From 3363828758597ed5e0f20172fccaad514b947031 Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Thu, 15 Jun 2017 11:29:03 +0300 Subject: [PATCH 01/34] IB/rxe: Use "foo *bar" instead of "foo * bar" Signed-off-by: Kamal Heib Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/sw/rxe/rxe.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/infiniband/sw/rxe/rxe.h b/drivers/infiniband/sw/rxe/rxe.h index 1ac5b8551a4d..6447d736d5a4 100644 --- a/drivers/infiniband/sw/rxe/rxe.h +++ b/drivers/infiniband/sw/rxe/rxe.h @@ -97,7 +97,7 @@ int rxe_rcv(struct sk_buff *skb); void rxe_dev_put(struct rxe_dev *rxe); struct rxe_dev *net_to_rxe(struct net_device *ndev); -struct rxe_dev *get_rxe_by_name(const char* name); +struct rxe_dev *get_rxe_by_name(const char *name); void rxe_port_up(struct rxe_dev *rxe); void rxe_port_down(struct rxe_dev *rxe); From c498e82e3c3231504831cd391970e63c1597a01b Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Thu, 15 Jun 2017 11:29:04 +0300 Subject: [PATCH 02/34] IB/rxe: Prefer 'unsigned int' to bare use of 'unsigned' Signed-off-by: Kamal Heib Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/sw/rxe/rxe_loc.h | 2 +- drivers/infiniband/sw/rxe/rxe_pool.c | 2 +- drivers/infiniband/sw/rxe/rxe_req.c | 2 +- 3 files changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/infiniband/sw/rxe/rxe_loc.h b/drivers/infiniband/sw/rxe/rxe_loc.h index d6299edf9a5b..6b65c0132289 100644 --- a/drivers/infiniband/sw/rxe/rxe_loc.h +++ b/drivers/infiniband/sw/rxe/rxe_loc.h @@ -250,7 +250,7 @@ void rxe_resp_queue_pkt(struct rxe_dev *rxe, void rxe_comp_queue_pkt(struct rxe_dev *rxe, struct rxe_qp *qp, struct sk_buff *skb); -static inline unsigned wr_opcode_mask(int opcode, struct rxe_qp *qp) +static inline unsigned int wr_opcode_mask(int opcode, struct rxe_qp *qp) { return rxe_wr_opcode_info[opcode].mask[qp->ibqp.qp_type]; } diff --git a/drivers/infiniband/sw/rxe/rxe_pool.c b/drivers/infiniband/sw/rxe/rxe_pool.c index 75d11ee635ec..c1b5f38f31a5 100644 --- a/drivers/infiniband/sw/rxe/rxe_pool.c +++ b/drivers/infiniband/sw/rxe/rxe_pool.c @@ -188,7 +188,7 @@ int rxe_pool_init( struct rxe_dev *rxe, struct rxe_pool *pool, enum rxe_elem_type type, - unsigned max_elem) + unsigned int max_elem) { int err = 0; size_t size = rxe_type_info[type].size; diff --git a/drivers/infiniband/sw/rxe/rxe_req.c b/drivers/infiniband/sw/rxe/rxe_req.c index 7ee465d1a1e1..db7161456f45 100644 --- a/drivers/infiniband/sw/rxe/rxe_req.c +++ b/drivers/infiniband/sw/rxe/rxe_req.c @@ -43,7 +43,7 @@ static int next_opcode(struct rxe_qp *qp, struct rxe_send_wqe *wqe, static inline void retry_first_write_send(struct rxe_qp *qp, struct rxe_send_wqe *wqe, - unsigned mask, int npsn) + unsigned int mask, int npsn) { int i; From c05d26647b06b49e286b11cbd14e1444f9acc1c5 Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Thu, 15 Jun 2017 11:29:05 +0300 Subject: [PATCH 03/34] IB/rxe: Use DEVICE_ATTR_RO macro to show parent field Use DEVICE_ATTR RO() macro and rename the show function accordingly. Signed-off-by: Kamal Heib Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/sw/rxe/rxe_verbs.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.c b/drivers/infiniband/sw/rxe/rxe_verbs.c index af90a7d42b96..ceab12a91d0c 100644 --- a/drivers/infiniband/sw/rxe/rxe_verbs.c +++ b/drivers/infiniband/sw/rxe/rxe_verbs.c @@ -1210,8 +1210,8 @@ static int rxe_detach_mcast(struct ib_qp *ibqp, union ib_gid *mgid, u16 mlid) return rxe_mcast_drop_grp_elem(rxe, qp, mgid); } -static ssize_t rxe_show_parent(struct device *device, - struct device_attribute *attr, char *buf) +static ssize_t parent_show(struct device *device, + struct device_attribute *attr, char *buf) { struct rxe_dev *rxe = container_of(device, struct rxe_dev, ib_dev.dev); @@ -1219,7 +1219,7 @@ static ssize_t rxe_show_parent(struct device *device, return snprintf(buf, 16, "%s\n", rxe_parent_name(rxe, 1)); } -static DEVICE_ATTR(parent, S_IRUGO, rxe_show_parent, NULL); +static DEVICE_ATTR_RO(parent); static struct device_attribute *rxe_dev_attributes[] = { &dev_attr_parent, From 61013828f6673fffc6ead0a62d52674ffc1b34dc Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Thu, 15 Jun 2017 11:29:06 +0300 Subject: [PATCH 04/34] IB/rxe: Use __func__ to print function's name Its better to use __func__ to print functions name instead of writing the name in the print statement. Signed-off-by: Kamal Heib Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/sw/rxe/rxe_task.c | 4 ++-- drivers/infiniband/sw/rxe/rxe_verbs.c | 6 +++--- 2 files changed, 5 insertions(+), 5 deletions(-) diff --git a/drivers/infiniband/sw/rxe/rxe_task.c b/drivers/infiniband/sw/rxe/rxe_task.c index d2a14a1bdc7f..ea3810b29273 100644 --- a/drivers/infiniband/sw/rxe/rxe_task.c +++ b/drivers/infiniband/sw/rxe/rxe_task.c @@ -78,7 +78,7 @@ void rxe_do_task(unsigned long data) default: spin_unlock_irqrestore(&task->state_lock, flags); - pr_warn("bad state = %d in rxe_do_task\n", task->state); + pr_warn("%s failed with bad state %d\n", __func__, task->state); return; } @@ -105,7 +105,7 @@ void rxe_do_task(unsigned long data) break; default: - pr_warn("bad state = %d in rxe_do_task\n", + pr_warn("%s failed with bad state %d\n", __func__, task->state); } spin_unlock_irqrestore(&task->state_lock, flags); diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.c b/drivers/infiniband/sw/rxe/rxe_verbs.c index ceab12a91d0c..aba3fa9ae599 100644 --- a/drivers/infiniband/sw/rxe/rxe_verbs.c +++ b/drivers/infiniband/sw/rxe/rxe_verbs.c @@ -1336,15 +1336,15 @@ int rxe_register_device(struct rxe_dev *rxe) err = ib_register_device(dev, NULL); if (err) { - pr_warn("rxe_register_device failed, err = %d\n", err); + pr_warn("%s failed with error %d\n", __func__, err); goto err1; } for (i = 0; i < ARRAY_SIZE(rxe_dev_attributes); ++i) { err = device_create_file(&dev->dev, rxe_dev_attributes[i]); if (err) { - pr_warn("device_create_file failed, i = %d, err = %d\n", - i, err); + pr_warn("%s failed with error %d for attr number %d\n", + __func__, err, i); goto err2; } } From b4fbec9673db22dcd3cea7ce04148af39fe19e12 Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Thu, 15 Jun 2017 11:29:07 +0300 Subject: [PATCH 05/34] IB/rxe: Constify static rxe_vm_ops Constify static rxe_vm_ops that is never modified. Signed-off-by: Kamal Heib Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/sw/rxe/rxe_mmap.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/infiniband/sw/rxe/rxe_mmap.c b/drivers/infiniband/sw/rxe/rxe_mmap.c index bd812e00988e..d22431e3a908 100644 --- a/drivers/infiniband/sw/rxe/rxe_mmap.c +++ b/drivers/infiniband/sw/rxe/rxe_mmap.c @@ -76,7 +76,7 @@ static void rxe_vma_close(struct vm_area_struct *vma) kref_put(&ip->ref, rxe_mmap_release); } -static struct vm_operations_struct rxe_vm_ops = { +static const struct vm_operations_struct rxe_vm_ops = { .open = rxe_vma_open, .close = rxe_vma_close, }; From bded747bb432bc5f7ad6d84ea747368b70ed9df2 Mon Sep 17 00:00:00 2001 From: Huy Nguyen Date: Tue, 30 May 2017 09:42:53 +0300 Subject: [PATCH 06/34] net/mlx5: Add raw ethernet local loopback firmware command Add support for raw ethernet local loopback firmware command. Signed-off-by: Huy Nguyen Reviewed-by: Daniel Jurgens Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- .../net/ethernet/mellanox/mlx5/core/vport.c | 62 +++++++++++++++++++ include/linux/mlx5/mlx5_ifc.h | 11 +++- include/linux/mlx5/vport.h | 3 +- 3 files changed, 72 insertions(+), 4 deletions(-) diff --git a/drivers/net/ethernet/mellanox/mlx5/core/vport.c b/drivers/net/ethernet/mellanox/mlx5/core/vport.c index 5abfec1c3399..d653b0025b13 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/vport.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/vport.c @@ -897,6 +897,68 @@ int mlx5_modify_nic_vport_promisc(struct mlx5_core_dev *mdev, } EXPORT_SYMBOL_GPL(mlx5_modify_nic_vport_promisc); +enum { + UC_LOCAL_LB, + MC_LOCAL_LB +}; + +int mlx5_nic_vport_update_local_lb(struct mlx5_core_dev *mdev, bool enable) +{ + int inlen = MLX5_ST_SZ_BYTES(modify_nic_vport_context_in); + void *in; + int err; + + mlx5_core_dbg(mdev, "%s local_lb\n", enable ? "enable" : "disable"); + in = kvzalloc(inlen, GFP_KERNEL); + if (!in) + return -ENOMEM; + + MLX5_SET(modify_nic_vport_context_in, in, + field_select.disable_mc_local_lb, 1); + MLX5_SET(modify_nic_vport_context_in, in, + nic_vport_context.disable_mc_local_lb, !enable); + + MLX5_SET(modify_nic_vport_context_in, in, + field_select.disable_uc_local_lb, 1); + MLX5_SET(modify_nic_vport_context_in, in, + nic_vport_context.disable_uc_local_lb, !enable); + + err = mlx5_modify_nic_vport_context(mdev, in, inlen); + + kvfree(in); + return err; +} +EXPORT_SYMBOL_GPL(mlx5_nic_vport_update_local_lb); + +int mlx5_nic_vport_query_local_lb(struct mlx5_core_dev *mdev, bool *status) +{ + int outlen = MLX5_ST_SZ_BYTES(query_nic_vport_context_out); + u32 *out; + int value; + int err; + + out = kzalloc(outlen, GFP_KERNEL); + if (!out) + return -ENOMEM; + + err = mlx5_query_nic_vport_context(mdev, 0, out, outlen); + if (err) + goto out; + + value = MLX5_GET(query_nic_vport_context_out, out, + nic_vport_context.disable_mc_local_lb) << MC_LOCAL_LB; + + value |= MLX5_GET(query_nic_vport_context_out, out, + nic_vport_context.disable_uc_local_lb) << UC_LOCAL_LB; + + *status = !value; + +out: + kfree(out); + return err; +} +EXPORT_SYMBOL_GPL(mlx5_nic_vport_query_local_lb); + enum mlx5_vport_roce_state { MLX5_VPORT_ROCE_DISABLED = 0, MLX5_VPORT_ROCE_ENABLED = 1, diff --git a/include/linux/mlx5/mlx5_ifc.h b/include/linux/mlx5/mlx5_ifc.h index 87869c04849a..57c75e8b3c19 100644 --- a/include/linux/mlx5/mlx5_ifc.h +++ b/include/linux/mlx5/mlx5_ifc.h @@ -1016,7 +1016,8 @@ struct mlx5_ifc_cmd_hca_cap_bits { u8 log_max_wq_sz[0x5]; u8 nic_vport_change_event[0x1]; - u8 reserved_at_3e1[0xa]; + u8 disable_local_lb[0x1]; + u8 reserved_at_3e2[0x9]; u8 log_max_vlan_list[0x5]; u8 reserved_at_3f0[0x3]; u8 log_max_current_mc_list[0x5]; @@ -2562,7 +2563,9 @@ struct mlx5_ifc_rmpc_bits { struct mlx5_ifc_nic_vport_context_bits { u8 reserved_at_0[0x5]; u8 min_wqe_inline_mode[0x3]; - u8 reserved_at_8[0x17]; + u8 reserved_at_8[0x15]; + u8 disable_mc_local_lb[0x1]; + u8 disable_uc_local_lb[0x1]; u8 roce_en[0x1]; u8 arm_change_event[0x1]; @@ -5229,7 +5232,9 @@ struct mlx5_ifc_modify_nic_vport_context_out_bits { }; struct mlx5_ifc_modify_nic_vport_field_select_bits { - u8 reserved_at_0[0x16]; + u8 reserved_at_0[0x14]; + u8 disable_uc_local_lb[0x1]; + u8 disable_mc_local_lb[0x1]; u8 node_guid[0x1]; u8 port_guid[0x1]; u8 min_inline[0x1]; diff --git a/include/linux/mlx5/vport.h b/include/linux/mlx5/vport.h index 656c70b65dd2..aaa0bb9e7655 100644 --- a/include/linux/mlx5/vport.h +++ b/include/linux/mlx5/vport.h @@ -114,5 +114,6 @@ int mlx5_core_modify_hca_vport_context(struct mlx5_core_dev *dev, u8 other_vport, u8 port_num, int vf, struct mlx5_hca_vport_context *req); - +int mlx5_nic_vport_update_local_lb(struct mlx5_core_dev *mdev, bool enable); +int mlx5_nic_vport_query_local_lb(struct mlx5_core_dev *mdev, bool *status); #endif /* __MLX5_VPORT_H__ */ From c85023e153e3824661d07307138fdeff41f6d86a Mon Sep 17 00:00:00 2001 From: Huy Nguyen Date: Tue, 30 May 2017 09:42:54 +0300 Subject: [PATCH 07/34] IB/mlx5: Add raw ethernet local loopback support Currently, unicast/multicast loopback raw ethernet (non-RDMA) packets are sent back to the vport. A unicast loopback packet is the packet with destination MAC address the same as the source MAC address. For multicast, the destination MAC address is in the vport's multicast filter list. Moreover, the local loopback is not needed if there is one or none user space context. After this patch, the raw ethernet unicast and multicast local loopback are disabled by default. When there is more than one user space context, the local loopback is enabled. Note that when local loopback is disabled, raw ethernet packets are not looped back to the vport and are forwarded to the next routing level (eswitch, or multihost switch, or out to the wire depending on the configuration). Signed-off-by: Huy Nguyen Reviewed-by: Daniel Jurgens Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx5/main.c | 51 +++++++++++++++++-- drivers/infiniband/hw/mlx5/mlx5_ib.h | 10 ++-- .../net/ethernet/mellanox/mlx5/core/main.c | 19 +++++++ 3 files changed, 73 insertions(+), 7 deletions(-) diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index a7f2e60085c4..85bf17a616d5 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -58,6 +58,7 @@ #include #include "mlx5_ib.h" #include "cmd.h" +#include #define DRIVER_NAME "mlx5_ib" #define DRIVER_VERSION "5.0-0" @@ -1187,6 +1188,45 @@ static int deallocate_uars(struct mlx5_ib_dev *dev, struct mlx5_ib_ucontext *con return 0; } +static int mlx5_ib_alloc_transport_domain(struct mlx5_ib_dev *dev, u32 *tdn) +{ + int err; + + err = mlx5_core_alloc_transport_domain(dev->mdev, tdn); + if (err) + return err; + + if ((MLX5_CAP_GEN(dev->mdev, port_type) != MLX5_CAP_PORT_TYPE_ETH) || + !MLX5_CAP_GEN(dev->mdev, disable_local_lb)) + return err; + + mutex_lock(&dev->lb_mutex); + dev->user_td++; + + if (dev->user_td == 2) + err = mlx5_nic_vport_update_local_lb(dev->mdev, true); + + mutex_unlock(&dev->lb_mutex); + return err; +} + +static void mlx5_ib_dealloc_transport_domain(struct mlx5_ib_dev *dev, u32 tdn) +{ + mlx5_core_dealloc_transport_domain(dev->mdev, tdn); + + if ((MLX5_CAP_GEN(dev->mdev, port_type) != MLX5_CAP_PORT_TYPE_ETH) || + !MLX5_CAP_GEN(dev->mdev, disable_local_lb)) + return; + + mutex_lock(&dev->lb_mutex); + dev->user_td--; + + if (dev->user_td < 2) + mlx5_nic_vport_update_local_lb(dev->mdev, false); + + mutex_unlock(&dev->lb_mutex); +} + static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev, struct ib_udata *udata) { @@ -1295,8 +1335,7 @@ static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev, mutex_init(&context->upd_xlt_page_mutex); if (MLX5_CAP_GEN(dev->mdev, log_max_transport_domain)) { - err = mlx5_core_alloc_transport_domain(dev->mdev, - &context->tdn); + err = mlx5_ib_alloc_transport_domain(dev, &context->tdn); if (err) goto out_page; } @@ -1362,7 +1401,7 @@ static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev, out_td: if (MLX5_CAP_GEN(dev->mdev, log_max_transport_domain)) - mlx5_core_dealloc_transport_domain(dev->mdev, context->tdn); + mlx5_ib_dealloc_transport_domain(dev, context->tdn); out_page: free_page(context->upd_xlt_page); @@ -1390,7 +1429,7 @@ static int mlx5_ib_dealloc_ucontext(struct ib_ucontext *ibcontext) bfregi = &context->bfregi; if (MLX5_CAP_GEN(dev->mdev, log_max_transport_domain)) - mlx5_core_dealloc_transport_domain(dev->mdev, context->tdn); + mlx5_ib_dealloc_transport_domain(dev, context->tdn); free_page(context->upd_xlt_page); deallocate_uars(dev, context); @@ -3797,6 +3836,10 @@ static void *mlx5_ib_add(struct mlx5_core_dev *mdev) goto err_umrc; } + if ((MLX5_CAP_GEN(mdev, port_type) == MLX5_CAP_PORT_TYPE_ETH) && + MLX5_CAP_GEN(mdev, disable_local_lb)) + mutex_init(&dev->lb_mutex); + dev->ib_active = true; return dev; diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index bdcf25410c99..d2b27647f939 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -652,9 +652,13 @@ struct mlx5_ib_dev { struct list_head qp_list; /* Array with num_ports elements */ struct mlx5_ib_port *port; - struct mlx5_sq_bfreg bfreg; - struct mlx5_sq_bfreg fp_bfreg; - u8 umr_fence; + struct mlx5_sq_bfreg bfreg; + struct mlx5_sq_bfreg fp_bfreg; + + /* protect the user_td */ + struct mutex lb_mutex; + u32 user_td; + u8 umr_fence; }; static inline struct mlx5_ib_cq *to_mibcq(struct mlx5_core_cq *mcq) diff --git a/drivers/net/ethernet/mellanox/mlx5/core/main.c b/drivers/net/ethernet/mellanox/mlx5/core/main.c index c065132b956d..3cec683fd70f 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/main.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/main.c @@ -47,6 +47,7 @@ #include #include #include +#include #ifdef CONFIG_RFS_ACCEL #include #endif @@ -579,6 +580,18 @@ static int set_hca_ctrl(struct mlx5_core_dev *dev) return err; } +static int mlx5_core_set_hca_defaults(struct mlx5_core_dev *dev) +{ + int ret = 0; + + /* Disable local_lb by default */ + if ((MLX5_CAP_GEN(dev, port_type) == MLX5_CAP_PORT_TYPE_ETH) && + MLX5_CAP_GEN(dev, disable_local_lb)) + ret = mlx5_nic_vport_update_local_lb(dev, false); + + return ret; +} + int mlx5_core_enable_hca(struct mlx5_core_dev *dev, u16 func_id) { u32 out[MLX5_ST_SZ_DW(enable_hca_out)] = {0}; @@ -1155,6 +1168,12 @@ static int mlx5_load_one(struct mlx5_core_dev *dev, struct mlx5_priv *priv, goto err_fs; } + err = mlx5_core_set_hca_defaults(dev); + if (err) { + dev_err(&pdev->dev, "Failed to set hca defaults\n"); + goto err_fs; + } + #ifdef CONFIG_MLX5_CORE_EN mlx5_eswitch_attach(dev->priv.eswitch); #endif From 2c43c5a036be7d03232a2b28f4a440a8cafe699f Mon Sep 17 00:00:00 2001 From: Huy Nguyen Date: Tue, 30 May 2017 09:42:55 +0300 Subject: [PATCH 08/34] net/mlx5e: Enable local loopback in loopback selftest Before running the ethtool's loopback selftest, we need to make sure that the local loopback is enabled. Signed-off-by: Huy Nguyen Reviewed-by: Daniel Jurgens Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- .../net/ethernet/mellanox/mlx5/core/en_selftest.c | 13 +++++++++++++ 1 file changed, 13 insertions(+) diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_selftest.c b/drivers/net/ethernet/mellanox/mlx5/core/en_selftest.c index 898759fcf9ec..1f1f8af87d4d 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_selftest.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_selftest.c @@ -189,6 +189,7 @@ struct mlx5e_lbt_priv { struct packet_type pt; struct completion comp; bool loopback_ok; + bool local_lb; }; static int @@ -236,6 +237,13 @@ static int mlx5e_test_loopback_setup(struct mlx5e_priv *priv, { int err = 0; + /* Temporarily enable local_lb */ + if (MLX5_CAP_GEN(priv->mdev, disable_local_lb)) { + mlx5_nic_vport_query_local_lb(priv->mdev, &lbtp->local_lb); + if (!lbtp->local_lb) + mlx5_nic_vport_update_local_lb(priv->mdev, true); + } + err = mlx5e_refresh_tirs(priv, true); if (err) return err; @@ -254,6 +262,11 @@ static int mlx5e_test_loopback_setup(struct mlx5e_priv *priv, static void mlx5e_test_loopback_cleanup(struct mlx5e_priv *priv, struct mlx5e_lbt_priv *lbtp) { + if (MLX5_CAP_GEN(priv->mdev, disable_local_lb)) { + if (!lbtp->local_lb) + mlx5_nic_vport_update_local_lb(priv->mdev, false); + } + dev_remove_pack(&lbtp->pt); mlx5e_refresh_tirs(priv, false); } From fd65f1b8eefdd20bd70cc920a7306e1e1b113d81 Mon Sep 17 00:00:00 2001 From: Moni Shoua Date: Tue, 30 May 2017 09:56:05 +0300 Subject: [PATCH 09/34] IB/mlx5: Change logic for dispatching IB events for port state The old logic ignored link state. This led to missing IB events like when link goes down on the switch while admin state is up or to redundant events like when admin state goes up while link is down. To fix that, probe the port state on NETDEV events and compare to last known state to decide if IB events needs to be dispatched. FIxes: 5ec8c83e3ad3 ("IB/mlx5: Port events in RoCE now rely on netdev events") Signed-off-by: Moni Shoua Reviewed-by: Noa Osherovich Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx5/main.c | 33 ++++++++++++++++++++++++++-- drivers/infiniband/hw/mlx5/mlx5_ib.h | 1 + 2 files changed, 32 insertions(+), 2 deletions(-) diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index 85bf17a616d5..bfd9117da10c 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -98,6 +98,20 @@ mlx5_ib_port_link_layer(struct ib_device *device, u8 port_num) return mlx5_port_type_cap_to_rdma_ll(port_type_cap); } +static int get_port_state(struct ib_device *ibdev, + u8 port_num, + enum ib_port_state *state) +{ + struct ib_port_attr attr; + int ret; + + memset(&attr, 0, sizeof(attr)); + ret = mlx5_ib_query_port(ibdev, port_num, &attr); + if (!ret) + *state = attr.state; + return ret; +} + static int mlx5_netdev_event(struct notifier_block *this, unsigned long event, void *ptr) { @@ -115,6 +129,7 @@ static int mlx5_netdev_event(struct notifier_block *this, write_unlock(&ibdev->roce.netdev_lock); break; + case NETDEV_CHANGE: case NETDEV_UP: case NETDEV_DOWN: { struct net_device *lag_ndev = mlx5_lag_get_roce_netdev(ibdev->mdev); @@ -128,10 +143,23 @@ static int mlx5_netdev_event(struct notifier_block *this, if ((upper == ndev || (!upper && ndev == ibdev->roce.netdev)) && ibdev->ib_active) { struct ib_event ibev = { }; + enum ib_port_state port_state; + if (get_port_state(&ibdev->ib_dev, 1, &port_state)) + return NOTIFY_DONE; + + if (ibdev->roce.last_port_state == port_state) + return NOTIFY_DONE; + + ibdev->roce.last_port_state = port_state; ibev.device = &ibdev->ib_dev; - ibev.event = (event == NETDEV_UP) ? - IB_EVENT_PORT_ACTIVE : IB_EVENT_PORT_ERR; + if (port_state == IB_PORT_DOWN) + ibev.event = IB_EVENT_PORT_ERR; + else if (port_state == IB_PORT_ACTIVE) + ibev.event = IB_EVENT_PORT_ACTIVE; + else + return NOTIFY_DONE; + ibev.element.port_num = 1; ib_dispatch_event(&ibev); } @@ -3793,6 +3821,7 @@ static void *mlx5_ib_add(struct mlx5_core_dev *mdev) err = mlx5_enable_eth(dev); if (err) goto err_free_port; + dev->roce.last_port_state = IB_PORT_DOWN; } err = create_dev_resources(&dev->devr); diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index d2b27647f939..5fb4547bbcb8 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -616,6 +616,7 @@ struct mlx5_roce { struct net_device *netdev; struct notifier_block nb; atomic_t next_port; + enum ib_port_state last_port_state; }; struct mlx5_ib_dev { From 4a2da0b8c0782816f3ae6846ae7942fcbb0f8172 Mon Sep 17 00:00:00 2001 From: Parav Pandit Date: Tue, 30 May 2017 10:05:15 +0300 Subject: [PATCH 10/34] IB/mlx5: Add debug control parameters for congestion control This patch adds debug control parameters for congestion control which can be read or written through debugfs. They are for reaction point and notification point nodes. These control parameters are as below: +------------------------------+-----------------------------------------+ | Name | Description | |------------------------------+-----------------------------------------| |rp_clamp_tgt_rate | When set target rate is updated to | | | current rate | |------------------------------+-----------------------------------------| |rp_clamp_tgt_rate_ati | When set update target rate based on | | | timer as well | |------------------------------+-----------------------------------------| |rp_time_reset | time between rate increase if no | | | CNP is received unit in usec | |------------------------------+-----------------------------------------| |rp_byte_reset | Number of bytes between rate inease if | | | no CNP is received | |------------------------------+-----------------------------------------| |rp_threshold | Threshold for reaction point rate | | | control | |------------------------------+-----------------------------------------| |rp_ai_rate | Rate for target rate, unit in Mbps | |------------------------------+-----------------------------------------| |rp_hai_rate | Rate for hyper increase state | | | unit in Mbps | |------------------------------+-----------------------------------------| |rp_min_dec_fac | Minimum factor by which the current | | | transmit rate can be changed when | | | processing a CNP, unit is percerntage | |------------------------------+-----------------------------------------| |rp_min_rate | Minimum value for rate limit, | | | unit in Mbps | |------------------------------+-----------------------------------------| |rp_rate_to_set_on_first_cnp | Rate that is set when first CNP is | | | received, unit is Mbps | |------------------------------+-----------------------------------------| |rp_dce_tcp_g | Used to calculate alpha | |------------------------------+-----------------------------------------| |rp_dce_tcp_rtt | Time between updates of alpha value, | | | unit is usec | |------------------------------+-----------------------------------------| |rp_rate_reduce_monitor_period | Minimum time between consecutive rate | | | reductions | |------------------------------+-----------------------------------------| |rp_initial_alpha_value | Initial value of alpha | |------------------------------+-----------------------------------------| |rp_gd | When CNP is received, flow rate is | | | reduced based on gd, rp_gd is given as | | | log2(rp_gd) | |------------------------------+-----------------------------------------| |np_cnp_dscp | dscp code point for generated cnp | |------------------------------+-----------------------------------------| |np_cnp_prio_mode | 802.1p priority for generated cnp | |------------------------------+-----------------------------------------| |np_cnp_prio | cnp priority mode | +------------------------------+-----------------------------------------+ Signed-off-by: Parav Pandit Reviewed-by: Daniel Jurgens Reviewed-by: Eli Cohen Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx5/Makefile | 2 +- drivers/infiniband/hw/mlx5/cmd.c | 20 ++ drivers/infiniband/hw/mlx5/cmd.h | 4 + drivers/infiniband/hw/mlx5/cong.c | 421 +++++++++++++++++++++++++++ drivers/infiniband/hw/mlx5/main.c | 9 +- drivers/infiniband/hw/mlx5/mlx5_ib.h | 37 +++ include/linux/mlx5/mlx5_ifc.h | 3 +- 7 files changed, 493 insertions(+), 3 deletions(-) create mode 100644 drivers/infiniband/hw/mlx5/cong.c diff --git a/drivers/infiniband/hw/mlx5/Makefile b/drivers/infiniband/hw/mlx5/Makefile index 90ad2adc752f..bc6299697dda 100644 --- a/drivers/infiniband/hw/mlx5/Makefile +++ b/drivers/infiniband/hw/mlx5/Makefile @@ -1,4 +1,4 @@ obj-$(CONFIG_MLX5_INFINIBAND) += mlx5_ib.o -mlx5_ib-y := main.o cq.o doorbell.o qp.o mem.o srq.o mr.o ah.o mad.o gsi.o ib_virt.o cmd.o +mlx5_ib-y := main.o cq.o doorbell.o qp.o mem.o srq.o mr.o ah.o mad.o gsi.o ib_virt.o cmd.o cong.o mlx5_ib-$(CONFIG_INFINIBAND_ON_DEMAND_PAGING) += odp.o diff --git a/drivers/infiniband/hw/mlx5/cmd.c b/drivers/infiniband/hw/mlx5/cmd.c index 18d5e1db93ed..470995fa38d2 100644 --- a/drivers/infiniband/hw/mlx5/cmd.c +++ b/drivers/infiniband/hw/mlx5/cmd.c @@ -57,3 +57,23 @@ int mlx5_cmd_query_cong_counter(struct mlx5_core_dev *dev, MLX5_SET(query_cong_statistics_in, in, clear, reset); return mlx5_cmd_exec(dev, in, sizeof(in), out, out_size); } + +int mlx5_cmd_query_cong_params(struct mlx5_core_dev *dev, int cong_point, + void *out, int out_size) +{ + u32 in[MLX5_ST_SZ_DW(query_cong_params_in)] = { }; + + MLX5_SET(query_cong_params_in, in, opcode, + MLX5_CMD_OP_QUERY_CONG_PARAMS); + MLX5_SET(query_cong_params_in, in, cong_protocol, cong_point); + + return mlx5_cmd_exec(dev, in, sizeof(in), out, out_size); +} + +int mlx5_cmd_modify_cong_params(struct mlx5_core_dev *dev, + void *in, int in_size) +{ + u32 out[MLX5_ST_SZ_DW(modify_cong_params_out)] = { }; + + return mlx5_cmd_exec(dev, in, in_size, out, sizeof(out)); +} diff --git a/drivers/infiniband/hw/mlx5/cmd.h b/drivers/infiniband/hw/mlx5/cmd.h index fa09228193a6..af4c24596274 100644 --- a/drivers/infiniband/hw/mlx5/cmd.h +++ b/drivers/infiniband/hw/mlx5/cmd.h @@ -39,4 +39,8 @@ int mlx5_cmd_null_mkey(struct mlx5_core_dev *dev, u32 *null_mkey); int mlx5_cmd_query_cong_counter(struct mlx5_core_dev *dev, bool reset, void *out, int out_size); +int mlx5_cmd_query_cong_params(struct mlx5_core_dev *dev, int cong_point, + void *out, int out_size); +int mlx5_cmd_modify_cong_params(struct mlx5_core_dev *mdev, + void *in, int in_size); #endif /* MLX5_IB_CMD_H */ diff --git a/drivers/infiniband/hw/mlx5/cong.c b/drivers/infiniband/hw/mlx5/cong.c new file mode 100644 index 000000000000..2d32b519bb61 --- /dev/null +++ b/drivers/infiniband/hw/mlx5/cong.c @@ -0,0 +1,421 @@ +/* + * Copyright (c) 2013-2017, Mellanox Technologies. 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. + */ + +#include + +#include "mlx5_ib.h" +#include "cmd.h" + +enum mlx5_ib_cong_node_type { + MLX5_IB_RROCE_ECN_RP = 1, + MLX5_IB_RROCE_ECN_NP = 2, +}; + +static const char * const mlx5_ib_dbg_cc_name[] = { + "rp_clamp_tgt_rate", + "rp_clamp_tgt_rate_ati", + "rp_time_reset", + "rp_byte_reset", + "rp_threshold", + "rp_ai_rate", + "rp_hai_rate", + "rp_min_dec_fac", + "rp_min_rate", + "rp_rate_to_set_on_first_cnp", + "rp_dce_tcp_g", + "rp_dce_tcp_rtt", + "rp_rate_reduce_monitor_period", + "rp_initial_alpha_value", + "rp_gd", + "np_cnp_dscp", + "np_cnp_prio_mode", + "np_cnp_prio", +}; + +#define MLX5_IB_RP_CLAMP_TGT_RATE_ATTR BIT(1) +#define MLX5_IB_RP_CLAMP_TGT_RATE_ATI_ATTR BIT(2) +#define MLX5_IB_RP_TIME_RESET_ATTR BIT(3) +#define MLX5_IB_RP_BYTE_RESET_ATTR BIT(4) +#define MLX5_IB_RP_THRESHOLD_ATTR BIT(5) +#define MLX5_IB_RP_AI_RATE_ATTR BIT(7) +#define MLX5_IB_RP_HAI_RATE_ATTR BIT(8) +#define MLX5_IB_RP_MIN_DEC_FAC_ATTR BIT(9) +#define MLX5_IB_RP_MIN_RATE_ATTR BIT(10) +#define MLX5_IB_RP_RATE_TO_SET_ON_FIRST_CNP_ATTR BIT(11) +#define MLX5_IB_RP_DCE_TCP_G_ATTR BIT(12) +#define MLX5_IB_RP_DCE_TCP_RTT_ATTR BIT(13) +#define MLX5_IB_RP_RATE_REDUCE_MONITOR_PERIOD_ATTR BIT(14) +#define MLX5_IB_RP_INITIAL_ALPHA_VALUE_ATTR BIT(15) +#define MLX5_IB_RP_GD_ATTR BIT(16) + +#define MLX5_IB_NP_CNP_DSCP_ATTR BIT(3) +#define MLX5_IB_NP_CNP_PRIO_MODE_ATTR BIT(4) + +static enum mlx5_ib_cong_node_type +mlx5_ib_param_to_node(enum mlx5_ib_dbg_cc_types param_offset) +{ + if (param_offset >= MLX5_IB_DBG_CC_RP_CLAMP_TGT_RATE && + param_offset <= MLX5_IB_DBG_CC_RP_GD) + return MLX5_IB_RROCE_ECN_RP; + else + return MLX5_IB_RROCE_ECN_NP; +} + +static u32 mlx5_get_cc_param_val(void *field, int offset) +{ + switch (offset) { + case MLX5_IB_DBG_CC_RP_CLAMP_TGT_RATE: + return MLX5_GET(cong_control_r_roce_ecn_rp, field, + clamp_tgt_rate); + case MLX5_IB_DBG_CC_RP_CLAMP_TGT_RATE_ATI: + return MLX5_GET(cong_control_r_roce_ecn_rp, field, + clamp_tgt_rate_after_time_inc); + case MLX5_IB_DBG_CC_RP_TIME_RESET: + return MLX5_GET(cong_control_r_roce_ecn_rp, field, + rpg_time_reset); + case MLX5_IB_DBG_CC_RP_BYTE_RESET: + return MLX5_GET(cong_control_r_roce_ecn_rp, field, + rpg_byte_reset); + case MLX5_IB_DBG_CC_RP_THRESHOLD: + return MLX5_GET(cong_control_r_roce_ecn_rp, field, + rpg_threshold); + case MLX5_IB_DBG_CC_RP_AI_RATE: + return MLX5_GET(cong_control_r_roce_ecn_rp, field, + rpg_ai_rate); + case MLX5_IB_DBG_CC_RP_HAI_RATE: + return MLX5_GET(cong_control_r_roce_ecn_rp, field, + rpg_hai_rate); + case MLX5_IB_DBG_CC_RP_MIN_DEC_FAC: + return MLX5_GET(cong_control_r_roce_ecn_rp, field, + rpg_min_dec_fac); + case MLX5_IB_DBG_CC_RP_MIN_RATE: + return MLX5_GET(cong_control_r_roce_ecn_rp, field, + rpg_min_rate); + case MLX5_IB_DBG_CC_RP_RATE_TO_SET_ON_FIRST_CNP: + return MLX5_GET(cong_control_r_roce_ecn_rp, field, + rate_to_set_on_first_cnp); + case MLX5_IB_DBG_CC_RP_DCE_TCP_G: + return MLX5_GET(cong_control_r_roce_ecn_rp, field, + dce_tcp_g); + case MLX5_IB_DBG_CC_RP_DCE_TCP_RTT: + return MLX5_GET(cong_control_r_roce_ecn_rp, field, + dce_tcp_rtt); + case MLX5_IB_DBG_CC_RP_RATE_REDUCE_MONITOR_PERIOD: + return MLX5_GET(cong_control_r_roce_ecn_rp, field, + rate_reduce_monitor_period); + case MLX5_IB_DBG_CC_RP_INITIAL_ALPHA_VALUE: + return MLX5_GET(cong_control_r_roce_ecn_rp, field, + initial_alpha_value); + case MLX5_IB_DBG_CC_RP_GD: + return MLX5_GET(cong_control_r_roce_ecn_rp, field, + rpg_gd); + case MLX5_IB_DBG_CC_NP_CNP_DSCP: + return MLX5_GET(cong_control_r_roce_ecn_np, field, + cnp_dscp); + case MLX5_IB_DBG_CC_NP_CNP_PRIO_MODE: + return MLX5_GET(cong_control_r_roce_ecn_np, field, + cnp_prio_mode); + case MLX5_IB_DBG_CC_NP_CNP_PRIO: + return MLX5_GET(cong_control_r_roce_ecn_np, field, + cnp_802p_prio); + default: + return 0; + } +} + +static void mlx5_ib_set_cc_param_mask_val(void *field, int offset, + u32 var, u32 *attr_mask) +{ + switch (offset) { + case MLX5_IB_DBG_CC_RP_CLAMP_TGT_RATE: + *attr_mask |= MLX5_IB_RP_CLAMP_TGT_RATE_ATTR; + MLX5_SET(cong_control_r_roce_ecn_rp, field, + clamp_tgt_rate, var); + break; + case MLX5_IB_DBG_CC_RP_CLAMP_TGT_RATE_ATI: + *attr_mask |= MLX5_IB_RP_CLAMP_TGT_RATE_ATI_ATTR; + MLX5_SET(cong_control_r_roce_ecn_rp, field, + clamp_tgt_rate_after_time_inc, var); + break; + case MLX5_IB_DBG_CC_RP_TIME_RESET: + *attr_mask |= MLX5_IB_RP_TIME_RESET_ATTR; + MLX5_SET(cong_control_r_roce_ecn_rp, field, + rpg_time_reset, var); + break; + case MLX5_IB_DBG_CC_RP_BYTE_RESET: + *attr_mask |= MLX5_IB_RP_BYTE_RESET_ATTR; + MLX5_SET(cong_control_r_roce_ecn_rp, field, + rpg_byte_reset, var); + break; + case MLX5_IB_DBG_CC_RP_THRESHOLD: + *attr_mask |= MLX5_IB_RP_THRESHOLD_ATTR; + MLX5_SET(cong_control_r_roce_ecn_rp, field, + rpg_threshold, var); + break; + case MLX5_IB_DBG_CC_RP_AI_RATE: + *attr_mask |= MLX5_IB_RP_AI_RATE_ATTR; + MLX5_SET(cong_control_r_roce_ecn_rp, field, + rpg_ai_rate, var); + break; + case MLX5_IB_DBG_CC_RP_HAI_RATE: + *attr_mask |= MLX5_IB_RP_HAI_RATE_ATTR; + MLX5_SET(cong_control_r_roce_ecn_rp, field, + rpg_hai_rate, var); + break; + case MLX5_IB_DBG_CC_RP_MIN_DEC_FAC: + *attr_mask |= MLX5_IB_RP_MIN_DEC_FAC_ATTR; + MLX5_SET(cong_control_r_roce_ecn_rp, field, + rpg_min_dec_fac, var); + break; + case MLX5_IB_DBG_CC_RP_MIN_RATE: + *attr_mask |= MLX5_IB_RP_MIN_RATE_ATTR; + MLX5_SET(cong_control_r_roce_ecn_rp, field, + rpg_min_rate, var); + break; + case MLX5_IB_DBG_CC_RP_RATE_TO_SET_ON_FIRST_CNP: + *attr_mask |= MLX5_IB_RP_RATE_TO_SET_ON_FIRST_CNP_ATTR; + MLX5_SET(cong_control_r_roce_ecn_rp, field, + rate_to_set_on_first_cnp, var); + break; + case MLX5_IB_DBG_CC_RP_DCE_TCP_G: + *attr_mask |= MLX5_IB_RP_DCE_TCP_G_ATTR; + MLX5_SET(cong_control_r_roce_ecn_rp, field, + dce_tcp_g, var); + break; + case MLX5_IB_DBG_CC_RP_DCE_TCP_RTT: + *attr_mask |= MLX5_IB_RP_DCE_TCP_RTT_ATTR; + MLX5_SET(cong_control_r_roce_ecn_rp, field, + dce_tcp_rtt, var); + break; + case MLX5_IB_DBG_CC_RP_RATE_REDUCE_MONITOR_PERIOD: + *attr_mask |= MLX5_IB_RP_RATE_REDUCE_MONITOR_PERIOD_ATTR; + MLX5_SET(cong_control_r_roce_ecn_rp, field, + rate_reduce_monitor_period, var); + break; + case MLX5_IB_DBG_CC_RP_INITIAL_ALPHA_VALUE: + *attr_mask |= MLX5_IB_RP_INITIAL_ALPHA_VALUE_ATTR; + MLX5_SET(cong_control_r_roce_ecn_rp, field, + initial_alpha_value, var); + break; + case MLX5_IB_DBG_CC_RP_GD: + *attr_mask |= MLX5_IB_RP_GD_ATTR; + MLX5_SET(cong_control_r_roce_ecn_rp, field, + rpg_gd, var); + break; + case MLX5_IB_DBG_CC_NP_CNP_DSCP: + *attr_mask |= MLX5_IB_NP_CNP_DSCP_ATTR; + MLX5_SET(cong_control_r_roce_ecn_np, field, cnp_dscp, var); + break; + case MLX5_IB_DBG_CC_NP_CNP_PRIO_MODE: + *attr_mask |= MLX5_IB_NP_CNP_PRIO_MODE_ATTR; + MLX5_SET(cong_control_r_roce_ecn_np, field, cnp_prio_mode, var); + break; + case MLX5_IB_DBG_CC_NP_CNP_PRIO: + *attr_mask |= MLX5_IB_NP_CNP_PRIO_MODE_ATTR; + MLX5_SET(cong_control_r_roce_ecn_np, field, cnp_prio_mode, 0); + MLX5_SET(cong_control_r_roce_ecn_np, field, cnp_802p_prio, var); + break; + } +} + +static int mlx5_ib_get_cc_params(struct mlx5_ib_dev *dev, int offset, u32 *var) +{ + int outlen = MLX5_ST_SZ_BYTES(query_cong_params_out); + void *out; + void *field; + int err; + enum mlx5_ib_cong_node_type node; + + out = kvzalloc(outlen, GFP_KERNEL); + if (!out) + return -ENOMEM; + + node = mlx5_ib_param_to_node(offset); + + err = mlx5_cmd_query_cong_params(dev->mdev, node, out, outlen); + if (err) + goto free; + + field = MLX5_ADDR_OF(query_cong_params_out, out, congestion_parameters); + *var = mlx5_get_cc_param_val(field, offset); + +free: + kvfree(out); + return err; +} + +static int mlx5_ib_set_cc_params(struct mlx5_ib_dev *dev, int offset, u32 var) +{ + int inlen = MLX5_ST_SZ_BYTES(modify_cong_params_in); + void *in; + void *field; + enum mlx5_ib_cong_node_type node; + u32 attr_mask = 0; + int err; + + in = kvzalloc(inlen, GFP_KERNEL); + if (!in) + return -ENOMEM; + + MLX5_SET(modify_cong_params_in, in, opcode, + MLX5_CMD_OP_MODIFY_CONG_PARAMS); + + node = mlx5_ib_param_to_node(offset); + MLX5_SET(modify_cong_params_in, in, cong_protocol, node); + + field = MLX5_ADDR_OF(modify_cong_params_in, in, congestion_parameters); + mlx5_ib_set_cc_param_mask_val(field, offset, var, &attr_mask); + + field = MLX5_ADDR_OF(modify_cong_params_in, in, field_select); + MLX5_SET(field_select_r_roce_rp, field, field_select_r_roce_rp, + attr_mask); + + err = mlx5_cmd_modify_cong_params(dev->mdev, in, inlen); + kvfree(in); + return err; +} + +static ssize_t set_param(struct file *filp, const char __user *buf, + size_t count, loff_t *pos) +{ + struct mlx5_ib_dbg_param *param = filp->private_data; + int offset = param->offset; + char lbuf[11] = { }; + u32 var; + int ret; + + if (count > sizeof(lbuf)) + return -EINVAL; + + if (copy_from_user(lbuf, buf, count)) + return -EFAULT; + + lbuf[sizeof(lbuf) - 1] = '\0'; + + if (kstrtou32(lbuf, 0, &var)) + return -EINVAL; + + ret = mlx5_ib_set_cc_params(param->dev, offset, var); + return ret ? ret : count; +} + +static ssize_t get_param(struct file *filp, char __user *buf, size_t count, + loff_t *pos) +{ + struct mlx5_ib_dbg_param *param = filp->private_data; + int offset = param->offset; + u32 var = 0; + int ret; + char lbuf[11]; + + if (*pos) + return 0; + + ret = mlx5_ib_get_cc_params(param->dev, offset, &var); + if (ret) + return ret; + + ret = snprintf(lbuf, sizeof(lbuf), "%d\n", var); + if (ret < 0) + return ret; + + if (copy_to_user(buf, lbuf, ret)) + return -EFAULT; + + *pos += ret; + return ret; +} + +static const struct file_operations dbg_cc_fops = { + .owner = THIS_MODULE, + .open = simple_open, + .write = set_param, + .read = get_param, +}; + +void mlx5_ib_cleanup_cong_debugfs(struct mlx5_ib_dev *dev) +{ + if (!mlx5_debugfs_root || + !dev->dbg_cc_params || + !dev->dbg_cc_params->root) + return; + + debugfs_remove_recursive(dev->dbg_cc_params->root); + kfree(dev->dbg_cc_params); + dev->dbg_cc_params = NULL; +} + +int mlx5_ib_init_cong_debugfs(struct mlx5_ib_dev *dev) +{ + struct mlx5_ib_dbg_cc_params *dbg_cc_params; + int i; + + if (!mlx5_debugfs_root) + goto out; + + if (!MLX5_CAP_GEN(dev->mdev, cc_query_allowed) || + !MLX5_CAP_GEN(dev->mdev, cc_modify_allowed)) + goto out; + + dbg_cc_params = kzalloc(sizeof(*dbg_cc_params), GFP_KERNEL); + if (!dbg_cc_params) + goto out; + + dev->dbg_cc_params = dbg_cc_params; + + dbg_cc_params->root = debugfs_create_dir("cc_params", + dev->mdev->priv.dbg_root); + if (!dbg_cc_params->root) + goto err; + + for (i = 0; i < MLX5_IB_DBG_CC_MAX; i++) { + dbg_cc_params->params[i].offset = i; + dbg_cc_params->params[i].dev = dev; + dbg_cc_params->params[i].dentry = + debugfs_create_file(mlx5_ib_dbg_cc_name[i], + 0600, dbg_cc_params->root, + &dbg_cc_params->params[i], + &dbg_cc_fops); + if (!dbg_cc_params->params[i].dentry) + goto err; + } +out: return 0; + +err: + mlx5_ib_warn(dev, "cong debugfs failure\n"); + mlx5_ib_cleanup_cong_debugfs(dev); + /* + * We don't want to fail driver if debugfs failed to initialize, + * so we are not forwarding error to the user. + */ + return 0; +} diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index bfd9117da10c..a903728627f1 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -3838,9 +3838,13 @@ static void *mlx5_ib_add(struct mlx5_core_dev *mdev) goto err_odp; } + err = mlx5_ib_init_cong_debugfs(dev); + if (err) + goto err_cnt; + dev->mdev->priv.uar = mlx5_get_uars_page(dev->mdev); if (!dev->mdev->priv.uar) - goto err_cnt; + goto err_cong; err = mlx5_alloc_bfreg(dev->mdev, &dev->bfreg, false, false); if (err) @@ -3889,6 +3893,8 @@ static void *mlx5_ib_add(struct mlx5_core_dev *mdev) mlx5_put_uars_page(dev->mdev, dev->mdev->priv.uar); err_cnt: + mlx5_ib_cleanup_cong_debugfs(dev); +err_cong: if (MLX5_CAP_GEN(dev->mdev, max_qp_cnt)) mlx5_ib_dealloc_counters(dev); @@ -3923,6 +3929,7 @@ static void mlx5_ib_remove(struct mlx5_core_dev *mdev, void *context) mlx5_free_bfreg(dev->mdev, &dev->fp_bfreg); mlx5_free_bfreg(dev->mdev, &dev->bfreg); mlx5_put_uars_page(dev->mdev, mdev->priv.uar); + mlx5_ib_cleanup_cong_debugfs(dev); if (MLX5_CAP_GEN(dev->mdev, max_qp_cnt)) mlx5_ib_dealloc_counters(dev); destroy_umrc_res(dev); diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index 5fb4547bbcb8..f0682f383b52 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -619,6 +619,39 @@ struct mlx5_roce { enum ib_port_state last_port_state; }; +struct mlx5_ib_dbg_param { + int offset; + struct mlx5_ib_dev *dev; + struct dentry *dentry; +}; + +enum mlx5_ib_dbg_cc_types { + MLX5_IB_DBG_CC_RP_CLAMP_TGT_RATE, + MLX5_IB_DBG_CC_RP_CLAMP_TGT_RATE_ATI, + MLX5_IB_DBG_CC_RP_TIME_RESET, + MLX5_IB_DBG_CC_RP_BYTE_RESET, + MLX5_IB_DBG_CC_RP_THRESHOLD, + MLX5_IB_DBG_CC_RP_AI_RATE, + MLX5_IB_DBG_CC_RP_HAI_RATE, + MLX5_IB_DBG_CC_RP_MIN_DEC_FAC, + MLX5_IB_DBG_CC_RP_MIN_RATE, + MLX5_IB_DBG_CC_RP_RATE_TO_SET_ON_FIRST_CNP, + MLX5_IB_DBG_CC_RP_DCE_TCP_G, + MLX5_IB_DBG_CC_RP_DCE_TCP_RTT, + MLX5_IB_DBG_CC_RP_RATE_REDUCE_MONITOR_PERIOD, + MLX5_IB_DBG_CC_RP_INITIAL_ALPHA_VALUE, + MLX5_IB_DBG_CC_RP_GD, + MLX5_IB_DBG_CC_NP_CNP_DSCP, + MLX5_IB_DBG_CC_NP_CNP_PRIO_MODE, + MLX5_IB_DBG_CC_NP_CNP_PRIO, + MLX5_IB_DBG_CC_MAX, +}; + +struct mlx5_ib_dbg_cc_params { + struct dentry *root; + struct mlx5_ib_dbg_param params[MLX5_IB_DBG_CC_MAX]; +}; + struct mlx5_ib_dev { struct ib_device ib_dev; struct mlx5_core_dev *mdev; @@ -655,6 +688,7 @@ struct mlx5_ib_dev { struct mlx5_ib_port *port; struct mlx5_sq_bfreg bfreg; struct mlx5_sq_bfreg fp_bfreg; + struct mlx5_ib_dbg_cc_params *dbg_cc_params; /* protect the user_td */ struct mutex lb_mutex; @@ -909,6 +943,9 @@ __be16 mlx5_get_roce_udp_sport(struct mlx5_ib_dev *dev, u8 port_num, int mlx5_get_roce_gid_type(struct mlx5_ib_dev *dev, u8 port_num, int index, enum ib_gid_type *gid_type); +void mlx5_ib_cleanup_cong_debugfs(struct mlx5_ib_dev *dev); +int mlx5_ib_init_cong_debugfs(struct mlx5_ib_dev *dev); + /* GSI QP helper functions */ struct ib_qp *mlx5_ib_gsi_create_qp(struct ib_pd *pd, struct ib_qp_init_attr *init_attr); diff --git a/include/linux/mlx5/mlx5_ifc.h b/include/linux/mlx5/mlx5_ifc.h index 57c75e8b3c19..3e697f672de5 100644 --- a/include/linux/mlx5/mlx5_ifc.h +++ b/include/linux/mlx5/mlx5_ifc.h @@ -1188,7 +1188,8 @@ struct mlx5_ifc_cong_control_r_roce_ecn_np_bits { u8 reserved_at_c0[0x12]; u8 cnp_dscp[0x6]; - u8 reserved_at_d8[0x5]; + u8 reserved_at_d8[0x4]; + u8 cnp_prio_mode[0x1]; u8 cnp_802p_prio[0x3]; u8 reserved_at_e0[0x720]; From 7ecf6d8ff154e6f7471ee537a400d43a5f3b1c57 Mon Sep 17 00:00:00 2001 From: Bodong Wang Date: Tue, 30 May 2017 10:18:24 +0300 Subject: [PATCH 11/34] IB/mlx5: Restore IB guid/policy for virtual functions When a user sets port_guid, node_guid or policy of an IB virtual function, save this information in "struct mlx5_vf_context". This information will be restored later when pci_resume is called. To make sure this works, one can use aer-inject to generate PCI errors on mlx5 devices and verify if relevant fields are restored after PCI resume. Signed-off-by: Bodong Wang Reviewed-by: Eli Cohen Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx5/ib_virt.c | 9 ++++ .../net/ethernet/mellanox/mlx5/core/sriov.c | 42 +++++++++++++++++++ include/linux/mlx5/driver.h | 17 ++++---- 3 files changed, 61 insertions(+), 7 deletions(-) diff --git a/drivers/infiniband/hw/mlx5/ib_virt.c b/drivers/infiniband/hw/mlx5/ib_virt.c index c1b9de800fe5..649a3364f838 100644 --- a/drivers/infiniband/hw/mlx5/ib_virt.c +++ b/drivers/infiniband/hw/mlx5/ib_virt.c @@ -96,6 +96,7 @@ int mlx5_ib_set_vf_link_state(struct ib_device *device, int vf, struct mlx5_ib_dev *dev = to_mdev(device); struct mlx5_core_dev *mdev = dev->mdev; struct mlx5_hca_vport_context *in; + struct mlx5_vf_context *vfs_ctx = mdev->priv.sriov.vfs_ctx; int err; in = kzalloc(sizeof(*in), GFP_KERNEL); @@ -109,6 +110,8 @@ int mlx5_ib_set_vf_link_state(struct ib_device *device, int vf, } in->field_select = MLX5_HCA_VPORT_SEL_STATE_POLICY; err = mlx5_core_modify_hca_vport_context(mdev, 1, 1, vf + 1, in); + if (!err) + vfs_ctx[vf].policy = in->policy; out: kfree(in); @@ -151,6 +154,7 @@ static int set_vf_node_guid(struct ib_device *device, int vf, u8 port, u64 guid) struct mlx5_ib_dev *dev = to_mdev(device); struct mlx5_core_dev *mdev = dev->mdev; struct mlx5_hca_vport_context *in; + struct mlx5_vf_context *vfs_ctx = mdev->priv.sriov.vfs_ctx; int err; in = kzalloc(sizeof(*in), GFP_KERNEL); @@ -160,6 +164,8 @@ static int set_vf_node_guid(struct ib_device *device, int vf, u8 port, u64 guid) in->field_select = MLX5_HCA_VPORT_SEL_NODE_GUID; in->node_guid = guid; err = mlx5_core_modify_hca_vport_context(mdev, 1, 1, vf + 1, in); + if (!err) + vfs_ctx[vf].node_guid = guid; kfree(in); return err; } @@ -169,6 +175,7 @@ static int set_vf_port_guid(struct ib_device *device, int vf, u8 port, u64 guid) struct mlx5_ib_dev *dev = to_mdev(device); struct mlx5_core_dev *mdev = dev->mdev; struct mlx5_hca_vport_context *in; + struct mlx5_vf_context *vfs_ctx = mdev->priv.sriov.vfs_ctx; int err; in = kzalloc(sizeof(*in), GFP_KERNEL); @@ -178,6 +185,8 @@ static int set_vf_port_guid(struct ib_device *device, int vf, u8 port, u64 guid) in->field_select = MLX5_HCA_VPORT_SEL_PORT_GUID; in->port_guid = guid; err = mlx5_core_modify_hca_vport_context(mdev, 1, 1, vf + 1, in); + if (!err) + vfs_ctx[vf].port_guid = guid; kfree(in); return err; } diff --git a/drivers/net/ethernet/mellanox/mlx5/core/sriov.c b/drivers/net/ethernet/mellanox/mlx5/core/sriov.c index bcdf7779c48d..090b29e05a6a 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/sriov.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/sriov.c @@ -32,6 +32,7 @@ #include #include +#include #include "mlx5_core.h" #ifdef CONFIG_MLX5_CORE_EN #include "eswitch.h" @@ -44,6 +45,38 @@ bool mlx5_sriov_is_enabled(struct mlx5_core_dev *dev) return !!sriov->num_vfs; } +static int sriov_restore_guids(struct mlx5_core_dev *dev, int vf) +{ + struct mlx5_core_sriov *sriov = &dev->priv.sriov; + struct mlx5_hca_vport_context *in; + int err = 0; + + /* Restore sriov guid and policy settings */ + if (sriov->vfs_ctx[vf].node_guid || + sriov->vfs_ctx[vf].port_guid || + sriov->vfs_ctx[vf].policy != MLX5_POLICY_INVALID) { + in = kzalloc(sizeof(*in), GFP_KERNEL); + if (!in) + return -ENOMEM; + + in->node_guid = sriov->vfs_ctx[vf].node_guid; + in->port_guid = sriov->vfs_ctx[vf].port_guid; + in->policy = sriov->vfs_ctx[vf].policy; + in->field_select = + !!(in->port_guid) * MLX5_HCA_VPORT_SEL_PORT_GUID | + !!(in->node_guid) * MLX5_HCA_VPORT_SEL_NODE_GUID | + !!(in->policy) * MLX5_HCA_VPORT_SEL_STATE_POLICY; + + err = mlx5_core_modify_hca_vport_context(dev, 1, 1, vf + 1, in); + if (err) + mlx5_core_warn(dev, "modify vport context failed, unable to restore VF %d settings\n", vf); + + kfree(in); + } + + return err; +} + static int mlx5_device_enable_sriov(struct mlx5_core_dev *dev, int num_vfs) { struct mlx5_core_sriov *sriov = &dev->priv.sriov; @@ -74,6 +107,15 @@ static int mlx5_device_enable_sriov(struct mlx5_core_dev *dev, int num_vfs) } sriov->vfs_ctx[vf].enabled = 1; sriov->enabled_vfs++; + if (MLX5_CAP_GEN(dev, port_type) == MLX5_CAP_PORT_TYPE_IB) { + err = sriov_restore_guids(dev, vf); + if (err) { + mlx5_core_warn(dev, + "failed to restore VF %d settings, err %d\n", + vf, err); + continue; + } + } mlx5_core_dbg(dev, "successfully enabled VF* %d\n", vf); } diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h index df6ce59a1f95..54221be5f69e 100644 --- a/include/linux/mlx5/driver.h +++ b/include/linux/mlx5/driver.h @@ -162,6 +162,13 @@ enum dbg_rsc_type { MLX5_DBG_RSC_CQ, }; +enum port_state_policy { + MLX5_POLICY_DOWN = 0, + MLX5_POLICY_UP = 1, + MLX5_POLICY_FOLLOW = 2, + MLX5_POLICY_INVALID = 0xffffffff +}; + struct mlx5_field_desc { struct dentry *dent; int i; @@ -525,6 +532,9 @@ struct mlx5_mkey_table { struct mlx5_vf_context { int enabled; + u64 port_guid; + u64 node_guid; + enum port_state_policy policy; }; struct mlx5_core_sriov { @@ -842,13 +852,6 @@ struct mlx5_pas { u8 log_sz; }; -enum port_state_policy { - MLX5_POLICY_DOWN = 0, - MLX5_POLICY_UP = 1, - MLX5_POLICY_FOLLOW = 2, - MLX5_POLICY_INVALID = 0xffffffff -}; - enum phy_port_state { MLX5_AAA_111 }; From 7d9336d80b0b35d3537d37ff35e08dcb425073ed Mon Sep 17 00:00:00 2001 From: Maor Gottlieb Date: Tue, 30 May 2017 10:29:10 +0300 Subject: [PATCH 12/34] IB/core: Introduce delay drop for a WQ Work queue which is created with IB_WQ_FLAGS_DELAY_DROP won't cause packet drops when there aren't receive WQEs, but will wait until posting of receive WQEs or for some period of time that the device was configured with. It includes: * Add a new creation flag to enable delay drop functionality in a WQ. * A new capability was introduced - IB_RAW_PACKET_CAP_DELAY_DROP, which is the device's ability to delay packet drops when there aren't receive WQEs. Signed-off-by: Maor Gottlieb Reviewed-by: Yishai Hadas Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- include/rdma/ib_verbs.h | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h index b5732432bb29..4a1444456af7 100644 --- a/include/rdma/ib_verbs.h +++ b/include/rdma/ib_verbs.h @@ -1546,6 +1546,10 @@ enum ib_raw_packet_caps { IB_RAW_PACKET_CAP_SCATTER_FCS = (1 << 1), /* Checksum offloads are supported (for both send and receive). */ IB_RAW_PACKET_CAP_IP_CSUM = (1 << 2), + /* When a packet is received for an RQ with no receive WQEs, the + * packet processing is delayed. + */ + IB_RAW_PACKET_CAP_DELAY_DROP = (1 << 3), }; enum ib_wq_type { @@ -1574,6 +1578,7 @@ struct ib_wq { enum ib_wq_flags { IB_WQ_FLAGS_CVLAN_STRIPPING = 1 << 0, IB_WQ_FLAGS_SCATTER_FCS = 1 << 1, + IB_WQ_FLAGS_DELAY_DROP = 1 << 2, }; struct ib_wq_init_attr { From c1e0bfc1312d0e06bdb24e6e4e7e10b0b4313ec6 Mon Sep 17 00:00:00 2001 From: Maor Gottlieb Date: Tue, 30 May 2017 10:29:11 +0300 Subject: [PATCH 13/34] net/mlx5: Introduce set delay drop command Add support to SET_DELAY_DROP command. This command will be used in downstream patches for delay packet drop. The timeout value should be indicated by delay_drop_timeout field. Packet processing will be delayed till timeout value passed or until more WQEs are posted. Setting this value to 0 disables the feature. Signed-off-by: Maor Gottlieb Reviewed-by: Yishai Hadas Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/net/ethernet/mellanox/mlx5/core/qp.c | 14 +++++++++++ include/linux/mlx5/mlx5_ifc.h | 25 +++++++++++++++++++- include/linux/mlx5/qp.h | 3 +++ 3 files changed, 41 insertions(+), 1 deletion(-) diff --git a/drivers/net/ethernet/mellanox/mlx5/core/qp.c b/drivers/net/ethernet/mellanox/mlx5/core/qp.c index 340f281c9801..db9e665ab104 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/qp.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/qp.c @@ -242,6 +242,20 @@ int mlx5_core_destroy_qp(struct mlx5_core_dev *dev, } EXPORT_SYMBOL_GPL(mlx5_core_destroy_qp); +int mlx5_core_set_delay_drop(struct mlx5_core_dev *dev, + u32 timeout_usec) +{ + u32 out[MLX5_ST_SZ_DW(set_delay_drop_params_out)] = {0}; + u32 in[MLX5_ST_SZ_DW(set_delay_drop_params_in)] = {0}; + + MLX5_SET(set_delay_drop_params_in, in, opcode, + MLX5_CMD_OP_SET_DELAY_DROP_PARAMS); + MLX5_SET(set_delay_drop_params_in, in, delay_drop_timeout, + timeout_usec / 100); + return mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out)); +} +EXPORT_SYMBOL_GPL(mlx5_core_set_delay_drop); + struct mbox_info { u32 *in; u32 *out; diff --git a/include/linux/mlx5/mlx5_ifc.h b/include/linux/mlx5/mlx5_ifc.h index 3e697f672de5..40c05d29d5bb 100644 --- a/include/linux/mlx5/mlx5_ifc.h +++ b/include/linux/mlx5/mlx5_ifc.h @@ -200,6 +200,7 @@ enum { MLX5_CMD_OP_QUERY_SQ = 0x907, MLX5_CMD_OP_CREATE_RQ = 0x908, MLX5_CMD_OP_MODIFY_RQ = 0x909, + MLX5_CMD_OP_SET_DELAY_DROP_PARAMS = 0x910, MLX5_CMD_OP_DESTROY_RQ = 0x90a, MLX5_CMD_OP_QUERY_RQ = 0x90b, MLX5_CMD_OP_CREATE_RMP = 0x90c, @@ -840,7 +841,7 @@ struct mlx5_ifc_cmd_hca_cap_bits { u8 retransmission_q_counters[0x1]; u8 reserved_at_183[0x1]; u8 modify_rq_counter_set_id[0x1]; - u8 reserved_at_185[0x1]; + u8 rq_delay_drop[0x1]; u8 max_qp_cnt[0xa]; u8 pkey_table_size[0x10]; @@ -5853,6 +5854,28 @@ struct mlx5_ifc_destroy_rq_in_bits { u8 reserved_at_60[0x20]; }; +struct mlx5_ifc_set_delay_drop_params_in_bits { + u8 opcode[0x10]; + u8 reserved_at_10[0x10]; + + u8 reserved_at_20[0x10]; + u8 op_mod[0x10]; + + u8 reserved_at_40[0x20]; + + u8 reserved_at_60[0x10]; + u8 delay_drop_timeout[0x10]; +}; + +struct mlx5_ifc_set_delay_drop_params_out_bits { + u8 status[0x8]; + u8 reserved_at_8[0x18]; + + u8 syndrome[0x20]; + + u8 reserved_at_40[0x40]; +}; + struct mlx5_ifc_destroy_rmp_out_bits { u8 status[0x8]; u8 reserved_at_8[0x18]; diff --git a/include/linux/mlx5/qp.h b/include/linux/mlx5/qp.h index 6f41270d80c0..fff4ec13f620 100644 --- a/include/linux/mlx5/qp.h +++ b/include/linux/mlx5/qp.h @@ -561,6 +561,9 @@ int mlx5_core_destroy_qp(struct mlx5_core_dev *dev, int mlx5_core_qp_query(struct mlx5_core_dev *dev, struct mlx5_core_qp *qp, u32 *out, int outlen); +int mlx5_core_set_delay_drop(struct mlx5_core_dev *dev, + u32 timeout_usec); + int mlx5_core_xrcd_alloc(struct mlx5_core_dev *dev, u32 *xrcdn); int mlx5_core_xrcd_dealloc(struct mlx5_core_dev *dev, u32 xrcdn); void mlx5_init_qp_table(struct mlx5_core_dev *dev); From 246ac9814c5b2c0e9916dca5fbf8d6a40245fad1 Mon Sep 17 00:00:00 2001 From: Maor Gottlieb Date: Tue, 30 May 2017 10:29:12 +0300 Subject: [PATCH 14/34] net/mlx5: Introduce general notification event When delay drop timeout is expired, the firmware raises general notification event of DELAY_DROP_TIMEOUT subtype. In addition the feature is disable so the driver have to reactivate the timeout. Signed-off-by: Maor Gottlieb Reviewed-by: Yishai Hadas Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/net/ethernet/mellanox/mlx5/core/eq.c | 23 ++++++++++++++++++++ include/linux/mlx5/device.h | 5 +++++ include/linux/mlx5/driver.h | 1 + include/linux/mlx5/mlx5_ifc.h | 3 ++- 4 files changed, 31 insertions(+), 1 deletion(-) diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eq.c b/drivers/net/ethernet/mellanox/mlx5/core/eq.c index af51a5d2b912..849417425811 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/eq.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/eq.c @@ -161,6 +161,8 @@ static const char *eqe_type_str(u8 type) return "MLX5_EVENT_TYPE_NIC_VPORT_CHANGE"; case MLX5_EVENT_TYPE_FPGA_ERROR: return "MLX5_EVENT_TYPE_FPGA_ERROR"; + case MLX5_EVENT_TYPE_GENERAL_EVENT: + return "MLX5_EVENT_TYPE_GENERAL_EVENT"; default: return "Unrecognized event"; } @@ -378,6 +380,20 @@ int mlx5_core_page_fault_resume(struct mlx5_core_dev *dev, u32 token, EXPORT_SYMBOL_GPL(mlx5_core_page_fault_resume); #endif +static void general_event_handler(struct mlx5_core_dev *dev, + struct mlx5_eqe *eqe) +{ + switch (eqe->sub_type) { + case MLX5_GENERAL_SUBTYPE_DELAY_DROP_TIMEOUT: + if (dev->event) + dev->event(dev, MLX5_DEV_EVENT_DELAY_DROP_TIMEOUT, 0); + break; + default: + mlx5_core_dbg(dev, "General event with unrecognized subtype: sub_type %d\n", + eqe->sub_type); + } +} + static irqreturn_t mlx5_eq_int(int irq, void *eq_ptr) { struct mlx5_eq *eq = eq_ptr; @@ -486,6 +502,9 @@ static irqreturn_t mlx5_eq_int(int irq, void *eq_ptr) mlx5_fpga_event(dev, eqe->type, &eqe->data.raw); break; + case MLX5_EVENT_TYPE_GENERAL_EVENT: + general_event_handler(dev, eqe); + break; default: mlx5_core_warn(dev, "Unhandled event 0x%x on EQ 0x%x\n", eqe->type, eq->eqn); @@ -693,6 +712,10 @@ int mlx5_start_eqs(struct mlx5_core_dev *dev) mlx5_core_is_pf(dev)) async_event_mask |= (1ull << MLX5_EVENT_TYPE_NIC_VPORT_CHANGE); + if (MLX5_CAP_GEN(dev, port_type) == MLX5_CAP_PORT_TYPE_ETH && + MLX5_CAP_GEN(dev, general_notification_event)) + async_event_mask |= (1ull << MLX5_EVENT_TYPE_GENERAL_EVENT); + if (MLX5_CAP_GEN(dev, port_module_event)) async_event_mask |= (1ull << MLX5_EVENT_TYPE_PORT_MODULE_EVENT); else diff --git a/include/linux/mlx5/device.h b/include/linux/mlx5/device.h index f31a0b5377e1..a47b9ab9f2c9 100644 --- a/include/linux/mlx5/device.h +++ b/include/linux/mlx5/device.h @@ -290,6 +290,7 @@ enum mlx5_event { MLX5_EVENT_TYPE_GPIO_EVENT = 0x15, MLX5_EVENT_TYPE_PORT_MODULE_EVENT = 0x16, MLX5_EVENT_TYPE_REMOTE_CONFIG = 0x19, + MLX5_EVENT_TYPE_GENERAL_EVENT = 0x22, MLX5_EVENT_TYPE_PPS_EVENT = 0x25, MLX5_EVENT_TYPE_DB_BF_CONGESTION = 0x1a, @@ -304,6 +305,10 @@ enum mlx5_event { MLX5_EVENT_TYPE_FPGA_ERROR = 0x20, }; +enum { + MLX5_GENERAL_SUBTYPE_DELAY_DROP_TIMEOUT = 0x1, +}; + enum { MLX5_PORT_CHANGE_SUBTYPE_DOWN = 1, MLX5_PORT_CHANGE_SUBTYPE_ACTIVE = 4, diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h index 54221be5f69e..758ef40f9316 100644 --- a/include/linux/mlx5/driver.h +++ b/include/linux/mlx5/driver.h @@ -192,6 +192,7 @@ enum mlx5_dev_event { MLX5_DEV_EVENT_GUID_CHANGE, MLX5_DEV_EVENT_CLIENT_REREG, MLX5_DEV_EVENT_PPS, + MLX5_DEV_EVENT_DELAY_DROP_TIMEOUT, }; enum mlx5_port_status { diff --git a/include/linux/mlx5/mlx5_ifc.h b/include/linux/mlx5/mlx5_ifc.h index 40c05d29d5bb..4bc57647fad8 100644 --- a/include/linux/mlx5/mlx5_ifc.h +++ b/include/linux/mlx5/mlx5_ifc.h @@ -874,7 +874,8 @@ struct mlx5_ifc_cmd_hca_cap_bits { u8 max_tc[0x4]; u8 reserved_at_1d0[0x1]; u8 dcbx[0x1]; - u8 reserved_at_1d2[0x3]; + u8 general_notification_event[0x1]; + u8 reserved_at_1d3[0x2]; u8 fpga[0x1]; u8 rol_s[0x1]; u8 rol_g[0x1]; From 03404e8ae652e02a5e3388224836cef53d7a0988 Mon Sep 17 00:00:00 2001 From: Maor Gottlieb Date: Tue, 30 May 2017 10:29:13 +0300 Subject: [PATCH 15/34] IB/mlx5: Add support to dropless RQ RQs that were configured for "delay drop" will prevent packet drops when their WQEs are depleted. Marking an RQ to be drop-less is done by setting delay_drop_en in RQ context using CREATE_RQ command. Since this feature is globally activated/deactivated by using the SET_DELAY_DROP command on all the marked RQs, we activated/deactivated it according to the number of RQs with 'delay_drop' enabled. When timeout is expired, then the feature is deactivated. Therefore the driver handles the delay drop timeout event and reactivate it. Signed-off-by: Maor Gottlieb Reviewed-by: Yishai Hadas Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx5/main.c | 60 ++++++++++++++++++++++++++-- drivers/infiniband/hw/mlx5/mlx5_ib.h | 19 +++++++++ drivers/infiniband/hw/mlx5/qp.c | 37 +++++++++++++++++ include/linux/mlx5/mlx5_ifc.h | 2 +- 4 files changed, 113 insertions(+), 5 deletions(-) diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index a903728627f1..ad4b12decc23 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -697,6 +697,10 @@ static int mlx5_ib_query_device(struct ib_device *ibdev, props->device_cap_flags |= IB_DEVICE_UD_TSO; } + if (MLX5_CAP_GEN(dev->mdev, rq_delay_drop) && + MLX5_CAP_GEN(dev->mdev, general_notification_event)) + props->raw_packet_caps |= IB_RAW_PACKET_CAP_DELAY_DROP; + if (MLX5_CAP_GEN(dev->mdev, eth_net_offloads) && MLX5_CAP_ETH(dev->mdev, scatter_fcs)) { /* Legacy bit to support old userspace libraries */ @@ -2752,6 +2756,24 @@ static void mlx5_ib_handle_internal_error(struct mlx5_ib_dev *ibdev) spin_unlock_irqrestore(&ibdev->reset_flow_resource_lock, flags); } +static void delay_drop_handler(struct work_struct *work) +{ + int err; + struct mlx5_ib_delay_drop *delay_drop = + container_of(work, struct mlx5_ib_delay_drop, + delay_drop_work); + + mutex_lock(&delay_drop->lock); + err = mlx5_core_set_delay_drop(delay_drop->dev->mdev, + delay_drop->timeout); + if (err) { + mlx5_ib_warn(delay_drop->dev, "Failed to set delay drop, timeout=%u\n", + delay_drop->timeout); + delay_drop->activate = false; + } + mutex_unlock(&delay_drop->lock); +} + static void mlx5_ib_event(struct mlx5_core_dev *dev, void *context, enum mlx5_dev_event event, unsigned long param) { @@ -2804,8 +2826,11 @@ static void mlx5_ib_event(struct mlx5_core_dev *dev, void *context, ibev.event = IB_EVENT_CLIENT_REREGISTER; port = (u8)param; break; + case MLX5_DEV_EVENT_DELAY_DROP_TIMEOUT: + schedule_work(&ibdev->delay_drop.delay_drop_work); + goto out; default: - return; + goto out; } ibev.device = &ibdev->ib_dev; @@ -2813,7 +2838,7 @@ static void mlx5_ib_event(struct mlx5_core_dev *dev, void *context, if (port < 1 || port > ibdev->num_ports) { mlx5_ib_warn(ibdev, "warning: event on port %d\n", port); - return; + goto out; } if (ibdev->ib_active) @@ -2821,6 +2846,9 @@ static void mlx5_ib_event(struct mlx5_core_dev *dev, void *context, if (fatal) ibdev->ib_active = false; + +out: + return; } static int set_has_smi_cap(struct mlx5_ib_dev *dev) @@ -3623,6 +3651,26 @@ mlx5_ib_alloc_rdma_netdev(struct ib_device *hca, return netdev; } +static void cancel_delay_drop(struct mlx5_ib_dev *dev) +{ + if (!(dev->ib_dev.attrs.raw_packet_caps & IB_RAW_PACKET_CAP_DELAY_DROP)) + return; + + cancel_work_sync(&dev->delay_drop.delay_drop_work); +} + +static void init_delay_drop(struct mlx5_ib_dev *dev) +{ + if (!(dev->ib_dev.attrs.raw_packet_caps & IB_RAW_PACKET_CAP_DELAY_DROP)) + return; + + mutex_init(&dev->delay_drop.lock); + dev->delay_drop.dev = dev; + dev->delay_drop.activate = false; + dev->delay_drop.timeout = MLX5_MAX_DELAY_DROP_TIMEOUT_MS * 1000; + INIT_WORK(&dev->delay_drop.delay_drop_work, delay_drop_handler); +} + static void *mlx5_ib_add(struct mlx5_core_dev *mdev) { struct mlx5_ib_dev *dev; @@ -3862,11 +3910,13 @@ static void *mlx5_ib_add(struct mlx5_core_dev *mdev) if (err) goto err_dev; + init_delay_drop(dev); + for (i = 0; i < ARRAY_SIZE(mlx5_class_attributes); i++) { err = device_create_file(&dev->ib_dev.dev, mlx5_class_attributes[i]); if (err) - goto err_umrc; + goto err_delay_drop; } if ((MLX5_CAP_GEN(mdev, port_type) == MLX5_CAP_PORT_TYPE_ETH) && @@ -3877,7 +3927,8 @@ static void *mlx5_ib_add(struct mlx5_core_dev *mdev) return dev; -err_umrc: +err_delay_drop: + cancel_delay_drop(dev); destroy_umrc_res(dev); err_dev: @@ -3924,6 +3975,7 @@ static void mlx5_ib_remove(struct mlx5_core_dev *mdev, void *context) struct mlx5_ib_dev *dev = context; enum rdma_link_layer ll = mlx5_ib_port_link_layer(&dev->ib_dev, 1); + cancel_delay_drop(dev); mlx5_remove_netdev_notifier(dev); ib_unregister_device(&dev->ib_dev); mlx5_free_bfreg(dev->mdev, &dev->fp_bfreg); diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index f0682f383b52..097f12dc65a6 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -247,6 +247,10 @@ struct mlx5_ib_wq { void *qend; }; +enum mlx5_ib_wq_flags { + MLX5_IB_WQ_FLAGS_DELAY_DROP = 0x1, +}; + struct mlx5_ib_rwq { struct ib_wq ibwq; struct mlx5_core_qp core_qp; @@ -264,6 +268,7 @@ struct mlx5_ib_rwq { u32 wqe_count; u32 wqe_shift; int wq_sig; + u32 create_flags; /* Use enum mlx5_ib_wq_flags */ }; enum { @@ -652,6 +657,19 @@ struct mlx5_ib_dbg_cc_params { struct mlx5_ib_dbg_param params[MLX5_IB_DBG_CC_MAX]; }; +enum { + MLX5_MAX_DELAY_DROP_TIMEOUT_MS = 100, +}; + +struct mlx5_ib_delay_drop { + struct mlx5_ib_dev *dev; + struct work_struct delay_drop_work; + /* serialize setting of delay drop */ + struct mutex lock; + u32 timeout; + bool activate; +}; + struct mlx5_ib_dev { struct ib_device ib_dev; struct mlx5_core_dev *mdev; @@ -688,6 +706,7 @@ struct mlx5_ib_dev { struct mlx5_ib_port *port; struct mlx5_sq_bfreg bfreg; struct mlx5_sq_bfreg fp_bfreg; + struct mlx5_ib_delay_drop delay_drop; struct mlx5_ib_dbg_cc_params *dbg_cc_params; /* protect the user_td */ diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c index 0889ff367c86..939553d5c25f 100644 --- a/drivers/infiniband/hw/mlx5/qp.c +++ b/drivers/infiniband/hw/mlx5/qp.c @@ -4597,6 +4597,24 @@ static void mlx5_ib_wq_event(struct mlx5_core_qp *core_qp, int type) } } +static int set_delay_drop(struct mlx5_ib_dev *dev) +{ + int err = 0; + + mutex_lock(&dev->delay_drop.lock); + if (dev->delay_drop.activate) + goto out; + + err = mlx5_core_set_delay_drop(dev->mdev, dev->delay_drop.timeout); + if (err) + goto out; + + dev->delay_drop.activate = true; +out: + mutex_unlock(&dev->delay_drop.lock); + return err; +} + static int create_rq(struct mlx5_ib_rwq *rwq, struct ib_pd *pd, struct ib_wq_init_attr *init_attr) { @@ -4651,9 +4669,28 @@ static int create_rq(struct mlx5_ib_rwq *rwq, struct ib_pd *pd, } MLX5_SET(rqc, rqc, scatter_fcs, 1); } + if (init_attr->create_flags & IB_WQ_FLAGS_DELAY_DROP) { + if (!(dev->ib_dev.attrs.raw_packet_caps & + IB_RAW_PACKET_CAP_DELAY_DROP)) { + mlx5_ib_dbg(dev, "Delay drop is not supported\n"); + err = -EOPNOTSUPP; + goto out; + } + MLX5_SET(rqc, rqc, delay_drop_en, 1); + } rq_pas0 = (__be64 *)MLX5_ADDR_OF(wq, wq, pas); mlx5_ib_populate_pas(dev, rwq->umem, rwq->page_shift, rq_pas0, 0); err = mlx5_core_create_rq_tracked(dev->mdev, in, inlen, &rwq->core_qp); + if (!err && init_attr->create_flags & IB_WQ_FLAGS_DELAY_DROP) { + err = set_delay_drop(dev); + if (err) { + mlx5_ib_warn(dev, "Failed to enable delay drop err=%d\n", + err); + mlx5_core_destroy_rq_tracked(dev->mdev, &rwq->core_qp); + } else { + rwq->create_flags |= MLX5_IB_WQ_FLAGS_DELAY_DROP; + } + } out: kvfree(in); return err; diff --git a/include/linux/mlx5/mlx5_ifc.h b/include/linux/mlx5/mlx5_ifc.h index 4bc57647fad8..f350688a12fa 100644 --- a/include/linux/mlx5/mlx5_ifc.h +++ b/include/linux/mlx5/mlx5_ifc.h @@ -2519,7 +2519,7 @@ enum { struct mlx5_ifc_rqc_bits { u8 rlky[0x1]; - u8 reserved_at_1[0x1]; + u8 delay_drop_en[0x1]; u8 scatter_fcs[0x1]; u8 vsd[0x1]; u8 mem_rq_type[0x4]; From fe248c3a5837848717ed566fb4aefe66f43a5e53 Mon Sep 17 00:00:00 2001 From: Maor Gottlieb Date: Tue, 30 May 2017 10:29:14 +0300 Subject: [PATCH 16/34] IB/mlx5: Add delay drop configuration and statistics Add debugfs interface for monitor the number of delay drop timeout events and the number of existing dropless RQs in the system. In addition add debugfs interface for configuring the global timeout value which is used in the SET_DELAY_DROP command. Signed-off-by: Maor Gottlieb Reviewed-by: Yishai Hadas Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx5/main.c | 103 +++++++++++++++++++++++++++ drivers/infiniband/hw/mlx5/mlx5_ib.h | 10 +++ drivers/infiniband/hw/mlx5/qp.c | 13 +++- 3 files changed, 123 insertions(+), 3 deletions(-) diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index ad4b12decc23..7455f95bf679 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -30,6 +30,7 @@ * SOFTWARE. */ +#include #include #include #include @@ -2763,6 +2764,8 @@ static void delay_drop_handler(struct work_struct *work) container_of(work, struct mlx5_ib_delay_drop, delay_drop_work); + atomic_inc(&delay_drop->events_cnt); + mutex_lock(&delay_drop->lock); err = mlx5_core_set_delay_drop(delay_drop->dev->mdev, delay_drop->timeout); @@ -3651,12 +3654,107 @@ mlx5_ib_alloc_rdma_netdev(struct ib_device *hca, return netdev; } +static void delay_drop_debugfs_cleanup(struct mlx5_ib_dev *dev) +{ + if (!dev->delay_drop.dbg) + return; + debugfs_remove_recursive(dev->delay_drop.dbg->dir_debugfs); + kfree(dev->delay_drop.dbg); + dev->delay_drop.dbg = NULL; +} + static void cancel_delay_drop(struct mlx5_ib_dev *dev) { if (!(dev->ib_dev.attrs.raw_packet_caps & IB_RAW_PACKET_CAP_DELAY_DROP)) return; cancel_work_sync(&dev->delay_drop.delay_drop_work); + delay_drop_debugfs_cleanup(dev); +} + +static ssize_t delay_drop_timeout_read(struct file *filp, char __user *buf, + size_t count, loff_t *pos) +{ + struct mlx5_ib_delay_drop *delay_drop = filp->private_data; + char lbuf[20]; + int len; + + len = snprintf(lbuf, sizeof(lbuf), "%u\n", delay_drop->timeout); + return simple_read_from_buffer(buf, count, pos, lbuf, len); +} + +static ssize_t delay_drop_timeout_write(struct file *filp, const char __user *buf, + size_t count, loff_t *pos) +{ + struct mlx5_ib_delay_drop *delay_drop = filp->private_data; + u32 timeout; + u32 var; + + if (kstrtouint_from_user(buf, count, 0, &var)) + return -EFAULT; + + timeout = min_t(u32, roundup(var, 100), MLX5_MAX_DELAY_DROP_TIMEOUT_MS * + 1000); + if (timeout != var) + mlx5_ib_dbg(delay_drop->dev, "Round delay drop timeout to %u usec\n", + timeout); + + delay_drop->timeout = timeout; + + return count; +} + +static const struct file_operations fops_delay_drop_timeout = { + .owner = THIS_MODULE, + .open = simple_open, + .write = delay_drop_timeout_write, + .read = delay_drop_timeout_read, +}; + +static int delay_drop_debugfs_init(struct mlx5_ib_dev *dev) +{ + struct mlx5_ib_dbg_delay_drop *dbg; + + if (!mlx5_debugfs_root) + return 0; + + dbg = kzalloc(sizeof(*dbg), GFP_KERNEL); + if (!dbg) + return -ENOMEM; + + dbg->dir_debugfs = + debugfs_create_dir("delay_drop", + dev->mdev->priv.dbg_root); + if (!dbg->dir_debugfs) + return -ENOMEM; + + dbg->events_cnt_debugfs = + debugfs_create_atomic_t("num_timeout_events", 0400, + dbg->dir_debugfs, + &dev->delay_drop.events_cnt); + if (!dbg->events_cnt_debugfs) + goto out_debugfs; + + dbg->rqs_cnt_debugfs = + debugfs_create_atomic_t("num_rqs", 0400, + dbg->dir_debugfs, + &dev->delay_drop.rqs_cnt); + if (!dbg->rqs_cnt_debugfs) + goto out_debugfs; + + dbg->timeout_debugfs = + debugfs_create_file("timeout", 0600, + dbg->dir_debugfs, + &dev->delay_drop, + &fops_delay_drop_timeout); + if (!dbg->timeout_debugfs) + goto out_debugfs; + + return 0; + +out_debugfs: + delay_drop_debugfs_cleanup(dev); + return -ENOMEM; } static void init_delay_drop(struct mlx5_ib_dev *dev) @@ -3669,6 +3767,11 @@ static void init_delay_drop(struct mlx5_ib_dev *dev) dev->delay_drop.activate = false; dev->delay_drop.timeout = MLX5_MAX_DELAY_DROP_TIMEOUT_MS * 1000; INIT_WORK(&dev->delay_drop.delay_drop_work, delay_drop_handler); + atomic_set(&dev->delay_drop.rqs_cnt, 0); + atomic_set(&dev->delay_drop.events_cnt, 0); + + if (delay_drop_debugfs_init(dev)) + mlx5_ib_warn(dev, "Failed to init delay drop debugfs\n"); } static void *mlx5_ib_add(struct mlx5_core_dev *mdev) diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index 097f12dc65a6..0316147a39a8 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -661,6 +661,13 @@ enum { MLX5_MAX_DELAY_DROP_TIMEOUT_MS = 100, }; +struct mlx5_ib_dbg_delay_drop { + struct dentry *dir_debugfs; + struct dentry *rqs_cnt_debugfs; + struct dentry *events_cnt_debugfs; + struct dentry *timeout_debugfs; +}; + struct mlx5_ib_delay_drop { struct mlx5_ib_dev *dev; struct work_struct delay_drop_work; @@ -668,6 +675,9 @@ struct mlx5_ib_delay_drop { struct mutex lock; u32 timeout; bool activate; + atomic_t events_cnt; + atomic_t rqs_cnt; + struct mlx5_ib_dbg_delay_drop *dbg; }; struct mlx5_ib_dev { diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c index 939553d5c25f..c5d8ec839e99 100644 --- a/drivers/infiniband/hw/mlx5/qp.c +++ b/drivers/infiniband/hw/mlx5/qp.c @@ -675,10 +675,14 @@ static int mlx5_ib_umem_get(struct mlx5_ib_dev *dev, return err; } -static void destroy_user_rq(struct ib_pd *pd, struct mlx5_ib_rwq *rwq) +static void destroy_user_rq(struct mlx5_ib_dev *dev, struct ib_pd *pd, + struct mlx5_ib_rwq *rwq) { struct mlx5_ib_ucontext *context; + if (rwq->create_flags & MLX5_IB_WQ_FLAGS_DELAY_DROP) + atomic_dec(&dev->delay_drop.rqs_cnt); + context = to_mucontext(pd->uobject->context); mlx5_ib_db_unmap_user(context, &rwq->db); if (rwq->umem) @@ -4612,6 +4616,9 @@ static int set_delay_drop(struct mlx5_ib_dev *dev) dev->delay_drop.activate = true; out: mutex_unlock(&dev->delay_drop.lock); + + if (!err) + atomic_inc(&dev->delay_drop.rqs_cnt); return err; } @@ -4824,7 +4831,7 @@ struct ib_wq *mlx5_ib_create_wq(struct ib_pd *pd, err_copy: mlx5_core_destroy_rq_tracked(dev->mdev, &rwq->core_qp); err_user_rq: - destroy_user_rq(pd, rwq); + destroy_user_rq(dev, pd, rwq); err: kfree(rwq); return ERR_PTR(err); @@ -4836,7 +4843,7 @@ int mlx5_ib_destroy_wq(struct ib_wq *wq) struct mlx5_ib_rwq *rwq = to_mrwq(wq); mlx5_core_destroy_rq_tracked(dev->mdev, &rwq->core_qp); - destroy_user_rq(wq->pd, rwq); + destroy_user_rq(dev, wq->pd, rwq); kfree(rwq); return 0; From 5236333592244557a19694a51337df6ac018f0a7 Mon Sep 17 00:00:00 2001 From: Noa Osherovich Date: Mon, 12 Jun 2017 11:14:02 +0300 Subject: [PATCH 17/34] IB/core: Fix the validations of a multicast LID in attach or detach operations RoCE Annex (A16.9.10/11) declares that during attach (detach) QP to a multicast group, if the QP is associated with a RoCE port, the multicast group MLID is unused and is ignored. During attach or detach multicast, when the QP is associated with a port, it is enough to check the port's link layer and validate the LID only if it is Infiniband. Otherwise, avoid validating the multicast LID. Fixes: 8561eae60ff9 ("IB/core: For multicast functions, verify that LIDs are multicast LIDs") Signed-off-by: Noa Osherovich Reviewed-by: Moni Shoua Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/core/verbs.c | 44 ++++++++++++++++++++++++++++++--- 1 file changed, 40 insertions(+), 4 deletions(-) diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c index fb98ed67d5bc..802bdc397a57 100644 --- a/drivers/infiniband/core/verbs.c +++ b/drivers/infiniband/core/verbs.c @@ -1569,6 +1569,44 @@ EXPORT_SYMBOL(ib_dealloc_fmr); /* Multicast groups */ +static bool is_valid_mcast_lid(struct ib_qp *qp, u16 lid) +{ + struct ib_qp_init_attr init_attr = {}; + struct ib_qp_attr attr = {}; + int num_eth_ports = 0; + int port; + + /* If QP state >= init, it is assigned to a port and we can check this + * port only. + */ + if (!ib_query_qp(qp, &attr, IB_QP_STATE | IB_QP_PORT, &init_attr)) { + if (attr.qp_state >= IB_QPS_INIT) { + if (qp->device->get_link_layer(qp->device, attr.port_num) != + IB_LINK_LAYER_INFINIBAND) + return true; + goto lid_check; + } + } + + /* Can't get a quick answer, iterate over all ports */ + for (port = 0; port < qp->device->phys_port_cnt; port++) + if (qp->device->get_link_layer(qp->device, port) != + IB_LINK_LAYER_INFINIBAND) + num_eth_ports++; + + /* If we have at lease one Ethernet port, RoCE annex declares that + * multicast LID should be ignored. We can't tell at this step if the + * QP belongs to an IB or Ethernet port. + */ + if (num_eth_ports) + return true; + + /* If all the ports are IB, we can check according to IB spec. */ +lid_check: + return !(lid < be16_to_cpu(IB_MULTICAST_LID_BASE) || + lid == be16_to_cpu(IB_LID_PERMISSIVE)); +} + int ib_attach_mcast(struct ib_qp *qp, union ib_gid *gid, u16 lid) { int ret; @@ -1576,8 +1614,7 @@ int ib_attach_mcast(struct ib_qp *qp, union ib_gid *gid, u16 lid) if (!qp->device->attach_mcast) return -ENOSYS; if (gid->raw[0] != 0xff || qp->qp_type != IB_QPT_UD || - lid < be16_to_cpu(IB_MULTICAST_LID_BASE) || - lid == be16_to_cpu(IB_LID_PERMISSIVE)) + !is_valid_mcast_lid(qp, lid)) return -EINVAL; ret = qp->device->attach_mcast(qp, gid, lid); @@ -1594,8 +1631,7 @@ int ib_detach_mcast(struct ib_qp *qp, union ib_gid *gid, u16 lid) if (!qp->device->detach_mcast) return -ENOSYS; if (gid->raw[0] != 0xff || qp->qp_type != IB_QPT_UD || - lid < be16_to_cpu(IB_MULTICAST_LID_BASE) || - lid == be16_to_cpu(IB_LID_PERMISSIVE)) + !is_valid_mcast_lid(qp, lid)) return -EINVAL; ret = qp->device->detach_mcast(qp, gid, lid); From be1d325a335840a86c133a56c6a911c368bac0fd Mon Sep 17 00:00:00 2001 From: Noa Osherovich Date: Mon, 12 Jun 2017 11:14:03 +0300 Subject: [PATCH 18/34] IB/core: Set RoCEv2 MGID according to spec RoCEv2 Annex states that for RoCEv2 over IPv4, the corresponding IPv4 address is encoded into the GID according to the following rule: GID= :ffff: Remove the 0xff0e prefix for RoCEv2 packets with IPv4 and leave it zeroed and change rdma_is_multicast_addr() to consider the new logic. Signed-off-by: Noa Osherovich Reviewed-by: Moni Shoua Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/core/cma.c | 13 +++++++------ drivers/infiniband/core/verbs.c | 10 ++++++---- include/rdma/ib_addr.h | 8 +++++++- 3 files changed, 20 insertions(+), 11 deletions(-) diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c index 0eb393237ba2..a8c2f0ccd225 100644 --- a/drivers/infiniband/core/cma.c +++ b/drivers/infiniband/core/cma.c @@ -3998,7 +3998,8 @@ static void iboe_mcast_work_handler(struct work_struct *work) kfree(mw); } -static void cma_iboe_set_mgid(struct sockaddr *addr, union ib_gid *mgid) +static void cma_iboe_set_mgid(struct sockaddr *addr, union ib_gid *mgid, + enum ib_gid_type gid_type) { struct sockaddr_in *sin = (struct sockaddr_in *)addr; struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)addr; @@ -4008,8 +4009,8 @@ static void cma_iboe_set_mgid(struct sockaddr *addr, union ib_gid *mgid) } else if (addr->sa_family == AF_INET6) { memcpy(mgid, &sin6->sin6_addr, sizeof *mgid); } else { - mgid->raw[0] = 0xff; - mgid->raw[1] = 0x0e; + mgid->raw[0] = (gid_type == IB_GID_TYPE_IB) ? 0xff : 0; + mgid->raw[1] = (gid_type == IB_GID_TYPE_IB) ? 0x0e : 0; mgid->raw[2] = 0; mgid->raw[3] = 0; mgid->raw[4] = 0; @@ -4050,7 +4051,9 @@ static int cma_iboe_join_multicast(struct rdma_id_private *id_priv, goto out1; } - cma_iboe_set_mgid(addr, &mc->multicast.ib->rec.mgid); + gid_type = id_priv->cma_dev->default_gid_type[id_priv->id.port_num - + rdma_start_port(id_priv->cma_dev->device)]; + cma_iboe_set_mgid(addr, &mc->multicast.ib->rec.mgid, gid_type); mc->multicast.ib->rec.pkey = cpu_to_be16(0xffff); if (id_priv->id.ps == RDMA_PS_UDP) @@ -4066,8 +4069,6 @@ static int cma_iboe_join_multicast(struct rdma_id_private *id_priv, mc->multicast.ib->rec.hop_limit = 1; mc->multicast.ib->rec.mtu = iboe_get_mtu(ndev->mtu); - gid_type = id_priv->cma_dev->default_gid_type[id_priv->id.port_num - - rdma_start_port(id_priv->cma_dev->device)]; if (addr->sa_family == AF_INET) { if (gid_type == IB_GID_TYPE_ROCE_UDP_ENCAP) { mc->multicast.ib->rec.hop_limit = IPV6_DEFAULT_HOPLIMIT; diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c index 802bdc397a57..30fdc3ae1bbd 100644 --- a/drivers/infiniband/core/verbs.c +++ b/drivers/infiniband/core/verbs.c @@ -1613,8 +1613,9 @@ int ib_attach_mcast(struct ib_qp *qp, union ib_gid *gid, u16 lid) if (!qp->device->attach_mcast) return -ENOSYS; - if (gid->raw[0] != 0xff || qp->qp_type != IB_QPT_UD || - !is_valid_mcast_lid(qp, lid)) + + 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); @@ -1630,8 +1631,9 @@ int ib_detach_mcast(struct ib_qp *qp, union ib_gid *gid, u16 lid) if (!qp->device->detach_mcast) return -ENOSYS; - if (gid->raw[0] != 0xff || qp->qp_type != IB_QPT_UD || - !is_valid_mcast_lid(qp, lid)) + + 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); diff --git a/include/rdma/ib_addr.h b/include/rdma/ib_addr.h index b73a14edc85e..7aca12188ef3 100644 --- a/include/rdma/ib_addr.h +++ b/include/rdma/ib_addr.h @@ -304,7 +304,13 @@ static inline void rdma_get_ll_mac(struct in6_addr *addr, u8 *mac) static inline int rdma_is_multicast_addr(struct in6_addr *addr) { - return addr->s6_addr[0] == 0xff; + u32 ipv4_addr; + + if (addr->s6_addr[0] == 0xff) + return 1; + + memcpy(&ipv4_addr, addr->s6_addr + 12, 4); + return (ipv6_addr_v4mapped(addr) && ipv4_is_multicast(ipv4_addr)); } static inline void rdma_get_mcast_mac(struct in6_addr *addr, u8 *mac) From 9636a56fa864464896bf7d1272c701f2b9a57737 Mon Sep 17 00:00:00 2001 From: Noa Osherovich Date: Mon, 12 Jun 2017 11:14:04 +0300 Subject: [PATCH 19/34] IB/core: Add support for RoCEv2 multicast When creating address handle from multicast GID, set MAC according to the appropriate formula instead of searching for it in the GID table: - For IPv4 multicast GID use ip_eth_mc_map(). - For IPv6 multicast GID use ipv6_eth_mc_map(). Signed-off-by: Noa Osherovich Reviewed-by: Moni Shoua Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/core/verbs.c | 12 ++++++++++++ 1 file changed, 12 insertions(+) diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c index 30fdc3ae1bbd..e8006677b01c 100644 --- a/drivers/infiniband/core/verbs.c +++ b/drivers/infiniband/core/verbs.c @@ -1244,6 +1244,18 @@ int ib_resolve_eth_dmac(struct ib_device *device, if (rdma_link_local_addr((struct in6_addr *)grh->dgid.raw)) { rdma_get_ll_mac((struct in6_addr *)grh->dgid.raw, ah_attr->roce.dmac); + return 0; + } + if (rdma_is_multicast_addr((struct in6_addr *)ah_attr->grh.dgid.raw)) { + if (ipv6_addr_v4mapped((struct in6_addr *)ah_attr->grh.dgid.raw)) { + __be32 addr = 0; + + memcpy(&addr, ah_attr->grh.dgid.raw + 12, 4); + ip_eth_mc_map(addr, (char *)ah_attr->roce.dmac); + } else { + ipv6_eth_mc_map((struct in6_addr *)ah_attr->grh.dgid.raw, + (char *)ah_attr->roce.dmac); + } } else { union ib_gid sgid; struct ib_gid_attr sgid_attr; From 02984cc7b3d62418bd72abacaf875c3a9eccdb66 Mon Sep 17 00:00:00 2001 From: Yishai Hadas Date: Thu, 8 Jun 2017 16:15:06 +0300 Subject: [PATCH 20/34] IB/core: Enable QP creation with a given source QP number Enable QP creation with a given source QP number. The created QP will use the source QPN as its wire QP number. This comes as a pre-patch for downstream patches in this series to allow user space applications to accelerate traffic which is typically handled by IPoIB ULP. Signed-off-by: Yishai Hadas Reviewed-by: Maor Gottlieb Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- include/rdma/ib_verbs.h | 2 ++ 1 file changed, 2 insertions(+) diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h index 4a1444456af7..76a74c783c50 100644 --- a/include/rdma/ib_verbs.h +++ b/include/rdma/ib_verbs.h @@ -1059,6 +1059,7 @@ enum ib_qp_create_flags { /* FREE = 1 << 7, */ IB_QP_CREATE_SCATTER_FCS = 1 << 8, IB_QP_CREATE_CVLAN_STRIPPING = 1 << 9, + IB_QP_CREATE_SOURCE_QPN = 1 << 10, /* reserve bits 26-31 for low level drivers' internal use */ IB_QP_CREATE_RESERVED_START = 1 << 26, IB_QP_CREATE_RESERVED_END = 1 << 31, @@ -1086,6 +1087,7 @@ struct ib_qp_init_attr { */ u8 port_num; struct ib_rwq_ind_table *rwq_ind_tbl; + u32 source_qpn; }; struct ib_qp_open_attr { From 2dee0e545894c23b1a2cc2019ac87dffb42e5984 Mon Sep 17 00:00:00 2001 From: Yishai Hadas Date: Thu, 8 Jun 2017 16:15:07 +0300 Subject: [PATCH 21/34] IB/uverbs: Enable QP creation with a given source QP number Enable QP creation with a given source QP number, the created QP will use the source QPN as its wire QP number. To create such a QP, root privileges (i.e. CAP_NET_RAW) are required from the user application. This comes as a pre-patch for downstream patches in this series to allow user space applications to accelerate traffic which is typically handled by IPoIB ULP. Signed-off-by: Yishai Hadas Reviewed-by: Maor Gottlieb Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/core/uverbs_cmd.c | 17 ++++++++++++++--- include/uapi/rdma/ib_user_verbs.h | 2 +- 2 files changed, 15 insertions(+), 4 deletions(-) diff --git a/drivers/infiniband/core/uverbs_cmd.c b/drivers/infiniband/core/uverbs_cmd.c index 2c98533a0203..60535c754db3 100644 --- a/drivers/infiniband/core/uverbs_cmd.c +++ b/drivers/infiniband/core/uverbs_cmd.c @@ -1383,8 +1383,9 @@ static int create_qp(struct ib_uverbs_file *file, attr.rwq_ind_tbl = ind_tbl; } - if ((cmd_sz >= offsetof(typeof(*cmd), reserved1) + - sizeof(cmd->reserved1)) && cmd->reserved1) { + if (cmd_sz > sizeof(*cmd) && + !ib_is_udata_cleared(ucore, sizeof(*cmd), + cmd_sz - sizeof(*cmd))) { ret = -EOPNOTSUPP; goto err_put; } @@ -1482,11 +1483,21 @@ static int create_qp(struct ib_uverbs_file *file, IB_QP_CREATE_MANAGED_SEND | IB_QP_CREATE_MANAGED_RECV | IB_QP_CREATE_SCATTER_FCS | - IB_QP_CREATE_CVLAN_STRIPPING)) { + IB_QP_CREATE_CVLAN_STRIPPING | + IB_QP_CREATE_SOURCE_QPN)) { ret = -EINVAL; goto err_put; } + if (attr.create_flags & IB_QP_CREATE_SOURCE_QPN) { + if (!capable(CAP_NET_RAW)) { + ret = -EPERM; + goto err_put; + } + + attr.source_qpn = cmd->source_qpn; + } + buf = (void *)cmd + sizeof(*cmd); if (cmd_sz > sizeof(*cmd)) if (!(buf[0] == 0 && !memcmp(buf, buf + 1, diff --git a/include/uapi/rdma/ib_user_verbs.h b/include/uapi/rdma/ib_user_verbs.h index 270c350bedc6..63656d2e8705 100644 --- a/include/uapi/rdma/ib_user_verbs.h +++ b/include/uapi/rdma/ib_user_verbs.h @@ -578,7 +578,7 @@ struct ib_uverbs_ex_create_qp { __u32 comp_mask; __u32 create_flags; __u32 rwq_ind_tbl_handle; - __u32 reserved1; + __u32 source_qpn; }; struct ib_uverbs_open_qp { From c2e53b2ce1ba351918ede492c0cb207f42e1228f Mon Sep 17 00:00:00 2001 From: Yishai Hadas Date: Thu, 8 Jun 2017 16:15:08 +0300 Subject: [PATCH 22/34] IB/mlx5: Add support for QP with a given source QPN Allow user space applications to accelerate send and receive traffic which is typically handled by IPoIB ULP by creating a UD QP with a given source QPN of the IPoIB UD QP. UD QP with a given source QPN should basically be similar to RAW QP from point of view of its created resources. However, - Its TIS should point to the source QPN. - Modify can be done only on its state as the transport attributes are managed by its source QP. This patch manages below: - Creating/destroying/modifying UD QP with a given source QPN. Signed-off-by: Yishai Hadas Reviewed-by: Maor Gottlieb Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx5/mlx5_ib.h | 2 + drivers/infiniband/hw/mlx5/qp.c | 72 ++++++++++++++++++++++------ 2 files changed, 59 insertions(+), 15 deletions(-) diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index 0316147a39a8..7ac991070020 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -383,6 +383,7 @@ struct mlx5_ib_qp { struct list_head cq_recv_list; struct list_head cq_send_list; u32 rate_limit; + u32 underlay_qpn; }; struct mlx5_ib_cq_buf { @@ -404,6 +405,7 @@ enum mlx5_ib_qp_flags { MLX5_IB_QP_CAP_SCATTER_FCS = 1 << 7, MLX5_IB_QP_RSS = 1 << 8, MLX5_IB_QP_CVLAN_STRIPPING = 1 << 9, + MLX5_IB_QP_UNDERLAY = 1 << 10, }; struct mlx5_umr_wr { diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c index c5d8ec839e99..5c7ce9bd466e 100644 --- a/drivers/infiniband/hw/mlx5/qp.c +++ b/drivers/infiniband/hw/mlx5/qp.c @@ -34,6 +34,7 @@ #include #include #include +#include #include "mlx5_ib.h" /* not supported currently */ @@ -453,7 +454,8 @@ static int set_user_buf_size(struct mlx5_ib_dev *dev, return -EINVAL; } - if (attr->qp_type == IB_QPT_RAW_PACKET) { + if (attr->qp_type == IB_QPT_RAW_PACKET || + qp->flags & MLX5_IB_QP_UNDERLAY) { base->ubuffer.buf_size = qp->rq.wqe_cnt << qp->rq.wqe_shift; qp->raw_packet_qp.sq.ubuffer.buf_size = qp->sq.wqe_cnt << 6; } else { @@ -1025,12 +1027,16 @@ static int is_connected(enum ib_qp_type qp_type) } static int create_raw_packet_qp_tis(struct mlx5_ib_dev *dev, + struct mlx5_ib_qp *qp, struct mlx5_ib_sq *sq, u32 tdn) { u32 in[MLX5_ST_SZ_DW(create_tis_in)] = {0}; void *tisc = MLX5_ADDR_OF(create_tis_in, in, ctx); MLX5_SET(tisc, tisc, transport_domain, tdn); + if (qp->flags & MLX5_IB_QP_UNDERLAY) + MLX5_SET(tisc, tisc, underlay_qpn, qp->underlay_qpn); + return mlx5_core_create_tis(dev->mdev, in, sizeof(in), &sq->tisn); } @@ -1233,7 +1239,7 @@ static int create_raw_packet_qp(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp, u32 tdn = mucontext->tdn; if (qp->sq.wqe_cnt) { - err = create_raw_packet_qp_tis(dev, sq, tdn); + err = create_raw_packet_qp_tis(dev, qp, sq, tdn); if (err) return err; @@ -1506,10 +1512,6 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd, u32 *in; int err; - base = init_attr->qp_type == IB_QPT_RAW_PACKET ? - &qp->raw_packet_qp.rq.base : - &qp->trans_qp.base; - mutex_init(&qp->mutex); spin_lock_init(&qp->sq.lock); spin_lock_init(&qp->rq.lock); @@ -1591,10 +1593,28 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd, qp->wq_sig = !!(ucmd.flags & MLX5_QP_FLAG_SIGNATURE); qp->scat_cqe = !!(ucmd.flags & MLX5_QP_FLAG_SCATTER_CQE); + + 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) != + MLX5_CAP_PORT_TYPE_IB) || + !mlx5_get_flow_namespace(dev->mdev, MLX5_FLOW_NAMESPACE_BYPASS)) { + mlx5_ib_dbg(dev, "Source QP option isn't supported\n"); + return -EOPNOTSUPP; + } + + qp->flags |= MLX5_IB_QP_UNDERLAY; + qp->underlay_qpn = init_attr->source_qpn; + } } else { qp->wq_sig = !!wq_signature; } + base = (init_attr->qp_type == IB_QPT_RAW_PACKET || + qp->flags & MLX5_IB_QP_UNDERLAY) ? + &qp->raw_packet_qp.rq.base : + &qp->trans_qp.base; + 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); @@ -1745,7 +1765,8 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd, qp->flags |= MLX5_IB_QP_LSO; } - if (init_attr->qp_type == IB_QPT_RAW_PACKET) { + if (init_attr->qp_type == IB_QPT_RAW_PACKET || + qp->flags & MLX5_IB_QP_UNDERLAY) { qp->raw_packet_qp.sq.ubuffer.buf_addr = ucmd.sq_buf_addr; raw_packet_qp_copy_info(qp, &qp->raw_packet_qp); err = create_raw_packet_qp(dev, qp, in, pd); @@ -1897,7 +1918,7 @@ static int modify_raw_packet_qp(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp, static void destroy_qp_common(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp) { struct mlx5_ib_cq *send_cq, *recv_cq; - struct mlx5_ib_qp_base *base = &qp->trans_qp.base; + struct mlx5_ib_qp_base *base; unsigned long flags; int err; @@ -1906,12 +1927,14 @@ static void destroy_qp_common(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp) return; } - base = qp->ibqp.qp_type == IB_QPT_RAW_PACKET ? + base = (qp->ibqp.qp_type == IB_QPT_RAW_PACKET || + qp->flags & MLX5_IB_QP_UNDERLAY) ? &qp->raw_packet_qp.rq.base : &qp->trans_qp.base; if (qp->state != IB_QPS_RESET) { - if (qp->ibqp.qp_type != IB_QPT_RAW_PACKET) { + if (qp->ibqp.qp_type != IB_QPT_RAW_PACKET && + !(qp->flags & MLX5_IB_QP_UNDERLAY)) { err = mlx5_core_qp_modify(dev->mdev, MLX5_CMD_OP_2RST_QP, 0, NULL, &base->mqp); @@ -1950,7 +1973,8 @@ static void destroy_qp_common(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp) mlx5_ib_unlock_cqs(send_cq, recv_cq); spin_unlock_irqrestore(&dev->reset_flow_resource_lock, flags); - if (qp->ibqp.qp_type == IB_QPT_RAW_PACKET) { + if (qp->ibqp.qp_type == IB_QPT_RAW_PACKET || + qp->flags & MLX5_IB_QP_UNDERLAY) { destroy_raw_packet_qp(dev, qp); } else { err = mlx5_core_destroy_qp(dev->mdev, &base->mqp); @@ -2706,7 +2730,8 @@ static int __mlx5_ib_modify_qp(struct ib_qp *ibqp, if (is_sqp(ibqp->qp_type)) { context->mtu_msgmax = (IB_MTU_256 << 5) | 8; - } else if (ibqp->qp_type == IB_QPT_UD || + } else if ((ibqp->qp_type == IB_QPT_UD && + !(qp->flags & MLX5_IB_QP_UNDERLAY)) || ibqp->qp_type == MLX5_IB_QPT_REG_UMR) { context->mtu_msgmax = (IB_MTU_4096 << 5) | 12; } else if (attr_mask & IB_QP_PATH_MTU) { @@ -2803,6 +2828,11 @@ static int __mlx5_ib_modify_qp(struct ib_qp *ibqp, if (cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT) { u8 port_num = (attr_mask & IB_QP_PORT ? attr->port_num : qp->port) - 1; + + /* Underlay port should be used - index 0 function per port */ + if (qp->flags & MLX5_IB_QP_UNDERLAY) + port_num = 0; + mibport = &dev->port[port_num]; context->qp_counter_set_usr_page |= cpu_to_be32((u32)(mibport->cnts.set_id) << 24); @@ -2828,7 +2858,8 @@ static int __mlx5_ib_modify_qp(struct ib_qp *ibqp, optpar = ib_mask_to_mlx5_opt(attr_mask); optpar &= opt_mask[mlx5_cur][mlx5_new][mlx5_st]; - if (qp->ibqp.qp_type == IB_QPT_RAW_PACKET) { + if (qp->ibqp.qp_type == IB_QPT_RAW_PACKET || + qp->flags & MLX5_IB_QP_UNDERLAY) { struct mlx5_modify_raw_qp_param raw_qp_param = {}; raw_qp_param.operation = op; @@ -2917,7 +2948,13 @@ int mlx5_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, ll = dev->ib_dev.get_link_layer(&dev->ib_dev, port); } - if (qp_type != MLX5_IB_QPT_REG_UMR && + if (qp->flags & MLX5_IB_QP_UNDERLAY) { + if (attr_mask & ~(IB_QP_STATE | IB_QP_CUR_STATE)) { + mlx5_ib_dbg(dev, "invalid attr_mask 0x%x when underlay QP is used\n", + attr_mask); + goto out; + } + } else if (qp_type != MLX5_IB_QPT_REG_UMR && !ib_modify_qp_is_ok(cur_state, new_state, qp_type, attr_mask, ll)) { mlx5_ib_dbg(dev, "invalid QP state transition from %d to %d, qp_type %d, attr_mask 0x%x\n", cur_state, new_state, ibqp->qp_type, attr_mask); @@ -4481,9 +4518,14 @@ int mlx5_ib_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *qp_attr, return mlx5_ib_gsi_query_qp(ibqp, qp_attr, qp_attr_mask, qp_init_attr); + /* Not all of output fields are applicable, make sure to zero them */ + memset(qp_init_attr, 0, sizeof(*qp_init_attr)); + memset(qp_attr, 0, sizeof(*qp_attr)); + mutex_lock(&qp->mutex); - if (qp->ibqp.qp_type == IB_QPT_RAW_PACKET) { + if (qp->ibqp.qp_type == IB_QPT_RAW_PACKET || + qp->flags & MLX5_IB_QP_UNDERLAY) { err = query_raw_packet_qp_state(dev, qp, &raw_packet_qp_state); if (err) goto out; From 81e308804ba6b389ead57b54a259154a36a560a8 Mon Sep 17 00:00:00 2001 From: Yishai Hadas Date: Thu, 8 Jun 2017 16:15:09 +0300 Subject: [PATCH 23/34] IB/mlx5: Add multicast flow steering support for underlay QP In order to add multicast flow steering support, there is need to block the attaching of mcg flow for underlay QP, recognize multicast IB_FLOW_SPEC_IPV4 based on its IP and enable creating/destroying flow for IB layer. Signed-off-by: Yishai Hadas Reviewed-by: Maor Gottlieb Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx5/main.c | 45 ++++++++++++++++++++++--------- 1 file changed, 32 insertions(+), 13 deletions(-) diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index 7455f95bf679..bf74b2ee2f9c 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -2102,21 +2102,32 @@ static int parse_flow_attr(struct mlx5_core_dev *mdev, u32 *match_c, */ static bool flow_is_multicast_only(struct ib_flow_attr *ib_attr) { - struct ib_flow_spec_eth *eth_spec; + union ib_flow_spec *flow_spec; if (ib_attr->type != IB_FLOW_ATTR_NORMAL || - ib_attr->size < sizeof(struct ib_flow_attr) + - sizeof(struct ib_flow_spec_eth) || ib_attr->num_of_specs < 1) return false; - eth_spec = (struct ib_flow_spec_eth *)(ib_attr + 1); - if (eth_spec->type != IB_FLOW_SPEC_ETH || - eth_spec->size != sizeof(*eth_spec)) - return false; + flow_spec = (union ib_flow_spec *)(ib_attr + 1); + if (flow_spec->type == IB_FLOW_SPEC_IPV4) { + struct ib_flow_spec_ipv4 *ipv4_spec; - return is_multicast_ether_addr(eth_spec->mask.dst_mac) && - is_multicast_ether_addr(eth_spec->val.dst_mac); + ipv4_spec = (struct ib_flow_spec_ipv4 *)flow_spec; + if (ipv4_is_multicast(ipv4_spec->val.dst_ip)) + return true; + + return false; + } + + if (flow_spec->type == IB_FLOW_SPEC_ETH) { + struct ib_flow_spec_eth *eth_spec; + + eth_spec = (struct ib_flow_spec_eth *)flow_spec; + return is_multicast_ether_addr(eth_spec->mask.dst_mac) && + is_multicast_ether_addr(eth_spec->val.dst_mac); + } + + return false; } static bool is_valid_ethertype(struct mlx5_core_dev *mdev, @@ -2594,8 +2605,14 @@ static struct ib_flow *mlx5_ib_create_flow(struct ib_qp *qp, static int mlx5_ib_mcg_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid) { struct mlx5_ib_dev *dev = to_mdev(ibqp->device); + struct mlx5_ib_qp *mqp = to_mqp(ibqp); int err; + if (mqp->flags & MLX5_IB_QP_UNDERLAY) { + mlx5_ib_dbg(dev, "Attaching a multi cast group to underlay QP is not supported\n"); + return -EOPNOTSUPP; + } + err = mlx5_core_attach_mcg(dev->mdev, gid, ibqp->qp_num); if (err) mlx5_ib_warn(dev, "failed attaching QPN 0x%x, MGID %pI6\n", @@ -3941,18 +3958,20 @@ static void *mlx5_ib_add(struct mlx5_core_dev *mdev) (1ull << IB_USER_VERBS_CMD_CLOSE_XRCD); } + 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_ib_port_link_layer(&dev->ib_dev, 1) == IB_LINK_LAYER_ETHERNET) { - dev->ib_dev.create_flow = mlx5_ib_create_flow; - dev->ib_dev.destroy_flow = mlx5_ib_destroy_flow; 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_FLOW) | - (1ull << IB_USER_VERBS_EX_CMD_DESTROY_FLOW) | (1ull << IB_USER_VERBS_EX_CMD_CREATE_WQ) | (1ull << IB_USER_VERBS_EX_CMD_MODIFY_WQ) | (1ull << IB_USER_VERBS_EX_CMD_DESTROY_WQ) | From 4ce749bd94f697772ac2be4fb7e7a92726e94bfb Mon Sep 17 00:00:00 2001 From: Yishai Hadas Date: Thu, 8 Jun 2017 16:15:10 +0300 Subject: [PATCH 24/34] net/mlx5: Report enhanced capabilities for IPoIB Report 'ipoib_enhanced_offloads' capabilities from the core layer, it will be used in the next patch from this series. Signed-off-by: Yishai Hadas Reviewed-by: Maor Gottlieb Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/net/ethernet/mellanox/mlx5/core/fw.c | 6 ++++++ include/linux/mlx5/device.h | 6 +++++- 2 files changed, 11 insertions(+), 1 deletion(-) diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fw.c b/drivers/net/ethernet/mellanox/mlx5/core/fw.c index fa33d59ab485..2c71557d1cee 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/fw.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/fw.c @@ -120,6 +120,12 @@ int mlx5_query_hca_caps(struct mlx5_core_dev *dev) return err; } + if (MLX5_CAP_GEN(dev, ipoib_enhanced_offloads)) { + err = mlx5_core_get_caps(dev, MLX5_CAP_IPOIB_ENHANCED_OFFLOADS); + if (err) + return err; + } + if (MLX5_CAP_GEN(dev, pg)) { err = mlx5_core_get_caps(dev, MLX5_CAP_ODP); if (err) diff --git a/include/linux/mlx5/device.h b/include/linux/mlx5/device.h index a47b9ab9f2c9..c13d71deaeca 100644 --- a/include/linux/mlx5/device.h +++ b/include/linux/mlx5/device.h @@ -973,7 +973,7 @@ enum mlx5_cap_type { MLX5_CAP_ATOMIC, MLX5_CAP_ROCE, MLX5_CAP_IPOIB_OFFLOADS, - MLX5_CAP_EOIB_OFFLOADS, + MLX5_CAP_IPOIB_ENHANCED_OFFLOADS, MLX5_CAP_FLOW_TABLE, MLX5_CAP_ESWITCH_FLOW_TABLE, MLX5_CAP_ESWITCH, @@ -1016,6 +1016,10 @@ enum mlx5_mcam_feature_groups { MLX5_GET(per_protocol_networking_offload_caps,\ mdev->caps.hca_max[MLX5_CAP_ETHERNET_OFFLOADS], cap) +#define MLX5_CAP_IPOIB_ENHANCED(mdev, cap) \ + MLX5_GET(per_protocol_networking_offload_caps,\ + mdev->caps.hca_cur[MLX5_CAP_IPOIB_ENHANCED_OFFLOADS], cap) + #define MLX5_CAP_ROCE(mdev, cap) \ MLX5_GET(roce_cap, mdev->caps.hca_cur[MLX5_CAP_ROCE], cap) From 1d54f89094e2d2067948550ec52aab93e15b4b0c Mon Sep 17 00:00:00 2001 From: Yishai Hadas Date: Thu, 8 Jun 2017 16:15:11 +0300 Subject: [PATCH 25/34] IB/mlx5: Report RX checksum capabilities for IPoIB Report RX checksum capabilities when 'ipoib_enhanced_offloads' capabilities are set and support checksum. Signed-off-by: Yishai Hadas Reviewed-by: Maor Gottlieb Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx5/main.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index bf74b2ee2f9c..2757d445b042 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -702,6 +702,10 @@ static int mlx5_ib_query_device(struct ib_device *ibdev, MLX5_CAP_GEN(dev->mdev, general_notification_event)) props->raw_packet_caps |= IB_RAW_PACKET_CAP_DELAY_DROP; + if (MLX5_CAP_GEN(mdev, ipoib_enhanced_offloads) && + MLX5_CAP_IPOIB_ENHANCED(mdev, csum_cap)) + props->device_cap_flags |= IB_DEVICE_UD_IP_CSUM; + if (MLX5_CAP_GEN(dev->mdev, eth_net_offloads) && MLX5_CAP_ETH(dev->mdev, scatter_fcs)) { /* Legacy bit to support old userspace libraries */ From 4c25b7a39005c9243a492b577c3e940eeac36a25 Mon Sep 17 00:00:00 2001 From: Majd Dibbiny Date: Mon, 12 Jun 2017 10:36:15 +0300 Subject: [PATCH 26/34] IB/mlx5: Fix cached MR allocation flow When we have a miss in one order of the mkey cache, we try to get an mkey from a higher order. We still need to check that the higher order can be used with UMR before using it. Otherwise, we will get an mkey with 0 entries and the post send operation that is used to fill it will complete with the following error: mlx5_0:dump_cqe:275:(pid 0): dump error cqe 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 0f007806 25000025 49ce59d2 Fixes: 49780d42dfc9 ("IB/mlx5: Expose MR cache for mlx5_ib") Cc: # v4.10+ Signed-off-by: Majd Dibbiny Reviewed-by: Ilya Lesokhin Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx5/mr.c | 18 +++++++++++++----- 1 file changed, 13 insertions(+), 5 deletions(-) diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c index 2c40a2e989d2..a0eb2f96179a 100644 --- a/drivers/infiniband/hw/mlx5/mr.c +++ b/drivers/infiniband/hw/mlx5/mr.c @@ -48,6 +48,7 @@ enum { #define MLX5_UMR_ALIGN 2048 static int clean_mr(struct mlx5_ib_mr *mr); +static int max_umr_order(struct mlx5_ib_dev *dev); static int use_umr(struct mlx5_ib_dev *dev, int order); static int unreg_umr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr); @@ -491,16 +492,18 @@ static struct mlx5_ib_mr *alloc_cached_mr(struct mlx5_ib_dev *dev, int order) struct mlx5_mr_cache *cache = &dev->cache; struct mlx5_ib_mr *mr = NULL; struct mlx5_cache_ent *ent; + int last_umr_cache_entry; int c; int i; c = order2idx(dev, order); - if (c < 0 || c > MAX_UMR_CACHE_ENTRY) { + last_umr_cache_entry = order2idx(dev, max_umr_order(dev)); + if (c < 0 || c > last_umr_cache_entry) { mlx5_ib_warn(dev, "order %d, cache index %d\n", order, c); return NULL; } - for (i = c; i < MAX_UMR_CACHE_ENTRY; i++) { + for (i = c; i <= last_umr_cache_entry; i++) { ent = &cache->ent[i]; mlx5_ib_dbg(dev, "order %d, cache index %d\n", ent->order, i); @@ -816,11 +819,16 @@ static int get_octo_len(u64 addr, u64 len, int page_size) return (npages + 1) / 2; } -static int use_umr(struct mlx5_ib_dev *dev, int order) +static int max_umr_order(struct mlx5_ib_dev *dev) { if (MLX5_CAP_GEN(dev->mdev, umr_extended_translation_offset)) - return order <= MAX_UMR_CACHE_ENTRY + 2; - return order <= MLX5_MAX_UMR_SHIFT; + return MAX_UMR_CACHE_ENTRY + 2; + return MLX5_MAX_UMR_SHIFT; +} + +static int use_umr(struct mlx5_ib_dev *dev, int order) +{ + return order <= max_umr_order(dev); } static int mr_umem_get(struct ib_pd *pd, u64 start, u64 length, From 3fffc82ad6c78fcc9d5d4eca089f00db14ab0358 Mon Sep 17 00:00:00 2001 From: Leon Romanovsky Date: Mon, 12 Jun 2017 10:36:16 +0300 Subject: [PATCH 27/34] IB/mlx5: Fix existence check for extended address vector The extended address vector is the highest bit in be32 variable, but it was compared with the lowest. This patch fixes the endianness of that check and removes already declared define. Fixes: 17d2f88f92ce ("IB/mlx5: Add ODP atomics support") Reviewed-by: Artemy Kovalyov Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx5/odp.c | 2 +- include/linux/mlx5/qp.h | 1 - 2 files changed, 1 insertion(+), 2 deletions(-) diff --git a/drivers/infiniband/hw/mlx5/odp.c b/drivers/infiniband/hw/mlx5/odp.c index ae0746754008..3d701c7a4c91 100644 --- a/drivers/infiniband/hw/mlx5/odp.c +++ b/drivers/infiniband/hw/mlx5/odp.c @@ -939,7 +939,7 @@ static int mlx5_ib_mr_initiator_pfault_handler( if (qp->ibqp.qp_type != IB_QPT_RC) { av = *wqe; - if (av->dqp_dct & be32_to_cpu(MLX5_WQE_AV_EXT)) + if (av->dqp_dct & cpu_to_be32(MLX5_EXTENDED_UD_AV)) *wqe += sizeof(struct mlx5_av); else *wqe += sizeof(struct mlx5_base_av); diff --git a/include/linux/mlx5/qp.h b/include/linux/mlx5/qp.h index fff4ec13f620..66d19b611fe4 100644 --- a/include/linux/mlx5/qp.h +++ b/include/linux/mlx5/qp.h @@ -212,7 +212,6 @@ struct mlx5_wqe_ctrl_seg { #define MLX5_WQE_CTRL_OPCODE_MASK 0xff #define MLX5_WQE_CTRL_WQE_INDEX_MASK 0x00ffff00 #define MLX5_WQE_CTRL_WQE_INDEX_SHIFT 8 -#define MLX5_WQE_AV_EXT 0x80000000 enum { MLX5_ETH_WQE_L3_INNER_CSUM = 1 << 4, From 58dcb60a226ee48cc66c96c27b751f06ec2bc5a9 Mon Sep 17 00:00:00 2001 From: Parav Pandit Date: Mon, 19 Jun 2017 07:19:37 +0300 Subject: [PATCH 28/34] IB/mlx5: Expose extended error counters This patch adds below requester and responder side error counters, which will be exposed by hardware counters interface and are supported as part of query Q counters command extension. +---------------------------+-------------------------------------+ | Name | Description | |---------------------------+-------------------------------------| |resp_local_length_error | Number of times responder detected | | | local length errors | |---------------------------+-------------------------------------| |resp_cqe_error | Number of CQEs completed with error | | | at responder | |---------------------------+-------------------------------------| |req_cqe_error | Number of CQEs completed with error | | | at requester | |---------------------------+-------------------------------------| |req_remote_invalid_request | Number of times requester detected | | | remote invalid request error | |---------------------------+-------------------------------------| |req_remote_access_error | Number of times requester detected | | | remote access error | |---------------------------+-------------------------------------| |resp_remote_access_error | Number of times responder detected | | | remote access error | |---------------------------+-------------------------------------| |resp_cqe_flush_error | Number of CQEs completed with | | | flushed with error at responder | |---------------------------+-------------------------------------| |req_cqe_flush_error | Number of CQEs completed with | | | flushed with error at requester | +---------------------------+-------------------------------------+ Signed-off-by: Parav Pandit Reviewed-by: Daniel Jurgens Reviewed-by: Eli Cohen Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx5/main.c | 22 ++++++++++++++++ include/linux/mlx5/mlx5_ifc.h | 44 +++++++++++++++++++++++++++++-- 2 files changed, 64 insertions(+), 2 deletions(-) diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index 2757d445b042..9dd9759459fb 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -3432,6 +3432,17 @@ static const struct mlx5_ib_counter cong_cnts[] = { INIT_CONG_COUNTER(np_cnp_sent), }; +static const struct mlx5_ib_counter extended_err_cnts[] = { + INIT_Q_COUNTER(resp_local_length_error), + INIT_Q_COUNTER(resp_cqe_error), + INIT_Q_COUNTER(req_cqe_error), + INIT_Q_COUNTER(req_remote_invalid_request), + INIT_Q_COUNTER(req_remote_access_errors), + INIT_Q_COUNTER(resp_remote_access_errors), + INIT_Q_COUNTER(resp_cqe_flush_error), + INIT_Q_COUNTER(req_cqe_flush_error), +}; + static void mlx5_ib_dealloc_counters(struct mlx5_ib_dev *dev) { unsigned int i; @@ -3456,6 +3467,10 @@ static int __mlx5_ib_alloc_counters(struct mlx5_ib_dev *dev, if (MLX5_CAP_GEN(dev->mdev, retransmission_q_counters)) num_counters += ARRAY_SIZE(retrans_q_cnts); + + if (MLX5_CAP_GEN(dev->mdev, enhanced_error_q_counters)) + num_counters += ARRAY_SIZE(extended_err_cnts); + cnts->num_q_counters = num_counters; if (MLX5_CAP_GEN(dev->mdev, cc_query_allowed)) { @@ -3505,6 +3520,13 @@ static void mlx5_ib_fill_counters(struct mlx5_ib_dev *dev, } } + if (MLX5_CAP_GEN(dev->mdev, enhanced_error_q_counters)) { + for (i = 0; i < ARRAY_SIZE(extended_err_cnts); i++, j++) { + names[j] = extended_err_cnts[i].name; + offsets[j] = extended_err_cnts[i].offset; + } + } + if (MLX5_CAP_GEN(dev->mdev, cc_query_allowed)) { for (i = 0; i < ARRAY_SIZE(cong_cnts); i++, j++) { names[j] = cong_cnts[i].name; diff --git a/include/linux/mlx5/mlx5_ifc.h b/include/linux/mlx5/mlx5_ifc.h index f350688a12fa..5bae70eb25af 100644 --- a/include/linux/mlx5/mlx5_ifc.h +++ b/include/linux/mlx5/mlx5_ifc.h @@ -858,7 +858,7 @@ struct mlx5_ifc_cmd_hca_cap_bits { u8 pcam_reg[0x1]; u8 local_ca_ack_delay[0x5]; u8 port_module_event[0x1]; - u8 reserved_at_1b1[0x1]; + u8 enhanced_error_q_counters[0x1]; u8 ports_check[0x1]; u8 reserved_at_1b3[0x1]; u8 disable_link_up[0x1]; @@ -3953,7 +3953,47 @@ struct mlx5_ifc_query_q_counter_out_bits { u8 local_ack_timeout_err[0x20]; - u8 reserved_at_320[0x4e0]; + u8 reserved_at_320[0xa0]; + + u8 resp_local_length_error[0x20]; + + u8 req_local_length_error[0x20]; + + u8 resp_local_qp_error[0x20]; + + u8 local_operation_error[0x20]; + + u8 resp_local_protection[0x20]; + + u8 req_local_protection[0x20]; + + u8 resp_cqe_error[0x20]; + + u8 req_cqe_error[0x20]; + + u8 req_mw_binding[0x20]; + + u8 req_bad_response[0x20]; + + u8 req_remote_invalid_request[0x20]; + + u8 resp_remote_invalid_request[0x20]; + + u8 req_remote_access_errors[0x20]; + + u8 resp_remote_access_errors[0x20]; + + u8 req_remote_operation_errors[0x20]; + + u8 req_transport_retries_exceeded[0x20]; + + u8 cq_overflow[0x20]; + + u8 resp_cqe_flush_error[0x20]; + + u8 req_cqe_flush_error[0x20]; + + u8 reserved_at_620[0x1e0]; }; struct mlx5_ifc_query_q_counter_in_bits { From ea30b966f7dd6bcfb20c98a7f99608c7bb10bfac Mon Sep 17 00:00:00 2001 From: Maor Gottlieb Date: Wed, 21 Jun 2017 09:26:28 +0300 Subject: [PATCH 29/34] IB/mlx4: Add inline-receive support When inline-receive is enabled, the HCA may write received data into the receive WQE. Inline-receive is enabled by setting its matching bit in the QP context and each single-packet message with payload not exceeding the receive WQE size will be delivered to the WQE. The completion report will indicate that the payload was placed to the WQE. It includes: 1) Return maximum supported size of inline-receive by the hardware in query_device vendor's data part. 2) Enable the feature when requested by the vendor data input. Signed-off-by: Maor Gottlieb Reviewed-by: Yishai Hadas Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx4/main.c | 7 ++++++ drivers/infiniband/hw/mlx4/mlx4_ib.h | 3 +++ drivers/infiniband/hw/mlx4/qp.c | 32 ++++++++++++++++++++++------ include/uapi/rdma/mlx4-abi.h | 3 ++- 4 files changed, 37 insertions(+), 8 deletions(-) diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c index d1b43cbbfea7..e8c290edb1e1 100644 --- a/drivers/infiniband/hw/mlx4/main.c +++ b/drivers/infiniband/hw/mlx4/main.c @@ -563,6 +563,13 @@ static int mlx4_ib_query_device(struct ib_device *ibdev, } } + if (uhw->outlen >= resp.response_length + + sizeof(resp.max_inl_recv_sz)) { + resp.response_length += sizeof(resp.max_inl_recv_sz); + resp.max_inl_recv_sz = dev->dev->caps.max_rq_sg * + sizeof(struct mlx4_wqe_data_seg); + } + if (uhw->outlen) { err = ib_copy_to_udata(uhw, &resp, resp.response_length); if (err) diff --git a/drivers/infiniband/hw/mlx4/mlx4_ib.h b/drivers/infiniband/hw/mlx4/mlx4_ib.h index 9db82e67e959..a6337f3161cf 100644 --- a/drivers/infiniband/hw/mlx4/mlx4_ib.h +++ b/drivers/infiniband/hw/mlx4/mlx4_ib.h @@ -318,6 +318,7 @@ struct mlx4_ib_qp { u8 sq_no_prefetch; u8 state; int mlx_type; + u32 inl_recv_sz; struct list_head gid_list; struct list_head steering_rules; struct mlx4_ib_buf *sqp_proxy_rcv; @@ -623,6 +624,8 @@ struct mlx4_uverbs_ex_query_device_resp { __u32 comp_mask; __u32 response_length; __u64 hca_core_clock_offset; + __u32 max_inl_recv_sz; + __u32 reserved; }; static inline struct mlx4_ib_dev *to_mdev(struct ib_device *ibdev) diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c index 75c0e6c5dd56..d1caa39a3943 100644 --- a/drivers/infiniband/hw/mlx4/qp.c +++ b/drivers/infiniband/hw/mlx4/qp.c @@ -377,7 +377,8 @@ 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) + int is_user, int has_rq, struct mlx4_ib_qp *qp, + u32 inl_recv_sz) { /* Sanity check RQ size before proceeding */ if (cap->max_recv_wr > dev->dev->caps.max_wqes - MLX4_IB_SQ_MAX_SPARE || @@ -385,18 +386,24 @@ static int set_rq_size(struct mlx4_ib_dev *dev, struct ib_qp_cap *cap, return -EINVAL; if (!has_rq) { - if (cap->max_recv_wr) + if (cap->max_recv_wr || inl_recv_sz) return -EINVAL; qp->rq.wqe_cnt = qp->rq.max_gs = 0; } else { + u32 max_inl_recv_sz = dev->dev->caps.max_rq_sg * + sizeof(struct mlx4_wqe_data_seg); + u32 wqe_size; + /* HW requires >= 1 RQ entry with >= 1 gather entry */ - if (is_user && (!cap->max_recv_wr || !cap->max_recv_sge)) + if (is_user && (!cap->max_recv_wr || !cap->max_recv_sge || + inl_recv_sz > max_inl_recv_sz)) return -EINVAL; qp->rq.wqe_cnt = roundup_pow_of_two(max(1U, cap->max_recv_wr)); qp->rq.max_gs = roundup_pow_of_two(max(1U, cap->max_recv_sge)); - qp->rq.wqe_shift = ilog2(qp->rq.max_gs * sizeof (struct mlx4_wqe_data_seg)); + wqe_size = qp->rq.max_gs * sizeof(struct mlx4_wqe_data_seg); + qp->rq.wqe_shift = ilog2(max_t(u32, wqe_size, inl_recv_sz)); } /* leave userspace return values as they were, so as not to break ABI */ @@ -719,9 +726,6 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd, if (init_attr->sq_sig_type == IB_SIGNAL_ALL_WR) qp->sq_signal_bits = cpu_to_be32(MLX4_WQE_CTRL_CQ_UPDATE); - err = set_rq_size(dev, &init_attr->cap, !!pd->uobject, qp_has_rq(init_attr), qp); - if (err) - goto err; if (pd->uobject) { struct mlx4_ib_create_qp ucmd; @@ -731,6 +735,12 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd, goto err; } + err = set_rq_size(dev, &init_attr->cap, !!pd->uobject, + qp_has_rq(init_attr), qp, ucmd.inl_recv_sz); + if (err) + goto err; + + qp->inl_recv_sz = ucmd.inl_recv_sz; qp->sq_no_prefetch = ucmd.sq_no_prefetch; err = set_user_sq_size(dev, qp, &ucmd); @@ -760,6 +770,11 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd, goto err_mtt; } } else { + err = set_rq_size(dev, &init_attr->cap, !!pd->uobject, + qp_has_rq(init_attr), qp, 0); + if (err) + goto err; + qp->sq_no_prefetch = 0; if (init_attr->create_flags & IB_QP_CREATE_IPOIB_UD_LSO) @@ -1651,6 +1666,9 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp, } } + if (qp->inl_recv_sz) + context->param3 |= cpu_to_be32(1 << 25); + if (ibqp->qp_type == IB_QPT_GSI || ibqp->qp_type == IB_QPT_SMI) context->mtu_msgmax = (IB_MTU_4096 << 5) | 11; else if (ibqp->qp_type == IB_QPT_RAW_PACKET) diff --git a/include/uapi/rdma/mlx4-abi.h b/include/uapi/rdma/mlx4-abi.h index af431752655c..bf3bdba2f326 100644 --- a/include/uapi/rdma/mlx4-abi.h +++ b/include/uapi/rdma/mlx4-abi.h @@ -101,7 +101,8 @@ struct mlx4_ib_create_qp { __u8 log_sq_bb_count; __u8 log_sq_stride; __u8 sq_no_prefetch; - __u8 reserved[5]; + __u32 inl_recv_sz; + __u8 reserved; }; #endif /* MLX4_ABI_USER_H */ From f3301870161ca293cd14b20a802c5646da02407f Mon Sep 17 00:00:00 2001 From: Moshe Shemesh Date: Wed, 21 Jun 2017 09:29:36 +0300 Subject: [PATCH 30/34] (IB, net)/mlx4: Add resource utilization support Adding visibility of resource usage of QPs, CQs and counters used by virtual functions. This feature will be used to give the PF administrator more data while debugging VF status. Usage info was added to ALLOC_RES command, to notify the PF if the resource which is being reserved or allocated for the VF will be used by kernel driver or by user verbs. Updated reservation and allocation functions of QP, CQ and counter with additional usage parameter. Signed-off-by: Moshe Shemesh Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx4/cq.c | 2 ++ drivers/infiniband/hw/mlx4/main.c | 6 ++++-- drivers/infiniband/hw/mlx4/qp.c | 13 +++++++++---- drivers/net/ethernet/mellanox/mlx4/cq.c | 7 ++++--- drivers/net/ethernet/mellanox/mlx4/en_cq.c | 1 + drivers/net/ethernet/mellanox/mlx4/en_netdev.c | 3 ++- drivers/net/ethernet/mellanox/mlx4/en_rx.c | 6 ++++-- drivers/net/ethernet/mellanox/mlx4/en_tx.c | 3 ++- drivers/net/ethernet/mellanox/mlx4/main.c | 7 ++++--- drivers/net/ethernet/mellanox/mlx4/qp.c | 5 +++-- include/linux/mlx4/device.h | 12 ++++++++++-- 11 files changed, 45 insertions(+), 20 deletions(-) diff --git a/drivers/infiniband/hw/mlx4/cq.c b/drivers/infiniband/hw/mlx4/cq.c index ff931c580557..95382faa7ad1 100644 --- a/drivers/infiniband/hw/mlx4/cq.c +++ b/drivers/infiniband/hw/mlx4/cq.c @@ -218,6 +218,7 @@ struct ib_cq *mlx4_ib_create_cq(struct ib_device *ibdev, goto err_mtt; uar = &to_mucontext(context)->uar; + cq->mcq.usage = MLX4_RES_USAGE_USER_VERBS; } else { err = mlx4_db_alloc(dev->dev, &cq->db, 1); if (err) @@ -233,6 +234,7 @@ struct ib_cq *mlx4_ib_create_cq(struct ib_device *ibdev, goto err_db; uar = &dev->priv_uar; + cq->mcq.usage = MLX4_RES_USAGE_DRIVER; } if (dev->eq_table) diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c index e8c290edb1e1..0944e224c0df 100644 --- a/drivers/infiniband/hw/mlx4/main.c +++ b/drivers/infiniband/hw/mlx4/main.c @@ -2779,7 +2779,8 @@ static void *mlx4_ib_add(struct mlx4_dev *dev) allocated = 0; if (mlx4_ib_port_link_layer(&ibdev->ib_dev, i + 1) == IB_LINK_LAYER_ETHERNET) { - err = mlx4_counter_alloc(ibdev->dev, &counter_index); + err = mlx4_counter_alloc(ibdev->dev, &counter_index, + MLX4_RES_USAGE_DRIVER); /* if failed to allocate a new counter, use default */ if (err) counter_index = @@ -2834,7 +2835,8 @@ static void *mlx4_ib_add(struct mlx4_dev *dev) ibdev->steer_qpn_count = MLX4_IB_UC_MAX_NUM_QPS; err = mlx4_qp_reserve_range(dev, ibdev->steer_qpn_count, MLX4_IB_UC_STEER_QPN_ALIGN, - &ibdev->steer_qpn_base, 0); + &ibdev->steer_qpn_base, 0, + MLX4_RES_USAGE_DRIVER); if (err) goto err_counter; diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c index d1caa39a3943..247b9132e9de 100644 --- a/drivers/infiniband/hw/mlx4/qp.c +++ b/drivers/infiniband/hw/mlx4/qp.c @@ -769,6 +769,7 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd, if (err) goto err_mtt; } + qp->mqp.usage = MLX4_RES_USAGE_USER_VERBS; } else { err = set_rq_size(dev, &init_attr->cap, !!pd->uobject, qp_has_rq(init_attr), qp, 0); @@ -841,6 +842,7 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd, err = -ENOMEM; goto err_wrid; } + qp->mqp.usage = MLX4_RES_USAGE_DRIVER; } if (sqpn) { @@ -860,13 +862,14 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd, (init_attr->cap.max_send_wr ? MLX4_RESERVE_ETH_BF_QP : 0) | (init_attr->cap.max_recv_wr ? - MLX4_RESERVE_A0_QP : 0)); + MLX4_RESERVE_A0_QP : 0), + qp->mqp.usage); else if (qp->flags & MLX4_IB_QP_NETIF) err = mlx4_ib_steer_qp_alloc(dev, 1, &qpn); else err = mlx4_qp_reserve_range(dev->dev, 1, 1, - &qpn, 0); + &qpn, 0, qp->mqp.usage); if (err) goto err_proxy; } @@ -1218,7 +1221,9 @@ static struct ib_qp *_mlx4_ib_create_qp(struct ib_pd *pd, if (udata) return ERR_PTR(-EINVAL); if (init_attr->create_flags & MLX4_IB_QP_CREATE_ROCE_V2_GSI) { - int res = mlx4_qp_reserve_range(to_mdev(pd->device)->dev, 1, 1, &sqpn, 0); + int res = mlx4_qp_reserve_range(to_mdev(pd->device)->dev, + 1, 1, &sqpn, 0, + MLX4_RES_USAGE_DRIVER); if (res) return ERR_PTR(res); @@ -1581,7 +1586,7 @@ static int create_qp_lb_counter(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp) !(dev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_LB_SRC_CHK)) return 0; - err = mlx4_counter_alloc(dev->dev, &tmp_idx); + err = mlx4_counter_alloc(dev->dev, &tmp_idx, MLX4_RES_USAGE_DRIVER); if (err) return err; diff --git a/drivers/net/ethernet/mellanox/mlx4/cq.c b/drivers/net/ethernet/mellanox/mlx4/cq.c index c56a511b918e..72eb50cd5ecd 100644 --- a/drivers/net/ethernet/mellanox/mlx4/cq.c +++ b/drivers/net/ethernet/mellanox/mlx4/cq.c @@ -241,13 +241,14 @@ int __mlx4_cq_alloc_icm(struct mlx4_dev *dev, int *cqn) return err; } -static int mlx4_cq_alloc_icm(struct mlx4_dev *dev, int *cqn) +static int mlx4_cq_alloc_icm(struct mlx4_dev *dev, int *cqn, u8 usage) { + u32 in_modifier = RES_CQ | (((u32)usage & 3) << 30); u64 out_param; int err; if (mlx4_is_mfunc(dev)) { - err = mlx4_cmd_imm(dev, 0, &out_param, RES_CQ, + err = mlx4_cmd_imm(dev, 0, &out_param, in_modifier, RES_OP_RESERVE_AND_MAP, MLX4_CMD_ALLOC_RES, MLX4_CMD_TIME_CLASS_A, MLX4_CMD_WRAPPED); if (err) @@ -303,7 +304,7 @@ int mlx4_cq_alloc(struct mlx4_dev *dev, int nent, cq->vector = vector; - err = mlx4_cq_alloc_icm(dev, &cq->cqn); + err = mlx4_cq_alloc_icm(dev, &cq->cqn, cq->usage); if (err) return err; diff --git a/drivers/net/ethernet/mellanox/mlx4/en_cq.c b/drivers/net/ethernet/mellanox/mlx4/en_cq.c index 85fe17e4dcfb..f849eec21824 100644 --- a/drivers/net/ethernet/mellanox/mlx4/en_cq.c +++ b/drivers/net/ethernet/mellanox/mlx4/en_cq.c @@ -140,6 +140,7 @@ int mlx4_en_activate_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq, (cq->type == RX && priv->hwtstamp_config.rx_filter)) timestamp_en = 1; + cq->mcq.usage = MLX4_RES_USAGE_DRIVER; err = mlx4_cq_alloc(mdev->dev, cq->size, &cq->wqres.mtt, &mdev->priv_uar, cq->wqres.db.dma, &cq->mcq, cq->vector, 0, timestamp_en); diff --git a/drivers/net/ethernet/mellanox/mlx4/en_netdev.c b/drivers/net/ethernet/mellanox/mlx4/en_netdev.c index 3a291fc1780a..e3e6d9fa69fd 100644 --- a/drivers/net/ethernet/mellanox/mlx4/en_netdev.c +++ b/drivers/net/ethernet/mellanox/mlx4/en_netdev.c @@ -651,7 +651,8 @@ static int mlx4_en_get_qp(struct mlx4_en_priv *priv) return 0; } - err = mlx4_qp_reserve_range(dev, 1, 1, qpn, MLX4_RESERVE_A0_QP); + err = mlx4_qp_reserve_range(dev, 1, 1, qpn, MLX4_RESERVE_A0_QP, + MLX4_RES_USAGE_DRIVER); en_dbg(DRV, priv, "Reserved qp %d\n", *qpn); if (err) { en_err(priv, "Failed to reserve qp for mac registration\n"); diff --git a/drivers/net/ethernet/mellanox/mlx4/en_rx.c b/drivers/net/ethernet/mellanox/mlx4/en_rx.c index 436f7689a032..ad1ffd5857cb 100644 --- a/drivers/net/ethernet/mellanox/mlx4/en_rx.c +++ b/drivers/net/ethernet/mellanox/mlx4/en_rx.c @@ -1081,7 +1081,8 @@ int mlx4_en_create_drop_qp(struct mlx4_en_priv *priv) u32 qpn; err = mlx4_qp_reserve_range(priv->mdev->dev, 1, 1, &qpn, - MLX4_RESERVE_A0_QP); + MLX4_RESERVE_A0_QP, + MLX4_RES_USAGE_DRIVER); if (err) { en_err(priv, "Failed reserving drop qpn\n"); return err; @@ -1127,7 +1128,8 @@ int mlx4_en_config_rss_steer(struct mlx4_en_priv *priv) flags = priv->rx_ring_num == 1 ? MLX4_RESERVE_A0_QP : 0; err = mlx4_qp_reserve_range(mdev->dev, priv->rx_ring_num, priv->rx_ring_num, - &rss_map->base_qpn, flags); + &rss_map->base_qpn, flags, + MLX4_RES_USAGE_DRIVER); if (err) { en_err(priv, "Failed reserving %d qps\n", priv->rx_ring_num); return err; diff --git a/drivers/net/ethernet/mellanox/mlx4/en_tx.c b/drivers/net/ethernet/mellanox/mlx4/en_tx.c index 73faa3d77921..a81db2582555 100644 --- a/drivers/net/ethernet/mellanox/mlx4/en_tx.c +++ b/drivers/net/ethernet/mellanox/mlx4/en_tx.c @@ -105,7 +105,8 @@ int mlx4_en_create_tx_ring(struct mlx4_en_priv *priv, (unsigned long long) ring->sp_wqres.buf.direct.map); err = mlx4_qp_reserve_range(mdev->dev, 1, 1, &ring->qpn, - MLX4_RESERVE_ETH_BF_QP); + MLX4_RESERVE_ETH_BF_QP, + MLX4_RES_USAGE_DRIVER); if (err) { en_err(priv, "failed reserving qp for TX ring\n"); goto err_hwq_res; diff --git a/drivers/net/ethernet/mellanox/mlx4/main.c b/drivers/net/ethernet/mellanox/mlx4/main.c index a27c9c13a36e..fb2591d0e735 100644 --- a/drivers/net/ethernet/mellanox/mlx4/main.c +++ b/drivers/net/ethernet/mellanox/mlx4/main.c @@ -2475,7 +2475,7 @@ static int mlx4_allocate_default_counters(struct mlx4_dev *dev) priv->def_counter[port] = -1; for (port = 0; port < dev->caps.num_ports; port++) { - err = mlx4_counter_alloc(dev, &idx); + err = mlx4_counter_alloc(dev, &idx, MLX4_RES_USAGE_DRIVER); if (!err || err == -ENOSPC) { priv->def_counter[port] = idx; @@ -2517,13 +2517,14 @@ int __mlx4_counter_alloc(struct mlx4_dev *dev, u32 *idx) return 0; } -int mlx4_counter_alloc(struct mlx4_dev *dev, u32 *idx) +int mlx4_counter_alloc(struct mlx4_dev *dev, u32 *idx, u8 usage) { + u32 in_modifier = RES_COUNTER | (((u32)usage & 3) << 30); u64 out_param; int err; if (mlx4_is_mfunc(dev)) { - err = mlx4_cmd_imm(dev, 0, &out_param, RES_COUNTER, + err = mlx4_cmd_imm(dev, 0, &out_param, in_modifier, RES_OP_RESERVE, MLX4_CMD_ALLOC_RES, MLX4_CMD_TIME_CLASS_A, MLX4_CMD_WRAPPED); if (!err) diff --git a/drivers/net/ethernet/mellanox/mlx4/qp.c b/drivers/net/ethernet/mellanox/mlx4/qp.c index 26747212526b..5e5b4475b85e 100644 --- a/drivers/net/ethernet/mellanox/mlx4/qp.c +++ b/drivers/net/ethernet/mellanox/mlx4/qp.c @@ -245,8 +245,9 @@ int __mlx4_qp_reserve_range(struct mlx4_dev *dev, int cnt, int align, } int mlx4_qp_reserve_range(struct mlx4_dev *dev, int cnt, int align, - int *base, u8 flags) + int *base, u8 flags, u8 usage) { + u32 in_modifier = RES_QP | (((u32)usage & 3) << 30); u64 in_param = 0; u64 out_param; int err; @@ -258,7 +259,7 @@ int mlx4_qp_reserve_range(struct mlx4_dev *dev, int cnt, int align, set_param_l(&in_param, (((u32)flags) << 24) | (u32)cnt); set_param_h(&in_param, align); err = mlx4_cmd_imm(dev, in_param, &out_param, - RES_QP, RES_OP_RESERVE, + in_modifier, RES_OP_RESERVE, MLX4_CMD_ALLOC_RES, MLX4_CMD_TIME_CLASS_A, MLX4_CMD_WRAPPED); if (err) diff --git a/include/linux/mlx4/device.h b/include/linux/mlx4/device.h index aad5d81dfb44..3607da001ad3 100644 --- a/include/linux/mlx4/device.h +++ b/include/linux/mlx4/device.h @@ -428,6 +428,12 @@ enum mlx4_steer_type { MLX4_NUM_STEERS }; +enum mlx4_resource_usage { + MLX4_RES_USAGE_NONE, + MLX4_RES_USAGE_DRIVER, + MLX4_RES_USAGE_USER_VERBS, +}; + enum { MLX4_NUM_FEXCH = 64 * 1024, }; @@ -748,6 +754,7 @@ struct mlx4_cq { } tasklet_ctx; int reset_notify_added; struct list_head reset_notify; + u8 usage; }; struct mlx4_qp { @@ -757,6 +764,7 @@ struct mlx4_qp { atomic_t refcount; struct completion free; + u8 usage; }; struct mlx4_srq { @@ -1120,7 +1128,7 @@ int mlx4_cq_alloc(struct mlx4_dev *dev, int nent, struct mlx4_mtt *mtt, unsigned vector, int collapsed, int timestamp_en); void mlx4_cq_free(struct mlx4_dev *dev, struct mlx4_cq *cq); int mlx4_qp_reserve_range(struct mlx4_dev *dev, int cnt, int align, - int *base, u8 flags); + int *base, u8 flags, u8 usage); void mlx4_qp_release_range(struct mlx4_dev *dev, int base_qpn, int cnt); int mlx4_qp_alloc(struct mlx4_dev *dev, int qpn, struct mlx4_qp *qp); @@ -1417,7 +1425,7 @@ int mlx4_get_phys_port_id(struct mlx4_dev *dev); int mlx4_wol_read(struct mlx4_dev *dev, u64 *config, int port); int mlx4_wol_write(struct mlx4_dev *dev, u64 config, int port); -int mlx4_counter_alloc(struct mlx4_dev *dev, u32 *idx); +int mlx4_counter_alloc(struct mlx4_dev *dev, u32 *idx, u8 usage); void mlx4_counter_free(struct mlx4_dev *dev, u32 idx); int mlx4_get_default_counter_index(struct mlx4_dev *dev, int port); From 400b1ebcfe31279895f1baa8ecaa390d9a4a0eef Mon Sep 17 00:00:00 2001 From: Guy Levi Date: Tue, 4 Jul 2017 16:24:24 +0300 Subject: [PATCH 31/34] IB/mlx4: Add support for WQ related verbs Support create/modify/destroy WQ related verbs. The base IB object to enable RSS functionality is a WQ (i.e. ib_wq). This patch implements the related WQ verbs as of create, modify and destroy. In downstream patches the WQ will be used as part of an indirection table (i.e. ib_rwq_ind_table) to enable RSS QP creation. Notes: ConnectX-3 hardware requires consecutive WQNs list as receive descriptor queues for the RSS QP. Hence, the driver manages consecutive ranges lists per context which the user must respect. Destroying the WQ does not return its WQN back to its range for reusing. However, destroying all WQs from the same range releases the range and in turn releases its WQNs for reusing. Since the WQ object is not a natural object in the hardware, the driver implements the WQ by the hardware QP. As such, the WQ inherits its port from its RSS QP parent upon its RST->INIT transition and by that time its state is applied to the hardware. Signed-off-by: Guy Levi Reviewed-by: Yishai Hadas Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx4/main.c | 24 ++ drivers/infiniband/hw/mlx4/mlx4_ib.h | 25 +- drivers/infiniband/hw/mlx4/qp.c | 503 ++++++++++++++++++++++++--- include/uapi/rdma/mlx4-abi.h | 14 + 4 files changed, 510 insertions(+), 56 deletions(-) diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c index 0944e224c0df..7c6f929ebd3e 100644 --- a/drivers/infiniband/hw/mlx4/main.c +++ b/drivers/infiniband/hw/mlx4/main.c @@ -81,6 +81,8 @@ static const char mlx4_ib_version[] = DRV_VERSION "\n"; static void do_slave_init(struct mlx4_ib_dev *ibdev, int slave, int do_init); +static enum rdma_link_layer mlx4_ib_port_link_layer(struct ib_device *device, + u8 port_num); static struct workqueue_struct *wq; @@ -552,6 +554,11 @@ static int mlx4_ib_query_device(struct ib_device *ibdev, props->timestamp_mask = 0xFFFFFFFFFFFFULL; props->max_ah = INT_MAX; + if ((dev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_RSS) && + (mlx4_ib_port_link_layer(ibdev, 1) == IB_LINK_LAYER_ETHERNET || + mlx4_ib_port_link_layer(ibdev, 2) == IB_LINK_LAYER_ETHERNET)) + props->max_wq_type_rq = props->max_qp; + if (!mlx4_is_slave(dev->dev)) err = mlx4_get_internal_clock_params(dev->dev, &clock_params); @@ -1076,6 +1083,9 @@ static struct ib_ucontext *mlx4_ib_alloc_ucontext(struct ib_device *ibdev, INIT_LIST_HEAD(&context->db_page_list); mutex_init(&context->db_page_mutex); + INIT_LIST_HEAD(&context->wqn_ranges_list); + mutex_init(&context->wqn_ranges_mutex); + if (ibdev->uverbs_abi_ver == MLX4_IB_UVERBS_NO_DEV_CAPS_ABI_VERSION) err = ib_copy_to_udata(udata, &resp_v3, sizeof(resp_v3)); else @@ -2720,6 +2730,20 @@ static void *mlx4_ib_add(struct mlx4_dev *dev) ibdev->ib_dev.get_dev_fw_str = get_fw_ver_str; ibdev->ib_dev.disassociate_ucontext = mlx4_ib_disassociate_ucontext; + 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.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); + } + 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; diff --git a/drivers/infiniband/hw/mlx4/mlx4_ib.h b/drivers/infiniband/hw/mlx4/mlx4_ib.h index a6337f3161cf..ba4e78c064d2 100644 --- a/drivers/infiniband/hw/mlx4/mlx4_ib.h +++ b/drivers/infiniband/hw/mlx4/mlx4_ib.h @@ -88,6 +88,8 @@ struct mlx4_ib_ucontext { struct list_head db_page_list; struct mutex db_page_mutex; struct mlx4_ib_vma_private_data hw_bar_info[HW_BAR_COUNT]; + struct list_head wqn_ranges_list; + struct mutex wqn_ranges_mutex; /* protect wqn_ranges_list */ }; struct mlx4_ib_pd { @@ -289,8 +291,19 @@ struct mlx4_roce_smac_vlan_info { int update_vid; }; +struct mlx4_wqn_range { + int base_wqn; + int size; + int refcount; + bool dirty; + struct list_head list; +}; + struct mlx4_ib_qp { - struct ib_qp ibqp; + union { + struct ib_qp ibqp; + struct ib_wq ibwq; + }; struct mlx4_qp mqp; struct mlx4_buf buf; @@ -329,6 +342,9 @@ struct mlx4_ib_qp { struct list_head cq_recv_list; struct list_head cq_send_list; struct counter_index *counter_index; + struct mlx4_wqn_range *wqn_range; + /* Number of RSS QP parents that uses this WQ */ + u32 rss_usecnt; }; struct mlx4_ib_srq { @@ -893,4 +909,11 @@ void mlx4_sched_ib_sl2vl_update_work(struct mlx4_ib_dev *ibdev, void mlx4_ib_sl2vl_update(struct mlx4_ib_dev *mdev, int port); +struct ib_wq *mlx4_ib_create_wq(struct ib_pd *pd, + struct ib_wq_init_attr *init_attr, + struct ib_udata *udata); +int mlx4_ib_destroy_wq(struct ib_wq *wq); +int mlx4_ib_modify_wq(struct ib_wq *wq, struct ib_wq_attr *wq_attr, + u32 wq_attr_mask, struct ib_udata *udata); + #endif /* MLX4_IB_H */ diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c index 247b9132e9de..65e4ec549368 100644 --- a/drivers/infiniband/hw/mlx4/qp.c +++ b/drivers/infiniband/hw/mlx4/qp.c @@ -116,6 +116,11 @@ static const __be32 mlx4_ib_opcode[] = { [IB_WR_MASKED_ATOMIC_FETCH_AND_ADD] = cpu_to_be32(MLX4_OPCODE_MASKED_ATOMIC_FA), }; +enum mlx4_ib_source_type { + MLX4_IB_QP_SRC = 0, + MLX4_IB_RWQ_SRC = 1, +}; + static struct mlx4_ib_sqp *to_msqp(struct mlx4_ib_qp *mqp) { return container_of(mqp, struct mlx4_ib_sqp, qp); @@ -330,6 +335,12 @@ static void mlx4_ib_qp_event(struct mlx4_qp *qp, enum mlx4_event type) } } +static void mlx4_ib_wq_event(struct mlx4_qp *qp, enum mlx4_event type) +{ + pr_warn_ratelimited("Unexpected event type %d on WQ 0x%06x. Events are not supported for WQs\n", + type, qp->qpn); +} + static int send_wqe_overhead(enum mlx4_ib_qp_type type, u32 flags) { /* @@ -639,7 +650,91 @@ static void mlx4_ib_free_qp_counter(struct mlx4_ib_dev *dev, qp->counter_index = NULL; } +/* + * This function allocates a WQN from a range which is consecutive and aligned + * to its size. In case the range is full, then it creates a new range and + * allocates WQN from it. The new range will be used for following allocations. + */ +static int mlx4_ib_alloc_wqn(struct mlx4_ib_ucontext *context, + struct mlx4_ib_qp *qp, int range_size, int *wqn) +{ + struct mlx4_ib_dev *dev = to_mdev(context->ibucontext.device); + struct mlx4_wqn_range *range; + int err = 0; + + mutex_lock(&context->wqn_ranges_mutex); + + range = list_first_entry_or_null(&context->wqn_ranges_list, + struct mlx4_wqn_range, list); + + if (!range || (range->refcount == range->size) || range->dirty) { + range = kzalloc(sizeof(*range), GFP_KERNEL); + if (!range) { + err = -ENOMEM; + goto out; + } + + err = mlx4_qp_reserve_range(dev->dev, range_size, + range_size, &range->base_wqn, 0, + qp->mqp.usage); + if (err) { + kfree(range); + goto out; + } + + range->size = range_size; + list_add(&range->list, &context->wqn_ranges_list); + } else if (range_size != 1) { + /* + * Requesting a new range (>1) when last range is still open, is + * not valid. + */ + err = -EINVAL; + goto out; + } + + qp->wqn_range = range; + + *wqn = range->base_wqn + range->refcount; + + range->refcount++; + +out: + mutex_unlock(&context->wqn_ranges_mutex); + + return err; +} + +static void mlx4_ib_release_wqn(struct mlx4_ib_ucontext *context, + struct mlx4_ib_qp *qp, bool dirty_release) +{ + struct mlx4_ib_dev *dev = to_mdev(context->ibucontext.device); + struct mlx4_wqn_range *range; + + mutex_lock(&context->wqn_ranges_mutex); + + range = qp->wqn_range; + + range->refcount--; + if (!range->refcount) { + mlx4_qp_release_range(dev->dev, range->base_wqn, + range->size); + list_del(&range->list); + kfree(range); + } else if (dirty_release) { + /* + * A range which one of its WQNs is destroyed, won't be able to be + * reused for further WQN allocations. + * The next created WQ will allocate a new range. + */ + range->dirty = 1; + } + + mutex_unlock(&context->wqn_ranges_mutex); +} + static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd, + enum mlx4_ib_source_type src, struct ib_qp_init_attr *init_attr, struct ib_udata *udata, int sqpn, struct mlx4_ib_qp **caller_qp) @@ -652,6 +747,7 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd, enum mlx4_ib_qp_type qp_type = (enum mlx4_ib_qp_type) init_attr->qp_type; struct mlx4_ib_cq *mcq; unsigned long flags; + int range_size = 0; /* When tunneling special qps, we use a plain UD qp */ if (sqpn) { @@ -728,27 +824,69 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd, if (pd->uobject) { - struct mlx4_ib_create_qp ucmd; + union { + struct mlx4_ib_create_qp qp; + struct mlx4_ib_create_wq wq; + } ucmd; + size_t copy_len; - if (ib_copy_from_udata(&ucmd, udata, sizeof ucmd)) { + copy_len = (src == MLX4_IB_QP_SRC) ? + sizeof(struct mlx4_ib_create_qp) : + min(sizeof(struct mlx4_ib_create_wq), udata->inlen); + + if (ib_copy_from_udata(&ucmd, udata, copy_len)) { err = -EFAULT; goto err; } + if (src == MLX4_IB_RWQ_SRC) { + if (ucmd.wq.comp_mask || ucmd.wq.reserved1 || + ucmd.wq.reserved[0] || ucmd.wq.reserved[1] || + ucmd.wq.reserved[2]) { + pr_debug("user command isn't supported\n"); + err = -EOPNOTSUPP; + goto err; + } + + if (ucmd.wq.log_range_size > + ilog2(dev->dev->caps.max_rss_tbl_sz)) { + pr_debug("WQN range size must be equal or smaller than %d\n", + dev->dev->caps.max_rss_tbl_sz); + err = -EOPNOTSUPP; + goto err; + } + range_size = 1 << ucmd.wq.log_range_size; + } else { + qp->inl_recv_sz = ucmd.qp.inl_recv_sz; + } + err = set_rq_size(dev, &init_attr->cap, !!pd->uobject, - qp_has_rq(init_attr), qp, ucmd.inl_recv_sz); + qp_has_rq(init_attr), qp, qp->inl_recv_sz); if (err) goto err; - qp->inl_recv_sz = ucmd.inl_recv_sz; - qp->sq_no_prefetch = ucmd.sq_no_prefetch; + if (src == MLX4_IB_QP_SRC) { + qp->sq_no_prefetch = ucmd.qp.sq_no_prefetch; - err = set_user_sq_size(dev, qp, &ucmd); - if (err) - goto err; + err = set_user_sq_size(dev, qp, + (struct mlx4_ib_create_qp *) + &ucmd); + if (err) + goto err; + } else { + qp->sq_no_prefetch = 1; + qp->sq.wqe_cnt = 1; + qp->sq.wqe_shift = MLX4_IB_MIN_SQ_STRIDE; + /* Allocated buffer expects to have at least that SQ + * size. + */ + qp->buf_size = (qp->rq.wqe_cnt << qp->rq.wqe_shift) + + (qp->sq.wqe_cnt << qp->sq.wqe_shift); + } - qp->umem = ib_umem_get(pd->uobject->context, ucmd.buf_addr, - qp->buf_size, 0, 0); + qp->umem = ib_umem_get(pd->uobject->context, + (src == MLX4_IB_QP_SRC) ? ucmd.qp.buf_addr : + ucmd.wq.buf_addr, qp->buf_size, 0, 0); if (IS_ERR(qp->umem)) { err = PTR_ERR(qp->umem); goto err; @@ -765,7 +903,8 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd, if (qp_has_rq(init_attr)) { err = mlx4_ib_db_map_user(to_mucontext(pd->uobject->context), - ucmd.db_addr, &qp->db); + (src == MLX4_IB_QP_SRC) ? ucmd.qp.db_addr : + ucmd.wq.db_addr, &qp->db); if (err) goto err_mtt; } @@ -853,6 +992,11 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd, goto err_wrid; } } + } else if (src == MLX4_IB_RWQ_SRC) { + err = mlx4_ib_alloc_wqn(to_mucontext(pd->uobject->context), qp, + range_size, &qpn); + if (err) + goto err_wrid; } else { /* Raw packet QPNs may not have bits 6,7 set in their qp_num; * otherwise, the WQE BlueFlame setup flow wrongly causes @@ -891,7 +1035,9 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd, */ qp->doorbell_qpn = swab32(qp->mqp.qpn << 8); - qp->mqp.event = mlx4_ib_qp_event; + qp->mqp.event = (src == MLX4_IB_QP_SRC) ? mlx4_ib_qp_event : + mlx4_ib_wq_event; + if (!*caller_qp) *caller_qp = qp; @@ -918,6 +1064,9 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd, if (!sqpn) { if (qp->flags & MLX4_IB_QP_NETIF) mlx4_ib_steer_qp_free(dev, qpn, 1); + else if (src == MLX4_IB_RWQ_SRC) + mlx4_ib_release_wqn(to_mucontext(pd->uobject->context), + qp, 0); else mlx4_qp_release_range(dev->dev, qpn, 1); } @@ -1016,7 +1165,7 @@ static struct mlx4_ib_pd *get_pd(struct mlx4_ib_qp *qp) return to_mpd(qp->ibqp.pd); } -static void get_cqs(struct mlx4_ib_qp *qp, +static void get_cqs(struct mlx4_ib_qp *qp, enum mlx4_ib_source_type src, struct mlx4_ib_cq **send_cq, struct mlx4_ib_cq **recv_cq) { switch (qp->ibqp.qp_type) { @@ -1029,14 +1178,16 @@ static void get_cqs(struct mlx4_ib_qp *qp, *recv_cq = *send_cq; break; default: - *send_cq = to_mcq(qp->ibqp.send_cq); - *recv_cq = to_mcq(qp->ibqp.recv_cq); + *recv_cq = (src == MLX4_IB_QP_SRC) ? to_mcq(qp->ibqp.recv_cq) : + to_mcq(qp->ibwq.cq); + *send_cq = (src == MLX4_IB_QP_SRC) ? to_mcq(qp->ibqp.send_cq) : + *recv_cq; break; } } static void destroy_qp_common(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp, - int is_user) + enum mlx4_ib_source_type src, int is_user) { struct mlx4_ib_cq *send_cq, *recv_cq; unsigned long flags; @@ -1069,7 +1220,7 @@ static void destroy_qp_common(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp, } } - get_cqs(qp, &send_cq, &recv_cq); + get_cqs(qp, src, &send_cq, &recv_cq); spin_lock_irqsave(&dev->reset_flow_resource_lock, flags); mlx4_ib_lock_cqs(send_cq, recv_cq); @@ -1095,6 +1246,9 @@ static void destroy_qp_common(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp, if (!is_sqp(dev, qp) && !is_tunnel_qp(dev, qp)) { if (qp->flags & MLX4_IB_QP_NETIF) mlx4_ib_steer_qp_free(dev, qp->mqp.qpn, 1); + else if (src == MLX4_IB_RWQ_SRC) + mlx4_ib_release_wqn(to_mucontext( + qp->ibwq.uobject->context), qp, 1); else mlx4_qp_release_range(dev->dev, qp->mqp.qpn, 1); } @@ -1102,9 +1256,12 @@ static void destroy_qp_common(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp, mlx4_mtt_cleanup(dev->dev, &qp->mtt); if (is_user) { - if (qp->rq.wqe_cnt) - mlx4_ib_db_unmap_user(to_mucontext(qp->ibqp.uobject->context), - &qp->db); + if (qp->rq.wqe_cnt) { + struct mlx4_ib_ucontext *mcontext = !src ? + to_mucontext(qp->ibqp.uobject->context) : + to_mucontext(qp->ibwq.uobject->context); + mlx4_ib_db_unmap_user(mcontext, &qp->db); + } ib_umem_release(qp->umem); } else { kvfree(qp->sq.wrid); @@ -1200,8 +1357,8 @@ static struct ib_qp *_mlx4_ib_create_qp(struct ib_pd *pd, /* fall through */ case IB_QPT_UD: { - err = create_qp_common(to_mdev(pd->device), pd, init_attr, - udata, 0, &qp); + err = create_qp_common(to_mdev(pd->device), pd, MLX4_IB_QP_SRC, + init_attr, udata, 0, &qp); if (err) { kfree(qp); return ERR_PTR(err); @@ -1231,8 +1388,8 @@ static struct ib_qp *_mlx4_ib_create_qp(struct ib_pd *pd, sqpn = get_sqp_num(to_mdev(pd->device), init_attr); } - err = create_qp_common(to_mdev(pd->device), pd, init_attr, udata, - sqpn, &qp); + err = create_qp_common(to_mdev(pd->device), pd, MLX4_IB_QP_SRC, + init_attr, udata, sqpn, &qp); if (err) return ERR_PTR(err); @@ -1303,7 +1460,7 @@ static int _mlx4_ib_destroy_qp(struct ib_qp *qp) mlx4_ib_free_qp_counter(dev, mqp); pd = get_pd(mqp); - destroy_qp_common(dev, mqp, !!pd->ibpd.uobject); + destroy_qp_common(dev, mqp, MLX4_IB_QP_SRC, !!pd->ibpd.uobject); if (is_sqp(dev, mqp)) kfree(to_msqp(mqp)); @@ -1626,12 +1783,15 @@ static u8 gid_type_to_qpc(enum ib_gid_type gid_type) } } -static int __mlx4_ib_modify_qp(struct ib_qp *ibqp, +static int __mlx4_ib_modify_qp(void *src, enum mlx4_ib_source_type src_type, const struct ib_qp_attr *attr, int attr_mask, enum ib_qp_state cur_state, enum ib_qp_state new_state) { - struct mlx4_ib_dev *dev = to_mdev(ibqp->device); - struct mlx4_ib_qp *qp = to_mqp(ibqp); + struct ib_uobject *ibuobject; + struct ib_srq *ibsrq; + enum ib_qp_type qp_type; + struct mlx4_ib_dev *dev; + struct mlx4_ib_qp *qp; struct mlx4_ib_pd *pd; struct mlx4_ib_cq *send_cq, *recv_cq; struct mlx4_qp_context *context; @@ -1641,6 +1801,28 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp, int err = -EINVAL; int counter_index; + if (src_type == MLX4_IB_RWQ_SRC) { + struct ib_wq *ibwq; + + ibwq = (struct ib_wq *)src; + ibuobject = ibwq->uobject; + ibsrq = NULL; + qp_type = IB_QPT_RAW_PACKET; + qp = to_mqp((struct ib_qp *)ibwq); + dev = to_mdev(ibwq->device); + pd = to_mpd(ibwq->pd); + } else { + struct ib_qp *ibqp; + + ibqp = (struct ib_qp *)src; + ibuobject = ibqp->uobject; + ibsrq = ibqp->srq; + qp_type = ibqp->qp_type; + qp = to_mqp(ibqp); + dev = to_mdev(ibqp->device); + pd = get_pd(qp); + } + /* APM is not supported under RoCE */ if (attr_mask & IB_QP_ALT_PATH && rdma_port_get_link_layer(&dev->ib_dev, qp->port) == @@ -1674,11 +1856,11 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp, if (qp->inl_recv_sz) context->param3 |= cpu_to_be32(1 << 25); - if (ibqp->qp_type == IB_QPT_GSI || ibqp->qp_type == IB_QPT_SMI) + if (qp_type == IB_QPT_GSI || qp_type == IB_QPT_SMI) context->mtu_msgmax = (IB_MTU_4096 << 5) | 11; - else if (ibqp->qp_type == IB_QPT_RAW_PACKET) + else if (qp_type == IB_QPT_RAW_PACKET) context->mtu_msgmax = (MLX4_RAW_QP_MTU << 5) | MLX4_RAW_QP_MSGMAX; - else if (ibqp->qp_type == IB_QPT_UD) { + else if (qp_type == IB_QPT_UD) { if (qp->flags & MLX4_IB_QP_LSO) context->mtu_msgmax = (IB_MTU_4096 << 5) | ilog2(dev->dev->caps.max_gso_sz); @@ -1708,14 +1890,15 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp, if (cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT) { context->sq_size_stride |= !!qp->sq_no_prefetch << 7; context->xrcd = cpu_to_be32((u32) qp->xrcdn); - if (ibqp->qp_type == IB_QPT_RAW_PACKET) + if (qp_type == IB_QPT_RAW_PACKET) context->param3 |= cpu_to_be32(1 << 30); } - if (qp->ibqp.uobject) + if (ibuobject) context->usr_page = cpu_to_be32( mlx4_to_hw_uar_index(dev->dev, - to_mucontext(ibqp->uobject->context)->uar.index)); + to_mucontext(ibuobject->context) + ->uar.index)); else context->usr_page = cpu_to_be32( mlx4_to_hw_uar_index(dev->dev, dev->priv_uar.index)); @@ -1759,7 +1942,7 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp, steer_qp = 1; } - if (ibqp->qp_type == IB_QPT_GSI) { + if (qp_type == IB_QPT_GSI) { enum ib_gid_type gid_type = qp->flags & MLX4_IB_ROCE_V2_GSI_QP ? IB_GID_TYPE_ROCE_UDP_ENCAP : IB_GID_TYPE_ROCE; u8 qpc_roce_mode = gid_type_to_qpc(gid_type); @@ -1776,7 +1959,7 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp, } if (attr_mask & IB_QP_AV) { - u8 port_num = mlx4_is_bonded(to_mdev(ibqp->device)->dev) ? 1 : + u8 port_num = mlx4_is_bonded(dev->dev) ? 1 : attr_mask & IB_QP_PORT ? attr->port_num : qp->port; union ib_gid gid; struct ib_gid_attr gid_attr = {.gid_type = IB_GID_TYPE_IB}; @@ -1791,7 +1974,7 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp, int index = rdma_ah_read_grh(&attr->ah_attr)->sgid_index; - status = ib_get_cached_gid(ibqp->device, port_num, + status = ib_get_cached_gid(&dev->ib_dev, port_num, index, &gid, &gid_attr); if (!status && !memcmp(&gid, &zgid, sizeof(gid))) status = -ENOENT; @@ -1848,15 +2031,14 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp, optpar |= MLX4_QP_OPTPAR_ALT_ADDR_PATH; } - pd = get_pd(qp); - get_cqs(qp, &send_cq, &recv_cq); + get_cqs(qp, src_type, &send_cq, &recv_cq); context->pd = cpu_to_be32(pd->pdn); context->cqn_send = cpu_to_be32(send_cq->mcq.cqn); context->cqn_recv = cpu_to_be32(recv_cq->mcq.cqn); context->params1 = cpu_to_be32(MLX4_IB_ACK_REQ_FREQ << 28); /* Set "fast registration enabled" for all kernel QPs */ - if (!qp->ibqp.uobject) + if (!ibuobject) context->params1 |= cpu_to_be32(1 << 11); if (attr_mask & IB_QP_RNR_RETRY) { @@ -1891,7 +2073,7 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp, optpar |= MLX4_QP_OPTPAR_RWE | MLX4_QP_OPTPAR_RRE | MLX4_QP_OPTPAR_RAE; } - if (ibqp->srq) + if (ibsrq) context->params2 |= cpu_to_be32(MLX4_QP_BIT_RIC); if (attr_mask & IB_QP_MIN_RNR_TIMER) { @@ -1922,17 +2104,19 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp, optpar |= MLX4_QP_OPTPAR_Q_KEY; } - if (ibqp->srq) - context->srqn = cpu_to_be32(1 << 24 | to_msrq(ibqp->srq)->msrq.srqn); + if (ibsrq) + context->srqn = cpu_to_be32(1 << 24 | + to_msrq(ibsrq)->msrq.srqn); - if (qp->rq.wqe_cnt && cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT) + if (qp->rq.wqe_cnt && + cur_state == IB_QPS_RESET && + new_state == IB_QPS_INIT) context->db_rec_addr = cpu_to_be64(qp->db.dma); if (cur_state == IB_QPS_INIT && new_state == IB_QPS_RTR && - (ibqp->qp_type == IB_QPT_GSI || ibqp->qp_type == IB_QPT_SMI || - ibqp->qp_type == IB_QPT_UD || - ibqp->qp_type == IB_QPT_RAW_PACKET)) { + (qp_type == IB_QPT_GSI || qp_type == IB_QPT_SMI || + qp_type == IB_QPT_UD || qp_type == IB_QPT_RAW_PACKET)) { context->pri_path.sched_queue = (qp->port - 1) << 6; if (qp->mlx4_ib_qp_type == MLX4_IB_QPT_SMI || qp->mlx4_ib_qp_type & @@ -1965,7 +2149,7 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp, } } - if (qp->ibqp.qp_type == IB_QPT_RAW_PACKET) { + if (qp_type == IB_QPT_RAW_PACKET) { context->pri_path.ackto = (context->pri_path.ackto & 0xf8) | MLX4_IB_LINK_TYPE_ETH; if (dev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) { @@ -1975,7 +2159,7 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp, } } - if (ibqp->qp_type == IB_QPT_UD && (new_state == IB_QPS_RTR)) { + if (qp_type == IB_QPT_UD && (new_state == IB_QPS_RTR)) { int is_eth = rdma_port_get_link_layer( &dev->ib_dev, qp->port) == IB_LINK_LAYER_ETHERNET; @@ -1985,14 +2169,15 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp, } } - if (cur_state == IB_QPS_RTS && new_state == IB_QPS_SQD && attr_mask & IB_QP_EN_SQD_ASYNC_NOTIFY && attr->en_sqd_async_notify) sqd_event = 1; else sqd_event = 0; - if (!ibqp->uobject && cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT) + if (!ibuobject && + cur_state == IB_QPS_RESET && + new_state == IB_QPS_INIT) context->rlkey_roce_mode |= (1 << 4); /* @@ -2001,7 +2186,9 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp, * headroom is stamped so that the hardware doesn't start * processing stale work requests. */ - if (!ibqp->uobject && cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT) { + if (!ibuobject && + cur_state == IB_QPS_RESET && + new_state == IB_QPS_INIT) { struct mlx4_wqe_ctrl_seg *ctrl; int i; @@ -2058,9 +2245,9 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp, * entries and reinitialize the QP. */ if (new_state == IB_QPS_RESET) { - if (!ibqp->uobject) { + if (!ibuobject) { mlx4_ib_cq_clean(recv_cq, qp->mqp.qpn, - ibqp->srq ? to_msrq(ibqp->srq) : NULL); + ibsrq ? to_msrq(ibsrq) : NULL); if (send_cq != recv_cq) mlx4_ib_cq_clean(send_cq, qp->mqp.qpn, NULL); @@ -2265,7 +2452,8 @@ static int _mlx4_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, goto out; } - err = __mlx4_ib_modify_qp(ibqp, attr, attr_mask, cur_state, new_state); + err = __mlx4_ib_modify_qp(ibqp, MLX4_IB_QP_SRC, attr, attr_mask, + cur_state, new_state); if (mlx4_is_bonded(dev->dev) && (attr_mask & IB_QP_PORT)) attr->port_num = 1; @@ -3550,3 +3738,208 @@ int mlx4_ib_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *qp_attr, int qp_attr return err; } +struct ib_wq *mlx4_ib_create_wq(struct ib_pd *pd, + struct ib_wq_init_attr *init_attr, + struct ib_udata *udata) +{ + struct mlx4_ib_dev *dev; + struct ib_qp_init_attr ib_qp_init_attr; + struct mlx4_ib_qp *qp; + struct mlx4_ib_create_wq ucmd; + int err, required_cmd_sz; + + if (!(udata && pd->uobject)) + return ERR_PTR(-EINVAL); + + required_cmd_sz = offsetof(typeof(ucmd), reserved) + + sizeof(ucmd.reserved); + if (udata->inlen < required_cmd_sz) { + pr_debug("invalid inlen\n"); + return ERR_PTR(-EINVAL); + } + + if (udata->inlen > sizeof(ucmd) && + !ib_is_udata_cleared(udata, sizeof(ucmd), + udata->inlen - sizeof(ucmd))) { + pr_debug("inlen is not supported\n"); + return ERR_PTR(-EOPNOTSUPP); + } + + if (udata->outlen) + return ERR_PTR(-EOPNOTSUPP); + + dev = to_mdev(pd->device); + + if (init_attr->wq_type != IB_WQT_RQ) { + pr_debug("unsupported wq type %d\n", init_attr->wq_type); + return ERR_PTR(-EOPNOTSUPP); + } + + if (init_attr->create_flags) { + pr_debug("unsupported create_flags %u\n", + init_attr->create_flags); + return ERR_PTR(-EOPNOTSUPP); + } + + qp = kzalloc(sizeof(*qp), GFP_KERNEL); + if (!qp) + return ERR_PTR(-ENOMEM); + + qp->pri.vid = 0xFFFF; + qp->alt.vid = 0xFFFF; + + memset(&ib_qp_init_attr, 0, sizeof(ib_qp_init_attr)); + ib_qp_init_attr.qp_context = init_attr->wq_context; + ib_qp_init_attr.qp_type = IB_QPT_RAW_PACKET; + ib_qp_init_attr.cap.max_recv_wr = init_attr->max_wr; + ib_qp_init_attr.cap.max_recv_sge = init_attr->max_sge; + ib_qp_init_attr.recv_cq = init_attr->cq; + ib_qp_init_attr.send_cq = ib_qp_init_attr.recv_cq; /* Dummy CQ */ + + err = create_qp_common(dev, pd, MLX4_IB_RWQ_SRC, &ib_qp_init_attr, + udata, 0, &qp); + if (err) { + kfree(qp); + return ERR_PTR(err); + } + + qp->ibwq.event_handler = init_attr->event_handler; + qp->ibwq.wq_num = qp->mqp.qpn; + qp->ibwq.state = IB_WQS_RESET; + + return &qp->ibwq; +} + +static int ib_wq2qp_state(enum ib_wq_state state) +{ + switch (state) { + case IB_WQS_RESET: + return IB_QPS_RESET; + case IB_WQS_RDY: + return IB_QPS_RTR; + default: + return IB_QPS_ERR; + } +} + +static int _mlx4_ib_modify_wq(struct ib_wq *ibwq, enum ib_wq_state new_state) +{ + struct mlx4_ib_qp *qp = to_mqp((struct ib_qp *)ibwq); + enum ib_qp_state qp_cur_state; + enum ib_qp_state qp_new_state; + int attr_mask; + int err; + + /* ib_qp.state represents the WQ HW state while ib_wq.state represents + * the WQ logic state. + */ + qp_cur_state = qp->state; + qp_new_state = ib_wq2qp_state(new_state); + + if (ib_wq2qp_state(new_state) == qp_cur_state) + return 0; + + if (new_state == IB_WQS_RDY) { + struct ib_qp_attr attr = {}; + + attr.port_num = qp->port; + attr_mask = IB_QP_PORT; + + err = __mlx4_ib_modify_qp(ibwq, MLX4_IB_RWQ_SRC, &attr, + attr_mask, IB_QPS_RESET, IB_QPS_INIT); + if (err) { + pr_debug("WQN=0x%06x failed to apply RST->INIT on the HW QP\n", + ibwq->wq_num); + return err; + } + + qp_cur_state = IB_QPS_INIT; + } + + attr_mask = 0; + err = __mlx4_ib_modify_qp(ibwq, MLX4_IB_RWQ_SRC, NULL, attr_mask, + qp_cur_state, qp_new_state); + + if (err && (qp_cur_state == IB_QPS_INIT)) { + qp_new_state = IB_QPS_RESET; + if (__mlx4_ib_modify_qp(ibwq, MLX4_IB_RWQ_SRC, NULL, + attr_mask, IB_QPS_INIT, IB_QPS_RESET)) { + pr_warn("WQN=0x%06x failed with reverting HW's resources failure\n", + ibwq->wq_num); + qp_new_state = IB_QPS_INIT; + } + } + + qp->state = qp_new_state; + + return err; +} + +int mlx4_ib_modify_wq(struct ib_wq *ibwq, struct ib_wq_attr *wq_attr, + u32 wq_attr_mask, struct ib_udata *udata) +{ + struct mlx4_ib_qp *qp = to_mqp((struct ib_qp *)ibwq); + struct mlx4_ib_modify_wq ucmd = {}; + size_t required_cmd_sz; + enum ib_wq_state cur_state, new_state; + int err = 0; + + required_cmd_sz = offsetof(typeof(ucmd), reserved) + + sizeof(ucmd.reserved); + if (udata->inlen < required_cmd_sz) + return -EINVAL; + + if (udata->inlen > sizeof(ucmd) && + !ib_is_udata_cleared(udata, sizeof(ucmd), + udata->inlen - sizeof(ucmd))) + return -EOPNOTSUPP; + + if (ib_copy_from_udata(&ucmd, udata, min(sizeof(ucmd), udata->inlen))) + return -EFAULT; + + if (ucmd.comp_mask || ucmd.reserved) + return -EOPNOTSUPP; + + if (wq_attr_mask & IB_WQ_FLAGS) + return -EOPNOTSUPP; + + cur_state = wq_attr_mask & IB_WQ_CUR_STATE ? wq_attr->curr_wq_state : + ibwq->state; + new_state = wq_attr_mask & IB_WQ_STATE ? wq_attr->wq_state : cur_state; + + if (cur_state < IB_WQS_RESET || cur_state > IB_WQS_ERR || + new_state < IB_WQS_RESET || new_state > IB_WQS_ERR) + return -EINVAL; + + if ((new_state == IB_WQS_RDY) && (cur_state == IB_WQS_ERR)) + return -EINVAL; + + if ((new_state == IB_WQS_ERR) && (cur_state == IB_WQS_RESET)) + return -EINVAL; + + /* Can update HW state only if a RSS QP has already associated to this + * WQ, so we can apply its port on the WQ. + */ + if (qp->rss_usecnt) + err = _mlx4_ib_modify_wq(ibwq, new_state); + + if (!err) + ibwq->state = new_state; + + return err; +} + +int mlx4_ib_destroy_wq(struct ib_wq *ibwq) +{ + struct mlx4_ib_dev *dev = to_mdev(ibwq->device); + struct mlx4_ib_qp *qp = to_mqp((struct ib_qp *)ibwq); + + if (qp->counter_index) + mlx4_ib_free_qp_counter(dev, qp); + + destroy_qp_common(dev, qp, MLX4_IB_RWQ_SRC, 1); + + kfree(qp); + + return 0; +} diff --git a/include/uapi/rdma/mlx4-abi.h b/include/uapi/rdma/mlx4-abi.h index bf3bdba2f326..c9702a5f0bda 100644 --- a/include/uapi/rdma/mlx4-abi.h +++ b/include/uapi/rdma/mlx4-abi.h @@ -105,4 +105,18 @@ struct mlx4_ib_create_qp { __u8 reserved; }; +struct mlx4_ib_create_wq { + __u64 buf_addr; + __u64 db_addr; + __u8 log_range_size; + __u8 reserved[3]; + __u32 comp_mask; + __u32 reserved1; +}; + +struct mlx4_ib_modify_wq { + __u32 comp_mask; + __u32 reserved; +}; + #endif /* MLX4_ABI_USER_H */ From b8d46ca035060e70f5f0da849d86720752d5aa17 Mon Sep 17 00:00:00 2001 From: Guy Levi Date: Tue, 4 Jul 2017 16:24:25 +0300 Subject: [PATCH 32/34] IB/mlx4: Add support for WQ indirection table related verbs To enable RSS functionality the IB indirection table object (i.e. ib_rwq_ind_table) should be used. This patch implements the related verbs as of create and destroy an indirection table. In downstream patches the indirection table will be used as part of RSS QP creation. Signed-off-by: Guy Levi Reviewed-by: Yishai Hadas Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx4/main.c | 12 +++-- drivers/infiniband/hw/mlx4/mlx4_ib.h | 6 +++ drivers/infiniband/hw/mlx4/qp.c | 70 ++++++++++++++++++++++++++++ include/uapi/rdma/mlx4-abi.h | 4 ++ 4 files changed, 89 insertions(+), 3 deletions(-) diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c index 7c6f929ebd3e..b42234571a8c 100644 --- a/drivers/infiniband/hw/mlx4/main.c +++ b/drivers/infiniband/hw/mlx4/main.c @@ -2738,10 +2738,16 @@ static void *mlx4_ib_add(struct mlx4_dev *dev) 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_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); } if (!mlx4_is_slave(ibdev->dev)) { diff --git a/drivers/infiniband/hw/mlx4/mlx4_ib.h b/drivers/infiniband/hw/mlx4/mlx4_ib.h index ba4e78c064d2..85525bc400a0 100644 --- a/drivers/infiniband/hw/mlx4/mlx4_ib.h +++ b/drivers/infiniband/hw/mlx4/mlx4_ib.h @@ -916,4 +916,10 @@ int mlx4_ib_destroy_wq(struct ib_wq *wq); int mlx4_ib_modify_wq(struct ib_wq *wq, struct ib_wq_attr *wq_attr, u32 wq_attr_mask, struct ib_udata *udata); +struct ib_rwq_ind_table +*mlx4_ib_create_rwq_ind_table(struct ib_device *device, + struct ib_rwq_ind_table_init_attr *init_attr, + struct ib_udata *udata); +int mlx4_ib_destroy_rwq_ind_table(struct ib_rwq_ind_table *wq_ind_table); + #endif /* MLX4_IB_H */ diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c index 65e4ec549368..519919d15474 100644 --- a/drivers/infiniband/hw/mlx4/qp.c +++ b/drivers/infiniband/hw/mlx4/qp.c @@ -3943,3 +3943,73 @@ int mlx4_ib_destroy_wq(struct ib_wq *ibwq) return 0; } + +struct ib_rwq_ind_table +*mlx4_ib_create_rwq_ind_table(struct ib_device *device, + struct ib_rwq_ind_table_init_attr *init_attr, + struct ib_udata *udata) +{ + struct ib_rwq_ind_table *rwq_ind_table; + struct mlx4_ib_create_rwq_ind_tbl_resp resp = {}; + unsigned int ind_tbl_size = 1 << init_attr->log_ind_tbl_size; + unsigned int base_wqn; + size_t min_resp_len; + int i; + int err; + + if (udata->inlen > 0 && + !ib_is_udata_cleared(udata, 0, + udata->inlen)) + return ERR_PTR(-EOPNOTSUPP); + + min_resp_len = offsetof(typeof(resp), reserved) + sizeof(resp.reserved); + if (udata->outlen && udata->outlen < min_resp_len) + return ERR_PTR(-EINVAL); + + if (ind_tbl_size > + device->attrs.rss_caps.max_rwq_indirection_table_size) { + pr_debug("log_ind_tbl_size = %d is bigger than supported = %d\n", + ind_tbl_size, + device->attrs.rss_caps.max_rwq_indirection_table_size); + return ERR_PTR(-EINVAL); + } + + base_wqn = init_attr->ind_tbl[0]->wq_num; + + if (base_wqn % ind_tbl_size) { + pr_debug("WQN=0x%x isn't aligned with indirection table size\n", + base_wqn); + return ERR_PTR(-EINVAL); + } + + for (i = 1; i < ind_tbl_size; i++) { + if (++base_wqn != init_attr->ind_tbl[i]->wq_num) { + pr_debug("indirection table's WQNs aren't consecutive\n"); + return ERR_PTR(-EINVAL); + } + } + + rwq_ind_table = kzalloc(sizeof(*rwq_ind_table), GFP_KERNEL); + if (!rwq_ind_table) + return ERR_PTR(-ENOMEM); + + if (udata->outlen) { + resp.response_length = offsetof(typeof(resp), response_length) + + sizeof(resp.response_length); + err = ib_copy_to_udata(udata, &resp, resp.response_length); + if (err) + goto err; + } + + return rwq_ind_table; + +err: + kfree(rwq_ind_table); + return ERR_PTR(err); +} + +int mlx4_ib_destroy_rwq_ind_table(struct ib_rwq_ind_table *ib_rwq_ind_tbl) +{ + kfree(ib_rwq_ind_tbl); + return 0; +} diff --git a/include/uapi/rdma/mlx4-abi.h b/include/uapi/rdma/mlx4-abi.h index c9702a5f0bda..5591d955ba00 100644 --- a/include/uapi/rdma/mlx4-abi.h +++ b/include/uapi/rdma/mlx4-abi.h @@ -119,4 +119,8 @@ struct mlx4_ib_modify_wq { __u32 reserved; }; +struct mlx4_ib_create_rwq_ind_tbl_resp { + __u32 response_length; + __u32 reserved; +}; #endif /* MLX4_ABI_USER_H */ From 3078f5f1bd8b6c8aef77b8ef4d49671fa6eb058e Mon Sep 17 00:00:00 2001 From: Guy Levi Date: Tue, 4 Jul 2017 16:24:26 +0300 Subject: [PATCH 33/34] IB/mlx4: Add support for RSS QP Add support to work with a RSS QP by using an indirection table object upon QP creation. Other related QP verbs (e.g. modify/destroy/query) were updated as well for that QP mode. Notes: - The RX hash properties are supplied as driver private data. - The RSS QP port is used on the associated WQs in its indirection table. Applying different ports during WQ life time is not allowed. - The expected RSS QP flow is: create, modify(RST->INIT), modify(RST->RTR), destroy. Signed-off-by: Guy Levi Reviewed-by: Yishai Hadas Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx4/mlx4_ib.h | 8 + drivers/infiniband/hw/mlx4/qp.c | 453 +++++++++++++++++++++++++-- include/uapi/rdma/mlx4-abi.h | 33 ++ 3 files changed, 472 insertions(+), 22 deletions(-) diff --git a/drivers/infiniband/hw/mlx4/mlx4_ib.h b/drivers/infiniband/hw/mlx4/mlx4_ib.h index 85525bc400a0..1fa19820355a 100644 --- a/drivers/infiniband/hw/mlx4/mlx4_ib.h +++ b/drivers/infiniband/hw/mlx4/mlx4_ib.h @@ -46,6 +46,7 @@ #include #include +#include #define MLX4_IB_DRV_NAME "mlx4_ib" @@ -299,6 +300,12 @@ struct mlx4_wqn_range { struct list_head list; }; +struct mlx4_ib_rss { + unsigned int base_qpn_tbl_sz; + u8 flags; + u8 rss_key[MLX4_EN_RSS_KEY_SIZE]; +}; + struct mlx4_ib_qp { union { struct ib_qp ibqp; @@ -345,6 +352,7 @@ struct mlx4_ib_qp { struct mlx4_wqn_range *wqn_range; /* Number of RSS QP parents that uses this WQ */ u32 rss_usecnt; + struct mlx4_ib_rss *rss_ctx; }; struct mlx4_ib_srq { diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c index 519919d15474..e42acfb20588 100644 --- a/drivers/infiniband/hw/mlx4/qp.c +++ b/drivers/infiniband/hw/mlx4/qp.c @@ -53,6 +53,7 @@ static void mlx4_ib_lock_cqs(struct mlx4_ib_cq *send_cq, struct mlx4_ib_cq *recv_cq); static void mlx4_ib_unlock_cqs(struct mlx4_ib_cq *send_cq, struct mlx4_ib_cq *recv_cq); +static int _mlx4_ib_modify_wq(struct ib_wq *ibwq, enum ib_wq_state new_state); enum { MLX4_IB_ACK_REQ_FREQ = 8, @@ -650,6 +651,212 @@ static void mlx4_ib_free_qp_counter(struct mlx4_ib_dev *dev, qp->counter_index = NULL; } +static int set_qp_rss(struct mlx4_ib_dev *dev, struct mlx4_ib_rss *rss_ctx, + struct ib_qp_init_attr *init_attr, + struct mlx4_ib_create_qp_rss *ucmd) +{ + rss_ctx->base_qpn_tbl_sz = init_attr->rwq_ind_tbl->ind_tbl[0]->wq_num | + (init_attr->rwq_ind_tbl->log_ind_tbl_size << 24); + + if ((ucmd->rx_hash_function == MLX4_IB_RX_HASH_FUNC_TOEPLITZ) && + (dev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_RSS_TOP)) { + memcpy(rss_ctx->rss_key, ucmd->rx_hash_key, + MLX4_EN_RSS_KEY_SIZE); + } else { + pr_debug("RX Hash function is not supported\n"); + return (-EOPNOTSUPP); + } + + if ((ucmd->rx_hash_fields_mask & MLX4_IB_RX_HASH_SRC_IPV4) && + (ucmd->rx_hash_fields_mask & MLX4_IB_RX_HASH_DST_IPV4)) { + rss_ctx->flags = MLX4_RSS_IPV4; + } else if ((ucmd->rx_hash_fields_mask & MLX4_IB_RX_HASH_SRC_IPV4) || + (ucmd->rx_hash_fields_mask & MLX4_IB_RX_HASH_DST_IPV4)) { + pr_debug("RX Hash fields_mask is not supported - both IPv4 SRC and DST must be set\n"); + return (-EOPNOTSUPP); + } + + if ((ucmd->rx_hash_fields_mask & MLX4_IB_RX_HASH_SRC_IPV6) && + (ucmd->rx_hash_fields_mask & MLX4_IB_RX_HASH_DST_IPV6)) { + rss_ctx->flags |= MLX4_RSS_IPV6; + } else if ((ucmd->rx_hash_fields_mask & MLX4_IB_RX_HASH_SRC_IPV6) || + (ucmd->rx_hash_fields_mask & MLX4_IB_RX_HASH_DST_IPV6)) { + pr_debug("RX Hash fields_mask is not supported - both IPv6 SRC and DST must be set\n"); + return (-EOPNOTSUPP); + } + + if ((ucmd->rx_hash_fields_mask & MLX4_IB_RX_HASH_SRC_PORT_UDP) && + (ucmd->rx_hash_fields_mask & MLX4_IB_RX_HASH_DST_PORT_UDP)) { + if (!(dev->dev->caps.flags & MLX4_DEV_CAP_FLAG_UDP_RSS)) { + pr_debug("RX Hash fields_mask for UDP is not supported\n"); + return (-EOPNOTSUPP); + } + + if (rss_ctx->flags & MLX4_RSS_IPV4) { + rss_ctx->flags |= MLX4_RSS_UDP_IPV4; + } else if (rss_ctx->flags & MLX4_RSS_IPV6) { + rss_ctx->flags |= MLX4_RSS_UDP_IPV6; + } else { + pr_debug("RX Hash fields_mask is not supported - UDP must be set with IPv4 or IPv6\n"); + return (-EOPNOTSUPP); + } + } else if ((ucmd->rx_hash_fields_mask & MLX4_IB_RX_HASH_SRC_PORT_UDP) || + (ucmd->rx_hash_fields_mask & MLX4_IB_RX_HASH_DST_PORT_UDP)) { + pr_debug("RX Hash fields_mask is not supported - both UDP SRC and DST must be set\n"); + return (-EOPNOTSUPP); + } + + if ((ucmd->rx_hash_fields_mask & MLX4_IB_RX_HASH_SRC_PORT_TCP) && + (ucmd->rx_hash_fields_mask & MLX4_IB_RX_HASH_DST_PORT_TCP)) { + if (rss_ctx->flags & MLX4_RSS_IPV4) { + rss_ctx->flags |= MLX4_RSS_TCP_IPV4; + } else if (rss_ctx->flags & MLX4_RSS_IPV6) { + rss_ctx->flags |= MLX4_RSS_TCP_IPV6; + } else { + pr_debug("RX Hash fields_mask is not supported - TCP must be set with IPv4 or IPv6\n"); + return (-EOPNOTSUPP); + } + + } else if ((ucmd->rx_hash_fields_mask & MLX4_IB_RX_HASH_SRC_PORT_TCP) || + (ucmd->rx_hash_fields_mask & MLX4_IB_RX_HASH_DST_PORT_TCP)) { + pr_debug("RX Hash fields_mask is not supported - both TCP SRC and DST must be set\n"); + return (-EOPNOTSUPP); + } + + return 0; +} + +static int create_qp_rss(struct mlx4_ib_dev *dev, struct ib_pd *ibpd, + struct ib_qp_init_attr *init_attr, + struct mlx4_ib_create_qp_rss *ucmd, + struct mlx4_ib_qp *qp) +{ + int qpn; + int err; + + qp->mqp.usage = MLX4_RES_USAGE_USER_VERBS; + + err = mlx4_qp_reserve_range(dev->dev, 1, 1, &qpn, 0, qp->mqp.usage); + if (err) + return err; + + err = mlx4_qp_alloc(dev->dev, qpn, &qp->mqp); + if (err) + goto err_qpn; + + mutex_init(&qp->mutex); + + INIT_LIST_HEAD(&qp->gid_list); + INIT_LIST_HEAD(&qp->steering_rules); + + qp->mlx4_ib_qp_type = MLX4_IB_QPT_RAW_ETHERTYPE; + qp->state = IB_QPS_RESET; + + /* Set dummy send resources to be compatible with HV and PRM */ + qp->sq_no_prefetch = 1; + qp->sq.wqe_cnt = 1; + qp->sq.wqe_shift = MLX4_IB_MIN_SQ_STRIDE; + qp->buf_size = qp->sq.wqe_cnt << MLX4_IB_MIN_SQ_STRIDE; + qp->mtt = (to_mqp( + (struct ib_qp *)init_attr->rwq_ind_tbl->ind_tbl[0]))->mtt; + + qp->rss_ctx = kzalloc(sizeof(*qp->rss_ctx), GFP_KERNEL); + if (!qp->rss_ctx) { + err = -ENOMEM; + goto err_qp_alloc; + } + + err = set_qp_rss(dev, qp->rss_ctx, init_attr, ucmd); + if (err) + goto err; + + return 0; + +err: + kfree(qp->rss_ctx); + +err_qp_alloc: + mlx4_qp_remove(dev->dev, &qp->mqp); + mlx4_qp_free(dev->dev, &qp->mqp); + +err_qpn: + mlx4_qp_release_range(dev->dev, qpn, 1); + return err; +} + +static struct ib_qp *_mlx4_ib_create_qp_rss(struct ib_pd *pd, + struct ib_qp_init_attr *init_attr, + struct ib_udata *udata) +{ + struct mlx4_ib_qp *qp; + struct mlx4_ib_create_qp_rss ucmd = {}; + size_t required_cmd_sz; + int err; + + if (!udata) { + pr_debug("RSS QP with NULL udata\n"); + return ERR_PTR(-EINVAL); + } + + if (udata->outlen) + return ERR_PTR(-EOPNOTSUPP); + + required_cmd_sz = offsetof(typeof(ucmd), reserved1) + + sizeof(ucmd.reserved1); + if (udata->inlen < required_cmd_sz) { + pr_debug("invalid inlen\n"); + return ERR_PTR(-EINVAL); + } + + if (ib_copy_from_udata(&ucmd, udata, min(sizeof(ucmd), udata->inlen))) { + pr_debug("copy failed\n"); + return ERR_PTR(-EFAULT); + } + + if (ucmd.comp_mask || ucmd.reserved1) + return ERR_PTR(-EOPNOTSUPP); + + if (udata->inlen > sizeof(ucmd) && + !ib_is_udata_cleared(udata, sizeof(ucmd), + udata->inlen - sizeof(ucmd))) { + pr_debug("inlen is not supported\n"); + return ERR_PTR(-EOPNOTSUPP); + } + + if (init_attr->qp_type != IB_QPT_RAW_PACKET) { + pr_debug("RSS QP with unsupported QP type %d\n", + init_attr->qp_type); + return ERR_PTR(-EOPNOTSUPP); + } + + if (init_attr->create_flags) { + pr_debug("RSS QP doesn't support create flags\n"); + return ERR_PTR(-EOPNOTSUPP); + } + + if (init_attr->send_cq || init_attr->cap.max_send_wr) { + pr_debug("RSS QP with unsupported send attributes\n"); + return ERR_PTR(-EOPNOTSUPP); + } + + qp = kzalloc(sizeof(*qp), GFP_KERNEL); + if (!qp) + return ERR_PTR(-ENOMEM); + + qp->pri.vid = 0xFFFF; + qp->alt.vid = 0xFFFF; + + err = create_qp_rss(to_mdev(pd->device), pd, init_attr, &ucmd, qp); + if (err) { + kfree(qp); + return ERR_PTR(err); + } + + qp->ibqp.qp_num = qp->mqp.qpn; + + return &qp->ibqp; +} + /* * This function allocates a WQN from a range which is consecutive and aligned * to its size. In case the range is full, then it creates a new range and @@ -1186,6 +1393,36 @@ static void get_cqs(struct mlx4_ib_qp *qp, enum mlx4_ib_source_type src, } } +static void destroy_qp_rss(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp) +{ + if (qp->state != IB_QPS_RESET) { + int i; + + for (i = 0; i < (1 << qp->ibqp.rwq_ind_tbl->log_ind_tbl_size); + i++) { + struct ib_wq *ibwq = qp->ibqp.rwq_ind_tbl->ind_tbl[i]; + struct mlx4_ib_qp *wq = to_mqp((struct ib_qp *)ibwq); + + mutex_lock(&wq->mutex); + + wq->rss_usecnt--; + + mutex_unlock(&wq->mutex); + } + + if (mlx4_qp_modify(dev->dev, NULL, to_mlx4_state(qp->state), + MLX4_QP_STATE_RST, NULL, 0, 0, &qp->mqp)) + pr_warn("modify QP %06x to RESET failed.\n", + qp->mqp.qpn); + } + + mlx4_qp_remove(dev->dev, &qp->mqp); + mlx4_qp_free(dev->dev, &qp->mqp); + mlx4_qp_release_range(dev->dev, qp->mqp.qpn, 1); + del_gid_entries(qp); + kfree(qp->rss_ctx); +} + static void destroy_qp_common(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp, enum mlx4_ib_source_type src, int is_user) { @@ -1303,6 +1540,9 @@ static struct ib_qp *_mlx4_ib_create_qp(struct ib_pd *pd, int sup_u_create_flags = MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK; u16 xrcdn = 0; + if (init_attr->rwq_ind_tbl) + return _mlx4_ib_create_qp_rss(pd, init_attr, udata); + /* * We only support LSO, vendor flag1, and multicast loopback blocking, * and only for kernel UD QPs. @@ -1444,7 +1684,6 @@ static int _mlx4_ib_destroy_qp(struct ib_qp *qp) { struct mlx4_ib_dev *dev = to_mdev(qp->device); struct mlx4_ib_qp *mqp = to_mqp(qp); - struct mlx4_ib_pd *pd; if (is_qp0(dev, mqp)) mlx4_CLOSE_PORT(dev->dev, mqp->port); @@ -1459,8 +1698,14 @@ static int _mlx4_ib_destroy_qp(struct ib_qp *qp) if (mqp->counter_index) mlx4_ib_free_qp_counter(dev, mqp); - pd = get_pd(mqp); - destroy_qp_common(dev, mqp, MLX4_IB_QP_SRC, !!pd->ibpd.uobject); + 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, !!pd->ibpd.uobject); + } if (is_sqp(dev, mqp)) kfree(to_msqp(mqp)); @@ -1783,12 +2028,116 @@ static u8 gid_type_to_qpc(enum ib_gid_type gid_type) } } +/* + * Go over all RSS QP's childes (WQs) and apply their HW state according to + * their logic state if the RSS QP is the first RSS QP associated for the WQ. + */ +static int bringup_rss_rwqs(struct ib_rwq_ind_table *ind_tbl, u8 port_num) +{ + int i; + int err; + + for (i = 0; i < (1 << ind_tbl->log_ind_tbl_size); i++) { + struct ib_wq *ibwq = ind_tbl->ind_tbl[i]; + struct mlx4_ib_qp *wq = to_mqp((struct ib_qp *)ibwq); + + mutex_lock(&wq->mutex); + + /* Mlx4_ib restrictions: + * WQ's is associated to a port according to the RSS QP it is + * associates to. + * In case the WQ is associated to a different port by another + * RSS QP, return a failure. + */ + if ((wq->rss_usecnt > 0) && (wq->port != port_num)) { + err = -EINVAL; + mutex_unlock(&wq->mutex); + break; + } + wq->port = port_num; + if ((wq->rss_usecnt == 0) && (ibwq->state == IB_WQS_RDY)) { + err = _mlx4_ib_modify_wq(ibwq, IB_WQS_RDY); + if (err) { + mutex_unlock(&wq->mutex); + break; + } + } + wq->rss_usecnt++; + + mutex_unlock(&wq->mutex); + } + + if (i && err) { + int j; + + for (j = (i - 1); j >= 0; j--) { + struct ib_wq *ibwq = ind_tbl->ind_tbl[j]; + struct mlx4_ib_qp *wq = to_mqp((struct ib_qp *)ibwq); + + mutex_lock(&wq->mutex); + + if ((wq->rss_usecnt == 1) && + (ibwq->state == IB_WQS_RDY)) + if (_mlx4_ib_modify_wq(ibwq, IB_WQS_RESET)) + pr_warn("failed to reverse WQN=0x%06x\n", + ibwq->wq_num); + wq->rss_usecnt--; + + mutex_unlock(&wq->mutex); + } + } + + return err; +} + +static void bring_down_rss_rwqs(struct ib_rwq_ind_table *ind_tbl) +{ + int i; + + for (i = 0; i < (1 << ind_tbl->log_ind_tbl_size); i++) { + struct ib_wq *ibwq = ind_tbl->ind_tbl[i]; + struct mlx4_ib_qp *wq = to_mqp((struct ib_qp *)ibwq); + + mutex_lock(&wq->mutex); + + if ((wq->rss_usecnt == 1) && (ibwq->state == IB_WQS_RDY)) + if (_mlx4_ib_modify_wq(ibwq, IB_WQS_RESET)) + pr_warn("failed to reverse WQN=%x\n", + ibwq->wq_num); + wq->rss_usecnt--; + + mutex_unlock(&wq->mutex); + } +} + +static void fill_qp_rss_context(struct mlx4_qp_context *context, + struct mlx4_ib_qp *qp) +{ + struct mlx4_rss_context *rss_context; + + rss_context = (void *)context + offsetof(struct mlx4_qp_context, + pri_path) + MLX4_RSS_OFFSET_IN_QPC_PRI_PATH; + + rss_context->base_qpn = cpu_to_be32(qp->rss_ctx->base_qpn_tbl_sz); + rss_context->default_qpn = + cpu_to_be32(qp->rss_ctx->base_qpn_tbl_sz & 0xffffff); + if (qp->rss_ctx->flags & (MLX4_RSS_UDP_IPV4 | MLX4_RSS_UDP_IPV6)) + rss_context->base_qpn_udp = rss_context->default_qpn; + rss_context->flags = qp->rss_ctx->flags; + /* Currently support just toeplitz */ + rss_context->hash_fn = MLX4_RSS_HASH_TOP; + + memcpy(rss_context->rss_key, qp->rss_ctx->rss_key, + MLX4_EN_RSS_KEY_SIZE); +} + static int __mlx4_ib_modify_qp(void *src, enum mlx4_ib_source_type src_type, const struct ib_qp_attr *attr, int attr_mask, enum ib_qp_state cur_state, enum ib_qp_state new_state) { struct ib_uobject *ibuobject; struct ib_srq *ibsrq; + struct ib_rwq_ind_table *rwq_ind_tbl; enum ib_qp_type qp_type; struct mlx4_ib_dev *dev; struct mlx4_ib_qp *qp; @@ -1804,23 +2153,25 @@ static int __mlx4_ib_modify_qp(void *src, enum mlx4_ib_source_type src_type, if (src_type == MLX4_IB_RWQ_SRC) { struct ib_wq *ibwq; - ibwq = (struct ib_wq *)src; - ibuobject = ibwq->uobject; - ibsrq = NULL; - qp_type = IB_QPT_RAW_PACKET; - qp = to_mqp((struct ib_qp *)ibwq); - dev = to_mdev(ibwq->device); - pd = to_mpd(ibwq->pd); + ibwq = (struct ib_wq *)src; + ibuobject = ibwq->uobject; + ibsrq = NULL; + rwq_ind_tbl = NULL; + qp_type = IB_QPT_RAW_PACKET; + qp = to_mqp((struct ib_qp *)ibwq); + dev = to_mdev(ibwq->device); + pd = to_mpd(ibwq->pd); } else { struct ib_qp *ibqp; - ibqp = (struct ib_qp *)src; - ibuobject = ibqp->uobject; - ibsrq = ibqp->srq; - qp_type = ibqp->qp_type; - qp = to_mqp(ibqp); - dev = to_mdev(ibqp->device); - pd = get_pd(qp); + ibqp = (struct ib_qp *)src; + ibuobject = ibqp->uobject; + ibsrq = ibqp->srq; + rwq_ind_tbl = ibqp->rwq_ind_tbl; + qp_type = ibqp->qp_type; + qp = to_mqp(ibqp); + dev = to_mdev(ibqp->device); + pd = get_pd(qp); } /* APM is not supported under RoCE */ @@ -1836,6 +2187,11 @@ static int __mlx4_ib_modify_qp(void *src, enum mlx4_ib_source_type src_type, context->flags = cpu_to_be32((to_mlx4_state(new_state) << 28) | (to_mlx4_st(dev, qp->mlx4_ib_qp_type) << 16)); + if (rwq_ind_tbl) { + fill_qp_rss_context(context, qp); + context->flags |= cpu_to_be32(1 << MLX4_RSS_QPC_FLAG_OFFSET); + } + if (!(attr_mask & IB_QP_PATH_MIG_STATE)) context->flags |= cpu_to_be32(MLX4_QP_PM_MIGRATED << 11); else { @@ -1876,9 +2232,11 @@ static int __mlx4_ib_modify_qp(void *src, enum mlx4_ib_source_type src_type, ilog2(dev->dev->caps.max_msg_sz); } - if (qp->rq.wqe_cnt) - context->rq_size_stride = ilog2(qp->rq.wqe_cnt) << 3; - context->rq_size_stride |= qp->rq.wqe_shift - 4; + if (!rwq_ind_tbl) { /* PRM RSS receive side should be left zeros */ + if (qp->rq.wqe_cnt) + context->rq_size_stride = ilog2(qp->rq.wqe_cnt) << 3; + context->rq_size_stride |= qp->rq.wqe_shift - 4; + } if (qp->sq.wqe_cnt) context->sq_size_stride = ilog2(qp->sq.wqe_cnt) << 3; @@ -2031,8 +2389,14 @@ static int __mlx4_ib_modify_qp(void *src, enum mlx4_ib_source_type src_type, optpar |= MLX4_QP_OPTPAR_ALT_ADDR_PATH; } - get_cqs(qp, src_type, &send_cq, &recv_cq); - context->pd = cpu_to_be32(pd->pdn); + context->pd = cpu_to_be32(pd->pdn); + + if (!rwq_ind_tbl) { + get_cqs(qp, src_type, &send_cq, &recv_cq); + } else { /* Set dummy CQs to be compatible with HV and PRM */ + send_cq = to_mcq(rwq_ind_tbl->ind_tbl[0]->cq); + recv_cq = send_cq; + } context->cqn_send = cpu_to_be32(send_cq->mcq.cqn); context->cqn_recv = cpu_to_be32(recv_cq->mcq.cqn); context->params1 = cpu_to_be32(MLX4_IB_ACK_REQ_FREQ << 28); @@ -2358,6 +2722,11 @@ static int __mlx4_ib_modify_qp(void *src, enum mlx4_ib_source_type src_type, return err; } +enum { + MLX4_IB_MODIFY_QP_RSS_SUP_ATTR_MSK = (IB_QP_STATE | + IB_QP_PORT), +}; + static int _mlx4_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, int attr_mask, struct ib_udata *udata) { @@ -2388,6 +2757,27 @@ static int _mlx4_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, goto out; } + if (ibqp->rwq_ind_tbl) { + if (!(((cur_state == IB_QPS_RESET) && + (new_state == IB_QPS_INIT)) || + ((cur_state == IB_QPS_INIT) && + (new_state == IB_QPS_RTR)))) { + pr_debug("qpn 0x%x: RSS QP unsupported transition %d to %d\n", + ibqp->qp_num, cur_state, new_state); + + err = -EOPNOTSUPP; + goto out; + } + + if (attr_mask & ~MLX4_IB_MODIFY_QP_RSS_SUP_ATTR_MSK) { + pr_debug("qpn 0x%x: RSS QP unsupported attribute mask 0x%x for transition %d to %d\n", + ibqp->qp_num, attr_mask, cur_state, new_state); + + err = -EOPNOTSUPP; + goto out; + } + } + if (mlx4_is_bonded(dev->dev) && (attr_mask & IB_QP_PORT)) { if ((cur_state == IB_QPS_RESET) && (new_state == IB_QPS_INIT)) { if ((ibqp->qp_type == IB_QPT_RC) || @@ -2452,9 +2842,18 @@ static int _mlx4_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, goto out; } + if (ibqp->rwq_ind_tbl && (new_state == IB_QPS_INIT)) { + err = bringup_rss_rwqs(ibqp->rwq_ind_tbl, attr->port_num); + if (err) + goto out; + } + err = __mlx4_ib_modify_qp(ibqp, MLX4_IB_QP_SRC, attr, attr_mask, cur_state, new_state); + if (ibqp->rwq_ind_tbl && err) + bring_down_rss_rwqs(ibqp->rwq_ind_tbl); + if (mlx4_is_bonded(dev->dev) && (attr_mask & IB_QP_PORT)) attr->port_num = 1; @@ -3643,6 +4042,9 @@ int mlx4_ib_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *qp_attr, int qp_attr int mlx4_state; int err = 0; + if (ibqp->rwq_ind_tbl) + return -EOPNOTSUPP; + mutex_lock(&qp->mutex); if (qp->state == IB_QPS_RESET) { @@ -3917,6 +4319,11 @@ int mlx4_ib_modify_wq(struct ib_wq *ibwq, struct ib_wq_attr *wq_attr, if ((new_state == IB_WQS_ERR) && (cur_state == IB_WQS_RESET)) return -EINVAL; + /* Need to protect against the parent RSS which also may modify WQ + * state. + */ + mutex_lock(&qp->mutex); + /* Can update HW state only if a RSS QP has already associated to this * WQ, so we can apply its port on the WQ. */ @@ -3926,6 +4333,8 @@ int mlx4_ib_modify_wq(struct ib_wq *ibwq, struct ib_wq_attr *wq_attr, if (!err) ibwq->state = new_state; + mutex_unlock(&qp->mutex); + return err; } diff --git a/include/uapi/rdma/mlx4-abi.h b/include/uapi/rdma/mlx4-abi.h index 5591d955ba00..d915cab37ec3 100644 --- a/include/uapi/rdma/mlx4-abi.h +++ b/include/uapi/rdma/mlx4-abi.h @@ -95,6 +95,16 @@ struct mlx4_ib_create_srq_resp { __u32 reserved; }; +struct mlx4_ib_create_qp_rss { + __u64 rx_hash_fields_mask; + __u8 rx_hash_function; + __u8 rx_key_len; + __u8 reserved[6]; + __u8 rx_hash_key[40]; + __u32 comp_mask; + __u32 reserved1; +}; + struct mlx4_ib_create_qp { __u64 buf_addr; __u64 db_addr; @@ -123,4 +133,27 @@ struct mlx4_ib_create_rwq_ind_tbl_resp { __u32 response_length; __u32 reserved; }; + +/* RX Hash function flags */ +enum mlx4_ib_rx_hash_function_flags { + MLX4_IB_RX_HASH_FUNC_TOEPLITZ = 1 << 0, +}; + +/* + * RX Hash flags, these flags allows to set which incoming packet's field should + * participates in RX Hash. Each flag represent certain packet's field, + * when the flag is set the field that is represented by the flag will + * participate in RX Hash calculation. + */ +enum mlx4_ib_rx_hash_fields { + MLX4_IB_RX_HASH_SRC_IPV4 = 1 << 0, + MLX4_IB_RX_HASH_DST_IPV4 = 1 << 1, + MLX4_IB_RX_HASH_SRC_IPV6 = 1 << 2, + MLX4_IB_RX_HASH_DST_IPV6 = 1 << 3, + MLX4_IB_RX_HASH_SRC_PORT_TCP = 1 << 4, + MLX4_IB_RX_HASH_DST_PORT_TCP = 1 << 5, + MLX4_IB_RX_HASH_SRC_PORT_UDP = 1 << 6, + MLX4_IB_RX_HASH_DST_PORT_UDP = 1 << 7 +}; + #endif /* MLX4_ABI_USER_H */ From 6afff1c7152f6f0787d7ad5de2e124f8643f0a45 Mon Sep 17 00:00:00 2001 From: Guy Levi Date: Tue, 4 Jul 2017 16:24:27 +0300 Subject: [PATCH 34/34] IB/mlx4: Expose RSS capabilities As a final step to support RSS feature, expose the RSS capabilities in query device verb. It includes: - Max rwq indirection tables. - Max rwq indirection table size. - Supported qp types. Signed-off-by: Guy Levi Reviewed-by: Yishai Hadas Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx4/main.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c index b42234571a8c..1dee0a51ef67 100644 --- a/drivers/infiniband/hw/mlx4/main.c +++ b/drivers/infiniband/hw/mlx4/main.c @@ -556,8 +556,13 @@ static int mlx4_ib_query_device(struct ib_device *ibdev, if ((dev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_RSS) && (mlx4_ib_port_link_layer(ibdev, 1) == IB_LINK_LAYER_ETHERNET || - mlx4_ib_port_link_layer(ibdev, 2) == IB_LINK_LAYER_ETHERNET)) + mlx4_ib_port_link_layer(ibdev, 2) == IB_LINK_LAYER_ETHERNET)) { + props->rss_caps.max_rwq_indirection_tables = props->max_qp; + props->rss_caps.max_rwq_indirection_table_size = + dev->dev->caps.max_rss_tbl_sz; + props->rss_caps.supported_qpts = 1 << IB_QPT_RAW_PACKET; props->max_wq_type_rq = props->max_qp; + } if (!mlx4_is_slave(dev->dev)) err = mlx4_get_internal_clock_params(dev->dev, &clock_params);