diff options
Diffstat (limited to 'drivers/net/ethernet/netronome/nfp/nfp_net.h')
| -rw-r--r-- | drivers/net/ethernet/netronome/nfp/nfp_net.h | 302 |
1 files changed, 174 insertions, 128 deletions
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_net.h b/drivers/net/ethernet/netronome/nfp/nfp_net.h index 0b1865e9f0b5..46764aeccb37 100644 --- a/drivers/net/ethernet/netronome/nfp/nfp_net.h +++ b/drivers/net/ethernet/netronome/nfp/nfp_net.h @@ -63,9 +63,6 @@ #define NFP_NET_Q0_BAR 2 #define NFP_NET_Q1_BAR 4 /* OBSOLETE */ -/* Max bits in DMA address */ -#define NFP_NET_MAX_DMA_BITS 40 - /* Default size for MTU and freelist buffer sizes */ #define NFP_NET_DEFAULT_MTU 1500U @@ -85,17 +82,15 @@ NFP_NET_MAX_TX_RINGS : NFP_NET_MAX_RX_RINGS) #define NFP_NET_MAX_IRQS (NFP_NET_NON_Q_VECTORS + NFP_NET_MAX_R_VECS) -#define NFP_NET_MIN_TX_DESCS 256 /* Min. # of Tx descs per ring */ -#define NFP_NET_MIN_RX_DESCS 256 /* Min. # of Rx descs per ring */ -#define NFP_NET_MAX_TX_DESCS (256 * 1024) /* Max. # of Tx descs per ring */ -#define NFP_NET_MAX_RX_DESCS (256 * 1024) /* Max. # of Rx descs per ring */ - #define NFP_NET_TX_DESCS_DEFAULT 4096 /* Default # of Tx descs per ring */ #define NFP_NET_RX_DESCS_DEFAULT 4096 /* Default # of Rx descs per ring */ #define NFP_NET_FL_BATCH 16 /* Add freelist in this Batch size */ #define NFP_NET_XDP_MAX_COMPLETE 2048 /* XDP bufs to reclaim in NAPI poll */ +/* MC definitions */ +#define NFP_NET_CFG_MAC_MC_MAX 1024 /* The maximum number of MC address per port*/ + /* Offload definitions */ #define NFP_NET_N_VXLAN_PORTS (NFP_NET_CFG_VXLAN_SZ / sizeof(__be16)) @@ -105,104 +100,61 @@ /* Forward declarations */ struct nfp_cpp; +struct nfp_dev_info; +struct nfp_dp_ops; struct nfp_eth_table_port; struct nfp_net; struct nfp_net_r_vector; struct nfp_port; +struct xsk_buff_pool; + +struct nfp_nfd3_tx_desc; +struct nfp_nfd3_tx_buf; + +struct nfp_nfdk_tx_desc; +struct nfp_nfdk_tx_buf; /* Convenience macro for wrapping descriptor index on ring size */ #define D_IDX(ring, idx) ((idx) & ((ring)->cnt - 1)) /* Convenience macro for writing dma address into RX/TX descriptors */ -#define nfp_desc_set_dma_addr(desc, dma_addr) \ +#define nfp_desc_set_dma_addr_40b(desc, dma_addr) \ do { \ - __typeof(desc) __d = (desc); \ + __typeof__(desc) __d = (desc); \ dma_addr_t __addr = (dma_addr); \ \ __d->dma_addr_lo = cpu_to_le32(lower_32_bits(__addr)); \ __d->dma_addr_hi = upper_32_bits(__addr) & 0xff; \ } while (0) -/* TX descriptor format */ - -#define PCIE_DESC_TX_EOP BIT(7) -#define PCIE_DESC_TX_OFFSET_MASK GENMASK(6, 0) -#define PCIE_DESC_TX_MSS_MASK GENMASK(13, 0) - -/* Flags in the host TX descriptor */ -#define PCIE_DESC_TX_CSUM BIT(7) -#define PCIE_DESC_TX_IP4_CSUM BIT(6) -#define PCIE_DESC_TX_TCP_CSUM BIT(5) -#define PCIE_DESC_TX_UDP_CSUM BIT(4) -#define PCIE_DESC_TX_VLAN BIT(3) -#define PCIE_DESC_TX_LSO BIT(2) -#define PCIE_DESC_TX_ENCAP BIT(1) -#define PCIE_DESC_TX_O_IP4_CSUM BIT(0) - -struct nfp_net_tx_desc { - union { - struct { - u8 dma_addr_hi; /* High bits of host buf address */ - __le16 dma_len; /* Length to DMA for this desc */ - u8 offset_eop; /* Offset in buf where pkt starts + - * highest bit is eop flag. - */ - __le32 dma_addr_lo; /* Low 32bit of host buf addr */ - - __le16 mss; /* MSS to be used for LSO */ - u8 lso_hdrlen; /* LSO, TCP payload offset */ - u8 flags; /* TX Flags, see @PCIE_DESC_TX_* */ - union { - struct { - u8 l3_offset; /* L3 header offset */ - u8 l4_offset; /* L4 header offset */ - }; - __le16 vlan; /* VLAN tag to add if indicated */ - }; - __le16 data_len; /* Length of frame + meta data */ - } __packed; - __le32 vals[4]; - __le64 vals8[2]; - }; -}; - -/** - * struct nfp_net_tx_buf - software TX buffer descriptor - * @skb: normal ring, sk_buff associated with this buffer - * @frag: XDP ring, page frag associated with this buffer - * @dma_addr: DMA mapping address of the buffer - * @fidx: Fragment index (-1 for the head and [0..nr_frags-1] for frags) - * @pkt_cnt: Number of packets to be produced out of the skb associated - * with this buffer (valid only on the head's buffer). - * Will be 1 for all non-TSO packets. - * @real_len: Number of bytes which to be produced out of the skb (valid only - * on the head's buffer). Equal to skb->len for non-TSO packets. - */ -struct nfp_net_tx_buf { - union { - struct sk_buff *skb; - void *frag; - }; - dma_addr_t dma_addr; - short int fidx; - u16 pkt_cnt; - u32 real_len; -}; +#define nfp_desc_set_dma_addr_48b(desc, dma_addr) \ + do { \ + __typeof__(desc) __d = (desc); \ + dma_addr_t __addr = (dma_addr); \ + \ + __d->dma_addr_hi = cpu_to_le16(upper_32_bits(__addr)); \ + __d->dma_addr_lo = cpu_to_le32(lower_32_bits(__addr)); \ + } while (0) /** * struct nfp_net_tx_ring - TX ring structure * @r_vec: Back pointer to ring vector structure * @idx: Ring index from Linux's perspective - * @qcidx: Queue Controller Peripheral (QCP) queue index for the TX queue + * @data_pending: number of bytes added to current block (NFDK only) * @qcp_q: Pointer to base of the QCP TX queue + * @txrwb: TX pointer write back area * @cnt: Size of the queue in number of descriptors * @wr_p: TX ring write pointer (free running) * @rd_p: TX ring read pointer (free running) * @qcp_rd_p: Local copy of QCP TX queue read pointer * @wr_ptr_add: Accumulated number of buffers to add to QCP write pointer * (used for .xmit_more delayed kick) - * @txbufs: Array of transmitted TX buffers, to free on transmit - * @txds: Virtual address of TX ring in host memory + * @txbufs: Array of transmitted TX buffers, to free on transmit (NFD3) + * @ktxbufs: Array of transmitted TX buffers, to free on transmit (NFDK) + * @txds: Virtual address of TX ring in host memory (NFD3) + * @ktxds: Virtual address of TX ring in host memory (NFDK) + * + * @qcidx: Queue Controller Peripheral (QCP) queue index for the TX queue * @dma: DMA address of the TX ring * @size: Size, in bytes, of the TX ring (needed to free) * @is_xdp: Is this a XDP TX ring? @@ -210,9 +162,10 @@ struct nfp_net_tx_buf { struct nfp_net_tx_ring { struct nfp_net_r_vector *r_vec; - u32 idx; - int qcidx; + u16 idx; + u16 data_pending; u8 __iomem *qcp_q; + u64 *txrwb; u32 cnt; u32 wr_p; @@ -221,8 +174,17 @@ struct nfp_net_tx_ring { u32 wr_ptr_add; - struct nfp_net_tx_buf *txbufs; - struct nfp_net_tx_desc *txds; + union { + struct nfp_nfd3_tx_buf *txbufs; + struct nfp_nfdk_tx_buf *ktxbufs; + }; + union { + struct nfp_nfd3_tx_desc *txds; + struct nfp_nfdk_tx_desc *ktxds; + }; + + /* Cold data follows */ + int qcidx; dma_addr_t dma; size_t size; @@ -266,8 +228,8 @@ struct nfp_net_tx_ring { struct nfp_net_rx_desc { union { struct { - u8 dma_addr_hi; /* High bits of the buf address */ - __le16 reserved; /* Must be zero */ + __le16 dma_addr_hi; /* High bits of the buf address */ + u8 reserved; /* Must be zero */ u8 meta_len_dd; /* Must be zero */ __le32 dma_addr_lo; /* Low bits of the buffer address */ @@ -289,6 +251,8 @@ struct nfp_net_rx_desc { }; #define NFP_NET_META_FIELD_MASK GENMASK(NFP_NET_META_FIELD_SIZE - 1, 0) +#define NFP_NET_VLAN_CTAG 0 +#define NFP_NET_VLAN_STAG 1 struct nfp_meta_parsed { u8 hash_type; @@ -297,6 +261,15 @@ struct nfp_meta_parsed { u32 mark; u32 portid; __wsum csum; + struct { + bool stripped; + u8 tpid; + u16 tci; + } vlan; + +#ifdef CONFIG_NFP_NET_IPSEC + u32 ipsec_saidx; +#endif }; struct nfp_net_rx_hash { @@ -315,6 +288,16 @@ struct nfp_net_rx_buf { }; /** + * struct nfp_net_xsk_rx_buf - software RX XSK buffer descriptor + * @dma_addr: DMA mapping address of the buffer + * @xdp: XSK buffer pool handle (for AF_XDP) + */ +struct nfp_net_xsk_rx_buf { + dma_addr_t dma_addr; + struct xdp_buff *xdp; +}; + +/** * struct nfp_net_rx_ring - RX ring structure * @r_vec: Back pointer to ring vector structure * @cnt: Size of the queue in number of descriptors @@ -324,6 +307,7 @@ struct nfp_net_rx_buf { * @fl_qcidx: Queue Controller Peripheral (QCP) queue index for the freelist * @qcp_fl: Pointer to base of the QCP freelist queue * @rxbufs: Array of transmitted FL/RX buffers + * @xsk_rxbufs: Array of transmitted FL/RX buffers (for AF_XDP) * @rxds: Virtual address of FL/RX ring in host memory * @xdp_rxq: RX-ring info avail for XDP * @dma: DMA address of the FL/RX ring @@ -342,6 +326,7 @@ struct nfp_net_rx_ring { u8 __iomem *qcp_fl; struct nfp_net_rx_buf *rxbufs; + struct nfp_net_xsk_rx_buf *xsk_rxbufs; struct nfp_net_rx_desc *rxds; struct xdp_rxq_info xdp_rxq; @@ -360,6 +345,7 @@ struct nfp_net_rx_ring { * @tx_ring: Pointer to TX ring * @rx_ring: Pointer to RX ring * @xdp_ring: Pointer to an extra TX ring for XDP + * @xsk_pool: XSK buffer pool active on vector queue pair (for AF_XDP) * @irq_entry: MSI-X table entry (use for talking to the device) * @event_ctr: Number of interrupt * @rx_dim: Dynamic interrupt moderation structure for RX @@ -431,6 +417,7 @@ struct nfp_net_r_vector { u64 rx_replace_buf_alloc_fail; struct nfp_net_tx_ring *xdp_ring; + struct xsk_buff_pool *xsk_pool; struct u64_stats_sync tx_sync; u64 tx_pkts; @@ -460,13 +447,17 @@ struct nfp_net_fw_version { u8 minor; u8 major; u8 class; - u8 resv; + + /* This byte can be exploited for more use, currently, + * BIT0: dp type, BIT[7:1]: reserved + */ + u8 extend; } __packed; static inline bool nfp_net_fw_ver_eq(struct nfp_net_fw_version *fw_ver, - u8 resv, u8 class, u8 major, u8 minor) + u8 extend, u8 class, u8 major, u8 minor) { - return fw_ver->resv == resv && + return fw_ver->extend == extend && fw_ver->class == class && fw_ver->major == major && fw_ver->minor == minor; @@ -488,19 +479,24 @@ struct nfp_stat_pair { * @rx_dma_off: Offset at which DMA packets (for XDP headroom) * @rx_offset: Offset in the RX buffers where packet data starts * @ctrl: Local copy of the control register/word. + * @ctrl_w1: Local copy of the control register/word1. * @fl_bufsz: Currently configured size of the freelist buffers * @xdp_prog: Installed XDP program * @tx_rings: Array of pre-allocated TX ring structures * @rx_rings: Array of pre-allocated RX ring structures * @ctrl_bar: Pointer to mapped control BAR * - * @txd_cnt: Size of the TX ring in number of descriptors - * @rxd_cnt: Size of the RX ring in number of descriptors + * @ops: Callbacks and parameters for this vNIC's NFD version + * @txrwb: TX pointer write back area (indexed by queue id) + * @txrwb_dma: TX pointer write back area DMA address + * @txd_cnt: Size of the TX ring in number of min size packets + * @rxd_cnt: Size of the RX ring in number of min size packets * @num_r_vecs: Number of used ring vectors * @num_tx_rings: Currently configured number of TX rings * @num_stack_tx_rings: Number of TX rings used by the stack (not XDP) * @num_rx_rings: Currently configured number of RX rings * @mtu: Device MTU + * @xsk_pools: XSK buffer pools, @max_r_vecs in size (for AF_XDP). */ struct nfp_net_dp { struct device *dev; @@ -516,6 +512,7 @@ struct nfp_net_dp { u32 rx_dma_off; u32 ctrl; + u32 ctrl_w1; u32 fl_bufsz; struct bpf_prog *xdp_prog; @@ -527,6 +524,11 @@ struct nfp_net_dp { /* Cold data follows */ + const struct nfp_dp_ops *ops; + + u64 *txrwb; + dma_addr_t txrwb_dma; + unsigned int txd_cnt; unsigned int rxd_cnt; @@ -537,14 +539,18 @@ struct nfp_net_dp { unsigned int num_rx_rings; unsigned int mtu; + + struct xsk_buff_pool **xsk_pools; }; /** * struct nfp_net - NFP network device structure * @dp: Datapath structure + * @dev_info: NFP ASIC params * @id: vNIC id within the PF (0 for VFs) * @fw_ver: Firmware version * @cap: Capabilities advertised by the Firmware + * @cap_w1: Extended capabilities word advertised by the Firmware * @max_mtu: Maximum support MTU advertised by the Firmware * @rss_hfunc: RSS selected hash function * @rss_cfg: RSS configuration @@ -587,6 +593,7 @@ struct nfp_net_dp { * @qcp_cfg: Pointer to QCP queue used for configuration notification * @tx_bar: Pointer to mapped TX queues * @rx_bar: Pointer to mapped FL/RX queues + * @xa_ipsec: IPsec xarray SA data * @tlv_caps: Parsed TLV capabilities * @ktls_tx_conn_cnt: Number of offloaded kTLS TX connections * @ktls_rx_conn_cnt: Number of offloaded kTLS RX connections @@ -610,16 +617,25 @@ struct nfp_net_dp { * @vnic_no_name: For non-port PF vNIC make ndo_get_phys_port_name return * -EOPNOTSUPP to keep backwards compatibility (set by app) * @port: Pointer to nfp_port structure if vNIC is a port + * @mbox_amsg: Asynchronously processed message via mailbox + * @mbox_amsg.lock: Protect message list + * @mbox_amsg.list: List of message to process + * @mbox_amsg.work: Work to process message asynchronously + * @fs: Flow steering + * @fs.count: Flow count + * @fs.list: List of flows * @app_priv: APP private data for this vNIC */ struct nfp_net { struct nfp_net_dp dp; + const struct nfp_dev_info *dev_info; struct nfp_net_fw_version fw_ver; u32 id; u32 cap; + u32 cap_w1; u32 max_mtu; u8 rss_hfunc; @@ -673,6 +689,10 @@ struct nfp_net { u8 __iomem *tx_bar; u8 __iomem *rx_bar; +#ifdef CONFIG_NFP_NET_IPSEC + struct xarray xa_ipsec; +#endif + struct nfp_net_tlv_caps tlv_caps; unsigned int ktls_tx_conn_cnt; @@ -705,9 +725,55 @@ struct nfp_net { struct nfp_port *port; + struct { + spinlock_t lock; + struct list_head list; + struct work_struct work; + } mbox_amsg; + + struct { + u16 count; + struct list_head list; + } fs; + void *app_priv; }; +struct nfp_fs_entry { + struct list_head node; + u32 flow_type; + u32 loc; + struct { + union { + struct { + __be32 sip4; + __be32 dip4; + }; + struct { + __be32 sip6[4]; + __be32 dip6[4]; + }; + }; + union { + __be16 l3_proto; + u8 l4_proto; + }; + __be16 sport; + __be16 dport; + } key, msk; + u64 action; +}; + +struct nfp_mbox_amsg_entry { + struct list_head list; + int (*cfg)(struct nfp_net *nn, struct nfp_mbox_amsg_entry *entry); + u32 cmd; + char msg[]; +}; + +int nfp_net_sched_mbox_amsg_work(struct nfp_net *nn, u32 cmd, const void *data, size_t len, + int (*cb)(struct nfp_net *, struct nfp_mbox_amsg_entry *)); + /* Functions to read/write from/to a BAR * Performs any endian conversion necessary. */ @@ -767,7 +833,6 @@ static inline void nn_pci_flush(struct nfp_net *nn) * either add to a pointer or to read the pointer value. */ #define NFP_QCP_QUEUE_ADDR_SZ 0x800 -#define NFP_QCP_QUEUE_AREA_SZ 0x80000 #define NFP_QCP_QUEUE_OFF(_x) ((_x) * NFP_QCP_QUEUE_ADDR_SZ) #define NFP_QCP_QUEUE_ADD_RPTR 0x0000 #define NFP_QCP_QUEUE_ADD_WPTR 0x0004 @@ -776,50 +841,21 @@ static inline void nn_pci_flush(struct nfp_net *nn) #define NFP_QCP_QUEUE_STS_HI 0x000c #define NFP_QCP_QUEUE_STS_HI_WRITEPTR_mask 0x3ffff -/* The offset of a QCP queues in the PCIe Target */ -#define NFP_PCIE_QUEUE(_q) (0x80000 + (NFP_QCP_QUEUE_ADDR_SZ * ((_q) & 0xff))) - /* nfp_qcp_ptr - Read or Write Pointer of a queue */ enum nfp_qcp_ptr { NFP_QCP_READ_PTR = 0, NFP_QCP_WRITE_PTR }; -/* There appear to be an *undocumented* upper limit on the value which - * one can add to a queue and that value is either 0x3f or 0x7f. We - * go with 0x3f as a conservative measure. - */ -#define NFP_QCP_MAX_ADD 0x3f - -static inline void _nfp_qcp_ptr_add(u8 __iomem *q, - enum nfp_qcp_ptr ptr, u32 val) -{ - u32 off; - - if (ptr == NFP_QCP_READ_PTR) - off = NFP_QCP_QUEUE_ADD_RPTR; - else - off = NFP_QCP_QUEUE_ADD_WPTR; - - while (val > NFP_QCP_MAX_ADD) { - writel(NFP_QCP_MAX_ADD, q + off); - val -= NFP_QCP_MAX_ADD; - } - - writel(val, q + off); -} - /** * nfp_qcp_rd_ptr_add() - Add the value to the read pointer of a queue * * @q: Base address for queue structure * @val: Value to add to the queue pointer - * - * If @val is greater than @NFP_QCP_MAX_ADD multiple writes are performed. */ static inline void nfp_qcp_rd_ptr_add(u8 __iomem *q, u32 val) { - _nfp_qcp_ptr_add(q, NFP_QCP_READ_PTR, val); + writel(val, q + NFP_QCP_QUEUE_ADD_RPTR); } /** @@ -827,12 +863,10 @@ static inline void nfp_qcp_rd_ptr_add(u8 __iomem *q, u32 val) * * @q: Base address for queue structure * @val: Value to add to the queue pointer - * - * If @val is greater than @NFP_QCP_MAX_ADD multiple writes are performed. */ static inline void nfp_qcp_wr_ptr_add(u8 __iomem *q, u32 val) { - _nfp_qcp_ptr_add(q, NFP_QCP_WRITE_PTR, val); + writel(val, q + NFP_QCP_QUEUE_ADD_WPTR); } static inline u32 _nfp_qcp_read(u8 __iomem *q, enum nfp_qcp_ptr ptr) @@ -875,6 +909,8 @@ static inline u32 nfp_qcp_wr_ptr_read(u8 __iomem *q) return _nfp_qcp_read(q, NFP_QCP_WRITE_PTR); } +u32 nfp_qcp_queue_offset(const struct nfp_dev_info *dev_info, u16 queue); + static inline bool nfp_net_is_data_vnic(struct nfp_net *nn) { WARN_ON_ONCE(!nn->dp.netdev && nn->port); @@ -921,16 +957,18 @@ static inline void nn_ctrl_bar_unlock(struct nfp_net *nn) /* Globals */ extern const char nfp_driver_version[]; -extern const struct net_device_ops nfp_net_netdev_ops; +extern const struct net_device_ops nfp_nfd3_netdev_ops; +extern const struct net_device_ops nfp_nfdk_netdev_ops; static inline bool nfp_netdev_is_nfp_net(struct net_device *netdev) { - return netdev->netdev_ops == &nfp_net_netdev_ops; + return netdev->netdev_ops == &nfp_nfd3_netdev_ops || + netdev->netdev_ops == &nfp_nfdk_netdev_ops; } -static inline int nfp_net_coalesce_para_check(u32 usecs, u32 pkts) +static inline int nfp_net_coalesce_para_check(u32 param) { - if ((usecs >= ((1 << 16) - 1)) || (pkts >= ((1 << 16) - 1))) + if (param >= ((1 << 16) - 1)) return -EINVAL; return 0; @@ -941,7 +979,8 @@ void nfp_net_get_fw_version(struct nfp_net_fw_version *fw_ver, void __iomem *ctrl_bar); struct nfp_net * -nfp_net_alloc(struct pci_dev *pdev, void __iomem *ctrl_bar, bool needs_netdev, +nfp_net_alloc(struct pci_dev *pdev, const struct nfp_dev_info *dev_info, + void __iomem *ctrl_bar, bool needs_netdev, unsigned int max_tx_rings, unsigned int max_rx_rings); void nfp_net_free(struct nfp_net *nn); @@ -972,11 +1011,18 @@ void nfp_net_irqs_disable(struct pci_dev *pdev); void nfp_net_irqs_assign(struct nfp_net *nn, struct msix_entry *irq_entries, unsigned int n); +struct sk_buff * +nfp_net_tls_tx(struct nfp_net_dp *dp, struct nfp_net_r_vector *r_vec, + struct sk_buff *skb, u64 *tls_handle, int *nr_frags); +void nfp_net_tls_tx_undo(struct sk_buff *skb, u64 tls_handle); struct nfp_net_dp *nfp_net_clone_dp(struct nfp_net *nn); int nfp_net_ring_reconfig(struct nfp_net *nn, struct nfp_net_dp *new, struct netlink_ext_ack *extack); +int nfp_net_fs_add_hw(struct nfp_net *nn, struct nfp_fs_entry *entry); +int nfp_net_fs_del_hw(struct nfp_net *nn, struct nfp_fs_entry *entry); + #ifdef CONFIG_NFP_DEBUG void nfp_net_debugfs_create(void); void nfp_net_debugfs_destroy(void); |