*/
#include <linux/log2.h>
+#include <linux/slab.h>
#include <rdma/ib_cache.h>
#include <rdma/ib_pack.h>
/*
* Largest possible UD header: send with GRH and immediate data.
*/
- MLX4_IB_UD_HEADER_SIZE = 72
+ MLX4_IB_UD_HEADER_SIZE = 72,
+ MLX4_IB_LSO_HEADER_SPARE = 128,
};
struct mlx4_ib_sqp {
};
enum {
- MLX4_IB_MIN_SQ_STRIDE = 6
+ MLX4_IB_MIN_SQ_STRIDE = 6,
+ MLX4_IB_CACHE_LINE_SIZE = 64,
};
static const __be32 mlx4_ib_opcode[] = {
- [IB_WR_SEND] = __constant_cpu_to_be32(MLX4_OPCODE_SEND),
- [IB_WR_LSO] = __constant_cpu_to_be32(MLX4_OPCODE_LSO),
- [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),
- [IB_WR_SEND_WITH_INV] = __constant_cpu_to_be32(MLX4_OPCODE_SEND_INVAL),
- [IB_WR_LOCAL_INV] = __constant_cpu_to_be32(MLX4_OPCODE_LOCAL_INVAL),
- [IB_WR_FAST_REG_MR] = __constant_cpu_to_be32(MLX4_OPCODE_FMR),
+ [IB_WR_SEND] = cpu_to_be32(MLX4_OPCODE_SEND),
+ [IB_WR_LSO] = cpu_to_be32(MLX4_OPCODE_LSO),
+ [IB_WR_SEND_WITH_IMM] = cpu_to_be32(MLX4_OPCODE_SEND_IMM),
+ [IB_WR_RDMA_WRITE] = cpu_to_be32(MLX4_OPCODE_RDMA_WRITE),
+ [IB_WR_RDMA_WRITE_WITH_IMM] = cpu_to_be32(MLX4_OPCODE_RDMA_WRITE_IMM),
+ [IB_WR_RDMA_READ] = cpu_to_be32(MLX4_OPCODE_RDMA_READ),
+ [IB_WR_ATOMIC_CMP_AND_SWP] = cpu_to_be32(MLX4_OPCODE_ATOMIC_CS),
+ [IB_WR_ATOMIC_FETCH_AND_ADD] = cpu_to_be32(MLX4_OPCODE_ATOMIC_FA),
+ [IB_WR_SEND_WITH_INV] = cpu_to_be32(MLX4_OPCODE_SEND_INVAL),
+ [IB_WR_LOCAL_INV] = cpu_to_be32(MLX4_OPCODE_LOCAL_INVAL),
+ [IB_WR_FAST_REG_MR] = cpu_to_be32(MLX4_OPCODE_FMR),
};
static struct mlx4_ib_sqp *to_msqp(struct mlx4_ib_qp *mqp)
case IB_QPT_UD:
return sizeof (struct mlx4_wqe_ctrl_seg) +
sizeof (struct mlx4_wqe_datagram_seg) +
- ((flags & MLX4_IB_QP_LSO) ? 64 : 0);
+ ((flags & MLX4_IB_QP_LSO) ? MLX4_IB_LSO_HEADER_SPARE : 0);
case IB_QPT_UC:
return sizeof (struct mlx4_wqe_ctrl_seg) +
sizeof (struct mlx4_wqe_raddr_seg);
* 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
+ * the QP buffer virtually contiguous. Thus we have to use
* constant-sized WRs to make sure a WR is always fully within
* a single page-sized chunk.
*
}
static void mlx4_ib_lock_cqs(struct mlx4_ib_cq *send_cq, struct mlx4_ib_cq *recv_cq)
+ __acquires(&send_cq->lock) __acquires(&recv_cq->lock)
{
- if (send_cq == recv_cq)
+ if (send_cq == recv_cq) {
spin_lock_irq(&send_cq->lock);
- else if (send_cq->mcq.cqn < recv_cq->mcq.cqn) {
+ __acquire(&recv_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 {
}
static void mlx4_ib_unlock_cqs(struct mlx4_ib_cq *send_cq, struct mlx4_ib_cq *recv_cq)
+ __releases(&send_cq->lock) __releases(&recv_cq->lock)
{
- if (send_cq == recv_cq)
+ if (send_cq == recv_cq) {
+ __release(&recv_cq->lock);
spin_unlock_irq(&send_cq->lock);
- else if (send_cq->mcq.cqn < recv_cq->mcq.cqn) {
+ } else if (send_cq->mcq.cqn < recv_cq->mcq.cqn) {
spin_unlock(&recv_cq->lock);
spin_unlock_irq(&send_cq->lock);
} else {
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);
static int build_mlx_header(struct mlx4_ib_sqp *sqp, struct ib_send_wr *wr,
void *wqe, unsigned *mlx_seg_len)
{
- struct ib_device *ib_dev = &to_mdev(sqp->qp.ibqp.device)->ib_dev;
+ struct ib_device *ib_dev = sqp->qp.ibqp.device;
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);
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);
+ ib_ud_header_init(send_size, mlx4_ib_ah_grh_present(ah), 0, &sqp->ud_header);
sqp->ud_header.lrh.service_level =
be32_to_cpu(ah->av.sl_tclass_flowlabel) >> 28;
int i;
for (i = 0; i < wr->wr.fast_reg.page_list_len; ++i)
- wr->wr.fast_reg.page_list->page_list[i] =
+ mfrpl->mapped_page_list[i] =
cpu_to_be64(wr->wr.fast_reg.page_list->page_list[i] |
MLX4_MTT_FLAG_PRESENT);
static int build_lso_seg(struct mlx4_wqe_lso_seg *wqe, struct ib_send_wr *wr,
struct mlx4_ib_qp *qp, unsigned *lso_seg_len,
- __be32 *lso_hdr_sz)
+ __be32 *lso_hdr_sz, __be32 *blh)
{
unsigned halign = ALIGN(sizeof *wqe + wr->wr.ud.hlen, 16);
- /*
- * This is a temporary limitation and will be removed in
- * a forthcoming FW release:
- */
- if (unlikely(halign > 64))
- return -EINVAL;
+ if (unlikely(halign > MLX4_IB_CACHE_LINE_SIZE))
+ *blh = cpu_to_be32(1 << 6);
if (unlikely(!(qp->flags & MLX4_IB_QP_LSO) &&
wr->num_sge > qp->sq.max_gs - (halign >> 4)))
__be32 dummy;
__be32 *lso_wqe;
__be32 uninitialized_var(lso_hdr_sz);
+ __be32 blh;
int i;
spin_lock_irqsave(&qp->sq.lock, flags);
for (nreq = 0; wr; ++nreq, wr = wr->next) {
lso_wqe = &dummy;
+ blh = 0;
if (mlx4_wq_overflow(&qp->sq, nreq, qp->ibqp.send_cq)) {
err = -ENOMEM;
break;
case IB_WR_LOCAL_INV:
+ ctrl->srcrb_flags |=
+ cpu_to_be32(MLX4_WQE_CTRL_STRONG_ORDER);
set_local_inv_seg(wqe, wr->ex.invalidate_rkey);
wqe += sizeof (struct mlx4_wqe_local_inval_seg);
size += sizeof (struct mlx4_wqe_local_inval_seg) / 16;
break;
case IB_WR_FAST_REG_MR:
+ ctrl->srcrb_flags |=
+ cpu_to_be32(MLX4_WQE_CTRL_STRONG_ORDER);
set_fmr_seg(wqe, wr);
wqe += sizeof (struct mlx4_wqe_fmr_seg);
size += sizeof (struct mlx4_wqe_fmr_seg) / 16;
size += sizeof (struct mlx4_wqe_datagram_seg) / 16;
if (wr->opcode == IB_WR_LSO) {
- err = build_lso_seg(wqe, wr, qp, &seglen, &lso_hdr_sz);
+ err = build_lso_seg(wqe, wr, qp, &seglen, &lso_hdr_sz, &blh);
if (unlikely(err)) {
*bad_wr = wr;
goto out;
}
ctrl->owner_opcode = mlx4_ib_opcode[wr->opcode] |
- (ind & qp->sq.wqe_cnt ? cpu_to_be32(1 << 31) : 0);
+ (ind & qp->sq.wqe_cnt ? cpu_to_be32(1 << 31) : 0) | blh;
stamp = ind + qp->sq_spare_wqes;
ind += DIV_ROUND_UP(size * 16, 1U << qp->sq.wqe_shift);
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)) {
+ if (mlx4_wq_overflow(&qp->rq, nreq, qp->ibqp.recv_cq)) {
err = -ENOMEM;
*bad_wr = wr;
goto out;
git://ftp.safe.ca / safe / jmp / linux-2.6 / blobdiff