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