Merge branches 'af_ib', 'cxgb4', 'misc', 'mlx5', 'ocrdma', 'qib' and 'srp' into for-next

Documentation/ABI/stable/sysfs-driver-ib_srp

@@ -54,6 +54,13 @@ Description:	Interface for making ib_srp connect to a new target.
 		  ib_srp. Specifying a value that exceeds cmd_sg_entries is
 		  only safe with partial memory descriptor list support enabled
 		  (allow_ext_sg=1).
+		* comp_vector, a number in the range 0..n-1 specifying the
+		  MSI-X completion vector. Some HCA's allocate multiple (n)
+		  MSI-X vectors per HCA port. If the IRQ affinity masks of
+		  these interrupts have been configured such that each MSI-X
+		  interrupt is handled by a different CPU then the comp_vector
+		  parameter can be used to spread the SRP completion workload
+		  over multiple CPU's.
 
 What:		/sys/class/infiniband_srp/srp-<hca>-<port_number>/ibdev
 Date:		January 2, 2006

MAINTAINERS

@@ -5365,6 +5365,28 @@ W:	http://linuxtv.org
 S:	Odd Fixes
 F:	drivers/media/radio/radio-miropcm20*
 
+Mellanox MLX5 core VPI driver
+M:	Eli Cohen <eli@mellanox.com>
+L:	netdev@vger.kernel.org
+L:	linux-rdma@vger.kernel.org
+W:	http://www.mellanox.com
+Q:	http://patchwork.ozlabs.org/project/netdev/list/
+Q:	http://patchwork.kernel.org/project/linux-rdma/list/
+T:	git://openfabrics.org/~eli/connect-ib.git
+S:	Supported
+F:	drivers/net/ethernet/mellanox/mlx5/core/
+F:	include/linux/mlx5/
+
+Mellanox MLX5 IB driver
+M:      Eli Cohen <eli@mellanox.com>
+L:      linux-rdma@vger.kernel.org
+W:      http://www.mellanox.com
+Q:      http://patchwork.kernel.org/project/linux-rdma/list/
+T:      git://openfabrics.org/~eli/connect-ib.git
+S:      Supported
+F:      include/linux/mlx5/
+F:      drivers/infiniband/hw/mlx5/
+
 MODULE SUPPORT
 M:	Rusty Russell <rusty@rustcorp.com.au>
 S:	Maintained

drivers/infiniband/Kconfig

@@ -50,6 +50,7 @@ source "drivers/infiniband/hw/amso1100/Kconfig"
 source "drivers/infiniband/hw/cxgb3/Kconfig"
 source "drivers/infiniband/hw/cxgb4/Kconfig"
 source "drivers/infiniband/hw/mlx4/Kconfig"
+source "drivers/infiniband/hw/mlx5/Kconfig"
 source "drivers/infiniband/hw/nes/Kconfig"
 source "drivers/infiniband/hw/ocrdma/Kconfig"
 

drivers/infiniband/Makefile

@@ -7,6 +7,7 @@ obj-$(CONFIG_INFINIBAND_AMSO1100)	+= hw/amso1100/
 obj-$(CONFIG_INFINIBAND_CXGB3)		+= hw/cxgb3/
 obj-$(CONFIG_INFINIBAND_CXGB4)		+= hw/cxgb4/
 obj-$(CONFIG_MLX4_INFINIBAND)		+= hw/mlx4/
+obj-$(CONFIG_MLX5_INFINIBAND)		+= hw/mlx5/
 obj-$(CONFIG_INFINIBAND_NES)		+= hw/nes/
 obj-$(CONFIG_INFINIBAND_OCRDMA)		+= hw/ocrdma/
 obj-$(CONFIG_INFINIBAND_IPOIB)		+= ulp/ipoib/

drivers/infiniband/core/sysfs.c

@@ -545,8 +545,10 @@ static int add_port(struct ib_device *device, int port_num,
 
 	p->gid_group.name  = "gids";
 	p->gid_group.attrs = alloc_group_attrs(show_port_gid, attr.gid_tbl_len);
-	if (!p->gid_group.attrs)
+	if (!p->gid_group.attrs) {
+		ret = -ENOMEM;
 		goto err_remove_pma;
+	}
 
 	ret = sysfs_create_group(&p->kobj, &p->gid_group);
 	if (ret)
@@ -555,8 +557,10 @@ static int add_port(struct ib_device *device, int port_num,
 	p->pkey_group.name  = "pkeys";
 	p->pkey_group.attrs = alloc_group_attrs(show_port_pkey,
 						attr.pkey_tbl_len);
-	if (!p->pkey_group.attrs)
+	if (!p->pkey_group.attrs) {
+		ret = -ENOMEM;
 		goto err_remove_gid;
+	}
 
 	ret = sysfs_create_group(&p->kobj, &p->pkey_group);
 	if (ret)

drivers/infiniband/core/uverbs_cmd.c

@@ -334,7 +334,7 @@ ssize_t ib_uverbs_get_context(struct ib_uverbs_file *file,
 
 	resp.num_comp_vectors = file->device->num_comp_vectors;
 
-	ret = get_unused_fd();
+	ret = get_unused_fd_flags(O_CLOEXEC);
 	if (ret < 0)
 		goto err_free;
 	resp.async_fd = ret;
@@ -1184,7 +1184,7 @@ ssize_t ib_uverbs_create_comp_channel(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
-	ret = get_unused_fd();
+	ret = get_unused_fd_flags(O_CLOEXEC);
 	if (ret < 0)
 		return ret;
 	resp.fd = ret;

drivers/infiniband/hw/cxgb3/iwch_qp.c

@@ -883,7 +883,8 @@ u16 iwch_rqes_posted(struct iwch_qp *qhp)
 {
 	union t3_wr *wqe = qhp->wq.queue;
 	u16 count = 0;
-	while ((count+1) != 0 && fw_riwrh_opcode((struct fw_riwrh *)wqe) == T3_WR_RCV) {
+
+	while (count < USHRT_MAX && fw_riwrh_opcode((struct fw_riwrh *)wqe) == T3_WR_RCV) {
 		count++;
 		wqe++;
 	}

drivers/infiniband/hw/ehca/ehca_main.c

@@ -211,6 +211,7 @@ static int ehca_create_slab_caches(void)
 	if (!ctblk_cache) {
 		ehca_gen_err("Cannot create ctblk SLAB cache.");
 		ehca_cleanup_small_qp_cache();
+		ret = -ENOMEM;
 		goto create_slab_caches6;
 	}
 #endif

drivers/infiniband/hw/mlx5/Kconfig

@@ -0,0 +1,10 @@
+config MLX5_INFINIBAND
+	tristate "Mellanox Connect-IB HCA support"
+	depends on NETDEVICES && ETHERNET && PCI && X86
+	select NET_VENDOR_MELLANOX
+	select MLX5_CORE
+	---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.

drivers/infiniband/hw/mlx5/Makefile

@@ -0,0 +1,3 @@
+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

drivers/infiniband/hw/mlx5/ah.c

@@ -0,0 +1,92 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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 ib_ah *create_ib_ah(struct ib_ah_attr *ah_attr,
+			   struct mlx5_ib_ah *ah)
+{
+	if (ah_attr->ah_flags & IB_AH_GRH) {
+		memcpy(ah->av.rgid, &ah_attr->grh.dgid, 16);
+		ah->av.grh_gid_fl = cpu_to_be32(ah_attr->grh.flow_label |
+						(1 << 30) |
+						ah_attr->grh.sgid_index << 20);
+		ah->av.hop_limit = ah_attr->grh.hop_limit;
+		ah->av.tclass = ah_attr->grh.traffic_class;
+	}
+
+	ah->av.rlid = cpu_to_be16(ah_attr->dlid);
+	ah->av.fl_mlid = ah_attr->src_path_bits & 0x7f;
+	ah->av.stat_rate_sl = (ah_attr->static_rate << 4) | (ah_attr->sl & 0xf);
+
+	return &ah->ibah;
+}
+
+struct ib_ah *mlx5_ib_create_ah(struct ib_pd *pd, struct ib_ah_attr *ah_attr)
+{
+	struct mlx5_ib_ah *ah;
+
+	ah = kzalloc(sizeof(*ah), GFP_ATOMIC);
+	if (!ah)
+		return ERR_PTR(-ENOMEM);
+
+	return create_ib_ah(ah_attr, ah); /* never fails */
+}
+
+int mlx5_ib_query_ah(struct ib_ah *ibah, struct ib_ah_attr *ah_attr)
+{
+	struct mlx5_ib_ah *ah = to_mah(ibah);
+	u32 tmp;
+
+	memset(ah_attr, 0, sizeof(*ah_attr));
+
+	tmp = be32_to_cpu(ah->av.grh_gid_fl);
+	if (tmp & (1 << 30)) {
+		ah_attr->ah_flags = IB_AH_GRH;
+		ah_attr->grh.sgid_index = (tmp >> 20) & 0xff;
+		ah_attr->grh.flow_label = tmp & 0xfffff;
+		memcpy(&ah_attr->grh.dgid, ah->av.rgid, 16);
+		ah_attr->grh.hop_limit = ah->av.hop_limit;
+		ah_attr->grh.traffic_class = ah->av.tclass;
+	}
+	ah_attr->dlid = be16_to_cpu(ah->av.rlid);
+	ah_attr->static_rate = ah->av.stat_rate_sl >> 4;
+	ah_attr->sl = ah->av.stat_rate_sl & 0xf;
+
+	return 0;
+}
+
+int mlx5_ib_destroy_ah(struct ib_ah *ah)
+{
+	kfree(to_mah(ah));
+	return 0;
+}

drivers/infiniband/hw/mlx5/cq.c

@@ -0,0 +1,843 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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 "mlx5_ib.h"
+#include "user.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_from_buf(struct mlx5_ib_cq_buf *buf, int n, int size)
+{
+	return mlx5_buf_offset(&buf->buf, n * size);
+}
+
+static void *get_cqe(struct mlx5_ib_cq *cq, int n)
+{
+	return get_cqe_from_buf(&cq->buf, n, cq->mcq.cqe_sz);
+}
+
+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;
+	return ((cqe64->op_own & MLX5_CQE_OWNER_MASK) ^
+		!!(n & (cq->ibcq.cqe + 1))) ? NULL : cqe;
+}
+
+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_FAST_REG_MR:
+		return IB_WC_FAST_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;
+	case MLX5_OPCODE_RDMA_WRITE:
+		wc->opcode    = IB_WC_RDMA_WRITE;
+		break;
+	case MLX5_OPCODE_SEND_IMM:
+		wc->wc_flags |= IB_WC_WITH_IMM;
+	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_BIND_MW:
+		wc->opcode    = IB_WC_BIND_MW;
+		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)
+{
+	struct mlx5_ib_dev *dev = to_mdev(qp->ibqp.device);
+	struct mlx5_ib_srq *srq;
+	struct mlx5_ib_wq *wq;
+	u16 wqe_ctr;
+	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 = 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->slid	   = be16_to_cpu(cqe->slid);
+	wc->sl		   = (be32_to_cpu(cqe->flags_rqpn) >> 24) & 0xf;
+	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;
+	wc->pkey_index     = be32_to_cpu(cqe->imm_inval_pkey) & 0xffff;
+}
+
+static void dump_cqe(struct mlx5_ib_dev *dev, struct mlx5_err_cqe *cqe)
+{
+	__be32 *p = (__be32 *)cqe;
+	int i;
+
+	mlx5_ib_warn(dev, "dump error cqe\n");
+	for (i = 0; i < sizeof(*cqe) / 16; i++, p += 4)
+		pr_info("%08x %08x %08x %08x\n", be32_to_cpu(p[0]),
+			be32_to_cpu(p[1]), be32_to_cpu(p[2]),
+			be32_to_cpu(p[3]));
+}
+
+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)
+{
+	int 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 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;
+	uint8_t opcode;
+	uint32_t qpn;
+	u16 wqe_ctr;
+	void *cqe;
+	int idx;
+
+	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();
+
+	/* TBD: resize CQ */
+
+	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);
+		if (unlikely(!mqp)) {
+			mlx5_ib_warn(dev, "CQE@CQ %06x for unknown QPN %6x\n",
+				     cq->mcq.cqn, qpn);
+			return -EINVAL;
+		}
+
+		*cur_qp = to_mibqp(mqp);
+	}
+
+	wc->qp  = &(*cur_qp)->ibqp;
+	opcode = cqe64->op_own >> 4;
+	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;
+	}
+
+	return 0;
+}
+
+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;
+	unsigned long flags;
+	int npolled;
+	int err = 0;
+
+	spin_lock_irqsave(&cq->lock, flags);
+
+	for (npolled = 0; npolled < num_entries; npolled++) {
+		err = mlx5_poll_one(cq, &cur_qp, wc + npolled);
+		if (err)
+			break;
+	}
+
+	if (npolled)
+		mlx5_cq_set_ci(&cq->mcq);
+
+	spin_unlock_irqrestore(&cq->lock, flags);
+
+	if (err == 0 || err == -EAGAIN)
+		return npolled;
+	else
+		return err;
+}
+
+int mlx5_ib_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags flags)
+{
+	mlx5_cq_arm(&to_mcq(ibcq)->mcq,
+		    (flags & IB_CQ_SOLICITED_MASK) == IB_CQ_SOLICITED ?
+		    MLX5_CQ_DB_REQ_NOT_SOL : MLX5_CQ_DB_REQ_NOT,
+		    to_mdev(ibcq->device)->mdev.priv.uuari.uars[0].map,
+		    MLX5_GET_DOORBELL_LOCK(&to_mdev(ibcq->device)->mdev.priv.cq_uar_lock));
+
+	return 0;
+}
+
+static int alloc_cq_buf(struct mlx5_ib_dev *dev, struct mlx5_ib_cq_buf *buf,
+			int nent, int cqe_size)
+{
+	int err;
+
+	err = mlx5_buf_alloc(&dev->mdev, nent * cqe_size,
+			     PAGE_SIZE * 2, &buf->buf);
+	if (err)
+		return err;
+
+	buf->cqe_size = cqe_size;
+
+	return 0;
+}
+
+static void free_cq_buf(struct mlx5_ib_dev *dev, struct mlx5_ib_cq_buf *buf)
+{
+	mlx5_buf_free(&dev->mdev, &buf->buf);
+}
+
+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, struct mlx5_create_cq_mbox_in **cqb,
+			  int *cqe_size, int *index, int *inlen)
+{
+	struct mlx5_ib_create_cq ucmd;
+	int page_shift;
+	int npages;
+	int ncont;
+	int err;
+
+	if (ib_copy_from_udata(&ucmd, udata, sizeof(ucmd)))
+		return -EFAULT;
+
+	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, &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 = sizeof(**cqb) + sizeof(*(*cqb)->pas) * ncont;
+	*cqb = mlx5_vzalloc(*inlen);
+	if (!*cqb) {
+		err = -ENOMEM;
+		goto err_db;
+	}
+	mlx5_ib_populate_pas(dev, cq->buf.umem, page_shift, (*cqb)->pas, 0);
+	(*cqb)->ctx.log_pg_sz = page_shift - PAGE_SHIFT;
+
+	*index = to_mucontext(context)->uuari.uars[0].index;
+
+	return 0;
+
+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_buf(struct mlx5_ib_cq *cq, int nent)
+{
+	int i;
+	void *cqe;
+	struct mlx5_cqe64 *cqe64;
+
+	for (i = 0; i < nent; i++) {
+		cqe = get_cqe(cq, i);
+		cqe64 = (cq->buf.cqe_size == 64) ? cqe : cqe + 64;
+		cqe64->op_own = 0xf1;
+	}
+}
+
+static int create_cq_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_cq *cq,
+			    int entries, int cqe_size,
+			    struct mlx5_create_cq_mbox_in **cqb,
+			    int *index, int *inlen)
+{
+	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.set_ci_db = 0;
+	*cq->mcq.arm_db    = 0;
+	cq->mcq.cqe_sz = cqe_size;
+
+	err = alloc_cq_buf(dev, &cq->buf, entries, cqe_size);
+	if (err)
+		goto err_db;
+
+	init_cq_buf(cq, entries);
+
+	*inlen = sizeof(**cqb) + sizeof(*(*cqb)->pas) * cq->buf.buf.npages;
+	*cqb = mlx5_vzalloc(*inlen);
+	if (!*cqb) {
+		err = -ENOMEM;
+		goto err_buf;
+	}
+	mlx5_fill_page_array(&cq->buf.buf, (*cqb)->pas);
+
+	(*cqb)->ctx.log_pg_sz = cq->buf.buf.page_shift - PAGE_SHIFT;
+	*index = dev->mdev.priv.uuari.uars[0].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);
+}
+
+struct ib_cq *mlx5_ib_create_cq(struct ib_device *ibdev, int entries,
+				int vector, struct ib_ucontext *context,
+				struct ib_udata *udata)
+{
+	struct mlx5_create_cq_mbox_in *cqb = NULL;
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+	struct mlx5_ib_cq *cq;
+	int uninitialized_var(index);
+	int uninitialized_var(inlen);
+	int cqe_size;
+	int irqn;
+	int eqn;
+	int err;
+
+	entries = roundup_pow_of_two(entries + 1);
+	if (entries < 1 || entries > dev->mdev.caps.max_cqes)
+		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;
+
+	if (context) {
+		err = create_cq_user(dev, udata, context, cq, entries,
+				     &cqb, &cqe_size, &index, &inlen);
+		if (err)
+			goto err_create;
+	} else {
+		/* for now choose 64 bytes till we have a proper interface */
+		cqe_size = 64;
+		err = create_cq_kernel(dev, cq, entries, cqe_size, &cqb,
+				       &index, &inlen);
+		if (err)
+			goto err_create;
+	}
+
+	cq->cqe_size = cqe_size;
+	cqb->ctx.cqe_sz_flags = cqe_sz_to_mlx_sz(cqe_size) << 5;
+	cqb->ctx.log_sz_usr_page = cpu_to_be32((ilog2(entries) << 24) | index);
+	err = mlx5_vector2eqn(dev, vector, &eqn, &irqn);
+	if (err)
+		goto err_cqb;
+
+	cqb->ctx.c_eqn = cpu_to_be16(eqn);
+	cqb->ctx.db_record_addr = cpu_to_be64(cq->db.dma);
+
+	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;
+	cq->mcq.comp  = mlx5_ib_cq_comp;
+	cq->mcq.event = mlx5_ib_cq_event;
+
+	if (context)
+		if (ib_copy_to_udata(udata, &cq->mcq.cqn, sizeof(__u32))) {
+			err = -EFAULT;
+			goto err_cmd;
+		}
+
+
+	mlx5_vfree(cqb);
+	return &cq->ibcq;
+
+err_cmd:
+	mlx5_core_destroy_cq(&dev->mdev, &cq->mcq);
+
+err_cqb:
+	mlx5_vfree(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, struct mlx5_ib_srq *srq,
+			u32 rsn)
+{
+	u32 lrsn;
+
+	if (srq)
+		lrsn = be32_to_cpu(cqe64->srqn) & 0xffffff;
+	else
+		lrsn = be32_to_cpu(cqe64->sop_drop_qpn) & 0xffffff;
+
+	return rsn == lrsn;
+}
+
+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, srq, rsn)) {
+			if (srq)
+				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)
+{
+	return -ENOSYS;
+}
+
+int mlx5_ib_resize_cq(struct ib_cq *ibcq, int entries, struct ib_udata *udata)
+{
+	return -ENOSYS;
+}
+
+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;
+}

drivers/infiniband/hw/mlx5/doorbell.c

@@ -0,0 +1,100 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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;
+	struct ib_umem_chunk *chunk;
+	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:
+	chunk = list_entry(page->umem->chunk_list.next, struct ib_umem_chunk, list);
+	db->dma		= sg_dma_address(chunk->page_list) + (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);
+}

drivers/infiniband/hw/mlx5/mad.c

@@ -0,0 +1,139 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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 <rdma/ib_mad.h>
+#include <rdma/ib_smi.h>
+#include "mlx5_ib.h"
+
+enum {
+	MLX5_IB_VENDOR_CLASS1 = 0x9,
+	MLX5_IB_VENDOR_CLASS2 = 0xa
+};
+
+int mlx5_MAD_IFC(struct mlx5_ib_dev *dev, int ignore_mkey, int ignore_bkey,
+		 int port, struct ib_wc *in_wc, struct ib_grh *in_grh,
+		 void *in_mad, void *response_mad)
+{
+	u8 op_modifier = 0;
+
+	/* 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);
+}
+
+int mlx5_ib_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num,
+			struct ib_wc *in_wc, struct ib_grh *in_grh,
+			struct ib_mad *in_mad, struct ib_mad *out_mad)
+{
+	u16 slid;
+	int err;
+
+	slid = in_wc ? 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;
+}
+
+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.caps.ext_port_cap[port - 1] = (!err && !packet_error) ?
+		MLX_EXT_PORT_CAP_FLAG_EXTENDED_PORT_INFO : 0;
+
+out:
+	kfree(in_mad);
+	kfree(out_mad);
+	return err;
+}

drivers/infiniband/hw/mlx5/mem.c

@@ -0,0 +1,162 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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 "mlx5_ib.h"
+
+/* @umem: umem object to scan
+ * @addr: ib virtual address requested by the user
+ * @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, int *count, int *shift,
+			int *ncont, int *order)
+{
+	struct ib_umem_chunk *chunk;
+	unsigned long tmp;
+	unsigned long m;
+	int i, j, k;
+	u64 base = 0;
+	int p = 0;
+	int skip;
+	int mask;
+	u64 len;
+	u64 pfn;
+
+	addr = addr >> PAGE_SHIFT;
+	tmp = (unsigned long)addr;
+	m = find_first_bit(&tmp, sizeof(tmp));
+	skip = 1 << m;
+	mask = skip - 1;
+	i = 0;
+	list_for_each_entry(chunk, &umem->chunk_list, list)
+		for (j = 0; j < chunk->nmap; j++) {
+			len = sg_dma_len(&chunk->page_list[j]) >> PAGE_SHIFT;
+			pfn = sg_dma_address(&chunk->page_list[j]) >> PAGE_SHIFT;
+			for (k = 0; k < len; k++) {
+				if (!(i & mask)) {
+					tmp = (unsigned long)pfn;
+					m = min(m, find_first_bit(&tmp, sizeof(tmp)));
+					skip = 1 << m;
+					mask = skip - 1;
+					base = pfn;
+					p = 0;
+				} else {
+					if (base + p != pfn) {
+						tmp = (unsigned long)p;
+						m = find_first_bit(&tmp, sizeof(tmp));
+						skip = 1 << m;
+						mask = skip - 1;
+						base = pfn;
+						p = 0;
+					}
+				}
+				p++;
+				i++;
+			}
+		}
+
+	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;
+}
+
+void mlx5_ib_populate_pas(struct mlx5_ib_dev *dev, struct ib_umem *umem,
+			  int page_shift, __be64 *pas, int umr)
+{
+	int shift = page_shift - PAGE_SHIFT;
+	int mask = (1 << shift) - 1;
+	struct ib_umem_chunk *chunk;
+	int i, j, k;
+	u64 cur = 0;
+	u64 base;
+	int len;
+
+	i = 0;
+	list_for_each_entry(chunk, &umem->chunk_list, list)
+		for (j = 0; j < chunk->nmap; j++) {
+			len = sg_dma_len(&chunk->page_list[j]) >> PAGE_SHIFT;
+			base = sg_dma_address(&chunk->page_list[j]);
+			for (k = 0; k < len; k++) {
+				if (!(i & mask)) {
+					cur = base + (k << PAGE_SHIFT);
+					if (umr)
+						cur |= 3;
+
+					pas[i >> shift] = cpu_to_be64(cur);
+					mlx5_ib_dbg(dev, "pas[%d] 0x%llx\n",
+						    i >> shift, be64_to_cpu(pas[i >> shift]));
+				}  else
+					mlx5_ib_dbg(dev, "=====> 0x%llx\n",
+						    base + (k << PAGE_SHIFT));
+				i++;
+			}
+		}
+}
+
+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 = 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;
+}

drivers/infiniband/hw/mlx5/mlx5_ib.h

@@ -0,0 +1,545 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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>
+
+#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)
+
+enum {
+	MLX5_IB_MMAP_CMD_SHIFT	= 8,
+	MLX5_IB_MMAP_CMD_MASK	= 0xff,
+};
+
+enum mlx5_ib_mmap_cmd {
+	MLX5_IB_MMAP_REGULAR_PAGE		= 0,
+	MLX5_IB_MMAP_GET_CONTIGUOUS_PAGES	= 1, /* always last */
+};
+
+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_latency_class {
+	MLX5_IB_LATENCY_CLASS_LOW,
+	MLX5_IB_LATENCY_CLASS_MEDIUM,
+	MLX5_IB_LATENCY_CLASS_HIGH,
+	MLX5_IB_LATENCY_CLASS_FAST_PATH
+};
+
+enum mlx5_ib_mad_ifc_flags {
+	MLX5_MAD_IFC_IGNORE_MKEY	= 1,
+	MLX5_MAD_IFC_IGNORE_BKEY	= 2,
+	MLX5_MAD_IFC_NET_VIEW		= 4,
+};
+
+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_uuar_info	uuari;
+};
+
+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;
+	u32			pa_lkey;
+};
+
+/* 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_UNREG	IB_SEND_RESERVED_START
+#define MLX5_IB_QPT_REG_UMR	IB_QPT_RESERVED1
+#define MLX5_IB_WR_UMR		IB_WR_RESERVED1
+
+struct wr_list {
+	u16	opcode;
+	u16	next;
+};
+
+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_QP_USER,
+	MLX5_QP_KERNEL,
+	MLX5_QP_EMPTY
+};
+
+struct mlx5_ib_qp {
+	struct ib_qp		ibqp;
+	struct mlx5_core_qp	mqp;
+	struct mlx5_buf		buf;
+
+	struct mlx5_db		db;
+	struct mlx5_ib_wq	rq;
+
+	u32			doorbell_qpn;
+	u8			sq_signal_bits;
+	u8			fm_cache;
+	int			sq_max_wqes_per_wr;
+	int			sq_spare_wqes;
+	struct mlx5_ib_wq	sq;
+
+	struct ib_umem	       *umem;
+	int			buf_size;
+
+	/* serialize qp state modifications
+	 */
+	struct mutex		mutex;
+	u16			xrcdn;
+	u32			flags;
+	u8			port;
+	u8			alt_port;
+	u8			atomic_rd_en;
+	u8			resp_depth;
+	u8			state;
+	int			mlx_type;
+	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			uuarn;
+
+	int			create_type;
+	u32			pa_lkey;
+};
+
+struct mlx5_ib_cq_buf {
+	struct mlx5_buf		buf;
+	struct ib_umem		*umem;
+	int			cqe_size;
+};
+
+enum mlx5_ib_qp_flags {
+	MLX5_IB_QP_BLOCK_MULTICAST_LOOPBACK     = 1 << 0,
+	MLX5_IB_QP_SIGNATURE_HANDLING           = 1 << 1,
+};
+
+struct mlx5_shared_mr_info {
+	int mr_id;
+	struct ib_umem		*umem;
+};
+
+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_resize *resize_buf;
+	struct ib_umem	       *resize_umem;
+	int			cqe_size;
+};
+
+struct mlx5_ib_srq {
+	struct ib_srq		ibsrq;
+	struct mlx5_core_srq	msrq;
+	struct mlx5_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;
+};
+
+struct mlx5_ib_mr {
+	struct ib_mr		ibmr;
+	struct mlx5_core_mr	mmr;
+	struct ib_umem	       *umem;
+	struct mlx5_shared_mr_info	*smr_info;
+	struct list_head	list;
+	int			order;
+	int			umred;
+	__be64			*pas;
+	dma_addr_t		dma;
+	int			npages;
+	struct completion	done;
+	enum ib_wc_status	status;
+};
+
+struct mlx5_ib_fast_reg_page_list {
+	struct ib_fast_reg_page_list	ibfrpl;
+	__be64			       *mapped_page_list;
+	dma_addr_t			map;
+};
+
+struct umr_common {
+	struct ib_pd	*pd;
+	struct ib_cq	*cq;
+	struct ib_qp	*qp;
+	struct ib_mr	*mr;
+	/* control access to UMR QP
+	 */
+	struct semaphore	sem;
+};
+
+enum {
+	MLX5_FMR_INVALID,
+	MLX5_FMR_VALID,
+	MLX5_FMR_BUSY,
+};
+
+struct mlx5_ib_fmr {
+	struct ib_fmr			ibfmr;
+	struct mlx5_core_mr		mr;
+	int				access_flags;
+	int				state;
+	/* protect fmr state
+	 */
+	spinlock_t			lock;
+	u64				wrid;
+	struct ib_send_wr		wr[2];
+	u8				page_shift;
+	struct ib_fast_reg_page_list	page_list;
+};
+
+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			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;
+};
+
+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_resources {
+	struct ib_cq	*c0;
+	struct ib_xrcd	*x0;
+	struct ib_xrcd	*x1;
+	struct ib_pd	*p0;
+	struct ib_srq	*s0;
+};
+
+struct mlx5_ib_dev {
+	struct ib_device		ib_dev;
+	struct mlx5_core_dev		mdev;
+	MLX5_DECLARE_DOORBELL_LOCK(uar_lock);
+	struct list_head		eqs_list;
+	int				num_ports;
+	int				num_comp_vectors;
+	/* serialize update of capability mask
+	 */
+	struct mutex			cap_mask_mutex;
+	bool				ib_active;
+	struct umr_common		umrc;
+	/* sync used page count stats
+	 */
+	spinlock_t			mr_lock;
+	struct mlx5_ib_resources	devr;
+	struct mlx5_mr_cache		cache;
+};
+
+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_fmr *to_mfmr(struct ib_fmr *ibfmr)
+{
+	return container_of(ibfmr, struct mlx5_ib_fmr, ibfmr);
+}
+
+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, mqp);
+}
+
+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_srq *to_mibsrq(struct mlx5_core_srq *msrq)
+{
+	return container_of(msrq, struct mlx5_ib_srq, msrq);
+}
+
+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_fast_reg_page_list *to_mfrpl(struct ib_fast_reg_page_list *ibfrpl)
+{
+	return container_of(ibfrpl, struct mlx5_ib_fast_reg_page_list, ibfrpl);
+}
+
+struct mlx5_ib_ah {
+	struct ib_ah		ibah;
+	struct mlx5_av		av;
+};
+
+static inline struct mlx5_ib_ah *to_mah(struct ib_ah *ibah)
+{
+	return container_of(ibah, struct mlx5_ib_ah, ibah);
+}
+
+static inline struct mlx5_ib_dev *mlx5_core2ibdev(struct mlx5_core_dev *dev)
+{
+	return container_of(dev, struct mlx5_ib_dev, mdev);
+}
+
+static inline struct mlx5_ib_dev *mlx5_pci2ibdev(struct pci_dev *pdev)
+{
+	return mlx5_core2ibdev(pci2mlx5_core_dev(pdev));
+}
+
+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,
+		 int port, struct ib_wc *in_wc, struct ib_grh *in_grh,
+		 void *in_mad, void *response_mad);
+struct ib_ah *create_ib_ah(struct ib_ah_attr *ah_attr,
+			   struct mlx5_ib_ah *ah);
+struct ib_ah *mlx5_ib_create_ah(struct ib_pd *pd, struct ib_ah_attr *ah_attr);
+int mlx5_ib_query_ah(struct ib_ah *ibah, struct ib_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, struct ib_recv_wr *wr,
+			  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);
+int mlx5_ib_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
+		      struct ib_send_wr **bad_wr);
+int mlx5_ib_post_recv(struct ib_qp *ibqp, struct ib_recv_wr *wr,
+		      struct ib_recv_wr **bad_wr);
+void *mlx5_get_send_wqe(struct mlx5_ib_qp *qp, int n);
+struct ib_cq *mlx5_ib_create_cq(struct ib_device *ibdev, int entries,
+				int vector, 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);
+int mlx5_ib_dereg_mr(struct ib_mr *ibmr);
+struct ib_mr *mlx5_ib_alloc_fast_reg_mr(struct ib_pd *pd,
+					int max_page_list_len);
+struct ib_fast_reg_page_list *mlx5_ib_alloc_fast_reg_page_list(struct ib_device *ibdev,
+							       int page_list_len);
+void mlx5_ib_free_fast_reg_page_list(struct ib_fast_reg_page_list *page_list);
+struct ib_fmr *mlx5_ib_fmr_alloc(struct ib_pd *pd, int acc,
+				 struct ib_fmr_attr *fmr_attr);
+int mlx5_ib_map_phys_fmr(struct ib_fmr *ibfmr, u64 *page_list,
+		      int npages, u64 iova);
+int mlx5_ib_unmap_fmr(struct list_head *fmr_list);
+int mlx5_ib_fmr_dealloc(struct ib_fmr *ibfmr);
+int mlx5_ib_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num,
+			struct ib_wc *in_wc, struct ib_grh *in_grh,
+			struct ib_mad *in_mad, struct ib_mad *out_mad);
+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_vector2eqn(struct mlx5_ib_dev *dev, int vector, int *eqn, int *irqn);
+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_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, 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, __be64 *pas, int umr);
+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);
+int mlx5_mr_ib_cont_pages(struct ib_umem *umem, u64 addr, int *count, int *shift);
+void mlx5_umr_cq_handler(struct ib_cq *cq, void *cq_context);
+
+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;
+}
+
+#endif /* MLX5_IB_H */

drivers/infiniband/hw/mlx5/srq.c

@@ -0,0 +1,473 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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 "mlx5_ib.h"
+#include "user.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_create_srq_mbox_in **in,
+			   struct ib_udata *udata, int buf_size, int *inlen)
+{
+	struct mlx5_ib_dev *dev = to_mdev(pd->device);
+	struct mlx5_ib_create_srq ucmd;
+	int err;
+	int npages;
+	int page_shift;
+	int ncont;
+	u32 offset;
+
+	if (ib_copy_from_udata(&ucmd, udata, sizeof(ucmd))) {
+		mlx5_ib_dbg(dev, "failed copy udata\n");
+		return -EFAULT;
+	}
+	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, &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;
+	}
+
+	*inlen = sizeof(**in) + sizeof(*(*in)->pas) * ncont;
+	*in = mlx5_vzalloc(*inlen);
+	if (!(*in)) {
+		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)->ctx.log_pg_sz = page_shift - PAGE_SHIFT;
+	(*in)->ctx.pgoff_cqn = cpu_to_be32(offset << 26);
+
+	return 0;
+
+err_in:
+	mlx5_vfree(*in);
+
+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_create_srq_mbox_in **in, int buf_size,
+			     int *inlen)
+{
+	int err;
+	int i;
+	struct mlx5_wqe_srq_next_seg *next;
+	int page_shift;
+	int npages;
+
+	err = mlx5_db_alloc(&dev->mdev, &srq->db);
+	if (err) {
+		mlx5_ib_warn(dev, "alloc dbell rec failed\n");
+		return err;
+	}
+
+	*srq->db.db = 0;
+
+	if (mlx5_buf_alloc(&dev->mdev, buf_size, PAGE_SIZE * 2, &srq->buf)) {
+		mlx5_ib_dbg(dev, "buf alloc failed\n");
+		err = -ENOMEM;
+		goto err_db;
+	}
+	page_shift = srq->buf.page_shift;
+
+	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));
+	}
+
+	npages = DIV_ROUND_UP(srq->buf.npages, 1 << (page_shift - PAGE_SHIFT));
+	mlx5_ib_dbg(dev, "buf_size %d, page_shift %d, npages %d, calc npages %d\n",
+		    buf_size, page_shift, srq->buf.npages, npages);
+	*inlen = sizeof(**in) + sizeof(*(*in)->pas) * npages;
+	*in = mlx5_vzalloc(*inlen);
+	if (!*in) {
+		err = -ENOMEM;
+		goto err_buf;
+	}
+	mlx5_fill_page_array(&srq->buf, (*in)->pas);
+
+	srq->wrid = kmalloc(srq->msrq.max * sizeof(u64), GFP_KERNEL);
+	if (!srq->wrid) {
+		mlx5_ib_dbg(dev, "kmalloc failed %lu\n",
+			    (unsigned long)(srq->msrq.max * sizeof(u64)));
+		err = -ENOMEM;
+		goto err_in;
+	}
+	srq->wq_sig = !!srq_signature;
+
+	(*in)->ctx.log_pg_sz = page_shift - PAGE_SHIFT;
+
+	return 0;
+
+err_in:
+	mlx5_vfree(*in);
+
+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)
+{
+	kfree(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;
+	int desc_size;
+	int buf_size;
+	int err;
+	struct mlx5_create_srq_mbox_in *uninitialized_var(in);
+	int uninitialized_var(inlen);
+	int is_xrc;
+	u32 flgs, xrcdn;
+
+	/* Sanity check SRQ size before proceeding */
+	if (init_attr->attr.max_wr >= dev->mdev.caps.max_srq_wqes) {
+		mlx5_ib_dbg(dev, "max_wr %d, cap %d\n",
+			    init_attr->attr.max_wr,
+			    dev->mdev.caps.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);
+	desc_size = roundup_pow_of_two(desc_size);
+	desc_size = max_t(int, 32, desc_size);
+	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;
+	mlx5_ib_dbg(dev, "desc_size 0x%x, req wr 0x%x, srq size 0x%x, max_gs 0x%x, max_avail_gather 0x%x\n",
+		    desc_size, init_attr->attr.max_wr, srq->msrq.max, srq->msrq.max_gs,
+		    srq->msrq.max_avail_gather);
+
+	if (pd->uobject)
+		err = create_srq_user(pd, srq, &in, udata, buf_size, &inlen);
+	else
+		err = create_srq_kernel(dev, srq, &in, buf_size, &inlen);
+
+	if (err) {
+		mlx5_ib_warn(dev, "create srq %s failed, err %d\n",
+			     pd->uobject ? "user" : "kernel", err);
+		goto err_srq;
+	}
+
+	is_xrc = (init_attr->srq_type == IB_SRQT_XRC);
+	in->ctx.state_log_sz = ilog2(srq->msrq.max);
+	flgs = ((srq->msrq.wqe_shift - 4) | (is_xrc << 5) | (srq->wq_sig << 7)) << 24;
+	xrcdn = 0;
+	if (is_xrc) {
+		xrcdn = to_mxrcd(init_attr->ext.xrc.xrcd)->xrcdn;
+		in->ctx.pgoff_cqn |= cpu_to_be32(to_mcq(init_attr->ext.xrc.cq)->mcq.cqn);
+	} else if (init_attr->srq_type == IB_SRQT_BASIC) {
+		xrcdn = to_mxrcd(dev->devr.x0)->xrcdn;
+		in->ctx.pgoff_cqn |= cpu_to_be32(to_mcq(dev->devr.c0)->mcq.cqn);
+	}
+
+	in->ctx.flags_xrcd = cpu_to_be32((flgs & 0xFF000000) | (xrcdn & 0xFFFFFF));
+
+	in->ctx.pd = cpu_to_be32(to_mpd(pd)->pdn);
+	in->ctx.db_record = cpu_to_be64(srq->db.dma);
+	err = mlx5_core_create_srq(&dev->mdev, &srq->msrq, in, inlen);
+	mlx5_vfree(in);
+	if (err) {
+		mlx5_ib_dbg(dev, "create SRQ failed, err %d\n", err);
+		goto err_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);
+	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_query_srq_mbox_out *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 = be16_to_cpu(out->ctx.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 {
+		kfree(msrq->wrid);
+		mlx5_buf_free(&dev->mdev, &msrq->buf);
+		mlx5_db_free(&dev->mdev, &msrq->db);
+	}
+
+	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, struct ib_recv_wr *wr,
+			  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;
+	unsigned long flags;
+	int err = 0;
+	int nreq;
+	int i;
+
+	spin_lock_irqsave(&srq->lock, flags);
+
+	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);
+	}
+
+	spin_unlock_irqrestore(&srq->lock, flags);
+
+	return err;
+}

drivers/infiniband/hw/mlx5/user.h

@@ -0,0 +1,121 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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_USER_H
+#define MLX5_IB_USER_H
+
+#include <linux/types.h>
+
+enum {
+	MLX5_QP_FLAG_SIGNATURE		= 1 << 0,
+	MLX5_QP_FLAG_SCATTER_CQE	= 1 << 1,
+};
+
+enum {
+	MLX5_SRQ_FLAG_SIGNATURE		= 1 << 0,
+};
+
+
+/* Increment this value if any changes that break userspace ABI
+ * compatibility are made.
+ */
+#define MLX5_IB_UVERBS_ABI_VERSION	1
+
+/* Make sure that all structs defined in this file remain laid out so
+ * that they pack the same way on 32-bit and 64-bit architectures (to
+ * avoid incompatibility between 32-bit userspace and 64-bit kernels).
+ * In particular do not use pointer types -- pass pointers in __u64
+ * instead.
+ */
+
+struct mlx5_ib_alloc_ucontext_req {
+	__u32	total_num_uuars;
+	__u32	num_low_latency_uuars;
+};
+
+struct mlx5_ib_alloc_ucontext_resp {
+	__u32	qp_tab_size;
+	__u32	bf_reg_size;
+	__u32	tot_uuars;
+	__u32	cache_line_size;
+	__u16	max_sq_desc_sz;
+	__u16	max_rq_desc_sz;
+	__u32	max_send_wqebb;
+	__u32	max_recv_wr;
+	__u32	max_srq_recv_wr;
+	__u16	num_ports;
+	__u16	reserved;
+};
+
+struct mlx5_ib_alloc_pd_resp {
+	__u32	pdn;
+};
+
+struct mlx5_ib_create_cq {
+	__u64	buf_addr;
+	__u64	db_addr;
+	__u32	cqe_size;
+};
+
+struct mlx5_ib_create_cq_resp {
+	__u32	cqn;
+	__u32	reserved;
+};
+
+struct mlx5_ib_resize_cq {
+	__u64	buf_addr;
+};
+
+struct mlx5_ib_create_srq {
+	__u64	buf_addr;
+	__u64	db_addr;
+	__u32	flags;
+};
+
+struct mlx5_ib_create_srq_resp {
+	__u32	srqn;
+	__u32	reserved;
+};
+
+struct mlx5_ib_create_qp {
+	__u64	buf_addr;
+	__u64	db_addr;
+	__u32	sq_wqe_count;
+	__u32	rq_wqe_count;
+	__u32	rq_wqe_shift;
+	__u32	flags;
+};
+
+struct mlx5_ib_create_qp_resp {
+	__u32	uuar_index;
+};
+#endif /* MLX5_IB_USER_H */

drivers/infiniband/hw/ocrdma/ocrdma.h

@@ -42,8 +42,6 @@
 #define OCRDMA_ROCE_DEV_VERSION "1.0.0"
 #define OCRDMA_NODE_DESC "Emulex OneConnect RoCE HCA"
 
-#define ocrdma_err(format, arg...) printk(KERN_ERR format, ##arg)
-
 #define OCRDMA_MAX_AH 512
 
 #define OCRDMA_UVERBS(CMD_NAME) (1ull << IB_USER_VERBS_CMD_##CMD_NAME)
@@ -97,7 +95,6 @@ struct ocrdma_queue_info {
 	u16 id;			/* qid, where to ring the doorbell. */
 	u16 head, tail;
 	bool created;
-	atomic_t used;		/* Number of valid elements in the queue */
 };
 
 struct ocrdma_eq {
@@ -198,7 +195,6 @@ struct ocrdma_cq {
 	struct ocrdma_ucontext *ucontext;
 	dma_addr_t pa;
 	u32 len;
-	atomic_t use_cnt;
 
 	/* head of all qp's sq and rq for which cqes need to be flushed
 	 * by the software.
@@ -210,7 +206,6 @@ struct ocrdma_pd {
 	struct ib_pd ibpd;
 	struct ocrdma_dev *dev;
 	struct ocrdma_ucontext *uctx;
-	atomic_t use_cnt;
 	u32 id;
 	int num_dpp_qp;
 	u32 dpp_page;
@@ -241,16 +236,16 @@ struct ocrdma_srq {
 	struct ib_srq ibsrq;
 	struct ocrdma_dev *dev;
 	u8 __iomem *db;
-	/* provide synchronization to multiple context(s) posting rqe */
-	spinlock_t q_lock ____cacheline_aligned;
-
 	struct ocrdma_qp_hwq_info rq;
-	struct ocrdma_pd *pd;
-	atomic_t use_cnt;
-	u32 id;
 	u64 *rqe_wr_id_tbl;
 	u32 *idx_bit_fields;
 	u32 bit_fields_len;
+
+	/* provide synchronization to multiple context(s) posting rqe */
+	spinlock_t q_lock ____cacheline_aligned;
+
+	struct ocrdma_pd *pd;
+	u32 id;
 };
 
 struct ocrdma_qp {
@@ -258,8 +253,6 @@ struct ocrdma_qp {
 	struct ocrdma_dev *dev;
 
 	u8 __iomem *sq_db;
-	/* provide synchronization to multiple context(s) posting wqe, rqe */
-	spinlock_t q_lock ____cacheline_aligned;
 	struct ocrdma_qp_hwq_info sq;
 	struct {
 		uint64_t wrid;
@@ -269,6 +262,9 @@ struct ocrdma_qp {
 		uint8_t  rsvd[3];
 	} *wqe_wr_id_tbl;
 	u32 max_inline_data;
+
+	/* provide synchronization to multiple context(s) posting wqe, rqe */
+	spinlock_t q_lock ____cacheline_aligned;
 	struct ocrdma_cq *sq_cq;
 	/* list maintained per CQ to flush SQ errors */
 	struct list_head sq_entry;
@@ -296,10 +292,6 @@ struct ocrdma_qp {
 	u8 *ird_q_va;
 };
 
-#define OCRDMA_GET_NUM_POSTED_SHIFT_VAL(qp) \
-	(((qp->dev->nic_info.dev_family == OCRDMA_GEN2_FAMILY) && \
-		(qp->id < 64)) ? 24 : 16)
-
 struct ocrdma_hw_mr {
 	struct ocrdma_dev *dev;
 	u32 lkey;
@@ -390,4 +382,43 @@ static inline struct ocrdma_srq *get_ocrdma_srq(struct ib_srq *ibsrq)
 	return container_of(ibsrq, struct ocrdma_srq, ibsrq);
 }
 
+
+static inline int ocrdma_get_num_posted_shift(struct ocrdma_qp *qp)
+{
+	return ((qp->dev->nic_info.dev_family == OCRDMA_GEN2_FAMILY &&
+		 qp->id < 64) ? 24 : 16);
+}
+
+static inline int is_cqe_valid(struct ocrdma_cq *cq, struct ocrdma_cqe *cqe)
+{
+	int cqe_valid;
+	cqe_valid = le32_to_cpu(cqe->flags_status_srcqpn) & OCRDMA_CQE_VALID;
+	return ((cqe_valid == cq->phase) ? 1 : 0);
+}
+
+static inline int is_cqe_for_sq(struct ocrdma_cqe *cqe)
+{
+	return (le32_to_cpu(cqe->flags_status_srcqpn) &
+		OCRDMA_CQE_QTYPE) ? 0 : 1;
+}
+
+static inline int is_cqe_invalidated(struct ocrdma_cqe *cqe)
+{
+	return (le32_to_cpu(cqe->flags_status_srcqpn) &
+		OCRDMA_CQE_INVALIDATE) ? 1 : 0;
+}
+
+static inline int is_cqe_imm(struct ocrdma_cqe *cqe)
+{
+	return (le32_to_cpu(cqe->flags_status_srcqpn) &
+		OCRDMA_CQE_IMM) ? 1 : 0;
+}
+
+static inline int is_cqe_wr_imm(struct ocrdma_cqe *cqe)
+{
+	return (le32_to_cpu(cqe->flags_status_srcqpn) &
+		OCRDMA_CQE_WRITE_IMM) ? 1 : 0;
+}
+
+
 #endif

drivers/infiniband/hw/ocrdma/ocrdma_hw.c

@@ -128,7 +128,6 @@ static inline struct ocrdma_mqe *ocrdma_get_mqe(struct ocrdma_dev *dev)
 static inline void ocrdma_mq_inc_head(struct ocrdma_dev *dev)
 {
 	dev->mq.sq.head = (dev->mq.sq.head + 1) & (OCRDMA_MQ_LEN - 1);
-	atomic_inc(&dev->mq.sq.used);
 }
 
 static inline void *ocrdma_get_mqe_rsp(struct ocrdma_dev *dev)
@@ -564,32 +563,19 @@ static int ocrdma_mbx_create_mq(struct ocrdma_dev *dev,
 	memset(cmd, 0, sizeof(*cmd));
 	num_pages = PAGES_4K_SPANNED(mq->va, mq->size);
 
-	if (dev->nic_info.dev_family == OCRDMA_GEN2_FAMILY) {
-		ocrdma_init_mch(&cmd->req, OCRDMA_CMD_CREATE_MQ,
-				OCRDMA_SUBSYS_COMMON, sizeof(*cmd));
-		cmd->v0.pages = num_pages;
-		cmd->v0.async_cqid_valid = OCRDMA_CREATE_MQ_ASYNC_CQ_VALID;
-		cmd->v0.async_cqid_valid = (cq->id << 1);
-		cmd->v0.cqid_ringsize |= (ocrdma_encoded_q_len(mq->len) <<
-					     OCRDMA_CREATE_MQ_RING_SIZE_SHIFT);
-		cmd->v0.cqid_ringsize |=
-			(cq->id << OCRDMA_CREATE_MQ_V0_CQ_ID_SHIFT);
-		cmd->v0.valid = OCRDMA_CREATE_MQ_VALID;
-		pa = &cmd->v0.pa[0];
-	} else {
-		ocrdma_init_mch(&cmd->req, OCRDMA_CMD_CREATE_MQ_EXT,
-				OCRDMA_SUBSYS_COMMON, sizeof(*cmd));
-		cmd->req.rsvd_version = 1;
-		cmd->v1.cqid_pages = num_pages;
-		cmd->v1.cqid_pages |= (cq->id << OCRDMA_CREATE_MQ_CQ_ID_SHIFT);
-		cmd->v1.async_cqid_valid = OCRDMA_CREATE_MQ_ASYNC_CQ_VALID;
-		cmd->v1.async_event_bitmap = Bit(20);
-		cmd->v1.async_cqid_ringsize = cq->id;
-		cmd->v1.async_cqid_ringsize |= (ocrdma_encoded_q_len(mq->len) <<
-					     OCRDMA_CREATE_MQ_RING_SIZE_SHIFT);
-		cmd->v1.valid = OCRDMA_CREATE_MQ_VALID;
-		pa = &cmd->v1.pa[0];
-	}
+	ocrdma_init_mch(&cmd->req, OCRDMA_CMD_CREATE_MQ_EXT,
+			OCRDMA_SUBSYS_COMMON, sizeof(*cmd));
+	cmd->req.rsvd_version = 1;
+	cmd->cqid_pages = num_pages;
+	cmd->cqid_pages |= (cq->id << OCRDMA_CREATE_MQ_CQ_ID_SHIFT);
+	cmd->async_cqid_valid = OCRDMA_CREATE_MQ_ASYNC_CQ_VALID;
+	cmd->async_event_bitmap = Bit(20);
+	cmd->async_cqid_ringsize = cq->id;
+	cmd->async_cqid_ringsize |= (ocrdma_encoded_q_len(mq->len) <<
+				OCRDMA_CREATE_MQ_RING_SIZE_SHIFT);
+	cmd->valid = OCRDMA_CREATE_MQ_VALID;
+	pa = &cmd->pa[0];
+
 	ocrdma_build_q_pages(pa, num_pages, mq->dma, PAGE_SIZE_4K);
 	status = be_roce_mcc_cmd(dev->nic_info.netdev,
 				 cmd, sizeof(*cmd), NULL, NULL);
@@ -745,7 +731,7 @@ static void ocrdma_dispatch_ibevent(struct ocrdma_dev *dev,
 		qp_event = 0;
 		srq_event = 0;
 		dev_event = 0;
-		ocrdma_err("%s() unknown type=0x%x\n", __func__, type);
+		pr_err("%s() unknown type=0x%x\n", __func__, type);
 		break;
 	}
 
@@ -775,8 +761,8 @@ static void ocrdma_process_acqe(struct ocrdma_dev *dev, void *ae_cqe)
 	if (evt_code == OCRDMA_ASYNC_EVE_CODE)
 		ocrdma_dispatch_ibevent(dev, cqe);
 	else
-		ocrdma_err("%s(%d) invalid evt code=0x%x\n",
-			   __func__, dev->id, evt_code);
+		pr_err("%s(%d) invalid evt code=0x%x\n", __func__,
+		       dev->id, evt_code);
 }
 
 static void ocrdma_process_mcqe(struct ocrdma_dev *dev, struct ocrdma_mcqe *cqe)
@@ -790,8 +776,8 @@ static void ocrdma_process_mcqe(struct ocrdma_dev *dev, struct ocrdma_mcqe *cqe)
 		dev->mqe_ctx.cmd_done = true;
 		wake_up(&dev->mqe_ctx.cmd_wait);
 	} else
-		ocrdma_err("%s() cqe for invalid tag0x%x.expected=0x%x\n",
-			   __func__, cqe->tag_lo, dev->mqe_ctx.tag);
+		pr_err("%s() cqe for invalid tag0x%x.expected=0x%x\n",
+		       __func__, cqe->tag_lo, dev->mqe_ctx.tag);
 }
 
 static int ocrdma_mq_cq_handler(struct ocrdma_dev *dev, u16 cq_id)
@@ -810,7 +796,7 @@ static int ocrdma_mq_cq_handler(struct ocrdma_dev *dev, u16 cq_id)
 		else if (cqe->valid_ae_cmpl_cons & OCRDMA_MCQE_CMPL_MASK)
 			ocrdma_process_mcqe(dev, cqe);
 		else
-			ocrdma_err("%s() cqe->compl is not set.\n", __func__);
+			pr_err("%s() cqe->compl is not set.\n", __func__);
 		memset(cqe, 0, sizeof(struct ocrdma_mcqe));
 		ocrdma_mcq_inc_tail(dev);
 	}
@@ -869,7 +855,7 @@ static void ocrdma_qp_cq_handler(struct ocrdma_dev *dev, u16 cq_idx)
 
 	cq = dev->cq_tbl[cq_idx];
 	if (cq == NULL) {
-		ocrdma_err("%s%d invalid id=0x%x\n", __func__, dev->id, cq_idx);
+		pr_err("%s%d invalid id=0x%x\n", __func__, dev->id, cq_idx);
 		return;
 	}
 	spin_lock_irqsave(&cq->cq_lock, flags);
@@ -971,7 +957,7 @@ static int ocrdma_mbx_cmd(struct ocrdma_dev *dev, struct ocrdma_mqe *mqe)
 	rsp = ocrdma_get_mqe_rsp(dev);
 	ocrdma_copy_le32_to_cpu(mqe, rsp, (sizeof(*mqe)));
 	if (cqe_status || ext_status) {
-		ocrdma_err
+		pr_err
 		    ("%s() opcode=0x%x, cqe_status=0x%x, ext_status=0x%x\n",
 		     __func__,
 		     (rsp->u.rsp.subsys_op & OCRDMA_MBX_RSP_OPCODE_MASK) >>
@@ -1353,8 +1339,8 @@ int ocrdma_mbx_create_cq(struct ocrdma_dev *dev, struct ocrdma_cq *cq,
 	if (dpp_cq)
 		return -EINVAL;
 	if (entries > dev->attr.max_cqe) {
-		ocrdma_err("%s(%d) max_cqe=0x%x, requester_cqe=0x%x\n",
-			   __func__, dev->id, dev->attr.max_cqe, entries);
+		pr_err("%s(%d) max_cqe=0x%x, requester_cqe=0x%x\n",
+		       __func__, dev->id, dev->attr.max_cqe, entries);
 		return -EINVAL;
 	}
 	if (dpp_cq && (dev->nic_info.dev_family != OCRDMA_GEN2_FAMILY))
@@ -1621,7 +1607,7 @@ int ocrdma_reg_mr(struct ocrdma_dev *dev,
 	status = ocrdma_mbx_reg_mr(dev, hwmr, pdid,
 				   cur_pbl_cnt, hwmr->pbe_size, last);
 	if (status) {
-		ocrdma_err("%s() status=%d\n", __func__, status);
+		pr_err("%s() status=%d\n", __func__, status);
 		return status;
 	}
 	/* if there is no more pbls to register then exit. */
@@ -1644,7 +1630,7 @@ int ocrdma_reg_mr(struct ocrdma_dev *dev,
 			break;
 	}
 	if (status)
-		ocrdma_err("%s() err. status=%d\n", __func__, status);
+		pr_err("%s() err. status=%d\n", __func__, status);
 
 	return status;
 }
@@ -1841,8 +1827,8 @@ static int ocrdma_set_create_qp_sq_cmd(struct ocrdma_create_qp_req *cmd,
 	status = ocrdma_build_q_conf(&max_wqe_allocated,
 		dev->attr.wqe_size, &hw_pages, &hw_page_size);
 	if (status) {
-		ocrdma_err("%s() req. max_send_wr=0x%x\n", __func__,
-			   max_wqe_allocated);
+		pr_err("%s() req. max_send_wr=0x%x\n", __func__,
+		       max_wqe_allocated);
 		return -EINVAL;
 	}
 	qp->sq.max_cnt = max_wqe_allocated;
@@ -1891,8 +1877,8 @@ static int ocrdma_set_create_qp_rq_cmd(struct ocrdma_create_qp_req *cmd,
 	status = ocrdma_build_q_conf(&max_rqe_allocated, dev->attr.rqe_size,
 				     &hw_pages, &hw_page_size);
 	if (status) {
-		ocrdma_err("%s() req. max_recv_wr=0x%x\n", __func__,
-			   attrs->cap.max_recv_wr + 1);
+		pr_err("%s() req. max_recv_wr=0x%x\n", __func__,
+		       attrs->cap.max_recv_wr + 1);
 		return status;
 	}
 	qp->rq.max_cnt = max_rqe_allocated;
@@ -1900,7 +1886,7 @@ static int ocrdma_set_create_qp_rq_cmd(struct ocrdma_create_qp_req *cmd,
 
 	qp->rq.va = dma_alloc_coherent(&pdev->dev, len, &pa, GFP_KERNEL);
 	if (!qp->rq.va)
-		return status;
+		return -ENOMEM;
 	memset(qp->rq.va, 0, len);
 	qp->rq.pa = pa;
 	qp->rq.len = len;
@@ -2087,10 +2073,10 @@ mbx_err:
 	if (qp->rq.va)
 		dma_free_coherent(&pdev->dev, qp->rq.len, qp->rq.va, qp->rq.pa);
 rq_err:
-	ocrdma_err("%s(%d) rq_err\n", __func__, dev->id);
+	pr_err("%s(%d) rq_err\n", __func__, dev->id);
 	dma_free_coherent(&pdev->dev, qp->sq.len, qp->sq.va, qp->sq.pa);
 sq_err:
-	ocrdma_err("%s(%d) sq_err\n", __func__, dev->id);
+	pr_err("%s(%d) sq_err\n", __func__, dev->id);
 	kfree(cmd);
 	return status;
 }
@@ -2127,7 +2113,7 @@ int ocrdma_resolve_dgid(struct ocrdma_dev *dev, union ib_gid *dgid,
 	else if (rdma_link_local_addr(&in6))
 		rdma_get_ll_mac(&in6, mac_addr);
 	else {
-		ocrdma_err("%s() fail to resolve mac_addr.\n", __func__);
+		pr_err("%s() fail to resolve mac_addr.\n", __func__);
 		return -EINVAL;
 	}
 	return 0;
@@ -2362,8 +2348,8 @@ int ocrdma_mbx_create_srq(struct ocrdma_srq *srq,
 				dev->attr.rqe_size,
 				&hw_pages, &hw_page_size);
 	if (status) {
-		ocrdma_err("%s() req. max_wr=0x%x\n", __func__,
-			   srq_attr->attr.max_wr);
+		pr_err("%s() req. max_wr=0x%x\n", __func__,
+		       srq_attr->attr.max_wr);
 		status = -EINVAL;
 		goto ret;
 	}
@@ -2614,7 +2600,7 @@ mq_err:
 	ocrdma_destroy_qp_eqs(dev);
 qpeq_err:
 	ocrdma_destroy_eq(dev, &dev->meq);
-	ocrdma_err("%s() status=%d\n", __func__, status);
+	pr_err("%s() status=%d\n", __func__, status);
 	return status;
 }
 

drivers/infiniband/hw/ocrdma/ocrdma_main.c

@@ -378,7 +378,7 @@ static int ocrdma_alloc_resources(struct ocrdma_dev *dev)
 	spin_lock_init(&dev->flush_q_lock);
 	return 0;
 alloc_err:
-	ocrdma_err("%s(%d) error.\n", __func__, dev->id);
+	pr_err("%s(%d) error.\n", __func__, dev->id);
 	return -ENOMEM;
 }
 
@@ -396,7 +396,7 @@ static struct ocrdma_dev *ocrdma_add(struct be_dev_info *dev_info)
 
 	dev = (struct ocrdma_dev *)ib_alloc_device(sizeof(struct ocrdma_dev));
 	if (!dev) {
-		ocrdma_err("Unable to allocate ib device\n");
+		pr_err("Unable to allocate ib device\n");
 		return NULL;
 	}
 	dev->mbx_cmd = kzalloc(sizeof(struct ocrdma_mqe_emb_cmd), GFP_KERNEL);
@@ -437,7 +437,7 @@ init_err:
 idr_err:
 	kfree(dev->mbx_cmd);
 	ib_dealloc_device(&dev->ibdev);
-	ocrdma_err("%s() leaving. ret=%d\n", __func__, status);
+	pr_err("%s() leaving. ret=%d\n", __func__, status);
 	return NULL;
 }
 

drivers/infiniband/hw/ocrdma/ocrdma_sli.h

@@ -608,16 +608,8 @@ enum {
 	OCRDMA_CREATE_MQ_ASYNC_CQ_VALID		= Bit(0)
 };
 
-struct ocrdma_create_mq_v0 {
-	u32 pages;
-	u32 cqid_ringsize;
-	u32 valid;
-	u32 async_cqid_valid;
-	u32 rsvd;
-	struct ocrdma_pa pa[8];
-} __packed;
-
-struct ocrdma_create_mq_v1 {
+struct ocrdma_create_mq_req {
+	struct ocrdma_mbx_hdr req;
 	u32 cqid_pages;
 	u32 async_event_bitmap;
 	u32 async_cqid_ringsize;
@@ -627,14 +619,6 @@ struct ocrdma_create_mq_v1 {
 	struct ocrdma_pa pa[8];
 } __packed;
 
-struct ocrdma_create_mq_req {
-	struct ocrdma_mbx_hdr req;
-	union {
-		struct ocrdma_create_mq_v0 v0;
-		struct ocrdma_create_mq_v1 v1;
-	};
-} __packed;
-
 struct ocrdma_create_mq_rsp {
 	struct ocrdma_mbx_rsp rsp;
 	u32 id;
@@ -1550,21 +1534,6 @@ struct ocrdma_cqe {
 	u32 flags_status_srcqpn;	/* w3 */
 } __packed;
 
-#define is_cqe_valid(cq, cqe) \
-	(((le32_to_cpu(cqe->flags_status_srcqpn) & OCRDMA_CQE_VALID)\
-	== cq->phase) ? 1 : 0)
-#define is_cqe_for_sq(cqe) \
-	((le32_to_cpu(cqe->flags_status_srcqpn) & OCRDMA_CQE_QTYPE) ? 0 : 1)
-#define is_cqe_for_rq(cqe) \
-	((le32_to_cpu(cqe->flags_status_srcqpn) & OCRDMA_CQE_QTYPE) ? 1 : 0)
-#define is_cqe_invalidated(cqe) \
-	((le32_to_cpu(cqe->flags_status_srcqpn) & OCRDMA_CQE_INVALIDATE) ? \
-	1 : 0)
-#define is_cqe_imm(cqe) \
-	((le32_to_cpu(cqe->flags_status_srcqpn) & OCRDMA_CQE_IMM) ? 1 : 0)
-#define is_cqe_wr_imm(cqe) \
-	((le32_to_cpu(cqe->flags_status_srcqpn) & OCRDMA_CQE_WRITE_IMM) ? 1 : 0)
-
 struct ocrdma_sge {
 	u32 addr_hi;
 	u32 addr_lo;

drivers/infiniband/hw/ocrdma/ocrdma_verbs.c

@@ -114,8 +114,8 @@ int ocrdma_query_port(struct ib_device *ibdev,
 
 	dev = get_ocrdma_dev(ibdev);
 	if (port > 1) {
-		ocrdma_err("%s(%d) invalid_port=0x%x\n", __func__,
-			   dev->id, port);
+		pr_err("%s(%d) invalid_port=0x%x\n", __func__,
+		       dev->id, port);
 		return -EINVAL;
 	}
 	netdev = dev->nic_info.netdev;
@@ -155,8 +155,7 @@ int ocrdma_modify_port(struct ib_device *ibdev, u8 port, int mask,
 
 	dev = get_ocrdma_dev(ibdev);
 	if (port > 1) {
-		ocrdma_err("%s(%d) invalid_port=0x%x\n", __func__,
-			   dev->id, port);
+		pr_err("%s(%d) invalid_port=0x%x\n", __func__, dev->id, port);
 		return -EINVAL;
 	}
 	return 0;
@@ -398,7 +397,6 @@ struct ib_pd *ocrdma_alloc_pd(struct ib_device *ibdev,
 		kfree(pd);
 		return ERR_PTR(status);
 	}
-	atomic_set(&pd->use_cnt, 0);
 
 	if (udata && context) {
 		status = ocrdma_copy_pd_uresp(pd, context, udata);
@@ -419,12 +417,6 @@ int ocrdma_dealloc_pd(struct ib_pd *ibpd)
 	int status;
 	u64 usr_db;
 
-	if (atomic_read(&pd->use_cnt)) {
-		ocrdma_err("%s(%d) pd=0x%x is in use.\n",
-			   __func__, dev->id, pd->id);
-		status = -EFAULT;
-		goto dealloc_err;
-	}
 	status = ocrdma_mbx_dealloc_pd(dev, pd);
 	if (pd->uctx) {
 		u64 dpp_db = dev->nic_info.dpp_unmapped_addr +
@@ -436,7 +428,6 @@ int ocrdma_dealloc_pd(struct ib_pd *ibpd)
 		ocrdma_del_mmap(pd->uctx, usr_db, dev->nic_info.db_page_size);
 	}
 	kfree(pd);
-dealloc_err:
 	return status;
 }
 
@@ -450,8 +441,8 @@ static struct ocrdma_mr *ocrdma_alloc_lkey(struct ib_pd *ibpd,
 	struct ocrdma_dev *dev = pd->dev;
 
 	if (acc & IB_ACCESS_REMOTE_WRITE && !(acc & IB_ACCESS_LOCAL_WRITE)) {
-		ocrdma_err("%s(%d) leaving err, invalid access rights\n",
-			   __func__, dev->id);
+		pr_err("%s(%d) leaving err, invalid access rights\n",
+		       __func__, dev->id);
 		return ERR_PTR(-EINVAL);
 	}
 
@@ -474,7 +465,6 @@ static struct ocrdma_mr *ocrdma_alloc_lkey(struct ib_pd *ibpd,
 		return ERR_PTR(-ENOMEM);
 	}
 	mr->pd = pd;
-	atomic_inc(&pd->use_cnt);
 	mr->ibmr.lkey = mr->hwmr.lkey;
 	if (mr->hwmr.remote_wr || mr->hwmr.remote_rd)
 		mr->ibmr.rkey = mr->hwmr.lkey;
@@ -664,7 +654,6 @@ struct ib_mr *ocrdma_reg_user_mr(struct ib_pd *ibpd, u64 start, u64 len,
 	if (status)
 		goto mbx_err;
 	mr->pd = pd;
-	atomic_inc(&pd->use_cnt);
 	mr->ibmr.lkey = mr->hwmr.lkey;
 	if (mr->hwmr.remote_wr || mr->hwmr.remote_rd)
 		mr->ibmr.rkey = mr->hwmr.lkey;
@@ -689,7 +678,6 @@ int ocrdma_dereg_mr(struct ib_mr *ib_mr)
 	if (mr->hwmr.fr_mr == 0)
 		ocrdma_free_mr_pbl_tbl(dev, &mr->hwmr);
 
-	atomic_dec(&mr->pd->use_cnt);
 	/* it could be user registered memory. */
 	if (mr->umem)
 		ib_umem_release(mr->umem);
@@ -714,8 +702,8 @@ static int ocrdma_copy_cq_uresp(struct ocrdma_cq *cq, struct ib_udata *udata,
 	uresp.phase_change = cq->phase_change ? 1 : 0;
 	status = ib_copy_to_udata(udata, &uresp, sizeof(uresp));
 	if (status) {
-		ocrdma_err("%s(%d) copy error cqid=0x%x.\n",
-			   __func__, cq->dev->id, cq->id);
+		pr_err("%s(%d) copy error cqid=0x%x.\n",
+		       __func__, cq->dev->id, cq->id);
 		goto err;
 	}
 	uctx = get_ocrdma_ucontext(ib_ctx);
@@ -752,7 +740,6 @@ struct ib_cq *ocrdma_create_cq(struct ib_device *ibdev, int entries, int vector,
 
 	spin_lock_init(&cq->cq_lock);
 	spin_lock_init(&cq->comp_handler_lock);
-	atomic_set(&cq->use_cnt, 0);
 	INIT_LIST_HEAD(&cq->sq_head);
 	INIT_LIST_HEAD(&cq->rq_head);
 	cq->dev = dev;
@@ -799,9 +786,6 @@ int ocrdma_destroy_cq(struct ib_cq *ibcq)
 	struct ocrdma_cq *cq = get_ocrdma_cq(ibcq);
 	struct ocrdma_dev *dev = cq->dev;
 
-	if (atomic_read(&cq->use_cnt))
-		return -EINVAL;
-
 	status = ocrdma_mbx_destroy_cq(dev, cq);
 
 	if (cq->ucontext) {
@@ -837,57 +821,56 @@ static int ocrdma_check_qp_params(struct ib_pd *ibpd, struct ocrdma_dev *dev,
 	if (attrs->qp_type != IB_QPT_GSI &&
 	    attrs->qp_type != IB_QPT_RC &&
 	    attrs->qp_type != IB_QPT_UD) {
-		ocrdma_err("%s(%d) unsupported qp type=0x%x requested\n",
-			   __func__, dev->id, attrs->qp_type);
+		pr_err("%s(%d) unsupported qp type=0x%x requested\n",
+		       __func__, dev->id, attrs->qp_type);
 		return -EINVAL;
 	}
 	if (attrs->cap.max_send_wr > dev->attr.max_wqe) {
-		ocrdma_err("%s(%d) unsupported send_wr=0x%x requested\n",
-			   __func__, dev->id, attrs->cap.max_send_wr);
-		ocrdma_err("%s(%d) supported send_wr=0x%x\n",
-			   __func__, dev->id, dev->attr.max_wqe);
+		pr_err("%s(%d) unsupported send_wr=0x%x requested\n",
+		       __func__, dev->id, attrs->cap.max_send_wr);
+		pr_err("%s(%d) supported send_wr=0x%x\n",
+		       __func__, dev->id, dev->attr.max_wqe);
 		return -EINVAL;
 	}
 	if (!attrs->srq && (attrs->cap.max_recv_wr > dev->attr.max_rqe)) {
-		ocrdma_err("%s(%d) unsupported recv_wr=0x%x requested\n",
-			   __func__, dev->id, attrs->cap.max_recv_wr);
-		ocrdma_err("%s(%d) supported recv_wr=0x%x\n",
-			   __func__, dev->id, dev->attr.max_rqe);
+		pr_err("%s(%d) unsupported recv_wr=0x%x requested\n",
+		       __func__, dev->id, attrs->cap.max_recv_wr);
+		pr_err("%s(%d) supported recv_wr=0x%x\n",
+		       __func__, dev->id, dev->attr.max_rqe);
 		return -EINVAL;
 	}
 	if (attrs->cap.max_inline_data > dev->attr.max_inline_data) {
-		ocrdma_err("%s(%d) unsupported inline data size=0x%x"
-			   " requested\n", __func__, dev->id,
-			   attrs->cap.max_inline_data);
-		ocrdma_err("%s(%d) supported inline data size=0x%x\n",
-			   __func__, dev->id, dev->attr.max_inline_data);
+		pr_err("%s(%d) unsupported inline data size=0x%x requested\n",
+		       __func__, dev->id, attrs->cap.max_inline_data);
+		pr_err("%s(%d) supported inline data size=0x%x\n",
+		       __func__, dev->id, dev->attr.max_inline_data);
 		return -EINVAL;
 	}
 	if (attrs->cap.max_send_sge > dev->attr.max_send_sge) {
-		ocrdma_err("%s(%d) unsupported send_sge=0x%x requested\n",
-			   __func__, dev->id, attrs->cap.max_send_sge);
-		ocrdma_err("%s(%d) supported send_sge=0x%x\n",
-			   __func__, dev->id, dev->attr.max_send_sge);
+		pr_err("%s(%d) unsupported send_sge=0x%x requested\n",
+		       __func__, dev->id, attrs->cap.max_send_sge);
+		pr_err("%s(%d) supported send_sge=0x%x\n",
+		       __func__, dev->id, dev->attr.max_send_sge);
 		return -EINVAL;
 	}
 	if (attrs->cap.max_recv_sge > dev->attr.max_recv_sge) {
-		ocrdma_err("%s(%d) unsupported recv_sge=0x%x requested\n",
-			   __func__, dev->id, attrs->cap.max_recv_sge);
-		ocrdma_err("%s(%d) supported recv_sge=0x%x\n",
-			   __func__, dev->id, dev->attr.max_recv_sge);
+		pr_err("%s(%d) unsupported recv_sge=0x%x requested\n",
+		       __func__, dev->id, attrs->cap.max_recv_sge);
+		pr_err("%s(%d) supported recv_sge=0x%x\n",
+		       __func__, dev->id, dev->attr.max_recv_sge);
 		return -EINVAL;
 	}
 	/* unprivileged user space cannot create special QP */
 	if (ibpd->uobject && attrs->qp_type == IB_QPT_GSI) {
-		ocrdma_err
+		pr_err
 		    ("%s(%d) Userspace can't create special QPs of type=0x%x\n",
 		     __func__, dev->id, attrs->qp_type);
 		return -EINVAL;
 	}
 	/* allow creating only one GSI type of QP */
 	if (attrs->qp_type == IB_QPT_GSI && dev->gsi_qp_created) {
-		ocrdma_err("%s(%d) GSI special QPs already created.\n",
-			   __func__, dev->id);
+		pr_err("%s(%d) GSI special QPs already created.\n",
+		       __func__, dev->id);
 		return -EINVAL;
 	}
 	/* verify consumer QPs are not trying to use GSI QP's CQ */
@@ -896,8 +879,8 @@ static int ocrdma_check_qp_params(struct ib_pd *ibpd, struct ocrdma_dev *dev,
 		    (dev->gsi_sqcq == get_ocrdma_cq(attrs->recv_cq)) ||
 		    (dev->gsi_rqcq == get_ocrdma_cq(attrs->send_cq)) ||
 		    (dev->gsi_rqcq == get_ocrdma_cq(attrs->recv_cq))) {
-			ocrdma_err("%s(%d) Consumer QP cannot use GSI CQs.\n",
-				   __func__, dev->id);
+			pr_err("%s(%d) Consumer QP cannot use GSI CQs.\n",
+			       __func__, dev->id);
 			return -EINVAL;
 		}
 	}
@@ -949,7 +932,7 @@ static int ocrdma_copy_qp_uresp(struct ocrdma_qp *qp,
 	}
 	status = ib_copy_to_udata(udata, &uresp, sizeof(uresp));
 	if (status) {
-		ocrdma_err("%s(%d) user copy error.\n", __func__, dev->id);
+		pr_err("%s(%d) user copy error.\n", __func__, dev->id);
 		goto err;
 	}
 	status = ocrdma_add_mmap(pd->uctx, uresp.sq_page_addr[0],
@@ -1023,15 +1006,6 @@ static void ocrdma_set_qp_init_params(struct ocrdma_qp *qp,
 	qp->state = OCRDMA_QPS_RST;
 }
 
-static void ocrdma_set_qp_use_cnt(struct ocrdma_qp *qp, struct ocrdma_pd *pd)
-{
-	atomic_inc(&pd->use_cnt);
-	atomic_inc(&qp->sq_cq->use_cnt);
-	atomic_inc(&qp->rq_cq->use_cnt);
-	if (qp->srq)
-		atomic_inc(&qp->srq->use_cnt);
-	qp->ibqp.qp_num = qp->id;
-}
 
 static void ocrdma_store_gsi_qp_cq(struct ocrdma_dev *dev,
 				   struct ib_qp_init_attr *attrs)
@@ -1099,7 +1073,7 @@ struct ib_qp *ocrdma_create_qp(struct ib_pd *ibpd,
 			goto cpy_err;
 	}
 	ocrdma_store_gsi_qp_cq(dev, attrs);
-	ocrdma_set_qp_use_cnt(qp, pd);
+	qp->ibqp.qp_num = qp->id;
 	mutex_unlock(&dev->dev_lock);
 	return &qp->ibqp;
 
@@ -1112,7 +1086,7 @@ mbx_err:
 	kfree(qp->wqe_wr_id_tbl);
 	kfree(qp->rqe_wr_id_tbl);
 	kfree(qp);
-	ocrdma_err("%s(%d) error=%d\n", __func__, dev->id, status);
+	pr_err("%s(%d) error=%d\n", __func__, dev->id, status);
 gen_err:
 	return ERR_PTR(status);
 }
@@ -1162,10 +1136,10 @@ int ocrdma_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 	spin_unlock_irqrestore(&qp->q_lock, flags);
 
 	if (!ib_modify_qp_is_ok(old_qps, new_qps, ibqp->qp_type, attr_mask)) {
-		ocrdma_err("%s(%d) invalid attribute mask=0x%x specified for "
-			   "qpn=0x%x of type=0x%x old_qps=0x%x, new_qps=0x%x\n",
-			   __func__, dev->id, attr_mask, qp->id, ibqp->qp_type,
-			   old_qps, new_qps);
+		pr_err("%s(%d) invalid attribute mask=0x%x specified for\n"
+		       "qpn=0x%x of type=0x%x old_qps=0x%x, new_qps=0x%x\n",
+		       __func__, dev->id, attr_mask, qp->id, ibqp->qp_type,
+		       old_qps, new_qps);
 		goto param_err;
 	}
 
@@ -1475,11 +1449,6 @@ int ocrdma_destroy_qp(struct ib_qp *ibqp)
 
 	ocrdma_del_flush_qp(qp);
 
-	atomic_dec(&qp->pd->use_cnt);
-	atomic_dec(&qp->sq_cq->use_cnt);
-	atomic_dec(&qp->rq_cq->use_cnt);
-	if (qp->srq)
-		atomic_dec(&qp->srq->use_cnt);
 	kfree(qp->wqe_wr_id_tbl);
 	kfree(qp->rqe_wr_id_tbl);
 	kfree(qp);
@@ -1565,14 +1534,12 @@ struct ib_srq *ocrdma_create_srq(struct ib_pd *ibpd,
 			goto arm_err;
 	}
 
-	atomic_set(&srq->use_cnt, 0);
 	if (udata) {
 		status = ocrdma_copy_srq_uresp(srq, udata);
 		if (status)
 			goto arm_err;
 	}
 
-	atomic_inc(&pd->use_cnt);
 	return &srq->ibsrq;
 
 arm_err:
@@ -1618,18 +1585,12 @@ int ocrdma_destroy_srq(struct ib_srq *ibsrq)
 
 	srq = get_ocrdma_srq(ibsrq);
 	dev = srq->dev;
-	if (atomic_read(&srq->use_cnt)) {
-		ocrdma_err("%s(%d) err, srq=0x%x in use\n",
-			   __func__, dev->id, srq->id);
-		return -EAGAIN;
-	}
 
 	status = ocrdma_mbx_destroy_srq(dev, srq);
 
 	if (srq->pd->uctx)
 		ocrdma_del_mmap(srq->pd->uctx, (u64) srq->rq.pa, srq->rq.len);
 
-	atomic_dec(&srq->pd->use_cnt);
 	kfree(srq->idx_bit_fields);
 	kfree(srq->rqe_wr_id_tbl);
 	kfree(srq);
@@ -1677,9 +1638,9 @@ static int ocrdma_build_inline_sges(struct ocrdma_qp *qp,
 {
 	if (wr->send_flags & IB_SEND_INLINE) {
 		if (wr->sg_list[0].length > qp->max_inline_data) {
-			ocrdma_err("%s() supported_len=0x%x,"
-				" unspported len req=0x%x\n", __func__,
-				qp->max_inline_data, wr->sg_list[0].length);
+			pr_err("%s() supported_len=0x%x,\n"
+			       " unspported len req=0x%x\n", __func__,
+			       qp->max_inline_data, wr->sg_list[0].length);
 			return -EINVAL;
 		}
 		memcpy(sge,
@@ -1773,12 +1734,14 @@ int ocrdma_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 	spin_lock_irqsave(&qp->q_lock, flags);
 	if (qp->state != OCRDMA_QPS_RTS && qp->state != OCRDMA_QPS_SQD) {
 		spin_unlock_irqrestore(&qp->q_lock, flags);
+		*bad_wr = wr;
 		return -EINVAL;
 	}
 
 	while (wr) {
 		if (ocrdma_hwq_free_cnt(&qp->sq) == 0 ||
 		    wr->num_sge > qp->sq.max_sges) {
+			*bad_wr = wr;
 			status = -ENOMEM;
 			break;
 		}
@@ -1856,7 +1819,7 @@ int ocrdma_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 
 static void ocrdma_ring_rq_db(struct ocrdma_qp *qp)
 {
-	u32 val = qp->rq.dbid | (1 << OCRDMA_GET_NUM_POSTED_SHIFT_VAL(qp));
+	u32 val = qp->rq.dbid | (1 << ocrdma_get_num_posted_shift(qp));
 
 	iowrite32(val, qp->rq_db);
 }
@@ -2094,8 +2057,8 @@ static void ocrdma_update_wc(struct ocrdma_qp *qp, struct ib_wc *ibwc,
 		break;
 	default:
 		ibwc->status = IB_WC_GENERAL_ERR;
-		ocrdma_err("%s() invalid opcode received = 0x%x\n",
-			   __func__, hdr->cw & OCRDMA_WQE_OPCODE_MASK);
+		pr_err("%s() invalid opcode received = 0x%x\n",
+		       __func__, hdr->cw & OCRDMA_WQE_OPCODE_MASK);
 		break;
 	};
 }

drivers/infiniband/hw/qib/Kconfig

@@ -5,3 +5,11 @@ config INFINIBAND_QIB
 	This is a low-level driver for Intel PCIe QLE InfiniBand host
 	channel adapters.  This driver does not support the Intel
 	HyperTransport card (model QHT7140).
+
+config INFINIBAND_QIB_DCA
+	bool "QIB DCA support"
+	depends on INFINIBAND_QIB && DCA && SMP && GENERIC_HARDIRQS && !(INFINIBAND_QIB=y && DCA=m)
+	default y
+	---help---
+	Setting this enables DCA support on some Intel chip sets
+	with the iba7322 HCA.

drivers/infiniband/hw/qib/Makefile

@@ -13,3 +13,4 @@ ib_qib-$(CONFIG_PCI_MSI) += qib_iba6120.o
 
 ib_qib-$(CONFIG_X86_64) += qib_wc_x86_64.o
 ib_qib-$(CONFIG_PPC64) += qib_wc_ppc64.o
+ib_qib-$(CONFIG_DEBUG_FS) += qib_debugfs.o

drivers/infiniband/hw/qib/qib.h

@@ -1,7 +1,7 @@
 #ifndef _QIB_KERNEL_H
 #define _QIB_KERNEL_H
 /*
- * Copyright (c) 2012 Intel Corporation.  All rights reserved.
+ * Copyright (c) 2012, 2013 Intel Corporation.  All rights reserved.
  * Copyright (c) 2006 - 2012 QLogic Corporation. All rights reserved.
  * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
  *
@@ -51,6 +51,7 @@
 #include <linux/completion.h>
 #include <linux/kref.h>
 #include <linux/sched.h>
+#include <linux/kthread.h>
 
 #include "qib_common.h"
 #include "qib_verbs.h"
@@ -114,6 +115,11 @@ struct qib_eep_log_mask {
 /*
  * Below contains all data related to a single context (formerly called port).
  */
+
+#ifdef CONFIG_DEBUG_FS
+struct qib_opcode_stats_perctx;
+#endif
+
 struct qib_ctxtdata {
 	void **rcvegrbuf;
 	dma_addr_t *rcvegrbuf_phys;
@@ -154,6 +160,8 @@ struct qib_ctxtdata {
 	 */
 	/* instead of calculating it */
 	unsigned ctxt;
+	/* local node of context */
+	int node_id;
 	/* non-zero if ctxt is being shared. */
 	u16 subctxt_cnt;
 	/* non-zero if ctxt is being shared. */
@@ -222,12 +230,15 @@ struct qib_ctxtdata {
 	u8 redirect_seq_cnt;
 	/* ctxt rcvhdrq head offset */
 	u32 head;
-	u32 pkt_count;
 	/* lookaside fields */
 	struct qib_qp *lookaside_qp;
 	u32 lookaside_qpn;
 	/* QPs waiting for context processing */
 	struct list_head qp_wait_list;
+#ifdef CONFIG_DEBUG_FS
+	/* verbs stats per CTX */
+	struct qib_opcode_stats_perctx *opstats;
+#endif
 };
 
 struct qib_sge_state;
@@ -428,9 +439,19 @@ struct qib_verbs_txreq {
 #define ACTIVITY_TIMER 5
 
 #define MAX_NAME_SIZE 64
+
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+struct qib_irq_notify;
+#endif
+
 struct qib_msix_entry {
 	struct msix_entry msix;
 	void *arg;
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+	int dca;
+	int rcv;
+	struct qib_irq_notify *notifier;
+#endif
 	char name[MAX_NAME_SIZE];
 	cpumask_var_t mask;
 };
@@ -828,6 +849,9 @@ struct qib_devdata {
 		struct qib_ctxtdata *);
 	void (*f_writescratch)(struct qib_devdata *, u32);
 	int (*f_tempsense_rd)(struct qib_devdata *, int regnum);
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+	int (*f_notify_dca)(struct qib_devdata *, unsigned long event);
+#endif
 
 	char *boardname; /* human readable board info */
 
@@ -1075,6 +1099,10 @@ struct qib_devdata {
 	u16 psxmitwait_check_rate;
 	/* high volume overflow errors defered to tasklet */
 	struct tasklet_struct error_tasklet;
+	/* per device cq worker */
+	struct kthread_worker *worker;
+
+	int assigned_node_id; /* NUMA node closest to HCA */
 };
 
 /* hol_state values */
@@ -1154,7 +1182,7 @@ int qib_create_rcvhdrq(struct qib_devdata *, struct qib_ctxtdata *);
 int qib_setup_eagerbufs(struct qib_ctxtdata *);
 void qib_set_ctxtcnt(struct qib_devdata *);
 int qib_create_ctxts(struct qib_devdata *dd);
-struct qib_ctxtdata *qib_create_ctxtdata(struct qib_pportdata *, u32);
+struct qib_ctxtdata *qib_create_ctxtdata(struct qib_pportdata *, u32, int);
 void qib_init_pportdata(struct qib_pportdata *, struct qib_devdata *, u8, u8);
 void qib_free_ctxtdata(struct qib_devdata *, struct qib_ctxtdata *);
 
@@ -1445,6 +1473,7 @@ extern unsigned qib_n_krcv_queues;
 extern unsigned qib_sdma_fetch_arb;
 extern unsigned qib_compat_ddr_negotiate;
 extern int qib_special_trigger;
+extern unsigned qib_numa_aware;
 
 extern struct mutex qib_mutex;
 
@@ -1474,27 +1503,23 @@ extern struct mutex qib_mutex;
  * first to avoid possible serial port delays from printk.
  */
 #define qib_early_err(dev, fmt, ...) \
-	do { \
-		dev_err(dev, fmt, ##__VA_ARGS__); \
-	} while (0)
+	dev_err(dev, fmt, ##__VA_ARGS__)
 
 #define qib_dev_err(dd, fmt, ...) \
-	do { \
-		dev_err(&(dd)->pcidev->dev, "%s: " fmt, \
-			qib_get_unit_name((dd)->unit), ##__VA_ARGS__); \
-	} while (0)
+	dev_err(&(dd)->pcidev->dev, "%s: " fmt, \
+		qib_get_unit_name((dd)->unit), ##__VA_ARGS__)
+
+#define qib_dev_warn(dd, fmt, ...) \
+	dev_warn(&(dd)->pcidev->dev, "%s: " fmt, \
+		qib_get_unit_name((dd)->unit), ##__VA_ARGS__)
 
 #define qib_dev_porterr(dd, port, fmt, ...) \
-	do { \
-		dev_err(&(dd)->pcidev->dev, "%s: IB%u:%u " fmt, \
-			qib_get_unit_name((dd)->unit), (dd)->unit, (port), \
-			##__VA_ARGS__); \
-	} while (0)
+	dev_err(&(dd)->pcidev->dev, "%s: IB%u:%u " fmt, \
+		qib_get_unit_name((dd)->unit), (dd)->unit, (port), \
+		##__VA_ARGS__)
 
 #define qib_devinfo(pcidev, fmt, ...) \
-	do { \
-		dev_info(&(pcidev)->dev, fmt, ##__VA_ARGS__); \
-	} while (0)
+	dev_info(&(pcidev)->dev, fmt, ##__VA_ARGS__)
 
 /*
  * this is used for formatting hw error messages...

drivers/infiniband/hw/qib/qib_common.h

@@ -279,7 +279,7 @@ struct qib_base_info {
  * may not be implemented; the user code must deal with this if it
  * cares, or it must abort after initialization reports the difference.
  */
-#define QIB_USER_SWMINOR 11
+#define QIB_USER_SWMINOR 12
 
 #define QIB_USER_SWVERSION ((QIB_USER_SWMAJOR << 16) | QIB_USER_SWMINOR)
 

drivers/infiniband/hw/qib/qib_cq.c

@@ -1,4 +1,5 @@
 /*
+ * Copyright (c) 2013 Intel Corporation.  All rights reserved.
  * Copyright (c) 2006, 2007, 2008, 2010 QLogic Corporation. All rights reserved.
  * Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
  *
@@ -34,8 +35,10 @@
 #include <linux/err.h>
 #include <linux/slab.h>
 #include <linux/vmalloc.h>
+#include <linux/kthread.h>
 
 #include "qib_verbs.h"
+#include "qib.h"
 
 /**
  * qib_cq_enter - add a new entry to the completion queue
@@ -102,13 +105,18 @@ void qib_cq_enter(struct qib_cq *cq, struct ib_wc *entry, int solicited)
 	if (cq->notify == IB_CQ_NEXT_COMP ||
 	    (cq->notify == IB_CQ_SOLICITED &&
 	     (solicited || entry->status != IB_WC_SUCCESS))) {
-		cq->notify = IB_CQ_NONE;
-		cq->triggered++;
+		struct kthread_worker *worker;
 		/*
 		 * This will cause send_complete() to be called in
 		 * another thread.
 		 */
-		queue_work(qib_cq_wq, &cq->comptask);
+		smp_rmb();
+		worker = cq->dd->worker;
+		if (likely(worker)) {
+			cq->notify = IB_CQ_NONE;
+			cq->triggered++;
+			queue_kthread_work(worker, &cq->comptask);
+		}
 	}
 
 	spin_unlock_irqrestore(&cq->lock, flags);
@@ -163,7 +171,7 @@ bail:
 	return npolled;
 }
 
-static void send_complete(struct work_struct *work)
+static void send_complete(struct kthread_work *work)
 {
 	struct qib_cq *cq = container_of(work, struct qib_cq, comptask);
 
@@ -287,11 +295,12 @@ struct ib_cq *qib_create_cq(struct ib_device *ibdev, int entries,
 	 * The number of entries should be >= the number requested or return
 	 * an error.
 	 */
+	cq->dd = dd_from_dev(dev);
 	cq->ibcq.cqe = entries;
 	cq->notify = IB_CQ_NONE;
 	cq->triggered = 0;
 	spin_lock_init(&cq->lock);
-	INIT_WORK(&cq->comptask, send_complete);
+	init_kthread_work(&cq->comptask, send_complete);
 	wc->head = 0;
 	wc->tail = 0;
 	cq->queue = wc;
@@ -323,7 +332,7 @@ int qib_destroy_cq(struct ib_cq *ibcq)
 	struct qib_ibdev *dev = to_idev(ibcq->device);
 	struct qib_cq *cq = to_icq(ibcq);
 
-	flush_work(&cq->comptask);
+	flush_kthread_work(&cq->comptask);
 	spin_lock(&dev->n_cqs_lock);
 	dev->n_cqs_allocated--;
 	spin_unlock(&dev->n_cqs_lock);
@@ -483,3 +492,49 @@ bail_free:
 bail:
 	return ret;
 }
+
+int qib_cq_init(struct qib_devdata *dd)
+{
+	int ret = 0;
+	int cpu;
+	struct task_struct *task;
+
+	if (dd->worker)
+		return 0;
+	dd->worker = kzalloc(sizeof(*dd->worker), GFP_KERNEL);
+	if (!dd->worker)
+		return -ENOMEM;
+	init_kthread_worker(dd->worker);
+	task = kthread_create_on_node(
+		kthread_worker_fn,
+		dd->worker,
+		dd->assigned_node_id,
+		"qib_cq%d", dd->unit);
+	if (IS_ERR(task))
+		goto task_fail;
+	cpu = cpumask_first(cpumask_of_node(dd->assigned_node_id));
+	kthread_bind(task, cpu);
+	wake_up_process(task);
+out:
+	return ret;
+task_fail:
+	ret = PTR_ERR(task);
+	kfree(dd->worker);
+	dd->worker = NULL;
+	goto out;
+}
+
+void qib_cq_exit(struct qib_devdata *dd)
+{
+	struct kthread_worker *worker;
+
+	worker = dd->worker;
+	if (!worker)
+		return;
+	/* blocks future queuing from send_complete() */
+	dd->worker = NULL;
+	smp_wmb();
+	flush_kthread_worker(worker);
+	kthread_stop(worker->task);
+	kfree(worker);
+}

drivers/infiniband/hw/qib/qib_debugfs.c

@@ -0,0 +1,283 @@
+#ifdef CONFIG_DEBUG_FS
+/*
+ * Copyright (c) 2013 Intel Corporation.  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/seq_file.h>
+#include <linux/kernel.h>
+#include <linux/export.h>
+
+#include "qib.h"
+#include "qib_verbs.h"
+#include "qib_debugfs.h"
+
+static struct dentry *qib_dbg_root;
+
+#define DEBUGFS_FILE(name) \
+static const struct seq_operations _##name##_seq_ops = { \
+	.start = _##name##_seq_start, \
+	.next  = _##name##_seq_next, \
+	.stop  = _##name##_seq_stop, \
+	.show  = _##name##_seq_show \
+}; \
+static int _##name##_open(struct inode *inode, struct file *s) \
+{ \
+	struct seq_file *seq; \
+	int ret; \
+	ret =  seq_open(s, &_##name##_seq_ops); \
+	if (ret) \
+		return ret; \
+	seq = s->private_data; \
+	seq->private = inode->i_private; \
+	return 0; \
+} \
+static const struct file_operations _##name##_file_ops = { \
+	.owner   = THIS_MODULE, \
+	.open    = _##name##_open, \
+	.read    = seq_read, \
+	.llseek  = seq_lseek, \
+	.release = seq_release \
+};
+
+#define DEBUGFS_FILE_CREATE(name) \
+do { \
+	struct dentry *ent; \
+	ent = debugfs_create_file(#name , 0400, ibd->qib_ibdev_dbg, \
+		ibd, &_##name##_file_ops); \
+	if (!ent) \
+		pr_warn("create of " #name " failed\n"); \
+} while (0)
+
+static void *_opcode_stats_seq_start(struct seq_file *s, loff_t *pos)
+{
+	struct qib_opcode_stats_perctx *opstats;
+
+	if (*pos >= ARRAY_SIZE(opstats->stats))
+		return NULL;
+	return pos;
+}
+
+static void *_opcode_stats_seq_next(struct seq_file *s, void *v, loff_t *pos)
+{
+	struct qib_opcode_stats_perctx *opstats;
+
+	++*pos;
+	if (*pos >= ARRAY_SIZE(opstats->stats))
+		return NULL;
+	return pos;
+}
+
+
+static void _opcode_stats_seq_stop(struct seq_file *s, void *v)
+{
+	/* nothing allocated */
+}
+
+static int _opcode_stats_seq_show(struct seq_file *s, void *v)
+{
+	loff_t *spos = v;
+	loff_t i = *spos, j;
+	u64 n_packets = 0, n_bytes = 0;
+	struct qib_ibdev *ibd = (struct qib_ibdev *)s->private;
+	struct qib_devdata *dd = dd_from_dev(ibd);
+
+	for (j = 0; j < dd->first_user_ctxt; j++) {
+		if (!dd->rcd[j])
+			continue;
+		n_packets += dd->rcd[j]->opstats->stats[i].n_packets;
+		n_bytes += dd->rcd[j]->opstats->stats[i].n_bytes;
+	}
+	if (!n_packets && !n_bytes)
+		return SEQ_SKIP;
+	seq_printf(s, "%02llx %llu/%llu\n", i,
+		(unsigned long long) n_packets,
+		(unsigned long long) n_bytes);
+
+	return 0;
+}
+
+DEBUGFS_FILE(opcode_stats)
+
+static void *_ctx_stats_seq_start(struct seq_file *s, loff_t *pos)
+{
+	struct qib_ibdev *ibd = (struct qib_ibdev *)s->private;
+	struct qib_devdata *dd = dd_from_dev(ibd);
+
+	if (!*pos)
+		return SEQ_START_TOKEN;
+	if (*pos >= dd->first_user_ctxt)
+		return NULL;
+	return pos;
+}
+
+static void *_ctx_stats_seq_next(struct seq_file *s, void *v, loff_t *pos)
+{
+	struct qib_ibdev *ibd = (struct qib_ibdev *)s->private;
+	struct qib_devdata *dd = dd_from_dev(ibd);
+
+	if (v == SEQ_START_TOKEN)
+		return pos;
+
+	++*pos;
+	if (*pos >= dd->first_user_ctxt)
+		return NULL;
+	return pos;
+}
+
+static void _ctx_stats_seq_stop(struct seq_file *s, void *v)
+{
+	/* nothing allocated */
+}
+
+static int _ctx_stats_seq_show(struct seq_file *s, void *v)
+{
+	loff_t *spos;
+	loff_t i, j;
+	u64 n_packets = 0;
+	struct qib_ibdev *ibd = (struct qib_ibdev *)s->private;
+	struct qib_devdata *dd = dd_from_dev(ibd);
+
+	if (v == SEQ_START_TOKEN) {
+		seq_puts(s, "Ctx:npkts\n");
+		return 0;
+	}
+
+	spos = v;
+	i = *spos;
+
+	if (!dd->rcd[i])
+		return SEQ_SKIP;
+
+	for (j = 0; j < ARRAY_SIZE(dd->rcd[i]->opstats->stats); j++)
+		n_packets += dd->rcd[i]->opstats->stats[j].n_packets;
+
+	if (!n_packets)
+		return SEQ_SKIP;
+
+	seq_printf(s, "  %llu:%llu\n", i, n_packets);
+	return 0;
+}
+
+DEBUGFS_FILE(ctx_stats)
+
+static void *_qp_stats_seq_start(struct seq_file *s, loff_t *pos)
+{
+	struct qib_qp_iter *iter;
+	loff_t n = *pos;
+
+	iter = qib_qp_iter_init(s->private);
+	if (!iter)
+		return NULL;
+
+	while (n--) {
+		if (qib_qp_iter_next(iter)) {
+			kfree(iter);
+			return NULL;
+		}
+	}
+
+	return iter;
+}
+
+static void *_qp_stats_seq_next(struct seq_file *s, void *iter_ptr,
+				   loff_t *pos)
+{
+	struct qib_qp_iter *iter = iter_ptr;
+
+	(*pos)++;
+
+	if (qib_qp_iter_next(iter)) {
+		kfree(iter);
+		return NULL;
+	}
+
+	return iter;
+}
+
+static void _qp_stats_seq_stop(struct seq_file *s, void *iter_ptr)
+{
+	/* nothing for now */
+}
+
+static int _qp_stats_seq_show(struct seq_file *s, void *iter_ptr)
+{
+	struct qib_qp_iter *iter = iter_ptr;
+
+	if (!iter)
+		return 0;
+
+	qib_qp_iter_print(s, iter);
+
+	return 0;
+}
+
+DEBUGFS_FILE(qp_stats)
+
+void qib_dbg_ibdev_init(struct qib_ibdev *ibd)
+{
+	char name[10];
+
+	snprintf(name, sizeof(name), "qib%d", dd_from_dev(ibd)->unit);
+	ibd->qib_ibdev_dbg = debugfs_create_dir(name, qib_dbg_root);
+	if (!ibd->qib_ibdev_dbg) {
+		pr_warn("create of %s failed\n", name);
+		return;
+	}
+	DEBUGFS_FILE_CREATE(opcode_stats);
+	DEBUGFS_FILE_CREATE(ctx_stats);
+	DEBUGFS_FILE_CREATE(qp_stats);
+	return;
+}
+
+void qib_dbg_ibdev_exit(struct qib_ibdev *ibd)
+{
+	if (!qib_dbg_root)
+		goto out;
+	debugfs_remove_recursive(ibd->qib_ibdev_dbg);
+out:
+	ibd->qib_ibdev_dbg = NULL;
+}
+
+void qib_dbg_init(void)
+{
+	qib_dbg_root = debugfs_create_dir(QIB_DRV_NAME, NULL);
+	if (!qib_dbg_root)
+		pr_warn("init of debugfs failed\n");
+}
+
+void qib_dbg_exit(void)
+{
+	debugfs_remove_recursive(qib_dbg_root);
+	qib_dbg_root = NULL;
+}
+
+#endif
+

drivers/infiniband/hw/qib/qib_debugfs.h

@@ -0,0 +1,45 @@
+#ifndef _QIB_DEBUGFS_H
+#define _QIB_DEBUGFS_H
+
+#ifdef CONFIG_DEBUG_FS
+/*
+ * Copyright (c) 2013 Intel Corporation.  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.
+ */
+
+struct qib_ibdev;
+void qib_dbg_ibdev_init(struct qib_ibdev *ibd);
+void qib_dbg_ibdev_exit(struct qib_ibdev *ibd);
+void qib_dbg_init(void);
+void qib_dbg_exit(void);
+
+#endif
+
+#endif                          /* _QIB_DEBUGFS_H */

drivers/infiniband/hw/qib/qib_driver.c

@@ -558,7 +558,6 @@ move_along:
 	}
 
 	rcd->head = l;
-	rcd->pkt_count += i;
 
 	/*
 	 * Iterate over all QPs waiting to respond.

drivers/infiniband/hw/qib/qib_file_ops.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2012 Intel Corporation. All rights reserved.
+ * Copyright (c) 2012, 2013 Intel Corporation. All rights reserved.
  * Copyright (c) 2006 - 2012 QLogic Corporation. All rights reserved.
  * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
  *
@@ -1155,6 +1155,49 @@ static unsigned int qib_poll(struct file *fp, struct poll_table_struct *pt)
 	return pollflag;
 }
 
+static void assign_ctxt_affinity(struct file *fp, struct qib_devdata *dd)
+{
+	struct qib_filedata *fd = fp->private_data;
+	const unsigned int weight = cpumask_weight(&current->cpus_allowed);
+	const struct cpumask *local_mask = cpumask_of_pcibus(dd->pcidev->bus);
+	int local_cpu;
+
+	/*
+	 * If process has NOT already set it's affinity, select and
+	 * reserve a processor for it on the local NUMA node.
+	 */
+	if ((weight >= qib_cpulist_count) &&
+		(cpumask_weight(local_mask) <= qib_cpulist_count)) {
+		for_each_cpu(local_cpu, local_mask)
+			if (!test_and_set_bit(local_cpu, qib_cpulist)) {
+				fd->rec_cpu_num = local_cpu;
+				return;
+			}
+	}
+
+	/*
+	 * If process has NOT already set it's affinity, select and
+	 * reserve a processor for it, as a rendevous for all
+	 * users of the driver.  If they don't actually later
+	 * set affinity to this cpu, or set it to some other cpu,
+	 * it just means that sooner or later we don't recommend
+	 * a cpu, and let the scheduler do it's best.
+	 */
+	if (weight >= qib_cpulist_count) {
+		int cpu;
+		cpu = find_first_zero_bit(qib_cpulist,
+					  qib_cpulist_count);
+		if (cpu == qib_cpulist_count)
+			qib_dev_err(dd,
+			"no cpus avail for affinity PID %u\n",
+			current->pid);
+		else {
+			__set_bit(cpu, qib_cpulist);
+			fd->rec_cpu_num = cpu;
+		}
+	}
+}
+
 /*
  * Check that userland and driver are compatible for subcontexts.
  */
@@ -1259,12 +1302,20 @@ bail:
 static int setup_ctxt(struct qib_pportdata *ppd, int ctxt,
 		      struct file *fp, const struct qib_user_info *uinfo)
 {
+	struct qib_filedata *fd = fp->private_data;
 	struct qib_devdata *dd = ppd->dd;
 	struct qib_ctxtdata *rcd;
 	void *ptmp = NULL;
 	int ret;
+	int numa_id;
 
-	rcd = qib_create_ctxtdata(ppd, ctxt);
+	assign_ctxt_affinity(fp, dd);
+
+	numa_id = qib_numa_aware ? ((fd->rec_cpu_num != -1) ?
+		cpu_to_node(fd->rec_cpu_num) :
+		numa_node_id()) : dd->assigned_node_id;
+
+	rcd = qib_create_ctxtdata(ppd, ctxt, numa_id);
 
 	/*
 	 * Allocate memory for use in qib_tid_update() at open to
@@ -1296,6 +1347,9 @@ static int setup_ctxt(struct qib_pportdata *ppd, int ctxt,
 	goto bail;
 
 bailerr:
+	if (fd->rec_cpu_num != -1)
+		__clear_bit(fd->rec_cpu_num, qib_cpulist);
+
 	dd->rcd[ctxt] = NULL;
 	kfree(rcd);
 	kfree(ptmp);
@@ -1485,6 +1539,57 @@ static int qib_open(struct inode *in, struct file *fp)
 	return fp->private_data ? 0 : -ENOMEM;
 }
 
+static int find_hca(unsigned int cpu, int *unit)
+{
+	int ret = 0, devmax, npresent, nup, ndev;
+
+	*unit = -1;
+
+	devmax = qib_count_units(&npresent, &nup);
+	if (!npresent) {
+		ret = -ENXIO;
+		goto done;
+	}
+	if (!nup) {
+		ret = -ENETDOWN;
+		goto done;
+	}
+	for (ndev = 0; ndev < devmax; ndev++) {
+		struct qib_devdata *dd = qib_lookup(ndev);
+		if (dd) {
+			if (pcibus_to_node(dd->pcidev->bus) < 0) {
+				ret = -EINVAL;
+				goto done;
+			}
+			if (cpu_to_node(cpu) ==
+				pcibus_to_node(dd->pcidev->bus)) {
+				*unit = ndev;
+				goto done;
+			}
+		}
+	}
+done:
+	return ret;
+}
+
+static int do_qib_user_sdma_queue_create(struct file *fp)
+{
+	struct qib_filedata *fd = fp->private_data;
+	struct qib_ctxtdata *rcd = fd->rcd;
+	struct qib_devdata *dd = rcd->dd;
+
+	if (dd->flags & QIB_HAS_SEND_DMA)
+
+		fd->pq = qib_user_sdma_queue_create(&dd->pcidev->dev,
+						    dd->unit,
+						    rcd->ctxt,
+						    fd->subctxt);
+		if (!fd->pq)
+			return -ENOMEM;
+
+	return 0;
+}
+
 /*
  * Get ctxt early, so can set affinity prior to memory allocation.
  */
@@ -1517,61 +1622,36 @@ static int qib_assign_ctxt(struct file *fp, const struct qib_user_info *uinfo)
 	if (qib_compatible_subctxts(swmajor, swminor) &&
 	    uinfo->spu_subctxt_cnt) {
 		ret = find_shared_ctxt(fp, uinfo);
-		if (ret) {
-			if (ret > 0)
-				ret = 0;
-			goto done_chk_sdma;
+		if (ret > 0) {
+			ret = do_qib_user_sdma_queue_create(fp);
+			if (!ret)
+				assign_ctxt_affinity(fp, (ctxt_fp(fp))->dd);
+			goto done_ok;
 		}
 	}
 
 	i_minor = iminor(file_inode(fp)) - QIB_USER_MINOR_BASE;
 	if (i_minor)
 		ret = find_free_ctxt(i_minor - 1, fp, uinfo);
-	else
+	else {
+		int unit;
+		const unsigned int cpu = cpumask_first(&current->cpus_allowed);
+		const unsigned int weight =
+			cpumask_weight(&current->cpus_allowed);
+
+		if (weight == 1 && !test_bit(cpu, qib_cpulist))
+			if (!find_hca(cpu, &unit) && unit >= 0)
+				if (!find_free_ctxt(unit, fp, uinfo)) {
+					ret = 0;
+					goto done_chk_sdma;
+				}
 		ret = get_a_ctxt(fp, uinfo, alg);
-
-done_chk_sdma:
-	if (!ret) {
-		struct qib_filedata *fd = fp->private_data;
-		const struct qib_ctxtdata *rcd = fd->rcd;
-		const struct qib_devdata *dd = rcd->dd;
-		unsigned int weight;
-
-		if (dd->flags & QIB_HAS_SEND_DMA) {
-			fd->pq = qib_user_sdma_queue_create(&dd->pcidev->dev,
-							    dd->unit,
-							    rcd->ctxt,
-							    fd->subctxt);
-			if (!fd->pq)
-				ret = -ENOMEM;
-		}
-
-		/*
-		 * If process has NOT already set it's affinity, select and
-		 * reserve a processor for it, as a rendezvous for all
-		 * users of the driver.  If they don't actually later
-		 * set affinity to this cpu, or set it to some other cpu,
-		 * it just means that sooner or later we don't recommend
-		 * a cpu, and let the scheduler do it's best.
-		 */
-		weight = cpumask_weight(tsk_cpus_allowed(current));
-		if (!ret && weight >= qib_cpulist_count) {
-			int cpu;
-			cpu = find_first_zero_bit(qib_cpulist,
-						  qib_cpulist_count);
-			if (cpu != qib_cpulist_count) {
-				__set_bit(cpu, qib_cpulist);
-				fd->rec_cpu_num = cpu;
-			}
-		} else if (weight == 1 &&
-			test_bit(cpumask_first(tsk_cpus_allowed(current)),
-				 qib_cpulist))
-			qib_devinfo(dd->pcidev,
-				"%s PID %u affinity set to cpu %d; already allocated\n",
-				current->comm, current->pid,
-				cpumask_first(tsk_cpus_allowed(current)));
 	}
 
+done_chk_sdma:
+	if (!ret)
+		ret = do_qib_user_sdma_queue_create(fp);
+done_ok:
 	mutex_unlock(&qib_mutex);
 
 done:

drivers/infiniband/hw/qib/qib_iba6120.c

@@ -3464,6 +3464,13 @@ static int qib_6120_tempsense_rd(struct qib_devdata *dd, int regnum)
 	return -ENXIO;
 }
 
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+static int qib_6120_notify_dca(struct qib_devdata *dd, unsigned long event)
+{
+	return 0;
+}
+#endif
+
 /* Dummy function, as 6120 boards never disable EEPROM Write */
 static int qib_6120_eeprom_wen(struct qib_devdata *dd, int wen)
 {
@@ -3539,6 +3546,9 @@ struct qib_devdata *qib_init_iba6120_funcs(struct pci_dev *pdev,
 	dd->f_xgxs_reset        = qib_6120_xgxs_reset;
 	dd->f_writescratch      = writescratch;
 	dd->f_tempsense_rd	= qib_6120_tempsense_rd;
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+	dd->f_notify_dca = qib_6120_notify_dca;
+#endif
 	/*
 	 * Do remaining pcie setup and save pcie values in dd.
 	 * Any error printing is already done by the init code.

drivers/infiniband/hw/qib/qib_iba7220.c

@@ -4513,6 +4513,13 @@ bail:
 	return ret;
 }
 
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+static int qib_7220_notify_dca(struct qib_devdata *dd, unsigned long event)
+{
+	return 0;
+}
+#endif
+
 /* Dummy function, as 7220 boards never disable EEPROM Write */
 static int qib_7220_eeprom_wen(struct qib_devdata *dd, int wen)
 {
@@ -4587,6 +4594,9 @@ struct qib_devdata *qib_init_iba7220_funcs(struct pci_dev *pdev,
 	dd->f_xgxs_reset        = qib_7220_xgxs_reset;
 	dd->f_writescratch      = writescratch;
 	dd->f_tempsense_rd	= qib_7220_tempsense_rd;
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+	dd->f_notify_dca = qib_7220_notify_dca;
+#endif
 	/*
 	 * Do remaining pcie setup and save pcie values in dd.
 	 * Any error printing is already done by the init code.

drivers/infiniband/hw/qib/qib_iba7322.c

@@ -44,6 +44,9 @@
 #include <linux/module.h>
 #include <rdma/ib_verbs.h>
 #include <rdma/ib_smi.h>
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+#include <linux/dca.h>
+#endif
 
 #include "qib.h"
 #include "qib_7322_regs.h"
@@ -519,6 +522,14 @@ static const u8 qib_7322_physportstate[0x20] = {
 	[0x17] = IB_PHYSPORTSTATE_CFG_TRAIN
 };
 
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+struct qib_irq_notify {
+	int rcv;
+	void *arg;
+	struct irq_affinity_notify notify;
+};
+#endif
+
 struct qib_chip_specific {
 	u64 __iomem *cregbase;
 	u64 *cntrs;
@@ -546,6 +557,12 @@ struct qib_chip_specific {
 	u32 lastbuf_for_pio;
 	u32 stay_in_freeze;
 	u32 recovery_ports_initted;
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+	u32 dca_ctrl;
+	int rhdr_cpu[18];
+	int sdma_cpu[2];
+	u64 dca_rcvhdr_ctrl[5]; /* B, C, D, E, F */
+#endif
 	struct qib_msix_entry *msix_entries;
 	unsigned long *sendchkenable;
 	unsigned long *sendgrhchk;
@@ -573,7 +590,7 @@ struct vendor_txdds_ent {
 static void write_tx_serdes_param(struct qib_pportdata *, struct txdds_ent *);
 
 #define TXDDS_TABLE_SZ 16 /* number of entries per speed in onchip table */
-#define TXDDS_EXTRA_SZ 13 /* number of extra tx settings entries */
+#define TXDDS_EXTRA_SZ 18 /* number of extra tx settings entries */
 #define TXDDS_MFG_SZ 2    /* number of mfg tx settings entries */
 #define SERDES_CHANS 4 /* yes, it's obvious, but one less magic number */
 
@@ -642,28 +659,76 @@ static struct {
 	irq_handler_t handler;
 	int lsb;
 	int port; /* 0 if not port-specific, else port # */
+	int dca;
 } irq_table[] = {
-	{ "", qib_7322intr, -1, 0 },
+	{ "", qib_7322intr, -1, 0, 0 },
 	{ " (buf avail)", qib_7322bufavail,
-		SYM_LSB(IntStatus, SendBufAvail), 0 },
+		SYM_LSB(IntStatus, SendBufAvail), 0, 0},
 	{ " (sdma 0)", sdma_intr,
-		SYM_LSB(IntStatus, SDmaInt_0), 1 },
+		SYM_LSB(IntStatus, SDmaInt_0), 1, 1 },
 	{ " (sdma 1)", sdma_intr,
-		SYM_LSB(IntStatus, SDmaInt_1), 2 },
+		SYM_LSB(IntStatus, SDmaInt_1), 2, 1 },
 	{ " (sdmaI 0)", sdma_idle_intr,
-		SYM_LSB(IntStatus, SDmaIdleInt_0), 1 },
+		SYM_LSB(IntStatus, SDmaIdleInt_0), 1, 1},
 	{ " (sdmaI 1)", sdma_idle_intr,
-		SYM_LSB(IntStatus, SDmaIdleInt_1), 2 },
+		SYM_LSB(IntStatus, SDmaIdleInt_1), 2, 1},
 	{ " (sdmaP 0)", sdma_progress_intr,
-		SYM_LSB(IntStatus, SDmaProgressInt_0), 1 },
+		SYM_LSB(IntStatus, SDmaProgressInt_0), 1, 1 },
 	{ " (sdmaP 1)", sdma_progress_intr,
-		SYM_LSB(IntStatus, SDmaProgressInt_1), 2 },
+		SYM_LSB(IntStatus, SDmaProgressInt_1), 2, 1 },
 	{ " (sdmaC 0)", sdma_cleanup_intr,
-		SYM_LSB(IntStatus, SDmaCleanupDone_0), 1 },
+		SYM_LSB(IntStatus, SDmaCleanupDone_0), 1, 0 },
 	{ " (sdmaC 1)", sdma_cleanup_intr,
-		SYM_LSB(IntStatus, SDmaCleanupDone_1), 2 },
+		SYM_LSB(IntStatus, SDmaCleanupDone_1), 2 , 0},
 };
 
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+
+static const struct dca_reg_map {
+	int     shadow_inx;
+	int     lsb;
+	u64     mask;
+	u16     regno;
+} dca_rcvhdr_reg_map[] = {
+	{ 0, SYM_LSB(DCACtrlB, RcvHdrq0DCAOPH),
+	   ~SYM_MASK(DCACtrlB, RcvHdrq0DCAOPH) , KREG_IDX(DCACtrlB) },
+	{ 0, SYM_LSB(DCACtrlB, RcvHdrq1DCAOPH),
+	   ~SYM_MASK(DCACtrlB, RcvHdrq1DCAOPH) , KREG_IDX(DCACtrlB) },
+	{ 0, SYM_LSB(DCACtrlB, RcvHdrq2DCAOPH),
+	   ~SYM_MASK(DCACtrlB, RcvHdrq2DCAOPH) , KREG_IDX(DCACtrlB) },
+	{ 0, SYM_LSB(DCACtrlB, RcvHdrq3DCAOPH),
+	   ~SYM_MASK(DCACtrlB, RcvHdrq3DCAOPH) , KREG_IDX(DCACtrlB) },
+	{ 1, SYM_LSB(DCACtrlC, RcvHdrq4DCAOPH),
+	   ~SYM_MASK(DCACtrlC, RcvHdrq4DCAOPH) , KREG_IDX(DCACtrlC) },
+	{ 1, SYM_LSB(DCACtrlC, RcvHdrq5DCAOPH),
+	   ~SYM_MASK(DCACtrlC, RcvHdrq5DCAOPH) , KREG_IDX(DCACtrlC) },
+	{ 1, SYM_LSB(DCACtrlC, RcvHdrq6DCAOPH),
+	   ~SYM_MASK(DCACtrlC, RcvHdrq6DCAOPH) , KREG_IDX(DCACtrlC) },
+	{ 1, SYM_LSB(DCACtrlC, RcvHdrq7DCAOPH),
+	   ~SYM_MASK(DCACtrlC, RcvHdrq7DCAOPH) , KREG_IDX(DCACtrlC) },
+	{ 2, SYM_LSB(DCACtrlD, RcvHdrq8DCAOPH),
+	   ~SYM_MASK(DCACtrlD, RcvHdrq8DCAOPH) , KREG_IDX(DCACtrlD) },
+	{ 2, SYM_LSB(DCACtrlD, RcvHdrq9DCAOPH),
+	   ~SYM_MASK(DCACtrlD, RcvHdrq9DCAOPH) , KREG_IDX(DCACtrlD) },
+	{ 2, SYM_LSB(DCACtrlD, RcvHdrq10DCAOPH),
+	   ~SYM_MASK(DCACtrlD, RcvHdrq10DCAOPH) , KREG_IDX(DCACtrlD) },
+	{ 2, SYM_LSB(DCACtrlD, RcvHdrq11DCAOPH),
+	   ~SYM_MASK(DCACtrlD, RcvHdrq11DCAOPH) , KREG_IDX(DCACtrlD) },
+	{ 3, SYM_LSB(DCACtrlE, RcvHdrq12DCAOPH),
+	   ~SYM_MASK(DCACtrlE, RcvHdrq12DCAOPH) , KREG_IDX(DCACtrlE) },
+	{ 3, SYM_LSB(DCACtrlE, RcvHdrq13DCAOPH),
+	   ~SYM_MASK(DCACtrlE, RcvHdrq13DCAOPH) , KREG_IDX(DCACtrlE) },
+	{ 3, SYM_LSB(DCACtrlE, RcvHdrq14DCAOPH),
+	   ~SYM_MASK(DCACtrlE, RcvHdrq14DCAOPH) , KREG_IDX(DCACtrlE) },
+	{ 3, SYM_LSB(DCACtrlE, RcvHdrq15DCAOPH),
+	   ~SYM_MASK(DCACtrlE, RcvHdrq15DCAOPH) , KREG_IDX(DCACtrlE) },
+	{ 4, SYM_LSB(DCACtrlF, RcvHdrq16DCAOPH),
+	   ~SYM_MASK(DCACtrlF, RcvHdrq16DCAOPH) , KREG_IDX(DCACtrlF) },
+	{ 4, SYM_LSB(DCACtrlF, RcvHdrq17DCAOPH),
+	   ~SYM_MASK(DCACtrlF, RcvHdrq17DCAOPH) , KREG_IDX(DCACtrlF) },
+};
+#endif
+
 /* ibcctrl bits */
 #define QLOGIC_IB_IBCC_LINKINITCMD_DISABLE 1
 /* cycle through TS1/TS2 till OK */
@@ -686,6 +751,13 @@ static void write_7322_init_portregs(struct qib_pportdata *);
 static void setup_7322_link_recovery(struct qib_pportdata *, u32);
 static void check_7322_rxe_status(struct qib_pportdata *);
 static u32 __iomem *qib_7322_getsendbuf(struct qib_pportdata *, u64, u32 *);
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+static void qib_setup_dca(struct qib_devdata *dd);
+static void setup_dca_notifier(struct qib_devdata *dd,
+			       struct qib_msix_entry *m);
+static void reset_dca_notifier(struct qib_devdata *dd,
+			       struct qib_msix_entry *m);
+#endif
 
 /**
  * qib_read_ureg32 - read 32-bit virtualized per-context register
@@ -2558,6 +2630,162 @@ static void qib_setup_7322_setextled(struct qib_pportdata *ppd, u32 on)
 		qib_write_kreg_port(ppd, krp_rcvpktledcnt, ledblink);
 }
 
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+
+static int qib_7322_notify_dca(struct qib_devdata *dd, unsigned long event)
+{
+	switch (event) {
+	case DCA_PROVIDER_ADD:
+		if (dd->flags & QIB_DCA_ENABLED)
+			break;
+		if (!dca_add_requester(&dd->pcidev->dev)) {
+			qib_devinfo(dd->pcidev, "DCA enabled\n");
+			dd->flags |= QIB_DCA_ENABLED;
+			qib_setup_dca(dd);
+		}
+		break;
+	case DCA_PROVIDER_REMOVE:
+		if (dd->flags & QIB_DCA_ENABLED) {
+			dca_remove_requester(&dd->pcidev->dev);
+			dd->flags &= ~QIB_DCA_ENABLED;
+			dd->cspec->dca_ctrl = 0;
+			qib_write_kreg(dd, KREG_IDX(DCACtrlA),
+				dd->cspec->dca_ctrl);
+		}
+		break;
+	}
+	return 0;
+}
+
+static void qib_update_rhdrq_dca(struct qib_ctxtdata *rcd, int cpu)
+{
+	struct qib_devdata *dd = rcd->dd;
+	struct qib_chip_specific *cspec = dd->cspec;
+
+	if (!(dd->flags & QIB_DCA_ENABLED))
+		return;
+	if (cspec->rhdr_cpu[rcd->ctxt] != cpu) {
+		const struct dca_reg_map *rmp;
+
+		cspec->rhdr_cpu[rcd->ctxt] = cpu;
+		rmp = &dca_rcvhdr_reg_map[rcd->ctxt];
+		cspec->dca_rcvhdr_ctrl[rmp->shadow_inx] &= rmp->mask;
+		cspec->dca_rcvhdr_ctrl[rmp->shadow_inx] |=
+			(u64) dca3_get_tag(&dd->pcidev->dev, cpu) << rmp->lsb;
+		qib_devinfo(dd->pcidev,
+			"Ctxt %d cpu %d dca %llx\n", rcd->ctxt, cpu,
+			(long long) cspec->dca_rcvhdr_ctrl[rmp->shadow_inx]);
+		qib_write_kreg(dd, rmp->regno,
+			       cspec->dca_rcvhdr_ctrl[rmp->shadow_inx]);
+		cspec->dca_ctrl |= SYM_MASK(DCACtrlA, RcvHdrqDCAEnable);
+		qib_write_kreg(dd, KREG_IDX(DCACtrlA), cspec->dca_ctrl);
+	}
+}
+
+static void qib_update_sdma_dca(struct qib_pportdata *ppd, int cpu)
+{
+	struct qib_devdata *dd = ppd->dd;
+	struct qib_chip_specific *cspec = dd->cspec;
+	unsigned pidx = ppd->port - 1;
+
+	if (!(dd->flags & QIB_DCA_ENABLED))
+		return;
+	if (cspec->sdma_cpu[pidx] != cpu) {
+		cspec->sdma_cpu[pidx] = cpu;
+		cspec->dca_rcvhdr_ctrl[4] &= ~(ppd->hw_pidx ?
+			SYM_MASK(DCACtrlF, SendDma1DCAOPH) :
+			SYM_MASK(DCACtrlF, SendDma0DCAOPH));
+		cspec->dca_rcvhdr_ctrl[4] |=
+			(u64) dca3_get_tag(&dd->pcidev->dev, cpu) <<
+				(ppd->hw_pidx ?
+					SYM_LSB(DCACtrlF, SendDma1DCAOPH) :
+					SYM_LSB(DCACtrlF, SendDma0DCAOPH));
+		qib_devinfo(dd->pcidev,
+			"sdma %d cpu %d dca %llx\n", ppd->hw_pidx, cpu,
+			(long long) cspec->dca_rcvhdr_ctrl[4]);
+		qib_write_kreg(dd, KREG_IDX(DCACtrlF),
+			       cspec->dca_rcvhdr_ctrl[4]);
+		cspec->dca_ctrl |= ppd->hw_pidx ?
+			SYM_MASK(DCACtrlA, SendDMAHead1DCAEnable) :
+			SYM_MASK(DCACtrlA, SendDMAHead0DCAEnable);
+		qib_write_kreg(dd, KREG_IDX(DCACtrlA), cspec->dca_ctrl);
+	}
+}
+
+static void qib_setup_dca(struct qib_devdata *dd)
+{
+	struct qib_chip_specific *cspec = dd->cspec;
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(cspec->rhdr_cpu); i++)
+		cspec->rhdr_cpu[i] = -1;
+	for (i = 0; i < ARRAY_SIZE(cspec->sdma_cpu); i++)
+		cspec->sdma_cpu[i] = -1;
+	cspec->dca_rcvhdr_ctrl[0] =
+		(1ULL << SYM_LSB(DCACtrlB, RcvHdrq0DCAXfrCnt)) |
+		(1ULL << SYM_LSB(DCACtrlB, RcvHdrq1DCAXfrCnt)) |
+		(1ULL << SYM_LSB(DCACtrlB, RcvHdrq2DCAXfrCnt)) |
+		(1ULL << SYM_LSB(DCACtrlB, RcvHdrq3DCAXfrCnt));
+	cspec->dca_rcvhdr_ctrl[1] =
+		(1ULL << SYM_LSB(DCACtrlC, RcvHdrq4DCAXfrCnt)) |
+		(1ULL << SYM_LSB(DCACtrlC, RcvHdrq5DCAXfrCnt)) |
+		(1ULL << SYM_LSB(DCACtrlC, RcvHdrq6DCAXfrCnt)) |
+		(1ULL << SYM_LSB(DCACtrlC, RcvHdrq7DCAXfrCnt));
+	cspec->dca_rcvhdr_ctrl[2] =
+		(1ULL << SYM_LSB(DCACtrlD, RcvHdrq8DCAXfrCnt)) |
+		(1ULL << SYM_LSB(DCACtrlD, RcvHdrq9DCAXfrCnt)) |
+		(1ULL << SYM_LSB(DCACtrlD, RcvHdrq10DCAXfrCnt)) |
+		(1ULL << SYM_LSB(DCACtrlD, RcvHdrq11DCAXfrCnt));
+	cspec->dca_rcvhdr_ctrl[3] =
+		(1ULL << SYM_LSB(DCACtrlE, RcvHdrq12DCAXfrCnt)) |
+		(1ULL << SYM_LSB(DCACtrlE, RcvHdrq13DCAXfrCnt)) |
+		(1ULL << SYM_LSB(DCACtrlE, RcvHdrq14DCAXfrCnt)) |
+		(1ULL << SYM_LSB(DCACtrlE, RcvHdrq15DCAXfrCnt));
+	cspec->dca_rcvhdr_ctrl[4] =
+		(1ULL << SYM_LSB(DCACtrlF, RcvHdrq16DCAXfrCnt)) |
+		(1ULL << SYM_LSB(DCACtrlF, RcvHdrq17DCAXfrCnt));
+	for (i = 0; i < ARRAY_SIZE(cspec->sdma_cpu); i++)
+		qib_write_kreg(dd, KREG_IDX(DCACtrlB) + i,
+			       cspec->dca_rcvhdr_ctrl[i]);
+	for (i = 0; i < cspec->num_msix_entries; i++)
+		setup_dca_notifier(dd, &cspec->msix_entries[i]);
+}
+
+static void qib_irq_notifier_notify(struct irq_affinity_notify *notify,
+			     const cpumask_t *mask)
+{
+	struct qib_irq_notify *n =
+		container_of(notify, struct qib_irq_notify, notify);
+	int cpu = cpumask_first(mask);
+
+	if (n->rcv) {
+		struct qib_ctxtdata *rcd = (struct qib_ctxtdata *)n->arg;
+		qib_update_rhdrq_dca(rcd, cpu);
+	} else {
+		struct qib_pportdata *ppd = (struct qib_pportdata *)n->arg;
+		qib_update_sdma_dca(ppd, cpu);
+	}
+}
+
+static void qib_irq_notifier_release(struct kref *ref)
+{
+	struct qib_irq_notify *n =
+		container_of(ref, struct qib_irq_notify, notify.kref);
+	struct qib_devdata *dd;
+
+	if (n->rcv) {
+		struct qib_ctxtdata *rcd = (struct qib_ctxtdata *)n->arg;
+		dd = rcd->dd;
+	} else {
+		struct qib_pportdata *ppd = (struct qib_pportdata *)n->arg;
+		dd = ppd->dd;
+	}
+	qib_devinfo(dd->pcidev,
+		"release on HCA notify 0x%p n 0x%p\n", ref, n);
+	kfree(n);
+}
+#endif
+
 /*
  * Disable MSIx interrupt if enabled, call generic MSIx code
  * to cleanup, and clear pending MSIx interrupts.
@@ -2575,6 +2803,9 @@ static void qib_7322_nomsix(struct qib_devdata *dd)
 
 		dd->cspec->num_msix_entries = 0;
 		for (i = 0; i < n; i++) {
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+			reset_dca_notifier(dd, &dd->cspec->msix_entries[i]);
+#endif
 			irq_set_affinity_hint(
 			  dd->cspec->msix_entries[i].msix.vector, NULL);
 			free_cpumask_var(dd->cspec->msix_entries[i].mask);
@@ -2602,6 +2833,15 @@ static void qib_setup_7322_cleanup(struct qib_devdata *dd)
 {
 	int i;
 
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+	if (dd->flags & QIB_DCA_ENABLED) {
+		dca_remove_requester(&dd->pcidev->dev);
+		dd->flags &= ~QIB_DCA_ENABLED;
+		dd->cspec->dca_ctrl = 0;
+		qib_write_kreg(dd, KREG_IDX(DCACtrlA), dd->cspec->dca_ctrl);
+	}
+#endif
+
 	qib_7322_free_irq(dd);
 	kfree(dd->cspec->cntrs);
 	kfree(dd->cspec->sendchkenable);
@@ -3068,6 +3308,53 @@ static irqreturn_t sdma_cleanup_intr(int irq, void *data)
 	return IRQ_HANDLED;
 }
 
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+
+static void reset_dca_notifier(struct qib_devdata *dd, struct qib_msix_entry *m)
+{
+	if (!m->dca)
+		return;
+	qib_devinfo(dd->pcidev,
+		"Disabling notifier on HCA %d irq %d\n",
+		dd->unit,
+		m->msix.vector);
+	irq_set_affinity_notifier(
+		m->msix.vector,
+		NULL);
+	m->notifier = NULL;
+}
+
+static void setup_dca_notifier(struct qib_devdata *dd, struct qib_msix_entry *m)
+{
+	struct qib_irq_notify *n;
+
+	if (!m->dca)
+		return;
+	n = kzalloc(sizeof(*n), GFP_KERNEL);
+	if (n) {
+		int ret;
+
+		m->notifier = n;
+		n->notify.irq = m->msix.vector;
+		n->notify.notify = qib_irq_notifier_notify;
+		n->notify.release = qib_irq_notifier_release;
+		n->arg = m->arg;
+		n->rcv = m->rcv;
+		qib_devinfo(dd->pcidev,
+			"set notifier irq %d rcv %d notify %p\n",
+			n->notify.irq, n->rcv, &n->notify);
+		ret = irq_set_affinity_notifier(
+				n->notify.irq,
+				&n->notify);
+		if (ret) {
+			m->notifier = NULL;
+			kfree(n);
+		}
+	}
+}
+
+#endif
+
 /*
  * Set up our chip-specific interrupt handler.
  * The interrupt type has already been setup, so
@@ -3149,6 +3436,9 @@ try_intx:
 		void *arg;
 		u64 val;
 		int lsb, reg, sh;
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+		int dca = 0;
+#endif
 
 		dd->cspec->msix_entries[msixnum].
 			name[sizeof(dd->cspec->msix_entries[msixnum].name) - 1]
@@ -3161,6 +3451,9 @@ try_intx:
 				arg = dd->pport + irq_table[i].port - 1;
 			} else
 				arg = dd;
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+			dca = irq_table[i].dca;
+#endif
 			lsb = irq_table[i].lsb;
 			handler = irq_table[i].handler;
 			snprintf(dd->cspec->msix_entries[msixnum].name,
@@ -3178,6 +3471,9 @@ try_intx:
 				continue;
 			if (qib_krcvq01_no_msi && ctxt < 2)
 				continue;
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+			dca = 1;
+#endif
 			lsb = QIB_I_RCVAVAIL_LSB + ctxt;
 			handler = qib_7322pintr;
 			snprintf(dd->cspec->msix_entries[msixnum].name,
@@ -3203,6 +3499,11 @@ try_intx:
 			goto try_intx;
 		}
 		dd->cspec->msix_entries[msixnum].arg = arg;
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+		dd->cspec->msix_entries[msixnum].dca = dca;
+		dd->cspec->msix_entries[msixnum].rcv =
+			handler == qib_7322pintr;
+#endif
 		if (lsb >= 0) {
 			reg = lsb / IBA7322_REDIRECT_VEC_PER_REG;
 			sh = (lsb % IBA7322_REDIRECT_VEC_PER_REG) *
@@ -6885,6 +7186,9 @@ struct qib_devdata *qib_init_iba7322_funcs(struct pci_dev *pdev,
 	dd->f_sdma_init_early   = qib_7322_sdma_init_early;
 	dd->f_writescratch      = writescratch;
 	dd->f_tempsense_rd	= qib_7322_tempsense_rd;
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+	dd->f_notify_dca	= qib_7322_notify_dca;
+#endif
 	/*
 	 * Do remaining PCIe setup and save PCIe values in dd.
 	 * Any error printing is already done by the init code.
@@ -6921,7 +7225,7 @@ struct qib_devdata *qib_init_iba7322_funcs(struct pci_dev *pdev,
 		actual_cnt -= dd->num_pports;
 
 	tabsize = actual_cnt;
-	dd->cspec->msix_entries = kmalloc(tabsize *
+	dd->cspec->msix_entries = kzalloc(tabsize *
 			sizeof(struct qib_msix_entry), GFP_KERNEL);
 	if (!dd->cspec->msix_entries) {
 		qib_dev_err(dd, "No memory for MSIx table\n");
@@ -6941,7 +7245,13 @@ struct qib_devdata *qib_init_iba7322_funcs(struct pci_dev *pdev,
 
 	/* clear diagctrl register, in case diags were running and crashed */
 	qib_write_kreg(dd, kr_hwdiagctrl, 0);
-
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+	if (!dca_add_requester(&pdev->dev)) {
+		qib_devinfo(dd->pcidev, "DCA enabled\n");
+		dd->flags |= QIB_DCA_ENABLED;
+		qib_setup_dca(dd);
+	}
+#endif
 	goto bail;
 
 bail_cleanup:
@@ -7156,15 +7466,20 @@ static const struct txdds_ent txdds_extra_sdr[TXDDS_EXTRA_SZ] = {
 	{  0, 0, 0,  1 },	/* QMH7342 backplane settings */
 	{  0, 0, 0,  2 },	/* QMH7342 backplane settings */
 	{  0, 0, 0,  2 },	/* QMH7342 backplane settings */
-	{  0, 0, 0, 11 },	/* QME7342 backplane settings */
-	{  0, 0, 0, 11 },	/* QME7342 backplane settings */
-	{  0, 0, 0, 11 },	/* QME7342 backplane settings */
-	{  0, 0, 0, 11 },	/* QME7342 backplane settings */
-	{  0, 0, 0, 11 },	/* QME7342 backplane settings */
-	{  0, 0, 0, 11 },	/* QME7342 backplane settings */
-	{  0, 0, 0, 11 },	/* QME7342 backplane settings */
 	{  0, 0, 0,  3 },	/* QMH7342 backplane settings */
 	{  0, 0, 0,  4 },	/* QMH7342 backplane settings */
+	{  0, 1, 4, 15 },	/* QME7342 backplane settings 1.0 */
+	{  0, 1, 3, 15 },	/* QME7342 backplane settings 1.0 */
+	{  0, 1, 0, 12 },	/* QME7342 backplane settings 1.0 */
+	{  0, 1, 0, 11 },	/* QME7342 backplane settings 1.0 */
+	{  0, 1, 0,  9 },	/* QME7342 backplane settings 1.0 */
+	{  0, 1, 0, 14 },	/* QME7342 backplane settings 1.0 */
+	{  0, 1, 2, 15 },	/* QME7342 backplane settings 1.0 */
+	{  0, 1, 0, 11 },       /* QME7342 backplane settings 1.1 */
+	{  0, 1, 0,  7 },       /* QME7342 backplane settings 1.1 */
+	{  0, 1, 0,  9 },       /* QME7342 backplane settings 1.1 */
+	{  0, 1, 0,  6 },       /* QME7342 backplane settings 1.1 */
+	{  0, 1, 0,  8 },       /* QME7342 backplane settings 1.1 */
 };
 
 static const struct txdds_ent txdds_extra_ddr[TXDDS_EXTRA_SZ] = {
@@ -7173,15 +7488,20 @@ static const struct txdds_ent txdds_extra_ddr[TXDDS_EXTRA_SZ] = {
 	{  0, 0, 0,  7 },	/* QMH7342 backplane settings */
 	{  0, 0, 0,  8 },	/* QMH7342 backplane settings */
 	{  0, 0, 0,  8 },	/* QMH7342 backplane settings */
-	{  0, 0, 0, 13 },	/* QME7342 backplane settings */
-	{  0, 0, 0, 13 },	/* QME7342 backplane settings */
-	{  0, 0, 0, 13 },	/* QME7342 backplane settings */
-	{  0, 0, 0, 13 },	/* QME7342 backplane settings */
-	{  0, 0, 0, 13 },	/* QME7342 backplane settings */
-	{  0, 0, 0, 13 },	/* QME7342 backplane settings */
-	{  0, 0, 0, 13 },	/* QME7342 backplane settings */
 	{  0, 0, 0,  9 },	/* QMH7342 backplane settings */
 	{  0, 0, 0, 10 },	/* QMH7342 backplane settings */
+	{  0, 1, 4, 15 },	/* QME7342 backplane settings 1.0 */
+	{  0, 1, 3, 15 },	/* QME7342 backplane settings 1.0 */
+	{  0, 1, 0, 12 },	/* QME7342 backplane settings 1.0 */
+	{  0, 1, 0, 11 },	/* QME7342 backplane settings 1.0 */
+	{  0, 1, 0,  9 },	/* QME7342 backplane settings 1.0 */
+	{  0, 1, 0, 14 },	/* QME7342 backplane settings 1.0 */
+	{  0, 1, 2, 15 },	/* QME7342 backplane settings 1.0 */
+	{  0, 1, 0, 11 },       /* QME7342 backplane settings 1.1 */
+	{  0, 1, 0,  7 },       /* QME7342 backplane settings 1.1 */
+	{  0, 1, 0,  9 },       /* QME7342 backplane settings 1.1 */
+	{  0, 1, 0,  6 },       /* QME7342 backplane settings 1.1 */
+	{  0, 1, 0,  8 },       /* QME7342 backplane settings 1.1 */
 };
 
 static const struct txdds_ent txdds_extra_qdr[TXDDS_EXTRA_SZ] = {
@@ -7190,15 +7510,20 @@ static const struct txdds_ent txdds_extra_qdr[TXDDS_EXTRA_SZ] = {
 	{  0, 1,  0,  5 },	/* QMH7342 backplane settings */
 	{  0, 1,  0,  6 },	/* QMH7342 backplane settings */
 	{  0, 1,  0,  8 },	/* QMH7342 backplane settings */
-	{  0, 1, 12, 10 },	/* QME7342 backplane setting */
-	{  0, 1, 12, 11 },	/* QME7342 backplane setting */
-	{  0, 1, 12, 12 },	/* QME7342 backplane setting */
-	{  0, 1, 12, 14 },	/* QME7342 backplane setting */
-	{  0, 1, 12,  6 },	/* QME7342 backplane setting */
-	{  0, 1, 12,  7 },	/* QME7342 backplane setting */
-	{  0, 1, 12,  8 },	/* QME7342 backplane setting */
 	{  0, 1,  0, 10 },	/* QMH7342 backplane settings */
 	{  0, 1,  0, 12 },	/* QMH7342 backplane settings */
+	{  0, 1,  4, 15 },	/* QME7342 backplane settings 1.0 */
+	{  0, 1,  3, 15 },	/* QME7342 backplane settings 1.0 */
+	{  0, 1,  0, 12 },	/* QME7342 backplane settings 1.0 */
+	{  0, 1,  0, 11 },	/* QME7342 backplane settings 1.0 */
+	{  0, 1,  0,  9 },	/* QME7342 backplane settings 1.0 */
+	{  0, 1,  0, 14 },	/* QME7342 backplane settings 1.0 */
+	{  0, 1,  2, 15 },	/* QME7342 backplane settings 1.0 */
+	{  0, 1,  0, 11 },      /* QME7342 backplane settings 1.1 */
+	{  0, 1,  0,  7 },      /* QME7342 backplane settings 1.1 */
+	{  0, 1,  0,  9 },      /* QME7342 backplane settings 1.1 */
+	{  0, 1,  0,  6 },      /* QME7342 backplane settings 1.1 */
+	{  0, 1,  0,  8 },      /* QME7342 backplane settings 1.1 */
 };
 
 static const struct txdds_ent txdds_extra_mfg[TXDDS_MFG_SZ] = {

drivers/infiniband/hw/qib/qib_init.c

@@ -39,10 +39,17 @@
 #include <linux/idr.h>
 #include <linux/module.h>
 #include <linux/printk.h>
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+#include <linux/dca.h>
+#endif
 
 #include "qib.h"
 #include "qib_common.h"
 #include "qib_mad.h"
+#ifdef CONFIG_DEBUG_FS
+#include "qib_debugfs.h"
+#include "qib_verbs.h"
+#endif
 
 #undef pr_fmt
 #define pr_fmt(fmt) QIB_DRV_NAME ": " fmt
@@ -64,6 +71,11 @@ ushort qib_cfgctxts;
 module_param_named(cfgctxts, qib_cfgctxts, ushort, S_IRUGO);
 MODULE_PARM_DESC(cfgctxts, "Set max number of contexts to use");
 
+unsigned qib_numa_aware;
+module_param_named(numa_aware, qib_numa_aware, uint, S_IRUGO);
+MODULE_PARM_DESC(numa_aware,
+	"0 -> PSM allocation close to HCA, 1 -> PSM allocation local to process");
+
 /*
  * If set, do not write to any regs if avoidable, hack to allow
  * check for deranged default register values.
@@ -89,8 +101,6 @@ unsigned qib_wc_pat = 1; /* default (1) is to use PAT, not MTRR */
 module_param_named(wc_pat, qib_wc_pat, uint, S_IRUGO);
 MODULE_PARM_DESC(wc_pat, "enable write-combining via PAT mechanism");
 
-struct workqueue_struct *qib_cq_wq;
-
 static void verify_interrupt(unsigned long);
 
 static struct idr qib_unit_table;
@@ -121,6 +131,11 @@ int qib_create_ctxts(struct qib_devdata *dd)
 {
 	unsigned i;
 	int ret;
+	int local_node_id = pcibus_to_node(dd->pcidev->bus);
+
+	if (local_node_id < 0)
+		local_node_id = numa_node_id();
+	dd->assigned_node_id = local_node_id;
 
 	/*
 	 * Allocate full ctxtcnt array, rather than just cfgctxts, because
@@ -143,7 +158,8 @@ int qib_create_ctxts(struct qib_devdata *dd)
 			continue;
 
 		ppd = dd->pport + (i % dd->num_pports);
-		rcd = qib_create_ctxtdata(ppd, i);
+
+		rcd = qib_create_ctxtdata(ppd, i, dd->assigned_node_id);
 		if (!rcd) {
 			qib_dev_err(dd,
 				"Unable to allocate ctxtdata for Kernel ctxt, failing\n");
@@ -161,20 +177,33 @@ done:
 /*
  * Common code for user and kernel context setup.
  */
-struct qib_ctxtdata *qib_create_ctxtdata(struct qib_pportdata *ppd, u32 ctxt)
+struct qib_ctxtdata *qib_create_ctxtdata(struct qib_pportdata *ppd, u32 ctxt,
+	int node_id)
 {
 	struct qib_devdata *dd = ppd->dd;
 	struct qib_ctxtdata *rcd;
 
-	rcd = kzalloc(sizeof(*rcd), GFP_KERNEL);
+	rcd = kzalloc_node(sizeof(*rcd), GFP_KERNEL, node_id);
 	if (rcd) {
 		INIT_LIST_HEAD(&rcd->qp_wait_list);
+		rcd->node_id = node_id;
 		rcd->ppd = ppd;
 		rcd->dd = dd;
 		rcd->cnt = 1;
 		rcd->ctxt = ctxt;
 		dd->rcd[ctxt] = rcd;
-
+#ifdef CONFIG_DEBUG_FS
+		if (ctxt < dd->first_user_ctxt) { /* N/A for PSM contexts */
+			rcd->opstats = kzalloc_node(sizeof(*rcd->opstats),
+				GFP_KERNEL, node_id);
+			if (!rcd->opstats) {
+				kfree(rcd);
+				qib_dev_err(dd,
+					"Unable to allocate per ctxt stats buffer\n");
+				return NULL;
+			}
+		}
+#endif
 		dd->f_init_ctxt(rcd);
 
 		/*
@@ -429,6 +458,7 @@ static int loadtime_init(struct qib_devdata *dd)
 	dd->intrchk_timer.function = verify_interrupt;
 	dd->intrchk_timer.data = (unsigned long) dd;
 
+	ret = qib_cq_init(dd);
 done:
 	return ret;
 }
@@ -944,6 +974,10 @@ void qib_free_ctxtdata(struct qib_devdata *dd, struct qib_ctxtdata *rcd)
 	vfree(rcd->subctxt_uregbase);
 	vfree(rcd->subctxt_rcvegrbuf);
 	vfree(rcd->subctxt_rcvhdr_base);
+#ifdef CONFIG_DEBUG_FS
+	kfree(rcd->opstats);
+	rcd->opstats = NULL;
+#endif
 	kfree(rcd);
 }
 
@@ -1033,7 +1067,6 @@ done:
 	dd->f_set_armlaunch(dd, 1);
 }
 
-
 void qib_free_devdata(struct qib_devdata *dd)
 {
 	unsigned long flags;
@@ -1043,6 +1076,9 @@ void qib_free_devdata(struct qib_devdata *dd)
 	list_del(&dd->list);
 	spin_unlock_irqrestore(&qib_devs_lock, flags);
 
+#ifdef CONFIG_DEBUG_FS
+	qib_dbg_ibdev_exit(&dd->verbs_dev);
+#endif
 	ib_dealloc_device(&dd->verbs_dev.ibdev);
 }
 
@@ -1066,6 +1102,10 @@ struct qib_devdata *qib_alloc_devdata(struct pci_dev *pdev, size_t extra)
 		goto bail;
 	}
 
+#ifdef CONFIG_DEBUG_FS
+	qib_dbg_ibdev_init(&dd->verbs_dev);
+#endif
+
 	idr_preload(GFP_KERNEL);
 	spin_lock_irqsave(&qib_devs_lock, flags);
 
@@ -1081,6 +1121,9 @@ struct qib_devdata *qib_alloc_devdata(struct pci_dev *pdev, size_t extra)
 	if (ret < 0) {
 		qib_early_err(&pdev->dev,
 			      "Could not allocate unit ID: error %d\n", -ret);
+#ifdef CONFIG_DEBUG_FS
+		qib_dbg_ibdev_exit(&dd->verbs_dev);
+#endif
 		ib_dealloc_device(&dd->verbs_dev.ibdev);
 		dd = ERR_PTR(ret);
 		goto bail;
@@ -1158,6 +1201,35 @@ struct pci_driver qib_driver = {
 	.err_handler = &qib_pci_err_handler,
 };
 
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+
+static int qib_notify_dca(struct notifier_block *, unsigned long, void *);
+static struct notifier_block dca_notifier = {
+	.notifier_call  = qib_notify_dca,
+	.next           = NULL,
+	.priority       = 0
+};
+
+static int qib_notify_dca_device(struct device *device, void *data)
+{
+	struct qib_devdata *dd = dev_get_drvdata(device);
+	unsigned long event = *(unsigned long *)data;
+
+	return dd->f_notify_dca(dd, event);
+}
+
+static int qib_notify_dca(struct notifier_block *nb, unsigned long event,
+					  void *p)
+{
+	int rval;
+
+	rval = driver_for_each_device(&qib_driver.driver, NULL,
+				      &event, qib_notify_dca_device);
+	return rval ? NOTIFY_BAD : NOTIFY_DONE;
+}
+
+#endif
+
 /*
  * Do all the generic driver unit- and chip-independent memory
  * allocation and initialization.
@@ -1170,22 +1242,22 @@ static int __init qlogic_ib_init(void)
 	if (ret)
 		goto bail;
 
-	qib_cq_wq = create_singlethread_workqueue("qib_cq");
-	if (!qib_cq_wq) {
-		ret = -ENOMEM;
-		goto bail_dev;
-	}
-
 	/*
 	 * These must be called before the driver is registered with
 	 * the PCI subsystem.
 	 */
 	idr_init(&qib_unit_table);
 
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+	dca_register_notify(&dca_notifier);
+#endif
+#ifdef CONFIG_DEBUG_FS
+	qib_dbg_init();
+#endif
 	ret = pci_register_driver(&qib_driver);
 	if (ret < 0) {
 		pr_err("Unable to register driver: error %d\n", -ret);
-		goto bail_unit;
+		goto bail_dev;
 	}
 
 	/* not fatal if it doesn't work */
@@ -1193,10 +1265,14 @@ static int __init qlogic_ib_init(void)
 		pr_err("Unable to register ipathfs\n");
 	goto bail; /* all OK */
 
-bail_unit:
-	idr_destroy(&qib_unit_table);
-	destroy_workqueue(qib_cq_wq);
 bail_dev:
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+	dca_unregister_notify(&dca_notifier);
+#endif
+#ifdef CONFIG_DEBUG_FS
+	qib_dbg_exit();
+#endif
+	idr_destroy(&qib_unit_table);
 	qib_dev_cleanup();
 bail:
 	return ret;
@@ -1217,9 +1293,13 @@ static void __exit qlogic_ib_cleanup(void)
 			"Unable to cleanup counter filesystem: error %d\n",
 			-ret);
 
+#ifdef CONFIG_INFINIBAND_QIB_DCA
+	dca_unregister_notify(&dca_notifier);
+#endif
 	pci_unregister_driver(&qib_driver);
-
-	destroy_workqueue(qib_cq_wq);
+#ifdef CONFIG_DEBUG_FS
+	qib_dbg_exit();
+#endif
 
 	qib_cpulist_count = 0;
 	kfree(qib_cpulist);
@@ -1311,6 +1391,7 @@ static void cleanup_device_data(struct qib_devdata *dd)
 	}
 	kfree(tmp);
 	kfree(dd->boardname);
+	qib_cq_exit(dd);
 }
 
 /*
@@ -1483,6 +1564,7 @@ static void qib_remove_one(struct pci_dev *pdev)
 int qib_create_rcvhdrq(struct qib_devdata *dd, struct qib_ctxtdata *rcd)
 {
 	unsigned amt;
+	int old_node_id;
 
 	if (!rcd->rcvhdrq) {
 		dma_addr_t phys_hdrqtail;
@@ -1492,9 +1574,13 @@ int qib_create_rcvhdrq(struct qib_devdata *dd, struct qib_ctxtdata *rcd)
 			    sizeof(u32), PAGE_SIZE);
 		gfp_flags = (rcd->ctxt >= dd->first_user_ctxt) ?
 			GFP_USER : GFP_KERNEL;
+
+		old_node_id = dev_to_node(&dd->pcidev->dev);
+		set_dev_node(&dd->pcidev->dev, rcd->node_id);
 		rcd->rcvhdrq = dma_alloc_coherent(
 			&dd->pcidev->dev, amt, &rcd->rcvhdrq_phys,
 			gfp_flags | __GFP_COMP);
+		set_dev_node(&dd->pcidev->dev, old_node_id);
 
 		if (!rcd->rcvhdrq) {
 			qib_dev_err(dd,
@@ -1510,9 +1596,11 @@ int qib_create_rcvhdrq(struct qib_devdata *dd, struct qib_ctxtdata *rcd)
 		}
 
 		if (!(dd->flags & QIB_NODMA_RTAIL)) {
+			set_dev_node(&dd->pcidev->dev, rcd->node_id);
 			rcd->rcvhdrtail_kvaddr = dma_alloc_coherent(
 				&dd->pcidev->dev, PAGE_SIZE, &phys_hdrqtail,
 				gfp_flags);
+			set_dev_node(&dd->pcidev->dev, old_node_id);
 			if (!rcd->rcvhdrtail_kvaddr)
 				goto bail_free;
 			rcd->rcvhdrqtailaddr_phys = phys_hdrqtail;
@@ -1556,6 +1644,7 @@ int qib_setup_eagerbufs(struct qib_ctxtdata *rcd)
 	unsigned e, egrcnt, egrperchunk, chunk, egrsize, egroff;
 	size_t size;
 	gfp_t gfp_flags;
+	int old_node_id;
 
 	/*
 	 * GFP_USER, but without GFP_FS, so buffer cache can be
@@ -1574,25 +1663,29 @@ int qib_setup_eagerbufs(struct qib_ctxtdata *rcd)
 	size = rcd->rcvegrbuf_size;
 	if (!rcd->rcvegrbuf) {
 		rcd->rcvegrbuf =
-			kzalloc(chunk * sizeof(rcd->rcvegrbuf[0]),
-				GFP_KERNEL);
+			kzalloc_node(chunk * sizeof(rcd->rcvegrbuf[0]),
+				GFP_KERNEL, rcd->node_id);
 		if (!rcd->rcvegrbuf)
 			goto bail;
 	}
 	if (!rcd->rcvegrbuf_phys) {
 		rcd->rcvegrbuf_phys =
-			kmalloc(chunk * sizeof(rcd->rcvegrbuf_phys[0]),
-				GFP_KERNEL);
+			kmalloc_node(chunk * sizeof(rcd->rcvegrbuf_phys[0]),
+				GFP_KERNEL, rcd->node_id);
 		if (!rcd->rcvegrbuf_phys)
 			goto bail_rcvegrbuf;
 	}
 	for (e = 0; e < rcd->rcvegrbuf_chunks; e++) {
 		if (rcd->rcvegrbuf[e])
 			continue;
+
+		old_node_id = dev_to_node(&dd->pcidev->dev);
+		set_dev_node(&dd->pcidev->dev, rcd->node_id);
 		rcd->rcvegrbuf[e] =
 			dma_alloc_coherent(&dd->pcidev->dev, size,
 					   &rcd->rcvegrbuf_phys[e],
 					   gfp_flags);
+		set_dev_node(&dd->pcidev->dev, old_node_id);
 		if (!rcd->rcvegrbuf[e])
 			goto bail_rcvegrbuf_phys;
 	}

drivers/infiniband/hw/qib/qib_qp.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2012 Intel Corporation.  All rights reserved.
+ * Copyright (c) 2012, 2013 Intel Corporation.  All rights reserved.
  * Copyright (c) 2006 - 2012 QLogic Corporation.  * All rights reserved.
  * Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
  *
@@ -35,6 +35,9 @@
 #include <linux/err.h>
 #include <linux/vmalloc.h>
 #include <linux/jhash.h>
+#ifdef CONFIG_DEBUG_FS
+#include <linux/seq_file.h>
+#endif
 
 #include "qib.h"
 
@@ -222,8 +225,8 @@ static void insert_qp(struct qib_ibdev *dev, struct qib_qp *qp)
 	unsigned long flags;
 	unsigned n = qpn_hash(dev, qp->ibqp.qp_num);
 
-	spin_lock_irqsave(&dev->qpt_lock, flags);
 	atomic_inc(&qp->refcount);
+	spin_lock_irqsave(&dev->qpt_lock, flags);
 
 	if (qp->ibqp.qp_num == 0)
 		rcu_assign_pointer(ibp->qp0, qp);
@@ -235,7 +238,6 @@ static void insert_qp(struct qib_ibdev *dev, struct qib_qp *qp)
 	}
 
 	spin_unlock_irqrestore(&dev->qpt_lock, flags);
-	synchronize_rcu();
 }
 
 /*
@@ -247,36 +249,39 @@ static void remove_qp(struct qib_ibdev *dev, struct qib_qp *qp)
 	struct qib_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
 	unsigned n = qpn_hash(dev, qp->ibqp.qp_num);
 	unsigned long flags;
+	int removed = 1;
 
 	spin_lock_irqsave(&dev->qpt_lock, flags);
 
 	if (rcu_dereference_protected(ibp->qp0,
 			lockdep_is_held(&dev->qpt_lock)) == qp) {
-		atomic_dec(&qp->refcount);
 		rcu_assign_pointer(ibp->qp0, NULL);
 	} else if (rcu_dereference_protected(ibp->qp1,
 			lockdep_is_held(&dev->qpt_lock)) == qp) {
-		atomic_dec(&qp->refcount);
 		rcu_assign_pointer(ibp->qp1, NULL);
 	} else {
 		struct qib_qp *q;
 		struct qib_qp __rcu **qpp;
 
+		removed = 0;
 		qpp = &dev->qp_table[n];
 		for (; (q = rcu_dereference_protected(*qpp,
 				lockdep_is_held(&dev->qpt_lock))) != NULL;
 				qpp = &q->next)
 			if (q == qp) {
-				atomic_dec(&qp->refcount);
 				rcu_assign_pointer(*qpp,
 					rcu_dereference_protected(qp->next,
 					 lockdep_is_held(&dev->qpt_lock)));
+				removed = 1;
 				break;
 			}
 	}
 
 	spin_unlock_irqrestore(&dev->qpt_lock, flags);
-	synchronize_rcu();
+	if (removed) {
+		synchronize_rcu();
+		atomic_dec(&qp->refcount);
+	}
 }
 
 /**
@@ -334,26 +339,25 @@ struct qib_qp *qib_lookup_qpn(struct qib_ibport *ibp, u32 qpn)
 {
 	struct qib_qp *qp = NULL;
 
+	rcu_read_lock();
 	if (unlikely(qpn <= 1)) {
-		rcu_read_lock();
 		if (qpn == 0)
 			qp = rcu_dereference(ibp->qp0);
 		else
 			qp = rcu_dereference(ibp->qp1);
+		if (qp)
+			atomic_inc(&qp->refcount);
 	} else {
 		struct qib_ibdev *dev = &ppd_from_ibp(ibp)->dd->verbs_dev;
 		unsigned n = qpn_hash(dev, qpn);
 
-		rcu_read_lock();
 		for (qp = rcu_dereference(dev->qp_table[n]); qp;
 			qp = rcu_dereference(qp->next))
-			if (qp->ibqp.qp_num == qpn)
+			if (qp->ibqp.qp_num == qpn) {
+				atomic_inc(&qp->refcount);
 				break;
+			}
 	}
-	if (qp)
-		if (unlikely(!atomic_inc_not_zero(&qp->refcount)))
-			qp = NULL;
-
 	rcu_read_unlock();
 	return qp;
 }
@@ -1286,3 +1290,94 @@ void qib_get_credit(struct qib_qp *qp, u32 aeth)
 		}
 	}
 }
+
+#ifdef CONFIG_DEBUG_FS
+
+struct qib_qp_iter {
+	struct qib_ibdev *dev;
+	struct qib_qp *qp;
+	int n;
+};
+
+struct qib_qp_iter *qib_qp_iter_init(struct qib_ibdev *dev)
+{
+	struct qib_qp_iter *iter;
+
+	iter = kzalloc(sizeof(*iter), GFP_KERNEL);
+	if (!iter)
+		return NULL;
+
+	iter->dev = dev;
+	if (qib_qp_iter_next(iter)) {
+		kfree(iter);
+		return NULL;
+	}
+
+	return iter;
+}
+
+int qib_qp_iter_next(struct qib_qp_iter *iter)
+{
+	struct qib_ibdev *dev = iter->dev;
+	int n = iter->n;
+	int ret = 1;
+	struct qib_qp *pqp = iter->qp;
+	struct qib_qp *qp;
+
+	rcu_read_lock();
+	for (; n < dev->qp_table_size; n++) {
+		if (pqp)
+			qp = rcu_dereference(pqp->next);
+		else
+			qp = rcu_dereference(dev->qp_table[n]);
+		pqp = qp;
+		if (qp) {
+			if (iter->qp)
+				atomic_dec(&iter->qp->refcount);
+			atomic_inc(&qp->refcount);
+			rcu_read_unlock();
+			iter->qp = qp;
+			iter->n = n;
+			return 0;
+		}
+	}
+	rcu_read_unlock();
+	if (iter->qp)
+		atomic_dec(&iter->qp->refcount);
+	return ret;
+}
+
+static const char * const qp_type_str[] = {
+	"SMI", "GSI", "RC", "UC", "UD",
+};
+
+void qib_qp_iter_print(struct seq_file *s, struct qib_qp_iter *iter)
+{
+	struct qib_swqe *wqe;
+	struct qib_qp *qp = iter->qp;
+
+	wqe = get_swqe_ptr(qp, qp->s_last);
+	seq_printf(s,
+		   "N %d QP%u %s %u %u %u f=%x %u %u %u %u %u PSN %x %x %x %x %x (%u %u %u %u %u %u) QP%u LID %x\n",
+		   iter->n,
+		   qp->ibqp.qp_num,
+		   qp_type_str[qp->ibqp.qp_type],
+		   qp->state,
+		   wqe->wr.opcode,
+		   qp->s_hdrwords,
+		   qp->s_flags,
+		   atomic_read(&qp->s_dma_busy),
+		   !list_empty(&qp->iowait),
+		   qp->timeout,
+		   wqe->ssn,
+		   qp->s_lsn,
+		   qp->s_last_psn,
+		   qp->s_psn, qp->s_next_psn,
+		   qp->s_sending_psn, qp->s_sending_hpsn,
+		   qp->s_last, qp->s_acked, qp->s_cur,
+		   qp->s_tail, qp->s_head, qp->s_size,
+		   qp->remote_qpn,
+		   qp->remote_ah_attr.dlid);
+}
+
+#endif

drivers/infiniband/hw/qib/qib_verbs.c

@@ -645,9 +645,11 @@ void qib_ib_rcv(struct qib_ctxtdata *rcd, void *rhdr, void *data, u32 tlen)
 	} else
 		goto drop;
 
-	opcode = be32_to_cpu(ohdr->bth[0]) >> 24;
-	ibp->opstats[opcode & 0x7f].n_bytes += tlen;
-	ibp->opstats[opcode & 0x7f].n_packets++;
+	opcode = (be32_to_cpu(ohdr->bth[0]) >> 24) & 0x7f;
+#ifdef CONFIG_DEBUG_FS
+	rcd->opstats->stats[opcode].n_bytes += tlen;
+	rcd->opstats->stats[opcode].n_packets++;
+#endif
 
 	/* Get the destination QP number. */
 	qp_num = be32_to_cpu(ohdr->bth[1]) & QIB_QPN_MASK;

drivers/infiniband/hw/qib/qib_verbs.h

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2012 Intel Corporation.  All rights reserved.
+ * Copyright (c) 2012, 2013 Intel Corporation.  All rights reserved.
  * Copyright (c) 2006 - 2012 QLogic Corporation. All rights reserved.
  * Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
  *
@@ -41,6 +41,7 @@
 #include <linux/interrupt.h>
 #include <linux/kref.h>
 #include <linux/workqueue.h>
+#include <linux/kthread.h>
 #include <linux/completion.h>
 #include <rdma/ib_pack.h>
 #include <rdma/ib_user_verbs.h>
@@ -267,7 +268,8 @@ struct qib_cq_wc {
  */
 struct qib_cq {
 	struct ib_cq ibcq;
-	struct work_struct comptask;
+	struct kthread_work comptask;
+	struct qib_devdata *dd;
 	spinlock_t lock; /* protect changes in this struct */
 	u8 notify;
 	u8 triggered;
@@ -658,6 +660,10 @@ struct qib_opcode_stats {
 	u64 n_bytes;            /* total number of bytes */
 };
 
+struct qib_opcode_stats_perctx {
+	struct qib_opcode_stats stats[128];
+};
+
 struct qib_ibport {
 	struct qib_qp __rcu *qp0;
 	struct qib_qp __rcu *qp1;
@@ -724,7 +730,6 @@ struct qib_ibport {
 	u8 vl_high_limit;
 	u8 sl_to_vl[16];
 
-	struct qib_opcode_stats opstats[128];
 };
 
 
@@ -768,6 +773,10 @@ struct qib_ibdev {
 	spinlock_t n_srqs_lock;
 	u32 n_mcast_grps_allocated; /* number of mcast groups allocated */
 	spinlock_t n_mcast_grps_lock;
+#ifdef CONFIG_DEBUG_FS
+	/* per HCA debugfs */
+	struct dentry *qib_ibdev_dbg;
+#endif
 };
 
 struct qib_verbs_counters {
@@ -832,8 +841,6 @@ static inline int qib_send_ok(struct qib_qp *qp)
 		 !(qp->s_flags & QIB_S_ANY_WAIT_SEND));
 }
 
-extern struct workqueue_struct *qib_cq_wq;
-
 /*
  * This must be called with s_lock held.
  */
@@ -910,6 +917,18 @@ void qib_init_qpn_table(struct qib_devdata *dd, struct qib_qpn_table *qpt);
 
 void qib_free_qpn_table(struct qib_qpn_table *qpt);
 
+#ifdef CONFIG_DEBUG_FS
+
+struct qib_qp_iter;
+
+struct qib_qp_iter *qib_qp_iter_init(struct qib_ibdev *dev);
+
+int qib_qp_iter_next(struct qib_qp_iter *iter);
+
+void qib_qp_iter_print(struct seq_file *s, struct qib_qp_iter *iter);
+
+#endif
+
 void qib_get_credit(struct qib_qp *qp, u32 aeth);
 
 unsigned qib_pkt_delay(u32 plen, u8 snd_mult, u8 rcv_mult);
@@ -972,6 +991,10 @@ int qib_query_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr);
 
 int qib_destroy_srq(struct ib_srq *ibsrq);
 
+int qib_cq_init(struct qib_devdata *dd);
+
+void qib_cq_exit(struct qib_devdata *dd);
+
 void qib_cq_enter(struct qib_cq *cq, struct ib_wc *entry, int sig);
 
 int qib_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry);

drivers/infiniband/ulp/srp/ib_srp.c

@@ -53,8 +53,8 @@
 
 #define DRV_NAME	"ib_srp"
 #define PFX		DRV_NAME ": "
-#define DRV_VERSION	"0.2"
-#define DRV_RELDATE	"November 1, 2005"
+#define DRV_VERSION	"1.0"
+#define DRV_RELDATE	"July 1, 2013"
 
 MODULE_AUTHOR("Roland Dreier");
 MODULE_DESCRIPTION("InfiniBand SCSI RDMA Protocol initiator "
@@ -231,14 +231,16 @@ static int srp_create_target_ib(struct srp_target_port *target)
 		return -ENOMEM;
 
 	recv_cq = ib_create_cq(target->srp_host->srp_dev->dev,
-			       srp_recv_completion, NULL, target, SRP_RQ_SIZE, 0);
+			       srp_recv_completion, NULL, target, SRP_RQ_SIZE,
+			       target->comp_vector);
 	if (IS_ERR(recv_cq)) {
 		ret = PTR_ERR(recv_cq);
 		goto err;
 	}
 
 	send_cq = ib_create_cq(target->srp_host->srp_dev->dev,
-			       srp_send_completion, NULL, target, SRP_SQ_SIZE, 0);
+			       srp_send_completion, NULL, target, SRP_SQ_SIZE,
+			       target->comp_vector);
 	if (IS_ERR(send_cq)) {
 		ret = PTR_ERR(send_cq);
 		goto err_recv_cq;
@@ -542,11 +544,11 @@ static void srp_remove_work(struct work_struct *work)
 
 	WARN_ON_ONCE(target->state != SRP_TARGET_REMOVED);
 
+	srp_remove_target(target);
+
 	spin_lock(&target->srp_host->target_lock);
 	list_del(&target->list);
 	spin_unlock(&target->srp_host->target_lock);
-
-	srp_remove_target(target);
 }
 
 static void srp_rport_delete(struct srp_rport *rport)
@@ -1744,18 +1746,25 @@ static int srp_abort(struct scsi_cmnd *scmnd)
 {
 	struct srp_target_port *target = host_to_target(scmnd->device->host);
 	struct srp_request *req = (struct srp_request *) scmnd->host_scribble;
+	int ret;
 
 	shost_printk(KERN_ERR, target->scsi_host, "SRP abort called\n");
 
 	if (!req || !srp_claim_req(target, req, scmnd))
 		return FAILED;
-	srp_send_tsk_mgmt(target, req->index, scmnd->device->lun,
-			  SRP_TSK_ABORT_TASK);
+	if (srp_send_tsk_mgmt(target, req->index, scmnd->device->lun,
+			      SRP_TSK_ABORT_TASK) == 0 ||
+	    target->transport_offline)
+		ret = SUCCESS;
+	else if (target->transport_offline)
+		ret = FAST_IO_FAIL;
+	else
+		ret = FAILED;
 	srp_free_req(target, req, scmnd, 0);
 	scmnd->result = DID_ABORT << 16;
 	scmnd->scsi_done(scmnd);
 
-	return SUCCESS;
+	return ret;
 }
 
 static int srp_reset_device(struct scsi_cmnd *scmnd)
@@ -1891,6 +1900,14 @@ static ssize_t show_local_ib_device(struct device *dev,
 	return sprintf(buf, "%s\n", target->srp_host->srp_dev->dev->name);
 }
 
+static ssize_t show_comp_vector(struct device *dev,
+				struct device_attribute *attr, char *buf)
+{
+	struct srp_target_port *target = host_to_target(class_to_shost(dev));
+
+	return sprintf(buf, "%d\n", target->comp_vector);
+}
+
 static ssize_t show_cmd_sg_entries(struct device *dev,
 				   struct device_attribute *attr, char *buf)
 {
@@ -1917,6 +1934,7 @@ static DEVICE_ATTR(req_lim,         S_IRUGO, show_req_lim,         NULL);
 static DEVICE_ATTR(zero_req_lim,    S_IRUGO, show_zero_req_lim,	   NULL);
 static DEVICE_ATTR(local_ib_port,   S_IRUGO, show_local_ib_port,   NULL);
 static DEVICE_ATTR(local_ib_device, S_IRUGO, show_local_ib_device, NULL);
+static DEVICE_ATTR(comp_vector,     S_IRUGO, show_comp_vector,     NULL);
 static DEVICE_ATTR(cmd_sg_entries,  S_IRUGO, show_cmd_sg_entries,  NULL);
 static DEVICE_ATTR(allow_ext_sg,    S_IRUGO, show_allow_ext_sg,    NULL);
 
@@ -1931,6 +1949,7 @@ static struct device_attribute *srp_host_attrs[] = {
 	&dev_attr_zero_req_lim,
 	&dev_attr_local_ib_port,
 	&dev_attr_local_ib_device,
+	&dev_attr_comp_vector,
 	&dev_attr_cmd_sg_entries,
 	&dev_attr_allow_ext_sg,
 	NULL
@@ -1946,6 +1965,7 @@ static struct scsi_host_template srp_template = {
 	.eh_abort_handler		= srp_abort,
 	.eh_device_reset_handler	= srp_reset_device,
 	.eh_host_reset_handler		= srp_reset_host,
+	.skip_settle_delay		= true,
 	.sg_tablesize			= SRP_DEF_SG_TABLESIZE,
 	.can_queue			= SRP_CMD_SQ_SIZE,
 	.this_id			= -1,
@@ -2001,6 +2021,36 @@ static struct class srp_class = {
 	.dev_release = srp_release_dev
 };
 
+/**
+ * srp_conn_unique() - check whether the connection to a target is unique
+ */
+static bool srp_conn_unique(struct srp_host *host,
+			    struct srp_target_port *target)
+{
+	struct srp_target_port *t;
+	bool ret = false;
+
+	if (target->state == SRP_TARGET_REMOVED)
+		goto out;
+
+	ret = true;
+
+	spin_lock(&host->target_lock);
+	list_for_each_entry(t, &host->target_list, list) {
+		if (t != target &&
+		    target->id_ext == t->id_ext &&
+		    target->ioc_guid == t->ioc_guid &&
+		    target->initiator_ext == t->initiator_ext) {
+			ret = false;
+			break;
+		}
+	}
+	spin_unlock(&host->target_lock);
+
+out:
+	return ret;
+}
+
 /*
  * Target ports are added by writing
  *
@@ -2023,6 +2073,7 @@ enum {
 	SRP_OPT_CMD_SG_ENTRIES	= 1 << 9,
 	SRP_OPT_ALLOW_EXT_SG	= 1 << 10,
 	SRP_OPT_SG_TABLESIZE	= 1 << 11,
+	SRP_OPT_COMP_VECTOR	= 1 << 12,
 	SRP_OPT_ALL		= (SRP_OPT_ID_EXT	|
 				   SRP_OPT_IOC_GUID	|
 				   SRP_OPT_DGID		|
@@ -2043,6 +2094,7 @@ static const match_table_t srp_opt_tokens = {
 	{ SRP_OPT_CMD_SG_ENTRIES,	"cmd_sg_entries=%u"	},
 	{ SRP_OPT_ALLOW_EXT_SG,		"allow_ext_sg=%u"	},
 	{ SRP_OPT_SG_TABLESIZE,		"sg_tablesize=%u"	},
+	{ SRP_OPT_COMP_VECTOR,		"comp_vector=%u"	},
 	{ SRP_OPT_ERR,			NULL 			}
 };
 
@@ -2198,6 +2250,14 @@ static int srp_parse_options(const char *buf, struct srp_target_port *target)
 			target->sg_tablesize = token;
 			break;
 
+		case SRP_OPT_COMP_VECTOR:
+			if (match_int(args, &token) || token < 0) {
+				pr_warn("bad comp_vector parameter '%s'\n", p);
+				goto out;
+			}
+			target->comp_vector = token;
+			break;
+
 		default:
 			pr_warn("unknown parameter or missing value '%s' in target creation request\n",
 				p);
@@ -2257,6 +2317,16 @@ static ssize_t srp_create_target(struct device *dev,
 	if (ret)
 		goto err;
 
+	if (!srp_conn_unique(target->srp_host, target)) {
+		shost_printk(KERN_INFO, target->scsi_host,
+			     PFX "Already connected to target port with id_ext=%016llx;ioc_guid=%016llx;initiator_ext=%016llx\n",
+			     be64_to_cpu(target->id_ext),
+			     be64_to_cpu(target->ioc_guid),
+			     be64_to_cpu(target->initiator_ext));
+		ret = -EEXIST;
+		goto err;
+	}
+
 	if (!host->srp_dev->fmr_pool && !target->allow_ext_sg &&
 				target->cmd_sg_cnt < target->sg_tablesize) {
 		pr_warn("No FMR pool and no external indirect descriptors, limiting sg_tablesize to cmd_sg_cnt\n");
@@ -2507,6 +2577,8 @@ static void srp_remove_one(struct ib_device *device)
 	struct srp_target_port *target;
 
 	srp_dev = ib_get_client_data(device, &srp_client);
+	if (!srp_dev)
+		return;
 
 	list_for_each_entry_safe(host, tmp_host, &srp_dev->dev_list, list) {
 		device_unregister(&host->dev);

drivers/infiniband/ulp/srp/ib_srp.h

@@ -156,6 +156,7 @@ struct srp_target_port {
 	char			target_name[32];
 	unsigned int		scsi_id;
 	unsigned int		sg_tablesize;
+	int			comp_vector;
 
 	struct ib_sa_path_rec	path;
 	__be16			orig_dgid[8];

drivers/net/ethernet/mellanox/Kconfig

@@ -19,5 +19,6 @@ config NET_VENDOR_MELLANOX
 if NET_VENDOR_MELLANOX
 
 source "drivers/net/ethernet/mellanox/mlx4/Kconfig"
+source "drivers/net/ethernet/mellanox/mlx5/core/Kconfig"
 
 endif # NET_VENDOR_MELLANOX

drivers/net/ethernet/mellanox/Makefile

@@ -3,3 +3,4 @@
 #
 
 obj-$(CONFIG_MLX4_CORE) += mlx4/
+obj-$(CONFIG_MLX5_CORE) += mlx5/core/

drivers/net/ethernet/mellanox/mlx5/core/Kconfig

@@ -0,0 +1,18 @@
+#
+# Mellanox driver configuration
+#
+
+config MLX5_CORE
+	tristate
+	depends on PCI && X86
+	default n
+
+config MLX5_DEBUG
+	bool "Verbose debugging output" if (MLX5_CORE && EXPERT)
+	depends on MLX5_CORE
+	default y
+	---help---
+	  This option causes debugging code to be compiled into the
+	  mlx5_core driver.  The output can be turned on via the
+	  debug_mask module parameter (which can also be set after
+	  the driver is loaded through sysfs).

drivers/net/ethernet/mellanox/mlx5/core/Makefile

@@ -0,0 +1,5 @@
+obj-$(CONFIG_MLX5_CORE)		+= mlx5_core.o
+
+mlx5_core-y :=	main.o cmd.o debugfs.o fw.o eq.o uar.o pagealloc.o \
+		health.o mcg.o cq.o srq.o alloc.o qp.o port.o mr.o pd.o   \
+		mad.o

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

@@ -0,0 +1,238 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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/errno.h>
+#include <linux/slab.h>
+#include <linux/mm.h>
+#include <linux/export.h>
+#include <linux/bitmap.h>
+#include <linux/dma-mapping.h>
+#include <linux/vmalloc.h>
+#include <linux/mlx5/driver.h>
+
+#include "mlx5_core.h"
+
+/* Handling for queue buffers -- we allocate a bunch of memory and
+ * register it in a memory region at HCA virtual address 0.  If the
+ * requested size is > max_direct, we split the allocation into
+ * multiple pages, so we don't require too much contiguous memory.
+ */
+
+int mlx5_buf_alloc(struct mlx5_core_dev *dev, int size, int max_direct,
+		   struct mlx5_buf *buf)
+{
+	dma_addr_t t;
+
+	buf->size = size;
+	if (size <= max_direct) {
+		buf->nbufs        = 1;
+		buf->npages       = 1;
+		buf->page_shift   = get_order(size) + PAGE_SHIFT;
+		buf->direct.buf   = dma_zalloc_coherent(&dev->pdev->dev,
+							size, &t, GFP_KERNEL);
+		if (!buf->direct.buf)
+			return -ENOMEM;
+
+		buf->direct.map = t;
+
+		while (t & ((1 << buf->page_shift) - 1)) {
+			--buf->page_shift;
+			buf->npages *= 2;
+		}
+	} else {
+		int i;
+
+		buf->direct.buf  = NULL;
+		buf->nbufs       = (size + PAGE_SIZE - 1) / PAGE_SIZE;
+		buf->npages      = buf->nbufs;
+		buf->page_shift  = PAGE_SHIFT;
+		buf->page_list   = kcalloc(buf->nbufs, sizeof(*buf->page_list),
+					   GFP_KERNEL);
+		if (!buf->page_list)
+			return -ENOMEM;
+
+		for (i = 0; i < buf->nbufs; i++) {
+			buf->page_list[i].buf =
+				dma_zalloc_coherent(&dev->pdev->dev, PAGE_SIZE,
+						    &t, GFP_KERNEL);
+			if (!buf->page_list[i].buf)
+				goto err_free;
+
+			buf->page_list[i].map = t;
+		}
+
+		if (BITS_PER_LONG == 64) {
+			struct page **pages;
+			pages = kmalloc(sizeof(*pages) * buf->nbufs, GFP_KERNEL);
+			if (!pages)
+				goto err_free;
+			for (i = 0; i < buf->nbufs; i++)
+				pages[i] = virt_to_page(buf->page_list[i].buf);
+			buf->direct.buf = vmap(pages, buf->nbufs, VM_MAP, PAGE_KERNEL);
+			kfree(pages);
+			if (!buf->direct.buf)
+				goto err_free;
+		}
+	}
+
+	return 0;
+
+err_free:
+	mlx5_buf_free(dev, buf);
+
+	return -ENOMEM;
+}
+EXPORT_SYMBOL_GPL(mlx5_buf_alloc);
+
+void mlx5_buf_free(struct mlx5_core_dev *dev, struct mlx5_buf *buf)
+{
+	int i;
+
+	if (buf->nbufs == 1)
+		dma_free_coherent(&dev->pdev->dev, buf->size, buf->direct.buf,
+				  buf->direct.map);
+	else {
+		if (BITS_PER_LONG == 64 && buf->direct.buf)
+			vunmap(buf->direct.buf);
+
+		for (i = 0; i < buf->nbufs; i++)
+			if (buf->page_list[i].buf)
+				dma_free_coherent(&dev->pdev->dev, PAGE_SIZE,
+						  buf->page_list[i].buf,
+						  buf->page_list[i].map);
+		kfree(buf->page_list);
+	}
+}
+EXPORT_SYMBOL_GPL(mlx5_buf_free);
+
+static struct mlx5_db_pgdir *mlx5_alloc_db_pgdir(struct device *dma_device)
+{
+	struct mlx5_db_pgdir *pgdir;
+
+	pgdir = kzalloc(sizeof(*pgdir), GFP_KERNEL);
+	if (!pgdir)
+		return NULL;
+
+	bitmap_fill(pgdir->bitmap, MLX5_DB_PER_PAGE);
+	pgdir->db_page = dma_alloc_coherent(dma_device, PAGE_SIZE,
+					    &pgdir->db_dma, GFP_KERNEL);
+	if (!pgdir->db_page) {
+		kfree(pgdir);
+		return NULL;
+	}
+
+	return pgdir;
+}
+
+static int mlx5_alloc_db_from_pgdir(struct mlx5_db_pgdir *pgdir,
+				    struct mlx5_db *db)
+{
+	int offset;
+	int i;
+
+	i = find_first_bit(pgdir->bitmap, MLX5_DB_PER_PAGE);
+	if (i >= MLX5_DB_PER_PAGE)
+		return -ENOMEM;
+
+	__clear_bit(i, pgdir->bitmap);
+
+	db->u.pgdir = pgdir;
+	db->index   = i;
+	offset = db->index * L1_CACHE_BYTES;
+	db->db      = pgdir->db_page + offset / sizeof(*pgdir->db_page);
+	db->dma     = pgdir->db_dma  + offset;
+
+	return 0;
+}
+
+int mlx5_db_alloc(struct mlx5_core_dev *dev, struct mlx5_db *db)
+{
+	struct mlx5_db_pgdir *pgdir;
+	int ret = 0;
+
+	mutex_lock(&dev->priv.pgdir_mutex);
+
+	list_for_each_entry(pgdir, &dev->priv.pgdir_list, list)
+		if (!mlx5_alloc_db_from_pgdir(pgdir, db))
+			goto out;
+
+	pgdir = mlx5_alloc_db_pgdir(&(dev->pdev->dev));
+	if (!pgdir) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	list_add(&pgdir->list, &dev->priv.pgdir_list);
+
+	/* This should never fail -- we just allocated an empty page: */
+	WARN_ON(mlx5_alloc_db_from_pgdir(pgdir, db));
+
+out:
+	mutex_unlock(&dev->priv.pgdir_mutex);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(mlx5_db_alloc);
+
+void mlx5_db_free(struct mlx5_core_dev *dev, struct mlx5_db *db)
+{
+	mutex_lock(&dev->priv.pgdir_mutex);
+
+	__set_bit(db->index, db->u.pgdir->bitmap);
+
+	if (bitmap_full(db->u.pgdir->bitmap, MLX5_DB_PER_PAGE)) {
+		dma_free_coherent(&(dev->pdev->dev), PAGE_SIZE,
+				  db->u.pgdir->db_page, db->u.pgdir->db_dma);
+		list_del(&db->u.pgdir->list);
+		kfree(db->u.pgdir);
+	}
+
+	mutex_unlock(&dev->priv.pgdir_mutex);
+}
+EXPORT_SYMBOL_GPL(mlx5_db_free);
+
+
+void mlx5_fill_page_array(struct mlx5_buf *buf, __be64 *pas)
+{
+	u64 addr;
+	int i;
+
+	for (i = 0; i < buf->npages; i++) {
+		if (buf->nbufs == 1)
+			addr = buf->direct.map + (i << buf->page_shift);
+		else
+			addr = buf->page_list[i].map;
+
+		pas[i] = cpu_to_be64(addr);
+	}
+}
+EXPORT_SYMBOL_GPL(mlx5_fill_page_array);

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

@@ -0,0 +1,224 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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/kernel.h>
+#include <linux/module.h>
+#include <linux/hardirq.h>
+#include <linux/mlx5/driver.h>
+#include <linux/mlx5/cmd.h>
+#include <rdma/ib_verbs.h>
+#include <linux/mlx5/cq.h>
+#include "mlx5_core.h"
+
+void mlx5_cq_completion(struct mlx5_core_dev *dev, u32 cqn)
+{
+	struct mlx5_core_cq *cq;
+	struct mlx5_cq_table *table = &dev->priv.cq_table;
+
+	spin_lock(&table->lock);
+	cq = radix_tree_lookup(&table->tree, cqn);
+	if (likely(cq))
+		atomic_inc(&cq->refcount);
+	spin_unlock(&table->lock);
+
+	if (!cq) {
+		mlx5_core_warn(dev, "Completion event for bogus CQ 0x%x\n", cqn);
+		return;
+	}
+
+	++cq->arm_sn;
+
+	cq->comp(cq);
+
+	if (atomic_dec_and_test(&cq->refcount))
+		complete(&cq->free);
+}
+
+void mlx5_cq_event(struct mlx5_core_dev *dev, u32 cqn, int event_type)
+{
+	struct mlx5_cq_table *table = &dev->priv.cq_table;
+	struct mlx5_core_cq *cq;
+
+	spin_lock(&table->lock);
+
+	cq = radix_tree_lookup(&table->tree, cqn);
+	if (cq)
+		atomic_inc(&cq->refcount);
+
+	spin_unlock(&table->lock);
+
+	if (!cq) {
+		mlx5_core_warn(dev, "Async event for bogus CQ 0x%x\n", cqn);
+		return;
+	}
+
+	cq->event(cq, event_type);
+
+	if (atomic_dec_and_test(&cq->refcount))
+		complete(&cq->free);
+}
+
+
+int mlx5_core_create_cq(struct mlx5_core_dev *dev, struct mlx5_core_cq *cq,
+			struct mlx5_create_cq_mbox_in *in, int inlen)
+{
+	int err;
+	struct mlx5_cq_table *table = &dev->priv.cq_table;
+	struct mlx5_create_cq_mbox_out out;
+	struct mlx5_destroy_cq_mbox_in din;
+	struct mlx5_destroy_cq_mbox_out dout;
+
+	in->hdr.opcode = cpu_to_be16(MLX5_CMD_OP_CREATE_CQ);
+	memset(&out, 0, sizeof(out));
+	err = mlx5_cmd_exec(dev, in, inlen, &out, sizeof(out));
+	if (err)
+		return err;
+
+	if (out.hdr.status)
+		return mlx5_cmd_status_to_err(&out.hdr);
+
+	cq->cqn = be32_to_cpu(out.cqn) & 0xffffff;
+	cq->cons_index = 0;
+	cq->arm_sn     = 0;
+	atomic_set(&cq->refcount, 1);
+	init_completion(&cq->free);
+
+	spin_lock_irq(&table->lock);
+	err = radix_tree_insert(&table->tree, cq->cqn, cq);
+	spin_unlock_irq(&table->lock);
+	if (err)
+		goto err_cmd;
+
+	cq->pid = current->pid;
+	err = mlx5_debug_cq_add(dev, cq);
+	if (err)
+		mlx5_core_dbg(dev, "failed adding CP 0x%x to debug file system\n",
+			      cq->cqn);
+
+	return 0;
+
+err_cmd:
+	memset(&din, 0, sizeof(din));
+	memset(&dout, 0, sizeof(dout));
+	din.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_DESTROY_CQ);
+	mlx5_cmd_exec(dev, &din, sizeof(din), &dout, sizeof(dout));
+	return err;
+}
+EXPORT_SYMBOL(mlx5_core_create_cq);
+
+int mlx5_core_destroy_cq(struct mlx5_core_dev *dev, struct mlx5_core_cq *cq)
+{
+	struct mlx5_cq_table *table = &dev->priv.cq_table;
+	struct mlx5_destroy_cq_mbox_in in;
+	struct mlx5_destroy_cq_mbox_out out;
+	struct mlx5_core_cq *tmp;
+	int err;
+
+	spin_lock_irq(&table->lock);
+	tmp = radix_tree_delete(&table->tree, cq->cqn);
+	spin_unlock_irq(&table->lock);
+	if (!tmp) {
+		mlx5_core_warn(dev, "cq 0x%x not found in tree\n", cq->cqn);
+		return -EINVAL;
+	}
+	if (tmp != cq) {
+		mlx5_core_warn(dev, "corruption on srqn 0x%x\n", cq->cqn);
+		return -EINVAL;
+	}
+
+	memset(&in, 0, sizeof(in));
+	memset(&out, 0, sizeof(out));
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_DESTROY_CQ);
+	in.cqn = cpu_to_be32(cq->cqn);
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), &out, sizeof(out));
+	if (err)
+		return err;
+
+	if (out.hdr.status)
+		return mlx5_cmd_status_to_err(&out.hdr);
+
+	synchronize_irq(cq->irqn);
+
+	mlx5_debug_cq_remove(dev, cq);
+	if (atomic_dec_and_test(&cq->refcount))
+		complete(&cq->free);
+	wait_for_completion(&cq->free);
+
+	return 0;
+}
+EXPORT_SYMBOL(mlx5_core_destroy_cq);
+
+int mlx5_core_query_cq(struct mlx5_core_dev *dev, struct mlx5_core_cq *cq,
+		       struct mlx5_query_cq_mbox_out *out)
+{
+	struct mlx5_query_cq_mbox_in in;
+	int err;
+
+	memset(&in, 0, sizeof(in));
+	memset(out, 0, sizeof(*out));
+
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_QUERY_CQ);
+	in.cqn = cpu_to_be32(cq->cqn);
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), out, sizeof(*out));
+	if (err)
+		return err;
+
+	if (out->hdr.status)
+		return mlx5_cmd_status_to_err(&out->hdr);
+
+	return err;
+}
+EXPORT_SYMBOL(mlx5_core_query_cq);
+
+
+int mlx5_core_modify_cq(struct mlx5_core_dev *dev, struct mlx5_core_cq *cq,
+			int type, struct mlx5_cq_modify_params *params)
+{
+	return -ENOSYS;
+}
+
+int mlx5_init_cq_table(struct mlx5_core_dev *dev)
+{
+	struct mlx5_cq_table *table = &dev->priv.cq_table;
+	int err;
+
+	spin_lock_init(&table->lock);
+	INIT_RADIX_TREE(&table->tree, GFP_ATOMIC);
+	err = mlx5_cq_debugfs_init(dev);
+
+	return err;
+}
+
+void mlx5_cleanup_cq_table(struct mlx5_core_dev *dev)
+{
+	mlx5_cq_debugfs_cleanup(dev);
+}

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

@@ -0,0 +1,587 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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/debugfs.h>
+#include <linux/mlx5/qp.h>
+#include <linux/mlx5/cq.h>
+#include <linux/mlx5/driver.h>
+#include "mlx5_core.h"
+
+enum {
+	QP_PID,
+	QP_STATE,
+	QP_XPORT,
+	QP_MTU,
+	QP_N_RECV,
+	QP_RECV_SZ,
+	QP_N_SEND,
+	QP_LOG_PG_SZ,
+	QP_RQPN,
+};
+
+static char *qp_fields[] = {
+	[QP_PID]	= "pid",
+	[QP_STATE]	= "state",
+	[QP_XPORT]	= "transport",
+	[QP_MTU]	= "mtu",
+	[QP_N_RECV]	= "num_recv",
+	[QP_RECV_SZ]	= "rcv_wqe_sz",
+	[QP_N_SEND]	= "num_send",
+	[QP_LOG_PG_SZ]	= "log2_page_sz",
+	[QP_RQPN]	= "remote_qpn",
+};
+
+enum {
+	EQ_NUM_EQES,
+	EQ_INTR,
+	EQ_LOG_PG_SZ,
+};
+
+static char *eq_fields[] = {
+	[EQ_NUM_EQES]	= "num_eqes",
+	[EQ_INTR]	= "intr",
+	[EQ_LOG_PG_SZ]	= "log_page_size",
+};
+
+enum {
+	CQ_PID,
+	CQ_NUM_CQES,
+	CQ_LOG_PG_SZ,
+};
+
+static char *cq_fields[] = {
+	[CQ_PID]	= "pid",
+	[CQ_NUM_CQES]	= "num_cqes",
+	[CQ_LOG_PG_SZ]	= "log_page_size",
+};
+
+struct dentry *mlx5_debugfs_root;
+EXPORT_SYMBOL(mlx5_debugfs_root);
+
+void mlx5_register_debugfs(void)
+{
+	mlx5_debugfs_root = debugfs_create_dir("mlx5", NULL);
+	if (IS_ERR_OR_NULL(mlx5_debugfs_root))
+		mlx5_debugfs_root = NULL;
+}
+
+void mlx5_unregister_debugfs(void)
+{
+	debugfs_remove(mlx5_debugfs_root);
+}
+
+int mlx5_qp_debugfs_init(struct mlx5_core_dev *dev)
+{
+	if (!mlx5_debugfs_root)
+		return 0;
+
+	atomic_set(&dev->num_qps, 0);
+
+	dev->priv.qp_debugfs = debugfs_create_dir("QPs",  dev->priv.dbg_root);
+	if (!dev->priv.qp_debugfs)
+		return -ENOMEM;
+
+	return 0;
+}
+
+void mlx5_qp_debugfs_cleanup(struct mlx5_core_dev *dev)
+{
+	if (!mlx5_debugfs_root)
+		return;
+
+	debugfs_remove_recursive(dev->priv.qp_debugfs);
+}
+
+int mlx5_eq_debugfs_init(struct mlx5_core_dev *dev)
+{
+	if (!mlx5_debugfs_root)
+		return 0;
+
+	dev->priv.eq_debugfs = debugfs_create_dir("EQs",  dev->priv.dbg_root);
+	if (!dev->priv.eq_debugfs)
+		return -ENOMEM;
+
+	return 0;
+}
+
+void mlx5_eq_debugfs_cleanup(struct mlx5_core_dev *dev)
+{
+	if (!mlx5_debugfs_root)
+		return;
+
+	debugfs_remove_recursive(dev->priv.eq_debugfs);
+}
+
+static ssize_t average_read(struct file *filp, char __user *buf, size_t count,
+			    loff_t *pos)
+{
+	struct mlx5_cmd_stats *stats;
+	u64 field = 0;
+	int ret;
+	int err;
+	char tbuf[22];
+
+	if (*pos)
+		return 0;
+
+	stats = filp->private_data;
+	spin_lock(&stats->lock);
+	if (stats->n)
+		field = stats->sum / stats->n;
+	spin_unlock(&stats->lock);
+	ret = snprintf(tbuf, sizeof(tbuf), "%llu\n", field);
+	if (ret > 0) {
+		err = copy_to_user(buf, tbuf, ret);
+		if (err)
+			return err;
+	}
+
+	*pos += ret;
+	return ret;
+}
+
+
+static ssize_t average_write(struct file *filp, const char __user *buf,
+			     size_t count, loff_t *pos)
+{
+	struct mlx5_cmd_stats *stats;
+
+	stats = filp->private_data;
+	spin_lock(&stats->lock);
+	stats->sum = 0;
+	stats->n = 0;
+	spin_unlock(&stats->lock);
+
+	*pos += count;
+
+	return count;
+}
+
+static const struct file_operations stats_fops = {
+	.owner	= THIS_MODULE,
+	.open	= simple_open,
+	.read	= average_read,
+	.write	= average_write,
+};
+
+int mlx5_cmdif_debugfs_init(struct mlx5_core_dev *dev)
+{
+	struct mlx5_cmd_stats *stats;
+	struct dentry **cmd;
+	const char *namep;
+	int err;
+	int i;
+
+	if (!mlx5_debugfs_root)
+		return 0;
+
+	cmd = &dev->priv.cmdif_debugfs;
+	*cmd = debugfs_create_dir("commands", dev->priv.dbg_root);
+	if (!*cmd)
+		return -ENOMEM;
+
+	for (i = 0; i < ARRAY_SIZE(dev->cmd.stats); i++) {
+		stats = &dev->cmd.stats[i];
+		namep = mlx5_command_str(i);
+		if (strcmp(namep, "unknown command opcode")) {
+			stats->root = debugfs_create_dir(namep, *cmd);
+			if (!stats->root) {
+				mlx5_core_warn(dev, "failed adding command %d\n",
+					       i);
+				err = -ENOMEM;
+				goto out;
+			}
+
+			stats->avg = debugfs_create_file("average", 0400,
+							 stats->root, stats,
+							 &stats_fops);
+			if (!stats->avg) {
+				mlx5_core_warn(dev, "failed creating debugfs file\n");
+				err = -ENOMEM;
+				goto out;
+			}
+
+			stats->count = debugfs_create_u64("n", 0400,
+							  stats->root,
+							  &stats->n);
+			if (!stats->count) {
+				mlx5_core_warn(dev, "failed creating debugfs file\n");
+				err = -ENOMEM;
+				goto out;
+			}
+		}
+	}
+
+	return 0;
+out:
+	debugfs_remove_recursive(dev->priv.cmdif_debugfs);
+	return err;
+}
+
+void mlx5_cmdif_debugfs_cleanup(struct mlx5_core_dev *dev)
+{
+	if (!mlx5_debugfs_root)
+		return;
+
+	debugfs_remove_recursive(dev->priv.cmdif_debugfs);
+}
+
+int mlx5_cq_debugfs_init(struct mlx5_core_dev *dev)
+{
+	if (!mlx5_debugfs_root)
+		return 0;
+
+	dev->priv.cq_debugfs = debugfs_create_dir("CQs",  dev->priv.dbg_root);
+	if (!dev->priv.cq_debugfs)
+		return -ENOMEM;
+
+	return 0;
+}
+
+void mlx5_cq_debugfs_cleanup(struct mlx5_core_dev *dev)
+{
+	if (!mlx5_debugfs_root)
+		return;
+
+	debugfs_remove_recursive(dev->priv.cq_debugfs);
+}
+
+static u64 qp_read_field(struct mlx5_core_dev *dev, struct mlx5_core_qp *qp,
+			 int index)
+{
+	struct mlx5_query_qp_mbox_out *out;
+	struct mlx5_qp_context *ctx;
+	u64 param = 0;
+	int err;
+	int no_sq;
+
+	out = kzalloc(sizeof(*out), GFP_KERNEL);
+	if (!out)
+		return param;
+
+	err = mlx5_core_qp_query(dev, qp, out, sizeof(*out));
+	if (err) {
+		mlx5_core_warn(dev, "failed to query qp\n");
+		goto out;
+	}
+
+	ctx = &out->ctx;
+	switch (index) {
+	case QP_PID:
+		param = qp->pid;
+		break;
+	case QP_STATE:
+		param = be32_to_cpu(ctx->flags) >> 28;
+		break;
+	case QP_XPORT:
+		param = (be32_to_cpu(ctx->flags) >> 16) & 0xff;
+		break;
+	case QP_MTU:
+		param = ctx->mtu_msgmax >> 5;
+		break;
+	case QP_N_RECV:
+		param = 1 << ((ctx->rq_size_stride >> 3) & 0xf);
+		break;
+	case QP_RECV_SZ:
+		param = 1 << ((ctx->rq_size_stride & 7) + 4);
+		break;
+	case QP_N_SEND:
+		no_sq = be16_to_cpu(ctx->sq_crq_size) >> 15;
+		if (!no_sq)
+			param = 1 << (be16_to_cpu(ctx->sq_crq_size) >> 11);
+		else
+			param = 0;
+		break;
+	case QP_LOG_PG_SZ:
+		param = (be32_to_cpu(ctx->log_pg_sz_remote_qpn) >> 24) & 0x1f;
+		param += 12;
+		break;
+	case QP_RQPN:
+		param = be32_to_cpu(ctx->log_pg_sz_remote_qpn) & 0xffffff;
+		break;
+	}
+
+out:
+	kfree(out);
+	return param;
+}
+
+static u64 eq_read_field(struct mlx5_core_dev *dev, struct mlx5_eq *eq,
+			 int index)
+{
+	struct mlx5_query_eq_mbox_out *out;
+	struct mlx5_eq_context *ctx;
+	u64 param = 0;
+	int err;
+
+	out = kzalloc(sizeof(*out), GFP_KERNEL);
+	if (!out)
+		return param;
+
+	ctx = &out->ctx;
+
+	err = mlx5_core_eq_query(dev, eq, out, sizeof(*out));
+	if (err) {
+		mlx5_core_warn(dev, "failed to query eq\n");
+		goto out;
+	}
+
+	switch (index) {
+	case EQ_NUM_EQES:
+		param = 1 << ((be32_to_cpu(ctx->log_sz_usr_page) >> 24) & 0x1f);
+		break;
+	case EQ_INTR:
+		param = ctx->intr;
+		break;
+	case EQ_LOG_PG_SZ:
+		param = (ctx->log_page_size & 0x1f) + 12;
+		break;
+	}
+
+out:
+	kfree(out);
+	return param;
+}
+
+static u64 cq_read_field(struct mlx5_core_dev *dev, struct mlx5_core_cq *cq,
+			 int index)
+{
+	struct mlx5_query_cq_mbox_out *out;
+	struct mlx5_cq_context *ctx;
+	u64 param = 0;
+	int err;
+
+	out = kzalloc(sizeof(*out), GFP_KERNEL);
+	if (!out)
+		return param;
+
+	ctx = &out->ctx;
+
+	err = mlx5_core_query_cq(dev, cq, out);
+	if (err) {
+		mlx5_core_warn(dev, "failed to query cq\n");
+		goto out;
+	}
+
+	switch (index) {
+	case CQ_PID:
+		param = cq->pid;
+		break;
+	case CQ_NUM_CQES:
+		param = 1 << ((be32_to_cpu(ctx->log_sz_usr_page) >> 24) & 0x1f);
+		break;
+	case CQ_LOG_PG_SZ:
+		param = (ctx->log_pg_sz & 0x1f) + 12;
+		break;
+	}
+
+out:
+	kfree(out);
+	return param;
+}
+
+static ssize_t dbg_read(struct file *filp, char __user *buf, size_t count,
+			loff_t *pos)
+{
+	struct mlx5_field_desc *desc;
+	struct mlx5_rsc_debug *d;
+	char tbuf[18];
+	u64 field;
+	int ret;
+	int err;
+
+	if (*pos)
+		return 0;
+
+	desc = filp->private_data;
+	d = (void *)(desc - desc->i) - sizeof(*d);
+	switch (d->type) {
+	case MLX5_DBG_RSC_QP:
+		field = qp_read_field(d->dev, d->object, desc->i);
+		break;
+
+	case MLX5_DBG_RSC_EQ:
+		field = eq_read_field(d->dev, d->object, desc->i);
+		break;
+
+	case MLX5_DBG_RSC_CQ:
+		field = cq_read_field(d->dev, d->object, desc->i);
+		break;
+
+	default:
+		mlx5_core_warn(d->dev, "invalid resource type %d\n", d->type);
+		return -EINVAL;
+	}
+
+	ret = snprintf(tbuf, sizeof(tbuf), "0x%llx\n", field);
+	if (ret > 0) {
+		err = copy_to_user(buf, tbuf, ret);
+		if (err)
+			return err;
+	}
+
+	*pos += ret;
+	return ret;
+}
+
+static const struct file_operations fops = {
+	.owner	= THIS_MODULE,
+	.open	= simple_open,
+	.read	= dbg_read,
+};
+
+static int add_res_tree(struct mlx5_core_dev *dev, enum dbg_rsc_type type,
+			struct dentry *root, struct mlx5_rsc_debug **dbg,
+			int rsn, char **field, int nfile, void *data)
+{
+	struct mlx5_rsc_debug *d;
+	char resn[32];
+	int err;
+	int i;
+
+	d = kzalloc(sizeof(*d) + nfile * sizeof(d->fields[0]), GFP_KERNEL);
+	if (!d)
+		return -ENOMEM;
+
+	d->dev = dev;
+	d->object = data;
+	d->type = type;
+	sprintf(resn, "0x%x", rsn);
+	d->root = debugfs_create_dir(resn,  root);
+	if (!d->root) {
+		err = -ENOMEM;
+		goto out_free;
+	}
+
+	for (i = 0; i < nfile; i++) {
+		d->fields[i].i = i;
+		d->fields[i].dent = debugfs_create_file(field[i], 0400,
+							d->root, &d->fields[i],
+							&fops);
+		if (!d->fields[i].dent) {
+			err = -ENOMEM;
+			goto out_rem;
+		}
+	}
+	*dbg = d;
+
+	return 0;
+out_rem:
+	debugfs_remove_recursive(d->root);
+
+out_free:
+	kfree(d);
+	return err;
+}
+
+static void rem_res_tree(struct mlx5_rsc_debug *d)
+{
+	debugfs_remove_recursive(d->root);
+	kfree(d);
+}
+
+int mlx5_debug_qp_add(struct mlx5_core_dev *dev, struct mlx5_core_qp *qp)
+{
+	int err;
+
+	if (!mlx5_debugfs_root)
+		return 0;
+
+	err = add_res_tree(dev, MLX5_DBG_RSC_QP, dev->priv.qp_debugfs,
+			   &qp->dbg, qp->qpn, qp_fields,
+			   ARRAY_SIZE(qp_fields), qp);
+	if (err)
+		qp->dbg = NULL;
+
+	return err;
+}
+
+void mlx5_debug_qp_remove(struct mlx5_core_dev *dev, struct mlx5_core_qp *qp)
+{
+	if (!mlx5_debugfs_root)
+		return;
+
+	if (qp->dbg)
+		rem_res_tree(qp->dbg);
+}
+
+
+int mlx5_debug_eq_add(struct mlx5_core_dev *dev, struct mlx5_eq *eq)
+{
+	int err;
+
+	if (!mlx5_debugfs_root)
+		return 0;
+
+	err = add_res_tree(dev, MLX5_DBG_RSC_EQ, dev->priv.eq_debugfs,
+			   &eq->dbg, eq->eqn, eq_fields,
+			   ARRAY_SIZE(eq_fields), eq);
+	if (err)
+		eq->dbg = NULL;
+
+	return err;
+}
+
+void mlx5_debug_eq_remove(struct mlx5_core_dev *dev, struct mlx5_eq *eq)
+{
+	if (!mlx5_debugfs_root)
+		return;
+
+	if (eq->dbg)
+		rem_res_tree(eq->dbg);
+}
+
+int mlx5_debug_cq_add(struct mlx5_core_dev *dev, struct mlx5_core_cq *cq)
+{
+	int err;
+
+	if (!mlx5_debugfs_root)
+		return 0;
+
+	err = add_res_tree(dev, MLX5_DBG_RSC_CQ, dev->priv.cq_debugfs,
+			   &cq->dbg, cq->cqn, cq_fields,
+			   ARRAY_SIZE(cq_fields), cq);
+	if (err)
+		cq->dbg = NULL;
+
+	return err;
+}
+
+void mlx5_debug_cq_remove(struct mlx5_core_dev *dev, struct mlx5_core_cq *cq)
+{
+	if (!mlx5_debugfs_root)
+		return;
+
+	if (cq->dbg)
+		rem_res_tree(cq->dbg);
+}

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

@@ -0,0 +1,521 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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/interrupt.h>
+#include <linux/module.h>
+#include <linux/mlx5/driver.h>
+#include <linux/mlx5/cmd.h>
+#include "mlx5_core.h"
+
+enum {
+	MLX5_EQE_SIZE		= sizeof(struct mlx5_eqe),
+	MLX5_EQE_OWNER_INIT_VAL	= 0x1,
+};
+
+enum {
+	MLX5_EQ_STATE_ARMED		= 0x9,
+	MLX5_EQ_STATE_FIRED		= 0xa,
+	MLX5_EQ_STATE_ALWAYS_ARMED	= 0xb,
+};
+
+enum {
+	MLX5_NUM_SPARE_EQE	= 0x80,
+	MLX5_NUM_ASYNC_EQE	= 0x100,
+	MLX5_NUM_CMD_EQE	= 32,
+};
+
+enum {
+	MLX5_EQ_DOORBEL_OFFSET	= 0x40,
+};
+
+#define MLX5_ASYNC_EVENT_MASK ((1ull << MLX5_EVENT_TYPE_PATH_MIG)	    | \
+			       (1ull << MLX5_EVENT_TYPE_COMM_EST)	    | \
+			       (1ull << MLX5_EVENT_TYPE_SQ_DRAINED)	    | \
+			       (1ull << MLX5_EVENT_TYPE_CQ_ERROR)	    | \
+			       (1ull << MLX5_EVENT_TYPE_WQ_CATAS_ERROR)	    | \
+			       (1ull << MLX5_EVENT_TYPE_PATH_MIG_FAILED)    | \
+			       (1ull << MLX5_EVENT_TYPE_WQ_INVAL_REQ_ERROR) | \
+			       (1ull << MLX5_EVENT_TYPE_WQ_ACCESS_ERROR)    | \
+			       (1ull << MLX5_EVENT_TYPE_PORT_CHANGE)	    | \
+			       (1ull << MLX5_EVENT_TYPE_SRQ_CATAS_ERROR)    | \
+			       (1ull << MLX5_EVENT_TYPE_SRQ_LAST_WQE)	    | \
+			       (1ull << MLX5_EVENT_TYPE_SRQ_RQ_LIMIT))
+
+struct map_eq_in {
+	u64	mask;
+	u32	reserved;
+	u32	unmap_eqn;
+};
+
+struct cre_des_eq {
+	u8	reserved[15];
+	u8	eqn;
+};
+
+static int mlx5_cmd_destroy_eq(struct mlx5_core_dev *dev, u8 eqn)
+{
+	struct mlx5_destroy_eq_mbox_in in;
+	struct mlx5_destroy_eq_mbox_out out;
+	int err;
+
+	memset(&in, 0, sizeof(in));
+	memset(&out, 0, sizeof(out));
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_DESTROY_EQ);
+	in.eqn = eqn;
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), &out, sizeof(out));
+	if (!err)
+		goto ex;
+
+	if (out.hdr.status)
+		err = mlx5_cmd_status_to_err(&out.hdr);
+
+ex:
+	return err;
+}
+
+static struct mlx5_eqe *get_eqe(struct mlx5_eq *eq, u32 entry)
+{
+	return mlx5_buf_offset(&eq->buf, entry * MLX5_EQE_SIZE);
+}
+
+static struct mlx5_eqe *next_eqe_sw(struct mlx5_eq *eq)
+{
+	struct mlx5_eqe *eqe = get_eqe(eq, eq->cons_index & (eq->nent - 1));
+
+	return ((eqe->owner & 1) ^ !!(eq->cons_index & eq->nent)) ? NULL : eqe;
+}
+
+static const char *eqe_type_str(u8 type)
+{
+	switch (type) {
+	case MLX5_EVENT_TYPE_COMP:
+		return "MLX5_EVENT_TYPE_COMP";
+	case MLX5_EVENT_TYPE_PATH_MIG:
+		return "MLX5_EVENT_TYPE_PATH_MIG";
+	case MLX5_EVENT_TYPE_COMM_EST:
+		return "MLX5_EVENT_TYPE_COMM_EST";
+	case MLX5_EVENT_TYPE_SQ_DRAINED:
+		return "MLX5_EVENT_TYPE_SQ_DRAINED";
+	case MLX5_EVENT_TYPE_SRQ_LAST_WQE:
+		return "MLX5_EVENT_TYPE_SRQ_LAST_WQE";
+	case MLX5_EVENT_TYPE_SRQ_RQ_LIMIT:
+		return "MLX5_EVENT_TYPE_SRQ_RQ_LIMIT";
+	case MLX5_EVENT_TYPE_CQ_ERROR:
+		return "MLX5_EVENT_TYPE_CQ_ERROR";
+	case MLX5_EVENT_TYPE_WQ_CATAS_ERROR:
+		return "MLX5_EVENT_TYPE_WQ_CATAS_ERROR";
+	case MLX5_EVENT_TYPE_PATH_MIG_FAILED:
+		return "MLX5_EVENT_TYPE_PATH_MIG_FAILED";
+	case MLX5_EVENT_TYPE_WQ_INVAL_REQ_ERROR:
+		return "MLX5_EVENT_TYPE_WQ_INVAL_REQ_ERROR";
+	case MLX5_EVENT_TYPE_WQ_ACCESS_ERROR:
+		return "MLX5_EVENT_TYPE_WQ_ACCESS_ERROR";
+	case MLX5_EVENT_TYPE_SRQ_CATAS_ERROR:
+		return "MLX5_EVENT_TYPE_SRQ_CATAS_ERROR";
+	case MLX5_EVENT_TYPE_INTERNAL_ERROR:
+		return "MLX5_EVENT_TYPE_INTERNAL_ERROR";
+	case MLX5_EVENT_TYPE_PORT_CHANGE:
+		return "MLX5_EVENT_TYPE_PORT_CHANGE";
+	case MLX5_EVENT_TYPE_GPIO_EVENT:
+		return "MLX5_EVENT_TYPE_GPIO_EVENT";
+	case MLX5_EVENT_TYPE_REMOTE_CONFIG:
+		return "MLX5_EVENT_TYPE_REMOTE_CONFIG";
+	case MLX5_EVENT_TYPE_DB_BF_CONGESTION:
+		return "MLX5_EVENT_TYPE_DB_BF_CONGESTION";
+	case MLX5_EVENT_TYPE_STALL_EVENT:
+		return "MLX5_EVENT_TYPE_STALL_EVENT";
+	case MLX5_EVENT_TYPE_CMD:
+		return "MLX5_EVENT_TYPE_CMD";
+	case MLX5_EVENT_TYPE_PAGE_REQUEST:
+		return "MLX5_EVENT_TYPE_PAGE_REQUEST";
+	default:
+		return "Unrecognized event";
+	}
+}
+
+static enum mlx5_dev_event port_subtype_event(u8 subtype)
+{
+	switch (subtype) {
+	case MLX5_PORT_CHANGE_SUBTYPE_DOWN:
+		return MLX5_DEV_EVENT_PORT_DOWN;
+	case MLX5_PORT_CHANGE_SUBTYPE_ACTIVE:
+		return MLX5_DEV_EVENT_PORT_UP;
+	case MLX5_PORT_CHANGE_SUBTYPE_INITIALIZED:
+		return MLX5_DEV_EVENT_PORT_INITIALIZED;
+	case MLX5_PORT_CHANGE_SUBTYPE_LID:
+		return MLX5_DEV_EVENT_LID_CHANGE;
+	case MLX5_PORT_CHANGE_SUBTYPE_PKEY:
+		return MLX5_DEV_EVENT_PKEY_CHANGE;
+	case MLX5_PORT_CHANGE_SUBTYPE_GUID:
+		return MLX5_DEV_EVENT_GUID_CHANGE;
+	case MLX5_PORT_CHANGE_SUBTYPE_CLIENT_REREG:
+		return MLX5_DEV_EVENT_CLIENT_REREG;
+	}
+	return -1;
+}
+
+static void eq_update_ci(struct mlx5_eq *eq, int arm)
+{
+	__be32 __iomem *addr = eq->doorbell + (arm ? 0 : 2);
+	u32 val = (eq->cons_index & 0xffffff) | (eq->eqn << 24);
+	__raw_writel((__force u32) cpu_to_be32(val), addr);
+	/* We still want ordering, just not swabbing, so add a barrier */
+	mb();
+}
+
+static int mlx5_eq_int(struct mlx5_core_dev *dev, struct mlx5_eq *eq)
+{
+	struct mlx5_eqe *eqe;
+	int eqes_found = 0;
+	int set_ci = 0;
+	u32 cqn;
+	u32 srqn;
+	u8 port;
+
+	while ((eqe = next_eqe_sw(eq))) {
+		/*
+		 * Make sure we read EQ entry contents after we've
+		 * checked the ownership bit.
+		 */
+		rmb();
+
+		mlx5_core_dbg(eq->dev, "eqn %d, eqe type %s\n", eq->eqn, eqe_type_str(eqe->type));
+		switch (eqe->type) {
+		case MLX5_EVENT_TYPE_COMP:
+			cqn = be32_to_cpu(eqe->data.comp.cqn) & 0xffffff;
+			mlx5_cq_completion(dev, cqn);
+			break;
+
+		case MLX5_EVENT_TYPE_PATH_MIG:
+		case MLX5_EVENT_TYPE_COMM_EST:
+		case MLX5_EVENT_TYPE_SQ_DRAINED:
+		case MLX5_EVENT_TYPE_SRQ_LAST_WQE:
+		case MLX5_EVENT_TYPE_WQ_CATAS_ERROR:
+		case MLX5_EVENT_TYPE_PATH_MIG_FAILED:
+		case MLX5_EVENT_TYPE_WQ_INVAL_REQ_ERROR:
+		case MLX5_EVENT_TYPE_WQ_ACCESS_ERROR:
+			mlx5_core_dbg(dev, "event %s(%d) arrived\n",
+				      eqe_type_str(eqe->type), eqe->type);
+			mlx5_qp_event(dev, be32_to_cpu(eqe->data.qp_srq.qp_srq_n) & 0xffffff,
+				      eqe->type);
+			break;
+
+		case MLX5_EVENT_TYPE_SRQ_RQ_LIMIT:
+		case MLX5_EVENT_TYPE_SRQ_CATAS_ERROR:
+			srqn = be32_to_cpu(eqe->data.qp_srq.qp_srq_n) & 0xffffff;
+			mlx5_core_dbg(dev, "SRQ event %s(%d): srqn 0x%x\n",
+				      eqe_type_str(eqe->type), eqe->type, srqn);
+			mlx5_srq_event(dev, srqn, eqe->type);
+			break;
+
+		case MLX5_EVENT_TYPE_CMD:
+			mlx5_cmd_comp_handler(dev, be32_to_cpu(eqe->data.cmd.vector));
+			break;
+
+		case MLX5_EVENT_TYPE_PORT_CHANGE:
+			port = (eqe->data.port.port >> 4) & 0xf;
+			switch (eqe->sub_type) {
+			case MLX5_PORT_CHANGE_SUBTYPE_DOWN:
+			case MLX5_PORT_CHANGE_SUBTYPE_ACTIVE:
+			case MLX5_PORT_CHANGE_SUBTYPE_LID:
+			case MLX5_PORT_CHANGE_SUBTYPE_PKEY:
+			case MLX5_PORT_CHANGE_SUBTYPE_GUID:
+			case MLX5_PORT_CHANGE_SUBTYPE_CLIENT_REREG:
+			case MLX5_PORT_CHANGE_SUBTYPE_INITIALIZED:
+				dev->event(dev, port_subtype_event(eqe->sub_type), &port);
+				break;
+			default:
+				mlx5_core_warn(dev, "Port event with unrecognized subtype: port %d, sub_type %d\n",
+					       port, eqe->sub_type);
+			}
+			break;
+		case MLX5_EVENT_TYPE_CQ_ERROR:
+			cqn = be32_to_cpu(eqe->data.cq_err.cqn) & 0xffffff;
+			mlx5_core_warn(dev, "CQ error on CQN 0x%x, syndrom 0x%x\n",
+				       cqn, eqe->data.cq_err.syndrome);
+			mlx5_cq_event(dev, cqn, eqe->type);
+			break;
+
+		case MLX5_EVENT_TYPE_PAGE_REQUEST:
+			{
+				u16 func_id = be16_to_cpu(eqe->data.req_pages.func_id);
+				s16 npages = be16_to_cpu(eqe->data.req_pages.num_pages);
+
+				mlx5_core_dbg(dev, "page request for func 0x%x, napges %d\n", func_id, npages);
+				mlx5_core_req_pages_handler(dev, func_id, npages);
+			}
+			break;
+
+
+		default:
+			mlx5_core_warn(dev, "Unhandled event 0x%x on EQ 0x%x\n", eqe->type, eq->eqn);
+			break;
+		}
+
+		++eq->cons_index;
+		eqes_found = 1;
+		++set_ci;
+
+		/* The HCA will think the queue has overflowed if we
+		 * don't tell it we've been processing events.  We
+		 * create our EQs with MLX5_NUM_SPARE_EQE extra
+		 * entries, so we must update our consumer index at
+		 * least that often.
+		 */
+		if (unlikely(set_ci >= MLX5_NUM_SPARE_EQE)) {
+			eq_update_ci(eq, 0);
+			set_ci = 0;
+		}
+	}
+
+	eq_update_ci(eq, 1);
+
+	return eqes_found;
+}
+
+static irqreturn_t mlx5_msix_handler(int irq, void *eq_ptr)
+{
+	struct mlx5_eq *eq = eq_ptr;
+	struct mlx5_core_dev *dev = eq->dev;
+
+	mlx5_eq_int(dev, eq);
+
+	/* MSI-X vectors always belong to us */
+	return IRQ_HANDLED;
+}
+
+static void init_eq_buf(struct mlx5_eq *eq)
+{
+	struct mlx5_eqe *eqe;
+	int i;
+
+	for (i = 0; i < eq->nent; i++) {
+		eqe = get_eqe(eq, i);
+		eqe->owner = MLX5_EQE_OWNER_INIT_VAL;
+	}
+}
+
+int mlx5_create_map_eq(struct mlx5_core_dev *dev, struct mlx5_eq *eq, u8 vecidx,
+		       int nent, u64 mask, const char *name, struct mlx5_uar *uar)
+{
+	struct mlx5_eq_table *table = &dev->priv.eq_table;
+	struct mlx5_create_eq_mbox_in *in;
+	struct mlx5_create_eq_mbox_out out;
+	int err;
+	int inlen;
+
+	eq->nent = roundup_pow_of_two(nent + MLX5_NUM_SPARE_EQE);
+	err = mlx5_buf_alloc(dev, eq->nent * MLX5_EQE_SIZE, 2 * PAGE_SIZE,
+			     &eq->buf);
+	if (err)
+		return err;
+
+	init_eq_buf(eq);
+
+	inlen = sizeof(*in) + sizeof(in->pas[0]) * eq->buf.npages;
+	in = mlx5_vzalloc(inlen);
+	if (!in) {
+		err = -ENOMEM;
+		goto err_buf;
+	}
+	memset(&out, 0, sizeof(out));
+
+	mlx5_fill_page_array(&eq->buf, in->pas);
+
+	in->hdr.opcode = cpu_to_be16(MLX5_CMD_OP_CREATE_EQ);
+	in->ctx.log_sz_usr_page = cpu_to_be32(ilog2(eq->nent) << 24 | uar->index);
+	in->ctx.intr = vecidx;
+	in->ctx.log_page_size = PAGE_SHIFT - 12;
+	in->events_mask = cpu_to_be64(mask);
+
+	err = mlx5_cmd_exec(dev, in, inlen, &out, sizeof(out));
+	if (err)
+		goto err_in;
+
+	if (out.hdr.status) {
+		err = mlx5_cmd_status_to_err(&out.hdr);
+		goto err_in;
+	}
+
+	eq->eqn = out.eq_number;
+	err = request_irq(table->msix_arr[vecidx].vector, mlx5_msix_handler, 0,
+			  name, eq);
+	if (err)
+		goto err_eq;
+
+	eq->irqn = vecidx;
+	eq->dev = dev;
+	eq->doorbell = uar->map + MLX5_EQ_DOORBEL_OFFSET;
+
+	err = mlx5_debug_eq_add(dev, eq);
+	if (err)
+		goto err_irq;
+
+	/* EQs are created in ARMED state
+	 */
+	eq_update_ci(eq, 1);
+
+	mlx5_vfree(in);
+	return 0;
+
+err_irq:
+	free_irq(table->msix_arr[vecidx].vector, eq);
+
+err_eq:
+	mlx5_cmd_destroy_eq(dev, eq->eqn);
+
+err_in:
+	mlx5_vfree(in);
+
+err_buf:
+	mlx5_buf_free(dev, &eq->buf);
+	return err;
+}
+EXPORT_SYMBOL_GPL(mlx5_create_map_eq);
+
+int mlx5_destroy_unmap_eq(struct mlx5_core_dev *dev, struct mlx5_eq *eq)
+{
+	struct mlx5_eq_table *table = &dev->priv.eq_table;
+	int err;
+
+	mlx5_debug_eq_remove(dev, eq);
+	free_irq(table->msix_arr[eq->irqn].vector, eq);
+	err = mlx5_cmd_destroy_eq(dev, eq->eqn);
+	if (err)
+		mlx5_core_warn(dev, "failed to destroy a previously created eq: eqn %d\n",
+			       eq->eqn);
+	mlx5_buf_free(dev, &eq->buf);
+
+	return err;
+}
+EXPORT_SYMBOL_GPL(mlx5_destroy_unmap_eq);
+
+int mlx5_eq_init(struct mlx5_core_dev *dev)
+{
+	int err;
+
+	spin_lock_init(&dev->priv.eq_table.lock);
+
+	err = mlx5_eq_debugfs_init(dev);
+
+	return err;
+}
+
+
+void mlx5_eq_cleanup(struct mlx5_core_dev *dev)
+{
+	mlx5_eq_debugfs_cleanup(dev);
+}
+
+int mlx5_start_eqs(struct mlx5_core_dev *dev)
+{
+	struct mlx5_eq_table *table = &dev->priv.eq_table;
+	int err;
+
+	err = mlx5_create_map_eq(dev, &table->cmd_eq, MLX5_EQ_VEC_CMD,
+				 MLX5_NUM_CMD_EQE, 1ull << MLX5_EVENT_TYPE_CMD,
+				 "mlx5_cmd_eq", &dev->priv.uuari.uars[0]);
+	if (err) {
+		mlx5_core_warn(dev, "failed to create cmd EQ %d\n", err);
+		return err;
+	}
+
+	mlx5_cmd_use_events(dev);
+
+	err = mlx5_create_map_eq(dev, &table->async_eq, MLX5_EQ_VEC_ASYNC,
+				 MLX5_NUM_ASYNC_EQE, MLX5_ASYNC_EVENT_MASK,
+				 "mlx5_async_eq", &dev->priv.uuari.uars[0]);
+	if (err) {
+		mlx5_core_warn(dev, "failed to create async EQ %d\n", err);
+		goto err1;
+	}
+
+	err = mlx5_create_map_eq(dev, &table->pages_eq,
+				 MLX5_EQ_VEC_PAGES,
+				 dev->caps.max_vf + 1,
+				 1 << MLX5_EVENT_TYPE_PAGE_REQUEST, "mlx5_pages_eq",
+				 &dev->priv.uuari.uars[0]);
+	if (err) {
+		mlx5_core_warn(dev, "failed to create pages EQ %d\n", err);
+		goto err2;
+	}
+
+	return err;
+
+err2:
+	mlx5_destroy_unmap_eq(dev, &table->async_eq);
+
+err1:
+	mlx5_cmd_use_polling(dev);
+	mlx5_destroy_unmap_eq(dev, &table->cmd_eq);
+	return err;
+}
+
+int mlx5_stop_eqs(struct mlx5_core_dev *dev)
+{
+	struct mlx5_eq_table *table = &dev->priv.eq_table;
+	int err;
+
+	err = mlx5_destroy_unmap_eq(dev, &table->pages_eq);
+	if (err)
+		return err;
+
+	mlx5_destroy_unmap_eq(dev, &table->async_eq);
+	mlx5_cmd_use_polling(dev);
+
+	err = mlx5_destroy_unmap_eq(dev, &table->cmd_eq);
+	if (err)
+		mlx5_cmd_use_events(dev);
+
+	return err;
+}
+
+int mlx5_core_eq_query(struct mlx5_core_dev *dev, struct mlx5_eq *eq,
+		       struct mlx5_query_eq_mbox_out *out, int outlen)
+{
+	struct mlx5_query_eq_mbox_in in;
+	int err;
+
+	memset(&in, 0, sizeof(in));
+	memset(out, 0, outlen);
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_QUERY_EQ);
+	in.eqn = eq->eqn;
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), out, outlen);
+	if (err)
+		return err;
+
+	if (out->hdr.status)
+		err = mlx5_cmd_status_to_err(&out->hdr);
+
+	return err;
+}
+EXPORT_SYMBOL_GPL(mlx5_core_eq_query);

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

@@ -0,0 +1,185 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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/driver.h>
+#include <linux/mlx5/cmd.h>
+#include <linux/module.h>
+#include "mlx5_core.h"
+
+int mlx5_cmd_query_adapter(struct mlx5_core_dev *dev)
+{
+	struct mlx5_cmd_query_adapter_mbox_out *out;
+	struct mlx5_cmd_query_adapter_mbox_in in;
+	int err;
+
+	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_ADAPTER);
+	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;
+	}
+
+	memcpy(dev->board_id, out->vsd_psid, sizeof(out->vsd_psid));
+
+out_out:
+	kfree(out);
+
+	return err;
+}
+
+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 = be16_to_cpu(out->hca_cap.max_qp_mcg);
+	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;
+}
+
+int mlx5_cmd_init_hca(struct mlx5_core_dev *dev)
+{
+	struct mlx5_cmd_init_hca_mbox_in in;
+	struct mlx5_cmd_init_hca_mbox_out out;
+	int err;
+
+	memset(&in, 0, sizeof(in));
+	memset(&out, 0, sizeof(out));
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_INIT_HCA);
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), &out, sizeof(out));
+	if (err)
+		return err;
+
+	if (out.hdr.status)
+		err = mlx5_cmd_status_to_err(&out.hdr);
+
+	return err;
+}
+
+int mlx5_cmd_teardown_hca(struct mlx5_core_dev *dev)
+{
+	struct mlx5_cmd_teardown_hca_mbox_in in;
+	struct mlx5_cmd_teardown_hca_mbox_out out;
+	int err;
+
+	memset(&in, 0, sizeof(in));
+	memset(&out, 0, sizeof(out));
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_TEARDOWN_HCA);
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), &out, sizeof(out));
+	if (err)
+		return err;
+
+	if (out.hdr.status)
+		err = mlx5_cmd_status_to_err(&out.hdr);
+
+	return err;
+}

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

@@ -0,0 +1,227 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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/kernel.h>
+#include <linux/module.h>
+#include <linux/random.h>
+#include <linux/vmalloc.h>
+#include <linux/mlx5/driver.h>
+#include <linux/mlx5/cmd.h>
+#include "mlx5_core.h"
+
+enum {
+	MLX5_HEALTH_POLL_INTERVAL	= 2 * HZ,
+	MAX_MISSES			= 3,
+};
+
+enum {
+	MLX5_HEALTH_SYNDR_FW_ERR		= 0x1,
+	MLX5_HEALTH_SYNDR_IRISC_ERR		= 0x7,
+	MLX5_HEALTH_SYNDR_CRC_ERR		= 0x9,
+	MLX5_HEALTH_SYNDR_FETCH_PCI_ERR		= 0xa,
+	MLX5_HEALTH_SYNDR_HW_FTL_ERR		= 0xb,
+	MLX5_HEALTH_SYNDR_ASYNC_EQ_OVERRUN_ERR	= 0xc,
+	MLX5_HEALTH_SYNDR_EQ_ERR		= 0xd,
+	MLX5_HEALTH_SYNDR_FFSER_ERR		= 0xf,
+};
+
+static DEFINE_SPINLOCK(health_lock);
+
+static LIST_HEAD(health_list);
+static struct work_struct health_work;
+
+static health_handler_t reg_handler;
+int mlx5_register_health_report_handler(health_handler_t handler)
+{
+	spin_lock_irq(&health_lock);
+	if (reg_handler) {
+		spin_unlock_irq(&health_lock);
+		return -EEXIST;
+	}
+	reg_handler = handler;
+	spin_unlock_irq(&health_lock);
+
+	return 0;
+}
+EXPORT_SYMBOL(mlx5_register_health_report_handler);
+
+void mlx5_unregister_health_report_handler(void)
+{
+	spin_lock_irq(&health_lock);
+	reg_handler = NULL;
+	spin_unlock_irq(&health_lock);
+}
+EXPORT_SYMBOL(mlx5_unregister_health_report_handler);
+
+static void health_care(struct work_struct *work)
+{
+	struct mlx5_core_health *health, *n;
+	struct mlx5_core_dev *dev;
+	struct mlx5_priv *priv;
+	LIST_HEAD(tlist);
+
+	spin_lock_irq(&health_lock);
+	list_splice_init(&health_list, &tlist);
+
+	spin_unlock_irq(&health_lock);
+
+	list_for_each_entry_safe(health, n, &tlist, list) {
+		priv = container_of(health, struct mlx5_priv, health);
+		dev = container_of(priv, struct mlx5_core_dev, priv);
+		mlx5_core_warn(dev, "handling bad device here\n");
+		spin_lock_irq(&health_lock);
+		if (reg_handler)
+			reg_handler(dev->pdev, health->health,
+				    sizeof(health->health));
+
+		list_del_init(&health->list);
+		spin_unlock_irq(&health_lock);
+	}
+}
+
+static const char *hsynd_str(u8 synd)
+{
+	switch (synd) {
+	case MLX5_HEALTH_SYNDR_FW_ERR:
+		return "firmware internal error";
+	case MLX5_HEALTH_SYNDR_IRISC_ERR:
+		return "irisc not responding";
+	case MLX5_HEALTH_SYNDR_CRC_ERR:
+		return "firmware CRC error";
+	case MLX5_HEALTH_SYNDR_FETCH_PCI_ERR:
+		return "ICM fetch PCI error";
+	case MLX5_HEALTH_SYNDR_HW_FTL_ERR:
+		return "HW fatal error\n";
+	case MLX5_HEALTH_SYNDR_ASYNC_EQ_OVERRUN_ERR:
+		return "async EQ buffer overrun";
+	case MLX5_HEALTH_SYNDR_EQ_ERR:
+		return "EQ error";
+	case MLX5_HEALTH_SYNDR_FFSER_ERR:
+		return "FFSER error";
+	default:
+		return "unrecognized error";
+	}
+}
+
+static u16 read_be16(__be16 __iomem *p)
+{
+	return swab16(readl((__force u16 __iomem *) p));
+}
+
+static u32 read_be32(__be32 __iomem *p)
+{
+	return swab32(readl((__force u32 __iomem *) p));
+}
+
+static void print_health_info(struct mlx5_core_dev *dev)
+{
+	struct mlx5_core_health *health = &dev->priv.health;
+	struct health_buffer __iomem *h = health->health;
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(h->assert_var); i++)
+		pr_info("assert_var[%d] 0x%08x\n", i, read_be32(h->assert_var + i));
+
+	pr_info("assert_exit_ptr 0x%08x\n", read_be32(&h->assert_exit_ptr));
+	pr_info("assert_callra 0x%08x\n", read_be32(&h->assert_callra));
+	pr_info("fw_ver 0x%08x\n", read_be32(&h->fw_ver));
+	pr_info("hw_id 0x%08x\n", read_be32(&h->hw_id));
+	pr_info("irisc_index %d\n", readb(&h->irisc_index));
+	pr_info("synd 0x%x: %s\n", readb(&h->synd), hsynd_str(readb(&h->synd)));
+	pr_info("ext_sync 0x%04x\n", read_be16(&h->ext_sync));
+}
+
+static void poll_health(unsigned long data)
+{
+	struct mlx5_core_dev *dev = (struct mlx5_core_dev *)data;
+	struct mlx5_core_health *health = &dev->priv.health;
+	unsigned long next;
+	u32 count;
+
+	count = ioread32be(health->health_counter);
+	if (count == health->prev)
+		++health->miss_counter;
+	else
+		health->miss_counter = 0;
+
+	health->prev = count;
+	if (health->miss_counter == MAX_MISSES) {
+		mlx5_core_err(dev, "device's health compromised\n");
+		print_health_info(dev);
+		spin_lock_irq(&health_lock);
+		list_add_tail(&health->list, &health_list);
+		spin_unlock_irq(&health_lock);
+
+		queue_work(mlx5_core_wq, &health_work);
+	} else {
+		get_random_bytes(&next, sizeof(next));
+		next %= HZ;
+		next += jiffies + MLX5_HEALTH_POLL_INTERVAL;
+		mod_timer(&health->timer, next);
+	}
+}
+
+void mlx5_start_health_poll(struct mlx5_core_dev *dev)
+{
+	struct mlx5_core_health *health = &dev->priv.health;
+
+	INIT_LIST_HEAD(&health->list);
+	init_timer(&health->timer);
+	health->health = &dev->iseg->health;
+	health->health_counter = &dev->iseg->health_counter;
+
+	health->timer.data = (unsigned long)dev;
+	health->timer.function = poll_health;
+	health->timer.expires = round_jiffies(jiffies + MLX5_HEALTH_POLL_INTERVAL);
+	add_timer(&health->timer);
+}
+
+void mlx5_stop_health_poll(struct mlx5_core_dev *dev)
+{
+	struct mlx5_core_health *health = &dev->priv.health;
+
+	del_timer_sync(&health->timer);
+
+	spin_lock_irq(&health_lock);
+	if (!list_empty(&health->list))
+		list_del_init(&health->list);
+	spin_unlock_irq(&health_lock);
+}
+
+void mlx5_health_cleanup(void)
+{
+}
+
+void  __init mlx5_health_init(void)
+{
+	INIT_WORK(&health_work, health_care);
+}

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

@@ -0,0 +1,78 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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/kernel.h>
+#include <linux/module.h>
+#include <linux/mlx5/driver.h>
+#include <linux/mlx5/cmd.h>
+#include "mlx5_core.h"
+
+int mlx5_core_mad_ifc(struct mlx5_core_dev *dev, void *inb, void *outb,
+		      u16 opmod, int port)
+{
+	struct mlx5_mad_ifc_mbox_in *in = NULL;
+	struct mlx5_mad_ifc_mbox_out *out = NULL;
+	int err;
+
+	in = kzalloc(sizeof(*in), GFP_KERNEL);
+	if (!in)
+		return -ENOMEM;
+
+	out = kzalloc(sizeof(*out), GFP_KERNEL);
+	if (!out) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	in->hdr.opcode = cpu_to_be16(MLX5_CMD_OP_MAD_IFC);
+	in->hdr.opmod = cpu_to_be16(opmod);
+	in->port = port;
+
+	memcpy(in->data, inb, sizeof(in->data));
+
+	err = mlx5_cmd_exec(dev, in, sizeof(*in), out, sizeof(*out));
+	if (err)
+		goto out;
+
+	if (out->hdr.status) {
+		err = mlx5_cmd_status_to_err(&out->hdr);
+		goto out;
+	}
+
+	memcpy(outb, out->data, sizeof(out->data));
+
+out:
+	kfree(out);
+	kfree(in);
+	return err;
+}
+EXPORT_SYMBOL_GPL(mlx5_core_mad_ifc);

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

@@ -0,0 +1,475 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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 <asm-generic/kmap_types.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/io-mapping.h>
+#include <linux/mlx5/driver.h>
+#include <linux/mlx5/cq.h>
+#include <linux/mlx5/qp.h>
+#include <linux/mlx5/srq.h>
+#include <linux/debugfs.h>
+#include "mlx5_core.h"
+
+#define DRIVER_NAME "mlx5_core"
+#define DRIVER_VERSION "1.0"
+#define DRIVER_RELDATE	"June 2013"
+
+MODULE_AUTHOR("Eli Cohen <eli@mellanox.com>");
+MODULE_DESCRIPTION("Mellanox ConnectX-IB HCA core library");
+MODULE_LICENSE("Dual BSD/GPL");
+MODULE_VERSION(DRIVER_VERSION);
+
+int mlx5_core_debug_mask;
+module_param_named(debug_mask, mlx5_core_debug_mask, int, 0644);
+MODULE_PARM_DESC(debug_mask, "debug mask: 1 = dump cmd data, 2 = dump cmd exec time, 3 = both. Default=0");
+
+struct workqueue_struct *mlx5_core_wq;
+
+static int set_dma_caps(struct pci_dev *pdev)
+{
+	int err;
+
+	err = pci_set_dma_mask(pdev, DMA_BIT_MASK(64));
+	if (err) {
+		dev_warn(&pdev->dev, "Warning: couldn't set 64-bit PCI DMA mask.\n");
+		err = pci_set_dma_mask(pdev, DMA_BIT_MASK(32));
+		if (err) {
+			dev_err(&pdev->dev, "Can't set PCI DMA mask, aborting.\n");
+			return err;
+		}
+	}
+
+	err = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(64));
+	if (err) {
+		dev_warn(&pdev->dev,
+			 "Warning: couldn't set 64-bit consistent PCI DMA mask.\n");
+		err = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(32));
+		if (err) {
+			dev_err(&pdev->dev,
+				"Can't set consistent PCI DMA mask, aborting.\n");
+			return err;
+		}
+	}
+
+	dma_set_max_seg_size(&pdev->dev, 2u * 1024 * 1024 * 1024);
+	return err;
+}
+
+static int request_bar(struct pci_dev *pdev)
+{
+	int err = 0;
+
+	if (!(pci_resource_flags(pdev, 0) & IORESOURCE_MEM)) {
+		dev_err(&pdev->dev, "Missing registers BAR, aborting.\n");
+		return -ENODEV;
+	}
+
+	err = pci_request_regions(pdev, DRIVER_NAME);
+	if (err)
+		dev_err(&pdev->dev, "Couldn't get PCI resources, aborting\n");
+
+	return err;
+}
+
+static void release_bar(struct pci_dev *pdev)
+{
+	pci_release_regions(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 nvec;
+	int err;
+	int i;
+
+	nvec = dev->caps.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;
+
+	table->msix_arr = kzalloc(nvec * sizeof(*table->msix_arr), GFP_KERNEL);
+	if (!table->msix_arr)
+		return -ENOMEM;
+
+	for (i = 0; i < nvec; i++)
+		table->msix_arr[i].entry = i;
+
+retry:
+	table->num_comp_vectors = nvec - MLX5_EQ_VEC_COMP_BASE;
+	err = pci_enable_msix(dev->pdev, table->msix_arr, nvec);
+	if (err <= 0) {
+		return err;
+	} else if (err > 2) {
+		nvec = err;
+		goto retry;
+	}
+
+	mlx5_core_dbg(dev, "received %d MSI vectors out of %d requested\n", err, nvec);
+
+	return 0;
+}
+
+static void mlx5_disable_msix(struct mlx5_core_dev *dev)
+{
+	struct mlx5_eq_table *table = &dev->priv.eq_table;
+
+	pci_disable_msix(dev->pdev);
+	kfree(table->msix_arr);
+}
+
+struct mlx5_reg_host_endianess {
+	u8	he;
+	u8      rsvd[15];
+};
+
+static int handle_hca_cap(struct mlx5_core_dev *dev)
+{
+	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;
+	struct mlx5_profile *prof = dev->profile;
+	u64 flags;
+	int csum = 1;
+	int err;
+
+	memset(&query_ctx, 0, sizeof(query_ctx));
+	query_out = kzalloc(sizeof(*query_out), GFP_KERNEL);
+	if (!query_out)
+		return -ENOMEM;
+
+	set_ctx = kzalloc(sizeof(*set_ctx), GFP_KERNEL);
+	if (!set_ctx) {
+		err = -ENOMEM;
+		goto query_ex;
+	}
+
+	query_ctx.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_QUERY_HCA_CAP);
+	query_ctx.hdr.opmod  = cpu_to_be16(0x1);
+	err = mlx5_cmd_exec(dev, &query_ctx, sizeof(query_ctx),
+				 query_out, sizeof(*query_out));
+	if (err)
+		goto query_ex;
+
+	err = mlx5_cmd_status_to_err(&query_out->hdr);
+	if (err) {
+		mlx5_core_warn(dev, "query hca cap failed, %d\n", err);
+		goto query_ex;
+	}
+
+	memcpy(&set_ctx->hca_cap, &query_out->hca_cap,
+	       sizeof(set_ctx->hca_cap));
+
+	if (prof->mask & MLX5_PROF_MASK_CMDIF_CSUM) {
+		csum = !!prof->cmdif_csum;
+		flags = be64_to_cpu(set_ctx->hca_cap.flags);
+		if (csum)
+			flags |= MLX5_DEV_CAP_FLAG_CMDIF_CSUM;
+		else
+			flags &= ~MLX5_DEV_CAP_FLAG_CMDIF_CSUM;
+
+		set_ctx->hca_cap.flags = cpu_to_be64(flags);
+	}
+
+	if (dev->profile->mask & MLX5_PROF_MASK_QP_SIZE)
+		set_ctx->hca_cap.log_max_qp = dev->profile->log_max_qp;
+
+	memset(&set_out, 0, sizeof(set_out));
+	set_ctx->hca_cap.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);
+		goto query_ex;
+	}
+
+	err = mlx5_cmd_status_to_err(&set_out.hdr);
+	if (err)
+		goto query_ex;
+
+	if (!csum)
+		dev->cmd.checksum_disabled = 1;
+
+query_ex:
+	kfree(query_out);
+	kfree(set_ctx);
+
+	return err;
+}
+
+static int set_hca_ctrl(struct mlx5_core_dev *dev)
+{
+	struct mlx5_reg_host_endianess he_in;
+	struct mlx5_reg_host_endianess he_out;
+	int err;
+
+	memset(&he_in, 0, sizeof(he_in));
+	he_in.he = MLX5_SET_HOST_ENDIANNESS;
+	err = mlx5_core_access_reg(dev, &he_in,  sizeof(he_in),
+					&he_out, sizeof(he_out),
+					MLX5_REG_HOST_ENDIANNESS, 0, 1);
+	return err;
+}
+
+int mlx5_dev_init(struct mlx5_core_dev *dev, struct pci_dev *pdev)
+{
+	struct mlx5_priv *priv = &dev->priv;
+	int err;
+
+	dev->pdev = pdev;
+	pci_set_drvdata(dev->pdev, dev);
+	strncpy(priv->name, dev_name(&pdev->dev), MLX5_MAX_NAME_LEN);
+	priv->name[MLX5_MAX_NAME_LEN - 1] = 0;
+
+	mutex_init(&priv->pgdir_mutex);
+	INIT_LIST_HEAD(&priv->pgdir_list);
+	spin_lock_init(&priv->mkey_lock);
+
+	priv->dbg_root = debugfs_create_dir(dev_name(&pdev->dev), mlx5_debugfs_root);
+	if (!priv->dbg_root)
+		return -ENOMEM;
+
+	err = pci_enable_device(pdev);
+	if (err) {
+		dev_err(&pdev->dev, "Cannot enable PCI device, aborting.\n");
+		goto err_dbg;
+	}
+
+	err = request_bar(pdev);
+	if (err) {
+		dev_err(&pdev->dev, "error requesting BARs, aborting.\n");
+		goto err_disable;
+	}
+
+	pci_set_master(pdev);
+
+	err = set_dma_caps(pdev);
+	if (err) {
+		dev_err(&pdev->dev, "Failed setting DMA capabilities mask, aborting\n");
+		goto err_clr_master;
+	}
+
+	dev->iseg_base = pci_resource_start(dev->pdev, 0);
+	dev->iseg = ioremap(dev->iseg_base, sizeof(*dev->iseg));
+	if (!dev->iseg) {
+		err = -ENOMEM;
+		dev_err(&pdev->dev, "Failed mapping initialization segment, aborting\n");
+		goto err_clr_master;
+	}
+	dev_info(&pdev->dev, "firmware version: %d.%d.%d\n", fw_rev_maj(dev),
+		 fw_rev_min(dev), fw_rev_sub(dev));
+
+	err = mlx5_cmd_init(dev);
+	if (err) {
+		dev_err(&pdev->dev, "Failed initializing command interface, aborting\n");
+		goto err_unmap;
+	}
+
+	mlx5_pagealloc_init(dev);
+	err = set_hca_ctrl(dev);
+	if (err) {
+		dev_err(&pdev->dev, "set_hca_ctrl failed\n");
+		goto err_pagealloc_cleanup;
+	}
+
+	err = handle_hca_cap(dev);
+	if (err) {
+		dev_err(&pdev->dev, "handle_hca_cap failed\n");
+		goto err_pagealloc_cleanup;
+	}
+
+	err = mlx5_satisfy_startup_pages(dev);
+	if (err) {
+		dev_err(&pdev->dev, "failed to allocate startup pages\n");
+		goto err_pagealloc_cleanup;
+	}
+
+	err = mlx5_pagealloc_start(dev);
+	if (err) {
+		dev_err(&pdev->dev, "mlx5_pagealloc_start failed\n");
+		goto err_reclaim_pages;
+	}
+
+	err = mlx5_cmd_init_hca(dev);
+	if (err) {
+		dev_err(&pdev->dev, "init hca failed\n");
+		goto err_pagealloc_stop;
+	}
+
+	mlx5_start_health_poll(dev);
+
+	err = mlx5_cmd_query_hca_cap(dev, &dev->caps);
+	if (err) {
+		dev_err(&pdev->dev, "query hca failed\n");
+		goto err_stop_poll;
+	}
+
+	err = mlx5_cmd_query_adapter(dev);
+	if (err) {
+		dev_err(&pdev->dev, "query adapter failed\n");
+		goto err_stop_poll;
+	}
+
+	err = mlx5_enable_msix(dev);
+	if (err) {
+		dev_err(&pdev->dev, "enable msix failed\n");
+		goto err_stop_poll;
+	}
+
+	err = mlx5_eq_init(dev);
+	if (err) {
+		dev_err(&pdev->dev, "failed to initialize eq\n");
+		goto disable_msix;
+	}
+
+	err = mlx5_alloc_uuars(dev, &priv->uuari);
+	if (err) {
+		dev_err(&pdev->dev, "Failed allocating uar, aborting\n");
+		goto err_eq_cleanup;
+	}
+
+	err = mlx5_start_eqs(dev);
+	if (err) {
+		dev_err(&pdev->dev, "Failed to start pages and async EQs\n");
+		goto err_free_uar;
+	}
+
+	MLX5_INIT_DOORBELL_LOCK(&priv->cq_uar_lock);
+
+	mlx5_init_cq_table(dev);
+	mlx5_init_qp_table(dev);
+	mlx5_init_srq_table(dev);
+
+	return 0;
+
+err_free_uar:
+	mlx5_free_uuars(dev, &priv->uuari);
+
+err_eq_cleanup:
+	mlx5_eq_cleanup(dev);
+
+disable_msix:
+	mlx5_disable_msix(dev);
+
+err_stop_poll:
+	mlx5_stop_health_poll(dev);
+	mlx5_cmd_teardown_hca(dev);
+
+err_pagealloc_stop:
+	mlx5_pagealloc_stop(dev);
+
+err_reclaim_pages:
+	mlx5_reclaim_startup_pages(dev);
+
+err_pagealloc_cleanup:
+	mlx5_pagealloc_cleanup(dev);
+	mlx5_cmd_cleanup(dev);
+
+err_unmap:
+	iounmap(dev->iseg);
+
+err_clr_master:
+	pci_clear_master(dev->pdev);
+	release_bar(dev->pdev);
+
+err_disable:
+	pci_disable_device(dev->pdev);
+
+err_dbg:
+	debugfs_remove(priv->dbg_root);
+	return err;
+}
+EXPORT_SYMBOL(mlx5_dev_init);
+
+void mlx5_dev_cleanup(struct mlx5_core_dev *dev)
+{
+	struct mlx5_priv *priv = &dev->priv;
+
+	mlx5_cleanup_srq_table(dev);
+	mlx5_cleanup_qp_table(dev);
+	mlx5_cleanup_cq_table(dev);
+	mlx5_stop_eqs(dev);
+	mlx5_free_uuars(dev, &priv->uuari);
+	mlx5_eq_cleanup(dev);
+	mlx5_disable_msix(dev);
+	mlx5_stop_health_poll(dev);
+	mlx5_cmd_teardown_hca(dev);
+	mlx5_pagealloc_stop(dev);
+	mlx5_reclaim_startup_pages(dev);
+	mlx5_pagealloc_cleanup(dev);
+	mlx5_cmd_cleanup(dev);
+	iounmap(dev->iseg);
+	pci_clear_master(dev->pdev);
+	release_bar(dev->pdev);
+	pci_disable_device(dev->pdev);
+	debugfs_remove(priv->dbg_root);
+}
+EXPORT_SYMBOL(mlx5_dev_cleanup);
+
+static int __init init(void)
+{
+	int err;
+
+	mlx5_register_debugfs();
+	mlx5_core_wq = create_singlethread_workqueue("mlx5_core_wq");
+	if (!mlx5_core_wq) {
+		err = -ENOMEM;
+		goto err_debug;
+	}
+	mlx5_health_init();
+
+	return 0;
+
+	mlx5_health_cleanup();
+err_debug:
+	mlx5_unregister_debugfs();
+	return err;
+}
+
+static void __exit cleanup(void)
+{
+	mlx5_health_cleanup();
+	destroy_workqueue(mlx5_core_wq);
+	mlx5_unregister_debugfs();
+}
+
+module_init(init);
+module_exit(cleanup);

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

@@ -0,0 +1,106 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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/kernel.h>
+#include <linux/module.h>
+#include <linux/mlx5/driver.h>
+#include <linux/mlx5/cmd.h>
+#include <rdma/ib_verbs.h>
+#include "mlx5_core.h"
+
+struct mlx5_attach_mcg_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	__be32			qpn;
+	__be32			rsvd;
+	u8			gid[16];
+};
+
+struct mlx5_attach_mcg_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvf[8];
+};
+
+struct mlx5_detach_mcg_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	__be32			qpn;
+	__be32			rsvd;
+	u8			gid[16];
+};
+
+struct mlx5_detach_mcg_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvf[8];
+};
+
+int mlx5_core_attach_mcg(struct mlx5_core_dev *dev, union ib_gid *mgid, u32 qpn)
+{
+	struct mlx5_attach_mcg_mbox_in in;
+	struct mlx5_attach_mcg_mbox_out out;
+	int err;
+
+	memset(&in, 0, sizeof(in));
+	memset(&out, 0, sizeof(out));
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_ATTACH_TO_MCG);
+	memcpy(in.gid, mgid, sizeof(*mgid));
+	in.qpn = cpu_to_be32(qpn);
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), &out, sizeof(out));
+	if (err)
+		return err;
+
+	if (out.hdr.status)
+		err = mlx5_cmd_status_to_err(&out.hdr);
+
+	return err;
+}
+EXPORT_SYMBOL(mlx5_core_attach_mcg);
+
+int mlx5_core_detach_mcg(struct mlx5_core_dev *dev, union ib_gid *mgid, u32 qpn)
+{
+	struct mlx5_detach_mcg_mbox_in in;
+	struct mlx5_detach_mcg_mbox_out out;
+	int err;
+
+	memset(&in, 0, sizeof(in));
+	memset(&out, 0, sizeof(out));
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_DETACH_FROM_MCG);
+	memcpy(in.gid, mgid, sizeof(*mgid));
+	in.qpn = cpu_to_be32(qpn);
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), &out, sizeof(out));
+	if (err)
+		return err;
+
+	if (out.hdr.status)
+		err = mlx5_cmd_status_to_err(&out.hdr);
+
+	return err;
+}
+EXPORT_SYMBOL(mlx5_core_detach_mcg);

drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h

@@ -0,0 +1,73 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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_CORE_H__
+#define __MLX5_CORE_H__
+
+#include <linux/types.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
+
+extern int mlx5_core_debug_mask;
+
+#define mlx5_core_dbg(dev, format, arg...)				       \
+pr_debug("%s:%s:%d:(pid %d): " format, (dev)->priv.name, __func__, __LINE__,   \
+	 current->pid, ##arg)
+
+#define mlx5_core_dbg_mask(dev, mask, format, arg...)			       \
+do {									       \
+	if ((mask) & mlx5_core_debug_mask)				       \
+		pr_debug("%s:%s:%d:(pid %d): " format, (dev)->priv.name,       \
+			 __func__, __LINE__, current->pid, ##arg);	       \
+} while (0)
+
+#define mlx5_core_err(dev, format, arg...) \
+pr_err("%s:%s:%d:(pid %d): " format, (dev)->priv.name, __func__, __LINE__,     \
+	current->pid, ##arg)
+
+#define mlx5_core_warn(dev, format, arg...) \
+pr_warn("%s:%s:%d:(pid %d): " format, (dev)->priv.name, __func__, __LINE__,    \
+	current->pid, ##arg)
+
+enum {
+	MLX5_CMD_DATA, /* print command payload only */
+	MLX5_CMD_TIME, /* print command execution time */
+};
+
+
+int mlx5_cmd_query_hca_cap(struct mlx5_core_dev *dev,
+			   struct mlx5_caps *caps);
+int mlx5_cmd_query_adapter(struct mlx5_core_dev *dev);
+int mlx5_cmd_init_hca(struct mlx5_core_dev *dev);
+int mlx5_cmd_teardown_hca(struct mlx5_core_dev *dev);
+
+#endif /* __MLX5_CORE_H__ */

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

@@ -0,0 +1,136 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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/kernel.h>
+#include <linux/module.h>
+#include <linux/mlx5/driver.h>
+#include <linux/mlx5/cmd.h>
+#include "mlx5_core.h"
+
+int mlx5_core_create_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mr *mr,
+			  struct mlx5_create_mkey_mbox_in *in, int inlen)
+{
+	struct mlx5_create_mkey_mbox_out out;
+	int err;
+	u8 key;
+
+	memset(&out, 0, sizeof(out));
+	spin_lock(&dev->priv.mkey_lock);
+	key = dev->priv.mkey_key++;
+	spin_unlock(&dev->priv.mkey_lock);
+	in->seg.qpn_mkey7_0 |= cpu_to_be32(key);
+	in->hdr.opcode = cpu_to_be16(MLX5_CMD_OP_CREATE_MKEY);
+	err = mlx5_cmd_exec(dev, in, inlen, &out, sizeof(out));
+	if (err) {
+		mlx5_core_dbg(dev, "cmd exec faile %d\n", err);
+		return err;
+	}
+
+	if (out.hdr.status) {
+		mlx5_core_dbg(dev, "status %d\n", out.hdr.status);
+		return mlx5_cmd_status_to_err(&out.hdr);
+	}
+
+	mr->key = mlx5_idx_to_mkey(be32_to_cpu(out.mkey) & 0xffffff) | key;
+	mlx5_core_dbg(dev, "out 0x%x, key 0x%x, mkey 0x%x\n", be32_to_cpu(out.mkey), key, mr->key);
+
+	return err;
+}
+EXPORT_SYMBOL(mlx5_core_create_mkey);
+
+int mlx5_core_destroy_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mr *mr)
+{
+	struct mlx5_destroy_mkey_mbox_in in;
+	struct mlx5_destroy_mkey_mbox_out out;
+	int err;
+
+	memset(&in, 0, sizeof(in));
+	memset(&out, 0, sizeof(out));
+
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_DESTROY_MKEY);
+	in.mkey = cpu_to_be32(mlx5_mkey_to_idx(mr->key));
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), &out, sizeof(out));
+	if (err)
+		return err;
+
+	if (out.hdr.status)
+		return mlx5_cmd_status_to_err(&out.hdr);
+
+	return err;
+}
+EXPORT_SYMBOL(mlx5_core_destroy_mkey);
+
+int mlx5_core_query_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mr *mr,
+			 struct mlx5_query_mkey_mbox_out *out, int outlen)
+{
+	struct mlx5_destroy_mkey_mbox_in in;
+	int err;
+
+	memset(&in, 0, sizeof(in));
+	memset(out, 0, outlen);
+
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_QUERY_MKEY);
+	in.mkey = cpu_to_be32(mlx5_mkey_to_idx(mr->key));
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), out, outlen);
+	if (err)
+		return err;
+
+	if (out->hdr.status)
+		return mlx5_cmd_status_to_err(&out->hdr);
+
+	return err;
+}
+EXPORT_SYMBOL(mlx5_core_query_mkey);
+
+int mlx5_core_dump_fill_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mr *mr,
+			     u32 *mkey)
+{
+	struct mlx5_query_special_ctxs_mbox_in in;
+	struct mlx5_query_special_ctxs_mbox_out out;
+	int err;
+
+	memset(&in, 0, sizeof(in));
+	memset(&out, 0, sizeof(out));
+
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_QUERY_SPECIAL_CONTEXTS);
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), &out, sizeof(out));
+	if (err)
+		return err;
+
+	if (out.hdr.status)
+		return mlx5_cmd_status_to_err(&out.hdr);
+
+	*mkey = be32_to_cpu(out.dump_fill_mkey);
+
+	return err;
+}
+EXPORT_SYMBOL(mlx5_core_dump_fill_mkey);

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

@@ -0,0 +1,435 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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 <asm-generic/kmap_types.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/mlx5/driver.h>
+#include <linux/mlx5/cmd.h>
+#include "mlx5_core.h"
+
+enum {
+	MLX5_PAGES_CANT_GIVE	= 0,
+	MLX5_PAGES_GIVE		= 1,
+	MLX5_PAGES_TAKE		= 2
+};
+
+struct mlx5_pages_req {
+	struct mlx5_core_dev *dev;
+	u32	func_id;
+	s16	npages;
+	struct work_struct work;
+};
+
+struct fw_page {
+	struct rb_node	rb_node;
+	u64		addr;
+	struct page	*page;
+	u16		func_id;
+};
+
+struct mlx5_query_pages_inbox {
+	struct mlx5_inbox_hdr	hdr;
+	u8			rsvd[8];
+};
+
+struct mlx5_query_pages_outbox {
+	struct mlx5_outbox_hdr	hdr;
+	u8			reserved[2];
+	__be16			func_id;
+	__be16			init_pages;
+	__be16			num_pages;
+};
+
+struct mlx5_manage_pages_inbox {
+	struct mlx5_inbox_hdr	hdr;
+	__be16			rsvd0;
+	__be16			func_id;
+	__be16			rsvd1;
+	__be16			num_entries;
+	u8			rsvd2[16];
+	__be64			pas[0];
+};
+
+struct mlx5_manage_pages_outbox {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvd0[2];
+	__be16			num_entries;
+	u8			rsvd1[20];
+	__be64			pas[0];
+};
+
+static int insert_page(struct mlx5_core_dev *dev, u64 addr, struct page *page, u16 func_id)
+{
+	struct rb_root *root = &dev->priv.page_root;
+	struct rb_node **new = &root->rb_node;
+	struct rb_node *parent = NULL;
+	struct fw_page *nfp;
+	struct fw_page *tfp;
+
+	while (*new) {
+		parent = *new;
+		tfp = rb_entry(parent, struct fw_page, rb_node);
+		if (tfp->addr < addr)
+			new = &parent->rb_left;
+		else if (tfp->addr > addr)
+			new = &parent->rb_right;
+		else
+			return -EEXIST;
+	}
+
+	nfp = kmalloc(sizeof(*nfp), GFP_KERNEL);
+	if (!nfp)
+		return -ENOMEM;
+
+	nfp->addr = addr;
+	nfp->page = page;
+	nfp->func_id = func_id;
+
+	rb_link_node(&nfp->rb_node, parent, new);
+	rb_insert_color(&nfp->rb_node, root);
+
+	return 0;
+}
+
+static struct page *remove_page(struct mlx5_core_dev *dev, u64 addr)
+{
+	struct rb_root *root = &dev->priv.page_root;
+	struct rb_node *tmp = root->rb_node;
+	struct page *result = NULL;
+	struct fw_page *tfp;
+
+	while (tmp) {
+		tfp = rb_entry(tmp, struct fw_page, rb_node);
+		if (tfp->addr < addr) {
+			tmp = tmp->rb_left;
+		} else if (tfp->addr > addr) {
+			tmp = tmp->rb_right;
+		} else {
+			rb_erase(&tfp->rb_node, root);
+			result = tfp->page;
+			kfree(tfp);
+			break;
+		}
+	}
+
+	return result;
+}
+
+static int mlx5_cmd_query_pages(struct mlx5_core_dev *dev, u16 *func_id,
+				s16 *pages, s16 *init_pages)
+{
+	struct mlx5_query_pages_inbox	in;
+	struct mlx5_query_pages_outbox	out;
+	int err;
+
+	memset(&in, 0, sizeof(in));
+	memset(&out, 0, sizeof(out));
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_QUERY_PAGES);
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), &out, sizeof(out));
+	if (err)
+		return err;
+
+	if (out.hdr.status)
+		return mlx5_cmd_status_to_err(&out.hdr);
+
+	if (pages)
+		*pages = be16_to_cpu(out.num_pages);
+	if (init_pages)
+		*init_pages = be16_to_cpu(out.init_pages);
+	*func_id = be16_to_cpu(out.func_id);
+
+	return err;
+}
+
+static int give_pages(struct mlx5_core_dev *dev, u16 func_id, int npages,
+		      int notify_fail)
+{
+	struct mlx5_manage_pages_inbox *in;
+	struct mlx5_manage_pages_outbox out;
+	struct page *page;
+	int inlen;
+	u64 addr;
+	int err;
+	int i;
+
+	inlen = sizeof(*in) + npages * sizeof(in->pas[0]);
+	in = mlx5_vzalloc(inlen);
+	if (!in) {
+		mlx5_core_warn(dev, "vzalloc failed %d\n", inlen);
+		return -ENOMEM;
+	}
+	memset(&out, 0, sizeof(out));
+
+	for (i = 0; i < npages; i++) {
+		page = alloc_page(GFP_HIGHUSER);
+		if (!page) {
+			err = -ENOMEM;
+			mlx5_core_warn(dev, "failed to allocate page\n");
+			goto out_alloc;
+		}
+		addr = dma_map_page(&dev->pdev->dev, page, 0,
+				    PAGE_SIZE, DMA_BIDIRECTIONAL);
+		if (dma_mapping_error(&dev->pdev->dev, addr)) {
+			mlx5_core_warn(dev, "failed dma mapping page\n");
+			__free_page(page);
+			err = -ENOMEM;
+			goto out_alloc;
+		}
+		err = insert_page(dev, addr, page, func_id);
+		if (err) {
+			mlx5_core_err(dev, "failed to track allocated page\n");
+			dma_unmap_page(&dev->pdev->dev, addr, PAGE_SIZE, DMA_BIDIRECTIONAL);
+			__free_page(page);
+			err = -ENOMEM;
+			goto out_alloc;
+		}
+		in->pas[i] = cpu_to_be64(addr);
+	}
+
+	in->hdr.opcode = cpu_to_be16(MLX5_CMD_OP_MANAGE_PAGES);
+	in->hdr.opmod = cpu_to_be16(MLX5_PAGES_GIVE);
+	in->func_id = cpu_to_be16(func_id);
+	in->num_entries = cpu_to_be16(npages);
+	err = mlx5_cmd_exec(dev, in, inlen, &out, sizeof(out));
+	mlx5_core_dbg(dev, "err %d\n", err);
+	if (err) {
+		mlx5_core_warn(dev, "func_id 0x%x, npages %d, err %d\n", func_id, npages, err);
+		goto out_alloc;
+	}
+	dev->priv.fw_pages += npages;
+
+	if (out.hdr.status) {
+		err = mlx5_cmd_status_to_err(&out.hdr);
+		if (err) {
+			mlx5_core_warn(dev, "func_id 0x%x, npages %d, status %d\n", func_id, npages, out.hdr.status);
+			goto out_alloc;
+		}
+	}
+
+	mlx5_core_dbg(dev, "err %d\n", err);
+
+	goto out_free;
+
+out_alloc:
+	if (notify_fail) {
+		memset(in, 0, inlen);
+		memset(&out, 0, sizeof(out));
+		in->hdr.opcode = cpu_to_be16(MLX5_CMD_OP_MANAGE_PAGES);
+		in->hdr.opmod = cpu_to_be16(MLX5_PAGES_CANT_GIVE);
+		if (mlx5_cmd_exec(dev, in, sizeof(*in), &out, sizeof(out)))
+			mlx5_core_warn(dev, "\n");
+	}
+	for (i--; i >= 0; i--) {
+		addr = be64_to_cpu(in->pas[i]);
+		page = remove_page(dev, addr);
+		if (!page) {
+			mlx5_core_err(dev, "BUG: can't remove page at addr 0x%llx\n",
+				      addr);
+			continue;
+		}
+		dma_unmap_page(&dev->pdev->dev, addr, PAGE_SIZE, DMA_BIDIRECTIONAL);
+		__free_page(page);
+	}
+
+out_free:
+	mlx5_vfree(in);
+	return err;
+}
+
+static int reclaim_pages(struct mlx5_core_dev *dev, u32 func_id, int npages,
+			 int *nclaimed)
+{
+	struct mlx5_manage_pages_inbox   in;
+	struct mlx5_manage_pages_outbox *out;
+	struct page *page;
+	int num_claimed;
+	int outlen;
+	u64 addr;
+	int err;
+	int i;
+
+	memset(&in, 0, sizeof(in));
+	outlen = sizeof(*out) + npages * sizeof(out->pas[0]);
+	out = mlx5_vzalloc(outlen);
+	if (!out)
+		return -ENOMEM;
+
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_MANAGE_PAGES);
+	in.hdr.opmod = cpu_to_be16(MLX5_PAGES_TAKE);
+	in.func_id = cpu_to_be16(func_id);
+	in.num_entries = cpu_to_be16(npages);
+	mlx5_core_dbg(dev, "npages %d, outlen %d\n", npages, outlen);
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), out, outlen);
+	if (err) {
+		mlx5_core_err(dev, "failed recliaming pages\n");
+		goto out_free;
+	}
+	dev->priv.fw_pages -= npages;
+
+	if (out->hdr.status) {
+		err = mlx5_cmd_status_to_err(&out->hdr);
+		goto out_free;
+	}
+
+	num_claimed = be16_to_cpu(out->num_entries);
+	if (nclaimed)
+		*nclaimed = num_claimed;
+
+	for (i = 0; i < num_claimed; i++) {
+		addr = be64_to_cpu(out->pas[i]);
+		page = remove_page(dev, addr);
+		if (!page) {
+			mlx5_core_warn(dev, "FW reported unknown DMA address 0x%llx\n", addr);
+		} else {
+			dma_unmap_page(&dev->pdev->dev, addr, PAGE_SIZE, DMA_BIDIRECTIONAL);
+			__free_page(page);
+		}
+	}
+
+out_free:
+	mlx5_vfree(out);
+	return err;
+}
+
+static void pages_work_handler(struct work_struct *work)
+{
+	struct mlx5_pages_req *req = container_of(work, struct mlx5_pages_req, work);
+	struct mlx5_core_dev *dev = req->dev;
+	int err = 0;
+
+	if (req->npages < 0)
+		err = reclaim_pages(dev, req->func_id, -1 * req->npages, NULL);
+	else if (req->npages > 0)
+		err = give_pages(dev, req->func_id, req->npages, 1);
+
+	if (err)
+		mlx5_core_warn(dev, "%s fail %d\n", req->npages < 0 ?
+			       "reclaim" : "give", err);
+
+	kfree(req);
+}
+
+void mlx5_core_req_pages_handler(struct mlx5_core_dev *dev, u16 func_id,
+				 s16 npages)
+{
+	struct mlx5_pages_req *req;
+
+	req = kzalloc(sizeof(*req), GFP_ATOMIC);
+	if (!req) {
+		mlx5_core_warn(dev, "failed to allocate pages request\n");
+		return;
+	}
+
+	req->dev = dev;
+	req->func_id = func_id;
+	req->npages = npages;
+	INIT_WORK(&req->work, pages_work_handler);
+	queue_work(dev->priv.pg_wq, &req->work);
+}
+
+int mlx5_satisfy_startup_pages(struct mlx5_core_dev *dev)
+{
+	s16 uninitialized_var(init_pages);
+	u16 uninitialized_var(func_id);
+	int err;
+
+	err = mlx5_cmd_query_pages(dev, &func_id, NULL, &init_pages);
+	if (err)
+		return err;
+
+	mlx5_core_dbg(dev, "requested %d init pages for func_id 0x%x\n", init_pages, func_id);
+
+	return give_pages(dev, func_id, init_pages, 0);
+}
+
+static int optimal_reclaimed_pages(void)
+{
+	struct mlx5_cmd_prot_block *block;
+	struct mlx5_cmd_layout *lay;
+	int ret;
+
+	ret = (sizeof(lay->in) + sizeof(block->data) -
+	       sizeof(struct mlx5_manage_pages_outbox)) / 8;
+
+	return ret;
+}
+
+int mlx5_reclaim_startup_pages(struct mlx5_core_dev *dev)
+{
+	unsigned long end = jiffies + msecs_to_jiffies(5000);
+	struct fw_page *fwp;
+	struct rb_node *p;
+	int err;
+
+	do {
+		p = rb_first(&dev->priv.page_root);
+		if (p) {
+			fwp = rb_entry(p, struct fw_page, rb_node);
+			err = reclaim_pages(dev, fwp->func_id, optimal_reclaimed_pages(), NULL);
+			if (err) {
+				mlx5_core_warn(dev, "failed reclaiming pages (%d)\n", err);
+				return err;
+			}
+		}
+		if (time_after(jiffies, end)) {
+			mlx5_core_warn(dev, "FW did not return all pages. giving up...\n");
+			break;
+		}
+	} while (p);
+
+	return 0;
+}
+
+void mlx5_pagealloc_init(struct mlx5_core_dev *dev)
+{
+	dev->priv.page_root = RB_ROOT;
+}
+
+void mlx5_pagealloc_cleanup(struct mlx5_core_dev *dev)
+{
+	/* nothing */
+}
+
+int mlx5_pagealloc_start(struct mlx5_core_dev *dev)
+{
+	dev->priv.pg_wq = create_singlethread_workqueue("mlx5_page_allocator");
+	if (!dev->priv.pg_wq)
+		return -ENOMEM;
+
+	return 0;
+}
+
+void mlx5_pagealloc_stop(struct mlx5_core_dev *dev)
+{
+	destroy_workqueue(dev->priv.pg_wq);
+}

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

@@ -0,0 +1,101 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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/kernel.h>
+#include <linux/module.h>
+#include <linux/mlx5/driver.h>
+#include <linux/mlx5/cmd.h>
+#include "mlx5_core.h"
+
+struct mlx5_alloc_pd_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	u8			rsvd[8];
+};
+
+struct mlx5_alloc_pd_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	__be32			pdn;
+	u8			rsvd[4];
+};
+
+struct mlx5_dealloc_pd_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	__be32			pdn;
+	u8			rsvd[4];
+};
+
+struct mlx5_dealloc_pd_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvd[8];
+};
+
+int mlx5_core_alloc_pd(struct mlx5_core_dev *dev, u32 *pdn)
+{
+	struct mlx5_alloc_pd_mbox_in	in;
+	struct mlx5_alloc_pd_mbox_out	out;
+	int err;
+
+	memset(&in, 0, sizeof(in));
+	memset(&out, 0, sizeof(out));
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_ALLOC_PD);
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), &out, sizeof(out));
+	if (err)
+		return err;
+
+	if (out.hdr.status)
+		return mlx5_cmd_status_to_err(&out.hdr);
+
+	*pdn = be32_to_cpu(out.pdn) & 0xffffff;
+	return err;
+}
+EXPORT_SYMBOL(mlx5_core_alloc_pd);
+
+int mlx5_core_dealloc_pd(struct mlx5_core_dev *dev, u32 pdn)
+{
+	struct mlx5_dealloc_pd_mbox_in	in;
+	struct mlx5_dealloc_pd_mbox_out	out;
+	int err;
+
+	memset(&in, 0, sizeof(in));
+	memset(&out, 0, sizeof(out));
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_DEALLOC_PD);
+	in.pdn = cpu_to_be32(pdn);
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), &out, sizeof(out));
+	if (err)
+		return err;
+
+	if (out.hdr.status)
+		return mlx5_cmd_status_to_err(&out.hdr);
+
+	return err;
+}
+EXPORT_SYMBOL(mlx5_core_dealloc_pd);

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

@@ -0,0 +1,104 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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/driver.h>
+#include <linux/mlx5/cmd.h>
+#include "mlx5_core.h"
+
+int mlx5_core_access_reg(struct mlx5_core_dev *dev, void *data_in,
+			 int size_in, void *data_out, int size_out,
+			 u16 reg_num, int arg, int write)
+{
+	struct mlx5_access_reg_mbox_in *in = NULL;
+	struct mlx5_access_reg_mbox_out *out = NULL;
+	int err = -ENOMEM;
+
+	in = mlx5_vzalloc(sizeof(*in) + size_in);
+	if (!in)
+		return -ENOMEM;
+
+	out = mlx5_vzalloc(sizeof(*out) + size_out);
+	if (!out)
+		goto ex1;
+
+	memcpy(in->data, data_in, size_in);
+	in->hdr.opcode = cpu_to_be16(MLX5_CMD_OP_ACCESS_REG);
+	in->hdr.opmod = cpu_to_be16(!write);
+	in->arg = cpu_to_be32(arg);
+	in->register_id = cpu_to_be16(reg_num);
+	err = mlx5_cmd_exec(dev, in, sizeof(*in) + size_in, out,
+			    sizeof(out) + size_out);
+	if (err)
+		goto ex2;
+
+	if (out->hdr.status)
+		err = mlx5_cmd_status_to_err(&out->hdr);
+
+	if (!err)
+		memcpy(data_out, out->data, size_out);
+
+ex2:
+	mlx5_vfree(out);
+ex1:
+	mlx5_vfree(in);
+	return err;
+}
+EXPORT_SYMBOL_GPL(mlx5_core_access_reg);
+
+
+struct mlx5_reg_pcap {
+	u8			rsvd0;
+	u8			port_num;
+	u8			rsvd1[2];
+	__be32			caps_127_96;
+	__be32			caps_95_64;
+	__be32			caps_63_32;
+	__be32			caps_31_0;
+};
+
+int mlx5_set_port_caps(struct mlx5_core_dev *dev, int port_num, u32 caps)
+{
+	struct mlx5_reg_pcap in;
+	struct mlx5_reg_pcap out;
+	int err;
+
+	memset(&in, 0, sizeof(in));
+	in.caps_127_96 = cpu_to_be32(caps);
+	in.port_num = port_num;
+
+	err = mlx5_core_access_reg(dev, &in, sizeof(in), &out,
+				   sizeof(out), MLX5_REG_PCAP, 0, 1);
+
+	return err;
+}
+EXPORT_SYMBOL_GPL(mlx5_set_port_caps);

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

@@ -0,0 +1,301 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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/gfp.h>
+#include <linux/export.h>
+#include <linux/mlx5/cmd.h>
+#include <linux/mlx5/qp.h>
+#include <linux/mlx5/driver.h>
+
+#include "mlx5_core.h"
+
+void mlx5_qp_event(struct mlx5_core_dev *dev, u32 qpn, int event_type)
+{
+	struct mlx5_qp_table *table = &dev->priv.qp_table;
+	struct mlx5_core_qp *qp;
+
+	spin_lock(&table->lock);
+
+	qp = radix_tree_lookup(&table->tree, qpn);
+	if (qp)
+		atomic_inc(&qp->refcount);
+
+	spin_unlock(&table->lock);
+
+	if (!qp) {
+		mlx5_core_warn(dev, "Async event for bogus QP 0x%x\n", qpn);
+		return;
+	}
+
+	qp->event(qp, event_type);
+
+	if (atomic_dec_and_test(&qp->refcount))
+		complete(&qp->free);
+}
+
+int mlx5_core_create_qp(struct mlx5_core_dev *dev,
+			struct mlx5_core_qp *qp,
+			struct mlx5_create_qp_mbox_in *in,
+			int inlen)
+{
+	struct mlx5_qp_table *table = &dev->priv.qp_table;
+	struct mlx5_create_qp_mbox_out out;
+	struct mlx5_destroy_qp_mbox_in din;
+	struct mlx5_destroy_qp_mbox_out dout;
+	int err;
+
+	memset(&dout, 0, sizeof(dout));
+	in->hdr.opcode = cpu_to_be16(MLX5_CMD_OP_CREATE_QP);
+
+	err = mlx5_cmd_exec(dev, in, inlen, &out, sizeof(out));
+	if (err) {
+		mlx5_core_warn(dev, "ret %d", err);
+		return err;
+	}
+
+	if (out.hdr.status) {
+		pr_warn("current num of QPs 0x%x\n", atomic_read(&dev->num_qps));
+		return mlx5_cmd_status_to_err(&out.hdr);
+	}
+
+	qp->qpn = be32_to_cpu(out.qpn) & 0xffffff;
+	mlx5_core_dbg(dev, "qpn = 0x%x\n", qp->qpn);
+
+	spin_lock_irq(&table->lock);
+	err = radix_tree_insert(&table->tree, qp->qpn, qp);
+	spin_unlock_irq(&table->lock);
+	if (err) {
+		mlx5_core_warn(dev, "err %d", err);
+		goto err_cmd;
+	}
+
+	err = mlx5_debug_qp_add(dev, qp);
+	if (err)
+		mlx5_core_dbg(dev, "failed adding QP 0x%x to debug file system\n",
+			      qp->qpn);
+
+	qp->pid = current->pid;
+	atomic_set(&qp->refcount, 1);
+	atomic_inc(&dev->num_qps);
+	init_completion(&qp->free);
+
+	return 0;
+
+err_cmd:
+	memset(&din, 0, sizeof(din));
+	memset(&dout, 0, sizeof(dout));
+	din.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_DESTROY_QP);
+	din.qpn = cpu_to_be32(qp->qpn);
+	mlx5_cmd_exec(dev, &din, sizeof(din), &out, sizeof(dout));
+
+	return err;
+}
+EXPORT_SYMBOL_GPL(mlx5_core_create_qp);
+
+int mlx5_core_destroy_qp(struct mlx5_core_dev *dev,
+			 struct mlx5_core_qp *qp)
+{
+	struct mlx5_destroy_qp_mbox_in in;
+	struct mlx5_destroy_qp_mbox_out out;
+	struct mlx5_qp_table *table = &dev->priv.qp_table;
+	unsigned long flags;
+	int err;
+
+	mlx5_debug_qp_remove(dev, qp);
+
+	spin_lock_irqsave(&table->lock, flags);
+	radix_tree_delete(&table->tree, qp->qpn);
+	spin_unlock_irqrestore(&table->lock, flags);
+
+	if (atomic_dec_and_test(&qp->refcount))
+		complete(&qp->free);
+	wait_for_completion(&qp->free);
+
+	memset(&in, 0, sizeof(in));
+	memset(&out, 0, sizeof(out));
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_DESTROY_QP);
+	in.qpn = cpu_to_be32(qp->qpn);
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), &out, sizeof(out));
+	if (err)
+		return err;
+
+	if (out.hdr.status)
+		return mlx5_cmd_status_to_err(&out.hdr);
+
+	atomic_dec(&dev->num_qps);
+	return 0;
+}
+EXPORT_SYMBOL_GPL(mlx5_core_destroy_qp);
+
+int mlx5_core_qp_modify(struct mlx5_core_dev *dev, enum mlx5_qp_state cur_state,
+			enum mlx5_qp_state new_state,
+			struct mlx5_modify_qp_mbox_in *in, int sqd_event,
+			struct mlx5_core_qp *qp)
+{
+	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_CMD_OP_RTS2SQD_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_RTS]	= MLX5_CMD_OP_SQD2RTS_QP,
+			[MLX5_QP_STATE_SQD]	= MLX5_CMD_OP_SQD2SQD_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_modify_qp_mbox_out out;
+	int err = 0;
+	u16 op;
+
+	if (cur_state >= MLX5_QP_NUM_STATE || new_state >= MLX5_QP_NUM_STATE ||
+	    !optab[cur_state][new_state])
+		return -EINVAL;
+
+	memset(&out, 0, sizeof(out));
+	op = optab[cur_state][new_state];
+	in->hdr.opcode = cpu_to_be16(op);
+	in->qpn = cpu_to_be32(qp->qpn);
+	err = mlx5_cmd_exec(dev, in, sizeof(*in), &out, sizeof(out));
+	if (err)
+		return err;
+
+	return mlx5_cmd_status_to_err(&out.hdr);
+}
+EXPORT_SYMBOL_GPL(mlx5_core_qp_modify);
+
+void mlx5_init_qp_table(struct mlx5_core_dev *dev)
+{
+	struct mlx5_qp_table *table = &dev->priv.qp_table;
+
+	spin_lock_init(&table->lock);
+	INIT_RADIX_TREE(&table->tree, GFP_ATOMIC);
+	mlx5_qp_debugfs_init(dev);
+}
+
+void mlx5_cleanup_qp_table(struct mlx5_core_dev *dev)
+{
+	mlx5_qp_debugfs_cleanup(dev);
+}
+
+int mlx5_core_qp_query(struct mlx5_core_dev *dev, struct mlx5_core_qp *qp,
+		       struct mlx5_query_qp_mbox_out *out, int outlen)
+{
+	struct mlx5_query_qp_mbox_in in;
+	int err;
+
+	memset(&in, 0, sizeof(in));
+	memset(out, 0, outlen);
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_QUERY_QP);
+	in.qpn = cpu_to_be32(qp->qpn);
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), out, outlen);
+	if (err)
+		return err;
+
+	if (out->hdr.status)
+		return mlx5_cmd_status_to_err(&out->hdr);
+
+	return err;
+}
+EXPORT_SYMBOL_GPL(mlx5_core_qp_query);
+
+int mlx5_core_xrcd_alloc(struct mlx5_core_dev *dev, u32 *xrcdn)
+{
+	struct mlx5_alloc_xrcd_mbox_in in;
+	struct mlx5_alloc_xrcd_mbox_out out;
+	int err;
+
+	memset(&in, 0, sizeof(in));
+	memset(&out, 0, sizeof(out));
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_ALLOC_XRCD);
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), &out, sizeof(out));
+	if (err)
+		return err;
+
+	if (out.hdr.status)
+		err = mlx5_cmd_status_to_err(&out.hdr);
+	else
+		*xrcdn = be32_to_cpu(out.xrcdn);
+
+	return err;
+}
+EXPORT_SYMBOL_GPL(mlx5_core_xrcd_alloc);
+
+int mlx5_core_xrcd_dealloc(struct mlx5_core_dev *dev, u32 xrcdn)
+{
+	struct mlx5_dealloc_xrcd_mbox_in in;
+	struct mlx5_dealloc_xrcd_mbox_out out;
+	int err;
+
+	memset(&in, 0, sizeof(in));
+	memset(&out, 0, sizeof(out));
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_DEALLOC_XRCD);
+	in.xrcdn = cpu_to_be32(xrcdn);
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), &out, sizeof(out));
+	if (err)
+		return err;
+
+	if (out.hdr.status)
+		err = mlx5_cmd_status_to_err(&out.hdr);
+
+	return err;
+}
+EXPORT_SYMBOL_GPL(mlx5_core_xrcd_dealloc);

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

@@ -0,0 +1,223 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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/kernel.h>
+#include <linux/module.h>
+#include <linux/mlx5/driver.h>
+#include <linux/mlx5/cmd.h>
+#include <linux/mlx5/srq.h>
+#include <rdma/ib_verbs.h>
+#include "mlx5_core.h"
+
+void mlx5_srq_event(struct mlx5_core_dev *dev, u32 srqn, int event_type)
+{
+	struct mlx5_srq_table *table = &dev->priv.srq_table;
+	struct mlx5_core_srq *srq;
+
+	spin_lock(&table->lock);
+
+	srq = radix_tree_lookup(&table->tree, srqn);
+	if (srq)
+		atomic_inc(&srq->refcount);
+
+	spin_unlock(&table->lock);
+
+	if (!srq) {
+		mlx5_core_warn(dev, "Async event for bogus SRQ 0x%08x\n", srqn);
+		return;
+	}
+
+	srq->event(srq, event_type);
+
+	if (atomic_dec_and_test(&srq->refcount))
+		complete(&srq->free);
+}
+
+struct mlx5_core_srq *mlx5_core_get_srq(struct mlx5_core_dev *dev, u32 srqn)
+{
+	struct mlx5_srq_table *table = &dev->priv.srq_table;
+	struct mlx5_core_srq *srq;
+
+	spin_lock(&table->lock);
+
+	srq = radix_tree_lookup(&table->tree, srqn);
+	if (srq)
+		atomic_inc(&srq->refcount);
+
+	spin_unlock(&table->lock);
+
+	return srq;
+}
+EXPORT_SYMBOL(mlx5_core_get_srq);
+
+int mlx5_core_create_srq(struct mlx5_core_dev *dev, struct mlx5_core_srq *srq,
+			 struct mlx5_create_srq_mbox_in *in, int inlen)
+{
+	struct mlx5_create_srq_mbox_out out;
+	struct mlx5_srq_table *table = &dev->priv.srq_table;
+	struct mlx5_destroy_srq_mbox_in din;
+	struct mlx5_destroy_srq_mbox_out dout;
+	int err;
+
+	memset(&out, 0, sizeof(out));
+	in->hdr.opcode = cpu_to_be16(MLX5_CMD_OP_CREATE_SRQ);
+	err = mlx5_cmd_exec(dev, in, inlen, &out, sizeof(out));
+	if (err)
+		return err;
+
+	if (out.hdr.status)
+		return mlx5_cmd_status_to_err(&out.hdr);
+
+	srq->srqn = be32_to_cpu(out.srqn) & 0xffffff;
+
+	atomic_set(&srq->refcount, 1);
+	init_completion(&srq->free);
+
+	spin_lock_irq(&table->lock);
+	err = radix_tree_insert(&table->tree, srq->srqn, srq);
+	spin_unlock_irq(&table->lock);
+	if (err) {
+		mlx5_core_warn(dev, "err %d, srqn 0x%x\n", err, srq->srqn);
+		goto err_cmd;
+	}
+
+	return 0;
+
+err_cmd:
+	memset(&din, 0, sizeof(din));
+	memset(&dout, 0, sizeof(dout));
+	din.srqn = cpu_to_be32(srq->srqn);
+	din.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_DESTROY_SRQ);
+	mlx5_cmd_exec(dev, &din, sizeof(din), &dout, sizeof(dout));
+	return err;
+}
+EXPORT_SYMBOL(mlx5_core_create_srq);
+
+int mlx5_core_destroy_srq(struct mlx5_core_dev *dev, struct mlx5_core_srq *srq)
+{
+	struct mlx5_destroy_srq_mbox_in in;
+	struct mlx5_destroy_srq_mbox_out out;
+	struct mlx5_srq_table *table = &dev->priv.srq_table;
+	struct mlx5_core_srq *tmp;
+	int err;
+
+	spin_lock_irq(&table->lock);
+	tmp = radix_tree_delete(&table->tree, srq->srqn);
+	spin_unlock_irq(&table->lock);
+	if (!tmp) {
+		mlx5_core_warn(dev, "srq 0x%x not found in tree\n", srq->srqn);
+		return -EINVAL;
+	}
+	if (tmp != srq) {
+		mlx5_core_warn(dev, "corruption on srqn 0x%x\n", srq->srqn);
+		return -EINVAL;
+	}
+
+	memset(&in, 0, sizeof(in));
+	memset(&out, 0, sizeof(out));
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_DESTROY_SRQ);
+	in.srqn = cpu_to_be32(srq->srqn);
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), &out, sizeof(out));
+	if (err)
+		return err;
+
+	if (out.hdr.status)
+		return mlx5_cmd_status_to_err(&out.hdr);
+
+	if (atomic_dec_and_test(&srq->refcount))
+		complete(&srq->free);
+	wait_for_completion(&srq->free);
+
+	return 0;
+}
+EXPORT_SYMBOL(mlx5_core_destroy_srq);
+
+int mlx5_core_query_srq(struct mlx5_core_dev *dev, struct mlx5_core_srq *srq,
+			struct mlx5_query_srq_mbox_out *out)
+{
+	struct mlx5_query_srq_mbox_in in;
+	int err;
+
+	memset(&in, 0, sizeof(in));
+	memset(out, 0, sizeof(*out));
+
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_QUERY_SRQ);
+	in.srqn = cpu_to_be32(srq->srqn);
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), out, sizeof(*out));
+	if (err)
+		return err;
+
+	if (out->hdr.status)
+		return mlx5_cmd_status_to_err(&out->hdr);
+
+	return err;
+}
+EXPORT_SYMBOL(mlx5_core_query_srq);
+
+int mlx5_core_arm_srq(struct mlx5_core_dev *dev, struct mlx5_core_srq *srq,
+		      u16 lwm, int is_srq)
+{
+	struct mlx5_arm_srq_mbox_in	in;
+	struct mlx5_arm_srq_mbox_out	out;
+	int err;
+
+	memset(&in, 0, sizeof(in));
+	memset(&out, 0, sizeof(out));
+
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_ARM_RQ);
+	in.hdr.opmod = cpu_to_be16(!!is_srq);
+	in.srqn = cpu_to_be32(srq->srqn);
+	in.lwm = cpu_to_be16(lwm);
+
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), &out, sizeof(out));
+	if (err)
+		return err;
+
+	if (out.hdr.status)
+		return mlx5_cmd_status_to_err(&out.hdr);
+
+	return err;
+}
+EXPORT_SYMBOL(mlx5_core_arm_srq);
+
+void mlx5_init_srq_table(struct mlx5_core_dev *dev)
+{
+	struct mlx5_srq_table *table = &dev->priv.srq_table;
+
+	spin_lock_init(&table->lock);
+	INIT_RADIX_TREE(&table->tree, GFP_ATOMIC);
+}
+
+void mlx5_cleanup_srq_table(struct mlx5_core_dev *dev)
+{
+	/* nothing */
+}

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

@@ -0,0 +1,223 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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/kernel.h>
+#include <linux/module.h>
+#include <linux/mlx5/driver.h>
+#include <linux/mlx5/cmd.h>
+#include "mlx5_core.h"
+
+enum {
+	NUM_DRIVER_UARS		= 4,
+	NUM_LOW_LAT_UUARS	= 4,
+};
+
+
+struct mlx5_alloc_uar_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	u8			rsvd[8];
+};
+
+struct mlx5_alloc_uar_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	__be32			uarn;
+	u8			rsvd[4];
+};
+
+struct mlx5_free_uar_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	__be32			uarn;
+	u8			rsvd[4];
+};
+
+struct mlx5_free_uar_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvd[8];
+};
+
+int mlx5_cmd_alloc_uar(struct mlx5_core_dev *dev, u32 *uarn)
+{
+	struct mlx5_alloc_uar_mbox_in	in;
+	struct mlx5_alloc_uar_mbox_out	out;
+	int err;
+
+	memset(&in, 0, sizeof(in));
+	memset(&out, 0, sizeof(out));
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_ALLOC_UAR);
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), &out, sizeof(out));
+	if (err)
+		goto ex;
+
+	if (out.hdr.status) {
+		err = mlx5_cmd_status_to_err(&out.hdr);
+		goto ex;
+	}
+
+	*uarn = be32_to_cpu(out.uarn) & 0xffffff;
+
+ex:
+	return err;
+}
+EXPORT_SYMBOL(mlx5_cmd_alloc_uar);
+
+int mlx5_cmd_free_uar(struct mlx5_core_dev *dev, u32 uarn)
+{
+	struct mlx5_free_uar_mbox_in	in;
+	struct mlx5_free_uar_mbox_out	out;
+	int err;
+
+	memset(&in, 0, sizeof(in));
+	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_DEALLOC_UAR);
+	in.uarn = cpu_to_be32(uarn);
+	err = mlx5_cmd_exec(dev, &in, sizeof(in), &out, sizeof(out));
+	if (err)
+		goto ex;
+
+	if (out.hdr.status)
+		err = mlx5_cmd_status_to_err(&out.hdr);
+
+ex:
+	return err;
+}
+EXPORT_SYMBOL(mlx5_cmd_free_uar);
+
+static int need_uuar_lock(int uuarn)
+{
+	int tot_uuars = NUM_DRIVER_UARS * MLX5_BF_REGS_PER_PAGE;
+
+	if (uuarn == 0 || tot_uuars - NUM_LOW_LAT_UUARS)
+		return 0;
+
+	return 1;
+}
+
+int mlx5_alloc_uuars(struct mlx5_core_dev *dev, struct mlx5_uuar_info *uuari)
+{
+	int tot_uuars = NUM_DRIVER_UARS * MLX5_BF_REGS_PER_PAGE;
+	struct mlx5_bf *bf;
+	phys_addr_t addr;
+	int err;
+	int i;
+
+	uuari->num_uars = NUM_DRIVER_UARS;
+	uuari->num_low_latency_uuars = NUM_LOW_LAT_UUARS;
+
+	mutex_init(&uuari->lock);
+	uuari->uars = kcalloc(uuari->num_uars, sizeof(*uuari->uars), GFP_KERNEL);
+	if (!uuari->uars)
+		return -ENOMEM;
+
+	uuari->bfs = kcalloc(tot_uuars, sizeof(*uuari->bfs), GFP_KERNEL);
+	if (!uuari->bfs) {
+		err = -ENOMEM;
+		goto out_uars;
+	}
+
+	uuari->bitmap = kcalloc(BITS_TO_LONGS(tot_uuars), sizeof(*uuari->bitmap),
+				GFP_KERNEL);
+	if (!uuari->bitmap) {
+		err = -ENOMEM;
+		goto out_bfs;
+	}
+
+	uuari->count = kcalloc(tot_uuars, sizeof(*uuari->count), GFP_KERNEL);
+	if (!uuari->count) {
+		err = -ENOMEM;
+		goto out_bitmap;
+	}
+
+	for (i = 0; i < uuari->num_uars; i++) {
+		err = mlx5_cmd_alloc_uar(dev, &uuari->uars[i].index);
+		if (err)
+			goto out_count;
+
+		addr = dev->iseg_base + ((phys_addr_t)(uuari->uars[i].index) << PAGE_SHIFT);
+		uuari->uars[i].map = ioremap(addr, PAGE_SIZE);
+		if (!uuari->uars[i].map) {
+			mlx5_cmd_free_uar(dev, uuari->uars[i].index);
+			goto out_count;
+		}
+		mlx5_core_dbg(dev, "allocated uar index 0x%x, mmaped at %p\n",
+			      uuari->uars[i].index, uuari->uars[i].map);
+	}
+
+	for (i = 0; i < tot_uuars; i++) {
+		bf = &uuari->bfs[i];
+
+		bf->buf_size = dev->caps.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 +
+			MLX5_BF_OFFSET;
+		bf->need_lock = need_uuar_lock(i);
+		spin_lock_init(&bf->lock);
+		spin_lock_init(&bf->lock32);
+		bf->uuarn = i;
+	}
+
+	return 0;
+
+out_count:
+	for (i--; i >= 0; i--) {
+		iounmap(uuari->uars[i].map);
+		mlx5_cmd_free_uar(dev, uuari->uars[i].index);
+	}
+	kfree(uuari->count);
+
+out_bitmap:
+	kfree(uuari->bitmap);
+
+out_bfs:
+	kfree(uuari->bfs);
+
+out_uars:
+	kfree(uuari->uars);
+	return err;
+}
+
+int mlx5_free_uuars(struct mlx5_core_dev *dev, struct mlx5_uuar_info *uuari)
+{
+	int i = uuari->num_uars;
+
+	for (i--; i >= 0; i--) {
+		iounmap(uuari->uars[i].map);
+		mlx5_cmd_free_uar(dev, uuari->uars[i].index);
+	}
+
+	kfree(uuari->count);
+	kfree(uuari->bitmap);
+	kfree(uuari->bfs);
+	kfree(uuari->uars);
+
+	return 0;
+}

include/linux/mlx5/cmd.h

@@ -0,0 +1,51 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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_CMD_H
+#define MLX5_CMD_H
+
+#include <linux/types.h>
+
+struct manage_pages_layout {
+	u64	ptr;
+	u32	reserved;
+	u16	num_entries;
+	u16	func_id;
+};
+
+
+struct mlx5_cmd_alloc_uar_imm_out {
+	u32	rsvd[3];
+	u32	uarn;
+};
+
+#endif /* MLX5_CMD_H */

include/linux/mlx5/cq.h

@@ -0,0 +1,165 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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_CORE_CQ_H
+#define MLX5_CORE_CQ_H
+
+#include <rdma/ib_verbs.h>
+#include <linux/mlx5/driver.h>
+
+
+struct mlx5_core_cq {
+	u32			cqn;
+	int			cqe_sz;
+	__be32		       *set_ci_db;
+	__be32		       *arm_db;
+	atomic_t		refcount;
+	struct completion	free;
+	unsigned		vector;
+	int			irqn;
+	void (*comp)		(struct mlx5_core_cq *);
+	void (*event)		(struct mlx5_core_cq *, enum mlx5_event);
+	struct mlx5_uar	       *uar;
+	u32			cons_index;
+	unsigned		arm_sn;
+	struct mlx5_rsc_debug	*dbg;
+	int			pid;
+};
+
+
+enum {
+	MLX5_CQE_SYNDROME_LOCAL_LENGTH_ERR		= 0x01,
+	MLX5_CQE_SYNDROME_LOCAL_QP_OP_ERR		= 0x02,
+	MLX5_CQE_SYNDROME_LOCAL_PROT_ERR		= 0x04,
+	MLX5_CQE_SYNDROME_WR_FLUSH_ERR			= 0x05,
+	MLX5_CQE_SYNDROME_MW_BIND_ERR			= 0x06,
+	MLX5_CQE_SYNDROME_BAD_RESP_ERR			= 0x10,
+	MLX5_CQE_SYNDROME_LOCAL_ACCESS_ERR		= 0x11,
+	MLX5_CQE_SYNDROME_REMOTE_INVAL_REQ_ERR		= 0x12,
+	MLX5_CQE_SYNDROME_REMOTE_ACCESS_ERR		= 0x13,
+	MLX5_CQE_SYNDROME_REMOTE_OP_ERR			= 0x14,
+	MLX5_CQE_SYNDROME_TRANSPORT_RETRY_EXC_ERR	= 0x15,
+	MLX5_CQE_SYNDROME_RNR_RETRY_EXC_ERR		= 0x16,
+	MLX5_CQE_SYNDROME_REMOTE_ABORTED_ERR		= 0x22,
+};
+
+enum {
+	MLX5_CQE_OWNER_MASK	= 1,
+	MLX5_CQE_REQ		= 0,
+	MLX5_CQE_RESP_WR_IMM	= 1,
+	MLX5_CQE_RESP_SEND	= 2,
+	MLX5_CQE_RESP_SEND_IMM	= 3,
+	MLX5_CQE_RESP_SEND_INV	= 4,
+	MLX5_CQE_RESIZE_CQ	= 0xff, /* TBD */
+	MLX5_CQE_REQ_ERR	= 13,
+	MLX5_CQE_RESP_ERR	= 14,
+};
+
+enum {
+	MLX5_CQ_MODIFY_RESEIZE = 0,
+	MLX5_CQ_MODIFY_MODER = 1,
+	MLX5_CQ_MODIFY_MAPPING = 2,
+};
+
+struct mlx5_cq_modify_params {
+	int	type;
+	union {
+		struct {
+			u32	page_offset;
+			u8	log_cq_size;
+		} resize;
+
+		struct {
+		} moder;
+
+		struct {
+		} mapping;
+	} params;
+};
+
+enum {
+	CQE_SIZE_64 = 0,
+	CQE_SIZE_128 = 1,
+};
+
+static inline int cqe_sz_to_mlx_sz(u8 size)
+{
+	return size == 64 ? CQE_SIZE_64 : CQE_SIZE_128;
+}
+
+static inline void mlx5_cq_set_ci(struct mlx5_core_cq *cq)
+{
+	*cq->set_ci_db = cpu_to_be32(cq->cons_index & 0xffffff);
+}
+
+enum {
+	MLX5_CQ_DB_REQ_NOT_SOL		= 1 << 24,
+	MLX5_CQ_DB_REQ_NOT		= 0 << 24
+};
+
+static inline void mlx5_cq_arm(struct mlx5_core_cq *cq, u32 cmd,
+			       void __iomem *uar_page,
+			       spinlock_t *doorbell_lock)
+{
+	__be32 doorbell[2];
+	u32 sn;
+	u32 ci;
+
+	sn = cq->arm_sn & 3;
+	ci = cq->cons_index & 0xffffff;
+
+	*cq->arm_db = cpu_to_be32(sn << 28 | cmd | ci);
+
+	/* Make sure that the doorbell record in host memory is
+	 * written before ringing the doorbell via PCI MMIO.
+	 */
+	wmb();
+
+	doorbell[0] = cpu_to_be32(sn << 28 | cmd | ci);
+	doorbell[1] = cpu_to_be32(cq->cqn);
+
+	mlx5_write64(doorbell, uar_page + MLX5_CQ_DOORBELL, doorbell_lock);
+}
+
+int mlx5_init_cq_table(struct mlx5_core_dev *dev);
+void mlx5_cleanup_cq_table(struct mlx5_core_dev *dev);
+int mlx5_core_create_cq(struct mlx5_core_dev *dev, struct mlx5_core_cq *cq,
+			struct mlx5_create_cq_mbox_in *in, int inlen);
+int mlx5_core_destroy_cq(struct mlx5_core_dev *dev, struct mlx5_core_cq *cq);
+int mlx5_core_query_cq(struct mlx5_core_dev *dev, struct mlx5_core_cq *cq,
+		       struct mlx5_query_cq_mbox_out *out);
+int mlx5_core_modify_cq(struct mlx5_core_dev *dev, struct mlx5_core_cq *cq,
+			int type, struct mlx5_cq_modify_params *params);
+int mlx5_debug_cq_add(struct mlx5_core_dev *dev, struct mlx5_core_cq *cq);
+void mlx5_debug_cq_remove(struct mlx5_core_dev *dev, struct mlx5_core_cq *cq);
+
+#endif /* MLX5_CORE_CQ_H */

include/linux/mlx5/device.h

@@ -0,0 +1,893 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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_DEVICE_H
+#define MLX5_DEVICE_H
+
+#include <linux/types.h>
+#include <rdma/ib_verbs.h>
+
+#if defined(__LITTLE_ENDIAN)
+#define MLX5_SET_HOST_ENDIANNESS	0
+#elif defined(__BIG_ENDIAN)
+#define MLX5_SET_HOST_ENDIANNESS	0x80
+#else
+#error Host endianness not defined
+#endif
+
+enum {
+	MLX5_MAX_COMMANDS		= 32,
+	MLX5_CMD_DATA_BLOCK_SIZE	= 512,
+	MLX5_PCI_CMD_XPORT		= 7,
+};
+
+enum {
+	MLX5_EXTENDED_UD_AV		= 0x80000000,
+};
+
+enum {
+	MLX5_CQ_STATE_ARMED		= 9,
+	MLX5_CQ_STATE_ALWAYS_ARMED	= 0xb,
+	MLX5_CQ_STATE_FIRED		= 0xa,
+};
+
+enum {
+	MLX5_STAT_RATE_OFFSET	= 5,
+};
+
+enum {
+	MLX5_INLINE_SEG = 0x80000000,
+};
+
+enum {
+	MLX5_PERM_LOCAL_READ	= 1 << 2,
+	MLX5_PERM_LOCAL_WRITE	= 1 << 3,
+	MLX5_PERM_REMOTE_READ	= 1 << 4,
+	MLX5_PERM_REMOTE_WRITE	= 1 << 5,
+	MLX5_PERM_ATOMIC	= 1 << 6,
+	MLX5_PERM_UMR_EN	= 1 << 7,
+};
+
+enum {
+	MLX5_PCIE_CTRL_SMALL_FENCE	= 1 << 0,
+	MLX5_PCIE_CTRL_RELAXED_ORDERING	= 1 << 2,
+	MLX5_PCIE_CTRL_NO_SNOOP		= 1 << 3,
+	MLX5_PCIE_CTRL_TLP_PROCE_EN	= 1 << 6,
+	MLX5_PCIE_CTRL_TPH_MASK		= 3 << 4,
+};
+
+enum {
+	MLX5_ACCESS_MODE_PA	= 0,
+	MLX5_ACCESS_MODE_MTT	= 1,
+	MLX5_ACCESS_MODE_KLM	= 2
+};
+
+enum {
+	MLX5_MKEY_REMOTE_INVAL	= 1 << 24,
+	MLX5_MKEY_FLAG_SYNC_UMR = 1 << 29,
+	MLX5_MKEY_BSF_EN	= 1 << 30,
+	MLX5_MKEY_LEN64		= 1 << 31,
+};
+
+enum {
+	MLX5_EN_RD	= (u64)1,
+	MLX5_EN_WR	= (u64)2
+};
+
+enum {
+	MLX5_BF_REGS_PER_PAGE	= 4,
+	MLX5_MAX_UAR_PAGES	= 1 << 8,
+	MLX5_MAX_UUARS		= MLX5_MAX_UAR_PAGES * MLX5_BF_REGS_PER_PAGE,
+};
+
+enum {
+	MLX5_MKEY_MASK_LEN		= 1ull << 0,
+	MLX5_MKEY_MASK_PAGE_SIZE	= 1ull << 1,
+	MLX5_MKEY_MASK_START_ADDR	= 1ull << 6,
+	MLX5_MKEY_MASK_PD		= 1ull << 7,
+	MLX5_MKEY_MASK_EN_RINVAL	= 1ull << 8,
+	MLX5_MKEY_MASK_BSF_EN		= 1ull << 12,
+	MLX5_MKEY_MASK_KEY		= 1ull << 13,
+	MLX5_MKEY_MASK_QPN		= 1ull << 14,
+	MLX5_MKEY_MASK_LR		= 1ull << 17,
+	MLX5_MKEY_MASK_LW		= 1ull << 18,
+	MLX5_MKEY_MASK_RR		= 1ull << 19,
+	MLX5_MKEY_MASK_RW		= 1ull << 20,
+	MLX5_MKEY_MASK_A		= 1ull << 21,
+	MLX5_MKEY_MASK_SMALL_FENCE	= 1ull << 23,
+	MLX5_MKEY_MASK_FREE		= 1ull << 29,
+};
+
+enum mlx5_event {
+	MLX5_EVENT_TYPE_COMP		   = 0x0,
+
+	MLX5_EVENT_TYPE_PATH_MIG	   = 0x01,
+	MLX5_EVENT_TYPE_COMM_EST	   = 0x02,
+	MLX5_EVENT_TYPE_SQ_DRAINED	   = 0x03,
+	MLX5_EVENT_TYPE_SRQ_LAST_WQE	   = 0x13,
+	MLX5_EVENT_TYPE_SRQ_RQ_LIMIT	   = 0x14,
+
+	MLX5_EVENT_TYPE_CQ_ERROR	   = 0x04,
+	MLX5_EVENT_TYPE_WQ_CATAS_ERROR	   = 0x05,
+	MLX5_EVENT_TYPE_PATH_MIG_FAILED	   = 0x07,
+	MLX5_EVENT_TYPE_WQ_INVAL_REQ_ERROR = 0x10,
+	MLX5_EVENT_TYPE_WQ_ACCESS_ERROR	   = 0x11,
+	MLX5_EVENT_TYPE_SRQ_CATAS_ERROR	   = 0x12,
+
+	MLX5_EVENT_TYPE_INTERNAL_ERROR	   = 0x08,
+	MLX5_EVENT_TYPE_PORT_CHANGE	   = 0x09,
+	MLX5_EVENT_TYPE_GPIO_EVENT	   = 0x15,
+	MLX5_EVENT_TYPE_REMOTE_CONFIG	   = 0x19,
+
+	MLX5_EVENT_TYPE_DB_BF_CONGESTION   = 0x1a,
+	MLX5_EVENT_TYPE_STALL_EVENT	   = 0x1b,
+
+	MLX5_EVENT_TYPE_CMD		   = 0x0a,
+	MLX5_EVENT_TYPE_PAGE_REQUEST	   = 0xb,
+};
+
+enum {
+	MLX5_PORT_CHANGE_SUBTYPE_DOWN		= 1,
+	MLX5_PORT_CHANGE_SUBTYPE_ACTIVE		= 4,
+	MLX5_PORT_CHANGE_SUBTYPE_INITIALIZED	= 5,
+	MLX5_PORT_CHANGE_SUBTYPE_LID		= 6,
+	MLX5_PORT_CHANGE_SUBTYPE_PKEY		= 7,
+	MLX5_PORT_CHANGE_SUBTYPE_GUID		= 8,
+	MLX5_PORT_CHANGE_SUBTYPE_CLIENT_REREG	= 9,
+};
+
+enum {
+	MLX5_DEV_CAP_FLAG_RC		= 1LL <<  0,
+	MLX5_DEV_CAP_FLAG_UC		= 1LL <<  1,
+	MLX5_DEV_CAP_FLAG_UD		= 1LL <<  2,
+	MLX5_DEV_CAP_FLAG_XRC		= 1LL <<  3,
+	MLX5_DEV_CAP_FLAG_SRQ		= 1LL <<  6,
+	MLX5_DEV_CAP_FLAG_BAD_PKEY_CNTR	= 1LL <<  8,
+	MLX5_DEV_CAP_FLAG_BAD_QKEY_CNTR	= 1LL <<  9,
+	MLX5_DEV_CAP_FLAG_APM		= 1LL << 17,
+	MLX5_DEV_CAP_FLAG_ATOMIC	= 1LL << 18,
+	MLX5_DEV_CAP_FLAG_ON_DMND_PG	= 1LL << 24,
+	MLX5_DEV_CAP_FLAG_RESIZE_SRQ	= 1LL << 32,
+	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	= 1LL << 46,
+};
+
+enum {
+	MLX5_OPCODE_NOP			= 0x00,
+	MLX5_OPCODE_SEND_INVAL		= 0x01,
+	MLX5_OPCODE_RDMA_WRITE		= 0x08,
+	MLX5_OPCODE_RDMA_WRITE_IMM	= 0x09,
+	MLX5_OPCODE_SEND		= 0x0a,
+	MLX5_OPCODE_SEND_IMM		= 0x0b,
+	MLX5_OPCODE_RDMA_READ		= 0x10,
+	MLX5_OPCODE_ATOMIC_CS		= 0x11,
+	MLX5_OPCODE_ATOMIC_FA		= 0x12,
+	MLX5_OPCODE_ATOMIC_MASKED_CS	= 0x14,
+	MLX5_OPCODE_ATOMIC_MASKED_FA	= 0x15,
+	MLX5_OPCODE_BIND_MW		= 0x18,
+	MLX5_OPCODE_CONFIG_CMD		= 0x1f,
+
+	MLX5_RECV_OPCODE_RDMA_WRITE_IMM	= 0x00,
+	MLX5_RECV_OPCODE_SEND		= 0x01,
+	MLX5_RECV_OPCODE_SEND_IMM	= 0x02,
+	MLX5_RECV_OPCODE_SEND_INVAL	= 0x03,
+
+	MLX5_CQE_OPCODE_ERROR		= 0x1e,
+	MLX5_CQE_OPCODE_RESIZE		= 0x16,
+
+	MLX5_OPCODE_SET_PSV		= 0x20,
+	MLX5_OPCODE_GET_PSV		= 0x21,
+	MLX5_OPCODE_CHECK_PSV		= 0x22,
+	MLX5_OPCODE_RGET_PSV		= 0x26,
+	MLX5_OPCODE_RCHECK_PSV		= 0x27,
+
+	MLX5_OPCODE_UMR			= 0x25,
+
+};
+
+enum {
+	MLX5_SET_PORT_RESET_QKEY	= 0,
+	MLX5_SET_PORT_GUID0		= 16,
+	MLX5_SET_PORT_NODE_GUID		= 17,
+	MLX5_SET_PORT_SYS_GUID		= 18,
+	MLX5_SET_PORT_GID_TABLE		= 19,
+	MLX5_SET_PORT_PKEY_TABLE	= 20,
+};
+
+enum {
+	MLX5_MAX_PAGE_SHIFT		= 31
+};
+
+struct mlx5_inbox_hdr {
+	__be16		opcode;
+	u8		rsvd[4];
+	__be16		opmod;
+};
+
+struct mlx5_outbox_hdr {
+	u8		status;
+	u8		rsvd[3];
+	__be32		syndrome;
+};
+
+struct mlx5_cmd_query_adapter_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	u8			rsvd[8];
+};
+
+struct mlx5_cmd_query_adapter_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvd0[24];
+	u8			intapin;
+	u8			rsvd1[13];
+	__be16			vsd_vendor_id;
+	u8			vsd[208];
+	u8			vsd_psid[16];
+};
+
+struct mlx5_hca_cap {
+	u8	rsvd1[16];
+	u8	log_max_srq_sz;
+	u8	log_max_qp_sz;
+	u8	rsvd2;
+	u8	log_max_qp;
+	u8	log_max_strq_sz;
+	u8	log_max_srqs;
+	u8	rsvd4[2];
+	u8	rsvd5;
+	u8	log_max_cq_sz;
+	u8	rsvd6;
+	u8	log_max_cq;
+	u8	log_max_eq_sz;
+	u8	log_max_mkey;
+	u8	rsvd7;
+	u8	log_max_eq;
+	u8	max_indirection;
+	u8	log_max_mrw_sz;
+	u8	log_max_bsf_list_sz;
+	u8	log_max_klm_list_sz;
+	u8	rsvd_8_0;
+	u8	log_max_ra_req_dc;
+	u8	rsvd_8_1;
+	u8	log_max_ra_res_dc;
+	u8	rsvd9;
+	u8	log_max_ra_req_qp;
+	u8	rsvd10;
+	u8	log_max_ra_res_qp;
+	u8	rsvd11[4];
+	__be16	max_qp_count;
+	__be16	rsvd12;
+	u8	rsvd13;
+	u8	local_ca_ack_delay;
+	u8	rsvd14;
+	u8	num_ports;
+	u8	log_max_msg;
+	u8	rsvd15[3];
+	__be16	stat_rate_support;
+	u8	rsvd16[2];
+	__be64	flags;
+	u8	rsvd17;
+	u8	uar_sz;
+	u8	rsvd18;
+	u8	log_pg_sz;
+	__be16	bf_log_bf_reg_size;
+	u8	rsvd19[4];
+	__be16	max_desc_sz_sq;
+	u8	rsvd20[2];
+	__be16	max_desc_sz_rq;
+	u8	rsvd21[2];
+	__be16	max_desc_sz_sq_dc;
+	u8	rsvd22[4];
+	__be16	max_qp_mcg;
+	u8	rsvd23;
+	u8	log_max_mcg;
+	u8	rsvd24;
+	u8	log_max_pd;
+	u8	rsvd25;
+	u8	log_max_xrcd;
+	u8	rsvd26[40];
+	__be32  uar_page_sz;
+	u8	rsvd27[28];
+	u8	log_msx_atomic_size_qp;
+	u8	rsvd28[2];
+	u8	log_msx_atomic_size_dc;
+	u8	rsvd29[76];
+};
+
+
+struct mlx5_cmd_query_hca_cap_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	u8			rsvd[8];
+};
+
+
+struct mlx5_cmd_query_hca_cap_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvd0[8];
+	struct mlx5_hca_cap     hca_cap;
+};
+
+
+struct mlx5_cmd_set_hca_cap_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	u8			rsvd[8];
+	struct mlx5_hca_cap     hca_cap;
+};
+
+
+struct mlx5_cmd_set_hca_cap_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvd0[8];
+};
+
+
+struct mlx5_cmd_init_hca_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	u8			rsvd0[2];
+	__be16			profile;
+	u8			rsvd1[4];
+};
+
+struct mlx5_cmd_init_hca_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvd[8];
+};
+
+struct mlx5_cmd_teardown_hca_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	u8			rsvd0[2];
+	__be16			profile;
+	u8			rsvd1[4];
+};
+
+struct mlx5_cmd_teardown_hca_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvd[8];
+};
+
+struct mlx5_cmd_layout {
+	u8		type;
+	u8		rsvd0[3];
+	__be32		inlen;
+	__be64		in_ptr;
+	__be32		in[4];
+	__be32		out[4];
+	__be64		out_ptr;
+	__be32		outlen;
+	u8		token;
+	u8		sig;
+	u8		rsvd1;
+	u8		status_own;
+};
+
+
+struct health_buffer {
+	__be32		assert_var[5];
+	__be32		rsvd0[3];
+	__be32		assert_exit_ptr;
+	__be32		assert_callra;
+	__be32		rsvd1[2];
+	__be32		fw_ver;
+	__be32		hw_id;
+	__be32		rsvd2;
+	u8		irisc_index;
+	u8		synd;
+	__be16		ext_sync;
+};
+
+struct mlx5_init_seg {
+	__be32			fw_rev;
+	__be32			cmdif_rev_fw_sub;
+	__be32			rsvd0[2];
+	__be32			cmdq_addr_h;
+	__be32			cmdq_addr_l_sz;
+	__be32			cmd_dbell;
+	__be32			rsvd1[121];
+	struct health_buffer	health;
+	__be32			rsvd2[884];
+	__be32			health_counter;
+	__be32			rsvd3[1023];
+	__be64			ieee1588_clk;
+	__be32			ieee1588_clk_type;
+	__be32			clr_intx;
+};
+
+struct mlx5_eqe_comp {
+	__be32	reserved[6];
+	__be32	cqn;
+};
+
+struct mlx5_eqe_qp_srq {
+	__be32	reserved[6];
+	__be32	qp_srq_n;
+};
+
+struct mlx5_eqe_cq_err {
+	__be32	cqn;
+	u8	reserved1[7];
+	u8	syndrome;
+};
+
+struct mlx5_eqe_dropped_packet {
+};
+
+struct mlx5_eqe_port_state {
+	u8	reserved0[8];
+	u8	port;
+};
+
+struct mlx5_eqe_gpio {
+	__be32	reserved0[2];
+	__be64	gpio_event;
+};
+
+struct mlx5_eqe_congestion {
+	u8	type;
+	u8	rsvd0;
+	u8	congestion_level;
+};
+
+struct mlx5_eqe_stall_vl {
+	u8	rsvd0[3];
+	u8	port_vl;
+};
+
+struct mlx5_eqe_cmd {
+	__be32	vector;
+	__be32	rsvd[6];
+};
+
+struct mlx5_eqe_page_req {
+	u8		rsvd0[2];
+	__be16		func_id;
+	u8		rsvd1[2];
+	__be16		num_pages;
+	__be32		rsvd2[5];
+};
+
+union ev_data {
+	__be32				raw[7];
+	struct mlx5_eqe_cmd		cmd;
+	struct mlx5_eqe_comp		comp;
+	struct mlx5_eqe_qp_srq		qp_srq;
+	struct mlx5_eqe_cq_err		cq_err;
+	struct mlx5_eqe_dropped_packet	dp;
+	struct mlx5_eqe_port_state	port;
+	struct mlx5_eqe_gpio		gpio;
+	struct mlx5_eqe_congestion	cong;
+	struct mlx5_eqe_stall_vl	stall_vl;
+	struct mlx5_eqe_page_req	req_pages;
+} __packed;
+
+struct mlx5_eqe {
+	u8		rsvd0;
+	u8		type;
+	u8		rsvd1;
+	u8		sub_type;
+	__be32		rsvd2[7];
+	union ev_data	data;
+	__be16		rsvd3;
+	u8		signature;
+	u8		owner;
+} __packed;
+
+struct mlx5_cmd_prot_block {
+	u8		data[MLX5_CMD_DATA_BLOCK_SIZE];
+	u8		rsvd0[48];
+	__be64		next;
+	__be32		block_num;
+	u8		rsvd1;
+	u8		token;
+	u8		ctrl_sig;
+	u8		sig;
+};
+
+struct mlx5_err_cqe {
+	u8	rsvd0[32];
+	__be32	srqn;
+	u8	rsvd1[18];
+	u8	vendor_err_synd;
+	u8	syndrome;
+	__be32	s_wqe_opcode_qpn;
+	__be16	wqe_counter;
+	u8	signature;
+	u8	op_own;
+};
+
+struct mlx5_cqe64 {
+	u8		rsvd0[17];
+	u8		ml_path;
+	u8		rsvd20[4];
+	__be16		slid;
+	__be32		flags_rqpn;
+	u8		rsvd28[4];
+	__be32		srqn;
+	__be32		imm_inval_pkey;
+	u8		rsvd40[4];
+	__be32		byte_cnt;
+	__be64		timestamp;
+	__be32		sop_drop_qpn;
+	__be16		wqe_counter;
+	u8		signature;
+	u8		op_own;
+};
+
+struct mlx5_wqe_srq_next_seg {
+	u8			rsvd0[2];
+	__be16			next_wqe_index;
+	u8			signature;
+	u8			rsvd1[11];
+};
+
+union mlx5_ext_cqe {
+	struct ib_grh	grh;
+	u8		inl[64];
+};
+
+struct mlx5_cqe128 {
+	union mlx5_ext_cqe	inl_grh;
+	struct mlx5_cqe64	cqe64;
+};
+
+struct mlx5_srq_ctx {
+	u8			state_log_sz;
+	u8			rsvd0[3];
+	__be32			flags_xrcd;
+	__be32			pgoff_cqn;
+	u8			rsvd1[4];
+	u8			log_pg_sz;
+	u8			rsvd2[7];
+	__be32			pd;
+	__be16			lwm;
+	__be16			wqe_cnt;
+	u8			rsvd3[8];
+	__be64			db_record;
+};
+
+struct mlx5_create_srq_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	__be32			input_srqn;
+	u8			rsvd0[4];
+	struct mlx5_srq_ctx	ctx;
+	u8			rsvd1[208];
+	__be64			pas[0];
+};
+
+struct mlx5_create_srq_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	__be32			srqn;
+	u8			rsvd[4];
+};
+
+struct mlx5_destroy_srq_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	__be32			srqn;
+	u8			rsvd[4];
+};
+
+struct mlx5_destroy_srq_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvd[8];
+};
+
+struct mlx5_query_srq_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	__be32			srqn;
+	u8			rsvd0[4];
+};
+
+struct mlx5_query_srq_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvd0[8];
+	struct mlx5_srq_ctx	ctx;
+	u8			rsvd1[32];
+	__be64			pas[0];
+};
+
+struct mlx5_arm_srq_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	__be32			srqn;
+	__be16			rsvd;
+	__be16			lwm;
+};
+
+struct mlx5_arm_srq_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvd[8];
+};
+
+struct mlx5_cq_context {
+	u8			status;
+	u8			cqe_sz_flags;
+	u8			st;
+	u8			rsvd3;
+	u8			rsvd4[6];
+	__be16			page_offset;
+	__be32			log_sz_usr_page;
+	__be16			cq_period;
+	__be16			cq_max_count;
+	__be16			rsvd20;
+	__be16			c_eqn;
+	u8			log_pg_sz;
+	u8			rsvd25[7];
+	__be32			last_notified_index;
+	__be32			solicit_producer_index;
+	__be32			consumer_counter;
+	__be32			producer_counter;
+	u8			rsvd48[8];
+	__be64			db_record_addr;
+};
+
+struct mlx5_create_cq_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	__be32			input_cqn;
+	u8			rsvdx[4];
+	struct mlx5_cq_context	ctx;
+	u8			rsvd6[192];
+	__be64			pas[0];
+};
+
+struct mlx5_create_cq_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	__be32			cqn;
+	u8			rsvd0[4];
+};
+
+struct mlx5_destroy_cq_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	__be32			cqn;
+	u8			rsvd0[4];
+};
+
+struct mlx5_destroy_cq_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvd0[8];
+};
+
+struct mlx5_query_cq_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	__be32			cqn;
+	u8			rsvd0[4];
+};
+
+struct mlx5_query_cq_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvd0[8];
+	struct mlx5_cq_context	ctx;
+	u8			rsvd6[16];
+	__be64			pas[0];
+};
+
+struct mlx5_eq_context {
+	u8			status;
+	u8			ec_oi;
+	u8			st;
+	u8			rsvd2[7];
+	__be16			page_pffset;
+	__be32			log_sz_usr_page;
+	u8			rsvd3[7];
+	u8			intr;
+	u8			log_page_size;
+	u8			rsvd4[15];
+	__be32			consumer_counter;
+	__be32			produser_counter;
+	u8			rsvd5[16];
+};
+
+struct mlx5_create_eq_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	u8			rsvd0[3];
+	u8			input_eqn;
+	u8			rsvd1[4];
+	struct mlx5_eq_context	ctx;
+	u8			rsvd2[8];
+	__be64			events_mask;
+	u8			rsvd3[176];
+	__be64			pas[0];
+};
+
+struct mlx5_create_eq_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvd0[3];
+	u8			eq_number;
+	u8			rsvd1[4];
+};
+
+struct mlx5_destroy_eq_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	u8			rsvd0[3];
+	u8			eqn;
+	u8			rsvd1[4];
+};
+
+struct mlx5_destroy_eq_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvd[8];
+};
+
+struct mlx5_map_eq_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	__be64			mask;
+	u8			mu;
+	u8			rsvd0[2];
+	u8			eqn;
+	u8			rsvd1[24];
+};
+
+struct mlx5_map_eq_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvd[8];
+};
+
+struct mlx5_query_eq_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	u8			rsvd0[3];
+	u8			eqn;
+	u8			rsvd1[4];
+};
+
+struct mlx5_query_eq_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvd[8];
+	struct mlx5_eq_context	ctx;
+};
+
+struct mlx5_mkey_seg {
+	/* This is a two bit field occupying bits 31-30.
+	 * bit 31 is always 0,
+	 * bit 30 is zero for regular MRs and 1 (e.g free) for UMRs that do not have tanslation
+	 */
+	u8		status;
+	u8		pcie_control;
+	u8		flags;
+	u8		version;
+	__be32		qpn_mkey7_0;
+	u8		rsvd1[4];
+	__be32		flags_pd;
+	__be64		start_addr;
+	__be64		len;
+	__be32		bsfs_octo_size;
+	u8		rsvd2[16];
+	__be32		xlt_oct_size;
+	u8		rsvd3[3];
+	u8		log2_page_size;
+	u8		rsvd4[4];
+};
+
+struct mlx5_query_special_ctxs_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	u8			rsvd[8];
+};
+
+struct mlx5_query_special_ctxs_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	__be32			dump_fill_mkey;
+	__be32			reserved_lkey;
+};
+
+struct mlx5_create_mkey_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	__be32			input_mkey_index;
+	u8			rsvd0[4];
+	struct mlx5_mkey_seg	seg;
+	u8			rsvd1[16];
+	__be32			xlat_oct_act_size;
+	__be32			bsf_coto_act_size;
+	u8			rsvd2[168];
+	__be64			pas[0];
+};
+
+struct mlx5_create_mkey_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	__be32			mkey;
+	u8			rsvd[4];
+};
+
+struct mlx5_destroy_mkey_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	__be32			mkey;
+	u8			rsvd[4];
+};
+
+struct mlx5_destroy_mkey_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvd[8];
+};
+
+struct mlx5_query_mkey_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	__be32			mkey;
+};
+
+struct mlx5_query_mkey_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	__be64			pas[0];
+};
+
+struct mlx5_modify_mkey_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	__be32			mkey;
+	__be64			pas[0];
+};
+
+struct mlx5_modify_mkey_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+};
+
+struct mlx5_dump_mkey_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+};
+
+struct mlx5_dump_mkey_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	__be32			mkey;
+};
+
+struct mlx5_mad_ifc_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	__be16			remote_lid;
+	u8			rsvd0;
+	u8			port;
+	u8			rsvd1[4];
+	u8			data[256];
+};
+
+struct mlx5_mad_ifc_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvd[8];
+	u8			data[256];
+};
+
+struct mlx5_access_reg_mbox_in {
+	struct mlx5_inbox_hdr		hdr;
+	u8				rsvd0[2];
+	__be16				register_id;
+	__be32				arg;
+	__be32				data[0];
+};
+
+struct mlx5_access_reg_mbox_out {
+	struct mlx5_outbox_hdr		hdr;
+	u8				rsvd[8];
+	__be32				data[0];
+};
+
+#define MLX5_ATTR_EXTENDED_PORT_INFO	cpu_to_be16(0xff90)
+
+enum {
+	MLX_EXT_PORT_CAP_FLAG_EXTENDED_PORT_INFO	= 1 <<  0
+};
+
+#endif /* MLX5_DEVICE_H */

include/linux/mlx5/doorbell.h

@@ -0,0 +1,79 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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_DOORBELL_H
+#define MLX5_DOORBELL_H
+
+#define MLX5_BF_OFFSET	      0x800
+#define MLX5_CQ_DOORBELL      0x20
+
+#if BITS_PER_LONG == 64
+/* Assume that we can just write a 64-bit doorbell atomically.  s390
+ * actually doesn't have writeq() but S/390 systems don't even have
+ * PCI so we won't worry about it.
+ */
+
+#define MLX5_DECLARE_DOORBELL_LOCK(name)
+#define MLX5_INIT_DOORBELL_LOCK(ptr)    do { } while (0)
+#define MLX5_GET_DOORBELL_LOCK(ptr)      (NULL)
+
+static inline void mlx5_write64(__be32 val[2], void __iomem *dest,
+				spinlock_t *doorbell_lock)
+{
+	__raw_writeq(*(u64 *)val, dest);
+}
+
+#else
+
+/* Just fall back to a spinlock to protect the doorbell if
+ * BITS_PER_LONG is 32 -- there's no portable way to do atomic 64-bit
+ * MMIO writes.
+ */
+
+#define MLX5_DECLARE_DOORBELL_LOCK(name) spinlock_t name;
+#define MLX5_INIT_DOORBELL_LOCK(ptr)     spin_lock_init(ptr)
+#define MLX5_GET_DOORBELL_LOCK(ptr)      (ptr)
+
+static inline void mlx5_write64(__be32 val[2], void __iomem *dest,
+				spinlock_t *doorbell_lock)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(doorbell_lock, flags);
+	__raw_writel((__force u32) val[0], dest);
+	__raw_writel((__force u32) val[1], dest + 4);
+	spin_unlock_irqrestore(doorbell_lock, flags);
+}
+
+#endif
+
+#endif /* MLX5_DOORBELL_H */

include/linux/mlx5/driver.h

@@ -0,0 +1,769 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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_DRIVER_H
+#define MLX5_DRIVER_H
+
+#include <linux/kernel.h>
+#include <linux/completion.h>
+#include <linux/pci.h>
+#include <linux/spinlock_types.h>
+#include <linux/semaphore.h>
+#include <linux/vmalloc.h>
+#include <linux/radix-tree.h>
+#include <linux/mlx5/device.h>
+#include <linux/mlx5/doorbell.h>
+
+enum {
+	MLX5_BOARD_ID_LEN = 64,
+	MLX5_MAX_NAME_LEN = 16,
+};
+
+enum {
+	/* one minute for the sake of bringup. Generally, commands must always
+	 * complete and we may need to increase this timeout value
+	 */
+	MLX5_CMD_TIMEOUT_MSEC	= 7200 * 1000,
+	MLX5_CMD_WQ_MAX_NAME	= 32,
+};
+
+enum {
+	CMD_OWNER_SW		= 0x0,
+	CMD_OWNER_HW		= 0x1,
+	CMD_STATUS_SUCCESS	= 0,
+};
+
+enum mlx5_sqp_t {
+	MLX5_SQP_SMI		= 0,
+	MLX5_SQP_GSI		= 1,
+	MLX5_SQP_IEEE_1588	= 2,
+	MLX5_SQP_SNIFFER	= 3,
+	MLX5_SQP_SYNC_UMR	= 4,
+};
+
+enum {
+	MLX5_MAX_PORTS	= 2,
+};
+
+enum {
+	MLX5_EQ_VEC_PAGES	 = 0,
+	MLX5_EQ_VEC_CMD		 = 1,
+	MLX5_EQ_VEC_ASYNC	 = 2,
+	MLX5_EQ_VEC_COMP_BASE,
+};
+
+enum {
+	MLX5_MAX_EQ_NAME	= 20
+};
+
+enum {
+	MLX5_ATOMIC_MODE_IB_COMP	= 1 << 16,
+	MLX5_ATOMIC_MODE_CX		= 2 << 16,
+	MLX5_ATOMIC_MODE_8B		= 3 << 16,
+	MLX5_ATOMIC_MODE_16B		= 4 << 16,
+	MLX5_ATOMIC_MODE_32B		= 5 << 16,
+	MLX5_ATOMIC_MODE_64B		= 6 << 16,
+	MLX5_ATOMIC_MODE_128B		= 7 << 16,
+	MLX5_ATOMIC_MODE_256B		= 8 << 16,
+};
+
+enum {
+	MLX5_CMD_OP_QUERY_HCA_CAP		= 0x100,
+	MLX5_CMD_OP_QUERY_ADAPTER		= 0x101,
+	MLX5_CMD_OP_INIT_HCA			= 0x102,
+	MLX5_CMD_OP_TEARDOWN_HCA		= 0x103,
+	MLX5_CMD_OP_QUERY_PAGES			= 0x107,
+	MLX5_CMD_OP_MANAGE_PAGES		= 0x108,
+	MLX5_CMD_OP_SET_HCA_CAP			= 0x109,
+
+	MLX5_CMD_OP_CREATE_MKEY			= 0x200,
+	MLX5_CMD_OP_QUERY_MKEY			= 0x201,
+	MLX5_CMD_OP_DESTROY_MKEY		= 0x202,
+	MLX5_CMD_OP_QUERY_SPECIAL_CONTEXTS	= 0x203,
+
+	MLX5_CMD_OP_CREATE_EQ			= 0x301,
+	MLX5_CMD_OP_DESTROY_EQ			= 0x302,
+	MLX5_CMD_OP_QUERY_EQ			= 0x303,
+
+	MLX5_CMD_OP_CREATE_CQ			= 0x400,
+	MLX5_CMD_OP_DESTROY_CQ			= 0x401,
+	MLX5_CMD_OP_QUERY_CQ			= 0x402,
+	MLX5_CMD_OP_MODIFY_CQ			= 0x403,
+
+	MLX5_CMD_OP_CREATE_QP			= 0x500,
+	MLX5_CMD_OP_DESTROY_QP			= 0x501,
+	MLX5_CMD_OP_RST2INIT_QP			= 0x502,
+	MLX5_CMD_OP_INIT2RTR_QP			= 0x503,
+	MLX5_CMD_OP_RTR2RTS_QP			= 0x504,
+	MLX5_CMD_OP_RTS2RTS_QP			= 0x505,
+	MLX5_CMD_OP_SQERR2RTS_QP		= 0x506,
+	MLX5_CMD_OP_2ERR_QP			= 0x507,
+	MLX5_CMD_OP_RTS2SQD_QP			= 0x508,
+	MLX5_CMD_OP_SQD2RTS_QP			= 0x509,
+	MLX5_CMD_OP_2RST_QP			= 0x50a,
+	MLX5_CMD_OP_QUERY_QP			= 0x50b,
+	MLX5_CMD_OP_CONF_SQP			= 0x50c,
+	MLX5_CMD_OP_MAD_IFC			= 0x50d,
+	MLX5_CMD_OP_INIT2INIT_QP		= 0x50e,
+	MLX5_CMD_OP_SUSPEND_QP			= 0x50f,
+	MLX5_CMD_OP_UNSUSPEND_QP		= 0x510,
+	MLX5_CMD_OP_SQD2SQD_QP			= 0x511,
+	MLX5_CMD_OP_ALLOC_QP_COUNTER_SET	= 0x512,
+	MLX5_CMD_OP_DEALLOC_QP_COUNTER_SET	= 0x513,
+	MLX5_CMD_OP_QUERY_QP_COUNTER_SET	= 0x514,
+
+	MLX5_CMD_OP_CREATE_PSV			= 0x600,
+	MLX5_CMD_OP_DESTROY_PSV			= 0x601,
+	MLX5_CMD_OP_QUERY_PSV			= 0x602,
+	MLX5_CMD_OP_QUERY_SIG_RULE_TABLE	= 0x603,
+	MLX5_CMD_OP_QUERY_BLOCK_SIZE_TABLE	= 0x604,
+
+	MLX5_CMD_OP_CREATE_SRQ			= 0x700,
+	MLX5_CMD_OP_DESTROY_SRQ			= 0x701,
+	MLX5_CMD_OP_QUERY_SRQ			= 0x702,
+	MLX5_CMD_OP_ARM_RQ			= 0x703,
+	MLX5_CMD_OP_RESIZE_SRQ			= 0x704,
+
+	MLX5_CMD_OP_ALLOC_PD			= 0x800,
+	MLX5_CMD_OP_DEALLOC_PD			= 0x801,
+	MLX5_CMD_OP_ALLOC_UAR			= 0x802,
+	MLX5_CMD_OP_DEALLOC_UAR			= 0x803,
+
+	MLX5_CMD_OP_ATTACH_TO_MCG		= 0x806,
+	MLX5_CMD_OP_DETACH_FROM_MCG		= 0x807,
+
+
+	MLX5_CMD_OP_ALLOC_XRCD			= 0x80e,
+	MLX5_CMD_OP_DEALLOC_XRCD		= 0x80f,
+
+	MLX5_CMD_OP_ACCESS_REG			= 0x805,
+	MLX5_CMD_OP_MAX				= 0x810,
+};
+
+enum {
+	MLX5_REG_PCAP		 = 0x5001,
+	MLX5_REG_PMTU		 = 0x5003,
+	MLX5_REG_PTYS		 = 0x5004,
+	MLX5_REG_PAOS		 = 0x5006,
+	MLX5_REG_PMAOS		 = 0x5012,
+	MLX5_REG_PUDE		 = 0x5009,
+	MLX5_REG_PMPE		 = 0x5010,
+	MLX5_REG_PELC		 = 0x500e,
+	MLX5_REG_PMLP		 = 0, /* TBD */
+	MLX5_REG_NODE_DESC	 = 0x6001,
+	MLX5_REG_HOST_ENDIANNESS = 0x7004,
+};
+
+enum dbg_rsc_type {
+	MLX5_DBG_RSC_QP,
+	MLX5_DBG_RSC_EQ,
+	MLX5_DBG_RSC_CQ,
+};
+
+struct mlx5_field_desc {
+	struct dentry	       *dent;
+	int			i;
+};
+
+struct mlx5_rsc_debug {
+	struct mlx5_core_dev   *dev;
+	void		       *object;
+	enum dbg_rsc_type	type;
+	struct dentry	       *root;
+	struct mlx5_field_desc	fields[0];
+};
+
+enum mlx5_dev_event {
+	MLX5_DEV_EVENT_SYS_ERROR,
+	MLX5_DEV_EVENT_PORT_UP,
+	MLX5_DEV_EVENT_PORT_DOWN,
+	MLX5_DEV_EVENT_PORT_INITIALIZED,
+	MLX5_DEV_EVENT_LID_CHANGE,
+	MLX5_DEV_EVENT_PKEY_CHANGE,
+	MLX5_DEV_EVENT_GUID_CHANGE,
+	MLX5_DEV_EVENT_CLIENT_REREG,
+};
+
+struct mlx5_uuar_info {
+	struct mlx5_uar	       *uars;
+	int			num_uars;
+	int			num_low_latency_uuars;
+	unsigned long	       *bitmap;
+	unsigned int	       *count;
+	struct mlx5_bf	       *bfs;
+
+	/*
+	 * protect uuar allocation data structs
+	 */
+	struct mutex		lock;
+};
+
+struct mlx5_bf {
+	void __iomem	       *reg;
+	void __iomem	       *regreg;
+	int			buf_size;
+	struct mlx5_uar	       *uar;
+	unsigned long		offset;
+	int			need_lock;
+	/* protect blue flame buffer selection when needed
+	 */
+	spinlock_t		lock;
+
+	/* serialize 64 bit writes when done as two 32 bit accesses
+	 */
+	spinlock_t		lock32;
+	int			uuarn;
+};
+
+struct mlx5_cmd_first {
+	__be32		data[4];
+};
+
+struct mlx5_cmd_msg {
+	struct list_head		list;
+	struct cache_ent	       *cache;
+	u32				len;
+	struct mlx5_cmd_first		first;
+	struct mlx5_cmd_mailbox	       *next;
+};
+
+struct mlx5_cmd_debug {
+	struct dentry	       *dbg_root;
+	struct dentry	       *dbg_in;
+	struct dentry	       *dbg_out;
+	struct dentry	       *dbg_outlen;
+	struct dentry	       *dbg_status;
+	struct dentry	       *dbg_run;
+	void		       *in_msg;
+	void		       *out_msg;
+	u8			status;
+	u16			inlen;
+	u16			outlen;
+};
+
+struct cache_ent {
+	/* protect block chain allocations
+	 */
+	spinlock_t		lock;
+	struct list_head	head;
+};
+
+struct cmd_msg_cache {
+	struct cache_ent	large;
+	struct cache_ent	med;
+
+};
+
+struct mlx5_cmd_stats {
+	u64		sum;
+	u64		n;
+	struct dentry  *root;
+	struct dentry  *avg;
+	struct dentry  *count;
+	/* protect command average calculations */
+	spinlock_t	lock;
+};
+
+struct mlx5_cmd {
+	void	       *cmd_buf;
+	dma_addr_t	dma;
+	u16		cmdif_rev;
+	u8		log_sz;
+	u8		log_stride;
+	int		max_reg_cmds;
+	int		events;
+	u32 __iomem    *vector;
+
+	/* protect command queue allocations
+	 */
+	spinlock_t	alloc_lock;
+
+	/* protect token allocations
+	 */
+	spinlock_t	token_lock;
+	u8		token;
+	unsigned long	bitmask;
+	char		wq_name[MLX5_CMD_WQ_MAX_NAME];
+	struct workqueue_struct *wq;
+	struct semaphore sem;
+	struct semaphore pages_sem;
+	int	mode;
+	struct mlx5_cmd_work_ent *ent_arr[MLX5_MAX_COMMANDS];
+	struct pci_pool *pool;
+	struct mlx5_cmd_debug dbg;
+	struct cmd_msg_cache cache;
+	int checksum_disabled;
+	struct mlx5_cmd_stats stats[MLX5_CMD_OP_MAX];
+};
+
+struct mlx5_port_caps {
+	int	gid_table_len;
+	int	pkey_table_len;
+};
+
+struct mlx5_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;
+	u32	max_cqes;
+	int	max_wqes;
+	int	max_sq_desc_sz;
+	int	max_rq_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;
+	int	max_srq_wqes;
+	int	bf_reg_size;
+	int	bf_regs_per_page;
+	struct mlx5_port_caps	port[MLX5_MAX_PORTS];
+	u8			ext_port_cap[MLX5_MAX_PORTS];
+	int	max_vf;
+	u32	reserved_lkey;
+	u8	local_ca_ack_delay;
+	u8	log_max_mcg;
+	u16	max_qp_mcg;
+	int	min_page_sz;
+};
+
+struct mlx5_cmd_mailbox {
+	void	       *buf;
+	dma_addr_t	dma;
+	struct mlx5_cmd_mailbox *next;
+};
+
+struct mlx5_buf_list {
+	void		       *buf;
+	dma_addr_t		map;
+};
+
+struct mlx5_buf {
+	struct mlx5_buf_list	direct;
+	struct mlx5_buf_list   *page_list;
+	int			nbufs;
+	int			npages;
+	int			page_shift;
+	int			size;
+};
+
+struct mlx5_eq {
+	struct mlx5_core_dev   *dev;
+	__be32 __iomem	       *doorbell;
+	u32			cons_index;
+	struct mlx5_buf		buf;
+	int			size;
+	u8			irqn;
+	u8			eqn;
+	int			nent;
+	u64			mask;
+	char			name[MLX5_MAX_EQ_NAME];
+	struct list_head	list;
+	int			index;
+	struct mlx5_rsc_debug	*dbg;
+};
+
+
+struct mlx5_core_mr {
+	u64			iova;
+	u64			size;
+	u32			key;
+	u32			pd;
+	u32			access;
+};
+
+struct mlx5_core_srq {
+	u32		srqn;
+	int		max;
+	int		max_gs;
+	int		max_avail_gather;
+	int		wqe_shift;
+	void (*event)	(struct mlx5_core_srq *, enum mlx5_event);
+
+	atomic_t		refcount;
+	struct completion	free;
+};
+
+struct mlx5_eq_table {
+	void __iomem	       *update_ci;
+	void __iomem	       *update_arm_ci;
+	struct list_head       *comp_eq_head;
+	struct mlx5_eq		pages_eq;
+	struct mlx5_eq		async_eq;
+	struct mlx5_eq		cmd_eq;
+	struct msix_entry	*msix_arr;
+	int			num_comp_vectors;
+	/* protect EQs list
+	 */
+	spinlock_t		lock;
+};
+
+struct mlx5_uar {
+	u32			index;
+	struct list_head	bf_list;
+	unsigned		free_bf_bmap;
+	void __iomem	       *wc_map;
+	void __iomem	       *map;
+};
+
+
+struct mlx5_core_health {
+	struct health_buffer __iomem   *health;
+	__be32 __iomem		       *health_counter;
+	struct timer_list		timer;
+	struct list_head		list;
+	u32				prev;
+	int				miss_counter;
+};
+
+struct mlx5_cq_table {
+	/* protect radix tree
+	 */
+	spinlock_t		lock;
+	struct radix_tree_root	tree;
+};
+
+struct mlx5_qp_table {
+	/* protect radix tree
+	 */
+	spinlock_t		lock;
+	struct radix_tree_root	tree;
+};
+
+struct mlx5_srq_table {
+	/* protect radix tree
+	 */
+	spinlock_t		lock;
+	struct radix_tree_root	tree;
+};
+
+struct mlx5_priv {
+	char			name[MLX5_MAX_NAME_LEN];
+	struct mlx5_eq_table	eq_table;
+	struct mlx5_uuar_info	uuari;
+	MLX5_DECLARE_DOORBELL_LOCK(cq_uar_lock);
+
+	/* pages stuff */
+	struct workqueue_struct *pg_wq;
+	struct rb_root		page_root;
+	int			fw_pages;
+	int			reg_pages;
+
+	struct mlx5_core_health health;
+
+	struct mlx5_srq_table	srq_table;
+
+	/* start: qp staff */
+	struct mlx5_qp_table	qp_table;
+	struct dentry	       *qp_debugfs;
+	struct dentry	       *eq_debugfs;
+	struct dentry	       *cq_debugfs;
+	struct dentry	       *cmdif_debugfs;
+	/* end: qp staff */
+
+	/* start: cq staff */
+	struct mlx5_cq_table	cq_table;
+	/* end: cq staff */
+
+	/* start: alloc staff */
+	struct mutex            pgdir_mutex;
+	struct list_head        pgdir_list;
+	/* end: alloc staff */
+	struct dentry	       *dbg_root;
+
+	/* protect mkey key part */
+	spinlock_t		mkey_lock;
+	u8			mkey_key;
+};
+
+struct mlx5_core_dev {
+	struct pci_dev	       *pdev;
+	u8			rev_id;
+	char			board_id[MLX5_BOARD_ID_LEN];
+	struct mlx5_cmd		cmd;
+	struct mlx5_caps	caps;
+	phys_addr_t		iseg_base;
+	struct mlx5_init_seg __iomem *iseg;
+	void			(*event) (struct mlx5_core_dev *dev,
+					  enum mlx5_dev_event event,
+					  void *data);
+	struct mlx5_priv	priv;
+	struct mlx5_profile	*profile;
+	atomic_t		num_qps;
+};
+
+struct mlx5_db {
+	__be32			*db;
+	union {
+		struct mlx5_db_pgdir		*pgdir;
+		struct mlx5_ib_user_db_page	*user_page;
+	}			u;
+	dma_addr_t		dma;
+	int			index;
+};
+
+enum {
+	MLX5_DB_PER_PAGE = PAGE_SIZE / L1_CACHE_BYTES,
+};
+
+enum {
+	MLX5_COMP_EQ_SIZE = 1024,
+};
+
+struct mlx5_db_pgdir {
+	struct list_head	list;
+	DECLARE_BITMAP(bitmap, MLX5_DB_PER_PAGE);
+	__be32		       *db_page;
+	dma_addr_t		db_dma;
+};
+
+typedef void (*mlx5_cmd_cbk_t)(int status, void *context);
+
+struct mlx5_cmd_work_ent {
+	struct mlx5_cmd_msg    *in;
+	struct mlx5_cmd_msg    *out;
+	mlx5_cmd_cbk_t		callback;
+	void		       *context;
+	int idx;
+	struct completion	done;
+	struct mlx5_cmd        *cmd;
+	struct work_struct	work;
+	struct mlx5_cmd_layout *lay;
+	int			ret;
+	int			page_queue;
+	u8			status;
+	u8			token;
+	struct timespec		ts1;
+	struct timespec		ts2;
+};
+
+struct mlx5_pas {
+	u64	pa;
+	u8	log_sz;
+};
+
+static inline void *mlx5_buf_offset(struct mlx5_buf *buf, int offset)
+{
+	if (likely(BITS_PER_LONG == 64 || buf->nbufs == 1))
+		return buf->direct.buf + offset;
+	else
+		return buf->page_list[offset >> PAGE_SHIFT].buf +
+			(offset & (PAGE_SIZE - 1));
+}
+
+extern struct workqueue_struct *mlx5_core_wq;
+
+#define STRUCT_FIELD(header, field) \
+	.struct_offset_bytes = offsetof(struct ib_unpacked_ ## header, field),      \
+	.struct_size_bytes   = sizeof((struct ib_unpacked_ ## header *)0)->field
+
+struct ib_field {
+	size_t struct_offset_bytes;
+	size_t struct_size_bytes;
+	int    offset_bits;
+	int    size_bits;
+};
+
+static inline struct mlx5_core_dev *pci2mlx5_core_dev(struct pci_dev *pdev)
+{
+	return pci_get_drvdata(pdev);
+}
+
+extern struct dentry *mlx5_debugfs_root;
+
+static inline u16 fw_rev_maj(struct mlx5_core_dev *dev)
+{
+	return ioread32be(&dev->iseg->fw_rev) & 0xffff;
+}
+
+static inline u16 fw_rev_min(struct mlx5_core_dev *dev)
+{
+	return ioread32be(&dev->iseg->fw_rev) >> 16;
+}
+
+static inline u16 fw_rev_sub(struct mlx5_core_dev *dev)
+{
+	return ioread32be(&dev->iseg->cmdif_rev_fw_sub) & 0xffff;
+}
+
+static inline u16 cmdif_rev(struct mlx5_core_dev *dev)
+{
+	return ioread32be(&dev->iseg->cmdif_rev_fw_sub) >> 16;
+}
+
+static inline void *mlx5_vzalloc(unsigned long size)
+{
+	void *rtn;
+
+	rtn = kzalloc(size, GFP_KERNEL | __GFP_NOWARN);
+	if (!rtn)
+		rtn = vzalloc(size);
+	return rtn;
+}
+
+static inline void mlx5_vfree(const void *addr)
+{
+	if (addr && is_vmalloc_addr(addr))
+		vfree(addr);
+	else
+		kfree(addr);
+}
+
+int mlx5_dev_init(struct mlx5_core_dev *dev, struct pci_dev *pdev);
+void mlx5_dev_cleanup(struct mlx5_core_dev *dev);
+int mlx5_cmd_init(struct mlx5_core_dev *dev);
+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_cmd_exec(struct mlx5_core_dev *dev, void *in, int in_size, void *out,
+		  int out_size);
+int mlx5_cmd_alloc_uar(struct mlx5_core_dev *dev, u32 *uarn);
+int mlx5_cmd_free_uar(struct mlx5_core_dev *dev, u32 uarn);
+int mlx5_alloc_uuars(struct mlx5_core_dev *dev, struct mlx5_uuar_info *uuari);
+int mlx5_free_uuars(struct mlx5_core_dev *dev, struct mlx5_uuar_info *uuari);
+void mlx5_health_cleanup(void);
+void  __init mlx5_health_init(void);
+void mlx5_start_health_poll(struct mlx5_core_dev *dev);
+void mlx5_stop_health_poll(struct mlx5_core_dev *dev);
+int mlx5_buf_alloc(struct mlx5_core_dev *dev, int size, int max_direct,
+		   struct mlx5_buf *buf);
+void mlx5_buf_free(struct mlx5_core_dev *dev, struct mlx5_buf *buf);
+struct mlx5_cmd_mailbox *mlx5_alloc_cmd_mailbox_chain(struct mlx5_core_dev *dev,
+						      gfp_t flags, int npages);
+void mlx5_free_cmd_mailbox_chain(struct mlx5_core_dev *dev,
+				 struct mlx5_cmd_mailbox *head);
+int mlx5_core_create_srq(struct mlx5_core_dev *dev, struct mlx5_core_srq *srq,
+			 struct mlx5_create_srq_mbox_in *in, int inlen);
+int mlx5_core_destroy_srq(struct mlx5_core_dev *dev, struct mlx5_core_srq *srq);
+int mlx5_core_query_srq(struct mlx5_core_dev *dev, struct mlx5_core_srq *srq,
+			struct mlx5_query_srq_mbox_out *out);
+int mlx5_core_arm_srq(struct mlx5_core_dev *dev, struct mlx5_core_srq *srq,
+		      u16 lwm, int is_srq);
+int mlx5_core_create_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mr *mr,
+			  struct mlx5_create_mkey_mbox_in *in, int inlen);
+int mlx5_core_destroy_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mr *mr);
+int mlx5_core_query_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mr *mr,
+			 struct mlx5_query_mkey_mbox_out *out, int outlen);
+int mlx5_core_dump_fill_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mr *mr,
+			     u32 *mkey);
+int mlx5_core_alloc_pd(struct mlx5_core_dev *dev, u32 *pdn);
+int mlx5_core_dealloc_pd(struct mlx5_core_dev *dev, u32 pdn);
+int mlx5_core_mad_ifc(struct mlx5_core_dev *dev, void *inb, void *outb,
+		      u16 opmod, int port);
+void mlx5_pagealloc_init(struct mlx5_core_dev *dev);
+void mlx5_pagealloc_cleanup(struct mlx5_core_dev *dev);
+int mlx5_pagealloc_start(struct mlx5_core_dev *dev);
+void mlx5_pagealloc_stop(struct mlx5_core_dev *dev);
+void mlx5_core_req_pages_handler(struct mlx5_core_dev *dev, u16 func_id,
+				 s16 npages);
+int mlx5_satisfy_startup_pages(struct mlx5_core_dev *dev);
+int mlx5_reclaim_startup_pages(struct mlx5_core_dev *dev);
+void mlx5_register_debugfs(void);
+void mlx5_unregister_debugfs(void);
+int mlx5_eq_init(struct mlx5_core_dev *dev);
+void mlx5_eq_cleanup(struct mlx5_core_dev *dev);
+void mlx5_fill_page_array(struct mlx5_buf *buf, __be64 *pas);
+void mlx5_cq_completion(struct mlx5_core_dev *dev, u32 cqn);
+void mlx5_qp_event(struct mlx5_core_dev *dev, u32 qpn, int event_type);
+void mlx5_srq_event(struct mlx5_core_dev *dev, u32 srqn, int event_type);
+struct mlx5_core_srq *mlx5_core_get_srq(struct mlx5_core_dev *dev, u32 srqn);
+void mlx5_cmd_comp_handler(struct mlx5_core_dev *dev, unsigned long vector);
+void mlx5_cq_event(struct mlx5_core_dev *dev, u32 cqn, int event_type);
+int mlx5_create_map_eq(struct mlx5_core_dev *dev, struct mlx5_eq *eq, u8 vecidx,
+		       int nent, u64 mask, const char *name, struct mlx5_uar *uar);
+int mlx5_destroy_unmap_eq(struct mlx5_core_dev *dev, struct mlx5_eq *eq);
+int mlx5_start_eqs(struct mlx5_core_dev *dev);
+int mlx5_stop_eqs(struct mlx5_core_dev *dev);
+int mlx5_core_attach_mcg(struct mlx5_core_dev *dev, union ib_gid *mgid, u32 qpn);
+int mlx5_core_detach_mcg(struct mlx5_core_dev *dev, union ib_gid *mgid, u32 qpn);
+
+int mlx5_qp_debugfs_init(struct mlx5_core_dev *dev);
+void mlx5_qp_debugfs_cleanup(struct mlx5_core_dev *dev);
+int mlx5_core_access_reg(struct mlx5_core_dev *dev, void *data_in,
+			 int size_in, void *data_out, int size_out,
+			 u16 reg_num, int arg, int write);
+int mlx5_set_port_caps(struct mlx5_core_dev *dev, int port_num, u32 caps);
+
+int mlx5_debug_eq_add(struct mlx5_core_dev *dev, struct mlx5_eq *eq);
+void mlx5_debug_eq_remove(struct mlx5_core_dev *dev, struct mlx5_eq *eq);
+int mlx5_core_eq_query(struct mlx5_core_dev *dev, struct mlx5_eq *eq,
+		       struct mlx5_query_eq_mbox_out *out, int outlen);
+int mlx5_eq_debugfs_init(struct mlx5_core_dev *dev);
+void mlx5_eq_debugfs_cleanup(struct mlx5_core_dev *dev);
+int mlx5_cq_debugfs_init(struct mlx5_core_dev *dev);
+void mlx5_cq_debugfs_cleanup(struct mlx5_core_dev *dev);
+int mlx5_db_alloc(struct mlx5_core_dev *dev, struct mlx5_db *db);
+void mlx5_db_free(struct mlx5_core_dev *dev, struct mlx5_db *db);
+
+typedef void (*health_handler_t)(struct pci_dev *pdev, struct health_buffer __iomem *buf, int size);
+int mlx5_register_health_report_handler(health_handler_t handler);
+void mlx5_unregister_health_report_handler(void);
+const char *mlx5_command_str(int command);
+int mlx5_cmdif_debugfs_init(struct mlx5_core_dev *dev);
+void mlx5_cmdif_debugfs_cleanup(struct mlx5_core_dev *dev);
+
+static inline u32 mlx5_mkey_to_idx(u32 mkey)
+{
+	return mkey >> 8;
+}
+
+static inline u32 mlx5_idx_to_mkey(u32 mkey_idx)
+{
+	return mkey_idx << 8;
+}
+
+enum {
+	MLX5_PROF_MASK_QP_SIZE		= (u64)1 << 0,
+	MLX5_PROF_MASK_CMDIF_CSUM	= (u64)1 << 1,
+	MLX5_PROF_MASK_MR_CACHE		= (u64)1 << 2,
+};
+
+enum {
+	MAX_MR_CACHE_ENTRIES    = 16,
+};
+
+struct mlx5_profile {
+	u64	mask;
+	u32	log_max_qp;
+	int	cmdif_csum;
+	struct {
+		int	size;
+		int	limit;
+	} mr_cache[MAX_MR_CACHE_ENTRIES];
+};
+
+#endif /* MLX5_DRIVER_H */

include/linux/mlx5/qp.h

@@ -0,0 +1,467 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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_QP_H
+#define MLX5_QP_H
+
+#include <linux/mlx5/device.h>
+#include <linux/mlx5/driver.h>
+
+#define MLX5_INVALID_LKEY	0x100
+
+enum mlx5_qp_optpar {
+	MLX5_QP_OPTPAR_ALT_ADDR_PATH		= 1 << 0,
+	MLX5_QP_OPTPAR_RRE			= 1 << 1,
+	MLX5_QP_OPTPAR_RAE			= 1 << 2,
+	MLX5_QP_OPTPAR_RWE			= 1 << 3,
+	MLX5_QP_OPTPAR_PKEY_INDEX		= 1 << 4,
+	MLX5_QP_OPTPAR_Q_KEY			= 1 << 5,
+	MLX5_QP_OPTPAR_RNR_TIMEOUT		= 1 << 6,
+	MLX5_QP_OPTPAR_PRIMARY_ADDR_PATH	= 1 << 7,
+	MLX5_QP_OPTPAR_SRA_MAX			= 1 << 8,
+	MLX5_QP_OPTPAR_RRA_MAX			= 1 << 9,
+	MLX5_QP_OPTPAR_PM_STATE			= 1 << 10,
+	MLX5_QP_OPTPAR_RETRY_COUNT		= 1 << 12,
+	MLX5_QP_OPTPAR_RNR_RETRY		= 1 << 13,
+	MLX5_QP_OPTPAR_ACK_TIMEOUT		= 1 << 14,
+	MLX5_QP_OPTPAR_PRI_PORT			= 1 << 16,
+	MLX5_QP_OPTPAR_SRQN			= 1 << 18,
+	MLX5_QP_OPTPAR_CQN_RCV			= 1 << 19,
+	MLX5_QP_OPTPAR_DC_HS			= 1 << 20,
+	MLX5_QP_OPTPAR_DC_KEY			= 1 << 21,
+};
+
+enum mlx5_qp_state {
+	MLX5_QP_STATE_RST			= 0,
+	MLX5_QP_STATE_INIT			= 1,
+	MLX5_QP_STATE_RTR			= 2,
+	MLX5_QP_STATE_RTS			= 3,
+	MLX5_QP_STATE_SQER			= 4,
+	MLX5_QP_STATE_SQD			= 5,
+	MLX5_QP_STATE_ERR			= 6,
+	MLX5_QP_STATE_SQ_DRAINING		= 7,
+	MLX5_QP_STATE_SUSPENDED			= 9,
+	MLX5_QP_NUM_STATE
+};
+
+enum {
+	MLX5_QP_ST_RC				= 0x0,
+	MLX5_QP_ST_UC				= 0x1,
+	MLX5_QP_ST_UD				= 0x2,
+	MLX5_QP_ST_XRC				= 0x3,
+	MLX5_QP_ST_MLX				= 0x4,
+	MLX5_QP_ST_DCI				= 0x5,
+	MLX5_QP_ST_DCT				= 0x6,
+	MLX5_QP_ST_QP0				= 0x7,
+	MLX5_QP_ST_QP1				= 0x8,
+	MLX5_QP_ST_RAW_ETHERTYPE		= 0x9,
+	MLX5_QP_ST_RAW_IPV6			= 0xa,
+	MLX5_QP_ST_SNIFFER			= 0xb,
+	MLX5_QP_ST_SYNC_UMR			= 0xe,
+	MLX5_QP_ST_PTP_1588			= 0xd,
+	MLX5_QP_ST_REG_UMR			= 0xc,
+	MLX5_QP_ST_MAX
+};
+
+enum {
+	MLX5_QP_PM_MIGRATED			= 0x3,
+	MLX5_QP_PM_ARMED			= 0x0,
+	MLX5_QP_PM_REARM			= 0x1
+};
+
+enum {
+	MLX5_NON_ZERO_RQ	= 0 << 24,
+	MLX5_SRQ_RQ		= 1 << 24,
+	MLX5_CRQ_RQ		= 2 << 24,
+	MLX5_ZERO_LEN_RQ	= 3 << 24
+};
+
+enum {
+	/* params1 */
+	MLX5_QP_BIT_SRE				= 1 << 15,
+	MLX5_QP_BIT_SWE				= 1 << 14,
+	MLX5_QP_BIT_SAE				= 1 << 13,
+	/* params2 */
+	MLX5_QP_BIT_RRE				= 1 << 15,
+	MLX5_QP_BIT_RWE				= 1 << 14,
+	MLX5_QP_BIT_RAE				= 1 << 13,
+	MLX5_QP_BIT_RIC				= 1 <<	4,
+};
+
+enum {
+	MLX5_WQE_CTRL_CQ_UPDATE		= 2 << 2,
+	MLX5_WQE_CTRL_SOLICITED		= 1 << 1,
+};
+
+enum {
+	MLX5_SEND_WQE_BB	= 64,
+};
+
+enum {
+	MLX5_WQE_FMR_PERM_LOCAL_READ	= 1 << 27,
+	MLX5_WQE_FMR_PERM_LOCAL_WRITE	= 1 << 28,
+	MLX5_WQE_FMR_PERM_REMOTE_READ	= 1 << 29,
+	MLX5_WQE_FMR_PERM_REMOTE_WRITE	= 1 << 30,
+	MLX5_WQE_FMR_PERM_ATOMIC	= 1 << 31
+};
+
+enum {
+	MLX5_FENCE_MODE_NONE			= 0 << 5,
+	MLX5_FENCE_MODE_INITIATOR_SMALL		= 1 << 5,
+	MLX5_FENCE_MODE_STRONG_ORDERING		= 3 << 5,
+	MLX5_FENCE_MODE_SMALL_AND_FENCE		= 4 << 5,
+};
+
+enum {
+	MLX5_QP_LAT_SENSITIVE	= 1 << 28,
+	MLX5_QP_ENABLE_SIG	= 1 << 31,
+};
+
+enum {
+	MLX5_RCV_DBR	= 0,
+	MLX5_SND_DBR	= 1,
+};
+
+struct mlx5_wqe_fmr_seg {
+	__be32			flags;
+	__be32			mem_key;
+	__be64			buf_list;
+	__be64			start_addr;
+	__be64			reg_len;
+	__be32			offset;
+	__be32			page_size;
+	u32			reserved[2];
+};
+
+struct mlx5_wqe_ctrl_seg {
+	__be32			opmod_idx_opcode;
+	__be32			qpn_ds;
+	u8			signature;
+	u8			rsvd[2];
+	u8			fm_ce_se;
+	__be32			imm;
+};
+
+struct mlx5_wqe_xrc_seg {
+	__be32			xrc_srqn;
+	u8			rsvd[12];
+};
+
+struct mlx5_wqe_masked_atomic_seg {
+	__be64			swap_add;
+	__be64			compare;
+	__be64			swap_add_mask;
+	__be64			compare_mask;
+};
+
+struct mlx5_av {
+	union {
+		struct {
+			__be32	qkey;
+			__be32	reserved;
+		} qkey;
+		__be64	dc_key;
+	} key;
+	__be32	dqp_dct;
+	u8	stat_rate_sl;
+	u8	fl_mlid;
+	__be16	rlid;
+	u8	reserved0[10];
+	u8	tclass;
+	u8	hop_limit;
+	__be32	grh_gid_fl;
+	u8	rgid[16];
+};
+
+struct mlx5_wqe_datagram_seg {
+	struct mlx5_av	av;
+};
+
+struct mlx5_wqe_raddr_seg {
+	__be64			raddr;
+	__be32			rkey;
+	u32			reserved;
+};
+
+struct mlx5_wqe_atomic_seg {
+	__be64			swap_add;
+	__be64			compare;
+};
+
+struct mlx5_wqe_data_seg {
+	__be32			byte_count;
+	__be32			lkey;
+	__be64			addr;
+};
+
+struct mlx5_wqe_umr_ctrl_seg {
+	u8		flags;
+	u8		rsvd0[3];
+	__be16		klm_octowords;
+	__be16		bsf_octowords;
+	__be64		mkey_mask;
+	u8		rsvd1[32];
+};
+
+struct mlx5_seg_set_psv {
+	__be32		psv_num;
+	__be16		syndrome;
+	__be16		status;
+	__be32		transient_sig;
+	__be32		ref_tag;
+};
+
+struct mlx5_seg_get_psv {
+	u8		rsvd[19];
+	u8		num_psv;
+	__be32		l_key;
+	__be64		va;
+	__be32		psv_index[4];
+};
+
+struct mlx5_seg_check_psv {
+	u8		rsvd0[2];
+	__be16		err_coalescing_op;
+	u8		rsvd1[2];
+	__be16		xport_err_op;
+	u8		rsvd2[2];
+	__be16		xport_err_mask;
+	u8		rsvd3[7];
+	u8		num_psv;
+	__be32		l_key;
+	__be64		va;
+	__be32		psv_index[4];
+};
+
+struct mlx5_rwqe_sig {
+	u8	rsvd0[4];
+	u8	signature;
+	u8	rsvd1[11];
+};
+
+struct mlx5_wqe_signature_seg {
+	u8	rsvd0[4];
+	u8	signature;
+	u8	rsvd1[11];
+};
+
+struct mlx5_wqe_inline_seg {
+	__be32	byte_count;
+};
+
+struct mlx5_core_qp {
+	void (*event)		(struct mlx5_core_qp *, int);
+	int			qpn;
+	atomic_t		refcount;
+	struct completion	free;
+	struct mlx5_rsc_debug	*dbg;
+	int			pid;
+};
+
+struct mlx5_qp_path {
+	u8			fl;
+	u8			rsvd3;
+	u8			free_ar;
+	u8			pkey_index;
+	u8			rsvd0;
+	u8			grh_mlid;
+	__be16			rlid;
+	u8			ackto_lt;
+	u8			mgid_index;
+	u8			static_rate;
+	u8			hop_limit;
+	__be32			tclass_flowlabel;
+	u8			rgid[16];
+	u8			rsvd1[4];
+	u8			sl;
+	u8			port;
+	u8			rsvd2[6];
+};
+
+struct mlx5_qp_context {
+	__be32			flags;
+	__be32			flags_pd;
+	u8			mtu_msgmax;
+	u8			rq_size_stride;
+	__be16			sq_crq_size;
+	__be32			qp_counter_set_usr_page;
+	__be32			wire_qpn;
+	__be32			log_pg_sz_remote_qpn;
+	struct			mlx5_qp_path pri_path;
+	struct			mlx5_qp_path alt_path;
+	__be32			params1;
+	u8			reserved2[4];
+	__be32			next_send_psn;
+	__be32			cqn_send;
+	u8			reserved3[8];
+	__be32			last_acked_psn;
+	__be32			ssn;
+	__be32			params2;
+	__be32			rnr_nextrecvpsn;
+	__be32			xrcd;
+	__be32			cqn_recv;
+	__be64			db_rec_addr;
+	__be32			qkey;
+	__be32			rq_type_srqn;
+	__be32			rmsn;
+	__be16			hw_sq_wqe_counter;
+	__be16			sw_sq_wqe_counter;
+	__be16			hw_rcyclic_byte_counter;
+	__be16			hw_rq_counter;
+	__be16			sw_rcyclic_byte_counter;
+	__be16			sw_rq_counter;
+	u8			rsvd0[5];
+	u8			cgs;
+	u8			cs_req;
+	u8			cs_res;
+	__be64			dc_access_key;
+	u8			rsvd1[24];
+};
+
+struct mlx5_create_qp_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	__be32			input_qpn;
+	u8			rsvd0[4];
+	__be32			opt_param_mask;
+	u8			rsvd1[4];
+	struct mlx5_qp_context	ctx;
+	u8			rsvd3[16];
+	__be64			pas[0];
+};
+
+struct mlx5_create_qp_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	__be32			qpn;
+	u8			rsvd0[4];
+};
+
+struct mlx5_destroy_qp_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	__be32			qpn;
+	u8			rsvd0[4];
+};
+
+struct mlx5_destroy_qp_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvd0[8];
+};
+
+struct mlx5_modify_qp_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	__be32			qpn;
+	u8			rsvd1[4];
+	__be32			optparam;
+	u8			rsvd0[4];
+	struct mlx5_qp_context	ctx;
+};
+
+struct mlx5_modify_qp_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvd0[8];
+};
+
+struct mlx5_query_qp_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	__be32			qpn;
+	u8			rsvd[4];
+};
+
+struct mlx5_query_qp_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvd1[8];
+	__be32			optparam;
+	u8			rsvd0[4];
+	struct mlx5_qp_context	ctx;
+	u8			rsvd2[16];
+	__be64			pas[0];
+};
+
+struct mlx5_conf_sqp_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	__be32			qpn;
+	u8			rsvd[3];
+	u8			type;
+};
+
+struct mlx5_conf_sqp_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvd[8];
+};
+
+struct mlx5_alloc_xrcd_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	u8			rsvd[8];
+};
+
+struct mlx5_alloc_xrcd_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	__be32			xrcdn;
+	u8			rsvd[4];
+};
+
+struct mlx5_dealloc_xrcd_mbox_in {
+	struct mlx5_inbox_hdr	hdr;
+	__be32			xrcdn;
+	u8			rsvd[4];
+};
+
+struct mlx5_dealloc_xrcd_mbox_out {
+	struct mlx5_outbox_hdr	hdr;
+	u8			rsvd[8];
+};
+
+static inline struct mlx5_core_qp *__mlx5_qp_lookup(struct mlx5_core_dev *dev, u32 qpn)
+{
+	return radix_tree_lookup(&dev->priv.qp_table.tree, qpn);
+}
+
+int mlx5_core_create_qp(struct mlx5_core_dev *dev,
+			struct mlx5_core_qp *qp,
+			struct mlx5_create_qp_mbox_in *in,
+			int inlen);
+int mlx5_core_qp_modify(struct mlx5_core_dev *dev, enum mlx5_qp_state cur_state,
+			enum mlx5_qp_state new_state,
+			struct mlx5_modify_qp_mbox_in *in, int sqd_event,
+			struct mlx5_core_qp *qp);
+int mlx5_core_destroy_qp(struct mlx5_core_dev *dev,
+			 struct mlx5_core_qp *qp);
+int mlx5_core_qp_query(struct mlx5_core_dev *dev, struct mlx5_core_qp *qp,
+		       struct mlx5_query_qp_mbox_out *out, int outlen);
+
+int mlx5_core_xrcd_alloc(struct mlx5_core_dev *dev, u32 *xrcdn);
+int mlx5_core_xrcd_dealloc(struct mlx5_core_dev *dev, u32 xrcdn);
+void mlx5_init_qp_table(struct mlx5_core_dev *dev);
+void mlx5_cleanup_qp_table(struct mlx5_core_dev *dev);
+int mlx5_debug_qp_add(struct mlx5_core_dev *dev, struct mlx5_core_qp *qp);
+void mlx5_debug_qp_remove(struct mlx5_core_dev *dev, struct mlx5_core_qp *qp);
+
+#endif /* MLX5_QP_H */

include/linux/mlx5/srq.h

@@ -0,0 +1,41 @@
+/*
+ * Copyright (c) 2013, Mellanox Technologies inc.  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_SRQ_H
+#define MLX5_SRQ_H
+
+#include <linux/mlx5/driver.h>
+
+void mlx5_init_srq_table(struct mlx5_core_dev *dev);
+void mlx5_cleanup_srq_table(struct mlx5_core_dev *dev);
+
+#endif /* MLX5_SRQ_H */

include/rdma/ib_verbs.h

@@ -610,7 +610,21 @@ enum ib_qp_type {
 	IB_QPT_RAW_PACKET = 8,
 	IB_QPT_XRC_INI = 9,
 	IB_QPT_XRC_TGT,
-	IB_QPT_MAX
+	IB_QPT_MAX,
+	/* Reserve a range for qp types internal to the low level driver.
+	 * These qp types will not be visible at the IB core layer, so the
+	 * IB_QPT_MAX usages should not be affected in the core layer
+	 */
+	IB_QPT_RESERVED1 = 0x1000,
+	IB_QPT_RESERVED2,
+	IB_QPT_RESERVED3,
+	IB_QPT_RESERVED4,
+	IB_QPT_RESERVED5,
+	IB_QPT_RESERVED6,
+	IB_QPT_RESERVED7,
+	IB_QPT_RESERVED8,
+	IB_QPT_RESERVED9,
+	IB_QPT_RESERVED10,
 };
 
 enum ib_qp_create_flags {
@@ -766,6 +780,19 @@ enum ib_wr_opcode {
 	IB_WR_MASKED_ATOMIC_CMP_AND_SWP,
 	IB_WR_MASKED_ATOMIC_FETCH_AND_ADD,
 	IB_WR_BIND_MW,
+	/* reserve values for low level drivers' internal use.
+	 * These values will not be used at all in the ib core layer.
+	 */
+	IB_WR_RESERVED1 = 0xf0,
+	IB_WR_RESERVED2,
+	IB_WR_RESERVED3,
+	IB_WR_RESERVED4,
+	IB_WR_RESERVED5,
+	IB_WR_RESERVED6,
+	IB_WR_RESERVED7,
+	IB_WR_RESERVED8,
+	IB_WR_RESERVED9,
+	IB_WR_RESERVED10,
 };
 
 enum ib_send_flags {
@@ -773,7 +800,11 @@ enum ib_send_flags {
 	IB_SEND_SIGNALED	= (1<<1),
 	IB_SEND_SOLICITED	= (1<<2),
 	IB_SEND_INLINE		= (1<<3),
-	IB_SEND_IP_CSUM		= (1<<4)
+	IB_SEND_IP_CSUM		= (1<<4),
+
+	/* reserve bits 26-31 for low level drivers' internal use */
+	IB_SEND_RESERVED_START	= (1 << 26),
+	IB_SEND_RESERVED_END	= (1 << 31),
 };
 
 struct ib_sge {