From bb6c96d72879fe1f674a804eb95b891def4ace61 Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Thu, 24 Sep 2015 10:34:21 +0300 Subject: [PATCH 1/7] xprtrdma: Replace global lkey with lkey local to PD The core API has changed so that devices that do not have a global DMA lkey automatically create an mr, per-PD, and make that lkey available. The global DMA lkey interface is going away in favor of the per-PD DMA lkey. The per-PD DMA lkey is always available. Convert xprtrdma to use the device's per-PD DMA lkey for regbufs, no matter which memory registration scheme is in use. Signed-off-by: Chuck Lever Signed-off-by: Sagi Grimberg Cc: linux-nfs Acked-by: Anna Schumaker Signed-off-by: Doug Ledford --- net/sunrpc/xprtrdma/fmr_ops.c | 19 ------------------- net/sunrpc/xprtrdma/frwr_ops.c | 5 ----- net/sunrpc/xprtrdma/physical_ops.c | 10 +--------- net/sunrpc/xprtrdma/verbs.c | 2 +- net/sunrpc/xprtrdma/xprt_rdma.h | 1 - 5 files changed, 2 insertions(+), 35 deletions(-) diff --git a/net/sunrpc/xprtrdma/fmr_ops.c b/net/sunrpc/xprtrdma/fmr_ops.c index cb25c89da623..f1e8dafbd507 100644 --- a/net/sunrpc/xprtrdma/fmr_ops.c +++ b/net/sunrpc/xprtrdma/fmr_ops.c @@ -39,25 +39,6 @@ static int fmr_op_open(struct rpcrdma_ia *ia, struct rpcrdma_ep *ep, struct rpcrdma_create_data_internal *cdata) { - struct ib_device_attr *devattr = &ia->ri_devattr; - struct ib_mr *mr; - - /* Obtain an lkey to use for the regbufs, which are - * protected from remote access. - */ - if (devattr->device_cap_flags & IB_DEVICE_LOCAL_DMA_LKEY) { - ia->ri_dma_lkey = ia->ri_device->local_dma_lkey; - } else { - mr = ib_get_dma_mr(ia->ri_pd, IB_ACCESS_LOCAL_WRITE); - if (IS_ERR(mr)) { - pr_err("%s: ib_get_dma_mr for failed with %lX\n", - __func__, PTR_ERR(mr)); - return -ENOMEM; - } - ia->ri_dma_lkey = ia->ri_dma_mr->lkey; - ia->ri_dma_mr = mr; - } - return 0; } diff --git a/net/sunrpc/xprtrdma/frwr_ops.c b/net/sunrpc/xprtrdma/frwr_ops.c index d6653f5d0830..5318951b3b53 100644 --- a/net/sunrpc/xprtrdma/frwr_ops.c +++ b/net/sunrpc/xprtrdma/frwr_ops.c @@ -189,11 +189,6 @@ frwr_op_open(struct rpcrdma_ia *ia, struct rpcrdma_ep *ep, struct ib_device_attr *devattr = &ia->ri_devattr; int depth, delta; - /* Obtain an lkey to use for the regbufs, which are - * protected from remote access. - */ - ia->ri_dma_lkey = ia->ri_device->local_dma_lkey; - ia->ri_max_frmr_depth = min_t(unsigned int, RPCRDMA_MAX_DATA_SEGS, devattr->max_fast_reg_page_list_len); diff --git a/net/sunrpc/xprtrdma/physical_ops.c b/net/sunrpc/xprtrdma/physical_ops.c index 72cf8b15bbb4..617b76f22154 100644 --- a/net/sunrpc/xprtrdma/physical_ops.c +++ b/net/sunrpc/xprtrdma/physical_ops.c @@ -23,7 +23,6 @@ static int physical_op_open(struct rpcrdma_ia *ia, struct rpcrdma_ep *ep, struct rpcrdma_create_data_internal *cdata) { - struct ib_device_attr *devattr = &ia->ri_devattr; struct ib_mr *mr; /* Obtain an rkey to use for RPC data payloads. @@ -37,15 +36,8 @@ physical_op_open(struct rpcrdma_ia *ia, struct rpcrdma_ep *ep, __func__, PTR_ERR(mr)); return -ENOMEM; } + ia->ri_dma_mr = mr; - - /* Obtain an lkey to use for regbufs. - */ - if (devattr->device_cap_flags & IB_DEVICE_LOCAL_DMA_LKEY) - ia->ri_dma_lkey = ia->ri_device->local_dma_lkey; - else - ia->ri_dma_lkey = ia->ri_dma_mr->lkey; - return 0; } diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c index 682996779970..eb081ad05e33 100644 --- a/net/sunrpc/xprtrdma/verbs.c +++ b/net/sunrpc/xprtrdma/verbs.c @@ -1252,7 +1252,7 @@ rpcrdma_alloc_regbuf(struct rpcrdma_ia *ia, size_t size, gfp_t flags) goto out_free; iov->length = size; - iov->lkey = ia->ri_dma_lkey; + iov->lkey = ia->ri_pd->local_dma_lkey; rb->rg_size = size; rb->rg_owner = NULL; return rb; diff --git a/net/sunrpc/xprtrdma/xprt_rdma.h b/net/sunrpc/xprtrdma/xprt_rdma.h index 02512221b8bc..c09414e6f91b 100644 --- a/net/sunrpc/xprtrdma/xprt_rdma.h +++ b/net/sunrpc/xprtrdma/xprt_rdma.h @@ -65,7 +65,6 @@ struct rpcrdma_ia { struct rdma_cm_id *ri_id; struct ib_pd *ri_pd; struct ib_mr *ri_dma_mr; - u32 ri_dma_lkey; struct completion ri_done; int ri_async_rc; unsigned int ri_max_frmr_depth; From 3cffd930171518821595839c5ce7036894ef0d74 Mon Sep 17 00:00:00 2001 From: Sagi Grimberg Date: Thu, 24 Sep 2015 10:34:22 +0300 Subject: [PATCH 2/7] IB/iser: Add module parameter for always register memory This module parameter forces memory registration even for a continuous memory region. It is true by default as sending an all-physical rkey with remote permissions might be insecure. Signed-off-by: Sagi Grimberg Signed-off-by: Doug Ledford --- drivers/infiniband/ulp/iser/iscsi_iser.c | 5 +++++ drivers/infiniband/ulp/iser/iscsi_iser.h | 1 + drivers/infiniband/ulp/iser/iser_memory.c | 18 ++++++++++++------ drivers/infiniband/ulp/iser/iser_verbs.c | 21 +++++++++++++-------- 4 files changed, 31 insertions(+), 14 deletions(-) diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c b/drivers/infiniband/ulp/iser/iscsi_iser.c index 1ace5d83a4d7..f58ff96b6cbb 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.c +++ b/drivers/infiniband/ulp/iser/iscsi_iser.c @@ -97,6 +97,11 @@ unsigned int iser_max_sectors = ISER_DEF_MAX_SECTORS; module_param_named(max_sectors, iser_max_sectors, uint, S_IRUGO | S_IWUSR); MODULE_PARM_DESC(max_sectors, "Max number of sectors in a single scsi command (default:1024"); +bool iser_always_reg = true; +module_param_named(always_register, iser_always_reg, bool, S_IRUGO); +MODULE_PARM_DESC(always_register, + "Always register memory, even for continuous memory regions (default:true)"); + bool iser_pi_enable = false; module_param_named(pi_enable, iser_pi_enable, bool, S_IRUGO); MODULE_PARM_DESC(pi_enable, "Enable T10-PI offload support (default:disabled)"); diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.h b/drivers/infiniband/ulp/iser/iscsi_iser.h index 86f6583485ef..a5edd6ede692 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.h +++ b/drivers/infiniband/ulp/iser/iscsi_iser.h @@ -611,6 +611,7 @@ extern int iser_debug_level; extern bool iser_pi_enable; extern int iser_pi_guard; extern unsigned int iser_max_sectors; +extern bool iser_always_reg; int iser_assign_reg_ops(struct iser_device *device); diff --git a/drivers/infiniband/ulp/iser/iser_memory.c b/drivers/infiniband/ulp/iser/iser_memory.c index 2493cc748db8..4c46d67d37a1 100644 --- a/drivers/infiniband/ulp/iser/iser_memory.c +++ b/drivers/infiniband/ulp/iser/iser_memory.c @@ -803,11 +803,12 @@ static int iser_reg_prot_sg(struct iscsi_iser_task *task, struct iser_data_buf *mem, struct iser_fr_desc *desc, + bool use_dma_key, struct iser_mem_reg *reg) { struct iser_device *device = task->iser_conn->ib_conn.device; - if (mem->dma_nents == 1) + if (use_dma_key) return iser_reg_dma(device, mem, reg); return device->reg_ops->reg_mem(task, mem, &desc->pi_ctx->rsc, reg); @@ -817,11 +818,12 @@ static int iser_reg_data_sg(struct iscsi_iser_task *task, struct iser_data_buf *mem, struct iser_fr_desc *desc, + bool use_dma_key, struct iser_mem_reg *reg) { struct iser_device *device = task->iser_conn->ib_conn.device; - if (mem->dma_nents == 1) + if (use_dma_key) return iser_reg_dma(device, mem, reg); return device->reg_ops->reg_mem(task, mem, &desc->rsc, reg); @@ -836,14 +838,17 @@ int iser_reg_rdma_mem(struct iscsi_iser_task *task, struct iser_mem_reg *reg = &task->rdma_reg[dir]; struct iser_mem_reg *data_reg; struct iser_fr_desc *desc = NULL; + bool use_dma_key; int err; err = iser_handle_unaligned_buf(task, mem, dir); if (unlikely(err)) return err; - if (mem->dma_nents != 1 || - scsi_get_prot_op(task->sc) != SCSI_PROT_NORMAL) { + use_dma_key = (mem->dma_nents == 1 && !iser_always_reg && + scsi_get_prot_op(task->sc) == SCSI_PROT_NORMAL); + + if (!use_dma_key) { desc = device->reg_ops->reg_desc_get(ib_conn); reg->mem_h = desc; } @@ -853,7 +858,7 @@ int iser_reg_rdma_mem(struct iscsi_iser_task *task, else data_reg = &task->desc.data_reg; - err = iser_reg_data_sg(task, mem, desc, data_reg); + err = iser_reg_data_sg(task, mem, desc, use_dma_key, data_reg); if (unlikely(err)) goto err_reg; @@ -866,7 +871,8 @@ int iser_reg_rdma_mem(struct iscsi_iser_task *task, if (unlikely(err)) goto err_reg; - err = iser_reg_prot_sg(task, mem, desc, prot_reg); + err = iser_reg_prot_sg(task, mem, desc, + use_dma_key, prot_reg); if (unlikely(err)) goto err_reg; } diff --git a/drivers/infiniband/ulp/iser/iser_verbs.c b/drivers/infiniband/ulp/iser/iser_verbs.c index ae70cc1463ac..85132d867bc8 100644 --- a/drivers/infiniband/ulp/iser/iser_verbs.c +++ b/drivers/infiniband/ulp/iser/iser_verbs.c @@ -133,11 +133,15 @@ static int iser_create_device_ib_res(struct iser_device *device) (unsigned long)comp); } - device->mr = ib_get_dma_mr(device->pd, IB_ACCESS_LOCAL_WRITE | - IB_ACCESS_REMOTE_WRITE | - IB_ACCESS_REMOTE_READ); - if (IS_ERR(device->mr)) - goto dma_mr_err; + if (!iser_always_reg) { + int access = IB_ACCESS_LOCAL_WRITE | + IB_ACCESS_REMOTE_WRITE | + IB_ACCESS_REMOTE_READ; + + device->mr = ib_get_dma_mr(device->pd, access); + if (IS_ERR(device->mr)) + goto dma_mr_err; + } INIT_IB_EVENT_HANDLER(&device->event_handler, device->ib_device, iser_event_handler); @@ -147,7 +151,8 @@ static int iser_create_device_ib_res(struct iser_device *device) return 0; handler_err: - ib_dereg_mr(device->mr); + if (device->mr) + ib_dereg_mr(device->mr); dma_mr_err: for (i = 0; i < device->comps_used; i++) tasklet_kill(&device->comps[i].tasklet); @@ -173,7 +178,6 @@ static int iser_create_device_ib_res(struct iser_device *device) static void iser_free_device_ib_res(struct iser_device *device) { int i; - BUG_ON(device->mr == NULL); for (i = 0; i < device->comps_used; i++) { struct iser_comp *comp = &device->comps[i]; @@ -184,7 +188,8 @@ static void iser_free_device_ib_res(struct iser_device *device) } (void)ib_unregister_event_handler(&device->event_handler); - (void)ib_dereg_mr(device->mr); + if (device->mr) + (void)ib_dereg_mr(device->mr); ib_dealloc_pd(device->pd); kfree(device->comps); From c6790aa9f4fdc26b1246ba36da2fd749663beb65 Mon Sep 17 00:00:00 2001 From: Sagi Grimberg Date: Thu, 24 Sep 2015 10:34:23 +0300 Subject: [PATCH 3/7] IB/mlx5: Remove support for IB_DEVICE_LOCAL_DMA_LKEY Commit 96249d70dd70 ("IB/core: Guarantee that a local_dma_lkey is available") allows ULPs that make use of the local dma key to keep working as before by allocating a DMA MR with local permissions and converted these consumers to use the MR associated with the PD rather then device->local_dma_lkey. ConnectIB has some known issues with memory registration using the local_dma_lkey (SEND, RDMA, RECV seems to work ok). Thus don't expose support for it (remove device->local_dma_lkey setting), and take advantage of the above commit such that no regression is introduced to working systems. The local_dma_lkey support will be restored in CX4 depending on FW capability query. Signed-off-by: Sagi Grimberg Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx5/main.c | 10 +-------- drivers/net/ethernet/mellanox/mlx5/core/fw.c | 22 -------------------- include/linux/mlx5/device.h | 11 ---------- include/linux/mlx5/driver.h | 1 - 4 files changed, 1 insertion(+), 43 deletions(-) diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index 41d6911e244e..0ab9625911a1 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -245,7 +245,6 @@ static int mlx5_ib_query_device(struct ib_device *ibdev, props->device_cap_flags |= IB_DEVICE_BAD_QKEY_CNTR; if (MLX5_CAP_GEN(mdev, apm)) props->device_cap_flags |= IB_DEVICE_AUTO_PATH_MIG; - props->device_cap_flags |= IB_DEVICE_LOCAL_DMA_LKEY; if (MLX5_CAP_GEN(mdev, xrc)) props->device_cap_flags |= IB_DEVICE_XRC; props->device_cap_flags |= IB_DEVICE_MEM_MGT_EXTENSIONS; @@ -1245,18 +1244,10 @@ static int create_dev_resources(struct mlx5_ib_resources *devr) struct ib_srq_init_attr attr; struct mlx5_ib_dev *dev; struct ib_cq_init_attr cq_attr = {.cqe = 1}; - u32 rsvd_lkey; int ret = 0; dev = container_of(devr, struct mlx5_ib_dev, devr); - ret = mlx5_core_query_special_context(dev->mdev, &rsvd_lkey); - if (ret) { - pr_err("Failed to query special context %d\n", ret); - return ret; - } - dev->ib_dev.local_dma_lkey = rsvd_lkey; - devr->p0 = mlx5_ib_alloc_pd(&dev->ib_dev, NULL, NULL); if (IS_ERR(devr->p0)) { ret = PTR_ERR(devr->p0); @@ -1418,6 +1409,7 @@ static void *mlx5_ib_add(struct mlx5_core_dev *mdev) strlcpy(dev->ib_dev.name, "mlx5_%d", IB_DEVICE_NAME_MAX); dev->ib_dev.owner = THIS_MODULE; dev->ib_dev.node_type = RDMA_NODE_IB_CA; + dev->ib_dev.local_dma_lkey = 0 /* not supported for now */; dev->num_ports = MLX5_CAP_GEN(mdev, num_ports); dev->ib_dev.phys_port_cnt = dev->num_ports; dev->ib_dev.num_comp_vectors = diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fw.c b/drivers/net/ethernet/mellanox/mlx5/core/fw.c index aa0d5ffe92d8..9335e5ae18cc 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/fw.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/fw.c @@ -200,25 +200,3 @@ int mlx5_cmd_teardown_hca(struct mlx5_core_dev *dev) return err; } - -int mlx5_core_query_special_context(struct mlx5_core_dev *dev, u32 *rsvd_lkey) -{ - struct mlx5_cmd_query_special_contexts_mbox_in in; - struct mlx5_cmd_query_special_contexts_mbox_out out; - int err; - - memset(&in, 0, sizeof(in)); - memset(&out, 0, sizeof(out)); - in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_QUERY_SPECIAL_CONTEXTS); - err = mlx5_cmd_exec(dev, &in, sizeof(in), &out, sizeof(out)); - if (err) - return err; - - if (out.hdr.status) - err = mlx5_cmd_status_to_err(&out.hdr); - - *rsvd_lkey = be32_to_cpu(out.resd_lkey); - - return err; -} -EXPORT_SYMBOL(mlx5_core_query_special_context); diff --git a/include/linux/mlx5/device.h b/include/linux/mlx5/device.h index 8eb3b19af2a4..250b1ff8b48d 100644 --- a/include/linux/mlx5/device.h +++ b/include/linux/mlx5/device.h @@ -402,17 +402,6 @@ struct mlx5_cmd_teardown_hca_mbox_out { u8 rsvd[8]; }; -struct mlx5_cmd_query_special_contexts_mbox_in { - struct mlx5_inbox_hdr hdr; - u8 rsvd[8]; -}; - -struct mlx5_cmd_query_special_contexts_mbox_out { - struct mlx5_outbox_hdr hdr; - __be32 dump_fill_mkey; - __be32 resd_lkey; -}; - struct mlx5_cmd_layout { u8 type; u8 rsvd0[3]; diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h index 27b53f9a24ad..8b6d6f2154a4 100644 --- a/include/linux/mlx5/driver.h +++ b/include/linux/mlx5/driver.h @@ -845,7 +845,6 @@ void *mlx5_get_protocol_dev(struct mlx5_core_dev *mdev, int protocol); int mlx5_register_interface(struct mlx5_interface *intf); void mlx5_unregister_interface(struct mlx5_interface *intf); int mlx5_core_query_vendor_id(struct mlx5_core_dev *mdev, u32 *vendor_id); -int mlx5_core_query_special_context(struct mlx5_core_dev *dev, u32 *rsvd_lkey); struct mlx5_profile { u64 mask; From 81fb5e26a9d05674c048803a20cb8f08a1b1c9b8 Mon Sep 17 00:00:00 2001 From: Sagi Grimberg Date: Thu, 24 Sep 2015 10:34:24 +0300 Subject: [PATCH 4/7] IB/mlx5: Remove pa_lkey usages Since mlx5 driver cannot rely on registration using the reserved lkey (global_dma_lkey) it used to allocate a private physical address lkey for each allocated pd. Commit 96249d70dd70 ("IB/core: Guarantee that a local_dma_lkey is available") just does it in the core layer so we can go ahead and use that. Signed-off-by: Sagi Grimberg Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx5/main.c | 57 ---------------------------- drivers/infiniband/hw/mlx5/mlx5_ib.h | 2 - drivers/infiniband/hw/mlx5/qp.c | 4 +- 3 files changed, 1 insertion(+), 62 deletions(-) diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index 0ab9625911a1..f1ccd40beae9 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -794,53 +794,6 @@ static int mlx5_ib_mmap(struct ib_ucontext *ibcontext, struct vm_area_struct *vm return 0; } -static int alloc_pa_mkey(struct mlx5_ib_dev *dev, u32 *key, u32 pdn) -{ - struct mlx5_create_mkey_mbox_in *in; - struct mlx5_mkey_seg *seg; - struct mlx5_core_mr mr; - int err; - - in = kzalloc(sizeof(*in), GFP_KERNEL); - if (!in) - return -ENOMEM; - - seg = &in->seg; - seg->flags = MLX5_PERM_LOCAL_READ | MLX5_ACCESS_MODE_PA; - seg->flags_pd = cpu_to_be32(pdn | MLX5_MKEY_LEN64); - seg->qpn_mkey7_0 = cpu_to_be32(0xffffff << 8); - seg->start_addr = 0; - - err = mlx5_core_create_mkey(dev->mdev, &mr, in, sizeof(*in), - NULL, NULL, NULL); - if (err) { - mlx5_ib_warn(dev, "failed to create mkey, %d\n", err); - goto err_in; - } - - kfree(in); - *key = mr.key; - - return 0; - -err_in: - kfree(in); - - return err; -} - -static void free_pa_mkey(struct mlx5_ib_dev *dev, u32 key) -{ - struct mlx5_core_mr mr; - int err; - - memset(&mr, 0, sizeof(mr)); - mr.key = key; - err = mlx5_core_destroy_mkey(dev->mdev, &mr); - if (err) - mlx5_ib_warn(dev, "failed to destroy mkey 0x%x\n", key); -} - static struct ib_pd *mlx5_ib_alloc_pd(struct ib_device *ibdev, struct ib_ucontext *context, struct ib_udata *udata) @@ -866,13 +819,6 @@ static struct ib_pd *mlx5_ib_alloc_pd(struct ib_device *ibdev, kfree(pd); return ERR_PTR(-EFAULT); } - } else { - err = alloc_pa_mkey(to_mdev(ibdev), &pd->pa_lkey, pd->pdn); - if (err) { - mlx5_core_dealloc_pd(to_mdev(ibdev)->mdev, pd->pdn); - kfree(pd); - return ERR_PTR(err); - } } return &pd->ibpd; @@ -883,9 +829,6 @@ static int mlx5_ib_dealloc_pd(struct ib_pd *pd) struct mlx5_ib_dev *mdev = to_mdev(pd->device); struct mlx5_ib_pd *mpd = to_mpd(pd); - if (!pd->uobject) - free_pa_mkey(mdev, mpd->pa_lkey); - mlx5_core_dealloc_pd(mdev->mdev, mpd->pdn); kfree(mpd); diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index bb8cda79e881..22123b79d550 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -103,7 +103,6 @@ static inline struct mlx5_ib_ucontext *to_mucontext(struct ib_ucontext *ibuconte struct mlx5_ib_pd { struct ib_pd ibpd; u32 pdn; - u32 pa_lkey; }; /* Use macros here so that don't have to duplicate @@ -213,7 +212,6 @@ struct mlx5_ib_qp { int uuarn; int create_type; - u32 pa_lkey; /* Store signature errors */ bool signature_en; diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c index c745c6c5e10d..6f521a3418e8 100644 --- a/drivers/infiniband/hw/mlx5/qp.c +++ b/drivers/infiniband/hw/mlx5/qp.c @@ -925,8 +925,6 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd, err = create_kernel_qp(dev, init_attr, qp, &in, &inlen); if (err) mlx5_ib_dbg(dev, "err %d\n", err); - else - qp->pa_lkey = to_mpd(pd)->pa_lkey; } if (err) @@ -2045,7 +2043,7 @@ static void set_frwr_pages(struct mlx5_wqe_data_seg *dseg, mfrpl->mapped_page_list[i] = cpu_to_be64(page_list[i] | perm); dseg->addr = cpu_to_be64(mfrpl->map); dseg->byte_count = cpu_to_be32(ALIGN(sizeof(u64) * wr->wr.fast_reg.page_list_len, 64)); - dseg->lkey = cpu_to_be32(pd->pa_lkey); + dseg->lkey = cpu_to_be32(pd->ibpd.local_dma_lkey); } static __be32 send_ieth(struct ib_send_wr *wr) From bd99b2e05c4df2a428e5c9dd338289089d0e26df Mon Sep 17 00:00:00 2001 From: Christoph Lameter Date: Thu, 24 Sep 2015 12:00:05 -0500 Subject: [PATCH 5/7] IB/ipoib: Expire sendonly multicast joins On neighbor expiration, check to see if the neighbor was actually a sendonly multicast join, and if so, leave the multicast group as we expire the neighbor. Signed-off-by: Christoph Lameter Signed-off-by: Doug Ledford --- drivers/infiniband/ulp/ipoib/ipoib.h | 2 ++ drivers/infiniband/ulp/ipoib/ipoib_main.c | 18 ++++++++++++++++++ drivers/infiniband/ulp/ipoib/ipoib_multicast.c | 4 ++-- 3 files changed, 22 insertions(+), 2 deletions(-) diff --git a/drivers/infiniband/ulp/ipoib/ipoib.h b/drivers/infiniband/ulp/ipoib/ipoib.h index ca2873698d75..4ff4e5290354 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib.h +++ b/drivers/infiniband/ulp/ipoib/ipoib.h @@ -548,6 +548,8 @@ void ipoib_path_iter_read(struct ipoib_path_iter *iter, int ipoib_mcast_attach(struct net_device *dev, u16 mlid, union ib_gid *mgid, int set_qkey); +int ipoib_mcast_leave(struct net_device *dev, struct ipoib_mcast *mcast); +struct ipoib_mcast *__ipoib_mcast_find(struct net_device *dev, void *mgid); int ipoib_init_qp(struct net_device *dev); int ipoib_transport_dev_init(struct net_device *dev, struct ib_device *ca); diff --git a/drivers/infiniband/ulp/ipoib/ipoib_main.c b/drivers/infiniband/ulp/ipoib/ipoib_main.c index 36536ce5a3e2..f74316e679d2 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_main.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c @@ -1149,6 +1149,9 @@ static void __ipoib_reap_neigh(struct ipoib_dev_priv *priv) unsigned long dt; unsigned long flags; int i; + LIST_HEAD(remove_list); + struct ipoib_mcast *mcast, *tmcast; + struct net_device *dev = priv->dev; if (test_bit(IPOIB_STOP_NEIGH_GC, &priv->flags)) return; @@ -1176,6 +1179,19 @@ static void __ipoib_reap_neigh(struct ipoib_dev_priv *priv) lockdep_is_held(&priv->lock))) != NULL) { /* was the neigh idle for two GC periods */ if (time_after(neigh_obsolete, neigh->alive)) { + u8 *mgid = neigh->daddr + 4; + + /* Is this multicast ? */ + if (*mgid == 0xff) { + mcast = __ipoib_mcast_find(dev, mgid); + + if (mcast && test_bit(IPOIB_MCAST_FLAG_SENDONLY, &mcast->flags)) { + list_del(&mcast->list); + rb_erase(&mcast->rb_node, &priv->multicast_tree); + list_add_tail(&mcast->list, &remove_list); + } + } + rcu_assign_pointer(*np, rcu_dereference_protected(neigh->hnext, lockdep_is_held(&priv->lock))); @@ -1191,6 +1207,8 @@ static void __ipoib_reap_neigh(struct ipoib_dev_priv *priv) out_unlock: spin_unlock_irqrestore(&priv->lock, flags); + list_for_each_entry_safe(mcast, tmcast, &remove_list, list) + ipoib_mcast_leave(dev, mcast); } static void ipoib_reap_neigh(struct work_struct *work) diff --git a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c index 09a1748f9d13..33131001fa24 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c @@ -153,7 +153,7 @@ static struct ipoib_mcast *ipoib_mcast_alloc(struct net_device *dev, return mcast; } -static struct ipoib_mcast *__ipoib_mcast_find(struct net_device *dev, void *mgid) +struct ipoib_mcast *__ipoib_mcast_find(struct net_device *dev, void *mgid) { struct ipoib_dev_priv *priv = netdev_priv(dev); struct rb_node *n = priv->multicast_tree.rb_node; @@ -675,7 +675,7 @@ int ipoib_mcast_stop_thread(struct net_device *dev) return 0; } -static int ipoib_mcast_leave(struct net_device *dev, struct ipoib_mcast *mcast) +int ipoib_mcast_leave(struct net_device *dev, struct ipoib_mcast *mcast) { struct ipoib_dev_priv *priv = netdev_priv(dev); int ret = 0; From c3852ab0e606212de523c1fb1e15adbf9f431619 Mon Sep 17 00:00:00 2001 From: Doug Ledford Date: Fri, 25 Sep 2015 14:35:01 -0400 Subject: [PATCH 6/7] IB/ipoib: Make sendonly multicast joins create the mcast group Since IPoIB should, as much as possible, emulate how multicast sends work on Ethernet for regular TCP/IP apps, there should be no requirement to subscribe to a multicast group before your sends are properly sent. However, due to the difference in how multicast is handled on InfiniBand, we must join the appropriate multicast group before we can send to it. Previously we tried not to trigger the auto-create feature of the subnet manager when doing this because we didn't have tracking of these sendonly groups and the auto-creation might never get undone. The previous patch added timing to these sendonly joins and allows us to leave them after a reasonable idle expiration time. So supply all of the information needed to auto-create group. Signed-off-by: Doug Ledford --- .../infiniband/ulp/ipoib/ipoib_multicast.c | 22 ++++++++++--------- 1 file changed, 12 insertions(+), 10 deletions(-) diff --git a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c index 33131001fa24..136cbefe00f8 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c @@ -508,17 +508,19 @@ static void ipoib_mcast_join(struct net_device *dev, struct ipoib_mcast *mcast) rec.hop_limit = priv->broadcast->mcmember.hop_limit; /* - * Historically Linux IPoIB has never properly supported SEND - * ONLY join. It emulated it by not providing all the required - * attributes, which is enough to prevent group creation and - * detect if there are full members or not. A major problem - * with supporting SEND ONLY is detecting when the group is - * auto-destroyed as IPoIB will cache the MLID.. + * Send-only IB Multicast joins do not work at the core + * IB layer yet, so we can't use them here. However, + * we are emulating an Ethernet multicast send, which + * does not require a multicast subscription and will + * still send properly. The most appropriate thing to + * do is to create the group if it doesn't exist as that + * most closely emulates the behavior, from a user space + * application perspecitive, of Ethernet multicast + * operation. For now, we do a full join, maybe later + * when the core IB layers support send only joins we + * will use them. */ -#if 1 - if (test_bit(IPOIB_MCAST_FLAG_SENDONLY, &mcast->flags)) - comp_mask &= ~IB_SA_MCMEMBER_REC_TRAFFIC_CLASS; -#else +#if 0 if (test_bit(IPOIB_MCAST_FLAG_SENDONLY, &mcast->flags)) rec.join_state = 4; #endif From 2866196f294954ce9fa226825c8c1eaa64c7da8a Mon Sep 17 00:00:00 2001 From: Doug Ledford Date: Fri, 25 Sep 2015 22:30:24 -0400 Subject: [PATCH 7/7] IB/ipoib: increase the max mcast backlog queue When performing sendonly joins, we queue the packets that trigger the join until the join completes. This may take on the order of hundreds of milliseconds. It is easy to have many more than three packets come in during that time. Expand the maximum queue depth in order to try and prevent dropped packets during the time it takes to join the multicast group. Signed-off-by: Doug Ledford --- drivers/infiniband/ulp/ipoib/ipoib.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/infiniband/ulp/ipoib/ipoib.h b/drivers/infiniband/ulp/ipoib/ipoib.h index 4ff4e5290354..4cd5428a2399 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib.h +++ b/drivers/infiniband/ulp/ipoib/ipoib.h @@ -80,7 +80,7 @@ enum { IPOIB_NUM_WC = 4, IPOIB_MAX_PATH_REC_QUEUE = 3, - IPOIB_MAX_MCAST_QUEUE = 3, + IPOIB_MAX_MCAST_QUEUE = 64, IPOIB_FLAG_OPER_UP = 0, IPOIB_FLAG_INITIALIZED = 1,