From 2e07a3d945851f0edc192336b3ac411e806c4da2 Mon Sep 17 00:00:00 2001 From: Wenpeng Liang Date: Sat, 30 Jan 2021 16:58:09 +0800 Subject: [PATCH] RDMA/hns: Refactor hns_roce_v2_post_srq_recv() The SRQ in the hns driver consists of the following four parts: * wqe buf: the buffer to store WQE. * wqe_idx buf: the cqe of SRQ may be not generated in the order of wqe, so the wqe_idx corresponding to the idle WQE needs to be pushed into the index queue which is a FIFO, then it instructs the hardware to obtain the corresponding WQE. * bitmap: bitmap is used to generate and release wqe_idx. When the user has a new WR, the driver finds the idx of the idle wqe in bitmap. When the CQE of wqe is generated, the driver will release the idx. * wr_id buf: wr_id buf is used to store the user's wr_id, then return it to the user when poll_cq verb is invoked. The process of post SRQ recv is refactored to make preceding code clearer. Link: https://lore.kernel.org/r/1611997090-48820-12-git-send-email-liweihang@huawei.com Signed-off-by: Wenpeng Liang Signed-off-by: Weihang Li Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 69 +++++++++++++++++++----------- 1 file changed, 43 insertions(+), 26 deletions(-) diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c index 7eba9b5..d5a63e4 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c @@ -869,13 +869,32 @@ static void hns_roce_free_srq_wqe(struct hns_roce_srq *srq, u32 wqe_index) spin_unlock(&srq->lock); } -int hns_roce_srqwq_overflow(struct hns_roce_srq *srq, u32 nreq) +static int hns_roce_srqwq_overflow(struct hns_roce_srq *srq) { struct hns_roce_idx_que *idx_que = &srq->idx_que; - unsigned int cur; - cur = idx_que->head - idx_que->tail; - return cur + nreq >= srq->wqe_cnt; + return idx_que->head - idx_que->tail >= srq->wqe_cnt; +} + +static int check_post_srq_valid(struct hns_roce_srq *srq, u32 max_sge, + const struct ib_recv_wr *wr) +{ + struct ib_device *ib_dev = srq->ibsrq.device; + + if (unlikely(wr->num_sge > max_sge)) { + ibdev_err(ib_dev, + "failed to check sge, wr->num_sge = %d, max_sge = %u.\n", + wr->num_sge, max_sge); + return -EINVAL; + } + + if (unlikely(hns_roce_srqwq_overflow(srq))) { + ibdev_err(ib_dev, + "failed to check srqwq status, srqwq is full.\n"); + return -ENOMEM; + } + + return 0; } static int get_srq_wqe_idx(struct hns_roce_srq *srq, u32 *wqe_idx) @@ -892,36 +911,40 @@ static int get_srq_wqe_idx(struct hns_roce_srq *srq, u32 *wqe_idx) return 0; } +static void fill_wqe_idx(struct hns_roce_srq *srq, unsigned int wqe_idx) +{ + struct hns_roce_idx_que *idx_que = &srq->idx_que; + unsigned int head; + __le32 *buf; + + head = idx_que->head & (srq->wqe_cnt - 1); + + buf = get_idx_buf(idx_que, head); + *buf = cpu_to_le32(wqe_idx); + + idx_que->head++; +} + static int hns_roce_v2_post_srq_recv(struct ib_srq *ibsrq, const struct ib_recv_wr *wr, const struct ib_recv_wr **bad_wr) { struct hns_roce_dev *hr_dev = to_hr_dev(ibsrq->device); struct hns_roce_srq *srq = to_hr_srq(ibsrq); - u32 wqe_idx, ind, nreq, max_sge; struct hns_roce_v2_db srq_db; unsigned long flags; - __le32 *srq_idx; int ret = 0; + u32 max_sge; + u32 wqe_idx; void *wqe; + u32 nreq; spin_lock_irqsave(&srq->lock, flags); - ind = srq->idx_que.head & (srq->wqe_cnt - 1); max_sge = srq->max_gs - srq->rsv_sge; - for (nreq = 0; wr; ++nreq, wr = wr->next) { - if (unlikely(wr->num_sge > max_sge)) { - ibdev_err(&hr_dev->ib_dev, - "srq: num_sge = %d, max_sge = %u.\n", - wr->num_sge, max_sge); - ret = -EINVAL; - *bad_wr = wr; - break; - } - - if (unlikely(hns_roce_srqwq_overflow(srq, nreq))) { - ret = -ENOMEM; + ret = check_post_srq_valid(srq, max_sge, wr); + if (ret) { *bad_wr = wr; break; } @@ -934,17 +957,11 @@ static int hns_roce_v2_post_srq_recv(struct ib_srq *ibsrq, wqe = get_srq_wqe_buf(srq, wqe_idx); fill_recv_sge_to_wqe(wr, wqe, max_sge, srq->rsv_sge); - - srq_idx = get_idx_buf(&srq->idx_que, ind); - *srq_idx = cpu_to_le32(wqe_idx); - + fill_wqe_idx(srq, wqe_idx); srq->wrid[wqe_idx] = wr->wr_id; - ind = (ind + 1) & (srq->wqe_cnt - 1); } if (likely(nreq)) { - srq->idx_que.head += nreq; - /* * Make sure that descriptors are written before * doorbell record. -- 2.7.4