X-Git-Url: http://ftp.safe.ca/?a=blobdiff_plain;f=drivers%2Finfiniband%2Fhw%2Fmlx4%2Fcq.c;h=299f20832ab64aaed174b67b3047710e74d95354;hb=00f7ec36c9324928e4cd23f02e6d8550f30c32ca;hp=b2a290c6703a4b59634daf93763b43ab4cbc5a86;hpb=225c7b1feef1b41170f7037a5b10a65cd8a42c54;p=safe%2Fjmp%2Flinux-2.6 diff --git a/drivers/infiniband/hw/mlx4/cq.c b/drivers/infiniband/hw/mlx4/cq.c index b2a290c..299f208 100644 --- a/drivers/infiniband/hw/mlx4/cq.c +++ b/drivers/infiniband/hw/mlx4/cq.c @@ -64,13 +64,7 @@ static void mlx4_ib_cq_event(struct mlx4_cq *cq, enum mlx4_event type) static void *get_cqe_from_buf(struct mlx4_ib_cq_buf *buf, int n) { - int offset = n * sizeof (struct mlx4_cqe); - - if (buf->buf.nbufs == 1) - return buf->buf.u.direct.buf + offset; - else - return buf->buf.u.page_list[offset >> PAGE_SHIFT].buf + - (offset & (PAGE_SIZE - 1)); + return mlx4_buf_offset(&buf->buf, n * sizeof (struct mlx4_cqe)); } static void *get_cqe(struct mlx4_ib_cq *cq, int n) @@ -91,6 +85,82 @@ static struct mlx4_cqe *next_cqe_sw(struct mlx4_ib_cq *cq) return get_sw_cqe(cq, cq->mcq.cons_index); } +int mlx4_ib_modify_cq(struct ib_cq *cq, u16 cq_count, u16 cq_period) +{ + struct mlx4_ib_cq *mcq = to_mcq(cq); + struct mlx4_ib_dev *dev = to_mdev(cq->device); + + return mlx4_cq_modify(dev->dev, &mcq->mcq, cq_count, cq_period); +} + +static int mlx4_ib_alloc_cq_buf(struct mlx4_ib_dev *dev, struct mlx4_ib_cq_buf *buf, int nent) +{ + int err; + + err = mlx4_buf_alloc(dev->dev, nent * sizeof(struct mlx4_cqe), + PAGE_SIZE * 2, &buf->buf); + + if (err) + goto out; + + err = mlx4_mtt_init(dev->dev, buf->buf.npages, buf->buf.page_shift, + &buf->mtt); + if (err) + goto err_buf; + + err = mlx4_buf_write_mtt(dev->dev, &buf->mtt, &buf->buf); + if (err) + goto err_mtt; + + return 0; + +err_mtt: + mlx4_mtt_cleanup(dev->dev, &buf->mtt); + +err_buf: + mlx4_buf_free(dev->dev, nent * sizeof(struct mlx4_cqe), + &buf->buf); + +out: + return err; +} + +static void mlx4_ib_free_cq_buf(struct mlx4_ib_dev *dev, struct mlx4_ib_cq_buf *buf, int cqe) +{ + mlx4_buf_free(dev->dev, (cqe + 1) * sizeof(struct mlx4_cqe), &buf->buf); +} + +static int mlx4_ib_get_cq_umem(struct mlx4_ib_dev *dev, struct ib_ucontext *context, + struct mlx4_ib_cq_buf *buf, struct ib_umem **umem, + u64 buf_addr, int cqe) +{ + int err; + + *umem = ib_umem_get(context, buf_addr, cqe * sizeof (struct mlx4_cqe), + IB_ACCESS_LOCAL_WRITE, 1); + if (IS_ERR(*umem)) + return PTR_ERR(*umem); + + err = mlx4_mtt_init(dev->dev, ib_umem_page_count(*umem), + ilog2((*umem)->page_size), &buf->mtt); + if (err) + goto err_buf; + + err = mlx4_ib_umem_write_mtt(dev, &buf->mtt, *umem); + if (err) + goto err_mtt; + + return 0; + +err_mtt: + mlx4_mtt_cleanup(dev->dev, &buf->mtt); + +err_buf: + ib_umem_release(*umem); + + return err; +} + struct ib_cq *mlx4_ib_create_cq(struct ib_device *ibdev, int entries, int vector, struct ib_ucontext *context, struct ib_udata *udata) @@ -98,7 +168,6 @@ struct ib_cq *mlx4_ib_create_cq(struct ib_device *ibdev, int entries, int vector struct mlx4_ib_dev *dev = to_mdev(ibdev); struct mlx4_ib_cq *cq; struct mlx4_uar *uar; - int buf_size; int err; if (entries < 1 || entries > dev->dev->caps.max_cqes) @@ -110,8 +179,10 @@ struct ib_cq *mlx4_ib_create_cq(struct ib_device *ibdev, int entries, int vector entries = roundup_pow_of_two(entries + 1); cq->ibcq.cqe = entries - 1; - buf_size = entries * sizeof (struct mlx4_cqe); + mutex_init(&cq->resize_mutex); spin_lock_init(&cq->lock); + cq->resize_buf = NULL; + cq->resize_umem = NULL; if (context) { struct mlx4_ib_create_cq ucmd; @@ -121,21 +192,10 @@ struct ib_cq *mlx4_ib_create_cq(struct ib_device *ibdev, int entries, int vector goto err_cq; } - cq->umem = ib_umem_get(context, ucmd.buf_addr, buf_size, - IB_ACCESS_LOCAL_WRITE); - if (IS_ERR(cq->umem)) { - err = PTR_ERR(cq->umem); - goto err_cq; - } - - err = mlx4_mtt_init(dev->dev, ib_umem_page_count(cq->umem), - ilog2(cq->umem->page_size), &cq->buf.mtt); + err = mlx4_ib_get_cq_umem(dev, context, &cq->buf, &cq->umem, + ucmd.buf_addr, entries); if (err) - goto err_buf; - - err = mlx4_ib_umem_write_mtt(dev, &cq->buf.mtt, cq->umem); - if (err) - goto err_mtt; + goto err_cq; err = mlx4_ib_db_map_user(to_mucontext(context), ucmd.db_addr, &cq->db); @@ -144,7 +204,7 @@ struct ib_cq *mlx4_ib_create_cq(struct ib_device *ibdev, int entries, int vector uar = &to_mucontext(context)->uar; } else { - err = mlx4_ib_db_alloc(dev, &cq->db, 1); + err = mlx4_db_alloc(dev->dev, &cq->db, 1); if (err) goto err_cq; @@ -153,25 +213,15 @@ struct ib_cq *mlx4_ib_create_cq(struct ib_device *ibdev, int entries, int vector *cq->mcq.set_ci_db = 0; *cq->mcq.arm_db = 0; - if (mlx4_buf_alloc(dev->dev, buf_size, PAGE_SIZE * 2, &cq->buf.buf)) { - err = -ENOMEM; - goto err_db; - } - - err = mlx4_mtt_init(dev->dev, cq->buf.buf.npages, cq->buf.buf.page_shift, - &cq->buf.mtt); + err = mlx4_ib_alloc_cq_buf(dev, &cq->buf, entries); if (err) - goto err_buf; - - err = mlx4_buf_write_mtt(dev->dev, &cq->buf.mtt, &cq->buf.buf); - if (err) - goto err_mtt; + goto err_db; uar = &dev->priv_uar; } err = mlx4_cq_alloc(dev->dev, entries, &cq->buf.mtt, uar, - cq->db.dma, &cq->mcq); + cq->db.dma, &cq->mcq, 0); if (err) goto err_dbmap; @@ -193,16 +243,14 @@ err_dbmap: err_mtt: mlx4_mtt_cleanup(dev->dev, &cq->buf.mtt); -err_buf: if (context) ib_umem_release(cq->umem); else - mlx4_buf_free(dev->dev, entries * sizeof (struct mlx4_cqe), - &cq->buf.buf); + mlx4_ib_free_cq_buf(dev, &cq->buf, cq->ibcq.cqe); err_db: if (!context) - mlx4_ib_db_free(dev, &cq->db); + mlx4_db_free(dev->dev, &cq->db); err_cq: kfree(cq); @@ -210,6 +258,170 @@ err_cq: return ERR_PTR(err); } +static int mlx4_alloc_resize_buf(struct mlx4_ib_dev *dev, struct mlx4_ib_cq *cq, + int entries) +{ + int err; + + if (cq->resize_buf) + return -EBUSY; + + cq->resize_buf = kmalloc(sizeof *cq->resize_buf, GFP_ATOMIC); + if (!cq->resize_buf) + return -ENOMEM; + + err = mlx4_ib_alloc_cq_buf(dev, &cq->resize_buf->buf, entries); + if (err) { + kfree(cq->resize_buf); + cq->resize_buf = NULL; + return err; + } + + cq->resize_buf->cqe = entries - 1; + + return 0; +} + +static int mlx4_alloc_resize_umem(struct mlx4_ib_dev *dev, struct mlx4_ib_cq *cq, + int entries, struct ib_udata *udata) +{ + struct mlx4_ib_resize_cq ucmd; + int err; + + if (cq->resize_umem) + return -EBUSY; + + if (ib_copy_from_udata(&ucmd, udata, sizeof ucmd)) + return -EFAULT; + + cq->resize_buf = kmalloc(sizeof *cq->resize_buf, GFP_ATOMIC); + if (!cq->resize_buf) + return -ENOMEM; + + err = mlx4_ib_get_cq_umem(dev, cq->umem->context, &cq->resize_buf->buf, + &cq->resize_umem, ucmd.buf_addr, entries); + if (err) { + kfree(cq->resize_buf); + cq->resize_buf = NULL; + return err; + } + + cq->resize_buf->cqe = entries - 1; + + return 0; +} + +static int mlx4_ib_get_outstanding_cqes(struct mlx4_ib_cq *cq) +{ + u32 i; + + i = cq->mcq.cons_index; + while (get_sw_cqe(cq, i & cq->ibcq.cqe)) + ++i; + + return i - cq->mcq.cons_index; +} + +static void mlx4_ib_cq_resize_copy_cqes(struct mlx4_ib_cq *cq) +{ + struct mlx4_cqe *cqe; + int i; + + i = cq->mcq.cons_index; + cqe = get_cqe(cq, i & cq->ibcq.cqe); + while ((cqe->owner_sr_opcode & MLX4_CQE_OPCODE_MASK) != MLX4_CQE_OPCODE_RESIZE) { + memcpy(get_cqe_from_buf(&cq->resize_buf->buf, + (i + 1) & cq->resize_buf->cqe), + get_cqe(cq, i & cq->ibcq.cqe), sizeof(struct mlx4_cqe)); + cqe = get_cqe(cq, ++i & cq->ibcq.cqe); + } + ++cq->mcq.cons_index; +} + +int mlx4_ib_resize_cq(struct ib_cq *ibcq, int entries, struct ib_udata *udata) +{ + struct mlx4_ib_dev *dev = to_mdev(ibcq->device); + struct mlx4_ib_cq *cq = to_mcq(ibcq); + int outst_cqe; + int err; + + mutex_lock(&cq->resize_mutex); + + if (entries < 1 || entries > dev->dev->caps.max_cqes) { + err = -EINVAL; + goto out; + } + + entries = roundup_pow_of_two(entries + 1); + if (entries == ibcq->cqe + 1) { + err = 0; + goto out; + } + + if (ibcq->uobject) { + err = mlx4_alloc_resize_umem(dev, cq, entries, udata); + if (err) + goto out; + } else { + /* Can't be smaller then the number of outstanding CQEs */ + outst_cqe = mlx4_ib_get_outstanding_cqes(cq); + if (entries < outst_cqe + 1) { + err = 0; + goto out; + } + + err = mlx4_alloc_resize_buf(dev, cq, entries); + if (err) + goto out; + } + + err = mlx4_cq_resize(dev->dev, &cq->mcq, entries, &cq->resize_buf->buf.mtt); + if (err) + goto err_buf; + + if (ibcq->uobject) { + cq->buf = cq->resize_buf->buf; + cq->ibcq.cqe = cq->resize_buf->cqe; + ib_umem_release(cq->umem); + cq->umem = cq->resize_umem; + + kfree(cq->resize_buf); + cq->resize_buf = NULL; + cq->resize_umem = NULL; + } else { + spin_lock_irq(&cq->lock); + if (cq->resize_buf) { + mlx4_ib_cq_resize_copy_cqes(cq); + mlx4_ib_free_cq_buf(dev, &cq->buf, cq->ibcq.cqe); + cq->buf = cq->resize_buf->buf; + cq->ibcq.cqe = cq->resize_buf->cqe; + + kfree(cq->resize_buf); + cq->resize_buf = NULL; + } + spin_unlock_irq(&cq->lock); + } + + goto out; + +err_buf: + if (!ibcq->uobject) + mlx4_ib_free_cq_buf(dev, &cq->resize_buf->buf, + cq->resize_buf->cqe); + + kfree(cq->resize_buf); + cq->resize_buf = NULL; + + if (cq->resize_umem) { + ib_umem_release(cq->resize_umem); + cq->resize_umem = NULL; + } + +out: + mutex_unlock(&cq->resize_mutex); + return err; +} + int mlx4_ib_destroy_cq(struct ib_cq *cq) { struct mlx4_ib_dev *dev = to_mdev(cq->device); @@ -222,9 +434,8 @@ int mlx4_ib_destroy_cq(struct ib_cq *cq) mlx4_ib_db_unmap_user(to_mucontext(cq->uobject->context), &mcq->db); ib_umem_release(mcq->umem); } else { - mlx4_buf_free(dev->dev, (cq->cqe + 1) * sizeof (struct mlx4_cqe), - &mcq->buf.buf); - mlx4_ib_db_free(dev, &mcq->db); + mlx4_ib_free_cq_buf(dev, &mcq->buf, cq->cqe); + mlx4_db_free(dev->dev, &mcq->db); } kfree(mcq); @@ -303,6 +514,20 @@ static void mlx4_ib_handle_error_cqe(struct mlx4_err_cqe *cqe, wc->vendor_err = cqe->vendor_err_syndrome; } +static int mlx4_ib_ipoib_csum_ok(__be32 status, __be16 checksum) +{ + return ((status & cpu_to_be32(MLX4_CQE_IPOIB_STATUS_IPV4 | + MLX4_CQE_IPOIB_STATUS_IPV4F | + MLX4_CQE_IPOIB_STATUS_IPV4OPT | + MLX4_CQE_IPOIB_STATUS_IPV6 | + MLX4_CQE_IPOIB_STATUS_IPOK)) == + cpu_to_be32(MLX4_CQE_IPOIB_STATUS_IPV4 | + MLX4_CQE_IPOIB_STATUS_IPOK)) && + (status & cpu_to_be32(MLX4_CQE_IPOIB_STATUS_UDP | + MLX4_CQE_IPOIB_STATUS_TCP)) && + checksum == cpu_to_be16(0xffff); +} + static int mlx4_ib_poll_one(struct mlx4_ib_cq *cq, struct mlx4_ib_qp **cur_qp, struct ib_wc *wc) @@ -313,8 +538,10 @@ static int mlx4_ib_poll_one(struct mlx4_ib_cq *cq, struct mlx4_ib_srq *srq; int is_send; int is_error; + u32 g_mlpath_rqpn; u16 wqe_ctr; +repoll: cqe = next_cqe_sw(cq); if (!cqe) return -EAGAIN; @@ -331,6 +558,28 @@ static int mlx4_ib_poll_one(struct mlx4_ib_cq *cq, is_error = (cqe->owner_sr_opcode & MLX4_CQE_OPCODE_MASK) == MLX4_CQE_OPCODE_ERROR; + if (unlikely((cqe->owner_sr_opcode & MLX4_CQE_OPCODE_MASK) == MLX4_OPCODE_NOP && + is_send)) { + printk(KERN_WARNING "Completion for NOP opcode detected!\n"); + return -EINVAL; + } + + /* Resize CQ in progress */ + if (unlikely((cqe->owner_sr_opcode & MLX4_CQE_OPCODE_MASK) == MLX4_CQE_OPCODE_RESIZE)) { + if (cq->resize_buf) { + struct mlx4_ib_dev *dev = to_mdev(cq->ibcq.device); + + mlx4_ib_free_cq_buf(dev, &cq->buf, cq->ibcq.cqe); + cq->buf = cq->resize_buf->buf; + cq->ibcq.cqe = cq->resize_buf->cqe; + + kfree(cq->resize_buf); + cq->resize_buf = NULL; + } + + goto repoll; + } + if (!*cur_qp || (be32_to_cpu(cqe->my_qpn) & 0xffffff) != (*cur_qp)->mqp.qpn) { /* @@ -353,9 +602,11 @@ static int mlx4_ib_poll_one(struct mlx4_ib_cq *cq, if (is_send) { wq = &(*cur_qp)->sq; - wqe_ctr = be16_to_cpu(cqe->wqe_index); - wq->tail += wqe_ctr - (u16) wq->tail; - wc->wr_id = wq->wrid[wq->tail & (wq->max - 1)]; + if (!(*cur_qp)->sq_signal_bits) { + wqe_ctr = be16_to_cpu(cqe->wqe_index); + wq->tail += (u16) (wqe_ctr - (u16) wq->tail); + } + wc->wr_id = wq->wrid[wq->tail & (wq->wqe_cnt - 1)]; ++wq->tail; } else if ((*cur_qp)->ibqp.srq) { srq = to_msrq((*cur_qp)->ibqp.srq); @@ -364,7 +615,7 @@ static int mlx4_ib_poll_one(struct mlx4_ib_cq *cq, mlx4_ib_free_srq_wqe(srq, wqe_ctr); } else { wq = &(*cur_qp)->rq; - wc->wr_id = wq->wrid[wq->tail & (wq->max - 1)]; + wc->wr_id = wq->wrid[wq->tail & (wq->wqe_cnt - 1)]; ++wq->tail; } @@ -389,7 +640,7 @@ static int mlx4_ib_poll_one(struct mlx4_ib_cq *cq, wc->opcode = IB_WC_SEND; break; case MLX4_OPCODE_RDMA_READ: - wc->opcode = IB_WC_SEND; + wc->opcode = IB_WC_RDMA_READ; wc->byte_len = be32_to_cpu(cqe->byte_cnt); break; case MLX4_OPCODE_ATOMIC_CS: @@ -403,34 +654,39 @@ static int mlx4_ib_poll_one(struct mlx4_ib_cq *cq, case MLX4_OPCODE_BIND_MW: wc->opcode = IB_WC_BIND_MW; break; + case MLX4_OPCODE_LSO: + wc->opcode = IB_WC_LSO; + break; } } else { wc->byte_len = be32_to_cpu(cqe->byte_cnt); switch (cqe->owner_sr_opcode & MLX4_CQE_OPCODE_MASK) { case MLX4_RECV_OPCODE_RDMA_WRITE_IMM: - wc->opcode = IB_WC_RECV_RDMA_WITH_IMM; - wc->wc_flags = IB_WC_WITH_IMM; - wc->imm_data = cqe->immed_rss_invalid; + wc->opcode = IB_WC_RECV_RDMA_WITH_IMM; + wc->wc_flags = IB_WC_WITH_IMM; + wc->ex.imm_data = cqe->immed_rss_invalid; break; case MLX4_RECV_OPCODE_SEND: wc->opcode = IB_WC_RECV; wc->wc_flags = 0; break; case MLX4_RECV_OPCODE_SEND_IMM: - wc->opcode = IB_WC_RECV; - wc->wc_flags = IB_WC_WITH_IMM; - wc->imm_data = cqe->immed_rss_invalid; + wc->opcode = IB_WC_RECV; + wc->wc_flags = IB_WC_WITH_IMM; + wc->ex.imm_data = cqe->immed_rss_invalid; break; } wc->slid = be16_to_cpu(cqe->rlid); wc->sl = cqe->sl >> 4; - wc->src_qp = be32_to_cpu(cqe->g_mlpath_rqpn) & 0xffffff; - wc->dlid_path_bits = (be32_to_cpu(cqe->g_mlpath_rqpn) >> 24) & 0x7f; - wc->wc_flags |= be32_to_cpu(cqe->g_mlpath_rqpn) & 0x80000000 ? - IB_WC_GRH : 0; - wc->pkey_index = be32_to_cpu(cqe->immed_rss_invalid) >> 16; + g_mlpath_rqpn = be32_to_cpu(cqe->g_mlpath_rqpn); + wc->src_qp = g_mlpath_rqpn & 0xffffff; + wc->dlid_path_bits = (g_mlpath_rqpn >> 24) & 0x7f; + wc->wc_flags |= g_mlpath_rqpn & 0x80000000 ? IB_WC_GRH : 0; + wc->pkey_index = be32_to_cpu(cqe->immed_rss_invalid) & 0x7f; + wc->csum_ok = mlx4_ib_ipoib_csum_ok(cqe->ipoib_status, + cqe->checksum); } return 0; @@ -478,7 +734,8 @@ void __mlx4_ib_cq_clean(struct mlx4_ib_cq *cq, u32 qpn, struct mlx4_ib_srq *srq) { u32 prod_index; int nfreed = 0; - struct mlx4_cqe *cqe; + struct mlx4_cqe *cqe, *dest; + u8 owner_bit; /* * First we need to find the current producer index, so we @@ -501,9 +758,13 @@ void __mlx4_ib_cq_clean(struct mlx4_ib_cq *cq, u32 qpn, struct mlx4_ib_srq *srq) if (srq && !(cqe->owner_sr_opcode & MLX4_CQE_IS_SEND_MASK)) mlx4_ib_free_srq_wqe(srq, be16_to_cpu(cqe->wqe_index)); ++nfreed; - } else if (nfreed) - memcpy(get_cqe(cq, (prod_index + nfreed) & cq->ibcq.cqe), - cqe, sizeof *cqe); + } else if (nfreed) { + dest = get_cqe(cq, (prod_index + nfreed) & cq->ibcq.cqe); + owner_bit = dest->owner_sr_opcode & MLX4_CQE_OWNER_MASK; + memcpy(dest, cqe, sizeof *cqe); + dest->owner_sr_opcode = owner_bit | + (dest->owner_sr_opcode & ~MLX4_CQE_OWNER_MASK); + } } if (nfreed) {