diff options
Diffstat (limited to 'drivers/net/ethernet/qlogic/qed/qed_dev.c')
-rw-r--r-- | drivers/net/ethernet/qlogic/qed/qed_dev.c | 1427 |
1 files changed, 1276 insertions, 151 deletions
diff --git a/drivers/net/ethernet/qlogic/qed/qed_dev.c b/drivers/net/ethernet/qlogic/qed/qed_dev.c index b7d100f6b..0e4f4a930 100644 --- a/drivers/net/ethernet/qlogic/qed/qed_dev.c +++ b/drivers/net/ethernet/qlogic/qed/qed_dev.c @@ -17,11 +17,13 @@ #include <linux/pci.h> #include <linux/slab.h> #include <linux/string.h> +#include <linux/vmalloc.h> #include <linux/etherdevice.h> #include <linux/qed/qed_chain.h> #include <linux/qed/qed_if.h> #include "qed.h" #include "qed_cxt.h" +#include "qed_dcbx.h" #include "qed_dev_api.h" #include "qed_hsi.h" #include "qed_hw.h" @@ -30,6 +32,11 @@ #include "qed_mcp.h" #include "qed_reg_addr.h" #include "qed_sp.h" +#include "qed_sriov.h" +#include "qed_vf.h" + +static spinlock_t qm_lock; +static bool qm_lock_init = false; /* API common to all protocols */ enum BAR_ID { @@ -40,10 +47,14 @@ enum BAR_ID { static u32 qed_hw_bar_size(struct qed_hwfn *p_hwfn, enum BAR_ID bar_id) { - u32 bar_reg = (bar_id == BAR_ID_0 ? - PGLUE_B_REG_PF_BAR0_SIZE : PGLUE_B_REG_PF_BAR1_SIZE); - u32 val = qed_rd(p_hwfn, p_hwfn->p_main_ptt, bar_reg); + u32 bar_reg = (bar_id == BAR_ID_0 ? + PGLUE_B_REG_PF_BAR0_SIZE : PGLUE_B_REG_PF_BAR1_SIZE); + u32 val; + + if (IS_VF(p_hwfn->cdev)) + return 1 << 17; + val = qed_rd(p_hwfn, p_hwfn->p_main_ptt, bar_reg); if (val) return 1 << (val + 15); @@ -105,12 +116,17 @@ static void qed_qm_info_free(struct qed_hwfn *p_hwfn) qm_info->qm_vport_params = NULL; kfree(qm_info->qm_port_params); qm_info->qm_port_params = NULL; + kfree(qm_info->wfq_data); + qm_info->wfq_data = NULL; } void qed_resc_free(struct qed_dev *cdev) { int i; + if (IS_VF(cdev)) + return; + kfree(cdev->fw_data); cdev->fw_data = NULL; @@ -134,22 +150,54 @@ void qed_resc_free(struct qed_dev *cdev) qed_eq_free(p_hwfn, p_hwfn->p_eq); qed_consq_free(p_hwfn, p_hwfn->p_consq); qed_int_free(p_hwfn); + qed_iov_free(p_hwfn); qed_dmae_info_free(p_hwfn); + qed_dcbx_info_free(p_hwfn, p_hwfn->p_dcbx_info); } } -static int qed_init_qm_info(struct qed_hwfn *p_hwfn) +static int qed_init_qm_info(struct qed_hwfn *p_hwfn, bool b_sleepable) { + u8 num_vports, vf_offset = 0, i, vport_id, num_ports, curr_queue = 0; struct qed_qm_info *qm_info = &p_hwfn->qm_info; struct init_qm_port_params *p_qm_port; - u8 num_vports, i, vport_id, num_ports; + bool init_rdma_offload_pq = false; + bool init_pure_ack_pq = false; + bool init_ooo_pq = false; u16 num_pqs, multi_cos_tcs = 1; - + u8 pf_wfq = qm_info->pf_wfq; + u32 pf_rl = qm_info->pf_rl; + u16 num_pf_rls = 0; + u16 num_vfs = 0; + +#ifdef CONFIG_QED_SRIOV + if (p_hwfn->cdev->p_iov_info) + num_vfs = p_hwfn->cdev->p_iov_info->total_vfs; +#endif memset(qm_info, 0, sizeof(*qm_info)); - num_pqs = multi_cos_tcs + 1; /* The '1' is for pure-LB */ + num_pqs = multi_cos_tcs + num_vfs + 1; /* The '1' is for pure-LB */ num_vports = (u8)RESC_NUM(p_hwfn, QED_VPORT); + if (p_hwfn->hw_info.personality == QED_PCI_ETH_ROCE) { + num_pqs++; /* for RoCE queue */ + init_rdma_offload_pq = true; + /* we subtract num_vfs because each require a rate limiter, + * and one default rate limiter + */ + if (p_hwfn->pf_params.rdma_pf_params.enable_dcqcn) + num_pf_rls = RESC_NUM(p_hwfn, QED_RL) - num_vfs - 1; + + num_pqs += num_pf_rls; + qm_info->num_pf_rls = (u8) num_pf_rls; + } + + if (p_hwfn->hw_info.personality == QED_PCI_ISCSI) { + num_pqs += 2; /* for iSCSI pure-ACK / OOO queue */ + init_pure_ack_pq = true; + init_ooo_pq = true; + } + /* Sanity checking that setup requires legal number of resources */ if (num_pqs > RESC_NUM(p_hwfn, QED_PQ)) { DP_ERR(p_hwfn, @@ -160,40 +208,107 @@ static int qed_init_qm_info(struct qed_hwfn *p_hwfn) /* PQs will be arranged as follows: First per-TC PQ then pure-LB quete. */ - qm_info->qm_pq_params = kzalloc(sizeof(*qm_info->qm_pq_params) * - num_pqs, GFP_KERNEL); + qm_info->qm_pq_params = kcalloc(num_pqs, + sizeof(struct init_qm_pq_params), + b_sleepable ? GFP_KERNEL : GFP_ATOMIC); if (!qm_info->qm_pq_params) goto alloc_err; - qm_info->qm_vport_params = kzalloc(sizeof(*qm_info->qm_vport_params) * - num_vports, GFP_KERNEL); + qm_info->qm_vport_params = kcalloc(num_vports, + sizeof(struct init_qm_vport_params), + b_sleepable ? GFP_KERNEL + : GFP_ATOMIC); if (!qm_info->qm_vport_params) goto alloc_err; - qm_info->qm_port_params = kzalloc(sizeof(*qm_info->qm_port_params) * - MAX_NUM_PORTS, GFP_KERNEL); + qm_info->qm_port_params = kcalloc(MAX_NUM_PORTS, + sizeof(struct init_qm_port_params), + b_sleepable ? GFP_KERNEL + : GFP_ATOMIC); if (!qm_info->qm_port_params) goto alloc_err; + qm_info->wfq_data = kcalloc(num_vports, sizeof(struct qed_wfq_data), + b_sleepable ? GFP_KERNEL : GFP_ATOMIC); + if (!qm_info->wfq_data) + goto alloc_err; + vport_id = (u8)RESC_START(p_hwfn, QED_VPORT); + /* First init rate limited queues */ + for (curr_queue = 0; curr_queue < num_pf_rls; curr_queue++) { + qm_info->qm_pq_params[curr_queue].vport_id = vport_id++; + qm_info->qm_pq_params[curr_queue].tc_id = + p_hwfn->hw_info.non_offload_tc; + qm_info->qm_pq_params[curr_queue].wrr_group = 1; + qm_info->qm_pq_params[curr_queue].rl_valid = 1; + } + /* First init per-TC PQs */ for (i = 0; i < multi_cos_tcs; i++) { - struct init_qm_pq_params *params = &qm_info->qm_pq_params[i]; - - params->vport_id = vport_id; - params->tc_id = p_hwfn->hw_info.non_offload_tc; - params->wrr_group = 1; + struct init_qm_pq_params *params = + &qm_info->qm_pq_params[curr_queue++]; + + if (p_hwfn->hw_info.personality == QED_PCI_ETH_ROCE || + p_hwfn->hw_info.personality == QED_PCI_ETH) { + params->vport_id = vport_id; + params->tc_id = p_hwfn->hw_info.non_offload_tc; + params->wrr_group = 1; + } else { + params->vport_id = vport_id; + params->tc_id = p_hwfn->hw_info.offload_tc; + params->wrr_group = 1; + } } /* Then init pure-LB PQ */ - qm_info->pure_lb_pq = i; - qm_info->qm_pq_params[i].vport_id = (u8)RESC_START(p_hwfn, QED_VPORT); - qm_info->qm_pq_params[i].tc_id = PURE_LB_TC; - qm_info->qm_pq_params[i].wrr_group = 1; - i++; + qm_info->pure_lb_pq = curr_queue; + qm_info->qm_pq_params[curr_queue].vport_id = + (u8) RESC_START(p_hwfn, QED_VPORT); + qm_info->qm_pq_params[curr_queue].tc_id = PURE_LB_TC; + qm_info->qm_pq_params[curr_queue].wrr_group = 1; + curr_queue++; qm_info->offload_pq = 0; + if (init_rdma_offload_pq) { + qm_info->offload_pq = curr_queue; + qm_info->qm_pq_params[curr_queue].vport_id = vport_id; + qm_info->qm_pq_params[curr_queue].tc_id = + p_hwfn->hw_info.offload_tc; + qm_info->qm_pq_params[curr_queue].wrr_group = 1; + curr_queue++; + } + + if (init_pure_ack_pq) { + qm_info->pure_ack_pq = curr_queue; + qm_info->qm_pq_params[curr_queue].vport_id = vport_id; + qm_info->qm_pq_params[curr_queue].tc_id = + p_hwfn->hw_info.offload_tc; + qm_info->qm_pq_params[curr_queue].wrr_group = 1; + curr_queue++; + } + + if (init_ooo_pq) { + qm_info->ooo_pq = curr_queue; + qm_info->qm_pq_params[curr_queue].vport_id = vport_id; + qm_info->qm_pq_params[curr_queue].tc_id = DCBX_ISCSI_OOO_TC; + qm_info->qm_pq_params[curr_queue].wrr_group = 1; + curr_queue++; + } + + /* Then init per-VF PQs */ + vf_offset = curr_queue; + for (i = 0; i < num_vfs; i++) { + /* First vport is used by the PF */ + qm_info->qm_pq_params[curr_queue].vport_id = vport_id + i + 1; + qm_info->qm_pq_params[curr_queue].tc_id = + p_hwfn->hw_info.non_offload_tc; + qm_info->qm_pq_params[curr_queue].wrr_group = 1; + qm_info->qm_pq_params[curr_queue].rl_valid = 1; + curr_queue++; + } + + qm_info->vf_queues_offset = vf_offset; qm_info->num_pqs = num_pqs; qm_info->num_vports = num_vports; @@ -202,7 +317,10 @@ static int qed_init_qm_info(struct qed_hwfn *p_hwfn) for (i = 0; i < num_ports; i++) { p_qm_port = &qm_info->qm_port_params[i]; p_qm_port->active = 1; - p_qm_port->num_active_phys_tcs = 4; + if (num_ports == 4) + p_qm_port->active_phys_tcs = 0x7; + else + p_qm_port->active_phys_tcs = 0x9f; p_qm_port->num_pbf_cmd_lines = PBF_MAX_CMD_LINES / num_ports; p_qm_port->num_btb_blocks = BTB_MAX_BLOCKS / num_ports; } @@ -211,29 +329,91 @@ static int qed_init_qm_info(struct qed_hwfn *p_hwfn) qm_info->start_pq = (u16)RESC_START(p_hwfn, QED_PQ); - qm_info->start_vport = (u8)RESC_START(p_hwfn, QED_VPORT); + qm_info->num_vf_pqs = num_vfs; + qm_info->start_vport = (u8) RESC_START(p_hwfn, QED_VPORT); + + for (i = 0; i < qm_info->num_vports; i++) + qm_info->qm_vport_params[i].vport_wfq = 1; - qm_info->pf_wfq = 0; - qm_info->pf_rl = 0; qm_info->vport_rl_en = 1; + qm_info->vport_wfq_en = 1; + qm_info->pf_rl = pf_rl; + qm_info->pf_wfq = pf_wfq; return 0; alloc_err: DP_NOTICE(p_hwfn, "Failed to allocate memory for QM params\n"); - kfree(qm_info->qm_pq_params); - kfree(qm_info->qm_vport_params); - kfree(qm_info->qm_port_params); - + qed_qm_info_free(p_hwfn); return -ENOMEM; } +/* This function reconfigures the QM pf on the fly. + * For this purpose we: + * 1. reconfigure the QM database + * 2. set new values to runtime arrat + * 3. send an sdm_qm_cmd through the rbc interface to stop the QM + * 4. activate init tool in QM_PF stage + * 5. send an sdm_qm_cmd through rbc interface to release the QM + */ +int qed_qm_reconf(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt) +{ + struct qed_qm_info *qm_info = &p_hwfn->qm_info; + bool b_rc; + int rc; + + /* qm_info is allocated in qed_init_qm_info() which is already called + * from qed_resc_alloc() or previous call of qed_qm_reconf(). + * The allocated size may change each init, so we free it before next + * allocation. + */ + qed_qm_info_free(p_hwfn); + + /* initialize qed's qm data structure */ + rc = qed_init_qm_info(p_hwfn, false); + if (rc) + return rc; + + /* stop PF's qm queues */ + spin_lock_bh(&qm_lock); + b_rc = qed_send_qm_stop_cmd(p_hwfn, p_ptt, false, true, + qm_info->start_pq, qm_info->num_pqs); + spin_unlock_bh(&qm_lock); + if (!b_rc) + return -EINVAL; + + /* clear the QM_PF runtime phase leftovers from previous init */ + qed_init_clear_rt_data(p_hwfn); + + /* prepare QM portion of runtime array */ + qed_qm_init_pf(p_hwfn); + + /* activate init tool on runtime array */ + rc = qed_init_run(p_hwfn, p_ptt, PHASE_QM_PF, p_hwfn->rel_pf_id, + p_hwfn->hw_info.hw_mode); + if (rc) + return rc; + + /* start PF's qm queues */ + spin_lock_bh(&qm_lock); + b_rc = qed_send_qm_stop_cmd(p_hwfn, p_ptt, true, true, + qm_info->start_pq, qm_info->num_pqs); + spin_unlock_bh(&qm_lock); + if (!b_rc) + return -EINVAL; + + return 0; +} + int qed_resc_alloc(struct qed_dev *cdev) { struct qed_consq *p_consq; struct qed_eq *p_eq; int i, rc = 0; + if (IS_VF(cdev)) + return rc; + cdev->fw_data = kzalloc(sizeof(*cdev->fw_data), GFP_KERNEL); if (!cdev->fw_data) return -ENOMEM; @@ -250,21 +430,20 @@ int qed_resc_alloc(struct qed_dev *cdev) if (!p_hwfn->p_tx_cids) { DP_NOTICE(p_hwfn, "Failed to allocate memory for Tx Cids\n"); - rc = -ENOMEM; - goto alloc_err; + goto alloc_no_mem; } p_hwfn->p_rx_cids = kzalloc(rx_size, GFP_KERNEL); if (!p_hwfn->p_rx_cids) { DP_NOTICE(p_hwfn, "Failed to allocate memory for Rx Cids\n"); - rc = -ENOMEM; - goto alloc_err; + goto alloc_no_mem; } } for_each_hwfn(cdev, i) { struct qed_hwfn *p_hwfn = &cdev->hwfns[i]; + u32 n_eqes, num_cons; /* First allocate the context manager structure */ rc = qed_cxt_mngr_alloc(p_hwfn); @@ -279,7 +458,7 @@ int qed_resc_alloc(struct qed_dev *cdev) goto alloc_err; /* Prepare and process QM requirements */ - rc = qed_init_qm_info(p_hwfn); + rc = qed_init_qm_info(p_hwfn, true); if (rc) goto alloc_err; @@ -308,19 +487,40 @@ int qed_resc_alloc(struct qed_dev *cdev) if (rc) goto alloc_err; + rc = qed_iov_alloc(p_hwfn); + if (rc) + goto alloc_err; + /* EQ */ - p_eq = qed_eq_alloc(p_hwfn, 256); - if (!p_eq) { - rc = -ENOMEM; + n_eqes = qed_chain_get_capacity(&p_hwfn->p_spq->chain); + if (p_hwfn->hw_info.personality == QED_PCI_ETH_ROCE) { + num_cons = qed_cxt_get_proto_cid_count(p_hwfn, + PROTOCOLID_ROCE, + 0) * 2; + n_eqes += num_cons + 2 * MAX_NUM_VFS_BB; + } else if (p_hwfn->hw_info.personality == QED_PCI_ISCSI) { + num_cons = + qed_cxt_get_proto_cid_count(p_hwfn, + PROTOCOLID_ISCSI, 0); + n_eqes += 2 * num_cons; + } + + if (n_eqes > 0xFFFF) { + DP_ERR(p_hwfn, + "Cannot allocate 0x%x EQ elements. The maximum of a u16 chain is 0x%x\n", + n_eqes, 0xFFFF); + rc = -EINVAL; goto alloc_err; } + + p_eq = qed_eq_alloc(p_hwfn, (u16) n_eqes); + if (!p_eq) + goto alloc_no_mem; p_hwfn->p_eq = p_eq; p_consq = qed_consq_alloc(p_hwfn); - if (!p_consq) { - rc = -ENOMEM; - goto alloc_err; - } + if (!p_consq) + goto alloc_no_mem; p_hwfn->p_consq = p_consq; /* DMA info initialization */ @@ -330,6 +530,14 @@ int qed_resc_alloc(struct qed_dev *cdev) "Failed to allocate memory for dmae_info structure\n"); goto alloc_err; } + + /* DCBX initialization */ + rc = qed_dcbx_info_alloc(p_hwfn); + if (rc) { + DP_NOTICE(p_hwfn, + "Failed to allocate memory for dcbx structure\n"); + goto alloc_err; + } } cdev->reset_stats = kzalloc(sizeof(*cdev->reset_stats), GFP_KERNEL); @@ -341,6 +549,8 @@ int qed_resc_alloc(struct qed_dev *cdev) return 0; +alloc_no_mem: + rc = -ENOMEM; alloc_err: qed_resc_free(cdev); return rc; @@ -350,6 +560,9 @@ void qed_resc_setup(struct qed_dev *cdev) { int i; + if (IS_VF(cdev)) + return; + for_each_hwfn(cdev, i) { struct qed_hwfn *p_hwfn = &cdev->hwfns[i]; @@ -365,14 +578,15 @@ void qed_resc_setup(struct qed_dev *cdev) p_hwfn->mcp_info->mfw_mb_length); qed_int_setup(p_hwfn, p_hwfn->p_main_ptt); + + qed_iov_setup(p_hwfn, p_hwfn->p_main_ptt); } } #define FINAL_CLEANUP_POLL_CNT (100) #define FINAL_CLEANUP_POLL_TIME (10) int qed_final_cleanup(struct qed_hwfn *p_hwfn, - struct qed_ptt *p_ptt, - u16 id) + struct qed_ptt *p_ptt, u16 id, bool is_vf) { u32 command = 0, addr, count = FINAL_CLEANUP_POLL_CNT; int rc = -EBUSY; @@ -380,6 +594,9 @@ int qed_final_cleanup(struct qed_hwfn *p_hwfn, addr = GTT_BAR0_MAP_REG_USDM_RAM + USTORM_FLR_FINAL_ACK_OFFSET(p_hwfn->rel_pf_id); + if (is_vf) + id += 0x10; + command |= X_FINAL_CLEANUP_AGG_INT << SDM_AGG_INT_COMP_PARAMS_AGG_INT_INDEX_SHIFT; command |= 1 << SDM_AGG_INT_COMP_PARAMS_AGG_VECTOR_ENABLE_SHIFT; @@ -453,7 +670,14 @@ static void qed_calc_hw_mode(struct qed_hwfn *p_hwfn) hw_mode |= 1 << MODE_ASIC; + if (p_hwfn->cdev->num_hwfns > 1) + hw_mode |= 1 << MODE_100G; + p_hwfn->hw_info.hw_mode = hw_mode; + + DP_VERBOSE(p_hwfn, (NETIF_MSG_PROBE | NETIF_MSG_IFUP), + "Configuring function for hw_mode: 0x%08x\n", + p_hwfn->hw_info.hw_mode); } /* Init run time data for all PFs on an engine. */ @@ -492,7 +716,10 @@ static int qed_hw_init_common(struct qed_hwfn *p_hwfn, struct qed_qm_info *qm_info = &p_hwfn->qm_info; struct qed_qm_common_rt_init_params params; struct qed_dev *cdev = p_hwfn->cdev; + u16 num_pfs, pf_id; + u32 concrete_fid; int rc = 0; + u8 vf_id; qed_init_cau_rt_data(cdev); @@ -538,9 +765,24 @@ static int qed_hw_init_common(struct qed_hwfn *p_hwfn, qed_wr(p_hwfn, p_ptt, PSWRQ2_REG_L2P_VALIDATE_VFID, 0); qed_wr(p_hwfn, p_ptt, PGLUE_B_REG_USE_CLIENTID_IN_TAG, 1); - /* Disable relaxed ordering in the PCI config space */ - qed_wr(p_hwfn, p_ptt, 0x20b4, - qed_rd(p_hwfn, p_ptt, 0x20b4) & ~0x10); + if (QED_IS_BB(p_hwfn->cdev)) { + num_pfs = NUM_OF_ENG_PFS(p_hwfn->cdev); + for (pf_id = 0; pf_id < num_pfs; pf_id++) { + qed_fid_pretend(p_hwfn, p_ptt, pf_id); + qed_wr(p_hwfn, p_ptt, PRS_REG_SEARCH_ROCE, 0x0); + qed_wr(p_hwfn, p_ptt, PRS_REG_SEARCH_TCP, 0x0); + } + /* pretend to original PF */ + qed_fid_pretend(p_hwfn, p_ptt, p_hwfn->rel_pf_id); + } + + for (vf_id = 0; vf_id < MAX_NUM_VFS_BB; vf_id++) { + concrete_fid = qed_vfid_to_concrete(p_hwfn, vf_id); + qed_fid_pretend(p_hwfn, p_ptt, (u16) concrete_fid); + qed_wr(p_hwfn, p_ptt, CCFC_REG_STRONG_ENABLE_VF, 0x1); + } + /* pretend to original PF */ + qed_fid_pretend(p_hwfn, p_ptt, p_hwfn->rel_pf_id); return rc; } @@ -551,13 +793,37 @@ static int qed_hw_init_port(struct qed_hwfn *p_hwfn, { int rc = 0; - rc = qed_init_run(p_hwfn, p_ptt, PHASE_PORT, p_hwfn->port_id, - hw_mode); + rc = qed_init_run(p_hwfn, p_ptt, PHASE_PORT, p_hwfn->port_id, hw_mode); + if (rc != 0) + return rc; + + if (hw_mode & (1 << MODE_MF_SI)) { + u8 pf_id = 0; + + if (!qed_hw_init_first_eth(p_hwfn, p_ptt, &pf_id)) { + DP_VERBOSE(p_hwfn, NETIF_MSG_IFUP, + "PF[%08x] is first eth on engine\n", pf_id); + + /* We should have configured BIT for ppfid, i.e., the + * relative function number in the port. But there's a + * bug in LLH in BB where the ppfid is actually engine + * based, so we need to take this into account. + */ + qed_wr(p_hwfn, p_ptt, + NIG_REG_LLH_TAGMAC_DEF_PF_VECTOR, 1 << pf_id); + } + + /* Take the protocol-based hit vector if there is a hit, + * otherwise take the other vector. + */ + qed_wr(p_hwfn, p_ptt, NIG_REG_LLH_CLS_TYPE_DUALMODE, 0x2); + } return rc; } static int qed_hw_init_pf(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt, + struct qed_tunn_start_params *p_tunn, int hw_mode, bool b_hw_start, enum qed_int_mode int_mode, @@ -574,7 +840,7 @@ static int qed_hw_init_pf(struct qed_hwfn *p_hwfn, p_hwfn->qm_info.pf_wfq = p_info->bandwidth_min; /* Update rate limit once we'll actually have a link */ - p_hwfn->qm_info.pf_rl = 100; + p_hwfn->qm_info.pf_rl = 100000; } qed_cxt_hw_init_pf(p_hwfn); @@ -598,12 +864,13 @@ static int qed_hw_init_pf(struct qed_hwfn *p_hwfn, } /* Protocl Configuration */ - STORE_RT_REG(p_hwfn, PRS_REG_SEARCH_TCP_RT_OFFSET, 0); + STORE_RT_REG(p_hwfn, PRS_REG_SEARCH_TCP_RT_OFFSET, + (p_hwfn->hw_info.personality == QED_PCI_ISCSI) ? 1 : 0); STORE_RT_REG(p_hwfn, PRS_REG_SEARCH_FCOE_RT_OFFSET, 0); STORE_RT_REG(p_hwfn, PRS_REG_SEARCH_ROCE_RT_OFFSET, 0); /* Cleanup chip from previous driver if such remains exist */ - rc = qed_final_cleanup(p_hwfn, p_ptt, rel_pf_id); + rc = qed_final_cleanup(p_hwfn, p_ptt, rel_pf_id, false); if (rc != 0) return rc; @@ -620,12 +887,28 @@ static int qed_hw_init_pf(struct qed_hwfn *p_hwfn, /* Pure runtime initializations - directly to the HW */ qed_int_igu_init_pure_rt(p_hwfn, p_ptt, true, true); + if (hw_mode & (1 << MODE_MF_SI)) { + u8 pf_id = 0; + u32 val = 0; + + if (!qed_hw_init_first_eth(p_hwfn, p_ptt, &pf_id)) { + if (p_hwfn->rel_pf_id == pf_id) { + DP_VERBOSE(p_hwfn, NETIF_MSG_IFUP, + "PF[%d] is first ETH on engine\n", + pf_id); + val = 1; + } + qed_wr(p_hwfn, p_ptt, PRS_REG_MSG_INFO, val); + } + } + if (b_hw_start) { /* enable interrupts */ qed_int_igu_enable(p_hwfn, p_ptt, int_mode); /* send function start command */ - rc = qed_sp_pf_start(p_hwfn, p_hwfn->cdev->mf_mode); + rc = qed_sp_pf_start(p_hwfn, p_tunn, p_hwfn->cdev->mf_mode, + allow_npar_tx_switch); if (rc) DP_NOTICE(p_hwfn, "Function start ramrod failed\n"); } @@ -672,6 +955,7 @@ static void qed_reset_mb_shadow(struct qed_hwfn *p_hwfn, } int qed_hw_init(struct qed_dev *cdev, + struct qed_tunn_start_params *p_tunn, bool b_hw_start, enum qed_int_mode int_mode, bool allow_npar_tx_switch, @@ -680,13 +964,25 @@ int qed_hw_init(struct qed_dev *cdev, u32 load_code, param; int rc, mfw_rc, i; - rc = qed_init_fw_data(cdev, bin_fw_data); - if (rc != 0) - return rc; + if ((int_mode == QED_INT_MODE_MSI) && (cdev->num_hwfns > 1)) { + DP_NOTICE(cdev, "MSI mode is not supported for CMT devices\n"); + return -EINVAL; + } + + if (IS_PF(cdev)) { + rc = qed_init_fw_data(cdev, bin_fw_data); + if (rc != 0) + return rc; + } for_each_hwfn(cdev, i) { struct qed_hwfn *p_hwfn = &cdev->hwfns[i]; + if (IS_VF(cdev)) { + p_hwfn->b_int_enabled = 1; + continue; + } + /* Enable DMAE in PXP */ rc = qed_change_pci_hwfn(p_hwfn, p_hwfn->p_main_ptt, true); @@ -708,6 +1004,11 @@ int qed_hw_init(struct qed_dev *cdev, p_hwfn->first_on_engine = (load_code == FW_MSG_CODE_DRV_LOAD_ENGINE); + if (!qm_lock_init) { + spin_lock_init(&qm_lock); + qm_lock_init = true; + } + switch (load_code) { case FW_MSG_CODE_DRV_LOAD_ENGINE: rc = qed_hw_init_common(p_hwfn, p_hwfn->p_main_ptt, @@ -724,7 +1025,7 @@ int qed_hw_init(struct qed_dev *cdev, /* Fall into */ case FW_MSG_CODE_DRV_LOAD_FUNCTION: rc = qed_hw_init_pf(p_hwfn, p_hwfn->p_main_ptt, - p_hwfn->hw_info.hw_mode, + p_tunn, p_hwfn->hw_info.hw_mode, b_hw_start, int_mode, allow_npar_tx_switch); break; @@ -749,6 +1050,20 @@ int qed_hw_init(struct qed_dev *cdev, return mfw_rc; } + /* send DCBX attention request command */ + DP_VERBOSE(p_hwfn, + QED_MSG_DCB, + "sending phony dcbx set command to trigger DCBx attention handling\n"); + mfw_rc = qed_mcp_cmd(p_hwfn, p_hwfn->p_main_ptt, + DRV_MSG_CODE_SET_DCBX, + 1 << DRV_MB_PARAM_DCBX_NOTIFY_SHIFT, + &load_code, ¶m); + if (mfw_rc) { + DP_NOTICE(p_hwfn, + "Failed to send DCBX attention request\n"); + return mfw_rc; + } + p_hwfn->hw_init_done = true; } @@ -811,6 +1126,11 @@ int qed_hw_stop(struct qed_dev *cdev) DP_VERBOSE(p_hwfn, NETIF_MSG_IFDOWN, "Stopping hw/fw\n"); + if (IS_VF(cdev)) { + qed_vf_pf_int_cleanup(p_hwfn); + continue; + } + /* mark the hw as uninitialized... */ p_hwfn->hw_init_done = false; @@ -842,15 +1162,16 @@ int qed_hw_stop(struct qed_dev *cdev) usleep_range(1000, 2000); } - /* Disable DMAE in PXP - in CMT, this should only be done for - * first hw-function, and only after all transactions have - * stopped for all active hw-functions. - */ - t_rc = qed_change_pci_hwfn(&cdev->hwfns[0], - cdev->hwfns[0].p_main_ptt, - false); - if (t_rc != 0) - rc = t_rc; + if (IS_PF(cdev)) { + /* Disable DMAE in PXP - in CMT, this should only be done for + * first hw-function, and only after all transactions have + * stopped for all active hw-functions. + */ + t_rc = qed_change_pci_hwfn(&cdev->hwfns[0], + cdev->hwfns[0].p_main_ptt, false); + if (t_rc != 0) + rc = t_rc; + } return rc; } @@ -861,7 +1182,12 @@ void qed_hw_stop_fastpath(struct qed_dev *cdev) for_each_hwfn(cdev, j) { struct qed_hwfn *p_hwfn = &cdev->hwfns[j]; - struct qed_ptt *p_ptt = p_hwfn->p_main_ptt; + struct qed_ptt *p_ptt = p_hwfn->p_main_ptt; + + if (IS_VF(cdev)) { + qed_vf_pf_int_cleanup(p_hwfn); + continue; + } DP_VERBOSE(p_hwfn, NETIF_MSG_IFDOWN, @@ -885,6 +1211,9 @@ void qed_hw_stop_fastpath(struct qed_dev *cdev) void qed_hw_start_fastpath(struct qed_hwfn *p_hwfn) { + if (IS_VF(p_hwfn->cdev)) + return; + /* Re-open incoming traffic */ qed_wr(p_hwfn, p_hwfn->p_main_ptt, NIG_REG_RX_LLH_BRB_GATE_DNTFWD_PERPF, 0x0); @@ -914,6 +1243,13 @@ int qed_hw_reset(struct qed_dev *cdev) for_each_hwfn(cdev, i) { struct qed_hwfn *p_hwfn = &cdev->hwfns[i]; + if (IS_VF(cdev)) { + rc = qed_vf_pf_reset(p_hwfn); + if (rc) + return rc; + continue; + } + DP_VERBOSE(p_hwfn, NETIF_MSG_IFDOWN, "Resetting hw/fw\n"); /* Check for incorrect states */ @@ -1006,16 +1342,23 @@ static void qed_hw_set_feat(struct qed_hwfn *p_hwfn) num_features); } -static void qed_hw_get_resc(struct qed_hwfn *p_hwfn) +static int qed_hw_get_resc(struct qed_hwfn *p_hwfn) { + u8 enabled_func_idx = p_hwfn->enabled_func_idx; u32 *resc_start = p_hwfn->hw_info.resc_start; + u8 num_funcs = p_hwfn->num_funcs_on_engine; u32 *resc_num = p_hwfn->hw_info.resc_num; struct qed_sb_cnt_info sb_cnt_info; - int num_funcs, i; - - num_funcs = MAX_NUM_PFS_BB; + int i, max_vf_vlan_filters; memset(&sb_cnt_info, 0, sizeof(sb_cnt_info)); + +#ifdef CONFIG_QED_SRIOV + max_vf_vlan_filters = QED_ETH_MAX_VF_NUM_VLAN_FILTERS; +#else + max_vf_vlan_filters = 0; +#endif + qed_int_get_num_sbs(p_hwfn, &sb_cnt_info); resc_num[QED_SB] = min_t(u32, @@ -1025,14 +1368,22 @@ static void qed_hw_get_resc(struct qed_hwfn *p_hwfn) resc_num[QED_VPORT] = MAX_NUM_VPORTS_BB / num_funcs; resc_num[QED_RSS_ENG] = ETH_RSS_ENGINE_NUM_BB / num_funcs; resc_num[QED_PQ] = MAX_QM_TX_QUEUES_BB / num_funcs; - resc_num[QED_RL] = 8; + resc_num[QED_RL] = min_t(u32, 64, resc_num[QED_VPORT]); resc_num[QED_MAC] = ETH_NUM_MAC_FILTERS / num_funcs; resc_num[QED_VLAN] = (ETH_NUM_VLAN_FILTERS - 1 /*For vlan0*/) / num_funcs; - resc_num[QED_ILT] = 950; + resc_num[QED_ILT] = PXP_NUM_ILT_RECORDS_BB / num_funcs; for (i = 0; i < QED_MAX_RESC; i++) - resc_start[i] = resc_num[i] * p_hwfn->rel_pf_id; + resc_start[i] = resc_num[i] * enabled_func_idx; + + /* Sanity for ILT */ + if (RESC_END(p_hwfn, QED_ILT) > PXP_NUM_ILT_RECORDS_BB) { + DP_NOTICE(p_hwfn, "Can't assign ILT pages [%08x,...,%08x]\n", + RESC_START(p_hwfn, QED_ILT), + RESC_END(p_hwfn, QED_ILT) - 1); + return -EINVAL; + } qed_hw_set_feat(p_hwfn); @@ -1062,6 +1413,8 @@ static void qed_hw_get_resc(struct qed_hwfn *p_hwfn) p_hwfn->hw_info.resc_start[QED_VLAN], p_hwfn->hw_info.resc_num[QED_ILT], p_hwfn->hw_info.resc_start[QED_ILT]); + + return 0; } static int qed_hw_get_nvm_info(struct qed_hwfn *p_hwfn, @@ -1091,31 +1444,31 @@ static int qed_hw_get_nvm_info(struct qed_hwfn *p_hwfn, switch ((core_cfg & NVM_CFG1_GLOB_NETWORK_PORT_MODE_MASK) >> NVM_CFG1_GLOB_NETWORK_PORT_MODE_OFFSET) { - case NVM_CFG1_GLOB_NETWORK_PORT_MODE_DE_2X40G: + case NVM_CFG1_GLOB_NETWORK_PORT_MODE_BB_2X40G: p_hwfn->hw_info.port_mode = QED_PORT_MODE_DE_2X40G; break; - case NVM_CFG1_GLOB_NETWORK_PORT_MODE_DE_2X50G: + case NVM_CFG1_GLOB_NETWORK_PORT_MODE_2X50G: p_hwfn->hw_info.port_mode = QED_PORT_MODE_DE_2X50G; break; - case NVM_CFG1_GLOB_NETWORK_PORT_MODE_DE_1X100G: + case NVM_CFG1_GLOB_NETWORK_PORT_MODE_BB_1X100G: p_hwfn->hw_info.port_mode = QED_PORT_MODE_DE_1X100G; break; - case NVM_CFG1_GLOB_NETWORK_PORT_MODE_DE_4X10G_F: + case NVM_CFG1_GLOB_NETWORK_PORT_MODE_4X10G_F: p_hwfn->hw_info.port_mode = QED_PORT_MODE_DE_4X10G_F; break; - case NVM_CFG1_GLOB_NETWORK_PORT_MODE_DE_4X10G_E: + case NVM_CFG1_GLOB_NETWORK_PORT_MODE_BB_4X10G_E: p_hwfn->hw_info.port_mode = QED_PORT_MODE_DE_4X10G_E; break; - case NVM_CFG1_GLOB_NETWORK_PORT_MODE_DE_4X20G: + case NVM_CFG1_GLOB_NETWORK_PORT_MODE_BB_4X20G: p_hwfn->hw_info.port_mode = QED_PORT_MODE_DE_4X20G; break; - case NVM_CFG1_GLOB_NETWORK_PORT_MODE_DE_1X40G: + case NVM_CFG1_GLOB_NETWORK_PORT_MODE_1X40G: p_hwfn->hw_info.port_mode = QED_PORT_MODE_DE_1X40G; break; - case NVM_CFG1_GLOB_NETWORK_PORT_MODE_DE_2X25G: + case NVM_CFG1_GLOB_NETWORK_PORT_MODE_2X25G: p_hwfn->hw_info.port_mode = QED_PORT_MODE_DE_2X25G; break; - case NVM_CFG1_GLOB_NETWORK_PORT_MODE_DE_1X25G: + case NVM_CFG1_GLOB_NETWORK_PORT_MODE_1X25G: p_hwfn->hw_info.port_mode = QED_PORT_MODE_DE_1X25G; break; default: @@ -1160,7 +1513,7 @@ static int qed_hw_get_nvm_info(struct qed_hwfn *p_hwfn, case NVM_CFG1_PORT_DRV_LINK_SPEED_50G: link->speed.forced_speed = 50000; break; - case NVM_CFG1_PORT_DRV_LINK_SPEED_100G: + case NVM_CFG1_PORT_DRV_LINK_SPEED_BB_100G: link->speed.forced_speed = 100000; break; default: @@ -1216,10 +1569,71 @@ static int qed_hw_get_nvm_info(struct qed_hwfn *p_hwfn, if (device_capabilities & NVM_CFG1_GLOB_DEVICE_CAPABILITIES_ETHERNET) __set_bit(QED_DEV_CAP_ETH, &p_hwfn->hw_info.device_capabilities); + if (device_capabilities & NVM_CFG1_GLOB_DEVICE_CAPABILITIES_ISCSI) + __set_bit(QED_DEV_CAP_ISCSI, + &p_hwfn->hw_info.device_capabilities); + if (device_capabilities & NVM_CFG1_GLOB_DEVICE_CAPABILITIES_ROCE) + __set_bit(QED_DEV_CAP_ROCE, + &p_hwfn->hw_info.device_capabilities); return qed_mcp_fill_shmem_func_info(p_hwfn, p_ptt); } +static void qed_get_num_funcs(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt) +{ + u8 num_funcs, enabled_func_idx = p_hwfn->rel_pf_id; + u32 reg_function_hide, tmp, eng_mask, low_pfs_mask; + + num_funcs = MAX_NUM_PFS_BB; + + /* Bit 0 of MISCS_REG_FUNCTION_HIDE indicates whether the bypass values + * in the other bits are selected. + * Bits 1-15 are for functions 1-15, respectively, and their value is + * '0' only for enabled functions (function 0 always exists and + * enabled). + * In case of CMT, only the "even" functions are enabled, and thus the + * number of functions for both hwfns is learnt from the same bits. + */ + reg_function_hide = qed_rd(p_hwfn, p_ptt, MISCS_REG_FUNCTION_HIDE); + + if (reg_function_hide & 0x1) { + if (QED_PATH_ID(p_hwfn) && p_hwfn->cdev->num_hwfns == 1) { + num_funcs = 0; + eng_mask = 0xaaaa; + } else { + num_funcs = 1; + eng_mask = 0x5554; + } + + /* Get the number of the enabled functions on the engine */ + tmp = (reg_function_hide ^ 0xffffffff) & eng_mask; + while (tmp) { + if (tmp & 0x1) + num_funcs++; + tmp >>= 0x1; + } + + /* Get the PF index within the enabled functions */ + low_pfs_mask = (0x1 << p_hwfn->abs_pf_id) - 1; + tmp = reg_function_hide & eng_mask & low_pfs_mask; + while (tmp) { + if (tmp & 0x1) + enabled_func_idx--; + tmp >>= 0x1; + } + } + + p_hwfn->num_funcs_on_engine = num_funcs; + p_hwfn->enabled_func_idx = enabled_func_idx; + + DP_VERBOSE(p_hwfn, + NETIF_MSG_PROBE, + "PF [rel_id %d, abs_id %d] within the %d enabled functions on the engine\n", + p_hwfn->rel_pf_id, + p_hwfn->abs_pf_id, + p_hwfn->num_funcs_on_engine); +} + static int qed_get_hw_info(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt, @@ -1228,6 +1642,13 @@ qed_get_hw_info(struct qed_hwfn *p_hwfn, u32 port_mode; int rc; + /* Since all information is common, only first hwfns should do this */ + if (IS_LEAD_HWFN(p_hwfn)) { + rc = qed_iov_hw_info(p_hwfn); + if (rc) + return rc; + } + /* Read the port mode */ port_mode = qed_rd(p_hwfn, p_ptt, CNIG_REG_NW_PORT_MODE_BB_B0); @@ -1271,9 +1692,9 @@ qed_get_hw_info(struct qed_hwfn *p_hwfn, p_hwfn->hw_info.personality = protocol; } - qed_hw_get_resc(p_hwfn); + qed_get_num_funcs(p_hwfn, p_ptt); - return rc; + return qed_hw_get_resc(p_hwfn); } static int qed_get_dev_info(struct qed_dev *cdev) @@ -1336,6 +1757,9 @@ static int qed_hw_prepare_single(struct qed_hwfn *p_hwfn, p_hwfn->regview = p_regview; p_hwfn->doorbells = p_doorbells; + if (IS_VF(p_hwfn->cdev)) + return qed_vf_hw_prepare(p_hwfn); + /* Validate that chip access is feasible */ if (REG_RD(p_hwfn, PXP_PF_ME_OPAQUE_ADDR) == 0xffffffff) { DP_ERR(p_hwfn, @@ -1387,6 +1811,8 @@ static int qed_hw_prepare_single(struct qed_hwfn *p_hwfn, return rc; err2: + if (IS_LEAD_HWFN(p_hwfn)) + qed_iov_free_hw_info(p_hwfn->cdev); qed_mcp_free(p_hwfn); err1: qed_hw_hwfn_free(p_hwfn); @@ -1401,7 +1827,8 @@ int qed_hw_prepare(struct qed_dev *cdev, int rc; /* Store the precompiled init data ptrs */ - qed_init_iro_array(cdev); + if (IS_PF(cdev)) + qed_init_iro_array(cdev); /* Initialize the first hwfn - will learn number of hwfns */ rc = qed_hw_prepare_single(p_hwfn, @@ -1433,9 +1860,11 @@ int qed_hw_prepare(struct qed_dev *cdev, * initiliazed hwfn 0. */ if (rc) { - qed_init_free(p_hwfn); - qed_mcp_free(p_hwfn); - qed_hw_hwfn_free(p_hwfn); + if (IS_PF(cdev)) { + qed_init_free(p_hwfn); + qed_mcp_free(p_hwfn); + qed_hw_hwfn_free(p_hwfn); + } } } @@ -1449,98 +1878,298 @@ void qed_hw_remove(struct qed_dev *cdev) for_each_hwfn(cdev, i) { struct qed_hwfn *p_hwfn = &cdev->hwfns[i]; + if (IS_VF(cdev)) { + qed_vf_pf_release(p_hwfn); + continue; + } + qed_init_free(p_hwfn); qed_hw_hwfn_free(p_hwfn); qed_mcp_free(p_hwfn); } + + qed_iov_free_hw_info(cdev); } -int qed_chain_alloc(struct qed_dev *cdev, - enum qed_chain_use_mode intended_use, - enum qed_chain_mode mode, - u16 num_elems, - size_t elem_size, - struct qed_chain *p_chain) +static void qed_chain_free_next_ptr(struct qed_dev *cdev, + struct qed_chain *p_chain) { - dma_addr_t p_pbl_phys = 0; - void *p_pbl_virt = NULL; + void *p_virt = p_chain->p_virt_addr, *p_virt_next = NULL; + dma_addr_t p_phys = p_chain->p_phys_addr, p_phys_next = 0; + struct qed_chain_next *p_next; + u32 size, i; + + if (!p_virt) + return; + + size = p_chain->elem_size * p_chain->usable_per_page; + + for (i = 0; i < p_chain->page_cnt; i++) { + if (!p_virt) + break; + + p_next = (struct qed_chain_next *)((u8 *)p_virt + size); + p_virt_next = p_next->next_virt; + p_phys_next = HILO_DMA_REGPAIR(p_next->next_phys); + + dma_free_coherent(&cdev->pdev->dev, + QED_CHAIN_PAGE_SIZE, p_virt, p_phys); + + p_virt = p_virt_next; + p_phys = p_phys_next; + } +} + +static void qed_chain_free_single(struct qed_dev *cdev, + struct qed_chain *p_chain) +{ + if (!p_chain->p_virt_addr) + return; + + dma_free_coherent(&cdev->pdev->dev, + QED_CHAIN_PAGE_SIZE, + p_chain->p_virt_addr, p_chain->p_phys_addr); +} + +static void qed_chain_free_pbl(struct qed_dev *cdev, struct qed_chain *p_chain) +{ + void **pp_virt_addr_tbl = p_chain->pbl.pp_virt_addr_tbl; + u32 page_cnt = p_chain->page_cnt, i, pbl_size; + u8 *p_pbl_virt = p_chain->pbl.p_virt_table; + + if (!pp_virt_addr_tbl) + return; + + if (!p_chain->pbl.p_virt_table) + goto out; + + for (i = 0; i < page_cnt; i++) { + if (!pp_virt_addr_tbl[i]) + break; + + dma_free_coherent(&cdev->pdev->dev, + QED_CHAIN_PAGE_SIZE, + pp_virt_addr_tbl[i], + *(dma_addr_t *)p_pbl_virt); + + p_pbl_virt += QED_CHAIN_PBL_ENTRY_SIZE; + } + + pbl_size = page_cnt * QED_CHAIN_PBL_ENTRY_SIZE; + dma_free_coherent(&cdev->pdev->dev, + pbl_size, + p_chain->pbl.p_virt_table, p_chain->pbl.p_phys_table); +out: + vfree(p_chain->pbl.pp_virt_addr_tbl); +} + +void qed_chain_free(struct qed_dev *cdev, struct qed_chain *p_chain) +{ + switch (p_chain->mode) { + case QED_CHAIN_MODE_NEXT_PTR: + qed_chain_free_next_ptr(cdev, p_chain); + break; + case QED_CHAIN_MODE_SINGLE: + qed_chain_free_single(cdev, p_chain); + break; + case QED_CHAIN_MODE_PBL: + qed_chain_free_pbl(cdev, p_chain); + break; + } +} + +static int +qed_chain_alloc_sanity_check(struct qed_dev *cdev, + enum qed_chain_cnt_type cnt_type, + size_t elem_size, u32 page_cnt) +{ + u64 chain_size = ELEMS_PER_PAGE(elem_size) * page_cnt; + + /* The actual chain size can be larger than the maximal possible value + * after rounding up the requested elements number to pages, and after + * taking into acount the unusuable elements (next-ptr elements). + * The size of a "u16" chain can be (U16_MAX + 1) since the chain + * size/capacity fields are of a u32 type. + */ + if ((cnt_type == QED_CHAIN_CNT_TYPE_U16 && + chain_size > 0x10000) || + (cnt_type == QED_CHAIN_CNT_TYPE_U32 && + chain_size > 0x100000000ULL)) { + DP_NOTICE(cdev, + "The actual chain size (0x%llx) is larger than the maximal possible value\n", + chain_size); + return -EINVAL; + } + + return 0; +} + +static int +qed_chain_alloc_next_ptr(struct qed_dev *cdev, struct qed_chain *p_chain) +{ + void *p_virt = NULL, *p_virt_prev = NULL; dma_addr_t p_phys = 0; - void *p_virt = NULL; - u16 page_cnt = 0; - size_t size; + u32 i; - if (mode == QED_CHAIN_MODE_SINGLE) - page_cnt = 1; - else - page_cnt = QED_CHAIN_PAGE_CNT(num_elems, elem_size, mode); + for (i = 0; i < p_chain->page_cnt; i++) { + p_virt = dma_alloc_coherent(&cdev->pdev->dev, + QED_CHAIN_PAGE_SIZE, + &p_phys, GFP_KERNEL); + if (!p_virt) { + DP_NOTICE(cdev, "Failed to allocate chain memory\n"); + return -ENOMEM; + } + + if (i == 0) { + qed_chain_init_mem(p_chain, p_virt, p_phys); + qed_chain_reset(p_chain); + } else { + qed_chain_init_next_ptr_elem(p_chain, p_virt_prev, + p_virt, p_phys); + } + + p_virt_prev = p_virt; + } + /* Last page's next element should point to the beginning of the + * chain. + */ + qed_chain_init_next_ptr_elem(p_chain, p_virt_prev, + p_chain->p_virt_addr, + p_chain->p_phys_addr); + + return 0; +} + +static int +qed_chain_alloc_single(struct qed_dev *cdev, struct qed_chain *p_chain) +{ + dma_addr_t p_phys = 0; + void *p_virt = NULL; - size = page_cnt * QED_CHAIN_PAGE_SIZE; p_virt = dma_alloc_coherent(&cdev->pdev->dev, - size, &p_phys, GFP_KERNEL); + QED_CHAIN_PAGE_SIZE, &p_phys, GFP_KERNEL); if (!p_virt) { - DP_NOTICE(cdev, "Failed to allocate chain mem\n"); - goto nomem; + DP_NOTICE(cdev, "Failed to allocate chain memory\n"); + return -ENOMEM; } - if (mode == QED_CHAIN_MODE_PBL) { - size = page_cnt * QED_CHAIN_PBL_ENTRY_SIZE; - p_pbl_virt = dma_alloc_coherent(&cdev->pdev->dev, - size, &p_pbl_phys, - GFP_KERNEL); - if (!p_pbl_virt) { - DP_NOTICE(cdev, "Failed to allocate chain pbl mem\n"); - goto nomem; - } + qed_chain_init_mem(p_chain, p_virt, p_phys); + qed_chain_reset(p_chain); - qed_chain_pbl_init(p_chain, p_virt, p_phys, page_cnt, - (u8)elem_size, intended_use, - p_pbl_phys, p_pbl_virt); - } else { - qed_chain_init(p_chain, p_virt, p_phys, page_cnt, - (u8)elem_size, intended_use, mode); + return 0; +} + +static int qed_chain_alloc_pbl(struct qed_dev *cdev, struct qed_chain *p_chain) +{ + u32 page_cnt = p_chain->page_cnt, size, i; + dma_addr_t p_phys = 0, p_pbl_phys = 0; + void **pp_virt_addr_tbl = NULL; + u8 *p_pbl_virt = NULL; + void *p_virt = NULL; + + size = page_cnt * sizeof(*pp_virt_addr_tbl); + pp_virt_addr_tbl = vmalloc(size); + if (!pp_virt_addr_tbl) { + DP_NOTICE(cdev, + "Failed to allocate memory for the chain virtual addresses table\n"); + return -ENOMEM; } + memset(pp_virt_addr_tbl, 0, size); - return 0; + /* The allocation of the PBL table is done with its full size, since it + * is expected to be successive. + * qed_chain_init_pbl_mem() is called even in a case of an allocation + * failure, since pp_virt_addr_tbl was previously allocated, and it + * should be saved to allow its freeing during the error flow. + */ + size = page_cnt * QED_CHAIN_PBL_ENTRY_SIZE; + p_pbl_virt = dma_alloc_coherent(&cdev->pdev->dev, + size, &p_pbl_phys, GFP_KERNEL); + qed_chain_init_pbl_mem(p_chain, p_pbl_virt, p_pbl_phys, + pp_virt_addr_tbl); + if (!p_pbl_virt) { + DP_NOTICE(cdev, "Failed to allocate chain pbl memory\n"); + return -ENOMEM; + } -nomem: - dma_free_coherent(&cdev->pdev->dev, - page_cnt * QED_CHAIN_PAGE_SIZE, - p_virt, p_phys); - dma_free_coherent(&cdev->pdev->dev, - page_cnt * QED_CHAIN_PBL_ENTRY_SIZE, - p_pbl_virt, p_pbl_phys); + for (i = 0; i < page_cnt; i++) { + p_virt = dma_alloc_coherent(&cdev->pdev->dev, + QED_CHAIN_PAGE_SIZE, + &p_phys, GFP_KERNEL); + if (!p_virt) { + DP_NOTICE(cdev, "Failed to allocate chain memory\n"); + return -ENOMEM; + } - return -ENOMEM; + if (i == 0) { + qed_chain_init_mem(p_chain, p_virt, p_phys); + qed_chain_reset(p_chain); + } + + /* Fill the PBL table with the physical address of the page */ + *(dma_addr_t *)p_pbl_virt = p_phys; + /* Keep the virtual address of the page */ + p_chain->pbl.pp_virt_addr_tbl[i] = p_virt; + + p_pbl_virt += QED_CHAIN_PBL_ENTRY_SIZE; + } + + return 0; } -void qed_chain_free(struct qed_dev *cdev, - struct qed_chain *p_chain) +int qed_chain_alloc(struct qed_dev *cdev, + enum qed_chain_use_mode intended_use, + enum qed_chain_mode mode, + enum qed_chain_cnt_type cnt_type, + u32 num_elems, size_t elem_size, struct qed_chain *p_chain) { - size_t size; + u32 page_cnt; + int rc = 0; - if (!p_chain->p_virt_addr) - return; + if (mode == QED_CHAIN_MODE_SINGLE) + page_cnt = 1; + else + page_cnt = QED_CHAIN_PAGE_CNT(num_elems, elem_size, mode); - if (p_chain->mode == QED_CHAIN_MODE_PBL) { - size = p_chain->page_cnt * QED_CHAIN_PBL_ENTRY_SIZE; - dma_free_coherent(&cdev->pdev->dev, size, - p_chain->pbl.p_virt_table, - p_chain->pbl.p_phys_table); + rc = qed_chain_alloc_sanity_check(cdev, cnt_type, elem_size, page_cnt); + if (rc) { + DP_NOTICE(cdev, + "Cannot allocate a chain with the given arguments:\n" + "[use_mode %d, mode %d, cnt_type %d, num_elems %d, elem_size %zu]\n", + intended_use, mode, cnt_type, num_elems, elem_size); + return rc; } - size = p_chain->page_cnt * QED_CHAIN_PAGE_SIZE; - dma_free_coherent(&cdev->pdev->dev, size, - p_chain->p_virt_addr, - p_chain->p_phys_addr); + qed_chain_init_params(p_chain, page_cnt, (u8) elem_size, intended_use, + mode, cnt_type); + + switch (mode) { + case QED_CHAIN_MODE_NEXT_PTR: + rc = qed_chain_alloc_next_ptr(cdev, p_chain); + break; + case QED_CHAIN_MODE_SINGLE: + rc = qed_chain_alloc_single(cdev, p_chain); + break; + case QED_CHAIN_MODE_PBL: + rc = qed_chain_alloc_pbl(cdev, p_chain); + break; + } + if (rc) + goto nomem; + + return 0; + +nomem: + qed_chain_free(cdev, p_chain); + return rc; } -int qed_fw_l2_queue(struct qed_hwfn *p_hwfn, - u16 src_id, u16 *dst_id) +int qed_fw_l2_queue(struct qed_hwfn *p_hwfn, u16 src_id, u16 *dst_id) { if (src_id >= RESC_NUM(p_hwfn, QED_L2_QUEUE)) { u16 min, max; - min = (u16)RESC_START(p_hwfn, QED_L2_QUEUE); + min = (u16) RESC_START(p_hwfn, QED_L2_QUEUE); max = min + RESC_NUM(p_hwfn, QED_L2_QUEUE); DP_NOTICE(p_hwfn, "l2_queue id [%d] is not valid, available indices [%d - %d]\n", @@ -1593,3 +2222,499 @@ int qed_fw_rss_eng(struct qed_hwfn *p_hwfn, return 0; } + +static int qed_set_coalesce(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt, + u32 hw_addr, void *p_eth_qzone, + size_t eth_qzone_size, u8 timeset) +{ + struct coalescing_timeset *p_coal_timeset; + + if (p_hwfn->cdev->int_coalescing_mode != QED_COAL_MODE_ENABLE) { + DP_NOTICE(p_hwfn, "Coalescing configuration not enabled\n"); + return -EINVAL; + } + + p_coal_timeset = p_eth_qzone; + memset(p_coal_timeset, 0, eth_qzone_size); + SET_FIELD(p_coal_timeset->value, COALESCING_TIMESET_TIMESET, timeset); + SET_FIELD(p_coal_timeset->value, COALESCING_TIMESET_VALID, 1); + qed_memcpy_to(p_hwfn, p_ptt, hw_addr, p_eth_qzone, eth_qzone_size); + + return 0; +} + +int qed_set_rxq_coalesce(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt, + u16 coalesce, u8 qid, u16 sb_id) +{ + struct ustorm_eth_queue_zone eth_qzone; + u8 timeset, timer_res; + u16 fw_qid = 0; + u32 address; + int rc; + + /* Coalesce = (timeset << timer-resolution), timeset is 7bit wide */ + if (coalesce <= 0x7F) { + timer_res = 0; + } else if (coalesce <= 0xFF) { + timer_res = 1; + } else if (coalesce <= 0x1FF) { + timer_res = 2; + } else { + DP_ERR(p_hwfn, "Invalid coalesce value - %d\n", coalesce); + return -EINVAL; + } + timeset = (u8)(coalesce >> timer_res); + + rc = qed_fw_l2_queue(p_hwfn, (u16)qid, &fw_qid); + if (rc) + return rc; + + rc = qed_int_set_timer_res(p_hwfn, p_ptt, timer_res, sb_id, false); + if (rc) + goto out; + + address = BAR0_MAP_REG_USDM_RAM + USTORM_ETH_QUEUE_ZONE_OFFSET(fw_qid); + + rc = qed_set_coalesce(p_hwfn, p_ptt, address, ð_qzone, + sizeof(struct ustorm_eth_queue_zone), timeset); + if (rc) + goto out; + + p_hwfn->cdev->rx_coalesce_usecs = coalesce; +out: + return rc; +} + +int qed_set_txq_coalesce(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt, + u16 coalesce, u8 qid, u16 sb_id) +{ + struct xstorm_eth_queue_zone eth_qzone; + u8 timeset, timer_res; + u16 fw_qid = 0; + u32 address; + int rc; + + /* Coalesce = (timeset << timer-resolution), timeset is 7bit wide */ + if (coalesce <= 0x7F) { + timer_res = 0; + } else if (coalesce <= 0xFF) { + timer_res = 1; + } else if (coalesce <= 0x1FF) { + timer_res = 2; + } else { + DP_ERR(p_hwfn, "Invalid coalesce value - %d\n", coalesce); + return -EINVAL; + } + timeset = (u8)(coalesce >> timer_res); + + rc = qed_fw_l2_queue(p_hwfn, (u16)qid, &fw_qid); + if (rc) + return rc; + + rc = qed_int_set_timer_res(p_hwfn, p_ptt, timer_res, sb_id, true); + if (rc) + goto out; + + address = BAR0_MAP_REG_XSDM_RAM + XSTORM_ETH_QUEUE_ZONE_OFFSET(fw_qid); + + rc = qed_set_coalesce(p_hwfn, p_ptt, address, ð_qzone, + sizeof(struct xstorm_eth_queue_zone), timeset); + if (rc) + goto out; + + p_hwfn->cdev->tx_coalesce_usecs = coalesce; +out: + return rc; +} + +/* Calculate final WFQ values for all vports and configure them. + * After this configuration each vport will have + * approx min rate = min_pf_rate * (vport_wfq / QED_WFQ_UNIT) + */ +static void qed_configure_wfq_for_all_vports(struct qed_hwfn *p_hwfn, + struct qed_ptt *p_ptt, + u32 min_pf_rate) +{ + struct init_qm_vport_params *vport_params; + int i; + + vport_params = p_hwfn->qm_info.qm_vport_params; + + for (i = 0; i < p_hwfn->qm_info.num_vports; i++) { + u32 wfq_speed = p_hwfn->qm_info.wfq_data[i].min_speed; + + vport_params[i].vport_wfq = (wfq_speed * QED_WFQ_UNIT) / + min_pf_rate; + qed_init_vport_wfq(p_hwfn, p_ptt, + vport_params[i].first_tx_pq_id, + vport_params[i].vport_wfq); + } +} + +static void qed_init_wfq_default_param(struct qed_hwfn *p_hwfn, + u32 min_pf_rate) + +{ + int i; + + for (i = 0; i < p_hwfn->qm_info.num_vports; i++) + p_hwfn->qm_info.qm_vport_params[i].vport_wfq = 1; +} + +static void qed_disable_wfq_for_all_vports(struct qed_hwfn *p_hwfn, + struct qed_ptt *p_ptt, + u32 min_pf_rate) +{ + struct init_qm_vport_params *vport_params; + int i; + + vport_params = p_hwfn->qm_info.qm_vport_params; + + for (i = 0; i < p_hwfn->qm_info.num_vports; i++) { + qed_init_wfq_default_param(p_hwfn, min_pf_rate); + qed_init_vport_wfq(p_hwfn, p_ptt, + vport_params[i].first_tx_pq_id, + vport_params[i].vport_wfq); + } +} + +/* This function performs several validations for WFQ + * configuration and required min rate for a given vport + * 1. req_rate must be greater than one percent of min_pf_rate. + * 2. req_rate should not cause other vports [not configured for WFQ explicitly] + * rates to get less than one percent of min_pf_rate. + * 3. total_req_min_rate [all vports min rate sum] shouldn't exceed min_pf_rate. + */ +static int qed_init_wfq_param(struct qed_hwfn *p_hwfn, + u16 vport_id, u32 req_rate, + u32 min_pf_rate) +{ + u32 total_req_min_rate = 0, total_left_rate = 0, left_rate_per_vp = 0; + int non_requested_count = 0, req_count = 0, i, num_vports; + + num_vports = p_hwfn->qm_info.num_vports; + + /* Accounting for the vports which are configured for WFQ explicitly */ + for (i = 0; i < num_vports; i++) { + u32 tmp_speed; + + if ((i != vport_id) && + p_hwfn->qm_info.wfq_data[i].configured) { + req_count++; + tmp_speed = p_hwfn->qm_info.wfq_data[i].min_speed; + total_req_min_rate += tmp_speed; + } + } + + /* Include current vport data as well */ + req_count++; + total_req_min_rate += req_rate; + non_requested_count = num_vports - req_count; + + if (req_rate < min_pf_rate / QED_WFQ_UNIT) { + DP_VERBOSE(p_hwfn, NETIF_MSG_LINK, + "Vport [%d] - Requested rate[%d Mbps] is less than one percent of configured PF min rate[%d Mbps]\n", + vport_id, req_rate, min_pf_rate); + return -EINVAL; + } + + if (num_vports > QED_WFQ_UNIT) { + DP_VERBOSE(p_hwfn, NETIF_MSG_LINK, + "Number of vports is greater than %d\n", + QED_WFQ_UNIT); + return -EINVAL; + } + + if (total_req_min_rate > min_pf_rate) { + DP_VERBOSE(p_hwfn, NETIF_MSG_LINK, + "Total requested min rate for all vports[%d Mbps] is greater than configured PF min rate[%d Mbps]\n", + total_req_min_rate, min_pf_rate); + return -EINVAL; + } + + total_left_rate = min_pf_rate - total_req_min_rate; + + left_rate_per_vp = total_left_rate / non_requested_count; + if (left_rate_per_vp < min_pf_rate / QED_WFQ_UNIT) { + DP_VERBOSE(p_hwfn, NETIF_MSG_LINK, + "Non WFQ configured vports rate [%d Mbps] is less than one percent of configured PF min rate[%d Mbps]\n", + left_rate_per_vp, min_pf_rate); + return -EINVAL; + } + + p_hwfn->qm_info.wfq_data[vport_id].min_speed = req_rate; + p_hwfn->qm_info.wfq_data[vport_id].configured = true; + + for (i = 0; i < num_vports; i++) { + if (p_hwfn->qm_info.wfq_data[i].configured) + continue; + + p_hwfn->qm_info.wfq_data[i].min_speed = left_rate_per_vp; + } + + return 0; +} + +static int __qed_configure_vport_wfq(struct qed_hwfn *p_hwfn, + struct qed_ptt *p_ptt, u16 vp_id, u32 rate) +{ + struct qed_mcp_link_state *p_link; + int rc = 0; + + p_link = &p_hwfn->cdev->hwfns[0].mcp_info->link_output; + + if (!p_link->min_pf_rate) { + p_hwfn->qm_info.wfq_data[vp_id].min_speed = rate; + p_hwfn->qm_info.wfq_data[vp_id].configured = true; + return rc; + } + + rc = qed_init_wfq_param(p_hwfn, vp_id, rate, p_link->min_pf_rate); + + if (rc == 0) + qed_configure_wfq_for_all_vports(p_hwfn, p_ptt, + p_link->min_pf_rate); + else + DP_NOTICE(p_hwfn, + "Validation failed while configuring min rate\n"); + + return rc; +} + +static int __qed_configure_vp_wfq_on_link_change(struct qed_hwfn *p_hwfn, + struct qed_ptt *p_ptt, + u32 min_pf_rate) +{ + bool use_wfq = false; + int rc = 0; + u16 i; + + /* Validate all pre configured vports for wfq */ + for (i = 0; i < p_hwfn->qm_info.num_vports; i++) { + u32 rate; + + if (!p_hwfn->qm_info.wfq_data[i].configured) + continue; + + rate = p_hwfn->qm_info.wfq_data[i].min_speed; + use_wfq = true; + + rc = qed_init_wfq_param(p_hwfn, i, rate, min_pf_rate); + if (rc) { + DP_NOTICE(p_hwfn, + "WFQ validation failed while configuring min rate\n"); + break; + } + } + + if (!rc && use_wfq) + qed_configure_wfq_for_all_vports(p_hwfn, p_ptt, min_pf_rate); + else + qed_disable_wfq_for_all_vports(p_hwfn, p_ptt, min_pf_rate); + + return rc; +} + +/* Main API for qed clients to configure vport min rate. + * vp_id - vport id in PF Range[0 - (total_num_vports_per_pf - 1)] + * rate - Speed in Mbps needs to be assigned to a given vport. + */ +int qed_configure_vport_wfq(struct qed_dev *cdev, u16 vp_id, u32 rate) +{ + int i, rc = -EINVAL; + + /* Currently not supported; Might change in future */ + if (cdev->num_hwfns > 1) { + DP_NOTICE(cdev, + "WFQ configuration is not supported for this device\n"); + return rc; + } + + for_each_hwfn(cdev, i) { + struct qed_hwfn *p_hwfn = &cdev->hwfns[i]; + struct qed_ptt *p_ptt; + + p_ptt = qed_ptt_acquire(p_hwfn); + if (!p_ptt) + return -EBUSY; + + rc = __qed_configure_vport_wfq(p_hwfn, p_ptt, vp_id, rate); + + if (rc) { + qed_ptt_release(p_hwfn, p_ptt); + return rc; + } + + qed_ptt_release(p_hwfn, p_ptt); + } + + return rc; +} + +/* API to configure WFQ from mcp link change */ +void qed_configure_vp_wfq_on_link_change(struct qed_dev *cdev, u32 min_pf_rate) +{ + int i; + + if (cdev->num_hwfns > 1) { + DP_VERBOSE(cdev, + NETIF_MSG_LINK, + "WFQ configuration is not supported for this device\n"); + return; + } + + for_each_hwfn(cdev, i) { + struct qed_hwfn *p_hwfn = &cdev->hwfns[i]; + + __qed_configure_vp_wfq_on_link_change(p_hwfn, + p_hwfn->p_dpc_ptt, + min_pf_rate); + } +} + +int __qed_configure_pf_max_bandwidth(struct qed_hwfn *p_hwfn, + struct qed_ptt *p_ptt, + struct qed_mcp_link_state *p_link, + u8 max_bw) +{ + int rc = 0; + + p_hwfn->mcp_info->func_info.bandwidth_max = max_bw; + + if (!p_link->line_speed && (max_bw != 100)) + return rc; + + p_link->speed = (p_link->line_speed * max_bw) / 100; + p_hwfn->qm_info.pf_rl = p_link->speed; + + /* Since the limiter also affects Tx-switched traffic, we don't want it + * to limit such traffic in case there's no actual limit. + * In that case, set limit to imaginary high boundary. + */ + if (max_bw == 100) + p_hwfn->qm_info.pf_rl = 100000; + + rc = qed_init_pf_rl(p_hwfn, p_ptt, p_hwfn->rel_pf_id, + p_hwfn->qm_info.pf_rl); + + DP_VERBOSE(p_hwfn, NETIF_MSG_LINK, + "Configured MAX bandwidth to be %08x Mb/sec\n", + p_link->speed); + + return rc; +} + +/* Main API to configure PF max bandwidth where bw range is [1 - 100] */ +int qed_configure_pf_max_bandwidth(struct qed_dev *cdev, u8 max_bw) +{ + int i, rc = -EINVAL; + + if (max_bw < 1 || max_bw > 100) { + DP_NOTICE(cdev, "PF max bw valid range is [1-100]\n"); + return rc; + } + + for_each_hwfn(cdev, i) { + struct qed_hwfn *p_hwfn = &cdev->hwfns[i]; + struct qed_hwfn *p_lead = QED_LEADING_HWFN(cdev); + struct qed_mcp_link_state *p_link; + struct qed_ptt *p_ptt; + + p_link = &p_lead->mcp_info->link_output; + + p_ptt = qed_ptt_acquire(p_hwfn); + if (!p_ptt) + return -EBUSY; + + rc = __qed_configure_pf_max_bandwidth(p_hwfn, p_ptt, + p_link, max_bw); + + qed_ptt_release(p_hwfn, p_ptt); + + if (rc) + break; + } + + return rc; +} + +int __qed_configure_pf_min_bandwidth(struct qed_hwfn *p_hwfn, + struct qed_ptt *p_ptt, + struct qed_mcp_link_state *p_link, + u8 min_bw) +{ + int rc = 0; + + p_hwfn->mcp_info->func_info.bandwidth_min = min_bw; + p_hwfn->qm_info.pf_wfq = min_bw; + + if (!p_link->line_speed) + return rc; + + p_link->min_pf_rate = (p_link->line_speed * min_bw) / 100; + + rc = qed_init_pf_wfq(p_hwfn, p_ptt, p_hwfn->rel_pf_id, min_bw); + + DP_VERBOSE(p_hwfn, NETIF_MSG_LINK, + "Configured MIN bandwidth to be %d Mb/sec\n", + p_link->min_pf_rate); + + return rc; +} + +/* Main API to configure PF min bandwidth where bw range is [1-100] */ +int qed_configure_pf_min_bandwidth(struct qed_dev *cdev, u8 min_bw) +{ + int i, rc = -EINVAL; + + if (min_bw < 1 || min_bw > 100) { + DP_NOTICE(cdev, "PF min bw valid range is [1-100]\n"); + return rc; + } + + for_each_hwfn(cdev, i) { + struct qed_hwfn *p_hwfn = &cdev->hwfns[i]; + struct qed_hwfn *p_lead = QED_LEADING_HWFN(cdev); + struct qed_mcp_link_state *p_link; + struct qed_ptt *p_ptt; + + p_link = &p_lead->mcp_info->link_output; + + p_ptt = qed_ptt_acquire(p_hwfn); + if (!p_ptt) + return -EBUSY; + + rc = __qed_configure_pf_min_bandwidth(p_hwfn, p_ptt, + p_link, min_bw); + if (rc) { + qed_ptt_release(p_hwfn, p_ptt); + return rc; + } + + if (p_link->min_pf_rate) { + u32 min_rate = p_link->min_pf_rate; + + rc = __qed_configure_vp_wfq_on_link_change(p_hwfn, + p_ptt, + min_rate); + } + + qed_ptt_release(p_hwfn, p_ptt); + } + + return rc; +} + +void qed_clean_wfq_db(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt) +{ + struct qed_mcp_link_state *p_link; + + p_link = &p_hwfn->mcp_info->link_output; + + if (p_link->min_pf_rate) + qed_disable_wfq_for_all_vports(p_hwfn, p_ptt, + p_link->min_pf_rate); + + memset(p_hwfn->qm_info.wfq_data, 0, + sizeof(*p_hwfn->qm_info.wfq_data) * p_hwfn->qm_info.num_vports); +} |