14.1 网络设备驱动结构

  • 网络协议接口层：硬件无关，标准收发函数dev_queue_xmit()和netif_rx();  注意，netif_rx是将接收到的数据给上层，有时也在驱动收到数据以后调用。
  • 网络设备接口层，net_device，统一接口名称，使上层独立于具体硬件。
  • 设备驱动功能层，实现net_device的各成员
  • 物理层

在整个以太网架构里，有两个数据结构非常重要，即sk_buff和net_device，后面两节有说明。

还有一些与内核交互的函数，需要掌握，如netif_start_queue()，netif_stop_queue()，netif_wakeup_queue()，netif_rx()，netif_carrier_on/off_ok()

14.1.1 sk_buff

　　以太网各层之间用sk_buff结构体传递数据，该结构体是很多函数的形参。

  1. #include <linux/skbuff.h>
  2.  
  3. /**
  4.  *    struct sk_buff - socket buffer
  5.  *    @next: Next buffer in list
  6.  *    @prev: Previous buffer in list
  7.  *    @tstamp: Time we arrived
  8.  *    @sk: Socket we are owned by
  9.  *    @dev: Device we arrived on/are leaving by
  10.  *    @cb: Control buffer. Free for use by every layer. Put private vars here
  11.  *    @_skb_refdst: destination entry (with norefcount bit)
  12.  *    @sp: the security path, used for xfrm
  13.  *    @len: Length of actual data
  14.  *    @data_len: Data length
  15.  *    @mac_len: Length of link layer header
  16.  *    @hdr_len: writable header length of cloned skb
  17.  *    @csum: Checksum (must include start/offset pair)
  18.  *    @csum_start: Offset from skb->head where checksumming should start
  19.  *    @csum_offset: Offset from csum_start where checksum should be stored
  20.  *    @priority: Packet queueing priority
  21.  *    @local_df: allow local fragmentation
  22.  *    @cloned: Head may be cloned (check refcnt to be sure)
  23.  *    @ip_summed: Driver fed us an IP checksum
  24.  *    @nohdr: Payload reference only, must not modify header
  25.  *    @nfctinfo: Relationship of this skb to the connection
  26.  *    @pkt_type: Packet class
  27.  *    @fclone: skbuff clone status
  28.  *    @ipvs_property: skbuff is owned by ipvs
  29.  *    @peeked: this packet has been seen already, so stats have been
  30.  *        done for it, don't do them again
  31.  *    @nf_trace: netfilter packet trace flag
  32.  *    @protocol: Packet protocol from driver
  33.  *    @destructor: Destruct function
  34.  *    @nfct: Associated connection, if any
  35.  *    @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
  36.  *    @skb_iif: ifindex of device we arrived on
  37.  *    @tc_index: Traffic control index
  38.  *    @tc_verd: traffic control verdict
  39.  *    @rxhash: the packet hash computed on receive
  40.  *    @queue_mapping: Queue mapping for multiqueue devices
  41.  *    @ndisc_nodetype: router type (from link layer)
  42.  *    @ooo_okay: allow the mapping of a socket to a queue to be changed
  43.  *    @l4_rxhash: indicate rxhash is a canonical 4-tuple hash over transport
  44.  *        ports.
  45.  *    @wifi_acked_valid: wifi_acked was set
  46.  *    @wifi_acked: whether frame was acked on wifi or not
  47.  *    @no_fcs:  Request NIC to treat last 4 bytes as Ethernet FCS
  48.  *    @dma_cookie: a cookie to one of several possible DMA operations
  49.  *        done by skb DMA functions
  50.   *    @napi_id: id of the NAPI struct this skb came from
  51.  *    @secmark: security marking
  52.  *    @mark: Generic packet mark
  53.  *    @dropcount: total number of sk_receive_queue overflows
  54.  *    @vlan_proto: vlan encapsulation protocol
  55.  *    @vlan_tci: vlan tag control information
  56.  *    @inner_protocol: Protocol (encapsulation)
  57.  *    @inner_transport_header: Inner transport layer header (encapsulation)
  58.  *    @inner_network_header: Network layer header (encapsulation)
  59.  *    @inner_mac_header: Link layer header (encapsulation)
  60.  *    @transport_header: Transport layer header
  61.  *    @network_header: Network layer header
  62.  *    @mac_header: Link layer header
  63.  *    @tail: Tail pointer
  64.  *    @end: End pointer
  65.  *    @head: Head of buffer
  66.  *    @data: Data head pointer
  67.  *    @truesize: Buffer size
  68.  *    @users: User count - see {datagram,tcp}.c
  69.  */
  70.  
  71. struct sk_buff {
  72.     /* These two members must be first. */
  73.     struct sk_buff        *next;
  74.     struct sk_buff        *prev;
  75.  
  76.     ktime_t            tstamp;
  77.  
  78.     struct sock        *sk;
  79.     struct net_device    *dev;
  80.  
  81.     /*
  82.      * This is the control buffer. It is free to use for every
  83.      * layer. Please put your private variables there. If you
  84.      * want to keep them across layers you have to do a skb_clone()
  85.      * first. This is owned by whoever has the skb queued ATM.
  86.      */
  87.     char            cb[] __aligned();
  88.  
  89.     unsigned long        _skb_refdst;
  90. #ifdef CONFIG_XFRM
  91.     struct    sec_path    *sp;
  92. #endif
  93.     unsigned int        len,　　　　// data段的长度
  94.                 data_len;
  95.     __u16            mac_len,
  96.                 hdr_len;
  97.     union {
  98.         __wsum        csum;
  99.         struct {
  100.             __u16    csum_start;
  101.             __u16    csum_offset;
  102.         };
  103.     };
  104.     __u32            priority;
  105.     kmemcheck_bitfield_begin(flags1);
  106.     __u8            local_df:,
  107.                 cloned:,
  108.                 ip_summed:,
  109.                 nohdr:,
  110.                 nfctinfo:;
  111.     __u8            pkt_type:,
  112.                 fclone:,
  113.                 ipvs_property:,
  114.                 peeked:,
  115.                 nf_trace:;
  116.     kmemcheck_bitfield_end(flags1);
  117.     __be16            protocol;
  118.  
  119.     void            (*destructor)(struct sk_buff *skb);
  120. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  121.     struct nf_conntrack    *nfct;
  122. #endif
  123. #ifdef CONFIG_BRIDGE_NETFILTER
  124.     struct nf_bridge_info    *nf_bridge;
  125. #endif
  126.  
  127.     int            skb_iif;
  128.  
  129.     __u32            rxhash;
  130.  
  131.     __be16            vlan_proto;
  132.     __u16            vlan_tci;
  133.  
  134. #ifdef CONFIG_NET_SCHED
  135.     __u16            tc_index;    /* traffic control index */
  136. #ifdef CONFIG_NET_CLS_ACT
  137.     __u16            tc_verd;    /* traffic control verdict */
  138. #endif
  139. #endif
  140.  
  141.     __u16            queue_mapping;
  142.     kmemcheck_bitfield_begin(flags2);
  143. #ifdef CONFIG_IPV6_NDISC_NODETYPE
  144.     __u8            ndisc_nodetype:;
  145. #endif
  146.     __u8            pfmemalloc:;
  147.     __u8            ooo_okay:;
  148.     __u8            l4_rxhash:;
  149.     __u8            wifi_acked_valid:;
  150.     __u8            wifi_acked:;
  151.     __u8            no_fcs:;
  152.     __u8            head_frag:;
  153.     /* Encapsulation protocol and NIC drivers should use
  154.      * this flag to indicate to each other if the skb contains
  155.      * encapsulated packet or not and maybe use the inner packet
  156.      * headers if needed
  157.      */
  158.     __u8            encapsulation:;
  159.     /* 6/8 bit hole (depending on ndisc_nodetype presence) */
  160.     kmemcheck_bitfield_end(flags2);
  161.  
  162. #if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
  163.     union {
  164.         unsigned int    napi_id;
  165.         dma_cookie_t    dma_cookie;
  166.     };
  167. #endif
  168. #ifdef CONFIG_NETWORK_SECMARK
  169.     __u32            secmark;
  170. #endif
  171.     union {
  172.         __u32        mark;
  173.         __u32        dropcount;
  174.         __u32        reserved_tailroom;
  175.     };
  176.  
  177.     __be16            inner_protocol;
  178.     __u16            inner_transport_header;
  179.     __u16            inner_network_header;
  180.     __u16            inner_mac_header;
  181.     __u16            transport_header;
  182.     __u16            network_header;
  183.     __u16            mac_header;
  184.     /* These elements must be at the end, see alloc_skb() for details.  */
  185.     sk_buff_data_t        tail;
  186.     sk_buff_data_t        end;
  187.     unsigned char        *head,
  188.                 *data;
  189.     unsigned int        truesize;
  190.     atomic_t        users;
  191. };

  • 接收时，各层去掉自己的协议，把数据给上层；
  • 发送时，各层添加自己的协议，最终给物理网口。
  • 上图指针可以动态调整，下面是若干函数
  1. /**   tail后移，即在data中增加数据
  2.  *    skb_put - add data to a buffer
  3.  *    @skb: buffer to use
  4.  *    @len: amount of data to add
  5.  *
  6.  *    This function extends the used data area of the buffer. If this would
  7.  *    exceed the total buffer size the kernel will panic. A pointer to the
  8.  *    first byte of the extra data is returned.
  9.  */
  10. unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
  11. {
  12.     unsigned char *tmp = skb_tail_pointer(skb);
  13.     SKB_LINEAR_ASSERT(skb);
  14.     skb->tail += len;
  15.     skb->len  += len;
  16.     if (unlikely(skb->tail > skb->end))
  17.         skb_over_panic(skb, len, __builtin_return_address());
  18.     return tmp;
  19. }
  20.  
  21. /**
  22.  *    skb_push - add data to the start of a buffer，data前移
  23.  *    @skb: buffer to use
  24.  *    @len: amount of data to add
  25.  *
  26.  *    This function extends the used data area of the buffer at the buffer
  27.  *    start. If this would exceed the total buffer headroom the kernel will
  28.  *    panic. A pointer to the first byte of the extra data is returned.
  29.  */
  30. unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
  31. {
  32.     skb->data -= len;
  33.     skb->len  += len;
  34.     if (unlikely(skb->data<skb->head))
  35.         skb_under_panic(skb, len, __builtin_return_address());
  36.     return skb->data;
  37. }
  38.  
  39. /**
  40.  *    skb_pull - remove data from the start of a buffer，data后移
  41.  *    @skb: buffer to use
  42.  *    @len: amount of data to remove
  43.  *
  44.  *    This function removes data from the start of a buffer, returning
  45.  *    the memory to the headroom. A pointer to the next data in the buffer
  46.  *    is returned. Once the data has been pulled future pushes will overwrite
  47.  *    the old data.
  48.  */
  49. unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
  50. {
  51.     return skb_pull_inline(skb, len);
  52. }
  53.  
  54. static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
  55. {
  56.     return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
  57. }
  58.  
  59. static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
  60. {
  61.     skb->len -= len;
  62.     BUG_ON(skb->len < skb->data_len);
  63.     return skb->data += len;
  64. }
  65. /**
  66.  *    skb_reserve - adjust headroom
  67.  *    @skb: buffer to alter
  68.  *    @len: bytes to move
  69.  *
  70.  *    Increase the headroom of an empty &sk_buff by reducing the tail
  71.  *    room. This is only allowed for an empty buffer.
  72.  */
  73. static inline void skb_reserve(struct sk_buff *skb, int len)
  74. {
  75.     skb->data += len;
  76.     skb->tail += len;
  77. }
  78.  
  79. // 例子：
    skb=alloc_skb(len+headspace, GFP_KERNEL);　　// 分配
    skb_reserve(skb, headspace);　　　　　　　　　 // tail=data=起始位置+headspace
    skb_put(skb,len);　　　　　　　　　　　　　　　　// tail += len
    memcpy_fromfs(skb->data,data,len);
    pass_to_m_protocol(skb);

  • sk_buff的动态分配和释放

  1. static inline struct sk_buff *alloc_skb(unsigned int size,
  2.                     gfp_t priority);/* legacy helper around netdev_alloc_skb() */
  3. static inline struct sk_buff *dev_alloc_skb(unsigned int length);void kfree_skb(struct sk_buff *skb);
    void dev_kfree_skb)(struct sk_buff * skb);
  4.  
  5. /*
  6.  * It is not allowed to call kfree_skb() or consume_skb() from hardware
  7.  * interrupt context or with hardware interrupts being disabled.
  8.  * (in_irq() || irqs_disabled())
  9.  *
  10.  * We provide four helpers that can be used in following contexts :
  11.  *
  12.  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
  13.  *  replacing kfree_skb(skb)
  14.  *
  15.  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
  16.  *  Typically used in place of consume_skb(skb) in TX completion path
  17.  *
  18.  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
  19.  *  replacing kfree_skb(skb)
  20.  *
  21.  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
  22.  *  and consumed a packet. Used in place of consume_skb(skb)
  23.  */
  24. static inline void dev_kfree_skb_irq(struct sk_buff *skb);
  25. static inline void dev_kfree_skb_any(struct sk_buff *skb);

14.1.2 net_device

net_device有点类似字符设备中的file_opretions,里面定义了很多标准成员函数，驱动需要实现里面的函数。

  1. /*
  2.  *    The DEVICE structure.
  3.  *    Actually, this whole structure is a big mistake.  It mixes I/O
  4.  *    data with strictly "high-level" data, and it has to know about
  5.  *    almost every data structure used in the INET module.
  6.  *
  7.  *    FIXME: cleanup struct net_device such that network protocol info
  8.  *    moves out.
  9.  */
  10.  
  11. struct net_device {
  12.  
  13.     /*
  14.      * This is the first field of the "visible" part of this structure
  15.      * (i.e. as seen by users in the "Space.c" file).  It is the name
  16.      * of the interface.
  17.      */
  18.     char            name[IFNAMSIZ];　　　　　　/* device name hash chain, please keep it close to name[] */
  19.     struct hlist_node    name_hlist;
  20.  
  21.     /* snmp alias */
  22.     char             *ifalias;
  23.  
  24.     /*
  25.      *    I/O specific fields
  26.      *    FIXME: Merge these and struct ifmap into one
  27.      */
  28.     unsigned long        mem_end;    /* shared mem end    */
  29.     unsigned long        mem_start;    /* shared mem start    */
  30.     unsigned long        base_addr;    /* device I/O address    */
  31.     int            irq;        /* device IRQ number    */
  32.  
  33.     /*
  34.      *    Some hardware also needs these fields, but they are not
  35.      *    part of the usual set specified in Space.c.
  36.      */
  37.  
  38.     unsigned long        state;
  39.  
  40.     struct list_head    dev_list;
  41.     struct list_head    napi_list;
  42.     struct list_head    unreg_list;
  43.     struct list_head    close_list;
  44.  
  45.     /* directly linked devices, like slaves for bonding */
  46.     struct {
  47.         struct list_head upper;
  48.         struct list_head lower;
  49.     } adj_list;
  50.  
  51.     /* all linked devices, *including* neighbours */
  52.     struct {
  53.         struct list_head upper;
  54.         struct list_head lower;
  55.     } all_adj_list;
  56.  
  57.     /* currently active device features */
  58.     netdev_features_t    features;
  59.     /* user-changeable features */
  60.     netdev_features_t    hw_features;
  61.     /* user-requested features */
  62.     netdev_features_t    wanted_features;
  63.     /* mask of features inheritable by VLAN devices */
  64.     netdev_features_t    vlan_features;
  65.     /* mask of features inherited by encapsulating devices
  66.      * This field indicates what encapsulation offloads
  67.      * the hardware is capable of doing, and drivers will
  68.      * need to set them appropriately.
  69.      */
  70.     netdev_features_t    hw_enc_features;
  71.     /* mask of fetures inheritable by MPLS */
  72.     netdev_features_t    mpls_features;
  73.  
  74.     /* Interface index. Unique device identifier    */
  75.     int            ifindex;
  76.     int            iflink;
  77.  
  78.     struct net_device_stats    stats;　　// 各种统计信息
  79.     atomic_long_t        rx_dropped; /* dropped packets by core network
  80.                          * Do not use this in drivers.
  81.                          */
  82.  
  83. #ifdef CONFIG_WIRELESS_EXT
  84.     /* List of functions to handle Wireless Extensions (instead of ioctl).
  85.      * See <net/iw_handler.h> for details. Jean II */
  86.     const struct iw_handler_def *    wireless_handlers;
  87.     /* Instance data managed by the core of Wireless Extensions. */
  88.     struct iw_public_data *    wireless_data;
  89. #endif
  90.     /* Management operations */
  91.     const struct net_device_ops *netdev_ops;　　// 具体函数，需驱动填充
  92.     const struct ethtool_ops *ethtool_ops;
  93.     const struct forwarding_accel_ops *fwd_ops;
  94.  
  95.     /* Hardware header description */
  96.     const struct header_ops *header_ops;
  97.  
  98.     unsigned int        flags;    /* interface flags (a la BSD)    ,接口标记，以IFF_开头，说明设备接口的能力和特性*/
  99.     unsigned int        priv_flags; /* Like 'flags' but invisible to userspace.
  100.                          * See if.h for definitions. */
  101.     unsigned short        gflags;
  102.     unsigned short        padded;    /* How much padding added by alloc_netdev() */
  103.  
  104.     unsigned char        operstate; /* RFC2863 operstate */
  105.     unsigned char        link_mode; /* mapping policy to operstate */
  106.  
  107.     unsigned char        if_port;    /* Selectable AUI, TP,..*/
  108.     unsigned char        dma;        /* DMA channel        */
  109.  
  110.     unsigned int        mtu;    /* interface MTU value，最单传输单元        */
  111.     unsigned short        type;    /* interface hardware type，硬件类型    */
  112.     unsigned short        hard_header_len;    /* hardware hdr length，Dmac+Smac+type=14    */
  113.  
  114.     /* extra head- and tailroom the hardware may need, but not in all cases
  115.      * can this be guaranteed, especially tailroom. Some cases also use
  116.      * LL_MAX_HEADER instead to allocate the skb.
  117.      */
  118.     unsigned short        needed_headroom;
  119.     unsigned short        needed_tailroom;
  120.  
  121.     /* Interface address info. */
  122.     unsigned char        perm_addr[MAX_ADDR_LEN]; /* permanent hw address */
  123.     unsigned char        addr_assign_type; /* hw address assignment type */
  124.     unsigned char        addr_len;    /* hardware address length    */
  125.     unsigned short        neigh_priv_len;
  126.     unsigned short          dev_id;        /* Used to differentiate devices
  127.                          * that share the same link
  128.                          * layer address
  129.                          */
  130.     spinlock_t        addr_list_lock;
  131.     struct netdev_hw_addr_list    uc;    /* Unicast mac addresses */
  132.     struct netdev_hw_addr_list    mc;    /* Multicast mac addresses */
  133.     struct netdev_hw_addr_list    dev_addrs; /* list of device
  134.                             * hw addresses
  135.                             */
  136. #ifdef CONFIG_SYSFS
  137.     struct kset        *queues_kset;
  138. #endif
  139.  
  140.     bool            uc_promisc;
  141.     unsigned int        promiscuity;
  142.     unsigned int        allmulti;
  143.  
  144.     /* Protocol specific pointers */
  145.  
  146. #if IS_ENABLED(CONFIG_VLAN_8021Q)
  147.     struct vlan_info __rcu    *vlan_info;    /* VLAN info */
  148. #endif
  149. #if IS_ENABLED(CONFIG_NET_DSA)
  150.     struct dsa_switch_tree    *dsa_ptr;    /* dsa specific data */
  151. #endif
  152. #if IS_ENABLED(CONFIG_TIPC)
  153.     struct tipc_bearer __rcu *tipc_ptr;    /* TIPC specific data */
  154. #endif
  155.     void             *atalk_ptr;    /* AppleTalk link     */
  156.     struct in_device __rcu    *ip_ptr;    /* IPv4 specific data    */
  157.     struct dn_dev __rcu     *dn_ptr;        /* DECnet specific data */
  158.     struct inet6_dev __rcu    *ip6_ptr;       /* IPv6 specific data */
  159.     void            *ax25_ptr;    /* AX.25 specific data */
  160.     struct wireless_dev    *ieee80211_ptr;    /* IEEE 802.11 specific data,
  161.                            assign before registering */
  162.  
  163. /*
  164.  * Cache lines mostly used on receive path (including eth_type_trans())
  165.  */
  166.     unsigned long        last_rx;    /* Time of last Rx
  167.                          * This should not be set in
  168.                          * drivers, unless really needed,
  169.                          * because network stack (bonding)
  170.                          * use it if/when necessary, to
  171.                          * avoid dirtying this cache line.
  172.                          */
  173.  
  174.     /* Interface address info used in eth_type_trans() */
  175.     unsigned char        *dev_addr;    /* hw address, (before bcast
  176.                            because most packets are
  177.                            unicast) */
  178.  
  179. #ifdef CONFIG_SYSFS
  180.     struct netdev_rx_queue    *_rx;
  181.  
  182.     /* Number of RX queues allocated at register_netdev() time */
  183.     unsigned int        num_rx_queues;
  184.  
  185.     /* Number of RX queues currently active in device */
  186.     unsigned int        real_num_rx_queues;
  187.  
  188. #endif
  189.  
  190.     rx_handler_func_t __rcu    *rx_handler;
  191.     void __rcu        *rx_handler_data;
  192.  
  193.     struct netdev_queue __rcu *ingress_queue;
  194.     unsigned char        broadcast[MAX_ADDR_LEN];    /* hw bcast add    */
  195.  
  196. /*
  197.  * Cache lines mostly used on transmit path
  198.  */
  199.     struct netdev_queue    *_tx ____cacheline_aligned_in_smp;
  200.  
  201.     /* Number of TX queues allocated at alloc_netdev_mq() time  */
  202.     unsigned int        num_tx_queues;
  203.  
  204.     /* Number of TX queues currently active in device  */
  205.     unsigned int        real_num_tx_queues;
  206.  
  207.     /* root qdisc from userspace point of view */
  208.     struct Qdisc        *qdisc;
  209.  
  210.     unsigned long        tx_queue_len;    /* Max frames per queue allowed */
  211.     spinlock_t        tx_global_lock;
  212.  
  213. #ifdef CONFIG_XPS
  214.     struct xps_dev_maps __rcu *xps_maps;
  215. #endif
  216. #ifdef CONFIG_RFS_ACCEL
  217.     /* CPU reverse-mapping for RX completion interrupts, indexed
  218.      * by RX queue number.  Assigned by driver.  This must only be
  219.      * set if the ndo_rx_flow_steer operation is defined. */
  220.     struct cpu_rmap        *rx_cpu_rmap;
  221. #endif
  222.  
  223.     /* These may be needed for future network-power-down code. */
  224.  
  225.     /*
  226.      * trans_start here is expensive for high speed devices on SMP,
  227.      * please use netdev_queue->trans_start instead.
  228.      */
  229.     unsigned long        trans_start;    /* Time (in jiffies) of last Tx    */
  230.  
  231.     int            watchdog_timeo; /* used by dev_watchdog() */
  232.     struct timer_list    watchdog_timer;
  233.  
  234.     /* Number of references to this device */
  235.     int __percpu        *pcpu_refcnt;
  236.  
  237.     /* delayed register/unregister */
  238.     struct list_head    todo_list;
  239.     /* device index hash chain */
  240.     struct hlist_node    index_hlist;
  241.  
  242.     struct list_head    link_watch_list;
  243.  
  244.     /* register/unregister state machine */
  245.     enum { NETREG_UNINITIALIZED=,
  246.            NETREG_REGISTERED,    /* completed register_netdevice */
  247.            NETREG_UNREGISTERING,    /* called unregister_netdevice */
  248.            NETREG_UNREGISTERED,    /* completed unregister todo */
  249.            NETREG_RELEASED,        /* called free_netdev */
  250.            NETREG_DUMMY,        /* dummy device for NAPI poll */
  251.     } reg_state:;
  252.  
  253.     bool dismantle; /* device is going do be freed */
  254.  
  255.     enum {
  256.         RTNL_LINK_INITIALIZED,
  257.         RTNL_LINK_INITIALIZING,
  258.     } rtnl_link_state:;
  259.  
  260.     /* Called from unregister, can be used to call free_netdev */
  261.     void (*destructor)(struct net_device *dev);
  262.  
  263. #ifdef CONFIG_NETPOLL
  264.     struct netpoll_info __rcu    *npinfo;
  265. #endif
  266.  
  267. #ifdef CONFIG_NET_NS
  268.     /* Network namespace this network device is inside */
  269.     struct net        *nd_net;
  270. #endif
  271.  
  272.     /* mid-layer private */
  273.     union {
  274.         void                *ml_priv;
  275.         struct pcpu_lstats __percpu    *lstats; /* loopback stats */
  276.         struct pcpu_sw_netstats __percpu    *tstats;
  277.         struct pcpu_dstats __percpu    *dstats; /* dummy stats */
  278.         struct pcpu_vstats __percpu    *vstats; /* veth stats */
  279.     };
  280.     /* GARP */
  281.     struct garp_port __rcu    *garp_port;
  282.     /* MRP */
  283.     struct mrp_port __rcu    *mrp_port;
  284.  
  285.     /* class/net/name entry */
  286.     struct device        dev;
  287.     /* space for optional device, statistics, and wireless sysfs groups */
  288.     const struct attribute_group *sysfs_groups[];
  289.     /* space for optional per-rx queue attributes */
  290.     const struct attribute_group *sysfs_rx_queue_group;
  291.  
  292.     /* rtnetlink link ops */
  293.     const struct rtnl_link_ops *rtnl_link_ops;
  294.  
  295.     /* for setting kernel sock attribute on TCP connection setup */
  296. #define GSO_MAX_SIZE        65536
  297.     unsigned int        gso_max_size;
  298. #define GSO_MAX_SEGS        65535
  299.     u16            gso_max_segs;
  300.  
  301. #ifdef CONFIG_DCB
  302.     /* Data Center Bridging netlink ops */
  303.     const struct dcbnl_rtnl_ops *dcbnl_ops;
  304. #endif
  305.     u8 num_tc;
  306.     struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
  307.     u8 prio_tc_map[TC_BITMASK + ];
  308.  
  309. #if IS_ENABLED(CONFIG_FCOE)
  310.     /* max exchange id for FCoE LRO by ddp */
  311.     unsigned int        fcoe_ddp_xid;
  312. #endif
  313. #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
  314.     struct netprio_map __rcu *priomap;
  315. #endif
  316.     /* phy device may attach itself for hardware timestamping */
  317.     struct phy_device *phydev;
  318.  
  319.     struct lock_class_key *qdisc_tx_busylock;
  320.  
  321.     /* group the device belongs to */
  322.     int group;
  323.  
  324.     struct pm_qos_request    pm_qos_req;
  325. };

* Standard interface flags (netdevice->flags). */
#define IFF_UP 0x1 /* interface is up */
#define IFF_BROADCAST 0x2 /* broadcast address valid */
#define IFF_DEBUG 0x4 /* turn on debugging */
#define IFF_LOOPBACK 0x8 /* is a loopback net */
#define IFF_POINTOPOINT 0x10 /* interface is has p-p link */
#define IFF_NOTRAILERS 0x20 /* avoid use of trailers */
#define IFF_RUNNING 0x40 /* interface RFC2863 OPER_UP */
#define IFF_NOARP 0x80 /* no ARP protocol */
#define IFF_PROMISC 0x100 /* receive all packets */
#define IFF_ALLMULTI 0x200 /* receive all multicast packets*/

#define IFF_MASTER 0x400 /* master of a load balancer */
#define IFF_SLAVE 0x800 /* slave of a load balancer */

#define IFF_MULTICAST 0x1000 /* Supports multicast */

#define IFF_PORTSEL 0x2000 /* can set media type */

#define IFF_AUTOMEDIA 0x4000 /* auto media select active */
#define IFF_DYNAMIC 0x8000 /* dialup device with changing addresses*/

  1.  

#define IFF_LOWER_UP 0x10000 /* driver signals L1 up */
#define IFF_DORMANT 0x20000 /* driver signals dormant */

  1.  

#define IFF_ECHO 0x40000 /* echo sent packets */

  1.  

#define IFF_VOLATILE (IFF_LOOPBACK|IFF_POINTOPOINT|IFF_BROADCAST|IFF_ECHO|\
IFF_MASTER|IFF_SLAVE|IFF_RUNNING|IFF_LOWER_UP|IFF_DORMANT)

  1.  

/* Private (from user) interface flags (netdevice->priv_flags). */
#define IFF_802_1Q_VLAN 0x1 /* 802.1Q VLAN device. */
#define IFF_EBRIDGE 0x2 /* Ethernet bridging device. */
#define IFF_SLAVE_INACTIVE 0x4 /* bonding slave not the curr. active */
#define IFF_MASTER_8023AD 0x8 /* bonding master, 802.3ad. */
#define IFF_MASTER_ALB 0x10 /* bonding master, balance-alb. */
#define IFF_BONDING 0x20 /* bonding master or slave */
#define IFF_SLAVE_NEEDARP 0x40 /* need ARPs for validation */
#define IFF_ISATAP 0x80 /* ISATAP interface (RFC4214) */
#define IFF_MASTER_ARPMON 0x100 /* bonding master, ARP mon in use */
#define IFF_WAN_HDLC 0x200 /* WAN HDLC device */
#define IFF_XMIT_DST_RELEASE 0x400 /* dev_hard_start_xmit() is allowed to
* release skb->dst
*/
#define IFF_DONT_BRIDGE 0x800 /* disallow bridging this ether dev */
#define IFF_DISABLE_NETPOLL 0x1000 /* disable netpoll at run-time */
#define IFF_MACVLAN_PORT 0x2000 /* device used as macvlan port */
#define IFF_BRIDGE_PORT 0x4000 /* device used as bridge port */
#define IFF_OVS_DATAPATH 0x8000 /* device used as Open vSwitch
* datapath port */
#define IFF_TX_SKB_SHARING 0x10000 /* The interface supports sharing
* skbs on transmit */
#define IFF_UNICAST_FLT 0x20000 /* Supports unicast filtering */
#define IFF_TEAM_PORT 0x40000 /* device used as team port */
#define IFF_SUPP_NOFCS 0x80000 /* device supports sending custom FCS */
#define IFF_LIVE_ADDR_CHANGE 0x100000 /* device supports hardware address
* change when it's running */
#define IFF_MACVLAN 0x200000 /* Macvlan device */

net_device中的一个重要结构体是net_device_ops,驱动需要填充里面的成员。

  1. /*
  2.  * This structure defines the management hooks for network devices.
  3.  * The following hooks can be defined; unless noted otherwise, they are
  4.  * optional and can be filled with a null pointer.
  5.  *
  6.  * int (*ndo_init)(struct net_device *dev);
  7.  *     This function is called once when network device is registered.
  8.  *     The network device can use this to any late stage initializaton
  9.  *     or semantic validattion. It can fail with an error code which will
  10.  *     be propogated back to register_netdev
  11.  *
  12.  * void (*ndo_uninit)(struct net_device *dev);
  13.  *     This function is called when device is unregistered or when registration
  14.  *     fails. It is not called if init fails.
  15.  *
  16.  * int (*ndo_open)(struct net_device *dev);一般在次获取设备需要的IO地址、IRQ、DMA通道等
  17.  *     This function is called when network device transistions to the up
  18.  *     state.
  19.  *
  20.  * int (*ndo_stop)(struct net_device *dev);
  21.  *     This function is called when network device transistions to the down
  22.  *     state.
  23.  *
  24.  * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
  25.  *                               struct net_device *dev);
  26.  *    Called when a packet needs to be transmitted.
  27.  *    Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
  28.  *        (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
  29.  *    Required can not be NULL.　　不能为空，必须实现，启动发送
  30.  *
  31.  * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
  32.  *                         void *accel_priv, select_queue_fallback_t fallback);
  33.  *    Called to decide which queue to when device supports multiple
  34.  *    transmit queues.
  35.  *
  36.  * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
  37.  *    This function is called to allow device receiver to make
  38.  *    changes to configuration when multicast or promiscious is enabled.
  39.  *
  40.  * void (*ndo_set_rx_mode)(struct net_device *dev);
  41.  *    This function is called device changes address list filtering.
  42.  *    If driver handles unicast address filtering, it should set
  43.  *    IFF_UNICAST_FLT to its priv_flags.
  44.  *
  45.  * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
  46.  *    This function  is called when the Media Access Control address
  47.  *    needs to be changed. If this interface is not defined, the
  48.  *    mac address can not be changed.
  49.  *
  50.  * int (*ndo_validate_addr)(struct net_device *dev);
  51.  *    Test if Media Access Control address is valid for the device.
  52.  *
  53.  * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
  54.  *    Called when a user request an ioctl which can't be handled by
  55.  *    the generic interface code. If not defined ioctl's return
  56.  *    not supported error code.
  57.  *
  58.  * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
  59.  *    Used to set network devices bus interface parameters. This interface
  60.  *    is retained for legacy reason, new devices should use the bus
  61.  *    interface (PCI) for low level management.　　配置接口，也可以改变设备IO地址和中断号
  62.  *
  63.  * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
  64.  *    Called when a user wants to change the Maximum Transfer Unit
  65.  *    of a device. If not defined, any request to change MTU will
  66.  *    will return an error.
  67.  *
  68.  * void (*ndo_tx_timeout)(struct net_device *dev);
  69.  *    Callback uses when the transmitter has not made any progress
  70.  *    for dev->watchdog ticks. 发送超时以后调用
  71.  *
  72.  * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
  73.  *                      struct rtnl_link_stats64 *storage);
  74.  * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);　　　　返回的结构体包含若干统计信息
  75.  *    Called when a user wants to get the network device usage
  76.  *    statistics. Drivers must do one of the following:
  77.  *    1. Define @ndo_get_stats64 to fill in a zero-initialised
  78.  *       rtnl_link_stats64 structure passed by the caller.
  79.  *    2. Define @ndo_get_stats to update a net_device_stats structure
  80.  *       (which should normally be dev->stats) and return a pointer to
  81.  *       it. The structure may be changed asynchronously only if each
  82.  *       field is written atomically.
  83.  *    3. Update dev->stats asynchronously and atomically, and define
  84.  *       neither operation.
  85.  *
  86.  * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid);
  87.  *    If device support VLAN filtering this function is called when a
  88.  *    VLAN id is registered.
  89.  *
  90.  * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
  91.  *    If device support VLAN filtering this function is called when a
  92.  *    VLAN id is unregistered.
  93.  *
  94.  * void (*ndo_poll_controller)(struct net_device *dev);
  95.  *
  96.  *    SR-IOV management functions.
  97.  * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
  98.  * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
  99.  * int (*ndo_set_vf_tx_rate)(struct net_device *dev, int vf, int rate);
  100.  * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
  101.  * int (*ndo_get_vf_config)(struct net_device *dev,
  102.  *                int vf, struct ifla_vf_info *ivf);
  103.  * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
  104.  * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
  105.  *              struct nlattr *port[]);
  106.  * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
  107.  * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
  108.  *     Called to setup 'tc' number of traffic classes in the net device. This
  109.  *     is always called from the stack with the rtnl lock held and netif tx
  110.  *     queues stopped. This allows the netdevice to perform queue management
  111.  *     safely.
  112.  *
  113.  *    Fiber Channel over Ethernet (FCoE) offload functions.
  114.  * int (*ndo_fcoe_enable)(struct net_device *dev);
  115.  *    Called when the FCoE protocol stack wants to start using LLD for FCoE
  116.  *    so the underlying device can perform whatever needed configuration or
  117.  *    initialization to support acceleration of FCoE traffic.
  118.  *
  119.  * int (*ndo_fcoe_disable)(struct net_device *dev);
  120.  *    Called when the FCoE protocol stack wants to stop using LLD for FCoE
  121.  *    so the underlying device can perform whatever needed clean-ups to
  122.  *    stop supporting acceleration of FCoE traffic.
  123.  *
  124.  * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
  125.  *                 struct scatterlist *sgl, unsigned int sgc);
  126.  *    Called when the FCoE Initiator wants to initialize an I/O that
  127.  *    is a possible candidate for Direct Data Placement (DDP). The LLD can
  128.  *    perform necessary setup and returns 1 to indicate the device is set up
  129.  *    successfully to perform DDP on this I/O, otherwise this returns 0.
  130.  *
  131.  * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
  132.  *    Called when the FCoE Initiator/Target is done with the DDPed I/O as
  133.  *    indicated by the FC exchange id 'xid', so the underlying device can
  134.  *    clean up and reuse resources for later DDP requests.
  135.  *
  136.  * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
  137.  *                  struct scatterlist *sgl, unsigned int sgc);
  138.  *    Called when the FCoE Target wants to initialize an I/O that
  139.  *    is a possible candidate for Direct Data Placement (DDP). The LLD can
  140.  *    perform necessary setup and returns 1 to indicate the device is set up
  141.  *    successfully to perform DDP on this I/O, otherwise this returns 0.
  142.  *
  143.  * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
  144.  *                   struct netdev_fcoe_hbainfo *hbainfo);
  145.  *    Called when the FCoE Protocol stack wants information on the underlying
  146.  *    device. This information is utilized by the FCoE protocol stack to
  147.  *    register attributes with Fiber Channel management service as per the
  148.  *    FC-GS Fabric Device Management Information(FDMI) specification.
  149.  *
  150.  * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
  151.  *    Called when the underlying device wants to override default World Wide
  152.  *    Name (WWN) generation mechanism in FCoE protocol stack to pass its own
  153.  *    World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
  154.  *    protocol stack to use.
  155.  *
  156.  *    RFS acceleration.
  157.  * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
  158.  *                u16 rxq_index, u32 flow_id);
  159.  *    Set hardware filter for RFS.  rxq_index is the target queue index;
  160.  *    flow_id is a flow ID to be passed to rps_may_expire_flow() later.
  161.  *    Return the filter ID on success, or a negative error code.
  162.  *
  163.  *    Slave management functions (for bridge, bonding, etc).
  164.  * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
  165.  *    Called to make another netdev an underling.
  166.  *
  167.  * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
  168.  *    Called to release previously enslaved netdev.
  169.  *
  170.  *      Feature/offload setting functions.
  171.  * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
  172.  *        netdev_features_t features);
  173.  *    Adjusts the requested feature flags according to device-specific
  174.  *    constraints, and returns the resulting flags. Must not modify
  175.  *    the device state.
  176.  *
  177.  * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
  178.  *    Called to update device configuration to new features. Passed
  179.  *    feature set might be less than what was returned by ndo_fix_features()).
  180.  *    Must return >0 or -errno if it changed dev->features itself.
  181.  *
  182.  * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
  183.  *              struct net_device *dev,
  184.  *              const unsigned char *addr, u16 flags)
  185.  *    Adds an FDB entry to dev for addr.
  186.  * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
  187.  *              struct net_device *dev,
  188.  *              const unsigned char *addr)
  189.  *    Deletes the FDB entry from dev coresponding to addr.
  190.  * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
  191.  *               struct net_device *dev, int idx)
  192.  *    Used to add FDB entries to dump requests. Implementers should add
  193.  *    entries to skb and update idx with the number of entries.
  194.  *
  195.  * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
  196.  * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
  197.  *                 struct net_device *dev, u32 filter_mask)
  198.  *
  199.  * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
  200.  *    Called to change device carrier. Soft-devices (like dummy, team, etc)
  201.  *    which do not represent real hardware may define this to allow their
  202.  *    userspace components to manage their virtual carrier state. Devices
  203.  *    that determine carrier state from physical hardware properties (eg
  204.  *    network cables) or protocol-dependent mechanisms (eg
  205.  *    USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
  206.  *
  207.  * int (*ndo_get_phys_port_id)(struct net_device *dev,
  208.  *                   struct netdev_phys_port_id *ppid);
  209.  *    Called to get ID of physical port of this device. If driver does
  210.  *    not implement this, it is assumed that the hw is not able to have
  211.  *    multiple net devices on single physical port.
  212.  *
  213.  * void (*ndo_add_vxlan_port)(struct  net_device *dev,
  214.  *                  sa_family_t sa_family, __be16 port);
  215.  *    Called by vxlan to notiy a driver about the UDP port and socket
  216.  *    address family that vxlan is listnening to. It is called only when
  217.  *    a new port starts listening. The operation is protected by the
  218.  *    vxlan_net->sock_lock.
  219.  *
  220.  * void (*ndo_del_vxlan_port)(struct  net_device *dev,
  221.  *                  sa_family_t sa_family, __be16 port);
  222.  *    Called by vxlan to notify the driver about a UDP port and socket
  223.  *    address family that vxlan is not listening to anymore. The operation
  224.  *    is protected by the vxlan_net->sock_lock.
  225.  *
  226.  * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
  227.  *                 struct net_device *dev)
  228.  *    Called by upper layer devices to accelerate switching or other
  229.  *    station functionality into hardware. 'pdev is the lowerdev
  230.  *    to use for the offload and 'dev' is the net device that will
  231.  *    back the offload. Returns a pointer to the private structure
  232.  *    the upper layer will maintain.
  233.  * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
  234.  *    Called by upper layer device to delete the station created
  235.  *    by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
  236.  *    the station and priv is the structure returned by the add
  237.  *    operation.
  238.  * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
  239.  *                      struct net_device *dev,
  240.  *                      void *priv);
  241.  *    Callback to use for xmit over the accelerated station. This
  242.  *    is used in place of ndo_start_xmit on accelerated net
  243.  *    devices.
  244.  */
  245. struct net_device_ops {
  246.     int            (*ndo_init)(struct net_device *dev);
  247.     void            (*ndo_uninit)(struct net_device *dev);
  248.     int            (*ndo_open)(struct net_device *dev);
  249.     int            (*ndo_stop)(struct net_device *dev);
  250.     netdev_tx_t        (*ndo_start_xmit) (struct sk_buff *skb,
  251.                            struct net_device *dev);
  252.     u16            (*ndo_select_queue)(struct net_device *dev,
  253.                             struct sk_buff *skb,
  254.                             void *accel_priv,
  255.                             select_queue_fallback_t fallback);
  256.     void            (*ndo_change_rx_flags)(struct net_device *dev,
  257.                                int flags);
  258.     void            (*ndo_set_rx_mode)(struct net_device *dev);
  259.     int            (*ndo_set_mac_address)(struct net_device *dev,
  260.                                void *addr);
  261.     int            (*ndo_validate_addr)(struct net_device *dev);
  262.     int            (*ndo_do_ioctl)(struct net_device *dev,
  263.                             struct ifreq *ifr, int cmd);
  264.     int            (*ndo_set_config)(struct net_device *dev,
  265.                               struct ifmap *map);
  266.     int            (*ndo_change_mtu)(struct net_device *dev,
  267.                           int new_mtu);
  268.     int            (*ndo_neigh_setup)(struct net_device *dev,
  269.                            struct neigh_parms *);
  270.     void            (*ndo_tx_timeout) (struct net_device *dev);
  271.  
  272.     struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
  273.                              struct rtnl_link_stats64 *storage);
  274.     struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
  275.  
  276.     int            (*ndo_vlan_rx_add_vid)(struct net_device *dev,
  277.                                __be16 proto, u16 vid);
  278.     int            (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
  279.                                 __be16 proto, u16 vid);
  280. #ifdef CONFIG_NET_POLL_CONTROLLER
  281.     void                    (*ndo_poll_controller)(struct net_device *dev);
  282.     int            (*ndo_netpoll_setup)(struct net_device *dev,
  283.                              struct netpoll_info *info,
  284.                              gfp_t gfp);
  285.     void            (*ndo_netpoll_cleanup)(struct net_device *dev);
  286. #endif
  287. #ifdef CONFIG_NET_RX_BUSY_POLL
  288.     int            (*ndo_busy_poll)(struct napi_struct *dev);
  289. #endif
  290.     int            (*ndo_set_vf_mac)(struct net_device *dev,
  291.                           int queue, u8 *mac);
  292.     int            (*ndo_set_vf_vlan)(struct net_device *dev,
  293.                            int queue, u16 vlan, u8 qos);
  294.     int            (*ndo_set_vf_tx_rate)(struct net_device *dev,
  295.                               int vf, int rate);
  296.     int            (*ndo_set_vf_spoofchk)(struct net_device *dev,
  297.                                int vf, bool setting);
  298.     int            (*ndo_get_vf_config)(struct net_device *dev,
  299.                              int vf,
  300.                              struct ifla_vf_info *ivf);
  301.     int            (*ndo_set_vf_link_state)(struct net_device *dev,
  302.                              int vf, int link_state);
  303.     int            (*ndo_set_vf_port)(struct net_device *dev,
  304.                            int vf,
  305.                            struct nlattr *port[]);
  306.     int            (*ndo_get_vf_port)(struct net_device *dev,
  307.                            int vf, struct sk_buff *skb);
  308.     int            (*ndo_setup_tc)(struct net_device *dev, u8 tc);
  309. #if IS_ENABLED(CONFIG_FCOE)
  310.     int            (*ndo_fcoe_enable)(struct net_device *dev);
  311.     int            (*ndo_fcoe_disable)(struct net_device *dev);
  312.     int            (*ndo_fcoe_ddp_setup)(struct net_device *dev,
  313.                               u16 xid,
  314.                               struct scatterlist *sgl,
  315.                               unsigned int sgc);
  316.     int            (*ndo_fcoe_ddp_done)(struct net_device *dev,
  317.                              u16 xid);
  318.     int            (*ndo_fcoe_ddp_target)(struct net_device *dev,
  319.                                u16 xid,
  320.                                struct scatterlist *sgl,
  321.                                unsigned int sgc);
  322.     int            (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
  323.                             struct netdev_fcoe_hbainfo *hbainfo);
  324. #endif
  325.  
  326. #if IS_ENABLED(CONFIG_LIBFCOE)
  327. #define NETDEV_FCOE_WWNN 0
  328. #define NETDEV_FCOE_WWPN 1
  329.     int            (*ndo_fcoe_get_wwn)(struct net_device *dev,
  330.                             u64 *wwn, int type);
  331. #endif
  332.  
  333. #ifdef CONFIG_RFS_ACCEL
  334.     int            (*ndo_rx_flow_steer)(struct net_device *dev,
  335.                              const struct sk_buff *skb,
  336.                              u16 rxq_index,
  337.                              u32 flow_id);
  338. #endif
  339.     int            (*ndo_add_slave)(struct net_device *dev,
  340.                          struct net_device *slave_dev);
  341.     int            (*ndo_del_slave)(struct net_device *dev,
  342.                          struct net_device *slave_dev);
  343.     netdev_features_t    (*ndo_fix_features)(struct net_device *dev,
  344.                             netdev_features_t features);
  345.     int            (*ndo_set_features)(struct net_device *dev,
  346.                             netdev_features_t features);
  347.     int            (*ndo_neigh_construct)(struct neighbour *n);
  348.     void            (*ndo_neigh_destroy)(struct neighbour *n);
  349.  
  350.     int            (*ndo_fdb_add)(struct ndmsg *ndm,
  351.                            struct nlattr *tb[],
  352.                            struct net_device *dev,
  353.                            const unsigned char *addr,
  354.                            u16 flags);
  355.     int            (*ndo_fdb_del)(struct ndmsg *ndm,
  356.                            struct nlattr *tb[],
  357.                            struct net_device *dev,
  358.                            const unsigned char *addr);
  359.     int            (*ndo_fdb_dump)(struct sk_buff *skb,
  360.                         struct netlink_callback *cb,
  361.                         struct net_device *dev,
  362.                         int idx);
  363.  
  364.     int            (*ndo_bridge_setlink)(struct net_device *dev,
  365.                               struct nlmsghdr *nlh);
  366.     int            (*ndo_bridge_getlink)(struct sk_buff *skb,
  367.                               u32 pid, u32 seq,
  368.                               struct net_device *dev,
  369.                               u32 filter_mask);
  370.     int            (*ndo_bridge_dellink)(struct net_device *dev,
  371.                               struct nlmsghdr *nlh);
  372.     int            (*ndo_change_carrier)(struct net_device *dev,
  373.                               bool new_carrier);
  374.     int            (*ndo_get_phys_port_id)(struct net_device *dev,
  375.                             struct netdev_phys_port_id *ppid);
  376.     void            (*ndo_add_vxlan_port)(struct  net_device *dev,
  377.                               sa_family_t sa_family,
  378.                               __be16 port);
  379.     void            (*ndo_del_vxlan_port)(struct  net_device *dev,
  380.                               sa_family_t sa_family,
  381.                               __be16 port);
  382.  
  383.     void*            (*ndo_dfwd_add_station)(struct net_device *pdev,
  384.                             struct net_device *dev);
  385.     void            (*ndo_dfwd_del_station)(struct net_device *pdev,
  386.                             void *priv);
  387.  
  388.     netdev_tx_t        (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
  389.                             struct net_device *dev,
  390.                             void *priv);
  391. };

14.2 注册与注销

14.2.1 注册与注销

register时，net_device的net_device_ops的ndo_init()会执行。

  1. /**
  2.  *    register_netdev    - register a network device
  3.  *    @dev: device to register
  4.  *
  5.  *    Take a completed network device structure and add it to the kernel
  6.  *    interfaces. A %NETDEV_REGISTER message is sent to the netdev notifier
  7.  *    chain. 0 is returned on success. A negative errno code is returned
  8.  *    on a failure to set up the device, or if the name is a duplicate.
  9.  *
  10.  *    This is a wrapper around register_netdevice that takes the rtnl semaphore
  11.  *    and expands the device name if you passed a format string to
  12.  *    alloc_netdev.
  13.  */
  14. int register_netdev(struct net_device *dev);
  15.  
  16. /**
  17.  *    unregister_netdev - remove device from the kernel
  18.  *    @dev: device
  19.  *
  20.  *    This function shuts down a device interface and removes it
  21.  *    from the kernel tables.
  22.  *
  23.  *    This is just a wrapper for unregister_netdevice that takes
  24.  *    the rtnl semaphore.  In general you want to use this and not
  25.  *    unregister_netdevice.
  26.  */
  27. void unregister_netdev(struct net_device *dev);

14.2.2 申请和释放

可以自己定义net_device结构体，也可以动态申请和释放空间。

  1. // sizeof_priv是自定义数据结构的大小
    // setup是动态申请后，自动执行的初始化函数，下面两个ether申请函数，都有默认的setup函数
    #define alloc_netdev(sizeof_priv, name, setup) \
  2.     alloc_netdev_mqs(sizeof_priv, name, setup, , )
  3. #define alloc_etherdev(sizeof_priv) alloc_etherdev_mq(sizeof_priv, 1)
  4. #define alloc_etherdev_mq(sizeof_priv, count) alloc_etherdev_mqs(sizeof_priv, count, count)
  5.  
  6. /**
  7.  * alloc_etherdev_mqs - Allocates and sets up an Ethernet device
  8.  * @sizeof_priv: Size of additional driver-private structure to be allocated
  9.  *    for this Ethernet device
  10.  * @txqs: The number of TX queues this device has.
  11.  * @rxqs: The number of RX queues this device has.
  12.  *
  13.  * Fill in the fields of the device structure with Ethernet-generic
  14.  * values. Basically does everything except registering the device.
  15.  *
  16.  * Constructs a new net device, complete with a private data area of
  17.  * size (sizeof_priv).  A 32-byte (not bit) alignment is enforced for
  18.  * this private data area.
  19.  */
  20.  
  21. struct net_device *alloc_etherdev_mqs(int sizeof_priv, unsigned int txqs,
  22.                       unsigned int rxqs)
  23. {
  24.     return alloc_netdev_mqs(sizeof_priv, "eth%d", ether_setup, txqs, rxqs);
  25. }
  26.  
  27. /**
  28.  *    alloc_netdev_mqs - allocate network device
  29.  *    @sizeof_priv:    size of private data to allocate space for
  30.  *    @name:        device name format string
  31.  *    @setup:        callback to initialize device
  32.  *    @txqs:        the number of TX subqueues to allocate
  33.  *    @rxqs:        the number of RX subqueues to allocate
  34.  *
  35.  *    Allocates a struct net_device with private data area for driver use
  36.  *    and performs basic initialization.  Also allocates subqueue structs
  37.  *    for each queue on the device.
  38.  */
  39. struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
  40.         void (*setup)(struct net_device *),
  41.         unsigned int txqs, unsigned int rxqs)
  42. {
  43.     struct net_device *dev;
  44.     size_t alloc_size;
  45.     struct net_device *p;
  46.  
  47.     BUG_ON(strlen(name) >= sizeof(dev->name));
  48.  
  49.     if (txqs < ) {
  50.         pr_err("alloc_netdev: Unable to allocate device with zero queues\n");
  51.         return NULL;
  52.     }
  53.  
  54. #ifdef CONFIG_SYSFS
  55.     if (rxqs < ) {
  56.         pr_err("alloc_netdev: Unable to allocate device with zero RX queues\n");
  57.         return NULL;
  58.     }
  59. #endif
  60.  
  61.     alloc_size = sizeof(struct net_device);
  62.     if (sizeof_priv) {
  63.         /* ensure 32-byte alignment of private area */
  64.         alloc_size = ALIGN(alloc_size, NETDEV_ALIGN);
  65.         alloc_size += sizeof_priv;
  66.     }
  67.     /* ensure 32-byte alignment of whole construct */
  68.     alloc_size += NETDEV_ALIGN - ;
  69.  
  70.     p = kzalloc(alloc_size, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
  71.     if (!p)
  72.         p = vzalloc(alloc_size);
  73.     if (!p)
  74.         return NULL;
  75.  
  76.     dev = PTR_ALIGN(p, NETDEV_ALIGN);
  77.     dev->padded = (char *)dev - (char *)p;
  78.  
  79.     dev->pcpu_refcnt = alloc_percpu(int);
  80.     if (!dev->pcpu_refcnt)
  81.         goto free_dev;
  82.  
  83.     if (dev_addr_init(dev))
  84.         goto free_pcpu;
  85.  
  86.     dev_mc_init(dev);
  87.     dev_uc_init(dev);
  88.  
  89.     dev_net_set(dev, &init_net);
  90.  
  91.     dev->gso_max_size = GSO_MAX_SIZE;
  92.     dev->gso_max_segs = GSO_MAX_SEGS;
  93.  
  94.     INIT_LIST_HEAD(&dev->napi_list);
  95.     INIT_LIST_HEAD(&dev->unreg_list);
  96.     INIT_LIST_HEAD(&dev->close_list);
  97.     INIT_LIST_HEAD(&dev->link_watch_list);
  98.     INIT_LIST_HEAD(&dev->adj_list.upper);
  99.     INIT_LIST_HEAD(&dev->adj_list.lower);
  100.     INIT_LIST_HEAD(&dev->all_adj_list.upper);
  101.     INIT_LIST_HEAD(&dev->all_adj_list.lower);
  102.     dev->priv_flags = IFF_XMIT_DST_RELEASE;
  103.     setup(dev);
  104.  
  105.     dev->num_tx_queues = txqs;
  106.     dev->real_num_tx_queues = txqs;
  107.     if (netif_alloc_netdev_queues(dev))
  108.         goto free_all;
  109.  
  110. #ifdef CONFIG_SYSFS
  111.     dev->num_rx_queues = rxqs;
  112.     dev->real_num_rx_queues = rxqs;
  113.     if (netif_alloc_rx_queues(dev))
  114.         goto free_all;
  115. #endif
  116.  
  117.     strcpy(dev->name, name);
  118.     dev->group = INIT_NETDEV_GROUP;
  119.     if (!dev->ethtool_ops)
  120.         dev->ethtool_ops = &default_ethtool_ops;
  121.     return dev;
  122.  
  123. free_all:
  124.     free_netdev(dev);
  125.     return NULL;
  126.  
  127. free_pcpu:
  128.     free_percpu(dev->pcpu_refcnt);
  129.     netif_free_tx_queues(dev);
  130. #ifdef CONFIG_SYSFS
  131.     kfree(dev->_rx);
  132. #endif
  133.  
  134. free_dev:
  135.     netdev_freemem(dev);
  136.     return NULL;
  137. }
  138.  
  139. //  释放net_device
    void free_netdev(struct net_device *dev);

模板：

  1. static int xxx_register(void)
  2. {
  3.     ...
  4.  
  5.     /* 分配 net_device 结构体并对其成员赋值 */
  6.     xxx_dev = alloc_netdev(sizeof(struct xxx_priv), "sn%d", xxx_init);
  7.     if (xxx_dev == NULL)
  8.         ... /* 分配 net_device 失败 */
  9.  
  10.     /* 注册 net_device 结构体 */
  11.     if ((result = register_netdev(xxx_dev)))
  12.     ...
  13. }
  14.  
  15. static void xxx_unregister(void)
  16. {
  17.     ...
  18.     /* 注销 net_device 结构体 */
  19.     unregister_netdev(xxx_dev);
  20.     /* 释放 net_device 结构体 */
  21.     free_netdev(xxx_dev);
  22. }

14.3 初始化ndo_init()

向内核register时，ndo_init()函数会被执行。

  1. int register_netdev(struct net_device *dev)
  2. {
  3.     int err;
  4.  
  5.     rtnl_lock();
  6.     err = register_netdevice(dev);
  7.     rtnl_unlock();
  8.     return err;
  9. }
  10.  
  11. int register_netdevice(struct net_device *dev)
  12. {
  13.     ...
  14.     /* Init, if this function is available */
  15.     if (dev->netdev_ops->ndo_init) {
  16.         ret = dev->netdev_ops->ndo_init(dev);
  17.         if (ret) {
  18.             if (ret > )
  19.                 ret = -EIO;
  20.             goto out;
  21.         }
  22.     }
  23.     ...
  24. }

ndo_init()要干的事：

  • 准备硬件
  • 初始化net_device结构体的相关内容
  • 获取私有指针，并初始化

ndo_init()模板：

  1. xxx_netdev_ops.ndo_init = xxx_init;
  1. void xxx_init(struct net_device *dev)
  2. {
  3.     /* 设备的私有信息结构体 */
  4.     struct xxx_priv *priv;
  5.  
  6.     /* 检查设备是否存在和设备所使用的硬件资源 */
  7.     xxx_hw_init();
  8.  
  9.     /* 初始化以太网设备的公用成员 */
  10.     ether_setup(dev);
  11.  
  12.     /* 设置设备的成员函数指针 */
  13.     ndev->netdev_ops = &xxx_netdev_ops;
  14.     ndev->ethtool_ops = &xxx_ethtool_ops;
  15.     dev->watchdog_timeo = timeout;
  16.  
  17.     /* 取得私有信息， 并初始化它 */
  18.     priv = netdev_priv(dev);
  19.     ... /* 初始化设备私有数据区 */
  20. }

14.4 打开和释放ndo_open()/ndo_stop()

ndo_open()的工作：

  • 使能硬件资源，申请IO区域、中断和DMA通道等；
  • 调用netif_start_queue()函数，激活设备发送队列

ndo_stop()的工作：

  • 调用netif_stop_queue()函数，停止设备发送队列
  • 释放IO区域、中断和DMA资源　　
  1. 原来这两个函数只针对 tx queue
    /**
  2.  *    netif_start_queue - allow transmit
  3.  *    @dev: network device
  4.  *
  5.  *    Allow upper layers to call the device hard_start_xmit routine.
  6.  */
  7. static inline void netif_start_queue(struct net_device *dev)
  8. {
  9.     netif_tx_start_queue(netdev_get_tx_queue(dev, ));
  10. }
  11. /**
  12.  *    netif_stop_queue - stop transmitted packets
  13.  *    @dev: network device
  14.  *
  15.  *    Stop upper layers calling the device hard_start_xmit routine.
  16.  *    Used for flow control when transmit resources are unavailable.
  17.  */
  18. static inline void netif_stop_queue(struct net_device *dev)
  19. {
  20.     netif_tx_stop_queue(netdev_get_tx_queue(dev, ));
  21. }

/**
* netif_wake_queue - restart transmit
* @dev: network device
*
* Allow upper layers to call the device hard_start_xmit routine.
* Used for flow control when transmit resources are available.
*/
static inline void netif_wake_queue(struct net_device *dev)
{
　　netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
}

  1.  

模板：

  1.  
  1. xxx_netdev_ops.ndo_open = xxx_open;
  1. xxx_netdev_ops.ndo_stop = xxx_stop;
  1. static int xxx_open(struct net_device *dev)
  2. {
  3.     /* 申请端口、 IRQ 等， 类似于 fops->open */
  4.     ret = request_irq(dev->irq, &xxx_interrupt, , dev->name, dev);
  5.     ...
  6.     netif_start_queue(dev);
  7.     ...
  8. }
  9.  
  10. static int xxx_stop(struct net_device *dev)
  11. {
  12.     /* 释放端口、 IRQ 等， 类似于 fops->close */
  13.     free_irq(dev->irq, dev);
  14.     ...
  15.     netif_stop_queue(dev); /* can't transmit any more */
  16.     ...
  17. }

14.5 发送 ndo_start_xmit()/ndo_tx_timeout()

ndo_start_xmit()主要流程：

  • 解析sk_buff，缓存有效数据
  • 校验数据长度，若小于ETH_ZLEN（60，不包含FCS），则尾部填充0
  • 控制硬件发送数据
  1. int xxx_tx(struct sk_buff *skb, struct net_device *dev)
  2. {
  3.     int len;
  4.     char *data, shortpkt[ETH_ZLEN];
  5.     if (xxx_send_available(...)) { /* 发送队列未满， 可以发送 */
  6.         /* 获得有效数据指针和长度 */
  7.         data = skb->data;
  8.         len = skb->len;
  9.         if (len < ETH_ZLEN) {
  10.         　　/* 如果帧长小于以太网帧最小长度， 补 0 */
  11.         　　memset(shortpkt, , ETH_ZLEN);
  12.        　　 memcpy(shortpkt, skb->data, skb->len);
  13.         　　len = ETH_ZLEN;
  14.        　　 data = shortpkt;
    　　　　  }
  15.  
  16.     　　 dev->trans_start = jiffies; /* 记录发送时间戳 */
  17.  
  18.    　　 if (avail) {/* 设置硬件寄存器， 让硬件把数据包发送出去 */
  19.         　　xxx_hw_tx(data, len, dev);
  20.     　　} else {
  21.         　　netif_stop_queue(dev);　　// 不一定非要这么搞，返回busy也可以，如果调用了，需要在TX结束中断或者超时中断里唤醒
  22.         　　...
  23.     　　}
  24. 　　}
    }
  25.  
  26. void xxx_tx_timeout(struct net_device *dev)
    {
    　　...
    　　netif_wake_queue(dev); /* 重新启动设备发送队列 */
  27. }

14.6 接收

没有固定形式，在需要的地方调用netif_rx()即可。

  1. static void xxx_interrupt(int irq, void *dev_id)
  2. {
  3.     ...
  4.     switch (status &ISQ_EVENT_MASK) {
  5.     case ISQ_RECEIVER_EVENT:
  6.         /* 获取数据包 */
  7.         xxx_rx(dev);
  8.         break;
  9.     /* 其他类型的中断 */
  10.     }
  11. }
  12.  
  13. static void xxx_rx(struct xxx_device *dev)
  14. {
  15.     ...
  16.     length = get_rev_len (...);
  17.     /* 分配新的套接字缓冲区 */
  18.     skb = dev_alloc_skb(length + );
  19.  
  20.     skb_reserve(skb, ); /* 对齐 */
  21.     skb->dev = dev;
  22.  
  23.     /* 读取硬件上接收到的数据 */
  24.     insw(ioaddr + RX_FRAME_PORT, skb_put(skb, length), length >> );
  25.     if (length &)
  26.     skb->data[length - ] = inw(ioaddr + RX_FRAME_PORT);
  27.  
  28.     /* 获取上层协议类型 */
  29.     skb->protocol = eth_type_trans(skb, dev);
  30.  
  31.     /* 把数据包交给上层 */
  32.     netif_rx(skb);
  33.  
  34.     /* 记录接收时间戳 */
  35.     dev->last_rx = jiffies;
  36.     ...
  37. }

14.7 连接状态

一般在定时中断里，检查并更新连接状态。

  1. static inline bool netif_carrier_ok(const struct net_device *dev);　　// 连接是否ok
  2. void netif_carrier_on(struct net_device *dev);　　　　　　　　　　　　　 // 改变连接状态，on
  3. void netif_carrier_off(struct net_device *dev);　　　　　　　　　　　　　// 改变连接状态，off

14.8 参数设置和统计数据

参数设置可以通过ioctl()，传入的描述符为socket，linux对命令做了统一规定，如下：

  1. /*路径： include/uapi/linux/sockios.h */
  2. /*
  3.  * INET        An implementation of the TCP/IP protocol suite for the LINUX
  4.  *        operating system.  INET is implemented using the  BSD Socket
  5.  *        interface as the means of communication with the user level.
  6.  *
  7.  *        Definitions of the socket-level I/O control calls.
  8.  *
  9.  * Version:    @(#)sockios.h    1.0.2    03/09/93
  10.  *
  11.  * Authors:    Ross Biro
  12.  *        Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
  13.  *
  14.  *        This program is free software; you can redistribute it and/or
  15.  *        modify it under the terms of the GNU General Public License
  16.  *        as published by the Free Software Foundation; either version
  17.  *        2 of the License, or (at your option) any later version.
  18.  */
  19. #ifndef _LINUX_SOCKIOS_H
  20. #define _LINUX_SOCKIOS_H
  21.  
  22. #include <asm/sockios.h>
  23.  
  24. /* Linux-specific socket ioctls */
  25. #define SIOCINQ        FIONREAD
  26. #define SIOCOUTQ    TIOCOUTQ        /* output queue size (not sent + not acked) */
  27.  
  28. /* Routing table calls. */
  29. #define SIOCADDRT    0x890B        /* add routing table entry    */
  30. #define SIOCDELRT    0x890C        /* delete routing table entry    */
  31. #define SIOCRTMSG    0x890D        /* call to routing system    */
  32.  
  33. /* Socket configuration controls. */
  34. #define SIOCGIFNAME    0x8910        /* get iface name        */
  35. #define SIOCSIFLINK    0x8911        /* set iface channel        */
  36. #define SIOCGIFCONF    0x8912        /* get iface list        */
  37. #define SIOCGIFFLAGS    0x8913        /* get flags            */
  38. #define SIOCSIFFLAGS    0x8914        /* set flags            */
  39. #define SIOCGIFADDR    0x8915        /* get PA address        */
  40. #define SIOCSIFADDR    0x8916        /* set PA address        */
  41. #define SIOCGIFDSTADDR    0x8917        /* get remote PA address    */
  42. #define SIOCSIFDSTADDR    0x8918        /* set remote PA address    */
  43. #define SIOCGIFBRDADDR    0x8919        /* get broadcast PA address    */
  44. #define SIOCSIFBRDADDR    0x891a        /* set broadcast PA address    */
  45. #define SIOCGIFNETMASK    0x891b        /* get network PA mask        */
  46. #define SIOCSIFNETMASK    0x891c        /* set network PA mask        */
  47. #define SIOCGIFMETRIC    0x891d        /* get metric            */
  48. #define SIOCSIFMETRIC    0x891e        /* set metric            */
  49. #define SIOCGIFMEM    0x891f        /* get memory address (BSD)    */
  50. #define SIOCSIFMEM    0x8920        /* set memory address (BSD)    */
  51. #define SIOCGIFMTU    0x8921        /* get MTU size            */
  52. #define SIOCSIFMTU    0x8922        /* set MTU size            */
  53. #define SIOCSIFNAME    0x8923        /* set interface name */
  54. #define    SIOCSIFHWADDR    0x8924        /* set hardware address     */
  55. #define SIOCGIFENCAP    0x8925        /* get/set encapsulations       */
  56. #define SIOCSIFENCAP    0x8926
  57. #define SIOCGIFHWADDR    0x8927        /* Get hardware address        */
  58. #define SIOCGIFSLAVE    0x8929        /* Driver slaving support    */
  59. #define SIOCSIFSLAVE    0x8930
  60. #define SIOCADDMULTI    0x8931        /* Multicast address lists    */
  61. #define SIOCDELMULTI    0x8932
  62. #define SIOCGIFINDEX    0x8933        /* name -> if_index mapping    */
  63. #define SIOGIFINDEX    SIOCGIFINDEX    /* misprint compatibility :-)    */
  64. #define SIOCSIFPFLAGS    0x8934        /* set/get extended flags set    */
  65. #define SIOCGIFPFLAGS    0x8935
  66. #define SIOCDIFADDR    0x8936        /* delete PA address        */
  67. #define    SIOCSIFHWBROADCAST    0x8937    /* set hardware broadcast addr    */
  68. #define SIOCGIFCOUNT    0x8938        /* get number of devices */
  69.  
  70. #define SIOCGIFBR    0x8940        /* Bridging support        */
  71. #define SIOCSIFBR    0x8941        /* Set bridging options     */
  72.  
  73. #define SIOCGIFTXQLEN    0x8942        /* Get the tx queue length    */
  74. #define SIOCSIFTXQLEN    0x8943        /* Set the tx queue length     */
  75.  
  76. /* SIOCGIFDIVERT was:    0x8944        Frame diversion support */
  77. /* SIOCSIFDIVERT was:    0x8945        Set frame diversion options */
  78.  
  79. #define SIOCETHTOOL    0x8946        /* Ethtool interface        */
  80.  
  81. #define SIOCGMIIPHY    0x8947        /* Get address of MII PHY in use. */
  82. #define SIOCGMIIREG    0x8948        /* Read MII PHY register.    */
  83. #define SIOCSMIIREG    0x8949        /* Write MII PHY register.    */
  84.  
  85. #define SIOCWANDEV    0x894A        /* get/set netdev parameters    */
  86.  
  87. #define SIOCOUTQNSD    0x894B        /* output queue size (not sent only) */
  88.  
  89. /* ARP cache control calls. */
  90.             /*  0x8950 - 0x8952  * obsolete calls, don't re-use */
  91. #define SIOCDARP    0x8953        /* delete ARP table entry    */
  92. #define SIOCGARP    0x8954        /* get ARP table entry        */
  93. #define SIOCSARP    0x8955        /* set ARP table entry        */
  94.  
  95. /* RARP cache control calls. */
  96. #define SIOCDRARP    0x8960        /* delete RARP table entry    */
  97. #define SIOCGRARP    0x8961        /* get RARP table entry        */
  98. #define SIOCSRARP    0x8962        /* set RARP table entry        */
  99.  
  100. /* Driver configuration calls */
  101.  
  102. #define SIOCGIFMAP    0x8970        /* Get device parameters    */
  103. #define SIOCSIFMAP    0x8971        /* Set device parameters    */
  104.  
  105. /* DLCI configuration calls */
  106.  
  107. #define SIOCADDDLCI    0x8980        /* Create new DLCI device    */
  108. #define SIOCDELDLCI    0x8981        /* Delete DLCI device        */
  109.  
  110. #define SIOCGIFVLAN    0x8982        /* 802.1Q VLAN support        */
  111. #define SIOCSIFVLAN    0x8983        /* Set 802.1Q VLAN options     */
  112.  
  113. /* bonding calls */
  114.  
  115. #define SIOCBONDENSLAVE    0x8990        /* enslave a device to the bond */
  116. #define SIOCBONDRELEASE 0x8991        /* release a slave from the bond*/
  117. #define SIOCBONDSETHWADDR      0x8992    /* set the hw addr of the bond  */
  118. #define SIOCBONDSLAVEINFOQUERY 0x8993   /* rtn info about slave state   */
  119. #define SIOCBONDINFOQUERY      0x8994    /* rtn info about bond state    */
  120. #define SIOCBONDCHANGEACTIVE   0x8995   /* update to a new active slave */
  121.  
  122. /* bridge calls */
  123. #define SIOCBRADDBR     0x89a0        /* create new bridge device     */
  124. #define SIOCBRDELBR     0x89a1        /* remove bridge device         */
  125. #define SIOCBRADDIF    0x89a2        /* add interface to bridge      */
  126. #define SIOCBRDELIF    0x89a3        /* remove interface from bridge */
  127.  
  128. /* hardware time stamping: parameters in linux/net_tstamp.h */
  129. #define SIOCSHWTSTAMP    0x89b0        /* set and get config        */
  130. #define SIOCGHWTSTAMP    0x89b1        /* get config            */
  131.  
  132. /* Device private ioctl calls */
  133.  
  134. /*
  135.  *    These 16 ioctls are available to devices via the do_ioctl() device
  136.  *    vector. Each device should include this file and redefine these names
  137.  *    as their own. Because these are device dependent it is a good idea
  138.  *    _NOT_ to issue them to random objects and hope.
  139.  *
  140.  *    THESE IOCTLS ARE _DEPRECATED_ AND WILL DISAPPEAR IN 2.5.X -DaveM
  141.  */
  142.  
  143. #define SIOCDEVPRIVATE    0x89F0    /* to 89FF */
  144.  
  145. /*
  146.  *    These 16 ioctl calls are protocol private
  147.  */
  148.  
  149. #define SIOCPROTOPRIVATE 0x89E0 /* to 89EF */
  150. #endif    /* _LINUX_SOCKIOS_H */
  1. ndo_get_stats()模板，在程序合适的地方对各种计数进行设置即可。
  1. struct net_device_stats *xxx_stats(struct net_device *dev)
  2. {
  3.     …
  4.     return &dev->stats;
  5. }
  6.  
  7. struct net_device_stats
  8. {
  9.     unsigned long rx_packets; /* 收到的数据包数 */
  10.     unsigned long tx_packets; /* 发送的数据包数 */
  11.     unsigned long rx_bytes; /* 收到的字节数 */
  12.     unsigned long tx_bytes; /* 发送的字节数 */
  13.     unsigned long rx_errors; /* 收到的错误数据包数 */
  14.     unsigned long tx_errors; /* 发生发送错误的数据包数 */
  15.      ...
  16. };

14.9 DM9000实例

14.10 总结

《linux设备驱动开发详解》笔记——14 linux网络设备驱动的更多相关文章

  1. linux设备驱动开发详解 笔记

      在目录的 Makefile 中关于 RTC_DRV_S3C 的编译脚本为: obj -$(CONFIG_RTC_DRV_S3C) += rtc-s3c.o 上述脚本意味着如果 RTC_DRV_S3 ...

  2. Linux设备驱动开发详解-Note(11)--- Linux 文件系统与设备文件系统(3)

    Linux 文件系统与设备文件系统(3) 成于坚持,败于止步 sysfs 文件系统与 Linux 设备模型 1．sysfs 文件系统 Linux 2.6 内核引入了 sysfs 文件系统,sysfs ...

  3. Linux设备驱动开发详解-Note(5)---Linux 内核及内核编程(1)

    Linux 内核及内核编程(1) 成于坚持,败于止步 Linux 2.6 内核的特点 Linux 2.6 相对于 Linux 2.4 有相当大的改进,主要体现在如下几个方面. 1．新的调度器 2.6 ...

  4. Linux设备驱动开发详解

    Linux设备驱动开发详解 http://download.csdn.net/detail/wuyouzi067/9581380

  5. 嵌入式Linux应用程序开发详解------（创建守护进程）

    嵌入式Linux应用程序开发详解 华清远见 本文只是阅读文摘. 创建一个守护进程的步骤: 1.创建一个子进程,然后退出父进程: 2.在子进程中使用创建新会话---setsid(): 3.改变当前工作目 ...

  6. 《linux设备驱动开发详解》笔记——6字符设备驱动

    6.1 字符设备驱动结构 先看看字符设备驱动的架构: 6.1.1 cdev cdev结构体是字符设备的核心数据结构,用于描述一个字符设备,cdev定义如下: #include <linux/cd ...

  7. 《linux设备驱动开发详解》笔记——10中断与时钟

    10.1 中断与定时器 中断一般有如下类型: 内部中断和外部中断:内部中断来自CPU,例如软件中断指令.溢出.除0错误等:外部中断有外部设备触发 可屏蔽中断和不可屏蔽中断 向量中断和非向量中断,ARM ...

  8. 《linux设备驱动开发详解》笔记——8阻塞与非阻塞IO

    8.1 阻塞与非阻塞IO 8.1.0 概述 阻塞:访问设备时,若不能获取资源,则进程挂起,进入睡眠状态:也就是进入等待队列 非阻塞:不能获取资源时,不睡眠,要么退出.要么一直查询:直接退出且无资源时, ...

  9. 《Linux设备驱动开发详解（第2版）》配套视频登录51cto教育频道

    http://edu.51cto.com/course/course_id-379-page-1.html http://edu.51cto.com/course/course_id-379-page ...

随机推荐

  1. php：比较两个txt文件，格式如下，分别取出a.txt有的b.txt没有的，b.txt有的a.txt没有的及两个都有的

    <?php /*比较两个txt文件,格式如下,分别取出a.txt有的b.txt没有的,b.txt有的a.txt没有的及两个都有的 * a.txt: * A * B * C * D * b.txt ...

  2. Identity Service

    Identity Service - 解析微软微服务架构eShopOnContainers(二)   接上一篇,众所周知一个网站的用户登录是非常重要,一站式的登录(SSO)也成了大家讨论的热点.微软在 ...

  3. Core中使用Razor视图引擎渲染视图为字符串 阅读目录

    Core中使用Razor视图引擎渲染视图为字符串 } <!DOCTYPE html> <html> <head> <title>Render view ...

  4. PIX 防火墙

    ---恢复内容开始--- 一 , PIX 防火墙的认识 PIX 是cisco 的硬件防火墙 硬件防火墙的工作速度快,使用方便. PIX 有很多型号,并发连接数是PIX防火墙的重要参数   PIX 25 ...

  5. Win10+VirtualBox+Openstack Mitaka

    首先VirtualBox安装的话,没有什么可演示的,去官网(https://www.virtualbox.org/wiki/Downloads)下载,或者可以去(https://www.virtual ...

  6. restful 风格 加上springmvc

    一.spring 版本:spring-framework-3.2.7.RELEASE 二.所需其它Jar包: 三.主要代码: ? 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 ...

  7. Ubuntu常用指令集

    Ubuntu Linux 操作系统常用命令详细介绍 ( 1)Udo apt-get install 软件名 安装软件命令 sudo nautilus 打开文件(有 root 权限)su root 切换 ...

  8. Linux大棚版vimrc配置

    Linux大棚版vimrc配置—V2.0版本,如下: [shell] $cat .vimrc “== “Author :roc “Website:roclinux.cn “Version:2.0 “= ...

  9. mysql 存储过程变量及循环的使用

    1.用游标循环 BEGIN -- 定义变量 -- 定义done DECLARE done INT; -- 定义 ammeter_id_bl DECLARE ammeter_id_bl DOUBLE; ...

  10. C#、VSTO讀取Excel類

    之前寫的類存在Excel進程不能結束的Bug,重寫ExcelReader類,類實例清理時Excel進程自動結束. class ExcelReader { // Excel Object public ...