// SPDX-License-Identifier: (GPL-2.0-only OR BSD-3-Clause) /* QLogic qed NIC Driver * Copyright (c) 2015-2017 QLogic Corporation * Copyright (c) 2019-2020 Marvell International Ltd. */ #include #include #include #include #include #include #include "qed_cxt.h" #include "qed_hw.h" #include "qed_ll2.h" #include "qed_rdma.h" #include "qed_reg_addr.h" #include "qed_sp.h" #include "qed_ooo.h" #define QED_IWARP_ORD_DEFAULT 32 #define QED_IWARP_IRD_DEFAULT 32 #define QED_IWARP_MAX_FW_MSS 4120 #define QED_EP_SIG 0xecabcdef struct mpa_v2_hdr { __be16 ird; __be16 ord; }; #define MPA_V2_PEER2PEER_MODEL 0x8000 #define MPA_V2_SEND_RTR 0x4000 /* on ird */ #define MPA_V2_READ_RTR 0x4000 /* on ord */ #define MPA_V2_WRITE_RTR 0x8000 #define MPA_V2_IRD_ORD_MASK 0x3FFF #define MPA_REV2(_mpa_rev) ((_mpa_rev) == MPA_NEGOTIATION_TYPE_ENHANCED) #define QED_IWARP_INVALID_TCP_CID 0xffffffff #define QED_IWARP_RCV_WND_SIZE_DEF_BB_2P (200 * 1024) #define QED_IWARP_RCV_WND_SIZE_DEF_BB_4P (100 * 1024) #define QED_IWARP_RCV_WND_SIZE_DEF_AH_2P (150 * 1024) #define QED_IWARP_RCV_WND_SIZE_DEF_AH_4P (90 * 1024) #define QED_IWARP_RCV_WND_SIZE_MIN (0xffff) #define TIMESTAMP_HEADER_SIZE (12) #define QED_IWARP_MAX_FIN_RT_DEFAULT (2) #define QED_IWARP_TS_EN BIT(0) #define QED_IWARP_DA_EN BIT(1) #define QED_IWARP_PARAM_CRC_NEEDED (1) #define QED_IWARP_PARAM_P2P (1) #define QED_IWARP_DEF_MAX_RT_TIME (0) #define QED_IWARP_DEF_CWND_FACTOR (4) #define QED_IWARP_DEF_KA_MAX_PROBE_CNT (5) #define QED_IWARP_DEF_KA_TIMEOUT (1200000) /* 20 min */ #define QED_IWARP_DEF_KA_INTERVAL (1000) /* 1 sec */ static int qed_iwarp_async_event(struct qed_hwfn *p_hwfn, u8 fw_event_code, u16 echo, union event_ring_data *data, u8 fw_return_code); /* Override devinfo with iWARP specific values */ void qed_iwarp_init_devinfo(struct qed_hwfn *p_hwfn) { struct qed_rdma_device *dev = p_hwfn->p_rdma_info->dev; dev->max_inline = IWARP_REQ_MAX_INLINE_DATA_SIZE; dev->max_qp = min_t(u32, IWARP_MAX_QPS, p_hwfn->p_rdma_info->num_qps) - QED_IWARP_PREALLOC_CNT; dev->max_cq = dev->max_qp; dev->max_qp_resp_rd_atomic_resc = QED_IWARP_IRD_DEFAULT; dev->max_qp_req_rd_atomic_resc = QED_IWARP_ORD_DEFAULT; } void qed_iwarp_init_hw(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt) { p_hwfn->rdma_prs_search_reg = PRS_REG_SEARCH_TCP; qed_wr(p_hwfn, p_ptt, p_hwfn->rdma_prs_search_reg, 1); p_hwfn->b_rdma_enabled_in_prs = true; } /* We have two cid maps, one for tcp which should be used only from passive * syn processing and replacing a pre-allocated ep in the list. The second * for active tcp and for QPs. */ static void qed_iwarp_cid_cleaned(struct qed_hwfn *p_hwfn, u32 cid) { cid -= qed_cxt_get_proto_cid_start(p_hwfn, p_hwfn->p_rdma_info->proto); spin_lock_bh(&p_hwfn->p_rdma_info->lock); if (cid < QED_IWARP_PREALLOC_CNT) qed_bmap_release_id(p_hwfn, &p_hwfn->p_rdma_info->tcp_cid_map, cid); else qed_bmap_release_id(p_hwfn, &p_hwfn->p_rdma_info->cid_map, cid); spin_unlock_bh(&p_hwfn->p_rdma_info->lock); } void qed_iwarp_init_fw_ramrod(struct qed_hwfn *p_hwfn, struct iwarp_init_func_ramrod_data *p_ramrod) { p_ramrod->iwarp.ll2_ooo_q_index = RESC_START(p_hwfn, QED_LL2_RAM_QUEUE) + p_hwfn->p_rdma_info->iwarp.ll2_ooo_handle; p_ramrod->tcp.max_fin_rt = QED_IWARP_MAX_FIN_RT_DEFAULT; return; } static int qed_iwarp_alloc_cid(struct qed_hwfn *p_hwfn, u32 *cid) { int rc; spin_lock_bh(&p_hwfn->p_rdma_info->lock); rc = qed_rdma_bmap_alloc_id(p_hwfn, &p_hwfn->p_rdma_info->cid_map, cid); spin_unlock_bh(&p_hwfn->p_rdma_info->lock); if (rc) { DP_NOTICE(p_hwfn, "Failed in allocating iwarp cid\n"); return rc; } *cid += qed_cxt_get_proto_cid_start(p_hwfn, p_hwfn->p_rdma_info->proto); rc = qed_cxt_dynamic_ilt_alloc(p_hwfn, QED_ELEM_CXT, *cid); if (rc) qed_iwarp_cid_cleaned(p_hwfn, *cid); return rc; } static void qed_iwarp_set_tcp_cid(struct qed_hwfn *p_hwfn, u32 cid) { cid -= qed_cxt_get_proto_cid_start(p_hwfn, p_hwfn->p_rdma_info->proto); spin_lock_bh(&p_hwfn->p_rdma_info->lock); qed_bmap_set_id(p_hwfn, &p_hwfn->p_rdma_info->tcp_cid_map, cid); spin_unlock_bh(&p_hwfn->p_rdma_info->lock); } /* This function allocates a cid for passive tcp (called from syn receive) * the reason it's separate from the regular cid allocation is because it * is assured that these cids already have ilt allocated. They are preallocated * to ensure that we won't need to allocate memory during syn processing */ static int qed_iwarp_alloc_tcp_cid(struct qed_hwfn *p_hwfn, u32 *cid) { int rc; spin_lock_bh(&p_hwfn->p_rdma_info->lock); rc = qed_rdma_bmap_alloc_id(p_hwfn, &p_hwfn->p_rdma_info->tcp_cid_map, cid); spin_unlock_bh(&p_hwfn->p_rdma_info->lock); if (rc) { DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "can't allocate iwarp tcp cid max-count=%d\n", p_hwfn->p_rdma_info->tcp_cid_map.max_count); *cid = QED_IWARP_INVALID_TCP_CID; return rc; } *cid += qed_cxt_get_proto_cid_start(p_hwfn, p_hwfn->p_rdma_info->proto); return 0; } int qed_iwarp_create_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp, struct qed_rdma_create_qp_out_params *out_params) { struct iwarp_create_qp_ramrod_data *p_ramrod; struct qed_sp_init_data init_data; struct qed_spq_entry *p_ent; u16 physical_queue; u32 cid; int rc; qp->shared_queue = dma_alloc_coherent(&p_hwfn->cdev->pdev->dev, IWARP_SHARED_QUEUE_PAGE_SIZE, &qp->shared_queue_phys_addr, GFP_KERNEL); if (!qp->shared_queue) return -ENOMEM; out_params->sq_pbl_virt = (u8 *)qp->shared_queue + IWARP_SHARED_QUEUE_PAGE_SQ_PBL_OFFSET; out_params->sq_pbl_phys = qp->shared_queue_phys_addr + IWARP_SHARED_QUEUE_PAGE_SQ_PBL_OFFSET; out_params->rq_pbl_virt = (u8 *)qp->shared_queue + IWARP_SHARED_QUEUE_PAGE_RQ_PBL_OFFSET; out_params->rq_pbl_phys = qp->shared_queue_phys_addr + IWARP_SHARED_QUEUE_PAGE_RQ_PBL_OFFSET; rc = qed_iwarp_alloc_cid(p_hwfn, &cid); if (rc) goto err1; qp->icid = (u16)cid; memset(&init_data, 0, sizeof(init_data)); init_data.opaque_fid = p_hwfn->hw_info.opaque_fid; init_data.cid = qp->icid; init_data.comp_mode = QED_SPQ_MODE_EBLOCK; rc = qed_sp_init_request(p_hwfn, &p_ent, IWARP_RAMROD_CMD_ID_CREATE_QP, PROTOCOLID_IWARP, &init_data); if (rc) goto err2; p_ramrod = &p_ent->ramrod.iwarp_create_qp; SET_FIELD(p_ramrod->flags, IWARP_CREATE_QP_RAMROD_DATA_FMR_AND_RESERVED_EN, qp->fmr_and_reserved_lkey); SET_FIELD(p_ramrod->flags, IWARP_CREATE_QP_RAMROD_DATA_SIGNALED_COMP, qp->signal_all); SET_FIELD(p_ramrod->flags, IWARP_CREATE_QP_RAMROD_DATA_RDMA_RD_EN, qp->incoming_rdma_read_en); SET_FIELD(p_ramrod->flags, IWARP_CREATE_QP_RAMROD_DATA_RDMA_WR_EN, qp->incoming_rdma_write_en); SET_FIELD(p_ramrod->flags, IWARP_CREATE_QP_RAMROD_DATA_ATOMIC_EN, qp->incoming_atomic_en); SET_FIELD(p_ramrod->flags, IWARP_CREATE_QP_RAMROD_DATA_SRQ_FLG, qp->use_srq); p_ramrod->pd = qp->pd; p_ramrod->sq_num_pages = qp->sq_num_pages; p_ramrod->rq_num_pages = qp->rq_num_pages; p_ramrod->srq_id.srq_idx = cpu_to_le16(qp->srq_id); p_ramrod->srq_id.opaque_fid = cpu_to_le16(p_hwfn->hw_info.opaque_fid); p_ramrod->qp_handle_for_cqe.hi = cpu_to_le32(qp->qp_handle.hi); p_ramrod->qp_handle_for_cqe.lo = cpu_to_le32(qp->qp_handle.lo); p_ramrod->cq_cid_for_sq = cpu_to_le32((p_hwfn->hw_info.opaque_fid << 16) | qp->sq_cq_id); p_ramrod->cq_cid_for_rq = cpu_to_le32((p_hwfn->hw_info.opaque_fid << 16) | qp->rq_cq_id); p_ramrod->dpi = cpu_to_le16(qp->dpi); physical_queue = qed_get_cm_pq_idx(p_hwfn, PQ_FLAGS_OFLD); p_ramrod->physical_q0 = cpu_to_le16(physical_queue); physical_queue = qed_get_cm_pq_idx(p_hwfn, PQ_FLAGS_ACK); p_ramrod->physical_q1 = cpu_to_le16(physical_queue); rc = qed_spq_post(p_hwfn, p_ent, NULL); if (rc) goto err2; return rc; err2: qed_iwarp_cid_cleaned(p_hwfn, cid); err1: dma_free_coherent(&p_hwfn->cdev->pdev->dev, IWARP_SHARED_QUEUE_PAGE_SIZE, qp->shared_queue, qp->shared_queue_phys_addr); return rc; } static int qed_iwarp_modify_fw(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp) { struct iwarp_modify_qp_ramrod_data *p_ramrod; struct qed_sp_init_data init_data; struct qed_spq_entry *p_ent; int rc; /* Get SPQ entry */ memset(&init_data, 0, sizeof(init_data)); init_data.cid = qp->icid; init_data.opaque_fid = p_hwfn->hw_info.opaque_fid; init_data.comp_mode = QED_SPQ_MODE_EBLOCK; rc = qed_sp_init_request(p_hwfn, &p_ent, IWARP_RAMROD_CMD_ID_MODIFY_QP, p_hwfn->p_rdma_info->proto, &init_data); if (rc) return rc; p_ramrod = &p_ent->ramrod.iwarp_modify_qp; SET_FIELD(p_ramrod->flags, IWARP_MODIFY_QP_RAMROD_DATA_STATE_TRANS_EN, 0x1); if (qp->iwarp_state == QED_IWARP_QP_STATE_CLOSING) p_ramrod->transition_to_state = IWARP_MODIFY_QP_STATE_CLOSING; else p_ramrod->transition_to_state = IWARP_MODIFY_QP_STATE_ERROR; rc = qed_spq_post(p_hwfn, p_ent, NULL); DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "QP(0x%x)rc=%d\n", qp->icid, rc); return rc; } enum qed_iwarp_qp_state qed_roce2iwarp_state(enum qed_roce_qp_state state) { switch (state) { case QED_ROCE_QP_STATE_RESET: case QED_ROCE_QP_STATE_INIT: case QED_ROCE_QP_STATE_RTR: return QED_IWARP_QP_STATE_IDLE; case QED_ROCE_QP_STATE_RTS: return QED_IWARP_QP_STATE_RTS; case QED_ROCE_QP_STATE_SQD: return QED_IWARP_QP_STATE_CLOSING; case QED_ROCE_QP_STATE_ERR: return QED_IWARP_QP_STATE_ERROR; case QED_ROCE_QP_STATE_SQE: return QED_IWARP_QP_STATE_TERMINATE; default: return QED_IWARP_QP_STATE_ERROR; } } static enum qed_roce_qp_state qed_iwarp2roce_state(enum qed_iwarp_qp_state state) { switch (state) { case QED_IWARP_QP_STATE_IDLE: return QED_ROCE_QP_STATE_INIT; case QED_IWARP_QP_STATE_RTS: return QED_ROCE_QP_STATE_RTS; case QED_IWARP_QP_STATE_TERMINATE: return QED_ROCE_QP_STATE_SQE; case QED_IWARP_QP_STATE_CLOSING: return QED_ROCE_QP_STATE_SQD; case QED_IWARP_QP_STATE_ERROR: return QED_ROCE_QP_STATE_ERR; default: return QED_ROCE_QP_STATE_ERR; } } static const char * const iwarp_state_names[] = { "IDLE", "RTS", "TERMINATE", "CLOSING", "ERROR", }; int qed_iwarp_modify_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp, enum qed_iwarp_qp_state new_state, bool internal) { enum qed_iwarp_qp_state prev_iw_state; bool modify_fw = false; int rc = 0; /* modify QP can be called from upper-layer or as a result of async * RST/FIN... therefore need to protect */ spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.qp_lock); prev_iw_state = qp->iwarp_state; if (prev_iw_state == new_state) { spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.qp_lock); return 0; } switch (prev_iw_state) { case QED_IWARP_QP_STATE_IDLE: switch (new_state) { case QED_IWARP_QP_STATE_RTS: qp->iwarp_state = QED_IWARP_QP_STATE_RTS; break; case QED_IWARP_QP_STATE_ERROR: qp->iwarp_state = QED_IWARP_QP_STATE_ERROR; if (!internal) modify_fw = true; break; default: break; } break; case QED_IWARP_QP_STATE_RTS: switch (new_state) { case QED_IWARP_QP_STATE_CLOSING: if (!internal) modify_fw = true; qp->iwarp_state = QED_IWARP_QP_STATE_CLOSING; break; case QED_IWARP_QP_STATE_ERROR: if (!internal) modify_fw = true; qp->iwarp_state = QED_IWARP_QP_STATE_ERROR; break; default: break; } break; case QED_IWARP_QP_STATE_ERROR: switch (new_state) { case QED_IWARP_QP_STATE_IDLE: qp->iwarp_state = new_state; break; case QED_IWARP_QP_STATE_CLOSING: /* could happen due to race... do nothing.... */ break; default: rc = -EINVAL; } break; case QED_IWARP_QP_STATE_TERMINATE: case QED_IWARP_QP_STATE_CLOSING: qp->iwarp_state = new_state; break; default: break; } DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "QP(0x%x) %s --> %s%s\n", qp->icid, iwarp_state_names[prev_iw_state], iwarp_state_names[qp->iwarp_state], internal ? "internal" : ""); spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.qp_lock); if (modify_fw) rc = qed_iwarp_modify_fw(p_hwfn, qp); return rc; } int qed_iwarp_fw_destroy(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp) { struct qed_sp_init_data init_data; struct qed_spq_entry *p_ent; int rc; /* Get SPQ entry */ memset(&init_data, 0, sizeof(init_data)); init_data.cid = qp->icid; init_data.opaque_fid = p_hwfn->hw_info.opaque_fid; init_data.comp_mode = QED_SPQ_MODE_EBLOCK; rc = qed_sp_init_request(p_hwfn, &p_ent, IWARP_RAMROD_CMD_ID_DESTROY_QP, p_hwfn->p_rdma_info->proto, &init_data); if (rc) return rc; rc = qed_spq_post(p_hwfn, p_ent, NULL); DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "QP(0x%x) rc = %d\n", qp->icid, rc); return rc; } static void qed_iwarp_destroy_ep(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep, bool remove_from_active_list) { dma_free_coherent(&p_hwfn->cdev->pdev->dev, sizeof(*ep->ep_buffer_virt), ep->ep_buffer_virt, ep->ep_buffer_phys); if (remove_from_active_list) { spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); list_del(&ep->list_entry); spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); } if (ep->qp) ep->qp->ep = NULL; kfree(ep); } int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp) { struct qed_iwarp_ep *ep = qp->ep; int wait_count = 0; int rc = 0; if (qp->iwarp_state != QED_IWARP_QP_STATE_ERROR) { rc = qed_iwarp_modify_qp(p_hwfn, qp, QED_IWARP_QP_STATE_ERROR, false); if (rc) return rc; } /* Make sure ep is closed before returning and freeing memory. */ if (ep) { while (READ_ONCE(ep->state) != QED_IWARP_EP_CLOSED && wait_count++ < 200) msleep(100); if (ep->state != QED_IWARP_EP_CLOSED) DP_NOTICE(p_hwfn, "ep state close timeout state=%x\n", ep->state); qed_iwarp_destroy_ep(p_hwfn, ep, false); } rc = qed_iwarp_fw_destroy(p_hwfn, qp); if (qp->shared_queue) dma_free_coherent(&p_hwfn->cdev->pdev->dev, IWARP_SHARED_QUEUE_PAGE_SIZE, qp->shared_queue, qp->shared_queue_phys_addr); return rc; } static int qed_iwarp_create_ep(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep **ep_out) { struct qed_iwarp_ep *ep; int rc; ep = kzalloc(sizeof(*ep), GFP_KERNEL); if (!ep) return -ENOMEM; ep->state = QED_IWARP_EP_INIT; ep->ep_buffer_virt = dma_alloc_coherent(&p_hwfn->cdev->pdev->dev, sizeof(*ep->ep_buffer_virt), &ep->ep_buffer_phys, GFP_KERNEL); if (!ep->ep_buffer_virt) { rc = -ENOMEM; goto err; } ep->sig = QED_EP_SIG; *ep_out = ep; return 0; err: kfree(ep); return rc; } static void qed_iwarp_print_tcp_ramrod(struct qed_hwfn *p_hwfn, struct iwarp_tcp_offload_ramrod_data *p_tcp_ramrod) { DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "local_mac=%x %x %x, remote_mac=%x %x %x\n", p_tcp_ramrod->tcp.local_mac_addr_lo, p_tcp_ramrod->tcp.local_mac_addr_mid, p_tcp_ramrod->tcp.local_mac_addr_hi, p_tcp_ramrod->tcp.remote_mac_addr_lo, p_tcp_ramrod->tcp.remote_mac_addr_mid, p_tcp_ramrod->tcp.remote_mac_addr_hi); if (p_tcp_ramrod->tcp.ip_version == TCP_IPV4) { DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "local_ip=%pI4h:%x, remote_ip=%pI4h:%x, vlan=%x\n", p_tcp_ramrod->tcp.local_ip, p_tcp_ramrod->tcp.local_port, p_tcp_ramrod->tcp.remote_ip, p_tcp_ramrod->tcp.remote_port, p_tcp_ramrod->tcp.vlan_id); } else { DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "local_ip=%pI6:%x, remote_ip=%pI6:%x, vlan=%x\n", p_tcp_ramrod->tcp.local_ip, p_tcp_ramrod->tcp.local_port, p_tcp_ramrod->tcp.remote_ip, p_tcp_ramrod->tcp.remote_port, p_tcp_ramrod->tcp.vlan_id); } DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "flow_label=%x, ttl=%x, tos_or_tc=%x, mss=%x, rcv_wnd_scale=%x, connect_mode=%x, flags=%x\n", p_tcp_ramrod->tcp.flow_label, p_tcp_ramrod->tcp.ttl, p_tcp_ramrod->tcp.tos_or_tc, p_tcp_ramrod->tcp.mss, p_tcp_ramrod->tcp.rcv_wnd_scale, p_tcp_ramrod->tcp.connect_mode, p_tcp_ramrod->tcp.flags); DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "syn_ip_payload_length=%x, lo=%x, hi=%x\n", p_tcp_ramrod->tcp.syn_ip_payload_length, p_tcp_ramrod->tcp.syn_phy_addr_lo, p_tcp_ramrod->tcp.syn_phy_addr_hi); } static int qed_iwarp_tcp_offload(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep) { struct qed_iwarp_info *iwarp_info = &p_hwfn->p_rdma_info->iwarp; struct iwarp_tcp_offload_ramrod_data *p_tcp_ramrod; struct tcp_offload_params_opt2 *tcp; struct qed_sp_init_data init_data; struct qed_spq_entry *p_ent; dma_addr_t async_output_phys; dma_addr_t in_pdata_phys; u16 physical_q; u8 tcp_flags; int rc; int i; memset(&init_data, 0, sizeof(init_data)); init_data.cid = ep->tcp_cid; init_data.opaque_fid = p_hwfn->hw_info.opaque_fid; if (ep->connect_mode == TCP_CONNECT_PASSIVE) init_data.comp_mode = QED_SPQ_MODE_CB; else init_data.comp_mode = QED_SPQ_MODE_EBLOCK; rc = qed_sp_init_request(p_hwfn, &p_ent, IWARP_RAMROD_CMD_ID_TCP_OFFLOAD, PROTOCOLID_IWARP, &init_data); if (rc) return rc; p_tcp_ramrod = &p_ent->ramrod.iwarp_tcp_offload; in_pdata_phys = ep->ep_buffer_phys + offsetof(struct qed_iwarp_ep_memory, in_pdata); DMA_REGPAIR_LE(p_tcp_ramrod->iwarp.incoming_ulp_buffer.addr, in_pdata_phys); p_tcp_ramrod->iwarp.incoming_ulp_buffer.len = cpu_to_le16(sizeof(ep->ep_buffer_virt->in_pdata)); async_output_phys = ep->ep_buffer_phys + offsetof(struct qed_iwarp_ep_memory, async_output); DMA_REGPAIR_LE(p_tcp_ramrod->iwarp.async_eqe_output_buf, async_output_phys); p_tcp_ramrod->iwarp.handle_for_async.hi = cpu_to_le32(PTR_HI(ep)); p_tcp_ramrod->iwarp.handle_for_async.lo = cpu_to_le32(PTR_LO(ep)); physical_q = qed_get_cm_pq_idx(p_hwfn, PQ_FLAGS_OFLD); p_tcp_ramrod->iwarp.physical_q0 = cpu_to_le16(physical_q); physical_q = qed_get_cm_pq_idx(p_hwfn, PQ_FLAGS_ACK); p_tcp_ramrod->iwarp.physical_q1 = cpu_to_le16(physical_q); p_tcp_ramrod->iwarp.mpa_mode = iwarp_info->mpa_rev; tcp = &p_tcp_ramrod->tcp; qed_set_fw_mac_addr(&tcp->remote_mac_addr_hi, &tcp->remote_mac_addr_mid, &tcp->remote_mac_addr_lo, ep->remote_mac_addr); qed_set_fw_mac_addr(&tcp->local_mac_addr_hi, &tcp->local_mac_addr_mid, &tcp->local_mac_addr_lo, ep->local_mac_addr); tcp->vlan_id = cpu_to_le16(ep->cm_info.vlan); tcp_flags = p_hwfn->p_rdma_info->iwarp.tcp_flags; tcp->flags = 0; SET_FIELD(tcp->flags, TCP_OFFLOAD_PARAMS_OPT2_TS_EN, !!(tcp_flags & QED_IWARP_TS_EN)); SET_FIELD(tcp->flags, TCP_OFFLOAD_PARAMS_OPT2_DA_EN, !!(tcp_flags & QED_IWARP_DA_EN)); tcp->ip_version = ep->cm_info.ip_version; for (i = 0; i < 4; i++) { tcp->remote_ip[i] = cpu_to_le32(ep->cm_info.remote_ip[i]); tcp->local_ip[i] = cpu_to_le32(ep->cm_info.local_ip[i]); } tcp->remote_port = cpu_to_le16(ep->cm_info.remote_port); tcp->local_port = cpu_to_le16(ep->cm_info.local_port); tcp->mss = cpu_to_le16(ep->mss); tcp->flow_label = 0; tcp->ttl = 0x40; tcp->tos_or_tc = 0; tcp->max_rt_time = QED_IWARP_DEF_MAX_RT_TIME; tcp->cwnd = QED_IWARP_DEF_CWND_FACTOR * tcp->mss; tcp->ka_max_probe_cnt = QED_IWARP_DEF_KA_MAX_PROBE_CNT; tcp->ka_timeout = QED_IWARP_DEF_KA_TIMEOUT; tcp->ka_interval = QED_IWARP_DEF_KA_INTERVAL; tcp->rcv_wnd_scale = (u8)p_hwfn->p_rdma_info->iwarp.rcv_wnd_scale; tcp->connect_mode = ep->connect_mode; if (ep->connect_mode == TCP_CONNECT_PASSIVE) { tcp->syn_ip_payload_length = cpu_to_le16(ep->syn_ip_payload_length); tcp->syn_phy_addr_hi = DMA_HI_LE(ep->syn_phy_addr); tcp->syn_phy_addr_lo = DMA_LO_LE(ep->syn_phy_addr); } qed_iwarp_print_tcp_ramrod(p_hwfn, p_tcp_ramrod); rc = qed_spq_post(p_hwfn, p_ent, NULL); DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "EP(0x%x) Offload completed rc=%d\n", ep->tcp_cid, rc); return rc; } static void qed_iwarp_mpa_received(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep) { struct qed_iwarp_info *iwarp_info = &p_hwfn->p_rdma_info->iwarp; struct qed_iwarp_cm_event_params params; struct mpa_v2_hdr *mpa_v2; union async_output *async_data; u16 mpa_ord, mpa_ird; u8 mpa_hdr_size = 0; u8 mpa_rev; async_data = &ep->ep_buffer_virt->async_output; mpa_rev = async_data->mpa_request.mpa_handshake_mode; DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "private_data_len=%x handshake_mode=%x private_data=(%x)\n", async_data->mpa_request.ulp_data_len, mpa_rev, *((u32 *)(ep->ep_buffer_virt->in_pdata))); if (mpa_rev == MPA_NEGOTIATION_TYPE_ENHANCED) { /* Read ord/ird values from private data buffer */ mpa_v2 = (struct mpa_v2_hdr *)ep->ep_buffer_virt->in_pdata; mpa_hdr_size = sizeof(*mpa_v2); mpa_ord = ntohs(mpa_v2->ord); mpa_ird = ntohs(mpa_v2->ird); /* Temprary store in cm_info incoming ord/ird requested, later * replace with negotiated value during accept */ ep->cm_info.ord = (u8)min_t(u16, (mpa_ord & MPA_V2_IRD_ORD_MASK), QED_IWARP_ORD_DEFAULT); ep->cm_info.ird = (u8)min_t(u16, (mpa_ird & MPA_V2_IRD_ORD_MASK), QED_IWARP_IRD_DEFAULT); /* Peer2Peer negotiation */ ep->rtr_type = MPA_RTR_TYPE_NONE; if (mpa_ird & MPA_V2_PEER2PEER_MODEL) { if (mpa_ord & MPA_V2_WRITE_RTR) ep->rtr_type |= MPA_RTR_TYPE_ZERO_WRITE; if (mpa_ord & MPA_V2_READ_RTR) ep->rtr_type |= MPA_RTR_TYPE_ZERO_READ; if (mpa_ird & MPA_V2_SEND_RTR) ep->rtr_type |= MPA_RTR_TYPE_ZERO_SEND; ep->rtr_type &= iwarp_info->rtr_type; /* if we're left with no match send our capabilities */ if (ep->rtr_type == MPA_RTR_TYPE_NONE) ep->rtr_type = iwarp_info->rtr_type; } ep->mpa_rev = MPA_NEGOTIATION_TYPE_ENHANCED; } else { ep->cm_info.ord = QED_IWARP_ORD_DEFAULT; ep->cm_info.ird = QED_IWARP_IRD_DEFAULT; ep->mpa_rev = MPA_NEGOTIATION_TYPE_BASIC; } DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "MPA_NEGOTIATE (v%d): ORD: 0x%x IRD: 0x%x rtr:0x%x ulp_data_len = %x mpa_hdr_size = %x\n", mpa_rev, ep->cm_info.ord, ep->cm_info.ird, ep->rtr_type, async_data->mpa_request.ulp_data_len, mpa_hdr_size); /* Strip mpa v2 hdr from private data before sending to upper layer */ ep->cm_info.private_data = ep->ep_buffer_virt->in_pdata + mpa_hdr_size; ep->cm_info.private_data_len = async_data->mpa_request.ulp_data_len - mpa_hdr_size; params.event = QED_IWARP_EVENT_MPA_REQUEST; params.cm_info = &ep->cm_info; params.ep_context = ep; params.status = 0; ep->state = QED_IWARP_EP_MPA_REQ_RCVD; ep->event_cb(ep->cb_context, ¶ms); } static int qed_iwarp_mpa_offload(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep) { struct iwarp_mpa_offload_ramrod_data *p_mpa_ramrod; struct mpa_outgoing_params *common; struct qed_iwarp_info *iwarp_info; struct qed_sp_init_data init_data; dma_addr_t async_output_phys; struct qed_spq_entry *p_ent; dma_addr_t out_pdata_phys; dma_addr_t in_pdata_phys; struct qed_rdma_qp *qp; bool reject; int rc; if (!ep) return -EINVAL; qp = ep->qp; reject = !qp; memset(&init_data, 0, sizeof(init_data)); init_data.cid = reject ? ep->tcp_cid : qp->icid; init_data.opaque_fid = p_hwfn->hw_info.opaque_fid; if (ep->connect_mode == TCP_CONNECT_ACTIVE) init_data.comp_mode = QED_SPQ_MODE_CB; else init_data.comp_mode = QED_SPQ_MODE_EBLOCK; rc = qed_sp_init_request(p_hwfn, &p_ent, IWARP_RAMROD_CMD_ID_MPA_OFFLOAD, PROTOCOLID_IWARP, &init_data); if (rc) return rc; p_mpa_ramrod = &p_ent->ramrod.iwarp_mpa_offload; common = &p_mpa_ramrod->common; out_pdata_phys = ep->ep_buffer_phys + offsetof(struct qed_iwarp_ep_memory, out_pdata); DMA_REGPAIR_LE(common->outgoing_ulp_buffer.addr, out_pdata_phys); common->outgoing_ulp_buffer.len = ep->cm_info.private_data_len; common->crc_needed = p_hwfn->p_rdma_info->iwarp.crc_needed; common->out_rq.ord = ep->cm_info.ord; common->out_rq.ird = ep->cm_info.ird; p_mpa_ramrod->tcp_cid = p_hwfn->hw_info.opaque_fid << 16 | ep->tcp_cid; in_pdata_phys = ep->ep_buffer_phys + offsetof(struct qed_iwarp_ep_memory, in_pdata); p_mpa_ramrod->tcp_connect_side = ep->connect_mode; DMA_REGPAIR_LE(p_mpa_ramrod->incoming_ulp_buffer.addr, in_pdata_phys); p_mpa_ramrod->incoming_ulp_buffer.len = cpu_to_le16(sizeof(ep->ep_buffer_virt->in_pdata)); async_output_phys = ep->ep_buffer_phys + offsetof(struct qed_iwarp_ep_memory, async_output); DMA_REGPAIR_LE(p_mpa_ramrod->async_eqe_output_buf, async_output_phys); p_mpa_ramrod->handle_for_async.hi = cpu_to_le32(PTR_HI(ep)); p_mpa_ramrod->handle_for_async.lo = cpu_to_le32(PTR_LO(ep)); if (!reject) { DMA_REGPAIR_LE(p_mpa_ramrod->shared_queue_addr, qp->shared_queue_phys_addr); p_mpa_ramrod->stats_counter_id = RESC_START(p_hwfn, QED_RDMA_STATS_QUEUE) + qp->stats_queue; } else { common->reject = 1; } iwarp_info = &p_hwfn->p_rdma_info->iwarp; p_mpa_ramrod->rcv_wnd = iwarp_info->rcv_wnd_size; p_mpa_ramrod->mode = ep->mpa_rev; SET_FIELD(p_mpa_ramrod->rtr_pref, IWARP_MPA_OFFLOAD_RAMROD_DATA_RTR_SUPPORTED, ep->rtr_type); ep->state = QED_IWARP_EP_MPA_OFFLOADED; rc = qed_spq_post(p_hwfn, p_ent, NULL); if (!reject) ep->cid = qp->icid; /* Now they're migrated. */ DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "QP(0x%x) EP(0x%x) MPA Offload rc = %d IRD=0x%x ORD=0x%x rtr_type=%d mpa_rev=%d reject=%d\n", reject ? 0xffff : qp->icid, ep->tcp_cid, rc, ep->cm_info.ird, ep->cm_info.ord, ep->rtr_type, ep->mpa_rev, reject); return rc; } static void qed_iwarp_return_ep(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep) { ep->state = QED_IWARP_EP_INIT; if (ep->qp) ep->qp->ep = NULL; ep->qp = NULL; memset(&ep->cm_info, 0, sizeof(ep->cm_info)); if (ep->tcp_cid == QED_IWARP_INVALID_TCP_CID) { /* We don't care about the return code, it's ok if tcp_cid * remains invalid...in this case we'll defer allocation */ qed_iwarp_alloc_tcp_cid(p_hwfn, &ep->tcp_cid); } spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); list_move_tail(&ep->list_entry, &p_hwfn->p_rdma_info->iwarp.ep_free_list); spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); } static void qed_iwarp_parse_private_data(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep) { struct mpa_v2_hdr *mpa_v2_params; union async_output *async_data; u16 mpa_ird, mpa_ord; u8 mpa_data_size = 0; if (MPA_REV2(p_hwfn->p_rdma_info->iwarp.mpa_rev)) { mpa_v2_params = (struct mpa_v2_hdr *)(ep->ep_buffer_virt->in_pdata); mpa_data_size = sizeof(*mpa_v2_params); mpa_ird = ntohs(mpa_v2_params->ird); mpa_ord = ntohs(mpa_v2_params->ord); ep->cm_info.ird = (u8)(mpa_ord & MPA_V2_IRD_ORD_MASK); ep->cm_info.ord = (u8)(mpa_ird & MPA_V2_IRD_ORD_MASK); } async_data = &ep->ep_buffer_virt->async_output; ep->cm_info.private_data = ep->ep_buffer_virt->in_pdata + mpa_data_size; ep->cm_info.private_data_len = async_data->mpa_response.ulp_data_len - mpa_data_size; } static void qed_iwarp_mpa_reply_arrived(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep) { struct qed_iwarp_cm_event_params params; if (ep->connect_mode == TCP_CONNECT_PASSIVE) { DP_NOTICE(p_hwfn, "MPA reply event not expected on passive side!\n"); return; } params.event = QED_IWARP_EVENT_ACTIVE_MPA_REPLY; qed_iwarp_parse_private_data(p_hwfn, ep); DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "MPA_NEGOTIATE (v%d): ORD: 0x%x IRD: 0x%x\n", ep->mpa_rev, ep->cm_info.ord, ep->cm_info.ird); params.cm_info = &ep->cm_info; params.ep_context = ep; params.status = 0; ep->mpa_reply_processed = true; ep->event_cb(ep->cb_context, ¶ms); } #define QED_IWARP_CONNECT_MODE_STRING(ep) \ ((ep)->connect_mode == TCP_CONNECT_PASSIVE) ? "Passive" : "Active" /* Called as a result of the event: * IWARP_EVENT_TYPE_ASYNC_MPA_HANDSHAKE_COMPLETE */ static void qed_iwarp_mpa_complete(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep, u8 fw_return_code) { struct qed_iwarp_cm_event_params params; if (ep->connect_mode == TCP_CONNECT_ACTIVE) params.event = QED_IWARP_EVENT_ACTIVE_COMPLETE; else params.event = QED_IWARP_EVENT_PASSIVE_COMPLETE; if (ep->connect_mode == TCP_CONNECT_ACTIVE && !ep->mpa_reply_processed) qed_iwarp_parse_private_data(p_hwfn, ep); DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "MPA_NEGOTIATE (v%d): ORD: 0x%x IRD: 0x%x\n", ep->mpa_rev, ep->cm_info.ord, ep->cm_info.ird); params.cm_info = &ep->cm_info; params.ep_context = ep; switch (fw_return_code) { case RDMA_RETURN_OK: ep->qp->max_rd_atomic_req = ep->cm_info.ord; ep->qp->max_rd_atomic_resp = ep->cm_info.ird; qed_iwarp_modify_qp(p_hwfn, ep->qp, QED_IWARP_QP_STATE_RTS, 1); ep->state = QED_IWARP_EP_ESTABLISHED; params.status = 0; break; case IWARP_CONN_ERROR_MPA_TIMEOUT: DP_NOTICE(p_hwfn, "%s(0x%x) MPA timeout\n", QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid); params.status = -EBUSY; break; case IWARP_CONN_ERROR_MPA_ERROR_REJECT: DP_NOTICE(p_hwfn, "%s(0x%x) MPA Reject\n", QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid); params.status = -ECONNREFUSED; break; case IWARP_CONN_ERROR_MPA_RST: DP_NOTICE(p_hwfn, "%s(0x%x) MPA reset(tcp cid: 0x%x)\n", QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid, ep->tcp_cid); params.status = -ECONNRESET; break; case IWARP_CONN_ERROR_MPA_FIN: DP_NOTICE(p_hwfn, "%s(0x%x) MPA received FIN\n", QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid); params.status = -ECONNREFUSED; break; case IWARP_CONN_ERROR_MPA_INSUF_IRD: DP_NOTICE(p_hwfn, "%s(0x%x) MPA insufficient ird\n", QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid); params.status = -ECONNREFUSED; break; case IWARP_CONN_ERROR_MPA_RTR_MISMATCH: DP_NOTICE(p_hwfn, "%s(0x%x) MPA RTR MISMATCH\n", QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid); params.status = -ECONNREFUSED; break; case IWARP_CONN_ERROR_MPA_INVALID_PACKET: DP_NOTICE(p_hwfn, "%s(0x%x) MPA Invalid Packet\n", QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid); params.status = -ECONNREFUSED; break; case IWARP_CONN_ERROR_MPA_LOCAL_ERROR: DP_NOTICE(p_hwfn, "%s(0x%x) MPA Local Error\n", QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid); params.status = -ECONNREFUSED; break; case IWARP_CONN_ERROR_MPA_TERMINATE: DP_NOTICE(p_hwfn, "%s(0x%x) MPA TERMINATE\n", QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid); params.status = -ECONNREFUSED; break; default: params.status = -ECONNRESET; break; } if (fw_return_code != RDMA_RETURN_OK) /* paired with READ_ONCE in destroy_qp */ smp_store_release(&ep->state, QED_IWARP_EP_CLOSED); ep->event_cb(ep->cb_context, ¶ms); /* on passive side, if there is no associated QP (REJECT) we need to * return the ep to the pool, (in the regular case we add an element * in accept instead of this one. * In both cases we need to remove it from the ep_list. */ if (fw_return_code != RDMA_RETURN_OK) { ep->tcp_cid = QED_IWARP_INVALID_TCP_CID; if ((ep->connect_mode == TCP_CONNECT_PASSIVE) && (!ep->qp)) { /* Rejected */ qed_iwarp_return_ep(p_hwfn, ep); } else { spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); list_del(&ep->list_entry); spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); } } } static void qed_iwarp_mpa_v2_set_private(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep, u8 *mpa_data_size) { struct mpa_v2_hdr *mpa_v2_params; u16 mpa_ird, mpa_ord; *mpa_data_size = 0; if (MPA_REV2(ep->mpa_rev)) { mpa_v2_params = (struct mpa_v2_hdr *)ep->ep_buffer_virt->out_pdata; *mpa_data_size = sizeof(*mpa_v2_params); mpa_ird = (u16)ep->cm_info.ird; mpa_ord = (u16)ep->cm_info.ord; if (ep->rtr_type != MPA_RTR_TYPE_NONE) { mpa_ird |= MPA_V2_PEER2PEER_MODEL; if (ep->rtr_type & MPA_RTR_TYPE_ZERO_SEND) mpa_ird |= MPA_V2_SEND_RTR; if (ep->rtr_type & MPA_RTR_TYPE_ZERO_WRITE) mpa_ord |= MPA_V2_WRITE_RTR; if (ep->rtr_type & MPA_RTR_TYPE_ZERO_READ) mpa_ord |= MPA_V2_READ_RTR; } mpa_v2_params->ird = htons(mpa_ird); mpa_v2_params->ord = htons(mpa_ord); DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "MPA_NEGOTIATE Header: [%x ord:%x ird] %x ord:%x ird:%x peer2peer:%x rtr_send:%x rtr_write:%x rtr_read:%x\n", mpa_v2_params->ird, mpa_v2_params->ord, *((u32 *)mpa_v2_params), mpa_ord & MPA_V2_IRD_ORD_MASK, mpa_ird & MPA_V2_IRD_ORD_MASK, !!(mpa_ird & MPA_V2_PEER2PEER_MODEL), !!(mpa_ird & MPA_V2_SEND_RTR), !!(mpa_ord & MPA_V2_WRITE_RTR), !!(mpa_ord & MPA_V2_READ_RTR)); } } int qed_iwarp_connect(void *rdma_cxt, struct qed_iwarp_connect_in *iparams, struct qed_iwarp_connect_out *oparams) { struct qed_hwfn *p_hwfn = rdma_cxt; struct qed_iwarp_info *iwarp_info; struct qed_iwarp_ep *ep; u8 mpa_data_size = 0; u32 cid; int rc; if ((iparams->cm_info.ord > QED_IWARP_ORD_DEFAULT) || (iparams->cm_info.ird > QED_IWARP_IRD_DEFAULT)) { DP_NOTICE(p_hwfn, "QP(0x%x) ERROR: Invalid ord(0x%x)/ird(0x%x)\n", iparams->qp->icid, iparams->cm_info.ord, iparams->cm_info.ird); return -EINVAL; } iwarp_info = &p_hwfn->p_rdma_info->iwarp; /* Allocate ep object */ rc = qed_iwarp_alloc_cid(p_hwfn, &cid); if (rc) return rc; rc = qed_iwarp_create_ep(p_hwfn, &ep); if (rc) goto err; ep->tcp_cid = cid; spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); list_add_tail(&ep->list_entry, &p_hwfn->p_rdma_info->iwarp.ep_list); spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); ep->qp = iparams->qp; ep->qp->ep = ep; ether_addr_copy(ep->remote_mac_addr, iparams->remote_mac_addr); ether_addr_copy(ep->local_mac_addr, iparams->local_mac_addr); memcpy(&ep->cm_info, &iparams->cm_info, sizeof(ep->cm_info)); ep->cm_info.ord = iparams->cm_info.ord; ep->cm_info.ird = iparams->cm_info.ird; ep->rtr_type = iwarp_info->rtr_type; if (!iwarp_info->peer2peer) ep->rtr_type = MPA_RTR_TYPE_NONE; if ((ep->rtr_type & MPA_RTR_TYPE_ZERO_READ) && (ep->cm_info.ord == 0)) ep->cm_info.ord = 1; ep->mpa_rev = iwarp_info->mpa_rev; qed_iwarp_mpa_v2_set_private(p_hwfn, ep, &mpa_data_size); ep->cm_info.private_data = ep->ep_buffer_virt->out_pdata; ep->cm_info.private_data_len = iparams->cm_info.private_data_len + mpa_data_size; memcpy((u8 *)ep->ep_buffer_virt->out_pdata + mpa_data_size, iparams->cm_info.private_data, iparams->cm_info.private_data_len); ep->mss = iparams->mss; ep->mss = min_t(u16, QED_IWARP_MAX_FW_MSS, ep->mss); ep->event_cb = iparams->event_cb; ep->cb_context = iparams->cb_context; ep->connect_mode = TCP_CONNECT_ACTIVE; oparams->ep_context = ep; rc = qed_iwarp_tcp_offload(p_hwfn, ep); DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "QP(0x%x) EP(0x%x) rc = %d\n", iparams->qp->icid, ep->tcp_cid, rc); if (rc) { qed_iwarp_destroy_ep(p_hwfn, ep, true); goto err; } return rc; err: qed_iwarp_cid_cleaned(p_hwfn, cid); return rc; } static struct qed_iwarp_ep *qed_iwarp_get_free_ep(struct qed_hwfn *p_hwfn) { struct qed_iwarp_ep *ep = NULL; int rc; spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); if (list_empty(&p_hwfn->p_rdma_info->iwarp.ep_free_list)) { DP_ERR(p_hwfn, "Ep list is empty\n"); goto out; } ep = list_first_entry(&p_hwfn->p_rdma_info->iwarp.ep_free_list, struct qed_iwarp_ep, list_entry); /* in some cases we could have failed allocating a tcp cid when added * from accept / failure... retry now..this is not the common case. */ if (ep->tcp_cid == QED_IWARP_INVALID_TCP_CID) { rc = qed_iwarp_alloc_tcp_cid(p_hwfn, &ep->tcp_cid); /* if we fail we could look for another entry with a valid * tcp_cid, but since we don't expect to reach this anyway * it's not worth the handling */ if (rc) { ep->tcp_cid = QED_IWARP_INVALID_TCP_CID; ep = NULL; goto out; } } list_del(&ep->list_entry); out: spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); return ep; } #define QED_IWARP_MAX_CID_CLEAN_TIME 100 #define QED_IWARP_MAX_NO_PROGRESS_CNT 5 /* This function waits for all the bits of a bmap to be cleared, as long as * there is progress ( i.e. the number of bits left to be cleared decreases ) * the function continues. */ static int qed_iwarp_wait_cid_map_cleared(struct qed_hwfn *p_hwfn, struct qed_bmap *bmap) { int prev_weight = 0; int wait_count = 0; int weight = 0; weight = bitmap_weight(bmap->bitmap, bmap->max_count); prev_weight = weight; while (weight) { msleep(QED_IWARP_MAX_CID_CLEAN_TIME); weight = bitmap_weight(bmap->bitmap, bmap->max_count); if (prev_weight == weight) { wait_count++; } else { prev_weight = weight; wait_count = 0; } if (wait_count > QED_IWARP_MAX_NO_PROGRESS_CNT) { DP_NOTICE(p_hwfn, "%s bitmap wait timed out (%d cids pending)\n", bmap->name, weight); return -EBUSY; } } return 0; } static int qed_iwarp_wait_for_all_cids(struct qed_hwfn *p_hwfn) { int rc; int i; rc = qed_iwarp_wait_cid_map_cleared(p_hwfn, &p_hwfn->p_rdma_info->tcp_cid_map); if (rc) return rc; /* Now free the tcp cids from the main cid map */ for (i = 0; i < QED_IWARP_PREALLOC_CNT; i++) qed_bmap_release_id(p_hwfn, &p_hwfn->p_rdma_info->cid_map, i); /* Now wait for all cids to be completed */ return qed_iwarp_wait_cid_map_cleared(p_hwfn, &p_hwfn->p_rdma_info->cid_map); } static void qed_iwarp_free_prealloc_ep(struct qed_hwfn *p_hwfn) { struct qed_iwarp_ep *ep; while (!list_empty(&p_hwfn->p_rdma_info->iwarp.ep_free_list)) { spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); ep = list_first_entry(&p_hwfn->p_rdma_info->iwarp.ep_free_list, struct qed_iwarp_ep, list_entry); if (!ep) { spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); break; } list_del(&ep->list_entry); spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); if (ep->tcp_cid != QED_IWARP_INVALID_TCP_CID) qed_iwarp_cid_cleaned(p_hwfn, ep->tcp_cid); qed_iwarp_destroy_ep(p_hwfn, ep, false); } } static int qed_iwarp_prealloc_ep(struct qed_hwfn *p_hwfn, bool init) { struct qed_iwarp_ep *ep; int rc = 0; int count; u32 cid; int i; count = init ? QED_IWARP_PREALLOC_CNT : 1; for (i = 0; i < count; i++) { rc = qed_iwarp_create_ep(p_hwfn, &ep); if (rc) return rc; /* During initialization we allocate from the main pool, * afterwards we allocate only from the tcp_cid. */ if (init) { rc = qed_iwarp_alloc_cid(p_hwfn, &cid); if (rc) goto err; qed_iwarp_set_tcp_cid(p_hwfn, cid); } else { /* We don't care about the return code, it's ok if * tcp_cid remains invalid...in this case we'll * defer allocation */ qed_iwarp_alloc_tcp_cid(p_hwfn, &cid); } ep->tcp_cid = cid; spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); list_add_tail(&ep->list_entry, &p_hwfn->p_rdma_info->iwarp.ep_free_list); spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); } return rc; err: qed_iwarp_destroy_ep(p_hwfn, ep, false); return rc; } int qed_iwarp_alloc(struct qed_hwfn *p_hwfn) { int rc; /* Allocate bitmap for tcp cid. These are used by passive side * to ensure it can allocate a tcp cid during dpc that was * pre-acquired and doesn't require dynamic allocation of ilt */ rc = qed_rdma_bmap_alloc(p_hwfn, &p_hwfn->p_rdma_info->tcp_cid_map, QED_IWARP_PREALLOC_CNT, "TCP_CID"); if (rc) { DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Failed to allocate tcp cid, rc = %d\n", rc); return rc; } INIT_LIST_HEAD(&p_hwfn->p_rdma_info->iwarp.ep_free_list); spin_lock_init(&p_hwfn->p_rdma_info->iwarp.iw_lock); rc = qed_iwarp_prealloc_ep(p_hwfn, true); if (rc) return rc; return qed_ooo_alloc(p_hwfn); } void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn) { struct qed_iwarp_info *iwarp_info = &p_hwfn->p_rdma_info->iwarp; qed_ooo_free(p_hwfn); qed_rdma_bmap_free(p_hwfn, &p_hwfn->p_rdma_info->tcp_cid_map, 1); kfree(iwarp_info->mpa_bufs); kfree(iwarp_info->partial_fpdus); kfree(iwarp_info->mpa_intermediate_buf); } int qed_iwarp_accept(void *rdma_cxt, struct qed_iwarp_accept_in *iparams) { struct qed_hwfn *p_hwfn = rdma_cxt; struct qed_iwarp_ep *ep; u8 mpa_data_size = 0; int rc; ep = iparams->ep_context; if (!ep) { DP_ERR(p_hwfn, "Ep Context receive in accept is NULL\n"); return -EINVAL; } DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "QP(0x%x) EP(0x%x)\n", iparams->qp->icid, ep->tcp_cid); if ((iparams->ord > QED_IWARP_ORD_DEFAULT) || (iparams->ird > QED_IWARP_IRD_DEFAULT)) { DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "QP(0x%x) EP(0x%x) ERROR: Invalid ord(0x%x)/ird(0x%x)\n", iparams->qp->icid, ep->tcp_cid, iparams->ord, iparams->ord); return -EINVAL; } qed_iwarp_prealloc_ep(p_hwfn, false); ep->cb_context = iparams->cb_context; ep->qp = iparams->qp; ep->qp->ep = ep; if (ep->mpa_rev == MPA_NEGOTIATION_TYPE_ENHANCED) { /* Negotiate ord/ird: if upperlayer requested ord larger than * ird advertised by remote, we need to decrease our ord */ if (iparams->ord > ep->cm_info.ird) iparams->ord = ep->cm_info.ird; if ((ep->rtr_type & MPA_RTR_TYPE_ZERO_READ) && (iparams->ird == 0)) iparams->ird = 1; } /* Update cm_info ord/ird to be negotiated values */ ep->cm_info.ord = iparams->ord; ep->cm_info.ird = iparams->ird; qed_iwarp_mpa_v2_set_private(p_hwfn, ep, &mpa_data_size); ep->cm_info.private_data = ep->ep_buffer_virt->out_pdata; ep->cm_info.private_data_len = iparams->private_data_len + mpa_data_size; memcpy((u8 *)ep->ep_buffer_virt->out_pdata + mpa_data_size, iparams->private_data, iparams->private_data_len); rc = qed_iwarp_mpa_offload(p_hwfn, ep); if (rc) qed_iwarp_modify_qp(p_hwfn, iparams->qp, QED_IWARP_QP_STATE_ERROR, 1); return rc; } int qed_iwarp_reject(void *rdma_cxt, struct qed_iwarp_reject_in *iparams) { struct qed_hwfn *p_hwfn = rdma_cxt; struct qed_iwarp_ep *ep; u8 mpa_data_size = 0; ep = iparams->ep_context; if (!ep) { DP_ERR(p_hwfn, "Ep Context receive in reject is NULL\n"); return -EINVAL; } DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "EP(0x%x)\n", ep->tcp_cid); ep->cb_context = iparams->cb_context; ep->qp = NULL; qed_iwarp_mpa_v2_set_private(p_hwfn, ep, &mpa_data_size); ep->cm_info.private_data = ep->ep_buffer_virt->out_pdata; ep->cm_info.private_data_len = iparams->private_data_len + mpa_data_size; memcpy((u8 *)ep->ep_buffer_virt->out_pdata + mpa_data_size, iparams->private_data, iparams->private_data_len); return qed_iwarp_mpa_offload(p_hwfn, ep); } static void qed_iwarp_print_cm_info(struct qed_hwfn *p_hwfn, struct qed_iwarp_cm_info *cm_info) { DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "ip_version = %d\n", cm_info->ip_version); if (cm_info->ip_version == QED_TCP_IPV4) DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "remote_ip %pI4h:%x, local_ip %pI4h:%x vlan=%x\n", cm_info->remote_ip, cm_info->remote_port, cm_info->local_ip, cm_info->local_port, cm_info->vlan); else DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "remote_ip %pI6:%x, local_ip %pI6:%x vlan=%x\n", cm_info->remote_ip, cm_info->remote_port, cm_info->local_ip, cm_info->local_port, cm_info->vlan); DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "private_data_len = %x ord = %d, ird = %d\n", cm_info->private_data_len, cm_info->ord, cm_info->ird); } static int qed_iwarp_ll2_post_rx(struct qed_hwfn *p_hwfn, struct qed_iwarp_ll2_buff *buf, u8 handle) { int rc; rc = qed_ll2_post_rx_buffer(p_hwfn, handle, buf->data_phys_addr, (u16)buf->buff_size, buf, 1); if (rc) { DP_NOTICE(p_hwfn, "Failed to repost rx buffer to ll2 rc = %d, handle=%d\n", rc, handle); dma_free_coherent(&p_hwfn->cdev->pdev->dev, buf->buff_size, buf->data, buf->data_phys_addr); kfree(buf); } return rc; } static bool qed_iwarp_ep_exists(struct qed_hwfn *p_hwfn, struct qed_iwarp_cm_info *cm_info) { struct qed_iwarp_ep *ep = NULL; bool found = false; list_for_each_entry(ep, &p_hwfn->p_rdma_info->iwarp.ep_list, list_entry) { if ((ep->cm_info.local_port == cm_info->local_port) && (ep->cm_info.remote_port == cm_info->remote_port) && (ep->cm_info.vlan == cm_info->vlan) && !memcmp(&ep->cm_info.local_ip, cm_info->local_ip, sizeof(cm_info->local_ip)) && !memcmp(&ep->cm_info.remote_ip, cm_info->remote_ip, sizeof(cm_info->remote_ip))) { found = true; break; } } if (found) { DP_NOTICE(p_hwfn, "SYN received on active connection - dropping\n"); qed_iwarp_print_cm_info(p_hwfn, cm_info); return true; } return false; } static struct qed_iwarp_listener * qed_iwarp_get_listener(struct qed_hwfn *p_hwfn, struct qed_iwarp_cm_info *cm_info) { struct qed_iwarp_listener *listener = NULL; static const u32 ip_zero[4] = { 0, 0, 0, 0 }; bool found = false; qed_iwarp_print_cm_info(p_hwfn, cm_info); list_for_each_entry(listener, &p_hwfn->p_rdma_info->iwarp.listen_list, list_entry) { if (listener->port == cm_info->local_port) { if (!memcmp(listener->ip_addr, ip_zero, sizeof(ip_zero))) { found = true; break; } if (!memcmp(listener->ip_addr, cm_info->local_ip, sizeof(cm_info->local_ip)) && (listener->vlan == cm_info->vlan)) { found = true; break; } } } if (found) { DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "listener found = %p\n", listener); return listener; } DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "listener not found\n"); return NULL; } static int qed_iwarp_parse_rx_pkt(struct qed_hwfn *p_hwfn, struct qed_iwarp_cm_info *cm_info, void *buf, u8 *remote_mac_addr, u8 *local_mac_addr, int *payload_len, int *tcp_start_offset) { struct vlan_ethhdr *vethh; bool vlan_valid = false; struct ipv6hdr *ip6h; struct ethhdr *ethh; struct tcphdr *tcph; struct iphdr *iph; int eth_hlen; int ip_hlen; int eth_type; int i; ethh = buf; eth_type = ntohs(ethh->h_proto); if (eth_type == ETH_P_8021Q) { vlan_valid = true; vethh = (struct vlan_ethhdr *)ethh; cm_info->vlan = ntohs(vethh->h_vlan_TCI) & VLAN_VID_MASK; eth_type = ntohs(vethh->h_vlan_encapsulated_proto); } eth_hlen = ETH_HLEN + (vlan_valid ? sizeof(u32) : 0); if (!ether_addr_equal(ethh->h_dest, p_hwfn->p_rdma_info->iwarp.mac_addr)) { DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Got unexpected mac %pM instead of %pM\n", ethh->h_dest, p_hwfn->p_rdma_info->iwarp.mac_addr); return -EINVAL; } ether_addr_copy(remote_mac_addr, ethh->h_source); ether_addr_copy(local_mac_addr, ethh->h_dest); DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "eth_type =%d source mac: %pM\n", eth_type, ethh->h_source); DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "eth_hlen=%d destination mac: %pM\n", eth_hlen, ethh->h_dest); iph = (struct iphdr *)((u8 *)(ethh) + eth_hlen); if (eth_type == ETH_P_IP) { if (iph->protocol != IPPROTO_TCP) { DP_NOTICE(p_hwfn, "Unexpected ip protocol on ll2 %x\n", iph->protocol); return -EINVAL; } cm_info->local_ip[0] = ntohl(iph->daddr); cm_info->remote_ip[0] = ntohl(iph->saddr); cm_info->ip_version = QED_TCP_IPV4; ip_hlen = (iph->ihl) * sizeof(u32); *payload_len = ntohs(iph->tot_len) - ip_hlen; } else if (eth_type == ETH_P_IPV6) { ip6h = (struct ipv6hdr *)iph; if (ip6h->nexthdr != IPPROTO_TCP) { DP_NOTICE(p_hwfn, "Unexpected ip protocol on ll2 %x\n", iph->protocol); return -EINVAL; } for (i = 0; i < 4; i++) { cm_info->local_ip[i] = ntohl(ip6h->daddr.in6_u.u6_addr32[i]); cm_info->remote_ip[i] = ntohl(ip6h->saddr.in6_u.u6_addr32[i]); } cm_info->ip_version = QED_TCP_IPV6; ip_hlen = sizeof(*ip6h); *payload_len = ntohs(ip6h->payload_len); } else { DP_NOTICE(p_hwfn, "Unexpected ethertype on ll2 %x\n", eth_type); return -EINVAL; } tcph = (struct tcphdr *)((u8 *)iph + ip_hlen); if (!tcph->syn) { DP_NOTICE(p_hwfn, "Only SYN type packet expected on this ll2 conn, iph->ihl=%d source=%d dest=%d\n", iph->ihl, tcph->source, tcph->dest); return -EINVAL; } cm_info->local_port = ntohs(tcph->dest); cm_info->remote_port = ntohs(tcph->source); qed_iwarp_print_cm_info(p_hwfn, cm_info); *tcp_start_offset = eth_hlen + ip_hlen; return 0; } static struct qed_iwarp_fpdu *qed_iwarp_get_curr_fpdu(struct qed_hwfn *p_hwfn, u16 cid) { struct qed_iwarp_info *iwarp_info = &p_hwfn->p_rdma_info->iwarp; struct qed_iwarp_fpdu *partial_fpdu; u32 idx; idx = cid - qed_cxt_get_proto_cid_start(p_hwfn, PROTOCOLID_IWARP); if (idx >= iwarp_info->max_num_partial_fpdus) { DP_ERR(p_hwfn, "Invalid cid %x max_num_partial_fpdus=%x\n", cid, iwarp_info->max_num_partial_fpdus); return NULL; } partial_fpdu = &iwarp_info->partial_fpdus[idx]; return partial_fpdu; } enum qed_iwarp_mpa_pkt_type { QED_IWARP_MPA_PKT_PACKED, QED_IWARP_MPA_PKT_PARTIAL, QED_IWARP_MPA_PKT_UNALIGNED }; #define QED_IWARP_INVALID_FPDU_LENGTH 0xffff #define QED_IWARP_MPA_FPDU_LENGTH_SIZE (2) #define QED_IWARP_MPA_CRC32_DIGEST_SIZE (4) /* Pad to multiple of 4 */ #define QED_IWARP_PDU_DATA_LEN_WITH_PAD(data_len) ALIGN(data_len, 4) #define QED_IWARP_FPDU_LEN_WITH_PAD(_mpa_len) \ (QED_IWARP_PDU_DATA_LEN_WITH_PAD((_mpa_len) + \ QED_IWARP_MPA_FPDU_LENGTH_SIZE) + \ QED_IWARP_MPA_CRC32_DIGEST_SIZE) /* fpdu can be fragmented over maximum 3 bds: header, partial mpa, unaligned */ #define QED_IWARP_MAX_BDS_PER_FPDU 3 static const char * const pkt_type_str[] = { "QED_IWARP_MPA_PKT_PACKED", "QED_IWARP_MPA_PKT_PARTIAL", "QED_IWARP_MPA_PKT_UNALIGNED" }; static int qed_iwarp_recycle_pkt(struct qed_hwfn *p_hwfn, struct qed_iwarp_fpdu *fpdu, struct qed_iwarp_ll2_buff *buf); static enum qed_iwarp_mpa_pkt_type qed_iwarp_mpa_classify(struct qed_hwfn *p_hwfn, struct qed_iwarp_fpdu *fpdu, u16 tcp_payload_len, u8 *mpa_data) { enum qed_iwarp_mpa_pkt_type pkt_type; u16 mpa_len; if (fpdu->incomplete_bytes) { pkt_type = QED_IWARP_MPA_PKT_UNALIGNED; goto out; } /* special case of one byte remaining... * lower byte will be read next packet */ if (tcp_payload_len == 1) { fpdu->fpdu_length = *mpa_data << BITS_PER_BYTE; pkt_type = QED_IWARP_MPA_PKT_PARTIAL; goto out; } mpa_len = ntohs(*((u16 *)(mpa_data))); fpdu->fpdu_length = QED_IWARP_FPDU_LEN_WITH_PAD(mpa_len); if (fpdu->fpdu_length <= tcp_payload_len) pkt_type = QED_IWARP_MPA_PKT_PACKED; else pkt_type = QED_IWARP_MPA_PKT_PARTIAL; out: DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "MPA_ALIGN: %s: fpdu_length=0x%x tcp_payload_len:0x%x\n", pkt_type_str[pkt_type], fpdu->fpdu_length, tcp_payload_len); return pkt_type; } static void qed_iwarp_init_fpdu(struct qed_iwarp_ll2_buff *buf, struct qed_iwarp_fpdu *fpdu, struct unaligned_opaque_data *pkt_data, u16 tcp_payload_size, u8 placement_offset) { fpdu->mpa_buf = buf; fpdu->pkt_hdr = buf->data_phys_addr + placement_offset; fpdu->pkt_hdr_size = pkt_data->tcp_payload_offset; fpdu->mpa_frag = buf->data_phys_addr + pkt_data->first_mpa_offset; fpdu->mpa_frag_virt = (u8 *)(buf->data) + pkt_data->first_mpa_offset; if (tcp_payload_size == 1) fpdu->incomplete_bytes = QED_IWARP_INVALID_FPDU_LENGTH; else if (tcp_payload_size < fpdu->fpdu_length) fpdu->incomplete_bytes = fpdu->fpdu_length - tcp_payload_size; else fpdu->incomplete_bytes = 0; /* complete fpdu */ fpdu->mpa_frag_len = fpdu->fpdu_length - fpdu->incomplete_bytes; } static int qed_iwarp_cp_pkt(struct qed_hwfn *p_hwfn, struct qed_iwarp_fpdu *fpdu, struct unaligned_opaque_data *pkt_data, struct qed_iwarp_ll2_buff *buf, u16 tcp_payload_size) { u8 *tmp_buf = p_hwfn->p_rdma_info->iwarp.mpa_intermediate_buf; int rc; /* need to copy the data from the partial packet stored in fpdu * to the new buf, for this we also need to move the data currently * placed on the buf. The assumption is that the buffer is big enough * since fpdu_length <= mss, we use an intermediate buffer since * we may need to copy the new data to an overlapping location */ if ((fpdu->mpa_frag_len + tcp_payload_size) > (u16)buf->buff_size) { DP_ERR(p_hwfn, "MPA ALIGN: Unexpected: buffer is not large enough for split fpdu buff_size = %d mpa_frag_len = %d, tcp_payload_size = %d, incomplete_bytes = %d\n", buf->buff_size, fpdu->mpa_frag_len, tcp_payload_size, fpdu->incomplete_bytes); return -EINVAL; } DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "MPA ALIGN Copying fpdu: [%p, %d] [%p, %d]\n", fpdu->mpa_frag_virt, fpdu->mpa_frag_len, (u8 *)(buf->data) + pkt_data->first_mpa_offset, tcp_payload_size); memcpy(tmp_buf, fpdu->mpa_frag_virt, fpdu->mpa_frag_len); memcpy(tmp_buf + fpdu->mpa_frag_len, (u8 *)(buf->data) + pkt_data->first_mpa_offset, tcp_payload_size); rc = qed_iwarp_recycle_pkt(p_hwfn, fpdu, fpdu->mpa_buf); if (rc) return rc; /* If we managed to post the buffer copy the data to the new buffer * o/w this will occur in the next round... */ memcpy((u8 *)(buf->data), tmp_buf, fpdu->mpa_frag_len + tcp_payload_size); fpdu->mpa_buf = buf; /* fpdu->pkt_hdr remains as is */ /* fpdu->mpa_frag is overridden with new buf */ fpdu->mpa_frag = buf->data_phys_addr; fpdu->mpa_frag_virt = buf->data; fpdu->mpa_frag_len += tcp_payload_size; fpdu->incomplete_bytes -= tcp_payload_size; DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "MPA ALIGN: split fpdu buff_size = %d mpa_frag_len = %d, tcp_payload_size = %d, incomplete_bytes = %d\n", buf->buff_size, fpdu->mpa_frag_len, tcp_payload_size, fpdu->incomplete_bytes); return 0; } static void qed_iwarp_update_fpdu_length(struct qed_hwfn *p_hwfn, struct qed_iwarp_fpdu *fpdu, u8 *mpa_data) { u16 mpa_len; /* Update incomplete packets if needed */ if (fpdu->incomplete_bytes == QED_IWARP_INVALID_FPDU_LENGTH) { /* Missing lower byte is now available */ mpa_len = fpdu->fpdu_length | *mpa_data; fpdu->fpdu_length = QED_IWARP_FPDU_LEN_WITH_PAD(mpa_len); /* one byte of hdr */ fpdu->mpa_frag_len = 1; fpdu->incomplete_bytes = fpdu->fpdu_length - 1; DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "MPA_ALIGN: Partial header mpa_len=%x fpdu_length=%x incomplete_bytes=%x\n", mpa_len, fpdu->fpdu_length, fpdu->incomplete_bytes); } } #define QED_IWARP_IS_RIGHT_EDGE(_curr_pkt) \ (GET_FIELD((_curr_pkt)->flags, \ UNALIGNED_OPAQUE_DATA_PKT_REACHED_WIN_RIGHT_EDGE)) /* This function is used to recycle a buffer using the ll2 drop option. It * uses the mechanism to ensure that all buffers posted to tx before this one * were completed. The buffer sent here will be sent as a cookie in the tx * completion function and can then be reposted to rx chain when done. The flow * that requires this is the flow where a FPDU splits over more than 3 tcp * segments. In this case the driver needs to re-post a rx buffer instead of * the one received, but driver can't simply repost a buffer it copied from * as there is a case where the buffer was originally a packed FPDU, and is * partially posted to FW. Driver needs to ensure FW is done with it. */ static int qed_iwarp_recycle_pkt(struct qed_hwfn *p_hwfn, struct qed_iwarp_fpdu *fpdu, struct qed_iwarp_ll2_buff *buf) { struct qed_ll2_tx_pkt_info tx_pkt; u8 ll2_handle; int rc; memset(&tx_pkt, 0, sizeof(tx_pkt)); tx_pkt.num_of_bds = 1; tx_pkt.tx_dest = QED_LL2_TX_DEST_DROP; tx_pkt.l4_hdr_offset_w = fpdu->pkt_hdr_size >> 2; tx_pkt.first_frag = fpdu->pkt_hdr; tx_pkt.first_frag_len = fpdu->pkt_hdr_size; buf->piggy_buf = NULL; tx_pkt.cookie = buf; ll2_handle = p_hwfn->p_rdma_info->iwarp.ll2_mpa_handle; rc = qed_ll2_prepare_tx_packet(p_hwfn, ll2_handle, &tx_pkt, true); if (rc) DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Can't drop packet rc=%d\n", rc); DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "MPA_ALIGN: send drop tx packet [%lx, 0x%x], buf=%p, rc=%d\n", (unsigned long int)tx_pkt.first_frag, tx_pkt.first_frag_len, buf, rc); return rc; } static int qed_iwarp_win_right_edge(struct qed_hwfn *p_hwfn, struct qed_iwarp_fpdu *fpdu) { struct qed_ll2_tx_pkt_info tx_pkt; u8 ll2_handle; int rc; memset(&tx_pkt, 0, sizeof(tx_pkt)); tx_pkt.num_of_bds = 1; tx_pkt.tx_dest = QED_LL2_TX_DEST_LB; tx_pkt.l4_hdr_offset_w = fpdu->pkt_hdr_size >> 2; tx_pkt.first_frag = fpdu->pkt_hdr; tx_pkt.first_frag_len = fpdu->pkt_hdr_size; tx_pkt.enable_ip_cksum = true; tx_pkt.enable_l4_cksum = true; tx_pkt.calc_ip_len = true; /* vlan overload with enum iwarp_ll2_tx_queues */ tx_pkt.vlan = IWARP_LL2_ALIGNED_RIGHT_TRIMMED_TX_QUEUE; ll2_handle = p_hwfn->p_rdma_info->iwarp.ll2_mpa_handle; rc = qed_ll2_prepare_tx_packet(p_hwfn, ll2_handle, &tx_pkt, true); if (rc) DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Can't send right edge rc=%d\n", rc); DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "MPA_ALIGN: Sent right edge FPDU num_bds=%d [%lx, 0x%x], rc=%d\n", tx_pkt.num_of_bds, (unsigned long int)tx_pkt.first_frag, tx_pkt.first_frag_len, rc); return rc; } static int qed_iwarp_send_fpdu(struct qed_hwfn *p_hwfn, struct qed_iwarp_fpdu *fpdu, struct unaligned_opaque_data *curr_pkt, struct qed_iwarp_ll2_buff *buf, u16 tcp_payload_size, enum qed_iwarp_mpa_pkt_type pkt_type) { struct qed_ll2_tx_pkt_info tx_pkt; u8 ll2_handle; int rc; memset(&tx_pkt, 0, sizeof(tx_pkt)); /* An unaligned packet means it's split over two tcp segments. So the * complete packet requires 3 bds, one for the header, one for the * part of the fpdu of the first tcp segment, and the last fragment * will point to the remainder of the fpdu. A packed pdu, requires only * two bds, one for the header and one for the data. */ tx_pkt.num_of_bds = (pkt_type == QED_IWARP_MPA_PKT_UNALIGNED) ? 3 : 2; tx_pkt.tx_dest = QED_LL2_TX_DEST_LB; tx_pkt.l4_hdr_offset_w = fpdu->pkt_hdr_size >> 2; /* offset in words */ /* Send the mpa_buf only with the last fpdu (in case of packed) */ if (pkt_type == QED_IWARP_MPA_PKT_UNALIGNED || tcp_payload_size <= fpdu->fpdu_length) tx_pkt.cookie = fpdu->mpa_buf; tx_pkt.first_frag = fpdu->pkt_hdr; tx_pkt.first_frag_len = fpdu->pkt_hdr_size; tx_pkt.enable_ip_cksum = true; tx_pkt.enable_l4_cksum = true; tx_pkt.calc_ip_len = true; /* vlan overload with enum iwarp_ll2_tx_queues */ tx_pkt.vlan = IWARP_LL2_ALIGNED_TX_QUEUE; /* special case of unaligned packet and not packed, need to send * both buffers as cookie to release. */ if (tcp_payload_size == fpdu->incomplete_bytes) fpdu->mpa_buf->piggy_buf = buf; ll2_handle = p_hwfn->p_rdma_info->iwarp.ll2_mpa_handle; /* Set first fragment to header */ rc = qed_ll2_prepare_tx_packet(p_hwfn, ll2_handle, &tx_pkt, true); if (rc) goto out; /* Set second fragment to first part of packet */ rc = qed_ll2_set_fragment_of_tx_packet(p_hwfn, ll2_handle, fpdu->mpa_frag, fpdu->mpa_frag_len); if (rc) goto out; if (!fpdu->incomplete_bytes) goto out; /* Set third fragment to second part of the packet */ rc = qed_ll2_set_fragment_of_tx_packet(p_hwfn, ll2_handle, buf->data_phys_addr + curr_pkt->first_mpa_offset, fpdu->incomplete_bytes); out: DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "MPA_ALIGN: Sent FPDU num_bds=%d first_frag_len=%x, mpa_frag_len=0x%x, incomplete_bytes:0x%x rc=%d\n", tx_pkt.num_of_bds, tx_pkt.first_frag_len, fpdu->mpa_frag_len, fpdu->incomplete_bytes, rc); return rc; } static void qed_iwarp_mpa_get_data(struct qed_hwfn *p_hwfn, struct unaligned_opaque_data *curr_pkt, u32 opaque_data0, u32 opaque_data1) { u64 opaque_data; opaque_data = HILO_64(opaque_data1, opaque_data0); *curr_pkt = *((struct unaligned_opaque_data *)&opaque_data); curr_pkt->first_mpa_offset = curr_pkt->tcp_payload_offset + le16_to_cpu(curr_pkt->first_mpa_offset); curr_pkt->cid = le32_to_cpu(curr_pkt->cid); } /* This function is called when an unaligned or incomplete MPA packet arrives * driver needs to align the packet, perhaps using previous data and send * it down to FW once it is aligned. */ static int qed_iwarp_process_mpa_pkt(struct qed_hwfn *p_hwfn, struct qed_iwarp_ll2_mpa_buf *mpa_buf) { struct unaligned_opaque_data *curr_pkt = &mpa_buf->data; struct qed_iwarp_ll2_buff *buf = mpa_buf->ll2_buf; enum qed_iwarp_mpa_pkt_type pkt_type; struct qed_iwarp_fpdu *fpdu; int rc = -EINVAL; u8 *mpa_data; fpdu = qed_iwarp_get_curr_fpdu(p_hwfn, curr_pkt->cid & 0xffff); if (!fpdu) { /* something corrupt with cid, post rx back */ DP_ERR(p_hwfn, "Invalid cid, drop and post back to rx cid=%x\n", curr_pkt->cid); goto err; } do { mpa_data = ((u8 *)(buf->data) + curr_pkt->first_mpa_offset); pkt_type = qed_iwarp_mpa_classify(p_hwfn, fpdu, mpa_buf->tcp_payload_len, mpa_data); switch (pkt_type) { case QED_IWARP_MPA_PKT_PARTIAL: qed_iwarp_init_fpdu(buf, fpdu, curr_pkt, mpa_buf->tcp_payload_len, mpa_buf->placement_offset); if (!QED_IWARP_IS_RIGHT_EDGE(curr_pkt)) { mpa_buf->tcp_payload_len = 0; break; } rc = qed_iwarp_win_right_edge(p_hwfn, fpdu); if (rc) { DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Can't send FPDU:reset rc=%d\n", rc); memset(fpdu, 0, sizeof(*fpdu)); break; } mpa_buf->tcp_payload_len = 0; break; case QED_IWARP_MPA_PKT_PACKED: qed_iwarp_init_fpdu(buf, fpdu, curr_pkt, mpa_buf->tcp_payload_len, mpa_buf->placement_offset); rc = qed_iwarp_send_fpdu(p_hwfn, fpdu, curr_pkt, buf, mpa_buf->tcp_payload_len, pkt_type); if (rc) { DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Can't send FPDU:reset rc=%d\n", rc); memset(fpdu, 0, sizeof(*fpdu)); break; } mpa_buf->tcp_payload_len -= fpdu->fpdu_length; curr_pkt->first_mpa_offset += fpdu->fpdu_length; break; case QED_IWARP_MPA_PKT_UNALIGNED: qed_iwarp_update_fpdu_length(p_hwfn, fpdu, mpa_data); if (mpa_buf->tcp_payload_len < fpdu->incomplete_bytes) { /* special handling of fpdu split over more * than 2 segments */ if (QED_IWARP_IS_RIGHT_EDGE(curr_pkt)) { rc = qed_iwarp_win_right_edge(p_hwfn, fpdu); /* packet will be re-processed later */ if (rc) return rc; } rc = qed_iwarp_cp_pkt(p_hwfn, fpdu, curr_pkt, buf, mpa_buf->tcp_payload_len); if (rc) /* packet will be re-processed later */ return rc; mpa_buf->tcp_payload_len = 0; break; } rc = qed_iwarp_send_fpdu(p_hwfn, fpdu, curr_pkt, buf, mpa_buf->tcp_payload_len, pkt_type); if (rc) { DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Can't send FPDU:delay rc=%d\n", rc); /* don't reset fpdu -> we need it for next * classify */ break; } mpa_buf->tcp_payload_len -= fpdu->incomplete_bytes; curr_pkt->first_mpa_offset += fpdu->incomplete_bytes; /* The framed PDU was sent - no more incomplete bytes */ fpdu->incomplete_bytes = 0; break; } } while (mpa_buf->tcp_payload_len && !rc); return rc; err: qed_iwarp_ll2_post_rx(p_hwfn, buf, p_hwfn->p_rdma_info->iwarp.ll2_mpa_handle); return rc; } static void qed_iwarp_process_pending_pkts(struct qed_hwfn *p_hwfn) { struct qed_iwarp_info *iwarp_info = &p_hwfn->p_rdma_info->iwarp; struct qed_iwarp_ll2_mpa_buf *mpa_buf = NULL; int rc; while (!list_empty(&iwarp_info->mpa_buf_pending_list)) { mpa_buf = list_first_entry(&iwarp_info->mpa_buf_pending_list, struct qed_iwarp_ll2_mpa_buf, list_entry); rc = qed_iwarp_process_mpa_pkt(p_hwfn, mpa_buf); /* busy means break and continue processing later, don't * remove the buf from the pending list. */ if (rc == -EBUSY) break; list_move_tail(&mpa_buf->list_entry, &iwarp_info->mpa_buf_list); if (rc) { /* different error, don't continue */ DP_NOTICE(p_hwfn, "process pkts failed rc=%d\n", rc); break; } } } static void qed_iwarp_ll2_comp_mpa_pkt(void *cxt, struct qed_ll2_comp_rx_data *data) { struct qed_iwarp_ll2_mpa_buf *mpa_buf; struct qed_iwarp_info *iwarp_info; struct qed_hwfn *p_hwfn = cxt; iwarp_info = &p_hwfn->p_rdma_info->iwarp; mpa_buf = list_first_entry(&iwarp_info->mpa_buf_list, struct qed_iwarp_ll2_mpa_buf, list_entry); if (!mpa_buf) { DP_ERR(p_hwfn, "No free mpa buf\n"); goto err; } list_del(&mpa_buf->list_entry); qed_iwarp_mpa_get_data(p_hwfn, &mpa_buf->data, data->opaque_data_0, data->opaque_data_1); DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "LL2 MPA CompRx payload_len:0x%x\tfirst_mpa_offset:0x%x\ttcp_payload_offset:0x%x\tflags:0x%x\tcid:0x%x\n", data->length.packet_length, mpa_buf->data.first_mpa_offset, mpa_buf->data.tcp_payload_offset, mpa_buf->data.flags, mpa_buf->data.cid); mpa_buf->ll2_buf = data->cookie; mpa_buf->tcp_payload_len = data->length.packet_length - mpa_buf->data.first_mpa_offset; mpa_buf->data.first_mpa_offset += data->u.placement_offset; mpa_buf->placement_offset = data->u.placement_offset; list_add_tail(&mpa_buf->list_entry, &iwarp_info->mpa_buf_pending_list); qed_iwarp_process_pending_pkts(p_hwfn); return; err: qed_iwarp_ll2_post_rx(p_hwfn, data->cookie, iwarp_info->ll2_mpa_handle); } static void qed_iwarp_ll2_comp_syn_pkt(void *cxt, struct qed_ll2_comp_rx_data *data) { struct qed_iwarp_ll2_buff *buf = data->cookie; struct qed_iwarp_listener *listener; struct qed_ll2_tx_pkt_info tx_pkt; struct qed_iwarp_cm_info cm_info; struct qed_hwfn *p_hwfn = cxt; u8 remote_mac_addr[ETH_ALEN]; u8 local_mac_addr[ETH_ALEN]; struct qed_iwarp_ep *ep; int tcp_start_offset; u8 ll2_syn_handle; int payload_len; u32 hdr_size; int rc; memset(&cm_info, 0, sizeof(cm_info)); ll2_syn_handle = p_hwfn->p_rdma_info->iwarp.ll2_syn_handle; /* Check if packet was received with errors... */ if (data->err_flags) { DP_NOTICE(p_hwfn, "Error received on SYN packet: 0x%x\n", data->err_flags); goto err; } if (GET_FIELD(data->parse_flags, PARSING_AND_ERR_FLAGS_L4CHKSMWASCALCULATED) && GET_FIELD(data->parse_flags, PARSING_AND_ERR_FLAGS_L4CHKSMERROR)) { DP_NOTICE(p_hwfn, "Syn packet received with checksum error\n"); goto err; } rc = qed_iwarp_parse_rx_pkt(p_hwfn, &cm_info, (u8 *)(buf->data) + data->u.placement_offset, remote_mac_addr, local_mac_addr, &payload_len, &tcp_start_offset); if (rc) goto err; /* Check if there is a listener for this 4-tuple+vlan */ listener = qed_iwarp_get_listener(p_hwfn, &cm_info); if (!listener) { DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "SYN received on tuple not listened on parse_flags=%d packet len=%d\n", data->parse_flags, data->length.packet_length); memset(&tx_pkt, 0, sizeof(tx_pkt)); tx_pkt.num_of_bds = 1; tx_pkt.l4_hdr_offset_w = (data->length.packet_length) >> 2; tx_pkt.tx_dest = QED_LL2_TX_DEST_LB; tx_pkt.first_frag = buf->data_phys_addr + data->u.placement_offset; tx_pkt.first_frag_len = data->length.packet_length; tx_pkt.cookie = buf; rc = qed_ll2_prepare_tx_packet(p_hwfn, ll2_syn_handle, &tx_pkt, true); if (rc) { DP_NOTICE(p_hwfn, "Can't post SYN back to chip rc=%d\n", rc); goto err; } return; } DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Received syn on listening port\n"); /* There may be an open ep on this connection if this is a syn * retrasnmit... need to make sure there isn't... */ if (qed_iwarp_ep_exists(p_hwfn, &cm_info)) goto err; ep = qed_iwarp_get_free_ep(p_hwfn); if (!ep) goto err; spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); list_add_tail(&ep->list_entry, &p_hwfn->p_rdma_info->iwarp.ep_list); spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); ether_addr_copy(ep->remote_mac_addr, remote_mac_addr); ether_addr_copy(ep->local_mac_addr, local_mac_addr); memcpy(&ep->cm_info, &cm_info, sizeof(ep->cm_info)); hdr_size = ((cm_info.ip_version == QED_TCP_IPV4) ? 40 : 60); ep->mss = p_hwfn->p_rdma_info->iwarp.max_mtu - hdr_size; ep->mss = min_t(u16, QED_IWARP_MAX_FW_MSS, ep->mss); ep->event_cb = listener->event_cb; ep->cb_context = listener->cb_context; ep->connect_mode = TCP_CONNECT_PASSIVE; ep->syn = buf; ep->syn_ip_payload_length = (u16)payload_len; ep->syn_phy_addr = buf->data_phys_addr + data->u.placement_offset + tcp_start_offset; rc = qed_iwarp_tcp_offload(p_hwfn, ep); if (rc) { qed_iwarp_return_ep(p_hwfn, ep); goto err; } return; err: qed_iwarp_ll2_post_rx(p_hwfn, buf, ll2_syn_handle); } static void qed_iwarp_ll2_rel_rx_pkt(void *cxt, u8 connection_handle, void *cookie, dma_addr_t rx_buf_addr, bool b_last_packet) { struct qed_iwarp_ll2_buff *buffer = cookie; struct qed_hwfn *p_hwfn = cxt; dma_free_coherent(&p_hwfn->cdev->pdev->dev, buffer->buff_size, buffer->data, buffer->data_phys_addr); kfree(buffer); } static void qed_iwarp_ll2_comp_tx_pkt(void *cxt, u8 connection_handle, void *cookie, dma_addr_t first_frag_addr, bool b_last_fragment, bool b_last_packet) { struct qed_iwarp_ll2_buff *buffer = cookie; struct qed_iwarp_ll2_buff *piggy; struct qed_hwfn *p_hwfn = cxt; if (!buffer) /* can happen in packed mpa unaligned... */ return; /* this was originally an rx packet, post it back */ piggy = buffer->piggy_buf; if (piggy) { buffer->piggy_buf = NULL; qed_iwarp_ll2_post_rx(p_hwfn, piggy, connection_handle); } qed_iwarp_ll2_post_rx(p_hwfn, buffer, connection_handle); if (connection_handle == p_hwfn->p_rdma_info->iwarp.ll2_mpa_handle) qed_iwarp_process_pending_pkts(p_hwfn); return; } static void qed_iwarp_ll2_rel_tx_pkt(void *cxt, u8 connection_handle, void *cookie, dma_addr_t first_frag_addr, bool b_last_fragment, bool b_last_packet) { struct qed_iwarp_ll2_buff *buffer = cookie; struct qed_hwfn *p_hwfn = cxt; if (!buffer) return; if (buffer->piggy_buf) { dma_free_coherent(&p_hwfn->cdev->pdev->dev, buffer->piggy_buf->buff_size, buffer->piggy_buf->data, buffer->piggy_buf->data_phys_addr); kfree(buffer->piggy_buf); } dma_free_coherent(&p_hwfn->cdev->pdev->dev, buffer->buff_size, buffer->data, buffer->data_phys_addr); kfree(buffer); } /* The only slowpath for iwarp ll2 is unalign flush. When this completion * is received, need to reset the FPDU. */ static void qed_iwarp_ll2_slowpath(void *cxt, u8 connection_handle, u32 opaque_data_0, u32 opaque_data_1) { struct unaligned_opaque_data unalign_data; struct qed_hwfn *p_hwfn = cxt; struct qed_iwarp_fpdu *fpdu; qed_iwarp_mpa_get_data(p_hwfn, &unalign_data, opaque_data_0, opaque_data_1); DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "(0x%x) Flush fpdu\n", unalign_data.cid); fpdu = qed_iwarp_get_curr_fpdu(p_hwfn, (u16)unalign_data.cid); if (fpdu) memset(fpdu, 0, sizeof(*fpdu)); } static int qed_iwarp_ll2_stop(struct qed_hwfn *p_hwfn) { struct qed_iwarp_info *iwarp_info = &p_hwfn->p_rdma_info->iwarp; int rc = 0; if (iwarp_info->ll2_syn_handle != QED_IWARP_HANDLE_INVAL) { rc = qed_ll2_terminate_connection(p_hwfn, iwarp_info->ll2_syn_handle); if (rc) DP_INFO(p_hwfn, "Failed to terminate syn connection\n"); qed_ll2_release_connection(p_hwfn, iwarp_info->ll2_syn_handle); iwarp_info->ll2_syn_handle = QED_IWARP_HANDLE_INVAL; } if (iwarp_info->ll2_ooo_handle != QED_IWARP_HANDLE_INVAL) { rc = qed_ll2_terminate_connection(p_hwfn, iwarp_info->ll2_ooo_handle); if (rc) DP_INFO(p_hwfn, "Failed to terminate ooo connection\n"); qed_ll2_release_connection(p_hwfn, iwarp_info->ll2_ooo_handle); iwarp_info->ll2_ooo_handle = QED_IWARP_HANDLE_INVAL; } if (iwarp_info->ll2_mpa_handle != QED_IWARP_HANDLE_INVAL) { rc = qed_ll2_terminate_connection(p_hwfn, iwarp_info->ll2_mpa_handle); if (rc) DP_INFO(p_hwfn, "Failed to terminate mpa connection\n"); qed_ll2_release_connection(p_hwfn, iwarp_info->ll2_mpa_handle); iwarp_info->ll2_mpa_handle = QED_IWARP_HANDLE_INVAL; } qed_llh_remove_mac_filter(p_hwfn->cdev, 0, p_hwfn->p_rdma_info->iwarp.mac_addr); return rc; } static int qed_iwarp_ll2_alloc_buffers(struct qed_hwfn *p_hwfn, int num_rx_bufs, int buff_size, u8 ll2_handle) { struct qed_iwarp_ll2_buff *buffer; int rc = 0; int i; for (i = 0; i < num_rx_bufs; i++) { buffer = kzalloc(sizeof(*buffer), GFP_KERNEL); if (!buffer) { rc = -ENOMEM; break; } buffer->data = dma_alloc_coherent(&p_hwfn->cdev->pdev->dev, buff_size, &buffer->data_phys_addr, GFP_KERNEL); if (!buffer->data) { kfree(buffer); rc = -ENOMEM; break; } buffer->buff_size = buff_size; rc = qed_iwarp_ll2_post_rx(p_hwfn, buffer, ll2_handle); if (rc) /* buffers will be deallocated by qed_ll2 */ break; } return rc; } #define QED_IWARP_MAX_BUF_SIZE(mtu) \ ALIGN((mtu) + ETH_HLEN + 2 * VLAN_HLEN + 2 + ETH_CACHE_LINE_SIZE, \ ETH_CACHE_LINE_SIZE) static int qed_iwarp_ll2_start(struct qed_hwfn *p_hwfn, struct qed_rdma_start_in_params *params, u32 rcv_wnd_size) { struct qed_iwarp_info *iwarp_info; struct qed_ll2_acquire_data data; struct qed_ll2_cbs cbs; u32 buff_size; u16 n_ooo_bufs; int rc = 0; int i; iwarp_info = &p_hwfn->p_rdma_info->iwarp; iwarp_info->ll2_syn_handle = QED_IWARP_HANDLE_INVAL; iwarp_info->ll2_ooo_handle = QED_IWARP_HANDLE_INVAL; iwarp_info->ll2_mpa_handle = QED_IWARP_HANDLE_INVAL; iwarp_info->max_mtu = params->max_mtu; ether_addr_copy(p_hwfn->p_rdma_info->iwarp.mac_addr, params->mac_addr); rc = qed_llh_add_mac_filter(p_hwfn->cdev, 0, params->mac_addr); if (rc) return rc; /* Start SYN connection */ cbs.rx_comp_cb = qed_iwarp_ll2_comp_syn_pkt; cbs.rx_release_cb = qed_iwarp_ll2_rel_rx_pkt; cbs.tx_comp_cb = qed_iwarp_ll2_comp_tx_pkt; cbs.tx_release_cb = qed_iwarp_ll2_rel_tx_pkt; cbs.slowpath_cb = NULL; cbs.cookie = p_hwfn; memset(&data, 0, sizeof(data)); data.input.conn_type = QED_LL2_TYPE_IWARP; /* SYN will use ctx based queues */ data.input.rx_conn_type = QED_LL2_RX_TYPE_CTX; data.input.mtu = params->max_mtu; data.input.rx_num_desc = QED_IWARP_LL2_SYN_RX_SIZE; data.input.tx_num_desc = QED_IWARP_LL2_SYN_TX_SIZE; data.input.tx_max_bds_per_packet = 1; /* will never be fragmented */ data.input.tx_tc = PKT_LB_TC; data.input.tx_dest = QED_LL2_TX_DEST_LB; data.p_connection_handle = &iwarp_info->ll2_syn_handle; data.cbs = &cbs; rc = qed_ll2_acquire_connection(p_hwfn, &data); if (rc) { DP_NOTICE(p_hwfn, "Failed to acquire LL2 connection\n"); qed_llh_remove_mac_filter(p_hwfn->cdev, 0, params->mac_addr); return rc; } rc = qed_ll2_establish_connection(p_hwfn, iwarp_info->ll2_syn_handle); if (rc) { DP_NOTICE(p_hwfn, "Failed to establish LL2 connection\n"); goto err; } buff_size = QED_IWARP_MAX_BUF_SIZE(params->max_mtu); rc = qed_iwarp_ll2_alloc_buffers(p_hwfn, QED_IWARP_LL2_SYN_RX_SIZE, buff_size, iwarp_info->ll2_syn_handle); if (rc) goto err; /* Start OOO connection */ data.input.conn_type = QED_LL2_TYPE_OOO; /* OOO/unaligned will use legacy ll2 queues (ram based) */ data.input.rx_conn_type = QED_LL2_RX_TYPE_LEGACY; data.input.mtu = params->max_mtu; n_ooo_bufs = (QED_IWARP_MAX_OOO * rcv_wnd_size) / iwarp_info->max_mtu; n_ooo_bufs = min_t(u32, n_ooo_bufs, QED_IWARP_LL2_OOO_MAX_RX_SIZE); data.input.rx_num_desc = n_ooo_bufs; data.input.rx_num_ooo_buffers = n_ooo_bufs; data.input.tx_max_bds_per_packet = 1; /* will never be fragmented */ data.input.tx_num_desc = QED_IWARP_LL2_OOO_DEF_TX_SIZE; data.p_connection_handle = &iwarp_info->ll2_ooo_handle; rc = qed_ll2_acquire_connection(p_hwfn, &data); if (rc) goto err; rc = qed_ll2_establish_connection(p_hwfn, iwarp_info->ll2_ooo_handle); if (rc) goto err; /* Start Unaligned MPA connection */ cbs.rx_comp_cb = qed_iwarp_ll2_comp_mpa_pkt; cbs.slowpath_cb = qed_iwarp_ll2_slowpath; memset(&data, 0, sizeof(data)); data.input.conn_type = QED_LL2_TYPE_IWARP; data.input.mtu = params->max_mtu; /* FW requires that once a packet arrives OOO, it must have at * least 2 rx buffers available on the unaligned connection * for handling the case that it is a partial fpdu. */ data.input.rx_num_desc = n_ooo_bufs * 2; data.input.tx_num_desc = data.input.rx_num_desc; data.input.tx_max_bds_per_packet = QED_IWARP_MAX_BDS_PER_FPDU; data.input.tx_tc = PKT_LB_TC; data.input.tx_dest = QED_LL2_TX_DEST_LB; data.p_connection_handle = &iwarp_info->ll2_mpa_handle; data.input.secondary_queue = true; data.cbs = &cbs; rc = qed_ll2_acquire_connection(p_hwfn, &data); if (rc) goto err; rc = qed_ll2_establish_connection(p_hwfn, iwarp_info->ll2_mpa_handle); if (rc) goto err; rc = qed_iwarp_ll2_alloc_buffers(p_hwfn, data.input.rx_num_desc, buff_size, iwarp_info->ll2_mpa_handle); if (rc) goto err; iwarp_info->partial_fpdus = kcalloc((u16)p_hwfn->p_rdma_info->num_qps, sizeof(*iwarp_info->partial_fpdus), GFP_KERNEL); if (!iwarp_info->partial_fpdus) goto err; iwarp_info->max_num_partial_fpdus = (u16)p_hwfn->p_rdma_info->num_qps; iwarp_info->mpa_intermediate_buf = kzalloc(buff_size, GFP_KERNEL); if (!iwarp_info->mpa_intermediate_buf) goto err; /* The mpa_bufs array serves for pending RX packets received on the * mpa ll2 that don't have place on the tx ring and require later * processing. We can't fail on allocation of such a struct therefore * we allocate enough to take care of all rx packets */ iwarp_info->mpa_bufs = kcalloc(data.input.rx_num_desc, sizeof(*iwarp_info->mpa_bufs), GFP_KERNEL); if (!iwarp_info->mpa_bufs) goto err; INIT_LIST_HEAD(&iwarp_info->mpa_buf_pending_list); INIT_LIST_HEAD(&iwarp_info->mpa_buf_list); for (i = 0; i < data.input.rx_num_desc; i++) list_add_tail(&iwarp_info->mpa_bufs[i].list_entry, &iwarp_info->mpa_buf_list); return rc; err: qed_iwarp_ll2_stop(p_hwfn); return rc; } static struct { u32 two_ports; u32 four_ports; } qed_iwarp_rcv_wnd_size[MAX_CHIP_IDS] = { {QED_IWARP_RCV_WND_SIZE_DEF_BB_2P, QED_IWARP_RCV_WND_SIZE_DEF_BB_4P}, {QED_IWARP_RCV_WND_SIZE_DEF_AH_2P, QED_IWARP_RCV_WND_SIZE_DEF_AH_4P} }; int qed_iwarp_setup(struct qed_hwfn *p_hwfn, struct qed_rdma_start_in_params *params) { struct qed_dev *cdev = p_hwfn->cdev; struct qed_iwarp_info *iwarp_info; enum chip_ids chip_id; u32 rcv_wnd_size; iwarp_info = &p_hwfn->p_rdma_info->iwarp; iwarp_info->tcp_flags = QED_IWARP_TS_EN; chip_id = QED_IS_BB(cdev) ? CHIP_BB : CHIP_K2; rcv_wnd_size = (qed_device_num_ports(cdev) == 4) ? qed_iwarp_rcv_wnd_size[chip_id].four_ports : qed_iwarp_rcv_wnd_size[chip_id].two_ports; /* value 0 is used for ilog2(QED_IWARP_RCV_WND_SIZE_MIN) */ iwarp_info->rcv_wnd_scale = ilog2(rcv_wnd_size) - ilog2(QED_IWARP_RCV_WND_SIZE_MIN); iwarp_info->rcv_wnd_size = rcv_wnd_size >> iwarp_info->rcv_wnd_scale; iwarp_info->crc_needed = QED_IWARP_PARAM_CRC_NEEDED; iwarp_info->mpa_rev = MPA_NEGOTIATION_TYPE_ENHANCED; iwarp_info->peer2peer = QED_IWARP_PARAM_P2P; iwarp_info->rtr_type = MPA_RTR_TYPE_ZERO_SEND | MPA_RTR_TYPE_ZERO_WRITE | MPA_RTR_TYPE_ZERO_READ; spin_lock_init(&p_hwfn->p_rdma_info->iwarp.qp_lock); INIT_LIST_HEAD(&p_hwfn->p_rdma_info->iwarp.ep_list); INIT_LIST_HEAD(&p_hwfn->p_rdma_info->iwarp.listen_list); qed_spq_register_async_cb(p_hwfn, PROTOCOLID_IWARP, qed_iwarp_async_event); qed_ooo_setup(p_hwfn); return qed_iwarp_ll2_start(p_hwfn, params, rcv_wnd_size); } int qed_iwarp_stop(struct qed_hwfn *p_hwfn) { int rc; qed_iwarp_free_prealloc_ep(p_hwfn); rc = qed_iwarp_wait_for_all_cids(p_hwfn); if (rc) return rc; return qed_iwarp_ll2_stop(p_hwfn); } static void qed_iwarp_qp_in_error(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep, u8 fw_return_code) { struct qed_iwarp_cm_event_params params; qed_iwarp_modify_qp(p_hwfn, ep->qp, QED_IWARP_QP_STATE_ERROR, true); params.event = QED_IWARP_EVENT_CLOSE; params.ep_context = ep; params.cm_info = &ep->cm_info; params.status = (fw_return_code == IWARP_QP_IN_ERROR_GOOD_CLOSE) ? 0 : -ECONNRESET; /* paired with READ_ONCE in destroy_qp */ smp_store_release(&ep->state, QED_IWARP_EP_CLOSED); spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); list_del(&ep->list_entry); spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); ep->event_cb(ep->cb_context, ¶ms); } static void qed_iwarp_exception_received(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep, int fw_ret_code) { struct qed_iwarp_cm_event_params params; bool event_cb = false; DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "EP(0x%x) fw_ret_code=%d\n", ep->cid, fw_ret_code); switch (fw_ret_code) { case IWARP_EXCEPTION_DETECTED_LLP_CLOSED: params.status = 0; params.event = QED_IWARP_EVENT_DISCONNECT; event_cb = true; break; case IWARP_EXCEPTION_DETECTED_LLP_RESET: params.status = -ECONNRESET; params.event = QED_IWARP_EVENT_DISCONNECT; event_cb = true; break; case IWARP_EXCEPTION_DETECTED_RQ_EMPTY: params.event = QED_IWARP_EVENT_RQ_EMPTY; event_cb = true; break; case IWARP_EXCEPTION_DETECTED_IRQ_FULL: params.event = QED_IWARP_EVENT_IRQ_FULL; event_cb = true; break; case IWARP_EXCEPTION_DETECTED_LLP_TIMEOUT: params.event = QED_IWARP_EVENT_LLP_TIMEOUT; event_cb = true; break; case IWARP_EXCEPTION_DETECTED_REMOTE_PROTECTION_ERROR: params.event = QED_IWARP_EVENT_REMOTE_PROTECTION_ERROR; event_cb = true; break; case IWARP_EXCEPTION_DETECTED_CQ_OVERFLOW: params.event = QED_IWARP_EVENT_CQ_OVERFLOW; event_cb = true; break; case IWARP_EXCEPTION_DETECTED_LOCAL_CATASTROPHIC: params.event = QED_IWARP_EVENT_QP_CATASTROPHIC; event_cb = true; break; case IWARP_EXCEPTION_DETECTED_LOCAL_ACCESS_ERROR: params.event = QED_IWARP_EVENT_LOCAL_ACCESS_ERROR; event_cb = true; break; case IWARP_EXCEPTION_DETECTED_REMOTE_OPERATION_ERROR: params.event = QED_IWARP_EVENT_REMOTE_OPERATION_ERROR; event_cb = true; break; case IWARP_EXCEPTION_DETECTED_TERMINATE_RECEIVED: params.event = QED_IWARP_EVENT_TERMINATE_RECEIVED; event_cb = true; break; default: DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Unhandled exception received...fw_ret_code=%d\n", fw_ret_code); break; } if (event_cb) { params.ep_context = ep; params.cm_info = &ep->cm_info; ep->event_cb(ep->cb_context, ¶ms); } } static void qed_iwarp_tcp_connect_unsuccessful(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep, u8 fw_return_code) { struct qed_iwarp_cm_event_params params; memset(¶ms, 0, sizeof(params)); params.event = QED_IWARP_EVENT_ACTIVE_COMPLETE; params.ep_context = ep; params.cm_info = &ep->cm_info; /* paired with READ_ONCE in destroy_qp */ smp_store_release(&ep->state, QED_IWARP_EP_CLOSED); switch (fw_return_code) { case IWARP_CONN_ERROR_TCP_CONNECT_INVALID_PACKET: DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "%s(0x%x) TCP connect got invalid packet\n", QED_IWARP_CONNECT_MODE_STRING(ep), ep->tcp_cid); params.status = -ECONNRESET; break; case IWARP_CONN_ERROR_TCP_CONNECTION_RST: DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "%s(0x%x) TCP Connection Reset\n", QED_IWARP_CONNECT_MODE_STRING(ep), ep->tcp_cid); params.status = -ECONNRESET; break; case IWARP_CONN_ERROR_TCP_CONNECT_TIMEOUT: DP_NOTICE(p_hwfn, "%s(0x%x) TCP timeout\n", QED_IWARP_CONNECT_MODE_STRING(ep), ep->tcp_cid); params.status = -EBUSY; break; case IWARP_CONN_ERROR_MPA_NOT_SUPPORTED_VER: DP_NOTICE(p_hwfn, "%s(0x%x) MPA not supported VER\n", QED_IWARP_CONNECT_MODE_STRING(ep), ep->tcp_cid); params.status = -ECONNREFUSED; break; case IWARP_CONN_ERROR_MPA_INVALID_PACKET: DP_NOTICE(p_hwfn, "%s(0x%x) MPA Invalid Packet\n", QED_IWARP_CONNECT_MODE_STRING(ep), ep->tcp_cid); params.status = -ECONNRESET; break; default: DP_ERR(p_hwfn, "%s(0x%x) Unexpected return code tcp connect: %d\n", QED_IWARP_CONNECT_MODE_STRING(ep), ep->tcp_cid, fw_return_code); params.status = -ECONNRESET; break; } if (ep->connect_mode == TCP_CONNECT_PASSIVE) { ep->tcp_cid = QED_IWARP_INVALID_TCP_CID; qed_iwarp_return_ep(p_hwfn, ep); } else { ep->event_cb(ep->cb_context, ¶ms); spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); list_del(&ep->list_entry); spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); } } static void qed_iwarp_connect_complete(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep, u8 fw_return_code) { u8 ll2_syn_handle = p_hwfn->p_rdma_info->iwarp.ll2_syn_handle; if (ep->connect_mode == TCP_CONNECT_PASSIVE) { /* Done with the SYN packet, post back to ll2 rx */ qed_iwarp_ll2_post_rx(p_hwfn, ep->syn, ll2_syn_handle); ep->syn = NULL; /* If connect failed - upper layer doesn't know about it */ if (fw_return_code == RDMA_RETURN_OK) qed_iwarp_mpa_received(p_hwfn, ep); else qed_iwarp_tcp_connect_unsuccessful(p_hwfn, ep, fw_return_code); } else { if (fw_return_code == RDMA_RETURN_OK) qed_iwarp_mpa_offload(p_hwfn, ep); else qed_iwarp_tcp_connect_unsuccessful(p_hwfn, ep, fw_return_code); } } static inline bool qed_iwarp_check_ep_ok(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep) { if (!ep || (ep->sig != QED_EP_SIG)) { DP_ERR(p_hwfn, "ERROR ON ASYNC ep=%p\n", ep); return false; } return true; } static int qed_iwarp_async_event(struct qed_hwfn *p_hwfn, u8 fw_event_code, u16 echo, union event_ring_data *data, u8 fw_return_code) { struct qed_rdma_events events = p_hwfn->p_rdma_info->events; struct regpair *fw_handle = &data->rdma_data.async_handle; struct qed_iwarp_ep *ep = NULL; u16 srq_offset; u16 srq_id; u16 cid; ep = (struct qed_iwarp_ep *)(uintptr_t)HILO_64(fw_handle->hi, fw_handle->lo); switch (fw_event_code) { case IWARP_EVENT_TYPE_ASYNC_CONNECT_COMPLETE: /* Async completion after TCP 3-way handshake */ if (!qed_iwarp_check_ep_ok(p_hwfn, ep)) return -EINVAL; DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "EP(0x%x) IWARP_EVENT_TYPE_ASYNC_CONNECT_COMPLETE fw_ret_code=%d\n", ep->tcp_cid, fw_return_code); qed_iwarp_connect_complete(p_hwfn, ep, fw_return_code); break; case IWARP_EVENT_TYPE_ASYNC_EXCEPTION_DETECTED: if (!qed_iwarp_check_ep_ok(p_hwfn, ep)) return -EINVAL; DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "QP(0x%x) IWARP_EVENT_TYPE_ASYNC_EXCEPTION_DETECTED fw_ret_code=%d\n", ep->cid, fw_return_code); qed_iwarp_exception_received(p_hwfn, ep, fw_return_code); break; case IWARP_EVENT_TYPE_ASYNC_QP_IN_ERROR_STATE: /* Async completion for Close Connection ramrod */ if (!qed_iwarp_check_ep_ok(p_hwfn, ep)) return -EINVAL; DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "QP(0x%x) IWARP_EVENT_TYPE_ASYNC_QP_IN_ERROR_STATE fw_ret_code=%d\n", ep->cid, fw_return_code); qed_iwarp_qp_in_error(p_hwfn, ep, fw_return_code); break; case IWARP_EVENT_TYPE_ASYNC_ENHANCED_MPA_REPLY_ARRIVED: /* Async event for active side only */ if (!qed_iwarp_check_ep_ok(p_hwfn, ep)) return -EINVAL; DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "QP(0x%x) IWARP_EVENT_TYPE_ASYNC_MPA_HANDSHAKE_MPA_REPLY_ARRIVED fw_ret_code=%d\n", ep->cid, fw_return_code); qed_iwarp_mpa_reply_arrived(p_hwfn, ep); break; case IWARP_EVENT_TYPE_ASYNC_MPA_HANDSHAKE_COMPLETE: if (!qed_iwarp_check_ep_ok(p_hwfn, ep)) return -EINVAL; DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "QP(0x%x) IWARP_EVENT_TYPE_ASYNC_MPA_HANDSHAKE_COMPLETE fw_ret_code=%d\n", ep->cid, fw_return_code); qed_iwarp_mpa_complete(p_hwfn, ep, fw_return_code); break; case IWARP_EVENT_TYPE_ASYNC_CID_CLEANED: cid = (u16)le32_to_cpu(fw_handle->lo); DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "(0x%x)IWARP_EVENT_TYPE_ASYNC_CID_CLEANED\n", cid); qed_iwarp_cid_cleaned(p_hwfn, cid); break; case IWARP_EVENT_TYPE_ASYNC_SRQ_EMPTY: DP_NOTICE(p_hwfn, "IWARP_EVENT_TYPE_ASYNC_SRQ_EMPTY\n"); srq_offset = p_hwfn->p_rdma_info->srq_id_offset; /* FW assigns value that is no greater than u16 */ srq_id = ((u16)le32_to_cpu(fw_handle->lo)) - srq_offset; events.affiliated_event(events.context, QED_IWARP_EVENT_SRQ_EMPTY, &srq_id); break; case IWARP_EVENT_TYPE_ASYNC_SRQ_LIMIT: DP_NOTICE(p_hwfn, "IWARP_EVENT_TYPE_ASYNC_SRQ_LIMIT\n"); srq_offset = p_hwfn->p_rdma_info->srq_id_offset; /* FW assigns value that is no greater than u16 */ srq_id = ((u16)le32_to_cpu(fw_handle->lo)) - srq_offset; events.affiliated_event(events.context, QED_IWARP_EVENT_SRQ_LIMIT, &srq_id); break; case IWARP_EVENT_TYPE_ASYNC_CQ_OVERFLOW: DP_NOTICE(p_hwfn, "IWARP_EVENT_TYPE_ASYNC_CQ_OVERFLOW\n"); p_hwfn->p_rdma_info->events.affiliated_event( p_hwfn->p_rdma_info->events.context, QED_IWARP_EVENT_CQ_OVERFLOW, (void *)fw_handle); break; default: DP_ERR(p_hwfn, "Received unexpected async iwarp event %d\n", fw_event_code); return -EINVAL; } return 0; } int qed_iwarp_create_listen(void *rdma_cxt, struct qed_iwarp_listen_in *iparams, struct qed_iwarp_listen_out *oparams) { struct qed_hwfn *p_hwfn = rdma_cxt; struct qed_iwarp_listener *listener; listener = kzalloc(sizeof(*listener), GFP_KERNEL); if (!listener) return -ENOMEM; listener->ip_version = iparams->ip_version; memcpy(listener->ip_addr, iparams->ip_addr, sizeof(listener->ip_addr)); listener->port = iparams->port; listener->vlan = iparams->vlan; listener->event_cb = iparams->event_cb; listener->cb_context = iparams->cb_context; listener->max_backlog = iparams->max_backlog; oparams->handle = listener; spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); list_add_tail(&listener->list_entry, &p_hwfn->p_rdma_info->iwarp.listen_list); spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "callback=%p handle=%p ip=%x:%x:%x:%x port=0x%x vlan=0x%x\n", listener->event_cb, listener, listener->ip_addr[0], listener->ip_addr[1], listener->ip_addr[2], listener->ip_addr[3], listener->port, listener->vlan); return 0; } int qed_iwarp_destroy_listen(void *rdma_cxt, void *handle) { struct qed_iwarp_listener *listener = handle; struct qed_hwfn *p_hwfn = rdma_cxt; DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "handle=%p\n", handle); spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); list_del(&listener->list_entry); spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock); kfree(listener); return 0; } int qed_iwarp_send_rtr(void *rdma_cxt, struct qed_iwarp_send_rtr_in *iparams) { struct qed_hwfn *p_hwfn = rdma_cxt; struct qed_sp_init_data init_data; struct qed_spq_entry *p_ent; struct qed_iwarp_ep *ep; struct qed_rdma_qp *qp; int rc; ep = iparams->ep_context; if (!ep) { DP_ERR(p_hwfn, "Ep Context receive in send_rtr is NULL\n"); return -EINVAL; } qp = ep->qp; DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "QP(0x%x) EP(0x%x)\n", qp->icid, ep->tcp_cid); memset(&init_data, 0, sizeof(init_data)); init_data.cid = qp->icid; init_data.opaque_fid = p_hwfn->hw_info.opaque_fid; init_data.comp_mode = QED_SPQ_MODE_CB; rc = qed_sp_init_request(p_hwfn, &p_ent, IWARP_RAMROD_CMD_ID_MPA_OFFLOAD_SEND_RTR, PROTOCOLID_IWARP, &init_data); if (rc) return rc; rc = qed_spq_post(p_hwfn, p_ent, NULL); DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "rc = 0x%x\n", rc); return rc; } void qed_iwarp_query_qp(struct qed_rdma_qp *qp, struct qed_rdma_query_qp_out_params *out_params) { out_params->state = qed_iwarp2roce_state(qp->iwarp_state); }