root/include/linux/netdevice.h

/* [<][>][^][v][top][bottom][index][help] */

INCLUDED FROM


DEFINITIONS

This source file includes following definitions.
  1. dev_xmit_complete
  2. napi_disable_pending
  3. napi_schedule
  4. napi_schedule_irqoff
  5. napi_reschedule
  6. napi_complete
  7. napi_enable
  8. napi_synchronize
  9. napi_if_scheduled_mark_missed
  10. net_has_fallback_tunnels
  11. netdev_queue_numa_node_read
  12. netdev_queue_numa_node_write
  13. rps_record_sock_flow
  14. netdev_phys_item_id_same
  15. netif_elide_gro
  16. netdev_get_prio_tc_map
  17. netdev_set_prio_tc_map
  18. netdev_get_num_tc
  19. netdev_get_sb_channel
  20. netdev_get_tx_queue
  21. skb_get_tx_queue
  22. netdev_for_each_tx_queue
  23. netdev_get_fwd_headroom
  24. netdev_set_rx_headroom
  25. netdev_reset_rx_headroom
  26. dev_net
  27. dev_net_set
  28. netdev_priv
  29. netif_tx_napi_add
  30. gro_recursion_inc_test
  31. call_gro_receive
  32. call_gro_receive_sk
  33. netdev_notifier_info_init
  34. netdev_notifier_info_to_dev
  35. netdev_notifier_info_to_extack
  36. next_net_device
  37. next_net_device_rcu
  38. first_net_device
  39. first_net_device_rcu
  40. unregister_netdevice
  41. skb_gro_offset
  42. skb_gro_len
  43. skb_gro_pull
  44. skb_gro_header_fast
  45. skb_gro_header_hard
  46. skb_gro_frag0_invalidate
  47. skb_gro_header_slow
  48. skb_gro_network_header
  49. skb_gro_postpull_rcsum
  50. skb_at_gro_remcsum_start
  51. __skb_gro_checksum_validate_needed
  52. __skb_gro_checksum_validate_complete
  53. skb_gro_incr_csum_unnecessary
  54. __skb_gro_checksum_convert_check
  55. __skb_gro_checksum_convert
  56. skb_gro_remcsum_init
  57. skb_gro_remcsum_process
  58. skb_gro_remcsum_cleanup
  59. skb_gro_flush_final
  60. skb_gro_flush_final_remcsum
  61. skb_gro_flush_final
  62. skb_gro_flush_final_remcsum
  63. dev_hard_header
  64. dev_parse_header
  65. dev_parse_header_protocol
  66. dev_validate_header
  67. unregister_gifconf
  68. input_queue_head_incr
  69. input_queue_tail_incr_save
  70. dev_recursion_level
  71. dev_xmit_recursion
  72. dev_xmit_recursion_inc
  73. dev_xmit_recursion_dec
  74. netif_tx_schedule_all
  75. netif_tx_start_queue
  76. netif_start_queue
  77. netif_tx_start_all_queues
  78. netif_wake_queue
  79. netif_tx_wake_all_queues
  80. netif_tx_stop_queue
  81. netif_stop_queue
  82. netif_tx_queue_stopped
  83. netif_queue_stopped
  84. netif_xmit_stopped
  85. netif_xmit_frozen_or_stopped
  86. netif_xmit_frozen_or_drv_stopped
  87. netdev_txq_bql_enqueue_prefetchw
  88. netdev_txq_bql_complete_prefetchw
  89. netdev_tx_sent_queue
  90. __netdev_tx_sent_queue
  91. netdev_sent_queue
  92. __netdev_sent_queue
  93. netdev_tx_completed_queue
  94. netdev_completed_queue
  95. netdev_tx_reset_queue
  96. netdev_reset_queue
  97. netdev_cap_txqueue
  98. netif_running
  99. netif_start_subqueue
  100. netif_stop_subqueue
  101. __netif_subqueue_stopped
  102. netif_subqueue_stopped
  103. netif_wake_subqueue
  104. netif_attr_test_mask
  105. netif_attr_test_online
  106. netif_attrmask_next
  107. netif_attrmask_next_and
  108. netif_set_xps_queue
  109. __netif_set_xps_queue
  110. netif_is_multiqueue
  111. netif_set_real_num_rx_queues
  112. __netif_get_rx_queue
  113. get_netdev_rx_queue_index
  114. dev_kfree_skb_irq
  115. dev_consume_skb_irq
  116. dev_kfree_skb_any
  117. dev_consume_skb_any
  118. napi_free_frags
  119. ____dev_forward_skb
  120. dev_put
  121. dev_hold
  122. netif_carrier_ok
  123. netif_dormant_on
  124. netif_dormant_off
  125. netif_dormant
  126. netif_oper_up
  127. netif_device_present
  128. netif_msg_init
  129. __netif_tx_lock
  130. __netif_tx_acquire
  131. __netif_tx_release
  132. __netif_tx_lock_bh
  133. __netif_tx_trylock
  134. __netif_tx_unlock
  135. __netif_tx_unlock_bh
  136. txq_trans_update
  137. netif_trans_update
  138. netif_tx_lock
  139. netif_tx_lock_bh
  140. netif_tx_unlock
  141. netif_tx_unlock_bh
  142. netif_tx_disable
  143. netif_addr_lock
  144. netif_addr_lock_bh
  145. netif_addr_unlock
  146. netif_addr_unlock_bh
  147. __dev_uc_sync
  148. __dev_uc_unsync
  149. __dev_mc_sync
  150. __dev_mc_unsync
  151. skb_gso_segment
  152. can_checksum_protocol
  153. netdev_rx_csum_fault
  154. __netdev_start_xmit
  155. netdev_xmit_more
  156. netdev_start_xmit
  157. netdev_class_create_file
  158. netdev_class_remove_file
  159. netdev_intersect_features
  160. netdev_get_wanted_features
  161. netdev_add_tso_features
  162. net_gso_ok
  163. skb_gso_ok
  164. netif_needs_gso
  165. netif_set_gso_max_size
  166. skb_gso_error_unwind
  167. netif_is_macsec
  168. netif_is_macvlan
  169. netif_is_macvlan_port
  170. netif_is_bond_master
  171. netif_is_bond_slave
  172. netif_supports_nofcs
  173. netif_has_l3_rx_handler
  174. netif_is_l3_master
  175. netif_is_l3_slave
  176. netif_is_bridge_master
  177. netif_is_bridge_port
  178. netif_is_ovs_master
  179. netif_is_ovs_port
  180. netif_is_team_master
  181. netif_is_team_port
  182. netif_is_lag_master
  183. netif_is_lag_port
  184. netif_is_rxfh_configured
  185. netif_is_failover
  186. netif_is_failover_slave
  187. netif_keep_dst
  188. netif_reduces_vlan_mtu
  189. netdev_name
  190. netdev_unregistering
  191. netdev_reg_state

   1 /* SPDX-License-Identifier: GPL-2.0-or-later */
   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 for the Interfaces handler.
   8  *
   9  * Version:     @(#)dev.h       1.0.10  08/12/93
  10  *
  11  * Authors:     Ross Biro
  12  *              Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
  13  *              Corey Minyard <wf-rch!minyard@relay.EU.net>
  14  *              Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
  15  *              Alan Cox, <alan@lxorguk.ukuu.org.uk>
  16  *              Bjorn Ekwall. <bj0rn@blox.se>
  17  *              Pekka Riikonen <priikone@poseidon.pspt.fi>
  18  *
  19  *              Moved to /usr/include/linux for NET3
  20  */
  21 #ifndef _LINUX_NETDEVICE_H
  22 #define _LINUX_NETDEVICE_H
  23 
  24 #include <linux/timer.h>
  25 #include <linux/bug.h>
  26 #include <linux/delay.h>
  27 #include <linux/atomic.h>
  28 #include <linux/prefetch.h>
  29 #include <asm/cache.h>
  30 #include <asm/byteorder.h>
  31 
  32 #include <linux/percpu.h>
  33 #include <linux/rculist.h>
  34 #include <linux/workqueue.h>
  35 #include <linux/dynamic_queue_limits.h>
  36 
  37 #include <linux/ethtool.h>
  38 #include <net/net_namespace.h>
  39 #ifdef CONFIG_DCB
  40 #include <net/dcbnl.h>
  41 #endif
  42 #include <net/netprio_cgroup.h>
  43 #include <net/xdp.h>
  44 
  45 #include <linux/netdev_features.h>
  46 #include <linux/neighbour.h>
  47 #include <uapi/linux/netdevice.h>
  48 #include <uapi/linux/if_bonding.h>
  49 #include <uapi/linux/pkt_cls.h>
  50 #include <linux/hashtable.h>
  51 
  52 struct netpoll_info;
  53 struct device;
  54 struct phy_device;
  55 struct dsa_port;
  56 
  57 struct sfp_bus;
  58 /* 802.11 specific */
  59 struct wireless_dev;
  60 /* 802.15.4 specific */
  61 struct wpan_dev;
  62 struct mpls_dev;
  63 /* UDP Tunnel offloads */
  64 struct udp_tunnel_info;
  65 struct bpf_prog;
  66 struct xdp_buff;
  67 
  68 void netdev_set_default_ethtool_ops(struct net_device *dev,
  69                                     const struct ethtool_ops *ops);
  70 
  71 /* Backlog congestion levels */
  72 #define NET_RX_SUCCESS          0       /* keep 'em coming, baby */
  73 #define NET_RX_DROP             1       /* packet dropped */
  74 
  75 #define MAX_NEST_DEV 8
  76 
  77 /*
  78  * Transmit return codes: transmit return codes originate from three different
  79  * namespaces:
  80  *
  81  * - qdisc return codes
  82  * - driver transmit return codes
  83  * - errno values
  84  *
  85  * Drivers are allowed to return any one of those in their hard_start_xmit()
  86  * function. Real network devices commonly used with qdiscs should only return
  87  * the driver transmit return codes though - when qdiscs are used, the actual
  88  * transmission happens asynchronously, so the value is not propagated to
  89  * higher layers. Virtual network devices transmit synchronously; in this case
  90  * the driver transmit return codes are consumed by dev_queue_xmit(), and all
  91  * others are propagated to higher layers.
  92  */
  93 
  94 /* qdisc ->enqueue() return codes. */
  95 #define NET_XMIT_SUCCESS        0x00
  96 #define NET_XMIT_DROP           0x01    /* skb dropped                  */
  97 #define NET_XMIT_CN             0x02    /* congestion notification      */
  98 #define NET_XMIT_MASK           0x0f    /* qdisc flags in net/sch_generic.h */
  99 
 100 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
 101  * indicates that the device will soon be dropping packets, or already drops
 102  * some packets of the same priority; prompting us to send less aggressively. */
 103 #define net_xmit_eval(e)        ((e) == NET_XMIT_CN ? 0 : (e))
 104 #define net_xmit_errno(e)       ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
 105 
 106 /* Driver transmit return codes */
 107 #define NETDEV_TX_MASK          0xf0
 108 
 109 enum netdev_tx {
 110         __NETDEV_TX_MIN  = INT_MIN,     /* make sure enum is signed */
 111         NETDEV_TX_OK     = 0x00,        /* driver took care of packet */
 112         NETDEV_TX_BUSY   = 0x10,        /* driver tx path was busy*/
 113 };
 114 typedef enum netdev_tx netdev_tx_t;
 115 
 116 /*
 117  * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
 118  * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
 119  */
 120 static inline bool dev_xmit_complete(int rc)
 121 {
 122         /*
 123          * Positive cases with an skb consumed by a driver:
 124          * - successful transmission (rc == NETDEV_TX_OK)
 125          * - error while transmitting (rc < 0)
 126          * - error while queueing to a different device (rc & NET_XMIT_MASK)
 127          */
 128         if (likely(rc < NET_XMIT_MASK))
 129                 return true;
 130 
 131         return false;
 132 }
 133 
 134 /*
 135  *      Compute the worst-case header length according to the protocols
 136  *      used.
 137  */
 138 
 139 #if defined(CONFIG_HYPERV_NET)
 140 # define LL_MAX_HEADER 128
 141 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
 142 # if defined(CONFIG_MAC80211_MESH)
 143 #  define LL_MAX_HEADER 128
 144 # else
 145 #  define LL_MAX_HEADER 96
 146 # endif
 147 #else
 148 # define LL_MAX_HEADER 32
 149 #endif
 150 
 151 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
 152     !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
 153 #define MAX_HEADER LL_MAX_HEADER
 154 #else
 155 #define MAX_HEADER (LL_MAX_HEADER + 48)
 156 #endif
 157 
 158 /*
 159  *      Old network device statistics. Fields are native words
 160  *      (unsigned long) so they can be read and written atomically.
 161  */
 162 
 163 struct net_device_stats {
 164         unsigned long   rx_packets;
 165         unsigned long   tx_packets;
 166         unsigned long   rx_bytes;
 167         unsigned long   tx_bytes;
 168         unsigned long   rx_errors;
 169         unsigned long   tx_errors;
 170         unsigned long   rx_dropped;
 171         unsigned long   tx_dropped;
 172         unsigned long   multicast;
 173         unsigned long   collisions;
 174         unsigned long   rx_length_errors;
 175         unsigned long   rx_over_errors;
 176         unsigned long   rx_crc_errors;
 177         unsigned long   rx_frame_errors;
 178         unsigned long   rx_fifo_errors;
 179         unsigned long   rx_missed_errors;
 180         unsigned long   tx_aborted_errors;
 181         unsigned long   tx_carrier_errors;
 182         unsigned long   tx_fifo_errors;
 183         unsigned long   tx_heartbeat_errors;
 184         unsigned long   tx_window_errors;
 185         unsigned long   rx_compressed;
 186         unsigned long   tx_compressed;
 187 };
 188 
 189 
 190 #include <linux/cache.h>
 191 #include <linux/skbuff.h>
 192 
 193 #ifdef CONFIG_RPS
 194 #include <linux/static_key.h>
 195 extern struct static_key_false rps_needed;
 196 extern struct static_key_false rfs_needed;
 197 #endif
 198 
 199 struct neighbour;
 200 struct neigh_parms;
 201 struct sk_buff;
 202 
 203 struct netdev_hw_addr {
 204         struct list_head        list;
 205         unsigned char           addr[MAX_ADDR_LEN];
 206         unsigned char           type;
 207 #define NETDEV_HW_ADDR_T_LAN            1
 208 #define NETDEV_HW_ADDR_T_SAN            2
 209 #define NETDEV_HW_ADDR_T_SLAVE          3
 210 #define NETDEV_HW_ADDR_T_UNICAST        4
 211 #define NETDEV_HW_ADDR_T_MULTICAST      5
 212         bool                    global_use;
 213         int                     sync_cnt;
 214         int                     refcount;
 215         int                     synced;
 216         struct rcu_head         rcu_head;
 217 };
 218 
 219 struct netdev_hw_addr_list {
 220         struct list_head        list;
 221         int                     count;
 222 };
 223 
 224 #define netdev_hw_addr_list_count(l) ((l)->count)
 225 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
 226 #define netdev_hw_addr_list_for_each(ha, l) \
 227         list_for_each_entry(ha, &(l)->list, list)
 228 
 229 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
 230 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
 231 #define netdev_for_each_uc_addr(ha, dev) \
 232         netdev_hw_addr_list_for_each(ha, &(dev)->uc)
 233 
 234 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
 235 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
 236 #define netdev_for_each_mc_addr(ha, dev) \
 237         netdev_hw_addr_list_for_each(ha, &(dev)->mc)
 238 
 239 struct hh_cache {
 240         unsigned int    hh_len;
 241         seqlock_t       hh_lock;
 242 
 243         /* cached hardware header; allow for machine alignment needs.        */
 244 #define HH_DATA_MOD     16
 245 #define HH_DATA_OFF(__len) \
 246         (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
 247 #define HH_DATA_ALIGN(__len) \
 248         (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
 249         unsigned long   hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
 250 };
 251 
 252 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
 253  * Alternative is:
 254  *   dev->hard_header_len ? (dev->hard_header_len +
 255  *                           (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
 256  *
 257  * We could use other alignment values, but we must maintain the
 258  * relationship HH alignment <= LL alignment.
 259  */
 260 #define LL_RESERVED_SPACE(dev) \
 261         ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
 262 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
 263         ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
 264 
 265 struct header_ops {
 266         int     (*create) (struct sk_buff *skb, struct net_device *dev,
 267                            unsigned short type, const void *daddr,
 268                            const void *saddr, unsigned int len);
 269         int     (*parse)(const struct sk_buff *skb, unsigned char *haddr);
 270         int     (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
 271         void    (*cache_update)(struct hh_cache *hh,
 272                                 const struct net_device *dev,
 273                                 const unsigned char *haddr);
 274         bool    (*validate)(const char *ll_header, unsigned int len);
 275         __be16  (*parse_protocol)(const struct sk_buff *skb);
 276 };
 277 
 278 /* These flag bits are private to the generic network queueing
 279  * layer; they may not be explicitly referenced by any other
 280  * code.
 281  */
 282 
 283 enum netdev_state_t {
 284         __LINK_STATE_START,
 285         __LINK_STATE_PRESENT,
 286         __LINK_STATE_NOCARRIER,
 287         __LINK_STATE_LINKWATCH_PENDING,
 288         __LINK_STATE_DORMANT,
 289 };
 290 
 291 
 292 /*
 293  * This structure holds boot-time configured netdevice settings. They
 294  * are then used in the device probing.
 295  */
 296 struct netdev_boot_setup {
 297         char name[IFNAMSIZ];
 298         struct ifmap map;
 299 };
 300 #define NETDEV_BOOT_SETUP_MAX 8
 301 
 302 int __init netdev_boot_setup(char *str);
 303 
 304 struct gro_list {
 305         struct list_head        list;
 306         int                     count;
 307 };
 308 
 309 /*
 310  * size of gro hash buckets, must less than bit number of
 311  * napi_struct::gro_bitmask
 312  */
 313 #define GRO_HASH_BUCKETS        8
 314 
 315 /*
 316  * Structure for NAPI scheduling similar to tasklet but with weighting
 317  */
 318 struct napi_struct {
 319         /* The poll_list must only be managed by the entity which
 320          * changes the state of the NAPI_STATE_SCHED bit.  This means
 321          * whoever atomically sets that bit can add this napi_struct
 322          * to the per-CPU poll_list, and whoever clears that bit
 323          * can remove from the list right before clearing the bit.
 324          */
 325         struct list_head        poll_list;
 326 
 327         unsigned long           state;
 328         int                     weight;
 329         unsigned long           gro_bitmask;
 330         int                     (*poll)(struct napi_struct *, int);
 331 #ifdef CONFIG_NETPOLL
 332         int                     poll_owner;
 333 #endif
 334         struct net_device       *dev;
 335         struct gro_list         gro_hash[GRO_HASH_BUCKETS];
 336         struct sk_buff          *skb;
 337         struct list_head        rx_list; /* Pending GRO_NORMAL skbs */
 338         int                     rx_count; /* length of rx_list */
 339         struct hrtimer          timer;
 340         struct list_head        dev_list;
 341         struct hlist_node       napi_hash_node;
 342         unsigned int            napi_id;
 343 };
 344 
 345 enum {
 346         NAPI_STATE_SCHED,       /* Poll is scheduled */
 347         NAPI_STATE_MISSED,      /* reschedule a napi */
 348         NAPI_STATE_DISABLE,     /* Disable pending */
 349         NAPI_STATE_NPSVC,       /* Netpoll - don't dequeue from poll_list */
 350         NAPI_STATE_HASHED,      /* In NAPI hash (busy polling possible) */
 351         NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
 352         NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
 353 };
 354 
 355 enum {
 356         NAPIF_STATE_SCHED        = BIT(NAPI_STATE_SCHED),
 357         NAPIF_STATE_MISSED       = BIT(NAPI_STATE_MISSED),
 358         NAPIF_STATE_DISABLE      = BIT(NAPI_STATE_DISABLE),
 359         NAPIF_STATE_NPSVC        = BIT(NAPI_STATE_NPSVC),
 360         NAPIF_STATE_HASHED       = BIT(NAPI_STATE_HASHED),
 361         NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
 362         NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
 363 };
 364 
 365 enum gro_result {
 366         GRO_MERGED,
 367         GRO_MERGED_FREE,
 368         GRO_HELD,
 369         GRO_NORMAL,
 370         GRO_DROP,
 371         GRO_CONSUMED,
 372 };
 373 typedef enum gro_result gro_result_t;
 374 
 375 /*
 376  * enum rx_handler_result - Possible return values for rx_handlers.
 377  * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
 378  * further.
 379  * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
 380  * case skb->dev was changed by rx_handler.
 381  * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
 382  * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
 383  *
 384  * rx_handlers are functions called from inside __netif_receive_skb(), to do
 385  * special processing of the skb, prior to delivery to protocol handlers.
 386  *
 387  * Currently, a net_device can only have a single rx_handler registered. Trying
 388  * to register a second rx_handler will return -EBUSY.
 389  *
 390  * To register a rx_handler on a net_device, use netdev_rx_handler_register().
 391  * To unregister a rx_handler on a net_device, use
 392  * netdev_rx_handler_unregister().
 393  *
 394  * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
 395  * do with the skb.
 396  *
 397  * If the rx_handler consumed the skb in some way, it should return
 398  * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
 399  * the skb to be delivered in some other way.
 400  *
 401  * If the rx_handler changed skb->dev, to divert the skb to another
 402  * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
 403  * new device will be called if it exists.
 404  *
 405  * If the rx_handler decides the skb should be ignored, it should return
 406  * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
 407  * are registered on exact device (ptype->dev == skb->dev).
 408  *
 409  * If the rx_handler didn't change skb->dev, but wants the skb to be normally
 410  * delivered, it should return RX_HANDLER_PASS.
 411  *
 412  * A device without a registered rx_handler will behave as if rx_handler
 413  * returned RX_HANDLER_PASS.
 414  */
 415 
 416 enum rx_handler_result {
 417         RX_HANDLER_CONSUMED,
 418         RX_HANDLER_ANOTHER,
 419         RX_HANDLER_EXACT,
 420         RX_HANDLER_PASS,
 421 };
 422 typedef enum rx_handler_result rx_handler_result_t;
 423 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
 424 
 425 void __napi_schedule(struct napi_struct *n);
 426 void __napi_schedule_irqoff(struct napi_struct *n);
 427 
 428 static inline bool napi_disable_pending(struct napi_struct *n)
 429 {
 430         return test_bit(NAPI_STATE_DISABLE, &n->state);
 431 }
 432 
 433 bool napi_schedule_prep(struct napi_struct *n);
 434 
 435 /**
 436  *      napi_schedule - schedule NAPI poll
 437  *      @n: NAPI context
 438  *
 439  * Schedule NAPI poll routine to be called if it is not already
 440  * running.
 441  */
 442 static inline void napi_schedule(struct napi_struct *n)
 443 {
 444         if (napi_schedule_prep(n))
 445                 __napi_schedule(n);
 446 }
 447 
 448 /**
 449  *      napi_schedule_irqoff - schedule NAPI poll
 450  *      @n: NAPI context
 451  *
 452  * Variant of napi_schedule(), assuming hard irqs are masked.
 453  */
 454 static inline void napi_schedule_irqoff(struct napi_struct *n)
 455 {
 456         if (napi_schedule_prep(n))
 457                 __napi_schedule_irqoff(n);
 458 }
 459 
 460 /* Try to reschedule poll. Called by dev->poll() after napi_complete().  */
 461 static inline bool napi_reschedule(struct napi_struct *napi)
 462 {
 463         if (napi_schedule_prep(napi)) {
 464                 __napi_schedule(napi);
 465                 return true;
 466         }
 467         return false;
 468 }
 469 
 470 bool napi_complete_done(struct napi_struct *n, int work_done);
 471 /**
 472  *      napi_complete - NAPI processing complete
 473  *      @n: NAPI context
 474  *
 475  * Mark NAPI processing as complete.
 476  * Consider using napi_complete_done() instead.
 477  * Return false if device should avoid rearming interrupts.
 478  */
 479 static inline bool napi_complete(struct napi_struct *n)
 480 {
 481         return napi_complete_done(n, 0);
 482 }
 483 
 484 /**
 485  *      napi_hash_del - remove a NAPI from global table
 486  *      @napi: NAPI context
 487  *
 488  * Warning: caller must observe RCU grace period
 489  * before freeing memory containing @napi, if
 490  * this function returns true.
 491  * Note: core networking stack automatically calls it
 492  * from netif_napi_del().
 493  * Drivers might want to call this helper to combine all
 494  * the needed RCU grace periods into a single one.
 495  */
 496 bool napi_hash_del(struct napi_struct *napi);
 497 
 498 /**
 499  *      napi_disable - prevent NAPI from scheduling
 500  *      @n: NAPI context
 501  *
 502  * Stop NAPI from being scheduled on this context.
 503  * Waits till any outstanding processing completes.
 504  */
 505 void napi_disable(struct napi_struct *n);
 506 
 507 /**
 508  *      napi_enable - enable NAPI scheduling
 509  *      @n: NAPI context
 510  *
 511  * Resume NAPI from being scheduled on this context.
 512  * Must be paired with napi_disable.
 513  */
 514 static inline void napi_enable(struct napi_struct *n)
 515 {
 516         BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
 517         smp_mb__before_atomic();
 518         clear_bit(NAPI_STATE_SCHED, &n->state);
 519         clear_bit(NAPI_STATE_NPSVC, &n->state);
 520 }
 521 
 522 /**
 523  *      napi_synchronize - wait until NAPI is not running
 524  *      @n: NAPI context
 525  *
 526  * Wait until NAPI is done being scheduled on this context.
 527  * Waits till any outstanding processing completes but
 528  * does not disable future activations.
 529  */
 530 static inline void napi_synchronize(const struct napi_struct *n)
 531 {
 532         if (IS_ENABLED(CONFIG_SMP))
 533                 while (test_bit(NAPI_STATE_SCHED, &n->state))
 534                         msleep(1);
 535         else
 536                 barrier();
 537 }
 538 
 539 /**
 540  *      napi_if_scheduled_mark_missed - if napi is running, set the
 541  *      NAPIF_STATE_MISSED
 542  *      @n: NAPI context
 543  *
 544  * If napi is running, set the NAPIF_STATE_MISSED, and return true if
 545  * NAPI is scheduled.
 546  **/
 547 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
 548 {
 549         unsigned long val, new;
 550 
 551         do {
 552                 val = READ_ONCE(n->state);
 553                 if (val & NAPIF_STATE_DISABLE)
 554                         return true;
 555 
 556                 if (!(val & NAPIF_STATE_SCHED))
 557                         return false;
 558 
 559                 new = val | NAPIF_STATE_MISSED;
 560         } while (cmpxchg(&n->state, val, new) != val);
 561 
 562         return true;
 563 }
 564 
 565 enum netdev_queue_state_t {
 566         __QUEUE_STATE_DRV_XOFF,
 567         __QUEUE_STATE_STACK_XOFF,
 568         __QUEUE_STATE_FROZEN,
 569 };
 570 
 571 #define QUEUE_STATE_DRV_XOFF    (1 << __QUEUE_STATE_DRV_XOFF)
 572 #define QUEUE_STATE_STACK_XOFF  (1 << __QUEUE_STATE_STACK_XOFF)
 573 #define QUEUE_STATE_FROZEN      (1 << __QUEUE_STATE_FROZEN)
 574 
 575 #define QUEUE_STATE_ANY_XOFF    (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
 576 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
 577                                         QUEUE_STATE_FROZEN)
 578 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
 579                                         QUEUE_STATE_FROZEN)
 580 
 581 /*
 582  * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue.  The
 583  * netif_tx_* functions below are used to manipulate this flag.  The
 584  * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
 585  * queue independently.  The netif_xmit_*stopped functions below are called
 586  * to check if the queue has been stopped by the driver or stack (either
 587  * of the XOFF bits are set in the state).  Drivers should not need to call
 588  * netif_xmit*stopped functions, they should only be using netif_tx_*.
 589  */
 590 
 591 struct netdev_queue {
 592 /*
 593  * read-mostly part
 594  */
 595         struct net_device       *dev;
 596         struct Qdisc __rcu      *qdisc;
 597         struct Qdisc            *qdisc_sleeping;
 598 #ifdef CONFIG_SYSFS
 599         struct kobject          kobj;
 600 #endif
 601 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
 602         int                     numa_node;
 603 #endif
 604         unsigned long           tx_maxrate;
 605         /*
 606          * Number of TX timeouts for this queue
 607          * (/sys/class/net/DEV/Q/trans_timeout)
 608          */
 609         unsigned long           trans_timeout;
 610 
 611         /* Subordinate device that the queue has been assigned to */
 612         struct net_device       *sb_dev;
 613 #ifdef CONFIG_XDP_SOCKETS
 614         struct xdp_umem         *umem;
 615 #endif
 616 /*
 617  * write-mostly part
 618  */
 619         spinlock_t              _xmit_lock ____cacheline_aligned_in_smp;
 620         int                     xmit_lock_owner;
 621         /*
 622          * Time (in jiffies) of last Tx
 623          */
 624         unsigned long           trans_start;
 625 
 626         unsigned long           state;
 627 
 628 #ifdef CONFIG_BQL
 629         struct dql              dql;
 630 #endif
 631 } ____cacheline_aligned_in_smp;
 632 
 633 extern int sysctl_fb_tunnels_only_for_init_net;
 634 extern int sysctl_devconf_inherit_init_net;
 635 
 636 static inline bool net_has_fallback_tunnels(const struct net *net)
 637 {
 638         return net == &init_net ||
 639                !IS_ENABLED(CONFIG_SYSCTL) ||
 640                !sysctl_fb_tunnels_only_for_init_net;
 641 }
 642 
 643 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
 644 {
 645 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
 646         return q->numa_node;
 647 #else
 648         return NUMA_NO_NODE;
 649 #endif
 650 }
 651 
 652 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
 653 {
 654 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
 655         q->numa_node = node;
 656 #endif
 657 }
 658 
 659 #ifdef CONFIG_RPS
 660 /*
 661  * This structure holds an RPS map which can be of variable length.  The
 662  * map is an array of CPUs.
 663  */
 664 struct rps_map {
 665         unsigned int len;
 666         struct rcu_head rcu;
 667         u16 cpus[0];
 668 };
 669 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
 670 
 671 /*
 672  * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
 673  * tail pointer for that CPU's input queue at the time of last enqueue, and
 674  * a hardware filter index.
 675  */
 676 struct rps_dev_flow {
 677         u16 cpu;
 678         u16 filter;
 679         unsigned int last_qtail;
 680 };
 681 #define RPS_NO_FILTER 0xffff
 682 
 683 /*
 684  * The rps_dev_flow_table structure contains a table of flow mappings.
 685  */
 686 struct rps_dev_flow_table {
 687         unsigned int mask;
 688         struct rcu_head rcu;
 689         struct rps_dev_flow flows[0];
 690 };
 691 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
 692     ((_num) * sizeof(struct rps_dev_flow)))
 693 
 694 /*
 695  * The rps_sock_flow_table contains mappings of flows to the last CPU
 696  * on which they were processed by the application (set in recvmsg).
 697  * Each entry is a 32bit value. Upper part is the high-order bits
 698  * of flow hash, lower part is CPU number.
 699  * rps_cpu_mask is used to partition the space, depending on number of
 700  * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
 701  * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
 702  * meaning we use 32-6=26 bits for the hash.
 703  */
 704 struct rps_sock_flow_table {
 705         u32     mask;
 706 
 707         u32     ents[0] ____cacheline_aligned_in_smp;
 708 };
 709 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
 710 
 711 #define RPS_NO_CPU 0xffff
 712 
 713 extern u32 rps_cpu_mask;
 714 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
 715 
 716 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
 717                                         u32 hash)
 718 {
 719         if (table && hash) {
 720                 unsigned int index = hash & table->mask;
 721                 u32 val = hash & ~rps_cpu_mask;
 722 
 723                 /* We only give a hint, preemption can change CPU under us */
 724                 val |= raw_smp_processor_id();
 725 
 726                 if (table->ents[index] != val)
 727                         table->ents[index] = val;
 728         }
 729 }
 730 
 731 #ifdef CONFIG_RFS_ACCEL
 732 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
 733                          u16 filter_id);
 734 #endif
 735 #endif /* CONFIG_RPS */
 736 
 737 /* This structure contains an instance of an RX queue. */
 738 struct netdev_rx_queue {
 739 #ifdef CONFIG_RPS
 740         struct rps_map __rcu            *rps_map;
 741         struct rps_dev_flow_table __rcu *rps_flow_table;
 742 #endif
 743         struct kobject                  kobj;
 744         struct net_device               *dev;
 745         struct xdp_rxq_info             xdp_rxq;
 746 #ifdef CONFIG_XDP_SOCKETS
 747         struct xdp_umem                 *umem;
 748 #endif
 749 } ____cacheline_aligned_in_smp;
 750 
 751 /*
 752  * RX queue sysfs structures and functions.
 753  */
 754 struct rx_queue_attribute {
 755         struct attribute attr;
 756         ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
 757         ssize_t (*store)(struct netdev_rx_queue *queue,
 758                          const char *buf, size_t len);
 759 };
 760 
 761 #ifdef CONFIG_XPS
 762 /*
 763  * This structure holds an XPS map which can be of variable length.  The
 764  * map is an array of queues.
 765  */
 766 struct xps_map {
 767         unsigned int len;
 768         unsigned int alloc_len;
 769         struct rcu_head rcu;
 770         u16 queues[0];
 771 };
 772 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
 773 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
 774        - sizeof(struct xps_map)) / sizeof(u16))
 775 
 776 /*
 777  * This structure holds all XPS maps for device.  Maps are indexed by CPU.
 778  */
 779 struct xps_dev_maps {
 780         struct rcu_head rcu;
 781         struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */
 782 };
 783 
 784 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) +      \
 785         (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
 786 
 787 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
 788         (_rxqs * (_tcs) * sizeof(struct xps_map *)))
 789 
 790 #endif /* CONFIG_XPS */
 791 
 792 #define TC_MAX_QUEUE    16
 793 #define TC_BITMASK      15
 794 /* HW offloaded queuing disciplines txq count and offset maps */
 795 struct netdev_tc_txq {
 796         u16 count;
 797         u16 offset;
 798 };
 799 
 800 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
 801 /*
 802  * This structure is to hold information about the device
 803  * configured to run FCoE protocol stack.
 804  */
 805 struct netdev_fcoe_hbainfo {
 806         char    manufacturer[64];
 807         char    serial_number[64];
 808         char    hardware_version[64];
 809         char    driver_version[64];
 810         char    optionrom_version[64];
 811         char    firmware_version[64];
 812         char    model[256];
 813         char    model_description[256];
 814 };
 815 #endif
 816 
 817 #define MAX_PHYS_ITEM_ID_LEN 32
 818 
 819 /* This structure holds a unique identifier to identify some
 820  * physical item (port for example) used by a netdevice.
 821  */
 822 struct netdev_phys_item_id {
 823         unsigned char id[MAX_PHYS_ITEM_ID_LEN];
 824         unsigned char id_len;
 825 };
 826 
 827 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
 828                                             struct netdev_phys_item_id *b)
 829 {
 830         return a->id_len == b->id_len &&
 831                memcmp(a->id, b->id, a->id_len) == 0;
 832 }
 833 
 834 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
 835                                        struct sk_buff *skb,
 836                                        struct net_device *sb_dev);
 837 
 838 enum tc_setup_type {
 839         TC_SETUP_QDISC_MQPRIO,
 840         TC_SETUP_CLSU32,
 841         TC_SETUP_CLSFLOWER,
 842         TC_SETUP_CLSMATCHALL,
 843         TC_SETUP_CLSBPF,
 844         TC_SETUP_BLOCK,
 845         TC_SETUP_QDISC_CBS,
 846         TC_SETUP_QDISC_RED,
 847         TC_SETUP_QDISC_PRIO,
 848         TC_SETUP_QDISC_MQ,
 849         TC_SETUP_QDISC_ETF,
 850         TC_SETUP_ROOT_QDISC,
 851         TC_SETUP_QDISC_GRED,
 852         TC_SETUP_QDISC_TAPRIO,
 853 };
 854 
 855 /* These structures hold the attributes of bpf state that are being passed
 856  * to the netdevice through the bpf op.
 857  */
 858 enum bpf_netdev_command {
 859         /* Set or clear a bpf program used in the earliest stages of packet
 860          * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
 861          * is responsible for calling bpf_prog_put on any old progs that are
 862          * stored. In case of error, the callee need not release the new prog
 863          * reference, but on success it takes ownership and must bpf_prog_put
 864          * when it is no longer used.
 865          */
 866         XDP_SETUP_PROG,
 867         XDP_SETUP_PROG_HW,
 868         XDP_QUERY_PROG,
 869         XDP_QUERY_PROG_HW,
 870         /* BPF program for offload callbacks, invoked at program load time. */
 871         BPF_OFFLOAD_MAP_ALLOC,
 872         BPF_OFFLOAD_MAP_FREE,
 873         XDP_SETUP_XSK_UMEM,
 874 };
 875 
 876 struct bpf_prog_offload_ops;
 877 struct netlink_ext_ack;
 878 struct xdp_umem;
 879 
 880 struct netdev_bpf {
 881         enum bpf_netdev_command command;
 882         union {
 883                 /* XDP_SETUP_PROG */
 884                 struct {
 885                         u32 flags;
 886                         struct bpf_prog *prog;
 887                         struct netlink_ext_ack *extack;
 888                 };
 889                 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */
 890                 struct {
 891                         u32 prog_id;
 892                         /* flags with which program was installed */
 893                         u32 prog_flags;
 894                 };
 895                 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
 896                 struct {
 897                         struct bpf_offloaded_map *offmap;
 898                 };
 899                 /* XDP_SETUP_XSK_UMEM */
 900                 struct {
 901                         struct xdp_umem *umem;
 902                         u16 queue_id;
 903                 } xsk;
 904         };
 905 };
 906 
 907 /* Flags for ndo_xsk_wakeup. */
 908 #define XDP_WAKEUP_RX (1 << 0)
 909 #define XDP_WAKEUP_TX (1 << 1)
 910 
 911 #ifdef CONFIG_XFRM_OFFLOAD
 912 struct xfrmdev_ops {
 913         int     (*xdo_dev_state_add) (struct xfrm_state *x);
 914         void    (*xdo_dev_state_delete) (struct xfrm_state *x);
 915         void    (*xdo_dev_state_free) (struct xfrm_state *x);
 916         bool    (*xdo_dev_offload_ok) (struct sk_buff *skb,
 917                                        struct xfrm_state *x);
 918         void    (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
 919 };
 920 #endif
 921 
 922 struct dev_ifalias {
 923         struct rcu_head rcuhead;
 924         char ifalias[];
 925 };
 926 
 927 struct devlink;
 928 struct tlsdev_ops;
 929 
 930 
 931 /*
 932  * This structure defines the management hooks for network devices.
 933  * The following hooks can be defined; unless noted otherwise, they are
 934  * optional and can be filled with a null pointer.
 935  *
 936  * int (*ndo_init)(struct net_device *dev);
 937  *     This function is called once when a network device is registered.
 938  *     The network device can use this for any late stage initialization
 939  *     or semantic validation. It can fail with an error code which will
 940  *     be propagated back to register_netdev.
 941  *
 942  * void (*ndo_uninit)(struct net_device *dev);
 943  *     This function is called when device is unregistered or when registration
 944  *     fails. It is not called if init fails.
 945  *
 946  * int (*ndo_open)(struct net_device *dev);
 947  *     This function is called when a network device transitions to the up
 948  *     state.
 949  *
 950  * int (*ndo_stop)(struct net_device *dev);
 951  *     This function is called when a network device transitions to the down
 952  *     state.
 953  *
 954  * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
 955  *                               struct net_device *dev);
 956  *      Called when a packet needs to be transmitted.
 957  *      Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
 958  *      the queue before that can happen; it's for obsolete devices and weird
 959  *      corner cases, but the stack really does a non-trivial amount
 960  *      of useless work if you return NETDEV_TX_BUSY.
 961  *      Required; cannot be NULL.
 962  *
 963  * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
 964  *                                         struct net_device *dev
 965  *                                         netdev_features_t features);
 966  *      Called by core transmit path to determine if device is capable of
 967  *      performing offload operations on a given packet. This is to give
 968  *      the device an opportunity to implement any restrictions that cannot
 969  *      be otherwise expressed by feature flags. The check is called with
 970  *      the set of features that the stack has calculated and it returns
 971  *      those the driver believes to be appropriate.
 972  *
 973  * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
 974  *                         struct net_device *sb_dev);
 975  *      Called to decide which queue to use when device supports multiple
 976  *      transmit queues.
 977  *
 978  * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
 979  *      This function is called to allow device receiver to make
 980  *      changes to configuration when multicast or promiscuous is enabled.
 981  *
 982  * void (*ndo_set_rx_mode)(struct net_device *dev);
 983  *      This function is called device changes address list filtering.
 984  *      If driver handles unicast address filtering, it should set
 985  *      IFF_UNICAST_FLT in its priv_flags.
 986  *
 987  * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
 988  *      This function  is called when the Media Access Control address
 989  *      needs to be changed. If this interface is not defined, the
 990  *      MAC address can not be changed.
 991  *
 992  * int (*ndo_validate_addr)(struct net_device *dev);
 993  *      Test if Media Access Control address is valid for the device.
 994  *
 995  * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
 996  *      Called when a user requests an ioctl which can't be handled by
 997  *      the generic interface code. If not defined ioctls return
 998  *      not supported error code.
 999  *
1000  * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1001  *      Used to set network devices bus interface parameters. This interface
1002  *      is retained for legacy reasons; new devices should use the bus
1003  *      interface (PCI) for low level management.
1004  *
1005  * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1006  *      Called when a user wants to change the Maximum Transfer Unit
1007  *      of a device.
1008  *
1009  * void (*ndo_tx_timeout)(struct net_device *dev);
1010  *      Callback used when the transmitter has not made any progress
1011  *      for dev->watchdog ticks.
1012  *
1013  * void (*ndo_get_stats64)(struct net_device *dev,
1014  *                         struct rtnl_link_stats64 *storage);
1015  * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1016  *      Called when a user wants to get the network device usage
1017  *      statistics. Drivers must do one of the following:
1018  *      1. Define @ndo_get_stats64 to fill in a zero-initialised
1019  *         rtnl_link_stats64 structure passed by the caller.
1020  *      2. Define @ndo_get_stats to update a net_device_stats structure
1021  *         (which should normally be dev->stats) and return a pointer to
1022  *         it. The structure may be changed asynchronously only if each
1023  *         field is written atomically.
1024  *      3. Update dev->stats asynchronously and atomically, and define
1025  *         neither operation.
1026  *
1027  * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1028  *      Return true if this device supports offload stats of this attr_id.
1029  *
1030  * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1031  *      void *attr_data)
1032  *      Get statistics for offload operations by attr_id. Write it into the
1033  *      attr_data pointer.
1034  *
1035  * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1036  *      If device supports VLAN filtering this function is called when a
1037  *      VLAN id is registered.
1038  *
1039  * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1040  *      If device supports VLAN filtering this function is called when a
1041  *      VLAN id is unregistered.
1042  *
1043  * void (*ndo_poll_controller)(struct net_device *dev);
1044  *
1045  *      SR-IOV management functions.
1046  * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1047  * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1048  *                        u8 qos, __be16 proto);
1049  * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1050  *                        int max_tx_rate);
1051  * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1052  * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1053  * int (*ndo_get_vf_config)(struct net_device *dev,
1054  *                          int vf, struct ifla_vf_info *ivf);
1055  * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1056  * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1057  *                        struct nlattr *port[]);
1058  *
1059  *      Enable or disable the VF ability to query its RSS Redirection Table and
1060  *      Hash Key. This is needed since on some devices VF share this information
1061  *      with PF and querying it may introduce a theoretical security risk.
1062  * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1063  * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1064  * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1065  *                     void *type_data);
1066  *      Called to setup any 'tc' scheduler, classifier or action on @dev.
1067  *      This is always called from the stack with the rtnl lock held and netif
1068  *      tx queues stopped. This allows the netdevice to perform queue
1069  *      management safely.
1070  *
1071  *      Fiber Channel over Ethernet (FCoE) offload functions.
1072  * int (*ndo_fcoe_enable)(struct net_device *dev);
1073  *      Called when the FCoE protocol stack wants to start using LLD for FCoE
1074  *      so the underlying device can perform whatever needed configuration or
1075  *      initialization to support acceleration of FCoE traffic.
1076  *
1077  * int (*ndo_fcoe_disable)(struct net_device *dev);
1078  *      Called when the FCoE protocol stack wants to stop using LLD for FCoE
1079  *      so the underlying device can perform whatever needed clean-ups to
1080  *      stop supporting acceleration of FCoE traffic.
1081  *
1082  * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1083  *                           struct scatterlist *sgl, unsigned int sgc);
1084  *      Called when the FCoE Initiator wants to initialize an I/O that
1085  *      is a possible candidate for Direct Data Placement (DDP). The LLD can
1086  *      perform necessary setup and returns 1 to indicate the device is set up
1087  *      successfully to perform DDP on this I/O, otherwise this returns 0.
1088  *
1089  * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
1090  *      Called when the FCoE Initiator/Target is done with the DDPed I/O as
1091  *      indicated by the FC exchange id 'xid', so the underlying device can
1092  *      clean up and reuse resources for later DDP requests.
1093  *
1094  * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1095  *                            struct scatterlist *sgl, unsigned int sgc);
1096  *      Called when the FCoE Target wants to initialize an I/O that
1097  *      is a possible candidate for Direct Data Placement (DDP). The LLD can
1098  *      perform necessary setup and returns 1 to indicate the device is set up
1099  *      successfully to perform DDP on this I/O, otherwise this returns 0.
1100  *
1101  * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1102  *                             struct netdev_fcoe_hbainfo *hbainfo);
1103  *      Called when the FCoE Protocol stack wants information on the underlying
1104  *      device. This information is utilized by the FCoE protocol stack to
1105  *      register attributes with Fiber Channel management service as per the
1106  *      FC-GS Fabric Device Management Information(FDMI) specification.
1107  *
1108  * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1109  *      Called when the underlying device wants to override default World Wide
1110  *      Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1111  *      World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1112  *      protocol stack to use.
1113  *
1114  *      RFS acceleration.
1115  * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1116  *                          u16 rxq_index, u32 flow_id);
1117  *      Set hardware filter for RFS.  rxq_index is the target queue index;
1118  *      flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1119  *      Return the filter ID on success, or a negative error code.
1120  *
1121  *      Slave management functions (for bridge, bonding, etc).
1122  * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1123  *      Called to make another netdev an underling.
1124  *
1125  * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1126  *      Called to release previously enslaved netdev.
1127  *
1128  *      Feature/offload setting functions.
1129  * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1130  *              netdev_features_t features);
1131  *      Adjusts the requested feature flags according to device-specific
1132  *      constraints, and returns the resulting flags. Must not modify
1133  *      the device state.
1134  *
1135  * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1136  *      Called to update device configuration to new features. Passed
1137  *      feature set might be less than what was returned by ndo_fix_features()).
1138  *      Must return >0 or -errno if it changed dev->features itself.
1139  *
1140  * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1141  *                    struct net_device *dev,
1142  *                    const unsigned char *addr, u16 vid, u16 flags,
1143  *                    struct netlink_ext_ack *extack);
1144  *      Adds an FDB entry to dev for addr.
1145  * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1146  *                    struct net_device *dev,
1147  *                    const unsigned char *addr, u16 vid)
1148  *      Deletes the FDB entry from dev coresponding to addr.
1149  * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1150  *                     struct net_device *dev, struct net_device *filter_dev,
1151  *                     int *idx)
1152  *      Used to add FDB entries to dump requests. Implementers should add
1153  *      entries to skb and update idx with the number of entries.
1154  *
1155  * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1156  *                           u16 flags, struct netlink_ext_ack *extack)
1157  * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1158  *                           struct net_device *dev, u32 filter_mask,
1159  *                           int nlflags)
1160  * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1161  *                           u16 flags);
1162  *
1163  * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1164  *      Called to change device carrier. Soft-devices (like dummy, team, etc)
1165  *      which do not represent real hardware may define this to allow their
1166  *      userspace components to manage their virtual carrier state. Devices
1167  *      that determine carrier state from physical hardware properties (eg
1168  *      network cables) or protocol-dependent mechanisms (eg
1169  *      USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1170  *
1171  * int (*ndo_get_phys_port_id)(struct net_device *dev,
1172  *                             struct netdev_phys_item_id *ppid);
1173  *      Called to get ID of physical port of this device. If driver does
1174  *      not implement this, it is assumed that the hw is not able to have
1175  *      multiple net devices on single physical port.
1176  *
1177  * int (*ndo_get_port_parent_id)(struct net_device *dev,
1178  *                               struct netdev_phys_item_id *ppid)
1179  *      Called to get the parent ID of the physical port of this device.
1180  *
1181  * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1182  *                            struct udp_tunnel_info *ti);
1183  *      Called by UDP tunnel to notify a driver about the UDP port and socket
1184  *      address family that a UDP tunnel is listnening to. It is called only
1185  *      when a new port starts listening. The operation is protected by the
1186  *      RTNL.
1187  *
1188  * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1189  *                            struct udp_tunnel_info *ti);
1190  *      Called by UDP tunnel to notify the driver about a UDP port and socket
1191  *      address family that the UDP tunnel is not listening to anymore. The
1192  *      operation is protected by the RTNL.
1193  *
1194  * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1195  *                               struct net_device *dev)
1196  *      Called by upper layer devices to accelerate switching or other
1197  *      station functionality into hardware. 'pdev is the lowerdev
1198  *      to use for the offload and 'dev' is the net device that will
1199  *      back the offload. Returns a pointer to the private structure
1200  *      the upper layer will maintain.
1201  * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1202  *      Called by upper layer device to delete the station created
1203  *      by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1204  *      the station and priv is the structure returned by the add
1205  *      operation.
1206  * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1207  *                           int queue_index, u32 maxrate);
1208  *      Called when a user wants to set a max-rate limitation of specific
1209  *      TX queue.
1210  * int (*ndo_get_iflink)(const struct net_device *dev);
1211  *      Called to get the iflink value of this device.
1212  * void (*ndo_change_proto_down)(struct net_device *dev,
1213  *                               bool proto_down);
1214  *      This function is used to pass protocol port error state information
1215  *      to the switch driver. The switch driver can react to the proto_down
1216  *      by doing a phys down on the associated switch port.
1217  * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1218  *      This function is used to get egress tunnel information for given skb.
1219  *      This is useful for retrieving outer tunnel header parameters while
1220  *      sampling packet.
1221  * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1222  *      This function is used to specify the headroom that the skb must
1223  *      consider when allocation skb during packet reception. Setting
1224  *      appropriate rx headroom value allows avoiding skb head copy on
1225  *      forward. Setting a negative value resets the rx headroom to the
1226  *      default value.
1227  * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1228  *      This function is used to set or query state related to XDP on the
1229  *      netdevice and manage BPF offload. See definition of
1230  *      enum bpf_netdev_command for details.
1231  * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1232  *                      u32 flags);
1233  *      This function is used to submit @n XDP packets for transmit on a
1234  *      netdevice. Returns number of frames successfully transmitted, frames
1235  *      that got dropped are freed/returned via xdp_return_frame().
1236  *      Returns negative number, means general error invoking ndo, meaning
1237  *      no frames were xmit'ed and core-caller will free all frames.
1238  * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1239  *      This function is used to wake up the softirq, ksoftirqd or kthread
1240  *      responsible for sending and/or receiving packets on a specific
1241  *      queue id bound to an AF_XDP socket. The flags field specifies if
1242  *      only RX, only Tx, or both should be woken up using the flags
1243  *      XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1244  * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1245  *      Get devlink port instance associated with a given netdev.
1246  *      Called with a reference on the netdevice and devlink locks only,
1247  *      rtnl_lock is not held.
1248  */
1249 struct net_device_ops {
1250         int                     (*ndo_init)(struct net_device *dev);
1251         void                    (*ndo_uninit)(struct net_device *dev);
1252         int                     (*ndo_open)(struct net_device *dev);
1253         int                     (*ndo_stop)(struct net_device *dev);
1254         netdev_tx_t             (*ndo_start_xmit)(struct sk_buff *skb,
1255                                                   struct net_device *dev);
1256         netdev_features_t       (*ndo_features_check)(struct sk_buff *skb,
1257                                                       struct net_device *dev,
1258                                                       netdev_features_t features);
1259         u16                     (*ndo_select_queue)(struct net_device *dev,
1260                                                     struct sk_buff *skb,
1261                                                     struct net_device *sb_dev);
1262         void                    (*ndo_change_rx_flags)(struct net_device *dev,
1263                                                        int flags);
1264         void                    (*ndo_set_rx_mode)(struct net_device *dev);
1265         int                     (*ndo_set_mac_address)(struct net_device *dev,
1266                                                        void *addr);
1267         int                     (*ndo_validate_addr)(struct net_device *dev);
1268         int                     (*ndo_do_ioctl)(struct net_device *dev,
1269                                                 struct ifreq *ifr, int cmd);
1270         int                     (*ndo_set_config)(struct net_device *dev,
1271                                                   struct ifmap *map);
1272         int                     (*ndo_change_mtu)(struct net_device *dev,
1273                                                   int new_mtu);
1274         int                     (*ndo_neigh_setup)(struct net_device *dev,
1275                                                    struct neigh_parms *);
1276         void                    (*ndo_tx_timeout) (struct net_device *dev);
1277 
1278         void                    (*ndo_get_stats64)(struct net_device *dev,
1279                                                    struct rtnl_link_stats64 *storage);
1280         bool                    (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1281         int                     (*ndo_get_offload_stats)(int attr_id,
1282                                                          const struct net_device *dev,
1283                                                          void *attr_data);
1284         struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1285 
1286         int                     (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1287                                                        __be16 proto, u16 vid);
1288         int                     (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1289                                                         __be16 proto, u16 vid);
1290 #ifdef CONFIG_NET_POLL_CONTROLLER
1291         void                    (*ndo_poll_controller)(struct net_device *dev);
1292         int                     (*ndo_netpoll_setup)(struct net_device *dev,
1293                                                      struct netpoll_info *info);
1294         void                    (*ndo_netpoll_cleanup)(struct net_device *dev);
1295 #endif
1296         int                     (*ndo_set_vf_mac)(struct net_device *dev,
1297                                                   int queue, u8 *mac);
1298         int                     (*ndo_set_vf_vlan)(struct net_device *dev,
1299                                                    int queue, u16 vlan,
1300                                                    u8 qos, __be16 proto);
1301         int                     (*ndo_set_vf_rate)(struct net_device *dev,
1302                                                    int vf, int min_tx_rate,
1303                                                    int max_tx_rate);
1304         int                     (*ndo_set_vf_spoofchk)(struct net_device *dev,
1305                                                        int vf, bool setting);
1306         int                     (*ndo_set_vf_trust)(struct net_device *dev,
1307                                                     int vf, bool setting);
1308         int                     (*ndo_get_vf_config)(struct net_device *dev,
1309                                                      int vf,
1310                                                      struct ifla_vf_info *ivf);
1311         int                     (*ndo_set_vf_link_state)(struct net_device *dev,
1312                                                          int vf, int link_state);
1313         int                     (*ndo_get_vf_stats)(struct net_device *dev,
1314                                                     int vf,
1315                                                     struct ifla_vf_stats
1316                                                     *vf_stats);
1317         int                     (*ndo_set_vf_port)(struct net_device *dev,
1318                                                    int vf,
1319                                                    struct nlattr *port[]);
1320         int                     (*ndo_get_vf_port)(struct net_device *dev,
1321                                                    int vf, struct sk_buff *skb);
1322         int                     (*ndo_set_vf_guid)(struct net_device *dev,
1323                                                    int vf, u64 guid,
1324                                                    int guid_type);
1325         int                     (*ndo_set_vf_rss_query_en)(
1326                                                    struct net_device *dev,
1327                                                    int vf, bool setting);
1328         int                     (*ndo_setup_tc)(struct net_device *dev,
1329                                                 enum tc_setup_type type,
1330                                                 void *type_data);
1331 #if IS_ENABLED(CONFIG_FCOE)
1332         int                     (*ndo_fcoe_enable)(struct net_device *dev);
1333         int                     (*ndo_fcoe_disable)(struct net_device *dev);
1334         int                     (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1335                                                       u16 xid,
1336                                                       struct scatterlist *sgl,
1337                                                       unsigned int sgc);
1338         int                     (*ndo_fcoe_ddp_done)(struct net_device *dev,
1339                                                      u16 xid);
1340         int                     (*ndo_fcoe_ddp_target)(struct net_device *dev,
1341                                                        u16 xid,
1342                                                        struct scatterlist *sgl,
1343                                                        unsigned int sgc);
1344         int                     (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1345                                                         struct netdev_fcoe_hbainfo *hbainfo);
1346 #endif
1347 
1348 #if IS_ENABLED(CONFIG_LIBFCOE)
1349 #define NETDEV_FCOE_WWNN 0
1350 #define NETDEV_FCOE_WWPN 1
1351         int                     (*ndo_fcoe_get_wwn)(struct net_device *dev,
1352                                                     u64 *wwn, int type);
1353 #endif
1354 
1355 #ifdef CONFIG_RFS_ACCEL
1356         int                     (*ndo_rx_flow_steer)(struct net_device *dev,
1357                                                      const struct sk_buff *skb,
1358                                                      u16 rxq_index,
1359                                                      u32 flow_id);
1360 #endif
1361         int                     (*ndo_add_slave)(struct net_device *dev,
1362                                                  struct net_device *slave_dev,
1363                                                  struct netlink_ext_ack *extack);
1364         int                     (*ndo_del_slave)(struct net_device *dev,
1365                                                  struct net_device *slave_dev);
1366         netdev_features_t       (*ndo_fix_features)(struct net_device *dev,
1367                                                     netdev_features_t features);
1368         int                     (*ndo_set_features)(struct net_device *dev,
1369                                                     netdev_features_t features);
1370         int                     (*ndo_neigh_construct)(struct net_device *dev,
1371                                                        struct neighbour *n);
1372         void                    (*ndo_neigh_destroy)(struct net_device *dev,
1373                                                      struct neighbour *n);
1374 
1375         int                     (*ndo_fdb_add)(struct ndmsg *ndm,
1376                                                struct nlattr *tb[],
1377                                                struct net_device *dev,
1378                                                const unsigned char *addr,
1379                                                u16 vid,
1380                                                u16 flags,
1381                                                struct netlink_ext_ack *extack);
1382         int                     (*ndo_fdb_del)(struct ndmsg *ndm,
1383                                                struct nlattr *tb[],
1384                                                struct net_device *dev,
1385                                                const unsigned char *addr,
1386                                                u16 vid);
1387         int                     (*ndo_fdb_dump)(struct sk_buff *skb,
1388                                                 struct netlink_callback *cb,
1389                                                 struct net_device *dev,
1390                                                 struct net_device *filter_dev,
1391                                                 int *idx);
1392         int                     (*ndo_fdb_get)(struct sk_buff *skb,
1393                                                struct nlattr *tb[],
1394                                                struct net_device *dev,
1395                                                const unsigned char *addr,
1396                                                u16 vid, u32 portid, u32 seq,
1397                                                struct netlink_ext_ack *extack);
1398         int                     (*ndo_bridge_setlink)(struct net_device *dev,
1399                                                       struct nlmsghdr *nlh,
1400                                                       u16 flags,
1401                                                       struct netlink_ext_ack *extack);
1402         int                     (*ndo_bridge_getlink)(struct sk_buff *skb,
1403                                                       u32 pid, u32 seq,
1404                                                       struct net_device *dev,
1405                                                       u32 filter_mask,
1406                                                       int nlflags);
1407         int                     (*ndo_bridge_dellink)(struct net_device *dev,
1408                                                       struct nlmsghdr *nlh,
1409                                                       u16 flags);
1410         int                     (*ndo_change_carrier)(struct net_device *dev,
1411                                                       bool new_carrier);
1412         int                     (*ndo_get_phys_port_id)(struct net_device *dev,
1413                                                         struct netdev_phys_item_id *ppid);
1414         int                     (*ndo_get_port_parent_id)(struct net_device *dev,
1415                                                           struct netdev_phys_item_id *ppid);
1416         int                     (*ndo_get_phys_port_name)(struct net_device *dev,
1417                                                           char *name, size_t len);
1418         void                    (*ndo_udp_tunnel_add)(struct net_device *dev,
1419                                                       struct udp_tunnel_info *ti);
1420         void                    (*ndo_udp_tunnel_del)(struct net_device *dev,
1421                                                       struct udp_tunnel_info *ti);
1422         void*                   (*ndo_dfwd_add_station)(struct net_device *pdev,
1423                                                         struct net_device *dev);
1424         void                    (*ndo_dfwd_del_station)(struct net_device *pdev,
1425                                                         void *priv);
1426 
1427         int                     (*ndo_set_tx_maxrate)(struct net_device *dev,
1428                                                       int queue_index,
1429                                                       u32 maxrate);
1430         int                     (*ndo_get_iflink)(const struct net_device *dev);
1431         int                     (*ndo_change_proto_down)(struct net_device *dev,
1432                                                          bool proto_down);
1433         int                     (*ndo_fill_metadata_dst)(struct net_device *dev,
1434                                                        struct sk_buff *skb);
1435         void                    (*ndo_set_rx_headroom)(struct net_device *dev,
1436                                                        int needed_headroom);
1437         int                     (*ndo_bpf)(struct net_device *dev,
1438                                            struct netdev_bpf *bpf);
1439         int                     (*ndo_xdp_xmit)(struct net_device *dev, int n,
1440                                                 struct xdp_frame **xdp,
1441                                                 u32 flags);
1442         int                     (*ndo_xsk_wakeup)(struct net_device *dev,
1443                                                   u32 queue_id, u32 flags);
1444         struct devlink_port *   (*ndo_get_devlink_port)(struct net_device *dev);
1445 };
1446 
1447 /**
1448  * enum net_device_priv_flags - &struct net_device priv_flags
1449  *
1450  * These are the &struct net_device, they are only set internally
1451  * by drivers and used in the kernel. These flags are invisible to
1452  * userspace; this means that the order of these flags can change
1453  * during any kernel release.
1454  *
1455  * You should have a pretty good reason to be extending these flags.
1456  *
1457  * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1458  * @IFF_EBRIDGE: Ethernet bridging device
1459  * @IFF_BONDING: bonding master or slave
1460  * @IFF_ISATAP: ISATAP interface (RFC4214)
1461  * @IFF_WAN_HDLC: WAN HDLC device
1462  * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1463  *      release skb->dst
1464  * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1465  * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1466  * @IFF_MACVLAN_PORT: device used as macvlan port
1467  * @IFF_BRIDGE_PORT: device used as bridge port
1468  * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1469  * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1470  * @IFF_UNICAST_FLT: Supports unicast filtering
1471  * @IFF_TEAM_PORT: device used as team port
1472  * @IFF_SUPP_NOFCS: device supports sending custom FCS
1473  * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1474  *      change when it's running
1475  * @IFF_MACVLAN: Macvlan device
1476  * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1477  *      underlying stacked devices
1478  * @IFF_L3MDEV_MASTER: device is an L3 master device
1479  * @IFF_NO_QUEUE: device can run without qdisc attached
1480  * @IFF_OPENVSWITCH: device is a Open vSwitch master
1481  * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1482  * @IFF_TEAM: device is a team device
1483  * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1484  * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1485  *      entity (i.e. the master device for bridged veth)
1486  * @IFF_MACSEC: device is a MACsec device
1487  * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1488  * @IFF_FAILOVER: device is a failover master device
1489  * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1490  * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1491  * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1492  */
1493 enum netdev_priv_flags {
1494         IFF_802_1Q_VLAN                 = 1<<0,
1495         IFF_EBRIDGE                     = 1<<1,
1496         IFF_BONDING                     = 1<<2,
1497         IFF_ISATAP                      = 1<<3,
1498         IFF_WAN_HDLC                    = 1<<4,
1499         IFF_XMIT_DST_RELEASE            = 1<<5,
1500         IFF_DONT_BRIDGE                 = 1<<6,
1501         IFF_DISABLE_NETPOLL             = 1<<7,
1502         IFF_MACVLAN_PORT                = 1<<8,
1503         IFF_BRIDGE_PORT                 = 1<<9,
1504         IFF_OVS_DATAPATH                = 1<<10,
1505         IFF_TX_SKB_SHARING              = 1<<11,
1506         IFF_UNICAST_FLT                 = 1<<12,
1507         IFF_TEAM_PORT                   = 1<<13,
1508         IFF_SUPP_NOFCS                  = 1<<14,
1509         IFF_LIVE_ADDR_CHANGE            = 1<<15,
1510         IFF_MACVLAN                     = 1<<16,
1511         IFF_XMIT_DST_RELEASE_PERM       = 1<<17,
1512         IFF_L3MDEV_MASTER               = 1<<18,
1513         IFF_NO_QUEUE                    = 1<<19,
1514         IFF_OPENVSWITCH                 = 1<<20,
1515         IFF_L3MDEV_SLAVE                = 1<<21,
1516         IFF_TEAM                        = 1<<22,
1517         IFF_RXFH_CONFIGURED             = 1<<23,
1518         IFF_PHONY_HEADROOM              = 1<<24,
1519         IFF_MACSEC                      = 1<<25,
1520         IFF_NO_RX_HANDLER               = 1<<26,
1521         IFF_FAILOVER                    = 1<<27,
1522         IFF_FAILOVER_SLAVE              = 1<<28,
1523         IFF_L3MDEV_RX_HANDLER           = 1<<29,
1524         IFF_LIVE_RENAME_OK              = 1<<30,
1525 };
1526 
1527 #define IFF_802_1Q_VLAN                 IFF_802_1Q_VLAN
1528 #define IFF_EBRIDGE                     IFF_EBRIDGE
1529 #define IFF_BONDING                     IFF_BONDING
1530 #define IFF_ISATAP                      IFF_ISATAP
1531 #define IFF_WAN_HDLC                    IFF_WAN_HDLC
1532 #define IFF_XMIT_DST_RELEASE            IFF_XMIT_DST_RELEASE
1533 #define IFF_DONT_BRIDGE                 IFF_DONT_BRIDGE
1534 #define IFF_DISABLE_NETPOLL             IFF_DISABLE_NETPOLL
1535 #define IFF_MACVLAN_PORT                IFF_MACVLAN_PORT
1536 #define IFF_BRIDGE_PORT                 IFF_BRIDGE_PORT
1537 #define IFF_OVS_DATAPATH                IFF_OVS_DATAPATH
1538 #define IFF_TX_SKB_SHARING              IFF_TX_SKB_SHARING
1539 #define IFF_UNICAST_FLT                 IFF_UNICAST_FLT
1540 #define IFF_TEAM_PORT                   IFF_TEAM_PORT
1541 #define IFF_SUPP_NOFCS                  IFF_SUPP_NOFCS
1542 #define IFF_LIVE_ADDR_CHANGE            IFF_LIVE_ADDR_CHANGE
1543 #define IFF_MACVLAN                     IFF_MACVLAN
1544 #define IFF_XMIT_DST_RELEASE_PERM       IFF_XMIT_DST_RELEASE_PERM
1545 #define IFF_L3MDEV_MASTER               IFF_L3MDEV_MASTER
1546 #define IFF_NO_QUEUE                    IFF_NO_QUEUE
1547 #define IFF_OPENVSWITCH                 IFF_OPENVSWITCH
1548 #define IFF_L3MDEV_SLAVE                IFF_L3MDEV_SLAVE
1549 #define IFF_TEAM                        IFF_TEAM
1550 #define IFF_RXFH_CONFIGURED             IFF_RXFH_CONFIGURED
1551 #define IFF_MACSEC                      IFF_MACSEC
1552 #define IFF_NO_RX_HANDLER               IFF_NO_RX_HANDLER
1553 #define IFF_FAILOVER                    IFF_FAILOVER
1554 #define IFF_FAILOVER_SLAVE              IFF_FAILOVER_SLAVE
1555 #define IFF_L3MDEV_RX_HANDLER           IFF_L3MDEV_RX_HANDLER
1556 #define IFF_LIVE_RENAME_OK              IFF_LIVE_RENAME_OK
1557 
1558 /**
1559  *      struct net_device - The DEVICE structure.
1560  *
1561  *      Actually, this whole structure is a big mistake.  It mixes I/O
1562  *      data with strictly "high-level" data, and it has to know about
1563  *      almost every data structure used in the INET module.
1564  *
1565  *      @name:  This is the first field of the "visible" part of this structure
1566  *              (i.e. as seen by users in the "Space.c" file).  It is the name
1567  *              of the interface.
1568  *
1569  *      @name_hlist:    Device name hash chain, please keep it close to name[]
1570  *      @ifalias:       SNMP alias
1571  *      @mem_end:       Shared memory end
1572  *      @mem_start:     Shared memory start
1573  *      @base_addr:     Device I/O address
1574  *      @irq:           Device IRQ number
1575  *
1576  *      @state:         Generic network queuing layer state, see netdev_state_t
1577  *      @dev_list:      The global list of network devices
1578  *      @napi_list:     List entry used for polling NAPI devices
1579  *      @unreg_list:    List entry  when we are unregistering the
1580  *                      device; see the function unregister_netdev
1581  *      @close_list:    List entry used when we are closing the device
1582  *      @ptype_all:     Device-specific packet handlers for all protocols
1583  *      @ptype_specific: Device-specific, protocol-specific packet handlers
1584  *
1585  *      @adj_list:      Directly linked devices, like slaves for bonding
1586  *      @features:      Currently active device features
1587  *      @hw_features:   User-changeable features
1588  *
1589  *      @wanted_features:       User-requested features
1590  *      @vlan_features:         Mask of features inheritable by VLAN devices
1591  *
1592  *      @hw_enc_features:       Mask of features inherited by encapsulating devices
1593  *                              This field indicates what encapsulation
1594  *                              offloads the hardware is capable of doing,
1595  *                              and drivers will need to set them appropriately.
1596  *
1597  *      @mpls_features: Mask of features inheritable by MPLS
1598  *
1599  *      @ifindex:       interface index
1600  *      @group:         The group the device belongs to
1601  *
1602  *      @stats:         Statistics struct, which was left as a legacy, use
1603  *                      rtnl_link_stats64 instead
1604  *
1605  *      @rx_dropped:    Dropped packets by core network,
1606  *                      do not use this in drivers
1607  *      @tx_dropped:    Dropped packets by core network,
1608  *                      do not use this in drivers
1609  *      @rx_nohandler:  nohandler dropped packets by core network on
1610  *                      inactive devices, do not use this in drivers
1611  *      @carrier_up_count:      Number of times the carrier has been up
1612  *      @carrier_down_count:    Number of times the carrier has been down
1613  *
1614  *      @wireless_handlers:     List of functions to handle Wireless Extensions,
1615  *                              instead of ioctl,
1616  *                              see <net/iw_handler.h> for details.
1617  *      @wireless_data: Instance data managed by the core of wireless extensions
1618  *
1619  *      @netdev_ops:    Includes several pointers to callbacks,
1620  *                      if one wants to override the ndo_*() functions
1621  *      @ethtool_ops:   Management operations
1622  *      @ndisc_ops:     Includes callbacks for different IPv6 neighbour
1623  *                      discovery handling. Necessary for e.g. 6LoWPAN.
1624  *      @header_ops:    Includes callbacks for creating,parsing,caching,etc
1625  *                      of Layer 2 headers.
1626  *
1627  *      @flags:         Interface flags (a la BSD)
1628  *      @priv_flags:    Like 'flags' but invisible to userspace,
1629  *                      see if.h for the definitions
1630  *      @gflags:        Global flags ( kept as legacy )
1631  *      @padded:        How much padding added by alloc_netdev()
1632  *      @operstate:     RFC2863 operstate
1633  *      @link_mode:     Mapping policy to operstate
1634  *      @if_port:       Selectable AUI, TP, ...
1635  *      @dma:           DMA channel
1636  *      @mtu:           Interface MTU value
1637  *      @min_mtu:       Interface Minimum MTU value
1638  *      @max_mtu:       Interface Maximum MTU value
1639  *      @type:          Interface hardware type
1640  *      @hard_header_len: Maximum hardware header length.
1641  *      @min_header_len:  Minimum hardware header length
1642  *
1643  *      @needed_headroom: Extra headroom the hardware may need, but not in all
1644  *                        cases can this be guaranteed
1645  *      @needed_tailroom: Extra tailroom the hardware may need, but not in all
1646  *                        cases can this be guaranteed. Some cases also use
1647  *                        LL_MAX_HEADER instead to allocate the skb
1648  *
1649  *      interface address info:
1650  *
1651  *      @perm_addr:             Permanent hw address
1652  *      @addr_assign_type:      Hw address assignment type
1653  *      @addr_len:              Hardware address length
1654  *      @upper_level:           Maximum depth level of upper devices.
1655  *      @lower_level:           Maximum depth level of lower devices.
1656  *      @neigh_priv_len:        Used in neigh_alloc()
1657  *      @dev_id:                Used to differentiate devices that share
1658  *                              the same link layer address
1659  *      @dev_port:              Used to differentiate devices that share
1660  *                              the same function
1661  *      @addr_list_lock:        XXX: need comments on this one
1662  *      @uc_promisc:            Counter that indicates promiscuous mode
1663  *                              has been enabled due to the need to listen to
1664  *                              additional unicast addresses in a device that
1665  *                              does not implement ndo_set_rx_mode()
1666  *      @uc:                    unicast mac addresses
1667  *      @mc:                    multicast mac addresses
1668  *      @dev_addrs:             list of device hw addresses
1669  *      @queues_kset:           Group of all Kobjects in the Tx and RX queues
1670  *      @promiscuity:           Number of times the NIC is told to work in
1671  *                              promiscuous mode; if it becomes 0 the NIC will
1672  *                              exit promiscuous mode
1673  *      @allmulti:              Counter, enables or disables allmulticast mode
1674  *
1675  *      @vlan_info:     VLAN info
1676  *      @dsa_ptr:       dsa specific data
1677  *      @tipc_ptr:      TIPC specific data
1678  *      @atalk_ptr:     AppleTalk link
1679  *      @ip_ptr:        IPv4 specific data
1680  *      @dn_ptr:        DECnet specific data
1681  *      @ip6_ptr:       IPv6 specific data
1682  *      @ax25_ptr:      AX.25 specific data
1683  *      @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1684  *
1685  *      @dev_addr:      Hw address (before bcast,
1686  *                      because most packets are unicast)
1687  *
1688  *      @_rx:                   Array of RX queues
1689  *      @num_rx_queues:         Number of RX queues
1690  *                              allocated at register_netdev() time
1691  *      @real_num_rx_queues:    Number of RX queues currently active in device
1692  *
1693  *      @rx_handler:            handler for received packets
1694  *      @rx_handler_data:       XXX: need comments on this one
1695  *      @miniq_ingress:         ingress/clsact qdisc specific data for
1696  *                              ingress processing
1697  *      @ingress_queue:         XXX: need comments on this one
1698  *      @broadcast:             hw bcast address
1699  *
1700  *      @rx_cpu_rmap:   CPU reverse-mapping for RX completion interrupts,
1701  *                      indexed by RX queue number. Assigned by driver.
1702  *                      This must only be set if the ndo_rx_flow_steer
1703  *                      operation is defined
1704  *      @index_hlist:           Device index hash chain
1705  *
1706  *      @_tx:                   Array of TX queues
1707  *      @num_tx_queues:         Number of TX queues allocated at alloc_netdev_mq() time
1708  *      @real_num_tx_queues:    Number of TX queues currently active in device
1709  *      @qdisc:                 Root qdisc from userspace point of view
1710  *      @tx_queue_len:          Max frames per queue allowed
1711  *      @tx_global_lock:        XXX: need comments on this one
1712  *
1713  *      @xps_maps:      XXX: need comments on this one
1714  *      @miniq_egress:          clsact qdisc specific data for
1715  *                              egress processing
1716  *      @watchdog_timeo:        Represents the timeout that is used by
1717  *                              the watchdog (see dev_watchdog())
1718  *      @watchdog_timer:        List of timers
1719  *
1720  *      @pcpu_refcnt:           Number of references to this device
1721  *      @todo_list:             Delayed register/unregister
1722  *      @link_watch_list:       XXX: need comments on this one
1723  *
1724  *      @reg_state:             Register/unregister state machine
1725  *      @dismantle:             Device is going to be freed
1726  *      @rtnl_link_state:       This enum represents the phases of creating
1727  *                              a new link
1728  *
1729  *      @needs_free_netdev:     Should unregister perform free_netdev?
1730  *      @priv_destructor:       Called from unregister
1731  *      @npinfo:                XXX: need comments on this one
1732  *      @nd_net:                Network namespace this network device is inside
1733  *
1734  *      @ml_priv:       Mid-layer private
1735  *      @lstats:        Loopback statistics
1736  *      @tstats:        Tunnel statistics
1737  *      @dstats:        Dummy statistics
1738  *      @vstats:        Virtual ethernet statistics
1739  *
1740  *      @garp_port:     GARP
1741  *      @mrp_port:      MRP
1742  *
1743  *      @dev:           Class/net/name entry
1744  *      @sysfs_groups:  Space for optional device, statistics and wireless
1745  *                      sysfs groups
1746  *
1747  *      @sysfs_rx_queue_group:  Space for optional per-rx queue attributes
1748  *      @rtnl_link_ops: Rtnl_link_ops
1749  *
1750  *      @gso_max_size:  Maximum size of generic segmentation offload
1751  *      @gso_max_segs:  Maximum number of segments that can be passed to the
1752  *                      NIC for GSO
1753  *
1754  *      @dcbnl_ops:     Data Center Bridging netlink ops
1755  *      @num_tc:        Number of traffic classes in the net device
1756  *      @tc_to_txq:     XXX: need comments on this one
1757  *      @prio_tc_map:   XXX: need comments on this one
1758  *
1759  *      @fcoe_ddp_xid:  Max exchange id for FCoE LRO by ddp
1760  *
1761  *      @priomap:       XXX: need comments on this one
1762  *      @phydev:        Physical device may attach itself
1763  *                      for hardware timestamping
1764  *      @sfp_bus:       attached &struct sfp_bus structure.
1765  *      @qdisc_tx_busylock_key: lockdep class annotating Qdisc->busylock
1766  *                              spinlock
1767  *      @qdisc_running_key:     lockdep class annotating Qdisc->running seqcount
1768  *      @qdisc_xmit_lock_key:   lockdep class annotating
1769  *                              netdev_queue->_xmit_lock spinlock
1770  *      @addr_list_lock_key:    lockdep class annotating
1771  *                              net_device->addr_list_lock spinlock
1772  *
1773  *      @proto_down:    protocol port state information can be sent to the
1774  *                      switch driver and used to set the phys state of the
1775  *                      switch port.
1776  *
1777  *      @wol_enabled:   Wake-on-LAN is enabled
1778  *
1779  *      FIXME: cleanup struct net_device such that network protocol info
1780  *      moves out.
1781  */
1782 
1783 struct net_device {
1784         char                    name[IFNAMSIZ];
1785         struct hlist_node       name_hlist;
1786         struct dev_ifalias      __rcu *ifalias;
1787         /*
1788          *      I/O specific fields
1789          *      FIXME: Merge these and struct ifmap into one
1790          */
1791         unsigned long           mem_end;
1792         unsigned long           mem_start;
1793         unsigned long           base_addr;
1794         int                     irq;
1795 
1796         /*
1797          *      Some hardware also needs these fields (state,dev_list,
1798          *      napi_list,unreg_list,close_list) but they are not
1799          *      part of the usual set specified in Space.c.
1800          */
1801 
1802         unsigned long           state;
1803 
1804         struct list_head        dev_list;
1805         struct list_head        napi_list;
1806         struct list_head        unreg_list;
1807         struct list_head        close_list;
1808         struct list_head        ptype_all;
1809         struct list_head        ptype_specific;
1810 
1811         struct {
1812                 struct list_head upper;
1813                 struct list_head lower;
1814         } adj_list;
1815 
1816         netdev_features_t       features;
1817         netdev_features_t       hw_features;
1818         netdev_features_t       wanted_features;
1819         netdev_features_t       vlan_features;
1820         netdev_features_t       hw_enc_features;
1821         netdev_features_t       mpls_features;
1822         netdev_features_t       gso_partial_features;
1823 
1824         int                     ifindex;
1825         int                     group;
1826 
1827         struct net_device_stats stats;
1828 
1829         atomic_long_t           rx_dropped;
1830         atomic_long_t           tx_dropped;
1831         atomic_long_t           rx_nohandler;
1832 
1833         /* Stats to monitor link on/off, flapping */
1834         atomic_t                carrier_up_count;
1835         atomic_t                carrier_down_count;
1836 
1837 #ifdef CONFIG_WIRELESS_EXT
1838         const struct iw_handler_def *wireless_handlers;
1839         struct iw_public_data   *wireless_data;
1840 #endif
1841         const struct net_device_ops *netdev_ops;
1842         const struct ethtool_ops *ethtool_ops;
1843 #ifdef CONFIG_NET_L3_MASTER_DEV
1844         const struct l3mdev_ops *l3mdev_ops;
1845 #endif
1846 #if IS_ENABLED(CONFIG_IPV6)
1847         const struct ndisc_ops *ndisc_ops;
1848 #endif
1849 
1850 #ifdef CONFIG_XFRM_OFFLOAD
1851         const struct xfrmdev_ops *xfrmdev_ops;
1852 #endif
1853 
1854 #if IS_ENABLED(CONFIG_TLS_DEVICE)
1855         const struct tlsdev_ops *tlsdev_ops;
1856 #endif
1857 
1858         const struct header_ops *header_ops;
1859 
1860         unsigned int            flags;
1861         unsigned int            priv_flags;
1862 
1863         unsigned short          gflags;
1864         unsigned short          padded;
1865 
1866         unsigned char           operstate;
1867         unsigned char           link_mode;
1868 
1869         unsigned char           if_port;
1870         unsigned char           dma;
1871 
1872         /* Note : dev->mtu is often read without holding a lock.
1873          * Writers usually hold RTNL.
1874          * It is recommended to use READ_ONCE() to annotate the reads,
1875          * and to use WRITE_ONCE() to annotate the writes.
1876          */
1877         unsigned int            mtu;
1878         unsigned int            min_mtu;
1879         unsigned int            max_mtu;
1880         unsigned short          type;
1881         unsigned short          hard_header_len;
1882         unsigned char           min_header_len;
1883 
1884         unsigned short          needed_headroom;
1885         unsigned short          needed_tailroom;
1886 
1887         /* Interface address info. */
1888         unsigned char           perm_addr[MAX_ADDR_LEN];
1889         unsigned char           addr_assign_type;
1890         unsigned char           addr_len;
1891         unsigned char           upper_level;
1892         unsigned char           lower_level;
1893         unsigned short          neigh_priv_len;
1894         unsigned short          dev_id;
1895         unsigned short          dev_port;
1896         spinlock_t              addr_list_lock;
1897         unsigned char           name_assign_type;
1898         bool                    uc_promisc;
1899         struct netdev_hw_addr_list      uc;
1900         struct netdev_hw_addr_list      mc;
1901         struct netdev_hw_addr_list      dev_addrs;
1902 
1903 #ifdef CONFIG_SYSFS
1904         struct kset             *queues_kset;
1905 #endif
1906         unsigned int            promiscuity;
1907         unsigned int            allmulti;
1908 
1909 
1910         /* Protocol-specific pointers */
1911 
1912 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1913         struct vlan_info __rcu  *vlan_info;
1914 #endif
1915 #if IS_ENABLED(CONFIG_NET_DSA)
1916         struct dsa_port         *dsa_ptr;
1917 #endif
1918 #if IS_ENABLED(CONFIG_TIPC)
1919         struct tipc_bearer __rcu *tipc_ptr;
1920 #endif
1921 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
1922         void                    *atalk_ptr;
1923 #endif
1924         struct in_device __rcu  *ip_ptr;
1925 #if IS_ENABLED(CONFIG_DECNET)
1926         struct dn_dev __rcu     *dn_ptr;
1927 #endif
1928         struct inet6_dev __rcu  *ip6_ptr;
1929 #if IS_ENABLED(CONFIG_AX25)
1930         void                    *ax25_ptr;
1931 #endif
1932         struct wireless_dev     *ieee80211_ptr;
1933         struct wpan_dev         *ieee802154_ptr;
1934 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1935         struct mpls_dev __rcu   *mpls_ptr;
1936 #endif
1937 
1938 /*
1939  * Cache lines mostly used on receive path (including eth_type_trans())
1940  */
1941         /* Interface address info used in eth_type_trans() */
1942         unsigned char           *dev_addr;
1943 
1944         struct netdev_rx_queue  *_rx;
1945         unsigned int            num_rx_queues;
1946         unsigned int            real_num_rx_queues;
1947 
1948         struct bpf_prog __rcu   *xdp_prog;
1949         unsigned long           gro_flush_timeout;
1950         rx_handler_func_t __rcu *rx_handler;
1951         void __rcu              *rx_handler_data;
1952 
1953 #ifdef CONFIG_NET_CLS_ACT
1954         struct mini_Qdisc __rcu *miniq_ingress;
1955 #endif
1956         struct netdev_queue __rcu *ingress_queue;
1957 #ifdef CONFIG_NETFILTER_INGRESS
1958         struct nf_hook_entries __rcu *nf_hooks_ingress;
1959 #endif
1960 
1961         unsigned char           broadcast[MAX_ADDR_LEN];
1962 #ifdef CONFIG_RFS_ACCEL
1963         struct cpu_rmap         *rx_cpu_rmap;
1964 #endif
1965         struct hlist_node       index_hlist;
1966 
1967 /*
1968  * Cache lines mostly used on transmit path
1969  */
1970         struct netdev_queue     *_tx ____cacheline_aligned_in_smp;
1971         unsigned int            num_tx_queues;
1972         unsigned int            real_num_tx_queues;
1973         struct Qdisc            *qdisc;
1974 #ifdef CONFIG_NET_SCHED
1975         DECLARE_HASHTABLE       (qdisc_hash, 4);
1976 #endif
1977         unsigned int            tx_queue_len;
1978         spinlock_t              tx_global_lock;
1979         int                     watchdog_timeo;
1980 
1981 #ifdef CONFIG_XPS
1982         struct xps_dev_maps __rcu *xps_cpus_map;
1983         struct xps_dev_maps __rcu *xps_rxqs_map;
1984 #endif
1985 #ifdef CONFIG_NET_CLS_ACT
1986         struct mini_Qdisc __rcu *miniq_egress;
1987 #endif
1988 
1989         /* These may be needed for future network-power-down code. */
1990         struct timer_list       watchdog_timer;
1991 
1992         int __percpu            *pcpu_refcnt;
1993         struct list_head        todo_list;
1994 
1995         struct list_head        link_watch_list;
1996 
1997         enum { NETREG_UNINITIALIZED=0,
1998                NETREG_REGISTERED,       /* completed register_netdevice */
1999                NETREG_UNREGISTERING,    /* called unregister_netdevice */
2000                NETREG_UNREGISTERED,     /* completed unregister todo */
2001                NETREG_RELEASED,         /* called free_netdev */
2002                NETREG_DUMMY,            /* dummy device for NAPI poll */
2003         } reg_state:8;
2004 
2005         bool dismantle;
2006 
2007         enum {
2008                 RTNL_LINK_INITIALIZED,
2009                 RTNL_LINK_INITIALIZING,
2010         } rtnl_link_state:16;
2011 
2012         bool needs_free_netdev;
2013         void (*priv_destructor)(struct net_device *dev);
2014 
2015 #ifdef CONFIG_NETPOLL
2016         struct netpoll_info __rcu       *npinfo;
2017 #endif
2018 
2019         possible_net_t                  nd_net;
2020 
2021         /* mid-layer private */
2022         union {
2023                 void                                    *ml_priv;
2024                 struct pcpu_lstats __percpu             *lstats;
2025                 struct pcpu_sw_netstats __percpu        *tstats;
2026                 struct pcpu_dstats __percpu             *dstats;
2027         };
2028 
2029 #if IS_ENABLED(CONFIG_GARP)
2030         struct garp_port __rcu  *garp_port;
2031 #endif
2032 #if IS_ENABLED(CONFIG_MRP)
2033         struct mrp_port __rcu   *mrp_port;
2034 #endif
2035 
2036         struct device           dev;
2037         const struct attribute_group *sysfs_groups[4];
2038         const struct attribute_group *sysfs_rx_queue_group;
2039 
2040         const struct rtnl_link_ops *rtnl_link_ops;
2041 
2042         /* for setting kernel sock attribute on TCP connection setup */
2043 #define GSO_MAX_SIZE            65536
2044         unsigned int            gso_max_size;
2045 #define GSO_MAX_SEGS            65535
2046         u16                     gso_max_segs;
2047 
2048 #ifdef CONFIG_DCB
2049         const struct dcbnl_rtnl_ops *dcbnl_ops;
2050 #endif
2051         s16                     num_tc;
2052         struct netdev_tc_txq    tc_to_txq[TC_MAX_QUEUE];
2053         u8                      prio_tc_map[TC_BITMASK + 1];
2054 
2055 #if IS_ENABLED(CONFIG_FCOE)
2056         unsigned int            fcoe_ddp_xid;
2057 #endif
2058 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2059         struct netprio_map __rcu *priomap;
2060 #endif
2061         struct phy_device       *phydev;
2062         struct sfp_bus          *sfp_bus;
2063         struct lock_class_key   qdisc_tx_busylock_key;
2064         struct lock_class_key   qdisc_running_key;
2065         struct lock_class_key   qdisc_xmit_lock_key;
2066         struct lock_class_key   addr_list_lock_key;
2067         bool                    proto_down;
2068         unsigned                wol_enabled:1;
2069 };
2070 #define to_net_dev(d) container_of(d, struct net_device, dev)
2071 
2072 static inline bool netif_elide_gro(const struct net_device *dev)
2073 {
2074         if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2075                 return true;
2076         return false;
2077 }
2078 
2079 #define NETDEV_ALIGN            32
2080 
2081 static inline
2082 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2083 {
2084         return dev->prio_tc_map[prio & TC_BITMASK];
2085 }
2086 
2087 static inline
2088 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2089 {
2090         if (tc >= dev->num_tc)
2091                 return -EINVAL;
2092 
2093         dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2094         return 0;
2095 }
2096 
2097 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2098 void netdev_reset_tc(struct net_device *dev);
2099 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2100 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2101 
2102 static inline
2103 int netdev_get_num_tc(struct net_device *dev)
2104 {
2105         return dev->num_tc;
2106 }
2107 
2108 void netdev_unbind_sb_channel(struct net_device *dev,
2109                               struct net_device *sb_dev);
2110 int netdev_bind_sb_channel_queue(struct net_device *dev,
2111                                  struct net_device *sb_dev,
2112                                  u8 tc, u16 count, u16 offset);
2113 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2114 static inline int netdev_get_sb_channel(struct net_device *dev)
2115 {
2116         return max_t(int, -dev->num_tc, 0);
2117 }
2118 
2119 static inline
2120 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2121                                          unsigned int index)
2122 {
2123         return &dev->_tx[index];
2124 }
2125 
2126 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2127                                                     const struct sk_buff *skb)
2128 {
2129         return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2130 }
2131 
2132 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2133                                             void (*f)(struct net_device *,
2134                                                       struct netdev_queue *,
2135                                                       void *),
2136                                             void *arg)
2137 {
2138         unsigned int i;
2139 
2140         for (i = 0; i < dev->num_tx_queues; i++)
2141                 f(dev, &dev->_tx[i], arg);
2142 }
2143 
2144 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2145                      struct net_device *sb_dev);
2146 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2147                                          struct sk_buff *skb,
2148                                          struct net_device *sb_dev);
2149 
2150 /* returns the headroom that the master device needs to take in account
2151  * when forwarding to this dev
2152  */
2153 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2154 {
2155         return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2156 }
2157 
2158 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2159 {
2160         if (dev->netdev_ops->ndo_set_rx_headroom)
2161                 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2162 }
2163 
2164 /* set the device rx headroom to the dev's default */
2165 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2166 {
2167         netdev_set_rx_headroom(dev, -1);
2168 }
2169 
2170 /*
2171  * Net namespace inlines
2172  */
2173 static inline
2174 struct net *dev_net(const struct net_device *dev)
2175 {
2176         return read_pnet(&dev->nd_net);
2177 }
2178 
2179 static inline
2180 void dev_net_set(struct net_device *dev, struct net *net)
2181 {
2182         write_pnet(&dev->nd_net, net);
2183 }
2184 
2185 /**
2186  *      netdev_priv - access network device private data
2187  *      @dev: network device
2188  *
2189  * Get network device private data
2190  */
2191 static inline void *netdev_priv(const struct net_device *dev)
2192 {
2193         return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2194 }
2195 
2196 /* Set the sysfs physical device reference for the network logical device
2197  * if set prior to registration will cause a symlink during initialization.
2198  */
2199 #define SET_NETDEV_DEV(net, pdev)       ((net)->dev.parent = (pdev))
2200 
2201 /* Set the sysfs device type for the network logical device to allow
2202  * fine-grained identification of different network device types. For
2203  * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2204  */
2205 #define SET_NETDEV_DEVTYPE(net, devtype)        ((net)->dev.type = (devtype))
2206 
2207 /* Default NAPI poll() weight
2208  * Device drivers are strongly advised to not use bigger value
2209  */
2210 #define NAPI_POLL_WEIGHT 64
2211 
2212 /**
2213  *      netif_napi_add - initialize a NAPI context
2214  *      @dev:  network device
2215  *      @napi: NAPI context
2216  *      @poll: polling function
2217  *      @weight: default weight
2218  *
2219  * netif_napi_add() must be used to initialize a NAPI context prior to calling
2220  * *any* of the other NAPI-related functions.
2221  */
2222 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2223                     int (*poll)(struct napi_struct *, int), int weight);
2224 
2225 /**
2226  *      netif_tx_napi_add - initialize a NAPI context
2227  *      @dev:  network device
2228  *      @napi: NAPI context
2229  *      @poll: polling function
2230  *      @weight: default weight
2231  *
2232  * This variant of netif_napi_add() should be used from drivers using NAPI
2233  * to exclusively poll a TX queue.
2234  * This will avoid we add it into napi_hash[], thus polluting this hash table.
2235  */
2236 static inline void netif_tx_napi_add(struct net_device *dev,
2237                                      struct napi_struct *napi,
2238                                      int (*poll)(struct napi_struct *, int),
2239                                      int weight)
2240 {
2241         set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2242         netif_napi_add(dev, napi, poll, weight);
2243 }
2244 
2245 /**
2246  *  netif_napi_del - remove a NAPI context
2247  *  @napi: NAPI context
2248  *
2249  *  netif_napi_del() removes a NAPI context from the network device NAPI list
2250  */
2251 void netif_napi_del(struct napi_struct *napi);
2252 
2253 struct napi_gro_cb {
2254         /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2255         void    *frag0;
2256 
2257         /* Length of frag0. */
2258         unsigned int frag0_len;
2259 
2260         /* This indicates where we are processing relative to skb->data. */
2261         int     data_offset;
2262 
2263         /* This is non-zero if the packet cannot be merged with the new skb. */
2264         u16     flush;
2265 
2266         /* Save the IP ID here and check when we get to the transport layer */
2267         u16     flush_id;
2268 
2269         /* Number of segments aggregated. */
2270         u16     count;
2271 
2272         /* Start offset for remote checksum offload */
2273         u16     gro_remcsum_start;
2274 
2275         /* jiffies when first packet was created/queued */
2276         unsigned long age;
2277 
2278         /* Used in ipv6_gro_receive() and foo-over-udp */
2279         u16     proto;
2280 
2281         /* This is non-zero if the packet may be of the same flow. */
2282         u8      same_flow:1;
2283 
2284         /* Used in tunnel GRO receive */
2285         u8      encap_mark:1;
2286 
2287         /* GRO checksum is valid */
2288         u8      csum_valid:1;
2289 
2290         /* Number of checksums via CHECKSUM_UNNECESSARY */
2291         u8      csum_cnt:3;
2292 
2293         /* Free the skb? */
2294         u8      free:2;
2295 #define NAPI_GRO_FREE             1
2296 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2297 
2298         /* Used in foo-over-udp, set in udp[46]_gro_receive */
2299         u8      is_ipv6:1;
2300 
2301         /* Used in GRE, set in fou/gue_gro_receive */
2302         u8      is_fou:1;
2303 
2304         /* Used to determine if flush_id can be ignored */
2305         u8      is_atomic:1;
2306 
2307         /* Number of gro_receive callbacks this packet already went through */
2308         u8 recursion_counter:4;
2309 
2310         /* 1 bit hole */
2311 
2312         /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2313         __wsum  csum;
2314 
2315         /* used in skb_gro_receive() slow path */
2316         struct sk_buff *last;
2317 };
2318 
2319 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2320 
2321 #define GRO_RECURSION_LIMIT 15
2322 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2323 {
2324         return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2325 }
2326 
2327 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2328 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2329                                                struct list_head *head,
2330                                                struct sk_buff *skb)
2331 {
2332         if (unlikely(gro_recursion_inc_test(skb))) {
2333                 NAPI_GRO_CB(skb)->flush |= 1;
2334                 return NULL;
2335         }
2336 
2337         return cb(head, skb);
2338 }
2339 
2340 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2341                                             struct sk_buff *);
2342 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2343                                                   struct sock *sk,
2344                                                   struct list_head *head,
2345                                                   struct sk_buff *skb)
2346 {
2347         if (unlikely(gro_recursion_inc_test(skb))) {
2348                 NAPI_GRO_CB(skb)->flush |= 1;
2349                 return NULL;
2350         }
2351 
2352         return cb(sk, head, skb);
2353 }
2354 
2355 struct packet_type {
2356         __be16                  type;   /* This is really htons(ether_type). */
2357         bool                    ignore_outgoing;
2358         struct net_device       *dev;   /* NULL is wildcarded here           */
2359         int                     (*func) (struct sk_buff *,
2360                                          struct net_device *,
2361                                          struct packet_type *,
2362                                          struct net_device *);
2363         void                    (*list_func) (struct list_head *,
2364                                               struct packet_type *,
2365                                               struct net_device *);
2366         bool                    (*id_match)(struct packet_type *ptype,
2367                                             struct sock *sk);
2368         void                    *af_packet_priv;
2369         struct list_head        list;
2370 };
2371 
2372 struct offload_callbacks {
2373         struct sk_buff          *(*gso_segment)(struct sk_buff *skb,
2374                                                 netdev_features_t features);
2375         struct sk_buff          *(*gro_receive)(struct list_head *head,
2376                                                 struct sk_buff *skb);
2377         int                     (*gro_complete)(struct sk_buff *skb, int nhoff);
2378 };
2379 
2380 struct packet_offload {
2381         __be16                   type;  /* This is really htons(ether_type). */
2382         u16                      priority;
2383         struct offload_callbacks callbacks;
2384         struct list_head         list;
2385 };
2386 
2387 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2388 struct pcpu_sw_netstats {
2389         u64     rx_packets;
2390         u64     rx_bytes;
2391         u64     tx_packets;
2392         u64     tx_bytes;
2393         struct u64_stats_sync   syncp;
2394 } __aligned(4 * sizeof(u64));
2395 
2396 struct pcpu_lstats {
2397         u64 packets;
2398         u64 bytes;
2399         struct u64_stats_sync syncp;
2400 } __aligned(2 * sizeof(u64));
2401 
2402 #define __netdev_alloc_pcpu_stats(type, gfp)                            \
2403 ({                                                                      \
2404         typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2405         if (pcpu_stats) {                                               \
2406                 int __cpu;                                              \
2407                 for_each_possible_cpu(__cpu) {                          \
2408                         typeof(type) *stat;                             \
2409                         stat = per_cpu_ptr(pcpu_stats, __cpu);          \
2410                         u64_stats_init(&stat->syncp);                   \
2411                 }                                                       \
2412         }                                                               \
2413         pcpu_stats;                                                     \
2414 })
2415 
2416 #define netdev_alloc_pcpu_stats(type)                                   \
2417         __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2418 
2419 enum netdev_lag_tx_type {
2420         NETDEV_LAG_TX_TYPE_UNKNOWN,
2421         NETDEV_LAG_TX_TYPE_RANDOM,
2422         NETDEV_LAG_TX_TYPE_BROADCAST,
2423         NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2424         NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2425         NETDEV_LAG_TX_TYPE_HASH,
2426 };
2427 
2428 enum netdev_lag_hash {
2429         NETDEV_LAG_HASH_NONE,
2430         NETDEV_LAG_HASH_L2,
2431         NETDEV_LAG_HASH_L34,
2432         NETDEV_LAG_HASH_L23,
2433         NETDEV_LAG_HASH_E23,
2434         NETDEV_LAG_HASH_E34,
2435         NETDEV_LAG_HASH_UNKNOWN,
2436 };
2437 
2438 struct netdev_lag_upper_info {
2439         enum netdev_lag_tx_type tx_type;
2440         enum netdev_lag_hash hash_type;
2441 };
2442 
2443 struct netdev_lag_lower_state_info {
2444         u8 link_up : 1,
2445            tx_enabled : 1;
2446 };
2447 
2448 #include <linux/notifier.h>
2449 
2450 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2451  * and the rtnetlink notification exclusion list in rtnetlink_event() when
2452  * adding new types.
2453  */
2454 enum netdev_cmd {
2455         NETDEV_UP       = 1,    /* For now you can't veto a device up/down */
2456         NETDEV_DOWN,
2457         NETDEV_REBOOT,          /* Tell a protocol stack a network interface
2458                                    detected a hardware crash and restarted
2459                                    - we can use this eg to kick tcp sessions
2460                                    once done */
2461         NETDEV_CHANGE,          /* Notify device state change */
2462         NETDEV_REGISTER,
2463         NETDEV_UNREGISTER,
2464         NETDEV_CHANGEMTU,       /* notify after mtu change happened */
2465         NETDEV_CHANGEADDR,      /* notify after the address change */
2466         NETDEV_PRE_CHANGEADDR,  /* notify before the address change */
2467         NETDEV_GOING_DOWN,
2468         NETDEV_CHANGENAME,
2469         NETDEV_FEAT_CHANGE,
2470         NETDEV_BONDING_FAILOVER,
2471         NETDEV_PRE_UP,
2472         NETDEV_PRE_TYPE_CHANGE,
2473         NETDEV_POST_TYPE_CHANGE,
2474         NETDEV_POST_INIT,
2475         NETDEV_RELEASE,
2476         NETDEV_NOTIFY_PEERS,
2477         NETDEV_JOIN,
2478         NETDEV_CHANGEUPPER,
2479         NETDEV_RESEND_IGMP,
2480         NETDEV_PRECHANGEMTU,    /* notify before mtu change happened */
2481         NETDEV_CHANGEINFODATA,
2482         NETDEV_BONDING_INFO,
2483         NETDEV_PRECHANGEUPPER,
2484         NETDEV_CHANGELOWERSTATE,
2485         NETDEV_UDP_TUNNEL_PUSH_INFO,
2486         NETDEV_UDP_TUNNEL_DROP_INFO,
2487         NETDEV_CHANGE_TX_QUEUE_LEN,
2488         NETDEV_CVLAN_FILTER_PUSH_INFO,
2489         NETDEV_CVLAN_FILTER_DROP_INFO,
2490         NETDEV_SVLAN_FILTER_PUSH_INFO,
2491         NETDEV_SVLAN_FILTER_DROP_INFO,
2492 };
2493 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2494 
2495 int register_netdevice_notifier(struct notifier_block *nb);
2496 int unregister_netdevice_notifier(struct notifier_block *nb);
2497 
2498 struct netdev_notifier_info {
2499         struct net_device       *dev;
2500         struct netlink_ext_ack  *extack;
2501 };
2502 
2503 struct netdev_notifier_info_ext {
2504         struct netdev_notifier_info info; /* must be first */
2505         union {
2506                 u32 mtu;
2507         } ext;
2508 };
2509 
2510 struct netdev_notifier_change_info {
2511         struct netdev_notifier_info info; /* must be first */
2512         unsigned int flags_changed;
2513 };
2514 
2515 struct netdev_notifier_changeupper_info {
2516         struct netdev_notifier_info info; /* must be first */
2517         struct net_device *upper_dev; /* new upper dev */
2518         bool master; /* is upper dev master */
2519         bool linking; /* is the notification for link or unlink */
2520         void *upper_info; /* upper dev info */
2521 };
2522 
2523 struct netdev_notifier_changelowerstate_info {
2524         struct netdev_notifier_info info; /* must be first */
2525         void *lower_state_info; /* is lower dev state */
2526 };
2527 
2528 struct netdev_notifier_pre_changeaddr_info {
2529         struct netdev_notifier_info info; /* must be first */
2530         const unsigned char *dev_addr;
2531 };
2532 
2533 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2534                                              struct net_device *dev)
2535 {
2536         info->dev = dev;
2537         info->extack = NULL;
2538 }
2539 
2540 static inline struct net_device *
2541 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2542 {
2543         return info->dev;
2544 }
2545 
2546 static inline struct netlink_ext_ack *
2547 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2548 {
2549         return info->extack;
2550 }
2551 
2552 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2553 
2554 
2555 extern rwlock_t                         dev_base_lock;          /* Device list lock */
2556 
2557 #define for_each_netdev(net, d)         \
2558                 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2559 #define for_each_netdev_reverse(net, d) \
2560                 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2561 #define for_each_netdev_rcu(net, d)             \
2562                 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2563 #define for_each_netdev_safe(net, d, n) \
2564                 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2565 #define for_each_netdev_continue(net, d)                \
2566                 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2567 #define for_each_netdev_continue_rcu(net, d)            \
2568         list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2569 #define for_each_netdev_in_bond_rcu(bond, slave)        \
2570                 for_each_netdev_rcu(&init_net, slave)   \
2571                         if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2572 #define net_device_entry(lh)    list_entry(lh, struct net_device, dev_list)
2573 
2574 static inline struct net_device *next_net_device(struct net_device *dev)
2575 {
2576         struct list_head *lh;
2577         struct net *net;
2578 
2579         net = dev_net(dev);
2580         lh = dev->dev_list.next;
2581         return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2582 }
2583 
2584 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2585 {
2586         struct list_head *lh;
2587         struct net *net;
2588 
2589         net = dev_net(dev);
2590         lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2591         return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2592 }
2593 
2594 static inline struct net_device *first_net_device(struct net *net)
2595 {
2596         return list_empty(&net->dev_base_head) ? NULL :
2597                 net_device_entry(net->dev_base_head.next);
2598 }
2599 
2600 static inline struct net_device *first_net_device_rcu(struct net *net)
2601 {
2602         struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2603 
2604         return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2605 }
2606 
2607 int netdev_boot_setup_check(struct net_device *dev);
2608 unsigned long netdev_boot_base(const char *prefix, int unit);
2609 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2610                                        const char *hwaddr);
2611 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2612 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2613 void dev_add_pack(struct packet_type *pt);
2614 void dev_remove_pack(struct packet_type *pt);
2615 void __dev_remove_pack(struct packet_type *pt);
2616 void dev_add_offload(struct packet_offload *po);
2617 void dev_remove_offload(struct packet_offload *po);
2618 
2619 int dev_get_iflink(const struct net_device *dev);
2620 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2621 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2622                                       unsigned short mask);
2623 struct net_device *dev_get_by_name(struct net *net, const char *name);
2624 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2625 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2626 int dev_alloc_name(struct net_device *dev, const char *name);
2627 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2628 void dev_close(struct net_device *dev);
2629 void dev_close_many(struct list_head *head, bool unlink);
2630 void dev_disable_lro(struct net_device *dev);
2631 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2632 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2633                      struct net_device *sb_dev);
2634 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2635                        struct net_device *sb_dev);
2636 int dev_queue_xmit(struct sk_buff *skb);
2637 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2638 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2639 int register_netdevice(struct net_device *dev);
2640 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2641 void unregister_netdevice_many(struct list_head *head);
2642 static inline void unregister_netdevice(struct net_device *dev)
2643 {
2644         unregister_netdevice_queue(dev, NULL);
2645 }
2646 
2647 int netdev_refcnt_read(const struct net_device *dev);
2648 void free_netdev(struct net_device *dev);
2649 void netdev_freemem(struct net_device *dev);
2650 void synchronize_net(void);
2651 int init_dummy_netdev(struct net_device *dev);
2652 
2653 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2654 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2655 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2656 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2657 int netdev_get_name(struct net *net, char *name, int ifindex);
2658 int dev_restart(struct net_device *dev);
2659 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2660 
2661 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2662 {
2663         return NAPI_GRO_CB(skb)->data_offset;
2664 }
2665 
2666 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2667 {
2668         return skb->len - NAPI_GRO_CB(skb)->data_offset;
2669 }
2670 
2671 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2672 {
2673         NAPI_GRO_CB(skb)->data_offset += len;
2674 }
2675 
2676 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2677                                         unsigned int offset)
2678 {
2679         return NAPI_GRO_CB(skb)->frag0 + offset;
2680 }
2681 
2682 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2683 {
2684         return NAPI_GRO_CB(skb)->frag0_len < hlen;
2685 }
2686 
2687 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2688 {
2689         NAPI_GRO_CB(skb)->frag0 = NULL;
2690         NAPI_GRO_CB(skb)->frag0_len = 0;
2691 }
2692 
2693 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2694                                         unsigned int offset)
2695 {
2696         if (!pskb_may_pull(skb, hlen))
2697                 return NULL;
2698 
2699         skb_gro_frag0_invalidate(skb);
2700         return skb->data + offset;
2701 }
2702 
2703 static inline void *skb_gro_network_header(struct sk_buff *skb)
2704 {
2705         return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2706                skb_network_offset(skb);
2707 }
2708 
2709 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2710                                         const void *start, unsigned int len)
2711 {
2712         if (NAPI_GRO_CB(skb)->csum_valid)
2713                 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2714                                                   csum_partial(start, len, 0));
2715 }
2716 
2717 /* GRO checksum functions. These are logical equivalents of the normal
2718  * checksum functions (in skbuff.h) except that they operate on the GRO
2719  * offsets and fields in sk_buff.
2720  */
2721 
2722 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2723 
2724 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2725 {
2726         return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2727 }
2728 
2729 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2730                                                       bool zero_okay,
2731                                                       __sum16 check)
2732 {
2733         return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2734                 skb_checksum_start_offset(skb) <
2735                  skb_gro_offset(skb)) &&
2736                 !skb_at_gro_remcsum_start(skb) &&
2737                 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2738                 (!zero_okay || check));
2739 }
2740 
2741 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2742                                                            __wsum psum)
2743 {
2744         if (NAPI_GRO_CB(skb)->csum_valid &&
2745             !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2746                 return 0;
2747 
2748         NAPI_GRO_CB(skb)->csum = psum;
2749 
2750         return __skb_gro_checksum_complete(skb);
2751 }
2752 
2753 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2754 {
2755         if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2756                 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2757                 NAPI_GRO_CB(skb)->csum_cnt--;
2758         } else {
2759                 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2760                  * verified a new top level checksum or an encapsulated one
2761                  * during GRO. This saves work if we fallback to normal path.
2762                  */
2763                 __skb_incr_checksum_unnecessary(skb);
2764         }
2765 }
2766 
2767 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check,       \
2768                                     compute_pseudo)                     \
2769 ({                                                                      \
2770         __sum16 __ret = 0;                                              \
2771         if (__skb_gro_checksum_validate_needed(skb, zero_okay, check))  \
2772                 __ret = __skb_gro_checksum_validate_complete(skb,       \
2773                                 compute_pseudo(skb, proto));            \
2774         if (!__ret)                                                     \
2775                 skb_gro_incr_csum_unnecessary(skb);                     \
2776         __ret;                                                          \
2777 })
2778 
2779 #define skb_gro_checksum_validate(skb, proto, compute_pseudo)           \
2780         __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2781 
2782 #define skb_gro_checksum_validate_zero_check(skb, proto, check,         \
2783                                              compute_pseudo)            \
2784         __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2785 
2786 #define skb_gro_checksum_simple_validate(skb)                           \
2787         __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2788 
2789 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2790 {
2791         return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2792                 !NAPI_GRO_CB(skb)->csum_valid);
2793 }
2794 
2795 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2796                                               __sum16 check, __wsum pseudo)
2797 {
2798         NAPI_GRO_CB(skb)->csum = ~pseudo;
2799         NAPI_GRO_CB(skb)->csum_valid = 1;
2800 }
2801 
2802 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2803 do {                                                                    \
2804         if (__skb_gro_checksum_convert_check(skb))                      \
2805                 __skb_gro_checksum_convert(skb, check,                  \
2806                                            compute_pseudo(skb, proto)); \
2807 } while (0)
2808 
2809 struct gro_remcsum {
2810         int offset;
2811         __wsum delta;
2812 };
2813 
2814 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2815 {
2816         grc->offset = 0;
2817         grc->delta = 0;
2818 }
2819 
2820 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2821                                             unsigned int off, size_t hdrlen,
2822                                             int start, int offset,
2823                                             struct gro_remcsum *grc,
2824                                             bool nopartial)
2825 {
2826         __wsum delta;
2827         size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2828 
2829         BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2830 
2831         if (!nopartial) {
2832                 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2833                 return ptr;
2834         }
2835 
2836         ptr = skb_gro_header_fast(skb, off);
2837         if (skb_gro_header_hard(skb, off + plen)) {
2838                 ptr = skb_gro_header_slow(skb, off + plen, off);
2839                 if (!ptr)
2840                         return NULL;
2841         }
2842 
2843         delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2844                                start, offset);
2845 
2846         /* Adjust skb->csum since we changed the packet */
2847         NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2848 
2849         grc->offset = off + hdrlen + offset;
2850         grc->delta = delta;
2851 
2852         return ptr;
2853 }
2854 
2855 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2856                                            struct gro_remcsum *grc)
2857 {
2858         void *ptr;
2859         size_t plen = grc->offset + sizeof(u16);
2860 
2861         if (!grc->delta)
2862                 return;
2863 
2864         ptr = skb_gro_header_fast(skb, grc->offset);
2865         if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2866                 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2867                 if (!ptr)
2868                         return;
2869         }
2870 
2871         remcsum_unadjust((__sum16 *)ptr, grc->delta);
2872 }
2873 
2874 #ifdef CONFIG_XFRM_OFFLOAD
2875 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2876 {
2877         if (PTR_ERR(pp) != -EINPROGRESS)
2878                 NAPI_GRO_CB(skb)->flush |= flush;
2879 }
2880 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2881                                                struct sk_buff *pp,
2882                                                int flush,
2883                                                struct gro_remcsum *grc)
2884 {
2885         if (PTR_ERR(pp) != -EINPROGRESS) {
2886                 NAPI_GRO_CB(skb)->flush |= flush;
2887                 skb_gro_remcsum_cleanup(skb, grc);
2888                 skb->remcsum_offload = 0;
2889         }
2890 }
2891 #else
2892 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2893 {
2894         NAPI_GRO_CB(skb)->flush |= flush;
2895 }
2896 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2897                                                struct sk_buff *pp,
2898                                                int flush,
2899                                                struct gro_remcsum *grc)
2900 {
2901         NAPI_GRO_CB(skb)->flush |= flush;
2902         skb_gro_remcsum_cleanup(skb, grc);
2903         skb->remcsum_offload = 0;
2904 }
2905 #endif
2906 
2907 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2908                                   unsigned short type,
2909                                   const void *daddr, const void *saddr,
2910                                   unsigned int len)
2911 {
2912         if (!dev->header_ops || !dev->header_ops->create)
2913                 return 0;
2914 
2915         return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2916 }
2917 
2918 static inline int dev_parse_header(const struct sk_buff *skb,
2919                                    unsigned char *haddr)
2920 {
2921         const struct net_device *dev = skb->dev;
2922 
2923         if (!dev->header_ops || !dev->header_ops->parse)
2924                 return 0;
2925         return dev->header_ops->parse(skb, haddr);
2926 }
2927 
2928 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
2929 {
2930         const struct net_device *dev = skb->dev;
2931 
2932         if (!dev->header_ops || !dev->header_ops->parse_protocol)
2933                 return 0;
2934         return dev->header_ops->parse_protocol(skb);
2935 }
2936 
2937 /* ll_header must have at least hard_header_len allocated */
2938 static inline bool dev_validate_header(const struct net_device *dev,
2939                                        char *ll_header, int len)
2940 {
2941         if (likely(len >= dev->hard_header_len))
2942                 return true;
2943         if (len < dev->min_header_len)
2944                 return false;
2945 
2946         if (capable(CAP_SYS_RAWIO)) {
2947                 memset(ll_header + len, 0, dev->hard_header_len - len);
2948                 return true;
2949         }
2950 
2951         if (dev->header_ops && dev->header_ops->validate)
2952                 return dev->header_ops->validate(ll_header, len);
2953 
2954         return false;
2955 }
2956 
2957 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
2958                            int len, int size);
2959 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2960 static inline int unregister_gifconf(unsigned int family)
2961 {
2962         return register_gifconf(family, NULL);
2963 }
2964 
2965 #ifdef CONFIG_NET_FLOW_LIMIT
2966 #define FLOW_LIMIT_HISTORY      (1 << 7)  /* must be ^2 and !overflow buckets */
2967 struct sd_flow_limit {
2968         u64                     count;
2969         unsigned int            num_buckets;
2970         unsigned int            history_head;
2971         u16                     history[FLOW_LIMIT_HISTORY];
2972         u8                      buckets[];
2973 };
2974 
2975 extern int netdev_flow_limit_table_len;
2976 #endif /* CONFIG_NET_FLOW_LIMIT */
2977 
2978 /*
2979  * Incoming packets are placed on per-CPU queues
2980  */
2981 struct softnet_data {
2982         struct list_head        poll_list;
2983         struct sk_buff_head     process_queue;
2984 
2985         /* stats */
2986         unsigned int            processed;
2987         unsigned int            time_squeeze;
2988         unsigned int            received_rps;
2989 #ifdef CONFIG_RPS
2990         struct softnet_data     *rps_ipi_list;
2991 #endif
2992 #ifdef CONFIG_NET_FLOW_LIMIT
2993         struct sd_flow_limit __rcu *flow_limit;
2994 #endif
2995         struct Qdisc            *output_queue;
2996         struct Qdisc            **output_queue_tailp;
2997         struct sk_buff          *completion_queue;
2998 #ifdef CONFIG_XFRM_OFFLOAD
2999         struct sk_buff_head     xfrm_backlog;
3000 #endif
3001         /* written and read only by owning cpu: */
3002         struct {
3003                 u16 recursion;
3004                 u8  more;
3005         } xmit;
3006 #ifdef CONFIG_RPS
3007         /* input_queue_head should be written by cpu owning this struct,
3008          * and only read by other cpus. Worth using a cache line.
3009          */
3010         unsigned int            input_queue_head ____cacheline_aligned_in_smp;
3011 
3012         /* Elements below can be accessed between CPUs for RPS/RFS */
3013         call_single_data_t      csd ____cacheline_aligned_in_smp;
3014         struct softnet_data     *rps_ipi_next;
3015         unsigned int            cpu;
3016         unsigned int            input_queue_tail;
3017 #endif
3018         unsigned int            dropped;
3019         struct sk_buff_head     input_pkt_queue;
3020         struct napi_struct      backlog;
3021 
3022 };
3023 
3024 static inline void input_queue_head_incr(struct softnet_data *sd)
3025 {
3026 #ifdef CONFIG_RPS
3027         sd->input_queue_head++;
3028 #endif
3029 }
3030 
3031 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3032                                               unsigned int *qtail)
3033 {
3034 #ifdef CONFIG_RPS
3035         *qtail = ++sd->input_queue_tail;
3036 #endif
3037 }
3038 
3039 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3040 
3041 static inline int dev_recursion_level(void)
3042 {
3043         return this_cpu_read(softnet_data.xmit.recursion);
3044 }
3045 
3046 #define XMIT_RECURSION_LIMIT    10
3047 static inline bool dev_xmit_recursion(void)
3048 {
3049         return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3050                         XMIT_RECURSION_LIMIT);
3051 }
3052 
3053 static inline void dev_xmit_recursion_inc(void)
3054 {
3055         __this_cpu_inc(softnet_data.xmit.recursion);
3056 }
3057 
3058 static inline void dev_xmit_recursion_dec(void)
3059 {
3060         __this_cpu_dec(softnet_data.xmit.recursion);
3061 }
3062 
3063 void __netif_schedule(struct Qdisc *q);
3064 void netif_schedule_queue(struct netdev_queue *txq);
3065 
3066 static inline void netif_tx_schedule_all(struct net_device *dev)
3067 {
3068         unsigned int i;
3069 
3070         for (i = 0; i < dev->num_tx_queues; i++)
3071                 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3072 }
3073 
3074 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3075 {
3076         clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3077 }
3078 
3079 /**
3080  *      netif_start_queue - allow transmit
3081  *      @dev: network device
3082  *
3083  *      Allow upper layers to call the device hard_start_xmit routine.
3084  */
3085 static inline void netif_start_queue(struct net_device *dev)
3086 {
3087         netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3088 }
3089 
3090 static inline void netif_tx_start_all_queues(struct net_device *dev)
3091 {
3092         unsigned int i;
3093 
3094         for (i = 0; i < dev->num_tx_queues; i++) {
3095                 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3096                 netif_tx_start_queue(txq);
3097         }
3098 }
3099 
3100 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3101 
3102 /**
3103  *      netif_wake_queue - restart transmit
3104  *      @dev: network device
3105  *
3106  *      Allow upper layers to call the device hard_start_xmit routine.
3107  *      Used for flow control when transmit resources are available.
3108  */
3109 static inline void netif_wake_queue(struct net_device *dev)
3110 {
3111         netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3112 }
3113 
3114 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3115 {
3116         unsigned int i;
3117 
3118         for (i = 0; i < dev->num_tx_queues; i++) {
3119                 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3120                 netif_tx_wake_queue(txq);
3121         }
3122 }
3123 
3124 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3125 {
3126         set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3127 }
3128 
3129 /**
3130  *      netif_stop_queue - stop transmitted packets
3131  *      @dev: network device
3132  *
3133  *      Stop upper layers calling the device hard_start_xmit routine.
3134  *      Used for flow control when transmit resources are unavailable.
3135  */
3136 static inline void netif_stop_queue(struct net_device *dev)
3137 {
3138         netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3139 }
3140 
3141 void netif_tx_stop_all_queues(struct net_device *dev);
3142 void netdev_update_lockdep_key(struct net_device *dev);
3143 
3144 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3145 {
3146         return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3147 }
3148 
3149 /**
3150  *      netif_queue_stopped - test if transmit queue is flowblocked
3151  *      @dev: network device
3152  *
3153  *      Test if transmit queue on device is currently unable to send.
3154  */
3155 static inline bool netif_queue_stopped(const struct net_device *dev)
3156 {
3157         return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3158 }
3159 
3160 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3161 {
3162         return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3163 }
3164 
3165 static inline bool
3166 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3167 {
3168         return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3169 }
3170 
3171 static inline bool
3172 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3173 {
3174         return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3175 }
3176 
3177 /**
3178  *      netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3179  *      @dev_queue: pointer to transmit queue
3180  *
3181  * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3182  * to give appropriate hint to the CPU.
3183  */
3184 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3185 {
3186 #ifdef CONFIG_BQL
3187         prefetchw(&dev_queue->dql.num_queued);
3188 #endif
3189 }
3190 
3191 /**
3192  *      netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3193  *      @dev_queue: pointer to transmit queue
3194  *
3195  * BQL enabled drivers might use this helper in their TX completion path,
3196  * to give appropriate hint to the CPU.
3197  */
3198 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3199 {
3200 #ifdef CONFIG_BQL
3201         prefetchw(&dev_queue->dql.limit);
3202 #endif
3203 }
3204 
3205 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3206                                         unsigned int bytes)
3207 {
3208 #ifdef CONFIG_BQL
3209         dql_queued(&dev_queue->dql, bytes);
3210 
3211         if (likely(dql_avail(&dev_queue->dql) >= 0))
3212                 return;
3213 
3214         set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3215 
3216         /*
3217          * The XOFF flag must be set before checking the dql_avail below,
3218          * because in netdev_tx_completed_queue we update the dql_completed
3219          * before checking the XOFF flag.
3220          */
3221         smp_mb();
3222 
3223         /* check again in case another CPU has just made room avail */
3224         if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3225                 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3226 #endif
3227 }
3228 
3229 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3230  * that they should not test BQL status themselves.
3231  * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3232  * skb of a batch.
3233  * Returns true if the doorbell must be used to kick the NIC.
3234  */
3235 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3236                                           unsigned int bytes,
3237                                           bool xmit_more)
3238 {
3239         if (xmit_more) {
3240 #ifdef CONFIG_BQL
3241                 dql_queued(&dev_queue->dql, bytes);
3242 #endif
3243                 return netif_tx_queue_stopped(dev_queue);
3244         }
3245         netdev_tx_sent_queue(dev_queue, bytes);
3246         return true;
3247 }
3248 
3249 /**
3250  *      netdev_sent_queue - report the number of bytes queued to hardware
3251  *      @dev: network device
3252  *      @bytes: number of bytes queued to the hardware device queue
3253  *
3254  *      Report the number of bytes queued for sending/completion to the network
3255  *      device hardware queue. @bytes should be a good approximation and should
3256  *      exactly match netdev_completed_queue() @bytes
3257  */
3258 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3259 {
3260         netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3261 }
3262 
3263 static inline bool __netdev_sent_queue(struct net_device *dev,
3264                                        unsigned int bytes,
3265                                        bool xmit_more)
3266 {
3267         return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3268                                       xmit_more);
3269 }
3270 
3271 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3272                                              unsigned int pkts, unsigned int bytes)
3273 {
3274 #ifdef CONFIG_BQL
3275         if (unlikely(!bytes))
3276                 return;
3277 
3278         dql_completed(&dev_queue->dql, bytes);
3279 
3280         /*
3281          * Without the memory barrier there is a small possiblity that
3282          * netdev_tx_sent_queue will miss the update and cause the queue to
3283          * be stopped forever
3284          */
3285         smp_mb();
3286 
3287         if (unlikely(dql_avail(&dev_queue->dql) < 0))
3288                 return;
3289 
3290         if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3291                 netif_schedule_queue(dev_queue);
3292 #endif
3293 }
3294 
3295 /**
3296  *      netdev_completed_queue - report bytes and packets completed by device
3297  *      @dev: network device
3298  *      @pkts: actual number of packets sent over the medium
3299  *      @bytes: actual number of bytes sent over the medium
3300  *
3301  *      Report the number of bytes and packets transmitted by the network device
3302  *      hardware queue over the physical medium, @bytes must exactly match the
3303  *      @bytes amount passed to netdev_sent_queue()
3304  */
3305 static inline void netdev_completed_queue(struct net_device *dev,
3306                                           unsigned int pkts, unsigned int bytes)
3307 {
3308         netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3309 }
3310 
3311 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3312 {
3313 #ifdef CONFIG_BQL
3314         clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3315         dql_reset(&q->dql);
3316 #endif
3317 }
3318 
3319 /**
3320  *      netdev_reset_queue - reset the packets and bytes count of a network device
3321  *      @dev_queue: network device
3322  *
3323  *      Reset the bytes and packet count of a network device and clear the
3324  *      software flow control OFF bit for this network device
3325  */
3326 static inline void netdev_reset_queue(struct net_device *dev_queue)
3327 {
3328         netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3329 }
3330 
3331 /**
3332  *      netdev_cap_txqueue - check if selected tx queue exceeds device queues
3333  *      @dev: network device
3334  *      @queue_index: given tx queue index
3335  *
3336  *      Returns 0 if given tx queue index >= number of device tx queues,
3337  *      otherwise returns the originally passed tx queue index.
3338  */
3339 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3340 {
3341         if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3342                 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3343                                      dev->name, queue_index,
3344                                      dev->real_num_tx_queues);
3345                 return 0;
3346         }
3347 
3348         return queue_index;
3349 }
3350 
3351 /**
3352  *      netif_running - test if up
3353  *      @dev: network device
3354  *
3355  *      Test if the device has been brought up.
3356  */
3357 static inline bool netif_running(const struct net_device *dev)
3358 {
3359         return test_bit(__LINK_STATE_START, &dev->state);
3360 }
3361 
3362 /*
3363  * Routines to manage the subqueues on a device.  We only need start,
3364  * stop, and a check if it's stopped.  All other device management is
3365  * done at the overall netdevice level.
3366  * Also test the device if we're multiqueue.
3367  */
3368 
3369 /**
3370  *      netif_start_subqueue - allow sending packets on subqueue
3371  *      @dev: network device
3372  *      @queue_index: sub queue index
3373  *
3374  * Start individual transmit queue of a device with multiple transmit queues.
3375  */
3376 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3377 {
3378         struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3379 
3380         netif_tx_start_queue(txq);
3381 }
3382 
3383 /**
3384  *      netif_stop_subqueue - stop sending packets on subqueue
3385  *      @dev: network device
3386  *      @queue_index: sub queue index
3387  *
3388  * Stop individual transmit queue of a device with multiple transmit queues.
3389  */
3390 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3391 {
3392         struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3393         netif_tx_stop_queue(txq);
3394 }
3395 
3396 /**
3397  *      netif_subqueue_stopped - test status of subqueue
3398  *      @dev: network device
3399  *      @queue_index: sub queue index
3400  *
3401  * Check individual transmit queue of a device with multiple transmit queues.
3402  */
3403 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3404                                             u16 queue_index)
3405 {
3406         struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3407 
3408         return netif_tx_queue_stopped(txq);
3409 }
3410 
3411 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3412                                           struct sk_buff *skb)
3413 {
3414         return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3415 }
3416 
3417 /**
3418  *      netif_wake_subqueue - allow sending packets on subqueue
3419  *      @dev: network device
3420  *      @queue_index: sub queue index
3421  *
3422  * Resume individual transmit queue of a device with multiple transmit queues.
3423  */
3424 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3425 {
3426         struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3427 
3428         netif_tx_wake_queue(txq);
3429 }
3430 
3431 #ifdef CONFIG_XPS
3432 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3433                         u16 index);
3434 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3435                           u16 index, bool is_rxqs_map);
3436 
3437 /**
3438  *      netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3439  *      @j: CPU/Rx queue index
3440  *      @mask: bitmask of all cpus/rx queues
3441  *      @nr_bits: number of bits in the bitmask
3442  *
3443  * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3444  */
3445 static inline bool netif_attr_test_mask(unsigned long j,
3446                                         const unsigned long *mask,
3447                                         unsigned int nr_bits)
3448 {
3449         cpu_max_bits_warn(j, nr_bits);
3450         return test_bit(j, mask);
3451 }
3452 
3453 /**
3454  *      netif_attr_test_online - Test for online CPU/Rx queue
3455  *      @j: CPU/Rx queue index
3456  *      @online_mask: bitmask for CPUs/Rx queues that are online
3457  *      @nr_bits: number of bits in the bitmask
3458  *
3459  * Returns true if a CPU/Rx queue is online.
3460  */
3461 static inline bool netif_attr_test_online(unsigned long j,
3462                                           const unsigned long *online_mask,
3463                                           unsigned int nr_bits)
3464 {
3465         cpu_max_bits_warn(j, nr_bits);
3466 
3467         if (online_mask)
3468                 return test_bit(j, online_mask);
3469 
3470         return (j < nr_bits);
3471 }
3472 
3473 /**
3474  *      netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3475  *      @n: CPU/Rx queue index
3476  *      @srcp: the cpumask/Rx queue mask pointer
3477  *      @nr_bits: number of bits in the bitmask
3478  *
3479  * Returns >= nr_bits if no further CPUs/Rx queues set.
3480  */
3481 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3482                                                unsigned int nr_bits)
3483 {
3484         /* -1 is a legal arg here. */
3485         if (n != -1)
3486                 cpu_max_bits_warn(n, nr_bits);
3487 
3488         if (srcp)
3489                 return find_next_bit(srcp, nr_bits, n + 1);
3490 
3491         return n + 1;
3492 }
3493 
3494 /**
3495  *      netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p
3496  *      @n: CPU/Rx queue index
3497  *      @src1p: the first CPUs/Rx queues mask pointer
3498  *      @src2p: the second CPUs/Rx queues mask pointer
3499  *      @nr_bits: number of bits in the bitmask
3500  *
3501  * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3502  */
3503 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3504                                           const unsigned long *src2p,
3505                                           unsigned int nr_bits)
3506 {
3507         /* -1 is a legal arg here. */
3508         if (n != -1)
3509                 cpu_max_bits_warn(n, nr_bits);
3510 
3511         if (src1p && src2p)
3512                 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3513         else if (src1p)
3514                 return find_next_bit(src1p, nr_bits, n + 1);
3515         else if (src2p)
3516                 return find_next_bit(src2p, nr_bits, n + 1);
3517 
3518         return n + 1;
3519 }
3520 #else
3521 static inline int netif_set_xps_queue(struct net_device *dev,
3522                                       const struct cpumask *mask,
3523                                       u16 index)
3524 {
3525         return 0;
3526 }
3527 
3528 static inline int __netif_set_xps_queue(struct net_device *dev,
3529                                         const unsigned long *mask,
3530                                         u16 index, bool is_rxqs_map)
3531 {
3532         return 0;
3533 }
3534 #endif
3535 
3536 /**
3537  *      netif_is_multiqueue - test if device has multiple transmit queues
3538  *      @dev: network device
3539  *
3540  * Check if device has multiple transmit queues
3541  */
3542 static inline bool netif_is_multiqueue(const struct net_device *dev)
3543 {
3544         return dev->num_tx_queues > 1;
3545 }
3546 
3547 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3548 
3549 #ifdef CONFIG_SYSFS
3550 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3551 #else
3552 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3553                                                 unsigned int rxqs)
3554 {
3555         dev->real_num_rx_queues = rxqs;
3556         return 0;
3557 }
3558 #endif
3559 
3560 static inline struct netdev_rx_queue *
3561 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3562 {
3563         return dev->_rx + rxq;
3564 }
3565 
3566 #ifdef CONFIG_SYSFS
3567 static inline unsigned int get_netdev_rx_queue_index(
3568                 struct netdev_rx_queue *queue)
3569 {
3570         struct net_device *dev = queue->dev;
3571         int index = queue - dev->_rx;
3572 
3573         BUG_ON(index >= dev->num_rx_queues);
3574         return index;
3575 }
3576 #endif
3577 
3578 #define DEFAULT_MAX_NUM_RSS_QUEUES      (8)
3579 int netif_get_num_default_rss_queues(void);
3580 
3581 enum skb_free_reason {
3582         SKB_REASON_CONSUMED,
3583         SKB_REASON_DROPPED,
3584 };
3585 
3586 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3587 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3588 
3589 /*
3590  * It is not allowed to call kfree_skb() or consume_skb() from hardware
3591  * interrupt context or with hardware interrupts being disabled.
3592  * (in_irq() || irqs_disabled())
3593  *
3594  * We provide four helpers that can be used in following contexts :
3595  *
3596  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3597  *  replacing kfree_skb(skb)
3598  *
3599  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3600  *  Typically used in place of consume_skb(skb) in TX completion path
3601  *
3602  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3603  *  replacing kfree_skb(skb)
3604  *
3605  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3606  *  and consumed a packet. Used in place of consume_skb(skb)
3607  */
3608 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3609 {
3610         __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3611 }
3612 
3613 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3614 {
3615         __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3616 }
3617 
3618 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3619 {
3620         __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3621 }
3622 
3623 static inline void dev_consume_skb_any(struct sk_buff *skb)
3624 {
3625         __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3626 }
3627 
3628 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3629 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3630 int netif_rx(struct sk_buff *skb);
3631 int netif_rx_ni(struct sk_buff *skb);
3632 int netif_receive_skb(struct sk_buff *skb);
3633 int netif_receive_skb_core(struct sk_buff *skb);
3634 void netif_receive_skb_list(struct list_head *head);
3635 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3636 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3637 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3638 gro_result_t napi_gro_frags(struct napi_struct *napi);
3639 struct packet_offload *gro_find_receive_by_type(__be16 type);
3640 struct packet_offload *gro_find_complete_by_type(__be16 type);
3641 
3642 static inline void napi_free_frags(struct napi_struct *napi)
3643 {
3644         kfree_skb(napi->skb);
3645         napi->skb = NULL;
3646 }
3647 
3648 bool netdev_is_rx_handler_busy(struct net_device *dev);
3649 int netdev_rx_handler_register(struct net_device *dev,
3650                                rx_handler_func_t *rx_handler,
3651                                void *rx_handler_data);
3652 void netdev_rx_handler_unregister(struct net_device *dev);
3653 
3654 bool dev_valid_name(const char *name);
3655 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3656                 bool *need_copyout);
3657 int dev_ifconf(struct net *net, struct ifconf *, int);
3658 int dev_ethtool(struct net *net, struct ifreq *);
3659 unsigned int dev_get_flags(const struct net_device *);
3660 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3661                        struct netlink_ext_ack *extack);
3662 int dev_change_flags(struct net_device *dev, unsigned int flags,
3663                      struct netlink_ext_ack *extack);
3664 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3665                         unsigned int gchanges);
3666 int dev_change_name(struct net_device *, const char *);
3667 int dev_set_alias(struct net_device *, const char *, size_t);
3668 int dev_get_alias(const struct net_device *, char *, size_t);
3669 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3670 int __dev_set_mtu(struct net_device *, int);
3671 int dev_validate_mtu(struct net_device *dev, int mtu,
3672                      struct netlink_ext_ack *extack);
3673 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3674                     struct netlink_ext_ack *extack);
3675 int dev_set_mtu(struct net_device *, int);
3676 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3677 void dev_set_group(struct net_device *, int);
3678 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3679                               struct netlink_ext_ack *extack);
3680 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3681                         struct netlink_ext_ack *extack);
3682 int dev_change_carrier(struct net_device *, bool new_carrier);
3683 int dev_get_phys_port_id(struct net_device *dev,
3684                          struct netdev_phys_item_id *ppid);
3685 int dev_get_phys_port_name(struct net_device *dev,
3686                            char *name, size_t len);
3687 int dev_get_port_parent_id(struct net_device *dev,
3688                            struct netdev_phys_item_id *ppid, bool recurse);
3689 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3690 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3691 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3692 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3693 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3694                                     struct netdev_queue *txq, int *ret);
3695 
3696 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3697 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3698                       int fd, u32 flags);
3699 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3700                     enum bpf_netdev_command cmd);
3701 int xdp_umem_query(struct net_device *dev, u16 queue_id);
3702 
3703 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3704 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3705 bool is_skb_forwardable(const struct net_device *dev,
3706                         const struct sk_buff *skb);
3707 
3708 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3709                                                struct sk_buff *skb)
3710 {
3711         if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3712             unlikely(!is_skb_forwardable(dev, skb))) {
3713                 atomic_long_inc(&dev->rx_dropped);
3714                 kfree_skb(skb);
3715                 return NET_RX_DROP;
3716         }
3717 
3718         skb_scrub_packet(skb, true);
3719         skb->priority = 0;
3720         return 0;
3721 }
3722 
3723 bool dev_nit_active(struct net_device *dev);
3724 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3725 
3726 extern int              netdev_budget;
3727 extern unsigned int     netdev_budget_usecs;
3728 
3729 /* Called by rtnetlink.c:rtnl_unlock() */
3730 void netdev_run_todo(void);
3731 
3732 /**
3733  *      dev_put - release reference to device
3734  *      @dev: network device
3735  *
3736  * Release reference to device to allow it to be freed.
3737  */
3738 static inline void dev_put(struct net_device *dev)
3739 {
3740         this_cpu_dec(*dev->pcpu_refcnt);
3741 }
3742 
3743 /**
3744  *      dev_hold - get reference to device
3745  *      @dev: network device
3746  *
3747  * Hold reference to device to keep it from being freed.
3748  */
3749 static inline void dev_hold(struct net_device *dev)
3750 {
3751         this_cpu_inc(*dev->pcpu_refcnt);
3752 }
3753 
3754 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3755  * and _off may be called from IRQ context, but it is caller
3756  * who is responsible for serialization of these calls.
3757  *
3758  * The name carrier is inappropriate, these functions should really be
3759  * called netif_lowerlayer_*() because they represent the state of any
3760  * kind of lower layer not just hardware media.
3761  */
3762 
3763 void linkwatch_init_dev(struct net_device *dev);
3764 void linkwatch_fire_event(struct net_device *dev);
3765 void linkwatch_forget_dev(struct net_device *dev);
3766 
3767 /**
3768  *      netif_carrier_ok - test if carrier present
3769  *      @dev: network device
3770  *
3771  * Check if carrier is present on device
3772  */
3773 static inline bool netif_carrier_ok(const struct net_device *dev)
3774 {
3775         return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3776 }
3777 
3778 unsigned long dev_trans_start(struct net_device *dev);
3779 
3780 void __netdev_watchdog_up(struct net_device *dev);
3781 
3782 void netif_carrier_on(struct net_device *dev);
3783 
3784 void netif_carrier_off(struct net_device *dev);
3785 
3786 /**
3787  *      netif_dormant_on - mark device as dormant.
3788  *      @dev: network device
3789  *
3790  * Mark device as dormant (as per RFC2863).
3791  *
3792  * The dormant state indicates that the relevant interface is not
3793  * actually in a condition to pass packets (i.e., it is not 'up') but is
3794  * in a "pending" state, waiting for some external event.  For "on-
3795  * demand" interfaces, this new state identifies the situation where the
3796  * interface is waiting for events to place it in the up state.
3797  */
3798 static inline void netif_dormant_on(struct net_device *dev)
3799 {
3800         if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3801                 linkwatch_fire_event(dev);
3802 }
3803 
3804 /**
3805  *      netif_dormant_off - set device as not dormant.
3806  *      @dev: network device
3807  *
3808  * Device is not in dormant state.
3809  */
3810 static inline void netif_dormant_off(struct net_device *dev)
3811 {
3812         if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3813                 linkwatch_fire_event(dev);
3814 }
3815 
3816 /**
3817  *      netif_dormant - test if device is dormant
3818  *      @dev: network device
3819  *
3820  * Check if device is dormant.
3821  */
3822 static inline bool netif_dormant(const struct net_device *dev)
3823 {
3824         return test_bit(__LINK_STATE_DORMANT, &dev->state);
3825 }
3826 
3827 
3828 /**
3829  *      netif_oper_up - test if device is operational
3830  *      @dev: network device
3831  *
3832  * Check if carrier is operational
3833  */
3834 static inline bool netif_oper_up(const struct net_device *dev)
3835 {
3836         return (dev->operstate == IF_OPER_UP ||
3837                 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3838 }
3839 
3840 /**
3841  *      netif_device_present - is device available or removed
3842  *      @dev: network device
3843  *
3844  * Check if device has not been removed from system.
3845  */
3846 static inline bool netif_device_present(struct net_device *dev)
3847 {
3848         return test_bit(__LINK_STATE_PRESENT, &dev->state);
3849 }
3850 
3851 void netif_device_detach(struct net_device *dev);
3852 
3853 void netif_device_attach(struct net_device *dev);
3854 
3855 /*
3856  * Network interface message level settings
3857  */
3858 
3859 enum {
3860         NETIF_MSG_DRV           = 0x0001,
3861         NETIF_MSG_PROBE         = 0x0002,
3862         NETIF_MSG_LINK          = 0x0004,
3863         NETIF_MSG_TIMER         = 0x0008,
3864         NETIF_MSG_IFDOWN        = 0x0010,
3865         NETIF_MSG_IFUP          = 0x0020,
3866         NETIF_MSG_RX_ERR        = 0x0040,
3867         NETIF_MSG_TX_ERR        = 0x0080,
3868         NETIF_MSG_TX_QUEUED     = 0x0100,
3869         NETIF_MSG_INTR          = 0x0200,
3870         NETIF_MSG_TX_DONE       = 0x0400,
3871         NETIF_MSG_RX_STATUS     = 0x0800,
3872         NETIF_MSG_PKTDATA       = 0x1000,
3873         NETIF_MSG_HW            = 0x2000,
3874         NETIF_MSG_WOL           = 0x4000,
3875 };
3876 
3877 #define netif_msg_drv(p)        ((p)->msg_enable & NETIF_MSG_DRV)
3878 #define netif_msg_probe(p)      ((p)->msg_enable & NETIF_MSG_PROBE)
3879 #define netif_msg_link(p)       ((p)->msg_enable & NETIF_MSG_LINK)
3880 #define netif_msg_timer(p)      ((p)->msg_enable & NETIF_MSG_TIMER)
3881 #define netif_msg_ifdown(p)     ((p)->msg_enable & NETIF_MSG_IFDOWN)
3882 #define netif_msg_ifup(p)       ((p)->msg_enable & NETIF_MSG_IFUP)
3883 #define netif_msg_rx_err(p)     ((p)->msg_enable & NETIF_MSG_RX_ERR)
3884 #define netif_msg_tx_err(p)     ((p)->msg_enable & NETIF_MSG_TX_ERR)
3885 #define netif_msg_tx_queued(p)  ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3886 #define netif_msg_intr(p)       ((p)->msg_enable & NETIF_MSG_INTR)
3887 #define netif_msg_tx_done(p)    ((p)->msg_enable & NETIF_MSG_TX_DONE)
3888 #define netif_msg_rx_status(p)  ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3889 #define netif_msg_pktdata(p)    ((p)->msg_enable & NETIF_MSG_PKTDATA)
3890 #define netif_msg_hw(p)         ((p)->msg_enable & NETIF_MSG_HW)
3891 #define netif_msg_wol(p)        ((p)->msg_enable & NETIF_MSG_WOL)
3892 
3893 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3894 {
3895         /* use default */
3896         if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3897                 return default_msg_enable_bits;
3898         if (debug_value == 0)   /* no output */
3899                 return 0;
3900         /* set low N bits */
3901         return (1U << debug_value) - 1;
3902 }
3903 
3904 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3905 {
3906         spin_lock(&txq->_xmit_lock);
3907         txq->xmit_lock_owner = cpu;
3908 }
3909 
3910 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3911 {
3912         __acquire(&txq->_xmit_lock);
3913         return true;
3914 }
3915 
3916 static inline void __netif_tx_release(struct netdev_queue *txq)
3917 {
3918         __release(&txq->_xmit_lock);
3919 }
3920 
3921 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3922 {
3923         spin_lock_bh(&txq->_xmit_lock);
3924         txq->xmit_lock_owner = smp_processor_id();
3925 }
3926 
3927 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3928 {
3929         bool ok = spin_trylock(&txq->_xmit_lock);
3930         if (likely(ok))
3931                 txq->xmit_lock_owner = smp_processor_id();
3932         return ok;
3933 }
3934 
3935 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3936 {
3937         txq->xmit_lock_owner = -1;
3938         spin_unlock(&txq->_xmit_lock);
3939 }
3940 
3941 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3942 {
3943         txq->xmit_lock_owner = -1;
3944         spin_unlock_bh(&txq->_xmit_lock);
3945 }
3946 
3947 static inline void txq_trans_update(struct netdev_queue *txq)
3948 {
3949         if (txq->xmit_lock_owner != -1)
3950                 txq->trans_start = jiffies;
3951 }
3952 
3953 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3954 static inline void netif_trans_update(struct net_device *dev)
3955 {
3956         struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3957 
3958         if (txq->trans_start != jiffies)
3959                 txq->trans_start = jiffies;
3960 }
3961 
3962 /**
3963  *      netif_tx_lock - grab network device transmit lock
3964  *      @dev: network device
3965  *
3966  * Get network device transmit lock
3967  */
3968 static inline void netif_tx_lock(struct net_device *dev)
3969 {
3970         unsigned int i;
3971         int cpu;
3972 
3973         spin_lock(&dev->tx_global_lock);
3974         cpu = smp_processor_id();
3975         for (i = 0; i < dev->num_tx_queues; i++) {
3976                 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3977 
3978                 /* We are the only thread of execution doing a
3979                  * freeze, but we have to grab the _xmit_lock in
3980                  * order to synchronize with threads which are in
3981                  * the ->hard_start_xmit() handler and already
3982                  * checked the frozen bit.
3983                  */
3984                 __netif_tx_lock(txq, cpu);
3985                 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3986                 __netif_tx_unlock(txq);
3987         }
3988 }
3989 
3990 static inline void netif_tx_lock_bh(struct net_device *dev)
3991 {
3992         local_bh_disable();
3993         netif_tx_lock(dev);
3994 }
3995 
3996 static inline void netif_tx_unlock(struct net_device *dev)
3997 {
3998         unsigned int i;
3999 
4000         for (i = 0; i < dev->num_tx_queues; i++) {
4001                 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4002 
4003                 /* No need to grab the _xmit_lock here.  If the
4004                  * queue is not stopped for another reason, we
4005                  * force a schedule.
4006                  */
4007                 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4008                 netif_schedule_queue(txq);
4009         }
4010         spin_unlock(&dev->tx_global_lock);
4011 }
4012 
4013 static inline void netif_tx_unlock_bh(struct net_device *dev)
4014 {
4015         netif_tx_unlock(dev);
4016         local_bh_enable();
4017 }
4018 
4019 #define HARD_TX_LOCK(dev, txq, cpu) {                   \
4020         if ((dev->features & NETIF_F_LLTX) == 0) {      \
4021                 __netif_tx_lock(txq, cpu);              \
4022         } else {                                        \
4023                 __netif_tx_acquire(txq);                \
4024         }                                               \
4025 }
4026 
4027 #define HARD_TX_TRYLOCK(dev, txq)                       \
4028         (((dev->features & NETIF_F_LLTX) == 0) ?        \
4029                 __netif_tx_trylock(txq) :               \
4030                 __netif_tx_acquire(txq))
4031 
4032 #define HARD_TX_UNLOCK(dev, txq) {                      \
4033         if ((dev->features & NETIF_F_LLTX) == 0) {      \
4034                 __netif_tx_unlock(txq);                 \
4035         } else {                                        \
4036                 __netif_tx_release(txq);                \
4037         }                                               \
4038 }
4039 
4040 static inline void netif_tx_disable(struct net_device *dev)
4041 {
4042         unsigned int i;
4043         int cpu;
4044 
4045         local_bh_disable();
4046         cpu = smp_processor_id();
4047         for (i = 0; i < dev->num_tx_queues; i++) {
4048                 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4049 
4050                 __netif_tx_lock(txq, cpu);
4051                 netif_tx_stop_queue(txq);
4052                 __netif_tx_unlock(txq);
4053         }
4054         local_bh_enable();
4055 }
4056 
4057 static inline void netif_addr_lock(struct net_device *dev)
4058 {
4059         spin_lock(&dev->addr_list_lock);
4060 }
4061 
4062 static inline void netif_addr_lock_bh(struct net_device *dev)
4063 {
4064         spin_lock_bh(&dev->addr_list_lock);
4065 }
4066 
4067 static inline void netif_addr_unlock(struct net_device *dev)
4068 {
4069         spin_unlock(&dev->addr_list_lock);
4070 }
4071 
4072 static inline void netif_addr_unlock_bh(struct net_device *dev)
4073 {
4074         spin_unlock_bh(&dev->addr_list_lock);
4075 }
4076 
4077 /*
4078  * dev_addrs walker. Should be used only for read access. Call with
4079  * rcu_read_lock held.
4080  */
4081 #define for_each_dev_addr(dev, ha) \
4082                 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4083 
4084 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4085 
4086 void ether_setup(struct net_device *dev);
4087 
4088 /* Support for loadable net-drivers */
4089 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4090                                     unsigned char name_assign_type,
4091                                     void (*setup)(struct net_device *),
4092                                     unsigned int txqs, unsigned int rxqs);
4093 int dev_get_valid_name(struct net *net, struct net_device *dev,
4094                        const char *name);
4095 
4096 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4097         alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4098 
4099 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4100         alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4101                          count)
4102 
4103 int register_netdev(struct net_device *dev);
4104 void unregister_netdev(struct net_device *dev);
4105 
4106 /* General hardware address lists handling functions */
4107 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4108                    struct netdev_hw_addr_list *from_list, int addr_len);
4109 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4110                       struct netdev_hw_addr_list *from_list, int addr_len);
4111 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4112                        struct net_device *dev,
4113                        int (*sync)(struct net_device *, const unsigned char *),
4114                        int (*unsync)(struct net_device *,
4115                                      const unsigned char *));
4116 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4117                            struct net_device *dev,
4118                            int (*sync)(struct net_device *,
4119                                        const unsigned char *, int),
4120                            int (*unsync)(struct net_device *,
4121                                          const unsigned char *, int));
4122 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4123                               struct net_device *dev,
4124                               int (*unsync)(struct net_device *,
4125                                             const unsigned char *, int));
4126 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4127                           struct net_device *dev,
4128                           int (*unsync)(struct net_device *,
4129                                         const unsigned char *));
4130 void __hw_addr_init(struct netdev_hw_addr_list *list);
4131 
4132 /* Functions used for device addresses handling */
4133 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4134                  unsigned char addr_type);
4135 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4136                  unsigned char addr_type);
4137 void dev_addr_flush(struct net_device *dev);
4138 int dev_addr_init(struct net_device *dev);
4139 
4140 /* Functions used for unicast addresses handling */
4141 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4142 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4143 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4144 int dev_uc_sync(struct net_device *to, struct net_device *from);
4145 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4146 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4147 void dev_uc_flush(struct net_device *dev);
4148 void dev_uc_init(struct net_device *dev);
4149 
4150 /**
4151  *  __dev_uc_sync - Synchonize device's unicast list
4152  *  @dev:  device to sync
4153  *  @sync: function to call if address should be added
4154  *  @unsync: function to call if address should be removed
4155  *
4156  *  Add newly added addresses to the interface, and release
4157  *  addresses that have been deleted.
4158  */
4159 static inline int __dev_uc_sync(struct net_device *dev,
4160                                 int (*sync)(struct net_device *,
4161                                             const unsigned char *),
4162                                 int (*unsync)(struct net_device *,
4163                                               const unsigned char *))
4164 {
4165         return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4166 }
4167 
4168 /**
4169  *  __dev_uc_unsync - Remove synchronized addresses from device
4170  *  @dev:  device to sync
4171  *  @unsync: function to call if address should be removed
4172  *
4173  *  Remove all addresses that were added to the device by dev_uc_sync().
4174  */
4175 static inline void __dev_uc_unsync(struct net_device *dev,
4176                                    int (*unsync)(struct net_device *,
4177                                                  const unsigned char *))
4178 {
4179         __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4180 }
4181 
4182 /* Functions used for multicast addresses handling */
4183 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4184 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4185 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4186 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4187 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4188 int dev_mc_sync(struct net_device *to, struct net_device *from);
4189 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4190 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4191 void dev_mc_flush(struct net_device *dev);
4192 void dev_mc_init(struct net_device *dev);
4193 
4194 /**
4195  *  __dev_mc_sync - Synchonize device's multicast list
4196  *  @dev:  device to sync
4197  *  @sync: function to call if address should be added
4198  *  @unsync: function to call if address should be removed
4199  *
4200  *  Add newly added addresses to the interface, and release
4201  *  addresses that have been deleted.
4202  */
4203 static inline int __dev_mc_sync(struct net_device *dev,
4204                                 int (*sync)(struct net_device *,
4205                                             const unsigned char *),
4206                                 int (*unsync)(struct net_device *,
4207                                               const unsigned char *))
4208 {
4209         return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4210 }
4211 
4212 /**
4213  *  __dev_mc_unsync - Remove synchronized addresses from device
4214  *  @dev:  device to sync
4215  *  @unsync: function to call if address should be removed
4216  *
4217  *  Remove all addresses that were added to the device by dev_mc_sync().
4218  */
4219 static inline void __dev_mc_unsync(struct net_device *dev,
4220                                    int (*unsync)(struct net_device *,
4221                                                  const unsigned char *))
4222 {
4223         __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4224 }
4225 
4226 /* Functions used for secondary unicast and multicast support */
4227 void dev_set_rx_mode(struct net_device *dev);
4228 void __dev_set_rx_mode(struct net_device *dev);
4229 int dev_set_promiscuity(struct net_device *dev, int inc);
4230 int dev_set_allmulti(struct net_device *dev, int inc);
4231 void netdev_state_change(struct net_device *dev);
4232 void netdev_notify_peers(struct net_device *dev);
4233 void netdev_features_change(struct net_device *dev);
4234 /* Load a device via the kmod */
4235 void dev_load(struct net *net, const char *name);
4236 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4237                                         struct rtnl_link_stats64 *storage);
4238 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4239                              const struct net_device_stats *netdev_stats);
4240 
4241 extern int              netdev_max_backlog;
4242 extern int              netdev_tstamp_prequeue;
4243 extern int              weight_p;
4244 extern int              dev_weight_rx_bias;
4245 extern int              dev_weight_tx_bias;
4246 extern int              dev_rx_weight;
4247 extern int              dev_tx_weight;
4248 extern int              gro_normal_batch;
4249 
4250 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4251 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4252                                                      struct list_head **iter);
4253 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4254                                                      struct list_head **iter);
4255 
4256 /* iterate through upper list, must be called under RCU read lock */
4257 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4258         for (iter = &(dev)->adj_list.upper, \
4259              updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4260              updev; \
4261              updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4262 
4263 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4264                                   int (*fn)(struct net_device *upper_dev,
4265                                             void *data),
4266                                   void *data);
4267 
4268 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4269                                   struct net_device *upper_dev);
4270 
4271 bool netdev_has_any_upper_dev(struct net_device *dev);
4272 
4273 void *netdev_lower_get_next_private(struct net_device *dev,
4274                                     struct list_head **iter);
4275 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4276                                         struct list_head **iter);
4277 
4278 #define netdev_for_each_lower_private(dev, priv, iter) \
4279         for (iter = (dev)->adj_list.lower.next, \
4280              priv = netdev_lower_get_next_private(dev, &(iter)); \
4281              priv; \
4282              priv = netdev_lower_get_next_private(dev, &(iter)))
4283 
4284 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4285         for (iter = &(dev)->adj_list.lower, \
4286              priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4287              priv; \
4288              priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4289 
4290 void *netdev_lower_get_next(struct net_device *dev,
4291                                 struct list_head **iter);
4292 
4293 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4294         for (iter = (dev)->adj_list.lower.next, \
4295              ldev = netdev_lower_get_next(dev, &(iter)); \
4296              ldev; \
4297              ldev = netdev_lower_get_next(dev, &(iter)))
4298 
4299 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4300                                              struct list_head **iter);
4301 int netdev_walk_all_lower_dev(struct net_device *dev,
4302                               int (*fn)(struct net_device *lower_dev,
4303                                         void *data),
4304                               void *data);
4305 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4306                                   int (*fn)(struct net_device *lower_dev,
4307                                             void *data),
4308                                   void *data);
4309 
4310 void *netdev_adjacent_get_private(struct list_head *adj_list);
4311 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4312 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4313 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4314 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4315                           struct netlink_ext_ack *extack);
4316 int netdev_master_upper_dev_link(struct net_device *dev,
4317                                  struct net_device *upper_dev,
4318                                  void *upper_priv, void *upper_info,
4319                                  struct netlink_ext_ack *extack);
4320 void netdev_upper_dev_unlink(struct net_device *dev,
4321                              struct net_device *upper_dev);
4322 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4323                                    struct net_device *new_dev,
4324                                    struct net_device *dev,
4325                                    struct netlink_ext_ack *extack);
4326 void netdev_adjacent_change_commit(struct net_device *old_dev,
4327                                    struct net_device *new_dev,
4328                                    struct net_device *dev);
4329 void netdev_adjacent_change_abort(struct net_device *old_dev,
4330                                   struct net_device *new_dev,
4331                                   struct net_device *dev);
4332 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4333 void *netdev_lower_dev_get_private(struct net_device *dev,
4334                                    struct net_device *lower_dev);
4335 void netdev_lower_state_changed(struct net_device *lower_dev,
4336                                 void *lower_state_info);
4337 
4338 /* RSS keys are 40 or 52 bytes long */
4339 #define NETDEV_RSS_KEY_LEN 52
4340 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4341 void netdev_rss_key_fill(void *buffer, size_t len);
4342 
4343 int skb_checksum_help(struct sk_buff *skb);
4344 int skb_crc32c_csum_help(struct sk_buff *skb);
4345 int skb_csum_hwoffload_help(struct sk_buff *skb,
4346                             const netdev_features_t features);
4347 
4348 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4349                                   netdev_features_t features, bool tx_path);
4350 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4351                                     netdev_features_t features);
4352 
4353 struct netdev_bonding_info {
4354         ifslave slave;
4355         ifbond  master;
4356 };
4357 
4358 struct netdev_notifier_bonding_info {
4359         struct netdev_notifier_info info; /* must be first */
4360         struct netdev_bonding_info  bonding_info;
4361 };
4362 
4363 void netdev_bonding_info_change(struct net_device *dev,
4364                                 struct netdev_bonding_info *bonding_info);
4365 
4366 static inline
4367 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4368 {
4369         return __skb_gso_segment(skb, features, true);
4370 }
4371 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4372 
4373 static inline bool can_checksum_protocol(netdev_features_t features,
4374                                          __be16 protocol)
4375 {
4376         if (protocol == htons(ETH_P_FCOE))
4377                 return !!(features & NETIF_F_FCOE_CRC);
4378 
4379         /* Assume this is an IP checksum (not SCTP CRC) */
4380 
4381         if (features & NETIF_F_HW_CSUM) {
4382                 /* Can checksum everything */
4383                 return true;
4384         }
4385 
4386         switch (protocol) {
4387         case htons(ETH_P_IP):
4388                 return !!(features & NETIF_F_IP_CSUM);
4389         case htons(ETH_P_IPV6):
4390                 return !!(features & NETIF_F_IPV6_CSUM);
4391         default:
4392                 return false;
4393         }
4394 }
4395 
4396 #ifdef CONFIG_BUG
4397 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4398 #else
4399 static inline void netdev_rx_csum_fault(struct net_device *dev,
4400                                         struct sk_buff *skb)
4401 {
4402 }
4403 #endif
4404 /* rx skb timestamps */
4405 void net_enable_timestamp(void);
4406 void net_disable_timestamp(void);
4407 
4408 #ifdef CONFIG_PROC_FS
4409 int __init dev_proc_init(void);
4410 #else
4411 #define dev_proc_init() 0
4412 #endif
4413 
4414 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4415                                               struct sk_buff *skb, struct net_device *dev,
4416                                               bool more)
4417 {
4418         __this_cpu_write(softnet_data.xmit.more, more);
4419         return ops->ndo_start_xmit(skb, dev);
4420 }
4421 
4422 static inline bool netdev_xmit_more(void)
4423 {
4424         return __this_cpu_read(softnet_data.xmit.more);
4425 }
4426 
4427 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4428                                             struct netdev_queue *txq, bool more)
4429 {
4430         const struct net_device_ops *ops = dev->netdev_ops;
4431         netdev_tx_t rc;
4432 
4433         rc = __netdev_start_xmit(ops, skb, dev, more);
4434         if (rc == NETDEV_TX_OK)
4435                 txq_trans_update(txq);
4436 
4437         return rc;
4438 }
4439 
4440 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4441                                 const void *ns);
4442 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4443                                  const void *ns);
4444 
4445 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4446 {
4447         return netdev_class_create_file_ns(class_attr, NULL);
4448 }
4449 
4450 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4451 {
4452         netdev_class_remove_file_ns(class_attr, NULL);
4453 }
4454 
4455 extern const struct kobj_ns_type_operations net_ns_type_operations;
4456 
4457 const char *netdev_drivername(const struct net_device *dev);
4458 
4459 void linkwatch_run_queue(void);
4460 
4461 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4462                                                           netdev_features_t f2)
4463 {
4464         if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4465                 if (f1 & NETIF_F_HW_CSUM)
4466                         f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4467                 else
4468                         f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4469         }
4470 
4471         return f1 & f2;
4472 }
4473 
4474 static inline netdev_features_t netdev_get_wanted_features(
4475         struct net_device *dev)
4476 {
4477         return (dev->features & ~dev->hw_features) | dev->wanted_features;
4478 }
4479 netdev_features_t netdev_increment_features(netdev_features_t all,
4480         netdev_features_t one, netdev_features_t mask);
4481 
4482 /* Allow TSO being used on stacked device :
4483  * Performing the GSO segmentation before last device
4484  * is a performance improvement.
4485  */
4486 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4487                                                         netdev_features_t mask)
4488 {
4489         return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4490 }
4491 
4492 int __netdev_update_features(struct net_device *dev);
4493 void netdev_update_features(struct net_device *dev);
4494 void netdev_change_features(struct net_device *dev);
4495 
4496 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4497                                         struct net_device *dev);
4498 
4499 netdev_features_t passthru_features_check(struct sk_buff *skb,
4500                                           struct net_device *dev,
4501                                           netdev_features_t features);
4502 netdev_features_t netif_skb_features(struct sk_buff *skb);
4503 
4504 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4505 {
4506         netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4507 
4508         /* check flags correspondence */
4509         BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4510         BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4511         BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4512         BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4513         BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4514         BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4515         BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4516         BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4517         BUILD_BUG_ON(SKB_GSO_IPXIP4  != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4518         BUILD_BUG_ON(SKB_GSO_IPXIP6  != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4519         BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4520         BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4521         BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4522         BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4523         BUILD_BUG_ON(SKB_GSO_SCTP    != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4524         BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4525         BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4526         BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4527 
4528         return (features & feature) == feature;
4529 }
4530 
4531 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4532 {
4533         return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4534                (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4535 }
4536 
4537 static inline bool netif_needs_gso(struct sk_buff *skb,
4538                                    netdev_features_t features)
4539 {
4540         return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4541                 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4542                          (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4543 }
4544 
4545 static inline void netif_set_gso_max_size(struct net_device *dev,
4546                                           unsigned int size)
4547 {
4548         dev->gso_max_size = size;
4549 }
4550 
4551 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4552                                         int pulled_hlen, u16 mac_offset,
4553                                         int mac_len)
4554 {
4555         skb->protocol = protocol;
4556         skb->encapsulation = 1;
4557         skb_push(skb, pulled_hlen);
4558         skb_reset_transport_header(skb);
4559         skb->mac_header = mac_offset;
4560         skb->network_header = skb->mac_header + mac_len;
4561         skb->mac_len = mac_len;
4562 }
4563 
4564 static inline bool netif_is_macsec(const struct net_device *dev)
4565 {
4566         return dev->priv_flags & IFF_MACSEC;
4567 }
4568 
4569 static inline bool netif_is_macvlan(const struct net_device *dev)
4570 {
4571         return dev->priv_flags & IFF_MACVLAN;
4572 }
4573 
4574 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4575 {
4576         return dev->priv_flags & IFF_MACVLAN_PORT;
4577 }
4578 
4579 static inline bool netif_is_bond_master(const struct net_device *dev)
4580 {
4581         return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4582 }
4583 
4584 static inline bool netif_is_bond_slave(const struct net_device *dev)
4585 {
4586         return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4587 }
4588 
4589 static inline bool netif_supports_nofcs(struct net_device *dev)
4590 {
4591         return dev->priv_flags & IFF_SUPP_NOFCS;
4592 }
4593 
4594 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4595 {
4596         return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4597 }
4598 
4599 static inline bool netif_is_l3_master(const struct net_device *dev)
4600 {
4601         return dev->priv_flags & IFF_L3MDEV_MASTER;
4602 }
4603 
4604 static inline bool netif_is_l3_slave(const struct net_device *dev)
4605 {
4606         return dev->priv_flags & IFF_L3MDEV_SLAVE;
4607 }
4608 
4609 static inline bool netif_is_bridge_master(const struct net_device *dev)
4610 {
4611         return dev->priv_flags & IFF_EBRIDGE;
4612 }
4613 
4614 static inline bool netif_is_bridge_port(const struct net_device *dev)
4615 {
4616         return dev->priv_flags & IFF_BRIDGE_PORT;
4617 }
4618 
4619 static inline bool netif_is_ovs_master(const struct net_device *dev)
4620 {
4621         return dev->priv_flags & IFF_OPENVSWITCH;
4622 }
4623 
4624 static inline bool netif_is_ovs_port(const struct net_device *dev)
4625 {
4626         return dev->priv_flags & IFF_OVS_DATAPATH;
4627 }
4628 
4629 static inline bool netif_is_team_master(const struct net_device *dev)
4630 {
4631         return dev->priv_flags & IFF_TEAM;
4632 }
4633 
4634 static inline bool netif_is_team_port(const struct net_device *dev)
4635 {
4636         return dev->priv_flags & IFF_TEAM_PORT;
4637 }
4638 
4639 static inline bool netif_is_lag_master(const struct net_device *dev)
4640 {
4641         return netif_is_bond_master(dev) || netif_is_team_master(dev);
4642 }
4643 
4644 static inline bool netif_is_lag_port(const struct net_device *dev)
4645 {
4646         return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4647 }
4648 
4649 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4650 {
4651         return dev->priv_flags & IFF_RXFH_CONFIGURED;
4652 }
4653 
4654 static inline bool netif_is_failover(const struct net_device *dev)
4655 {
4656         return dev->priv_flags & IFF_FAILOVER;
4657 }
4658 
4659 static inline bool netif_is_failover_slave(const struct net_device *dev)
4660 {
4661         return dev->priv_flags & IFF_FAILOVER_SLAVE;
4662 }
4663 
4664 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4665 static inline void netif_keep_dst(struct net_device *dev)
4666 {
4667         dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4668 }
4669 
4670 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4671 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4672 {
4673         /* TODO: reserve and use an additional IFF bit, if we get more users */
4674         return dev->priv_flags & IFF_MACSEC;
4675 }
4676 
4677 extern struct pernet_operations __net_initdata loopback_net_ops;
4678 
4679 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4680 
4681 /* netdev_printk helpers, similar to dev_printk */
4682 
4683 static inline const char *netdev_name(const struct net_device *dev)
4684 {
4685         if (!dev->name[0] || strchr(dev->name, '%'))
4686                 return "(unnamed net_device)";
4687         return dev->name;
4688 }
4689 
4690 static inline bool netdev_unregistering(const struct net_device *dev)
4691 {
4692         return dev->reg_state == NETREG_UNREGISTERING;
4693 }
4694 
4695 static inline const char *netdev_reg_state(const struct net_device *dev)
4696 {
4697         switch (dev->reg_state) {
4698         case NETREG_UNINITIALIZED: return " (uninitialized)";
4699         case NETREG_REGISTERED: return "";
4700         case NETREG_UNREGISTERING: return " (unregistering)";
4701         case NETREG_UNREGISTERED: return " (unregistered)";
4702         case NETREG_RELEASED: return " (released)";
4703         case NETREG_DUMMY: return " (dummy)";
4704         }
4705 
4706         WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4707         return " (unknown)";
4708 }
4709 
4710 __printf(3, 4) __cold
4711 void netdev_printk(const char *level, const struct net_device *dev,
4712                    const char *format, ...);
4713 __printf(2, 3) __cold
4714 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4715 __printf(2, 3) __cold
4716 void netdev_alert(const struct net_device *dev, const char *format, ...);
4717 __printf(2, 3) __cold
4718 void netdev_crit(const struct net_device *dev, const char *format, ...);
4719 __printf(2, 3) __cold
4720 void netdev_err(const struct net_device *dev, const char *format, ...);
4721 __printf(2, 3) __cold
4722 void netdev_warn(const struct net_device *dev, const char *format, ...);
4723 __printf(2, 3) __cold
4724 void netdev_notice(const struct net_device *dev, const char *format, ...);
4725 __printf(2, 3) __cold
4726 void netdev_info(const struct net_device *dev, const char *format, ...);
4727 
4728 #define netdev_level_once(level, dev, fmt, ...)                 \
4729 do {                                                            \
4730         static bool __print_once __read_mostly;                 \
4731                                                                 \
4732         if (!__print_once) {                                    \
4733                 __print_once = true;                            \
4734                 netdev_printk(level, dev, fmt, ##__VA_ARGS__);  \
4735         }                                                       \
4736 } while (0)
4737 
4738 #define netdev_emerg_once(dev, fmt, ...) \
4739         netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4740 #define netdev_alert_once(dev, fmt, ...) \
4741         netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4742 #define netdev_crit_once(dev, fmt, ...) \
4743         netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4744 #define netdev_err_once(dev, fmt, ...) \
4745         netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4746 #define netdev_warn_once(dev, fmt, ...) \
4747         netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4748 #define netdev_notice_once(dev, fmt, ...) \
4749         netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4750 #define netdev_info_once(dev, fmt, ...) \
4751         netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4752 
4753 #define MODULE_ALIAS_NETDEV(device) \
4754         MODULE_ALIAS("netdev-" device)
4755 
4756 #if defined(CONFIG_DYNAMIC_DEBUG)
4757 #define netdev_dbg(__dev, format, args...)                      \
4758 do {                                                            \
4759         dynamic_netdev_dbg(__dev, format, ##args);              \
4760 } while (0)
4761 #elif defined(DEBUG)
4762 #define netdev_dbg(__dev, format, args...)                      \
4763         netdev_printk(KERN_DEBUG, __dev, format, ##args)
4764 #else
4765 #define netdev_dbg(__dev, format, args...)                      \
4766 ({                                                              \
4767         if (0)                                                  \
4768                 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4769 })
4770 #endif
4771 
4772 #if defined(VERBOSE_DEBUG)
4773 #define netdev_vdbg     netdev_dbg
4774 #else
4775 
4776 #define netdev_vdbg(dev, format, args...)                       \
4777 ({                                                              \
4778         if (0)                                                  \
4779                 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4780         0;                                                      \
4781 })
4782 #endif
4783 
4784 /*
4785  * netdev_WARN() acts like dev_printk(), but with the key difference
4786  * of using a WARN/WARN_ON to get the message out, including the
4787  * file/line information and a backtrace.
4788  */
4789 #define netdev_WARN(dev, format, args...)                       \
4790         WARN(1, "netdevice: %s%s: " format, netdev_name(dev),   \
4791              netdev_reg_state(dev), ##args)
4792 
4793 #define netdev_WARN_ONCE(dev, format, args...)                          \
4794         WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev),      \
4795                   netdev_reg_state(dev), ##args)
4796 
4797 /* netif printk helpers, similar to netdev_printk */
4798 
4799 #define netif_printk(priv, type, level, dev, fmt, args...)      \
4800 do {                                                            \
4801         if (netif_msg_##type(priv))                             \
4802                 netdev_printk(level, (dev), fmt, ##args);       \
4803 } while (0)
4804 
4805 #define netif_level(level, priv, type, dev, fmt, args...)       \
4806 do {                                                            \
4807         if (netif_msg_##type(priv))                             \
4808                 netdev_##level(dev, fmt, ##args);               \
4809 } while (0)
4810 
4811 #define netif_emerg(priv, type, dev, fmt, args...)              \
4812         netif_level(emerg, priv, type, dev, fmt, ##args)
4813 #define netif_alert(priv, type, dev, fmt, args...)              \
4814         netif_level(alert, priv, type, dev, fmt, ##args)
4815 #define netif_crit(priv, type, dev, fmt, args...)               \
4816         netif_level(crit, priv, type, dev, fmt, ##args)
4817 #define netif_err(priv, type, dev, fmt, args...)                \
4818         netif_level(err, priv, type, dev, fmt, ##args)
4819 #define netif_warn(priv, type, dev, fmt, args...)               \
4820         netif_level(warn, priv, type, dev, fmt, ##args)
4821 #define netif_notice(priv, type, dev, fmt, args...)             \
4822         netif_level(notice, priv, type, dev, fmt, ##args)
4823 #define netif_info(priv, type, dev, fmt, args...)               \
4824         netif_level(info, priv, type, dev, fmt, ##args)
4825 
4826 #if defined(CONFIG_DYNAMIC_DEBUG)
4827 #define netif_dbg(priv, type, netdev, format, args...)          \
4828 do {                                                            \
4829         if (netif_msg_##type(priv))                             \
4830                 dynamic_netdev_dbg(netdev, format, ##args);     \
4831 } while (0)
4832 #elif defined(DEBUG)
4833 #define netif_dbg(priv, type, dev, format, args...)             \
4834         netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4835 #else
4836 #define netif_dbg(priv, type, dev, format, args...)                     \
4837 ({                                                                      \
4838         if (0)                                                          \
4839                 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4840         0;                                                              \
4841 })
4842 #endif
4843 
4844 /* if @cond then downgrade to debug, else print at @level */
4845 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...)     \
4846         do {                                                              \
4847                 if (cond)                                                 \
4848                         netif_dbg(priv, type, netdev, fmt, ##args);       \
4849                 else                                                      \
4850                         netif_ ## level(priv, type, netdev, fmt, ##args); \
4851         } while (0)
4852 
4853 #if defined(VERBOSE_DEBUG)
4854 #define netif_vdbg      netif_dbg
4855 #else
4856 #define netif_vdbg(priv, type, dev, format, args...)            \
4857 ({                                                              \
4858         if (0)                                                  \
4859                 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4860         0;                                                      \
4861 })
4862 #endif
4863 
4864 /*
4865  *      The list of packet types we will receive (as opposed to discard)
4866  *      and the routines to invoke.
4867  *
4868  *      Why 16. Because with 16 the only overlap we get on a hash of the
4869  *      low nibble of the protocol value is RARP/SNAP/X.25.
4870  *
4871  *              0800    IP
4872  *              0001    802.3
4873  *              0002    AX.25
4874  *              0004    802.2
4875  *              8035    RARP
4876  *              0005    SNAP
4877  *              0805    X.25
4878  *              0806    ARP
4879  *              8137    IPX
4880  *              0009    Localtalk
4881  *              86DD    IPv6
4882  */
4883 #define PTYPE_HASH_SIZE (16)
4884 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4885 
4886 extern struct net_device *blackhole_netdev;
4887 
4888 #endif  /* _LINUX_NETDEVICE_H */

/* [<][>][^][v][top][bottom][index][help] */