at v6.16 177 kB view raw
1/* SPDX-License-Identifier: GPL-2.0-or-later */ 2/* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 13 * Corey Minyard <wf-rch!minyard@relay.EU.net> 14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov> 15 * Alan Cox, <alan@lxorguk.ukuu.org.uk> 16 * Bjorn Ekwall. <bj0rn@blox.se> 17 * Pekka Riikonen <priikone@poseidon.pspt.fi> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21#ifndef _LINUX_NETDEVICE_H 22#define _LINUX_NETDEVICE_H 23 24#include <linux/timer.h> 25#include <linux/bug.h> 26#include <linux/delay.h> 27#include <linux/atomic.h> 28#include <linux/prefetch.h> 29#include <asm/cache.h> 30#include <asm/byteorder.h> 31#include <asm/local.h> 32 33#include <linux/percpu.h> 34#include <linux/rculist.h> 35#include <linux/workqueue.h> 36#include <linux/dynamic_queue_limits.h> 37 38#include <net/net_namespace.h> 39#ifdef CONFIG_DCB 40#include <net/dcbnl.h> 41#endif 42#include <net/netprio_cgroup.h> 43#include <linux/netdev_features.h> 44#include <linux/neighbour.h> 45#include <linux/netdevice_xmit.h> 46#include <uapi/linux/netdevice.h> 47#include <uapi/linux/if_bonding.h> 48#include <uapi/linux/pkt_cls.h> 49#include <uapi/linux/netdev.h> 50#include <linux/hashtable.h> 51#include <linux/rbtree.h> 52#include <net/net_trackers.h> 53#include <net/net_debug.h> 54#include <net/dropreason-core.h> 55#include <net/neighbour_tables.h> 56 57struct netpoll_info; 58struct device; 59struct ethtool_ops; 60struct kernel_hwtstamp_config; 61struct phy_device; 62struct dsa_port; 63struct ip_tunnel_parm_kern; 64struct macsec_context; 65struct macsec_ops; 66struct netdev_config; 67struct netdev_name_node; 68struct sd_flow_limit; 69struct sfp_bus; 70/* 802.11 specific */ 71struct wireless_dev; 72/* 802.15.4 specific */ 73struct wpan_dev; 74struct mpls_dev; 75/* UDP Tunnel offloads */ 76struct udp_tunnel_info; 77struct udp_tunnel_nic_info; 78struct udp_tunnel_nic; 79struct bpf_prog; 80struct xdp_buff; 81struct xdp_frame; 82struct xdp_metadata_ops; 83struct xdp_md; 84struct ethtool_netdev_state; 85struct phy_link_topology; 86struct hwtstamp_provider; 87 88typedef u32 xdp_features_t; 89 90void synchronize_net(void); 91void netdev_set_default_ethtool_ops(struct net_device *dev, 92 const struct ethtool_ops *ops); 93void netdev_sw_irq_coalesce_default_on(struct net_device *dev); 94 95/* Backlog congestion levels */ 96#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 97#define NET_RX_DROP 1 /* packet dropped */ 98 99#define MAX_NEST_DEV 8 100 101/* 102 * Transmit return codes: transmit return codes originate from three different 103 * namespaces: 104 * 105 * - qdisc return codes 106 * - driver transmit return codes 107 * - errno values 108 * 109 * Drivers are allowed to return any one of those in their hard_start_xmit() 110 * function. Real network devices commonly used with qdiscs should only return 111 * the driver transmit return codes though - when qdiscs are used, the actual 112 * transmission happens asynchronously, so the value is not propagated to 113 * higher layers. Virtual network devices transmit synchronously; in this case 114 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 115 * others are propagated to higher layers. 116 */ 117 118/* qdisc ->enqueue() return codes. */ 119#define NET_XMIT_SUCCESS 0x00 120#define NET_XMIT_DROP 0x01 /* skb dropped */ 121#define NET_XMIT_CN 0x02 /* congestion notification */ 122#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 123 124/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 125 * indicates that the device will soon be dropping packets, or already drops 126 * some packets of the same priority; prompting us to send less aggressively. */ 127#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 128#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 129 130/* Driver transmit return codes */ 131#define NETDEV_TX_MASK 0xf0 132 133enum netdev_tx { 134 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 135 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 136 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 137}; 138typedef enum netdev_tx netdev_tx_t; 139 140/* 141 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 142 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 143 */ 144static inline bool dev_xmit_complete(int rc) 145{ 146 /* 147 * Positive cases with an skb consumed by a driver: 148 * - successful transmission (rc == NETDEV_TX_OK) 149 * - error while transmitting (rc < 0) 150 * - error while queueing to a different device (rc & NET_XMIT_MASK) 151 */ 152 if (likely(rc < NET_XMIT_MASK)) 153 return true; 154 155 return false; 156} 157 158/* 159 * Compute the worst-case header length according to the protocols 160 * used. 161 */ 162 163#if defined(CONFIG_HYPERV_NET) 164# define LL_MAX_HEADER 128 165#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 166# if defined(CONFIG_MAC80211_MESH) 167# define LL_MAX_HEADER 128 168# else 169# define LL_MAX_HEADER 96 170# endif 171#else 172# define LL_MAX_HEADER 32 173#endif 174 175#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 176 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 177#define MAX_HEADER LL_MAX_HEADER 178#else 179#define MAX_HEADER (LL_MAX_HEADER + 48) 180#endif 181 182/* 183 * Old network device statistics. Fields are native words 184 * (unsigned long) so they can be read and written atomically. 185 */ 186 187#define NET_DEV_STAT(FIELD) \ 188 union { \ 189 unsigned long FIELD; \ 190 atomic_long_t __##FIELD; \ 191 } 192 193struct net_device_stats { 194 NET_DEV_STAT(rx_packets); 195 NET_DEV_STAT(tx_packets); 196 NET_DEV_STAT(rx_bytes); 197 NET_DEV_STAT(tx_bytes); 198 NET_DEV_STAT(rx_errors); 199 NET_DEV_STAT(tx_errors); 200 NET_DEV_STAT(rx_dropped); 201 NET_DEV_STAT(tx_dropped); 202 NET_DEV_STAT(multicast); 203 NET_DEV_STAT(collisions); 204 NET_DEV_STAT(rx_length_errors); 205 NET_DEV_STAT(rx_over_errors); 206 NET_DEV_STAT(rx_crc_errors); 207 NET_DEV_STAT(rx_frame_errors); 208 NET_DEV_STAT(rx_fifo_errors); 209 NET_DEV_STAT(rx_missed_errors); 210 NET_DEV_STAT(tx_aborted_errors); 211 NET_DEV_STAT(tx_carrier_errors); 212 NET_DEV_STAT(tx_fifo_errors); 213 NET_DEV_STAT(tx_heartbeat_errors); 214 NET_DEV_STAT(tx_window_errors); 215 NET_DEV_STAT(rx_compressed); 216 NET_DEV_STAT(tx_compressed); 217}; 218#undef NET_DEV_STAT 219 220/* per-cpu stats, allocated on demand. 221 * Try to fit them in a single cache line, for dev_get_stats() sake. 222 */ 223struct net_device_core_stats { 224 unsigned long rx_dropped; 225 unsigned long tx_dropped; 226 unsigned long rx_nohandler; 227 unsigned long rx_otherhost_dropped; 228} __aligned(4 * sizeof(unsigned long)); 229 230#include <linux/cache.h> 231#include <linux/skbuff.h> 232 233struct neighbour; 234struct neigh_parms; 235struct sk_buff; 236 237struct netdev_hw_addr { 238 struct list_head list; 239 struct rb_node node; 240 unsigned char addr[MAX_ADDR_LEN]; 241 unsigned char type; 242#define NETDEV_HW_ADDR_T_LAN 1 243#define NETDEV_HW_ADDR_T_SAN 2 244#define NETDEV_HW_ADDR_T_UNICAST 3 245#define NETDEV_HW_ADDR_T_MULTICAST 4 246 bool global_use; 247 int sync_cnt; 248 int refcount; 249 int synced; 250 struct rcu_head rcu_head; 251}; 252 253struct netdev_hw_addr_list { 254 struct list_head list; 255 int count; 256 257 /* Auxiliary tree for faster lookup on addition and deletion */ 258 struct rb_root tree; 259}; 260 261#define netdev_hw_addr_list_count(l) ((l)->count) 262#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 263#define netdev_hw_addr_list_for_each(ha, l) \ 264 list_for_each_entry(ha, &(l)->list, list) 265 266#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 267#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 268#define netdev_for_each_uc_addr(ha, dev) \ 269 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 270#define netdev_for_each_synced_uc_addr(_ha, _dev) \ 271 netdev_for_each_uc_addr((_ha), (_dev)) \ 272 if ((_ha)->sync_cnt) 273 274#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 275#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 276#define netdev_for_each_mc_addr(ha, dev) \ 277 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 278#define netdev_for_each_synced_mc_addr(_ha, _dev) \ 279 netdev_for_each_mc_addr((_ha), (_dev)) \ 280 if ((_ha)->sync_cnt) 281 282struct hh_cache { 283 unsigned int hh_len; 284 seqlock_t hh_lock; 285 286 /* cached hardware header; allow for machine alignment needs. */ 287#define HH_DATA_MOD 16 288#define HH_DATA_OFF(__len) \ 289 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 290#define HH_DATA_ALIGN(__len) \ 291 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 292 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 293}; 294 295/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 296 * Alternative is: 297 * dev->hard_header_len ? (dev->hard_header_len + 298 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 299 * 300 * We could use other alignment values, but we must maintain the 301 * relationship HH alignment <= LL alignment. 302 */ 303#define LL_RESERVED_SPACE(dev) \ 304 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \ 305 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 306#define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 307 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \ 308 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 309 310struct header_ops { 311 int (*create) (struct sk_buff *skb, struct net_device *dev, 312 unsigned short type, const void *daddr, 313 const void *saddr, unsigned int len); 314 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 315 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 316 void (*cache_update)(struct hh_cache *hh, 317 const struct net_device *dev, 318 const unsigned char *haddr); 319 bool (*validate)(const char *ll_header, unsigned int len); 320 __be16 (*parse_protocol)(const struct sk_buff *skb); 321}; 322 323/* These flag bits are private to the generic network queueing 324 * layer; they may not be explicitly referenced by any other 325 * code. 326 */ 327 328enum netdev_state_t { 329 __LINK_STATE_START, 330 __LINK_STATE_PRESENT, 331 __LINK_STATE_NOCARRIER, 332 __LINK_STATE_LINKWATCH_PENDING, 333 __LINK_STATE_DORMANT, 334 __LINK_STATE_TESTING, 335}; 336 337struct gro_list { 338 struct list_head list; 339 int count; 340}; 341 342/* 343 * size of gro hash buckets, must be <= the number of bits in 344 * gro_node::bitmask 345 */ 346#define GRO_HASH_BUCKETS 8 347 348/** 349 * struct gro_node - structure to support Generic Receive Offload 350 * @bitmask: bitmask to indicate used buckets in @hash 351 * @hash: hashtable of pending aggregated skbs, separated by flows 352 * @rx_list: list of pending ``GRO_NORMAL`` skbs 353 * @rx_count: cached current length of @rx_list 354 * @cached_napi_id: napi_struct::napi_id cached for hotpath, 0 for standalone 355 */ 356struct gro_node { 357 unsigned long bitmask; 358 struct gro_list hash[GRO_HASH_BUCKETS]; 359 struct list_head rx_list; 360 u32 rx_count; 361 u32 cached_napi_id; 362}; 363 364/* 365 * Structure for per-NAPI config 366 */ 367struct napi_config { 368 u64 gro_flush_timeout; 369 u64 irq_suspend_timeout; 370 u32 defer_hard_irqs; 371 cpumask_t affinity_mask; 372 unsigned int napi_id; 373}; 374 375/* 376 * Structure for NAPI scheduling similar to tasklet but with weighting 377 */ 378struct napi_struct { 379 /* The poll_list must only be managed by the entity which 380 * changes the state of the NAPI_STATE_SCHED bit. This means 381 * whoever atomically sets that bit can add this napi_struct 382 * to the per-CPU poll_list, and whoever clears that bit 383 * can remove from the list right before clearing the bit. 384 */ 385 struct list_head poll_list; 386 387 unsigned long state; 388 int weight; 389 u32 defer_hard_irqs_count; 390 int (*poll)(struct napi_struct *, int); 391#ifdef CONFIG_NETPOLL 392 /* CPU actively polling if netpoll is configured */ 393 int poll_owner; 394#endif 395 /* CPU on which NAPI has been scheduled for processing */ 396 int list_owner; 397 struct net_device *dev; 398 struct sk_buff *skb; 399 struct gro_node gro; 400 struct hrtimer timer; 401 /* all fields past this point are write-protected by netdev_lock */ 402 struct task_struct *thread; 403 unsigned long gro_flush_timeout; 404 unsigned long irq_suspend_timeout; 405 u32 defer_hard_irqs; 406 /* control-path-only fields follow */ 407 u32 napi_id; 408 struct list_head dev_list; 409 struct hlist_node napi_hash_node; 410 int irq; 411 struct irq_affinity_notify notify; 412 int napi_rmap_idx; 413 int index; 414 struct napi_config *config; 415}; 416 417enum { 418 NAPI_STATE_SCHED, /* Poll is scheduled */ 419 NAPI_STATE_MISSED, /* reschedule a napi */ 420 NAPI_STATE_DISABLE, /* Disable pending */ 421 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 422 NAPI_STATE_LISTED, /* NAPI added to system lists */ 423 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 424 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 425 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 426 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 427 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 428 NAPI_STATE_HAS_NOTIFIER, /* Napi has an IRQ notifier */ 429}; 430 431enum { 432 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 433 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 434 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 435 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 436 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 437 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 438 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 439 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 440 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 441 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 442 NAPIF_STATE_HAS_NOTIFIER = BIT(NAPI_STATE_HAS_NOTIFIER), 443}; 444 445enum gro_result { 446 GRO_MERGED, 447 GRO_MERGED_FREE, 448 GRO_HELD, 449 GRO_NORMAL, 450 GRO_CONSUMED, 451}; 452typedef enum gro_result gro_result_t; 453 454/* 455 * enum rx_handler_result - Possible return values for rx_handlers. 456 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 457 * further. 458 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 459 * case skb->dev was changed by rx_handler. 460 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 461 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 462 * 463 * rx_handlers are functions called from inside __netif_receive_skb(), to do 464 * special processing of the skb, prior to delivery to protocol handlers. 465 * 466 * Currently, a net_device can only have a single rx_handler registered. Trying 467 * to register a second rx_handler will return -EBUSY. 468 * 469 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 470 * To unregister a rx_handler on a net_device, use 471 * netdev_rx_handler_unregister(). 472 * 473 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 474 * do with the skb. 475 * 476 * If the rx_handler consumed the skb in some way, it should return 477 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 478 * the skb to be delivered in some other way. 479 * 480 * If the rx_handler changed skb->dev, to divert the skb to another 481 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 482 * new device will be called if it exists. 483 * 484 * If the rx_handler decides the skb should be ignored, it should return 485 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 486 * are registered on exact device (ptype->dev == skb->dev). 487 * 488 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 489 * delivered, it should return RX_HANDLER_PASS. 490 * 491 * A device without a registered rx_handler will behave as if rx_handler 492 * returned RX_HANDLER_PASS. 493 */ 494 495enum rx_handler_result { 496 RX_HANDLER_CONSUMED, 497 RX_HANDLER_ANOTHER, 498 RX_HANDLER_EXACT, 499 RX_HANDLER_PASS, 500}; 501typedef enum rx_handler_result rx_handler_result_t; 502typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 503 504void __napi_schedule(struct napi_struct *n); 505void __napi_schedule_irqoff(struct napi_struct *n); 506 507static inline bool napi_disable_pending(struct napi_struct *n) 508{ 509 return test_bit(NAPI_STATE_DISABLE, &n->state); 510} 511 512static inline bool napi_prefer_busy_poll(struct napi_struct *n) 513{ 514 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 515} 516 517/** 518 * napi_is_scheduled - test if NAPI is scheduled 519 * @n: NAPI context 520 * 521 * This check is "best-effort". With no locking implemented, 522 * a NAPI can be scheduled or terminate right after this check 523 * and produce not precise results. 524 * 525 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled 526 * should not be used normally and napi_schedule should be 527 * used instead. 528 * 529 * Use only if the driver really needs to check if a NAPI 530 * is scheduled for example in the context of delayed timer 531 * that can be skipped if a NAPI is already scheduled. 532 * 533 * Return: True if NAPI is scheduled, False otherwise. 534 */ 535static inline bool napi_is_scheduled(struct napi_struct *n) 536{ 537 return test_bit(NAPI_STATE_SCHED, &n->state); 538} 539 540bool napi_schedule_prep(struct napi_struct *n); 541 542/** 543 * napi_schedule - schedule NAPI poll 544 * @n: NAPI context 545 * 546 * Schedule NAPI poll routine to be called if it is not already 547 * running. 548 * Return: true if we schedule a NAPI or false if not. 549 * Refer to napi_schedule_prep() for additional reason on why 550 * a NAPI might not be scheduled. 551 */ 552static inline bool napi_schedule(struct napi_struct *n) 553{ 554 if (napi_schedule_prep(n)) { 555 __napi_schedule(n); 556 return true; 557 } 558 559 return false; 560} 561 562/** 563 * napi_schedule_irqoff - schedule NAPI poll 564 * @n: NAPI context 565 * 566 * Variant of napi_schedule(), assuming hard irqs are masked. 567 */ 568static inline void napi_schedule_irqoff(struct napi_struct *n) 569{ 570 if (napi_schedule_prep(n)) 571 __napi_schedule_irqoff(n); 572} 573 574/** 575 * napi_complete_done - NAPI processing complete 576 * @n: NAPI context 577 * @work_done: number of packets processed 578 * 579 * Mark NAPI processing as complete. Should only be called if poll budget 580 * has not been completely consumed. 581 * Prefer over napi_complete(). 582 * Return: false if device should avoid rearming interrupts. 583 */ 584bool napi_complete_done(struct napi_struct *n, int work_done); 585 586static inline bool napi_complete(struct napi_struct *n) 587{ 588 return napi_complete_done(n, 0); 589} 590 591int dev_set_threaded(struct net_device *dev, bool threaded); 592 593void napi_disable(struct napi_struct *n); 594void napi_disable_locked(struct napi_struct *n); 595 596void napi_enable(struct napi_struct *n); 597void napi_enable_locked(struct napi_struct *n); 598 599/** 600 * napi_synchronize - wait until NAPI is not running 601 * @n: NAPI context 602 * 603 * Wait until NAPI is done being scheduled on this context. 604 * Waits till any outstanding processing completes but 605 * does not disable future activations. 606 */ 607static inline void napi_synchronize(const struct napi_struct *n) 608{ 609 if (IS_ENABLED(CONFIG_SMP)) 610 while (test_bit(NAPI_STATE_SCHED, &n->state)) 611 msleep(1); 612 else 613 barrier(); 614} 615 616/** 617 * napi_if_scheduled_mark_missed - if napi is running, set the 618 * NAPIF_STATE_MISSED 619 * @n: NAPI context 620 * 621 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 622 * NAPI is scheduled. 623 **/ 624static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 625{ 626 unsigned long val, new; 627 628 val = READ_ONCE(n->state); 629 do { 630 if (val & NAPIF_STATE_DISABLE) 631 return true; 632 633 if (!(val & NAPIF_STATE_SCHED)) 634 return false; 635 636 new = val | NAPIF_STATE_MISSED; 637 } while (!try_cmpxchg(&n->state, &val, new)); 638 639 return true; 640} 641 642enum netdev_queue_state_t { 643 __QUEUE_STATE_DRV_XOFF, 644 __QUEUE_STATE_STACK_XOFF, 645 __QUEUE_STATE_FROZEN, 646}; 647 648#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 649#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 650#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 651 652#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 653#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 654 QUEUE_STATE_FROZEN) 655#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 656 QUEUE_STATE_FROZEN) 657 658/* 659 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 660 * netif_tx_* functions below are used to manipulate this flag. The 661 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 662 * queue independently. The netif_xmit_*stopped functions below are called 663 * to check if the queue has been stopped by the driver or stack (either 664 * of the XOFF bits are set in the state). Drivers should not need to call 665 * netif_xmit*stopped functions, they should only be using netif_tx_*. 666 */ 667 668struct netdev_queue { 669/* 670 * read-mostly part 671 */ 672 struct net_device *dev; 673 netdevice_tracker dev_tracker; 674 675 struct Qdisc __rcu *qdisc; 676 struct Qdisc __rcu *qdisc_sleeping; 677#ifdef CONFIG_SYSFS 678 struct kobject kobj; 679 const struct attribute_group **groups; 680#endif 681 unsigned long tx_maxrate; 682 /* 683 * Number of TX timeouts for this queue 684 * (/sys/class/net/DEV/Q/trans_timeout) 685 */ 686 atomic_long_t trans_timeout; 687 688 /* Subordinate device that the queue has been assigned to */ 689 struct net_device *sb_dev; 690#ifdef CONFIG_XDP_SOCKETS 691 /* "ops protected", see comment about net_device::lock */ 692 struct xsk_buff_pool *pool; 693#endif 694 695/* 696 * write-mostly part 697 */ 698#ifdef CONFIG_BQL 699 struct dql dql; 700#endif 701 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 702 int xmit_lock_owner; 703 /* 704 * Time (in jiffies) of last Tx 705 */ 706 unsigned long trans_start; 707 708 unsigned long state; 709 710/* 711 * slow- / control-path part 712 */ 713 /* NAPI instance for the queue 714 * "ops protected", see comment about net_device::lock 715 */ 716 struct napi_struct *napi; 717 718#if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 719 int numa_node; 720#endif 721} ____cacheline_aligned_in_smp; 722 723extern int sysctl_fb_tunnels_only_for_init_net; 724extern int sysctl_devconf_inherit_init_net; 725 726/* 727 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 728 * == 1 : For initns only 729 * == 2 : For none. 730 */ 731static inline bool net_has_fallback_tunnels(const struct net *net) 732{ 733#if IS_ENABLED(CONFIG_SYSCTL) 734 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net); 735 736 return !fb_tunnels_only_for_init_net || 737 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1); 738#else 739 return true; 740#endif 741} 742 743static inline int net_inherit_devconf(void) 744{ 745#if IS_ENABLED(CONFIG_SYSCTL) 746 return READ_ONCE(sysctl_devconf_inherit_init_net); 747#else 748 return 0; 749#endif 750} 751 752static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 753{ 754#if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 755 return q->numa_node; 756#else 757 return NUMA_NO_NODE; 758#endif 759} 760 761static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 762{ 763#if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 764 q->numa_node = node; 765#endif 766} 767 768#ifdef CONFIG_RFS_ACCEL 769bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 770 u16 filter_id); 771#endif 772 773/* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 774enum xps_map_type { 775 XPS_CPUS = 0, 776 XPS_RXQS, 777 XPS_MAPS_MAX, 778}; 779 780#ifdef CONFIG_XPS 781/* 782 * This structure holds an XPS map which can be of variable length. The 783 * map is an array of queues. 784 */ 785struct xps_map { 786 unsigned int len; 787 unsigned int alloc_len; 788 struct rcu_head rcu; 789 u16 queues[]; 790}; 791#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 792#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 793 - sizeof(struct xps_map)) / sizeof(u16)) 794 795/* 796 * This structure holds all XPS maps for device. Maps are indexed by CPU. 797 * 798 * We keep track of the number of cpus/rxqs used when the struct is allocated, 799 * in nr_ids. This will help not accessing out-of-bound memory. 800 * 801 * We keep track of the number of traffic classes used when the struct is 802 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 803 * not crossing its upper bound, as the original dev->num_tc can be updated in 804 * the meantime. 805 */ 806struct xps_dev_maps { 807 struct rcu_head rcu; 808 unsigned int nr_ids; 809 s16 num_tc; 810 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 811}; 812 813#define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 814 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 815 816#define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 817 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 818 819#endif /* CONFIG_XPS */ 820 821#define TC_MAX_QUEUE 16 822#define TC_BITMASK 15 823/* HW offloaded queuing disciplines txq count and offset maps */ 824struct netdev_tc_txq { 825 u16 count; 826 u16 offset; 827}; 828 829#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 830/* 831 * This structure is to hold information about the device 832 * configured to run FCoE protocol stack. 833 */ 834struct netdev_fcoe_hbainfo { 835 char manufacturer[64]; 836 char serial_number[64]; 837 char hardware_version[64]; 838 char driver_version[64]; 839 char optionrom_version[64]; 840 char firmware_version[64]; 841 char model[256]; 842 char model_description[256]; 843}; 844#endif 845 846#define MAX_PHYS_ITEM_ID_LEN 32 847 848/* This structure holds a unique identifier to identify some 849 * physical item (port for example) used by a netdevice. 850 */ 851struct netdev_phys_item_id { 852 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 853 unsigned char id_len; 854}; 855 856static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 857 struct netdev_phys_item_id *b) 858{ 859 return a->id_len == b->id_len && 860 memcmp(a->id, b->id, a->id_len) == 0; 861} 862 863typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 864 struct sk_buff *skb, 865 struct net_device *sb_dev); 866 867enum net_device_path_type { 868 DEV_PATH_ETHERNET = 0, 869 DEV_PATH_VLAN, 870 DEV_PATH_BRIDGE, 871 DEV_PATH_PPPOE, 872 DEV_PATH_DSA, 873 DEV_PATH_MTK_WDMA, 874}; 875 876struct net_device_path { 877 enum net_device_path_type type; 878 const struct net_device *dev; 879 union { 880 struct { 881 u16 id; 882 __be16 proto; 883 u8 h_dest[ETH_ALEN]; 884 } encap; 885 struct { 886 enum { 887 DEV_PATH_BR_VLAN_KEEP, 888 DEV_PATH_BR_VLAN_TAG, 889 DEV_PATH_BR_VLAN_UNTAG, 890 DEV_PATH_BR_VLAN_UNTAG_HW, 891 } vlan_mode; 892 u16 vlan_id; 893 __be16 vlan_proto; 894 } bridge; 895 struct { 896 int port; 897 u16 proto; 898 } dsa; 899 struct { 900 u8 wdma_idx; 901 u8 queue; 902 u16 wcid; 903 u8 bss; 904 u8 amsdu; 905 } mtk_wdma; 906 }; 907}; 908 909#define NET_DEVICE_PATH_STACK_MAX 5 910#define NET_DEVICE_PATH_VLAN_MAX 2 911 912struct net_device_path_stack { 913 int num_paths; 914 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 915}; 916 917struct net_device_path_ctx { 918 const struct net_device *dev; 919 u8 daddr[ETH_ALEN]; 920 921 int num_vlans; 922 struct { 923 u16 id; 924 __be16 proto; 925 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 926}; 927 928enum tc_setup_type { 929 TC_QUERY_CAPS, 930 TC_SETUP_QDISC_MQPRIO, 931 TC_SETUP_CLSU32, 932 TC_SETUP_CLSFLOWER, 933 TC_SETUP_CLSMATCHALL, 934 TC_SETUP_CLSBPF, 935 TC_SETUP_BLOCK, 936 TC_SETUP_QDISC_CBS, 937 TC_SETUP_QDISC_RED, 938 TC_SETUP_QDISC_PRIO, 939 TC_SETUP_QDISC_MQ, 940 TC_SETUP_QDISC_ETF, 941 TC_SETUP_ROOT_QDISC, 942 TC_SETUP_QDISC_GRED, 943 TC_SETUP_QDISC_TAPRIO, 944 TC_SETUP_FT, 945 TC_SETUP_QDISC_ETS, 946 TC_SETUP_QDISC_TBF, 947 TC_SETUP_QDISC_FIFO, 948 TC_SETUP_QDISC_HTB, 949 TC_SETUP_ACT, 950}; 951 952/* These structures hold the attributes of bpf state that are being passed 953 * to the netdevice through the bpf op. 954 */ 955enum bpf_netdev_command { 956 /* Set or clear a bpf program used in the earliest stages of packet 957 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 958 * is responsible for calling bpf_prog_put on any old progs that are 959 * stored. In case of error, the callee need not release the new prog 960 * reference, but on success it takes ownership and must bpf_prog_put 961 * when it is no longer used. 962 */ 963 XDP_SETUP_PROG, 964 XDP_SETUP_PROG_HW, 965 /* BPF program for offload callbacks, invoked at program load time. */ 966 BPF_OFFLOAD_MAP_ALLOC, 967 BPF_OFFLOAD_MAP_FREE, 968 XDP_SETUP_XSK_POOL, 969}; 970 971struct bpf_prog_offload_ops; 972struct netlink_ext_ack; 973struct xdp_umem; 974struct xdp_dev_bulk_queue; 975struct bpf_xdp_link; 976 977enum bpf_xdp_mode { 978 XDP_MODE_SKB = 0, 979 XDP_MODE_DRV = 1, 980 XDP_MODE_HW = 2, 981 __MAX_XDP_MODE 982}; 983 984struct bpf_xdp_entity { 985 struct bpf_prog *prog; 986 struct bpf_xdp_link *link; 987}; 988 989struct netdev_bpf { 990 enum bpf_netdev_command command; 991 union { 992 /* XDP_SETUP_PROG */ 993 struct { 994 u32 flags; 995 struct bpf_prog *prog; 996 struct netlink_ext_ack *extack; 997 }; 998 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 999 struct { 1000 struct bpf_offloaded_map *offmap; 1001 }; 1002 /* XDP_SETUP_XSK_POOL */ 1003 struct { 1004 struct xsk_buff_pool *pool; 1005 u16 queue_id; 1006 } xsk; 1007 }; 1008}; 1009 1010/* Flags for ndo_xsk_wakeup. */ 1011#define XDP_WAKEUP_RX (1 << 0) 1012#define XDP_WAKEUP_TX (1 << 1) 1013 1014#ifdef CONFIG_XFRM_OFFLOAD 1015struct xfrmdev_ops { 1016 int (*xdo_dev_state_add)(struct net_device *dev, 1017 struct xfrm_state *x, 1018 struct netlink_ext_ack *extack); 1019 void (*xdo_dev_state_delete)(struct net_device *dev, 1020 struct xfrm_state *x); 1021 void (*xdo_dev_state_free)(struct net_device *dev, 1022 struct xfrm_state *x); 1023 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 1024 struct xfrm_state *x); 1025 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1026 void (*xdo_dev_state_update_stats) (struct xfrm_state *x); 1027 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack); 1028 void (*xdo_dev_policy_delete) (struct xfrm_policy *x); 1029 void (*xdo_dev_policy_free) (struct xfrm_policy *x); 1030}; 1031#endif 1032 1033struct dev_ifalias { 1034 struct rcu_head rcuhead; 1035 char ifalias[]; 1036}; 1037 1038struct devlink; 1039struct tlsdev_ops; 1040 1041struct netdev_net_notifier { 1042 struct list_head list; 1043 struct notifier_block *nb; 1044}; 1045 1046/* 1047 * This structure defines the management hooks for network devices. 1048 * The following hooks can be defined; unless noted otherwise, they are 1049 * optional and can be filled with a null pointer. 1050 * 1051 * int (*ndo_init)(struct net_device *dev); 1052 * This function is called once when a network device is registered. 1053 * The network device can use this for any late stage initialization 1054 * or semantic validation. It can fail with an error code which will 1055 * be propagated back to register_netdev. 1056 * 1057 * void (*ndo_uninit)(struct net_device *dev); 1058 * This function is called when device is unregistered or when registration 1059 * fails. It is not called if init fails. 1060 * 1061 * int (*ndo_open)(struct net_device *dev); 1062 * This function is called when a network device transitions to the up 1063 * state. 1064 * 1065 * int (*ndo_stop)(struct net_device *dev); 1066 * This function is called when a network device transitions to the down 1067 * state. 1068 * 1069 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1070 * struct net_device *dev); 1071 * Called when a packet needs to be transmitted. 1072 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1073 * the queue before that can happen; it's for obsolete devices and weird 1074 * corner cases, but the stack really does a non-trivial amount 1075 * of useless work if you return NETDEV_TX_BUSY. 1076 * Required; cannot be NULL. 1077 * 1078 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1079 * struct net_device *dev 1080 * netdev_features_t features); 1081 * Called by core transmit path to determine if device is capable of 1082 * performing offload operations on a given packet. This is to give 1083 * the device an opportunity to implement any restrictions that cannot 1084 * be otherwise expressed by feature flags. The check is called with 1085 * the set of features that the stack has calculated and it returns 1086 * those the driver believes to be appropriate. 1087 * 1088 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1089 * struct net_device *sb_dev); 1090 * Called to decide which queue to use when device supports multiple 1091 * transmit queues. 1092 * 1093 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1094 * This function is called to allow device receiver to make 1095 * changes to configuration when multicast or promiscuous is enabled. 1096 * 1097 * void (*ndo_set_rx_mode)(struct net_device *dev); 1098 * This function is called device changes address list filtering. 1099 * If driver handles unicast address filtering, it should set 1100 * IFF_UNICAST_FLT in its priv_flags. 1101 * 1102 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1103 * This function is called when the Media Access Control address 1104 * needs to be changed. If this interface is not defined, the 1105 * MAC address can not be changed. 1106 * 1107 * int (*ndo_validate_addr)(struct net_device *dev); 1108 * Test if Media Access Control address is valid for the device. 1109 * 1110 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1111 * Old-style ioctl entry point. This is used internally by the 1112 * ieee802154 subsystem but is no longer called by the device 1113 * ioctl handler. 1114 * 1115 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1116 * Used by the bonding driver for its device specific ioctls: 1117 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1118 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1119 * 1120 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1121 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1122 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1123 * 1124 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1125 * Used to set network devices bus interface parameters. This interface 1126 * is retained for legacy reasons; new devices should use the bus 1127 * interface (PCI) for low level management. 1128 * 1129 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1130 * Called when a user wants to change the Maximum Transfer Unit 1131 * of a device. 1132 * 1133 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1134 * Callback used when the transmitter has not made any progress 1135 * for dev->watchdog ticks. 1136 * 1137 * void (*ndo_get_stats64)(struct net_device *dev, 1138 * struct rtnl_link_stats64 *storage); 1139 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1140 * Called when a user wants to get the network device usage 1141 * statistics. Drivers must do one of the following: 1142 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1143 * rtnl_link_stats64 structure passed by the caller. 1144 * 2. Define @ndo_get_stats to update a net_device_stats structure 1145 * (which should normally be dev->stats) and return a pointer to 1146 * it. The structure may be changed asynchronously only if each 1147 * field is written atomically. 1148 * 3. Update dev->stats asynchronously and atomically, and define 1149 * neither operation. 1150 * 1151 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1152 * Return true if this device supports offload stats of this attr_id. 1153 * 1154 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1155 * void *attr_data) 1156 * Get statistics for offload operations by attr_id. Write it into the 1157 * attr_data pointer. 1158 * 1159 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1160 * If device supports VLAN filtering this function is called when a 1161 * VLAN id is registered. 1162 * 1163 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1164 * If device supports VLAN filtering this function is called when a 1165 * VLAN id is unregistered. 1166 * 1167 * void (*ndo_poll_controller)(struct net_device *dev); 1168 * 1169 * SR-IOV management functions. 1170 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1171 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1172 * u8 qos, __be16 proto); 1173 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1174 * int max_tx_rate); 1175 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1176 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1177 * int (*ndo_get_vf_config)(struct net_device *dev, 1178 * int vf, struct ifla_vf_info *ivf); 1179 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1180 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1181 * struct nlattr *port[]); 1182 * 1183 * Enable or disable the VF ability to query its RSS Redirection Table and 1184 * Hash Key. This is needed since on some devices VF share this information 1185 * with PF and querying it may introduce a theoretical security risk. 1186 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1187 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1188 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1189 * void *type_data); 1190 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1191 * This is always called from the stack with the rtnl lock held and netif 1192 * tx queues stopped. This allows the netdevice to perform queue 1193 * management safely. 1194 * 1195 * Fiber Channel over Ethernet (FCoE) offload functions. 1196 * int (*ndo_fcoe_enable)(struct net_device *dev); 1197 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1198 * so the underlying device can perform whatever needed configuration or 1199 * initialization to support acceleration of FCoE traffic. 1200 * 1201 * int (*ndo_fcoe_disable)(struct net_device *dev); 1202 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1203 * so the underlying device can perform whatever needed clean-ups to 1204 * stop supporting acceleration of FCoE traffic. 1205 * 1206 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1207 * struct scatterlist *sgl, unsigned int sgc); 1208 * Called when the FCoE Initiator wants to initialize an I/O that 1209 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1210 * perform necessary setup and returns 1 to indicate the device is set up 1211 * successfully to perform DDP on this I/O, otherwise this returns 0. 1212 * 1213 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1214 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1215 * indicated by the FC exchange id 'xid', so the underlying device can 1216 * clean up and reuse resources for later DDP requests. 1217 * 1218 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1219 * struct scatterlist *sgl, unsigned int sgc); 1220 * Called when the FCoE Target wants to initialize an I/O that 1221 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1222 * perform necessary setup and returns 1 to indicate the device is set up 1223 * successfully to perform DDP on this I/O, otherwise this returns 0. 1224 * 1225 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1226 * struct netdev_fcoe_hbainfo *hbainfo); 1227 * Called when the FCoE Protocol stack wants information on the underlying 1228 * device. This information is utilized by the FCoE protocol stack to 1229 * register attributes with Fiber Channel management service as per the 1230 * FC-GS Fabric Device Management Information(FDMI) specification. 1231 * 1232 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1233 * Called when the underlying device wants to override default World Wide 1234 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1235 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1236 * protocol stack to use. 1237 * 1238 * RFS acceleration. 1239 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1240 * u16 rxq_index, u32 flow_id); 1241 * Set hardware filter for RFS. rxq_index is the target queue index; 1242 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1243 * Return the filter ID on success, or a negative error code. 1244 * 1245 * Slave management functions (for bridge, bonding, etc). 1246 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1247 * Called to make another netdev an underling. 1248 * 1249 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1250 * Called to release previously enslaved netdev. 1251 * 1252 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1253 * struct sk_buff *skb, 1254 * bool all_slaves); 1255 * Get the xmit slave of master device. If all_slaves is true, function 1256 * assume all the slaves can transmit. 1257 * 1258 * Feature/offload setting functions. 1259 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1260 * netdev_features_t features); 1261 * Adjusts the requested feature flags according to device-specific 1262 * constraints, and returns the resulting flags. Must not modify 1263 * the device state. 1264 * 1265 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1266 * Called to update device configuration to new features. Passed 1267 * feature set might be less than what was returned by ndo_fix_features()). 1268 * Must return >0 or -errno if it changed dev->features itself. 1269 * 1270 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1271 * struct net_device *dev, 1272 * const unsigned char *addr, u16 vid, u16 flags, 1273 * bool *notified, struct netlink_ext_ack *extack); 1274 * Adds an FDB entry to dev for addr. 1275 * Callee shall set *notified to true if it sent any appropriate 1276 * notification(s). Otherwise core will send a generic one. 1277 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1278 * struct net_device *dev, 1279 * const unsigned char *addr, u16 vid 1280 * bool *notified, struct netlink_ext_ack *extack); 1281 * Deletes the FDB entry from dev corresponding to addr. 1282 * Callee shall set *notified to true if it sent any appropriate 1283 * notification(s). Otherwise core will send a generic one. 1284 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev, 1285 * struct netlink_ext_ack *extack); 1286 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1287 * struct net_device *dev, struct net_device *filter_dev, 1288 * int *idx) 1289 * Used to add FDB entries to dump requests. Implementers should add 1290 * entries to skb and update idx with the number of entries. 1291 * 1292 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[], 1293 * u16 nlmsg_flags, struct netlink_ext_ack *extack); 1294 * Adds an MDB entry to dev. 1295 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[], 1296 * struct netlink_ext_ack *extack); 1297 * Deletes the MDB entry from dev. 1298 * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[], 1299 * struct netlink_ext_ack *extack); 1300 * Bulk deletes MDB entries from dev. 1301 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb, 1302 * struct netlink_callback *cb); 1303 * Dumps MDB entries from dev. The first argument (marker) in the netlink 1304 * callback is used by core rtnetlink code. 1305 * 1306 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1307 * u16 flags, struct netlink_ext_ack *extack) 1308 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1309 * struct net_device *dev, u32 filter_mask, 1310 * int nlflags) 1311 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1312 * u16 flags); 1313 * 1314 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1315 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1316 * which do not represent real hardware may define this to allow their 1317 * userspace components to manage their virtual carrier state. Devices 1318 * that determine carrier state from physical hardware properties (eg 1319 * network cables) or protocol-dependent mechanisms (eg 1320 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1321 * 1322 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1323 * struct netdev_phys_item_id *ppid); 1324 * Called to get ID of physical port of this device. If driver does 1325 * not implement this, it is assumed that the hw is not able to have 1326 * multiple net devices on single physical port. 1327 * 1328 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1329 * struct netdev_phys_item_id *ppid) 1330 * Called to get the parent ID of the physical port of this device. 1331 * 1332 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1333 * struct net_device *dev) 1334 * Called by upper layer devices to accelerate switching or other 1335 * station functionality into hardware. 'pdev is the lowerdev 1336 * to use for the offload and 'dev' is the net device that will 1337 * back the offload. Returns a pointer to the private structure 1338 * the upper layer will maintain. 1339 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1340 * Called by upper layer device to delete the station created 1341 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1342 * the station and priv is the structure returned by the add 1343 * operation. 1344 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1345 * int queue_index, u32 maxrate); 1346 * Called when a user wants to set a max-rate limitation of specific 1347 * TX queue. 1348 * int (*ndo_get_iflink)(const struct net_device *dev); 1349 * Called to get the iflink value of this device. 1350 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1351 * This function is used to get egress tunnel information for given skb. 1352 * This is useful for retrieving outer tunnel header parameters while 1353 * sampling packet. 1354 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1355 * This function is used to specify the headroom that the skb must 1356 * consider when allocation skb during packet reception. Setting 1357 * appropriate rx headroom value allows avoiding skb head copy on 1358 * forward. Setting a negative value resets the rx headroom to the 1359 * default value. 1360 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1361 * This function is used to set or query state related to XDP on the 1362 * netdevice and manage BPF offload. See definition of 1363 * enum bpf_netdev_command for details. 1364 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1365 * u32 flags); 1366 * This function is used to submit @n XDP packets for transmit on a 1367 * netdevice. Returns number of frames successfully transmitted, frames 1368 * that got dropped are freed/returned via xdp_return_frame(). 1369 * Returns negative number, means general error invoking ndo, meaning 1370 * no frames were xmit'ed and core-caller will free all frames. 1371 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1372 * struct xdp_buff *xdp); 1373 * Get the xmit slave of master device based on the xdp_buff. 1374 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1375 * This function is used to wake up the softirq, ksoftirqd or kthread 1376 * responsible for sending and/or receiving packets on a specific 1377 * queue id bound to an AF_XDP socket. The flags field specifies if 1378 * only RX, only Tx, or both should be woken up using the flags 1379 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1380 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm_kern *p, 1381 * int cmd); 1382 * Add, change, delete or get information on an IPv4 tunnel. 1383 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1384 * If a device is paired with a peer device, return the peer instance. 1385 * The caller must be under RCU read context. 1386 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1387 * Get the forwarding path to reach the real device from the HW destination address 1388 * ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1389 * const struct skb_shared_hwtstamps *hwtstamps, 1390 * bool cycles); 1391 * Get hardware timestamp based on normal/adjustable time or free running 1392 * cycle counter. This function is required if physical clock supports a 1393 * free running cycle counter. 1394 * 1395 * int (*ndo_hwtstamp_get)(struct net_device *dev, 1396 * struct kernel_hwtstamp_config *kernel_config); 1397 * Get the currently configured hardware timestamping parameters for the 1398 * NIC device. 1399 * 1400 * int (*ndo_hwtstamp_set)(struct net_device *dev, 1401 * struct kernel_hwtstamp_config *kernel_config, 1402 * struct netlink_ext_ack *extack); 1403 * Change the hardware timestamping parameters for NIC device. 1404 */ 1405struct net_device_ops { 1406 int (*ndo_init)(struct net_device *dev); 1407 void (*ndo_uninit)(struct net_device *dev); 1408 int (*ndo_open)(struct net_device *dev); 1409 int (*ndo_stop)(struct net_device *dev); 1410 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1411 struct net_device *dev); 1412 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1413 struct net_device *dev, 1414 netdev_features_t features); 1415 u16 (*ndo_select_queue)(struct net_device *dev, 1416 struct sk_buff *skb, 1417 struct net_device *sb_dev); 1418 void (*ndo_change_rx_flags)(struct net_device *dev, 1419 int flags); 1420 void (*ndo_set_rx_mode)(struct net_device *dev); 1421 int (*ndo_set_mac_address)(struct net_device *dev, 1422 void *addr); 1423 int (*ndo_validate_addr)(struct net_device *dev); 1424 int (*ndo_do_ioctl)(struct net_device *dev, 1425 struct ifreq *ifr, int cmd); 1426 int (*ndo_eth_ioctl)(struct net_device *dev, 1427 struct ifreq *ifr, int cmd); 1428 int (*ndo_siocbond)(struct net_device *dev, 1429 struct ifreq *ifr, int cmd); 1430 int (*ndo_siocwandev)(struct net_device *dev, 1431 struct if_settings *ifs); 1432 int (*ndo_siocdevprivate)(struct net_device *dev, 1433 struct ifreq *ifr, 1434 void __user *data, int cmd); 1435 int (*ndo_set_config)(struct net_device *dev, 1436 struct ifmap *map); 1437 int (*ndo_change_mtu)(struct net_device *dev, 1438 int new_mtu); 1439 int (*ndo_neigh_setup)(struct net_device *dev, 1440 struct neigh_parms *); 1441 void (*ndo_tx_timeout) (struct net_device *dev, 1442 unsigned int txqueue); 1443 1444 void (*ndo_get_stats64)(struct net_device *dev, 1445 struct rtnl_link_stats64 *storage); 1446 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1447 int (*ndo_get_offload_stats)(int attr_id, 1448 const struct net_device *dev, 1449 void *attr_data); 1450 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1451 1452 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1453 __be16 proto, u16 vid); 1454 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1455 __be16 proto, u16 vid); 1456#ifdef CONFIG_NET_POLL_CONTROLLER 1457 void (*ndo_poll_controller)(struct net_device *dev); 1458 int (*ndo_netpoll_setup)(struct net_device *dev); 1459 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1460#endif 1461 int (*ndo_set_vf_mac)(struct net_device *dev, 1462 int queue, u8 *mac); 1463 int (*ndo_set_vf_vlan)(struct net_device *dev, 1464 int queue, u16 vlan, 1465 u8 qos, __be16 proto); 1466 int (*ndo_set_vf_rate)(struct net_device *dev, 1467 int vf, int min_tx_rate, 1468 int max_tx_rate); 1469 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1470 int vf, bool setting); 1471 int (*ndo_set_vf_trust)(struct net_device *dev, 1472 int vf, bool setting); 1473 int (*ndo_get_vf_config)(struct net_device *dev, 1474 int vf, 1475 struct ifla_vf_info *ivf); 1476 int (*ndo_set_vf_link_state)(struct net_device *dev, 1477 int vf, int link_state); 1478 int (*ndo_get_vf_stats)(struct net_device *dev, 1479 int vf, 1480 struct ifla_vf_stats 1481 *vf_stats); 1482 int (*ndo_set_vf_port)(struct net_device *dev, 1483 int vf, 1484 struct nlattr *port[]); 1485 int (*ndo_get_vf_port)(struct net_device *dev, 1486 int vf, struct sk_buff *skb); 1487 int (*ndo_get_vf_guid)(struct net_device *dev, 1488 int vf, 1489 struct ifla_vf_guid *node_guid, 1490 struct ifla_vf_guid *port_guid); 1491 int (*ndo_set_vf_guid)(struct net_device *dev, 1492 int vf, u64 guid, 1493 int guid_type); 1494 int (*ndo_set_vf_rss_query_en)( 1495 struct net_device *dev, 1496 int vf, bool setting); 1497 int (*ndo_setup_tc)(struct net_device *dev, 1498 enum tc_setup_type type, 1499 void *type_data); 1500#if IS_ENABLED(CONFIG_FCOE) 1501 int (*ndo_fcoe_enable)(struct net_device *dev); 1502 int (*ndo_fcoe_disable)(struct net_device *dev); 1503 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1504 u16 xid, 1505 struct scatterlist *sgl, 1506 unsigned int sgc); 1507 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1508 u16 xid); 1509 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1510 u16 xid, 1511 struct scatterlist *sgl, 1512 unsigned int sgc); 1513 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1514 struct netdev_fcoe_hbainfo *hbainfo); 1515#endif 1516 1517#if IS_ENABLED(CONFIG_LIBFCOE) 1518#define NETDEV_FCOE_WWNN 0 1519#define NETDEV_FCOE_WWPN 1 1520 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1521 u64 *wwn, int type); 1522#endif 1523 1524#ifdef CONFIG_RFS_ACCEL 1525 int (*ndo_rx_flow_steer)(struct net_device *dev, 1526 const struct sk_buff *skb, 1527 u16 rxq_index, 1528 u32 flow_id); 1529#endif 1530 int (*ndo_add_slave)(struct net_device *dev, 1531 struct net_device *slave_dev, 1532 struct netlink_ext_ack *extack); 1533 int (*ndo_del_slave)(struct net_device *dev, 1534 struct net_device *slave_dev); 1535 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1536 struct sk_buff *skb, 1537 bool all_slaves); 1538 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1539 struct sock *sk); 1540 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1541 netdev_features_t features); 1542 int (*ndo_set_features)(struct net_device *dev, 1543 netdev_features_t features); 1544 int (*ndo_neigh_construct)(struct net_device *dev, 1545 struct neighbour *n); 1546 void (*ndo_neigh_destroy)(struct net_device *dev, 1547 struct neighbour *n); 1548 1549 int (*ndo_fdb_add)(struct ndmsg *ndm, 1550 struct nlattr *tb[], 1551 struct net_device *dev, 1552 const unsigned char *addr, 1553 u16 vid, 1554 u16 flags, 1555 bool *notified, 1556 struct netlink_ext_ack *extack); 1557 int (*ndo_fdb_del)(struct ndmsg *ndm, 1558 struct nlattr *tb[], 1559 struct net_device *dev, 1560 const unsigned char *addr, 1561 u16 vid, 1562 bool *notified, 1563 struct netlink_ext_ack *extack); 1564 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, 1565 struct net_device *dev, 1566 struct netlink_ext_ack *extack); 1567 int (*ndo_fdb_dump)(struct sk_buff *skb, 1568 struct netlink_callback *cb, 1569 struct net_device *dev, 1570 struct net_device *filter_dev, 1571 int *idx); 1572 int (*ndo_fdb_get)(struct sk_buff *skb, 1573 struct nlattr *tb[], 1574 struct net_device *dev, 1575 const unsigned char *addr, 1576 u16 vid, u32 portid, u32 seq, 1577 struct netlink_ext_ack *extack); 1578 int (*ndo_mdb_add)(struct net_device *dev, 1579 struct nlattr *tb[], 1580 u16 nlmsg_flags, 1581 struct netlink_ext_ack *extack); 1582 int (*ndo_mdb_del)(struct net_device *dev, 1583 struct nlattr *tb[], 1584 struct netlink_ext_ack *extack); 1585 int (*ndo_mdb_del_bulk)(struct net_device *dev, 1586 struct nlattr *tb[], 1587 struct netlink_ext_ack *extack); 1588 int (*ndo_mdb_dump)(struct net_device *dev, 1589 struct sk_buff *skb, 1590 struct netlink_callback *cb); 1591 int (*ndo_mdb_get)(struct net_device *dev, 1592 struct nlattr *tb[], u32 portid, 1593 u32 seq, 1594 struct netlink_ext_ack *extack); 1595 int (*ndo_bridge_setlink)(struct net_device *dev, 1596 struct nlmsghdr *nlh, 1597 u16 flags, 1598 struct netlink_ext_ack *extack); 1599 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1600 u32 pid, u32 seq, 1601 struct net_device *dev, 1602 u32 filter_mask, 1603 int nlflags); 1604 int (*ndo_bridge_dellink)(struct net_device *dev, 1605 struct nlmsghdr *nlh, 1606 u16 flags); 1607 int (*ndo_change_carrier)(struct net_device *dev, 1608 bool new_carrier); 1609 int (*ndo_get_phys_port_id)(struct net_device *dev, 1610 struct netdev_phys_item_id *ppid); 1611 int (*ndo_get_port_parent_id)(struct net_device *dev, 1612 struct netdev_phys_item_id *ppid); 1613 int (*ndo_get_phys_port_name)(struct net_device *dev, 1614 char *name, size_t len); 1615 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1616 struct net_device *dev); 1617 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1618 void *priv); 1619 1620 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1621 int queue_index, 1622 u32 maxrate); 1623 int (*ndo_get_iflink)(const struct net_device *dev); 1624 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1625 struct sk_buff *skb); 1626 void (*ndo_set_rx_headroom)(struct net_device *dev, 1627 int needed_headroom); 1628 int (*ndo_bpf)(struct net_device *dev, 1629 struct netdev_bpf *bpf); 1630 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1631 struct xdp_frame **xdp, 1632 u32 flags); 1633 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1634 struct xdp_buff *xdp); 1635 int (*ndo_xsk_wakeup)(struct net_device *dev, 1636 u32 queue_id, u32 flags); 1637 int (*ndo_tunnel_ctl)(struct net_device *dev, 1638 struct ip_tunnel_parm_kern *p, 1639 int cmd); 1640 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1641 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1642 struct net_device_path *path); 1643 ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1644 const struct skb_shared_hwtstamps *hwtstamps, 1645 bool cycles); 1646 int (*ndo_hwtstamp_get)(struct net_device *dev, 1647 struct kernel_hwtstamp_config *kernel_config); 1648 int (*ndo_hwtstamp_set)(struct net_device *dev, 1649 struct kernel_hwtstamp_config *kernel_config, 1650 struct netlink_ext_ack *extack); 1651 1652#if IS_ENABLED(CONFIG_NET_SHAPER) 1653 /** 1654 * @net_shaper_ops: Device shaping offload operations 1655 * see include/net/net_shapers.h 1656 */ 1657 const struct net_shaper_ops *net_shaper_ops; 1658#endif 1659}; 1660 1661/** 1662 * enum netdev_priv_flags - &struct net_device priv_flags 1663 * 1664 * These are the &struct net_device, they are only set internally 1665 * by drivers and used in the kernel. These flags are invisible to 1666 * userspace; this means that the order of these flags can change 1667 * during any kernel release. 1668 * 1669 * You should add bitfield booleans after either net_device::priv_flags 1670 * (hotpath) or ::threaded (slowpath) instead of extending these flags. 1671 * 1672 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1673 * @IFF_EBRIDGE: Ethernet bridging device 1674 * @IFF_BONDING: bonding master or slave 1675 * @IFF_ISATAP: ISATAP interface (RFC4214) 1676 * @IFF_WAN_HDLC: WAN HDLC device 1677 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1678 * release skb->dst 1679 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1680 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1681 * @IFF_MACVLAN_PORT: device used as macvlan port 1682 * @IFF_BRIDGE_PORT: device used as bridge port 1683 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1684 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1685 * @IFF_UNICAST_FLT: Supports unicast filtering 1686 * @IFF_TEAM_PORT: device used as team port 1687 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1688 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1689 * change when it's running 1690 * @IFF_MACVLAN: Macvlan device 1691 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1692 * underlying stacked devices 1693 * @IFF_L3MDEV_MASTER: device is an L3 master device 1694 * @IFF_NO_QUEUE: device can run without qdisc attached 1695 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1696 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1697 * @IFF_TEAM: device is a team device 1698 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1699 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1700 * entity (i.e. the master device for bridged veth) 1701 * @IFF_MACSEC: device is a MACsec device 1702 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1703 * @IFF_FAILOVER: device is a failover master device 1704 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1705 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1706 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf 1707 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1708 * skb_headlen(skb) == 0 (data starts from frag0) 1709 */ 1710enum netdev_priv_flags { 1711 IFF_802_1Q_VLAN = 1<<0, 1712 IFF_EBRIDGE = 1<<1, 1713 IFF_BONDING = 1<<2, 1714 IFF_ISATAP = 1<<3, 1715 IFF_WAN_HDLC = 1<<4, 1716 IFF_XMIT_DST_RELEASE = 1<<5, 1717 IFF_DONT_BRIDGE = 1<<6, 1718 IFF_DISABLE_NETPOLL = 1<<7, 1719 IFF_MACVLAN_PORT = 1<<8, 1720 IFF_BRIDGE_PORT = 1<<9, 1721 IFF_OVS_DATAPATH = 1<<10, 1722 IFF_TX_SKB_SHARING = 1<<11, 1723 IFF_UNICAST_FLT = 1<<12, 1724 IFF_TEAM_PORT = 1<<13, 1725 IFF_SUPP_NOFCS = 1<<14, 1726 IFF_LIVE_ADDR_CHANGE = 1<<15, 1727 IFF_MACVLAN = 1<<16, 1728 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1729 IFF_L3MDEV_MASTER = 1<<18, 1730 IFF_NO_QUEUE = 1<<19, 1731 IFF_OPENVSWITCH = 1<<20, 1732 IFF_L3MDEV_SLAVE = 1<<21, 1733 IFF_TEAM = 1<<22, 1734 IFF_RXFH_CONFIGURED = 1<<23, 1735 IFF_PHONY_HEADROOM = 1<<24, 1736 IFF_MACSEC = 1<<25, 1737 IFF_NO_RX_HANDLER = 1<<26, 1738 IFF_FAILOVER = 1<<27, 1739 IFF_FAILOVER_SLAVE = 1<<28, 1740 IFF_L3MDEV_RX_HANDLER = 1<<29, 1741 IFF_NO_ADDRCONF = BIT_ULL(30), 1742 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31), 1743}; 1744 1745/* Specifies the type of the struct net_device::ml_priv pointer */ 1746enum netdev_ml_priv_type { 1747 ML_PRIV_NONE, 1748 ML_PRIV_CAN, 1749}; 1750 1751enum netdev_stat_type { 1752 NETDEV_PCPU_STAT_NONE, 1753 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */ 1754 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */ 1755 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */ 1756}; 1757 1758enum netdev_reg_state { 1759 NETREG_UNINITIALIZED = 0, 1760 NETREG_REGISTERED, /* completed register_netdevice */ 1761 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1762 NETREG_UNREGISTERED, /* completed unregister todo */ 1763 NETREG_RELEASED, /* called free_netdev */ 1764 NETREG_DUMMY, /* dummy device for NAPI poll */ 1765}; 1766 1767/** 1768 * struct net_device - The DEVICE structure. 1769 * 1770 * Actually, this whole structure is a big mistake. It mixes I/O 1771 * data with strictly "high-level" data, and it has to know about 1772 * almost every data structure used in the INET module. 1773 * 1774 * @priv_flags: flags invisible to userspace defined as bits, see 1775 * enum netdev_priv_flags for the definitions 1776 * @lltx: device supports lockless Tx. Deprecated for real HW 1777 * drivers. Mainly used by logical interfaces, such as 1778 * bonding and tunnels 1779 * @netmem_tx: device support netmem_tx. 1780 * 1781 * @name: This is the first field of the "visible" part of this structure 1782 * (i.e. as seen by users in the "Space.c" file). It is the name 1783 * of the interface. 1784 * 1785 * @name_node: Name hashlist node 1786 * @ifalias: SNMP alias 1787 * @mem_end: Shared memory end 1788 * @mem_start: Shared memory start 1789 * @base_addr: Device I/O address 1790 * @irq: Device IRQ number 1791 * 1792 * @state: Generic network queuing layer state, see netdev_state_t 1793 * @dev_list: The global list of network devices 1794 * @napi_list: List entry used for polling NAPI devices 1795 * @unreg_list: List entry when we are unregistering the 1796 * device; see the function unregister_netdev 1797 * @close_list: List entry used when we are closing the device 1798 * @ptype_all: Device-specific packet handlers for all protocols 1799 * @ptype_specific: Device-specific, protocol-specific packet handlers 1800 * 1801 * @adj_list: Directly linked devices, like slaves for bonding 1802 * @features: Currently active device features 1803 * @hw_features: User-changeable features 1804 * 1805 * @wanted_features: User-requested features 1806 * @vlan_features: Mask of features inheritable by VLAN devices 1807 * 1808 * @hw_enc_features: Mask of features inherited by encapsulating devices 1809 * This field indicates what encapsulation 1810 * offloads the hardware is capable of doing, 1811 * and drivers will need to set them appropriately. 1812 * 1813 * @mpls_features: Mask of features inheritable by MPLS 1814 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1815 * 1816 * @ifindex: interface index 1817 * @group: The group the device belongs to 1818 * 1819 * @stats: Statistics struct, which was left as a legacy, use 1820 * rtnl_link_stats64 instead 1821 * 1822 * @core_stats: core networking counters, 1823 * do not use this in drivers 1824 * @carrier_up_count: Number of times the carrier has been up 1825 * @carrier_down_count: Number of times the carrier has been down 1826 * 1827 * @wireless_handlers: List of functions to handle Wireless Extensions, 1828 * instead of ioctl, 1829 * see <net/iw_handler.h> for details. 1830 * 1831 * @netdev_ops: Includes several pointers to callbacks, 1832 * if one wants to override the ndo_*() functions 1833 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks. 1834 * @xsk_tx_metadata_ops: Includes pointers to AF_XDP TX metadata callbacks. 1835 * @ethtool_ops: Management operations 1836 * @l3mdev_ops: Layer 3 master device operations 1837 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1838 * discovery handling. Necessary for e.g. 6LoWPAN. 1839 * @xfrmdev_ops: Transformation offload operations 1840 * @tlsdev_ops: Transport Layer Security offload operations 1841 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1842 * of Layer 2 headers. 1843 * 1844 * @flags: Interface flags (a la BSD) 1845 * @xdp_features: XDP capability supported by the device 1846 * @gflags: Global flags ( kept as legacy ) 1847 * @priv_len: Size of the ->priv flexible array 1848 * @priv: Flexible array containing private data 1849 * @operstate: RFC2863 operstate 1850 * @link_mode: Mapping policy to operstate 1851 * @if_port: Selectable AUI, TP, ... 1852 * @dma: DMA channel 1853 * @mtu: Interface MTU value 1854 * @min_mtu: Interface Minimum MTU value 1855 * @max_mtu: Interface Maximum MTU value 1856 * @type: Interface hardware type 1857 * @hard_header_len: Maximum hardware header length. 1858 * @min_header_len: Minimum hardware header length 1859 * 1860 * @needed_headroom: Extra headroom the hardware may need, but not in all 1861 * cases can this be guaranteed 1862 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1863 * cases can this be guaranteed. Some cases also use 1864 * LL_MAX_HEADER instead to allocate the skb 1865 * 1866 * interface address info: 1867 * 1868 * @perm_addr: Permanent hw address 1869 * @addr_assign_type: Hw address assignment type 1870 * @addr_len: Hardware address length 1871 * @upper_level: Maximum depth level of upper devices. 1872 * @lower_level: Maximum depth level of lower devices. 1873 * @neigh_priv_len: Used in neigh_alloc() 1874 * @dev_id: Used to differentiate devices that share 1875 * the same link layer address 1876 * @dev_port: Used to differentiate devices that share 1877 * the same function 1878 * @addr_list_lock: XXX: need comments on this one 1879 * @name_assign_type: network interface name assignment type 1880 * @uc_promisc: Counter that indicates promiscuous mode 1881 * has been enabled due to the need to listen to 1882 * additional unicast addresses in a device that 1883 * does not implement ndo_set_rx_mode() 1884 * @uc: unicast mac addresses 1885 * @mc: multicast mac addresses 1886 * @dev_addrs: list of device hw addresses 1887 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1888 * @promiscuity: Number of times the NIC is told to work in 1889 * promiscuous mode; if it becomes 0 the NIC will 1890 * exit promiscuous mode 1891 * @allmulti: Counter, enables or disables allmulticast mode 1892 * 1893 * @vlan_info: VLAN info 1894 * @dsa_ptr: dsa specific data 1895 * @tipc_ptr: TIPC specific data 1896 * @atalk_ptr: AppleTalk link 1897 * @ip_ptr: IPv4 specific data 1898 * @ip6_ptr: IPv6 specific data 1899 * @ax25_ptr: AX.25 specific data 1900 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1901 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1902 * device struct 1903 * @mpls_ptr: mpls_dev struct pointer 1904 * @mctp_ptr: MCTP specific data 1905 * 1906 * @dev_addr: Hw address (before bcast, 1907 * because most packets are unicast) 1908 * 1909 * @_rx: Array of RX queues 1910 * @num_rx_queues: Number of RX queues 1911 * allocated at register_netdev() time 1912 * @real_num_rx_queues: Number of RX queues currently active in device 1913 * @xdp_prog: XDP sockets filter program pointer 1914 * 1915 * @rx_handler: handler for received packets 1916 * @rx_handler_data: XXX: need comments on this one 1917 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing 1918 * @ingress_queue: XXX: need comments on this one 1919 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1920 * @broadcast: hw bcast address 1921 * 1922 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1923 * indexed by RX queue number. Assigned by driver. 1924 * This must only be set if the ndo_rx_flow_steer 1925 * operation is defined 1926 * @index_hlist: Device index hash chain 1927 * 1928 * @_tx: Array of TX queues 1929 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1930 * @real_num_tx_queues: Number of TX queues currently active in device 1931 * @qdisc: Root qdisc from userspace point of view 1932 * @tx_queue_len: Max frames per queue allowed 1933 * @tx_global_lock: XXX: need comments on this one 1934 * @xdp_bulkq: XDP device bulk queue 1935 * @xps_maps: all CPUs/RXQs maps for XPS device 1936 * 1937 * @xps_maps: XXX: need comments on this one 1938 * @tcx_egress: BPF & clsact qdisc specific data for egress processing 1939 * @nf_hooks_egress: netfilter hooks executed for egress packets 1940 * @qdisc_hash: qdisc hash table 1941 * @watchdog_timeo: Represents the timeout that is used by 1942 * the watchdog (see dev_watchdog()) 1943 * @watchdog_timer: List of timers 1944 * 1945 * @proto_down_reason: reason a netdev interface is held down 1946 * @pcpu_refcnt: Number of references to this device 1947 * @dev_refcnt: Number of references to this device 1948 * @refcnt_tracker: Tracker directory for tracked references to this device 1949 * @todo_list: Delayed register/unregister 1950 * @link_watch_list: XXX: need comments on this one 1951 * 1952 * @reg_state: Register/unregister state machine 1953 * @dismantle: Device is going to be freed 1954 * @needs_free_netdev: Should unregister perform free_netdev? 1955 * @priv_destructor: Called from unregister 1956 * @npinfo: XXX: need comments on this one 1957 * @nd_net: Network namespace this network device is inside 1958 * protected by @lock 1959 * 1960 * @ml_priv: Mid-layer private 1961 * @ml_priv_type: Mid-layer private type 1962 * 1963 * @pcpu_stat_type: Type of device statistics which the core should 1964 * allocate/free: none, lstats, tstats, dstats. none 1965 * means the driver is handling statistics allocation/ 1966 * freeing internally. 1967 * @lstats: Loopback statistics: packets, bytes 1968 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes 1969 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes 1970 * 1971 * @garp_port: GARP 1972 * @mrp_port: MRP 1973 * 1974 * @dm_private: Drop monitor private 1975 * 1976 * @dev: Class/net/name entry 1977 * @sysfs_groups: Space for optional device, statistics and wireless 1978 * sysfs groups 1979 * 1980 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1981 * @rtnl_link_ops: Rtnl_link_ops 1982 * @stat_ops: Optional ops for queue-aware statistics 1983 * @queue_mgmt_ops: Optional ops for queue management 1984 * 1985 * @gso_max_size: Maximum size of generic segmentation offload 1986 * @tso_max_size: Device (as in HW) limit on the max TSO request size 1987 * @gso_max_segs: Maximum number of segments that can be passed to the 1988 * NIC for GSO 1989 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count 1990 * @gso_ipv4_max_size: Maximum size of generic segmentation offload, 1991 * for IPv4. 1992 * 1993 * @dcbnl_ops: Data Center Bridging netlink ops 1994 * @num_tc: Number of traffic classes in the net device 1995 * @tc_to_txq: XXX: need comments on this one 1996 * @prio_tc_map: XXX: need comments on this one 1997 * 1998 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1999 * 2000 * @priomap: XXX: need comments on this one 2001 * @link_topo: Physical link topology tracking attached PHYs 2002 * @phydev: Physical device may attach itself 2003 * for hardware timestamping 2004 * @sfp_bus: attached &struct sfp_bus structure. 2005 * 2006 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 2007 * 2008 * @proto_down: protocol port state information can be sent to the 2009 * switch driver and used to set the phys state of the 2010 * switch port. 2011 * 2012 * @threaded: napi threaded mode is enabled 2013 * 2014 * @irq_affinity_auto: driver wants the core to store and re-assign the IRQ 2015 * affinity. Set by netif_enable_irq_affinity(), then 2016 * the driver must create a persistent napi by 2017 * netif_napi_add_config() and finally bind the napi to 2018 * IRQ (via netif_napi_set_irq()). 2019 * 2020 * @rx_cpu_rmap_auto: driver wants the core to manage the ARFS rmap. 2021 * Set by calling netif_enable_cpu_rmap(). 2022 * 2023 * @see_all_hwtstamp_requests: device wants to see calls to 2024 * ndo_hwtstamp_set() for all timestamp requests 2025 * regardless of source, even if those aren't 2026 * HWTSTAMP_SOURCE_NETDEV 2027 * @change_proto_down: device supports setting carrier via IFLA_PROTO_DOWN 2028 * @netns_immutable: interface can't change network namespaces 2029 * @fcoe_mtu: device supports maximum FCoE MTU, 2158 bytes 2030 * 2031 * @net_notifier_list: List of per-net netdev notifier block 2032 * that follow this device when it is moved 2033 * to another network namespace. 2034 * 2035 * @macsec_ops: MACsec offloading ops 2036 * 2037 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 2038 * offload capabilities of the device 2039 * @udp_tunnel_nic: UDP tunnel offload state 2040 * @ethtool: ethtool related state 2041 * @xdp_state: stores info on attached XDP BPF programs 2042 * 2043 * @nested_level: Used as a parameter of spin_lock_nested() of 2044 * dev->addr_list_lock. 2045 * @unlink_list: As netif_addr_lock() can be called recursively, 2046 * keep a list of interfaces to be deleted. 2047 * @gro_max_size: Maximum size of aggregated packet in generic 2048 * receive offload (GRO) 2049 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic 2050 * receive offload (GRO), for IPv4. 2051 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP 2052 * zero copy driver 2053 * 2054 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 2055 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 2056 * @watchdog_dev_tracker: refcount tracker used by watchdog. 2057 * @dev_registered_tracker: tracker for reference held while 2058 * registered 2059 * @offload_xstats_l3: L3 HW stats for this netdevice. 2060 * 2061 * @devlink_port: Pointer to related devlink port structure. 2062 * Assigned by a driver before netdev registration using 2063 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static 2064 * during the time netdevice is registered. 2065 * 2066 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem, 2067 * where the clock is recovered. 2068 * 2069 * @max_pacing_offload_horizon: max EDT offload horizon in nsec. 2070 * @napi_config: An array of napi_config structures containing per-NAPI 2071 * settings. 2072 * @gro_flush_timeout: timeout for GRO layer in NAPI 2073 * @napi_defer_hard_irqs: If not zero, provides a counter that would 2074 * allow to avoid NIC hard IRQ, on busy queues. 2075 * 2076 * @neighbours: List heads pointing to this device's neighbours' 2077 * dev_list, one per address-family. 2078 * @hwprov: Tracks which PTP performs hardware packet time stamping. 2079 * 2080 * FIXME: cleanup struct net_device such that network protocol info 2081 * moves out. 2082 */ 2083 2084struct net_device { 2085 /* Cacheline organization can be found documented in 2086 * Documentation/networking/net_cachelines/net_device.rst. 2087 * Please update the document when adding new fields. 2088 */ 2089 2090 /* TX read-mostly hotpath */ 2091 __cacheline_group_begin(net_device_read_tx); 2092 struct_group(priv_flags_fast, 2093 unsigned long priv_flags:32; 2094 unsigned long lltx:1; 2095 unsigned long netmem_tx:1; 2096 ); 2097 const struct net_device_ops *netdev_ops; 2098 const struct header_ops *header_ops; 2099 struct netdev_queue *_tx; 2100 netdev_features_t gso_partial_features; 2101 unsigned int real_num_tx_queues; 2102 unsigned int gso_max_size; 2103 unsigned int gso_ipv4_max_size; 2104 u16 gso_max_segs; 2105 s16 num_tc; 2106 /* Note : dev->mtu is often read without holding a lock. 2107 * Writers usually hold RTNL. 2108 * It is recommended to use READ_ONCE() to annotate the reads, 2109 * and to use WRITE_ONCE() to annotate the writes. 2110 */ 2111 unsigned int mtu; 2112 unsigned short needed_headroom; 2113 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2114#ifdef CONFIG_XPS 2115 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2116#endif 2117#ifdef CONFIG_NETFILTER_EGRESS 2118 struct nf_hook_entries __rcu *nf_hooks_egress; 2119#endif 2120#ifdef CONFIG_NET_XGRESS 2121 struct bpf_mprog_entry __rcu *tcx_egress; 2122#endif 2123 __cacheline_group_end(net_device_read_tx); 2124 2125 /* TXRX read-mostly hotpath */ 2126 __cacheline_group_begin(net_device_read_txrx); 2127 union { 2128 struct pcpu_lstats __percpu *lstats; 2129 struct pcpu_sw_netstats __percpu *tstats; 2130 struct pcpu_dstats __percpu *dstats; 2131 }; 2132 unsigned long state; 2133 unsigned int flags; 2134 unsigned short hard_header_len; 2135 netdev_features_t features; 2136 struct inet6_dev __rcu *ip6_ptr; 2137 __cacheline_group_end(net_device_read_txrx); 2138 2139 /* RX read-mostly hotpath */ 2140 __cacheline_group_begin(net_device_read_rx); 2141 struct bpf_prog __rcu *xdp_prog; 2142 struct list_head ptype_specific; 2143 int ifindex; 2144 unsigned int real_num_rx_queues; 2145 struct netdev_rx_queue *_rx; 2146 unsigned int gro_max_size; 2147 unsigned int gro_ipv4_max_size; 2148 rx_handler_func_t __rcu *rx_handler; 2149 void __rcu *rx_handler_data; 2150 possible_net_t nd_net; 2151#ifdef CONFIG_NETPOLL 2152 struct netpoll_info __rcu *npinfo; 2153#endif 2154#ifdef CONFIG_NET_XGRESS 2155 struct bpf_mprog_entry __rcu *tcx_ingress; 2156#endif 2157 __cacheline_group_end(net_device_read_rx); 2158 2159 char name[IFNAMSIZ]; 2160 struct netdev_name_node *name_node; 2161 struct dev_ifalias __rcu *ifalias; 2162 /* 2163 * I/O specific fields 2164 * FIXME: Merge these and struct ifmap into one 2165 */ 2166 unsigned long mem_end; 2167 unsigned long mem_start; 2168 unsigned long base_addr; 2169 2170 /* 2171 * Some hardware also needs these fields (state,dev_list, 2172 * napi_list,unreg_list,close_list) but they are not 2173 * part of the usual set specified in Space.c. 2174 */ 2175 2176 2177 struct list_head dev_list; 2178 struct list_head napi_list; 2179 struct list_head unreg_list; 2180 struct list_head close_list; 2181 struct list_head ptype_all; 2182 2183 struct { 2184 struct list_head upper; 2185 struct list_head lower; 2186 } adj_list; 2187 2188 /* Read-mostly cache-line for fast-path access */ 2189 xdp_features_t xdp_features; 2190 const struct xdp_metadata_ops *xdp_metadata_ops; 2191 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops; 2192 unsigned short gflags; 2193 2194 unsigned short needed_tailroom; 2195 2196 netdev_features_t hw_features; 2197 netdev_features_t wanted_features; 2198 netdev_features_t vlan_features; 2199 netdev_features_t hw_enc_features; 2200 netdev_features_t mpls_features; 2201 2202 unsigned int min_mtu; 2203 unsigned int max_mtu; 2204 unsigned short type; 2205 unsigned char min_header_len; 2206 unsigned char name_assign_type; 2207 2208 int group; 2209 2210 struct net_device_stats stats; /* not used by modern drivers */ 2211 2212 struct net_device_core_stats __percpu *core_stats; 2213 2214 /* Stats to monitor link on/off, flapping */ 2215 atomic_t carrier_up_count; 2216 atomic_t carrier_down_count; 2217 2218#ifdef CONFIG_WIRELESS_EXT 2219 const struct iw_handler_def *wireless_handlers; 2220#endif 2221 const struct ethtool_ops *ethtool_ops; 2222#ifdef CONFIG_NET_L3_MASTER_DEV 2223 const struct l3mdev_ops *l3mdev_ops; 2224#endif 2225#if IS_ENABLED(CONFIG_IPV6) 2226 const struct ndisc_ops *ndisc_ops; 2227#endif 2228 2229#ifdef CONFIG_XFRM_OFFLOAD 2230 const struct xfrmdev_ops *xfrmdev_ops; 2231#endif 2232 2233#if IS_ENABLED(CONFIG_TLS_DEVICE) 2234 const struct tlsdev_ops *tlsdev_ops; 2235#endif 2236 2237 unsigned int operstate; 2238 unsigned char link_mode; 2239 2240 unsigned char if_port; 2241 unsigned char dma; 2242 2243 /* Interface address info. */ 2244 unsigned char perm_addr[MAX_ADDR_LEN]; 2245 unsigned char addr_assign_type; 2246 unsigned char addr_len; 2247 unsigned char upper_level; 2248 unsigned char lower_level; 2249 2250 unsigned short neigh_priv_len; 2251 unsigned short dev_id; 2252 unsigned short dev_port; 2253 int irq; 2254 u32 priv_len; 2255 2256 spinlock_t addr_list_lock; 2257 2258 struct netdev_hw_addr_list uc; 2259 struct netdev_hw_addr_list mc; 2260 struct netdev_hw_addr_list dev_addrs; 2261 2262#ifdef CONFIG_SYSFS 2263 struct kset *queues_kset; 2264#endif 2265#ifdef CONFIG_LOCKDEP 2266 struct list_head unlink_list; 2267#endif 2268 unsigned int promiscuity; 2269 unsigned int allmulti; 2270 bool uc_promisc; 2271#ifdef CONFIG_LOCKDEP 2272 unsigned char nested_level; 2273#endif 2274 2275 2276 /* Protocol-specific pointers */ 2277 struct in_device __rcu *ip_ptr; 2278 /** @fib_nh_head: nexthops associated with this netdev */ 2279 struct hlist_head fib_nh_head; 2280 2281#if IS_ENABLED(CONFIG_VLAN_8021Q) 2282 struct vlan_info __rcu *vlan_info; 2283#endif 2284#if IS_ENABLED(CONFIG_NET_DSA) 2285 struct dsa_port *dsa_ptr; 2286#endif 2287#if IS_ENABLED(CONFIG_TIPC) 2288 struct tipc_bearer __rcu *tipc_ptr; 2289#endif 2290#if IS_ENABLED(CONFIG_ATALK) 2291 void *atalk_ptr; 2292#endif 2293#if IS_ENABLED(CONFIG_AX25) 2294 struct ax25_dev __rcu *ax25_ptr; 2295#endif 2296#if IS_ENABLED(CONFIG_CFG80211) 2297 struct wireless_dev *ieee80211_ptr; 2298#endif 2299#if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2300 struct wpan_dev *ieee802154_ptr; 2301#endif 2302#if IS_ENABLED(CONFIG_MPLS_ROUTING) 2303 struct mpls_dev __rcu *mpls_ptr; 2304#endif 2305#if IS_ENABLED(CONFIG_MCTP) 2306 struct mctp_dev __rcu *mctp_ptr; 2307#endif 2308 2309/* 2310 * Cache lines mostly used on receive path (including eth_type_trans()) 2311 */ 2312 /* Interface address info used in eth_type_trans() */ 2313 const unsigned char *dev_addr; 2314 2315 unsigned int num_rx_queues; 2316#define GRO_LEGACY_MAX_SIZE 65536u 2317/* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2318 * and shinfo->gso_segs is a 16bit field. 2319 */ 2320#define GRO_MAX_SIZE (8 * 65535u) 2321 unsigned int xdp_zc_max_segs; 2322 struct netdev_queue __rcu *ingress_queue; 2323#ifdef CONFIG_NETFILTER_INGRESS 2324 struct nf_hook_entries __rcu *nf_hooks_ingress; 2325#endif 2326 2327 unsigned char broadcast[MAX_ADDR_LEN]; 2328#ifdef CONFIG_RFS_ACCEL 2329 struct cpu_rmap *rx_cpu_rmap; 2330#endif 2331 struct hlist_node index_hlist; 2332 2333/* 2334 * Cache lines mostly used on transmit path 2335 */ 2336 unsigned int num_tx_queues; 2337 struct Qdisc __rcu *qdisc; 2338 unsigned int tx_queue_len; 2339 spinlock_t tx_global_lock; 2340 2341 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2342 2343#ifdef CONFIG_NET_SCHED 2344 DECLARE_HASHTABLE (qdisc_hash, 4); 2345#endif 2346 /* These may be needed for future network-power-down code. */ 2347 struct timer_list watchdog_timer; 2348 int watchdog_timeo; 2349 2350 u32 proto_down_reason; 2351 2352 struct list_head todo_list; 2353 2354#ifdef CONFIG_PCPU_DEV_REFCNT 2355 int __percpu *pcpu_refcnt; 2356#else 2357 refcount_t dev_refcnt; 2358#endif 2359 struct ref_tracker_dir refcnt_tracker; 2360 2361 struct list_head link_watch_list; 2362 2363 u8 reg_state; 2364 2365 bool dismantle; 2366 2367 /** @moving_ns: device is changing netns, protected by @lock */ 2368 bool moving_ns; 2369 /** @rtnl_link_initializing: Device being created, suppress events */ 2370 bool rtnl_link_initializing; 2371 2372 bool needs_free_netdev; 2373 void (*priv_destructor)(struct net_device *dev); 2374 2375 /* mid-layer private */ 2376 void *ml_priv; 2377 enum netdev_ml_priv_type ml_priv_type; 2378 2379 enum netdev_stat_type pcpu_stat_type:8; 2380 2381#if IS_ENABLED(CONFIG_GARP) 2382 struct garp_port __rcu *garp_port; 2383#endif 2384#if IS_ENABLED(CONFIG_MRP) 2385 struct mrp_port __rcu *mrp_port; 2386#endif 2387#if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2388 struct dm_hw_stat_delta __rcu *dm_private; 2389#endif 2390 struct device dev; 2391 const struct attribute_group *sysfs_groups[4]; 2392 const struct attribute_group *sysfs_rx_queue_group; 2393 2394 const struct rtnl_link_ops *rtnl_link_ops; 2395 2396 const struct netdev_stat_ops *stat_ops; 2397 2398 const struct netdev_queue_mgmt_ops *queue_mgmt_ops; 2399 2400 /* for setting kernel sock attribute on TCP connection setup */ 2401#define GSO_MAX_SEGS 65535u 2402#define GSO_LEGACY_MAX_SIZE 65536u 2403/* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2404 * and shinfo->gso_segs is a 16bit field. 2405 */ 2406#define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2407 2408#define TSO_LEGACY_MAX_SIZE 65536 2409#define TSO_MAX_SIZE UINT_MAX 2410 unsigned int tso_max_size; 2411#define TSO_MAX_SEGS U16_MAX 2412 u16 tso_max_segs; 2413 2414#ifdef CONFIG_DCB 2415 const struct dcbnl_rtnl_ops *dcbnl_ops; 2416#endif 2417 u8 prio_tc_map[TC_BITMASK + 1]; 2418 2419#if IS_ENABLED(CONFIG_FCOE) 2420 unsigned int fcoe_ddp_xid; 2421#endif 2422#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2423 struct netprio_map __rcu *priomap; 2424#endif 2425 struct phy_link_topology *link_topo; 2426 struct phy_device *phydev; 2427 struct sfp_bus *sfp_bus; 2428 struct lock_class_key *qdisc_tx_busylock; 2429 bool proto_down; 2430 bool threaded; 2431 bool irq_affinity_auto; 2432 bool rx_cpu_rmap_auto; 2433 2434 /* priv_flags_slow, ungrouped to save space */ 2435 unsigned long see_all_hwtstamp_requests:1; 2436 unsigned long change_proto_down:1; 2437 unsigned long netns_immutable:1; 2438 unsigned long fcoe_mtu:1; 2439 2440 struct list_head net_notifier_list; 2441 2442#if IS_ENABLED(CONFIG_MACSEC) 2443 /* MACsec management functions */ 2444 const struct macsec_ops *macsec_ops; 2445#endif 2446 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2447 struct udp_tunnel_nic *udp_tunnel_nic; 2448 2449 /** @cfg: net_device queue-related configuration */ 2450 struct netdev_config *cfg; 2451 /** 2452 * @cfg_pending: same as @cfg but when device is being actively 2453 * reconfigured includes any changes to the configuration 2454 * requested by the user, but which may or may not be rejected. 2455 */ 2456 struct netdev_config *cfg_pending; 2457 struct ethtool_netdev_state *ethtool; 2458 2459 /* protected by rtnl_lock */ 2460 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2461 2462 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2463 netdevice_tracker linkwatch_dev_tracker; 2464 netdevice_tracker watchdog_dev_tracker; 2465 netdevice_tracker dev_registered_tracker; 2466 struct rtnl_hw_stats64 *offload_xstats_l3; 2467 2468 struct devlink_port *devlink_port; 2469 2470#if IS_ENABLED(CONFIG_DPLL) 2471 struct dpll_pin __rcu *dpll_pin; 2472#endif 2473#if IS_ENABLED(CONFIG_PAGE_POOL) 2474 /** @page_pools: page pools created for this netdevice */ 2475 struct hlist_head page_pools; 2476#endif 2477 2478 /** @irq_moder: dim parameters used if IS_ENABLED(CONFIG_DIMLIB). */ 2479 struct dim_irq_moder *irq_moder; 2480 2481 u64 max_pacing_offload_horizon; 2482 struct napi_config *napi_config; 2483 unsigned long gro_flush_timeout; 2484 u32 napi_defer_hard_irqs; 2485 2486 /** 2487 * @up: copy of @state's IFF_UP, but safe to read with just @lock. 2488 * May report false negatives while the device is being opened 2489 * or closed (@lock does not protect .ndo_open, or .ndo_close). 2490 */ 2491 bool up; 2492 2493 /** 2494 * @request_ops_lock: request the core to run all @netdev_ops and 2495 * @ethtool_ops under the @lock. 2496 */ 2497 bool request_ops_lock; 2498 2499 /** 2500 * @lock: netdev-scope lock, protects a small selection of fields. 2501 * Should always be taken using netdev_lock() / netdev_unlock() helpers. 2502 * Drivers are free to use it for other protection. 2503 * 2504 * For the drivers that implement shaper or queue API, the scope 2505 * of this lock is expanded to cover most ndo/queue/ethtool/sysfs 2506 * operations. Drivers may opt-in to this behavior by setting 2507 * @request_ops_lock. 2508 * 2509 * @lock protection mixes with rtnl_lock in multiple ways, fields are 2510 * either: 2511 * 2512 * - simply protected by the instance @lock; 2513 * 2514 * - double protected - writers hold both locks, readers hold either; 2515 * 2516 * - ops protected - protected by the lock held around the NDOs 2517 * and other callbacks, that is the instance lock on devices for 2518 * which netdev_need_ops_lock() returns true, otherwise by rtnl_lock; 2519 * 2520 * - double ops protected - always protected by rtnl_lock but for 2521 * devices for which netdev_need_ops_lock() returns true - also 2522 * the instance lock. 2523 * 2524 * Simply protects: 2525 * @gro_flush_timeout, @napi_defer_hard_irqs, @napi_list, 2526 * @net_shaper_hierarchy, @reg_state, @threaded 2527 * 2528 * Double protects: 2529 * @up, @moving_ns, @nd_net, @xdp_features 2530 * 2531 * Double ops protects: 2532 * @real_num_rx_queues, @real_num_tx_queues 2533 * 2534 * Also protects some fields in: 2535 * struct napi_struct, struct netdev_queue, struct netdev_rx_queue 2536 * 2537 * Ordering: take after rtnl_lock. 2538 */ 2539 struct mutex lock; 2540 2541#if IS_ENABLED(CONFIG_NET_SHAPER) 2542 /** 2543 * @net_shaper_hierarchy: data tracking the current shaper status 2544 * see include/net/net_shapers.h 2545 */ 2546 struct net_shaper_hierarchy *net_shaper_hierarchy; 2547#endif 2548 2549 struct hlist_head neighbours[NEIGH_NR_TABLES]; 2550 2551 struct hwtstamp_provider __rcu *hwprov; 2552 2553 u8 priv[] ____cacheline_aligned 2554 __counted_by(priv_len); 2555} ____cacheline_aligned; 2556#define to_net_dev(d) container_of(d, struct net_device, dev) 2557 2558/* 2559 * Driver should use this to assign devlink port instance to a netdevice 2560 * before it registers the netdevice. Therefore devlink_port is static 2561 * during the netdev lifetime after it is registered. 2562 */ 2563#define SET_NETDEV_DEVLINK_PORT(dev, port) \ 2564({ \ 2565 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \ 2566 ((dev)->devlink_port = (port)); \ 2567}) 2568 2569static inline bool netif_elide_gro(const struct net_device *dev) 2570{ 2571 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2572 return true; 2573 return false; 2574} 2575 2576#define NETDEV_ALIGN 32 2577 2578static inline 2579int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2580{ 2581 return dev->prio_tc_map[prio & TC_BITMASK]; 2582} 2583 2584static inline 2585int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2586{ 2587 if (tc >= dev->num_tc) 2588 return -EINVAL; 2589 2590 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2591 return 0; 2592} 2593 2594int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2595void netdev_reset_tc(struct net_device *dev); 2596int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2597int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2598 2599static inline 2600int netdev_get_num_tc(struct net_device *dev) 2601{ 2602 return dev->num_tc; 2603} 2604 2605static inline void net_prefetch(void *p) 2606{ 2607 prefetch(p); 2608#if L1_CACHE_BYTES < 128 2609 prefetch((u8 *)p + L1_CACHE_BYTES); 2610#endif 2611} 2612 2613static inline void net_prefetchw(void *p) 2614{ 2615 prefetchw(p); 2616#if L1_CACHE_BYTES < 128 2617 prefetchw((u8 *)p + L1_CACHE_BYTES); 2618#endif 2619} 2620 2621void netdev_unbind_sb_channel(struct net_device *dev, 2622 struct net_device *sb_dev); 2623int netdev_bind_sb_channel_queue(struct net_device *dev, 2624 struct net_device *sb_dev, 2625 u8 tc, u16 count, u16 offset); 2626int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2627static inline int netdev_get_sb_channel(struct net_device *dev) 2628{ 2629 return max_t(int, -dev->num_tc, 0); 2630} 2631 2632static inline 2633struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2634 unsigned int index) 2635{ 2636 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues); 2637 return &dev->_tx[index]; 2638} 2639 2640static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2641 const struct sk_buff *skb) 2642{ 2643 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2644} 2645 2646static inline void netdev_for_each_tx_queue(struct net_device *dev, 2647 void (*f)(struct net_device *, 2648 struct netdev_queue *, 2649 void *), 2650 void *arg) 2651{ 2652 unsigned int i; 2653 2654 for (i = 0; i < dev->num_tx_queues; i++) 2655 f(dev, &dev->_tx[i], arg); 2656} 2657 2658u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2659 struct net_device *sb_dev); 2660struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2661 struct sk_buff *skb, 2662 struct net_device *sb_dev); 2663 2664/* returns the headroom that the master device needs to take in account 2665 * when forwarding to this dev 2666 */ 2667static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2668{ 2669 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2670} 2671 2672static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2673{ 2674 if (dev->netdev_ops->ndo_set_rx_headroom) 2675 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2676} 2677 2678/* set the device rx headroom to the dev's default */ 2679static inline void netdev_reset_rx_headroom(struct net_device *dev) 2680{ 2681 netdev_set_rx_headroom(dev, -1); 2682} 2683 2684static inline void *netdev_get_ml_priv(struct net_device *dev, 2685 enum netdev_ml_priv_type type) 2686{ 2687 if (dev->ml_priv_type != type) 2688 return NULL; 2689 2690 return dev->ml_priv; 2691} 2692 2693static inline void netdev_set_ml_priv(struct net_device *dev, 2694 void *ml_priv, 2695 enum netdev_ml_priv_type type) 2696{ 2697 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2698 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2699 dev->ml_priv_type, type); 2700 WARN(!dev->ml_priv_type && dev->ml_priv, 2701 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2702 2703 dev->ml_priv = ml_priv; 2704 dev->ml_priv_type = type; 2705} 2706 2707/* 2708 * Net namespace inlines 2709 */ 2710static inline 2711struct net *dev_net(const struct net_device *dev) 2712{ 2713 return read_pnet(&dev->nd_net); 2714} 2715 2716static inline 2717struct net *dev_net_rcu(const struct net_device *dev) 2718{ 2719 return read_pnet_rcu(&dev->nd_net); 2720} 2721 2722static inline 2723void dev_net_set(struct net_device *dev, struct net *net) 2724{ 2725 write_pnet(&dev->nd_net, net); 2726} 2727 2728/** 2729 * netdev_priv - access network device private data 2730 * @dev: network device 2731 * 2732 * Get network device private data 2733 */ 2734static inline void *netdev_priv(const struct net_device *dev) 2735{ 2736 return (void *)dev->priv; 2737} 2738 2739/* Set the sysfs physical device reference for the network logical device 2740 * if set prior to registration will cause a symlink during initialization. 2741 */ 2742#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2743 2744/* Set the sysfs device type for the network logical device to allow 2745 * fine-grained identification of different network device types. For 2746 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2747 */ 2748#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2749 2750void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index, 2751 enum netdev_queue_type type, 2752 struct napi_struct *napi); 2753 2754static inline void netdev_lock(struct net_device *dev) 2755{ 2756 mutex_lock(&dev->lock); 2757} 2758 2759static inline void netdev_unlock(struct net_device *dev) 2760{ 2761 mutex_unlock(&dev->lock); 2762} 2763/* Additional netdev_lock()-related helpers are in net/netdev_lock.h */ 2764 2765void netif_napi_set_irq_locked(struct napi_struct *napi, int irq); 2766 2767static inline void netif_napi_set_irq(struct napi_struct *napi, int irq) 2768{ 2769 netdev_lock(napi->dev); 2770 netif_napi_set_irq_locked(napi, irq); 2771 netdev_unlock(napi->dev); 2772} 2773 2774/* Default NAPI poll() weight 2775 * Device drivers are strongly advised to not use bigger value 2776 */ 2777#define NAPI_POLL_WEIGHT 64 2778 2779void netif_napi_add_weight_locked(struct net_device *dev, 2780 struct napi_struct *napi, 2781 int (*poll)(struct napi_struct *, int), 2782 int weight); 2783 2784static inline void 2785netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2786 int (*poll)(struct napi_struct *, int), int weight) 2787{ 2788 netdev_lock(dev); 2789 netif_napi_add_weight_locked(dev, napi, poll, weight); 2790 netdev_unlock(dev); 2791} 2792 2793/** 2794 * netif_napi_add() - initialize a NAPI context 2795 * @dev: network device 2796 * @napi: NAPI context 2797 * @poll: polling function 2798 * 2799 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2800 * *any* of the other NAPI-related functions. 2801 */ 2802static inline void 2803netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2804 int (*poll)(struct napi_struct *, int)) 2805{ 2806 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2807} 2808 2809static inline void 2810netif_napi_add_locked(struct net_device *dev, struct napi_struct *napi, 2811 int (*poll)(struct napi_struct *, int)) 2812{ 2813 netif_napi_add_weight_locked(dev, napi, poll, NAPI_POLL_WEIGHT); 2814} 2815 2816static inline void 2817netif_napi_add_tx_weight(struct net_device *dev, 2818 struct napi_struct *napi, 2819 int (*poll)(struct napi_struct *, int), 2820 int weight) 2821{ 2822 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2823 netif_napi_add_weight(dev, napi, poll, weight); 2824} 2825 2826static inline void 2827netif_napi_add_config_locked(struct net_device *dev, struct napi_struct *napi, 2828 int (*poll)(struct napi_struct *, int), int index) 2829{ 2830 napi->index = index; 2831 napi->config = &dev->napi_config[index]; 2832 netif_napi_add_weight_locked(dev, napi, poll, NAPI_POLL_WEIGHT); 2833} 2834 2835/** 2836 * netif_napi_add_config - initialize a NAPI context with persistent config 2837 * @dev: network device 2838 * @napi: NAPI context 2839 * @poll: polling function 2840 * @index: the NAPI index 2841 */ 2842static inline void 2843netif_napi_add_config(struct net_device *dev, struct napi_struct *napi, 2844 int (*poll)(struct napi_struct *, int), int index) 2845{ 2846 netdev_lock(dev); 2847 netif_napi_add_config_locked(dev, napi, poll, index); 2848 netdev_unlock(dev); 2849} 2850 2851/** 2852 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2853 * @dev: network device 2854 * @napi: NAPI context 2855 * @poll: polling function 2856 * 2857 * This variant of netif_napi_add() should be used from drivers using NAPI 2858 * to exclusively poll a TX queue. 2859 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2860 */ 2861static inline void netif_napi_add_tx(struct net_device *dev, 2862 struct napi_struct *napi, 2863 int (*poll)(struct napi_struct *, int)) 2864{ 2865 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2866} 2867 2868void __netif_napi_del_locked(struct napi_struct *napi); 2869 2870/** 2871 * __netif_napi_del - remove a NAPI context 2872 * @napi: NAPI context 2873 * 2874 * Warning: caller must observe RCU grace period before freeing memory 2875 * containing @napi. Drivers might want to call this helper to combine 2876 * all the needed RCU grace periods into a single one. 2877 */ 2878static inline void __netif_napi_del(struct napi_struct *napi) 2879{ 2880 netdev_lock(napi->dev); 2881 __netif_napi_del_locked(napi); 2882 netdev_unlock(napi->dev); 2883} 2884 2885static inline void netif_napi_del_locked(struct napi_struct *napi) 2886{ 2887 __netif_napi_del_locked(napi); 2888 synchronize_net(); 2889} 2890 2891/** 2892 * netif_napi_del - remove a NAPI context 2893 * @napi: NAPI context 2894 * 2895 * netif_napi_del() removes a NAPI context from the network device NAPI list 2896 */ 2897static inline void netif_napi_del(struct napi_struct *napi) 2898{ 2899 __netif_napi_del(napi); 2900 synchronize_net(); 2901} 2902 2903int netif_enable_cpu_rmap(struct net_device *dev, unsigned int num_irqs); 2904void netif_set_affinity_auto(struct net_device *dev); 2905 2906struct packet_type { 2907 __be16 type; /* This is really htons(ether_type). */ 2908 bool ignore_outgoing; 2909 struct net_device *dev; /* NULL is wildcarded here */ 2910 netdevice_tracker dev_tracker; 2911 int (*func) (struct sk_buff *, 2912 struct net_device *, 2913 struct packet_type *, 2914 struct net_device *); 2915 void (*list_func) (struct list_head *, 2916 struct packet_type *, 2917 struct net_device *); 2918 bool (*id_match)(struct packet_type *ptype, 2919 struct sock *sk); 2920 struct net *af_packet_net; 2921 void *af_packet_priv; 2922 struct list_head list; 2923}; 2924 2925struct offload_callbacks { 2926 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2927 netdev_features_t features); 2928 struct sk_buff *(*gro_receive)(struct list_head *head, 2929 struct sk_buff *skb); 2930 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2931}; 2932 2933struct packet_offload { 2934 __be16 type; /* This is really htons(ether_type). */ 2935 u16 priority; 2936 struct offload_callbacks callbacks; 2937 struct list_head list; 2938}; 2939 2940/* often modified stats are per-CPU, other are shared (netdev->stats) */ 2941struct pcpu_sw_netstats { 2942 u64_stats_t rx_packets; 2943 u64_stats_t rx_bytes; 2944 u64_stats_t tx_packets; 2945 u64_stats_t tx_bytes; 2946 struct u64_stats_sync syncp; 2947} __aligned(4 * sizeof(u64)); 2948 2949struct pcpu_dstats { 2950 u64_stats_t rx_packets; 2951 u64_stats_t rx_bytes; 2952 u64_stats_t tx_packets; 2953 u64_stats_t tx_bytes; 2954 u64_stats_t rx_drops; 2955 u64_stats_t tx_drops; 2956 struct u64_stats_sync syncp; 2957} __aligned(8 * sizeof(u64)); 2958 2959struct pcpu_lstats { 2960 u64_stats_t packets; 2961 u64_stats_t bytes; 2962 struct u64_stats_sync syncp; 2963} __aligned(2 * sizeof(u64)); 2964 2965void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2966 2967static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2968{ 2969 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2970 2971 u64_stats_update_begin(&tstats->syncp); 2972 u64_stats_add(&tstats->rx_bytes, len); 2973 u64_stats_inc(&tstats->rx_packets); 2974 u64_stats_update_end(&tstats->syncp); 2975} 2976 2977static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2978 unsigned int packets, 2979 unsigned int len) 2980{ 2981 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2982 2983 u64_stats_update_begin(&tstats->syncp); 2984 u64_stats_add(&tstats->tx_bytes, len); 2985 u64_stats_add(&tstats->tx_packets, packets); 2986 u64_stats_update_end(&tstats->syncp); 2987} 2988 2989static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2990{ 2991 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2992 2993 u64_stats_update_begin(&lstats->syncp); 2994 u64_stats_add(&lstats->bytes, len); 2995 u64_stats_inc(&lstats->packets); 2996 u64_stats_update_end(&lstats->syncp); 2997} 2998 2999static inline void dev_dstats_rx_add(struct net_device *dev, 3000 unsigned int len) 3001{ 3002 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3003 3004 u64_stats_update_begin(&dstats->syncp); 3005 u64_stats_inc(&dstats->rx_packets); 3006 u64_stats_add(&dstats->rx_bytes, len); 3007 u64_stats_update_end(&dstats->syncp); 3008} 3009 3010static inline void dev_dstats_rx_dropped(struct net_device *dev) 3011{ 3012 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3013 3014 u64_stats_update_begin(&dstats->syncp); 3015 u64_stats_inc(&dstats->rx_drops); 3016 u64_stats_update_end(&dstats->syncp); 3017} 3018 3019static inline void dev_dstats_tx_add(struct net_device *dev, 3020 unsigned int len) 3021{ 3022 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3023 3024 u64_stats_update_begin(&dstats->syncp); 3025 u64_stats_inc(&dstats->tx_packets); 3026 u64_stats_add(&dstats->tx_bytes, len); 3027 u64_stats_update_end(&dstats->syncp); 3028} 3029 3030static inline void dev_dstats_tx_dropped(struct net_device *dev) 3031{ 3032 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3033 3034 u64_stats_update_begin(&dstats->syncp); 3035 u64_stats_inc(&dstats->tx_drops); 3036 u64_stats_update_end(&dstats->syncp); 3037} 3038 3039#define __netdev_alloc_pcpu_stats(type, gfp) \ 3040({ \ 3041 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 3042 if (pcpu_stats) { \ 3043 int __cpu; \ 3044 for_each_possible_cpu(__cpu) { \ 3045 typeof(type) *stat; \ 3046 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 3047 u64_stats_init(&stat->syncp); \ 3048 } \ 3049 } \ 3050 pcpu_stats; \ 3051}) 3052 3053#define netdev_alloc_pcpu_stats(type) \ 3054 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 3055 3056#define devm_netdev_alloc_pcpu_stats(dev, type) \ 3057({ \ 3058 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 3059 if (pcpu_stats) { \ 3060 int __cpu; \ 3061 for_each_possible_cpu(__cpu) { \ 3062 typeof(type) *stat; \ 3063 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 3064 u64_stats_init(&stat->syncp); \ 3065 } \ 3066 } \ 3067 pcpu_stats; \ 3068}) 3069 3070enum netdev_lag_tx_type { 3071 NETDEV_LAG_TX_TYPE_UNKNOWN, 3072 NETDEV_LAG_TX_TYPE_RANDOM, 3073 NETDEV_LAG_TX_TYPE_BROADCAST, 3074 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 3075 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 3076 NETDEV_LAG_TX_TYPE_HASH, 3077}; 3078 3079enum netdev_lag_hash { 3080 NETDEV_LAG_HASH_NONE, 3081 NETDEV_LAG_HASH_L2, 3082 NETDEV_LAG_HASH_L34, 3083 NETDEV_LAG_HASH_L23, 3084 NETDEV_LAG_HASH_E23, 3085 NETDEV_LAG_HASH_E34, 3086 NETDEV_LAG_HASH_VLAN_SRCMAC, 3087 NETDEV_LAG_HASH_UNKNOWN, 3088}; 3089 3090struct netdev_lag_upper_info { 3091 enum netdev_lag_tx_type tx_type; 3092 enum netdev_lag_hash hash_type; 3093}; 3094 3095struct netdev_lag_lower_state_info { 3096 u8 link_up : 1, 3097 tx_enabled : 1; 3098}; 3099 3100#include <linux/notifier.h> 3101 3102/* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 3103 * and the rtnetlink notification exclusion list in rtnetlink_event() when 3104 * adding new types. 3105 */ 3106enum netdev_cmd { 3107 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 3108 NETDEV_DOWN, 3109 NETDEV_REBOOT, /* Tell a protocol stack a network interface 3110 detected a hardware crash and restarted 3111 - we can use this eg to kick tcp sessions 3112 once done */ 3113 NETDEV_CHANGE, /* Notify device state change */ 3114 NETDEV_REGISTER, 3115 NETDEV_UNREGISTER, 3116 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 3117 NETDEV_CHANGEADDR, /* notify after the address change */ 3118 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 3119 NETDEV_GOING_DOWN, 3120 NETDEV_CHANGENAME, 3121 NETDEV_FEAT_CHANGE, 3122 NETDEV_BONDING_FAILOVER, 3123 NETDEV_PRE_UP, 3124 NETDEV_PRE_TYPE_CHANGE, 3125 NETDEV_POST_TYPE_CHANGE, 3126 NETDEV_POST_INIT, 3127 NETDEV_PRE_UNINIT, 3128 NETDEV_RELEASE, 3129 NETDEV_NOTIFY_PEERS, 3130 NETDEV_JOIN, 3131 NETDEV_CHANGEUPPER, 3132 NETDEV_RESEND_IGMP, 3133 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 3134 NETDEV_CHANGEINFODATA, 3135 NETDEV_BONDING_INFO, 3136 NETDEV_PRECHANGEUPPER, 3137 NETDEV_CHANGELOWERSTATE, 3138 NETDEV_UDP_TUNNEL_PUSH_INFO, 3139 NETDEV_UDP_TUNNEL_DROP_INFO, 3140 NETDEV_CHANGE_TX_QUEUE_LEN, 3141 NETDEV_CVLAN_FILTER_PUSH_INFO, 3142 NETDEV_CVLAN_FILTER_DROP_INFO, 3143 NETDEV_SVLAN_FILTER_PUSH_INFO, 3144 NETDEV_SVLAN_FILTER_DROP_INFO, 3145 NETDEV_OFFLOAD_XSTATS_ENABLE, 3146 NETDEV_OFFLOAD_XSTATS_DISABLE, 3147 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 3148 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 3149 NETDEV_XDP_FEAT_CHANGE, 3150}; 3151const char *netdev_cmd_to_name(enum netdev_cmd cmd); 3152 3153int register_netdevice_notifier(struct notifier_block *nb); 3154int unregister_netdevice_notifier(struct notifier_block *nb); 3155int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 3156int unregister_netdevice_notifier_net(struct net *net, 3157 struct notifier_block *nb); 3158int register_netdevice_notifier_dev_net(struct net_device *dev, 3159 struct notifier_block *nb, 3160 struct netdev_net_notifier *nn); 3161int unregister_netdevice_notifier_dev_net(struct net_device *dev, 3162 struct notifier_block *nb, 3163 struct netdev_net_notifier *nn); 3164 3165struct netdev_notifier_info { 3166 struct net_device *dev; 3167 struct netlink_ext_ack *extack; 3168}; 3169 3170struct netdev_notifier_info_ext { 3171 struct netdev_notifier_info info; /* must be first */ 3172 union { 3173 u32 mtu; 3174 } ext; 3175}; 3176 3177struct netdev_notifier_change_info { 3178 struct netdev_notifier_info info; /* must be first */ 3179 unsigned int flags_changed; 3180}; 3181 3182struct netdev_notifier_changeupper_info { 3183 struct netdev_notifier_info info; /* must be first */ 3184 struct net_device *upper_dev; /* new upper dev */ 3185 bool master; /* is upper dev master */ 3186 bool linking; /* is the notification for link or unlink */ 3187 void *upper_info; /* upper dev info */ 3188}; 3189 3190struct netdev_notifier_changelowerstate_info { 3191 struct netdev_notifier_info info; /* must be first */ 3192 void *lower_state_info; /* is lower dev state */ 3193}; 3194 3195struct netdev_notifier_pre_changeaddr_info { 3196 struct netdev_notifier_info info; /* must be first */ 3197 const unsigned char *dev_addr; 3198}; 3199 3200enum netdev_offload_xstats_type { 3201 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 3202}; 3203 3204struct netdev_notifier_offload_xstats_info { 3205 struct netdev_notifier_info info; /* must be first */ 3206 enum netdev_offload_xstats_type type; 3207 3208 union { 3209 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 3210 struct netdev_notifier_offload_xstats_rd *report_delta; 3211 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 3212 struct netdev_notifier_offload_xstats_ru *report_used; 3213 }; 3214}; 3215 3216int netdev_offload_xstats_enable(struct net_device *dev, 3217 enum netdev_offload_xstats_type type, 3218 struct netlink_ext_ack *extack); 3219int netdev_offload_xstats_disable(struct net_device *dev, 3220 enum netdev_offload_xstats_type type); 3221bool netdev_offload_xstats_enabled(const struct net_device *dev, 3222 enum netdev_offload_xstats_type type); 3223int netdev_offload_xstats_get(struct net_device *dev, 3224 enum netdev_offload_xstats_type type, 3225 struct rtnl_hw_stats64 *stats, bool *used, 3226 struct netlink_ext_ack *extack); 3227void 3228netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 3229 const struct rtnl_hw_stats64 *stats); 3230void 3231netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 3232void netdev_offload_xstats_push_delta(struct net_device *dev, 3233 enum netdev_offload_xstats_type type, 3234 const struct rtnl_hw_stats64 *stats); 3235 3236static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 3237 struct net_device *dev) 3238{ 3239 info->dev = dev; 3240 info->extack = NULL; 3241} 3242 3243static inline struct net_device * 3244netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 3245{ 3246 return info->dev; 3247} 3248 3249static inline struct netlink_ext_ack * 3250netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 3251{ 3252 return info->extack; 3253} 3254 3255int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 3256int call_netdevice_notifiers_info(unsigned long val, 3257 struct netdev_notifier_info *info); 3258 3259#define for_each_netdev(net, d) \ 3260 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 3261#define for_each_netdev_reverse(net, d) \ 3262 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 3263#define for_each_netdev_rcu(net, d) \ 3264 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 3265#define for_each_netdev_safe(net, d, n) \ 3266 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 3267#define for_each_netdev_continue(net, d) \ 3268 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 3269#define for_each_netdev_continue_reverse(net, d) \ 3270 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 3271 dev_list) 3272#define for_each_netdev_continue_rcu(net, d) \ 3273 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 3274#define for_each_netdev_in_bond_rcu(bond, slave) \ 3275 for_each_netdev_rcu(dev_net_rcu(bond), slave) \ 3276 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 3277#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 3278 3279#define for_each_netdev_dump(net, d, ifindex) \ 3280 for (; (d = xa_find(&(net)->dev_by_index, &ifindex, \ 3281 ULONG_MAX, XA_PRESENT)); ifindex++) 3282 3283static inline struct net_device *next_net_device(struct net_device *dev) 3284{ 3285 struct list_head *lh; 3286 struct net *net; 3287 3288 net = dev_net(dev); 3289 lh = dev->dev_list.next; 3290 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3291} 3292 3293static inline struct net_device *next_net_device_rcu(struct net_device *dev) 3294{ 3295 struct list_head *lh; 3296 struct net *net; 3297 3298 net = dev_net(dev); 3299 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 3300 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3301} 3302 3303static inline struct net_device *first_net_device(struct net *net) 3304{ 3305 return list_empty(&net->dev_base_head) ? NULL : 3306 net_device_entry(net->dev_base_head.next); 3307} 3308 3309static inline struct net_device *first_net_device_rcu(struct net *net) 3310{ 3311 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 3312 3313 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3314} 3315 3316int netdev_boot_setup_check(struct net_device *dev); 3317struct net_device *dev_getbyhwaddr(struct net *net, unsigned short type, 3318 const char *hwaddr); 3319struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 3320 const char *hwaddr); 3321struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 3322void dev_add_pack(struct packet_type *pt); 3323void dev_remove_pack(struct packet_type *pt); 3324void __dev_remove_pack(struct packet_type *pt); 3325void dev_add_offload(struct packet_offload *po); 3326void dev_remove_offload(struct packet_offload *po); 3327 3328int dev_get_iflink(const struct net_device *dev); 3329int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 3330int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 3331 struct net_device_path_stack *stack); 3332struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 3333 unsigned short mask); 3334struct net_device *dev_get_by_name(struct net *net, const char *name); 3335struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 3336struct net_device *__dev_get_by_name(struct net *net, const char *name); 3337bool netdev_name_in_use(struct net *net, const char *name); 3338int dev_alloc_name(struct net_device *dev, const char *name); 3339int netif_open(struct net_device *dev, struct netlink_ext_ack *extack); 3340int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 3341void netif_close(struct net_device *dev); 3342void dev_close(struct net_device *dev); 3343void dev_close_many(struct list_head *head, bool unlink); 3344void netif_disable_lro(struct net_device *dev); 3345void dev_disable_lro(struct net_device *dev); 3346int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 3347u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 3348 struct net_device *sb_dev); 3349 3350int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3351int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3352 3353static inline int dev_queue_xmit(struct sk_buff *skb) 3354{ 3355 return __dev_queue_xmit(skb, NULL); 3356} 3357 3358static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3359 struct net_device *sb_dev) 3360{ 3361 return __dev_queue_xmit(skb, sb_dev); 3362} 3363 3364static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3365{ 3366 int ret; 3367 3368 ret = __dev_direct_xmit(skb, queue_id); 3369 if (!dev_xmit_complete(ret)) 3370 kfree_skb(skb); 3371 return ret; 3372} 3373 3374int register_netdevice(struct net_device *dev); 3375void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3376void unregister_netdevice_many(struct list_head *head); 3377static inline void unregister_netdevice(struct net_device *dev) 3378{ 3379 unregister_netdevice_queue(dev, NULL); 3380} 3381 3382int netdev_refcnt_read(const struct net_device *dev); 3383void free_netdev(struct net_device *dev); 3384 3385struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3386 struct sk_buff *skb, 3387 bool all_slaves); 3388struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3389 struct sock *sk); 3390struct net_device *dev_get_by_index(struct net *net, int ifindex); 3391struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3392struct net_device *netdev_get_by_index(struct net *net, int ifindex, 3393 netdevice_tracker *tracker, gfp_t gfp); 3394struct net_device *netdev_get_by_name(struct net *net, const char *name, 3395 netdevice_tracker *tracker, gfp_t gfp); 3396struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3397void netdev_copy_name(struct net_device *dev, char *name); 3398 3399static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3400 unsigned short type, 3401 const void *daddr, const void *saddr, 3402 unsigned int len) 3403{ 3404 if (!dev->header_ops || !dev->header_ops->create) 3405 return 0; 3406 3407 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3408} 3409 3410static inline int dev_parse_header(const struct sk_buff *skb, 3411 unsigned char *haddr) 3412{ 3413 const struct net_device *dev = skb->dev; 3414 3415 if (!dev->header_ops || !dev->header_ops->parse) 3416 return 0; 3417 return dev->header_ops->parse(skb, haddr); 3418} 3419 3420static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3421{ 3422 const struct net_device *dev = skb->dev; 3423 3424 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3425 return 0; 3426 return dev->header_ops->parse_protocol(skb); 3427} 3428 3429/* ll_header must have at least hard_header_len allocated */ 3430static inline bool dev_validate_header(const struct net_device *dev, 3431 char *ll_header, int len) 3432{ 3433 if (likely(len >= dev->hard_header_len)) 3434 return true; 3435 if (len < dev->min_header_len) 3436 return false; 3437 3438 if (capable(CAP_SYS_RAWIO)) { 3439 memset(ll_header + len, 0, dev->hard_header_len - len); 3440 return true; 3441 } 3442 3443 if (dev->header_ops && dev->header_ops->validate) 3444 return dev->header_ops->validate(ll_header, len); 3445 3446 return false; 3447} 3448 3449static inline bool dev_has_header(const struct net_device *dev) 3450{ 3451 return dev->header_ops && dev->header_ops->create; 3452} 3453 3454/* 3455 * Incoming packets are placed on per-CPU queues 3456 */ 3457struct softnet_data { 3458 struct list_head poll_list; 3459 struct sk_buff_head process_queue; 3460 local_lock_t process_queue_bh_lock; 3461 3462 /* stats */ 3463 unsigned int processed; 3464 unsigned int time_squeeze; 3465#ifdef CONFIG_RPS 3466 struct softnet_data *rps_ipi_list; 3467#endif 3468 3469 unsigned int received_rps; 3470 bool in_net_rx_action; 3471 bool in_napi_threaded_poll; 3472 3473#ifdef CONFIG_NET_FLOW_LIMIT 3474 struct sd_flow_limit __rcu *flow_limit; 3475#endif 3476 struct Qdisc *output_queue; 3477 struct Qdisc **output_queue_tailp; 3478 struct sk_buff *completion_queue; 3479#ifdef CONFIG_XFRM_OFFLOAD 3480 struct sk_buff_head xfrm_backlog; 3481#endif 3482 /* written and read only by owning cpu: */ 3483 struct netdev_xmit xmit; 3484#ifdef CONFIG_RPS 3485 /* input_queue_head should be written by cpu owning this struct, 3486 * and only read by other cpus. Worth using a cache line. 3487 */ 3488 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3489 3490 /* Elements below can be accessed between CPUs for RPS/RFS */ 3491 call_single_data_t csd ____cacheline_aligned_in_smp; 3492 struct softnet_data *rps_ipi_next; 3493 unsigned int cpu; 3494 unsigned int input_queue_tail; 3495#endif 3496 struct sk_buff_head input_pkt_queue; 3497 struct napi_struct backlog; 3498 3499 atomic_t dropped ____cacheline_aligned_in_smp; 3500 3501 /* Another possibly contended cache line */ 3502 spinlock_t defer_lock ____cacheline_aligned_in_smp; 3503 int defer_count; 3504 int defer_ipi_scheduled; 3505 struct sk_buff *defer_list; 3506 call_single_data_t defer_csd; 3507}; 3508 3509DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3510 3511struct page_pool_bh { 3512 struct page_pool *pool; 3513 local_lock_t bh_lock; 3514}; 3515DECLARE_PER_CPU(struct page_pool_bh, system_page_pool); 3516 3517#ifndef CONFIG_PREEMPT_RT 3518static inline int dev_recursion_level(void) 3519{ 3520 return this_cpu_read(softnet_data.xmit.recursion); 3521} 3522#else 3523static inline int dev_recursion_level(void) 3524{ 3525 return current->net_xmit.recursion; 3526} 3527 3528#endif 3529 3530void __netif_schedule(struct Qdisc *q); 3531void netif_schedule_queue(struct netdev_queue *txq); 3532 3533static inline void netif_tx_schedule_all(struct net_device *dev) 3534{ 3535 unsigned int i; 3536 3537 for (i = 0; i < dev->num_tx_queues; i++) 3538 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3539} 3540 3541static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3542{ 3543 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3544} 3545 3546/** 3547 * netif_start_queue - allow transmit 3548 * @dev: network device 3549 * 3550 * Allow upper layers to call the device hard_start_xmit routine. 3551 */ 3552static inline void netif_start_queue(struct net_device *dev) 3553{ 3554 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3555} 3556 3557static inline void netif_tx_start_all_queues(struct net_device *dev) 3558{ 3559 unsigned int i; 3560 3561 for (i = 0; i < dev->num_tx_queues; i++) { 3562 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3563 netif_tx_start_queue(txq); 3564 } 3565} 3566 3567void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3568 3569/** 3570 * netif_wake_queue - restart transmit 3571 * @dev: network device 3572 * 3573 * Allow upper layers to call the device hard_start_xmit routine. 3574 * Used for flow control when transmit resources are available. 3575 */ 3576static inline void netif_wake_queue(struct net_device *dev) 3577{ 3578 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3579} 3580 3581static inline void netif_tx_wake_all_queues(struct net_device *dev) 3582{ 3583 unsigned int i; 3584 3585 for (i = 0; i < dev->num_tx_queues; i++) { 3586 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3587 netif_tx_wake_queue(txq); 3588 } 3589} 3590 3591static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3592{ 3593 /* Paired with READ_ONCE() from dev_watchdog() */ 3594 WRITE_ONCE(dev_queue->trans_start, jiffies); 3595 3596 /* This barrier is paired with smp_mb() from dev_watchdog() */ 3597 smp_mb__before_atomic(); 3598 3599 /* Must be an atomic op see netif_txq_try_stop() */ 3600 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3601} 3602 3603/** 3604 * netif_stop_queue - stop transmitted packets 3605 * @dev: network device 3606 * 3607 * Stop upper layers calling the device hard_start_xmit routine. 3608 * Used for flow control when transmit resources are unavailable. 3609 */ 3610static inline void netif_stop_queue(struct net_device *dev) 3611{ 3612 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3613} 3614 3615void netif_tx_stop_all_queues(struct net_device *dev); 3616 3617static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3618{ 3619 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3620} 3621 3622/** 3623 * netif_queue_stopped - test if transmit queue is flowblocked 3624 * @dev: network device 3625 * 3626 * Test if transmit queue on device is currently unable to send. 3627 */ 3628static inline bool netif_queue_stopped(const struct net_device *dev) 3629{ 3630 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3631} 3632 3633static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3634{ 3635 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3636} 3637 3638static inline bool 3639netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3640{ 3641 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3642} 3643 3644static inline bool 3645netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3646{ 3647 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3648} 3649 3650/** 3651 * netdev_queue_set_dql_min_limit - set dql minimum limit 3652 * @dev_queue: pointer to transmit queue 3653 * @min_limit: dql minimum limit 3654 * 3655 * Forces xmit_more() to return true until the minimum threshold 3656 * defined by @min_limit is reached (or until the tx queue is 3657 * empty). Warning: to be use with care, misuse will impact the 3658 * latency. 3659 */ 3660static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3661 unsigned int min_limit) 3662{ 3663#ifdef CONFIG_BQL 3664 dev_queue->dql.min_limit = min_limit; 3665#endif 3666} 3667 3668static inline int netdev_queue_dql_avail(const struct netdev_queue *txq) 3669{ 3670#ifdef CONFIG_BQL 3671 /* Non-BQL migrated drivers will return 0, too. */ 3672 return dql_avail(&txq->dql); 3673#else 3674 return 0; 3675#endif 3676} 3677 3678/** 3679 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3680 * @dev_queue: pointer to transmit queue 3681 * 3682 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3683 * to give appropriate hint to the CPU. 3684 */ 3685static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3686{ 3687#ifdef CONFIG_BQL 3688 prefetchw(&dev_queue->dql.num_queued); 3689#endif 3690} 3691 3692/** 3693 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3694 * @dev_queue: pointer to transmit queue 3695 * 3696 * BQL enabled drivers might use this helper in their TX completion path, 3697 * to give appropriate hint to the CPU. 3698 */ 3699static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3700{ 3701#ifdef CONFIG_BQL 3702 prefetchw(&dev_queue->dql.limit); 3703#endif 3704} 3705 3706/** 3707 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3708 * @dev_queue: network device queue 3709 * @bytes: number of bytes queued to the device queue 3710 * 3711 * Report the number of bytes queued for sending/completion to the network 3712 * device hardware queue. @bytes should be a good approximation and should 3713 * exactly match netdev_completed_queue() @bytes. 3714 * This is typically called once per packet, from ndo_start_xmit(). 3715 */ 3716static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3717 unsigned int bytes) 3718{ 3719#ifdef CONFIG_BQL 3720 dql_queued(&dev_queue->dql, bytes); 3721 3722 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3723 return; 3724 3725 /* Paired with READ_ONCE() from dev_watchdog() */ 3726 WRITE_ONCE(dev_queue->trans_start, jiffies); 3727 3728 /* This barrier is paired with smp_mb() from dev_watchdog() */ 3729 smp_mb__before_atomic(); 3730 3731 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3732 3733 /* 3734 * The XOFF flag must be set before checking the dql_avail below, 3735 * because in netdev_tx_completed_queue we update the dql_completed 3736 * before checking the XOFF flag. 3737 */ 3738 smp_mb__after_atomic(); 3739 3740 /* check again in case another CPU has just made room avail */ 3741 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3742 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3743#endif 3744} 3745 3746/* Variant of netdev_tx_sent_queue() for drivers that are aware 3747 * that they should not test BQL status themselves. 3748 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3749 * skb of a batch. 3750 * Returns true if the doorbell must be used to kick the NIC. 3751 */ 3752static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3753 unsigned int bytes, 3754 bool xmit_more) 3755{ 3756 if (xmit_more) { 3757#ifdef CONFIG_BQL 3758 dql_queued(&dev_queue->dql, bytes); 3759#endif 3760 return netif_tx_queue_stopped(dev_queue); 3761 } 3762 netdev_tx_sent_queue(dev_queue, bytes); 3763 return true; 3764} 3765 3766/** 3767 * netdev_sent_queue - report the number of bytes queued to hardware 3768 * @dev: network device 3769 * @bytes: number of bytes queued to the hardware device queue 3770 * 3771 * Report the number of bytes queued for sending/completion to the network 3772 * device hardware queue#0. @bytes should be a good approximation and should 3773 * exactly match netdev_completed_queue() @bytes. 3774 * This is typically called once per packet, from ndo_start_xmit(). 3775 */ 3776static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3777{ 3778 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3779} 3780 3781static inline bool __netdev_sent_queue(struct net_device *dev, 3782 unsigned int bytes, 3783 bool xmit_more) 3784{ 3785 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3786 xmit_more); 3787} 3788 3789/** 3790 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3791 * @dev_queue: network device queue 3792 * @pkts: number of packets (currently ignored) 3793 * @bytes: number of bytes dequeued from the device queue 3794 * 3795 * Must be called at most once per TX completion round (and not per 3796 * individual packet), so that BQL can adjust its limits appropriately. 3797 */ 3798static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3799 unsigned int pkts, unsigned int bytes) 3800{ 3801#ifdef CONFIG_BQL 3802 if (unlikely(!bytes)) 3803 return; 3804 3805 dql_completed(&dev_queue->dql, bytes); 3806 3807 /* 3808 * Without the memory barrier there is a small possibility that 3809 * netdev_tx_sent_queue will miss the update and cause the queue to 3810 * be stopped forever 3811 */ 3812 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */ 3813 3814 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3815 return; 3816 3817 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3818 netif_schedule_queue(dev_queue); 3819#endif 3820} 3821 3822/** 3823 * netdev_completed_queue - report bytes and packets completed by device 3824 * @dev: network device 3825 * @pkts: actual number of packets sent over the medium 3826 * @bytes: actual number of bytes sent over the medium 3827 * 3828 * Report the number of bytes and packets transmitted by the network device 3829 * hardware queue over the physical medium, @bytes must exactly match the 3830 * @bytes amount passed to netdev_sent_queue() 3831 */ 3832static inline void netdev_completed_queue(struct net_device *dev, 3833 unsigned int pkts, unsigned int bytes) 3834{ 3835 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3836} 3837 3838static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3839{ 3840#ifdef CONFIG_BQL 3841 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3842 dql_reset(&q->dql); 3843#endif 3844} 3845 3846/** 3847 * netdev_tx_reset_subqueue - reset the BQL stats and state of a netdev queue 3848 * @dev: network device 3849 * @qid: stack index of the queue to reset 3850 */ 3851static inline void netdev_tx_reset_subqueue(const struct net_device *dev, 3852 u32 qid) 3853{ 3854 netdev_tx_reset_queue(netdev_get_tx_queue(dev, qid)); 3855} 3856 3857/** 3858 * netdev_reset_queue - reset the packets and bytes count of a network device 3859 * @dev_queue: network device 3860 * 3861 * Reset the bytes and packet count of a network device and clear the 3862 * software flow control OFF bit for this network device 3863 */ 3864static inline void netdev_reset_queue(struct net_device *dev_queue) 3865{ 3866 netdev_tx_reset_subqueue(dev_queue, 0); 3867} 3868 3869/** 3870 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3871 * @dev: network device 3872 * @queue_index: given tx queue index 3873 * 3874 * Returns 0 if given tx queue index >= number of device tx queues, 3875 * otherwise returns the originally passed tx queue index. 3876 */ 3877static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3878{ 3879 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3880 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3881 dev->name, queue_index, 3882 dev->real_num_tx_queues); 3883 return 0; 3884 } 3885 3886 return queue_index; 3887} 3888 3889/** 3890 * netif_running - test if up 3891 * @dev: network device 3892 * 3893 * Test if the device has been brought up. 3894 */ 3895static inline bool netif_running(const struct net_device *dev) 3896{ 3897 return test_bit(__LINK_STATE_START, &dev->state); 3898} 3899 3900/* 3901 * Routines to manage the subqueues on a device. We only need start, 3902 * stop, and a check if it's stopped. All other device management is 3903 * done at the overall netdevice level. 3904 * Also test the device if we're multiqueue. 3905 */ 3906 3907/** 3908 * netif_start_subqueue - allow sending packets on subqueue 3909 * @dev: network device 3910 * @queue_index: sub queue index 3911 * 3912 * Start individual transmit queue of a device with multiple transmit queues. 3913 */ 3914static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3915{ 3916 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3917 3918 netif_tx_start_queue(txq); 3919} 3920 3921/** 3922 * netif_stop_subqueue - stop sending packets on subqueue 3923 * @dev: network device 3924 * @queue_index: sub queue index 3925 * 3926 * Stop individual transmit queue of a device with multiple transmit queues. 3927 */ 3928static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3929{ 3930 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3931 netif_tx_stop_queue(txq); 3932} 3933 3934/** 3935 * __netif_subqueue_stopped - test status of subqueue 3936 * @dev: network device 3937 * @queue_index: sub queue index 3938 * 3939 * Check individual transmit queue of a device with multiple transmit queues. 3940 */ 3941static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3942 u16 queue_index) 3943{ 3944 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3945 3946 return netif_tx_queue_stopped(txq); 3947} 3948 3949/** 3950 * netif_subqueue_stopped - test status of subqueue 3951 * @dev: network device 3952 * @skb: sub queue buffer pointer 3953 * 3954 * Check individual transmit queue of a device with multiple transmit queues. 3955 */ 3956static inline bool netif_subqueue_stopped(const struct net_device *dev, 3957 struct sk_buff *skb) 3958{ 3959 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3960} 3961 3962/** 3963 * netif_wake_subqueue - allow sending packets on subqueue 3964 * @dev: network device 3965 * @queue_index: sub queue index 3966 * 3967 * Resume individual transmit queue of a device with multiple transmit queues. 3968 */ 3969static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3970{ 3971 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3972 3973 netif_tx_wake_queue(txq); 3974} 3975 3976#ifdef CONFIG_XPS 3977int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3978 u16 index); 3979int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3980 u16 index, enum xps_map_type type); 3981 3982/** 3983 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3984 * @j: CPU/Rx queue index 3985 * @mask: bitmask of all cpus/rx queues 3986 * @nr_bits: number of bits in the bitmask 3987 * 3988 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3989 */ 3990static inline bool netif_attr_test_mask(unsigned long j, 3991 const unsigned long *mask, 3992 unsigned int nr_bits) 3993{ 3994 cpu_max_bits_warn(j, nr_bits); 3995 return test_bit(j, mask); 3996} 3997 3998/** 3999 * netif_attr_test_online - Test for online CPU/Rx queue 4000 * @j: CPU/Rx queue index 4001 * @online_mask: bitmask for CPUs/Rx queues that are online 4002 * @nr_bits: number of bits in the bitmask 4003 * 4004 * Returns: true if a CPU/Rx queue is online. 4005 */ 4006static inline bool netif_attr_test_online(unsigned long j, 4007 const unsigned long *online_mask, 4008 unsigned int nr_bits) 4009{ 4010 cpu_max_bits_warn(j, nr_bits); 4011 4012 if (online_mask) 4013 return test_bit(j, online_mask); 4014 4015 return (j < nr_bits); 4016} 4017 4018/** 4019 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 4020 * @n: CPU/Rx queue index 4021 * @srcp: the cpumask/Rx queue mask pointer 4022 * @nr_bits: number of bits in the bitmask 4023 * 4024 * Returns: next (after n) CPU/Rx queue index in the mask; 4025 * >= nr_bits if no further CPUs/Rx queues set. 4026 */ 4027static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 4028 unsigned int nr_bits) 4029{ 4030 /* -1 is a legal arg here. */ 4031 if (n != -1) 4032 cpu_max_bits_warn(n, nr_bits); 4033 4034 if (srcp) 4035 return find_next_bit(srcp, nr_bits, n + 1); 4036 4037 return n + 1; 4038} 4039 4040/** 4041 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 4042 * @n: CPU/Rx queue index 4043 * @src1p: the first CPUs/Rx queues mask pointer 4044 * @src2p: the second CPUs/Rx queues mask pointer 4045 * @nr_bits: number of bits in the bitmask 4046 * 4047 * Returns: next (after n) CPU/Rx queue index set in both masks; 4048 * >= nr_bits if no further CPUs/Rx queues set in both. 4049 */ 4050static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 4051 const unsigned long *src2p, 4052 unsigned int nr_bits) 4053{ 4054 /* -1 is a legal arg here. */ 4055 if (n != -1) 4056 cpu_max_bits_warn(n, nr_bits); 4057 4058 if (src1p && src2p) 4059 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 4060 else if (src1p) 4061 return find_next_bit(src1p, nr_bits, n + 1); 4062 else if (src2p) 4063 return find_next_bit(src2p, nr_bits, n + 1); 4064 4065 return n + 1; 4066} 4067#else 4068static inline int netif_set_xps_queue(struct net_device *dev, 4069 const struct cpumask *mask, 4070 u16 index) 4071{ 4072 return 0; 4073} 4074 4075static inline int __netif_set_xps_queue(struct net_device *dev, 4076 const unsigned long *mask, 4077 u16 index, enum xps_map_type type) 4078{ 4079 return 0; 4080} 4081#endif 4082 4083/** 4084 * netif_is_multiqueue - test if device has multiple transmit queues 4085 * @dev: network device 4086 * 4087 * Check if device has multiple transmit queues 4088 */ 4089static inline bool netif_is_multiqueue(const struct net_device *dev) 4090{ 4091 return dev->num_tx_queues > 1; 4092} 4093 4094int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 4095int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 4096int netif_set_real_num_queues(struct net_device *dev, 4097 unsigned int txq, unsigned int rxq); 4098 4099int netif_get_num_default_rss_queues(void); 4100 4101void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason); 4102void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason); 4103 4104/* 4105 * It is not allowed to call kfree_skb() or consume_skb() from hardware 4106 * interrupt context or with hardware interrupts being disabled. 4107 * (in_hardirq() || irqs_disabled()) 4108 * 4109 * We provide four helpers that can be used in following contexts : 4110 * 4111 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 4112 * replacing kfree_skb(skb) 4113 * 4114 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 4115 * Typically used in place of consume_skb(skb) in TX completion path 4116 * 4117 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 4118 * replacing kfree_skb(skb) 4119 * 4120 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 4121 * and consumed a packet. Used in place of consume_skb(skb) 4122 */ 4123static inline void dev_kfree_skb_irq(struct sk_buff *skb) 4124{ 4125 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 4126} 4127 4128static inline void dev_consume_skb_irq(struct sk_buff *skb) 4129{ 4130 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED); 4131} 4132 4133static inline void dev_kfree_skb_any(struct sk_buff *skb) 4134{ 4135 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 4136} 4137 4138static inline void dev_consume_skb_any(struct sk_buff *skb) 4139{ 4140 dev_kfree_skb_any_reason(skb, SKB_CONSUMED); 4141} 4142 4143u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 4144 const struct bpf_prog *xdp_prog); 4145void generic_xdp_tx(struct sk_buff *skb, const struct bpf_prog *xdp_prog); 4146int do_xdp_generic(const struct bpf_prog *xdp_prog, struct sk_buff **pskb); 4147int netif_rx(struct sk_buff *skb); 4148int __netif_rx(struct sk_buff *skb); 4149 4150int netif_receive_skb(struct sk_buff *skb); 4151int netif_receive_skb_core(struct sk_buff *skb); 4152void netif_receive_skb_list_internal(struct list_head *head); 4153void netif_receive_skb_list(struct list_head *head); 4154gro_result_t gro_receive_skb(struct gro_node *gro, struct sk_buff *skb); 4155 4156static inline gro_result_t napi_gro_receive(struct napi_struct *napi, 4157 struct sk_buff *skb) 4158{ 4159 return gro_receive_skb(&napi->gro, skb); 4160} 4161 4162struct sk_buff *napi_get_frags(struct napi_struct *napi); 4163gro_result_t napi_gro_frags(struct napi_struct *napi); 4164 4165static inline void napi_free_frags(struct napi_struct *napi) 4166{ 4167 kfree_skb(napi->skb); 4168 napi->skb = NULL; 4169} 4170 4171bool netdev_is_rx_handler_busy(struct net_device *dev); 4172int netdev_rx_handler_register(struct net_device *dev, 4173 rx_handler_func_t *rx_handler, 4174 void *rx_handler_data); 4175void netdev_rx_handler_unregister(struct net_device *dev); 4176 4177bool dev_valid_name(const char *name); 4178static inline bool is_socket_ioctl_cmd(unsigned int cmd) 4179{ 4180 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 4181} 4182int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 4183int put_user_ifreq(struct ifreq *ifr, void __user *arg); 4184int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 4185 void __user *data, bool *need_copyout); 4186int dev_ifconf(struct net *net, struct ifconf __user *ifc); 4187int dev_eth_ioctl(struct net_device *dev, 4188 struct ifreq *ifr, unsigned int cmd); 4189int generic_hwtstamp_get_lower(struct net_device *dev, 4190 struct kernel_hwtstamp_config *kernel_cfg); 4191int generic_hwtstamp_set_lower(struct net_device *dev, 4192 struct kernel_hwtstamp_config *kernel_cfg, 4193 struct netlink_ext_ack *extack); 4194int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 4195unsigned int dev_get_flags(const struct net_device *); 4196int __dev_change_flags(struct net_device *dev, unsigned int flags, 4197 struct netlink_ext_ack *extack); 4198int netif_change_flags(struct net_device *dev, unsigned int flags, 4199 struct netlink_ext_ack *extack); 4200int dev_change_flags(struct net_device *dev, unsigned int flags, 4201 struct netlink_ext_ack *extack); 4202int netif_set_alias(struct net_device *dev, const char *alias, size_t len); 4203int dev_set_alias(struct net_device *, const char *, size_t); 4204int dev_get_alias(const struct net_device *, char *, size_t); 4205int __dev_change_net_namespace(struct net_device *dev, struct net *net, 4206 const char *pat, int new_ifindex, 4207 struct netlink_ext_ack *extack); 4208int dev_change_net_namespace(struct net_device *dev, struct net *net, 4209 const char *pat); 4210int __dev_set_mtu(struct net_device *, int); 4211int netif_set_mtu(struct net_device *dev, int new_mtu); 4212int dev_set_mtu(struct net_device *, int); 4213int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 4214 struct netlink_ext_ack *extack); 4215int netif_set_mac_address(struct net_device *dev, struct sockaddr_storage *ss, 4216 struct netlink_ext_ack *extack); 4217int dev_set_mac_address(struct net_device *dev, struct sockaddr_storage *ss, 4218 struct netlink_ext_ack *extack); 4219int dev_set_mac_address_user(struct net_device *dev, struct sockaddr_storage *ss, 4220 struct netlink_ext_ack *extack); 4221int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 4222int dev_get_port_parent_id(struct net_device *dev, 4223 struct netdev_phys_item_id *ppid, bool recurse); 4224bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 4225 4226struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 4227struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 4228 struct netdev_queue *txq, int *ret); 4229 4230int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 4231u8 dev_xdp_prog_count(struct net_device *dev); 4232int netif_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf); 4233int dev_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf); 4234u8 dev_xdp_sb_prog_count(struct net_device *dev); 4235u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 4236 4237u32 dev_get_min_mp_channel_count(const struct net_device *dev); 4238 4239int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4240int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4241int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 4242bool is_skb_forwardable(const struct net_device *dev, 4243 const struct sk_buff *skb); 4244 4245static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 4246 const struct sk_buff *skb, 4247 const bool check_mtu) 4248{ 4249 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 4250 unsigned int len; 4251 4252 if (!(dev->flags & IFF_UP)) 4253 return false; 4254 4255 if (!check_mtu) 4256 return true; 4257 4258 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 4259 if (skb->len <= len) 4260 return true; 4261 4262 /* if TSO is enabled, we don't care about the length as the packet 4263 * could be forwarded without being segmented before 4264 */ 4265 if (skb_is_gso(skb)) 4266 return true; 4267 4268 return false; 4269} 4270 4271void netdev_core_stats_inc(struct net_device *dev, u32 offset); 4272 4273#define DEV_CORE_STATS_INC(FIELD) \ 4274static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 4275{ \ 4276 netdev_core_stats_inc(dev, \ 4277 offsetof(struct net_device_core_stats, FIELD)); \ 4278} 4279DEV_CORE_STATS_INC(rx_dropped) 4280DEV_CORE_STATS_INC(tx_dropped) 4281DEV_CORE_STATS_INC(rx_nohandler) 4282DEV_CORE_STATS_INC(rx_otherhost_dropped) 4283#undef DEV_CORE_STATS_INC 4284 4285static __always_inline int ____dev_forward_skb(struct net_device *dev, 4286 struct sk_buff *skb, 4287 const bool check_mtu) 4288{ 4289 if (skb_orphan_frags(skb, GFP_ATOMIC) || 4290 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 4291 dev_core_stats_rx_dropped_inc(dev); 4292 kfree_skb(skb); 4293 return NET_RX_DROP; 4294 } 4295 4296 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 4297 skb->priority = 0; 4298 return 0; 4299} 4300 4301bool dev_nit_active_rcu(const struct net_device *dev); 4302static inline bool dev_nit_active(const struct net_device *dev) 4303{ 4304 bool ret; 4305 4306 rcu_read_lock(); 4307 ret = dev_nit_active_rcu(dev); 4308 rcu_read_unlock(); 4309 return ret; 4310} 4311 4312void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 4313 4314static inline void __dev_put(struct net_device *dev) 4315{ 4316 if (dev) { 4317#ifdef CONFIG_PCPU_DEV_REFCNT 4318 this_cpu_dec(*dev->pcpu_refcnt); 4319#else 4320 refcount_dec(&dev->dev_refcnt); 4321#endif 4322 } 4323} 4324 4325static inline void __dev_hold(struct net_device *dev) 4326{ 4327 if (dev) { 4328#ifdef CONFIG_PCPU_DEV_REFCNT 4329 this_cpu_inc(*dev->pcpu_refcnt); 4330#else 4331 refcount_inc(&dev->dev_refcnt); 4332#endif 4333 } 4334} 4335 4336static inline void __netdev_tracker_alloc(struct net_device *dev, 4337 netdevice_tracker *tracker, 4338 gfp_t gfp) 4339{ 4340#ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4341 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 4342#endif 4343} 4344 4345/* netdev_tracker_alloc() can upgrade a prior untracked reference 4346 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4347 */ 4348static inline void netdev_tracker_alloc(struct net_device *dev, 4349 netdevice_tracker *tracker, gfp_t gfp) 4350{ 4351#ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4352 refcount_dec(&dev->refcnt_tracker.no_tracker); 4353 __netdev_tracker_alloc(dev, tracker, gfp); 4354#endif 4355} 4356 4357static inline void netdev_tracker_free(struct net_device *dev, 4358 netdevice_tracker *tracker) 4359{ 4360#ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4361 ref_tracker_free(&dev->refcnt_tracker, tracker); 4362#endif 4363} 4364 4365static inline void netdev_hold(struct net_device *dev, 4366 netdevice_tracker *tracker, gfp_t gfp) 4367{ 4368 if (dev) { 4369 __dev_hold(dev); 4370 __netdev_tracker_alloc(dev, tracker, gfp); 4371 } 4372} 4373 4374static inline void netdev_put(struct net_device *dev, 4375 netdevice_tracker *tracker) 4376{ 4377 if (dev) { 4378 netdev_tracker_free(dev, tracker); 4379 __dev_put(dev); 4380 } 4381} 4382 4383/** 4384 * dev_hold - get reference to device 4385 * @dev: network device 4386 * 4387 * Hold reference to device to keep it from being freed. 4388 * Try using netdev_hold() instead. 4389 */ 4390static inline void dev_hold(struct net_device *dev) 4391{ 4392 netdev_hold(dev, NULL, GFP_ATOMIC); 4393} 4394 4395/** 4396 * dev_put - release reference to device 4397 * @dev: network device 4398 * 4399 * Release reference to device to allow it to be freed. 4400 * Try using netdev_put() instead. 4401 */ 4402static inline void dev_put(struct net_device *dev) 4403{ 4404 netdev_put(dev, NULL); 4405} 4406 4407DEFINE_FREE(dev_put, struct net_device *, if (_T) dev_put(_T)) 4408 4409static inline void netdev_ref_replace(struct net_device *odev, 4410 struct net_device *ndev, 4411 netdevice_tracker *tracker, 4412 gfp_t gfp) 4413{ 4414 if (odev) 4415 netdev_tracker_free(odev, tracker); 4416 4417 __dev_hold(ndev); 4418 __dev_put(odev); 4419 4420 if (ndev) 4421 __netdev_tracker_alloc(ndev, tracker, gfp); 4422} 4423 4424/* Carrier loss detection, dial on demand. The functions netif_carrier_on 4425 * and _off may be called from IRQ context, but it is caller 4426 * who is responsible for serialization of these calls. 4427 * 4428 * The name carrier is inappropriate, these functions should really be 4429 * called netif_lowerlayer_*() because they represent the state of any 4430 * kind of lower layer not just hardware media. 4431 */ 4432void linkwatch_fire_event(struct net_device *dev); 4433 4434/** 4435 * linkwatch_sync_dev - sync linkwatch for the given device 4436 * @dev: network device to sync linkwatch for 4437 * 4438 * Sync linkwatch for the given device, removing it from the 4439 * pending work list (if queued). 4440 */ 4441void linkwatch_sync_dev(struct net_device *dev); 4442void __linkwatch_sync_dev(struct net_device *dev); 4443 4444/** 4445 * netif_carrier_ok - test if carrier present 4446 * @dev: network device 4447 * 4448 * Check if carrier is present on device 4449 */ 4450static inline bool netif_carrier_ok(const struct net_device *dev) 4451{ 4452 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4453} 4454 4455unsigned long dev_trans_start(struct net_device *dev); 4456 4457void netdev_watchdog_up(struct net_device *dev); 4458 4459void netif_carrier_on(struct net_device *dev); 4460void netif_carrier_off(struct net_device *dev); 4461void netif_carrier_event(struct net_device *dev); 4462 4463/** 4464 * netif_dormant_on - mark device as dormant. 4465 * @dev: network device 4466 * 4467 * Mark device as dormant (as per RFC2863). 4468 * 4469 * The dormant state indicates that the relevant interface is not 4470 * actually in a condition to pass packets (i.e., it is not 'up') but is 4471 * in a "pending" state, waiting for some external event. For "on- 4472 * demand" interfaces, this new state identifies the situation where the 4473 * interface is waiting for events to place it in the up state. 4474 */ 4475static inline void netif_dormant_on(struct net_device *dev) 4476{ 4477 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4478 linkwatch_fire_event(dev); 4479} 4480 4481/** 4482 * netif_dormant_off - set device as not dormant. 4483 * @dev: network device 4484 * 4485 * Device is not in dormant state. 4486 */ 4487static inline void netif_dormant_off(struct net_device *dev) 4488{ 4489 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4490 linkwatch_fire_event(dev); 4491} 4492 4493/** 4494 * netif_dormant - test if device is dormant 4495 * @dev: network device 4496 * 4497 * Check if device is dormant. 4498 */ 4499static inline bool netif_dormant(const struct net_device *dev) 4500{ 4501 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4502} 4503 4504 4505/** 4506 * netif_testing_on - mark device as under test. 4507 * @dev: network device 4508 * 4509 * Mark device as under test (as per RFC2863). 4510 * 4511 * The testing state indicates that some test(s) must be performed on 4512 * the interface. After completion, of the test, the interface state 4513 * will change to up, dormant, or down, as appropriate. 4514 */ 4515static inline void netif_testing_on(struct net_device *dev) 4516{ 4517 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4518 linkwatch_fire_event(dev); 4519} 4520 4521/** 4522 * netif_testing_off - set device as not under test. 4523 * @dev: network device 4524 * 4525 * Device is not in testing state. 4526 */ 4527static inline void netif_testing_off(struct net_device *dev) 4528{ 4529 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4530 linkwatch_fire_event(dev); 4531} 4532 4533/** 4534 * netif_testing - test if device is under test 4535 * @dev: network device 4536 * 4537 * Check if device is under test 4538 */ 4539static inline bool netif_testing(const struct net_device *dev) 4540{ 4541 return test_bit(__LINK_STATE_TESTING, &dev->state); 4542} 4543 4544 4545/** 4546 * netif_oper_up - test if device is operational 4547 * @dev: network device 4548 * 4549 * Check if carrier is operational 4550 */ 4551static inline bool netif_oper_up(const struct net_device *dev) 4552{ 4553 unsigned int operstate = READ_ONCE(dev->operstate); 4554 4555 return operstate == IF_OPER_UP || 4556 operstate == IF_OPER_UNKNOWN /* backward compat */; 4557} 4558 4559/** 4560 * netif_device_present - is device available or removed 4561 * @dev: network device 4562 * 4563 * Check if device has not been removed from system. 4564 */ 4565static inline bool netif_device_present(const struct net_device *dev) 4566{ 4567 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4568} 4569 4570void netif_device_detach(struct net_device *dev); 4571 4572void netif_device_attach(struct net_device *dev); 4573 4574/* 4575 * Network interface message level settings 4576 */ 4577 4578enum { 4579 NETIF_MSG_DRV_BIT, 4580 NETIF_MSG_PROBE_BIT, 4581 NETIF_MSG_LINK_BIT, 4582 NETIF_MSG_TIMER_BIT, 4583 NETIF_MSG_IFDOWN_BIT, 4584 NETIF_MSG_IFUP_BIT, 4585 NETIF_MSG_RX_ERR_BIT, 4586 NETIF_MSG_TX_ERR_BIT, 4587 NETIF_MSG_TX_QUEUED_BIT, 4588 NETIF_MSG_INTR_BIT, 4589 NETIF_MSG_TX_DONE_BIT, 4590 NETIF_MSG_RX_STATUS_BIT, 4591 NETIF_MSG_PKTDATA_BIT, 4592 NETIF_MSG_HW_BIT, 4593 NETIF_MSG_WOL_BIT, 4594 4595 /* When you add a new bit above, update netif_msg_class_names array 4596 * in net/ethtool/common.c 4597 */ 4598 NETIF_MSG_CLASS_COUNT, 4599}; 4600/* Both ethtool_ops interface and internal driver implementation use u32 */ 4601static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4602 4603#define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4604#define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4605 4606#define NETIF_MSG_DRV __NETIF_MSG(DRV) 4607#define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4608#define NETIF_MSG_LINK __NETIF_MSG(LINK) 4609#define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4610#define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4611#define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4612#define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4613#define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4614#define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4615#define NETIF_MSG_INTR __NETIF_MSG(INTR) 4616#define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4617#define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4618#define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4619#define NETIF_MSG_HW __NETIF_MSG(HW) 4620#define NETIF_MSG_WOL __NETIF_MSG(WOL) 4621 4622#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4623#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4624#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4625#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4626#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4627#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4628#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4629#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4630#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4631#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4632#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4633#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4634#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4635#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4636#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4637 4638static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4639{ 4640 /* use default */ 4641 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4642 return default_msg_enable_bits; 4643 if (debug_value == 0) /* no output */ 4644 return 0; 4645 /* set low N bits */ 4646 return (1U << debug_value) - 1; 4647} 4648 4649static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4650{ 4651 spin_lock(&txq->_xmit_lock); 4652 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4653 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4654} 4655 4656static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4657{ 4658 __acquire(&txq->_xmit_lock); 4659 return true; 4660} 4661 4662static inline void __netif_tx_release(struct netdev_queue *txq) 4663{ 4664 __release(&txq->_xmit_lock); 4665} 4666 4667static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4668{ 4669 spin_lock_bh(&txq->_xmit_lock); 4670 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4671 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4672} 4673 4674static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4675{ 4676 bool ok = spin_trylock(&txq->_xmit_lock); 4677 4678 if (likely(ok)) { 4679 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4680 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4681 } 4682 return ok; 4683} 4684 4685static inline void __netif_tx_unlock(struct netdev_queue *txq) 4686{ 4687 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4688 WRITE_ONCE(txq->xmit_lock_owner, -1); 4689 spin_unlock(&txq->_xmit_lock); 4690} 4691 4692static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4693{ 4694 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4695 WRITE_ONCE(txq->xmit_lock_owner, -1); 4696 spin_unlock_bh(&txq->_xmit_lock); 4697} 4698 4699/* 4700 * txq->trans_start can be read locklessly from dev_watchdog() 4701 */ 4702static inline void txq_trans_update(const struct net_device *dev, 4703 struct netdev_queue *txq) 4704{ 4705 if (!dev->lltx) 4706 WRITE_ONCE(txq->trans_start, jiffies); 4707} 4708 4709static inline void txq_trans_cond_update(struct netdev_queue *txq) 4710{ 4711 unsigned long now = jiffies; 4712 4713 if (READ_ONCE(txq->trans_start) != now) 4714 WRITE_ONCE(txq->trans_start, now); 4715} 4716 4717/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4718static inline void netif_trans_update(struct net_device *dev) 4719{ 4720 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4721 4722 txq_trans_cond_update(txq); 4723} 4724 4725/** 4726 * netif_tx_lock - grab network device transmit lock 4727 * @dev: network device 4728 * 4729 * Get network device transmit lock 4730 */ 4731void netif_tx_lock(struct net_device *dev); 4732 4733static inline void netif_tx_lock_bh(struct net_device *dev) 4734{ 4735 local_bh_disable(); 4736 netif_tx_lock(dev); 4737} 4738 4739void netif_tx_unlock(struct net_device *dev); 4740 4741static inline void netif_tx_unlock_bh(struct net_device *dev) 4742{ 4743 netif_tx_unlock(dev); 4744 local_bh_enable(); 4745} 4746 4747#define HARD_TX_LOCK(dev, txq, cpu) { \ 4748 if (!(dev)->lltx) { \ 4749 __netif_tx_lock(txq, cpu); \ 4750 } else { \ 4751 __netif_tx_acquire(txq); \ 4752 } \ 4753} 4754 4755#define HARD_TX_TRYLOCK(dev, txq) \ 4756 (!(dev)->lltx ? \ 4757 __netif_tx_trylock(txq) : \ 4758 __netif_tx_acquire(txq)) 4759 4760#define HARD_TX_UNLOCK(dev, txq) { \ 4761 if (!(dev)->lltx) { \ 4762 __netif_tx_unlock(txq); \ 4763 } else { \ 4764 __netif_tx_release(txq); \ 4765 } \ 4766} 4767 4768static inline void netif_tx_disable(struct net_device *dev) 4769{ 4770 unsigned int i; 4771 int cpu; 4772 4773 local_bh_disable(); 4774 cpu = smp_processor_id(); 4775 spin_lock(&dev->tx_global_lock); 4776 for (i = 0; i < dev->num_tx_queues; i++) { 4777 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4778 4779 __netif_tx_lock(txq, cpu); 4780 netif_tx_stop_queue(txq); 4781 __netif_tx_unlock(txq); 4782 } 4783 spin_unlock(&dev->tx_global_lock); 4784 local_bh_enable(); 4785} 4786 4787static inline void netif_addr_lock(struct net_device *dev) 4788{ 4789 unsigned char nest_level = 0; 4790 4791#ifdef CONFIG_LOCKDEP 4792 nest_level = dev->nested_level; 4793#endif 4794 spin_lock_nested(&dev->addr_list_lock, nest_level); 4795} 4796 4797static inline void netif_addr_lock_bh(struct net_device *dev) 4798{ 4799 unsigned char nest_level = 0; 4800 4801#ifdef CONFIG_LOCKDEP 4802 nest_level = dev->nested_level; 4803#endif 4804 local_bh_disable(); 4805 spin_lock_nested(&dev->addr_list_lock, nest_level); 4806} 4807 4808static inline void netif_addr_unlock(struct net_device *dev) 4809{ 4810 spin_unlock(&dev->addr_list_lock); 4811} 4812 4813static inline void netif_addr_unlock_bh(struct net_device *dev) 4814{ 4815 spin_unlock_bh(&dev->addr_list_lock); 4816} 4817 4818/* 4819 * dev_addrs walker. Should be used only for read access. Call with 4820 * rcu_read_lock held. 4821 */ 4822#define for_each_dev_addr(dev, ha) \ 4823 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4824 4825/* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4826 4827void ether_setup(struct net_device *dev); 4828 4829/* Allocate dummy net_device */ 4830struct net_device *alloc_netdev_dummy(int sizeof_priv); 4831 4832/* Support for loadable net-drivers */ 4833struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4834 unsigned char name_assign_type, 4835 void (*setup)(struct net_device *), 4836 unsigned int txqs, unsigned int rxqs); 4837#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4838 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4839 4840#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4841 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4842 count) 4843 4844int register_netdev(struct net_device *dev); 4845void unregister_netdev(struct net_device *dev); 4846 4847int devm_register_netdev(struct device *dev, struct net_device *ndev); 4848 4849/* General hardware address lists handling functions */ 4850int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4851 struct netdev_hw_addr_list *from_list, int addr_len); 4852int __hw_addr_sync_multiple(struct netdev_hw_addr_list *to_list, 4853 struct netdev_hw_addr_list *from_list, 4854 int addr_len); 4855void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4856 struct netdev_hw_addr_list *from_list, int addr_len); 4857int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4858 struct net_device *dev, 4859 int (*sync)(struct net_device *, const unsigned char *), 4860 int (*unsync)(struct net_device *, 4861 const unsigned char *)); 4862int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4863 struct net_device *dev, 4864 int (*sync)(struct net_device *, 4865 const unsigned char *, int), 4866 int (*unsync)(struct net_device *, 4867 const unsigned char *, int)); 4868void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4869 struct net_device *dev, 4870 int (*unsync)(struct net_device *, 4871 const unsigned char *, int)); 4872void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4873 struct net_device *dev, 4874 int (*unsync)(struct net_device *, 4875 const unsigned char *)); 4876void __hw_addr_init(struct netdev_hw_addr_list *list); 4877 4878/* Functions used for device addresses handling */ 4879void dev_addr_mod(struct net_device *dev, unsigned int offset, 4880 const void *addr, size_t len); 4881 4882static inline void 4883__dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4884{ 4885 dev_addr_mod(dev, 0, addr, len); 4886} 4887 4888static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4889{ 4890 __dev_addr_set(dev, addr, dev->addr_len); 4891} 4892 4893int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4894 unsigned char addr_type); 4895int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4896 unsigned char addr_type); 4897 4898/* Functions used for unicast addresses handling */ 4899int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4900int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4901int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4902int dev_uc_sync(struct net_device *to, struct net_device *from); 4903int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4904void dev_uc_unsync(struct net_device *to, struct net_device *from); 4905void dev_uc_flush(struct net_device *dev); 4906void dev_uc_init(struct net_device *dev); 4907 4908/** 4909 * __dev_uc_sync - Synchronize device's unicast list 4910 * @dev: device to sync 4911 * @sync: function to call if address should be added 4912 * @unsync: function to call if address should be removed 4913 * 4914 * Add newly added addresses to the interface, and release 4915 * addresses that have been deleted. 4916 */ 4917static inline int __dev_uc_sync(struct net_device *dev, 4918 int (*sync)(struct net_device *, 4919 const unsigned char *), 4920 int (*unsync)(struct net_device *, 4921 const unsigned char *)) 4922{ 4923 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4924} 4925 4926/** 4927 * __dev_uc_unsync - Remove synchronized addresses from device 4928 * @dev: device to sync 4929 * @unsync: function to call if address should be removed 4930 * 4931 * Remove all addresses that were added to the device by dev_uc_sync(). 4932 */ 4933static inline void __dev_uc_unsync(struct net_device *dev, 4934 int (*unsync)(struct net_device *, 4935 const unsigned char *)) 4936{ 4937 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4938} 4939 4940/* Functions used for multicast addresses handling */ 4941int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4942int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4943int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4944int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4945int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4946int dev_mc_sync(struct net_device *to, struct net_device *from); 4947int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4948void dev_mc_unsync(struct net_device *to, struct net_device *from); 4949void dev_mc_flush(struct net_device *dev); 4950void dev_mc_init(struct net_device *dev); 4951 4952/** 4953 * __dev_mc_sync - Synchronize device's multicast list 4954 * @dev: device to sync 4955 * @sync: function to call if address should be added 4956 * @unsync: function to call if address should be removed 4957 * 4958 * Add newly added addresses to the interface, and release 4959 * addresses that have been deleted. 4960 */ 4961static inline int __dev_mc_sync(struct net_device *dev, 4962 int (*sync)(struct net_device *, 4963 const unsigned char *), 4964 int (*unsync)(struct net_device *, 4965 const unsigned char *)) 4966{ 4967 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4968} 4969 4970/** 4971 * __dev_mc_unsync - Remove synchronized addresses from device 4972 * @dev: device to sync 4973 * @unsync: function to call if address should be removed 4974 * 4975 * Remove all addresses that were added to the device by dev_mc_sync(). 4976 */ 4977static inline void __dev_mc_unsync(struct net_device *dev, 4978 int (*unsync)(struct net_device *, 4979 const unsigned char *)) 4980{ 4981 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4982} 4983 4984/* Functions used for secondary unicast and multicast support */ 4985void dev_set_rx_mode(struct net_device *dev); 4986int netif_set_promiscuity(struct net_device *dev, int inc); 4987int dev_set_promiscuity(struct net_device *dev, int inc); 4988int netif_set_allmulti(struct net_device *dev, int inc, bool notify); 4989int dev_set_allmulti(struct net_device *dev, int inc); 4990void netif_state_change(struct net_device *dev); 4991void netdev_state_change(struct net_device *dev); 4992void __netdev_notify_peers(struct net_device *dev); 4993void netdev_notify_peers(struct net_device *dev); 4994void netdev_features_change(struct net_device *dev); 4995/* Load a device via the kmod */ 4996void dev_load(struct net *net, const char *name); 4997struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4998 struct rtnl_link_stats64 *storage); 4999void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 5000 const struct net_device_stats *netdev_stats); 5001void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 5002 const struct pcpu_sw_netstats __percpu *netstats); 5003void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 5004 5005enum { 5006 NESTED_SYNC_IMM_BIT, 5007 NESTED_SYNC_TODO_BIT, 5008}; 5009 5010#define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 5011#define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 5012 5013#define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 5014#define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 5015 5016struct netdev_nested_priv { 5017 unsigned char flags; 5018 void *data; 5019}; 5020 5021bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 5022struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 5023 struct list_head **iter); 5024 5025/* iterate through upper list, must be called under RCU read lock */ 5026#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 5027 for (iter = &(dev)->adj_list.upper, \ 5028 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 5029 updev; \ 5030 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 5031 5032int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 5033 int (*fn)(struct net_device *upper_dev, 5034 struct netdev_nested_priv *priv), 5035 struct netdev_nested_priv *priv); 5036 5037bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 5038 struct net_device *upper_dev); 5039 5040bool netdev_has_any_upper_dev(struct net_device *dev); 5041 5042void *netdev_lower_get_next_private(struct net_device *dev, 5043 struct list_head **iter); 5044void *netdev_lower_get_next_private_rcu(struct net_device *dev, 5045 struct list_head **iter); 5046 5047#define netdev_for_each_lower_private(dev, priv, iter) \ 5048 for (iter = (dev)->adj_list.lower.next, \ 5049 priv = netdev_lower_get_next_private(dev, &(iter)); \ 5050 priv; \ 5051 priv = netdev_lower_get_next_private(dev, &(iter))) 5052 5053#define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 5054 for (iter = &(dev)->adj_list.lower, \ 5055 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 5056 priv; \ 5057 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 5058 5059void *netdev_lower_get_next(struct net_device *dev, 5060 struct list_head **iter); 5061 5062#define netdev_for_each_lower_dev(dev, ldev, iter) \ 5063 for (iter = (dev)->adj_list.lower.next, \ 5064 ldev = netdev_lower_get_next(dev, &(iter)); \ 5065 ldev; \ 5066 ldev = netdev_lower_get_next(dev, &(iter))) 5067 5068struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 5069 struct list_head **iter); 5070int netdev_walk_all_lower_dev(struct net_device *dev, 5071 int (*fn)(struct net_device *lower_dev, 5072 struct netdev_nested_priv *priv), 5073 struct netdev_nested_priv *priv); 5074int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 5075 int (*fn)(struct net_device *lower_dev, 5076 struct netdev_nested_priv *priv), 5077 struct netdev_nested_priv *priv); 5078 5079void *netdev_adjacent_get_private(struct list_head *adj_list); 5080void *netdev_lower_get_first_private_rcu(struct net_device *dev); 5081struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 5082struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 5083int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 5084 struct netlink_ext_ack *extack); 5085int netdev_master_upper_dev_link(struct net_device *dev, 5086 struct net_device *upper_dev, 5087 void *upper_priv, void *upper_info, 5088 struct netlink_ext_ack *extack); 5089void netdev_upper_dev_unlink(struct net_device *dev, 5090 struct net_device *upper_dev); 5091int netdev_adjacent_change_prepare(struct net_device *old_dev, 5092 struct net_device *new_dev, 5093 struct net_device *dev, 5094 struct netlink_ext_ack *extack); 5095void netdev_adjacent_change_commit(struct net_device *old_dev, 5096 struct net_device *new_dev, 5097 struct net_device *dev); 5098void netdev_adjacent_change_abort(struct net_device *old_dev, 5099 struct net_device *new_dev, 5100 struct net_device *dev); 5101void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 5102void *netdev_lower_dev_get_private(struct net_device *dev, 5103 struct net_device *lower_dev); 5104void netdev_lower_state_changed(struct net_device *lower_dev, 5105 void *lower_state_info); 5106 5107/* RSS keys are 40 or 52 bytes long */ 5108#define NETDEV_RSS_KEY_LEN 52 5109extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 5110void netdev_rss_key_fill(void *buffer, size_t len); 5111 5112int skb_checksum_help(struct sk_buff *skb); 5113int skb_crc32c_csum_help(struct sk_buff *skb); 5114int skb_csum_hwoffload_help(struct sk_buff *skb, 5115 const netdev_features_t features); 5116 5117struct netdev_bonding_info { 5118 ifslave slave; 5119 ifbond master; 5120}; 5121 5122struct netdev_notifier_bonding_info { 5123 struct netdev_notifier_info info; /* must be first */ 5124 struct netdev_bonding_info bonding_info; 5125}; 5126 5127void netdev_bonding_info_change(struct net_device *dev, 5128 struct netdev_bonding_info *bonding_info); 5129 5130#if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 5131void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 5132#else 5133static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 5134 const void *data) 5135{ 5136} 5137#endif 5138 5139__be16 skb_network_protocol(struct sk_buff *skb, int *depth); 5140 5141static inline bool can_checksum_protocol(netdev_features_t features, 5142 __be16 protocol) 5143{ 5144 if (protocol == htons(ETH_P_FCOE)) 5145 return !!(features & NETIF_F_FCOE_CRC); 5146 5147 /* Assume this is an IP checksum (not SCTP CRC) */ 5148 5149 if (features & NETIF_F_HW_CSUM) { 5150 /* Can checksum everything */ 5151 return true; 5152 } 5153 5154 switch (protocol) { 5155 case htons(ETH_P_IP): 5156 return !!(features & NETIF_F_IP_CSUM); 5157 case htons(ETH_P_IPV6): 5158 return !!(features & NETIF_F_IPV6_CSUM); 5159 default: 5160 return false; 5161 } 5162} 5163 5164#ifdef CONFIG_BUG 5165void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 5166#else 5167static inline void netdev_rx_csum_fault(struct net_device *dev, 5168 struct sk_buff *skb) 5169{ 5170} 5171#endif 5172/* rx skb timestamps */ 5173void net_enable_timestamp(void); 5174void net_disable_timestamp(void); 5175 5176static inline ktime_t netdev_get_tstamp(struct net_device *dev, 5177 const struct skb_shared_hwtstamps *hwtstamps, 5178 bool cycles) 5179{ 5180 const struct net_device_ops *ops = dev->netdev_ops; 5181 5182 if (ops->ndo_get_tstamp) 5183 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 5184 5185 return hwtstamps->hwtstamp; 5186} 5187 5188#ifndef CONFIG_PREEMPT_RT 5189static inline void netdev_xmit_set_more(bool more) 5190{ 5191 __this_cpu_write(softnet_data.xmit.more, more); 5192} 5193 5194static inline bool netdev_xmit_more(void) 5195{ 5196 return __this_cpu_read(softnet_data.xmit.more); 5197} 5198#else 5199static inline void netdev_xmit_set_more(bool more) 5200{ 5201 current->net_xmit.more = more; 5202} 5203 5204static inline bool netdev_xmit_more(void) 5205{ 5206 return current->net_xmit.more; 5207} 5208#endif 5209 5210static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 5211 struct sk_buff *skb, struct net_device *dev, 5212 bool more) 5213{ 5214 netdev_xmit_set_more(more); 5215 return ops->ndo_start_xmit(skb, dev); 5216} 5217 5218static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 5219 struct netdev_queue *txq, bool more) 5220{ 5221 const struct net_device_ops *ops = dev->netdev_ops; 5222 netdev_tx_t rc; 5223 5224 rc = __netdev_start_xmit(ops, skb, dev, more); 5225 if (rc == NETDEV_TX_OK) 5226 txq_trans_update(dev, txq); 5227 5228 return rc; 5229} 5230 5231int netdev_class_create_file_ns(const struct class_attribute *class_attr, 5232 const void *ns); 5233void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 5234 const void *ns); 5235 5236extern const struct kobj_ns_type_operations net_ns_type_operations; 5237 5238const char *netdev_drivername(const struct net_device *dev); 5239 5240static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 5241 netdev_features_t f2) 5242{ 5243 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 5244 if (f1 & NETIF_F_HW_CSUM) 5245 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5246 else 5247 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5248 } 5249 5250 return f1 & f2; 5251} 5252 5253static inline netdev_features_t netdev_get_wanted_features( 5254 struct net_device *dev) 5255{ 5256 return (dev->features & ~dev->hw_features) | dev->wanted_features; 5257} 5258netdev_features_t netdev_increment_features(netdev_features_t all, 5259 netdev_features_t one, netdev_features_t mask); 5260 5261/* Allow TSO being used on stacked device : 5262 * Performing the GSO segmentation before last device 5263 * is a performance improvement. 5264 */ 5265static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 5266 netdev_features_t mask) 5267{ 5268 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 5269} 5270 5271int __netdev_update_features(struct net_device *dev); 5272void netdev_update_features(struct net_device *dev); 5273void netdev_change_features(struct net_device *dev); 5274 5275void netif_stacked_transfer_operstate(const struct net_device *rootdev, 5276 struct net_device *dev); 5277 5278netdev_features_t passthru_features_check(struct sk_buff *skb, 5279 struct net_device *dev, 5280 netdev_features_t features); 5281netdev_features_t netif_skb_features(struct sk_buff *skb); 5282void skb_warn_bad_offload(const struct sk_buff *skb); 5283 5284static inline bool net_gso_ok(netdev_features_t features, int gso_type) 5285{ 5286 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 5287 5288 /* check flags correspondence */ 5289 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 5290 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 5291 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 5292 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 5293 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 5294 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 5295 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 5296 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 5297 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 5298 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 5299 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 5300 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 5301 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 5302 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 5303 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 5304 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 5305 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 5306 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 5307 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 5308 BUILD_BUG_ON(SKB_GSO_TCP_ACCECN != 5309 (NETIF_F_GSO_ACCECN >> NETIF_F_GSO_SHIFT)); 5310 5311 return (features & feature) == feature; 5312} 5313 5314static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 5315{ 5316 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 5317 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 5318} 5319 5320static inline bool netif_needs_gso(struct sk_buff *skb, 5321 netdev_features_t features) 5322{ 5323 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 5324 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 5325 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 5326} 5327 5328void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 5329void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 5330void netif_inherit_tso_max(struct net_device *to, 5331 const struct net_device *from); 5332 5333static inline unsigned int 5334netif_get_gro_max_size(const struct net_device *dev, const struct sk_buff *skb) 5335{ 5336 /* pairs with WRITE_ONCE() in netif_set_gro(_ipv4)_max_size() */ 5337 return skb->protocol == htons(ETH_P_IPV6) ? 5338 READ_ONCE(dev->gro_max_size) : 5339 READ_ONCE(dev->gro_ipv4_max_size); 5340} 5341 5342static inline unsigned int 5343netif_get_gso_max_size(const struct net_device *dev, const struct sk_buff *skb) 5344{ 5345 /* pairs with WRITE_ONCE() in netif_set_gso(_ipv4)_max_size() */ 5346 return skb->protocol == htons(ETH_P_IPV6) ? 5347 READ_ONCE(dev->gso_max_size) : 5348 READ_ONCE(dev->gso_ipv4_max_size); 5349} 5350 5351static inline bool netif_is_macsec(const struct net_device *dev) 5352{ 5353 return dev->priv_flags & IFF_MACSEC; 5354} 5355 5356static inline bool netif_is_macvlan(const struct net_device *dev) 5357{ 5358 return dev->priv_flags & IFF_MACVLAN; 5359} 5360 5361static inline bool netif_is_macvlan_port(const struct net_device *dev) 5362{ 5363 return dev->priv_flags & IFF_MACVLAN_PORT; 5364} 5365 5366static inline bool netif_is_bond_master(const struct net_device *dev) 5367{ 5368 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5369} 5370 5371static inline bool netif_is_bond_slave(const struct net_device *dev) 5372{ 5373 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5374} 5375 5376static inline bool netif_supports_nofcs(struct net_device *dev) 5377{ 5378 return dev->priv_flags & IFF_SUPP_NOFCS; 5379} 5380 5381static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5382{ 5383 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5384} 5385 5386static inline bool netif_is_l3_master(const struct net_device *dev) 5387{ 5388 return dev->priv_flags & IFF_L3MDEV_MASTER; 5389} 5390 5391static inline bool netif_is_l3_slave(const struct net_device *dev) 5392{ 5393 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5394} 5395 5396static inline int dev_sdif(const struct net_device *dev) 5397{ 5398#ifdef CONFIG_NET_L3_MASTER_DEV 5399 if (netif_is_l3_slave(dev)) 5400 return dev->ifindex; 5401#endif 5402 return 0; 5403} 5404 5405static inline bool netif_is_bridge_master(const struct net_device *dev) 5406{ 5407 return dev->priv_flags & IFF_EBRIDGE; 5408} 5409 5410static inline bool netif_is_bridge_port(const struct net_device *dev) 5411{ 5412 return dev->priv_flags & IFF_BRIDGE_PORT; 5413} 5414 5415static inline bool netif_is_ovs_master(const struct net_device *dev) 5416{ 5417 return dev->priv_flags & IFF_OPENVSWITCH; 5418} 5419 5420static inline bool netif_is_ovs_port(const struct net_device *dev) 5421{ 5422 return dev->priv_flags & IFF_OVS_DATAPATH; 5423} 5424 5425static inline bool netif_is_any_bridge_master(const struct net_device *dev) 5426{ 5427 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev); 5428} 5429 5430static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5431{ 5432 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5433} 5434 5435static inline bool netif_is_team_master(const struct net_device *dev) 5436{ 5437 return dev->priv_flags & IFF_TEAM; 5438} 5439 5440static inline bool netif_is_team_port(const struct net_device *dev) 5441{ 5442 return dev->priv_flags & IFF_TEAM_PORT; 5443} 5444 5445static inline bool netif_is_lag_master(const struct net_device *dev) 5446{ 5447 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5448} 5449 5450static inline bool netif_is_lag_port(const struct net_device *dev) 5451{ 5452 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5453} 5454 5455static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5456{ 5457 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5458} 5459 5460static inline bool netif_is_failover(const struct net_device *dev) 5461{ 5462 return dev->priv_flags & IFF_FAILOVER; 5463} 5464 5465static inline bool netif_is_failover_slave(const struct net_device *dev) 5466{ 5467 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5468} 5469 5470/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5471static inline void netif_keep_dst(struct net_device *dev) 5472{ 5473 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5474} 5475 5476/* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5477static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5478{ 5479 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5480 return netif_is_macsec(dev); 5481} 5482 5483extern struct pernet_operations __net_initdata loopback_net_ops; 5484 5485/* Logging, debugging and troubleshooting/diagnostic helpers. */ 5486 5487/* netdev_printk helpers, similar to dev_printk */ 5488 5489static inline const char *netdev_name(const struct net_device *dev) 5490{ 5491 if (!dev->name[0] || strchr(dev->name, '%')) 5492 return "(unnamed net_device)"; 5493 return dev->name; 5494} 5495 5496static inline const char *netdev_reg_state(const struct net_device *dev) 5497{ 5498 u8 reg_state = READ_ONCE(dev->reg_state); 5499 5500 switch (reg_state) { 5501 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5502 case NETREG_REGISTERED: return ""; 5503 case NETREG_UNREGISTERING: return " (unregistering)"; 5504 case NETREG_UNREGISTERED: return " (unregistered)"; 5505 case NETREG_RELEASED: return " (released)"; 5506 case NETREG_DUMMY: return " (dummy)"; 5507 } 5508 5509 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state); 5510 return " (unknown)"; 5511} 5512 5513#define MODULE_ALIAS_NETDEV(device) \ 5514 MODULE_ALIAS("netdev-" device) 5515 5516/* 5517 * netdev_WARN() acts like dev_printk(), but with the key difference 5518 * of using a WARN/WARN_ON to get the message out, including the 5519 * file/line information and a backtrace. 5520 */ 5521#define netdev_WARN(dev, format, args...) \ 5522 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5523 netdev_reg_state(dev), ##args) 5524 5525#define netdev_WARN_ONCE(dev, format, args...) \ 5526 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5527 netdev_reg_state(dev), ##args) 5528 5529/* 5530 * The list of packet types we will receive (as opposed to discard) 5531 * and the routines to invoke. 5532 * 5533 * Why 16. Because with 16 the only overlap we get on a hash of the 5534 * low nibble of the protocol value is RARP/SNAP/X.25. 5535 * 5536 * 0800 IP 5537 * 0001 802.3 5538 * 0002 AX.25 5539 * 0004 802.2 5540 * 8035 RARP 5541 * 0005 SNAP 5542 * 0805 X.25 5543 * 0806 ARP 5544 * 8137 IPX 5545 * 0009 Localtalk 5546 * 86DD IPv6 5547 */ 5548#define PTYPE_HASH_SIZE (16) 5549#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5550 5551extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5552 5553extern struct net_device *blackhole_netdev; 5554 5555/* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */ 5556#define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD) 5557#define DEV_STATS_ADD(DEV, FIELD, VAL) \ 5558 atomic_long_add((VAL), &(DEV)->stats.__##FIELD) 5559#define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD) 5560 5561#endif /* _LINUX_NETDEVICE_H */