From a372173bf314d374da4dd1155549d8ca7fc44709 Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Thu, 14 Jan 2021 21:14:23 +0200 Subject: RDMA/cxgb4: Fix the reported max_recv_sge value The max_recv_sge value is wrongly reported when calling query_qp, This is happening due to a typo when assigning the max_recv_sge value, the value of sq_max_sges was assigned instead of rq_max_sges. Fixes: 3e5c02c9ef9a ("iw_cxgb4: Support query_qp() verb") Link: https://lore.kernel.org/r/20210114191423.423529-1-kamalheib1@gmail.com Signed-off-by: Kamal Heib Reviewed-by: Potnuri Bharat Teja Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/cxgb4/qp.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers') diff --git a/drivers/infiniband/hw/cxgb4/qp.c b/drivers/infiniband/hw/cxgb4/qp.c index a7401398cb34..d109bb3822a5 100644 --- a/drivers/infiniband/hw/cxgb4/qp.c +++ b/drivers/infiniband/hw/cxgb4/qp.c @@ -2474,7 +2474,7 @@ int c4iw_ib_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, init_attr->cap.max_send_wr = qhp->attr.sq_num_entries; init_attr->cap.max_recv_wr = qhp->attr.rq_num_entries; init_attr->cap.max_send_sge = qhp->attr.sq_max_sges; - init_attr->cap.max_recv_sge = qhp->attr.sq_max_sges; + init_attr->cap.max_recv_sge = qhp->attr.rq_max_sges; init_attr->cap.max_inline_data = T4_MAX_SEND_INLINE; init_attr->sq_sig_type = qhp->sq_sig_all ? IB_SIGNAL_ALL_WR : 0; return 0; -- cgit v1.2.3 From 9f206f7398f6f6ec7dd0198c045c2459b4f720b6 Mon Sep 17 00:00:00 2001 From: Bryan Tan Date: Mon, 18 Jan 2021 19:16:29 -0800 Subject: RDMA/vmw_pvrdma: Fix network_hdr_type reported in WC The PVRDMA device HW interface defines network_hdr_type according to an old definition of the internal kernel rdma_network_type enum that has since changed, resulting in the wrong rdma_network_type being reported. Fix this by explicitly defining the enum used by the PVRDMA device and adding a function to convert the pvrdma_network_type to rdma_network_type enum. Cc: stable@vger.kernel.org # 5.10+ Fixes: 1c15b4f2a42f ("RDMA/core: Modify enum ib_gid_type and enum rdma_network_type") Link: https://lore.kernel.org/r/1611026189-17943-1-git-send-email-bryantan@vmware.com Reviewed-by: Adit Ranadive Signed-off-by: Bryan Tan Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/vmw_pvrdma/pvrdma.h | 14 ++++++++++++++ drivers/infiniband/hw/vmw_pvrdma/pvrdma_cq.c | 2 +- include/uapi/rdma/vmw_pvrdma-abi.h | 7 +++++++ 3 files changed, 22 insertions(+), 1 deletion(-) (limited to 'drivers') diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma.h b/drivers/infiniband/hw/vmw_pvrdma/pvrdma.h index c142f5e7f25f..de57f2fed743 100644 --- a/drivers/infiniband/hw/vmw_pvrdma/pvrdma.h +++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma.h @@ -509,6 +509,20 @@ static inline int ib_send_flags_to_pvrdma(int flags) return flags & PVRDMA_MASK(PVRDMA_SEND_FLAGS_MAX); } +static inline int pvrdma_network_type_to_ib(enum pvrdma_network_type type) +{ + switch (type) { + case PVRDMA_NETWORK_ROCE_V1: + return RDMA_NETWORK_ROCE_V1; + case PVRDMA_NETWORK_IPV4: + return RDMA_NETWORK_IPV4; + case PVRDMA_NETWORK_IPV6: + return RDMA_NETWORK_IPV6; + default: + return RDMA_NETWORK_IPV6; + } +} + void pvrdma_qp_cap_to_ib(struct ib_qp_cap *dst, const struct pvrdma_qp_cap *src); void ib_qp_cap_to_pvrdma(struct pvrdma_qp_cap *dst, diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_cq.c b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_cq.c index a119ac3e103c..6aa40bd2fd52 100644 --- a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_cq.c +++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_cq.c @@ -367,7 +367,7 @@ retry: wc->dlid_path_bits = cqe->dlid_path_bits; wc->port_num = cqe->port_num; wc->vendor_err = cqe->vendor_err; - wc->network_hdr_type = cqe->network_hdr_type; + wc->network_hdr_type = pvrdma_network_type_to_ib(cqe->network_hdr_type); /* Update shared ring state */ pvrdma_idx_ring_inc(&cq->ring_state->rx.cons_head, cq->ibcq.cqe); diff --git a/include/uapi/rdma/vmw_pvrdma-abi.h b/include/uapi/rdma/vmw_pvrdma-abi.h index f8b638c73371..901a4fd72c09 100644 --- a/include/uapi/rdma/vmw_pvrdma-abi.h +++ b/include/uapi/rdma/vmw_pvrdma-abi.h @@ -133,6 +133,13 @@ enum pvrdma_wc_flags { PVRDMA_WC_FLAGS_MAX = PVRDMA_WC_WITH_NETWORK_HDR_TYPE, }; +enum pvrdma_network_type { + PVRDMA_NETWORK_IB, + PVRDMA_NETWORK_ROCE_V1 = PVRDMA_NETWORK_IB, + PVRDMA_NETWORK_IPV4, + PVRDMA_NETWORK_IPV6 +}; + struct pvrdma_alloc_ucontext_resp { __u32 qp_tab_size; __u32 reserved; -- cgit v1.2.3 From 9293d3fcb70583f2c786f04ca788af026b7c4c5c Mon Sep 17 00:00:00 2001 From: Yangyang Li Date: Tue, 19 Jan 2021 17:28:33 +0800 Subject: RDMA/hns: Use mutex instead of spinlock for ida allocation GFP_KERNEL may cause ida_alloc_range() to sleep, but the spinlock covering this function is not allowed to sleep, so the spinlock needs to be changed to mutex. As there is a certain chance of memory allocation failure, GFP_ATOMIC is not suitable for QP allocation scenarios. Fixes: 71586dd20010 ("RDMA/hns: Create QP with selected QPN for bank load balance") Link: https://lore.kernel.org/r/1611048513-28663-1-git-send-email-liweihang@huawei.com Signed-off-by: Yangyang Li Signed-off-by: Weihang Li Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/hns/hns_roce_device.h | 2 +- drivers/infiniband/hw/hns/hns_roce_qp.c | 11 ++++++----- 2 files changed, 7 insertions(+), 6 deletions(-) (limited to 'drivers') diff --git a/drivers/infiniband/hw/hns/hns_roce_device.h b/drivers/infiniband/hw/hns/hns_roce_device.h index 55d538625e36..ad8253245a85 100644 --- a/drivers/infiniband/hw/hns/hns_roce_device.h +++ b/drivers/infiniband/hw/hns/hns_roce_device.h @@ -532,7 +532,7 @@ struct hns_roce_qp_table { struct hns_roce_hem_table sccc_table; struct mutex scc_mutex; struct hns_roce_bank bank[HNS_ROCE_QP_BANK_NUM]; - spinlock_t bank_lock; + struct mutex bank_mutex; }; struct hns_roce_cq_table { diff --git a/drivers/infiniband/hw/hns/hns_roce_qp.c b/drivers/infiniband/hw/hns/hns_roce_qp.c index d8e2fe5558d2..1116371adf74 100644 --- a/drivers/infiniband/hw/hns/hns_roce_qp.c +++ b/drivers/infiniband/hw/hns/hns_roce_qp.c @@ -209,7 +209,7 @@ static int alloc_qpn(struct hns_roce_dev *hr_dev, struct hns_roce_qp *hr_qp) hr_qp->doorbell_qpn = 1; } else { - spin_lock(&qp_table->bank_lock); + mutex_lock(&qp_table->bank_mutex); bankid = get_least_load_bankid_for_qp(qp_table->bank); ret = alloc_qpn_with_bankid(&qp_table->bank[bankid], bankid, @@ -217,12 +217,12 @@ static int alloc_qpn(struct hns_roce_dev *hr_dev, struct hns_roce_qp *hr_qp) if (ret) { ibdev_err(&hr_dev->ib_dev, "failed to alloc QPN, ret = %d\n", ret); - spin_unlock(&qp_table->bank_lock); + mutex_unlock(&qp_table->bank_mutex); return ret; } qp_table->bank[bankid].inuse++; - spin_unlock(&qp_table->bank_lock); + mutex_unlock(&qp_table->bank_mutex); hr_qp->doorbell_qpn = (u32)num; } @@ -408,9 +408,9 @@ static void free_qpn(struct hns_roce_dev *hr_dev, struct hns_roce_qp *hr_qp) ida_free(&hr_dev->qp_table.bank[bankid].ida, hr_qp->qpn >> 3); - spin_lock(&hr_dev->qp_table.bank_lock); + mutex_lock(&hr_dev->qp_table.bank_mutex); hr_dev->qp_table.bank[bankid].inuse--; - spin_unlock(&hr_dev->qp_table.bank_lock); + mutex_unlock(&hr_dev->qp_table.bank_mutex); } static int set_rq_size(struct hns_roce_dev *hr_dev, struct ib_qp_cap *cap, @@ -1371,6 +1371,7 @@ int hns_roce_init_qp_table(struct hns_roce_dev *hr_dev) unsigned int i; mutex_init(&qp_table->scc_mutex); + mutex_init(&qp_table->bank_mutex); xa_init(&hr_dev->qp_table_xa); reserved_from_bot = hr_dev->caps.reserved_qps; -- cgit v1.2.3 From de641d74fb00f5b32f054ee154e31fb037e0db88 Mon Sep 17 00:00:00 2001 From: Parav Pandit Date: Sun, 17 Jan 2021 11:26:33 +0200 Subject: Revert "RDMA/mlx5: Fix devlink deadlock on net namespace deletion" This reverts commit fbdd0049d98d44914fc57d4b91f867f4996c787b. Due to commit in fixes tag, netdevice events were received only in one net namespace of mlx5_core_dev. Due to this when netdevice events arrive in net namespace other than net namespace of mlx5_core_dev, they are missed. This results in empty GID table due to RDMA device being detached from its net device. Hence, revert back to receive netdevice events in all net namespaces to restore back RDMA functionality in non init_net net namespace. The deadlock will have to be addressed in another patch. Fixes: fbdd0049d98d ("RDMA/mlx5: Fix devlink deadlock on net namespace deletion") Link: https://lore.kernel.org/r/20210117092633.10690-1-leon@kernel.org Signed-off-by: Parav Pandit Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/main.c | 6 ++---- drivers/net/ethernet/mellanox/mlx5/core/lib/mlx5.h | 5 +++++ include/linux/mlx5/driver.h | 18 ------------------ 3 files changed, 7 insertions(+), 22 deletions(-) (limited to 'drivers') diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index d26f3f3e0462..aabdc07e4753 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -3311,8 +3311,7 @@ static int mlx5_add_netdev_notifier(struct mlx5_ib_dev *dev, u8 port_num) int err; dev->port[port_num].roce.nb.notifier_call = mlx5_netdev_event; - err = register_netdevice_notifier_net(mlx5_core_net(dev->mdev), - &dev->port[port_num].roce.nb); + err = register_netdevice_notifier(&dev->port[port_num].roce.nb); if (err) { dev->port[port_num].roce.nb.notifier_call = NULL; return err; @@ -3324,8 +3323,7 @@ static int mlx5_add_netdev_notifier(struct mlx5_ib_dev *dev, u8 port_num) static void mlx5_remove_netdev_notifier(struct mlx5_ib_dev *dev, u8 port_num) { if (dev->port[port_num].roce.nb.notifier_call) { - unregister_netdevice_notifier_net(mlx5_core_net(dev->mdev), - &dev->port[port_num].roce.nb); + unregister_netdevice_notifier(&dev->port[port_num].roce.nb); dev->port[port_num].roce.nb.notifier_call = NULL; } } diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lib/mlx5.h b/drivers/net/ethernet/mellanox/mlx5/core/lib/mlx5.h index 3a9fa629503f..d046db7bb047 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/lib/mlx5.h +++ b/drivers/net/ethernet/mellanox/mlx5/core/lib/mlx5.h @@ -90,4 +90,9 @@ int mlx5_create_encryption_key(struct mlx5_core_dev *mdev, u32 key_type, u32 *p_key_id); void mlx5_destroy_encryption_key(struct mlx5_core_dev *mdev, u32 key_id); +static inline struct net *mlx5_core_net(struct mlx5_core_dev *dev) +{ + return devlink_net(priv_to_devlink(dev)); +} + #endif diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h index f93bfe7473aa..4672e12f1aa5 100644 --- a/include/linux/mlx5/driver.h +++ b/include/linux/mlx5/driver.h @@ -1209,22 +1209,4 @@ static inline bool mlx5_is_roce_enabled(struct mlx5_core_dev *dev) return val.vbool; } -/** - * mlx5_core_net - Provide net namespace of the mlx5_core_dev - * @dev: mlx5 core device - * - * mlx5_core_net() returns the net namespace of mlx5 core device. - * This can be called only in below described limited context. - * (a) When a devlink instance for mlx5_core is registered and - * when devlink reload operation is disabled. - * or - * (b) during devlink reload reload_down() and reload_up callbacks - * where it is ensured that devlink instance's net namespace is - * stable. - */ -static inline struct net *mlx5_core_net(struct mlx5_core_dev *dev) -{ - return devlink_net(priv_to_devlink(dev)); -} - #endif /* MLX5_DRIVER_H */ -- cgit v1.2.3 From f068cb1db2cb40c9782874df7b08c684106cf609 Mon Sep 17 00:00:00 2001 From: Joe Perches Date: Fri, 15 Jan 2021 16:36:50 -0800 Subject: RDMA/usnic: Fix misuse of sysfs_emit_at In commit e28bf1f03b01 ("RDMA: Convert various random sprintf sysfs _show uses to sysfs_emit") I mistakenly used len = sysfs_emit_at to overwrite the last trailing space of potentially multiple entry output. Instead use a more common style by removing the trailing space from the output formats and adding a prefixing space to the contination formats and converting the final terminating output newline from the defective len = sysfs_emit_at(buf, len, "\n"); to the now appropriate and typical len += sysfs_emit_at(buf, len, "\n"); Fixes: e28bf1f03b01 ("RDMA: Convert various random sprintf sysfs _show uses to sysfs_emit") Link: https://lore.kernel.org/r/5eb794b9c9bca0494d94b2b209f1627fa4e7b555.camel@perches.com Reported-by: James Bottomley Signed-off-by: Joe Perches Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/usnic/usnic_ib_sysfs.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) (limited to 'drivers') diff --git a/drivers/infiniband/hw/usnic/usnic_ib_sysfs.c b/drivers/infiniband/hw/usnic/usnic_ib_sysfs.c index e59615a4c9d9..586b0e52ba7f 100644 --- a/drivers/infiniband/hw/usnic/usnic_ib_sysfs.c +++ b/drivers/infiniband/hw/usnic/usnic_ib_sysfs.c @@ -214,7 +214,7 @@ static ssize_t summary_show(struct usnic_ib_qp_grp *qp_grp, char *buf) struct usnic_vnic_res *vnic_res; int len; - len = sysfs_emit(buf, "QPN: %d State: (%s) PID: %u VF Idx: %hu ", + len = sysfs_emit(buf, "QPN: %d State: (%s) PID: %u VF Idx: %hu", qp_grp->ibqp.qp_num, usnic_ib_qp_grp_state_to_string(qp_grp->state), qp_grp->owner_pid, @@ -224,14 +224,13 @@ static ssize_t summary_show(struct usnic_ib_qp_grp *qp_grp, char *buf) res_chunk = qp_grp->res_chunk_list[i]; for (j = 0; j < res_chunk->cnt; j++) { vnic_res = res_chunk->res[j]; - len += sysfs_emit_at( - buf, len, "%s[%d] ", + len += sysfs_emit_at(buf, len, " %s[%d]", usnic_vnic_res_type_to_str(vnic_res->type), vnic_res->vnic_idx); } } - len = sysfs_emit_at(buf, len, "\n"); + len += sysfs_emit_at(buf, len, "\n"); return len; } -- cgit v1.2.3 From f1b0a8ea9f12b8ade0dbe40dd57e4ffa9a30ed93 Mon Sep 17 00:00:00 2001 From: Martin Wilck Date: Wed, 20 Jan 2021 17:19:13 +0100 Subject: Revert "RDMA/rxe: Remove VLAN code leftovers from RXE" This reverts commit b2d2440430c0fdd5e0cad3efd6d1c9e3d3d02e5b. It's true that creating rxe on top of 802.1q interfaces doesn't work. Thus, commit fd49ddaf7e26 ("RDMA/rxe: prevent rxe creation on top of vlan interface") was absolutely correct. But b2d2440430c0 was incorrect assuming that with this change, RDMA and VLAN don't work togehter at all. It just has to be set up differently. Rather than creating rxe on top of the VLAN interface, rxe must be created on top of the physical interface. RDMA then works just fine through VLAN interfaces on top of that physical interface, via the "upper device" logic. This is hard to see in the rxe logic because it never talks about vlan, but instead rxe carefully selects upper vlan netdevices when working with packets which in turn imply certain vlan tagging. This is all done correctly and interacts with the gid table with VLAN support the same as real HW does. b2d2440430c0 broke this setup deliberately and should thus be reverted. Also, b2d2440430c0 removed rxe_dma_device(), so adapt the revert to discard that hunk. Fixes: b2d2440430c0 ("RDMA/rxe: Remove VLAN code leftovers from RXE") Link: https://lore.kernel.org/r/20210120161913.7347-1-mwilck@suse.com Signed-off-by: Martin Wilck Signed-off-by: Jason Gunthorpe --- drivers/infiniband/sw/rxe/rxe_net.c | 6 ++++++ drivers/infiniband/sw/rxe/rxe_resp.c | 5 +++++ 2 files changed, 11 insertions(+) (limited to 'drivers') diff --git a/drivers/infiniband/sw/rxe/rxe_net.c b/drivers/infiniband/sw/rxe/rxe_net.c index c4b06ced30a7..943914c2a50c 100644 --- a/drivers/infiniband/sw/rxe/rxe_net.c +++ b/drivers/infiniband/sw/rxe/rxe_net.c @@ -8,6 +8,7 @@ #include #include #include +#include #include #include #include @@ -153,9 +154,14 @@ static int rxe_udp_encap_recv(struct sock *sk, struct sk_buff *skb) { struct udphdr *udph; struct net_device *ndev = skb->dev; + struct net_device *rdev = ndev; struct rxe_dev *rxe = rxe_get_dev_from_net(ndev); struct rxe_pkt_info *pkt = SKB_TO_PKT(skb); + if (!rxe && is_vlan_dev(rdev)) { + rdev = vlan_dev_real_dev(ndev); + rxe = rxe_get_dev_from_net(rdev); + } if (!rxe) goto drop; diff --git a/drivers/infiniband/sw/rxe/rxe_resp.c b/drivers/infiniband/sw/rxe/rxe_resp.c index 5a098083a9d2..c7e3b6a4af38 100644 --- a/drivers/infiniband/sw/rxe/rxe_resp.c +++ b/drivers/infiniband/sw/rxe/rxe_resp.c @@ -872,6 +872,11 @@ static enum resp_states do_complete(struct rxe_qp *qp, else wc->network_hdr_type = RDMA_NETWORK_IPV6; + if (is_vlan_dev(skb->dev)) { + wc->wc_flags |= IB_WC_WITH_VLAN; + wc->vlan_id = vlan_dev_vlan_id(skb->dev); + } + if (pkt->mask & RXE_IMMDT_MASK) { wc->wc_flags |= IB_WC_WITH_IMM; wc->ex.imm_data = immdt_imm(pkt); -- cgit v1.2.3