From 6d10550c84f51cb7c449c51d1a6fd31185a6ba6c Mon Sep 17 00:00:00 2001 From: Allen Pais Date: Mon, 22 Oct 2018 19:30:43 +0530 Subject: RDMA/hns: prefer dma_zalloc_coherent to dma_alloc_coherent/memse This is a mechanical transformation, no change in logic. Signed-off-by: Allen Pais Signed-off-by: Doug Ledford --- drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 9 +++------ 1 file changed, 3 insertions(+), 6 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c index a4c62ae23a9a..d4609e7eb3c0 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c @@ -4972,13 +4972,12 @@ static int hns_roce_mhop_alloc_eq(struct hns_roce_dev *hr_dev, eqe_alloc = i * (buf_chk_sz / eq->eqe_size); size = (eq->entries - eqe_alloc) * eq->eqe_size; } - eq->buf[i] = dma_alloc_coherent(dev, size, + eq->buf[i] = dma_zalloc_coherent(dev, size, &(eq->buf_dma[i]), GFP_KERNEL); if (!eq->buf[i]) goto err_dma_alloc_buf; - memset(eq->buf[i], 0, size); *(eq->bt_l0 + i) = eq->buf_dma[i]; eq_buf_cnt++; @@ -5008,13 +5007,12 @@ static int hns_roce_mhop_alloc_eq(struct hns_roce_dev *hr_dev, size = (eq->entries - eqe_alloc) * eq->eqe_size; } - eq->buf[idx] = dma_alloc_coherent(dev, size, + eq->buf[idx] = dma_zalloc_coherent(dev, size, &(eq->buf_dma[idx]), GFP_KERNEL); if (!eq->buf[idx]) goto err_dma_alloc_buf; - memset(eq->buf[idx], 0, size); *(eq->bt_l1[i] + j) = eq->buf_dma[idx]; eq_buf_cnt++; @@ -5124,7 +5122,7 @@ static int hns_roce_v2_create_eq(struct hns_roce_dev *hr_dev, goto free_cmd_mbox; } - eq->buf_list->buf = dma_alloc_coherent(dev, buf_chk_sz, + eq->buf_list->buf = dma_zalloc_coherent(dev, buf_chk_sz, &(eq->buf_list->map), GFP_KERNEL); if (!eq->buf_list->buf) { @@ -5132,7 +5130,6 @@ static int hns_roce_v2_create_eq(struct hns_roce_dev *hr_dev, goto err_alloc_buf; } - memset(eq->buf_list->buf, 0, buf_chk_sz); } else { ret = hns_roce_mhop_alloc_eq(hr_dev, eq); if (ret) { -- cgit v1.2.3-59-g8ed1b From 9afc97c29b032af9a4112c2f4a02d5313b4dc71f Mon Sep 17 00:00:00 2001 From: Sagi Grimberg Date: Thu, 1 Nov 2018 09:13:12 -0700 Subject: mlx5: remove support for ib_get_vector_affinity Devices that does not use managed affinity can not export a vector affinity as the consumer relies on having a static mapping it can map to upper layer affinity (e.g. sw queues). If the driver allows the user to set the device irq affinity, then the affinitization of a long term existing entites is not relevant. For example, nvme-rdma controllers queue-irq affinitization is determined at init time so if the irq affinity changes over time, we are no longer aligned. Signed-off-by: Sagi Grimberg Acked-by: Leon Romanovsky Signed-off-by: Doug Ledford Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/main.c | 9 --------- include/linux/mlx5/driver.h | 6 ------ 2 files changed, 15 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index e9c428071df3..d4f1cac067cb 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -5332,14 +5332,6 @@ static void init_delay_drop(struct mlx5_ib_dev *dev) mlx5_ib_warn(dev, "Failed to init delay drop debugfs\n"); } -static const struct cpumask * -mlx5_ib_get_vector_affinity(struct ib_device *ibdev, int comp_vector) -{ - struct mlx5_ib_dev *dev = to_mdev(ibdev); - - return mlx5_get_vector_affinity_hint(dev->mdev, comp_vector); -} - /* The mlx5_ib_multiport_mutex should be held when calling this function */ static void mlx5_ib_unbind_slave_port(struct mlx5_ib_dev *ibdev, struct mlx5_ib_multiport_info *mpi) @@ -5845,7 +5837,6 @@ int mlx5_ib_stage_caps_init(struct mlx5_ib_dev *dev) dev->ib_dev.map_mr_sg = mlx5_ib_map_mr_sg; dev->ib_dev.check_mr_status = mlx5_ib_check_mr_status; dev->ib_dev.get_dev_fw_str = get_dev_fw_str; - dev->ib_dev.get_vector_affinity = mlx5_ib_get_vector_affinity; if (MLX5_CAP_GEN(mdev, ipoib_enhanced_offloads) && IS_ENABLED(CONFIG_MLX5_CORE_IPOIB)) dev->ib_dev.rdma_netdev_get_params = mlx5_ib_rn_get_params; diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h index aa5963b5d38e..6b6b3d9dcc00 100644 --- a/include/linux/mlx5/driver.h +++ b/include/linux/mlx5/driver.h @@ -1306,10 +1306,4 @@ enum { MLX5_TRIGGERED_CMD_COMP = (u64)1 << 32, }; -static inline const struct cpumask * -mlx5_get_vector_affinity_hint(struct mlx5_core_dev *dev, int vector) -{ - return dev->priv.irq_info[vector].mask; -} - #endif /* MLX5_DRIVER_H */ -- cgit v1.2.3-59-g8ed1b From 759ace7832802eaefbca821b2b43a44ab896b449 Mon Sep 17 00:00:00 2001 From: Sagi Grimberg Date: Thu, 1 Nov 2018 13:08:07 -0700 Subject: i40iw: remove support for ib_get_vector_affinity Devices that does not use managed affinity can not export a vector affinity as the consumer relies on having a static mapping it can map to upper layer affinity (e.g. sw queues). If the driver allows the user to set the device irq affinity, then the affinitization of a long term existing entites is not relevant. For example, nvme-rdma controllers queue-irq affinitization is determined at init time so if the irq affinity changes over time, we are no longer aligned. Signed-off-by: Sagi Grimberg Acked-by: Shiraz Saleem Signed-off-by: Doug Ledford Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/i40iw/i40iw_verbs.c | 20 -------------------- 1 file changed, 20 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/i40iw/i40iw_verbs.c b/drivers/infiniband/hw/i40iw/i40iw_verbs.c index 102875872bea..a773d1edf7fd 100644 --- a/drivers/infiniband/hw/i40iw/i40iw_verbs.c +++ b/drivers/infiniband/hw/i40iw/i40iw_verbs.c @@ -2721,25 +2721,6 @@ static int i40iw_query_pkey(struct ib_device *ibdev, return 0; } -/** - * i40iw_get_vector_affinity - report IRQ affinity mask - * @ibdev: IB device - * @comp_vector: completion vector index - */ -static const struct cpumask *i40iw_get_vector_affinity(struct ib_device *ibdev, - int comp_vector) -{ - struct i40iw_device *iwdev = to_iwdev(ibdev); - struct i40iw_msix_vector *msix_vec; - - if (iwdev->msix_shared) - msix_vec = &iwdev->iw_msixtbl[comp_vector]; - else - msix_vec = &iwdev->iw_msixtbl[comp_vector + 1]; - - return irq_get_affinity_mask(msix_vec->irq); -} - /** * i40iw_init_rdma_device - initialization of iwarp device * @iwdev: iwarp device @@ -2832,7 +2813,6 @@ static struct i40iw_ib_device *i40iw_init_rdma_device(struct i40iw_device *iwdev iwibdev->ibdev.req_notify_cq = i40iw_req_notify_cq; iwibdev->ibdev.post_send = i40iw_post_send; iwibdev->ibdev.post_recv = i40iw_post_recv; - iwibdev->ibdev.get_vector_affinity = i40iw_get_vector_affinity; return iwibdev; } -- cgit v1.2.3-59-g8ed1b From 5ec0304cdc35585085292d91867f15ec38afb134 Mon Sep 17 00:00:00 2001 From: Artemy Kovalyov Date: Mon, 5 Nov 2018 08:12:07 +0200 Subject: IB/mlx5: Allow modify AV in DCI QP to RTR This is required so the user can set the SL on the DC QP. Signed-off-by: Artemy Kovalyov Reviewed-by: Yossi Itigin Reviewed-by: Majd Dibbiny Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/qp.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c index 6841c0f9237f..f76f21ecb958 100644 --- a/drivers/infiniband/hw/mlx5/qp.c +++ b/drivers/infiniband/hw/mlx5/qp.c @@ -3516,7 +3516,7 @@ static bool modify_dci_qp_is_ok(enum ib_qp_state cur_state, enum ib_qp_state new return is_valid_mask(attr_mask, req, opt); } else if (cur_state == IB_QPS_INIT && new_state == IB_QPS_RTR) { req |= IB_QP_PATH_MTU; - opt = IB_QP_PKEY_INDEX; + opt = IB_QP_PKEY_INDEX | IB_QP_AV; return is_valid_mask(attr_mask, req, opt); } else if (cur_state == IB_QPS_RTR && new_state == IB_QPS_RTS) { req |= IB_QP_TIMEOUT | IB_QP_RETRY_CNT | IB_QP_RNR_RETRY | -- cgit v1.2.3-59-g8ed1b From fb670347559bc5e370c008e2673cd67a3e80b189 Mon Sep 17 00:00:00 2001 From: Sabyasachi Gupta Date: Fri, 9 Nov 2018 22:20:29 +0530 Subject: infiniband/hw/cxgb3/cxio_hal.c: Use dma_zalloc_coherent Replaced dma_alloc_coherent + memset with dma_zalloc_coherent Signed-off-by: Sabyasachi Gupta Acked-by: Steve Wise Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/cxgb3/cxio_hal.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/cxgb3/cxio_hal.c b/drivers/infiniband/hw/cxgb3/cxio_hal.c index dcb4bba522ba..df4f7a3f043d 100644 --- a/drivers/infiniband/hw/cxgb3/cxio_hal.c +++ b/drivers/infiniband/hw/cxgb3/cxio_hal.c @@ -291,13 +291,12 @@ int cxio_create_qp(struct cxio_rdev *rdev_p, u32 kernel_domain, if (!wq->sq) goto err3; - wq->queue = dma_alloc_coherent(&(rdev_p->rnic_info.pdev->dev), + wq->queue = dma_zalloc_coherent(&(rdev_p->rnic_info.pdev->dev), depth * sizeof(union t3_wr), &(wq->dma_addr), GFP_KERNEL); if (!wq->queue) goto err4; - memset(wq->queue, 0, depth * sizeof(union t3_wr)); dma_unmap_addr_set(wq, mapping, wq->dma_addr); wq->doorbell = (void __iomem *)rdev_p->rnic_info.kdb_addr; if (!kernel_domain) -- cgit v1.2.3-59-g8ed1b From 2b91fdfd13f5d3c3a5800731b4c46b0b5d69feaf Mon Sep 17 00:00:00 2001 From: Sabyasachi Gupta Date: Mon, 12 Nov 2018 20:51:59 +0530 Subject: infiniband/hw/cxgb4/qp.c: Use dma_zalloc_coherent Replaced dma_alloc_coherent + memset with dma_zalloc_coherent Signed-off-by: Sabyasachi Gupta Acked-by: Steve Wise Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/cxgb4/qp.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/cxgb4/qp.c b/drivers/infiniband/hw/cxgb4/qp.c index 13478f3b7057..5a8030bd4208 100644 --- a/drivers/infiniband/hw/cxgb4/qp.c +++ b/drivers/infiniband/hw/cxgb4/qp.c @@ -2564,13 +2564,12 @@ static int alloc_srq_queue(struct c4iw_srq *srq, struct c4iw_dev_ucontext *uctx, wq->rqt_abs_idx = (wq->rqt_hwaddr - rdev->lldi.vr->rq.start) >> T4_RQT_ENTRY_SHIFT; - wq->queue = dma_alloc_coherent(&rdev->lldi.pdev->dev, + wq->queue = dma_zalloc_coherent(&rdev->lldi.pdev->dev, wq->memsize, &wq->dma_addr, GFP_KERNEL); if (!wq->queue) goto err_free_rqtpool; - memset(wq->queue, 0, wq->memsize); dma_unmap_addr_set(wq, mapping, wq->dma_addr); wq->bar2_va = c4iw_bar2_addrs(rdev, wq->qid, CXGB4_BAR2_QTYPE_EGRESS, -- cgit v1.2.3-59-g8ed1b From 9828ca654b52848e7eb7dcc9b0994ff130dd4546 Mon Sep 17 00:00:00 2001 From: Steve Wise Date: Sat, 10 Nov 2018 05:27:39 -0800 Subject: iw_cxgb4: only reconnect with MPAv1 if the peer aborts Only retry connection setup with MPAv1 if the peer actually aborted the connection upon receiving the MPAv2 start message. This avoids retrying with MPAv1 in the case where the connection was aborted due to retransmit timeouts. Fixes: d2fe99e86bb2 ("RDMA/cxgb4: Add support for MPAv2 Enhanced RDMA Negotiation") Signed-off-by: Steve Wise Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/cxgb4/cm.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/cxgb4/cm.c b/drivers/infiniband/hw/cxgb4/cm.c index 615413bd3e8d..659175c9ae91 100644 --- a/drivers/infiniband/hw/cxgb4/cm.c +++ b/drivers/infiniband/hw/cxgb4/cm.c @@ -2795,7 +2795,8 @@ static int peer_abort(struct c4iw_dev *dev, struct sk_buff *skb) break; case MPA_REQ_SENT: (void)stop_ep_timer(ep); - if (mpa_rev == 1 || (mpa_rev == 2 && ep->tried_with_mpa_v1)) + if (status != CPL_ERR_CONN_RESET || mpa_rev == 1 || + (mpa_rev == 2 && ep->tried_with_mpa_v1)) connect_reply_upcall(ep, -ECONNRESET); else { /* -- cgit v1.2.3-59-g8ed1b From fc6526fba130dcbd496b96a9abf75a9382da95da Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Mon, 12 Nov 2018 12:41:55 +0100 Subject: mlx4: trigger IB events needed by SMC The mlx4 driver does not trigger an IB_EVENT_PORT_ACTIVE when the RoCE network interface is activated. When SMC determines the RoCE device port to be used, it checks the port states. This patch triggers IB events for NETDEV_UP and NETDEV_DOWN. Signed-off-by: Ursula Braun Acked-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx4/main.c | 27 +++++++++++++++++++++++++++ drivers/infiniband/hw/mlx4/mlx4_ib.h | 1 + 2 files changed, 28 insertions(+) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c index 0def2323459c..b73b5fa1822a 100644 --- a/drivers/infiniband/hw/mlx4/main.c +++ b/drivers/infiniband/hw/mlx4/main.c @@ -2352,6 +2352,32 @@ static void mlx4_ib_scan_netdevs(struct mlx4_ib_dev *ibdev, event == NETDEV_UP || event == NETDEV_CHANGE)) update_qps_port = port; + if (dev == iboe->netdevs[port - 1] && + (event == NETDEV_UP || event == NETDEV_DOWN)) { + enum ib_port_state port_state; + struct ib_event ibev = { }; + + if (ib_get_cached_port_state(&ibdev->ib_dev, port, + &port_state)) + continue; + + if (event == NETDEV_UP && + (port_state != IB_PORT_ACTIVE || + iboe->last_port_state[port - 1] != IB_PORT_DOWN)) + continue; + if (event == NETDEV_DOWN && + (port_state != IB_PORT_DOWN || + iboe->last_port_state[port - 1] != IB_PORT_ACTIVE)) + continue; + iboe->last_port_state[port - 1] = port_state; + + ibev.device = &ibdev->ib_dev; + ibev.element.port_num = port; + ibev.event = event == NETDEV_UP ? IB_EVENT_PORT_ACTIVE : + IB_EVENT_PORT_ERR; + ib_dispatch_event(&ibev); + } + } spin_unlock_bh(&iboe->lock); @@ -2710,6 +2736,7 @@ static void *mlx4_ib_add(struct mlx4_dev *dev) for (i = 0; i < ibdev->num_ports; ++i) { mutex_init(&ibdev->counters_table[i].mutex); INIT_LIST_HEAD(&ibdev->counters_table[i].counters_list); + iboe->last_port_state[i] = IB_PORT_DOWN; } num_req_counters = mlx4_is_bonded(dev) ? 1 : ibdev->num_ports; diff --git a/drivers/infiniband/hw/mlx4/mlx4_ib.h b/drivers/infiniband/hw/mlx4/mlx4_ib.h index 8850dfc3826d..5cb52424912e 100644 --- a/drivers/infiniband/hw/mlx4/mlx4_ib.h +++ b/drivers/infiniband/hw/mlx4/mlx4_ib.h @@ -519,6 +519,7 @@ struct mlx4_ib_iboe { atomic64_t mac[MLX4_MAX_PORTS]; struct notifier_block nb; struct mlx4_port_gid_table gids[MLX4_MAX_PORTS]; + enum ib_port_state last_port_state[MLX4_MAX_PORTS]; }; struct pkey_mgt { -- cgit v1.2.3-59-g8ed1b From dfb631a187b9b04c066df60e28adf05334112ca6 Mon Sep 17 00:00:00 2001 From: Jason Gunthorpe Date: Mon, 12 Nov 2018 22:59:49 +0200 Subject: RDMA/mlx5: Do not generate the uabi specs unconditionally For DM there is no reason not to add the spec for the START_OFFSET, if DM is not supported then ib_dev.alloc_dm is already set to NULL which ensures we do not call the method. For IPSEC, the core code should be setting ib_dev.create_flow_action_esp to NULL to disable it, not relying on wonky manipulation of the specs. Signed-off-by: Jason Gunthorpe Signed-off-by: Leon Romanovsky --- drivers/infiniband/hw/mlx5/main.c | 17 +++++++++-------- 1 file changed, 9 insertions(+), 8 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index 3e9057c1fedb..f12e045981fc 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -5555,12 +5555,8 @@ static int populate_specs_root(struct mlx5_ib_dev *dev) const struct uverbs_object_tree_def **trees = dev->driver_trees; size_t num_trees = 0; - if (mlx5_accel_ipsec_device_caps(dev->mdev) & - MLX5_ACCEL_IPSEC_CAP_DEVICE) - trees[num_trees++] = &mlx5_ib_flow_action; - - if (MLX5_CAP_DEV_MEM(dev->mdev, memic)) - trees[num_trees++] = &mlx5_ib_dm; + trees[num_trees++] = &mlx5_ib_flow_action; + trees[num_trees++] = &mlx5_ib_dm; if (MLX5_CAP_GEN_64(dev->mdev, general_obj_types) & MLX5_GENERAL_OBJ_TYPES_CAP_UCTX) @@ -5878,9 +5874,14 @@ int mlx5_ib_stage_caps_init(struct mlx5_ib_dev *dev) dev->ib_dev.uverbs_ex_cmd_mask |= (1ull << IB_USER_VERBS_EX_CMD_CREATE_FLOW) | (1ull << IB_USER_VERBS_EX_CMD_DESTROY_FLOW); - dev->ib_dev.create_flow_action_esp = mlx5_ib_create_flow_action_esp; + if (mlx5_accel_ipsec_device_caps(dev->mdev) & + MLX5_ACCEL_IPSEC_CAP_DEVICE) { + dev->ib_dev.create_flow_action_esp = + mlx5_ib_create_flow_action_esp; + dev->ib_dev.modify_flow_action_esp = + mlx5_ib_modify_flow_action_esp; + } dev->ib_dev.destroy_flow_action = mlx5_ib_destroy_flow_action; - dev->ib_dev.modify_flow_action_esp = mlx5_ib_modify_flow_action_esp; dev->ib_dev.driver_id = RDMA_DRIVER_MLX5; dev->ib_dev.create_counters = mlx5_ib_create_counters; dev->ib_dev.destroy_counters = mlx5_ib_destroy_counters; -- cgit v1.2.3-59-g8ed1b From 0cbf432db405289216747a8d31d74bab2452337c Mon Sep 17 00:00:00 2001 From: Jason Gunthorpe Date: Mon, 12 Nov 2018 22:59:50 +0200 Subject: RDMA/uverbs: Use a linear list to describe the compiled-in uapi The 'tree' data structure is very hard to build at compile time, and this makes it very limited. The new radix tree based compiler can handle a more complex input language that does not require the compiler to perfectly group everything into a neat tree structure. Instead use a simple list to describe to input, where the list elements can be of various different 'opcodes' instructing the radix compiler what to do. Start out with opcodes chaining to other definition lists and chaining to the existing 'tree' definition. Replace the very top level of the 'object tree' with this list type and get rid of struct uverbs_object_tree_def and DECLARE_UVERBS_OBJECT_TREE. Signed-off-by: Jason Gunthorpe Signed-off-by: Leon Romanovsky --- drivers/infiniband/core/rdma_core.h | 7 +- drivers/infiniband/core/uverbs_main.c | 2 +- drivers/infiniband/core/uverbs_std_types.c | 41 ++++----- drivers/infiniband/core/uverbs_uapi.c | 143 +++++++++++++++++------------ drivers/infiniband/hw/mlx5/devx.c | 15 ++- drivers/infiniband/hw/mlx5/flow.c | 20 ++-- drivers/infiniband/hw/mlx5/main.c | 31 +++++-- drivers/infiniband/hw/mlx5/mlx5_ib.h | 12 +-- include/rdma/ib_verbs.h | 2 +- include/rdma/uverbs_ioctl.h | 53 +++++++---- include/rdma/uverbs_named_ioctl.h | 11 +-- include/rdma/uverbs_std_types.h | 9 -- 12 files changed, 183 insertions(+), 163 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/core/rdma_core.h b/drivers/infiniband/core/rdma_core.h index 4886d2bba7c7..e39e9da1ff71 100644 --- a/drivers/infiniband/core/rdma_core.h +++ b/drivers/infiniband/core/rdma_core.h @@ -152,9 +152,8 @@ uapi_get_object(struct uverbs_api *uapi, u16 object_id) } char *uapi_key_format(char *S, unsigned int key); -struct uverbs_api *uverbs_alloc_api( - const struct uverbs_object_tree_def *const *driver_specs, - enum rdma_driver_id driver_id); +struct uverbs_api *uverbs_alloc_api(const struct uapi_definition *driver_def, + enum rdma_driver_id driver_id); void uverbs_disassociate_api_pre(struct ib_uverbs_device *uverbs_dev); void uverbs_disassociate_api(struct uverbs_api *uapi); void uverbs_destroy_api(struct uverbs_api *uapi); @@ -162,4 +161,6 @@ void uapi_compute_bundle_size(struct uverbs_api_ioctl_method *method_elm, unsigned int num_attrs); void uverbs_user_mmap_disassociate(struct ib_uverbs_file *ufile); +extern const struct uapi_definition uverbs_def_obj_intf[]; + #endif /* RDMA_CORE_H */ diff --git a/drivers/infiniband/core/uverbs_main.c b/drivers/infiniband/core/uverbs_main.c index 6d373f5515b7..9a3b88d0095a 100644 --- a/drivers/infiniband/core/uverbs_main.c +++ b/drivers/infiniband/core/uverbs_main.c @@ -1224,7 +1224,7 @@ static int ib_uverbs_create_uapi(struct ib_device *device, { struct uverbs_api *uapi; - uapi = uverbs_alloc_api(device->driver_specs, device->driver_id); + uapi = uverbs_alloc_api(device->driver_def, device->driver_id); if (IS_ERR(uapi)) return PTR_ERR(uapi); diff --git a/drivers/infiniband/core/uverbs_std_types.c b/drivers/infiniband/core/uverbs_std_types.c index 203cc96ac6f5..2db188ed825f 100644 --- a/drivers/infiniband/core/uverbs_std_types.c +++ b/drivers/infiniband/core/uverbs_std_types.c @@ -262,25 +262,22 @@ DECLARE_UVERBS_NAMED_OBJECT(UVERBS_OBJECT_PD, DECLARE_UVERBS_GLOBAL_METHODS(UVERBS_OBJECT_DEVICE); -DECLARE_UVERBS_OBJECT_TREE(uverbs_default_objects, - &UVERBS_OBJECT(UVERBS_OBJECT_DEVICE), - &UVERBS_OBJECT(UVERBS_OBJECT_PD), - &UVERBS_OBJECT(UVERBS_OBJECT_MR), - &UVERBS_OBJECT(UVERBS_OBJECT_COMP_CHANNEL), - &UVERBS_OBJECT(UVERBS_OBJECT_CQ), - &UVERBS_OBJECT(UVERBS_OBJECT_QP), - &UVERBS_OBJECT(UVERBS_OBJECT_AH), - &UVERBS_OBJECT(UVERBS_OBJECT_MW), - &UVERBS_OBJECT(UVERBS_OBJECT_SRQ), - &UVERBS_OBJECT(UVERBS_OBJECT_FLOW), - &UVERBS_OBJECT(UVERBS_OBJECT_WQ), - &UVERBS_OBJECT(UVERBS_OBJECT_RWQ_IND_TBL), - &UVERBS_OBJECT(UVERBS_OBJECT_XRCD), - &UVERBS_OBJECT(UVERBS_OBJECT_FLOW_ACTION), - &UVERBS_OBJECT(UVERBS_OBJECT_DM), - &UVERBS_OBJECT(UVERBS_OBJECT_COUNTERS)); - -const struct uverbs_object_tree_def *uverbs_default_get_objects(void) -{ - return &uverbs_default_objects; -} +const struct uapi_definition uverbs_def_obj_intf[] = { + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_DEVICE), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_PD), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_MR), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_COMP_CHANNEL), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_CQ), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_QP), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_AH), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_MW), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_SRQ), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_FLOW), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_WQ), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_RWQ_IND_TBL), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_XRCD), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_FLOW_ACTION), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_DM), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_COUNTERS), + {} +}; diff --git a/drivers/infiniband/core/uverbs_uapi.c b/drivers/infiniband/core/uverbs_uapi.c index 86f3fc5e04b4..67e8e96adb05 100644 --- a/drivers/infiniband/core/uverbs_uapi.c +++ b/drivers/infiniband/core/uverbs_uapi.c @@ -98,72 +98,96 @@ static int uapi_merge_method(struct uverbs_api *uapi, return 0; } -static int uapi_merge_tree(struct uverbs_api *uapi, - const struct uverbs_object_tree_def *tree, - bool is_driver) +static int uapi_merge_obj_tree(struct uverbs_api *uapi, + const struct uverbs_object_def *obj, + bool is_driver) { - unsigned int i, j; + struct uverbs_api_object *obj_elm; + unsigned int i; + u32 obj_key; int rc; - if (!tree->objects) + obj_key = uapi_key_obj(obj->id); + obj_elm = uapi_add_elm(uapi, obj_key, sizeof(*obj_elm)); + if (IS_ERR(obj_elm)) { + if (obj_elm != ERR_PTR(-EEXIST)) + return PTR_ERR(obj_elm); + + /* This occurs when a driver uses ADD_UVERBS_METHODS */ + if (WARN_ON(obj->type_attrs)) + return -EINVAL; + obj_elm = radix_tree_lookup(&uapi->radix, obj_key); + if (WARN_ON(!obj_elm)) + return -EINVAL; + } else { + obj_elm->type_attrs = obj->type_attrs; + if (obj->type_attrs) { + obj_elm->type_class = obj->type_attrs->type_class; + /* + * Today drivers are only permitted to use idr_class + * types. They cannot use FD types because we + * currently have no way to revoke the fops pointer + * after device disassociation. + */ + if (WARN_ON(is_driver && obj->type_attrs->type_class != + &uverbs_idr_class)) + return -EINVAL; + } + } + + if (!obj->methods) return 0; - for (i = 0; i != tree->num_objects; i++) { - const struct uverbs_object_def *obj = (*tree->objects)[i]; - struct uverbs_api_object *obj_elm; - u32 obj_key; + for (i = 0; i != obj->num_methods; i++) { + const struct uverbs_method_def *method = (*obj->methods)[i]; - if (!obj) + if (!method) continue; - obj_key = uapi_key_obj(obj->id); - obj_elm = uapi_add_elm(uapi, obj_key, sizeof(*obj_elm)); - if (IS_ERR(obj_elm)) { - if (obj_elm != ERR_PTR(-EEXIST)) - return PTR_ERR(obj_elm); + rc = uapi_merge_method(uapi, obj_elm, obj_key, method, + is_driver); + if (rc) + return rc; + } - /* This occurs when a driver uses ADD_UVERBS_METHODS */ - if (WARN_ON(obj->type_attrs)) - return -EINVAL; - obj_elm = radix_tree_lookup(&uapi->radix, obj_key); - if (WARN_ON(!obj_elm)) - return -EINVAL; - } else { - obj_elm->type_attrs = obj->type_attrs; - if (obj->type_attrs) { - obj_elm->type_class = - obj->type_attrs->type_class; - /* - * Today drivers are only permitted to use - * idr_class types. They cannot use FD types - * because we currently have no way to revoke - * the fops pointer after device - * disassociation. - */ - if (WARN_ON(is_driver && - obj->type_attrs->type_class != - &uverbs_idr_class)) - return -EINVAL; - } - } + return 0; +} + +static int uapi_merge_def(struct uverbs_api *uapi, + const struct uapi_definition *def_list, + bool is_driver) +{ + const struct uapi_definition *def = def_list; + int rc; + + if (!def_list) + return 0; - if (!obj->methods) + for (;; def++) { + switch ((enum uapi_definition_kind)def->kind) { + case UAPI_DEF_CHAIN: + rc = uapi_merge_def(uapi, def->chain, is_driver); + if (rc) + return rc; continue; - for (j = 0; j != obj->num_methods; j++) { - const struct uverbs_method_def *method = - (*obj->methods)[j]; - if (!method) - continue; + case UAPI_DEF_CHAIN_OBJ_TREE: + if (WARN_ON(def->object_start.object_id != + def->chain_obj_tree->id)) + return -EINVAL; - rc = uapi_merge_method(uapi, obj_elm, obj_key, method, - is_driver); + rc = uapi_merge_obj_tree(uapi, def->chain_obj_tree, + is_driver); if (rc) return rc; + continue; + + case UAPI_DEF_END: + return 0; } + WARN_ON(true); + return -EINVAL; } - - return 0; } static int @@ -263,9 +287,13 @@ void uverbs_destroy_api(struct uverbs_api *uapi) kfree(uapi); } -struct uverbs_api *uverbs_alloc_api( - const struct uverbs_object_tree_def *const *driver_specs, - enum rdma_driver_id driver_id) +static const struct uapi_definition uverbs_core_api[] = { + UAPI_DEF_CHAIN(uverbs_def_obj_intf), + {}, +}; + +struct uverbs_api *uverbs_alloc_api(const struct uapi_definition *driver_def, + enum rdma_driver_id driver_id) { struct uverbs_api *uapi; int rc; @@ -277,15 +305,12 @@ struct uverbs_api *uverbs_alloc_api( INIT_RADIX_TREE(&uapi->radix, GFP_KERNEL); uapi->driver_id = driver_id; - rc = uapi_merge_tree(uapi, uverbs_default_get_objects(), false); + rc = uapi_merge_def(uapi, uverbs_core_api, false); + if (rc) + goto err; + rc = uapi_merge_def(uapi, driver_def, true); if (rc) goto err; - - for (; driver_specs && *driver_specs; driver_specs++) { - rc = uapi_merge_tree(uapi, *driver_specs, true); - if (rc) - goto err; - } rc = uapi_finalize(uapi); if (rc) diff --git a/drivers/infiniband/hw/mlx5/devx.c b/drivers/infiniband/hw/mlx5/devx.c index 61aab7c0c513..992aca00f53c 100644 --- a/drivers/infiniband/hw/mlx5/devx.c +++ b/drivers/infiniband/hw/mlx5/devx.c @@ -1323,12 +1323,9 @@ DECLARE_UVERBS_NAMED_OBJECT(MLX5_IB_OBJECT_DEVX_UMEM, &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_UMEM_REG), &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_UMEM_DEREG)); -DECLARE_UVERBS_OBJECT_TREE(devx_objects, - &UVERBS_OBJECT(MLX5_IB_OBJECT_DEVX), - &UVERBS_OBJECT(MLX5_IB_OBJECT_DEVX_OBJ), - &UVERBS_OBJECT(MLX5_IB_OBJECT_DEVX_UMEM)); - -const struct uverbs_object_tree_def *mlx5_ib_get_devx_tree(void) -{ - return &devx_objects; -} +const struct uapi_definition mlx5_ib_devx_defs[] = { + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(MLX5_IB_OBJECT_DEVX), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(MLX5_IB_OBJECT_DEVX_OBJ), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(MLX5_IB_OBJECT_DEVX_UMEM), + {}, +}; diff --git a/drivers/infiniband/hw/mlx5/flow.c b/drivers/infiniband/hw/mlx5/flow.c index f86cdcafdafc..497ef18771f5 100644 --- a/drivers/infiniband/hw/mlx5/flow.c +++ b/drivers/infiniband/hw/mlx5/flow.c @@ -610,16 +610,10 @@ DECLARE_UVERBS_NAMED_OBJECT(MLX5_IB_OBJECT_FLOW_MATCHER, &UVERBS_METHOD(MLX5_IB_METHOD_FLOW_MATCHER_CREATE), &UVERBS_METHOD(MLX5_IB_METHOD_FLOW_MATCHER_DESTROY)); -DECLARE_UVERBS_OBJECT_TREE(flow_objects, - &UVERBS_OBJECT(MLX5_IB_OBJECT_FLOW_MATCHER)); - -int mlx5_ib_get_flow_trees(const struct uverbs_object_tree_def **root) -{ - int i = 0; - - root[i++] = &flow_objects; - root[i++] = &mlx5_ib_fs; - root[i++] = &mlx5_ib_flow_actions; - - return i; -} +const struct uapi_definition mlx5_ib_flow_defs[] = { + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(MLX5_IB_OBJECT_FLOW_MATCHER), + UAPI_DEF_CHAIN_OBJ_TREE(UVERBS_OBJECT_FLOW, &mlx5_ib_fs), + UAPI_DEF_CHAIN_OBJ_TREE(UVERBS_OBJECT_FLOW_ACTION, + &mlx5_ib_flow_actions), + {}, +}; diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index f12e045981fc..30a35a8ae0bb 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -5550,23 +5550,34 @@ ADD_UVERBS_ATTRIBUTES_SIMPLE( UVERBS_ATTR_FLAGS_IN(MLX5_IB_ATTR_CREATE_FLOW_ACTION_FLAGS, enum mlx5_ib_uapi_flow_action_flags)); +static const struct uapi_definition mlx5_ib_defs[] = { +#if IS_ENABLED(CONFIG_INFINIBAND_USER_ACCESS) + UAPI_DEF_CHAIN(mlx5_ib_flow_defs), +#endif + + UAPI_DEF_CHAIN_OBJ_TREE(UVERBS_OBJECT_FLOW_ACTION, + &mlx5_ib_flow_action), + UAPI_DEF_CHAIN_OBJ_TREE(UVERBS_OBJECT_DM, &mlx5_ib_dm), + {} +}; + static int populate_specs_root(struct mlx5_ib_dev *dev) { - const struct uverbs_object_tree_def **trees = dev->driver_trees; - size_t num_trees = 0; - - trees[num_trees++] = &mlx5_ib_flow_action; - trees[num_trees++] = &mlx5_ib_dm; + struct uapi_definition *defs = dev->driver_defs; +#if IS_ENABLED(CONFIG_INFINIBAND_USER_ACCESS) if (MLX5_CAP_GEN_64(dev->mdev, general_obj_types) & MLX5_GENERAL_OBJ_TYPES_CAP_UCTX) - trees[num_trees++] = mlx5_ib_get_devx_tree(); + *defs++ = (struct uapi_definition)UAPI_DEF_CHAIN( + mlx5_ib_devx_defs); +#endif - num_trees += mlx5_ib_get_flow_trees(trees + num_trees); + *defs++ = (struct uapi_definition)UAPI_DEF_CHAIN(mlx5_ib_defs); + *defs++ = (struct uapi_definition){}; + WARN_ON(defs - dev->driver_defs >= ARRAY_SIZE(dev->driver_defs)); - WARN_ON(num_trees >= ARRAY_SIZE(dev->driver_trees)); - trees[num_trees] = NULL; - dev->ib_dev.driver_specs = trees; + if (IS_ENABLED(CONFIG_INFINIBAND_USER_ACCESS)) + dev->ib_dev.driver_def = dev->driver_defs; return 0; } diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index 27999fd32356..3663664ecee8 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -891,7 +891,7 @@ struct mlx5_ib_pf_eq { struct mlx5_ib_dev { struct ib_device ib_dev; - const struct uverbs_object_tree_def *driver_trees[7]; + struct uapi_definition driver_defs[7]; struct mlx5_core_dev *mdev; struct mlx5_roce roce[MLX5_MAX_PORTS]; int num_ports; @@ -1264,29 +1264,23 @@ void mlx5_ib_put_native_port_mdev(struct mlx5_ib_dev *dev, int mlx5_ib_devx_create(struct mlx5_ib_dev *dev); void mlx5_ib_devx_destroy(struct mlx5_ib_dev *dev, u16 uid); const struct uverbs_object_tree_def *mlx5_ib_get_devx_tree(void); +extern const struct uapi_definition mlx5_ib_devx_defs[]; +extern const struct uapi_definition mlx5_ib_flow_defs[]; struct mlx5_ib_flow_handler *mlx5_ib_raw_fs_rule_add( struct mlx5_ib_dev *dev, struct mlx5_ib_flow_matcher *fs_matcher, struct mlx5_flow_act *flow_act, void *cmd_in, int inlen, int dest_id, int dest_type); bool mlx5_ib_devx_is_flow_dest(void *obj, int *dest_id, int *dest_type); -int mlx5_ib_get_flow_trees(const struct uverbs_object_tree_def **root); void mlx5_ib_destroy_flow_action_raw(struct mlx5_ib_flow_action *maction); #else static inline int mlx5_ib_devx_create(struct mlx5_ib_dev *dev) { return -EOPNOTSUPP; }; static inline void mlx5_ib_devx_destroy(struct mlx5_ib_dev *dev, u16 uid) {} -static inline const struct uverbs_object_tree_def * -mlx5_ib_get_devx_tree(void) { return NULL; } static inline bool mlx5_ib_devx_is_flow_dest(void *obj, int *dest_id, int *dest_type) { return false; } -static inline int -mlx5_ib_get_flow_trees(const struct uverbs_object_tree_def **root) -{ - return 0; -} static inline void mlx5_ib_destroy_flow_action_raw(struct mlx5_ib_flow_action *maction) { diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h index 86313e1e9210..1af7bf34c04c 100644 --- a/include/rdma/ib_verbs.h +++ b/include/rdma/ib_verbs.h @@ -2600,7 +2600,7 @@ struct ib_device { const struct cpumask *(*get_vector_affinity)(struct ib_device *ibdev, int comp_vector); - const struct uverbs_object_tree_def *const *driver_specs; + const struct uapi_definition *driver_def; enum rdma_driver_id driver_id; }; diff --git a/include/rdma/uverbs_ioctl.h b/include/rdma/uverbs_ioctl.h index 84d3d15f1f38..9fa0cca45783 100644 --- a/include/rdma/uverbs_ioctl.h +++ b/include/rdma/uverbs_ioctl.h @@ -296,11 +296,43 @@ struct uverbs_object_def { const struct uverbs_method_def * const (*methods)[]; }; -struct uverbs_object_tree_def { - size_t num_objects; - const struct uverbs_object_def * const (*objects)[]; +enum uapi_definition_kind { + UAPI_DEF_END = 0, + UAPI_DEF_CHAIN_OBJ_TREE, + UAPI_DEF_CHAIN, }; +struct uapi_definition { + u8 kind; + union { + struct { + u16 object_id; + } object_start; + }; + + union { + const struct uapi_definition *chain; + const struct uverbs_object_def *chain_obj_tree; + }; +}; + +/* Include another struct uapi_definition in this one */ +#define UAPI_DEF_CHAIN(_def_var) \ + { \ + .kind = UAPI_DEF_CHAIN, .chain = _def_var, \ + } + +/* Temporary until the tree base description is replaced */ +#define UAPI_DEF_CHAIN_OBJ_TREE(_object_enum, _object_ptr) \ + { \ + .kind = UAPI_DEF_CHAIN_OBJ_TREE, \ + .object_start = { .object_id = _object_enum }, \ + .chain_obj_tree = _object_ptr, \ + } +#define UAPI_DEF_CHAIN_OBJ_TREE_NAMED(_object_enum, ...) \ + UAPI_DEF_CHAIN_OBJ_TREE(_object_enum, &UVERBS_OBJECT(_object_enum)), \ + ##__VA_ARGS__ + /* * ======================================= * Attribute Specifications @@ -438,21 +470,6 @@ struct uverbs_object_tree_def { UVERBS_ATTR_MIN_SIZE(0), \ UA_OPTIONAL) -/* - * ======================================= - * Declaration helpers - * ======================================= - */ - -#define DECLARE_UVERBS_OBJECT_TREE(_name, ...) \ - static const struct uverbs_object_def *const _name##_ptr[] = { \ - __VA_ARGS__, \ - }; \ - static const struct uverbs_object_tree_def _name = { \ - .num_objects = ARRAY_SIZE(_name##_ptr), \ - .objects = &_name##_ptr, \ - } - /* ================================================= * Parsing infrastructure * ================================================= diff --git a/include/rdma/uverbs_named_ioctl.h b/include/rdma/uverbs_named_ioctl.h index b3b21733cc55..b586e7c41d4a 100644 --- a/include/rdma/uverbs_named_ioctl.h +++ b/include/rdma/uverbs_named_ioctl.h @@ -102,18 +102,11 @@ #define ADD_UVERBS_METHODS(_name, _object_id, ...) \ static const struct uverbs_method_def *const UVERBS_OBJECT_METHODS( \ _object_id)[] = { __VA_ARGS__ }; \ - static const struct uverbs_object_def _name##_struct = { \ + static const struct uverbs_object_def _name = { \ .id = _object_id, \ .num_methods = ARRAY_SIZE(UVERBS_OBJECT_METHODS(_object_id)), \ .methods = &UVERBS_OBJECT_METHODS(_object_id) \ - }; \ - static const struct uverbs_object_def *const _name##_ptrs[] = { \ - &_name##_struct, \ - }; \ - static const struct uverbs_object_tree_def _name = { \ - .num_objects = 1, \ - .objects = &_name##_ptrs, \ - } + }; /* Used by drivers to declare a complete parsing tree for a single method that * differs only in having additional driver specific attributes. diff --git a/include/rdma/uverbs_std_types.h b/include/rdma/uverbs_std_types.h index 3db2802fbc68..e3154f8d3cda 100644 --- a/include/rdma/uverbs_std_types.h +++ b/include/rdma/uverbs_std_types.h @@ -37,15 +37,6 @@ #include #include -#if IS_ENABLED(CONFIG_INFINIBAND_USER_ACCESS) -const struct uverbs_object_tree_def *uverbs_default_get_objects(void); -#else -static inline const struct uverbs_object_tree_def *uverbs_default_get_objects(void) -{ - return NULL; -} -#endif - /* Returns _id, or causes a compile error if _id is not a u32. * * The uobj APIs should only be used with the write based uAPI to access -- cgit v1.2.3-59-g8ed1b From 36e235c8829935a59d4652c878cffb08229205c2 Mon Sep 17 00:00:00 2001 From: Jason Gunthorpe Date: Mon, 12 Nov 2018 22:59:53 +0200 Subject: RDMA/mlx5: Use the uapi disablement APIs instead of code Rely on UAPI_DEF_IS_OBJ_SUPPORTED instead of manipulating the contents of the driver's definition list. Signed-off-by: Jason Gunthorpe Signed-off-by: Leon Romanovsky --- drivers/infiniband/hw/mlx5/devx.c | 20 +++++++++++++++++--- drivers/infiniband/hw/mlx5/main.c | 36 ++++-------------------------------- drivers/infiniband/hw/mlx5/mlx5_ib.h | 2 -- 3 files changed, 21 insertions(+), 37 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/devx.c b/drivers/infiniband/hw/mlx5/devx.c index 992aca00f53c..ee8db8d9e919 100644 --- a/drivers/infiniband/hw/mlx5/devx.c +++ b/drivers/infiniband/hw/mlx5/devx.c @@ -1323,9 +1323,23 @@ DECLARE_UVERBS_NAMED_OBJECT(MLX5_IB_OBJECT_DEVX_UMEM, &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_UMEM_REG), &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_UMEM_DEREG)); +static bool devx_is_supported(struct ib_device *device) +{ + struct mlx5_ib_dev *dev = to_mdev(device); + + return !dev->rep && MLX5_CAP_GEN_64(dev->mdev, general_obj_types) & + MLX5_GENERAL_OBJ_TYPES_CAP_UCTX; +} + const struct uapi_definition mlx5_ib_devx_defs[] = { - UAPI_DEF_CHAIN_OBJ_TREE_NAMED(MLX5_IB_OBJECT_DEVX), - UAPI_DEF_CHAIN_OBJ_TREE_NAMED(MLX5_IB_OBJECT_DEVX_OBJ), - UAPI_DEF_CHAIN_OBJ_TREE_NAMED(MLX5_IB_OBJECT_DEVX_UMEM), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED( + MLX5_IB_OBJECT_DEVX, + UAPI_DEF_IS_OBJ_SUPPORTED(devx_is_supported)), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED( + MLX5_IB_OBJECT_DEVX_OBJ, + UAPI_DEF_IS_OBJ_SUPPORTED(devx_is_supported)), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED( + MLX5_IB_OBJECT_DEVX_UMEM, + UAPI_DEF_IS_OBJ_SUPPORTED(devx_is_supported)), {}, }; diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index 30a35a8ae0bb..0707ede7dcdd 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -5552,6 +5552,7 @@ ADD_UVERBS_ATTRIBUTES_SIMPLE( static const struct uapi_definition mlx5_ib_defs[] = { #if IS_ENABLED(CONFIG_INFINIBAND_USER_ACCESS) + UAPI_DEF_CHAIN(mlx5_ib_devx_defs), UAPI_DEF_CHAIN(mlx5_ib_flow_defs), #endif @@ -5561,27 +5562,6 @@ static const struct uapi_definition mlx5_ib_defs[] = { {} }; -static int populate_specs_root(struct mlx5_ib_dev *dev) -{ - struct uapi_definition *defs = dev->driver_defs; - -#if IS_ENABLED(CONFIG_INFINIBAND_USER_ACCESS) - if (MLX5_CAP_GEN_64(dev->mdev, general_obj_types) & - MLX5_GENERAL_OBJ_TYPES_CAP_UCTX) - *defs++ = (struct uapi_definition)UAPI_DEF_CHAIN( - mlx5_ib_devx_defs); -#endif - - *defs++ = (struct uapi_definition)UAPI_DEF_CHAIN(mlx5_ib_defs); - *defs++ = (struct uapi_definition){}; - WARN_ON(defs - dev->driver_defs >= ARRAY_SIZE(dev->driver_defs)); - - if (IS_ENABLED(CONFIG_INFINIBAND_USER_ACCESS)) - dev->ib_dev.driver_def = dev->driver_defs; - - return 0; -} - static int mlx5_ib_read_counters(struct ib_counters *counters, struct ib_counters_read_attr *read_attr, struct uverbs_attr_bundle *attrs) @@ -5898,6 +5878,9 @@ int mlx5_ib_stage_caps_init(struct mlx5_ib_dev *dev) dev->ib_dev.destroy_counters = mlx5_ib_destroy_counters; dev->ib_dev.read_counters = mlx5_ib_read_counters; + if (IS_ENABLED(CONFIG_INFINIBAND_USER_ACCESS)) + dev->ib_dev.driver_def = mlx5_ib_defs; + err = init_node_data(dev); if (err) return err; @@ -6110,11 +6093,6 @@ void mlx5_ib_stage_bfrag_cleanup(struct mlx5_ib_dev *dev) mlx5_free_bfreg(dev->mdev, &dev->bfreg); } -static int mlx5_ib_stage_populate_specs(struct mlx5_ib_dev *dev) -{ - return populate_specs_root(dev); -} - int mlx5_ib_stage_ib_reg_init(struct mlx5_ib_dev *dev) { const char *name; @@ -6249,9 +6227,6 @@ static const struct mlx5_ib_profile pf_profile = { STAGE_CREATE(MLX5_IB_STAGE_PRE_IB_REG_UMR, NULL, mlx5_ib_stage_pre_ib_reg_umr_cleanup), - STAGE_CREATE(MLX5_IB_STAGE_SPECS, - mlx5_ib_stage_populate_specs, - NULL), STAGE_CREATE(MLX5_IB_STAGE_IB_REG, mlx5_ib_stage_ib_reg_init, mlx5_ib_stage_ib_reg_cleanup), @@ -6294,9 +6269,6 @@ static const struct mlx5_ib_profile nic_rep_profile = { STAGE_CREATE(MLX5_IB_STAGE_PRE_IB_REG_UMR, NULL, mlx5_ib_stage_pre_ib_reg_umr_cleanup), - STAGE_CREATE(MLX5_IB_STAGE_SPECS, - mlx5_ib_stage_populate_specs, - NULL), STAGE_CREATE(MLX5_IB_STAGE_IB_REG, mlx5_ib_stage_ib_reg_init, mlx5_ib_stage_ib_reg_cleanup), diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index 3663664ecee8..9b434246d4e3 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -781,7 +781,6 @@ enum mlx5_ib_stages { MLX5_IB_STAGE_UAR, MLX5_IB_STAGE_BFREG, MLX5_IB_STAGE_PRE_IB_REG_UMR, - MLX5_IB_STAGE_SPECS, MLX5_IB_STAGE_IB_REG, MLX5_IB_STAGE_POST_IB_REG_UMR, MLX5_IB_STAGE_DELAY_DROP, @@ -891,7 +890,6 @@ struct mlx5_ib_pf_eq { struct mlx5_ib_dev { struct ib_device ib_dev; - struct uapi_definition driver_defs[7]; struct mlx5_core_dev *mdev; struct mlx5_roce roce[MLX5_MAX_PORTS]; int num_ports; -- cgit v1.2.3-59-g8ed1b From d2c9d9abe117a470753283bf5884fea6f4af15d4 Mon Sep 17 00:00:00 2001 From: Colin Ian King Date: Mon, 26 Nov 2018 16:23:20 +0000 Subject: IB/qib: fix spelling mistake "colescing" -> "coalescing" There is a spelling mistake in the module description text, fix it. Signed-off-by: Colin Ian King Reviewed-by: Dennis Dalessandro Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/qib/qib_pcie.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/qib/qib_pcie.c b/drivers/infiniband/hw/qib/qib_pcie.c index 30595b358d8f..864f2af171f7 100644 --- a/drivers/infiniband/hw/qib/qib_pcie.c +++ b/drivers/infiniband/hw/qib/qib_pcie.c @@ -387,7 +387,7 @@ void qib_pcie_reenable(struct qib_devdata *dd, u16 cmd, u8 iline, u8 cline) static int qib_pcie_coalesce; module_param_named(pcie_coalesce, qib_pcie_coalesce, int, S_IRUGO); -MODULE_PARM_DESC(pcie_coalesce, "tune PCIe colescing on some Intel chipsets"); +MODULE_PARM_DESC(pcie_coalesce, "tune PCIe coalescing on some Intel chipsets"); /* * Enable PCIe completion and data coalescing, on Intel 5x00 and 7300 -- cgit v1.2.3-59-g8ed1b From 8313c10fa8be032fccc1e757bccc21207f533127 Mon Sep 17 00:00:00 2001 From: Jason Gunthorpe Date: Sun, 25 Nov 2018 20:51:13 +0200 Subject: RDMA/uverbs: Replace ib_uverbs_file with uverbs_attr_bundle for write Now that we can add meta-data to the description of write() methods we need to pass the uverbs_attr_bundle into all write based handlers so future patches can use it as a container for any new data transferred out of the core. This is the first step to bringing the write() and ioctl() methods to a common interface signature. This is a simple search/replace, and we push the attr down into the uobj and other APIs to keep changes minimal. Signed-off-by: Jason Gunthorpe Signed-off-by: Leon Romanovsky --- drivers/infiniband/core/rdma_core.c | 13 +- drivers/infiniband/core/rdma_core.h | 8 +- drivers/infiniband/core/uverbs_cmd.c | 301 +++++++++++++++++----------------- drivers/infiniband/core/uverbs_main.c | 12 +- drivers/infiniband/core/uverbs_uapi.c | 4 +- drivers/infiniband/hw/mlx5/devx.c | 2 +- include/rdma/ib_verbs.h | 2 +- include/rdma/uverbs_ioctl.h | 10 +- include/rdma/uverbs_std_types.h | 48 +++--- 9 files changed, 208 insertions(+), 192 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/core/rdma_core.c b/drivers/infiniband/core/rdma_core.c index 752a55c6bdce..cf671a643d4c 100644 --- a/drivers/infiniband/core/rdma_core.c +++ b/drivers/infiniband/core/rdma_core.c @@ -224,12 +224,14 @@ out_unlock: * uverbs_put_destroy. */ struct ib_uobject *__uobj_get_destroy(const struct uverbs_api_object *obj, - u32 id, struct ib_uverbs_file *ufile) + u32 id, + const struct uverbs_attr_bundle *attrs) { struct ib_uobject *uobj; int ret; - uobj = rdma_lookup_get_uobject(obj, ufile, id, UVERBS_LOOKUP_DESTROY); + uobj = rdma_lookup_get_uobject(obj, attrs->ufile, id, + UVERBS_LOOKUP_DESTROY); if (IS_ERR(uobj)) return uobj; @@ -248,11 +250,12 @@ struct ib_uobject *__uobj_get_destroy(const struct uverbs_api_object *obj, * the uobj. */ int __uobj_perform_destroy(const struct uverbs_api_object *obj, u32 id, - struct ib_uverbs_file *ufile, int success_res) + const struct uverbs_attr_bundle *attrs, + int success_res) { struct ib_uobject *uobj; - uobj = __uobj_get_destroy(obj, id, ufile); + uobj = __uobj_get_destroy(obj, id, attrs); if (IS_ERR(uobj)) return PTR_ERR(uobj); @@ -267,7 +270,7 @@ static struct ib_uobject *alloc_uobj(struct ib_uverbs_file *ufile, struct ib_uobject *uobj; struct ib_ucontext *ucontext; - ucontext = ib_uverbs_get_ucontext(ufile); + ucontext = ib_uverbs_get_ucontext_file(ufile); if (IS_ERR(ucontext)) return ERR_CAST(ucontext); diff --git a/drivers/infiniband/core/rdma_core.h b/drivers/infiniband/core/rdma_core.h index f9039cfb28b4..edd299174d95 100644 --- a/drivers/infiniband/core/rdma_core.h +++ b/drivers/infiniband/core/rdma_core.h @@ -137,10 +137,10 @@ struct uverbs_api_ioctl_method { }; struct uverbs_api_write_method { - ssize_t (*handler)(struct ib_uverbs_file *file, const char __user *buf, - int in_len, int out_len); - int (*handler_ex)(struct ib_uverbs_file *file, struct ib_udata *ucore, - struct ib_udata *uhw); + ssize_t (*handler)(struct uverbs_attr_bundle *attrs, + const char __user *buf, int in_len, int out_len); + int (*handler_ex)(struct uverbs_attr_bundle *attrs, + struct ib_udata *ucore, struct ib_udata *uhw); u8 disabled:1; u8 is_ex:1; }; diff --git a/drivers/infiniband/core/uverbs_cmd.c b/drivers/infiniband/core/uverbs_cmd.c index e2fcaf398a4d..a49926a63ce0 100644 --- a/drivers/infiniband/core/uverbs_cmd.c +++ b/drivers/infiniband/core/uverbs_cmd.c @@ -48,10 +48,10 @@ #include "core_priv.h" static struct ib_uverbs_completion_event_file * -_ib_uverbs_lookup_comp_file(s32 fd, struct ib_uverbs_file *ufile) +_ib_uverbs_lookup_comp_file(s32 fd, const struct uverbs_attr_bundle *attrs) { struct ib_uobject *uobj = ufd_get_read(UVERBS_OBJECT_COMP_CHANNEL, - fd, ufile); + fd, attrs); if (IS_ERR(uobj)) return (void *)uobj; @@ -65,10 +65,11 @@ _ib_uverbs_lookup_comp_file(s32 fd, struct ib_uverbs_file *ufile) #define ib_uverbs_lookup_comp_file(_fd, _ufile) \ _ib_uverbs_lookup_comp_file((_fd)*typecheck(s32, _fd), _ufile) -static ssize_t ib_uverbs_get_context(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_get_context(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { + struct ib_uverbs_file *file = attrs->ufile; struct ib_uverbs_get_context cmd; struct ib_uverbs_get_context_resp resp; struct ib_udata udata; @@ -224,7 +225,7 @@ static void copy_query_dev_fields(struct ib_ucontext *ucontext, resp->phys_port_cnt = ib_dev->phys_port_cnt; } -static ssize_t ib_uverbs_query_device(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_query_device(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -232,7 +233,7 @@ static ssize_t ib_uverbs_query_device(struct ib_uverbs_file *file, struct ib_uverbs_query_device_resp resp; struct ib_ucontext *ucontext; - ucontext = ib_uverbs_get_ucontext(file); + ucontext = ib_uverbs_get_ucontext(attrs); if (IS_ERR(ucontext)) return PTR_ERR(ucontext); @@ -272,7 +273,7 @@ static u32 make_port_cap_flags(const struct ib_port_attr *attr) return res; } -static ssize_t ib_uverbs_query_port(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_query_port(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -283,7 +284,7 @@ static ssize_t ib_uverbs_query_port(struct ib_uverbs_file *file, struct ib_ucontext *ucontext; struct ib_device *ib_dev; - ucontext = ib_uverbs_get_ucontext(file); + ucontext = ib_uverbs_get_ucontext(attrs); if (IS_ERR(ucontext)) return PTR_ERR(ucontext); ib_dev = ucontext->device; @@ -337,7 +338,7 @@ static ssize_t ib_uverbs_query_port(struct ib_uverbs_file *file, return in_len; } -static ssize_t ib_uverbs_alloc_pd(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_alloc_pd(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -360,7 +361,7 @@ static ssize_t ib_uverbs_alloc_pd(struct ib_uverbs_file *file, in_len - sizeof(cmd) - sizeof(struct ib_uverbs_cmd_hdr), out_len - sizeof(resp)); - uobj = uobj_alloc(UVERBS_OBJECT_PD, file, &ib_dev); + uobj = uobj_alloc(UVERBS_OBJECT_PD, attrs, &ib_dev); if (IS_ERR(uobj)) return PTR_ERR(uobj); @@ -396,7 +397,7 @@ err: return ret; } -static ssize_t ib_uverbs_dealloc_pd(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_dealloc_pd(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -405,7 +406,7 @@ static ssize_t ib_uverbs_dealloc_pd(struct ib_uverbs_file *file, if (copy_from_user(&cmd, buf, sizeof cmd)) return -EFAULT; - return uobj_perform_destroy(UVERBS_OBJECT_PD, cmd.pd_handle, file, + return uobj_perform_destroy(UVERBS_OBJECT_PD, cmd.pd_handle, attrs, in_len); } @@ -494,10 +495,11 @@ static void xrcd_table_delete(struct ib_uverbs_device *dev, } } -static ssize_t ib_uverbs_open_xrcd(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_open_xrcd(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { + struct ib_uverbs_device *ibudev = attrs->ufile->device; struct ib_uverbs_open_xrcd cmd; struct ib_uverbs_open_xrcd_resp resp; struct ib_udata udata; @@ -520,7 +522,7 @@ static ssize_t ib_uverbs_open_xrcd(struct ib_uverbs_file *file, in_len - sizeof(cmd) - sizeof(struct ib_uverbs_cmd_hdr), out_len - sizeof(resp)); - mutex_lock(&file->device->xrcd_tree_mutex); + mutex_lock(&ibudev->xrcd_tree_mutex); if (cmd.fd != -1) { /* search for file descriptor */ @@ -531,7 +533,7 @@ static ssize_t ib_uverbs_open_xrcd(struct ib_uverbs_file *file, } inode = file_inode(f.file); - xrcd = find_xrcd(file->device, inode); + xrcd = find_xrcd(ibudev, inode); if (!xrcd && !(cmd.oflags & O_CREAT)) { /* no file descriptor. Need CREATE flag */ ret = -EAGAIN; @@ -544,7 +546,7 @@ static ssize_t ib_uverbs_open_xrcd(struct ib_uverbs_file *file, } } - obj = (struct ib_uxrcd_object *)uobj_alloc(UVERBS_OBJECT_XRCD, file, + obj = (struct ib_uxrcd_object *)uobj_alloc(UVERBS_OBJECT_XRCD, attrs, &ib_dev); if (IS_ERR(obj)) { ret = PTR_ERR(obj); @@ -574,7 +576,7 @@ static ssize_t ib_uverbs_open_xrcd(struct ib_uverbs_file *file, if (inode) { if (new_xrcd) { /* create new inode/xrcd table entry */ - ret = xrcd_table_insert(file->device, inode, xrcd); + ret = xrcd_table_insert(ibudev, inode, xrcd); if (ret) goto err_dealloc_xrcd; } @@ -589,14 +591,14 @@ static ssize_t ib_uverbs_open_xrcd(struct ib_uverbs_file *file, if (f.file) fdput(f); - mutex_unlock(&file->device->xrcd_tree_mutex); + mutex_unlock(&ibudev->xrcd_tree_mutex); return uobj_alloc_commit(&obj->uobject, in_len); err_copy: if (inode) { if (new_xrcd) - xrcd_table_delete(file->device, inode); + xrcd_table_delete(ibudev, inode); atomic_dec(&xrcd->usecnt); } @@ -610,12 +612,12 @@ err_tree_mutex_unlock: if (f.file) fdput(f); - mutex_unlock(&file->device->xrcd_tree_mutex); + mutex_unlock(&ibudev->xrcd_tree_mutex); return ret; } -static ssize_t ib_uverbs_close_xrcd(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_close_xrcd(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -624,7 +626,7 @@ static ssize_t ib_uverbs_close_xrcd(struct ib_uverbs_file *file, if (copy_from_user(&cmd, buf, sizeof cmd)) return -EFAULT; - return uobj_perform_destroy(UVERBS_OBJECT_XRCD, cmd.xrcd_handle, file, + return uobj_perform_destroy(UVERBS_OBJECT_XRCD, cmd.xrcd_handle, attrs, in_len); } @@ -653,7 +655,7 @@ int ib_uverbs_dealloc_xrcd(struct ib_uobject *uobject, return ret; } -static ssize_t ib_uverbs_reg_mr(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_reg_mr(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { struct ib_uverbs_reg_mr cmd; @@ -683,11 +685,11 @@ static ssize_t ib_uverbs_reg_mr(struct ib_uverbs_file *file, if (ret) return ret; - uobj = uobj_alloc(UVERBS_OBJECT_MR, file, &ib_dev); + uobj = uobj_alloc(UVERBS_OBJECT_MR, attrs, &ib_dev); if (IS_ERR(uobj)) return PTR_ERR(uobj); - pd = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd.pd_handle, file); + pd = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd.pd_handle, attrs); if (!pd) { ret = -EINVAL; goto err_free; @@ -744,7 +746,7 @@ err_free: return ret; } -static ssize_t ib_uverbs_rereg_mr(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_rereg_mr(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -776,7 +778,7 @@ static ssize_t ib_uverbs_rereg_mr(struct ib_uverbs_file *file, (cmd.start & ~PAGE_MASK) != (cmd.hca_va & ~PAGE_MASK))) return -EINVAL; - uobj = uobj_get_write(UVERBS_OBJECT_MR, cmd.mr_handle, file); + uobj = uobj_get_write(UVERBS_OBJECT_MR, cmd.mr_handle, attrs); if (IS_ERR(uobj)) return PTR_ERR(uobj); @@ -795,7 +797,7 @@ static ssize_t ib_uverbs_rereg_mr(struct ib_uverbs_file *file, if (cmd.flags & IB_MR_REREG_PD) { pd = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd.pd_handle, - file); + attrs); if (!pd) { ret = -EINVAL; goto put_uobjs; @@ -835,7 +837,7 @@ put_uobjs: return ret; } -static ssize_t ib_uverbs_dereg_mr(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_dereg_mr(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -844,11 +846,11 @@ static ssize_t ib_uverbs_dereg_mr(struct ib_uverbs_file *file, if (copy_from_user(&cmd, buf, sizeof cmd)) return -EFAULT; - return uobj_perform_destroy(UVERBS_OBJECT_MR, cmd.mr_handle, file, + return uobj_perform_destroy(UVERBS_OBJECT_MR, cmd.mr_handle, attrs, in_len); } -static ssize_t ib_uverbs_alloc_mw(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_alloc_mw(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -867,11 +869,11 @@ static ssize_t ib_uverbs_alloc_mw(struct ib_uverbs_file *file, if (copy_from_user(&cmd, buf, sizeof(cmd))) return -EFAULT; - uobj = uobj_alloc(UVERBS_OBJECT_MW, file, &ib_dev); + uobj = uobj_alloc(UVERBS_OBJECT_MW, attrs, &ib_dev); if (IS_ERR(uobj)) return PTR_ERR(uobj); - pd = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd.pd_handle, file); + pd = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd.pd_handle, attrs); if (!pd) { ret = -EINVAL; goto err_free; @@ -916,7 +918,7 @@ err_free: return ret; } -static ssize_t ib_uverbs_dealloc_mw(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_dealloc_mw(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -925,11 +927,11 @@ static ssize_t ib_uverbs_dealloc_mw(struct ib_uverbs_file *file, if (copy_from_user(&cmd, buf, sizeof(cmd))) return -EFAULT; - return uobj_perform_destroy(UVERBS_OBJECT_MW, cmd.mw_handle, file, + return uobj_perform_destroy(UVERBS_OBJECT_MW, cmd.mw_handle, attrs, in_len); } -static ssize_t ib_uverbs_create_comp_channel(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_create_comp_channel(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -945,7 +947,7 @@ static ssize_t ib_uverbs_create_comp_channel(struct ib_uverbs_file *file, if (copy_from_user(&cmd, buf, sizeof cmd)) return -EFAULT; - uobj = uobj_alloc(UVERBS_OBJECT_COMP_CHANNEL, file, &ib_dev); + uobj = uobj_alloc(UVERBS_OBJECT_COMP_CHANNEL, attrs, &ib_dev); if (IS_ERR(uobj)) return PTR_ERR(uobj); @@ -963,12 +965,12 @@ static ssize_t ib_uverbs_create_comp_channel(struct ib_uverbs_file *file, return uobj_alloc_commit(uobj, in_len); } -static struct ib_ucq_object *create_cq(struct ib_uverbs_file *file, +static struct ib_ucq_object *create_cq(struct uverbs_attr_bundle *attrs, struct ib_udata *ucore, struct ib_udata *uhw, struct ib_uverbs_ex_create_cq *cmd, size_t cmd_sz, - int (*cb)(struct ib_uverbs_file *file, + int (*cb)(struct uverbs_attr_bundle *attrs, struct ib_ucq_object *obj, struct ib_uverbs_ex_create_cq_resp *resp, struct ib_udata *udata, @@ -983,16 +985,16 @@ static struct ib_ucq_object *create_cq(struct ib_uverbs_file *file, struct ib_cq_init_attr attr = {}; struct ib_device *ib_dev; - if (cmd->comp_vector >= file->device->num_comp_vectors) + if (cmd->comp_vector >= attrs->ufile->device->num_comp_vectors) return ERR_PTR(-EINVAL); - obj = (struct ib_ucq_object *)uobj_alloc(UVERBS_OBJECT_CQ, file, + obj = (struct ib_ucq_object *)uobj_alloc(UVERBS_OBJECT_CQ, attrs, &ib_dev); if (IS_ERR(obj)) return obj; if (cmd->comp_channel >= 0) { - ev_file = ib_uverbs_lookup_comp_file(cmd->comp_channel, file); + ev_file = ib_uverbs_lookup_comp_file(cmd->comp_channel, attrs); if (IS_ERR(ev_file)) { ret = PTR_ERR(ev_file); goto err; @@ -1035,7 +1037,7 @@ static struct ib_ucq_object *create_cq(struct ib_uverbs_file *file, cq->res.type = RDMA_RESTRACK_CQ; rdma_restrack_add(&cq->res); - ret = cb(file, obj, &resp, ucore, context); + ret = cb(attrs, obj, &resp, ucore, context); if (ret) goto err_cb; @@ -1049,7 +1051,7 @@ err_cb: err_file: if (ev_file) - ib_uverbs_release_ucq(file, ev_file, obj); + ib_uverbs_release_ucq(attrs->ufile, ev_file, obj); err: uobj_alloc_abort(&obj->uobject); @@ -1057,7 +1059,7 @@ err: return ERR_PTR(ret); } -static int ib_uverbs_create_cq_cb(struct ib_uverbs_file *file, +static int ib_uverbs_create_cq_cb(struct uverbs_attr_bundle *attrs, struct ib_ucq_object *obj, struct ib_uverbs_ex_create_cq_resp *resp, struct ib_udata *ucore, void *context) @@ -1068,7 +1070,7 @@ static int ib_uverbs_create_cq_cb(struct ib_uverbs_file *file, return 0; } -static ssize_t ib_uverbs_create_cq(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_create_cq(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -1099,7 +1101,7 @@ static ssize_t ib_uverbs_create_cq(struct ib_uverbs_file *file, cmd_ex.comp_vector = cmd.comp_vector; cmd_ex.comp_channel = cmd.comp_channel; - obj = create_cq(file, &ucore, &uhw, &cmd_ex, + obj = create_cq(attrs, &ucore, &uhw, &cmd_ex, offsetof(typeof(cmd_ex), comp_channel) + sizeof(cmd.comp_channel), ib_uverbs_create_cq_cb, NULL); @@ -1110,7 +1112,7 @@ static ssize_t ib_uverbs_create_cq(struct ib_uverbs_file *file, return in_len; } -static int ib_uverbs_ex_create_cq_cb(struct ib_uverbs_file *file, +static int ib_uverbs_ex_create_cq_cb(struct uverbs_attr_bundle *attrs, struct ib_ucq_object *obj, struct ib_uverbs_ex_create_cq_resp *resp, struct ib_udata *ucore, void *context) @@ -1121,7 +1123,7 @@ static int ib_uverbs_ex_create_cq_cb(struct ib_uverbs_file *file, return 0; } -static int ib_uverbs_ex_create_cq(struct ib_uverbs_file *file, +static int ib_uverbs_ex_create_cq(struct uverbs_attr_bundle *attrs, struct ib_udata *ucore, struct ib_udata *uhw) { struct ib_uverbs_ex_create_cq_resp resp; @@ -1146,14 +1148,14 @@ static int ib_uverbs_ex_create_cq(struct ib_uverbs_file *file, sizeof(resp.response_length))) return -ENOSPC; - obj = create_cq(file, ucore, uhw, &cmd, + obj = create_cq(attrs, ucore, uhw, &cmd, min(ucore->inlen, sizeof(cmd)), ib_uverbs_ex_create_cq_cb, NULL); return PTR_ERR_OR_ZERO(obj); } -static ssize_t ib_uverbs_resize_cq(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_resize_cq(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -1171,7 +1173,7 @@ static ssize_t ib_uverbs_resize_cq(struct ib_uverbs_file *file, in_len - sizeof(cmd) - sizeof(struct ib_uverbs_cmd_hdr), out_len - sizeof(resp)); - cq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, cmd.cq_handle, file); + cq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, cmd.cq_handle, attrs); if (!cq) return -EINVAL; @@ -1220,7 +1222,7 @@ static int copy_wc_to_user(struct ib_device *ib_dev, void __user *dest, return 0; } -static ssize_t ib_uverbs_poll_cq(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_poll_cq(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -1235,7 +1237,7 @@ static ssize_t ib_uverbs_poll_cq(struct ib_uverbs_file *file, if (copy_from_user(&cmd, buf, sizeof cmd)) return -EFAULT; - cq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, cmd.cq_handle, file); + cq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, cmd.cq_handle, attrs); if (!cq) return -EINVAL; @@ -1271,7 +1273,7 @@ out_put: return ret; } -static ssize_t ib_uverbs_req_notify_cq(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_req_notify_cq(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -1281,7 +1283,7 @@ static ssize_t ib_uverbs_req_notify_cq(struct ib_uverbs_file *file, if (copy_from_user(&cmd, buf, sizeof cmd)) return -EFAULT; - cq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, cmd.cq_handle, file); + cq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, cmd.cq_handle, attrs); if (!cq) return -EINVAL; @@ -1293,7 +1295,7 @@ static ssize_t ib_uverbs_req_notify_cq(struct ib_uverbs_file *file, return in_len; } -static ssize_t ib_uverbs_destroy_cq(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_destroy_cq(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -1305,7 +1307,7 @@ static ssize_t ib_uverbs_destroy_cq(struct ib_uverbs_file *file, if (copy_from_user(&cmd, buf, sizeof cmd)) return -EFAULT; - uobj = uobj_get_destroy(UVERBS_OBJECT_CQ, cmd.cq_handle, file); + uobj = uobj_get_destroy(UVERBS_OBJECT_CQ, cmd.cq_handle, attrs); if (IS_ERR(uobj)) return PTR_ERR(uobj); @@ -1322,12 +1324,12 @@ static ssize_t ib_uverbs_destroy_cq(struct ib_uverbs_file *file, return in_len; } -static int create_qp(struct ib_uverbs_file *file, +static int create_qp(struct uverbs_attr_bundle *attrs, struct ib_udata *ucore, struct ib_udata *uhw, struct ib_uverbs_ex_create_qp *cmd, size_t cmd_sz, - int (*cb)(struct ib_uverbs_file *file, + int (*cb)(struct uverbs_attr_bundle *attrs, struct ib_uverbs_ex_create_qp_resp *resp, struct ib_udata *udata), void *context) @@ -1351,7 +1353,7 @@ static int create_qp(struct ib_uverbs_file *file, if (cmd->qp_type == IB_QPT_RAW_PACKET && !capable(CAP_NET_RAW)) return -EPERM; - obj = (struct ib_uqp_object *)uobj_alloc(UVERBS_OBJECT_QP, file, + obj = (struct ib_uqp_object *)uobj_alloc(UVERBS_OBJECT_QP, attrs, &ib_dev); if (IS_ERR(obj)) return PTR_ERR(obj); @@ -1364,7 +1366,7 @@ static int create_qp(struct ib_uverbs_file *file, (cmd->comp_mask & IB_UVERBS_CREATE_QP_MASK_IND_TABLE)) { ind_tbl = uobj_get_obj_read(rwq_ind_table, UVERBS_OBJECT_RWQ_IND_TBL, - cmd->rwq_ind_tbl_handle, file); + cmd->rwq_ind_tbl_handle, attrs); if (!ind_tbl) { ret = -EINVAL; goto err_put; @@ -1390,7 +1392,7 @@ static int create_qp(struct ib_uverbs_file *file, if (cmd->qp_type == IB_QPT_XRC_TGT) { xrcd_uobj = uobj_get_read(UVERBS_OBJECT_XRCD, cmd->pd_handle, - file); + attrs); if (IS_ERR(xrcd_uobj)) { ret = -EINVAL; @@ -1410,7 +1412,7 @@ static int create_qp(struct ib_uverbs_file *file, } else { if (cmd->is_srq) { srq = uobj_get_obj_read(srq, UVERBS_OBJECT_SRQ, - cmd->srq_handle, file); + cmd->srq_handle, attrs); if (!srq || srq->srq_type == IB_SRQT_XRC) { ret = -EINVAL; goto err_put; @@ -1421,7 +1423,7 @@ static int create_qp(struct ib_uverbs_file *file, if (cmd->recv_cq_handle != cmd->send_cq_handle) { rcq = uobj_get_obj_read( cq, UVERBS_OBJECT_CQ, - cmd->recv_cq_handle, file); + cmd->recv_cq_handle, attrs); if (!rcq) { ret = -EINVAL; goto err_put; @@ -1432,11 +1434,11 @@ static int create_qp(struct ib_uverbs_file *file, if (has_sq) scq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, - cmd->send_cq_handle, file); + cmd->send_cq_handle, attrs); if (!ind_tbl) rcq = rcq ?: scq; pd = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd->pd_handle, - file); + attrs); if (!pd || (!scq && has_sq)) { ret = -EINVAL; goto err_put; @@ -1446,7 +1448,7 @@ static int create_qp(struct ib_uverbs_file *file, } attr.event_handler = ib_uverbs_qp_event_handler; - attr.qp_context = file; + attr.qp_context = attrs->ufile; attr.send_cq = scq; attr.recv_cq = rcq; attr.srq = srq; @@ -1554,7 +1556,7 @@ static int create_qp(struct ib_uverbs_file *file, resp.response_length = offsetof(typeof(resp), response_length) + sizeof(resp.response_length); - ret = cb(file, &resp, ucore); + ret = cb(attrs, &resp, ucore); if (ret) goto err_cb; @@ -1598,7 +1600,7 @@ err_put: return ret; } -static int ib_uverbs_create_qp_cb(struct ib_uverbs_file *file, +static int ib_uverbs_create_qp_cb(struct uverbs_attr_bundle *attrs, struct ib_uverbs_ex_create_qp_resp *resp, struct ib_udata *ucore) { @@ -1608,7 +1610,7 @@ static int ib_uverbs_create_qp_cb(struct ib_uverbs_file *file, return 0; } -static ssize_t ib_uverbs_create_qp(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_create_qp(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -1647,7 +1649,7 @@ static ssize_t ib_uverbs_create_qp(struct ib_uverbs_file *file, cmd_ex.qp_type = cmd.qp_type; cmd_ex.is_srq = cmd.is_srq; - err = create_qp(file, &ucore, &uhw, &cmd_ex, + err = create_qp(attrs, &ucore, &uhw, &cmd_ex, offsetof(typeof(cmd_ex), is_srq) + sizeof(cmd.is_srq), ib_uverbs_create_qp_cb, NULL); @@ -1658,7 +1660,7 @@ static ssize_t ib_uverbs_create_qp(struct ib_uverbs_file *file, return in_len; } -static int ib_uverbs_ex_create_qp_cb(struct ib_uverbs_file *file, +static int ib_uverbs_ex_create_qp_cb(struct uverbs_attr_bundle *attrs, struct ib_uverbs_ex_create_qp_resp *resp, struct ib_udata *ucore) { @@ -1668,7 +1670,7 @@ static int ib_uverbs_ex_create_qp_cb(struct ib_uverbs_file *file, return 0; } -static int ib_uverbs_ex_create_qp(struct ib_uverbs_file *file, +static int ib_uverbs_ex_create_qp(struct uverbs_attr_bundle *attrs, struct ib_udata *ucore, struct ib_udata *uhw) { struct ib_uverbs_ex_create_qp_resp resp; @@ -1693,7 +1695,7 @@ static int ib_uverbs_ex_create_qp(struct ib_uverbs_file *file, sizeof(resp.response_length))) return -ENOSPC; - err = create_qp(file, ucore, uhw, &cmd, + err = create_qp(attrs, ucore, uhw, &cmd, min(ucore->inlen, sizeof(cmd)), ib_uverbs_ex_create_qp_cb, NULL); @@ -1703,7 +1705,7 @@ static int ib_uverbs_ex_create_qp(struct ib_uverbs_file *file, return 0; } -static ssize_t ib_uverbs_open_qp(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_open_qp(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -1729,12 +1731,12 @@ static ssize_t ib_uverbs_open_qp(struct ib_uverbs_file *file, in_len - sizeof(cmd) - sizeof(struct ib_uverbs_cmd_hdr), out_len - sizeof(resp)); - obj = (struct ib_uqp_object *)uobj_alloc(UVERBS_OBJECT_QP, file, + obj = (struct ib_uqp_object *)uobj_alloc(UVERBS_OBJECT_QP, attrs, &ib_dev); if (IS_ERR(obj)) return PTR_ERR(obj); - xrcd_uobj = uobj_get_read(UVERBS_OBJECT_XRCD, cmd.pd_handle, file); + xrcd_uobj = uobj_get_read(UVERBS_OBJECT_XRCD, cmd.pd_handle, attrs); if (IS_ERR(xrcd_uobj)) { ret = -EINVAL; goto err_put; @@ -1747,7 +1749,7 @@ static ssize_t ib_uverbs_open_qp(struct ib_uverbs_file *file, } attr.event_handler = ib_uverbs_qp_event_handler; - attr.qp_context = file; + attr.qp_context = attrs->ufile; attr.qp_num = cmd.qpn; attr.qp_type = cmd.qp_type; @@ -1811,7 +1813,7 @@ static void copy_ah_attr_to_uverbs(struct ib_uverbs_qp_dest *uverb_attr, uverb_attr->port_num = rdma_ah_get_port_num(rdma_attr); } -static ssize_t ib_uverbs_query_qp(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_query_qp(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -1832,7 +1834,7 @@ static ssize_t ib_uverbs_query_qp(struct ib_uverbs_file *file, goto out; } - qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd.qp_handle, file); + qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd.qp_handle, attrs); if (!qp) { ret = -EINVAL; goto out; @@ -1926,7 +1928,7 @@ static void copy_ah_attr_from_uverbs(struct ib_device *dev, rdma_ah_set_make_grd(rdma_attr, false); } -static int modify_qp(struct ib_uverbs_file *file, +static int modify_qp(struct uverbs_attr_bundle *attrs, struct ib_uverbs_ex_modify_qp *cmd, struct ib_udata *udata) { struct ib_qp_attr *attr; @@ -1937,7 +1939,8 @@ static int modify_qp(struct ib_uverbs_file *file, if (!attr) return -ENOMEM; - qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd->base.qp_handle, file); + qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd->base.qp_handle, + attrs); if (!qp) { ret = -EINVAL; goto out; @@ -2084,7 +2087,7 @@ out: return ret; } -static ssize_t ib_uverbs_modify_qp(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_modify_qp(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -2103,14 +2106,14 @@ static ssize_t ib_uverbs_modify_qp(struct ib_uverbs_file *file, in_len - sizeof(cmd.base) - sizeof(struct ib_uverbs_cmd_hdr), out_len); - ret = modify_qp(file, &cmd, &udata); + ret = modify_qp(attrs, &cmd, &udata); if (ret) return ret; return in_len; } -static int ib_uverbs_ex_modify_qp(struct ib_uverbs_file *file, +static int ib_uverbs_ex_modify_qp(struct uverbs_attr_bundle *attrs, struct ib_udata *ucore, struct ib_udata *uhw) { struct ib_uverbs_ex_modify_qp cmd = {}; @@ -2139,12 +2142,12 @@ static int ib_uverbs_ex_modify_qp(struct ib_uverbs_file *file, return -EOPNOTSUPP; } - ret = modify_qp(file, &cmd, uhw); + ret = modify_qp(attrs, &cmd, uhw); return ret; } -static ssize_t ib_uverbs_destroy_qp(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_destroy_qp(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -2156,7 +2159,7 @@ static ssize_t ib_uverbs_destroy_qp(struct ib_uverbs_file *file, if (copy_from_user(&cmd, buf, sizeof cmd)) return -EFAULT; - uobj = uobj_get_destroy(UVERBS_OBJECT_QP, cmd.qp_handle, file); + uobj = uobj_get_destroy(UVERBS_OBJECT_QP, cmd.qp_handle, attrs); if (IS_ERR(uobj)) return PTR_ERR(uobj); @@ -2182,7 +2185,7 @@ static void *alloc_wr(size_t wr_size, __u32 num_sge) num_sge * sizeof (struct ib_sge), GFP_KERNEL); } -static ssize_t ib_uverbs_post_send(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_post_send(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -2211,7 +2214,7 @@ static ssize_t ib_uverbs_post_send(struct ib_uverbs_file *file, if (!user_wr) return -ENOMEM; - qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd.qp_handle, file); + qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd.qp_handle, attrs); if (!qp) goto out; @@ -2248,7 +2251,7 @@ static ssize_t ib_uverbs_post_send(struct ib_uverbs_file *file, } ud->ah = uobj_get_obj_read(ah, UVERBS_OBJECT_AH, - user_wr->wr.ud.ah, file); + user_wr->wr.ud.ah, attrs); if (!ud->ah) { kfree(ud); ret = -EINVAL; @@ -2462,7 +2465,7 @@ err: return ERR_PTR(ret); } -static ssize_t ib_uverbs_post_recv(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_post_recv(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -2482,7 +2485,7 @@ static ssize_t ib_uverbs_post_recv(struct ib_uverbs_file *file, if (IS_ERR(wr)) return PTR_ERR(wr); - qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd.qp_handle, file); + qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd.qp_handle, attrs); if (!qp) goto out; @@ -2511,7 +2514,7 @@ out: return ret ? ret : in_len; } -static ssize_t ib_uverbs_post_srq_recv(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_post_srq_recv(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -2531,7 +2534,7 @@ static ssize_t ib_uverbs_post_srq_recv(struct ib_uverbs_file *file, if (IS_ERR(wr)) return PTR_ERR(wr); - srq = uobj_get_obj_read(srq, UVERBS_OBJECT_SRQ, cmd.srq_handle, file); + srq = uobj_get_obj_read(srq, UVERBS_OBJECT_SRQ, cmd.srq_handle, attrs); if (!srq) goto out; @@ -2560,7 +2563,7 @@ out: return ret ? ret : in_len; } -static ssize_t ib_uverbs_create_ah(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_create_ah(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -2585,7 +2588,7 @@ static ssize_t ib_uverbs_create_ah(struct ib_uverbs_file *file, in_len - sizeof(cmd) - sizeof(struct ib_uverbs_cmd_hdr), out_len - sizeof(resp)); - uobj = uobj_alloc(UVERBS_OBJECT_AH, file, &ib_dev); + uobj = uobj_alloc(UVERBS_OBJECT_AH, attrs, &ib_dev); if (IS_ERR(uobj)) return PTR_ERR(uobj); @@ -2594,7 +2597,7 @@ static ssize_t ib_uverbs_create_ah(struct ib_uverbs_file *file, goto err; } - pd = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd.pd_handle, file); + pd = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd.pd_handle, attrs); if (!pd) { ret = -EINVAL; goto err; @@ -2649,7 +2652,7 @@ err: return ret; } -static ssize_t ib_uverbs_destroy_ah(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_destroy_ah(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -2658,11 +2661,11 @@ static ssize_t ib_uverbs_destroy_ah(struct ib_uverbs_file *file, if (copy_from_user(&cmd, buf, sizeof cmd)) return -EFAULT; - return uobj_perform_destroy(UVERBS_OBJECT_AH, cmd.ah_handle, file, + return uobj_perform_destroy(UVERBS_OBJECT_AH, cmd.ah_handle, attrs, in_len); } -static ssize_t ib_uverbs_attach_mcast(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_attach_mcast(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -2675,7 +2678,7 @@ static ssize_t ib_uverbs_attach_mcast(struct ib_uverbs_file *file, if (copy_from_user(&cmd, buf, sizeof cmd)) return -EFAULT; - qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd.qp_handle, file); + qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd.qp_handle, attrs); if (!qp) return -EINVAL; @@ -2711,7 +2714,7 @@ out_put: return ret ? ret : in_len; } -static ssize_t ib_uverbs_detach_mcast(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_detach_mcast(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -2725,7 +2728,7 @@ static ssize_t ib_uverbs_detach_mcast(struct ib_uverbs_file *file, if (copy_from_user(&cmd, buf, sizeof cmd)) return -EFAULT; - qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd.qp_handle, file); + qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd.qp_handle, attrs); if (!qp) return -EINVAL; @@ -2830,7 +2833,7 @@ void flow_resources_add(struct ib_uflow_resources *uflow_res, } EXPORT_SYMBOL(flow_resources_add); -static int kern_spec_to_ib_spec_action(struct ib_uverbs_file *ufile, +static int kern_spec_to_ib_spec_action(const struct uverbs_attr_bundle *attrs, struct ib_uverbs_flow_spec *kern_spec, union ib_flow_spec *ib_spec, struct ib_uflow_resources *uflow_res) @@ -2859,7 +2862,7 @@ static int kern_spec_to_ib_spec_action(struct ib_uverbs_file *ufile, ib_spec->action.act = uobj_get_obj_read(flow_action, UVERBS_OBJECT_FLOW_ACTION, kern_spec->action.handle, - ufile); + attrs); if (!ib_spec->action.act) return -EINVAL; ib_spec->action.size = @@ -2877,7 +2880,7 @@ static int kern_spec_to_ib_spec_action(struct ib_uverbs_file *ufile, uobj_get_obj_read(counters, UVERBS_OBJECT_COUNTERS, kern_spec->flow_count.handle, - ufile); + attrs); if (!ib_spec->flow_count.counters) return -EINVAL; ib_spec->flow_count.size = @@ -3058,7 +3061,7 @@ static int kern_spec_to_ib_spec_filter(struct ib_uverbs_flow_spec *kern_spec, kern_filter_sz, ib_spec); } -static int kern_spec_to_ib_spec(struct ib_uverbs_file *ufile, +static int kern_spec_to_ib_spec(struct uverbs_attr_bundle *attrs, struct ib_uverbs_flow_spec *kern_spec, union ib_flow_spec *ib_spec, struct ib_uflow_resources *uflow_res) @@ -3067,13 +3070,13 @@ static int kern_spec_to_ib_spec(struct ib_uverbs_file *ufile, return -EINVAL; if (kern_spec->type >= IB_FLOW_SPEC_ACTION_TAG) - return kern_spec_to_ib_spec_action(ufile, kern_spec, ib_spec, + return kern_spec_to_ib_spec_action(attrs, kern_spec, ib_spec, uflow_res); else return kern_spec_to_ib_spec_filter(kern_spec, ib_spec); } -static int ib_uverbs_ex_create_wq(struct ib_uverbs_file *file, +static int ib_uverbs_ex_create_wq(struct uverbs_attr_bundle *attrs, struct ib_udata *ucore, struct ib_udata *uhw) { struct ib_uverbs_ex_create_wq cmd = {}; @@ -3109,18 +3112,18 @@ static int ib_uverbs_ex_create_wq(struct ib_uverbs_file *file, if (cmd.comp_mask) return -EOPNOTSUPP; - obj = (struct ib_uwq_object *)uobj_alloc(UVERBS_OBJECT_WQ, file, + obj = (struct ib_uwq_object *)uobj_alloc(UVERBS_OBJECT_WQ, attrs, &ib_dev); if (IS_ERR(obj)) return PTR_ERR(obj); - pd = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd.pd_handle, file); + pd = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd.pd_handle, attrs); if (!pd) { err = -EINVAL; goto err_uobj; } - cq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, cmd.cq_handle, file); + cq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, cmd.cq_handle, attrs); if (!cq) { err = -EINVAL; goto err_put_pd; @@ -3129,7 +3132,7 @@ static int ib_uverbs_ex_create_wq(struct ib_uverbs_file *file, wq_init_attr.cq = cq; wq_init_attr.max_sge = cmd.max_sge; wq_init_attr.max_wr = cmd.max_wr; - wq_init_attr.wq_context = file; + wq_init_attr.wq_context = attrs->ufile; wq_init_attr.wq_type = cmd.wq_type; wq_init_attr.event_handler = ib_uverbs_wq_event_handler; if (ucore->inlen >= (offsetof(typeof(cmd), create_flags) + @@ -3184,7 +3187,7 @@ err_uobj: return err; } -static int ib_uverbs_ex_destroy_wq(struct ib_uverbs_file *file, +static int ib_uverbs_ex_destroy_wq(struct uverbs_attr_bundle *attrs, struct ib_udata *ucore, struct ib_udata *uhw) { struct ib_uverbs_ex_destroy_wq cmd = {}; @@ -3217,7 +3220,7 @@ static int ib_uverbs_ex_destroy_wq(struct ib_uverbs_file *file, return -EOPNOTSUPP; resp.response_length = required_resp_len; - uobj = uobj_get_destroy(UVERBS_OBJECT_WQ, cmd.wq_handle, file); + uobj = uobj_get_destroy(UVERBS_OBJECT_WQ, cmd.wq_handle, attrs); if (IS_ERR(uobj)) return PTR_ERR(uobj); @@ -3229,7 +3232,7 @@ static int ib_uverbs_ex_destroy_wq(struct ib_uverbs_file *file, return ib_copy_to_udata(ucore, &resp, resp.response_length); } -static int ib_uverbs_ex_modify_wq(struct ib_uverbs_file *file, +static int ib_uverbs_ex_modify_wq(struct uverbs_attr_bundle *attrs, struct ib_udata *ucore, struct ib_udata *uhw) { struct ib_uverbs_ex_modify_wq cmd = {}; @@ -3257,7 +3260,7 @@ static int ib_uverbs_ex_modify_wq(struct ib_uverbs_file *file, if (cmd.attr_mask > (IB_WQ_STATE | IB_WQ_CUR_STATE | IB_WQ_FLAGS)) return -EINVAL; - wq = uobj_get_obj_read(wq, UVERBS_OBJECT_WQ, cmd.wq_handle, file); + wq = uobj_get_obj_read(wq, UVERBS_OBJECT_WQ, cmd.wq_handle, attrs); if (!wq) return -EINVAL; @@ -3272,7 +3275,7 @@ static int ib_uverbs_ex_modify_wq(struct ib_uverbs_file *file, return ret; } -static int ib_uverbs_ex_create_rwq_ind_table(struct ib_uverbs_file *file, +static int ib_uverbs_ex_create_rwq_ind_table(struct uverbs_attr_bundle *attrs, struct ib_udata *ucore, struct ib_udata *uhw) { @@ -3347,7 +3350,7 @@ static int ib_uverbs_ex_create_rwq_ind_table(struct ib_uverbs_file *file, for (num_read_wqs = 0; num_read_wqs < num_wq_handles; num_read_wqs++) { wq = uobj_get_obj_read(wq, UVERBS_OBJECT_WQ, - wqs_handles[num_read_wqs], file); + wqs_handles[num_read_wqs], attrs); if (!wq) { err = -EINVAL; goto put_wqs; @@ -3356,7 +3359,7 @@ static int ib_uverbs_ex_create_rwq_ind_table(struct ib_uverbs_file *file, wqs[num_read_wqs] = wq; } - uobj = uobj_alloc(UVERBS_OBJECT_RWQ_IND_TBL, file, &ib_dev); + uobj = uobj_alloc(UVERBS_OBJECT_RWQ_IND_TBL, attrs, &ib_dev); if (IS_ERR(uobj)) { err = PTR_ERR(uobj); goto put_wqs; @@ -3411,7 +3414,7 @@ err_free: return err; } -static int ib_uverbs_ex_destroy_rwq_ind_table(struct ib_uverbs_file *file, +static int ib_uverbs_ex_destroy_rwq_ind_table(struct uverbs_attr_bundle *attrs, struct ib_udata *ucore, struct ib_udata *uhw) { @@ -3437,10 +3440,10 @@ static int ib_uverbs_ex_destroy_rwq_ind_table(struct ib_uverbs_file *file, return -EOPNOTSUPP; return uobj_perform_destroy(UVERBS_OBJECT_RWQ_IND_TBL, - cmd.ind_tbl_handle, file, 0); + cmd.ind_tbl_handle, attrs, 0); } -static int ib_uverbs_ex_create_flow(struct ib_uverbs_file *file, +static int ib_uverbs_ex_create_flow(struct uverbs_attr_bundle *attrs, struct ib_udata *ucore, struct ib_udata *uhw) { @@ -3512,13 +3515,13 @@ static int ib_uverbs_ex_create_flow(struct ib_uverbs_file *file, kern_flow_attr = &cmd.flow_attr; } - uobj = uobj_alloc(UVERBS_OBJECT_FLOW, file, &ib_dev); + uobj = uobj_alloc(UVERBS_OBJECT_FLOW, attrs, &ib_dev); if (IS_ERR(uobj)) { err = PTR_ERR(uobj); goto err_free_attr; } - qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd.qp_handle, file); + qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd.qp_handle, attrs); if (!qp) { err = -EINVAL; goto err_uobj; @@ -3555,7 +3558,7 @@ static int ib_uverbs_ex_create_flow(struct ib_uverbs_file *file, cmd.flow_attr.size >= kern_spec->size; i++) { err = kern_spec_to_ib_spec( - file, (struct ib_uverbs_flow_spec *)kern_spec, + attrs, (struct ib_uverbs_flow_spec *)kern_spec, ib_spec, uflow_res); if (err) goto err_free; @@ -3613,7 +3616,7 @@ err_free_attr: return err; } -static int ib_uverbs_ex_destroy_flow(struct ib_uverbs_file *file, +static int ib_uverbs_ex_destroy_flow(struct uverbs_attr_bundle *attrs, struct ib_udata *ucore, struct ib_udata *uhw) { @@ -3630,11 +3633,11 @@ static int ib_uverbs_ex_destroy_flow(struct ib_uverbs_file *file, if (cmd.comp_mask) return -EINVAL; - return uobj_perform_destroy(UVERBS_OBJECT_FLOW, cmd.flow_handle, file, + return uobj_perform_destroy(UVERBS_OBJECT_FLOW, cmd.flow_handle, attrs, 0); } -static int __uverbs_create_xsrq(struct ib_uverbs_file *file, +static int __uverbs_create_xsrq(struct uverbs_attr_bundle *attrs, struct ib_uverbs_create_xsrq *cmd, struct ib_udata *udata) { @@ -3647,7 +3650,7 @@ static int __uverbs_create_xsrq(struct ib_uverbs_file *file, int ret; struct ib_device *ib_dev; - obj = (struct ib_usrq_object *)uobj_alloc(UVERBS_OBJECT_SRQ, file, + obj = (struct ib_usrq_object *)uobj_alloc(UVERBS_OBJECT_SRQ, attrs, &ib_dev); if (IS_ERR(obj)) return PTR_ERR(obj); @@ -3657,7 +3660,7 @@ static int __uverbs_create_xsrq(struct ib_uverbs_file *file, if (cmd->srq_type == IB_SRQT_XRC) { xrcd_uobj = uobj_get_read(UVERBS_OBJECT_XRCD, cmd->xrcd_handle, - file); + attrs); if (IS_ERR(xrcd_uobj)) { ret = -EINVAL; goto err; @@ -3675,21 +3678,21 @@ static int __uverbs_create_xsrq(struct ib_uverbs_file *file, if (ib_srq_has_cq(cmd->srq_type)) { attr.ext.cq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, - cmd->cq_handle, file); + cmd->cq_handle, attrs); if (!attr.ext.cq) { ret = -EINVAL; goto err_put_xrcd; } } - pd = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd->pd_handle, file); + pd = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd->pd_handle, attrs); if (!pd) { ret = -EINVAL; goto err_put_cq; } attr.event_handler = ib_uverbs_srq_event_handler; - attr.srq_context = file; + attr.srq_context = attrs->ufile; attr.srq_type = cmd->srq_type; attr.attr.max_wr = cmd->max_wr; attr.attr.max_sge = cmd->max_sge; @@ -3770,7 +3773,7 @@ err: return ret; } -static ssize_t ib_uverbs_create_srq(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_create_srq(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -3800,14 +3803,14 @@ static ssize_t ib_uverbs_create_srq(struct ib_uverbs_file *file, in_len - sizeof(cmd) - sizeof(struct ib_uverbs_cmd_hdr), out_len - sizeof(resp)); - ret = __uverbs_create_xsrq(file, &xcmd, &udata); + ret = __uverbs_create_xsrq(attrs, &xcmd, &udata); if (ret) return ret; return in_len; } -static ssize_t ib_uverbs_create_xsrq(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_create_xsrq(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -3827,14 +3830,14 @@ static ssize_t ib_uverbs_create_xsrq(struct ib_uverbs_file *file, in_len - sizeof(cmd) - sizeof(struct ib_uverbs_cmd_hdr), out_len - sizeof(resp)); - ret = __uverbs_create_xsrq(file, &cmd, &udata); + ret = __uverbs_create_xsrq(attrs, &cmd, &udata); if (ret) return ret; return in_len; } -static ssize_t ib_uverbs_modify_srq(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_modify_srq(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -3850,7 +3853,7 @@ static ssize_t ib_uverbs_modify_srq(struct ib_uverbs_file *file, ib_uverbs_init_udata(&udata, buf + sizeof cmd, NULL, in_len - sizeof cmd, out_len); - srq = uobj_get_obj_read(srq, UVERBS_OBJECT_SRQ, cmd.srq_handle, file); + srq = uobj_get_obj_read(srq, UVERBS_OBJECT_SRQ, cmd.srq_handle, attrs); if (!srq) return -EINVAL; @@ -3864,7 +3867,7 @@ static ssize_t ib_uverbs_modify_srq(struct ib_uverbs_file *file, return ret ? ret : in_len; } -static ssize_t ib_uverbs_query_srq(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_query_srq(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -3880,7 +3883,7 @@ static ssize_t ib_uverbs_query_srq(struct ib_uverbs_file *file, if (copy_from_user(&cmd, buf, sizeof cmd)) return -EFAULT; - srq = uobj_get_obj_read(srq, UVERBS_OBJECT_SRQ, cmd.srq_handle, file); + srq = uobj_get_obj_read(srq, UVERBS_OBJECT_SRQ, cmd.srq_handle, attrs); if (!srq) return -EINVAL; @@ -3903,7 +3906,7 @@ static ssize_t ib_uverbs_query_srq(struct ib_uverbs_file *file, return in_len; } -static ssize_t ib_uverbs_destroy_srq(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_destroy_srq(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { @@ -3915,7 +3918,7 @@ static ssize_t ib_uverbs_destroy_srq(struct ib_uverbs_file *file, if (copy_from_user(&cmd, buf, sizeof cmd)) return -EFAULT; - uobj = uobj_get_destroy(UVERBS_OBJECT_SRQ, cmd.srq_handle, file); + uobj = uobj_get_destroy(UVERBS_OBJECT_SRQ, cmd.srq_handle, attrs); if (IS_ERR(uobj)) return PTR_ERR(uobj); @@ -3931,7 +3934,7 @@ static ssize_t ib_uverbs_destroy_srq(struct ib_uverbs_file *file, return in_len; } -static int ib_uverbs_ex_query_device(struct ib_uverbs_file *file, +static int ib_uverbs_ex_query_device(struct uverbs_attr_bundle *attrs, struct ib_udata *ucore, struct ib_udata *uhw) { @@ -3942,7 +3945,7 @@ static int ib_uverbs_ex_query_device(struct ib_uverbs_file *file, struct ib_device *ib_dev; int err; - ucontext = ib_uverbs_get_ucontext(file); + ucontext = ib_uverbs_get_ucontext(attrs); if (IS_ERR(ucontext)) return PTR_ERR(ucontext); ib_dev = ucontext->device; @@ -4055,7 +4058,7 @@ end: return err; } -static int ib_uverbs_ex_modify_cq(struct ib_uverbs_file *file, +static int ib_uverbs_ex_modify_cq(struct uverbs_attr_bundle *attrs, struct ib_udata *ucore, struct ib_udata *uhw) { struct ib_uverbs_ex_modify_cq cmd = {}; @@ -4084,7 +4087,7 @@ static int ib_uverbs_ex_modify_cq(struct ib_uverbs_file *file, if (cmd.attr_mask > IB_CQ_MODERATE) return -EOPNOTSUPP; - cq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, cmd.cq_handle, file); + cq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, cmd.cq_handle, attrs); if (!cq) return -EINVAL; diff --git a/drivers/infiniband/core/uverbs_main.c b/drivers/infiniband/core/uverbs_main.c index 6a1284c8a07d..c5d245133b51 100644 --- a/drivers/infiniband/core/uverbs_main.c +++ b/drivers/infiniband/core/uverbs_main.c @@ -81,7 +81,7 @@ static void ib_uverbs_remove_one(struct ib_device *device, void *client_data); * Must be called with the ufile->device->disassociate_srcu held, and the lock * must be held until use of the ucontext is finished. */ -struct ib_ucontext *ib_uverbs_get_ucontext(struct ib_uverbs_file *ufile) +struct ib_ucontext *ib_uverbs_get_ucontext_file(struct ib_uverbs_file *ufile) { /* * We do not hold the hw_destroy_rwsem lock for this flow, instead @@ -99,7 +99,7 @@ struct ib_ucontext *ib_uverbs_get_ucontext(struct ib_uverbs_file *ufile) return ucontext; } -EXPORT_SYMBOL(ib_uverbs_get_ucontext); +EXPORT_SYMBOL(ib_uverbs_get_ucontext_file); int uverbs_dealloc_mw(struct ib_mw *mw) { @@ -632,6 +632,7 @@ static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf, struct uverbs_api *uapi = file->device->uapi; struct ib_uverbs_ex_cmd_hdr ex_hdr; struct ib_uverbs_cmd_hdr hdr; + struct uverbs_attr_bundle bundle; int srcu_key; ssize_t ret; @@ -666,8 +667,9 @@ static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf, buf += sizeof(hdr); + bundle.ufile = file; if (!method_elm->is_ex) { - ret = method_elm->handler(file, buf, hdr.in_words * 4, + ret = method_elm->handler(&bundle, buf, hdr.in_words * 4, hdr.out_words * 4); } else { struct ib_udata ucore; @@ -685,7 +687,7 @@ static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf, ex_hdr.provider_in_words * 8, ex_hdr.provider_out_words * 8); - ret = method_elm->handler_ex(file, &ucore, &uhw); + ret = method_elm->handler_ex(&bundle, &ucore, &uhw); ret = (ret) ? : count; } @@ -701,7 +703,7 @@ static int ib_uverbs_mmap(struct file *filp, struct vm_area_struct *vma) int srcu_key; srcu_key = srcu_read_lock(&file->device->disassociate_srcu); - ucontext = ib_uverbs_get_ucontext(file); + ucontext = ib_uverbs_get_ucontext_file(file); if (IS_ERR(ucontext)) { ret = PTR_ERR(ucontext); goto out; diff --git a/drivers/infiniband/core/uverbs_uapi.c b/drivers/infiniband/core/uverbs_uapi.c index a92adbe2b034..ad72d726efa9 100644 --- a/drivers/infiniband/core/uverbs_uapi.c +++ b/drivers/infiniband/core/uverbs_uapi.c @@ -8,14 +8,14 @@ #include "rdma_core.h" #include "uverbs.h" -static ssize_t ib_uverbs_notsupp(struct ib_uverbs_file *file, +static ssize_t ib_uverbs_notsupp(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len) { return -EOPNOTSUPP; } -static int ib_uverbs_ex_notsupp(struct ib_uverbs_file *file, +static int ib_uverbs_ex_notsupp(struct uverbs_attr_bundle *attrs, struct ib_udata *ucore, struct ib_udata *uhw) { return -EOPNOTSUPP; diff --git a/drivers/infiniband/hw/mlx5/devx.c b/drivers/infiniband/hw/mlx5/devx.c index ee8db8d9e919..8ccb655db1ec 100644 --- a/drivers/infiniband/hw/mlx5/devx.c +++ b/drivers/infiniband/hw/mlx5/devx.c @@ -42,7 +42,7 @@ struct devx_umem_reg_cmd { static struct mlx5_ib_ucontext *devx_ufile2uctx(struct ib_uverbs_file *file) { - return to_mucontext(ib_uverbs_get_ucontext(file)); + return to_mucontext(ib_uverbs_get_ucontext_file(file)); } int mlx5_ib_devx_create(struct mlx5_ib_dev *dev) diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h index ee8a6008e222..0b7269870571 100644 --- a/include/rdma/ib_verbs.h +++ b/include/rdma/ib_verbs.h @@ -4207,7 +4207,7 @@ ib_get_vector_affinity(struct ib_device *device, int comp_vector) */ void rdma_roce_rescan_device(struct ib_device *ibdev); -struct ib_ucontext *ib_uverbs_get_ucontext(struct ib_uverbs_file *ufile); +struct ib_ucontext *ib_uverbs_get_ucontext_file(struct ib_uverbs_file *ufile); int uverbs_destroy_def_handler(struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs); diff --git a/include/rdma/uverbs_ioctl.h b/include/rdma/uverbs_ioctl.h index 963dd6b70cde..f744691cd9ae 100644 --- a/include/rdma/uverbs_ioctl.h +++ b/include/rdma/uverbs_ioctl.h @@ -368,10 +368,10 @@ struct uapi_definition { union { bool (*func_is_supported)(struct ib_device *device); - ssize_t (*func_write)(struct ib_uverbs_file *file, + ssize_t (*func_write)(struct uverbs_attr_bundle *attrs, const char __user *buf, int in_len, int out_len); - int (*func_write_ex)(struct ib_uverbs_file *file, + int (*func_write_ex)(struct uverbs_attr_bundle *attrs, struct ib_udata *ucore, struct ib_udata *uhw); const struct uapi_definition *chain; @@ -806,6 +806,12 @@ static inline int _uverbs_copy_from_or_zero(void *to, #define uverbs_copy_from_or_zero(to, attrs_bundle, idx) \ _uverbs_copy_from_or_zero(to, attrs_bundle, idx, sizeof(*to)) +static inline struct ib_ucontext * +ib_uverbs_get_ucontext(const struct uverbs_attr_bundle *attrs) +{ + return ib_uverbs_get_ucontext_file(attrs->ufile); +} + #if IS_ENABLED(CONFIG_INFINIBAND_USER_ACCESS) int uverbs_get_flags64(u64 *to, const struct uverbs_attr_bundle *attrs_bundle, size_t idx, u64 allowed_bits); diff --git a/include/rdma/uverbs_std_types.h b/include/rdma/uverbs_std_types.h index e3154f8d3cda..a6358c36bba0 100644 --- a/include/rdma/uverbs_std_types.h +++ b/include/rdma/uverbs_std_types.h @@ -45,15 +45,15 @@ */ #define _uobj_check_id(_id) ((_id) * typecheck(u32, _id)) -#define uobj_get_type(_ufile, _object) \ - uapi_get_object((_ufile)->device->uapi, _object) +#define uobj_get_type(_attrs, _object) \ + uapi_get_object((_attrs)->ufile->device->uapi, _object) -#define uobj_get_read(_type, _id, _ufile) \ - rdma_lookup_get_uobject(uobj_get_type(_ufile, _type), _ufile, \ +#define uobj_get_read(_type, _id, _attrs) \ + rdma_lookup_get_uobject(uobj_get_type(_attrs, _type), (_attrs)->ufile, \ _uobj_check_id(_id), UVERBS_LOOKUP_READ) -#define ufd_get_read(_type, _fdnum, _ufile) \ - rdma_lookup_get_uobject(uobj_get_type(_ufile, _type), _ufile, \ +#define ufd_get_read(_type, _fdnum, _attrs) \ + rdma_lookup_get_uobject(uobj_get_type(_attrs, _type), (_attrs)->ufile, \ (_fdnum)*typecheck(s32, _fdnum), \ UVERBS_LOOKUP_READ) @@ -63,26 +63,28 @@ static inline void *_uobj_get_obj_read(struct ib_uobject *uobj) return NULL; return uobj->object; } -#define uobj_get_obj_read(_object, _type, _id, _ufile) \ +#define uobj_get_obj_read(_object, _type, _id, _attrs) \ ((struct ib_##_object *)_uobj_get_obj_read( \ - uobj_get_read(_type, _id, _ufile))) + uobj_get_read(_type, _id, _attrs))) -#define uobj_get_write(_type, _id, _ufile) \ - rdma_lookup_get_uobject(uobj_get_type(_ufile, _type), _ufile, \ +#define uobj_get_write(_type, _id, _attrs) \ + rdma_lookup_get_uobject(uobj_get_type(_attrs, _type), (_attrs)->ufile, \ _uobj_check_id(_id), UVERBS_LOOKUP_WRITE) int __uobj_perform_destroy(const struct uverbs_api_object *obj, u32 id, - struct ib_uverbs_file *ufile, int success_res); -#define uobj_perform_destroy(_type, _id, _ufile, _success_res) \ - __uobj_perform_destroy(uobj_get_type(_ufile, _type), \ - _uobj_check_id(_id), _ufile, _success_res) + const struct uverbs_attr_bundle *attrs, + int success_res); +#define uobj_perform_destroy(_type, _id, _attrs, _success_res) \ + __uobj_perform_destroy(uobj_get_type(_attrs, _type), \ + _uobj_check_id(_id), _attrs, _success_res) struct ib_uobject *__uobj_get_destroy(const struct uverbs_api_object *obj, - u32 id, struct ib_uverbs_file *ufile); + u32 id, + const struct uverbs_attr_bundle *attrs); -#define uobj_get_destroy(_type, _id, _ufile) \ - __uobj_get_destroy(uobj_get_type(_ufile, _type), _uobj_check_id(_id), \ - _ufile) +#define uobj_get_destroy(_type, _id, _attrs) \ + __uobj_get_destroy(uobj_get_type(_attrs, _type), _uobj_check_id(_id), \ + _attrs) static inline void uobj_put_destroy(struct ib_uobject *uobj) { @@ -118,18 +120,18 @@ static inline void uobj_alloc_abort(struct ib_uobject *uobj) } static inline struct ib_uobject * -__uobj_alloc(const struct uverbs_api_object *obj, struct ib_uverbs_file *ufile, - struct ib_device **ib_dev) +__uobj_alloc(const struct uverbs_api_object *obj, + struct uverbs_attr_bundle *attrs, struct ib_device **ib_dev) { - struct ib_uobject *uobj = rdma_alloc_begin_uobject(obj, ufile); + struct ib_uobject *uobj = rdma_alloc_begin_uobject(obj, attrs->ufile); if (!IS_ERR(uobj)) *ib_dev = uobj->context->device; return uobj; } -#define uobj_alloc(_type, _ufile, _ib_dev) \ - __uobj_alloc(uobj_get_type(_ufile, _type), _ufile, _ib_dev) +#define uobj_alloc(_type, _attrs, _ib_dev) \ + __uobj_alloc(uobj_get_type(_attrs, _type), _attrs, _ib_dev) static inline void uverbs_flow_action_fill_action(struct ib_flow_action *action, struct ib_uobject *uobj, -- cgit v1.2.3-59-g8ed1b From 15a1b4becba886176aa1426604548c34904fd054 Mon Sep 17 00:00:00 2001 From: Jason Gunthorpe Date: Sun, 25 Nov 2018 20:51:15 +0200 Subject: RDMA/uverbs: Do not pass ib_uverbs_file to ioctl methods The uverbs_attr_bundle already contains this pointer, and most methods don't actually need it. Get rid of the redundant function argument. Signed-off-by: Jason Gunthorpe Signed-off-by: Leon Romanovsky --- drivers/infiniband/core/rdma_core.h | 3 +-- drivers/infiniband/core/uverbs_ioctl.c | 7 +++--- drivers/infiniband/core/uverbs_std_types.c | 3 +-- .../infiniband/core/uverbs_std_types_counters.c | 4 ++-- drivers/infiniband/core/uverbs_std_types_cq.c | 6 +++--- drivers/infiniband/core/uverbs_std_types_dm.c | 5 ++--- .../infiniband/core/uverbs_std_types_flow_action.c | 10 ++++----- drivers/infiniband/core/uverbs_std_types_mr.c | 2 +- drivers/infiniband/hw/mlx5/devx.c | 25 +++++++++++----------- drivers/infiniband/hw/mlx5/flow.c | 6 ++---- include/rdma/ib_verbs.h | 4 ++-- include/rdma/uverbs_ioctl.h | 3 +-- 12 files changed, 35 insertions(+), 43 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/core/rdma_core.h b/drivers/infiniband/core/rdma_core.h index 59edd7514095..ce323439a0f9 100644 --- a/drivers/infiniband/core/rdma_core.h +++ b/drivers/infiniband/core/rdma_core.h @@ -125,8 +125,7 @@ struct uverbs_api_object { }; struct uverbs_api_ioctl_method { - int (__rcu *handler)(struct ib_uverbs_file *ufile, - struct uverbs_attr_bundle *ctx); + int(__rcu *handler)(struct uverbs_attr_bundle *attrs); DECLARE_BITMAP(attr_mandatory, UVERBS_API_ATTR_BKEY_LEN); u16 bundle_size; u8 use_stack:1; diff --git a/drivers/infiniband/core/uverbs_ioctl.c b/drivers/infiniband/core/uverbs_ioctl.c index b0e493e8d860..5c04149ce775 100644 --- a/drivers/infiniband/core/uverbs_ioctl.c +++ b/drivers/infiniband/core/uverbs_ioctl.c @@ -404,8 +404,7 @@ static int uverbs_set_attr(struct bundle_priv *pbundle, static int ib_uverbs_run_method(struct bundle_priv *pbundle, unsigned int num_attrs) { - int (*handler)(struct ib_uverbs_file *ufile, - struct uverbs_attr_bundle *ctx); + int (*handler)(struct uverbs_attr_bundle *attrs); size_t uattrs_size = array_size(sizeof(*pbundle->uattrs), num_attrs); unsigned int destroy_bkey = pbundle->method_elm->destroy_bkey; unsigned int i; @@ -445,10 +444,10 @@ static int ib_uverbs_run_method(struct bundle_priv *pbundle, return ret; __clear_bit(destroy_bkey, pbundle->uobj_finalize); - ret = handler(pbundle->bundle.ufile, &pbundle->bundle); + ret = handler(&pbundle->bundle); uobj_put_destroy(destroy_attr->uobject); } else { - ret = handler(pbundle->bundle.ufile, &pbundle->bundle); + ret = handler(&pbundle->bundle); } /* diff --git a/drivers/infiniband/core/uverbs_std_types.c b/drivers/infiniband/core/uverbs_std_types.c index 9e1abeaed57d..063aff9e7a04 100644 --- a/drivers/infiniband/core/uverbs_std_types.c +++ b/drivers/infiniband/core/uverbs_std_types.c @@ -210,8 +210,7 @@ static int uverbs_hot_unplug_completion_event_file(struct ib_uobject *uobj, return 0; }; -int uverbs_destroy_def_handler(struct ib_uverbs_file *file, - struct uverbs_attr_bundle *attrs) +int uverbs_destroy_def_handler(struct uverbs_attr_bundle *attrs) { return 0; } diff --git a/drivers/infiniband/core/uverbs_std_types_counters.c b/drivers/infiniband/core/uverbs_std_types_counters.c index c4a78abffc10..8835bad5c6dd 100644 --- a/drivers/infiniband/core/uverbs_std_types_counters.c +++ b/drivers/infiniband/core/uverbs_std_types_counters.c @@ -48,7 +48,7 @@ static int uverbs_free_counters(struct ib_uobject *uobject, } static int UVERBS_HANDLER(UVERBS_METHOD_COUNTERS_CREATE)( - struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs) + struct uverbs_attr_bundle *attrs) { struct ib_uobject *uobj = uverbs_attr_get_uobject( attrs, UVERBS_ATTR_CREATE_COUNTERS_HANDLE); @@ -82,7 +82,7 @@ err_create_counters: } static int UVERBS_HANDLER(UVERBS_METHOD_COUNTERS_READ)( - struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs) + struct uverbs_attr_bundle *attrs) { struct ib_counters_read_attr read_attr = {}; const struct uverbs_attr *uattr; diff --git a/drivers/infiniband/core/uverbs_std_types_cq.c b/drivers/infiniband/core/uverbs_std_types_cq.c index f94fd877867b..859d1e862034 100644 --- a/drivers/infiniband/core/uverbs_std_types_cq.c +++ b/drivers/infiniband/core/uverbs_std_types_cq.c @@ -58,7 +58,7 @@ static int uverbs_free_cq(struct ib_uobject *uobject, } static int UVERBS_HANDLER(UVERBS_METHOD_CQ_CREATE)( - struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs) + struct uverbs_attr_bundle *attrs) { struct ib_ucq_object *obj = container_of( uverbs_attr_get_uobject(attrs, UVERBS_ATTR_CREATE_CQ_HANDLE), @@ -101,7 +101,7 @@ static int UVERBS_HANDLER(UVERBS_METHOD_CQ_CREATE)( uverbs_uobject_get(ev_file_uobj); } - if (attr.comp_vector >= file->device->num_comp_vectors) { + if (attr.comp_vector >= attrs->ufile->device->num_comp_vectors) { ret = -EINVAL; goto err_event_file; } @@ -173,7 +173,7 @@ DECLARE_UVERBS_NAMED_METHOD( UVERBS_ATTR_UHW()); static int UVERBS_HANDLER(UVERBS_METHOD_CQ_DESTROY)( - struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs) + struct uverbs_attr_bundle *attrs) { struct ib_uobject *uobj = uverbs_attr_get_uobject(attrs, UVERBS_ATTR_DESTROY_CQ_HANDLE); diff --git a/drivers/infiniband/core/uverbs_std_types_dm.c b/drivers/infiniband/core/uverbs_std_types_dm.c index c40353331d90..658261b8f08e 100644 --- a/drivers/infiniband/core/uverbs_std_types_dm.c +++ b/drivers/infiniband/core/uverbs_std_types_dm.c @@ -46,9 +46,8 @@ static int uverbs_free_dm(struct ib_uobject *uobject, return dm->device->dealloc_dm(dm); } -static int -UVERBS_HANDLER(UVERBS_METHOD_DM_ALLOC)(struct ib_uverbs_file *file, - struct uverbs_attr_bundle *attrs) +static int UVERBS_HANDLER(UVERBS_METHOD_DM_ALLOC)( + struct uverbs_attr_bundle *attrs) { struct ib_dm_alloc_attr attr = {}; struct ib_uobject *uobj = diff --git a/drivers/infiniband/core/uverbs_std_types_flow_action.c b/drivers/infiniband/core/uverbs_std_types_flow_action.c index 24ac332fc3e0..e4d01fb5335d 100644 --- a/drivers/infiniband/core/uverbs_std_types_flow_action.c +++ b/drivers/infiniband/core/uverbs_std_types_flow_action.c @@ -223,7 +223,6 @@ struct ib_flow_action_esp_attr { #define ESP_LAST_SUPPORTED_FLAG IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ESN_NEW_WINDOW static int parse_flow_action_esp(struct ib_device *ib_dev, - struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs, struct ib_flow_action_esp_attr *esp_attr, bool is_modify) @@ -305,7 +304,7 @@ static int parse_flow_action_esp(struct ib_device *ib_dev, } static int UVERBS_HANDLER(UVERBS_METHOD_FLOW_ACTION_ESP_CREATE)( - struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs) + struct uverbs_attr_bundle *attrs) { struct ib_uobject *uobj = uverbs_attr_get_uobject( attrs, UVERBS_ATTR_CREATE_FLOW_ACTION_ESP_HANDLE); @@ -317,7 +316,7 @@ static int UVERBS_HANDLER(UVERBS_METHOD_FLOW_ACTION_ESP_CREATE)( if (!ib_dev->create_flow_action_esp) return -EOPNOTSUPP; - ret = parse_flow_action_esp(ib_dev, file, attrs, &esp_attr, false); + ret = parse_flow_action_esp(ib_dev, attrs, &esp_attr, false); if (ret) return ret; @@ -333,7 +332,7 @@ static int UVERBS_HANDLER(UVERBS_METHOD_FLOW_ACTION_ESP_CREATE)( } static int UVERBS_HANDLER(UVERBS_METHOD_FLOW_ACTION_ESP_MODIFY)( - struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs) + struct uverbs_attr_bundle *attrs) { struct ib_uobject *uobj = uverbs_attr_get_uobject( attrs, UVERBS_ATTR_MODIFY_FLOW_ACTION_ESP_HANDLE); @@ -344,8 +343,7 @@ static int UVERBS_HANDLER(UVERBS_METHOD_FLOW_ACTION_ESP_MODIFY)( if (!action->device->modify_flow_action_esp) return -EOPNOTSUPP; - ret = parse_flow_action_esp(action->device, file, attrs, &esp_attr, - true); + ret = parse_flow_action_esp(action->device, attrs, &esp_attr, true); if (ret) return ret; diff --git a/drivers/infiniband/core/uverbs_std_types_mr.c b/drivers/infiniband/core/uverbs_std_types_mr.c index c400fd382b0f..70ea48cfc047 100644 --- a/drivers/infiniband/core/uverbs_std_types_mr.c +++ b/drivers/infiniband/core/uverbs_std_types_mr.c @@ -40,7 +40,7 @@ static int uverbs_free_mr(struct ib_uobject *uobject, } static int UVERBS_HANDLER(UVERBS_METHOD_DM_MR_REG)( - struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs) + struct uverbs_attr_bundle *attrs) { struct ib_dm_mr_attr attr = {}; struct ib_uobject *uobj = diff --git a/drivers/infiniband/hw/mlx5/devx.c b/drivers/infiniband/hw/mlx5/devx.c index 8ccb655db1ec..9dfc8593ad43 100644 --- a/drivers/infiniband/hw/mlx5/devx.c +++ b/drivers/infiniband/hw/mlx5/devx.c @@ -40,9 +40,10 @@ struct devx_umem_reg_cmd { u32 out[MLX5_ST_SZ_DW(general_obj_out_cmd_hdr)]; }; -static struct mlx5_ib_ucontext *devx_ufile2uctx(struct ib_uverbs_file *file) +static struct mlx5_ib_ucontext * +devx_ufile2uctx(const struct uverbs_attr_bundle *attrs) { - return to_mucontext(ib_uverbs_get_ucontext_file(file)); + return to_mucontext(ib_uverbs_get_ucontext(attrs)); } int mlx5_ib_devx_create(struct mlx5_ib_dev *dev) @@ -603,7 +604,7 @@ static bool devx_is_general_cmd(void *in) } static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_QUERY_EQN)( - struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs) + struct uverbs_attr_bundle *attrs) { struct mlx5_ib_ucontext *c; struct mlx5_ib_dev *dev; @@ -616,7 +617,7 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_QUERY_EQN)( MLX5_IB_ATTR_DEVX_QUERY_EQN_USER_VEC)) return -EFAULT; - c = devx_ufile2uctx(file); + c = devx_ufile2uctx(attrs); if (IS_ERR(c)) return PTR_ERR(c); dev = to_mdev(c->ibucontext.device); @@ -653,14 +654,14 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_QUERY_EQN)( * queue or arm its CQ for event generation), no further harm is expected. */ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_QUERY_UAR)( - struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs) + struct uverbs_attr_bundle *attrs) { struct mlx5_ib_ucontext *c; struct mlx5_ib_dev *dev; u32 user_idx; s32 dev_idx; - c = devx_ufile2uctx(file); + c = devx_ufile2uctx(attrs); if (IS_ERR(c)) return PTR_ERR(c); dev = to_mdev(c->ibucontext.device); @@ -681,7 +682,7 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_QUERY_UAR)( } static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OTHER)( - struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs) + struct uverbs_attr_bundle *attrs) { struct mlx5_ib_ucontext *c; struct mlx5_ib_dev *dev; @@ -693,7 +694,7 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OTHER)( int err; int uid; - c = devx_ufile2uctx(file); + c = devx_ufile2uctx(attrs); if (IS_ERR(c)) return PTR_ERR(c); dev = to_mdev(c->ibucontext.device); @@ -908,7 +909,7 @@ static int devx_obj_cleanup(struct ib_uobject *uobject, } static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_CREATE)( - struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs) + struct uverbs_attr_bundle *attrs) { void *cmd_in = uverbs_attr_get_alloced_ptr(attrs, MLX5_IB_ATTR_DEVX_OBJ_CREATE_CMD_IN); int cmd_out_len = uverbs_attr_get_len(attrs, @@ -970,7 +971,7 @@ obj_free: } static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_MODIFY)( - struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs) + struct uverbs_attr_bundle *attrs) { void *cmd_in = uverbs_attr_get_alloced_ptr(attrs, MLX5_IB_ATTR_DEVX_OBJ_MODIFY_CMD_IN); int cmd_out_len = uverbs_attr_get_len(attrs, @@ -1011,7 +1012,7 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_MODIFY)( } static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_QUERY)( - struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs) + struct uverbs_attr_bundle *attrs) { void *cmd_in = uverbs_attr_get_alloced_ptr(attrs, MLX5_IB_ATTR_DEVX_OBJ_QUERY_CMD_IN); int cmd_out_len = uverbs_attr_get_len(attrs, @@ -1125,7 +1126,7 @@ static void devx_umem_reg_cmd_build(struct mlx5_ib_dev *dev, } static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_UMEM_REG)( - struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs) + struct uverbs_attr_bundle *attrs) { struct devx_umem_reg_cmd cmd; struct devx_umem *obj; diff --git a/drivers/infiniband/hw/mlx5/flow.c b/drivers/infiniband/hw/mlx5/flow.c index 497ef18771f5..0144ac5fde1a 100644 --- a/drivers/infiniband/hw/mlx5/flow.c +++ b/drivers/infiniband/hw/mlx5/flow.c @@ -60,7 +60,7 @@ static const struct uverbs_attr_spec mlx5_ib_flow_type[] = { #define MLX5_IB_CREATE_FLOW_MAX_FLOW_ACTIONS 2 static int UVERBS_HANDLER(MLX5_IB_METHOD_CREATE_FLOW)( - struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs) + struct uverbs_attr_bundle *attrs) { struct mlx5_flow_act flow_act = {.flow_tag = MLX5_FS_DEFAULT_FLOW_TAG}; struct mlx5_ib_flow_handler *flow_handler; @@ -194,7 +194,7 @@ static int flow_matcher_cleanup(struct ib_uobject *uobject, } static int UVERBS_HANDLER(MLX5_IB_METHOD_FLOW_MATCHER_CREATE)( - struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs) + struct uverbs_attr_bundle *attrs) { struct ib_uobject *uobj = uverbs_attr_get_uobject( attrs, MLX5_IB_ATTR_FLOW_MATCHER_CREATE_HANDLE); @@ -313,7 +313,6 @@ static bool mlx5_ib_modify_header_supported(struct mlx5_ib_dev *dev) } static int UVERBS_HANDLER(MLX5_IB_METHOD_FLOW_ACTION_CREATE_MODIFY_HEADER)( - struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs) { struct ib_uobject *uobj = uverbs_attr_get_uobject( @@ -435,7 +434,6 @@ static int mlx5_ib_flow_action_create_packet_reformat_ctx( } static int UVERBS_HANDLER(MLX5_IB_METHOD_FLOW_ACTION_CREATE_PACKET_REFORMAT)( - struct ib_uverbs_file *file, struct uverbs_attr_bundle *attrs) { struct ib_uobject *uobj = uverbs_attr_get_uobject(attrs, diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h index 0b7269870571..92633c15125b 100644 --- a/include/rdma/ib_verbs.h +++ b/include/rdma/ib_verbs.h @@ -4209,8 +4209,8 @@ void rdma_roce_rescan_device(struct ib_device *ibdev); struct ib_ucontext *ib_uverbs_get_ucontext_file(struct ib_uverbs_file *ufile); -int uverbs_destroy_def_handler(struct ib_uverbs_file *file, - struct uverbs_attr_bundle *attrs); + +int uverbs_destroy_def_handler(struct uverbs_attr_bundle *attrs); struct net_device *rdma_alloc_netdev(struct ib_device *device, u8 port_num, enum rdma_netdev_t type, const char *name, diff --git a/include/rdma/uverbs_ioctl.h b/include/rdma/uverbs_ioctl.h index 2a59d6ad9f62..e0b1a77b1c2c 100644 --- a/include/rdma/uverbs_ioctl.h +++ b/include/rdma/uverbs_ioctl.h @@ -327,8 +327,7 @@ struct uverbs_method_def { u32 flags; size_t num_attrs; const struct uverbs_attr_def * const (*attrs)[]; - int (*handler)(struct ib_uverbs_file *ufile, - struct uverbs_attr_bundle *ctx); + int (*handler)(struct uverbs_attr_bundle *attrs); }; struct uverbs_object_def { -- cgit v1.2.3-59-g8ed1b From 901018f29e65c2c442b09b9913473dd8cf8d3cee Mon Sep 17 00:00:00 2001 From: Colin Ian King Date: Wed, 28 Nov 2018 15:11:16 +0000 Subject: RDMA/drivers: Fix spelling mistake "initalize" -> "initialize" Fix spelling mistake in usnic_err error message Signed-off-by: Colin Ian King Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/usnic/usnic_ib_main.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/usnic/usnic_ib_main.c b/drivers/infiniband/hw/usnic/usnic_ib_main.c index 73bd00f8d2c8..413fa5732e2b 100644 --- a/drivers/infiniband/hw/usnic/usnic_ib_main.c +++ b/drivers/infiniband/hw/usnic/usnic_ib_main.c @@ -649,7 +649,7 @@ static int __init usnic_ib_init(void) err = usnic_uiom_init(DRV_NAME); if (err) { - usnic_err("Unable to initalize umem with err %d\n", err); + usnic_err("Unable to initialize umem with err %d\n", err); return err; } -- cgit v1.2.3-59-g8ed1b From d12c416dd131df990036b9b9136d98e5682bb1a7 Mon Sep 17 00:00:00 2001 From: Colin Ian King Date: Thu, 29 Nov 2018 10:42:13 +0000 Subject: IB/usnic: fix spelling mistake "miniumum" -> "minimum" There is a spelling mistake in a usnic_err error message, fix it. Signed-off-by: Colin Ian King Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/usnic/usnic_ib_qp_grp.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/usnic/usnic_ib_qp_grp.c b/drivers/infiniband/hw/usnic/usnic_ib_qp_grp.c index bf5136533d49..0cdb156e165e 100644 --- a/drivers/infiniband/hw/usnic/usnic_ib_qp_grp.c +++ b/drivers/infiniband/hw/usnic/usnic_ib_qp_grp.c @@ -681,7 +681,7 @@ usnic_ib_qp_grp_create(struct usnic_fwd_dev *ufdev, struct usnic_ib_vf *vf, err = usnic_vnic_res_spec_satisfied(&min_transport_spec[transport], res_spec); if (err) { - usnic_err("Spec does not meet miniumum req for transport %d\n", + usnic_err("Spec does not meet minimum req for transport %d\n", transport); log_spec(res_spec); return ERR_PTR(err); -- cgit v1.2.3-59-g8ed1b From 67810e8c3c01ed80b5118391a8939ff24f08acd3 Mon Sep 17 00:00:00 2001 From: Leon Romanovsky Date: Thu, 29 Nov 2018 14:15:28 +0200 Subject: RDMA/qib: Remove all occurrences of BUG_ON() QIB driver was added in 2010 with many BUG_ON(), most of them were cleaned out after years of development and usages. It looks like that it is safe now to remove rest of BUG_ONs. Signed-off-by: Leon Romanovsky Acked-by: Dennis Dalessandro Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/qib/qib_iba6120.c | 1 - drivers/infiniband/hw/qib/qib_iba7220.c | 2 -- drivers/infiniband/hw/qib/qib_iba7322.c | 3 --- drivers/infiniband/hw/qib/qib_init.c | 1 - drivers/infiniband/hw/qib/qib_sdma.c | 1 - drivers/infiniband/hw/qib/qib_ud.c | 1 - drivers/infiniband/hw/qib/qib_user_sdma.c | 1 - drivers/infiniband/hw/qib/qib_verbs.c | 3 --- 8 files changed, 13 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/qib/qib_iba6120.c b/drivers/infiniband/hw/qib/qib_iba6120.c index fb1ff59f40bd..cdbf707fa267 100644 --- a/drivers/infiniband/hw/qib/qib_iba6120.c +++ b/drivers/infiniband/hw/qib/qib_iba6120.c @@ -3237,7 +3237,6 @@ static int init_6120_variables(struct qib_devdata *dd) /* we always allocate at least 2048 bytes for eager buffers */ ret = ib_mtu_enum_to_int(qib_ibmtu); dd->rcvegrbufsize = ret != -1 ? max(ret, 2048) : QIB_DEFAULT_MTU; - BUG_ON(!is_power_of_2(dd->rcvegrbufsize)); dd->rcvegrbufsize_shift = ilog2(dd->rcvegrbufsize); qib_6120_tidtemplate(dd); diff --git a/drivers/infiniband/hw/qib/qib_iba7220.c b/drivers/infiniband/hw/qib/qib_iba7220.c index 163a57a88742..9fde45538f6e 100644 --- a/drivers/infiniband/hw/qib/qib_iba7220.c +++ b/drivers/infiniband/hw/qib/qib_iba7220.c @@ -4043,7 +4043,6 @@ static int qib_init_7220_variables(struct qib_devdata *dd) /* we always allocate at least 2048 bytes for eager buffers */ ret = ib_mtu_enum_to_int(qib_ibmtu); dd->rcvegrbufsize = ret != -1 ? max(ret, 2048) : QIB_DEFAULT_MTU; - BUG_ON(!is_power_of_2(dd->rcvegrbufsize)); dd->rcvegrbufsize_shift = ilog2(dd->rcvegrbufsize); qib_7220_tidtemplate(dd); @@ -4252,7 +4251,6 @@ static int init_sdma_7220_regs(struct qib_pportdata *ppd) unsigned word = i / 64; unsigned bit = i & 63; - BUG_ON(word >= 3); senddmabufmask[word] |= 1ULL << bit; } qib_write_kreg(dd, kr_senddmabufmask0, senddmabufmask[0]); diff --git a/drivers/infiniband/hw/qib/qib_iba7322.c b/drivers/infiniband/hw/qib/qib_iba7322.c index bf5e222eed8e..17d6b24b3473 100644 --- a/drivers/infiniband/hw/qib/qib_iba7322.c +++ b/drivers/infiniband/hw/qib/qib_iba7322.c @@ -1382,7 +1382,6 @@ static void err_decode(char *msg, size_t len, u64 errs, *msg++ = ','; len--; } - BUG_ON(!msp->sz); /* msp->sz counts the nul */ took = min_t(size_t, msp->sz - (size_t)1, len); memcpy(msg, msp->msg, took); @@ -6599,7 +6598,6 @@ static int qib_init_7322_variables(struct qib_devdata *dd) /* we always allocate at least 2048 bytes for eager buffers */ dd->rcvegrbufsize = max(mtu, 2048); - BUG_ON(!is_power_of_2(dd->rcvegrbufsize)); dd->rcvegrbufsize_shift = ilog2(dd->rcvegrbufsize); qib_7322_tidtemplate(dd); @@ -6904,7 +6902,6 @@ static int init_sdma_7322_regs(struct qib_pportdata *ppd) unsigned word = erstbuf / BITS_PER_LONG; unsigned bit = erstbuf & (BITS_PER_LONG - 1); - BUG_ON(word >= 3); senddmabufmask[word] |= 1ULL << bit; } qib_write_kreg_port(ppd, krp_senddmabufmask0, senddmabufmask[0]); diff --git a/drivers/infiniband/hw/qib/qib_init.c b/drivers/infiniband/hw/qib/qib_init.c index d7cdc77d6306..9fd69903ca57 100644 --- a/drivers/infiniband/hw/qib/qib_init.c +++ b/drivers/infiniband/hw/qib/qib_init.c @@ -209,7 +209,6 @@ struct qib_ctxtdata *qib_create_ctxtdata(struct qib_pportdata *ppd, u32 ctxt, rcd->rcvegrbuf_chunks = (rcd->rcvegrcnt + rcd->rcvegrbufs_perchunk - 1) / rcd->rcvegrbufs_perchunk; - BUG_ON(!is_power_of_2(rcd->rcvegrbufs_perchunk)); rcd->rcvegrbufs_perchunk_shift = ilog2(rcd->rcvegrbufs_perchunk); } diff --git a/drivers/infiniband/hw/qib/qib_sdma.c b/drivers/infiniband/hw/qib/qib_sdma.c index 757d4c9d713d..0b6ca424c11d 100644 --- a/drivers/infiniband/hw/qib/qib_sdma.c +++ b/drivers/infiniband/hw/qib/qib_sdma.c @@ -572,7 +572,6 @@ retry: len = sge->length; if (len > sge->sge_length) len = sge->sge_length; - BUG_ON(len == 0); dw = (len + 3) >> 2; addr = dma_map_single(&ppd->dd->pcidev->dev, sge->vaddr, dw << 2, DMA_TO_DEVICE); diff --git a/drivers/infiniband/hw/qib/qib_ud.c b/drivers/infiniband/hw/qib/qib_ud.c index 4d4c31ea4e2d..868da0ece7ba 100644 --- a/drivers/infiniband/hw/qib/qib_ud.c +++ b/drivers/infiniband/hw/qib/qib_ud.c @@ -178,7 +178,6 @@ static void qib_ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe) len = length; if (len > sge->sge_length) len = sge->sge_length; - BUG_ON(len == 0); rvt_copy_sge(qp, &qp->r_sge, sge->vaddr, len, true, false); sge->vaddr += len; sge->length -= len; diff --git a/drivers/infiniband/hw/qib/qib_user_sdma.c b/drivers/infiniband/hw/qib/qib_user_sdma.c index 926f3c8eba69..31c523b2a9f5 100644 --- a/drivers/infiniband/hw/qib/qib_user_sdma.c +++ b/drivers/infiniband/hw/qib/qib_user_sdma.c @@ -237,7 +237,6 @@ qib_user_sdma_queue_create(struct device *dev, int unit, int ctxt, int sctxt) ret = qib_user_sdma_rb_insert(&qib_user_sdma_rb_root, sdma_rb_node); - BUG_ON(ret == 0); } pq->sdma_rb_node = sdma_rb_node; diff --git a/drivers/infiniband/hw/qib/qib_verbs.c b/drivers/infiniband/hw/qib/qib_verbs.c index 4b0f5761a646..8914abdd7584 100644 --- a/drivers/infiniband/hw/qib/qib_verbs.c +++ b/drivers/infiniband/hw/qib/qib_verbs.c @@ -150,7 +150,6 @@ static u32 qib_count_sge(struct rvt_sge_state *ss, u32 length) len = length; if (len > sge.sge_length) len = sge.sge_length; - BUG_ON(len == 0); if (((long) sge.vaddr & (sizeof(u32) - 1)) || (len != length && (len & (sizeof(u32) - 1)))) { ndesc = 0; @@ -193,7 +192,6 @@ static void qib_copy_from_sge(void *data, struct rvt_sge_state *ss, u32 length) len = length; if (len > sge->sge_length) len = sge->sge_length; - BUG_ON(len == 0); memcpy(data, sge->vaddr, len); sge->vaddr += len; sge->length -= len; @@ -449,7 +447,6 @@ static void copy_io(u32 __iomem *piobuf, struct rvt_sge_state *ss, len = length; if (len > ss->sge.sge_length) len = ss->sge.sge_length; - BUG_ON(len == 0); /* If the source address is not aligned, try to align it. */ off = (unsigned long)ss->sge.vaddr & (sizeof(u32) - 1); if (off) { -- cgit v1.2.3-59-g8ed1b From bfc5d839184f53cc16d551873f9254f2d4d493be Mon Sep 17 00:00:00 2001 From: Mark Bloch Date: Tue, 20 Nov 2018 20:31:08 +0200 Subject: RDMA/mlx5: Attach a DEVX counter via raw flow creation Allow a user to attach a DEVX counter via mlx5 raw flow creation. In order to attach a counter we introduce a new attribute: MLX5_IB_ATTR_CREATE_FLOW_ARR_COUNTERS_DEVX A counter can be attached to multiple flow steering rules. Signed-off-by: Mark Bloch Reviewed-by: Yishai Hadas Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/devx.c | 15 +++++++++++++++ drivers/infiniband/hw/mlx5/flow.c | 17 ++++++++++++++++- drivers/infiniband/hw/mlx5/main.c | 29 ++++++++++++++++++++--------- drivers/infiniband/hw/mlx5/mlx5_ib.h | 6 ++++-- include/uapi/rdma/mlx5_user_ioctl_cmds.h | 1 + 5 files changed, 56 insertions(+), 12 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/devx.c b/drivers/infiniband/hw/mlx5/devx.c index 9dfc8593ad43..0aa2ee732eaa 100644 --- a/drivers/infiniband/hw/mlx5/devx.c +++ b/drivers/infiniband/hw/mlx5/devx.c @@ -107,6 +107,21 @@ bool mlx5_ib_devx_is_flow_dest(void *obj, int *dest_id, int *dest_type) } } +bool mlx5_ib_devx_is_flow_counter(void *obj, u32 *counter_id) +{ + struct devx_obj *devx_obj = obj; + u16 opcode = MLX5_GET(general_obj_in_cmd_hdr, devx_obj->dinbox, opcode); + + if (opcode == MLX5_CMD_OP_DEALLOC_FLOW_COUNTER) { + *counter_id = MLX5_GET(dealloc_flow_counter_in, + devx_obj->dinbox, + flow_counter_id); + return true; + } + + return false; +} + /* * As the obj_id in the firmware is not globally unique the object type * must be considered upon checking for a valid object id. diff --git a/drivers/infiniband/hw/mlx5/flow.c b/drivers/infiniband/hw/mlx5/flow.c index 0144ac5fde1a..793afa404442 100644 --- a/drivers/infiniband/hw/mlx5/flow.c +++ b/drivers/infiniband/hw/mlx5/flow.c @@ -77,6 +77,7 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_CREATE_FLOW)( uverbs_attr_get_uobject(attrs, MLX5_IB_ATTR_CREATE_FLOW_HANDLE); struct mlx5_ib_dev *dev = to_mdev(uobj->context->device); int len, ret, i; + u32 counter_id = 0; if (!capable(CAP_NET_RAW)) return -EPERM; @@ -128,6 +129,15 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_CREATE_FLOW)( dest_type = MLX5_FLOW_DESTINATION_TYPE_PORT; } + len = uverbs_attr_get_uobjs_arr(attrs, + MLX5_IB_ATTR_CREATE_FLOW_ARR_COUNTERS_DEVX, &arr_flow_actions); + if (len) { + devx_obj = arr_flow_actions[0]->object; + + if (!mlx5_ib_devx_is_flow_counter(devx_obj, &counter_id)) + return -EINVAL; + flow_act.action |= MLX5_FLOW_CONTEXT_ACTION_COUNT; + } if (dev->rep) return -ENOTSUPP; @@ -164,6 +174,7 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_CREATE_FLOW)( } flow_handler = mlx5_ib_raw_fs_rule_add(dev, fs_matcher, &flow_act, + counter_id, cmd_in, inlen, dest_id, dest_type); if (IS_ERR(flow_handler)) { @@ -524,7 +535,11 @@ DECLARE_UVERBS_NAMED_METHOD( UA_OPTIONAL), UVERBS_ATTR_PTR_IN(MLX5_IB_ATTR_CREATE_FLOW_TAG, UVERBS_ATTR_TYPE(u32), - UA_OPTIONAL)); + UA_OPTIONAL), + UVERBS_ATTR_IDRS_ARR(MLX5_IB_ATTR_CREATE_FLOW_ARR_COUNTERS_DEVX, + MLX5_IB_OBJECT_DEVX_OBJ, + UVERBS_ACCESS_READ, 1, 1, + UA_OPTIONAL)); DECLARE_UVERBS_NAMED_METHOD_DESTROY( MLX5_IB_METHOD_DESTROY_FLOW, diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index 0707ede7dcdd..5236169c42d0 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -3713,7 +3713,8 @@ _create_raw_flow_rule(struct mlx5_ib_dev *dev, struct mlx5_flow_destination *dst, struct mlx5_ib_flow_matcher *fs_matcher, struct mlx5_flow_act *flow_act, - void *cmd_in, int inlen) + void *cmd_in, int inlen, + int dst_num) { struct mlx5_ib_flow_handler *handler; struct mlx5_flow_spec *spec; @@ -3735,7 +3736,7 @@ _create_raw_flow_rule(struct mlx5_ib_dev *dev, spec->match_criteria_enable = fs_matcher->match_criteria_enable; handler->rule = mlx5_add_flow_rules(ft, spec, - flow_act, dst, 1); + flow_act, dst, dst_num); if (IS_ERR(handler->rule)) { err = PTR_ERR(handler->rule); @@ -3798,12 +3799,14 @@ struct mlx5_ib_flow_handler * mlx5_ib_raw_fs_rule_add(struct mlx5_ib_dev *dev, struct mlx5_ib_flow_matcher *fs_matcher, struct mlx5_flow_act *flow_act, + u32 counter_id, void *cmd_in, int inlen, int dest_id, int dest_type) { struct mlx5_flow_destination *dst; struct mlx5_ib_flow_prio *ft_prio; struct mlx5_ib_flow_handler *handler; + int dst_num = 0; bool mcast; int err; @@ -3813,7 +3816,7 @@ mlx5_ib_raw_fs_rule_add(struct mlx5_ib_dev *dev, if (fs_matcher->priority > MLX5_IB_FLOW_LAST_PRIO) return ERR_PTR(-ENOMEM); - dst = kzalloc(sizeof(*dst), GFP_KERNEL); + dst = kzalloc(sizeof(*dst) * 2, GFP_KERNEL); if (!dst) return ERR_PTR(-ENOMEM); @@ -3827,20 +3830,28 @@ mlx5_ib_raw_fs_rule_add(struct mlx5_ib_dev *dev, } if (dest_type == MLX5_FLOW_DESTINATION_TYPE_TIR) { - dst->type = dest_type; - dst->tir_num = dest_id; + dst[dst_num].type = dest_type; + dst[dst_num].tir_num = dest_id; flow_act->action |= MLX5_FLOW_CONTEXT_ACTION_FWD_DEST; } else if (dest_type == MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE) { - dst->type = MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE_NUM; - dst->ft_num = dest_id; + dst[dst_num].type = MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE_NUM; + dst[dst_num].ft_num = dest_id; flow_act->action |= MLX5_FLOW_CONTEXT_ACTION_FWD_DEST; } else { - dst->type = MLX5_FLOW_DESTINATION_TYPE_PORT; + dst[dst_num].type = MLX5_FLOW_DESTINATION_TYPE_PORT; flow_act->action |= MLX5_FLOW_CONTEXT_ACTION_ALLOW; } + dst_num++; + + if (flow_act->action & MLX5_FLOW_CONTEXT_ACTION_COUNT) { + dst[dst_num].type = MLX5_FLOW_DESTINATION_TYPE_COUNTER; + dst[dst_num].counter_id = counter_id; + dst_num++; + } + handler = _create_raw_flow_rule(dev, ft_prio, dst, fs_matcher, flow_act, - cmd_in, inlen); + cmd_in, inlen, dst_num); if (IS_ERR(handler)) { err = PTR_ERR(handler); diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index 9b434246d4e3..a2b35a1a5031 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -1266,9 +1266,11 @@ extern const struct uapi_definition mlx5_ib_devx_defs[]; extern const struct uapi_definition mlx5_ib_flow_defs[]; struct mlx5_ib_flow_handler *mlx5_ib_raw_fs_rule_add( struct mlx5_ib_dev *dev, struct mlx5_ib_flow_matcher *fs_matcher, - struct mlx5_flow_act *flow_act, void *cmd_in, int inlen, - int dest_id, int dest_type); + struct mlx5_flow_act *flow_act, u32 counter_id, + void *cmd_in, int inlen, int dest_id, int dest_type); bool mlx5_ib_devx_is_flow_dest(void *obj, int *dest_id, int *dest_type); +bool mlx5_ib_devx_is_flow_counter(void *obj, u32 *counter_id); +int mlx5_ib_get_flow_trees(const struct uverbs_object_tree_def **root); void mlx5_ib_destroy_flow_action_raw(struct mlx5_ib_flow_action *maction); #else static inline int diff --git a/include/uapi/rdma/mlx5_user_ioctl_cmds.h b/include/uapi/rdma/mlx5_user_ioctl_cmds.h index 408e220034de..b8d121d457f1 100644 --- a/include/uapi/rdma/mlx5_user_ioctl_cmds.h +++ b/include/uapi/rdma/mlx5_user_ioctl_cmds.h @@ -158,6 +158,7 @@ enum mlx5_ib_create_flow_attrs { MLX5_IB_ATTR_CREATE_FLOW_MATCHER, MLX5_IB_ATTR_CREATE_FLOW_ARR_FLOW_ACTIONS, MLX5_IB_ATTR_CREATE_FLOW_TAG, + MLX5_IB_ATTR_CREATE_FLOW_ARR_COUNTERS_DEVX, }; enum mlx5_ib_destoy_flow_attrs { -- cgit v1.2.3-59-g8ed1b From 20e5a59b2e64a01f8e0957727887564a4d004970 Mon Sep 17 00:00:00 2001 From: Guy Levi Date: Mon, 26 Nov 2018 08:15:39 +0200 Subject: IB/mlx5: Use fragmented SRQ's buffer for in-kernel users The current implementation of create SRQ requires contiguous memory, such a requirement is problematic once the memory is fragmented or the system is low in memory, it causes failures in dma_zalloc_coherent(). This patch takes the advantage of the new mlx5_core API which allocates a fragmented buffer, and makes the SRQ creation much more resilient to memory fragmentation. Data-path code was adapted to the fact that WQEs can cross buffers. Signed-off-by: Guy Levi Reviewed-by: Majd Dibbiny Reviewed-by: Yishai Hadas Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/mlx5_ib.h | 1 + drivers/infiniband/hw/mlx5/srq.c | 14 +++++++++----- 2 files changed, 10 insertions(+), 5 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index a2b35a1a5031..caed1fe76102 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -523,6 +523,7 @@ struct mlx5_ib_srq { struct mlx5_core_srq msrq; struct mlx5_frag_buf buf; struct mlx5_db db; + struct mlx5_frag_buf_ctrl fbc; u64 *wrid; /* protect SRQ hanlding */ diff --git a/drivers/infiniband/hw/mlx5/srq.c b/drivers/infiniband/hw/mlx5/srq.c index d012e7dbcc38..cc8273d0f66a 100644 --- a/drivers/infiniband/hw/mlx5/srq.c +++ b/drivers/infiniband/hw/mlx5/srq.c @@ -44,7 +44,7 @@ static int srq_signature; static void *get_wqe(struct mlx5_ib_srq *srq, int n) { - return mlx5_buf_offset(&srq->buf, n << srq->msrq.wqe_shift); + return mlx5_frag_buf_get_wqe(&srq->fbc, n); } static void mlx5_ib_srq_event(struct mlx5_core_srq *srq, enum mlx5_event type) @@ -173,12 +173,16 @@ static int create_srq_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_srq *srq, return err; } - if (mlx5_buf_alloc(dev->mdev, buf_size, &srq->buf)) { + if (mlx5_frag_buf_alloc_node(dev->mdev, buf_size, &srq->buf, + dev->mdev->priv.numa_node)) { mlx5_ib_dbg(dev, "buf alloc failed\n"); err = -ENOMEM; goto err_db; } + mlx5_init_fbc(srq->buf.frags, srq->msrq.wqe_shift, ilog2(srq->msrq.max), + &srq->fbc); + srq->head = 0; srq->tail = srq->msrq.max - 1; srq->wqe_ctr = 0; @@ -195,7 +199,7 @@ static int create_srq_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_srq *srq, err = -ENOMEM; goto err_buf; } - mlx5_fill_page_array(&srq->buf, in->pas); + mlx5_fill_page_frag_array(&srq->buf, in->pas); srq->wrid = kvmalloc_array(srq->msrq.max, sizeof(u64), GFP_KERNEL); if (!srq->wrid) { @@ -215,7 +219,7 @@ err_in: kvfree(in->pas); err_buf: - mlx5_buf_free(dev->mdev, &srq->buf); + mlx5_frag_buf_free(dev->mdev, &srq->buf); err_db: mlx5_db_free(dev->mdev, &srq->db); @@ -232,7 +236,7 @@ static void destroy_srq_user(struct ib_pd *pd, struct mlx5_ib_srq *srq) static void destroy_srq_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_srq *srq) { kvfree(srq->wrid); - mlx5_buf_free(dev->mdev, &srq->buf); + mlx5_frag_buf_free(dev->mdev, &srq->buf); mlx5_db_free(dev->mdev, &srq->db); } -- cgit v1.2.3-59-g8ed1b From 34f4c9554d8b2a7d2deb9503e9373b598ee3279f Mon Sep 17 00:00:00 2001 From: Guy Levi Date: Mon, 26 Nov 2018 08:15:50 +0200 Subject: IB/mlx5: Use fragmented QP's buffer for in-kernel users The current implementation of create QP requires contiguous memory, such a requirement is problematic once the memory is fragmented or the system is low in memory, it causes failures in dma_zalloc_coherent(). This patch takes advantage of the new mlx5_core API which allocates a fragmented buffer. This makes the QP creation much more resilient to memory fragmentation. Data-path code was adapted to the fact that WQEs can cross buffers. We also use the opportunity to fix some cosmetic legacy coding convention errors which were in the feature scope. Signed-off-by: Guy Levi Reviewed-by: Majd Dibbiny Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/mlx5_ib.h | 3 +- drivers/infiniband/hw/mlx5/qp.c | 405 +++++++++++++++++++++-------------- 2 files changed, 241 insertions(+), 167 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index caed1fe76102..93772434b9e3 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -257,6 +257,7 @@ enum mlx5_ib_rq_flags { }; struct mlx5_ib_wq { + struct mlx5_frag_buf_ctrl fbc; u64 *wrid; u32 *wr_data; struct wr_list *w_list; @@ -275,7 +276,7 @@ struct mlx5_ib_wq { unsigned tail; u16 cur_post; u16 last_poll; - void *qend; + void *cur_edge; }; enum mlx5_ib_wq_flags { diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c index f76f21ecb958..67adc09c6fa7 100644 --- a/drivers/infiniband/hw/mlx5/qp.c +++ b/drivers/infiniband/hw/mlx5/qp.c @@ -108,21 +108,6 @@ static int is_sqp(enum ib_qp_type qp_type) return is_qp0(qp_type) || is_qp1(qp_type); } -static void *get_wqe(struct mlx5_ib_qp *qp, int offset) -{ - return mlx5_buf_offset(&qp->buf, offset); -} - -static void *get_recv_wqe(struct mlx5_ib_qp *qp, int n) -{ - return get_wqe(qp, qp->rq.offset + (n << qp->rq.wqe_shift)); -} - -void *mlx5_get_send_wqe(struct mlx5_ib_qp *qp, int n) -{ - return get_wqe(qp, qp->sq.offset + (n << MLX5_IB_SQ_STRIDE)); -} - /** * mlx5_ib_read_user_wqe() - Copy a user-space WQE to kernel space. * @@ -917,6 +902,30 @@ static void destroy_qp_user(struct mlx5_ib_dev *dev, struct ib_pd *pd, mlx5_ib_free_bfreg(dev, &context->bfregi, qp->bfregn); } +/* get_sq_edge - Get the next nearby edge. + * + * An 'edge' is defined as the first following address after the end + * of the fragment or the SQ. Accordingly, during the WQE construction + * which repetitively increases the pointer to write the next data, it + * simply should check if it gets to an edge. + * + * @sq - SQ buffer. + * @idx - Stride index in the SQ buffer. + * + * Return: + * The new edge. + */ +static void *get_sq_edge(struct mlx5_ib_wq *sq, u32 idx) +{ + void *fragment_end; + + fragment_end = mlx5_frag_buf_get_wqe + (&sq->fbc, + mlx5_frag_buf_get_idx_last_contig_stride(&sq->fbc, idx)); + + return fragment_end + MLX5_SEND_WQE_BB; +} + static int create_kernel_qp(struct mlx5_ib_dev *dev, struct ib_qp_init_attr *init_attr, struct mlx5_ib_qp *qp, @@ -955,13 +964,29 @@ static int create_kernel_qp(struct mlx5_ib_dev *dev, qp->sq.offset = qp->rq.wqe_cnt << qp->rq.wqe_shift; base->ubuffer.buf_size = err + (qp->rq.wqe_cnt << qp->rq.wqe_shift); - err = mlx5_buf_alloc(dev->mdev, base->ubuffer.buf_size, &qp->buf); + err = mlx5_frag_buf_alloc_node(dev->mdev, base->ubuffer.buf_size, + &qp->buf, dev->mdev->priv.numa_node); if (err) { mlx5_ib_dbg(dev, "err %d\n", err); return err; } - qp->sq.qend = mlx5_get_send_wqe(qp, qp->sq.wqe_cnt); + if (qp->rq.wqe_cnt) + mlx5_init_fbc(qp->buf.frags, qp->rq.wqe_shift, + ilog2(qp->rq.wqe_cnt), &qp->rq.fbc); + + if (qp->sq.wqe_cnt) { + int sq_strides_offset = (qp->sq.offset & (PAGE_SIZE - 1)) / + MLX5_SEND_WQE_BB; + mlx5_init_fbc_offset(qp->buf.frags + + (qp->sq.offset / PAGE_SIZE), + ilog2(MLX5_SEND_WQE_BB), + ilog2(qp->sq.wqe_cnt), + sq_strides_offset, &qp->sq.fbc); + + qp->sq.cur_edge = get_sq_edge(&qp->sq, 0); + } + *inlen = MLX5_ST_SZ_BYTES(create_qp_in) + MLX5_FLD_SZ_BYTES(create_qp_in, pas[0]) * qp->buf.npages; *in = kvzalloc(*inlen, GFP_KERNEL); @@ -983,8 +1008,9 @@ static int create_kernel_qp(struct mlx5_ib_dev *dev, qp->flags |= MLX5_IB_QP_SQPN_QP1; } - mlx5_fill_page_array(&qp->buf, - (__be64 *)MLX5_ADDR_OF(create_qp_in, *in, pas)); + mlx5_fill_page_frag_array(&qp->buf, + (__be64 *)MLX5_ADDR_OF(create_qp_in, + *in, pas)); err = mlx5_db_alloc(dev->mdev, &qp->db); if (err) { @@ -1024,7 +1050,7 @@ err_free: kvfree(*in); err_buf: - mlx5_buf_free(dev->mdev, &qp->buf); + mlx5_frag_buf_free(dev->mdev, &qp->buf); return err; } @@ -1036,7 +1062,7 @@ static void destroy_qp_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp) kvfree(qp->sq.wr_data); kvfree(qp->rq.wrid); mlx5_db_free(dev->mdev, &qp->db); - mlx5_buf_free(dev->mdev, &qp->buf); + mlx5_frag_buf_free(dev->mdev, &qp->buf); } static u32 get_rx_type(struct mlx5_ib_qp *qp, struct ib_qp_init_attr *attr) @@ -3476,6 +3502,8 @@ static int __mlx5_ib_modify_qp(struct ib_qp *ibqp, qp->sq.head = 0; qp->sq.tail = 0; qp->sq.cur_post = 0; + if (qp->sq.wqe_cnt) + qp->sq.cur_edge = get_sq_edge(&qp->sq, 0); qp->sq.last_poll = 0; qp->db.db[MLX5_RCV_DBR] = 0; qp->db.db[MLX5_SND_DBR] = 0; @@ -3750,6 +3778,62 @@ out: return err; } +static void _handle_post_send_edge(struct mlx5_ib_wq *sq, void **seg, + u32 wqe_sz, void **cur_edge) +{ + u32 idx; + + idx = (sq->cur_post + (wqe_sz >> 2)) & (sq->wqe_cnt - 1); + *cur_edge = get_sq_edge(sq, idx); + + *seg = mlx5_frag_buf_get_wqe(&sq->fbc, idx); +} + +/* handle_post_send_edge - Check if we get to SQ edge. If yes, update to the + * next nearby edge and get new address translation for current WQE position. + * @sq - SQ buffer. + * @seg: Current WQE position (16B aligned). + * @wqe_sz: Total current WQE size [16B]. + * @cur_edge: Updated current edge. + */ +static inline void handle_post_send_edge(struct mlx5_ib_wq *sq, void **seg, + u32 wqe_sz, void **cur_edge) +{ + if (likely(*seg != *cur_edge)) + return; + + _handle_post_send_edge(sq, seg, wqe_sz, cur_edge); +} + +/* memcpy_send_wqe - copy data from src to WQE and update the relevant WQ's + * pointers. At the end @seg is aligned to 16B regardless the copied size. + * @sq - SQ buffer. + * @cur_edge: Updated current edge. + * @seg: Current WQE position (16B aligned). + * @wqe_sz: Total current WQE size [16B]. + * @src: Pointer to copy from. + * @n: Number of bytes to copy. + */ +static inline void memcpy_send_wqe(struct mlx5_ib_wq *sq, void **cur_edge, + void **seg, u32 *wqe_sz, const void *src, + size_t n) +{ + while (likely(n)) { + size_t leftlen = *cur_edge - *seg; + size_t copysz = min_t(size_t, leftlen, n); + size_t stride; + + memcpy(*seg, src, copysz); + + n -= copysz; + src += copysz; + stride = !n ? ALIGN(copysz, 16) : copysz; + *seg += stride; + *wqe_sz += stride >> 4; + handle_post_send_edge(sq, seg, *wqe_sz, cur_edge); + } +} + static int mlx5_wq_overflow(struct mlx5_ib_wq *wq, int nreq, struct ib_cq *ib_cq) { struct mlx5_ib_cq *cq; @@ -3775,11 +3859,10 @@ static __always_inline void set_raddr_seg(struct mlx5_wqe_raddr_seg *rseg, rseg->reserved = 0; } -static void *set_eth_seg(struct mlx5_wqe_eth_seg *eseg, - const struct ib_send_wr *wr, void *qend, - struct mlx5_ib_qp *qp, int *size) +static void set_eth_seg(const struct ib_send_wr *wr, struct mlx5_ib_qp *qp, + void **seg, int *size, void **cur_edge) { - void *seg = eseg; + struct mlx5_wqe_eth_seg *eseg = *seg; memset(eseg, 0, sizeof(struct mlx5_wqe_eth_seg)); @@ -3787,45 +3870,41 @@ static void *set_eth_seg(struct mlx5_wqe_eth_seg *eseg, eseg->cs_flags = MLX5_ETH_WQE_L3_CSUM | MLX5_ETH_WQE_L4_CSUM; - seg += sizeof(struct mlx5_wqe_eth_seg); - *size += sizeof(struct mlx5_wqe_eth_seg) / 16; - if (wr->opcode == IB_WR_LSO) { struct ib_ud_wr *ud_wr = container_of(wr, struct ib_ud_wr, wr); - int size_of_inl_hdr_start = sizeof(eseg->inline_hdr.start); - u64 left, leftlen, copysz; + size_t left, copysz; void *pdata = ud_wr->header; + size_t stride; left = ud_wr->hlen; eseg->mss = cpu_to_be16(ud_wr->mss); eseg->inline_hdr.sz = cpu_to_be16(left); - /* - * check if there is space till the end of queue, if yes, - * copy all in one shot, otherwise copy till the end of queue, - * rollback and than the copy the left + /* memcpy_send_wqe should get a 16B align address. Hence, we + * first copy up to the current edge and then, if needed, + * fall-through to memcpy_send_wqe. */ - leftlen = qend - (void *)eseg->inline_hdr.start; - copysz = min_t(u64, leftlen, left); - - memcpy(seg - size_of_inl_hdr_start, pdata, copysz); - - if (likely(copysz > size_of_inl_hdr_start)) { - seg += ALIGN(copysz - size_of_inl_hdr_start, 16); - *size += ALIGN(copysz - size_of_inl_hdr_start, 16) / 16; - } - - if (unlikely(copysz < left)) { /* the last wqe in the queue */ - seg = mlx5_get_send_wqe(qp, 0); + copysz = min_t(u64, *cur_edge - (void *)eseg->inline_hdr.start, + left); + memcpy(eseg->inline_hdr.start, pdata, copysz); + stride = ALIGN(sizeof(struct mlx5_wqe_eth_seg) - + sizeof(eseg->inline_hdr.start) + copysz, 16); + *size += stride / 16; + *seg += stride; + + if (copysz < left) { + handle_post_send_edge(&qp->sq, seg, *size, cur_edge); left -= copysz; pdata += copysz; - memcpy(seg, pdata, left); - seg += ALIGN(left, 16); - *size += ALIGN(left, 16) / 16; + memcpy_send_wqe(&qp->sq, cur_edge, seg, size, pdata, + left); } + + return; } - return seg; + *seg += sizeof(struct mlx5_wqe_eth_seg); + *size += sizeof(struct mlx5_wqe_eth_seg) / 16; } static void set_datagram_seg(struct mlx5_wqe_datagram_seg *dseg, @@ -4084,24 +4163,6 @@ static void set_reg_data_seg(struct mlx5_wqe_data_seg *dseg, dseg->lkey = cpu_to_be32(pd->ibpd.local_dma_lkey); } -static void set_reg_umr_inline_seg(void *seg, struct mlx5_ib_qp *qp, - struct mlx5_ib_mr *mr, int mr_list_size) -{ - void *qend = qp->sq.qend; - void *addr = mr->descs; - int copy; - - if (unlikely(seg + mr_list_size > qend)) { - copy = qend - seg; - memcpy(seg, addr, copy); - addr += copy; - mr_list_size -= copy; - seg = mlx5_get_send_wqe(qp, 0); - } - memcpy(seg, addr, mr_list_size); - seg += mr_list_size; -} - static __be32 send_ieth(const struct ib_send_wr *wr) { switch (wr->opcode) { @@ -4135,40 +4196,48 @@ static u8 wq_sig(void *wqe) } static int set_data_inl_seg(struct mlx5_ib_qp *qp, const struct ib_send_wr *wr, - void *wqe, int *sz) + void **wqe, int *wqe_sz, void **cur_edge) { struct mlx5_wqe_inline_seg *seg; - void *qend = qp->sq.qend; - void *addr; + size_t offset; int inl = 0; - int copy; - int len; int i; - seg = wqe; - wqe += sizeof(*seg); + seg = *wqe; + *wqe += sizeof(*seg); + offset = sizeof(*seg); + for (i = 0; i < wr->num_sge; i++) { - addr = (void *)(unsigned long)(wr->sg_list[i].addr); - len = wr->sg_list[i].length; + size_t len = wr->sg_list[i].length; + void *addr = (void *)(unsigned long)(wr->sg_list[i].addr); + inl += len; if (unlikely(inl > qp->max_inline_data)) return -ENOMEM; - if (unlikely(wqe + len > qend)) { - copy = qend - wqe; - memcpy(wqe, addr, copy); - addr += copy; - len -= copy; - wqe = mlx5_get_send_wqe(qp, 0); + while (likely(len)) { + size_t leftlen; + size_t copysz; + + handle_post_send_edge(&qp->sq, wqe, + *wqe_sz + (offset >> 4), + cur_edge); + + leftlen = *cur_edge - *wqe; + copysz = min_t(size_t, leftlen, len); + + memcpy(*wqe, addr, copysz); + len -= copysz; + addr += copysz; + *wqe += copysz; + offset += copysz; } - memcpy(wqe, addr, len); - wqe += len; } seg->byte_count = cpu_to_be32(inl | MLX5_INLINE_SEG); - *sz = ALIGN(inl + sizeof(seg->byte_count), 16) / 16; + *wqe_sz += ALIGN(inl + sizeof(seg->byte_count), 16) / 16; return 0; } @@ -4281,7 +4350,8 @@ static int mlx5_set_bsf(struct ib_mr *sig_mr, } static int set_sig_data_segment(const struct ib_sig_handover_wr *wr, - struct mlx5_ib_qp *qp, void **seg, int *size) + struct mlx5_ib_qp *qp, void **seg, + int *size, void **cur_edge) { struct ib_sig_attrs *sig_attrs = wr->sig_attrs; struct ib_mr *sig_mr = wr->sig_mr; @@ -4365,8 +4435,7 @@ static int set_sig_data_segment(const struct ib_sig_handover_wr *wr, *seg += wqe_size; *size += wqe_size / 16; - if (unlikely((*seg == qp->sq.qend))) - *seg = mlx5_get_send_wqe(qp, 0); + handle_post_send_edge(&qp->sq, seg, *size, cur_edge); bsf = *seg; ret = mlx5_set_bsf(sig_mr, sig_attrs, bsf, data_len); @@ -4375,8 +4444,7 @@ static int set_sig_data_segment(const struct ib_sig_handover_wr *wr, *seg += sizeof(*bsf); *size += sizeof(*bsf) / 16; - if (unlikely((*seg == qp->sq.qend))) - *seg = mlx5_get_send_wqe(qp, 0); + handle_post_send_edge(&qp->sq, seg, *size, cur_edge); return 0; } @@ -4414,7 +4482,8 @@ static void set_sig_umr_segment(struct mlx5_wqe_umr_ctrl_seg *umr, static int set_sig_umr_wr(const struct ib_send_wr *send_wr, - struct mlx5_ib_qp *qp, void **seg, int *size) + struct mlx5_ib_qp *qp, void **seg, int *size, + void **cur_edge) { const struct ib_sig_handover_wr *wr = sig_handover_wr(send_wr); struct mlx5_ib_mr *sig_mr = to_mmr(wr->sig_mr); @@ -4446,16 +4515,14 @@ static int set_sig_umr_wr(const struct ib_send_wr *send_wr, set_sig_umr_segment(*seg, xlt_size); *seg += sizeof(struct mlx5_wqe_umr_ctrl_seg); *size += sizeof(struct mlx5_wqe_umr_ctrl_seg) / 16; - if (unlikely((*seg == qp->sq.qend))) - *seg = mlx5_get_send_wqe(qp, 0); + handle_post_send_edge(&qp->sq, seg, *size, cur_edge); set_sig_mkey_segment(*seg, wr, xlt_size, region_len, pdn); *seg += sizeof(struct mlx5_mkey_seg); *size += sizeof(struct mlx5_mkey_seg) / 16; - if (unlikely((*seg == qp->sq.qend))) - *seg = mlx5_get_send_wqe(qp, 0); + handle_post_send_edge(&qp->sq, seg, *size, cur_edge); - ret = set_sig_data_segment(wr, qp, seg, size); + ret = set_sig_data_segment(wr, qp, seg, size, cur_edge); if (ret) return ret; @@ -4492,11 +4559,11 @@ static int set_psv_wr(struct ib_sig_domain *domain, static int set_reg_wr(struct mlx5_ib_qp *qp, const struct ib_reg_wr *wr, - void **seg, int *size) + void **seg, int *size, void **cur_edge) { struct mlx5_ib_mr *mr = to_mmr(wr->mr); struct mlx5_ib_pd *pd = to_mpd(qp->ibqp.pd); - int mr_list_size = mr->ndescs * mr->desc_size; + size_t mr_list_size = mr->ndescs * mr->desc_size; bool umr_inline = mr_list_size <= MLX5_IB_SQ_UMR_INLINE_THRESHOLD; if (unlikely(wr->wr.send_flags & IB_SEND_INLINE)) { @@ -4508,18 +4575,17 @@ static int set_reg_wr(struct mlx5_ib_qp *qp, set_reg_umr_seg(*seg, mr, umr_inline); *seg += sizeof(struct mlx5_wqe_umr_ctrl_seg); *size += sizeof(struct mlx5_wqe_umr_ctrl_seg) / 16; - if (unlikely((*seg == qp->sq.qend))) - *seg = mlx5_get_send_wqe(qp, 0); + handle_post_send_edge(&qp->sq, seg, *size, cur_edge); set_reg_mkey_seg(*seg, mr, wr->key, wr->access); *seg += sizeof(struct mlx5_mkey_seg); *size += sizeof(struct mlx5_mkey_seg) / 16; - if (unlikely((*seg == qp->sq.qend))) - *seg = mlx5_get_send_wqe(qp, 0); + handle_post_send_edge(&qp->sq, seg, *size, cur_edge); if (umr_inline) { - set_reg_umr_inline_seg(*seg, qp, mr, mr_list_size); - *size += get_xlt_octo(mr_list_size); + memcpy_send_wqe(&qp->sq, cur_edge, seg, size, mr->descs, + mr_list_size); + *size = ALIGN(*size, MLX5_SEND_WQE_BB >> 4); } else { set_reg_data_seg(*seg, mr, pd); *seg += sizeof(struct mlx5_wqe_data_seg); @@ -4528,32 +4594,31 @@ static int set_reg_wr(struct mlx5_ib_qp *qp, return 0; } -static void set_linv_wr(struct mlx5_ib_qp *qp, void **seg, int *size) +static void set_linv_wr(struct mlx5_ib_qp *qp, void **seg, int *size, + void **cur_edge) { set_linv_umr_seg(*seg); *seg += sizeof(struct mlx5_wqe_umr_ctrl_seg); *size += sizeof(struct mlx5_wqe_umr_ctrl_seg) / 16; - if (unlikely((*seg == qp->sq.qend))) - *seg = mlx5_get_send_wqe(qp, 0); + handle_post_send_edge(&qp->sq, seg, *size, cur_edge); set_linv_mkey_seg(*seg); *seg += sizeof(struct mlx5_mkey_seg); *size += sizeof(struct mlx5_mkey_seg) / 16; - if (unlikely((*seg == qp->sq.qend))) - *seg = mlx5_get_send_wqe(qp, 0); + handle_post_send_edge(&qp->sq, seg, *size, cur_edge); } -static void dump_wqe(struct mlx5_ib_qp *qp, int idx, int size_16) +static void dump_wqe(struct mlx5_ib_qp *qp, u32 idx, int size_16) { __be32 *p = NULL; - int tidx = idx; + u32 tidx = idx; int i, j; - pr_debug("dump wqe at %p\n", mlx5_get_send_wqe(qp, tidx)); + pr_debug("dump WQE index %u:\n", idx); for (i = 0, j = 0; i < size_16 * 4; i += 4, j += 4) { if ((i & 0xf) == 0) { - void *buf = mlx5_get_send_wqe(qp, tidx); tidx = (tidx + 1) & (qp->sq.wqe_cnt - 1); - p = buf; + p = mlx5_frag_buf_get_wqe(&qp->sq.fbc, tidx); + pr_debug("WQBB at %p:\n", (void *)p); j = 0; } pr_debug("%08x %08x %08x %08x\n", be32_to_cpu(p[j]), @@ -4563,15 +4628,16 @@ static void dump_wqe(struct mlx5_ib_qp *qp, int idx, int size_16) } static int __begin_wqe(struct mlx5_ib_qp *qp, void **seg, - struct mlx5_wqe_ctrl_seg **ctrl, - const struct ib_send_wr *wr, unsigned *idx, - int *size, int nreq, bool send_signaled, bool solicited) + struct mlx5_wqe_ctrl_seg **ctrl, + const struct ib_send_wr *wr, unsigned int *idx, + int *size, void **cur_edge, int nreq, + bool send_signaled, bool solicited) { if (unlikely(mlx5_wq_overflow(&qp->sq, nreq, qp->ibqp.send_cq))) return -ENOMEM; *idx = qp->sq.cur_post & (qp->sq.wqe_cnt - 1); - *seg = mlx5_get_send_wqe(qp, *idx); + *seg = mlx5_frag_buf_get_wqe(&qp->sq.fbc, *idx); *ctrl = *seg; *(uint32_t *)(*seg + 8) = 0; (*ctrl)->imm = send_ieth(wr); @@ -4581,6 +4647,7 @@ static int __begin_wqe(struct mlx5_ib_qp *qp, void **seg, *seg += sizeof(**ctrl); *size = sizeof(**ctrl) / 16; + *cur_edge = qp->sq.cur_edge; return 0; } @@ -4588,17 +4655,18 @@ static int __begin_wqe(struct mlx5_ib_qp *qp, void **seg, static int begin_wqe(struct mlx5_ib_qp *qp, void **seg, struct mlx5_wqe_ctrl_seg **ctrl, const struct ib_send_wr *wr, unsigned *idx, - int *size, int nreq) + int *size, void **cur_edge, int nreq) { - return __begin_wqe(qp, seg, ctrl, wr, idx, size, nreq, + return __begin_wqe(qp, seg, ctrl, wr, idx, size, cur_edge, nreq, wr->send_flags & IB_SEND_SIGNALED, wr->send_flags & IB_SEND_SOLICITED); } static void finish_wqe(struct mlx5_ib_qp *qp, struct mlx5_wqe_ctrl_seg *ctrl, - u8 size, unsigned idx, u64 wr_id, - int nreq, u8 fence, u32 mlx5_opcode) + void *seg, u8 size, void *cur_edge, + unsigned int idx, u64 wr_id, int nreq, u8 fence, + u32 mlx5_opcode) { u8 opmod = 0; @@ -4614,6 +4682,15 @@ static void finish_wqe(struct mlx5_ib_qp *qp, qp->sq.wqe_head[idx] = qp->sq.head + nreq; qp->sq.cur_post += DIV_ROUND_UP(size * 16, MLX5_SEND_WQE_BB); qp->sq.w_list[idx].next = qp->sq.cur_post; + + /* We save the edge which was possibly updated during the WQE + * construction, into SQ's cache. + */ + seg = PTR_ALIGN(seg, MLX5_SEND_WQE_BB); + qp->sq.cur_edge = (unlikely(seg == cur_edge)) ? + get_sq_edge(&qp->sq, qp->sq.cur_post & + (qp->sq.wqe_cnt - 1)) : + cur_edge; } static int _mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr, @@ -4624,11 +4701,10 @@ static int _mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr, struct mlx5_core_dev *mdev = dev->mdev; struct mlx5_ib_qp *qp; struct mlx5_ib_mr *mr; - struct mlx5_wqe_data_seg *dpseg; struct mlx5_wqe_xrc_seg *xrc; struct mlx5_bf *bf; + void *cur_edge; int uninitialized_var(size); - void *qend; unsigned long flags; unsigned idx; int err = 0; @@ -4650,7 +4726,6 @@ static int _mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr, qp = to_mqp(ibqp); bf = &qp->bf; - qend = qp->sq.qend; spin_lock_irqsave(&qp->sq.lock, flags); @@ -4670,7 +4745,8 @@ static int _mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr, goto out; } - err = begin_wqe(qp, &seg, &ctrl, wr, &idx, &size, nreq); + err = begin_wqe(qp, &seg, &ctrl, wr, &idx, &size, &cur_edge, + nreq); if (err) { mlx5_ib_warn(dev, "\n"); err = -ENOMEM; @@ -4719,14 +4795,15 @@ static int _mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr, case IB_WR_LOCAL_INV: qp->sq.wr_data[idx] = IB_WR_LOCAL_INV; ctrl->imm = cpu_to_be32(wr->ex.invalidate_rkey); - set_linv_wr(qp, &seg, &size); + set_linv_wr(qp, &seg, &size, &cur_edge); num_sge = 0; break; case IB_WR_REG_MR: qp->sq.wr_data[idx] = IB_WR_REG_MR; ctrl->imm = cpu_to_be32(reg_wr(wr)->key); - err = set_reg_wr(qp, reg_wr(wr), &seg, &size); + err = set_reg_wr(qp, reg_wr(wr), &seg, &size, + &cur_edge); if (err) { *bad_wr = wr; goto out; @@ -4739,21 +4816,24 @@ static int _mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr, mr = to_mmr(sig_handover_wr(wr)->sig_mr); ctrl->imm = cpu_to_be32(mr->ibmr.rkey); - err = set_sig_umr_wr(wr, qp, &seg, &size); + err = set_sig_umr_wr(wr, qp, &seg, &size, + &cur_edge); if (err) { mlx5_ib_warn(dev, "\n"); *bad_wr = wr; goto out; } - finish_wqe(qp, ctrl, size, idx, wr->wr_id, nreq, - fence, MLX5_OPCODE_UMR); + finish_wqe(qp, ctrl, seg, size, cur_edge, idx, + wr->wr_id, nreq, fence, + MLX5_OPCODE_UMR); /* * SET_PSV WQEs are not signaled and solicited * on error */ err = __begin_wqe(qp, &seg, &ctrl, wr, &idx, - &size, nreq, false, true); + &size, &cur_edge, nreq, false, + true); if (err) { mlx5_ib_warn(dev, "\n"); err = -ENOMEM; @@ -4770,10 +4850,12 @@ static int _mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr, goto out; } - finish_wqe(qp, ctrl, size, idx, wr->wr_id, nreq, - fence, MLX5_OPCODE_SET_PSV); + finish_wqe(qp, ctrl, seg, size, cur_edge, idx, + wr->wr_id, nreq, fence, + MLX5_OPCODE_SET_PSV); err = __begin_wqe(qp, &seg, &ctrl, wr, &idx, - &size, nreq, false, true); + &size, &cur_edge, nreq, false, + true); if (err) { mlx5_ib_warn(dev, "\n"); err = -ENOMEM; @@ -4790,8 +4872,9 @@ static int _mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr, goto out; } - finish_wqe(qp, ctrl, size, idx, wr->wr_id, nreq, - fence, MLX5_OPCODE_SET_PSV); + finish_wqe(qp, ctrl, seg, size, cur_edge, idx, + wr->wr_id, nreq, fence, + MLX5_OPCODE_SET_PSV); qp->next_fence = MLX5_FENCE_MODE_INITIATOR_SMALL; num_sge = 0; goto skip_psv; @@ -4828,16 +4911,14 @@ static int _mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr, set_datagram_seg(seg, wr); seg += sizeof(struct mlx5_wqe_datagram_seg); size += sizeof(struct mlx5_wqe_datagram_seg) / 16; - if (unlikely((seg == qend))) - seg = mlx5_get_send_wqe(qp, 0); + handle_post_send_edge(&qp->sq, &seg, size, &cur_edge); + break; case IB_QPT_UD: set_datagram_seg(seg, wr); seg += sizeof(struct mlx5_wqe_datagram_seg); size += sizeof(struct mlx5_wqe_datagram_seg) / 16; - - if (unlikely((seg == qend))) - seg = mlx5_get_send_wqe(qp, 0); + handle_post_send_edge(&qp->sq, &seg, size, &cur_edge); /* handle qp that supports ud offload */ if (qp->flags & IB_QP_CREATE_IPOIB_UD_LSO) { @@ -4847,11 +4928,9 @@ static int _mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr, memset(pad, 0, sizeof(struct mlx5_wqe_eth_pad)); seg += sizeof(struct mlx5_wqe_eth_pad); size += sizeof(struct mlx5_wqe_eth_pad) / 16; - - seg = set_eth_seg(seg, wr, qend, qp, &size); - - if (unlikely((seg == qend))) - seg = mlx5_get_send_wqe(qp, 0); + set_eth_seg(wr, qp, &seg, &size, &cur_edge); + handle_post_send_edge(&qp->sq, &seg, size, + &cur_edge); } break; case MLX5_IB_QPT_REG_UMR: @@ -4867,13 +4946,11 @@ static int _mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr, goto out; seg += sizeof(struct mlx5_wqe_umr_ctrl_seg); size += sizeof(struct mlx5_wqe_umr_ctrl_seg) / 16; - if (unlikely((seg == qend))) - seg = mlx5_get_send_wqe(qp, 0); + handle_post_send_edge(&qp->sq, &seg, size, &cur_edge); set_reg_mkey_segment(seg, wr); seg += sizeof(struct mlx5_mkey_seg); size += sizeof(struct mlx5_mkey_seg) / 16; - if (unlikely((seg == qend))) - seg = mlx5_get_send_wqe(qp, 0); + handle_post_send_edge(&qp->sq, &seg, size, &cur_edge); break; default: @@ -4881,33 +4958,29 @@ static int _mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr, } if (wr->send_flags & IB_SEND_INLINE && num_sge) { - int uninitialized_var(sz); - - err = set_data_inl_seg(qp, wr, seg, &sz); + err = set_data_inl_seg(qp, wr, &seg, &size, &cur_edge); if (unlikely(err)) { mlx5_ib_warn(dev, "\n"); *bad_wr = wr; goto out; } - size += sz; } else { - dpseg = seg; for (i = 0; i < num_sge; i++) { - if (unlikely(dpseg == qend)) { - seg = mlx5_get_send_wqe(qp, 0); - dpseg = seg; - } + handle_post_send_edge(&qp->sq, &seg, size, + &cur_edge); if (likely(wr->sg_list[i].length)) { - set_data_ptr_seg(dpseg, wr->sg_list + i); + set_data_ptr_seg + ((struct mlx5_wqe_data_seg *)seg, + wr->sg_list + i); size += sizeof(struct mlx5_wqe_data_seg) / 16; - dpseg++; + seg += sizeof(struct mlx5_wqe_data_seg); } } } qp->next_fence = next_fence; - finish_wqe(qp, ctrl, size, idx, wr->wr_id, nreq, fence, - mlx5_ib_opcode[wr->opcode]); + finish_wqe(qp, ctrl, seg, size, cur_edge, idx, wr->wr_id, nreq, + fence, mlx5_ib_opcode[wr->opcode]); skip_psv: if (0) dump_wqe(qp, idx, size); @@ -4993,7 +5066,7 @@ static int _mlx5_ib_post_recv(struct ib_qp *ibqp, const struct ib_recv_wr *wr, goto out; } - scat = get_recv_wqe(qp, ind); + scat = mlx5_frag_buf_get_wqe(&qp->rq.fbc, ind); if (qp->wq_sig) scat++; -- cgit v1.2.3-59-g8ed1b From 34613eb1d2ad051cb02cbcf5675482340cbb9bf0 Mon Sep 17 00:00:00 2001 From: Yishai Hadas Date: Mon, 26 Nov 2018 08:28:35 +0200 Subject: IB/mlx5: Enable modify and query verbs objects via DEVX Enables modify and query verbs objects via the DEVX interface. To support this the above DEVX handlers were changed to get any object type via the UVERBS_IDR_ANY_OBJECT mechanism. The type checking and handling is done per object as part of the driver code. Signed-off-by: Yishai Hadas Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx5/devx.c | 108 +++++++++++++++++++++++++++++++++----- 1 file changed, 96 insertions(+), 12 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/devx.c b/drivers/infiniband/hw/mlx5/devx.c index 0aa2ee732eaa..f80b78aab4da 100644 --- a/drivers/infiniband/hw/mlx5/devx.c +++ b/drivers/infiniband/hw/mlx5/devx.c @@ -9,6 +9,7 @@ #include #include #include +#include #include #include #include "mlx5_ib.h" @@ -132,7 +133,7 @@ static u64 get_enc_obj_id(u16 opcode, u32 obj_id) return ((u64)opcode << 32) | obj_id; } -static int devx_is_valid_obj_id(struct devx_obj *obj, const void *in) +static u64 devx_get_obj_id(const void *in) { u16 opcode = MLX5_GET(general_obj_in_cmd_hdr, in, opcode); u64 obj_id; @@ -336,13 +337,96 @@ static int devx_is_valid_obj_id(struct devx_obj *obj, const void *in) MLX5_GET(arm_xrq_in, in, xrqn)); break; default: + obj_id = 0; + } + + return obj_id; +} + +static bool devx_is_valid_obj_id(struct ib_uobject *uobj, const void *in) +{ + u64 obj_id = devx_get_obj_id(in); + + if (!obj_id) return false; + + switch (uobj_get_object_id(uobj)) { + case UVERBS_OBJECT_CQ: + return get_enc_obj_id(MLX5_CMD_OP_CREATE_CQ, + to_mcq(uobj->object)->mcq.cqn) == + obj_id; + + case UVERBS_OBJECT_SRQ: + { + struct mlx5_core_srq *srq = &(to_msrq(uobj->object)->msrq); + struct mlx5_ib_dev *dev = to_mdev(uobj->context->device); + u16 opcode; + + switch (srq->common.res) { + case MLX5_RES_XSRQ: + opcode = MLX5_CMD_OP_CREATE_XRC_SRQ; + break; + case MLX5_RES_XRQ: + opcode = MLX5_CMD_OP_CREATE_XRQ; + break; + default: + if (!dev->mdev->issi) + opcode = MLX5_CMD_OP_CREATE_SRQ; + else + opcode = MLX5_CMD_OP_CREATE_RMP; + } + + return get_enc_obj_id(opcode, + to_msrq(uobj->object)->msrq.srqn) == + obj_id; } - if (obj_id == obj->obj_id) - return true; + case UVERBS_OBJECT_QP: + { + struct mlx5_ib_qp *qp = to_mqp(uobj->object); + enum ib_qp_type qp_type = qp->ibqp.qp_type; + + if (qp_type == IB_QPT_RAW_PACKET || + (qp->flags & MLX5_IB_QP_UNDERLAY)) { + struct mlx5_ib_raw_packet_qp *raw_packet_qp = + &qp->raw_packet_qp; + struct mlx5_ib_rq *rq = &raw_packet_qp->rq; + struct mlx5_ib_sq *sq = &raw_packet_qp->sq; + + return (get_enc_obj_id(MLX5_CMD_OP_CREATE_RQ, + rq->base.mqp.qpn) == obj_id || + get_enc_obj_id(MLX5_CMD_OP_CREATE_SQ, + sq->base.mqp.qpn) == obj_id || + get_enc_obj_id(MLX5_CMD_OP_CREATE_TIR, + rq->tirn) == obj_id || + get_enc_obj_id(MLX5_CMD_OP_CREATE_TIS, + sq->tisn) == obj_id); + } + + if (qp_type == MLX5_IB_QPT_DCT) + return get_enc_obj_id(MLX5_CMD_OP_CREATE_DCT, + qp->dct.mdct.mqp.qpn) == obj_id; + + return get_enc_obj_id(MLX5_CMD_OP_CREATE_QP, + qp->ibqp.qp_num) == obj_id; + } - return false; + case UVERBS_OBJECT_WQ: + return get_enc_obj_id(MLX5_CMD_OP_CREATE_RQ, + to_mrwq(uobj->object)->core_qp.qpn) == + obj_id; + + case UVERBS_OBJECT_RWQ_IND_TBL: + return get_enc_obj_id(MLX5_CMD_OP_CREATE_RQT, + to_mrwq_ind_table(uobj->object)->rqtn) == + obj_id; + + case MLX5_IB_OBJECT_DEVX_OBJ: + return ((struct devx_obj *)uobj->object)->obj_id == obj_id; + + default: + return false; + } } static void devx_set_umem_valid(const void *in) @@ -994,7 +1078,7 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_MODIFY)( struct ib_uobject *uobj = uverbs_attr_get_uobject(attrs, MLX5_IB_ATTR_DEVX_OBJ_MODIFY_HANDLE); struct mlx5_ib_ucontext *c = to_mucontext(uobj->context); - struct devx_obj *obj = uobj->object; + struct mlx5_ib_dev *mdev = to_mdev(uobj->context->device); void *cmd_out; int err; int uid; @@ -1006,7 +1090,7 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_MODIFY)( if (!devx_is_obj_modify_cmd(cmd_in)) return -EINVAL; - if (!devx_is_valid_obj_id(obj, cmd_in)) + if (!devx_is_valid_obj_id(uobj, cmd_in)) return -EINVAL; cmd_out = uverbs_zalloc(attrs, cmd_out_len); @@ -1016,7 +1100,7 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_MODIFY)( MLX5_SET(general_obj_in_cmd_hdr, cmd_in, uid, uid); devx_set_umem_valid(cmd_in); - err = mlx5_cmd_exec(obj->mdev, cmd_in, + err = mlx5_cmd_exec(mdev->mdev, cmd_in, uverbs_attr_get_len(attrs, MLX5_IB_ATTR_DEVX_OBJ_MODIFY_CMD_IN), cmd_out, cmd_out_len); if (err) @@ -1035,10 +1119,10 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_QUERY)( struct ib_uobject *uobj = uverbs_attr_get_uobject(attrs, MLX5_IB_ATTR_DEVX_OBJ_QUERY_HANDLE); struct mlx5_ib_ucontext *c = to_mucontext(uobj->context); - struct devx_obj *obj = uobj->object; void *cmd_out; int err; int uid; + struct mlx5_ib_dev *mdev = to_mdev(uobj->context->device); uid = devx_get_uid(c, cmd_in); if (uid < 0) @@ -1047,7 +1131,7 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_QUERY)( if (!devx_is_obj_query_cmd(cmd_in)) return -EINVAL; - if (!devx_is_valid_obj_id(obj, cmd_in)) + if (!devx_is_valid_obj_id(uobj, cmd_in)) return -EINVAL; cmd_out = uverbs_zalloc(attrs, cmd_out_len); @@ -1055,7 +1139,7 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_QUERY)( return PTR_ERR(cmd_out); MLX5_SET(general_obj_in_cmd_hdr, cmd_in, uid, uid); - err = mlx5_cmd_exec(obj->mdev, cmd_in, + err = mlx5_cmd_exec(mdev->mdev, cmd_in, uverbs_attr_get_len(attrs, MLX5_IB_ATTR_DEVX_OBJ_QUERY_CMD_IN), cmd_out, cmd_out_len); if (err) @@ -1293,7 +1377,7 @@ DECLARE_UVERBS_NAMED_METHOD_DESTROY( DECLARE_UVERBS_NAMED_METHOD( MLX5_IB_METHOD_DEVX_OBJ_MODIFY, UVERBS_ATTR_IDR(MLX5_IB_ATTR_DEVX_OBJ_MODIFY_HANDLE, - MLX5_IB_OBJECT_DEVX_OBJ, + UVERBS_IDR_ANY_OBJECT, UVERBS_ACCESS_WRITE, UA_MANDATORY), UVERBS_ATTR_PTR_IN( @@ -1309,7 +1393,7 @@ DECLARE_UVERBS_NAMED_METHOD( DECLARE_UVERBS_NAMED_METHOD( MLX5_IB_METHOD_DEVX_OBJ_QUERY, UVERBS_ATTR_IDR(MLX5_IB_ATTR_DEVX_OBJ_QUERY_HANDLE, - MLX5_IB_OBJECT_DEVX_OBJ, + UVERBS_IDR_ANY_OBJECT, UVERBS_ACCESS_READ, UA_MANDATORY), UVERBS_ATTR_PTR_IN( -- cgit v1.2.3-59-g8ed1b From fb98153bbf28b627fe52f41e658ae39fa67d2684 Mon Sep 17 00:00:00 2001 From: Yishai Hadas Date: Mon, 26 Nov 2018 08:28:36 +0200 Subject: IB/mlx5: Enforce DEVX privilege by firmware Enforce DEVX privilege by firmware, this enables future device functionality without the need to make driver changes unless a new privilege type will be introduced. Signed-off-by: Yishai Hadas Reviewed-by: Artemy Kovalyov Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx5/devx.c | 17 +++++++++-------- drivers/infiniband/hw/mlx5/main.c | 4 ++-- drivers/infiniband/hw/mlx5/mlx5_ib.h | 5 +++-- 3 files changed, 14 insertions(+), 12 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/devx.c b/drivers/infiniband/hw/mlx5/devx.c index f80b78aab4da..80053324dd31 100644 --- a/drivers/infiniband/hw/mlx5/devx.c +++ b/drivers/infiniband/hw/mlx5/devx.c @@ -47,24 +47,31 @@ devx_ufile2uctx(const struct uverbs_attr_bundle *attrs) return to_mucontext(ib_uverbs_get_ucontext(attrs)); } -int mlx5_ib_devx_create(struct mlx5_ib_dev *dev) +int mlx5_ib_devx_create(struct mlx5_ib_dev *dev, bool is_user) { u32 in[MLX5_ST_SZ_DW(create_uctx_in)] = {0}; u32 out[MLX5_ST_SZ_DW(general_obj_out_cmd_hdr)] = {0}; u64 general_obj_types; - void *hdr; + void *hdr, *uctx; int err; u16 uid; + u32 cap = 0; hdr = MLX5_ADDR_OF(create_uctx_in, in, hdr); + uctx = MLX5_ADDR_OF(create_uctx_in, in, uctx); general_obj_types = MLX5_CAP_GEN_64(dev->mdev, general_obj_types); if (!(general_obj_types & MLX5_GENERAL_OBJ_TYPES_CAP_UCTX) || !(general_obj_types & MLX5_GENERAL_OBJ_TYPES_CAP_UMEM)) return -EINVAL; + if (is_user && capable(CAP_NET_RAW) && + (MLX5_CAP_GEN(dev->mdev, uctx_cap) & MLX5_UCTX_CAP_RAW_TX)) + cap |= MLX5_UCTX_CAP_RAW_TX; + MLX5_SET(general_obj_in_cmd_hdr, hdr, opcode, MLX5_CMD_OP_CREATE_GENERAL_OBJECT); MLX5_SET(general_obj_in_cmd_hdr, hdr, obj_type, MLX5_OBJ_TYPE_UCTX); + MLX5_SET(uctx, uctx, cap, cap); err = mlx5_cmd_exec(dev->mdev, in, sizeof(in), out, sizeof(out)); if (err) @@ -672,9 +679,6 @@ static int devx_get_uid(struct mlx5_ib_ucontext *c, void *cmd_in) if (!c->devx_uid) return -EINVAL; - if (!capable(CAP_NET_RAW)) - return -EPERM; - return c->devx_uid; } static bool devx_is_general_cmd(void *in) @@ -1239,9 +1243,6 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_UMEM_REG)( if (!c->devx_uid) return -EINVAL; - if (!capable(CAP_NET_RAW)) - return -EPERM; - obj = kzalloc(sizeof(struct devx_umem), GFP_KERNEL); if (!obj) return -ENOMEM; diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index b3986bc961ca..2b09e6896e5a 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -1763,7 +1763,7 @@ static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev, #endif if (req.flags & MLX5_IB_ALLOC_UCTX_DEVX) { - err = mlx5_ib_devx_create(dev); + err = mlx5_ib_devx_create(dev, true); if (err < 0) goto out_uars; context->devx_uid = err; @@ -6234,7 +6234,7 @@ static int mlx5_ib_stage_devx_init(struct mlx5_ib_dev *dev) { int uid; - uid = mlx5_ib_devx_create(dev); + uid = mlx5_ib_devx_create(dev, false); if (uid > 0) dev->devx_whitelist_uid = uid; diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index 59e1664a107f..4d33965369cc 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -1268,7 +1268,7 @@ void mlx5_ib_put_native_port_mdev(struct mlx5_ib_dev *dev, u8 port_num); #if IS_ENABLED(CONFIG_INFINIBAND_USER_ACCESS) -int mlx5_ib_devx_create(struct mlx5_ib_dev *dev); +int mlx5_ib_devx_create(struct mlx5_ib_dev *dev, bool is_user); void mlx5_ib_devx_destroy(struct mlx5_ib_dev *dev, u16 uid); const struct uverbs_object_tree_def *mlx5_ib_get_devx_tree(void); extern const struct uapi_definition mlx5_ib_devx_defs[]; @@ -1283,7 +1283,8 @@ int mlx5_ib_get_flow_trees(const struct uverbs_object_tree_def **root); void mlx5_ib_destroy_flow_action_raw(struct mlx5_ib_flow_action *maction); #else static inline int -mlx5_ib_devx_create(struct mlx5_ib_dev *dev) { return -EOPNOTSUPP; }; +mlx5_ib_devx_create(struct mlx5_ib_dev *dev, + bool is_user) { return -EOPNOTSUPP; } static inline void mlx5_ib_devx_destroy(struct mlx5_ib_dev *dev, u16 uid) {} static inline bool mlx5_ib_devx_is_flow_dest(void *obj, int *dest_id, int *dest_type) -- cgit v1.2.3-59-g8ed1b From 719598c98d1961e78e2ad514a2cc15deb5e41db5 Mon Sep 17 00:00:00 2001 From: Yishai Hadas Date: Mon, 26 Nov 2018 08:28:37 +0200 Subject: IB/mlx5: Update the supported DEVX commands Update the supported DEVX commands, it includes adding to the query/modify command's list and to the encoding handling. In addition, a valid range for general commands was added to be used for future commands. Signed-off-by: Yishai Hadas Reviewed-by: Artemy Kovalyov Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx5/devx.c | 17 +++++++++++++++++ include/linux/mlx5/mlx5_ifc.h | 10 ++++++++++ 2 files changed, 27 insertions(+) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/devx.c b/drivers/infiniband/hw/mlx5/devx.c index 80053324dd31..5271469aad10 100644 --- a/drivers/infiniband/hw/mlx5/devx.c +++ b/drivers/infiniband/hw/mlx5/devx.c @@ -314,6 +314,8 @@ static u64 devx_get_obj_id(const void *in) MLX5_GET(query_dct_in, in, dctn)); break; case MLX5_CMD_OP_QUERY_XRQ: + case MLX5_CMD_OP_QUERY_XRQ_DC_PARAMS_ENTRY: + case MLX5_CMD_OP_QUERY_XRQ_ERROR_PARAMS: obj_id = get_enc_obj_id(MLX5_CMD_OP_CREATE_XRQ, MLX5_GET(query_xrq_in, in, xrqn)); break; @@ -340,9 +342,16 @@ static u64 devx_get_obj_id(const void *in) MLX5_GET(drain_dct_in, in, dctn)); break; case MLX5_CMD_OP_ARM_XRQ: + case MLX5_CMD_OP_SET_XRQ_DC_PARAMS_ENTRY: obj_id = get_enc_obj_id(MLX5_CMD_OP_CREATE_XRQ, MLX5_GET(arm_xrq_in, in, xrqn)); break; + case MLX5_CMD_OP_QUERY_PACKET_REFORMAT_CONTEXT: + obj_id = get_enc_obj_id + (MLX5_CMD_OP_ALLOC_PACKET_REFORMAT_CONTEXT, + MLX5_GET(query_packet_reformat_context_in, + in, packet_reformat_id)); + break; default: obj_id = 0; } @@ -601,6 +610,7 @@ static bool devx_is_obj_modify_cmd(const void *in) case MLX5_CMD_OP_DRAIN_DCT: case MLX5_CMD_OP_ARM_DCT_FOR_KEY_VIOLATION: case MLX5_CMD_OP_ARM_XRQ: + case MLX5_CMD_OP_SET_XRQ_DC_PARAMS_ENTRY: return true; case MLX5_CMD_OP_SET_FLOW_TABLE_ENTRY: { @@ -642,6 +652,9 @@ static bool devx_is_obj_query_cmd(const void *in) case MLX5_CMD_OP_QUERY_XRC_SRQ: case MLX5_CMD_OP_QUERY_DCT: case MLX5_CMD_OP_QUERY_XRQ: + case MLX5_CMD_OP_QUERY_XRQ_DC_PARAMS_ENTRY: + case MLX5_CMD_OP_QUERY_XRQ_ERROR_PARAMS: + case MLX5_CMD_OP_QUERY_PACKET_REFORMAT_CONTEXT: return true; default: return false; @@ -685,6 +698,10 @@ static bool devx_is_general_cmd(void *in) { u16 opcode = MLX5_GET(general_obj_in_cmd_hdr, in, opcode); + if (opcode >= MLX5_CMD_OP_GENERAL_START && + opcode < MLX5_CMD_OP_GENERAL_END) + return true; + switch (opcode) { case MLX5_CMD_OP_QUERY_HCA_CAP: case MLX5_CMD_OP_QUERY_HCA_VPORT_CONTEXT: diff --git a/include/linux/mlx5/mlx5_ifc.h b/include/linux/mlx5/mlx5_ifc.h index ece1b606c909..171d68663640 100644 --- a/include/linux/mlx5/mlx5_ifc.h +++ b/include/linux/mlx5/mlx5_ifc.h @@ -144,6 +144,9 @@ enum { MLX5_CMD_OP_DESTROY_XRQ = 0x718, MLX5_CMD_OP_QUERY_XRQ = 0x719, MLX5_CMD_OP_ARM_XRQ = 0x71a, + MLX5_CMD_OP_QUERY_XRQ_DC_PARAMS_ENTRY = 0x725, + MLX5_CMD_OP_SET_XRQ_DC_PARAMS_ENTRY = 0x726, + MLX5_CMD_OP_QUERY_XRQ_ERROR_PARAMS = 0x727, MLX5_CMD_OP_QUERY_VPORT_STATE = 0x750, MLX5_CMD_OP_MODIFY_VPORT_STATE = 0x751, MLX5_CMD_OP_QUERY_ESW_VPORT_CONTEXT = 0x752, @@ -245,6 +248,7 @@ enum { MLX5_CMD_OP_MODIFY_FLOW_TABLE = 0x93c, MLX5_CMD_OP_ALLOC_PACKET_REFORMAT_CONTEXT = 0x93d, MLX5_CMD_OP_DEALLOC_PACKET_REFORMAT_CONTEXT = 0x93e, + MLX5_CMD_OP_QUERY_PACKET_REFORMAT_CONTEXT = 0x93f, MLX5_CMD_OP_ALLOC_MODIFY_HEADER_CONTEXT = 0x940, MLX5_CMD_OP_DEALLOC_MODIFY_HEADER_CONTEXT = 0x941, MLX5_CMD_OP_QUERY_MODIFY_HEADER_CONTEXT = 0x942, @@ -260,6 +264,12 @@ enum { MLX5_CMD_OP_MAX }; +/* Valid range for general commands that don't work over an object */ +enum { + MLX5_CMD_OP_GENERAL_START = 0xb00, + MLX5_CMD_OP_GENERAL_END = 0xd00, +}; + struct mlx5_ifc_flow_table_fields_supported_bits { u8 outer_dmac[0x1]; u8 outer_smac[0x1]; -- cgit v1.2.3-59-g8ed1b From 5aa3771ded54894ce34f4ec6bc2bb403e6771eb2 Mon Sep 17 00:00:00 2001 From: Yishai Hadas Date: Mon, 26 Nov 2018 08:28:38 +0200 Subject: IB/mlx5: Allow XRC usage via verbs in DEVX context Allows XRC usage from the verbs flow in a DEVX context. As XRCD is some shared kernel resource between processes it should be created with UID=0 to point on that. As a result once XRC QP/SRQ are created they must be used as well with UID=0 so that firmware will allow the XRCD usage. Signed-off-by: Yishai Hadas Reviewed-by: Artemy Kovalyov Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx5/mlx5_ib.h | 1 - drivers/infiniband/hw/mlx5/qp.c | 12 +++++------- drivers/infiniband/hw/mlx5/srq.c | 2 +- 3 files changed, 6 insertions(+), 9 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index 4d33965369cc..24cb2f793210 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -543,7 +543,6 @@ struct mlx5_ib_srq { struct mlx5_ib_xrcd { struct ib_xrcd ibxrcd; u32 xrcdn; - u16 uid; }; enum mlx5_ib_mtt_access_flags { diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c index 67adc09c6fa7..48f2e1fbeff8 100644 --- a/drivers/infiniband/hw/mlx5/qp.c +++ b/drivers/infiniband/hw/mlx5/qp.c @@ -775,6 +775,7 @@ static int create_user_qp(struct mlx5_ib_dev *dev, struct ib_pd *pd, __be64 *pas; void *qpc; int err; + u16 uid; err = ib_copy_from_udata(&ucmd, udata, sizeof(ucmd)); if (err) { @@ -836,7 +837,8 @@ static int create_user_qp(struct mlx5_ib_dev *dev, struct ib_pd *pd, goto err_umem; } - MLX5_SET(create_qp_in, *in, uid, to_mpd(pd)->uid); + uid = (attr->qp_type != IB_QPT_XRC_TGT) ? to_mpd(pd)->uid : 0; + MLX5_SET(create_qp_in, *in, uid, uid); pas = (__be64 *)MLX5_ADDR_OF(create_qp_in, *in, pas); if (ubuffer->umem) mlx5_ib_populate_pas(dev, ubuffer->umem, page_shift, pas, 0); @@ -5514,7 +5516,6 @@ struct ib_xrcd *mlx5_ib_alloc_xrcd(struct ib_device *ibdev, struct mlx5_ib_dev *dev = to_mdev(ibdev); struct mlx5_ib_xrcd *xrcd; int err; - u16 uid; if (!MLX5_CAP_GEN(dev->mdev, xrc)) return ERR_PTR(-ENOSYS); @@ -5523,14 +5524,12 @@ struct ib_xrcd *mlx5_ib_alloc_xrcd(struct ib_device *ibdev, if (!xrcd) return ERR_PTR(-ENOMEM); - uid = context ? to_mucontext(context)->devx_uid : 0; - err = mlx5_cmd_xrcd_alloc(dev->mdev, &xrcd->xrcdn, uid); + err = mlx5_cmd_xrcd_alloc(dev->mdev, &xrcd->xrcdn, 0); if (err) { kfree(xrcd); return ERR_PTR(-ENOMEM); } - xrcd->uid = uid; return &xrcd->ibxrcd; } @@ -5538,10 +5537,9 @@ int mlx5_ib_dealloc_xrcd(struct ib_xrcd *xrcd) { struct mlx5_ib_dev *dev = to_mdev(xrcd->device); u32 xrcdn = to_mxrcd(xrcd)->xrcdn; - u16 uid = to_mxrcd(xrcd)->uid; int err; - err = mlx5_cmd_xrcd_dealloc(dev->mdev, xrcdn, uid); + err = mlx5_cmd_xrcd_dealloc(dev->mdev, xrcdn, 0); if (err) mlx5_ib_warn(dev, "failed to dealloc xrcdn 0x%x\n", xrcdn); diff --git a/drivers/infiniband/hw/mlx5/srq.c b/drivers/infiniband/hw/mlx5/srq.c index b3aef0eb39cb..0413b10dea71 100644 --- a/drivers/infiniband/hw/mlx5/srq.c +++ b/drivers/infiniband/hw/mlx5/srq.c @@ -113,7 +113,7 @@ static int create_srq_user(struct ib_pd *pd, struct mlx5_ib_srq *srq, in->log_page_size = page_shift - MLX5_ADAPTER_PAGE_SHIFT; in->page_offset = offset; - in->uid = to_mpd(pd)->uid; + in->uid = (in->type != IB_SRQT_XRC) ? to_mpd(pd)->uid : 0; if (MLX5_CAP_GEN(dev->mdev, cqe_version) == MLX5_CQE_VERSION_V1 && in->type != IB_SRQT_BASIC) in->user_index = uidx; -- cgit v1.2.3-59-g8ed1b From d16da11992d44fb54d99b71408b201311d81e0af Mon Sep 17 00:00:00 2001 From: Lijun Ou Date: Sat, 24 Nov 2018 16:49:19 +0800 Subject: RDMA/hns: Eanble SRQ capacity for hip08 This patch configures the flags for enabling the SRQ(Share Receive Queue) capacity as well as update the verb of querying device for setting srq specifications. Signed-off-by: Lijun Ou Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hns/hns_roce_device.h | 4 ++++ drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 7 ++++++- drivers/infiniband/hw/hns/hns_roce_hw_v2.h | 3 +++ drivers/infiniband/hw/hns/hns_roce_main.c | 5 +++++ 4 files changed, 18 insertions(+), 1 deletion(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/hns/hns_roce_device.h b/drivers/infiniband/hw/hns/hns_roce_device.h index d39bdfdb5de9..42ff400248c3 100644 --- a/drivers/infiniband/hw/hns/hns_roce_device.h +++ b/drivers/infiniband/hw/hns/hns_roce_device.h @@ -196,6 +196,7 @@ enum { HNS_ROCE_CAP_FLAG_RQ_INLINE = BIT(2), HNS_ROCE_CAP_FLAG_RECORD_DB = BIT(3), HNS_ROCE_CAP_FLAG_SQ_RECORD_DB = BIT(4), + HNS_ROCE_CAP_FLAG_SRQ = BIT(5), HNS_ROCE_CAP_FLAG_MW = BIT(7), HNS_ROCE_CAP_FLAG_FRMR = BIT(8), HNS_ROCE_CAP_FLAG_ATOMIC = BIT(10), @@ -680,6 +681,9 @@ struct hns_roce_caps { int num_qps; /* 256k */ int reserved_qps; u32 max_wqes; /* 16k */ + u32 max_srqs; + u32 max_srq_wrs; + u32 max_srq_sges; u32 max_sq_desc_sz; /* 64 */ u32 max_rq_desc_sz; /* 64 */ u32 max_srq_desc_sz; diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c index d4609e7eb3c0..2fec1f2f1552 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c @@ -1354,8 +1354,13 @@ static int hns_roce_v2_profile(struct hns_roce_dev *hr_dev) caps->local_ca_ack_delay = 0; caps->max_mtu = IB_MTU_4096; + caps->max_srqs = HNS_ROCE_V2_MAX_SRQ; + caps->max_srq_wrs = HNS_ROCE_V2_MAX_SRQ_WR; + caps->max_srq_sges = HNS_ROCE_V2_MAX_SRQ_SGE; + if (hr_dev->pci_dev->revision == 0x21) - caps->flags |= HNS_ROCE_CAP_FLAG_ATOMIC; + caps->flags |= HNS_ROCE_CAP_FLAG_ATOMIC | + HNS_ROCE_CAP_FLAG_SRQ; ret = hns_roce_v2_set_bt(hr_dev); if (ret) diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.h b/drivers/infiniband/hw/hns/hns_roce_hw_v2.h index 8bc820635bbd..24a485103397 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.h +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.h @@ -46,6 +46,9 @@ #define HNS_ROCE_V2_MAX_QP_NUM 0x2000 #define HNS_ROCE_V2_MAX_WQE_NUM 0x8000 +#define HNS_ROCE_V2_MAX_SRQ 0x100000 +#define HNS_ROCE_V2_MAX_SRQ_WR 0x8000 +#define HNS_ROCE_V2_MAX_SRQ_SGE 0x100 #define HNS_ROCE_V2_MAX_CQ_NUM 0x8000 #define HNS_ROCE_V2_MAX_CQE_NUM 0x10000 #define HNS_ROCE_V2_MAX_RQ_SGE_NUM 0x100 diff --git a/drivers/infiniband/hw/hns/hns_roce_main.c b/drivers/infiniband/hw/hns/hns_roce_main.c index 1b3ee514f2ef..f6f288f36e04 100644 --- a/drivers/infiniband/hw/hns/hns_roce_main.c +++ b/drivers/infiniband/hw/hns/hns_roce_main.c @@ -220,6 +220,11 @@ static int hns_roce_query_device(struct ib_device *ib_dev, IB_ATOMIC_HCA : IB_ATOMIC_NONE; props->max_pkeys = 1; props->local_ca_ack_delay = hr_dev->caps.local_ca_ack_delay; + if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_SRQ) { + props->max_srq = hr_dev->caps.max_srqs; + props->max_srq_wr = hr_dev->caps.max_srq_wrs; + props->max_srq_sge = hr_dev->caps.max_srq_sges; + } return 0; } -- cgit v1.2.3-59-g8ed1b From 5c1f167af112746eb97621f12b4053d4f66c0357 Mon Sep 17 00:00:00 2001 From: Lijun Ou Date: Sat, 24 Nov 2018 16:49:20 +0800 Subject: RDMA/hns: Init SRQ table for hip08 This patch inits hem resource for SRQ table, includes SRQWQE and SRQWQE index resource. Signed-off-by: Lijun Ou Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hns/Makefile | 2 +- drivers/infiniband/hw/hns/hns_roce_alloc.c | 2 + drivers/infiniband/hw/hns/hns_roce_device.h | 23 +++++++++++ drivers/infiniband/hw/hns/hns_roce_hem.c | 9 +++++ drivers/infiniband/hw/hns/hns_roce_hem.h | 2 + drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 8 ++++ drivers/infiniband/hw/hns/hns_roce_hw_v2.h | 6 +++ drivers/infiniband/hw/hns/hns_roce_main.c | 63 +++++++++++++++++++++++++++++ drivers/infiniband/hw/hns/hns_roce_srq.c | 26 ++++++++++++ 9 files changed, 140 insertions(+), 1 deletion(-) create mode 100644 drivers/infiniband/hw/hns/hns_roce_srq.c (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/hns/Makefile b/drivers/infiniband/hw/hns/Makefile index cf03404b9d58..004c88b32e13 100644 --- a/drivers/infiniband/hw/hns/Makefile +++ b/drivers/infiniband/hw/hns/Makefile @@ -7,7 +7,7 @@ ccflags-y := -Idrivers/net/ethernet/hisilicon/hns3 obj-$(CONFIG_INFINIBAND_HNS) += hns-roce.o hns-roce-objs := hns_roce_main.o hns_roce_cmd.o hns_roce_pd.o \ hns_roce_ah.o hns_roce_hem.o hns_roce_mr.o hns_roce_qp.o \ - hns_roce_cq.o hns_roce_alloc.o hns_roce_db.o + hns_roce_cq.o hns_roce_alloc.o hns_roce_db.o hns_roce_srq.o obj-$(CONFIG_INFINIBAND_HNS_HIP06) += hns-roce-hw-v1.o hns-roce-hw-v1-objs := hns_roce_hw_v1.o obj-$(CONFIG_INFINIBAND_HNS_HIP08) += hns-roce-hw-v2.o diff --git a/drivers/infiniband/hw/hns/hns_roce_alloc.c b/drivers/infiniband/hw/hns/hns_roce_alloc.c index 46f65f9f59d0..6300033a448f 100644 --- a/drivers/infiniband/hw/hns/hns_roce_alloc.c +++ b/drivers/infiniband/hw/hns/hns_roce_alloc.c @@ -239,6 +239,8 @@ err_free: void hns_roce_cleanup_bitmap(struct hns_roce_dev *hr_dev) { + if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_SRQ) + hns_roce_cleanup_srq_table(hr_dev); hns_roce_cleanup_qp_table(hr_dev); hns_roce_cleanup_cq_table(hr_dev); hns_roce_cleanup_mr_table(hr_dev); diff --git a/drivers/infiniband/hw/hns/hns_roce_device.h b/drivers/infiniband/hw/hns/hns_roce_device.h index 42ff400248c3..5d79d8070a62 100644 --- a/drivers/infiniband/hw/hns/hns_roce_device.h +++ b/drivers/infiniband/hw/hns/hns_roce_device.h @@ -205,6 +205,8 @@ enum { enum hns_roce_mtt_type { MTT_TYPE_WQE, MTT_TYPE_CQE, + MTT_TYPE_SRQWQE, + MTT_TYPE_IDX }; enum { @@ -340,6 +342,10 @@ struct hns_roce_mr_table { struct hns_roce_hem_table mtpt_table; struct hns_roce_buddy mtt_cqe_buddy; struct hns_roce_hem_table mtt_cqe_table; + struct hns_roce_buddy mtt_srqwqe_buddy; + struct hns_roce_hem_table mtt_srqwqe_table; + struct hns_roce_buddy mtt_idx_buddy; + struct hns_roce_hem_table mtt_idx_table; }; struct hns_roce_wq { @@ -454,6 +460,12 @@ struct hns_roce_cq_table { struct hns_roce_hem_table table; }; +struct hns_roce_srq_table { + struct hns_roce_bitmap bitmap; + struct xarray xa; + struct hns_roce_hem_table table; +}; + struct hns_roce_raq_table { struct hns_roce_buf_list *e_raq_buf; }; @@ -680,6 +692,8 @@ struct hns_roce_caps { u32 max_extend_sg; int num_qps; /* 256k */ int reserved_qps; + u32 max_srq_sg; + int num_srqs; u32 max_wqes; /* 16k */ u32 max_srqs; u32 max_srq_wrs; @@ -694,12 +708,16 @@ struct hns_roce_caps { int min_cqes; u32 min_wqes; int reserved_cqs; + int reserved_srqs; + u32 max_srqwqes; int num_aeq_vectors; /* 1 */ int num_comp_vectors; int num_other_vectors; int num_mtpts; u32 num_mtt_segs; u32 num_cqe_segs; + u32 num_srqwqe_segs; + u32 num_idx_segs; int reserved_mrws; int reserved_uars; int num_pds; @@ -713,6 +731,8 @@ struct hns_roce_caps { int irrl_entry_sz; int trrl_entry_sz; int cqc_entry_sz; + int srqc_entry_sz; + int idx_entry_sz; u32 pbl_ba_pg_sz; u32 pbl_buf_pg_sz; u32 pbl_hop_num; @@ -843,6 +863,7 @@ struct hns_roce_dev { struct hns_roce_uar_table uar_table; struct hns_roce_mr_table mr_table; struct hns_roce_cq_table cq_table; + struct hns_roce_srq_table srq_table; struct hns_roce_qp_table qp_table; struct hns_roce_eq_table eq_table; @@ -955,12 +976,14 @@ int hns_roce_init_mr_table(struct hns_roce_dev *hr_dev); int hns_roce_init_eq_table(struct hns_roce_dev *hr_dev); int hns_roce_init_cq_table(struct hns_roce_dev *hr_dev); int hns_roce_init_qp_table(struct hns_roce_dev *hr_dev); +int hns_roce_init_srq_table(struct hns_roce_dev *hr_dev); void hns_roce_cleanup_pd_table(struct hns_roce_dev *hr_dev); void hns_roce_cleanup_mr_table(struct hns_roce_dev *hr_dev); void hns_roce_cleanup_eq_table(struct hns_roce_dev *hr_dev); void hns_roce_cleanup_cq_table(struct hns_roce_dev *hr_dev); void hns_roce_cleanup_qp_table(struct hns_roce_dev *hr_dev); +void hns_roce_cleanup_srq_table(struct hns_roce_dev *hr_dev); int hns_roce_bitmap_alloc(struct hns_roce_bitmap *bitmap, unsigned long *obj); void hns_roce_bitmap_free(struct hns_roce_bitmap *bitmap, unsigned long obj, diff --git a/drivers/infiniband/hw/hns/hns_roce_hem.c b/drivers/infiniband/hw/hns/hns_roce_hem.c index f6faefed96e8..388b0406e41f 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hem.c +++ b/drivers/infiniband/hw/hns/hns_roce_hem.c @@ -1041,6 +1041,15 @@ void hns_roce_cleanup_hem_table(struct hns_roce_dev *hr_dev, void hns_roce_cleanup_hem(struct hns_roce_dev *hr_dev) { + if ((hr_dev->caps.num_idx_segs)) + hns_roce_cleanup_hem_table(hr_dev, + &hr_dev->mr_table.mtt_idx_table); + if (hr_dev->caps.num_srqwqe_segs) + hns_roce_cleanup_hem_table(hr_dev, + &hr_dev->mr_table.mtt_srqwqe_table); + if (hr_dev->caps.srqc_entry_sz) + hns_roce_cleanup_hem_table(hr_dev, + &hr_dev->srq_table.table); hns_roce_cleanup_hem_table(hr_dev, &hr_dev->cq_table.table); if (hr_dev->caps.trrl_entry_sz) hns_roce_cleanup_hem_table(hr_dev, diff --git a/drivers/infiniband/hw/hns/hns_roce_hem.h b/drivers/infiniband/hw/hns/hns_roce_hem.h index e8850d59e780..a650278c6fbd 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hem.h +++ b/drivers/infiniband/hw/hns/hns_roce_hem.h @@ -48,6 +48,8 @@ enum { /* UNMAP HEM */ HEM_TYPE_MTT, HEM_TYPE_CQE, + HEM_TYPE_SRQWQE, + HEM_TYPE_IDX, HEM_TYPE_IRRL, HEM_TYPE_TRRL, }; diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c index 2fec1f2f1552..215f5ca377cc 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c @@ -1276,11 +1276,14 @@ static int hns_roce_v2_profile(struct hns_roce_dev *hr_dev) caps->num_qps = HNS_ROCE_V2_MAX_QP_NUM; caps->max_wqes = HNS_ROCE_V2_MAX_WQE_NUM; caps->num_cqs = HNS_ROCE_V2_MAX_CQ_NUM; + caps->num_srqs = HNS_ROCE_V2_MAX_SRQ_NUM; caps->max_cqes = HNS_ROCE_V2_MAX_CQE_NUM; + caps->max_srqwqes = HNS_ROCE_V2_MAX_SRQWQE_NUM; caps->max_sq_sg = HNS_ROCE_V2_MAX_SQ_SGE_NUM; caps->max_extend_sg = HNS_ROCE_V2_MAX_EXTEND_SGE_NUM; caps->max_rq_sg = HNS_ROCE_V2_MAX_RQ_SGE_NUM; caps->max_sq_inline = HNS_ROCE_V2_MAX_SQ_INLINE; + caps->max_srq_sg = HNS_ROCE_V2_MAX_SRQ_SGE_NUM; caps->num_uars = HNS_ROCE_V2_UAR_NUM; caps->phy_num_uars = HNS_ROCE_V2_PHY_UAR_NUM; caps->num_aeq_vectors = HNS_ROCE_V2_AEQE_VEC_NUM; @@ -1289,6 +1292,8 @@ static int hns_roce_v2_profile(struct hns_roce_dev *hr_dev) caps->num_mtpts = HNS_ROCE_V2_MAX_MTPT_NUM; caps->num_mtt_segs = HNS_ROCE_V2_MAX_MTT_SEGS; caps->num_cqe_segs = HNS_ROCE_V2_MAX_CQE_SEGS; + caps->num_srqwqe_segs = HNS_ROCE_V2_MAX_SRQWQE_SEGS; + caps->num_idx_segs = HNS_ROCE_V2_MAX_IDX_SEGS; caps->num_pds = HNS_ROCE_V2_MAX_PD_NUM; caps->max_qp_init_rdma = HNS_ROCE_V2_MAX_QP_INIT_RDMA; caps->max_qp_dest_rdma = HNS_ROCE_V2_MAX_QP_DEST_RDMA; @@ -1299,8 +1304,10 @@ static int hns_roce_v2_profile(struct hns_roce_dev *hr_dev) caps->irrl_entry_sz = HNS_ROCE_V2_IRRL_ENTRY_SZ; caps->trrl_entry_sz = HNS_ROCE_V2_TRRL_ENTRY_SZ; caps->cqc_entry_sz = HNS_ROCE_V2_CQC_ENTRY_SZ; + caps->srqc_entry_sz = HNS_ROCE_V2_SRQC_ENTRY_SZ; caps->mtpt_entry_sz = HNS_ROCE_V2_MTPT_ENTRY_SZ; caps->mtt_entry_sz = HNS_ROCE_V2_MTT_ENTRY_SZ; + caps->idx_entry_sz = 4; caps->cq_entry_sz = HNS_ROCE_V2_CQE_ENTRY_SIZE; caps->page_size_cap = HNS_ROCE_V2_PAGE_SIZE_SUPPORTED; caps->reserved_lkey = 0; @@ -1308,6 +1315,7 @@ static int hns_roce_v2_profile(struct hns_roce_dev *hr_dev) caps->reserved_mrws = 1; caps->reserved_uars = 0; caps->reserved_cqs = 0; + caps->reserved_srqs = 0; caps->reserved_qps = HNS_ROCE_V2_RSV_QPS; caps->qpc_ba_pg_sz = 0; diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.h b/drivers/infiniband/hw/hns/hns_roce_hw_v2.h index 24a485103397..383bab5e75ef 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.h +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.h @@ -50,9 +50,12 @@ #define HNS_ROCE_V2_MAX_SRQ_WR 0x8000 #define HNS_ROCE_V2_MAX_SRQ_SGE 0x100 #define HNS_ROCE_V2_MAX_CQ_NUM 0x8000 +#define HNS_ROCE_V2_MAX_SRQ_NUM 0x100000 #define HNS_ROCE_V2_MAX_CQE_NUM 0x10000 +#define HNS_ROCE_V2_MAX_SRQWQE_NUM 0x8000 #define HNS_ROCE_V2_MAX_RQ_SGE_NUM 0x100 #define HNS_ROCE_V2_MAX_SQ_SGE_NUM 0xff +#define HNS_ROCE_V2_MAX_SRQ_SGE_NUM 0x100 #define HNS_ROCE_V2_MAX_EXTEND_SGE_NUM 0x200000 #define HNS_ROCE_V2_MAX_SQ_INLINE 0x20 #define HNS_ROCE_V2_UAR_NUM 256 @@ -64,6 +67,8 @@ #define HNS_ROCE_V2_MAX_MTPT_NUM 0x8000 #define HNS_ROCE_V2_MAX_MTT_SEGS 0x1000000 #define HNS_ROCE_V2_MAX_CQE_SEGS 0x1000000 +#define HNS_ROCE_V2_MAX_SRQWQE_SEGS 0x1000000 +#define HNS_ROCE_V2_MAX_IDX_SEGS 0x1000000 #define HNS_ROCE_V2_MAX_PD_NUM 0x1000000 #define HNS_ROCE_V2_MAX_QP_INIT_RDMA 128 #define HNS_ROCE_V2_MAX_QP_DEST_RDMA 128 @@ -74,6 +79,7 @@ #define HNS_ROCE_V2_IRRL_ENTRY_SZ 64 #define HNS_ROCE_V2_TRRL_ENTRY_SZ 48 #define HNS_ROCE_V2_CQC_ENTRY_SZ 64 +#define HNS_ROCE_V2_SRQC_ENTRY_SZ 64 #define HNS_ROCE_V2_MTPT_ENTRY_SZ 64 #define HNS_ROCE_V2_MTT_ENTRY_SZ 64 #define HNS_ROCE_V2_CQE_ENTRY_SIZE 32 diff --git a/drivers/infiniband/hw/hns/hns_roce_main.c b/drivers/infiniband/hw/hns/hns_roce_main.c index f6f288f36e04..d183f13d206f 100644 --- a/drivers/infiniband/hw/hns/hns_roce_main.c +++ b/drivers/infiniband/hw/hns/hns_roce_main.c @@ -651,8 +651,58 @@ static int hns_roce_init_hem(struct hns_roce_dev *hr_dev) goto err_unmap_trrl; } + if (hr_dev->caps.srqc_entry_sz) { + ret = hns_roce_init_hem_table(hr_dev, &hr_dev->srq_table.table, + HEM_TYPE_SRQC, + hr_dev->caps.srqc_entry_sz, + hr_dev->caps.num_srqs, 1); + if (ret) { + dev_err(dev, + "Failed to init SRQ context memory, aborting.\n"); + goto err_unmap_cq; + } + } + + if (hr_dev->caps.num_srqwqe_segs) { + ret = hns_roce_init_hem_table(hr_dev, + &hr_dev->mr_table.mtt_srqwqe_table, + HEM_TYPE_SRQWQE, + hr_dev->caps.mtt_entry_sz, + hr_dev->caps.num_srqwqe_segs, 1); + if (ret) { + dev_err(dev, + "Failed to init MTT srqwqe memory, aborting.\n"); + goto err_unmap_srq; + } + } + + if (hr_dev->caps.num_idx_segs) { + ret = hns_roce_init_hem_table(hr_dev, + &hr_dev->mr_table.mtt_idx_table, + HEM_TYPE_IDX, + hr_dev->caps.idx_entry_sz, + hr_dev->caps.num_idx_segs, 1); + if (ret) { + dev_err(dev, + "Failed to init MTT idx memory, aborting.\n"); + goto err_unmap_srqwqe; + } + } + return 0; +err_unmap_srqwqe: + if (hr_dev->caps.num_srqwqe_segs) + hns_roce_cleanup_hem_table(hr_dev, + &hr_dev->mr_table.mtt_srqwqe_table); + +err_unmap_srq: + if (hr_dev->caps.srqc_entry_sz) + hns_roce_cleanup_hem_table(hr_dev, &hr_dev->srq_table.table); + +err_unmap_cq: + hns_roce_cleanup_hem_table(hr_dev, &hr_dev->cq_table.table); + err_unmap_trrl: if (hr_dev->caps.trrl_entry_sz) hns_roce_cleanup_hem_table(hr_dev, @@ -732,8 +782,21 @@ static int hns_roce_setup_hca(struct hns_roce_dev *hr_dev) goto err_cq_table_free; } + if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_SRQ) { + ret = hns_roce_init_srq_table(hr_dev); + if (ret) { + dev_err(dev, + "Failed to init share receive queue table.\n"); + goto err_qp_table_free; + } + } + return 0; +err_qp_table_free: + if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_SRQ) + hns_roce_cleanup_qp_table(hr_dev); + err_cq_table_free: hns_roce_cleanup_cq_table(hr_dev); diff --git a/drivers/infiniband/hw/hns/hns_roce_srq.c b/drivers/infiniband/hw/hns/hns_roce_srq.c new file mode 100644 index 000000000000..d8a86130febe --- /dev/null +++ b/drivers/infiniband/hw/hns/hns_roce_srq.c @@ -0,0 +1,26 @@ +// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB +/* + * Copyright (c) 2018 Hisilicon Limited. + */ + +#include +#include +#include "hns_roce_device.h" +#include "hns_roce_cmd.h" +#include "hns_roce_hem.h" + +int hns_roce_init_srq_table(struct hns_roce_dev *hr_dev) +{ + struct hns_roce_srq_table *srq_table = &hr_dev->srq_table; + + xa_init(&srq_table->xa); + + return hns_roce_bitmap_init(&srq_table->bitmap, hr_dev->caps.num_srqs, + hr_dev->caps.num_srqs - 1, + hr_dev->caps.reserved_srqs, 0); +} + +void hns_roce_cleanup_srq_table(struct hns_roce_dev *hr_dev) +{ + hns_roce_bitmap_cleanup(&hr_dev->srq_table.bitmap); +} -- cgit v1.2.3-59-g8ed1b From c7bcb13442e1bd8a4cc41c74540dd2e2f7058d16 Mon Sep 17 00:00:00 2001 From: Lijun Ou Date: Sat, 24 Nov 2018 16:49:21 +0800 Subject: RDMA/hns: Add SRQ support for hip08 kernel mode This patch implements the SRQ(Share Receive Queue) verbs and update the poll cq verbs to deal with SRQ complentions. Signed-off-by: Lijun Ou Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hns/hns_roce_cmd.h | 4 + drivers/infiniband/hw/hns/hns_roce_device.h | 58 +++- drivers/infiniband/hw/hns/hns_roce_hem.c | 32 ++- drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 384 ++++++++++++++++++++++++-- drivers/infiniband/hw/hns/hns_roce_hw_v2.h | 86 ++++++ drivers/infiniband/hw/hns/hns_roce_main.c | 15 + drivers/infiniband/hw/hns/hns_roce_mr.c | 137 +++++++++- drivers/infiniband/hw/hns/hns_roce_qp.c | 21 +- drivers/infiniband/hw/hns/hns_roce_srq.c | 407 ++++++++++++++++++++++++++++ include/uapi/rdma/hns-abi.h | 6 + 10 files changed, 1098 insertions(+), 52 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/hns/hns_roce_cmd.h b/drivers/infiniband/hw/hns/hns_roce_cmd.h index 9549ae51a0dd..927701df5eff 100644 --- a/drivers/infiniband/hw/hns/hns_roce_cmd.h +++ b/drivers/infiniband/hw/hns/hns_roce_cmd.h @@ -120,6 +120,10 @@ enum { HNS_ROCE_CMD_SQD2RTS_QP = 0x20, HNS_ROCE_CMD_2RST_QP = 0x21, HNS_ROCE_CMD_QUERY_QP = 0x22, + HNS_ROCE_CMD_SW2HW_SRQ = 0x70, + HNS_ROCE_CMD_MODIFY_SRQC = 0x72, + HNS_ROCE_CMD_QUERY_SRQC = 0x73, + HNS_ROCE_CMD_HW2SW_SRQ = 0x74, }; int hns_roce_cmd_mbox(struct hns_roce_dev *hr_dev, u64 in_param, u64 out_param, diff --git a/drivers/infiniband/hw/hns/hns_roce_device.h b/drivers/infiniband/hw/hns/hns_roce_device.h index 5d79d8070a62..5a40746cb2e3 100644 --- a/drivers/infiniband/hw/hns/hns_roce_device.h +++ b/drivers/infiniband/hw/hns/hns_roce_device.h @@ -111,6 +111,9 @@ #define PAGES_SHIFT_24 24 #define PAGES_SHIFT_32 32 +#define HNS_ROCE_IDX_QUE_ENTRY_SZ 4 +#define SRQ_DB_REG 0x230 + enum { HNS_ROCE_SUPPORT_RQ_RECORD_DB = 1 << 0, HNS_ROCE_SUPPORT_SQ_RECORD_DB = 1 << 1, @@ -436,9 +439,37 @@ struct hns_roce_cq { struct completion free; }; +struct hns_roce_idx_que { + struct hns_roce_buf idx_buf; + int entry_sz; + u32 buf_size; + struct ib_umem *umem; + struct hns_roce_mtt mtt; + u64 *bitmap; +}; + struct hns_roce_srq { struct ib_srq ibsrq; - int srqn; + void (*event)(struct hns_roce_srq *srq, enum hns_roce_event event); + unsigned long srqn; + int max; + int max_gs; + int wqe_shift; + void __iomem *db_reg_l; + + atomic_t refcount; + struct completion free; + + struct hns_roce_buf buf; + u64 *wrid; + struct ib_umem *umem; + struct hns_roce_mtt mtt; + struct hns_roce_idx_que idx_que; + spinlock_t lock; + int head; + int tail; + u16 wqe_ctr; + struct mutex mutex; }; struct hns_roce_uar_table { @@ -761,6 +792,12 @@ struct hns_roce_caps { u32 cqe_ba_pg_sz; u32 cqe_buf_pg_sz; u32 cqe_hop_num; + u32 srqwqe_ba_pg_sz; + u32 srqwqe_buf_pg_sz; + u32 srqwqe_hop_num; + u32 idx_ba_pg_sz; + u32 idx_buf_pg_sz; + u32 idx_hop_num; u32 eqe_ba_pg_sz; u32 eqe_buf_pg_sz; u32 eqe_hop_num; @@ -829,6 +866,17 @@ struct hns_roce_hw { int (*modify_cq)(struct ib_cq *cq, u16 cq_count, u16 cq_period); int (*init_eq)(struct hns_roce_dev *hr_dev); void (*cleanup_eq)(struct hns_roce_dev *hr_dev); + void (*write_srqc)(struct hns_roce_dev *hr_dev, + struct hns_roce_srq *srq, u32 pdn, u16 xrcd, u32 cqn, + void *mb_buf, u64 *mtts_wqe, u64 *mtts_idx, + dma_addr_t dma_handle_wqe, + dma_addr_t dma_handle_idx); + int (*modify_srq)(struct ib_srq *ibsrq, struct ib_srq_attr *srq_attr, + enum ib_srq_attr_mask srq_attr_mask, + struct ib_udata *udata); + int (*query_srq)(struct ib_srq *ibsrq, struct ib_srq_attr *attr); + int (*post_srq_recv)(struct ib_srq *ibsrq, const struct ib_recv_wr *wr, + const struct ib_recv_wr **bad_wr); }; struct hns_roce_dev { @@ -1038,6 +1086,14 @@ int hns_roce_buf_alloc(struct hns_roce_dev *hr_dev, u32 size, u32 max_direct, int hns_roce_ib_umem_write_mtt(struct hns_roce_dev *hr_dev, struct hns_roce_mtt *mtt, struct ib_umem *umem); +struct ib_srq *hns_roce_create_srq(struct ib_pd *pd, + struct ib_srq_init_attr *srq_init_attr, + struct ib_udata *udata); +int hns_roce_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *srq_attr, + enum ib_srq_attr_mask srq_attr_mask, + struct ib_udata *udata); +int hns_roce_destroy_srq(struct ib_srq *ibsrq); + struct ib_qp *hns_roce_create_qp(struct ib_pd *ib_pd, struct ib_qp_init_attr *init_attr, struct ib_udata *udata); diff --git a/drivers/infiniband/hw/hns/hns_roce_hem.c b/drivers/infiniband/hw/hns/hns_roce_hem.c index 388b0406e41f..4cdbcafa5915 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hem.c +++ b/drivers/infiniband/hw/hns/hns_roce_hem.c @@ -46,7 +46,9 @@ bool hns_roce_check_whether_mhop(struct hns_roce_dev *hr_dev, u32 type) (hr_dev->caps.cqc_hop_num && type == HEM_TYPE_CQC) || (hr_dev->caps.srqc_hop_num && type == HEM_TYPE_SRQC) || (hr_dev->caps.cqe_hop_num && type == HEM_TYPE_CQE) || - (hr_dev->caps.mtt_hop_num && type == HEM_TYPE_MTT)) + (hr_dev->caps.mtt_hop_num && type == HEM_TYPE_MTT) || + (hr_dev->caps.srqwqe_hop_num && type == HEM_TYPE_SRQWQE) || + (hr_dev->caps.idx_hop_num && type == HEM_TYPE_IDX)) return true; return false; @@ -147,6 +149,22 @@ int hns_roce_calc_hem_mhop(struct hns_roce_dev *hr_dev, mhop->ba_l0_num = mhop->bt_chunk_size / 8; mhop->hop_num = hr_dev->caps.cqe_hop_num; break; + case HEM_TYPE_SRQWQE: + mhop->buf_chunk_size = 1 << (hr_dev->caps.srqwqe_buf_pg_sz + + PAGE_SHIFT); + mhop->bt_chunk_size = 1 << (hr_dev->caps.srqwqe_ba_pg_sz + + PAGE_SHIFT); + mhop->ba_l0_num = mhop->bt_chunk_size / 8; + mhop->hop_num = hr_dev->caps.srqwqe_hop_num; + break; + case HEM_TYPE_IDX: + mhop->buf_chunk_size = 1 << (hr_dev->caps.idx_buf_pg_sz + + PAGE_SHIFT); + mhop->bt_chunk_size = 1 << (hr_dev->caps.idx_ba_pg_sz + + PAGE_SHIFT); + mhop->ba_l0_num = mhop->bt_chunk_size / 8; + mhop->hop_num = hr_dev->caps.idx_hop_num; + break; default: dev_err(dev, "Table %d not support multi-hop addressing!\n", table->type); @@ -906,6 +924,18 @@ int hns_roce_init_hem_table(struct hns_roce_dev *hr_dev, bt_chunk_size = buf_chunk_size; hop_num = hr_dev->caps.cqe_hop_num; break; + case HEM_TYPE_SRQWQE: + buf_chunk_size = 1 << (hr_dev->caps.srqwqe_ba_pg_sz + + PAGE_SHIFT); + bt_chunk_size = buf_chunk_size; + hop_num = hr_dev->caps.srqwqe_hop_num; + break; + case HEM_TYPE_IDX: + buf_chunk_size = 1 << (hr_dev->caps.idx_ba_pg_sz + + PAGE_SHIFT); + bt_chunk_size = buf_chunk_size; + hop_num = hr_dev->caps.idx_hop_num; + break; default: dev_err(dev, "Table %d not support to init hem table here!\n", diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c index 215f5ca377cc..6c9baf99894e 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c @@ -1339,6 +1339,12 @@ static int hns_roce_v2_profile(struct hns_roce_dev *hr_dev) caps->cqe_ba_pg_sz = 0; caps->cqe_buf_pg_sz = 0; caps->cqe_hop_num = HNS_ROCE_CQE_HOP_NUM; + caps->srqwqe_ba_pg_sz = 0; + caps->srqwqe_buf_pg_sz = 0; + caps->srqwqe_hop_num = HNS_ROCE_SRQWQE_HOP_NUM; + caps->idx_ba_pg_sz = 0; + caps->idx_buf_pg_sz = 0; + caps->idx_hop_num = HNS_ROCE_IDX_HOP_NUM; caps->eqe_ba_pg_sz = 0; caps->eqe_buf_pg_sz = 0; caps->eqe_hop_num = HNS_ROCE_EQE_HOP_NUM; @@ -2028,6 +2034,27 @@ static struct hns_roce_v2_cqe *next_cqe_sw_v2(struct hns_roce_cq *hr_cq) return get_sw_cqe_v2(hr_cq, hr_cq->cons_index); } +static void *get_srq_wqe(struct hns_roce_srq *srq, int n) +{ + return hns_roce_buf_offset(&srq->buf, n << srq->wqe_shift); +} + +static void hns_roce_free_srq_wqe(struct hns_roce_srq *srq, int wqe_index) +{ + u32 bitmap_num; + int bit_num; + + /* always called with interrupts disabled. */ + spin_lock(&srq->lock); + + bitmap_num = wqe_index / (sizeof(u64) * 8); + bit_num = wqe_index % (sizeof(u64) * 8); + srq->idx_que.bitmap[bitmap_num] |= (1ULL << bit_num); + srq->tail++; + + spin_unlock(&srq->lock); +} + static void hns_roce_v2_cq_set_ci(struct hns_roce_cq *hr_cq, u32 cons_index) { *hr_cq->set_ci_db = cons_index & 0xffffff; @@ -2039,6 +2066,7 @@ static void __hns_roce_v2_cq_clean(struct hns_roce_cq *hr_cq, u32 qpn, struct hns_roce_v2_cqe *cqe, *dest; u32 prod_index; int nfreed = 0; + int wqe_index; u8 owner_bit; for (prod_index = hr_cq->cons_index; get_sw_cqe_v2(hr_cq, prod_index); @@ -2056,7 +2084,13 @@ static void __hns_roce_v2_cq_clean(struct hns_roce_cq *hr_cq, u32 qpn, if ((roce_get_field(cqe->byte_16, V2_CQE_BYTE_16_LCL_QPN_M, V2_CQE_BYTE_16_LCL_QPN_S) & HNS_ROCE_V2_CQE_QPN_MASK) == qpn) { - /* In v1 engine, not support SRQ */ + if (srq && + roce_get_bit(cqe->byte_4, V2_CQE_BYTE_4_S_R_S)) { + wqe_index = roce_get_field(cqe->byte_4, + V2_CQE_BYTE_4_WQE_INDX_M, + V2_CQE_BYTE_4_WQE_INDX_S); + hns_roce_free_srq_wqe(srq, wqe_index); + } ++nfreed; } else if (nfreed) { dest = get_cqe_v2(hr_cq, (prod_index + nfreed) & @@ -2233,6 +2267,7 @@ static int hns_roce_handle_recv_inl_wqe(struct hns_roce_v2_cqe *cqe, static int hns_roce_v2_poll_one(struct hns_roce_cq *hr_cq, struct hns_roce_qp **cur_qp, struct ib_wc *wc) { + struct hns_roce_srq *srq = NULL; struct hns_roce_dev *hr_dev; struct hns_roce_v2_cqe *cqe; struct hns_roce_qp *hr_qp; @@ -2275,6 +2310,37 @@ static int hns_roce_v2_poll_one(struct hns_roce_cq *hr_cq, wc->qp = &(*cur_qp)->ibqp; wc->vendor_err = 0; + if (is_send) { + wq = &(*cur_qp)->sq; + if ((*cur_qp)->sq_signal_bits) { + /* + * If sg_signal_bit is 1, + * firstly tail pointer updated to wqe + * which current cqe correspond to + */ + wqe_ctr = (u16)roce_get_field(cqe->byte_4, + V2_CQE_BYTE_4_WQE_INDX_M, + V2_CQE_BYTE_4_WQE_INDX_S); + wq->tail += (wqe_ctr - (u16)wq->tail) & + (wq->wqe_cnt - 1); + } + + wc->wr_id = wq->wrid[wq->tail & (wq->wqe_cnt - 1)]; + ++wq->tail; + } else if ((*cur_qp)->ibqp.srq) { + srq = to_hr_srq((*cur_qp)->ibqp.srq); + wqe_ctr = le16_to_cpu(roce_get_field(cqe->byte_4, + V2_CQE_BYTE_4_WQE_INDX_M, + V2_CQE_BYTE_4_WQE_INDX_S)); + wc->wr_id = srq->wrid[wqe_ctr]; + hns_roce_free_srq_wqe(srq, wqe_ctr); + } else { + /* Update tail pointer, record wr_id */ + wq = &(*cur_qp)->rq; + wc->wr_id = wq->wrid[wq->tail & (wq->wqe_cnt - 1)]; + ++wq->tail; + } + status = roce_get_field(cqe->byte_4, V2_CQE_BYTE_4_STATUS_M, V2_CQE_BYTE_4_STATUS_S); switch (status & HNS_ROCE_V2_CQE_STATUS_MASK) { @@ -2394,23 +2460,6 @@ static int hns_roce_v2_poll_one(struct hns_roce_cq *hr_cq, wc->status = IB_WC_GENERAL_ERR; break; } - - wq = &(*cur_qp)->sq; - if ((*cur_qp)->sq_signal_bits) { - /* - * If sg_signal_bit is 1, - * firstly tail pointer updated to wqe - * which current cqe correspond to - */ - wqe_ctr = (u16)roce_get_field(cqe->byte_4, - V2_CQE_BYTE_4_WQE_INDX_M, - V2_CQE_BYTE_4_WQE_INDX_S); - wq->tail += (wqe_ctr - (u16)wq->tail) & - (wq->wqe_cnt - 1); - } - - wc->wr_id = wq->wrid[wq->tail & (wq->wqe_cnt - 1)]; - ++wq->tail; } else { /* RQ correspond to CQE */ wc->byte_len = le32_to_cpu(cqe->byte_cnt); @@ -2455,11 +2504,6 @@ static int hns_roce_v2_poll_one(struct hns_roce_cq *hr_cq, return -EAGAIN; } - /* Update tail pointer, record wr_id */ - wq = &(*cur_qp)->rq; - wc->wr_id = wq->wrid[wq->tail & (wq->wqe_cnt - 1)]; - ++wq->tail; - wc->sl = (u8)roce_get_field(cqe->byte_32, V2_CQE_BYTE_32_SL_M, V2_CQE_BYTE_32_SL_S); wc->src_qp = (u8)roce_get_field(cqe->byte_32, @@ -2768,6 +2812,8 @@ static void modify_qp_reset_to_init(struct ib_qp *ibqp, roce_set_field(context->byte_20_smac_sgid_idx, V2_QPC_BYTE_20_RQ_SHIFT_M, V2_QPC_BYTE_20_RQ_SHIFT_S, + (hr_qp->ibqp.qp_type == IB_QPT_XRC_INI || + hr_qp->ibqp.qp_type == IB_QPT_XRC_TGT || ibqp->srq) ? 0 : ilog2((unsigned int)hr_qp->rq.wqe_cnt)); roce_set_field(qpc_mask->byte_20_smac_sgid_idx, V2_QPC_BYTE_20_RQ_SHIFT_M, V2_QPC_BYTE_20_RQ_SHIFT_S, 0); @@ -3109,6 +3155,8 @@ static void modify_qp_init_to_init(struct ib_qp *ibqp, roce_set_field(context->byte_20_smac_sgid_idx, V2_QPC_BYTE_20_RQ_SHIFT_M, V2_QPC_BYTE_20_RQ_SHIFT_S, + (hr_qp->ibqp.qp_type == IB_QPT_XRC_INI || + hr_qp->ibqp.qp_type == IB_QPT_XRC_TGT || ibqp->srq) ? 0 : ilog2((unsigned int)hr_qp->rq.wqe_cnt)); roce_set_field(qpc_mask->byte_20_smac_sgid_idx, V2_QPC_BYTE_20_RQ_SHIFT_M, V2_QPC_BYTE_20_RQ_SHIFT_S, 0); @@ -3810,6 +3858,11 @@ static int hns_roce_v2_modify_qp(struct ib_qp *ibqp, if (attr_mask & (IB_QP_ACCESS_FLAGS | IB_QP_MAX_DEST_RD_ATOMIC)) set_access_flags(hr_qp, context, qpc_mask, attr, attr_mask); + roce_set_bit(context->byte_108_rx_reqepsn, V2_QPC_BYTE_108_INV_CREDIT_S, + ibqp->srq ? 1 : 0); + roce_set_bit(qpc_mask->byte_108_rx_reqepsn, + V2_QPC_BYTE_108_INV_CREDIT_S, 0); + /* Every status migrate must change state */ roce_set_field(context->byte_60_qpst_tempid, V2_QPC_BYTE_60_QP_ST_M, V2_QPC_BYTE_60_QP_ST_S, new_state); @@ -4095,7 +4148,8 @@ static int hns_roce_v2_destroy_qp_common(struct hns_roce_dev *hr_dev, hns_roce_free_db(hr_dev, &hr_qp->rdb); } - if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_RQ_INLINE) { + if ((hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_RQ_INLINE) && + hr_qp->rq.wqe_cnt) { kfree(hr_qp->rq_inl_buf.wqe_list[0].sg_list); kfree(hr_qp->rq_inl_buf.wqe_list); } @@ -5350,6 +5404,284 @@ static void hns_roce_v2_cleanup_eq_table(struct hns_roce_dev *hr_dev) destroy_workqueue(hr_dev->irq_workq); } +static void hns_roce_v2_write_srqc(struct hns_roce_dev *hr_dev, + struct hns_roce_srq *srq, u32 pdn, u16 xrcd, + u32 cqn, void *mb_buf, u64 *mtts_wqe, + u64 *mtts_idx, dma_addr_t dma_handle_wqe, + dma_addr_t dma_handle_idx) +{ + struct hns_roce_srq_context *srq_context; + + srq_context = mb_buf; + memset(srq_context, 0, sizeof(*srq_context)); + + roce_set_field(srq_context->byte_4_srqn_srqst, SRQC_BYTE_4_SRQ_ST_M, + SRQC_BYTE_4_SRQ_ST_S, 1); + + roce_set_field(srq_context->byte_4_srqn_srqst, + SRQC_BYTE_4_SRQ_WQE_HOP_NUM_M, + SRQC_BYTE_4_SRQ_WQE_HOP_NUM_S, + (hr_dev->caps.srqwqe_hop_num == HNS_ROCE_HOP_NUM_0 ? 0 : + hr_dev->caps.srqwqe_hop_num)); + roce_set_field(srq_context->byte_4_srqn_srqst, + SRQC_BYTE_4_SRQ_SHIFT_M, SRQC_BYTE_4_SRQ_SHIFT_S, + ilog2(srq->max)); + + roce_set_field(srq_context->byte_4_srqn_srqst, SRQC_BYTE_4_SRQN_M, + SRQC_BYTE_4_SRQN_S, srq->srqn); + + roce_set_field(srq_context->byte_8_limit_wl, SRQC_BYTE_8_SRQ_LIMIT_WL_M, + SRQC_BYTE_8_SRQ_LIMIT_WL_S, 0); + + roce_set_field(srq_context->byte_12_xrcd, SRQC_BYTE_12_SRQ_XRCD_M, + SRQC_BYTE_12_SRQ_XRCD_S, xrcd); + + srq_context->wqe_bt_ba = cpu_to_le32((u32)(dma_handle_wqe >> 3)); + + roce_set_field(srq_context->byte_24_wqe_bt_ba, + SRQC_BYTE_24_SRQ_WQE_BT_BA_M, + SRQC_BYTE_24_SRQ_WQE_BT_BA_S, + cpu_to_le32(dma_handle_wqe >> 35)); + + roce_set_field(srq_context->byte_28_rqws_pd, SRQC_BYTE_28_PD_M, + SRQC_BYTE_28_PD_S, pdn); + roce_set_field(srq_context->byte_28_rqws_pd, SRQC_BYTE_28_RQWS_M, + SRQC_BYTE_28_RQWS_S, srq->max_gs <= 0 ? 0 : + fls(srq->max_gs - 1)); + + srq_context->idx_bt_ba = (u32)(dma_handle_idx >> 3); + srq_context->idx_bt_ba = cpu_to_le32(srq_context->idx_bt_ba); + roce_set_field(srq_context->rsv_idx_bt_ba, + SRQC_BYTE_36_SRQ_IDX_BT_BA_M, + SRQC_BYTE_36_SRQ_IDX_BT_BA_S, + cpu_to_le32(dma_handle_idx >> 35)); + + srq_context->idx_cur_blk_addr = (u32)(mtts_idx[0] >> PAGE_ADDR_SHIFT); + srq_context->idx_cur_blk_addr = + cpu_to_le32(srq_context->idx_cur_blk_addr); + roce_set_field(srq_context->byte_44_idxbufpgsz_addr, + SRQC_BYTE_44_SRQ_IDX_CUR_BLK_ADDR_M, + SRQC_BYTE_44_SRQ_IDX_CUR_BLK_ADDR_S, + cpu_to_le32((mtts_idx[0]) >> (32 + PAGE_ADDR_SHIFT))); + roce_set_field(srq_context->byte_44_idxbufpgsz_addr, + SRQC_BYTE_44_SRQ_IDX_HOP_NUM_M, + SRQC_BYTE_44_SRQ_IDX_HOP_NUM_S, + hr_dev->caps.idx_hop_num == HNS_ROCE_HOP_NUM_0 ? 0 : + hr_dev->caps.idx_hop_num); + + roce_set_field(srq_context->byte_44_idxbufpgsz_addr, + SRQC_BYTE_44_SRQ_IDX_BA_PG_SZ_M, + SRQC_BYTE_44_SRQ_IDX_BA_PG_SZ_S, + hr_dev->caps.idx_ba_pg_sz); + roce_set_field(srq_context->byte_44_idxbufpgsz_addr, + SRQC_BYTE_44_SRQ_IDX_BUF_PG_SZ_M, + SRQC_BYTE_44_SRQ_IDX_BUF_PG_SZ_S, + hr_dev->caps.idx_buf_pg_sz); + + srq_context->idx_nxt_blk_addr = (u32)(mtts_idx[1] >> PAGE_ADDR_SHIFT); + srq_context->idx_nxt_blk_addr = + cpu_to_le32(srq_context->idx_nxt_blk_addr); + roce_set_field(srq_context->rsv_idxnxtblkaddr, + SRQC_BYTE_52_SRQ_IDX_NXT_BLK_ADDR_M, + SRQC_BYTE_52_SRQ_IDX_NXT_BLK_ADDR_S, + cpu_to_le32((mtts_idx[1]) >> (32 + PAGE_ADDR_SHIFT))); + roce_set_field(srq_context->byte_56_xrc_cqn, + SRQC_BYTE_56_SRQ_XRC_CQN_M, SRQC_BYTE_56_SRQ_XRC_CQN_S, + cqn); + roce_set_field(srq_context->byte_56_xrc_cqn, + SRQC_BYTE_56_SRQ_WQE_BA_PG_SZ_M, + SRQC_BYTE_56_SRQ_WQE_BA_PG_SZ_S, + hr_dev->caps.srqwqe_ba_pg_sz + PG_SHIFT_OFFSET); + roce_set_field(srq_context->byte_56_xrc_cqn, + SRQC_BYTE_56_SRQ_WQE_BUF_PG_SZ_M, + SRQC_BYTE_56_SRQ_WQE_BUF_PG_SZ_S, + hr_dev->caps.srqwqe_buf_pg_sz + PG_SHIFT_OFFSET); + + roce_set_bit(srq_context->db_record_addr_record_en, + SRQC_BYTE_60_SRQ_RECORD_EN_S, 0); +} + +static int hns_roce_v2_modify_srq(struct ib_srq *ibsrq, + struct ib_srq_attr *srq_attr, + enum ib_srq_attr_mask srq_attr_mask, + struct ib_udata *udata) +{ + struct hns_roce_dev *hr_dev = to_hr_dev(ibsrq->device); + struct hns_roce_srq *srq = to_hr_srq(ibsrq); + struct hns_roce_srq_context *srq_context; + struct hns_roce_srq_context *srqc_mask; + struct hns_roce_cmd_mailbox *mailbox; + int ret; + + if (srq_attr_mask & IB_SRQ_LIMIT) { + if (srq_attr->srq_limit >= srq->max) + return -EINVAL; + + mailbox = hns_roce_alloc_cmd_mailbox(hr_dev); + if (IS_ERR(mailbox)) + return PTR_ERR(mailbox); + + srq_context = mailbox->buf; + srqc_mask = (struct hns_roce_srq_context *)mailbox->buf + 1; + + memset(srqc_mask, 0xff, sizeof(*srqc_mask)); + + roce_set_field(srq_context->byte_8_limit_wl, + SRQC_BYTE_8_SRQ_LIMIT_WL_M, + SRQC_BYTE_8_SRQ_LIMIT_WL_S, srq_attr->srq_limit); + roce_set_field(srqc_mask->byte_8_limit_wl, + SRQC_BYTE_8_SRQ_LIMIT_WL_M, + SRQC_BYTE_8_SRQ_LIMIT_WL_S, 0); + + ret = hns_roce_cmd_mbox(hr_dev, mailbox->dma, 0, srq->srqn, 0, + HNS_ROCE_CMD_MODIFY_SRQC, + HNS_ROCE_CMD_TIMEOUT_MSECS); + hns_roce_free_cmd_mailbox(hr_dev, mailbox); + if (ret) { + dev_err(hr_dev->dev, + "MODIFY SRQ Failed to cmd mailbox.\n"); + return ret; + } + } + + return 0; +} + +int hns_roce_v2_query_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr) +{ + struct hns_roce_dev *hr_dev = to_hr_dev(ibsrq->device); + struct hns_roce_srq *srq = to_hr_srq(ibsrq); + struct hns_roce_srq_context *srq_context; + struct hns_roce_cmd_mailbox *mailbox; + int limit_wl; + int ret; + + mailbox = hns_roce_alloc_cmd_mailbox(hr_dev); + if (IS_ERR(mailbox)) + return PTR_ERR(mailbox); + + srq_context = mailbox->buf; + ret = hns_roce_cmd_mbox(hr_dev, 0, mailbox->dma, srq->srqn, 0, + HNS_ROCE_CMD_QUERY_SRQC, + HNS_ROCE_CMD_TIMEOUT_MSECS); + if (ret) { + dev_err(hr_dev->dev, "QUERY SRQ cmd process error\n"); + goto out; + } + + limit_wl = roce_get_field(srq_context->byte_8_limit_wl, + SRQC_BYTE_8_SRQ_LIMIT_WL_M, + SRQC_BYTE_8_SRQ_LIMIT_WL_S); + + attr->srq_limit = limit_wl; + attr->max_wr = srq->max - 1; + attr->max_sge = srq->max_gs; + + memcpy(srq_context, mailbox->buf, sizeof(*srq_context)); + +out: + hns_roce_free_cmd_mailbox(hr_dev, mailbox); + return ret; +} + +static int find_empty_entry(struct hns_roce_idx_que *idx_que) +{ + int bit_num; + int i; + + /* bitmap[i] is set zero if all bits are allocated */ + for (i = 0; idx_que->bitmap[i] == 0; ++i) + ; + bit_num = ffs(idx_que->bitmap[i]); + idx_que->bitmap[i] &= ~(1ULL << (bit_num - 1)); + + return i * sizeof(u64) * 8 + (bit_num - 1); +} + +static void fill_idx_queue(struct hns_roce_idx_que *idx_que, + int cur_idx, int wqe_idx) +{ + unsigned int *addr; + + addr = (unsigned int *)hns_roce_buf_offset(&idx_que->idx_buf, + cur_idx * idx_que->entry_sz); + *addr = wqe_idx; +} + +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_srq *srq = to_hr_srq(ibsrq); + struct hns_roce_v2_wqe_data_seg *dseg; + struct hns_roce_v2_db srq_db; + unsigned long flags; + int ret = 0; + int wqe_idx; + void *wqe; + int nreq; + int ind; + int i; + + spin_lock_irqsave(&srq->lock, flags); + + ind = srq->head & (srq->max - 1); + + for (nreq = 0; wr; ++nreq, wr = wr->next) { + if (unlikely(wr->num_sge > srq->max_gs)) { + ret = -EINVAL; + *bad_wr = wr; + break; + } + + if (unlikely(srq->head == srq->tail)) { + ret = -ENOMEM; + *bad_wr = wr; + break; + } + + wqe_idx = find_empty_entry(&srq->idx_que); + fill_idx_queue(&srq->idx_que, ind, wqe_idx); + wqe = get_srq_wqe(srq, wqe_idx); + dseg = (struct hns_roce_v2_wqe_data_seg *)wqe; + + for (i = 0; i < wr->num_sge; ++i) { + dseg[i].len = cpu_to_le32(wr->sg_list[i].length); + dseg[i].lkey = cpu_to_le32(wr->sg_list[i].lkey); + dseg[i].addr = cpu_to_le64(wr->sg_list[i].addr); + } + + if (i < srq->max_gs) { + dseg->len = 0; + dseg->lkey = cpu_to_le32(0x100); + dseg->addr = 0; + } + + srq->wrid[wqe_idx] = wr->wr_id; + ind = (ind + 1) & (srq->max - 1); + } + + if (likely(nreq)) { + srq->head += nreq; + + /* + * Make sure that descriptors are written before + * doorbell record. + */ + wmb(); + + srq_db.byte_4 = HNS_ROCE_V2_SRQ_DB << 24 | srq->srqn; + srq_db.parameter = srq->head; + + hns_roce_write64_k((__le32 *)&srq_db, srq->db_reg_l); + + } + + spin_unlock_irqrestore(&srq->lock, flags); + + return ret; +} + static const struct hns_roce_hw hns_roce_hw_v2 = { .cmq_init = hns_roce_v2_cmq_init, .cmq_exit = hns_roce_v2_cmq_exit, @@ -5377,6 +5709,10 @@ static const struct hns_roce_hw hns_roce_hw_v2 = { .poll_cq = hns_roce_v2_poll_cq, .init_eq = hns_roce_v2_init_eq_table, .cleanup_eq = hns_roce_v2_cleanup_eq_table, + .write_srqc = hns_roce_v2_write_srqc, + .modify_srq = hns_roce_v2_modify_srq, + .query_srq = hns_roce_v2_query_srq, + .post_srq_recv = hns_roce_v2_post_srq_recv, }; static const struct pci_device_id hns_roce_hw_v2_pci_tbl[] = { diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.h b/drivers/infiniband/hw/hns/hns_roce_hw_v2.h index 383bab5e75ef..c48301c6fe5d 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.h +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.h @@ -93,8 +93,10 @@ #define HNS_ROCE_CONTEXT_HOP_NUM 1 #define HNS_ROCE_MTT_HOP_NUM 1 #define HNS_ROCE_CQE_HOP_NUM 1 +#define HNS_ROCE_SRQWQE_HOP_NUM 1 #define HNS_ROCE_PBL_HOP_NUM 2 #define HNS_ROCE_EQE_HOP_NUM 2 +#define HNS_ROCE_IDX_HOP_NUM 1 #define HNS_ROCE_V2_GID_INDEX_NUM 256 @@ -334,6 +336,90 @@ struct hns_roce_v2_cq_context { #define V2_CQC_BYTE_64_SE_CQE_IDX_S 0 #define V2_CQC_BYTE_64_SE_CQE_IDX_M GENMASK(23, 0) +struct hns_roce_srq_context { + __le32 byte_4_srqn_srqst; + __le32 byte_8_limit_wl; + __le32 byte_12_xrcd; + __le32 byte_16_pi_ci; + __le32 wqe_bt_ba; + __le32 byte_24_wqe_bt_ba; + __le32 byte_28_rqws_pd; + __le32 idx_bt_ba; + __le32 rsv_idx_bt_ba; + __le32 idx_cur_blk_addr; + __le32 byte_44_idxbufpgsz_addr; + __le32 idx_nxt_blk_addr; + __le32 rsv_idxnxtblkaddr; + __le32 byte_56_xrc_cqn; + __le32 db_record_addr_record_en; + __le32 db_record_addr; +}; + +#define SRQC_BYTE_4_SRQ_ST_S 0 +#define SRQC_BYTE_4_SRQ_ST_M GENMASK(1, 0) + +#define SRQC_BYTE_4_SRQ_WQE_HOP_NUM_S 2 +#define SRQC_BYTE_4_SRQ_WQE_HOP_NUM_M GENMASK(3, 2) + +#define SRQC_BYTE_4_SRQ_SHIFT_S 4 +#define SRQC_BYTE_4_SRQ_SHIFT_M GENMASK(7, 4) + +#define SRQC_BYTE_4_SRQN_S 8 +#define SRQC_BYTE_4_SRQN_M GENMASK(31, 8) + +#define SRQC_BYTE_8_SRQ_LIMIT_WL_S 0 +#define SRQC_BYTE_8_SRQ_LIMIT_WL_M GENMASK(15, 0) + +#define SRQC_BYTE_12_SRQ_XRCD_S 0 +#define SRQC_BYTE_12_SRQ_XRCD_M GENMASK(23, 0) + +#define SRQC_BYTE_16_SRQ_PRODUCER_IDX_S 0 +#define SRQC_BYTE_16_SRQ_PRODUCER_IDX_M GENMASK(15, 0) + +#define SRQC_BYTE_16_SRQ_CONSUMER_IDX_S 0 +#define SRQC_BYTE_16_SRQ_CONSUMER_IDX_M GENMASK(31, 16) + +#define SRQC_BYTE_24_SRQ_WQE_BT_BA_S 0 +#define SRQC_BYTE_24_SRQ_WQE_BT_BA_M GENMASK(28, 0) + +#define SRQC_BYTE_28_PD_S 0 +#define SRQC_BYTE_28_PD_M GENMASK(23, 0) + +#define SRQC_BYTE_28_RQWS_S 24 +#define SRQC_BYTE_28_RQWS_M GENMASK(27, 24) + +#define SRQC_BYTE_36_SRQ_IDX_BT_BA_S 0 +#define SRQC_BYTE_36_SRQ_IDX_BT_BA_M GENMASK(28, 0) + +#define SRQC_BYTE_44_SRQ_IDX_CUR_BLK_ADDR_S 0 +#define SRQC_BYTE_44_SRQ_IDX_CUR_BLK_ADDR_M GENMASK(19, 0) + +#define SRQC_BYTE_44_SRQ_IDX_HOP_NUM_S 22 +#define SRQC_BYTE_44_SRQ_IDX_HOP_NUM_M GENMASK(23, 22) + +#define SRQC_BYTE_44_SRQ_IDX_BA_PG_SZ_S 24 +#define SRQC_BYTE_44_SRQ_IDX_BA_PG_SZ_M GENMASK(27, 24) + +#define SRQC_BYTE_44_SRQ_IDX_BUF_PG_SZ_S 28 +#define SRQC_BYTE_44_SRQ_IDX_BUF_PG_SZ_M GENMASK(31, 28) + +#define SRQC_BYTE_52_SRQ_IDX_NXT_BLK_ADDR_S 0 +#define SRQC_BYTE_52_SRQ_IDX_NXT_BLK_ADDR_M GENMASK(19, 0) + +#define SRQC_BYTE_56_SRQ_XRC_CQN_S 0 +#define SRQC_BYTE_56_SRQ_XRC_CQN_M GENMASK(23, 0) + +#define SRQC_BYTE_56_SRQ_WQE_BA_PG_SZ_S 24 +#define SRQC_BYTE_56_SRQ_WQE_BA_PG_SZ_M GENMASK(27, 24) + +#define SRQC_BYTE_56_SRQ_WQE_BUF_PG_SZ_S 28 +#define SRQC_BYTE_56_SRQ_WQE_BUF_PG_SZ_M GENMASK(31, 28) + +#define SRQC_BYTE_60_SRQ_RECORD_EN_S 0 + +#define SRQC_BYTE_60_SRQ_DB_RECORD_ADDR_S 1 +#define SRQC_BYTE_60_SRQ_DB_RECORD_ADDR_M GENMASK(31, 1) + enum{ V2_MPT_ST_VALID = 0x1, V2_MPT_ST_FREE = 0x2, diff --git a/drivers/infiniband/hw/hns/hns_roce_main.c b/drivers/infiniband/hw/hns/hns_roce_main.c index d183f13d206f..65ba43cee810 100644 --- a/drivers/infiniband/hw/hns/hns_roce_main.c +++ b/drivers/infiniband/hw/hns/hns_roce_main.c @@ -546,6 +546,21 @@ static int hns_roce_register_device(struct hns_roce_dev *hr_dev) ib_dev->map_mr_sg = hns_roce_map_mr_sg; } + /* SRQ */ + if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_SRQ) { + ib_dev->create_srq = hns_roce_create_srq; + ib_dev->modify_srq = hr_dev->hw->modify_srq; + ib_dev->query_srq = hr_dev->hw->query_srq; + ib_dev->destroy_srq = hns_roce_destroy_srq; + ib_dev->post_srq_recv = hr_dev->hw->post_srq_recv; + ib_dev->uverbs_cmd_mask |= + (1ULL << IB_USER_VERBS_CMD_CREATE_SRQ) | + (1ULL << IB_USER_VERBS_CMD_MODIFY_SRQ) | + (1ULL << IB_USER_VERBS_CMD_QUERY_SRQ) | + (1ULL << IB_USER_VERBS_CMD_DESTROY_SRQ) | + (1ULL << IB_USER_VERBS_CMD_POST_SRQ_RECV); + } + /* OTHERS */ ib_dev->get_port_immutable = hns_roce_port_immutable; ib_dev->disassociate_ucontext = hns_roce_disassociate_ucontext; diff --git a/drivers/infiniband/hw/hns/hns_roce_mr.c b/drivers/infiniband/hw/hns/hns_roce_mr.c index 521ad2aa3a4e..fabc95dbdf97 100644 --- a/drivers/infiniband/hw/hns/hns_roce_mr.c +++ b/drivers/infiniband/hw/hns/hns_roce_mr.c @@ -184,12 +184,27 @@ static int hns_roce_alloc_mtt_range(struct hns_roce_dev *hr_dev, int order, struct hns_roce_buddy *buddy; int ret; - if (mtt_type == MTT_TYPE_WQE) { + switch (mtt_type) { + case MTT_TYPE_WQE: buddy = &mr_table->mtt_buddy; table = &mr_table->mtt_table; - } else { + break; + case MTT_TYPE_CQE: buddy = &mr_table->mtt_cqe_buddy; table = &mr_table->mtt_cqe_table; + break; + case MTT_TYPE_SRQWQE: + buddy = &mr_table->mtt_srqwqe_buddy; + table = &mr_table->mtt_srqwqe_table; + break; + case MTT_TYPE_IDX: + buddy = &mr_table->mtt_idx_buddy; + table = &mr_table->mtt_idx_table; + break; + default: + dev_err(hr_dev->dev, "Unsupport MTT table type: %d\n", + mtt_type); + return -EINVAL; } ret = hns_roce_buddy_alloc(buddy, order, seg); @@ -242,18 +257,40 @@ void hns_roce_mtt_cleanup(struct hns_roce_dev *hr_dev, struct hns_roce_mtt *mtt) if (mtt->order < 0) return; - if (mtt->mtt_type == MTT_TYPE_WQE) { + switch (mtt->mtt_type) { + case MTT_TYPE_WQE: hns_roce_buddy_free(&mr_table->mtt_buddy, mtt->first_seg, mtt->order); hns_roce_table_put_range(hr_dev, &mr_table->mtt_table, mtt->first_seg, mtt->first_seg + (1 << mtt->order) - 1); - } else { + break; + case MTT_TYPE_CQE: hns_roce_buddy_free(&mr_table->mtt_cqe_buddy, mtt->first_seg, mtt->order); hns_roce_table_put_range(hr_dev, &mr_table->mtt_cqe_table, mtt->first_seg, mtt->first_seg + (1 << mtt->order) - 1); + break; + case MTT_TYPE_SRQWQE: + hns_roce_buddy_free(&mr_table->mtt_srqwqe_buddy, mtt->first_seg, + mtt->order); + hns_roce_table_put_range(hr_dev, &mr_table->mtt_srqwqe_table, + mtt->first_seg, + mtt->first_seg + (1 << mtt->order) - 1); + break; + case MTT_TYPE_IDX: + hns_roce_buddy_free(&mr_table->mtt_idx_buddy, mtt->first_seg, + mtt->order); + hns_roce_table_put_range(hr_dev, &mr_table->mtt_idx_table, + mtt->first_seg, + mtt->first_seg + (1 << mtt->order) - 1); + break; + default: + dev_err(hr_dev->dev, + "Unsupport mtt type %d, clean mtt failed\n", + mtt->mtt_type); + break; } } EXPORT_SYMBOL_GPL(hns_roce_mtt_cleanup); @@ -713,10 +750,26 @@ static int hns_roce_write_mtt_chunk(struct hns_roce_dev *hr_dev, u32 bt_page_size; u32 i; - if (mtt->mtt_type == MTT_TYPE_WQE) + switch (mtt->mtt_type) { + case MTT_TYPE_WQE: + table = &hr_dev->mr_table.mtt_table; bt_page_size = 1 << (hr_dev->caps.mtt_ba_pg_sz + PAGE_SHIFT); - else + break; + case MTT_TYPE_CQE: + table = &hr_dev->mr_table.mtt_cqe_table; bt_page_size = 1 << (hr_dev->caps.cqe_ba_pg_sz + PAGE_SHIFT); + break; + case MTT_TYPE_SRQWQE: + table = &hr_dev->mr_table.mtt_srqwqe_table; + bt_page_size = 1 << (hr_dev->caps.srqwqe_ba_pg_sz + PAGE_SHIFT); + break; + case MTT_TYPE_IDX: + table = &hr_dev->mr_table.mtt_idx_table; + bt_page_size = 1 << (hr_dev->caps.idx_ba_pg_sz + PAGE_SHIFT); + break; + default: + return -EINVAL; + } /* All MTTs must fit in the same page */ if (start_index / (bt_page_size / sizeof(u64)) != @@ -726,11 +779,6 @@ static int hns_roce_write_mtt_chunk(struct hns_roce_dev *hr_dev, if (start_index & (HNS_ROCE_MTT_ENTRY_PER_SEG - 1)) return -EINVAL; - if (mtt->mtt_type == MTT_TYPE_WQE) - table = &hr_dev->mr_table.mtt_table; - else - table = &hr_dev->mr_table.mtt_cqe_table; - mtts = hns_roce_table_find(hr_dev, table, mtt->first_seg + s / hr_dev->caps.mtt_entry_sz, &dma_handle); @@ -759,10 +807,25 @@ static int hns_roce_write_mtt(struct hns_roce_dev *hr_dev, if (mtt->order < 0) return -EINVAL; - if (mtt->mtt_type == MTT_TYPE_WQE) + switch (mtt->mtt_type) { + case MTT_TYPE_WQE: bt_page_size = 1 << (hr_dev->caps.mtt_ba_pg_sz + PAGE_SHIFT); - else + break; + case MTT_TYPE_CQE: bt_page_size = 1 << (hr_dev->caps.cqe_ba_pg_sz + PAGE_SHIFT); + break; + case MTT_TYPE_SRQWQE: + bt_page_size = 1 << (hr_dev->caps.srqwqe_ba_pg_sz + PAGE_SHIFT); + break; + case MTT_TYPE_IDX: + bt_page_size = 1 << (hr_dev->caps.idx_ba_pg_sz + PAGE_SHIFT); + break; + default: + dev_err(hr_dev->dev, + "Unsupport mtt type %d, write mtt failed\n", + mtt->mtt_type); + return -EINVAL; + } while (npages > 0) { chunk = min_t(int, bt_page_size / sizeof(u64), npages); @@ -828,8 +891,31 @@ int hns_roce_init_mr_table(struct hns_roce_dev *hr_dev) if (ret) goto err_buddy_cqe; } + + if (hr_dev->caps.num_srqwqe_segs) { + ret = hns_roce_buddy_init(&mr_table->mtt_srqwqe_buddy, + ilog2(hr_dev->caps.num_srqwqe_segs)); + if (ret) + goto err_buddy_srqwqe; + } + + if (hr_dev->caps.num_idx_segs) { + ret = hns_roce_buddy_init(&mr_table->mtt_idx_buddy, + ilog2(hr_dev->caps.num_idx_segs)); + if (ret) + goto err_buddy_idx; + } + return 0; +err_buddy_idx: + if (hr_dev->caps.num_srqwqe_segs) + hns_roce_buddy_cleanup(&mr_table->mtt_srqwqe_buddy); + +err_buddy_srqwqe: + if (hns_roce_check_whether_mhop(hr_dev, HEM_TYPE_CQE)) + hns_roce_buddy_cleanup(&mr_table->mtt_cqe_buddy); + err_buddy_cqe: hns_roce_buddy_cleanup(&mr_table->mtt_buddy); @@ -842,6 +928,10 @@ void hns_roce_cleanup_mr_table(struct hns_roce_dev *hr_dev) { struct hns_roce_mr_table *mr_table = &hr_dev->mr_table; + if (hr_dev->caps.num_idx_segs) + hns_roce_buddy_cleanup(&mr_table->mtt_idx_buddy); + if (hr_dev->caps.num_srqwqe_segs) + hns_roce_buddy_cleanup(&mr_table->mtt_srqwqe_buddy); hns_roce_buddy_cleanup(&mr_table->mtt_buddy); if (hns_roce_check_whether_mhop(hr_dev, HEM_TYPE_CQE)) hns_roce_buddy_cleanup(&mr_table->mtt_cqe_buddy); @@ -897,8 +987,25 @@ int hns_roce_ib_umem_write_mtt(struct hns_roce_dev *hr_dev, u32 bt_page_size; u32 n; - order = mtt->mtt_type == MTT_TYPE_WQE ? hr_dev->caps.mtt_ba_pg_sz : - hr_dev->caps.cqe_ba_pg_sz; + switch (mtt->mtt_type) { + case MTT_TYPE_WQE: + order = hr_dev->caps.mtt_ba_pg_sz; + break; + case MTT_TYPE_CQE: + order = hr_dev->caps.cqe_ba_pg_sz; + break; + case MTT_TYPE_SRQWQE: + order = hr_dev->caps.srqwqe_ba_pg_sz; + break; + case MTT_TYPE_IDX: + order = hr_dev->caps.idx_ba_pg_sz; + break; + default: + dev_err(dev, "Unsupport mtt type %d, write mtt failed\n", + mtt->mtt_type); + return -EINVAL; + } + bt_page_size = 1 << (order + PAGE_SHIFT); pages = (u64 *) __get_free_pages(GFP_KERNEL, order); diff --git a/drivers/infiniband/hw/hns/hns_roce_qp.c b/drivers/infiniband/hw/hns/hns_roce_qp.c index 5ebf481a39d9..52d2b299b3be 100644 --- a/drivers/infiniband/hw/hns/hns_roce_qp.c +++ b/drivers/infiniband/hw/hns/hns_roce_qp.c @@ -280,7 +280,7 @@ void hns_roce_release_range_qp(struct hns_roce_dev *hr_dev, int base_qpn, EXPORT_SYMBOL_GPL(hns_roce_release_range_qp); static int hns_roce_set_rq_size(struct hns_roce_dev *hr_dev, - struct ib_qp_cap *cap, int is_user, int has_srq, + struct ib_qp_cap *cap, int is_user, int has_rq, struct hns_roce_qp *hr_qp) { struct device *dev = hr_dev->dev; @@ -294,14 +294,12 @@ static int hns_roce_set_rq_size(struct hns_roce_dev *hr_dev, return -EINVAL; } - /* If srq exit, set zero for relative number of rq */ - if (has_srq) { - if (cap->max_recv_wr) { - dev_dbg(dev, "srq no need config max_recv_wr\n"); - return -EINVAL; - } - - hr_qp->rq.wqe_cnt = hr_qp->rq.max_gs = 0; + /* If srq exist, set zero for relative number of rq */ + if (!has_rq) { + hr_qp->rq.wqe_cnt = 0; + hr_qp->rq.max_gs = 0; + cap->max_recv_wr = 0; + cap->max_recv_sge = 0; } else { if (is_user && (!cap->max_recv_wr || !cap->max_recv_sge)) { dev_err(dev, "user space no need config max_recv_wr max_recv_sge\n"); @@ -563,13 +561,14 @@ static int hns_roce_create_qp_common(struct hns_roce_dev *hr_dev, hr_qp->sq_signal_bits = cpu_to_le32(IB_SIGNAL_REQ_WR); ret = hns_roce_set_rq_size(hr_dev, &init_attr->cap, !!ib_pd->uobject, - !!init_attr->srq, hr_qp); + hns_roce_qp_has_rq(init_attr), hr_qp); if (ret) { dev_err(dev, "hns_roce_set_rq_size failed\n"); goto err_out; } - if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_RQ_INLINE) { + if ((hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_RQ_INLINE) && + hns_roce_qp_has_rq(init_attr)) { /* allocate recv inline buf */ hr_qp->rq_inl_buf.wqe_list = kcalloc(hr_qp->rq.wqe_cnt, sizeof(struct hns_roce_rinl_wqe), diff --git a/drivers/infiniband/hw/hns/hns_roce_srq.c b/drivers/infiniband/hw/hns/hns_roce_srq.c index d8a86130febe..46732d25a8a8 100644 --- a/drivers/infiniband/hw/hns/hns_roce_srq.c +++ b/drivers/infiniband/hw/hns/hns_roce_srq.c @@ -9,6 +9,413 @@ #include "hns_roce_cmd.h" #include "hns_roce_hem.h" +static void hns_roce_ib_srq_event(struct hns_roce_srq *srq, + enum hns_roce_event event_type) +{ + struct hns_roce_dev *hr_dev = to_hr_dev(srq->ibsrq.device); + struct ib_srq *ibsrq = &srq->ibsrq; + struct ib_event event; + + if (ibsrq->event_handler) { + event.device = ibsrq->device; + event.element.srq = ibsrq; + switch (event_type) { + case HNS_ROCE_EVENT_TYPE_SRQ_LIMIT_REACH: + event.event = IB_EVENT_SRQ_LIMIT_REACHED; + break; + case HNS_ROCE_EVENT_TYPE_SRQ_CATAS_ERROR: + event.event = IB_EVENT_SRQ_ERR; + break; + default: + dev_err(hr_dev->dev, + "hns_roce:Unexpected event type 0x%x on SRQ %06lx\n", + event_type, srq->srqn); + return; + } + + ibsrq->event_handler(&event, ibsrq->srq_context); + } +} + +static int hns_roce_sw2hw_srq(struct hns_roce_dev *dev, + struct hns_roce_cmd_mailbox *mailbox, + unsigned long srq_num) +{ + return hns_roce_cmd_mbox(dev, mailbox->dma, 0, srq_num, 0, + HNS_ROCE_CMD_SW2HW_SRQ, + HNS_ROCE_CMD_TIMEOUT_MSECS); +} + +static int hns_roce_hw2sw_srq(struct hns_roce_dev *dev, + struct hns_roce_cmd_mailbox *mailbox, + unsigned long srq_num) +{ + return hns_roce_cmd_mbox(dev, 0, mailbox ? mailbox->dma : 0, srq_num, + mailbox ? 0 : 1, HNS_ROCE_CMD_HW2SW_SRQ, + HNS_ROCE_CMD_TIMEOUT_MSECS); +} + +int hns_roce_srq_alloc(struct hns_roce_dev *hr_dev, u32 pdn, u32 cqn, u16 xrcd, + struct hns_roce_mtt *hr_mtt, u64 db_rec_addr, + struct hns_roce_srq *srq) +{ + struct hns_roce_srq_table *srq_table = &hr_dev->srq_table; + struct hns_roce_cmd_mailbox *mailbox; + dma_addr_t dma_handle_wqe; + dma_addr_t dma_handle_idx; + u64 *mtts_wqe; + u64 *mtts_idx; + int ret; + + /* Get the physical address of srq buf */ + mtts_wqe = hns_roce_table_find(hr_dev, + &hr_dev->mr_table.mtt_srqwqe_table, + srq->mtt.first_seg, + &dma_handle_wqe); + if (!mtts_wqe) { + dev_err(hr_dev->dev, + "SRQ alloc.Failed to find srq buf addr.\n"); + return -EINVAL; + } + + /* Get physical address of idx que buf */ + mtts_idx = hns_roce_table_find(hr_dev, &hr_dev->mr_table.mtt_idx_table, + srq->idx_que.mtt.first_seg, + &dma_handle_idx); + if (!mtts_idx) { + dev_err(hr_dev->dev, + "SRQ alloc.Failed to find idx que buf addr.\n"); + return -EINVAL; + } + + ret = hns_roce_bitmap_alloc(&srq_table->bitmap, &srq->srqn); + if (ret == -1) { + dev_err(hr_dev->dev, "SRQ alloc.Failed to alloc index.\n"); + return -ENOMEM; + } + + ret = hns_roce_table_get(hr_dev, &srq_table->table, srq->srqn); + if (ret) + goto err_out; + + ret = xa_err(xa_store(&srq_table->xa, srq->srqn, srq, GFP_KERNEL)); + if (ret) + goto err_put; + + mailbox = hns_roce_alloc_cmd_mailbox(hr_dev); + if (IS_ERR(mailbox)) { + ret = PTR_ERR(mailbox); + goto err_xa; + } + + hr_dev->hw->write_srqc(hr_dev, srq, pdn, xrcd, cqn, mailbox->buf, + mtts_wqe, mtts_idx, dma_handle_wqe, + dma_handle_idx); + + ret = hns_roce_sw2hw_srq(hr_dev, mailbox, srq->srqn); + hns_roce_free_cmd_mailbox(hr_dev, mailbox); + if (ret) + goto err_xa; + + atomic_set(&srq->refcount, 1); + init_completion(&srq->free); + return ret; + +err_xa: + xa_erase(&srq_table->xa, srq->srqn); + +err_put: + hns_roce_table_put(hr_dev, &srq_table->table, srq->srqn); + +err_out: + hns_roce_bitmap_free(&srq_table->bitmap, srq->srqn, BITMAP_NO_RR); + return ret; +} + +void hns_roce_srq_free(struct hns_roce_dev *hr_dev, struct hns_roce_srq *srq) +{ + struct hns_roce_srq_table *srq_table = &hr_dev->srq_table; + int ret; + + ret = hns_roce_hw2sw_srq(hr_dev, NULL, srq->srqn); + if (ret) + dev_err(hr_dev->dev, "HW2SW_SRQ failed (%d) for CQN %06lx\n", + ret, srq->srqn); + + xa_erase(&srq_table->xa, srq->srqn); + + if (atomic_dec_and_test(&srq->refcount)) + complete(&srq->free); + wait_for_completion(&srq->free); + + hns_roce_table_put(hr_dev, &srq_table->table, srq->srqn); + hns_roce_bitmap_free(&srq_table->bitmap, srq->srqn, BITMAP_NO_RR); +} + +static int hns_roce_create_idx_que(struct ib_pd *pd, struct hns_roce_srq *srq, + u32 page_shift) +{ + struct hns_roce_dev *hr_dev = to_hr_dev(pd->device); + struct hns_roce_idx_que *idx_que = &srq->idx_que; + u32 bitmap_num; + int i; + + bitmap_num = HNS_ROCE_ALOGN_UP(srq->max, 8 * sizeof(u64)); + + idx_que->bitmap = kcalloc(1, bitmap_num / 8, GFP_KERNEL); + if (!idx_que->bitmap) + return -ENOMEM; + + bitmap_num = bitmap_num / (8 * sizeof(u64)); + + idx_que->buf_size = srq->idx_que.buf_size; + + if (hns_roce_buf_alloc(hr_dev, idx_que->buf_size, (1 << page_shift) * 2, + &idx_que->idx_buf, page_shift)) { + kfree(idx_que->bitmap); + return -ENOMEM; + } + + for (i = 0; i < bitmap_num; i++) + idx_que->bitmap[i] = ~(0UL); + + return 0; +} + +struct ib_srq *hns_roce_create_srq(struct ib_pd *pd, + struct ib_srq_init_attr *srq_init_attr, + struct ib_udata *udata) +{ + struct hns_roce_dev *hr_dev = to_hr_dev(pd->device); + struct hns_roce_srq *srq; + int srq_desc_size; + int srq_buf_size; + u32 page_shift; + int ret = 0; + u32 npages; + u32 cqn; + + /* Check the actual SRQ wqe and SRQ sge num */ + if (srq_init_attr->attr.max_wr >= hr_dev->caps.max_srq_wrs || + srq_init_attr->attr.max_sge > hr_dev->caps.max_srq_sges) + return ERR_PTR(-EINVAL); + + srq = kzalloc(sizeof(*srq), GFP_KERNEL); + if (!srq) + return ERR_PTR(-ENOMEM); + + mutex_init(&srq->mutex); + spin_lock_init(&srq->lock); + + srq->max = roundup_pow_of_two(srq_init_attr->attr.max_wr + 1); + srq->max_gs = srq_init_attr->attr.max_sge; + + srq_desc_size = max(16, 16 * srq->max_gs); + + srq->wqe_shift = ilog2(srq_desc_size); + + srq_buf_size = srq->max * srq_desc_size; + + srq->idx_que.entry_sz = HNS_ROCE_IDX_QUE_ENTRY_SZ; + srq->idx_que.buf_size = srq->max * srq->idx_que.entry_sz; + srq->mtt.mtt_type = MTT_TYPE_SRQWQE; + srq->idx_que.mtt.mtt_type = MTT_TYPE_IDX; + + if (udata) { + struct hns_roce_ib_create_srq ucmd; + + if (ib_copy_from_udata(&ucmd, udata, sizeof(ucmd))) { + ret = -EFAULT; + goto err_srq; + } + + srq->umem = ib_umem_get(pd->uobject->context, ucmd.buf_addr, + srq_buf_size, 0, 0); + if (IS_ERR(srq->umem)) { + ret = PTR_ERR(srq->umem); + goto err_srq; + } + + if (hr_dev->caps.srqwqe_buf_pg_sz) { + npages = (ib_umem_page_count(srq->umem) + + (1 << hr_dev->caps.srqwqe_buf_pg_sz) - 1) / + (1 << hr_dev->caps.srqwqe_buf_pg_sz); + page_shift = PAGE_SHIFT + hr_dev->caps.srqwqe_buf_pg_sz; + ret = hns_roce_mtt_init(hr_dev, npages, + page_shift, + &srq->mtt); + } else + ret = hns_roce_mtt_init(hr_dev, + ib_umem_page_count(srq->umem), + srq->umem->page_shift, + &srq->mtt); + if (ret) + goto err_buf; + + ret = hns_roce_ib_umem_write_mtt(hr_dev, &srq->mtt, srq->umem); + if (ret) + goto err_srq_mtt; + + /* config index queue BA */ + srq->idx_que.umem = ib_umem_get(pd->uobject->context, + ucmd.que_addr, + srq->idx_que.buf_size, 0, 0); + if (IS_ERR(srq->idx_que.umem)) { + dev_err(hr_dev->dev, + "ib_umem_get error for index queue\n"); + goto err_srq_mtt; + } + + if (hr_dev->caps.idx_buf_pg_sz) { + npages = (ib_umem_page_count(srq->idx_que.umem) + + (1 << hr_dev->caps.idx_buf_pg_sz) - 1) / + (1 << hr_dev->caps.idx_buf_pg_sz); + page_shift = PAGE_SHIFT + hr_dev->caps.idx_buf_pg_sz; + ret = hns_roce_mtt_init(hr_dev, npages, + page_shift, &srq->idx_que.mtt); + } else { + ret = hns_roce_mtt_init(hr_dev, + ib_umem_page_count(srq->idx_que.umem), + srq->idx_que.umem->page_shift, + &srq->idx_que.mtt); + } + + if (ret) { + dev_err(hr_dev->dev, + "hns_roce_mtt_init error for idx que\n"); + goto err_idx_mtt; + } + + ret = hns_roce_ib_umem_write_mtt(hr_dev, &srq->idx_que.mtt, + srq->idx_que.umem); + if (ret) { + dev_err(hr_dev->dev, + "hns_roce_ib_umem_write_mtt error for idx que\n"); + goto err_idx_buf; + } + } else { + page_shift = PAGE_SHIFT + hr_dev->caps.srqwqe_buf_pg_sz; + if (hns_roce_buf_alloc(hr_dev, srq_buf_size, + (1 << page_shift) * 2, + &srq->buf, page_shift)) { + ret = -ENOMEM; + goto err_srq; + } + + srq->head = 0; + srq->tail = srq->max - 1; + + ret = hns_roce_mtt_init(hr_dev, srq->buf.npages, + srq->buf.page_shift, &srq->mtt); + if (ret) + goto err_buf; + + ret = hns_roce_buf_write_mtt(hr_dev, &srq->mtt, &srq->buf); + if (ret) + goto err_srq_mtt; + + page_shift = PAGE_SHIFT + hr_dev->caps.idx_buf_pg_sz; + ret = hns_roce_create_idx_que(pd, srq, page_shift); + if (ret) { + dev_err(hr_dev->dev, "Create idx queue fail(%d)!\n", + ret); + goto err_srq_mtt; + } + + /* Init mtt table for idx_que */ + ret = hns_roce_mtt_init(hr_dev, srq->idx_que.idx_buf.npages, + srq->idx_que.idx_buf.page_shift, + &srq->idx_que.mtt); + if (ret) + goto err_create_idx; + + /* Write buffer address into the mtt table */ + ret = hns_roce_buf_write_mtt(hr_dev, &srq->idx_que.mtt, + &srq->idx_que.idx_buf); + if (ret) + goto err_idx_buf; + + srq->wrid = kvmalloc_array(srq->max, sizeof(u64), GFP_KERNEL); + if (!srq->wrid) { + ret = -ENOMEM; + goto err_idx_buf; + } + } + + cqn = ib_srq_has_cq(srq_init_attr->srq_type) ? + to_hr_cq(srq_init_attr->ext.cq)->cqn : 0; + + srq->db_reg_l = hr_dev->reg_base + SRQ_DB_REG; + + ret = hns_roce_srq_alloc(hr_dev, to_hr_pd(pd)->pdn, cqn, 0, + &srq->mtt, 0, srq); + if (ret) + goto err_wrid; + + srq->event = hns_roce_ib_srq_event; + srq->ibsrq.ext.xrc.srq_num = srq->srqn; + + if (pd->uobject) { + if (ib_copy_to_udata(udata, &srq->srqn, sizeof(__u32))) { + ret = -EFAULT; + goto err_wrid; + } + } + + return &srq->ibsrq; + +err_wrid: + kvfree(srq->wrid); + +err_idx_buf: + hns_roce_mtt_cleanup(hr_dev, &srq->idx_que.mtt); + +err_idx_mtt: + if (udata) + ib_umem_release(srq->idx_que.umem); + +err_create_idx: + hns_roce_buf_free(hr_dev, srq->idx_que.buf_size, + &srq->idx_que.idx_buf); + kfree(srq->idx_que.bitmap); + +err_srq_mtt: + hns_roce_mtt_cleanup(hr_dev, &srq->mtt); + +err_buf: + if (udata) + ib_umem_release(srq->umem); + else + hns_roce_buf_free(hr_dev, srq_buf_size, &srq->buf); + +err_srq: + kfree(srq); + return ERR_PTR(ret); +} + +int hns_roce_destroy_srq(struct ib_srq *ibsrq) +{ + struct hns_roce_dev *hr_dev = to_hr_dev(ibsrq->device); + struct hns_roce_srq *srq = to_hr_srq(ibsrq); + + hns_roce_srq_free(hr_dev, srq); + hns_roce_mtt_cleanup(hr_dev, &srq->mtt); + + if (ibsrq->uobject) { + hns_roce_mtt_cleanup(hr_dev, &srq->idx_que.mtt); + ib_umem_release(srq->idx_que.umem); + ib_umem_release(srq->umem); + } else { + kvfree(srq->wrid); + hns_roce_buf_free(hr_dev, srq->max << srq->wqe_shift, + &srq->buf); + } + + kfree(srq); + + return 0; +} + int hns_roce_init_srq_table(struct hns_roce_dev *hr_dev) { struct hns_roce_srq_table *srq_table = &hr_dev->srq_table; diff --git a/include/uapi/rdma/hns-abi.h b/include/uapi/rdma/hns-abi.h index c1f87735514f..ef3c7ec793a7 100644 --- a/include/uapi/rdma/hns-abi.h +++ b/include/uapi/rdma/hns-abi.h @@ -46,6 +46,12 @@ struct hns_roce_ib_create_cq_resp { __aligned_u64 cap_flags; }; +struct hns_roce_ib_create_srq { + __aligned_u64 buf_addr; + __aligned_u64 db_addr; + __aligned_u64 que_addr; +}; + struct hns_roce_ib_create_qp { __aligned_u64 buf_addr; __aligned_u64 db_addr; -- cgit v1.2.3-59-g8ed1b From 81fce6291d9999cee692e4118134a8c850b60857 Mon Sep 17 00:00:00 2001 From: Lijun Ou Date: Sat, 24 Nov 2018 16:49:22 +0800 Subject: RDMA/hns: Add SRQ asynchronous event support This patch implements the process flow of SRQ asynchronous event. Signed-off-by: Lijun Ou Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hns/hns_roce_device.h | 7 +++++++ drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 7 ++++++- drivers/infiniband/hw/hns/hns_roce_srq.c | 23 +++++++++++++++++++++++ 3 files changed, 36 insertions(+), 1 deletion(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/hns/hns_roce_device.h b/drivers/infiniband/hw/hns/hns_roce_device.h index 5a40746cb2e3..779dd4c409cb 100644 --- a/drivers/infiniband/hw/hns/hns_roce_device.h +++ b/drivers/infiniband/hw/hns/hns_roce_device.h @@ -646,6 +646,12 @@ struct hns_roce_aeqe { u32 rsv1; } qp_event; + struct { + __le32 srq; + u32 rsv0; + u32 rsv1; + } srq_event; + struct { __le32 cq; u32 rsv0; @@ -1135,6 +1141,7 @@ void hns_roce_free_db(struct hns_roce_dev *hr_dev, struct hns_roce_db *db); void hns_roce_cq_completion(struct hns_roce_dev *hr_dev, u32 cqn); void hns_roce_cq_event(struct hns_roce_dev *hr_dev, u32 cqn, int event_type); void hns_roce_qp_event(struct hns_roce_dev *hr_dev, u32 qpn, int event_type); +void hns_roce_srq_event(struct hns_roce_dev *hr_dev, u32 srqn, int event_type); int hns_get_gid_index(struct hns_roce_dev *hr_dev, u8 port, int gid_index); int hns_roce_init(struct hns_roce_dev *hr_dev); void hns_roce_exit(struct hns_roce_dev *hr_dev); diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c index 6c9baf99894e..835b78371294 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c @@ -4459,6 +4459,7 @@ static int hns_roce_v2_aeq_int(struct hns_roce_dev *hr_dev, int aeqe_found = 0; int event_type; int sub_type; + u32 srqn; u32 qpn; u32 cqn; @@ -4481,6 +4482,9 @@ static int hns_roce_v2_aeq_int(struct hns_roce_dev *hr_dev, cqn = roce_get_field(aeqe->event.cq_event.cq, HNS_ROCE_V2_AEQE_EVENT_QUEUE_NUM_M, HNS_ROCE_V2_AEQE_EVENT_QUEUE_NUM_S); + srqn = roce_get_field(aeqe->event.srq_event.srq, + HNS_ROCE_V2_AEQE_EVENT_QUEUE_NUM_M, + HNS_ROCE_V2_AEQE_EVENT_QUEUE_NUM_S); switch (event_type) { case HNS_ROCE_EVENT_TYPE_PATH_MIG: @@ -4488,13 +4492,14 @@ static int hns_roce_v2_aeq_int(struct hns_roce_dev *hr_dev, case HNS_ROCE_EVENT_TYPE_COMM_EST: case HNS_ROCE_EVENT_TYPE_SQ_DRAINED: case HNS_ROCE_EVENT_TYPE_WQ_CATAS_ERROR: + case HNS_ROCE_EVENT_TYPE_SRQ_LAST_WQE_REACH: case HNS_ROCE_EVENT_TYPE_INV_REQ_LOCAL_WQ_ERROR: case HNS_ROCE_EVENT_TYPE_LOCAL_WQ_ACCESS_ERROR: hns_roce_qp_event(hr_dev, qpn, event_type); break; case HNS_ROCE_EVENT_TYPE_SRQ_LIMIT_REACH: - case HNS_ROCE_EVENT_TYPE_SRQ_LAST_WQE_REACH: case HNS_ROCE_EVENT_TYPE_SRQ_CATAS_ERROR: + hns_roce_srq_event(hr_dev, srqn, event_type); break; case HNS_ROCE_EVENT_TYPE_CQ_ACCESS_ERROR: case HNS_ROCE_EVENT_TYPE_CQ_OVERFLOW: diff --git a/drivers/infiniband/hw/hns/hns_roce_srq.c b/drivers/infiniband/hw/hns/hns_roce_srq.c index 46732d25a8a8..463df60094e8 100644 --- a/drivers/infiniband/hw/hns/hns_roce_srq.c +++ b/drivers/infiniband/hw/hns/hns_roce_srq.c @@ -9,6 +9,29 @@ #include "hns_roce_cmd.h" #include "hns_roce_hem.h" +void hns_roce_srq_event(struct hns_roce_dev *hr_dev, u32 srqn, int event_type) +{ + struct hns_roce_srq_table *srq_table = &hr_dev->srq_table; + struct hns_roce_srq *srq; + + xa_lock(&srq_table->xa); + srq = xa_load(&srq_table->xa, srqn & (hr_dev->caps.num_srqs - 1)); + if (srq) + atomic_inc(&srq->refcount); + xa_unlock(&srq_table->xa); + + if (!srq) { + dev_warn(hr_dev->dev, "Async event for bogus SRQ %08x\n", srqn); + return; + } + + srq->event(srq, event_type); + + if (atomic_dec_and_test(&srq->refcount)) + complete(&srq->free); +} +EXPORT_SYMBOL_GPL(hns_roce_srq_event); + static void hns_roce_ib_srq_event(struct hns_roce_srq *srq, enum hns_roce_event event_type) { -- cgit v1.2.3-59-g8ed1b From 937488a85986faa743d12456970a0cbe83e3b04e Mon Sep 17 00:00:00 2001 From: Kaike Wan Date: Wed, 28 Nov 2018 10:14:32 -0800 Subject: IB/hfi1: Dump pio info for non-user send contexts This patch dumps the pio info for non-user send contexts to assist debugging in the field. Reviewed-by: Mike Marciniczyn Reviewed-by: Mike Ruhl Signed-off-by: Kaike Wan Signed-off-by: Dennis Dalessandro Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hfi1/chip_registers.h | 4 +++ drivers/infiniband/hw/hfi1/debugfs.c | 49 +++++++++++++++++++++++++++++ drivers/infiniband/hw/hfi1/pio.c | 25 +++++++++++++++ drivers/infiniband/hw/hfi1/pio.h | 3 ++ 4 files changed, 81 insertions(+) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/hfi1/chip_registers.h b/drivers/infiniband/hw/hfi1/chip_registers.h index c6163a347e93..c0800ea5a3f8 100644 --- a/drivers/infiniband/hw/hfi1/chip_registers.h +++ b/drivers/infiniband/hw/hfi1/chip_registers.h @@ -935,6 +935,10 @@ #define SEND_CTXT_CREDIT_CTRL_THRESHOLD_MASK 0x7FFull #define SEND_CTXT_CREDIT_CTRL_THRESHOLD_SHIFT 0 #define SEND_CTXT_CREDIT_CTRL_THRESHOLD_SMASK 0x7FFull +#define SEND_CTXT_CREDIT_STATUS (TXE + 0x000000100018) +#define SEND_CTXT_CREDIT_STATUS_CURRENT_FREE_COUNTER_MASK 0x7FFull +#define SEND_CTXT_CREDIT_STATUS_CURRENT_FREE_COUNTER_SHIFT 32 +#define SEND_CTXT_CREDIT_STATUS_LAST_RETURNED_COUNTER_SMASK 0x7FFull #define SEND_CTXT_CREDIT_FORCE (TXE + 0x000000100028) #define SEND_CTXT_CREDIT_FORCE_FORCE_RETURN_SMASK 0x1ull #define SEND_CTXT_CREDIT_RETURN_ADDR (TXE + 0x000000100020) diff --git a/drivers/infiniband/hw/hfi1/debugfs.c b/drivers/infiniband/hw/hfi1/debugfs.c index 9f992ae36c89..0a557795563c 100644 --- a/drivers/infiniband/hw/hfi1/debugfs.c +++ b/drivers/infiniband/hw/hfi1/debugfs.c @@ -407,6 +407,54 @@ DEBUGFS_SEQ_FILE_OPS(rcds); DEBUGFS_SEQ_FILE_OPEN(rcds) DEBUGFS_FILE_OPS(rcds); +static void *_pios_seq_start(struct seq_file *s, loff_t *pos) +{ + struct hfi1_ibdev *ibd; + struct hfi1_devdata *dd; + + ibd = (struct hfi1_ibdev *)s->private; + dd = dd_from_dev(ibd); + if (!dd->send_contexts || *pos >= dd->num_send_contexts) + return NULL; + return pos; +} + +static void *_pios_seq_next(struct seq_file *s, void *v, loff_t *pos) +{ + struct hfi1_ibdev *ibd = (struct hfi1_ibdev *)s->private; + struct hfi1_devdata *dd = dd_from_dev(ibd); + + ++*pos; + if (!dd->send_contexts || *pos >= dd->num_send_contexts) + return NULL; + return pos; +} + +static void _pios_seq_stop(struct seq_file *s, void *v) +{ +} + +static int _pios_seq_show(struct seq_file *s, void *v) +{ + struct hfi1_ibdev *ibd = (struct hfi1_ibdev *)s->private; + struct hfi1_devdata *dd = dd_from_dev(ibd); + struct send_context_info *sci; + loff_t *spos = v; + loff_t i = *spos; + unsigned long flags; + + spin_lock_irqsave(&dd->sc_lock, flags); + sci = &dd->send_contexts[i]; + if (sci && sci->type != SC_USER && sci->allocated && sci->sc) + seqfile_dump_sci(s, i, sci); + spin_unlock_irqrestore(&dd->sc_lock, flags); + return 0; +} + +DEBUGFS_SEQ_FILE_OPS(pios); +DEBUGFS_SEQ_FILE_OPEN(pios) +DEBUGFS_FILE_OPS(pios); + /* read the per-device counters */ static ssize_t dev_counters_read(struct file *file, char __user *buf, size_t count, loff_t *ppos) @@ -1143,6 +1191,7 @@ void hfi1_dbg_ibdev_init(struct hfi1_ibdev *ibd) DEBUGFS_SEQ_FILE_CREATE(qp_stats, ibd->hfi1_ibdev_dbg, ibd); DEBUGFS_SEQ_FILE_CREATE(sdes, ibd->hfi1_ibdev_dbg, ibd); DEBUGFS_SEQ_FILE_CREATE(rcds, ibd->hfi1_ibdev_dbg, ibd); + DEBUGFS_SEQ_FILE_CREATE(pios, ibd->hfi1_ibdev_dbg, ibd); DEBUGFS_SEQ_FILE_CREATE(sdma_cpu_list, ibd->hfi1_ibdev_dbg, ibd); /* dev counter files */ for (i = 0; i < ARRAY_SIZE(cntr_ops); i++) diff --git a/drivers/infiniband/hw/hfi1/pio.c b/drivers/infiniband/hw/hfi1/pio.c index 9ab50d2308dc..6d5d0d017e6b 100644 --- a/drivers/infiniband/hw/hfi1/pio.c +++ b/drivers/infiniband/hw/hfi1/pio.c @@ -2137,3 +2137,28 @@ void free_credit_return(struct hfi1_devdata *dd) kfree(dd->cr_base); dd->cr_base = NULL; } + +void seqfile_dump_sci(struct seq_file *s, u32 i, + struct send_context_info *sci) +{ + struct send_context *sc = sci->sc; + u64 reg; + + seq_printf(s, "SCI %u: type %u base %u credits %u\n", + i, sci->type, sci->base, sci->credits); + seq_printf(s, " flags 0x%x sw_inx %u hw_ctxt %u grp %u\n", + sc->flags, sc->sw_index, sc->hw_context, sc->group); + seq_printf(s, " sr_size %u credits %u sr_head %u sr_tail %u\n", + sc->sr_size, sc->credits, sc->sr_head, sc->sr_tail); + seq_printf(s, " fill %lu free %lu fill_wrap %u alloc_free %lu\n", + sc->fill, sc->free, sc->fill_wrap, sc->alloc_free); + seq_printf(s, " credit_intr_count %u credit_ctrl 0x%llx\n", + sc->credit_intr_count, sc->credit_ctrl); + reg = read_kctxt_csr(sc->dd, sc->hw_context, SC(CREDIT_STATUS)); + seq_printf(s, " *hw_free %llu CurrentFree %llu LastReturned %llu\n", + (le64_to_cpu(*sc->hw_free) & CR_COUNTER_SMASK) >> + CR_COUNTER_SHIFT, + (reg >> SC(CREDIT_STATUS_CURRENT_FREE_COUNTER_SHIFT)) & + SC(CREDIT_STATUS_CURRENT_FREE_COUNTER_MASK), + reg & SC(CREDIT_STATUS_LAST_RETURNED_COUNTER_SMASK)); +} diff --git a/drivers/infiniband/hw/hfi1/pio.h b/drivers/infiniband/hw/hfi1/pio.h index aaf372c3e5d6..bf1afb0077a7 100644 --- a/drivers/infiniband/hw/hfi1/pio.h +++ b/drivers/infiniband/hw/hfi1/pio.h @@ -329,4 +329,7 @@ void seg_pio_copy_start(struct pio_buf *pbuf, u64 pbc, void seg_pio_copy_mid(struct pio_buf *pbuf, const void *from, size_t nbytes); void seg_pio_copy_end(struct pio_buf *pbuf); +void seqfile_dump_sci(struct seq_file *s, u32 i, + struct send_context_info *sci); + #endif /* _PIO_H */ -- cgit v1.2.3-59-g8ed1b From c1a797c0818e0122c7ec8422edd971cfec9b15ea Mon Sep 17 00:00:00 2001 From: Kaike Wan Date: Wed, 28 Nov 2018 10:19:04 -0800 Subject: IB/hfi1: Ignore LNI errors before DC8051 transitions to Polling state When it is requested to change its physical state back to Offline while in the process to go up, DC8051 will set the ERROR field in the DC8051_DBG_ERR_INFO_SET_BY_8051 register. This ERROR field will remain until the next time when DC8051 transitions from Offline to Polling. Subsequently, when the host requests DC8051 to change its physical state to Polling again, it may receive a DC8051 interrupt with the stale ERROR field still in DC8051_DBG_ERR_INFO_SET_BY_8051. If the host link state has been changed to Polling, this stale ERROR will force the host to transition to Offline state, resulting in a vicious cycle of Polling ->Offline->Polling->Offline. On the other hand, if the host link state is still Offline when the stale ERROR is received, the stale ERROR will be ignored, and the link will come up correctly. This patch implements the correct behavior by changing host link state to Polling only after DC8051 changes its physical state to Polling. Reviewed-by: Mike Marciniszyn Signed-off-by: Krzysztof Goreczny Signed-off-by: Kaike Wan Signed-off-by: Dennis Dalessandro Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hfi1/chip.c | 47 ++++++++++++++++++++++++++++++++++++++- 1 file changed, 46 insertions(+), 1 deletion(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/hfi1/chip.c b/drivers/infiniband/hw/hfi1/chip.c index 9b20479dc710..385c33745c9f 100644 --- a/drivers/infiniband/hw/hfi1/chip.c +++ b/drivers/infiniband/hw/hfi1/chip.c @@ -1072,6 +1072,8 @@ static void log_state_transition(struct hfi1_pportdata *ppd, u32 state); static void log_physical_state(struct hfi1_pportdata *ppd, u32 state); static int wait_physical_linkstate(struct hfi1_pportdata *ppd, u32 state, int msecs); +static int wait_phys_link_out_of_offline(struct hfi1_pportdata *ppd, + int msecs); static void read_planned_down_reason_code(struct hfi1_devdata *dd, u8 *pdrrc); static void read_link_down_reason(struct hfi1_devdata *dd, u8 *ldr); static void handle_temp_err(struct hfi1_devdata *dd); @@ -10770,13 +10772,15 @@ int set_link_state(struct hfi1_pportdata *ppd, u32 state) break; ppd->port_error_action = 0; - ppd->host_link_state = HLS_DN_POLL; if (quick_linkup) { /* quick linkup does not go into polling */ ret = do_quick_linkup(dd); } else { ret1 = set_physical_link_state(dd, PLS_POLLING); + if (!ret1) + ret1 = wait_phys_link_out_of_offline(ppd, + 3000); if (ret1 != HCMD_SUCCESS) { dd_dev_err(dd, "Failed to transition to Polling link state, return 0x%x\n", @@ -10784,6 +10788,14 @@ int set_link_state(struct hfi1_pportdata *ppd, u32 state) ret = -EINVAL; } } + + /* + * Change the host link state after requesting DC8051 to + * change its physical state so that we can ignore any + * interrupt with stale LNI(XX) error, which will not be + * cleared until DC8051 transitions to Polling state. + */ + ppd->host_link_state = HLS_DN_POLL; ppd->offline_disabled_reason = HFI1_ODR_MASK(OPA_LINKDOWN_REASON_NONE); /* @@ -12927,6 +12939,39 @@ static int wait_phys_link_offline_substates(struct hfi1_pportdata *ppd, return read_state; } +/* + * wait_phys_link_out_of_offline - wait for any out of offline state + * @ppd: port device + * @msecs: the number of milliseconds to wait + * + * Wait up to msecs milliseconds for any out of offline physical link + * state change to occur. + * Returns 0 if at least one state is reached, otherwise -ETIMEDOUT. + */ +static int wait_phys_link_out_of_offline(struct hfi1_pportdata *ppd, + int msecs) +{ + u32 read_state; + unsigned long timeout; + + timeout = jiffies + msecs_to_jiffies(msecs); + while (1) { + read_state = read_physical_state(ppd->dd); + if ((read_state & 0xF0) != PLS_OFFLINE) + break; + if (time_after(jiffies, timeout)) { + dd_dev_err(ppd->dd, + "timeout waiting for phy link out of offline. Read state 0x%x, %dms\n", + read_state, msecs); + return -ETIMEDOUT; + } + usleep_range(1950, 2050); /* sleep 2ms-ish */ + } + + log_state_transition(ppd, read_state); + return read_state; +} + #define CLEAR_STATIC_RATE_CONTROL_SMASK(r) \ (r &= ~SEND_CTXT_CHECK_ENABLE_DISALLOW_PBC_STATIC_RATE_CONTROL_SMASK) -- cgit v1.2.3-59-g8ed1b From fe4dd4239277486ca3a468e7bbeafd7ef3a5634e Mon Sep 17 00:00:00 2001 From: Mitko Haralanov Date: Wed, 28 Nov 2018 10:19:15 -0800 Subject: IB/hfi1: Correctly process FECN and BECN in packets A CA is supposed to ignore FECN bits in multicast, ACK, and CNP packets. This patch corrects the behavior of the HFI1 driver in this regard by ignoring FECNs in those packet types. While fixing the above behavior, fix the extraction of the FECN and BECN bits from the packet headers for both 9B and 16B packets. Furthermore, this patch corrects the driver's response to a FECN in RDMA READ RESPONSE packets. Instead of sending an "empty" ACK, the driver now sends a CNP packet. While editing that code path, add the missing trace for CNP packets. Fixes: 88733e3b8450 ("IB/hfi1: Add 16B UD support") Fixes: f59fb9e05109 ("IB/hfi1: Fix handling of FECN marked multicast packet") Reviewed-by: Kaike Wan Reviewed-by: Mike Marciniszyn Reviewed-by: Dennis Dalessandro Signed-off-by: Mitko Haralanov Signed-off-by: Dennis Dalessandro Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hfi1/driver.c | 70 +++++++++++++++++++++++++------------ drivers/infiniband/hw/hfi1/hfi.h | 35 +++++++++++++------ drivers/infiniband/hw/hfi1/rc.c | 30 ++++++---------- drivers/infiniband/hw/hfi1/uc.c | 2 +- drivers/infiniband/hw/hfi1/ud.c | 33 ++++++++++------- 5 files changed, 104 insertions(+), 66 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/hfi1/driver.c b/drivers/infiniband/hw/hfi1/driver.c index a41f85558312..d5277c23cba6 100644 --- a/drivers/infiniband/hw/hfi1/driver.c +++ b/drivers/infiniband/hw/hfi1/driver.c @@ -430,40 +430,60 @@ static const hfi1_handle_cnp hfi1_handle_cnp_tbl[2] = { [HFI1_PKT_TYPE_16B] = &return_cnp_16B }; -void hfi1_process_ecn_slowpath(struct rvt_qp *qp, struct hfi1_packet *pkt, - bool do_cnp) +/** + * hfi1_process_ecn_slowpath - Process FECN or BECN bits + * @qp: The packet's destination QP + * @pkt: The packet itself. + * @prescan: Is the caller the RXQ prescan + * + * Process the packet's FECN or BECN bits. By now, the packet + * has already been evaluated whether processing of those bit should + * be done. + * The significance of the @prescan argument is that if the caller + * is the RXQ prescan, a CNP will be send out instead of waiting for the + * normal packet processing to send an ACK with BECN set (or a CNP). + */ +bool hfi1_process_ecn_slowpath(struct rvt_qp *qp, struct hfi1_packet *pkt, + bool prescan) { struct hfi1_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num); struct hfi1_pportdata *ppd = ppd_from_ibp(ibp); struct ib_other_headers *ohdr = pkt->ohdr; struct ib_grh *grh = pkt->grh; - u32 rqpn = 0, bth1; + u32 rqpn = 0; u16 pkey; u32 rlid, slid, dlid = 0; - u8 hdr_type, sc, svc_type; - bool is_mcast = false; + u8 hdr_type, sc, svc_type, opcode; + bool is_mcast = false, ignore_fecn = false, do_cnp = false, + fecn, becn; /* can be called from prescan */ if (pkt->etype == RHF_RCV_TYPE_BYPASS) { - is_mcast = hfi1_is_16B_mcast(dlid); pkey = hfi1_16B_get_pkey(pkt->hdr); sc = hfi1_16B_get_sc(pkt->hdr); dlid = hfi1_16B_get_dlid(pkt->hdr); slid = hfi1_16B_get_slid(pkt->hdr); + is_mcast = hfi1_is_16B_mcast(dlid); + opcode = ib_bth_get_opcode(ohdr); hdr_type = HFI1_PKT_TYPE_16B; + fecn = hfi1_16B_get_fecn(pkt->hdr); + becn = hfi1_16B_get_becn(pkt->hdr); } else { - is_mcast = (dlid > be16_to_cpu(IB_MULTICAST_LID_BASE)) && - (dlid != be16_to_cpu(IB_LID_PERMISSIVE)); pkey = ib_bth_get_pkey(ohdr); sc = hfi1_9B_get_sc5(pkt->hdr, pkt->rhf); - dlid = ib_get_dlid(pkt->hdr); + dlid = qp->ibqp.qp_type != IB_QPT_UD ? ib_get_dlid(pkt->hdr) : + ppd->lid; slid = ib_get_slid(pkt->hdr); + is_mcast = (dlid > be16_to_cpu(IB_MULTICAST_LID_BASE)) && + (dlid != be16_to_cpu(IB_LID_PERMISSIVE)); + opcode = ib_bth_get_opcode(ohdr); hdr_type = HFI1_PKT_TYPE_9B; + fecn = ib_bth_get_fecn(ohdr); + becn = ib_bth_get_becn(ohdr); } switch (qp->ibqp.qp_type) { case IB_QPT_UD: - dlid = ppd->lid; rlid = slid; rqpn = ib_get_sqpn(pkt->ohdr); svc_type = IB_CC_SVCTYPE_UD; @@ -485,22 +505,31 @@ void hfi1_process_ecn_slowpath(struct rvt_qp *qp, struct hfi1_packet *pkt, svc_type = IB_CC_SVCTYPE_RC; break; default: - return; + return false; } - bth1 = be32_to_cpu(ohdr->bth[1]); + ignore_fecn = is_mcast || (opcode == IB_OPCODE_CNP) || + (opcode == IB_OPCODE_RC_ACKNOWLEDGE); + /* + * ACKNOWLEDGE packets do not get a CNP but this will be + * guarded by ignore_fecn above. + */ + do_cnp = prescan || + (opcode >= IB_OPCODE_RC_RDMA_READ_RESPONSE_FIRST && + opcode <= IB_OPCODE_RC_ATOMIC_ACKNOWLEDGE); + /* Call appropriate CNP handler */ - if (do_cnp && (bth1 & IB_FECN_SMASK)) + if (!ignore_fecn && do_cnp && fecn) hfi1_handle_cnp_tbl[hdr_type](ibp, qp, rqpn, pkey, dlid, rlid, sc, grh); - if (!is_mcast && (bth1 & IB_BECN_SMASK)) { - u32 lqpn = bth1 & RVT_QPN_MASK; + if (becn) { + u32 lqpn = be32_to_cpu(ohdr->bth[1]) & RVT_QPN_MASK; u8 sl = ibp->sc_to_sl[sc]; process_becn(ppd, sl, rlid, lqpn, rqpn, svc_type); } - + return !ignore_fecn && fecn; } struct ps_mdata { @@ -599,7 +628,6 @@ static void __prescan_rxq(struct hfi1_packet *packet) struct rvt_dev_info *rdi = &rcd->dd->verbs_dev.rdi; u64 rhf = rhf_to_cpu(rhf_addr); u32 etype = rhf_rcv_type(rhf), qpn, bth1; - int is_ecn = 0; u8 lnh; if (ps_done(&mdata, rhf, rcd)) @@ -625,12 +653,10 @@ static void __prescan_rxq(struct hfi1_packet *packet) goto next; /* just in case */ } - bth1 = be32_to_cpu(packet->ohdr->bth[1]); - is_ecn = !!(bth1 & (IB_FECN_SMASK | IB_BECN_SMASK)); - - if (!is_ecn) + if (!hfi1_may_ecn(packet)) goto next; + bth1 = be32_to_cpu(packet->ohdr->bth[1]); qpn = bth1 & RVT_QPN_MASK; rcu_read_lock(); qp = rvt_lookup_qpn(rdi, &ibp->rvp, qpn); @@ -640,7 +666,7 @@ static void __prescan_rxq(struct hfi1_packet *packet) goto next; } - process_ecn(qp, packet, true); + hfi1_process_ecn_slowpath(qp, packet, true); rcu_read_unlock(); /* turn off BECN, FECN */ diff --git a/drivers/infiniband/hw/hfi1/hfi.h b/drivers/infiniband/hw/hfi1/hfi.h index 1401b6ea4a28..ca572ad6061c 100644 --- a/drivers/infiniband/hw/hfi1/hfi.h +++ b/drivers/infiniband/hw/hfi1/hfi.h @@ -1802,13 +1802,20 @@ static inline struct hfi1_ibport *rcd_to_iport(struct hfi1_ctxtdata *rcd) return &rcd->ppd->ibport_data; } -void hfi1_process_ecn_slowpath(struct rvt_qp *qp, struct hfi1_packet *pkt, - bool do_cnp); -static inline bool process_ecn(struct rvt_qp *qp, struct hfi1_packet *pkt, - bool do_cnp) +/** + * hfi1_may_ecn - Check whether FECN or BECN processing should be done + * @pkt: the packet to be evaluated + * + * Check whether the FECN or BECN bits in the packet's header are + * enabled, depending on packet type. + * + * This function only checks for FECN and BECN bits. Additional checks + * are done in the slowpath (hfi1_process_ecn_slowpath()) in order to + * ensure correct handling. + */ +static inline bool hfi1_may_ecn(struct hfi1_packet *pkt) { - bool becn; - bool fecn; + bool fecn, becn; if (pkt->etype == RHF_RCV_TYPE_BYPASS) { fecn = hfi1_16B_get_fecn(pkt->hdr); @@ -1817,10 +1824,18 @@ static inline bool process_ecn(struct rvt_qp *qp, struct hfi1_packet *pkt, fecn = ib_bth_get_fecn(pkt->ohdr); becn = ib_bth_get_becn(pkt->ohdr); } - if (unlikely(fecn || becn)) { - hfi1_process_ecn_slowpath(qp, pkt, do_cnp); - return fecn; - } + return fecn || becn; +} + +bool hfi1_process_ecn_slowpath(struct rvt_qp *qp, struct hfi1_packet *pkt, + bool prescan); +static inline bool process_ecn(struct rvt_qp *qp, struct hfi1_packet *pkt) +{ + bool do_work; + + do_work = hfi1_may_ecn(pkt); + if (unlikely(do_work)) + return hfi1_process_ecn_slowpath(qp, pkt, false); return false; } diff --git a/drivers/infiniband/hw/hfi1/rc.c b/drivers/infiniband/hw/hfi1/rc.c index 188aa4f686a0..a677f5a53360 100644 --- a/drivers/infiniband/hw/hfi1/rc.c +++ b/drivers/infiniband/hw/hfi1/rc.c @@ -2049,8 +2049,7 @@ void hfi1_rc_rcv(struct hfi1_packet *packet) struct ib_reth *reth; unsigned long flags; int ret; - bool is_fecn = false; - bool copy_last = false; + bool copy_last = false, fecn; u32 rkey; u8 extra_bytes = pad + packet->extra_byte + (SIZE_OF_CRC << 2); @@ -2059,7 +2058,7 @@ void hfi1_rc_rcv(struct hfi1_packet *packet) if (hfi1_ruc_check_hdr(ibp, packet)) return; - is_fecn = process_ecn(qp, packet, false); + fecn = process_ecn(qp, packet); /* * Process responses (ACKs) before anything else. Note that the @@ -2070,8 +2069,6 @@ void hfi1_rc_rcv(struct hfi1_packet *packet) if (opcode >= OP(RDMA_READ_RESPONSE_FIRST) && opcode <= OP(ATOMIC_ACKNOWLEDGE)) { rc_rcv_resp(packet); - if (is_fecn) - goto send_ack; return; } @@ -2347,11 +2344,11 @@ send_last: /* Schedule the send engine. */ qp->s_flags |= RVT_S_RESP_PENDING; + if (fecn) + qp->s_flags |= RVT_S_ECN; hfi1_schedule_send(qp); spin_unlock_irqrestore(&qp->s_lock, flags); - if (is_fecn) - goto send_ack; return; } @@ -2413,11 +2410,11 @@ send_last: /* Schedule the send engine. */ qp->s_flags |= RVT_S_RESP_PENDING; + if (fecn) + qp->s_flags |= RVT_S_ECN; hfi1_schedule_send(qp); spin_unlock_irqrestore(&qp->s_lock, flags); - if (is_fecn) - goto send_ack; return; } @@ -2430,16 +2427,9 @@ send_last: qp->r_ack_psn = psn; qp->r_nak_state = 0; /* Send an ACK if requested or required. */ - if (psn & IB_BTH_REQ_ACK) { - if (packet->numpkt == 0) { - rc_cancel_ack(qp); - goto send_ack; - } - if (qp->r_adefered >= HFI1_PSN_CREDIT) { - rc_cancel_ack(qp); - goto send_ack; - } - if (unlikely(is_fecn)) { + if (psn & IB_BTH_REQ_ACK || fecn) { + if (packet->numpkt == 0 || fecn || + qp->r_adefered >= HFI1_PSN_CREDIT) { rc_cancel_ack(qp); goto send_ack; } @@ -2480,7 +2470,7 @@ nack_acc: qp->r_nak_state = IB_NAK_REMOTE_ACCESS_ERROR; qp->r_ack_psn = qp->r_psn; send_ack: - hfi1_send_rc_ack(packet, is_fecn); + hfi1_send_rc_ack(packet, fecn); } void hfi1_rc_hdrerr( diff --git a/drivers/infiniband/hw/hfi1/uc.c b/drivers/infiniband/hw/hfi1/uc.c index 6aca0c5a7f97..6ba47037c424 100644 --- a/drivers/infiniband/hw/hfi1/uc.c +++ b/drivers/infiniband/hw/hfi1/uc.c @@ -321,7 +321,7 @@ void hfi1_uc_rcv(struct hfi1_packet *packet) if (hfi1_ruc_check_hdr(ibp, packet)) return; - process_ecn(qp, packet, true); + process_ecn(qp, packet); psn = ib_bth_get_psn(ohdr); /* Compare the PSN verses the expected PSN. */ diff --git a/drivers/infiniband/hw/hfi1/ud.c b/drivers/infiniband/hw/hfi1/ud.c index 4baa8f4d49de..88242fe95eaa 100644 --- a/drivers/infiniband/hw/hfi1/ud.c +++ b/drivers/infiniband/hw/hfi1/ud.c @@ -51,6 +51,7 @@ #include "hfi.h" #include "mad.h" #include "verbs_txreq.h" +#include "trace_ibhdrs.h" #include "qp.h" /* We support only two types - 9B and 16B for now */ @@ -656,18 +657,19 @@ void return_cnp_16B(struct hfi1_ibport *ibp, struct rvt_qp *qp, u32 bth0, plen, vl, hwords = 7; u16 len; u8 l4; - struct hfi1_16b_header hdr; + struct hfi1_opa_header hdr; struct ib_other_headers *ohdr; struct pio_buf *pbuf; struct send_context *ctxt = qp_to_send_context(qp, sc5); struct hfi1_pportdata *ppd = ppd_from_ibp(ibp); u32 nwords; + hdr.hdr_type = HFI1_PKT_TYPE_16B; /* Populate length */ nwords = ((hfi1_get_16b_padding(hwords << 2, 0) + SIZE_OF_LT) >> 2) + SIZE_OF_CRC; if (old_grh) { - struct ib_grh *grh = &hdr.u.l.grh; + struct ib_grh *grh = &hdr.opah.u.l.grh; grh->version_tclass_flow = old_grh->version_tclass_flow; grh->paylen = cpu_to_be16( @@ -675,11 +677,11 @@ void return_cnp_16B(struct hfi1_ibport *ibp, struct rvt_qp *qp, grh->hop_limit = 0xff; grh->sgid = old_grh->dgid; grh->dgid = old_grh->sgid; - ohdr = &hdr.u.l.oth; + ohdr = &hdr.opah.u.l.oth; l4 = OPA_16B_L4_IB_GLOBAL; hwords += sizeof(struct ib_grh) / sizeof(u32); } else { - ohdr = &hdr.u.oth; + ohdr = &hdr.opah.u.oth; l4 = OPA_16B_L4_IB_LOCAL; } @@ -693,7 +695,7 @@ void return_cnp_16B(struct hfi1_ibport *ibp, struct rvt_qp *qp, /* Convert dwords to flits */ len = (hwords + nwords) >> 1; - hfi1_make_16b_hdr(&hdr, slid, dlid, len, pkey, 1, 0, l4, sc5); + hfi1_make_16b_hdr(&hdr.opah, slid, dlid, len, pkey, 1, 0, l4, sc5); plen = 2 /* PBC */ + hwords + nwords; pbc_flags |= PBC_PACKET_BYPASS | PBC_INSERT_BYPASS_ICRC; @@ -701,9 +703,11 @@ void return_cnp_16B(struct hfi1_ibport *ibp, struct rvt_qp *qp, pbc = create_pbc(ppd, pbc_flags, qp->srate_mbps, vl, plen); if (ctxt) { pbuf = sc_buffer_alloc(ctxt, plen, NULL, NULL); - if (pbuf) + if (pbuf) { + trace_pio_output_ibhdr(ppd->dd, &hdr, sc5); ppd->dd->pio_inline_send(ppd->dd, pbuf, pbc, &hdr, hwords); + } } } @@ -715,14 +719,15 @@ void return_cnp(struct hfi1_ibport *ibp, struct rvt_qp *qp, u32 remote_qpn, u32 bth0, plen, vl, hwords = 5; u16 lrh0; u8 sl = ibp->sc_to_sl[sc5]; - struct ib_header hdr; + struct hfi1_opa_header hdr; struct ib_other_headers *ohdr; struct pio_buf *pbuf; struct send_context *ctxt = qp_to_send_context(qp, sc5); struct hfi1_pportdata *ppd = ppd_from_ibp(ibp); + hdr.hdr_type = HFI1_PKT_TYPE_9B; if (old_grh) { - struct ib_grh *grh = &hdr.u.l.grh; + struct ib_grh *grh = &hdr.ibh.u.l.grh; grh->version_tclass_flow = old_grh->version_tclass_flow; grh->paylen = cpu_to_be16( @@ -730,11 +735,11 @@ void return_cnp(struct hfi1_ibport *ibp, struct rvt_qp *qp, u32 remote_qpn, grh->hop_limit = 0xff; grh->sgid = old_grh->dgid; grh->dgid = old_grh->sgid; - ohdr = &hdr.u.l.oth; + ohdr = &hdr.ibh.u.l.oth; lrh0 = HFI1_LRH_GRH; hwords += sizeof(struct ib_grh) / sizeof(u32); } else { - ohdr = &hdr.u.oth; + ohdr = &hdr.ibh.u.oth; lrh0 = HFI1_LRH_BTH; } @@ -746,16 +751,18 @@ void return_cnp(struct hfi1_ibport *ibp, struct rvt_qp *qp, u32 remote_qpn, ohdr->bth[1] = cpu_to_be32(remote_qpn | (1 << IB_BECN_SHIFT)); ohdr->bth[2] = 0; /* PSN 0 */ - hfi1_make_ib_hdr(&hdr, lrh0, hwords + SIZE_OF_CRC, dlid, slid); + hfi1_make_ib_hdr(&hdr.ibh, lrh0, hwords + SIZE_OF_CRC, dlid, slid); plen = 2 /* PBC */ + hwords; pbc_flags |= (ib_is_sc5(sc5) << PBC_DC_INFO_SHIFT); vl = sc_to_vlt(ppd->dd, sc5); pbc = create_pbc(ppd, pbc_flags, qp->srate_mbps, vl, plen); if (ctxt) { pbuf = sc_buffer_alloc(ctxt, plen, NULL, NULL); - if (pbuf) + if (pbuf) { + trace_pio_output_ibhdr(ppd->dd, &hdr, sc5); ppd->dd->pio_inline_send(ppd->dd, pbuf, pbc, &hdr, hwords); + } } } @@ -912,7 +919,7 @@ void hfi1_ud_rcv(struct hfi1_packet *packet) src_qp = hfi1_16B_get_src_qpn(packet->mgmt); } - process_ecn(qp, packet, (opcode != IB_OPCODE_CNP)); + process_ecn(qp, packet); /* * Get the number of bytes the message was padded by * and drop incomplete packets. -- cgit v1.2.3-59-g8ed1b From dd6c6a5a2e1e7be615c81ca6d44c2e89e22cb463 Mon Sep 17 00:00:00 2001 From: "Michael J. Ruhl" Date: Wed, 28 Nov 2018 10:19:25 -0800 Subject: IB/hfi1: Limit VNIC use of SDMA engines to the available count VNIC assumes that all SDMA engines have been configured for use. This is not necessarily true (i.e. if the count was constrained by the module parameter). Update VNICs usage to use the configured count, rather than the hardware count. Reviewed-by: Mike Marciniszyn Reviewed-by: Gary Leshner Signed-off-by: Michael J. Ruhl Signed-off-by: Dennis Dalessandro Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hfi1/vnic_main.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/hfi1/vnic_main.c b/drivers/infiniband/hw/hfi1/vnic_main.c index c9876d9e3cb9..a922db58be14 100644 --- a/drivers/infiniband/hw/hfi1/vnic_main.c +++ b/drivers/infiniband/hw/hfi1/vnic_main.c @@ -816,14 +816,14 @@ struct net_device *hfi1_vnic_alloc_rn(struct ib_device *device, size = sizeof(struct opa_vnic_rdma_netdev) + sizeof(*vinfo); netdev = alloc_netdev_mqs(size, name, name_assign_type, setup, - chip_sdma_engines(dd), dd->num_vnic_contexts); + dd->num_sdma, dd->num_vnic_contexts); if (!netdev) return ERR_PTR(-ENOMEM); rn = netdev_priv(netdev); vinfo = opa_vnic_dev_priv(netdev); vinfo->dd = dd; - vinfo->num_tx_q = chip_sdma_engines(dd); + vinfo->num_tx_q = dd->num_sdma; vinfo->num_rx_q = dd->num_vnic_contexts; vinfo->netdev = netdev; rn->free_rdma_netdev = hfi1_vnic_free_rn; -- cgit v1.2.3-59-g8ed1b From dbc2970caef74e8ff41923d302aa6fb5a4812d0e Mon Sep 17 00:00:00 2001 From: "Michael J. Ruhl" Date: Wed, 28 Nov 2018 10:19:36 -0800 Subject: IB/hfi1: Incorrect sizing of sge for PIO will OOPs An incorrect sge sizing in the HFI PIO path will cause an OOPs similar to this: BUG: unable to handle kernel NULL pointer dereference at (null) IP: [] hfi1_verbs_send_pio+0x3d8/0x530 [hfi1] PGD 0 Oops: 0000 1 SMP Call Trace: ? hfi1_verbs_send_dma+0xad0/0xad0 [hfi1] hfi1_verbs_send+0xdf/0x250 [hfi1] ? make_rc_ack+0xa80/0xa80 [hfi1] hfi1_do_send+0x192/0x430 [hfi1] hfi1_do_send_from_rvt+0x10/0x20 [hfi1] rvt_post_send+0x369/0x820 [rdmavt] ib_uverbs_post_send+0x317/0x570 [ib_uverbs] ib_uverbs_write+0x26f/0x420 [ib_uverbs] ? security_file_permission+0x21/0xa0 vfs_write+0xbd/0x1e0 ? mntput+0x24/0x40 SyS_write+0x7f/0xe0 system_call_fastpath+0x16/0x1b Fix by adding the missing sizing check to correctly determine the sge length. Fixes: 7724105686e7 ("IB/hfi1: add driver files") Reviewed-by: Mike Marciniszyn Signed-off-by: Michael J. Ruhl Signed-off-by: Dennis Dalessandro Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hfi1/verbs.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/hfi1/verbs.c b/drivers/infiniband/hw/hfi1/verbs.c index 48e11e510358..10210208ac7c 100644 --- a/drivers/infiniband/hw/hfi1/verbs.c +++ b/drivers/infiniband/hw/hfi1/verbs.c @@ -919,6 +919,8 @@ int hfi1_verbs_send_pio(struct rvt_qp *qp, struct hfi1_pkt_state *ps, if (slen > len) slen = len; + if (slen > ss->sge.sge_length) + slen = ss->sge.sge_length; rvt_update_sge(ss, slen, false); seg_pio_copy_mid(pbuf, addr, slen); len -= slen; -- cgit v1.2.3-59-g8ed1b From ff8b67fccdb65402df78a1695c38be805252cf8e Mon Sep 17 00:00:00 2001 From: Ashutosh Dixit Date: Wed, 28 Nov 2018 10:19:47 -0800 Subject: IB/hfi1: Consider LMC in 16B/bypass ingress packet check Ingress packet check for 16B/bypass packets should consider the port LMC. Not doing this will result in packets sent to the LMC LIDs getting dropped. The check is implemented in HW for 9B packets. Reviewed-by: Mike Ruhl Signed-off-by: Ashutosh Dixit Signed-off-by: Dennis Dalessandro Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hfi1/driver.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/hfi1/driver.c b/drivers/infiniband/hw/hfi1/driver.c index d5277c23cba6..a8ad70730203 100644 --- a/drivers/infiniband/hw/hfi1/driver.c +++ b/drivers/infiniband/hw/hfi1/driver.c @@ -1426,7 +1426,7 @@ static int hfi1_bypass_ingress_pkt_check(struct hfi1_packet *packet) if ((!(hfi1_is_16B_mcast(packet->dlid))) && (packet->dlid != opa_get_lid(be32_to_cpu(OPA_LID_PERMISSIVE), 16B))) { - if (packet->dlid != ppd->lid) + if ((packet->dlid & ~((1 << ppd->lmc) - 1)) != ppd->lid) return -EINVAL; } -- cgit v1.2.3-59-g8ed1b From ca95f802ef5139722acc8d30aeaab6fe5bbe939e Mon Sep 17 00:00:00 2001 From: Kaike Wan Date: Wed, 28 Nov 2018 10:22:09 -0800 Subject: IB/hfi1: Unreserve a reserved request when it is completed Currently, When a reserved operation is completed, its entry in the send queue will not be unreserved, which leads to the miscalculation of qp->s_avail and thus the triggering of a WARN_ON call trace. This patch fixes the problem by unreserving the reserved operation when it is completed. Fixes: 856cc4c237ad ("IB/hfi1: Add the capability for reserved operations") Reviewed-by: Mike Marciniszyn Signed-off-by: Kaike Wan Signed-off-by: Dennis Dalessandro Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hfi1/rc.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/hfi1/rc.c b/drivers/infiniband/hw/hfi1/rc.c index a677f5a53360..be603f35d7e4 100644 --- a/drivers/infiniband/hw/hfi1/rc.c +++ b/drivers/infiniband/hw/hfi1/rc.c @@ -1157,6 +1157,7 @@ void hfi1_rc_send_complete(struct rvt_qp *qp, struct hfi1_opa_header *opah) if (cmp_psn(wqe->lpsn, qp->s_sending_psn) >= 0 && cmp_psn(qp->s_sending_psn, qp->s_sending_hpsn) <= 0) break; + rvt_qp_wqe_unreserve(qp, wqe); s_last = qp->s_last; trace_hfi1_qp_send_completion(qp, wqe, s_last); if (++s_last >= qp->s_size) @@ -1209,6 +1210,7 @@ static struct rvt_swqe *do_rc_completion(struct rvt_qp *qp, u32 s_last; rvt_put_swqe(wqe); + rvt_qp_wqe_unreserve(qp, wqe); s_last = qp->s_last; trace_hfi1_qp_send_completion(qp, wqe, s_last); if (++s_last >= qp->s_size) -- cgit v1.2.3-59-g8ed1b From d2e9ace47aac92a465c4ad8e0cd1f5f8422a117e Mon Sep 17 00:00:00 2001 From: Kaike Wan Date: Wed, 28 Nov 2018 10:22:20 -0800 Subject: IB/hfi1: Add OPFN and TID RDMA capability bits The OPFN and TID RDMA capability bits are added to allow users to control which feature is enabled and disabled. Reviewed-by: Mike Marciniszyn Signed-off-by: Mitko Haralanov Signed-off-by: Kaike Wan Signed-off-by: Dennis Dalessandro Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hfi1/common.h | 19 +++++++++++-------- include/uapi/rdma/hfi/hfi1_user.h | 6 +++--- 2 files changed, 14 insertions(+), 11 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/hfi1/common.h b/drivers/infiniband/hw/hfi1/common.h index 7108d4d92259..40d3cfb58bd1 100644 --- a/drivers/infiniband/hw/hfi1/common.h +++ b/drivers/infiniband/hw/hfi1/common.h @@ -1,5 +1,5 @@ /* - * Copyright(c) 2015, 2016 Intel Corporation. + * Copyright(c) 2015 - 2018 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or * redistributing this file, you may do so under either license. @@ -136,18 +136,21 @@ HFI1_CAP_ALLOW_PERM_JKEY | \ HFI1_CAP_STATIC_RATE_CTRL | \ HFI1_CAP_PRINT_UNIMPL | \ - HFI1_CAP_TID_UNMAP) + HFI1_CAP_TID_UNMAP | \ + HFI1_CAP_OPFN) /* * A set of capability bits that are "global" and are not allowed to be * set in the user bitmask. */ #define HFI1_CAP_RESERVED_MASK ((HFI1_CAP_SDMA | \ - HFI1_CAP_USE_SDMA_HEAD | \ - HFI1_CAP_EXTENDED_PSN | \ - HFI1_CAP_PRINT_UNIMPL | \ - HFI1_CAP_NO_INTEGRITY | \ - HFI1_CAP_PKEY_CHECK) << \ - HFI1_CAP_USER_SHIFT) + HFI1_CAP_USE_SDMA_HEAD | \ + HFI1_CAP_EXTENDED_PSN | \ + HFI1_CAP_PRINT_UNIMPL | \ + HFI1_CAP_NO_INTEGRITY | \ + HFI1_CAP_PKEY_CHECK | \ + HFI1_CAP_TID_RDMA | \ + HFI1_CAP_OPFN) << \ + HFI1_CAP_USER_SHIFT) /* * Set of capabilities that need to be enabled for kernel context in * order to be allowed for user contexts, as well. diff --git a/include/uapi/rdma/hfi/hfi1_user.h b/include/uapi/rdma/hfi/hfi1_user.h index c6a984c0c881..01ac5853d9ac 100644 --- a/include/uapi/rdma/hfi/hfi1_user.h +++ b/include/uapi/rdma/hfi/hfi1_user.h @@ -6,7 +6,7 @@ * * GPL LICENSE SUMMARY * - * Copyright(c) 2015 Intel Corporation. + * Copyright(c) 2015 - 2018 Intel Corporation. * * This program is free software; you can redistribute it and/or modify * it under the terms of version 2 of the GNU General Public License as @@ -95,7 +95,7 @@ #define HFI1_CAP_SDMA_AHG (1UL << 2) /* Enable SDMA AHG support */ #define HFI1_CAP_EXTENDED_PSN (1UL << 3) /* Enable Extended PSN support */ #define HFI1_CAP_HDRSUPP (1UL << 4) /* Enable Header Suppression */ -/* 1UL << 5 unused */ +#define HFI1_CAP_TID_RDMA (1UL << 5) /* Enable TID RDMA operations */ #define HFI1_CAP_USE_SDMA_HEAD (1UL << 6) /* DMA Hdr Q tail vs. use CSR */ #define HFI1_CAP_MULTI_PKT_EGR (1UL << 7) /* Enable multi-packet Egr buffs*/ #define HFI1_CAP_NODROP_RHQ_FULL (1UL << 8) /* Don't drop on Hdr Q full */ @@ -106,7 +106,7 @@ #define HFI1_CAP_NO_INTEGRITY (1UL << 13) /* Enable ctxt integrity checks */ #define HFI1_CAP_PKEY_CHECK (1UL << 14) /* Enable ctxt PKey checking */ #define HFI1_CAP_STATIC_RATE_CTRL (1UL << 15) /* Allow PBC.StaticRateControl */ -/* 1UL << 16 unused */ +#define HFI1_CAP_OPFN (1UL << 16) /* Enable the OPFN protocol */ #define HFI1_CAP_SDMA_HEAD_CHECK (1UL << 17) /* SDMA head checking */ #define HFI1_CAP_EARLY_CREDIT_RETURN (1UL << 18) /* early credit return */ -- cgit v1.2.3-59-g8ed1b From 5190f052a3654aa1120ea4f9ff3bfac430459893 Mon Sep 17 00:00:00 2001 From: Mike Marciniszyn Date: Wed, 28 Nov 2018 10:22:31 -0800 Subject: IB/hfi1: Allow the driver to initialize QP priv struct This patch adds an interface to allow the driver to initialize the QP priv struct when the QP is created and after the qpn has been assigned. A field is added to the QP priv struct to reference the rcd and two new files are added to contain the function to initialize the rcd field so that more TID RDMA related code can be added here later. Signed-off-by: Mike Marciniszyn Signed-off-by: Kaike Wan Signed-off-by: Dennis Dalessandro Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hfi1/Makefile | 1 + drivers/infiniband/hw/hfi1/tid_rdma.c | 48 +++++++++++++++++++++++++++++++++++ drivers/infiniband/hw/hfi1/tid_rdma.h | 13 ++++++++++ drivers/infiniband/hw/hfi1/verbs.c | 1 + drivers/infiniband/hw/hfi1/verbs.h | 2 ++ drivers/infiniband/sw/rdmavt/qp.c | 9 ++++++- include/rdma/rdma_vt.h | 7 +++++ 7 files changed, 80 insertions(+), 1 deletion(-) create mode 100644 drivers/infiniband/hw/hfi1/tid_rdma.c create mode 100644 drivers/infiniband/hw/hfi1/tid_rdma.h (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/hfi1/Makefile b/drivers/infiniband/hw/hfi1/Makefile index ff790390c91a..3ce9dc8c3463 100644 --- a/drivers/infiniband/hw/hfi1/Makefile +++ b/drivers/infiniband/hw/hfi1/Makefile @@ -34,6 +34,7 @@ hfi1-y := \ ruc.o \ sdma.o \ sysfs.o \ + tid_rdma.o \ trace.o \ uc.o \ ud.o \ diff --git a/drivers/infiniband/hw/hfi1/tid_rdma.c b/drivers/infiniband/hw/hfi1/tid_rdma.c new file mode 100644 index 000000000000..da1ecb68a928 --- /dev/null +++ b/drivers/infiniband/hw/hfi1/tid_rdma.c @@ -0,0 +1,48 @@ +// SPDX-License-Identifier: (GPL-2.0 OR BSD-3-Clause) +/* + * Copyright(c) 2018 Intel Corporation. + * + */ + +#include "hfi.h" +#include "verbs.h" +#include "tid_rdma.h" + +/** + * qp_to_rcd - determine the receive context used by a qp + * @qp - the qp + * + * This routine returns the receive context associated + * with a a qp's qpn. + * + * Returns the context. + */ +static struct hfi1_ctxtdata *qp_to_rcd(struct rvt_dev_info *rdi, + struct rvt_qp *qp) +{ + struct hfi1_ibdev *verbs_dev = container_of(rdi, + struct hfi1_ibdev, + rdi); + struct hfi1_devdata *dd = container_of(verbs_dev, + struct hfi1_devdata, + verbs_dev); + unsigned int ctxt; + + if (qp->ibqp.qp_num == 0) + ctxt = 0; + else + ctxt = ((qp->ibqp.qp_num >> dd->qos_shift) % + (dd->n_krcv_queues - 1)) + 1; + + return dd->rcd[ctxt]; +} + +int hfi1_qp_priv_init(struct rvt_dev_info *rdi, struct rvt_qp *qp, + struct ib_qp_init_attr *init_attr) +{ + struct hfi1_qp_priv *qpriv = qp->priv; + + qpriv->rcd = qp_to_rcd(rdi, qp); + + return 0; +} diff --git a/drivers/infiniband/hw/hfi1/tid_rdma.h b/drivers/infiniband/hw/hfi1/tid_rdma.h new file mode 100644 index 000000000000..6fcd3adcdcc3 --- /dev/null +++ b/drivers/infiniband/hw/hfi1/tid_rdma.h @@ -0,0 +1,13 @@ +/* SPDX-License-Identifier: (GPL-2.0 OR BSD-3-Clause) */ +/* + * Copyright(c) 2018 Intel Corporation. + * + */ +#ifndef HFI1_TID_RDMA_H +#define HFI1_TID_RDMA_H + +int hfi1_qp_priv_init(struct rvt_dev_info *rdi, struct rvt_qp *qp, + struct ib_qp_init_attr *init_attr); + +#endif /* HFI1_TID_RDMA_H */ + diff --git a/drivers/infiniband/hw/hfi1/verbs.c b/drivers/infiniband/hw/hfi1/verbs.c index 10210208ac7c..910147dbed59 100644 --- a/drivers/infiniband/hw/hfi1/verbs.c +++ b/drivers/infiniband/hw/hfi1/verbs.c @@ -1706,6 +1706,7 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd) dd->verbs_dev.rdi.dparms.max_mad_size = OPA_MGMT_MAD_SIZE; dd->verbs_dev.rdi.driver_f.qp_priv_alloc = qp_priv_alloc; + dd->verbs_dev.rdi.driver_f.qp_priv_init = hfi1_qp_priv_init; dd->verbs_dev.rdi.driver_f.qp_priv_free = qp_priv_free; dd->verbs_dev.rdi.driver_f.free_all_qps = free_all_qps; dd->verbs_dev.rdi.driver_f.notify_qp_reset = notify_qp_reset; diff --git a/drivers/infiniband/hw/hfi1/verbs.h b/drivers/infiniband/hw/hfi1/verbs.h index 64c9054db5f3..1ad0b14bdb3c 100644 --- a/drivers/infiniband/hw/hfi1/verbs.h +++ b/drivers/infiniband/hw/hfi1/verbs.h @@ -71,6 +71,7 @@ struct hfi1_devdata; struct hfi1_packet; #include "iowait.h" +#include "tid_rdma.h" #define HFI1_MAX_RDMA_ATOMIC 16 @@ -156,6 +157,7 @@ struct hfi1_qp_priv { struct hfi1_ahg_info *s_ahg; /* ahg info for next header */ struct sdma_engine *s_sde; /* current sde */ struct send_context *s_sendcontext; /* current sendcontext */ + struct hfi1_ctxtdata *rcd; /* QP's receive context */ u8 s_sc; /* SC[0..4] for next packet */ struct iowait s_iowait; struct rvt_qp *owner; diff --git a/drivers/infiniband/sw/rdmavt/qp.c b/drivers/infiniband/sw/rdmavt/qp.c index 1735deb1a9d4..a1bd8cfc2c25 100644 --- a/drivers/infiniband/sw/rdmavt/qp.c +++ b/drivers/infiniband/sw/rdmavt/qp.c @@ -1,5 +1,5 @@ /* - * Copyright(c) 2016, 2017 Intel Corporation. + * Copyright(c) 2016 - 2018 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or * redistributing this file, you may do so under either license. @@ -1094,6 +1094,13 @@ struct ib_qp *rvt_create_qp(struct ib_pd *ibpd, qp->ibqp.qp_num = err; qp->port_num = init_attr->port_num; rvt_init_qp(rdi, qp, init_attr->qp_type); + if (rdi->driver_f.qp_priv_init) { + err = rdi->driver_f.qp_priv_init(rdi, qp, init_attr); + if (err) { + ret = ERR_PTR(err); + goto bail_rq_wq; + } + } break; default: diff --git a/include/rdma/rdma_vt.h b/include/rdma/rdma_vt.h index 3584d0816fcd..dd0ed8048bb4 100644 --- a/include/rdma/rdma_vt.h +++ b/include/rdma/rdma_vt.h @@ -268,6 +268,13 @@ struct rvt_driver_provided { */ void * (*qp_priv_alloc)(struct rvt_dev_info *rdi, struct rvt_qp *qp); + /* + * Init a struture allocated with qp_priv_alloc(). This should be + * called after all qp fields have been initialized in rdmavt. + */ + int (*qp_priv_init)(struct rvt_dev_info *rdi, struct rvt_qp *qp, + struct ib_qp_init_attr *init_attr); + /* * Free the driver's private qp structure. */ -- cgit v1.2.3-59-g8ed1b From 18912c4524385dd6532c682cb9d4f6aa39ba8d47 Mon Sep 17 00:00:00 2001 From: Mike Marciniszyn Date: Wed, 28 Nov 2018 10:32:48 -0800 Subject: IB/hfi1: Close VNIC sdma_progress sleep window The call to sdma_progress() is called outside the wait lock. In this case, there is a race condition where sdma_progress() can return false and the sdma_engine can idle. If that happens, there will be no more sdma interrupts to cause the wakeup and the vnic_sdma xmit will hang. Fix by moving the lock to enclose the sdma_progress() call. Also, delete the tx_retry. The need for this was removed by: commit bcad29137a97 ("IB/hfi1: Serve the most starved iowait entry first") Fixes: 64551ede6cd1 ("IB/hfi1: VNIC SDMA support") Reviewed-by: Gary Leshner Signed-off-by: Mike Marciniszyn Signed-off-by: Dennis Dalessandro Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hfi1/vnic_sdma.c | 15 +++++---------- 1 file changed, 5 insertions(+), 10 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/hfi1/vnic_sdma.c b/drivers/infiniband/hw/hfi1/vnic_sdma.c index 97bd940a056a..77a344792c8a 100644 --- a/drivers/infiniband/hw/hfi1/vnic_sdma.c +++ b/drivers/infiniband/hw/hfi1/vnic_sdma.c @@ -57,7 +57,6 @@ #define HFI1_VNIC_TXREQ_NAME_LEN 32 #define HFI1_VNIC_SDMA_DESC_WTRMRK 64 -#define HFI1_VNIC_SDMA_RETRY_COUNT 1 /* * struct vnic_txreq - VNIC transmit descriptor @@ -67,7 +66,6 @@ * @pad: pad buffer * @plen: pad length * @pbc_val: pbc value - * @retry_count: tx retry count */ struct vnic_txreq { struct sdma_txreq txreq; @@ -77,8 +75,6 @@ struct vnic_txreq { unsigned char pad[HFI1_VNIC_MAX_PAD]; u16 plen; __le64 pbc_val; - - u32 retry_count; }; static void vnic_sdma_complete(struct sdma_txreq *txreq, @@ -196,7 +192,6 @@ int hfi1_vnic_send_dma(struct hfi1_devdata *dd, u8 q_idx, ret = build_vnic_tx_desc(sde, tx, pbc); if (unlikely(ret)) goto free_desc; - tx->retry_count = 0; ret = sdma_send_txreq(sde, iowait_get_ib_work(&vnic_sdma->wait), &tx->txreq, vnic_sdma->pkts_sent); @@ -238,14 +233,14 @@ static int hfi1_vnic_sdma_sleep(struct sdma_engine *sde, struct hfi1_vnic_sdma *vnic_sdma = container_of(wait->iow, struct hfi1_vnic_sdma, wait); struct hfi1_ibdev *dev = &vnic_sdma->dd->verbs_dev; - struct vnic_txreq *tx = container_of(txreq, struct vnic_txreq, txreq); - if (sdma_progress(sde, seq, txreq)) - if (tx->retry_count++ < HFI1_VNIC_SDMA_RETRY_COUNT) - return -EAGAIN; + write_seqlock(&dev->iowait_lock); + if (sdma_progress(sde, seq, txreq)) { + write_sequnlock(&dev->iowait_lock); + return -EAGAIN; + } vnic_sdma->state = HFI1_VNIC_SDMA_Q_DEFERRED; - write_seqlock(&dev->iowait_lock); if (list_empty(&vnic_sdma->wait.list)) iowait_queue(pkts_sent, wait->iow, &sde->dmawait); write_sequnlock(&dev->iowait_lock); -- cgit v1.2.3-59-g8ed1b From 9aefcabe579bca06325ad9e577a36816f57386ff Mon Sep 17 00:00:00 2001 From: Mike Marciniszyn Date: Wed, 28 Nov 2018 10:33:00 -0800 Subject: IB/hfi1: Reduce lock contention on iowait_lock for sdma and pio Commit 4e045572e2c2 ("IB/hfi1: Add unique txwait_lock for txreq events") laid the ground work to support per resource waiting locking. This patch adds that with a lock unique to each sdma engine and pio sendcontext and makes necessary changes for verbs, PSM, and vnic to use the new locks. This is particularly beneficial for smaller messages that will exhaust resources at a faster rate. Fixes: 7724105686e7 ("IB/hfi1: add driver files") Reviewed-by: Gary Leshner Signed-off-by: Mike Marciniszyn Signed-off-by: Dennis Dalessandro Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hfi1/pio.c | 6 +++--- drivers/infiniband/hw/hfi1/pio.h | 2 ++ drivers/infiniband/hw/hfi1/qp.c | 20 ++++++++------------ drivers/infiniband/hw/hfi1/sdma.c | 10 +++++----- drivers/infiniband/hw/hfi1/sdma.h | 1 + drivers/infiniband/hw/hfi1/user_sdma.c | 5 ++--- drivers/infiniband/hw/hfi1/verbs.c | 7 +++---- drivers/infiniband/hw/hfi1/vnic_sdma.c | 7 +++---- 8 files changed, 27 insertions(+), 31 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/hfi1/pio.c b/drivers/infiniband/hw/hfi1/pio.c index 6d5d0d017e6b..dd5a5c030066 100644 --- a/drivers/infiniband/hw/hfi1/pio.c +++ b/drivers/infiniband/hw/hfi1/pio.c @@ -742,6 +742,7 @@ struct send_context *sc_alloc(struct hfi1_devdata *dd, int type, spin_lock_init(&sc->alloc_lock); spin_lock_init(&sc->release_lock); spin_lock_init(&sc->credit_ctrl_lock); + seqlock_init(&sc->waitlock); INIT_LIST_HEAD(&sc->piowait); INIT_WORK(&sc->halt_work, sc_halted); init_waitqueue_head(&sc->halt_wait); @@ -1593,7 +1594,6 @@ void hfi1_sc_wantpiobuf_intr(struct send_context *sc, u32 needint) static void sc_piobufavail(struct send_context *sc) { struct hfi1_devdata *dd = sc->dd; - struct hfi1_ibdev *dev = &dd->verbs_dev; struct list_head *list; struct rvt_qp *qps[PIO_WAIT_BATCH_SIZE]; struct rvt_qp *qp; @@ -1612,7 +1612,7 @@ static void sc_piobufavail(struct send_context *sc) * could end up with QPs on the wait list with the interrupt * disabled. */ - write_seqlock_irqsave(&dev->iowait_lock, flags); + write_seqlock_irqsave(&sc->waitlock, flags); while (!list_empty(list)) { struct iowait *wait; @@ -1636,7 +1636,7 @@ static void sc_piobufavail(struct send_context *sc) if (!list_empty(list)) hfi1_sc_wantpiobuf_intr(sc, 1); } - write_sequnlock_irqrestore(&dev->iowait_lock, flags); + write_sequnlock_irqrestore(&sc->waitlock, flags); /* Wake up the most starved one first */ if (n) diff --git a/drivers/infiniband/hw/hfi1/pio.h b/drivers/infiniband/hw/hfi1/pio.h index bf1afb0077a7..c9a58b642bdd 100644 --- a/drivers/infiniband/hw/hfi1/pio.h +++ b/drivers/infiniband/hw/hfi1/pio.h @@ -127,6 +127,8 @@ struct send_context { volatile __le64 *hw_free; /* HW free counter */ /* list for PIO waiters */ struct list_head piowait ____cacheline_aligned_in_smp; + seqlock_t waitlock; + spinlock_t credit_ctrl_lock ____cacheline_aligned_in_smp; u32 credit_intr_count; /* count of credit intr users */ u64 credit_ctrl; /* cache for credit control */ diff --git a/drivers/infiniband/hw/hfi1/qp.c b/drivers/infiniband/hw/hfi1/qp.c index 6f3bc4dab858..e32fbfe029bc 100644 --- a/drivers/infiniband/hw/hfi1/qp.c +++ b/drivers/infiniband/hw/hfi1/qp.c @@ -368,20 +368,18 @@ bool _hfi1_schedule_send(struct rvt_qp *qp) static void qp_pio_drain(struct rvt_qp *qp) { - struct hfi1_ibdev *dev; struct hfi1_qp_priv *priv = qp->priv; if (!priv->s_sendcontext) return; - dev = to_idev(qp->ibqp.device); while (iowait_pio_pending(&priv->s_iowait)) { - write_seqlock_irq(&dev->iowait_lock); + write_seqlock_irq(&priv->s_sendcontext->waitlock); hfi1_sc_wantpiobuf_intr(priv->s_sendcontext, 1); - write_sequnlock_irq(&dev->iowait_lock); + write_sequnlock_irq(&priv->s_sendcontext->waitlock); iowait_pio_drain(&priv->s_iowait); - write_seqlock_irq(&dev->iowait_lock); + write_seqlock_irq(&priv->s_sendcontext->waitlock); hfi1_sc_wantpiobuf_intr(priv->s_sendcontext, 0); - write_sequnlock_irq(&dev->iowait_lock); + write_sequnlock_irq(&priv->s_sendcontext->waitlock); } } @@ -452,7 +450,6 @@ static int iowait_sleep( struct hfi1_qp_priv *priv; unsigned long flags; int ret = 0; - struct hfi1_ibdev *dev; qp = tx->qp; priv = qp->priv; @@ -465,9 +462,8 @@ static int iowait_sleep( * buffer and undoing the side effects of the copy. */ /* Make a common routine? */ - dev = &sde->dd->verbs_dev; list_add_tail(&stx->list, &wait->tx_head); - write_seqlock(&dev->iowait_lock); + write_seqlock(&sde->waitlock); if (sdma_progress(sde, seq, stx)) goto eagain; if (list_empty(&priv->s_iowait.list)) { @@ -478,11 +474,11 @@ static int iowait_sleep( qp->s_flags |= RVT_S_WAIT_DMA_DESC; iowait_queue(pkts_sent, &priv->s_iowait, &sde->dmawait); - priv->s_iowait.lock = &dev->iowait_lock; + priv->s_iowait.lock = &sde->waitlock; trace_hfi1_qpsleep(qp, RVT_S_WAIT_DMA_DESC); rvt_get_qp(qp); } - write_sequnlock(&dev->iowait_lock); + write_sequnlock(&sde->waitlock); hfi1_qp_unbusy(qp, wait); spin_unlock_irqrestore(&qp->s_lock, flags); ret = -EBUSY; @@ -492,7 +488,7 @@ static int iowait_sleep( } return ret; eagain: - write_sequnlock(&dev->iowait_lock); + write_sequnlock(&sde->waitlock); spin_unlock_irqrestore(&qp->s_lock, flags); list_del_init(&stx->list); return -EAGAIN; diff --git a/drivers/infiniband/hw/hfi1/sdma.c b/drivers/infiniband/hw/hfi1/sdma.c index 891d2386d1ca..b84356e1a4c1 100644 --- a/drivers/infiniband/hw/hfi1/sdma.c +++ b/drivers/infiniband/hw/hfi1/sdma.c @@ -1424,6 +1424,7 @@ int sdma_init(struct hfi1_devdata *dd, u8 port) seqlock_init(&sde->head_lock); spin_lock_init(&sde->senddmactrl_lock); spin_lock_init(&sde->flushlist_lock); + seqlock_init(&sde->waitlock); /* insure there is always a zero bit */ sde->ahg_bits = 0xfffffffe00000000ULL; @@ -1758,7 +1759,6 @@ static void sdma_desc_avail(struct sdma_engine *sde, uint avail) struct iowait *wait, *nw; struct iowait *waits[SDMA_WAIT_BATCH_SIZE]; uint i, n = 0, seq, max_idx = 0; - struct hfi1_ibdev *dev = &sde->dd->verbs_dev; u8 max_starved_cnt = 0; #ifdef CONFIG_SDMA_VERBOSITY @@ -1768,10 +1768,10 @@ static void sdma_desc_avail(struct sdma_engine *sde, uint avail) #endif do { - seq = read_seqbegin(&dev->iowait_lock); + seq = read_seqbegin(&sde->waitlock); if (!list_empty(&sde->dmawait)) { /* at least one item */ - write_seqlock(&dev->iowait_lock); + write_seqlock(&sde->waitlock); /* Harvest waiters wanting DMA descriptors */ list_for_each_entry_safe( wait, @@ -1794,10 +1794,10 @@ static void sdma_desc_avail(struct sdma_engine *sde, uint avail) list_del_init(&wait->list); waits[n++] = wait; } - write_sequnlock(&dev->iowait_lock); + write_sequnlock(&sde->waitlock); break; } - } while (read_seqretry(&dev->iowait_lock, seq)); + } while (read_seqretry(&sde->waitlock, seq)); /* Schedule the most starved one first */ if (n) diff --git a/drivers/infiniband/hw/hfi1/sdma.h b/drivers/infiniband/hw/hfi1/sdma.h index 6dc63d7c5685..1e2e40f79cb2 100644 --- a/drivers/infiniband/hw/hfi1/sdma.h +++ b/drivers/infiniband/hw/hfi1/sdma.h @@ -382,6 +382,7 @@ struct sdma_engine { u64 progress_int_cnt; /* private: */ + seqlock_t waitlock; struct list_head dmawait; /* CONFIG SDMA for now, just blindly duplicate */ diff --git a/drivers/infiniband/hw/hfi1/user_sdma.c b/drivers/infiniband/hw/hfi1/user_sdma.c index 3f0aadccd9f6..e5e7fad09f32 100644 --- a/drivers/infiniband/hw/hfi1/user_sdma.c +++ b/drivers/infiniband/hw/hfi1/user_sdma.c @@ -130,7 +130,6 @@ static int defer_packet_queue( { struct hfi1_user_sdma_pkt_q *pq = container_of(wait->iow, struct hfi1_user_sdma_pkt_q, busy); - struct hfi1_ibdev *dev = &pq->dd->verbs_dev; struct user_sdma_txreq *tx = container_of(txreq, struct user_sdma_txreq, txreq); @@ -144,10 +143,10 @@ static int defer_packet_queue( * it is supposed to be enqueued. */ xchg(&pq->state, SDMA_PKT_Q_DEFERRED); - write_seqlock(&dev->iowait_lock); + write_seqlock(&sde->waitlock); if (list_empty(&pq->busy.list)) iowait_queue(pkts_sent, &pq->busy, &sde->dmawait); - write_sequnlock(&dev->iowait_lock); + write_sequnlock(&sde->waitlock); return -EBUSY; eagain: return -EAGAIN; diff --git a/drivers/infiniband/hw/hfi1/verbs.c b/drivers/infiniband/hw/hfi1/verbs.c index 910147dbed59..28f754d2a844 100644 --- a/drivers/infiniband/hw/hfi1/verbs.c +++ b/drivers/infiniband/hw/hfi1/verbs.c @@ -765,7 +765,6 @@ static int pio_wait(struct rvt_qp *qp, { struct hfi1_qp_priv *priv = qp->priv; struct hfi1_devdata *dd = sc->dd; - struct hfi1_ibdev *dev = &dd->verbs_dev; unsigned long flags; int ret = 0; @@ -777,7 +776,7 @@ static int pio_wait(struct rvt_qp *qp, */ spin_lock_irqsave(&qp->s_lock, flags); if (ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK) { - write_seqlock(&dev->iowait_lock); + write_seqlock(&sc->waitlock); list_add_tail(&ps->s_txreq->txreq.list, &ps->wait->tx_head); if (list_empty(&priv->s_iowait.list)) { @@ -790,14 +789,14 @@ static int pio_wait(struct rvt_qp *qp, was_empty = list_empty(&sc->piowait); iowait_queue(ps->pkts_sent, &priv->s_iowait, &sc->piowait); - priv->s_iowait.lock = &dev->iowait_lock; + priv->s_iowait.lock = &sc->waitlock; trace_hfi1_qpsleep(qp, RVT_S_WAIT_PIO); rvt_get_qp(qp); /* counting: only call wantpiobuf_intr if first user */ if (was_empty) hfi1_sc_wantpiobuf_intr(sc, 1); } - write_sequnlock(&dev->iowait_lock); + write_sequnlock(&sc->waitlock); hfi1_qp_unbusy(qp, ps->wait); ret = -EBUSY; } diff --git a/drivers/infiniband/hw/hfi1/vnic_sdma.c b/drivers/infiniband/hw/hfi1/vnic_sdma.c index 77a344792c8a..1f81c480e028 100644 --- a/drivers/infiniband/hw/hfi1/vnic_sdma.c +++ b/drivers/infiniband/hw/hfi1/vnic_sdma.c @@ -232,18 +232,17 @@ static int hfi1_vnic_sdma_sleep(struct sdma_engine *sde, { struct hfi1_vnic_sdma *vnic_sdma = container_of(wait->iow, struct hfi1_vnic_sdma, wait); - struct hfi1_ibdev *dev = &vnic_sdma->dd->verbs_dev; - write_seqlock(&dev->iowait_lock); + write_seqlock(&sde->waitlock); if (sdma_progress(sde, seq, txreq)) { - write_sequnlock(&dev->iowait_lock); + write_sequnlock(&sde->waitlock); return -EAGAIN; } vnic_sdma->state = HFI1_VNIC_SDMA_Q_DEFERRED; if (list_empty(&vnic_sdma->wait.list)) iowait_queue(pkts_sent, wait->iow, &sde->dmawait); - write_sequnlock(&dev->iowait_lock); + write_sequnlock(&sde->waitlock); return -EBUSY; } -- cgit v1.2.3-59-g8ed1b From 0fbc9b8b4ea3f688a5da141a64f97aa33ad02ae9 Mon Sep 17 00:00:00 2001 From: Qian Cai Date: Thu, 29 Nov 2018 21:18:07 -0500 Subject: mlx4: Use snprintf instead of complicated strcpy This fixes a compilation warning in sysfs.c drivers/infiniband/hw/mlx4/sysfs.c:360:2: warning: 'strncpy' output may be truncated copying 8 bytes from a string of length 31 [-Wstringop-truncation] By eliminating the temporary stack buffer. Signed-off-by: Qian Cai Reviewed-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx4/sysfs.c | 12 ++++-------- 1 file changed, 4 insertions(+), 8 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx4/sysfs.c b/drivers/infiniband/hw/mlx4/sysfs.c index 752bdd536130..ea1f3a081b05 100644 --- a/drivers/infiniband/hw/mlx4/sysfs.c +++ b/drivers/infiniband/hw/mlx4/sysfs.c @@ -353,16 +353,12 @@ err: static void get_name(struct mlx4_ib_dev *dev, char *name, int i, int max) { - char base_name[9]; - - /* pci_name format is: bus:dev:func -> xxxx:yy:zz.n */ - strlcpy(name, pci_name(dev->dev->persist->pdev), max); - strncpy(base_name, name, 8); /*till xxxx:yy:*/ - base_name[8] = '\0'; - /* with no ARI only 3 last bits are used so when the fn is higher than 8 + /* pci_name format is: bus:dev:func -> xxxx:yy:zz.n + * with no ARI only 3 last bits are used so when the fn is higher than 8 * need to add it to the dev num, so count in the last number will be * modulo 8 */ - sprintf(name, "%s%.2d.%d", base_name, (i/8), (i%8)); + snprintf(name, max, "%.8s%.2d.%d", pci_name(dev->dev->persist->pdev), + i / 8, i % 8); } struct mlx4_port { -- cgit v1.2.3-59-g8ed1b From 419822c8b83f37ea4934d5583b5c5fbe9493cfcf Mon Sep 17 00:00:00 2001 From: Alex Vesker Date: Tue, 4 Dec 2018 15:34:05 +0200 Subject: IB/mlx5: Enable TX on a DEVX flow table Flow table can be passed as a DEVX object which is a valid destination in an EGRESS flow. Fix the original code to allow that. Fixes: a7ee18bdee83 ("RDMA/mlx5: Allow creating a matcher for a NIC TX flow table") Signed-off-by: Alex Vesker Reviewed-by: Yishai Hadas Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/flow.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/flow.c b/drivers/infiniband/hw/mlx5/flow.c index 793afa404442..b73756bd38bd 100644 --- a/drivers/infiniband/hw/mlx5/flow.c +++ b/drivers/infiniband/hw/mlx5/flow.c @@ -93,10 +93,6 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_CREATE_FLOW)( ((dest_devx && dest_qp) || (!dest_devx && !dest_qp))) return -EINVAL; - if (fs_matcher->ns_type == MLX5_FLOW_NAMESPACE_EGRESS && - (dest_devx || dest_qp)) - return -EINVAL; - if (dest_devx) { devx_obj = uverbs_attr_get_obj( attrs, MLX5_IB_ATTR_CREATE_FLOW_DEST_DEVX); @@ -141,6 +137,10 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_CREATE_FLOW)( if (dev->rep) return -ENOTSUPP; + if (dest_type == MLX5_FLOW_DESTINATION_TYPE_TIR && + fs_matcher->ns_type == MLX5_FLOW_NAMESPACE_EGRESS) + return -EINVAL; + cmd_in = uverbs_attr_get_alloced_ptr( attrs, MLX5_IB_ATTR_CREATE_FLOW_MATCH_VALUE); inlen = uverbs_attr_get_len(attrs, -- cgit v1.2.3-59-g8ed1b From 569c665150156e44ecbd92af47a6d3fd4e2e4690 Mon Sep 17 00:00:00 2001 From: Danit Goldberg Date: Fri, 30 Nov 2018 13:22:05 +0200 Subject: IB/mlx5: Add packet based credit mode support The device can support two credit modes, message based (default) and packet based. In order to enable packet based mode, the QP should be created with special flag that indicates this. This patch adds support for the new DV QP creation flag that can be used for RC QPs in order to change the credit mode. Signed-off-by: Danit Goldberg Reviewed-by: Majd Dibbiny Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/mlx5_ib.h | 1 + drivers/infiniband/hw/mlx5/qp.c | 15 +++++++++++++-- include/uapi/rdma/mlx5-abi.h | 1 + 3 files changed, 15 insertions(+), 2 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index 861b68f2e330..3e034bc85bde 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -461,6 +461,7 @@ enum mlx5_ib_qp_flags { MLX5_IB_QP_UNDERLAY = 1 << 10, MLX5_IB_QP_PCI_WRITE_END_PADDING = 1 << 11, MLX5_IB_QP_TUNNEL_OFFLOAD = 1 << 12, + MLX5_IB_QP_PACKET_BASED_CREDIT = 1 << 13, }; struct mlx5_umr_wr { diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c index 6841c0f9237f..d5095fcd4cda 100644 --- a/drivers/infiniband/hw/mlx5/qp.c +++ b/drivers/infiniband/hw/mlx5/qp.c @@ -1889,7 +1889,8 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd, MLX5_QP_FLAG_BFREG_INDEX | MLX5_QP_FLAG_TYPE_DCT | MLX5_QP_FLAG_TYPE_DCI | - MLX5_QP_FLAG_ALLOW_SCATTER_CQE)) + MLX5_QP_FLAG_ALLOW_SCATTER_CQE | + MLX5_QP_FLAG_PACKET_BASED_CREDIT_MODE)) return -EINVAL; err = get_qp_user_index(to_mucontext(pd->uobject->context), @@ -1925,6 +1926,15 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd, qp->flags_en |= MLX5_QP_FLAG_TIR_ALLOW_SELF_LB_MC; } + if (ucmd.flags & MLX5_QP_FLAG_PACKET_BASED_CREDIT_MODE) { + if (init_attr->qp_type != IB_QPT_RC || + !MLX5_CAP_GEN(dev->mdev, qp_packet_based)) { + mlx5_ib_dbg(dev, "packet based credit mode isn't supported\n"); + return -EOPNOTSUPP; + } + qp->flags |= MLX5_IB_QP_PACKET_BASED_CREDIT; + } + if (init_attr->create_flags & IB_QP_CREATE_SOURCE_QPN) { if (init_attr->qp_type != IB_QPT_UD || (MLX5_CAP_GEN(dev->mdev, port_type) != @@ -2021,7 +2031,8 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd, MLX5_SET(qpc, qpc, cd_slave_send, 1); if (qp->flags & MLX5_IB_QP_MANAGED_RECV) MLX5_SET(qpc, qpc, cd_slave_receive, 1); - + if (qp->flags & MLX5_IB_QP_PACKET_BASED_CREDIT) + MLX5_SET(qpc, qpc, req_e2e_credit_mode, 1); if (qp->scat_cqe && is_connected(init_attr->qp_type)) { configure_responder_scat_cqe(init_attr, qpc); configure_requester_scat_cqe(dev, init_attr, diff --git a/include/uapi/rdma/mlx5-abi.h b/include/uapi/rdma/mlx5-abi.h index 8fa9f90e2bb1..4af581170f62 100644 --- a/include/uapi/rdma/mlx5-abi.h +++ b/include/uapi/rdma/mlx5-abi.h @@ -48,6 +48,7 @@ enum { MLX5_QP_FLAG_TIR_ALLOW_SELF_LB_UC = 1 << 6, MLX5_QP_FLAG_TIR_ALLOW_SELF_LB_MC = 1 << 7, MLX5_QP_FLAG_ALLOW_SCATTER_CQE = 1 << 8, + MLX5_QP_FLAG_PACKET_BASED_CREDIT_MODE = 1 << 9, }; enum { -- cgit v1.2.3-59-g8ed1b From 7e11b911b520de6a3189fafa94740f5fde2a2c98 Mon Sep 17 00:00:00 2001 From: Danit Goldberg Date: Fri, 30 Nov 2018 13:22:06 +0200 Subject: IB/mlx5: Report packet based credit mode device capability Report packet based credit mode capability via the mlx5 DV interface. Signed-off-by: Danit Goldberg Reviewed-by: Majd Dibbiny Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/main.c | 3 +++ include/uapi/rdma/mlx5-abi.h | 1 + 2 files changed, 4 insertions(+) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index 96515a8c9d2c..f985d0d9b883 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -1018,6 +1018,9 @@ static int mlx5_ib_query_device(struct ib_device *ibdev, if (MLX5_CAP_GEN(mdev, cqe_128_always)) resp.flags |= MLX5_IB_QUERY_DEV_RESP_FLAGS_CQE_128B_PAD; + if (MLX5_CAP_GEN(mdev, qp_packet_based)) + resp.flags |= + MLX5_IB_QUERY_DEV_RESP_PACKET_BASED_CREDIT_MODE; } if (field_avail(typeof(resp), sw_parsing_caps, diff --git a/include/uapi/rdma/mlx5-abi.h b/include/uapi/rdma/mlx5-abi.h index 4af581170f62..87b3198f4b5d 100644 --- a/include/uapi/rdma/mlx5-abi.h +++ b/include/uapi/rdma/mlx5-abi.h @@ -237,6 +237,7 @@ enum mlx5_ib_query_dev_resp_flags { /* Support 128B CQE compression */ MLX5_IB_QUERY_DEV_RESP_FLAGS_CQE_128B_COMP = 1 << 0, MLX5_IB_QUERY_DEV_RESP_FLAGS_CQE_128B_PAD = 1 << 1, + MLX5_IB_QUERY_DEV_RESP_PACKET_BASED_CREDIT_MODE = 1 << 2, }; enum mlx5_ib_tunnel_offloads { -- cgit v1.2.3-59-g8ed1b From 9f5071010347b2134e1139dca2396a0cea8130dc Mon Sep 17 00:00:00 2001 From: Lijun Ou Date: Sat, 8 Dec 2018 18:40:07 +0800 Subject: RDMA/hns: Init qp context when modify qp from reset to init It needs to clear qp context previous when init qp context. Otherwise, the newly created qp context residue has the contents of the qp context before the uninstall, and the qp context content is disordered, causing the task to fail. Signed-off-by: Lijun Ou Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 1 + 1 file changed, 1 insertion(+) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c index 835b78371294..1cffe74b5fc8 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c @@ -3695,6 +3695,7 @@ static int hns_roce_v2_modify_qp(struct ib_qp *ibqp, */ memset(qpc_mask, 0xff, sizeof(*qpc_mask)); if (cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT) { + memset(qpc_mask, 0, sizeof(*qpc_mask)); modify_qp_reset_to_init(ibqp, attr, attr_mask, context, qpc_mask); } else if (cur_state == IB_QPS_INIT && new_state == IB_QPS_INIT) { -- cgit v1.2.3-59-g8ed1b From 233673e422de70d6ba32e7b41dd75e47f036d03b Mon Sep 17 00:00:00 2001 From: Lijun Ou Date: Sat, 8 Dec 2018 18:40:08 +0800 Subject: RDMA/hns: Encapsulate and simplify qp state transition This patch move the codes of qp state transition into the new function as well as simplify the logic for other qp states transition. Signed-off-by: Lijun Ou Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 31 +++++++++++++++--------------- 1 file changed, 16 insertions(+), 15 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c index 1cffe74b5fc8..ca71e1079d8a 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c @@ -3670,6 +3670,21 @@ static int modify_qp_rtr_to_rts(struct ib_qp *ibqp, return 0; } +static inline bool hns_roce_v2_check_qp_stat(enum ib_qp_state cur_state, + enum ib_qp_state new_state) +{ + + if ((cur_state != IB_QPS_RESET && + (new_state == IB_QPS_ERR || new_state == IB_QPS_RESET)) || + ((cur_state == IB_QPS_RTS || cur_state == IB_QPS_SQD) && + (new_state == IB_QPS_RTS || new_state == IB_QPS_SQD)) || + (cur_state == IB_QPS_SQE && new_state == IB_QPS_RTS)) + return true; + + return false; + +} + static int hns_roce_v2_modify_qp(struct ib_qp *ibqp, const struct ib_qp_attr *attr, int attr_mask, enum ib_qp_state cur_state, @@ -3711,21 +3726,7 @@ static int hns_roce_v2_modify_qp(struct ib_qp *ibqp, qpc_mask); if (ret) goto out; - } else if ((cur_state == IB_QPS_RTS && new_state == IB_QPS_RTS) || - (cur_state == IB_QPS_SQE && new_state == IB_QPS_RTS) || - (cur_state == IB_QPS_RTS && new_state == IB_QPS_SQD) || - (cur_state == IB_QPS_SQD && new_state == IB_QPS_SQD) || - (cur_state == IB_QPS_SQD && new_state == IB_QPS_RTS) || - (cur_state == IB_QPS_INIT && new_state == IB_QPS_RESET) || - (cur_state == IB_QPS_RTR && new_state == IB_QPS_RESET) || - (cur_state == IB_QPS_RTS && new_state == IB_QPS_RESET) || - (cur_state == IB_QPS_ERR && new_state == IB_QPS_RESET) || - (cur_state == IB_QPS_INIT && new_state == IB_QPS_ERR) || - (cur_state == IB_QPS_RTR && new_state == IB_QPS_ERR) || - (cur_state == IB_QPS_RTS && new_state == IB_QPS_ERR) || - (cur_state == IB_QPS_SQD && new_state == IB_QPS_ERR) || - (cur_state == IB_QPS_SQE && new_state == IB_QPS_ERR) || - (cur_state == IB_QPS_ERR && new_state == IB_QPS_ERR)) { + } else if (hns_roce_v2_check_qp_stat(cur_state, new_state)) { /* Nothing */ ; } else { -- cgit v1.2.3-59-g8ed1b From 4af07f01f7a787ba5158352b98c9e3cb74995a1c Mon Sep 17 00:00:00 2001 From: Lijun Ou Date: Sat, 8 Dec 2018 18:40:09 +0800 Subject: RDMA/hns: Fix the bug while use multi-hop of pbl It will prevent multiply overflow when defines the pbl for u64 type. Signed-off-by: Lijun Ou Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hns/hns_roce_mr.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/hns/hns_roce_mr.c b/drivers/infiniband/hw/hns/hns_roce_mr.c index fabc95dbdf97..ee5991bd4171 100644 --- a/drivers/infiniband/hw/hns/hns_roce_mr.c +++ b/drivers/infiniband/hw/hns/hns_roce_mr.c @@ -1128,14 +1128,14 @@ struct ib_mr *hns_roce_reg_user_mr(struct ib_pd *pd, u64 start, u64 length, goto err_umem; } } else { - int pbl_size = 1; + u64 pbl_size = 1; bt_size = (1 << (hr_dev->caps.pbl_ba_pg_sz + PAGE_SHIFT)) / 8; for (i = 0; i < hr_dev->caps.pbl_hop_num; i++) pbl_size *= bt_size; if (n > pbl_size) { dev_err(dev, - " MR len %lld err. MR page num is limited to %d!\n", + " MR len %lld err. MR page num is limited to %lld!\n", length, pbl_size); ret = -EINVAL; goto err_umem; -- cgit v1.2.3-59-g8ed1b From f747b689451821fdf4bea53998feb38000595095 Mon Sep 17 00:00:00 2001 From: Lijun Ou Date: Sat, 8 Dec 2018 18:40:10 +0800 Subject: RDMA/hns: Update posting & querying mailbox This patch updates the implementation of the mailbox command interface by using command queue instead of operating registers. With this update, the software can be well decoupled with the hardware. Signed-off-by: Lijun Ou Signed-off-by: Yixian Liu Signed-off-by: Wei Hu (Xavier) Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hns/hns_roce_common.h | 3 -- drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 69 ++++++++++++++++++----------- drivers/infiniband/hw/hns/hns_roce_hw_v2.h | 28 +++++++----- 3 files changed, 59 insertions(+), 41 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/hns/hns_roce_common.h b/drivers/infiniband/hw/hns/hns_roce_common.h index 93d4b4ec002d..f4c92a7ac1ce 100644 --- a/drivers/infiniband/hw/hns/hns_roce_common.h +++ b/drivers/infiniband/hw/hns/hns_roce_common.h @@ -376,9 +376,6 @@ #define ROCEE_RX_CMQ_TAIL_REG 0x07024 #define ROCEE_RX_CMQ_HEAD_REG 0x07028 -#define ROCEE_VF_MB_CFG0_REG 0x40 -#define ROCEE_VF_MB_STATUS_REG 0x58 - #define ROCEE_VF_EQ_DB_CFG0_REG 0x238 #define ROCEE_VF_EQ_DB_CFG1_REG 0x23C diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c index ca71e1079d8a..9b8c81839766 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c @@ -1606,30 +1606,62 @@ static void hns_roce_v2_exit(struct hns_roce_dev *hr_dev) hns_roce_free_link_table(hr_dev, &priv->tsq); } +static int hns_roce_query_mbox_status(struct hns_roce_dev *hr_dev) +{ + struct hns_roce_cmq_desc desc; + struct hns_roce_mbox_status *mb_st = + (struct hns_roce_mbox_status *)desc.data; + enum hns_roce_cmd_return_status status; + + hns_roce_cmq_setup_basic_desc(&desc, HNS_ROCE_OPC_QUERY_MB_ST, true); + + status = hns_roce_cmq_send(hr_dev, &desc, 1); + if (status) + return status; + + return cpu_to_le32(mb_st->mb_status_hw_run); +} + static int hns_roce_v2_cmd_pending(struct hns_roce_dev *hr_dev) { - u32 status = readl(hr_dev->reg_base + ROCEE_VF_MB_STATUS_REG); + u32 status = hns_roce_query_mbox_status(hr_dev); return status >> HNS_ROCE_HW_RUN_BIT_SHIFT; } static int hns_roce_v2_cmd_complete(struct hns_roce_dev *hr_dev) { - u32 status = readl(hr_dev->reg_base + ROCEE_VF_MB_STATUS_REG); + u32 status = hns_roce_query_mbox_status(hr_dev); return status & HNS_ROCE_HW_MB_STATUS_MASK; } +static int hns_roce_mbox_post(struct hns_roce_dev *hr_dev, u64 in_param, + u64 out_param, u32 in_modifier, u8 op_modifier, + u16 op, u16 token, int event) +{ + struct hns_roce_cmq_desc desc; + struct hns_roce_post_mbox *mb = (struct hns_roce_post_mbox *)desc.data; + + hns_roce_cmq_setup_basic_desc(&desc, HNS_ROCE_OPC_POST_MB, false); + + mb->in_param_l = cpu_to_le64(in_param); + mb->in_param_h = cpu_to_le64(in_param) >> 32; + mb->out_param_l = cpu_to_le64(out_param); + mb->out_param_h = cpu_to_le64(out_param) >> 32; + mb->cmd_tag = cpu_to_le32(in_modifier << 8 | op); + mb->token_event_en = cpu_to_le32(event << 16 | token); + + return hns_roce_cmq_send(hr_dev, &desc, 1); +} + static int hns_roce_v2_post_mbox(struct hns_roce_dev *hr_dev, u64 in_param, u64 out_param, u32 in_modifier, u8 op_modifier, u16 op, u16 token, int event) { struct device *dev = hr_dev->dev; - u32 __iomem *hcr = (u32 __iomem *)(hr_dev->reg_base + - ROCEE_VF_MB_CFG0_REG); unsigned long end; - u32 val0 = 0; - u32 val1 = 0; + int ret; end = msecs_to_jiffies(HNS_ROCE_V2_GO_BIT_TIMEOUT_MSECS) + jiffies; while (hns_roce_v2_cmd_pending(hr_dev)) { @@ -1641,27 +1673,12 @@ static int hns_roce_v2_post_mbox(struct hns_roce_dev *hr_dev, u64 in_param, cond_resched(); } - roce_set_field(val0, HNS_ROCE_VF_MB4_TAG_MASK, - HNS_ROCE_VF_MB4_TAG_SHIFT, in_modifier); - roce_set_field(val0, HNS_ROCE_VF_MB4_CMD_MASK, - HNS_ROCE_VF_MB4_CMD_SHIFT, op); - roce_set_field(val1, HNS_ROCE_VF_MB5_EVENT_MASK, - HNS_ROCE_VF_MB5_EVENT_SHIFT, event); - roce_set_field(val1, HNS_ROCE_VF_MB5_TOKEN_MASK, - HNS_ROCE_VF_MB5_TOKEN_SHIFT, token); - - writeq(in_param, hcr + 0); - writeq(out_param, hcr + 2); - - /* Memory barrier */ - wmb(); - - writel(val0, hcr + 4); - writel(val1, hcr + 5); - - mmiowb(); + ret = hns_roce_mbox_post(hr_dev, in_param, out_param, in_modifier, + op_modifier, op, token, event); + if (ret) + dev_err(dev, "Post mailbox fail(%d)\n", ret); - return 0; + return ret; } static int hns_roce_v2_chk_mbox(struct hns_roce_dev *hr_dev, diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.h b/drivers/infiniband/hw/hns/hns_roce_hw_v2.h index c48301c6fe5d..7898d0600931 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.h +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.h @@ -224,6 +224,8 @@ enum hns_roce_opcode_type { HNS_ROCE_OPC_CFG_TMOUT_LLM = 0x8404, HNS_ROCE_OPC_CFG_SGID_TB = 0x8500, HNS_ROCE_OPC_CFG_SMAC_TB = 0x8501, + HNS_ROCE_OPC_POST_MB = 0x8504, + HNS_ROCE_OPC_QUERY_MB_ST = 0x8505, HNS_ROCE_OPC_CFG_BT_ATTR = 0x8506, }; @@ -1384,6 +1386,20 @@ struct hns_roce_vf_res_b { #define VF_RES_B_DATA_3_VF_SL_NUM_S 16 #define VF_RES_B_DATA_3_VF_SL_NUM_M GENMASK(19, 16) +struct hns_roce_post_mbox { + __le32 in_param_l; + __le32 in_param_h; + __le32 out_param_l; + __le32 out_param_h; + __le32 cmd_tag; + __le32 token_event_en; +}; + +struct hns_roce_mbox_status { + __le32 mb_status_hw_run; + __le32 rsv[5]; +}; + struct hns_roce_cfg_bt_attr { __le32 vf_qpc_cfg; __le32 vf_srqc_cfg; @@ -1467,18 +1483,6 @@ struct hns_roce_cmq_desc { #define HNS_ROCE_HW_RUN_BIT_SHIFT 31 #define HNS_ROCE_HW_MB_STATUS_MASK 0xFF -#define HNS_ROCE_VF_MB4_TAG_MASK 0xFFFFFF00 -#define HNS_ROCE_VF_MB4_TAG_SHIFT 8 - -#define HNS_ROCE_VF_MB4_CMD_MASK 0xFF -#define HNS_ROCE_VF_MB4_CMD_SHIFT 0 - -#define HNS_ROCE_VF_MB5_EVENT_MASK 0x10000 -#define HNS_ROCE_VF_MB5_EVENT_SHIFT 16 - -#define HNS_ROCE_VF_MB5_TOKEN_MASK 0xFFFF -#define HNS_ROCE_VF_MB5_TOKEN_SHIFT 0 - struct hns_roce_v2_cmq_ring { dma_addr_t desc_dma_addr; struct hns_roce_cmq_desc *desc; -- cgit v1.2.3-59-g8ed1b From 0c1c3880442f07532780533457ba0784790d8de2 Mon Sep 17 00:00:00 2001 From: Lijun Ou Date: Sat, 8 Dec 2018 18:40:11 +0800 Subject: RDMA/hns: Bugfix for RoCE loopback test This patch implements a cmdq to enable the loopback of ssu module according to the modified hardware desgin. The ssu consists of ingress unit, packet buffer and programmable packet process unit. if the loopback bit of ssu is not enabled, the roce packet with loopback bit will fail. Signed-off-by: Lijun Ou Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 36 ++++++++++++++++++++++++++++++ drivers/infiniband/hw/hns/hns_roce_hw_v2.h | 19 ++++++++++++++++ 2 files changed, 55 insertions(+) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c index 9b8c81839766..04fe5cc8b4ff 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c @@ -1082,6 +1082,33 @@ static int hns_roce_query_pf_resource(struct hns_roce_dev *hr_dev) return 0; } +static int hns_roce_set_vf_switch_param(struct hns_roce_dev *hr_dev, + int vf_id) +{ + struct hns_roce_cmq_desc desc; + struct hns_roce_vf_switch *swt; + int ret; + + swt = (struct hns_roce_vf_switch *)desc.data; + hns_roce_cmq_setup_basic_desc(&desc, HNS_SWITCH_PARAMETER_CFG, true); + swt->rocee_sel |= cpu_to_le16(HNS_ICL_SWITCH_CMD_ROCEE_SEL); + roce_set_field(swt->fun_id, + VF_SWITCH_DATA_FUN_ID_VF_ID_M, + VF_SWITCH_DATA_FUN_ID_VF_ID_S, + vf_id); + ret = hns_roce_cmq_send(hr_dev, &desc, 1); + if (ret) + return ret; + desc.flag = + cpu_to_le16(HNS_ROCE_CMD_FLAG_NO_INTR | HNS_ROCE_CMD_FLAG_IN); + desc.flag &= cpu_to_le16(~HNS_ROCE_CMD_FLAG_WR); + roce_set_bit(swt->cfg, VF_SWITCH_DATA_CFG_ALW_LPBK_S, 1); + roce_set_bit(swt->cfg, VF_SWITCH_DATA_CFG_ALW_LCL_LPBK_S, 1); + roce_set_bit(swt->cfg, VF_SWITCH_DATA_CFG_ALW_DST_OVRD_S, 1); + + return hns_roce_cmq_send(hr_dev, &desc, 1); +} + static int hns_roce_alloc_vf_resource(struct hns_roce_dev *hr_dev) { struct hns_roce_cmq_desc desc[2]; @@ -1269,6 +1296,15 @@ static int hns_roce_v2_profile(struct hns_roce_dev *hr_dev) return ret; } + if (hr_dev->pci_dev->revision == 0x21) { + ret = hns_roce_set_vf_switch_param(hr_dev, 0); + if (ret) { + dev_err(hr_dev->dev, + "Set function switch param fail, ret = %d.\n", + ret); + return ret; + } + } hr_dev->vendor_part_id = hr_dev->pci_dev->device; hr_dev->sys_image_guid = be64_to_cpu(hr_dev->ib_dev.node_guid); diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.h b/drivers/infiniband/hw/hns/hns_roce_hw_v2.h index 7898d0600931..b72d0443c835 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.h +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.h @@ -124,6 +124,8 @@ ((step_idx == 0 && hop_num == HNS_ROCE_HOP_NUM_0) || \ (step_idx == 1 && hop_num == 1) || \ (step_idx == 2 && hop_num == 2)) +#define HNS_ICL_SWITCH_CMD_ROCEE_SEL_SHIFT 0 +#define HNS_ICL_SWITCH_CMD_ROCEE_SEL BIT(HNS_ICL_SWITCH_CMD_ROCEE_SEL_SHIFT) #define CMD_CSQ_DESC_NUM 1024 #define CMD_CRQ_DESC_NUM 1024 @@ -227,6 +229,7 @@ enum hns_roce_opcode_type { HNS_ROCE_OPC_POST_MB = 0x8504, HNS_ROCE_OPC_QUERY_MB_ST = 0x8505, HNS_ROCE_OPC_CFG_BT_ATTR = 0x8506, + HNS_SWITCH_PARAMETER_CFG = 0x1033, }; enum { @@ -1386,6 +1389,22 @@ struct hns_roce_vf_res_b { #define VF_RES_B_DATA_3_VF_SL_NUM_S 16 #define VF_RES_B_DATA_3_VF_SL_NUM_M GENMASK(19, 16) +struct hns_roce_vf_switch { + __le32 rocee_sel; + __le32 fun_id; + __le32 cfg; + __le32 resv1; + __le32 resv2; + __le32 resv3; +}; + +#define VF_SWITCH_DATA_FUN_ID_VF_ID_S 3 +#define VF_SWITCH_DATA_FUN_ID_VF_ID_M GENMASK(10, 3) + +#define VF_SWITCH_DATA_CFG_ALW_LPBK_S 1 +#define VF_SWITCH_DATA_CFG_ALW_LCL_LPBK_S 2 +#define VF_SWITCH_DATA_CFG_ALW_DST_OVRD_S 3 + struct hns_roce_post_mbox { __le32 in_param_l; __le32 in_param_h; -- cgit v1.2.3-59-g8ed1b From a5a5d1993696419e7d5357fc3128e53d219d382e Mon Sep 17 00:00:00 2001 From: Michael Guralnik Date: Sun, 9 Dec 2018 11:49:50 +0200 Subject: IB/core: Add new IB rates Add the new rates that were added to Infiniband spec as part of HDR and 2x support. Signed-off-by: Michael Guralnik Reviewed-by: Majd Dibbiny Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/core/verbs.c | 12 ++++++++++++ drivers/infiniband/hw/mlx5/qp.c | 2 +- include/rdma/ib_verbs.h | 6 +++++- 3 files changed, 18 insertions(+), 2 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c index 178899e3ce73..fb2fc0c7ecfb 100644 --- a/drivers/infiniband/core/verbs.c +++ b/drivers/infiniband/core/verbs.c @@ -141,6 +141,10 @@ __attribute_const__ int ib_rate_to_mult(enum ib_rate rate) case IB_RATE_100_GBPS: return 40; case IB_RATE_200_GBPS: return 80; case IB_RATE_300_GBPS: return 120; + case IB_RATE_28_GBPS: return 11; + case IB_RATE_50_GBPS: return 20; + case IB_RATE_400_GBPS: return 160; + case IB_RATE_600_GBPS: return 240; default: return -1; } } @@ -166,6 +170,10 @@ __attribute_const__ enum ib_rate mult_to_ib_rate(int mult) case 40: return IB_RATE_100_GBPS; case 80: return IB_RATE_200_GBPS; case 120: return IB_RATE_300_GBPS; + case 11: return IB_RATE_28_GBPS; + case 20: return IB_RATE_50_GBPS; + case 160: return IB_RATE_400_GBPS; + case 240: return IB_RATE_600_GBPS; default: return IB_RATE_PORT_CURRENT; } } @@ -191,6 +199,10 @@ __attribute_const__ int ib_rate_to_mbps(enum ib_rate rate) case IB_RATE_100_GBPS: return 103125; case IB_RATE_200_GBPS: return 206250; case IB_RATE_300_GBPS: return 309375; + case IB_RATE_28_GBPS: return 28125; + case IB_RATE_50_GBPS: return 53125; + case IB_RATE_400_GBPS: return 425000; + case IB_RATE_600_GBPS: return 637500; default: return -1; } } diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c index 7789367114f7..07a27b8cc838 100644 --- a/drivers/infiniband/hw/mlx5/qp.c +++ b/drivers/infiniband/hw/mlx5/qp.c @@ -2703,7 +2703,7 @@ static int ib_rate_to_mlx5(struct mlx5_ib_dev *dev, u8 rate) if (rate == IB_RATE_PORT_CURRENT) return 0; - if (rate < IB_RATE_2_5_GBPS || rate > IB_RATE_300_GBPS) + if (rate < IB_RATE_2_5_GBPS || rate > IB_RATE_600_GBPS) return -EINVAL; while (rate != IB_RATE_PORT_CURRENT && diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h index a7b839312671..89eead636e68 100644 --- a/include/rdma/ib_verbs.h +++ b/include/rdma/ib_verbs.h @@ -732,7 +732,11 @@ enum ib_rate { IB_RATE_25_GBPS = 15, IB_RATE_100_GBPS = 16, IB_RATE_200_GBPS = 17, - IB_RATE_300_GBPS = 18 + IB_RATE_300_GBPS = 18, + IB_RATE_28_GBPS = 19, + IB_RATE_50_GBPS = 20, + IB_RATE_400_GBPS = 21, + IB_RATE_600_GBPS = 22, }; /** -- cgit v1.2.3-59-g8ed1b From 4106a758f791de11502cc6be89c971735cab360f Mon Sep 17 00:00:00 2001 From: Michael Guralnik Date: Sun, 9 Dec 2018 11:49:51 +0200 Subject: IB/mlx5: Report CapabilityMask2 in ib_query_port CapabilityMask2 exists when IB_PORT_CAP_MASK2_SUP is set in the original capability mask. In such cases, query its value and report it in query port. Signed-off-by: Michael Guralnik Reviewed-by: Majd Dibbiny Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/mad.c | 4 ++++ drivers/infiniband/hw/mlx5/main.c | 3 +++ include/linux/mlx5/driver.h | 4 ++-- 3 files changed, 9 insertions(+), 2 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/mad.c b/drivers/infiniband/hw/mlx5/mad.c index 32a9e9228b13..5806724450d5 100644 --- a/drivers/infiniband/hw/mlx5/mad.c +++ b/drivers/infiniband/hw/mlx5/mad.c @@ -568,6 +568,10 @@ int mlx5_query_mad_ifc_port(struct ib_device *ibdev, u8 port, props->max_vl_num = out_mad->data[37] >> 4; props->init_type_reply = out_mad->data[41] >> 4; + if (props->port_cap_flags & IB_PORT_CAP_MASK2_SUP) + props->port_cap_flags2 = + be16_to_cpup((__be16 *)(out_mad->data + 60)); + /* Check if extended speeds (EDR/FDR/...) are supported */ if (props->port_cap_flags & IB_PORT_EXTENDED_SPEEDS_SUP) { ext_active_speed = out_mad->data[62] >> 4; diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index 1b2e5465b882..935de3d400ea 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -1228,6 +1228,9 @@ static int mlx5_query_hca_port(struct ib_device *ibdev, u8 port, props->subnet_timeout = rep->subnet_timeout; props->init_type_reply = rep->init_type_reply; + if (props->port_cap_flags & IB_PORT_CAP_MASK2_SUP) + props->port_cap_flags2 = rep->cap_mask2; + err = mlx5_query_port_link_width_oper(mdev, &ib_link_width_oper, port); if (err) goto out; diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h index 584d8a5df7eb..b090a96f87df 100644 --- a/include/linux/mlx5/driver.h +++ b/include/linux/mlx5/driver.h @@ -749,8 +749,8 @@ struct mlx5_hca_vport_context { u64 node_guid; u32 cap_mask1; u32 cap_mask1_perm; - u32 cap_mask2; - u32 cap_mask2_perm; + u16 cap_mask2; + u16 cap_mask2_perm; u16 lid; u8 init_type_reply; /* bitmask: see ib spec 14.2.5.6 InitTypeReply */ u8 lmc; -- cgit v1.2.3-59-g8ed1b From b874155a5fa450bf5e4b5303cbfff05958805bf8 Mon Sep 17 00:00:00 2001 From: Michael Guralnik Date: Sun, 9 Dec 2018 11:49:52 +0200 Subject: IB/mlx5: Add HDR speed support to query port Report HDR speed when HDR is supported in CapabilityMask2 and the actual speed is HDR. Signed-off-by: Michael Guralnik Reviewed-by: Majd Dibbiny Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/mad.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/mad.c b/drivers/infiniband/hw/mlx5/mad.c index 5806724450d5..a2735f246d5c 100644 --- a/drivers/infiniband/hw/mlx5/mad.c +++ b/drivers/infiniband/hw/mlx5/mad.c @@ -583,6 +583,11 @@ int mlx5_query_mad_ifc_port(struct ib_device *ibdev, u8 port, case 2: props->active_speed = 32; /* EDR */ break; + case 4: + if (props->port_cap_flags & IB_PORT_CAP_MASK2_SUP && + props->port_cap_flags2 & IB_PORT_LINK_SPEED_HDR_SUP) + props->active_speed = IB_SPEED_HDR; + break; } } -- cgit v1.2.3-59-g8ed1b From d764970bcea34b2d907ef1b9fe29d09b17099552 Mon Sep 17 00:00:00 2001 From: Michael Guralnik Date: Sun, 9 Dec 2018 11:49:54 +0200 Subject: IB/mlx5: Add 2X width support to query_port Report to the user 2x width over MAD interface. Signed-off-by: Michael Guralnik Reviewed-by: Majd Dibbiny Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/mad.c | 6 +++++- drivers/infiniband/hw/mlx5/main.c | 2 ++ 2 files changed, 7 insertions(+), 1 deletion(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/mad.c b/drivers/infiniband/hw/mlx5/mad.c index a2735f246d5c..00e5d29517ca 100644 --- a/drivers/infiniband/hw/mlx5/mad.c +++ b/drivers/infiniband/hw/mlx5/mad.c @@ -568,10 +568,14 @@ int mlx5_query_mad_ifc_port(struct ib_device *ibdev, u8 port, props->max_vl_num = out_mad->data[37] >> 4; props->init_type_reply = out_mad->data[41] >> 4; - if (props->port_cap_flags & IB_PORT_CAP_MASK2_SUP) + if (props->port_cap_flags & IB_PORT_CAP_MASK2_SUP) { props->port_cap_flags2 = be16_to_cpup((__be16 *)(out_mad->data + 60)); + if (props->port_cap_flags2 & IB_PORT_LINK_WIDTH_2X_SUP) + props->active_width = out_mad->data[31] & 0x1f; + } + /* Check if extended speeds (EDR/FDR/...) are supported */ if (props->port_cap_flags & IB_PORT_EXTENDED_SPEEDS_SUP) { ext_active_speed = out_mad->data[62] >> 4; diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index 5edd8d5f151f..af091f363b81 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -1108,6 +1108,8 @@ static void translate_active_width(struct ib_device *ibdev, u8 active_width, if (active_width & MLX5_IB_WIDTH_1X) *ib_width = IB_WIDTH_1X; + else if (active_width & MLX5_IB_WIDTH_2X) + *ib_width = IB_WIDTH_2X; else if (active_width & MLX5_IB_WIDTH_4X) *ib_width = IB_WIDTH_4X; else if (active_width & MLX5_IB_WIDTH_8X) -- cgit v1.2.3-59-g8ed1b From f94e02ddfd88a511ed64fb68905e35d727635fea Mon Sep 17 00:00:00 2001 From: YueHaibing Date: Mon, 10 Dec 2018 15:27:59 +0800 Subject: IB/mlx5: Remove duplicated include from mlx5_ib.h Remove duplicated include. Signed-off-by: YueHaibing Acked-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/mlx5_ib.h | 1 - 1 file changed, 1 deletion(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index 7145f512f948..8fd3de05cdba 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -41,7 +41,6 @@ #include #include #include -#include #include #include #include -- cgit v1.2.3-59-g8ed1b From a1462351b590a4fa8eed6e5bf14245425635be0d Mon Sep 17 00:00:00 2001 From: Leon Romanovsky Date: Mon, 10 Dec 2018 11:19:49 +0200 Subject: RDMA/mlx5: Fail early if user tries to create flows on IB representors IB representors don't support creation of RAW ethernet QP flows. Disable them by reusing existing RDMA/core support macros. We do it for both creation and matcher because latter is not usable if no flow creation is available. Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/flow.c | 16 ++++++++++++---- include/rdma/uverbs_ioctl.h | 9 +++++---- 2 files changed, 17 insertions(+), 8 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/flow.c b/drivers/infiniband/hw/mlx5/flow.c index b73756bd38bd..b840d4eeba84 100644 --- a/drivers/infiniband/hw/mlx5/flow.c +++ b/drivers/infiniband/hw/mlx5/flow.c @@ -134,8 +134,6 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_CREATE_FLOW)( return -EINVAL; flow_act.action |= MLX5_FLOW_CONTEXT_ACTION_COUNT; } - if (dev->rep) - return -ENOTSUPP; if (dest_type == MLX5_FLOW_DESTINATION_TYPE_TIR && fs_matcher->ns_type == MLX5_FLOW_NAMESPACE_EGRESS) @@ -623,9 +621,19 @@ DECLARE_UVERBS_NAMED_OBJECT(MLX5_IB_OBJECT_FLOW_MATCHER, &UVERBS_METHOD(MLX5_IB_METHOD_FLOW_MATCHER_CREATE), &UVERBS_METHOD(MLX5_IB_METHOD_FLOW_MATCHER_DESTROY)); +static bool flow_is_supported(struct ib_device *device) +{ + return !to_mdev(device)->rep; +} + const struct uapi_definition mlx5_ib_flow_defs[] = { - UAPI_DEF_CHAIN_OBJ_TREE_NAMED(MLX5_IB_OBJECT_FLOW_MATCHER), - UAPI_DEF_CHAIN_OBJ_TREE(UVERBS_OBJECT_FLOW, &mlx5_ib_fs), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED( + MLX5_IB_OBJECT_FLOW_MATCHER, + UAPI_DEF_IS_OBJ_SUPPORTED(flow_is_supported)), + UAPI_DEF_CHAIN_OBJ_TREE( + UVERBS_OBJECT_FLOW, + &mlx5_ib_fs, + UAPI_DEF_IS_OBJ_SUPPORTED(flow_is_supported)), UAPI_DEF_CHAIN_OBJ_TREE(UVERBS_OBJECT_FLOW_ACTION, &mlx5_ib_flow_actions), {}, diff --git a/include/rdma/uverbs_ioctl.h b/include/rdma/uverbs_ioctl.h index 2f56844fb7da..7f83265f6298 100644 --- a/include/rdma/uverbs_ioctl.h +++ b/include/rdma/uverbs_ioctl.h @@ -454,15 +454,16 @@ struct uapi_definition { } /* Temporary until the tree base description is replaced */ -#define UAPI_DEF_CHAIN_OBJ_TREE(_object_enum, _object_ptr) \ +#define UAPI_DEF_CHAIN_OBJ_TREE(_object_enum, _object_ptr, ...) \ { \ .kind = UAPI_DEF_CHAIN_OBJ_TREE, \ .object_start = { .object_id = _object_enum }, \ .chain_obj_tree = _object_ptr, \ - } -#define UAPI_DEF_CHAIN_OBJ_TREE_NAMED(_object_enum, ...) \ - UAPI_DEF_CHAIN_OBJ_TREE(_object_enum, &UVERBS_OBJECT(_object_enum)), \ + }, \ ##__VA_ARGS__ +#define UAPI_DEF_CHAIN_OBJ_TREE_NAMED(_object_enum, ...) \ + UAPI_DEF_CHAIN_OBJ_TREE(_object_enum, &UVERBS_OBJECT(_object_enum), \ + ##__VA_ARGS__) /* * ======================================= -- cgit v1.2.3-59-g8ed1b From 59590b8ad2eea9543a639880fc0d2c139b66d164 Mon Sep 17 00:00:00 2001 From: Yuval Shaia Date: Sun, 9 Dec 2018 13:06:10 +0200 Subject: IB/{mlx5,ocrdma,qedr,rxe}: Omit port validation from IB verbs RDMA core layer already make sure port is valid, no need to check it here again. For the pkey validation this depends on commit b3ac5742fead ("RDMA/core: Validate port number in query_pkey verb") Signed-off-by: Yuval Shaia Acked-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/mad.c | 5 ----- drivers/infiniband/hw/ocrdma/ocrdma_verbs.c | 12 ------------ drivers/infiniband/hw/qedr/verbs.c | 12 ------------ drivers/infiniband/sw/rxe/rxe_verbs.c | 22 +--------------------- 4 files changed, 1 insertion(+), 50 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/mad.c b/drivers/infiniband/hw/mlx5/mad.c index 00e5d29517ca..558638468edb 100644 --- a/drivers/infiniband/hw/mlx5/mad.c +++ b/drivers/infiniband/hw/mlx5/mad.c @@ -526,11 +526,6 @@ int mlx5_query_mad_ifc_port(struct ib_device *ibdev, u8 port, int ext_active_speed; int err = -ENOMEM; - if (port < 1 || port > dev->num_ports) { - mlx5_ib_warn(dev, "invalid port number %d\n", port); - return -EINVAL; - } - in_mad = kzalloc(sizeof(*in_mad), GFP_KERNEL); out_mad = kmalloc(sizeof(*out_mad), GFP_KERNEL); if (!in_mad || !out_mad) diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c b/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c index 06d2a7f3304c..1a5c048e1d67 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c @@ -177,11 +177,6 @@ int ocrdma_query_port(struct ib_device *ibdev, /* props being zeroed by the caller, avoid zeroing it here */ dev = get_ocrdma_dev(ibdev); - if (port > 1) { - pr_err("%s(%d) invalid_port=0x%x\n", __func__, - dev->id, port); - return -EINVAL; - } netdev = dev->nic_info.netdev; if (netif_running(netdev) && netif_oper_up(netdev)) { port_state = IB_PORT_ACTIVE; @@ -215,13 +210,6 @@ int ocrdma_query_port(struct ib_device *ibdev, int ocrdma_modify_port(struct ib_device *ibdev, u8 port, int mask, struct ib_port_modify *props) { - struct ocrdma_dev *dev; - - dev = get_ocrdma_dev(ibdev); - if (port > 1) { - pr_err("%s(%d) invalid_port=0x%x\n", __func__, dev->id, port); - return -EINVAL; - } return 0; } diff --git a/drivers/infiniband/hw/qedr/verbs.c b/drivers/infiniband/hw/qedr/verbs.c index 82ee4b4a7084..28c6d5b11487 100644 --- a/drivers/infiniband/hw/qedr/verbs.c +++ b/drivers/infiniband/hw/qedr/verbs.c @@ -216,10 +216,6 @@ int qedr_query_port(struct ib_device *ibdev, u8 port, struct ib_port_attr *attr) struct qed_rdma_port *rdma_port; dev = get_qedr_dev(ibdev); - if (port > 1) { - DP_ERR(dev, "invalid_port=0x%x\n", port); - return -EINVAL; - } if (!dev->rdma_ctx) { DP_ERR(dev, "rdma_ctx is NULL\n"); @@ -263,14 +259,6 @@ int qedr_query_port(struct ib_device *ibdev, u8 port, struct ib_port_attr *attr) int qedr_modify_port(struct ib_device *ibdev, u8 port, int mask, struct ib_port_modify *props) { - struct qedr_dev *dev; - - dev = get_qedr_dev(ibdev); - if (port > 1) { - DP_ERR(dev, "invalid_port=0x%x\n", port); - return -EINVAL; - } - return 0; } diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.c b/drivers/infiniband/sw/rxe/rxe_verbs.c index 30817c79ba96..e4df2d60fe8b 100644 --- a/drivers/infiniband/sw/rxe/rxe_verbs.c +++ b/drivers/infiniband/sw/rxe/rxe_verbs.c @@ -56,12 +56,7 @@ static int rxe_query_port(struct ib_device *dev, { struct rxe_dev *rxe = to_rdev(dev); struct rxe_port *port; - int rc = -EINVAL; - - if (unlikely(port_num != 1)) { - pr_warn("invalid port_number %d\n", port_num); - goto out; - } + int rc; port = &rxe->port; @@ -81,7 +76,6 @@ static int rxe_query_port(struct ib_device *dev, mutex_unlock(&rxe->usdev_lock); -out: return rc; } @@ -104,12 +98,6 @@ static int rxe_query_pkey(struct ib_device *device, struct rxe_dev *rxe = to_rdev(device); struct rxe_port *port; - if (unlikely(port_num != 1)) { - dev_warn(device->dev.parent, "invalid port_num = %d\n", - port_num); - goto err1; - } - port = &rxe->port; if (unlikely(index >= port->attr.pkey_tbl_len)) { @@ -147,11 +135,6 @@ static int rxe_modify_port(struct ib_device *dev, struct rxe_dev *rxe = to_rdev(dev); struct rxe_port *port; - if (unlikely(port_num != 1)) { - pr_warn("invalid port_num = %d\n", port_num); - goto err1; - } - port = &rxe->port; port->attr.port_cap_flags |= attr->set_port_cap_mask; @@ -161,9 +144,6 @@ static int rxe_modify_port(struct ib_device *dev, port->attr.qkey_viol_cntr = 0; return 0; - -err1: - return -EINVAL; } static enum rdma_link_layer rxe_get_link_layer(struct ib_device *dev, -- cgit v1.2.3-59-g8ed1b From a276a4d93bf1580d737f38d1810e5f4b166f3edd Mon Sep 17 00:00:00 2001 From: Gal Pressman Date: Mon, 10 Dec 2018 17:17:25 +0200 Subject: RDMA/vmw_pvrdma: Use atomic memory allocation in create AH Create address handle callback should not sleep, use GFP_ATOMIC instead of GFP_KERNEL for memory allocation. Fixes: 29c8d9eba550 ("IB: Add vmw_pvrdma driver") Cc: Adit Ranadive Signed-off-by: Gal Pressman Reviewed-by: Yuval Shaia Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.c b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.c index b65d10b0a875..f4cb5cf26006 100644 --- a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.c +++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.c @@ -555,7 +555,7 @@ struct ib_ah *pvrdma_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, if (!atomic_add_unless(&dev->num_ahs, 1, dev->dsr->caps.max_ah)) return ERR_PTR(-ENOMEM); - ah = kzalloc(sizeof(*ah), GFP_KERNEL); + ah = kzalloc(sizeof(*ah), GFP_ATOMIC); if (!ah) { atomic_dec(&dev->num_ahs); return ERR_PTR(-ENOMEM); -- cgit v1.2.3-59-g8ed1b From 8cc0698f46ca9cb38e77323e1074f182aac4040a Mon Sep 17 00:00:00 2001 From: Leon Romanovsky Date: Tue, 11 Dec 2018 12:04:42 +0200 Subject: RDMA/ocrdma: Use PCI-ID as an identification in debugfs In opposite to current implementation of identification based on device name, PCI-ID identification provides stable names suitable for device rename. Change ocrdma debugfs representation to use PCI-ID. Signed-off-by: Leon Romanovsky Reviewed-by: Yuval Shaia Acked-by: Selvin Xavier Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/ocrdma/ocrdma_stats.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_stats.c b/drivers/infiniband/hw/ocrdma/ocrdma_stats.c index 290d776edf48..dd15474b19b7 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_stats.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_stats.c @@ -760,12 +760,13 @@ static const struct file_operations ocrdma_dbg_ops = { void ocrdma_add_port_stats(struct ocrdma_dev *dev) { + const struct pci_dev *pdev = dev->nic_info.pdev; + if (!ocrdma_dbgfs_dir) return; /* Create post stats base dir */ - dev->dir = - debugfs_create_dir(dev_name(&dev->ibdev.dev), ocrdma_dbgfs_dir); + dev->dir = debugfs_create_dir(pci_name(pdev), ocrdma_dbgfs_dir); if (!dev->dir) goto err; -- cgit v1.2.3-59-g8ed1b From 2dd8e44cb4f3893f26ca29e9593450f289a0b48a Mon Sep 17 00:00:00 2001 From: Yuval Shaia Date: Tue, 11 Dec 2018 12:26:35 +0200 Subject: IB/mlx4: Remove unneeded NULL check NULL check for kfree is unnecessary, remove it. Fixes: b42dde478bca ("IB/mlx4: Rework special QP creation error path") Signed-off-by: Yuval Shaia Reviewed-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx4/qp.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c index 0711ca1dfb8f..74ef7a0ec0b7 100644 --- a/drivers/infiniband/hw/mlx4/qp.c +++ b/drivers/infiniband/hw/mlx4/qp.c @@ -1211,10 +1211,10 @@ err_db: mlx4_db_free(dev->dev, &qp->db); err: - if (sqp) - kfree(sqp); - else if (!*caller_qp) + if (!sqp && !*caller_qp) kfree(qp); + kfree(sqp); + return err; } -- cgit v1.2.3-59-g8ed1b From 9615f86be981a41a830ae088856da3e5ac362ab6 Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Mon, 10 Dec 2018 21:09:31 +0200 Subject: RDMA/bnxt_re: Initialize ib_device_ops struct Initialize ib_device_ops with the supported operations using ib_set_device_ops(). Signed-off-by: Kamal Heib Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/bnxt_re/main.c | 96 +++++++++++++++++------------------- 1 file changed, 45 insertions(+), 51 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/bnxt_re/main.c b/drivers/infiniband/hw/bnxt_re/main.c index 77f095e5fbe3..31cbef57a19b 100644 --- a/drivers/infiniband/hw/bnxt_re/main.c +++ b/drivers/infiniband/hw/bnxt_re/main.c @@ -568,6 +568,50 @@ static void bnxt_re_unregister_ib(struct bnxt_re_dev *rdev) ib_unregister_device(&rdev->ibdev); } +static const struct ib_device_ops bnxt_re_dev_ops = { + .add_gid = bnxt_re_add_gid, + .alloc_hw_stats = bnxt_re_ib_alloc_hw_stats, + .alloc_mr = bnxt_re_alloc_mr, + .alloc_pd = bnxt_re_alloc_pd, + .alloc_ucontext = bnxt_re_alloc_ucontext, + .create_ah = bnxt_re_create_ah, + .create_cq = bnxt_re_create_cq, + .create_qp = bnxt_re_create_qp, + .create_srq = bnxt_re_create_srq, + .dealloc_pd = bnxt_re_dealloc_pd, + .dealloc_ucontext = bnxt_re_dealloc_ucontext, + .del_gid = bnxt_re_del_gid, + .dereg_mr = bnxt_re_dereg_mr, + .destroy_ah = bnxt_re_destroy_ah, + .destroy_cq = bnxt_re_destroy_cq, + .destroy_qp = bnxt_re_destroy_qp, + .destroy_srq = bnxt_re_destroy_srq, + .get_dev_fw_str = bnxt_re_query_fw_str, + .get_dma_mr = bnxt_re_get_dma_mr, + .get_hw_stats = bnxt_re_ib_get_hw_stats, + .get_link_layer = bnxt_re_get_link_layer, + .get_netdev = bnxt_re_get_netdev, + .get_port_immutable = bnxt_re_get_port_immutable, + .map_mr_sg = bnxt_re_map_mr_sg, + .mmap = bnxt_re_mmap, + .modify_ah = bnxt_re_modify_ah, + .modify_device = bnxt_re_modify_device, + .modify_qp = bnxt_re_modify_qp, + .modify_srq = bnxt_re_modify_srq, + .poll_cq = bnxt_re_poll_cq, + .post_recv = bnxt_re_post_recv, + .post_send = bnxt_re_post_send, + .post_srq_recv = bnxt_re_post_srq_recv, + .query_ah = bnxt_re_query_ah, + .query_device = bnxt_re_query_device, + .query_pkey = bnxt_re_query_pkey, + .query_port = bnxt_re_query_port, + .query_qp = bnxt_re_query_qp, + .query_srq = bnxt_re_query_srq, + .reg_user_mr = bnxt_re_reg_user_mr, + .req_notify_cq = bnxt_re_req_notify_cq, +}; + static int bnxt_re_register_ib(struct bnxt_re_dev *rdev) { struct ib_device *ibdev = &rdev->ibdev; @@ -614,60 +658,10 @@ static int bnxt_re_register_ib(struct bnxt_re_dev *rdev) (1ull << IB_USER_VERBS_CMD_DESTROY_AH); /* POLL_CQ and REQ_NOTIFY_CQ is directly handled in libbnxt_re */ - /* Kernel verbs */ - ibdev->query_device = bnxt_re_query_device; - ibdev->modify_device = bnxt_re_modify_device; - - ibdev->query_port = bnxt_re_query_port; - ibdev->get_port_immutable = bnxt_re_get_port_immutable; - ibdev->get_dev_fw_str = bnxt_re_query_fw_str; - ibdev->query_pkey = bnxt_re_query_pkey; - ibdev->get_netdev = bnxt_re_get_netdev; - ibdev->add_gid = bnxt_re_add_gid; - ibdev->del_gid = bnxt_re_del_gid; - ibdev->get_link_layer = bnxt_re_get_link_layer; - - ibdev->alloc_pd = bnxt_re_alloc_pd; - ibdev->dealloc_pd = bnxt_re_dealloc_pd; - - ibdev->create_ah = bnxt_re_create_ah; - ibdev->modify_ah = bnxt_re_modify_ah; - ibdev->query_ah = bnxt_re_query_ah; - ibdev->destroy_ah = bnxt_re_destroy_ah; - - ibdev->create_srq = bnxt_re_create_srq; - ibdev->modify_srq = bnxt_re_modify_srq; - ibdev->query_srq = bnxt_re_query_srq; - ibdev->destroy_srq = bnxt_re_destroy_srq; - ibdev->post_srq_recv = bnxt_re_post_srq_recv; - - ibdev->create_qp = bnxt_re_create_qp; - ibdev->modify_qp = bnxt_re_modify_qp; - ibdev->query_qp = bnxt_re_query_qp; - ibdev->destroy_qp = bnxt_re_destroy_qp; - - ibdev->post_send = bnxt_re_post_send; - ibdev->post_recv = bnxt_re_post_recv; - - ibdev->create_cq = bnxt_re_create_cq; - ibdev->destroy_cq = bnxt_re_destroy_cq; - ibdev->poll_cq = bnxt_re_poll_cq; - ibdev->req_notify_cq = bnxt_re_req_notify_cq; - - ibdev->get_dma_mr = bnxt_re_get_dma_mr; - ibdev->dereg_mr = bnxt_re_dereg_mr; - ibdev->alloc_mr = bnxt_re_alloc_mr; - ibdev->map_mr_sg = bnxt_re_map_mr_sg; - - ibdev->reg_user_mr = bnxt_re_reg_user_mr; - ibdev->alloc_ucontext = bnxt_re_alloc_ucontext; - ibdev->dealloc_ucontext = bnxt_re_dealloc_ucontext; - ibdev->mmap = bnxt_re_mmap; - ibdev->get_hw_stats = bnxt_re_ib_get_hw_stats; - ibdev->alloc_hw_stats = bnxt_re_ib_alloc_hw_stats; rdma_set_device_sysfs_group(ibdev, &bnxt_re_dev_attr_group); ibdev->driver_id = RDMA_DRIVER_BNXT_RE; + ib_set_device_ops(ibdev, &bnxt_re_dev_ops); return ib_register_device(ibdev, "bnxt_re%d", NULL); } -- cgit v1.2.3-59-g8ed1b From 071b2ca40ae31bf1ce0c4a6edb5fc52988dafa54 Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Mon, 10 Dec 2018 21:09:32 +0200 Subject: RDMA/cxgb3: Initialize ib_device_ops struct Initialize ib_device_ops with the supported operations using ib_set_device_ops(). Signed-off-by: Kamal Heib Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/cxgb3/iwch_provider.c | 64 +++++++++++++++-------------- 1 file changed, 34 insertions(+), 30 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c index ebbec02cebe0..7a1dc83ba588 100644 --- a/drivers/infiniband/hw/cxgb3/iwch_provider.c +++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c @@ -1317,6 +1317,39 @@ static void get_dev_fw_ver_str(struct ib_device *ibdev, char *str) snprintf(str, IB_FW_VERSION_NAME_MAX, "%s", info.fw_version); } +static const struct ib_device_ops iwch_dev_ops = { + .alloc_hw_stats = iwch_alloc_stats, + .alloc_mr = iwch_alloc_mr, + .alloc_mw = iwch_alloc_mw, + .alloc_pd = iwch_allocate_pd, + .alloc_ucontext = iwch_alloc_ucontext, + .create_cq = iwch_create_cq, + .create_qp = iwch_create_qp, + .dealloc_mw = iwch_dealloc_mw, + .dealloc_pd = iwch_deallocate_pd, + .dealloc_ucontext = iwch_dealloc_ucontext, + .dereg_mr = iwch_dereg_mr, + .destroy_cq = iwch_destroy_cq, + .destroy_qp = iwch_destroy_qp, + .get_dev_fw_str = get_dev_fw_ver_str, + .get_dma_mr = iwch_get_dma_mr, + .get_hw_stats = iwch_get_mib, + .get_port_immutable = iwch_port_immutable, + .map_mr_sg = iwch_map_mr_sg, + .mmap = iwch_mmap, + .modify_qp = iwch_ib_modify_qp, + .poll_cq = iwch_poll_cq, + .post_recv = iwch_post_receive, + .post_send = iwch_post_send, + .query_device = iwch_query_device, + .query_gid = iwch_query_gid, + .query_pkey = iwch_query_pkey, + .query_port = iwch_query_port, + .reg_user_mr = iwch_reg_user_mr, + .req_notify_cq = iwch_arm_cq, + .resize_cq = iwch_resize_cq, +}; + int iwch_register_device(struct iwch_dev *dev) { int ret; @@ -1356,37 +1389,7 @@ int iwch_register_device(struct iwch_dev *dev) dev->ibdev.phys_port_cnt = dev->rdev.port_info.nports; dev->ibdev.num_comp_vectors = 1; dev->ibdev.dev.parent = &dev->rdev.rnic_info.pdev->dev; - dev->ibdev.query_device = iwch_query_device; - dev->ibdev.query_port = iwch_query_port; - dev->ibdev.query_pkey = iwch_query_pkey; - dev->ibdev.query_gid = iwch_query_gid; - dev->ibdev.alloc_ucontext = iwch_alloc_ucontext; - dev->ibdev.dealloc_ucontext = iwch_dealloc_ucontext; - dev->ibdev.mmap = iwch_mmap; - dev->ibdev.alloc_pd = iwch_allocate_pd; - dev->ibdev.dealloc_pd = iwch_deallocate_pd; - dev->ibdev.create_qp = iwch_create_qp; - dev->ibdev.modify_qp = iwch_ib_modify_qp; - dev->ibdev.destroy_qp = iwch_destroy_qp; - dev->ibdev.create_cq = iwch_create_cq; - dev->ibdev.destroy_cq = iwch_destroy_cq; - dev->ibdev.resize_cq = iwch_resize_cq; - dev->ibdev.poll_cq = iwch_poll_cq; - dev->ibdev.get_dma_mr = iwch_get_dma_mr; - dev->ibdev.reg_user_mr = iwch_reg_user_mr; - dev->ibdev.dereg_mr = iwch_dereg_mr; - dev->ibdev.alloc_mw = iwch_alloc_mw; - dev->ibdev.dealloc_mw = iwch_dealloc_mw; - dev->ibdev.alloc_mr = iwch_alloc_mr; - dev->ibdev.map_mr_sg = iwch_map_mr_sg; - dev->ibdev.req_notify_cq = iwch_arm_cq; - dev->ibdev.post_send = iwch_post_send; - dev->ibdev.post_recv = iwch_post_receive; - dev->ibdev.alloc_hw_stats = iwch_alloc_stats; - dev->ibdev.get_hw_stats = iwch_get_mib; dev->ibdev.uverbs_abi_ver = IWCH_UVERBS_ABI_VERSION; - dev->ibdev.get_port_immutable = iwch_port_immutable; - dev->ibdev.get_dev_fw_str = get_dev_fw_ver_str; dev->ibdev.iwcm = kmalloc(sizeof(struct iw_cm_verbs), GFP_KERNEL); if (!dev->ibdev.iwcm) @@ -1405,6 +1408,7 @@ int iwch_register_device(struct iwch_dev *dev) dev->ibdev.driver_id = RDMA_DRIVER_CXGB3; rdma_set_device_sysfs_group(&dev->ibdev, &iwch_attr_group); + ib_set_device_ops(&dev->ibdev, &iwch_dev_ops); ret = ib_register_device(&dev->ibdev, "cxgb3_%d", NULL); if (ret) kfree(dev->ibdev.iwcm); -- cgit v1.2.3-59-g8ed1b From dad3b05d050a8ce5825b2bd7c680fc127b3c2c69 Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Mon, 10 Dec 2018 21:09:33 +0200 Subject: RDMA/cxgb4: Initialize ib_device_ops struct Initialize ib_device_ops with the supported operations using ib_set_device_ops(). Signed-off-by: Kamal Heib Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/cxgb4/provider.c | 74 ++++++++++++++++++---------------- 1 file changed, 39 insertions(+), 35 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/cxgb4/provider.c b/drivers/infiniband/hw/cxgb4/provider.c index cbb3c0ddd990..586b0c37481f 100644 --- a/drivers/infiniband/hw/cxgb4/provider.c +++ b/drivers/infiniband/hw/cxgb4/provider.c @@ -531,6 +531,44 @@ static int fill_res_entry(struct sk_buff *msg, struct rdma_restrack_entry *res) c4iw_restrack_funcs[res->type](msg, res) : 0; } +static const struct ib_device_ops c4iw_dev_ops = { + .alloc_hw_stats = c4iw_alloc_stats, + .alloc_mr = c4iw_alloc_mr, + .alloc_mw = c4iw_alloc_mw, + .alloc_pd = c4iw_allocate_pd, + .alloc_ucontext = c4iw_alloc_ucontext, + .create_cq = c4iw_create_cq, + .create_qp = c4iw_create_qp, + .create_srq = c4iw_create_srq, + .dealloc_mw = c4iw_dealloc_mw, + .dealloc_pd = c4iw_deallocate_pd, + .dealloc_ucontext = c4iw_dealloc_ucontext, + .dereg_mr = c4iw_dereg_mr, + .destroy_cq = c4iw_destroy_cq, + .destroy_qp = c4iw_destroy_qp, + .destroy_srq = c4iw_destroy_srq, + .get_dev_fw_str = get_dev_fw_str, + .get_dma_mr = c4iw_get_dma_mr, + .get_hw_stats = c4iw_get_mib, + .get_netdev = get_netdev, + .get_port_immutable = c4iw_port_immutable, + .map_mr_sg = c4iw_map_mr_sg, + .mmap = c4iw_mmap, + .modify_qp = c4iw_ib_modify_qp, + .modify_srq = c4iw_modify_srq, + .poll_cq = c4iw_poll_cq, + .post_recv = c4iw_post_receive, + .post_send = c4iw_post_send, + .post_srq_recv = c4iw_post_srq_recv, + .query_device = c4iw_query_device, + .query_gid = c4iw_query_gid, + .query_pkey = c4iw_query_pkey, + .query_port = c4iw_query_port, + .query_qp = c4iw_ib_query_qp, + .reg_user_mr = c4iw_reg_user_mr, + .req_notify_cq = c4iw_arm_cq, +}; + void c4iw_register_device(struct work_struct *work) { int ret; @@ -573,42 +611,7 @@ void c4iw_register_device(struct work_struct *work) dev->ibdev.phys_port_cnt = dev->rdev.lldi.nports; dev->ibdev.num_comp_vectors = dev->rdev.lldi.nciq; dev->ibdev.dev.parent = &dev->rdev.lldi.pdev->dev; - dev->ibdev.query_device = c4iw_query_device; - dev->ibdev.query_port = c4iw_query_port; - dev->ibdev.query_pkey = c4iw_query_pkey; - dev->ibdev.query_gid = c4iw_query_gid; - dev->ibdev.alloc_ucontext = c4iw_alloc_ucontext; - dev->ibdev.dealloc_ucontext = c4iw_dealloc_ucontext; - dev->ibdev.mmap = c4iw_mmap; - dev->ibdev.alloc_pd = c4iw_allocate_pd; - dev->ibdev.dealloc_pd = c4iw_deallocate_pd; - dev->ibdev.create_qp = c4iw_create_qp; - dev->ibdev.modify_qp = c4iw_ib_modify_qp; - dev->ibdev.query_qp = c4iw_ib_query_qp; - dev->ibdev.destroy_qp = c4iw_destroy_qp; - dev->ibdev.create_srq = c4iw_create_srq; - dev->ibdev.modify_srq = c4iw_modify_srq; - dev->ibdev.destroy_srq = c4iw_destroy_srq; - dev->ibdev.create_cq = c4iw_create_cq; - dev->ibdev.destroy_cq = c4iw_destroy_cq; - dev->ibdev.poll_cq = c4iw_poll_cq; - dev->ibdev.get_dma_mr = c4iw_get_dma_mr; - dev->ibdev.reg_user_mr = c4iw_reg_user_mr; - dev->ibdev.dereg_mr = c4iw_dereg_mr; - dev->ibdev.alloc_mw = c4iw_alloc_mw; - dev->ibdev.dealloc_mw = c4iw_dealloc_mw; - dev->ibdev.alloc_mr = c4iw_alloc_mr; - dev->ibdev.map_mr_sg = c4iw_map_mr_sg; - dev->ibdev.req_notify_cq = c4iw_arm_cq; - dev->ibdev.post_send = c4iw_post_send; - dev->ibdev.post_recv = c4iw_post_receive; - dev->ibdev.post_srq_recv = c4iw_post_srq_recv; - dev->ibdev.alloc_hw_stats = c4iw_alloc_stats; - dev->ibdev.get_hw_stats = c4iw_get_mib; dev->ibdev.uverbs_abi_ver = C4IW_UVERBS_ABI_VERSION; - dev->ibdev.get_port_immutable = c4iw_port_immutable; - dev->ibdev.get_dev_fw_str = get_dev_fw_str; - dev->ibdev.get_netdev = get_netdev; dev->ibdev.iwcm = kmalloc(sizeof(struct iw_cm_verbs), GFP_KERNEL); if (!dev->ibdev.iwcm) { @@ -630,6 +633,7 @@ void c4iw_register_device(struct work_struct *work) rdma_set_device_sysfs_group(&dev->ibdev, &c4iw_attr_group); dev->ibdev.driver_id = RDMA_DRIVER_CXGB4; + ib_set_device_ops(&dev->ibdev, &c4iw_dev_ops); ret = ib_register_device(&dev->ibdev, "cxgb4_%d", NULL); if (ret) goto err_kfree_iwcm; -- cgit v1.2.3-59-g8ed1b From e3c320caa10713e1276cd2b813a6cdac37810e51 Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Mon, 10 Dec 2018 21:09:34 +0200 Subject: RDMA/hfi1: Initialize ib_device_ops struct Initialize ib_device_ops with the supported operations using ib_set_device_ops(). Signed-off-by: Kamal Heib Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hfi1/verbs.c | 18 +++++++++++------- 1 file changed, 11 insertions(+), 7 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/hfi1/verbs.c b/drivers/infiniband/hw/hfi1/verbs.c index 28f754d2a844..e42162ea59f6 100644 --- a/drivers/infiniband/hw/hfi1/verbs.c +++ b/drivers/infiniband/hw/hfi1/verbs.c @@ -1617,6 +1617,16 @@ static int get_hw_stats(struct ib_device *ibdev, struct rdma_hw_stats *stats, return count; } +static const struct ib_device_ops hfi1_dev_ops = { + .alloc_hw_stats = alloc_hw_stats, + .alloc_rdma_netdev = hfi1_vnic_alloc_rn, + .get_dev_fw_str = hfi1_get_dev_fw_str, + .get_hw_stats = get_hw_stats, + .modify_device = modify_device, + /* keep process mad in the driver */ + .process_mad = hfi1_process_mad, +}; + /** * hfi1_register_ib_device - register our device with the infiniband core * @dd: the device data structure @@ -1660,14 +1670,8 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd) ibdev->owner = THIS_MODULE; ibdev->phys_port_cnt = dd->num_pports; ibdev->dev.parent = &dd->pcidev->dev; - ibdev->modify_device = modify_device; - ibdev->alloc_hw_stats = alloc_hw_stats; - ibdev->get_hw_stats = get_hw_stats; - ibdev->alloc_rdma_netdev = hfi1_vnic_alloc_rn; - /* keep process mad in the driver */ - ibdev->process_mad = hfi1_process_mad; - ibdev->get_dev_fw_str = hfi1_get_dev_fw_str; + ib_set_device_ops(ibdev, &hfi1_dev_ops); strlcpy(ibdev->node_desc, init_utsname()->nodename, sizeof(ibdev->node_desc)); -- cgit v1.2.3-59-g8ed1b From 7f645a58d00155b93800fdd4e157c8f58fb3122f Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Mon, 10 Dec 2018 21:09:35 +0200 Subject: RDMA/hns: Initialize ib_device_ops struct Initialize ib_device_ops with the supported operations using ib_set_device_ops(). Signed-off-by: Kamal Heib Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hns/hns_roce_device.h | 2 + drivers/infiniband/hw/hns/hns_roce_hw_v1.c | 11 +++ drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 18 +++++ drivers/infiniband/hw/hns/hns_roce_main.c | 114 ++++++++++++++-------------- 4 files changed, 87 insertions(+), 58 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/hns/hns_roce_device.h b/drivers/infiniband/hw/hns/hns_roce_device.h index 779dd4c409cb..67609cc6a45e 100644 --- a/drivers/infiniband/hw/hns/hns_roce_device.h +++ b/drivers/infiniband/hw/hns/hns_roce_device.h @@ -883,6 +883,8 @@ struct hns_roce_hw { int (*query_srq)(struct ib_srq *ibsrq, struct ib_srq_attr *attr); int (*post_srq_recv)(struct ib_srq *ibsrq, const struct ib_recv_wr *wr, const struct ib_recv_wr **bad_wr); + const struct ib_device_ops *hns_roce_dev_ops; + const struct ib_device_ops *hns_roce_dev_srq_ops; }; struct hns_roce_dev { diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c index ca05810c92dc..d17a7ce3c93a 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c @@ -4793,6 +4793,16 @@ static void hns_roce_v1_cleanup_eq_table(struct hns_roce_dev *hr_dev) kfree(eq_table->eq); } +static const struct ib_device_ops hns_roce_v1_dev_ops = { + .destroy_qp = hns_roce_v1_destroy_qp, + .modify_cq = hns_roce_v1_modify_cq, + .poll_cq = hns_roce_v1_poll_cq, + .post_recv = hns_roce_v1_post_recv, + .post_send = hns_roce_v1_post_send, + .query_qp = hns_roce_v1_query_qp, + .req_notify_cq = hns_roce_v1_req_notify_cq, +}; + static const struct hns_roce_hw hns_roce_hw_v1 = { .reset = hns_roce_v1_reset, .hw_profile = hns_roce_v1_profile, @@ -4818,6 +4828,7 @@ static const struct hns_roce_hw hns_roce_hw_v1 = { .destroy_cq = hns_roce_v1_destroy_cq, .init_eq = hns_roce_v1_init_eq_table, .cleanup_eq = hns_roce_v1_cleanup_eq_table, + .hns_roce_dev_ops = &hns_roce_v1_dev_ops, }; static const struct of_device_id hns_roce_of_match[] = { diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c index 5007fe7af2f9..22236ebf033d 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c @@ -5734,6 +5734,22 @@ static int hns_roce_v2_post_srq_recv(struct ib_srq *ibsrq, return ret; } +static const struct ib_device_ops hns_roce_v2_dev_ops = { + .destroy_qp = hns_roce_v2_destroy_qp, + .modify_cq = hns_roce_v2_modify_cq, + .poll_cq = hns_roce_v2_poll_cq, + .post_recv = hns_roce_v2_post_recv, + .post_send = hns_roce_v2_post_send, + .query_qp = hns_roce_v2_query_qp, + .req_notify_cq = hns_roce_v2_req_notify_cq, +}; + +static const struct ib_device_ops hns_roce_v2_dev_srq_ops = { + .modify_srq = hns_roce_v2_modify_srq, + .post_srq_recv = hns_roce_v2_post_srq_recv, + .query_srq = hns_roce_v2_query_srq, +}; + static const struct hns_roce_hw hns_roce_hw_v2 = { .cmq_init = hns_roce_v2_cmq_init, .cmq_exit = hns_roce_v2_cmq_exit, @@ -5765,6 +5781,8 @@ static const struct hns_roce_hw hns_roce_hw_v2 = { .modify_srq = hns_roce_v2_modify_srq, .query_srq = hns_roce_v2_query_srq, .post_srq_recv = hns_roce_v2_post_srq_recv, + .hns_roce_dev_ops = &hns_roce_v2_dev_ops, + .hns_roce_dev_srq_ops = &hns_roce_v2_dev_srq_ops, }; static const struct pci_device_id hns_roce_hw_v2_pci_tbl[] = { diff --git a/drivers/infiniband/hw/hns/hns_roce_main.c b/drivers/infiniband/hw/hns/hns_roce_main.c index 65ba43cee810..c79054ba9495 100644 --- a/drivers/infiniband/hw/hns/hns_roce_main.c +++ b/drivers/infiniband/hw/hns/hns_roce_main.c @@ -445,6 +445,54 @@ static void hns_roce_unregister_device(struct hns_roce_dev *hr_dev) ib_unregister_device(&hr_dev->ib_dev); } +static const struct ib_device_ops hns_roce_dev_ops = { + .add_gid = hns_roce_add_gid, + .alloc_pd = hns_roce_alloc_pd, + .alloc_ucontext = hns_roce_alloc_ucontext, + .create_ah = hns_roce_create_ah, + .create_cq = hns_roce_ib_create_cq, + .create_qp = hns_roce_create_qp, + .dealloc_pd = hns_roce_dealloc_pd, + .dealloc_ucontext = hns_roce_dealloc_ucontext, + .del_gid = hns_roce_del_gid, + .dereg_mr = hns_roce_dereg_mr, + .destroy_ah = hns_roce_destroy_ah, + .destroy_cq = hns_roce_ib_destroy_cq, + .disassociate_ucontext = hns_roce_disassociate_ucontext, + .get_dma_mr = hns_roce_get_dma_mr, + .get_link_layer = hns_roce_get_link_layer, + .get_netdev = hns_roce_get_netdev, + .get_port_immutable = hns_roce_port_immutable, + .mmap = hns_roce_mmap, + .modify_device = hns_roce_modify_device, + .modify_port = hns_roce_modify_port, + .modify_qp = hns_roce_modify_qp, + .query_ah = hns_roce_query_ah, + .query_device = hns_roce_query_device, + .query_pkey = hns_roce_query_pkey, + .query_port = hns_roce_query_port, + .reg_user_mr = hns_roce_reg_user_mr, +}; + +static const struct ib_device_ops hns_roce_dev_mr_ops = { + .rereg_user_mr = hns_roce_rereg_user_mr, +}; + +static const struct ib_device_ops hns_roce_dev_mw_ops = { + .alloc_mw = hns_roce_alloc_mw, + .dealloc_mw = hns_roce_dealloc_mw, +}; + +static const struct ib_device_ops hns_roce_dev_frmr_ops = { + .alloc_mr = hns_roce_alloc_mr, + .map_mr_sg = hns_roce_map_mr_sg, +}; + +static const struct ib_device_ops hns_roce_dev_srq_ops = { + .create_srq = hns_roce_create_srq, + .destroy_srq = hns_roce_destroy_srq, +}; + static int hns_roce_register_device(struct hns_roce_dev *hr_dev) { int ret; @@ -484,88 +532,38 @@ static int hns_roce_register_device(struct hns_roce_dev *hr_dev) ib_dev->uverbs_ex_cmd_mask |= (1ULL << IB_USER_VERBS_EX_CMD_MODIFY_CQ); - /* HCA||device||port */ - ib_dev->modify_device = hns_roce_modify_device; - ib_dev->query_device = hns_roce_query_device; - ib_dev->query_port = hns_roce_query_port; - ib_dev->modify_port = hns_roce_modify_port; - ib_dev->get_link_layer = hns_roce_get_link_layer; - ib_dev->get_netdev = hns_roce_get_netdev; - ib_dev->add_gid = hns_roce_add_gid; - ib_dev->del_gid = hns_roce_del_gid; - ib_dev->query_pkey = hns_roce_query_pkey; - ib_dev->alloc_ucontext = hns_roce_alloc_ucontext; - ib_dev->dealloc_ucontext = hns_roce_dealloc_ucontext; - ib_dev->mmap = hns_roce_mmap; - - /* PD */ - ib_dev->alloc_pd = hns_roce_alloc_pd; - ib_dev->dealloc_pd = hns_roce_dealloc_pd; - - /* AH */ - ib_dev->create_ah = hns_roce_create_ah; - ib_dev->query_ah = hns_roce_query_ah; - ib_dev->destroy_ah = hns_roce_destroy_ah; - - /* QP */ - ib_dev->create_qp = hns_roce_create_qp; - ib_dev->modify_qp = hns_roce_modify_qp; - ib_dev->query_qp = hr_dev->hw->query_qp; - ib_dev->destroy_qp = hr_dev->hw->destroy_qp; - ib_dev->post_send = hr_dev->hw->post_send; - ib_dev->post_recv = hr_dev->hw->post_recv; - - /* CQ */ - ib_dev->create_cq = hns_roce_ib_create_cq; - ib_dev->modify_cq = hr_dev->hw->modify_cq; - ib_dev->destroy_cq = hns_roce_ib_destroy_cq; - ib_dev->req_notify_cq = hr_dev->hw->req_notify_cq; - ib_dev->poll_cq = hr_dev->hw->poll_cq; - - /* MR */ - ib_dev->get_dma_mr = hns_roce_get_dma_mr; - ib_dev->reg_user_mr = hns_roce_reg_user_mr; - ib_dev->dereg_mr = hns_roce_dereg_mr; if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_REREG_MR) { - ib_dev->rereg_user_mr = hns_roce_rereg_user_mr; ib_dev->uverbs_cmd_mask |= (1ULL << IB_USER_VERBS_CMD_REREG_MR); + ib_set_device_ops(ib_dev, &hns_roce_dev_mr_ops); } /* MW */ if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_MW) { - ib_dev->alloc_mw = hns_roce_alloc_mw; - ib_dev->dealloc_mw = hns_roce_dealloc_mw; ib_dev->uverbs_cmd_mask |= (1ULL << IB_USER_VERBS_CMD_ALLOC_MW) | (1ULL << IB_USER_VERBS_CMD_DEALLOC_MW); + ib_set_device_ops(ib_dev, &hns_roce_dev_mw_ops); } /* FRMR */ - if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_FRMR) { - ib_dev->alloc_mr = hns_roce_alloc_mr; - ib_dev->map_mr_sg = hns_roce_map_mr_sg; - } + if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_FRMR) + ib_set_device_ops(ib_dev, &hns_roce_dev_frmr_ops); /* SRQ */ if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_SRQ) { - ib_dev->create_srq = hns_roce_create_srq; - ib_dev->modify_srq = hr_dev->hw->modify_srq; - ib_dev->query_srq = hr_dev->hw->query_srq; - ib_dev->destroy_srq = hns_roce_destroy_srq; - ib_dev->post_srq_recv = hr_dev->hw->post_srq_recv; ib_dev->uverbs_cmd_mask |= (1ULL << IB_USER_VERBS_CMD_CREATE_SRQ) | (1ULL << IB_USER_VERBS_CMD_MODIFY_SRQ) | (1ULL << IB_USER_VERBS_CMD_QUERY_SRQ) | (1ULL << IB_USER_VERBS_CMD_DESTROY_SRQ) | (1ULL << IB_USER_VERBS_CMD_POST_SRQ_RECV); + ib_set_device_ops(ib_dev, &hns_roce_dev_srq_ops); + ib_set_device_ops(ib_dev, hr_dev->hw->hns_roce_dev_srq_ops); } - /* OTHERS */ - ib_dev->get_port_immutable = hns_roce_port_immutable; - ib_dev->disassociate_ucontext = hns_roce_disassociate_ucontext; - ib_dev->driver_id = RDMA_DRIVER_HNS; + ib_set_device_ops(ib_dev, hr_dev->hw->hns_roce_dev_ops); + ib_set_device_ops(ib_dev, &hns_roce_dev_ops); ret = ib_register_device(ib_dev, "hns_%d", NULL); if (ret) { dev_err(dev, "ib_register_device failed!\n"); -- cgit v1.2.3-59-g8ed1b From 15644f57cb6659349bc1ca7419fa89a28fbcfe61 Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Mon, 10 Dec 2018 21:09:36 +0200 Subject: RDMA/i40iw: Initialize ib_device_ops struct Initialize ib_device_ops with the supported operations using ib_set_device_ops(). Signed-off-by: Kamal Heib Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/i40iw/i40iw_verbs.c | 63 ++++++++++++++++--------------- 1 file changed, 33 insertions(+), 30 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/i40iw/i40iw_verbs.c b/drivers/infiniband/hw/i40iw/i40iw_verbs.c index a773d1edf7fd..9e42ac2db3ca 100644 --- a/drivers/infiniband/hw/i40iw/i40iw_verbs.c +++ b/drivers/infiniband/hw/i40iw/i40iw_verbs.c @@ -2721,6 +2721,39 @@ static int i40iw_query_pkey(struct ib_device *ibdev, return 0; } +static const struct ib_device_ops i40iw_dev_ops = { + .alloc_hw_stats = i40iw_alloc_hw_stats, + .alloc_mr = i40iw_alloc_mr, + .alloc_pd = i40iw_alloc_pd, + .alloc_ucontext = i40iw_alloc_ucontext, + .create_cq = i40iw_create_cq, + .create_qp = i40iw_create_qp, + .dealloc_pd = i40iw_dealloc_pd, + .dealloc_ucontext = i40iw_dealloc_ucontext, + .dereg_mr = i40iw_dereg_mr, + .destroy_cq = i40iw_destroy_cq, + .destroy_qp = i40iw_destroy_qp, + .drain_rq = i40iw_drain_rq, + .drain_sq = i40iw_drain_sq, + .get_dev_fw_str = i40iw_get_dev_fw_str, + .get_dma_mr = i40iw_get_dma_mr, + .get_hw_stats = i40iw_get_hw_stats, + .get_port_immutable = i40iw_port_immutable, + .map_mr_sg = i40iw_map_mr_sg, + .mmap = i40iw_mmap, + .modify_qp = i40iw_modify_qp, + .poll_cq = i40iw_poll_cq, + .post_recv = i40iw_post_recv, + .post_send = i40iw_post_send, + .query_device = i40iw_query_device, + .query_gid = i40iw_query_gid, + .query_pkey = i40iw_query_pkey, + .query_port = i40iw_query_port, + .query_qp = i40iw_query_qp, + .reg_user_mr = i40iw_reg_user_mr, + .req_notify_cq = i40iw_req_notify_cq, +}; + /** * i40iw_init_rdma_device - initialization of iwarp device * @iwdev: iwarp device @@ -2767,30 +2800,6 @@ static struct i40iw_ib_device *i40iw_init_rdma_device(struct i40iw_device *iwdev iwibdev->ibdev.phys_port_cnt = 1; iwibdev->ibdev.num_comp_vectors = iwdev->ceqs_count; iwibdev->ibdev.dev.parent = &pcidev->dev; - iwibdev->ibdev.query_port = i40iw_query_port; - iwibdev->ibdev.query_pkey = i40iw_query_pkey; - iwibdev->ibdev.query_gid = i40iw_query_gid; - iwibdev->ibdev.alloc_ucontext = i40iw_alloc_ucontext; - iwibdev->ibdev.dealloc_ucontext = i40iw_dealloc_ucontext; - iwibdev->ibdev.mmap = i40iw_mmap; - iwibdev->ibdev.alloc_pd = i40iw_alloc_pd; - iwibdev->ibdev.dealloc_pd = i40iw_dealloc_pd; - iwibdev->ibdev.create_qp = i40iw_create_qp; - iwibdev->ibdev.modify_qp = i40iw_modify_qp; - iwibdev->ibdev.query_qp = i40iw_query_qp; - iwibdev->ibdev.destroy_qp = i40iw_destroy_qp; - iwibdev->ibdev.create_cq = i40iw_create_cq; - iwibdev->ibdev.destroy_cq = i40iw_destroy_cq; - iwibdev->ibdev.get_dma_mr = i40iw_get_dma_mr; - iwibdev->ibdev.reg_user_mr = i40iw_reg_user_mr; - iwibdev->ibdev.dereg_mr = i40iw_dereg_mr; - iwibdev->ibdev.alloc_hw_stats = i40iw_alloc_hw_stats; - iwibdev->ibdev.get_hw_stats = i40iw_get_hw_stats; - iwibdev->ibdev.query_device = i40iw_query_device; - iwibdev->ibdev.drain_sq = i40iw_drain_sq; - iwibdev->ibdev.drain_rq = i40iw_drain_rq; - iwibdev->ibdev.alloc_mr = i40iw_alloc_mr; - iwibdev->ibdev.map_mr_sg = i40iw_map_mr_sg; iwibdev->ibdev.iwcm = kzalloc(sizeof(*iwibdev->ibdev.iwcm), GFP_KERNEL); if (!iwibdev->ibdev.iwcm) { ib_dealloc_device(&iwibdev->ibdev); @@ -2807,12 +2816,6 @@ static struct i40iw_ib_device *i40iw_init_rdma_device(struct i40iw_device *iwdev iwibdev->ibdev.iwcm->destroy_listen = i40iw_destroy_listen; memcpy(iwibdev->ibdev.iwcm->ifname, netdev->name, sizeof(iwibdev->ibdev.iwcm->ifname)); - iwibdev->ibdev.get_port_immutable = i40iw_port_immutable; - iwibdev->ibdev.get_dev_fw_str = i40iw_get_dev_fw_str; - iwibdev->ibdev.poll_cq = i40iw_poll_cq; - iwibdev->ibdev.req_notify_cq = i40iw_req_notify_cq; - iwibdev->ibdev.post_send = i40iw_post_send; - iwibdev->ibdev.post_recv = i40iw_post_recv; return iwibdev; } -- cgit v1.2.3-59-g8ed1b From 4725c4ba8d1c81c0b44a7a4f9b7f21bb0331e705 Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Mon, 10 Dec 2018 21:09:37 +0200 Subject: RDMA/mlx4: Initialize ib_device_ops struct Initialize ib_device_ops with the supported operations using ib_set_device_ops(). Signed-off-by: Kamal Heib Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx4/main.c | 178 +++++++++++++++++++++----------------- 1 file changed, 99 insertions(+), 79 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c index b73b5fa1822a..1f15ec3e2b83 100644 --- a/drivers/infiniband/hw/mlx4/main.c +++ b/drivers/infiniband/hw/mlx4/main.c @@ -2220,6 +2220,11 @@ static void mlx4_ib_fill_diag_counters(struct mlx4_ib_dev *ibdev, } } +static const struct ib_device_ops mlx4_ib_hw_stats_ops = { + .alloc_hw_stats = mlx4_ib_alloc_hw_stats, + .get_hw_stats = mlx4_ib_get_hw_stats, +}; + static int mlx4_ib_alloc_diag_counters(struct mlx4_ib_dev *ibdev) { struct mlx4_ib_diag_counters *diag = ibdev->diag_counters; @@ -2246,8 +2251,7 @@ static int mlx4_ib_alloc_diag_counters(struct mlx4_ib_dev *ibdev) diag[i].offset, i); } - ibdev->ib_dev.get_hw_stats = mlx4_ib_get_hw_stats; - ibdev->ib_dev.alloc_hw_stats = mlx4_ib_alloc_hw_stats; + ib_set_device_ops(&ibdev->ib_dev, &mlx4_ib_hw_stats_ops); return 0; @@ -2525,6 +2529,88 @@ static void get_fw_ver_str(struct ib_device *device, char *str) (int) dev->dev->caps.fw_ver & 0xffff); } +static const struct ib_device_ops mlx4_ib_dev_ops = { + .add_gid = mlx4_ib_add_gid, + .alloc_mr = mlx4_ib_alloc_mr, + .alloc_pd = mlx4_ib_alloc_pd, + .alloc_ucontext = mlx4_ib_alloc_ucontext, + .attach_mcast = mlx4_ib_mcg_attach, + .create_ah = mlx4_ib_create_ah, + .create_cq = mlx4_ib_create_cq, + .create_qp = mlx4_ib_create_qp, + .create_srq = mlx4_ib_create_srq, + .dealloc_pd = mlx4_ib_dealloc_pd, + .dealloc_ucontext = mlx4_ib_dealloc_ucontext, + .del_gid = mlx4_ib_del_gid, + .dereg_mr = mlx4_ib_dereg_mr, + .destroy_ah = mlx4_ib_destroy_ah, + .destroy_cq = mlx4_ib_destroy_cq, + .destroy_qp = mlx4_ib_destroy_qp, + .destroy_srq = mlx4_ib_destroy_srq, + .detach_mcast = mlx4_ib_mcg_detach, + .disassociate_ucontext = mlx4_ib_disassociate_ucontext, + .drain_rq = mlx4_ib_drain_rq, + .drain_sq = mlx4_ib_drain_sq, + .get_dev_fw_str = get_fw_ver_str, + .get_dma_mr = mlx4_ib_get_dma_mr, + .get_link_layer = mlx4_ib_port_link_layer, + .get_netdev = mlx4_ib_get_netdev, + .get_port_immutable = mlx4_port_immutable, + .map_mr_sg = mlx4_ib_map_mr_sg, + .mmap = mlx4_ib_mmap, + .modify_cq = mlx4_ib_modify_cq, + .modify_device = mlx4_ib_modify_device, + .modify_port = mlx4_ib_modify_port, + .modify_qp = mlx4_ib_modify_qp, + .modify_srq = mlx4_ib_modify_srq, + .poll_cq = mlx4_ib_poll_cq, + .post_recv = mlx4_ib_post_recv, + .post_send = mlx4_ib_post_send, + .post_srq_recv = mlx4_ib_post_srq_recv, + .process_mad = mlx4_ib_process_mad, + .query_ah = mlx4_ib_query_ah, + .query_device = mlx4_ib_query_device, + .query_gid = mlx4_ib_query_gid, + .query_pkey = mlx4_ib_query_pkey, + .query_port = mlx4_ib_query_port, + .query_qp = mlx4_ib_query_qp, + .query_srq = mlx4_ib_query_srq, + .reg_user_mr = mlx4_ib_reg_user_mr, + .req_notify_cq = mlx4_ib_arm_cq, + .rereg_user_mr = mlx4_ib_rereg_user_mr, + .resize_cq = mlx4_ib_resize_cq, +}; + +static const struct ib_device_ops mlx4_ib_dev_wq_ops = { + .create_rwq_ind_table = mlx4_ib_create_rwq_ind_table, + .create_wq = mlx4_ib_create_wq, + .destroy_rwq_ind_table = mlx4_ib_destroy_rwq_ind_table, + .destroy_wq = mlx4_ib_destroy_wq, + .modify_wq = mlx4_ib_modify_wq, +}; + +static const struct ib_device_ops mlx4_ib_dev_fmr_ops = { + .alloc_fmr = mlx4_ib_fmr_alloc, + .dealloc_fmr = mlx4_ib_fmr_dealloc, + .map_phys_fmr = mlx4_ib_map_phys_fmr, + .unmap_fmr = mlx4_ib_unmap_fmr, +}; + +static const struct ib_device_ops mlx4_ib_dev_mw_ops = { + .alloc_mw = mlx4_ib_alloc_mw, + .dealloc_mw = mlx4_ib_dealloc_mw, +}; + +static const struct ib_device_ops mlx4_ib_dev_xrc_ops = { + .alloc_xrcd = mlx4_ib_alloc_xrcd, + .dealloc_xrcd = mlx4_ib_dealloc_xrcd, +}; + +static const struct ib_device_ops mlx4_ib_dev_fs_ops = { + .create_flow = mlx4_ib_create_flow, + .destroy_flow = mlx4_ib_destroy_flow, +}; + static void *mlx4_ib_add(struct mlx4_dev *dev) { struct mlx4_ib_dev *ibdev; @@ -2580,9 +2666,6 @@ static void *mlx4_ib_add(struct mlx4_dev *dev) 1 : ibdev->num_ports; ibdev->ib_dev.num_comp_vectors = dev->caps.num_comp_vectors; ibdev->ib_dev.dev.parent = &dev->persist->pdev->dev; - ibdev->ib_dev.get_netdev = mlx4_ib_get_netdev; - ibdev->ib_dev.add_gid = mlx4_ib_add_gid; - ibdev->ib_dev.del_gid = mlx4_ib_del_gid; if (dev->caps.userspace_caps) ibdev->ib_dev.uverbs_abi_ver = MLX4_IB_UVERBS_ABI_VERSION; @@ -2615,116 +2698,53 @@ static void *mlx4_ib_add(struct mlx4_dev *dev) (1ull << IB_USER_VERBS_CMD_CREATE_XSRQ) | (1ull << IB_USER_VERBS_CMD_OPEN_QP); - ibdev->ib_dev.query_device = mlx4_ib_query_device; - ibdev->ib_dev.query_port = mlx4_ib_query_port; - ibdev->ib_dev.get_link_layer = mlx4_ib_port_link_layer; - ibdev->ib_dev.query_gid = mlx4_ib_query_gid; - ibdev->ib_dev.query_pkey = mlx4_ib_query_pkey; - ibdev->ib_dev.modify_device = mlx4_ib_modify_device; - ibdev->ib_dev.modify_port = mlx4_ib_modify_port; - ibdev->ib_dev.alloc_ucontext = mlx4_ib_alloc_ucontext; - ibdev->ib_dev.dealloc_ucontext = mlx4_ib_dealloc_ucontext; - ibdev->ib_dev.mmap = mlx4_ib_mmap; - ibdev->ib_dev.alloc_pd = mlx4_ib_alloc_pd; - ibdev->ib_dev.dealloc_pd = mlx4_ib_dealloc_pd; - ibdev->ib_dev.create_ah = mlx4_ib_create_ah; - ibdev->ib_dev.query_ah = mlx4_ib_query_ah; - ibdev->ib_dev.destroy_ah = mlx4_ib_destroy_ah; - ibdev->ib_dev.create_srq = mlx4_ib_create_srq; - ibdev->ib_dev.modify_srq = mlx4_ib_modify_srq; - ibdev->ib_dev.query_srq = mlx4_ib_query_srq; - ibdev->ib_dev.destroy_srq = mlx4_ib_destroy_srq; - ibdev->ib_dev.post_srq_recv = mlx4_ib_post_srq_recv; - ibdev->ib_dev.create_qp = mlx4_ib_create_qp; - ibdev->ib_dev.modify_qp = mlx4_ib_modify_qp; - ibdev->ib_dev.query_qp = mlx4_ib_query_qp; - ibdev->ib_dev.destroy_qp = mlx4_ib_destroy_qp; - ibdev->ib_dev.drain_sq = mlx4_ib_drain_sq; - ibdev->ib_dev.drain_rq = mlx4_ib_drain_rq; - ibdev->ib_dev.post_send = mlx4_ib_post_send; - ibdev->ib_dev.post_recv = mlx4_ib_post_recv; - ibdev->ib_dev.create_cq = mlx4_ib_create_cq; - ibdev->ib_dev.modify_cq = mlx4_ib_modify_cq; - ibdev->ib_dev.resize_cq = mlx4_ib_resize_cq; - ibdev->ib_dev.destroy_cq = mlx4_ib_destroy_cq; - ibdev->ib_dev.poll_cq = mlx4_ib_poll_cq; - ibdev->ib_dev.req_notify_cq = mlx4_ib_arm_cq; - ibdev->ib_dev.get_dma_mr = mlx4_ib_get_dma_mr; - ibdev->ib_dev.reg_user_mr = mlx4_ib_reg_user_mr; - ibdev->ib_dev.rereg_user_mr = mlx4_ib_rereg_user_mr; - ibdev->ib_dev.dereg_mr = mlx4_ib_dereg_mr; - ibdev->ib_dev.alloc_mr = mlx4_ib_alloc_mr; - ibdev->ib_dev.map_mr_sg = mlx4_ib_map_mr_sg; - ibdev->ib_dev.attach_mcast = mlx4_ib_mcg_attach; - ibdev->ib_dev.detach_mcast = mlx4_ib_mcg_detach; - ibdev->ib_dev.process_mad = mlx4_ib_process_mad; - ibdev->ib_dev.get_port_immutable = mlx4_port_immutable; - ibdev->ib_dev.get_dev_fw_str = get_fw_ver_str; - ibdev->ib_dev.disassociate_ucontext = mlx4_ib_disassociate_ucontext; - + ib_set_device_ops(&ibdev->ib_dev, &mlx4_ib_dev_ops); ibdev->ib_dev.uverbs_ex_cmd_mask |= - (1ull << IB_USER_VERBS_EX_CMD_MODIFY_CQ); + (1ull << IB_USER_VERBS_EX_CMD_MODIFY_CQ) | + (1ull << IB_USER_VERBS_EX_CMD_QUERY_DEVICE) | + (1ull << IB_USER_VERBS_EX_CMD_CREATE_CQ) | + (1ull << IB_USER_VERBS_EX_CMD_CREATE_QP); if ((dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_RSS) && ((mlx4_ib_port_link_layer(&ibdev->ib_dev, 1) == IB_LINK_LAYER_ETHERNET) || (mlx4_ib_port_link_layer(&ibdev->ib_dev, 2) == IB_LINK_LAYER_ETHERNET))) { - ibdev->ib_dev.create_wq = mlx4_ib_create_wq; - ibdev->ib_dev.modify_wq = mlx4_ib_modify_wq; - ibdev->ib_dev.destroy_wq = mlx4_ib_destroy_wq; - ibdev->ib_dev.create_rwq_ind_table = - mlx4_ib_create_rwq_ind_table; - ibdev->ib_dev.destroy_rwq_ind_table = - mlx4_ib_destroy_rwq_ind_table; ibdev->ib_dev.uverbs_ex_cmd_mask |= (1ull << IB_USER_VERBS_EX_CMD_CREATE_WQ) | (1ull << IB_USER_VERBS_EX_CMD_MODIFY_WQ) | (1ull << IB_USER_VERBS_EX_CMD_DESTROY_WQ) | (1ull << IB_USER_VERBS_EX_CMD_CREATE_RWQ_IND_TBL) | (1ull << IB_USER_VERBS_EX_CMD_DESTROY_RWQ_IND_TBL); + ib_set_device_ops(&ibdev->ib_dev, &mlx4_ib_dev_wq_ops); } - if (!mlx4_is_slave(ibdev->dev)) { - ibdev->ib_dev.alloc_fmr = mlx4_ib_fmr_alloc; - ibdev->ib_dev.map_phys_fmr = mlx4_ib_map_phys_fmr; - ibdev->ib_dev.unmap_fmr = mlx4_ib_unmap_fmr; - ibdev->ib_dev.dealloc_fmr = mlx4_ib_fmr_dealloc; - } + if (!mlx4_is_slave(ibdev->dev)) + ib_set_device_ops(&ibdev->ib_dev, &mlx4_ib_dev_fmr_ops); if (dev->caps.flags & MLX4_DEV_CAP_FLAG_MEM_WINDOW || dev->caps.bmme_flags & MLX4_BMME_FLAG_TYPE_2_WIN) { - ibdev->ib_dev.alloc_mw = mlx4_ib_alloc_mw; - ibdev->ib_dev.dealloc_mw = mlx4_ib_dealloc_mw; - ibdev->ib_dev.uverbs_cmd_mask |= (1ull << IB_USER_VERBS_CMD_ALLOC_MW) | (1ull << IB_USER_VERBS_CMD_DEALLOC_MW); + ib_set_device_ops(&ibdev->ib_dev, &mlx4_ib_dev_mw_ops); } if (dev->caps.flags & MLX4_DEV_CAP_FLAG_XRC) { - ibdev->ib_dev.alloc_xrcd = mlx4_ib_alloc_xrcd; - ibdev->ib_dev.dealloc_xrcd = mlx4_ib_dealloc_xrcd; ibdev->ib_dev.uverbs_cmd_mask |= (1ull << IB_USER_VERBS_CMD_OPEN_XRCD) | (1ull << IB_USER_VERBS_CMD_CLOSE_XRCD); + ib_set_device_ops(&ibdev->ib_dev, &mlx4_ib_dev_xrc_ops); } if (check_flow_steering_support(dev)) { ibdev->steering_support = MLX4_STEERING_MODE_DEVICE_MANAGED; - ibdev->ib_dev.create_flow = mlx4_ib_create_flow; - ibdev->ib_dev.destroy_flow = mlx4_ib_destroy_flow; - ibdev->ib_dev.uverbs_ex_cmd_mask |= (1ull << IB_USER_VERBS_EX_CMD_CREATE_FLOW) | (1ull << IB_USER_VERBS_EX_CMD_DESTROY_FLOW); + ib_set_device_ops(&ibdev->ib_dev, &mlx4_ib_dev_fs_ops); } - ibdev->ib_dev.uverbs_ex_cmd_mask |= - (1ull << IB_USER_VERBS_EX_CMD_QUERY_DEVICE) | - (1ull << IB_USER_VERBS_EX_CMD_CREATE_CQ) | - (1ull << IB_USER_VERBS_EX_CMD_CREATE_QP); - mlx4_ib_alloc_eqs(dev, ibdev); spin_lock_init(&iboe->lock); -- cgit v1.2.3-59-g8ed1b From 96458233ee73b6082f75c1a55c9a4ad6ea7913d5 Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Mon, 10 Dec 2018 21:09:38 +0200 Subject: RDMA/mlx5: Initialize ib_device_ops struct Initialize ib_device_ops with the supported operations using ib_set_device_ops(). Signed-off-by: Kamal Heib Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/main.c | 227 ++++++++++++++++++++++---------------- 1 file changed, 131 insertions(+), 96 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index af091f363b81..47d9cd260846 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -5812,6 +5812,94 @@ static void mlx5_ib_stage_flow_db_cleanup(struct mlx5_ib_dev *dev) kfree(dev->flow_db); } +static const struct ib_device_ops mlx5_ib_dev_ops = { + .add_gid = mlx5_ib_add_gid, + .alloc_mr = mlx5_ib_alloc_mr, + .alloc_pd = mlx5_ib_alloc_pd, + .alloc_ucontext = mlx5_ib_alloc_ucontext, + .attach_mcast = mlx5_ib_mcg_attach, + .check_mr_status = mlx5_ib_check_mr_status, + .create_ah = mlx5_ib_create_ah, + .create_counters = mlx5_ib_create_counters, + .create_cq = mlx5_ib_create_cq, + .create_flow = mlx5_ib_create_flow, + .create_qp = mlx5_ib_create_qp, + .create_srq = mlx5_ib_create_srq, + .dealloc_pd = mlx5_ib_dealloc_pd, + .dealloc_ucontext = mlx5_ib_dealloc_ucontext, + .del_gid = mlx5_ib_del_gid, + .dereg_mr = mlx5_ib_dereg_mr, + .destroy_ah = mlx5_ib_destroy_ah, + .destroy_counters = mlx5_ib_destroy_counters, + .destroy_cq = mlx5_ib_destroy_cq, + .destroy_flow = mlx5_ib_destroy_flow, + .destroy_flow_action = mlx5_ib_destroy_flow_action, + .destroy_qp = mlx5_ib_destroy_qp, + .destroy_srq = mlx5_ib_destroy_srq, + .detach_mcast = mlx5_ib_mcg_detach, + .disassociate_ucontext = mlx5_ib_disassociate_ucontext, + .drain_rq = mlx5_ib_drain_rq, + .drain_sq = mlx5_ib_drain_sq, + .get_dev_fw_str = get_dev_fw_str, + .get_dma_mr = mlx5_ib_get_dma_mr, + .get_link_layer = mlx5_ib_port_link_layer, + .map_mr_sg = mlx5_ib_map_mr_sg, + .mmap = mlx5_ib_mmap, + .modify_cq = mlx5_ib_modify_cq, + .modify_device = mlx5_ib_modify_device, + .modify_port = mlx5_ib_modify_port, + .modify_qp = mlx5_ib_modify_qp, + .modify_srq = mlx5_ib_modify_srq, + .poll_cq = mlx5_ib_poll_cq, + .post_recv = mlx5_ib_post_recv, + .post_send = mlx5_ib_post_send, + .post_srq_recv = mlx5_ib_post_srq_recv, + .process_mad = mlx5_ib_process_mad, + .query_ah = mlx5_ib_query_ah, + .query_device = mlx5_ib_query_device, + .query_gid = mlx5_ib_query_gid, + .query_pkey = mlx5_ib_query_pkey, + .query_qp = mlx5_ib_query_qp, + .query_srq = mlx5_ib_query_srq, + .read_counters = mlx5_ib_read_counters, + .reg_user_mr = mlx5_ib_reg_user_mr, + .req_notify_cq = mlx5_ib_arm_cq, + .rereg_user_mr = mlx5_ib_rereg_user_mr, + .resize_cq = mlx5_ib_resize_cq, +}; + +static const struct ib_device_ops mlx5_ib_dev_flow_ipsec_ops = { + .create_flow_action_esp = mlx5_ib_create_flow_action_esp, + .modify_flow_action_esp = mlx5_ib_modify_flow_action_esp, +}; + +static const struct ib_device_ops mlx5_ib_dev_ipoib_enhanced_ops = { + .rdma_netdev_get_params = mlx5_ib_rn_get_params, +}; + +static const struct ib_device_ops mlx5_ib_dev_sriov_ops = { + .get_vf_config = mlx5_ib_get_vf_config, + .get_vf_stats = mlx5_ib_get_vf_stats, + .set_vf_guid = mlx5_ib_set_vf_guid, + .set_vf_link_state = mlx5_ib_set_vf_link_state, +}; + +static const struct ib_device_ops mlx5_ib_dev_mw_ops = { + .alloc_mw = mlx5_ib_alloc_mw, + .dealloc_mw = mlx5_ib_dealloc_mw, +}; + +static const struct ib_device_ops mlx5_ib_dev_xrc_ops = { + .alloc_xrcd = mlx5_ib_alloc_xrcd, + .dealloc_xrcd = mlx5_ib_dealloc_xrcd, +}; + +static const struct ib_device_ops mlx5_ib_dev_dm_ops = { + .alloc_dm = mlx5_ib_alloc_dm, + .dealloc_dm = mlx5_ib_dealloc_dm, + .reg_dm_mr = mlx5_ib_reg_dm_mr, +}; + int mlx5_ib_stage_caps_init(struct mlx5_ib_dev *dev) { struct mlx5_core_dev *mdev = dev->mdev; @@ -5850,108 +5938,42 @@ int mlx5_ib_stage_caps_init(struct mlx5_ib_dev *dev) (1ull << IB_USER_VERBS_EX_CMD_CREATE_CQ) | (1ull << IB_USER_VERBS_EX_CMD_CREATE_QP) | (1ull << IB_USER_VERBS_EX_CMD_MODIFY_QP) | - (1ull << IB_USER_VERBS_EX_CMD_MODIFY_CQ); - - dev->ib_dev.query_device = mlx5_ib_query_device; - dev->ib_dev.get_link_layer = mlx5_ib_port_link_layer; - dev->ib_dev.query_gid = mlx5_ib_query_gid; - dev->ib_dev.add_gid = mlx5_ib_add_gid; - dev->ib_dev.del_gid = mlx5_ib_del_gid; - dev->ib_dev.query_pkey = mlx5_ib_query_pkey; - dev->ib_dev.modify_device = mlx5_ib_modify_device; - dev->ib_dev.modify_port = mlx5_ib_modify_port; - dev->ib_dev.alloc_ucontext = mlx5_ib_alloc_ucontext; - dev->ib_dev.dealloc_ucontext = mlx5_ib_dealloc_ucontext; - dev->ib_dev.mmap = mlx5_ib_mmap; - dev->ib_dev.alloc_pd = mlx5_ib_alloc_pd; - dev->ib_dev.dealloc_pd = mlx5_ib_dealloc_pd; - dev->ib_dev.create_ah = mlx5_ib_create_ah; - dev->ib_dev.query_ah = mlx5_ib_query_ah; - dev->ib_dev.destroy_ah = mlx5_ib_destroy_ah; - dev->ib_dev.create_srq = mlx5_ib_create_srq; - dev->ib_dev.modify_srq = mlx5_ib_modify_srq; - dev->ib_dev.query_srq = mlx5_ib_query_srq; - dev->ib_dev.destroy_srq = mlx5_ib_destroy_srq; - dev->ib_dev.post_srq_recv = mlx5_ib_post_srq_recv; - dev->ib_dev.create_qp = mlx5_ib_create_qp; - dev->ib_dev.modify_qp = mlx5_ib_modify_qp; - dev->ib_dev.query_qp = mlx5_ib_query_qp; - dev->ib_dev.destroy_qp = mlx5_ib_destroy_qp; - dev->ib_dev.drain_sq = mlx5_ib_drain_sq; - dev->ib_dev.drain_rq = mlx5_ib_drain_rq; - dev->ib_dev.post_send = mlx5_ib_post_send; - dev->ib_dev.post_recv = mlx5_ib_post_recv; - dev->ib_dev.create_cq = mlx5_ib_create_cq; - dev->ib_dev.modify_cq = mlx5_ib_modify_cq; - dev->ib_dev.resize_cq = mlx5_ib_resize_cq; - dev->ib_dev.destroy_cq = mlx5_ib_destroy_cq; - dev->ib_dev.poll_cq = mlx5_ib_poll_cq; - dev->ib_dev.req_notify_cq = mlx5_ib_arm_cq; - dev->ib_dev.get_dma_mr = mlx5_ib_get_dma_mr; - dev->ib_dev.reg_user_mr = mlx5_ib_reg_user_mr; - dev->ib_dev.rereg_user_mr = mlx5_ib_rereg_user_mr; - dev->ib_dev.dereg_mr = mlx5_ib_dereg_mr; - dev->ib_dev.attach_mcast = mlx5_ib_mcg_attach; - dev->ib_dev.detach_mcast = mlx5_ib_mcg_detach; - dev->ib_dev.process_mad = mlx5_ib_process_mad; - dev->ib_dev.alloc_mr = mlx5_ib_alloc_mr; - dev->ib_dev.map_mr_sg = mlx5_ib_map_mr_sg; - dev->ib_dev.check_mr_status = mlx5_ib_check_mr_status; - dev->ib_dev.get_dev_fw_str = get_dev_fw_str; + (1ull << IB_USER_VERBS_EX_CMD_MODIFY_CQ) | + (1ull << IB_USER_VERBS_EX_CMD_CREATE_FLOW) | + (1ull << IB_USER_VERBS_EX_CMD_DESTROY_FLOW); + if (MLX5_CAP_GEN(mdev, ipoib_enhanced_offloads) && IS_ENABLED(CONFIG_MLX5_CORE_IPOIB)) - dev->ib_dev.rdma_netdev_get_params = mlx5_ib_rn_get_params; - - if (mlx5_core_is_pf(mdev)) { - dev->ib_dev.get_vf_config = mlx5_ib_get_vf_config; - dev->ib_dev.set_vf_link_state = mlx5_ib_set_vf_link_state; - dev->ib_dev.get_vf_stats = mlx5_ib_get_vf_stats; - dev->ib_dev.set_vf_guid = mlx5_ib_set_vf_guid; - } + ib_set_device_ops(&dev->ib_dev, + &mlx5_ib_dev_ipoib_enhanced_ops); - dev->ib_dev.disassociate_ucontext = mlx5_ib_disassociate_ucontext; + if (mlx5_core_is_pf(mdev)) + ib_set_device_ops(&dev->ib_dev, &mlx5_ib_dev_sriov_ops); dev->umr_fence = mlx5_get_umr_fence(MLX5_CAP_GEN(mdev, umr_fence)); if (MLX5_CAP_GEN(mdev, imaicl)) { - dev->ib_dev.alloc_mw = mlx5_ib_alloc_mw; - dev->ib_dev.dealloc_mw = mlx5_ib_dealloc_mw; dev->ib_dev.uverbs_cmd_mask |= (1ull << IB_USER_VERBS_CMD_ALLOC_MW) | (1ull << IB_USER_VERBS_CMD_DEALLOC_MW); + ib_set_device_ops(&dev->ib_dev, &mlx5_ib_dev_mw_ops); } if (MLX5_CAP_GEN(mdev, xrc)) { - dev->ib_dev.alloc_xrcd = mlx5_ib_alloc_xrcd; - dev->ib_dev.dealloc_xrcd = mlx5_ib_dealloc_xrcd; dev->ib_dev.uverbs_cmd_mask |= (1ull << IB_USER_VERBS_CMD_OPEN_XRCD) | (1ull << IB_USER_VERBS_CMD_CLOSE_XRCD); + ib_set_device_ops(&dev->ib_dev, &mlx5_ib_dev_xrc_ops); } - if (MLX5_CAP_DEV_MEM(mdev, memic)) { - dev->ib_dev.alloc_dm = mlx5_ib_alloc_dm; - dev->ib_dev.dealloc_dm = mlx5_ib_dealloc_dm; - dev->ib_dev.reg_dm_mr = mlx5_ib_reg_dm_mr; - } + if (MLX5_CAP_DEV_MEM(mdev, memic)) + ib_set_device_ops(&dev->ib_dev, &mlx5_ib_dev_dm_ops); - dev->ib_dev.create_flow = mlx5_ib_create_flow; - dev->ib_dev.destroy_flow = mlx5_ib_destroy_flow; - dev->ib_dev.uverbs_ex_cmd_mask |= - (1ull << IB_USER_VERBS_EX_CMD_CREATE_FLOW) | - (1ull << IB_USER_VERBS_EX_CMD_DESTROY_FLOW); if (mlx5_accel_ipsec_device_caps(dev->mdev) & - MLX5_ACCEL_IPSEC_CAP_DEVICE) { - dev->ib_dev.create_flow_action_esp = - mlx5_ib_create_flow_action_esp; - dev->ib_dev.modify_flow_action_esp = - mlx5_ib_modify_flow_action_esp; - } - dev->ib_dev.destroy_flow_action = mlx5_ib_destroy_flow_action; + MLX5_ACCEL_IPSEC_CAP_DEVICE) + ib_set_device_ops(&dev->ib_dev, &mlx5_ib_dev_flow_ipsec_ops); dev->ib_dev.driver_id = RDMA_DRIVER_MLX5; - dev->ib_dev.create_counters = mlx5_ib_create_counters; - dev->ib_dev.destroy_counters = mlx5_ib_destroy_counters; - dev->ib_dev.read_counters = mlx5_ib_read_counters; + ib_set_device_ops(&dev->ib_dev, &mlx5_ib_dev_ops); if (IS_ENABLED(CONFIG_INFINIBAND_USER_ACCESS)) dev->ib_dev.driver_def = mlx5_ib_defs; @@ -5968,22 +5990,37 @@ int mlx5_ib_stage_caps_init(struct mlx5_ib_dev *dev) return 0; } +static const struct ib_device_ops mlx5_ib_dev_port_ops = { + .get_port_immutable = mlx5_port_immutable, + .query_port = mlx5_ib_query_port, +}; + static int mlx5_ib_stage_non_default_cb(struct mlx5_ib_dev *dev) { - dev->ib_dev.get_port_immutable = mlx5_port_immutable; - dev->ib_dev.query_port = mlx5_ib_query_port; - + ib_set_device_ops(&dev->ib_dev, &mlx5_ib_dev_port_ops); return 0; } +static const struct ib_device_ops mlx5_ib_dev_port_rep_ops = { + .get_port_immutable = mlx5_port_rep_immutable, + .query_port = mlx5_ib_rep_query_port, +}; + int mlx5_ib_stage_rep_non_default_cb(struct mlx5_ib_dev *dev) { - dev->ib_dev.get_port_immutable = mlx5_port_rep_immutable; - dev->ib_dev.query_port = mlx5_ib_rep_query_port; - + ib_set_device_ops(&dev->ib_dev, &mlx5_ib_dev_port_rep_ops); return 0; } +static const struct ib_device_ops mlx5_ib_dev_common_roce_ops = { + .create_rwq_ind_table = mlx5_ib_create_rwq_ind_table, + .create_wq = mlx5_ib_create_wq, + .destroy_rwq_ind_table = mlx5_ib_destroy_rwq_ind_table, + .destroy_wq = mlx5_ib_destroy_wq, + .get_netdev = mlx5_ib_get_netdev, + .modify_wq = mlx5_ib_modify_wq, +}; + static int mlx5_ib_stage_common_roce_init(struct mlx5_ib_dev *dev) { u8 port_num; @@ -5995,19 +6032,13 @@ static int mlx5_ib_stage_common_roce_init(struct mlx5_ib_dev *dev) dev->roce[i].last_port_state = IB_PORT_DOWN; } - dev->ib_dev.get_netdev = mlx5_ib_get_netdev; - dev->ib_dev.create_wq = mlx5_ib_create_wq; - dev->ib_dev.modify_wq = mlx5_ib_modify_wq; - dev->ib_dev.destroy_wq = mlx5_ib_destroy_wq; - dev->ib_dev.create_rwq_ind_table = mlx5_ib_create_rwq_ind_table; - dev->ib_dev.destroy_rwq_ind_table = mlx5_ib_destroy_rwq_ind_table; - dev->ib_dev.uverbs_ex_cmd_mask |= (1ull << IB_USER_VERBS_EX_CMD_CREATE_WQ) | (1ull << IB_USER_VERBS_EX_CMD_MODIFY_WQ) | (1ull << IB_USER_VERBS_EX_CMD_DESTROY_WQ) | (1ull << IB_USER_VERBS_EX_CMD_CREATE_RWQ_IND_TBL) | (1ull << IB_USER_VERBS_EX_CMD_DESTROY_RWQ_IND_TBL); + ib_set_device_ops(&dev->ib_dev, &mlx5_ib_dev_common_roce_ops); port_num = mlx5_core_native_port_num(dev->mdev) - 1; @@ -6106,11 +6137,15 @@ void mlx5_ib_stage_odp_cleanup(struct mlx5_ib_dev *dev) mlx5_ib_odp_cleanup_one(dev); } +static const struct ib_device_ops mlx5_ib_dev_hw_stats_ops = { + .alloc_hw_stats = mlx5_ib_alloc_hw_stats, + .get_hw_stats = mlx5_ib_get_hw_stats, +}; + int mlx5_ib_stage_counters_init(struct mlx5_ib_dev *dev) { if (MLX5_CAP_GEN(dev->mdev, max_qp_cnt)) { - dev->ib_dev.get_hw_stats = mlx5_ib_get_hw_stats; - dev->ib_dev.alloc_hw_stats = mlx5_ib_alloc_hw_stats; + ib_set_device_ops(&dev->ib_dev, &mlx5_ib_dev_hw_stats_ops); return mlx5_ib_alloc_counters(dev); } -- cgit v1.2.3-59-g8ed1b From 56e2a43136b77205fc4a33f8915d6a289396668a Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Mon, 10 Dec 2018 21:09:39 +0200 Subject: RDMA/mthca: Initialize ib_device_ops struct Initialize ib_device_ops with the supported operations using ib_set_device_ops(). Signed-off-by: Kamal Heib Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mthca/mthca_provider.c | 139 +++++++++++++++++---------- 1 file changed, 88 insertions(+), 51 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mthca/mthca_provider.c b/drivers/infiniband/hw/mthca/mthca_provider.c index 691c6f048938..c697ec54ea5f 100644 --- a/drivers/infiniband/hw/mthca/mthca_provider.c +++ b/drivers/infiniband/hw/mthca/mthca_provider.c @@ -1193,6 +1193,81 @@ static void get_dev_fw_str(struct ib_device *device, char *str) (int) dev->fw_ver & 0xffff); } +static const struct ib_device_ops mthca_dev_ops = { + .alloc_pd = mthca_alloc_pd, + .alloc_ucontext = mthca_alloc_ucontext, + .attach_mcast = mthca_multicast_attach, + .create_ah = mthca_ah_create, + .create_cq = mthca_create_cq, + .create_qp = mthca_create_qp, + .dealloc_pd = mthca_dealloc_pd, + .dealloc_ucontext = mthca_dealloc_ucontext, + .dereg_mr = mthca_dereg_mr, + .destroy_ah = mthca_ah_destroy, + .destroy_cq = mthca_destroy_cq, + .destroy_qp = mthca_destroy_qp, + .detach_mcast = mthca_multicast_detach, + .get_dev_fw_str = get_dev_fw_str, + .get_dma_mr = mthca_get_dma_mr, + .get_port_immutable = mthca_port_immutable, + .mmap = mthca_mmap_uar, + .modify_device = mthca_modify_device, + .modify_port = mthca_modify_port, + .modify_qp = mthca_modify_qp, + .poll_cq = mthca_poll_cq, + .process_mad = mthca_process_mad, + .query_ah = mthca_ah_query, + .query_device = mthca_query_device, + .query_gid = mthca_query_gid, + .query_pkey = mthca_query_pkey, + .query_port = mthca_query_port, + .query_qp = mthca_query_qp, + .reg_user_mr = mthca_reg_user_mr, + .resize_cq = mthca_resize_cq, +}; + +static const struct ib_device_ops mthca_dev_arbel_srq_ops = { + .create_srq = mthca_create_srq, + .destroy_srq = mthca_destroy_srq, + .modify_srq = mthca_modify_srq, + .post_srq_recv = mthca_arbel_post_srq_recv, + .query_srq = mthca_query_srq, +}; + +static const struct ib_device_ops mthca_dev_tavor_srq_ops = { + .create_srq = mthca_create_srq, + .destroy_srq = mthca_destroy_srq, + .modify_srq = mthca_modify_srq, + .post_srq_recv = mthca_tavor_post_srq_recv, + .query_srq = mthca_query_srq, +}; + +static const struct ib_device_ops mthca_dev_arbel_fmr_ops = { + .alloc_fmr = mthca_alloc_fmr, + .dealloc_fmr = mthca_dealloc_fmr, + .map_phys_fmr = mthca_arbel_map_phys_fmr, + .unmap_fmr = mthca_unmap_fmr, +}; + +static const struct ib_device_ops mthca_dev_tavor_fmr_ops = { + .alloc_fmr = mthca_alloc_fmr, + .dealloc_fmr = mthca_dealloc_fmr, + .map_phys_fmr = mthca_tavor_map_phys_fmr, + .unmap_fmr = mthca_unmap_fmr, +}; + +static const struct ib_device_ops mthca_dev_arbel_ops = { + .post_recv = mthca_arbel_post_receive, + .post_send = mthca_arbel_post_send, + .req_notify_cq = mthca_arbel_arm_cq, +}; + +static const struct ib_device_ops mthca_dev_tavor_ops = { + .post_recv = mthca_tavor_post_receive, + .post_send = mthca_tavor_post_send, + .req_notify_cq = mthca_tavor_arm_cq, +}; + int mthca_register_device(struct mthca_dev *dev) { int ret; @@ -1226,26 +1301,8 @@ int mthca_register_device(struct mthca_dev *dev) dev->ib_dev.phys_port_cnt = dev->limits.num_ports; dev->ib_dev.num_comp_vectors = 1; dev->ib_dev.dev.parent = &dev->pdev->dev; - dev->ib_dev.query_device = mthca_query_device; - dev->ib_dev.query_port = mthca_query_port; - dev->ib_dev.modify_device = mthca_modify_device; - dev->ib_dev.modify_port = mthca_modify_port; - dev->ib_dev.query_pkey = mthca_query_pkey; - dev->ib_dev.query_gid = mthca_query_gid; - dev->ib_dev.alloc_ucontext = mthca_alloc_ucontext; - dev->ib_dev.dealloc_ucontext = mthca_dealloc_ucontext; - dev->ib_dev.mmap = mthca_mmap_uar; - dev->ib_dev.alloc_pd = mthca_alloc_pd; - dev->ib_dev.dealloc_pd = mthca_dealloc_pd; - dev->ib_dev.create_ah = mthca_ah_create; - dev->ib_dev.query_ah = mthca_ah_query; - dev->ib_dev.destroy_ah = mthca_ah_destroy; if (dev->mthca_flags & MTHCA_FLAG_SRQ) { - dev->ib_dev.create_srq = mthca_create_srq; - dev->ib_dev.modify_srq = mthca_modify_srq; - dev->ib_dev.query_srq = mthca_query_srq; - dev->ib_dev.destroy_srq = mthca_destroy_srq; dev->ib_dev.uverbs_cmd_mask |= (1ull << IB_USER_VERBS_CMD_CREATE_SRQ) | (1ull << IB_USER_VERBS_CMD_MODIFY_SRQ) | @@ -1253,48 +1310,28 @@ int mthca_register_device(struct mthca_dev *dev) (1ull << IB_USER_VERBS_CMD_DESTROY_SRQ); if (mthca_is_memfree(dev)) - dev->ib_dev.post_srq_recv = mthca_arbel_post_srq_recv; + ib_set_device_ops(&dev->ib_dev, + &mthca_dev_arbel_srq_ops); else - dev->ib_dev.post_srq_recv = mthca_tavor_post_srq_recv; + ib_set_device_ops(&dev->ib_dev, + &mthca_dev_tavor_srq_ops); } - dev->ib_dev.create_qp = mthca_create_qp; - dev->ib_dev.modify_qp = mthca_modify_qp; - dev->ib_dev.query_qp = mthca_query_qp; - dev->ib_dev.destroy_qp = mthca_destroy_qp; - dev->ib_dev.create_cq = mthca_create_cq; - dev->ib_dev.resize_cq = mthca_resize_cq; - dev->ib_dev.destroy_cq = mthca_destroy_cq; - dev->ib_dev.poll_cq = mthca_poll_cq; - dev->ib_dev.get_dma_mr = mthca_get_dma_mr; - dev->ib_dev.reg_user_mr = mthca_reg_user_mr; - dev->ib_dev.dereg_mr = mthca_dereg_mr; - dev->ib_dev.get_port_immutable = mthca_port_immutable; - dev->ib_dev.get_dev_fw_str = get_dev_fw_str; - if (dev->mthca_flags & MTHCA_FLAG_FMR) { - dev->ib_dev.alloc_fmr = mthca_alloc_fmr; - dev->ib_dev.unmap_fmr = mthca_unmap_fmr; - dev->ib_dev.dealloc_fmr = mthca_dealloc_fmr; if (mthca_is_memfree(dev)) - dev->ib_dev.map_phys_fmr = mthca_arbel_map_phys_fmr; + ib_set_device_ops(&dev->ib_dev, + &mthca_dev_arbel_fmr_ops); else - dev->ib_dev.map_phys_fmr = mthca_tavor_map_phys_fmr; + ib_set_device_ops(&dev->ib_dev, + &mthca_dev_tavor_fmr_ops); } - dev->ib_dev.attach_mcast = mthca_multicast_attach; - dev->ib_dev.detach_mcast = mthca_multicast_detach; - dev->ib_dev.process_mad = mthca_process_mad; + ib_set_device_ops(&dev->ib_dev, &mthca_dev_ops); - if (mthca_is_memfree(dev)) { - dev->ib_dev.req_notify_cq = mthca_arbel_arm_cq; - dev->ib_dev.post_send = mthca_arbel_post_send; - dev->ib_dev.post_recv = mthca_arbel_post_receive; - } else { - dev->ib_dev.req_notify_cq = mthca_tavor_arm_cq; - dev->ib_dev.post_send = mthca_tavor_post_send; - dev->ib_dev.post_recv = mthca_tavor_post_receive; - } + if (mthca_is_memfree(dev)) + ib_set_device_ops(&dev->ib_dev, &mthca_dev_arbel_ops); + else + ib_set_device_ops(&dev->ib_dev, &mthca_dev_tavor_ops); mutex_init(&dev->cap_mask_mutex); -- cgit v1.2.3-59-g8ed1b From 5a6c6e71acb069f0d7707fbc9b29a93cbe241d46 Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Mon, 10 Dec 2018 21:09:40 +0200 Subject: RDMA/nes: Initialize ib_device_ops struct Initialize ib_device_ops with the supported operations using ib_set_device_ops(). Signed-off-by: Kamal Heib Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/nes/nes_verbs.c | 67 ++++++++++++++++++----------------- 1 file changed, 35 insertions(+), 32 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/nes/nes_verbs.c b/drivers/infiniband/hw/nes/nes_verbs.c index 92d1cadd4cfd..f9d510431900 100644 --- a/drivers/infiniband/hw/nes/nes_verbs.c +++ b/drivers/infiniband/hw/nes/nes_verbs.c @@ -3627,6 +3627,39 @@ static void get_dev_fw_str(struct ib_device *dev, char *str) (nesvnic->nesdev->nesadapter->firmware_version & 0x000000ff)); } +static const struct ib_device_ops nes_dev_ops = { + .alloc_mr = nes_alloc_mr, + .alloc_mw = nes_alloc_mw, + .alloc_pd = nes_alloc_pd, + .alloc_ucontext = nes_alloc_ucontext, + .create_cq = nes_create_cq, + .create_qp = nes_create_qp, + .dealloc_mw = nes_dealloc_mw, + .dealloc_pd = nes_dealloc_pd, + .dealloc_ucontext = nes_dealloc_ucontext, + .dereg_mr = nes_dereg_mr, + .destroy_cq = nes_destroy_cq, + .destroy_qp = nes_destroy_qp, + .drain_rq = nes_drain_rq, + .drain_sq = nes_drain_sq, + .get_dev_fw_str = get_dev_fw_str, + .get_dma_mr = nes_get_dma_mr, + .get_port_immutable = nes_port_immutable, + .map_mr_sg = nes_map_mr_sg, + .mmap = nes_mmap, + .modify_qp = nes_modify_qp, + .poll_cq = nes_poll_cq, + .post_recv = nes_post_recv, + .post_send = nes_post_send, + .query_device = nes_query_device, + .query_gid = nes_query_gid, + .query_pkey = nes_query_pkey, + .query_port = nes_query_port, + .query_qp = nes_query_qp, + .reg_user_mr = nes_reg_user_mr, + .req_notify_cq = nes_req_notify_cq, +}; + /** * nes_init_ofa_device */ @@ -3673,36 +3706,6 @@ struct nes_ib_device *nes_init_ofa_device(struct net_device *netdev) nesibdev->ibdev.phys_port_cnt = 1; nesibdev->ibdev.num_comp_vectors = 1; nesibdev->ibdev.dev.parent = &nesdev->pcidev->dev; - nesibdev->ibdev.query_device = nes_query_device; - nesibdev->ibdev.query_port = nes_query_port; - nesibdev->ibdev.query_pkey = nes_query_pkey; - nesibdev->ibdev.query_gid = nes_query_gid; - nesibdev->ibdev.alloc_ucontext = nes_alloc_ucontext; - nesibdev->ibdev.dealloc_ucontext = nes_dealloc_ucontext; - nesibdev->ibdev.mmap = nes_mmap; - nesibdev->ibdev.alloc_pd = nes_alloc_pd; - nesibdev->ibdev.dealloc_pd = nes_dealloc_pd; - nesibdev->ibdev.create_qp = nes_create_qp; - nesibdev->ibdev.modify_qp = nes_modify_qp; - nesibdev->ibdev.query_qp = nes_query_qp; - nesibdev->ibdev.destroy_qp = nes_destroy_qp; - nesibdev->ibdev.create_cq = nes_create_cq; - nesibdev->ibdev.destroy_cq = nes_destroy_cq; - nesibdev->ibdev.poll_cq = nes_poll_cq; - nesibdev->ibdev.get_dma_mr = nes_get_dma_mr; - nesibdev->ibdev.reg_user_mr = nes_reg_user_mr; - nesibdev->ibdev.dereg_mr = nes_dereg_mr; - nesibdev->ibdev.alloc_mw = nes_alloc_mw; - nesibdev->ibdev.dealloc_mw = nes_dealloc_mw; - - nesibdev->ibdev.alloc_mr = nes_alloc_mr; - nesibdev->ibdev.map_mr_sg = nes_map_mr_sg; - - nesibdev->ibdev.req_notify_cq = nes_req_notify_cq; - nesibdev->ibdev.post_send = nes_post_send; - nesibdev->ibdev.post_recv = nes_post_recv; - nesibdev->ibdev.drain_sq = nes_drain_sq; - nesibdev->ibdev.drain_rq = nes_drain_rq; nesibdev->ibdev.iwcm = kzalloc(sizeof(*nesibdev->ibdev.iwcm), GFP_KERNEL); if (nesibdev->ibdev.iwcm == NULL) { @@ -3717,8 +3720,8 @@ struct nes_ib_device *nes_init_ofa_device(struct net_device *netdev) nesibdev->ibdev.iwcm->reject = nes_reject; nesibdev->ibdev.iwcm->create_listen = nes_create_listen; nesibdev->ibdev.iwcm->destroy_listen = nes_destroy_listen; - nesibdev->ibdev.get_port_immutable = nes_port_immutable; - nesibdev->ibdev.get_dev_fw_str = get_dev_fw_str; + + ib_set_device_ops(&nesibdev->ibdev, &nes_dev_ops); memcpy(nesibdev->ibdev.iwcm->ifname, netdev->name, sizeof(nesibdev->ibdev.iwcm->ifname)); -- cgit v1.2.3-59-g8ed1b From a263c1241a18975756fcdbb6ccdb6460c89b04a5 Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Mon, 10 Dec 2018 21:09:41 +0200 Subject: RDMA/ocrdma: Initialize ib_device_ops struct Initialize ib_device_ops with the supported operations using ib_set_device_ops(). Signed-off-by: Kamal Heib Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/ocrdma/ocrdma_main.c | 92 +++++++++++++++--------------- 1 file changed, 46 insertions(+), 46 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_main.c b/drivers/infiniband/hw/ocrdma/ocrdma_main.c index 873cc7f6fe61..1f393842453a 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_main.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_main.c @@ -143,6 +143,50 @@ static const struct attribute_group ocrdma_attr_group = { .attrs = ocrdma_attributes, }; +static const struct ib_device_ops ocrdma_dev_ops = { + .alloc_mr = ocrdma_alloc_mr, + .alloc_pd = ocrdma_alloc_pd, + .alloc_ucontext = ocrdma_alloc_ucontext, + .create_ah = ocrdma_create_ah, + .create_cq = ocrdma_create_cq, + .create_qp = ocrdma_create_qp, + .dealloc_pd = ocrdma_dealloc_pd, + .dealloc_ucontext = ocrdma_dealloc_ucontext, + .dereg_mr = ocrdma_dereg_mr, + .destroy_ah = ocrdma_destroy_ah, + .destroy_cq = ocrdma_destroy_cq, + .destroy_qp = ocrdma_destroy_qp, + .get_dev_fw_str = get_dev_fw_str, + .get_dma_mr = ocrdma_get_dma_mr, + .get_link_layer = ocrdma_link_layer, + .get_netdev = ocrdma_get_netdev, + .get_port_immutable = ocrdma_port_immutable, + .map_mr_sg = ocrdma_map_mr_sg, + .mmap = ocrdma_mmap, + .modify_port = ocrdma_modify_port, + .modify_qp = ocrdma_modify_qp, + .poll_cq = ocrdma_poll_cq, + .post_recv = ocrdma_post_recv, + .post_send = ocrdma_post_send, + .process_mad = ocrdma_process_mad, + .query_ah = ocrdma_query_ah, + .query_device = ocrdma_query_device, + .query_pkey = ocrdma_query_pkey, + .query_port = ocrdma_query_port, + .query_qp = ocrdma_query_qp, + .reg_user_mr = ocrdma_reg_user_mr, + .req_notify_cq = ocrdma_arm_cq, + .resize_cq = ocrdma_resize_cq, +}; + +static const struct ib_device_ops ocrdma_dev_srq_ops = { + .create_srq = ocrdma_create_srq, + .destroy_srq = ocrdma_destroy_srq, + .modify_srq = ocrdma_modify_srq, + .post_srq_recv = ocrdma_post_srq_recv, + .query_srq = ocrdma_query_srq, +}; + static int ocrdma_register_device(struct ocrdma_dev *dev) { ocrdma_get_guid(dev, (u8 *)&dev->ibdev.node_guid); @@ -182,50 +226,10 @@ static int ocrdma_register_device(struct ocrdma_dev *dev) dev->ibdev.phys_port_cnt = 1; dev->ibdev.num_comp_vectors = dev->eq_cnt; - /* mandatory verbs. */ - dev->ibdev.query_device = ocrdma_query_device; - dev->ibdev.query_port = ocrdma_query_port; - dev->ibdev.modify_port = ocrdma_modify_port; - dev->ibdev.get_netdev = ocrdma_get_netdev; - dev->ibdev.get_link_layer = ocrdma_link_layer; - dev->ibdev.alloc_pd = ocrdma_alloc_pd; - dev->ibdev.dealloc_pd = ocrdma_dealloc_pd; - - dev->ibdev.create_cq = ocrdma_create_cq; - dev->ibdev.destroy_cq = ocrdma_destroy_cq; - dev->ibdev.resize_cq = ocrdma_resize_cq; - - dev->ibdev.create_qp = ocrdma_create_qp; - dev->ibdev.modify_qp = ocrdma_modify_qp; - dev->ibdev.query_qp = ocrdma_query_qp; - dev->ibdev.destroy_qp = ocrdma_destroy_qp; - - dev->ibdev.query_pkey = ocrdma_query_pkey; - dev->ibdev.create_ah = ocrdma_create_ah; - dev->ibdev.destroy_ah = ocrdma_destroy_ah; - dev->ibdev.query_ah = ocrdma_query_ah; - - dev->ibdev.poll_cq = ocrdma_poll_cq; - dev->ibdev.post_send = ocrdma_post_send; - dev->ibdev.post_recv = ocrdma_post_recv; - dev->ibdev.req_notify_cq = ocrdma_arm_cq; - - dev->ibdev.get_dma_mr = ocrdma_get_dma_mr; - dev->ibdev.dereg_mr = ocrdma_dereg_mr; - dev->ibdev.reg_user_mr = ocrdma_reg_user_mr; - - dev->ibdev.alloc_mr = ocrdma_alloc_mr; - dev->ibdev.map_mr_sg = ocrdma_map_mr_sg; - /* mandatory to support user space verbs consumer. */ - dev->ibdev.alloc_ucontext = ocrdma_alloc_ucontext; - dev->ibdev.dealloc_ucontext = ocrdma_dealloc_ucontext; - dev->ibdev.mmap = ocrdma_mmap; dev->ibdev.dev.parent = &dev->nic_info.pdev->dev; - dev->ibdev.process_mad = ocrdma_process_mad; - dev->ibdev.get_port_immutable = ocrdma_port_immutable; - dev->ibdev.get_dev_fw_str = get_dev_fw_str; + ib_set_device_ops(&dev->ibdev, &ocrdma_dev_ops); if (ocrdma_get_asic_type(dev) == OCRDMA_ASIC_GEN_SKH_R) { dev->ibdev.uverbs_cmd_mask |= @@ -235,11 +239,7 @@ static int ocrdma_register_device(struct ocrdma_dev *dev) OCRDMA_UVERBS(DESTROY_SRQ) | OCRDMA_UVERBS(POST_SRQ_RECV); - dev->ibdev.create_srq = ocrdma_create_srq; - dev->ibdev.modify_srq = ocrdma_modify_srq; - dev->ibdev.query_srq = ocrdma_query_srq; - dev->ibdev.destroy_srq = ocrdma_destroy_srq; - dev->ibdev.post_srq_recv = ocrdma_post_srq_recv; + ib_set_device_ops(&dev->ibdev, &ocrdma_dev_srq_ops); } rdma_set_device_sysfs_group(&dev->ibdev, &ocrdma_attr_group); dev->ibdev.driver_id = RDMA_DRIVER_OCRDMA; -- cgit v1.2.3-59-g8ed1b From bd59461e57b4163dbba8aaa131a6c2a61e1bdaab Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Mon, 10 Dec 2018 21:09:42 +0200 Subject: RDMA/qedr: Initialize ib_device_ops struct Initialize ib_device_ops with the supported operations using ib_set_device_ops(). Signed-off-by: Kamal Heib Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/qedr/main.c | 103 +++++++++++++++++++------------------- 1 file changed, 52 insertions(+), 51 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/qedr/main.c b/drivers/infiniband/hw/qedr/main.c index 8d6ff9df49fe..75940e2a8791 100644 --- a/drivers/infiniband/hw/qedr/main.c +++ b/drivers/infiniband/hw/qedr/main.c @@ -160,12 +160,16 @@ static const struct attribute_group qedr_attr_group = { .attrs = qedr_attributes, }; +static const struct ib_device_ops qedr_iw_dev_ops = { + .get_port_immutable = qedr_iw_port_immutable, + .query_gid = qedr_iw_query_gid, +}; + static int qedr_iw_register_device(struct qedr_dev *dev) { dev->ibdev.node_type = RDMA_NODE_RNIC; - dev->ibdev.query_gid = qedr_iw_query_gid; - dev->ibdev.get_port_immutable = qedr_iw_port_immutable; + ib_set_device_ops(&dev->ibdev, &qedr_iw_dev_ops); dev->ibdev.iwcm = kzalloc(sizeof(*dev->ibdev.iwcm), GFP_KERNEL); if (!dev->ibdev.iwcm) @@ -186,13 +190,56 @@ static int qedr_iw_register_device(struct qedr_dev *dev) return 0; } +static const struct ib_device_ops qedr_roce_dev_ops = { + .get_port_immutable = qedr_roce_port_immutable, +}; + static void qedr_roce_register_device(struct qedr_dev *dev) { dev->ibdev.node_type = RDMA_NODE_IB_CA; - dev->ibdev.get_port_immutable = qedr_roce_port_immutable; + ib_set_device_ops(&dev->ibdev, &qedr_roce_dev_ops); } +static const struct ib_device_ops qedr_dev_ops = { + .alloc_mr = qedr_alloc_mr, + .alloc_pd = qedr_alloc_pd, + .alloc_ucontext = qedr_alloc_ucontext, + .create_ah = qedr_create_ah, + .create_cq = qedr_create_cq, + .create_qp = qedr_create_qp, + .create_srq = qedr_create_srq, + .dealloc_pd = qedr_dealloc_pd, + .dealloc_ucontext = qedr_dealloc_ucontext, + .dereg_mr = qedr_dereg_mr, + .destroy_ah = qedr_destroy_ah, + .destroy_cq = qedr_destroy_cq, + .destroy_qp = qedr_destroy_qp, + .destroy_srq = qedr_destroy_srq, + .get_dev_fw_str = qedr_get_dev_fw_str, + .get_dma_mr = qedr_get_dma_mr, + .get_link_layer = qedr_link_layer, + .get_netdev = qedr_get_netdev, + .map_mr_sg = qedr_map_mr_sg, + .mmap = qedr_mmap, + .modify_port = qedr_modify_port, + .modify_qp = qedr_modify_qp, + .modify_srq = qedr_modify_srq, + .poll_cq = qedr_poll_cq, + .post_recv = qedr_post_recv, + .post_send = qedr_post_send, + .post_srq_recv = qedr_post_srq_recv, + .process_mad = qedr_process_mad, + .query_device = qedr_query_device, + .query_pkey = qedr_query_pkey, + .query_port = qedr_query_port, + .query_qp = qedr_query_qp, + .query_srq = qedr_query_srq, + .reg_user_mr = qedr_reg_user_mr, + .req_notify_cq = qedr_arm_cq, + .resize_cq = qedr_resize_cq, +}; + static int qedr_register_device(struct qedr_dev *dev) { int rc; @@ -237,57 +284,11 @@ static int qedr_register_device(struct qedr_dev *dev) dev->ibdev.phys_port_cnt = 1; dev->ibdev.num_comp_vectors = dev->num_cnq; - - dev->ibdev.query_device = qedr_query_device; - dev->ibdev.query_port = qedr_query_port; - dev->ibdev.modify_port = qedr_modify_port; - - dev->ibdev.alloc_ucontext = qedr_alloc_ucontext; - dev->ibdev.dealloc_ucontext = qedr_dealloc_ucontext; - dev->ibdev.mmap = qedr_mmap; - - dev->ibdev.alloc_pd = qedr_alloc_pd; - dev->ibdev.dealloc_pd = qedr_dealloc_pd; - - dev->ibdev.create_cq = qedr_create_cq; - dev->ibdev.destroy_cq = qedr_destroy_cq; - dev->ibdev.resize_cq = qedr_resize_cq; - dev->ibdev.req_notify_cq = qedr_arm_cq; - - dev->ibdev.create_qp = qedr_create_qp; - dev->ibdev.modify_qp = qedr_modify_qp; - dev->ibdev.query_qp = qedr_query_qp; - dev->ibdev.destroy_qp = qedr_destroy_qp; - - dev->ibdev.create_srq = qedr_create_srq; - dev->ibdev.destroy_srq = qedr_destroy_srq; - dev->ibdev.modify_srq = qedr_modify_srq; - dev->ibdev.query_srq = qedr_query_srq; - dev->ibdev.post_srq_recv = qedr_post_srq_recv; - dev->ibdev.query_pkey = qedr_query_pkey; - - dev->ibdev.create_ah = qedr_create_ah; - dev->ibdev.destroy_ah = qedr_destroy_ah; - - dev->ibdev.get_dma_mr = qedr_get_dma_mr; - dev->ibdev.dereg_mr = qedr_dereg_mr; - dev->ibdev.reg_user_mr = qedr_reg_user_mr; - dev->ibdev.alloc_mr = qedr_alloc_mr; - dev->ibdev.map_mr_sg = qedr_map_mr_sg; - - dev->ibdev.poll_cq = qedr_poll_cq; - dev->ibdev.post_send = qedr_post_send; - dev->ibdev.post_recv = qedr_post_recv; - - dev->ibdev.process_mad = qedr_process_mad; - - dev->ibdev.get_netdev = qedr_get_netdev; - dev->ibdev.dev.parent = &dev->pdev->dev; - dev->ibdev.get_link_layer = qedr_link_layer; - dev->ibdev.get_dev_fw_str = qedr_get_dev_fw_str; rdma_set_device_sysfs_group(&dev->ibdev, &qedr_attr_group); + ib_set_device_ops(&dev->ibdev, &qedr_dev_ops); + dev->ibdev.driver_id = RDMA_DRIVER_QEDR; return ib_register_device(&dev->ibdev, "qedr%d", NULL); } -- cgit v1.2.3-59-g8ed1b From 16b0ba9571f99a3beb9e13e7340b04f34608c93a Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Mon, 10 Dec 2018 21:09:43 +0200 Subject: RDMA/qib: Initialize ib_device_ops struct Initialize ib_device_ops with the supported operations using ib_set_device_ops(). Signed-off-by: Kamal Heib Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/qib/qib_verbs.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/qib/qib_verbs.c b/drivers/infiniband/hw/qib/qib_verbs.c index 8914abdd7584..611a6b5ef83f 100644 --- a/drivers/infiniband/hw/qib/qib_verbs.c +++ b/drivers/infiniband/hw/qib/qib_verbs.c @@ -1493,6 +1493,11 @@ static void qib_fill_device_attr(struct qib_devdata *dd) dd->verbs_dev.rdi.wc_opcode = ib_qib_wc_opcode; } +static const struct ib_device_ops qib_dev_ops = { + .modify_device = qib_modify_device, + .process_mad = qib_process_mad, +}; + /** * qib_register_ib_device - register our device with the infiniband core * @dd: the device data structure @@ -1555,8 +1560,6 @@ int qib_register_ib_device(struct qib_devdata *dd) ibdev->node_guid = ppd->guid; ibdev->phys_port_cnt = dd->num_pports; ibdev->dev.parent = &dd->pcidev->dev; - ibdev->modify_device = qib_modify_device; - ibdev->process_mad = qib_process_mad; snprintf(ibdev->node_desc, sizeof(ibdev->node_desc), "Intel Infiniband HCA %s", init_utsname()->nodename); @@ -1624,6 +1627,7 @@ int qib_register_ib_device(struct qib_devdata *dd) } rdma_set_device_sysfs_group(&dd->verbs_dev.rdi.ibdev, &qib_attr_group); + ib_set_device_ops(ibdev, &qib_dev_ops); ret = rvt_register_device(&dd->verbs_dev.rdi, RDMA_DRIVER_QIB); if (ret) goto err_tx; -- cgit v1.2.3-59-g8ed1b From e761058190b49e1d672bc5fef8cfefdec3b7d555 Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Mon, 10 Dec 2018 21:09:44 +0200 Subject: RDMA/usnic: Initialize ib_device_ops struct Initialize ib_device_ops with the supported operations using ib_set_device_ops(). Signed-off-by: Kamal Heib Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/usnic/usnic_ib_main.c | 61 +++++++++++++++-------------- 1 file changed, 32 insertions(+), 29 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/usnic/usnic_ib_main.c b/drivers/infiniband/hw/usnic/usnic_ib_main.c index 413fa5732e2b..b2323a52a0dd 100644 --- a/drivers/infiniband/hw/usnic/usnic_ib_main.c +++ b/drivers/infiniband/hw/usnic/usnic_ib_main.c @@ -330,6 +330,37 @@ static void usnic_get_dev_fw_str(struct ib_device *device, char *str) snprintf(str, IB_FW_VERSION_NAME_MAX, "%s", info.fw_version); } +static const struct ib_device_ops usnic_dev_ops = { + .alloc_pd = usnic_ib_alloc_pd, + .alloc_ucontext = usnic_ib_alloc_ucontext, + .create_ah = usnic_ib_create_ah, + .create_cq = usnic_ib_create_cq, + .create_qp = usnic_ib_create_qp, + .dealloc_pd = usnic_ib_dealloc_pd, + .dealloc_ucontext = usnic_ib_dealloc_ucontext, + .dereg_mr = usnic_ib_dereg_mr, + .destroy_ah = usnic_ib_destroy_ah, + .destroy_cq = usnic_ib_destroy_cq, + .destroy_qp = usnic_ib_destroy_qp, + .get_dev_fw_str = usnic_get_dev_fw_str, + .get_dma_mr = usnic_ib_get_dma_mr, + .get_link_layer = usnic_ib_port_link_layer, + .get_netdev = usnic_get_netdev, + .get_port_immutable = usnic_port_immutable, + .mmap = usnic_ib_mmap, + .modify_qp = usnic_ib_modify_qp, + .poll_cq = usnic_ib_poll_cq, + .post_recv = usnic_ib_post_recv, + .post_send = usnic_ib_post_send, + .query_device = usnic_ib_query_device, + .query_gid = usnic_ib_query_gid, + .query_pkey = usnic_ib_query_pkey, + .query_port = usnic_ib_query_port, + .query_qp = usnic_ib_query_qp, + .reg_user_mr = usnic_ib_reg_mr, + .req_notify_cq = usnic_ib_req_notify_cq, +}; + /* Start of PF discovery section */ static void *usnic_ib_device_add(struct pci_dev *dev) { @@ -386,35 +417,7 @@ static void *usnic_ib_device_add(struct pci_dev *dev) (1ull << IB_USER_VERBS_CMD_DETACH_MCAST) | (1ull << IB_USER_VERBS_CMD_OPEN_QP); - us_ibdev->ib_dev.query_device = usnic_ib_query_device; - us_ibdev->ib_dev.query_port = usnic_ib_query_port; - us_ibdev->ib_dev.query_pkey = usnic_ib_query_pkey; - us_ibdev->ib_dev.query_gid = usnic_ib_query_gid; - us_ibdev->ib_dev.get_netdev = usnic_get_netdev; - us_ibdev->ib_dev.get_link_layer = usnic_ib_port_link_layer; - us_ibdev->ib_dev.alloc_pd = usnic_ib_alloc_pd; - us_ibdev->ib_dev.dealloc_pd = usnic_ib_dealloc_pd; - us_ibdev->ib_dev.create_qp = usnic_ib_create_qp; - us_ibdev->ib_dev.modify_qp = usnic_ib_modify_qp; - us_ibdev->ib_dev.query_qp = usnic_ib_query_qp; - us_ibdev->ib_dev.destroy_qp = usnic_ib_destroy_qp; - us_ibdev->ib_dev.create_cq = usnic_ib_create_cq; - us_ibdev->ib_dev.destroy_cq = usnic_ib_destroy_cq; - us_ibdev->ib_dev.reg_user_mr = usnic_ib_reg_mr; - us_ibdev->ib_dev.dereg_mr = usnic_ib_dereg_mr; - us_ibdev->ib_dev.alloc_ucontext = usnic_ib_alloc_ucontext; - us_ibdev->ib_dev.dealloc_ucontext = usnic_ib_dealloc_ucontext; - us_ibdev->ib_dev.mmap = usnic_ib_mmap; - us_ibdev->ib_dev.create_ah = usnic_ib_create_ah; - us_ibdev->ib_dev.destroy_ah = usnic_ib_destroy_ah; - us_ibdev->ib_dev.post_send = usnic_ib_post_send; - us_ibdev->ib_dev.post_recv = usnic_ib_post_recv; - us_ibdev->ib_dev.poll_cq = usnic_ib_poll_cq; - us_ibdev->ib_dev.req_notify_cq = usnic_ib_req_notify_cq; - us_ibdev->ib_dev.get_dma_mr = usnic_ib_get_dma_mr; - us_ibdev->ib_dev.get_port_immutable = usnic_port_immutable; - us_ibdev->ib_dev.get_dev_fw_str = usnic_get_dev_fw_str; - + ib_set_device_ops(&us_ibdev->ib_dev, &usnic_dev_ops); us_ibdev->ib_dev.driver_id = RDMA_DRIVER_USNIC; rdma_set_device_sysfs_group(&us_ibdev->ib_dev, &usnic_attr_group); -- cgit v1.2.3-59-g8ed1b From 20a6b5886102bac96e7dc5f1c6ffb6543ad4b479 Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Mon, 10 Dec 2018 21:09:45 +0200 Subject: RDMA/vmw_pvrdma: Initialize ib_device_ops struct Initialize ib_device_ops with the supported operations using ib_set_device_ops(). Signed-off-by: Kamal Heib Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c | 82 ++++++++++++++------------ 1 file changed, 45 insertions(+), 37 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c index 398443f43dc3..eaa109dbc96a 100644 --- a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c +++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c @@ -161,6 +161,49 @@ static struct net_device *pvrdma_get_netdev(struct ib_device *ibdev, return netdev; } +static const struct ib_device_ops pvrdma_dev_ops = { + .add_gid = pvrdma_add_gid, + .alloc_mr = pvrdma_alloc_mr, + .alloc_pd = pvrdma_alloc_pd, + .alloc_ucontext = pvrdma_alloc_ucontext, + .create_ah = pvrdma_create_ah, + .create_cq = pvrdma_create_cq, + .create_qp = pvrdma_create_qp, + .dealloc_pd = pvrdma_dealloc_pd, + .dealloc_ucontext = pvrdma_dealloc_ucontext, + .del_gid = pvrdma_del_gid, + .dereg_mr = pvrdma_dereg_mr, + .destroy_ah = pvrdma_destroy_ah, + .destroy_cq = pvrdma_destroy_cq, + .destroy_qp = pvrdma_destroy_qp, + .get_dev_fw_str = pvrdma_get_fw_ver_str, + .get_dma_mr = pvrdma_get_dma_mr, + .get_link_layer = pvrdma_port_link_layer, + .get_netdev = pvrdma_get_netdev, + .get_port_immutable = pvrdma_port_immutable, + .map_mr_sg = pvrdma_map_mr_sg, + .mmap = pvrdma_mmap, + .modify_port = pvrdma_modify_port, + .modify_qp = pvrdma_modify_qp, + .poll_cq = pvrdma_poll_cq, + .post_recv = pvrdma_post_recv, + .post_send = pvrdma_post_send, + .query_device = pvrdma_query_device, + .query_gid = pvrdma_query_gid, + .query_pkey = pvrdma_query_pkey, + .query_port = pvrdma_query_port, + .query_qp = pvrdma_query_qp, + .reg_user_mr = pvrdma_reg_user_mr, + .req_notify_cq = pvrdma_req_notify_cq, +}; + +static const struct ib_device_ops pvrdma_dev_srq_ops = { + .create_srq = pvrdma_create_srq, + .destroy_srq = pvrdma_destroy_srq, + .modify_srq = pvrdma_modify_srq, + .query_srq = pvrdma_query_srq, +}; + static int pvrdma_register_device(struct pvrdma_dev *dev) { int ret = -1; @@ -197,39 +240,7 @@ static int pvrdma_register_device(struct pvrdma_dev *dev) dev->ib_dev.node_type = RDMA_NODE_IB_CA; dev->ib_dev.phys_port_cnt = dev->dsr->caps.phys_port_cnt; - dev->ib_dev.query_device = pvrdma_query_device; - dev->ib_dev.query_port = pvrdma_query_port; - dev->ib_dev.query_gid = pvrdma_query_gid; - dev->ib_dev.query_pkey = pvrdma_query_pkey; - dev->ib_dev.modify_port = pvrdma_modify_port; - dev->ib_dev.alloc_ucontext = pvrdma_alloc_ucontext; - dev->ib_dev.dealloc_ucontext = pvrdma_dealloc_ucontext; - dev->ib_dev.mmap = pvrdma_mmap; - dev->ib_dev.alloc_pd = pvrdma_alloc_pd; - dev->ib_dev.dealloc_pd = pvrdma_dealloc_pd; - dev->ib_dev.create_ah = pvrdma_create_ah; - dev->ib_dev.destroy_ah = pvrdma_destroy_ah; - dev->ib_dev.create_qp = pvrdma_create_qp; - dev->ib_dev.modify_qp = pvrdma_modify_qp; - dev->ib_dev.query_qp = pvrdma_query_qp; - dev->ib_dev.destroy_qp = pvrdma_destroy_qp; - dev->ib_dev.post_send = pvrdma_post_send; - dev->ib_dev.post_recv = pvrdma_post_recv; - dev->ib_dev.create_cq = pvrdma_create_cq; - dev->ib_dev.destroy_cq = pvrdma_destroy_cq; - dev->ib_dev.poll_cq = pvrdma_poll_cq; - dev->ib_dev.req_notify_cq = pvrdma_req_notify_cq; - dev->ib_dev.get_dma_mr = pvrdma_get_dma_mr; - dev->ib_dev.reg_user_mr = pvrdma_reg_user_mr; - dev->ib_dev.dereg_mr = pvrdma_dereg_mr; - dev->ib_dev.alloc_mr = pvrdma_alloc_mr; - dev->ib_dev.map_mr_sg = pvrdma_map_mr_sg; - dev->ib_dev.add_gid = pvrdma_add_gid; - dev->ib_dev.del_gid = pvrdma_del_gid; - dev->ib_dev.get_netdev = pvrdma_get_netdev; - dev->ib_dev.get_port_immutable = pvrdma_port_immutable; - dev->ib_dev.get_link_layer = pvrdma_port_link_layer; - dev->ib_dev.get_dev_fw_str = pvrdma_get_fw_ver_str; + ib_set_device_ops(&dev->ib_dev, &pvrdma_dev_ops); mutex_init(&dev->port_mutex); spin_lock_init(&dev->desc_lock); @@ -255,10 +266,7 @@ static int pvrdma_register_device(struct pvrdma_dev *dev) (1ull << IB_USER_VERBS_CMD_DESTROY_SRQ) | (1ull << IB_USER_VERBS_CMD_POST_SRQ_RECV); - dev->ib_dev.create_srq = pvrdma_create_srq; - dev->ib_dev.modify_srq = pvrdma_modify_srq; - dev->ib_dev.query_srq = pvrdma_query_srq; - dev->ib_dev.destroy_srq = pvrdma_destroy_srq; + ib_set_device_ops(&dev->ib_dev, &pvrdma_dev_srq_ops); dev->srq_tbl = kcalloc(dev->dsr->caps.max_srq, sizeof(struct pvrdma_srq *), -- cgit v1.2.3-59-g8ed1b From 3023a1e93656c02b8d6a3a46e712b815843fa514 Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Mon, 10 Dec 2018 21:09:48 +0200 Subject: RDMA: Start use ib_device_ops Make all the required change to start use the ib_device_ops structure. Signed-off-by: Kamal Heib Signed-off-by: Jason Gunthorpe --- drivers/infiniband/core/cache.c | 12 +- drivers/infiniband/core/core_priv.h | 12 +- drivers/infiniband/core/cq.c | 6 +- drivers/infiniband/core/device.c | 211 +++++++------- drivers/infiniband/core/fmr_pool.c | 4 +- drivers/infiniband/core/mad.c | 22 +- drivers/infiniband/core/nldev.c | 4 +- drivers/infiniband/core/opa_smi.h | 4 +- drivers/infiniband/core/rdma_core.c | 6 +- drivers/infiniband/core/security.c | 8 +- drivers/infiniband/core/smi.h | 4 +- drivers/infiniband/core/sysfs.c | 28 +- drivers/infiniband/core/ucm.c | 2 +- drivers/infiniband/core/uverbs_cmd.c | 60 ++-- drivers/infiniband/core/uverbs_main.c | 14 +- drivers/infiniband/core/uverbs_std_types.c | 2 +- .../infiniband/core/uverbs_std_types_counters.c | 10 +- drivers/infiniband/core/uverbs_std_types_cq.c | 6 +- drivers/infiniband/core/uverbs_std_types_dm.c | 6 +- .../infiniband/core/uverbs_std_types_flow_action.c | 14 +- drivers/infiniband/core/uverbs_std_types_mr.c | 4 +- drivers/infiniband/core/uverbs_uapi.c | 3 +- drivers/infiniband/core/verbs.c | 159 +++++------ drivers/infiniband/hw/i40iw/i40iw_cm.c | 2 +- drivers/infiniband/hw/mlx4/alias_GUID.c | 2 +- drivers/infiniband/hw/mlx5/main.c | 2 +- drivers/infiniband/hw/nes/nes_cm.c | 2 +- drivers/infiniband/sw/rdmavt/vt.c | 16 +- drivers/infiniband/ulp/ipoib/ipoib_main.c | 4 +- drivers/infiniband/ulp/iser/iser_memory.c | 4 +- drivers/infiniband/ulp/opa_vnic/opa_vnic_netdev.c | 8 +- drivers/infiniband/ulp/srp/ib_srp.c | 6 +- fs/cifs/smbdirect.c | 2 +- include/rdma/ib_verbs.h | 303 +-------------------- include/rdma/uverbs_ioctl.h | 12 +- net/rds/ib.c | 4 +- net/sunrpc/xprtrdma/fmr_ops.c | 2 +- 37 files changed, 353 insertions(+), 617 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/core/cache.c b/drivers/infiniband/core/cache.c index 5b2fce4a7091..22e20ed5a393 100644 --- a/drivers/infiniband/core/cache.c +++ b/drivers/infiniband/core/cache.c @@ -217,7 +217,7 @@ static void free_gid_entry_locked(struct ib_gid_table_entry *entry) if (rdma_cap_roce_gid_table(device, port_num) && entry->state != GID_TABLE_ENTRY_INVALID) - device->del_gid(&entry->attr, &entry->context); + device->ops.del_gid(&entry->attr, &entry->context); write_lock_irq(&table->rwlock); @@ -324,7 +324,7 @@ static int add_roce_gid(struct ib_gid_table_entry *entry) return -EINVAL; } if (rdma_cap_roce_gid_table(attr->device, attr->port_num)) { - ret = attr->device->add_gid(attr, &entry->context); + ret = attr->device->ops.add_gid(attr, &entry->context); if (ret) { dev_err(&attr->device->dev, "%s GID add failed port=%d index=%d\n", @@ -548,8 +548,8 @@ int ib_cache_gid_add(struct ib_device *ib_dev, u8 port, unsigned long mask; int ret; - if (ib_dev->get_netdev) { - idev = ib_dev->get_netdev(ib_dev, port); + if (ib_dev->ops.get_netdev) { + idev = ib_dev->ops.get_netdev(ib_dev, port); if (idev && attr->ndev != idev) { union ib_gid default_gid; @@ -1296,9 +1296,9 @@ static int config_non_roce_gid_cache(struct ib_device *device, mutex_lock(&table->lock); for (i = 0; i < gid_tbl_len; ++i) { - if (!device->query_gid) + if (!device->ops.query_gid) continue; - ret = device->query_gid(device, port, i, &gid_attr.gid); + ret = device->ops.query_gid(device, port, i, &gid_attr.gid); if (ret) { dev_warn(&device->dev, "query_gid failed (%d) for index %d\n", ret, diff --git a/drivers/infiniband/core/core_priv.h b/drivers/infiniband/core/core_priv.h index cc7535c5e192..cea92624f9d4 100644 --- a/drivers/infiniband/core/core_priv.h +++ b/drivers/infiniband/core/core_priv.h @@ -215,10 +215,10 @@ static inline int ib_security_modify_qp(struct ib_qp *qp, int qp_attr_mask, struct ib_udata *udata) { - return qp->device->modify_qp(qp->real_qp, - qp_attr, - qp_attr_mask, - udata); + return qp->device->ops.modify_qp(qp->real_qp, + qp_attr, + qp_attr_mask, + udata); } static inline int ib_create_qp_security(struct ib_qp *qp, @@ -280,10 +280,10 @@ static inline struct ib_qp *_ib_create_qp(struct ib_device *dev, { struct ib_qp *qp; - if (!dev->create_qp) + if (!dev->ops.create_qp) return ERR_PTR(-EOPNOTSUPP); - qp = dev->create_qp(pd, attr, udata); + qp = dev->ops.create_qp(pd, attr, udata); if (IS_ERR(qp)) return qp; diff --git a/drivers/infiniband/core/cq.c b/drivers/infiniband/core/cq.c index b1e5365ddafa..7fb4f64ae933 100644 --- a/drivers/infiniband/core/cq.c +++ b/drivers/infiniband/core/cq.c @@ -145,7 +145,7 @@ struct ib_cq *__ib_alloc_cq(struct ib_device *dev, void *private, struct ib_cq *cq; int ret = -ENOMEM; - cq = dev->create_cq(dev, &cq_attr, NULL, NULL); + cq = dev->ops.create_cq(dev, &cq_attr, NULL, NULL); if (IS_ERR(cq)) return cq; @@ -193,7 +193,7 @@ out_free_wc: kfree(cq->wc); rdma_restrack_del(&cq->res); out_destroy_cq: - cq->device->destroy_cq(cq); + cq->device->ops.destroy_cq(cq); return ERR_PTR(ret); } EXPORT_SYMBOL(__ib_alloc_cq); @@ -225,7 +225,7 @@ void ib_free_cq(struct ib_cq *cq) kfree(cq->wc); rdma_restrack_del(&cq->res); - ret = cq->device->destroy_cq(cq); + ret = cq->device->ops.destroy_cq(cq); WARN_ON_ONCE(ret); } EXPORT_SYMBOL(ib_free_cq); diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c index 108c62d875af..47ab34ee1a9d 100644 --- a/drivers/infiniband/core/device.c +++ b/drivers/infiniband/core/device.c @@ -96,7 +96,7 @@ static struct notifier_block ibdev_lsm_nb = { static int ib_device_check_mandatory(struct ib_device *device) { -#define IB_MANDATORY_FUNC(x) { offsetof(struct ib_device, x), #x } +#define IB_MANDATORY_FUNC(x) { offsetof(struct ib_device_ops, x), #x } static const struct { size_t offset; char *name; @@ -122,7 +122,8 @@ static int ib_device_check_mandatory(struct ib_device *device) int i; for (i = 0; i < ARRAY_SIZE(mandatory_table); ++i) { - if (!*(void **) ((void *) device + mandatory_table[i].offset)) { + if (!*(void **) ((void *) &device->ops + + mandatory_table[i].offset)) { dev_warn(&device->dev, "Device is missing mandatory function %s\n", mandatory_table[i].name); @@ -373,8 +374,8 @@ static int read_port_immutable(struct ib_device *device) return -ENOMEM; for (port = start_port; port <= end_port; ++port) { - ret = device->get_port_immutable(device, port, - &device->port_immutable[port]); + ret = device->ops.get_port_immutable( + device, port, &device->port_immutable[port]); if (ret) return ret; @@ -386,8 +387,8 @@ static int read_port_immutable(struct ib_device *device) void ib_get_device_fw_str(struct ib_device *dev, char *str) { - if (dev->get_dev_fw_str) - dev->get_dev_fw_str(dev, str); + if (dev->ops.get_dev_fw_str) + dev->ops.get_dev_fw_str(dev, str); else str[0] = '\0'; } @@ -536,7 +537,7 @@ static int setup_device(struct ib_device *device) } memset(&device->attrs, 0, sizeof(device->attrs)); - ret = device->query_device(device, &device->attrs, &uhw); + ret = device->ops.query_device(device, &device->attrs, &uhw); if (ret) { dev_warn(&device->dev, "Couldn't query the device attributes\n"); @@ -923,14 +924,14 @@ int ib_query_port(struct ib_device *device, return -EINVAL; memset(port_attr, 0, sizeof(*port_attr)); - err = device->query_port(device, port_num, port_attr); + err = device->ops.query_port(device, port_num, port_attr); if (err || port_attr->subnet_prefix) return err; if (rdma_port_get_link_layer(device, port_num) != IB_LINK_LAYER_INFINIBAND) return 0; - err = device->query_gid(device, port_num, 0, &gid); + err = device->ops.query_gid(device, port_num, 0, &gid); if (err) return err; @@ -964,8 +965,8 @@ void ib_enum_roce_netdev(struct ib_device *ib_dev, if (rdma_protocol_roce(ib_dev, port)) { struct net_device *idev = NULL; - if (ib_dev->get_netdev) - idev = ib_dev->get_netdev(ib_dev, port); + if (ib_dev->ops.get_netdev) + idev = ib_dev->ops.get_netdev(ib_dev, port); if (idev && idev->reg_state >= NETREG_UNREGISTERED) { @@ -1045,7 +1046,7 @@ int ib_query_pkey(struct ib_device *device, if (!rdma_is_port_valid(device, port_num)) return -EINVAL; - return device->query_pkey(device, port_num, index, pkey); + return device->ops.query_pkey(device, port_num, index, pkey); } EXPORT_SYMBOL(ib_query_pkey); @@ -1062,11 +1063,11 @@ int ib_modify_device(struct ib_device *device, int device_modify_mask, struct ib_device_modify *device_modify) { - if (!device->modify_device) + if (!device->ops.modify_device) return -ENOSYS; - return device->modify_device(device, device_modify_mask, - device_modify); + return device->ops.modify_device(device, device_modify_mask, + device_modify); } EXPORT_SYMBOL(ib_modify_device); @@ -1090,9 +1091,10 @@ int ib_modify_port(struct ib_device *device, if (!rdma_is_port_valid(device, port_num)) return -EINVAL; - if (device->modify_port) - rc = device->modify_port(device, port_num, port_modify_mask, - port_modify); + if (device->ops.modify_port) + rc = device->ops.modify_port(device, port_num, + port_modify_mask, + port_modify); else rc = rdma_protocol_roce(device, port_num) ? 0 : -ENOSYS; return rc; @@ -1221,6 +1223,7 @@ EXPORT_SYMBOL(ib_get_net_dev_by_params); void ib_set_device_ops(struct ib_device *dev, const struct ib_device_ops *ops) { + struct ib_device_ops *dev_ops = &dev->ops; #define SET_DEVICE_OP(ptr, name) \ do { \ if (ops->name) \ @@ -1228,92 +1231,92 @@ void ib_set_device_ops(struct ib_device *dev, const struct ib_device_ops *ops) (ptr)->name = ops->name; \ } while (0) - SET_DEVICE_OP(dev, add_gid); - SET_DEVICE_OP(dev, alloc_dm); - SET_DEVICE_OP(dev, alloc_fmr); - SET_DEVICE_OP(dev, alloc_hw_stats); - SET_DEVICE_OP(dev, alloc_mr); - SET_DEVICE_OP(dev, alloc_mw); - SET_DEVICE_OP(dev, alloc_pd); - SET_DEVICE_OP(dev, alloc_rdma_netdev); - SET_DEVICE_OP(dev, alloc_ucontext); - SET_DEVICE_OP(dev, alloc_xrcd); - SET_DEVICE_OP(dev, attach_mcast); - SET_DEVICE_OP(dev, check_mr_status); - SET_DEVICE_OP(dev, create_ah); - SET_DEVICE_OP(dev, create_counters); - SET_DEVICE_OP(dev, create_cq); - SET_DEVICE_OP(dev, create_flow); - SET_DEVICE_OP(dev, create_flow_action_esp); - SET_DEVICE_OP(dev, create_qp); - SET_DEVICE_OP(dev, create_rwq_ind_table); - SET_DEVICE_OP(dev, create_srq); - SET_DEVICE_OP(dev, create_wq); - SET_DEVICE_OP(dev, dealloc_dm); - SET_DEVICE_OP(dev, dealloc_fmr); - SET_DEVICE_OP(dev, dealloc_mw); - SET_DEVICE_OP(dev, dealloc_pd); - SET_DEVICE_OP(dev, dealloc_ucontext); - SET_DEVICE_OP(dev, dealloc_xrcd); - SET_DEVICE_OP(dev, del_gid); - SET_DEVICE_OP(dev, dereg_mr); - SET_DEVICE_OP(dev, destroy_ah); - SET_DEVICE_OP(dev, destroy_counters); - SET_DEVICE_OP(dev, destroy_cq); - SET_DEVICE_OP(dev, destroy_flow); - SET_DEVICE_OP(dev, destroy_flow_action); - SET_DEVICE_OP(dev, destroy_qp); - SET_DEVICE_OP(dev, destroy_rwq_ind_table); - SET_DEVICE_OP(dev, destroy_srq); - SET_DEVICE_OP(dev, destroy_wq); - SET_DEVICE_OP(dev, detach_mcast); - SET_DEVICE_OP(dev, disassociate_ucontext); - SET_DEVICE_OP(dev, drain_rq); - SET_DEVICE_OP(dev, drain_sq); - SET_DEVICE_OP(dev, get_dev_fw_str); - SET_DEVICE_OP(dev, get_dma_mr); - SET_DEVICE_OP(dev, get_hw_stats); - SET_DEVICE_OP(dev, get_link_layer); - SET_DEVICE_OP(dev, get_netdev); - SET_DEVICE_OP(dev, get_port_immutable); - SET_DEVICE_OP(dev, get_vector_affinity); - SET_DEVICE_OP(dev, get_vf_config); - SET_DEVICE_OP(dev, get_vf_stats); - SET_DEVICE_OP(dev, map_mr_sg); - SET_DEVICE_OP(dev, map_phys_fmr); - SET_DEVICE_OP(dev, mmap); - SET_DEVICE_OP(dev, modify_ah); - SET_DEVICE_OP(dev, modify_cq); - SET_DEVICE_OP(dev, modify_device); - SET_DEVICE_OP(dev, modify_flow_action_esp); - SET_DEVICE_OP(dev, modify_port); - SET_DEVICE_OP(dev, modify_qp); - SET_DEVICE_OP(dev, modify_srq); - SET_DEVICE_OP(dev, modify_wq); - SET_DEVICE_OP(dev, peek_cq); - SET_DEVICE_OP(dev, poll_cq); - SET_DEVICE_OP(dev, post_recv); - SET_DEVICE_OP(dev, post_send); - SET_DEVICE_OP(dev, post_srq_recv); - SET_DEVICE_OP(dev, process_mad); - SET_DEVICE_OP(dev, query_ah); - SET_DEVICE_OP(dev, query_device); - SET_DEVICE_OP(dev, query_gid); - SET_DEVICE_OP(dev, query_pkey); - SET_DEVICE_OP(dev, query_port); - SET_DEVICE_OP(dev, query_qp); - SET_DEVICE_OP(dev, query_srq); - SET_DEVICE_OP(dev, rdma_netdev_get_params); - SET_DEVICE_OP(dev, read_counters); - SET_DEVICE_OP(dev, reg_dm_mr); - SET_DEVICE_OP(dev, reg_user_mr); - SET_DEVICE_OP(dev, req_ncomp_notif); - SET_DEVICE_OP(dev, req_notify_cq); - SET_DEVICE_OP(dev, rereg_user_mr); - SET_DEVICE_OP(dev, resize_cq); - SET_DEVICE_OP(dev, set_vf_guid); - SET_DEVICE_OP(dev, set_vf_link_state); - SET_DEVICE_OP(dev, unmap_fmr); + SET_DEVICE_OP(dev_ops, add_gid); + SET_DEVICE_OP(dev_ops, alloc_dm); + SET_DEVICE_OP(dev_ops, alloc_fmr); + SET_DEVICE_OP(dev_ops, alloc_hw_stats); + SET_DEVICE_OP(dev_ops, alloc_mr); + SET_DEVICE_OP(dev_ops, alloc_mw); + SET_DEVICE_OP(dev_ops, alloc_pd); + SET_DEVICE_OP(dev_ops, alloc_rdma_netdev); + SET_DEVICE_OP(dev_ops, alloc_ucontext); + SET_DEVICE_OP(dev_ops, alloc_xrcd); + SET_DEVICE_OP(dev_ops, attach_mcast); + SET_DEVICE_OP(dev_ops, check_mr_status); + SET_DEVICE_OP(dev_ops, create_ah); + SET_DEVICE_OP(dev_ops, create_counters); + SET_DEVICE_OP(dev_ops, create_cq); + SET_DEVICE_OP(dev_ops, create_flow); + SET_DEVICE_OP(dev_ops, create_flow_action_esp); + SET_DEVICE_OP(dev_ops, create_qp); + SET_DEVICE_OP(dev_ops, create_rwq_ind_table); + SET_DEVICE_OP(dev_ops, create_srq); + SET_DEVICE_OP(dev_ops, create_wq); + SET_DEVICE_OP(dev_ops, dealloc_dm); + SET_DEVICE_OP(dev_ops, dealloc_fmr); + SET_DEVICE_OP(dev_ops, dealloc_mw); + SET_DEVICE_OP(dev_ops, dealloc_pd); + SET_DEVICE_OP(dev_ops, dealloc_ucontext); + SET_DEVICE_OP(dev_ops, dealloc_xrcd); + SET_DEVICE_OP(dev_ops, del_gid); + SET_DEVICE_OP(dev_ops, dereg_mr); + SET_DEVICE_OP(dev_ops, destroy_ah); + SET_DEVICE_OP(dev_ops, destroy_counters); + SET_DEVICE_OP(dev_ops, destroy_cq); + SET_DEVICE_OP(dev_ops, destroy_flow); + SET_DEVICE_OP(dev_ops, destroy_flow_action); + SET_DEVICE_OP(dev_ops, destroy_qp); + SET_DEVICE_OP(dev_ops, destroy_rwq_ind_table); + SET_DEVICE_OP(dev_ops, destroy_srq); + SET_DEVICE_OP(dev_ops, destroy_wq); + SET_DEVICE_OP(dev_ops, detach_mcast); + SET_DEVICE_OP(dev_ops, disassociate_ucontext); + SET_DEVICE_OP(dev_ops, drain_rq); + SET_DEVICE_OP(dev_ops, drain_sq); + SET_DEVICE_OP(dev_ops, get_dev_fw_str); + SET_DEVICE_OP(dev_ops, get_dma_mr); + SET_DEVICE_OP(dev_ops, get_hw_stats); + SET_DEVICE_OP(dev_ops, get_link_layer); + SET_DEVICE_OP(dev_ops, get_netdev); + SET_DEVICE_OP(dev_ops, get_port_immutable); + SET_DEVICE_OP(dev_ops, get_vector_affinity); + SET_DEVICE_OP(dev_ops, get_vf_config); + SET_DEVICE_OP(dev_ops, get_vf_stats); + SET_DEVICE_OP(dev_ops, map_mr_sg); + SET_DEVICE_OP(dev_ops, map_phys_fmr); + SET_DEVICE_OP(dev_ops, mmap); + SET_DEVICE_OP(dev_ops, modify_ah); + SET_DEVICE_OP(dev_ops, modify_cq); + SET_DEVICE_OP(dev_ops, modify_device); + SET_DEVICE_OP(dev_ops, modify_flow_action_esp); + SET_DEVICE_OP(dev_ops, modify_port); + SET_DEVICE_OP(dev_ops, modify_qp); + SET_DEVICE_OP(dev_ops, modify_srq); + SET_DEVICE_OP(dev_ops, modify_wq); + SET_DEVICE_OP(dev_ops, peek_cq); + SET_DEVICE_OP(dev_ops, poll_cq); + SET_DEVICE_OP(dev_ops, post_recv); + SET_DEVICE_OP(dev_ops, post_send); + SET_DEVICE_OP(dev_ops, post_srq_recv); + SET_DEVICE_OP(dev_ops, process_mad); + SET_DEVICE_OP(dev_ops, query_ah); + SET_DEVICE_OP(dev_ops, query_device); + SET_DEVICE_OP(dev_ops, query_gid); + SET_DEVICE_OP(dev_ops, query_pkey); + SET_DEVICE_OP(dev_ops, query_port); + SET_DEVICE_OP(dev_ops, query_qp); + SET_DEVICE_OP(dev_ops, query_srq); + SET_DEVICE_OP(dev_ops, rdma_netdev_get_params); + SET_DEVICE_OP(dev_ops, read_counters); + SET_DEVICE_OP(dev_ops, reg_dm_mr); + SET_DEVICE_OP(dev_ops, reg_user_mr); + SET_DEVICE_OP(dev_ops, req_ncomp_notif); + SET_DEVICE_OP(dev_ops, req_notify_cq); + SET_DEVICE_OP(dev_ops, rereg_user_mr); + SET_DEVICE_OP(dev_ops, resize_cq); + SET_DEVICE_OP(dev_ops, set_vf_guid); + SET_DEVICE_OP(dev_ops, set_vf_link_state); + SET_DEVICE_OP(dev_ops, unmap_fmr); } EXPORT_SYMBOL(ib_set_device_ops); diff --git a/drivers/infiniband/core/fmr_pool.c b/drivers/infiniband/core/fmr_pool.c index b00dfd2ad31e..7d841b689a1e 100644 --- a/drivers/infiniband/core/fmr_pool.c +++ b/drivers/infiniband/core/fmr_pool.c @@ -211,8 +211,8 @@ struct ib_fmr_pool *ib_create_fmr_pool(struct ib_pd *pd, return ERR_PTR(-EINVAL); device = pd->device; - if (!device->alloc_fmr || !device->dealloc_fmr || - !device->map_phys_fmr || !device->unmap_fmr) { + if (!device->ops.alloc_fmr || !device->ops.dealloc_fmr || + !device->ops.map_phys_fmr || !device->ops.unmap_fmr) { dev_info(&device->dev, "Device does not support FMRs\n"); return ERR_PTR(-ENOSYS); } diff --git a/drivers/infiniband/core/mad.c b/drivers/infiniband/core/mad.c index d7025cd5be28..7870823bac47 100644 --- a/drivers/infiniband/core/mad.c +++ b/drivers/infiniband/core/mad.c @@ -888,10 +888,10 @@ static int handle_outgoing_dr_smp(struct ib_mad_agent_private *mad_agent_priv, } /* No GRH for DR SMP */ - ret = device->process_mad(device, 0, port_num, &mad_wc, NULL, - (const struct ib_mad_hdr *)smp, mad_size, - (struct ib_mad_hdr *)mad_priv->mad, - &mad_size, &out_mad_pkey_index); + ret = device->ops.process_mad(device, 0, port_num, &mad_wc, NULL, + (const struct ib_mad_hdr *)smp, mad_size, + (struct ib_mad_hdr *)mad_priv->mad, + &mad_size, &out_mad_pkey_index); switch (ret) { case IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_REPLY: @@ -2305,14 +2305,12 @@ static void ib_mad_recv_done(struct ib_cq *cq, struct ib_wc *wc) } /* Give driver "right of first refusal" on incoming MAD */ - if (port_priv->device->process_mad) { - ret = port_priv->device->process_mad(port_priv->device, 0, - port_priv->port_num, - wc, &recv->grh, - (const struct ib_mad_hdr *)recv->mad, - recv->mad_size, - (struct ib_mad_hdr *)response->mad, - &mad_size, &resp_mad_pkey_index); + if (port_priv->device->ops.process_mad) { + ret = port_priv->device->ops.process_mad( + port_priv->device, 0, port_priv->port_num, wc, + &recv->grh, (const struct ib_mad_hdr *)recv->mad, + recv->mad_size, (struct ib_mad_hdr *)response->mad, + &mad_size, &resp_mad_pkey_index); if (opa) wc->pkey_index = resp_mad_pkey_index; diff --git a/drivers/infiniband/core/nldev.c b/drivers/infiniband/core/nldev.c index 9abbadb9e366..093bbfcdc53b 100644 --- a/drivers/infiniband/core/nldev.c +++ b/drivers/infiniband/core/nldev.c @@ -259,8 +259,8 @@ static int fill_port_info(struct sk_buff *msg, if (nla_put_u8(msg, RDMA_NLDEV_ATTR_PORT_PHYS_STATE, attr.phys_state)) return -EMSGSIZE; - if (device->get_netdev) - netdev = device->get_netdev(device, port); + if (device->ops.get_netdev) + netdev = device->ops.get_netdev(device, port); if (netdev && net_eq(dev_net(netdev), net)) { ret = nla_put_u32(msg, diff --git a/drivers/infiniband/core/opa_smi.h b/drivers/infiniband/core/opa_smi.h index 3bfab3505a29..af4879bdf3d6 100644 --- a/drivers/infiniband/core/opa_smi.h +++ b/drivers/infiniband/core/opa_smi.h @@ -55,7 +55,7 @@ static inline enum smi_action opa_smi_check_local_smp(struct opa_smp *smp, { /* C14-9:3 -- We're at the end of the DR segment of path */ /* C14-9:4 -- Hop Pointer = Hop Count + 1 -> give to SMA/SM */ - return (device->process_mad && + return (device->ops.process_mad && !opa_get_smp_direction(smp) && (smp->hop_ptr == smp->hop_cnt + 1)) ? IB_SMI_HANDLE : IB_SMI_DISCARD; @@ -70,7 +70,7 @@ static inline enum smi_action opa_smi_check_local_returning_smp(struct opa_smp * { /* C14-13:3 -- We're at the end of the DR segment of path */ /* C14-13:4 -- Hop Pointer == 0 -> give to SM */ - return (device->process_mad && + return (device->ops.process_mad && opa_get_smp_direction(smp) && !smp->hop_ptr) ? IB_SMI_HANDLE : IB_SMI_DISCARD; } diff --git a/drivers/infiniband/core/rdma_core.c b/drivers/infiniband/core/rdma_core.c index 7d2f1ef75025..6c4747e61d2b 100644 --- a/drivers/infiniband/core/rdma_core.c +++ b/drivers/infiniband/core/rdma_core.c @@ -820,8 +820,8 @@ static void ufile_destroy_ucontext(struct ib_uverbs_file *ufile, */ if (reason == RDMA_REMOVE_DRIVER_REMOVE) { uverbs_user_mmap_disassociate(ufile); - if (ib_dev->disassociate_ucontext) - ib_dev->disassociate_ucontext(ucontext); + if (ib_dev->ops.disassociate_ucontext) + ib_dev->ops.disassociate_ucontext(ucontext); } ib_rdmacg_uncharge(&ucontext->cg_obj, ib_dev, @@ -833,7 +833,7 @@ static void ufile_destroy_ucontext(struct ib_uverbs_file *ufile, * FIXME: Drivers are not permitted to fail dealloc_ucontext, remove * the error return. */ - ret = ib_dev->dealloc_ucontext(ucontext); + ret = ib_dev->ops.dealloc_ucontext(ucontext); WARN_ON(ret); ufile->ucontext = NULL; diff --git a/drivers/infiniband/core/security.c b/drivers/infiniband/core/security.c index 1143c0448666..1efadbccf394 100644 --- a/drivers/infiniband/core/security.c +++ b/drivers/infiniband/core/security.c @@ -626,10 +626,10 @@ int ib_security_modify_qp(struct ib_qp *qp, } if (!ret) - ret = real_qp->device->modify_qp(real_qp, - qp_attr, - qp_attr_mask, - udata); + ret = real_qp->device->ops.modify_qp(real_qp, + qp_attr, + qp_attr_mask, + udata); if (new_pps) { /* Clean up the lists and free the appropriate diff --git a/drivers/infiniband/core/smi.h b/drivers/infiniband/core/smi.h index 33c91c8a16e9..91d9b353ab85 100644 --- a/drivers/infiniband/core/smi.h +++ b/drivers/infiniband/core/smi.h @@ -67,7 +67,7 @@ static inline enum smi_action smi_check_local_smp(struct ib_smp *smp, { /* C14-9:3 -- We're at the end of the DR segment of path */ /* C14-9:4 -- Hop Pointer = Hop Count + 1 -> give to SMA/SM */ - return ((device->process_mad && + return ((device->ops.process_mad && !ib_get_smp_direction(smp) && (smp->hop_ptr == smp->hop_cnt + 1)) ? IB_SMI_HANDLE : IB_SMI_DISCARD); @@ -82,7 +82,7 @@ static inline enum smi_action smi_check_local_returning_smp(struct ib_smp *smp, { /* C14-13:3 -- We're at the end of the DR segment of path */ /* C14-13:4 -- Hop Pointer == 0 -> give to SM */ - return ((device->process_mad && + return ((device->ops.process_mad && ib_get_smp_direction(smp) && !smp->hop_ptr) ? IB_SMI_HANDLE : IB_SMI_DISCARD); } diff --git a/drivers/infiniband/core/sysfs.c b/drivers/infiniband/core/sysfs.c index 6fcce2c206c6..80f68eb0ba5c 100644 --- a/drivers/infiniband/core/sysfs.c +++ b/drivers/infiniband/core/sysfs.c @@ -462,7 +462,7 @@ static int get_perf_mad(struct ib_device *dev, int port_num, __be16 attr, u16 out_mad_pkey_index = 0; ssize_t ret; - if (!dev->process_mad) + if (!dev->ops.process_mad) return -ENOSYS; in_mad = kzalloc(sizeof *in_mad, GFP_KERNEL); @@ -481,11 +481,11 @@ static int get_perf_mad(struct ib_device *dev, int port_num, __be16 attr, if (attr != IB_PMA_CLASS_PORT_INFO) in_mad->data[41] = port_num; /* PortSelect field */ - if ((dev->process_mad(dev, IB_MAD_IGNORE_MKEY, - port_num, NULL, NULL, - (const struct ib_mad_hdr *)in_mad, mad_size, - (struct ib_mad_hdr *)out_mad, &mad_size, - &out_mad_pkey_index) & + if ((dev->ops.process_mad(dev, IB_MAD_IGNORE_MKEY, + port_num, NULL, NULL, + (const struct ib_mad_hdr *)in_mad, mad_size, + (struct ib_mad_hdr *)out_mad, &mad_size, + &out_mad_pkey_index) & (IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_REPLY)) != (IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_REPLY)) { ret = -EINVAL; @@ -786,7 +786,7 @@ static int update_hw_stats(struct ib_device *dev, struct rdma_hw_stats *stats, if (time_is_after_eq_jiffies(stats->timestamp + stats->lifespan)) return 0; - ret = dev->get_hw_stats(dev, stats, port_num, index); + ret = dev->ops.get_hw_stats(dev, stats, port_num, index); if (ret < 0) return ret; if (ret == stats->num_counters) @@ -946,7 +946,7 @@ static void setup_hw_stats(struct ib_device *device, struct ib_port *port, struct rdma_hw_stats *stats; int i, ret; - stats = device->alloc_hw_stats(device, port_num); + stats = device->ops.alloc_hw_stats(device, port_num); if (!stats) return; @@ -964,8 +964,8 @@ static void setup_hw_stats(struct ib_device *device, struct ib_port *port, if (!hsag) goto err_free_stats; - ret = device->get_hw_stats(device, stats, port_num, - stats->num_counters); + ret = device->ops.get_hw_stats(device, stats, port_num, + stats->num_counters); if (ret != stats->num_counters) goto err_free_hsag; @@ -1057,7 +1057,7 @@ static int add_port(struct ib_device *device, int port_num, goto err_put; } - if (device->process_mad) { + if (device->ops.process_mad) { p->pma_table = get_counter_table(device, port_num); ret = sysfs_create_group(&p->kobj, p->pma_table); if (ret) @@ -1124,7 +1124,7 @@ static int add_port(struct ib_device *device, int port_num, * port, so holder should be device. Therefore skip per port conunter * initialization. */ - if (device->alloc_hw_stats && port_num) + if (device->ops.alloc_hw_stats && port_num) setup_hw_stats(device, p, port_num); list_add_tail(&p->kobj.entry, &device->port_list); @@ -1245,7 +1245,7 @@ static ssize_t node_desc_store(struct device *device, struct ib_device_modify desc = {}; int ret; - if (!dev->modify_device) + if (!dev->ops.modify_device) return -EIO; memcpy(desc.node_desc, buf, min_t(int, count, IB_DEVICE_NODE_DESC_MAX)); @@ -1341,7 +1341,7 @@ int ib_device_register_sysfs(struct ib_device *device, } } - if (device->alloc_hw_stats) + if (device->ops.alloc_hw_stats) setup_hw_stats(device, NULL, 0); return 0; diff --git a/drivers/infiniband/core/ucm.c b/drivers/infiniband/core/ucm.c index 73332b9a25b5..7541fbaf58a3 100644 --- a/drivers/infiniband/core/ucm.c +++ b/drivers/infiniband/core/ucm.c @@ -1242,7 +1242,7 @@ static void ib_ucm_add_one(struct ib_device *device) dev_t base; struct ib_ucm_device *ucm_dev; - if (!device->alloc_ucontext || !rdma_cap_ib_cm(device, 1)) + if (!device->ops.alloc_ucontext || !rdma_cap_ib_cm(device, 1)) return; ucm_dev = kzalloc(sizeof *ucm_dev, GFP_KERNEL); diff --git a/drivers/infiniband/core/uverbs_cmd.c b/drivers/infiniband/core/uverbs_cmd.c index 7f10eade7653..357d33120ca4 100644 --- a/drivers/infiniband/core/uverbs_cmd.c +++ b/drivers/infiniband/core/uverbs_cmd.c @@ -220,7 +220,7 @@ static int ib_uverbs_get_context(struct uverbs_attr_bundle *attrs) if (ret) goto err; - ucontext = ib_dev->alloc_ucontext(ib_dev, &attrs->driver_udata); + ucontext = ib_dev->ops.alloc_ucontext(ib_dev, &attrs->driver_udata); if (IS_ERR(ucontext)) { ret = PTR_ERR(ucontext); goto err_alloc; @@ -282,7 +282,7 @@ err_fd: put_unused_fd(resp.async_fd); err_free: - ib_dev->dealloc_ucontext(ucontext); + ib_dev->ops.dealloc_ucontext(ucontext); err_alloc: ib_rdmacg_uncharge(&cg_obj, ib_dev, RDMACG_RESOURCE_HCA_HANDLE); @@ -457,7 +457,7 @@ static int ib_uverbs_alloc_pd(struct uverbs_attr_bundle *attrs) if (IS_ERR(uobj)) return PTR_ERR(uobj); - pd = ib_dev->alloc_pd(ib_dev, uobj->context, &attrs->driver_udata); + pd = ib_dev->ops.alloc_pd(ib_dev, uobj->context, &attrs->driver_udata); if (IS_ERR(pd)) { ret = PTR_ERR(pd); goto err; @@ -634,8 +634,8 @@ static int ib_uverbs_open_xrcd(struct uverbs_attr_bundle *attrs) } if (!xrcd) { - xrcd = ib_dev->alloc_xrcd(ib_dev, obj->uobject.context, - &attrs->driver_udata); + xrcd = ib_dev->ops.alloc_xrcd(ib_dev, obj->uobject.context, + &attrs->driver_udata); if (IS_ERR(xrcd)) { ret = PTR_ERR(xrcd); goto err; @@ -774,8 +774,9 @@ static int ib_uverbs_reg_mr(struct uverbs_attr_bundle *attrs) } } - mr = pd->device->reg_user_mr(pd, cmd.start, cmd.length, cmd.hca_va, - cmd.access_flags, &attrs->driver_udata); + mr = pd->device->ops.reg_user_mr(pd, cmd.start, cmd.length, cmd.hca_va, + cmd.access_flags, + &attrs->driver_udata); if (IS_ERR(mr)) { ret = PTR_ERR(mr); goto err_put; @@ -864,9 +865,10 @@ static int ib_uverbs_rereg_mr(struct uverbs_attr_bundle *attrs) } old_pd = mr->pd; - ret = mr->device->rereg_user_mr(mr, cmd.flags, cmd.start, cmd.length, - cmd.hca_va, cmd.access_flags, pd, - &attrs->driver_udata); + ret = mr->device->ops.rereg_user_mr(mr, cmd.flags, cmd.start, + cmd.length, cmd.hca_va, + cmd.access_flags, pd, + &attrs->driver_udata); if (!ret) { if (cmd.flags & IB_MR_REREG_PD) { atomic_inc(&pd->usecnt); @@ -929,7 +931,7 @@ static int ib_uverbs_alloc_mw(struct uverbs_attr_bundle *attrs) goto err_free; } - mw = pd->device->alloc_mw(pd, cmd.mw_type, &attrs->driver_udata); + mw = pd->device->ops.alloc_mw(pd, cmd.mw_type, &attrs->driver_udata); if (IS_ERR(mw)) { ret = PTR_ERR(mw); goto err_put; @@ -1043,8 +1045,8 @@ static struct ib_ucq_object *create_cq(struct uverbs_attr_bundle *attrs, attr.comp_vector = cmd->comp_vector; attr.flags = cmd->flags; - cq = ib_dev->create_cq(ib_dev, &attr, obj->uobject.context, - &attrs->driver_udata); + cq = ib_dev->ops.create_cq(ib_dev, &attr, obj->uobject.context, + &attrs->driver_udata); if (IS_ERR(cq)) { ret = PTR_ERR(cq); goto err_file; @@ -1144,7 +1146,7 @@ static int ib_uverbs_resize_cq(struct uverbs_attr_bundle *attrs) if (!cq) return -EINVAL; - ret = cq->device->resize_cq(cq, cmd.cqe, &attrs->driver_udata); + ret = cq->device->ops.resize_cq(cq, cmd.cqe, &attrs->driver_udata); if (ret) goto out; @@ -2188,7 +2190,7 @@ static int ib_uverbs_post_send(struct uverbs_attr_bundle *attrs) } resp.bad_wr = 0; - ret = qp->device->post_send(qp->real_qp, wr, &bad_wr); + ret = qp->device->ops.post_send(qp->real_qp, wr, &bad_wr); if (ret) for (next = wr; next; next = next->next) { ++resp.bad_wr; @@ -2341,7 +2343,7 @@ static int ib_uverbs_post_recv(struct uverbs_attr_bundle *attrs) } resp.bad_wr = 0; - ret = qp->device->post_recv(qp->real_qp, wr, &bad_wr); + ret = qp->device->ops.post_recv(qp->real_qp, wr, &bad_wr); uobj_put_obj_read(qp); if (ret) { @@ -2391,7 +2393,7 @@ static int ib_uverbs_post_srq_recv(struct uverbs_attr_bundle *attrs) } resp.bad_wr = 0; - ret = srq->device->post_srq_recv(srq, wr, &bad_wr); + ret = srq->device->ops.post_srq_recv(srq, wr, &bad_wr); uobj_put_obj_read(srq); @@ -2961,7 +2963,7 @@ static int ib_uverbs_ex_create_wq(struct uverbs_attr_bundle *attrs) obj->uevent.events_reported = 0; INIT_LIST_HEAD(&obj->uevent.event_list); - wq = pd->device->create_wq(pd, &wq_init_attr, &attrs->driver_udata); + wq = pd->device->ops.create_wq(pd, &wq_init_attr, &attrs->driver_udata); if (IS_ERR(wq)) { err = PTR_ERR(wq); goto err_put_cq; @@ -3061,8 +3063,8 @@ static int ib_uverbs_ex_modify_wq(struct uverbs_attr_bundle *attrs) wq_attr.flags = cmd.flags; wq_attr.flags_mask = cmd.flags_mask; } - ret = wq->device->modify_wq(wq, &wq_attr, cmd.attr_mask, - &attrs->driver_udata); + ret = wq->device->ops.modify_wq(wq, &wq_attr, cmd.attr_mask, + &attrs->driver_udata); uobj_put_obj_read(wq); return ret; } @@ -3135,8 +3137,8 @@ static int ib_uverbs_ex_create_rwq_ind_table(struct uverbs_attr_bundle *attrs) init_attr.log_ind_tbl_size = cmd.log_ind_tbl_size; init_attr.ind_tbl = wqs; - rwq_ind_tbl = ib_dev->create_rwq_ind_table(ib_dev, &init_attr, - &attrs->driver_udata); + rwq_ind_tbl = ib_dev->ops.create_rwq_ind_table(ib_dev, &init_attr, + &attrs->driver_udata); if (IS_ERR(rwq_ind_tbl)) { err = PTR_ERR(rwq_ind_tbl); @@ -3323,8 +3325,8 @@ static int ib_uverbs_ex_create_flow(struct uverbs_attr_bundle *attrs) goto err_free; } - flow_id = qp->device->create_flow(qp, flow_attr, IB_FLOW_DOMAIN_USER, - &attrs->driver_udata); + flow_id = qp->device->ops.create_flow( + qp, flow_attr, IB_FLOW_DOMAIN_USER, &attrs->driver_udata); if (IS_ERR(flow_id)) { err = PTR_ERR(flow_id); @@ -3346,7 +3348,7 @@ static int ib_uverbs_ex_create_flow(struct uverbs_attr_bundle *attrs) kfree(kern_flow_attr); return uobj_alloc_commit(uobj); err_copy: - if (!qp->device->destroy_flow(flow_id)) + if (!qp->device->ops.destroy_flow(flow_id)) atomic_dec(&qp->usecnt); err_free: ib_uverbs_flow_resources_free(uflow_res); @@ -3441,7 +3443,7 @@ static int __uverbs_create_xsrq(struct uverbs_attr_bundle *attrs, obj->uevent.events_reported = 0; INIT_LIST_HEAD(&obj->uevent.event_list); - srq = pd->device->create_srq(pd, &attr, udata); + srq = pd->device->ops.create_srq(pd, &attr, udata); if (IS_ERR(srq)) { ret = PTR_ERR(srq); goto err_put; @@ -3563,8 +3565,8 @@ static int ib_uverbs_modify_srq(struct uverbs_attr_bundle *attrs) attr.max_wr = cmd.max_wr; attr.srq_limit = cmd.srq_limit; - ret = srq->device->modify_srq(srq, &attr, cmd.attr_mask, - &attrs->driver_udata); + ret = srq->device->ops.modify_srq(srq, &attr, cmd.attr_mask, + &attrs->driver_udata); uobj_put_obj_read(srq); @@ -3652,7 +3654,7 @@ static int ib_uverbs_ex_query_device(struct uverbs_attr_bundle *attrs) if (cmd.reserved) return -EINVAL; - err = ib_dev->query_device(ib_dev, &attr, &attrs->driver_udata); + err = ib_dev->ops.query_device(ib_dev, &attr, &attrs->driver_udata); if (err) return err; diff --git a/drivers/infiniband/core/uverbs_main.c b/drivers/infiniband/core/uverbs_main.c index 96a5f89bbb75..9f9172eb1512 100644 --- a/drivers/infiniband/core/uverbs_main.c +++ b/drivers/infiniband/core/uverbs_main.c @@ -106,7 +106,7 @@ int uverbs_dealloc_mw(struct ib_mw *mw) struct ib_pd *pd = mw->pd; int ret; - ret = mw->device->dealloc_mw(mw); + ret = mw->device->ops.dealloc_mw(mw); if (!ret) atomic_dec(&pd->usecnt); return ret; @@ -197,7 +197,7 @@ void ib_uverbs_release_file(struct kref *ref) srcu_key = srcu_read_lock(&file->device->disassociate_srcu); ib_dev = srcu_dereference(file->device->ib_dev, &file->device->disassociate_srcu); - if (ib_dev && !ib_dev->disassociate_ucontext) + if (ib_dev && !ib_dev->ops.disassociate_ucontext) module_put(ib_dev->owner); srcu_read_unlock(&file->device->disassociate_srcu, srcu_key); @@ -774,7 +774,7 @@ static int ib_uverbs_mmap(struct file *filp, struct vm_area_struct *vma) goto out; } - ret = ucontext->device->mmap(ucontext, vma); + ret = ucontext->device->ops.mmap(ucontext, vma); out: srcu_read_unlock(&file->device->disassociate_srcu, srcu_key); return ret; @@ -1036,7 +1036,7 @@ static int ib_uverbs_open(struct inode *inode, struct file *filp) /* In case IB device supports disassociate ucontext, there is no hard * dependency between uverbs device and its low level device. */ - module_dependent = !(ib_dev->disassociate_ucontext); + module_dependent = !(ib_dev->ops.disassociate_ucontext); if (module_dependent) { if (!try_module_get(ib_dev->owner)) { @@ -1203,7 +1203,7 @@ static void ib_uverbs_add_one(struct ib_device *device) struct ib_uverbs_device *uverbs_dev; int ret; - if (!device->alloc_ucontext) + if (!device->ops.alloc_ucontext) return; uverbs_dev = kzalloc(sizeof(*uverbs_dev), GFP_KERNEL); @@ -1249,7 +1249,7 @@ static void ib_uverbs_add_one(struct ib_device *device) dev_set_name(&uverbs_dev->dev, "uverbs%d", uverbs_dev->devnum); cdev_init(&uverbs_dev->cdev, - device->mmap ? &uverbs_mmap_fops : &uverbs_fops); + device->ops.mmap ? &uverbs_mmap_fops : &uverbs_fops); uverbs_dev->cdev.owner = THIS_MODULE; ret = cdev_device_add(&uverbs_dev->cdev, &uverbs_dev->dev); @@ -1337,7 +1337,7 @@ static void ib_uverbs_remove_one(struct ib_device *device, void *client_data) cdev_device_del(&uverbs_dev->cdev, &uverbs_dev->dev); ida_free(&uverbs_ida, uverbs_dev->devnum); - if (device->disassociate_ucontext) { + if (device->ops.disassociate_ucontext) { /* We disassociate HW resources and immediately return. * Userspace will see a EIO errno for all future access. * Upon returning, ib_device may be freed internally and is not diff --git a/drivers/infiniband/core/uverbs_std_types.c b/drivers/infiniband/core/uverbs_std_types.c index 063aff9e7a04..424f325f8cba 100644 --- a/drivers/infiniband/core/uverbs_std_types.c +++ b/drivers/infiniband/core/uverbs_std_types.c @@ -54,7 +54,7 @@ static int uverbs_free_flow(struct ib_uobject *uobject, struct ib_qp *qp = flow->qp; int ret; - ret = flow->device->destroy_flow(flow); + ret = flow->device->ops.destroy_flow(flow); if (!ret) { if (qp) atomic_dec(&qp->usecnt); diff --git a/drivers/infiniband/core/uverbs_std_types_counters.c b/drivers/infiniband/core/uverbs_std_types_counters.c index 8835bad5c6dd..309c5e80988d 100644 --- a/drivers/infiniband/core/uverbs_std_types_counters.c +++ b/drivers/infiniband/core/uverbs_std_types_counters.c @@ -44,7 +44,7 @@ static int uverbs_free_counters(struct ib_uobject *uobject, if (ret) return ret; - return counters->device->destroy_counters(counters); + return counters->device->ops.destroy_counters(counters); } static int UVERBS_HANDLER(UVERBS_METHOD_COUNTERS_CREATE)( @@ -61,10 +61,10 @@ static int UVERBS_HANDLER(UVERBS_METHOD_COUNTERS_CREATE)( * have the ability to remove methods from parse tree once * such condition is met. */ - if (!ib_dev->create_counters) + if (!ib_dev->ops.create_counters) return -EOPNOTSUPP; - counters = ib_dev->create_counters(ib_dev, attrs); + counters = ib_dev->ops.create_counters(ib_dev, attrs); if (IS_ERR(counters)) { ret = PTR_ERR(counters); goto err_create_counters; @@ -90,7 +90,7 @@ static int UVERBS_HANDLER(UVERBS_METHOD_COUNTERS_READ)( uverbs_attr_get_obj(attrs, UVERBS_ATTR_READ_COUNTERS_HANDLE); int ret; - if (!counters->device->read_counters) + if (!counters->device->ops.read_counters) return -EOPNOTSUPP; if (!atomic_read(&counters->usecnt)) @@ -109,7 +109,7 @@ static int UVERBS_HANDLER(UVERBS_METHOD_COUNTERS_READ)( if (IS_ERR(read_attr.counters_buff)) return PTR_ERR(read_attr.counters_buff); - ret = counters->device->read_counters(counters, &read_attr, attrs); + ret = counters->device->ops.read_counters(counters, &read_attr, attrs); if (ret) return ret; diff --git a/drivers/infiniband/core/uverbs_std_types_cq.c b/drivers/infiniband/core/uverbs_std_types_cq.c index 859518eab583..42df59635a3c 100644 --- a/drivers/infiniband/core/uverbs_std_types_cq.c +++ b/drivers/infiniband/core/uverbs_std_types_cq.c @@ -71,7 +71,7 @@ static int UVERBS_HANDLER(UVERBS_METHOD_CQ_CREATE)( struct ib_uverbs_completion_event_file *ev_file = NULL; struct ib_uobject *ev_file_uobj; - if (!ib_dev->create_cq || !ib_dev->destroy_cq) + if (!ib_dev->ops.create_cq || !ib_dev->ops.destroy_cq) return -EOPNOTSUPP; ret = uverbs_copy_from(&attr.comp_vector, attrs, @@ -110,8 +110,8 @@ static int UVERBS_HANDLER(UVERBS_METHOD_CQ_CREATE)( INIT_LIST_HEAD(&obj->comp_list); INIT_LIST_HEAD(&obj->async_list); - cq = ib_dev->create_cq(ib_dev, &attr, obj->uobject.context, - &attrs->driver_udata); + cq = ib_dev->ops.create_cq(ib_dev, &attr, obj->uobject.context, + &attrs->driver_udata); if (IS_ERR(cq)) { ret = PTR_ERR(cq); goto err_event_file; diff --git a/drivers/infiniband/core/uverbs_std_types_dm.c b/drivers/infiniband/core/uverbs_std_types_dm.c index 658261b8f08e..2ef70637bee1 100644 --- a/drivers/infiniband/core/uverbs_std_types_dm.c +++ b/drivers/infiniband/core/uverbs_std_types_dm.c @@ -43,7 +43,7 @@ static int uverbs_free_dm(struct ib_uobject *uobject, if (ret) return ret; - return dm->device->dealloc_dm(dm); + return dm->device->ops.dealloc_dm(dm); } static int UVERBS_HANDLER(UVERBS_METHOD_DM_ALLOC)( @@ -57,7 +57,7 @@ static int UVERBS_HANDLER(UVERBS_METHOD_DM_ALLOC)( struct ib_dm *dm; int ret; - if (!ib_dev->alloc_dm) + if (!ib_dev->ops.alloc_dm) return -EOPNOTSUPP; ret = uverbs_copy_from(&attr.length, attrs, @@ -70,7 +70,7 @@ static int UVERBS_HANDLER(UVERBS_METHOD_DM_ALLOC)( if (ret) return ret; - dm = ib_dev->alloc_dm(ib_dev, uobj->context, &attr, attrs); + dm = ib_dev->ops.alloc_dm(ib_dev, uobj->context, &attr, attrs); if (IS_ERR(dm)) return PTR_ERR(dm); diff --git a/drivers/infiniband/core/uverbs_std_types_flow_action.c b/drivers/infiniband/core/uverbs_std_types_flow_action.c index e4d01fb5335d..4962b87fa600 100644 --- a/drivers/infiniband/core/uverbs_std_types_flow_action.c +++ b/drivers/infiniband/core/uverbs_std_types_flow_action.c @@ -43,7 +43,7 @@ static int uverbs_free_flow_action(struct ib_uobject *uobject, if (ret) return ret; - return action->device->destroy_flow_action(action); + return action->device->ops.destroy_flow_action(action); } static u64 esp_flags_uverbs_to_verbs(struct uverbs_attr_bundle *attrs, @@ -313,7 +313,7 @@ static int UVERBS_HANDLER(UVERBS_METHOD_FLOW_ACTION_ESP_CREATE)( struct ib_flow_action *action; struct ib_flow_action_esp_attr esp_attr = {}; - if (!ib_dev->create_flow_action_esp) + if (!ib_dev->ops.create_flow_action_esp) return -EOPNOTSUPP; ret = parse_flow_action_esp(ib_dev, attrs, &esp_attr, false); @@ -321,7 +321,8 @@ static int UVERBS_HANDLER(UVERBS_METHOD_FLOW_ACTION_ESP_CREATE)( return ret; /* No need to check as this attribute is marked as MANDATORY */ - action = ib_dev->create_flow_action_esp(ib_dev, &esp_attr.hdr, attrs); + action = ib_dev->ops.create_flow_action_esp(ib_dev, &esp_attr.hdr, + attrs); if (IS_ERR(action)) return PTR_ERR(action); @@ -340,7 +341,7 @@ static int UVERBS_HANDLER(UVERBS_METHOD_FLOW_ACTION_ESP_MODIFY)( int ret; struct ib_flow_action_esp_attr esp_attr = {}; - if (!action->device->modify_flow_action_esp) + if (!action->device->ops.modify_flow_action_esp) return -EOPNOTSUPP; ret = parse_flow_action_esp(action->device, attrs, &esp_attr, true); @@ -350,8 +351,9 @@ static int UVERBS_HANDLER(UVERBS_METHOD_FLOW_ACTION_ESP_MODIFY)( if (action->type != IB_FLOW_ACTION_ESP) return -EINVAL; - return action->device->modify_flow_action_esp(action, &esp_attr.hdr, - attrs); + return action->device->ops.modify_flow_action_esp(action, + &esp_attr.hdr, + attrs); } static const struct uverbs_attr_spec uverbs_flow_action_esp_keymat[] = { diff --git a/drivers/infiniband/core/uverbs_std_types_mr.c b/drivers/infiniband/core/uverbs_std_types_mr.c index 70ea48cfc047..cafb49a45515 100644 --- a/drivers/infiniband/core/uverbs_std_types_mr.c +++ b/drivers/infiniband/core/uverbs_std_types_mr.c @@ -54,7 +54,7 @@ static int UVERBS_HANDLER(UVERBS_METHOD_DM_MR_REG)( struct ib_mr *mr; int ret; - if (!ib_dev->reg_dm_mr) + if (!ib_dev->ops.reg_dm_mr) return -EOPNOTSUPP; ret = uverbs_copy_from(&attr.offset, attrs, UVERBS_ATTR_REG_DM_MR_OFFSET); @@ -83,7 +83,7 @@ static int UVERBS_HANDLER(UVERBS_METHOD_DM_MR_REG)( attr.length > dm->length - attr.offset) return -EINVAL; - mr = pd->device->reg_dm_mr(pd, dm, &attr, attrs); + mr = pd->device->ops.reg_dm_mr(pd, dm, &attr, attrs); if (IS_ERR(mr)) return PTR_ERR(mr); diff --git a/drivers/infiniband/core/uverbs_uapi.c b/drivers/infiniband/core/uverbs_uapi.c index 0136c1d78a0f..830e48fe5e65 100644 --- a/drivers/infiniband/core/uverbs_uapi.c +++ b/drivers/infiniband/core/uverbs_uapi.c @@ -300,7 +300,8 @@ static int uapi_merge_def(struct uverbs_api *uapi, struct ib_device *ibdev, return 0; case UAPI_DEF_IS_SUPPORTED_DEV_FN: { - void **ibdev_fn = (void *)ibdev + def->needs_fn_offset; + void **ibdev_fn = + (void *)(&ibdev->ops) + def->needs_fn_offset; if (*ibdev_fn) continue; diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c index fb2fc0c7ecfb..92dbc758f6c9 100644 --- a/drivers/infiniband/core/verbs.c +++ b/drivers/infiniband/core/verbs.c @@ -226,8 +226,8 @@ EXPORT_SYMBOL(rdma_node_get_transport); enum rdma_link_layer rdma_port_get_link_layer(struct ib_device *device, u8 port_num) { enum rdma_transport_type lt; - if (device->get_link_layer) - return device->get_link_layer(device, port_num); + if (device->ops.get_link_layer) + return device->ops.get_link_layer(device, port_num); lt = rdma_node_get_transport(device->node_type); if (lt == RDMA_TRANSPORT_IB) @@ -255,7 +255,7 @@ struct ib_pd *__ib_alloc_pd(struct ib_device *device, unsigned int flags, struct ib_pd *pd; int mr_access_flags = 0; - pd = device->alloc_pd(device, NULL, NULL); + pd = device->ops.alloc_pd(device, NULL, NULL); if (IS_ERR(pd)) return pd; @@ -282,7 +282,7 @@ struct ib_pd *__ib_alloc_pd(struct ib_device *device, unsigned int flags, if (mr_access_flags) { struct ib_mr *mr; - mr = pd->device->get_dma_mr(pd, mr_access_flags); + mr = pd->device->ops.get_dma_mr(pd, mr_access_flags); if (IS_ERR(mr)) { ib_dealloc_pd(pd); return ERR_CAST(mr); @@ -319,7 +319,7 @@ void ib_dealloc_pd(struct ib_pd *pd) int ret; if (pd->__internal_mr) { - ret = pd->device->dereg_mr(pd->__internal_mr); + ret = pd->device->ops.dereg_mr(pd->__internal_mr); WARN_ON(ret); pd->__internal_mr = NULL; } @@ -331,7 +331,7 @@ void ib_dealloc_pd(struct ib_pd *pd) rdma_restrack_del(&pd->res); /* Making delalloc_pd a void return is a WIP, no driver should return an error here. */ - ret = pd->device->dealloc_pd(pd); + ret = pd->device->ops.dealloc_pd(pd); WARN_ONCE(ret, "Infiniband HW driver failed dealloc_pd"); } EXPORT_SYMBOL(ib_dealloc_pd); @@ -491,10 +491,10 @@ static struct ib_ah *_rdma_create_ah(struct ib_pd *pd, { struct ib_ah *ah; - if (!pd->device->create_ah) + if (!pd->device->ops.create_ah) return ERR_PTR(-EOPNOTSUPP); - ah = pd->device->create_ah(pd, ah_attr, udata); + ah = pd->device->ops.create_ah(pd, ah_attr, udata); if (!IS_ERR(ah)) { ah->device = pd->device; @@ -900,8 +900,8 @@ int rdma_modify_ah(struct ib_ah *ah, struct rdma_ah_attr *ah_attr) if (ret) return ret; - ret = ah->device->modify_ah ? - ah->device->modify_ah(ah, ah_attr) : + ret = ah->device->ops.modify_ah ? + ah->device->ops.modify_ah(ah, ah_attr) : -EOPNOTSUPP; ah->sgid_attr = rdma_update_sgid_attr(ah_attr, ah->sgid_attr); @@ -914,8 +914,8 @@ int rdma_query_ah(struct ib_ah *ah, struct rdma_ah_attr *ah_attr) { ah_attr->grh.sgid_attr = NULL; - return ah->device->query_ah ? - ah->device->query_ah(ah, ah_attr) : + return ah->device->ops.query_ah ? + ah->device->ops.query_ah(ah, ah_attr) : -EOPNOTSUPP; } EXPORT_SYMBOL(rdma_query_ah); @@ -927,7 +927,7 @@ int rdma_destroy_ah(struct ib_ah *ah) int ret; pd = ah->pd; - ret = ah->device->destroy_ah(ah); + ret = ah->device->ops.destroy_ah(ah); if (!ret) { atomic_dec(&pd->usecnt); if (sgid_attr) @@ -945,10 +945,10 @@ struct ib_srq *ib_create_srq(struct ib_pd *pd, { struct ib_srq *srq; - if (!pd->device->create_srq) + if (!pd->device->ops.create_srq) return ERR_PTR(-EOPNOTSUPP); - srq = pd->device->create_srq(pd, srq_init_attr, NULL); + srq = pd->device->ops.create_srq(pd, srq_init_attr, NULL); if (!IS_ERR(srq)) { srq->device = pd->device; @@ -977,17 +977,17 @@ int ib_modify_srq(struct ib_srq *srq, struct ib_srq_attr *srq_attr, enum ib_srq_attr_mask srq_attr_mask) { - return srq->device->modify_srq ? - srq->device->modify_srq(srq, srq_attr, srq_attr_mask, NULL) : - -EOPNOTSUPP; + return srq->device->ops.modify_srq ? + srq->device->ops.modify_srq(srq, srq_attr, srq_attr_mask, + NULL) : -EOPNOTSUPP; } EXPORT_SYMBOL(ib_modify_srq); int ib_query_srq(struct ib_srq *srq, struct ib_srq_attr *srq_attr) { - return srq->device->query_srq ? - srq->device->query_srq(srq, srq_attr) : -EOPNOTSUPP; + return srq->device->ops.query_srq ? + srq->device->ops.query_srq(srq, srq_attr) : -EOPNOTSUPP; } EXPORT_SYMBOL(ib_query_srq); @@ -1009,7 +1009,7 @@ int ib_destroy_srq(struct ib_srq *srq) if (srq_type == IB_SRQT_XRC) xrcd = srq->ext.xrc.xrcd; - ret = srq->device->destroy_srq(srq); + ret = srq->device->ops.destroy_srq(srq); if (!ret) { atomic_dec(&pd->usecnt); if (srq_type == IB_SRQT_XRC) @@ -1118,7 +1118,7 @@ static struct ib_qp *ib_create_xrc_qp(struct ib_qp *qp, if (!IS_ERR(qp)) __ib_insert_xrcd_qp(qp_init_attr->xrcd, real_qp); else - real_qp->device->destroy_qp(real_qp); + real_qp->device->ops.destroy_qp(real_qp); return qp; } @@ -1704,10 +1704,10 @@ int ib_get_eth_speed(struct ib_device *dev, u8 port_num, u8 *speed, u8 *width) if (rdma_port_get_link_layer(dev, port_num) != IB_LINK_LAYER_ETHERNET) return -EINVAL; - if (!dev->get_netdev) + if (!dev->ops.get_netdev) return -EOPNOTSUPP; - netdev = dev->get_netdev(dev, port_num); + netdev = dev->ops.get_netdev(dev, port_num); if (!netdev) return -ENODEV; @@ -1765,9 +1765,9 @@ int ib_query_qp(struct ib_qp *qp, qp_attr->ah_attr.grh.sgid_attr = NULL; qp_attr->alt_ah_attr.grh.sgid_attr = NULL; - return qp->device->query_qp ? - qp->device->query_qp(qp->real_qp, qp_attr, qp_attr_mask, qp_init_attr) : - -EOPNOTSUPP; + return qp->device->ops.query_qp ? + qp->device->ops.query_qp(qp->real_qp, qp_attr, qp_attr_mask, + qp_init_attr) : -EOPNOTSUPP; } EXPORT_SYMBOL(ib_query_qp); @@ -1853,7 +1853,7 @@ int ib_destroy_qp(struct ib_qp *qp) rdma_rw_cleanup_mrs(qp); rdma_restrack_del(&qp->res); - ret = qp->device->destroy_qp(qp); + ret = qp->device->ops.destroy_qp(qp); if (!ret) { if (alt_path_sgid_attr) rdma_put_gid_attr(alt_path_sgid_attr); @@ -1891,7 +1891,7 @@ struct ib_cq *__ib_create_cq(struct ib_device *device, { struct ib_cq *cq; - cq = device->create_cq(device, cq_attr, NULL, NULL); + cq = device->ops.create_cq(device, cq_attr, NULL, NULL); if (!IS_ERR(cq)) { cq->device = device; @@ -1911,8 +1911,9 @@ EXPORT_SYMBOL(__ib_create_cq); int rdma_set_cq_moderation(struct ib_cq *cq, u16 cq_count, u16 cq_period) { - return cq->device->modify_cq ? - cq->device->modify_cq(cq, cq_count, cq_period) : -EOPNOTSUPP; + return cq->device->ops.modify_cq ? + cq->device->ops.modify_cq(cq, cq_count, + cq_period) : -EOPNOTSUPP; } EXPORT_SYMBOL(rdma_set_cq_moderation); @@ -1922,14 +1923,14 @@ int ib_destroy_cq(struct ib_cq *cq) return -EBUSY; rdma_restrack_del(&cq->res); - return cq->device->destroy_cq(cq); + return cq->device->ops.destroy_cq(cq); } EXPORT_SYMBOL(ib_destroy_cq); int ib_resize_cq(struct ib_cq *cq, int cqe) { - return cq->device->resize_cq ? - cq->device->resize_cq(cq, cqe, NULL) : -EOPNOTSUPP; + return cq->device->ops.resize_cq ? + cq->device->ops.resize_cq(cq, cqe, NULL) : -EOPNOTSUPP; } EXPORT_SYMBOL(ib_resize_cq); @@ -1942,7 +1943,7 @@ int ib_dereg_mr(struct ib_mr *mr) int ret; rdma_restrack_del(&mr->res); - ret = mr->device->dereg_mr(mr); + ret = mr->device->ops.dereg_mr(mr); if (!ret) { atomic_dec(&pd->usecnt); if (dm) @@ -1971,10 +1972,10 @@ struct ib_mr *ib_alloc_mr(struct ib_pd *pd, { struct ib_mr *mr; - if (!pd->device->alloc_mr) + if (!pd->device->ops.alloc_mr) return ERR_PTR(-EOPNOTSUPP); - mr = pd->device->alloc_mr(pd, mr_type, max_num_sg); + mr = pd->device->ops.alloc_mr(pd, mr_type, max_num_sg); if (!IS_ERR(mr)) { mr->device = pd->device; mr->pd = pd; @@ -1998,10 +1999,10 @@ struct ib_fmr *ib_alloc_fmr(struct ib_pd *pd, { struct ib_fmr *fmr; - if (!pd->device->alloc_fmr) + if (!pd->device->ops.alloc_fmr) return ERR_PTR(-EOPNOTSUPP); - fmr = pd->device->alloc_fmr(pd, mr_access_flags, fmr_attr); + fmr = pd->device->ops.alloc_fmr(pd, mr_access_flags, fmr_attr); if (!IS_ERR(fmr)) { fmr->device = pd->device; fmr->pd = pd; @@ -2020,7 +2021,7 @@ int ib_unmap_fmr(struct list_head *fmr_list) return 0; fmr = list_entry(fmr_list->next, struct ib_fmr, list); - return fmr->device->unmap_fmr(fmr_list); + return fmr->device->ops.unmap_fmr(fmr_list); } EXPORT_SYMBOL(ib_unmap_fmr); @@ -2030,7 +2031,7 @@ int ib_dealloc_fmr(struct ib_fmr *fmr) int ret; pd = fmr->pd; - ret = fmr->device->dealloc_fmr(fmr); + ret = fmr->device->ops.dealloc_fmr(fmr); if (!ret) atomic_dec(&pd->usecnt); @@ -2082,14 +2083,14 @@ int ib_attach_mcast(struct ib_qp *qp, union ib_gid *gid, u16 lid) { int ret; - if (!qp->device->attach_mcast) + if (!qp->device->ops.attach_mcast) return -EOPNOTSUPP; if (!rdma_is_multicast_addr((struct in6_addr *)gid->raw) || qp->qp_type != IB_QPT_UD || !is_valid_mcast_lid(qp, lid)) return -EINVAL; - ret = qp->device->attach_mcast(qp, gid, lid); + ret = qp->device->ops.attach_mcast(qp, gid, lid); if (!ret) atomic_inc(&qp->usecnt); return ret; @@ -2100,14 +2101,14 @@ int ib_detach_mcast(struct ib_qp *qp, union ib_gid *gid, u16 lid) { int ret; - if (!qp->device->detach_mcast) + if (!qp->device->ops.detach_mcast) return -EOPNOTSUPP; if (!rdma_is_multicast_addr((struct in6_addr *)gid->raw) || qp->qp_type != IB_QPT_UD || !is_valid_mcast_lid(qp, lid)) return -EINVAL; - ret = qp->device->detach_mcast(qp, gid, lid); + ret = qp->device->ops.detach_mcast(qp, gid, lid); if (!ret) atomic_dec(&qp->usecnt); return ret; @@ -2118,10 +2119,10 @@ struct ib_xrcd *__ib_alloc_xrcd(struct ib_device *device, const char *caller) { struct ib_xrcd *xrcd; - if (!device->alloc_xrcd) + if (!device->ops.alloc_xrcd) return ERR_PTR(-EOPNOTSUPP); - xrcd = device->alloc_xrcd(device, NULL, NULL); + xrcd = device->ops.alloc_xrcd(device, NULL, NULL); if (!IS_ERR(xrcd)) { xrcd->device = device; xrcd->inode = NULL; @@ -2149,7 +2150,7 @@ int ib_dealloc_xrcd(struct ib_xrcd *xrcd) return ret; } - return xrcd->device->dealloc_xrcd(xrcd); + return xrcd->device->ops.dealloc_xrcd(xrcd); } EXPORT_SYMBOL(ib_dealloc_xrcd); @@ -2172,10 +2173,10 @@ struct ib_wq *ib_create_wq(struct ib_pd *pd, { struct ib_wq *wq; - if (!pd->device->create_wq) + if (!pd->device->ops.create_wq) return ERR_PTR(-EOPNOTSUPP); - wq = pd->device->create_wq(pd, wq_attr, NULL); + wq = pd->device->ops.create_wq(pd, wq_attr, NULL); if (!IS_ERR(wq)) { wq->event_handler = wq_attr->event_handler; wq->wq_context = wq_attr->wq_context; @@ -2205,7 +2206,7 @@ int ib_destroy_wq(struct ib_wq *wq) if (atomic_read(&wq->usecnt)) return -EBUSY; - err = wq->device->destroy_wq(wq); + err = wq->device->ops.destroy_wq(wq); if (!err) { atomic_dec(&pd->usecnt); atomic_dec(&cq->usecnt); @@ -2227,10 +2228,10 @@ int ib_modify_wq(struct ib_wq *wq, struct ib_wq_attr *wq_attr, { int err; - if (!wq->device->modify_wq) + if (!wq->device->ops.modify_wq) return -EOPNOTSUPP; - err = wq->device->modify_wq(wq, wq_attr, wq_attr_mask, NULL); + err = wq->device->ops.modify_wq(wq, wq_attr, wq_attr_mask, NULL); return err; } EXPORT_SYMBOL(ib_modify_wq); @@ -2252,12 +2253,12 @@ struct ib_rwq_ind_table *ib_create_rwq_ind_table(struct ib_device *device, int i; u32 table_size; - if (!device->create_rwq_ind_table) + if (!device->ops.create_rwq_ind_table) return ERR_PTR(-EOPNOTSUPP); table_size = (1 << init_attr->log_ind_tbl_size); - rwq_ind_table = device->create_rwq_ind_table(device, - init_attr, NULL); + rwq_ind_table = device->ops.create_rwq_ind_table(device, + init_attr, NULL); if (IS_ERR(rwq_ind_table)) return rwq_ind_table; @@ -2287,7 +2288,7 @@ int ib_destroy_rwq_ind_table(struct ib_rwq_ind_table *rwq_ind_table) if (atomic_read(&rwq_ind_table->usecnt)) return -EBUSY; - err = rwq_ind_table->device->destroy_rwq_ind_table(rwq_ind_table); + err = rwq_ind_table->device->ops.destroy_rwq_ind_table(rwq_ind_table); if (!err) { for (i = 0; i < table_size; i++) atomic_dec(&ind_tbl[i]->usecnt); @@ -2300,48 +2301,50 @@ EXPORT_SYMBOL(ib_destroy_rwq_ind_table); int ib_check_mr_status(struct ib_mr *mr, u32 check_mask, struct ib_mr_status *mr_status) { - return mr->device->check_mr_status ? - mr->device->check_mr_status(mr, check_mask, mr_status) : -EOPNOTSUPP; + if (!mr->device->ops.check_mr_status) + return -EOPNOTSUPP; + + return mr->device->ops.check_mr_status(mr, check_mask, mr_status); } EXPORT_SYMBOL(ib_check_mr_status); int ib_set_vf_link_state(struct ib_device *device, int vf, u8 port, int state) { - if (!device->set_vf_link_state) + if (!device->ops.set_vf_link_state) return -EOPNOTSUPP; - return device->set_vf_link_state(device, vf, port, state); + return device->ops.set_vf_link_state(device, vf, port, state); } EXPORT_SYMBOL(ib_set_vf_link_state); int ib_get_vf_config(struct ib_device *device, int vf, u8 port, struct ifla_vf_info *info) { - if (!device->get_vf_config) + if (!device->ops.get_vf_config) return -EOPNOTSUPP; - return device->get_vf_config(device, vf, port, info); + return device->ops.get_vf_config(device, vf, port, info); } EXPORT_SYMBOL(ib_get_vf_config); int ib_get_vf_stats(struct ib_device *device, int vf, u8 port, struct ifla_vf_stats *stats) { - if (!device->get_vf_stats) + if (!device->ops.get_vf_stats) return -EOPNOTSUPP; - return device->get_vf_stats(device, vf, port, stats); + return device->ops.get_vf_stats(device, vf, port, stats); } EXPORT_SYMBOL(ib_get_vf_stats); int ib_set_vf_guid(struct ib_device *device, int vf, u8 port, u64 guid, int type) { - if (!device->set_vf_guid) + if (!device->ops.set_vf_guid) return -EOPNOTSUPP; - return device->set_vf_guid(device, vf, port, guid, type); + return device->ops.set_vf_guid(device, vf, port, guid, type); } EXPORT_SYMBOL(ib_set_vf_guid); @@ -2373,12 +2376,12 @@ EXPORT_SYMBOL(ib_set_vf_guid); int ib_map_mr_sg(struct ib_mr *mr, struct scatterlist *sg, int sg_nents, unsigned int *sg_offset, unsigned int page_size) { - if (unlikely(!mr->device->map_mr_sg)) + if (unlikely(!mr->device->ops.map_mr_sg)) return -EOPNOTSUPP; mr->page_size = page_size; - return mr->device->map_mr_sg(mr, sg, sg_nents, sg_offset); + return mr->device->ops.map_mr_sg(mr, sg, sg_nents, sg_offset); } EXPORT_SYMBOL(ib_map_mr_sg); @@ -2577,8 +2580,8 @@ static void __ib_drain_rq(struct ib_qp *qp) */ void ib_drain_sq(struct ib_qp *qp) { - if (qp->device->drain_sq) - qp->device->drain_sq(qp); + if (qp->device->ops.drain_sq) + qp->device->ops.drain_sq(qp); else __ib_drain_sq(qp); } @@ -2605,8 +2608,8 @@ EXPORT_SYMBOL(ib_drain_sq); */ void ib_drain_rq(struct ib_qp *qp) { - if (qp->device->drain_rq) - qp->device->drain_rq(qp); + if (qp->device->ops.drain_rq) + qp->device->ops.drain_rq(qp); else __ib_drain_rq(qp); } @@ -2644,10 +2647,11 @@ struct net_device *rdma_alloc_netdev(struct ib_device *device, u8 port_num, struct net_device *netdev; int rc; - if (!device->rdma_netdev_get_params) + if (!device->ops.rdma_netdev_get_params) return ERR_PTR(-EOPNOTSUPP); - rc = device->rdma_netdev_get_params(device, port_num, type, ¶ms); + rc = device->ops.rdma_netdev_get_params(device, port_num, type, + ¶ms); if (rc) return ERR_PTR(rc); @@ -2669,10 +2673,11 @@ int rdma_init_netdev(struct ib_device *device, u8 port_num, struct rdma_netdev_alloc_params params; int rc; - if (!device->rdma_netdev_get_params) + if (!device->ops.rdma_netdev_get_params) return -EOPNOTSUPP; - rc = device->rdma_netdev_get_params(device, port_num, type, ¶ms); + rc = device->ops.rdma_netdev_get_params(device, port_num, type, + ¶ms); if (rc) return rc; diff --git a/drivers/infiniband/hw/i40iw/i40iw_cm.c b/drivers/infiniband/hw/i40iw/i40iw_cm.c index 771eb6bd0785..ef137c40205c 100644 --- a/drivers/infiniband/hw/i40iw/i40iw_cm.c +++ b/drivers/infiniband/hw/i40iw/i40iw_cm.c @@ -3478,7 +3478,7 @@ static void i40iw_qp_disconnect(struct i40iw_qp *iwqp) /* Need to free the Last Streaming Mode Message */ if (iwqp->ietf_mem.va) { if (iwqp->lsmm_mr) - iwibdev->ibdev.dereg_mr(iwqp->lsmm_mr); + iwibdev->ibdev.ops.dereg_mr(iwqp->lsmm_mr); i40iw_free_dma_mem(iwdev->sc_dev.hw, &iwqp->ietf_mem); } } diff --git a/drivers/infiniband/hw/mlx4/alias_GUID.c b/drivers/infiniband/hw/mlx4/alias_GUID.c index 155b4dfc0ae8..782499abcd98 100644 --- a/drivers/infiniband/hw/mlx4/alias_GUID.c +++ b/drivers/infiniband/hw/mlx4/alias_GUID.c @@ -849,7 +849,7 @@ int mlx4_ib_init_alias_guid_service(struct mlx4_ib_dev *dev) spin_lock_init(&dev->sriov.alias_guid.ag_work_lock); for (i = 1; i <= dev->num_ports; ++i) { - if (dev->ib_dev.query_gid(&dev->ib_dev , i, 0, &gid)) { + if (dev->ib_dev.ops.query_gid(&dev->ib_dev, i, 0, &gid)) { ret = -EFAULT; goto err_unregister; } diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index 47d9cd260846..5a4e23105b0c 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -150,7 +150,7 @@ static int get_port_state(struct ib_device *ibdev, int ret; memset(&attr, 0, sizeof(attr)); - ret = ibdev->query_port(ibdev, port_num, &attr); + ret = ibdev->ops.query_port(ibdev, port_num, &attr); if (!ret) *state = attr.state; return ret; diff --git a/drivers/infiniband/hw/nes/nes_cm.c b/drivers/infiniband/hw/nes/nes_cm.c index 2b67ace5b614..032883180f65 100644 --- a/drivers/infiniband/hw/nes/nes_cm.c +++ b/drivers/infiniband/hw/nes/nes_cm.c @@ -3033,7 +3033,7 @@ static int nes_disconnect(struct nes_qp *nesqp, int abrupt) /* Need to free the Last Streaming Mode Message */ if (nesqp->ietf_frame) { if (nesqp->lsmm_mr) - nesibdev->ibdev.dereg_mr(nesqp->lsmm_mr); + nesibdev->ibdev.ops.dereg_mr(nesqp->lsmm_mr); pci_free_consistent(nesdev->pcidev, nesqp->private_data_len + nesqp->ietf_frame_size, nesqp->ietf_frame, nesqp->ietf_frame_pbase); diff --git a/drivers/infiniband/sw/rdmavt/vt.c b/drivers/infiniband/sw/rdmavt/vt.c index c52b38fe2416..aef3aa3fe667 100644 --- a/drivers/infiniband/sw/rdmavt/vt.c +++ b/drivers/infiniband/sw/rdmavt/vt.c @@ -456,31 +456,31 @@ static noinline int check_support(struct rvt_dev_info *rdi, int verb) * rdmavt does not support modify device currently drivers must * provide. */ - if (!rdi->ibdev.modify_device) + if (!rdi->ibdev.ops.modify_device) return -EOPNOTSUPP; break; case QUERY_PORT: - if (!rdi->ibdev.query_port) + if (!rdi->ibdev.ops.query_port) if (!rdi->driver_f.query_port_state) return -EINVAL; break; case MODIFY_PORT: - if (!rdi->ibdev.modify_port) + if (!rdi->ibdev.ops.modify_port) if (!rdi->driver_f.cap_mask_chg || !rdi->driver_f.shut_down_port) return -EINVAL; break; case QUERY_GID: - if (!rdi->ibdev.query_gid) + if (!rdi->ibdev.ops.query_gid) if (!rdi->driver_f.get_guid_be) return -EINVAL; break; case CREATE_QP: - if (!rdi->ibdev.create_qp) + if (!rdi->ibdev.ops.create_qp) if (!rdi->driver_f.qp_priv_alloc || !rdi->driver_f.qp_priv_free || !rdi->driver_f.notify_qp_reset || @@ -491,7 +491,7 @@ static noinline int check_support(struct rvt_dev_info *rdi, int verb) break; case MODIFY_QP: - if (!rdi->ibdev.modify_qp) + if (!rdi->ibdev.ops.modify_qp) if (!rdi->driver_f.notify_qp_reset || !rdi->driver_f.schedule_send || !rdi->driver_f.get_pmtu_from_attr || @@ -505,7 +505,7 @@ static noinline int check_support(struct rvt_dev_info *rdi, int verb) break; case DESTROY_QP: - if (!rdi->ibdev.destroy_qp) + if (!rdi->ibdev.ops.destroy_qp) if (!rdi->driver_f.qp_priv_free || !rdi->driver_f.notify_qp_reset || !rdi->driver_f.flush_qp_waiters || @@ -515,7 +515,7 @@ static noinline int check_support(struct rvt_dev_info *rdi, int verb) break; case POST_SEND: - if (!rdi->ibdev.post_send) + if (!rdi->ibdev.ops.post_send) if (!rdi->driver_f.schedule_send || !rdi->driver_f.do_send || !rdi->post_parms) diff --git a/drivers/infiniband/ulp/ipoib/ipoib_main.c b/drivers/infiniband/ulp/ipoib/ipoib_main.c index 8710214594d8..5224c42f9d08 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_main.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c @@ -2453,8 +2453,8 @@ static struct net_device *ipoib_add_port(const char *format, return ERR_PTR(result); } - if (hca->rdma_netdev_get_params) { - int rc = hca->rdma_netdev_get_params(hca, port, + if (hca->ops.rdma_netdev_get_params) { + int rc = hca->ops.rdma_netdev_get_params(hca, port, RDMA_NETDEV_IPOIB, ¶ms); diff --git a/drivers/infiniband/ulp/iser/iser_memory.c b/drivers/infiniband/ulp/iser/iser_memory.c index dbe97c02848c..e9b7efc302d0 100644 --- a/drivers/infiniband/ulp/iser/iser_memory.c +++ b/drivers/infiniband/ulp/iser/iser_memory.c @@ -77,8 +77,8 @@ int iser_assign_reg_ops(struct iser_device *device) struct ib_device *ib_dev = device->ib_device; /* Assign function handles - based on FMR support */ - if (ib_dev->alloc_fmr && ib_dev->dealloc_fmr && - ib_dev->map_phys_fmr && ib_dev->unmap_fmr) { + if (ib_dev->ops.alloc_fmr && ib_dev->ops.dealloc_fmr && + ib_dev->ops.map_phys_fmr && ib_dev->ops.unmap_fmr) { iser_info("FMR supported, using FMR for registration\n"); device->reg_ops = &fmr_ops; } else if (ib_dev->attrs.device_cap_flags & IB_DEVICE_MEM_MGT_EXTENSIONS) { diff --git a/drivers/infiniband/ulp/opa_vnic/opa_vnic_netdev.c b/drivers/infiniband/ulp/opa_vnic/opa_vnic_netdev.c index 61558788b3fa..ae70cd18903e 100644 --- a/drivers/infiniband/ulp/opa_vnic/opa_vnic_netdev.c +++ b/drivers/infiniband/ulp/opa_vnic/opa_vnic_netdev.c @@ -330,10 +330,10 @@ struct opa_vnic_adapter *opa_vnic_add_netdev(struct ib_device *ibdev, struct rdma_netdev *rn; int rc; - netdev = ibdev->alloc_rdma_netdev(ibdev, port_num, - RDMA_NETDEV_OPA_VNIC, - "veth%d", NET_NAME_UNKNOWN, - ether_setup); + netdev = ibdev->ops.alloc_rdma_netdev(ibdev, port_num, + RDMA_NETDEV_OPA_VNIC, + "veth%d", NET_NAME_UNKNOWN, + ether_setup); if (!netdev) return ERR_PTR(-ENOMEM); else if (IS_ERR(netdev)) diff --git a/drivers/infiniband/ulp/srp/ib_srp.c b/drivers/infiniband/ulp/srp/ib_srp.c index eed0eb3bb04c..e58146d020bc 100644 --- a/drivers/infiniband/ulp/srp/ib_srp.c +++ b/drivers/infiniband/ulp/srp/ib_srp.c @@ -4063,8 +4063,10 @@ static void srp_add_one(struct ib_device *device) srp_dev->max_pages_per_mr = min_t(u64, SRP_MAX_PAGES_PER_MR, max_pages_per_mr); - srp_dev->has_fmr = (device->alloc_fmr && device->dealloc_fmr && - device->map_phys_fmr && device->unmap_fmr); + srp_dev->has_fmr = (device->ops.alloc_fmr && + device->ops.dealloc_fmr && + device->ops.map_phys_fmr && + device->ops.unmap_fmr); srp_dev->has_fr = (attr->device_cap_flags & IB_DEVICE_MEM_MGT_EXTENSIONS); if (!never_register && !srp_dev->has_fmr && !srp_dev->has_fr) { diff --git a/fs/cifs/smbdirect.c b/fs/cifs/smbdirect.c index e94a8d1d08a3..a568dac7b3a1 100644 --- a/fs/cifs/smbdirect.c +++ b/fs/cifs/smbdirect.c @@ -1724,7 +1724,7 @@ static struct smbd_connection *_smbd_get_connection( info->responder_resources); /* Need to send IRD/ORD in private data for iWARP */ - info->id->device->get_port_immutable( + info->id->device->ops.get_port_immutable( info->id->device, info->id->port_num, &port_immutable); if (port_immutable.core_cap_flags & RDMA_CORE_PORT_IWARP) { ird_ord_hdr[0] = info->responder_resources; diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h index 60315fd05411..5b3b51f00f48 100644 --- a/include/rdma/ib_verbs.h +++ b/include/rdma/ib_verbs.h @@ -2507,7 +2507,7 @@ struct ib_device_ops { struct ib_device { /* Do not access @dma_device directly from ULP nor from HW drivers. */ struct device *dma_device; - + struct ib_device_ops ops; char name[IB_DEVICE_NAME_MAX]; struct list_head event_handler_list; @@ -2532,273 +2532,6 @@ struct ib_device { struct iw_cm_verbs *iwcm; - /** - * alloc_hw_stats - Allocate a struct rdma_hw_stats and fill in the - * driver initialized data. The struct is kfree()'ed by the sysfs - * core when the device is removed. A lifespan of -1 in the return - * struct tells the core to set a default lifespan. - */ - struct rdma_hw_stats *(*alloc_hw_stats)(struct ib_device *device, - u8 port_num); - /** - * get_hw_stats - Fill in the counter value(s) in the stats struct. - * @index - The index in the value array we wish to have updated, or - * num_counters if we want all stats updated - * Return codes - - * < 0 - Error, no counters updated - * index - Updated the single counter pointed to by index - * num_counters - Updated all counters (will reset the timestamp - * and prevent further calls for lifespan milliseconds) - * Drivers are allowed to update all counters in lieu of just the - * one given in index at their option - */ - int (*get_hw_stats)(struct ib_device *device, - struct rdma_hw_stats *stats, - u8 port, int index); - int (*query_device)(struct ib_device *device, - struct ib_device_attr *device_attr, - struct ib_udata *udata); - int (*query_port)(struct ib_device *device, - u8 port_num, - struct ib_port_attr *port_attr); - enum rdma_link_layer (*get_link_layer)(struct ib_device *device, - u8 port_num); - /* When calling get_netdev, the HW vendor's driver should return the - * net device of device @device at port @port_num or NULL if such - * a net device doesn't exist. The vendor driver should call dev_hold - * on this net device. The HW vendor's device driver must guarantee - * that this function returns NULL before the net device has finished - * NETDEV_UNREGISTER state. - */ - struct net_device *(*get_netdev)(struct ib_device *device, - u8 port_num); - /* query_gid should be return GID value for @device, when @port_num - * link layer is either IB or iWarp. It is no-op if @port_num port - * is RoCE link layer. - */ - int (*query_gid)(struct ib_device *device, - u8 port_num, int index, - union ib_gid *gid); - /* When calling add_gid, the HW vendor's driver should add the gid - * of device of port at gid index available at @attr. Meta-info of - * that gid (for example, the network device related to this gid) is - * available at @attr. @context allows the HW vendor driver to store - * extra information together with a GID entry. The HW vendor driver may - * allocate memory to contain this information and store it in @context - * when a new GID entry is written to. Params are consistent until the - * next call of add_gid or delete_gid. The function should return 0 on - * success or error otherwise. The function could be called - * concurrently for different ports. This function is only called when - * roce_gid_table is used. - */ - int (*add_gid)(const struct ib_gid_attr *attr, - void **context); - /* When calling del_gid, the HW vendor's driver should delete the - * gid of device @device at gid index gid_index of port port_num - * available in @attr. - * Upon the deletion of a GID entry, the HW vendor must free any - * allocated memory. The caller will clear @context afterwards. - * This function is only called when roce_gid_table is used. - */ - int (*del_gid)(const struct ib_gid_attr *attr, - void **context); - int (*query_pkey)(struct ib_device *device, - u8 port_num, u16 index, u16 *pkey); - int (*modify_device)(struct ib_device *device, - int device_modify_mask, - struct ib_device_modify *device_modify); - int (*modify_port)(struct ib_device *device, - u8 port_num, int port_modify_mask, - struct ib_port_modify *port_modify); - struct ib_ucontext * (*alloc_ucontext)(struct ib_device *device, - struct ib_udata *udata); - int (*dealloc_ucontext)(struct ib_ucontext *context); - int (*mmap)(struct ib_ucontext *context, - struct vm_area_struct *vma); - struct ib_pd * (*alloc_pd)(struct ib_device *device, - struct ib_ucontext *context, - struct ib_udata *udata); - int (*dealloc_pd)(struct ib_pd *pd); - struct ib_ah * (*create_ah)(struct ib_pd *pd, - struct rdma_ah_attr *ah_attr, - struct ib_udata *udata); - int (*modify_ah)(struct ib_ah *ah, - struct rdma_ah_attr *ah_attr); - int (*query_ah)(struct ib_ah *ah, - struct rdma_ah_attr *ah_attr); - int (*destroy_ah)(struct ib_ah *ah); - struct ib_srq * (*create_srq)(struct ib_pd *pd, - struct ib_srq_init_attr *srq_init_attr, - struct ib_udata *udata); - int (*modify_srq)(struct ib_srq *srq, - struct ib_srq_attr *srq_attr, - enum ib_srq_attr_mask srq_attr_mask, - struct ib_udata *udata); - int (*query_srq)(struct ib_srq *srq, - struct ib_srq_attr *srq_attr); - int (*destroy_srq)(struct ib_srq *srq); - int (*post_srq_recv)(struct ib_srq *srq, - const struct ib_recv_wr *recv_wr, - const struct ib_recv_wr **bad_recv_wr); - struct ib_qp * (*create_qp)(struct ib_pd *pd, - struct ib_qp_init_attr *qp_init_attr, - struct ib_udata *udata); - int (*modify_qp)(struct ib_qp *qp, - struct ib_qp_attr *qp_attr, - int qp_attr_mask, - struct ib_udata *udata); - int (*query_qp)(struct ib_qp *qp, - struct ib_qp_attr *qp_attr, - int qp_attr_mask, - struct ib_qp_init_attr *qp_init_attr); - int (*destroy_qp)(struct ib_qp *qp); - int (*post_send)(struct ib_qp *qp, - const struct ib_send_wr *send_wr, - const struct ib_send_wr **bad_send_wr); - int (*post_recv)(struct ib_qp *qp, - const struct ib_recv_wr *recv_wr, - const struct ib_recv_wr **bad_recv_wr); - struct ib_cq * (*create_cq)(struct ib_device *device, - const struct ib_cq_init_attr *attr, - struct ib_ucontext *context, - struct ib_udata *udata); - int (*modify_cq)(struct ib_cq *cq, u16 cq_count, - u16 cq_period); - int (*destroy_cq)(struct ib_cq *cq); - int (*resize_cq)(struct ib_cq *cq, int cqe, - struct ib_udata *udata); - int (*poll_cq)(struct ib_cq *cq, int num_entries, - struct ib_wc *wc); - int (*peek_cq)(struct ib_cq *cq, int wc_cnt); - int (*req_notify_cq)(struct ib_cq *cq, - enum ib_cq_notify_flags flags); - int (*req_ncomp_notif)(struct ib_cq *cq, - int wc_cnt); - struct ib_mr * (*get_dma_mr)(struct ib_pd *pd, - int mr_access_flags); - struct ib_mr * (*reg_user_mr)(struct ib_pd *pd, - u64 start, u64 length, - u64 virt_addr, - int mr_access_flags, - struct ib_udata *udata); - int (*rereg_user_mr)(struct ib_mr *mr, - int flags, - u64 start, u64 length, - u64 virt_addr, - int mr_access_flags, - struct ib_pd *pd, - struct ib_udata *udata); - int (*dereg_mr)(struct ib_mr *mr); - struct ib_mr * (*alloc_mr)(struct ib_pd *pd, - enum ib_mr_type mr_type, - u32 max_num_sg); - int (*map_mr_sg)(struct ib_mr *mr, - struct scatterlist *sg, - int sg_nents, - unsigned int *sg_offset); - struct ib_mw * (*alloc_mw)(struct ib_pd *pd, - enum ib_mw_type type, - struct ib_udata *udata); - int (*dealloc_mw)(struct ib_mw *mw); - struct ib_fmr * (*alloc_fmr)(struct ib_pd *pd, - int mr_access_flags, - struct ib_fmr_attr *fmr_attr); - int (*map_phys_fmr)(struct ib_fmr *fmr, - u64 *page_list, int list_len, - u64 iova); - int (*unmap_fmr)(struct list_head *fmr_list); - int (*dealloc_fmr)(struct ib_fmr *fmr); - int (*attach_mcast)(struct ib_qp *qp, - union ib_gid *gid, - u16 lid); - int (*detach_mcast)(struct ib_qp *qp, - union ib_gid *gid, - u16 lid); - int (*process_mad)(struct ib_device *device, - int process_mad_flags, - u8 port_num, - const struct ib_wc *in_wc, - const struct ib_grh *in_grh, - const struct ib_mad_hdr *in_mad, - size_t in_mad_size, - struct ib_mad_hdr *out_mad, - size_t *out_mad_size, - u16 *out_mad_pkey_index); - struct ib_xrcd * (*alloc_xrcd)(struct ib_device *device, - struct ib_ucontext *ucontext, - struct ib_udata *udata); - int (*dealloc_xrcd)(struct ib_xrcd *xrcd); - struct ib_flow * (*create_flow)(struct ib_qp *qp, - struct ib_flow_attr - *flow_attr, - int domain, - struct ib_udata *udata); - int (*destroy_flow)(struct ib_flow *flow_id); - int (*check_mr_status)(struct ib_mr *mr, u32 check_mask, - struct ib_mr_status *mr_status); - void (*disassociate_ucontext)(struct ib_ucontext *ibcontext); - void (*drain_rq)(struct ib_qp *qp); - void (*drain_sq)(struct ib_qp *qp); - int (*set_vf_link_state)(struct ib_device *device, int vf, u8 port, - int state); - int (*get_vf_config)(struct ib_device *device, int vf, u8 port, - struct ifla_vf_info *ivf); - int (*get_vf_stats)(struct ib_device *device, int vf, u8 port, - struct ifla_vf_stats *stats); - int (*set_vf_guid)(struct ib_device *device, int vf, u8 port, u64 guid, - int type); - struct ib_wq * (*create_wq)(struct ib_pd *pd, - struct ib_wq_init_attr *init_attr, - struct ib_udata *udata); - int (*destroy_wq)(struct ib_wq *wq); - int (*modify_wq)(struct ib_wq *wq, - struct ib_wq_attr *attr, - u32 wq_attr_mask, - struct ib_udata *udata); - struct ib_rwq_ind_table * (*create_rwq_ind_table)(struct ib_device *device, - struct ib_rwq_ind_table_init_attr *init_attr, - struct ib_udata *udata); - int (*destroy_rwq_ind_table)(struct ib_rwq_ind_table *wq_ind_table); - struct ib_flow_action * (*create_flow_action_esp)(struct ib_device *device, - const struct ib_flow_action_attrs_esp *attr, - struct uverbs_attr_bundle *attrs); - int (*destroy_flow_action)(struct ib_flow_action *action); - int (*modify_flow_action_esp)(struct ib_flow_action *action, - const struct ib_flow_action_attrs_esp *attr, - struct uverbs_attr_bundle *attrs); - struct ib_dm * (*alloc_dm)(struct ib_device *device, - struct ib_ucontext *context, - struct ib_dm_alloc_attr *attr, - struct uverbs_attr_bundle *attrs); - int (*dealloc_dm)(struct ib_dm *dm); - struct ib_mr * (*reg_dm_mr)(struct ib_pd *pd, struct ib_dm *dm, - struct ib_dm_mr_attr *attr, - struct uverbs_attr_bundle *attrs); - struct ib_counters * (*create_counters)(struct ib_device *device, - struct uverbs_attr_bundle *attrs); - int (*destroy_counters)(struct ib_counters *counters); - int (*read_counters)(struct ib_counters *counters, - struct ib_counters_read_attr *counters_read_attr, - struct uverbs_attr_bundle *attrs); - - /** - * rdma netdev operation - * - * Driver implementing alloc_rdma_netdev or rdma_netdev_get_params - * must return -EOPNOTSUPP if it doesn't support the specified type. - */ - struct net_device *(*alloc_rdma_netdev)( - struct ib_device *device, - u8 port_num, - enum rdma_netdev_t type, - const char *name, - unsigned char name_assign_type, - void (*setup)(struct net_device *)); - - int (*rdma_netdev_get_params)(struct ib_device *device, u8 port_num, - enum rdma_netdev_t type, - struct rdma_netdev_alloc_params *params); - struct module *owner; struct device dev; /* First group for device attributes, @@ -2840,17 +2573,6 @@ struct ib_device { */ struct rdma_restrack_root res; - /** - * The following mandatory functions are used only at device - * registration. Keep functions such as these at the end of this - * structure to avoid cache line misses when accessing struct ib_device - * in fast paths. - */ - int (*get_port_immutable)(struct ib_device *, u8, struct ib_port_immutable *); - void (*get_dev_fw_str)(struct ib_device *, char *str); - const struct cpumask *(*get_vector_affinity)(struct ib_device *ibdev, - int comp_vector); - const struct uapi_definition *driver_def; enum rdma_driver_id driver_id; /* @@ -3365,7 +3087,7 @@ static inline bool rdma_cap_roce_gid_table(const struct ib_device *device, u8 port_num) { return rdma_protocol_roce(device, port_num) && - device->add_gid && device->del_gid; + device->ops.add_gid && device->ops.del_gid; } /* @@ -3589,7 +3311,8 @@ static inline int ib_post_srq_recv(struct ib_srq *srq, { const struct ib_recv_wr *dummy; - return srq->device->post_srq_recv(srq, recv_wr, bad_recv_wr ? : &dummy); + return srq->device->ops.post_srq_recv(srq, recv_wr, + bad_recv_wr ? : &dummy); } /** @@ -3692,7 +3415,7 @@ static inline int ib_post_send(struct ib_qp *qp, { const struct ib_send_wr *dummy; - return qp->device->post_send(qp, send_wr, bad_send_wr ? : &dummy); + return qp->device->ops.post_send(qp, send_wr, bad_send_wr ? : &dummy); } /** @@ -3709,7 +3432,7 @@ static inline int ib_post_recv(struct ib_qp *qp, { const struct ib_recv_wr *dummy; - return qp->device->post_recv(qp, recv_wr, bad_recv_wr ? : &dummy); + return qp->device->ops.post_recv(qp, recv_wr, bad_recv_wr ? : &dummy); } struct ib_cq *__ib_alloc_cq(struct ib_device *dev, void *private, @@ -3782,7 +3505,7 @@ int ib_destroy_cq(struct ib_cq *cq); static inline int ib_poll_cq(struct ib_cq *cq, int num_entries, struct ib_wc *wc) { - return cq->device->poll_cq(cq, num_entries, wc); + return cq->device->ops.poll_cq(cq, num_entries, wc); } /** @@ -3815,7 +3538,7 @@ static inline int ib_poll_cq(struct ib_cq *cq, int num_entries, static inline int ib_req_notify_cq(struct ib_cq *cq, enum ib_cq_notify_flags flags) { - return cq->device->req_notify_cq(cq, flags); + return cq->device->ops.req_notify_cq(cq, flags); } /** @@ -3827,8 +3550,8 @@ static inline int ib_req_notify_cq(struct ib_cq *cq, */ static inline int ib_req_ncomp_notif(struct ib_cq *cq, int wc_cnt) { - return cq->device->req_ncomp_notif ? - cq->device->req_ncomp_notif(cq, wc_cnt) : + return cq->device->ops.req_ncomp_notif ? + cq->device->ops.req_ncomp_notif(cq, wc_cnt) : -ENOSYS; } @@ -4092,7 +3815,7 @@ static inline int ib_map_phys_fmr(struct ib_fmr *fmr, u64 *page_list, int list_len, u64 iova) { - return fmr->device->map_phys_fmr(fmr, page_list, list_len, iova); + return fmr->device->ops.map_phys_fmr(fmr, page_list, list_len, iova); } /** @@ -4445,10 +4168,10 @@ static inline const struct cpumask * ib_get_vector_affinity(struct ib_device *device, int comp_vector) { if (comp_vector < 0 || comp_vector >= device->num_comp_vectors || - !device->get_vector_affinity) + !device->ops.get_vector_affinity) return NULL; - return device->get_vector_affinity(device, comp_vector); + return device->ops.get_vector_affinity(device, comp_vector); } diff --git a/include/rdma/uverbs_ioctl.h b/include/rdma/uverbs_ioctl.h index 7f83265f6298..839a857aa329 100644 --- a/include/rdma/uverbs_ioctl.h +++ b/include/rdma/uverbs_ioctl.h @@ -419,10 +419,10 @@ struct uapi_definition { .kind = UAPI_DEF_IS_SUPPORTED_DEV_FN, \ .scope = UAPI_SCOPE_OBJECT, \ .needs_fn_offset = \ - offsetof(struct ib_device, ibdev_fn) + \ + offsetof(struct ib_device_ops, ibdev_fn) + \ BUILD_BUG_ON_ZERO( \ - sizeof(((struct ib_device *)0)->ibdev_fn) != \ - sizeof(void *)), \ + sizeof(((struct ib_device_ops *)0)->ibdev_fn) != \ + sizeof(void *)), \ } /* @@ -434,10 +434,10 @@ struct uapi_definition { .kind = UAPI_DEF_IS_SUPPORTED_DEV_FN, \ .scope = UAPI_SCOPE_METHOD, \ .needs_fn_offset = \ - offsetof(struct ib_device, ibdev_fn) + \ + offsetof(struct ib_device_ops, ibdev_fn) + \ BUILD_BUG_ON_ZERO( \ - sizeof(((struct ib_device *)0)->ibdev_fn) != \ - sizeof(void *)), \ + sizeof(((struct ib_device_ops *)0)->ibdev_fn) != \ + sizeof(void *)), \ } /* Call a function to determine if the entire object is supported or not */ diff --git a/net/rds/ib.c b/net/rds/ib.c index eba75c1ba359..9d7b7586f240 100644 --- a/net/rds/ib.c +++ b/net/rds/ib.c @@ -148,8 +148,8 @@ static void rds_ib_add_one(struct ib_device *device) has_fr = (device->attrs.device_cap_flags & IB_DEVICE_MEM_MGT_EXTENSIONS); - has_fmr = (device->alloc_fmr && device->dealloc_fmr && - device->map_phys_fmr && device->unmap_fmr); + has_fmr = (device->ops.alloc_fmr && device->ops.dealloc_fmr && + device->ops.map_phys_fmr && device->ops.unmap_fmr); rds_ibdev->use_fastreg = (has_fr && !has_fmr); rds_ibdev->fmr_max_remaps = device->attrs.max_map_per_fmr?: 32; diff --git a/net/sunrpc/xprtrdma/fmr_ops.c b/net/sunrpc/xprtrdma/fmr_ops.c index 7f5632cd5a48..fd8fea59fe92 100644 --- a/net/sunrpc/xprtrdma/fmr_ops.c +++ b/net/sunrpc/xprtrdma/fmr_ops.c @@ -41,7 +41,7 @@ enum { bool fmr_is_supported(struct rpcrdma_ia *ia) { - if (!ia->ri_device->alloc_fmr) { + if (!ia->ri_device->ops.alloc_fmr) { pr_info("rpcrdma: 'fmr' mode is not supported by device %s\n", ia->ri_device->name); return false; -- cgit v1.2.3-59-g8ed1b From 8e3b688301863fba914883e3531b406c68cb7501 Mon Sep 17 00:00:00 2001 From: Leon Romanovsky Date: Wed, 12 Dec 2018 19:45:53 +0200 Subject: RDMA/mlx5: Delete unreachable handle_atomic code by simplifying SW completion Handle atomic was left as unimplemented from 2013, remove the code till this part will be developed. Remove the dead code by simplifying SW completion logic which is supposed to be the same for send and receive paths. Fixes: e126ba97dba9 ("mlx5: Add driver for Mellanox Connect-IB adapters") Reported-by: Stephen Rothwell Signed-off-by: Leon Romanovsky Tested-by: Stephen Rothwell # compile tested Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/cq.c | 102 ++--------------------------------- drivers/infiniband/hw/mlx5/mlx5_ib.h | 2 - drivers/infiniband/hw/mlx5/qp.c | 1 - 3 files changed, 5 insertions(+), 100 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/cq.c b/drivers/infiniband/hw/mlx5/cq.c index 0b99f7d0630d..95a29e85522e 100644 --- a/drivers/infiniband/hw/mlx5/cq.c +++ b/drivers/infiniband/hw/mlx5/cq.c @@ -330,67 +330,6 @@ static void mlx5_handle_error_cqe(struct mlx5_ib_dev *dev, dump_cqe(dev, cqe); } -static int is_atomic_response(struct mlx5_ib_qp *qp, uint16_t idx) -{ - /* TBD: waiting decision - */ - return 0; -} - -static void *mlx5_get_atomic_laddr(struct mlx5_ib_qp *qp, uint16_t idx) -{ - struct mlx5_wqe_data_seg *dpseg; - void *addr; - - dpseg = mlx5_get_send_wqe(qp, idx) + sizeof(struct mlx5_wqe_ctrl_seg) + - sizeof(struct mlx5_wqe_raddr_seg) + - sizeof(struct mlx5_wqe_atomic_seg); - addr = (void *)(unsigned long)be64_to_cpu(dpseg->addr); - return addr; -} - -static void handle_atomic(struct mlx5_ib_qp *qp, struct mlx5_cqe64 *cqe64, - uint16_t idx) -{ - void *addr; - int byte_count; - int i; - - if (!is_atomic_response(qp, idx)) - return; - - byte_count = be32_to_cpu(cqe64->byte_cnt); - addr = mlx5_get_atomic_laddr(qp, idx); - - if (byte_count == 4) { - *(uint32_t *)addr = be32_to_cpu(*((__be32 *)addr)); - } else { - for (i = 0; i < byte_count; i += 8) { - *(uint64_t *)addr = be64_to_cpu(*((__be64 *)addr)); - addr += 8; - } - } - - return; -} - -static void handle_atomics(struct mlx5_ib_qp *qp, struct mlx5_cqe64 *cqe64, - u16 tail, u16 head) -{ - u16 idx; - - do { - idx = tail & (qp->sq.wqe_cnt - 1); - handle_atomic(qp, cqe64, idx); - if (idx == head) - break; - - tail = qp->sq.w_list[idx].next; - } while (1); - tail = qp->sq.w_list[idx].next; - qp->sq.last_poll = tail; -} - static void free_cq_buf(struct mlx5_ib_dev *dev, struct mlx5_ib_cq_buf *buf) { mlx5_frag_buf_free(dev->mdev, &buf->frag_buf); @@ -428,45 +367,15 @@ static void get_sig_err_item(struct mlx5_sig_err_cqe *cqe, item->key = be32_to_cpu(cqe->mkey); } -static void sw_send_comp(struct mlx5_ib_qp *qp, int num_entries, - struct ib_wc *wc, int *npolled) -{ - struct mlx5_ib_wq *wq; - unsigned int cur; - unsigned int idx; - int np; - int i; - - wq = &qp->sq; - cur = wq->head - wq->tail; - np = *npolled; - - if (cur == 0) - return; - - for (i = 0; i < cur && np < num_entries; i++) { - idx = wq->last_poll & (wq->wqe_cnt - 1); - wc->wr_id = wq->wrid[idx]; - wc->status = IB_WC_WR_FLUSH_ERR; - wc->vendor_err = MLX5_CQE_SYNDROME_WR_FLUSH_ERR; - wq->tail++; - np++; - wc->qp = &qp->ibqp; - wc++; - wq->last_poll = wq->w_list[idx].next; - } - *npolled = np; -} - -static void sw_recv_comp(struct mlx5_ib_qp *qp, int num_entries, - struct ib_wc *wc, int *npolled) +static void sw_comp(struct mlx5_ib_qp *qp, int num_entries, struct ib_wc *wc, + int *npolled, int is_send) { struct mlx5_ib_wq *wq; unsigned int cur; int np; int i; - wq = &qp->rq; + wq = (is_send) ? &qp->sq : &qp->rq; cur = wq->head - wq->tail; np = *npolled; @@ -493,13 +402,13 @@ static void mlx5_ib_poll_sw_comp(struct mlx5_ib_cq *cq, int num_entries, *npolled = 0; /* Find uncompleted WQEs belonging to that cq and return mmics ones */ list_for_each_entry(qp, &cq->list_send_qp, cq_send_list) { - sw_send_comp(qp, num_entries, wc + *npolled, npolled); + sw_comp(qp, num_entries, wc + *npolled, npolled, true); if (*npolled >= num_entries) return; } list_for_each_entry(qp, &cq->list_recv_qp, cq_recv_list) { - sw_recv_comp(qp, num_entries, wc + *npolled, npolled); + sw_comp(qp, num_entries, wc + *npolled, npolled, false); if (*npolled >= num_entries) return; } @@ -567,7 +476,6 @@ repoll: wqe_ctr = be16_to_cpu(cqe64->wqe_counter); idx = wqe_ctr & (wq->wqe_cnt - 1); handle_good_req(wc, cqe64, wq, idx); - handle_atomics(*cur_qp, cqe64, wq->last_poll, idx); wc->wr_id = wq->wrid[idx]; wq->tail = wq->wqe_head[idx] + 1; wc->status = IB_WC_SUCCESS; diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index 8fd3de05cdba..1285ac11bb70 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -275,7 +275,6 @@ struct mlx5_ib_wq { unsigned head; unsigned tail; u16 cur_post; - u16 last_poll; void *cur_edge; }; @@ -1070,7 +1069,6 @@ int mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr, const struct ib_send_wr **bad_wr); int mlx5_ib_post_recv(struct ib_qp *ibqp, const struct ib_recv_wr *wr, const struct ib_recv_wr **bad_wr); -void *mlx5_get_send_wqe(struct mlx5_ib_qp *qp, int n); int mlx5_ib_read_user_wqe(struct mlx5_ib_qp *qp, int send, int wqe_index, void *buffer, u32 length, struct mlx5_ib_qp_base *base); diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c index 4a85748aeeea..2ceb8f68f605 100644 --- a/drivers/infiniband/hw/mlx5/qp.c +++ b/drivers/infiniband/hw/mlx5/qp.c @@ -3516,7 +3516,6 @@ static int __mlx5_ib_modify_qp(struct ib_qp *ibqp, qp->sq.cur_post = 0; if (qp->sq.wqe_cnt) qp->sq.cur_edge = get_sq_edge(&qp->sq, 0); - qp->sq.last_poll = 0; qp->db.db[MLX5_RCV_DBR] = 0; qp->db.db[MLX5_SND_DBR] = 0; } -- cgit v1.2.3-59-g8ed1b From b81a327dbc0d07c219ac48b7daf155ea6a51672b Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Tue, 18 Dec 2018 22:55:07 +0200 Subject: RDMA/i40iw: Make sure to initialize ib_device_ops The initialization of the ib_device_ops was dropped by mistake when rebasing the ib_device_ops series, this patch fixes that. Fixes: 15644f57cb66 ("RDMA/i40iw: Initialize ib_device_ops struct") Signed-off-by: Kamal Heib Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/i40iw/i40iw_verbs.c | 1 + 1 file changed, 1 insertion(+) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/i40iw/i40iw_verbs.c b/drivers/infiniband/hw/i40iw/i40iw_verbs.c index 9e42ac2db3ca..475f32cf447e 100644 --- a/drivers/infiniband/hw/i40iw/i40iw_verbs.c +++ b/drivers/infiniband/hw/i40iw/i40iw_verbs.c @@ -2816,6 +2816,7 @@ static struct i40iw_ib_device *i40iw_init_rdma_device(struct i40iw_device *iwdev iwibdev->ibdev.iwcm->destroy_listen = i40iw_destroy_listen; memcpy(iwibdev->ibdev.iwcm->ifname, netdev->name, sizeof(iwibdev->ibdev.iwcm->ifname)); + ib_set_device_ops(&iwibdev->ibdev, &i40iw_dev_ops); return iwibdev; } -- cgit v1.2.3-59-g8ed1b From ac2f7e623d47fd68d2517f191696fbc96a5d7270 Mon Sep 17 00:00:00 2001 From: Gal Pressman Date: Tue, 18 Dec 2018 17:57:32 +0200 Subject: RDMA/mlx5: Fix function name typo 'fileds' -> 'fields' Fix typo in 'set_mr_fileds' -> 'set_mr_fields'. Signed-off-by: Gal Pressman Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/mr.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c index 9b195d65a13e..8e072f84a323 100644 --- a/drivers/infiniband/hw/mlx5/mr.c +++ b/drivers/infiniband/hw/mlx5/mr.c @@ -1211,7 +1211,7 @@ err_1: return ERR_PTR(err); } -static void set_mr_fileds(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr, +static void set_mr_fields(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr, int npages, u64 length, int access_flags) { mr->npages = npages; @@ -1267,7 +1267,7 @@ static struct ib_mr *mlx5_ib_get_memic_mr(struct ib_pd *pd, u64 memic_addr, kfree(in); mr->umem = NULL; - set_mr_fileds(dev, mr, 0, length, acc); + set_mr_fields(dev, mr, 0, length, acc); return &mr->ibmr; @@ -1369,7 +1369,7 @@ struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length, mlx5_ib_dbg(dev, "mkey 0x%x\n", mr->mmkey.key); mr->umem = umem; - set_mr_fileds(dev, mr, npages, length, access_flags); + set_mr_fields(dev, mr, npages, length, access_flags); #ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING update_odp_mr(mr); @@ -1536,7 +1536,7 @@ int mlx5_ib_rereg_user_mr(struct ib_mr *ib_mr, int flags, u64 start, goto err; } - set_mr_fileds(dev, mr, npages, len, access_flags); + set_mr_fields(dev, mr, npages, len, access_flags); #ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING update_odp_mr(mr); -- cgit v1.2.3-59-g8ed1b From 5050ae5fa3d54c8e83e1e447cc7e3591110a7f57 Mon Sep 17 00:00:00 2001 From: Dan Carpenter Date: Mon, 17 Dec 2018 10:05:36 +0300 Subject: IB/qib: Fix an error code in qib_sdma_verbs_send() We accidentally return success on this error path. Fixes: f931551bafe1 ("IB/qib: Add new qib driver for QLogic PCIe InfiniBand adapters") Signed-off-by: Dan Carpenter Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/qib/qib_sdma.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/qib/qib_sdma.c b/drivers/infiniband/hw/qib/qib_sdma.c index 0b6ca424c11d..3d64081c4819 100644 --- a/drivers/infiniband/hw/qib/qib_sdma.c +++ b/drivers/infiniband/hw/qib/qib_sdma.c @@ -575,8 +575,10 @@ retry: dw = (len + 3) >> 2; addr = dma_map_single(&ppd->dd->pcidev->dev, sge->vaddr, dw << 2, DMA_TO_DEVICE); - if (dma_mapping_error(&ppd->dd->pcidev->dev, addr)) + if (dma_mapping_error(&ppd->dd->pcidev->dev, addr)) { + ret = -ENOMEM; goto unmap; + } sdmadesc[0] = 0; make_sdma_desc(ppd, sdmadesc, (u64) addr, dw, dwoffset); /* SDmaUseLargeBuf has to be set in every descriptor */ -- cgit v1.2.3-59-g8ed1b From e9dfa53a399545056cf97cf8c996d71dcebb4d7e Mon Sep 17 00:00:00 2001 From: Dan Carpenter Date: Mon, 17 Dec 2018 10:08:15 +0300 Subject: RDMA/hns: Fix an error code in hns_roce_create_srq() The function accidentally returns success on this error path. Fixes: c7bcb13442e1 ("RDMA/hns: Add SRQ support for hip08 kernel mode") Signed-off-by: Dan Carpenter Acked-by: Lijun Ou Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hns/hns_roce_srq.c | 1 + 1 file changed, 1 insertion(+) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/hns/hns_roce_srq.c b/drivers/infiniband/hw/hns/hns_roce_srq.c index 463df60094e8..6377e734e28e 100644 --- a/drivers/infiniband/hw/hns/hns_roce_srq.c +++ b/drivers/infiniband/hw/hns/hns_roce_srq.c @@ -286,6 +286,7 @@ struct ib_srq *hns_roce_create_srq(struct ib_pd *pd, if (IS_ERR(srq->idx_que.umem)) { dev_err(hr_dev->dev, "ib_umem_get error for index queue\n"); + ret = PTR_ERR(srq->idx_que.umem); goto err_srq_mtt; } -- cgit v1.2.3-59-g8ed1b From 350b4c8ac1b4c07909ab7f8ad4303fff2505ab7c Mon Sep 17 00:00:00 2001 From: Yuval Shaia Date: Tue, 11 Dec 2018 10:36:47 +0200 Subject: IB/mlx4: Utilize macro to calculate SQ spare size The macro MLX4_IB_SQ_HEADROOM calculates the spare room needed to be left. Use it instead of hard-coding the HW prefetch size. Signed-off-by: Yuval Shaia Reviewed-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx4/qp.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c index 74ef7a0ec0b7..651d8594dc3f 100644 --- a/drivers/infiniband/hw/mlx4/qp.c +++ b/drivers/infiniband/hw/mlx4/qp.c @@ -401,7 +401,7 @@ static int set_kernel_sq_size(struct mlx4_ib_dev *dev, struct ib_qp_cap *cap, * 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) + 1; + qp->sq_spare_wqes = MLX4_IB_SQ_HEADROOM(qp->sq.wqe_shift); qp->sq.wqe_cnt = roundup_pow_of_two(cap->max_send_wr + qp->sq_spare_wqes); -- cgit v1.2.3-59-g8ed1b From cbfdd442c43eab8c62bf2ea5127511cd39e9046d Mon Sep 17 00:00:00 2001 From: Moni Shoua Date: Tue, 11 Dec 2018 13:37:51 +0200 Subject: IB/uverbs: Add helper to get array size from ptr attribute When the parser of an ioctl command has the knowledge that a ptr attribute in a bundle represents an array of structures, it is useful for it to know the number of elements in the array. This is done by dividing the attribute length with the element size. Signed-off-by: Moni Shoua Reviewed-by: Guy Levi Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/flow.c | 12 +++++------- include/rdma/uverbs_ioctl.h | 22 ++++++++++++++++++++++ 2 files changed, 27 insertions(+), 7 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/flow.c b/drivers/infiniband/hw/mlx5/flow.c index b840d4eeba84..95cf4b9fc65d 100644 --- a/drivers/infiniband/hw/mlx5/flow.c +++ b/drivers/infiniband/hw/mlx5/flow.c @@ -331,7 +331,6 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_FLOW_ACTION_CREATE_MODIFY_HEADER)( struct ib_flow_action *action; size_t num_actions; void *in; - int len; int ret; if (!mlx5_ib_modify_header_supported(mdev)) @@ -339,18 +338,17 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_FLOW_ACTION_CREATE_MODIFY_HEADER)( in = uverbs_attr_get_alloced_ptr(attrs, MLX5_IB_ATTR_CREATE_MODIFY_HEADER_ACTIONS_PRM); - len = uverbs_attr_get_len(attrs, - MLX5_IB_ATTR_CREATE_MODIFY_HEADER_ACTIONS_PRM); - if (len % MLX5_UN_SZ_BYTES(set_action_in_add_action_in_auto)) - return -EINVAL; + num_actions = uverbs_attr_ptr_get_array_size( + attrs, MLX5_IB_ATTR_CREATE_MODIFY_HEADER_ACTIONS_PRM, + MLX5_UN_SZ_BYTES(set_action_in_add_action_in_auto)); + if (num_actions < 0) + return num_actions; ret = uverbs_get_const(&ft_type, attrs, MLX5_IB_ATTR_CREATE_MODIFY_HEADER_FT_TYPE); if (ret) return ret; - - num_actions = len / MLX5_UN_SZ_BYTES(set_action_in_add_action_in_auto), action = mlx5_ib_create_modify_header(mdev, ft_type, num_actions, in); if (IS_ERR(action)) return PTR_ERR(action); diff --git a/include/rdma/uverbs_ioctl.h b/include/rdma/uverbs_ioctl.h index 839a857aa329..bf07cd6336d3 100644 --- a/include/rdma/uverbs_ioctl.h +++ b/include/rdma/uverbs_ioctl.h @@ -719,6 +719,28 @@ uverbs_attr_get_len(const struct uverbs_attr_bundle *attrs_bundle, u16 idx) return attr->ptr_attr.len; } +/* + * uverbs_attr_ptr_get_array_size() - Get array size pointer by a ptr + * attribute. + * @attrs: The attribute bundle + * @idx: The ID of the attribute + * @elem_size: The size of the element in the array + */ +static inline int +uverbs_attr_ptr_get_array_size(struct uverbs_attr_bundle *attrs, u16 idx, + size_t elem_size) +{ + int size = uverbs_attr_get_len(attrs, idx); + + if (size < 0) + return size; + + if (size % elem_size) + return -EINVAL; + + return size / elem_size; +} + /** * uverbs_attr_get_uobjs_arr() - Provides array's properties for attribute for * UVERBS_ATTR_TYPE_IDRS_ARRAY. -- cgit v1.2.3-59-g8ed1b From 813e90b1aeaa550641332625174d57edb15bc8bd Mon Sep 17 00:00:00 2001 From: Moni Shoua Date: Tue, 11 Dec 2018 13:37:53 +0200 Subject: IB/mlx5: Add advise_mr() support The verb advise_mr() is used to give advice to the kernel about an address range that belongs to a MR. Implement the verb and register it on the device. The current implementation supports the only known advice to date, prefetch. Signed-off-by: Moni Shoua Reviewed-by: Guy Levi Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/main.c | 8 +++ drivers/infiniband/hw/mlx5/mlx5_ib.h | 18 +++++ drivers/infiniband/hw/mlx5/mr.c | 15 +++++ drivers/infiniband/hw/mlx5/odp.c | 125 ++++++++++++++++++++++++++++++++--- 4 files changed, 157 insertions(+), 9 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index 5a4e23105b0c..d7b56222fea3 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -5712,6 +5712,8 @@ void mlx5_ib_stage_init_cleanup(struct mlx5_ib_dev *dev) mlx5_ib_cleanup_multiport_master(dev); #ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING cleanup_srcu_struct(&dev->mr_srcu); + drain_workqueue(dev->advise_mr_wq); + destroy_workqueue(dev->advise_mr_wq); #endif kfree(dev->port); } @@ -5766,6 +5768,12 @@ int mlx5_ib_stage_init_init(struct mlx5_ib_dev *dev) dev->memic.dev = mdev; #ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING + dev->advise_mr_wq = alloc_ordered_workqueue("mlx5_ib_advise_mr_wq", 0); + if (!dev->advise_mr_wq) { + err = -ENOMEM; + goto err_free_port; + } + err = init_srcu_struct(&dev->mr_srcu); if (err) goto err_free_port; diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index 1285ac11bb70..f245b5d8a3bc 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -923,6 +923,7 @@ struct mlx5_ib_dev { */ struct srcu_struct mr_srcu; u32 null_mkey; + struct workqueue_struct *advise_mr_wq; #endif struct mlx5_ib_flow_db *flow_db; /* protect resources needed as part of reset flow */ @@ -1085,6 +1086,12 @@ struct ib_mr *mlx5_ib_get_dma_mr(struct ib_pd *pd, int acc); struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length, u64 virt_addr, int access_flags, struct ib_udata *udata); +int mlx5_ib_advise_mr(struct ib_pd *pd, + enum ib_uverbs_advise_mr_advice advice, + u32 flags, + struct ib_sge *sg_list, + u32 num_sge, + struct uverbs_attr_bundle *attrs); struct ib_mw *mlx5_ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type, struct ib_udata *udata); int mlx5_ib_dealloc_mw(struct ib_mw *mw); @@ -1182,6 +1189,10 @@ void mlx5_ib_invalidate_range(struct ib_umem_odp *umem_odp, unsigned long start, void mlx5_odp_init_mr_cache_entry(struct mlx5_cache_ent *ent); void mlx5_odp_populate_klm(struct mlx5_klm *pklm, size_t offset, size_t nentries, struct mlx5_ib_mr *mr, int flags); + +int mlx5_ib_advise_mr_prefetch(struct ib_pd *pd, + enum ib_uverbs_advise_mr_advice advice, + u32 flags, struct ib_sge *sg_list, u32 num_sge); #else /* CONFIG_INFINIBAND_ON_DEMAND_PAGING */ static inline void mlx5_ib_internal_fill_odp_caps(struct mlx5_ib_dev *dev) { @@ -1197,6 +1208,13 @@ static inline void mlx5_odp_populate_klm(struct mlx5_klm *pklm, size_t offset, size_t nentries, struct mlx5_ib_mr *mr, int flags) {} +static int mlx5_ib_advise_mr_prefetch(struct ib_pd *pd, + enum ib_uverbs_advise_mr_advice advice, + u32 flags, struct ib_sge *sg_list, + u32 num_sge) +{ + return -EOPNOTSUPP; +} #endif /* CONFIG_INFINIBAND_ON_DEMAND_PAGING */ /* Needed for rep profile */ diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c index 8e072f84a323..fd6ea1f75085 100644 --- a/drivers/infiniband/hw/mlx5/mr.c +++ b/drivers/infiniband/hw/mlx5/mr.c @@ -1280,6 +1280,21 @@ err_free: return ERR_PTR(err); } +int mlx5_ib_advise_mr(struct ib_pd *pd, + enum ib_uverbs_advise_mr_advice advice, + u32 flags, + struct ib_sge *sg_list, + u32 num_sge, + struct uverbs_attr_bundle *attrs) +{ + if (advice != IB_UVERBS_ADVISE_MR_ADVICE_PREFETCH && + advice != IB_UVERBS_ADVISE_MR_ADVICE_PREFETCH_WRITE) + return -EOPNOTSUPP; + + return mlx5_ib_advise_mr_prefetch(pd, advice, flags, + sg_list, num_sge); +} + struct ib_mr *mlx5_ib_reg_dm_mr(struct ib_pd *pd, struct ib_dm *dm, struct ib_dm_mr_attr *attr, struct uverbs_attr_bundle *attrs) diff --git a/drivers/infiniband/hw/mlx5/odp.c b/drivers/infiniband/hw/mlx5/odp.c index 4ead8c0fff5a..80fa2438db8f 100644 --- a/drivers/infiniband/hw/mlx5/odp.c +++ b/drivers/infiniband/hw/mlx5/odp.c @@ -549,10 +549,15 @@ void mlx5_ib_free_implicit_mr(struct mlx5_ib_mr *imr) wait_event(imr->q_leaf_free, !atomic_read(&imr->num_leaf_free)); } +#define MLX5_PF_FLAGS_PREFETCH BIT(0) +#define MLX5_PF_FLAGS_DOWNGRADE BIT(1) static int pagefault_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr, - u64 io_virt, size_t bcnt, u32 *bytes_mapped) + u64 io_virt, size_t bcnt, u32 *bytes_mapped, + u32 flags) { struct ib_umem_odp *odp_mr = to_ib_umem_odp(mr->umem); + bool downgrade = flags & MLX5_PF_FLAGS_DOWNGRADE; + bool prefetch = flags & MLX5_PF_FLAGS_PREFETCH; u64 access_mask = ODP_READ_ALLOWED_BIT; int npages = 0, page_shift, np; u64 start_idx, page_mask; @@ -579,7 +584,15 @@ next_mr: page_mask = ~(BIT(page_shift) - 1); start_idx = (io_virt - (mr->mmkey.iova & page_mask)) >> page_shift; - if (mr->umem->writable) + if (prefetch && !downgrade && !mr->umem->writable) { + /* prefetch with write-access must + * be supported by the MR + */ + ret = -EINVAL; + goto out; + } + + if (mr->umem->writable && !downgrade) access_mask |= ODP_WRITE_ALLOWED_BIT; current_seq = READ_ONCE(odp->notifiers_seq); @@ -684,12 +697,13 @@ struct pf_frame { * -EFAULT when there's an error mapping the requested pages. The caller will * abort the page fault handling. */ -static int pagefault_single_data_segment(struct mlx5_ib_dev *dev, - u32 key, u64 io_virt, size_t bcnt, +static int pagefault_single_data_segment(struct mlx5_ib_dev *dev, u32 key, + u64 io_virt, size_t bcnt, u32 *bytes_committed, - u32 *bytes_mapped) + u32 *bytes_mapped, u32 flags) { int npages = 0, srcu_key, ret, i, outlen, cur_outlen = 0, depth = 0; + bool prefetch = flags & MLX5_PF_FLAGS_PREFETCH; struct pf_frame *head = NULL, *frame; struct mlx5_core_mkey *mmkey; struct mlx5_ib_mw *mw; @@ -711,6 +725,12 @@ next_mr: goto srcu_unlock; } + if (prefetch && mmkey->type != MLX5_MKEY_MR) { + mlx5_ib_dbg(dev, "prefetch is allowed only for MR\n"); + ret = -EINVAL; + goto srcu_unlock; + } + switch (mmkey->type) { case MLX5_MKEY_MR: mr = container_of(mmkey, struct mlx5_ib_mr, mmkey); @@ -720,6 +740,11 @@ next_mr: goto srcu_unlock; } + if (prefetch && !mr->umem->is_odp) { + ret = -EINVAL; + goto srcu_unlock; + } + if (!mr->umem->is_odp) { mlx5_ib_dbg(dev, "skipping non ODP MR (lkey=0x%06x) in page fault handler.\n", key); @@ -729,7 +754,7 @@ next_mr: goto srcu_unlock; } - ret = pagefault_mr(dev, mr, io_virt, bcnt, bytes_mapped); + ret = pagefault_mr(dev, mr, io_virt, bcnt, bytes_mapped, flags); if (ret < 0) goto srcu_unlock; @@ -906,7 +931,7 @@ static int pagefault_data_segments(struct mlx5_ib_dev *dev, ret = pagefault_single_data_segment(dev, key, io_virt, bcnt, &pfault->bytes_committed, - bytes_mapped); + bytes_mapped, 0); if (ret < 0) break; npages += ret; @@ -1217,7 +1242,8 @@ static void mlx5_ib_mr_rdma_pfault_handler(struct mlx5_ib_dev *dev, } ret = pagefault_single_data_segment(dev, rkey, address, length, - &pfault->bytes_committed, NULL); + &pfault->bytes_committed, NULL, + 0); if (ret == -EAGAIN) { /* We're racing with an invalidation, don't prefetch */ prefetch_activated = 0; @@ -1244,7 +1270,8 @@ static void mlx5_ib_mr_rdma_pfault_handler(struct mlx5_ib_dev *dev, ret = pagefault_single_data_segment(dev, rkey, address, prefetch_len, - &bytes_committed, NULL); + &bytes_committed, NULL, + 0); if (ret < 0 && ret != -EAGAIN) { mlx5_ib_dbg(dev, "Prefetch failed. ret: %d, QP 0x%x, address: 0x%.16llx, length = 0x%.16x\n", ret, pfault->token, address, prefetch_len); @@ -1493,10 +1520,17 @@ void mlx5_odp_init_mr_cache_entry(struct mlx5_cache_ent *ent) } } +static const struct ib_device_ops mlx5_ib_dev_odp_ops = { + .advise_mr = mlx5_ib_advise_mr, +}; + int mlx5_ib_odp_init_one(struct mlx5_ib_dev *dev) { int ret = 0; + if (dev->odp_caps.general_caps & IB_ODP_SUPPORT) + ib_set_device_ops(&dev->ib_dev, &mlx5_ib_dev_odp_ops); + if (dev->odp_caps.general_caps & IB_ODP_SUPPORT_IMPLICIT) { ret = mlx5_cmd_null_mkey(dev->mdev, &dev->null_mkey); if (ret) { @@ -1528,3 +1562,76 @@ int mlx5_ib_odp_init(void) return 0; } + +struct prefetch_mr_work { + struct work_struct work; + struct mlx5_ib_dev *dev; + u32 pf_flags; + u32 num_sge; + struct ib_sge sg_list[0]; +}; + +static int mlx5_ib_prefetch_sg_list(struct mlx5_ib_dev *dev, u32 pf_flags, + struct ib_sge *sg_list, u32 num_sge) +{ + int i; + + for (i = 0; i < num_sge; ++i) { + struct ib_sge *sg = &sg_list[i]; + int bytes_committed = 0; + int ret; + + ret = pagefault_single_data_segment(dev, sg->lkey, sg->addr, + sg->length, + &bytes_committed, NULL, + pf_flags); + if (ret < 0) + return ret; + } + return 0; +} + +static void mlx5_ib_prefetch_mr_work(struct work_struct *work) +{ + struct prefetch_mr_work *w = + container_of(work, struct prefetch_mr_work, work); + + if (w->dev->ib_dev.reg_state == IB_DEV_REGISTERED) + mlx5_ib_prefetch_sg_list(w->dev, w->pf_flags, w->sg_list, + w->num_sge); + + kfree(w); +} + +int mlx5_ib_advise_mr_prefetch(struct ib_pd *pd, + enum ib_uverbs_advise_mr_advice advice, + u32 flags, struct ib_sge *sg_list, u32 num_sge) +{ + struct mlx5_ib_dev *dev = to_mdev(pd->device); + u32 pf_flags = MLX5_PF_FLAGS_PREFETCH; + struct prefetch_mr_work *work; + + if (advice == IB_UVERBS_ADVISE_MR_ADVICE_PREFETCH) + pf_flags |= MLX5_PF_FLAGS_DOWNGRADE; + + if (flags & IB_UVERBS_ADVISE_MR_FLAG_FLUSH) + return mlx5_ib_prefetch_sg_list(dev, pf_flags, sg_list, + num_sge); + + if (dev->ib_dev.reg_state != IB_DEV_REGISTERED) + return -ENODEV; + + work = kvzalloc(struct_size(work, sg_list, num_sge), GFP_KERNEL); + if (!work) + return -ENOMEM; + + memcpy(work->sg_list, sg_list, num_sge * sizeof(struct ib_sge)); + + work->dev = dev; + work->pf_flags = pf_flags; + work->num_sge = num_sge; + + INIT_WORK(&work->work, mlx5_ib_prefetch_mr_work); + schedule_work(&work->work); + return 0; +} -- cgit v1.2.3-59-g8ed1b From e00b64f7c54c4cbd88143bbd43e7c3d61a090e5c Mon Sep 17 00:00:00 2001 From: Shamir Rabinovitch Date: Mon, 17 Dec 2018 17:15:18 +0200 Subject: RDMA: Cleanup undesired pd->uobject usage Drivers should be using udata to determine if a method is invoked from user space or kernel space. A pd does not necessarily say a different objects is kernel or user. Transforming the tests to use udata eliminates a large number of uobject references from the drivers. Signed-off-by: Shamir Rabinovitch Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/bnxt_re/ib_verbs.c | 4 +-- drivers/infiniband/hw/cxgb3/iwch_provider.c | 2 +- drivers/infiniband/hw/cxgb4/qp.c | 4 +-- drivers/infiniband/hw/hns/hns_roce_hw_v1.c | 2 +- drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 4 +-- drivers/infiniband/hw/hns/hns_roce_qp.c | 18 +++++------ drivers/infiniband/hw/hns/hns_roce_srq.c | 2 +- drivers/infiniband/hw/i40iw/i40iw_verbs.c | 45 +++++++++++++-------------- drivers/infiniband/hw/mlx4/qp.c | 20 ++++++------ drivers/infiniband/hw/mlx4/srq.c | 10 +++--- drivers/infiniband/hw/mlx5/qp.c | 10 +++--- drivers/infiniband/hw/mlx5/srq.c | 8 ++--- drivers/infiniband/hw/mthca/mthca_dev.h | 9 ++++-- drivers/infiniband/hw/mthca/mthca_provider.c | 16 +++++----- drivers/infiniband/hw/mthca/mthca_qp.c | 22 +++++++------ drivers/infiniband/hw/mthca/mthca_srq.c | 29 +++++++++-------- drivers/infiniband/hw/nes/nes_verbs.c | 4 +-- drivers/infiniband/hw/ocrdma/ocrdma_verbs.c | 7 +++-- drivers/infiniband/hw/qedr/verbs.c | 9 +++--- drivers/infiniband/hw/vmw_pvrdma/pvrdma_qp.c | 2 +- drivers/infiniband/hw/vmw_pvrdma/pvrdma_srq.c | 2 +- drivers/infiniband/sw/rxe/rxe_loc.h | 2 +- drivers/infiniband/sw/rxe/rxe_qp.c | 5 +-- drivers/infiniband/sw/rxe/rxe_verbs.c | 2 +- 24 files changed, 125 insertions(+), 113 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.c b/drivers/infiniband/hw/bnxt_re/ib_verbs.c index 54fdd4cf5288..611bacd00b80 100644 --- a/drivers/infiniband/hw/bnxt_re/ib_verbs.c +++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.c @@ -698,7 +698,7 @@ struct ib_ah *bnxt_re_create_ah(struct ib_pd *ib_pd, ah->qplib_ah.flow_label = grh->flow_label; ah->qplib_ah.hop_limit = grh->hop_limit; ah->qplib_ah.sl = rdma_ah_get_sl(ah_attr); - if (ib_pd->uobject && + if (udata && !rdma_is_multicast_addr((struct in6_addr *) grh->dgid.raw) && !rdma_link_local_addr((struct in6_addr *) @@ -729,7 +729,7 @@ struct ib_ah *bnxt_re_create_ah(struct ib_pd *ib_pd, } /* Write AVID to shared page. */ - if (ib_pd->uobject) { + if (udata) { struct ib_ucontext *ib_uctx = ib_pd->uobject->context; struct bnxt_re_ucontext *uctx; unsigned long flag; diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c index 7a1dc83ba588..b34b1a1bd94b 100644 --- a/drivers/infiniband/hw/cxgb3/iwch_provider.c +++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c @@ -836,7 +836,7 @@ static struct ib_qp *iwch_create_qp(struct ib_pd *pd, * Kernel users need more wq space for fastreg WRs which can take * 2 WR fragments. */ - ucontext = pd->uobject ? to_iwch_ucontext(pd->uobject->context) : NULL; + ucontext = udata ? to_iwch_ucontext(pd->uobject->context) : NULL; if (!ucontext && wqsize < (rqsize + (2 * sqsize))) wqsize = roundup_pow_of_two(rqsize + roundup_pow_of_two(attrs->cap.max_send_wr * 2)); diff --git a/drivers/infiniband/hw/cxgb4/qp.c b/drivers/infiniband/hw/cxgb4/qp.c index 5a8030bd4208..981ff5cfb5d1 100644 --- a/drivers/infiniband/hw/cxgb4/qp.c +++ b/drivers/infiniband/hw/cxgb4/qp.c @@ -2163,7 +2163,7 @@ struct ib_qp *c4iw_create_qp(struct ib_pd *pd, struct ib_qp_init_attr *attrs, if (sqsize < 8) sqsize = 8; - ucontext = pd->uobject ? to_c4iw_ucontext(pd->uobject->context) : NULL; + ucontext = udata ? to_c4iw_ucontext(pd->uobject->context) : NULL; qhp = kzalloc(sizeof(*qhp), GFP_KERNEL); if (!qhp) @@ -2712,7 +2712,7 @@ struct ib_srq *c4iw_create_srq(struct ib_pd *pd, struct ib_srq_init_attr *attrs, rqsize = attrs->attr.max_wr + 1; rqsize = roundup_pow_of_two(max_t(u16, rqsize, 16)); - ucontext = pd->uobject ? to_c4iw_ucontext(pd->uobject->context) : NULL; + ucontext = udata ? to_c4iw_ucontext(pd->uobject->context) : NULL; srq = kzalloc(sizeof(*srq), GFP_KERNEL); if (!srq) diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c index d17a7ce3c93a..b74c742b000c 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c @@ -3926,7 +3926,7 @@ int hns_roce_v1_destroy_qp(struct ib_qp *ibqp) struct hns_roce_qp_work *qp_work; struct hns_roce_v1_priv *priv; struct hns_roce_cq *send_cq, *recv_cq; - int is_user = !!ibqp->pd->uobject; + bool is_user = ibqp->uobject; int is_timeout = 0; int ret; diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c index 22236ebf033d..3a669451cf86 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c @@ -4133,7 +4133,7 @@ out: static int hns_roce_v2_destroy_qp_common(struct hns_roce_dev *hr_dev, struct hns_roce_qp *hr_qp, - int is_user) + bool is_user) { struct hns_roce_cq *send_cq, *recv_cq; struct device *dev = hr_dev->dev; @@ -4210,7 +4210,7 @@ static int hns_roce_v2_destroy_qp(struct ib_qp *ibqp) struct hns_roce_qp *hr_qp = to_hr_qp(ibqp); int ret; - ret = hns_roce_v2_destroy_qp_common(hr_dev, hr_qp, !!ibqp->pd->uobject); + ret = hns_roce_v2_destroy_qp_common(hr_dev, hr_qp, ibqp->uobject); if (ret) { dev_err(hr_dev->dev, "Destroy qp failed(%d)\n", ret); return ret; diff --git a/drivers/infiniband/hw/hns/hns_roce_qp.c b/drivers/infiniband/hw/hns/hns_roce_qp.c index 52d2b299b3be..54031c5b53fa 100644 --- a/drivers/infiniband/hw/hns/hns_roce_qp.c +++ b/drivers/infiniband/hw/hns/hns_roce_qp.c @@ -280,7 +280,7 @@ void hns_roce_release_range_qp(struct hns_roce_dev *hr_dev, int base_qpn, EXPORT_SYMBOL_GPL(hns_roce_release_range_qp); static int hns_roce_set_rq_size(struct hns_roce_dev *hr_dev, - struct ib_qp_cap *cap, int is_user, int has_rq, + struct ib_qp_cap *cap, bool is_user, int has_rq, struct hns_roce_qp *hr_qp) { struct device *dev = hr_dev->dev; @@ -560,7 +560,7 @@ static int hns_roce_create_qp_common(struct hns_roce_dev *hr_dev, else hr_qp->sq_signal_bits = cpu_to_le32(IB_SIGNAL_REQ_WR); - ret = hns_roce_set_rq_size(hr_dev, &init_attr->cap, !!ib_pd->uobject, + ret = hns_roce_set_rq_size(hr_dev, &init_attr->cap, udata, hns_roce_qp_has_rq(init_attr), hr_qp); if (ret) { dev_err(dev, "hns_roce_set_rq_size failed\n"); @@ -598,7 +598,7 @@ static int hns_roce_create_qp_common(struct hns_roce_dev *hr_dev, init_attr->cap.max_recv_sge]; } - if (ib_pd->uobject) { + if (udata) { if (ib_copy_from_udata(&ucmd, udata, sizeof(ucmd))) { dev_err(dev, "ib_copy_from_udata error for create qp\n"); ret = -EFAULT; @@ -783,7 +783,7 @@ static int hns_roce_create_qp_common(struct hns_roce_dev *hr_dev, else hr_qp->doorbell_qpn = cpu_to_le64(hr_qp->qpn); - if (ib_pd->uobject && (udata->outlen >= sizeof(resp)) && + if (udata && (udata->outlen >= sizeof(resp)) && (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_RECORD_DB)) { /* indicate kernel supports rq record db */ @@ -810,7 +810,7 @@ err_qpn: hns_roce_release_range_qp(hr_dev, qpn, 1); err_wrid: - if (ib_pd->uobject) { + if (udata) { if ((hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_RECORD_DB) && (udata->outlen >= sizeof(resp)) && hns_roce_qp_has_rq(init_attr)) @@ -823,7 +823,7 @@ err_wrid: } err_sq_dbmap: - if (ib_pd->uobject) + if (udata) if ((hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_SQ_RECORD_DB) && (udata->inlen >= sizeof(ucmd)) && (udata->outlen >= sizeof(resp)) && @@ -836,13 +836,13 @@ err_mtt: hns_roce_mtt_cleanup(hr_dev, &hr_qp->mtt); err_buf: - if (ib_pd->uobject) + if (hr_qp->umem) ib_umem_release(hr_qp->umem); else hns_roce_buf_free(hr_dev, hr_qp->buff_size, &hr_qp->hr_buf); err_db: - if (!ib_pd->uobject && hns_roce_qp_has_rq(init_attr) && + if (!udata && hns_roce_qp_has_rq(init_attr) && (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_RECORD_DB)) hns_roce_free_db(hr_dev, &hr_qp->rdb); @@ -888,7 +888,7 @@ struct ib_qp *hns_roce_create_qp(struct ib_pd *pd, } case IB_QPT_GSI: { /* Userspace is not allowed to create special QPs: */ - if (pd->uobject) { + if (udata) { dev_err(dev, "not support usr space GSI\n"); return ERR_PTR(-EINVAL); } diff --git a/drivers/infiniband/hw/hns/hns_roce_srq.c b/drivers/infiniband/hw/hns/hns_roce_srq.c index 6377e734e28e..960b1946c365 100644 --- a/drivers/infiniband/hw/hns/hns_roce_srq.c +++ b/drivers/infiniband/hw/hns/hns_roce_srq.c @@ -379,7 +379,7 @@ struct ib_srq *hns_roce_create_srq(struct ib_pd *pd, srq->event = hns_roce_ib_srq_event; srq->ibsrq.ext.xrc.srq_num = srq->srqn; - if (pd->uobject) { + if (udata) { if (ib_copy_to_udata(udata, &srq->srqn, sizeof(__u32))) { ret = -EFAULT; goto err_wrid; diff --git a/drivers/infiniband/hw/i40iw/i40iw_verbs.c b/drivers/infiniband/hw/i40iw/i40iw_verbs.c index 475f32cf447e..0b675b0742c2 100644 --- a/drivers/infiniband/hw/i40iw/i40iw_verbs.c +++ b/drivers/infiniband/hw/i40iw/i40iw_verbs.c @@ -673,28 +673,26 @@ static struct ib_qp *i40iw_create_qp(struct ib_pd *ibpd, goto error; } iwqp->ctx_info.qp_compl_ctx = req.user_compl_ctx; - if (ibpd->uobject && ibpd->uobject->context) { - iwqp->user_mode = 1; - ucontext = to_ucontext(ibpd->uobject->context); - - if (req.user_wqe_buffers) { - struct i40iw_pbl *iwpbl; - - spin_lock_irqsave( - &ucontext->qp_reg_mem_list_lock, flags); - iwpbl = i40iw_get_pbl( - (unsigned long)req.user_wqe_buffers, - &ucontext->qp_reg_mem_list); - spin_unlock_irqrestore( - &ucontext->qp_reg_mem_list_lock, flags); - - if (!iwpbl) { - err_code = -ENODATA; - i40iw_pr_err("no pbl info\n"); - goto error; - } - memcpy(&iwqp->iwpbl, iwpbl, sizeof(iwqp->iwpbl)); + iwqp->user_mode = 1; + ucontext = to_ucontext(ibpd->uobject->context); + + if (req.user_wqe_buffers) { + struct i40iw_pbl *iwpbl; + + spin_lock_irqsave( + &ucontext->qp_reg_mem_list_lock, flags); + iwpbl = i40iw_get_pbl( + (unsigned long)req.user_wqe_buffers, + &ucontext->qp_reg_mem_list); + spin_unlock_irqrestore( + &ucontext->qp_reg_mem_list_lock, flags); + + if (!iwpbl) { + err_code = -ENODATA; + i40iw_pr_err("no pbl info\n"); + goto error; } + memcpy(&iwqp->iwpbl, iwpbl, sizeof(iwqp->iwpbl)); } err_code = i40iw_setup_virt_qp(iwdev, iwqp, &init_info); } else { @@ -768,7 +766,7 @@ static struct ib_qp *i40iw_create_qp(struct ib_pd *ibpd, iwdev->qp_table[qp_num] = iwqp; i40iw_add_pdusecount(iwqp->iwpd); i40iw_add_devusecount(iwdev); - if (ibpd->uobject && udata) { + if (udata) { memset(&uresp, 0, sizeof(uresp)); uresp.actual_sq_size = sq_size; uresp.actual_rq_size = rq_size; @@ -2092,7 +2090,8 @@ static int i40iw_dereg_mr(struct ib_mr *ib_mr) ib_umem_release(iwmr->region); if (iwmr->type != IW_MEMREG_TYPE_MEM) { - if (ibpd->uobject) { + /* region is released. only test for userness. */ + if (iwmr->region) { struct i40iw_ucontext *ucontext; ucontext = to_ucontext(ibpd->uobject->context); diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c index 651d8594dc3f..24ee30f1cb45 100644 --- a/drivers/infiniband/hw/mlx4/qp.c +++ b/drivers/infiniband/hw/mlx4/qp.c @@ -323,7 +323,7 @@ static int send_wqe_overhead(enum mlx4_ib_qp_type type, u32 flags) } static int set_rq_size(struct mlx4_ib_dev *dev, struct ib_qp_cap *cap, - int is_user, int has_rq, struct mlx4_ib_qp *qp, + bool is_user, int has_rq, struct mlx4_ib_qp *qp, u32 inl_recv_sz) { /* Sanity check RQ size before proceeding */ @@ -942,7 +942,7 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd, qp->sq_signal_bits = cpu_to_be32(MLX4_WQE_CTRL_CQ_UPDATE); - if (pd->uobject) { + if (udata) { union { struct mlx4_ib_create_qp qp; struct mlx4_ib_create_wq wq; @@ -991,7 +991,7 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd, qp->flags |= MLX4_IB_QP_SCATTER_FCS; } - err = set_rq_size(dev, &init_attr->cap, !!pd->uobject, + err = set_rq_size(dev, &init_attr->cap, udata, qp_has_rq(init_attr), qp, qp->inl_recv_sz); if (err) goto err; @@ -1043,7 +1043,7 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd, } qp->mqp.usage = MLX4_RES_USAGE_USER_VERBS; } else { - err = set_rq_size(dev, &init_attr->cap, !!pd->uobject, + err = set_rq_size(dev, &init_attr->cap, udata, qp_has_rq(init_attr), qp, 0); if (err) goto err; @@ -1189,7 +1189,7 @@ err_proxy: if (qp->mlx4_ib_qp_type == MLX4_IB_QPT_PROXY_GSI) free_proxy_bufs(pd->device, qp); err_wrid: - if (pd->uobject) { + if (udata) { if (qp_has_rq(init_attr)) mlx4_ib_db_unmap_user(to_mucontext(pd->uobject->context), &qp->db); } else { @@ -1201,13 +1201,13 @@ err_mtt: mlx4_mtt_cleanup(dev->dev, &qp->mtt); err_buf: - if (pd->uobject) + if (qp->umem) ib_umem_release(qp->umem); else mlx4_buf_free(dev->dev, qp->buf_size, &qp->buf); err_db: - if (!pd->uobject && qp_has_rq(init_attr)) + if (!udata && qp_has_rq(init_attr)) mlx4_db_free(dev->dev, &qp->db); err: @@ -1332,7 +1332,7 @@ static void destroy_qp_rss(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp) } static void destroy_qp_common(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp, - enum mlx4_ib_source_type src, int is_user) + enum mlx4_ib_source_type src, bool is_user) { struct mlx4_ib_cq *send_cq, *recv_cq; unsigned long flags; @@ -1612,7 +1612,7 @@ static int _mlx4_ib_destroy_qp(struct ib_qp *qp) struct mlx4_ib_pd *pd; pd = get_pd(mqp); - destroy_qp_common(dev, mqp, MLX4_IB_QP_SRC, !!pd->ibpd.uobject); + destroy_qp_common(dev, mqp, MLX4_IB_QP_SRC, qp->uobject); } if (is_sqp(dev, mqp)) @@ -4044,7 +4044,7 @@ struct ib_wq *mlx4_ib_create_wq(struct ib_pd *pd, struct mlx4_ib_create_wq ucmd; int err, required_cmd_sz; - if (!(udata && pd->uobject)) + if (!udata) return ERR_PTR(-EINVAL); required_cmd_sz = offsetof(typeof(ucmd), comp_mask) + diff --git a/drivers/infiniband/hw/mlx4/srq.c b/drivers/infiniband/hw/mlx4/srq.c index 3731b31c3653..4456f1b8921d 100644 --- a/drivers/infiniband/hw/mlx4/srq.c +++ b/drivers/infiniband/hw/mlx4/srq.c @@ -105,7 +105,7 @@ struct ib_srq *mlx4_ib_create_srq(struct ib_pd *pd, buf_size = srq->msrq.max * desc_size; - if (pd->uobject) { + if (udata) { struct mlx4_ib_create_srq ucmd; if (ib_copy_from_udata(&ucmd, udata, sizeof ucmd)) { @@ -191,7 +191,7 @@ struct ib_srq *mlx4_ib_create_srq(struct ib_pd *pd, srq->msrq.event = mlx4_ib_srq_event; srq->ibsrq.ext.xrc.srq_num = srq->msrq.srqn; - if (pd->uobject) + if (udata) if (ib_copy_to_udata(udata, &srq->msrq.srqn, sizeof (__u32))) { err = -EFAULT; goto err_wrid; @@ -202,7 +202,7 @@ struct ib_srq *mlx4_ib_create_srq(struct ib_pd *pd, return &srq->ibsrq; err_wrid: - if (pd->uobject) + if (udata) mlx4_ib_db_unmap_user(to_mucontext(pd->uobject->context), &srq->db); else kvfree(srq->wrid); @@ -211,13 +211,13 @@ err_mtt: mlx4_mtt_cleanup(dev->dev, &srq->mtt); err_buf: - if (pd->uobject) + if (srq->umem) ib_umem_release(srq->umem); else mlx4_buf_free(dev->dev, buf_size, &srq->buf); err_db: - if (!pd->uobject) + if (!udata) mlx4_db_free(dev->dev, &srq->db); err_srq: diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c index 2ceb8f68f605..b26ddb147643 100644 --- a/drivers/infiniband/hw/mlx5/qp.c +++ b/drivers/infiniband/hw/mlx5/qp.c @@ -1904,7 +1904,7 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd, qp->flags |= MLX5_IB_QP_CVLAN_STRIPPING; } - if (pd && pd->uobject) { + if (udata) { if (ib_copy_from_udata(&ucmd, udata, sizeof(ucmd))) { mlx5_ib_dbg(dev, "copy failed\n"); return -EFAULT; @@ -1986,14 +1986,14 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd, qp->has_rq = qp_has_rq(init_attr); err = set_rq_size(dev, &init_attr->cap, qp->has_rq, - qp, (pd && pd->uobject) ? &ucmd : NULL); + qp, udata ? &ucmd : NULL); if (err) { mlx5_ib_dbg(dev, "err %d\n", err); return err; } if (pd) { - if (pd->uobject) { + if (udata) { __u32 max_wqes = 1 << MLX5_CAP_GEN(mdev, log_max_qp_sz); mlx5_ib_dbg(dev, "requested sq_wqe_count (%d)\n", ucmd.sq_wqe_count); @@ -2064,7 +2064,7 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd, if (qp->scat_cqe && is_connected(init_attr->qp_type)) { configure_responder_scat_cqe(init_attr, qpc); configure_requester_scat_cqe(dev, init_attr, - (pd && pd->uobject) ? &ucmd : NULL, + udata ? &ucmd : NULL, qpc); } @@ -2504,7 +2504,7 @@ struct ib_qp *mlx5_ib_create_qp(struct ib_pd *pd, dev = to_mdev(pd->device); if (init_attr->qp_type == IB_QPT_RAW_PACKET) { - if (!pd->uobject) { + if (!udata) { mlx5_ib_dbg(dev, "Raw Packet QP is not supported for kernel consumers\n"); return ERR_PTR(-EINVAL); } else if (!to_mucontext(pd->uobject->context)->cqe_version) { diff --git a/drivers/infiniband/hw/mlx5/srq.c b/drivers/infiniband/hw/mlx5/srq.c index 0413b10dea71..4e8d18009f58 100644 --- a/drivers/infiniband/hw/mlx5/srq.c +++ b/drivers/infiniband/hw/mlx5/srq.c @@ -260,14 +260,14 @@ struct ib_srq *mlx5_ib_create_srq(struct ib_pd *pd, } in.type = init_attr->srq_type; - if (pd->uobject) + if (udata) err = create_srq_user(pd, srq, &in, udata, buf_size); else err = create_srq_kernel(dev, srq, &in, buf_size); if (err) { mlx5_ib_warn(dev, "create srq %s failed, err %d\n", - pd->uobject ? "user" : "kernel", err); + udata ? "user" : "kernel", err); goto err_srq; } @@ -312,7 +312,7 @@ struct ib_srq *mlx5_ib_create_srq(struct ib_pd *pd, srq->msrq.event = mlx5_ib_srq_event; srq->ibsrq.ext.xrc.srq_num = srq->msrq.srqn; - if (pd->uobject) + if (udata) if (ib_copy_to_udata(udata, &srq->msrq.srqn, sizeof(__u32))) { mlx5_ib_dbg(dev, "copy to user failed\n"); err = -EFAULT; @@ -327,7 +327,7 @@ err_core: mlx5_cmd_destroy_srq(dev, &srq->msrq); err_usr_kern_srq: - if (pd->uobject) + if (udata) destroy_srq_user(pd, srq); else destroy_srq_kernel(dev, srq); diff --git a/drivers/infiniband/hw/mthca/mthca_dev.h b/drivers/infiniband/hw/mthca/mthca_dev.h index 220a3e4717a3..bfd4eebc1182 100644 --- a/drivers/infiniband/hw/mthca/mthca_dev.h +++ b/drivers/infiniband/hw/mthca/mthca_dev.h @@ -510,7 +510,8 @@ int mthca_alloc_cq_buf(struct mthca_dev *dev, struct mthca_cq_buf *buf, int nent void mthca_free_cq_buf(struct mthca_dev *dev, struct mthca_cq_buf *buf, int cqe); int mthca_alloc_srq(struct mthca_dev *dev, struct mthca_pd *pd, - struct ib_srq_attr *attr, struct mthca_srq *srq); + struct ib_srq_attr *attr, struct mthca_srq *srq, + struct ib_udata *udata); void mthca_free_srq(struct mthca_dev *dev, struct mthca_srq *srq); int mthca_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr, enum ib_srq_attr_mask attr_mask, struct ib_udata *udata); @@ -547,7 +548,8 @@ int mthca_alloc_qp(struct mthca_dev *dev, enum ib_qp_type type, enum ib_sig_type send_policy, struct ib_qp_cap *cap, - struct mthca_qp *qp); + struct mthca_qp *qp, + struct ib_udata *udata); int mthca_alloc_sqp(struct mthca_dev *dev, struct mthca_pd *pd, struct mthca_cq *send_cq, @@ -556,7 +558,8 @@ int mthca_alloc_sqp(struct mthca_dev *dev, struct ib_qp_cap *cap, int qpn, int port, - struct mthca_sqp *sqp); + struct mthca_sqp *sqp, + struct ib_udata *udata); void mthca_free_qp(struct mthca_dev *dev, struct mthca_qp *qp); int mthca_create_ah(struct mthca_dev *dev, struct mthca_pd *pd, diff --git a/drivers/infiniband/hw/mthca/mthca_provider.c b/drivers/infiniband/hw/mthca/mthca_provider.c index c697ec54ea5f..443521cf8107 100644 --- a/drivers/infiniband/hw/mthca/mthca_provider.c +++ b/drivers/infiniband/hw/mthca/mthca_provider.c @@ -455,7 +455,7 @@ static struct ib_srq *mthca_create_srq(struct ib_pd *pd, if (!srq) return ERR_PTR(-ENOMEM); - if (pd->uobject) { + if (udata) { context = to_mucontext(pd->uobject->context); if (ib_copy_from_udata(&ucmd, udata, sizeof ucmd)) { @@ -475,9 +475,9 @@ static struct ib_srq *mthca_create_srq(struct ib_pd *pd, } err = mthca_alloc_srq(to_mdev(pd->device), to_mpd(pd), - &init_attr->attr, srq); + &init_attr->attr, srq, udata); - if (err && pd->uobject) + if (err && udata) mthca_unmap_user_db(to_mdev(pd->device), &context->uar, context->db_tab, ucmd.db_index); @@ -537,7 +537,7 @@ static struct ib_qp *mthca_create_qp(struct ib_pd *pd, if (!qp) return ERR_PTR(-ENOMEM); - if (pd->uobject) { + if (udata) { context = to_mucontext(pd->uobject->context); if (ib_copy_from_udata(&ucmd, udata, sizeof ucmd)) { @@ -574,9 +574,9 @@ static struct ib_qp *mthca_create_qp(struct ib_pd *pd, to_mcq(init_attr->send_cq), to_mcq(init_attr->recv_cq), init_attr->qp_type, init_attr->sq_sig_type, - &init_attr->cap, qp); + &init_attr->cap, qp, udata); - if (err && pd->uobject) { + if (err && udata) { context = to_mucontext(pd->uobject->context); mthca_unmap_user_db(to_mdev(pd->device), @@ -596,7 +596,7 @@ static struct ib_qp *mthca_create_qp(struct ib_pd *pd, case IB_QPT_GSI: { /* Don't allow userspace to create special QPs */ - if (pd->uobject) + if (udata) return ERR_PTR(-EINVAL); qp = kmalloc(sizeof (struct mthca_sqp), GFP_KERNEL); @@ -610,7 +610,7 @@ static struct ib_qp *mthca_create_qp(struct ib_pd *pd, to_mcq(init_attr->recv_cq), init_attr->sq_sig_type, &init_attr->cap, qp->ibqp.qp_num, init_attr->port_num, - to_msqp(qp)); + to_msqp(qp), udata); break; } default: diff --git a/drivers/infiniband/hw/mthca/mthca_qp.c b/drivers/infiniband/hw/mthca/mthca_qp.c index 9d178ee3c96a..4e5b5cc17f1d 100644 --- a/drivers/infiniband/hw/mthca/mthca_qp.c +++ b/drivers/infiniband/hw/mthca/mthca_qp.c @@ -981,7 +981,8 @@ static void mthca_adjust_qp_caps(struct mthca_dev *dev, */ static int mthca_alloc_wqe_buf(struct mthca_dev *dev, struct mthca_pd *pd, - struct mthca_qp *qp) + struct mthca_qp *qp, + struct ib_udata *udata) { int size; int err = -ENOMEM; @@ -1048,7 +1049,7 @@ static int mthca_alloc_wqe_buf(struct mthca_dev *dev, * allocate anything. All we need is to calculate the WQE * sizes and the send_wqe_offset, so we're done now. */ - if (pd->ibpd.uobject) + if (udata) return 0; size = PAGE_ALIGN(qp->send_wqe_offset + @@ -1155,7 +1156,8 @@ static int mthca_alloc_qp_common(struct mthca_dev *dev, struct mthca_cq *send_cq, struct mthca_cq *recv_cq, enum ib_sig_type send_policy, - struct mthca_qp *qp) + struct mthca_qp *qp, + struct ib_udata *udata) { int ret; int i; @@ -1178,7 +1180,7 @@ static int mthca_alloc_qp_common(struct mthca_dev *dev, if (ret) return ret; - ret = mthca_alloc_wqe_buf(dev, pd, qp); + ret = mthca_alloc_wqe_buf(dev, pd, qp, udata); if (ret) { mthca_unmap_memfree(dev, qp); return ret; @@ -1191,7 +1193,7 @@ static int mthca_alloc_qp_common(struct mthca_dev *dev, * will be allocated and buffers will be initialized in * userspace. */ - if (pd->ibpd.uobject) + if (udata) return 0; ret = mthca_alloc_memfree(dev, qp); @@ -1285,7 +1287,8 @@ int mthca_alloc_qp(struct mthca_dev *dev, enum ib_qp_type type, enum ib_sig_type send_policy, struct ib_qp_cap *cap, - struct mthca_qp *qp) + struct mthca_qp *qp, + struct ib_udata *udata) { int err; @@ -1308,7 +1311,7 @@ int mthca_alloc_qp(struct mthca_dev *dev, qp->port = 0; err = mthca_alloc_qp_common(dev, pd, send_cq, recv_cq, - send_policy, qp); + send_policy, qp, udata); if (err) { mthca_free(&dev->qp_table.alloc, qp->qpn); return err; @@ -1360,7 +1363,8 @@ int mthca_alloc_sqp(struct mthca_dev *dev, struct ib_qp_cap *cap, int qpn, int port, - struct mthca_sqp *sqp) + struct mthca_sqp *sqp, + struct ib_udata *udata) { u32 mqpn = qpn * 2 + dev->qp_table.sqp_start + port - 1; int err; @@ -1391,7 +1395,7 @@ int mthca_alloc_sqp(struct mthca_dev *dev, sqp->qp.transport = MLX; err = mthca_alloc_qp_common(dev, pd, send_cq, recv_cq, - send_policy, &sqp->qp); + send_policy, &sqp->qp, udata); if (err) goto err_out_free; diff --git a/drivers/infiniband/hw/mthca/mthca_srq.c b/drivers/infiniband/hw/mthca/mthca_srq.c index 9a3fc6fb0d7e..b8333c79e3fa 100644 --- a/drivers/infiniband/hw/mthca/mthca_srq.c +++ b/drivers/infiniband/hw/mthca/mthca_srq.c @@ -95,7 +95,8 @@ static inline int *wqe_to_link(void *wqe) static void mthca_tavor_init_srq_context(struct mthca_dev *dev, struct mthca_pd *pd, struct mthca_srq *srq, - struct mthca_tavor_srq_context *context) + struct mthca_tavor_srq_context *context, + bool is_user) { memset(context, 0, sizeof *context); @@ -103,7 +104,7 @@ static void mthca_tavor_init_srq_context(struct mthca_dev *dev, context->state_pd = cpu_to_be32(pd->pd_num); context->lkey = cpu_to_be32(srq->mr.ibmr.lkey); - if (pd->ibpd.uobject) + if (is_user) context->uar = cpu_to_be32(to_mucontext(pd->ibpd.uobject->context)->uar.index); else @@ -113,7 +114,8 @@ static void mthca_tavor_init_srq_context(struct mthca_dev *dev, static void mthca_arbel_init_srq_context(struct mthca_dev *dev, struct mthca_pd *pd, struct mthca_srq *srq, - struct mthca_arbel_srq_context *context) + struct mthca_arbel_srq_context *context, + bool is_user) { int logsize, max; @@ -129,7 +131,7 @@ static void mthca_arbel_init_srq_context(struct mthca_dev *dev, context->lkey = cpu_to_be32(srq->mr.ibmr.lkey); context->db_index = cpu_to_be32(srq->db_index); context->logstride_usrpage = cpu_to_be32((srq->wqe_shift - 4) << 29); - if (pd->ibpd.uobject) + if (is_user) context->logstride_usrpage |= cpu_to_be32(to_mucontext(pd->ibpd.uobject->context)->uar.index); else @@ -145,14 +147,14 @@ static void mthca_free_srq_buf(struct mthca_dev *dev, struct mthca_srq *srq) } static int mthca_alloc_srq_buf(struct mthca_dev *dev, struct mthca_pd *pd, - struct mthca_srq *srq) + struct mthca_srq *srq, struct ib_udata *udata) { struct mthca_data_seg *scatter; void *wqe; int err; int i; - if (pd->ibpd.uobject) + if (udata) return 0; srq->wrid = kmalloc_array(srq->max, sizeof(u64), GFP_KERNEL); @@ -197,7 +199,8 @@ static int mthca_alloc_srq_buf(struct mthca_dev *dev, struct mthca_pd *pd, } int mthca_alloc_srq(struct mthca_dev *dev, struct mthca_pd *pd, - struct ib_srq_attr *attr, struct mthca_srq *srq) + struct ib_srq_attr *attr, struct mthca_srq *srq, + struct ib_udata *udata) { struct mthca_mailbox *mailbox; int ds; @@ -235,7 +238,7 @@ int mthca_alloc_srq(struct mthca_dev *dev, struct mthca_pd *pd, if (err) goto err_out; - if (!pd->ibpd.uobject) { + if (!udata) { srq->db_index = mthca_alloc_db(dev, MTHCA_DB_TYPE_SRQ, srq->srqn, &srq->db); if (srq->db_index < 0) { @@ -251,7 +254,7 @@ int mthca_alloc_srq(struct mthca_dev *dev, struct mthca_pd *pd, goto err_out_db; } - err = mthca_alloc_srq_buf(dev, pd, srq); + err = mthca_alloc_srq_buf(dev, pd, srq, udata); if (err) goto err_out_mailbox; @@ -261,9 +264,9 @@ int mthca_alloc_srq(struct mthca_dev *dev, struct mthca_pd *pd, mutex_init(&srq->mutex); if (mthca_is_memfree(dev)) - mthca_arbel_init_srq_context(dev, pd, srq, mailbox->buf); + mthca_arbel_init_srq_context(dev, pd, srq, mailbox->buf, udata); else - mthca_tavor_init_srq_context(dev, pd, srq, mailbox->buf); + mthca_tavor_init_srq_context(dev, pd, srq, mailbox->buf, udata); err = mthca_SW2HW_SRQ(dev, mailbox, srq->srqn); @@ -297,14 +300,14 @@ err_out_free_srq: mthca_warn(dev, "HW2SW_SRQ failed (%d)\n", err); err_out_free_buf: - if (!pd->ibpd.uobject) + if (!udata) mthca_free_srq_buf(dev, srq); err_out_mailbox: mthca_free_mailbox(dev, mailbox); err_out_db: - if (!pd->ibpd.uobject && mthca_is_memfree(dev)) + if (!udata && mthca_is_memfree(dev)) mthca_free_db(dev, MTHCA_DB_TYPE_SRQ, srq->db_index); err_out_icm: diff --git a/drivers/infiniband/hw/nes/nes_verbs.c b/drivers/infiniband/hw/nes/nes_verbs.c index f9d510431900..4e7f08ee1907 100644 --- a/drivers/infiniband/hw/nes/nes_verbs.c +++ b/drivers/infiniband/hw/nes/nes_verbs.c @@ -1066,7 +1066,7 @@ static struct ib_qp *nes_create_qp(struct ib_pd *ibpd, } if (req.user_qp_buffer) nesqp->nesuqp_addr = req.user_qp_buffer; - if ((ibpd->uobject) && (ibpd->uobject->context)) { + if (udata && (ibpd->uobject->context)) { nesqp->user_mode = 1; nes_ucontext = to_nesucontext(ibpd->uobject->context); if (virt_wqs) { @@ -1257,7 +1257,7 @@ static struct ib_qp *nes_create_qp(struct ib_pd *ibpd, nes_put_cqp_request(nesdev, cqp_request); - if (ibpd->uobject) { + if (udata) { uresp.mmap_sq_db_index = nesqp->mmap_sq_db_index; uresp.mmap_rq_db_index = 0; uresp.actual_sq_size = sq_size; diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c b/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c index 1a5c048e1d67..c46bed0c5513 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c @@ -1157,7 +1157,8 @@ static void ocrdma_del_qpn_map(struct ocrdma_dev *dev, struct ocrdma_qp *qp) } static int ocrdma_check_qp_params(struct ib_pd *ibpd, struct ocrdma_dev *dev, - struct ib_qp_init_attr *attrs) + struct ib_qp_init_attr *attrs, + struct ib_udata *udata) { if ((attrs->qp_type != IB_QPT_GSI) && (attrs->qp_type != IB_QPT_RC) && @@ -1205,7 +1206,7 @@ static int ocrdma_check_qp_params(struct ib_pd *ibpd, struct ocrdma_dev *dev, return -EINVAL; } /* unprivileged user space cannot create special QP */ - if (ibpd->uobject && attrs->qp_type == IB_QPT_GSI) { + if (udata && attrs->qp_type == IB_QPT_GSI) { pr_err ("%s(%d) Userspace can't create special QPs of type=0x%x\n", __func__, dev->id, attrs->qp_type); @@ -1362,7 +1363,7 @@ struct ib_qp *ocrdma_create_qp(struct ib_pd *ibpd, struct ocrdma_create_qp_ureq ureq; u16 dpp_credit_lmt, dpp_offset; - status = ocrdma_check_qp_params(ibpd, dev, attrs); + status = ocrdma_check_qp_params(ibpd, dev, attrs, udata); if (status) goto gen_err; diff --git a/drivers/infiniband/hw/qedr/verbs.c b/drivers/infiniband/hw/qedr/verbs.c index 28c6d5b11487..8056121e9f69 100644 --- a/drivers/infiniband/hw/qedr/verbs.c +++ b/drivers/infiniband/hw/qedr/verbs.c @@ -1136,7 +1136,8 @@ static inline int get_gid_info_from_table(struct ib_qp *ibqp, } static int qedr_check_qp_attrs(struct ib_pd *ibpd, struct qedr_dev *dev, - struct ib_qp_init_attr *attrs) + struct ib_qp_init_attr *attrs, + struct ib_udata *udata) { struct qedr_device_attr *qattr = &dev->attr; @@ -1177,7 +1178,7 @@ static int qedr_check_qp_attrs(struct ib_pd *ibpd, struct qedr_dev *dev, } /* Unprivileged user space cannot create special QP */ - if (ibpd->uobject && attrs->qp_type == IB_QPT_GSI) { + if (udata && attrs->qp_type == IB_QPT_GSI) { DP_ERR(dev, "create qp: userspace can't create special QPs of type=0x%x\n", attrs->qp_type); @@ -1540,7 +1541,7 @@ int qedr_destroy_srq(struct ib_srq *ibsrq) in_params.srq_id = srq->srq_id; dev->ops->rdma_destroy_srq(dev->rdma_ctx, &in_params); - if (ibsrq->pd->uobject) + if (ibsrq->uobject) qedr_free_srq_user_params(srq); else qedr_free_srq_kernel_params(srq); @@ -1993,7 +1994,7 @@ struct ib_qp *qedr_create_qp(struct ib_pd *ibpd, DP_DEBUG(dev, QEDR_MSG_QP, "create qp: called from %s, pd=%p\n", udata ? "user library" : "kernel", pd); - rc = qedr_check_qp_attrs(ibpd, dev, attrs); + rc = qedr_check_qp_attrs(ibpd, dev, attrs, udata); if (rc) return ERR_PTR(rc); diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_qp.c b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_qp.c index cf22f57a9f0d..3acf74cbe266 100644 --- a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_qp.c +++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_qp.c @@ -249,7 +249,7 @@ struct ib_qp *pvrdma_create_qp(struct ib_pd *pd, init_completion(&qp->free); qp->state = IB_QPS_RESET; - qp->is_kernel = !(pd->uobject && udata); + qp->is_kernel = !udata; if (!qp->is_kernel) { dev_dbg(&dev->pdev->dev, diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_srq.c b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_srq.c index dc0ce877c7a3..06ba7c7a2235 100644 --- a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_srq.c +++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_srq.c @@ -111,7 +111,7 @@ struct ib_srq *pvrdma_create_srq(struct ib_pd *pd, unsigned long flags; int ret; - if (!(pd->uobject && udata)) { + if (!udata) { /* No support for kernel clients. */ dev_warn(&dev->pdev->dev, "no shared receive queue support for kernel client\n"); diff --git a/drivers/infiniband/sw/rxe/rxe_loc.h b/drivers/infiniband/sw/rxe/rxe_loc.h index a675c9f2b427..01b74597b36a 100644 --- a/drivers/infiniband/sw/rxe/rxe_loc.h +++ b/drivers/infiniband/sw/rxe/rxe_loc.h @@ -157,7 +157,7 @@ int rxe_qp_chk_init(struct rxe_dev *rxe, struct ib_qp_init_attr *init); int rxe_qp_from_init(struct rxe_dev *rxe, struct rxe_qp *qp, struct rxe_pd *pd, struct ib_qp_init_attr *init, struct rxe_create_qp_resp __user *uresp, - struct ib_pd *ibpd); + struct ib_pd *ibpd, struct ib_udata *udata); int rxe_qp_to_init(struct rxe_qp *qp, struct ib_qp_init_attr *init); diff --git a/drivers/infiniband/sw/rxe/rxe_qp.c b/drivers/infiniband/sw/rxe/rxe_qp.c index 2ca4ffe5015f..fd86fd2fbb26 100644 --- a/drivers/infiniband/sw/rxe/rxe_qp.c +++ b/drivers/infiniband/sw/rxe/rxe_qp.c @@ -336,13 +336,14 @@ static int rxe_qp_init_resp(struct rxe_dev *rxe, struct rxe_qp *qp, int rxe_qp_from_init(struct rxe_dev *rxe, struct rxe_qp *qp, struct rxe_pd *pd, struct ib_qp_init_attr *init, struct rxe_create_qp_resp __user *uresp, - struct ib_pd *ibpd) + struct ib_pd *ibpd, + struct ib_udata *udata) { int err; struct rxe_cq *rcq = to_rcq(init->recv_cq); struct rxe_cq *scq = to_rcq(init->send_cq); struct rxe_srq *srq = init->srq ? to_rsrq(init->srq) : NULL; - struct ib_ucontext *context = ibpd->uobject ? ibpd->uobject->context : NULL; + struct ib_ucontext *context = udata ? ibpd->uobject->context : NULL; rxe_add_ref(pd); rxe_add_ref(rcq); diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.c b/drivers/infiniband/sw/rxe/rxe_verbs.c index ecd0824b7eac..8d15b1e11a2f 100644 --- a/drivers/infiniband/sw/rxe/rxe_verbs.c +++ b/drivers/infiniband/sw/rxe/rxe_verbs.c @@ -486,7 +486,7 @@ static struct ib_qp *rxe_create_qp(struct ib_pd *ibpd, rxe_add_index(qp); - err = rxe_qp_from_init(rxe, qp, pd, init, uresp, ibpd); + err = rxe_qp_from_init(rxe, qp, pd, init, uresp, ibpd, udata); if (err) goto err3; -- cgit v1.2.3-59-g8ed1b From c9e585ebdc28a334b8100e560c9aaad6d4525a6e Mon Sep 17 00:00:00 2001 From: Doug Ledford Date: Wed, 19 Dec 2018 13:43:17 -0500 Subject: IB/mlx5: Fix compile issue when ODP disabled When CONFIG_INFINIBAND_ON_DEMAND_PAGING is not enabled, we were getting build failures for defined but not used code. Fix that. Fixes: 813e90b1aeaa ("IB/mlx5: Add advise_mr() support") Signed-off-by: Jason Gunthorpe Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx5/mlx5_ib.h | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index f245b5d8a3bc..9b4e2554889a 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -1208,10 +1208,10 @@ static inline void mlx5_odp_populate_klm(struct mlx5_klm *pklm, size_t offset, size_t nentries, struct mlx5_ib_mr *mr, int flags) {} -static int mlx5_ib_advise_mr_prefetch(struct ib_pd *pd, - enum ib_uverbs_advise_mr_advice advice, - u32 flags, struct ib_sge *sg_list, - u32 num_sge) +static inline int +mlx5_ib_advise_mr_prefetch(struct ib_pd *pd, + enum ib_uverbs_advise_mr_advice advice, u32 flags, + struct ib_sge *sg_list, u32 num_sge) { return -EOPNOTSUPP; } -- cgit v1.2.3-59-g8ed1b From b090c4e3a07c33ffdf95fb7601551b38fc2a4bbb Mon Sep 17 00:00:00 2001 From: Gal Pressman Date: Wed, 12 Dec 2018 11:09:05 +0200 Subject: RDMA: Mark if create address handle is in a sleepable context Introduce a 'flags' field to create address handle callback and add a flag that marks whether the callback is executed in an atomic context or not. This will allow drivers to wait for completion instead of polling for it when it is allowed. Signed-off-by: Gal Pressman Signed-off-by: Jason Gunthorpe --- drivers/infiniband/core/cm.c | 2 +- drivers/infiniband/core/sa_query.c | 3 ++- drivers/infiniband/core/verbs.c | 15 ++++++++++----- drivers/infiniband/hw/bnxt_re/ib_verbs.c | 1 + drivers/infiniband/hw/bnxt_re/ib_verbs.h | 1 + drivers/infiniband/hw/hfi1/mad.c | 2 +- drivers/infiniband/hw/hns/hns_roce_ah.c | 1 + drivers/infiniband/hw/hns/hns_roce_device.h | 1 + drivers/infiniband/hw/mlx4/ah.c | 4 ++-- drivers/infiniband/hw/mlx4/mad.c | 4 ++-- drivers/infiniband/hw/mlx4/mlx4_ib.h | 2 +- drivers/infiniband/hw/mlx5/ah.c | 2 +- drivers/infiniband/hw/mlx5/mlx5_ib.h | 2 +- drivers/infiniband/hw/mthca/mthca_mad.c | 2 +- drivers/infiniband/hw/mthca/mthca_provider.c | 1 + drivers/infiniband/hw/ocrdma/ocrdma_ah.c | 2 +- drivers/infiniband/hw/ocrdma/ocrdma_ah.h | 2 +- drivers/infiniband/hw/qedr/verbs.c | 2 +- drivers/infiniband/hw/qedr/verbs.h | 2 +- drivers/infiniband/hw/qib/qib_verbs.c | 2 +- drivers/infiniband/hw/usnic/usnic_ib_verbs.c | 1 + drivers/infiniband/hw/usnic/usnic_ib_verbs.h | 1 + drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.c | 3 ++- drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.h | 2 +- drivers/infiniband/sw/rdmavt/ah.c | 2 ++ drivers/infiniband/sw/rdmavt/ah.h | 1 + drivers/infiniband/sw/rxe/rxe_verbs.c | 1 + drivers/infiniband/ulp/ipoib/ipoib_ib.c | 2 +- drivers/infiniband/ulp/opa_vnic/opa_vnic_vema.c | 2 +- include/rdma/ib_verbs.h | 11 +++++++++-- 30 files changed, 52 insertions(+), 27 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c index edb2cb758be7..cf5b3c4314bb 100644 --- a/drivers/infiniband/core/cm.c +++ b/drivers/infiniband/core/cm.c @@ -343,7 +343,7 @@ static int cm_alloc_msg(struct cm_id_private *cm_id_priv, ret = -ENODEV; goto out; } - ah = rdma_create_ah(mad_agent->qp->pd, &av->ah_attr); + ah = rdma_create_ah(mad_agent->qp->pd, &av->ah_attr, 0); if (IS_ERR(ah)) { ret = PTR_ERR(ah); goto out; diff --git a/drivers/infiniband/core/sa_query.c b/drivers/infiniband/core/sa_query.c index be5ba5e15496..fb96d9df1fdd 100644 --- a/drivers/infiniband/core/sa_query.c +++ b/drivers/infiniband/core/sa_query.c @@ -2276,7 +2276,8 @@ static void update_sm_ah(struct work_struct *work) cpu_to_be64(IB_SA_WELL_KNOWN_GUID)); } - new_ah->ah = rdma_create_ah(port->agent->qp->pd, &ah_attr); + new_ah->ah = rdma_create_ah(port->agent->qp->pd, &ah_attr, + RDMA_CREATE_AH_SLEEPABLE); if (IS_ERR(new_ah->ah)) { pr_warn("Couldn't create new SM AH\n"); kfree(new_ah); diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c index bc0b3a150e3c..4e2df239447e 100644 --- a/drivers/infiniband/core/verbs.c +++ b/drivers/infiniband/core/verbs.c @@ -487,14 +487,17 @@ rdma_update_sgid_attr(struct rdma_ah_attr *ah_attr, static struct ib_ah *_rdma_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, + u32 flags, struct ib_udata *udata) { struct ib_ah *ah; + might_sleep_if(flags & RDMA_CREATE_AH_SLEEPABLE); + if (!pd->device->ops.create_ah) return ERR_PTR(-EOPNOTSUPP); - ah = pd->device->ops.create_ah(pd, ah_attr, udata); + ah = pd->device->ops.create_ah(pd, ah_attr, flags, udata); if (!IS_ERR(ah)) { ah->device = pd->device; @@ -514,12 +517,14 @@ static struct ib_ah *_rdma_create_ah(struct ib_pd *pd, * given address vector. * @pd: The protection domain associated with the address handle. * @ah_attr: The attributes of the address vector. + * @flags: Create address handle flags (see enum rdma_create_ah_flags). * * It returns 0 on success and returns appropriate error code on error. * The address handle is used to reference a local or global destination * in all UD QP post sends. */ -struct ib_ah *rdma_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr) +struct ib_ah *rdma_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, + u32 flags) { const struct ib_gid_attr *old_sgid_attr; struct ib_ah *ah; @@ -529,7 +534,7 @@ struct ib_ah *rdma_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr) if (ret) return ERR_PTR(ret); - ah = _rdma_create_ah(pd, ah_attr, NULL); + ah = _rdma_create_ah(pd, ah_attr, flags, NULL); rdma_unfill_sgid_attr(ah_attr, old_sgid_attr); return ah; @@ -569,7 +574,7 @@ struct ib_ah *rdma_create_user_ah(struct ib_pd *pd, } } - ah = _rdma_create_ah(pd, ah_attr, udata); + ah = _rdma_create_ah(pd, ah_attr, RDMA_CREATE_AH_SLEEPABLE, udata); out: rdma_unfill_sgid_attr(ah_attr, old_sgid_attr); @@ -881,7 +886,7 @@ struct ib_ah *ib_create_ah_from_wc(struct ib_pd *pd, const struct ib_wc *wc, if (ret) return ERR_PTR(ret); - ah = rdma_create_ah(pd, &ah_attr); + ah = rdma_create_ah(pd, &ah_attr, RDMA_CREATE_AH_SLEEPABLE); rdma_destroy_ah_attr(&ah_attr); return ah; diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.c b/drivers/infiniband/hw/bnxt_re/ib_verbs.c index 611bacd00b80..4b37bc3edc81 100644 --- a/drivers/infiniband/hw/bnxt_re/ib_verbs.c +++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.c @@ -664,6 +664,7 @@ int bnxt_re_destroy_ah(struct ib_ah *ib_ah) struct ib_ah *bnxt_re_create_ah(struct ib_pd *ib_pd, struct rdma_ah_attr *ah_attr, + u32 flags, struct ib_udata *udata) { struct bnxt_re_pd *pd = container_of(ib_pd, struct bnxt_re_pd, ib_pd); diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.h b/drivers/infiniband/hw/bnxt_re/ib_verbs.h index aa33e7b82c84..6675db4b4ba4 100644 --- a/drivers/infiniband/hw/bnxt_re/ib_verbs.h +++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.h @@ -169,6 +169,7 @@ struct ib_pd *bnxt_re_alloc_pd(struct ib_device *ibdev, int bnxt_re_dealloc_pd(struct ib_pd *pd); struct ib_ah *bnxt_re_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, + u32 flags, struct ib_udata *udata); int bnxt_re_modify_ah(struct ib_ah *ah, struct rdma_ah_attr *ah_attr); int bnxt_re_query_ah(struct ib_ah *ah, struct rdma_ah_attr *ah_attr); diff --git a/drivers/infiniband/hw/hfi1/mad.c b/drivers/infiniband/hw/hfi1/mad.c index 88a0cf930136..4228393e6c4c 100644 --- a/drivers/infiniband/hw/hfi1/mad.c +++ b/drivers/infiniband/hw/hfi1/mad.c @@ -305,7 +305,7 @@ static struct ib_ah *hfi1_create_qp0_ah(struct hfi1_ibport *ibp, u32 dlid) rcu_read_lock(); qp0 = rcu_dereference(ibp->rvp.qp[0]); if (qp0) - ah = rdma_create_ah(qp0->ibqp.pd, &attr); + ah = rdma_create_ah(qp0->ibqp.pd, &attr, 0); rcu_read_unlock(); return ah; } diff --git a/drivers/infiniband/hw/hns/hns_roce_ah.c b/drivers/infiniband/hw/hns/hns_roce_ah.c index 9990dc9eb96a..dae4c54f021d 100644 --- a/drivers/infiniband/hw/hns/hns_roce_ah.c +++ b/drivers/infiniband/hw/hns/hns_roce_ah.c @@ -41,6 +41,7 @@ struct ib_ah *hns_roce_create_ah(struct ib_pd *ibpd, struct rdma_ah_attr *ah_attr, + u32 flags, struct ib_udata *udata) { struct hns_roce_dev *hr_dev = to_hr_dev(ibpd->device); diff --git a/drivers/infiniband/hw/hns/hns_roce_device.h b/drivers/infiniband/hw/hns/hns_roce_device.h index 67609cc6a45e..b2c46821c283 100644 --- a/drivers/infiniband/hw/hns/hns_roce_device.h +++ b/drivers/infiniband/hw/hns/hns_roce_device.h @@ -1056,6 +1056,7 @@ void hns_roce_bitmap_free_range(struct hns_roce_bitmap *bitmap, struct ib_ah *hns_roce_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, + u32 flags, struct ib_udata *udata); int hns_roce_query_ah(struct ib_ah *ibah, struct rdma_ah_attr *ah_attr); int hns_roce_destroy_ah(struct ib_ah *ah); diff --git a/drivers/infiniband/hw/mlx4/ah.c b/drivers/infiniband/hw/mlx4/ah.c index e9e3a6f390db..6c0e6bcf1bff 100644 --- a/drivers/infiniband/hw/mlx4/ah.c +++ b/drivers/infiniband/hw/mlx4/ah.c @@ -144,7 +144,7 @@ static struct ib_ah *create_iboe_ah(struct ib_pd *pd, } struct ib_ah *mlx4_ib_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, - struct ib_udata *udata) + u32 flags, struct ib_udata *udata) { struct mlx4_ib_ah *ah; @@ -189,7 +189,7 @@ struct ib_ah *mlx4_ib_create_ah_slave(struct ib_pd *pd, slave_attr.grh.sgid_attr = NULL; slave_attr.grh.sgid_index = slave_sgid_index; - ah = mlx4_ib_create_ah(pd, &slave_attr, NULL); + ah = mlx4_ib_create_ah(pd, &slave_attr, 0, NULL); if (IS_ERR(ah)) return ah; diff --git a/drivers/infiniband/hw/mlx4/mad.c b/drivers/infiniband/hw/mlx4/mad.c index 8942f5f7f04d..b227f7c2fca2 100644 --- a/drivers/infiniband/hw/mlx4/mad.c +++ b/drivers/infiniband/hw/mlx4/mad.c @@ -202,7 +202,7 @@ static void update_sm_ah(struct mlx4_ib_dev *dev, u8 port_num, u16 lid, u8 sl) rdma_ah_set_port_num(&ah_attr, port_num); new_ah = rdma_create_ah(dev->send_agent[port_num - 1][0]->qp->pd, - &ah_attr); + &ah_attr, 0); if (IS_ERR(new_ah)) return; @@ -567,7 +567,7 @@ int mlx4_ib_send_to_slave(struct mlx4_ib_dev *dev, int slave, u8 port, return -EINVAL; rdma_ah_set_grh(&attr, &dgid, 0, 0, 0, 0); } - ah = rdma_create_ah(tun_ctx->pd, &attr); + ah = rdma_create_ah(tun_ctx->pd, &attr, 0); if (IS_ERR(ah)) return -ENOMEM; diff --git a/drivers/infiniband/hw/mlx4/mlx4_ib.h b/drivers/infiniband/hw/mlx4/mlx4_ib.h index 5cb52424912e..7b827b185d5f 100644 --- a/drivers/infiniband/hw/mlx4/mlx4_ib.h +++ b/drivers/infiniband/hw/mlx4/mlx4_ib.h @@ -754,7 +754,7 @@ void __mlx4_ib_cq_clean(struct mlx4_ib_cq *cq, u32 qpn, struct mlx4_ib_srq *srq) void mlx4_ib_cq_clean(struct mlx4_ib_cq *cq, u32 qpn, struct mlx4_ib_srq *srq); struct ib_ah *mlx4_ib_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, - struct ib_udata *udata); + u32 flags, struct ib_udata *udata); struct ib_ah *mlx4_ib_create_ah_slave(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, int slave_sgid_index, u8 *s_mac, diff --git a/drivers/infiniband/hw/mlx5/ah.c b/drivers/infiniband/hw/mlx5/ah.c index ffd03bf1a71e..536c2c067b94 100644 --- a/drivers/infiniband/hw/mlx5/ah.c +++ b/drivers/infiniband/hw/mlx5/ah.c @@ -72,7 +72,7 @@ static struct ib_ah *create_ib_ah(struct mlx5_ib_dev *dev, } struct ib_ah *mlx5_ib_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, - struct ib_udata *udata) + u32 flags, struct ib_udata *udata) { struct mlx5_ib_ah *ah; diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index 9b4e2554889a..ebf700298acb 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -1042,7 +1042,7 @@ int mlx5_MAD_IFC(struct mlx5_ib_dev *dev, int ignore_mkey, int ignore_bkey, u8 port, const struct ib_wc *in_wc, const struct ib_grh *in_grh, const void *in_mad, void *response_mad); struct ib_ah *mlx5_ib_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, - struct ib_udata *udata); + u32 flags, struct ib_udata *udata); int mlx5_ib_query_ah(struct ib_ah *ibah, struct rdma_ah_attr *ah_attr); int mlx5_ib_destroy_ah(struct ib_ah *ah); struct ib_srq *mlx5_ib_create_srq(struct ib_pd *pd, diff --git a/drivers/infiniband/hw/mthca/mthca_mad.c b/drivers/infiniband/hw/mthca/mthca_mad.c index 2e5dc0a67cfc..4a5916d8f7da 100644 --- a/drivers/infiniband/hw/mthca/mthca_mad.c +++ b/drivers/infiniband/hw/mthca/mthca_mad.c @@ -89,7 +89,7 @@ static void update_sm_ah(struct mthca_dev *dev, rdma_ah_set_port_num(&ah_attr, port_num); new_ah = rdma_create_ah(dev->send_agent[port_num - 1][0]->qp->pd, - &ah_attr); + &ah_attr, 0); if (IS_ERR(new_ah)) return; diff --git a/drivers/infiniband/hw/mthca/mthca_provider.c b/drivers/infiniband/hw/mthca/mthca_provider.c index 443521cf8107..c8a7c3765558 100644 --- a/drivers/infiniband/hw/mthca/mthca_provider.c +++ b/drivers/infiniband/hw/mthca/mthca_provider.c @@ -412,6 +412,7 @@ static int mthca_dealloc_pd(struct ib_pd *pd) static struct ib_ah *mthca_ah_create(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, + u32 flags, struct ib_udata *udata) { diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_ah.c b/drivers/infiniband/hw/ocrdma/ocrdma_ah.c index 58188fe5aed2..62c502220724 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_ah.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_ah.c @@ -157,7 +157,7 @@ static inline int set_av_attr(struct ocrdma_dev *dev, struct ocrdma_ah *ah, } struct ib_ah *ocrdma_create_ah(struct ib_pd *ibpd, struct rdma_ah_attr *attr, - struct ib_udata *udata) + u32 flags, struct ib_udata *udata) { u32 *ahid_addr; int status; diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_ah.h b/drivers/infiniband/hw/ocrdma/ocrdma_ah.h index c0c32c9b80ae..c407022d885c 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_ah.h +++ b/drivers/infiniband/hw/ocrdma/ocrdma_ah.h @@ -52,7 +52,7 @@ enum { }; struct ib_ah *ocrdma_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, - struct ib_udata *udata); + u32 flags, struct ib_udata *udata); int ocrdma_destroy_ah(struct ib_ah *ah); int ocrdma_query_ah(struct ib_ah *ah, struct rdma_ah_attr *ah_attr); diff --git a/drivers/infiniband/hw/qedr/verbs.c b/drivers/infiniband/hw/qedr/verbs.c index 8056121e9f69..bb53c08b86c1 100644 --- a/drivers/infiniband/hw/qedr/verbs.c +++ b/drivers/infiniband/hw/qedr/verbs.c @@ -2615,7 +2615,7 @@ int qedr_destroy_qp(struct ib_qp *ibqp) } struct ib_ah *qedr_create_ah(struct ib_pd *ibpd, struct rdma_ah_attr *attr, - struct ib_udata *udata) + u32 flags, struct ib_udata *udata) { struct qedr_ah *ah; diff --git a/drivers/infiniband/hw/qedr/verbs.h b/drivers/infiniband/hw/qedr/verbs.h index 0b7d0124b16c..80ac864dbc78 100644 --- a/drivers/infiniband/hw/qedr/verbs.h +++ b/drivers/infiniband/hw/qedr/verbs.h @@ -76,7 +76,7 @@ int qedr_destroy_srq(struct ib_srq *ibsrq); int qedr_post_srq_recv(struct ib_srq *ibsrq, const struct ib_recv_wr *wr, const struct ib_recv_wr **bad_recv_wr); struct ib_ah *qedr_create_ah(struct ib_pd *ibpd, struct rdma_ah_attr *attr, - struct ib_udata *udata); + u32 flags, struct ib_udata *udata); int qedr_destroy_ah(struct ib_ah *ibah); int qedr_dereg_mr(struct ib_mr *); diff --git a/drivers/infiniband/hw/qib/qib_verbs.c b/drivers/infiniband/hw/qib/qib_verbs.c index 611a6b5ef83f..276304f611ab 100644 --- a/drivers/infiniband/hw/qib/qib_verbs.c +++ b/drivers/infiniband/hw/qib/qib_verbs.c @@ -1362,7 +1362,7 @@ struct ib_ah *qib_create_qp0_ah(struct qib_ibport *ibp, u16 dlid) rcu_read_lock(); qp0 = rcu_dereference(ibp->rvp.qp[0]); if (qp0) - ah = rdma_create_ah(qp0->ibqp.pd, &attr); + ah = rdma_create_ah(qp0->ibqp.pd, &attr, 0); rcu_read_unlock(); return ah; } diff --git a/drivers/infiniband/hw/usnic/usnic_ib_verbs.c b/drivers/infiniband/hw/usnic/usnic_ib_verbs.c index 0b91ff36768a..fa477140f548 100644 --- a/drivers/infiniband/hw/usnic/usnic_ib_verbs.c +++ b/drivers/infiniband/hw/usnic/usnic_ib_verbs.c @@ -760,6 +760,7 @@ int usnic_ib_mmap(struct ib_ucontext *context, /* In ib callbacks section - Start of stub funcs */ struct ib_ah *usnic_ib_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, + u32 flags, struct ib_udata *udata) { diff --git a/drivers/infiniband/hw/usnic/usnic_ib_verbs.h b/drivers/infiniband/hw/usnic/usnic_ib_verbs.h index 2a2c9beb715f..066f53636d0b 100644 --- a/drivers/infiniband/hw/usnic/usnic_ib_verbs.h +++ b/drivers/infiniband/hw/usnic/usnic_ib_verbs.h @@ -77,6 +77,7 @@ int usnic_ib_mmap(struct ib_ucontext *context, struct vm_area_struct *vma); struct ib_ah *usnic_ib_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, + u32 flags, struct ib_udata *udata); int usnic_ib_destroy_ah(struct ib_ah *ah); diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.c b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.c index f4cb5cf26006..08ea4b9e4c6c 100644 --- a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.c +++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.c @@ -533,11 +533,12 @@ int pvrdma_dealloc_pd(struct ib_pd *pd) * @pd: the protection domain * @ah_attr: the attributes of the AH * @udata: user data blob + * @flags: create address handle flags (see enum rdma_create_ah_flags) * * @return: the ib_ah pointer on success, otherwise errno. */ struct ib_ah *pvrdma_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, - struct ib_udata *udata) + u32 flags, struct ib_udata *udata) { struct pvrdma_dev *dev = to_vdev(pd->device); struct pvrdma_ah *ah; diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.h b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.h index b2e3ab50cb08..d872e6aa159c 100644 --- a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.h +++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.h @@ -420,7 +420,7 @@ int pvrdma_destroy_cq(struct ib_cq *cq); int pvrdma_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *wc); int pvrdma_req_notify_cq(struct ib_cq *cq, enum ib_cq_notify_flags flags); struct ib_ah *pvrdma_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, - struct ib_udata *udata); + u32 flags, struct ib_udata *udata); int pvrdma_destroy_ah(struct ib_ah *ah); struct ib_srq *pvrdma_create_srq(struct ib_pd *pd, diff --git a/drivers/infiniband/sw/rdmavt/ah.c b/drivers/infiniband/sw/rdmavt/ah.c index 084bb4baebb5..43e46877d300 100644 --- a/drivers/infiniband/sw/rdmavt/ah.c +++ b/drivers/infiniband/sw/rdmavt/ah.c @@ -91,6 +91,7 @@ EXPORT_SYMBOL(rvt_check_ah); * rvt_create_ah - create an address handle * @pd: the protection domain * @ah_attr: the attributes of the AH + * @create_flags: create address handle flags (see enum rdma_create_ah_flags) * @udata: pointer to user's input output buffer information. * * This may be called from interrupt context. @@ -99,6 +100,7 @@ EXPORT_SYMBOL(rvt_check_ah); */ struct ib_ah *rvt_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, + u32 create_flags, struct ib_udata *udata) { struct rvt_ah *ah; diff --git a/drivers/infiniband/sw/rdmavt/ah.h b/drivers/infiniband/sw/rdmavt/ah.h index 25271b48a683..70f8d089f740 100644 --- a/drivers/infiniband/sw/rdmavt/ah.h +++ b/drivers/infiniband/sw/rdmavt/ah.h @@ -52,6 +52,7 @@ struct ib_ah *rvt_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, + u32 create_flags, struct ib_udata *udata); int rvt_destroy_ah(struct ib_ah *ibah); int rvt_modify_ah(struct ib_ah *ibah, struct rdma_ah_attr *ah_attr); diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.c b/drivers/infiniband/sw/rxe/rxe_verbs.c index 8d15b1e11a2f..9c0ae9a26319 100644 --- a/drivers/infiniband/sw/rxe/rxe_verbs.c +++ b/drivers/infiniband/sw/rxe/rxe_verbs.c @@ -219,6 +219,7 @@ static void rxe_init_av(struct rxe_dev *rxe, struct rdma_ah_attr *attr, static struct ib_ah *rxe_create_ah(struct ib_pd *ibpd, struct rdma_ah_attr *attr, + u32 flags, struct ib_udata *udata) { diff --git a/drivers/infiniband/ulp/ipoib/ipoib_ib.c b/drivers/infiniband/ulp/ipoib/ipoib_ib.c index 9006a13af1de..1ecf21006f72 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_ib.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_ib.c @@ -66,7 +66,7 @@ struct ipoib_ah *ipoib_create_ah(struct net_device *dev, ah->last_send = 0; kref_init(&ah->ref); - vah = rdma_create_ah(pd, attr); + vah = rdma_create_ah(pd, attr, RDMA_CREATE_AH_SLEEPABLE); if (IS_ERR(vah)) { kfree(ah); ah = (struct ipoib_ah *)vah; diff --git a/drivers/infiniband/ulp/opa_vnic/opa_vnic_vema.c b/drivers/infiniband/ulp/opa_vnic/opa_vnic_vema.c index d119d9afa845..2f4f63847c10 100644 --- a/drivers/infiniband/ulp/opa_vnic/opa_vnic_vema.c +++ b/drivers/infiniband/ulp/opa_vnic/opa_vnic_vema.c @@ -777,7 +777,7 @@ void opa_vnic_vema_send_trap(struct opa_vnic_adapter *adapter, } rdma_ah_set_dlid(&ah_attr, trap_lid); - ah = rdma_create_ah(port->mad_agent->qp->pd, &ah_attr); + ah = rdma_create_ah(port->mad_agent->qp->pd, &ah_attr, 0); if (IS_ERR(ah)) { c_err("%s:Couldn't create new AH = %p\n", __func__, ah); c_err("%s:dlid = %d, sl = %d, port = %d\n", __func__, diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h index 0ec15d673d92..86ae878e1de4 100644 --- a/include/rdma/ib_verbs.h +++ b/include/rdma/ib_verbs.h @@ -2377,7 +2377,7 @@ struct ib_device_ops { struct ib_udata *udata); int (*dealloc_pd)(struct ib_pd *pd); struct ib_ah *(*create_ah)(struct ib_pd *pd, - struct rdma_ah_attr *ah_attr, + struct rdma_ah_attr *ah_attr, u32 flags, struct ib_udata *udata); int (*modify_ah)(struct ib_ah *ah, struct rdma_ah_attr *ah_attr); int (*query_ah)(struct ib_ah *ah, struct rdma_ah_attr *ah_attr); @@ -3151,15 +3151,22 @@ struct ib_pd *__ib_alloc_pd(struct ib_device *device, unsigned int flags, __ib_alloc_pd((device), (flags), KBUILD_MODNAME) void ib_dealloc_pd(struct ib_pd *pd); +enum rdma_create_ah_flags { + /* In a sleepable context */ + RDMA_CREATE_AH_SLEEPABLE = BIT(0), +}; + /** * rdma_create_ah - Creates an address handle for the given address vector. * @pd: The protection domain associated with the address handle. * @ah_attr: The attributes of the address vector. + * @flags: Create address handle flags (see enum rdma_create_ah_flags). * * The address handle is used to reference a local or global destination * in all UD QP post sends. */ -struct ib_ah *rdma_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr); +struct ib_ah *rdma_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, + u32 flags); /** * rdma_create_user_ah - Creates an address handle for the given address vector. -- cgit v1.2.3-59-g8ed1b From 2553ba217eea37dc6291635ecddb883fb5c36a8b Mon Sep 17 00:00:00 2001 From: Gal Pressman Date: Wed, 12 Dec 2018 11:09:06 +0200 Subject: RDMA: Mark if destroy address handle is in a sleepable context Introduce a 'flags' field to destroy address handle callback and add a flag that marks whether the callback is executed in an atomic context or not. This will allow drivers to wait for completion instead of polling for it when it is allowed. Signed-off-by: Gal Pressman Signed-off-by: Jason Gunthorpe --- drivers/infiniband/core/agent.c | 4 ++-- drivers/infiniband/core/cm.c | 4 ++-- drivers/infiniband/core/mad_rmpp.c | 11 ++++++----- drivers/infiniband/core/sa_query.c | 2 +- drivers/infiniband/core/user_mad.c | 4 ++-- drivers/infiniband/core/uverbs_cmd.c | 2 +- drivers/infiniband/core/uverbs_std_types.c | 3 ++- drivers/infiniband/core/verbs.c | 6 ++++-- drivers/infiniband/hw/bnxt_re/ib_verbs.c | 2 +- drivers/infiniband/hw/bnxt_re/ib_verbs.h | 2 +- drivers/infiniband/hw/hns/hns_roce_ah.c | 2 +- drivers/infiniband/hw/hns/hns_roce_device.h | 2 +- drivers/infiniband/hw/mlx4/ah.c | 2 +- drivers/infiniband/hw/mlx4/mad.c | 24 ++++++++++++------------ drivers/infiniband/hw/mlx4/mlx4_ib.h | 2 +- drivers/infiniband/hw/mlx5/ah.c | 2 +- drivers/infiniband/hw/mlx5/mlx5_ib.h | 2 +- drivers/infiniband/hw/mthca/mthca_mad.c | 5 +++-- drivers/infiniband/hw/mthca/mthca_provider.c | 2 +- drivers/infiniband/hw/ocrdma/ocrdma_ah.c | 2 +- drivers/infiniband/hw/ocrdma/ocrdma_ah.h | 2 +- drivers/infiniband/hw/qedr/verbs.c | 2 +- drivers/infiniband/hw/qedr/verbs.h | 2 +- drivers/infiniband/hw/qib/qib_mad.c | 3 ++- drivers/infiniband/hw/usnic/usnic_ib_verbs.c | 2 +- drivers/infiniband/hw/usnic/usnic_ib_verbs.h | 2 +- drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.c | 3 ++- drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.h | 2 +- drivers/infiniband/sw/rdmavt/ah.c | 3 ++- drivers/infiniband/sw/rdmavt/ah.h | 2 +- drivers/infiniband/sw/rdmavt/mad.c | 3 ++- drivers/infiniband/sw/rxe/rxe_verbs.c | 2 +- drivers/infiniband/ulp/ipoib/ipoib_ib.c | 2 +- drivers/infiniband/ulp/opa_vnic/opa_vnic_vema.c | 6 +++--- drivers/infiniband/ulp/srpt/ib_srpt.c | 4 ++-- include/rdma/ib_verbs.h | 10 ++++++++-- 36 files changed, 75 insertions(+), 60 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/core/agent.c b/drivers/infiniband/core/agent.c index 324ef85a13b6..f82b4260de42 100644 --- a/drivers/infiniband/core/agent.c +++ b/drivers/infiniband/core/agent.c @@ -137,13 +137,13 @@ void agent_send_response(const struct ib_mad_hdr *mad_hdr, const struct ib_grh * err2: ib_free_send_mad(send_buf); err1: - rdma_destroy_ah(ah); + rdma_destroy_ah(ah, RDMA_DESTROY_AH_SLEEPABLE); } static void agent_send_handler(struct ib_mad_agent *mad_agent, struct ib_mad_send_wc *mad_send_wc) { - rdma_destroy_ah(mad_send_wc->send_buf->ah); + rdma_destroy_ah(mad_send_wc->send_buf->ah, RDMA_DESTROY_AH_SLEEPABLE); ib_free_send_mad(mad_send_wc->send_buf); } diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c index cf5b3c4314bb..37980c7564c0 100644 --- a/drivers/infiniband/core/cm.c +++ b/drivers/infiniband/core/cm.c @@ -355,7 +355,7 @@ static int cm_alloc_msg(struct cm_id_private *cm_id_priv, GFP_ATOMIC, IB_MGMT_BASE_VERSION); if (IS_ERR(m)) { - rdma_destroy_ah(ah); + rdma_destroy_ah(ah, 0); ret = PTR_ERR(m); goto out; } @@ -400,7 +400,7 @@ static int cm_create_response_msg_ah(struct cm_port *port, static void cm_free_msg(struct ib_mad_send_buf *msg) { if (msg->ah) - rdma_destroy_ah(msg->ah); + rdma_destroy_ah(msg->ah, 0); if (msg->context[0]) cm_deref_id(msg->context[0]); ib_free_send_mad(msg); diff --git a/drivers/infiniband/core/mad_rmpp.c b/drivers/infiniband/core/mad_rmpp.c index e5cf09c66fe6..5ec57abc0849 100644 --- a/drivers/infiniband/core/mad_rmpp.c +++ b/drivers/infiniband/core/mad_rmpp.c @@ -81,7 +81,7 @@ static void destroy_rmpp_recv(struct mad_rmpp_recv *rmpp_recv) { deref_rmpp_recv(rmpp_recv); wait_for_completion(&rmpp_recv->comp); - rdma_destroy_ah(rmpp_recv->ah); + rdma_destroy_ah(rmpp_recv->ah, RDMA_DESTROY_AH_SLEEPABLE); kfree(rmpp_recv); } @@ -171,7 +171,7 @@ static struct ib_mad_send_buf *alloc_response_msg(struct ib_mad_agent *agent, hdr_len, 0, GFP_KERNEL, IB_MGMT_BASE_VERSION); if (IS_ERR(msg)) - rdma_destroy_ah(ah); + rdma_destroy_ah(ah, RDMA_DESTROY_AH_SLEEPABLE); else { msg->ah = ah; msg->context[0] = ah; @@ -201,7 +201,7 @@ static void ack_ds_ack(struct ib_mad_agent_private *agent, ret = ib_post_send_mad(msg, NULL); if (ret) { - rdma_destroy_ah(msg->ah); + rdma_destroy_ah(msg->ah, RDMA_DESTROY_AH_SLEEPABLE); ib_free_send_mad(msg); } } @@ -209,7 +209,8 @@ static void ack_ds_ack(struct ib_mad_agent_private *agent, void ib_rmpp_send_handler(struct ib_mad_send_wc *mad_send_wc) { if (mad_send_wc->send_buf->context[0] == mad_send_wc->send_buf->ah) - rdma_destroy_ah(mad_send_wc->send_buf->ah); + rdma_destroy_ah(mad_send_wc->send_buf->ah, + RDMA_DESTROY_AH_SLEEPABLE); ib_free_send_mad(mad_send_wc->send_buf); } @@ -237,7 +238,7 @@ static void nack_recv(struct ib_mad_agent_private *agent, ret = ib_post_send_mad(msg, NULL); if (ret) { - rdma_destroy_ah(msg->ah); + rdma_destroy_ah(msg->ah, RDMA_DESTROY_AH_SLEEPABLE); ib_free_send_mad(msg); } } diff --git a/drivers/infiniband/core/sa_query.c b/drivers/infiniband/core/sa_query.c index fb96d9df1fdd..97e6d7b69abf 100644 --- a/drivers/infiniband/core/sa_query.c +++ b/drivers/infiniband/core/sa_query.c @@ -1147,7 +1147,7 @@ static void free_sm_ah(struct kref *kref) { struct ib_sa_sm_ah *sm_ah = container_of(kref, struct ib_sa_sm_ah, ref); - rdma_destroy_ah(sm_ah->ah); + rdma_destroy_ah(sm_ah->ah, 0); kfree(sm_ah); } diff --git a/drivers/infiniband/core/user_mad.c b/drivers/infiniband/core/user_mad.c index f55f48f6b272..ee3b1f7482b6 100644 --- a/drivers/infiniband/core/user_mad.c +++ b/drivers/infiniband/core/user_mad.c @@ -205,7 +205,7 @@ static void send_handler(struct ib_mad_agent *agent, struct ib_umad_packet *packet = send_wc->send_buf->context[0]; dequeue_send(file, packet); - rdma_destroy_ah(packet->msg->ah); + rdma_destroy_ah(packet->msg->ah, RDMA_DESTROY_AH_SLEEPABLE); ib_free_send_mad(packet->msg); if (send_wc->status == IB_WC_RESP_TIMEOUT_ERR) { @@ -621,7 +621,7 @@ err_send: err_msg: ib_free_send_mad(packet->msg); err_ah: - rdma_destroy_ah(ah); + rdma_destroy_ah(ah, RDMA_DESTROY_AH_SLEEPABLE); err_up: mutex_unlock(&file->mutex); err: diff --git a/drivers/infiniband/core/uverbs_cmd.c b/drivers/infiniband/core/uverbs_cmd.c index 17be53da21b0..4dc454d35f32 100644 --- a/drivers/infiniband/core/uverbs_cmd.c +++ b/drivers/infiniband/core/uverbs_cmd.c @@ -2486,7 +2486,7 @@ static int ib_uverbs_create_ah(struct uverbs_attr_bundle *attrs) return uobj_alloc_commit(uobj); err_copy: - rdma_destroy_ah(ah); + rdma_destroy_ah(ah, RDMA_DESTROY_AH_SLEEPABLE); err_put: uobj_put_obj_read(pd); diff --git a/drivers/infiniband/core/uverbs_std_types.c b/drivers/infiniband/core/uverbs_std_types.c index 8e975f5b1f01..cbc72312eb41 100644 --- a/drivers/infiniband/core/uverbs_std_types.c +++ b/drivers/infiniband/core/uverbs_std_types.c @@ -42,7 +42,8 @@ static int uverbs_free_ah(struct ib_uobject *uobject, enum rdma_remove_reason why) { - return rdma_destroy_ah((struct ib_ah *)uobject->object); + return rdma_destroy_ah((struct ib_ah *)uobject->object, + RDMA_DESTROY_AH_SLEEPABLE); } static int uverbs_free_flow(struct ib_uobject *uobject, diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c index 4e2df239447e..ac011836bb54 100644 --- a/drivers/infiniband/core/verbs.c +++ b/drivers/infiniband/core/verbs.c @@ -925,14 +925,16 @@ int rdma_query_ah(struct ib_ah *ah, struct rdma_ah_attr *ah_attr) } EXPORT_SYMBOL(rdma_query_ah); -int rdma_destroy_ah(struct ib_ah *ah) +int rdma_destroy_ah(struct ib_ah *ah, u32 flags) { const struct ib_gid_attr *sgid_attr = ah->sgid_attr; struct ib_pd *pd; int ret; + might_sleep_if(flags & RDMA_DESTROY_AH_SLEEPABLE); + pd = ah->pd; - ret = ah->device->ops.destroy_ah(ah); + ret = ah->device->ops.destroy_ah(ah, flags); if (!ret) { atomic_dec(&pd->usecnt); if (sgid_attr) diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.c b/drivers/infiniband/hw/bnxt_re/ib_verbs.c index 4b37bc3edc81..ece45e7aee20 100644 --- a/drivers/infiniband/hw/bnxt_re/ib_verbs.c +++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.c @@ -647,7 +647,7 @@ fail: } /* Address Handles */ -int bnxt_re_destroy_ah(struct ib_ah *ib_ah) +int bnxt_re_destroy_ah(struct ib_ah *ib_ah, u32 flags) { struct bnxt_re_ah *ah = container_of(ib_ah, struct bnxt_re_ah, ib_ah); struct bnxt_re_dev *rdev = ah->rdev; diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.h b/drivers/infiniband/hw/bnxt_re/ib_verbs.h index 6675db4b4ba4..c4af72604b4f 100644 --- a/drivers/infiniband/hw/bnxt_re/ib_verbs.h +++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.h @@ -173,7 +173,7 @@ struct ib_ah *bnxt_re_create_ah(struct ib_pd *pd, struct ib_udata *udata); int bnxt_re_modify_ah(struct ib_ah *ah, struct rdma_ah_attr *ah_attr); int bnxt_re_query_ah(struct ib_ah *ah, struct rdma_ah_attr *ah_attr); -int bnxt_re_destroy_ah(struct ib_ah *ah); +int bnxt_re_destroy_ah(struct ib_ah *ah, u32 flags); struct ib_srq *bnxt_re_create_srq(struct ib_pd *pd, struct ib_srq_init_attr *srq_init_attr, struct ib_udata *udata); diff --git a/drivers/infiniband/hw/hns/hns_roce_ah.c b/drivers/infiniband/hw/hns/hns_roce_ah.c index dae4c54f021d..b3c8c45ec1e3 100644 --- a/drivers/infiniband/hw/hns/hns_roce_ah.c +++ b/drivers/infiniband/hw/hns/hns_roce_ah.c @@ -111,7 +111,7 @@ int hns_roce_query_ah(struct ib_ah *ibah, struct rdma_ah_attr *ah_attr) return 0; } -int hns_roce_destroy_ah(struct ib_ah *ah) +int hns_roce_destroy_ah(struct ib_ah *ah, u32 flags) { kfree(to_hr_ah(ah)); diff --git a/drivers/infiniband/hw/hns/hns_roce_device.h b/drivers/infiniband/hw/hns/hns_roce_device.h index b2c46821c283..509e467843f6 100644 --- a/drivers/infiniband/hw/hns/hns_roce_device.h +++ b/drivers/infiniband/hw/hns/hns_roce_device.h @@ -1059,7 +1059,7 @@ struct ib_ah *hns_roce_create_ah(struct ib_pd *pd, u32 flags, struct ib_udata *udata); int hns_roce_query_ah(struct ib_ah *ibah, struct rdma_ah_attr *ah_attr); -int hns_roce_destroy_ah(struct ib_ah *ah); +int hns_roce_destroy_ah(struct ib_ah *ah, u32 flags); struct ib_pd *hns_roce_alloc_pd(struct ib_device *ib_dev, struct ib_ucontext *context, diff --git a/drivers/infiniband/hw/mlx4/ah.c b/drivers/infiniband/hw/mlx4/ah.c index 6c0e6bcf1bff..1672808262ba 100644 --- a/drivers/infiniband/hw/mlx4/ah.c +++ b/drivers/infiniband/hw/mlx4/ah.c @@ -250,7 +250,7 @@ int mlx4_ib_query_ah(struct ib_ah *ibah, struct rdma_ah_attr *ah_attr) return 0; } -int mlx4_ib_destroy_ah(struct ib_ah *ah) +int mlx4_ib_destroy_ah(struct ib_ah *ah, u32 flags) { kfree(to_mah(ah)); return 0; diff --git a/drivers/infiniband/hw/mlx4/mad.c b/drivers/infiniband/hw/mlx4/mad.c index b227f7c2fca2..25439da8976c 100644 --- a/drivers/infiniband/hw/mlx4/mad.c +++ b/drivers/infiniband/hw/mlx4/mad.c @@ -208,7 +208,7 @@ static void update_sm_ah(struct mlx4_ib_dev *dev, u8 port_num, u16 lid, u8 sl) spin_lock_irqsave(&dev->sm_lock, flags); if (dev->sm_ah[port_num - 1]) - rdma_destroy_ah(dev->sm_ah[port_num - 1]); + rdma_destroy_ah(dev->sm_ah[port_num - 1], 0); dev->sm_ah[port_num - 1] = new_ah; spin_unlock_irqrestore(&dev->sm_lock, flags); } @@ -584,7 +584,7 @@ int mlx4_ib_send_to_slave(struct mlx4_ib_dev *dev, int slave, u8 port, tun_mad = (struct mlx4_rcv_tunnel_mad *) (tun_qp->tx_ring[tun_tx_ix].buf.addr); if (tun_qp->tx_ring[tun_tx_ix].ah) - rdma_destroy_ah(tun_qp->tx_ring[tun_tx_ix].ah); + rdma_destroy_ah(tun_qp->tx_ring[tun_tx_ix].ah, 0); tun_qp->tx_ring[tun_tx_ix].ah = ah; ib_dma_sync_single_for_cpu(&dev->ib_dev, tun_qp->tx_ring[tun_tx_ix].buf.map, @@ -657,7 +657,7 @@ int mlx4_ib_send_to_slave(struct mlx4_ib_dev *dev, int slave, u8 port, spin_unlock(&tun_qp->tx_lock); tun_qp->tx_ring[tun_tx_ix].ah = NULL; end: - rdma_destroy_ah(ah); + rdma_destroy_ah(ah, 0); return ret; } @@ -1024,7 +1024,7 @@ static void send_handler(struct ib_mad_agent *agent, struct ib_mad_send_wc *mad_send_wc) { if (mad_send_wc->send_buf->context[0]) - rdma_destroy_ah(mad_send_wc->send_buf->context[0]); + rdma_destroy_ah(mad_send_wc->send_buf->context[0], 0); ib_free_send_mad(mad_send_wc->send_buf); } @@ -1079,7 +1079,7 @@ void mlx4_ib_mad_cleanup(struct mlx4_ib_dev *dev) } if (dev->sm_ah[p]) - rdma_destroy_ah(dev->sm_ah[p]); + rdma_destroy_ah(dev->sm_ah[p], 0); } } @@ -1411,7 +1411,7 @@ int mlx4_ib_send_to_wire(struct mlx4_ib_dev *dev, int slave, u8 port, sqp_mad = (struct mlx4_mad_snd_buf *) (sqp->tx_ring[wire_tx_ix].buf.addr); if (sqp->tx_ring[wire_tx_ix].ah) - rdma_destroy_ah(sqp->tx_ring[wire_tx_ix].ah); + rdma_destroy_ah(sqp->tx_ring[wire_tx_ix].ah, 0); sqp->tx_ring[wire_tx_ix].ah = ah; ib_dma_sync_single_for_cpu(&dev->ib_dev, sqp->tx_ring[wire_tx_ix].buf.map, @@ -1450,7 +1450,7 @@ int mlx4_ib_send_to_wire(struct mlx4_ib_dev *dev, int slave, u8 port, spin_unlock(&sqp->tx_lock); sqp->tx_ring[wire_tx_ix].ah = NULL; out: - mlx4_ib_destroy_ah(ah); + mlx4_ib_destroy_ah(ah, 0); return ret; } @@ -1716,7 +1716,7 @@ static void mlx4_ib_free_pv_qp_bufs(struct mlx4_ib_demux_pv_ctx *ctx, tx_buf_size, DMA_TO_DEVICE); kfree(tun_qp->tx_ring[i].buf.addr); if (tun_qp->tx_ring[i].ah) - rdma_destroy_ah(tun_qp->tx_ring[i].ah); + rdma_destroy_ah(tun_qp->tx_ring[i].ah, 0); } kfree(tun_qp->tx_ring); kfree(tun_qp->ring); @@ -1749,7 +1749,7 @@ static void mlx4_ib_tunnel_comp_worker(struct work_struct *work) "wrid=0x%llx, status=0x%x\n", wc.wr_id, wc.status); rdma_destroy_ah(tun_qp->tx_ring[wc.wr_id & - (MLX4_NUM_TUNNEL_BUFS - 1)].ah); + (MLX4_NUM_TUNNEL_BUFS - 1)].ah, 0); tun_qp->tx_ring[wc.wr_id & (MLX4_NUM_TUNNEL_BUFS - 1)].ah = NULL; spin_lock(&tun_qp->tx_lock); @@ -1766,7 +1766,7 @@ static void mlx4_ib_tunnel_comp_worker(struct work_struct *work) ctx->slave, wc.status, wc.wr_id); if (!MLX4_TUN_IS_RECV(wc.wr_id)) { rdma_destroy_ah(tun_qp->tx_ring[wc.wr_id & - (MLX4_NUM_TUNNEL_BUFS - 1)].ah); + (MLX4_NUM_TUNNEL_BUFS - 1)].ah, 0); tun_qp->tx_ring[wc.wr_id & (MLX4_NUM_TUNNEL_BUFS - 1)].ah = NULL; spin_lock(&tun_qp->tx_lock); @@ -1903,7 +1903,7 @@ static void mlx4_ib_sqp_comp_worker(struct work_struct *work) switch (wc.opcode) { case IB_WC_SEND: rdma_destroy_ah(sqp->tx_ring[wc.wr_id & - (MLX4_NUM_TUNNEL_BUFS - 1)].ah); + (MLX4_NUM_TUNNEL_BUFS - 1)].ah, 0); sqp->tx_ring[wc.wr_id & (MLX4_NUM_TUNNEL_BUFS - 1)].ah = NULL; spin_lock(&sqp->tx_lock); @@ -1932,7 +1932,7 @@ static void mlx4_ib_sqp_comp_worker(struct work_struct *work) ctx->slave, wc.status, wc.wr_id); if (!MLX4_TUN_IS_RECV(wc.wr_id)) { rdma_destroy_ah(sqp->tx_ring[wc.wr_id & - (MLX4_NUM_TUNNEL_BUFS - 1)].ah); + (MLX4_NUM_TUNNEL_BUFS - 1)].ah, 0); sqp->tx_ring[wc.wr_id & (MLX4_NUM_TUNNEL_BUFS - 1)].ah = NULL; spin_lock(&sqp->tx_lock); diff --git a/drivers/infiniband/hw/mlx4/mlx4_ib.h b/drivers/infiniband/hw/mlx4/mlx4_ib.h index 7b827b185d5f..e491f3eda6e7 100644 --- a/drivers/infiniband/hw/mlx4/mlx4_ib.h +++ b/drivers/infiniband/hw/mlx4/mlx4_ib.h @@ -760,7 +760,7 @@ struct ib_ah *mlx4_ib_create_ah_slave(struct ib_pd *pd, int slave_sgid_index, u8 *s_mac, u16 vlan_tag); int mlx4_ib_query_ah(struct ib_ah *ibah, struct rdma_ah_attr *ah_attr); -int mlx4_ib_destroy_ah(struct ib_ah *ah); +int mlx4_ib_destroy_ah(struct ib_ah *ah, u32 flags); struct ib_srq *mlx4_ib_create_srq(struct ib_pd *pd, struct ib_srq_init_attr *init_attr, diff --git a/drivers/infiniband/hw/mlx5/ah.c b/drivers/infiniband/hw/mlx5/ah.c index 536c2c067b94..420ae0897333 100644 --- a/drivers/infiniband/hw/mlx5/ah.c +++ b/drivers/infiniband/hw/mlx5/ah.c @@ -131,7 +131,7 @@ int mlx5_ib_query_ah(struct ib_ah *ibah, struct rdma_ah_attr *ah_attr) return 0; } -int mlx5_ib_destroy_ah(struct ib_ah *ah) +int mlx5_ib_destroy_ah(struct ib_ah *ah, u32 flags) { kfree(to_mah(ah)); return 0; diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index ebf700298acb..96e8fa1109f5 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -1044,7 +1044,7 @@ int mlx5_MAD_IFC(struct mlx5_ib_dev *dev, int ignore_mkey, int ignore_bkey, struct ib_ah *mlx5_ib_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, u32 flags, struct ib_udata *udata); int mlx5_ib_query_ah(struct ib_ah *ibah, struct rdma_ah_attr *ah_attr); -int mlx5_ib_destroy_ah(struct ib_ah *ah); +int mlx5_ib_destroy_ah(struct ib_ah *ah, u32 flags); struct ib_srq *mlx5_ib_create_srq(struct ib_pd *pd, struct ib_srq_init_attr *init_attr, struct ib_udata *udata); diff --git a/drivers/infiniband/hw/mthca/mthca_mad.c b/drivers/infiniband/hw/mthca/mthca_mad.c index 4a5916d8f7da..7ad517da4917 100644 --- a/drivers/infiniband/hw/mthca/mthca_mad.c +++ b/drivers/infiniband/hw/mthca/mthca_mad.c @@ -95,7 +95,7 @@ static void update_sm_ah(struct mthca_dev *dev, spin_lock_irqsave(&dev->sm_lock, flags); if (dev->sm_ah[port_num - 1]) - rdma_destroy_ah(dev->sm_ah[port_num - 1]); + rdma_destroy_ah(dev->sm_ah[port_num - 1], 0); dev->sm_ah[port_num - 1] = new_ah; spin_unlock_irqrestore(&dev->sm_lock, flags); } @@ -347,6 +347,7 @@ void mthca_free_agents(struct mthca_dev *dev) } if (dev->sm_ah[p]) - rdma_destroy_ah(dev->sm_ah[p]); + rdma_destroy_ah(dev->sm_ah[p], + RDMA_DESTROY_AH_SLEEPABLE); } } diff --git a/drivers/infiniband/hw/mthca/mthca_provider.c b/drivers/infiniband/hw/mthca/mthca_provider.c index c8a7c3765558..82cb6b71ac7c 100644 --- a/drivers/infiniband/hw/mthca/mthca_provider.c +++ b/drivers/infiniband/hw/mthca/mthca_provider.c @@ -432,7 +432,7 @@ static struct ib_ah *mthca_ah_create(struct ib_pd *pd, return &ah->ibah; } -static int mthca_ah_destroy(struct ib_ah *ah) +static int mthca_ah_destroy(struct ib_ah *ah, u32 flags) { mthca_destroy_ah(to_mdev(ah->device), to_mah(ah)); kfree(ah); diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_ah.c b/drivers/infiniband/hw/ocrdma/ocrdma_ah.c index 62c502220724..a7295322efbc 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_ah.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_ah.c @@ -219,7 +219,7 @@ av_err: return ERR_PTR(status); } -int ocrdma_destroy_ah(struct ib_ah *ibah) +int ocrdma_destroy_ah(struct ib_ah *ibah, u32 flags) { struct ocrdma_ah *ah = get_ocrdma_ah(ibah); struct ocrdma_dev *dev = get_ocrdma_dev(ibah->device); diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_ah.h b/drivers/infiniband/hw/ocrdma/ocrdma_ah.h index c407022d885c..eb996e14b520 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_ah.h +++ b/drivers/infiniband/hw/ocrdma/ocrdma_ah.h @@ -53,7 +53,7 @@ enum { struct ib_ah *ocrdma_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, u32 flags, struct ib_udata *udata); -int ocrdma_destroy_ah(struct ib_ah *ah); +int ocrdma_destroy_ah(struct ib_ah *ah, u32 flags); int ocrdma_query_ah(struct ib_ah *ah, struct rdma_ah_attr *ah_attr); int ocrdma_process_mad(struct ib_device *, diff --git a/drivers/infiniband/hw/qedr/verbs.c b/drivers/infiniband/hw/qedr/verbs.c index bb53c08b86c1..b342a70e2814 100644 --- a/drivers/infiniband/hw/qedr/verbs.c +++ b/drivers/infiniband/hw/qedr/verbs.c @@ -2628,7 +2628,7 @@ struct ib_ah *qedr_create_ah(struct ib_pd *ibpd, struct rdma_ah_attr *attr, return &ah->ibah; } -int qedr_destroy_ah(struct ib_ah *ibah) +int qedr_destroy_ah(struct ib_ah *ibah, u32 flags) { struct qedr_ah *ah = get_qedr_ah(ibah); diff --git a/drivers/infiniband/hw/qedr/verbs.h b/drivers/infiniband/hw/qedr/verbs.h index 80ac864dbc78..1852b7012bf4 100644 --- a/drivers/infiniband/hw/qedr/verbs.h +++ b/drivers/infiniband/hw/qedr/verbs.h @@ -77,7 +77,7 @@ int qedr_post_srq_recv(struct ib_srq *ibsrq, const struct ib_recv_wr *wr, const struct ib_recv_wr **bad_recv_wr); struct ib_ah *qedr_create_ah(struct ib_pd *ibpd, struct rdma_ah_attr *attr, u32 flags, struct ib_udata *udata); -int qedr_destroy_ah(struct ib_ah *ibah); +int qedr_destroy_ah(struct ib_ah *ibah, u32 flags); int qedr_dereg_mr(struct ib_mr *); struct ib_mr *qedr_get_dma_mr(struct ib_pd *, int acc); diff --git a/drivers/infiniband/hw/qib/qib_mad.c b/drivers/infiniband/hw/qib/qib_mad.c index 4845d000c22f..f92faf5ec369 100644 --- a/drivers/infiniband/hw/qib/qib_mad.c +++ b/drivers/infiniband/hw/qib/qib_mad.c @@ -2494,5 +2494,6 @@ void qib_notify_free_mad_agent(struct rvt_dev_info *rdi, int port_idx) del_timer_sync(&dd->pport[port_idx].cong_stats.timer); if (dd->pport[port_idx].ibport_data.smi_ah) - rdma_destroy_ah(&dd->pport[port_idx].ibport_data.smi_ah->ibah); + rdma_destroy_ah(&dd->pport[port_idx].ibport_data.smi_ah->ibah, + RDMA_DESTROY_AH_SLEEPABLE); } diff --git a/drivers/infiniband/hw/usnic/usnic_ib_verbs.c b/drivers/infiniband/hw/usnic/usnic_ib_verbs.c index fa477140f548..1b9f69d65111 100644 --- a/drivers/infiniband/hw/usnic/usnic_ib_verbs.c +++ b/drivers/infiniband/hw/usnic/usnic_ib_verbs.c @@ -768,7 +768,7 @@ struct ib_ah *usnic_ib_create_ah(struct ib_pd *pd, return ERR_PTR(-EPERM); } -int usnic_ib_destroy_ah(struct ib_ah *ah) +int usnic_ib_destroy_ah(struct ib_ah *ah, u32 flags) { usnic_dbg("\n"); return -EINVAL; diff --git a/drivers/infiniband/hw/usnic/usnic_ib_verbs.h b/drivers/infiniband/hw/usnic/usnic_ib_verbs.h index 066f53636d0b..e33144261b9a 100644 --- a/drivers/infiniband/hw/usnic/usnic_ib_verbs.h +++ b/drivers/infiniband/hw/usnic/usnic_ib_verbs.h @@ -80,7 +80,7 @@ struct ib_ah *usnic_ib_create_ah(struct ib_pd *pd, u32 flags, struct ib_udata *udata); -int usnic_ib_destroy_ah(struct ib_ah *ah); +int usnic_ib_destroy_ah(struct ib_ah *ah, u32 flags); int usnic_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr, const struct ib_send_wr **bad_wr); int usnic_ib_post_recv(struct ib_qp *ibqp, const struct ib_recv_wr *wr, diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.c b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.c index 08ea4b9e4c6c..4d238d0e484b 100644 --- a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.c +++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.c @@ -582,10 +582,11 @@ struct ib_ah *pvrdma_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, /** * pvrdma_destroy_ah - destroy an address handle * @ah: the address handle to destroyed + * @flags: destroy address handle flags (see enum rdma_destroy_ah_flags) * * @return: 0 on success. */ -int pvrdma_destroy_ah(struct ib_ah *ah) +int pvrdma_destroy_ah(struct ib_ah *ah, u32 flags) { struct pvrdma_dev *dev = to_vdev(ah->device); diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.h b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.h index d872e6aa159c..f7f758d60110 100644 --- a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.h +++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.h @@ -421,7 +421,7 @@ int pvrdma_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *wc); int pvrdma_req_notify_cq(struct ib_cq *cq, enum ib_cq_notify_flags flags); struct ib_ah *pvrdma_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, u32 flags, struct ib_udata *udata); -int pvrdma_destroy_ah(struct ib_ah *ah); +int pvrdma_destroy_ah(struct ib_ah *ah, u32 flags); struct ib_srq *pvrdma_create_srq(struct ib_pd *pd, struct ib_srq_init_attr *init_attr, diff --git a/drivers/infiniband/sw/rdmavt/ah.c b/drivers/infiniband/sw/rdmavt/ah.c index 43e46877d300..fc10e4e26ca7 100644 --- a/drivers/infiniband/sw/rdmavt/ah.c +++ b/drivers/infiniband/sw/rdmavt/ah.c @@ -137,10 +137,11 @@ struct ib_ah *rvt_create_ah(struct ib_pd *pd, /** * rvt_destory_ah - Destory an address handle * @ibah: address handle + * @destroy_flags: destroy address handle flags (see enum rdma_destroy_ah_flags) * * Return: 0 on success */ -int rvt_destroy_ah(struct ib_ah *ibah) +int rvt_destroy_ah(struct ib_ah *ibah, u32 destroy_flags) { struct rvt_dev_info *dev = ib_to_rvt(ibah->device); struct rvt_ah *ah = ibah_to_rvtah(ibah); diff --git a/drivers/infiniband/sw/rdmavt/ah.h b/drivers/infiniband/sw/rdmavt/ah.h index 70f8d089f740..72431a618d5d 100644 --- a/drivers/infiniband/sw/rdmavt/ah.h +++ b/drivers/infiniband/sw/rdmavt/ah.h @@ -54,7 +54,7 @@ struct ib_ah *rvt_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, u32 create_flags, struct ib_udata *udata); -int rvt_destroy_ah(struct ib_ah *ibah); +int rvt_destroy_ah(struct ib_ah *ibah, u32 destroy_flags); int rvt_modify_ah(struct ib_ah *ibah, struct rdma_ah_attr *ah_attr); int rvt_query_ah(struct ib_ah *ibah, struct rdma_ah_attr *ah_attr); diff --git a/drivers/infiniband/sw/rdmavt/mad.c b/drivers/infiniband/sw/rdmavt/mad.c index d6981dc04adb..108c71e3ac23 100644 --- a/drivers/infiniband/sw/rdmavt/mad.c +++ b/drivers/infiniband/sw/rdmavt/mad.c @@ -160,7 +160,8 @@ void rvt_free_mad_agents(struct rvt_dev_info *rdi) ib_unregister_mad_agent(agent); } if (rvp->sm_ah) { - rdma_destroy_ah(&rvp->sm_ah->ibah); + rdma_destroy_ah(&rvp->sm_ah->ibah, + RDMA_DESTROY_AH_SLEEPABLE); rvp->sm_ah = NULL; } diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.c b/drivers/infiniband/sw/rxe/rxe_verbs.c index 9c0ae9a26319..b20e6e0415f5 100644 --- a/drivers/infiniband/sw/rxe/rxe_verbs.c +++ b/drivers/infiniband/sw/rxe/rxe_verbs.c @@ -267,7 +267,7 @@ static int rxe_query_ah(struct ib_ah *ibah, struct rdma_ah_attr *attr) return 0; } -static int rxe_destroy_ah(struct ib_ah *ibah) +static int rxe_destroy_ah(struct ib_ah *ibah, u32 flags) { struct rxe_ah *ah = to_rah(ibah); diff --git a/drivers/infiniband/ulp/ipoib/ipoib_ib.c b/drivers/infiniband/ulp/ipoib/ipoib_ib.c index 1ecf21006f72..6d35570092d6 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_ib.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_ib.c @@ -678,7 +678,7 @@ static void __ipoib_reap_ah(struct net_device *dev) list_for_each_entry_safe(ah, tah, &priv->dead_ahs, list) if ((int) priv->tx_tail - (int) ah->last_send >= 0) { list_del(&ah->list); - rdma_destroy_ah(ah->ah); + rdma_destroy_ah(ah->ah, 0); kfree(ah); } diff --git a/drivers/infiniband/ulp/opa_vnic/opa_vnic_vema.c b/drivers/infiniband/ulp/opa_vnic/opa_vnic_vema.c index 2f4f63847c10..560e4f2d466e 100644 --- a/drivers/infiniband/ulp/opa_vnic/opa_vnic_vema.c +++ b/drivers/infiniband/ulp/opa_vnic/opa_vnic_vema.c @@ -606,7 +606,7 @@ static void vema_set(struct opa_vnic_vema_port *port, static void vema_send(struct ib_mad_agent *mad_agent, struct ib_mad_send_wc *mad_wc) { - rdma_destroy_ah(mad_wc->send_buf->ah); + rdma_destroy_ah(mad_wc->send_buf->ah, RDMA_DESTROY_AH_SLEEPABLE); ib_free_send_mad(mad_wc->send_buf); } @@ -680,7 +680,7 @@ static void vema_recv(struct ib_mad_agent *mad_agent, ib_free_send_mad(rsp); err_rsp: - rdma_destroy_ah(ah); + rdma_destroy_ah(ah, RDMA_DESTROY_AH_SLEEPABLE); free_recv_mad: ib_free_recv_mad(mad_wc); } @@ -848,7 +848,7 @@ void opa_vnic_vema_send_trap(struct opa_vnic_adapter *adapter, } err_sndbuf: - rdma_destroy_ah(ah); + rdma_destroy_ah(ah, 0); err_exit: v_err("Aborting trap\n"); } diff --git a/drivers/infiniband/ulp/srpt/ib_srpt.c b/drivers/infiniband/ulp/srpt/ib_srpt.c index bddd0ad9dabf..b2f63a6dadb6 100644 --- a/drivers/infiniband/ulp/srpt/ib_srpt.c +++ b/drivers/infiniband/ulp/srpt/ib_srpt.c @@ -458,7 +458,7 @@ static void srpt_mgmt_method_get(struct srpt_port *sp, struct ib_mad *rq_mad, static void srpt_mad_send_handler(struct ib_mad_agent *mad_agent, struct ib_mad_send_wc *mad_wc) { - rdma_destroy_ah(mad_wc->send_buf->ah); + rdma_destroy_ah(mad_wc->send_buf->ah, RDMA_DESTROY_AH_SLEEPABLE); ib_free_send_mad(mad_wc->send_buf); } @@ -525,7 +525,7 @@ static void srpt_mad_recv_handler(struct ib_mad_agent *mad_agent, ib_free_send_mad(rsp); err_rsp: - rdma_destroy_ah(ah); + rdma_destroy_ah(ah, RDMA_DESTROY_AH_SLEEPABLE); err: ib_free_recv_mad(mad_wc); } diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h index 86ae878e1de4..a3ceed3a040a 100644 --- a/include/rdma/ib_verbs.h +++ b/include/rdma/ib_verbs.h @@ -2381,7 +2381,7 @@ struct ib_device_ops { struct ib_udata *udata); int (*modify_ah)(struct ib_ah *ah, struct rdma_ah_attr *ah_attr); int (*query_ah)(struct ib_ah *ah, struct rdma_ah_attr *ah_attr); - int (*destroy_ah)(struct ib_ah *ah); + int (*destroy_ah)(struct ib_ah *ah, u32 flags); struct ib_srq *(*create_srq)(struct ib_pd *pd, struct ib_srq_init_attr *srq_init_attr, struct ib_udata *udata); @@ -3256,11 +3256,17 @@ int rdma_modify_ah(struct ib_ah *ah, struct rdma_ah_attr *ah_attr); */ int rdma_query_ah(struct ib_ah *ah, struct rdma_ah_attr *ah_attr); +enum rdma_destroy_ah_flags { + /* In a sleepable context */ + RDMA_DESTROY_AH_SLEEPABLE = BIT(0), +}; + /** * rdma_destroy_ah - Destroys an address handle. * @ah: The address handle to destroy. + * @flags: Destroy address handle flags (see enum rdma_destroy_ah_flags). */ -int rdma_destroy_ah(struct ib_ah *ah); +int rdma_destroy_ah(struct ib_ah *ah, u32 flags); /** * ib_create_srq - Creates a SRQ associated with the specified protection -- cgit v1.2.3-59-g8ed1b From 90e3edd8cca453c6042bb99eff3515cb70cfb2ba Mon Sep 17 00:00:00 2001 From: Gal Pressman Date: Wed, 12 Dec 2018 11:09:07 +0200 Subject: RDMA/bnxt_re: Make use of create AH sleepable flag When in a sleepable (non-atomic) context, wait for firmware completion instead of polling for it. Signed-off-by: Gal Pressman Acked-by: Selvin Xavier Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/bnxt_re/ib_verbs.c | 5 +++-- drivers/infiniband/hw/bnxt_re/qplib_sp.c | 5 +++-- drivers/infiniband/hw/bnxt_re/qplib_sp.h | 3 ++- 3 files changed, 8 insertions(+), 5 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.c b/drivers/infiniband/hw/bnxt_re/ib_verbs.c index ece45e7aee20..7f934c2d90eb 100644 --- a/drivers/infiniband/hw/bnxt_re/ib_verbs.c +++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.c @@ -723,7 +723,8 @@ struct ib_ah *bnxt_re_create_ah(struct ib_pd *ib_pd, } memcpy(ah->qplib_ah.dmac, ah_attr->roce.dmac, ETH_ALEN); - rc = bnxt_qplib_create_ah(&rdev->qplib_res, &ah->qplib_ah); + rc = bnxt_qplib_create_ah(&rdev->qplib_res, &ah->qplib_ah, + !(flags & RDMA_CREATE_AH_SLEEPABLE)); if (rc) { dev_err(rdev_to_dev(rdev), "Failed to allocate HW AH"); goto fail; @@ -959,7 +960,7 @@ static struct bnxt_re_ah *bnxt_re_create_shadow_qp_ah /* Have DMAC same as SMAC */ ether_addr_copy(ah->qplib_ah.dmac, rdev->netdev->dev_addr); - rc = bnxt_qplib_create_ah(&rdev->qplib_res, &ah->qplib_ah); + rc = bnxt_qplib_create_ah(&rdev->qplib_res, &ah->qplib_ah, false); if (rc) { dev_err(rdev_to_dev(rdev), "Failed to allocate HW AH for Shadow QP"); diff --git a/drivers/infiniband/hw/bnxt_re/qplib_sp.c b/drivers/infiniband/hw/bnxt_re/qplib_sp.c index 5216b5f844cc..8f4b19ea409c 100644 --- a/drivers/infiniband/hw/bnxt_re/qplib_sp.c +++ b/drivers/infiniband/hw/bnxt_re/qplib_sp.c @@ -488,7 +488,8 @@ int bnxt_qplib_add_pkey(struct bnxt_qplib_res *res, } /* AH */ -int bnxt_qplib_create_ah(struct bnxt_qplib_res *res, struct bnxt_qplib_ah *ah) +int bnxt_qplib_create_ah(struct bnxt_qplib_res *res, struct bnxt_qplib_ah *ah, + bool block) { struct bnxt_qplib_rcfw *rcfw = res->rcfw; struct cmdq_create_ah req; @@ -522,7 +523,7 @@ int bnxt_qplib_create_ah(struct bnxt_qplib_res *res, struct bnxt_qplib_ah *ah) req.dest_mac[2] = cpu_to_le16(temp16[2]); rc = bnxt_qplib_rcfw_send_message(rcfw, (void *)&req, (void *)&resp, - NULL, 1); + NULL, block); if (rc) return rc; diff --git a/drivers/infiniband/hw/bnxt_re/qplib_sp.h b/drivers/infiniband/hw/bnxt_re/qplib_sp.h index 8079d7f5a008..53145aa1e7ff 100644 --- a/drivers/infiniband/hw/bnxt_re/qplib_sp.h +++ b/drivers/infiniband/hw/bnxt_re/qplib_sp.h @@ -241,7 +241,8 @@ int bnxt_qplib_get_dev_attr(struct bnxt_qplib_rcfw *rcfw, int bnxt_qplib_set_func_resources(struct bnxt_qplib_res *res, struct bnxt_qplib_rcfw *rcfw, struct bnxt_qplib_ctx *ctx); -int bnxt_qplib_create_ah(struct bnxt_qplib_res *res, struct bnxt_qplib_ah *ah); +int bnxt_qplib_create_ah(struct bnxt_qplib_res *res, struct bnxt_qplib_ah *ah, + bool block); int bnxt_qplib_destroy_ah(struct bnxt_qplib_res *res, struct bnxt_qplib_ah *ah); int bnxt_qplib_alloc_mrw(struct bnxt_qplib_res *res, struct bnxt_qplib_mrw *mrw); -- cgit v1.2.3-59-g8ed1b From 50c582de1d39231af31c881a8444333bff6f11a8 Mon Sep 17 00:00:00 2001 From: Gal Pressman Date: Wed, 12 Dec 2018 11:09:08 +0200 Subject: RDMA/bnxt_re: Make use of destroy AH sleepable flag When in a sleepable (non-atomic) context, wait for firmware completion instead of polling for it. Signed-off-by: Gal Pressman Acked-by: Selvin Xavier Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/bnxt_re/ib_verbs.c | 5 +++-- drivers/infiniband/hw/bnxt_re/qplib_sp.c | 5 +++-- drivers/infiniband/hw/bnxt_re/qplib_sp.h | 3 ++- 3 files changed, 8 insertions(+), 5 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.c b/drivers/infiniband/hw/bnxt_re/ib_verbs.c index 7f934c2d90eb..1e2515e2eb62 100644 --- a/drivers/infiniband/hw/bnxt_re/ib_verbs.c +++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.c @@ -653,7 +653,8 @@ int bnxt_re_destroy_ah(struct ib_ah *ib_ah, u32 flags) struct bnxt_re_dev *rdev = ah->rdev; int rc; - rc = bnxt_qplib_destroy_ah(&rdev->qplib_res, &ah->qplib_ah); + rc = bnxt_qplib_destroy_ah(&rdev->qplib_res, &ah->qplib_ah, + !(flags & RDMA_DESTROY_AH_SLEEPABLE)); if (rc) { dev_err(rdev_to_dev(rdev), "Failed to destroy HW AH"); return rc; @@ -820,7 +821,7 @@ int bnxt_re_destroy_qp(struct ib_qp *ib_qp) if (ib_qp->qp_type == IB_QPT_GSI && rdev->qp1_sqp) { rc = bnxt_qplib_destroy_ah(&rdev->qplib_res, - &rdev->sqp_ah->qplib_ah); + &rdev->sqp_ah->qplib_ah, false); if (rc) { dev_err(rdev_to_dev(rdev), "Failed to destroy HW AH for shadow QP"); diff --git a/drivers/infiniband/hw/bnxt_re/qplib_sp.c b/drivers/infiniband/hw/bnxt_re/qplib_sp.c index 8f4b19ea409c..be03b5738f71 100644 --- a/drivers/infiniband/hw/bnxt_re/qplib_sp.c +++ b/drivers/infiniband/hw/bnxt_re/qplib_sp.c @@ -531,7 +531,8 @@ int bnxt_qplib_create_ah(struct bnxt_qplib_res *res, struct bnxt_qplib_ah *ah, return 0; } -int bnxt_qplib_destroy_ah(struct bnxt_qplib_res *res, struct bnxt_qplib_ah *ah) +int bnxt_qplib_destroy_ah(struct bnxt_qplib_res *res, struct bnxt_qplib_ah *ah, + bool block) { struct bnxt_qplib_rcfw *rcfw = res->rcfw; struct cmdq_destroy_ah req; @@ -545,7 +546,7 @@ int bnxt_qplib_destroy_ah(struct bnxt_qplib_res *res, struct bnxt_qplib_ah *ah) req.ah_cid = cpu_to_le32(ah->id); rc = bnxt_qplib_rcfw_send_message(rcfw, (void *)&req, (void *)&resp, - NULL, 1); + NULL, block); if (rc) return rc; return 0; diff --git a/drivers/infiniband/hw/bnxt_re/qplib_sp.h b/drivers/infiniband/hw/bnxt_re/qplib_sp.h index 53145aa1e7ff..39454b3f738d 100644 --- a/drivers/infiniband/hw/bnxt_re/qplib_sp.h +++ b/drivers/infiniband/hw/bnxt_re/qplib_sp.h @@ -243,7 +243,8 @@ int bnxt_qplib_set_func_resources(struct bnxt_qplib_res *res, struct bnxt_qplib_ctx *ctx); int bnxt_qplib_create_ah(struct bnxt_qplib_res *res, struct bnxt_qplib_ah *ah, bool block); -int bnxt_qplib_destroy_ah(struct bnxt_qplib_res *res, struct bnxt_qplib_ah *ah); +int bnxt_qplib_destroy_ah(struct bnxt_qplib_res *res, struct bnxt_qplib_ah *ah, + bool block); int bnxt_qplib_alloc_mrw(struct bnxt_qplib_res *res, struct bnxt_qplib_mrw *mrw); int bnxt_qplib_dereg_mrw(struct bnxt_qplib_res *res, struct bnxt_qplib_mrw *mrw, -- cgit v1.2.3-59-g8ed1b From 8036e90f92aae2784b855a0007ae2d8154d28b3c Mon Sep 17 00:00:00 2001 From: Parvi Kaustubhi Date: Tue, 11 Dec 2018 14:15:42 -0800 Subject: IB/usnic: Fix potential deadlock Acquiring the rtnl lock while holding usdev_lock could result in a deadlock. For example: usnic_ib_query_port() | mutex_lock(&us_ibdev->usdev_lock) | ib_get_eth_speed() | rtnl_lock() rtnl_lock() | usnic_ib_netdevice_event() | mutex_lock(&us_ibdev->usdev_lock) This commit moves the usdev_lock acquisition after the rtnl lock has been released. This is safe to do because usdev_lock is not protecting anything being accessed in ib_get_eth_speed(). Hence, the correct order of holding locks (rtnl -> usdev_lock) is not violated. Signed-off-by: Parvi Kaustubhi Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/usnic/usnic_ib_verbs.c | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/usnic/usnic_ib_verbs.c b/drivers/infiniband/hw/usnic/usnic_ib_verbs.c index 1b9f69d65111..1d4abef17e38 100644 --- a/drivers/infiniband/hw/usnic/usnic_ib_verbs.c +++ b/drivers/infiniband/hw/usnic/usnic_ib_verbs.c @@ -336,13 +336,16 @@ int usnic_ib_query_port(struct ib_device *ibdev, u8 port, usnic_dbg("\n"); - mutex_lock(&us_ibdev->usdev_lock); if (ib_get_eth_speed(ibdev, port, &props->active_speed, - &props->active_width)) { - mutex_unlock(&us_ibdev->usdev_lock); + &props->active_width)) return -EINVAL; - } + /* + * usdev_lock is acquired after (and not before) ib_get_eth_speed call + * because acquiring rtnl_lock in ib_get_eth_speed, while holding + * usdev_lock could lead to a deadlock. + */ + mutex_lock(&us_ibdev->usdev_lock); /* props being zeroed by the caller, avoid zeroing it here */ props->lid = 0; -- cgit v1.2.3-59-g8ed1b From 2b827ea1926b5ad7ac3e9ba8651ff99181d6c7f9 Mon Sep 17 00:00:00 2001 From: Selvin Xavier Date: Wed, 12 Dec 2018 01:56:23 -0800 Subject: RDMA/bnxt_re: Query HWRM Interface version from FW Get HWRM interface major, minor, build and patch version from FW for checking the FW/Driver compatibility. Signed-off-by: Selvin Xavier Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/bnxt_re/main.c | 31 +++++++++++++++++++++++++++++++ drivers/infiniband/hw/bnxt_re/qplib_res.h | 1 + 2 files changed, 32 insertions(+) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/bnxt_re/main.c b/drivers/infiniband/hw/bnxt_re/main.c index 31cbef57a19b..1d18b77940f2 100644 --- a/drivers/infiniband/hw/bnxt_re/main.c +++ b/drivers/infiniband/hw/bnxt_re/main.c @@ -1197,6 +1197,35 @@ static int bnxt_re_setup_qos(struct bnxt_re_dev *rdev) return 0; } +static void bnxt_re_query_hwrm_intf_version(struct bnxt_re_dev *rdev) +{ + struct bnxt_en_dev *en_dev = rdev->en_dev; + struct hwrm_ver_get_output resp = {0}; + struct hwrm_ver_get_input req = {0}; + struct bnxt_fw_msg fw_msg; + int rc = 0; + + memset(&fw_msg, 0, sizeof(fw_msg)); + bnxt_re_init_hwrm_hdr(rdev, (void *)&req, + HWRM_VER_GET, -1, -1); + req.hwrm_intf_maj = HWRM_VERSION_MAJOR; + req.hwrm_intf_min = HWRM_VERSION_MINOR; + req.hwrm_intf_upd = HWRM_VERSION_UPDATE; + bnxt_re_fill_fw_msg(&fw_msg, (void *)&req, sizeof(req), (void *)&resp, + sizeof(resp), DFLT_HWRM_CMD_TIMEOUT); + rc = en_dev->en_ops->bnxt_send_fw_msg(en_dev, BNXT_ROCE_ULP, &fw_msg); + if (rc) { + dev_err(rdev_to_dev(rdev), + "Failed to query HW version, rc = 0x%x", rc); + return; + } + rdev->qplib_ctx.hwrm_intf_ver = + (u64)resp.hwrm_intf_major << 48 | + (u64)resp.hwrm_intf_minor << 32 | + (u64)resp.hwrm_intf_build << 16 | + resp.hwrm_intf_patch; +} + static void bnxt_re_ib_unreg(struct bnxt_re_dev *rdev) { int rc; @@ -1279,6 +1308,8 @@ static int bnxt_re_ib_reg(struct bnxt_re_dev *rdev) } set_bit(BNXT_RE_FLAG_GOT_MSIX, &rdev->flags); + bnxt_re_query_hwrm_intf_version(rdev); + /* Establish RCFW Communication Channel to initialize the context * memory for the function and all child VFs */ diff --git a/drivers/infiniband/hw/bnxt_re/qplib_res.h b/drivers/infiniband/hw/bnxt_re/qplib_res.h index 2e5c052da5a9..1e80aa7bbcce 100644 --- a/drivers/infiniband/hw/bnxt_re/qplib_res.h +++ b/drivers/infiniband/hw/bnxt_re/qplib_res.h @@ -177,6 +177,7 @@ struct bnxt_qplib_ctx { struct bnxt_qplib_hwq tqm_tbl[MAX_TQM_ALLOC_REQ]; struct bnxt_qplib_stats stats; struct bnxt_qplib_vf_res vf_res; + u64 hwrm_intf_ver; }; struct bnxt_qplib_res { -- cgit v1.2.3-59-g8ed1b From bd1c24ccf9eb070510c5da2fe0b56899c9a52c96 Mon Sep 17 00:00:00 2001 From: Devesh Sharma Date: Wed, 12 Dec 2018 01:56:24 -0800 Subject: RDMA/bnxt_re: Increase depth of control path command queue Increasing the depth of control path command queue to 8K entries to handle burst of commands. This feature needs support from FW and the driver/fw compatibility is checked from the interface version number. Signed-off-by: Devesh Sharma Signed-off-by: Selvin Xavier Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/bnxt_re/main.c | 1 + drivers/infiniband/hw/bnxt_re/qplib_rcfw.c | 34 +++++++++++------ drivers/infiniband/hw/bnxt_re/qplib_rcfw.h | 59 +++++++++++++++++++++++------- 3 files changed, 68 insertions(+), 26 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/bnxt_re/main.c b/drivers/infiniband/hw/bnxt_re/main.c index 1d18b77940f2..e7a997f2a537 100644 --- a/drivers/infiniband/hw/bnxt_re/main.c +++ b/drivers/infiniband/hw/bnxt_re/main.c @@ -1314,6 +1314,7 @@ static int bnxt_re_ib_reg(struct bnxt_re_dev *rdev) * memory for the function and all child VFs */ rc = bnxt_qplib_alloc_rcfw_channel(rdev->en_dev->pdev, &rdev->rcfw, + &rdev->qplib_ctx, BNXT_RE_MAX_QPC_COUNT); if (rc) { pr_err("Failed to allocate RCFW Channel: %#x\n", rc); diff --git a/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c b/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c index be4e33e9f962..326805461265 100644 --- a/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c +++ b/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c @@ -58,7 +58,7 @@ static int __wait_for_resp(struct bnxt_qplib_rcfw *rcfw, u16 cookie) u16 cbit; int rc; - cbit = cookie % RCFW_MAX_OUTSTANDING_CMD; + cbit = cookie % rcfw->cmdq_depth; rc = wait_event_timeout(rcfw->waitq, !test_bit(cbit, rcfw->cmdq_bitmap), msecs_to_jiffies(RCFW_CMD_WAIT_TIME_MS)); @@ -70,7 +70,7 @@ static int __block_for_resp(struct bnxt_qplib_rcfw *rcfw, u16 cookie) u32 count = RCFW_BLOCKED_CMD_WAIT_COUNT; u16 cbit; - cbit = cookie % RCFW_MAX_OUTSTANDING_CMD; + cbit = cookie % rcfw->cmdq_depth; if (!test_bit(cbit, rcfw->cmdq_bitmap)) goto done; do { @@ -86,6 +86,7 @@ static int __send_message(struct bnxt_qplib_rcfw *rcfw, struct cmdq_base *req, { struct bnxt_qplib_cmdqe *cmdqe, **cmdq_ptr; struct bnxt_qplib_hwq *cmdq = &rcfw->cmdq; + u32 cmdq_depth = rcfw->cmdq_depth; struct bnxt_qplib_crsq *crsqe; u32 sw_prod, cmdq_prod; unsigned long flags; @@ -124,7 +125,7 @@ static int __send_message(struct bnxt_qplib_rcfw *rcfw, struct cmdq_base *req, cookie = rcfw->seq_num & RCFW_MAX_COOKIE_VALUE; - cbit = cookie % RCFW_MAX_OUTSTANDING_CMD; + cbit = cookie % rcfw->cmdq_depth; if (is_block) cookie |= RCFW_CMD_IS_BLOCKING; @@ -153,7 +154,8 @@ static int __send_message(struct bnxt_qplib_rcfw *rcfw, struct cmdq_base *req, do { /* Locate the next cmdq slot */ sw_prod = HWQ_CMP(cmdq->prod, cmdq); - cmdqe = &cmdq_ptr[get_cmdq_pg(sw_prod)][get_cmdq_idx(sw_prod)]; + cmdqe = &cmdq_ptr[get_cmdq_pg(sw_prod, cmdq_depth)] + [get_cmdq_idx(sw_prod, cmdq_depth)]; if (!cmdqe) { dev_err(&rcfw->pdev->dev, "RCFW request failed with no cmdqe!\n"); @@ -326,7 +328,7 @@ static int bnxt_qplib_process_qp_event(struct bnxt_qplib_rcfw *rcfw, mcookie = qp_event->cookie; blocked = cookie & RCFW_CMD_IS_BLOCKING; cookie &= RCFW_MAX_COOKIE_VALUE; - cbit = cookie % RCFW_MAX_OUTSTANDING_CMD; + cbit = cookie % rcfw->cmdq_depth; crsqe = &rcfw->crsqe_tbl[cbit]; if (crsqe->resp && crsqe->resp->cookie == mcookie) { @@ -555,6 +557,7 @@ void bnxt_qplib_free_rcfw_channel(struct bnxt_qplib_rcfw *rcfw) int bnxt_qplib_alloc_rcfw_channel(struct pci_dev *pdev, struct bnxt_qplib_rcfw *rcfw, + struct bnxt_qplib_ctx *ctx, int qp_tbl_sz) { rcfw->pdev = pdev; @@ -567,11 +570,18 @@ int bnxt_qplib_alloc_rcfw_channel(struct pci_dev *pdev, "HW channel CREQ allocation failed\n"); goto fail; } - rcfw->cmdq.max_elements = BNXT_QPLIB_CMDQE_MAX_CNT; - if (bnxt_qplib_alloc_init_hwq(rcfw->pdev, &rcfw->cmdq, NULL, 0, - &rcfw->cmdq.max_elements, - BNXT_QPLIB_CMDQE_UNITS, 0, PAGE_SIZE, - HWQ_TYPE_CTX)) { + if (ctx->hwrm_intf_ver < HWRM_VERSION_RCFW_CMDQ_DEPTH_CHECK) + rcfw->cmdq_depth = BNXT_QPLIB_CMDQE_MAX_CNT_256; + else + rcfw->cmdq_depth = BNXT_QPLIB_CMDQE_MAX_CNT_8192; + + rcfw->cmdq.max_elements = rcfw->cmdq_depth; + if (bnxt_qplib_alloc_init_hwq + (rcfw->pdev, &rcfw->cmdq, NULL, 0, + &rcfw->cmdq.max_elements, + BNXT_QPLIB_CMDQE_UNITS, 0, + bnxt_qplib_cmdqe_page_size(rcfw->cmdq_depth), + HWQ_TYPE_CTX)) { dev_err(&rcfw->pdev->dev, "HW channel CMDQ allocation failed\n"); goto fail; @@ -674,7 +684,7 @@ int bnxt_qplib_enable_rcfw_channel(struct pci_dev *pdev, /* General */ rcfw->seq_num = 0; set_bit(FIRMWARE_FIRST_FLAG, &rcfw->flags); - bmap_size = BITS_TO_LONGS(RCFW_MAX_OUTSTANDING_CMD * + bmap_size = BITS_TO_LONGS(rcfw->cmdq_depth * sizeof(unsigned long)); rcfw->cmdq_bitmap = kzalloc(bmap_size, GFP_KERNEL); if (!rcfw->cmdq_bitmap) @@ -734,7 +744,7 @@ int bnxt_qplib_enable_rcfw_channel(struct pci_dev *pdev, init.cmdq_pbl = cpu_to_le64(rcfw->cmdq.pbl[PBL_LVL_0].pg_map_arr[0]); init.cmdq_size_cmdq_lvl = cpu_to_le16( - ((BNXT_QPLIB_CMDQE_MAX_CNT << CMDQ_INIT_CMDQ_SIZE_SFT) & + ((rcfw->cmdq_depth << CMDQ_INIT_CMDQ_SIZE_SFT) & CMDQ_INIT_CMDQ_SIZE_MASK) | ((rcfw->cmdq.level << CMDQ_INIT_CMDQ_LVL_SFT) & CMDQ_INIT_CMDQ_LVL_MASK)); diff --git a/drivers/infiniband/hw/bnxt_re/qplib_rcfw.h b/drivers/infiniband/hw/bnxt_re/qplib_rcfw.h index 9a8687dc0a79..be0ef0e8c53e 100644 --- a/drivers/infiniband/hw/bnxt_re/qplib_rcfw.h +++ b/drivers/infiniband/hw/bnxt_re/qplib_rcfw.h @@ -63,32 +63,60 @@ #define RCFW_CMD_WAIT_TIME_MS 20000 /* 20 Seconds timeout */ +/* Cmdq contains a fix number of a 16-Byte slots */ +struct bnxt_qplib_cmdqe { + u8 data[16]; +}; + /* CMDQ elements */ -#define BNXT_QPLIB_CMDQE_MAX_CNT 256 +#define BNXT_QPLIB_CMDQE_MAX_CNT_256 256 +#define BNXT_QPLIB_CMDQE_MAX_CNT_8192 8192 #define BNXT_QPLIB_CMDQE_UNITS sizeof(struct bnxt_qplib_cmdqe) -#define BNXT_QPLIB_CMDQE_CNT_PER_PG (PAGE_SIZE / BNXT_QPLIB_CMDQE_UNITS) +#define BNXT_QPLIB_CMDQE_BYTES(depth) ((depth) * BNXT_QPLIB_CMDQE_UNITS) + +static inline u32 bnxt_qplib_cmdqe_npages(u32 depth) +{ + u32 npages; + + npages = BNXT_QPLIB_CMDQE_BYTES(depth) / PAGE_SIZE; + if (BNXT_QPLIB_CMDQE_BYTES(depth) % PAGE_SIZE) + npages++; + return npages; +} + +static inline u32 bnxt_qplib_cmdqe_page_size(u32 depth) +{ + return (bnxt_qplib_cmdqe_npages(depth) * PAGE_SIZE); +} + +static inline u32 bnxt_qplib_cmdqe_cnt_per_pg(u32 depth) +{ + return (bnxt_qplib_cmdqe_page_size(depth) / + BNXT_QPLIB_CMDQE_UNITS); +} -#define MAX_CMDQ_IDX (BNXT_QPLIB_CMDQE_MAX_CNT - 1) -#define MAX_CMDQ_IDX_PER_PG (BNXT_QPLIB_CMDQE_CNT_PER_PG - 1) +#define MAX_CMDQ_IDX(depth) ((depth) - 1) + +static inline u32 bnxt_qplib_max_cmdq_idx_per_pg(u32 depth) +{ + return (bnxt_qplib_cmdqe_cnt_per_pg(depth) - 1); +} -#define RCFW_MAX_OUTSTANDING_CMD BNXT_QPLIB_CMDQE_MAX_CNT #define RCFW_MAX_COOKIE_VALUE 0x7FFF #define RCFW_CMD_IS_BLOCKING 0x8000 #define RCFW_BLOCKED_CMD_WAIT_COUNT 0x4E20 -/* Cmdq contains a fix number of a 16-Byte slots */ -struct bnxt_qplib_cmdqe { - u8 data[16]; -}; +#define HWRM_VERSION_RCFW_CMDQ_DEPTH_CHECK 0x1000900020011ULL -static inline u32 get_cmdq_pg(u32 val) +static inline u32 get_cmdq_pg(u32 val, u32 depth) { - return (val & ~MAX_CMDQ_IDX_PER_PG) / BNXT_QPLIB_CMDQE_CNT_PER_PG; + return (val & ~(bnxt_qplib_max_cmdq_idx_per_pg(depth))) / + (bnxt_qplib_cmdqe_cnt_per_pg(depth)); } -static inline u32 get_cmdq_idx(u32 val) +static inline u32 get_cmdq_idx(u32 val, u32 depth) { - return val & MAX_CMDQ_IDX_PER_PG; + return val & (bnxt_qplib_max_cmdq_idx_per_pg(depth)); } /* Crsq buf is 1024-Byte */ @@ -194,11 +222,14 @@ struct bnxt_qplib_rcfw { struct bnxt_qplib_qp_node *qp_tbl; u64 oos_prev; u32 init_oos_stats; + u32 cmdq_depth; }; void bnxt_qplib_free_rcfw_channel(struct bnxt_qplib_rcfw *rcfw); int bnxt_qplib_alloc_rcfw_channel(struct pci_dev *pdev, - struct bnxt_qplib_rcfw *rcfw, int qp_tbl_sz); + struct bnxt_qplib_rcfw *rcfw, + struct bnxt_qplib_ctx *ctx, + int qp_tbl_sz); void bnxt_qplib_rcfw_stop_irq(struct bnxt_qplib_rcfw *rcfw, bool kill); void bnxt_qplib_disable_rcfw_channel(struct bnxt_qplib_rcfw *rcfw); int bnxt_qplib_rcfw_start_irq(struct bnxt_qplib_rcfw *rcfw, int msix_vector, -- cgit v1.2.3-59-g8ed1b From 425518cc5ede0d7a71886816368e4bef60e418db Mon Sep 17 00:00:00 2001 From: Yishai Hadas Date: Wed, 19 Dec 2018 16:28:11 +0200 Subject: IB/mlx5: Use uid as part of alloc/dealloc transport domain Use uid as part of alloc/dealloc transport domain to let firmware manages the resources correctly. Fixes: d2d19121ae2f ("IB/mlx5: Set uid as part of TD commands") Signed-off-by: Yishai Hadas Reviewed-by: Artemy Kovalyov Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/cmd.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/cmd.c b/drivers/infiniband/hw/mlx5/cmd.c index ca060a2e2b36..59d569c6a13a 100644 --- a/drivers/infiniband/hw/mlx5/cmd.c +++ b/drivers/infiniband/hw/mlx5/cmd.c @@ -240,6 +240,7 @@ int mlx5_cmd_alloc_transport_domain(struct mlx5_core_dev *dev, u32 *tdn, MLX5_SET(alloc_transport_domain_in, in, opcode, MLX5_CMD_OP_ALLOC_TRANSPORT_DOMAIN); + MLX5_SET(alloc_transport_domain_in, in, uid, uid); err = mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out)); if (!err) @@ -257,6 +258,7 @@ void mlx5_cmd_dealloc_transport_domain(struct mlx5_core_dev *dev, u32 tdn, MLX5_SET(dealloc_transport_domain_in, in, opcode, MLX5_CMD_OP_DEALLOC_TRANSPORT_DOMAIN); + MLX5_SET(dealloc_transport_domain_in, in, uid, uid); MLX5_SET(dealloc_transport_domain_in, in, transport_domain, tdn); mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out)); } -- cgit v1.2.3-59-g8ed1b From 6e3722baac048fdf95b867c5ee7e270718e8630d Mon Sep 17 00:00:00 2001 From: Yishai Hadas Date: Wed, 19 Dec 2018 16:28:15 +0200 Subject: IB/mlx5: Use the correct commands for UMEM and UCTX allocation During testing the command format was changed to close a security hole. Revise the driver to use the command format that will actually be supported in GA firmware. Both the UMEM and UCTX are intended only for use by the kernel and cannot be executed using a general command. Since the UMEM and CTX are not part of the general object the caps bits were moved to be some log_xxx location in the general HCA caps. The firmware code was adapted as well to match the above. Fixes: a8b92ca1b0e5 ("IB/mlx5: Introduce DEVX") Signed-off-by: Yishai Hadas Reviewed-by: Achiad Shochat Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/devx.c | 34 ++++++++++----------- include/linux/mlx5/mlx5_ifc.h | 62 ++++++++++++++++++++++++++------------- 2 files changed, 57 insertions(+), 39 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/devx.c b/drivers/infiniband/hw/mlx5/devx.c index 5271469aad10..dcc7c974173f 100644 --- a/drivers/infiniband/hw/mlx5/devx.c +++ b/drivers/infiniband/hw/mlx5/devx.c @@ -51,26 +51,21 @@ int mlx5_ib_devx_create(struct mlx5_ib_dev *dev, bool is_user) { u32 in[MLX5_ST_SZ_DW(create_uctx_in)] = {0}; u32 out[MLX5_ST_SZ_DW(general_obj_out_cmd_hdr)] = {0}; - u64 general_obj_types; - void *hdr, *uctx; + void *uctx; int err; u16 uid; u32 cap = 0; - hdr = MLX5_ADDR_OF(create_uctx_in, in, hdr); - uctx = MLX5_ADDR_OF(create_uctx_in, in, uctx); - - general_obj_types = MLX5_CAP_GEN_64(dev->mdev, general_obj_types); - if (!(general_obj_types & MLX5_GENERAL_OBJ_TYPES_CAP_UCTX) || - !(general_obj_types & MLX5_GENERAL_OBJ_TYPES_CAP_UMEM)) + /* 0 means not supported */ + if (!MLX5_CAP_GEN(dev->mdev, log_max_uctx)) return -EINVAL; + uctx = MLX5_ADDR_OF(create_uctx_in, in, uctx); if (is_user && capable(CAP_NET_RAW) && (MLX5_CAP_GEN(dev->mdev, uctx_cap) & MLX5_UCTX_CAP_RAW_TX)) cap |= MLX5_UCTX_CAP_RAW_TX; - MLX5_SET(general_obj_in_cmd_hdr, hdr, opcode, MLX5_CMD_OP_CREATE_GENERAL_OBJECT); - MLX5_SET(general_obj_in_cmd_hdr, hdr, obj_type, MLX5_OBJ_TYPE_UCTX); + MLX5_SET(create_uctx_in, in, opcode, MLX5_CMD_OP_CREATE_UCTX); MLX5_SET(uctx, uctx, cap, cap); err = mlx5_cmd_exec(dev->mdev, in, sizeof(in), out, sizeof(out)); @@ -83,12 +78,11 @@ int mlx5_ib_devx_create(struct mlx5_ib_dev *dev, bool is_user) void mlx5_ib_devx_destroy(struct mlx5_ib_dev *dev, u16 uid) { - u32 in[MLX5_ST_SZ_DW(general_obj_in_cmd_hdr)] = {0}; + u32 in[MLX5_ST_SZ_DW(destroy_uctx_in)] = {0}; u32 out[MLX5_ST_SZ_DW(general_obj_out_cmd_hdr)] = {0}; - MLX5_SET(general_obj_in_cmd_hdr, in, opcode, MLX5_CMD_OP_DESTROY_GENERAL_OBJECT); - MLX5_SET(general_obj_in_cmd_hdr, in, obj_type, MLX5_OBJ_TYPE_UCTX); - MLX5_SET(general_obj_in_cmd_hdr, in, obj_id, uid); + MLX5_SET(destroy_uctx_in, in, opcode, MLX5_CMD_OP_DESTROY_UCTX); + MLX5_SET(destroy_uctx_in, in, uid, uid); mlx5_cmd_exec(dev->mdev, in, sizeof(in), out, sizeof(out)); } @@ -861,6 +855,10 @@ static void devx_obj_build_destroy_cmd(void *in, void *out, void *din, MLX5_SET(general_obj_in_cmd_hdr, din, obj_type, obj_type); break; + case MLX5_CMD_OP_CREATE_UMEM: + MLX5_SET(general_obj_in_cmd_hdr, din, opcode, + MLX5_CMD_OP_DESTROY_UMEM); + break; case MLX5_CMD_OP_CREATE_MKEY: MLX5_SET(general_obj_in_cmd_hdr, din, opcode, MLX5_CMD_OP_DESTROY_MKEY); break; @@ -1234,8 +1232,7 @@ static void devx_umem_reg_cmd_build(struct mlx5_ib_dev *dev, umem = MLX5_ADDR_OF(create_umem_in, cmd->in, umem); mtt = (__be64 *)MLX5_ADDR_OF(umem, umem, mtt); - MLX5_SET(general_obj_in_cmd_hdr, cmd->in, opcode, MLX5_CMD_OP_CREATE_GENERAL_OBJECT); - MLX5_SET(general_obj_in_cmd_hdr, cmd->in, obj_type, MLX5_OBJ_TYPE_UMEM); + MLX5_SET(create_umem_in, cmd->in, opcode, MLX5_CMD_OP_CREATE_UMEM); MLX5_SET64(umem, umem, num_of_mtt, obj->ncont); MLX5_SET(umem, umem, log_page_size, obj->page_shift - MLX5_ADAPTER_PAGE_SHIFT); @@ -1274,7 +1271,7 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_UMEM_REG)( devx_umem_reg_cmd_build(dev, obj, &cmd); - MLX5_SET(general_obj_in_cmd_hdr, cmd.in, uid, c->devx_uid); + MLX5_SET(create_umem_in, cmd.in, uid, c->devx_uid); err = mlx5_cmd_exec(dev->mdev, cmd.in, cmd.inlen, cmd.out, sizeof(cmd.out)); if (err) @@ -1445,8 +1442,7 @@ static bool devx_is_supported(struct ib_device *device) { struct mlx5_ib_dev *dev = to_mdev(device); - return !dev->rep && MLX5_CAP_GEN_64(dev->mdev, general_obj_types) & - MLX5_GENERAL_OBJ_TYPES_CAP_UCTX; + return !dev->rep && MLX5_CAP_GEN(dev->mdev, log_max_uctx); } const struct uapi_definition mlx5_ib_devx_defs[] = { diff --git a/include/linux/mlx5/mlx5_ifc.h b/include/linux/mlx5/mlx5_ifc.h index 0bca5a6387e9..5ae0b0b9914a 100644 --- a/include/linux/mlx5/mlx5_ifc.h +++ b/include/linux/mlx5/mlx5_ifc.h @@ -75,16 +75,6 @@ enum { MLX5_SET_HCA_CAP_OP_MOD_ATOMIC = 0x3, }; -enum { - MLX5_GENERAL_OBJ_TYPES_CAP_UCTX = (1ULL << 4), - MLX5_GENERAL_OBJ_TYPES_CAP_UMEM = (1ULL << 5), -}; - -enum { - MLX5_OBJ_TYPE_UCTX = 0x0004, - MLX5_OBJ_TYPE_UMEM = 0x0005, -}; - enum { MLX5_SHARED_RESOURCE_UID = 0xffff, }; @@ -267,6 +257,10 @@ enum { MLX5_CMD_OP_MODIFY_GENERAL_OBJECT = 0xa01, MLX5_CMD_OP_QUERY_GENERAL_OBJECT = 0xa02, MLX5_CMD_OP_DESTROY_GENERAL_OBJECT = 0xa03, + MLX5_CMD_OP_CREATE_UCTX = 0xa04, + MLX5_CMD_OP_DESTROY_UCTX = 0xa06, + MLX5_CMD_OP_CREATE_UMEM = 0xa08, + MLX5_CMD_OP_DESTROY_UMEM = 0xa0a, MLX5_CMD_OP_MAX }; @@ -1191,7 +1185,10 @@ struct mlx5_ifc_cmd_hca_cap_bits { u8 reserved_at_440[0x20]; - u8 reserved_at_460[0x10]; + u8 reserved_at_460[0x3]; + u8 log_max_uctx[0x5]; + u8 reserved_at_468[0x3]; + u8 log_max_umem[0x5]; u8 max_num_eqs[0x10]; u8 reserved_at_480[0x3]; @@ -9400,9 +9397,9 @@ struct mlx5_ifc_general_obj_out_cmd_hdr_bits { }; struct mlx5_ifc_umem_bits { - u8 modify_field_select[0x40]; + u8 reserved_at_0[0x80]; - u8 reserved_at_40[0x5b]; + u8 reserved_at_80[0x1b]; u8 log_page_size[0x5]; u8 page_offset[0x20]; @@ -9413,21 +9410,46 @@ struct mlx5_ifc_umem_bits { }; struct mlx5_ifc_uctx_bits { - u8 modify_field_select[0x40]; - u8 cap[0x20]; - u8 reserved_at_60[0x1a0]; + u8 reserved_at_20[0x160]; }; struct mlx5_ifc_create_umem_in_bits { - struct mlx5_ifc_general_obj_in_cmd_hdr_bits hdr; - struct mlx5_ifc_umem_bits umem; + u8 opcode[0x10]; + u8 uid[0x10]; + + u8 reserved_at_20[0x10]; + u8 op_mod[0x10]; + + u8 reserved_at_40[0x40]; + + struct mlx5_ifc_umem_bits umem; }; struct mlx5_ifc_create_uctx_in_bits { - struct mlx5_ifc_general_obj_in_cmd_hdr_bits hdr; - struct mlx5_ifc_uctx_bits uctx; + u8 opcode[0x10]; + u8 reserved_at_10[0x10]; + + u8 reserved_at_20[0x10]; + u8 op_mod[0x10]; + + u8 reserved_at_40[0x40]; + + struct mlx5_ifc_uctx_bits uctx; +}; + +struct mlx5_ifc_destroy_uctx_in_bits { + u8 opcode[0x10]; + u8 reserved_at_10[0x10]; + + u8 reserved_at_20[0x10]; + u8 op_mod[0x10]; + + u8 reserved_at_40[0x10]; + u8 uid[0x10]; + + u8 reserved_at_60[0x20]; }; struct mlx5_ifc_mtrc_string_db_param_bits { -- cgit v1.2.3-59-g8ed1b From 842a9c837e2017127250a93204368f13e660ed9d Mon Sep 17 00:00:00 2001 From: Or Gerlitz Date: Tue, 11 Dec 2018 18:10:43 +0200 Subject: IB/mlx5: Simplify netdev unbinding When dealing with netdev unregister events, we just need to know that this is our currently bounded netdev. There's no need to do any further checks/queries. This patch doesn't change any functionality. Signed-off-by: Or Gerlitz Reviewed-by: Mark Bloch Signed-off-by: Saeed Mahameed Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/main.c | 14 +++++++++----- 1 file changed, 9 insertions(+), 5 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index 2a0526d7ff13..9b40ec73cc64 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -172,7 +172,6 @@ static int mlx5_netdev_event(struct notifier_block *this, switch (event) { case NETDEV_REGISTER: - case NETDEV_UNREGISTER: write_lock(&roce->netdev_lock); if (ibdev->rep) { struct mlx5_eswitch *esw = ibdev->mdev->priv.eswitch; @@ -181,15 +180,20 @@ static int mlx5_netdev_event(struct notifier_block *this, rep_ndev = mlx5_ib_get_rep_netdev(esw, ibdev->rep->vport); if (rep_ndev == ndev) - roce->netdev = (event == NETDEV_UNREGISTER) ? - NULL : ndev; + roce->netdev = ndev; } else if (ndev->dev.parent == &mdev->pdev->dev) { - roce->netdev = (event == NETDEV_UNREGISTER) ? - NULL : ndev; + roce->netdev = ndev; } write_unlock(&roce->netdev_lock); break; + case NETDEV_UNREGISTER: + write_lock(&roce->netdev_lock); + if (roce->netdev == ndev) + roce->netdev = NULL; + write_unlock(&roce->netdev_lock); + break; + case NETDEV_CHANGE: case NETDEV_UP: case NETDEV_DOWN: { -- cgit v1.2.3-59-g8ed1b From bb7e22a8ab00ff9ba911a45ba8784cef9e6d6f7a Mon Sep 17 00:00:00 2001 From: Huy Nguyen Date: Tue, 18 Dec 2018 14:15:56 +0200 Subject: IB/mlx5: Fix long EEH recover time with NVMe offloads On NVMe offloads connection with many IO queues, EEH takes long time to recover. The culprit is the synchronize_srcu in the destroy_mkey. The solution is to use synchronize_srcu only for ODP mkey. Fixes: b4cfe447d47b ("IB/mlx5: Implement on demand paging by adding support for MMU notifiers") Signed-off-by: Huy Nguyen Reviewed-by: Daniel Jurgens Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/mr.c | 19 ++++++++++++++++--- 1 file changed, 16 insertions(+), 3 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c index fd6ea1f75085..1bd8c1b1dba1 100644 --- a/drivers/infiniband/hw/mlx5/mr.c +++ b/drivers/infiniband/hw/mlx5/mr.c @@ -73,7 +73,8 @@ static int destroy_mkey(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr) #ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING /* Wait until all page fault handlers using the mr complete. */ - synchronize_srcu(&dev->mr_srcu); + if (mr->umem && mr->umem->is_odp) + synchronize_srcu(&dev->mr_srcu); #endif return err; @@ -237,6 +238,9 @@ static void remove_keys(struct mlx5_ib_dev *dev, int c, int num) { struct mlx5_mr_cache *cache = &dev->cache; struct mlx5_cache_ent *ent = &cache->ent[c]; +#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING + bool odp_mkey_exist = false; +#endif struct mlx5_ib_mr *tmp_mr; struct mlx5_ib_mr *mr; LIST_HEAD(del_list); @@ -249,6 +253,10 @@ static void remove_keys(struct mlx5_ib_dev *dev, int c, int num) break; } mr = list_first_entry(&ent->head, struct mlx5_ib_mr, list); +#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING + if (mr->umem && mr->umem->is_odp) + odp_mkey_exist = true; +#endif list_move(&mr->list, &del_list); ent->cur--; ent->size--; @@ -257,7 +265,8 @@ static void remove_keys(struct mlx5_ib_dev *dev, int c, int num) } #ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING - synchronize_srcu(&dev->mr_srcu); + if (odp_mkey_exist) + synchronize_srcu(&dev->mr_srcu); #endif list_for_each_entry_safe(mr, tmp_mr, &del_list, list) { @@ -572,6 +581,7 @@ static void clean_keys(struct mlx5_ib_dev *dev, int c) { struct mlx5_mr_cache *cache = &dev->cache; struct mlx5_cache_ent *ent = &cache->ent[c]; + bool odp_mkey_exist = false; struct mlx5_ib_mr *tmp_mr; struct mlx5_ib_mr *mr; LIST_HEAD(del_list); @@ -584,6 +594,8 @@ static void clean_keys(struct mlx5_ib_dev *dev, int c) break; } mr = list_first_entry(&ent->head, struct mlx5_ib_mr, list); + if (mr->umem && mr->umem->is_odp) + odp_mkey_exist = true; list_move(&mr->list, &del_list); ent->cur--; ent->size--; @@ -592,7 +604,8 @@ static void clean_keys(struct mlx5_ib_dev *dev, int c) } #ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING - synchronize_srcu(&dev->mr_srcu); + if (odp_mkey_exist) + synchronize_srcu(&dev->mr_srcu); #endif list_for_each_entry_safe(mr, tmp_mr, &del_list, list) { -- cgit v1.2.3-59-g8ed1b From e7c4d8e6045171896d763c551b90c672fb5ab42b Mon Sep 17 00:00:00 2001 From: YueHaibing Date: Fri, 21 Dec 2018 02:05:57 +0000 Subject: IB/mlx4: Remove set but not used variable 'pd' Fixes gcc '-Wunused-but-set-variable' warning: drivers/infiniband/hw/mlx4/qp.c: In function '_mlx4_ib_destroy_qp': drivers/infiniband/hw/mlx4/qp.c:1612:22: warning: variable 'pd' set but not used [-Wunused-but-set-variable] Fixes: e00b64f7c54c ("RDMA: Cleanup undesired pd->uobject usage") Signed-off-by: YueHaibing Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx4/qp.c | 3 --- 1 file changed, 3 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c index 24ee30f1cb45..971e9a9ebdaf 100644 --- a/drivers/infiniband/hw/mlx4/qp.c +++ b/drivers/infiniband/hw/mlx4/qp.c @@ -1609,9 +1609,6 @@ static int _mlx4_ib_destroy_qp(struct ib_qp *qp) if (qp->rwq_ind_tbl) { destroy_qp_rss(dev, mqp); } else { - struct mlx4_ib_pd *pd; - - pd = get_pd(mqp); destroy_qp_common(dev, mqp, MLX4_IB_QP_SRC, qp->uobject); } -- cgit v1.2.3-59-g8ed1b From 623d154305c06c83924b8fd2afbfdeba3b396fb7 Mon Sep 17 00:00:00 2001 From: Jason Gunthorpe Date: Thu, 20 Dec 2018 16:39:26 -0700 Subject: IB/mlx5: Fix wrong error unwind The destroy_workqueue on error unwind is missing, and the code jumps to the wrong exit label. Fixes: 813e90b1aeaa ("IB/mlx5: Add advise_mr() support") Acked-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/main.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index 9b40ec73cc64..75edb0804358 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -5775,12 +5775,14 @@ int mlx5_ib_stage_init_init(struct mlx5_ib_dev *dev) dev->advise_mr_wq = alloc_ordered_workqueue("mlx5_ib_advise_mr_wq", 0); if (!dev->advise_mr_wq) { err = -ENOMEM; - goto err_free_port; + goto err_mp; } err = init_srcu_struct(&dev->mr_srcu); - if (err) - goto err_free_port; + if (err) { + destroy_workqueue(dev->advise_mr_wq); + goto err_mp; + } #endif return 0; -- cgit v1.2.3-59-g8ed1b From aa74be6eeade1caf3ef945ff1812059f3d7dc947 Mon Sep 17 00:00:00 2001 From: Yishai Hadas Date: Sun, 9 Dec 2018 12:52:36 +0200 Subject: IB/mlx5: Allocate the per-port Q counter shared when DEVX is supported The per-port Q counter is some kernel resource and as such may be used by few UID(s) upon DEVX usage. To enable using it for QP/RQ when DEVX context is used need to allocate it with a sharing mode indication to let firmware allows its usage. The UID = 0xffff was chosen to mark it. Signed-off-by: Yishai Hadas Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/cmd.c | 17 +++++++++++++++++ drivers/infiniband/hw/mlx5/cmd.h | 2 ++ drivers/infiniband/hw/mlx5/main.c | 9 +++++++-- 3 files changed, 26 insertions(+), 2 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/cmd.c b/drivers/infiniband/hw/mlx5/cmd.c index 59d569c6a13a..356bccc715ee 100644 --- a/drivers/infiniband/hw/mlx5/cmd.c +++ b/drivers/infiniband/hw/mlx5/cmd.c @@ -328,3 +328,20 @@ int mlx5_cmd_xrcd_dealloc(struct mlx5_core_dev *dev, u32 xrcdn, u16 uid) MLX5_SET(dealloc_xrcd_in, in, uid, uid); return mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out)); } + +int mlx5_cmd_alloc_q_counter(struct mlx5_core_dev *dev, u16 *counter_id, + u16 uid) +{ + u32 in[MLX5_ST_SZ_DW(alloc_q_counter_in)] = {0}; + u32 out[MLX5_ST_SZ_DW(alloc_q_counter_out)] = {0}; + int err; + + MLX5_SET(alloc_q_counter_in, in, opcode, MLX5_CMD_OP_ALLOC_Q_COUNTER); + MLX5_SET(alloc_q_counter_in, in, uid, uid); + + err = mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out)); + if (!err) + *counter_id = MLX5_GET(alloc_q_counter_out, out, + counter_set_id); + return err; +} diff --git a/drivers/infiniband/hw/mlx5/cmd.h b/drivers/infiniband/hw/mlx5/cmd.h index c03c56455534..1e76dc67a369 100644 --- a/drivers/infiniband/hw/mlx5/cmd.h +++ b/drivers/infiniband/hw/mlx5/cmd.h @@ -61,4 +61,6 @@ int mlx5_cmd_detach_mcg(struct mlx5_core_dev *dev, union ib_gid *mgid, u32 qpn, u16 uid); int mlx5_cmd_xrcd_alloc(struct mlx5_core_dev *dev, u32 *xrcdn, u16 uid); int mlx5_cmd_xrcd_dealloc(struct mlx5_core_dev *dev, u32 xrcdn, u16 uid); +int mlx5_cmd_alloc_q_counter(struct mlx5_core_dev *dev, u16 *counter_id, + u16 uid); #endif /* MLX5_IB_CMD_H */ diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index 75edb0804358..a57183413e8b 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -5115,6 +5115,9 @@ static int mlx5_ib_alloc_counters(struct mlx5_ib_dev *dev) { int err = 0; int i; + bool is_shared; + + is_shared = MLX5_CAP_GEN(dev->mdev, log_max_uctx) != 0; for (i = 0; i < dev->num_ports; i++) { err = __mlx5_ib_alloc_counters(dev, &dev->port[i].cnts); @@ -5124,8 +5127,10 @@ static int mlx5_ib_alloc_counters(struct mlx5_ib_dev *dev) mlx5_ib_fill_counters(dev, dev->port[i].cnts.names, dev->port[i].cnts.offsets); - err = mlx5_core_alloc_q_counter(dev->mdev, - &dev->port[i].cnts.set_id); + err = mlx5_cmd_alloc_q_counter(dev->mdev, + &dev->port[i].cnts.set_id, + is_shared ? + MLX5_SHARED_RESOURCE_UID : 0); if (err) { mlx5_ib_warn(dev, "couldn't allocate queue counter for port %d, err %d\n", -- cgit v1.2.3-59-g8ed1b From 58f7c0bfb4e111006ba376df7ab266a6d42d64ba Mon Sep 17 00:00:00 2001 From: Dan Carpenter Date: Sat, 22 Dec 2018 10:43:32 +0300 Subject: RDMA/mlx5: Signedness bug in UVERBS_HANDLER() The "num_actions" variable needs to be signed for the error handling to work. The maximum number of actions is less than 256 so int type is large enough for that. Fixes: cbfdd442c43e ("IB/uverbs: Add helper to get array size from ptr attribute") Signed-off-by: Dan Carpenter Acked-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/flow.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/flow.c b/drivers/infiniband/hw/mlx5/flow.c index 95cf4b9fc65d..e8a1e4498e3f 100644 --- a/drivers/infiniband/hw/mlx5/flow.c +++ b/drivers/infiniband/hw/mlx5/flow.c @@ -329,7 +329,7 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_FLOW_ACTION_CREATE_MODIFY_HEADER)( struct mlx5_ib_dev *mdev = to_mdev(uobj->context->device); enum mlx5_ib_uapi_flow_table_type ft_type; struct ib_flow_action *action; - size_t num_actions; + int num_actions; void *in; int ret; -- cgit v1.2.3-59-g8ed1b