linux/include/linux/netdevice.h
<<
>>
Prefs
   1/*
   2 * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3 *              operating system.  INET is implemented using the  BSD Socket
   4 *              interface as the means of communication with the user level.
   5 *
   6 *              Definitions for the Interfaces handler.
   7 *
   8 * Version:     @(#)dev.h       1.0.10  08/12/93
   9 *
  10 * Authors:     Ross Biro
  11 *              Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
  12 *              Corey Minyard <wf-rch!minyard@relay.EU.net>
  13 *              Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
  14 *              Alan Cox, <alan@lxorguk.ukuu.org.uk>
  15 *              Bjorn Ekwall. <bj0rn@blox.se>
  16 *              Pekka Riikonen <priikone@poseidon.pspt.fi>
  17 *
  18 *              This program is free software; you can redistribute it and/or
  19 *              modify it under the terms of the GNU General Public License
  20 *              as published by the Free Software Foundation; either version
  21 *              2 of the License, or (at your option) any later version.
  22 *
  23 *              Moved to /usr/include/linux for NET3
  24 */
  25#ifndef _LINUX_NETDEVICE_H
  26#define _LINUX_NETDEVICE_H
  27
  28#include <linux/timer.h>
  29#include <linux/bug.h>
  30#include <linux/delay.h>
  31#include <linux/atomic.h>
  32#include <linux/prefetch.h>
  33#include <asm/cache.h>
  34#include <asm/byteorder.h>
  35
  36#include <linux/percpu.h>
  37#include <linux/rculist.h>
  38#include <linux/dmaengine.h>
  39#include <linux/workqueue.h>
  40#include <linux/dynamic_queue_limits.h>
  41
  42#include <linux/ethtool.h>
  43#include <net/net_namespace.h>
  44#include <net/dsa.h>
  45#ifdef CONFIG_DCB
  46#include <net/dcbnl.h>
  47#endif
  48#include <net/netprio_cgroup.h>
  49
  50#include <linux/netdev_features.h>
  51#include <linux/neighbour.h>
  52#include <uapi/linux/netdevice.h>
  53#include <uapi/linux/if_bonding.h>
  54#include <uapi/linux/pkt_cls.h>
  55#include <linux/hashtable.h>
  56
  57struct netpoll_info;
  58struct device;
  59struct phy_device;
  60/* 802.11 specific */
  61struct wireless_dev;
  62/* 802.15.4 specific */
  63struct wpan_dev;
  64struct mpls_dev;
  65/* UDP Tunnel offloads */
  66struct udp_tunnel_info;
  67struct bpf_prog;
  68
  69void netdev_set_default_ethtool_ops(struct net_device *dev,
  70                                    const struct ethtool_ops *ops);
  71
  72/* Backlog congestion levels */
  73#define NET_RX_SUCCESS          0       /* keep 'em coming, baby */
  74#define NET_RX_DROP             1       /* packet dropped */
  75
  76/*
  77 * Transmit return codes: transmit return codes originate from three different
  78 * namespaces:
  79 *
  80 * - qdisc return codes
  81 * - driver transmit return codes
  82 * - errno values
  83 *
  84 * Drivers are allowed to return any one of those in their hard_start_xmit()
  85 * function. Real network devices commonly used with qdiscs should only return
  86 * the driver transmit return codes though - when qdiscs are used, the actual
  87 * transmission happens asynchronously, so the value is not propagated to
  88 * higher layers. Virtual network devices transmit synchronously; in this case
  89 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
  90 * others are propagated to higher layers.
  91 */
  92
  93/* qdisc ->enqueue() return codes. */
  94#define NET_XMIT_SUCCESS        0x00
  95#define NET_XMIT_DROP           0x01    /* skb dropped                  */
  96#define NET_XMIT_CN             0x02    /* congestion notification      */
  97#define NET_XMIT_MASK           0x0f    /* qdisc flags in net/sch_generic.h */
  98
  99/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
 100 * indicates that the device will soon be dropping packets, or already drops
 101 * some packets of the same priority; prompting us to send less aggressively. */
 102#define net_xmit_eval(e)        ((e) == NET_XMIT_CN ? 0 : (e))
 103#define net_xmit_errno(e)       ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
 104
 105/* Driver transmit return codes */
 106#define NETDEV_TX_MASK          0xf0
 107
 108enum netdev_tx {
 109        __NETDEV_TX_MIN  = INT_MIN,     /* make sure enum is signed */
 110        NETDEV_TX_OK     = 0x00,        /* driver took care of packet */
 111        NETDEV_TX_BUSY   = 0x10,        /* driver tx path was busy*/
 112};
 113typedef enum netdev_tx netdev_tx_t;
 114
 115/*
 116 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
 117 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
 118 */
 119static inline bool dev_xmit_complete(int rc)
 120{
 121        /*
 122         * Positive cases with an skb consumed by a driver:
 123         * - successful transmission (rc == NETDEV_TX_OK)
 124         * - error while transmitting (rc < 0)
 125         * - error while queueing to a different device (rc & NET_XMIT_MASK)
 126         */
 127        if (likely(rc < NET_XMIT_MASK))
 128                return true;
 129
 130        return false;
 131}
 132
 133/*
 134 *      Compute the worst-case header length according to the protocols
 135 *      used.
 136 */
 137
 138#if defined(CONFIG_HYPERV_NET)
 139# define LL_MAX_HEADER 128
 140#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
 141# if defined(CONFIG_MAC80211_MESH)
 142#  define LL_MAX_HEADER 128
 143# else
 144#  define LL_MAX_HEADER 96
 145# endif
 146#else
 147# define LL_MAX_HEADER 32
 148#endif
 149
 150#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
 151    !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
 152#define MAX_HEADER LL_MAX_HEADER
 153#else
 154#define MAX_HEADER (LL_MAX_HEADER + 48)
 155#endif
 156
 157/*
 158 *      Old network device statistics. Fields are native words
 159 *      (unsigned long) so they can be read and written atomically.
 160 */
 161
 162struct net_device_stats {
 163        unsigned long   rx_packets;
 164        unsigned long   tx_packets;
 165        unsigned long   rx_bytes;
 166        unsigned long   tx_bytes;
 167        unsigned long   rx_errors;
 168        unsigned long   tx_errors;
 169        unsigned long   rx_dropped;
 170        unsigned long   tx_dropped;
 171        unsigned long   multicast;
 172        unsigned long   collisions;
 173        unsigned long   rx_length_errors;
 174        unsigned long   rx_over_errors;
 175        unsigned long   rx_crc_errors;
 176        unsigned long   rx_frame_errors;
 177        unsigned long   rx_fifo_errors;
 178        unsigned long   rx_missed_errors;
 179        unsigned long   tx_aborted_errors;
 180        unsigned long   tx_carrier_errors;
 181        unsigned long   tx_fifo_errors;
 182        unsigned long   tx_heartbeat_errors;
 183        unsigned long   tx_window_errors;
 184        unsigned long   rx_compressed;
 185        unsigned long   tx_compressed;
 186};
 187
 188
 189#include <linux/cache.h>
 190#include <linux/skbuff.h>
 191
 192#ifdef CONFIG_RPS
 193#include <linux/static_key.h>
 194extern struct static_key rps_needed;
 195extern struct static_key rfs_needed;
 196#endif
 197
 198struct neighbour;
 199struct neigh_parms;
 200struct sk_buff;
 201
 202struct netdev_hw_addr {
 203        struct list_head        list;
 204        unsigned char           addr[MAX_ADDR_LEN];
 205        unsigned char           type;
 206#define NETDEV_HW_ADDR_T_LAN            1
 207#define NETDEV_HW_ADDR_T_SAN            2
 208#define NETDEV_HW_ADDR_T_SLAVE          3
 209#define NETDEV_HW_ADDR_T_UNICAST        4
 210#define NETDEV_HW_ADDR_T_MULTICAST      5
 211        bool                    global_use;
 212        int                     sync_cnt;
 213        int                     refcount;
 214        int                     synced;
 215        struct rcu_head         rcu_head;
 216};
 217
 218struct netdev_hw_addr_list {
 219        struct list_head        list;
 220        int                     count;
 221};
 222
 223#define netdev_hw_addr_list_count(l) ((l)->count)
 224#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
 225#define netdev_hw_addr_list_for_each(ha, l) \
 226        list_for_each_entry(ha, &(l)->list, list)
 227
 228#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
 229#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
 230#define netdev_for_each_uc_addr(ha, dev) \
 231        netdev_hw_addr_list_for_each(ha, &(dev)->uc)
 232
 233#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
 234#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
 235#define netdev_for_each_mc_addr(ha, dev) \
 236        netdev_hw_addr_list_for_each(ha, &(dev)->mc)
 237
 238struct hh_cache {
 239        u16             hh_len;
 240        u16             __pad;
 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
 265struct 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};
 276
 277/* These flag bits are private to the generic network queueing
 278 * layer; they may not be explicitly referenced by any other
 279 * code.
 280 */
 281
 282enum netdev_state_t {
 283        __LINK_STATE_START,
 284        __LINK_STATE_PRESENT,
 285        __LINK_STATE_NOCARRIER,
 286        __LINK_STATE_LINKWATCH_PENDING,
 287        __LINK_STATE_DORMANT,
 288};
 289
 290
 291/*
 292 * This structure holds boot-time configured netdevice settings. They
 293 * are then used in the device probing.
 294 */
 295struct netdev_boot_setup {
 296        char name[IFNAMSIZ];
 297        struct ifmap map;
 298};
 299#define NETDEV_BOOT_SETUP_MAX 8
 300
 301int __init netdev_boot_setup(char *str);
 302
 303/*
 304 * Structure for NAPI scheduling similar to tasklet but with weighting
 305 */
 306struct napi_struct {
 307        /* The poll_list must only be managed by the entity which
 308         * changes the state of the NAPI_STATE_SCHED bit.  This means
 309         * whoever atomically sets that bit can add this napi_struct
 310         * to the per-CPU poll_list, and whoever clears that bit
 311         * can remove from the list right before clearing the bit.
 312         */
 313        struct list_head        poll_list;
 314
 315        unsigned long           state;
 316        int                     weight;
 317        unsigned int            gro_count;
 318        int                     (*poll)(struct napi_struct *, int);
 319#ifdef CONFIG_NETPOLL
 320        int                     poll_owner;
 321#endif
 322        struct net_device       *dev;
 323        struct sk_buff          *gro_list;
 324        struct sk_buff          *skb;
 325        struct hrtimer          timer;
 326        struct list_head        dev_list;
 327        struct hlist_node       napi_hash_node;
 328        unsigned int            napi_id;
 329};
 330
 331enum {
 332        NAPI_STATE_SCHED,       /* Poll is scheduled */
 333        NAPI_STATE_MISSED,      /* reschedule a napi */
 334        NAPI_STATE_DISABLE,     /* Disable pending */
 335        NAPI_STATE_NPSVC,       /* Netpoll - don't dequeue from poll_list */
 336        NAPI_STATE_HASHED,      /* In NAPI hash (busy polling possible) */
 337        NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
 338        NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
 339};
 340
 341enum {
 342        NAPIF_STATE_SCHED        = BIT(NAPI_STATE_SCHED),
 343        NAPIF_STATE_MISSED       = BIT(NAPI_STATE_MISSED),
 344        NAPIF_STATE_DISABLE      = BIT(NAPI_STATE_DISABLE),
 345        NAPIF_STATE_NPSVC        = BIT(NAPI_STATE_NPSVC),
 346        NAPIF_STATE_HASHED       = BIT(NAPI_STATE_HASHED),
 347        NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
 348        NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
 349};
 350
 351enum gro_result {
 352        GRO_MERGED,
 353        GRO_MERGED_FREE,
 354        GRO_HELD,
 355        GRO_NORMAL,
 356        GRO_DROP,
 357        GRO_CONSUMED,
 358};
 359typedef enum gro_result gro_result_t;
 360
 361/*
 362 * enum rx_handler_result - Possible return values for rx_handlers.
 363 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
 364 * further.
 365 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
 366 * case skb->dev was changed by rx_handler.
 367 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
 368 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
 369 *
 370 * rx_handlers are functions called from inside __netif_receive_skb(), to do
 371 * special processing of the skb, prior to delivery to protocol handlers.
 372 *
 373 * Currently, a net_device can only have a single rx_handler registered. Trying
 374 * to register a second rx_handler will return -EBUSY.
 375 *
 376 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
 377 * To unregister a rx_handler on a net_device, use
 378 * netdev_rx_handler_unregister().
 379 *
 380 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
 381 * do with the skb.
 382 *
 383 * If the rx_handler consumed the skb in some way, it should return
 384 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
 385 * the skb to be delivered in some other way.
 386 *
 387 * If the rx_handler changed skb->dev, to divert the skb to another
 388 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
 389 * new device will be called if it exists.
 390 *
 391 * If the rx_handler decides the skb should be ignored, it should return
 392 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
 393 * are registered on exact device (ptype->dev == skb->dev).
 394 *
 395 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
 396 * delivered, it should return RX_HANDLER_PASS.
 397 *
 398 * A device without a registered rx_handler will behave as if rx_handler
 399 * returned RX_HANDLER_PASS.
 400 */
 401
 402enum rx_handler_result {
 403        RX_HANDLER_CONSUMED,
 404        RX_HANDLER_ANOTHER,
 405        RX_HANDLER_EXACT,
 406        RX_HANDLER_PASS,
 407};
 408typedef enum rx_handler_result rx_handler_result_t;
 409typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
 410
 411void __napi_schedule(struct napi_struct *n);
 412void __napi_schedule_irqoff(struct napi_struct *n);
 413
 414static inline bool napi_disable_pending(struct napi_struct *n)
 415{
 416        return test_bit(NAPI_STATE_DISABLE, &n->state);
 417}
 418
 419bool napi_schedule_prep(struct napi_struct *n);
 420
 421/**
 422 *      napi_schedule - schedule NAPI poll
 423 *      @n: NAPI context
 424 *
 425 * Schedule NAPI poll routine to be called if it is not already
 426 * running.
 427 */
 428static inline void napi_schedule(struct napi_struct *n)
 429{
 430        if (napi_schedule_prep(n))
 431                __napi_schedule(n);
 432}
 433
 434/**
 435 *      napi_schedule_irqoff - schedule NAPI poll
 436 *      @n: NAPI context
 437 *
 438 * Variant of napi_schedule(), assuming hard irqs are masked.
 439 */
 440static inline void napi_schedule_irqoff(struct napi_struct *n)
 441{
 442        if (napi_schedule_prep(n))
 443                __napi_schedule_irqoff(n);
 444}
 445
 446/* Try to reschedule poll. Called by dev->poll() after napi_complete().  */
 447static inline bool napi_reschedule(struct napi_struct *napi)
 448{
 449        if (napi_schedule_prep(napi)) {
 450                __napi_schedule(napi);
 451                return true;
 452        }
 453        return false;
 454}
 455
 456bool napi_complete_done(struct napi_struct *n, int work_done);
 457/**
 458 *      napi_complete - NAPI processing complete
 459 *      @n: NAPI context
 460 *
 461 * Mark NAPI processing as complete.
 462 * Consider using napi_complete_done() instead.
 463 * Return false if device should avoid rearming interrupts.
 464 */
 465static inline bool napi_complete(struct napi_struct *n)
 466{
 467        return napi_complete_done(n, 0);
 468}
 469
 470/**
 471 *      napi_hash_del - remove a NAPI from global table
 472 *      @napi: NAPI context
 473 *
 474 * Warning: caller must observe RCU grace period
 475 * before freeing memory containing @napi, if
 476 * this function returns true.
 477 * Note: core networking stack automatically calls it
 478 * from netif_napi_del().
 479 * Drivers might want to call this helper to combine all
 480 * the needed RCU grace periods into a single one.
 481 */
 482bool napi_hash_del(struct napi_struct *napi);
 483
 484/**
 485 *      napi_disable - prevent NAPI from scheduling
 486 *      @n: NAPI context
 487 *
 488 * Stop NAPI from being scheduled on this context.
 489 * Waits till any outstanding processing completes.
 490 */
 491void napi_disable(struct napi_struct *n);
 492
 493/**
 494 *      napi_enable - enable NAPI scheduling
 495 *      @n: NAPI context
 496 *
 497 * Resume NAPI from being scheduled on this context.
 498 * Must be paired with napi_disable.
 499 */
 500static inline void napi_enable(struct napi_struct *n)
 501{
 502        BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
 503        smp_mb__before_atomic();
 504        clear_bit(NAPI_STATE_SCHED, &n->state);
 505        clear_bit(NAPI_STATE_NPSVC, &n->state);
 506}
 507
 508/**
 509 *      napi_synchronize - wait until NAPI is not running
 510 *      @n: NAPI context
 511 *
 512 * Wait until NAPI is done being scheduled on this context.
 513 * Waits till any outstanding processing completes but
 514 * does not disable future activations.
 515 */
 516static inline void napi_synchronize(const struct napi_struct *n)
 517{
 518        if (IS_ENABLED(CONFIG_SMP))
 519                while (test_bit(NAPI_STATE_SCHED, &n->state))
 520                        msleep(1);
 521        else
 522                barrier();
 523}
 524
 525enum netdev_queue_state_t {
 526        __QUEUE_STATE_DRV_XOFF,
 527        __QUEUE_STATE_STACK_XOFF,
 528        __QUEUE_STATE_FROZEN,
 529};
 530
 531#define QUEUE_STATE_DRV_XOFF    (1 << __QUEUE_STATE_DRV_XOFF)
 532#define QUEUE_STATE_STACK_XOFF  (1 << __QUEUE_STATE_STACK_XOFF)
 533#define QUEUE_STATE_FROZEN      (1 << __QUEUE_STATE_FROZEN)
 534
 535#define QUEUE_STATE_ANY_XOFF    (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
 536#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
 537                                        QUEUE_STATE_FROZEN)
 538#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
 539                                        QUEUE_STATE_FROZEN)
 540
 541/*
 542 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue.  The
 543 * netif_tx_* functions below are used to manipulate this flag.  The
 544 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
 545 * queue independently.  The netif_xmit_*stopped functions below are called
 546 * to check if the queue has been stopped by the driver or stack (either
 547 * of the XOFF bits are set in the state).  Drivers should not need to call
 548 * netif_xmit*stopped functions, they should only be using netif_tx_*.
 549 */
 550
 551struct netdev_queue {
 552/*
 553 * read-mostly part
 554 */
 555        struct net_device       *dev;
 556        struct Qdisc __rcu      *qdisc;
 557        struct Qdisc            *qdisc_sleeping;
 558#ifdef CONFIG_SYSFS
 559        struct kobject          kobj;
 560#endif
 561#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
 562        int                     numa_node;
 563#endif
 564        unsigned long           tx_maxrate;
 565        /*
 566         * Number of TX timeouts for this queue
 567         * (/sys/class/net/DEV/Q/trans_timeout)
 568         */
 569        unsigned long           trans_timeout;
 570/*
 571 * write-mostly part
 572 */
 573        spinlock_t              _xmit_lock ____cacheline_aligned_in_smp;
 574        int                     xmit_lock_owner;
 575        /*
 576         * Time (in jiffies) of last Tx
 577         */
 578        unsigned long           trans_start;
 579
 580        unsigned long           state;
 581
 582#ifdef CONFIG_BQL
 583        struct dql              dql;
 584#endif
 585} ____cacheline_aligned_in_smp;
 586
 587static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
 588{
 589#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
 590        return q->numa_node;
 591#else
 592        return NUMA_NO_NODE;
 593#endif
 594}
 595
 596static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
 597{
 598#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
 599        q->numa_node = node;
 600#endif
 601}
 602
 603#ifdef CONFIG_RPS
 604/*
 605 * This structure holds an RPS map which can be of variable length.  The
 606 * map is an array of CPUs.
 607 */
 608struct rps_map {
 609        unsigned int len;
 610        struct rcu_head rcu;
 611        u16 cpus[0];
 612};
 613#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
 614
 615/*
 616 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
 617 * tail pointer for that CPU's input queue at the time of last enqueue, and
 618 * a hardware filter index.
 619 */
 620struct rps_dev_flow {
 621        u16 cpu;
 622        u16 filter;
 623        unsigned int last_qtail;
 624};
 625#define RPS_NO_FILTER 0xffff
 626
 627/*
 628 * The rps_dev_flow_table structure contains a table of flow mappings.
 629 */
 630struct rps_dev_flow_table {
 631        unsigned int mask;
 632        struct rcu_head rcu;
 633        struct rps_dev_flow flows[0];
 634};
 635#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
 636    ((_num) * sizeof(struct rps_dev_flow)))
 637
 638/*
 639 * The rps_sock_flow_table contains mappings of flows to the last CPU
 640 * on which they were processed by the application (set in recvmsg).
 641 * Each entry is a 32bit value. Upper part is the high-order bits
 642 * of flow hash, lower part is CPU number.
 643 * rps_cpu_mask is used to partition the space, depending on number of
 644 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
 645 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
 646 * meaning we use 32-6=26 bits for the hash.
 647 */
 648struct rps_sock_flow_table {
 649        u32     mask;
 650
 651        u32     ents[0] ____cacheline_aligned_in_smp;
 652};
 653#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
 654
 655#define RPS_NO_CPU 0xffff
 656
 657extern u32 rps_cpu_mask;
 658extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
 659
 660static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
 661                                        u32 hash)
 662{
 663        if (table && hash) {
 664                unsigned int index = hash & table->mask;
 665                u32 val = hash & ~rps_cpu_mask;
 666
 667                /* We only give a hint, preemption can change CPU under us */
 668                val |= raw_smp_processor_id();
 669
 670                if (table->ents[index] != val)
 671                        table->ents[index] = val;
 672        }
 673}
 674
 675#ifdef CONFIG_RFS_ACCEL
 676bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
 677                         u16 filter_id);
 678#endif
 679#endif /* CONFIG_RPS */
 680
 681/* This structure contains an instance of an RX queue. */
 682struct netdev_rx_queue {
 683#ifdef CONFIG_RPS
 684        struct rps_map __rcu            *rps_map;
 685        struct rps_dev_flow_table __rcu *rps_flow_table;
 686#endif
 687        struct kobject                  kobj;
 688        struct net_device               *dev;
 689} ____cacheline_aligned_in_smp;
 690
 691/*
 692 * RX queue sysfs structures and functions.
 693 */
 694struct rx_queue_attribute {
 695        struct attribute attr;
 696        ssize_t (*show)(struct netdev_rx_queue *queue,
 697            struct rx_queue_attribute *attr, char *buf);
 698        ssize_t (*store)(struct netdev_rx_queue *queue,
 699            struct rx_queue_attribute *attr, const char *buf, size_t len);
 700};
 701
 702#ifdef CONFIG_XPS
 703/*
 704 * This structure holds an XPS map which can be of variable length.  The
 705 * map is an array of queues.
 706 */
 707struct xps_map {
 708        unsigned int len;
 709        unsigned int alloc_len;
 710        struct rcu_head rcu;
 711        u16 queues[0];
 712};
 713#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
 714#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
 715       - sizeof(struct xps_map)) / sizeof(u16))
 716
 717/*
 718 * This structure holds all XPS maps for device.  Maps are indexed by CPU.
 719 */
 720struct xps_dev_maps {
 721        struct rcu_head rcu;
 722        struct xps_map __rcu *cpu_map[0];
 723};
 724#define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) +          \
 725        (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
 726#endif /* CONFIG_XPS */
 727
 728#define TC_MAX_QUEUE    16
 729#define TC_BITMASK      15
 730/* HW offloaded queuing disciplines txq count and offset maps */
 731struct netdev_tc_txq {
 732        u16 count;
 733        u16 offset;
 734};
 735
 736#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
 737/*
 738 * This structure is to hold information about the device
 739 * configured to run FCoE protocol stack.
 740 */
 741struct netdev_fcoe_hbainfo {
 742        char    manufacturer[64];
 743        char    serial_number[64];
 744        char    hardware_version[64];
 745        char    driver_version[64];
 746        char    optionrom_version[64];
 747        char    firmware_version[64];
 748        char    model[256];
 749        char    model_description[256];
 750};
 751#endif
 752
 753#define MAX_PHYS_ITEM_ID_LEN 32
 754
 755/* This structure holds a unique identifier to identify some
 756 * physical item (port for example) used by a netdevice.
 757 */
 758struct netdev_phys_item_id {
 759        unsigned char id[MAX_PHYS_ITEM_ID_LEN];
 760        unsigned char id_len;
 761};
 762
 763static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
 764                                            struct netdev_phys_item_id *b)
 765{
 766        return a->id_len == b->id_len &&
 767               memcmp(a->id, b->id, a->id_len) == 0;
 768}
 769
 770typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
 771                                       struct sk_buff *skb);
 772
 773/* These structures hold the attributes of qdisc and classifiers
 774 * that are being passed to the netdevice through the setup_tc op.
 775 */
 776enum {
 777        TC_SETUP_MQPRIO,
 778        TC_SETUP_CLSU32,
 779        TC_SETUP_CLSFLOWER,
 780        TC_SETUP_MATCHALL,
 781        TC_SETUP_CLSBPF,
 782};
 783
 784struct tc_cls_u32_offload;
 785
 786struct tc_to_netdev {
 787        unsigned int type;
 788        union {
 789                u8 tc;
 790                struct tc_cls_u32_offload *cls_u32;
 791                struct tc_cls_flower_offload *cls_flower;
 792                struct tc_cls_matchall_offload *cls_mall;
 793                struct tc_cls_bpf_offload *cls_bpf;
 794        };
 795        bool egress_dev;
 796};
 797
 798/* These structures hold the attributes of xdp state that are being passed
 799 * to the netdevice through the xdp op.
 800 */
 801enum xdp_netdev_command {
 802        /* Set or clear a bpf program used in the earliest stages of packet
 803         * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
 804         * is responsible for calling bpf_prog_put on any old progs that are
 805         * stored. In case of error, the callee need not release the new prog
 806         * reference, but on success it takes ownership and must bpf_prog_put
 807         * when it is no longer used.
 808         */
 809        XDP_SETUP_PROG,
 810        /* Check if a bpf program is set on the device.  The callee should
 811         * return true if a program is currently attached and running.
 812         */
 813        XDP_QUERY_PROG,
 814};
 815
 816struct netdev_xdp {
 817        enum xdp_netdev_command command;
 818        union {
 819                /* XDP_SETUP_PROG */
 820                struct bpf_prog *prog;
 821                /* XDP_QUERY_PROG */
 822                bool prog_attached;
 823        };
 824};
 825
 826/*
 827 * This structure defines the management hooks for network devices.
 828 * The following hooks can be defined; unless noted otherwise, they are
 829 * optional and can be filled with a null pointer.
 830 *
 831 * int (*ndo_init)(struct net_device *dev);
 832 *     This function is called once when a network device is registered.
 833 *     The network device can use this for any late stage initialization
 834 *     or semantic validation. It can fail with an error code which will
 835 *     be propagated back to register_netdev.
 836 *
 837 * void (*ndo_uninit)(struct net_device *dev);
 838 *     This function is called when device is unregistered or when registration
 839 *     fails. It is not called if init fails.
 840 *
 841 * int (*ndo_open)(struct net_device *dev);
 842 *     This function is called when a network device transitions to the up
 843 *     state.
 844 *
 845 * int (*ndo_stop)(struct net_device *dev);
 846 *     This function is called when a network device transitions to the down
 847 *     state.
 848 *
 849 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
 850 *                               struct net_device *dev);
 851 *      Called when a packet needs to be transmitted.
 852 *      Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
 853 *      the queue before that can happen; it's for obsolete devices and weird
 854 *      corner cases, but the stack really does a non-trivial amount
 855 *      of useless work if you return NETDEV_TX_BUSY.
 856 *      Required; cannot be NULL.
 857 *
 858 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
 859 *                                         struct net_device *dev
 860 *                                         netdev_features_t features);
 861 *      Called by core transmit path to determine if device is capable of
 862 *      performing offload operations on a given packet. This is to give
 863 *      the device an opportunity to implement any restrictions that cannot
 864 *      be otherwise expressed by feature flags. The check is called with
 865 *      the set of features that the stack has calculated and it returns
 866 *      those the driver believes to be appropriate.
 867 *
 868 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
 869 *                         void *accel_priv, select_queue_fallback_t fallback);
 870 *      Called to decide which queue to use when device supports multiple
 871 *      transmit queues.
 872 *
 873 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
 874 *      This function is called to allow device receiver to make
 875 *      changes to configuration when multicast or promiscuous is enabled.
 876 *
 877 * void (*ndo_set_rx_mode)(struct net_device *dev);
 878 *      This function is called device changes address list filtering.
 879 *      If driver handles unicast address filtering, it should set
 880 *      IFF_UNICAST_FLT in its priv_flags.
 881 *
 882 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
 883 *      This function  is called when the Media Access Control address
 884 *      needs to be changed. If this interface is not defined, the
 885 *      MAC address can not be changed.
 886 *
 887 * int (*ndo_validate_addr)(struct net_device *dev);
 888 *      Test if Media Access Control address is valid for the device.
 889 *
 890 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
 891 *      Called when a user requests an ioctl which can't be handled by
 892 *      the generic interface code. If not defined ioctls return
 893 *      not supported error code.
 894 *
 895 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
 896 *      Used to set network devices bus interface parameters. This interface
 897 *      is retained for legacy reasons; new devices should use the bus
 898 *      interface (PCI) for low level management.
 899 *
 900 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
 901 *      Called when a user wants to change the Maximum Transfer Unit
 902 *      of a device. If not defined, any request to change MTU will
 903 *      will return an error.
 904 *
 905 * void (*ndo_tx_timeout)(struct net_device *dev);
 906 *      Callback used when the transmitter has not made any progress
 907 *      for dev->watchdog ticks.
 908 *
 909 * void (*ndo_get_stats64)(struct net_device *dev,
 910 *                         struct rtnl_link_stats64 *storage);
 911 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
 912 *      Called when a user wants to get the network device usage
 913 *      statistics. Drivers must do one of the following:
 914 *      1. Define @ndo_get_stats64 to fill in a zero-initialised
 915 *         rtnl_link_stats64 structure passed by the caller.
 916 *      2. Define @ndo_get_stats to update a net_device_stats structure
 917 *         (which should normally be dev->stats) and return a pointer to
 918 *         it. The structure may be changed asynchronously only if each
 919 *         field is written atomically.
 920 *      3. Update dev->stats asynchronously and atomically, and define
 921 *         neither operation.
 922 *
 923 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
 924 *      Return true if this device supports offload stats of this attr_id.
 925 *
 926 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
 927 *      void *attr_data)
 928 *      Get statistics for offload operations by attr_id. Write it into the
 929 *      attr_data pointer.
 930 *
 931 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
 932 *      If device supports VLAN filtering this function is called when a
 933 *      VLAN id is registered.
 934 *
 935 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
 936 *      If device supports VLAN filtering this function is called when a
 937 *      VLAN id is unregistered.
 938 *
 939 * void (*ndo_poll_controller)(struct net_device *dev);
 940 *
 941 *      SR-IOV management functions.
 942 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
 943 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
 944 *                        u8 qos, __be16 proto);
 945 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
 946 *                        int max_tx_rate);
 947 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
 948 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
 949 * int (*ndo_get_vf_config)(struct net_device *dev,
 950 *                          int vf, struct ifla_vf_info *ivf);
 951 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
 952 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
 953 *                        struct nlattr *port[]);
 954 *
 955 *      Enable or disable the VF ability to query its RSS Redirection Table and
 956 *      Hash Key. This is needed since on some devices VF share this information
 957 *      with PF and querying it may introduce a theoretical security risk.
 958 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
 959 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
 960 * int (*ndo_setup_tc)(struct net_device *dev, u32 handle,
 961 *                     __be16 protocol, struct tc_to_netdev *tc);
 962 *      Called to setup any 'tc' scheduler, classifier or action on @dev.
 963 *      This is always called from the stack with the rtnl lock held and netif
 964 *      tx queues stopped. This allows the netdevice to perform queue
 965 *      management safely.
 966 *
 967 *      Fiber Channel over Ethernet (FCoE) offload functions.
 968 * int (*ndo_fcoe_enable)(struct net_device *dev);
 969 *      Called when the FCoE protocol stack wants to start using LLD for FCoE
 970 *      so the underlying device can perform whatever needed configuration or
 971 *      initialization to support acceleration of FCoE traffic.
 972 *
 973 * int (*ndo_fcoe_disable)(struct net_device *dev);
 974 *      Called when the FCoE protocol stack wants to stop using LLD for FCoE
 975 *      so the underlying device can perform whatever needed clean-ups to
 976 *      stop supporting acceleration of FCoE traffic.
 977 *
 978 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
 979 *                           struct scatterlist *sgl, unsigned int sgc);
 980 *      Called when the FCoE Initiator wants to initialize an I/O that
 981 *      is a possible candidate for Direct Data Placement (DDP). The LLD can
 982 *      perform necessary setup and returns 1 to indicate the device is set up
 983 *      successfully to perform DDP on this I/O, otherwise this returns 0.
 984 *
 985 * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
 986 *      Called when the FCoE Initiator/Target is done with the DDPed I/O as
 987 *      indicated by the FC exchange id 'xid', so the underlying device can
 988 *      clean up and reuse resources for later DDP requests.
 989 *
 990 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
 991 *                            struct scatterlist *sgl, unsigned int sgc);
 992 *      Called when the FCoE Target wants to initialize an I/O that
 993 *      is a possible candidate for Direct Data Placement (DDP). The LLD can
 994 *      perform necessary setup and returns 1 to indicate the device is set up
 995 *      successfully to perform DDP on this I/O, otherwise this returns 0.
 996 *
 997 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
 998 *                             struct netdev_fcoe_hbainfo *hbainfo);
 999 *      Called when the FCoE Protocol stack wants information on the underlying
1000 *      device. This information is utilized by the FCoE protocol stack to
1001 *      register attributes with Fiber Channel management service as per the
1002 *      FC-GS Fabric Device Management Information(FDMI) specification.
1003 *
1004 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1005 *      Called when the underlying device wants to override default World Wide
1006 *      Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1007 *      World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1008 *      protocol stack to use.
1009 *
1010 *      RFS acceleration.
1011 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1012 *                          u16 rxq_index, u32 flow_id);
1013 *      Set hardware filter for RFS.  rxq_index is the target queue index;
1014 *      flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1015 *      Return the filter ID on success, or a negative error code.
1016 *
1017 *      Slave management functions (for bridge, bonding, etc).
1018 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1019 *      Called to make another netdev an underling.
1020 *
1021 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1022 *      Called to release previously enslaved netdev.
1023 *
1024 *      Feature/offload setting functions.
1025 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1026 *              netdev_features_t features);
1027 *      Adjusts the requested feature flags according to device-specific
1028 *      constraints, and returns the resulting flags. Must not modify
1029 *      the device state.
1030 *
1031 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1032 *      Called to update device configuration to new features. Passed
1033 *      feature set might be less than what was returned by ndo_fix_features()).
1034 *      Must return >0 or -errno if it changed dev->features itself.
1035 *
1036 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1037 *                    struct net_device *dev,
1038 *                    const unsigned char *addr, u16 vid, u16 flags)
1039 *      Adds an FDB entry to dev for addr.
1040 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1041 *                    struct net_device *dev,
1042 *                    const unsigned char *addr, u16 vid)
1043 *      Deletes the FDB entry from dev coresponding to addr.
1044 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1045 *                     struct net_device *dev, struct net_device *filter_dev,
1046 *                     int *idx)
1047 *      Used to add FDB entries to dump requests. Implementers should add
1048 *      entries to skb and update idx with the number of entries.
1049 *
1050 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1051 *                           u16 flags)
1052 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1053 *                           struct net_device *dev, u32 filter_mask,
1054 *                           int nlflags)
1055 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1056 *                           u16 flags);
1057 *
1058 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1059 *      Called to change device carrier. Soft-devices (like dummy, team, etc)
1060 *      which do not represent real hardware may define this to allow their
1061 *      userspace components to manage their virtual carrier state. Devices
1062 *      that determine carrier state from physical hardware properties (eg
1063 *      network cables) or protocol-dependent mechanisms (eg
1064 *      USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1065 *
1066 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1067 *                             struct netdev_phys_item_id *ppid);
1068 *      Called to get ID of physical port of this device. If driver does
1069 *      not implement this, it is assumed that the hw is not able to have
1070 *      multiple net devices on single physical port.
1071 *
1072 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1073 *                            struct udp_tunnel_info *ti);
1074 *      Called by UDP tunnel to notify a driver about the UDP port and socket
1075 *      address family that a UDP tunnel is listnening to. It is called only
1076 *      when a new port starts listening. The operation is protected by the
1077 *      RTNL.
1078 *
1079 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1080 *                            struct udp_tunnel_info *ti);
1081 *      Called by UDP tunnel to notify the driver about a UDP port and socket
1082 *      address family that the UDP tunnel is not listening to anymore. The
1083 *      operation is protected by the RTNL.
1084 *
1085 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1086 *                               struct net_device *dev)
1087 *      Called by upper layer devices to accelerate switching or other
1088 *      station functionality into hardware. 'pdev is the lowerdev
1089 *      to use for the offload and 'dev' is the net device that will
1090 *      back the offload. Returns a pointer to the private structure
1091 *      the upper layer will maintain.
1092 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1093 *      Called by upper layer device to delete the station created
1094 *      by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1095 *      the station and priv is the structure returned by the add
1096 *      operation.
1097 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1098 *                                    struct net_device *dev,
1099 *                                    void *priv);
1100 *      Callback to use for xmit over the accelerated station. This
1101 *      is used in place of ndo_start_xmit on accelerated net
1102 *      devices.
1103 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1104 *                           int queue_index, u32 maxrate);
1105 *      Called when a user wants to set a max-rate limitation of specific
1106 *      TX queue.
1107 * int (*ndo_get_iflink)(const struct net_device *dev);
1108 *      Called to get the iflink value of this device.
1109 * void (*ndo_change_proto_down)(struct net_device *dev,
1110 *                               bool proto_down);
1111 *      This function is used to pass protocol port error state information
1112 *      to the switch driver. The switch driver can react to the proto_down
1113 *      by doing a phys down on the associated switch port.
1114 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1115 *      This function is used to get egress tunnel information for given skb.
1116 *      This is useful for retrieving outer tunnel header parameters while
1117 *      sampling packet.
1118 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1119 *      This function is used to specify the headroom that the skb must
1120 *      consider when allocation skb during packet reception. Setting
1121 *      appropriate rx headroom value allows avoiding skb head copy on
1122 *      forward. Setting a negative value resets the rx headroom to the
1123 *      default value.
1124 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1125 *      This function is used to set or query state related to XDP on the
1126 *      netdevice. See definition of enum xdp_netdev_command for details.
1127 *
1128 */
1129struct net_device_ops {
1130        int                     (*ndo_init)(struct net_device *dev);
1131        void                    (*ndo_uninit)(struct net_device *dev);
1132        int                     (*ndo_open)(struct net_device *dev);
1133        int                     (*ndo_stop)(struct net_device *dev);
1134        netdev_tx_t             (*ndo_start_xmit)(struct sk_buff *skb,
1135                                                  struct net_device *dev);
1136        netdev_features_t       (*ndo_features_check)(struct sk_buff *skb,
1137                                                      struct net_device *dev,
1138                                                      netdev_features_t features);
1139        u16                     (*ndo_select_queue)(struct net_device *dev,
1140                                                    struct sk_buff *skb,
1141                                                    void *accel_priv,
1142                                                    select_queue_fallback_t fallback);
1143        void                    (*ndo_change_rx_flags)(struct net_device *dev,
1144                                                       int flags);
1145        void                    (*ndo_set_rx_mode)(struct net_device *dev);
1146        int                     (*ndo_set_mac_address)(struct net_device *dev,
1147                                                       void *addr);
1148        int                     (*ndo_validate_addr)(struct net_device *dev);
1149        int                     (*ndo_do_ioctl)(struct net_device *dev,
1150                                                struct ifreq *ifr, int cmd);
1151        int                     (*ndo_set_config)(struct net_device *dev,
1152                                                  struct ifmap *map);
1153        int                     (*ndo_change_mtu)(struct net_device *dev,
1154                                                  int new_mtu);
1155        int                     (*ndo_neigh_setup)(struct net_device *dev,
1156                                                   struct neigh_parms *);
1157        void                    (*ndo_tx_timeout) (struct net_device *dev);
1158
1159        void                    (*ndo_get_stats64)(struct net_device *dev,
1160                                                   struct rtnl_link_stats64 *storage);
1161        bool                    (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1162        int                     (*ndo_get_offload_stats)(int attr_id,
1163                                                         const struct net_device *dev,
1164                                                         void *attr_data);
1165        struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1166
1167        int                     (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1168                                                       __be16 proto, u16 vid);
1169        int                     (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1170                                                        __be16 proto, u16 vid);
1171#ifdef CONFIG_NET_POLL_CONTROLLER
1172        void                    (*ndo_poll_controller)(struct net_device *dev);
1173        int                     (*ndo_netpoll_setup)(struct net_device *dev,
1174                                                     struct netpoll_info *info);
1175        void                    (*ndo_netpoll_cleanup)(struct net_device *dev);
1176#endif
1177        int                     (*ndo_set_vf_mac)(struct net_device *dev,
1178                                                  int queue, u8 *mac);
1179        int                     (*ndo_set_vf_vlan)(struct net_device *dev,
1180                                                   int queue, u16 vlan,
1181                                                   u8 qos, __be16 proto);
1182        int                     (*ndo_set_vf_rate)(struct net_device *dev,
1183                                                   int vf, int min_tx_rate,
1184                                                   int max_tx_rate);
1185        int                     (*ndo_set_vf_spoofchk)(struct net_device *dev,
1186                                                       int vf, bool setting);
1187        int                     (*ndo_set_vf_trust)(struct net_device *dev,
1188                                                    int vf, bool setting);
1189        int                     (*ndo_get_vf_config)(struct net_device *dev,
1190                                                     int vf,
1191                                                     struct ifla_vf_info *ivf);
1192        int                     (*ndo_set_vf_link_state)(struct net_device *dev,
1193                                                         int vf, int link_state);
1194        int                     (*ndo_get_vf_stats)(struct net_device *dev,
1195                                                    int vf,
1196                                                    struct ifla_vf_stats
1197                                                    *vf_stats);
1198        int                     (*ndo_set_vf_port)(struct net_device *dev,
1199                                                   int vf,
1200                                                   struct nlattr *port[]);
1201        int                     (*ndo_get_vf_port)(struct net_device *dev,
1202                                                   int vf, struct sk_buff *skb);
1203        int                     (*ndo_set_vf_guid)(struct net_device *dev,
1204                                                   int vf, u64 guid,
1205                                                   int guid_type);
1206        int                     (*ndo_set_vf_rss_query_en)(
1207                                                   struct net_device *dev,
1208                                                   int vf, bool setting);
1209        int                     (*ndo_setup_tc)(struct net_device *dev,
1210                                                u32 handle,
1211                                                __be16 protocol,
1212                                                struct tc_to_netdev *tc);
1213#if IS_ENABLED(CONFIG_FCOE)
1214        int                     (*ndo_fcoe_enable)(struct net_device *dev);
1215        int                     (*ndo_fcoe_disable)(struct net_device *dev);
1216        int                     (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1217                                                      u16 xid,
1218                                                      struct scatterlist *sgl,
1219                                                      unsigned int sgc);
1220        int                     (*ndo_fcoe_ddp_done)(struct net_device *dev,
1221                                                     u16 xid);
1222        int                     (*ndo_fcoe_ddp_target)(struct net_device *dev,
1223                                                       u16 xid,
1224                                                       struct scatterlist *sgl,
1225                                                       unsigned int sgc);
1226        int                     (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1227                                                        struct netdev_fcoe_hbainfo *hbainfo);
1228#endif
1229
1230#if IS_ENABLED(CONFIG_LIBFCOE)
1231#define NETDEV_FCOE_WWNN 0
1232#define NETDEV_FCOE_WWPN 1
1233        int                     (*ndo_fcoe_get_wwn)(struct net_device *dev,
1234                                                    u64 *wwn, int type);
1235#endif
1236
1237#ifdef CONFIG_RFS_ACCEL
1238        int                     (*ndo_rx_flow_steer)(struct net_device *dev,
1239                                                     const struct sk_buff *skb,
1240                                                     u16 rxq_index,
1241                                                     u32 flow_id);
1242#endif
1243        int                     (*ndo_add_slave)(struct net_device *dev,
1244                                                 struct net_device *slave_dev);
1245        int                     (*ndo_del_slave)(struct net_device *dev,
1246                                                 struct net_device *slave_dev);
1247        netdev_features_t       (*ndo_fix_features)(struct net_device *dev,
1248                                                    netdev_features_t features);
1249        int                     (*ndo_set_features)(struct net_device *dev,
1250                                                    netdev_features_t features);
1251        int                     (*ndo_neigh_construct)(struct net_device *dev,
1252                                                       struct neighbour *n);
1253        void                    (*ndo_neigh_destroy)(struct net_device *dev,
1254                                                     struct neighbour *n);
1255
1256        int                     (*ndo_fdb_add)(struct ndmsg *ndm,
1257                                               struct nlattr *tb[],
1258                                               struct net_device *dev,
1259                                               const unsigned char *addr,
1260                                               u16 vid,
1261                                               u16 flags);
1262        int                     (*ndo_fdb_del)(struct ndmsg *ndm,
1263                                               struct nlattr *tb[],
1264                                               struct net_device *dev,
1265                                               const unsigned char *addr,
1266                                               u16 vid);
1267        int                     (*ndo_fdb_dump)(struct sk_buff *skb,
1268                                                struct netlink_callback *cb,
1269                                                struct net_device *dev,
1270                                                struct net_device *filter_dev,
1271                                                int *idx);
1272
1273        int                     (*ndo_bridge_setlink)(struct net_device *dev,
1274                                                      struct nlmsghdr *nlh,
1275                                                      u16 flags);
1276        int                     (*ndo_bridge_getlink)(struct sk_buff *skb,
1277                                                      u32 pid, u32 seq,
1278                                                      struct net_device *dev,
1279                                                      u32 filter_mask,
1280                                                      int nlflags);
1281        int                     (*ndo_bridge_dellink)(struct net_device *dev,
1282                                                      struct nlmsghdr *nlh,
1283                                                      u16 flags);
1284        int                     (*ndo_change_carrier)(struct net_device *dev,
1285                                                      bool new_carrier);
1286        int                     (*ndo_get_phys_port_id)(struct net_device *dev,
1287                                                        struct netdev_phys_item_id *ppid);
1288        int                     (*ndo_get_phys_port_name)(struct net_device *dev,
1289                                                          char *name, size_t len);
1290        void                    (*ndo_udp_tunnel_add)(struct net_device *dev,
1291                                                      struct udp_tunnel_info *ti);
1292        void                    (*ndo_udp_tunnel_del)(struct net_device *dev,
1293                                                      struct udp_tunnel_info *ti);
1294        void*                   (*ndo_dfwd_add_station)(struct net_device *pdev,
1295                                                        struct net_device *dev);
1296        void                    (*ndo_dfwd_del_station)(struct net_device *pdev,
1297                                                        void *priv);
1298
1299        netdev_tx_t             (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1300                                                        struct net_device *dev,
1301                                                        void *priv);
1302        int                     (*ndo_get_lock_subclass)(struct net_device *dev);
1303        int                     (*ndo_set_tx_maxrate)(struct net_device *dev,
1304                                                      int queue_index,
1305                                                      u32 maxrate);
1306        int                     (*ndo_get_iflink)(const struct net_device *dev);
1307        int                     (*ndo_change_proto_down)(struct net_device *dev,
1308                                                         bool proto_down);
1309        int                     (*ndo_fill_metadata_dst)(struct net_device *dev,
1310                                                       struct sk_buff *skb);
1311        void                    (*ndo_set_rx_headroom)(struct net_device *dev,
1312                                                       int needed_headroom);
1313        int                     (*ndo_xdp)(struct net_device *dev,
1314                                           struct netdev_xdp *xdp);
1315};
1316
1317/**
1318 * enum net_device_priv_flags - &struct net_device priv_flags
1319 *
1320 * These are the &struct net_device, they are only set internally
1321 * by drivers and used in the kernel. These flags are invisible to
1322 * userspace; this means that the order of these flags can change
1323 * during any kernel release.
1324 *
1325 * You should have a pretty good reason to be extending these flags.
1326 *
1327 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1328 * @IFF_EBRIDGE: Ethernet bridging device
1329 * @IFF_BONDING: bonding master or slave
1330 * @IFF_ISATAP: ISATAP interface (RFC4214)
1331 * @IFF_WAN_HDLC: WAN HDLC device
1332 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1333 *      release skb->dst
1334 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1335 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1336 * @IFF_MACVLAN_PORT: device used as macvlan port
1337 * @IFF_BRIDGE_PORT: device used as bridge port
1338 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1339 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1340 * @IFF_UNICAST_FLT: Supports unicast filtering
1341 * @IFF_TEAM_PORT: device used as team port
1342 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1343 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1344 *      change when it's running
1345 * @IFF_MACVLAN: Macvlan device
1346 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1347 *      underlying stacked devices
1348 * @IFF_IPVLAN_MASTER: IPvlan master device
1349 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1350 * @IFF_L3MDEV_MASTER: device is an L3 master device
1351 * @IFF_NO_QUEUE: device can run without qdisc attached
1352 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1353 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1354 * @IFF_TEAM: device is a team device
1355 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1356 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1357 *      entity (i.e. the master device for bridged veth)
1358 * @IFF_MACSEC: device is a MACsec device
1359 */
1360enum netdev_priv_flags {
1361        IFF_802_1Q_VLAN                 = 1<<0,
1362        IFF_EBRIDGE                     = 1<<1,
1363        IFF_BONDING                     = 1<<2,
1364        IFF_ISATAP                      = 1<<3,
1365        IFF_WAN_HDLC                    = 1<<4,
1366        IFF_XMIT_DST_RELEASE            = 1<<5,
1367        IFF_DONT_BRIDGE                 = 1<<6,
1368        IFF_DISABLE_NETPOLL             = 1<<7,
1369        IFF_MACVLAN_PORT                = 1<<8,
1370        IFF_BRIDGE_PORT                 = 1<<9,
1371        IFF_OVS_DATAPATH                = 1<<10,
1372        IFF_TX_SKB_SHARING              = 1<<11,
1373        IFF_UNICAST_FLT                 = 1<<12,
1374        IFF_TEAM_PORT                   = 1<<13,
1375        IFF_SUPP_NOFCS                  = 1<<14,
1376        IFF_LIVE_ADDR_CHANGE            = 1<<15,
1377        IFF_MACVLAN                     = 1<<16,
1378        IFF_XMIT_DST_RELEASE_PERM       = 1<<17,
1379        IFF_IPVLAN_MASTER               = 1<<18,
1380        IFF_IPVLAN_SLAVE                = 1<<19,
1381        IFF_L3MDEV_MASTER               = 1<<20,
1382        IFF_NO_QUEUE                    = 1<<21,
1383        IFF_OPENVSWITCH                 = 1<<22,
1384        IFF_L3MDEV_SLAVE                = 1<<23,
1385        IFF_TEAM                        = 1<<24,
1386        IFF_RXFH_CONFIGURED             = 1<<25,
1387        IFF_PHONY_HEADROOM              = 1<<26,
1388        IFF_MACSEC                      = 1<<27,
1389};
1390
1391#define IFF_802_1Q_VLAN                 IFF_802_1Q_VLAN
1392#define IFF_EBRIDGE                     IFF_EBRIDGE
1393#define IFF_BONDING                     IFF_BONDING
1394#define IFF_ISATAP                      IFF_ISATAP
1395#define IFF_WAN_HDLC                    IFF_WAN_HDLC
1396#define IFF_XMIT_DST_RELEASE            IFF_XMIT_DST_RELEASE
1397#define IFF_DONT_BRIDGE                 IFF_DONT_BRIDGE
1398#define IFF_DISABLE_NETPOLL             IFF_DISABLE_NETPOLL
1399#define IFF_MACVLAN_PORT                IFF_MACVLAN_PORT
1400#define IFF_BRIDGE_PORT                 IFF_BRIDGE_PORT
1401#define IFF_OVS_DATAPATH                IFF_OVS_DATAPATH
1402#define IFF_TX_SKB_SHARING              IFF_TX_SKB_SHARING
1403#define IFF_UNICAST_FLT                 IFF_UNICAST_FLT
1404#define IFF_TEAM_PORT                   IFF_TEAM_PORT
1405#define IFF_SUPP_NOFCS                  IFF_SUPP_NOFCS
1406#define IFF_LIVE_ADDR_CHANGE            IFF_LIVE_ADDR_CHANGE
1407#define IFF_MACVLAN                     IFF_MACVLAN
1408#define IFF_XMIT_DST_RELEASE_PERM       IFF_XMIT_DST_RELEASE_PERM
1409#define IFF_IPVLAN_MASTER               IFF_IPVLAN_MASTER
1410#define IFF_IPVLAN_SLAVE                IFF_IPVLAN_SLAVE
1411#define IFF_L3MDEV_MASTER               IFF_L3MDEV_MASTER
1412#define IFF_NO_QUEUE                    IFF_NO_QUEUE
1413#define IFF_OPENVSWITCH                 IFF_OPENVSWITCH
1414#define IFF_L3MDEV_SLAVE                IFF_L3MDEV_SLAVE
1415#define IFF_TEAM                        IFF_TEAM
1416#define IFF_RXFH_CONFIGURED             IFF_RXFH_CONFIGURED
1417#define IFF_MACSEC                      IFF_MACSEC
1418
1419/**
1420 *      struct net_device - The DEVICE structure.
1421 *              Actually, this whole structure is a big mistake.  It mixes I/O
1422 *              data with strictly "high-level" data, and it has to know about
1423 *              almost every data structure used in the INET module.
1424 *
1425 *      @name:  This is the first field of the "visible" part of this structure
1426 *              (i.e. as seen by users in the "Space.c" file).  It is the name
1427 *              of the interface.
1428 *
1429 *      @name_hlist:    Device name hash chain, please keep it close to name[]
1430 *      @ifalias:       SNMP alias
1431 *      @mem_end:       Shared memory end
1432 *      @mem_start:     Shared memory start
1433 *      @base_addr:     Device I/O address
1434 *      @irq:           Device IRQ number
1435 *
1436 *      @carrier_changes:       Stats to monitor carrier on<->off transitions
1437 *
1438 *      @state:         Generic network queuing layer state, see netdev_state_t
1439 *      @dev_list:      The global list of network devices
1440 *      @napi_list:     List entry used for polling NAPI devices
1441 *      @unreg_list:    List entry  when we are unregistering the
1442 *                      device; see the function unregister_netdev
1443 *      @close_list:    List entry used when we are closing the device
1444 *      @ptype_all:     Device-specific packet handlers for all protocols
1445 *      @ptype_specific: Device-specific, protocol-specific packet handlers
1446 *
1447 *      @adj_list:      Directly linked devices, like slaves for bonding
1448 *      @features:      Currently active device features
1449 *      @hw_features:   User-changeable features
1450 *
1451 *      @wanted_features:       User-requested features
1452 *      @vlan_features:         Mask of features inheritable by VLAN devices
1453 *
1454 *      @hw_enc_features:       Mask of features inherited by encapsulating devices
1455 *                              This field indicates what encapsulation
1456 *                              offloads the hardware is capable of doing,
1457 *                              and drivers will need to set them appropriately.
1458 *
1459 *      @mpls_features: Mask of features inheritable by MPLS
1460 *
1461 *      @ifindex:       interface index
1462 *      @group:         The group the device belongs to
1463 *
1464 *      @stats:         Statistics struct, which was left as a legacy, use
1465 *                      rtnl_link_stats64 instead
1466 *
1467 *      @rx_dropped:    Dropped packets by core network,
1468 *                      do not use this in drivers
1469 *      @tx_dropped:    Dropped packets by core network,
1470 *                      do not use this in drivers
1471 *      @rx_nohandler:  nohandler dropped packets by core network on
1472 *                      inactive devices, do not use this in drivers
1473 *
1474 *      @wireless_handlers:     List of functions to handle Wireless Extensions,
1475 *                              instead of ioctl,
1476 *                              see <net/iw_handler.h> for details.
1477 *      @wireless_data: Instance data managed by the core of wireless extensions
1478 *
1479 *      @netdev_ops:    Includes several pointers to callbacks,
1480 *                      if one wants to override the ndo_*() functions
1481 *      @ethtool_ops:   Management operations
1482 *      @ndisc_ops:     Includes callbacks for different IPv6 neighbour
1483 *                      discovery handling. Necessary for e.g. 6LoWPAN.
1484 *      @header_ops:    Includes callbacks for creating,parsing,caching,etc
1485 *                      of Layer 2 headers.
1486 *
1487 *      @flags:         Interface flags (a la BSD)
1488 *      @priv_flags:    Like 'flags' but invisible to userspace,
1489 *                      see if.h for the definitions
1490 *      @gflags:        Global flags ( kept as legacy )
1491 *      @padded:        How much padding added by alloc_netdev()
1492 *      @operstate:     RFC2863 operstate
1493 *      @link_mode:     Mapping policy to operstate
1494 *      @if_port:       Selectable AUI, TP, ...
1495 *      @dma:           DMA channel
1496 *      @mtu:           Interface MTU value
1497 *      @min_mtu:       Interface Minimum MTU value
1498 *      @max_mtu:       Interface Maximum MTU value
1499 *      @type:          Interface hardware type
1500 *      @hard_header_len: Maximum hardware header length.
1501 *      @min_header_len:  Minimum hardware header length
1502 *
1503 *      @needed_headroom: Extra headroom the hardware may need, but not in all
1504 *                        cases can this be guaranteed
1505 *      @needed_tailroom: Extra tailroom the hardware may need, but not in all
1506 *                        cases can this be guaranteed. Some cases also use
1507 *                        LL_MAX_HEADER instead to allocate the skb
1508 *
1509 *      interface address info:
1510 *
1511 *      @perm_addr:             Permanent hw address
1512 *      @addr_assign_type:      Hw address assignment type
1513 *      @addr_len:              Hardware address length
1514 *      @neigh_priv_len:        Used in neigh_alloc()
1515 *      @dev_id:                Used to differentiate devices that share
1516 *                              the same link layer address
1517 *      @dev_port:              Used to differentiate devices that share
1518 *                              the same function
1519 *      @addr_list_lock:        XXX: need comments on this one
1520 *      @uc_promisc:            Counter that indicates promiscuous mode
1521 *                              has been enabled due to the need to listen to
1522 *                              additional unicast addresses in a device that
1523 *                              does not implement ndo_set_rx_mode()
1524 *      @uc:                    unicast mac addresses
1525 *      @mc:                    multicast mac addresses
1526 *      @dev_addrs:             list of device hw addresses
1527 *      @queues_kset:           Group of all Kobjects in the Tx and RX queues
1528 *      @promiscuity:           Number of times the NIC is told to work in
1529 *                              promiscuous mode; if it becomes 0 the NIC will
1530 *                              exit promiscuous mode
1531 *      @allmulti:              Counter, enables or disables allmulticast mode
1532 *
1533 *      @vlan_info:     VLAN info
1534 *      @dsa_ptr:       dsa specific data
1535 *      @tipc_ptr:      TIPC specific data
1536 *      @atalk_ptr:     AppleTalk link
1537 *      @ip_ptr:        IPv4 specific data
1538 *      @dn_ptr:        DECnet specific data
1539 *      @ip6_ptr:       IPv6 specific data
1540 *      @ax25_ptr:      AX.25 specific data
1541 *      @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1542 *
1543 *      @dev_addr:      Hw address (before bcast,
1544 *                      because most packets are unicast)
1545 *
1546 *      @_rx:                   Array of RX queues
1547 *      @num_rx_queues:         Number of RX queues
1548 *                              allocated at register_netdev() time
1549 *      @real_num_rx_queues:    Number of RX queues currently active in device
1550 *
1551 *      @rx_handler:            handler for received packets
1552 *      @rx_handler_data:       XXX: need comments on this one
1553 *      @ingress_queue:         XXX: need comments on this one
1554 *      @broadcast:             hw bcast address
1555 *
1556 *      @rx_cpu_rmap:   CPU reverse-mapping for RX completion interrupts,
1557 *                      indexed by RX queue number. Assigned by driver.
1558 *                      This must only be set if the ndo_rx_flow_steer
1559 *                      operation is defined
1560 *      @index_hlist:           Device index hash chain
1561 *
1562 *      @_tx:                   Array of TX queues
1563 *      @num_tx_queues:         Number of TX queues allocated at alloc_netdev_mq() time
1564 *      @real_num_tx_queues:    Number of TX queues currently active in device
1565 *      @qdisc:                 Root qdisc from userspace point of view
1566 *      @tx_queue_len:          Max frames per queue allowed
1567 *      @tx_global_lock:        XXX: need comments on this one
1568 *
1569 *      @xps_maps:      XXX: need comments on this one
1570 *
1571 *      @watchdog_timeo:        Represents the timeout that is used by
1572 *                              the watchdog (see dev_watchdog())
1573 *      @watchdog_timer:        List of timers
1574 *
1575 *      @pcpu_refcnt:           Number of references to this device
1576 *      @todo_list:             Delayed register/unregister
1577 *      @link_watch_list:       XXX: need comments on this one
1578 *
1579 *      @reg_state:             Register/unregister state machine
1580 *      @dismantle:             Device is going to be freed
1581 *      @rtnl_link_state:       This enum represents the phases of creating
1582 *                              a new link
1583 *
1584 *      @destructor:            Called from unregister,
1585 *                              can be used to call free_netdev
1586 *      @npinfo:                XXX: need comments on this one
1587 *      @nd_net:                Network namespace this network device is inside
1588 *
1589 *      @ml_priv:       Mid-layer private
1590 *      @lstats:        Loopback statistics
1591 *      @tstats:        Tunnel statistics
1592 *      @dstats:        Dummy statistics
1593 *      @vstats:        Virtual ethernet statistics
1594 *
1595 *      @garp_port:     GARP
1596 *      @mrp_port:      MRP
1597 *
1598 *      @dev:           Class/net/name entry
1599 *      @sysfs_groups:  Space for optional device, statistics and wireless
1600 *                      sysfs groups
1601 *
1602 *      @sysfs_rx_queue_group:  Space for optional per-rx queue attributes
1603 *      @rtnl_link_ops: Rtnl_link_ops
1604 *
1605 *      @gso_max_size:  Maximum size of generic segmentation offload
1606 *      @gso_max_segs:  Maximum number of segments that can be passed to the
1607 *                      NIC for GSO
1608 *
1609 *      @dcbnl_ops:     Data Center Bridging netlink ops
1610 *      @num_tc:        Number of traffic classes in the net device
1611 *      @tc_to_txq:     XXX: need comments on this one
1612 *      @prio_tc_map:   XXX: need comments on this one
1613 *
1614 *      @fcoe_ddp_xid:  Max exchange id for FCoE LRO by ddp
1615 *
1616 *      @priomap:       XXX: need comments on this one
1617 *      @phydev:        Physical device may attach itself
1618 *                      for hardware timestamping
1619 *
1620 *      @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1621 *      @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1622 *
1623 *      @proto_down:    protocol port state information can be sent to the
1624 *                      switch driver and used to set the phys state of the
1625 *                      switch port.
1626 *
1627 *      FIXME: cleanup struct net_device such that network protocol info
1628 *      moves out.
1629 */
1630
1631struct net_device {
1632        char                    name[IFNAMSIZ];
1633        struct hlist_node       name_hlist;
1634        char                    *ifalias;
1635        /*
1636         *      I/O specific fields
1637         *      FIXME: Merge these and struct ifmap into one
1638         */
1639        unsigned long           mem_end;
1640        unsigned long           mem_start;
1641        unsigned long           base_addr;
1642        int                     irq;
1643
1644        atomic_t                carrier_changes;
1645
1646        /*
1647         *      Some hardware also needs these fields (state,dev_list,
1648         *      napi_list,unreg_list,close_list) but they are not
1649         *      part of the usual set specified in Space.c.
1650         */
1651
1652        unsigned long           state;
1653
1654        struct list_head        dev_list;
1655        struct list_head        napi_list;
1656        struct list_head        unreg_list;
1657        struct list_head        close_list;
1658        struct list_head        ptype_all;
1659        struct list_head        ptype_specific;
1660
1661        struct {
1662                struct list_head upper;
1663                struct list_head lower;
1664        } adj_list;
1665
1666        netdev_features_t       features;
1667        netdev_features_t       hw_features;
1668        netdev_features_t       wanted_features;
1669        netdev_features_t       vlan_features;
1670        netdev_features_t       hw_enc_features;
1671        netdev_features_t       mpls_features;
1672        netdev_features_t       gso_partial_features;
1673
1674        int                     ifindex;
1675        int                     group;
1676
1677        struct net_device_stats stats;
1678
1679        atomic_long_t           rx_dropped;
1680        atomic_long_t           tx_dropped;
1681        atomic_long_t           rx_nohandler;
1682
1683#ifdef CONFIG_WIRELESS_EXT
1684        const struct iw_handler_def *wireless_handlers;
1685        struct iw_public_data   *wireless_data;
1686#endif
1687        const struct net_device_ops *netdev_ops;
1688        const struct ethtool_ops *ethtool_ops;
1689#ifdef CONFIG_NET_SWITCHDEV
1690        const struct switchdev_ops *switchdev_ops;
1691#endif
1692#ifdef CONFIG_NET_L3_MASTER_DEV
1693        const struct l3mdev_ops *l3mdev_ops;
1694#endif
1695#if IS_ENABLED(CONFIG_IPV6)
1696        const struct ndisc_ops *ndisc_ops;
1697#endif
1698
1699        const struct header_ops *header_ops;
1700
1701        unsigned int            flags;
1702        unsigned int            priv_flags;
1703
1704        unsigned short          gflags;
1705        unsigned short          padded;
1706
1707        unsigned char           operstate;
1708        unsigned char           link_mode;
1709
1710        unsigned char           if_port;
1711        unsigned char           dma;
1712
1713        unsigned int            mtu;
1714        unsigned int            min_mtu;
1715        unsigned int            max_mtu;
1716        unsigned short          type;
1717        unsigned short          hard_header_len;
1718        unsigned short          min_header_len;
1719
1720        unsigned short          needed_headroom;
1721        unsigned short          needed_tailroom;
1722
1723        /* Interface address info. */
1724        unsigned char           perm_addr[MAX_ADDR_LEN];
1725        unsigned char           addr_assign_type;
1726        unsigned char           addr_len;
1727        unsigned short          neigh_priv_len;
1728        unsigned short          dev_id;
1729        unsigned short          dev_port;
1730        spinlock_t              addr_list_lock;
1731        unsigned char           name_assign_type;
1732        bool                    uc_promisc;
1733        struct netdev_hw_addr_list      uc;
1734        struct netdev_hw_addr_list      mc;
1735        struct netdev_hw_addr_list      dev_addrs;
1736
1737#ifdef CONFIG_SYSFS
1738        struct kset             *queues_kset;
1739#endif
1740        unsigned int            promiscuity;
1741        unsigned int            allmulti;
1742
1743
1744        /* Protocol-specific pointers */
1745
1746#if IS_ENABLED(CONFIG_VLAN_8021Q)
1747        struct vlan_info __rcu  *vlan_info;
1748#endif
1749#if IS_ENABLED(CONFIG_NET_DSA)
1750        struct dsa_switch_tree  *dsa_ptr;
1751#endif
1752#if IS_ENABLED(CONFIG_TIPC)
1753        struct tipc_bearer __rcu *tipc_ptr;
1754#endif
1755        void                    *atalk_ptr;
1756        struct in_device __rcu  *ip_ptr;
1757        struct dn_dev __rcu     *dn_ptr;
1758        struct inet6_dev __rcu  *ip6_ptr;
1759        void                    *ax25_ptr;
1760        struct wireless_dev     *ieee80211_ptr;
1761        struct wpan_dev         *ieee802154_ptr;
1762#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1763        struct mpls_dev __rcu   *mpls_ptr;
1764#endif
1765
1766/*
1767 * Cache lines mostly used on receive path (including eth_type_trans())
1768 */
1769        /* Interface address info used in eth_type_trans() */
1770        unsigned char           *dev_addr;
1771
1772#ifdef CONFIG_SYSFS
1773        struct netdev_rx_queue  *_rx;
1774
1775        unsigned int            num_rx_queues;
1776        unsigned int            real_num_rx_queues;
1777#endif
1778
1779        unsigned long           gro_flush_timeout;
1780        rx_handler_func_t __rcu *rx_handler;
1781        void __rcu              *rx_handler_data;
1782
1783#ifdef CONFIG_NET_CLS_ACT
1784        struct tcf_proto __rcu  *ingress_cl_list;
1785#endif
1786        struct netdev_queue __rcu *ingress_queue;
1787#ifdef CONFIG_NETFILTER_INGRESS
1788        struct nf_hook_entry __rcu *nf_hooks_ingress;
1789#endif
1790
1791        unsigned char           broadcast[MAX_ADDR_LEN];
1792#ifdef CONFIG_RFS_ACCEL
1793        struct cpu_rmap         *rx_cpu_rmap;
1794#endif
1795        struct hlist_node       index_hlist;
1796
1797/*
1798 * Cache lines mostly used on transmit path
1799 */
1800        struct netdev_queue     *_tx ____cacheline_aligned_in_smp;
1801        unsigned int            num_tx_queues;
1802        unsigned int            real_num_tx_queues;
1803        struct Qdisc            *qdisc;
1804#ifdef CONFIG_NET_SCHED
1805        DECLARE_HASHTABLE       (qdisc_hash, 4);
1806#endif
1807        unsigned long           tx_queue_len;
1808        spinlock_t              tx_global_lock;
1809        int                     watchdog_timeo;
1810
1811#ifdef CONFIG_XPS
1812        struct xps_dev_maps __rcu *xps_maps;
1813#endif
1814#ifdef CONFIG_NET_CLS_ACT
1815        struct tcf_proto __rcu  *egress_cl_list;
1816#endif
1817
1818        /* These may be needed for future network-power-down code. */
1819        struct timer_list       watchdog_timer;
1820
1821        int __percpu            *pcpu_refcnt;
1822        struct list_head        todo_list;
1823
1824        struct list_head        link_watch_list;
1825
1826        enum { NETREG_UNINITIALIZED=0,
1827               NETREG_REGISTERED,       /* completed register_netdevice */
1828               NETREG_UNREGISTERING,    /* called unregister_netdevice */
1829               NETREG_UNREGISTERED,     /* completed unregister todo */
1830               NETREG_RELEASED,         /* called free_netdev */
1831               NETREG_DUMMY,            /* dummy device for NAPI poll */
1832        } reg_state:8;
1833
1834        bool dismantle;
1835
1836        enum {
1837                RTNL_LINK_INITIALIZED,
1838                RTNL_LINK_INITIALIZING,
1839        } rtnl_link_state:16;
1840
1841        void (*destructor)(struct net_device *dev);
1842
1843#ifdef CONFIG_NETPOLL
1844        struct netpoll_info __rcu       *npinfo;
1845#endif
1846
1847        possible_net_t                  nd_net;
1848
1849        /* mid-layer private */
1850        union {
1851                void                                    *ml_priv;
1852                struct pcpu_lstats __percpu             *lstats;
1853                struct pcpu_sw_netstats __percpu        *tstats;
1854                struct pcpu_dstats __percpu             *dstats;
1855                struct pcpu_vstats __percpu             *vstats;
1856        };
1857
1858#if IS_ENABLED(CONFIG_GARP)
1859        struct garp_port __rcu  *garp_port;
1860#endif
1861#if IS_ENABLED(CONFIG_MRP)
1862        struct mrp_port __rcu   *mrp_port;
1863#endif
1864
1865        struct device           dev;
1866        const struct attribute_group *sysfs_groups[4];
1867        const struct attribute_group *sysfs_rx_queue_group;
1868
1869        const struct rtnl_link_ops *rtnl_link_ops;
1870
1871        /* for setting kernel sock attribute on TCP connection setup */
1872#define GSO_MAX_SIZE            65536
1873        unsigned int            gso_max_size;
1874#define GSO_MAX_SEGS            65535
1875        u16                     gso_max_segs;
1876
1877#ifdef CONFIG_DCB
1878        const struct dcbnl_rtnl_ops *dcbnl_ops;
1879#endif
1880        u8                      num_tc;
1881        struct netdev_tc_txq    tc_to_txq[TC_MAX_QUEUE];
1882        u8                      prio_tc_map[TC_BITMASK + 1];
1883
1884#if IS_ENABLED(CONFIG_FCOE)
1885        unsigned int            fcoe_ddp_xid;
1886#endif
1887#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1888        struct netprio_map __rcu *priomap;
1889#endif
1890        struct phy_device       *phydev;
1891        struct lock_class_key   *qdisc_tx_busylock;
1892        struct lock_class_key   *qdisc_running_key;
1893        bool                    proto_down;
1894};
1895#define to_net_dev(d) container_of(d, struct net_device, dev)
1896
1897#define NETDEV_ALIGN            32
1898
1899static inline
1900int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1901{
1902        return dev->prio_tc_map[prio & TC_BITMASK];
1903}
1904
1905static inline
1906int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1907{
1908        if (tc >= dev->num_tc)
1909                return -EINVAL;
1910
1911        dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1912        return 0;
1913}
1914
1915int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
1916void netdev_reset_tc(struct net_device *dev);
1917int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1918int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
1919
1920static inline
1921int netdev_get_num_tc(struct net_device *dev)
1922{
1923        return dev->num_tc;
1924}
1925
1926static inline
1927struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1928                                         unsigned int index)
1929{
1930        return &dev->_tx[index];
1931}
1932
1933static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1934                                                    const struct sk_buff *skb)
1935{
1936        return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1937}
1938
1939static inline void netdev_for_each_tx_queue(struct net_device *dev,
1940                                            void (*f)(struct net_device *,
1941                                                      struct netdev_queue *,
1942                                                      void *),
1943                                            void *arg)
1944{
1945        unsigned int i;
1946
1947        for (i = 0; i < dev->num_tx_queues; i++)
1948                f(dev, &dev->_tx[i], arg);
1949}
1950
1951#define netdev_lockdep_set_classes(dev)                         \
1952{                                                               \
1953        static struct lock_class_key qdisc_tx_busylock_key;     \
1954        static struct lock_class_key qdisc_running_key;         \
1955        static struct lock_class_key qdisc_xmit_lock_key;       \
1956        static struct lock_class_key dev_addr_list_lock_key;    \
1957        unsigned int i;                                         \
1958                                                                \
1959        (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key;      \
1960        (dev)->qdisc_running_key = &qdisc_running_key;          \
1961        lockdep_set_class(&(dev)->addr_list_lock,               \
1962                          &dev_addr_list_lock_key);             \
1963        for (i = 0; i < (dev)->num_tx_queues; i++)              \
1964                lockdep_set_class(&(dev)->_tx[i]._xmit_lock,    \
1965                                  &qdisc_xmit_lock_key);        \
1966}
1967
1968struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1969                                    struct sk_buff *skb,
1970                                    void *accel_priv);
1971
1972/* returns the headroom that the master device needs to take in account
1973 * when forwarding to this dev
1974 */
1975static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1976{
1977        return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1978}
1979
1980static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1981{
1982        if (dev->netdev_ops->ndo_set_rx_headroom)
1983                dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1984}
1985
1986/* set the device rx headroom to the dev's default */
1987static inline void netdev_reset_rx_headroom(struct net_device *dev)
1988{
1989        netdev_set_rx_headroom(dev, -1);
1990}
1991
1992/*
1993 * Net namespace inlines
1994 */
1995static inline
1996struct net *dev_net(const struct net_device *dev)
1997{
1998        return read_pnet(&dev->nd_net);
1999}
2000
2001static inline
2002void dev_net_set(struct net_device *dev, struct net *net)
2003{
2004        write_pnet(&dev->nd_net, net);
2005}
2006
2007static inline bool netdev_uses_dsa(struct net_device *dev)
2008{
2009#if IS_ENABLED(CONFIG_NET_DSA)
2010        if (dev->dsa_ptr != NULL)
2011                return dsa_uses_tagged_protocol(dev->dsa_ptr);
2012#endif
2013        return false;
2014}
2015
2016/**
2017 *      netdev_priv - access network device private data
2018 *      @dev: network device
2019 *
2020 * Get network device private data
2021 */
2022static inline void *netdev_priv(const struct net_device *dev)
2023{
2024        return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2025}
2026
2027/* Set the sysfs physical device reference for the network logical device
2028 * if set prior to registration will cause a symlink during initialization.
2029 */
2030#define SET_NETDEV_DEV(net, pdev)       ((net)->dev.parent = (pdev))
2031
2032/* Set the sysfs device type for the network logical device to allow
2033 * fine-grained identification of different network device types. For
2034 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2035 */
2036#define SET_NETDEV_DEVTYPE(net, devtype)        ((net)->dev.type = (devtype))
2037
2038/* Default NAPI poll() weight
2039 * Device drivers are strongly advised to not use bigger value
2040 */
2041#define NAPI_POLL_WEIGHT 64
2042
2043/**
2044 *      netif_napi_add - initialize a NAPI context
2045 *      @dev:  network device
2046 *      @napi: NAPI context
2047 *      @poll: polling function
2048 *      @weight: default weight
2049 *
2050 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2051 * *any* of the other NAPI-related functions.
2052 */
2053void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2054                    int (*poll)(struct napi_struct *, int), int weight);
2055
2056/**
2057 *      netif_tx_napi_add - initialize a NAPI context
2058 *      @dev:  network device
2059 *      @napi: NAPI context
2060 *      @poll: polling function
2061 *      @weight: default weight
2062 *
2063 * This variant of netif_napi_add() should be used from drivers using NAPI
2064 * to exclusively poll a TX queue.
2065 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2066 */
2067static inline void netif_tx_napi_add(struct net_device *dev,
2068                                     struct napi_struct *napi,
2069                                     int (*poll)(struct napi_struct *, int),
2070                                     int weight)
2071{
2072        set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2073        netif_napi_add(dev, napi, poll, weight);
2074}
2075
2076/**
2077 *  netif_napi_del - remove a NAPI context
2078 *  @napi: NAPI context
2079 *
2080 *  netif_napi_del() removes a NAPI context from the network device NAPI list
2081 */
2082void netif_napi_del(struct napi_struct *napi);
2083
2084struct napi_gro_cb {
2085        /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2086        void    *frag0;
2087
2088        /* Length of frag0. */
2089        unsigned int frag0_len;
2090
2091        /* This indicates where we are processing relative to skb->data. */
2092        int     data_offset;
2093
2094        /* This is non-zero if the packet cannot be merged with the new skb. */
2095        u16     flush;
2096
2097        /* Save the IP ID here and check when we get to the transport layer */
2098        u16     flush_id;
2099
2100        /* Number of segments aggregated. */
2101        u16     count;
2102
2103        /* Start offset for remote checksum offload */
2104        u16     gro_remcsum_start;
2105
2106        /* jiffies when first packet was created/queued */
2107        unsigned long age;
2108
2109        /* Used in ipv6_gro_receive() and foo-over-udp */
2110        u16     proto;
2111
2112        /* This is non-zero if the packet may be of the same flow. */
2113        u8      same_flow:1;
2114
2115        /* Used in tunnel GRO receive */
2116        u8      encap_mark:1;
2117
2118        /* GRO checksum is valid */
2119        u8      csum_valid:1;
2120
2121        /* Number of checksums via CHECKSUM_UNNECESSARY */
2122        u8      csum_cnt:3;
2123
2124        /* Free the skb? */
2125        u8      free:2;
2126#define NAPI_GRO_FREE             1
2127#define NAPI_GRO_FREE_STOLEN_HEAD 2
2128
2129        /* Used in foo-over-udp, set in udp[46]_gro_receive */
2130        u8      is_ipv6:1;
2131
2132        /* Used in GRE, set in fou/gue_gro_receive */
2133        u8      is_fou:1;
2134
2135        /* Used to determine if flush_id can be ignored */
2136        u8      is_atomic:1;
2137
2138        /* Number of gro_receive callbacks this packet already went through */
2139        u8 recursion_counter:4;
2140
2141        /* 1 bit hole */
2142
2143        /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2144        __wsum  csum;
2145
2146        /* used in skb_gro_receive() slow path */
2147        struct sk_buff *last;
2148};
2149
2150#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2151
2152#define GRO_RECURSION_LIMIT 15
2153static inline int gro_recursion_inc_test(struct sk_buff *skb)
2154{
2155        return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2156}
2157
2158typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2159static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2160                                                struct sk_buff **head,
2161                                                struct sk_buff *skb)
2162{
2163        if (unlikely(gro_recursion_inc_test(skb))) {
2164                NAPI_GRO_CB(skb)->flush |= 1;
2165                return NULL;
2166        }
2167
2168        return cb(head, skb);
2169}
2170
2171typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2172                                             struct sk_buff *);
2173static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2174                                                   struct sock *sk,
2175                                                   struct sk_buff **head,
2176                                                   struct sk_buff *skb)
2177{
2178        if (unlikely(gro_recursion_inc_test(skb))) {
2179                NAPI_GRO_CB(skb)->flush |= 1;
2180                return NULL;
2181        }
2182
2183        return cb(sk, head, skb);
2184}
2185
2186struct packet_type {
2187        __be16                  type;   /* This is really htons(ether_type). */
2188        struct net_device       *dev;   /* NULL is wildcarded here           */
2189        int                     (*func) (struct sk_buff *,
2190                                         struct net_device *,
2191                                         struct packet_type *,
2192                                         struct net_device *);
2193        bool                    (*id_match)(struct packet_type *ptype,
2194                                            struct sock *sk);
2195        void                    *af_packet_priv;
2196        struct list_head        list;
2197};
2198
2199struct offload_callbacks {
2200        struct sk_buff          *(*gso_segment)(struct sk_buff *skb,
2201                                                netdev_features_t features);
2202        struct sk_buff          **(*gro_receive)(struct sk_buff **head,
2203                                                 struct sk_buff *skb);
2204        int                     (*gro_complete)(struct sk_buff *skb, int nhoff);
2205};
2206
2207struct packet_offload {
2208        __be16                   type;  /* This is really htons(ether_type). */
2209        u16                      priority;
2210        struct offload_callbacks callbacks;
2211        struct list_head         list;
2212};
2213
2214/* often modified stats are per-CPU, other are shared (netdev->stats) */
2215struct pcpu_sw_netstats {
2216        u64     rx_packets;
2217        u64     rx_bytes;
2218        u64     tx_packets;
2219        u64     tx_bytes;
2220        struct u64_stats_sync   syncp;
2221};
2222
2223#define __netdev_alloc_pcpu_stats(type, gfp)                            \
2224({                                                                      \
2225        typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2226        if (pcpu_stats) {                                               \
2227                int __cpu;                                              \
2228                for_each_possible_cpu(__cpu) {                          \
2229                        typeof(type) *stat;                             \
2230                        stat = per_cpu_ptr(pcpu_stats, __cpu);          \
2231                        u64_stats_init(&stat->syncp);                   \
2232                }                                                       \
2233        }                                                               \
2234        pcpu_stats;                                                     \
2235})
2236
2237#define netdev_alloc_pcpu_stats(type)                                   \
2238        __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2239
2240enum netdev_lag_tx_type {
2241        NETDEV_LAG_TX_TYPE_UNKNOWN,
2242        NETDEV_LAG_TX_TYPE_RANDOM,
2243        NETDEV_LAG_TX_TYPE_BROADCAST,
2244        NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2245        NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2246        NETDEV_LAG_TX_TYPE_HASH,
2247};
2248
2249struct netdev_lag_upper_info {
2250        enum netdev_lag_tx_type tx_type;
2251};
2252
2253struct netdev_lag_lower_state_info {
2254        u8 link_up : 1,
2255           tx_enabled : 1;
2256};
2257
2258#include <linux/notifier.h>
2259
2260/* netdevice notifier chain. Please remember to update the rtnetlink
2261 * notification exclusion list in rtnetlink_event() when adding new
2262 * types.
2263 */
2264#define NETDEV_UP       0x0001  /* For now you can't veto a device up/down */
2265#define NETDEV_DOWN     0x0002
2266#define NETDEV_REBOOT   0x0003  /* Tell a protocol stack a network interface
2267                                   detected a hardware crash and restarted
2268                                   - we can use this eg to kick tcp sessions
2269                                   once done */
2270#define NETDEV_CHANGE   0x0004  /* Notify device state change */
2271#define NETDEV_REGISTER 0x0005
2272#define NETDEV_UNREGISTER       0x0006
2273#define NETDEV_CHANGEMTU        0x0007 /* notify after mtu change happened */
2274#define NETDEV_CHANGEADDR       0x0008
2275#define NETDEV_GOING_DOWN       0x0009
2276#define NETDEV_CHANGENAME       0x000A
2277#define NETDEV_FEAT_CHANGE      0x000B
2278#define NETDEV_BONDING_FAILOVER 0x000C
2279#define NETDEV_PRE_UP           0x000D
2280#define NETDEV_PRE_TYPE_CHANGE  0x000E
2281#define NETDEV_POST_TYPE_CHANGE 0x000F
2282#define NETDEV_POST_INIT        0x0010
2283#define NETDEV_UNREGISTER_FINAL 0x0011
2284#define NETDEV_RELEASE          0x0012
2285#define NETDEV_NOTIFY_PEERS     0x0013
2286#define NETDEV_JOIN             0x0014
2287#define NETDEV_CHANGEUPPER      0x0015
2288#define NETDEV_RESEND_IGMP      0x0016
2289#define NETDEV_PRECHANGEMTU     0x0017 /* notify before mtu change happened */
2290#define NETDEV_CHANGEINFODATA   0x0018
2291#define NETDEV_BONDING_INFO     0x0019
2292#define NETDEV_PRECHANGEUPPER   0x001A
2293#define NETDEV_CHANGELOWERSTATE 0x001B
2294#define NETDEV_UDP_TUNNEL_PUSH_INFO     0x001C
2295#define NETDEV_CHANGE_TX_QUEUE_LEN      0x001E
2296
2297int register_netdevice_notifier(struct notifier_block *nb);
2298int unregister_netdevice_notifier(struct notifier_block *nb);
2299
2300struct netdev_notifier_info {
2301        struct net_device *dev;
2302};
2303
2304struct netdev_notifier_change_info {
2305        struct netdev_notifier_info info; /* must be first */
2306        unsigned int flags_changed;
2307};
2308
2309struct netdev_notifier_changeupper_info {
2310        struct netdev_notifier_info info; /* must be first */
2311        struct net_device *upper_dev; /* new upper dev */
2312        bool master; /* is upper dev master */
2313        bool linking; /* is the notification for link or unlink */
2314        void *upper_info; /* upper dev info */
2315};
2316
2317struct netdev_notifier_changelowerstate_info {
2318        struct netdev_notifier_info info; /* must be first */
2319        void *lower_state_info; /* is lower dev state */
2320};
2321
2322static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2323                                             struct net_device *dev)
2324{
2325        info->dev = dev;
2326}
2327
2328static inline struct net_device *
2329netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2330{
2331        return info->dev;
2332}
2333
2334int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2335
2336
2337extern rwlock_t                         dev_base_lock;          /* Device list lock */
2338
2339#define for_each_netdev(net, d)         \
2340                list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2341#define for_each_netdev_reverse(net, d) \
2342                list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2343#define for_each_netdev_rcu(net, d)             \
2344                list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2345#define for_each_netdev_safe(net, d, n) \
2346                list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2347#define for_each_netdev_continue(net, d)                \
2348                list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2349#define for_each_netdev_continue_rcu(net, d)            \
2350        list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2351#define for_each_netdev_in_bond_rcu(bond, slave)        \
2352                for_each_netdev_rcu(&init_net, slave)   \
2353                        if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2354#define net_device_entry(lh)    list_entry(lh, struct net_device, dev_list)
2355
2356static inline struct net_device *next_net_device(struct net_device *dev)
2357{
2358        struct list_head *lh;
2359        struct net *net;
2360
2361        net = dev_net(dev);
2362        lh = dev->dev_list.next;
2363        return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2364}
2365
2366static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2367{
2368        struct list_head *lh;
2369        struct net *net;
2370
2371        net = dev_net(dev);
2372        lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2373        return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2374}
2375
2376static inline struct net_device *first_net_device(struct net *net)
2377{
2378        return list_empty(&net->dev_base_head) ? NULL :
2379                net_device_entry(net->dev_base_head.next);
2380}
2381
2382static inline struct net_device *first_net_device_rcu(struct net *net)
2383{
2384        struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2385
2386        return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2387}
2388
2389int netdev_boot_setup_check(struct net_device *dev);
2390unsigned long netdev_boot_base(const char *prefix, int unit);
2391struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2392                                       const char *hwaddr);
2393struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2394struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2395void dev_add_pack(struct packet_type *pt);
2396void dev_remove_pack(struct packet_type *pt);
2397void __dev_remove_pack(struct packet_type *pt);
2398void dev_add_offload(struct packet_offload *po);
2399void dev_remove_offload(struct packet_offload *po);
2400
2401int dev_get_iflink(const struct net_device *dev);
2402int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2403struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2404                                      unsigned short mask);
2405struct net_device *dev_get_by_name(struct net *net, const char *name);
2406struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2407struct net_device *__dev_get_by_name(struct net *net, const char *name);
2408int dev_alloc_name(struct net_device *dev, const char *name);
2409int dev_open(struct net_device *dev);
2410int dev_close(struct net_device *dev);
2411int dev_close_many(struct list_head *head, bool unlink);
2412void dev_disable_lro(struct net_device *dev);
2413int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2414int dev_queue_xmit(struct sk_buff *skb);
2415int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2416int register_netdevice(struct net_device *dev);
2417void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2418void unregister_netdevice_many(struct list_head *head);
2419static inline void unregister_netdevice(struct net_device *dev)
2420{
2421        unregister_netdevice_queue(dev, NULL);
2422}
2423
2424int netdev_refcnt_read(const struct net_device *dev);
2425void free_netdev(struct net_device *dev);
2426void netdev_freemem(struct net_device *dev);
2427void synchronize_net(void);
2428int init_dummy_netdev(struct net_device *dev);
2429
2430DECLARE_PER_CPU(int, xmit_recursion);
2431#define XMIT_RECURSION_LIMIT    10
2432
2433static inline int dev_recursion_level(void)
2434{
2435        return this_cpu_read(xmit_recursion);
2436}
2437
2438struct net_device *dev_get_by_index(struct net *net, int ifindex);
2439struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2440struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2441int netdev_get_name(struct net *net, char *name, int ifindex);
2442int dev_restart(struct net_device *dev);
2443int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2444
2445static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2446{
2447        return NAPI_GRO_CB(skb)->data_offset;
2448}
2449
2450static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2451{
2452        return skb->len - NAPI_GRO_CB(skb)->data_offset;
2453}
2454
2455static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2456{
2457        NAPI_GRO_CB(skb)->data_offset += len;
2458}
2459
2460static inline void *skb_gro_header_fast(struct sk_buff *skb,
2461                                        unsigned int offset)
2462{
2463        return NAPI_GRO_CB(skb)->frag0 + offset;
2464}
2465
2466static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2467{
2468        return NAPI_GRO_CB(skb)->frag0_len < hlen;
2469}
2470
2471static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2472{
2473        NAPI_GRO_CB(skb)->frag0 = NULL;
2474        NAPI_GRO_CB(skb)->frag0_len = 0;
2475}
2476
2477static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2478                                        unsigned int offset)
2479{
2480        if (!pskb_may_pull(skb, hlen))
2481                return NULL;
2482
2483        skb_gro_frag0_invalidate(skb);
2484        return skb->data + offset;
2485}
2486
2487static inline void *skb_gro_network_header(struct sk_buff *skb)
2488{
2489        return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2490               skb_network_offset(skb);
2491}
2492
2493static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2494                                        const void *start, unsigned int len)
2495{
2496        if (NAPI_GRO_CB(skb)->csum_valid)
2497                NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2498                                                  csum_partial(start, len, 0));
2499}
2500
2501/* GRO checksum functions. These are logical equivalents of the normal
2502 * checksum functions (in skbuff.h) except that they operate on the GRO
2503 * offsets and fields in sk_buff.
2504 */
2505
2506__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2507
2508static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2509{
2510        return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2511}
2512
2513static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2514                                                      bool zero_okay,
2515                                                      __sum16 check)
2516{
2517        return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2518                skb_checksum_start_offset(skb) <
2519                 skb_gro_offset(skb)) &&
2520                !skb_at_gro_remcsum_start(skb) &&
2521                NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2522                (!zero_okay || check));
2523}
2524
2525static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2526                                                           __wsum psum)
2527{
2528        if (NAPI_GRO_CB(skb)->csum_valid &&
2529            !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2530                return 0;
2531
2532        NAPI_GRO_CB(skb)->csum = psum;
2533
2534        return __skb_gro_checksum_complete(skb);
2535}
2536
2537static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2538{
2539        if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2540                /* Consume a checksum from CHECKSUM_UNNECESSARY */
2541                NAPI_GRO_CB(skb)->csum_cnt--;
2542        } else {
2543                /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2544                 * verified a new top level checksum or an encapsulated one
2545                 * during GRO. This saves work if we fallback to normal path.
2546                 */
2547                __skb_incr_checksum_unnecessary(skb);
2548        }
2549}
2550
2551#define __skb_gro_checksum_validate(skb, proto, zero_okay, check,       \
2552                                    compute_pseudo)                     \
2553({                                                                      \
2554        __sum16 __ret = 0;                                              \
2555        if (__skb_gro_checksum_validate_needed(skb, zero_okay, check))  \
2556                __ret = __skb_gro_checksum_validate_complete(skb,       \
2557                                compute_pseudo(skb, proto));            \
2558        if (__ret)                                                      \
2559                __skb_mark_checksum_bad(skb);                           \
2560        else                                                            \
2561                skb_gro_incr_csum_unnecessary(skb);                     \
2562        __ret;                                                          \
2563})
2564
2565#define skb_gro_checksum_validate(skb, proto, compute_pseudo)           \
2566        __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2567
2568#define skb_gro_checksum_validate_zero_check(skb, proto, check,         \
2569                                             compute_pseudo)            \
2570        __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2571
2572#define skb_gro_checksum_simple_validate(skb)                           \
2573        __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2574
2575static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2576{
2577        return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2578                !NAPI_GRO_CB(skb)->csum_valid);
2579}
2580
2581static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2582                                              __sum16 check, __wsum pseudo)
2583{
2584        NAPI_GRO_CB(skb)->csum = ~pseudo;
2585        NAPI_GRO_CB(skb)->csum_valid = 1;
2586}
2587
2588#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2589do {                                                                    \
2590        if (__skb_gro_checksum_convert_check(skb))                      \
2591                __skb_gro_checksum_convert(skb, check,                  \
2592                                           compute_pseudo(skb, proto)); \
2593} while (0)
2594
2595struct gro_remcsum {
2596        int offset;
2597        __wsum delta;
2598};
2599
2600static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2601{
2602        grc->offset = 0;
2603        grc->delta = 0;
2604}
2605
2606static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2607                                            unsigned int off, size_t hdrlen,
2608                                            int start, int offset,
2609                                            struct gro_remcsum *grc,
2610                                            bool nopartial)
2611{
2612        __wsum delta;
2613        size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2614
2615        BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2616
2617        if (!nopartial) {
2618                NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2619                return ptr;
2620        }
2621
2622        ptr = skb_gro_header_fast(skb, off);
2623        if (skb_gro_header_hard(skb, off + plen)) {
2624                ptr = skb_gro_header_slow(skb, off + plen, off);
2625                if (!ptr)
2626                        return NULL;
2627        }
2628
2629        delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2630                               start, offset);
2631
2632        /* Adjust skb->csum since we changed the packet */
2633        NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2634
2635        grc->offset = off + hdrlen + offset;
2636        grc->delta = delta;
2637
2638        return ptr;
2639}
2640
2641static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2642                                           struct gro_remcsum *grc)
2643{
2644        void *ptr;
2645        size_t plen = grc->offset + sizeof(u16);
2646
2647        if (!grc->delta)
2648                return;
2649
2650        ptr = skb_gro_header_fast(skb, grc->offset);
2651        if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2652                ptr = skb_gro_header_slow(skb, plen, grc->offset);
2653                if (!ptr)
2654                        return;
2655        }
2656
2657        remcsum_unadjust((__sum16 *)ptr, grc->delta);
2658}
2659
2660#ifdef CONFIG_XFRM_OFFLOAD
2661static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2662{
2663        if (PTR_ERR(pp) != -EINPROGRESS)
2664                NAPI_GRO_CB(skb)->flush |= flush;
2665}
2666#else
2667static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2668{
2669        NAPI_GRO_CB(skb)->flush |= flush;
2670}
2671#endif
2672
2673static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2674                                  unsigned short type,
2675                                  const void *daddr, const void *saddr,
2676                                  unsigned int len)
2677{
2678        if (!dev->header_ops || !dev->header_ops->create)
2679                return 0;
2680
2681        return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2682}
2683
2684static inline int dev_parse_header(const struct sk_buff *skb,
2685                                   unsigned char *haddr)
2686{
2687        const struct net_device *dev = skb->dev;
2688
2689        if (!dev->header_ops || !dev->header_ops->parse)
2690                return 0;
2691        return dev->header_ops->parse(skb, haddr);
2692}
2693
2694/* ll_header must have at least hard_header_len allocated */
2695static inline bool dev_validate_header(const struct net_device *dev,
2696                                       char *ll_header, int len)
2697{
2698        if (likely(len >= dev->hard_header_len))
2699                return true;
2700        if (len < dev->min_header_len)
2701                return false;
2702
2703        if (capable(CAP_SYS_RAWIO)) {
2704                memset(ll_header + len, 0, dev->hard_header_len - len);
2705                return true;
2706        }
2707
2708        if (dev->header_ops && dev->header_ops->validate)
2709                return dev->header_ops->validate(ll_header, len);
2710
2711        return false;
2712}
2713
2714typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2715int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2716static inline int unregister_gifconf(unsigned int family)
2717{
2718        return register_gifconf(family, NULL);
2719}
2720
2721#ifdef CONFIG_NET_FLOW_LIMIT
2722#define FLOW_LIMIT_HISTORY      (1 << 7)  /* must be ^2 and !overflow buckets */
2723struct sd_flow_limit {
2724        u64                     count;
2725        unsigned int            num_buckets;
2726        unsigned int            history_head;
2727        u16                     history[FLOW_LIMIT_HISTORY];
2728        u8                      buckets[];
2729};
2730
2731extern int netdev_flow_limit_table_len;
2732#endif /* CONFIG_NET_FLOW_LIMIT */
2733
2734/*
2735 * Incoming packets are placed on per-CPU queues
2736 */
2737struct softnet_data {
2738        struct list_head        poll_list;
2739        struct sk_buff_head     process_queue;
2740
2741        /* stats */
2742        unsigned int            processed;
2743        unsigned int            time_squeeze;
2744        unsigned int            received_rps;
2745#ifdef CONFIG_RPS
2746        struct softnet_data     *rps_ipi_list;
2747#endif
2748#ifdef CONFIG_NET_FLOW_LIMIT
2749        struct sd_flow_limit __rcu *flow_limit;
2750#endif
2751        struct Qdisc            *output_queue;
2752        struct Qdisc            **output_queue_tailp;
2753        struct sk_buff          *completion_queue;
2754
2755#ifdef CONFIG_RPS
2756        /* input_queue_head should be written by cpu owning this struct,
2757         * and only read by other cpus. Worth using a cache line.
2758         */
2759        unsigned int            input_queue_head ____cacheline_aligned_in_smp;
2760
2761        /* Elements below can be accessed between CPUs for RPS/RFS */
2762        struct call_single_data csd ____cacheline_aligned_in_smp;
2763        struct softnet_data     *rps_ipi_next;
2764        unsigned int            cpu;
2765        unsigned int            input_queue_tail;
2766#endif
2767        unsigned int            dropped;
2768        struct sk_buff_head     input_pkt_queue;
2769        struct napi_struct      backlog;
2770
2771};
2772
2773static inline void input_queue_head_incr(struct softnet_data *sd)
2774{
2775#ifdef CONFIG_RPS
2776        sd->input_queue_head++;
2777#endif
2778}
2779
2780static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2781                                              unsigned int *qtail)
2782{
2783#ifdef CONFIG_RPS
2784        *qtail = ++sd->input_queue_tail;
2785#endif
2786}
2787
2788DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2789
2790void __netif_schedule(struct Qdisc *q);
2791void netif_schedule_queue(struct netdev_queue *txq);
2792
2793static inline void netif_tx_schedule_all(struct net_device *dev)
2794{
2795        unsigned int i;
2796
2797        for (i = 0; i < dev->num_tx_queues; i++)
2798                netif_schedule_queue(netdev_get_tx_queue(dev, i));
2799}
2800
2801static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2802{
2803        clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2804}
2805
2806/**
2807 *      netif_start_queue - allow transmit
2808 *      @dev: network device
2809 *
2810 *      Allow upper layers to call the device hard_start_xmit routine.
2811 */
2812static inline void netif_start_queue(struct net_device *dev)
2813{
2814        netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2815}
2816
2817static inline void netif_tx_start_all_queues(struct net_device *dev)
2818{
2819        unsigned int i;
2820
2821        for (i = 0; i < dev->num_tx_queues; i++) {
2822                struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2823                netif_tx_start_queue(txq);
2824        }
2825}
2826
2827void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2828
2829/**
2830 *      netif_wake_queue - restart transmit
2831 *      @dev: network device
2832 *
2833 *      Allow upper layers to call the device hard_start_xmit routine.
2834 *      Used for flow control when transmit resources are available.
2835 */
2836static inline void netif_wake_queue(struct net_device *dev)
2837{
2838        netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2839}
2840
2841static inline void netif_tx_wake_all_queues(struct net_device *dev)
2842{
2843        unsigned int i;
2844
2845        for (i = 0; i < dev->num_tx_queues; i++) {
2846                struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2847                netif_tx_wake_queue(txq);
2848        }
2849}
2850
2851static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2852{
2853        set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2854}
2855
2856/**
2857 *      netif_stop_queue - stop transmitted packets
2858 *      @dev: network device
2859 *
2860 *      Stop upper layers calling the device hard_start_xmit routine.
2861 *      Used for flow control when transmit resources are unavailable.
2862 */
2863static inline void netif_stop_queue(struct net_device *dev)
2864{
2865        netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2866}
2867
2868void netif_tx_stop_all_queues(struct net_device *dev);
2869
2870static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2871{
2872        return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2873}
2874
2875/**
2876 *      netif_queue_stopped - test if transmit queue is flowblocked
2877 *      @dev: network device
2878 *
2879 *      Test if transmit queue on device is currently unable to send.
2880 */
2881static inline bool netif_queue_stopped(const struct net_device *dev)
2882{
2883        return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2884}
2885
2886static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2887{
2888        return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2889}
2890
2891static inline bool
2892netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2893{
2894        return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2895}
2896
2897static inline bool
2898netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2899{
2900        return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2901}
2902
2903/**
2904 *      netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2905 *      @dev_queue: pointer to transmit queue
2906 *
2907 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2908 * to give appropriate hint to the CPU.
2909 */
2910static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2911{
2912#ifdef CONFIG_BQL
2913        prefetchw(&dev_queue->dql.num_queued);
2914#endif
2915}
2916
2917/**
2918 *      netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2919 *      @dev_queue: pointer to transmit queue
2920 *
2921 * BQL enabled drivers might use this helper in their TX completion path,
2922 * to give appropriate hint to the CPU.
2923 */
2924static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2925{
2926#ifdef CONFIG_BQL
2927        prefetchw(&dev_queue->dql.limit);
2928#endif
2929}
2930
2931static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2932                                        unsigned int bytes)
2933{
2934#ifdef CONFIG_BQL
2935        dql_queued(&dev_queue->dql, bytes);
2936
2937        if (likely(dql_avail(&dev_queue->dql) >= 0))
2938                return;
2939
2940        set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2941
2942        /*
2943         * The XOFF flag must be set before checking the dql_avail below,
2944         * because in netdev_tx_completed_queue we update the dql_completed
2945         * before checking the XOFF flag.
2946         */
2947        smp_mb();
2948
2949        /* check again in case another CPU has just made room avail */
2950        if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2951                clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2952#endif
2953}
2954
2955/**
2956 *      netdev_sent_queue - report the number of bytes queued to hardware
2957 *      @dev: network device
2958 *      @bytes: number of bytes queued to the hardware device queue
2959 *
2960 *      Report the number of bytes queued for sending/completion to the network
2961 *      device hardware queue. @bytes should be a good approximation and should
2962 *      exactly match netdev_completed_queue() @bytes
2963 */
2964static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2965{
2966        netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2967}
2968
2969static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2970                                             unsigned int pkts, unsigned int bytes)
2971{
2972#ifdef CONFIG_BQL
2973        if (unlikely(!bytes))
2974                return;
2975
2976        dql_completed(&dev_queue->dql, bytes);
2977
2978        /*
2979         * Without the memory barrier there is a small possiblity that
2980         * netdev_tx_sent_queue will miss the update and cause the queue to
2981         * be stopped forever
2982         */
2983        smp_mb();
2984
2985        if (dql_avail(&dev_queue->dql) < 0)
2986                return;
2987
2988        if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2989                netif_schedule_queue(dev_queue);
2990#endif
2991}
2992
2993/**
2994 *      netdev_completed_queue - report bytes and packets completed by device
2995 *      @dev: network device
2996 *      @pkts: actual number of packets sent over the medium
2997 *      @bytes: actual number of bytes sent over the medium
2998 *
2999 *      Report the number of bytes and packets transmitted by the network device
3000 *      hardware queue over the physical medium, @bytes must exactly match the
3001 *      @bytes amount passed to netdev_sent_queue()
3002 */
3003static inline void netdev_completed_queue(struct net_device *dev,
3004                                          unsigned int pkts, unsigned int bytes)
3005{
3006        netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3007}
3008
3009static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3010{
3011#ifdef CONFIG_BQL
3012        clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3013        dql_reset(&q->dql);
3014#endif
3015}
3016
3017/**
3018 *      netdev_reset_queue - reset the packets and bytes count of a network device
3019 *      @dev_queue: network device
3020 *
3021 *      Reset the bytes and packet count of a network device and clear the
3022 *      software flow control OFF bit for this network device
3023 */
3024static inline void netdev_reset_queue(struct net_device *dev_queue)
3025{
3026        netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3027}
3028
3029/**
3030 *      netdev_cap_txqueue - check if selected tx queue exceeds device queues
3031 *      @dev: network device
3032 *      @queue_index: given tx queue index
3033 *
3034 *      Returns 0 if given tx queue index >= number of device tx queues,
3035 *      otherwise returns the originally passed tx queue index.
3036 */
3037static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3038{
3039        if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3040                net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3041                                     dev->name, queue_index,
3042                                     dev->real_num_tx_queues);
3043                return 0;
3044        }
3045
3046        return queue_index;
3047}
3048
3049/**
3050 *      netif_running - test if up
3051 *      @dev: network device
3052 *
3053 *      Test if the device has been brought up.
3054 */
3055static inline bool netif_running(const struct net_device *dev)
3056{
3057        return test_bit(__LINK_STATE_START, &dev->state);
3058}
3059
3060/*
3061 * Routines to manage the subqueues on a device.  We only need start,
3062 * stop, and a check if it's stopped.  All other device management is
3063 * done at the overall netdevice level.
3064 * Also test the device if we're multiqueue.
3065 */
3066
3067/**
3068 *      netif_start_subqueue - allow sending packets on subqueue
3069 *      @dev: network device
3070 *      @queue_index: sub queue index
3071 *
3072 * Start individual transmit queue of a device with multiple transmit queues.
3073 */
3074static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3075{
3076        struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3077
3078        netif_tx_start_queue(txq);
3079}
3080
3081/**
3082 *      netif_stop_subqueue - stop sending packets on subqueue
3083 *      @dev: network device
3084 *      @queue_index: sub queue index
3085 *
3086 * Stop individual transmit queue of a device with multiple transmit queues.
3087 */
3088static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3089{
3090        struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3091        netif_tx_stop_queue(txq);
3092}
3093
3094/**
3095 *      netif_subqueue_stopped - test status of subqueue
3096 *      @dev: network device
3097 *      @queue_index: sub queue index
3098 *
3099 * Check individual transmit queue of a device with multiple transmit queues.
3100 */
3101static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3102                                            u16 queue_index)
3103{
3104        struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3105
3106        return netif_tx_queue_stopped(txq);
3107}
3108
3109static inline bool netif_subqueue_stopped(const struct net_device *dev,
3110                                          struct sk_buff *skb)
3111{
3112        return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3113}
3114
3115/**
3116 *      netif_wake_subqueue - allow sending packets on subqueue
3117 *      @dev: network device
3118 *      @queue_index: sub queue index
3119 *
3120 * Resume individual transmit queue of a device with multiple transmit queues.
3121 */
3122static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3123{
3124        struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3125
3126        netif_tx_wake_queue(txq);
3127}
3128
3129#ifdef CONFIG_XPS
3130int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3131                        u16 index);
3132#else
3133static inline int netif_set_xps_queue(struct net_device *dev,
3134                                      const struct cpumask *mask,
3135                                      u16 index)
3136{
3137        return 0;
3138}
3139#endif
3140
3141u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3142                  unsigned int num_tx_queues);
3143
3144/*
3145 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3146 * as a distribution range limit for the returned value.
3147 */
3148static inline u16 skb_tx_hash(const struct net_device *dev,
3149                              struct sk_buff *skb)
3150{
3151        return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3152}
3153
3154/**
3155 *      netif_is_multiqueue - test if device has multiple transmit queues
3156 *      @dev: network device
3157 *
3158 * Check if device has multiple transmit queues
3159 */
3160static inline bool netif_is_multiqueue(const struct net_device *dev)
3161{
3162        return dev->num_tx_queues > 1;
3163}
3164
3165int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3166
3167#ifdef CONFIG_SYSFS
3168int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3169#else
3170static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3171                                                unsigned int rxq)
3172{
3173        return 0;
3174}
3175#endif
3176
3177#ifdef CONFIG_SYSFS
3178static inline unsigned int get_netdev_rx_queue_index(
3179                struct netdev_rx_queue *queue)
3180{
3181        struct net_device *dev = queue->dev;
3182        int index = queue - dev->_rx;
3183
3184        BUG_ON(index >= dev->num_rx_queues);
3185        return index;
3186}
3187#endif
3188
3189#define DEFAULT_MAX_NUM_RSS_QUEUES      (8)
3190int netif_get_num_default_rss_queues(void);
3191
3192enum skb_free_reason {
3193        SKB_REASON_CONSUMED,
3194        SKB_REASON_DROPPED,
3195};
3196
3197void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3198void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3199
3200/*
3201 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3202 * interrupt context or with hardware interrupts being disabled.
3203 * (in_irq() || irqs_disabled())
3204 *
3205 * We provide four helpers that can be used in following contexts :
3206 *
3207 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3208 *  replacing kfree_skb(skb)
3209 *
3210 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3211 *  Typically used in place of consume_skb(skb) in TX completion path
3212 *
3213 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3214 *  replacing kfree_skb(skb)
3215 *
3216 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3217 *  and consumed a packet. Used in place of consume_skb(skb)
3218 */
3219static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3220{
3221        __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3222}
3223
3224static inline void dev_consume_skb_irq(struct sk_buff *skb)
3225{
3226        __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3227}
3228
3229static inline void dev_kfree_skb_any(struct sk_buff *skb)
3230{
3231        __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3232}
3233
3234static inline void dev_consume_skb_any(struct sk_buff *skb)
3235{
3236        __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3237}
3238
3239int netif_rx(struct sk_buff *skb);
3240int netif_rx_ni(struct sk_buff *skb);
3241int netif_receive_skb(struct sk_buff *skb);
3242gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3243void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3244struct sk_buff *napi_get_frags(struct napi_struct *napi);
3245gro_result_t napi_gro_frags(struct napi_struct *napi);
3246struct packet_offload *gro_find_receive_by_type(__be16 type);
3247struct packet_offload *gro_find_complete_by_type(__be16 type);
3248
3249static inline void napi_free_frags(struct napi_struct *napi)
3250{
3251        kfree_skb(napi->skb);
3252        napi->skb = NULL;
3253}
3254
3255bool netdev_is_rx_handler_busy(struct net_device *dev);
3256int netdev_rx_handler_register(struct net_device *dev,
3257                               rx_handler_func_t *rx_handler,
3258                               void *rx_handler_data);
3259void netdev_rx_handler_unregister(struct net_device *dev);
3260
3261bool dev_valid_name(const char *name);
3262int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3263int dev_ethtool(struct net *net, struct ifreq *);
3264unsigned int dev_get_flags(const struct net_device *);
3265int __dev_change_flags(struct net_device *, unsigned int flags);
3266int dev_change_flags(struct net_device *, unsigned int);
3267void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3268                        unsigned int gchanges);
3269int dev_change_name(struct net_device *, const char *);
3270int dev_set_alias(struct net_device *, const char *, size_t);
3271int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3272int dev_set_mtu(struct net_device *, int);
3273void dev_set_group(struct net_device *, int);
3274int dev_set_mac_address(struct net_device *, struct sockaddr *);
3275int dev_change_carrier(struct net_device *, bool new_carrier);
3276int dev_get_phys_port_id(struct net_device *dev,
3277                         struct netdev_phys_item_id *ppid);
3278int dev_get_phys_port_name(struct net_device *dev,
3279                           char *name, size_t len);
3280int dev_change_proto_down(struct net_device *dev, bool proto_down);
3281int dev_change_xdp_fd(struct net_device *dev, int fd, u32 flags);
3282struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3283struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3284                                    struct netdev_queue *txq, int *ret);
3285int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3286int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3287bool is_skb_forwardable(const struct net_device *dev,
3288                        const struct sk_buff *skb);
3289
3290static __always_inline int ____dev_forward_skb(struct net_device *dev,
3291                                               struct sk_buff *skb)
3292{
3293        if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3294            unlikely(!is_skb_forwardable(dev, skb))) {
3295                atomic_long_inc(&dev->rx_dropped);
3296                kfree_skb(skb);
3297                return NET_RX_DROP;
3298        }
3299
3300        skb_scrub_packet(skb, true);
3301        skb->priority = 0;
3302        return 0;
3303}
3304
3305void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3306
3307extern int              netdev_budget;
3308
3309/* Called by rtnetlink.c:rtnl_unlock() */
3310void netdev_run_todo(void);
3311
3312/**
3313 *      dev_put - release reference to device
3314 *      @dev: network device
3315 *
3316 * Release reference to device to allow it to be freed.
3317 */
3318static inline void dev_put(struct net_device *dev)
3319{
3320        this_cpu_dec(*dev->pcpu_refcnt);
3321}
3322
3323/**
3324 *      dev_hold - get reference to device
3325 *      @dev: network device
3326 *
3327 * Hold reference to device to keep it from being freed.
3328 */
3329static inline void dev_hold(struct net_device *dev)
3330{
3331        this_cpu_inc(*dev->pcpu_refcnt);
3332}
3333
3334/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3335 * and _off may be called from IRQ context, but it is caller
3336 * who is responsible for serialization of these calls.
3337 *
3338 * The name carrier is inappropriate, these functions should really be
3339 * called netif_lowerlayer_*() because they represent the state of any
3340 * kind of lower layer not just hardware media.
3341 */
3342
3343void linkwatch_init_dev(struct net_device *dev);
3344void linkwatch_fire_event(struct net_device *dev);
3345void linkwatch_forget_dev(struct net_device *dev);
3346
3347/**
3348 *      netif_carrier_ok - test if carrier present
3349 *      @dev: network device
3350 *
3351 * Check if carrier is present on device
3352 */
3353static inline bool netif_carrier_ok(const struct net_device *dev)
3354{
3355        return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3356}
3357
3358unsigned long dev_trans_start(struct net_device *dev);
3359
3360void __netdev_watchdog_up(struct net_device *dev);
3361
3362void netif_carrier_on(struct net_device *dev);
3363
3364void netif_carrier_off(struct net_device *dev);
3365
3366/**
3367 *      netif_dormant_on - mark device as dormant.
3368 *      @dev: network device
3369 *
3370 * Mark device as dormant (as per RFC2863).
3371 *
3372 * The dormant state indicates that the relevant interface is not
3373 * actually in a condition to pass packets (i.e., it is not 'up') but is
3374 * in a "pending" state, waiting for some external event.  For "on-
3375 * demand" interfaces, this new state identifies the situation where the
3376 * interface is waiting for events to place it in the up state.
3377 */
3378static inline void netif_dormant_on(struct net_device *dev)
3379{
3380        if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3381                linkwatch_fire_event(dev);
3382}
3383
3384/**
3385 *      netif_dormant_off - set device as not dormant.
3386 *      @dev: network device
3387 *
3388 * Device is not in dormant state.
3389 */
3390static inline void netif_dormant_off(struct net_device *dev)
3391{
3392        if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3393                linkwatch_fire_event(dev);
3394}
3395
3396/**
3397 *      netif_dormant - test if carrier present
3398 *      @dev: network device
3399 *
3400 * Check if carrier is present on device
3401 */
3402static inline bool netif_dormant(const struct net_device *dev)
3403{
3404        return test_bit(__LINK_STATE_DORMANT, &dev->state);
3405}
3406
3407
3408/**
3409 *      netif_oper_up - test if device is operational
3410 *      @dev: network device
3411 *
3412 * Check if carrier is operational
3413 */
3414static inline bool netif_oper_up(const struct net_device *dev)
3415{
3416        return (dev->operstate == IF_OPER_UP ||
3417                dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3418}
3419
3420/**
3421 *      netif_device_present - is device available or removed
3422 *      @dev: network device
3423 *
3424 * Check if device has not been removed from system.
3425 */
3426static inline bool netif_device_present(struct net_device *dev)
3427{
3428        return test_bit(__LINK_STATE_PRESENT, &dev->state);
3429}
3430
3431void netif_device_detach(struct net_device *dev);
3432
3433void netif_device_attach(struct net_device *dev);
3434
3435/*
3436 * Network interface message level settings
3437 */
3438
3439enum {
3440        NETIF_MSG_DRV           = 0x0001,
3441        NETIF_MSG_PROBE         = 0x0002,
3442        NETIF_MSG_LINK          = 0x0004,
3443        NETIF_MSG_TIMER         = 0x0008,
3444        NETIF_MSG_IFDOWN        = 0x0010,
3445        NETIF_MSG_IFUP          = 0x0020,
3446        NETIF_MSG_RX_ERR        = 0x0040,
3447        NETIF_MSG_TX_ERR        = 0x0080,
3448        NETIF_MSG_TX_QUEUED     = 0x0100,
3449        NETIF_MSG_INTR          = 0x0200,
3450        NETIF_MSG_TX_DONE       = 0x0400,
3451        NETIF_MSG_RX_STATUS     = 0x0800,
3452        NETIF_MSG_PKTDATA       = 0x1000,
3453        NETIF_MSG_HW            = 0x2000,
3454        NETIF_MSG_WOL           = 0x4000,
3455};
3456
3457#define netif_msg_drv(p)        ((p)->msg_enable & NETIF_MSG_DRV)
3458#define netif_msg_probe(p)      ((p)->msg_enable & NETIF_MSG_PROBE)
3459#define netif_msg_link(p)       ((p)->msg_enable & NETIF_MSG_LINK)
3460#define netif_msg_timer(p)      ((p)->msg_enable & NETIF_MSG_TIMER)
3461#define netif_msg_ifdown(p)     ((p)->msg_enable & NETIF_MSG_IFDOWN)
3462#define netif_msg_ifup(p)       ((p)->msg_enable & NETIF_MSG_IFUP)
3463#define netif_msg_rx_err(p)     ((p)->msg_enable & NETIF_MSG_RX_ERR)
3464#define netif_msg_tx_err(p)     ((p)->msg_enable & NETIF_MSG_TX_ERR)
3465#define netif_msg_tx_queued(p)  ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3466#define netif_msg_intr(p)       ((p)->msg_enable & NETIF_MSG_INTR)
3467#define netif_msg_tx_done(p)    ((p)->msg_enable & NETIF_MSG_TX_DONE)
3468#define netif_msg_rx_status(p)  ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3469#define netif_msg_pktdata(p)    ((p)->msg_enable & NETIF_MSG_PKTDATA)
3470#define netif_msg_hw(p)         ((p)->msg_enable & NETIF_MSG_HW)
3471#define netif_msg_wol(p)        ((p)->msg_enable & NETIF_MSG_WOL)
3472
3473static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3474{
3475        /* use default */
3476        if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3477                return default_msg_enable_bits;
3478        if (debug_value == 0)   /* no output */
3479                return 0;
3480        /* set low N bits */
3481        return (1 << debug_value) - 1;
3482}
3483
3484static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3485{
3486        spin_lock(&txq->_xmit_lock);
3487        txq->xmit_lock_owner = cpu;
3488}
3489
3490static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3491{
3492        __acquire(&txq->_xmit_lock);
3493        return true;
3494}
3495
3496static inline void __netif_tx_release(struct netdev_queue *txq)
3497{
3498        __release(&txq->_xmit_lock);
3499}
3500
3501static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3502{
3503        spin_lock_bh(&txq->_xmit_lock);
3504        txq->xmit_lock_owner = smp_processor_id();
3505}
3506
3507static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3508{
3509        bool ok = spin_trylock(&txq->_xmit_lock);
3510        if (likely(ok))
3511                txq->xmit_lock_owner = smp_processor_id();
3512        return ok;
3513}
3514
3515static inline void __netif_tx_unlock(struct netdev_queue *txq)
3516{
3517        txq->xmit_lock_owner = -1;
3518        spin_unlock(&txq->_xmit_lock);
3519}
3520
3521static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3522{
3523        txq->xmit_lock_owner = -1;
3524        spin_unlock_bh(&txq->_xmit_lock);
3525}
3526
3527static inline void txq_trans_update(struct netdev_queue *txq)
3528{
3529        if (txq->xmit_lock_owner != -1)
3530                txq->trans_start = jiffies;
3531}
3532
3533/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3534static inline void netif_trans_update(struct net_device *dev)
3535{
3536        struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3537
3538        if (txq->trans_start != jiffies)
3539                txq->trans_start = jiffies;
3540}
3541
3542/**
3543 *      netif_tx_lock - grab network device transmit lock
3544 *      @dev: network device
3545 *
3546 * Get network device transmit lock
3547 */
3548static inline void netif_tx_lock(struct net_device *dev)
3549{
3550        unsigned int i;
3551        int cpu;
3552
3553        spin_lock(&dev->tx_global_lock);
3554        cpu = smp_processor_id();
3555        for (i = 0; i < dev->num_tx_queues; i++) {
3556                struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3557
3558                /* We are the only thread of execution doing a
3559                 * freeze, but we have to grab the _xmit_lock in
3560                 * order to synchronize with threads which are in
3561                 * the ->hard_start_xmit() handler and already
3562                 * checked the frozen bit.
3563                 */
3564                __netif_tx_lock(txq, cpu);
3565                set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3566                __netif_tx_unlock(txq);
3567        }
3568}
3569
3570static inline void netif_tx_lock_bh(struct net_device *dev)
3571{
3572        local_bh_disable();
3573        netif_tx_lock(dev);
3574}
3575
3576static inline void netif_tx_unlock(struct net_device *dev)
3577{
3578        unsigned int i;
3579
3580        for (i = 0; i < dev->num_tx_queues; i++) {
3581                struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3582
3583                /* No need to grab the _xmit_lock here.  If the
3584                 * queue is not stopped for another reason, we
3585                 * force a schedule.
3586                 */
3587                clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3588                netif_schedule_queue(txq);
3589        }
3590        spin_unlock(&dev->tx_global_lock);
3591}
3592
3593static inline void netif_tx_unlock_bh(struct net_device *dev)
3594{
3595        netif_tx_unlock(dev);
3596        local_bh_enable();
3597}
3598
3599#define HARD_TX_LOCK(dev, txq, cpu) {                   \
3600        if ((dev->features & NETIF_F_LLTX) == 0) {      \
3601                __netif_tx_lock(txq, cpu);              \
3602        } else {                                        \
3603                __netif_tx_acquire(txq);                \
3604        }                                               \
3605}
3606
3607#define HARD_TX_TRYLOCK(dev, txq)                       \
3608        (((dev->features & NETIF_F_LLTX) == 0) ?        \
3609                __netif_tx_trylock(txq) :               \
3610                __netif_tx_acquire(txq))
3611
3612#define HARD_TX_UNLOCK(dev, txq) {                      \
3613        if ((dev->features & NETIF_F_LLTX) == 0) {      \
3614                __netif_tx_unlock(txq);                 \
3615        } else {                                        \
3616                __netif_tx_release(txq);                \
3617        }                                               \
3618}
3619
3620static inline void netif_tx_disable(struct net_device *dev)
3621{
3622        unsigned int i;
3623        int cpu;
3624
3625        local_bh_disable();
3626        cpu = smp_processor_id();
3627        for (i = 0; i < dev->num_tx_queues; i++) {
3628                struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3629
3630                __netif_tx_lock(txq, cpu);
3631                netif_tx_stop_queue(txq);
3632                __netif_tx_unlock(txq);
3633        }
3634        local_bh_enable();
3635}
3636
3637static inline void netif_addr_lock(struct net_device *dev)
3638{
3639        spin_lock(&dev->addr_list_lock);
3640}
3641
3642static inline void netif_addr_lock_nested(struct net_device *dev)
3643{
3644        int subclass = SINGLE_DEPTH_NESTING;
3645
3646        if (dev->netdev_ops->ndo_get_lock_subclass)
3647                subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3648
3649        spin_lock_nested(&dev->addr_list_lock, subclass);
3650}
3651
3652static inline void netif_addr_lock_bh(struct net_device *dev)
3653{
3654        spin_lock_bh(&dev->addr_list_lock);
3655}
3656
3657static inline void netif_addr_unlock(struct net_device *dev)
3658{
3659        spin_unlock(&dev->addr_list_lock);
3660}
3661
3662static inline void netif_addr_unlock_bh(struct net_device *dev)
3663{
3664        spin_unlock_bh(&dev->addr_list_lock);
3665}
3666
3667/*
3668 * dev_addrs walker. Should be used only for read access. Call with
3669 * rcu_read_lock held.
3670 */
3671#define for_each_dev_addr(dev, ha) \
3672                list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3673
3674/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3675
3676void ether_setup(struct net_device *dev);
3677
3678/* Support for loadable net-drivers */
3679struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3680                                    unsigned char name_assign_type,
3681                                    void (*setup)(struct net_device *),
3682                                    unsigned int txqs, unsigned int rxqs);
3683#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3684        alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3685
3686#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3687        alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3688                         count)
3689
3690int register_netdev(struct net_device *dev);
3691void unregister_netdev(struct net_device *dev);
3692
3693/* General hardware address lists handling functions */
3694int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3695                   struct netdev_hw_addr_list *from_list, int addr_len);
3696void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3697                      struct netdev_hw_addr_list *from_list, int addr_len);
3698int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3699                       struct net_device *dev,
3700                       int (*sync)(struct net_device *, const unsigned char *),
3701                       int (*unsync)(struct net_device *,
3702                                     const unsigned char *));
3703void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3704                          struct net_device *dev,
3705                          int (*unsync)(struct net_device *,
3706                                        const unsigned char *));
3707void __hw_addr_init(struct netdev_hw_addr_list *list);
3708
3709/* Functions used for device addresses handling */
3710int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3711                 unsigned char addr_type);
3712int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3713                 unsigned char addr_type);
3714void dev_addr_flush(struct net_device *dev);
3715int dev_addr_init(struct net_device *dev);
3716
3717/* Functions used for unicast addresses handling */
3718int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3719int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3720int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3721int dev_uc_sync(struct net_device *to, struct net_device *from);
3722int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3723void dev_uc_unsync(struct net_device *to, struct net_device *from);
3724void dev_uc_flush(struct net_device *dev);
3725void dev_uc_init(struct net_device *dev);
3726
3727/**
3728 *  __dev_uc_sync - Synchonize device's unicast list
3729 *  @dev:  device to sync
3730 *  @sync: function to call if address should be added
3731 *  @unsync: function to call if address should be removed
3732 *
3733 *  Add newly added addresses to the interface, and release
3734 *  addresses that have been deleted.
3735 */
3736static inline int __dev_uc_sync(struct net_device *dev,
3737                                int (*sync)(struct net_device *,
3738                                            const unsigned char *),
3739                                int (*unsync)(struct net_device *,
3740                                              const unsigned char *))
3741{
3742        return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3743}
3744
3745/**
3746 *  __dev_uc_unsync - Remove synchronized addresses from device
3747 *  @dev:  device to sync
3748 *  @unsync: function to call if address should be removed
3749 *
3750 *  Remove all addresses that were added to the device by dev_uc_sync().
3751 */
3752static inline void __dev_uc_unsync(struct net_device *dev,
3753                                   int (*unsync)(struct net_device *,
3754                                                 const unsigned char *))
3755{
3756        __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3757}
3758
3759/* Functions used for multicast addresses handling */
3760int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3761int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3762int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3763int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3764int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3765int dev_mc_sync(struct net_device *to, struct net_device *from);
3766int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3767void dev_mc_unsync(struct net_device *to, struct net_device *from);
3768void dev_mc_flush(struct net_device *dev);
3769void dev_mc_init(struct net_device *dev);
3770
3771/**
3772 *  __dev_mc_sync - Synchonize device's multicast list
3773 *  @dev:  device to sync
3774 *  @sync: function to call if address should be added
3775 *  @unsync: function to call if address should be removed
3776 *
3777 *  Add newly added addresses to the interface, and release
3778 *  addresses that have been deleted.
3779 */
3780static inline int __dev_mc_sync(struct net_device *dev,
3781                                int (*sync)(struct net_device *,
3782                                            const unsigned char *),
3783                                int (*unsync)(struct net_device *,
3784                                              const unsigned char *))
3785{
3786        return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3787}
3788
3789/**
3790 *  __dev_mc_unsync - Remove synchronized addresses from device
3791 *  @dev:  device to sync
3792 *  @unsync: function to call if address should be removed
3793 *
3794 *  Remove all addresses that were added to the device by dev_mc_sync().
3795 */
3796static inline void __dev_mc_unsync(struct net_device *dev,
3797                                   int (*unsync)(struct net_device *,
3798                                                 const unsigned char *))
3799{
3800        __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3801}
3802
3803/* Functions used for secondary unicast and multicast support */
3804void dev_set_rx_mode(struct net_device *dev);
3805void __dev_set_rx_mode(struct net_device *dev);
3806int dev_set_promiscuity(struct net_device *dev, int inc);
3807int dev_set_allmulti(struct net_device *dev, int inc);
3808void netdev_state_change(struct net_device *dev);
3809void netdev_notify_peers(struct net_device *dev);
3810void netdev_features_change(struct net_device *dev);
3811/* Load a device via the kmod */
3812void dev_load(struct net *net, const char *name);
3813struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3814                                        struct rtnl_link_stats64 *storage);
3815void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3816                             const struct net_device_stats *netdev_stats);
3817
3818extern int              netdev_max_backlog;
3819extern int              netdev_tstamp_prequeue;
3820extern int              weight_p;
3821extern int              dev_weight_rx_bias;
3822extern int              dev_weight_tx_bias;
3823extern int              dev_rx_weight;
3824extern int              dev_tx_weight;
3825
3826bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3827struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3828                                                     struct list_head **iter);
3829struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3830                                                     struct list_head **iter);
3831
3832/* iterate through upper list, must be called under RCU read lock */
3833#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3834        for (iter = &(dev)->adj_list.upper, \
3835             updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3836             updev; \
3837             updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3838
3839int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3840                                  int (*fn)(struct net_device *upper_dev,
3841                                            void *data),
3842                                  void *data);
3843
3844bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3845                                  struct net_device *upper_dev);
3846
3847void *netdev_lower_get_next_private(struct net_device *dev,
3848                                    struct list_head **iter);
3849void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3850                                        struct list_head **iter);
3851
3852#define netdev_for_each_lower_private(dev, priv, iter) \
3853        for (iter = (dev)->adj_list.lower.next, \
3854             priv = netdev_lower_get_next_private(dev, &(iter)); \
3855             priv; \
3856             priv = netdev_lower_get_next_private(dev, &(iter)))
3857
3858#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3859        for (iter = &(dev)->adj_list.lower, \
3860             priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3861             priv; \
3862             priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3863
3864void *netdev_lower_get_next(struct net_device *dev,
3865                                struct list_head **iter);
3866
3867#define netdev_for_each_lower_dev(dev, ldev, iter) \
3868        for (iter = (dev)->adj_list.lower.next, \
3869             ldev = netdev_lower_get_next(dev, &(iter)); \
3870             ldev; \
3871             ldev = netdev_lower_get_next(dev, &(iter)))
3872
3873struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3874                                             struct list_head **iter);
3875struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3876                                                 struct list_head **iter);
3877
3878int netdev_walk_all_lower_dev(struct net_device *dev,
3879                              int (*fn)(struct net_device *lower_dev,
3880                                        void *data),
3881                              void *data);
3882int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3883                                  int (*fn)(struct net_device *lower_dev,
3884                                            void *data),
3885                                  void *data);
3886
3887void *netdev_adjacent_get_private(struct list_head *adj_list);
3888void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3889struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3890struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3891int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3892int netdev_master_upper_dev_link(struct net_device *dev,
3893                                 struct net_device *upper_dev,
3894                                 void *upper_priv, void *upper_info);
3895void netdev_upper_dev_unlink(struct net_device *dev,
3896                             struct net_device *upper_dev);
3897void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3898void *netdev_lower_dev_get_private(struct net_device *dev,
3899                                   struct net_device *lower_dev);
3900void netdev_lower_state_changed(struct net_device *lower_dev,
3901                                void *lower_state_info);
3902
3903/* RSS keys are 40 or 52 bytes long */
3904#define NETDEV_RSS_KEY_LEN 52
3905extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3906void netdev_rss_key_fill(void *buffer, size_t len);
3907
3908int dev_get_nest_level(struct net_device *dev);
3909int skb_checksum_help(struct sk_buff *skb);
3910struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3911                                  netdev_features_t features, bool tx_path);
3912struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3913                                    netdev_features_t features);
3914
3915struct netdev_bonding_info {
3916        ifslave slave;
3917        ifbond  master;
3918};
3919
3920struct netdev_notifier_bonding_info {
3921        struct netdev_notifier_info info; /* must be first */
3922        struct netdev_bonding_info  bonding_info;
3923};
3924
3925void netdev_bonding_info_change(struct net_device *dev,
3926                                struct netdev_bonding_info *bonding_info);
3927
3928static inline
3929struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3930{
3931        return __skb_gso_segment(skb, features, true);
3932}
3933__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3934
3935static inline bool can_checksum_protocol(netdev_features_t features,
3936                                         __be16 protocol)
3937{
3938        if (protocol == htons(ETH_P_FCOE))
3939                return !!(features & NETIF_F_FCOE_CRC);
3940
3941        /* Assume this is an IP checksum (not SCTP CRC) */
3942
3943        if (features & NETIF_F_HW_CSUM) {
3944                /* Can checksum everything */
3945                return true;
3946        }
3947
3948        switch (protocol) {
3949        case htons(ETH_P_IP):
3950                return !!(features & NETIF_F_IP_CSUM);
3951        case htons(ETH_P_IPV6):
3952                return !!(features & NETIF_F_IPV6_CSUM);
3953        default:
3954                return false;
3955        }
3956}
3957
3958#ifdef CONFIG_BUG
3959void netdev_rx_csum_fault(struct net_device *dev);
3960#else
3961static inline void netdev_rx_csum_fault(struct net_device *dev)
3962{
3963}
3964#endif
3965/* rx skb timestamps */
3966void net_enable_timestamp(void);
3967void net_disable_timestamp(void);
3968
3969#ifdef CONFIG_PROC_FS
3970int __init dev_proc_init(void);
3971#else
3972#define dev_proc_init() 0
3973#endif
3974
3975static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3976                                              struct sk_buff *skb, struct net_device *dev,
3977                                              bool more)
3978{
3979        skb->xmit_more = more ? 1 : 0;
3980        return ops->ndo_start_xmit(skb, dev);
3981}
3982
3983static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3984                                            struct netdev_queue *txq, bool more)
3985{
3986        const struct net_device_ops *ops = dev->netdev_ops;
3987        int rc;
3988
3989        rc = __netdev_start_xmit(ops, skb, dev, more);
3990        if (rc == NETDEV_TX_OK)
3991                txq_trans_update(txq);
3992
3993        return rc;
3994}
3995
3996int netdev_class_create_file_ns(struct class_attribute *class_attr,
3997                                const void *ns);
3998void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3999                                 const void *ns);
4000
4001static inline int netdev_class_create_file(struct class_attribute *class_attr)
4002{
4003        return netdev_class_create_file_ns(class_attr, NULL);
4004}
4005
4006static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4007{
4008        netdev_class_remove_file_ns(class_attr, NULL);
4009}
4010
4011extern struct kobj_ns_type_operations net_ns_type_operations;
4012
4013const char *netdev_drivername(const struct net_device *dev);
4014
4015void linkwatch_run_queue(void);
4016
4017static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4018                                                          netdev_features_t f2)
4019{
4020        if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4021                if (f1 & NETIF_F_HW_CSUM)
4022                        f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4023                else
4024                        f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4025        }
4026
4027        return f1 & f2;
4028}
4029
4030static inline netdev_features_t netdev_get_wanted_features(
4031        struct net_device *dev)
4032{
4033        return (dev->features & ~dev->hw_features) | dev->wanted_features;
4034}
4035netdev_features_t netdev_increment_features(netdev_features_t all,
4036        netdev_features_t one, netdev_features_t mask);
4037
4038/* Allow TSO being used on stacked device :
4039 * Performing the GSO segmentation before last device
4040 * is a performance improvement.
4041 */
4042static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4043                                                        netdev_features_t mask)
4044{
4045        return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4046}
4047
4048int __netdev_update_features(struct net_device *dev);
4049void netdev_update_features(struct net_device *dev);
4050void netdev_change_features(struct net_device *dev);
4051
4052void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4053                                        struct net_device *dev);
4054
4055netdev_features_t passthru_features_check(struct sk_buff *skb,
4056                                          struct net_device *dev,
4057                                          netdev_features_t features);
4058netdev_features_t netif_skb_features(struct sk_buff *skb);
4059
4060static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4061{
4062        netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4063
4064        /* check flags correspondence */
4065        BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4066        BUILD_BUG_ON(SKB_GSO_UDP     != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4067        BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4068        BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4069        BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4070        BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4071        BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4072        BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4073        BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4074        BUILD_BUG_ON(SKB_GSO_IPXIP4  != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4075        BUILD_BUG_ON(SKB_GSO_IPXIP6  != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4076        BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4077        BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4078        BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4079        BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4080        BUILD_BUG_ON(SKB_GSO_SCTP    != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4081
4082        return (features & feature) == feature;
4083}
4084
4085static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4086{
4087        return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4088               (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4089}
4090
4091static inline bool netif_needs_gso(struct sk_buff *skb,
4092                                   netdev_features_t features)
4093{
4094        return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4095                unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4096                         (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4097}
4098
4099static inline void netif_set_gso_max_size(struct net_device *dev,
4100                                          unsigned int size)
4101{
4102        dev->gso_max_size = size;
4103}
4104
4105static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4106                                        int pulled_hlen, u16 mac_offset,
4107                                        int mac_len)
4108{
4109        skb->protocol = protocol;
4110        skb->encapsulation = 1;
4111        skb_push(skb, pulled_hlen);
4112        skb_reset_transport_header(skb);
4113        skb->mac_header = mac_offset;
4114        skb->network_header = skb->mac_header + mac_len;
4115        skb->mac_len = mac_len;
4116}
4117
4118static inline bool netif_is_macsec(const struct net_device *dev)
4119{
4120        return dev->priv_flags & IFF_MACSEC;
4121}
4122
4123static inline bool netif_is_macvlan(const struct net_device *dev)
4124{
4125        return dev->priv_flags & IFF_MACVLAN;
4126}
4127
4128static inline bool netif_is_macvlan_port(const struct net_device *dev)
4129{
4130        return dev->priv_flags & IFF_MACVLAN_PORT;
4131}
4132
4133static inline bool netif_is_ipvlan(const struct net_device *dev)
4134{
4135        return dev->priv_flags & IFF_IPVLAN_SLAVE;
4136}
4137
4138static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4139{
4140        return dev->priv_flags & IFF_IPVLAN_MASTER;
4141}
4142
4143static inline bool netif_is_bond_master(const struct net_device *dev)
4144{
4145        return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4146}
4147
4148static inline bool netif_is_bond_slave(const struct net_device *dev)
4149{
4150        return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4151}
4152
4153static inline bool netif_supports_nofcs(struct net_device *dev)
4154{
4155        return dev->priv_flags & IFF_SUPP_NOFCS;
4156}
4157
4158static inline bool netif_is_l3_master(const struct net_device *dev)
4159{
4160        return dev->priv_flags & IFF_L3MDEV_MASTER;
4161}
4162
4163static inline bool netif_is_l3_slave(const struct net_device *dev)
4164{
4165        return dev->priv_flags & IFF_L3MDEV_SLAVE;
4166}
4167
4168static inline bool netif_is_bridge_master(const struct net_device *dev)
4169{
4170        return dev->priv_flags & IFF_EBRIDGE;
4171}
4172
4173static inline bool netif_is_bridge_port(const struct net_device *dev)
4174{
4175        return dev->priv_flags & IFF_BRIDGE_PORT;
4176}
4177
4178static inline bool netif_is_ovs_master(const struct net_device *dev)
4179{
4180        return dev->priv_flags & IFF_OPENVSWITCH;
4181}
4182
4183static inline bool netif_is_team_master(const struct net_device *dev)
4184{
4185        return dev->priv_flags & IFF_TEAM;
4186}
4187
4188static inline bool netif_is_team_port(const struct net_device *dev)
4189{
4190        return dev->priv_flags & IFF_TEAM_PORT;
4191}
4192
4193static inline bool netif_is_lag_master(const struct net_device *dev)
4194{
4195        return netif_is_bond_master(dev) || netif_is_team_master(dev);
4196}
4197
4198static inline bool netif_is_lag_port(const struct net_device *dev)
4199{
4200        return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4201}
4202
4203static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4204{
4205        return dev->priv_flags & IFF_RXFH_CONFIGURED;
4206}
4207
4208/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4209static inline void netif_keep_dst(struct net_device *dev)
4210{
4211        dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4212}
4213
4214/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4215static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4216{
4217        /* TODO: reserve and use an additional IFF bit, if we get more users */
4218        return dev->priv_flags & IFF_MACSEC;
4219}
4220
4221extern struct pernet_operations __net_initdata loopback_net_ops;
4222
4223/* Logging, debugging and troubleshooting/diagnostic helpers. */
4224
4225/* netdev_printk helpers, similar to dev_printk */
4226
4227static inline const char *netdev_name(const struct net_device *dev)
4228{
4229        if (!dev->name[0] || strchr(dev->name, '%'))
4230                return "(unnamed net_device)";
4231        return dev->name;
4232}
4233
4234static inline const char *netdev_reg_state(const struct net_device *dev)
4235{
4236        switch (dev->reg_state) {
4237        case NETREG_UNINITIALIZED: return " (uninitialized)";
4238        case NETREG_REGISTERED: return "";
4239        case NETREG_UNREGISTERING: return " (unregistering)";
4240        case NETREG_UNREGISTERED: return " (unregistered)";
4241        case NETREG_RELEASED: return " (released)";
4242        case NETREG_DUMMY: return " (dummy)";
4243        }
4244
4245        WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4246        return " (unknown)";
4247}
4248
4249__printf(3, 4)
4250void netdev_printk(const char *level, const struct net_device *dev,
4251                   const char *format, ...);
4252__printf(2, 3)
4253void netdev_emerg(const struct net_device *dev, const char *format, ...);
4254__printf(2, 3)
4255void netdev_alert(const struct net_device *dev, const char *format, ...);
4256__printf(2, 3)
4257void netdev_crit(const struct net_device *dev, const char *format, ...);
4258__printf(2, 3)
4259void netdev_err(const struct net_device *dev, const char *format, ...);
4260__printf(2, 3)
4261void netdev_warn(const struct net_device *dev, const char *format, ...);
4262__printf(2, 3)
4263void netdev_notice(const struct net_device *dev, const char *format, ...);
4264__printf(2, 3)
4265void netdev_info(const struct net_device *dev, const char *format, ...);
4266
4267#define MODULE_ALIAS_NETDEV(device) \
4268        MODULE_ALIAS("netdev-" device)
4269
4270#if defined(CONFIG_DYNAMIC_DEBUG)
4271#define netdev_dbg(__dev, format, args...)                      \
4272do {                                                            \
4273        dynamic_netdev_dbg(__dev, format, ##args);              \
4274} while (0)
4275#elif defined(DEBUG)
4276#define netdev_dbg(__dev, format, args...)                      \
4277        netdev_printk(KERN_DEBUG, __dev, format, ##args)
4278#else
4279#define netdev_dbg(__dev, format, args...)                      \
4280({                                                              \
4281        if (0)                                                  \
4282                netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4283})
4284#endif
4285
4286#if defined(VERBOSE_DEBUG)
4287#define netdev_vdbg     netdev_dbg
4288#else
4289
4290#define netdev_vdbg(dev, format, args...)                       \
4291({                                                              \
4292        if (0)                                                  \
4293                netdev_printk(KERN_DEBUG, dev, format, ##args); \
4294        0;                                                      \
4295})
4296#endif
4297
4298/*
4299 * netdev_WARN() acts like dev_printk(), but with the key difference
4300 * of using a WARN/WARN_ON to get the message out, including the
4301 * file/line information and a backtrace.
4302 */
4303#define netdev_WARN(dev, format, args...)                       \
4304        WARN(1, "netdevice: %s%s\n" format, netdev_name(dev),   \
4305             netdev_reg_state(dev), ##args)
4306
4307/* netif printk helpers, similar to netdev_printk */
4308
4309#define netif_printk(priv, type, level, dev, fmt, args...)      \
4310do {                                                            \
4311        if (netif_msg_##type(priv))                             \
4312                netdev_printk(level, (dev), fmt, ##args);       \
4313} while (0)
4314
4315#define netif_level(level, priv, type, dev, fmt, args...)       \
4316do {                                                            \
4317        if (netif_msg_##type(priv))                             \
4318                netdev_##level(dev, fmt, ##args);               \
4319} while (0)
4320
4321#define netif_emerg(priv, type, dev, fmt, args...)              \
4322        netif_level(emerg, priv, type, dev, fmt, ##args)
4323#define netif_alert(priv, type, dev, fmt, args...)              \
4324        netif_level(alert, priv, type, dev, fmt, ##args)
4325#define netif_crit(priv, type, dev, fmt, args...)               \
4326        netif_level(crit, priv, type, dev, fmt, ##args)
4327#define netif_err(priv, type, dev, fmt, args...)                \
4328        netif_level(err, priv, type, dev, fmt, ##args)
4329#define netif_warn(priv, type, dev, fmt, args...)               \
4330        netif_level(warn, priv, type, dev, fmt, ##args)
4331#define netif_notice(priv, type, dev, fmt, args...)             \
4332        netif_level(notice, priv, type, dev, fmt, ##args)
4333#define netif_info(priv, type, dev, fmt, args...)               \
4334        netif_level(info, priv, type, dev, fmt, ##args)
4335
4336#if defined(CONFIG_DYNAMIC_DEBUG)
4337#define netif_dbg(priv, type, netdev, format, args...)          \
4338do {                                                            \
4339        if (netif_msg_##type(priv))                             \
4340                dynamic_netdev_dbg(netdev, format, ##args);     \
4341} while (0)
4342#elif defined(DEBUG)
4343#define netif_dbg(priv, type, dev, format, args...)             \
4344        netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4345#else
4346#define netif_dbg(priv, type, dev, format, args...)                     \
4347({                                                                      \
4348        if (0)                                                          \
4349                netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4350        0;                                                              \
4351})
4352#endif
4353
4354/* if @cond then downgrade to debug, else print at @level */
4355#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...)     \
4356        do {                                                              \
4357                if (cond)                                                 \
4358                        netif_dbg(priv, type, netdev, fmt, ##args);       \
4359                else                                                      \
4360                        netif_ ## level(priv, type, netdev, fmt, ##args); \
4361        } while (0)
4362
4363#if defined(VERBOSE_DEBUG)
4364#define netif_vdbg      netif_dbg
4365#else
4366#define netif_vdbg(priv, type, dev, format, args...)            \
4367({                                                              \
4368        if (0)                                                  \
4369                netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4370        0;                                                      \
4371})
4372#endif
4373
4374/*
4375 *      The list of packet types we will receive (as opposed to discard)
4376 *      and the routines to invoke.
4377 *
4378 *      Why 16. Because with 16 the only overlap we get on a hash of the
4379 *      low nibble of the protocol value is RARP/SNAP/X.25.
4380 *
4381 *      NOTE:  That is no longer true with the addition of VLAN tags.  Not
4382 *             sure which should go first, but I bet it won't make much
4383 *             difference if we are running VLANs.  The good news is that
4384 *             this protocol won't be in the list unless compiled in, so
4385 *             the average user (w/out VLANs) will not be adversely affected.
4386 *             --BLG
4387 *
4388 *              0800    IP
4389 *              8100    802.1Q VLAN
4390 *              0001    802.3
4391 *              0002    AX.25
4392 *              0004    802.2
4393 *              8035    RARP
4394 *              0005    SNAP
4395 *              0805    X.25
4396 *              0806    ARP
4397 *              8137    IPX
4398 *              0009    Localtalk
4399 *              86DD    IPv6
4400 */
4401#define PTYPE_HASH_SIZE (16)
4402#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4403
4404#endif  /* _LINUX_NETDEVICE_H */
4405