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