common/cnxk: convert warning to debug print
[dpdk.git] / drivers / net / hns3 / hns3_common.c
index 85316d3..dcdc609 100644 (file)
@@ -3,9 +3,154 @@
  */
 
 #include <rte_kvargs.h>
+#include <rte_bus_pci.h>
+#include <ethdev_pci.h>
+#include <rte_pci.h>
 
-#include "hns3_logs.h"
 #include "hns3_common.h"
+#include "hns3_logs.h"
+#include "hns3_regs.h"
+#include "hns3_rxtx.h"
+
+int
+hns3_fw_version_get(struct rte_eth_dev *eth_dev, char *fw_version,
+                   size_t fw_size)
+{
+       struct hns3_adapter *hns = eth_dev->data->dev_private;
+       struct hns3_hw *hw = &hns->hw;
+       uint32_t version = hw->fw_version;
+       int ret;
+
+       ret = snprintf(fw_version, fw_size, "%lu.%lu.%lu.%lu",
+                      hns3_get_field(version, HNS3_FW_VERSION_BYTE3_M,
+                                     HNS3_FW_VERSION_BYTE3_S),
+                      hns3_get_field(version, HNS3_FW_VERSION_BYTE2_M,
+                                     HNS3_FW_VERSION_BYTE2_S),
+                      hns3_get_field(version, HNS3_FW_VERSION_BYTE1_M,
+                                     HNS3_FW_VERSION_BYTE1_S),
+                      hns3_get_field(version, HNS3_FW_VERSION_BYTE0_M,
+                                     HNS3_FW_VERSION_BYTE0_S));
+       if (ret < 0)
+               return -EINVAL;
+
+       ret += 1; /* add the size of '\0' */
+       if (fw_size < (size_t)ret)
+               return ret;
+       else
+               return 0;
+}
+
+int
+hns3_dev_infos_get(struct rte_eth_dev *eth_dev, struct rte_eth_dev_info *info)
+{
+       struct hns3_adapter *hns = eth_dev->data->dev_private;
+       struct hns3_hw *hw = &hns->hw;
+       uint16_t queue_num = hw->tqps_num;
+
+       /*
+        * In interrupt mode, 'max_rx_queues' is set based on the number of
+        * MSI-X interrupt resources of the hardware.
+        */
+       if (hw->data->dev_conf.intr_conf.rxq == 1)
+               queue_num = hw->intr_tqps_num;
+
+       info->max_rx_queues = queue_num;
+       info->max_tx_queues = hw->tqps_num;
+       info->max_rx_pktlen = HNS3_MAX_FRAME_LEN; /* CRC included */
+       info->min_rx_bufsize = HNS3_MIN_BD_BUF_SIZE;
+       info->max_mtu = info->max_rx_pktlen - HNS3_ETH_OVERHEAD;
+       info->max_lro_pkt_size = HNS3_MAX_LRO_SIZE;
+       info->rx_offload_capa = (RTE_ETH_RX_OFFLOAD_IPV4_CKSUM |
+                                RTE_ETH_RX_OFFLOAD_TCP_CKSUM |
+                                RTE_ETH_RX_OFFLOAD_UDP_CKSUM |
+                                RTE_ETH_RX_OFFLOAD_SCTP_CKSUM |
+                                RTE_ETH_RX_OFFLOAD_OUTER_IPV4_CKSUM |
+                                RTE_ETH_RX_OFFLOAD_OUTER_UDP_CKSUM |
+                                RTE_ETH_RX_OFFLOAD_SCATTER |
+                                RTE_ETH_RX_OFFLOAD_VLAN_STRIP |
+                                RTE_ETH_RX_OFFLOAD_VLAN_FILTER |
+                                RTE_ETH_RX_OFFLOAD_RSS_HASH |
+                                RTE_ETH_RX_OFFLOAD_TCP_LRO);
+       info->tx_offload_capa = (RTE_ETH_TX_OFFLOAD_OUTER_IPV4_CKSUM |
+                                RTE_ETH_TX_OFFLOAD_IPV4_CKSUM |
+                                RTE_ETH_TX_OFFLOAD_TCP_CKSUM |
+                                RTE_ETH_TX_OFFLOAD_UDP_CKSUM |
+                                RTE_ETH_TX_OFFLOAD_SCTP_CKSUM |
+                                RTE_ETH_TX_OFFLOAD_MULTI_SEGS |
+                                RTE_ETH_TX_OFFLOAD_TCP_TSO |
+                                RTE_ETH_TX_OFFLOAD_VXLAN_TNL_TSO |
+                                RTE_ETH_TX_OFFLOAD_GRE_TNL_TSO |
+                                RTE_ETH_TX_OFFLOAD_GENEVE_TNL_TSO |
+                                RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE |
+                                RTE_ETH_TX_OFFLOAD_VLAN_INSERT);
+
+       if (!hw->port_base_vlan_cfg.state)
+               info->tx_offload_capa |= RTE_ETH_TX_OFFLOAD_QINQ_INSERT;
+
+       if (hns3_dev_get_support(hw, OUTER_UDP_CKSUM))
+               info->tx_offload_capa |= RTE_ETH_TX_OFFLOAD_OUTER_UDP_CKSUM;
+
+       if (hns3_dev_get_support(hw, INDEP_TXRX))
+               info->dev_capa = RTE_ETH_DEV_CAPA_RUNTIME_RX_QUEUE_SETUP |
+                                RTE_ETH_DEV_CAPA_RUNTIME_TX_QUEUE_SETUP;
+       info->dev_capa &= ~RTE_ETH_DEV_CAPA_FLOW_RULE_KEEP;
+
+       if (hns3_dev_get_support(hw, PTP))
+               info->rx_offload_capa |= RTE_ETH_RX_OFFLOAD_TIMESTAMP;
+
+       info->rx_desc_lim = (struct rte_eth_desc_lim) {
+               .nb_max = HNS3_MAX_RING_DESC,
+               .nb_min = HNS3_MIN_RING_DESC,
+               .nb_align = HNS3_ALIGN_RING_DESC,
+       };
+
+       info->tx_desc_lim = (struct rte_eth_desc_lim) {
+               .nb_max = HNS3_MAX_RING_DESC,
+               .nb_min = HNS3_MIN_RING_DESC,
+               .nb_align = HNS3_ALIGN_RING_DESC,
+               .nb_seg_max = HNS3_MAX_TSO_BD_PER_PKT,
+               .nb_mtu_seg_max = hw->max_non_tso_bd_num,
+       };
+
+       info->default_rxconf = (struct rte_eth_rxconf) {
+               .rx_free_thresh = HNS3_DEFAULT_RX_FREE_THRESH,
+               /*
+                * If there are no available Rx buffer descriptors, incoming
+                * packets are always dropped by hardware based on hns3 network
+                * engine.
+                */
+               .rx_drop_en = 1,
+               .offloads = 0,
+       };
+       info->default_txconf = (struct rte_eth_txconf) {
+               .tx_rs_thresh = HNS3_DEFAULT_TX_RS_THRESH,
+               .offloads = 0,
+       };
+
+       info->reta_size = hw->rss_ind_tbl_size;
+       info->hash_key_size = HNS3_RSS_KEY_SIZE;
+       info->flow_type_rss_offloads = HNS3_ETH_RSS_SUPPORT;
+
+       info->default_rxportconf.burst_size = HNS3_DEFAULT_PORT_CONF_BURST_SIZE;
+       info->default_txportconf.burst_size = HNS3_DEFAULT_PORT_CONF_BURST_SIZE;
+       info->default_rxportconf.nb_queues = HNS3_DEFAULT_PORT_CONF_QUEUES_NUM;
+       info->default_txportconf.nb_queues = HNS3_DEFAULT_PORT_CONF_QUEUES_NUM;
+       info->default_rxportconf.ring_size = HNS3_DEFAULT_RING_DESC;
+       info->default_txportconf.ring_size = HNS3_DEFAULT_RING_DESC;
+
+       /*
+        * Next is the PF/VF difference section.
+        */
+       if (!hns->is_vf) {
+               info->max_mac_addrs = HNS3_UC_MACADDR_NUM;
+               info->rx_offload_capa |= RTE_ETH_RX_OFFLOAD_KEEP_CRC;
+               info->speed_capa = hns3_get_speed_capa(hw);
+       } else {
+               info->max_mac_addrs = HNS3_VF_UC_MACADDR_NUM;
+       }
+
+       return 0;
+}
 
 static int
 hns3_parse_io_hint_func(const char *key, const char *value, void *extra_args)
@@ -54,7 +199,7 @@ hns3_parse_dev_caps_mask(const char *key, const char *value, void *extra_args)
 
        RTE_SET_USED(key);
 
-       val = strtoull(value, NULL, 16);
+       val = strtoull(value, NULL, HNS3_CONVERT_TO_HEXADECIMAL);
        *(uint64_t *)extra_args = val;
 
        return 0;
@@ -67,7 +212,13 @@ hns3_parse_mbx_time_limit(const char *key, const char *value, void *extra_args)
 
        RTE_SET_USED(key);
 
-       val = strtoul(value, NULL, 10);
+       val = strtoul(value, NULL, HNS3_CONVERT_TO_DECIMAL);
+
+       /*
+        * 500ms is empirical value in process of mailbox communication. If
+        * the delay value is set to one lower than the empirical value, mailbox
+        * communication may fail.
+        */
        if (val > HNS3_MBX_DEF_TIME_LIMIT_MS && val <= UINT16_MAX)
                *(uint16_t *)extra_args = val;
 
@@ -85,6 +236,12 @@ hns3_parse_devargs(struct rte_eth_dev *dev)
        uint64_t dev_caps_mask = 0;
        struct rte_kvargs *kvlist;
 
+       /* Set default value of runtime config parameters. */
+       hns->rx_func_hint = HNS3_IO_FUNC_HINT_NONE;
+       hns->tx_func_hint = HNS3_IO_FUNC_HINT_NONE;
+       hns->dev_caps_mask = 0;
+       hns->mbx_time_limit_ms = HNS3_MBX_DEF_TIME_LIMIT_MS;
+
        if (dev->device->devargs == NULL)
                return;
 
@@ -116,6 +273,11 @@ hns3_parse_devargs(struct rte_eth_dev *dev)
                hns3_warn(hw, "parsed %s = 0x%" PRIx64 ".",
                          HNS3_DEVARG_DEV_CAPS_MASK, dev_caps_mask);
        hns->dev_caps_mask = dev_caps_mask;
+
+       if (mbx_time_limit_ms != HNS3_MBX_DEF_TIME_LIMIT_MS)
+               hns3_warn(hw, "parsed %s = %u.", HNS3_DEVARG_MBX_TIME_LIMIT_MS,
+                               mbx_time_limit_ms);
+       hns->mbx_time_limit_ms = mbx_time_limit_ms;
 }
 
 void
@@ -154,10 +316,10 @@ hns3_clock_gettime_ms(void)
 void hns3_ether_format_addr(char *buf, uint16_t size,
                            const struct rte_ether_addr *ether_addr)
 {
-       snprintf(buf, size, "%02X:**:**:**:%02X:%02X",
-               ether_addr->addr_bytes[0],
-               ether_addr->addr_bytes[4],
-               ether_addr->addr_bytes[5]);
+       (void)snprintf(buf, size, "%02X:**:**:**:%02X:%02X",
+                       ether_addr->addr_bytes[0],
+                       ether_addr->addr_bytes[4],
+                       ether_addr->addr_bytes[5]);
 }
 
 static int
@@ -424,3 +586,260 @@ hns3_remove_mac_addr(struct rte_eth_dev *dev, uint32_t idx)
                         ret);
        }
 }
+
+int
+hns3_init_mac_addrs(struct rte_eth_dev *dev)
+{
+       struct hns3_hw *hw = HNS3_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+       struct hns3_adapter *hns = HNS3_DEV_HW_TO_ADAPTER(hw);
+       const char *memory_name = hns->is_vf ? "hns3vf-mac" : "hns3-mac";
+       uint16_t mac_addrs_capa = hns->is_vf ? HNS3_VF_UC_MACADDR_NUM :
+                                               HNS3_UC_MACADDR_NUM;
+       char mac_str[RTE_ETHER_ADDR_FMT_SIZE];
+       struct rte_ether_addr *eth_addr;
+
+       /* Allocate memory for storing MAC addresses */
+       dev->data->mac_addrs = rte_zmalloc(memory_name,
+                               sizeof(struct rte_ether_addr) * mac_addrs_capa,
+                               0);
+       if (dev->data->mac_addrs == NULL) {
+               hns3_err(hw, "failed to allocate %zx bytes needed to store MAC addresses",
+                            sizeof(struct rte_ether_addr) * mac_addrs_capa);
+               return -ENOMEM;
+       }
+
+       eth_addr = (struct rte_ether_addr *)hw->mac.mac_addr;
+       if (!hns->is_vf) {
+               if (!rte_is_valid_assigned_ether_addr(eth_addr)) {
+                       rte_eth_random_addr(hw->mac.mac_addr);
+                       hns3_ether_format_addr(mac_str, RTE_ETHER_ADDR_FMT_SIZE,
+                               (struct rte_ether_addr *)hw->mac.mac_addr);
+                       hns3_warn(hw, "default mac_addr from firmware is an invalid "
+                                 "unicast address, using random MAC address %s",
+                                 mac_str);
+               }
+       } else {
+               /*
+                * The hns3 PF ethdev driver in kernel support setting VF MAC
+                * address on the host by "ip link set ..." command. To avoid
+                * some incorrect scenes, for example, hns3 VF PMD driver fails
+                * to receive and send packets after user configure the MAC
+                * address by using the "ip link set ..." command, hns3 VF PMD
+                * driver keep the same MAC address strategy as the hns3 kernel
+                * ethdev driver in the initialization. If user configure a MAC
+                * address by the ip command for VF device, then hns3 VF PMD
+                * driver will start with it, otherwise start with a random MAC
+                * address in the initialization.
+                */
+               if (rte_is_zero_ether_addr(eth_addr))
+                       rte_eth_random_addr(hw->mac.mac_addr);
+       }
+
+       rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.mac_addr,
+                           &dev->data->mac_addrs[0]);
+
+       return 0;
+}
+
+int
+hns3_init_ring_with_vector(struct hns3_hw *hw)
+{
+       uint16_t vec;
+       int ret;
+       int i;
+
+       /*
+        * In hns3 network engine, vector 0 is always the misc interrupt of this
+        * function, vector 1~N can be used respectively for the queues of the
+        * function. Tx and Rx queues with the same number share the interrupt
+        * vector. In the initialization clearing the all hardware mapping
+        * relationship configurations between queues and interrupt vectors is
+        * needed, so some error caused by the residual configurations, such as
+        * the unexpected Tx interrupt, can be avoid.
+        */
+       vec = hw->num_msi - 1; /* vector 0 for misc interrupt, not for queue */
+       if (hw->intr.mapping_mode == HNS3_INTR_MAPPING_VEC_RSV_ONE)
+               vec = vec - 1; /* the last interrupt is reserved */
+       hw->intr_tqps_num = RTE_MIN(vec, hw->tqps_num);
+       for (i = 0; i < hw->intr_tqps_num; i++) {
+               /*
+                * Set gap limiter/rate limiter/quantity limiter algorithm
+                * configuration for interrupt coalesce of queue's interrupt.
+                */
+               hns3_set_queue_intr_gl(hw, i, HNS3_RING_GL_RX,
+                                      HNS3_TQP_INTR_GL_DEFAULT);
+               hns3_set_queue_intr_gl(hw, i, HNS3_RING_GL_TX,
+                                      HNS3_TQP_INTR_GL_DEFAULT);
+               hns3_set_queue_intr_rl(hw, i, HNS3_TQP_INTR_RL_DEFAULT);
+               /*
+                * QL(quantity limiter) is not used currently, just set 0 to
+                * close it.
+                */
+               hns3_set_queue_intr_ql(hw, i, HNS3_TQP_INTR_QL_DEFAULT);
+
+               ret = hw->ops.bind_ring_with_vector(hw, vec, false,
+                                                   HNS3_RING_TYPE_TX, i);
+               if (ret) {
+                       PMD_INIT_LOG(ERR, "fail to unbind TX ring(%d) with "
+                                         "vector: %u, ret=%d", i, vec, ret);
+                       return ret;
+               }
+
+               ret = hw->ops.bind_ring_with_vector(hw, vec, false,
+                                                   HNS3_RING_TYPE_RX, i);
+               if (ret) {
+                       PMD_INIT_LOG(ERR, "fail to unbind RX ring(%d) with "
+                                         "vector: %u, ret=%d", i, vec, ret);
+                       return ret;
+               }
+       }
+
+       return 0;
+}
+
+int
+hns3_map_rx_interrupt(struct rte_eth_dev *dev)
+{
+       struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+       struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
+       struct hns3_hw *hw = HNS3_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+       uint16_t base = RTE_INTR_VEC_ZERO_OFFSET;
+       uint16_t vec = RTE_INTR_VEC_ZERO_OFFSET;
+       uint32_t intr_vector;
+       uint16_t q_id;
+       int ret;
+
+       /*
+        * hns3 needs a separate interrupt to be used as event interrupt which
+        * could not be shared with task queue pair, so KERNEL drivers need
+        * support multiple interrupt vectors.
+        */
+       if (dev->data->dev_conf.intr_conf.rxq == 0 ||
+           !rte_intr_cap_multiple(intr_handle))
+               return 0;
+
+       rte_intr_disable(intr_handle);
+       intr_vector = hw->used_rx_queues;
+       /* creates event fd for each intr vector when MSIX is used */
+       if (rte_intr_efd_enable(intr_handle, intr_vector))
+               return -EINVAL;
+
+       /* Allocate vector list */
+       if (rte_intr_vec_list_alloc(intr_handle, "intr_vec",
+                                   hw->used_rx_queues)) {
+               hns3_err(hw, "failed to allocate %u rx_queues intr_vec",
+                        hw->used_rx_queues);
+               ret = -ENOMEM;
+               goto alloc_intr_vec_error;
+       }
+
+       if (rte_intr_allow_others(intr_handle)) {
+               vec = RTE_INTR_VEC_RXTX_OFFSET;
+               base = RTE_INTR_VEC_RXTX_OFFSET;
+       }
+
+       for (q_id = 0; q_id < hw->used_rx_queues; q_id++) {
+               ret = hw->ops.bind_ring_with_vector(hw, vec, true,
+                                                   HNS3_RING_TYPE_RX, q_id);
+               if (ret)
+                       goto bind_vector_error;
+
+               if (rte_intr_vec_list_index_set(intr_handle, q_id, vec))
+                       goto bind_vector_error;
+               /*
+                * If there are not enough efds (e.g. not enough interrupt),
+                * remaining queues will be bond to the last interrupt.
+                */
+               if (vec < base + rte_intr_nb_efd_get(intr_handle) - 1)
+                       vec++;
+       }
+       rte_intr_enable(intr_handle);
+       return 0;
+
+bind_vector_error:
+       rte_intr_vec_list_free(intr_handle);
+alloc_intr_vec_error:
+       rte_intr_efd_disable(intr_handle);
+       return ret;
+}
+
+void
+hns3_unmap_rx_interrupt(struct rte_eth_dev *dev)
+{
+       struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+       struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
+       struct hns3_adapter *hns = dev->data->dev_private;
+       struct hns3_hw *hw = &hns->hw;
+       uint8_t base = RTE_INTR_VEC_ZERO_OFFSET;
+       uint8_t vec = RTE_INTR_VEC_ZERO_OFFSET;
+       uint16_t q_id;
+
+       if (dev->data->dev_conf.intr_conf.rxq == 0)
+               return;
+
+       /* unmap the ring with vector */
+       if (rte_intr_allow_others(intr_handle)) {
+               vec = RTE_INTR_VEC_RXTX_OFFSET;
+               base = RTE_INTR_VEC_RXTX_OFFSET;
+       }
+       if (rte_intr_dp_is_en(intr_handle)) {
+               for (q_id = 0; q_id < hw->used_rx_queues; q_id++) {
+                       (void)hw->ops.bind_ring_with_vector(hw, vec, false,
+                                                       HNS3_RING_TYPE_RX,
+                                                       q_id);
+                       if (vec < base + rte_intr_nb_efd_get(intr_handle) - 1)
+                               vec++;
+               }
+       }
+       /* Clean datapath event and queue/vec mapping */
+       rte_intr_efd_disable(intr_handle);
+       rte_intr_vec_list_free(intr_handle);
+}
+
+int
+hns3_restore_rx_interrupt(struct hns3_hw *hw)
+{
+       struct rte_eth_dev *dev = &rte_eth_devices[hw->data->port_id];
+       struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+       struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
+       uint16_t q_id;
+       int ret;
+
+       if (dev->data->dev_conf.intr_conf.rxq == 0)
+               return 0;
+
+       if (rte_intr_dp_is_en(intr_handle)) {
+               for (q_id = 0; q_id < hw->used_rx_queues; q_id++) {
+                       ret = hw->ops.bind_ring_with_vector(hw,
+                               rte_intr_vec_list_index_get(intr_handle,
+                                                                  q_id),
+                               true, HNS3_RING_TYPE_RX, q_id);
+                       if (ret)
+                               return ret;
+               }
+       }
+
+       return 0;
+}
+
+int
+hns3_get_pci_revision_id(struct hns3_hw *hw, uint8_t *revision_id)
+{
+       struct rte_pci_device *pci_dev;
+       struct rte_eth_dev *eth_dev;
+       uint8_t revision;
+       int ret;
+
+       eth_dev = &rte_eth_devices[hw->data->port_id];
+       pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
+       ret = rte_pci_read_config(pci_dev, &revision, HNS3_PCI_REVISION_ID_LEN,
+                                 HNS3_PCI_REVISION_ID);
+       if (ret != HNS3_PCI_REVISION_ID_LEN) {
+               hns3_err(hw, "failed to read pci revision id, ret = %d", ret);
+               return -EIO;
+       }
+
+       *revision_id = revision;
+
+       return 0;
+}