
 
 	switch (init_attr->qp_type) {
 	case IB_QPT_XRC_TGT:
 	case IB_QPT_XRC_INI:
-		if (!(dev->mdev->caps.flags & MLX5_DEV_CAP_FLAG_XRC)) {
+		if (!(gen->flags & MLX5_DEV_CAP_FLAG_XRC)) {
 			mlx5_ib_dbg(dev, "XRC not supported\n");
 			return ERR_PTR(-ENOSYS);
 		}
@@ -1272,6 +1282,9 @@ enum {
 
 static int ib_rate_to_mlx5(struct mlx5_ib_dev *dev, u8 rate)
 {
+	struct mlx5_general_caps *gen;
+
+	gen = &dev->mdev->caps.gen;
 	if (rate == IB_RATE_PORT_CURRENT) {
 		return 0;
 	} else if (rate < IB_RATE_2_5_GBPS || rate > IB_RATE_300_GBPS) {
@@ -1279,7 +1292,7 @@ static int ib_rate_to_mlx5(struct mlx5_ib_dev *dev, u8 rate)
 	} else {
 		while (rate != IB_RATE_2_5_GBPS &&
 		       !(1 << (rate + MLX5_STAT_RATE_OFFSET) &
-			 dev->mdev->caps.stat_rate_support))
+			 gen->stat_rate_support))
 			--rate;
 	}
 
@@ -1290,8 +1303,10 @@ static int mlx5_set_path(struct mlx5_ib_dev *dev, const struct ib_ah_attr *ah,
 			 struct mlx5_qp_path *path, u8 port, int attr_mask,
 			 u32 path_flags, const struct ib_qp_attr *attr)
 {
+	struct mlx5_general_caps *gen;
 	int err;
 
+	gen = &dev->mdev->caps.gen;
 	path->fl = (path_flags & MLX5_PATH_FLAG_FL) ? 0x80 : 0;
 	path->free_ar = (path_flags & MLX5_PATH_FLAG_FREE_AR) ? 0x80 : 0;
 
@@ -1318,9 +1333,9 @@ static int mlx5_set_path(struct mlx5_ib_dev *dev, const struct ib_ah_attr *ah,
 	path->port = port;
 
 	if (ah->ah_flags & IB_AH_GRH) {
-		if (ah->grh.sgid_index >= dev->mdev->caps.port[port - 1].gid_table_len) {
+		if (ah->grh.sgid_index >= gen->port[port - 1].gid_table_len) {
 			pr_err(KERN_ERR "sgid_index (%u) too large. max is %d\n",
-			       ah->grh.sgid_index, dev->mdev->caps.port[port - 1].gid_table_len);
+			       ah->grh.sgid_index, gen->port[port - 1].gid_table_len);
 			return -EINVAL;
 		}
 
@@ -1492,6 +1507,7 @@ static int __mlx5_ib_modify_qp(struct ib_qp *ibqp,
 	struct mlx5_ib_qp *qp = to_mqp(ibqp);
 	struct mlx5_ib_cq *send_cq, *recv_cq;
 	struct mlx5_qp_context *context;
+	struct mlx5_general_caps *gen;
 	struct mlx5_modify_qp_mbox_in *in;
 	struct mlx5_ib_pd *pd;
 	enum mlx5_qp_state mlx5_cur, mlx5_new;
@@ -1500,6 +1516,7 @@ static int __mlx5_ib_modify_qp(struct ib_qp *ibqp,
 	int mlx5_st;
 	int err;
 
+	gen = &dev->mdev->caps.gen;
 	in = kzalloc(sizeof(*in), GFP_KERNEL);
 	if (!in)
 		return -ENOMEM;
@@ -1539,7 +1556,7 @@ static int __mlx5_ib_modify_qp(struct ib_qp *ibqp,
 			err = -EINVAL;
 			goto out;
 		}
-		context->mtu_msgmax = (attr->path_mtu << 5) | dev->mdev->caps.log_max_msg;
+		context->mtu_msgmax = (attr->path_mtu << 5) | gen->log_max_msg;
 	}
 
 	if (attr_mask & IB_QP_DEST_QPN)
@@ -1685,9 +1702,11 @@ int mlx5_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 	struct mlx5_ib_dev *dev = to_mdev(ibqp->device);
 	struct mlx5_ib_qp *qp = to_mqp(ibqp);
 	enum ib_qp_state cur_state, new_state;
+	struct mlx5_general_caps *gen;
 	int err = -EINVAL;
 	int port;
 
+	gen = &dev->mdev->caps.gen;
 	mutex_lock(&qp->mutex);
 
 	cur_state = attr_mask & IB_QP_CUR_STATE ? attr->cur_qp_state : qp->state;
@@ -1699,21 +1718,21 @@ int mlx5_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 		goto out;
 
 	if ((attr_mask & IB_QP_PORT) &&
-	    (attr->port_num == 0 || attr->port_num > dev->mdev->caps.num_ports))
+	    (attr->port_num == 0 || attr->port_num > gen->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->caps.port[port - 1].pkey_table_len)
+		if (attr->pkey_index >= gen->port[port - 1].pkey_table_len)
 			goto out;
 	}
 
 	if (attr_mask & IB_QP_MAX_QP_RD_ATOMIC &&
-	    attr->max_rd_atomic > dev->mdev->caps.max_ra_res_qp)
+	    attr->max_rd_atomic > (1 << gen->log_max_ra_res_qp))
 		goto out;
 
 	if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC &&
-	    attr->max_dest_rd_atomic > dev->mdev->caps.max_ra_req_qp)
+	    attr->max_dest_rd_atomic > (1 << gen->log_max_ra_req_qp))
 		goto out;
 
 	if (cur_state == new_state && cur_state == IB_QPS_RESET) {
@@ -2893,7 +2912,8 @@ static void to_ib_ah_attr(struct mlx5_ib_dev *ibdev, struct ib_ah_attr *ib_ah_at
 	memset(ib_ah_attr, 0, sizeof(*ib_ah_attr));
 	ib_ah_attr->port_num	  = path->port;
 
-	if (ib_ah_attr->port_num == 0 || ib_ah_attr->port_num > dev->caps.num_ports)
+	if (ib_ah_attr->port_num == 0 ||
+	    ib_ah_attr->port_num > dev->caps.gen.num_ports)
 		return;
 
 	ib_ah_attr->sl = path->sl & 0xf;
@@ -3011,10 +3031,12 @@ struct ib_xrcd *mlx5_ib_alloc_xrcd(struct ib_device *ibdev,
 					  struct ib_udata *udata)
 {
 	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+	struct mlx5_general_caps *gen;
 	struct mlx5_ib_xrcd *xrcd;
 	int err;
 
-	if (!(dev->mdev->caps.flags & MLX5_DEV_CAP_FLAG_XRC))
+	gen = &dev->mdev->caps.gen;
+	if (!(gen->flags & MLX5_DEV_CAP_FLAG_XRC))
 		return ERR_PTR(-ENOSYS);
 
 	xrcd = kmalloc(sizeof(*xrcd), GFP_KERNEL);
diff --git a/drivers/infiniband/hw/mlx5/srq.c b/drivers/infiniband/hw/mlx5/srq.c
index 70bd131ba646..97cc1baaa8e3 100644
--- a/drivers/infiniband/hw/mlx5/srq.c
+++ b/drivers/infiniband/hw/mlx5/srq.c
@@ -238,6 +238,7 @@ struct ib_srq *mlx5_ib_create_srq(struct ib_pd *pd,
 				  struct ib_udata *udata)
 {
 	struct mlx5_ib_dev *dev = to_mdev(pd->device);
+	struct mlx5_general_caps *gen;
 	struct mlx5_ib_srq *srq;
 	int desc_size;
 	int buf_size;
@@ -247,11 +248,12 @@ struct ib_srq *mlx5_ib_create_srq(struct ib_pd *pd,
 	int is_xrc;
 	u32 flgs, xrcdn;
 
+	gen = &dev->mdev->caps.gen;
 	/* Sanity check SRQ size before proceeding */
-	if (init_attr->attr.max_wr >= dev->mdev->caps.max_srq_wqes) {
+	if (init_attr->attr.max_wr >= gen->max_srq_wqes) {
 		mlx5_ib_dbg(dev, "max_wr %d, cap %d\n",
 			    init_attr->attr.max_wr,
-			    dev->mdev->caps.max_srq_wqes);
+			    gen->max_srq_wqes);
 		return ERR_PTR(-EINVAL);
 	}
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/cmd.c b/drivers/net/ethernet/mellanox/mlx5/core/cmd.c
index 65a7da69e2ac..6eb0f85cf872 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/cmd.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/cmd.c
@@ -1538,16 +1538,9 @@ static const char *cmd_status_str(u8 status)
 	}
 }
 
-int mlx5_cmd_status_to_err(struct mlx5_outbox_hdr *hdr)
+static int cmd_status_to_err(u8 status)
 {
-	if (!hdr->status)
-		return 0;
-
-	pr_warn("command failed, status %s(0x%x), syndrome 0x%x\n",
-		cmd_status_str(hdr->status), hdr->status,
-		be32_to_cpu(hdr->syndrome));
-
-	switch (hdr->status) {
+	switch (status) {
 	case MLX5_CMD_STAT_OK:				return 0;
 	case MLX5_CMD_STAT_INT_ERR:			return -EIO;
 	case MLX5_CMD_STAT_BAD_OP_ERR:			return -EINVAL;
@@ -1567,3 +1560,16 @@ int mlx5_cmd_status_to_err(struct mlx5_outbox_hdr *hdr)
 	default:					return -EIO;
 	}
 }
+
+/* this will be available till all the commands use set/get macros */
+int mlx5_cmd_status_to_err(struct mlx5_outbox_hdr *hdr)
+{
+	if (!hdr->status)
+		return 0;
+
+	pr_warn("command failed, status %s(0x%x), syndrome 0x%x\n",
+		cmd_status_str(hdr->status), hdr->status,
+		be32_to_cpu(hdr->syndrome));
+
+	return cmd_status_to_err(hdr->status);
+}
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eq.c b/drivers/net/ethernet/mellanox/mlx5/core/eq.c
index 4e8bd0b34bb0..11b9b840ad4d 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eq.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eq.c
@@ -468,7 +468,7 @@ int mlx5_start_eqs(struct mlx5_core_dev *dev)
 
 	err = mlx5_create_map_eq(dev, &table->pages_eq,
 				 MLX5_EQ_VEC_PAGES,
-				 dev->caps.max_vf + 1,
+				 dev->caps.gen.max_vf + 1,
 				 1 << MLX5_EVENT_TYPE_PAGE_REQUEST, "mlx5_pages_eq",
 				 &dev->priv.uuari.uars[0]);
 	if (err) {
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fw.c b/drivers/net/ethernet/mellanox/mlx5/core/fw.c
index f012658b6a92..087c4c797deb 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fw.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fw.c
@@ -64,86 +64,9 @@ out_out:
 	return err;
 }
 
-int mlx5_cmd_query_hca_cap(struct mlx5_core_dev *dev,
-			   struct mlx5_caps *caps)
+int mlx5_cmd_query_hca_cap(struct mlx5_core_dev *dev, struct mlx5_caps *caps)
 {
-	struct mlx5_cmd_query_hca_cap_mbox_out *out;
-	struct mlx5_cmd_query_hca_cap_mbox_in in;
-	struct mlx5_query_special_ctxs_mbox_out ctx_out;
-	struct mlx5_query_special_ctxs_mbox_in ctx_in;
-	int err;
-	u16 t16;
-
-	out = kzalloc(sizeof(*out), GFP_KERNEL);
-	if (!out)
-		return -ENOMEM;
-
-	memset(&in, 0, sizeof(in));
-	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_QUERY_HCA_CAP);
-	in.hdr.opmod  = cpu_to_be16(0x1);
-	err = mlx5_cmd_exec(dev, &in, sizeof(in), out, sizeof(*out));
-	if (err)
-		goto out_out;
-
-	if (out->hdr.status) {
-		err = mlx5_cmd_status_to_err(&out->hdr);
-		goto out_out;
-	}
-
-
-	caps->log_max_eq = out->hca_cap.log_max_eq & 0xf;
-	caps->max_cqes = 1 << out->hca_cap.log_max_cq_sz;
-	caps->max_wqes = 1 << out->hca_cap.log_max_qp_sz;
-	caps->max_sq_desc_sz = be16_to_cpu(out->hca_cap.max_desc_sz_sq);
-	caps->max_rq_desc_sz = be16_to_cpu(out->hca_cap.max_desc_sz_rq);
-	caps->flags = be64_to_cpu(out->hca_cap.flags);
-	caps->stat_rate_support = be16_to_cpu(out->hca_cap.stat_rate_support);
-	caps->log_max_msg = out->hca_cap.log_max_msg & 0x1f;
-	caps->num_ports = out->hca_cap.num_ports & 0xf;
-	caps->log_max_cq = out->hca_cap.log_max_cq & 0x1f;
-	if (caps->num_ports > MLX5_MAX_PORTS) {
-		mlx5_core_err(dev, "device has %d ports while the driver supports max %d ports\n",
-			      caps->num_ports, MLX5_MAX_PORTS);
-		err = -EINVAL;
-		goto out_out;
-	}
-	caps->log_max_qp = out->hca_cap.log_max_qp & 0x1f;
-	caps->log_max_mkey = out->hca_cap.log_max_mkey & 0x3f;
-	caps->log_max_pd = out->hca_cap.log_max_pd & 0x1f;
-	caps->log_max_srq = out->hca_cap.log_max_srqs & 0x1f;
-	caps->local_ca_ack_delay = out->hca_cap.local_ca_ack_delay & 0x1f;
-	caps->log_max_mcg = out->hca_cap.log_max_mcg;
-	caps->max_qp_mcg = be32_to_cpu(out->hca_cap.max_qp_mcg) & 0xffffff;
-	caps->max_ra_res_qp = 1 << (out->hca_cap.log_max_ra_res_qp & 0x3f);
-	caps->max_ra_req_qp = 1 << (out->hca_cap.log_max_ra_req_qp & 0x3f);
-	caps->max_srq_wqes = 1 << out->hca_cap.log_max_srq_sz;
-	t16 = be16_to_cpu(out->hca_cap.bf_log_bf_reg_size);
-	if (t16 & 0x8000) {
-		caps->bf_reg_size = 1 << (t16 & 0x1f);
-		caps->bf_regs_per_page = MLX5_BF_REGS_PER_PAGE;
-	} else {
-		caps->bf_reg_size = 0;
-		caps->bf_regs_per_page = 0;
-	}
-	caps->min_page_sz = ~(u32)((1 << out->hca_cap.log_pg_sz) - 1);
-
-	memset(&ctx_in, 0, sizeof(ctx_in));
-	memset(&ctx_out, 0, sizeof(ctx_out));
-	ctx_in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_QUERY_SPECIAL_CONTEXTS);
-	err = mlx5_cmd_exec(dev, &ctx_in, sizeof(ctx_in),
-				 &ctx_out, sizeof(ctx_out));
-	if (err)
-		goto out_out;
-
-	if (ctx_out.hdr.status)
-		err = mlx5_cmd_status_to_err(&ctx_out.hdr);
-
-	caps->reserved_lkey = be32_to_cpu(ctx_out.reserved_lkey);
-
-out_out:
-	kfree(out);
-
-	return err;
+	return mlx5_core_get_caps(dev, caps, HCA_CAP_OPMOD_GET_CUR);
 }
 
 int mlx5_cmd_init_hca(struct mlx5_core_dev *dev)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/main.c b/drivers/net/ethernet/mellanox/mlx5/core/main.c
index f2716cc1f51d..d9f74618befa 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/main.c
@@ -207,11 +207,11 @@ static void release_bar(struct pci_dev *pdev)
 static int mlx5_enable_msix(struct mlx5_core_dev *dev)
 {
 	struct mlx5_eq_table *table = &dev->priv.eq_table;
-	int num_eqs = 1 << dev->caps.log_max_eq;
+	int num_eqs = 1 << dev->caps.gen.log_max_eq;
 	int nvec;
 	int i;
 
-	nvec = dev->caps.num_ports * num_online_cpus() + MLX5_EQ_VEC_COMP_BASE;
+	nvec = dev->caps.gen.num_ports * num_online_cpus() + MLX5_EQ_VEC_COMP_BASE;
 	nvec = min_t(int, nvec, num_eqs);
 	if (nvec <= MLX5_EQ_VEC_COMP_BASE)
 		return -ENOMEM;
@@ -250,13 +250,34 @@ struct mlx5_reg_host_endianess {
 #define CAP_MASK(pos, size) ((u64)((1 << (size)) - 1) << (pos))
 
 enum {
-	MLX5_CAP_BITS_RW_MASK	= CAP_MASK(MLX5_CAP_OFF_CMDIF_CSUM, 2) |
-				  CAP_MASK(MLX5_CAP_OFF_DCT, 1),
+	MLX5_CAP_BITS_RW_MASK = CAP_MASK(MLX5_CAP_OFF_CMDIF_CSUM, 2) |
+				MLX5_DEV_CAP_FLAG_DCT,
 };
 
+static u16 to_fw_pkey_sz(u32 size)
+{
+	switch (size) {
+	case 128:
+		return 0;
+	case 256:
+		return 1;
+	case 512:
+		return 2;
+	case 1024:
+		return 3;
+	case 2048:
+		return 4;
+	case 4096:
+		return 5;
+	default:
+		pr_warn("invalid pkey table size %d\n", size);
+		return 0;
+	}
+}
+
 /* selectively copy writable fields clearing any reserved area
  */
-static void copy_rw_fields(struct mlx5_hca_cap *to, struct mlx5_hca_cap *from)
+static void copy_rw_fields(struct mlx5_hca_cap *to, struct mlx5_general_caps *from)
 {
 	u64 v64;
 
@@ -265,76 +286,172 @@ static void copy_rw_fields(struct mlx5_hca_cap *to, struct mlx5_hca_cap *from)
 	to->log_max_ra_res_dc = from->log_max_ra_res_dc & 0x3f;
 	to->log_max_ra_req_qp = from->log_max_ra_req_qp & 0x3f;
 	to->log_max_ra_res_qp = from->log_max_ra_res_qp & 0x3f;
-	to->log_max_atomic_size_qp = from->log_max_atomic_size_qp;
-	to->log_max_atomic_size_dc = from->log_max_atomic_size_dc;
-	v64 = be64_to_cpu(from->flags) & MLX5_CAP_BITS_RW_MASK;
+	to->pkey_table_size = cpu_to_be16(to_fw_pkey_sz(from->pkey_table_size));
+	v64 = from->flags & MLX5_CAP_BITS_RW_MASK;
 	to->flags = cpu_to_be64(v64);
 }
 
-enum {
-	HCA_CAP_OPMOD_GET_MAX	= 0,
-	HCA_CAP_OPMOD_GET_CUR	= 1,
-};
+static u16 get_pkey_table_size(int pkey)
+{
+	if (pkey > MLX5_MAX_LOG_PKEY_TABLE)
+		return 0;
 
-static int handle_hca_cap(struct mlx5_core_dev *dev)
+	return MLX5_MIN_PKEY_TABLE_SIZE << pkey;
+}
+
+static void fw2drv_caps(struct mlx5_caps *caps,
+			struct mlx5_cmd_query_hca_cap_mbox_out *out)
 {
-	struct mlx5_cmd_query_hca_cap_mbox_out *query_out = NULL;
-	struct mlx5_cmd_set_hca_cap_mbox_in *set_ctx = NULL;
-	struct mlx5_cmd_query_hca_cap_mbox_in query_ctx;
-	struct mlx5_cmd_set_hca_cap_mbox_out set_out;
-	u64 flags;
+	struct mlx5_general_caps *gen = &caps->gen;
+	u16 t16;
+
+	gen->max_srq_wqes = 1 << out->hca_cap.log_max_srq_sz;
+	gen->max_wqes = 1 << out->hca_cap.log_max_qp_sz;
+	gen->log_max_qp = out->hca_cap.log_max_qp & 0x1f;
+	gen->log_max_strq = out->hca_cap.log_max_strq_sz;
+	gen->log_max_srq = out->hca_cap.log_max_srqs & 0x1f;
+	gen->max_cqes = 1 << out->hca_cap.log_max_cq_sz;
+	gen->log_max_cq = out->hca_cap.log_max_cq & 0x1f;
+	gen->max_eqes = out->hca_cap.log_max_eq_sz;
+	gen->log_max_mkey = out->hca_cap.log_max_mkey & 0x3f;
+	gen->log_max_eq = out->hca_cap.log_max_eq & 0xf;
+	gen->max_indirection = out->hca_cap.max_indirection;
+	gen->log_max_mrw_sz = out->hca_cap.log_max_mrw_sz;
+	gen->log_max_bsf_list_size = 0;
+	gen->log_max_klm_list_size = 0;
+	gen->log_max_ra_req_dc = out->hca_cap.log_max_ra_req_dc;
+	gen->log_max_ra_res_dc = out->hca_cap.log_max_ra_res_dc;
+	gen->log_max_ra_req_qp = out->hca_cap.log_max_ra_req_qp;
+	gen->log_max_ra_res_qp = out->hca_cap.log_max_ra_res_qp;
+	gen->max_qp_counters = be16_to_cpu(out->hca_cap.max_qp_count);
+	gen->pkey_table_size = get_pkey_table_size(be16_to_cpu(out->hca_cap.pkey_table_size));
+	gen->local_ca_ack_delay = out->hca_cap.local_ca_ack_delay & 0x1f;
+	gen->num_ports = out->hca_cap.num_ports & 0xf;
+	gen->log_max_msg = out->hca_cap.log_max_msg & 0x1f;
+	gen->stat_rate_support = be16_to_cpu(out->hca_cap.stat_rate_support);
+	gen->flags = be64_to_cpu(out->hca_cap.flags);
+	pr_debug("flags = 0x%llx\n", gen->flags);
+	gen->uar_sz = out->hca_cap.uar_sz;
+	gen->min_log_pg_sz = out->hca_cap.log_pg_sz;
+
+	t16 = be16_to_cpu(out->hca_cap.bf_log_bf_reg_size);
+	if (t16 & 0x8000) {
+		gen->bf_reg_size = 1 << (t16 & 0x1f);
+		gen->bf_regs_per_page = MLX5_BF_REGS_PER_PAGE;
+	} else {
+		gen->bf_reg_size = 0;
+		gen->bf_regs_per_page = 0;
+	}
+	gen->max_sq_desc_sz = be16_to_cpu(out->hca_cap.max_desc_sz_sq);
+	gen->max_rq_desc_sz = be16_to_cpu(out->hca_cap.max_desc_sz_rq);
+	gen->max_qp_mcg = be32_to_cpu(out->hca_cap.max_qp_mcg) & 0xffffff;
+	gen->log_max_pd = out->hca_cap.log_max_pd & 0x1f;
+	gen->log_max_xrcd = out->hca_cap.log_max_xrcd;
+	gen->log_uar_page_sz = be16_to_cpu(out->hca_cap.log_uar_page_sz);
+}
+
+static const char *caps_opmod_str(u16 opmod)
+{
+	switch (opmod) {
+	case HCA_CAP_OPMOD_GET_MAX:
+		return "GET_MAX";
+	case HCA_CAP_OPMOD_GET_CUR:
+		return "GET_CUR";
+	default:
+		return "Invalid";
+	}
+}
+
+int mlx5_core_get_caps(struct mlx5_core_dev *dev, struct mlx5_caps *caps,
+		       u16 opmod)
+{
+	struct mlx5_cmd_query_hca_cap_mbox_out *out;
+	struct mlx5_cmd_query_hca_cap_mbox_in in;
 	int err;
 
-	memset(&query_ctx, 0, sizeof(query_ctx));
-	query_out = kzalloc(sizeof(*query_out), GFP_KERNEL);
-	if (!query_out)
+	memset(&in, 0, sizeof(in));
+	out = kzalloc(sizeof(*out), GFP_KERNEL);
+	if (!out)
 		return -ENOMEM;
 
-	set_ctx = kzalloc(sizeof(*set_ctx), GFP_KERNEL);
-	if (!set_ctx) {
-		err = -ENOMEM;
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_QUERY_HCA_CAP);
+	in.hdr.opmod  = cpu_to_be16(opmod);
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), out, sizeof(*out));
+
+	err = mlx5_cmd_status_to_err(&out->hdr);
+	if (err) {
+		mlx5_core_warn(dev, "query max hca cap failed, %d\n", err);
 		goto query_ex;
 	}
+	mlx5_core_dbg(dev, "%s\n", caps_opmod_str(opmod));
+	fw2drv_caps(caps, out);
 
-	query_ctx.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_QUERY_HCA_CAP);
-	query_ctx.hdr.opmod  = cpu_to_be16(HCA_CAP_OPMOD_GET_CUR);
-	err = mlx5_cmd_exec(dev, &query_ctx, sizeof(query_ctx),
-				 query_out, sizeof(*query_out));
+query_ex:
+	kfree(out);
+	return err;
+}
+
+static int set_caps(struct mlx5_core_dev *dev,
+		    struct mlx5_cmd_set_hca_cap_mbox_in *in)
+{
+	struct mlx5_cmd_set_hca_cap_mbox_out out;
+	int err;
+
+	memset(&out, 0, sizeof(out));
+
+	in->hdr.opcode = cpu_to_be16(MLX5_CMD_OP_SET_HCA_CAP);
+	err = mlx5_cmd_exec(dev, in, sizeof(*in), &out, sizeof(out));
 	if (err)
-		goto query_ex;
+		return err;
 
-	err = mlx5_cmd_status_to_err(&query_out->hdr);
-	if (err) {
-		mlx5_core_warn(dev, "query hca cap failed, %d\n", err);
+	err = mlx5_cmd_status_to_err(&out.hdr);
+
+	return err;
+}
+
+static int handle_hca_cap(struct mlx5_core_dev *dev)
+{
+	struct mlx5_cmd_set_hca_cap_mbox_in *set_ctx = NULL;
+	struct mlx5_profile *prof = dev->profile;
+	struct mlx5_caps *cur_caps = NULL;
+	struct mlx5_caps *max_caps = NULL;
+	int err = -ENOMEM;
+
+	set_ctx = kzalloc(sizeof(*set_ctx), GFP_KERNEL);
+	if (!set_ctx)
 		goto query_ex;
-	}
 
-	copy_rw_fields(&set_ctx->hca_cap, &query_out->hca_cap);
+	max_caps = kzalloc(sizeof(*max_caps), GFP_KERNEL);
+	if (!max_caps)
+		goto query_ex;
 
-	if (dev->profile && dev->profile->mask & MLX5_PROF_MASK_QP_SIZE)
-		set_ctx->hca_cap.log_max_qp = dev->profile->log_max_qp;
+	cur_caps = kzalloc(sizeof(*cur_caps), GFP_KERNEL);
+	if (!cur_caps)
+		goto query_ex;
 
-	flags = be64_to_cpu(query_out->hca_cap.flags);
-	/* disable checksum */
-	flags &= ~MLX5_DEV_CAP_FLAG_CMDIF_CSUM;
-
-	set_ctx->hca_cap.flags = cpu_to_be64(flags);
-	memset(&set_out, 0, sizeof(set_out));
-	set_ctx->hca_cap.log_uar_page_sz = cpu_to_be16(PAGE_SHIFT - 12);
-	set_ctx->hdr.opcode = cpu_to_be16(MLX5_CMD_OP_SET_HCA_CAP);
-	err = mlx5_cmd_exec(dev, set_ctx, sizeof(*set_ctx),
-				 &set_out, sizeof(set_out));
-	if (err) {
-		mlx5_core_warn(dev, "set hca cap failed, %d\n", err);
+	err = mlx5_core_get_caps(dev, max_caps, HCA_CAP_OPMOD_GET_MAX);
+	if (err)
 		goto query_ex;
-	}
 
-	err = mlx5_cmd_status_to_err(&set_out.hdr);
+	err = mlx5_core_get_caps(dev, cur_caps, HCA_CAP_OPMOD_GET_CUR);
 	if (err)
 		goto query_ex;
 
+	/* we limit the size of the pkey table to 128 entries for now */
+	cur_caps->gen.pkey_table_size = 128;
+
+	if (prof->mask & MLX5_PROF_MASK_QP_SIZE)
+		cur_caps->gen.log_max_qp = prof->log_max_qp;
+
+	/* disable checksum */
+	cur_caps->gen.flags &= ~MLX5_DEV_CAP_FLAG_CMDIF_CSUM;
+
+	copy_rw_fields(&set_ctx->hca_cap, &cur_caps->gen);
+	err = set_caps(dev, set_ctx);
+
 query_ex:
-	kfree(query_out);
+	kfree(cur_caps);
+	kfree(max_caps);
 	kfree(set_ctx);
 
 	return err;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/uar.c b/drivers/net/ethernet/mellanox/mlx5/core/uar.c
index 68f5d9c77c7b..0a6348cefc01 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/uar.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/uar.c
@@ -174,11 +174,11 @@ int mlx5_alloc_uuars(struct mlx5_core_dev *dev, struct mlx5_uuar_info *uuari)
 	for (i = 0; i < tot_uuars; i++) {
 		bf = &uuari->bfs[i];
 
-		bf->buf_size = dev->caps.bf_reg_size / 2;
+		bf->buf_size = dev->caps.gen.bf_reg_size / 2;
 		bf->uar = &uuari->uars[i / MLX5_BF_REGS_PER_PAGE];
 		bf->regreg = uuari->uars[i / MLX5_BF_REGS_PER_PAGE].map;
 		bf->reg = NULL; /* Add WC support */
-		bf->offset = (i % MLX5_BF_REGS_PER_PAGE) * dev->caps.bf_reg_size +
+		bf->offset = (i % MLX5_BF_REGS_PER_PAGE) * dev->caps.gen.bf_reg_size +
 			MLX5_BF_OFFSET;
 		bf->need_lock = need_uuar_lock(i);
 		spin_lock_init(&bf->lock);
diff --git a/include/linux/mlx5/device.h b/include/linux/mlx5/device.h
index 334947151dfc..dce01fd854a8 100644
--- a/include/linux/mlx5/device.h
+++ b/include/linux/mlx5/device.h
@@ -71,6 +71,11 @@ enum {
 };
 
 enum {
+	MLX5_MIN_PKEY_TABLE_SIZE = 128,
+	MLX5_MAX_LOG_PKEY_TABLE  = 5,
+};
+
+enum {
 	MLX5_PERM_LOCAL_READ	= 1 << 2,
 	MLX5_PERM_LOCAL_WRITE	= 1 << 3,
 	MLX5_PERM_REMOTE_READ	= 1 << 4,
@@ -184,10 +189,10 @@ enum {
 	MLX5_DEV_CAP_FLAG_CQ_MODER	= 1LL << 29,
 	MLX5_DEV_CAP_FLAG_RESIZE_CQ	= 1LL << 30,
 	MLX5_DEV_CAP_FLAG_RESIZE_SRQ	= 1LL << 32,
+	MLX5_DEV_CAP_FLAG_DCT		= 1LL << 37,
 	MLX5_DEV_CAP_FLAG_REMOTE_FENCE	= 1LL << 38,
 	MLX5_DEV_CAP_FLAG_TLP_HINTS	= 1LL << 39,
 	MLX5_DEV_CAP_FLAG_SIG_HAND_OVER	= 1LL << 40,
-	MLX5_DEV_CAP_FLAG_DCT		= 1LL << 41,
 	MLX5_DEV_CAP_FLAG_CMDIF_CSUM	= 3LL << 46,
 };
 
@@ -243,10 +248,14 @@ enum {
 };
 
 enum {
-	MLX5_CAP_OFF_DCT		= 41,
 	MLX5_CAP_OFF_CMDIF_CSUM		= 46,
 };
 
+enum {
+	HCA_CAP_OPMOD_GET_MAX	= 0,
+	HCA_CAP_OPMOD_GET_CUR	= 1,
+};
+
 struct mlx5_inbox_hdr {
 	__be16		opcode;
 	u8		rsvd[4];
@@ -303,9 +312,10 @@ struct mlx5_hca_cap {
 	u8	log_max_ra_req_qp;
 	u8	rsvd10;
 	u8	log_max_ra_res_qp;
-	u8	rsvd11[4];
+	u8	pad_cap;
+	u8	rsvd11[3];
 	__be16	max_qp_count;
-	__be16	rsvd12;
+	__be16	pkey_table_size;
 	u8	rsvd13;
 	u8	local_ca_ack_delay;
 	u8	rsvd14;
@@ -335,11 +345,7 @@ struct mlx5_hca_cap {
 	u8	log_max_xrcd;
 	u8	rsvd25[42];
 	__be16  log_uar_page_sz;
-	u8	rsvd26[28];
-	u8	log_max_atomic_size_qp;
-	u8	rsvd27[2];
-	u8	log_max_atomic_size_dc;
-	u8	rsvd28[76];
+	u8	rsvd26[108];
 };
 
 
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index b88e9b46d957..45a2add747e0 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -335,23 +335,30 @@ struct mlx5_port_caps {
 	int	pkey_table_len;
 };
 
-struct mlx5_caps {
+struct mlx5_general_caps {
 	u8	log_max_eq;
 	u8	log_max_cq;
 	u8	log_max_qp;
 	u8	log_max_mkey;
 	u8	log_max_pd;
 	u8	log_max_srq;
+	u8	log_max_strq;
+	u8	log_max_mrw_sz;
+	u8	log_max_bsf_list_size;
+	u8	log_max_klm_list_size;
 	u32	max_cqes;
 	int	max_wqes;
+	u32	max_eqes;
+	u32	max_indirection;
 	int	max_sq_desc_sz;
 	int	max_rq_desc_sz;
+	int	max_dc_sq_desc_sz;
 	u64	flags;
 	u16	stat_rate_support;
 	int	log_max_msg;
 	int	num_ports;
-	int	max_ra_res_qp;
-	int	max_ra_req_qp;
+	u8	log_max_ra_res_qp;
+	u8	log_max_ra_req_qp;
 	int	max_srq_wqes;
 	int	bf_reg_size;
 	int	bf_regs_per_page;
@@ -363,6 +370,19 @@ struct mlx5_caps {
 	u8	log_max_mcg;
 	u32	max_qp_mcg;
 	int	min_page_sz;
+	int	pd_cap;
+	u32	max_qp_counters;
+	u32	pkey_table_size;
+	u8	log_max_ra_req_dc;
+	u8	log_max_ra_res_dc;
+	u32	uar_sz;
+	u8	min_log_pg_sz;
+	u8	log_max_xrcd;
+	u16	log_uar_page_sz;
+};
+
+struct mlx5_caps {
+	struct mlx5_general_caps gen;
 };
 
 struct mlx5_cmd_mailbox {
@@ -695,6 +715,8 @@ void mlx5_cmd_cleanup(struct mlx5_core_dev *dev);
 void mlx5_cmd_use_events(struct mlx5_core_dev *dev);
 void mlx5_cmd_use_polling(struct mlx5_core_dev *dev);
 int mlx5_cmd_status_to_err(struct mlx5_outbox_hdr *hdr);
+int mlx5_core_get_caps(struct mlx5_core_dev *dev, struct mlx5_caps *caps,
+		       u16 opmod);
 int mlx5_cmd_exec(struct mlx5_core_dev *dev, void *in, int in_size, void *out,
 		  int out_size);
 int mlx5_cmd_exec_cb(struct mlx5_core_dev *dev, void *in, int in_size,