net/mlx5: fix RSS expansion for patterns with ICMP item
[dpdk.git] / drivers / net / iavf / iavf.h
index fe25d80..025ab3f 100644 (file)
@@ -5,7 +5,11 @@
 #ifndef _IAVF_ETHDEV_H_
 #define _IAVF_ETHDEV_H_
 
+#include <sys/queue.h>
+
 #include <rte_kvargs.h>
+#include <rte_tm_driver.h>
+
 #include <iavf_prototype.h>
 #include <iavf_adminq_cmd.h>
 #include <iavf_type.h>
 
 #define IAVF_AQ_LEN               32
 #define IAVF_AQ_BUF_SZ            4096
-#define IAVF_RESET_WAIT_CNT       50
+#define IAVF_RESET_WAIT_CNT       500
 #define IAVF_BUF_SIZE_MIN         1024
 #define IAVF_FRAME_SIZE_MAX       9728
 #define IAVF_QUEUE_BASE_ADDR_UNIT 128
 
-#define IAVF_MAX_NUM_QUEUES       16
+#define IAVF_MAX_NUM_QUEUES_DFLT        16
+#define IAVF_MAX_NUM_QUEUES_LV          256
+#define IAVF_CFG_Q_NUM_PER_BUF          32
+#define IAVF_IRQ_MAP_NUM_PER_BUF        128
+#define IAVF_RXTX_QUEUE_CHUNKS_NUM      2
 
 #define IAVF_NUM_MACADDR_MAX      64
 
+#define IAVF_DEV_WATCHDOG_PERIOD     0
+
 #define IAVF_DEFAULT_RX_PTHRESH      8
 #define IAVF_DEFAULT_RX_HTHRESH      8
 #define IAVF_DEFAULT_RX_WTHRESH      0
        VIRTCHNL_VF_OFFLOAD_RX_POLLING)
 
 #define IAVF_RSS_OFFLOAD_ALL ( \
-       ETH_RSS_FRAG_IPV4 |         \
-       ETH_RSS_NONFRAG_IPV4_TCP |  \
-       ETH_RSS_NONFRAG_IPV4_UDP |  \
-       ETH_RSS_NONFRAG_IPV4_SCTP | \
-       ETH_RSS_NONFRAG_IPV4_OTHER)
+       RTE_ETH_RSS_IPV4 | \
+       RTE_ETH_RSS_FRAG_IPV4 |         \
+       RTE_ETH_RSS_NONFRAG_IPV4_TCP |  \
+       RTE_ETH_RSS_NONFRAG_IPV4_UDP |  \
+       RTE_ETH_RSS_NONFRAG_IPV4_SCTP | \
+       RTE_ETH_RSS_NONFRAG_IPV4_OTHER | \
+       RTE_ETH_RSS_IPV6 | \
+       RTE_ETH_RSS_FRAG_IPV6 | \
+       RTE_ETH_RSS_NONFRAG_IPV6_TCP | \
+       RTE_ETH_RSS_NONFRAG_IPV6_UDP | \
+       RTE_ETH_RSS_NONFRAG_IPV6_SCTP | \
+       RTE_ETH_RSS_NONFRAG_IPV6_OTHER)
 
 #define IAVF_MISC_VEC_ID                RTE_INTR_VEC_ZERO_OFFSET
 #define IAVF_RX_VEC_START               RTE_INTR_VEC_RXTX_OFFSET
 #define IAVF_QUEUE_ITR_INTERVAL_DEFAULT 32 /* 32 us */
 #define IAVF_QUEUE_ITR_INTERVAL_MAX     8160 /* 8160 us */
 
+#define IAVF_ALARM_INTERVAL 50000 /* us */
+
 /* The overhead from MTU to max frame size.
  * Considering QinQ packet, the VLAN tag needs to be counted twice.
  */
-#define IAVF_VLAN_TAG_SIZE               4
 #define IAVF_ETH_OVERHEAD \
-       (RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN + IAVF_VLAN_TAG_SIZE * 2)
+       (RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN + RTE_VLAN_HLEN * 2)
+#define IAVF_ETH_MAX_LEN (RTE_ETHER_MTU + IAVF_ETH_OVERHEAD)
 
 #define IAVF_32_BIT_WIDTH (CHAR_BIT * 4)
 #define IAVF_48_BIT_WIDTH (CHAR_BIT * 6)
 #define IAVF_48_BIT_MASK  RTE_LEN2MASK(IAVF_48_BIT_WIDTH, uint64_t)
 
+#define IAVF_RX_DESC_EXT_STATUS_FLEXBH_MASK  0x03
+#define IAVF_RX_DESC_EXT_STATUS_FLEXBH_FD_ID 0x01
+
+#define IAVF_BITS_PER_BYTE 8
+
+#define IAVF_VLAN_TAG_PCP_OFFSET 13
+
+#define IAVF_L2TPV2_FLAGS_LEN  0x4000
+
 struct iavf_adapter;
 struct iavf_rx_queue;
 struct iavf_tx_queue;
 
+
+struct iavf_ipsec_crypto_stats {
+       uint64_t icount;
+       uint64_t ibytes;
+       struct {
+               uint64_t count;
+               uint64_t sad_miss;
+               uint64_t not_processed;
+               uint64_t icv_check;
+               uint64_t ipsec_length;
+               uint64_t misc;
+       } ierrors;
+};
+
+struct iavf_eth_xstats {
+       struct virtchnl_eth_stats eth_stats;
+       struct iavf_ipsec_crypto_stats ips_stats;
+};
+
 /* Structure that defines a VSI, associated with a adapter. */
 struct iavf_vsi {
        struct iavf_adapter *adapter; /* Backreference to associated adapter */
@@ -80,11 +127,89 @@ struct iavf_vsi {
        uint16_t max_macaddrs;   /* Maximum number of MAC addresses */
        uint16_t base_vector;
        uint16_t msix_intr;      /* The MSIX interrupt binds to VSI */
-       struct virtchnl_eth_stats eth_stats_offset;
+       struct iavf_eth_xstats eth_stats_offset;
+};
+
+struct rte_flow;
+TAILQ_HEAD(iavf_flow_list, rte_flow);
+
+struct iavf_flow_parser_node;
+TAILQ_HEAD(iavf_parser_list, iavf_flow_parser_node);
+
+struct iavf_fdir_conf {
+       struct virtchnl_fdir_add add_fltr;
+       struct virtchnl_fdir_del del_fltr;
+       uint64_t input_set;
+       uint32_t flow_id;
+       uint32_t mark_flag;
+};
+
+struct iavf_fdir_info {
+       struct iavf_fdir_conf conf;
+};
+
+struct iavf_qv_map {
+       uint16_t queue_id;
+       uint16_t vector_id;
+};
+
+/* Message type read in admin queue from PF */
+enum iavf_aq_result {
+       IAVF_MSG_ERR = -1, /* Meet error when accessing admin queue */
+       IAVF_MSG_NON,      /* Read nothing from admin queue */
+       IAVF_MSG_SYS,      /* Read system msg from admin queue */
+       IAVF_MSG_CMD,      /* Read async command result */
+};
+
+/* Struct to store Traffic Manager node configuration. */
+struct iavf_tm_node {
+       TAILQ_ENTRY(iavf_tm_node) node;
+       uint32_t id;
+       uint32_t tc;
+       uint32_t priority;
+       uint32_t weight;
+       uint32_t reference_count;
+       struct iavf_tm_node *parent;
+       struct iavf_tm_shaper_profile *shaper_profile;
+       struct rte_tm_node_params params;
+};
+
+TAILQ_HEAD(iavf_tm_node_list, iavf_tm_node);
+
+struct iavf_tm_shaper_profile {
+       TAILQ_ENTRY(iavf_tm_shaper_profile) node;
+       uint32_t shaper_profile_id;
+       uint32_t reference_count;
+       struct rte_tm_shaper_params profile;
+};
+
+TAILQ_HEAD(iavf_shaper_profile_list, iavf_tm_shaper_profile);
+
+/* node type of Traffic Manager */
+enum iavf_tm_node_type {
+       IAVF_TM_NODE_TYPE_PORT,
+       IAVF_TM_NODE_TYPE_TC,
+       IAVF_TM_NODE_TYPE_QUEUE,
+       IAVF_TM_NODE_TYPE_MAX,
+};
+
+/* Struct to store all the Traffic Manager configuration. */
+struct iavf_tm_conf {
+       struct iavf_tm_node *root; /* root node - vf vsi */
+       struct iavf_tm_node_list tc_list; /* node list for all the TCs */
+       struct iavf_tm_node_list queue_list; /* node list for all the queues */
+       struct iavf_shaper_profile_list shaper_profile_list;
+       uint32_t nb_tc_node;
+       uint32_t nb_queue_node;
+       bool committed;
 };
 
-/* TODO: is that correct to assume the max number to be 16 ?*/
-#define IAVF_MAX_MSIX_VECTORS   16
+/* Struct to store queue TC mapping. Queue is continuous in one TC */
+struct iavf_qtc_map {
+       uint8_t tc;
+       uint16_t start_queue_id;
+       uint16_t queue_count;
+};
 
 /* Structure to store private data specific for VF instance. */
 struct iavf_info {
@@ -97,41 +222,100 @@ struct iavf_info {
        struct virtchnl_version_info virtchnl_version;
        struct virtchnl_vf_resource *vf_res; /* VF resource */
        struct virtchnl_vsi_resource *vsi_res; /* LAN VSI */
-
+       struct virtchnl_vlan_caps vlan_v2_caps;
+       uint64_t supported_rxdid;
+       uint8_t *proto_xtr; /* proto xtr type for all queues */
        volatile enum virtchnl_ops pend_cmd; /* pending command not finished */
-       uint32_t cmd_retval; /* return value of the cmd response from PF */
+       uint32_t pend_cmd_count;
+       int cmd_retval; /* return value of the cmd response from PF */
        uint8_t *aq_resp; /* buffer to store the adminq response from PF */
 
+       /** iAVF watchdog enable */
+       bool watchdog_enabled;
+
        /* Event from pf */
        bool dev_closed;
        bool link_up;
        uint32_t link_speed;
 
+       /* Multicast addrs */
+       struct rte_ether_addr mc_addrs[IAVF_NUM_MACADDR_MAX];
+       uint16_t mc_addrs_num;   /* Multicast mac addresses number */
+
        struct iavf_vsi vsi;
-       bool vf_reset;
+       bool vf_reset;  /* true for VF reset pending, false for no VF reset */
        uint64_t flags;
 
        uint8_t *rss_lut;
        uint8_t *rss_key;
+       uint64_t rss_hf;
        uint16_t nb_msix;   /* number of MSI-X interrupts on Rx */
        uint16_t msix_base; /* msix vector base from */
-       /* queue bitmask for each vector */
-       uint16_t rxq_map[IAVF_MAX_MSIX_VECTORS];
+       uint16_t max_rss_qregion; /* max RSS queue region supported by PF */
+       struct iavf_qv_map *qv_map; /* queue vector mapping */
+       struct iavf_flow_list flow_list;
+       rte_spinlock_t flow_ops_lock;
+       struct iavf_parser_list rss_parser_list;
+       struct iavf_parser_list dist_parser_list;
+       struct iavf_parser_list ipsec_crypto_parser_list;
+
+       struct iavf_fdir_info fdir; /* flow director info */
+       /* indicate large VF support enabled or not */
+       bool lv_enabled;
+
+       struct virtchnl_qos_cap_list *qos_cap;
+       struct iavf_qtc_map *qtc_map;
+       struct iavf_tm_conf tm_conf;
+
+       struct rte_eth_dev *eth_dev;
+
+       uint32_t ptp_caps;
+       rte_spinlock_t phc_time_aq_lock;
+};
+
+#define IAVF_MAX_PKT_TYPE 1024
+
+#define IAVF_MAX_QUEUE_NUM  2048
+
+enum iavf_proto_xtr_type {
+       IAVF_PROTO_XTR_NONE,
+       IAVF_PROTO_XTR_VLAN,
+       IAVF_PROTO_XTR_IPV4,
+       IAVF_PROTO_XTR_IPV6,
+       IAVF_PROTO_XTR_IPV6_FLOW,
+       IAVF_PROTO_XTR_TCP,
+       IAVF_PROTO_XTR_IP_OFFSET,
+       IAVF_PROTO_XTR_IPSEC_CRYPTO_SAID,
+       IAVF_PROTO_XTR_MAX,
+};
+
+/**
+ * Cache devargs parse result.
+ */
+struct iavf_devargs {
+       uint8_t proto_xtr_dflt;
+       uint8_t proto_xtr[IAVF_MAX_QUEUE_NUM];
+       uint16_t quanta_size;
 };
 
-#define IAVF_MAX_PKT_TYPE 256
+struct iavf_security_ctx;
 
 /* Structure to store private data for each VF instance. */
 struct iavf_adapter {
        struct iavf_hw hw;
-       struct rte_eth_dev *eth_dev;
+       struct rte_eth_dev_data *dev_data;
        struct iavf_info vf;
+       struct iavf_security_ctx *security_ctx;
 
        bool rx_bulk_alloc_allowed;
        /* For vector PMD */
        bool rx_vec_allowed;
        bool tx_vec_allowed;
+       uint32_t ptype_tbl[IAVF_MAX_PKT_TYPE] __rte_cache_min_aligned;
        bool stopped;
+       bool closed;
+       uint16_t fdir_ref_cnt;
+       struct iavf_devargs devargs;
 };
 
 /* IAVF_DEV_PRIVATE_TO */
@@ -141,14 +325,14 @@ struct iavf_adapter {
        (&((struct iavf_adapter *)adapter)->vf)
 #define IAVF_DEV_PRIVATE_TO_HW(adapter) \
        (&((struct iavf_adapter *)adapter)->hw)
+#define IAVF_DEV_PRIVATE_TO_IAVF_SECURITY_CTX(adapter) \
+       (((struct iavf_adapter *)adapter)->security_ctx)
 
 /* IAVF_VSI_TO */
 #define IAVF_VSI_TO_HW(vsi) \
        (&(((struct iavf_vsi *)vsi)->adapter->hw))
 #define IAVF_VSI_TO_VF(vsi) \
        (&(((struct iavf_vsi *)vsi)->adapter->vf))
-#define IAVF_VSI_TO_ETH_DEV(vsi) \
-       (((struct iavf_vsi *)vsi)->adapter->eth_dev)
 
 static inline void
 iavf_init_adminq_parameter(struct iavf_hw *hw)
@@ -182,7 +366,7 @@ struct iavf_cmd_info {
  * _atomic_set_cmd successfully.
  */
 static inline void
-_notify_cmd(struct iavf_info *vf, uint32_t msg_ret)
+_notify_cmd(struct iavf_info *vf, int msg_ret)
 {
        vf->cmd_retval = msg_ret;
        rte_wmb();
@@ -204,14 +388,33 @@ _clear_cmd(struct iavf_info *vf)
 static inline int
 _atomic_set_cmd(struct iavf_info *vf, enum virtchnl_ops ops)
 {
-       int ret = rte_atomic32_cmpset(&vf->pend_cmd, VIRTCHNL_OP_UNKNOWN, ops);
+       enum virtchnl_ops op_unk = VIRTCHNL_OP_UNKNOWN;
+       int ret = __atomic_compare_exchange(&vf->pend_cmd, &op_unk, &ops,
+                       0, __ATOMIC_ACQUIRE, __ATOMIC_ACQUIRE);
 
        if (!ret)
                PMD_DRV_LOG(ERR, "There is incomplete cmd %d", vf->pend_cmd);
 
+       __atomic_store_n(&vf->pend_cmd_count, 1, __ATOMIC_RELAXED);
+
        return !ret;
 }
 
+/* Check there is pending cmd in execution. If none, set new command. */
+static inline int
+_atomic_set_async_response_cmd(struct iavf_info *vf, enum virtchnl_ops ops)
+{
+       enum virtchnl_ops op_unk = VIRTCHNL_OP_UNKNOWN;
+       int ret = __atomic_compare_exchange(&vf->pend_cmd, &op_unk, &ops,
+                       0, __ATOMIC_ACQUIRE, __ATOMIC_ACQUIRE);
+
+       if (!ret)
+               PMD_DRV_LOG(ERR, "There is incomplete cmd %d", vf->pend_cmd);
+
+       __atomic_store_n(&vf->pend_cmd_count, 2, __ATOMIC_RELAXED);
+
+       return !ret;
+}
 int iavf_check_api_version(struct iavf_adapter *adapter);
 int iavf_get_vf_resource(struct iavf_adapter *adapter);
 void iavf_handle_virtchnl_msg(struct rte_eth_dev *dev);
@@ -219,20 +422,64 @@ int iavf_enable_vlan_strip(struct iavf_adapter *adapter);
 int iavf_disable_vlan_strip(struct iavf_adapter *adapter);
 int iavf_switch_queue(struct iavf_adapter *adapter, uint16_t qid,
                     bool rx, bool on);
+int iavf_switch_queue_lv(struct iavf_adapter *adapter, uint16_t qid,
+                    bool rx, bool on);
 int iavf_enable_queues(struct iavf_adapter *adapter);
+int iavf_enable_queues_lv(struct iavf_adapter *adapter);
 int iavf_disable_queues(struct iavf_adapter *adapter);
+int iavf_disable_queues_lv(struct iavf_adapter *adapter);
 int iavf_configure_rss_lut(struct iavf_adapter *adapter);
 int iavf_configure_rss_key(struct iavf_adapter *adapter);
-int iavf_configure_queues(struct iavf_adapter *adapter);
+int iavf_configure_queues(struct iavf_adapter *adapter,
+                       uint16_t num_queue_pairs, uint16_t index);
+int iavf_get_supported_rxdid(struct iavf_adapter *adapter);
+int iavf_config_vlan_strip_v2(struct iavf_adapter *adapter, bool enable);
+int iavf_config_vlan_insert_v2(struct iavf_adapter *adapter, bool enable);
+int iavf_add_del_vlan_v2(struct iavf_adapter *adapter, uint16_t vlanid,
+                        bool add);
+int iavf_get_vlan_offload_caps_v2(struct iavf_adapter *adapter);
 int iavf_config_irq_map(struct iavf_adapter *adapter);
+int iavf_config_irq_map_lv(struct iavf_adapter *adapter, uint16_t num,
+                       uint16_t index);
 void iavf_add_del_all_mac_addr(struct iavf_adapter *adapter, bool add);
 int iavf_dev_link_update(struct rte_eth_dev *dev,
                        __rte_unused int wait_to_complete);
+void iavf_dev_alarm_handler(void *param);
 int iavf_query_stats(struct iavf_adapter *adapter,
                    struct virtchnl_eth_stats **pstats);
 int iavf_config_promisc(struct iavf_adapter *adapter, bool enable_unicast,
                       bool enable_multicast);
 int iavf_add_del_eth_addr(struct iavf_adapter *adapter,
-                        struct rte_ether_addr *addr, bool add);
+                        struct rte_ether_addr *addr, bool add, uint8_t type);
 int iavf_add_del_vlan(struct iavf_adapter *adapter, uint16_t vlanid, bool add);
+int iavf_fdir_add(struct iavf_adapter *adapter, struct iavf_fdir_conf *filter);
+int iavf_fdir_del(struct iavf_adapter *adapter, struct iavf_fdir_conf *filter);
+int iavf_fdir_check(struct iavf_adapter *adapter,
+               struct iavf_fdir_conf *filter);
+int iavf_add_del_rss_cfg(struct iavf_adapter *adapter,
+                        struct virtchnl_rss_cfg *rss_cfg, bool add);
+int iavf_get_hena_caps(struct iavf_adapter *adapter, uint64_t *caps);
+int iavf_set_hena(struct iavf_adapter *adapter, uint64_t hena);
+int iavf_rss_hash_set(struct iavf_adapter *ad, uint64_t rss_hf, bool add);
+int iavf_add_del_mc_addr_list(struct iavf_adapter *adapter,
+                       struct rte_ether_addr *mc_addrs,
+                       uint32_t mc_addrs_num, bool add);
+int iavf_request_queues(struct rte_eth_dev *dev, uint16_t num);
+int iavf_get_max_rss_queue_region(struct iavf_adapter *adapter);
+int iavf_get_qos_cap(struct iavf_adapter *adapter);
+int iavf_set_q_bw(struct rte_eth_dev *dev,
+                 struct virtchnl_queues_bw_cfg *q_bw, uint16_t size);
+int iavf_set_q_tc_map(struct rte_eth_dev *dev,
+                       struct virtchnl_queue_tc_mapping *q_tc_mapping,
+                       uint16_t size);
+int iavf_set_vf_quanta_size(struct iavf_adapter *adapter, u16 start_queue_id,
+                           u16 num_queues);
+void iavf_tm_conf_init(struct rte_eth_dev *dev);
+void iavf_tm_conf_uninit(struct rte_eth_dev *dev);
+int iavf_ipsec_crypto_request(struct iavf_adapter *adapter,
+               uint8_t *msg, size_t msg_len,
+               uint8_t *resp_msg, size_t resp_msg_len);
+extern const struct rte_tm_ops iavf_tm_ops;
+int iavf_get_ptp_cap(struct iavf_adapter *adapter);
+int iavf_get_phc_time(struct iavf_rx_queue *rxq);
 #endif /* _IAVF_ETHDEV_H_ */