at v3.11 94 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/pm_qos.h> 29#include <linux/timer.h> 30#include <linux/bug.h> 31#include <linux/delay.h> 32#include <linux/atomic.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 54struct netpoll_info; 55struct device; 56struct phy_device; 57/* 802.11 specific */ 58struct wireless_dev; 59 /* source back-compat hooks */ 60#define SET_ETHTOOL_OPS(netdev,ops) \ 61 ( (netdev)->ethtool_ops = (ops) ) 62 63extern void netdev_set_default_ethtool_ops(struct net_device *dev, 64 const struct ethtool_ops *ops); 65 66/* hardware address assignment types */ 67#define NET_ADDR_PERM 0 /* address is permanent (default) */ 68#define NET_ADDR_RANDOM 1 /* address is generated randomly */ 69#define NET_ADDR_STOLEN 2 /* address is stolen from other device */ 70#define NET_ADDR_SET 3 /* address is set using 71 * dev_set_mac_address() */ 72 73/* Backlog congestion levels */ 74#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 75#define NET_RX_DROP 1 /* packet dropped */ 76 77/* 78 * Transmit return codes: transmit return codes originate from three different 79 * namespaces: 80 * 81 * - qdisc return codes 82 * - driver transmit return codes 83 * - errno values 84 * 85 * Drivers are allowed to return any one of those in their hard_start_xmit() 86 * function. Real network devices commonly used with qdiscs should only return 87 * the driver transmit return codes though - when qdiscs are used, the actual 88 * transmission happens asynchronously, so the value is not propagated to 89 * higher layers. Virtual network devices transmit synchronously, in this case 90 * the driver transmit return codes are consumed by dev_queue_xmit(), all 91 * others are propagated to higher layers. 92 */ 93 94/* qdisc ->enqueue() return codes. */ 95#define NET_XMIT_SUCCESS 0x00 96#define NET_XMIT_DROP 0x01 /* skb dropped */ 97#define NET_XMIT_CN 0x02 /* congestion notification */ 98#define NET_XMIT_POLICED 0x03 /* skb is shot by police */ 99#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 100 101/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 102 * indicates that the device will soon be dropping packets, or already drops 103 * some packets of the same priority; prompting us to send less aggressively. */ 104#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 105#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 106 107/* Driver transmit return codes */ 108#define NETDEV_TX_MASK 0xf0 109 110enum netdev_tx { 111 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 112 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 113 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 114 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */ 115}; 116typedef enum netdev_tx netdev_tx_t; 117 118/* 119 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 120 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 121 */ 122static inline bool dev_xmit_complete(int rc) 123{ 124 /* 125 * Positive cases with an skb consumed by a driver: 126 * - successful transmission (rc == NETDEV_TX_OK) 127 * - error while transmitting (rc < 0) 128 * - error while queueing to a different device (rc & NET_XMIT_MASK) 129 */ 130 if (likely(rc < NET_XMIT_MASK)) 131 return true; 132 133 return false; 134} 135 136/* 137 * Compute the worst case header length according to the protocols 138 * used. 139 */ 140 141#if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 142# if defined(CONFIG_MAC80211_MESH) 143# define LL_MAX_HEADER 128 144# else 145# define LL_MAX_HEADER 96 146# endif 147#else 148# define LL_MAX_HEADER 32 149#endif 150 151#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 153#define MAX_HEADER LL_MAX_HEADER 154#else 155#define MAX_HEADER (LL_MAX_HEADER + 48) 156#endif 157 158/* 159 * Old network device statistics. Fields are native words 160 * (unsigned long) so they can be read and written atomically. 161 */ 162 163struct net_device_stats { 164 unsigned long rx_packets; 165 unsigned long tx_packets; 166 unsigned long rx_bytes; 167 unsigned long tx_bytes; 168 unsigned long rx_errors; 169 unsigned long tx_errors; 170 unsigned long rx_dropped; 171 unsigned long tx_dropped; 172 unsigned long multicast; 173 unsigned long collisions; 174 unsigned long rx_length_errors; 175 unsigned long rx_over_errors; 176 unsigned long rx_crc_errors; 177 unsigned long rx_frame_errors; 178 unsigned long rx_fifo_errors; 179 unsigned long rx_missed_errors; 180 unsigned long tx_aborted_errors; 181 unsigned long tx_carrier_errors; 182 unsigned long tx_fifo_errors; 183 unsigned long tx_heartbeat_errors; 184 unsigned long tx_window_errors; 185 unsigned long rx_compressed; 186 unsigned long tx_compressed; 187}; 188 189 190#include <linux/cache.h> 191#include <linux/skbuff.h> 192 193#ifdef CONFIG_RPS 194#include <linux/static_key.h> 195extern struct static_key rps_needed; 196#endif 197 198struct neighbour; 199struct neigh_parms; 200struct sk_buff; 201 202struct netdev_hw_addr { 203 struct list_head list; 204 unsigned char addr[MAX_ADDR_LEN]; 205 unsigned char type; 206#define NETDEV_HW_ADDR_T_LAN 1 207#define NETDEV_HW_ADDR_T_SAN 2 208#define NETDEV_HW_ADDR_T_SLAVE 3 209#define NETDEV_HW_ADDR_T_UNICAST 4 210#define NETDEV_HW_ADDR_T_MULTICAST 5 211 bool global_use; 212 int sync_cnt; 213 int refcount; 214 int synced; 215 struct rcu_head rcu_head; 216}; 217 218struct netdev_hw_addr_list { 219 struct list_head list; 220 int count; 221}; 222 223#define netdev_hw_addr_list_count(l) ((l)->count) 224#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 225#define netdev_hw_addr_list_for_each(ha, l) \ 226 list_for_each_entry(ha, &(l)->list, list) 227 228#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 229#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 230#define netdev_for_each_uc_addr(ha, dev) \ 231 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 232 233#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 234#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 235#define netdev_for_each_mc_addr(ha, dev) \ 236 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 237 238struct hh_cache { 239 u16 hh_len; 240 u16 __pad; 241 seqlock_t hh_lock; 242 243 /* cached hardware header; allow for machine alignment needs. */ 244#define HH_DATA_MOD 16 245#define HH_DATA_OFF(__len) \ 246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 247#define HH_DATA_ALIGN(__len) \ 248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 249 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 250}; 251 252/* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much. 253 * Alternative is: 254 * dev->hard_header_len ? (dev->hard_header_len + 255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 256 * 257 * We could use other alignment values, but we must maintain the 258 * relationship HH alignment <= LL alignment. 259 */ 260#define LL_RESERVED_SPACE(dev) \ 261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 262#define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 264 265struct header_ops { 266 int (*create) (struct sk_buff *skb, struct net_device *dev, 267 unsigned short type, const void *daddr, 268 const void *saddr, unsigned int len); 269 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 270 int (*rebuild)(struct sk_buff *skb); 271 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 272 void (*cache_update)(struct hh_cache *hh, 273 const struct net_device *dev, 274 const unsigned char *haddr); 275}; 276 277/* These flag bits are private to the generic network queueing 278 * layer, they may not be explicitly referenced by any other 279 * code. 280 */ 281 282enum netdev_state_t { 283 __LINK_STATE_START, 284 __LINK_STATE_PRESENT, 285 __LINK_STATE_NOCARRIER, 286 __LINK_STATE_LINKWATCH_PENDING, 287 __LINK_STATE_DORMANT, 288}; 289 290 291/* 292 * This structure holds at boot time configured netdevice settings. They 293 * are then used in the device probing. 294 */ 295struct netdev_boot_setup { 296 char name[IFNAMSIZ]; 297 struct ifmap map; 298}; 299#define NETDEV_BOOT_SETUP_MAX 8 300 301extern int __init netdev_boot_setup(char *str); 302 303/* 304 * Structure for NAPI scheduling similar to tasklet but with weighting 305 */ 306struct napi_struct { 307 /* The poll_list must only be managed by the entity which 308 * changes the state of the NAPI_STATE_SCHED bit. This means 309 * whoever atomically sets that bit can add this napi_struct 310 * to the per-cpu poll_list, and whoever clears that bit 311 * can remove from the list right before clearing the bit. 312 */ 313 struct list_head poll_list; 314 315 unsigned long state; 316 int weight; 317 unsigned int gro_count; 318 int (*poll)(struct napi_struct *, int); 319#ifdef CONFIG_NETPOLL 320 spinlock_t poll_lock; 321 int poll_owner; 322#endif 323 struct net_device *dev; 324 struct sk_buff *gro_list; 325 struct sk_buff *skb; 326 struct list_head dev_list; 327 struct hlist_node napi_hash_node; 328 unsigned int napi_id; 329}; 330 331enum { 332 NAPI_STATE_SCHED, /* Poll is scheduled */ 333 NAPI_STATE_DISABLE, /* Disable pending */ 334 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 335 NAPI_STATE_HASHED, /* In NAPI hash */ 336}; 337 338enum gro_result { 339 GRO_MERGED, 340 GRO_MERGED_FREE, 341 GRO_HELD, 342 GRO_NORMAL, 343 GRO_DROP, 344}; 345typedef enum gro_result gro_result_t; 346 347/* 348 * enum rx_handler_result - Possible return values for rx_handlers. 349 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 350 * further. 351 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 352 * case skb->dev was changed by rx_handler. 353 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 354 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called. 355 * 356 * rx_handlers are functions called from inside __netif_receive_skb(), to do 357 * special processing of the skb, prior to delivery to protocol handlers. 358 * 359 * Currently, a net_device can only have a single rx_handler registered. Trying 360 * to register a second rx_handler will return -EBUSY. 361 * 362 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 363 * To unregister a rx_handler on a net_device, use 364 * netdev_rx_handler_unregister(). 365 * 366 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 367 * do with the skb. 368 * 369 * If the rx_handler consumed to skb in some way, it should return 370 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 371 * the skb to be delivered in some other ways. 372 * 373 * If the rx_handler changed skb->dev, to divert the skb to another 374 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 375 * new device will be called if it exists. 376 * 377 * If the rx_handler consider the skb should be ignored, it should return 378 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 379 * are registered on exact device (ptype->dev == skb->dev). 380 * 381 * If the rx_handler didn't changed skb->dev, but want the skb to be normally 382 * delivered, it should return RX_HANDLER_PASS. 383 * 384 * A device without a registered rx_handler will behave as if rx_handler 385 * returned RX_HANDLER_PASS. 386 */ 387 388enum rx_handler_result { 389 RX_HANDLER_CONSUMED, 390 RX_HANDLER_ANOTHER, 391 RX_HANDLER_EXACT, 392 RX_HANDLER_PASS, 393}; 394typedef enum rx_handler_result rx_handler_result_t; 395typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 396 397extern void __napi_schedule(struct napi_struct *n); 398 399static inline bool napi_disable_pending(struct napi_struct *n) 400{ 401 return test_bit(NAPI_STATE_DISABLE, &n->state); 402} 403 404/** 405 * napi_schedule_prep - check if napi can be scheduled 406 * @n: napi context 407 * 408 * Test if NAPI routine is already running, and if not mark 409 * it as running. This is used as a condition variable 410 * insure only one NAPI poll instance runs. We also make 411 * sure there is no pending NAPI disable. 412 */ 413static inline bool napi_schedule_prep(struct napi_struct *n) 414{ 415 return !napi_disable_pending(n) && 416 !test_and_set_bit(NAPI_STATE_SCHED, &n->state); 417} 418 419/** 420 * napi_schedule - schedule NAPI poll 421 * @n: napi context 422 * 423 * Schedule NAPI poll routine to be called if it is not already 424 * running. 425 */ 426static inline void napi_schedule(struct napi_struct *n) 427{ 428 if (napi_schedule_prep(n)) 429 __napi_schedule(n); 430} 431 432/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 433static inline bool napi_reschedule(struct napi_struct *napi) 434{ 435 if (napi_schedule_prep(napi)) { 436 __napi_schedule(napi); 437 return true; 438 } 439 return false; 440} 441 442/** 443 * napi_complete - NAPI processing complete 444 * @n: napi context 445 * 446 * Mark NAPI processing as complete. 447 */ 448extern void __napi_complete(struct napi_struct *n); 449extern void napi_complete(struct napi_struct *n); 450 451/** 452 * napi_by_id - lookup a NAPI by napi_id 453 * @napi_id: hashed napi_id 454 * 455 * lookup @napi_id in napi_hash table 456 * must be called under rcu_read_lock() 457 */ 458extern struct napi_struct *napi_by_id(unsigned int napi_id); 459 460/** 461 * napi_hash_add - add a NAPI to global hashtable 462 * @napi: napi context 463 * 464 * generate a new napi_id and store a @napi under it in napi_hash 465 */ 466extern void napi_hash_add(struct napi_struct *napi); 467 468/** 469 * napi_hash_del - remove a NAPI from global table 470 * @napi: napi context 471 * 472 * Warning: caller must observe rcu grace period 473 * before freeing memory containing @napi 474 */ 475extern void napi_hash_del(struct napi_struct *napi); 476 477/** 478 * napi_disable - prevent NAPI from scheduling 479 * @n: napi context 480 * 481 * Stop NAPI from being scheduled on this context. 482 * Waits till any outstanding processing completes. 483 */ 484static inline void napi_disable(struct napi_struct *n) 485{ 486 set_bit(NAPI_STATE_DISABLE, &n->state); 487 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state)) 488 msleep(1); 489 clear_bit(NAPI_STATE_DISABLE, &n->state); 490} 491 492/** 493 * napi_enable - enable NAPI scheduling 494 * @n: napi context 495 * 496 * Resume NAPI from being scheduled on this context. 497 * Must be paired with napi_disable. 498 */ 499static inline void napi_enable(struct napi_struct *n) 500{ 501 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 502 smp_mb__before_clear_bit(); 503 clear_bit(NAPI_STATE_SCHED, &n->state); 504} 505 506#ifdef CONFIG_SMP 507/** 508 * napi_synchronize - wait until NAPI is not running 509 * @n: napi context 510 * 511 * Wait until NAPI is done being scheduled on this context. 512 * Waits till any outstanding processing completes but 513 * does not disable future activations. 514 */ 515static inline void napi_synchronize(const struct napi_struct *n) 516{ 517 while (test_bit(NAPI_STATE_SCHED, &n->state)) 518 msleep(1); 519} 520#else 521# define napi_synchronize(n) barrier() 522#endif 523 524enum netdev_queue_state_t { 525 __QUEUE_STATE_DRV_XOFF, 526 __QUEUE_STATE_STACK_XOFF, 527 __QUEUE_STATE_FROZEN, 528#define QUEUE_STATE_ANY_XOFF ((1 << __QUEUE_STATE_DRV_XOFF) | \ 529 (1 << __QUEUE_STATE_STACK_XOFF)) 530#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 531 (1 << __QUEUE_STATE_FROZEN)) 532}; 533/* 534 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 535 * netif_tx_* functions below are used to manipulate this flag. The 536 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 537 * queue independently. The netif_xmit_*stopped functions below are called 538 * to check if the queue has been stopped by the driver or stack (either 539 * of the XOFF bits are set in the state). Drivers should not need to call 540 * netif_xmit*stopped functions, they should only be using netif_tx_*. 541 */ 542 543struct netdev_queue { 544/* 545 * read mostly part 546 */ 547 struct net_device *dev; 548 struct Qdisc *qdisc; 549 struct Qdisc *qdisc_sleeping; 550#ifdef CONFIG_SYSFS 551 struct kobject kobj; 552#endif 553#if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 554 int numa_node; 555#endif 556/* 557 * write mostly part 558 */ 559 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 560 int xmit_lock_owner; 561 /* 562 * please use this field instead of dev->trans_start 563 */ 564 unsigned long trans_start; 565 566 /* 567 * Number of TX timeouts for this queue 568 * (/sys/class/net/DEV/Q/trans_timeout) 569 */ 570 unsigned long trans_timeout; 571 572 unsigned long state; 573 574#ifdef CONFIG_BQL 575 struct dql dql; 576#endif 577} ____cacheline_aligned_in_smp; 578 579static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 580{ 581#if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 582 return q->numa_node; 583#else 584 return NUMA_NO_NODE; 585#endif 586} 587 588static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 589{ 590#if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 591 q->numa_node = node; 592#endif 593} 594 595#ifdef CONFIG_RPS 596/* 597 * This structure holds an RPS map which can be of variable length. The 598 * map is an array of CPUs. 599 */ 600struct rps_map { 601 unsigned int len; 602 struct rcu_head rcu; 603 u16 cpus[0]; 604}; 605#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 606 607/* 608 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 609 * tail pointer for that CPU's input queue at the time of last enqueue, and 610 * a hardware filter index. 611 */ 612struct rps_dev_flow { 613 u16 cpu; 614 u16 filter; 615 unsigned int last_qtail; 616}; 617#define RPS_NO_FILTER 0xffff 618 619/* 620 * The rps_dev_flow_table structure contains a table of flow mappings. 621 */ 622struct rps_dev_flow_table { 623 unsigned int mask; 624 struct rcu_head rcu; 625 struct rps_dev_flow flows[0]; 626}; 627#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 628 ((_num) * sizeof(struct rps_dev_flow))) 629 630/* 631 * The rps_sock_flow_table contains mappings of flows to the last CPU 632 * on which they were processed by the application (set in recvmsg). 633 */ 634struct rps_sock_flow_table { 635 unsigned int mask; 636 u16 ents[0]; 637}; 638#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \ 639 ((_num) * sizeof(u16))) 640 641#define RPS_NO_CPU 0xffff 642 643static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 644 u32 hash) 645{ 646 if (table && hash) { 647 unsigned int cpu, index = hash & table->mask; 648 649 /* We only give a hint, preemption can change cpu under us */ 650 cpu = raw_smp_processor_id(); 651 652 if (table->ents[index] != cpu) 653 table->ents[index] = cpu; 654 } 655} 656 657static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table, 658 u32 hash) 659{ 660 if (table && hash) 661 table->ents[hash & table->mask] = RPS_NO_CPU; 662} 663 664extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 665 666#ifdef CONFIG_RFS_ACCEL 667extern bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, 668 u32 flow_id, u16 filter_id); 669#endif 670 671/* This structure contains an instance of an RX queue. */ 672struct netdev_rx_queue { 673 struct rps_map __rcu *rps_map; 674 struct rps_dev_flow_table __rcu *rps_flow_table; 675 struct kobject kobj; 676 struct net_device *dev; 677} ____cacheline_aligned_in_smp; 678#endif /* CONFIG_RPS */ 679 680#ifdef CONFIG_XPS 681/* 682 * This structure holds an XPS map which can be of variable length. The 683 * map is an array of queues. 684 */ 685struct xps_map { 686 unsigned int len; 687 unsigned int alloc_len; 688 struct rcu_head rcu; 689 u16 queues[0]; 690}; 691#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 692#define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \ 693 / sizeof(u16)) 694 695/* 696 * This structure holds all XPS maps for device. Maps are indexed by CPU. 697 */ 698struct xps_dev_maps { 699 struct rcu_head rcu; 700 struct xps_map __rcu *cpu_map[0]; 701}; 702#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \ 703 (nr_cpu_ids * sizeof(struct xps_map *))) 704#endif /* CONFIG_XPS */ 705 706#define TC_MAX_QUEUE 16 707#define TC_BITMASK 15 708/* HW offloaded queuing disciplines txq count and offset maps */ 709struct netdev_tc_txq { 710 u16 count; 711 u16 offset; 712}; 713 714#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 715/* 716 * This structure is to hold information about the device 717 * configured to run FCoE protocol stack. 718 */ 719struct netdev_fcoe_hbainfo { 720 char manufacturer[64]; 721 char serial_number[64]; 722 char hardware_version[64]; 723 char driver_version[64]; 724 char optionrom_version[64]; 725 char firmware_version[64]; 726 char model[256]; 727 char model_description[256]; 728}; 729#endif 730 731/* 732 * This structure defines the management hooks for network devices. 733 * The following hooks can be defined; unless noted otherwise, they are 734 * optional and can be filled with a null pointer. 735 * 736 * int (*ndo_init)(struct net_device *dev); 737 * This function is called once when network device is registered. 738 * The network device can use this to any late stage initializaton 739 * or semantic validattion. It can fail with an error code which will 740 * be propogated back to register_netdev 741 * 742 * void (*ndo_uninit)(struct net_device *dev); 743 * This function is called when device is unregistered or when registration 744 * fails. It is not called if init fails. 745 * 746 * int (*ndo_open)(struct net_device *dev); 747 * This function is called when network device transistions to the up 748 * state. 749 * 750 * int (*ndo_stop)(struct net_device *dev); 751 * This function is called when network device transistions to the down 752 * state. 753 * 754 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 755 * struct net_device *dev); 756 * Called when a packet needs to be transmitted. 757 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY. 758 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX) 759 * Required can not be NULL. 760 * 761 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb); 762 * Called to decide which queue to when device supports multiple 763 * transmit queues. 764 * 765 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 766 * This function is called to allow device receiver to make 767 * changes to configuration when multicast or promiscious is enabled. 768 * 769 * void (*ndo_set_rx_mode)(struct net_device *dev); 770 * This function is called device changes address list filtering. 771 * If driver handles unicast address filtering, it should set 772 * IFF_UNICAST_FLT to its priv_flags. 773 * 774 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 775 * This function is called when the Media Access Control address 776 * needs to be changed. If this interface is not defined, the 777 * mac address can not be changed. 778 * 779 * int (*ndo_validate_addr)(struct net_device *dev); 780 * Test if Media Access Control address is valid for the device. 781 * 782 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 783 * Called when a user request an ioctl which can't be handled by 784 * the generic interface code. If not defined ioctl's return 785 * not supported error code. 786 * 787 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 788 * Used to set network devices bus interface parameters. This interface 789 * is retained for legacy reason, new devices should use the bus 790 * interface (PCI) for low level management. 791 * 792 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 793 * Called when a user wants to change the Maximum Transfer Unit 794 * of a device. If not defined, any request to change MTU will 795 * will return an error. 796 * 797 * void (*ndo_tx_timeout)(struct net_device *dev); 798 * Callback uses when the transmitter has not made any progress 799 * for dev->watchdog ticks. 800 * 801 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev, 802 * struct rtnl_link_stats64 *storage); 803 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 804 * Called when a user wants to get the network device usage 805 * statistics. Drivers must do one of the following: 806 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 807 * rtnl_link_stats64 structure passed by the caller. 808 * 2. Define @ndo_get_stats to update a net_device_stats structure 809 * (which should normally be dev->stats) and return a pointer to 810 * it. The structure may be changed asynchronously only if each 811 * field is written atomically. 812 * 3. Update dev->stats asynchronously and atomically, and define 813 * neither operation. 814 * 815 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid); 816 * If device support VLAN filtering this function is called when a 817 * VLAN id is registered. 818 * 819 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid); 820 * If device support VLAN filtering this function is called when a 821 * VLAN id is unregistered. 822 * 823 * void (*ndo_poll_controller)(struct net_device *dev); 824 * 825 * SR-IOV management functions. 826 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 827 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos); 828 * int (*ndo_set_vf_tx_rate)(struct net_device *dev, int vf, int rate); 829 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 830 * int (*ndo_get_vf_config)(struct net_device *dev, 831 * int vf, struct ifla_vf_info *ivf); 832 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 833 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 834 * struct nlattr *port[]); 835 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 836 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc) 837 * Called to setup 'tc' number of traffic classes in the net device. This 838 * is always called from the stack with the rtnl lock held and netif tx 839 * queues stopped. This allows the netdevice to perform queue management 840 * safely. 841 * 842 * Fiber Channel over Ethernet (FCoE) offload functions. 843 * int (*ndo_fcoe_enable)(struct net_device *dev); 844 * Called when the FCoE protocol stack wants to start using LLD for FCoE 845 * so the underlying device can perform whatever needed configuration or 846 * initialization to support acceleration of FCoE traffic. 847 * 848 * int (*ndo_fcoe_disable)(struct net_device *dev); 849 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 850 * so the underlying device can perform whatever needed clean-ups to 851 * stop supporting acceleration of FCoE traffic. 852 * 853 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 854 * struct scatterlist *sgl, unsigned int sgc); 855 * Called when the FCoE Initiator wants to initialize an I/O that 856 * is a possible candidate for Direct Data Placement (DDP). The LLD can 857 * perform necessary setup and returns 1 to indicate the device is set up 858 * successfully to perform DDP on this I/O, otherwise this returns 0. 859 * 860 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 861 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 862 * indicated by the FC exchange id 'xid', so the underlying device can 863 * clean up and reuse resources for later DDP requests. 864 * 865 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 866 * struct scatterlist *sgl, unsigned int sgc); 867 * Called when the FCoE Target wants to initialize an I/O that 868 * is a possible candidate for Direct Data Placement (DDP). The LLD can 869 * perform necessary setup and returns 1 to indicate the device is set up 870 * successfully to perform DDP on this I/O, otherwise this returns 0. 871 * 872 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 873 * struct netdev_fcoe_hbainfo *hbainfo); 874 * Called when the FCoE Protocol stack wants information on the underlying 875 * device. This information is utilized by the FCoE protocol stack to 876 * register attributes with Fiber Channel management service as per the 877 * FC-GS Fabric Device Management Information(FDMI) specification. 878 * 879 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 880 * Called when the underlying device wants to override default World Wide 881 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 882 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 883 * protocol stack to use. 884 * 885 * RFS acceleration. 886 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 887 * u16 rxq_index, u32 flow_id); 888 * Set hardware filter for RFS. rxq_index is the target queue index; 889 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 890 * Return the filter ID on success, or a negative error code. 891 * 892 * Slave management functions (for bridge, bonding, etc). 893 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 894 * Called to make another netdev an underling. 895 * 896 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 897 * Called to release previously enslaved netdev. 898 * 899 * Feature/offload setting functions. 900 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 901 * netdev_features_t features); 902 * Adjusts the requested feature flags according to device-specific 903 * constraints, and returns the resulting flags. Must not modify 904 * the device state. 905 * 906 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 907 * Called to update device configuration to new features. Passed 908 * feature set might be less than what was returned by ndo_fix_features()). 909 * Must return >0 or -errno if it changed dev->features itself. 910 * 911 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 912 * struct net_device *dev, 913 * const unsigned char *addr, u16 flags) 914 * Adds an FDB entry to dev for addr. 915 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 916 * struct net_device *dev, 917 * const unsigned char *addr) 918 * Deletes the FDB entry from dev coresponding to addr. 919 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 920 * struct net_device *dev, int idx) 921 * Used to add FDB entries to dump requests. Implementers should add 922 * entries to skb and update idx with the number of entries. 923 * 924 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh) 925 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 926 * struct net_device *dev, u32 filter_mask) 927 * 928 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 929 * Called to change device carrier. Soft-devices (like dummy, team, etc) 930 * which do not represent real hardware may define this to allow their 931 * userspace components to manage their virtual carrier state. Devices 932 * that determine carrier state from physical hardware properties (eg 933 * network cables) or protocol-dependent mechanisms (eg 934 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 935 */ 936struct net_device_ops { 937 int (*ndo_init)(struct net_device *dev); 938 void (*ndo_uninit)(struct net_device *dev); 939 int (*ndo_open)(struct net_device *dev); 940 int (*ndo_stop)(struct net_device *dev); 941 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb, 942 struct net_device *dev); 943 u16 (*ndo_select_queue)(struct net_device *dev, 944 struct sk_buff *skb); 945 void (*ndo_change_rx_flags)(struct net_device *dev, 946 int flags); 947 void (*ndo_set_rx_mode)(struct net_device *dev); 948 int (*ndo_set_mac_address)(struct net_device *dev, 949 void *addr); 950 int (*ndo_validate_addr)(struct net_device *dev); 951 int (*ndo_do_ioctl)(struct net_device *dev, 952 struct ifreq *ifr, int cmd); 953 int (*ndo_set_config)(struct net_device *dev, 954 struct ifmap *map); 955 int (*ndo_change_mtu)(struct net_device *dev, 956 int new_mtu); 957 int (*ndo_neigh_setup)(struct net_device *dev, 958 struct neigh_parms *); 959 void (*ndo_tx_timeout) (struct net_device *dev); 960 961 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev, 962 struct rtnl_link_stats64 *storage); 963 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 964 965 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 966 __be16 proto, u16 vid); 967 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 968 __be16 proto, u16 vid); 969#ifdef CONFIG_NET_POLL_CONTROLLER 970 void (*ndo_poll_controller)(struct net_device *dev); 971 int (*ndo_netpoll_setup)(struct net_device *dev, 972 struct netpoll_info *info, 973 gfp_t gfp); 974 void (*ndo_netpoll_cleanup)(struct net_device *dev); 975#endif 976#ifdef CONFIG_NET_RX_BUSY_POLL 977 int (*ndo_busy_poll)(struct napi_struct *dev); 978#endif 979 int (*ndo_set_vf_mac)(struct net_device *dev, 980 int queue, u8 *mac); 981 int (*ndo_set_vf_vlan)(struct net_device *dev, 982 int queue, u16 vlan, u8 qos); 983 int (*ndo_set_vf_tx_rate)(struct net_device *dev, 984 int vf, int rate); 985 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 986 int vf, bool setting); 987 int (*ndo_get_vf_config)(struct net_device *dev, 988 int vf, 989 struct ifla_vf_info *ivf); 990 int (*ndo_set_vf_link_state)(struct net_device *dev, 991 int vf, int link_state); 992 int (*ndo_set_vf_port)(struct net_device *dev, 993 int vf, 994 struct nlattr *port[]); 995 int (*ndo_get_vf_port)(struct net_device *dev, 996 int vf, struct sk_buff *skb); 997 int (*ndo_setup_tc)(struct net_device *dev, u8 tc); 998#if IS_ENABLED(CONFIG_FCOE) 999 int (*ndo_fcoe_enable)(struct net_device *dev); 1000 int (*ndo_fcoe_disable)(struct net_device *dev); 1001 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1002 u16 xid, 1003 struct scatterlist *sgl, 1004 unsigned int sgc); 1005 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1006 u16 xid); 1007 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1008 u16 xid, 1009 struct scatterlist *sgl, 1010 unsigned int sgc); 1011 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1012 struct netdev_fcoe_hbainfo *hbainfo); 1013#endif 1014 1015#if IS_ENABLED(CONFIG_LIBFCOE) 1016#define NETDEV_FCOE_WWNN 0 1017#define NETDEV_FCOE_WWPN 1 1018 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1019 u64 *wwn, int type); 1020#endif 1021 1022#ifdef CONFIG_RFS_ACCEL 1023 int (*ndo_rx_flow_steer)(struct net_device *dev, 1024 const struct sk_buff *skb, 1025 u16 rxq_index, 1026 u32 flow_id); 1027#endif 1028 int (*ndo_add_slave)(struct net_device *dev, 1029 struct net_device *slave_dev); 1030 int (*ndo_del_slave)(struct net_device *dev, 1031 struct net_device *slave_dev); 1032 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1033 netdev_features_t features); 1034 int (*ndo_set_features)(struct net_device *dev, 1035 netdev_features_t features); 1036 int (*ndo_neigh_construct)(struct neighbour *n); 1037 void (*ndo_neigh_destroy)(struct neighbour *n); 1038 1039 int (*ndo_fdb_add)(struct ndmsg *ndm, 1040 struct nlattr *tb[], 1041 struct net_device *dev, 1042 const unsigned char *addr, 1043 u16 flags); 1044 int (*ndo_fdb_del)(struct ndmsg *ndm, 1045 struct nlattr *tb[], 1046 struct net_device *dev, 1047 const unsigned char *addr); 1048 int (*ndo_fdb_dump)(struct sk_buff *skb, 1049 struct netlink_callback *cb, 1050 struct net_device *dev, 1051 int idx); 1052 1053 int (*ndo_bridge_setlink)(struct net_device *dev, 1054 struct nlmsghdr *nlh); 1055 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1056 u32 pid, u32 seq, 1057 struct net_device *dev, 1058 u32 filter_mask); 1059 int (*ndo_bridge_dellink)(struct net_device *dev, 1060 struct nlmsghdr *nlh); 1061 int (*ndo_change_carrier)(struct net_device *dev, 1062 bool new_carrier); 1063}; 1064 1065/* 1066 * The DEVICE structure. 1067 * Actually, this whole structure is a big mistake. It mixes I/O 1068 * data with strictly "high-level" data, and it has to know about 1069 * almost every data structure used in the INET module. 1070 * 1071 * FIXME: cleanup struct net_device such that network protocol info 1072 * moves out. 1073 */ 1074 1075struct net_device { 1076 1077 /* 1078 * This is the first field of the "visible" part of this structure 1079 * (i.e. as seen by users in the "Space.c" file). It is the name 1080 * of the interface. 1081 */ 1082 char name[IFNAMSIZ]; 1083 1084 /* device name hash chain, please keep it close to name[] */ 1085 struct hlist_node name_hlist; 1086 1087 /* snmp alias */ 1088 char *ifalias; 1089 1090 /* 1091 * I/O specific fields 1092 * FIXME: Merge these and struct ifmap into one 1093 */ 1094 unsigned long mem_end; /* shared mem end */ 1095 unsigned long mem_start; /* shared mem start */ 1096 unsigned long base_addr; /* device I/O address */ 1097 unsigned int irq; /* device IRQ number */ 1098 1099 /* 1100 * Some hardware also needs these fields, but they are not 1101 * part of the usual set specified in Space.c. 1102 */ 1103 1104 unsigned long state; 1105 1106 struct list_head dev_list; 1107 struct list_head napi_list; 1108 struct list_head unreg_list; 1109 struct list_head upper_dev_list; /* List of upper devices */ 1110 1111 1112 /* currently active device features */ 1113 netdev_features_t features; 1114 /* user-changeable features */ 1115 netdev_features_t hw_features; 1116 /* user-requested features */ 1117 netdev_features_t wanted_features; 1118 /* mask of features inheritable by VLAN devices */ 1119 netdev_features_t vlan_features; 1120 /* mask of features inherited by encapsulating devices 1121 * This field indicates what encapsulation offloads 1122 * the hardware is capable of doing, and drivers will 1123 * need to set them appropriately. 1124 */ 1125 netdev_features_t hw_enc_features; 1126 /* mask of fetures inheritable by MPLS */ 1127 netdev_features_t mpls_features; 1128 1129 /* Interface index. Unique device identifier */ 1130 int ifindex; 1131 int iflink; 1132 1133 struct net_device_stats stats; 1134 atomic_long_t rx_dropped; /* dropped packets by core network 1135 * Do not use this in drivers. 1136 */ 1137 1138#ifdef CONFIG_WIRELESS_EXT 1139 /* List of functions to handle Wireless Extensions (instead of ioctl). 1140 * See <net/iw_handler.h> for details. Jean II */ 1141 const struct iw_handler_def * wireless_handlers; 1142 /* Instance data managed by the core of Wireless Extensions. */ 1143 struct iw_public_data * wireless_data; 1144#endif 1145 /* Management operations */ 1146 const struct net_device_ops *netdev_ops; 1147 const struct ethtool_ops *ethtool_ops; 1148 1149 /* Hardware header description */ 1150 const struct header_ops *header_ops; 1151 1152 unsigned int flags; /* interface flags (a la BSD) */ 1153 unsigned int priv_flags; /* Like 'flags' but invisible to userspace. 1154 * See if.h for definitions. */ 1155 unsigned short gflags; 1156 unsigned short padded; /* How much padding added by alloc_netdev() */ 1157 1158 unsigned char operstate; /* RFC2863 operstate */ 1159 unsigned char link_mode; /* mapping policy to operstate */ 1160 1161 unsigned char if_port; /* Selectable AUI, TP,..*/ 1162 unsigned char dma; /* DMA channel */ 1163 1164 unsigned int mtu; /* interface MTU value */ 1165 unsigned short type; /* interface hardware type */ 1166 unsigned short hard_header_len; /* hardware hdr length */ 1167 1168 /* extra head- and tailroom the hardware may need, but not in all cases 1169 * can this be guaranteed, especially tailroom. Some cases also use 1170 * LL_MAX_HEADER instead to allocate the skb. 1171 */ 1172 unsigned short needed_headroom; 1173 unsigned short needed_tailroom; 1174 1175 /* Interface address info. */ 1176 unsigned char perm_addr[MAX_ADDR_LEN]; /* permanent hw address */ 1177 unsigned char addr_assign_type; /* hw address assignment type */ 1178 unsigned char addr_len; /* hardware address length */ 1179 unsigned char neigh_priv_len; 1180 unsigned short dev_id; /* Used to differentiate devices 1181 * that share the same link 1182 * layer address 1183 */ 1184 spinlock_t addr_list_lock; 1185 struct netdev_hw_addr_list uc; /* Unicast mac addresses */ 1186 struct netdev_hw_addr_list mc; /* Multicast mac addresses */ 1187 struct netdev_hw_addr_list dev_addrs; /* list of device 1188 * hw addresses 1189 */ 1190#ifdef CONFIG_SYSFS 1191 struct kset *queues_kset; 1192#endif 1193 1194 bool uc_promisc; 1195 unsigned int promiscuity; 1196 unsigned int allmulti; 1197 1198 1199 /* Protocol specific pointers */ 1200 1201#if IS_ENABLED(CONFIG_VLAN_8021Q) 1202 struct vlan_info __rcu *vlan_info; /* VLAN info */ 1203#endif 1204#if IS_ENABLED(CONFIG_NET_DSA) 1205 struct dsa_switch_tree *dsa_ptr; /* dsa specific data */ 1206#endif 1207 void *atalk_ptr; /* AppleTalk link */ 1208 struct in_device __rcu *ip_ptr; /* IPv4 specific data */ 1209 struct dn_dev __rcu *dn_ptr; /* DECnet specific data */ 1210 struct inet6_dev __rcu *ip6_ptr; /* IPv6 specific data */ 1211 void *ax25_ptr; /* AX.25 specific data */ 1212 struct wireless_dev *ieee80211_ptr; /* IEEE 802.11 specific data, 1213 assign before registering */ 1214 1215/* 1216 * Cache lines mostly used on receive path (including eth_type_trans()) 1217 */ 1218 unsigned long last_rx; /* Time of last Rx 1219 * This should not be set in 1220 * drivers, unless really needed, 1221 * because network stack (bonding) 1222 * use it if/when necessary, to 1223 * avoid dirtying this cache line. 1224 */ 1225 1226 /* Interface address info used in eth_type_trans() */ 1227 unsigned char *dev_addr; /* hw address, (before bcast 1228 because most packets are 1229 unicast) */ 1230 1231 1232#ifdef CONFIG_RPS 1233 struct netdev_rx_queue *_rx; 1234 1235 /* Number of RX queues allocated at register_netdev() time */ 1236 unsigned int num_rx_queues; 1237 1238 /* Number of RX queues currently active in device */ 1239 unsigned int real_num_rx_queues; 1240 1241#endif 1242 1243 rx_handler_func_t __rcu *rx_handler; 1244 void __rcu *rx_handler_data; 1245 1246 struct netdev_queue __rcu *ingress_queue; 1247 unsigned char broadcast[MAX_ADDR_LEN]; /* hw bcast add */ 1248 1249 1250/* 1251 * Cache lines mostly used on transmit path 1252 */ 1253 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1254 1255 /* Number of TX queues allocated at alloc_netdev_mq() time */ 1256 unsigned int num_tx_queues; 1257 1258 /* Number of TX queues currently active in device */ 1259 unsigned int real_num_tx_queues; 1260 1261 /* root qdisc from userspace point of view */ 1262 struct Qdisc *qdisc; 1263 1264 unsigned long tx_queue_len; /* Max frames per queue allowed */ 1265 spinlock_t tx_global_lock; 1266 1267#ifdef CONFIG_XPS 1268 struct xps_dev_maps __rcu *xps_maps; 1269#endif 1270#ifdef CONFIG_RFS_ACCEL 1271 /* CPU reverse-mapping for RX completion interrupts, indexed 1272 * by RX queue number. Assigned by driver. This must only be 1273 * set if the ndo_rx_flow_steer operation is defined. */ 1274 struct cpu_rmap *rx_cpu_rmap; 1275#endif 1276 1277 /* These may be needed for future network-power-down code. */ 1278 1279 /* 1280 * trans_start here is expensive for high speed devices on SMP, 1281 * please use netdev_queue->trans_start instead. 1282 */ 1283 unsigned long trans_start; /* Time (in jiffies) of last Tx */ 1284 1285 int watchdog_timeo; /* used by dev_watchdog() */ 1286 struct timer_list watchdog_timer; 1287 1288 /* Number of references to this device */ 1289 int __percpu *pcpu_refcnt; 1290 1291 /* delayed register/unregister */ 1292 struct list_head todo_list; 1293 /* device index hash chain */ 1294 struct hlist_node index_hlist; 1295 1296 struct list_head link_watch_list; 1297 1298 /* register/unregister state machine */ 1299 enum { NETREG_UNINITIALIZED=0, 1300 NETREG_REGISTERED, /* completed register_netdevice */ 1301 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1302 NETREG_UNREGISTERED, /* completed unregister todo */ 1303 NETREG_RELEASED, /* called free_netdev */ 1304 NETREG_DUMMY, /* dummy device for NAPI poll */ 1305 } reg_state:8; 1306 1307 bool dismantle; /* device is going do be freed */ 1308 1309 enum { 1310 RTNL_LINK_INITIALIZED, 1311 RTNL_LINK_INITIALIZING, 1312 } rtnl_link_state:16; 1313 1314 /* Called from unregister, can be used to call free_netdev */ 1315 void (*destructor)(struct net_device *dev); 1316 1317#ifdef CONFIG_NETPOLL 1318 struct netpoll_info __rcu *npinfo; 1319#endif 1320 1321#ifdef CONFIG_NET_NS 1322 /* Network namespace this network device is inside */ 1323 struct net *nd_net; 1324#endif 1325 1326 /* mid-layer private */ 1327 union { 1328 void *ml_priv; 1329 struct pcpu_lstats __percpu *lstats; /* loopback stats */ 1330 struct pcpu_tstats __percpu *tstats; /* tunnel stats */ 1331 struct pcpu_dstats __percpu *dstats; /* dummy stats */ 1332 struct pcpu_vstats __percpu *vstats; /* veth stats */ 1333 }; 1334 /* GARP */ 1335 struct garp_port __rcu *garp_port; 1336 /* MRP */ 1337 struct mrp_port __rcu *mrp_port; 1338 1339 /* class/net/name entry */ 1340 struct device dev; 1341 /* space for optional device, statistics, and wireless sysfs groups */ 1342 const struct attribute_group *sysfs_groups[4]; 1343 1344 /* rtnetlink link ops */ 1345 const struct rtnl_link_ops *rtnl_link_ops; 1346 1347 /* for setting kernel sock attribute on TCP connection setup */ 1348#define GSO_MAX_SIZE 65536 1349 unsigned int gso_max_size; 1350#define GSO_MAX_SEGS 65535 1351 u16 gso_max_segs; 1352 1353#ifdef CONFIG_DCB 1354 /* Data Center Bridging netlink ops */ 1355 const struct dcbnl_rtnl_ops *dcbnl_ops; 1356#endif 1357 u8 num_tc; 1358 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 1359 u8 prio_tc_map[TC_BITMASK + 1]; 1360 1361#if IS_ENABLED(CONFIG_FCOE) 1362 /* max exchange id for FCoE LRO by ddp */ 1363 unsigned int fcoe_ddp_xid; 1364#endif 1365#if IS_ENABLED(CONFIG_NETPRIO_CGROUP) 1366 struct netprio_map __rcu *priomap; 1367#endif 1368 /* phy device may attach itself for hardware timestamping */ 1369 struct phy_device *phydev; 1370 1371 struct lock_class_key *qdisc_tx_busylock; 1372 1373 /* group the device belongs to */ 1374 int group; 1375 1376 struct pm_qos_request pm_qos_req; 1377}; 1378#define to_net_dev(d) container_of(d, struct net_device, dev) 1379 1380#define NETDEV_ALIGN 32 1381 1382static inline 1383int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 1384{ 1385 return dev->prio_tc_map[prio & TC_BITMASK]; 1386} 1387 1388static inline 1389int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 1390{ 1391 if (tc >= dev->num_tc) 1392 return -EINVAL; 1393 1394 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 1395 return 0; 1396} 1397 1398static inline 1399void netdev_reset_tc(struct net_device *dev) 1400{ 1401 dev->num_tc = 0; 1402 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq)); 1403 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map)); 1404} 1405 1406static inline 1407int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset) 1408{ 1409 if (tc >= dev->num_tc) 1410 return -EINVAL; 1411 1412 dev->tc_to_txq[tc].count = count; 1413 dev->tc_to_txq[tc].offset = offset; 1414 return 0; 1415} 1416 1417static inline 1418int netdev_set_num_tc(struct net_device *dev, u8 num_tc) 1419{ 1420 if (num_tc > TC_MAX_QUEUE) 1421 return -EINVAL; 1422 1423 dev->num_tc = num_tc; 1424 return 0; 1425} 1426 1427static inline 1428int netdev_get_num_tc(struct net_device *dev) 1429{ 1430 return dev->num_tc; 1431} 1432 1433static inline 1434struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 1435 unsigned int index) 1436{ 1437 return &dev->_tx[index]; 1438} 1439 1440static inline void netdev_for_each_tx_queue(struct net_device *dev, 1441 void (*f)(struct net_device *, 1442 struct netdev_queue *, 1443 void *), 1444 void *arg) 1445{ 1446 unsigned int i; 1447 1448 for (i = 0; i < dev->num_tx_queues; i++) 1449 f(dev, &dev->_tx[i], arg); 1450} 1451 1452extern struct netdev_queue *netdev_pick_tx(struct net_device *dev, 1453 struct sk_buff *skb); 1454extern u16 __netdev_pick_tx(struct net_device *dev, struct sk_buff *skb); 1455 1456/* 1457 * Net namespace inlines 1458 */ 1459static inline 1460struct net *dev_net(const struct net_device *dev) 1461{ 1462 return read_pnet(&dev->nd_net); 1463} 1464 1465static inline 1466void dev_net_set(struct net_device *dev, struct net *net) 1467{ 1468#ifdef CONFIG_NET_NS 1469 release_net(dev->nd_net); 1470 dev->nd_net = hold_net(net); 1471#endif 1472} 1473 1474static inline bool netdev_uses_dsa_tags(struct net_device *dev) 1475{ 1476#ifdef CONFIG_NET_DSA_TAG_DSA 1477 if (dev->dsa_ptr != NULL) 1478 return dsa_uses_dsa_tags(dev->dsa_ptr); 1479#endif 1480 1481 return 0; 1482} 1483 1484static inline bool netdev_uses_trailer_tags(struct net_device *dev) 1485{ 1486#ifdef CONFIG_NET_DSA_TAG_TRAILER 1487 if (dev->dsa_ptr != NULL) 1488 return dsa_uses_trailer_tags(dev->dsa_ptr); 1489#endif 1490 1491 return 0; 1492} 1493 1494/** 1495 * netdev_priv - access network device private data 1496 * @dev: network device 1497 * 1498 * Get network device private data 1499 */ 1500static inline void *netdev_priv(const struct net_device *dev) 1501{ 1502 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 1503} 1504 1505/* Set the sysfs physical device reference for the network logical device 1506 * if set prior to registration will cause a symlink during initialization. 1507 */ 1508#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 1509 1510/* Set the sysfs device type for the network logical device to allow 1511 * fin grained indentification of different network device types. For 1512 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc. 1513 */ 1514#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 1515 1516/* Default NAPI poll() weight 1517 * Device drivers are strongly advised to not use bigger value 1518 */ 1519#define NAPI_POLL_WEIGHT 64 1520 1521/** 1522 * netif_napi_add - initialize a napi context 1523 * @dev: network device 1524 * @napi: napi context 1525 * @poll: polling function 1526 * @weight: default weight 1527 * 1528 * netif_napi_add() must be used to initialize a napi context prior to calling 1529 * *any* of the other napi related functions. 1530 */ 1531void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 1532 int (*poll)(struct napi_struct *, int), int weight); 1533 1534/** 1535 * netif_napi_del - remove a napi context 1536 * @napi: napi context 1537 * 1538 * netif_napi_del() removes a napi context from the network device napi list 1539 */ 1540void netif_napi_del(struct napi_struct *napi); 1541 1542struct napi_gro_cb { 1543 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 1544 void *frag0; 1545 1546 /* Length of frag0. */ 1547 unsigned int frag0_len; 1548 1549 /* This indicates where we are processing relative to skb->data. */ 1550 int data_offset; 1551 1552 /* This is non-zero if the packet cannot be merged with the new skb. */ 1553 int flush; 1554 1555 /* Number of segments aggregated. */ 1556 u16 count; 1557 1558 /* This is non-zero if the packet may be of the same flow. */ 1559 u8 same_flow; 1560 1561 /* Free the skb? */ 1562 u8 free; 1563#define NAPI_GRO_FREE 1 1564#define NAPI_GRO_FREE_STOLEN_HEAD 2 1565 1566 /* jiffies when first packet was created/queued */ 1567 unsigned long age; 1568 1569 /* Used in ipv6_gro_receive() */ 1570 int proto; 1571 1572 /* used in skb_gro_receive() slow path */ 1573 struct sk_buff *last; 1574}; 1575 1576#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 1577 1578struct packet_type { 1579 __be16 type; /* This is really htons(ether_type). */ 1580 struct net_device *dev; /* NULL is wildcarded here */ 1581 int (*func) (struct sk_buff *, 1582 struct net_device *, 1583 struct packet_type *, 1584 struct net_device *); 1585 bool (*id_match)(struct packet_type *ptype, 1586 struct sock *sk); 1587 void *af_packet_priv; 1588 struct list_head list; 1589}; 1590 1591struct offload_callbacks { 1592 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 1593 netdev_features_t features); 1594 int (*gso_send_check)(struct sk_buff *skb); 1595 struct sk_buff **(*gro_receive)(struct sk_buff **head, 1596 struct sk_buff *skb); 1597 int (*gro_complete)(struct sk_buff *skb); 1598}; 1599 1600struct packet_offload { 1601 __be16 type; /* This is really htons(ether_type). */ 1602 struct offload_callbacks callbacks; 1603 struct list_head list; 1604}; 1605 1606#include <linux/notifier.h> 1607 1608/* netdevice notifier chain. Please remember to update the rtnetlink 1609 * notification exclusion list in rtnetlink_event() when adding new 1610 * types. 1611 */ 1612#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */ 1613#define NETDEV_DOWN 0x0002 1614#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface 1615 detected a hardware crash and restarted 1616 - we can use this eg to kick tcp sessions 1617 once done */ 1618#define NETDEV_CHANGE 0x0004 /* Notify device state change */ 1619#define NETDEV_REGISTER 0x0005 1620#define NETDEV_UNREGISTER 0x0006 1621#define NETDEV_CHANGEMTU 0x0007 1622#define NETDEV_CHANGEADDR 0x0008 1623#define NETDEV_GOING_DOWN 0x0009 1624#define NETDEV_CHANGENAME 0x000A 1625#define NETDEV_FEAT_CHANGE 0x000B 1626#define NETDEV_BONDING_FAILOVER 0x000C 1627#define NETDEV_PRE_UP 0x000D 1628#define NETDEV_PRE_TYPE_CHANGE 0x000E 1629#define NETDEV_POST_TYPE_CHANGE 0x000F 1630#define NETDEV_POST_INIT 0x0010 1631#define NETDEV_UNREGISTER_FINAL 0x0011 1632#define NETDEV_RELEASE 0x0012 1633#define NETDEV_NOTIFY_PEERS 0x0013 1634#define NETDEV_JOIN 0x0014 1635#define NETDEV_CHANGEUPPER 0x0015 1636 1637extern int register_netdevice_notifier(struct notifier_block *nb); 1638extern int unregister_netdevice_notifier(struct notifier_block *nb); 1639 1640struct netdev_notifier_info { 1641 struct net_device *dev; 1642}; 1643 1644struct netdev_notifier_change_info { 1645 struct netdev_notifier_info info; /* must be first */ 1646 unsigned int flags_changed; 1647}; 1648 1649static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 1650 struct net_device *dev) 1651{ 1652 info->dev = dev; 1653} 1654 1655static inline struct net_device * 1656netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 1657{ 1658 return info->dev; 1659} 1660 1661extern int call_netdevice_notifiers_info(unsigned long val, struct net_device *dev, 1662 struct netdev_notifier_info *info); 1663extern int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 1664 1665 1666extern rwlock_t dev_base_lock; /* Device list lock */ 1667 1668extern seqcount_t devnet_rename_seq; /* Device rename seq */ 1669 1670 1671#define for_each_netdev(net, d) \ 1672 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 1673#define for_each_netdev_reverse(net, d) \ 1674 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 1675#define for_each_netdev_rcu(net, d) \ 1676 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 1677#define for_each_netdev_safe(net, d, n) \ 1678 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 1679#define for_each_netdev_continue(net, d) \ 1680 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 1681#define for_each_netdev_continue_rcu(net, d) \ 1682 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 1683#define for_each_netdev_in_bond_rcu(bond, slave) \ 1684 for_each_netdev_rcu(&init_net, slave) \ 1685 if (netdev_master_upper_dev_get_rcu(slave) == bond) 1686#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 1687 1688static inline struct net_device *next_net_device(struct net_device *dev) 1689{ 1690 struct list_head *lh; 1691 struct net *net; 1692 1693 net = dev_net(dev); 1694 lh = dev->dev_list.next; 1695 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 1696} 1697 1698static inline struct net_device *next_net_device_rcu(struct net_device *dev) 1699{ 1700 struct list_head *lh; 1701 struct net *net; 1702 1703 net = dev_net(dev); 1704 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 1705 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 1706} 1707 1708static inline struct net_device *first_net_device(struct net *net) 1709{ 1710 return list_empty(&net->dev_base_head) ? NULL : 1711 net_device_entry(net->dev_base_head.next); 1712} 1713 1714static inline struct net_device *first_net_device_rcu(struct net *net) 1715{ 1716 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 1717 1718 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 1719} 1720 1721extern int netdev_boot_setup_check(struct net_device *dev); 1722extern unsigned long netdev_boot_base(const char *prefix, int unit); 1723extern struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 1724 const char *hwaddr); 1725extern struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 1726extern struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 1727extern void dev_add_pack(struct packet_type *pt); 1728extern void dev_remove_pack(struct packet_type *pt); 1729extern void __dev_remove_pack(struct packet_type *pt); 1730extern void dev_add_offload(struct packet_offload *po); 1731extern void dev_remove_offload(struct packet_offload *po); 1732extern void __dev_remove_offload(struct packet_offload *po); 1733 1734extern struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short flags, 1735 unsigned short mask); 1736extern struct net_device *dev_get_by_name(struct net *net, const char *name); 1737extern struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 1738extern struct net_device *__dev_get_by_name(struct net *net, const char *name); 1739extern int dev_alloc_name(struct net_device *dev, const char *name); 1740extern int dev_open(struct net_device *dev); 1741extern int dev_close(struct net_device *dev); 1742extern void dev_disable_lro(struct net_device *dev); 1743extern int dev_loopback_xmit(struct sk_buff *newskb); 1744extern int dev_queue_xmit(struct sk_buff *skb); 1745extern int register_netdevice(struct net_device *dev); 1746extern void unregister_netdevice_queue(struct net_device *dev, 1747 struct list_head *head); 1748extern void unregister_netdevice_many(struct list_head *head); 1749static inline void unregister_netdevice(struct net_device *dev) 1750{ 1751 unregister_netdevice_queue(dev, NULL); 1752} 1753 1754extern int netdev_refcnt_read(const struct net_device *dev); 1755extern void free_netdev(struct net_device *dev); 1756extern void synchronize_net(void); 1757extern int init_dummy_netdev(struct net_device *dev); 1758 1759extern struct net_device *dev_get_by_index(struct net *net, int ifindex); 1760extern struct net_device *__dev_get_by_index(struct net *net, int ifindex); 1761extern struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 1762extern int netdev_get_name(struct net *net, char *name, int ifindex); 1763extern int dev_restart(struct net_device *dev); 1764#ifdef CONFIG_NETPOLL_TRAP 1765extern int netpoll_trap(void); 1766#endif 1767extern int skb_gro_receive(struct sk_buff **head, 1768 struct sk_buff *skb); 1769 1770static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 1771{ 1772 return NAPI_GRO_CB(skb)->data_offset; 1773} 1774 1775static inline unsigned int skb_gro_len(const struct sk_buff *skb) 1776{ 1777 return skb->len - NAPI_GRO_CB(skb)->data_offset; 1778} 1779 1780static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 1781{ 1782 NAPI_GRO_CB(skb)->data_offset += len; 1783} 1784 1785static inline void *skb_gro_header_fast(struct sk_buff *skb, 1786 unsigned int offset) 1787{ 1788 return NAPI_GRO_CB(skb)->frag0 + offset; 1789} 1790 1791static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 1792{ 1793 return NAPI_GRO_CB(skb)->frag0_len < hlen; 1794} 1795 1796static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 1797 unsigned int offset) 1798{ 1799 if (!pskb_may_pull(skb, hlen)) 1800 return NULL; 1801 1802 NAPI_GRO_CB(skb)->frag0 = NULL; 1803 NAPI_GRO_CB(skb)->frag0_len = 0; 1804 return skb->data + offset; 1805} 1806 1807static inline void *skb_gro_mac_header(struct sk_buff *skb) 1808{ 1809 return NAPI_GRO_CB(skb)->frag0 ?: skb_mac_header(skb); 1810} 1811 1812static inline void *skb_gro_network_header(struct sk_buff *skb) 1813{ 1814 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 1815 skb_network_offset(skb); 1816} 1817 1818static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 1819 unsigned short type, 1820 const void *daddr, const void *saddr, 1821 unsigned int len) 1822{ 1823 if (!dev->header_ops || !dev->header_ops->create) 1824 return 0; 1825 1826 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 1827} 1828 1829static inline int dev_parse_header(const struct sk_buff *skb, 1830 unsigned char *haddr) 1831{ 1832 const struct net_device *dev = skb->dev; 1833 1834 if (!dev->header_ops || !dev->header_ops->parse) 1835 return 0; 1836 return dev->header_ops->parse(skb, haddr); 1837} 1838 1839typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len); 1840extern int register_gifconf(unsigned int family, gifconf_func_t * gifconf); 1841static inline int unregister_gifconf(unsigned int family) 1842{ 1843 return register_gifconf(family, NULL); 1844} 1845 1846#ifdef CONFIG_NET_FLOW_LIMIT 1847#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 1848struct sd_flow_limit { 1849 u64 count; 1850 unsigned int num_buckets; 1851 unsigned int history_head; 1852 u16 history[FLOW_LIMIT_HISTORY]; 1853 u8 buckets[]; 1854}; 1855 1856extern int netdev_flow_limit_table_len; 1857#endif /* CONFIG_NET_FLOW_LIMIT */ 1858 1859/* 1860 * Incoming packets are placed on per-cpu queues 1861 */ 1862struct softnet_data { 1863 struct Qdisc *output_queue; 1864 struct Qdisc **output_queue_tailp; 1865 struct list_head poll_list; 1866 struct sk_buff *completion_queue; 1867 struct sk_buff_head process_queue; 1868 1869 /* stats */ 1870 unsigned int processed; 1871 unsigned int time_squeeze; 1872 unsigned int cpu_collision; 1873 unsigned int received_rps; 1874 1875#ifdef CONFIG_RPS 1876 struct softnet_data *rps_ipi_list; 1877 1878 /* Elements below can be accessed between CPUs for RPS */ 1879 struct call_single_data csd ____cacheline_aligned_in_smp; 1880 struct softnet_data *rps_ipi_next; 1881 unsigned int cpu; 1882 unsigned int input_queue_head; 1883 unsigned int input_queue_tail; 1884#endif 1885 unsigned int dropped; 1886 struct sk_buff_head input_pkt_queue; 1887 struct napi_struct backlog; 1888 1889#ifdef CONFIG_NET_FLOW_LIMIT 1890 struct sd_flow_limit __rcu *flow_limit; 1891#endif 1892}; 1893 1894static inline void input_queue_head_incr(struct softnet_data *sd) 1895{ 1896#ifdef CONFIG_RPS 1897 sd->input_queue_head++; 1898#endif 1899} 1900 1901static inline void input_queue_tail_incr_save(struct softnet_data *sd, 1902 unsigned int *qtail) 1903{ 1904#ifdef CONFIG_RPS 1905 *qtail = ++sd->input_queue_tail; 1906#endif 1907} 1908 1909DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 1910 1911extern void __netif_schedule(struct Qdisc *q); 1912 1913static inline void netif_schedule_queue(struct netdev_queue *txq) 1914{ 1915 if (!(txq->state & QUEUE_STATE_ANY_XOFF)) 1916 __netif_schedule(txq->qdisc); 1917} 1918 1919static inline void netif_tx_schedule_all(struct net_device *dev) 1920{ 1921 unsigned int i; 1922 1923 for (i = 0; i < dev->num_tx_queues; i++) 1924 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 1925} 1926 1927static inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 1928{ 1929 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 1930} 1931 1932/** 1933 * netif_start_queue - allow transmit 1934 * @dev: network device 1935 * 1936 * Allow upper layers to call the device hard_start_xmit routine. 1937 */ 1938static inline void netif_start_queue(struct net_device *dev) 1939{ 1940 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 1941} 1942 1943static inline void netif_tx_start_all_queues(struct net_device *dev) 1944{ 1945 unsigned int i; 1946 1947 for (i = 0; i < dev->num_tx_queues; i++) { 1948 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 1949 netif_tx_start_queue(txq); 1950 } 1951} 1952 1953static inline void netif_tx_wake_queue(struct netdev_queue *dev_queue) 1954{ 1955#ifdef CONFIG_NETPOLL_TRAP 1956 if (netpoll_trap()) { 1957 netif_tx_start_queue(dev_queue); 1958 return; 1959 } 1960#endif 1961 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state)) 1962 __netif_schedule(dev_queue->qdisc); 1963} 1964 1965/** 1966 * netif_wake_queue - restart transmit 1967 * @dev: network device 1968 * 1969 * Allow upper layers to call the device hard_start_xmit routine. 1970 * Used for flow control when transmit resources are available. 1971 */ 1972static inline void netif_wake_queue(struct net_device *dev) 1973{ 1974 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 1975} 1976 1977static inline void netif_tx_wake_all_queues(struct net_device *dev) 1978{ 1979 unsigned int i; 1980 1981 for (i = 0; i < dev->num_tx_queues; i++) { 1982 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 1983 netif_tx_wake_queue(txq); 1984 } 1985} 1986 1987static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 1988{ 1989 if (WARN_ON(!dev_queue)) { 1990 pr_info("netif_stop_queue() cannot be called before register_netdev()\n"); 1991 return; 1992 } 1993 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 1994} 1995 1996/** 1997 * netif_stop_queue - stop transmitted packets 1998 * @dev: network device 1999 * 2000 * Stop upper layers calling the device hard_start_xmit routine. 2001 * Used for flow control when transmit resources are unavailable. 2002 */ 2003static inline void netif_stop_queue(struct net_device *dev) 2004{ 2005 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 2006} 2007 2008static inline void netif_tx_stop_all_queues(struct net_device *dev) 2009{ 2010 unsigned int i; 2011 2012 for (i = 0; i < dev->num_tx_queues; i++) { 2013 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2014 netif_tx_stop_queue(txq); 2015 } 2016} 2017 2018static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 2019{ 2020 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2021} 2022 2023/** 2024 * netif_queue_stopped - test if transmit queue is flowblocked 2025 * @dev: network device 2026 * 2027 * Test if transmit queue on device is currently unable to send. 2028 */ 2029static inline bool netif_queue_stopped(const struct net_device *dev) 2030{ 2031 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 2032} 2033 2034static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 2035{ 2036 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 2037} 2038 2039static inline bool netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 2040{ 2041 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 2042} 2043 2044static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 2045 unsigned int bytes) 2046{ 2047#ifdef CONFIG_BQL 2048 dql_queued(&dev_queue->dql, bytes); 2049 2050 if (likely(dql_avail(&dev_queue->dql) >= 0)) 2051 return; 2052 2053 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2054 2055 /* 2056 * The XOFF flag must be set before checking the dql_avail below, 2057 * because in netdev_tx_completed_queue we update the dql_completed 2058 * before checking the XOFF flag. 2059 */ 2060 smp_mb(); 2061 2062 /* check again in case another CPU has just made room avail */ 2063 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 2064 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2065#endif 2066} 2067 2068static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 2069{ 2070 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 2071} 2072 2073static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 2074 unsigned int pkts, unsigned int bytes) 2075{ 2076#ifdef CONFIG_BQL 2077 if (unlikely(!bytes)) 2078 return; 2079 2080 dql_completed(&dev_queue->dql, bytes); 2081 2082 /* 2083 * Without the memory barrier there is a small possiblity that 2084 * netdev_tx_sent_queue will miss the update and cause the queue to 2085 * be stopped forever 2086 */ 2087 smp_mb(); 2088 2089 if (dql_avail(&dev_queue->dql) < 0) 2090 return; 2091 2092 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 2093 netif_schedule_queue(dev_queue); 2094#endif 2095} 2096 2097static inline void netdev_completed_queue(struct net_device *dev, 2098 unsigned int pkts, unsigned int bytes) 2099{ 2100 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 2101} 2102 2103static inline void netdev_tx_reset_queue(struct netdev_queue *q) 2104{ 2105#ifdef CONFIG_BQL 2106 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 2107 dql_reset(&q->dql); 2108#endif 2109} 2110 2111static inline void netdev_reset_queue(struct net_device *dev_queue) 2112{ 2113 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 2114} 2115 2116/** 2117 * netif_running - test if up 2118 * @dev: network device 2119 * 2120 * Test if the device has been brought up. 2121 */ 2122static inline bool netif_running(const struct net_device *dev) 2123{ 2124 return test_bit(__LINK_STATE_START, &dev->state); 2125} 2126 2127/* 2128 * Routines to manage the subqueues on a device. We only need start 2129 * stop, and a check if it's stopped. All other device management is 2130 * done at the overall netdevice level. 2131 * Also test the device if we're multiqueue. 2132 */ 2133 2134/** 2135 * netif_start_subqueue - allow sending packets on subqueue 2136 * @dev: network device 2137 * @queue_index: sub queue index 2138 * 2139 * Start individual transmit queue of a device with multiple transmit queues. 2140 */ 2141static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 2142{ 2143 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2144 2145 netif_tx_start_queue(txq); 2146} 2147 2148/** 2149 * netif_stop_subqueue - stop sending packets on subqueue 2150 * @dev: network device 2151 * @queue_index: sub queue index 2152 * 2153 * Stop individual transmit queue of a device with multiple transmit queues. 2154 */ 2155static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 2156{ 2157 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2158#ifdef CONFIG_NETPOLL_TRAP 2159 if (netpoll_trap()) 2160 return; 2161#endif 2162 netif_tx_stop_queue(txq); 2163} 2164 2165/** 2166 * netif_subqueue_stopped - test status of subqueue 2167 * @dev: network device 2168 * @queue_index: sub queue index 2169 * 2170 * Check individual transmit queue of a device with multiple transmit queues. 2171 */ 2172static inline bool __netif_subqueue_stopped(const struct net_device *dev, 2173 u16 queue_index) 2174{ 2175 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2176 2177 return netif_tx_queue_stopped(txq); 2178} 2179 2180static inline bool netif_subqueue_stopped(const struct net_device *dev, 2181 struct sk_buff *skb) 2182{ 2183 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 2184} 2185 2186/** 2187 * netif_wake_subqueue - allow sending packets on subqueue 2188 * @dev: network device 2189 * @queue_index: sub queue index 2190 * 2191 * Resume individual transmit queue of a device with multiple transmit queues. 2192 */ 2193static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 2194{ 2195 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2196#ifdef CONFIG_NETPOLL_TRAP 2197 if (netpoll_trap()) 2198 return; 2199#endif 2200 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state)) 2201 __netif_schedule(txq->qdisc); 2202} 2203 2204#ifdef CONFIG_XPS 2205extern int netif_set_xps_queue(struct net_device *dev, struct cpumask *mask, 2206 u16 index); 2207#else 2208static inline int netif_set_xps_queue(struct net_device *dev, 2209 struct cpumask *mask, 2210 u16 index) 2211{ 2212 return 0; 2213} 2214#endif 2215 2216/* 2217 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used 2218 * as a distribution range limit for the returned value. 2219 */ 2220static inline u16 skb_tx_hash(const struct net_device *dev, 2221 const struct sk_buff *skb) 2222{ 2223 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues); 2224} 2225 2226/** 2227 * netif_is_multiqueue - test if device has multiple transmit queues 2228 * @dev: network device 2229 * 2230 * Check if device has multiple transmit queues 2231 */ 2232static inline bool netif_is_multiqueue(const struct net_device *dev) 2233{ 2234 return dev->num_tx_queues > 1; 2235} 2236 2237extern int netif_set_real_num_tx_queues(struct net_device *dev, 2238 unsigned int txq); 2239 2240#ifdef CONFIG_RPS 2241extern int netif_set_real_num_rx_queues(struct net_device *dev, 2242 unsigned int rxq); 2243#else 2244static inline int netif_set_real_num_rx_queues(struct net_device *dev, 2245 unsigned int rxq) 2246{ 2247 return 0; 2248} 2249#endif 2250 2251static inline int netif_copy_real_num_queues(struct net_device *to_dev, 2252 const struct net_device *from_dev) 2253{ 2254 int err; 2255 2256 err = netif_set_real_num_tx_queues(to_dev, 2257 from_dev->real_num_tx_queues); 2258 if (err) 2259 return err; 2260#ifdef CONFIG_RPS 2261 return netif_set_real_num_rx_queues(to_dev, 2262 from_dev->real_num_rx_queues); 2263#else 2264 return 0; 2265#endif 2266} 2267 2268#define DEFAULT_MAX_NUM_RSS_QUEUES (8) 2269extern int netif_get_num_default_rss_queues(void); 2270 2271/* Use this variant when it is known for sure that it 2272 * is executing from hardware interrupt context or with hardware interrupts 2273 * disabled. 2274 */ 2275extern void dev_kfree_skb_irq(struct sk_buff *skb); 2276 2277/* Use this variant in places where it could be invoked 2278 * from either hardware interrupt or other context, with hardware interrupts 2279 * either disabled or enabled. 2280 */ 2281extern void dev_kfree_skb_any(struct sk_buff *skb); 2282 2283extern int netif_rx(struct sk_buff *skb); 2284extern int netif_rx_ni(struct sk_buff *skb); 2285extern int netif_receive_skb(struct sk_buff *skb); 2286extern gro_result_t napi_gro_receive(struct napi_struct *napi, 2287 struct sk_buff *skb); 2288extern void napi_gro_flush(struct napi_struct *napi, bool flush_old); 2289extern struct sk_buff * napi_get_frags(struct napi_struct *napi); 2290extern gro_result_t napi_gro_frags(struct napi_struct *napi); 2291 2292static inline void napi_free_frags(struct napi_struct *napi) 2293{ 2294 kfree_skb(napi->skb); 2295 napi->skb = NULL; 2296} 2297 2298extern int netdev_rx_handler_register(struct net_device *dev, 2299 rx_handler_func_t *rx_handler, 2300 void *rx_handler_data); 2301extern void netdev_rx_handler_unregister(struct net_device *dev); 2302 2303extern bool dev_valid_name(const char *name); 2304extern int dev_ioctl(struct net *net, unsigned int cmd, void __user *); 2305extern int dev_ethtool(struct net *net, struct ifreq *); 2306extern unsigned int dev_get_flags(const struct net_device *); 2307extern int __dev_change_flags(struct net_device *, unsigned int flags); 2308extern int dev_change_flags(struct net_device *, unsigned int); 2309extern void __dev_notify_flags(struct net_device *, unsigned int old_flags); 2310extern int dev_change_name(struct net_device *, const char *); 2311extern int dev_set_alias(struct net_device *, const char *, size_t); 2312extern int dev_change_net_namespace(struct net_device *, 2313 struct net *, const char *); 2314extern int dev_set_mtu(struct net_device *, int); 2315extern void dev_set_group(struct net_device *, int); 2316extern int dev_set_mac_address(struct net_device *, 2317 struct sockaddr *); 2318extern int dev_change_carrier(struct net_device *, 2319 bool new_carrier); 2320extern int dev_hard_start_xmit(struct sk_buff *skb, 2321 struct net_device *dev, 2322 struct netdev_queue *txq); 2323extern int dev_forward_skb(struct net_device *dev, 2324 struct sk_buff *skb); 2325 2326extern int netdev_budget; 2327 2328/* Called by rtnetlink.c:rtnl_unlock() */ 2329extern void netdev_run_todo(void); 2330 2331/** 2332 * dev_put - release reference to device 2333 * @dev: network device 2334 * 2335 * Release reference to device to allow it to be freed. 2336 */ 2337static inline void dev_put(struct net_device *dev) 2338{ 2339 this_cpu_dec(*dev->pcpu_refcnt); 2340} 2341 2342/** 2343 * dev_hold - get reference to device 2344 * @dev: network device 2345 * 2346 * Hold reference to device to keep it from being freed. 2347 */ 2348static inline void dev_hold(struct net_device *dev) 2349{ 2350 this_cpu_inc(*dev->pcpu_refcnt); 2351} 2352 2353/* Carrier loss detection, dial on demand. The functions netif_carrier_on 2354 * and _off may be called from IRQ context, but it is caller 2355 * who is responsible for serialization of these calls. 2356 * 2357 * The name carrier is inappropriate, these functions should really be 2358 * called netif_lowerlayer_*() because they represent the state of any 2359 * kind of lower layer not just hardware media. 2360 */ 2361 2362extern void linkwatch_init_dev(struct net_device *dev); 2363extern void linkwatch_fire_event(struct net_device *dev); 2364extern void linkwatch_forget_dev(struct net_device *dev); 2365 2366/** 2367 * netif_carrier_ok - test if carrier present 2368 * @dev: network device 2369 * 2370 * Check if carrier is present on device 2371 */ 2372static inline bool netif_carrier_ok(const struct net_device *dev) 2373{ 2374 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 2375} 2376 2377extern unsigned long dev_trans_start(struct net_device *dev); 2378 2379extern void __netdev_watchdog_up(struct net_device *dev); 2380 2381extern void netif_carrier_on(struct net_device *dev); 2382 2383extern void netif_carrier_off(struct net_device *dev); 2384 2385/** 2386 * netif_dormant_on - mark device as dormant. 2387 * @dev: network device 2388 * 2389 * Mark device as dormant (as per RFC2863). 2390 * 2391 * The dormant state indicates that the relevant interface is not 2392 * actually in a condition to pass packets (i.e., it is not 'up') but is 2393 * in a "pending" state, waiting for some external event. For "on- 2394 * demand" interfaces, this new state identifies the situation where the 2395 * interface is waiting for events to place it in the up state. 2396 * 2397 */ 2398static inline void netif_dormant_on(struct net_device *dev) 2399{ 2400 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 2401 linkwatch_fire_event(dev); 2402} 2403 2404/** 2405 * netif_dormant_off - set device as not dormant. 2406 * @dev: network device 2407 * 2408 * Device is not in dormant state. 2409 */ 2410static inline void netif_dormant_off(struct net_device *dev) 2411{ 2412 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 2413 linkwatch_fire_event(dev); 2414} 2415 2416/** 2417 * netif_dormant - test if carrier present 2418 * @dev: network device 2419 * 2420 * Check if carrier is present on device 2421 */ 2422static inline bool netif_dormant(const struct net_device *dev) 2423{ 2424 return test_bit(__LINK_STATE_DORMANT, &dev->state); 2425} 2426 2427 2428/** 2429 * netif_oper_up - test if device is operational 2430 * @dev: network device 2431 * 2432 * Check if carrier is operational 2433 */ 2434static inline bool netif_oper_up(const struct net_device *dev) 2435{ 2436 return (dev->operstate == IF_OPER_UP || 2437 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 2438} 2439 2440/** 2441 * netif_device_present - is device available or removed 2442 * @dev: network device 2443 * 2444 * Check if device has not been removed from system. 2445 */ 2446static inline bool netif_device_present(struct net_device *dev) 2447{ 2448 return test_bit(__LINK_STATE_PRESENT, &dev->state); 2449} 2450 2451extern void netif_device_detach(struct net_device *dev); 2452 2453extern void netif_device_attach(struct net_device *dev); 2454 2455/* 2456 * Network interface message level settings 2457 */ 2458 2459enum { 2460 NETIF_MSG_DRV = 0x0001, 2461 NETIF_MSG_PROBE = 0x0002, 2462 NETIF_MSG_LINK = 0x0004, 2463 NETIF_MSG_TIMER = 0x0008, 2464 NETIF_MSG_IFDOWN = 0x0010, 2465 NETIF_MSG_IFUP = 0x0020, 2466 NETIF_MSG_RX_ERR = 0x0040, 2467 NETIF_MSG_TX_ERR = 0x0080, 2468 NETIF_MSG_TX_QUEUED = 0x0100, 2469 NETIF_MSG_INTR = 0x0200, 2470 NETIF_MSG_TX_DONE = 0x0400, 2471 NETIF_MSG_RX_STATUS = 0x0800, 2472 NETIF_MSG_PKTDATA = 0x1000, 2473 NETIF_MSG_HW = 0x2000, 2474 NETIF_MSG_WOL = 0x4000, 2475}; 2476 2477#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 2478#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 2479#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 2480#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 2481#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 2482#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 2483#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 2484#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 2485#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 2486#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 2487#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 2488#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 2489#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 2490#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 2491#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 2492 2493static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 2494{ 2495 /* use default */ 2496 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 2497 return default_msg_enable_bits; 2498 if (debug_value == 0) /* no output */ 2499 return 0; 2500 /* set low N bits */ 2501 return (1 << debug_value) - 1; 2502} 2503 2504static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 2505{ 2506 spin_lock(&txq->_xmit_lock); 2507 txq->xmit_lock_owner = cpu; 2508} 2509 2510static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 2511{ 2512 spin_lock_bh(&txq->_xmit_lock); 2513 txq->xmit_lock_owner = smp_processor_id(); 2514} 2515 2516static inline bool __netif_tx_trylock(struct netdev_queue *txq) 2517{ 2518 bool ok = spin_trylock(&txq->_xmit_lock); 2519 if (likely(ok)) 2520 txq->xmit_lock_owner = smp_processor_id(); 2521 return ok; 2522} 2523 2524static inline void __netif_tx_unlock(struct netdev_queue *txq) 2525{ 2526 txq->xmit_lock_owner = -1; 2527 spin_unlock(&txq->_xmit_lock); 2528} 2529 2530static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 2531{ 2532 txq->xmit_lock_owner = -1; 2533 spin_unlock_bh(&txq->_xmit_lock); 2534} 2535 2536static inline void txq_trans_update(struct netdev_queue *txq) 2537{ 2538 if (txq->xmit_lock_owner != -1) 2539 txq->trans_start = jiffies; 2540} 2541 2542/** 2543 * netif_tx_lock - grab network device transmit lock 2544 * @dev: network device 2545 * 2546 * Get network device transmit lock 2547 */ 2548static inline void netif_tx_lock(struct net_device *dev) 2549{ 2550 unsigned int i; 2551 int cpu; 2552 2553 spin_lock(&dev->tx_global_lock); 2554 cpu = smp_processor_id(); 2555 for (i = 0; i < dev->num_tx_queues; i++) { 2556 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2557 2558 /* We are the only thread of execution doing a 2559 * freeze, but we have to grab the _xmit_lock in 2560 * order to synchronize with threads which are in 2561 * the ->hard_start_xmit() handler and already 2562 * checked the frozen bit. 2563 */ 2564 __netif_tx_lock(txq, cpu); 2565 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 2566 __netif_tx_unlock(txq); 2567 } 2568} 2569 2570static inline void netif_tx_lock_bh(struct net_device *dev) 2571{ 2572 local_bh_disable(); 2573 netif_tx_lock(dev); 2574} 2575 2576static inline void netif_tx_unlock(struct net_device *dev) 2577{ 2578 unsigned int i; 2579 2580 for (i = 0; i < dev->num_tx_queues; i++) { 2581 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2582 2583 /* No need to grab the _xmit_lock here. If the 2584 * queue is not stopped for another reason, we 2585 * force a schedule. 2586 */ 2587 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 2588 netif_schedule_queue(txq); 2589 } 2590 spin_unlock(&dev->tx_global_lock); 2591} 2592 2593static inline void netif_tx_unlock_bh(struct net_device *dev) 2594{ 2595 netif_tx_unlock(dev); 2596 local_bh_enable(); 2597} 2598 2599#define HARD_TX_LOCK(dev, txq, cpu) { \ 2600 if ((dev->features & NETIF_F_LLTX) == 0) { \ 2601 __netif_tx_lock(txq, cpu); \ 2602 } \ 2603} 2604 2605#define HARD_TX_UNLOCK(dev, txq) { \ 2606 if ((dev->features & NETIF_F_LLTX) == 0) { \ 2607 __netif_tx_unlock(txq); \ 2608 } \ 2609} 2610 2611static inline void netif_tx_disable(struct net_device *dev) 2612{ 2613 unsigned int i; 2614 int cpu; 2615 2616 local_bh_disable(); 2617 cpu = smp_processor_id(); 2618 for (i = 0; i < dev->num_tx_queues; i++) { 2619 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2620 2621 __netif_tx_lock(txq, cpu); 2622 netif_tx_stop_queue(txq); 2623 __netif_tx_unlock(txq); 2624 } 2625 local_bh_enable(); 2626} 2627 2628static inline void netif_addr_lock(struct net_device *dev) 2629{ 2630 spin_lock(&dev->addr_list_lock); 2631} 2632 2633static inline void netif_addr_lock_nested(struct net_device *dev) 2634{ 2635 spin_lock_nested(&dev->addr_list_lock, SINGLE_DEPTH_NESTING); 2636} 2637 2638static inline void netif_addr_lock_bh(struct net_device *dev) 2639{ 2640 spin_lock_bh(&dev->addr_list_lock); 2641} 2642 2643static inline void netif_addr_unlock(struct net_device *dev) 2644{ 2645 spin_unlock(&dev->addr_list_lock); 2646} 2647 2648static inline void netif_addr_unlock_bh(struct net_device *dev) 2649{ 2650 spin_unlock_bh(&dev->addr_list_lock); 2651} 2652 2653/* 2654 * dev_addrs walker. Should be used only for read access. Call with 2655 * rcu_read_lock held. 2656 */ 2657#define for_each_dev_addr(dev, ha) \ 2658 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 2659 2660/* These functions live elsewhere (drivers/net/net_init.c, but related) */ 2661 2662extern void ether_setup(struct net_device *dev); 2663 2664/* Support for loadable net-drivers */ 2665extern struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 2666 void (*setup)(struct net_device *), 2667 unsigned int txqs, unsigned int rxqs); 2668#define alloc_netdev(sizeof_priv, name, setup) \ 2669 alloc_netdev_mqs(sizeof_priv, name, setup, 1, 1) 2670 2671#define alloc_netdev_mq(sizeof_priv, name, setup, count) \ 2672 alloc_netdev_mqs(sizeof_priv, name, setup, count, count) 2673 2674extern int register_netdev(struct net_device *dev); 2675extern void unregister_netdev(struct net_device *dev); 2676 2677/* General hardware address lists handling functions */ 2678extern int __hw_addr_add_multiple(struct netdev_hw_addr_list *to_list, 2679 struct netdev_hw_addr_list *from_list, 2680 int addr_len, unsigned char addr_type); 2681extern void __hw_addr_del_multiple(struct netdev_hw_addr_list *to_list, 2682 struct netdev_hw_addr_list *from_list, 2683 int addr_len, unsigned char addr_type); 2684extern int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 2685 struct netdev_hw_addr_list *from_list, 2686 int addr_len); 2687extern void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 2688 struct netdev_hw_addr_list *from_list, 2689 int addr_len); 2690extern void __hw_addr_flush(struct netdev_hw_addr_list *list); 2691extern void __hw_addr_init(struct netdev_hw_addr_list *list); 2692 2693/* Functions used for device addresses handling */ 2694extern int dev_addr_add(struct net_device *dev, const unsigned char *addr, 2695 unsigned char addr_type); 2696extern int dev_addr_del(struct net_device *dev, const unsigned char *addr, 2697 unsigned char addr_type); 2698extern int dev_addr_add_multiple(struct net_device *to_dev, 2699 struct net_device *from_dev, 2700 unsigned char addr_type); 2701extern int dev_addr_del_multiple(struct net_device *to_dev, 2702 struct net_device *from_dev, 2703 unsigned char addr_type); 2704extern void dev_addr_flush(struct net_device *dev); 2705extern int dev_addr_init(struct net_device *dev); 2706 2707/* Functions used for unicast addresses handling */ 2708extern int dev_uc_add(struct net_device *dev, const unsigned char *addr); 2709extern int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 2710extern int dev_uc_del(struct net_device *dev, const unsigned char *addr); 2711extern int dev_uc_sync(struct net_device *to, struct net_device *from); 2712extern int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 2713extern void dev_uc_unsync(struct net_device *to, struct net_device *from); 2714extern void dev_uc_flush(struct net_device *dev); 2715extern void dev_uc_init(struct net_device *dev); 2716 2717/* Functions used for multicast addresses handling */ 2718extern int dev_mc_add(struct net_device *dev, const unsigned char *addr); 2719extern int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 2720extern int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 2721extern int dev_mc_del(struct net_device *dev, const unsigned char *addr); 2722extern int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 2723extern int dev_mc_sync(struct net_device *to, struct net_device *from); 2724extern int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 2725extern void dev_mc_unsync(struct net_device *to, struct net_device *from); 2726extern void dev_mc_flush(struct net_device *dev); 2727extern void dev_mc_init(struct net_device *dev); 2728 2729/* Functions used for secondary unicast and multicast support */ 2730extern void dev_set_rx_mode(struct net_device *dev); 2731extern void __dev_set_rx_mode(struct net_device *dev); 2732extern int dev_set_promiscuity(struct net_device *dev, int inc); 2733extern int dev_set_allmulti(struct net_device *dev, int inc); 2734extern void netdev_state_change(struct net_device *dev); 2735extern void netdev_notify_peers(struct net_device *dev); 2736extern void netdev_features_change(struct net_device *dev); 2737/* Load a device via the kmod */ 2738extern void dev_load(struct net *net, const char *name); 2739extern struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 2740 struct rtnl_link_stats64 *storage); 2741extern void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 2742 const struct net_device_stats *netdev_stats); 2743 2744extern int netdev_max_backlog; 2745extern int netdev_tstamp_prequeue; 2746extern int weight_p; 2747extern int bpf_jit_enable; 2748 2749extern bool netdev_has_upper_dev(struct net_device *dev, 2750 struct net_device *upper_dev); 2751extern bool netdev_has_any_upper_dev(struct net_device *dev); 2752extern struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 2753extern struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 2754extern int netdev_upper_dev_link(struct net_device *dev, 2755 struct net_device *upper_dev); 2756extern int netdev_master_upper_dev_link(struct net_device *dev, 2757 struct net_device *upper_dev); 2758extern void netdev_upper_dev_unlink(struct net_device *dev, 2759 struct net_device *upper_dev); 2760extern int skb_checksum_help(struct sk_buff *skb); 2761extern struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 2762 netdev_features_t features, bool tx_path); 2763extern struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 2764 netdev_features_t features); 2765 2766static inline 2767struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 2768{ 2769 return __skb_gso_segment(skb, features, true); 2770} 2771__be16 skb_network_protocol(struct sk_buff *skb); 2772 2773static inline bool can_checksum_protocol(netdev_features_t features, 2774 __be16 protocol) 2775{ 2776 return ((features & NETIF_F_GEN_CSUM) || 2777 ((features & NETIF_F_V4_CSUM) && 2778 protocol == htons(ETH_P_IP)) || 2779 ((features & NETIF_F_V6_CSUM) && 2780 protocol == htons(ETH_P_IPV6)) || 2781 ((features & NETIF_F_FCOE_CRC) && 2782 protocol == htons(ETH_P_FCOE))); 2783} 2784 2785#ifdef CONFIG_BUG 2786extern void netdev_rx_csum_fault(struct net_device *dev); 2787#else 2788static inline void netdev_rx_csum_fault(struct net_device *dev) 2789{ 2790} 2791#endif 2792/* rx skb timestamps */ 2793extern void net_enable_timestamp(void); 2794extern void net_disable_timestamp(void); 2795 2796#ifdef CONFIG_PROC_FS 2797extern int __init dev_proc_init(void); 2798#else 2799#define dev_proc_init() 0 2800#endif 2801 2802extern int netdev_class_create_file(struct class_attribute *class_attr); 2803extern void netdev_class_remove_file(struct class_attribute *class_attr); 2804 2805extern struct kobj_ns_type_operations net_ns_type_operations; 2806 2807extern const char *netdev_drivername(const struct net_device *dev); 2808 2809extern void linkwatch_run_queue(void); 2810 2811static inline netdev_features_t netdev_get_wanted_features( 2812 struct net_device *dev) 2813{ 2814 return (dev->features & ~dev->hw_features) | dev->wanted_features; 2815} 2816netdev_features_t netdev_increment_features(netdev_features_t all, 2817 netdev_features_t one, netdev_features_t mask); 2818 2819/* Allow TSO being used on stacked device : 2820 * Performing the GSO segmentation before last device 2821 * is a performance improvement. 2822 */ 2823static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 2824 netdev_features_t mask) 2825{ 2826 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 2827} 2828 2829int __netdev_update_features(struct net_device *dev); 2830void netdev_update_features(struct net_device *dev); 2831void netdev_change_features(struct net_device *dev); 2832 2833void netif_stacked_transfer_operstate(const struct net_device *rootdev, 2834 struct net_device *dev); 2835 2836netdev_features_t netif_skb_features(struct sk_buff *skb); 2837 2838static inline bool net_gso_ok(netdev_features_t features, int gso_type) 2839{ 2840 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT; 2841 2842 /* check flags correspondence */ 2843 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 2844 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT)); 2845 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 2846 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 2847 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 2848 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 2849 2850 return (features & feature) == feature; 2851} 2852 2853static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 2854{ 2855 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 2856 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 2857} 2858 2859static inline bool netif_needs_gso(struct sk_buff *skb, 2860 netdev_features_t features) 2861{ 2862 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 2863 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 2864 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 2865} 2866 2867static inline void netif_set_gso_max_size(struct net_device *dev, 2868 unsigned int size) 2869{ 2870 dev->gso_max_size = size; 2871} 2872 2873static inline bool netif_is_bond_master(struct net_device *dev) 2874{ 2875 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 2876} 2877 2878static inline bool netif_is_bond_slave(struct net_device *dev) 2879{ 2880 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 2881} 2882 2883static inline bool netif_supports_nofcs(struct net_device *dev) 2884{ 2885 return dev->priv_flags & IFF_SUPP_NOFCS; 2886} 2887 2888extern struct pernet_operations __net_initdata loopback_net_ops; 2889 2890/* Logging, debugging and troubleshooting/diagnostic helpers. */ 2891 2892/* netdev_printk helpers, similar to dev_printk */ 2893 2894static inline const char *netdev_name(const struct net_device *dev) 2895{ 2896 if (dev->reg_state != NETREG_REGISTERED) 2897 return "(unregistered net_device)"; 2898 return dev->name; 2899} 2900 2901extern __printf(3, 4) 2902int netdev_printk(const char *level, const struct net_device *dev, 2903 const char *format, ...); 2904extern __printf(2, 3) 2905int netdev_emerg(const struct net_device *dev, const char *format, ...); 2906extern __printf(2, 3) 2907int netdev_alert(const struct net_device *dev, const char *format, ...); 2908extern __printf(2, 3) 2909int netdev_crit(const struct net_device *dev, const char *format, ...); 2910extern __printf(2, 3) 2911int netdev_err(const struct net_device *dev, const char *format, ...); 2912extern __printf(2, 3) 2913int netdev_warn(const struct net_device *dev, const char *format, ...); 2914extern __printf(2, 3) 2915int netdev_notice(const struct net_device *dev, const char *format, ...); 2916extern __printf(2, 3) 2917int netdev_info(const struct net_device *dev, const char *format, ...); 2918 2919#define MODULE_ALIAS_NETDEV(device) \ 2920 MODULE_ALIAS("netdev-" device) 2921 2922#if defined(CONFIG_DYNAMIC_DEBUG) 2923#define netdev_dbg(__dev, format, args...) \ 2924do { \ 2925 dynamic_netdev_dbg(__dev, format, ##args); \ 2926} while (0) 2927#elif defined(DEBUG) 2928#define netdev_dbg(__dev, format, args...) \ 2929 netdev_printk(KERN_DEBUG, __dev, format, ##args) 2930#else 2931#define netdev_dbg(__dev, format, args...) \ 2932({ \ 2933 if (0) \ 2934 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 2935 0; \ 2936}) 2937#endif 2938 2939#if defined(VERBOSE_DEBUG) 2940#define netdev_vdbg netdev_dbg 2941#else 2942 2943#define netdev_vdbg(dev, format, args...) \ 2944({ \ 2945 if (0) \ 2946 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 2947 0; \ 2948}) 2949#endif 2950 2951/* 2952 * netdev_WARN() acts like dev_printk(), but with the key difference 2953 * of using a WARN/WARN_ON to get the message out, including the 2954 * file/line information and a backtrace. 2955 */ 2956#define netdev_WARN(dev, format, args...) \ 2957 WARN(1, "netdevice: %s\n" format, netdev_name(dev), ##args); 2958 2959/* netif printk helpers, similar to netdev_printk */ 2960 2961#define netif_printk(priv, type, level, dev, fmt, args...) \ 2962do { \ 2963 if (netif_msg_##type(priv)) \ 2964 netdev_printk(level, (dev), fmt, ##args); \ 2965} while (0) 2966 2967#define netif_level(level, priv, type, dev, fmt, args...) \ 2968do { \ 2969 if (netif_msg_##type(priv)) \ 2970 netdev_##level(dev, fmt, ##args); \ 2971} while (0) 2972 2973#define netif_emerg(priv, type, dev, fmt, args...) \ 2974 netif_level(emerg, priv, type, dev, fmt, ##args) 2975#define netif_alert(priv, type, dev, fmt, args...) \ 2976 netif_level(alert, priv, type, dev, fmt, ##args) 2977#define netif_crit(priv, type, dev, fmt, args...) \ 2978 netif_level(crit, priv, type, dev, fmt, ##args) 2979#define netif_err(priv, type, dev, fmt, args...) \ 2980 netif_level(err, priv, type, dev, fmt, ##args) 2981#define netif_warn(priv, type, dev, fmt, args...) \ 2982 netif_level(warn, priv, type, dev, fmt, ##args) 2983#define netif_notice(priv, type, dev, fmt, args...) \ 2984 netif_level(notice, priv, type, dev, fmt, ##args) 2985#define netif_info(priv, type, dev, fmt, args...) \ 2986 netif_level(info, priv, type, dev, fmt, ##args) 2987 2988#if defined(CONFIG_DYNAMIC_DEBUG) 2989#define netif_dbg(priv, type, netdev, format, args...) \ 2990do { \ 2991 if (netif_msg_##type(priv)) \ 2992 dynamic_netdev_dbg(netdev, format, ##args); \ 2993} while (0) 2994#elif defined(DEBUG) 2995#define netif_dbg(priv, type, dev, format, args...) \ 2996 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 2997#else 2998#define netif_dbg(priv, type, dev, format, args...) \ 2999({ \ 3000 if (0) \ 3001 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 3002 0; \ 3003}) 3004#endif 3005 3006#if defined(VERBOSE_DEBUG) 3007#define netif_vdbg netif_dbg 3008#else 3009#define netif_vdbg(priv, type, dev, format, args...) \ 3010({ \ 3011 if (0) \ 3012 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 3013 0; \ 3014}) 3015#endif 3016 3017/* 3018 * The list of packet types we will receive (as opposed to discard) 3019 * and the routines to invoke. 3020 * 3021 * Why 16. Because with 16 the only overlap we get on a hash of the 3022 * low nibble of the protocol value is RARP/SNAP/X.25. 3023 * 3024 * NOTE: That is no longer true with the addition of VLAN tags. Not 3025 * sure which should go first, but I bet it won't make much 3026 * difference if we are running VLANs. The good news is that 3027 * this protocol won't be in the list unless compiled in, so 3028 * the average user (w/out VLANs) will not be adversely affected. 3029 * --BLG 3030 * 3031 * 0800 IP 3032 * 8100 802.1Q VLAN 3033 * 0001 802.3 3034 * 0002 AX.25 3035 * 0004 802.2 3036 * 8035 RARP 3037 * 0005 SNAP 3038 * 0805 X.25 3039 * 0806 ARP 3040 * 8137 IPX 3041 * 0009 Localtalk 3042 * 86DD IPv6 3043 */ 3044#define PTYPE_HASH_SIZE (16) 3045#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 3046 3047#endif /* _LINUX_NETDEVICE_H */