at v4.5 132 kB view raw
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 55struct netpoll_info; 56struct device; 57struct phy_device; 58/* 802.11 specific */ 59struct wireless_dev; 60/* 802.15.4 specific */ 61struct wpan_dev; 62struct mpls_dev; 63 64void netdev_set_default_ethtool_ops(struct net_device *dev, 65 const struct ethtool_ops *ops); 66 67/* Backlog congestion levels */ 68#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 69#define NET_RX_DROP 1 /* packet dropped */ 70 71/* 72 * Transmit return codes: transmit return codes originate from three different 73 * namespaces: 74 * 75 * - qdisc return codes 76 * - driver transmit return codes 77 * - errno values 78 * 79 * Drivers are allowed to return any one of those in their hard_start_xmit() 80 * function. Real network devices commonly used with qdiscs should only return 81 * the driver transmit return codes though - when qdiscs are used, the actual 82 * transmission happens asynchronously, so the value is not propagated to 83 * higher layers. Virtual network devices transmit synchronously, in this case 84 * the driver transmit return codes are consumed by dev_queue_xmit(), all 85 * others are propagated to higher layers. 86 */ 87 88/* qdisc ->enqueue() return codes. */ 89#define NET_XMIT_SUCCESS 0x00 90#define NET_XMIT_DROP 0x01 /* skb dropped */ 91#define NET_XMIT_CN 0x02 /* congestion notification */ 92#define NET_XMIT_POLICED 0x03 /* skb is shot by police */ 93#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 94 95/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 96 * indicates that the device will soon be dropping packets, or already drops 97 * some packets of the same priority; prompting us to send less aggressively. */ 98#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 99#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 100 101/* Driver transmit return codes */ 102#define NETDEV_TX_MASK 0xf0 103 104enum netdev_tx { 105 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 106 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 107 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 108 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */ 109}; 110typedef enum netdev_tx netdev_tx_t; 111 112/* 113 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 114 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 115 */ 116static inline bool dev_xmit_complete(int rc) 117{ 118 /* 119 * Positive cases with an skb consumed by a driver: 120 * - successful transmission (rc == NETDEV_TX_OK) 121 * - error while transmitting (rc < 0) 122 * - error while queueing to a different device (rc & NET_XMIT_MASK) 123 */ 124 if (likely(rc < NET_XMIT_MASK)) 125 return true; 126 127 return false; 128} 129 130/* 131 * Compute the worst case header length according to the protocols 132 * used. 133 */ 134 135#if defined(CONFIG_HYPERV_NET) 136# define LL_MAX_HEADER 128 137#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 138# if defined(CONFIG_MAC80211_MESH) 139# define LL_MAX_HEADER 128 140# else 141# define LL_MAX_HEADER 96 142# endif 143#else 144# define LL_MAX_HEADER 32 145#endif 146 147#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 148 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 149#define MAX_HEADER LL_MAX_HEADER 150#else 151#define MAX_HEADER (LL_MAX_HEADER + 48) 152#endif 153 154/* 155 * Old network device statistics. Fields are native words 156 * (unsigned long) so they can be read and written atomically. 157 */ 158 159struct net_device_stats { 160 unsigned long rx_packets; 161 unsigned long tx_packets; 162 unsigned long rx_bytes; 163 unsigned long tx_bytes; 164 unsigned long rx_errors; 165 unsigned long tx_errors; 166 unsigned long rx_dropped; 167 unsigned long tx_dropped; 168 unsigned long multicast; 169 unsigned long collisions; 170 unsigned long rx_length_errors; 171 unsigned long rx_over_errors; 172 unsigned long rx_crc_errors; 173 unsigned long rx_frame_errors; 174 unsigned long rx_fifo_errors; 175 unsigned long rx_missed_errors; 176 unsigned long tx_aborted_errors; 177 unsigned long tx_carrier_errors; 178 unsigned long tx_fifo_errors; 179 unsigned long tx_heartbeat_errors; 180 unsigned long tx_window_errors; 181 unsigned long rx_compressed; 182 unsigned long tx_compressed; 183}; 184 185 186#include <linux/cache.h> 187#include <linux/skbuff.h> 188 189#ifdef CONFIG_RPS 190#include <linux/static_key.h> 191extern struct static_key rps_needed; 192#endif 193 194struct neighbour; 195struct neigh_parms; 196struct sk_buff; 197 198struct netdev_hw_addr { 199 struct list_head list; 200 unsigned char addr[MAX_ADDR_LEN]; 201 unsigned char type; 202#define NETDEV_HW_ADDR_T_LAN 1 203#define NETDEV_HW_ADDR_T_SAN 2 204#define NETDEV_HW_ADDR_T_SLAVE 3 205#define NETDEV_HW_ADDR_T_UNICAST 4 206#define NETDEV_HW_ADDR_T_MULTICAST 5 207 bool global_use; 208 int sync_cnt; 209 int refcount; 210 int synced; 211 struct rcu_head rcu_head; 212}; 213 214struct netdev_hw_addr_list { 215 struct list_head list; 216 int count; 217}; 218 219#define netdev_hw_addr_list_count(l) ((l)->count) 220#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 221#define netdev_hw_addr_list_for_each(ha, l) \ 222 list_for_each_entry(ha, &(l)->list, list) 223 224#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 225#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 226#define netdev_for_each_uc_addr(ha, dev) \ 227 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 228 229#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 230#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 231#define netdev_for_each_mc_addr(ha, dev) \ 232 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 233 234struct hh_cache { 235 u16 hh_len; 236 u16 __pad; 237 seqlock_t hh_lock; 238 239 /* cached hardware header; allow for machine alignment needs. */ 240#define HH_DATA_MOD 16 241#define HH_DATA_OFF(__len) \ 242 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 243#define HH_DATA_ALIGN(__len) \ 244 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 245 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 246}; 247 248/* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much. 249 * Alternative is: 250 * dev->hard_header_len ? (dev->hard_header_len + 251 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 252 * 253 * We could use other alignment values, but we must maintain the 254 * relationship HH alignment <= LL alignment. 255 */ 256#define LL_RESERVED_SPACE(dev) \ 257 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 258#define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 259 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 260 261struct header_ops { 262 int (*create) (struct sk_buff *skb, struct net_device *dev, 263 unsigned short type, const void *daddr, 264 const void *saddr, unsigned int len); 265 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 266 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 267 void (*cache_update)(struct hh_cache *hh, 268 const struct net_device *dev, 269 const unsigned char *haddr); 270}; 271 272/* These flag bits are private to the generic network queueing 273 * layer, they may not be explicitly referenced by any other 274 * code. 275 */ 276 277enum netdev_state_t { 278 __LINK_STATE_START, 279 __LINK_STATE_PRESENT, 280 __LINK_STATE_NOCARRIER, 281 __LINK_STATE_LINKWATCH_PENDING, 282 __LINK_STATE_DORMANT, 283}; 284 285 286/* 287 * This structure holds at boot time configured netdevice settings. They 288 * are then used in the device probing. 289 */ 290struct netdev_boot_setup { 291 char name[IFNAMSIZ]; 292 struct ifmap map; 293}; 294#define NETDEV_BOOT_SETUP_MAX 8 295 296int __init netdev_boot_setup(char *str); 297 298/* 299 * Structure for NAPI scheduling similar to tasklet but with weighting 300 */ 301struct napi_struct { 302 /* The poll_list must only be managed by the entity which 303 * changes the state of the NAPI_STATE_SCHED bit. This means 304 * whoever atomically sets that bit can add this napi_struct 305 * to the per-cpu poll_list, and whoever clears that bit 306 * can remove from the list right before clearing the bit. 307 */ 308 struct list_head poll_list; 309 310 unsigned long state; 311 int weight; 312 unsigned int gro_count; 313 int (*poll)(struct napi_struct *, int); 314#ifdef CONFIG_NETPOLL 315 spinlock_t poll_lock; 316 int poll_owner; 317#endif 318 struct net_device *dev; 319 struct sk_buff *gro_list; 320 struct sk_buff *skb; 321 struct hrtimer timer; 322 struct list_head dev_list; 323 struct hlist_node napi_hash_node; 324 unsigned int napi_id; 325}; 326 327enum { 328 NAPI_STATE_SCHED, /* Poll is scheduled */ 329 NAPI_STATE_DISABLE, /* Disable pending */ 330 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 331 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */ 332 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */ 333}; 334 335enum gro_result { 336 GRO_MERGED, 337 GRO_MERGED_FREE, 338 GRO_HELD, 339 GRO_NORMAL, 340 GRO_DROP, 341}; 342typedef enum gro_result gro_result_t; 343 344/* 345 * enum rx_handler_result - Possible return values for rx_handlers. 346 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 347 * further. 348 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 349 * case skb->dev was changed by rx_handler. 350 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 351 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called. 352 * 353 * rx_handlers are functions called from inside __netif_receive_skb(), to do 354 * special processing of the skb, prior to delivery to protocol handlers. 355 * 356 * Currently, a net_device can only have a single rx_handler registered. Trying 357 * to register a second rx_handler will return -EBUSY. 358 * 359 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 360 * To unregister a rx_handler on a net_device, use 361 * netdev_rx_handler_unregister(). 362 * 363 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 364 * do with the skb. 365 * 366 * If the rx_handler consumed to skb in some way, it should return 367 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 368 * the skb to be delivered in some other ways. 369 * 370 * If the rx_handler changed skb->dev, to divert the skb to another 371 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 372 * new device will be called if it exists. 373 * 374 * If the rx_handler consider the skb should be ignored, it should return 375 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 376 * are registered on exact device (ptype->dev == skb->dev). 377 * 378 * If the rx_handler didn't changed skb->dev, but want the skb to be normally 379 * delivered, it should return RX_HANDLER_PASS. 380 * 381 * A device without a registered rx_handler will behave as if rx_handler 382 * returned RX_HANDLER_PASS. 383 */ 384 385enum rx_handler_result { 386 RX_HANDLER_CONSUMED, 387 RX_HANDLER_ANOTHER, 388 RX_HANDLER_EXACT, 389 RX_HANDLER_PASS, 390}; 391typedef enum rx_handler_result rx_handler_result_t; 392typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 393 394void __napi_schedule(struct napi_struct *n); 395void __napi_schedule_irqoff(struct napi_struct *n); 396 397static inline bool napi_disable_pending(struct napi_struct *n) 398{ 399 return test_bit(NAPI_STATE_DISABLE, &n->state); 400} 401 402/** 403 * napi_schedule_prep - check if napi can be scheduled 404 * @n: napi context 405 * 406 * Test if NAPI routine is already running, and if not mark 407 * it as running. This is used as a condition variable 408 * insure only one NAPI poll instance runs. We also make 409 * sure there is no pending NAPI disable. 410 */ 411static inline bool napi_schedule_prep(struct napi_struct *n) 412{ 413 return !napi_disable_pending(n) && 414 !test_and_set_bit(NAPI_STATE_SCHED, &n->state); 415} 416 417/** 418 * napi_schedule - schedule NAPI poll 419 * @n: napi context 420 * 421 * Schedule NAPI poll routine to be called if it is not already 422 * running. 423 */ 424static inline void napi_schedule(struct napi_struct *n) 425{ 426 if (napi_schedule_prep(n)) 427 __napi_schedule(n); 428} 429 430/** 431 * napi_schedule_irqoff - schedule NAPI poll 432 * @n: napi context 433 * 434 * Variant of napi_schedule(), assuming hard irqs are masked. 435 */ 436static inline void napi_schedule_irqoff(struct napi_struct *n) 437{ 438 if (napi_schedule_prep(n)) 439 __napi_schedule_irqoff(n); 440} 441 442/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 443static inline bool napi_reschedule(struct napi_struct *napi) 444{ 445 if (napi_schedule_prep(napi)) { 446 __napi_schedule(napi); 447 return true; 448 } 449 return false; 450} 451 452void __napi_complete(struct napi_struct *n); 453void napi_complete_done(struct napi_struct *n, int work_done); 454/** 455 * napi_complete - NAPI processing complete 456 * @n: napi context 457 * 458 * Mark NAPI processing as complete. 459 * Consider using napi_complete_done() instead. 460 */ 461static inline void napi_complete(struct napi_struct *n) 462{ 463 return napi_complete_done(n, 0); 464} 465 466/** 467 * napi_hash_add - add a NAPI to global hashtable 468 * @napi: napi context 469 * 470 * generate a new napi_id and store a @napi under it in napi_hash 471 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL) 472 * Note: This is normally automatically done from netif_napi_add(), 473 * so might disappear in a future linux version. 474 */ 475void napi_hash_add(struct napi_struct *napi); 476 477/** 478 * napi_hash_del - remove a NAPI from global table 479 * @napi: napi context 480 * 481 * Warning: caller must observe rcu grace period 482 * before freeing memory containing @napi, if 483 * this function returns true. 484 * Note: core networking stack automatically calls it 485 * from netif_napi_del() 486 * Drivers might want to call this helper to combine all 487 * the needed rcu grace periods into a single one. 488 */ 489bool napi_hash_del(struct napi_struct *napi); 490 491/** 492 * napi_disable - prevent NAPI from scheduling 493 * @n: napi context 494 * 495 * Stop NAPI from being scheduled on this context. 496 * Waits till any outstanding processing completes. 497 */ 498void napi_disable(struct napi_struct *n); 499 500/** 501 * napi_enable - enable NAPI scheduling 502 * @n: napi context 503 * 504 * Resume NAPI from being scheduled on this context. 505 * Must be paired with napi_disable. 506 */ 507static inline void napi_enable(struct napi_struct *n) 508{ 509 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 510 smp_mb__before_atomic(); 511 clear_bit(NAPI_STATE_SCHED, &n->state); 512 clear_bit(NAPI_STATE_NPSVC, &n->state); 513} 514 515/** 516 * napi_synchronize - wait until NAPI is not running 517 * @n: napi context 518 * 519 * Wait until NAPI is done being scheduled on this context. 520 * Waits till any outstanding processing completes but 521 * does not disable future activations. 522 */ 523static inline void napi_synchronize(const struct napi_struct *n) 524{ 525 if (IS_ENABLED(CONFIG_SMP)) 526 while (test_bit(NAPI_STATE_SCHED, &n->state)) 527 msleep(1); 528 else 529 barrier(); 530} 531 532enum netdev_queue_state_t { 533 __QUEUE_STATE_DRV_XOFF, 534 __QUEUE_STATE_STACK_XOFF, 535 __QUEUE_STATE_FROZEN, 536}; 537 538#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 539#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 540#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 541 542#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 543#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 544 QUEUE_STATE_FROZEN) 545#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 546 QUEUE_STATE_FROZEN) 547 548/* 549 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 550 * netif_tx_* functions below are used to manipulate this flag. The 551 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 552 * queue independently. The netif_xmit_*stopped functions below are called 553 * to check if the queue has been stopped by the driver or stack (either 554 * of the XOFF bits are set in the state). Drivers should not need to call 555 * netif_xmit*stopped functions, they should only be using netif_tx_*. 556 */ 557 558struct netdev_queue { 559/* 560 * read mostly part 561 */ 562 struct net_device *dev; 563 struct Qdisc __rcu *qdisc; 564 struct Qdisc *qdisc_sleeping; 565#ifdef CONFIG_SYSFS 566 struct kobject kobj; 567#endif 568#if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 569 int numa_node; 570#endif 571/* 572 * write mostly part 573 */ 574 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 575 int xmit_lock_owner; 576 /* 577 * please use this field instead of dev->trans_start 578 */ 579 unsigned long trans_start; 580 581 /* 582 * Number of TX timeouts for this queue 583 * (/sys/class/net/DEV/Q/trans_timeout) 584 */ 585 unsigned long trans_timeout; 586 587 unsigned long state; 588 589#ifdef CONFIG_BQL 590 struct dql dql; 591#endif 592 unsigned long tx_maxrate; 593} ____cacheline_aligned_in_smp; 594 595static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 596{ 597#if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 598 return q->numa_node; 599#else 600 return NUMA_NO_NODE; 601#endif 602} 603 604static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 605{ 606#if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 607 q->numa_node = node; 608#endif 609} 610 611#ifdef CONFIG_RPS 612/* 613 * This structure holds an RPS map which can be of variable length. The 614 * map is an array of CPUs. 615 */ 616struct rps_map { 617 unsigned int len; 618 struct rcu_head rcu; 619 u16 cpus[0]; 620}; 621#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 622 623/* 624 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 625 * tail pointer for that CPU's input queue at the time of last enqueue, and 626 * a hardware filter index. 627 */ 628struct rps_dev_flow { 629 u16 cpu; 630 u16 filter; 631 unsigned int last_qtail; 632}; 633#define RPS_NO_FILTER 0xffff 634 635/* 636 * The rps_dev_flow_table structure contains a table of flow mappings. 637 */ 638struct rps_dev_flow_table { 639 unsigned int mask; 640 struct rcu_head rcu; 641 struct rps_dev_flow flows[0]; 642}; 643#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 644 ((_num) * sizeof(struct rps_dev_flow))) 645 646/* 647 * The rps_sock_flow_table contains mappings of flows to the last CPU 648 * on which they were processed by the application (set in recvmsg). 649 * Each entry is a 32bit value. Upper part is the high order bits 650 * of flow hash, lower part is cpu number. 651 * rps_cpu_mask is used to partition the space, depending on number of 652 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 653 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f, 654 * meaning we use 32-6=26 bits for the hash. 655 */ 656struct rps_sock_flow_table { 657 u32 mask; 658 659 u32 ents[0] ____cacheline_aligned_in_smp; 660}; 661#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 662 663#define RPS_NO_CPU 0xffff 664 665extern u32 rps_cpu_mask; 666extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 667 668static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 669 u32 hash) 670{ 671 if (table && hash) { 672 unsigned int index = hash & table->mask; 673 u32 val = hash & ~rps_cpu_mask; 674 675 /* We only give a hint, preemption can change cpu under us */ 676 val |= raw_smp_processor_id(); 677 678 if (table->ents[index] != val) 679 table->ents[index] = val; 680 } 681} 682 683#ifdef CONFIG_RFS_ACCEL 684bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 685 u16 filter_id); 686#endif 687#endif /* CONFIG_RPS */ 688 689/* This structure contains an instance of an RX queue. */ 690struct netdev_rx_queue { 691#ifdef CONFIG_RPS 692 struct rps_map __rcu *rps_map; 693 struct rps_dev_flow_table __rcu *rps_flow_table; 694#endif 695 struct kobject kobj; 696 struct net_device *dev; 697} ____cacheline_aligned_in_smp; 698 699/* 700 * RX queue sysfs structures and functions. 701 */ 702struct rx_queue_attribute { 703 struct attribute attr; 704 ssize_t (*show)(struct netdev_rx_queue *queue, 705 struct rx_queue_attribute *attr, char *buf); 706 ssize_t (*store)(struct netdev_rx_queue *queue, 707 struct rx_queue_attribute *attr, const char *buf, size_t len); 708}; 709 710#ifdef CONFIG_XPS 711/* 712 * This structure holds an XPS map which can be of variable length. The 713 * map is an array of queues. 714 */ 715struct xps_map { 716 unsigned int len; 717 unsigned int alloc_len; 718 struct rcu_head rcu; 719 u16 queues[0]; 720}; 721#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 722#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 723 - sizeof(struct xps_map)) / sizeof(u16)) 724 725/* 726 * This structure holds all XPS maps for device. Maps are indexed by CPU. 727 */ 728struct xps_dev_maps { 729 struct rcu_head rcu; 730 struct xps_map __rcu *cpu_map[0]; 731}; 732#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \ 733 (nr_cpu_ids * sizeof(struct xps_map *))) 734#endif /* CONFIG_XPS */ 735 736#define TC_MAX_QUEUE 16 737#define TC_BITMASK 15 738/* HW offloaded queuing disciplines txq count and offset maps */ 739struct netdev_tc_txq { 740 u16 count; 741 u16 offset; 742}; 743 744#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 745/* 746 * This structure is to hold information about the device 747 * configured to run FCoE protocol stack. 748 */ 749struct netdev_fcoe_hbainfo { 750 char manufacturer[64]; 751 char serial_number[64]; 752 char hardware_version[64]; 753 char driver_version[64]; 754 char optionrom_version[64]; 755 char firmware_version[64]; 756 char model[256]; 757 char model_description[256]; 758}; 759#endif 760 761#define MAX_PHYS_ITEM_ID_LEN 32 762 763/* This structure holds a unique identifier to identify some 764 * physical item (port for example) used by a netdevice. 765 */ 766struct netdev_phys_item_id { 767 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 768 unsigned char id_len; 769}; 770 771static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 772 struct netdev_phys_item_id *b) 773{ 774 return a->id_len == b->id_len && 775 memcmp(a->id, b->id, a->id_len) == 0; 776} 777 778typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 779 struct sk_buff *skb); 780 781/* 782 * This structure defines the management hooks for network devices. 783 * The following hooks can be defined; unless noted otherwise, they are 784 * optional and can be filled with a null pointer. 785 * 786 * int (*ndo_init)(struct net_device *dev); 787 * This function is called once when network device is registered. 788 * The network device can use this to any late stage initializaton 789 * or semantic validattion. It can fail with an error code which will 790 * be propogated back to register_netdev 791 * 792 * void (*ndo_uninit)(struct net_device *dev); 793 * This function is called when device is unregistered or when registration 794 * fails. It is not called if init fails. 795 * 796 * int (*ndo_open)(struct net_device *dev); 797 * This function is called when network device transistions to the up 798 * state. 799 * 800 * int (*ndo_stop)(struct net_device *dev); 801 * This function is called when network device transistions to the down 802 * state. 803 * 804 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 805 * struct net_device *dev); 806 * Called when a packet needs to be transmitted. 807 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 808 * the queue before that can happen; it's for obsolete devices and weird 809 * corner cases, but the stack really does a non-trivial amount 810 * of useless work if you return NETDEV_TX_BUSY. 811 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX) 812 * Required can not be NULL. 813 * 814 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 815 * netdev_features_t features); 816 * Adjusts the requested feature flags according to device-specific 817 * constraints, and returns the resulting flags. Must not modify 818 * the device state. 819 * 820 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 821 * void *accel_priv, select_queue_fallback_t fallback); 822 * Called to decide which queue to when device supports multiple 823 * transmit queues. 824 * 825 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 826 * This function is called to allow device receiver to make 827 * changes to configuration when multicast or promiscious is enabled. 828 * 829 * void (*ndo_set_rx_mode)(struct net_device *dev); 830 * This function is called device changes address list filtering. 831 * If driver handles unicast address filtering, it should set 832 * IFF_UNICAST_FLT to its priv_flags. 833 * 834 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 835 * This function is called when the Media Access Control address 836 * needs to be changed. If this interface is not defined, the 837 * mac address can not be changed. 838 * 839 * int (*ndo_validate_addr)(struct net_device *dev); 840 * Test if Media Access Control address is valid for the device. 841 * 842 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 843 * Called when a user request an ioctl which can't be handled by 844 * the generic interface code. If not defined ioctl's return 845 * not supported error code. 846 * 847 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 848 * Used to set network devices bus interface parameters. This interface 849 * is retained for legacy reason, new devices should use the bus 850 * interface (PCI) for low level management. 851 * 852 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 853 * Called when a user wants to change the Maximum Transfer Unit 854 * of a device. If not defined, any request to change MTU will 855 * will return an error. 856 * 857 * void (*ndo_tx_timeout)(struct net_device *dev); 858 * Callback uses when the transmitter has not made any progress 859 * for dev->watchdog ticks. 860 * 861 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev, 862 * struct rtnl_link_stats64 *storage); 863 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 864 * Called when a user wants to get the network device usage 865 * statistics. Drivers must do one of the following: 866 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 867 * rtnl_link_stats64 structure passed by the caller. 868 * 2. Define @ndo_get_stats to update a net_device_stats structure 869 * (which should normally be dev->stats) and return a pointer to 870 * it. The structure may be changed asynchronously only if each 871 * field is written atomically. 872 * 3. Update dev->stats asynchronously and atomically, and define 873 * neither operation. 874 * 875 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 876 * If device support VLAN filtering this function is called when a 877 * VLAN id is registered. 878 * 879 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 880 * If device support VLAN filtering this function is called when a 881 * VLAN id is unregistered. 882 * 883 * void (*ndo_poll_controller)(struct net_device *dev); 884 * 885 * SR-IOV management functions. 886 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 887 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos); 888 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 889 * int max_tx_rate); 890 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 891 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 892 * int (*ndo_get_vf_config)(struct net_device *dev, 893 * int vf, struct ifla_vf_info *ivf); 894 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 895 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 896 * struct nlattr *port[]); 897 * 898 * Enable or disable the VF ability to query its RSS Redirection Table and 899 * Hash Key. This is needed since on some devices VF share this information 900 * with PF and querying it may adduce a theoretical security risk. 901 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 902 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 903 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc) 904 * Called to setup 'tc' number of traffic classes in the net device. This 905 * is always called from the stack with the rtnl lock held and netif tx 906 * queues stopped. This allows the netdevice to perform queue management 907 * safely. 908 * 909 * Fiber Channel over Ethernet (FCoE) offload functions. 910 * int (*ndo_fcoe_enable)(struct net_device *dev); 911 * Called when the FCoE protocol stack wants to start using LLD for FCoE 912 * so the underlying device can perform whatever needed configuration or 913 * initialization to support acceleration of FCoE traffic. 914 * 915 * int (*ndo_fcoe_disable)(struct net_device *dev); 916 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 917 * so the underlying device can perform whatever needed clean-ups to 918 * stop supporting acceleration of FCoE traffic. 919 * 920 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 921 * struct scatterlist *sgl, unsigned int sgc); 922 * Called when the FCoE Initiator wants to initialize an I/O that 923 * is a possible candidate for Direct Data Placement (DDP). The LLD can 924 * perform necessary setup and returns 1 to indicate the device is set up 925 * successfully to perform DDP on this I/O, otherwise this returns 0. 926 * 927 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 928 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 929 * indicated by the FC exchange id 'xid', so the underlying device can 930 * clean up and reuse resources for later DDP requests. 931 * 932 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 933 * struct scatterlist *sgl, unsigned int sgc); 934 * Called when the FCoE Target wants to initialize an I/O that 935 * is a possible candidate for Direct Data Placement (DDP). The LLD can 936 * perform necessary setup and returns 1 to indicate the device is set up 937 * successfully to perform DDP on this I/O, otherwise this returns 0. 938 * 939 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 940 * struct netdev_fcoe_hbainfo *hbainfo); 941 * Called when the FCoE Protocol stack wants information on the underlying 942 * device. This information is utilized by the FCoE protocol stack to 943 * register attributes with Fiber Channel management service as per the 944 * FC-GS Fabric Device Management Information(FDMI) specification. 945 * 946 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 947 * Called when the underlying device wants to override default World Wide 948 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 949 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 950 * protocol stack to use. 951 * 952 * RFS acceleration. 953 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 954 * u16 rxq_index, u32 flow_id); 955 * Set hardware filter for RFS. rxq_index is the target queue index; 956 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 957 * Return the filter ID on success, or a negative error code. 958 * 959 * Slave management functions (for bridge, bonding, etc). 960 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 961 * Called to make another netdev an underling. 962 * 963 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 964 * Called to release previously enslaved netdev. 965 * 966 * Feature/offload setting functions. 967 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 968 * Called to update device configuration to new features. Passed 969 * feature set might be less than what was returned by ndo_fix_features()). 970 * Must return >0 or -errno if it changed dev->features itself. 971 * 972 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 973 * struct net_device *dev, 974 * const unsigned char *addr, u16 vid, u16 flags) 975 * Adds an FDB entry to dev for addr. 976 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 977 * struct net_device *dev, 978 * const unsigned char *addr, u16 vid) 979 * Deletes the FDB entry from dev coresponding to addr. 980 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 981 * struct net_device *dev, struct net_device *filter_dev, 982 * int idx) 983 * Used to add FDB entries to dump requests. Implementers should add 984 * entries to skb and update idx with the number of entries. 985 * 986 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 987 * u16 flags) 988 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 989 * struct net_device *dev, u32 filter_mask, 990 * int nlflags) 991 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 992 * u16 flags); 993 * 994 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 995 * Called to change device carrier. Soft-devices (like dummy, team, etc) 996 * which do not represent real hardware may define this to allow their 997 * userspace components to manage their virtual carrier state. Devices 998 * that determine carrier state from physical hardware properties (eg 999 * network cables) or protocol-dependent mechanisms (eg 1000 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1001 * 1002 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1003 * struct netdev_phys_item_id *ppid); 1004 * Called to get ID of physical port of this device. If driver does 1005 * not implement this, it is assumed that the hw is not able to have 1006 * multiple net devices on single physical port. 1007 * 1008 * void (*ndo_add_vxlan_port)(struct net_device *dev, 1009 * sa_family_t sa_family, __be16 port); 1010 * Called by vxlan to notiy a driver about the UDP port and socket 1011 * address family that vxlan is listnening to. It is called only when 1012 * a new port starts listening. The operation is protected by the 1013 * vxlan_net->sock_lock. 1014 * 1015 * void (*ndo_add_geneve_port)(struct net_device *dev, 1016 * sa_family_t sa_family, __be16 port); 1017 * Called by geneve to notify a driver about the UDP port and socket 1018 * address family that geneve is listnening to. It is called only when 1019 * a new port starts listening. The operation is protected by the 1020 * geneve_net->sock_lock. 1021 * 1022 * void (*ndo_del_geneve_port)(struct net_device *dev, 1023 * sa_family_t sa_family, __be16 port); 1024 * Called by geneve to notify the driver about a UDP port and socket 1025 * address family that geneve is not listening to anymore. The operation 1026 * is protected by the geneve_net->sock_lock. 1027 * 1028 * void (*ndo_del_vxlan_port)(struct net_device *dev, 1029 * sa_family_t sa_family, __be16 port); 1030 * Called by vxlan to notify the driver about a UDP port and socket 1031 * address family that vxlan is not listening to anymore. The operation 1032 * is protected by the vxlan_net->sock_lock. 1033 * 1034 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1035 * struct net_device *dev) 1036 * Called by upper layer devices to accelerate switching or other 1037 * station functionality into hardware. 'pdev is the lowerdev 1038 * to use for the offload and 'dev' is the net device that will 1039 * back the offload. Returns a pointer to the private structure 1040 * the upper layer will maintain. 1041 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1042 * Called by upper layer device to delete the station created 1043 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1044 * the station and priv is the structure returned by the add 1045 * operation. 1046 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb, 1047 * struct net_device *dev, 1048 * void *priv); 1049 * Callback to use for xmit over the accelerated station. This 1050 * is used in place of ndo_start_xmit on accelerated net 1051 * devices. 1052 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb, 1053 * struct net_device *dev 1054 * netdev_features_t features); 1055 * Called by core transmit path to determine if device is capable of 1056 * performing offload operations on a given packet. This is to give 1057 * the device an opportunity to implement any restrictions that cannot 1058 * be otherwise expressed by feature flags. The check is called with 1059 * the set of features that the stack has calculated and it returns 1060 * those the driver believes to be appropriate. 1061 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1062 * int queue_index, u32 maxrate); 1063 * Called when a user wants to set a max-rate limitation of specific 1064 * TX queue. 1065 * int (*ndo_get_iflink)(const struct net_device *dev); 1066 * Called to get the iflink value of this device. 1067 * void (*ndo_change_proto_down)(struct net_device *dev, 1068 * bool proto_down); 1069 * This function is used to pass protocol port error state information 1070 * to the switch driver. The switch driver can react to the proto_down 1071 * by doing a phys down on the associated switch port. 1072 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1073 * This function is used to get egress tunnel information for given skb. 1074 * This is useful for retrieving outer tunnel header parameters while 1075 * sampling packet. 1076 * 1077 */ 1078struct net_device_ops { 1079 int (*ndo_init)(struct net_device *dev); 1080 void (*ndo_uninit)(struct net_device *dev); 1081 int (*ndo_open)(struct net_device *dev); 1082 int (*ndo_stop)(struct net_device *dev); 1083 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1084 struct net_device *dev); 1085 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1086 struct net_device *dev, 1087 netdev_features_t features); 1088 u16 (*ndo_select_queue)(struct net_device *dev, 1089 struct sk_buff *skb, 1090 void *accel_priv, 1091 select_queue_fallback_t fallback); 1092 void (*ndo_change_rx_flags)(struct net_device *dev, 1093 int flags); 1094 void (*ndo_set_rx_mode)(struct net_device *dev); 1095 int (*ndo_set_mac_address)(struct net_device *dev, 1096 void *addr); 1097 int (*ndo_validate_addr)(struct net_device *dev); 1098 int (*ndo_do_ioctl)(struct net_device *dev, 1099 struct ifreq *ifr, int cmd); 1100 int (*ndo_set_config)(struct net_device *dev, 1101 struct ifmap *map); 1102 int (*ndo_change_mtu)(struct net_device *dev, 1103 int new_mtu); 1104 int (*ndo_neigh_setup)(struct net_device *dev, 1105 struct neigh_parms *); 1106 void (*ndo_tx_timeout) (struct net_device *dev); 1107 1108 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev, 1109 struct rtnl_link_stats64 *storage); 1110 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1111 1112 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1113 __be16 proto, u16 vid); 1114 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1115 __be16 proto, u16 vid); 1116#ifdef CONFIG_NET_POLL_CONTROLLER 1117 void (*ndo_poll_controller)(struct net_device *dev); 1118 int (*ndo_netpoll_setup)(struct net_device *dev, 1119 struct netpoll_info *info); 1120 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1121#endif 1122#ifdef CONFIG_NET_RX_BUSY_POLL 1123 int (*ndo_busy_poll)(struct napi_struct *dev); 1124#endif 1125 int (*ndo_set_vf_mac)(struct net_device *dev, 1126 int queue, u8 *mac); 1127 int (*ndo_set_vf_vlan)(struct net_device *dev, 1128 int queue, u16 vlan, u8 qos); 1129 int (*ndo_set_vf_rate)(struct net_device *dev, 1130 int vf, int min_tx_rate, 1131 int max_tx_rate); 1132 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1133 int vf, bool setting); 1134 int (*ndo_set_vf_trust)(struct net_device *dev, 1135 int vf, bool setting); 1136 int (*ndo_get_vf_config)(struct net_device *dev, 1137 int vf, 1138 struct ifla_vf_info *ivf); 1139 int (*ndo_set_vf_link_state)(struct net_device *dev, 1140 int vf, int link_state); 1141 int (*ndo_get_vf_stats)(struct net_device *dev, 1142 int vf, 1143 struct ifla_vf_stats 1144 *vf_stats); 1145 int (*ndo_set_vf_port)(struct net_device *dev, 1146 int vf, 1147 struct nlattr *port[]); 1148 int (*ndo_get_vf_port)(struct net_device *dev, 1149 int vf, struct sk_buff *skb); 1150 int (*ndo_set_vf_rss_query_en)( 1151 struct net_device *dev, 1152 int vf, bool setting); 1153 int (*ndo_setup_tc)(struct net_device *dev, u8 tc); 1154#if IS_ENABLED(CONFIG_FCOE) 1155 int (*ndo_fcoe_enable)(struct net_device *dev); 1156 int (*ndo_fcoe_disable)(struct net_device *dev); 1157 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1158 u16 xid, 1159 struct scatterlist *sgl, 1160 unsigned int sgc); 1161 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1162 u16 xid); 1163 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1164 u16 xid, 1165 struct scatterlist *sgl, 1166 unsigned int sgc); 1167 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1168 struct netdev_fcoe_hbainfo *hbainfo); 1169#endif 1170 1171#if IS_ENABLED(CONFIG_LIBFCOE) 1172#define NETDEV_FCOE_WWNN 0 1173#define NETDEV_FCOE_WWPN 1 1174 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1175 u64 *wwn, int type); 1176#endif 1177 1178#ifdef CONFIG_RFS_ACCEL 1179 int (*ndo_rx_flow_steer)(struct net_device *dev, 1180 const struct sk_buff *skb, 1181 u16 rxq_index, 1182 u32 flow_id); 1183#endif 1184 int (*ndo_add_slave)(struct net_device *dev, 1185 struct net_device *slave_dev); 1186 int (*ndo_del_slave)(struct net_device *dev, 1187 struct net_device *slave_dev); 1188 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1189 netdev_features_t features); 1190 int (*ndo_set_features)(struct net_device *dev, 1191 netdev_features_t features); 1192 int (*ndo_neigh_construct)(struct neighbour *n); 1193 void (*ndo_neigh_destroy)(struct neighbour *n); 1194 1195 int (*ndo_fdb_add)(struct ndmsg *ndm, 1196 struct nlattr *tb[], 1197 struct net_device *dev, 1198 const unsigned char *addr, 1199 u16 vid, 1200 u16 flags); 1201 int (*ndo_fdb_del)(struct ndmsg *ndm, 1202 struct nlattr *tb[], 1203 struct net_device *dev, 1204 const unsigned char *addr, 1205 u16 vid); 1206 int (*ndo_fdb_dump)(struct sk_buff *skb, 1207 struct netlink_callback *cb, 1208 struct net_device *dev, 1209 struct net_device *filter_dev, 1210 int idx); 1211 1212 int (*ndo_bridge_setlink)(struct net_device *dev, 1213 struct nlmsghdr *nlh, 1214 u16 flags); 1215 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1216 u32 pid, u32 seq, 1217 struct net_device *dev, 1218 u32 filter_mask, 1219 int nlflags); 1220 int (*ndo_bridge_dellink)(struct net_device *dev, 1221 struct nlmsghdr *nlh, 1222 u16 flags); 1223 int (*ndo_change_carrier)(struct net_device *dev, 1224 bool new_carrier); 1225 int (*ndo_get_phys_port_id)(struct net_device *dev, 1226 struct netdev_phys_item_id *ppid); 1227 int (*ndo_get_phys_port_name)(struct net_device *dev, 1228 char *name, size_t len); 1229 void (*ndo_add_vxlan_port)(struct net_device *dev, 1230 sa_family_t sa_family, 1231 __be16 port); 1232 void (*ndo_del_vxlan_port)(struct net_device *dev, 1233 sa_family_t sa_family, 1234 __be16 port); 1235 void (*ndo_add_geneve_port)(struct net_device *dev, 1236 sa_family_t sa_family, 1237 __be16 port); 1238 void (*ndo_del_geneve_port)(struct net_device *dev, 1239 sa_family_t sa_family, 1240 __be16 port); 1241 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1242 struct net_device *dev); 1243 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1244 void *priv); 1245 1246 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb, 1247 struct net_device *dev, 1248 void *priv); 1249 int (*ndo_get_lock_subclass)(struct net_device *dev); 1250 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1251 int queue_index, 1252 u32 maxrate); 1253 int (*ndo_get_iflink)(const struct net_device *dev); 1254 int (*ndo_change_proto_down)(struct net_device *dev, 1255 bool proto_down); 1256 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1257 struct sk_buff *skb); 1258}; 1259 1260/** 1261 * enum net_device_priv_flags - &struct net_device priv_flags 1262 * 1263 * These are the &struct net_device, they are only set internally 1264 * by drivers and used in the kernel. These flags are invisible to 1265 * userspace, this means that the order of these flags can change 1266 * during any kernel release. 1267 * 1268 * You should have a pretty good reason to be extending these flags. 1269 * 1270 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1271 * @IFF_EBRIDGE: Ethernet bridging device 1272 * @IFF_BONDING: bonding master or slave 1273 * @IFF_ISATAP: ISATAP interface (RFC4214) 1274 * @IFF_WAN_HDLC: WAN HDLC device 1275 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1276 * release skb->dst 1277 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1278 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1279 * @IFF_MACVLAN_PORT: device used as macvlan port 1280 * @IFF_BRIDGE_PORT: device used as bridge port 1281 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1282 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1283 * @IFF_UNICAST_FLT: Supports unicast filtering 1284 * @IFF_TEAM_PORT: device used as team port 1285 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1286 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1287 * change when it's running 1288 * @IFF_MACVLAN: Macvlan device 1289 * @IFF_L3MDEV_MASTER: device is an L3 master device 1290 * @IFF_NO_QUEUE: device can run without qdisc attached 1291 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1292 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1293 * @IFF_TEAM: device is a team device 1294 */ 1295enum netdev_priv_flags { 1296 IFF_802_1Q_VLAN = 1<<0, 1297 IFF_EBRIDGE = 1<<1, 1298 IFF_BONDING = 1<<2, 1299 IFF_ISATAP = 1<<3, 1300 IFF_WAN_HDLC = 1<<4, 1301 IFF_XMIT_DST_RELEASE = 1<<5, 1302 IFF_DONT_BRIDGE = 1<<6, 1303 IFF_DISABLE_NETPOLL = 1<<7, 1304 IFF_MACVLAN_PORT = 1<<8, 1305 IFF_BRIDGE_PORT = 1<<9, 1306 IFF_OVS_DATAPATH = 1<<10, 1307 IFF_TX_SKB_SHARING = 1<<11, 1308 IFF_UNICAST_FLT = 1<<12, 1309 IFF_TEAM_PORT = 1<<13, 1310 IFF_SUPP_NOFCS = 1<<14, 1311 IFF_LIVE_ADDR_CHANGE = 1<<15, 1312 IFF_MACVLAN = 1<<16, 1313 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1314 IFF_IPVLAN_MASTER = 1<<18, 1315 IFF_IPVLAN_SLAVE = 1<<19, 1316 IFF_L3MDEV_MASTER = 1<<20, 1317 IFF_NO_QUEUE = 1<<21, 1318 IFF_OPENVSWITCH = 1<<22, 1319 IFF_L3MDEV_SLAVE = 1<<23, 1320 IFF_TEAM = 1<<24, 1321}; 1322 1323#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1324#define IFF_EBRIDGE IFF_EBRIDGE 1325#define IFF_BONDING IFF_BONDING 1326#define IFF_ISATAP IFF_ISATAP 1327#define IFF_WAN_HDLC IFF_WAN_HDLC 1328#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1329#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1330#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1331#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1332#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1333#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1334#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1335#define IFF_UNICAST_FLT IFF_UNICAST_FLT 1336#define IFF_TEAM_PORT IFF_TEAM_PORT 1337#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1338#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1339#define IFF_MACVLAN IFF_MACVLAN 1340#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1341#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER 1342#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE 1343#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1344#define IFF_NO_QUEUE IFF_NO_QUEUE 1345#define IFF_OPENVSWITCH IFF_OPENVSWITCH 1346#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1347#define IFF_TEAM IFF_TEAM 1348 1349/** 1350 * struct net_device - The DEVICE structure. 1351 * Actually, this whole structure is a big mistake. It mixes I/O 1352 * data with strictly "high-level" data, and it has to know about 1353 * almost every data structure used in the INET module. 1354 * 1355 * @name: This is the first field of the "visible" part of this structure 1356 * (i.e. as seen by users in the "Space.c" file). It is the name 1357 * of the interface. 1358 * 1359 * @name_hlist: Device name hash chain, please keep it close to name[] 1360 * @ifalias: SNMP alias 1361 * @mem_end: Shared memory end 1362 * @mem_start: Shared memory start 1363 * @base_addr: Device I/O address 1364 * @irq: Device IRQ number 1365 * 1366 * @carrier_changes: Stats to monitor carrier on<->off transitions 1367 * 1368 * @state: Generic network queuing layer state, see netdev_state_t 1369 * @dev_list: The global list of network devices 1370 * @napi_list: List entry, that is used for polling napi devices 1371 * @unreg_list: List entry, that is used, when we are unregistering the 1372 * device, see the function unregister_netdev 1373 * @close_list: List entry, that is used, when we are closing the device 1374 * 1375 * @adj_list: Directly linked devices, like slaves for bonding 1376 * @all_adj_list: All linked devices, *including* neighbours 1377 * @features: Currently active device features 1378 * @hw_features: User-changeable features 1379 * 1380 * @wanted_features: User-requested features 1381 * @vlan_features: Mask of features inheritable by VLAN devices 1382 * 1383 * @hw_enc_features: Mask of features inherited by encapsulating devices 1384 * This field indicates what encapsulation 1385 * offloads the hardware is capable of doing, 1386 * and drivers will need to set them appropriately. 1387 * 1388 * @mpls_features: Mask of features inheritable by MPLS 1389 * 1390 * @ifindex: interface index 1391 * @group: The group, that the device belongs to 1392 * 1393 * @stats: Statistics struct, which was left as a legacy, use 1394 * rtnl_link_stats64 instead 1395 * 1396 * @rx_dropped: Dropped packets by core network, 1397 * do not use this in drivers 1398 * @tx_dropped: Dropped packets by core network, 1399 * do not use this in drivers 1400 * 1401 * @wireless_handlers: List of functions to handle Wireless Extensions, 1402 * instead of ioctl, 1403 * see <net/iw_handler.h> for details. 1404 * @wireless_data: Instance data managed by the core of wireless extensions 1405 * 1406 * @netdev_ops: Includes several pointers to callbacks, 1407 * if one wants to override the ndo_*() functions 1408 * @ethtool_ops: Management operations 1409 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1410 * of Layer 2 headers. 1411 * 1412 * @flags: Interface flags (a la BSD) 1413 * @priv_flags: Like 'flags' but invisible to userspace, 1414 * see if.h for the definitions 1415 * @gflags: Global flags ( kept as legacy ) 1416 * @padded: How much padding added by alloc_netdev() 1417 * @operstate: RFC2863 operstate 1418 * @link_mode: Mapping policy to operstate 1419 * @if_port: Selectable AUI, TP, ... 1420 * @dma: DMA channel 1421 * @mtu: Interface MTU value 1422 * @type: Interface hardware type 1423 * @hard_header_len: Hardware header length, which means that this is the 1424 * minimum size of a packet. 1425 * 1426 * @needed_headroom: Extra headroom the hardware may need, but not in all 1427 * cases can this be guaranteed 1428 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1429 * cases can this be guaranteed. Some cases also use 1430 * LL_MAX_HEADER instead to allocate the skb 1431 * 1432 * interface address info: 1433 * 1434 * @perm_addr: Permanent hw address 1435 * @addr_assign_type: Hw address assignment type 1436 * @addr_len: Hardware address length 1437 * @neigh_priv_len; Used in neigh_alloc(), 1438 * initialized only in atm/clip.c 1439 * @dev_id: Used to differentiate devices that share 1440 * the same link layer address 1441 * @dev_port: Used to differentiate devices that share 1442 * the same function 1443 * @addr_list_lock: XXX: need comments on this one 1444 * @uc_promisc: Counter, that indicates, that promiscuous mode 1445 * has been enabled due to the need to listen to 1446 * additional unicast addresses in a device that 1447 * does not implement ndo_set_rx_mode() 1448 * @uc: unicast mac addresses 1449 * @mc: multicast mac addresses 1450 * @dev_addrs: list of device hw addresses 1451 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1452 * @promiscuity: Number of times, the NIC is told to work in 1453 * Promiscuous mode, if it becomes 0 the NIC will 1454 * exit from working in Promiscuous mode 1455 * @allmulti: Counter, enables or disables allmulticast mode 1456 * 1457 * @vlan_info: VLAN info 1458 * @dsa_ptr: dsa specific data 1459 * @tipc_ptr: TIPC specific data 1460 * @atalk_ptr: AppleTalk link 1461 * @ip_ptr: IPv4 specific data 1462 * @dn_ptr: DECnet specific data 1463 * @ip6_ptr: IPv6 specific data 1464 * @ax25_ptr: AX.25 specific data 1465 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1466 * 1467 * @last_rx: Time of last Rx 1468 * @dev_addr: Hw address (before bcast, 1469 * because most packets are unicast) 1470 * 1471 * @_rx: Array of RX queues 1472 * @num_rx_queues: Number of RX queues 1473 * allocated at register_netdev() time 1474 * @real_num_rx_queues: Number of RX queues currently active in device 1475 * 1476 * @rx_handler: handler for received packets 1477 * @rx_handler_data: XXX: need comments on this one 1478 * @ingress_queue: XXX: need comments on this one 1479 * @broadcast: hw bcast address 1480 * 1481 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1482 * indexed by RX queue number. Assigned by driver. 1483 * This must only be set if the ndo_rx_flow_steer 1484 * operation is defined 1485 * @index_hlist: Device index hash chain 1486 * 1487 * @_tx: Array of TX queues 1488 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1489 * @real_num_tx_queues: Number of TX queues currently active in device 1490 * @qdisc: Root qdisc from userspace point of view 1491 * @tx_queue_len: Max frames per queue allowed 1492 * @tx_global_lock: XXX: need comments on this one 1493 * 1494 * @xps_maps: XXX: need comments on this one 1495 * 1496 * @offload_fwd_mark: Offload device fwding mark 1497 * 1498 * @trans_start: Time (in jiffies) of last Tx 1499 * @watchdog_timeo: Represents the timeout that is used by 1500 * the watchdog ( see dev_watchdog() ) 1501 * @watchdog_timer: List of timers 1502 * 1503 * @pcpu_refcnt: Number of references to this device 1504 * @todo_list: Delayed register/unregister 1505 * @link_watch_list: XXX: need comments on this one 1506 * 1507 * @reg_state: Register/unregister state machine 1508 * @dismantle: Device is going to be freed 1509 * @rtnl_link_state: This enum represents the phases of creating 1510 * a new link 1511 * 1512 * @destructor: Called from unregister, 1513 * can be used to call free_netdev 1514 * @npinfo: XXX: need comments on this one 1515 * @nd_net: Network namespace this network device is inside 1516 * 1517 * @ml_priv: Mid-layer private 1518 * @lstats: Loopback statistics 1519 * @tstats: Tunnel statistics 1520 * @dstats: Dummy statistics 1521 * @vstats: Virtual ethernet statistics 1522 * 1523 * @garp_port: GARP 1524 * @mrp_port: MRP 1525 * 1526 * @dev: Class/net/name entry 1527 * @sysfs_groups: Space for optional device, statistics and wireless 1528 * sysfs groups 1529 * 1530 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1531 * @rtnl_link_ops: Rtnl_link_ops 1532 * 1533 * @gso_max_size: Maximum size of generic segmentation offload 1534 * @gso_max_segs: Maximum number of segments that can be passed to the 1535 * NIC for GSO 1536 * @gso_min_segs: Minimum number of segments that can be passed to the 1537 * NIC for GSO 1538 * 1539 * @dcbnl_ops: Data Center Bridging netlink ops 1540 * @num_tc: Number of traffic classes in the net device 1541 * @tc_to_txq: XXX: need comments on this one 1542 * @prio_tc_map XXX: need comments on this one 1543 * 1544 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1545 * 1546 * @priomap: XXX: need comments on this one 1547 * @phydev: Physical device may attach itself 1548 * for hardware timestamping 1549 * 1550 * @qdisc_tx_busylock: XXX: need comments on this one 1551 * 1552 * @proto_down: protocol port state information can be sent to the 1553 * switch driver and used to set the phys state of the 1554 * switch port. 1555 * 1556 * FIXME: cleanup struct net_device such that network protocol info 1557 * moves out. 1558 */ 1559 1560struct net_device { 1561 char name[IFNAMSIZ]; 1562 struct hlist_node name_hlist; 1563 char *ifalias; 1564 /* 1565 * I/O specific fields 1566 * FIXME: Merge these and struct ifmap into one 1567 */ 1568 unsigned long mem_end; 1569 unsigned long mem_start; 1570 unsigned long base_addr; 1571 int irq; 1572 1573 atomic_t carrier_changes; 1574 1575 /* 1576 * Some hardware also needs these fields (state,dev_list, 1577 * napi_list,unreg_list,close_list) but they are not 1578 * part of the usual set specified in Space.c. 1579 */ 1580 1581 unsigned long state; 1582 1583 struct list_head dev_list; 1584 struct list_head napi_list; 1585 struct list_head unreg_list; 1586 struct list_head close_list; 1587 struct list_head ptype_all; 1588 struct list_head ptype_specific; 1589 1590 struct { 1591 struct list_head upper; 1592 struct list_head lower; 1593 } adj_list; 1594 1595 struct { 1596 struct list_head upper; 1597 struct list_head lower; 1598 } all_adj_list; 1599 1600 netdev_features_t features; 1601 netdev_features_t hw_features; 1602 netdev_features_t wanted_features; 1603 netdev_features_t vlan_features; 1604 netdev_features_t hw_enc_features; 1605 netdev_features_t mpls_features; 1606 1607 int ifindex; 1608 int group; 1609 1610 struct net_device_stats stats; 1611 1612 atomic_long_t rx_dropped; 1613 atomic_long_t tx_dropped; 1614 1615#ifdef CONFIG_WIRELESS_EXT 1616 const struct iw_handler_def * wireless_handlers; 1617 struct iw_public_data * wireless_data; 1618#endif 1619 const struct net_device_ops *netdev_ops; 1620 const struct ethtool_ops *ethtool_ops; 1621#ifdef CONFIG_NET_SWITCHDEV 1622 const struct switchdev_ops *switchdev_ops; 1623#endif 1624#ifdef CONFIG_NET_L3_MASTER_DEV 1625 const struct l3mdev_ops *l3mdev_ops; 1626#endif 1627 1628 const struct header_ops *header_ops; 1629 1630 unsigned int flags; 1631 unsigned int priv_flags; 1632 1633 unsigned short gflags; 1634 unsigned short padded; 1635 1636 unsigned char operstate; 1637 unsigned char link_mode; 1638 1639 unsigned char if_port; 1640 unsigned char dma; 1641 1642 unsigned int mtu; 1643 unsigned short type; 1644 unsigned short hard_header_len; 1645 1646 unsigned short needed_headroom; 1647 unsigned short needed_tailroom; 1648 1649 /* Interface address info. */ 1650 unsigned char perm_addr[MAX_ADDR_LEN]; 1651 unsigned char addr_assign_type; 1652 unsigned char addr_len; 1653 unsigned short neigh_priv_len; 1654 unsigned short dev_id; 1655 unsigned short dev_port; 1656 spinlock_t addr_list_lock; 1657 unsigned char name_assign_type; 1658 bool uc_promisc; 1659 struct netdev_hw_addr_list uc; 1660 struct netdev_hw_addr_list mc; 1661 struct netdev_hw_addr_list dev_addrs; 1662 1663#ifdef CONFIG_SYSFS 1664 struct kset *queues_kset; 1665#endif 1666 unsigned int promiscuity; 1667 unsigned int allmulti; 1668 1669 1670 /* Protocol specific pointers */ 1671 1672#if IS_ENABLED(CONFIG_VLAN_8021Q) 1673 struct vlan_info __rcu *vlan_info; 1674#endif 1675#if IS_ENABLED(CONFIG_NET_DSA) 1676 struct dsa_switch_tree *dsa_ptr; 1677#endif 1678#if IS_ENABLED(CONFIG_TIPC) 1679 struct tipc_bearer __rcu *tipc_ptr; 1680#endif 1681 void *atalk_ptr; 1682 struct in_device __rcu *ip_ptr; 1683 struct dn_dev __rcu *dn_ptr; 1684 struct inet6_dev __rcu *ip6_ptr; 1685 void *ax25_ptr; 1686 struct wireless_dev *ieee80211_ptr; 1687 struct wpan_dev *ieee802154_ptr; 1688#if IS_ENABLED(CONFIG_MPLS_ROUTING) 1689 struct mpls_dev __rcu *mpls_ptr; 1690#endif 1691 1692/* 1693 * Cache lines mostly used on receive path (including eth_type_trans()) 1694 */ 1695 unsigned long last_rx; 1696 1697 /* Interface address info used in eth_type_trans() */ 1698 unsigned char *dev_addr; 1699 1700 1701#ifdef CONFIG_SYSFS 1702 struct netdev_rx_queue *_rx; 1703 1704 unsigned int num_rx_queues; 1705 unsigned int real_num_rx_queues; 1706 1707#endif 1708 1709 unsigned long gro_flush_timeout; 1710 rx_handler_func_t __rcu *rx_handler; 1711 void __rcu *rx_handler_data; 1712 1713#ifdef CONFIG_NET_CLS_ACT 1714 struct tcf_proto __rcu *ingress_cl_list; 1715#endif 1716 struct netdev_queue __rcu *ingress_queue; 1717#ifdef CONFIG_NETFILTER_INGRESS 1718 struct list_head nf_hooks_ingress; 1719#endif 1720 1721 unsigned char broadcast[MAX_ADDR_LEN]; 1722#ifdef CONFIG_RFS_ACCEL 1723 struct cpu_rmap *rx_cpu_rmap; 1724#endif 1725 struct hlist_node index_hlist; 1726 1727/* 1728 * Cache lines mostly used on transmit path 1729 */ 1730 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1731 unsigned int num_tx_queues; 1732 unsigned int real_num_tx_queues; 1733 struct Qdisc *qdisc; 1734 unsigned long tx_queue_len; 1735 spinlock_t tx_global_lock; 1736 int watchdog_timeo; 1737 1738#ifdef CONFIG_XPS 1739 struct xps_dev_maps __rcu *xps_maps; 1740#endif 1741#ifdef CONFIG_NET_CLS_ACT 1742 struct tcf_proto __rcu *egress_cl_list; 1743#endif 1744#ifdef CONFIG_NET_SWITCHDEV 1745 u32 offload_fwd_mark; 1746#endif 1747 1748 /* These may be needed for future network-power-down code. */ 1749 1750 /* 1751 * trans_start here is expensive for high speed devices on SMP, 1752 * please use netdev_queue->trans_start instead. 1753 */ 1754 unsigned long trans_start; 1755 1756 struct timer_list watchdog_timer; 1757 1758 int __percpu *pcpu_refcnt; 1759 struct list_head todo_list; 1760 1761 struct list_head link_watch_list; 1762 1763 enum { NETREG_UNINITIALIZED=0, 1764 NETREG_REGISTERED, /* completed register_netdevice */ 1765 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1766 NETREG_UNREGISTERED, /* completed unregister todo */ 1767 NETREG_RELEASED, /* called free_netdev */ 1768 NETREG_DUMMY, /* dummy device for NAPI poll */ 1769 } reg_state:8; 1770 1771 bool dismantle; 1772 1773 enum { 1774 RTNL_LINK_INITIALIZED, 1775 RTNL_LINK_INITIALIZING, 1776 } rtnl_link_state:16; 1777 1778 void (*destructor)(struct net_device *dev); 1779 1780#ifdef CONFIG_NETPOLL 1781 struct netpoll_info __rcu *npinfo; 1782#endif 1783 1784 possible_net_t nd_net; 1785 1786 /* mid-layer private */ 1787 union { 1788 void *ml_priv; 1789 struct pcpu_lstats __percpu *lstats; 1790 struct pcpu_sw_netstats __percpu *tstats; 1791 struct pcpu_dstats __percpu *dstats; 1792 struct pcpu_vstats __percpu *vstats; 1793 }; 1794 1795 struct garp_port __rcu *garp_port; 1796 struct mrp_port __rcu *mrp_port; 1797 1798 struct device dev; 1799 const struct attribute_group *sysfs_groups[4]; 1800 const struct attribute_group *sysfs_rx_queue_group; 1801 1802 const struct rtnl_link_ops *rtnl_link_ops; 1803 1804 /* for setting kernel sock attribute on TCP connection setup */ 1805#define GSO_MAX_SIZE 65536 1806 unsigned int gso_max_size; 1807#define GSO_MAX_SEGS 65535 1808 u16 gso_max_segs; 1809 u16 gso_min_segs; 1810#ifdef CONFIG_DCB 1811 const struct dcbnl_rtnl_ops *dcbnl_ops; 1812#endif 1813 u8 num_tc; 1814 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 1815 u8 prio_tc_map[TC_BITMASK + 1]; 1816 1817#if IS_ENABLED(CONFIG_FCOE) 1818 unsigned int fcoe_ddp_xid; 1819#endif 1820#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 1821 struct netprio_map __rcu *priomap; 1822#endif 1823 struct phy_device *phydev; 1824 struct lock_class_key *qdisc_tx_busylock; 1825 bool proto_down; 1826}; 1827#define to_net_dev(d) container_of(d, struct net_device, dev) 1828 1829#define NETDEV_ALIGN 32 1830 1831static inline 1832int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 1833{ 1834 return dev->prio_tc_map[prio & TC_BITMASK]; 1835} 1836 1837static inline 1838int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 1839{ 1840 if (tc >= dev->num_tc) 1841 return -EINVAL; 1842 1843 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 1844 return 0; 1845} 1846 1847static inline 1848void netdev_reset_tc(struct net_device *dev) 1849{ 1850 dev->num_tc = 0; 1851 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq)); 1852 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map)); 1853} 1854 1855static inline 1856int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset) 1857{ 1858 if (tc >= dev->num_tc) 1859 return -EINVAL; 1860 1861 dev->tc_to_txq[tc].count = count; 1862 dev->tc_to_txq[tc].offset = offset; 1863 return 0; 1864} 1865 1866static inline 1867int netdev_set_num_tc(struct net_device *dev, u8 num_tc) 1868{ 1869 if (num_tc > TC_MAX_QUEUE) 1870 return -EINVAL; 1871 1872 dev->num_tc = num_tc; 1873 return 0; 1874} 1875 1876static inline 1877int netdev_get_num_tc(struct net_device *dev) 1878{ 1879 return dev->num_tc; 1880} 1881 1882static inline 1883struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 1884 unsigned int index) 1885{ 1886 return &dev->_tx[index]; 1887} 1888 1889static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 1890 const struct sk_buff *skb) 1891{ 1892 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 1893} 1894 1895static inline void netdev_for_each_tx_queue(struct net_device *dev, 1896 void (*f)(struct net_device *, 1897 struct netdev_queue *, 1898 void *), 1899 void *arg) 1900{ 1901 unsigned int i; 1902 1903 for (i = 0; i < dev->num_tx_queues; i++) 1904 f(dev, &dev->_tx[i], arg); 1905} 1906 1907struct netdev_queue *netdev_pick_tx(struct net_device *dev, 1908 struct sk_buff *skb, 1909 void *accel_priv); 1910 1911/* 1912 * Net namespace inlines 1913 */ 1914static inline 1915struct net *dev_net(const struct net_device *dev) 1916{ 1917 return read_pnet(&dev->nd_net); 1918} 1919 1920static inline 1921void dev_net_set(struct net_device *dev, struct net *net) 1922{ 1923 write_pnet(&dev->nd_net, net); 1924} 1925 1926static inline bool netdev_uses_dsa(struct net_device *dev) 1927{ 1928#if IS_ENABLED(CONFIG_NET_DSA) 1929 if (dev->dsa_ptr != NULL) 1930 return dsa_uses_tagged_protocol(dev->dsa_ptr); 1931#endif 1932 return false; 1933} 1934 1935/** 1936 * netdev_priv - access network device private data 1937 * @dev: network device 1938 * 1939 * Get network device private data 1940 */ 1941static inline void *netdev_priv(const struct net_device *dev) 1942{ 1943 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 1944} 1945 1946/* Set the sysfs physical device reference for the network logical device 1947 * if set prior to registration will cause a symlink during initialization. 1948 */ 1949#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 1950 1951/* Set the sysfs device type for the network logical device to allow 1952 * fine-grained identification of different network device types. For 1953 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc. 1954 */ 1955#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 1956 1957/* Default NAPI poll() weight 1958 * Device drivers are strongly advised to not use bigger value 1959 */ 1960#define NAPI_POLL_WEIGHT 64 1961 1962/** 1963 * netif_napi_add - initialize a napi context 1964 * @dev: network device 1965 * @napi: napi context 1966 * @poll: polling function 1967 * @weight: default weight 1968 * 1969 * netif_napi_add() must be used to initialize a napi context prior to calling 1970 * *any* of the other napi related functions. 1971 */ 1972void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 1973 int (*poll)(struct napi_struct *, int), int weight); 1974 1975/** 1976 * netif_tx_napi_add - initialize a napi context 1977 * @dev: network device 1978 * @napi: napi context 1979 * @poll: polling function 1980 * @weight: default weight 1981 * 1982 * This variant of netif_napi_add() should be used from drivers using NAPI 1983 * to exclusively poll a TX queue. 1984 * This will avoid we add it into napi_hash[], thus polluting this hash table. 1985 */ 1986static inline void netif_tx_napi_add(struct net_device *dev, 1987 struct napi_struct *napi, 1988 int (*poll)(struct napi_struct *, int), 1989 int weight) 1990{ 1991 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 1992 netif_napi_add(dev, napi, poll, weight); 1993} 1994 1995/** 1996 * netif_napi_del - remove a napi context 1997 * @napi: napi context 1998 * 1999 * netif_napi_del() removes a napi context from the network device napi list 2000 */ 2001void netif_napi_del(struct napi_struct *napi); 2002 2003struct napi_gro_cb { 2004 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2005 void *frag0; 2006 2007 /* Length of frag0. */ 2008 unsigned int frag0_len; 2009 2010 /* This indicates where we are processing relative to skb->data. */ 2011 int data_offset; 2012 2013 /* This is non-zero if the packet cannot be merged with the new skb. */ 2014 u16 flush; 2015 2016 /* Save the IP ID here and check when we get to the transport layer */ 2017 u16 flush_id; 2018 2019 /* Number of segments aggregated. */ 2020 u16 count; 2021 2022 /* Start offset for remote checksum offload */ 2023 u16 gro_remcsum_start; 2024 2025 /* jiffies when first packet was created/queued */ 2026 unsigned long age; 2027 2028 /* Used in ipv6_gro_receive() and foo-over-udp */ 2029 u16 proto; 2030 2031 /* This is non-zero if the packet may be of the same flow. */ 2032 u8 same_flow:1; 2033 2034 /* Used in udp_gro_receive */ 2035 u8 udp_mark:1; 2036 2037 /* GRO checksum is valid */ 2038 u8 csum_valid:1; 2039 2040 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2041 u8 csum_cnt:3; 2042 2043 /* Free the skb? */ 2044 u8 free:2; 2045#define NAPI_GRO_FREE 1 2046#define NAPI_GRO_FREE_STOLEN_HEAD 2 2047 2048 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2049 u8 is_ipv6:1; 2050 2051 /* 7 bit hole */ 2052 2053 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2054 __wsum csum; 2055 2056 /* used in skb_gro_receive() slow path */ 2057 struct sk_buff *last; 2058}; 2059 2060#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2061 2062struct packet_type { 2063 __be16 type; /* This is really htons(ether_type). */ 2064 struct net_device *dev; /* NULL is wildcarded here */ 2065 int (*func) (struct sk_buff *, 2066 struct net_device *, 2067 struct packet_type *, 2068 struct net_device *); 2069 bool (*id_match)(struct packet_type *ptype, 2070 struct sock *sk); 2071 void *af_packet_priv; 2072 struct list_head list; 2073}; 2074 2075struct offload_callbacks { 2076 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2077 netdev_features_t features); 2078 struct sk_buff **(*gro_receive)(struct sk_buff **head, 2079 struct sk_buff *skb); 2080 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2081}; 2082 2083struct packet_offload { 2084 __be16 type; /* This is really htons(ether_type). */ 2085 u16 priority; 2086 struct offload_callbacks callbacks; 2087 struct list_head list; 2088}; 2089 2090struct udp_offload; 2091 2092struct udp_offload_callbacks { 2093 struct sk_buff **(*gro_receive)(struct sk_buff **head, 2094 struct sk_buff *skb, 2095 struct udp_offload *uoff); 2096 int (*gro_complete)(struct sk_buff *skb, 2097 int nhoff, 2098 struct udp_offload *uoff); 2099}; 2100 2101struct udp_offload { 2102 __be16 port; 2103 u8 ipproto; 2104 struct udp_offload_callbacks callbacks; 2105}; 2106 2107/* often modified stats are per cpu, other are shared (netdev->stats) */ 2108struct pcpu_sw_netstats { 2109 u64 rx_packets; 2110 u64 rx_bytes; 2111 u64 tx_packets; 2112 u64 tx_bytes; 2113 struct u64_stats_sync syncp; 2114}; 2115 2116#define __netdev_alloc_pcpu_stats(type, gfp) \ 2117({ \ 2118 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2119 if (pcpu_stats) { \ 2120 int __cpu; \ 2121 for_each_possible_cpu(__cpu) { \ 2122 typeof(type) *stat; \ 2123 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2124 u64_stats_init(&stat->syncp); \ 2125 } \ 2126 } \ 2127 pcpu_stats; \ 2128}) 2129 2130#define netdev_alloc_pcpu_stats(type) \ 2131 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2132 2133enum netdev_lag_tx_type { 2134 NETDEV_LAG_TX_TYPE_UNKNOWN, 2135 NETDEV_LAG_TX_TYPE_RANDOM, 2136 NETDEV_LAG_TX_TYPE_BROADCAST, 2137 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2138 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2139 NETDEV_LAG_TX_TYPE_HASH, 2140}; 2141 2142struct netdev_lag_upper_info { 2143 enum netdev_lag_tx_type tx_type; 2144}; 2145 2146struct netdev_lag_lower_state_info { 2147 u8 link_up : 1, 2148 tx_enabled : 1; 2149}; 2150 2151#include <linux/notifier.h> 2152 2153/* netdevice notifier chain. Please remember to update the rtnetlink 2154 * notification exclusion list in rtnetlink_event() when adding new 2155 * types. 2156 */ 2157#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */ 2158#define NETDEV_DOWN 0x0002 2159#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface 2160 detected a hardware crash and restarted 2161 - we can use this eg to kick tcp sessions 2162 once done */ 2163#define NETDEV_CHANGE 0x0004 /* Notify device state change */ 2164#define NETDEV_REGISTER 0x0005 2165#define NETDEV_UNREGISTER 0x0006 2166#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */ 2167#define NETDEV_CHANGEADDR 0x0008 2168#define NETDEV_GOING_DOWN 0x0009 2169#define NETDEV_CHANGENAME 0x000A 2170#define NETDEV_FEAT_CHANGE 0x000B 2171#define NETDEV_BONDING_FAILOVER 0x000C 2172#define NETDEV_PRE_UP 0x000D 2173#define NETDEV_PRE_TYPE_CHANGE 0x000E 2174#define NETDEV_POST_TYPE_CHANGE 0x000F 2175#define NETDEV_POST_INIT 0x0010 2176#define NETDEV_UNREGISTER_FINAL 0x0011 2177#define NETDEV_RELEASE 0x0012 2178#define NETDEV_NOTIFY_PEERS 0x0013 2179#define NETDEV_JOIN 0x0014 2180#define NETDEV_CHANGEUPPER 0x0015 2181#define NETDEV_RESEND_IGMP 0x0016 2182#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */ 2183#define NETDEV_CHANGEINFODATA 0x0018 2184#define NETDEV_BONDING_INFO 0x0019 2185#define NETDEV_PRECHANGEUPPER 0x001A 2186#define NETDEV_CHANGELOWERSTATE 0x001B 2187 2188int register_netdevice_notifier(struct notifier_block *nb); 2189int unregister_netdevice_notifier(struct notifier_block *nb); 2190 2191struct netdev_notifier_info { 2192 struct net_device *dev; 2193}; 2194 2195struct netdev_notifier_change_info { 2196 struct netdev_notifier_info info; /* must be first */ 2197 unsigned int flags_changed; 2198}; 2199 2200struct netdev_notifier_changeupper_info { 2201 struct netdev_notifier_info info; /* must be first */ 2202 struct net_device *upper_dev; /* new upper dev */ 2203 bool master; /* is upper dev master */ 2204 bool linking; /* is the nofication for link or unlink */ 2205 void *upper_info; /* upper dev info */ 2206}; 2207 2208struct netdev_notifier_changelowerstate_info { 2209 struct netdev_notifier_info info; /* must be first */ 2210 void *lower_state_info; /* is lower dev state */ 2211}; 2212 2213static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2214 struct net_device *dev) 2215{ 2216 info->dev = dev; 2217} 2218 2219static inline struct net_device * 2220netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2221{ 2222 return info->dev; 2223} 2224 2225int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2226 2227 2228extern rwlock_t dev_base_lock; /* Device list lock */ 2229 2230#define for_each_netdev(net, d) \ 2231 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2232#define for_each_netdev_reverse(net, d) \ 2233 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2234#define for_each_netdev_rcu(net, d) \ 2235 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2236#define for_each_netdev_safe(net, d, n) \ 2237 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2238#define for_each_netdev_continue(net, d) \ 2239 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2240#define for_each_netdev_continue_rcu(net, d) \ 2241 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2242#define for_each_netdev_in_bond_rcu(bond, slave) \ 2243 for_each_netdev_rcu(&init_net, slave) \ 2244 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2245#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2246 2247static inline struct net_device *next_net_device(struct net_device *dev) 2248{ 2249 struct list_head *lh; 2250 struct net *net; 2251 2252 net = dev_net(dev); 2253 lh = dev->dev_list.next; 2254 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2255} 2256 2257static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2258{ 2259 struct list_head *lh; 2260 struct net *net; 2261 2262 net = dev_net(dev); 2263 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2264 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2265} 2266 2267static inline struct net_device *first_net_device(struct net *net) 2268{ 2269 return list_empty(&net->dev_base_head) ? NULL : 2270 net_device_entry(net->dev_base_head.next); 2271} 2272 2273static inline struct net_device *first_net_device_rcu(struct net *net) 2274{ 2275 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2276 2277 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2278} 2279 2280int netdev_boot_setup_check(struct net_device *dev); 2281unsigned long netdev_boot_base(const char *prefix, int unit); 2282struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2283 const char *hwaddr); 2284struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2285struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2286void dev_add_pack(struct packet_type *pt); 2287void dev_remove_pack(struct packet_type *pt); 2288void __dev_remove_pack(struct packet_type *pt); 2289void dev_add_offload(struct packet_offload *po); 2290void dev_remove_offload(struct packet_offload *po); 2291 2292int dev_get_iflink(const struct net_device *dev); 2293int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2294struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2295 unsigned short mask); 2296struct net_device *dev_get_by_name(struct net *net, const char *name); 2297struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2298struct net_device *__dev_get_by_name(struct net *net, const char *name); 2299int dev_alloc_name(struct net_device *dev, const char *name); 2300int dev_open(struct net_device *dev); 2301int dev_close(struct net_device *dev); 2302int dev_close_many(struct list_head *head, bool unlink); 2303void dev_disable_lro(struct net_device *dev); 2304int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2305int dev_queue_xmit(struct sk_buff *skb); 2306int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv); 2307int register_netdevice(struct net_device *dev); 2308void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2309void unregister_netdevice_many(struct list_head *head); 2310static inline void unregister_netdevice(struct net_device *dev) 2311{ 2312 unregister_netdevice_queue(dev, NULL); 2313} 2314 2315int netdev_refcnt_read(const struct net_device *dev); 2316void free_netdev(struct net_device *dev); 2317void netdev_freemem(struct net_device *dev); 2318void synchronize_net(void); 2319int init_dummy_netdev(struct net_device *dev); 2320 2321DECLARE_PER_CPU(int, xmit_recursion); 2322static inline int dev_recursion_level(void) 2323{ 2324 return this_cpu_read(xmit_recursion); 2325} 2326 2327struct net_device *dev_get_by_index(struct net *net, int ifindex); 2328struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2329struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2330int netdev_get_name(struct net *net, char *name, int ifindex); 2331int dev_restart(struct net_device *dev); 2332int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb); 2333 2334static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2335{ 2336 return NAPI_GRO_CB(skb)->data_offset; 2337} 2338 2339static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2340{ 2341 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2342} 2343 2344static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2345{ 2346 NAPI_GRO_CB(skb)->data_offset += len; 2347} 2348 2349static inline void *skb_gro_header_fast(struct sk_buff *skb, 2350 unsigned int offset) 2351{ 2352 return NAPI_GRO_CB(skb)->frag0 + offset; 2353} 2354 2355static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2356{ 2357 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2358} 2359 2360static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2361 unsigned int offset) 2362{ 2363 if (!pskb_may_pull(skb, hlen)) 2364 return NULL; 2365 2366 NAPI_GRO_CB(skb)->frag0 = NULL; 2367 NAPI_GRO_CB(skb)->frag0_len = 0; 2368 return skb->data + offset; 2369} 2370 2371static inline void *skb_gro_network_header(struct sk_buff *skb) 2372{ 2373 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2374 skb_network_offset(skb); 2375} 2376 2377static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2378 const void *start, unsigned int len) 2379{ 2380 if (NAPI_GRO_CB(skb)->csum_valid) 2381 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2382 csum_partial(start, len, 0)); 2383} 2384 2385/* GRO checksum functions. These are logical equivalents of the normal 2386 * checksum functions (in skbuff.h) except that they operate on the GRO 2387 * offsets and fields in sk_buff. 2388 */ 2389 2390__sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2391 2392static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2393{ 2394 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2395} 2396 2397static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2398 bool zero_okay, 2399 __sum16 check) 2400{ 2401 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2402 skb_checksum_start_offset(skb) < 2403 skb_gro_offset(skb)) && 2404 !skb_at_gro_remcsum_start(skb) && 2405 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2406 (!zero_okay || check)); 2407} 2408 2409static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2410 __wsum psum) 2411{ 2412 if (NAPI_GRO_CB(skb)->csum_valid && 2413 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2414 return 0; 2415 2416 NAPI_GRO_CB(skb)->csum = psum; 2417 2418 return __skb_gro_checksum_complete(skb); 2419} 2420 2421static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2422{ 2423 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2424 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2425 NAPI_GRO_CB(skb)->csum_cnt--; 2426 } else { 2427 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2428 * verified a new top level checksum or an encapsulated one 2429 * during GRO. This saves work if we fallback to normal path. 2430 */ 2431 __skb_incr_checksum_unnecessary(skb); 2432 } 2433} 2434 2435#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2436 compute_pseudo) \ 2437({ \ 2438 __sum16 __ret = 0; \ 2439 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2440 __ret = __skb_gro_checksum_validate_complete(skb, \ 2441 compute_pseudo(skb, proto)); \ 2442 if (__ret) \ 2443 __skb_mark_checksum_bad(skb); \ 2444 else \ 2445 skb_gro_incr_csum_unnecessary(skb); \ 2446 __ret; \ 2447}) 2448 2449#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2450 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2451 2452#define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2453 compute_pseudo) \ 2454 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2455 2456#define skb_gro_checksum_simple_validate(skb) \ 2457 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2458 2459static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2460{ 2461 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2462 !NAPI_GRO_CB(skb)->csum_valid); 2463} 2464 2465static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2466 __sum16 check, __wsum pseudo) 2467{ 2468 NAPI_GRO_CB(skb)->csum = ~pseudo; 2469 NAPI_GRO_CB(skb)->csum_valid = 1; 2470} 2471 2472#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \ 2473do { \ 2474 if (__skb_gro_checksum_convert_check(skb)) \ 2475 __skb_gro_checksum_convert(skb, check, \ 2476 compute_pseudo(skb, proto)); \ 2477} while (0) 2478 2479struct gro_remcsum { 2480 int offset; 2481 __wsum delta; 2482}; 2483 2484static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2485{ 2486 grc->offset = 0; 2487 grc->delta = 0; 2488} 2489 2490static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2491 unsigned int off, size_t hdrlen, 2492 int start, int offset, 2493 struct gro_remcsum *grc, 2494 bool nopartial) 2495{ 2496 __wsum delta; 2497 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2498 2499 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2500 2501 if (!nopartial) { 2502 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2503 return ptr; 2504 } 2505 2506 ptr = skb_gro_header_fast(skb, off); 2507 if (skb_gro_header_hard(skb, off + plen)) { 2508 ptr = skb_gro_header_slow(skb, off + plen, off); 2509 if (!ptr) 2510 return NULL; 2511 } 2512 2513 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2514 start, offset); 2515 2516 /* Adjust skb->csum since we changed the packet */ 2517 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2518 2519 grc->offset = off + hdrlen + offset; 2520 grc->delta = delta; 2521 2522 return ptr; 2523} 2524 2525static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2526 struct gro_remcsum *grc) 2527{ 2528 void *ptr; 2529 size_t plen = grc->offset + sizeof(u16); 2530 2531 if (!grc->delta) 2532 return; 2533 2534 ptr = skb_gro_header_fast(skb, grc->offset); 2535 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 2536 ptr = skb_gro_header_slow(skb, plen, grc->offset); 2537 if (!ptr) 2538 return; 2539 } 2540 2541 remcsum_unadjust((__sum16 *)ptr, grc->delta); 2542} 2543 2544struct skb_csum_offl_spec { 2545 __u16 ipv4_okay:1, 2546 ipv6_okay:1, 2547 encap_okay:1, 2548 ip_options_okay:1, 2549 ext_hdrs_okay:1, 2550 tcp_okay:1, 2551 udp_okay:1, 2552 sctp_okay:1, 2553 vlan_okay:1, 2554 no_encapped_ipv6:1, 2555 no_not_encapped:1; 2556}; 2557 2558bool __skb_csum_offload_chk(struct sk_buff *skb, 2559 const struct skb_csum_offl_spec *spec, 2560 bool *csum_encapped, 2561 bool csum_help); 2562 2563static inline bool skb_csum_offload_chk(struct sk_buff *skb, 2564 const struct skb_csum_offl_spec *spec, 2565 bool *csum_encapped, 2566 bool csum_help) 2567{ 2568 if (skb->ip_summed != CHECKSUM_PARTIAL) 2569 return false; 2570 2571 return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help); 2572} 2573 2574static inline bool skb_csum_offload_chk_help(struct sk_buff *skb, 2575 const struct skb_csum_offl_spec *spec) 2576{ 2577 bool csum_encapped; 2578 2579 return skb_csum_offload_chk(skb, spec, &csum_encapped, true); 2580} 2581 2582static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb) 2583{ 2584 static const struct skb_csum_offl_spec csum_offl_spec = { 2585 .ipv4_okay = 1, 2586 .ip_options_okay = 1, 2587 .ipv6_okay = 1, 2588 .vlan_okay = 1, 2589 .tcp_okay = 1, 2590 .udp_okay = 1, 2591 }; 2592 2593 return skb_csum_offload_chk_help(skb, &csum_offl_spec); 2594} 2595 2596static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb) 2597{ 2598 static const struct skb_csum_offl_spec csum_offl_spec = { 2599 .ipv4_okay = 1, 2600 .ip_options_okay = 1, 2601 .tcp_okay = 1, 2602 .udp_okay = 1, 2603 .vlan_okay = 1, 2604 }; 2605 2606 return skb_csum_offload_chk_help(skb, &csum_offl_spec); 2607} 2608 2609static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2610 unsigned short type, 2611 const void *daddr, const void *saddr, 2612 unsigned int len) 2613{ 2614 if (!dev->header_ops || !dev->header_ops->create) 2615 return 0; 2616 2617 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2618} 2619 2620static inline int dev_parse_header(const struct sk_buff *skb, 2621 unsigned char *haddr) 2622{ 2623 const struct net_device *dev = skb->dev; 2624 2625 if (!dev->header_ops || !dev->header_ops->parse) 2626 return 0; 2627 return dev->header_ops->parse(skb, haddr); 2628} 2629 2630typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len); 2631int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2632static inline int unregister_gifconf(unsigned int family) 2633{ 2634 return register_gifconf(family, NULL); 2635} 2636 2637#ifdef CONFIG_NET_FLOW_LIMIT 2638#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2639struct sd_flow_limit { 2640 u64 count; 2641 unsigned int num_buckets; 2642 unsigned int history_head; 2643 u16 history[FLOW_LIMIT_HISTORY]; 2644 u8 buckets[]; 2645}; 2646 2647extern int netdev_flow_limit_table_len; 2648#endif /* CONFIG_NET_FLOW_LIMIT */ 2649 2650/* 2651 * Incoming packets are placed on per-cpu queues 2652 */ 2653struct softnet_data { 2654 struct list_head poll_list; 2655 struct sk_buff_head process_queue; 2656 2657 /* stats */ 2658 unsigned int processed; 2659 unsigned int time_squeeze; 2660 unsigned int cpu_collision; 2661 unsigned int received_rps; 2662#ifdef CONFIG_RPS 2663 struct softnet_data *rps_ipi_list; 2664#endif 2665#ifdef CONFIG_NET_FLOW_LIMIT 2666 struct sd_flow_limit __rcu *flow_limit; 2667#endif 2668 struct Qdisc *output_queue; 2669 struct Qdisc **output_queue_tailp; 2670 struct sk_buff *completion_queue; 2671 2672#ifdef CONFIG_RPS 2673 /* Elements below can be accessed between CPUs for RPS */ 2674 struct call_single_data csd ____cacheline_aligned_in_smp; 2675 struct softnet_data *rps_ipi_next; 2676 unsigned int cpu; 2677 unsigned int input_queue_head; 2678 unsigned int input_queue_tail; 2679#endif 2680 unsigned int dropped; 2681 struct sk_buff_head input_pkt_queue; 2682 struct napi_struct backlog; 2683 2684}; 2685 2686static inline void input_queue_head_incr(struct softnet_data *sd) 2687{ 2688#ifdef CONFIG_RPS 2689 sd->input_queue_head++; 2690#endif 2691} 2692 2693static inline void input_queue_tail_incr_save(struct softnet_data *sd, 2694 unsigned int *qtail) 2695{ 2696#ifdef CONFIG_RPS 2697 *qtail = ++sd->input_queue_tail; 2698#endif 2699} 2700 2701DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 2702 2703void __netif_schedule(struct Qdisc *q); 2704void netif_schedule_queue(struct netdev_queue *txq); 2705 2706static inline void netif_tx_schedule_all(struct net_device *dev) 2707{ 2708 unsigned int i; 2709 2710 for (i = 0; i < dev->num_tx_queues; i++) 2711 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 2712} 2713 2714static inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 2715{ 2716 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2717} 2718 2719/** 2720 * netif_start_queue - allow transmit 2721 * @dev: network device 2722 * 2723 * Allow upper layers to call the device hard_start_xmit routine. 2724 */ 2725static inline void netif_start_queue(struct net_device *dev) 2726{ 2727 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 2728} 2729 2730static inline void netif_tx_start_all_queues(struct net_device *dev) 2731{ 2732 unsigned int i; 2733 2734 for (i = 0; i < dev->num_tx_queues; i++) { 2735 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2736 netif_tx_start_queue(txq); 2737 } 2738} 2739 2740void netif_tx_wake_queue(struct netdev_queue *dev_queue); 2741 2742/** 2743 * netif_wake_queue - restart transmit 2744 * @dev: network device 2745 * 2746 * Allow upper layers to call the device hard_start_xmit routine. 2747 * Used for flow control when transmit resources are available. 2748 */ 2749static inline void netif_wake_queue(struct net_device *dev) 2750{ 2751 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 2752} 2753 2754static inline void netif_tx_wake_all_queues(struct net_device *dev) 2755{ 2756 unsigned int i; 2757 2758 for (i = 0; i < dev->num_tx_queues; i++) { 2759 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2760 netif_tx_wake_queue(txq); 2761 } 2762} 2763 2764static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 2765{ 2766 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2767} 2768 2769/** 2770 * netif_stop_queue - stop transmitted packets 2771 * @dev: network device 2772 * 2773 * Stop upper layers calling the device hard_start_xmit routine. 2774 * Used for flow control when transmit resources are unavailable. 2775 */ 2776static inline void netif_stop_queue(struct net_device *dev) 2777{ 2778 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 2779} 2780 2781void netif_tx_stop_all_queues(struct net_device *dev); 2782 2783static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 2784{ 2785 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2786} 2787 2788/** 2789 * netif_queue_stopped - test if transmit queue is flowblocked 2790 * @dev: network device 2791 * 2792 * Test if transmit queue on device is currently unable to send. 2793 */ 2794static inline bool netif_queue_stopped(const struct net_device *dev) 2795{ 2796 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 2797} 2798 2799static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 2800{ 2801 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 2802} 2803 2804static inline bool 2805netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 2806{ 2807 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 2808} 2809 2810static inline bool 2811netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 2812{ 2813 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 2814} 2815 2816/** 2817 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 2818 * @dev_queue: pointer to transmit queue 2819 * 2820 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 2821 * to give appropriate hint to the cpu. 2822 */ 2823static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 2824{ 2825#ifdef CONFIG_BQL 2826 prefetchw(&dev_queue->dql.num_queued); 2827#endif 2828} 2829 2830/** 2831 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 2832 * @dev_queue: pointer to transmit queue 2833 * 2834 * BQL enabled drivers might use this helper in their TX completion path, 2835 * to give appropriate hint to the cpu. 2836 */ 2837static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 2838{ 2839#ifdef CONFIG_BQL 2840 prefetchw(&dev_queue->dql.limit); 2841#endif 2842} 2843 2844static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 2845 unsigned int bytes) 2846{ 2847#ifdef CONFIG_BQL 2848 dql_queued(&dev_queue->dql, bytes); 2849 2850 if (likely(dql_avail(&dev_queue->dql) >= 0)) 2851 return; 2852 2853 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2854 2855 /* 2856 * The XOFF flag must be set before checking the dql_avail below, 2857 * because in netdev_tx_completed_queue we update the dql_completed 2858 * before checking the XOFF flag. 2859 */ 2860 smp_mb(); 2861 2862 /* check again in case another CPU has just made room avail */ 2863 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 2864 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2865#endif 2866} 2867 2868/** 2869 * netdev_sent_queue - report the number of bytes queued to hardware 2870 * @dev: network device 2871 * @bytes: number of bytes queued to the hardware device queue 2872 * 2873 * Report the number of bytes queued for sending/completion to the network 2874 * device hardware queue. @bytes should be a good approximation and should 2875 * exactly match netdev_completed_queue() @bytes 2876 */ 2877static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 2878{ 2879 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 2880} 2881 2882static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 2883 unsigned int pkts, unsigned int bytes) 2884{ 2885#ifdef CONFIG_BQL 2886 if (unlikely(!bytes)) 2887 return; 2888 2889 dql_completed(&dev_queue->dql, bytes); 2890 2891 /* 2892 * Without the memory barrier there is a small possiblity that 2893 * netdev_tx_sent_queue will miss the update and cause the queue to 2894 * be stopped forever 2895 */ 2896 smp_mb(); 2897 2898 if (dql_avail(&dev_queue->dql) < 0) 2899 return; 2900 2901 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 2902 netif_schedule_queue(dev_queue); 2903#endif 2904} 2905 2906/** 2907 * netdev_completed_queue - report bytes and packets completed by device 2908 * @dev: network device 2909 * @pkts: actual number of packets sent over the medium 2910 * @bytes: actual number of bytes sent over the medium 2911 * 2912 * Report the number of bytes and packets transmitted by the network device 2913 * hardware queue over the physical medium, @bytes must exactly match the 2914 * @bytes amount passed to netdev_sent_queue() 2915 */ 2916static inline void netdev_completed_queue(struct net_device *dev, 2917 unsigned int pkts, unsigned int bytes) 2918{ 2919 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 2920} 2921 2922static inline void netdev_tx_reset_queue(struct netdev_queue *q) 2923{ 2924#ifdef CONFIG_BQL 2925 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 2926 dql_reset(&q->dql); 2927#endif 2928} 2929 2930/** 2931 * netdev_reset_queue - reset the packets and bytes count of a network device 2932 * @dev_queue: network device 2933 * 2934 * Reset the bytes and packet count of a network device and clear the 2935 * software flow control OFF bit for this network device 2936 */ 2937static inline void netdev_reset_queue(struct net_device *dev_queue) 2938{ 2939 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 2940} 2941 2942/** 2943 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 2944 * @dev: network device 2945 * @queue_index: given tx queue index 2946 * 2947 * Returns 0 if given tx queue index >= number of device tx queues, 2948 * otherwise returns the originally passed tx queue index. 2949 */ 2950static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 2951{ 2952 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 2953 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 2954 dev->name, queue_index, 2955 dev->real_num_tx_queues); 2956 return 0; 2957 } 2958 2959 return queue_index; 2960} 2961 2962/** 2963 * netif_running - test if up 2964 * @dev: network device 2965 * 2966 * Test if the device has been brought up. 2967 */ 2968static inline bool netif_running(const struct net_device *dev) 2969{ 2970 return test_bit(__LINK_STATE_START, &dev->state); 2971} 2972 2973/* 2974 * Routines to manage the subqueues on a device. We only need start 2975 * stop, and a check if it's stopped. All other device management is 2976 * done at the overall netdevice level. 2977 * Also test the device if we're multiqueue. 2978 */ 2979 2980/** 2981 * netif_start_subqueue - allow sending packets on subqueue 2982 * @dev: network device 2983 * @queue_index: sub queue index 2984 * 2985 * Start individual transmit queue of a device with multiple transmit queues. 2986 */ 2987static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 2988{ 2989 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2990 2991 netif_tx_start_queue(txq); 2992} 2993 2994/** 2995 * netif_stop_subqueue - stop sending packets on subqueue 2996 * @dev: network device 2997 * @queue_index: sub queue index 2998 * 2999 * Stop individual transmit queue of a device with multiple transmit queues. 3000 */ 3001static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3002{ 3003 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3004 netif_tx_stop_queue(txq); 3005} 3006 3007/** 3008 * netif_subqueue_stopped - test status of subqueue 3009 * @dev: network device 3010 * @queue_index: sub queue index 3011 * 3012 * Check individual transmit queue of a device with multiple transmit queues. 3013 */ 3014static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3015 u16 queue_index) 3016{ 3017 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3018 3019 return netif_tx_queue_stopped(txq); 3020} 3021 3022static inline bool netif_subqueue_stopped(const struct net_device *dev, 3023 struct sk_buff *skb) 3024{ 3025 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3026} 3027 3028void netif_wake_subqueue(struct net_device *dev, u16 queue_index); 3029 3030#ifdef CONFIG_XPS 3031int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3032 u16 index); 3033#else 3034static inline int netif_set_xps_queue(struct net_device *dev, 3035 const struct cpumask *mask, 3036 u16 index) 3037{ 3038 return 0; 3039} 3040#endif 3041 3042u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb, 3043 unsigned int num_tx_queues); 3044 3045/* 3046 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used 3047 * as a distribution range limit for the returned value. 3048 */ 3049static inline u16 skb_tx_hash(const struct net_device *dev, 3050 struct sk_buff *skb) 3051{ 3052 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues); 3053} 3054 3055/** 3056 * netif_is_multiqueue - test if device has multiple transmit queues 3057 * @dev: network device 3058 * 3059 * Check if device has multiple transmit queues 3060 */ 3061static inline bool netif_is_multiqueue(const struct net_device *dev) 3062{ 3063 return dev->num_tx_queues > 1; 3064} 3065 3066int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3067 3068#ifdef CONFIG_SYSFS 3069int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3070#else 3071static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3072 unsigned int rxq) 3073{ 3074 return 0; 3075} 3076#endif 3077 3078#ifdef CONFIG_SYSFS 3079static inline unsigned int get_netdev_rx_queue_index( 3080 struct netdev_rx_queue *queue) 3081{ 3082 struct net_device *dev = queue->dev; 3083 int index = queue - dev->_rx; 3084 3085 BUG_ON(index >= dev->num_rx_queues); 3086 return index; 3087} 3088#endif 3089 3090#define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3091int netif_get_num_default_rss_queues(void); 3092 3093enum skb_free_reason { 3094 SKB_REASON_CONSUMED, 3095 SKB_REASON_DROPPED, 3096}; 3097 3098void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3099void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3100 3101/* 3102 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3103 * interrupt context or with hardware interrupts being disabled. 3104 * (in_irq() || irqs_disabled()) 3105 * 3106 * We provide four helpers that can be used in following contexts : 3107 * 3108 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3109 * replacing kfree_skb(skb) 3110 * 3111 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3112 * Typically used in place of consume_skb(skb) in TX completion path 3113 * 3114 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3115 * replacing kfree_skb(skb) 3116 * 3117 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3118 * and consumed a packet. Used in place of consume_skb(skb) 3119 */ 3120static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3121{ 3122 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3123} 3124 3125static inline void dev_consume_skb_irq(struct sk_buff *skb) 3126{ 3127 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3128} 3129 3130static inline void dev_kfree_skb_any(struct sk_buff *skb) 3131{ 3132 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3133} 3134 3135static inline void dev_consume_skb_any(struct sk_buff *skb) 3136{ 3137 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3138} 3139 3140int netif_rx(struct sk_buff *skb); 3141int netif_rx_ni(struct sk_buff *skb); 3142int netif_receive_skb(struct sk_buff *skb); 3143gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3144void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3145struct sk_buff *napi_get_frags(struct napi_struct *napi); 3146gro_result_t napi_gro_frags(struct napi_struct *napi); 3147struct packet_offload *gro_find_receive_by_type(__be16 type); 3148struct packet_offload *gro_find_complete_by_type(__be16 type); 3149 3150static inline void napi_free_frags(struct napi_struct *napi) 3151{ 3152 kfree_skb(napi->skb); 3153 napi->skb = NULL; 3154} 3155 3156int netdev_rx_handler_register(struct net_device *dev, 3157 rx_handler_func_t *rx_handler, 3158 void *rx_handler_data); 3159void netdev_rx_handler_unregister(struct net_device *dev); 3160 3161bool dev_valid_name(const char *name); 3162int dev_ioctl(struct net *net, unsigned int cmd, void __user *); 3163int dev_ethtool(struct net *net, struct ifreq *); 3164unsigned int dev_get_flags(const struct net_device *); 3165int __dev_change_flags(struct net_device *, unsigned int flags); 3166int dev_change_flags(struct net_device *, unsigned int); 3167void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3168 unsigned int gchanges); 3169int dev_change_name(struct net_device *, const char *); 3170int dev_set_alias(struct net_device *, const char *, size_t); 3171int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3172int dev_set_mtu(struct net_device *, int); 3173void dev_set_group(struct net_device *, int); 3174int dev_set_mac_address(struct net_device *, struct sockaddr *); 3175int dev_change_carrier(struct net_device *, bool new_carrier); 3176int dev_get_phys_port_id(struct net_device *dev, 3177 struct netdev_phys_item_id *ppid); 3178int dev_get_phys_port_name(struct net_device *dev, 3179 char *name, size_t len); 3180int dev_change_proto_down(struct net_device *dev, bool proto_down); 3181struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev); 3182struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3183 struct netdev_queue *txq, int *ret); 3184int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3185int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3186bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb); 3187 3188extern int netdev_budget; 3189 3190/* Called by rtnetlink.c:rtnl_unlock() */ 3191void netdev_run_todo(void); 3192 3193/** 3194 * dev_put - release reference to device 3195 * @dev: network device 3196 * 3197 * Release reference to device to allow it to be freed. 3198 */ 3199static inline void dev_put(struct net_device *dev) 3200{ 3201 this_cpu_dec(*dev->pcpu_refcnt); 3202} 3203 3204/** 3205 * dev_hold - get reference to device 3206 * @dev: network device 3207 * 3208 * Hold reference to device to keep it from being freed. 3209 */ 3210static inline void dev_hold(struct net_device *dev) 3211{ 3212 this_cpu_inc(*dev->pcpu_refcnt); 3213} 3214 3215/* Carrier loss detection, dial on demand. The functions netif_carrier_on 3216 * and _off may be called from IRQ context, but it is caller 3217 * who is responsible for serialization of these calls. 3218 * 3219 * The name carrier is inappropriate, these functions should really be 3220 * called netif_lowerlayer_*() because they represent the state of any 3221 * kind of lower layer not just hardware media. 3222 */ 3223 3224void linkwatch_init_dev(struct net_device *dev); 3225void linkwatch_fire_event(struct net_device *dev); 3226void linkwatch_forget_dev(struct net_device *dev); 3227 3228/** 3229 * netif_carrier_ok - test if carrier present 3230 * @dev: network device 3231 * 3232 * Check if carrier is present on device 3233 */ 3234static inline bool netif_carrier_ok(const struct net_device *dev) 3235{ 3236 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3237} 3238 3239unsigned long dev_trans_start(struct net_device *dev); 3240 3241void __netdev_watchdog_up(struct net_device *dev); 3242 3243void netif_carrier_on(struct net_device *dev); 3244 3245void netif_carrier_off(struct net_device *dev); 3246 3247/** 3248 * netif_dormant_on - mark device as dormant. 3249 * @dev: network device 3250 * 3251 * Mark device as dormant (as per RFC2863). 3252 * 3253 * The dormant state indicates that the relevant interface is not 3254 * actually in a condition to pass packets (i.e., it is not 'up') but is 3255 * in a "pending" state, waiting for some external event. For "on- 3256 * demand" interfaces, this new state identifies the situation where the 3257 * interface is waiting for events to place it in the up state. 3258 * 3259 */ 3260static inline void netif_dormant_on(struct net_device *dev) 3261{ 3262 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3263 linkwatch_fire_event(dev); 3264} 3265 3266/** 3267 * netif_dormant_off - set device as not dormant. 3268 * @dev: network device 3269 * 3270 * Device is not in dormant state. 3271 */ 3272static inline void netif_dormant_off(struct net_device *dev) 3273{ 3274 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3275 linkwatch_fire_event(dev); 3276} 3277 3278/** 3279 * netif_dormant - test if carrier present 3280 * @dev: network device 3281 * 3282 * Check if carrier is present on device 3283 */ 3284static inline bool netif_dormant(const struct net_device *dev) 3285{ 3286 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3287} 3288 3289 3290/** 3291 * netif_oper_up - test if device is operational 3292 * @dev: network device 3293 * 3294 * Check if carrier is operational 3295 */ 3296static inline bool netif_oper_up(const struct net_device *dev) 3297{ 3298 return (dev->operstate == IF_OPER_UP || 3299 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3300} 3301 3302/** 3303 * netif_device_present - is device available or removed 3304 * @dev: network device 3305 * 3306 * Check if device has not been removed from system. 3307 */ 3308static inline bool netif_device_present(struct net_device *dev) 3309{ 3310 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3311} 3312 3313void netif_device_detach(struct net_device *dev); 3314 3315void netif_device_attach(struct net_device *dev); 3316 3317/* 3318 * Network interface message level settings 3319 */ 3320 3321enum { 3322 NETIF_MSG_DRV = 0x0001, 3323 NETIF_MSG_PROBE = 0x0002, 3324 NETIF_MSG_LINK = 0x0004, 3325 NETIF_MSG_TIMER = 0x0008, 3326 NETIF_MSG_IFDOWN = 0x0010, 3327 NETIF_MSG_IFUP = 0x0020, 3328 NETIF_MSG_RX_ERR = 0x0040, 3329 NETIF_MSG_TX_ERR = 0x0080, 3330 NETIF_MSG_TX_QUEUED = 0x0100, 3331 NETIF_MSG_INTR = 0x0200, 3332 NETIF_MSG_TX_DONE = 0x0400, 3333 NETIF_MSG_RX_STATUS = 0x0800, 3334 NETIF_MSG_PKTDATA = 0x1000, 3335 NETIF_MSG_HW = 0x2000, 3336 NETIF_MSG_WOL = 0x4000, 3337}; 3338 3339#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3340#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3341#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3342#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3343#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3344#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3345#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3346#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3347#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3348#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3349#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3350#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3351#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3352#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3353#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3354 3355static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3356{ 3357 /* use default */ 3358 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3359 return default_msg_enable_bits; 3360 if (debug_value == 0) /* no output */ 3361 return 0; 3362 /* set low N bits */ 3363 return (1 << debug_value) - 1; 3364} 3365 3366static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 3367{ 3368 spin_lock(&txq->_xmit_lock); 3369 txq->xmit_lock_owner = cpu; 3370} 3371 3372static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 3373{ 3374 spin_lock_bh(&txq->_xmit_lock); 3375 txq->xmit_lock_owner = smp_processor_id(); 3376} 3377 3378static inline bool __netif_tx_trylock(struct netdev_queue *txq) 3379{ 3380 bool ok = spin_trylock(&txq->_xmit_lock); 3381 if (likely(ok)) 3382 txq->xmit_lock_owner = smp_processor_id(); 3383 return ok; 3384} 3385 3386static inline void __netif_tx_unlock(struct netdev_queue *txq) 3387{ 3388 txq->xmit_lock_owner = -1; 3389 spin_unlock(&txq->_xmit_lock); 3390} 3391 3392static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 3393{ 3394 txq->xmit_lock_owner = -1; 3395 spin_unlock_bh(&txq->_xmit_lock); 3396} 3397 3398static inline void txq_trans_update(struct netdev_queue *txq) 3399{ 3400 if (txq->xmit_lock_owner != -1) 3401 txq->trans_start = jiffies; 3402} 3403 3404/** 3405 * netif_tx_lock - grab network device transmit lock 3406 * @dev: network device 3407 * 3408 * Get network device transmit lock 3409 */ 3410static inline void netif_tx_lock(struct net_device *dev) 3411{ 3412 unsigned int i; 3413 int cpu; 3414 3415 spin_lock(&dev->tx_global_lock); 3416 cpu = smp_processor_id(); 3417 for (i = 0; i < dev->num_tx_queues; i++) { 3418 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3419 3420 /* We are the only thread of execution doing a 3421 * freeze, but we have to grab the _xmit_lock in 3422 * order to synchronize with threads which are in 3423 * the ->hard_start_xmit() handler and already 3424 * checked the frozen bit. 3425 */ 3426 __netif_tx_lock(txq, cpu); 3427 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 3428 __netif_tx_unlock(txq); 3429 } 3430} 3431 3432static inline void netif_tx_lock_bh(struct net_device *dev) 3433{ 3434 local_bh_disable(); 3435 netif_tx_lock(dev); 3436} 3437 3438static inline void netif_tx_unlock(struct net_device *dev) 3439{ 3440 unsigned int i; 3441 3442 for (i = 0; i < dev->num_tx_queues; i++) { 3443 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3444 3445 /* No need to grab the _xmit_lock here. If the 3446 * queue is not stopped for another reason, we 3447 * force a schedule. 3448 */ 3449 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 3450 netif_schedule_queue(txq); 3451 } 3452 spin_unlock(&dev->tx_global_lock); 3453} 3454 3455static inline void netif_tx_unlock_bh(struct net_device *dev) 3456{ 3457 netif_tx_unlock(dev); 3458 local_bh_enable(); 3459} 3460 3461#define HARD_TX_LOCK(dev, txq, cpu) { \ 3462 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3463 __netif_tx_lock(txq, cpu); \ 3464 } \ 3465} 3466 3467#define HARD_TX_TRYLOCK(dev, txq) \ 3468 (((dev->features & NETIF_F_LLTX) == 0) ? \ 3469 __netif_tx_trylock(txq) : \ 3470 true ) 3471 3472#define HARD_TX_UNLOCK(dev, txq) { \ 3473 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3474 __netif_tx_unlock(txq); \ 3475 } \ 3476} 3477 3478static inline void netif_tx_disable(struct net_device *dev) 3479{ 3480 unsigned int i; 3481 int cpu; 3482 3483 local_bh_disable(); 3484 cpu = smp_processor_id(); 3485 for (i = 0; i < dev->num_tx_queues; i++) { 3486 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3487 3488 __netif_tx_lock(txq, cpu); 3489 netif_tx_stop_queue(txq); 3490 __netif_tx_unlock(txq); 3491 } 3492 local_bh_enable(); 3493} 3494 3495static inline void netif_addr_lock(struct net_device *dev) 3496{ 3497 spin_lock(&dev->addr_list_lock); 3498} 3499 3500static inline void netif_addr_lock_nested(struct net_device *dev) 3501{ 3502 int subclass = SINGLE_DEPTH_NESTING; 3503 3504 if (dev->netdev_ops->ndo_get_lock_subclass) 3505 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev); 3506 3507 spin_lock_nested(&dev->addr_list_lock, subclass); 3508} 3509 3510static inline void netif_addr_lock_bh(struct net_device *dev) 3511{ 3512 spin_lock_bh(&dev->addr_list_lock); 3513} 3514 3515static inline void netif_addr_unlock(struct net_device *dev) 3516{ 3517 spin_unlock(&dev->addr_list_lock); 3518} 3519 3520static inline void netif_addr_unlock_bh(struct net_device *dev) 3521{ 3522 spin_unlock_bh(&dev->addr_list_lock); 3523} 3524 3525/* 3526 * dev_addrs walker. Should be used only for read access. Call with 3527 * rcu_read_lock held. 3528 */ 3529#define for_each_dev_addr(dev, ha) \ 3530 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 3531 3532/* These functions live elsewhere (drivers/net/net_init.c, but related) */ 3533 3534void ether_setup(struct net_device *dev); 3535 3536/* Support for loadable net-drivers */ 3537struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 3538 unsigned char name_assign_type, 3539 void (*setup)(struct net_device *), 3540 unsigned int txqs, unsigned int rxqs); 3541#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 3542 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 3543 3544#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 3545 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 3546 count) 3547 3548int register_netdev(struct net_device *dev); 3549void unregister_netdev(struct net_device *dev); 3550 3551/* General hardware address lists handling functions */ 3552int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 3553 struct netdev_hw_addr_list *from_list, int addr_len); 3554void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 3555 struct netdev_hw_addr_list *from_list, int addr_len); 3556int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 3557 struct net_device *dev, 3558 int (*sync)(struct net_device *, const unsigned char *), 3559 int (*unsync)(struct net_device *, 3560 const unsigned char *)); 3561void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 3562 struct net_device *dev, 3563 int (*unsync)(struct net_device *, 3564 const unsigned char *)); 3565void __hw_addr_init(struct netdev_hw_addr_list *list); 3566 3567/* Functions used for device addresses handling */ 3568int dev_addr_add(struct net_device *dev, const unsigned char *addr, 3569 unsigned char addr_type); 3570int dev_addr_del(struct net_device *dev, const unsigned char *addr, 3571 unsigned char addr_type); 3572void dev_addr_flush(struct net_device *dev); 3573int dev_addr_init(struct net_device *dev); 3574 3575/* Functions used for unicast addresses handling */ 3576int dev_uc_add(struct net_device *dev, const unsigned char *addr); 3577int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 3578int dev_uc_del(struct net_device *dev, const unsigned char *addr); 3579int dev_uc_sync(struct net_device *to, struct net_device *from); 3580int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 3581void dev_uc_unsync(struct net_device *to, struct net_device *from); 3582void dev_uc_flush(struct net_device *dev); 3583void dev_uc_init(struct net_device *dev); 3584 3585/** 3586 * __dev_uc_sync - Synchonize device's unicast list 3587 * @dev: device to sync 3588 * @sync: function to call if address should be added 3589 * @unsync: function to call if address should be removed 3590 * 3591 * Add newly added addresses to the interface, and release 3592 * addresses that have been deleted. 3593 **/ 3594static inline int __dev_uc_sync(struct net_device *dev, 3595 int (*sync)(struct net_device *, 3596 const unsigned char *), 3597 int (*unsync)(struct net_device *, 3598 const unsigned char *)) 3599{ 3600 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 3601} 3602 3603/** 3604 * __dev_uc_unsync - Remove synchronized addresses from device 3605 * @dev: device to sync 3606 * @unsync: function to call if address should be removed 3607 * 3608 * Remove all addresses that were added to the device by dev_uc_sync(). 3609 **/ 3610static inline void __dev_uc_unsync(struct net_device *dev, 3611 int (*unsync)(struct net_device *, 3612 const unsigned char *)) 3613{ 3614 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 3615} 3616 3617/* Functions used for multicast addresses handling */ 3618int dev_mc_add(struct net_device *dev, const unsigned char *addr); 3619int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 3620int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 3621int dev_mc_del(struct net_device *dev, const unsigned char *addr); 3622int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 3623int dev_mc_sync(struct net_device *to, struct net_device *from); 3624int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 3625void dev_mc_unsync(struct net_device *to, struct net_device *from); 3626void dev_mc_flush(struct net_device *dev); 3627void dev_mc_init(struct net_device *dev); 3628 3629/** 3630 * __dev_mc_sync - Synchonize device's multicast list 3631 * @dev: device to sync 3632 * @sync: function to call if address should be added 3633 * @unsync: function to call if address should be removed 3634 * 3635 * Add newly added addresses to the interface, and release 3636 * addresses that have been deleted. 3637 **/ 3638static inline int __dev_mc_sync(struct net_device *dev, 3639 int (*sync)(struct net_device *, 3640 const unsigned char *), 3641 int (*unsync)(struct net_device *, 3642 const unsigned char *)) 3643{ 3644 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 3645} 3646 3647/** 3648 * __dev_mc_unsync - Remove synchronized addresses from device 3649 * @dev: device to sync 3650 * @unsync: function to call if address should be removed 3651 * 3652 * Remove all addresses that were added to the device by dev_mc_sync(). 3653 **/ 3654static inline void __dev_mc_unsync(struct net_device *dev, 3655 int (*unsync)(struct net_device *, 3656 const unsigned char *)) 3657{ 3658 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 3659} 3660 3661/* Functions used for secondary unicast and multicast support */ 3662void dev_set_rx_mode(struct net_device *dev); 3663void __dev_set_rx_mode(struct net_device *dev); 3664int dev_set_promiscuity(struct net_device *dev, int inc); 3665int dev_set_allmulti(struct net_device *dev, int inc); 3666void netdev_state_change(struct net_device *dev); 3667void netdev_notify_peers(struct net_device *dev); 3668void netdev_features_change(struct net_device *dev); 3669/* Load a device via the kmod */ 3670void dev_load(struct net *net, const char *name); 3671struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 3672 struct rtnl_link_stats64 *storage); 3673void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 3674 const struct net_device_stats *netdev_stats); 3675 3676extern int netdev_max_backlog; 3677extern int netdev_tstamp_prequeue; 3678extern int weight_p; 3679extern int bpf_jit_enable; 3680 3681bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 3682struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 3683 struct list_head **iter); 3684struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 3685 struct list_head **iter); 3686 3687/* iterate through upper list, must be called under RCU read lock */ 3688#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 3689 for (iter = &(dev)->adj_list.upper, \ 3690 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 3691 updev; \ 3692 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 3693 3694/* iterate through upper list, must be called under RCU read lock */ 3695#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \ 3696 for (iter = &(dev)->all_adj_list.upper, \ 3697 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \ 3698 updev; \ 3699 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter))) 3700 3701void *netdev_lower_get_next_private(struct net_device *dev, 3702 struct list_head **iter); 3703void *netdev_lower_get_next_private_rcu(struct net_device *dev, 3704 struct list_head **iter); 3705 3706#define netdev_for_each_lower_private(dev, priv, iter) \ 3707 for (iter = (dev)->adj_list.lower.next, \ 3708 priv = netdev_lower_get_next_private(dev, &(iter)); \ 3709 priv; \ 3710 priv = netdev_lower_get_next_private(dev, &(iter))) 3711 3712#define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 3713 for (iter = &(dev)->adj_list.lower, \ 3714 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 3715 priv; \ 3716 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 3717 3718void *netdev_lower_get_next(struct net_device *dev, 3719 struct list_head **iter); 3720#define netdev_for_each_lower_dev(dev, ldev, iter) \ 3721 for (iter = (dev)->adj_list.lower.next, \ 3722 ldev = netdev_lower_get_next(dev, &(iter)); \ 3723 ldev; \ 3724 ldev = netdev_lower_get_next(dev, &(iter))) 3725 3726void *netdev_adjacent_get_private(struct list_head *adj_list); 3727void *netdev_lower_get_first_private_rcu(struct net_device *dev); 3728struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 3729struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 3730int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev); 3731int netdev_master_upper_dev_link(struct net_device *dev, 3732 struct net_device *upper_dev, 3733 void *upper_priv, void *upper_info); 3734void netdev_upper_dev_unlink(struct net_device *dev, 3735 struct net_device *upper_dev); 3736void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 3737void *netdev_lower_dev_get_private(struct net_device *dev, 3738 struct net_device *lower_dev); 3739void netdev_lower_state_changed(struct net_device *lower_dev, 3740 void *lower_state_info); 3741 3742/* RSS keys are 40 or 52 bytes long */ 3743#define NETDEV_RSS_KEY_LEN 52 3744extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN]; 3745void netdev_rss_key_fill(void *buffer, size_t len); 3746 3747int dev_get_nest_level(struct net_device *dev, 3748 bool (*type_check)(const struct net_device *dev)); 3749int skb_checksum_help(struct sk_buff *skb); 3750struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 3751 netdev_features_t features, bool tx_path); 3752struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 3753 netdev_features_t features); 3754 3755struct netdev_bonding_info { 3756 ifslave slave; 3757 ifbond master; 3758}; 3759 3760struct netdev_notifier_bonding_info { 3761 struct netdev_notifier_info info; /* must be first */ 3762 struct netdev_bonding_info bonding_info; 3763}; 3764 3765void netdev_bonding_info_change(struct net_device *dev, 3766 struct netdev_bonding_info *bonding_info); 3767 3768static inline 3769struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 3770{ 3771 return __skb_gso_segment(skb, features, true); 3772} 3773__be16 skb_network_protocol(struct sk_buff *skb, int *depth); 3774 3775static inline bool can_checksum_protocol(netdev_features_t features, 3776 __be16 protocol) 3777{ 3778 if (protocol == htons(ETH_P_FCOE)) 3779 return !!(features & NETIF_F_FCOE_CRC); 3780 3781 /* Assume this is an IP checksum (not SCTP CRC) */ 3782 3783 if (features & NETIF_F_HW_CSUM) { 3784 /* Can checksum everything */ 3785 return true; 3786 } 3787 3788 switch (protocol) { 3789 case htons(ETH_P_IP): 3790 return !!(features & NETIF_F_IP_CSUM); 3791 case htons(ETH_P_IPV6): 3792 return !!(features & NETIF_F_IPV6_CSUM); 3793 default: 3794 return false; 3795 } 3796} 3797 3798/* Map an ethertype into IP protocol if possible */ 3799static inline int eproto_to_ipproto(int eproto) 3800{ 3801 switch (eproto) { 3802 case htons(ETH_P_IP): 3803 return IPPROTO_IP; 3804 case htons(ETH_P_IPV6): 3805 return IPPROTO_IPV6; 3806 default: 3807 return -1; 3808 } 3809} 3810 3811#ifdef CONFIG_BUG 3812void netdev_rx_csum_fault(struct net_device *dev); 3813#else 3814static inline void netdev_rx_csum_fault(struct net_device *dev) 3815{ 3816} 3817#endif 3818/* rx skb timestamps */ 3819void net_enable_timestamp(void); 3820void net_disable_timestamp(void); 3821 3822#ifdef CONFIG_PROC_FS 3823int __init dev_proc_init(void); 3824#else 3825#define dev_proc_init() 0 3826#endif 3827 3828static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 3829 struct sk_buff *skb, struct net_device *dev, 3830 bool more) 3831{ 3832 skb->xmit_more = more ? 1 : 0; 3833 return ops->ndo_start_xmit(skb, dev); 3834} 3835 3836static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 3837 struct netdev_queue *txq, bool more) 3838{ 3839 const struct net_device_ops *ops = dev->netdev_ops; 3840 int rc; 3841 3842 rc = __netdev_start_xmit(ops, skb, dev, more); 3843 if (rc == NETDEV_TX_OK) 3844 txq_trans_update(txq); 3845 3846 return rc; 3847} 3848 3849int netdev_class_create_file_ns(struct class_attribute *class_attr, 3850 const void *ns); 3851void netdev_class_remove_file_ns(struct class_attribute *class_attr, 3852 const void *ns); 3853 3854static inline int netdev_class_create_file(struct class_attribute *class_attr) 3855{ 3856 return netdev_class_create_file_ns(class_attr, NULL); 3857} 3858 3859static inline void netdev_class_remove_file(struct class_attribute *class_attr) 3860{ 3861 netdev_class_remove_file_ns(class_attr, NULL); 3862} 3863 3864extern struct kobj_ns_type_operations net_ns_type_operations; 3865 3866const char *netdev_drivername(const struct net_device *dev); 3867 3868void linkwatch_run_queue(void); 3869 3870static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 3871 netdev_features_t f2) 3872{ 3873 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 3874 if (f1 & NETIF_F_HW_CSUM) 3875 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 3876 else 3877 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 3878 } 3879 3880 return f1 & f2; 3881} 3882 3883static inline netdev_features_t netdev_get_wanted_features( 3884 struct net_device *dev) 3885{ 3886 return (dev->features & ~dev->hw_features) | dev->wanted_features; 3887} 3888netdev_features_t netdev_increment_features(netdev_features_t all, 3889 netdev_features_t one, netdev_features_t mask); 3890 3891/* Allow TSO being used on stacked device : 3892 * Performing the GSO segmentation before last device 3893 * is a performance improvement. 3894 */ 3895static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 3896 netdev_features_t mask) 3897{ 3898 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 3899} 3900 3901int __netdev_update_features(struct net_device *dev); 3902void netdev_update_features(struct net_device *dev); 3903void netdev_change_features(struct net_device *dev); 3904 3905void netif_stacked_transfer_operstate(const struct net_device *rootdev, 3906 struct net_device *dev); 3907 3908netdev_features_t passthru_features_check(struct sk_buff *skb, 3909 struct net_device *dev, 3910 netdev_features_t features); 3911netdev_features_t netif_skb_features(struct sk_buff *skb); 3912 3913static inline bool net_gso_ok(netdev_features_t features, int gso_type) 3914{ 3915 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT; 3916 3917 /* check flags correspondence */ 3918 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 3919 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT)); 3920 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 3921 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 3922 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 3923 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 3924 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 3925 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 3926 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT)); 3927 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT)); 3928 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 3929 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 3930 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 3931 3932 return (features & feature) == feature; 3933} 3934 3935static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 3936{ 3937 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 3938 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 3939} 3940 3941static inline bool netif_needs_gso(struct sk_buff *skb, 3942 netdev_features_t features) 3943{ 3944 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 3945 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 3946 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 3947} 3948 3949static inline void netif_set_gso_max_size(struct net_device *dev, 3950 unsigned int size) 3951{ 3952 dev->gso_max_size = size; 3953} 3954 3955static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 3956 int pulled_hlen, u16 mac_offset, 3957 int mac_len) 3958{ 3959 skb->protocol = protocol; 3960 skb->encapsulation = 1; 3961 skb_push(skb, pulled_hlen); 3962 skb_reset_transport_header(skb); 3963 skb->mac_header = mac_offset; 3964 skb->network_header = skb->mac_header + mac_len; 3965 skb->mac_len = mac_len; 3966} 3967 3968static inline bool netif_is_macvlan(const struct net_device *dev) 3969{ 3970 return dev->priv_flags & IFF_MACVLAN; 3971} 3972 3973static inline bool netif_is_macvlan_port(const struct net_device *dev) 3974{ 3975 return dev->priv_flags & IFF_MACVLAN_PORT; 3976} 3977 3978static inline bool netif_is_ipvlan(const struct net_device *dev) 3979{ 3980 return dev->priv_flags & IFF_IPVLAN_SLAVE; 3981} 3982 3983static inline bool netif_is_ipvlan_port(const struct net_device *dev) 3984{ 3985 return dev->priv_flags & IFF_IPVLAN_MASTER; 3986} 3987 3988static inline bool netif_is_bond_master(const struct net_device *dev) 3989{ 3990 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 3991} 3992 3993static inline bool netif_is_bond_slave(const struct net_device *dev) 3994{ 3995 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 3996} 3997 3998static inline bool netif_supports_nofcs(struct net_device *dev) 3999{ 4000 return dev->priv_flags & IFF_SUPP_NOFCS; 4001} 4002 4003static inline bool netif_is_l3_master(const struct net_device *dev) 4004{ 4005 return dev->priv_flags & IFF_L3MDEV_MASTER; 4006} 4007 4008static inline bool netif_is_l3_slave(const struct net_device *dev) 4009{ 4010 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4011} 4012 4013static inline bool netif_is_bridge_master(const struct net_device *dev) 4014{ 4015 return dev->priv_flags & IFF_EBRIDGE; 4016} 4017 4018static inline bool netif_is_bridge_port(const struct net_device *dev) 4019{ 4020 return dev->priv_flags & IFF_BRIDGE_PORT; 4021} 4022 4023static inline bool netif_is_ovs_master(const struct net_device *dev) 4024{ 4025 return dev->priv_flags & IFF_OPENVSWITCH; 4026} 4027 4028static inline bool netif_is_team_master(const struct net_device *dev) 4029{ 4030 return dev->priv_flags & IFF_TEAM; 4031} 4032 4033static inline bool netif_is_team_port(const struct net_device *dev) 4034{ 4035 return dev->priv_flags & IFF_TEAM_PORT; 4036} 4037 4038static inline bool netif_is_lag_master(const struct net_device *dev) 4039{ 4040 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4041} 4042 4043static inline bool netif_is_lag_port(const struct net_device *dev) 4044{ 4045 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4046} 4047 4048/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4049static inline void netif_keep_dst(struct net_device *dev) 4050{ 4051 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4052} 4053 4054extern struct pernet_operations __net_initdata loopback_net_ops; 4055 4056/* Logging, debugging and troubleshooting/diagnostic helpers. */ 4057 4058/* netdev_printk helpers, similar to dev_printk */ 4059 4060static inline const char *netdev_name(const struct net_device *dev) 4061{ 4062 if (!dev->name[0] || strchr(dev->name, '%')) 4063 return "(unnamed net_device)"; 4064 return dev->name; 4065} 4066 4067static inline const char *netdev_reg_state(const struct net_device *dev) 4068{ 4069 switch (dev->reg_state) { 4070 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4071 case NETREG_REGISTERED: return ""; 4072 case NETREG_UNREGISTERING: return " (unregistering)"; 4073 case NETREG_UNREGISTERED: return " (unregistered)"; 4074 case NETREG_RELEASED: return " (released)"; 4075 case NETREG_DUMMY: return " (dummy)"; 4076 } 4077 4078 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4079 return " (unknown)"; 4080} 4081 4082__printf(3, 4) 4083void netdev_printk(const char *level, const struct net_device *dev, 4084 const char *format, ...); 4085__printf(2, 3) 4086void netdev_emerg(const struct net_device *dev, const char *format, ...); 4087__printf(2, 3) 4088void netdev_alert(const struct net_device *dev, const char *format, ...); 4089__printf(2, 3) 4090void netdev_crit(const struct net_device *dev, const char *format, ...); 4091__printf(2, 3) 4092void netdev_err(const struct net_device *dev, const char *format, ...); 4093__printf(2, 3) 4094void netdev_warn(const struct net_device *dev, const char *format, ...); 4095__printf(2, 3) 4096void netdev_notice(const struct net_device *dev, const char *format, ...); 4097__printf(2, 3) 4098void netdev_info(const struct net_device *dev, const char *format, ...); 4099 4100#define MODULE_ALIAS_NETDEV(device) \ 4101 MODULE_ALIAS("netdev-" device) 4102 4103#if defined(CONFIG_DYNAMIC_DEBUG) 4104#define netdev_dbg(__dev, format, args...) \ 4105do { \ 4106 dynamic_netdev_dbg(__dev, format, ##args); \ 4107} while (0) 4108#elif defined(DEBUG) 4109#define netdev_dbg(__dev, format, args...) \ 4110 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4111#else 4112#define netdev_dbg(__dev, format, args...) \ 4113({ \ 4114 if (0) \ 4115 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4116}) 4117#endif 4118 4119#if defined(VERBOSE_DEBUG) 4120#define netdev_vdbg netdev_dbg 4121#else 4122 4123#define netdev_vdbg(dev, format, args...) \ 4124({ \ 4125 if (0) \ 4126 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 4127 0; \ 4128}) 4129#endif 4130 4131/* 4132 * netdev_WARN() acts like dev_printk(), but with the key difference 4133 * of using a WARN/WARN_ON to get the message out, including the 4134 * file/line information and a backtrace. 4135 */ 4136#define netdev_WARN(dev, format, args...) \ 4137 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \ 4138 netdev_reg_state(dev), ##args) 4139 4140/* netif printk helpers, similar to netdev_printk */ 4141 4142#define netif_printk(priv, type, level, dev, fmt, args...) \ 4143do { \ 4144 if (netif_msg_##type(priv)) \ 4145 netdev_printk(level, (dev), fmt, ##args); \ 4146} while (0) 4147 4148#define netif_level(level, priv, type, dev, fmt, args...) \ 4149do { \ 4150 if (netif_msg_##type(priv)) \ 4151 netdev_##level(dev, fmt, ##args); \ 4152} while (0) 4153 4154#define netif_emerg(priv, type, dev, fmt, args...) \ 4155 netif_level(emerg, priv, type, dev, fmt, ##args) 4156#define netif_alert(priv, type, dev, fmt, args...) \ 4157 netif_level(alert, priv, type, dev, fmt, ##args) 4158#define netif_crit(priv, type, dev, fmt, args...) \ 4159 netif_level(crit, priv, type, dev, fmt, ##args) 4160#define netif_err(priv, type, dev, fmt, args...) \ 4161 netif_level(err, priv, type, dev, fmt, ##args) 4162#define netif_warn(priv, type, dev, fmt, args...) \ 4163 netif_level(warn, priv, type, dev, fmt, ##args) 4164#define netif_notice(priv, type, dev, fmt, args...) \ 4165 netif_level(notice, priv, type, dev, fmt, ##args) 4166#define netif_info(priv, type, dev, fmt, args...) \ 4167 netif_level(info, priv, type, dev, fmt, ##args) 4168 4169#if defined(CONFIG_DYNAMIC_DEBUG) 4170#define netif_dbg(priv, type, netdev, format, args...) \ 4171do { \ 4172 if (netif_msg_##type(priv)) \ 4173 dynamic_netdev_dbg(netdev, format, ##args); \ 4174} while (0) 4175#elif defined(DEBUG) 4176#define netif_dbg(priv, type, dev, format, args...) \ 4177 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 4178#else 4179#define netif_dbg(priv, type, dev, format, args...) \ 4180({ \ 4181 if (0) \ 4182 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4183 0; \ 4184}) 4185#endif 4186 4187#if defined(VERBOSE_DEBUG) 4188#define netif_vdbg netif_dbg 4189#else 4190#define netif_vdbg(priv, type, dev, format, args...) \ 4191({ \ 4192 if (0) \ 4193 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4194 0; \ 4195}) 4196#endif 4197 4198/* 4199 * The list of packet types we will receive (as opposed to discard) 4200 * and the routines to invoke. 4201 * 4202 * Why 16. Because with 16 the only overlap we get on a hash of the 4203 * low nibble of the protocol value is RARP/SNAP/X.25. 4204 * 4205 * NOTE: That is no longer true with the addition of VLAN tags. Not 4206 * sure which should go first, but I bet it won't make much 4207 * difference if we are running VLANs. The good news is that 4208 * this protocol won't be in the list unless compiled in, so 4209 * the average user (w/out VLANs) will not be adversely affected. 4210 * --BLG 4211 * 4212 * 0800 IP 4213 * 8100 802.1Q VLAN 4214 * 0001 802.3 4215 * 0002 AX.25 4216 * 0004 802.2 4217 * 8035 RARP 4218 * 0005 SNAP 4219 * 0805 X.25 4220 * 0806 ARP 4221 * 8137 IPX 4222 * 0009 Localtalk 4223 * 86DD IPv6 4224 */ 4225#define PTYPE_HASH_SIZE (16) 4226#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 4227 4228#endif /* _LINUX_NETDEVICE_H */