From patchwork Tue Nov 4 10:01:27 2014 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: "Chen, Jing D" X-Patchwork-Id: 1122 Return-Path: X-Original-To: patchwork@dpdk.org Delivered-To: patchwork@dpdk.org Received: from [92.243.14.124] (localhost [IPv6:::1]) by dpdk.org (Postfix) with ESMTP id 88BC87F40; Tue, 4 Nov 2014 10:53:01 +0100 (CET) Received: from mga01.intel.com (mga01.intel.com [192.55.52.88]) by dpdk.org (Postfix) with ESMTP id B68457E75 for ; Tue, 4 Nov 2014 10:52:57 +0100 (CET) Received: from fmsmga001.fm.intel.com ([10.253.24.23]) by fmsmga101.fm.intel.com with ESMTP; 04 Nov 2014 02:02:14 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.07,312,1413270000"; d="scan'208";a="616716409" Received: from shvmail01.sh.intel.com ([10.239.29.42]) by fmsmga001.fm.intel.com with ESMTP; 04 Nov 2014 02:01:48 -0800 Received: from shecgisg003.sh.intel.com (shecgisg003.sh.intel.com [10.239.29.90]) by shvmail01.sh.intel.com with ESMTP id sA4A1kYL020749; Tue, 4 Nov 2014 18:01:46 +0800 Received: from shecgisg003.sh.intel.com (localhost [127.0.0.1]) by shecgisg003.sh.intel.com (8.13.6/8.13.6/SuSE Linux 0.8) with ESMTP id sA4A1i8E029023; Tue, 4 Nov 2014 18:01:46 +0800 Received: (from jingche2@localhost) by shecgisg003.sh.intel.com (8.13.6/8.13.6/Submit) id sA4A1iHQ029019; Tue, 4 Nov 2014 18:01:44 +0800 From: "Chen Jing D(Mark)" To: dev@dpdk.org Date: Tue, 4 Nov 2014 18:01:27 +0800 Message-Id: <1415095289-28961-5-git-send-email-jing.d.chen@intel.com> X-Mailer: git-send-email 1.7.12.2 In-Reply-To: <1415095289-28961-1-git-send-email-jing.d.chen@intel.com> References: <1411478047-1251-2-git-send-email-jing.d.chen@intel.com> <1415095289-28961-1-git-send-email-jing.d.chen@intel.com> Subject: [dpdk-dev] [PATCH v3 4/6] i40e: add VMDQ support X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: patches and discussions about DPDK List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" From: "Chen Jing D(Mark)" The change includes several parts: 1. Get maximum number of VMDQ pools supported in dev_init. 2. Fill VMDQ info in i40e_dev_info_get. 3. Setup VMDQ pools in i40e_dev_configure. 4. i40e_vsi_setup change to support creation of VMDQ VSI. Signed-off-by: Chen Jing D(Mark) --- config/common_bsdapp | 1 + config/common_linuxapp | 1 + lib/librte_pmd_i40e/i40e_ethdev.c | 236 ++++++++++++++++++++++++++++++++----- lib/librte_pmd_i40e/i40e_ethdev.h | 17 +++- 4 files changed, 225 insertions(+), 30 deletions(-) diff --git a/config/common_bsdapp b/config/common_bsdapp index eebd05b..9dc9f56 100644 --- a/config/common_bsdapp +++ b/config/common_bsdapp @@ -183,6 +183,7 @@ CONFIG_RTE_LIBRTE_I40E_PF_DISABLE_STRIP_CRC=y CONFIG_RTE_LIBRTE_I40E_RX_ALLOW_BULK_ALLOC=n CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC=n CONFIG_RTE_LIBRTE_I40E_QUEUE_NUM_PER_VF=4 +CONFIG_RTE_LIBRTE_I40E_QUEUE_NUM_PER_VM=4 # interval up to 8160 us, aligned to 2 (or default value) CONFIG_RTE_LIBRTE_I40E_ITR_INTERVAL=-1 diff --git a/config/common_linuxapp b/config/common_linuxapp index c5751bd..8be79c3 100644 --- a/config/common_linuxapp +++ b/config/common_linuxapp @@ -206,6 +206,7 @@ CONFIG_RTE_LIBRTE_I40E_PF_DISABLE_STRIP_CRC=n CONFIG_RTE_LIBRTE_I40E_RX_ALLOW_BULK_ALLOC=y CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC=n CONFIG_RTE_LIBRTE_I40E_QUEUE_NUM_PER_VF=4 +CONFIG_RTE_LIBRTE_I40E_QUEUE_NUM_PER_VM=4 # interval up to 8160 us, aligned to 2 (or default value) CONFIG_RTE_LIBRTE_I40E_ITR_INTERVAL=-1 diff --git a/lib/librte_pmd_i40e/i40e_ethdev.c b/lib/librte_pmd_i40e/i40e_ethdev.c index 661d146..020881f 100644 --- a/lib/librte_pmd_i40e/i40e_ethdev.c +++ b/lib/librte_pmd_i40e/i40e_ethdev.c @@ -163,6 +163,7 @@ static int i40e_get_cap(struct i40e_hw *hw); static int i40e_pf_parameter_init(struct rte_eth_dev *dev); static int i40e_pf_setup(struct i40e_pf *pf); static int i40e_vsi_init(struct i40e_vsi *vsi); +static int i40e_vmdq_setup(struct rte_eth_dev *dev); static void i40e_stat_update_32(struct i40e_hw *hw, uint32_t reg, bool offset_loaded, uint64_t *offset, uint64_t *stat); static void i40e_stat_update_48(struct i40e_hw *hw, @@ -275,21 +276,11 @@ static struct eth_driver rte_i40e_pmd = { }; static inline int -i40e_prev_power_of_2(int n) +i40e_align_floor(int n) { - int p = n; - - --p; - p |= p >> 1; - p |= p >> 2; - p |= p >> 4; - p |= p >> 8; - p |= p >> 16; - if (p == (n - 1)) - return n; - p >>= 1; - - return ++p; + if (n == 0) + return 0; + return (1 << (sizeof(n) * CHAR_BIT - 1 - __builtin_clz(n))); } static inline int @@ -506,7 +497,7 @@ eth_i40e_dev_init(__rte_unused struct eth_driver *eth_drv, if (!dev->data->mac_addrs) { PMD_INIT_LOG(ERR, "Failed to allocated memory " "for storing mac address"); - goto err_get_mac_addr; + goto err_mac_alloc; } ether_addr_copy((struct ether_addr *)hw->mac.perm_addr, &dev->data->mac_addrs[0]); @@ -527,8 +518,9 @@ eth_i40e_dev_init(__rte_unused struct eth_driver *eth_drv, return 0; +err_mac_alloc: + i40e_vsi_release(pf->main_vsi); err_setup_pf_switch: - rte_free(pf->main_vsi); err_get_mac_addr: err_configure_lan_hmc: (void)i40e_shutdown_lan_hmc(hw); @@ -547,6 +539,27 @@ err_get_capabilities: static int i40e_dev_configure(struct rte_eth_dev *dev) { + int ret; + enum rte_eth_rx_mq_mode mq_mode = dev->data->dev_conf.rxmode.mq_mode; + + /* VMDQ setup. + * Needs to move VMDQ setting out of i40e_pf_config_mq_rx() as VMDQ and + * RSS setting have different requirements. + * General PMD driver call sequence are NIC init, configure, + * rx/tx_queue_setup and dev_start. In rx/tx_queue_setup() function, it + * will try to lookup the VSI that specific queue belongs to if VMDQ + * applicable. So, VMDQ setting has to be done before + * rx/tx_queue_setup(). This function is good to place vmdq_setup. + * For RSS setting, it will try to calculate actual configured RX queue + * number, which will be available after rx_queue_setup(). dev_start() + * function is good to place RSS setup. + */ + if (mq_mode & ETH_MQ_RX_VMDQ_FLAG) { + ret = i40e_vmdq_setup(dev); + if (ret) + return ret; + } + return i40e_dev_init_vlan(dev); } @@ -1431,6 +1444,15 @@ i40e_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info) .txq_flags = ETH_TXQ_FLAGS_NOMULTSEGS | ETH_TXQ_FLAGS_NOOFFLOADS, }; + if (pf->flags | I40E_FLAG_VMDQ) { + dev_info->max_vmdq_pools = pf->max_nb_vmdq_vsi; + dev_info->vmdq_queue_base = dev_info->max_rx_queues; + dev_info->vmdq_queue_num = pf->vmdq_nb_qps * + pf->max_nb_vmdq_vsi; + dev_info->vmdq_pool_base = I40E_VMDQ_POOL_BASE; + dev_info->max_rx_queues += dev_info->vmdq_queue_num; + dev_info->max_tx_queues += dev_info->vmdq_queue_num; + } } static int @@ -1972,7 +1994,7 @@ i40e_pf_parameter_init(struct rte_eth_dev *dev) { struct i40e_pf *pf = I40E_DEV_PRIVATE_TO_PF(dev->data->dev_private); struct i40e_hw *hw = I40E_PF_TO_HW(pf); - uint16_t sum_queues = 0, sum_vsis; + uint16_t sum_queues = 0, sum_vsis, left_queues; /* First check if FW support SRIOV */ if (dev->pci_dev->max_vfs && !hw->func_caps.sr_iov_1_1) { @@ -1988,7 +2010,7 @@ i40e_pf_parameter_init(struct rte_eth_dev *dev) pf->flags |= I40E_FLAG_RSS; pf->lan_nb_qps = RTE_MIN(hw->func_caps.num_tx_qp, (uint32_t)(1 << hw->func_caps.rss_table_entry_width)); - pf->lan_nb_qps = i40e_prev_power_of_2(pf->lan_nb_qps); + pf->lan_nb_qps = i40e_align_floor(pf->lan_nb_qps); } else pf->lan_nb_qps = 1; sum_queues = pf->lan_nb_qps; @@ -2022,11 +2044,19 @@ i40e_pf_parameter_init(struct rte_eth_dev *dev) if (hw->func_caps.vmdq) { pf->flags |= I40E_FLAG_VMDQ; - pf->vmdq_nb_qps = I40E_DEFAULT_QP_NUM_VMDQ; - sum_queues += pf->vmdq_nb_qps; - sum_vsis += 1; - PMD_INIT_LOG(INFO, "VMDQ queue pairs:%u", pf->vmdq_nb_qps); + pf->vmdq_nb_qps = RTE_LIBRTE_I40E_QUEUE_NUM_PER_VM; + pf->max_nb_vmdq_vsi = 1; + /* + * If VMDQ available, assume a single VSI can be created. Will adjust + * later. + */ + sum_queues += pf->vmdq_nb_qps * pf->max_nb_vmdq_vsi; + sum_vsis += pf->max_nb_vmdq_vsi; + } else { + pf->vmdq_nb_qps = 0; + pf->max_nb_vmdq_vsi = 0; } + pf->nb_cfg_vmdq_vsi = 0; if (hw->func_caps.fd) { pf->flags |= I40E_FLAG_FDIR; @@ -2047,6 +2077,22 @@ i40e_pf_parameter_init(struct rte_eth_dev *dev) return -EINVAL; } + /* Adjust VMDQ setting to support as many VMs as possible */ + if (pf->flags & I40E_FLAG_VMDQ) { + left_queues = hw->func_caps.num_rx_qp - sum_queues; + + pf->max_nb_vmdq_vsi += RTE_MIN(left_queues / pf->vmdq_nb_qps, + pf->max_num_vsi - sum_vsis); + + /* Limit the max VMDQ number that rte_ether that can support */ + pf->max_nb_vmdq_vsi = RTE_MIN(pf->max_nb_vmdq_vsi, + ETH_64_POOLS - 1); + + PMD_INIT_LOG(INFO, "Max VMDQ VSI num:%u", + pf->max_nb_vmdq_vsi); + PMD_INIT_LOG(INFO, "VMDQ queue pairs:%u", pf->vmdq_nb_qps); + } + /* Each VSI occupy 1 MSIX interrupt at least, plus IRQ0 for misc intr * cause */ if (sum_vsis > hw->func_caps.num_msix_vectors - 1) { @@ -2439,7 +2485,7 @@ i40e_vsi_config_tc_queue_mapping(struct i40e_vsi *vsi, vsi->enabled_tc = enabled_tcmap; /* Number of queues per enabled TC */ - qpnum_per_tc = i40e_prev_power_of_2(vsi->nb_qps / total_tc); + qpnum_per_tc = i40e_align_floor(vsi->nb_qps / total_tc); qpnum_per_tc = RTE_MIN(qpnum_per_tc, I40E_MAX_Q_PER_TC); bsf = rte_bsf32(qpnum_per_tc); @@ -2752,6 +2798,9 @@ i40e_vsi_setup(struct i40e_pf *pf, case I40E_VSI_SRIOV : vsi->nb_qps = pf->vf_nb_qps; break; + case I40E_VSI_VMDQ2: + vsi->nb_qps = pf->vmdq_nb_qps; + break; default: goto fail_mem; } @@ -2893,8 +2942,44 @@ i40e_vsi_setup(struct i40e_pf *pf, * Since VSI is not created yet, only configure parameter, * will add vsi below. */ - } - else { + } else if (type == I40E_VSI_VMDQ2) { + memset(&ctxt, 0, sizeof(ctxt)); + /* + * For other VSI, the uplink_seid equals to uplink VSI's + * uplink_seid since they share same VEB + */ + vsi->uplink_seid = uplink_vsi->uplink_seid; + ctxt.pf_num = hw->pf_id; + ctxt.vf_num = 0; + ctxt.uplink_seid = vsi->uplink_seid; + ctxt.connection_type = 0x1; + ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2; + + ctxt.info.valid_sections |= + rte_cpu_to_le_16(I40E_AQ_VSI_PROP_SWITCH_VALID); + /* user_param carries flag to enable loop back */ + if (user_param) { + ctxt.info.switch_id = + rte_cpu_to_le_16(I40E_AQ_VSI_SW_ID_FLAG_LOCAL_LB); + ctxt.info.switch_id |= + rte_cpu_to_le_16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB); + } + + /* Configure port/vlan */ + ctxt.info.valid_sections |= + rte_cpu_to_le_16(I40E_AQ_VSI_PROP_VLAN_VALID); + ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL; + ret = i40e_vsi_config_tc_queue_mapping(vsi, &ctxt.info, + I40E_DEFAULT_TCMAP); + if (ret != I40E_SUCCESS) { + PMD_DRV_LOG(ERR, "Failed to configure " + "TC queue mapping"); + goto fail_msix_alloc; + } + ctxt.info.up_enable_bits = I40E_DEFAULT_TCMAP; + ctxt.info.valid_sections |= + rte_cpu_to_le_16(I40E_AQ_VSI_PROP_SCHED_VALID); + } else { PMD_DRV_LOG(ERR, "VSI: Not support other type VSI yet"); goto fail_msix_alloc; } @@ -3069,7 +3154,6 @@ i40e_pf_setup(struct i40e_pf *pf) { struct i40e_hw *hw = I40E_PF_TO_HW(pf); struct i40e_filter_control_settings settings; - struct rte_eth_dev_data *dev_data = pf->dev_data; struct i40e_vsi *vsi; int ret; @@ -3091,8 +3175,6 @@ i40e_pf_setup(struct i40e_pf *pf) return I40E_ERR_NOT_READY; } pf->main_vsi = vsi; - dev_data->nb_rx_queues = vsi->nb_qps; - dev_data->nb_tx_queues = vsi->nb_qps; /* Configure filter control */ memset(&settings, 0, sizeof(settings)); @@ -3363,6 +3445,102 @@ i40e_vsi_init(struct i40e_vsi *vsi) return err; } +static int +i40e_vmdq_setup(struct rte_eth_dev *dev) +{ + struct rte_eth_conf *conf = &dev->data->dev_conf; + struct i40e_pf *pf = I40E_DEV_PRIVATE_TO_PF(dev->data->dev_private); + int i, err, conf_vsis, j, loop; + struct i40e_vsi *vsi; + struct i40e_vmdq_info *vmdq_info; + struct rte_eth_vmdq_rx_conf *vmdq_conf; + struct i40e_hw *hw = I40E_PF_TO_HW(pf); + + /* + * Disable interrupt to avoid message from VF. Furthermore, it will + * avoid race condition in VSI creation/destroy. + */ + i40e_pf_disable_irq0(hw); + + if ((pf->flags & I40E_FLAG_VMDQ) == 0) { + PMD_INIT_LOG(ERR, "FW doesn't support VMDQ"); + return -ENOTSUP; + } + + conf_vsis = conf->rx_adv_conf.vmdq_rx_conf.nb_queue_pools; + if (conf_vsis > pf->max_nb_vmdq_vsi) { + PMD_INIT_LOG(ERR, "VMDQ config: %u, max support:%u", + conf->rx_adv_conf.vmdq_rx_conf.nb_queue_pools, + pf->max_nb_vmdq_vsi); + return -ENOTSUP; + } + + if (pf->vmdq != NULL) { + PMD_INIT_LOG(INFO, "VMDQ already configured"); + return 0; + } + + pf->vmdq = rte_zmalloc("vmdq_info_struct", + sizeof(*vmdq_info) * conf_vsis, 0); + + if (pf->vmdq == NULL) { + PMD_INIT_LOG(ERR, "Failed to allocate memory"); + return -ENOMEM; + } + + vmdq_conf = &conf->rx_adv_conf.vmdq_rx_conf; + + /* Create VMDQ VSI */ + for (i = 0; i < conf_vsis; i++) { + vsi = i40e_vsi_setup(pf, I40E_VSI_VMDQ2, pf->main_vsi, + vmdq_conf->enable_loop_back); + if (vsi == NULL) { + PMD_INIT_LOG(ERR, "Failed to create VMDQ VSI"); + err = -1; + goto err_vsi_setup; + } + vmdq_info = &pf->vmdq[i]; + vmdq_info->pf = pf; + vmdq_info->vsi = vsi; + } + pf->nb_cfg_vmdq_vsi = conf_vsis; + + /* Configure Vlan */ + loop = sizeof(vmdq_conf->pool_map[0].pools) * CHAR_BIT; + for (i = 0; i < vmdq_conf->nb_pool_maps; i++) { + for (j = 0; j < loop && j < pf->nb_cfg_vmdq_vsi; j++) { + if (vmdq_conf->pool_map[i].pools & (1UL << j)) { + PMD_INIT_LOG(INFO, "Add vlan %u to vmdq pool %u", + vmdq_conf->pool_map[i].vlan_id, j); + + err = i40e_vsi_add_vlan(pf->vmdq[j].vsi, + vmdq_conf->pool_map[i].vlan_id); + if (err) { + PMD_INIT_LOG(ERR, "Failed to add vlan"); + err = -1; + goto err_vsi_setup; + } + } + } + } + + i40e_pf_enable_irq0(hw); + + return 0; + +err_vsi_setup: + for (i = 0; i < conf_vsis; i++) + if (pf->vmdq[i].vsi == NULL) + break; + else + i40e_vsi_release(pf->vmdq[i].vsi); + + rte_free(pf->vmdq); + pf->vmdq = NULL; + i40e_pf_enable_irq0(hw); + return err; +} + static void i40e_stat_update_32(struct i40e_hw *hw, uint32_t reg, @@ -4639,7 +4817,7 @@ i40e_pf_config_rss(struct i40e_pf *pf) struct i40e_hw *hw = I40E_PF_TO_HW(pf); struct rte_eth_rss_conf rss_conf; uint32_t i, lut = 0; - uint16_t j, num = i40e_prev_power_of_2(pf->dev_data->nb_rx_queues); + uint16_t j, num = i40e_align_floor(pf->dev_data->nb_rx_queues); for (i = 0, j = 0; i < hw->func_caps.rss_table_size; i++, j++) { if (j == num) diff --git a/lib/librte_pmd_i40e/i40e_ethdev.h b/lib/librte_pmd_i40e/i40e_ethdev.h index e61d258..69512cd 100644 --- a/lib/librte_pmd_i40e/i40e_ethdev.h +++ b/lib/librte_pmd_i40e/i40e_ethdev.h @@ -47,13 +47,15 @@ #define I40E_QUEUE_BASE_ADDR_UNIT 128 /* number of VSIs and queue default setting */ #define I40E_MAX_QP_NUM_PER_VF 16 -#define I40E_DEFAULT_QP_NUM_VMDQ 64 #define I40E_DEFAULT_QP_NUM_FDIR 64 #define I40E_UINT32_BIT_SIZE (CHAR_BIT * sizeof(uint32_t)) #define I40E_VFTA_SIZE (4096 / I40E_UINT32_BIT_SIZE) /* Default TC traffic in case DCB is not enabled */ #define I40E_DEFAULT_TCMAP 0x1 +/* Always assign pool 0 to main VSI, VMDQ will start from 1 */ +#define I40E_VMDQ_POOL_BASE 1 + /* i40e flags */ #define I40E_FLAG_RSS (1ULL << 0) #define I40E_FLAG_DCB (1ULL << 1) @@ -233,6 +235,14 @@ struct i40e_pf_vf { }; /* + * Structure to store private data for VMDQ instance + */ +struct i40e_vmdq_info { + struct i40e_pf *pf; + struct i40e_vsi *vsi; +}; + +/* * Structure to store private data specific for PF instance. */ struct i40e_pf { @@ -264,6 +274,11 @@ struct i40e_pf { /* store VXLAN UDP ports */ uint16_t vxlan_ports[I40E_MAX_PF_UDP_OFFLOAD_PORTS]; uint16_t vxlan_bitmap; /* Vxlan bit mask */ + + /* VMDQ related info */ + uint16_t max_nb_vmdq_vsi; /* Max number of VMDQ VSIs supported */ + uint16_t nb_cfg_vmdq_vsi; /* number of VMDQ VSIs configured */ + struct i40e_vmdq_info *vmdq; }; enum pending_msg {