linux/include/linux/netdevice.h
<<
>>
Prefs
   1/*
   2 * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3 *              operating system.  INET is implemented using the  BSD Socket
   4 *              interface as the means of communication with the user level.
   5 *
   6 *              Definitions for the Interfaces handler.
   7 *
   8 * Version:     @(#)dev.h       1.0.10  08/12/93
   9 *
  10 * Authors:     Ross Biro
  11 *              Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
  12 *              Corey Minyard <wf-rch!minyard@relay.EU.net>
  13 *              Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
  14 *              Alan Cox, <alan@lxorguk.ukuu.org.uk>
  15 *              Bjorn Ekwall. <bj0rn@blox.se>
  16 *              Pekka Riikonen <priikone@poseidon.pspt.fi>
  17 *
  18 *              This program is free software; you can redistribute it and/or
  19 *              modify it under the terms of the GNU General Public License
  20 *              as published by the Free Software Foundation; either version
  21 *              2 of the License, or (at your option) any later version.
  22 *
  23 *              Moved to /usr/include/linux for NET3
  24 */
  25#ifndef _LINUX_NETDEVICE_H
  26#define _LINUX_NETDEVICE_H
  27
  28#include <linux/timer.h>
  29#include <linux/bug.h>
  30#include <linux/delay.h>
  31#include <linux/atomic.h>
  32#include <linux/prefetch.h>
  33#include <asm/cache.h>
  34#include <asm/byteorder.h>
  35
  36#include <linux/percpu.h>
  37#include <linux/rculist.h>
  38#include <linux/dmaengine.h>
  39#include <linux/workqueue.h>
  40#include <linux/dynamic_queue_limits.h>
  41
  42#include <linux/ethtool.h>
  43#include <net/net_namespace.h>
  44#include <net/dsa.h>
  45#ifdef CONFIG_DCB
  46#include <net/dcbnl.h>
  47#endif
  48#include <net/netprio_cgroup.h>
  49
  50#include <linux/netdev_features.h>
  51#include <linux/neighbour.h>
  52#include <uapi/linux/netdevice.h>
  53#include <uapi/linux/if_bonding.h>
  54#include <uapi/linux/pkt_cls.h>
  55
  56struct netpoll_info;
  57struct device;
  58struct phy_device;
  59/* 802.11 specific */
  60struct wireless_dev;
  61/* 802.15.4 specific */
  62struct wpan_dev;
  63struct mpls_dev;
  64
  65void netdev_set_default_ethtool_ops(struct net_device *dev,
  66                                    const struct ethtool_ops *ops);
  67
  68/* Backlog congestion levels */
  69#define NET_RX_SUCCESS          0       /* keep 'em coming, baby */
  70#define NET_RX_DROP             1       /* packet dropped */
  71
  72/*
  73 * Transmit return codes: transmit return codes originate from three different
  74 * namespaces:
  75 *
  76 * - qdisc return codes
  77 * - driver transmit return codes
  78 * - errno values
  79 *
  80 * Drivers are allowed to return any one of those in their hard_start_xmit()
  81 * function. Real network devices commonly used with qdiscs should only return
  82 * the driver transmit return codes though - when qdiscs are used, the actual
  83 * transmission happens asynchronously, so the value is not propagated to
  84 * higher layers. Virtual network devices transmit synchronously; in this case
  85 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
  86 * others are propagated to higher layers.
  87 */
  88
  89/* qdisc ->enqueue() return codes. */
  90#define NET_XMIT_SUCCESS        0x00
  91#define NET_XMIT_DROP           0x01    /* skb dropped                  */
  92#define NET_XMIT_CN             0x02    /* congestion notification      */
  93#define NET_XMIT_POLICED        0x03    /* skb is shot by police        */
  94#define NET_XMIT_MASK           0x0f    /* qdisc flags in net/sch_generic.h */
  95
  96/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
  97 * indicates that the device will soon be dropping packets, or already drops
  98 * some packets of the same priority; prompting us to send less aggressively. */
  99#define net_xmit_eval(e)        ((e) == NET_XMIT_CN ? 0 : (e))
 100#define net_xmit_errno(e)       ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
 101
 102/* Driver transmit return codes */
 103#define NETDEV_TX_MASK          0xf0
 104
 105enum netdev_tx {
 106        __NETDEV_TX_MIN  = INT_MIN,     /* make sure enum is signed */
 107        NETDEV_TX_OK     = 0x00,        /* driver took care of packet */
 108        NETDEV_TX_BUSY   = 0x10,        /* driver tx path was busy*/
 109        NETDEV_TX_LOCKED = 0x20,        /* driver tx lock was already taken */
 110};
 111typedef enum netdev_tx netdev_tx_t;
 112
 113/*
 114 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
 115 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
 116 */
 117static inline bool dev_xmit_complete(int rc)
 118{
 119        /*
 120         * Positive cases with an skb consumed by a driver:
 121         * - successful transmission (rc == NETDEV_TX_OK)
 122         * - error while transmitting (rc < 0)
 123         * - error while queueing to a different device (rc & NET_XMIT_MASK)
 124         */
 125        if (likely(rc < NET_XMIT_MASK))
 126                return true;
 127
 128        return false;
 129}
 130
 131/*
 132 *      Compute the worst-case header length according to the protocols
 133 *      used.
 134 */
 135
 136#if defined(CONFIG_HYPERV_NET)
 137# define LL_MAX_HEADER 128
 138#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
 139# if defined(CONFIG_MAC80211_MESH)
 140#  define LL_MAX_HEADER 128
 141# else
 142#  define LL_MAX_HEADER 96
 143# endif
 144#else
 145# define LL_MAX_HEADER 32
 146#endif
 147
 148#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
 149    !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
 150#define MAX_HEADER LL_MAX_HEADER
 151#else
 152#define MAX_HEADER (LL_MAX_HEADER + 48)
 153#endif
 154
 155/*
 156 *      Old network device statistics. Fields are native words
 157 *      (unsigned long) so they can be read and written atomically.
 158 */
 159
 160struct net_device_stats {
 161        unsigned long   rx_packets;
 162        unsigned long   tx_packets;
 163        unsigned long   rx_bytes;
 164        unsigned long   tx_bytes;
 165        unsigned long   rx_errors;
 166        unsigned long   tx_errors;
 167        unsigned long   rx_dropped;
 168        unsigned long   tx_dropped;
 169        unsigned long   multicast;
 170        unsigned long   collisions;
 171        unsigned long   rx_length_errors;
 172        unsigned long   rx_over_errors;
 173        unsigned long   rx_crc_errors;
 174        unsigned long   rx_frame_errors;
 175        unsigned long   rx_fifo_errors;
 176        unsigned long   rx_missed_errors;
 177        unsigned long   tx_aborted_errors;
 178        unsigned long   tx_carrier_errors;
 179        unsigned long   tx_fifo_errors;
 180        unsigned long   tx_heartbeat_errors;
 181        unsigned long   tx_window_errors;
 182        unsigned long   rx_compressed;
 183        unsigned long   tx_compressed;
 184};
 185
 186
 187#include <linux/cache.h>
 188#include <linux/skbuff.h>
 189
 190#ifdef CONFIG_RPS
 191#include <linux/static_key.h>
 192extern struct static_key rps_needed;
 193#endif
 194
 195struct neighbour;
 196struct neigh_parms;
 197struct sk_buff;
 198
 199struct netdev_hw_addr {
 200        struct list_head        list;
 201        unsigned char           addr[MAX_ADDR_LEN];
 202        unsigned char           type;
 203#define NETDEV_HW_ADDR_T_LAN            1
 204#define NETDEV_HW_ADDR_T_SAN            2
 205#define NETDEV_HW_ADDR_T_SLAVE          3
 206#define NETDEV_HW_ADDR_T_UNICAST        4
 207#define NETDEV_HW_ADDR_T_MULTICAST      5
 208        bool                    global_use;
 209        int                     sync_cnt;
 210        int                     refcount;
 211        int                     synced;
 212        struct rcu_head         rcu_head;
 213};
 214
 215struct netdev_hw_addr_list {
 216        struct list_head        list;
 217        int                     count;
 218};
 219
 220#define netdev_hw_addr_list_count(l) ((l)->count)
 221#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
 222#define netdev_hw_addr_list_for_each(ha, l) \
 223        list_for_each_entry(ha, &(l)->list, list)
 224
 225#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
 226#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
 227#define netdev_for_each_uc_addr(ha, dev) \
 228        netdev_hw_addr_list_for_each(ha, &(dev)->uc)
 229
 230#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
 231#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
 232#define netdev_for_each_mc_addr(ha, dev) \
 233        netdev_hw_addr_list_for_each(ha, &(dev)->mc)
 234
 235struct hh_cache {
 236        u16             hh_len;
 237        u16             __pad;
 238        seqlock_t       hh_lock;
 239
 240        /* cached hardware header; allow for machine alignment needs.        */
 241#define HH_DATA_MOD     16
 242#define HH_DATA_OFF(__len) \
 243        (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
 244#define HH_DATA_ALIGN(__len) \
 245        (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
 246        unsigned long   hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
 247};
 248
 249/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
 250 * Alternative is:
 251 *   dev->hard_header_len ? (dev->hard_header_len +
 252 *                           (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
 253 *
 254 * We could use other alignment values, but we must maintain the
 255 * relationship HH alignment <= LL alignment.
 256 */
 257#define LL_RESERVED_SPACE(dev) \
 258        ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
 259#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
 260        ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
 261
 262struct header_ops {
 263        int     (*create) (struct sk_buff *skb, struct net_device *dev,
 264                           unsigned short type, const void *daddr,
 265                           const void *saddr, unsigned int len);
 266        int     (*parse)(const struct sk_buff *skb, unsigned char *haddr);
 267        int     (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
 268        void    (*cache_update)(struct hh_cache *hh,
 269                                const struct net_device *dev,
 270                                const unsigned char *haddr);
 271        bool    (*validate)(const char *ll_header, unsigned int len);
 272};
 273
 274/* These flag bits are private to the generic network queueing
 275 * layer; they may not be explicitly referenced by any other
 276 * code.
 277 */
 278
 279enum netdev_state_t {
 280        __LINK_STATE_START,
 281        __LINK_STATE_PRESENT,
 282        __LINK_STATE_NOCARRIER,
 283        __LINK_STATE_LINKWATCH_PENDING,
 284        __LINK_STATE_DORMANT,
 285};
 286
 287
 288/*
 289 * This structure holds boot-time configured netdevice settings. They
 290 * are then used in the device probing.
 291 */
 292struct netdev_boot_setup {
 293        char name[IFNAMSIZ];
 294        struct ifmap map;
 295};
 296#define NETDEV_BOOT_SETUP_MAX 8
 297
 298int __init netdev_boot_setup(char *str);
 299
 300/*
 301 * Structure for NAPI scheduling similar to tasklet but with weighting
 302 */
 303struct napi_struct {
 304        /* The poll_list must only be managed by the entity which
 305         * changes the state of the NAPI_STATE_SCHED bit.  This means
 306         * whoever atomically sets that bit can add this napi_struct
 307         * to the per-CPU poll_list, and whoever clears that bit
 308         * can remove from the list right before clearing the bit.
 309         */
 310        struct list_head        poll_list;
 311
 312        unsigned long           state;
 313        int                     weight;
 314        unsigned int            gro_count;
 315        int                     (*poll)(struct napi_struct *, int);
 316#ifdef CONFIG_NETPOLL
 317        spinlock_t              poll_lock;
 318        int                     poll_owner;
 319#endif
 320        struct net_device       *dev;
 321        struct sk_buff          *gro_list;
 322        struct sk_buff          *skb;
 323        struct hrtimer          timer;
 324        struct list_head        dev_list;
 325        struct hlist_node       napi_hash_node;
 326        unsigned int            napi_id;
 327};
 328
 329enum {
 330        NAPI_STATE_SCHED,       /* Poll is scheduled */
 331        NAPI_STATE_DISABLE,     /* Disable pending */
 332        NAPI_STATE_NPSVC,       /* Netpoll - don't dequeue from poll_list */
 333        NAPI_STATE_HASHED,      /* In NAPI hash (busy polling possible) */
 334        NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
 335};
 336
 337enum gro_result {
 338        GRO_MERGED,
 339        GRO_MERGED_FREE,
 340        GRO_HELD,
 341        GRO_NORMAL,
 342        GRO_DROP,
 343};
 344typedef enum gro_result gro_result_t;
 345
 346/*
 347 * enum rx_handler_result - Possible return values for rx_handlers.
 348 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
 349 * further.
 350 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
 351 * case skb->dev was changed by rx_handler.
 352 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
 353 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
 354 *
 355 * rx_handlers are functions called from inside __netif_receive_skb(), to do
 356 * special processing of the skb, prior to delivery to protocol handlers.
 357 *
 358 * Currently, a net_device can only have a single rx_handler registered. Trying
 359 * to register a second rx_handler will return -EBUSY.
 360 *
 361 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
 362 * To unregister a rx_handler on a net_device, use
 363 * netdev_rx_handler_unregister().
 364 *
 365 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
 366 * do with the skb.
 367 *
 368 * If the rx_handler consumed the skb in some way, it should return
 369 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
 370 * the skb to be delivered in some other way.
 371 *
 372 * If the rx_handler changed skb->dev, to divert the skb to another
 373 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
 374 * new device will be called if it exists.
 375 *
 376 * If the rx_handler decides the skb should be ignored, it should return
 377 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
 378 * are registered on exact device (ptype->dev == skb->dev).
 379 *
 380 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
 381 * delivered, it should return RX_HANDLER_PASS.
 382 *
 383 * A device without a registered rx_handler will behave as if rx_handler
 384 * returned RX_HANDLER_PASS.
 385 */
 386
 387enum rx_handler_result {
 388        RX_HANDLER_CONSUMED,
 389        RX_HANDLER_ANOTHER,
 390        RX_HANDLER_EXACT,
 391        RX_HANDLER_PASS,
 392};
 393typedef enum rx_handler_result rx_handler_result_t;
 394typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
 395
 396void __napi_schedule(struct napi_struct *n);
 397void __napi_schedule_irqoff(struct napi_struct *n);
 398
 399static inline bool napi_disable_pending(struct napi_struct *n)
 400{
 401        return test_bit(NAPI_STATE_DISABLE, &n->state);
 402}
 403
 404/**
 405 *      napi_schedule_prep - check if NAPI can be scheduled
 406 *      @n: NAPI context
 407 *
 408 * Test if NAPI routine is already running, and if not mark
 409 * it as running.  This is used as a condition variable to
 410 * insure only one NAPI poll instance runs.  We also make
 411 * sure there is no pending NAPI disable.
 412 */
 413static inline bool napi_schedule_prep(struct napi_struct *n)
 414{
 415        return !napi_disable_pending(n) &&
 416                !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
 417}
 418
 419/**
 420 *      napi_schedule - schedule NAPI poll
 421 *      @n: NAPI context
 422 *
 423 * Schedule NAPI poll routine to be called if it is not already
 424 * running.
 425 */
 426static inline void napi_schedule(struct napi_struct *n)
 427{
 428        if (napi_schedule_prep(n))
 429                __napi_schedule(n);
 430}
 431
 432/**
 433 *      napi_schedule_irqoff - schedule NAPI poll
 434 *      @n: NAPI context
 435 *
 436 * Variant of napi_schedule(), assuming hard irqs are masked.
 437 */
 438static inline void napi_schedule_irqoff(struct napi_struct *n)
 439{
 440        if (napi_schedule_prep(n))
 441                __napi_schedule_irqoff(n);
 442}
 443
 444/* Try to reschedule poll. Called by dev->poll() after napi_complete().  */
 445static inline bool napi_reschedule(struct napi_struct *napi)
 446{
 447        if (napi_schedule_prep(napi)) {
 448                __napi_schedule(napi);
 449                return true;
 450        }
 451        return false;
 452}
 453
 454void __napi_complete(struct napi_struct *n);
 455void napi_complete_done(struct napi_struct *n, int work_done);
 456/**
 457 *      napi_complete - NAPI processing complete
 458 *      @n: NAPI context
 459 *
 460 * Mark NAPI processing as complete.
 461 * Consider using napi_complete_done() instead.
 462 */
 463static inline void napi_complete(struct napi_struct *n)
 464{
 465        return napi_complete_done(n, 0);
 466}
 467
 468/**
 469 *      napi_hash_add - add a NAPI to global hashtable
 470 *      @napi: NAPI context
 471 *
 472 * Generate a new napi_id and store a @napi under it in napi_hash.
 473 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL).
 474 * Note: This is normally automatically done from netif_napi_add(),
 475 * so might disappear in a future Linux version.
 476 */
 477void napi_hash_add(struct napi_struct *napi);
 478
 479/**
 480 *      napi_hash_del - remove a NAPI from global table
 481 *      @napi: NAPI context
 482 *
 483 * Warning: caller must observe RCU grace period
 484 * before freeing memory containing @napi, if
 485 * this function returns true.
 486 * Note: core networking stack automatically calls it
 487 * from netif_napi_del().
 488 * Drivers might want to call this helper to combine all
 489 * the needed RCU grace periods into a single one.
 490 */
 491bool napi_hash_del(struct napi_struct *napi);
 492
 493/**
 494 *      napi_disable - prevent NAPI from scheduling
 495 *      @n: NAPI context
 496 *
 497 * Stop NAPI from being scheduled on this context.
 498 * Waits till any outstanding processing completes.
 499 */
 500void napi_disable(struct napi_struct *n);
 501
 502/**
 503 *      napi_enable - enable NAPI scheduling
 504 *      @n: NAPI context
 505 *
 506 * Resume NAPI from being scheduled on this context.
 507 * Must be paired with napi_disable.
 508 */
 509static inline void napi_enable(struct napi_struct *n)
 510{
 511        BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
 512        smp_mb__before_atomic();
 513        clear_bit(NAPI_STATE_SCHED, &n->state);
 514        clear_bit(NAPI_STATE_NPSVC, &n->state);
 515}
 516
 517/**
 518 *      napi_synchronize - wait until NAPI is not running
 519 *      @n: NAPI context
 520 *
 521 * Wait until NAPI is done being scheduled on this context.
 522 * Waits till any outstanding processing completes but
 523 * does not disable future activations.
 524 */
 525static inline void napi_synchronize(const struct napi_struct *n)
 526{
 527        if (IS_ENABLED(CONFIG_SMP))
 528                while (test_bit(NAPI_STATE_SCHED, &n->state))
 529                        msleep(1);
 530        else
 531                barrier();
 532}
 533
 534enum netdev_queue_state_t {
 535        __QUEUE_STATE_DRV_XOFF,
 536        __QUEUE_STATE_STACK_XOFF,
 537        __QUEUE_STATE_FROZEN,
 538};
 539
 540#define QUEUE_STATE_DRV_XOFF    (1 << __QUEUE_STATE_DRV_XOFF)
 541#define QUEUE_STATE_STACK_XOFF  (1 << __QUEUE_STATE_STACK_XOFF)
 542#define QUEUE_STATE_FROZEN      (1 << __QUEUE_STATE_FROZEN)
 543
 544#define QUEUE_STATE_ANY_XOFF    (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
 545#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
 546                                        QUEUE_STATE_FROZEN)
 547#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
 548                                        QUEUE_STATE_FROZEN)
 549
 550/*
 551 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue.  The
 552 * netif_tx_* functions below are used to manipulate this flag.  The
 553 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
 554 * queue independently.  The netif_xmit_*stopped functions below are called
 555 * to check if the queue has been stopped by the driver or stack (either
 556 * of the XOFF bits are set in the state).  Drivers should not need to call
 557 * netif_xmit*stopped functions, they should only be using netif_tx_*.
 558 */
 559
 560struct netdev_queue {
 561/*
 562 * read-mostly part
 563 */
 564        struct net_device       *dev;
 565        struct Qdisc __rcu      *qdisc;
 566        struct Qdisc            *qdisc_sleeping;
 567#ifdef CONFIG_SYSFS
 568        struct kobject          kobj;
 569#endif
 570#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
 571        int                     numa_node;
 572#endif
 573/*
 574 * write-mostly part
 575 */
 576        spinlock_t              _xmit_lock ____cacheline_aligned_in_smp;
 577        int                     xmit_lock_owner;
 578        /*
 579         * please use this field instead of dev->trans_start
 580         */
 581        unsigned long           trans_start;
 582
 583        /*
 584         * Number of TX timeouts for this queue
 585         * (/sys/class/net/DEV/Q/trans_timeout)
 586         */
 587        unsigned long           trans_timeout;
 588
 589        unsigned long           state;
 590
 591#ifdef CONFIG_BQL
 592        struct dql              dql;
 593#endif
 594        unsigned long           tx_maxrate;
 595} ____cacheline_aligned_in_smp;
 596
 597static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
 598{
 599#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
 600        return q->numa_node;
 601#else
 602        return NUMA_NO_NODE;
 603#endif
 604}
 605
 606static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
 607{
 608#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
 609        q->numa_node = node;
 610#endif
 611}
 612
 613#ifdef CONFIG_RPS
 614/*
 615 * This structure holds an RPS map which can be of variable length.  The
 616 * map is an array of CPUs.
 617 */
 618struct rps_map {
 619        unsigned int len;
 620        struct rcu_head rcu;
 621        u16 cpus[0];
 622};
 623#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
 624
 625/*
 626 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
 627 * tail pointer for that CPU's input queue at the time of last enqueue, and
 628 * a hardware filter index.
 629 */
 630struct rps_dev_flow {
 631        u16 cpu;
 632        u16 filter;
 633        unsigned int last_qtail;
 634};
 635#define RPS_NO_FILTER 0xffff
 636
 637/*
 638 * The rps_dev_flow_table structure contains a table of flow mappings.
 639 */
 640struct rps_dev_flow_table {
 641        unsigned int mask;
 642        struct rcu_head rcu;
 643        struct rps_dev_flow flows[0];
 644};
 645#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
 646    ((_num) * sizeof(struct rps_dev_flow)))
 647
 648/*
 649 * The rps_sock_flow_table contains mappings of flows to the last CPU
 650 * on which they were processed by the application (set in recvmsg).
 651 * Each entry is a 32bit value. Upper part is the high-order bits
 652 * of flow hash, lower part is CPU number.
 653 * rps_cpu_mask is used to partition the space, depending on number of
 654 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
 655 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
 656 * meaning we use 32-6=26 bits for the hash.
 657 */
 658struct rps_sock_flow_table {
 659        u32     mask;
 660
 661        u32     ents[0] ____cacheline_aligned_in_smp;
 662};
 663#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
 664
 665#define RPS_NO_CPU 0xffff
 666
 667extern u32 rps_cpu_mask;
 668extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
 669
 670static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
 671                                        u32 hash)
 672{
 673        if (table && hash) {
 674                unsigned int index = hash & table->mask;
 675                u32 val = hash & ~rps_cpu_mask;
 676
 677                /* We only give a hint, preemption can change CPU under us */
 678                val |= raw_smp_processor_id();
 679
 680                if (table->ents[index] != val)
 681                        table->ents[index] = val;
 682        }
 683}
 684
 685#ifdef CONFIG_RFS_ACCEL
 686bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
 687                         u16 filter_id);
 688#endif
 689#endif /* CONFIG_RPS */
 690
 691/* This structure contains an instance of an RX queue. */
 692struct netdev_rx_queue {
 693#ifdef CONFIG_RPS
 694        struct rps_map __rcu            *rps_map;
 695        struct rps_dev_flow_table __rcu *rps_flow_table;
 696#endif
 697        struct kobject                  kobj;
 698        struct net_device               *dev;
 699} ____cacheline_aligned_in_smp;
 700
 701/*
 702 * RX queue sysfs structures and functions.
 703 */
 704struct rx_queue_attribute {
 705        struct attribute attr;
 706        ssize_t (*show)(struct netdev_rx_queue *queue,
 707            struct rx_queue_attribute *attr, char *buf);
 708        ssize_t (*store)(struct netdev_rx_queue *queue,
 709            struct rx_queue_attribute *attr, const char *buf, size_t len);
 710};
 711
 712#ifdef CONFIG_XPS
 713/*
 714 * This structure holds an XPS map which can be of variable length.  The
 715 * map is an array of queues.
 716 */
 717struct xps_map {
 718        unsigned int len;
 719        unsigned int alloc_len;
 720        struct rcu_head rcu;
 721        u16 queues[0];
 722};
 723#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
 724#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
 725       - sizeof(struct xps_map)) / sizeof(u16))
 726
 727/*
 728 * This structure holds all XPS maps for device.  Maps are indexed by CPU.
 729 */
 730struct xps_dev_maps {
 731        struct rcu_head rcu;
 732        struct xps_map __rcu *cpu_map[0];
 733};
 734#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) +                \
 735    (nr_cpu_ids * sizeof(struct xps_map *)))
 736#endif /* CONFIG_XPS */
 737
 738#define TC_MAX_QUEUE    16
 739#define TC_BITMASK      15
 740/* HW offloaded queuing disciplines txq count and offset maps */
 741struct netdev_tc_txq {
 742        u16 count;
 743        u16 offset;
 744};
 745
 746#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
 747/*
 748 * This structure is to hold information about the device
 749 * configured to run FCoE protocol stack.
 750 */
 751struct netdev_fcoe_hbainfo {
 752        char    manufacturer[64];
 753        char    serial_number[64];
 754        char    hardware_version[64];
 755        char    driver_version[64];
 756        char    optionrom_version[64];
 757        char    firmware_version[64];
 758        char    model[256];
 759        char    model_description[256];
 760};
 761#endif
 762
 763#define MAX_PHYS_ITEM_ID_LEN 32
 764
 765/* This structure holds a unique identifier to identify some
 766 * physical item (port for example) used by a netdevice.
 767 */
 768struct netdev_phys_item_id {
 769        unsigned char id[MAX_PHYS_ITEM_ID_LEN];
 770        unsigned char id_len;
 771};
 772
 773static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
 774                                            struct netdev_phys_item_id *b)
 775{
 776        return a->id_len == b->id_len &&
 777               memcmp(a->id, b->id, a->id_len) == 0;
 778}
 779
 780typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
 781                                       struct sk_buff *skb);
 782
 783/* These structures hold the attributes of qdisc and classifiers
 784 * that are being passed to the netdevice through the setup_tc op.
 785 */
 786enum {
 787        TC_SETUP_MQPRIO,
 788        TC_SETUP_CLSU32,
 789        TC_SETUP_CLSFLOWER,
 790};
 791
 792struct tc_cls_u32_offload;
 793
 794struct tc_to_netdev {
 795        unsigned int type;
 796        union {
 797                u8 tc;
 798                struct tc_cls_u32_offload *cls_u32;
 799                struct tc_cls_flower_offload *cls_flower;
 800        };
 801};
 802
 803
 804/*
 805 * This structure defines the management hooks for network devices.
 806 * The following hooks can be defined; unless noted otherwise, they are
 807 * optional and can be filled with a null pointer.
 808 *
 809 * int (*ndo_init)(struct net_device *dev);
 810 *     This function is called once when a network device is registered.
 811 *     The network device can use this for any late stage initialization
 812 *     or semantic validation. It can fail with an error code which will
 813 *     be propagated back to register_netdev.
 814 *
 815 * void (*ndo_uninit)(struct net_device *dev);
 816 *     This function is called when device is unregistered or when registration
 817 *     fails. It is not called if init fails.
 818 *
 819 * int (*ndo_open)(struct net_device *dev);
 820 *     This function is called when a network device transitions to the up
 821 *     state.
 822 *
 823 * int (*ndo_stop)(struct net_device *dev);
 824 *     This function is called when a network device transitions to the down
 825 *     state.
 826 *
 827 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
 828 *                               struct net_device *dev);
 829 *      Called when a packet needs to be transmitted.
 830 *      Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
 831 *      the queue before that can happen; it's for obsolete devices and weird
 832 *      corner cases, but the stack really does a non-trivial amount
 833 *      of useless work if you return NETDEV_TX_BUSY.
 834 *        (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
 835 *      Required; cannot be NULL.
 836 *
 837 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
 838 *              netdev_features_t features);
 839 *      Adjusts the requested feature flags according to device-specific
 840 *      constraints, and returns the resulting flags. Must not modify
 841 *      the device state.
 842 *
 843 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
 844 *                         void *accel_priv, select_queue_fallback_t fallback);
 845 *      Called to decide which queue to use when device supports multiple
 846 *      transmit queues.
 847 *
 848 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
 849 *      This function is called to allow device receiver to make
 850 *      changes to configuration when multicast or promiscuous is enabled.
 851 *
 852 * void (*ndo_set_rx_mode)(struct net_device *dev);
 853 *      This function is called device changes address list filtering.
 854 *      If driver handles unicast address filtering, it should set
 855 *      IFF_UNICAST_FLT in its priv_flags.
 856 *
 857 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
 858 *      This function  is called when the Media Access Control address
 859 *      needs to be changed. If this interface is not defined, the
 860 *      MAC address can not be changed.
 861 *
 862 * int (*ndo_validate_addr)(struct net_device *dev);
 863 *      Test if Media Access Control address is valid for the device.
 864 *
 865 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
 866 *      Called when a user requests an ioctl which can't be handled by
 867 *      the generic interface code. If not defined ioctls return
 868 *      not supported error code.
 869 *
 870 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
 871 *      Used to set network devices bus interface parameters. This interface
 872 *      is retained for legacy reasons; new devices should use the bus
 873 *      interface (PCI) for low level management.
 874 *
 875 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
 876 *      Called when a user wants to change the Maximum Transfer Unit
 877 *      of a device. If not defined, any request to change MTU will
 878 *      will return an error.
 879 *
 880 * void (*ndo_tx_timeout)(struct net_device *dev);
 881 *      Callback used when the transmitter has not made any progress
 882 *      for dev->watchdog ticks.
 883 *
 884 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
 885 *                      struct rtnl_link_stats64 *storage);
 886 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
 887 *      Called when a user wants to get the network device usage
 888 *      statistics. Drivers must do one of the following:
 889 *      1. Define @ndo_get_stats64 to fill in a zero-initialised
 890 *         rtnl_link_stats64 structure passed by the caller.
 891 *      2. Define @ndo_get_stats to update a net_device_stats structure
 892 *         (which should normally be dev->stats) and return a pointer to
 893 *         it. The structure may be changed asynchronously only if each
 894 *         field is written atomically.
 895 *      3. Update dev->stats asynchronously and atomically, and define
 896 *         neither operation.
 897 *
 898 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
 899 *      If device supports VLAN filtering this function is called when a
 900 *      VLAN id is registered.
 901 *
 902 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
 903 *      If device supports VLAN filtering this function is called when a
 904 *      VLAN id is unregistered.
 905 *
 906 * void (*ndo_poll_controller)(struct net_device *dev);
 907 *
 908 *      SR-IOV management functions.
 909 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
 910 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
 911 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
 912 *                        int max_tx_rate);
 913 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
 914 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
 915 * int (*ndo_get_vf_config)(struct net_device *dev,
 916 *                          int vf, struct ifla_vf_info *ivf);
 917 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
 918 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
 919 *                        struct nlattr *port[]);
 920 *
 921 *      Enable or disable the VF ability to query its RSS Redirection Table and
 922 *      Hash Key. This is needed since on some devices VF share this information
 923 *      with PF and querying it may introduce a theoretical security risk.
 924 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
 925 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
 926 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
 927 *      Called to setup 'tc' number of traffic classes in the net device. This
 928 *      is always called from the stack with the rtnl lock held and netif tx
 929 *      queues stopped. This allows the netdevice to perform queue management
 930 *      safely.
 931 *
 932 *      Fiber Channel over Ethernet (FCoE) offload functions.
 933 * int (*ndo_fcoe_enable)(struct net_device *dev);
 934 *      Called when the FCoE protocol stack wants to start using LLD for FCoE
 935 *      so the underlying device can perform whatever needed configuration or
 936 *      initialization to support acceleration of FCoE traffic.
 937 *
 938 * int (*ndo_fcoe_disable)(struct net_device *dev);
 939 *      Called when the FCoE protocol stack wants to stop using LLD for FCoE
 940 *      so the underlying device can perform whatever needed clean-ups to
 941 *      stop supporting acceleration of FCoE traffic.
 942 *
 943 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
 944 *                           struct scatterlist *sgl, unsigned int sgc);
 945 *      Called when the FCoE Initiator wants to initialize an I/O that
 946 *      is a possible candidate for Direct Data Placement (DDP). The LLD can
 947 *      perform necessary setup and returns 1 to indicate the device is set up
 948 *      successfully to perform DDP on this I/O, otherwise this returns 0.
 949 *
 950 * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
 951 *      Called when the FCoE Initiator/Target is done with the DDPed I/O as
 952 *      indicated by the FC exchange id 'xid', so the underlying device can
 953 *      clean up and reuse resources for later DDP requests.
 954 *
 955 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
 956 *                            struct scatterlist *sgl, unsigned int sgc);
 957 *      Called when the FCoE Target wants to initialize an I/O that
 958 *      is a possible candidate for Direct Data Placement (DDP). The LLD can
 959 *      perform necessary setup and returns 1 to indicate the device is set up
 960 *      successfully to perform DDP on this I/O, otherwise this returns 0.
 961 *
 962 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
 963 *                             struct netdev_fcoe_hbainfo *hbainfo);
 964 *      Called when the FCoE Protocol stack wants information on the underlying
 965 *      device. This information is utilized by the FCoE protocol stack to
 966 *      register attributes with Fiber Channel management service as per the
 967 *      FC-GS Fabric Device Management Information(FDMI) specification.
 968 *
 969 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
 970 *      Called when the underlying device wants to override default World Wide
 971 *      Name (WWN) generation mechanism in FCoE protocol stack to pass its own
 972 *      World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
 973 *      protocol stack to use.
 974 *
 975 *      RFS acceleration.
 976 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
 977 *                          u16 rxq_index, u32 flow_id);
 978 *      Set hardware filter for RFS.  rxq_index is the target queue index;
 979 *      flow_id is a flow ID to be passed to rps_may_expire_flow() later.
 980 *      Return the filter ID on success, or a negative error code.
 981 *
 982 *      Slave management functions (for bridge, bonding, etc).
 983 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
 984 *      Called to make another netdev an underling.
 985 *
 986 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
 987 *      Called to release previously enslaved netdev.
 988 *
 989 *      Feature/offload setting functions.
 990 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
 991 *      Called to update device configuration to new features. Passed
 992 *      feature set might be less than what was returned by ndo_fix_features()).
 993 *      Must return >0 or -errno if it changed dev->features itself.
 994 *
 995 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
 996 *                    struct net_device *dev,
 997 *                    const unsigned char *addr, u16 vid, u16 flags)
 998 *      Adds an FDB entry to dev for addr.
 999 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1000 *                    struct net_device *dev,
1001 *                    const unsigned char *addr, u16 vid)
1002 *      Deletes the FDB entry from dev coresponding to addr.
1003 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1004 *                     struct net_device *dev, struct net_device *filter_dev,
1005 *                     int idx)
1006 *      Used to add FDB entries to dump requests. Implementers should add
1007 *      entries to skb and update idx with the number of entries.
1008 *
1009 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1010 *                           u16 flags)
1011 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1012 *                           struct net_device *dev, u32 filter_mask,
1013 *                           int nlflags)
1014 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1015 *                           u16 flags);
1016 *
1017 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1018 *      Called to change device carrier. Soft-devices (like dummy, team, etc)
1019 *      which do not represent real hardware may define this to allow their
1020 *      userspace components to manage their virtual carrier state. Devices
1021 *      that determine carrier state from physical hardware properties (eg
1022 *      network cables) or protocol-dependent mechanisms (eg
1023 *      USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1024 *
1025 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1026 *                             struct netdev_phys_item_id *ppid);
1027 *      Called to get ID of physical port of this device. If driver does
1028 *      not implement this, it is assumed that the hw is not able to have
1029 *      multiple net devices on single physical port.
1030 *
1031 * void (*ndo_add_vxlan_port)(struct  net_device *dev,
1032 *                            sa_family_t sa_family, __be16 port);
1033 *      Called by vxlan to notify a driver about the UDP port and socket
1034 *      address family that vxlan is listening to. It is called only when
1035 *      a new port starts listening. The operation is protected by the
1036 *      vxlan_net->sock_lock.
1037 *
1038 * void (*ndo_add_geneve_port)(struct net_device *dev,
1039 *                             sa_family_t sa_family, __be16 port);
1040 *      Called by geneve to notify a driver about the UDP port and socket
1041 *      address family that geneve is listnening to. It is called only when
1042 *      a new port starts listening. The operation is protected by the
1043 *      geneve_net->sock_lock.
1044 *
1045 * void (*ndo_del_geneve_port)(struct net_device *dev,
1046 *                             sa_family_t sa_family, __be16 port);
1047 *      Called by geneve to notify the driver about a UDP port and socket
1048 *      address family that geneve is not listening to anymore. The operation
1049 *      is protected by the geneve_net->sock_lock.
1050 *
1051 * void (*ndo_del_vxlan_port)(struct  net_device *dev,
1052 *                            sa_family_t sa_family, __be16 port);
1053 *      Called by vxlan to notify the driver about a UDP port and socket
1054 *      address family that vxlan is not listening to anymore. The operation
1055 *      is protected by the vxlan_net->sock_lock.
1056 *
1057 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1058 *                               struct net_device *dev)
1059 *      Called by upper layer devices to accelerate switching or other
1060 *      station functionality into hardware. 'pdev is the lowerdev
1061 *      to use for the offload and 'dev' is the net device that will
1062 *      back the offload. Returns a pointer to the private structure
1063 *      the upper layer will maintain.
1064 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1065 *      Called by upper layer device to delete the station created
1066 *      by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1067 *      the station and priv is the structure returned by the add
1068 *      operation.
1069 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1070 *                                    struct net_device *dev,
1071 *                                    void *priv);
1072 *      Callback to use for xmit over the accelerated station. This
1073 *      is used in place of ndo_start_xmit on accelerated net
1074 *      devices.
1075 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1076 *                                         struct net_device *dev
1077 *                                         netdev_features_t features);
1078 *      Called by core transmit path to determine if device is capable of
1079 *      performing offload operations on a given packet. This is to give
1080 *      the device an opportunity to implement any restrictions that cannot
1081 *      be otherwise expressed by feature flags. The check is called with
1082 *      the set of features that the stack has calculated and it returns
1083 *      those the driver believes to be appropriate.
1084 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1085 *                           int queue_index, u32 maxrate);
1086 *      Called when a user wants to set a max-rate limitation of specific
1087 *      TX queue.
1088 * int (*ndo_get_iflink)(const struct net_device *dev);
1089 *      Called to get the iflink value of this device.
1090 * void (*ndo_change_proto_down)(struct net_device *dev,
1091 *                               bool proto_down);
1092 *      This function is used to pass protocol port error state information
1093 *      to the switch driver. The switch driver can react to the proto_down
1094 *      by doing a phys down on the associated switch port.
1095 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1096 *      This function is used to get egress tunnel information for given skb.
1097 *      This is useful for retrieving outer tunnel header parameters while
1098 *      sampling packet.
1099 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1100 *      This function is used to specify the headroom that the skb must
1101 *      consider when allocation skb during packet reception. Setting
1102 *      appropriate rx headroom value allows avoiding skb head copy on
1103 *      forward. Setting a negative value resets the rx headroom to the
1104 *      default value.
1105 *
1106 */
1107struct net_device_ops {
1108        int                     (*ndo_init)(struct net_device *dev);
1109        void                    (*ndo_uninit)(struct net_device *dev);
1110        int                     (*ndo_open)(struct net_device *dev);
1111        int                     (*ndo_stop)(struct net_device *dev);
1112        netdev_tx_t             (*ndo_start_xmit)(struct sk_buff *skb,
1113                                                  struct net_device *dev);
1114        netdev_features_t       (*ndo_features_check)(struct sk_buff *skb,
1115                                                      struct net_device *dev,
1116                                                      netdev_features_t features);
1117        u16                     (*ndo_select_queue)(struct net_device *dev,
1118                                                    struct sk_buff *skb,
1119                                                    void *accel_priv,
1120                                                    select_queue_fallback_t fallback);
1121        void                    (*ndo_change_rx_flags)(struct net_device *dev,
1122                                                       int flags);
1123        void                    (*ndo_set_rx_mode)(struct net_device *dev);
1124        int                     (*ndo_set_mac_address)(struct net_device *dev,
1125                                                       void *addr);
1126        int                     (*ndo_validate_addr)(struct net_device *dev);
1127        int                     (*ndo_do_ioctl)(struct net_device *dev,
1128                                                struct ifreq *ifr, int cmd);
1129        int                     (*ndo_set_config)(struct net_device *dev,
1130                                                  struct ifmap *map);
1131        int                     (*ndo_change_mtu)(struct net_device *dev,
1132                                                  int new_mtu);
1133        int                     (*ndo_neigh_setup)(struct net_device *dev,
1134                                                   struct neigh_parms *);
1135        void                    (*ndo_tx_timeout) (struct net_device *dev);
1136
1137        struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1138                                                     struct rtnl_link_stats64 *storage);
1139        struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1140
1141        int                     (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1142                                                       __be16 proto, u16 vid);
1143        int                     (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1144                                                        __be16 proto, u16 vid);
1145#ifdef CONFIG_NET_POLL_CONTROLLER
1146        void                    (*ndo_poll_controller)(struct net_device *dev);
1147        int                     (*ndo_netpoll_setup)(struct net_device *dev,
1148                                                     struct netpoll_info *info);
1149        void                    (*ndo_netpoll_cleanup)(struct net_device *dev);
1150#endif
1151#ifdef CONFIG_NET_RX_BUSY_POLL
1152        int                     (*ndo_busy_poll)(struct napi_struct *dev);
1153#endif
1154        int                     (*ndo_set_vf_mac)(struct net_device *dev,
1155                                                  int queue, u8 *mac);
1156        int                     (*ndo_set_vf_vlan)(struct net_device *dev,
1157                                                   int queue, u16 vlan, u8 qos);
1158        int                     (*ndo_set_vf_rate)(struct net_device *dev,
1159                                                   int vf, int min_tx_rate,
1160                                                   int max_tx_rate);
1161        int                     (*ndo_set_vf_spoofchk)(struct net_device *dev,
1162                                                       int vf, bool setting);
1163        int                     (*ndo_set_vf_trust)(struct net_device *dev,
1164                                                    int vf, bool setting);
1165        int                     (*ndo_get_vf_config)(struct net_device *dev,
1166                                                     int vf,
1167                                                     struct ifla_vf_info *ivf);
1168        int                     (*ndo_set_vf_link_state)(struct net_device *dev,
1169                                                         int vf, int link_state);
1170        int                     (*ndo_get_vf_stats)(struct net_device *dev,
1171                                                    int vf,
1172                                                    struct ifla_vf_stats
1173                                                    *vf_stats);
1174        int                     (*ndo_set_vf_port)(struct net_device *dev,
1175                                                   int vf,
1176                                                   struct nlattr *port[]);
1177        int                     (*ndo_get_vf_port)(struct net_device *dev,
1178                                                   int vf, struct sk_buff *skb);
1179        int                     (*ndo_set_vf_guid)(struct net_device *dev,
1180                                                   int vf, u64 guid,
1181                                                   int guid_type);
1182        int                     (*ndo_set_vf_rss_query_en)(
1183                                                   struct net_device *dev,
1184                                                   int vf, bool setting);
1185        int                     (*ndo_setup_tc)(struct net_device *dev,
1186                                                u32 handle,
1187                                                __be16 protocol,
1188                                                struct tc_to_netdev *tc);
1189#if IS_ENABLED(CONFIG_FCOE)
1190        int                     (*ndo_fcoe_enable)(struct net_device *dev);
1191        int                     (*ndo_fcoe_disable)(struct net_device *dev);
1192        int                     (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1193                                                      u16 xid,
1194                                                      struct scatterlist *sgl,
1195                                                      unsigned int sgc);
1196        int                     (*ndo_fcoe_ddp_done)(struct net_device *dev,
1197                                                     u16 xid);
1198        int                     (*ndo_fcoe_ddp_target)(struct net_device *dev,
1199                                                       u16 xid,
1200                                                       struct scatterlist *sgl,
1201                                                       unsigned int sgc);
1202        int                     (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1203                                                        struct netdev_fcoe_hbainfo *hbainfo);
1204#endif
1205
1206#if IS_ENABLED(CONFIG_LIBFCOE)
1207#define NETDEV_FCOE_WWNN 0
1208#define NETDEV_FCOE_WWPN 1
1209        int                     (*ndo_fcoe_get_wwn)(struct net_device *dev,
1210                                                    u64 *wwn, int type);
1211#endif
1212
1213#ifdef CONFIG_RFS_ACCEL
1214        int                     (*ndo_rx_flow_steer)(struct net_device *dev,
1215                                                     const struct sk_buff *skb,
1216                                                     u16 rxq_index,
1217                                                     u32 flow_id);
1218#endif
1219        int                     (*ndo_add_slave)(struct net_device *dev,
1220                                                 struct net_device *slave_dev);
1221        int                     (*ndo_del_slave)(struct net_device *dev,
1222                                                 struct net_device *slave_dev);
1223        netdev_features_t       (*ndo_fix_features)(struct net_device *dev,
1224                                                    netdev_features_t features);
1225        int                     (*ndo_set_features)(struct net_device *dev,
1226                                                    netdev_features_t features);
1227        int                     (*ndo_neigh_construct)(struct neighbour *n);
1228        void                    (*ndo_neigh_destroy)(struct neighbour *n);
1229
1230        int                     (*ndo_fdb_add)(struct ndmsg *ndm,
1231                                               struct nlattr *tb[],
1232                                               struct net_device *dev,
1233                                               const unsigned char *addr,
1234                                               u16 vid,
1235                                               u16 flags);
1236        int                     (*ndo_fdb_del)(struct ndmsg *ndm,
1237                                               struct nlattr *tb[],
1238                                               struct net_device *dev,
1239                                               const unsigned char *addr,
1240                                               u16 vid);
1241        int                     (*ndo_fdb_dump)(struct sk_buff *skb,
1242                                                struct netlink_callback *cb,
1243                                                struct net_device *dev,
1244                                                struct net_device *filter_dev,
1245                                                int idx);
1246
1247        int                     (*ndo_bridge_setlink)(struct net_device *dev,
1248                                                      struct nlmsghdr *nlh,
1249                                                      u16 flags);
1250        int                     (*ndo_bridge_getlink)(struct sk_buff *skb,
1251                                                      u32 pid, u32 seq,
1252                                                      struct net_device *dev,
1253                                                      u32 filter_mask,
1254                                                      int nlflags);
1255        int                     (*ndo_bridge_dellink)(struct net_device *dev,
1256                                                      struct nlmsghdr *nlh,
1257                                                      u16 flags);
1258        int                     (*ndo_change_carrier)(struct net_device *dev,
1259                                                      bool new_carrier);
1260        int                     (*ndo_get_phys_port_id)(struct net_device *dev,
1261                                                        struct netdev_phys_item_id *ppid);
1262        int                     (*ndo_get_phys_port_name)(struct net_device *dev,
1263                                                          char *name, size_t len);
1264        void                    (*ndo_add_vxlan_port)(struct  net_device *dev,
1265                                                      sa_family_t sa_family,
1266                                                      __be16 port);
1267        void                    (*ndo_del_vxlan_port)(struct  net_device *dev,
1268                                                      sa_family_t sa_family,
1269                                                      __be16 port);
1270        void                    (*ndo_add_geneve_port)(struct  net_device *dev,
1271                                                       sa_family_t sa_family,
1272                                                       __be16 port);
1273        void                    (*ndo_del_geneve_port)(struct  net_device *dev,
1274                                                       sa_family_t sa_family,
1275                                                       __be16 port);
1276        void*                   (*ndo_dfwd_add_station)(struct net_device *pdev,
1277                                                        struct net_device *dev);
1278        void                    (*ndo_dfwd_del_station)(struct net_device *pdev,
1279                                                        void *priv);
1280
1281        netdev_tx_t             (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1282                                                        struct net_device *dev,
1283                                                        void *priv);
1284        int                     (*ndo_get_lock_subclass)(struct net_device *dev);
1285        int                     (*ndo_set_tx_maxrate)(struct net_device *dev,
1286                                                      int queue_index,
1287                                                      u32 maxrate);
1288        int                     (*ndo_get_iflink)(const struct net_device *dev);
1289        int                     (*ndo_change_proto_down)(struct net_device *dev,
1290                                                         bool proto_down);
1291        int                     (*ndo_fill_metadata_dst)(struct net_device *dev,
1292                                                       struct sk_buff *skb);
1293        void                    (*ndo_set_rx_headroom)(struct net_device *dev,
1294                                                       int needed_headroom);
1295};
1296
1297/**
1298 * enum net_device_priv_flags - &struct net_device priv_flags
1299 *
1300 * These are the &struct net_device, they are only set internally
1301 * by drivers and used in the kernel. These flags are invisible to
1302 * userspace; this means that the order of these flags can change
1303 * during any kernel release.
1304 *
1305 * You should have a pretty good reason to be extending these flags.
1306 *
1307 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1308 * @IFF_EBRIDGE: Ethernet bridging device
1309 * @IFF_BONDING: bonding master or slave
1310 * @IFF_ISATAP: ISATAP interface (RFC4214)
1311 * @IFF_WAN_HDLC: WAN HDLC device
1312 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1313 *      release skb->dst
1314 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1315 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1316 * @IFF_MACVLAN_PORT: device used as macvlan port
1317 * @IFF_BRIDGE_PORT: device used as bridge port
1318 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1319 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1320 * @IFF_UNICAST_FLT: Supports unicast filtering
1321 * @IFF_TEAM_PORT: device used as team port
1322 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1323 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1324 *      change when it's running
1325 * @IFF_MACVLAN: Macvlan device
1326 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1327 *      underlying stacked devices
1328 * @IFF_IPVLAN_MASTER: IPvlan master device
1329 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1330 * @IFF_L3MDEV_MASTER: device is an L3 master device
1331 * @IFF_NO_QUEUE: device can run without qdisc attached
1332 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1333 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1334 * @IFF_TEAM: device is a team device
1335 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1336 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1337 *      entity (i.e. the master device for bridged veth)
1338 * @IFF_MACSEC: device is a MACsec device
1339 */
1340enum netdev_priv_flags {
1341        IFF_802_1Q_VLAN                 = 1<<0,
1342        IFF_EBRIDGE                     = 1<<1,
1343        IFF_BONDING                     = 1<<2,
1344        IFF_ISATAP                      = 1<<3,
1345        IFF_WAN_HDLC                    = 1<<4,
1346        IFF_XMIT_DST_RELEASE            = 1<<5,
1347        IFF_DONT_BRIDGE                 = 1<<6,
1348        IFF_DISABLE_NETPOLL             = 1<<7,
1349        IFF_MACVLAN_PORT                = 1<<8,
1350        IFF_BRIDGE_PORT                 = 1<<9,
1351        IFF_OVS_DATAPATH                = 1<<10,
1352        IFF_TX_SKB_SHARING              = 1<<11,
1353        IFF_UNICAST_FLT                 = 1<<12,
1354        IFF_TEAM_PORT                   = 1<<13,
1355        IFF_SUPP_NOFCS                  = 1<<14,
1356        IFF_LIVE_ADDR_CHANGE            = 1<<15,
1357        IFF_MACVLAN                     = 1<<16,
1358        IFF_XMIT_DST_RELEASE_PERM       = 1<<17,
1359        IFF_IPVLAN_MASTER               = 1<<18,
1360        IFF_IPVLAN_SLAVE                = 1<<19,
1361        IFF_L3MDEV_MASTER               = 1<<20,
1362        IFF_NO_QUEUE                    = 1<<21,
1363        IFF_OPENVSWITCH                 = 1<<22,
1364        IFF_L3MDEV_SLAVE                = 1<<23,
1365        IFF_TEAM                        = 1<<24,
1366        IFF_RXFH_CONFIGURED             = 1<<25,
1367        IFF_PHONY_HEADROOM              = 1<<26,
1368        IFF_MACSEC                      = 1<<27,
1369};
1370
1371#define IFF_802_1Q_VLAN                 IFF_802_1Q_VLAN
1372#define IFF_EBRIDGE                     IFF_EBRIDGE
1373#define IFF_BONDING                     IFF_BONDING
1374#define IFF_ISATAP                      IFF_ISATAP
1375#define IFF_WAN_HDLC                    IFF_WAN_HDLC
1376#define IFF_XMIT_DST_RELEASE            IFF_XMIT_DST_RELEASE
1377#define IFF_DONT_BRIDGE                 IFF_DONT_BRIDGE
1378#define IFF_DISABLE_NETPOLL             IFF_DISABLE_NETPOLL
1379#define IFF_MACVLAN_PORT                IFF_MACVLAN_PORT
1380#define IFF_BRIDGE_PORT                 IFF_BRIDGE_PORT
1381#define IFF_OVS_DATAPATH                IFF_OVS_DATAPATH
1382#define IFF_TX_SKB_SHARING              IFF_TX_SKB_SHARING
1383#define IFF_UNICAST_FLT                 IFF_UNICAST_FLT
1384#define IFF_TEAM_PORT                   IFF_TEAM_PORT
1385#define IFF_SUPP_NOFCS                  IFF_SUPP_NOFCS
1386#define IFF_LIVE_ADDR_CHANGE            IFF_LIVE_ADDR_CHANGE
1387#define IFF_MACVLAN                     IFF_MACVLAN
1388#define IFF_XMIT_DST_RELEASE_PERM       IFF_XMIT_DST_RELEASE_PERM
1389#define IFF_IPVLAN_MASTER               IFF_IPVLAN_MASTER
1390#define IFF_IPVLAN_SLAVE                IFF_IPVLAN_SLAVE
1391#define IFF_L3MDEV_MASTER               IFF_L3MDEV_MASTER
1392#define IFF_NO_QUEUE                    IFF_NO_QUEUE
1393#define IFF_OPENVSWITCH                 IFF_OPENVSWITCH
1394#define IFF_L3MDEV_SLAVE                IFF_L3MDEV_SLAVE
1395#define IFF_TEAM                        IFF_TEAM
1396#define IFF_RXFH_CONFIGURED             IFF_RXFH_CONFIGURED
1397#define IFF_MACSEC                      IFF_MACSEC
1398
1399/**
1400 *      struct net_device - The DEVICE structure.
1401 *              Actually, this whole structure is a big mistake.  It mixes I/O
1402 *              data with strictly "high-level" data, and it has to know about
1403 *              almost every data structure used in the INET module.
1404 *
1405 *      @name:  This is the first field of the "visible" part of this structure
1406 *              (i.e. as seen by users in the "Space.c" file).  It is the name
1407 *              of the interface.
1408 *
1409 *      @name_hlist:    Device name hash chain, please keep it close to name[]
1410 *      @ifalias:       SNMP alias
1411 *      @mem_end:       Shared memory end
1412 *      @mem_start:     Shared memory start
1413 *      @base_addr:     Device I/O address
1414 *      @irq:           Device IRQ number
1415 *
1416 *      @carrier_changes:       Stats to monitor carrier on<->off transitions
1417 *
1418 *      @state:         Generic network queuing layer state, see netdev_state_t
1419 *      @dev_list:      The global list of network devices
1420 *      @napi_list:     List entry used for polling NAPI devices
1421 *      @unreg_list:    List entry  when we are unregistering the
1422 *                      device; see the function unregister_netdev
1423 *      @close_list:    List entry used when we are closing the device
1424 *      @ptype_all:     Device-specific packet handlers for all protocols
1425 *      @ptype_specific: Device-specific, protocol-specific packet handlers
1426 *
1427 *      @adj_list:      Directly linked devices, like slaves for bonding
1428 *      @all_adj_list:  All linked devices, *including* neighbours
1429 *      @features:      Currently active device features
1430 *      @hw_features:   User-changeable features
1431 *
1432 *      @wanted_features:       User-requested features
1433 *      @vlan_features:         Mask of features inheritable by VLAN devices
1434 *
1435 *      @hw_enc_features:       Mask of features inherited by encapsulating devices
1436 *                              This field indicates what encapsulation
1437 *                              offloads the hardware is capable of doing,
1438 *                              and drivers will need to set them appropriately.
1439 *
1440 *      @mpls_features: Mask of features inheritable by MPLS
1441 *
1442 *      @ifindex:       interface index
1443 *      @group:         The group the device belongs to
1444 *
1445 *      @stats:         Statistics struct, which was left as a legacy, use
1446 *                      rtnl_link_stats64 instead
1447 *
1448 *      @rx_dropped:    Dropped packets by core network,
1449 *                      do not use this in drivers
1450 *      @tx_dropped:    Dropped packets by core network,
1451 *                      do not use this in drivers
1452 *      @rx_nohandler:  nohandler dropped packets by core network on
1453 *                      inactive devices, do not use this in drivers
1454 *
1455 *      @wireless_handlers:     List of functions to handle Wireless Extensions,
1456 *                              instead of ioctl,
1457 *                              see <net/iw_handler.h> for details.
1458 *      @wireless_data: Instance data managed by the core of wireless extensions
1459 *
1460 *      @netdev_ops:    Includes several pointers to callbacks,
1461 *                      if one wants to override the ndo_*() functions
1462 *      @ethtool_ops:   Management operations
1463 *      @header_ops:    Includes callbacks for creating,parsing,caching,etc
1464 *                      of Layer 2 headers.
1465 *
1466 *      @flags:         Interface flags (a la BSD)
1467 *      @priv_flags:    Like 'flags' but invisible to userspace,
1468 *                      see if.h for the definitions
1469 *      @gflags:        Global flags ( kept as legacy )
1470 *      @padded:        How much padding added by alloc_netdev()
1471 *      @operstate:     RFC2863 operstate
1472 *      @link_mode:     Mapping policy to operstate
1473 *      @if_port:       Selectable AUI, TP, ...
1474 *      @dma:           DMA channel
1475 *      @mtu:           Interface MTU value
1476 *      @type:          Interface hardware type
1477 *      @hard_header_len: Maximum hardware header length.
1478 *
1479 *      @needed_headroom: Extra headroom the hardware may need, but not in all
1480 *                        cases can this be guaranteed
1481 *      @needed_tailroom: Extra tailroom the hardware may need, but not in all
1482 *                        cases can this be guaranteed. Some cases also use
1483 *                        LL_MAX_HEADER instead to allocate the skb
1484 *
1485 *      interface address info:
1486 *
1487 *      @perm_addr:             Permanent hw address
1488 *      @addr_assign_type:      Hw address assignment type
1489 *      @addr_len:              Hardware address length
1490 *      @neigh_priv_len;        Used in neigh_alloc(),
1491 *                              initialized only in atm/clip.c
1492 *      @dev_id:                Used to differentiate devices that share
1493 *                              the same link layer address
1494 *      @dev_port:              Used to differentiate devices that share
1495 *                              the same function
1496 *      @addr_list_lock:        XXX: need comments on this one
1497 *      @uc_promisc:            Counter that indicates promiscuous mode
1498 *                              has been enabled due to the need to listen to
1499 *                              additional unicast addresses in a device that
1500 *                              does not implement ndo_set_rx_mode()
1501 *      @uc:                    unicast mac addresses
1502 *      @mc:                    multicast mac addresses
1503 *      @dev_addrs:             list of device hw addresses
1504 *      @queues_kset:           Group of all Kobjects in the Tx and RX queues
1505 *      @promiscuity:           Number of times the NIC is told to work in
1506 *                              promiscuous mode; if it becomes 0 the NIC will
1507 *                              exit promiscuous mode
1508 *      @allmulti:              Counter, enables or disables allmulticast mode
1509 *
1510 *      @vlan_info:     VLAN info
1511 *      @dsa_ptr:       dsa specific data
1512 *      @tipc_ptr:      TIPC specific data
1513 *      @atalk_ptr:     AppleTalk link
1514 *      @ip_ptr:        IPv4 specific data
1515 *      @dn_ptr:        DECnet specific data
1516 *      @ip6_ptr:       IPv6 specific data
1517 *      @ax25_ptr:      AX.25 specific data
1518 *      @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1519 *
1520 *      @last_rx:       Time of last Rx
1521 *      @dev_addr:      Hw address (before bcast,
1522 *                      because most packets are unicast)
1523 *
1524 *      @_rx:                   Array of RX queues
1525 *      @num_rx_queues:         Number of RX queues
1526 *                              allocated at register_netdev() time
1527 *      @real_num_rx_queues:    Number of RX queues currently active in device
1528 *
1529 *      @rx_handler:            handler for received packets
1530 *      @rx_handler_data:       XXX: need comments on this one
1531 *      @ingress_queue:         XXX: need comments on this one
1532 *      @broadcast:             hw bcast address
1533 *
1534 *      @rx_cpu_rmap:   CPU reverse-mapping for RX completion interrupts,
1535 *                      indexed by RX queue number. Assigned by driver.
1536 *                      This must only be set if the ndo_rx_flow_steer
1537 *                      operation is defined
1538 *      @index_hlist:           Device index hash chain
1539 *
1540 *      @_tx:                   Array of TX queues
1541 *      @num_tx_queues:         Number of TX queues allocated at alloc_netdev_mq() time
1542 *      @real_num_tx_queues:    Number of TX queues currently active in device
1543 *      @qdisc:                 Root qdisc from userspace point of view
1544 *      @tx_queue_len:          Max frames per queue allowed
1545 *      @tx_global_lock:        XXX: need comments on this one
1546 *
1547 *      @xps_maps:      XXX: need comments on this one
1548 *
1549 *      @offload_fwd_mark:      Offload device fwding mark
1550 *
1551 *      @trans_start:           Time (in jiffies) of last Tx
1552 *      @watchdog_timeo:        Represents the timeout that is used by
1553 *                              the watchdog (see dev_watchdog())
1554 *      @watchdog_timer:        List of timers
1555 *
1556 *      @pcpu_refcnt:           Number of references to this device
1557 *      @todo_list:             Delayed register/unregister
1558 *      @link_watch_list:       XXX: need comments on this one
1559 *
1560 *      @reg_state:             Register/unregister state machine
1561 *      @dismantle:             Device is going to be freed
1562 *      @rtnl_link_state:       This enum represents the phases of creating
1563 *                              a new link
1564 *
1565 *      @destructor:            Called from unregister,
1566 *                              can be used to call free_netdev
1567 *      @npinfo:                XXX: need comments on this one
1568 *      @nd_net:                Network namespace this network device is inside
1569 *
1570 *      @ml_priv:       Mid-layer private
1571 *      @lstats:        Loopback statistics
1572 *      @tstats:        Tunnel statistics
1573 *      @dstats:        Dummy statistics
1574 *      @vstats:        Virtual ethernet statistics
1575 *
1576 *      @garp_port:     GARP
1577 *      @mrp_port:      MRP
1578 *
1579 *      @dev:           Class/net/name entry
1580 *      @sysfs_groups:  Space for optional device, statistics and wireless
1581 *                      sysfs groups
1582 *
1583 *      @sysfs_rx_queue_group:  Space for optional per-rx queue attributes
1584 *      @rtnl_link_ops: Rtnl_link_ops
1585 *
1586 *      @gso_max_size:  Maximum size of generic segmentation offload
1587 *      @gso_max_segs:  Maximum number of segments that can be passed to the
1588 *                      NIC for GSO
1589 *      @gso_min_segs:  Minimum number of segments that can be passed to the
1590 *                      NIC for GSO
1591 *
1592 *      @dcbnl_ops:     Data Center Bridging netlink ops
1593 *      @num_tc:        Number of traffic classes in the net device
1594 *      @tc_to_txq:     XXX: need comments on this one
1595 *      @prio_tc_map    XXX: need comments on this one
1596 *
1597 *      @fcoe_ddp_xid:  Max exchange id for FCoE LRO by ddp
1598 *
1599 *      @priomap:       XXX: need comments on this one
1600 *      @phydev:        Physical device may attach itself
1601 *                      for hardware timestamping
1602 *
1603 *      @qdisc_tx_busylock:     XXX: need comments on this one
1604 *
1605 *      @proto_down:    protocol port state information can be sent to the
1606 *                      switch driver and used to set the phys state of the
1607 *                      switch port.
1608 *
1609 *      FIXME: cleanup struct net_device such that network protocol info
1610 *      moves out.
1611 */
1612
1613struct net_device {
1614        char                    name[IFNAMSIZ];
1615        struct hlist_node       name_hlist;
1616        char                    *ifalias;
1617        /*
1618         *      I/O specific fields
1619         *      FIXME: Merge these and struct ifmap into one
1620         */
1621        unsigned long           mem_end;
1622        unsigned long           mem_start;
1623        unsigned long           base_addr;
1624        int                     irq;
1625
1626        atomic_t                carrier_changes;
1627
1628        /*
1629         *      Some hardware also needs these fields (state,dev_list,
1630         *      napi_list,unreg_list,close_list) but they are not
1631         *      part of the usual set specified in Space.c.
1632         */
1633
1634        unsigned long           state;
1635
1636        struct list_head        dev_list;
1637        struct list_head        napi_list;
1638        struct list_head        unreg_list;
1639        struct list_head        close_list;
1640        struct list_head        ptype_all;
1641        struct list_head        ptype_specific;
1642
1643        struct {
1644                struct list_head upper;
1645                struct list_head lower;
1646        } adj_list;
1647
1648        struct {
1649                struct list_head upper;
1650                struct list_head lower;
1651        } all_adj_list;
1652
1653        netdev_features_t       features;
1654        netdev_features_t       hw_features;
1655        netdev_features_t       wanted_features;
1656        netdev_features_t       vlan_features;
1657        netdev_features_t       hw_enc_features;
1658        netdev_features_t       mpls_features;
1659
1660        int                     ifindex;
1661        int                     group;
1662
1663        struct net_device_stats stats;
1664
1665        atomic_long_t           rx_dropped;
1666        atomic_long_t           tx_dropped;
1667        atomic_long_t           rx_nohandler;
1668
1669#ifdef CONFIG_WIRELESS_EXT
1670        const struct iw_handler_def *wireless_handlers;
1671        struct iw_public_data   *wireless_data;
1672#endif
1673        const struct net_device_ops *netdev_ops;
1674        const struct ethtool_ops *ethtool_ops;
1675#ifdef CONFIG_NET_SWITCHDEV
1676        const struct switchdev_ops *switchdev_ops;
1677#endif
1678#ifdef CONFIG_NET_L3_MASTER_DEV
1679        const struct l3mdev_ops *l3mdev_ops;
1680#endif
1681
1682        const struct header_ops *header_ops;
1683
1684        unsigned int            flags;
1685        unsigned int            priv_flags;
1686
1687        unsigned short          gflags;
1688        unsigned short          padded;
1689
1690        unsigned char           operstate;
1691        unsigned char           link_mode;
1692
1693        unsigned char           if_port;
1694        unsigned char           dma;
1695
1696        unsigned int            mtu;
1697        unsigned short          type;
1698        unsigned short          hard_header_len;
1699
1700        unsigned short          needed_headroom;
1701        unsigned short          needed_tailroom;
1702
1703        /* Interface address info. */
1704        unsigned char           perm_addr[MAX_ADDR_LEN];
1705        unsigned char           addr_assign_type;
1706        unsigned char           addr_len;
1707        unsigned short          neigh_priv_len;
1708        unsigned short          dev_id;
1709        unsigned short          dev_port;
1710        spinlock_t              addr_list_lock;
1711        unsigned char           name_assign_type;
1712        bool                    uc_promisc;
1713        struct netdev_hw_addr_list      uc;
1714        struct netdev_hw_addr_list      mc;
1715        struct netdev_hw_addr_list      dev_addrs;
1716
1717#ifdef CONFIG_SYSFS
1718        struct kset             *queues_kset;
1719#endif
1720        unsigned int            promiscuity;
1721        unsigned int            allmulti;
1722
1723
1724        /* Protocol-specific pointers */
1725
1726#if IS_ENABLED(CONFIG_VLAN_8021Q)
1727        struct vlan_info __rcu  *vlan_info;
1728#endif
1729#if IS_ENABLED(CONFIG_NET_DSA)
1730        struct dsa_switch_tree  *dsa_ptr;
1731#endif
1732#if IS_ENABLED(CONFIG_TIPC)
1733        struct tipc_bearer __rcu *tipc_ptr;
1734#endif
1735        void                    *atalk_ptr;
1736        struct in_device __rcu  *ip_ptr;
1737        struct dn_dev __rcu     *dn_ptr;
1738        struct inet6_dev __rcu  *ip6_ptr;
1739        void                    *ax25_ptr;
1740        struct wireless_dev     *ieee80211_ptr;
1741        struct wpan_dev         *ieee802154_ptr;
1742#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1743        struct mpls_dev __rcu   *mpls_ptr;
1744#endif
1745
1746/*
1747 * Cache lines mostly used on receive path (including eth_type_trans())
1748 */
1749        unsigned long           last_rx;
1750
1751        /* Interface address info used in eth_type_trans() */
1752        unsigned char           *dev_addr;
1753
1754#ifdef CONFIG_SYSFS
1755        struct netdev_rx_queue  *_rx;
1756
1757        unsigned int            num_rx_queues;
1758        unsigned int            real_num_rx_queues;
1759#endif
1760
1761        unsigned long           gro_flush_timeout;
1762        rx_handler_func_t __rcu *rx_handler;
1763        void __rcu              *rx_handler_data;
1764
1765#ifdef CONFIG_NET_CLS_ACT
1766        struct tcf_proto __rcu  *ingress_cl_list;
1767#endif
1768        struct netdev_queue __rcu *ingress_queue;
1769#ifdef CONFIG_NETFILTER_INGRESS
1770        struct list_head        nf_hooks_ingress;
1771#endif
1772
1773        unsigned char           broadcast[MAX_ADDR_LEN];
1774#ifdef CONFIG_RFS_ACCEL
1775        struct cpu_rmap         *rx_cpu_rmap;
1776#endif
1777        struct hlist_node       index_hlist;
1778
1779/*
1780 * Cache lines mostly used on transmit path
1781 */
1782        struct netdev_queue     *_tx ____cacheline_aligned_in_smp;
1783        unsigned int            num_tx_queues;
1784        unsigned int            real_num_tx_queues;
1785        struct Qdisc            *qdisc;
1786        unsigned long           tx_queue_len;
1787        spinlock_t              tx_global_lock;
1788        int                     watchdog_timeo;
1789
1790#ifdef CONFIG_XPS
1791        struct xps_dev_maps __rcu *xps_maps;
1792#endif
1793#ifdef CONFIG_NET_CLS_ACT
1794        struct tcf_proto __rcu  *egress_cl_list;
1795#endif
1796#ifdef CONFIG_NET_SWITCHDEV
1797        u32                     offload_fwd_mark;
1798#endif
1799
1800        /* These may be needed for future network-power-down code. */
1801
1802        /*
1803         * trans_start here is expensive for high speed devices on SMP,
1804         * please use netdev_queue->trans_start instead.
1805         */
1806        unsigned long           trans_start;
1807
1808        struct timer_list       watchdog_timer;
1809
1810        int __percpu            *pcpu_refcnt;
1811        struct list_head        todo_list;
1812
1813        struct list_head        link_watch_list;
1814
1815        enum { NETREG_UNINITIALIZED=0,
1816               NETREG_REGISTERED,       /* completed register_netdevice */
1817               NETREG_UNREGISTERING,    /* called unregister_netdevice */
1818               NETREG_UNREGISTERED,     /* completed unregister todo */
1819               NETREG_RELEASED,         /* called free_netdev */
1820               NETREG_DUMMY,            /* dummy device for NAPI poll */
1821        } reg_state:8;
1822
1823        bool dismantle;
1824
1825        enum {
1826                RTNL_LINK_INITIALIZED,
1827                RTNL_LINK_INITIALIZING,
1828        } rtnl_link_state:16;
1829
1830        void (*destructor)(struct net_device *dev);
1831
1832#ifdef CONFIG_NETPOLL
1833        struct netpoll_info __rcu       *npinfo;
1834#endif
1835
1836        possible_net_t                  nd_net;
1837
1838        /* mid-layer private */
1839        union {
1840                void                                    *ml_priv;
1841                struct pcpu_lstats __percpu             *lstats;
1842                struct pcpu_sw_netstats __percpu        *tstats;
1843                struct pcpu_dstats __percpu             *dstats;
1844                struct pcpu_vstats __percpu             *vstats;
1845        };
1846
1847        struct garp_port __rcu  *garp_port;
1848        struct mrp_port __rcu   *mrp_port;
1849
1850        struct device           dev;
1851        const struct attribute_group *sysfs_groups[4];
1852        const struct attribute_group *sysfs_rx_queue_group;
1853
1854        const struct rtnl_link_ops *rtnl_link_ops;
1855
1856        /* for setting kernel sock attribute on TCP connection setup */
1857#define GSO_MAX_SIZE            65536
1858        unsigned int            gso_max_size;
1859#define GSO_MAX_SEGS            65535
1860        u16                     gso_max_segs;
1861        u16                     gso_min_segs;
1862#ifdef CONFIG_DCB
1863        const struct dcbnl_rtnl_ops *dcbnl_ops;
1864#endif
1865        u8                      num_tc;
1866        struct netdev_tc_txq    tc_to_txq[TC_MAX_QUEUE];
1867        u8                      prio_tc_map[TC_BITMASK + 1];
1868
1869#if IS_ENABLED(CONFIG_FCOE)
1870        unsigned int            fcoe_ddp_xid;
1871#endif
1872#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1873        struct netprio_map __rcu *priomap;
1874#endif
1875        struct phy_device       *phydev;
1876        struct lock_class_key   *qdisc_tx_busylock;
1877        bool                    proto_down;
1878};
1879#define to_net_dev(d) container_of(d, struct net_device, dev)
1880
1881#define NETDEV_ALIGN            32
1882
1883static inline
1884int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1885{
1886        return dev->prio_tc_map[prio & TC_BITMASK];
1887}
1888
1889static inline
1890int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1891{
1892        if (tc >= dev->num_tc)
1893                return -EINVAL;
1894
1895        dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1896        return 0;
1897}
1898
1899static inline
1900void netdev_reset_tc(struct net_device *dev)
1901{
1902        dev->num_tc = 0;
1903        memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1904        memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1905}
1906
1907static inline
1908int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1909{
1910        if (tc >= dev->num_tc)
1911                return -EINVAL;
1912
1913        dev->tc_to_txq[tc].count = count;
1914        dev->tc_to_txq[tc].offset = offset;
1915        return 0;
1916}
1917
1918static inline
1919int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1920{
1921        if (num_tc > TC_MAX_QUEUE)
1922                return -EINVAL;
1923
1924        dev->num_tc = num_tc;
1925        return 0;
1926}
1927
1928static inline
1929int netdev_get_num_tc(struct net_device *dev)
1930{
1931        return dev->num_tc;
1932}
1933
1934static inline
1935struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1936                                         unsigned int index)
1937{
1938        return &dev->_tx[index];
1939}
1940
1941static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1942                                                    const struct sk_buff *skb)
1943{
1944        return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1945}
1946
1947static inline void netdev_for_each_tx_queue(struct net_device *dev,
1948                                            void (*f)(struct net_device *,
1949                                                      struct netdev_queue *,
1950                                                      void *),
1951                                            void *arg)
1952{
1953        unsigned int i;
1954
1955        for (i = 0; i < dev->num_tx_queues; i++)
1956                f(dev, &dev->_tx[i], arg);
1957}
1958
1959struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1960                                    struct sk_buff *skb,
1961                                    void *accel_priv);
1962
1963/* returns the headroom that the master device needs to take in account
1964 * when forwarding to this dev
1965 */
1966static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1967{
1968        return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1969}
1970
1971static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1972{
1973        if (dev->netdev_ops->ndo_set_rx_headroom)
1974                dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1975}
1976
1977/* set the device rx headroom to the dev's default */
1978static inline void netdev_reset_rx_headroom(struct net_device *dev)
1979{
1980        netdev_set_rx_headroom(dev, -1);
1981}
1982
1983/*
1984 * Net namespace inlines
1985 */
1986static inline
1987struct net *dev_net(const struct net_device *dev)
1988{
1989        return read_pnet(&dev->nd_net);
1990}
1991
1992static inline
1993void dev_net_set(struct net_device *dev, struct net *net)
1994{
1995        write_pnet(&dev->nd_net, net);
1996}
1997
1998static inline bool netdev_uses_dsa(struct net_device *dev)
1999{
2000#if IS_ENABLED(CONFIG_NET_DSA)
2001        if (dev->dsa_ptr != NULL)
2002                return dsa_uses_tagged_protocol(dev->dsa_ptr);
2003#endif
2004        return false;
2005}
2006
2007/**
2008 *      netdev_priv - access network device private data
2009 *      @dev: network device
2010 *
2011 * Get network device private data
2012 */
2013static inline void *netdev_priv(const struct net_device *dev)
2014{
2015        return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2016}
2017
2018/* Set the sysfs physical device reference for the network logical device
2019 * if set prior to registration will cause a symlink during initialization.
2020 */
2021#define SET_NETDEV_DEV(net, pdev)       ((net)->dev.parent = (pdev))
2022
2023/* Set the sysfs device type for the network logical device to allow
2024 * fine-grained identification of different network device types. For
2025 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2026 */
2027#define SET_NETDEV_DEVTYPE(net, devtype)        ((net)->dev.type = (devtype))
2028
2029/* Default NAPI poll() weight
2030 * Device drivers are strongly advised to not use bigger value
2031 */
2032#define NAPI_POLL_WEIGHT 64
2033
2034/**
2035 *      netif_napi_add - initialize a NAPI context
2036 *      @dev:  network device
2037 *      @napi: NAPI context
2038 *      @poll: polling function
2039 *      @weight: default weight
2040 *
2041 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2042 * *any* of the other NAPI-related functions.
2043 */
2044void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2045                    int (*poll)(struct napi_struct *, int), int weight);
2046
2047/**
2048 *      netif_tx_napi_add - initialize a NAPI context
2049 *      @dev:  network device
2050 *      @napi: NAPI context
2051 *      @poll: polling function
2052 *      @weight: default weight
2053 *
2054 * This variant of netif_napi_add() should be used from drivers using NAPI
2055 * to exclusively poll a TX queue.
2056 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2057 */
2058static inline void netif_tx_napi_add(struct net_device *dev,
2059                                     struct napi_struct *napi,
2060                                     int (*poll)(struct napi_struct *, int),
2061                                     int weight)
2062{
2063        set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2064        netif_napi_add(dev, napi, poll, weight);
2065}
2066
2067/**
2068 *  netif_napi_del - remove a NAPI context
2069 *  @napi: NAPI context
2070 *
2071 *  netif_napi_del() removes a NAPI context from the network device NAPI list
2072 */
2073void netif_napi_del(struct napi_struct *napi);
2074
2075struct napi_gro_cb {
2076        /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2077        void    *frag0;
2078
2079        /* Length of frag0. */
2080        unsigned int frag0_len;
2081
2082        /* This indicates where we are processing relative to skb->data. */
2083        int     data_offset;
2084
2085        /* This is non-zero if the packet cannot be merged with the new skb. */
2086        u16     flush;
2087
2088        /* Save the IP ID here and check when we get to the transport layer */
2089        u16     flush_id;
2090
2091        /* Number of segments aggregated. */
2092        u16     count;
2093
2094        /* Start offset for remote checksum offload */
2095        u16     gro_remcsum_start;
2096
2097        /* jiffies when first packet was created/queued */
2098        unsigned long age;
2099
2100        /* Used in ipv6_gro_receive() and foo-over-udp */
2101        u16     proto;
2102
2103        /* This is non-zero if the packet may be of the same flow. */
2104        u8      same_flow:1;
2105
2106        /* Used in tunnel GRO receive */
2107        u8      encap_mark:1;
2108
2109        /* GRO checksum is valid */
2110        u8      csum_valid:1;
2111
2112        /* Number of checksums via CHECKSUM_UNNECESSARY */
2113        u8      csum_cnt:3;
2114
2115        /* Free the skb? */
2116        u8      free:2;
2117#define NAPI_GRO_FREE             1
2118#define NAPI_GRO_FREE_STOLEN_HEAD 2
2119
2120        /* Used in foo-over-udp, set in udp[46]_gro_receive */
2121        u8      is_ipv6:1;
2122
2123        /* Used in GRE, set in fou/gue_gro_receive */
2124        u8      is_fou:1;
2125
2126        /* 6 bit hole */
2127
2128        /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2129        __wsum  csum;
2130
2131        /* used in skb_gro_receive() slow path */
2132        struct sk_buff *last;
2133};
2134
2135#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2136
2137struct packet_type {
2138        __be16                  type;   /* This is really htons(ether_type). */
2139        struct net_device       *dev;   /* NULL is wildcarded here           */
2140        int                     (*func) (struct sk_buff *,
2141                                         struct net_device *,
2142                                         struct packet_type *,
2143                                         struct net_device *);
2144        bool                    (*id_match)(struct packet_type *ptype,
2145                                            struct sock *sk);
2146        void                    *af_packet_priv;
2147        struct list_head        list;
2148};
2149
2150struct offload_callbacks {
2151        struct sk_buff          *(*gso_segment)(struct sk_buff *skb,
2152                                                netdev_features_t features);
2153        struct sk_buff          **(*gro_receive)(struct sk_buff **head,
2154                                                 struct sk_buff *skb);
2155        int                     (*gro_complete)(struct sk_buff *skb, int nhoff);
2156};
2157
2158struct packet_offload {
2159        __be16                   type;  /* This is really htons(ether_type). */
2160        u16                      priority;
2161        struct offload_callbacks callbacks;
2162        struct list_head         list;
2163};
2164
2165struct udp_offload;
2166
2167/* 'skb->encapsulation' is set before gro_complete() is called.  gro_complete()
2168 * must set 'skb->inner_mac_header' to the beginning of tunnel payload.
2169 */
2170struct udp_offload_callbacks {
2171        struct sk_buff          **(*gro_receive)(struct sk_buff **head,
2172                                                 struct sk_buff *skb,
2173                                                 struct udp_offload *uoff);
2174        int                     (*gro_complete)(struct sk_buff *skb,
2175                                                int nhoff,
2176                                                struct udp_offload *uoff);
2177};
2178
2179struct udp_offload {
2180        __be16                   port;
2181        u8                       ipproto;
2182        struct udp_offload_callbacks callbacks;
2183};
2184
2185/* often modified stats are per-CPU, other are shared (netdev->stats) */
2186struct pcpu_sw_netstats {
2187        u64     rx_packets;
2188        u64     rx_bytes;
2189        u64     tx_packets;
2190        u64     tx_bytes;
2191        struct u64_stats_sync   syncp;
2192};
2193
2194#define __netdev_alloc_pcpu_stats(type, gfp)                            \
2195({                                                                      \
2196        typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2197        if (pcpu_stats) {                                               \
2198                int __cpu;                                              \
2199                for_each_possible_cpu(__cpu) {                          \
2200                        typeof(type) *stat;                             \
2201                        stat = per_cpu_ptr(pcpu_stats, __cpu);          \
2202                        u64_stats_init(&stat->syncp);                   \
2203                }                                                       \
2204        }                                                               \
2205        pcpu_stats;                                                     \
2206})
2207
2208#define netdev_alloc_pcpu_stats(type)                                   \
2209        __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2210
2211enum netdev_lag_tx_type {
2212        NETDEV_LAG_TX_TYPE_UNKNOWN,
2213        NETDEV_LAG_TX_TYPE_RANDOM,
2214        NETDEV_LAG_TX_TYPE_BROADCAST,
2215        NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2216        NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2217        NETDEV_LAG_TX_TYPE_HASH,
2218};
2219
2220struct netdev_lag_upper_info {
2221        enum netdev_lag_tx_type tx_type;
2222};
2223
2224struct netdev_lag_lower_state_info {
2225        u8 link_up : 1,
2226           tx_enabled : 1;
2227};
2228
2229#include <linux/notifier.h>
2230
2231/* netdevice notifier chain. Please remember to update the rtnetlink
2232 * notification exclusion list in rtnetlink_event() when adding new
2233 * types.
2234 */
2235#define NETDEV_UP       0x0001  /* For now you can't veto a device up/down */
2236#define NETDEV_DOWN     0x0002
2237#define NETDEV_REBOOT   0x0003  /* Tell a protocol stack a network interface
2238                                   detected a hardware crash and restarted
2239                                   - we can use this eg to kick tcp sessions
2240                                   once done */
2241#define NETDEV_CHANGE   0x0004  /* Notify device state change */
2242#define NETDEV_REGISTER 0x0005
2243#define NETDEV_UNREGISTER       0x0006
2244#define NETDEV_CHANGEMTU        0x0007 /* notify after mtu change happened */
2245#define NETDEV_CHANGEADDR       0x0008
2246#define NETDEV_GOING_DOWN       0x0009
2247#define NETDEV_CHANGENAME       0x000A
2248#define NETDEV_FEAT_CHANGE      0x000B
2249#define NETDEV_BONDING_FAILOVER 0x000C
2250#define NETDEV_PRE_UP           0x000D
2251#define NETDEV_PRE_TYPE_CHANGE  0x000E
2252#define NETDEV_POST_TYPE_CHANGE 0x000F
2253#define NETDEV_POST_INIT        0x0010
2254#define NETDEV_UNREGISTER_FINAL 0x0011
2255#define NETDEV_RELEASE          0x0012
2256#define NETDEV_NOTIFY_PEERS     0x0013
2257#define NETDEV_JOIN             0x0014
2258#define NETDEV_CHANGEUPPER      0x0015
2259#define NETDEV_RESEND_IGMP      0x0016
2260#define NETDEV_PRECHANGEMTU     0x0017 /* notify before mtu change happened */
2261#define NETDEV_CHANGEINFODATA   0x0018
2262#define NETDEV_BONDING_INFO     0x0019
2263#define NETDEV_PRECHANGEUPPER   0x001A
2264#define NETDEV_CHANGELOWERSTATE 0x001B
2265
2266int register_netdevice_notifier(struct notifier_block *nb);
2267int unregister_netdevice_notifier(struct notifier_block *nb);
2268
2269struct netdev_notifier_info {
2270        struct net_device *dev;
2271};
2272
2273struct netdev_notifier_change_info {
2274        struct netdev_notifier_info info; /* must be first */
2275        unsigned int flags_changed;
2276};
2277
2278struct netdev_notifier_changeupper_info {
2279        struct netdev_notifier_info info; /* must be first */
2280        struct net_device *upper_dev; /* new upper dev */
2281        bool master; /* is upper dev master */
2282        bool linking; /* is the notification for link or unlink */
2283        void *upper_info; /* upper dev info */
2284};
2285
2286struct netdev_notifier_changelowerstate_info {
2287        struct netdev_notifier_info info; /* must be first */
2288        void *lower_state_info; /* is lower dev state */
2289};
2290
2291static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2292                                             struct net_device *dev)
2293{
2294        info->dev = dev;
2295}
2296
2297static inline struct net_device *
2298netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2299{
2300        return info->dev;
2301}
2302
2303int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2304
2305
2306extern rwlock_t                         dev_base_lock;          /* Device list lock */
2307
2308#define for_each_netdev(net, d)         \
2309                list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2310#define for_each_netdev_reverse(net, d) \
2311                list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2312#define for_each_netdev_rcu(net, d)             \
2313                list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2314#define for_each_netdev_safe(net, d, n) \
2315                list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2316#define for_each_netdev_continue(net, d)                \
2317                list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2318#define for_each_netdev_continue_rcu(net, d)            \
2319        list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2320#define for_each_netdev_in_bond_rcu(bond, slave)        \
2321                for_each_netdev_rcu(&init_net, slave)   \
2322                        if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2323#define net_device_entry(lh)    list_entry(lh, struct net_device, dev_list)
2324
2325static inline struct net_device *next_net_device(struct net_device *dev)
2326{
2327        struct list_head *lh;
2328        struct net *net;
2329
2330        net = dev_net(dev);
2331        lh = dev->dev_list.next;
2332        return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2333}
2334
2335static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2336{
2337        struct list_head *lh;
2338        struct net *net;
2339
2340        net = dev_net(dev);
2341        lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2342        return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2343}
2344
2345static inline struct net_device *first_net_device(struct net *net)
2346{
2347        return list_empty(&net->dev_base_head) ? NULL :
2348                net_device_entry(net->dev_base_head.next);
2349}
2350
2351static inline struct net_device *first_net_device_rcu(struct net *net)
2352{
2353        struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2354
2355        return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2356}
2357
2358int netdev_boot_setup_check(struct net_device *dev);
2359unsigned long netdev_boot_base(const char *prefix, int unit);
2360struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2361                                       const char *hwaddr);
2362struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2363struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2364void dev_add_pack(struct packet_type *pt);
2365void dev_remove_pack(struct packet_type *pt);
2366void __dev_remove_pack(struct packet_type *pt);
2367void dev_add_offload(struct packet_offload *po);
2368void dev_remove_offload(struct packet_offload *po);
2369
2370int dev_get_iflink(const struct net_device *dev);
2371int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2372struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2373                                      unsigned short mask);
2374struct net_device *dev_get_by_name(struct net *net, const char *name);
2375struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2376struct net_device *__dev_get_by_name(struct net *net, const char *name);
2377int dev_alloc_name(struct net_device *dev, const char *name);
2378int dev_open(struct net_device *dev);
2379int dev_close(struct net_device *dev);
2380int dev_close_many(struct list_head *head, bool unlink);
2381void dev_disable_lro(struct net_device *dev);
2382int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2383int dev_queue_xmit(struct sk_buff *skb);
2384int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2385int register_netdevice(struct net_device *dev);
2386void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2387void unregister_netdevice_many(struct list_head *head);
2388static inline void unregister_netdevice(struct net_device *dev)
2389{
2390        unregister_netdevice_queue(dev, NULL);
2391}
2392
2393int netdev_refcnt_read(const struct net_device *dev);
2394void free_netdev(struct net_device *dev);
2395void netdev_freemem(struct net_device *dev);
2396void synchronize_net(void);
2397int init_dummy_netdev(struct net_device *dev);
2398
2399DECLARE_PER_CPU(int, xmit_recursion);
2400static inline int dev_recursion_level(void)
2401{
2402        return this_cpu_read(xmit_recursion);
2403}
2404
2405struct net_device *dev_get_by_index(struct net *net, int ifindex);
2406struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2407struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2408int netdev_get_name(struct net *net, char *name, int ifindex);
2409int dev_restart(struct net_device *dev);
2410int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2411
2412static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2413{
2414        return NAPI_GRO_CB(skb)->data_offset;
2415}
2416
2417static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2418{
2419        return skb->len - NAPI_GRO_CB(skb)->data_offset;
2420}
2421
2422static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2423{
2424        NAPI_GRO_CB(skb)->data_offset += len;
2425}
2426
2427static inline void *skb_gro_header_fast(struct sk_buff *skb,
2428                                        unsigned int offset)
2429{
2430        return NAPI_GRO_CB(skb)->frag0 + offset;
2431}
2432
2433static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2434{
2435        return NAPI_GRO_CB(skb)->frag0_len < hlen;
2436}
2437
2438static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2439                                        unsigned int offset)
2440{
2441        if (!pskb_may_pull(skb, hlen))
2442                return NULL;
2443
2444        NAPI_GRO_CB(skb)->frag0 = NULL;
2445        NAPI_GRO_CB(skb)->frag0_len = 0;
2446        return skb->data + offset;
2447}
2448
2449static inline void *skb_gro_network_header(struct sk_buff *skb)
2450{
2451        return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2452               skb_network_offset(skb);
2453}
2454
2455static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2456                                        const void *start, unsigned int len)
2457{
2458        if (NAPI_GRO_CB(skb)->csum_valid)
2459                NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2460                                                  csum_partial(start, len, 0));
2461}
2462
2463/* GRO checksum functions. These are logical equivalents of the normal
2464 * checksum functions (in skbuff.h) except that they operate on the GRO
2465 * offsets and fields in sk_buff.
2466 */
2467
2468__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2469
2470static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2471{
2472        return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2473}
2474
2475static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2476                                                      bool zero_okay,
2477                                                      __sum16 check)
2478{
2479        return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2480                skb_checksum_start_offset(skb) <
2481                 skb_gro_offset(skb)) &&
2482                !skb_at_gro_remcsum_start(skb) &&
2483                NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2484                (!zero_okay || check));
2485}
2486
2487static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2488                                                           __wsum psum)
2489{
2490        if (NAPI_GRO_CB(skb)->csum_valid &&
2491            !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2492                return 0;
2493
2494        NAPI_GRO_CB(skb)->csum = psum;
2495
2496        return __skb_gro_checksum_complete(skb);
2497}
2498
2499static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2500{
2501        if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2502                /* Consume a checksum from CHECKSUM_UNNECESSARY */
2503                NAPI_GRO_CB(skb)->csum_cnt--;
2504        } else {
2505                /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2506                 * verified a new top level checksum or an encapsulated one
2507                 * during GRO. This saves work if we fallback to normal path.
2508                 */
2509                __skb_incr_checksum_unnecessary(skb);
2510        }
2511}
2512
2513#define __skb_gro_checksum_validate(skb, proto, zero_okay, check,       \
2514                                    compute_pseudo)                     \
2515({                                                                      \
2516        __sum16 __ret = 0;                                              \
2517        if (__skb_gro_checksum_validate_needed(skb, zero_okay, check))  \
2518                __ret = __skb_gro_checksum_validate_complete(skb,       \
2519                                compute_pseudo(skb, proto));            \
2520        if (__ret)                                                      \
2521                __skb_mark_checksum_bad(skb);                           \
2522        else                                                            \
2523                skb_gro_incr_csum_unnecessary(skb);                     \
2524        __ret;                                                          \
2525})
2526
2527#define skb_gro_checksum_validate(skb, proto, compute_pseudo)           \
2528        __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2529
2530#define skb_gro_checksum_validate_zero_check(skb, proto, check,         \
2531                                             compute_pseudo)            \
2532        __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2533
2534#define skb_gro_checksum_simple_validate(skb)                           \
2535        __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2536
2537static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2538{
2539        return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2540                !NAPI_GRO_CB(skb)->csum_valid);
2541}
2542
2543static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2544                                              __sum16 check, __wsum pseudo)
2545{
2546        NAPI_GRO_CB(skb)->csum = ~pseudo;
2547        NAPI_GRO_CB(skb)->csum_valid = 1;
2548}
2549
2550#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2551do {                                                                    \
2552        if (__skb_gro_checksum_convert_check(skb))                      \
2553                __skb_gro_checksum_convert(skb, check,                  \
2554                                           compute_pseudo(skb, proto)); \
2555} while (0)
2556
2557struct gro_remcsum {
2558        int offset;
2559        __wsum delta;
2560};
2561
2562static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2563{
2564        grc->offset = 0;
2565        grc->delta = 0;
2566}
2567
2568static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2569                                            unsigned int off, size_t hdrlen,
2570                                            int start, int offset,
2571                                            struct gro_remcsum *grc,
2572                                            bool nopartial)
2573{
2574        __wsum delta;
2575        size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2576
2577        BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2578
2579        if (!nopartial) {
2580                NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2581                return ptr;
2582        }
2583
2584        ptr = skb_gro_header_fast(skb, off);
2585        if (skb_gro_header_hard(skb, off + plen)) {
2586                ptr = skb_gro_header_slow(skb, off + plen, off);
2587                if (!ptr)
2588                        return NULL;
2589        }
2590
2591        delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2592                               start, offset);
2593
2594        /* Adjust skb->csum since we changed the packet */
2595        NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2596
2597        grc->offset = off + hdrlen + offset;
2598        grc->delta = delta;
2599
2600        return ptr;
2601}
2602
2603static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2604                                           struct gro_remcsum *grc)
2605{
2606        void *ptr;
2607        size_t plen = grc->offset + sizeof(u16);
2608
2609        if (!grc->delta)
2610                return;
2611
2612        ptr = skb_gro_header_fast(skb, grc->offset);
2613        if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2614                ptr = skb_gro_header_slow(skb, plen, grc->offset);
2615                if (!ptr)
2616                        return;
2617        }
2618
2619        remcsum_unadjust((__sum16 *)ptr, grc->delta);
2620}
2621
2622struct skb_csum_offl_spec {
2623        __u16           ipv4_okay:1,
2624                        ipv6_okay:1,
2625                        encap_okay:1,
2626                        ip_options_okay:1,
2627                        ext_hdrs_okay:1,
2628                        tcp_okay:1,
2629                        udp_okay:1,
2630                        sctp_okay:1,
2631                        vlan_okay:1,
2632                        no_encapped_ipv6:1,
2633                        no_not_encapped:1;
2634};
2635
2636bool __skb_csum_offload_chk(struct sk_buff *skb,
2637                            const struct skb_csum_offl_spec *spec,
2638                            bool *csum_encapped,
2639                            bool csum_help);
2640
2641static inline bool skb_csum_offload_chk(struct sk_buff *skb,
2642                                        const struct skb_csum_offl_spec *spec,
2643                                        bool *csum_encapped,
2644                                        bool csum_help)
2645{
2646        if (skb->ip_summed != CHECKSUM_PARTIAL)
2647                return false;
2648
2649        return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help);
2650}
2651
2652static inline bool skb_csum_offload_chk_help(struct sk_buff *skb,
2653                                             const struct skb_csum_offl_spec *spec)
2654{
2655        bool csum_encapped;
2656
2657        return skb_csum_offload_chk(skb, spec, &csum_encapped, true);
2658}
2659
2660static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb)
2661{
2662        static const struct skb_csum_offl_spec csum_offl_spec = {
2663                .ipv4_okay = 1,
2664                .ip_options_okay = 1,
2665                .ipv6_okay = 1,
2666                .vlan_okay = 1,
2667                .tcp_okay = 1,
2668                .udp_okay = 1,
2669        };
2670
2671        return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2672}
2673
2674static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb)
2675{
2676        static const struct skb_csum_offl_spec csum_offl_spec = {
2677                .ipv4_okay = 1,
2678                .ip_options_okay = 1,
2679                .tcp_okay = 1,
2680                .udp_okay = 1,
2681                .vlan_okay = 1,
2682        };
2683
2684        return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2685}
2686
2687static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2688                                  unsigned short type,
2689                                  const void *daddr, const void *saddr,
2690                                  unsigned int len)
2691{
2692        if (!dev->header_ops || !dev->header_ops->create)
2693                return 0;
2694
2695        return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2696}
2697
2698static inline int dev_parse_header(const struct sk_buff *skb,
2699                                   unsigned char *haddr)
2700{
2701        const struct net_device *dev = skb->dev;
2702
2703        if (!dev->header_ops || !dev->header_ops->parse)
2704                return 0;
2705        return dev->header_ops->parse(skb, haddr);
2706}
2707
2708/* ll_header must have at least hard_header_len allocated */
2709static inline bool dev_validate_header(const struct net_device *dev,
2710                                       char *ll_header, int len)
2711{
2712        if (likely(len >= dev->hard_header_len))
2713                return true;
2714
2715        if (capable(CAP_SYS_RAWIO)) {
2716                memset(ll_header + len, 0, dev->hard_header_len - len);
2717                return true;
2718        }
2719
2720        if (dev->header_ops && dev->header_ops->validate)
2721                return dev->header_ops->validate(ll_header, len);
2722
2723        return false;
2724}
2725
2726typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2727int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2728static inline int unregister_gifconf(unsigned int family)
2729{
2730        return register_gifconf(family, NULL);
2731}
2732
2733#ifdef CONFIG_NET_FLOW_LIMIT
2734#define FLOW_LIMIT_HISTORY      (1 << 7)  /* must be ^2 and !overflow buckets */
2735struct sd_flow_limit {
2736        u64                     count;
2737        unsigned int            num_buckets;
2738        unsigned int            history_head;
2739        u16                     history[FLOW_LIMIT_HISTORY];
2740        u8                      buckets[];
2741};
2742
2743extern int netdev_flow_limit_table_len;
2744#endif /* CONFIG_NET_FLOW_LIMIT */
2745
2746/*
2747 * Incoming packets are placed on per-CPU queues
2748 */
2749struct softnet_data {
2750        struct list_head        poll_list;
2751        struct sk_buff_head     process_queue;
2752
2753        /* stats */
2754        unsigned int            processed;
2755        unsigned int            time_squeeze;
2756        unsigned int            cpu_collision;
2757        unsigned int            received_rps;
2758#ifdef CONFIG_RPS
2759        struct softnet_data     *rps_ipi_list;
2760#endif
2761#ifdef CONFIG_NET_FLOW_LIMIT
2762        struct sd_flow_limit __rcu *flow_limit;
2763#endif
2764        struct Qdisc            *output_queue;
2765        struct Qdisc            **output_queue_tailp;
2766        struct sk_buff          *completion_queue;
2767
2768#ifdef CONFIG_RPS
2769        /* Elements below can be accessed between CPUs for RPS */
2770        struct call_single_data csd ____cacheline_aligned_in_smp;
2771        struct softnet_data     *rps_ipi_next;
2772        unsigned int            cpu;
2773        unsigned int            input_queue_head;
2774        unsigned int            input_queue_tail;
2775#endif
2776        unsigned int            dropped;
2777        struct sk_buff_head     input_pkt_queue;
2778        struct napi_struct      backlog;
2779
2780};
2781
2782static inline void input_queue_head_incr(struct softnet_data *sd)
2783{
2784#ifdef CONFIG_RPS
2785        sd->input_queue_head++;
2786#endif
2787}
2788
2789static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2790                                              unsigned int *qtail)
2791{
2792#ifdef CONFIG_RPS
2793        *qtail = ++sd->input_queue_tail;
2794#endif
2795}
2796
2797DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2798
2799void __netif_schedule(struct Qdisc *q);
2800void netif_schedule_queue(struct netdev_queue *txq);
2801
2802static inline void netif_tx_schedule_all(struct net_device *dev)
2803{
2804        unsigned int i;
2805
2806        for (i = 0; i < dev->num_tx_queues; i++)
2807                netif_schedule_queue(netdev_get_tx_queue(dev, i));
2808}
2809
2810static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2811{
2812        clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2813}
2814
2815/**
2816 *      netif_start_queue - allow transmit
2817 *      @dev: network device
2818 *
2819 *      Allow upper layers to call the device hard_start_xmit routine.
2820 */
2821static inline void netif_start_queue(struct net_device *dev)
2822{
2823        netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2824}
2825
2826static inline void netif_tx_start_all_queues(struct net_device *dev)
2827{
2828        unsigned int i;
2829
2830        for (i = 0; i < dev->num_tx_queues; i++) {
2831                struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2832                netif_tx_start_queue(txq);
2833        }
2834}
2835
2836void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2837
2838/**
2839 *      netif_wake_queue - restart transmit
2840 *      @dev: network device
2841 *
2842 *      Allow upper layers to call the device hard_start_xmit routine.
2843 *      Used for flow control when transmit resources are available.
2844 */
2845static inline void netif_wake_queue(struct net_device *dev)
2846{
2847        netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2848}
2849
2850static inline void netif_tx_wake_all_queues(struct net_device *dev)
2851{
2852        unsigned int i;
2853
2854        for (i = 0; i < dev->num_tx_queues; i++) {
2855                struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2856                netif_tx_wake_queue(txq);
2857        }
2858}
2859
2860static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2861{
2862        set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2863}
2864
2865/**
2866 *      netif_stop_queue - stop transmitted packets
2867 *      @dev: network device
2868 *
2869 *      Stop upper layers calling the device hard_start_xmit routine.
2870 *      Used for flow control when transmit resources are unavailable.
2871 */
2872static inline void netif_stop_queue(struct net_device *dev)
2873{
2874        netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2875}
2876
2877void netif_tx_stop_all_queues(struct net_device *dev);
2878
2879static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2880{
2881        return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2882}
2883
2884/**
2885 *      netif_queue_stopped - test if transmit queue is flowblocked
2886 *      @dev: network device
2887 *
2888 *      Test if transmit queue on device is currently unable to send.
2889 */
2890static inline bool netif_queue_stopped(const struct net_device *dev)
2891{
2892        return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2893}
2894
2895static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2896{
2897        return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2898}
2899
2900static inline bool
2901netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2902{
2903        return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2904}
2905
2906static inline bool
2907netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2908{
2909        return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2910}
2911
2912/**
2913 *      netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2914 *      @dev_queue: pointer to transmit queue
2915 *
2916 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2917 * to give appropriate hint to the CPU.
2918 */
2919static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2920{
2921#ifdef CONFIG_BQL
2922        prefetchw(&dev_queue->dql.num_queued);
2923#endif
2924}
2925
2926/**
2927 *      netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2928 *      @dev_queue: pointer to transmit queue
2929 *
2930 * BQL enabled drivers might use this helper in their TX completion path,
2931 * to give appropriate hint to the CPU.
2932 */
2933static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2934{
2935#ifdef CONFIG_BQL
2936        prefetchw(&dev_queue->dql.limit);
2937#endif
2938}
2939
2940static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2941                                        unsigned int bytes)
2942{
2943#ifdef CONFIG_BQL
2944        dql_queued(&dev_queue->dql, bytes);
2945
2946        if (likely(dql_avail(&dev_queue->dql) >= 0))
2947                return;
2948
2949        set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2950
2951        /*
2952         * The XOFF flag must be set before checking the dql_avail below,
2953         * because in netdev_tx_completed_queue we update the dql_completed
2954         * before checking the XOFF flag.
2955         */
2956        smp_mb();
2957
2958        /* check again in case another CPU has just made room avail */
2959        if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2960                clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2961#endif
2962}
2963
2964/**
2965 *      netdev_sent_queue - report the number of bytes queued to hardware
2966 *      @dev: network device
2967 *      @bytes: number of bytes queued to the hardware device queue
2968 *
2969 *      Report the number of bytes queued for sending/completion to the network
2970 *      device hardware queue. @bytes should be a good approximation and should
2971 *      exactly match netdev_completed_queue() @bytes
2972 */
2973static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2974{
2975        netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2976}
2977
2978static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2979                                             unsigned int pkts, unsigned int bytes)
2980{
2981#ifdef CONFIG_BQL
2982        if (unlikely(!bytes))
2983                return;
2984
2985        dql_completed(&dev_queue->dql, bytes);
2986
2987        /*
2988         * Without the memory barrier there is a small possiblity that
2989         * netdev_tx_sent_queue will miss the update and cause the queue to
2990         * be stopped forever
2991         */
2992        smp_mb();
2993
2994        if (dql_avail(&dev_queue->dql) < 0)
2995                return;
2996
2997        if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2998                netif_schedule_queue(dev_queue);
2999#endif
3000}
3001
3002/**
3003 *      netdev_completed_queue - report bytes and packets completed by device
3004 *      @dev: network device
3005 *      @pkts: actual number of packets sent over the medium
3006 *      @bytes: actual number of bytes sent over the medium
3007 *
3008 *      Report the number of bytes and packets transmitted by the network device
3009 *      hardware queue over the physical medium, @bytes must exactly match the
3010 *      @bytes amount passed to netdev_sent_queue()
3011 */
3012static inline void netdev_completed_queue(struct net_device *dev,
3013                                          unsigned int pkts, unsigned int bytes)
3014{
3015        netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3016}
3017
3018static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3019{
3020#ifdef CONFIG_BQL
3021        clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3022        dql_reset(&q->dql);
3023#endif
3024}
3025
3026/**
3027 *      netdev_reset_queue - reset the packets and bytes count of a network device
3028 *      @dev_queue: network device
3029 *
3030 *      Reset the bytes and packet count of a network device and clear the
3031 *      software flow control OFF bit for this network device
3032 */
3033static inline void netdev_reset_queue(struct net_device *dev_queue)
3034{
3035        netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3036}
3037
3038/**
3039 *      netdev_cap_txqueue - check if selected tx queue exceeds device queues
3040 *      @dev: network device
3041 *      @queue_index: given tx queue index
3042 *
3043 *      Returns 0 if given tx queue index >= number of device tx queues,
3044 *      otherwise returns the originally passed tx queue index.
3045 */
3046static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3047{
3048        if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3049                net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3050                                     dev->name, queue_index,
3051                                     dev->real_num_tx_queues);
3052                return 0;
3053        }
3054
3055        return queue_index;
3056}
3057
3058/**
3059 *      netif_running - test if up
3060 *      @dev: network device
3061 *
3062 *      Test if the device has been brought up.
3063 */
3064static inline bool netif_running(const struct net_device *dev)
3065{
3066        return test_bit(__LINK_STATE_START, &dev->state);
3067}
3068
3069/*
3070 * Routines to manage the subqueues on a device.  We only need start,
3071 * stop, and a check if it's stopped.  All other device management is
3072 * done at the overall netdevice level.
3073 * Also test the device if we're multiqueue.
3074 */
3075
3076/**
3077 *      netif_start_subqueue - allow sending packets on subqueue
3078 *      @dev: network device
3079 *      @queue_index: sub queue index
3080 *
3081 * Start individual transmit queue of a device with multiple transmit queues.
3082 */
3083static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3084{
3085        struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3086
3087        netif_tx_start_queue(txq);
3088}
3089
3090/**
3091 *      netif_stop_subqueue - stop sending packets on subqueue
3092 *      @dev: network device
3093 *      @queue_index: sub queue index
3094 *
3095 * Stop individual transmit queue of a device with multiple transmit queues.
3096 */
3097static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3098{
3099        struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3100        netif_tx_stop_queue(txq);
3101}
3102
3103/**
3104 *      netif_subqueue_stopped - test status of subqueue
3105 *      @dev: network device
3106 *      @queue_index: sub queue index
3107 *
3108 * Check individual transmit queue of a device with multiple transmit queues.
3109 */
3110static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3111                                            u16 queue_index)
3112{
3113        struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3114
3115        return netif_tx_queue_stopped(txq);
3116}
3117
3118static inline bool netif_subqueue_stopped(const struct net_device *dev,
3119                                          struct sk_buff *skb)
3120{
3121        return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3122}
3123
3124void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
3125
3126#ifdef CONFIG_XPS
3127int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3128                        u16 index);
3129#else
3130static inline int netif_set_xps_queue(struct net_device *dev,
3131                                      const struct cpumask *mask,
3132                                      u16 index)
3133{
3134        return 0;
3135}
3136#endif
3137
3138u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3139                  unsigned int num_tx_queues);
3140
3141/*
3142 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3143 * as a distribution range limit for the returned value.
3144 */
3145static inline u16 skb_tx_hash(const struct net_device *dev,
3146                              struct sk_buff *skb)
3147{
3148        return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3149}
3150
3151/**
3152 *      netif_is_multiqueue - test if device has multiple transmit queues
3153 *      @dev: network device
3154 *
3155 * Check if device has multiple transmit queues
3156 */
3157static inline bool netif_is_multiqueue(const struct net_device *dev)
3158{
3159        return dev->num_tx_queues > 1;
3160}
3161
3162int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3163
3164#ifdef CONFIG_SYSFS
3165int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3166#else
3167static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3168                                                unsigned int rxq)
3169{
3170        return 0;
3171}
3172#endif
3173
3174#ifdef CONFIG_SYSFS
3175static inline unsigned int get_netdev_rx_queue_index(
3176                struct netdev_rx_queue *queue)
3177{
3178        struct net_device *dev = queue->dev;
3179        int index = queue - dev->_rx;
3180
3181        BUG_ON(index >= dev->num_rx_queues);
3182        return index;
3183}
3184#endif
3185
3186#define DEFAULT_MAX_NUM_RSS_QUEUES      (8)
3187int netif_get_num_default_rss_queues(void);
3188
3189enum skb_free_reason {
3190        SKB_REASON_CONSUMED,
3191        SKB_REASON_DROPPED,
3192};
3193
3194void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3195void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3196
3197/*
3198 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3199 * interrupt context or with hardware interrupts being disabled.
3200 * (in_irq() || irqs_disabled())
3201 *
3202 * We provide four helpers that can be used in following contexts :
3203 *
3204 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3205 *  replacing kfree_skb(skb)
3206 *
3207 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3208 *  Typically used in place of consume_skb(skb) in TX completion path
3209 *
3210 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3211 *  replacing kfree_skb(skb)
3212 *
3213 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3214 *  and consumed a packet. Used in place of consume_skb(skb)
3215 */
3216static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3217{
3218        __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3219}
3220
3221static inline void dev_consume_skb_irq(struct sk_buff *skb)
3222{
3223        __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3224}
3225
3226static inline void dev_kfree_skb_any(struct sk_buff *skb)
3227{
3228        __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3229}
3230
3231static inline void dev_consume_skb_any(struct sk_buff *skb)
3232{
3233        __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3234}
3235
3236int netif_rx(struct sk_buff *skb);
3237int netif_rx_ni(struct sk_buff *skb);
3238int netif_receive_skb(struct sk_buff *skb);
3239gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3240void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3241struct sk_buff *napi_get_frags(struct napi_struct *napi);
3242gro_result_t napi_gro_frags(struct napi_struct *napi);
3243struct packet_offload *gro_find_receive_by_type(__be16 type);
3244struct packet_offload *gro_find_complete_by_type(__be16 type);
3245
3246static inline void napi_free_frags(struct napi_struct *napi)
3247{
3248        kfree_skb(napi->skb);
3249        napi->skb = NULL;
3250}
3251
3252int netdev_rx_handler_register(struct net_device *dev,
3253                               rx_handler_func_t *rx_handler,
3254                               void *rx_handler_data);
3255void netdev_rx_handler_unregister(struct net_device *dev);
3256
3257bool dev_valid_name(const char *name);
3258int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3259int dev_ethtool(struct net *net, struct ifreq *);
3260unsigned int dev_get_flags(const struct net_device *);
3261int __dev_change_flags(struct net_device *, unsigned int flags);
3262int dev_change_flags(struct net_device *, unsigned int);
3263void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3264                        unsigned int gchanges);
3265int dev_change_name(struct net_device *, const char *);
3266int dev_set_alias(struct net_device *, const char *, size_t);
3267int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3268int dev_set_mtu(struct net_device *, int);
3269void dev_set_group(struct net_device *, int);
3270int dev_set_mac_address(struct net_device *, struct sockaddr *);
3271int dev_change_carrier(struct net_device *, bool new_carrier);
3272int dev_get_phys_port_id(struct net_device *dev,
3273                         struct netdev_phys_item_id *ppid);
3274int dev_get_phys_port_name(struct net_device *dev,
3275                           char *name, size_t len);
3276int dev_change_proto_down(struct net_device *dev, bool proto_down);
3277struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3278struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3279                                    struct netdev_queue *txq, int *ret);
3280int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3281int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3282bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
3283
3284extern int              netdev_budget;
3285
3286/* Called by rtnetlink.c:rtnl_unlock() */
3287void netdev_run_todo(void);
3288
3289/**
3290 *      dev_put - release reference to device
3291 *      @dev: network device
3292 *
3293 * Release reference to device to allow it to be freed.
3294 */
3295static inline void dev_put(struct net_device *dev)
3296{
3297        this_cpu_dec(*dev->pcpu_refcnt);
3298}
3299
3300/**
3301 *      dev_hold - get reference to device
3302 *      @dev: network device
3303 *
3304 * Hold reference to device to keep it from being freed.
3305 */
3306static inline void dev_hold(struct net_device *dev)
3307{
3308        this_cpu_inc(*dev->pcpu_refcnt);
3309}
3310
3311/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3312 * and _off may be called from IRQ context, but it is caller
3313 * who is responsible for serialization of these calls.
3314 *
3315 * The name carrier is inappropriate, these functions should really be
3316 * called netif_lowerlayer_*() because they represent the state of any
3317 * kind of lower layer not just hardware media.
3318 */
3319
3320void linkwatch_init_dev(struct net_device *dev);
3321void linkwatch_fire_event(struct net_device *dev);
3322void linkwatch_forget_dev(struct net_device *dev);
3323
3324/**
3325 *      netif_carrier_ok - test if carrier present
3326 *      @dev: network device
3327 *
3328 * Check if carrier is present on device
3329 */
3330static inline bool netif_carrier_ok(const struct net_device *dev)
3331{
3332        return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3333}
3334
3335unsigned long dev_trans_start(struct net_device *dev);
3336
3337void __netdev_watchdog_up(struct net_device *dev);
3338
3339void netif_carrier_on(struct net_device *dev);
3340
3341void netif_carrier_off(struct net_device *dev);
3342
3343/**
3344 *      netif_dormant_on - mark device as dormant.
3345 *      @dev: network device
3346 *
3347 * Mark device as dormant (as per RFC2863).
3348 *
3349 * The dormant state indicates that the relevant interface is not
3350 * actually in a condition to pass packets (i.e., it is not 'up') but is
3351 * in a "pending" state, waiting for some external event.  For "on-
3352 * demand" interfaces, this new state identifies the situation where the
3353 * interface is waiting for events to place it in the up state.
3354 */
3355static inline void netif_dormant_on(struct net_device *dev)
3356{
3357        if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3358                linkwatch_fire_event(dev);
3359}
3360
3361/**
3362 *      netif_dormant_off - set device as not dormant.
3363 *      @dev: network device
3364 *
3365 * Device is not in dormant state.
3366 */
3367static inline void netif_dormant_off(struct net_device *dev)
3368{
3369        if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3370                linkwatch_fire_event(dev);
3371}
3372
3373/**
3374 *      netif_dormant - test if carrier present
3375 *      @dev: network device
3376 *
3377 * Check if carrier is present on device
3378 */
3379static inline bool netif_dormant(const struct net_device *dev)
3380{
3381        return test_bit(__LINK_STATE_DORMANT, &dev->state);
3382}
3383
3384
3385/**
3386 *      netif_oper_up - test if device is operational
3387 *      @dev: network device
3388 *
3389 * Check if carrier is operational
3390 */
3391static inline bool netif_oper_up(const struct net_device *dev)
3392{
3393        return (dev->operstate == IF_OPER_UP ||
3394                dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3395}
3396
3397/**
3398 *      netif_device_present - is device available or removed
3399 *      @dev: network device
3400 *
3401 * Check if device has not been removed from system.
3402 */
3403static inline bool netif_device_present(struct net_device *dev)
3404{
3405        return test_bit(__LINK_STATE_PRESENT, &dev->state);
3406}
3407
3408void netif_device_detach(struct net_device *dev);
3409
3410void netif_device_attach(struct net_device *dev);
3411
3412/*
3413 * Network interface message level settings
3414 */
3415
3416enum {
3417        NETIF_MSG_DRV           = 0x0001,
3418        NETIF_MSG_PROBE         = 0x0002,
3419        NETIF_MSG_LINK          = 0x0004,
3420        NETIF_MSG_TIMER         = 0x0008,
3421        NETIF_MSG_IFDOWN        = 0x0010,
3422        NETIF_MSG_IFUP          = 0x0020,
3423        NETIF_MSG_RX_ERR        = 0x0040,
3424        NETIF_MSG_TX_ERR        = 0x0080,
3425        NETIF_MSG_TX_QUEUED     = 0x0100,
3426        NETIF_MSG_INTR          = 0x0200,
3427        NETIF_MSG_TX_DONE       = 0x0400,
3428        NETIF_MSG_RX_STATUS     = 0x0800,
3429        NETIF_MSG_PKTDATA       = 0x1000,
3430        NETIF_MSG_HW            = 0x2000,
3431        NETIF_MSG_WOL           = 0x4000,
3432};
3433
3434#define netif_msg_drv(p)        ((p)->msg_enable & NETIF_MSG_DRV)
3435#define netif_msg_probe(p)      ((p)->msg_enable & NETIF_MSG_PROBE)
3436#define netif_msg_link(p)       ((p)->msg_enable & NETIF_MSG_LINK)
3437#define netif_msg_timer(p)      ((p)->msg_enable & NETIF_MSG_TIMER)
3438#define netif_msg_ifdown(p)     ((p)->msg_enable & NETIF_MSG_IFDOWN)
3439#define netif_msg_ifup(p)       ((p)->msg_enable & NETIF_MSG_IFUP)
3440#define netif_msg_rx_err(p)     ((p)->msg_enable & NETIF_MSG_RX_ERR)
3441#define netif_msg_tx_err(p)     ((p)->msg_enable & NETIF_MSG_TX_ERR)
3442#define netif_msg_tx_queued(p)  ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3443#define netif_msg_intr(p)       ((p)->msg_enable & NETIF_MSG_INTR)
3444#define netif_msg_tx_done(p)    ((p)->msg_enable & NETIF_MSG_TX_DONE)
3445#define netif_msg_rx_status(p)  ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3446#define netif_msg_pktdata(p)    ((p)->msg_enable & NETIF_MSG_PKTDATA)
3447#define netif_msg_hw(p)         ((p)->msg_enable & NETIF_MSG_HW)
3448#define netif_msg_wol(p)        ((p)->msg_enable & NETIF_MSG_WOL)
3449
3450static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3451{
3452        /* use default */
3453        if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3454                return default_msg_enable_bits;
3455        if (debug_value == 0)   /* no output */
3456                return 0;
3457        /* set low N bits */
3458        return (1 << debug_value) - 1;
3459}
3460
3461static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3462{
3463        spin_lock(&txq->_xmit_lock);
3464        txq->xmit_lock_owner = cpu;
3465}
3466
3467static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3468{
3469        spin_lock_bh(&txq->_xmit_lock);
3470        txq->xmit_lock_owner = smp_processor_id();
3471}
3472
3473static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3474{
3475        bool ok = spin_trylock(&txq->_xmit_lock);
3476        if (likely(ok))
3477                txq->xmit_lock_owner = smp_processor_id();
3478        return ok;
3479}
3480
3481static inline void __netif_tx_unlock(struct netdev_queue *txq)
3482{
3483        txq->xmit_lock_owner = -1;
3484        spin_unlock(&txq->_xmit_lock);
3485}
3486
3487static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3488{
3489        txq->xmit_lock_owner = -1;
3490        spin_unlock_bh(&txq->_xmit_lock);
3491}
3492
3493static inline void txq_trans_update(struct netdev_queue *txq)
3494{
3495        if (txq->xmit_lock_owner != -1)
3496                txq->trans_start = jiffies;
3497}
3498
3499/**
3500 *      netif_tx_lock - grab network device transmit lock
3501 *      @dev: network device
3502 *
3503 * Get network device transmit lock
3504 */
3505static inline void netif_tx_lock(struct net_device *dev)
3506{
3507        unsigned int i;
3508        int cpu;
3509
3510        spin_lock(&dev->tx_global_lock);
3511        cpu = smp_processor_id();
3512        for (i = 0; i < dev->num_tx_queues; i++) {
3513                struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3514
3515                /* We are the only thread of execution doing a
3516                 * freeze, but we have to grab the _xmit_lock in
3517                 * order to synchronize with threads which are in
3518                 * the ->hard_start_xmit() handler and already
3519                 * checked the frozen bit.
3520                 */
3521                __netif_tx_lock(txq, cpu);
3522                set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3523                __netif_tx_unlock(txq);
3524        }
3525}
3526
3527static inline void netif_tx_lock_bh(struct net_device *dev)
3528{
3529        local_bh_disable();
3530        netif_tx_lock(dev);
3531}
3532
3533static inline void netif_tx_unlock(struct net_device *dev)
3534{
3535        unsigned int i;
3536
3537        for (i = 0; i < dev->num_tx_queues; i++) {
3538                struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3539
3540                /* No need to grab the _xmit_lock here.  If the
3541                 * queue is not stopped for another reason, we
3542                 * force a schedule.
3543                 */
3544                clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3545                netif_schedule_queue(txq);
3546        }
3547        spin_unlock(&dev->tx_global_lock);
3548}
3549
3550static inline void netif_tx_unlock_bh(struct net_device *dev)
3551{
3552        netif_tx_unlock(dev);
3553        local_bh_enable();
3554}
3555
3556#define HARD_TX_LOCK(dev, txq, cpu) {                   \
3557        if ((dev->features & NETIF_F_LLTX) == 0) {      \
3558                __netif_tx_lock(txq, cpu);              \
3559        }                                               \
3560}
3561
3562#define HARD_TX_TRYLOCK(dev, txq)                       \
3563        (((dev->features & NETIF_F_LLTX) == 0) ?        \
3564                __netif_tx_trylock(txq) :               \
3565                true )
3566
3567#define HARD_TX_UNLOCK(dev, txq) {                      \
3568        if ((dev->features & NETIF_F_LLTX) == 0) {      \
3569                __netif_tx_unlock(txq);                 \
3570        }                                               \
3571}
3572
3573static inline void netif_tx_disable(struct net_device *dev)
3574{
3575        unsigned int i;
3576        int cpu;
3577
3578        local_bh_disable();
3579        cpu = smp_processor_id();
3580        for (i = 0; i < dev->num_tx_queues; i++) {
3581                struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3582
3583                __netif_tx_lock(txq, cpu);
3584                netif_tx_stop_queue(txq);
3585                __netif_tx_unlock(txq);
3586        }
3587        local_bh_enable();
3588}
3589
3590static inline void netif_addr_lock(struct net_device *dev)
3591{
3592        spin_lock(&dev->addr_list_lock);
3593}
3594
3595static inline void netif_addr_lock_nested(struct net_device *dev)
3596{
3597        int subclass = SINGLE_DEPTH_NESTING;
3598
3599        if (dev->netdev_ops->ndo_get_lock_subclass)
3600                subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3601
3602        spin_lock_nested(&dev->addr_list_lock, subclass);
3603}
3604
3605static inline void netif_addr_lock_bh(struct net_device *dev)
3606{
3607        spin_lock_bh(&dev->addr_list_lock);
3608}
3609
3610static inline void netif_addr_unlock(struct net_device *dev)
3611{
3612        spin_unlock(&dev->addr_list_lock);
3613}
3614
3615static inline void netif_addr_unlock_bh(struct net_device *dev)
3616{
3617        spin_unlock_bh(&dev->addr_list_lock);
3618}
3619
3620/*
3621 * dev_addrs walker. Should be used only for read access. Call with
3622 * rcu_read_lock held.
3623 */
3624#define for_each_dev_addr(dev, ha) \
3625                list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3626
3627/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3628
3629void ether_setup(struct net_device *dev);
3630
3631/* Support for loadable net-drivers */
3632struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3633                                    unsigned char name_assign_type,
3634                                    void (*setup)(struct net_device *),
3635                                    unsigned int txqs, unsigned int rxqs);
3636#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3637        alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3638
3639#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3640        alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3641                         count)
3642
3643int register_netdev(struct net_device *dev);
3644void unregister_netdev(struct net_device *dev);
3645
3646/* General hardware address lists handling functions */
3647int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3648                   struct netdev_hw_addr_list *from_list, int addr_len);
3649void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3650                      struct netdev_hw_addr_list *from_list, int addr_len);
3651int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3652                       struct net_device *dev,
3653                       int (*sync)(struct net_device *, const unsigned char *),
3654                       int (*unsync)(struct net_device *,
3655                                     const unsigned char *));
3656void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3657                          struct net_device *dev,
3658                          int (*unsync)(struct net_device *,
3659                                        const unsigned char *));
3660void __hw_addr_init(struct netdev_hw_addr_list *list);
3661
3662/* Functions used for device addresses handling */
3663int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3664                 unsigned char addr_type);
3665int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3666                 unsigned char addr_type);
3667void dev_addr_flush(struct net_device *dev);
3668int dev_addr_init(struct net_device *dev);
3669
3670/* Functions used for unicast addresses handling */
3671int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3672int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3673int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3674int dev_uc_sync(struct net_device *to, struct net_device *from);
3675int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3676void dev_uc_unsync(struct net_device *to, struct net_device *from);
3677void dev_uc_flush(struct net_device *dev);
3678void dev_uc_init(struct net_device *dev);
3679
3680/**
3681 *  __dev_uc_sync - Synchonize device's unicast list
3682 *  @dev:  device to sync
3683 *  @sync: function to call if address should be added
3684 *  @unsync: function to call if address should be removed
3685 *
3686 *  Add newly added addresses to the interface, and release
3687 *  addresses that have been deleted.
3688 */
3689static inline int __dev_uc_sync(struct net_device *dev,
3690                                int (*sync)(struct net_device *,
3691                                            const unsigned char *),
3692                                int (*unsync)(struct net_device *,
3693                                              const unsigned char *))
3694{
3695        return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3696}
3697
3698/**
3699 *  __dev_uc_unsync - Remove synchronized addresses from device
3700 *  @dev:  device to sync
3701 *  @unsync: function to call if address should be removed
3702 *
3703 *  Remove all addresses that were added to the device by dev_uc_sync().
3704 */
3705static inline void __dev_uc_unsync(struct net_device *dev,
3706                                   int (*unsync)(struct net_device *,
3707                                                 const unsigned char *))
3708{
3709        __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3710}
3711
3712/* Functions used for multicast addresses handling */
3713int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3714int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3715int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3716int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3717int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3718int dev_mc_sync(struct net_device *to, struct net_device *from);
3719int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3720void dev_mc_unsync(struct net_device *to, struct net_device *from);
3721void dev_mc_flush(struct net_device *dev);
3722void dev_mc_init(struct net_device *dev);
3723
3724/**
3725 *  __dev_mc_sync - Synchonize device's multicast list
3726 *  @dev:  device to sync
3727 *  @sync: function to call if address should be added
3728 *  @unsync: function to call if address should be removed
3729 *
3730 *  Add newly added addresses to the interface, and release
3731 *  addresses that have been deleted.
3732 */
3733static inline int __dev_mc_sync(struct net_device *dev,
3734                                int (*sync)(struct net_device *,
3735                                            const unsigned char *),
3736                                int (*unsync)(struct net_device *,
3737                                              const unsigned char *))
3738{
3739        return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3740}
3741
3742/**
3743 *  __dev_mc_unsync - Remove synchronized addresses from device
3744 *  @dev:  device to sync
3745 *  @unsync: function to call if address should be removed
3746 *
3747 *  Remove all addresses that were added to the device by dev_mc_sync().
3748 */
3749static inline void __dev_mc_unsync(struct net_device *dev,
3750                                   int (*unsync)(struct net_device *,
3751                                                 const unsigned char *))
3752{
3753        __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3754}
3755
3756/* Functions used for secondary unicast and multicast support */
3757void dev_set_rx_mode(struct net_device *dev);
3758void __dev_set_rx_mode(struct net_device *dev);
3759int dev_set_promiscuity(struct net_device *dev, int inc);
3760int dev_set_allmulti(struct net_device *dev, int inc);
3761void netdev_state_change(struct net_device *dev);
3762void netdev_notify_peers(struct net_device *dev);
3763void netdev_features_change(struct net_device *dev);
3764/* Load a device via the kmod */
3765void dev_load(struct net *net, const char *name);
3766struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3767                                        struct rtnl_link_stats64 *storage);
3768void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3769                             const struct net_device_stats *netdev_stats);
3770
3771extern int              netdev_max_backlog;
3772extern int              netdev_tstamp_prequeue;
3773extern int              weight_p;
3774extern int              bpf_jit_enable;
3775
3776bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3777struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3778                                                     struct list_head **iter);
3779struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3780                                                     struct list_head **iter);
3781
3782/* iterate through upper list, must be called under RCU read lock */
3783#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3784        for (iter = &(dev)->adj_list.upper, \
3785             updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3786             updev; \
3787             updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3788
3789/* iterate through upper list, must be called under RCU read lock */
3790#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3791        for (iter = &(dev)->all_adj_list.upper, \
3792             updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3793             updev; \
3794             updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3795
3796void *netdev_lower_get_next_private(struct net_device *dev,
3797                                    struct list_head **iter);
3798void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3799                                        struct list_head **iter);
3800
3801#define netdev_for_each_lower_private(dev, priv, iter) \
3802        for (iter = (dev)->adj_list.lower.next, \
3803             priv = netdev_lower_get_next_private(dev, &(iter)); \
3804             priv; \
3805             priv = netdev_lower_get_next_private(dev, &(iter)))
3806
3807#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3808        for (iter = &(dev)->adj_list.lower, \
3809             priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3810             priv; \
3811             priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3812
3813void *netdev_lower_get_next(struct net_device *dev,
3814                                struct list_head **iter);
3815#define netdev_for_each_lower_dev(dev, ldev, iter) \
3816        for (iter = (dev)->adj_list.lower.next, \
3817             ldev = netdev_lower_get_next(dev, &(iter)); \
3818             ldev; \
3819             ldev = netdev_lower_get_next(dev, &(iter)))
3820
3821void *netdev_adjacent_get_private(struct list_head *adj_list);
3822void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3823struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3824struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3825int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3826int netdev_master_upper_dev_link(struct net_device *dev,
3827                                 struct net_device *upper_dev,
3828                                 void *upper_priv, void *upper_info);
3829void netdev_upper_dev_unlink(struct net_device *dev,
3830                             struct net_device *upper_dev);
3831void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3832void *netdev_lower_dev_get_private(struct net_device *dev,
3833                                   struct net_device *lower_dev);
3834void netdev_lower_state_changed(struct net_device *lower_dev,
3835                                void *lower_state_info);
3836
3837/* RSS keys are 40 or 52 bytes long */
3838#define NETDEV_RSS_KEY_LEN 52
3839extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3840void netdev_rss_key_fill(void *buffer, size_t len);
3841
3842int dev_get_nest_level(struct net_device *dev,
3843                       bool (*type_check)(const struct net_device *dev));
3844int skb_checksum_help(struct sk_buff *skb);
3845struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3846                                  netdev_features_t features, bool tx_path);
3847struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3848                                    netdev_features_t features);
3849
3850struct netdev_bonding_info {
3851        ifslave slave;
3852        ifbond  master;
3853};
3854
3855struct netdev_notifier_bonding_info {
3856        struct netdev_notifier_info info; /* must be first */
3857        struct netdev_bonding_info  bonding_info;
3858};
3859
3860void netdev_bonding_info_change(struct net_device *dev,
3861                                struct netdev_bonding_info *bonding_info);
3862
3863static inline
3864struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3865{
3866        return __skb_gso_segment(skb, features, true);
3867}
3868__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3869
3870static inline bool can_checksum_protocol(netdev_features_t features,
3871                                         __be16 protocol)
3872{
3873        if (protocol == htons(ETH_P_FCOE))
3874                return !!(features & NETIF_F_FCOE_CRC);
3875
3876        /* Assume this is an IP checksum (not SCTP CRC) */
3877
3878        if (features & NETIF_F_HW_CSUM) {
3879                /* Can checksum everything */
3880                return true;
3881        }
3882
3883        switch (protocol) {
3884        case htons(ETH_P_IP):
3885                return !!(features & NETIF_F_IP_CSUM);
3886        case htons(ETH_P_IPV6):
3887                return !!(features & NETIF_F_IPV6_CSUM);
3888        default:
3889                return false;
3890        }
3891}
3892
3893/* Map an ethertype into IP protocol if possible */
3894static inline int eproto_to_ipproto(int eproto)
3895{
3896        switch (eproto) {
3897        case htons(ETH_P_IP):
3898                return IPPROTO_IP;
3899        case htons(ETH_P_IPV6):
3900                return IPPROTO_IPV6;
3901        default:
3902                return -1;
3903        }
3904}
3905
3906#ifdef CONFIG_BUG
3907void netdev_rx_csum_fault(struct net_device *dev);
3908#else
3909static inline void netdev_rx_csum_fault(struct net_device *dev)
3910{
3911}
3912#endif
3913/* rx skb timestamps */
3914void net_enable_timestamp(void);
3915void net_disable_timestamp(void);
3916
3917#ifdef CONFIG_PROC_FS
3918int __init dev_proc_init(void);
3919#else
3920#define dev_proc_init() 0
3921#endif
3922
3923static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3924                                              struct sk_buff *skb, struct net_device *dev,
3925                                              bool more)
3926{
3927        skb->xmit_more = more ? 1 : 0;
3928        return ops->ndo_start_xmit(skb, dev);
3929}
3930
3931static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3932                                            struct netdev_queue *txq, bool more)
3933{
3934        const struct net_device_ops *ops = dev->netdev_ops;
3935        int rc;
3936
3937        rc = __netdev_start_xmit(ops, skb, dev, more);
3938        if (rc == NETDEV_TX_OK)
3939                txq_trans_update(txq);
3940
3941        return rc;
3942}
3943
3944int netdev_class_create_file_ns(struct class_attribute *class_attr,
3945                                const void *ns);
3946void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3947                                 const void *ns);
3948
3949static inline int netdev_class_create_file(struct class_attribute *class_attr)
3950{
3951        return netdev_class_create_file_ns(class_attr, NULL);
3952}
3953
3954static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3955{
3956        netdev_class_remove_file_ns(class_attr, NULL);
3957}
3958
3959extern struct kobj_ns_type_operations net_ns_type_operations;
3960
3961const char *netdev_drivername(const struct net_device *dev);
3962
3963void linkwatch_run_queue(void);
3964
3965static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3966                                                          netdev_features_t f2)
3967{
3968        if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
3969                if (f1 & NETIF_F_HW_CSUM)
3970                        f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
3971                else
3972                        f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
3973        }
3974
3975        return f1 & f2;
3976}
3977
3978static inline netdev_features_t netdev_get_wanted_features(
3979        struct net_device *dev)
3980{
3981        return (dev->features & ~dev->hw_features) | dev->wanted_features;
3982}
3983netdev_features_t netdev_increment_features(netdev_features_t all,
3984        netdev_features_t one, netdev_features_t mask);
3985
3986/* Allow TSO being used on stacked device :
3987 * Performing the GSO segmentation before last device
3988 * is a performance improvement.
3989 */
3990static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3991                                                        netdev_features_t mask)
3992{
3993        return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3994}
3995
3996int __netdev_update_features(struct net_device *dev);
3997void netdev_update_features(struct net_device *dev);
3998void netdev_change_features(struct net_device *dev);
3999
4000void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4001                                        struct net_device *dev);
4002
4003netdev_features_t passthru_features_check(struct sk_buff *skb,
4004                                          struct net_device *dev,
4005                                          netdev_features_t features);
4006netdev_features_t netif_skb_features(struct sk_buff *skb);
4007
4008static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4009{
4010        netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4011
4012        /* check flags correspondence */
4013        BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4014        BUILD_BUG_ON(SKB_GSO_UDP     != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4015        BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4016        BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4017        BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4018        BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4019        BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4020        BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4021        BUILD_BUG_ON(SKB_GSO_IPIP    != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
4022        BUILD_BUG_ON(SKB_GSO_SIT     != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
4023        BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4024        BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4025        BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4026
4027        return (features & feature) == feature;
4028}
4029
4030static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4031{
4032        return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4033               (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4034}
4035
4036static inline bool netif_needs_gso(struct sk_buff *skb,
4037                                   netdev_features_t features)
4038{
4039        return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4040                unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4041                         (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4042}
4043
4044static inline void netif_set_gso_max_size(struct net_device *dev,
4045                                          unsigned int size)
4046{
4047        dev->gso_max_size = size;
4048}
4049
4050static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4051                                        int pulled_hlen, u16 mac_offset,
4052                                        int mac_len)
4053{
4054        skb->protocol = protocol;
4055        skb->encapsulation = 1;
4056        skb_push(skb, pulled_hlen);
4057        skb_reset_transport_header(skb);
4058        skb->mac_header = mac_offset;
4059        skb->network_header = skb->mac_header + mac_len;
4060        skb->mac_len = mac_len;
4061}
4062
4063static inline bool netif_is_macsec(const struct net_device *dev)
4064{
4065        return dev->priv_flags & IFF_MACSEC;
4066}
4067
4068static inline bool netif_is_macvlan(const struct net_device *dev)
4069{
4070        return dev->priv_flags & IFF_MACVLAN;
4071}
4072
4073static inline bool netif_is_macvlan_port(const struct net_device *dev)
4074{
4075        return dev->priv_flags & IFF_MACVLAN_PORT;
4076}
4077
4078static inline bool netif_is_ipvlan(const struct net_device *dev)
4079{
4080        return dev->priv_flags & IFF_IPVLAN_SLAVE;
4081}
4082
4083static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4084{
4085        return dev->priv_flags & IFF_IPVLAN_MASTER;
4086}
4087
4088static inline bool netif_is_bond_master(const struct net_device *dev)
4089{
4090        return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4091}
4092
4093static inline bool netif_is_bond_slave(const struct net_device *dev)
4094{
4095        return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4096}
4097
4098static inline bool netif_supports_nofcs(struct net_device *dev)
4099{
4100        return dev->priv_flags & IFF_SUPP_NOFCS;
4101}
4102
4103static inline bool netif_is_l3_master(const struct net_device *dev)
4104{
4105        return dev->priv_flags & IFF_L3MDEV_MASTER;
4106}
4107
4108static inline bool netif_is_l3_slave(const struct net_device *dev)
4109{
4110        return dev->priv_flags & IFF_L3MDEV_SLAVE;
4111}
4112
4113static inline bool netif_is_bridge_master(const struct net_device *dev)
4114{
4115        return dev->priv_flags & IFF_EBRIDGE;
4116}
4117
4118static inline bool netif_is_bridge_port(const struct net_device *dev)
4119{
4120        return dev->priv_flags & IFF_BRIDGE_PORT;
4121}
4122
4123static inline bool netif_is_ovs_master(const struct net_device *dev)
4124{
4125        return dev->priv_flags & IFF_OPENVSWITCH;
4126}
4127
4128static inline bool netif_is_team_master(const struct net_device *dev)
4129{
4130        return dev->priv_flags & IFF_TEAM;
4131}
4132
4133static inline bool netif_is_team_port(const struct net_device *dev)
4134{
4135        return dev->priv_flags & IFF_TEAM_PORT;
4136}
4137
4138static inline bool netif_is_lag_master(const struct net_device *dev)
4139{
4140        return netif_is_bond_master(dev) || netif_is_team_master(dev);
4141}
4142
4143static inline bool netif_is_lag_port(const struct net_device *dev)
4144{
4145        return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4146}
4147
4148static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4149{
4150        return dev->priv_flags & IFF_RXFH_CONFIGURED;
4151}
4152
4153/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4154static inline void netif_keep_dst(struct net_device *dev)
4155{
4156        dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4157}
4158
4159extern struct pernet_operations __net_initdata loopback_net_ops;
4160
4161/* Logging, debugging and troubleshooting/diagnostic helpers. */
4162
4163/* netdev_printk helpers, similar to dev_printk */
4164
4165static inline const char *netdev_name(const struct net_device *dev)
4166{
4167        if (!dev->name[0] || strchr(dev->name, '%'))
4168                return "(unnamed net_device)";
4169        return dev->name;
4170}
4171
4172static inline const char *netdev_reg_state(const struct net_device *dev)
4173{
4174        switch (dev->reg_state) {
4175        case NETREG_UNINITIALIZED: return " (uninitialized)";
4176        case NETREG_REGISTERED: return "";
4177        case NETREG_UNREGISTERING: return " (unregistering)";
4178        case NETREG_UNREGISTERED: return " (unregistered)";
4179        case NETREG_RELEASED: return " (released)";
4180        case NETREG_DUMMY: return " (dummy)";
4181        }
4182
4183        WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4184        return " (unknown)";
4185}
4186
4187__printf(3, 4)
4188void netdev_printk(const char *level, const struct net_device *dev,
4189                   const char *format, ...);
4190__printf(2, 3)
4191void netdev_emerg(const struct net_device *dev, const char *format, ...);
4192__printf(2, 3)
4193void netdev_alert(const struct net_device *dev, const char *format, ...);
4194__printf(2, 3)
4195void netdev_crit(const struct net_device *dev, const char *format, ...);
4196__printf(2, 3)
4197void netdev_err(const struct net_device *dev, const char *format, ...);
4198__printf(2, 3)
4199void netdev_warn(const struct net_device *dev, const char *format, ...);
4200__printf(2, 3)
4201void netdev_notice(const struct net_device *dev, const char *format, ...);
4202__printf(2, 3)
4203void netdev_info(const struct net_device *dev, const char *format, ...);
4204
4205#define MODULE_ALIAS_NETDEV(device) \
4206        MODULE_ALIAS("netdev-" device)
4207
4208#if defined(CONFIG_DYNAMIC_DEBUG)
4209#define netdev_dbg(__dev, format, args...)                      \
4210do {                                                            \
4211        dynamic_netdev_dbg(__dev, format, ##args);              \
4212} while (0)
4213#elif defined(DEBUG)
4214#define netdev_dbg(__dev, format, args...)                      \
4215        netdev_printk(KERN_DEBUG, __dev, format, ##args)
4216#else
4217#define netdev_dbg(__dev, format, args...)                      \
4218({                                                              \
4219        if (0)                                                  \
4220                netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4221})
4222#endif
4223
4224#if defined(VERBOSE_DEBUG)
4225#define netdev_vdbg     netdev_dbg
4226#else
4227
4228#define netdev_vdbg(dev, format, args...)                       \
4229({                                                              \
4230        if (0)                                                  \
4231                netdev_printk(KERN_DEBUG, dev, format, ##args); \
4232        0;                                                      \
4233})
4234#endif
4235
4236/*
4237 * netdev_WARN() acts like dev_printk(), but with the key difference
4238 * of using a WARN/WARN_ON to get the message out, including the
4239 * file/line information and a backtrace.
4240 */
4241#define netdev_WARN(dev, format, args...)                       \
4242        WARN(1, "netdevice: %s%s\n" format, netdev_name(dev),   \
4243             netdev_reg_state(dev), ##args)
4244
4245/* netif printk helpers, similar to netdev_printk */
4246
4247#define netif_printk(priv, type, level, dev, fmt, args...)      \
4248do {                                                            \
4249        if (netif_msg_##type(priv))                             \
4250                netdev_printk(level, (dev), fmt, ##args);       \
4251} while (0)
4252
4253#define netif_level(level, priv, type, dev, fmt, args...)       \
4254do {                                                            \
4255        if (netif_msg_##type(priv))                             \
4256                netdev_##level(dev, fmt, ##args);               \
4257} while (0)
4258
4259#define netif_emerg(priv, type, dev, fmt, args...)              \
4260        netif_level(emerg, priv, type, dev, fmt, ##args)
4261#define netif_alert(priv, type, dev, fmt, args...)              \
4262        netif_level(alert, priv, type, dev, fmt, ##args)
4263#define netif_crit(priv, type, dev, fmt, args...)               \
4264        netif_level(crit, priv, type, dev, fmt, ##args)
4265#define netif_err(priv, type, dev, fmt, args...)                \
4266        netif_level(err, priv, type, dev, fmt, ##args)
4267#define netif_warn(priv, type, dev, fmt, args...)               \
4268        netif_level(warn, priv, type, dev, fmt, ##args)
4269#define netif_notice(priv, type, dev, fmt, args...)             \
4270        netif_level(notice, priv, type, dev, fmt, ##args)
4271#define netif_info(priv, type, dev, fmt, args...)               \
4272        netif_level(info, priv, type, dev, fmt, ##args)
4273
4274#if defined(CONFIG_DYNAMIC_DEBUG)
4275#define netif_dbg(priv, type, netdev, format, args...)          \
4276do {                                                            \
4277        if (netif_msg_##type(priv))                             \
4278                dynamic_netdev_dbg(netdev, format, ##args);     \
4279} while (0)
4280#elif defined(DEBUG)
4281#define netif_dbg(priv, type, dev, format, args...)             \
4282        netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4283#else
4284#define netif_dbg(priv, type, dev, format, args...)                     \
4285({                                                                      \
4286        if (0)                                                          \
4287                netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4288        0;                                                              \
4289})
4290#endif
4291
4292#if defined(VERBOSE_DEBUG)
4293#define netif_vdbg      netif_dbg
4294#else
4295#define netif_vdbg(priv, type, dev, format, args...)            \
4296({                                                              \
4297        if (0)                                                  \
4298                netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4299        0;                                                      \
4300})
4301#endif
4302
4303/*
4304 *      The list of packet types we will receive (as opposed to discard)
4305 *      and the routines to invoke.
4306 *
4307 *      Why 16. Because with 16 the only overlap we get on a hash of the
4308 *      low nibble of the protocol value is RARP/SNAP/X.25.
4309 *
4310 *      NOTE:  That is no longer true with the addition of VLAN tags.  Not
4311 *             sure which should go first, but I bet it won't make much
4312 *             difference if we are running VLANs.  The good news is that
4313 *             this protocol won't be in the list unless compiled in, so
4314 *             the average user (w/out VLANs) will not be adversely affected.
4315 *             --BLG
4316 *
4317 *              0800    IP
4318 *              8100    802.1Q VLAN
4319 *              0001    802.3
4320 *              0002    AX.25
4321 *              0004    802.2
4322 *              8035    RARP
4323 *              0005    SNAP
4324 *              0805    X.25
4325 *              0806    ARP
4326 *              8137    IPX
4327 *              0009    Localtalk
4328 *              86DD    IPv6
4329 */
4330#define PTYPE_HASH_SIZE (16)
4331#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4332
4333#endif  /* _LINUX_NETDEVICE_H */
4334