From mboxrd@z Thu Jan 1 00:00:00 1970 From: Yuval Shaia Subject: Re: [PATCH v2 03/15] IB/pvrdma: Add support for Completion Queues Date: Mon, 18 Jul 2016 17:12:22 +0300 Message-ID: <20160718141221.GA20068@yuval-lap.uk.oracle.com> References: <1468352205-9137-1-git-send-email-aditr@vmware.com> <1468352205-9137-4-git-send-email-aditr@vmware.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <1468352205-9137-4-git-send-email-aditr-pghWNbHTmq7QT0dZR+AlfA@public.gmane.org> Sender: linux-rdma-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: Adit Ranadive Cc: dledford-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org, linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, pv-drivers-pghWNbHTmq7QT0dZR+AlfA@public.gmane.org, jhansen-pghWNbHTmq7QT0dZR+AlfA@public.gmane.org, asarwade-pghWNbHTmq7QT0dZR+AlfA@public.gmane.org, georgezhang-pghWNbHTmq7QT0dZR+AlfA@public.gmane.org, bryantan-pghWNbHTmq7QT0dZR+AlfA@public.gmane.org List-Id: linux-rdma@vger.kernel.org On Tue, Jul 12, 2016 at 12:36:33PM -0700, Adit Ranadive wrote: > This patch adds the support for creating and destroying completion queues > on the paravirtual RDMA device. > > Reviewed-by: Jorgen Hansen > Reviewed-by: George Zhang > Reviewed-by: Aditya Sarwade > Reviewed-by: Bryan Tan > Signed-off-by: Adit Ranadive > --- > drivers/infiniband/hw/pvrdma/pvrdma_cq.c | 436 +++++++++++++++++++++++++++++++ > 1 file changed, 436 insertions(+) > create mode 100644 drivers/infiniband/hw/pvrdma/pvrdma_cq.c > > diff --git a/drivers/infiniband/hw/pvrdma/pvrdma_cq.c b/drivers/infiniband/hw/pvrdma/pvrdma_cq.c > new file mode 100644 > index 0000000..9a4b42c > --- /dev/null > +++ b/drivers/infiniband/hw/pvrdma/pvrdma_cq.c > @@ -0,0 +1,436 @@ > +/* > + * Copyright (c) 2012-2016 VMware, Inc. All rights reserved. > + * > + * This program is free software; you can redistribute it and/or > + * modify it under the terms of EITHER the GNU General Public License > + * version 2 as published by the Free Software Foundation or the BSD > + * 2-Clause License. This program is distributed in the hope that it > + * will be useful, but WITHOUT ANY WARRANTY; WITHOUT EVEN THE IMPLIED > + * WARRANTY OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. > + * See the GNU General Public License version 2 for more details at > + * http://www.gnu.org/licenses/old-licenses/gpl-2.0.en.html. > + * > + * You should have received a copy of the GNU General Public License > + * along with this program available in the file COPYING in the main > + * directory of this source tree. > + * > + * The BSD 2-Clause License > + * > + * 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. > + * > + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS > + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT > + * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS > + * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE > + * COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, > + * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES > + * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR > + * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) > + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, > + * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) > + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED > + * OF THE POSSIBILITY OF SUCH DAMAGE. > + */ > + > +#include > +#include > +#include > +#include > +#include > +#include > + > +#include "pvrdma.h" > +#include "pvrdma_user.h" > + > +/** > + * pvrdma_req_notify_cq - request notification for a completion queue > + * @ibcq: the completion queue > + * @notify_flags: notification flags > + * > + * @return: 0 for success. > + */ > +int pvrdma_req_notify_cq(struct ib_cq *ibcq, > + enum ib_cq_notify_flags notify_flags) > +{ > + struct pvrdma_dev *dev = to_vdev(ibcq->device); > + struct pvrdma_cq *cq = to_vcq(ibcq); > + u32 val = cq->cq_handle; > + > + val |= (notify_flags & IB_CQ_SOLICITED_MASK) == IB_CQ_SOLICITED ? > + PVRDMA_UAR_CQ_ARM_SOL : PVRDMA_UAR_CQ_ARM; > + > + writel(cpu_to_le32(val), dev->driver_uar.map + PVRDMA_UAR_CQ_OFFSET); > + > + return 0; > +} > + > +/** > + * pvrdma_create_cq - create completion queue > + * @ibdev: the device > + * @attr: completion queue attributes > + * @context: user context > + * @udata: user data > + * > + * @return: ib_cq completion queue pointer on success, > + * otherwise returns negative errno. > + */ > +struct ib_cq *pvrdma_create_cq(struct ib_device *ibdev, > + const struct ib_cq_init_attr *attr, > + struct ib_ucontext *context, > + struct ib_udata *udata) > +{ > + int entries = attr->cqe; > + struct pvrdma_dev *dev = to_vdev(ibdev); > + struct pvrdma_cq *cq; > + int ret; > + int npages; > + unsigned long flags; > + union pvrdma_cmd_req req; > + union pvrdma_cmd_resp rsp; > + struct pvrdma_cmd_create_cq *cmd = &req.create_cq; > + struct pvrdma_cmd_create_cq_resp *resp = &rsp.create_cq_resp; > + struct pvrdma_create_cq ucmd; > + > + BUILD_BUG_ON(sizeof(struct pvrdma_cqe) != 64); > + > + entries = roundup_pow_of_two(entries); > + if (entries < 1 || entries > dev->dsr->caps.max_cqe) > + return ERR_PTR(-EINVAL); > + > + if (!atomic_add_unless(&dev->num_cqs, 1, dev->dsr->caps.max_cq)) > + return ERR_PTR(-EINVAL); > + > + cq = kzalloc(sizeof(*cq), GFP_KERNEL); > + if (!cq) { > + atomic_dec(&dev->num_cqs); > + return ERR_PTR(-ENOMEM); > + } > + > + cq->ibcq.cqe = entries; > + > + if (context) { > + if (ib_copy_from_udata(&ucmd, udata, sizeof(ucmd))) { > + ret = -EFAULT; > + goto err_cq; > + } > + > + cq->umem = ib_umem_get(context, ucmd.buf_addr, ucmd.buf_size, > + IB_ACCESS_LOCAL_WRITE, 1); > + if (IS_ERR(cq->umem)) { > + ret = PTR_ERR(cq->umem); > + goto err_cq; > + } > + > + npages = ib_umem_page_count(cq->umem); > + } else { > + cq->is_kernel = true; > + > + /* One extra page for shared ring state */ > + npages = 1 + (entries * sizeof(struct pvrdma_cqe) + > + PAGE_SIZE - 1) / PAGE_SIZE; > + > + /* Skip header page. */ > + cq->offset = PAGE_SIZE; > + } > + > + if (npages < 0 || npages > PVRDMA_PAGE_DIR_MAX_PAGES) { > + dev_warn(&dev->pdev->dev, > + "overflow pages in completion queue\n"); > + ret = -EINVAL; > + goto err_umem; > + } > + > + ret = pvrdma_page_dir_init(dev, &cq->pdir, npages, cq->is_kernel); > + if (ret) { > + dev_warn(&dev->pdev->dev, > + "could not allocate page directory\n"); > + goto err_umem; > + } > + > + if (cq->is_kernel) { > + /* Ring state is always the first page. */ > + cq->ring_state = cq->pdir.pages[0]; > + } else { > + pvrdma_page_dir_insert_umem(&cq->pdir, cq->umem, 0); > + } > + > + atomic_set(&cq->refcnt, 1); > + init_waitqueue_head(&cq->wait); > + spin_lock_init(&cq->cq_lock); > + > + memset(cmd, 0, sizeof(*cmd)); > + cmd->hdr.cmd = PVRDMA_CMD_CREATE_CQ; > + cmd->nchunks = npages; > + cmd->ctx_handle = (context) ? > + (u64)to_vucontext(context)->ctx_handle : 0; > + cmd->cqe = entries; > + cmd->pdir_dma = cq->pdir.dir_dma; > + ret = pvrdma_cmd_post(dev, &req, true, &rsp); > + > + if (ret < 0 || rsp.hdr.ack != PVRDMA_CMD_CREATE_CQ_RESP) { > + dev_warn(&dev->pdev->dev, > + "could not create completion queue\n"); > + goto err_page_dir; > + } > + > + cq->ibcq.cqe = resp->cqe; > + cq->cq_handle = resp->cq_handle; > + spin_lock_irqsave(&dev->cq_tbl_lock, flags); > + dev->cq_tbl[cq->cq_handle % dev->dsr->caps.max_cq] = cq; > + spin_unlock_irqrestore(&dev->cq_tbl_lock, flags); > + > + if (context) { > + cq->uar = &(to_vucontext(context)->uar); > + > + /* Copy udata back. */ > + if (ib_copy_to_udata(udata, &cq->cq_handle, sizeof(__u32))) { > + dev_warn(&dev->pdev->dev, > + "failed to copy back udata\n"); > + ret = -EINVAL; > + goto err_page_dir; > + } > + } > + > + return &cq->ibcq; > + > +err_page_dir: > + pvrdma_page_dir_cleanup(dev, &cq->pdir); > +err_umem: > + if (context) > + ib_umem_release(cq->umem); > +err_cq: > + atomic_dec(&dev->num_cqs); > + kfree(cq); > + > + return ERR_PTR(ret); > +} > + > +static void pvrdma_free_cq(struct pvrdma_dev *dev, struct pvrdma_cq *cq) > +{ > + atomic_dec(&cq->refcnt); > + wait_event(cq->wait, !atomic_read(&cq->refcnt)); > + > + if (!cq->is_kernel) > + ib_umem_release(cq->umem); > + > + pvrdma_page_dir_cleanup(dev, &cq->pdir); > + kfree(cq); > +} > + > +/** > + * pvrdma_destroy_cq - destroy completion queue > + * @cq: the completion queue to destroy. > + * > + * @return: 0 for success. > + */ > +int pvrdma_destroy_cq(struct ib_cq *cq) > +{ > + struct pvrdma_cq *vcq = to_vcq(cq); > + union pvrdma_cmd_req req; > + union pvrdma_cmd_resp rsp; > + struct pvrdma_cmd_destroy_cq *cmd = &req.destroy_cq; > + struct pvrdma_dev *dev = to_vdev(cq->device); > + unsigned long flags; > + int ret; > + > + memset(cmd, 0, sizeof(*cmd)); > + cmd->hdr.cmd = PVRDMA_CMD_DESTROY_CQ; > + cmd->cq_handle = vcq->cq_handle; > + > + ret = pvrdma_cmd_post(dev, &req, false, &rsp); > + if (ret < 0) > + dev_warn(&dev->pdev->dev, > + "could not destroy completion queue\n"); > + > + /* free cq's resources */ > + spin_lock_irqsave(&dev->cq_tbl_lock, flags); > + dev->cq_tbl[vcq->cq_handle] = NULL; > + spin_unlock_irqrestore(&dev->cq_tbl_lock, flags); > + > + pvrdma_free_cq(dev, vcq); > + atomic_dec(&dev->num_cqs); > + > + return ret; > +} > + > +/** > + * pvrdma_modify_cq - modify the CQ moderation parameters > + * @ibcq: the CQ to modify > + * @cq_count: number of CQEs that will trigger an event > + * @cq_period: max period of time in usec before triggering an event > + * > + * @return: -EOPNOTSUPP as CQ resize is not supported. > + */ > +int pvrdma_modify_cq(struct ib_cq *cq, u16 cq_count, u16 cq_period) > +{ > + return -EOPNOTSUPP; > +} > + > +static inline struct pvrdma_cqe *get_cqe(struct pvrdma_cq *cq, int i) > +{ > + return (struct pvrdma_cqe *)pvrdma_page_dir_get_ptr( > + &cq->pdir, > + cq->offset + > + sizeof(struct pvrdma_cqe) * i); > +} > + > +void pvrdma_flush_cqe(struct pvrdma_qp *qp, struct pvrdma_cq *cq) > +{ > + int head; > + int has_data; > + > + if (!cq->is_kernel) > + return; > + > + /* Lock held */ > + has_data = pvrdma_idx_ring_has_data(&cq->ring_state->rx, > + cq->ibcq.cqe, &head); For easy review (and probably maintenance) it looks to me that we need to preserve some sort of order where function/structures/union/enum etc will be define in preceding patch in a patch-set. This function implemented in patch #11 while we are in patch #3. > + if (unlikely(has_data > 0)) { > + int items; > + int curr; > + int tail = pvrdma_idx(&cq->ring_state->rx.prod_tail, > + cq->ibcq.cqe); > + struct pvrdma_cqe *cqe; > + struct pvrdma_cqe *curr_cqe; > + > + items = (tail > head) ? (tail - head) : > + (cq->ibcq.cqe - head + tail); > + curr = --tail; > + while (items-- > 0) { > + if (curr < 0) > + curr = cq->ibcq.cqe - 1; > + if (tail < 0) > + tail = cq->ibcq.cqe - 1; > + curr_cqe = get_cqe(cq, curr); > + if ((curr_cqe->qp & 0xFFFF) != qp->qp_handle) { > + if (curr != tail) { > + cqe = get_cqe(cq, tail); > + *cqe = *curr_cqe; > + } > + tail--; > + } else { > + pvrdma_idx_ring_inc( > + &cq->ring_state->rx.cons_head, > + cq->ibcq.cqe); > + } > + curr--; > + } > + } > +} > + > +static int pvrdma_poll_one(struct pvrdma_cq *cq, struct pvrdma_qp **cur_qp, > + struct ib_wc *wc) > +{ > + struct pvrdma_dev *dev = to_vdev(cq->ibcq.device); > + int has_data; > + unsigned int head; > + bool tried = false; > + struct pvrdma_cqe *cqe; > + > +retry: > + has_data = pvrdma_idx_ring_has_data(&cq->ring_state->rx, > + cq->ibcq.cqe, &head); > + if (has_data == 0) { > + u32 val; > + > + if (tried) > + return -EAGAIN; > + > + /* Pass down POLL to give physical HCA a chance to poll. */ > + val = cq->cq_handle | PVRDMA_UAR_CQ_POLL; > + writel(cpu_to_le32(val), > + dev->driver_uar.map + PVRDMA_UAR_CQ_OFFSET); > + > + tried = true; > + goto retry; > + } else if (has_data == -1) { Is -1 represent fatal unrecoverable error or that next call to pvrdma_idx_ring_has_data might succeeds? > + return -EINVAL; > + } > + > + cqe = get_cqe(cq, head); > + > + /* Ensure cqe is valid. */ > + rmb(); > + if (dev->qp_tbl[cqe->qp & 0xffff]) > + *cur_qp = (struct pvrdma_qp *)dev->qp_tbl[cqe->qp & 0xffff]; > + else > + return -EINVAL; Is this can happen when QP was deleted while WR was processing? Or that in first place it was pushed with invalid QP? Can you explain why you choose EINVAL and not EAGAIN? i.e we polled the "corrupted" CQE, next CEQ should be fine. Reading pvrdma_poll_cq looks like that if it was requested to poll 10 CQE for example and the 9th fails - the function will fail them all. > + > + wc->opcode = pvrdma_wc_opcode_to_ib(cqe->opcode); > + wc->status = pvrdma_wc_status_to_ib(cqe->status); > + wc->wr_id = cqe->wr_id; > + wc->qp = &(*cur_qp)->ibqp; > + wc->byte_len = cqe->byte_len; > + wc->ex.imm_data = cqe->imm_data; > + wc->src_qp = cqe->src_qp; > + wc->wc_flags = pvrdma_wc_flags_to_ib(cqe->wc_flags); > + wc->pkey_index = cqe->pkey_index; > + wc->slid = cqe->slid; > + wc->sl = cqe->sl; > + wc->dlid_path_bits = cqe->dlid_path_bits; > + wc->port_num = cqe->port_num; > + wc->vendor_err = 0; > + > + /* Update shared ring state */ > + pvrdma_idx_ring_inc(&cq->ring_state->rx.cons_head, cq->ibcq.cqe); > + > + return 0; > +} > + > +/** > + * pvrdma_poll_cq - poll for work completion queue entries > + * @ibcq: completion queue > + * @num_entries: the maximum number of entries > + * @entry: pointer to work completion array > + * > + * @return: number of polled completion entries > + */ > +int pvrdma_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *wc) > +{ > + struct pvrdma_cq *cq = to_vcq(ibcq); > + struct pvrdma_qp *cur_qp = NULL; > + unsigned long flags; > + int npolled; > + int ret; > + > + if (num_entries < 1) > + return -EINVAL; > + > + spin_lock_irqsave(&cq->cq_lock, flags); > + for (npolled = 0; npolled < num_entries; ++npolled) { > + ret = pvrdma_poll_one(cq, &cur_qp, wc + npolled); > + if (ret) > + break; > + } > + > + spin_unlock_irqrestore(&cq->cq_lock, flags); > + > + if ((ret == 0) || (ret == -EAGAIN)) > + return npolled; > + else > + return ret; > +} > + > +/** > + * pvrdma_resize_cq - resize CQ > + * @ibcq: the completion queue > + * @entries: CQ entries > + * @udata: user data > + * > + * @return: -EOPNOTSUPP as CQ resize is not supported. > + */ > +int pvrdma_resize_cq(struct ib_cq *ibcq, int entries, struct ib_udata *udata) > +{ > + return -EOPNOTSUPP; > +} > -- > 2.7.4 > > -- > To unsubscribe from this list: send the line "unsubscribe linux-rdma" in > the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org > More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org More majordomo info at http://vger.kernel.org/majordomo-info.html