drivers/infiniband/hw/mlx4/qp.c

/*
 * Copyright (c) 2007 Cisco Systems, 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/log2.h>

#include <rdma/ib_cache.h>
#include <rdma/ib_pack.h>

#include <linux/mlx4/qp.h>

#include "mlx4_ib.h"
#include "user.h"

enum {
	MLX4_IB_ACK_REQ_FREQ	= 8,
};

enum {
	MLX4_IB_DEFAULT_SCHED_QUEUE	= 0x83,
	MLX4_IB_DEFAULT_QP0_SCHED_QUEUE	= 0x3f
};

enum {
	/*
	 * Largest possible UD header: send with GRH and immediate data.
	 */
	MLX4_IB_UD_HEADER_SIZE		= 72
};

struct mlx4_ib_sqp {
	struct mlx4_ib_qp	qp;
	int			pkey_index;
	u32			qkey;
	u32			send_psn;
	struct ib_ud_header	ud_header;
	u8			header_buf[MLX4_IB_UD_HEADER_SIZE];
};

enum {
	MLX4_IB_MIN_SQ_STRIDE = 6
};

static const __be32 mlx4_ib_opcode[] = {
	[IB_WR_SEND]			= __constant_cpu_to_be32(MLX4_OPCODE_SEND),
	[IB_WR_SEND_WITH_IMM]		= __constant_cpu_to_be32(MLX4_OPCODE_SEND_IMM),
	[IB_WR_RDMA_WRITE]		= __constant_cpu_to_be32(MLX4_OPCODE_RDMA_WRITE),
	[IB_WR_RDMA_WRITE_WITH_IMM]	= __constant_cpu_to_be32(MLX4_OPCODE_RDMA_WRITE_IMM),
	[IB_WR_RDMA_READ]		= __constant_cpu_to_be32(MLX4_OPCODE_RDMA_READ),
	[IB_WR_ATOMIC_CMP_AND_SWP]	= __constant_cpu_to_be32(MLX4_OPCODE_ATOMIC_CS),
	[IB_WR_ATOMIC_FETCH_AND_ADD]	= __constant_cpu_to_be32(MLX4_OPCODE_ATOMIC_FA),
};

static struct mlx4_ib_sqp *to_msqp(struct mlx4_ib_qp *mqp)
{
	return container_of(mqp, struct mlx4_ib_sqp, qp);
}

static int is_sqp(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp)
{
	return qp->mqp.qpn >= dev->dev->caps.sqp_start &&
		qp->mqp.qpn <= dev->dev->caps.sqp_start + 3;
}

static int is_qp0(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp)
{
	return qp->mqp.qpn >= dev->dev->caps.sqp_start &&
		qp->mqp.qpn <= dev->dev->caps.sqp_start + 1;
}

static void *get_wqe(struct mlx4_ib_qp *qp, int offset)
{
	return mlx4_buf_offset(&qp->buf, offset);
}

static void *get_recv_wqe(struct mlx4_ib_qp *qp, int n)
{
	return get_wqe(qp, qp->rq.offset + (n << qp->rq.wqe_shift));
}

static void *get_send_wqe(struct mlx4_ib_qp *qp, int n)
{
	return get_wqe(qp, qp->sq.offset + (n << qp->sq.wqe_shift));
}

/*
 * Stamp a SQ WQE so that it is invalid if prefetched by marking the
 * first four bytes of every 64 byte chunk with
 *     0x7FFFFFF | (invalid_ownership_value << 31).
 *
 * When the max work request size is less than or equal to the WQE
 * basic block size, as an optimization, we can stamp all WQEs with
 * 0xffffffff, and skip the very first chunk of each WQE.
 */
static void stamp_send_wqe(struct mlx4_ib_qp *qp, int n, int size)
{
	u32 *wqe;
	int i;
	int s;
	int ind;
	void *buf;
	__be32 stamp;

	s = roundup(size, 1U << qp->sq.wqe_shift);
	if (qp->sq_max_wqes_per_wr > 1) {
		for (i = 0; i < s; i += 64) {
			ind = (i >> qp->sq.wqe_shift) + n;
			stamp = ind & qp->sq.wqe_cnt ? cpu_to_be32(0x7fffffff) :
						       cpu_to_be32(0xffffffff);
			buf = get_send_wqe(qp, ind & (qp->sq.wqe_cnt - 1));
			wqe = buf + (i & ((1 << qp->sq.wqe_shift) - 1));
			*wqe = stamp;
		}
	} else {
		buf = get_send_wqe(qp, n & (qp->sq.wqe_cnt - 1));
		for (i = 64; i < s; i += 64) {
			wqe = buf + i;
			*wqe = 0xffffffff;
		}
	}
}

static void post_nop_wqe(struct mlx4_ib_qp *qp, int n, int size)
{
	struct mlx4_wqe_ctrl_seg *ctrl;
	struct mlx4_wqe_inline_seg *inl;
	void *wqe;
	int s;

	ctrl = wqe = get_send_wqe(qp, n & (qp->sq.wqe_cnt - 1));
	s = sizeof(struct mlx4_wqe_ctrl_seg);

	if (qp->ibqp.qp_type == IB_QPT_UD) {
		struct mlx4_wqe_datagram_seg *dgram = wqe + sizeof *ctrl;
		struct mlx4_av *av = (struct mlx4_av *)dgram->av;
		memset(dgram, 0, sizeof *dgram);
		av->port_pd = cpu_to_be32((qp->port << 24) | to_mpd(qp->ibqp.pd)->pdn);
		s += sizeof(struct mlx4_wqe_datagram_seg);
	}

	/* Pad the remainder of the WQE with an inline data segment. */
	if (size > s) {
		inl = wqe + s;
		inl->byte_count = cpu_to_be32(1 << 31 | (size - s - sizeof *inl));
	}
	ctrl->srcrb_flags = 0;
	ctrl->fence_size = size / 16;
	/*
	 * Make sure descriptor is fully written before setting ownership bit
	 * (because HW can start executing as soon as we do).
	 */
	wmb();

	ctrl->owner_opcode = cpu_to_be32(MLX4_OPCODE_NOP | MLX4_WQE_CTRL_NEC) |
		(n & qp->sq.wqe_cnt ? cpu_to_be32(1 << 31) : 0);

	stamp_send_wqe(qp, n + qp->sq_spare_wqes, size);
}

/* Post NOP WQE to prevent wrap-around in the middle of WR */
static inline unsigned pad_wraparound(struct mlx4_ib_qp *qp, int ind)
{
	unsigned s = qp->sq.wqe_cnt - (ind & (qp->sq.wqe_cnt - 1));
	if (unlikely(s < qp->sq_max_wqes_per_wr)) {
		post_nop_wqe(qp, ind, s << qp->sq.wqe_shift);
		ind += s;
	}
	return ind;
}

static void mlx4_ib_qp_event(struct mlx4_qp *qp, enum mlx4_event type)
{
	struct ib_event event;
	struct ib_qp *ibqp = &to_mibqp(qp)->ibqp;

	if (type == MLX4_EVENT_TYPE_PATH_MIG)
		to_mibqp(qp)->port = to_mibqp(qp)->alt_port;

	if (ibqp->event_handler) {
		event.device     = ibqp->device;
		event.element.qp = ibqp;
		switch (type) {
		case MLX4_EVENT_TYPE_PATH_MIG:
			event.event = IB_EVENT_PATH_MIG;
			break;
		case MLX4_EVENT_TYPE_COMM_EST:
			event.event = IB_EVENT_COMM_EST;
			break;
		case MLX4_EVENT_TYPE_SQ_DRAINED:
			event.event = IB_EVENT_SQ_DRAINED;
			break;
		case MLX4_EVENT_TYPE_SRQ_QP_LAST_WQE:
			event.event = IB_EVENT_QP_LAST_WQE_REACHED;
			break;
		case MLX4_EVENT_TYPE_WQ_CATAS_ERROR:
			event.event = IB_EVENT_QP_FATAL;
			break;
		case MLX4_EVENT_TYPE_PATH_MIG_FAILED:
			event.event = IB_EVENT_PATH_MIG_ERR;
			break;
		case MLX4_EVENT_TYPE_WQ_INVAL_REQ_ERROR:
			event.event = IB_EVENT_QP_REQ_ERR;
			break;
		case MLX4_EVENT_TYPE_WQ_ACCESS_ERROR:
			event.event = IB_EVENT_QP_ACCESS_ERR;
			break;
		default:
			printk(KERN_WARNING "mlx4_ib: Unexpected event type %d "
			       "on QP %06x\n", type, qp->qpn);
			return;
		}

		ibqp->event_handler(&event, ibqp->qp_context);
	}
}

static int send_wqe_overhead(enum ib_qp_type type)
{
	/*
	 * UD WQEs must have a datagram segment.
	 * RC and UC WQEs might have a remote address segment.
	 * MLX WQEs need two extra inline data segments (for the UD
	 * header and space for the ICRC).
	 */
	switch (type) {
	case IB_QPT_UD:
		return sizeof (struct mlx4_wqe_ctrl_seg) +
			sizeof (struct mlx4_wqe_datagram_seg);
	case IB_QPT_UC:
		return sizeof (struct mlx4_wqe_ctrl_seg) +
			sizeof (struct mlx4_wqe_raddr_seg);
	case IB_QPT_RC:
		return sizeof (struct mlx4_wqe_ctrl_seg) +
			sizeof (struct mlx4_wqe_atomic_seg) +
			sizeof (struct mlx4_wqe_raddr_seg);
	case IB_QPT_SMI:
	case IB_QPT_GSI:
		return sizeof (struct mlx4_wqe_ctrl_seg) +
			ALIGN(MLX4_IB_UD_HEADER_SIZE +
			      DIV_ROUND_UP(MLX4_IB_UD_HEADER_SIZE,
					   MLX4_INLINE_ALIGN) *
			      sizeof (struct mlx4_wqe_inline_seg),
			      sizeof (struct mlx4_wqe_data_seg)) +
			ALIGN(4 +
			      sizeof (struct mlx4_wqe_inline_seg),
			      sizeof (struct mlx4_wqe_data_seg));
	default:
		return sizeof (struct mlx4_wqe_ctrl_seg);
	}
}

static int set_rq_size(struct mlx4_ib_dev *dev, struct ib_qp_cap *cap,
		       int is_user, int has_srq, struct mlx4_ib_qp *qp)
{
	/* Sanity check RQ size before proceeding */
	if (cap->max_recv_wr  > dev->dev->caps.max_wqes  ||
	    cap->max_recv_sge > dev->dev->caps.max_rq_sg)
		return -EINVAL;

	if (has_srq) {
		/* QPs attached to an SRQ should have no RQ */
		if (cap->max_recv_wr)
			return -EINVAL;

		qp->rq.wqe_cnt = qp->rq.max_gs = 0;
	} else {
		/* HW requires >= 1 RQ entry with >= 1 gather entry */
		if (is_user && (!cap->max_recv_wr || !cap->max_recv_sge))
			return -EINVAL;

		qp->rq.wqe_cnt	 = roundup_pow_of_two(max(1U, cap->max_recv_wr));
		qp->rq.max_gs	 = roundup_pow_of_two(max(1U, cap->max_recv_sge));
		qp->rq.wqe_shift = ilog2(qp->rq.max_gs * sizeof (struct mlx4_wqe_data_seg));
	}

	cap->max_recv_wr  = qp->rq.max_post = qp->rq.wqe_cnt;
	cap->max_recv_sge = qp->rq.max_gs;

	return 0;
}

static int set_kernel_sq_size(struct mlx4_ib_dev *dev, struct ib_qp_cap *cap,
			      enum ib_qp_type type, struct mlx4_ib_qp *qp)
{
	int s;

	/* Sanity check SQ size before proceeding */
	if (cap->max_send_wr	 > dev->dev->caps.max_wqes  ||
	    cap->max_send_sge	 > dev->dev->caps.max_sq_sg ||
	    cap->max_inline_data + send_wqe_overhead(type) +
	    sizeof (struct mlx4_wqe_inline_seg) > dev->dev->caps.max_sq_desc_sz)
		return -EINVAL;

	/*
	 * For MLX transport we need 2 extra S/G entries:
	 * one for the header and one for the checksum at the end
	 */
	if ((type == IB_QPT_SMI || type == IB_QPT_GSI) &&
	    cap->max_send_sge + 2 > dev->dev->caps.max_sq_sg)
		return -EINVAL;

	s = max(cap->max_send_sge * sizeof (struct mlx4_wqe_data_seg),
		cap->max_inline_data + sizeof (struct mlx4_wqe_inline_seg)) +
		send_wqe_overhead(type);

	/*
	 * Hermon supports shrinking WQEs, such that a single work
	 * request can include multiple units of 1 << wqe_shift.  This
	 * way, work requests can differ in size, and do not have to
	 * be a power of 2 in size, saving memory and speeding up send
	 * WR posting.  Unfortunately, if we do this then the
	 * wqe_index field in CQEs can't be used to look up the WR ID
	 * anymore, so we do this only if selective signaling is off.
	 *
	 * Further, on 32-bit platforms, we can't use vmap() to make
	 * the QP buffer virtually contigious.  Thus we have to use
	 * constant-sized WRs to make sure a WR is always fully within
	 * a single page-sized chunk.
	 *
	 * Finally, we use NOP work requests to pad the end of the
	 * work queue, to avoid wrap-around in the middle of WR.  We
	 * set NEC bit to avoid getting completions with error for
	 * these NOP WRs, but since NEC is only supported starting
	 * with firmware 2.2.232, we use constant-sized WRs for older
	 * firmware.
	 *
	 * And, since MLX QPs only support SEND, we use constant-sized
	 * WRs in this case.
	 *
	 * We look for the smallest value of wqe_shift such that the
	 * resulting number of wqes does not exceed device
	 * capabilities.
	 *
	 * We set WQE size to at least 64 bytes, this way stamping
	 * invalidates each WQE.
	 */
	if (dev->dev->caps.fw_ver >= MLX4_FW_VER_WQE_CTRL_NEC &&
	    qp->sq_signal_bits && BITS_PER_LONG == 64 &&
	    type != IB_QPT_SMI && type != IB_QPT_GSI)
		qp->sq.wqe_shift = ilog2(64);
	else
		qp->sq.wqe_shift = ilog2(roundup_pow_of_two(s));

	for (;;) {
		if (1 << qp->sq.wqe_shift > dev->dev->caps.max_sq_desc_sz)
			return -EINVAL;

		qp->sq_max_wqes_per_wr = DIV_ROUND_UP(s, 1U << qp->sq.wqe_shift);

		/*
		 * We need to leave 2 KB + 1 WR of headroom in the SQ to
		 * allow HW to prefetch.
		 */
		qp->sq_spare_wqes = (2048 >> qp->sq.wqe_shift) + qp->sq_max_wqes_per_wr;
		qp->sq.wqe_cnt = roundup_pow_of_two(cap->max_send_wr *
						    qp->sq_max_wqes_per_wr +
						    qp->sq_spare_wqes);

		if (qp->sq.wqe_cnt <= dev->dev->caps.max_wqes)
			break;

		if (qp->sq_max_wqes_per_wr <= 1)
			return -EINVAL;

		++qp->sq.wqe_shift;
	}

	qp->sq.max_gs = ((qp->sq_max_wqes_per_wr << qp->sq.wqe_shift) -
			 send_wqe_overhead(type)) / sizeof (struct mlx4_wqe_data_seg);

	qp->buf_size = (qp->rq.wqe_cnt << qp->rq.wqe_shift) +
		(qp->sq.wqe_cnt << qp->sq.wqe_shift);
	if (qp->rq.wqe_shift > qp->sq.wqe_shift) {
		qp->rq.offset = 0;
		qp->sq.offset = qp->rq.wqe_cnt << qp->rq.wqe_shift;
	} else {
		qp->rq.offset = qp->sq.wqe_cnt << qp->sq.wqe_shift;
		qp->sq.offset = 0;
	}

	cap->max_send_wr  = qp->sq.max_post =
		(qp->sq.wqe_cnt - qp->sq_spare_wqes) / qp->sq_max_wqes_per_wr;
	cap->max_send_sge = qp->sq.max_gs;
	/* We don't support inline sends for kernel QPs (yet) */
	cap->max_inline_data = 0;

	return 0;
}

static int set_user_sq_size(struct mlx4_ib_dev *dev,
			    struct mlx4_ib_qp *qp,
			    struct mlx4_ib_create_qp *ucmd)
{
	/* Sanity check SQ size before proceeding */
	if ((1 << ucmd->log_sq_bb_count) > dev->dev->caps.max_wqes	 ||
	    ucmd->log_sq_stride >
		ilog2(roundup_pow_of_two(dev->dev->caps.max_sq_desc_sz)) ||
	    ucmd->log_sq_stride < MLX4_IB_MIN_SQ_STRIDE)
		return -EINVAL;

	qp->sq.wqe_cnt   = 1 << ucmd->log_sq_bb_count;
	qp->sq.wqe_shift = ucmd->log_sq_stride;

	qp->buf_size = (qp->rq.wqe_cnt << qp->rq.wqe_shift) +
		(qp->sq.wqe_cnt << qp->sq.wqe_shift);

	return 0;
}

static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
			    struct ib_qp_init_attr *init_attr,
			    struct ib_udata *udata, int sqpn, struct mlx4_ib_qp *qp)
{
	int err;

	mutex_init(&qp->mutex);
	spin_lock_init(&qp->sq.lock);
	spin_lock_init(&qp->rq.lock);

	qp->state	 = IB_QPS_RESET;
	qp->atomic_rd_en = 0;
	qp->resp_depth   = 0;

	qp->rq.head	    = 0;
	qp->rq.tail	    = 0;
	qp->sq.head	    = 0;
	qp->sq.tail	    = 0;
	qp->sq_next_wqe     = 0;

	if (init_attr->sq_sig_type == IB_SIGNAL_ALL_WR)
		qp->sq_signal_bits = cpu_to_be32(MLX4_WQE_CTRL_CQ_UPDATE);
	else
		qp->sq_signal_bits = 0;

	err = set_rq_size(dev, &init_attr->cap, !!pd->uobject, !!init_attr->srq, qp);
	if (err)
		goto err;

	if (pd->uobject) {
		struct mlx4_ib_create_qp ucmd;

		if (ib_copy_from_udata(&ucmd, udata, sizeof ucmd)) {
			err = -EFAULT;
			goto err;
		}

		qp->sq_no_prefetch = ucmd.sq_no_prefetch;

		err = set_user_sq_size(dev, qp, &ucmd);
		if (err)
			goto err;

		qp->umem = ib_umem_get(pd->uobject->context, ucmd.buf_addr,
				       qp->buf_size, 0);
		if (IS_ERR(qp->umem)) {
			err = PTR_ERR(qp->umem);
			goto err;
		}

		err = mlx4_mtt_init(dev->dev, ib_umem_page_count(qp->umem),
				    ilog2(qp->umem->page_size), &qp->mtt);
		if (err)
			goto err_buf;

		err = mlx4_ib_umem_write_mtt(dev, &qp->mtt, qp->umem);
		if (err)
			goto err_mtt;

		if (!init_attr->srq) {
			err = mlx4_ib_db_map_user(to_mucontext(pd->uobject->context),
						  ucmd.db_addr, &qp->db);
			if (err)
				goto err_mtt;
		}
	} else {
		qp->sq_no_prefetch = 0;

		err = set_kernel_sq_size(dev, &init_attr->cap, init_attr->qp_type, qp);
		if (err)
			goto err;

		if (!init_attr->srq) {
			err = mlx4_ib_db_alloc(dev, &qp->db, 0);
			if (err)
				goto err;

			*qp->db.db = 0;
		}

		if (mlx4_buf_alloc(dev->dev, qp->buf_size, PAGE_SIZE * 2, &qp->buf)) {
			err = -ENOMEM;
			goto err_db;
		}

		err = mlx4_mtt_init(dev->dev, qp->buf.npages, qp->buf.page_shift,
				    &qp->mtt);
		if (err)
			goto err_buf;

		err = mlx4_buf_write_mtt(dev->dev, &qp->mtt, &qp->buf);
		if (err)
			goto err_mtt;

		qp->sq.wrid  = kmalloc(qp->sq.wqe_cnt * sizeof (u64), GFP_KERNEL);
		qp->rq.wrid  = kmalloc(qp->rq.wqe_cnt * sizeof (u64), GFP_KERNEL);

		if (!qp->sq.wrid || !qp->rq.wrid) {
			err = -ENOMEM;
			goto err_wrid;
		}
	}

	err = mlx4_qp_alloc(dev->dev, sqpn, &qp->mqp);
	if (err)
		goto err_wrid;

	/*
	 * Hardware wants QPN written in big-endian order (after
	 * shifting) for send doorbell.  Precompute this value to save
	 * a little bit when posting sends.
	 */
	qp->doorbell_qpn = swab32(qp->mqp.qpn << 8);

	qp->mqp.event = mlx4_ib_qp_event;

	return 0;

err_wrid:
	if (pd->uobject) {
		if (!init_attr->srq)
			mlx4_ib_db_unmap_user(to_mucontext(pd->uobject->context),
					      &qp->db);
	} else {
		kfree(qp->sq.wrid);
		kfree(qp->rq.wrid);
	}

err_mtt:
	mlx4_mtt_cleanup(dev->dev, &qp->mtt);

err_buf:
	if (pd->uobject)
		ib_umem_release(qp->umem);
	else
		mlx4_buf_free(dev->dev, qp->buf_size, &qp->buf);

err_db:
	if (!pd->uobject && !init_attr->srq)
		mlx4_ib_db_free(dev, &qp->db);

err:
	return err;
}

static enum mlx4_qp_state to_mlx4_state(enum ib_qp_state state)
{
	switch (state) {
	case IB_QPS_RESET:	return MLX4_QP_STATE_RST;
	case IB_QPS_INIT:	return MLX4_QP_STATE_INIT;
	case IB_QPS_RTR:	return MLX4_QP_STATE_RTR;
	case IB_QPS_RTS:	return MLX4_QP_STATE_RTS;
	case IB_QPS_SQD:	return MLX4_QP_STATE_SQD;
	case IB_QPS_SQE:	return MLX4_QP_STATE_SQER;
	case IB_QPS_ERR:	return MLX4_QP_STATE_ERR;
	default:		return -1;
	}
}

static void mlx4_ib_lock_cqs(struct mlx4_ib_cq *send_cq, struct mlx4_ib_cq *recv_cq)
{
	if (send_cq == recv_cq)
		spin_lock_irq(&send_cq->lock);
	else if (send_cq->mcq.cqn < recv_cq->mcq.cqn) {
		spin_lock_irq(&send_cq->lock);
		spin_lock_nested(&recv_cq->lock, SINGLE_DEPTH_NESTING);
	} else {
		spin_lock_irq(&recv_cq->lock);
		spin_lock_nested(&send_cq->lock, SINGLE_DEPTH_NESTING);
	}
}

static void mlx4_ib_unlock_cqs(struct mlx4_ib_cq *send_cq, struct mlx4_ib_cq *recv_cq)
{
	if (send_cq == recv_cq)
		spin_unlock_irq(&send_cq->lock);
	else if (send_cq->mcq.cqn < recv_cq->mcq.cqn) {
		spin_unlock(&recv_cq->lock);
		spin_unlock_irq(&send_cq->lock);
	} else {
		spin_unlock(&send_cq->lock);
		spin_unlock_irq(&recv_cq->lock);
	}
}

static void destroy_qp_common(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp,
			      int is_user)
{
	struct mlx4_ib_cq *send_cq, *recv_cq;

	if (qp->state != IB_QPS_RESET)
		if (mlx4_qp_modify(dev->dev, NULL, to_mlx4_state(qp->state),
				   MLX4_QP_STATE_RST, NULL, 0, 0, &qp->mqp))
			printk(KERN_WARNING "mlx4_ib: modify QP %06x to RESET failed.\n",
			       qp->mqp.qpn);

	send_cq = to_mcq(qp->ibqp.send_cq);
	recv_cq = to_mcq(qp->ibqp.recv_cq);

	mlx4_ib_lock_cqs(send_cq, recv_cq);

	if (!is_user) {
		__mlx4_ib_cq_clean(recv_cq, qp->mqp.qpn,
				 qp->ibqp.srq ? to_msrq(qp->ibqp.srq): NULL);
		if (send_cq != recv_cq)
			__mlx4_ib_cq_clean(send_cq, qp->mqp.qpn, NULL);
	}

	mlx4_qp_remove(dev->dev, &qp->mqp);

	mlx4_ib_unlock_cqs(send_cq, recv_cq);

	mlx4_qp_free(dev->dev, &qp->mqp);
	mlx4_mtt_cleanup(dev->dev, &qp->mtt);

	if (is_user) {
		if (!qp->ibqp.srq)
			mlx4_ib_db_unmap_user(to_mucontext(qp->ibqp.uobject->context),
					      &qp->db);
		ib_umem_release(qp->umem);
	} else {
		kfree(qp->sq.wrid);
		kfree(qp->rq.wrid);
		mlx4_buf_free(dev->dev, qp->buf_size, &qp->buf);
		if (!qp->ibqp.srq)
			mlx4_ib_db_free(dev, &qp->db);
	}
}

struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
				struct ib_qp_init_attr *init_attr,
				struct ib_udata *udata)
{
	struct mlx4_ib_dev *dev = to_mdev(pd->device);
	struct mlx4_ib_sqp *sqp;
	struct mlx4_ib_qp *qp;
	int err;

	switch (init_attr->qp_type) {
	case IB_QPT_RC:
	case IB_QPT_UC:
	case IB_QPT_UD:
	{
		qp = kmalloc(sizeof *qp, GFP_KERNEL);
		if (!qp)
			return ERR_PTR(-ENOMEM);

		err = create_qp_common(dev, pd, init_attr, udata, 0, qp);
		if (err) {
			kfree(qp);
			return ERR_PTR(err);
		}

		qp->ibqp.qp_num = qp->mqp.qpn;

		break;
	}
	case IB_QPT_SMI:
	case IB_QPT_GSI:
	{
		/* Userspace is not allowed to create special QPs: */
		if (pd->uobject)
			return ERR_PTR(-EINVAL);

		sqp = kmalloc(sizeof *sqp, GFP_KERNEL);
		if (!sqp)
			return ERR_PTR(-ENOMEM);

		qp = &sqp->qp;

		err = create_qp_common(dev, pd, init_attr, udata,
				       dev->dev->caps.sqp_start +
				       (init_attr->qp_type == IB_QPT_SMI ? 0 : 2) +
				       init_attr->port_num - 1,
				       qp);
		if (err) {
			kfree(sqp);
			return ERR_PTR(err);
		}

		qp->port	= init_attr->port_num;
		qp->ibqp.qp_num = init_attr->qp_type == IB_QPT_SMI ? 0 : 1;

		break;
	}
	default:
		/* Don't support raw QPs */
		return ERR_PTR(-EINVAL);
	}

	return &qp->ibqp;
}

int mlx4_ib_destroy_qp(struct ib_qp *qp)
{
	struct mlx4_ib_dev *dev = to_mdev(qp->device);
	struct mlx4_ib_qp *mqp = to_mqp(qp);

	if (is_qp0(dev, mqp))
		mlx4_CLOSE_PORT(dev->dev, mqp->port);

	destroy_qp_common(dev, mqp, !!qp->pd->uobject);

	if (is_sqp(dev, mqp))
		kfree(to_msqp(mqp));
	else
		kfree(mqp);

	return 0;
}

static int to_mlx4_st(enum ib_qp_type type)
{
	switch (type) {
	case IB_QPT_RC:		return MLX4_QP_ST_RC;
	case IB_QPT_UC:		return MLX4_QP_ST_UC;
	case IB_QPT_UD:		return MLX4_QP_ST_UD;
	case IB_QPT_SMI:
	case IB_QPT_GSI:	return MLX4_QP_ST_MLX;
	default:		return -1;
	}
}

static __be32 to_mlx4_access_flags(struct mlx4_ib_qp *qp, const struct ib_qp_attr *attr,
				   int attr_mask)
{
	u8 dest_rd_atomic;
	u32 access_flags;
	u32 hw_access_flags = 0;

	if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC)
		dest_rd_atomic = attr->max_dest_rd_atomic;
	else
		dest_rd_atomic = qp->resp_depth;

	if (attr_mask & IB_QP_ACCESS_FLAGS)
		access_flags = attr->qp_access_flags;
	else
		access_flags = qp->atomic_rd_en;

	if (!dest_rd_atomic)
		access_flags &= IB_ACCESS_REMOTE_WRITE;

	if (access_flags & IB_ACCESS_REMOTE_READ)
		hw_access_flags |= MLX4_QP_BIT_RRE;
	if (access_flags & IB_ACCESS_REMOTE_ATOMIC)
		hw_access_flags |= MLX4_QP_BIT_RAE;
	if (access_flags & IB_ACCESS_REMOTE_WRITE)
		hw_access_flags |= MLX4_QP_BIT_RWE;

	return cpu_to_be32(hw_access_flags);
}

static void store_sqp_attrs(struct mlx4_ib_sqp *sqp, const struct ib_qp_attr *attr,
			    int attr_mask)
{
	if (attr_mask & IB_QP_PKEY_INDEX)
		sqp->pkey_index = attr->pkey_index;
	if (attr_mask & IB_QP_QKEY)
		sqp->qkey = attr->qkey;
	if (attr_mask & IB_QP_SQ_PSN)
		sqp->send_psn = attr->sq_psn;
}

static void mlx4_set_sched(struct mlx4_qp_path *path, u8 port)
{
	path->sched_queue = (path->sched_queue & 0xbf) | ((port - 1) << 6);
}

static int mlx4_set_path(struct mlx4_ib_dev *dev, const struct ib_ah_attr *ah,
			 struct mlx4_qp_path *path, u8 port)
{
	path->grh_mylmc     = ah->src_path_bits & 0x7f;
	path->rlid	    = cpu_to_be16(ah->dlid);
	if (ah->static_rate) {
		path->static_rate = ah->static_rate + MLX4_STAT_RATE_OFFSET;
		while (path->static_rate > IB_RATE_2_5_GBPS + MLX4_STAT_RATE_OFFSET &&
		       !(1 << path->static_rate & dev->dev->caps.stat_rate_support))
			--path->static_rate;
	} else
		path->static_rate = 0;
	path->counter_index = 0xff;

	if (ah->ah_flags & IB_AH_GRH) {
		if (ah->grh.sgid_index >= dev->dev->caps.gid_table_len[port]) {
			printk(KERN_ERR "sgid_index (%u) too large. max is %d\n",
			       ah->grh.sgid_index, dev->dev->caps.gid_table_len[port] - 1);
			return -1;
		}

		path->grh_mylmc |= 1 << 7;
		path->mgid_index = ah->grh.sgid_index;
		path->hop_limit  = ah->grh.hop_limit;
		path->tclass_flowlabel =
			cpu_to_be32((ah->grh.traffic_class << 20) |
				    (ah->grh.flow_label));
		memcpy(path->rgid, ah->grh.dgid.raw, 16);
	}

	path->sched_queue = MLX4_IB_DEFAULT_SCHED_QUEUE |
		((port - 1) << 6) | ((ah->sl & 0xf) << 2);

	return 0;
}

static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
			       const struct ib_qp_attr *attr, int attr_mask,
			       enum ib_qp_state cur_state, enum ib_qp_state new_state)
{
	struct mlx4_ib_dev *dev = to_mdev(ibqp->device);
	struct mlx4_ib_qp *qp = to_mqp(ibqp);
	struct mlx4_qp_context *context;
	enum mlx4_qp_optpar optpar = 0;
	int sqd_event;
	int err = -EINVAL;

	context = kzalloc(sizeof *context, GFP_KERNEL);
	if (!context)
		return -ENOMEM;

	context->flags = cpu_to_be32((to_mlx4_state(new_state) << 28) |
				     (to_mlx4_st(ibqp->qp_type) << 16));
	context->flags     |= cpu_to_be32(1 << 8); /* DE? */

	if (!(attr_mask & IB_QP_PATH_MIG_STATE))
		context->flags |= cpu_to_be32(MLX4_QP_PM_MIGRATED << 11);
	else {
		optpar |= MLX4_QP_OPTPAR_PM_STATE;
		switch (attr->path_mig_state) {
		case IB_MIG_MIGRATED:
			context->flags |= cpu_to_be32(MLX4_QP_PM_MIGRATED << 11);
			break;
		case IB_MIG_REARM:
			context->flags |= cpu_to_be32(MLX4_QP_PM_REARM << 11);
			break;
		case IB_MIG_ARMED:
			context->flags |= cpu_to_be32(MLX4_QP_PM_ARMED << 11);
			break;
		}
	}

	if (ibqp->qp_type == IB_QPT_GSI || ibqp->qp_type == IB_QPT_SMI ||
	    ibqp->qp_type == IB_QPT_UD)
		context->mtu_msgmax = (IB_MTU_4096 << 5) | 11;
	else if (attr_mask & IB_QP_PATH_MTU) {
		if (attr->path_mtu < IB_MTU_256 || attr->path_mtu > IB_MTU_4096) {
			printk(KERN_ERR "path MTU (%u) is invalid\n",
			       attr->path_mtu);
			goto out;
		}
		context->mtu_msgmax = (attr->path_mtu << 5) | 31;
	}

	if (qp->rq.wqe_cnt)
		context->rq_size_stride = ilog2(qp->rq.wqe_cnt) << 3;
	context->rq_size_stride |= qp->rq.wqe_shift - 4;

	if (qp->sq.wqe_cnt)
		context->sq_size_stride = ilog2(qp->sq.wqe_cnt) << 3;
	context->sq_size_stride |= qp->sq.wqe_shift - 4;

	if (cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT)
		context->sq_size_stride |= !!qp->sq_no_prefetch << 7;

	if (qp->ibqp.uobject)
		context->usr_page = cpu_to_be32(to_mucontext(ibqp->uobject->context)->uar.index);
	else
		context->usr_page = cpu_to_be32(dev->priv_uar.index);

	if (attr_mask & IB_QP_DEST_QPN)
		context->remote_qpn = cpu_to_be32(attr->dest_qp_num);

	if (attr_mask & IB_QP_PORT) {
		if (cur_state == IB_QPS_SQD && new_state == IB_QPS_SQD &&
		    !(attr_mask & IB_QP_AV)) {
			mlx4_set_sched(&context->pri_path, attr->port_num);
			optpar |= MLX4_QP_OPTPAR_SCHED_QUEUE;
		}
	}

	if (attr_mask & IB_QP_PKEY_INDEX) {
		context->pri_path.pkey_index = attr->pkey_index;
		optpar |= MLX4_QP_OPTPAR_PKEY_INDEX;
	}

	if (attr_mask & IB_QP_AV) {
		if (mlx4_set_path(dev, &attr->ah_attr, &context->pri_path,
				  attr_mask & IB_QP_PORT ? attr->port_num : qp->port))
			goto out;

		optpar |= (MLX4_QP_OPTPAR_PRIMARY_ADDR_PATH |
			   MLX4_QP_OPTPAR_SCHED_QUEUE);
	}

	if (attr_mask & IB_QP_TIMEOUT) {
		context->pri_path.ackto = attr->timeout << 3;
		optpar |= MLX4_QP_OPTPAR_ACK_TIMEOUT;
	}

	if (attr_mask & IB_QP_ALT_PATH) {
		if (attr->alt_port_num == 0 ||
		    attr->alt_port_num > dev->dev->caps.num_ports)
			goto out;

		if (attr->alt_pkey_index >=
		    dev->dev->caps.pkey_table_len[attr->alt_port_num])
			goto out;

		if (mlx4_set_path(dev, &attr->alt_ah_attr, &context->alt_path,
				  attr->alt_port_num))
			goto out;

		context->alt_path.pkey_index = attr->alt_pkey_index;
		context->alt_path.ackto = attr->alt_timeout << 3;
		optpar |= MLX4_QP_OPTPAR_ALT_ADDR_PATH;
	}

	context->pd	    = cpu_to_be32(to_mpd(ibqp->pd)->pdn);
	context->params1    = cpu_to_be32(MLX4_IB_ACK_REQ_FREQ << 28);

	if (attr_mask & IB_QP_RNR_RETRY) {
		context->params1 |= cpu_to_be32(attr->rnr_retry << 13);
		optpar |= MLX4_QP_OPTPAR_RNR_RETRY;
	}

	if (attr_mask & IB_QP_RETRY_CNT) {
		context->params1 |= cpu_to_be32(attr->retry_cnt << 16);
		optpar |= MLX4_QP_OPTPAR_RETRY_COUNT;
	}

	if (attr_mask & IB_QP_MAX_QP_RD_ATOMIC) {
		if (attr->max_rd_atomic)
			context->params1 |=
				cpu_to_be32(fls(attr->max_rd_atomic - 1) << 21);
		optpar |= MLX4_QP_OPTPAR_SRA_MAX;
	}

	if (attr_mask & IB_QP_SQ_PSN)
		context->next_send_psn = cpu_to_be32(attr->sq_psn);

	context->cqn_send = cpu_to_be32(to_mcq(ibqp->send_cq)->mcq.cqn);

	if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC) {
		if (attr->max_dest_rd_atomic)
			context->params2 |=
				cpu_to_be32(fls(attr->max_dest_rd_atomic - 1) << 21);
		optpar |= MLX4_QP_OPTPAR_RRA_MAX;
	}

	if (attr_mask & (IB_QP_ACCESS_FLAGS | IB_QP_MAX_DEST_RD_ATOMIC)) {
		context->params2 |= to_mlx4_access_flags(qp, attr, attr_mask);
		optpar |= MLX4_QP_OPTPAR_RWE | MLX4_QP_OPTPAR_RRE | MLX4_QP_OPTPAR_RAE;
	}

	if (ibqp->srq)
		context->params2 |= cpu_to_be32(MLX4_QP_BIT_RIC);

	if (attr_mask & IB_QP_MIN_RNR_TIMER) {
		context->rnr_nextrecvpsn |= cpu_to_be32(attr->min_rnr_timer << 24);
		optpar |= MLX4_QP_OPTPAR_RNR_TIMEOUT;
	}
	if (attr_mask & IB_QP_RQ_PSN)
		context->rnr_nextrecvpsn |= cpu_to_be32(attr->rq_psn);

	context->cqn_recv = cpu_to_be32(to_mcq(ibqp->recv_cq)->mcq.cqn);

	if (attr_mask & IB_QP_QKEY) {
		context->qkey = cpu_to_be32(attr->qkey);
		optpar |= MLX4_QP_OPTPAR_Q_KEY;
	}

	if (ibqp->srq)
		context->srqn = cpu_to_be32(1 << 24 | to_msrq(ibqp->srq)->msrq.srqn);

	if (!ibqp->srq && cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT)
		context->db_rec_addr = cpu_to_be64(qp->db.dma);

	if (cur_state == IB_QPS_INIT &&
	    new_state == IB_QPS_RTR  &&
	    (ibqp->qp_type == IB_QPT_GSI || ibqp->qp_type == IB_QPT_SMI ||
	     ibqp->qp_type == IB_QPT_UD)) {
		context->pri_path.sched_queue = (qp->port - 1) << 6;
		if (is_qp0(dev, qp))
			context->pri_path.sched_queue |= MLX4_IB_DEFAULT_QP0_SCHED_QUEUE;
		else
			context->pri_path.sched_queue |= MLX4_IB_DEFAULT_SCHED_QUEUE;
	}

	if (cur_state == IB_QPS_RTS && new_state == IB_QPS_SQD	&&
	    attr_mask & IB_QP_EN_SQD_ASYNC_NOTIFY && attr->en_sqd_async_notify)
		sqd_event = 1;
	else
		sqd_event = 0;

	/*
	 * Before passing a kernel QP to the HW, make sure that the
	 * ownership bits of the send queue are set and the SQ
	 * headroom is stamped so that the hardware doesn't start
	 * processing stale work requests.
	 */
	if (!ibqp->uobject && cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT) {
		struct mlx4_wqe_ctrl_seg *ctrl;
		int i;

		for (i = 0; i < qp->sq.wqe_cnt; ++i) {
			ctrl = get_send_wqe(qp, i);
			ctrl->owner_opcode = cpu_to_be32(1 << 31);

			stamp_send_wqe(qp, i, 1 << qp->sq.wqe_shift);
		}
	}

	err = mlx4_qp_modify(dev->dev, &qp->mtt, to_mlx4_state(cur_state),
			     to_mlx4_state(new_state), context, optpar,
			     sqd_event, &qp->mqp);
	if (err)
		goto out;

	qp->state = new_state;

	if (attr_mask & IB_QP_ACCESS_FLAGS)
		qp->atomic_rd_en = attr->qp_access_flags;
	if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC)
		qp->resp_depth = attr->max_dest_rd_atomic;
	if (attr_mask & IB_QP_PORT)
		qp->port = attr->port_num;
	if (attr_mask & IB_QP_ALT_PATH)
		qp->alt_port = attr->alt_port_num;

	if (is_sqp(dev, qp))
		store_sqp_attrs(to_msqp(qp), attr, attr_mask);

	/*
	 * If we moved QP0 to RTR, bring the IB link up; if we moved
	 * QP0 to RESET or ERROR, bring the link back down.
	 */
	if (is_qp0(dev, qp)) {
		if (cur_state != IB_QPS_RTR && new_state == IB_QPS_RTR)
			if (mlx4_INIT_PORT(dev->dev, qp->port))
				printk(KERN_WARNING "INIT_PORT failed for port %d\n",
				       qp->port);

		if (cur_state != IB_QPS_RESET && cur_state != IB_QPS_ERR &&
		    (new_state == IB_QPS_RESET || new_state == IB_QPS_ERR))
			mlx4_CLOSE_PORT(dev->dev, qp->port);
	}

	/*
	 * If we moved a kernel QP to RESET, clean up all old CQ
	 * entries and reinitialize the QP.
	 */
	if (new_state == IB_QPS_RESET && !ibqp->uobject) {
		mlx4_ib_cq_clean(to_mcq(ibqp->recv_cq), qp->mqp.qpn,
				 ibqp->srq ? to_msrq(ibqp->srq): NULL);
		if (ibqp->send_cq != ibqp->recv_cq)
			mlx4_ib_cq_clean(to_mcq(ibqp->send_cq), qp->mqp.qpn, NULL);

		qp->rq.head = 0;
		qp->rq.tail = 0;
		qp->sq.head = 0;
		qp->sq.tail = 0;
		qp->sq_next_wqe = 0;
		if (!ibqp->srq)
			*qp->db.db  = 0;
	}

out:
	kfree(context);
	return err;
}

static const struct ib_qp_attr mlx4_ib_qp_attr = { .port_num = 1 };
static const int mlx4_ib_qp_attr_mask_table[IB_QPT_UD + 1] = {
		[IB_QPT_UD]  = (IB_QP_PKEY_INDEX		|
				IB_QP_PORT			|
				IB_QP_QKEY),
		[IB_QPT_UC]  = (IB_QP_PKEY_INDEX		|
				IB_QP_PORT			|
				IB_QP_ACCESS_FLAGS),
		[IB_QPT_RC]  = (IB_QP_PKEY_INDEX		|
				IB_QP_PORT			|
				IB_QP_ACCESS_FLAGS),
		[IB_QPT_SMI] = (IB_QP_PKEY_INDEX		|
				IB_QP_QKEY),
		[IB_QPT_GSI] = (IB_QP_PKEY_INDEX		|
				IB_QP_QKEY),
};

int mlx4_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
		      int attr_mask, struct ib_udata *udata)
{
	struct mlx4_ib_dev *dev = to_mdev(ibqp->device);
	struct mlx4_ib_qp *qp = to_mqp(ibqp);
	enum ib_qp_state cur_state, new_state;
	int err = -EINVAL;

	mutex_lock(&qp->mutex);

	cur_state = attr_mask & IB_QP_CUR_STATE ? attr->cur_qp_state : qp->state;
	new_state = attr_mask & IB_QP_STATE ? attr->qp_state : cur_state;

	if (!ib_modify_qp_is_ok(cur_state, new_state, ibqp->qp_type, attr_mask))
		goto out;

	if ((attr_mask & IB_QP_PORT) &&
	    (attr->port_num == 0 || attr->port_num > dev->dev->caps.num_ports)) {
		goto out;
	}

	if (attr_mask & IB_QP_PKEY_INDEX) {
		int p = attr_mask & IB_QP_PORT ? attr->port_num : qp->port;
		if (attr->pkey_index >= dev->dev->caps.pkey_table_len[p])
			goto out;
	}

	if (attr_mask & IB_QP_MAX_QP_RD_ATOMIC &&
	    attr->max_rd_atomic > dev->dev->caps.max_qp_init_rdma) {
		goto out;
	}

	if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC &&
	    attr->max_dest_rd_atomic > dev->dev->caps.max_qp_dest_rdma) {
		goto out;
	}

	if (cur_state == new_state && cur_state == IB_QPS_RESET) {
		err = 0;
		goto out;
	}

	if (cur_state == IB_QPS_RESET && new_state == IB_QPS_ERR) {
		err = __mlx4_ib_modify_qp(ibqp, &mlx4_ib_qp_attr,
					  mlx4_ib_qp_attr_mask_table[ibqp->qp_type],
					  IB_QPS_RESET, IB_QPS_INIT);
		if (err)
			goto out;
		cur_state = IB_QPS_INIT;
	}

	err = __mlx4_ib_modify_qp(ibqp, attr, attr_mask, cur_state, new_state);

out:
	mutex_unlock(&qp->mutex);
	return err;
}

static int build_mlx_header(struct mlx4_ib_sqp *sqp, struct ib_send_wr *wr,
			    void *wqe)
{
	struct ib_device *ib_dev = &to_mdev(sqp->qp.ibqp.device)->ib_dev;
	struct mlx4_wqe_mlx_seg *mlx = wqe;
	struct mlx4_wqe_inline_seg *inl = wqe + sizeof *mlx;
	struct mlx4_ib_ah *ah = to_mah(wr->wr.ud.ah);
	u16 pkey;
	int send_size;
	int header_size;
	int spc;
	int i;

	send_size = 0;
	for (i = 0; i < wr->num_sge; ++i)
		send_size += wr->sg_list[i].length;

	ib_ud_header_init(send_size, mlx4_ib_ah_grh_present(ah), &sqp->ud_header);

	sqp->ud_header.lrh.service_level   =
		be32_to_cpu(ah->av.sl_tclass_flowlabel) >> 28;
	sqp->ud_header.lrh.destination_lid = ah->av.dlid;
	sqp->ud_header.lrh.source_lid      = cpu_to_be16(ah->av.g_slid & 0x7f);
	if (mlx4_ib_ah_grh_present(ah)) {
		sqp->ud_header.grh.traffic_class =
			(be32_to_cpu(ah->av.sl_tclass_flowlabel) >> 20) & 0xff;
		sqp->ud_header.grh.flow_label    =
			ah->av.sl_tclass_flowlabel & cpu_to_be32(0xfffff);
		sqp->ud_header.grh.hop_limit     = ah->av.hop_limit;
		ib_get_cached_gid(ib_dev, be32_to_cpu(ah->av.port_pd) >> 24,
				  ah->av.gid_index, &sqp->ud_header.grh.source_gid);
		memcpy(sqp->ud_header.grh.destination_gid.raw,
		       ah->av.dgid, 16);
	}

	mlx->flags &= cpu_to_be32(MLX4_WQE_CTRL_CQ_UPDATE);
	mlx->flags |= cpu_to_be32((!sqp->qp.ibqp.qp_num ? MLX4_WQE_MLX_VL15 : 0) |
				  (sqp->ud_header.lrh.destination_lid ==
				   IB_LID_PERMISSIVE ? MLX4_WQE_MLX_SLR : 0) |
				  (sqp->ud_header.lrh.service_level << 8));
	mlx->rlid   = sqp->ud_header.lrh.destination_lid;

	switch (wr->opcode) {
	case IB_WR_SEND:
		sqp->ud_header.bth.opcode	 = IB_OPCODE_UD_SEND_ONLY;
		sqp->ud_header.immediate_present = 0;
		break;
	case IB_WR_SEND_WITH_IMM:
		sqp->ud_header.bth.opcode	 = IB_OPCODE_UD_SEND_ONLY_WITH_IMMEDIATE;
		sqp->ud_header.immediate_present = 1;
		sqp->ud_header.immediate_data    = wr->imm_data;
		break;
	default:
		return -EINVAL;
	}

	sqp->ud_header.lrh.virtual_lane    = !sqp->qp.ibqp.qp_num ? 15 : 0;
	if (sqp->ud_header.lrh.destination_lid == IB_LID_PERMISSIVE)
		sqp->ud_header.lrh.source_lid = IB_LID_PERMISSIVE;
	sqp->ud_header.bth.solicited_event = !!(wr->send_flags & IB_SEND_SOLICITED);
	if (!sqp->qp.ibqp.qp_num)
		ib_get_cached_pkey(ib_dev, sqp->qp.port, sqp->pkey_index, &pkey);
	else
		ib_get_cached_pkey(ib_dev, sqp->qp.port, wr->wr.ud.pkey_index, &pkey);
	sqp->ud_header.bth.pkey = cpu_to_be16(pkey);
	sqp->ud_header.bth.destination_qpn = cpu_to_be32(wr->wr.ud.remote_qpn);
	sqp->ud_header.bth.psn = cpu_to_be32((sqp->send_psn++) & ((1 << 24) - 1));
	sqp->ud_header.deth.qkey = cpu_to_be32(wr->wr.ud.remote_qkey & 0x80000000 ?
					       sqp->qkey : wr->wr.ud.remote_qkey);
	sqp->ud_header.deth.source_qpn = cpu_to_be32(sqp->qp.ibqp.qp_num);

	header_size = ib_ud_header_pack(&sqp->ud_header, sqp->header_buf);

	if (0) {
		printk(KERN_ERR "built UD header of size %d:\n", header_size);
		for (i = 0; i < header_size / 4; ++i) {
			if (i % 8 == 0)
				printk("  [%02x] ", i * 4);
			printk(" %08x",
			       be32_to_cpu(((__be32 *) sqp->header_buf)[i]));
			if ((i + 1) % 8 == 0)
				printk("\n");
		}
		printk("\n");
	}

	/*
	 * Inline data segments may not cross a 64 byte boundary.  If
	 * our UD header is bigger than the space available up to the
	 * next 64 byte boundary in the WQE, use two inline data
	 * segments to hold the UD header.
	 */
	spc = MLX4_INLINE_ALIGN -
		((unsigned long) (inl + 1) & (MLX4_INLINE_ALIGN - 1));
	if (header_size <= spc) {
		inl->byte_count = cpu_to_be32(1 << 31 | header_size);
		memcpy(inl + 1, sqp->header_buf, header_size);
		i = 1;
	} else {
		inl->byte_count = cpu_to_be32(1 << 31 | spc);
		memcpy(inl + 1, sqp->header_buf, spc);

		inl = (void *) (inl + 1) + spc;
		memcpy(inl + 1, sqp->header_buf + spc, header_size - spc);
		/*
		 * Need a barrier here to make sure all the data is
		 * visible before the byte_count field is set.
		 * Otherwise the HCA prefetcher could grab the 64-byte
		 * chunk with this inline segment and get a valid (!=
		 * 0xffffffff) byte count but stale data, and end up
		 * generating a packet with bad headers.
		 *
		 * The first inline segment's byte_count field doesn't
		 * need a barrier, because it comes after a
		 * control/MLX segment and therefore is at an offset
		 * of 16 mod 64.
		 */
		wmb();
		inl->byte_count = cpu_to_be32(1 << 31 | (header_size - spc));
		i = 2;
	}

	return ALIGN(i * sizeof (struct mlx4_wqe_inline_seg) + header_size, 16);
}

static int mlx4_wq_overflow(struct mlx4_ib_wq *wq, int nreq, struct ib_cq *ib_cq)
{
	unsigned cur;
	struct mlx4_ib_cq *cq;

	cur = wq->head - wq->tail;
	if (likely(cur + nreq < wq->max_post))
		return 0;

	cq = to_mcq(ib_cq);
	spin_lock(&cq->lock);
	cur = wq->head - wq->tail;
	spin_unlock(&cq->lock);

	return cur + nreq >= wq->max_post;
}

static __always_inline void set_raddr_seg(struct mlx4_wqe_raddr_seg *rseg,
					  u64 remote_addr, u32 rkey)
{
	rseg->raddr    = cpu_to_be64(remote_addr);
	rseg->rkey     = cpu_to_be32(rkey);
	rseg->reserved = 0;
}

static void set_atomic_seg(struct mlx4_wqe_atomic_seg *aseg, struct ib_send_wr *wr)
{
	if (wr->opcode == IB_WR_ATOMIC_CMP_AND_SWP) {
		aseg->swap_add = cpu_to_be64(wr->wr.atomic.swap);
		aseg->compare  = cpu_to_be64(wr->wr.atomic.compare_add);
	} else {
		aseg->swap_add = cpu_to_be64(wr->wr.atomic.compare_add);
		aseg->compare  = 0;
	}

}

static void set_datagram_seg(struct mlx4_wqe_datagram_seg *dseg,
			     struct ib_send_wr *wr)
{
	memcpy(dseg->av, &to_mah(wr->wr.ud.ah)->av, sizeof (struct mlx4_av));
	dseg->dqpn = cpu_to_be32(wr->wr.ud.remote_qpn);
	dseg->qkey = cpu_to_be32(wr->wr.ud.remote_qkey);
}

static void set_mlx_icrc_seg(void *dseg)
{
	u32 *t = dseg;
	struct mlx4_wqe_inline_seg *iseg = dseg;

	t[1] = 0;

	/*
	 * Need a barrier here before writing the byte_count field to
	 * make sure that all the data is visible before the
	 * byte_count field is set.  Otherwise, if the segment begins
	 * a new cacheline, the HCA prefetcher could grab the 64-byte
	 * chunk and get a valid (!= * 0xffffffff) byte count but
	 * stale data, and end up sending the wrong data.
	 */
	wmb();

	iseg->byte_count = cpu_to_be32((1 << 31) | 4);
}

static void set_data_seg(struct mlx4_wqe_data_seg *dseg, struct ib_sge *sg)
{
	dseg->lkey       = cpu_to_be32(sg->lkey);
	dseg->addr       = cpu_to_be64(sg->addr);

	/*
	 * Need a barrier here before writing the byte_count field to
	 * make sure that all the data is visible before the
	 * byte_count field is set.  Otherwise, if the segment begins
	 * a new cacheline, the HCA prefetcher could grab the 64-byte
	 * chunk and get a valid (!= * 0xffffffff) byte count but
	 * stale data, and end up sending the wrong data.
	 */
	wmb();

	dseg->byte_count = cpu_to_be32(sg->length);
}

static void __set_data_seg(struct mlx4_wqe_data_seg *dseg, struct ib_sge *sg)
{
	dseg->byte_count = cpu_to_be32(sg->length);
	dseg->lkey       = cpu_to_be32(sg->lkey);
	dseg->addr       = cpu_to_be64(sg->addr);
}

int mlx4_ib_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
		      struct ib_send_wr **bad_wr)
{
	struct mlx4_ib_qp *qp = to_mqp(ibqp);
	void *wqe;
	struct mlx4_wqe_ctrl_seg *ctrl;
	struct mlx4_wqe_data_seg *dseg;
	unsigned long flags;
	int nreq;
	int err = 0;
	unsigned ind;
	int uninitialized_var(stamp);
	int uninitialized_var(size);
	int i;

	spin_lock_irqsave(&qp->sq.lock, flags);

	ind = qp->sq_next_wqe;

	for (nreq = 0; wr; ++nreq, wr = wr->next) {
		if (mlx4_wq_overflow(&qp->sq, nreq, qp->ibqp.send_cq)) {
			err = -ENOMEM;
			*bad_wr = wr;
			goto out;
		}

		if (unlikely(wr->num_sge > qp->sq.max_gs)) {
			err = -EINVAL;
			*bad_wr = wr;
			goto out;
		}

		ctrl = wqe = get_send_wqe(qp, ind & (qp->sq.wqe_cnt - 1));
		qp->sq.wrid[(qp->sq.head + nreq) & (qp->sq.wqe_cnt - 1)] = wr->wr_id;

		ctrl->srcrb_flags =
			(wr->send_flags & IB_SEND_SIGNALED ?
			 cpu_to_be32(MLX4_WQE_CTRL_CQ_UPDATE) : 0) |
			(wr->send_flags & IB_SEND_SOLICITED ?
			 cpu_to_be32(MLX4_WQE_CTRL_SOLICITED) : 0) |
			qp->sq_signal_bits;

		if (wr->opcode == IB_WR_SEND_WITH_IMM ||
		    wr->opcode == IB_WR_RDMA_WRITE_WITH_IMM)
			ctrl->imm = wr->imm_data;
		else
			ctrl->imm = 0;

		wqe += sizeof *ctrl;
		size = sizeof *ctrl / 16;

		switch (ibqp->qp_type) {
		case IB_QPT_RC:
		case IB_QPT_UC:
			switch (wr->opcode) {
			case IB_WR_ATOMIC_CMP_AND_SWP:
			case IB_WR_ATOMIC_FETCH_AND_ADD:
				set_raddr_seg(wqe, wr->wr.atomic.remote_addr,
					      wr->wr.atomic.rkey);
				wqe  += sizeof (struct mlx4_wqe_raddr_seg);

				set_atomic_seg(wqe, wr);
				wqe  += sizeof (struct mlx4_wqe_atomic_seg);

				size += (sizeof (struct mlx4_wqe_raddr_seg) +
					 sizeof (struct mlx4_wqe_atomic_seg)) / 16;

				break;

			case IB_WR_RDMA_READ:
			case IB_WR_RDMA_WRITE:
			case IB_WR_RDMA_WRITE_WITH_IMM:
				set_raddr_seg(wqe, wr->wr.rdma.remote_addr,
					      wr->wr.rdma.rkey);
				wqe  += sizeof (struct mlx4_wqe_raddr_seg);
				size += sizeof (struct mlx4_wqe_raddr_seg) / 16;
				break;

			default:
				/* No extra segments required for sends */
				break;
			}
			break;

		case IB_QPT_UD:
			set_datagram_seg(wqe, wr);
			wqe  += sizeof (struct mlx4_wqe_datagram_seg);
			size += sizeof (struct mlx4_wqe_datagram_seg) / 16;
			break;

		case IB_QPT_SMI:
		case IB_QPT_GSI:
			err = build_mlx_header(to_msqp(qp), wr, ctrl);
			if (err < 0) {
				*bad_wr = wr;
				goto out;
			}
			wqe  += err;
			size += err / 16;

			err = 0;
			break;

		default:
			break;
		}

		/*
		 * Write data segments in reverse order, so as to
		 * overwrite cacheline stamp last within each
		 * cacheline.  This avoids issues with WQE
		 * prefetching.
		 */

		dseg = wqe;
		dseg += wr->num_sge - 1;
		size += wr->num_sge * (sizeof (struct mlx4_wqe_data_seg) / 16);

		/* Add one more inline data segment for ICRC for MLX sends */
		if (unlikely(qp->ibqp.qp_type == IB_QPT_SMI ||
			     qp->ibqp.qp_type == IB_QPT_GSI)) {
			set_mlx_icrc_seg(dseg + 1);
			size += sizeof (struct mlx4_wqe_data_seg) / 16;
		}

		for (i = wr->num_sge - 1; i >= 0; --i, --dseg)
			set_data_seg(dseg, wr->sg_list + i);

		ctrl->fence_size = (wr->send_flags & IB_SEND_FENCE ?
				    MLX4_WQE_CTRL_FENCE : 0) | size;

		/*
		 * Make sure descriptor is fully written before
		 * setting ownership bit (because HW can start
		 * executing as soon as we do).
		 */
		wmb();

		if (wr->opcode < 0 || wr->opcode >= ARRAY_SIZE(mlx4_ib_opcode)) {
			err = -EINVAL;
			goto out;
		}

		ctrl->owner_opcode = mlx4_ib_opcode[wr->opcode] |
			(ind & qp->sq.wqe_cnt ? cpu_to_be32(1 << 31) : 0);

		stamp = ind + qp->sq_spare_wqes;
		ind += DIV_ROUND_UP(size * 16, 1U << qp->sq.wqe_shift);

		/*
		 * We can improve latency by not stamping the last
		 * send queue WQE until after ringing the doorbell, so
		 * only stamp here if there are still more WQEs to post.
		 *
		 * Same optimization applies to padding with NOP wqe
		 * in case of WQE shrinking (used to prevent wrap-around
		 * in the middle of WR).
		 */
		if (wr->next) {
			stamp_send_wqe(qp, stamp, size * 16);
			ind = pad_wraparound(qp, ind);
		}

	}

out:
	if (likely(nreq)) {
		qp->sq.head += nreq;

		/*
		 * Make sure that descriptors are written before
		 * doorbell record.
		 */
		wmb();

		writel(qp->doorbell_qpn,
		       to_mdev(ibqp->device)->uar_map + MLX4_SEND_DOORBELL);

		/*
		 * Make sure doorbells don't leak out of SQ spinlock
		 * and reach the HCA out of order.
		 */
		mmiowb();

		stamp_send_wqe(qp, stamp, size * 16);

		ind = pad_wraparound(qp, ind);
		qp->sq_next_wqe = ind;
	}

	spin_unlock_irqrestore(&qp->sq.lock, flags);

	return err;
}

int mlx4_ib_post_recv(struct ib_qp *ibqp, struct ib_recv_wr *wr,
		      struct ib_recv_wr **bad_wr)
{
	struct mlx4_ib_qp *qp = to_mqp(ibqp);
	struct mlx4_wqe_data_seg *scat;
	unsigned long flags;
	int err = 0;
	int nreq;
	int ind;
	int i;

	spin_lock_irqsave(&qp->rq.lock, flags);

	ind = qp->rq.head & (qp->rq.wqe_cnt - 1);

	for (nreq = 0; wr; ++nreq, wr = wr->next) {
		if (mlx4_wq_overflow(&qp->rq, nreq, qp->ibqp.send_cq)) {
			err = -ENOMEM;
			*bad_wr = wr;
			goto out;
		}

		if (unlikely(wr->num_sge > qp->rq.max_gs)) {
			err = -EINVAL;
			*bad_wr = wr;
			goto out;
		}

		scat = get_recv_wqe(qp, ind);

		for (i = 0; i < wr->num_sge; ++i)
			__set_data_seg(scat + i, wr->sg_list + i);

		if (i < qp->rq.max_gs) {
			scat[i].byte_count = 0;
			scat[i].lkey       = cpu_to_be32(MLX4_INVALID_LKEY);
			scat[i].addr       = 0;
		}

		qp->rq.wrid[ind] = wr->wr_id;

		ind = (ind + 1) & (qp->rq.wqe_cnt - 1);
	}

out:
	if (likely(nreq)) {
		qp->rq.head += nreq;

		/*
		 * Make sure that descriptors are written before
		 * doorbell record.
		 */
		wmb();

		*qp->db.db = cpu_to_be32(qp->rq.head & 0xffff);
	}

	spin_unlock_irqrestore(&qp->rq.lock, flags);

	return err;
}

static inline enum ib_qp_state to_ib_qp_state(enum mlx4_qp_state mlx4_state)
{
	switch (mlx4_state) {
	case MLX4_QP_STATE_RST:      return IB_QPS_RESET;
	case MLX4_QP_STATE_INIT:     return IB_QPS_INIT;
	case MLX4_QP_STATE_RTR:      return IB_QPS_RTR;
	case MLX4_QP_STATE_RTS:      return IB_QPS_RTS;
	case MLX4_QP_STATE_SQ_DRAINING:
	case MLX4_QP_STATE_SQD:      return IB_QPS_SQD;
	case MLX4_QP_STATE_SQER:     return IB_QPS_SQE;
	case MLX4_QP_STATE_ERR:      return IB_QPS_ERR;
	default:		     return -1;
	}
}

static inline enum ib_mig_state to_ib_mig_state(int mlx4_mig_state)
{
	switch (mlx4_mig_state) {
	case MLX4_QP_PM_ARMED:		return IB_MIG_ARMED;
	case MLX4_QP_PM_REARM:		return IB_MIG_REARM;
	case MLX4_QP_PM_MIGRATED:	return IB_MIG_MIGRATED;
	default: return -1;
	}
}

static int to_ib_qp_access_flags(int mlx4_flags)
{
	int ib_flags = 0;

	if (mlx4_flags & MLX4_QP_BIT_RRE)
		ib_flags |= IB_ACCESS_REMOTE_READ;
	if (mlx4_flags & MLX4_QP_BIT_RWE)
		ib_flags |= IB_ACCESS_REMOTE_WRITE;
	if (mlx4_flags & MLX4_QP_BIT_RAE)
		ib_flags |= IB_ACCESS_REMOTE_ATOMIC;

	return ib_flags;
}

static void to_ib_ah_attr(struct mlx4_dev *dev, struct ib_ah_attr *ib_ah_attr,
				struct mlx4_qp_path *path)
{
	memset(ib_ah_attr, 0, sizeof *ib_ah_attr);
	ib_ah_attr->port_num	  = path->sched_queue & 0x40 ? 2 : 1;

	if (ib_ah_attr->port_num == 0 || ib_ah_attr->port_num > dev->caps.num_ports)
		return;

	ib_ah_attr->dlid	  = be16_to_cpu(path->rlid);
	ib_ah_attr->sl		  = (path->sched_queue >> 2) & 0xf;
	ib_ah_attr->src_path_bits = path->grh_mylmc & 0x7f;
	ib_ah_attr->static_rate   = path->static_rate ? path->static_rate - 5 : 0;
	ib_ah_attr->ah_flags      = (path->grh_mylmc & (1 << 7)) ? IB_AH_GRH : 0;
	if (ib_ah_attr->ah_flags) {
		ib_ah_attr->grh.sgid_index = path->mgid_index;
		ib_ah_attr->grh.hop_limit  = path->hop_limit;
		ib_ah_attr->grh.traffic_class =
			(be32_to_cpu(path->tclass_flowlabel) >> 20) & 0xff;
		ib_ah_attr->grh.flow_label =
			be32_to_cpu(path->tclass_flowlabel) & 0xfffff;
		memcpy(ib_ah_attr->grh.dgid.raw,
			path->rgid, sizeof ib_ah_attr->grh.dgid.raw);
	}
}

int mlx4_ib_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *qp_attr, int qp_attr_mask,
		     struct ib_qp_init_attr *qp_init_attr)
{
	struct mlx4_ib_dev *dev = to_mdev(ibqp->device);
	struct mlx4_ib_qp *qp = to_mqp(ibqp);
	struct mlx4_qp_context context;
	int mlx4_state;
	int err;

	if (qp->state == IB_QPS_RESET) {
		qp_attr->qp_state = IB_QPS_RESET;
		goto done;
	}

	err = mlx4_qp_query(dev->dev, &qp->mqp, &context);
	if (err)
		return -EINVAL;

	mlx4_state = be32_to_cpu(context.flags) >> 28;

	qp_attr->qp_state	     = to_ib_qp_state(mlx4_state);
	qp_attr->path_mtu	     = context.mtu_msgmax >> 5;
	qp_attr->path_mig_state	     =
		to_ib_mig_state((be32_to_cpu(context.flags) >> 11) & 0x3);
	qp_attr->qkey		     = be32_to_cpu(context.qkey);
	qp_attr->rq_psn		     = be32_to_cpu(context.rnr_nextrecvpsn) & 0xffffff;
	qp_attr->sq_psn		     = be32_to_cpu(context.next_send_psn) & 0xffffff;
	qp_attr->dest_qp_num	     = be32_to_cpu(context.remote_qpn) & 0xffffff;
	qp_attr->qp_access_flags     =
		to_ib_qp_access_flags(be32_to_cpu(context.params2));

	if (qp->ibqp.qp_type == IB_QPT_RC || qp->ibqp.qp_type == IB_QPT_UC) {
		to_ib_ah_attr(dev->dev, &qp_attr->ah_attr, &context.pri_path);
		to_ib_ah_attr(dev->dev, &qp_attr->alt_ah_attr, &context.alt_path);
		qp_attr->alt_pkey_index = context.alt_path.pkey_index & 0x7f;
		qp_attr->alt_port_num	= qp_attr->alt_ah_attr.port_num;
	}

	qp_attr->pkey_index = context.pri_path.pkey_index & 0x7f;
	if (qp_attr->qp_state == IB_QPS_INIT)
		qp_attr->port_num = qp->port;
	else
		qp_attr->port_num = context.pri_path.sched_queue & 0x40 ? 2 : 1;

	/* qp_attr->en_sqd_async_notify is only applicable in modify qp */
	qp_attr->sq_draining = mlx4_state == MLX4_QP_STATE_SQ_DRAINING;

	qp_attr->max_rd_atomic = 1 << ((be32_to_cpu(context.params1) >> 21) & 0x7);

	qp_attr->max_dest_rd_atomic =
		1 << ((be32_to_cpu(context.params2) >> 21) & 0x7);
	qp_attr->min_rnr_timer	    =
		(be32_to_cpu(context.rnr_nextrecvpsn) >> 24) & 0x1f;
	qp_attr->timeout	    = context.pri_path.ackto >> 3;
	qp_attr->retry_cnt	    = (be32_to_cpu(context.params1) >> 16) & 0x7;
	qp_attr->rnr_retry	    = (be32_to_cpu(context.params1) >> 13) & 0x7;
	qp_attr->alt_timeout	    = context.alt_path.ackto >> 3;

done:
	qp_attr->cur_qp_state	     = qp_attr->qp_state;
	qp_attr->cap.max_recv_wr     = qp->rq.wqe_cnt;
	qp_attr->cap.max_recv_sge    = qp->rq.max_gs;

	if (!ibqp->uobject) {
		qp_attr->cap.max_send_wr  = qp->sq.wqe_cnt;
		qp_attr->cap.max_send_sge = qp->sq.max_gs;
	} else {
		qp_attr->cap.max_send_wr  = 0;
		qp_attr->cap.max_send_sge = 0;
	}

	/*
	 * We don't support inline sends for kernel QPs (yet), and we
	 * don't know what userspace's value should be.
	 */
	qp_attr->cap.max_inline_data = 0;

	qp_init_attr->cap	     = qp_attr->cap;

	return 0;
}