The branch main has been updated by jhb: URL: https://cgit.FreeBSD.org/src/commit/?id=75cde38e949345c9d63e400bd4707d6eec817941
commit 75cde38e949345c9d63e400bd4707d6eec817941 Author: John Baldwin <[email protected]> AuthorDate: 2026-06-24 19:30:56 +0000 Commit: John Baldwin <[email protected]> CommitDate: 2026-06-24 19:30:56 +0000 OFED: Various changes from Linux 4.17 This contains changes from the following Linux commits: a3b641af72ba RDMA/CM: move rdma_id_private to cma_priv.h 6612b4983f7e IB/core: Fix comments of GID query functions b26c4a1138df IB/{core, ipoib}: Simplify ib_find_gid() for unused ndev 5ac08a341303 IB/cma: Use rdma_protocol_roce() and remove cma_protocol_roce_dev_port() a9c06aeba997 IB/core: Remove rdma_resolve_ip_route() as exported symbol a22af59ea9a5 IB/cm: Add and use a helper function to add cm_id's to the port list cb12a8e2fa5f IB/cm: Introduce and use helper function to get cm_port from path 0a5141593567 IB/core: Refactor ib_init_ah_attr_from_path() for RoCE e41a7c41947d IB/core: Move rdma_addr_find_l2_eth_by_grh to core_priv.h 98f1f4e0ed26 IB/core: Refer to RoCE port property instead of GID table property 114cc9c4b182 IB/cma: Resolve route only while receiving CM requests 9137108cc3d6 RDMA/rdma_cm: Fix use after free race with process_one_req 190fb9c4d130 IB/core: Refer to RoCE port property to decide building cache 22d24f75a193 IB/core: Search GID only for IB link layer 4ab7cb4bf362 IB/core: Refer to RoCE port property instead of GID table property 3401857ea347 IB/core: Generate GID change event regardless of RoCE GID table property 218b9e3eb8b5 RDMA/cma: Move rdma_cm_state to cma_priv.h 8435168d50e6 RDMA/ucma: Don't allow setting RDMA_OPTION_IB_PATH without an RDMA device 72e1ff0fb7e0 RDMA/core: Update query_gid documentation for HCA drivers 0e1f9b924471 RDMA/providers: Simplify query_gid callback of RoCE providers 598ff6bae689 IB/core: Refactor GID modify code for RoCE f35faa4ba956 IB/core: Simplify ib_query_gid to always refer to cache 14169e333e71 IB/providers: Avoid zero GID check for RoCE 414448d249d8 RDMA: Use ib_gid_attr during GID modification ee6548d1d98d RDMA/rdma_cm: Delete rdma_addr_client 44e75052bc2a RDMA/rdma_cm: Make rdma_addr_cancel into a fence e19c0d237873 RDMA/rdma_cm: Remove process_req and timer sorting a66ed149b0da IB/core: Don't allow default GID addition at non reseved slots dc5640f294e4 IB/core: Fix deleting default GIDs when changing mac adddress 22c01ee4b8a8 IB/core: Fix to avoid deleting IPv6 look alike default GIDs 9aa169213d11 RDMA/cma: Do not query GID during QP state transition to RTR aec05afe641b IB/core: Remove redundant return Tested by: Wafa Hamzah <[email protected]> (mlx5_ib) Tested by: John Baldwin <[email protected]> (iw_cxgbe) Sponsored by: Chelsio Communications --- sys/dev/bnxt/bnxt_re/ib_verbs.c | 22 +- sys/dev/bnxt/bnxt_re/ib_verbs.h | 6 +- sys/dev/bnxt/bnxt_re/main.c | 1 - sys/dev/irdma/fbsd_kcompat.h | 6 +- sys/dev/irdma/irdma_kcompat.c | 31 +- sys/dev/irdma/irdma_verbs.c | 1 - sys/dev/mlx4/mlx4_ib/mlx4_ib_main.c | 49 +-- sys/dev/mlx4/mlx4_ib/mlx4_ib_qp.c | 2 - sys/dev/mlx5/mlx5_ib/mlx5_ib_main.c | 13 +- sys/dev/qlnx/qlnxr/qlnxr_os.c | 1 - sys/dev/qlnx/qlnxr/qlnxr_verbs.c | 38 -- sys/dev/qlnx/qlnxr/qlnxr_verbs.h | 5 - sys/ofed/drivers/infiniband/core/cma_priv.h | 95 +++++ sys/ofed/drivers/infiniband/core/core_priv.h | 10 + sys/ofed/drivers/infiniband/core/ib_addr.c | 148 +++---- sys/ofed/drivers/infiniband/core/ib_cache.c | 529 ++++++++++++++----------- sys/ofed/drivers/infiniband/core/ib_cm.c | 53 ++- sys/ofed/drivers/infiniband/core/ib_cma.c | 77 +--- sys/ofed/drivers/infiniband/core/ib_device.c | 18 +- sys/ofed/drivers/infiniband/core/ib_sa_query.c | 192 ++++----- sys/ofed/drivers/infiniband/core/ib_sysfs.c | 16 +- sys/ofed/drivers/infiniband/core/ib_ucma.c | 3 + sys/ofed/drivers/infiniband/core/ib_umem.c | 2 - sys/ofed/include/rdma/ib_addr.h | 29 +- sys/ofed/include/rdma/ib_cache.h | 29 -- sys/ofed/include/rdma/ib_sa.h | 13 + sys/ofed/include/rdma/ib_verbs.h | 44 +- sys/ofed/include/rdma/rdma_cm.h | 14 - 28 files changed, 696 insertions(+), 751 deletions(-) diff --git a/sys/dev/bnxt/bnxt_re/ib_verbs.c b/sys/dev/bnxt/bnxt_re/ib_verbs.c index eacd5c1abab5..516d0bb8cddf 100644 --- a/sys/dev/bnxt/bnxt_re/ib_verbs.c +++ b/sys/dev/bnxt/bnxt_re/ib_verbs.c @@ -443,12 +443,11 @@ int bnxt_re_query_gid(struct ib_device *ibdev, u8 port_num, return rc; } -int bnxt_re_del_gid(struct ib_device *ibdev, u8 port_num, - unsigned int index, void **context) +int bnxt_re_del_gid(const struct ib_gid_attr *attr, void **context) { int rc = 0; struct bnxt_re_gid_ctx *ctx, **ctx_tbl; - struct bnxt_re_dev *rdev = to_bnxt_re_dev(ibdev, ibdev); + struct bnxt_re_dev *rdev = to_bnxt_re_dev(attr->device, ibdev); struct bnxt_qplib_sgid_tbl *sgid_tbl = &rdev->qplib_res.sgid_tbl; struct bnxt_qplib_gid *gid_to_del; u16 vlan_id = 0xFFFF; @@ -482,14 +481,14 @@ int bnxt_re_del_gid(struct ib_device *ibdev, u8 port_num, dev_dbg(rdev_to_dev(rdev), "Trying to delete GID0 while QP1 is alive\n"); if (!ctx->refcnt) { - rdev->gid_map[index] = -1; + rdev->gid_map[attr->index] = -1; ctx_tbl = sgid_tbl->ctx; ctx_tbl[ctx->idx] = NULL; kfree(ctx); } return 0; } - rdev->gid_map[index] = -1; + rdev->gid_map[attr->index] = -1; if (!ctx->refcnt) { rc = bnxt_qplib_del_sgid(sgid_tbl, gid_to_del, vlan_id, true); @@ -510,15 +509,14 @@ int bnxt_re_del_gid(struct ib_device *ibdev, u8 port_num, return rc; } -int bnxt_re_add_gid(struct ib_device *ibdev, u8 port_num, - unsigned int index, const union ib_gid *gid, +int bnxt_re_add_gid(const union ib_gid *gid, const struct ib_gid_attr *attr, void **context) { int rc; u32 tbl_idx = 0; u16 vlan_id = 0xFFFF; struct bnxt_re_gid_ctx *ctx, **ctx_tbl; - struct bnxt_re_dev *rdev = to_bnxt_re_dev(ibdev, ibdev); + struct bnxt_re_dev *rdev = to_bnxt_re_dev(attr->device, ibdev); struct bnxt_qplib_sgid_tbl *sgid_tbl = &rdev->qplib_res.sgid_tbl; if ((attr->ndev) && is_vlan_dev(attr->ndev)) vlan_id = vlan_dev_vlan_id(attr->ndev); @@ -540,8 +538,8 @@ int bnxt_re_add_gid(struct ib_device *ibdev, u8 port_num, ctx_tbl[tbl_idx]->refcnt++; } *context = ctx_tbl[tbl_idx]; - /* tbl_idx is the HW table index and index is the stack index */ - rdev->gid_map[index] = tbl_idx; + /* tbl_idx is the HW table index and attr->index is the stack index */ + rdev->gid_map[attr->index] = tbl_idx; return 0; } else if (rc < 0) { dev_err(rdev_to_dev(rdev), "Add GID failed rc = 0x%x\n", rc); @@ -556,8 +554,8 @@ int bnxt_re_add_gid(struct ib_device *ibdev, u8 port_num, ctx->idx = tbl_idx; ctx->refcnt = 1; ctx_tbl[tbl_idx] = ctx; - /* tbl_idx is the HW table index and index is the stack index */ - rdev->gid_map[index] = tbl_idx; + /* tbl_idx is the HW table index and attr->index is the stack index */ + rdev->gid_map[attr->index] = tbl_idx; *context = ctx; } return rc; diff --git a/sys/dev/bnxt/bnxt_re/ib_verbs.h b/sys/dev/bnxt/bnxt_re/ib_verbs.h index 332c88e07a3d..4158aabb8fef 100644 --- a/sys/dev/bnxt/bnxt_re/ib_verbs.h +++ b/sys/dev/bnxt/bnxt_re/ib_verbs.h @@ -295,10 +295,8 @@ int bnxt_re_get_port_immutable(struct ib_device *ibdev, u8 port_num, void bnxt_re_compat_qfwstr(void); int bnxt_re_query_pkey(struct ib_device *ibdev, u8 port_num, u16 index, u16 *pkey); -int bnxt_re_del_gid(struct ib_device *ibdev, u8 port_num, - unsigned int index, void **context); -int bnxt_re_add_gid(struct ib_device *ibdev, u8 port_num, - unsigned int index, const union ib_gid *gid, +int bnxt_re_del_gid(const struct ib_gid_attr *attr, void **context); +int bnxt_re_add_gid(const union ib_gid *gid, const struct ib_gid_attr *attr, void **context); int bnxt_re_query_gid(struct ib_device *ibdev, u8 port_num, int index, union ib_gid *gid); diff --git a/sys/dev/bnxt/bnxt_re/main.c b/sys/dev/bnxt/bnxt_re/main.c index 991931bdd396..e6ec286006fa 100644 --- a/sys/dev/bnxt/bnxt_re/main.c +++ b/sys/dev/bnxt/bnxt_re/main.c @@ -2179,7 +2179,6 @@ static int bnxt_re_register_ib(struct bnxt_re_dev *rdev) ibdev->modify_port = bnxt_re_modify_port; ibdev->get_port_immutable = bnxt_re_get_port_immutable; ibdev->query_pkey = bnxt_re_query_pkey; - ibdev->query_gid = bnxt_re_query_gid; ibdev->get_netdev = bnxt_re_get_netdev; ibdev->add_gid = bnxt_re_add_gid; ibdev->del_gid = bnxt_re_del_gid; diff --git a/sys/dev/irdma/fbsd_kcompat.h b/sys/dev/irdma/fbsd_kcompat.h index 8dcbbd8e5144..8f9d7f86e481 100644 --- a/sys/dev/irdma/fbsd_kcompat.h +++ b/sys/dev/irdma/fbsd_kcompat.h @@ -156,8 +156,6 @@ int irdma_iw_port_immutable(struct ib_device *ibdev, u8 port_num, struct ib_port_immutable *immutable); int irdma_query_gid(struct ib_device *ibdev, u8 port, int index, union ib_gid *gid); -int irdma_query_gid_roce(struct ib_device *ibdev, u8 port, int index, - union ib_gid *gid); int irdma_query_pkey(struct ib_device *ibdev, u8 port, u16 index, u16 *pkey); int irdma_query_port(struct ib_device *ibdev, u8 port, @@ -255,9 +253,9 @@ int irdma_alloc_ucontext(struct ib_ucontext *uctx, struct ib_udata *udata); void irdma_dealloc_ucontext(struct ib_ucontext *context); int irdma_alloc_pd(struct ib_pd *pd, struct ib_udata *udata); void irdma_dealloc_pd(struct ib_pd *ibpd, struct ib_udata *udata); -int irdma_add_gid(struct ib_device *, u8, unsigned int, const union ib_gid *, +int irdma_add_gid(const union ib_gid *, const struct ib_gid_attr *, void **); -int irdma_del_gid(struct ib_device *, u8, unsigned int, void **); +int irdma_del_gid(const struct ib_gid_attr *, void **); struct ib_device *ib_device_get_by_netdev(struct ifnet *ndev, int driver_id); void ib_device_put(struct ib_device *device); void ib_unregister_device_put(struct ib_device *device); diff --git a/sys/dev/irdma/irdma_kcompat.c b/sys/dev/irdma/irdma_kcompat.c index a320d614d127..4472c61dd0e0 100644 --- a/sys/dev/irdma/irdma_kcompat.c +++ b/sys/dev/irdma/irdma_kcompat.c @@ -77,10 +77,7 @@ irdma_get_dev_fw_str(struct ib_device *dev, } int -irdma_add_gid(struct ib_device *device, - u8 port_num, - unsigned int index, - const union ib_gid *gid, +irdma_add_gid(const union ib_gid *gid, const struct ib_gid_attr *attr, void **context) { @@ -88,9 +85,7 @@ irdma_add_gid(struct ib_device *device, } int -irdma_del_gid(struct ib_device *device, - u8 port_num, - unsigned int index, +irdma_del_gid(const struct ib_gid_attr *attr, void **context) { return 0; @@ -1516,28 +1511,6 @@ ib_unregister_device_put(struct ib_device *device) ib_unregister_device(device); } -/** - * irdma_query_gid_roce - Query port GID for Roce - * @ibdev: device pointer from stack - * @port: port number - * @index: Entry index - * @gid: Global ID - */ -int -irdma_query_gid_roce(struct ib_device *ibdev, u8 port, int index, - union ib_gid *gid) -{ - int ret; - - ret = ib_get_cached_gid(ibdev, port, index, gid, NULL); - if (ret == -EAGAIN) { - memcpy(gid, &zgid, sizeof(*gid)); - return 0; - } - - return ret; -} - /** * irdma_modify_port - modify port attributes * @ibdev: device pointer from stack diff --git a/sys/dev/irdma/irdma_verbs.c b/sys/dev/irdma/irdma_verbs.c index eea241873522..6eba47e696c1 100644 --- a/sys/dev/irdma/irdma_verbs.c +++ b/sys/dev/irdma/irdma_verbs.c @@ -3328,7 +3328,6 @@ irdma_set_device_roce_ops(struct ib_device *ibdev) dev_ops->get_port_immutable = irdma_roce_port_immutable; dev_ops->modify_qp = irdma_modify_qp_roce; dev_ops->query_ah = irdma_query_ah; - dev_ops->query_gid = irdma_query_gid_roce; dev_ops->query_pkey = irdma_query_pkey; ibdev->add_gid = irdma_add_gid; ibdev->del_gid = irdma_del_gid; diff --git a/sys/dev/mlx4/mlx4_ib/mlx4_ib_main.c b/sys/dev/mlx4/mlx4_ib/mlx4_ib_main.c index ae82f8c1a187..9f51bac12d51 100644 --- a/sys/dev/mlx4/mlx4_ib/mlx4_ib_main.c +++ b/sys/dev/mlx4/mlx4_ib/mlx4_ib_main.c @@ -252,14 +252,11 @@ static int mlx4_ib_update_gids(struct gid_entry *gids, return mlx4_ib_update_gids_v1(gids, ibdev, port_num); } -static int mlx4_ib_add_gid(struct ib_device *device, - u8 port_num, - unsigned int index, - const union ib_gid *gid, +static int mlx4_ib_add_gid(const union ib_gid *gid, const struct ib_gid_attr *attr, void **context) { - struct mlx4_ib_dev *ibdev = to_mdev(device); + struct mlx4_ib_dev *ibdev = to_mdev(attr->device); struct mlx4_ib_iboe *iboe = &ibdev->iboe; struct mlx4_port_gid_table *port_gid_table; int free = -1, found = -1; @@ -268,16 +265,16 @@ static int mlx4_ib_add_gid(struct ib_device *device, int i; struct gid_entry *gids = NULL; - if (!rdma_cap_roce_gid_table(device, port_num)) + if (!rdma_cap_roce_gid_table(attr->device, attr->port_num)) return -EINVAL; - if (port_num > MLX4_MAX_PORTS) + if (attr->port_num > MLX4_MAX_PORTS) return -EINVAL; if (!context) return -EINVAL; - port_gid_table = &iboe->gids[port_num - 1]; + port_gid_table = &iboe->gids[attr->port_num - 1]; spin_lock_bh(&iboe->lock); for (i = 0; i < MLX4_MAX_PORT_GIDS; ++i) { if (!memcmp(&port_gid_table->gids[i].gid, gid, sizeof(*gid)) && @@ -324,33 +321,30 @@ static int mlx4_ib_add_gid(struct ib_device *device, spin_unlock_bh(&iboe->lock); if (!ret && hw_update) { - ret = mlx4_ib_update_gids(gids, ibdev, port_num); + ret = mlx4_ib_update_gids(gids, ibdev, attr->port_num); kfree(gids); } return ret; } -static int mlx4_ib_del_gid(struct ib_device *device, - u8 port_num, - unsigned int index, - void **context) +static int mlx4_ib_del_gid(const struct ib_gid_attr *attr, void **context) { struct gid_cache_context *ctx = *context; - struct mlx4_ib_dev *ibdev = to_mdev(device); + struct mlx4_ib_dev *ibdev = to_mdev(attr->device); struct mlx4_ib_iboe *iboe = &ibdev->iboe; struct mlx4_port_gid_table *port_gid_table; int ret = 0; int hw_update = 0; struct gid_entry *gids = NULL; - if (!rdma_cap_roce_gid_table(device, port_num)) + if (!rdma_cap_roce_gid_table(attr->device, attr->port_num)) return -EINVAL; - if (port_num > MLX4_MAX_PORTS) + if (attr->port_num > MLX4_MAX_PORTS) return -EINVAL; - port_gid_table = &iboe->gids[port_num - 1]; + port_gid_table = &iboe->gids[attr->port_num - 1]; spin_lock_bh(&iboe->lock); if (ctx) { ctx->refcount--; @@ -382,7 +376,7 @@ static int mlx4_ib_del_gid(struct ib_device *device, spin_unlock_bh(&iboe->lock); if (!ret && hw_update) { - ret = mlx4_ib_update_gids(gids, ibdev, port_num); + ret = mlx4_ib_update_gids(gids, ibdev, attr->port_num); kfree(gids); } return ret; @@ -417,9 +411,6 @@ int mlx4_ib_gid_index_to_real_index(struct mlx4_ib_dev *ibdev, if (attr.ndev) if_rele(attr.ndev); - if (!memcmp(&gid, &zgid, sizeof(gid))) - return -EINVAL; - spin_lock_irqsave(&iboe->lock, flags); port_gid_table = &iboe->gids[port_num - 1]; @@ -825,24 +816,10 @@ out: static int mlx4_ib_query_gid(struct ib_device *ibdev, u8 port, int index, union ib_gid *gid) { - int ret; - if (rdma_protocol_ib(ibdev, port)) return __mlx4_ib_query_gid(ibdev, port, index, gid, 0); - if (!rdma_protocol_roce(ibdev, port)) - return -ENODEV; - - if (!rdma_cap_roce_gid_table(ibdev, port)) - return -ENODEV; - - ret = ib_get_cached_gid(ibdev, port, index, gid, NULL); - if (ret == -EAGAIN) { - memcpy(gid, &zgid, sizeof(*gid)); - return 0; - } - - return ret; + return 0; } static int mlx4_ib_query_sl2vl(struct ib_device *ibdev, u8 port, u64 *sl2vl_tbl) diff --git a/sys/dev/mlx4/mlx4_ib/mlx4_ib_qp.c b/sys/dev/mlx4/mlx4_ib/mlx4_ib_qp.c index c4ab30c8bcb6..42606907a2e3 100644 --- a/sys/dev/mlx4/mlx4_ib/mlx4_ib_qp.c +++ b/sys/dev/mlx4/mlx4_ib/mlx4_ib_qp.c @@ -1787,8 +1787,6 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp, status = ib_get_cached_gid(ibqp->device, port_num, index, &gid, &gid_attr); - if (!status && !memcmp(&gid, &zgid, sizeof(gid))) - status = -ENOENT; if (!status && gid_attr.ndev) { vlan = rdma_vlan_dev_vlan_id(gid_attr.ndev); memcpy(smac, if_getlladdr(gid_attr.ndev), ETH_ALEN); diff --git a/sys/dev/mlx5/mlx5_ib/mlx5_ib_main.c b/sys/dev/mlx5/mlx5_ib/mlx5_ib_main.c index cf1e5f4324d8..a1159f91829d 100644 --- a/sys/dev/mlx5/mlx5_ib/mlx5_ib_main.c +++ b/sys/dev/mlx5/mlx5_ib/mlx5_ib_main.c @@ -428,18 +428,19 @@ static int set_roce_addr(struct ib_device *device, u8 port_num, return mlx5_cmd_exec(dev->mdev, in, sizeof(in), out, sizeof(out)); } -static int mlx5_ib_add_gid(struct ib_device *device, u8 port_num, - unsigned int index, const union ib_gid *gid, +static int mlx5_ib_add_gid(const union ib_gid *gid, const struct ib_gid_attr *attr, __always_unused void **context) { - return set_roce_addr(device, port_num, index, gid, attr); + return set_roce_addr(attr->device, attr->port_num, attr->index, gid, + attr); } -static int mlx5_ib_del_gid(struct ib_device *device, u8 port_num, - unsigned int index, __always_unused void **context) +static int mlx5_ib_del_gid(const struct ib_gid_attr *attr, + __always_unused void **context) { - return set_roce_addr(device, port_num, index, NULL, NULL); + return set_roce_addr(attr->device, attr->port_num, attr->index, NULL, + NULL); } __be16 mlx5_get_roce_udp_sport(struct mlx5_ib_dev *dev, u8 port_num, diff --git a/sys/dev/qlnx/qlnxr/qlnxr_os.c b/sys/dev/qlnx/qlnxr/qlnxr_os.c index fd32080af599..c2f87f91e450 100644 --- a/sys/dev/qlnx/qlnxr/qlnxr_os.c +++ b/sys/dev/qlnx/qlnxr/qlnxr_os.c @@ -197,7 +197,6 @@ qlnxr_register_device(qlnxr_dev_t *dev) ibdev->query_gid = qlnxr_iw_query_gid; } else { ibdev->node_type = RDMA_NODE_IB_CA; - ibdev->query_gid = qlnxr_query_gid; ibdev->uverbs_cmd_mask |= (1ull << IB_USER_VERBS_CMD_CREATE_SRQ) | (1ull << IB_USER_VERBS_CMD_DESTROY_SRQ) | diff --git a/sys/dev/qlnx/qlnxr/qlnxr_verbs.c b/sys/dev/qlnx/qlnxr/qlnxr_verbs.c index d27052fa56bf..6f640d2b274d 100644 --- a/sys/dev/qlnx/qlnxr/qlnxr_verbs.c +++ b/sys/dev/qlnx/qlnxr/qlnxr_verbs.c @@ -118,44 +118,6 @@ qlnxr_iw_query_gid(struct ib_device *ibdev, u8 port, int index, return 0; } -int -qlnxr_query_gid(struct ib_device *ibdev, u8 port, int index, - union ib_gid *sgid) -{ - struct qlnxr_dev *dev; - qlnx_host_t *ha; - - dev = get_qlnxr_dev(ibdev); - ha = dev->ha; - QL_DPRINT12(ha, "enter index: %d\n", index); -#if 0 - int ret = 0; - /* @@@: if DEFINE_ROCE_GID_TABLE to be used here */ - //if (!rdma_cap_roce_gid_table(ibdev, port)) { - if (!(rdma_protocol_roce(ibdev, port) && - ibdev->add_gid && ibdev->del_gid)) { - QL_DPRINT11(ha, "acquire gid failed\n"); - return -ENODEV; - } - - ret = ib_get_cached_gid(ibdev, port, index, sgid, NULL); - if (ret == -EAGAIN) { - memcpy(sgid, &zgid, sizeof(*sgid)); - return 0; - } -#endif - if ((index >= QLNXR_MAX_SGID) || (index < 0)) { - QL_DPRINT12(ha, "invalid gid index %d\n", index); - memset(sgid, 0, sizeof(*sgid)); - return -EINVAL; - } - memcpy(sgid, &dev->sgid_tbl[index], sizeof(*sgid)); - - QL_DPRINT12(ha, "exit : %p\n", sgid); - - return 0; -} - int qlnxr_create_srq(struct ib_srq *ibsrq, struct ib_srq_init_attr *init_attr, diff --git a/sys/dev/qlnx/qlnxr/qlnxr_verbs.h b/sys/dev/qlnx/qlnxr/qlnxr_verbs.h index dbb5e8602952..7be308ad6d9d 100644 --- a/sys/dev/qlnx/qlnxr/qlnxr_verbs.h +++ b/sys/dev/qlnx/qlnxr/qlnxr_verbs.h @@ -34,11 +34,6 @@ extern int qlnxr_iw_query_gid(struct ib_device *, int index, union ib_gid *gid); -extern int qlnxr_query_gid(struct ib_device *, - u8 port, - int index, - union ib_gid *gid); - extern int qlnxr_create_srq(struct ib_srq *ibsrq, struct ib_srq_init_attr *, struct ib_udata *); diff --git a/sys/ofed/drivers/infiniband/core/cma_priv.h b/sys/ofed/drivers/infiniband/core/cma_priv.h new file mode 100644 index 000000000000..ba054364d205 --- /dev/null +++ b/sys/ofed/drivers/infiniband/core/cma_priv.h @@ -0,0 +1,95 @@ +/* + * Copyright (c) 2005 Voltaire Inc. All rights reserved. + * Copyright (c) 2002-2005, Network Appliance, Inc. All rights reserved. + * Copyright (c) 1999-2005, Mellanox Technologies, Inc. All rights reserved. + * Copyright (c) 2005-2006 Intel Corporation. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#ifndef _CMA_PRIV_H +#define _CMA_PRIV_H + +enum rdma_cm_state { + RDMA_CM_IDLE, + RDMA_CM_ADDR_QUERY, + RDMA_CM_ADDR_RESOLVED, + RDMA_CM_ROUTE_QUERY, + RDMA_CM_ROUTE_RESOLVED, + RDMA_CM_CONNECT, + RDMA_CM_DISCONNECT, + RDMA_CM_ADDR_BOUND, + RDMA_CM_LISTEN, + RDMA_CM_DEVICE_REMOVAL, + RDMA_CM_DESTROYING +}; + +struct rdma_id_private { + struct rdma_cm_id id; + + struct rdma_bind_list *bind_list; + struct hlist_node node; + struct list_head list; /* listen_any_list or cma_device.list */ + struct list_head listen_list; /* per device listens */ + struct cma_device *cma_dev; + struct list_head mc_list; + + int internal_id; + enum rdma_cm_state state; + spinlock_t lock; + struct mutex qp_mutex; + + struct completion comp; + atomic_t refcount; + struct mutex handler_mutex; + + int backlog; + int timeout_ms; + struct ib_sa_query *query; + int query_id; + union { + struct ib_cm_id *ib; + struct iw_cm_id *iw; + } cm_id; + + u32 seq_num; + u32 qkey; + u32 qp_num; + pid_t owner; + u32 options; + u8 srq; + u8 tos; + u8 tos_set:1; + u8 timeout_set:1; + u8 reuseaddr; + u8 afonly; + u8 timeout; + enum ib_gid_type gid_type; +}; +#endif /* _CMA_PRIV_H */ diff --git a/sys/ofed/drivers/infiniband/core/core_priv.h b/sys/ofed/drivers/infiniband/core/core_priv.h index 508119e9259d..3a61defb540b 100644 --- a/sys/ofed/drivers/infiniband/core/core_priv.h +++ b/sys/ofed/drivers/infiniband/core/core_priv.h @@ -183,6 +183,16 @@ static inline struct ib_qp *_ib_create_qp(struct ib_device *dev, return qp; } +struct rdma_dev_addr; +int rdma_resolve_ip_route(struct sockaddr *src_addr, + const struct sockaddr *dst_addr, + struct rdma_dev_addr *addr); + +int rdma_addr_find_l2_eth_by_grh(const union ib_gid *sgid, + const union ib_gid *dgid, + u8 *dmac, if_t ndev, + int *hoplimit); + struct rdma_umap_priv { struct vm_area_struct *vma; struct list_head list; diff --git a/sys/ofed/drivers/infiniband/core/ib_addr.c b/sys/ofed/drivers/infiniband/core/ib_addr.c index 2a00e449ecda..893fd0bb2b37 100644 --- a/sys/ofed/drivers/infiniband/core/ib_addr.c +++ b/sys/ofed/drivers/infiniband/core/ib_addr.c @@ -61,7 +61,6 @@ struct addr_req { struct sockaddr_storage src_addr; struct sockaddr_storage dst_addr; struct rdma_dev_addr *addr; - struct rdma_addr_client *client; void *context; void (*callback)(int status, struct sockaddr *src_addr, struct rdma_dev_addr *addr, void *context); @@ -70,11 +69,8 @@ struct addr_req { int status; }; -static void process_req(struct work_struct *work); - -static DEFINE_MUTEX(lock); +static DEFINE_SPINLOCK(lock); static LIST_HEAD(req_list); -static DECLARE_DELAYED_WORK(work, process_req); static struct workqueue_struct *addr_wq; int rdma_addr_size(struct sockaddr *addr) @@ -108,28 +104,6 @@ int rdma_addr_size_kss(struct sockaddr_storage *addr) } EXPORT_SYMBOL(rdma_addr_size_kss); -static struct rdma_addr_client self; - -void rdma_addr_register_client(struct rdma_addr_client *client) -{ - atomic_set(&client->refcount, 1); - init_completion(&client->comp); -} -EXPORT_SYMBOL(rdma_addr_register_client); - -static inline void put_client(struct rdma_addr_client *client) -{ - if (atomic_dec_and_test(&client->refcount)) - complete(&client->comp); -} - -void rdma_addr_unregister_client(struct rdma_addr_client *client) -{ - put_client(client); - wait_for_completion(&client->comp); -} -EXPORT_SYMBOL(rdma_addr_unregister_client); - static inline void rdma_copy_addr_sub(u8 *dst, const u8 *src, unsigned min, unsigned max) { @@ -211,7 +185,7 @@ int rdma_translate_ip(const struct sockaddr *addr, } EXPORT_SYMBOL(rdma_translate_ip); -static void set_timeout(struct delayed_work *delayed_work, unsigned long time) +static void set_timeout(struct addr_req *req, unsigned long time) { unsigned long delay; @@ -221,23 +195,15 @@ static void set_timeout(struct delayed_work *delayed_work, unsigned long time) else if (delay > hz) delay = hz; - mod_delayed_work(addr_wq, delayed_work, delay); + mod_delayed_work(addr_wq, &req->work, delay); } static void queue_req(struct addr_req *req) { - struct addr_req *temp_req; - - mutex_lock(&lock); - list_for_each_entry_reverse(temp_req, &req_list, list) { - if (time_after_eq(req->timeout, temp_req->timeout)) - break; - } - - list_add(&req->list, &temp_req->list); - - set_timeout(&req->work, req->timeout); - mutex_unlock(&lock); + spin_lock_bh(&lock); + list_add_tail(&req->list, &req_list); + set_timeout(req, req->timeout); + spin_unlock_bh(&lock); } #if defined(INET) || defined(INET6) @@ -714,7 +680,6 @@ static void process_one_req(struct work_struct *_work) struct addr_req *req; struct sockaddr *src_in, *dst_in; - mutex_lock(&lock); req = container_of(_work, struct addr_req, work.work); if (req->status == -ENODATA) { @@ -725,62 +690,33 @@ static void process_one_req(struct work_struct *_work) req->status = -ETIMEDOUT; } else if (req->status == -ENODATA) { /* requeue the work for retrying again */ - set_timeout(&req->work, req->timeout); - mutex_unlock(&lock); + spin_lock_bh(&lock); + if (!list_empty(&req->list)) + set_timeout(req, req->timeout); + spin_unlock_bh(&lock); return; } } - list_del(&req->list); - mutex_unlock(&lock); req->callback(req->status, (struct sockaddr *)&req->src_addr, req->addr, req->context); - put_client(req->client); - kfree(req); -} - -static void process_req(struct work_struct *work) -{ - struct addr_req *req, *temp_req; - struct sockaddr *src_in, *dst_in; - struct list_head done_list; - - INIT_LIST_HEAD(&done_list); - - mutex_lock(&lock); - list_for_each_entry_safe(req, temp_req, &req_list, list) { - if (req->status == -ENODATA) { - src_in = (struct sockaddr *) &req->src_addr; - dst_in = (struct sockaddr *) &req->dst_addr; - req->status = addr_resolve(src_in, dst_in, req->addr); - if (req->status && time_after_eq(jiffies, req->timeout)) - req->status = -ETIMEDOUT; - else if (req->status == -ENODATA) { - set_timeout(&req->work, req->timeout); - continue; - } - } - list_move_tail(&req->list, &done_list); - } + req->callback = NULL; - mutex_unlock(&lock); - - list_for_each_entry_safe(req, temp_req, &done_list, list) { - list_del(&req->list); - /* It is safe to cancel other work items from this work item - * because at a time there can be only one work item running - * with this single threaded work queue. + spin_lock_bh(&lock); + if (!list_empty(&req->list)) { + /* + * Although the work will normally have been canceled by the + * workqueue, it can still be requeued as long as it is on the + * req_list. */ cancel_delayed_work(&req->work); - req->callback(req->status, (struct sockaddr *) &req->src_addr, - req->addr, req->context); - put_client(req->client); + list_del_init(&req->list); kfree(req); } + spin_unlock_bh(&lock); } -int rdma_resolve_ip(struct rdma_addr_client *client, - struct sockaddr *src_addr, struct sockaddr *dst_addr, +int rdma_resolve_ip(struct sockaddr *src_addr, struct sockaddr *dst_addr, struct rdma_dev_addr *addr, int timeout_ms, void (*callback)(int status, struct sockaddr *src_addr, struct rdma_dev_addr *addr, void *context), @@ -812,8 +748,6 @@ int rdma_resolve_ip(struct rdma_addr_client *client, req->addr = addr; req->callback = callback; req->context = context; - req->client = client; - atomic_inc(&client->refcount); INIT_DELAYED_WORK(&req->work, process_one_req); req->status = addr_resolve(src_in, dst_in, addr); @@ -828,7 +762,6 @@ int rdma_resolve_ip(struct rdma_addr_client *client, break; default: ret = req->status; - atomic_dec(&client->refcount); goto err; } return ret; @@ -856,23 +789,40 @@ int rdma_resolve_ip_route(struct sockaddr *src_addr, return addr_resolve(src_in, dst_addr, addr); } -EXPORT_SYMBOL(rdma_resolve_ip_route); void rdma_addr_cancel(struct rdma_dev_addr *addr) { struct addr_req *req, *temp_req; + struct addr_req *found = NULL; - mutex_lock(&lock); + spin_lock_bh(&lock); list_for_each_entry_safe(req, temp_req, &req_list, list) { if (req->addr == addr) { - req->status = -ECANCELED; - req->timeout = jiffies; - list_move(&req->list, &req_list); - set_timeout(&req->work, req->timeout); + /* + * Removing from the list means we take ownership of + * the req + */ + list_del_init(&req->list); + found = req; break; } } - mutex_unlock(&lock); + spin_unlock_bh(&lock); + + if (!found) + return; + + /* + * sync canceling the work after removing it from the req_list + * guarentees no work is running and none will be started. + */ + cancel_delayed_work_sync(&found->work); + + if (found->callback) + found->callback(-ECANCELED, (struct sockaddr *)&found->src_addr, + found->addr, found->context); + + kfree(found); } EXPORT_SYMBOL(rdma_addr_cancel); @@ -907,8 +857,8 @@ int rdma_addr_find_l2_eth_by_grh(const union ib_gid *sgid, dev_addr.net = dev_net(ndev); init_completion(&ctx.comp); - ret = rdma_resolve_ip(&self, &sgid_addr._sockaddr, &dgid_addr._sockaddr, - &dev_addr, 1000, resolve_cb, &ctx); + ret = rdma_resolve_ip(&sgid_addr._sockaddr, &dgid_addr._sockaddr, + &dev_addr, 1000, resolve_cb, &ctx); if (ret) return ret; @@ -929,13 +879,11 @@ int addr_init(void) if (!addr_wq) return -ENOMEM; - rdma_addr_register_client(&self); - return 0; } void addr_cleanup(void) { - rdma_addr_unregister_client(&self); destroy_workqueue(addr_wq); + WARN_ON(!list_empty(&req_list)); } diff --git a/sys/ofed/drivers/infiniband/core/ib_cache.c b/sys/ofed/drivers/infiniband/core/ib_cache.c index 878b2915f25d..1b7f0dfd4761 100644 --- a/sys/ofed/drivers/infiniband/core/ib_cache.c +++ b/sys/ofed/drivers/infiniband/core/ib_cache.c @@ -62,8 +62,6 @@ struct ib_update_work { union ib_gid zgid; EXPORT_SYMBOL(zgid); -static const struct ib_gid_attr zattr; - enum gid_attr_find_mask { GID_ATTR_FIND_MASK_GID = 1UL << 0, GID_ATTR_FIND_MASK_NETDEV = 1UL << 1, @@ -76,15 +74,6 @@ enum gid_table_entry_props { GID_TABLE_ENTRY_DEFAULT = 1UL << 1, }; -enum gid_table_write_action { - GID_TABLE_WRITE_ACTION_ADD, - GID_TABLE_WRITE_ACTION_DEL, - /* MODIFY only updates the GID table. Currently only used by - * ib_cache_update. - */ - GID_TABLE_WRITE_ACTION_MODIFY -}; - struct ib_gid_table_entry { unsigned long props; union ib_gid gid; @@ -103,31 +92,26 @@ struct ib_gid_table { * (a) Find the GID * (b) Delete it. * - * Add/delete should be carried out atomically. - * This is done by locking this mutex from multiple - * writers. We don't need this lock for IB, as the MAD - * layer replaces all entries. All data_vec entries - * are locked by this lock. **/ - struct mutex lock; - /* This lock protects the table entries from being - * read and written simultaneously. + /* Any writer to data_vec must hold this lock and the write side of + * rwlock. readers must hold only rwlock. All writers must be in a + * sleepable context. */ + struct mutex lock; + /* rwlock protects data_vec[ix]->props. */ rwlock_t rwlock; struct ib_gid_table_entry *data_vec; }; static void dispatch_gid_change_event(struct ib_device *ib_dev, u8 port) { - if (rdma_cap_roce_gid_table(ib_dev, port)) { - struct ib_event event; + struct ib_event event; - event.device = ib_dev; - event.element.port_num = port; - event.event = IB_EVENT_GID_CHANGE; + event.device = ib_dev; + event.element.port_num = port; + event.event = IB_EVENT_GID_CHANGE; - ib_dispatch_event(&event); - } + ib_dispatch_event(&event); } static const char * const gid_type_str[] = { @@ -168,94 +152,127 @@ int ib_cache_gid_parse_type_str(const char *buf) *** 1609 LINES SKIPPED ***
