at v4.1-rc2 124 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 <linux/prefetch.h> 34#include <asm/cache.h> 35#include <asm/byteorder.h> 36 37#include <linux/percpu.h> 38#include <linux/rculist.h> 39#include <linux/dmaengine.h> 40#include <linux/workqueue.h> 41#include <linux/dynamic_queue_limits.h> 42 43#include <linux/ethtool.h> 44#include <net/net_namespace.h> 45#include <net/dsa.h> 46#ifdef CONFIG_DCB 47#include <net/dcbnl.h> 48#endif 49#include <net/netprio_cgroup.h> 50 51#include <linux/netdev_features.h> 52#include <linux/neighbour.h> 53#include <uapi/linux/netdevice.h> 54#include <uapi/linux/if_bonding.h> 55 56struct netpoll_info; 57struct device; 58struct phy_device; 59/* 802.11 specific */ 60struct wireless_dev; 61/* 802.15.4 specific */ 62struct wpan_dev; 63struct mpls_dev; 64 65void netdev_set_default_ethtool_ops(struct net_device *dev, 66 const struct ethtool_ops *ops); 67 68/* Backlog congestion levels */ 69#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 70#define NET_RX_DROP 1 /* packet dropped */ 71 72/* 73 * Transmit return codes: transmit return codes originate from three different 74 * namespaces: 75 * 76 * - qdisc return codes 77 * - driver transmit return codes 78 * - errno values 79 * 80 * Drivers are allowed to return any one of those in their hard_start_xmit() 81 * function. Real network devices commonly used with qdiscs should only return 82 * the driver transmit return codes though - when qdiscs are used, the actual 83 * transmission happens asynchronously, so the value is not propagated to 84 * higher layers. Virtual network devices transmit synchronously, in this case 85 * the driver transmit return codes are consumed by dev_queue_xmit(), all 86 * others are propagated to higher layers. 87 */ 88 89/* qdisc ->enqueue() return codes. */ 90#define NET_XMIT_SUCCESS 0x00 91#define NET_XMIT_DROP 0x01 /* skb dropped */ 92#define NET_XMIT_CN 0x02 /* congestion notification */ 93#define NET_XMIT_POLICED 0x03 /* skb is shot by police */ 94#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 95 96/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 97 * indicates that the device will soon be dropping packets, or already drops 98 * some packets of the same priority; prompting us to send less aggressively. */ 99#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 100#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 101 102/* Driver transmit return codes */ 103#define NETDEV_TX_MASK 0xf0 104 105enum netdev_tx { 106 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 107 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 108 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 109 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */ 110}; 111typedef enum netdev_tx netdev_tx_t; 112 113/* 114 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 115 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 116 */ 117static inline bool dev_xmit_complete(int rc) 118{ 119 /* 120 * Positive cases with an skb consumed by a driver: 121 * - successful transmission (rc == NETDEV_TX_OK) 122 * - error while transmitting (rc < 0) 123 * - error while queueing to a different device (rc & NET_XMIT_MASK) 124 */ 125 if (likely(rc < NET_XMIT_MASK)) 126 return true; 127 128 return false; 129} 130 131/* 132 * Compute the worst case header length according to the protocols 133 * used. 134 */ 135 136#if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 137# if defined(CONFIG_MAC80211_MESH) 138# define LL_MAX_HEADER 128 139# else 140# define LL_MAX_HEADER 96 141# endif 142#else 143# define LL_MAX_HEADER 32 144#endif 145 146#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 147 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 148#define MAX_HEADER LL_MAX_HEADER 149#else 150#define MAX_HEADER (LL_MAX_HEADER + 48) 151#endif 152 153/* 154 * Old network device statistics. Fields are native words 155 * (unsigned long) so they can be read and written atomically. 156 */ 157 158struct net_device_stats { 159 unsigned long rx_packets; 160 unsigned long tx_packets; 161 unsigned long rx_bytes; 162 unsigned long tx_bytes; 163 unsigned long rx_errors; 164 unsigned long tx_errors; 165 unsigned long rx_dropped; 166 unsigned long tx_dropped; 167 unsigned long multicast; 168 unsigned long collisions; 169 unsigned long rx_length_errors; 170 unsigned long rx_over_errors; 171 unsigned long rx_crc_errors; 172 unsigned long rx_frame_errors; 173 unsigned long rx_fifo_errors; 174 unsigned long rx_missed_errors; 175 unsigned long tx_aborted_errors; 176 unsigned long tx_carrier_errors; 177 unsigned long tx_fifo_errors; 178 unsigned long tx_heartbeat_errors; 179 unsigned long tx_window_errors; 180 unsigned long rx_compressed; 181 unsigned long tx_compressed; 182}; 183 184 185#include <linux/cache.h> 186#include <linux/skbuff.h> 187 188#ifdef CONFIG_RPS 189#include <linux/static_key.h> 190extern struct static_key rps_needed; 191#endif 192 193struct neighbour; 194struct neigh_parms; 195struct sk_buff; 196 197struct netdev_hw_addr { 198 struct list_head list; 199 unsigned char addr[MAX_ADDR_LEN]; 200 unsigned char type; 201#define NETDEV_HW_ADDR_T_LAN 1 202#define NETDEV_HW_ADDR_T_SAN 2 203#define NETDEV_HW_ADDR_T_SLAVE 3 204#define NETDEV_HW_ADDR_T_UNICAST 4 205#define NETDEV_HW_ADDR_T_MULTICAST 5 206 bool global_use; 207 int sync_cnt; 208 int refcount; 209 int synced; 210 struct rcu_head rcu_head; 211}; 212 213struct netdev_hw_addr_list { 214 struct list_head list; 215 int count; 216}; 217 218#define netdev_hw_addr_list_count(l) ((l)->count) 219#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 220#define netdev_hw_addr_list_for_each(ha, l) \ 221 list_for_each_entry(ha, &(l)->list, list) 222 223#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 224#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 225#define netdev_for_each_uc_addr(ha, dev) \ 226 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 227 228#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 229#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 230#define netdev_for_each_mc_addr(ha, dev) \ 231 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 232 233struct hh_cache { 234 u16 hh_len; 235 u16 __pad; 236 seqlock_t hh_lock; 237 238 /* cached hardware header; allow for machine alignment needs. */ 239#define HH_DATA_MOD 16 240#define HH_DATA_OFF(__len) \ 241 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 242#define HH_DATA_ALIGN(__len) \ 243 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 244 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 245}; 246 247/* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much. 248 * Alternative is: 249 * dev->hard_header_len ? (dev->hard_header_len + 250 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 251 * 252 * We could use other alignment values, but we must maintain the 253 * relationship HH alignment <= LL alignment. 254 */ 255#define LL_RESERVED_SPACE(dev) \ 256 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 257#define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 258 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 259 260struct header_ops { 261 int (*create) (struct sk_buff *skb, struct net_device *dev, 262 unsigned short type, const void *daddr, 263 const void *saddr, unsigned int len); 264 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 265 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 266 void (*cache_update)(struct hh_cache *hh, 267 const struct net_device *dev, 268 const unsigned char *haddr); 269}; 270 271/* These flag bits are private to the generic network queueing 272 * layer, they may not be explicitly referenced by any other 273 * code. 274 */ 275 276enum netdev_state_t { 277 __LINK_STATE_START, 278 __LINK_STATE_PRESENT, 279 __LINK_STATE_NOCARRIER, 280 __LINK_STATE_LINKWATCH_PENDING, 281 __LINK_STATE_DORMANT, 282}; 283 284 285/* 286 * This structure holds at boot time configured netdevice settings. They 287 * are then used in the device probing. 288 */ 289struct netdev_boot_setup { 290 char name[IFNAMSIZ]; 291 struct ifmap map; 292}; 293#define NETDEV_BOOT_SETUP_MAX 8 294 295int __init netdev_boot_setup(char *str); 296 297/* 298 * Structure for NAPI scheduling similar to tasklet but with weighting 299 */ 300struct napi_struct { 301 /* The poll_list must only be managed by the entity which 302 * changes the state of the NAPI_STATE_SCHED bit. This means 303 * whoever atomically sets that bit can add this napi_struct 304 * to the per-cpu poll_list, and whoever clears that bit 305 * can remove from the list right before clearing the bit. 306 */ 307 struct list_head poll_list; 308 309 unsigned long state; 310 int weight; 311 unsigned int gro_count; 312 int (*poll)(struct napi_struct *, int); 313#ifdef CONFIG_NETPOLL 314 spinlock_t poll_lock; 315 int poll_owner; 316#endif 317 struct net_device *dev; 318 struct sk_buff *gro_list; 319 struct sk_buff *skb; 320 struct hrtimer timer; 321 struct list_head dev_list; 322 struct hlist_node napi_hash_node; 323 unsigned int napi_id; 324}; 325 326enum { 327 NAPI_STATE_SCHED, /* Poll is scheduled */ 328 NAPI_STATE_DISABLE, /* Disable pending */ 329 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 330 NAPI_STATE_HASHED, /* In NAPI hash */ 331}; 332 333enum gro_result { 334 GRO_MERGED, 335 GRO_MERGED_FREE, 336 GRO_HELD, 337 GRO_NORMAL, 338 GRO_DROP, 339}; 340typedef enum gro_result gro_result_t; 341 342/* 343 * enum rx_handler_result - Possible return values for rx_handlers. 344 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 345 * further. 346 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 347 * case skb->dev was changed by rx_handler. 348 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 349 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called. 350 * 351 * rx_handlers are functions called from inside __netif_receive_skb(), to do 352 * special processing of the skb, prior to delivery to protocol handlers. 353 * 354 * Currently, a net_device can only have a single rx_handler registered. Trying 355 * to register a second rx_handler will return -EBUSY. 356 * 357 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 358 * To unregister a rx_handler on a net_device, use 359 * netdev_rx_handler_unregister(). 360 * 361 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 362 * do with the skb. 363 * 364 * If the rx_handler consumed to skb in some way, it should return 365 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 366 * the skb to be delivered in some other ways. 367 * 368 * If the rx_handler changed skb->dev, to divert the skb to another 369 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 370 * new device will be called if it exists. 371 * 372 * If the rx_handler consider the skb should be ignored, it should return 373 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 374 * are registered on exact device (ptype->dev == skb->dev). 375 * 376 * If the rx_handler didn't changed skb->dev, but want the skb to be normally 377 * delivered, it should return RX_HANDLER_PASS. 378 * 379 * A device without a registered rx_handler will behave as if rx_handler 380 * returned RX_HANDLER_PASS. 381 */ 382 383enum rx_handler_result { 384 RX_HANDLER_CONSUMED, 385 RX_HANDLER_ANOTHER, 386 RX_HANDLER_EXACT, 387 RX_HANDLER_PASS, 388}; 389typedef enum rx_handler_result rx_handler_result_t; 390typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 391 392void __napi_schedule(struct napi_struct *n); 393void __napi_schedule_irqoff(struct napi_struct *n); 394 395static inline bool napi_disable_pending(struct napi_struct *n) 396{ 397 return test_bit(NAPI_STATE_DISABLE, &n->state); 398} 399 400/** 401 * napi_schedule_prep - check if napi can be scheduled 402 * @n: napi context 403 * 404 * Test if NAPI routine is already running, and if not mark 405 * it as running. This is used as a condition variable 406 * insure only one NAPI poll instance runs. We also make 407 * sure there is no pending NAPI disable. 408 */ 409static inline bool napi_schedule_prep(struct napi_struct *n) 410{ 411 return !napi_disable_pending(n) && 412 !test_and_set_bit(NAPI_STATE_SCHED, &n->state); 413} 414 415/** 416 * napi_schedule - schedule NAPI poll 417 * @n: napi context 418 * 419 * Schedule NAPI poll routine to be called if it is not already 420 * running. 421 */ 422static inline void napi_schedule(struct napi_struct *n) 423{ 424 if (napi_schedule_prep(n)) 425 __napi_schedule(n); 426} 427 428/** 429 * napi_schedule_irqoff - schedule NAPI poll 430 * @n: napi context 431 * 432 * Variant of napi_schedule(), assuming hard irqs are masked. 433 */ 434static inline void napi_schedule_irqoff(struct napi_struct *n) 435{ 436 if (napi_schedule_prep(n)) 437 __napi_schedule_irqoff(n); 438} 439 440/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 441static inline bool napi_reschedule(struct napi_struct *napi) 442{ 443 if (napi_schedule_prep(napi)) { 444 __napi_schedule(napi); 445 return true; 446 } 447 return false; 448} 449 450void __napi_complete(struct napi_struct *n); 451void napi_complete_done(struct napi_struct *n, int work_done); 452/** 453 * napi_complete - NAPI processing complete 454 * @n: napi context 455 * 456 * Mark NAPI processing as complete. 457 * Consider using napi_complete_done() instead. 458 */ 459static inline void napi_complete(struct napi_struct *n) 460{ 461 return napi_complete_done(n, 0); 462} 463 464/** 465 * napi_by_id - lookup a NAPI by napi_id 466 * @napi_id: hashed napi_id 467 * 468 * lookup @napi_id in napi_hash table 469 * must be called under rcu_read_lock() 470 */ 471struct napi_struct *napi_by_id(unsigned int napi_id); 472 473/** 474 * napi_hash_add - add a NAPI to global hashtable 475 * @napi: napi context 476 * 477 * generate a new napi_id and store a @napi under it in napi_hash 478 */ 479void napi_hash_add(struct napi_struct *napi); 480 481/** 482 * napi_hash_del - remove a NAPI from global table 483 * @napi: napi context 484 * 485 * Warning: caller must observe rcu grace period 486 * before freeing memory containing @napi 487 */ 488void napi_hash_del(struct napi_struct *napi); 489 490/** 491 * napi_disable - prevent NAPI from scheduling 492 * @n: napi context 493 * 494 * Stop NAPI from being scheduled on this context. 495 * Waits till any outstanding processing completes. 496 */ 497void napi_disable(struct napi_struct *n); 498 499/** 500 * napi_enable - enable NAPI scheduling 501 * @n: napi context 502 * 503 * Resume NAPI from being scheduled on this context. 504 * Must be paired with napi_disable. 505 */ 506static inline void napi_enable(struct napi_struct *n) 507{ 508 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 509 smp_mb__before_atomic(); 510 clear_bit(NAPI_STATE_SCHED, &n->state); 511} 512 513#ifdef CONFIG_SMP 514/** 515 * napi_synchronize - wait until NAPI is not running 516 * @n: napi context 517 * 518 * Wait until NAPI is done being scheduled on this context. 519 * Waits till any outstanding processing completes but 520 * does not disable future activations. 521 */ 522static inline void napi_synchronize(const struct napi_struct *n) 523{ 524 while (test_bit(NAPI_STATE_SCHED, &n->state)) 525 msleep(1); 526} 527#else 528# define napi_synchronize(n) barrier() 529#endif 530 531enum netdev_queue_state_t { 532 __QUEUE_STATE_DRV_XOFF, 533 __QUEUE_STATE_STACK_XOFF, 534 __QUEUE_STATE_FROZEN, 535}; 536 537#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 538#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 539#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 540 541#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 542#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 543 QUEUE_STATE_FROZEN) 544#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 545 QUEUE_STATE_FROZEN) 546 547/* 548 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 549 * netif_tx_* functions below are used to manipulate this flag. The 550 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 551 * queue independently. The netif_xmit_*stopped functions below are called 552 * to check if the queue has been stopped by the driver or stack (either 553 * of the XOFF bits are set in the state). Drivers should not need to call 554 * netif_xmit*stopped functions, they should only be using netif_tx_*. 555 */ 556 557struct netdev_queue { 558/* 559 * read mostly part 560 */ 561 struct net_device *dev; 562 struct Qdisc __rcu *qdisc; 563 struct Qdisc *qdisc_sleeping; 564#ifdef CONFIG_SYSFS 565 struct kobject kobj; 566#endif 567#if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 568 int numa_node; 569#endif 570/* 571 * write mostly part 572 */ 573 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 574 int xmit_lock_owner; 575 /* 576 * please use this field instead of dev->trans_start 577 */ 578 unsigned long trans_start; 579 580 /* 581 * Number of TX timeouts for this queue 582 * (/sys/class/net/DEV/Q/trans_timeout) 583 */ 584 unsigned long trans_timeout; 585 586 unsigned long state; 587 588#ifdef CONFIG_BQL 589 struct dql dql; 590#endif 591 unsigned long tx_maxrate; 592} ____cacheline_aligned_in_smp; 593 594static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 595{ 596#if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 597 return q->numa_node; 598#else 599 return NUMA_NO_NODE; 600#endif 601} 602 603static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 604{ 605#if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 606 q->numa_node = node; 607#endif 608} 609 610#ifdef CONFIG_RPS 611/* 612 * This structure holds an RPS map which can be of variable length. The 613 * map is an array of CPUs. 614 */ 615struct rps_map { 616 unsigned int len; 617 struct rcu_head rcu; 618 u16 cpus[0]; 619}; 620#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 621 622/* 623 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 624 * tail pointer for that CPU's input queue at the time of last enqueue, and 625 * a hardware filter index. 626 */ 627struct rps_dev_flow { 628 u16 cpu; 629 u16 filter; 630 unsigned int last_qtail; 631}; 632#define RPS_NO_FILTER 0xffff 633 634/* 635 * The rps_dev_flow_table structure contains a table of flow mappings. 636 */ 637struct rps_dev_flow_table { 638 unsigned int mask; 639 struct rcu_head rcu; 640 struct rps_dev_flow flows[0]; 641}; 642#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 643 ((_num) * sizeof(struct rps_dev_flow))) 644 645/* 646 * The rps_sock_flow_table contains mappings of flows to the last CPU 647 * on which they were processed by the application (set in recvmsg). 648 * Each entry is a 32bit value. Upper part is the high order bits 649 * of flow hash, lower part is cpu number. 650 * rps_cpu_mask is used to partition the space, depending on number of 651 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 652 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f, 653 * meaning we use 32-6=26 bits for the hash. 654 */ 655struct rps_sock_flow_table { 656 u32 mask; 657 658 u32 ents[0] ____cacheline_aligned_in_smp; 659}; 660#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 661 662#define RPS_NO_CPU 0xffff 663 664extern u32 rps_cpu_mask; 665extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 666 667static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 668 u32 hash) 669{ 670 if (table && hash) { 671 unsigned int index = hash & table->mask; 672 u32 val = hash & ~rps_cpu_mask; 673 674 /* We only give a hint, preemption can change cpu under us */ 675 val |= raw_smp_processor_id(); 676 677 if (table->ents[index] != val) 678 table->ents[index] = val; 679 } 680} 681 682#ifdef CONFIG_RFS_ACCEL 683bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 684 u16 filter_id); 685#endif 686#endif /* CONFIG_RPS */ 687 688/* This structure contains an instance of an RX queue. */ 689struct netdev_rx_queue { 690#ifdef CONFIG_RPS 691 struct rps_map __rcu *rps_map; 692 struct rps_dev_flow_table __rcu *rps_flow_table; 693#endif 694 struct kobject kobj; 695 struct net_device *dev; 696} ____cacheline_aligned_in_smp; 697 698/* 699 * RX queue sysfs structures and functions. 700 */ 701struct rx_queue_attribute { 702 struct attribute attr; 703 ssize_t (*show)(struct netdev_rx_queue *queue, 704 struct rx_queue_attribute *attr, char *buf); 705 ssize_t (*store)(struct netdev_rx_queue *queue, 706 struct rx_queue_attribute *attr, const char *buf, size_t len); 707}; 708 709#ifdef CONFIG_XPS 710/* 711 * This structure holds an XPS map which can be of variable length. The 712 * map is an array of queues. 713 */ 714struct xps_map { 715 unsigned int len; 716 unsigned int alloc_len; 717 struct rcu_head rcu; 718 u16 queues[0]; 719}; 720#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 721#define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \ 722 / sizeof(u16)) 723 724/* 725 * This structure holds all XPS maps for device. Maps are indexed by CPU. 726 */ 727struct xps_dev_maps { 728 struct rcu_head rcu; 729 struct xps_map __rcu *cpu_map[0]; 730}; 731#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \ 732 (nr_cpu_ids * sizeof(struct xps_map *))) 733#endif /* CONFIG_XPS */ 734 735#define TC_MAX_QUEUE 16 736#define TC_BITMASK 15 737/* HW offloaded queuing disciplines txq count and offset maps */ 738struct netdev_tc_txq { 739 u16 count; 740 u16 offset; 741}; 742 743#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 744/* 745 * This structure is to hold information about the device 746 * configured to run FCoE protocol stack. 747 */ 748struct netdev_fcoe_hbainfo { 749 char manufacturer[64]; 750 char serial_number[64]; 751 char hardware_version[64]; 752 char driver_version[64]; 753 char optionrom_version[64]; 754 char firmware_version[64]; 755 char model[256]; 756 char model_description[256]; 757}; 758#endif 759 760#define MAX_PHYS_ITEM_ID_LEN 32 761 762/* This structure holds a unique identifier to identify some 763 * physical item (port for example) used by a netdevice. 764 */ 765struct netdev_phys_item_id { 766 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 767 unsigned char id_len; 768}; 769 770typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 771 struct sk_buff *skb); 772 773/* 774 * This structure defines the management hooks for network devices. 775 * The following hooks can be defined; unless noted otherwise, they are 776 * optional and can be filled with a null pointer. 777 * 778 * int (*ndo_init)(struct net_device *dev); 779 * This function is called once when network device is registered. 780 * The network device can use this to any late stage initializaton 781 * or semantic validattion. It can fail with an error code which will 782 * be propogated back to register_netdev 783 * 784 * void (*ndo_uninit)(struct net_device *dev); 785 * This function is called when device is unregistered or when registration 786 * fails. It is not called if init fails. 787 * 788 * int (*ndo_open)(struct net_device *dev); 789 * This function is called when network device transistions to the up 790 * state. 791 * 792 * int (*ndo_stop)(struct net_device *dev); 793 * This function is called when network device transistions to the down 794 * state. 795 * 796 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 797 * struct net_device *dev); 798 * Called when a packet needs to be transmitted. 799 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 800 * the queue before that can happen; it's for obsolete devices and weird 801 * corner cases, but the stack really does a non-trivial amount 802 * of useless work if you return NETDEV_TX_BUSY. 803 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX) 804 * Required can not be NULL. 805 * 806 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 807 * void *accel_priv, select_queue_fallback_t fallback); 808 * Called to decide which queue to when device supports multiple 809 * transmit queues. 810 * 811 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 812 * This function is called to allow device receiver to make 813 * changes to configuration when multicast or promiscious is enabled. 814 * 815 * void (*ndo_set_rx_mode)(struct net_device *dev); 816 * This function is called device changes address list filtering. 817 * If driver handles unicast address filtering, it should set 818 * IFF_UNICAST_FLT to its priv_flags. 819 * 820 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 821 * This function is called when the Media Access Control address 822 * needs to be changed. If this interface is not defined, the 823 * mac address can not be changed. 824 * 825 * int (*ndo_validate_addr)(struct net_device *dev); 826 * Test if Media Access Control address is valid for the device. 827 * 828 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 829 * Called when a user request an ioctl which can't be handled by 830 * the generic interface code. If not defined ioctl's return 831 * not supported error code. 832 * 833 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 834 * Used to set network devices bus interface parameters. This interface 835 * is retained for legacy reason, new devices should use the bus 836 * interface (PCI) for low level management. 837 * 838 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 839 * Called when a user wants to change the Maximum Transfer Unit 840 * of a device. If not defined, any request to change MTU will 841 * will return an error. 842 * 843 * void (*ndo_tx_timeout)(struct net_device *dev); 844 * Callback uses when the transmitter has not made any progress 845 * for dev->watchdog ticks. 846 * 847 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev, 848 * struct rtnl_link_stats64 *storage); 849 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 850 * Called when a user wants to get the network device usage 851 * statistics. Drivers must do one of the following: 852 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 853 * rtnl_link_stats64 structure passed by the caller. 854 * 2. Define @ndo_get_stats to update a net_device_stats structure 855 * (which should normally be dev->stats) and return a pointer to 856 * it. The structure may be changed asynchronously only if each 857 * field is written atomically. 858 * 3. Update dev->stats asynchronously and atomically, and define 859 * neither operation. 860 * 861 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 862 * If device support VLAN filtering this function is called when a 863 * VLAN id is registered. 864 * 865 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 866 * If device support VLAN filtering this function is called when a 867 * VLAN id is unregistered. 868 * 869 * void (*ndo_poll_controller)(struct net_device *dev); 870 * 871 * SR-IOV management functions. 872 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 873 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos); 874 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 875 * int max_tx_rate); 876 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 877 * int (*ndo_get_vf_config)(struct net_device *dev, 878 * int vf, struct ifla_vf_info *ivf); 879 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 880 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 881 * struct nlattr *port[]); 882 * 883 * Enable or disable the VF ability to query its RSS Redirection Table and 884 * Hash Key. This is needed since on some devices VF share this information 885 * with PF and querying it may adduce a theoretical security risk. 886 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 887 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 888 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc) 889 * Called to setup 'tc' number of traffic classes in the net device. This 890 * is always called from the stack with the rtnl lock held and netif tx 891 * queues stopped. This allows the netdevice to perform queue management 892 * safely. 893 * 894 * Fiber Channel over Ethernet (FCoE) offload functions. 895 * int (*ndo_fcoe_enable)(struct net_device *dev); 896 * Called when the FCoE protocol stack wants to start using LLD for FCoE 897 * so the underlying device can perform whatever needed configuration or 898 * initialization to support acceleration of FCoE traffic. 899 * 900 * int (*ndo_fcoe_disable)(struct net_device *dev); 901 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 902 * so the underlying device can perform whatever needed clean-ups to 903 * stop supporting acceleration of FCoE traffic. 904 * 905 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 906 * struct scatterlist *sgl, unsigned int sgc); 907 * Called when the FCoE Initiator wants to initialize an I/O that 908 * is a possible candidate for Direct Data Placement (DDP). The LLD can 909 * perform necessary setup and returns 1 to indicate the device is set up 910 * successfully to perform DDP on this I/O, otherwise this returns 0. 911 * 912 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 913 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 914 * indicated by the FC exchange id 'xid', so the underlying device can 915 * clean up and reuse resources for later DDP requests. 916 * 917 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 918 * struct scatterlist *sgl, unsigned int sgc); 919 * Called when the FCoE Target wants to initialize an I/O that 920 * is a possible candidate for Direct Data Placement (DDP). The LLD can 921 * perform necessary setup and returns 1 to indicate the device is set up 922 * successfully to perform DDP on this I/O, otherwise this returns 0. 923 * 924 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 925 * struct netdev_fcoe_hbainfo *hbainfo); 926 * Called when the FCoE Protocol stack wants information on the underlying 927 * device. This information is utilized by the FCoE protocol stack to 928 * register attributes with Fiber Channel management service as per the 929 * FC-GS Fabric Device Management Information(FDMI) specification. 930 * 931 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 932 * Called when the underlying device wants to override default World Wide 933 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 934 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 935 * protocol stack to use. 936 * 937 * RFS acceleration. 938 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 939 * u16 rxq_index, u32 flow_id); 940 * Set hardware filter for RFS. rxq_index is the target queue index; 941 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 942 * Return the filter ID on success, or a negative error code. 943 * 944 * Slave management functions (for bridge, bonding, etc). 945 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 946 * Called to make another netdev an underling. 947 * 948 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 949 * Called to release previously enslaved netdev. 950 * 951 * Feature/offload setting functions. 952 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 953 * netdev_features_t features); 954 * Adjusts the requested feature flags according to device-specific 955 * constraints, and returns the resulting flags. Must not modify 956 * the device state. 957 * 958 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 959 * Called to update device configuration to new features. Passed 960 * feature set might be less than what was returned by ndo_fix_features()). 961 * Must return >0 or -errno if it changed dev->features itself. 962 * 963 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 964 * struct net_device *dev, 965 * const unsigned char *addr, u16 vid, u16 flags) 966 * Adds an FDB entry to dev for addr. 967 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 968 * struct net_device *dev, 969 * const unsigned char *addr, u16 vid) 970 * Deletes the FDB entry from dev coresponding to addr. 971 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 972 * struct net_device *dev, struct net_device *filter_dev, 973 * int idx) 974 * Used to add FDB entries to dump requests. Implementers should add 975 * entries to skb and update idx with the number of entries. 976 * 977 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 978 * u16 flags) 979 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 980 * struct net_device *dev, u32 filter_mask, 981 * int nlflags) 982 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 983 * u16 flags); 984 * 985 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 986 * Called to change device carrier. Soft-devices (like dummy, team, etc) 987 * which do not represent real hardware may define this to allow their 988 * userspace components to manage their virtual carrier state. Devices 989 * that determine carrier state from physical hardware properties (eg 990 * network cables) or protocol-dependent mechanisms (eg 991 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 992 * 993 * int (*ndo_get_phys_port_id)(struct net_device *dev, 994 * struct netdev_phys_item_id *ppid); 995 * Called to get ID of physical port of this device. If driver does 996 * not implement this, it is assumed that the hw is not able to have 997 * multiple net devices on single physical port. 998 * 999 * void (*ndo_add_vxlan_port)(struct net_device *dev, 1000 * sa_family_t sa_family, __be16 port); 1001 * Called by vxlan to notiy a driver about the UDP port and socket 1002 * address family that vxlan is listnening to. It is called only when 1003 * a new port starts listening. The operation is protected by the 1004 * vxlan_net->sock_lock. 1005 * 1006 * void (*ndo_del_vxlan_port)(struct net_device *dev, 1007 * sa_family_t sa_family, __be16 port); 1008 * Called by vxlan to notify the driver about a UDP port and socket 1009 * address family that vxlan is not listening to anymore. The operation 1010 * is protected by the vxlan_net->sock_lock. 1011 * 1012 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1013 * struct net_device *dev) 1014 * Called by upper layer devices to accelerate switching or other 1015 * station functionality into hardware. 'pdev is the lowerdev 1016 * to use for the offload and 'dev' is the net device that will 1017 * back the offload. Returns a pointer to the private structure 1018 * the upper layer will maintain. 1019 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1020 * Called by upper layer device to delete the station created 1021 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1022 * the station and priv is the structure returned by the add 1023 * operation. 1024 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb, 1025 * struct net_device *dev, 1026 * void *priv); 1027 * Callback to use for xmit over the accelerated station. This 1028 * is used in place of ndo_start_xmit on accelerated net 1029 * devices. 1030 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb, 1031 * struct net_device *dev 1032 * netdev_features_t features); 1033 * Called by core transmit path to determine if device is capable of 1034 * performing offload operations on a given packet. This is to give 1035 * the device an opportunity to implement any restrictions that cannot 1036 * be otherwise expressed by feature flags. The check is called with 1037 * the set of features that the stack has calculated and it returns 1038 * those the driver believes to be appropriate. 1039 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1040 * int queue_index, u32 maxrate); 1041 * Called when a user wants to set a max-rate limitation of specific 1042 * TX queue. 1043 * int (*ndo_get_iflink)(const struct net_device *dev); 1044 * Called to get the iflink value of this device. 1045 */ 1046struct net_device_ops { 1047 int (*ndo_init)(struct net_device *dev); 1048 void (*ndo_uninit)(struct net_device *dev); 1049 int (*ndo_open)(struct net_device *dev); 1050 int (*ndo_stop)(struct net_device *dev); 1051 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb, 1052 struct net_device *dev); 1053 u16 (*ndo_select_queue)(struct net_device *dev, 1054 struct sk_buff *skb, 1055 void *accel_priv, 1056 select_queue_fallback_t fallback); 1057 void (*ndo_change_rx_flags)(struct net_device *dev, 1058 int flags); 1059 void (*ndo_set_rx_mode)(struct net_device *dev); 1060 int (*ndo_set_mac_address)(struct net_device *dev, 1061 void *addr); 1062 int (*ndo_validate_addr)(struct net_device *dev); 1063 int (*ndo_do_ioctl)(struct net_device *dev, 1064 struct ifreq *ifr, int cmd); 1065 int (*ndo_set_config)(struct net_device *dev, 1066 struct ifmap *map); 1067 int (*ndo_change_mtu)(struct net_device *dev, 1068 int new_mtu); 1069 int (*ndo_neigh_setup)(struct net_device *dev, 1070 struct neigh_parms *); 1071 void (*ndo_tx_timeout) (struct net_device *dev); 1072 1073 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev, 1074 struct rtnl_link_stats64 *storage); 1075 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1076 1077 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1078 __be16 proto, u16 vid); 1079 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1080 __be16 proto, u16 vid); 1081#ifdef CONFIG_NET_POLL_CONTROLLER 1082 void (*ndo_poll_controller)(struct net_device *dev); 1083 int (*ndo_netpoll_setup)(struct net_device *dev, 1084 struct netpoll_info *info); 1085 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1086#endif 1087#ifdef CONFIG_NET_RX_BUSY_POLL 1088 int (*ndo_busy_poll)(struct napi_struct *dev); 1089#endif 1090 int (*ndo_set_vf_mac)(struct net_device *dev, 1091 int queue, u8 *mac); 1092 int (*ndo_set_vf_vlan)(struct net_device *dev, 1093 int queue, u16 vlan, u8 qos); 1094 int (*ndo_set_vf_rate)(struct net_device *dev, 1095 int vf, int min_tx_rate, 1096 int max_tx_rate); 1097 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1098 int vf, bool setting); 1099 int (*ndo_get_vf_config)(struct net_device *dev, 1100 int vf, 1101 struct ifla_vf_info *ivf); 1102 int (*ndo_set_vf_link_state)(struct net_device *dev, 1103 int vf, int link_state); 1104 int (*ndo_set_vf_port)(struct net_device *dev, 1105 int vf, 1106 struct nlattr *port[]); 1107 int (*ndo_get_vf_port)(struct net_device *dev, 1108 int vf, struct sk_buff *skb); 1109 int (*ndo_set_vf_rss_query_en)( 1110 struct net_device *dev, 1111 int vf, bool setting); 1112 int (*ndo_setup_tc)(struct net_device *dev, u8 tc); 1113#if IS_ENABLED(CONFIG_FCOE) 1114 int (*ndo_fcoe_enable)(struct net_device *dev); 1115 int (*ndo_fcoe_disable)(struct net_device *dev); 1116 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1117 u16 xid, 1118 struct scatterlist *sgl, 1119 unsigned int sgc); 1120 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1121 u16 xid); 1122 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1123 u16 xid, 1124 struct scatterlist *sgl, 1125 unsigned int sgc); 1126 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1127 struct netdev_fcoe_hbainfo *hbainfo); 1128#endif 1129 1130#if IS_ENABLED(CONFIG_LIBFCOE) 1131#define NETDEV_FCOE_WWNN 0 1132#define NETDEV_FCOE_WWPN 1 1133 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1134 u64 *wwn, int type); 1135#endif 1136 1137#ifdef CONFIG_RFS_ACCEL 1138 int (*ndo_rx_flow_steer)(struct net_device *dev, 1139 const struct sk_buff *skb, 1140 u16 rxq_index, 1141 u32 flow_id); 1142#endif 1143 int (*ndo_add_slave)(struct net_device *dev, 1144 struct net_device *slave_dev); 1145 int (*ndo_del_slave)(struct net_device *dev, 1146 struct net_device *slave_dev); 1147 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1148 netdev_features_t features); 1149 int (*ndo_set_features)(struct net_device *dev, 1150 netdev_features_t features); 1151 int (*ndo_neigh_construct)(struct neighbour *n); 1152 void (*ndo_neigh_destroy)(struct neighbour *n); 1153 1154 int (*ndo_fdb_add)(struct ndmsg *ndm, 1155 struct nlattr *tb[], 1156 struct net_device *dev, 1157 const unsigned char *addr, 1158 u16 vid, 1159 u16 flags); 1160 int (*ndo_fdb_del)(struct ndmsg *ndm, 1161 struct nlattr *tb[], 1162 struct net_device *dev, 1163 const unsigned char *addr, 1164 u16 vid); 1165 int (*ndo_fdb_dump)(struct sk_buff *skb, 1166 struct netlink_callback *cb, 1167 struct net_device *dev, 1168 struct net_device *filter_dev, 1169 int idx); 1170 1171 int (*ndo_bridge_setlink)(struct net_device *dev, 1172 struct nlmsghdr *nlh, 1173 u16 flags); 1174 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1175 u32 pid, u32 seq, 1176 struct net_device *dev, 1177 u32 filter_mask, 1178 int nlflags); 1179 int (*ndo_bridge_dellink)(struct net_device *dev, 1180 struct nlmsghdr *nlh, 1181 u16 flags); 1182 int (*ndo_change_carrier)(struct net_device *dev, 1183 bool new_carrier); 1184 int (*ndo_get_phys_port_id)(struct net_device *dev, 1185 struct netdev_phys_item_id *ppid); 1186 int (*ndo_get_phys_port_name)(struct net_device *dev, 1187 char *name, size_t len); 1188 void (*ndo_add_vxlan_port)(struct net_device *dev, 1189 sa_family_t sa_family, 1190 __be16 port); 1191 void (*ndo_del_vxlan_port)(struct net_device *dev, 1192 sa_family_t sa_family, 1193 __be16 port); 1194 1195 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1196 struct net_device *dev); 1197 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1198 void *priv); 1199 1200 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb, 1201 struct net_device *dev, 1202 void *priv); 1203 int (*ndo_get_lock_subclass)(struct net_device *dev); 1204 netdev_features_t (*ndo_features_check) (struct sk_buff *skb, 1205 struct net_device *dev, 1206 netdev_features_t features); 1207 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1208 int queue_index, 1209 u32 maxrate); 1210 int (*ndo_get_iflink)(const struct net_device *dev); 1211}; 1212 1213/** 1214 * enum net_device_priv_flags - &struct net_device priv_flags 1215 * 1216 * These are the &struct net_device, they are only set internally 1217 * by drivers and used in the kernel. These flags are invisible to 1218 * userspace, this means that the order of these flags can change 1219 * during any kernel release. 1220 * 1221 * You should have a pretty good reason to be extending these flags. 1222 * 1223 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1224 * @IFF_EBRIDGE: Ethernet bridging device 1225 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active 1226 * @IFF_MASTER_8023AD: bonding master, 802.3ad 1227 * @IFF_MASTER_ALB: bonding master, balance-alb 1228 * @IFF_BONDING: bonding master or slave 1229 * @IFF_SLAVE_NEEDARP: need ARPs for validation 1230 * @IFF_ISATAP: ISATAP interface (RFC4214) 1231 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use 1232 * @IFF_WAN_HDLC: WAN HDLC device 1233 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1234 * release skb->dst 1235 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1236 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1237 * @IFF_MACVLAN_PORT: device used as macvlan port 1238 * @IFF_BRIDGE_PORT: device used as bridge port 1239 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1240 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1241 * @IFF_UNICAST_FLT: Supports unicast filtering 1242 * @IFF_TEAM_PORT: device used as team port 1243 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1244 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1245 * change when it's running 1246 * @IFF_MACVLAN: Macvlan device 1247 */ 1248enum netdev_priv_flags { 1249 IFF_802_1Q_VLAN = 1<<0, 1250 IFF_EBRIDGE = 1<<1, 1251 IFF_SLAVE_INACTIVE = 1<<2, 1252 IFF_MASTER_8023AD = 1<<3, 1253 IFF_MASTER_ALB = 1<<4, 1254 IFF_BONDING = 1<<5, 1255 IFF_SLAVE_NEEDARP = 1<<6, 1256 IFF_ISATAP = 1<<7, 1257 IFF_MASTER_ARPMON = 1<<8, 1258 IFF_WAN_HDLC = 1<<9, 1259 IFF_XMIT_DST_RELEASE = 1<<10, 1260 IFF_DONT_BRIDGE = 1<<11, 1261 IFF_DISABLE_NETPOLL = 1<<12, 1262 IFF_MACVLAN_PORT = 1<<13, 1263 IFF_BRIDGE_PORT = 1<<14, 1264 IFF_OVS_DATAPATH = 1<<15, 1265 IFF_TX_SKB_SHARING = 1<<16, 1266 IFF_UNICAST_FLT = 1<<17, 1267 IFF_TEAM_PORT = 1<<18, 1268 IFF_SUPP_NOFCS = 1<<19, 1269 IFF_LIVE_ADDR_CHANGE = 1<<20, 1270 IFF_MACVLAN = 1<<21, 1271 IFF_XMIT_DST_RELEASE_PERM = 1<<22, 1272 IFF_IPVLAN_MASTER = 1<<23, 1273 IFF_IPVLAN_SLAVE = 1<<24, 1274}; 1275 1276#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1277#define IFF_EBRIDGE IFF_EBRIDGE 1278#define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE 1279#define IFF_MASTER_8023AD IFF_MASTER_8023AD 1280#define IFF_MASTER_ALB IFF_MASTER_ALB 1281#define IFF_BONDING IFF_BONDING 1282#define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP 1283#define IFF_ISATAP IFF_ISATAP 1284#define IFF_MASTER_ARPMON IFF_MASTER_ARPMON 1285#define IFF_WAN_HDLC IFF_WAN_HDLC 1286#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1287#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1288#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1289#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1290#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1291#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1292#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1293#define IFF_UNICAST_FLT IFF_UNICAST_FLT 1294#define IFF_TEAM_PORT IFF_TEAM_PORT 1295#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1296#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1297#define IFF_MACVLAN IFF_MACVLAN 1298#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1299#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER 1300#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE 1301 1302/** 1303 * struct net_device - The DEVICE structure. 1304 * Actually, this whole structure is a big mistake. It mixes I/O 1305 * data with strictly "high-level" data, and it has to know about 1306 * almost every data structure used in the INET module. 1307 * 1308 * @name: This is the first field of the "visible" part of this structure 1309 * (i.e. as seen by users in the "Space.c" file). It is the name 1310 * of the interface. 1311 * 1312 * @name_hlist: Device name hash chain, please keep it close to name[] 1313 * @ifalias: SNMP alias 1314 * @mem_end: Shared memory end 1315 * @mem_start: Shared memory start 1316 * @base_addr: Device I/O address 1317 * @irq: Device IRQ number 1318 * 1319 * @carrier_changes: Stats to monitor carrier on<->off transitions 1320 * 1321 * @state: Generic network queuing layer state, see netdev_state_t 1322 * @dev_list: The global list of network devices 1323 * @napi_list: List entry, that is used for polling napi devices 1324 * @unreg_list: List entry, that is used, when we are unregistering the 1325 * device, see the function unregister_netdev 1326 * @close_list: List entry, that is used, when we are closing the device 1327 * 1328 * @adj_list: Directly linked devices, like slaves for bonding 1329 * @all_adj_list: All linked devices, *including* neighbours 1330 * @features: Currently active device features 1331 * @hw_features: User-changeable features 1332 * 1333 * @wanted_features: User-requested features 1334 * @vlan_features: Mask of features inheritable by VLAN devices 1335 * 1336 * @hw_enc_features: Mask of features inherited by encapsulating devices 1337 * This field indicates what encapsulation 1338 * offloads the hardware is capable of doing, 1339 * and drivers will need to set them appropriately. 1340 * 1341 * @mpls_features: Mask of features inheritable by MPLS 1342 * 1343 * @ifindex: interface index 1344 * @group: The group, that the device belongs to 1345 * 1346 * @stats: Statistics struct, which was left as a legacy, use 1347 * rtnl_link_stats64 instead 1348 * 1349 * @rx_dropped: Dropped packets by core network, 1350 * do not use this in drivers 1351 * @tx_dropped: Dropped packets by core network, 1352 * do not use this in drivers 1353 * 1354 * @wireless_handlers: List of functions to handle Wireless Extensions, 1355 * instead of ioctl, 1356 * see <net/iw_handler.h> for details. 1357 * @wireless_data: Instance data managed by the core of wireless extensions 1358 * 1359 * @netdev_ops: Includes several pointers to callbacks, 1360 * if one wants to override the ndo_*() functions 1361 * @ethtool_ops: Management operations 1362 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1363 * of Layer 2 headers. 1364 * 1365 * @flags: Interface flags (a la BSD) 1366 * @priv_flags: Like 'flags' but invisible to userspace, 1367 * see if.h for the definitions 1368 * @gflags: Global flags ( kept as legacy ) 1369 * @padded: How much padding added by alloc_netdev() 1370 * @operstate: RFC2863 operstate 1371 * @link_mode: Mapping policy to operstate 1372 * @if_port: Selectable AUI, TP, ... 1373 * @dma: DMA channel 1374 * @mtu: Interface MTU value 1375 * @type: Interface hardware type 1376 * @hard_header_len: Hardware header length 1377 * 1378 * @needed_headroom: Extra headroom the hardware may need, but not in all 1379 * cases can this be guaranteed 1380 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1381 * cases can this be guaranteed. Some cases also use 1382 * LL_MAX_HEADER instead to allocate the skb 1383 * 1384 * interface address info: 1385 * 1386 * @perm_addr: Permanent hw address 1387 * @addr_assign_type: Hw address assignment type 1388 * @addr_len: Hardware address length 1389 * @neigh_priv_len; Used in neigh_alloc(), 1390 * initialized only in atm/clip.c 1391 * @dev_id: Used to differentiate devices that share 1392 * the same link layer address 1393 * @dev_port: Used to differentiate devices that share 1394 * the same function 1395 * @addr_list_lock: XXX: need comments on this one 1396 * @uc_promisc: Counter, that indicates, that promiscuous mode 1397 * has been enabled due to the need to listen to 1398 * additional unicast addresses in a device that 1399 * does not implement ndo_set_rx_mode() 1400 * @uc: unicast mac addresses 1401 * @mc: multicast mac addresses 1402 * @dev_addrs: list of device hw addresses 1403 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1404 * @promiscuity: Number of times, the NIC is told to work in 1405 * Promiscuous mode, if it becomes 0 the NIC will 1406 * exit from working in Promiscuous mode 1407 * @allmulti: Counter, enables or disables allmulticast mode 1408 * 1409 * @vlan_info: VLAN info 1410 * @dsa_ptr: dsa specific data 1411 * @tipc_ptr: TIPC specific data 1412 * @atalk_ptr: AppleTalk link 1413 * @ip_ptr: IPv4 specific data 1414 * @dn_ptr: DECnet specific data 1415 * @ip6_ptr: IPv6 specific data 1416 * @ax25_ptr: AX.25 specific data 1417 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1418 * 1419 * @last_rx: Time of last Rx 1420 * @dev_addr: Hw address (before bcast, 1421 * because most packets are unicast) 1422 * 1423 * @_rx: Array of RX queues 1424 * @num_rx_queues: Number of RX queues 1425 * allocated at register_netdev() time 1426 * @real_num_rx_queues: Number of RX queues currently active in device 1427 * 1428 * @rx_handler: handler for received packets 1429 * @rx_handler_data: XXX: need comments on this one 1430 * @ingress_queue: XXX: need comments on this one 1431 * @broadcast: hw bcast address 1432 * 1433 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1434 * indexed by RX queue number. Assigned by driver. 1435 * This must only be set if the ndo_rx_flow_steer 1436 * operation is defined 1437 * @index_hlist: Device index hash chain 1438 * 1439 * @_tx: Array of TX queues 1440 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1441 * @real_num_tx_queues: Number of TX queues currently active in device 1442 * @qdisc: Root qdisc from userspace point of view 1443 * @tx_queue_len: Max frames per queue allowed 1444 * @tx_global_lock: XXX: need comments on this one 1445 * 1446 * @xps_maps: XXX: need comments on this one 1447 * 1448 * @trans_start: Time (in jiffies) of last Tx 1449 * @watchdog_timeo: Represents the timeout that is used by 1450 * the watchdog ( see dev_watchdog() ) 1451 * @watchdog_timer: List of timers 1452 * 1453 * @pcpu_refcnt: Number of references to this device 1454 * @todo_list: Delayed register/unregister 1455 * @link_watch_list: XXX: need comments on this one 1456 * 1457 * @reg_state: Register/unregister state machine 1458 * @dismantle: Device is going to be freed 1459 * @rtnl_link_state: This enum represents the phases of creating 1460 * a new link 1461 * 1462 * @destructor: Called from unregister, 1463 * can be used to call free_netdev 1464 * @npinfo: XXX: need comments on this one 1465 * @nd_net: Network namespace this network device is inside 1466 * 1467 * @ml_priv: Mid-layer private 1468 * @lstats: Loopback statistics 1469 * @tstats: Tunnel statistics 1470 * @dstats: Dummy statistics 1471 * @vstats: Virtual ethernet statistics 1472 * 1473 * @garp_port: GARP 1474 * @mrp_port: MRP 1475 * 1476 * @dev: Class/net/name entry 1477 * @sysfs_groups: Space for optional device, statistics and wireless 1478 * sysfs groups 1479 * 1480 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1481 * @rtnl_link_ops: Rtnl_link_ops 1482 * 1483 * @gso_max_size: Maximum size of generic segmentation offload 1484 * @gso_max_segs: Maximum number of segments that can be passed to the 1485 * NIC for GSO 1486 * @gso_min_segs: Minimum number of segments that can be passed to the 1487 * NIC for GSO 1488 * 1489 * @dcbnl_ops: Data Center Bridging netlink ops 1490 * @num_tc: Number of traffic classes in the net device 1491 * @tc_to_txq: XXX: need comments on this one 1492 * @prio_tc_map XXX: need comments on this one 1493 * 1494 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1495 * 1496 * @priomap: XXX: need comments on this one 1497 * @phydev: Physical device may attach itself 1498 * for hardware timestamping 1499 * 1500 * @qdisc_tx_busylock: XXX: need comments on this one 1501 * 1502 * @pm_qos_req: Power Management QoS object 1503 * 1504 * FIXME: cleanup struct net_device such that network protocol info 1505 * moves out. 1506 */ 1507 1508struct net_device { 1509 char name[IFNAMSIZ]; 1510 struct hlist_node name_hlist; 1511 char *ifalias; 1512 /* 1513 * I/O specific fields 1514 * FIXME: Merge these and struct ifmap into one 1515 */ 1516 unsigned long mem_end; 1517 unsigned long mem_start; 1518 unsigned long base_addr; 1519 int irq; 1520 1521 atomic_t carrier_changes; 1522 1523 /* 1524 * Some hardware also needs these fields (state,dev_list, 1525 * napi_list,unreg_list,close_list) but they are not 1526 * part of the usual set specified in Space.c. 1527 */ 1528 1529 unsigned long state; 1530 1531 struct list_head dev_list; 1532 struct list_head napi_list; 1533 struct list_head unreg_list; 1534 struct list_head close_list; 1535 struct list_head ptype_all; 1536 struct list_head ptype_specific; 1537 1538 struct { 1539 struct list_head upper; 1540 struct list_head lower; 1541 } adj_list; 1542 1543 struct { 1544 struct list_head upper; 1545 struct list_head lower; 1546 } all_adj_list; 1547 1548 netdev_features_t features; 1549 netdev_features_t hw_features; 1550 netdev_features_t wanted_features; 1551 netdev_features_t vlan_features; 1552 netdev_features_t hw_enc_features; 1553 netdev_features_t mpls_features; 1554 1555 int ifindex; 1556 int group; 1557 1558 struct net_device_stats stats; 1559 1560 atomic_long_t rx_dropped; 1561 atomic_long_t tx_dropped; 1562 1563#ifdef CONFIG_WIRELESS_EXT 1564 const struct iw_handler_def * wireless_handlers; 1565 struct iw_public_data * wireless_data; 1566#endif 1567 const struct net_device_ops *netdev_ops; 1568 const struct ethtool_ops *ethtool_ops; 1569#ifdef CONFIG_NET_SWITCHDEV 1570 const struct swdev_ops *swdev_ops; 1571#endif 1572 1573 const struct header_ops *header_ops; 1574 1575 unsigned int flags; 1576 unsigned int priv_flags; 1577 1578 unsigned short gflags; 1579 unsigned short padded; 1580 1581 unsigned char operstate; 1582 unsigned char link_mode; 1583 1584 unsigned char if_port; 1585 unsigned char dma; 1586 1587 unsigned int mtu; 1588 unsigned short type; 1589 unsigned short hard_header_len; 1590 1591 unsigned short needed_headroom; 1592 unsigned short needed_tailroom; 1593 1594 /* Interface address info. */ 1595 unsigned char perm_addr[MAX_ADDR_LEN]; 1596 unsigned char addr_assign_type; 1597 unsigned char addr_len; 1598 unsigned short neigh_priv_len; 1599 unsigned short dev_id; 1600 unsigned short dev_port; 1601 spinlock_t addr_list_lock; 1602 unsigned char name_assign_type; 1603 bool uc_promisc; 1604 struct netdev_hw_addr_list uc; 1605 struct netdev_hw_addr_list mc; 1606 struct netdev_hw_addr_list dev_addrs; 1607 1608#ifdef CONFIG_SYSFS 1609 struct kset *queues_kset; 1610#endif 1611 unsigned int promiscuity; 1612 unsigned int allmulti; 1613 1614 1615 /* Protocol specific pointers */ 1616 1617#if IS_ENABLED(CONFIG_VLAN_8021Q) 1618 struct vlan_info __rcu *vlan_info; 1619#endif 1620#if IS_ENABLED(CONFIG_NET_DSA) 1621 struct dsa_switch_tree *dsa_ptr; 1622#endif 1623#if IS_ENABLED(CONFIG_TIPC) 1624 struct tipc_bearer __rcu *tipc_ptr; 1625#endif 1626 void *atalk_ptr; 1627 struct in_device __rcu *ip_ptr; 1628 struct dn_dev __rcu *dn_ptr; 1629 struct inet6_dev __rcu *ip6_ptr; 1630 void *ax25_ptr; 1631 struct wireless_dev *ieee80211_ptr; 1632 struct wpan_dev *ieee802154_ptr; 1633#if IS_ENABLED(CONFIG_MPLS_ROUTING) 1634 struct mpls_dev __rcu *mpls_ptr; 1635#endif 1636 1637/* 1638 * Cache lines mostly used on receive path (including eth_type_trans()) 1639 */ 1640 unsigned long last_rx; 1641 1642 /* Interface address info used in eth_type_trans() */ 1643 unsigned char *dev_addr; 1644 1645 1646#ifdef CONFIG_SYSFS 1647 struct netdev_rx_queue *_rx; 1648 1649 unsigned int num_rx_queues; 1650 unsigned int real_num_rx_queues; 1651 1652#endif 1653 1654 unsigned long gro_flush_timeout; 1655 rx_handler_func_t __rcu *rx_handler; 1656 void __rcu *rx_handler_data; 1657 1658 struct netdev_queue __rcu *ingress_queue; 1659 unsigned char broadcast[MAX_ADDR_LEN]; 1660#ifdef CONFIG_RFS_ACCEL 1661 struct cpu_rmap *rx_cpu_rmap; 1662#endif 1663 struct hlist_node index_hlist; 1664 1665/* 1666 * Cache lines mostly used on transmit path 1667 */ 1668 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1669 unsigned int num_tx_queues; 1670 unsigned int real_num_tx_queues; 1671 struct Qdisc *qdisc; 1672 unsigned long tx_queue_len; 1673 spinlock_t tx_global_lock; 1674 int watchdog_timeo; 1675 1676#ifdef CONFIG_XPS 1677 struct xps_dev_maps __rcu *xps_maps; 1678#endif 1679 1680 /* These may be needed for future network-power-down code. */ 1681 1682 /* 1683 * trans_start here is expensive for high speed devices on SMP, 1684 * please use netdev_queue->trans_start instead. 1685 */ 1686 unsigned long trans_start; 1687 1688 struct timer_list watchdog_timer; 1689 1690 int __percpu *pcpu_refcnt; 1691 struct list_head todo_list; 1692 1693 struct list_head link_watch_list; 1694 1695 enum { NETREG_UNINITIALIZED=0, 1696 NETREG_REGISTERED, /* completed register_netdevice */ 1697 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1698 NETREG_UNREGISTERED, /* completed unregister todo */ 1699 NETREG_RELEASED, /* called free_netdev */ 1700 NETREG_DUMMY, /* dummy device for NAPI poll */ 1701 } reg_state:8; 1702 1703 bool dismantle; 1704 1705 enum { 1706 RTNL_LINK_INITIALIZED, 1707 RTNL_LINK_INITIALIZING, 1708 } rtnl_link_state:16; 1709 1710 void (*destructor)(struct net_device *dev); 1711 1712#ifdef CONFIG_NETPOLL 1713 struct netpoll_info __rcu *npinfo; 1714#endif 1715 1716 possible_net_t nd_net; 1717 1718 /* mid-layer private */ 1719 union { 1720 void *ml_priv; 1721 struct pcpu_lstats __percpu *lstats; 1722 struct pcpu_sw_netstats __percpu *tstats; 1723 struct pcpu_dstats __percpu *dstats; 1724 struct pcpu_vstats __percpu *vstats; 1725 }; 1726 1727 struct garp_port __rcu *garp_port; 1728 struct mrp_port __rcu *mrp_port; 1729 1730 struct device dev; 1731 const struct attribute_group *sysfs_groups[4]; 1732 const struct attribute_group *sysfs_rx_queue_group; 1733 1734 const struct rtnl_link_ops *rtnl_link_ops; 1735 1736 /* for setting kernel sock attribute on TCP connection setup */ 1737#define GSO_MAX_SIZE 65536 1738 unsigned int gso_max_size; 1739#define GSO_MAX_SEGS 65535 1740 u16 gso_max_segs; 1741 u16 gso_min_segs; 1742#ifdef CONFIG_DCB 1743 const struct dcbnl_rtnl_ops *dcbnl_ops; 1744#endif 1745 u8 num_tc; 1746 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 1747 u8 prio_tc_map[TC_BITMASK + 1]; 1748 1749#if IS_ENABLED(CONFIG_FCOE) 1750 unsigned int fcoe_ddp_xid; 1751#endif 1752#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 1753 struct netprio_map __rcu *priomap; 1754#endif 1755 struct phy_device *phydev; 1756 struct lock_class_key *qdisc_tx_busylock; 1757}; 1758#define to_net_dev(d) container_of(d, struct net_device, dev) 1759 1760#define NETDEV_ALIGN 32 1761 1762static inline 1763int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 1764{ 1765 return dev->prio_tc_map[prio & TC_BITMASK]; 1766} 1767 1768static inline 1769int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 1770{ 1771 if (tc >= dev->num_tc) 1772 return -EINVAL; 1773 1774 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 1775 return 0; 1776} 1777 1778static inline 1779void netdev_reset_tc(struct net_device *dev) 1780{ 1781 dev->num_tc = 0; 1782 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq)); 1783 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map)); 1784} 1785 1786static inline 1787int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset) 1788{ 1789 if (tc >= dev->num_tc) 1790 return -EINVAL; 1791 1792 dev->tc_to_txq[tc].count = count; 1793 dev->tc_to_txq[tc].offset = offset; 1794 return 0; 1795} 1796 1797static inline 1798int netdev_set_num_tc(struct net_device *dev, u8 num_tc) 1799{ 1800 if (num_tc > TC_MAX_QUEUE) 1801 return -EINVAL; 1802 1803 dev->num_tc = num_tc; 1804 return 0; 1805} 1806 1807static inline 1808int netdev_get_num_tc(struct net_device *dev) 1809{ 1810 return dev->num_tc; 1811} 1812 1813static inline 1814struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 1815 unsigned int index) 1816{ 1817 return &dev->_tx[index]; 1818} 1819 1820static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 1821 const struct sk_buff *skb) 1822{ 1823 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 1824} 1825 1826static inline void netdev_for_each_tx_queue(struct net_device *dev, 1827 void (*f)(struct net_device *, 1828 struct netdev_queue *, 1829 void *), 1830 void *arg) 1831{ 1832 unsigned int i; 1833 1834 for (i = 0; i < dev->num_tx_queues; i++) 1835 f(dev, &dev->_tx[i], arg); 1836} 1837 1838struct netdev_queue *netdev_pick_tx(struct net_device *dev, 1839 struct sk_buff *skb, 1840 void *accel_priv); 1841 1842/* 1843 * Net namespace inlines 1844 */ 1845static inline 1846struct net *dev_net(const struct net_device *dev) 1847{ 1848 return read_pnet(&dev->nd_net); 1849} 1850 1851static inline 1852void dev_net_set(struct net_device *dev, struct net *net) 1853{ 1854 write_pnet(&dev->nd_net, net); 1855} 1856 1857static inline bool netdev_uses_dsa(struct net_device *dev) 1858{ 1859#if IS_ENABLED(CONFIG_NET_DSA) 1860 if (dev->dsa_ptr != NULL) 1861 return dsa_uses_tagged_protocol(dev->dsa_ptr); 1862#endif 1863 return false; 1864} 1865 1866/** 1867 * netdev_priv - access network device private data 1868 * @dev: network device 1869 * 1870 * Get network device private data 1871 */ 1872static inline void *netdev_priv(const struct net_device *dev) 1873{ 1874 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 1875} 1876 1877/* Set the sysfs physical device reference for the network logical device 1878 * if set prior to registration will cause a symlink during initialization. 1879 */ 1880#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 1881 1882/* Set the sysfs device type for the network logical device to allow 1883 * fine-grained identification of different network device types. For 1884 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc. 1885 */ 1886#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 1887 1888/* Default NAPI poll() weight 1889 * Device drivers are strongly advised to not use bigger value 1890 */ 1891#define NAPI_POLL_WEIGHT 64 1892 1893/** 1894 * netif_napi_add - initialize a napi context 1895 * @dev: network device 1896 * @napi: napi context 1897 * @poll: polling function 1898 * @weight: default weight 1899 * 1900 * netif_napi_add() must be used to initialize a napi context prior to calling 1901 * *any* of the other napi related functions. 1902 */ 1903void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 1904 int (*poll)(struct napi_struct *, int), int weight); 1905 1906/** 1907 * netif_napi_del - remove a napi context 1908 * @napi: napi context 1909 * 1910 * netif_napi_del() removes a napi context from the network device napi list 1911 */ 1912void netif_napi_del(struct napi_struct *napi); 1913 1914struct napi_gro_cb { 1915 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 1916 void *frag0; 1917 1918 /* Length of frag0. */ 1919 unsigned int frag0_len; 1920 1921 /* This indicates where we are processing relative to skb->data. */ 1922 int data_offset; 1923 1924 /* This is non-zero if the packet cannot be merged with the new skb. */ 1925 u16 flush; 1926 1927 /* Save the IP ID here and check when we get to the transport layer */ 1928 u16 flush_id; 1929 1930 /* Number of segments aggregated. */ 1931 u16 count; 1932 1933 /* Start offset for remote checksum offload */ 1934 u16 gro_remcsum_start; 1935 1936 /* jiffies when first packet was created/queued */ 1937 unsigned long age; 1938 1939 /* Used in ipv6_gro_receive() and foo-over-udp */ 1940 u16 proto; 1941 1942 /* This is non-zero if the packet may be of the same flow. */ 1943 u8 same_flow:1; 1944 1945 /* Used in udp_gro_receive */ 1946 u8 udp_mark:1; 1947 1948 /* GRO checksum is valid */ 1949 u8 csum_valid:1; 1950 1951 /* Number of checksums via CHECKSUM_UNNECESSARY */ 1952 u8 csum_cnt:3; 1953 1954 /* Free the skb? */ 1955 u8 free:2; 1956#define NAPI_GRO_FREE 1 1957#define NAPI_GRO_FREE_STOLEN_HEAD 2 1958 1959 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 1960 u8 is_ipv6:1; 1961 1962 /* 7 bit hole */ 1963 1964 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 1965 __wsum csum; 1966 1967 /* used in skb_gro_receive() slow path */ 1968 struct sk_buff *last; 1969}; 1970 1971#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 1972 1973struct packet_type { 1974 __be16 type; /* This is really htons(ether_type). */ 1975 struct net_device *dev; /* NULL is wildcarded here */ 1976 int (*func) (struct sk_buff *, 1977 struct net_device *, 1978 struct packet_type *, 1979 struct net_device *); 1980 bool (*id_match)(struct packet_type *ptype, 1981 struct sock *sk); 1982 void *af_packet_priv; 1983 struct list_head list; 1984}; 1985 1986struct offload_callbacks { 1987 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 1988 netdev_features_t features); 1989 struct sk_buff **(*gro_receive)(struct sk_buff **head, 1990 struct sk_buff *skb); 1991 int (*gro_complete)(struct sk_buff *skb, int nhoff); 1992}; 1993 1994struct packet_offload { 1995 __be16 type; /* This is really htons(ether_type). */ 1996 struct offload_callbacks callbacks; 1997 struct list_head list; 1998}; 1999 2000struct udp_offload; 2001 2002struct udp_offload_callbacks { 2003 struct sk_buff **(*gro_receive)(struct sk_buff **head, 2004 struct sk_buff *skb, 2005 struct udp_offload *uoff); 2006 int (*gro_complete)(struct sk_buff *skb, 2007 int nhoff, 2008 struct udp_offload *uoff); 2009}; 2010 2011struct udp_offload { 2012 __be16 port; 2013 u8 ipproto; 2014 struct udp_offload_callbacks callbacks; 2015}; 2016 2017/* often modified stats are per cpu, other are shared (netdev->stats) */ 2018struct pcpu_sw_netstats { 2019 u64 rx_packets; 2020 u64 rx_bytes; 2021 u64 tx_packets; 2022 u64 tx_bytes; 2023 struct u64_stats_sync syncp; 2024}; 2025 2026#define netdev_alloc_pcpu_stats(type) \ 2027({ \ 2028 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \ 2029 if (pcpu_stats) { \ 2030 int __cpu; \ 2031 for_each_possible_cpu(__cpu) { \ 2032 typeof(type) *stat; \ 2033 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2034 u64_stats_init(&stat->syncp); \ 2035 } \ 2036 } \ 2037 pcpu_stats; \ 2038}) 2039 2040#include <linux/notifier.h> 2041 2042/* netdevice notifier chain. Please remember to update the rtnetlink 2043 * notification exclusion list in rtnetlink_event() when adding new 2044 * types. 2045 */ 2046#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */ 2047#define NETDEV_DOWN 0x0002 2048#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface 2049 detected a hardware crash and restarted 2050 - we can use this eg to kick tcp sessions 2051 once done */ 2052#define NETDEV_CHANGE 0x0004 /* Notify device state change */ 2053#define NETDEV_REGISTER 0x0005 2054#define NETDEV_UNREGISTER 0x0006 2055#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */ 2056#define NETDEV_CHANGEADDR 0x0008 2057#define NETDEV_GOING_DOWN 0x0009 2058#define NETDEV_CHANGENAME 0x000A 2059#define NETDEV_FEAT_CHANGE 0x000B 2060#define NETDEV_BONDING_FAILOVER 0x000C 2061#define NETDEV_PRE_UP 0x000D 2062#define NETDEV_PRE_TYPE_CHANGE 0x000E 2063#define NETDEV_POST_TYPE_CHANGE 0x000F 2064#define NETDEV_POST_INIT 0x0010 2065#define NETDEV_UNREGISTER_FINAL 0x0011 2066#define NETDEV_RELEASE 0x0012 2067#define NETDEV_NOTIFY_PEERS 0x0013 2068#define NETDEV_JOIN 0x0014 2069#define NETDEV_CHANGEUPPER 0x0015 2070#define NETDEV_RESEND_IGMP 0x0016 2071#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */ 2072#define NETDEV_CHANGEINFODATA 0x0018 2073#define NETDEV_BONDING_INFO 0x0019 2074 2075int register_netdevice_notifier(struct notifier_block *nb); 2076int unregister_netdevice_notifier(struct notifier_block *nb); 2077 2078struct netdev_notifier_info { 2079 struct net_device *dev; 2080}; 2081 2082struct netdev_notifier_change_info { 2083 struct netdev_notifier_info info; /* must be first */ 2084 unsigned int flags_changed; 2085}; 2086 2087static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2088 struct net_device *dev) 2089{ 2090 info->dev = dev; 2091} 2092 2093static inline struct net_device * 2094netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2095{ 2096 return info->dev; 2097} 2098 2099int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2100 2101 2102extern rwlock_t dev_base_lock; /* Device list lock */ 2103 2104#define for_each_netdev(net, d) \ 2105 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2106#define for_each_netdev_reverse(net, d) \ 2107 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2108#define for_each_netdev_rcu(net, d) \ 2109 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2110#define for_each_netdev_safe(net, d, n) \ 2111 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2112#define for_each_netdev_continue(net, d) \ 2113 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2114#define for_each_netdev_continue_rcu(net, d) \ 2115 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2116#define for_each_netdev_in_bond_rcu(bond, slave) \ 2117 for_each_netdev_rcu(&init_net, slave) \ 2118 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2119#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2120 2121static inline struct net_device *next_net_device(struct net_device *dev) 2122{ 2123 struct list_head *lh; 2124 struct net *net; 2125 2126 net = dev_net(dev); 2127 lh = dev->dev_list.next; 2128 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2129} 2130 2131static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2132{ 2133 struct list_head *lh; 2134 struct net *net; 2135 2136 net = dev_net(dev); 2137 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2138 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2139} 2140 2141static inline struct net_device *first_net_device(struct net *net) 2142{ 2143 return list_empty(&net->dev_base_head) ? NULL : 2144 net_device_entry(net->dev_base_head.next); 2145} 2146 2147static inline struct net_device *first_net_device_rcu(struct net *net) 2148{ 2149 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2150 2151 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2152} 2153 2154int netdev_boot_setup_check(struct net_device *dev); 2155unsigned long netdev_boot_base(const char *prefix, int unit); 2156struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2157 const char *hwaddr); 2158struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2159struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2160void dev_add_pack(struct packet_type *pt); 2161void dev_remove_pack(struct packet_type *pt); 2162void __dev_remove_pack(struct packet_type *pt); 2163void dev_add_offload(struct packet_offload *po); 2164void dev_remove_offload(struct packet_offload *po); 2165 2166int dev_get_iflink(const struct net_device *dev); 2167struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2168 unsigned short mask); 2169struct net_device *dev_get_by_name(struct net *net, const char *name); 2170struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2171struct net_device *__dev_get_by_name(struct net *net, const char *name); 2172int dev_alloc_name(struct net_device *dev, const char *name); 2173int dev_open(struct net_device *dev); 2174int dev_close(struct net_device *dev); 2175int dev_close_many(struct list_head *head, bool unlink); 2176void dev_disable_lro(struct net_device *dev); 2177int dev_loopback_xmit(struct sock *sk, struct sk_buff *newskb); 2178int dev_queue_xmit_sk(struct sock *sk, struct sk_buff *skb); 2179static inline int dev_queue_xmit(struct sk_buff *skb) 2180{ 2181 return dev_queue_xmit_sk(skb->sk, skb); 2182} 2183int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv); 2184int register_netdevice(struct net_device *dev); 2185void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2186void unregister_netdevice_many(struct list_head *head); 2187static inline void unregister_netdevice(struct net_device *dev) 2188{ 2189 unregister_netdevice_queue(dev, NULL); 2190} 2191 2192int netdev_refcnt_read(const struct net_device *dev); 2193void free_netdev(struct net_device *dev); 2194void netdev_freemem(struct net_device *dev); 2195void synchronize_net(void); 2196int init_dummy_netdev(struct net_device *dev); 2197 2198DECLARE_PER_CPU(int, xmit_recursion); 2199static inline int dev_recursion_level(void) 2200{ 2201 return this_cpu_read(xmit_recursion); 2202} 2203 2204struct net_device *dev_get_by_index(struct net *net, int ifindex); 2205struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2206struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2207int netdev_get_name(struct net *net, char *name, int ifindex); 2208int dev_restart(struct net_device *dev); 2209int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb); 2210 2211static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2212{ 2213 return NAPI_GRO_CB(skb)->data_offset; 2214} 2215 2216static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2217{ 2218 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2219} 2220 2221static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2222{ 2223 NAPI_GRO_CB(skb)->data_offset += len; 2224} 2225 2226static inline void *skb_gro_header_fast(struct sk_buff *skb, 2227 unsigned int offset) 2228{ 2229 return NAPI_GRO_CB(skb)->frag0 + offset; 2230} 2231 2232static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2233{ 2234 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2235} 2236 2237static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2238 unsigned int offset) 2239{ 2240 if (!pskb_may_pull(skb, hlen)) 2241 return NULL; 2242 2243 NAPI_GRO_CB(skb)->frag0 = NULL; 2244 NAPI_GRO_CB(skb)->frag0_len = 0; 2245 return skb->data + offset; 2246} 2247 2248static inline void *skb_gro_network_header(struct sk_buff *skb) 2249{ 2250 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2251 skb_network_offset(skb); 2252} 2253 2254static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2255 const void *start, unsigned int len) 2256{ 2257 if (NAPI_GRO_CB(skb)->csum_valid) 2258 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2259 csum_partial(start, len, 0)); 2260} 2261 2262/* GRO checksum functions. These are logical equivalents of the normal 2263 * checksum functions (in skbuff.h) except that they operate on the GRO 2264 * offsets and fields in sk_buff. 2265 */ 2266 2267__sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2268 2269static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2270{ 2271 return (NAPI_GRO_CB(skb)->gro_remcsum_start - skb_headroom(skb) == 2272 skb_gro_offset(skb)); 2273} 2274 2275static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2276 bool zero_okay, 2277 __sum16 check) 2278{ 2279 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2280 skb_checksum_start_offset(skb) < 2281 skb_gro_offset(skb)) && 2282 !skb_at_gro_remcsum_start(skb) && 2283 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2284 (!zero_okay || check)); 2285} 2286 2287static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2288 __wsum psum) 2289{ 2290 if (NAPI_GRO_CB(skb)->csum_valid && 2291 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2292 return 0; 2293 2294 NAPI_GRO_CB(skb)->csum = psum; 2295 2296 return __skb_gro_checksum_complete(skb); 2297} 2298 2299static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2300{ 2301 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2302 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2303 NAPI_GRO_CB(skb)->csum_cnt--; 2304 } else { 2305 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2306 * verified a new top level checksum or an encapsulated one 2307 * during GRO. This saves work if we fallback to normal path. 2308 */ 2309 __skb_incr_checksum_unnecessary(skb); 2310 } 2311} 2312 2313#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2314 compute_pseudo) \ 2315({ \ 2316 __sum16 __ret = 0; \ 2317 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2318 __ret = __skb_gro_checksum_validate_complete(skb, \ 2319 compute_pseudo(skb, proto)); \ 2320 if (__ret) \ 2321 __skb_mark_checksum_bad(skb); \ 2322 else \ 2323 skb_gro_incr_csum_unnecessary(skb); \ 2324 __ret; \ 2325}) 2326 2327#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2328 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2329 2330#define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2331 compute_pseudo) \ 2332 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2333 2334#define skb_gro_checksum_simple_validate(skb) \ 2335 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2336 2337static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2338{ 2339 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2340 !NAPI_GRO_CB(skb)->csum_valid); 2341} 2342 2343static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2344 __sum16 check, __wsum pseudo) 2345{ 2346 NAPI_GRO_CB(skb)->csum = ~pseudo; 2347 NAPI_GRO_CB(skb)->csum_valid = 1; 2348} 2349 2350#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \ 2351do { \ 2352 if (__skb_gro_checksum_convert_check(skb)) \ 2353 __skb_gro_checksum_convert(skb, check, \ 2354 compute_pseudo(skb, proto)); \ 2355} while (0) 2356 2357struct gro_remcsum { 2358 int offset; 2359 __wsum delta; 2360}; 2361 2362static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2363{ 2364 grc->offset = 0; 2365 grc->delta = 0; 2366} 2367 2368static inline void skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2369 int start, int offset, 2370 struct gro_remcsum *grc, 2371 bool nopartial) 2372{ 2373 __wsum delta; 2374 2375 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2376 2377 if (!nopartial) { 2378 NAPI_GRO_CB(skb)->gro_remcsum_start = 2379 ((unsigned char *)ptr + start) - skb->head; 2380 return; 2381 } 2382 2383 delta = remcsum_adjust(ptr, NAPI_GRO_CB(skb)->csum, start, offset); 2384 2385 /* Adjust skb->csum since we changed the packet */ 2386 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2387 2388 grc->offset = (ptr + offset) - (void *)skb->head; 2389 grc->delta = delta; 2390} 2391 2392static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2393 struct gro_remcsum *grc) 2394{ 2395 if (!grc->delta) 2396 return; 2397 2398 remcsum_unadjust((__sum16 *)(skb->head + grc->offset), grc->delta); 2399} 2400 2401static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2402 unsigned short type, 2403 const void *daddr, const void *saddr, 2404 unsigned int len) 2405{ 2406 if (!dev->header_ops || !dev->header_ops->create) 2407 return 0; 2408 2409 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2410} 2411 2412static inline int dev_parse_header(const struct sk_buff *skb, 2413 unsigned char *haddr) 2414{ 2415 const struct net_device *dev = skb->dev; 2416 2417 if (!dev->header_ops || !dev->header_ops->parse) 2418 return 0; 2419 return dev->header_ops->parse(skb, haddr); 2420} 2421 2422typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len); 2423int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2424static inline int unregister_gifconf(unsigned int family) 2425{ 2426 return register_gifconf(family, NULL); 2427} 2428 2429#ifdef CONFIG_NET_FLOW_LIMIT 2430#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2431struct sd_flow_limit { 2432 u64 count; 2433 unsigned int num_buckets; 2434 unsigned int history_head; 2435 u16 history[FLOW_LIMIT_HISTORY]; 2436 u8 buckets[]; 2437}; 2438 2439extern int netdev_flow_limit_table_len; 2440#endif /* CONFIG_NET_FLOW_LIMIT */ 2441 2442/* 2443 * Incoming packets are placed on per-cpu queues 2444 */ 2445struct softnet_data { 2446 struct list_head poll_list; 2447 struct sk_buff_head process_queue; 2448 2449 /* stats */ 2450 unsigned int processed; 2451 unsigned int time_squeeze; 2452 unsigned int cpu_collision; 2453 unsigned int received_rps; 2454#ifdef CONFIG_RPS 2455 struct softnet_data *rps_ipi_list; 2456#endif 2457#ifdef CONFIG_NET_FLOW_LIMIT 2458 struct sd_flow_limit __rcu *flow_limit; 2459#endif 2460 struct Qdisc *output_queue; 2461 struct Qdisc **output_queue_tailp; 2462 struct sk_buff *completion_queue; 2463 2464#ifdef CONFIG_RPS 2465 /* Elements below can be accessed between CPUs for RPS */ 2466 struct call_single_data csd ____cacheline_aligned_in_smp; 2467 struct softnet_data *rps_ipi_next; 2468 unsigned int cpu; 2469 unsigned int input_queue_head; 2470 unsigned int input_queue_tail; 2471#endif 2472 unsigned int dropped; 2473 struct sk_buff_head input_pkt_queue; 2474 struct napi_struct backlog; 2475 2476}; 2477 2478static inline void input_queue_head_incr(struct softnet_data *sd) 2479{ 2480#ifdef CONFIG_RPS 2481 sd->input_queue_head++; 2482#endif 2483} 2484 2485static inline void input_queue_tail_incr_save(struct softnet_data *sd, 2486 unsigned int *qtail) 2487{ 2488#ifdef CONFIG_RPS 2489 *qtail = ++sd->input_queue_tail; 2490#endif 2491} 2492 2493DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 2494 2495void __netif_schedule(struct Qdisc *q); 2496void netif_schedule_queue(struct netdev_queue *txq); 2497 2498static inline void netif_tx_schedule_all(struct net_device *dev) 2499{ 2500 unsigned int i; 2501 2502 for (i = 0; i < dev->num_tx_queues; i++) 2503 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 2504} 2505 2506static inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 2507{ 2508 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2509} 2510 2511/** 2512 * netif_start_queue - allow transmit 2513 * @dev: network device 2514 * 2515 * Allow upper layers to call the device hard_start_xmit routine. 2516 */ 2517static inline void netif_start_queue(struct net_device *dev) 2518{ 2519 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 2520} 2521 2522static inline void netif_tx_start_all_queues(struct net_device *dev) 2523{ 2524 unsigned int i; 2525 2526 for (i = 0; i < dev->num_tx_queues; i++) { 2527 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2528 netif_tx_start_queue(txq); 2529 } 2530} 2531 2532void netif_tx_wake_queue(struct netdev_queue *dev_queue); 2533 2534/** 2535 * netif_wake_queue - restart transmit 2536 * @dev: network device 2537 * 2538 * Allow upper layers to call the device hard_start_xmit routine. 2539 * Used for flow control when transmit resources are available. 2540 */ 2541static inline void netif_wake_queue(struct net_device *dev) 2542{ 2543 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 2544} 2545 2546static inline void netif_tx_wake_all_queues(struct net_device *dev) 2547{ 2548 unsigned int i; 2549 2550 for (i = 0; i < dev->num_tx_queues; i++) { 2551 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2552 netif_tx_wake_queue(txq); 2553 } 2554} 2555 2556static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 2557{ 2558 if (WARN_ON(!dev_queue)) { 2559 pr_info("netif_stop_queue() cannot be called before register_netdev()\n"); 2560 return; 2561 } 2562 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2563} 2564 2565/** 2566 * netif_stop_queue - stop transmitted packets 2567 * @dev: network device 2568 * 2569 * Stop upper layers calling the device hard_start_xmit routine. 2570 * Used for flow control when transmit resources are unavailable. 2571 */ 2572static inline void netif_stop_queue(struct net_device *dev) 2573{ 2574 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 2575} 2576 2577static inline void netif_tx_stop_all_queues(struct net_device *dev) 2578{ 2579 unsigned int i; 2580 2581 for (i = 0; i < dev->num_tx_queues; i++) { 2582 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2583 netif_tx_stop_queue(txq); 2584 } 2585} 2586 2587static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 2588{ 2589 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2590} 2591 2592/** 2593 * netif_queue_stopped - test if transmit queue is flowblocked 2594 * @dev: network device 2595 * 2596 * Test if transmit queue on device is currently unable to send. 2597 */ 2598static inline bool netif_queue_stopped(const struct net_device *dev) 2599{ 2600 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 2601} 2602 2603static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 2604{ 2605 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 2606} 2607 2608static inline bool 2609netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 2610{ 2611 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 2612} 2613 2614static inline bool 2615netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 2616{ 2617 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 2618} 2619 2620/** 2621 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 2622 * @dev_queue: pointer to transmit queue 2623 * 2624 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 2625 * to give appropriate hint to the cpu. 2626 */ 2627static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 2628{ 2629#ifdef CONFIG_BQL 2630 prefetchw(&dev_queue->dql.num_queued); 2631#endif 2632} 2633 2634/** 2635 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 2636 * @dev_queue: pointer to transmit queue 2637 * 2638 * BQL enabled drivers might use this helper in their TX completion path, 2639 * to give appropriate hint to the cpu. 2640 */ 2641static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 2642{ 2643#ifdef CONFIG_BQL 2644 prefetchw(&dev_queue->dql.limit); 2645#endif 2646} 2647 2648static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 2649 unsigned int bytes) 2650{ 2651#ifdef CONFIG_BQL 2652 dql_queued(&dev_queue->dql, bytes); 2653 2654 if (likely(dql_avail(&dev_queue->dql) >= 0)) 2655 return; 2656 2657 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2658 2659 /* 2660 * The XOFF flag must be set before checking the dql_avail below, 2661 * because in netdev_tx_completed_queue we update the dql_completed 2662 * before checking the XOFF flag. 2663 */ 2664 smp_mb(); 2665 2666 /* check again in case another CPU has just made room avail */ 2667 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 2668 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2669#endif 2670} 2671 2672/** 2673 * netdev_sent_queue - report the number of bytes queued to hardware 2674 * @dev: network device 2675 * @bytes: number of bytes queued to the hardware device queue 2676 * 2677 * Report the number of bytes queued for sending/completion to the network 2678 * device hardware queue. @bytes should be a good approximation and should 2679 * exactly match netdev_completed_queue() @bytes 2680 */ 2681static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 2682{ 2683 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 2684} 2685 2686static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 2687 unsigned int pkts, unsigned int bytes) 2688{ 2689#ifdef CONFIG_BQL 2690 if (unlikely(!bytes)) 2691 return; 2692 2693 dql_completed(&dev_queue->dql, bytes); 2694 2695 /* 2696 * Without the memory barrier there is a small possiblity that 2697 * netdev_tx_sent_queue will miss the update and cause the queue to 2698 * be stopped forever 2699 */ 2700 smp_mb(); 2701 2702 if (dql_avail(&dev_queue->dql) < 0) 2703 return; 2704 2705 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 2706 netif_schedule_queue(dev_queue); 2707#endif 2708} 2709 2710/** 2711 * netdev_completed_queue - report bytes and packets completed by device 2712 * @dev: network device 2713 * @pkts: actual number of packets sent over the medium 2714 * @bytes: actual number of bytes sent over the medium 2715 * 2716 * Report the number of bytes and packets transmitted by the network device 2717 * hardware queue over the physical medium, @bytes must exactly match the 2718 * @bytes amount passed to netdev_sent_queue() 2719 */ 2720static inline void netdev_completed_queue(struct net_device *dev, 2721 unsigned int pkts, unsigned int bytes) 2722{ 2723 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 2724} 2725 2726static inline void netdev_tx_reset_queue(struct netdev_queue *q) 2727{ 2728#ifdef CONFIG_BQL 2729 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 2730 dql_reset(&q->dql); 2731#endif 2732} 2733 2734/** 2735 * netdev_reset_queue - reset the packets and bytes count of a network device 2736 * @dev_queue: network device 2737 * 2738 * Reset the bytes and packet count of a network device and clear the 2739 * software flow control OFF bit for this network device 2740 */ 2741static inline void netdev_reset_queue(struct net_device *dev_queue) 2742{ 2743 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 2744} 2745 2746/** 2747 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 2748 * @dev: network device 2749 * @queue_index: given tx queue index 2750 * 2751 * Returns 0 if given tx queue index >= number of device tx queues, 2752 * otherwise returns the originally passed tx queue index. 2753 */ 2754static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 2755{ 2756 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 2757 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 2758 dev->name, queue_index, 2759 dev->real_num_tx_queues); 2760 return 0; 2761 } 2762 2763 return queue_index; 2764} 2765 2766/** 2767 * netif_running - test if up 2768 * @dev: network device 2769 * 2770 * Test if the device has been brought up. 2771 */ 2772static inline bool netif_running(const struct net_device *dev) 2773{ 2774 return test_bit(__LINK_STATE_START, &dev->state); 2775} 2776 2777/* 2778 * Routines to manage the subqueues on a device. We only need start 2779 * stop, and a check if it's stopped. All other device management is 2780 * done at the overall netdevice level. 2781 * Also test the device if we're multiqueue. 2782 */ 2783 2784/** 2785 * netif_start_subqueue - allow sending packets on subqueue 2786 * @dev: network device 2787 * @queue_index: sub queue index 2788 * 2789 * Start individual transmit queue of a device with multiple transmit queues. 2790 */ 2791static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 2792{ 2793 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2794 2795 netif_tx_start_queue(txq); 2796} 2797 2798/** 2799 * netif_stop_subqueue - stop sending packets on subqueue 2800 * @dev: network device 2801 * @queue_index: sub queue index 2802 * 2803 * Stop individual transmit queue of a device with multiple transmit queues. 2804 */ 2805static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 2806{ 2807 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2808 netif_tx_stop_queue(txq); 2809} 2810 2811/** 2812 * netif_subqueue_stopped - test status of subqueue 2813 * @dev: network device 2814 * @queue_index: sub queue index 2815 * 2816 * Check individual transmit queue of a device with multiple transmit queues. 2817 */ 2818static inline bool __netif_subqueue_stopped(const struct net_device *dev, 2819 u16 queue_index) 2820{ 2821 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2822 2823 return netif_tx_queue_stopped(txq); 2824} 2825 2826static inline bool netif_subqueue_stopped(const struct net_device *dev, 2827 struct sk_buff *skb) 2828{ 2829 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 2830} 2831 2832void netif_wake_subqueue(struct net_device *dev, u16 queue_index); 2833 2834#ifdef CONFIG_XPS 2835int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 2836 u16 index); 2837#else 2838static inline int netif_set_xps_queue(struct net_device *dev, 2839 const struct cpumask *mask, 2840 u16 index) 2841{ 2842 return 0; 2843} 2844#endif 2845 2846/* 2847 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used 2848 * as a distribution range limit for the returned value. 2849 */ 2850static inline u16 skb_tx_hash(const struct net_device *dev, 2851 struct sk_buff *skb) 2852{ 2853 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues); 2854} 2855 2856/** 2857 * netif_is_multiqueue - test if device has multiple transmit queues 2858 * @dev: network device 2859 * 2860 * Check if device has multiple transmit queues 2861 */ 2862static inline bool netif_is_multiqueue(const struct net_device *dev) 2863{ 2864 return dev->num_tx_queues > 1; 2865} 2866 2867int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 2868 2869#ifdef CONFIG_SYSFS 2870int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 2871#else 2872static inline int netif_set_real_num_rx_queues(struct net_device *dev, 2873 unsigned int rxq) 2874{ 2875 return 0; 2876} 2877#endif 2878 2879#ifdef CONFIG_SYSFS 2880static inline unsigned int get_netdev_rx_queue_index( 2881 struct netdev_rx_queue *queue) 2882{ 2883 struct net_device *dev = queue->dev; 2884 int index = queue - dev->_rx; 2885 2886 BUG_ON(index >= dev->num_rx_queues); 2887 return index; 2888} 2889#endif 2890 2891#define DEFAULT_MAX_NUM_RSS_QUEUES (8) 2892int netif_get_num_default_rss_queues(void); 2893 2894enum skb_free_reason { 2895 SKB_REASON_CONSUMED, 2896 SKB_REASON_DROPPED, 2897}; 2898 2899void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 2900void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 2901 2902/* 2903 * It is not allowed to call kfree_skb() or consume_skb() from hardware 2904 * interrupt context or with hardware interrupts being disabled. 2905 * (in_irq() || irqs_disabled()) 2906 * 2907 * We provide four helpers that can be used in following contexts : 2908 * 2909 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 2910 * replacing kfree_skb(skb) 2911 * 2912 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 2913 * Typically used in place of consume_skb(skb) in TX completion path 2914 * 2915 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 2916 * replacing kfree_skb(skb) 2917 * 2918 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 2919 * and consumed a packet. Used in place of consume_skb(skb) 2920 */ 2921static inline void dev_kfree_skb_irq(struct sk_buff *skb) 2922{ 2923 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 2924} 2925 2926static inline void dev_consume_skb_irq(struct sk_buff *skb) 2927{ 2928 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 2929} 2930 2931static inline void dev_kfree_skb_any(struct sk_buff *skb) 2932{ 2933 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 2934} 2935 2936static inline void dev_consume_skb_any(struct sk_buff *skb) 2937{ 2938 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 2939} 2940 2941int netif_rx(struct sk_buff *skb); 2942int netif_rx_ni(struct sk_buff *skb); 2943int netif_receive_skb_sk(struct sock *sk, struct sk_buff *skb); 2944static inline int netif_receive_skb(struct sk_buff *skb) 2945{ 2946 return netif_receive_skb_sk(skb->sk, skb); 2947} 2948gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 2949void napi_gro_flush(struct napi_struct *napi, bool flush_old); 2950struct sk_buff *napi_get_frags(struct napi_struct *napi); 2951gro_result_t napi_gro_frags(struct napi_struct *napi); 2952struct packet_offload *gro_find_receive_by_type(__be16 type); 2953struct packet_offload *gro_find_complete_by_type(__be16 type); 2954 2955static inline void napi_free_frags(struct napi_struct *napi) 2956{ 2957 kfree_skb(napi->skb); 2958 napi->skb = NULL; 2959} 2960 2961int netdev_rx_handler_register(struct net_device *dev, 2962 rx_handler_func_t *rx_handler, 2963 void *rx_handler_data); 2964void netdev_rx_handler_unregister(struct net_device *dev); 2965 2966bool dev_valid_name(const char *name); 2967int dev_ioctl(struct net *net, unsigned int cmd, void __user *); 2968int dev_ethtool(struct net *net, struct ifreq *); 2969unsigned int dev_get_flags(const struct net_device *); 2970int __dev_change_flags(struct net_device *, unsigned int flags); 2971int dev_change_flags(struct net_device *, unsigned int); 2972void __dev_notify_flags(struct net_device *, unsigned int old_flags, 2973 unsigned int gchanges); 2974int dev_change_name(struct net_device *, const char *); 2975int dev_set_alias(struct net_device *, const char *, size_t); 2976int dev_change_net_namespace(struct net_device *, struct net *, const char *); 2977int dev_set_mtu(struct net_device *, int); 2978void dev_set_group(struct net_device *, int); 2979int dev_set_mac_address(struct net_device *, struct sockaddr *); 2980int dev_change_carrier(struct net_device *, bool new_carrier); 2981int dev_get_phys_port_id(struct net_device *dev, 2982 struct netdev_phys_item_id *ppid); 2983int dev_get_phys_port_name(struct net_device *dev, 2984 char *name, size_t len); 2985struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev); 2986struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 2987 struct netdev_queue *txq, int *ret); 2988int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 2989int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 2990bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb); 2991 2992extern int netdev_budget; 2993 2994/* Called by rtnetlink.c:rtnl_unlock() */ 2995void netdev_run_todo(void); 2996 2997/** 2998 * dev_put - release reference to device 2999 * @dev: network device 3000 * 3001 * Release reference to device to allow it to be freed. 3002 */ 3003static inline void dev_put(struct net_device *dev) 3004{ 3005 this_cpu_dec(*dev->pcpu_refcnt); 3006} 3007 3008/** 3009 * dev_hold - get reference to device 3010 * @dev: network device 3011 * 3012 * Hold reference to device to keep it from being freed. 3013 */ 3014static inline void dev_hold(struct net_device *dev) 3015{ 3016 this_cpu_inc(*dev->pcpu_refcnt); 3017} 3018 3019/* Carrier loss detection, dial on demand. The functions netif_carrier_on 3020 * and _off may be called from IRQ context, but it is caller 3021 * who is responsible for serialization of these calls. 3022 * 3023 * The name carrier is inappropriate, these functions should really be 3024 * called netif_lowerlayer_*() because they represent the state of any 3025 * kind of lower layer not just hardware media. 3026 */ 3027 3028void linkwatch_init_dev(struct net_device *dev); 3029void linkwatch_fire_event(struct net_device *dev); 3030void linkwatch_forget_dev(struct net_device *dev); 3031 3032/** 3033 * netif_carrier_ok - test if carrier present 3034 * @dev: network device 3035 * 3036 * Check if carrier is present on device 3037 */ 3038static inline bool netif_carrier_ok(const struct net_device *dev) 3039{ 3040 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3041} 3042 3043unsigned long dev_trans_start(struct net_device *dev); 3044 3045void __netdev_watchdog_up(struct net_device *dev); 3046 3047void netif_carrier_on(struct net_device *dev); 3048 3049void netif_carrier_off(struct net_device *dev); 3050 3051/** 3052 * netif_dormant_on - mark device as dormant. 3053 * @dev: network device 3054 * 3055 * Mark device as dormant (as per RFC2863). 3056 * 3057 * The dormant state indicates that the relevant interface is not 3058 * actually in a condition to pass packets (i.e., it is not 'up') but is 3059 * in a "pending" state, waiting for some external event. For "on- 3060 * demand" interfaces, this new state identifies the situation where the 3061 * interface is waiting for events to place it in the up state. 3062 * 3063 */ 3064static inline void netif_dormant_on(struct net_device *dev) 3065{ 3066 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3067 linkwatch_fire_event(dev); 3068} 3069 3070/** 3071 * netif_dormant_off - set device as not dormant. 3072 * @dev: network device 3073 * 3074 * Device is not in dormant state. 3075 */ 3076static inline void netif_dormant_off(struct net_device *dev) 3077{ 3078 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3079 linkwatch_fire_event(dev); 3080} 3081 3082/** 3083 * netif_dormant - test if carrier present 3084 * @dev: network device 3085 * 3086 * Check if carrier is present on device 3087 */ 3088static inline bool netif_dormant(const struct net_device *dev) 3089{ 3090 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3091} 3092 3093 3094/** 3095 * netif_oper_up - test if device is operational 3096 * @dev: network device 3097 * 3098 * Check if carrier is operational 3099 */ 3100static inline bool netif_oper_up(const struct net_device *dev) 3101{ 3102 return (dev->operstate == IF_OPER_UP || 3103 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3104} 3105 3106/** 3107 * netif_device_present - is device available or removed 3108 * @dev: network device 3109 * 3110 * Check if device has not been removed from system. 3111 */ 3112static inline bool netif_device_present(struct net_device *dev) 3113{ 3114 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3115} 3116 3117void netif_device_detach(struct net_device *dev); 3118 3119void netif_device_attach(struct net_device *dev); 3120 3121/* 3122 * Network interface message level settings 3123 */ 3124 3125enum { 3126 NETIF_MSG_DRV = 0x0001, 3127 NETIF_MSG_PROBE = 0x0002, 3128 NETIF_MSG_LINK = 0x0004, 3129 NETIF_MSG_TIMER = 0x0008, 3130 NETIF_MSG_IFDOWN = 0x0010, 3131 NETIF_MSG_IFUP = 0x0020, 3132 NETIF_MSG_RX_ERR = 0x0040, 3133 NETIF_MSG_TX_ERR = 0x0080, 3134 NETIF_MSG_TX_QUEUED = 0x0100, 3135 NETIF_MSG_INTR = 0x0200, 3136 NETIF_MSG_TX_DONE = 0x0400, 3137 NETIF_MSG_RX_STATUS = 0x0800, 3138 NETIF_MSG_PKTDATA = 0x1000, 3139 NETIF_MSG_HW = 0x2000, 3140 NETIF_MSG_WOL = 0x4000, 3141}; 3142 3143#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3144#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3145#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3146#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3147#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3148#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3149#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3150#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3151#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3152#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3153#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3154#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3155#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3156#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3157#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3158 3159static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3160{ 3161 /* use default */ 3162 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3163 return default_msg_enable_bits; 3164 if (debug_value == 0) /* no output */ 3165 return 0; 3166 /* set low N bits */ 3167 return (1 << debug_value) - 1; 3168} 3169 3170static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 3171{ 3172 spin_lock(&txq->_xmit_lock); 3173 txq->xmit_lock_owner = cpu; 3174} 3175 3176static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 3177{ 3178 spin_lock_bh(&txq->_xmit_lock); 3179 txq->xmit_lock_owner = smp_processor_id(); 3180} 3181 3182static inline bool __netif_tx_trylock(struct netdev_queue *txq) 3183{ 3184 bool ok = spin_trylock(&txq->_xmit_lock); 3185 if (likely(ok)) 3186 txq->xmit_lock_owner = smp_processor_id(); 3187 return ok; 3188} 3189 3190static inline void __netif_tx_unlock(struct netdev_queue *txq) 3191{ 3192 txq->xmit_lock_owner = -1; 3193 spin_unlock(&txq->_xmit_lock); 3194} 3195 3196static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 3197{ 3198 txq->xmit_lock_owner = -1; 3199 spin_unlock_bh(&txq->_xmit_lock); 3200} 3201 3202static inline void txq_trans_update(struct netdev_queue *txq) 3203{ 3204 if (txq->xmit_lock_owner != -1) 3205 txq->trans_start = jiffies; 3206} 3207 3208/** 3209 * netif_tx_lock - grab network device transmit lock 3210 * @dev: network device 3211 * 3212 * Get network device transmit lock 3213 */ 3214static inline void netif_tx_lock(struct net_device *dev) 3215{ 3216 unsigned int i; 3217 int cpu; 3218 3219 spin_lock(&dev->tx_global_lock); 3220 cpu = smp_processor_id(); 3221 for (i = 0; i < dev->num_tx_queues; i++) { 3222 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3223 3224 /* We are the only thread of execution doing a 3225 * freeze, but we have to grab the _xmit_lock in 3226 * order to synchronize with threads which are in 3227 * the ->hard_start_xmit() handler and already 3228 * checked the frozen bit. 3229 */ 3230 __netif_tx_lock(txq, cpu); 3231 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 3232 __netif_tx_unlock(txq); 3233 } 3234} 3235 3236static inline void netif_tx_lock_bh(struct net_device *dev) 3237{ 3238 local_bh_disable(); 3239 netif_tx_lock(dev); 3240} 3241 3242static inline void netif_tx_unlock(struct net_device *dev) 3243{ 3244 unsigned int i; 3245 3246 for (i = 0; i < dev->num_tx_queues; i++) { 3247 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3248 3249 /* No need to grab the _xmit_lock here. If the 3250 * queue is not stopped for another reason, we 3251 * force a schedule. 3252 */ 3253 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 3254 netif_schedule_queue(txq); 3255 } 3256 spin_unlock(&dev->tx_global_lock); 3257} 3258 3259static inline void netif_tx_unlock_bh(struct net_device *dev) 3260{ 3261 netif_tx_unlock(dev); 3262 local_bh_enable(); 3263} 3264 3265#define HARD_TX_LOCK(dev, txq, cpu) { \ 3266 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3267 __netif_tx_lock(txq, cpu); \ 3268 } \ 3269} 3270 3271#define HARD_TX_TRYLOCK(dev, txq) \ 3272 (((dev->features & NETIF_F_LLTX) == 0) ? \ 3273 __netif_tx_trylock(txq) : \ 3274 true ) 3275 3276#define HARD_TX_UNLOCK(dev, txq) { \ 3277 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3278 __netif_tx_unlock(txq); \ 3279 } \ 3280} 3281 3282static inline void netif_tx_disable(struct net_device *dev) 3283{ 3284 unsigned int i; 3285 int cpu; 3286 3287 local_bh_disable(); 3288 cpu = smp_processor_id(); 3289 for (i = 0; i < dev->num_tx_queues; i++) { 3290 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3291 3292 __netif_tx_lock(txq, cpu); 3293 netif_tx_stop_queue(txq); 3294 __netif_tx_unlock(txq); 3295 } 3296 local_bh_enable(); 3297} 3298 3299static inline void netif_addr_lock(struct net_device *dev) 3300{ 3301 spin_lock(&dev->addr_list_lock); 3302} 3303 3304static inline void netif_addr_lock_nested(struct net_device *dev) 3305{ 3306 int subclass = SINGLE_DEPTH_NESTING; 3307 3308 if (dev->netdev_ops->ndo_get_lock_subclass) 3309 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev); 3310 3311 spin_lock_nested(&dev->addr_list_lock, subclass); 3312} 3313 3314static inline void netif_addr_lock_bh(struct net_device *dev) 3315{ 3316 spin_lock_bh(&dev->addr_list_lock); 3317} 3318 3319static inline void netif_addr_unlock(struct net_device *dev) 3320{ 3321 spin_unlock(&dev->addr_list_lock); 3322} 3323 3324static inline void netif_addr_unlock_bh(struct net_device *dev) 3325{ 3326 spin_unlock_bh(&dev->addr_list_lock); 3327} 3328 3329/* 3330 * dev_addrs walker. Should be used only for read access. Call with 3331 * rcu_read_lock held. 3332 */ 3333#define for_each_dev_addr(dev, ha) \ 3334 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 3335 3336/* These functions live elsewhere (drivers/net/net_init.c, but related) */ 3337 3338void ether_setup(struct net_device *dev); 3339 3340/* Support for loadable net-drivers */ 3341struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 3342 unsigned char name_assign_type, 3343 void (*setup)(struct net_device *), 3344 unsigned int txqs, unsigned int rxqs); 3345#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 3346 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 3347 3348#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 3349 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 3350 count) 3351 3352int register_netdev(struct net_device *dev); 3353void unregister_netdev(struct net_device *dev); 3354 3355/* General hardware address lists handling functions */ 3356int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 3357 struct netdev_hw_addr_list *from_list, int addr_len); 3358void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 3359 struct netdev_hw_addr_list *from_list, int addr_len); 3360int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 3361 struct net_device *dev, 3362 int (*sync)(struct net_device *, const unsigned char *), 3363 int (*unsync)(struct net_device *, 3364 const unsigned char *)); 3365void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 3366 struct net_device *dev, 3367 int (*unsync)(struct net_device *, 3368 const unsigned char *)); 3369void __hw_addr_init(struct netdev_hw_addr_list *list); 3370 3371/* Functions used for device addresses handling */ 3372int dev_addr_add(struct net_device *dev, const unsigned char *addr, 3373 unsigned char addr_type); 3374int dev_addr_del(struct net_device *dev, const unsigned char *addr, 3375 unsigned char addr_type); 3376void dev_addr_flush(struct net_device *dev); 3377int dev_addr_init(struct net_device *dev); 3378 3379/* Functions used for unicast addresses handling */ 3380int dev_uc_add(struct net_device *dev, const unsigned char *addr); 3381int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 3382int dev_uc_del(struct net_device *dev, const unsigned char *addr); 3383int dev_uc_sync(struct net_device *to, struct net_device *from); 3384int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 3385void dev_uc_unsync(struct net_device *to, struct net_device *from); 3386void dev_uc_flush(struct net_device *dev); 3387void dev_uc_init(struct net_device *dev); 3388 3389/** 3390 * __dev_uc_sync - Synchonize device's unicast list 3391 * @dev: device to sync 3392 * @sync: function to call if address should be added 3393 * @unsync: function to call if address should be removed 3394 * 3395 * Add newly added addresses to the interface, and release 3396 * addresses that have been deleted. 3397 **/ 3398static inline int __dev_uc_sync(struct net_device *dev, 3399 int (*sync)(struct net_device *, 3400 const unsigned char *), 3401 int (*unsync)(struct net_device *, 3402 const unsigned char *)) 3403{ 3404 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 3405} 3406 3407/** 3408 * __dev_uc_unsync - Remove synchronized addresses from device 3409 * @dev: device to sync 3410 * @unsync: function to call if address should be removed 3411 * 3412 * Remove all addresses that were added to the device by dev_uc_sync(). 3413 **/ 3414static inline void __dev_uc_unsync(struct net_device *dev, 3415 int (*unsync)(struct net_device *, 3416 const unsigned char *)) 3417{ 3418 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 3419} 3420 3421/* Functions used for multicast addresses handling */ 3422int dev_mc_add(struct net_device *dev, const unsigned char *addr); 3423int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 3424int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 3425int dev_mc_del(struct net_device *dev, const unsigned char *addr); 3426int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 3427int dev_mc_sync(struct net_device *to, struct net_device *from); 3428int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 3429void dev_mc_unsync(struct net_device *to, struct net_device *from); 3430void dev_mc_flush(struct net_device *dev); 3431void dev_mc_init(struct net_device *dev); 3432 3433/** 3434 * __dev_mc_sync - Synchonize device's multicast list 3435 * @dev: device to sync 3436 * @sync: function to call if address should be added 3437 * @unsync: function to call if address should be removed 3438 * 3439 * Add newly added addresses to the interface, and release 3440 * addresses that have been deleted. 3441 **/ 3442static inline int __dev_mc_sync(struct net_device *dev, 3443 int (*sync)(struct net_device *, 3444 const unsigned char *), 3445 int (*unsync)(struct net_device *, 3446 const unsigned char *)) 3447{ 3448 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 3449} 3450 3451/** 3452 * __dev_mc_unsync - Remove synchronized addresses from device 3453 * @dev: device to sync 3454 * @unsync: function to call if address should be removed 3455 * 3456 * Remove all addresses that were added to the device by dev_mc_sync(). 3457 **/ 3458static inline void __dev_mc_unsync(struct net_device *dev, 3459 int (*unsync)(struct net_device *, 3460 const unsigned char *)) 3461{ 3462 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 3463} 3464 3465/* Functions used for secondary unicast and multicast support */ 3466void dev_set_rx_mode(struct net_device *dev); 3467void __dev_set_rx_mode(struct net_device *dev); 3468int dev_set_promiscuity(struct net_device *dev, int inc); 3469int dev_set_allmulti(struct net_device *dev, int inc); 3470void netdev_state_change(struct net_device *dev); 3471void netdev_notify_peers(struct net_device *dev); 3472void netdev_features_change(struct net_device *dev); 3473/* Load a device via the kmod */ 3474void dev_load(struct net *net, const char *name); 3475struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 3476 struct rtnl_link_stats64 *storage); 3477void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 3478 const struct net_device_stats *netdev_stats); 3479 3480extern int netdev_max_backlog; 3481extern int netdev_tstamp_prequeue; 3482extern int weight_p; 3483extern int bpf_jit_enable; 3484 3485bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 3486struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 3487 struct list_head **iter); 3488struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 3489 struct list_head **iter); 3490 3491/* iterate through upper list, must be called under RCU read lock */ 3492#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 3493 for (iter = &(dev)->adj_list.upper, \ 3494 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 3495 updev; \ 3496 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 3497 3498/* iterate through upper list, must be called under RCU read lock */ 3499#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \ 3500 for (iter = &(dev)->all_adj_list.upper, \ 3501 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \ 3502 updev; \ 3503 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter))) 3504 3505void *netdev_lower_get_next_private(struct net_device *dev, 3506 struct list_head **iter); 3507void *netdev_lower_get_next_private_rcu(struct net_device *dev, 3508 struct list_head **iter); 3509 3510#define netdev_for_each_lower_private(dev, priv, iter) \ 3511 for (iter = (dev)->adj_list.lower.next, \ 3512 priv = netdev_lower_get_next_private(dev, &(iter)); \ 3513 priv; \ 3514 priv = netdev_lower_get_next_private(dev, &(iter))) 3515 3516#define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 3517 for (iter = &(dev)->adj_list.lower, \ 3518 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 3519 priv; \ 3520 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 3521 3522void *netdev_lower_get_next(struct net_device *dev, 3523 struct list_head **iter); 3524#define netdev_for_each_lower_dev(dev, ldev, iter) \ 3525 for (iter = &(dev)->adj_list.lower, \ 3526 ldev = netdev_lower_get_next(dev, &(iter)); \ 3527 ldev; \ 3528 ldev = netdev_lower_get_next(dev, &(iter))) 3529 3530void *netdev_adjacent_get_private(struct list_head *adj_list); 3531void *netdev_lower_get_first_private_rcu(struct net_device *dev); 3532struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 3533struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 3534int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev); 3535int netdev_master_upper_dev_link(struct net_device *dev, 3536 struct net_device *upper_dev); 3537int netdev_master_upper_dev_link_private(struct net_device *dev, 3538 struct net_device *upper_dev, 3539 void *private); 3540void netdev_upper_dev_unlink(struct net_device *dev, 3541 struct net_device *upper_dev); 3542void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 3543void *netdev_lower_dev_get_private(struct net_device *dev, 3544 struct net_device *lower_dev); 3545 3546/* RSS keys are 40 or 52 bytes long */ 3547#define NETDEV_RSS_KEY_LEN 52 3548extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN]; 3549void netdev_rss_key_fill(void *buffer, size_t len); 3550 3551int dev_get_nest_level(struct net_device *dev, 3552 bool (*type_check)(struct net_device *dev)); 3553int skb_checksum_help(struct sk_buff *skb); 3554struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 3555 netdev_features_t features, bool tx_path); 3556struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 3557 netdev_features_t features); 3558 3559struct netdev_bonding_info { 3560 ifslave slave; 3561 ifbond master; 3562}; 3563 3564struct netdev_notifier_bonding_info { 3565 struct netdev_notifier_info info; /* must be first */ 3566 struct netdev_bonding_info bonding_info; 3567}; 3568 3569void netdev_bonding_info_change(struct net_device *dev, 3570 struct netdev_bonding_info *bonding_info); 3571 3572static inline 3573struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 3574{ 3575 return __skb_gso_segment(skb, features, true); 3576} 3577__be16 skb_network_protocol(struct sk_buff *skb, int *depth); 3578 3579static inline bool can_checksum_protocol(netdev_features_t features, 3580 __be16 protocol) 3581{ 3582 return ((features & NETIF_F_GEN_CSUM) || 3583 ((features & NETIF_F_V4_CSUM) && 3584 protocol == htons(ETH_P_IP)) || 3585 ((features & NETIF_F_V6_CSUM) && 3586 protocol == htons(ETH_P_IPV6)) || 3587 ((features & NETIF_F_FCOE_CRC) && 3588 protocol == htons(ETH_P_FCOE))); 3589} 3590 3591#ifdef CONFIG_BUG 3592void netdev_rx_csum_fault(struct net_device *dev); 3593#else 3594static inline void netdev_rx_csum_fault(struct net_device *dev) 3595{ 3596} 3597#endif 3598/* rx skb timestamps */ 3599void net_enable_timestamp(void); 3600void net_disable_timestamp(void); 3601 3602#ifdef CONFIG_PROC_FS 3603int __init dev_proc_init(void); 3604#else 3605#define dev_proc_init() 0 3606#endif 3607 3608static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 3609 struct sk_buff *skb, struct net_device *dev, 3610 bool more) 3611{ 3612 skb->xmit_more = more ? 1 : 0; 3613 return ops->ndo_start_xmit(skb, dev); 3614} 3615 3616static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 3617 struct netdev_queue *txq, bool more) 3618{ 3619 const struct net_device_ops *ops = dev->netdev_ops; 3620 int rc; 3621 3622 rc = __netdev_start_xmit(ops, skb, dev, more); 3623 if (rc == NETDEV_TX_OK) 3624 txq_trans_update(txq); 3625 3626 return rc; 3627} 3628 3629int netdev_class_create_file_ns(struct class_attribute *class_attr, 3630 const void *ns); 3631void netdev_class_remove_file_ns(struct class_attribute *class_attr, 3632 const void *ns); 3633 3634static inline int netdev_class_create_file(struct class_attribute *class_attr) 3635{ 3636 return netdev_class_create_file_ns(class_attr, NULL); 3637} 3638 3639static inline void netdev_class_remove_file(struct class_attribute *class_attr) 3640{ 3641 netdev_class_remove_file_ns(class_attr, NULL); 3642} 3643 3644extern struct kobj_ns_type_operations net_ns_type_operations; 3645 3646const char *netdev_drivername(const struct net_device *dev); 3647 3648void linkwatch_run_queue(void); 3649 3650static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 3651 netdev_features_t f2) 3652{ 3653 if (f1 & NETIF_F_GEN_CSUM) 3654 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM); 3655 if (f2 & NETIF_F_GEN_CSUM) 3656 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM); 3657 f1 &= f2; 3658 if (f1 & NETIF_F_GEN_CSUM) 3659 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM); 3660 3661 return f1; 3662} 3663 3664static inline netdev_features_t netdev_get_wanted_features( 3665 struct net_device *dev) 3666{ 3667 return (dev->features & ~dev->hw_features) | dev->wanted_features; 3668} 3669netdev_features_t netdev_increment_features(netdev_features_t all, 3670 netdev_features_t one, netdev_features_t mask); 3671 3672/* Allow TSO being used on stacked device : 3673 * Performing the GSO segmentation before last device 3674 * is a performance improvement. 3675 */ 3676static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 3677 netdev_features_t mask) 3678{ 3679 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 3680} 3681 3682int __netdev_update_features(struct net_device *dev); 3683void netdev_update_features(struct net_device *dev); 3684void netdev_change_features(struct net_device *dev); 3685 3686void netif_stacked_transfer_operstate(const struct net_device *rootdev, 3687 struct net_device *dev); 3688 3689netdev_features_t passthru_features_check(struct sk_buff *skb, 3690 struct net_device *dev, 3691 netdev_features_t features); 3692netdev_features_t netif_skb_features(struct sk_buff *skb); 3693 3694static inline bool net_gso_ok(netdev_features_t features, int gso_type) 3695{ 3696 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT; 3697 3698 /* check flags correspondence */ 3699 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 3700 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT)); 3701 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 3702 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 3703 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 3704 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 3705 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 3706 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 3707 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT)); 3708 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT)); 3709 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 3710 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 3711 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 3712 3713 return (features & feature) == feature; 3714} 3715 3716static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 3717{ 3718 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 3719 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 3720} 3721 3722static inline bool netif_needs_gso(struct sk_buff *skb, 3723 netdev_features_t features) 3724{ 3725 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 3726 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 3727 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 3728} 3729 3730static inline void netif_set_gso_max_size(struct net_device *dev, 3731 unsigned int size) 3732{ 3733 dev->gso_max_size = size; 3734} 3735 3736static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 3737 int pulled_hlen, u16 mac_offset, 3738 int mac_len) 3739{ 3740 skb->protocol = protocol; 3741 skb->encapsulation = 1; 3742 skb_push(skb, pulled_hlen); 3743 skb_reset_transport_header(skb); 3744 skb->mac_header = mac_offset; 3745 skb->network_header = skb->mac_header + mac_len; 3746 skb->mac_len = mac_len; 3747} 3748 3749static inline bool netif_is_macvlan(struct net_device *dev) 3750{ 3751 return dev->priv_flags & IFF_MACVLAN; 3752} 3753 3754static inline bool netif_is_macvlan_port(struct net_device *dev) 3755{ 3756 return dev->priv_flags & IFF_MACVLAN_PORT; 3757} 3758 3759static inline bool netif_is_ipvlan(struct net_device *dev) 3760{ 3761 return dev->priv_flags & IFF_IPVLAN_SLAVE; 3762} 3763 3764static inline bool netif_is_ipvlan_port(struct net_device *dev) 3765{ 3766 return dev->priv_flags & IFF_IPVLAN_MASTER; 3767} 3768 3769static inline bool netif_is_bond_master(struct net_device *dev) 3770{ 3771 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 3772} 3773 3774static inline bool netif_is_bond_slave(struct net_device *dev) 3775{ 3776 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 3777} 3778 3779static inline bool netif_supports_nofcs(struct net_device *dev) 3780{ 3781 return dev->priv_flags & IFF_SUPP_NOFCS; 3782} 3783 3784/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 3785static inline void netif_keep_dst(struct net_device *dev) 3786{ 3787 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 3788} 3789 3790extern struct pernet_operations __net_initdata loopback_net_ops; 3791 3792/* Logging, debugging and troubleshooting/diagnostic helpers. */ 3793 3794/* netdev_printk helpers, similar to dev_printk */ 3795 3796static inline const char *netdev_name(const struct net_device *dev) 3797{ 3798 if (!dev->name[0] || strchr(dev->name, '%')) 3799 return "(unnamed net_device)"; 3800 return dev->name; 3801} 3802 3803static inline const char *netdev_reg_state(const struct net_device *dev) 3804{ 3805 switch (dev->reg_state) { 3806 case NETREG_UNINITIALIZED: return " (uninitialized)"; 3807 case NETREG_REGISTERED: return ""; 3808 case NETREG_UNREGISTERING: return " (unregistering)"; 3809 case NETREG_UNREGISTERED: return " (unregistered)"; 3810 case NETREG_RELEASED: return " (released)"; 3811 case NETREG_DUMMY: return " (dummy)"; 3812 } 3813 3814 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 3815 return " (unknown)"; 3816} 3817 3818__printf(3, 4) 3819void netdev_printk(const char *level, const struct net_device *dev, 3820 const char *format, ...); 3821__printf(2, 3) 3822void netdev_emerg(const struct net_device *dev, const char *format, ...); 3823__printf(2, 3) 3824void netdev_alert(const struct net_device *dev, const char *format, ...); 3825__printf(2, 3) 3826void netdev_crit(const struct net_device *dev, const char *format, ...); 3827__printf(2, 3) 3828void netdev_err(const struct net_device *dev, const char *format, ...); 3829__printf(2, 3) 3830void netdev_warn(const struct net_device *dev, const char *format, ...); 3831__printf(2, 3) 3832void netdev_notice(const struct net_device *dev, const char *format, ...); 3833__printf(2, 3) 3834void netdev_info(const struct net_device *dev, const char *format, ...); 3835 3836#define MODULE_ALIAS_NETDEV(device) \ 3837 MODULE_ALIAS("netdev-" device) 3838 3839#if defined(CONFIG_DYNAMIC_DEBUG) 3840#define netdev_dbg(__dev, format, args...) \ 3841do { \ 3842 dynamic_netdev_dbg(__dev, format, ##args); \ 3843} while (0) 3844#elif defined(DEBUG) 3845#define netdev_dbg(__dev, format, args...) \ 3846 netdev_printk(KERN_DEBUG, __dev, format, ##args) 3847#else 3848#define netdev_dbg(__dev, format, args...) \ 3849({ \ 3850 if (0) \ 3851 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 3852}) 3853#endif 3854 3855#if defined(VERBOSE_DEBUG) 3856#define netdev_vdbg netdev_dbg 3857#else 3858 3859#define netdev_vdbg(dev, format, args...) \ 3860({ \ 3861 if (0) \ 3862 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 3863 0; \ 3864}) 3865#endif 3866 3867/* 3868 * netdev_WARN() acts like dev_printk(), but with the key difference 3869 * of using a WARN/WARN_ON to get the message out, including the 3870 * file/line information and a backtrace. 3871 */ 3872#define netdev_WARN(dev, format, args...) \ 3873 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \ 3874 netdev_reg_state(dev), ##args) 3875 3876/* netif printk helpers, similar to netdev_printk */ 3877 3878#define netif_printk(priv, type, level, dev, fmt, args...) \ 3879do { \ 3880 if (netif_msg_##type(priv)) \ 3881 netdev_printk(level, (dev), fmt, ##args); \ 3882} while (0) 3883 3884#define netif_level(level, priv, type, dev, fmt, args...) \ 3885do { \ 3886 if (netif_msg_##type(priv)) \ 3887 netdev_##level(dev, fmt, ##args); \ 3888} while (0) 3889 3890#define netif_emerg(priv, type, dev, fmt, args...) \ 3891 netif_level(emerg, priv, type, dev, fmt, ##args) 3892#define netif_alert(priv, type, dev, fmt, args...) \ 3893 netif_level(alert, priv, type, dev, fmt, ##args) 3894#define netif_crit(priv, type, dev, fmt, args...) \ 3895 netif_level(crit, priv, type, dev, fmt, ##args) 3896#define netif_err(priv, type, dev, fmt, args...) \ 3897 netif_level(err, priv, type, dev, fmt, ##args) 3898#define netif_warn(priv, type, dev, fmt, args...) \ 3899 netif_level(warn, priv, type, dev, fmt, ##args) 3900#define netif_notice(priv, type, dev, fmt, args...) \ 3901 netif_level(notice, priv, type, dev, fmt, ##args) 3902#define netif_info(priv, type, dev, fmt, args...) \ 3903 netif_level(info, priv, type, dev, fmt, ##args) 3904 3905#if defined(CONFIG_DYNAMIC_DEBUG) 3906#define netif_dbg(priv, type, netdev, format, args...) \ 3907do { \ 3908 if (netif_msg_##type(priv)) \ 3909 dynamic_netdev_dbg(netdev, format, ##args); \ 3910} while (0) 3911#elif defined(DEBUG) 3912#define netif_dbg(priv, type, dev, format, args...) \ 3913 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 3914#else 3915#define netif_dbg(priv, type, dev, format, args...) \ 3916({ \ 3917 if (0) \ 3918 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 3919 0; \ 3920}) 3921#endif 3922 3923#if defined(VERBOSE_DEBUG) 3924#define netif_vdbg netif_dbg 3925#else 3926#define netif_vdbg(priv, type, dev, format, args...) \ 3927({ \ 3928 if (0) \ 3929 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 3930 0; \ 3931}) 3932#endif 3933 3934/* 3935 * The list of packet types we will receive (as opposed to discard) 3936 * and the routines to invoke. 3937 * 3938 * Why 16. Because with 16 the only overlap we get on a hash of the 3939 * low nibble of the protocol value is RARP/SNAP/X.25. 3940 * 3941 * NOTE: That is no longer true with the addition of VLAN tags. Not 3942 * sure which should go first, but I bet it won't make much 3943 * difference if we are running VLANs. The good news is that 3944 * this protocol won't be in the list unless compiled in, so 3945 * the average user (w/out VLANs) will not be adversely affected. 3946 * --BLG 3947 * 3948 * 0800 IP 3949 * 8100 802.1Q VLAN 3950 * 0001 802.3 3951 * 0002 AX.25 3952 * 0004 802.2 3953 * 8035 RARP 3954 * 0005 SNAP 3955 * 0805 X.25 3956 * 0806 ARP 3957 * 8137 IPX 3958 * 0009 Localtalk 3959 * 86DD IPv6 3960 */ 3961#define PTYPE_HASH_SIZE (16) 3962#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 3963 3964#endif /* _LINUX_NETDEVICE_H */