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 #include <net/xdp.h> 44 45 #include <linux/netdev_features.h> 46 #include <linux/neighbour.h> 47 #include <uapi/linux/netdevice.h> 48 #include <uapi/linux/if_bonding.h> 49 #include <uapi/linux/pkt_cls.h> 50 #include <linux/hashtable.h> 51 #include <linux/rbtree.h> 52 #include <net/net_trackers.h> 53 #include <net/net_debug.h> 54 55 struct netpoll_info; 56 struct device; 57 struct ethtool_ops; 58 struct phy_device; 59 struct dsa_port; 60 struct ip_tunnel_parm; 61 struct macsec_context; 62 struct macsec_ops; 63 struct netdev_name_node; 64 struct sd_flow_limit; 65 struct sfp_bus; 66 /* 802.11 specific */ 67 struct wireless_dev; 68 /* 802.15.4 specific */ 69 struct wpan_dev; 70 struct mpls_dev; 71 /* UDP Tunnel offloads */ 72 struct udp_tunnel_info; 73 struct udp_tunnel_nic_info; 74 struct udp_tunnel_nic; 75 struct bpf_prog; 76 struct xdp_buff; 77 struct xdp_md; 78 79 void synchronize_net(void); 80 void netdev_set_default_ethtool_ops(struct net_device *dev, 81 const struct ethtool_ops *ops); 82 void netdev_sw_irq_coalesce_default_on(struct net_device *dev); 83 84 /* Backlog congestion levels */ 85 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 86 #define NET_RX_DROP 1 /* packet dropped */ 87 88 #define MAX_NEST_DEV 8 89 90 /* 91 * Transmit return codes: transmit return codes originate from three different 92 * namespaces: 93 * 94 * - qdisc return codes 95 * - driver transmit return codes 96 * - errno values 97 * 98 * Drivers are allowed to return any one of those in their hard_start_xmit() 99 * function. Real network devices commonly used with qdiscs should only return 100 * the driver transmit return codes though - when qdiscs are used, the actual 101 * transmission happens asynchronously, so the value is not propagated to 102 * higher layers. Virtual network devices transmit synchronously; in this case 103 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 104 * others are propagated to higher layers. 105 */ 106 107 /* qdisc ->enqueue() return codes. */ 108 #define NET_XMIT_SUCCESS 0x00 109 #define NET_XMIT_DROP 0x01 /* skb dropped */ 110 #define NET_XMIT_CN 0x02 /* congestion notification */ 111 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 112 113 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 114 * indicates that the device will soon be dropping packets, or already drops 115 * some packets of the same priority; prompting us to send less aggressively. */ 116 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 117 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 118 119 /* Driver transmit return codes */ 120 #define NETDEV_TX_MASK 0xf0 121 122 enum netdev_tx { 123 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 124 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 125 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 126 }; 127 typedef enum netdev_tx netdev_tx_t; 128 129 /* 130 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 131 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 132 */ 133 static inline bool dev_xmit_complete(int rc) 134 { 135 /* 136 * Positive cases with an skb consumed by a driver: 137 * - successful transmission (rc == NETDEV_TX_OK) 138 * - error while transmitting (rc < 0) 139 * - error while queueing to a different device (rc & NET_XMIT_MASK) 140 */ 141 if (likely(rc < NET_XMIT_MASK)) 142 return true; 143 144 return false; 145 } 146 147 /* 148 * Compute the worst-case header length according to the protocols 149 * used. 150 */ 151 152 #if defined(CONFIG_HYPERV_NET) 153 # define LL_MAX_HEADER 128 154 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 155 # if defined(CONFIG_MAC80211_MESH) 156 # define LL_MAX_HEADER 128 157 # else 158 # define LL_MAX_HEADER 96 159 # endif 160 #else 161 # define LL_MAX_HEADER 32 162 #endif 163 164 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 165 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 166 #define MAX_HEADER LL_MAX_HEADER 167 #else 168 #define MAX_HEADER (LL_MAX_HEADER + 48) 169 #endif 170 171 /* 172 * Old network device statistics. Fields are native words 173 * (unsigned long) so they can be read and written atomically. 174 */ 175 176 #define NET_DEV_STAT(FIELD) \ 177 union { \ 178 unsigned long FIELD; \ 179 atomic_long_t __##FIELD; \ 180 } 181 182 struct net_device_stats { 183 NET_DEV_STAT(rx_packets); 184 NET_DEV_STAT(tx_packets); 185 NET_DEV_STAT(rx_bytes); 186 NET_DEV_STAT(tx_bytes); 187 NET_DEV_STAT(rx_errors); 188 NET_DEV_STAT(tx_errors); 189 NET_DEV_STAT(rx_dropped); 190 NET_DEV_STAT(tx_dropped); 191 NET_DEV_STAT(multicast); 192 NET_DEV_STAT(collisions); 193 NET_DEV_STAT(rx_length_errors); 194 NET_DEV_STAT(rx_over_errors); 195 NET_DEV_STAT(rx_crc_errors); 196 NET_DEV_STAT(rx_frame_errors); 197 NET_DEV_STAT(rx_fifo_errors); 198 NET_DEV_STAT(rx_missed_errors); 199 NET_DEV_STAT(tx_aborted_errors); 200 NET_DEV_STAT(tx_carrier_errors); 201 NET_DEV_STAT(tx_fifo_errors); 202 NET_DEV_STAT(tx_heartbeat_errors); 203 NET_DEV_STAT(tx_window_errors); 204 NET_DEV_STAT(rx_compressed); 205 NET_DEV_STAT(tx_compressed); 206 }; 207 #undef NET_DEV_STAT 208 209 /* per-cpu stats, allocated on demand. 210 * Try to fit them in a single cache line, for dev_get_stats() sake. 211 */ 212 struct net_device_core_stats { 213 unsigned long rx_dropped; 214 unsigned long tx_dropped; 215 unsigned long rx_nohandler; 216 unsigned long rx_otherhost_dropped; 217 } __aligned(4 * sizeof(unsigned long)); 218 219 #include <linux/cache.h> 220 #include <linux/skbuff.h> 221 222 #ifdef CONFIG_RPS 223 #include <linux/static_key.h> 224 extern struct static_key_false rps_needed; 225 extern struct static_key_false rfs_needed; 226 #endif 227 228 struct neighbour; 229 struct neigh_parms; 230 struct sk_buff; 231 232 struct netdev_hw_addr { 233 struct list_head list; 234 struct rb_node node; 235 unsigned char addr[MAX_ADDR_LEN]; 236 unsigned char type; 237 #define NETDEV_HW_ADDR_T_LAN 1 238 #define NETDEV_HW_ADDR_T_SAN 2 239 #define NETDEV_HW_ADDR_T_UNICAST 3 240 #define NETDEV_HW_ADDR_T_MULTICAST 4 241 bool global_use; 242 int sync_cnt; 243 int refcount; 244 int synced; 245 struct rcu_head rcu_head; 246 }; 247 248 struct netdev_hw_addr_list { 249 struct list_head list; 250 int count; 251 252 /* Auxiliary tree for faster lookup on addition and deletion */ 253 struct rb_root tree; 254 }; 255 256 #define netdev_hw_addr_list_count(l) ((l)->count) 257 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 258 #define netdev_hw_addr_list_for_each(ha, l) \ 259 list_for_each_entry(ha, &(l)->list, list) 260 261 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 262 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 263 #define netdev_for_each_uc_addr(ha, dev) \ 264 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 265 #define netdev_for_each_synced_uc_addr(_ha, _dev) \ 266 netdev_for_each_uc_addr((_ha), (_dev)) \ 267 if ((_ha)->sync_cnt) 268 269 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 270 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 271 #define netdev_for_each_mc_addr(ha, dev) \ 272 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 273 #define netdev_for_each_synced_mc_addr(_ha, _dev) \ 274 netdev_for_each_mc_addr((_ha), (_dev)) \ 275 if ((_ha)->sync_cnt) 276 277 struct hh_cache { 278 unsigned int hh_len; 279 seqlock_t hh_lock; 280 281 /* cached hardware header; allow for machine alignment needs. */ 282 #define HH_DATA_MOD 16 283 #define HH_DATA_OFF(__len) \ 284 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 285 #define HH_DATA_ALIGN(__len) \ 286 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 287 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 288 }; 289 290 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 291 * Alternative is: 292 * dev->hard_header_len ? (dev->hard_header_len + 293 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 294 * 295 * We could use other alignment values, but we must maintain the 296 * relationship HH alignment <= LL alignment. 297 */ 298 #define LL_RESERVED_SPACE(dev) \ 299 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 300 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 301 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 302 303 struct header_ops { 304 int (*create) (struct sk_buff *skb, struct net_device *dev, 305 unsigned short type, const void *daddr, 306 const void *saddr, unsigned int len); 307 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 308 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 309 void (*cache_update)(struct hh_cache *hh, 310 const struct net_device *dev, 311 const unsigned char *haddr); 312 bool (*validate)(const char *ll_header, unsigned int len); 313 __be16 (*parse_protocol)(const struct sk_buff *skb); 314 }; 315 316 /* These flag bits are private to the generic network queueing 317 * layer; they may not be explicitly referenced by any other 318 * code. 319 */ 320 321 enum netdev_state_t { 322 __LINK_STATE_START, 323 __LINK_STATE_PRESENT, 324 __LINK_STATE_NOCARRIER, 325 __LINK_STATE_LINKWATCH_PENDING, 326 __LINK_STATE_DORMANT, 327 __LINK_STATE_TESTING, 328 }; 329 330 struct gro_list { 331 struct list_head list; 332 int count; 333 }; 334 335 /* 336 * size of gro hash buckets, must less than bit number of 337 * napi_struct::gro_bitmask 338 */ 339 #define GRO_HASH_BUCKETS 8 340 341 /* 342 * Structure for NAPI scheduling similar to tasklet but with weighting 343 */ 344 struct napi_struct { 345 /* The poll_list must only be managed by the entity which 346 * changes the state of the NAPI_STATE_SCHED bit. This means 347 * whoever atomically sets that bit can add this napi_struct 348 * to the per-CPU poll_list, and whoever clears that bit 349 * can remove from the list right before clearing the bit. 350 */ 351 struct list_head poll_list; 352 353 unsigned long state; 354 int weight; 355 int defer_hard_irqs_count; 356 unsigned long gro_bitmask; 357 int (*poll)(struct napi_struct *, int); 358 #ifdef CONFIG_NETPOLL 359 int poll_owner; 360 #endif 361 struct net_device *dev; 362 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 363 struct sk_buff *skb; 364 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 365 int rx_count; /* length of rx_list */ 366 struct hrtimer timer; 367 struct list_head dev_list; 368 struct hlist_node napi_hash_node; 369 unsigned int napi_id; 370 struct task_struct *thread; 371 }; 372 373 enum { 374 NAPI_STATE_SCHED, /* Poll is scheduled */ 375 NAPI_STATE_MISSED, /* reschedule a napi */ 376 NAPI_STATE_DISABLE, /* Disable pending */ 377 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 378 NAPI_STATE_LISTED, /* NAPI added to system lists */ 379 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 380 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 381 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 382 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 383 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 384 }; 385 386 enum { 387 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 388 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 389 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 390 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 391 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 392 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 393 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 394 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 395 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 396 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 397 }; 398 399 enum gro_result { 400 GRO_MERGED, 401 GRO_MERGED_FREE, 402 GRO_HELD, 403 GRO_NORMAL, 404 GRO_CONSUMED, 405 }; 406 typedef enum gro_result gro_result_t; 407 408 /* 409 * enum rx_handler_result - Possible return values for rx_handlers. 410 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 411 * further. 412 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 413 * case skb->dev was changed by rx_handler. 414 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 415 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 416 * 417 * rx_handlers are functions called from inside __netif_receive_skb(), to do 418 * special processing of the skb, prior to delivery to protocol handlers. 419 * 420 * Currently, a net_device can only have a single rx_handler registered. Trying 421 * to register a second rx_handler will return -EBUSY. 422 * 423 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 424 * To unregister a rx_handler on a net_device, use 425 * netdev_rx_handler_unregister(). 426 * 427 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 428 * do with the skb. 429 * 430 * If the rx_handler consumed the skb in some way, it should return 431 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 432 * the skb to be delivered in some other way. 433 * 434 * If the rx_handler changed skb->dev, to divert the skb to another 435 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 436 * new device will be called if it exists. 437 * 438 * If the rx_handler decides the skb should be ignored, it should return 439 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 440 * are registered on exact device (ptype->dev == skb->dev). 441 * 442 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 443 * delivered, it should return RX_HANDLER_PASS. 444 * 445 * A device without a registered rx_handler will behave as if rx_handler 446 * returned RX_HANDLER_PASS. 447 */ 448 449 enum rx_handler_result { 450 RX_HANDLER_CONSUMED, 451 RX_HANDLER_ANOTHER, 452 RX_HANDLER_EXACT, 453 RX_HANDLER_PASS, 454 }; 455 typedef enum rx_handler_result rx_handler_result_t; 456 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 457 458 void __napi_schedule(struct napi_struct *n); 459 void __napi_schedule_irqoff(struct napi_struct *n); 460 461 static inline bool napi_disable_pending(struct napi_struct *n) 462 { 463 return test_bit(NAPI_STATE_DISABLE, &n->state); 464 } 465 466 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 467 { 468 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 469 } 470 471 bool napi_schedule_prep(struct napi_struct *n); 472 473 /** 474 * napi_schedule - schedule NAPI poll 475 * @n: NAPI context 476 * 477 * Schedule NAPI poll routine to be called if it is not already 478 * running. 479 */ 480 static inline void napi_schedule(struct napi_struct *n) 481 { 482 if (napi_schedule_prep(n)) 483 __napi_schedule(n); 484 } 485 486 /** 487 * napi_schedule_irqoff - schedule NAPI poll 488 * @n: NAPI context 489 * 490 * Variant of napi_schedule(), assuming hard irqs are masked. 491 */ 492 static inline void napi_schedule_irqoff(struct napi_struct *n) 493 { 494 if (napi_schedule_prep(n)) 495 __napi_schedule_irqoff(n); 496 } 497 498 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 499 static inline bool napi_reschedule(struct napi_struct *napi) 500 { 501 if (napi_schedule_prep(napi)) { 502 __napi_schedule(napi); 503 return true; 504 } 505 return false; 506 } 507 508 bool napi_complete_done(struct napi_struct *n, int work_done); 509 /** 510 * napi_complete - NAPI processing complete 511 * @n: NAPI context 512 * 513 * Mark NAPI processing as complete. 514 * Consider using napi_complete_done() instead. 515 * Return false if device should avoid rearming interrupts. 516 */ 517 static inline bool napi_complete(struct napi_struct *n) 518 { 519 return napi_complete_done(n, 0); 520 } 521 522 int dev_set_threaded(struct net_device *dev, bool threaded); 523 524 /** 525 * napi_disable - prevent NAPI from scheduling 526 * @n: NAPI context 527 * 528 * Stop NAPI from being scheduled on this context. 529 * Waits till any outstanding processing completes. 530 */ 531 void napi_disable(struct napi_struct *n); 532 533 void napi_enable(struct napi_struct *n); 534 535 /** 536 * napi_synchronize - wait until NAPI is not running 537 * @n: NAPI context 538 * 539 * Wait until NAPI is done being scheduled on this context. 540 * Waits till any outstanding processing completes but 541 * does not disable future activations. 542 */ 543 static inline void napi_synchronize(const struct napi_struct *n) 544 { 545 if (IS_ENABLED(CONFIG_SMP)) 546 while (test_bit(NAPI_STATE_SCHED, &n->state)) 547 msleep(1); 548 else 549 barrier(); 550 } 551 552 /** 553 * napi_if_scheduled_mark_missed - if napi is running, set the 554 * NAPIF_STATE_MISSED 555 * @n: NAPI context 556 * 557 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 558 * NAPI is scheduled. 559 **/ 560 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 561 { 562 unsigned long val, new; 563 564 val = READ_ONCE(n->state); 565 do { 566 if (val & NAPIF_STATE_DISABLE) 567 return true; 568 569 if (!(val & NAPIF_STATE_SCHED)) 570 return false; 571 572 new = val | NAPIF_STATE_MISSED; 573 } while (!try_cmpxchg(&n->state, &val, new)); 574 575 return true; 576 } 577 578 enum netdev_queue_state_t { 579 __QUEUE_STATE_DRV_XOFF, 580 __QUEUE_STATE_STACK_XOFF, 581 __QUEUE_STATE_FROZEN, 582 }; 583 584 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 585 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 586 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 587 588 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 589 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 590 QUEUE_STATE_FROZEN) 591 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 592 QUEUE_STATE_FROZEN) 593 594 /* 595 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 596 * netif_tx_* functions below are used to manipulate this flag. The 597 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 598 * queue independently. The netif_xmit_*stopped functions below are called 599 * to check if the queue has been stopped by the driver or stack (either 600 * of the XOFF bits are set in the state). Drivers should not need to call 601 * netif_xmit*stopped functions, they should only be using netif_tx_*. 602 */ 603 604 struct netdev_queue { 605 /* 606 * read-mostly part 607 */ 608 struct net_device *dev; 609 netdevice_tracker dev_tracker; 610 611 struct Qdisc __rcu *qdisc; 612 struct Qdisc *qdisc_sleeping; 613 #ifdef CONFIG_SYSFS 614 struct kobject kobj; 615 #endif 616 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 617 int numa_node; 618 #endif 619 unsigned long tx_maxrate; 620 /* 621 * Number of TX timeouts for this queue 622 * (/sys/class/net/DEV/Q/trans_timeout) 623 */ 624 atomic_long_t trans_timeout; 625 626 /* Subordinate device that the queue has been assigned to */ 627 struct net_device *sb_dev; 628 #ifdef CONFIG_XDP_SOCKETS 629 struct xsk_buff_pool *pool; 630 #endif 631 /* 632 * write-mostly part 633 */ 634 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 635 int xmit_lock_owner; 636 /* 637 * Time (in jiffies) of last Tx 638 */ 639 unsigned long trans_start; 640 641 unsigned long state; 642 643 #ifdef CONFIG_BQL 644 struct dql dql; 645 #endif 646 } ____cacheline_aligned_in_smp; 647 648 extern int sysctl_fb_tunnels_only_for_init_net; 649 extern int sysctl_devconf_inherit_init_net; 650 651 /* 652 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 653 * == 1 : For initns only 654 * == 2 : For none. 655 */ 656 static inline bool net_has_fallback_tunnels(const struct net *net) 657 { 658 #if IS_ENABLED(CONFIG_SYSCTL) 659 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net); 660 661 return !fb_tunnels_only_for_init_net || 662 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1); 663 #else 664 return true; 665 #endif 666 } 667 668 static inline int net_inherit_devconf(void) 669 { 670 #if IS_ENABLED(CONFIG_SYSCTL) 671 return READ_ONCE(sysctl_devconf_inherit_init_net); 672 #else 673 return 0; 674 #endif 675 } 676 677 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 678 { 679 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 680 return q->numa_node; 681 #else 682 return NUMA_NO_NODE; 683 #endif 684 } 685 686 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 687 { 688 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 689 q->numa_node = node; 690 #endif 691 } 692 693 #ifdef CONFIG_RPS 694 /* 695 * This structure holds an RPS map which can be of variable length. The 696 * map is an array of CPUs. 697 */ 698 struct rps_map { 699 unsigned int len; 700 struct rcu_head rcu; 701 u16 cpus[]; 702 }; 703 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 704 705 /* 706 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 707 * tail pointer for that CPU's input queue at the time of last enqueue, and 708 * a hardware filter index. 709 */ 710 struct rps_dev_flow { 711 u16 cpu; 712 u16 filter; 713 unsigned int last_qtail; 714 }; 715 #define RPS_NO_FILTER 0xffff 716 717 /* 718 * The rps_dev_flow_table structure contains a table of flow mappings. 719 */ 720 struct rps_dev_flow_table { 721 unsigned int mask; 722 struct rcu_head rcu; 723 struct rps_dev_flow flows[]; 724 }; 725 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 726 ((_num) * sizeof(struct rps_dev_flow))) 727 728 /* 729 * The rps_sock_flow_table contains mappings of flows to the last CPU 730 * on which they were processed by the application (set in recvmsg). 731 * Each entry is a 32bit value. Upper part is the high-order bits 732 * of flow hash, lower part is CPU number. 733 * rps_cpu_mask is used to partition the space, depending on number of 734 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 735 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 736 * meaning we use 32-6=26 bits for the hash. 737 */ 738 struct rps_sock_flow_table { 739 u32 mask; 740 741 u32 ents[] ____cacheline_aligned_in_smp; 742 }; 743 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 744 745 #define RPS_NO_CPU 0xffff 746 747 extern u32 rps_cpu_mask; 748 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 749 750 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 751 u32 hash) 752 { 753 if (table && hash) { 754 unsigned int index = hash & table->mask; 755 u32 val = hash & ~rps_cpu_mask; 756 757 /* We only give a hint, preemption can change CPU under us */ 758 val |= raw_smp_processor_id(); 759 760 if (table->ents[index] != val) 761 table->ents[index] = val; 762 } 763 } 764 765 #ifdef CONFIG_RFS_ACCEL 766 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 767 u16 filter_id); 768 #endif 769 #endif /* CONFIG_RPS */ 770 771 /* This structure contains an instance of an RX queue. */ 772 struct netdev_rx_queue { 773 struct xdp_rxq_info xdp_rxq; 774 #ifdef CONFIG_RPS 775 struct rps_map __rcu *rps_map; 776 struct rps_dev_flow_table __rcu *rps_flow_table; 777 #endif 778 struct kobject kobj; 779 struct net_device *dev; 780 netdevice_tracker dev_tracker; 781 782 #ifdef CONFIG_XDP_SOCKETS 783 struct xsk_buff_pool *pool; 784 #endif 785 } ____cacheline_aligned_in_smp; 786 787 /* 788 * RX queue sysfs structures and functions. 789 */ 790 struct rx_queue_attribute { 791 struct attribute attr; 792 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 793 ssize_t (*store)(struct netdev_rx_queue *queue, 794 const char *buf, size_t len); 795 }; 796 797 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 798 enum xps_map_type { 799 XPS_CPUS = 0, 800 XPS_RXQS, 801 XPS_MAPS_MAX, 802 }; 803 804 #ifdef CONFIG_XPS 805 /* 806 * This structure holds an XPS map which can be of variable length. The 807 * map is an array of queues. 808 */ 809 struct xps_map { 810 unsigned int len; 811 unsigned int alloc_len; 812 struct rcu_head rcu; 813 u16 queues[]; 814 }; 815 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 816 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 817 - sizeof(struct xps_map)) / sizeof(u16)) 818 819 /* 820 * This structure holds all XPS maps for device. Maps are indexed by CPU. 821 * 822 * We keep track of the number of cpus/rxqs used when the struct is allocated, 823 * in nr_ids. This will help not accessing out-of-bound memory. 824 * 825 * We keep track of the number of traffic classes used when the struct is 826 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 827 * not crossing its upper bound, as the original dev->num_tc can be updated in 828 * the meantime. 829 */ 830 struct xps_dev_maps { 831 struct rcu_head rcu; 832 unsigned int nr_ids; 833 s16 num_tc; 834 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 835 }; 836 837 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 838 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 839 840 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 841 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 842 843 #endif /* CONFIG_XPS */ 844 845 #define TC_MAX_QUEUE 16 846 #define TC_BITMASK 15 847 /* HW offloaded queuing disciplines txq count and offset maps */ 848 struct netdev_tc_txq { 849 u16 count; 850 u16 offset; 851 }; 852 853 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 854 /* 855 * This structure is to hold information about the device 856 * configured to run FCoE protocol stack. 857 */ 858 struct netdev_fcoe_hbainfo { 859 char manufacturer[64]; 860 char serial_number[64]; 861 char hardware_version[64]; 862 char driver_version[64]; 863 char optionrom_version[64]; 864 char firmware_version[64]; 865 char model[256]; 866 char model_description[256]; 867 }; 868 #endif 869 870 #define MAX_PHYS_ITEM_ID_LEN 32 871 872 /* This structure holds a unique identifier to identify some 873 * physical item (port for example) used by a netdevice. 874 */ 875 struct netdev_phys_item_id { 876 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 877 unsigned char id_len; 878 }; 879 880 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 881 struct netdev_phys_item_id *b) 882 { 883 return a->id_len == b->id_len && 884 memcmp(a->id, b->id, a->id_len) == 0; 885 } 886 887 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 888 struct sk_buff *skb, 889 struct net_device *sb_dev); 890 891 enum net_device_path_type { 892 DEV_PATH_ETHERNET = 0, 893 DEV_PATH_VLAN, 894 DEV_PATH_BRIDGE, 895 DEV_PATH_PPPOE, 896 DEV_PATH_DSA, 897 DEV_PATH_MTK_WDMA, 898 }; 899 900 struct net_device_path { 901 enum net_device_path_type type; 902 const struct net_device *dev; 903 union { 904 struct { 905 u16 id; 906 __be16 proto; 907 u8 h_dest[ETH_ALEN]; 908 } encap; 909 struct { 910 enum { 911 DEV_PATH_BR_VLAN_KEEP, 912 DEV_PATH_BR_VLAN_TAG, 913 DEV_PATH_BR_VLAN_UNTAG, 914 DEV_PATH_BR_VLAN_UNTAG_HW, 915 } vlan_mode; 916 u16 vlan_id; 917 __be16 vlan_proto; 918 } bridge; 919 struct { 920 int port; 921 u16 proto; 922 } dsa; 923 struct { 924 u8 wdma_idx; 925 u8 queue; 926 u16 wcid; 927 u8 bss; 928 } mtk_wdma; 929 }; 930 }; 931 932 #define NET_DEVICE_PATH_STACK_MAX 5 933 #define NET_DEVICE_PATH_VLAN_MAX 2 934 935 struct net_device_path_stack { 936 int num_paths; 937 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 938 }; 939 940 struct net_device_path_ctx { 941 const struct net_device *dev; 942 u8 daddr[ETH_ALEN]; 943 944 int num_vlans; 945 struct { 946 u16 id; 947 __be16 proto; 948 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 949 }; 950 951 enum tc_setup_type { 952 TC_QUERY_CAPS, 953 TC_SETUP_QDISC_MQPRIO, 954 TC_SETUP_CLSU32, 955 TC_SETUP_CLSFLOWER, 956 TC_SETUP_CLSMATCHALL, 957 TC_SETUP_CLSBPF, 958 TC_SETUP_BLOCK, 959 TC_SETUP_QDISC_CBS, 960 TC_SETUP_QDISC_RED, 961 TC_SETUP_QDISC_PRIO, 962 TC_SETUP_QDISC_MQ, 963 TC_SETUP_QDISC_ETF, 964 TC_SETUP_ROOT_QDISC, 965 TC_SETUP_QDISC_GRED, 966 TC_SETUP_QDISC_TAPRIO, 967 TC_SETUP_FT, 968 TC_SETUP_QDISC_ETS, 969 TC_SETUP_QDISC_TBF, 970 TC_SETUP_QDISC_FIFO, 971 TC_SETUP_QDISC_HTB, 972 TC_SETUP_ACT, 973 }; 974 975 /* These structures hold the attributes of bpf state that are being passed 976 * to the netdevice through the bpf op. 977 */ 978 enum bpf_netdev_command { 979 /* Set or clear a bpf program used in the earliest stages of packet 980 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 981 * is responsible for calling bpf_prog_put on any old progs that are 982 * stored. In case of error, the callee need not release the new prog 983 * reference, but on success it takes ownership and must bpf_prog_put 984 * when it is no longer used. 985 */ 986 XDP_SETUP_PROG, 987 XDP_SETUP_PROG_HW, 988 /* BPF program for offload callbacks, invoked at program load time. */ 989 BPF_OFFLOAD_MAP_ALLOC, 990 BPF_OFFLOAD_MAP_FREE, 991 XDP_SETUP_XSK_POOL, 992 }; 993 994 struct bpf_prog_offload_ops; 995 struct netlink_ext_ack; 996 struct xdp_umem; 997 struct xdp_dev_bulk_queue; 998 struct bpf_xdp_link; 999 1000 enum bpf_xdp_mode { 1001 XDP_MODE_SKB = 0, 1002 XDP_MODE_DRV = 1, 1003 XDP_MODE_HW = 2, 1004 __MAX_XDP_MODE 1005 }; 1006 1007 struct bpf_xdp_entity { 1008 struct bpf_prog *prog; 1009 struct bpf_xdp_link *link; 1010 }; 1011 1012 struct netdev_bpf { 1013 enum bpf_netdev_command command; 1014 union { 1015 /* XDP_SETUP_PROG */ 1016 struct { 1017 u32 flags; 1018 struct bpf_prog *prog; 1019 struct netlink_ext_ack *extack; 1020 }; 1021 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 1022 struct { 1023 struct bpf_offloaded_map *offmap; 1024 }; 1025 /* XDP_SETUP_XSK_POOL */ 1026 struct { 1027 struct xsk_buff_pool *pool; 1028 u16 queue_id; 1029 } xsk; 1030 }; 1031 }; 1032 1033 /* Flags for ndo_xsk_wakeup. */ 1034 #define XDP_WAKEUP_RX (1 << 0) 1035 #define XDP_WAKEUP_TX (1 << 1) 1036 1037 #ifdef CONFIG_XFRM_OFFLOAD 1038 struct xfrmdev_ops { 1039 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack); 1040 void (*xdo_dev_state_delete) (struct xfrm_state *x); 1041 void (*xdo_dev_state_free) (struct xfrm_state *x); 1042 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 1043 struct xfrm_state *x); 1044 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1045 void (*xdo_dev_state_update_curlft) (struct xfrm_state *x); 1046 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack); 1047 void (*xdo_dev_policy_delete) (struct xfrm_policy *x); 1048 void (*xdo_dev_policy_free) (struct xfrm_policy *x); 1049 }; 1050 #endif 1051 1052 struct dev_ifalias { 1053 struct rcu_head rcuhead; 1054 char ifalias[]; 1055 }; 1056 1057 struct devlink; 1058 struct tlsdev_ops; 1059 1060 struct netdev_net_notifier { 1061 struct list_head list; 1062 struct notifier_block *nb; 1063 }; 1064 1065 /* 1066 * This structure defines the management hooks for network devices. 1067 * The following hooks can be defined; unless noted otherwise, they are 1068 * optional and can be filled with a null pointer. 1069 * 1070 * int (*ndo_init)(struct net_device *dev); 1071 * This function is called once when a network device is registered. 1072 * The network device can use this for any late stage initialization 1073 * or semantic validation. It can fail with an error code which will 1074 * be propagated back to register_netdev. 1075 * 1076 * void (*ndo_uninit)(struct net_device *dev); 1077 * This function is called when device is unregistered or when registration 1078 * fails. It is not called if init fails. 1079 * 1080 * int (*ndo_open)(struct net_device *dev); 1081 * This function is called when a network device transitions to the up 1082 * state. 1083 * 1084 * int (*ndo_stop)(struct net_device *dev); 1085 * This function is called when a network device transitions to the down 1086 * state. 1087 * 1088 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1089 * struct net_device *dev); 1090 * Called when a packet needs to be transmitted. 1091 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1092 * the queue before that can happen; it's for obsolete devices and weird 1093 * corner cases, but the stack really does a non-trivial amount 1094 * of useless work if you return NETDEV_TX_BUSY. 1095 * Required; cannot be NULL. 1096 * 1097 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1098 * struct net_device *dev 1099 * netdev_features_t features); 1100 * Called by core transmit path to determine if device is capable of 1101 * performing offload operations on a given packet. This is to give 1102 * the device an opportunity to implement any restrictions that cannot 1103 * be otherwise expressed by feature flags. The check is called with 1104 * the set of features that the stack has calculated and it returns 1105 * those the driver believes to be appropriate. 1106 * 1107 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1108 * struct net_device *sb_dev); 1109 * Called to decide which queue to use when device supports multiple 1110 * transmit queues. 1111 * 1112 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1113 * This function is called to allow device receiver to make 1114 * changes to configuration when multicast or promiscuous is enabled. 1115 * 1116 * void (*ndo_set_rx_mode)(struct net_device *dev); 1117 * This function is called device changes address list filtering. 1118 * If driver handles unicast address filtering, it should set 1119 * IFF_UNICAST_FLT in its priv_flags. 1120 * 1121 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1122 * This function is called when the Media Access Control address 1123 * needs to be changed. If this interface is not defined, the 1124 * MAC address can not be changed. 1125 * 1126 * int (*ndo_validate_addr)(struct net_device *dev); 1127 * Test if Media Access Control address is valid for the device. 1128 * 1129 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1130 * Old-style ioctl entry point. This is used internally by the 1131 * appletalk and ieee802154 subsystems but is no longer called by 1132 * the device ioctl handler. 1133 * 1134 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1135 * Used by the bonding driver for its device specific ioctls: 1136 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1137 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1138 * 1139 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1140 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1141 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1142 * 1143 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1144 * Used to set network devices bus interface parameters. This interface 1145 * is retained for legacy reasons; new devices should use the bus 1146 * interface (PCI) for low level management. 1147 * 1148 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1149 * Called when a user wants to change the Maximum Transfer Unit 1150 * of a device. 1151 * 1152 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1153 * Callback used when the transmitter has not made any progress 1154 * for dev->watchdog ticks. 1155 * 1156 * void (*ndo_get_stats64)(struct net_device *dev, 1157 * struct rtnl_link_stats64 *storage); 1158 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1159 * Called when a user wants to get the network device usage 1160 * statistics. Drivers must do one of the following: 1161 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1162 * rtnl_link_stats64 structure passed by the caller. 1163 * 2. Define @ndo_get_stats to update a net_device_stats structure 1164 * (which should normally be dev->stats) and return a pointer to 1165 * it. The structure may be changed asynchronously only if each 1166 * field is written atomically. 1167 * 3. Update dev->stats asynchronously and atomically, and define 1168 * neither operation. 1169 * 1170 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1171 * Return true if this device supports offload stats of this attr_id. 1172 * 1173 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1174 * void *attr_data) 1175 * Get statistics for offload operations by attr_id. Write it into the 1176 * attr_data pointer. 1177 * 1178 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1179 * If device supports VLAN filtering this function is called when a 1180 * VLAN id is registered. 1181 * 1182 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1183 * If device supports VLAN filtering this function is called when a 1184 * VLAN id is unregistered. 1185 * 1186 * void (*ndo_poll_controller)(struct net_device *dev); 1187 * 1188 * SR-IOV management functions. 1189 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1190 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1191 * u8 qos, __be16 proto); 1192 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1193 * int max_tx_rate); 1194 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1195 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1196 * int (*ndo_get_vf_config)(struct net_device *dev, 1197 * int vf, struct ifla_vf_info *ivf); 1198 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1199 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1200 * struct nlattr *port[]); 1201 * 1202 * Enable or disable the VF ability to query its RSS Redirection Table and 1203 * Hash Key. This is needed since on some devices VF share this information 1204 * with PF and querying it may introduce a theoretical security risk. 1205 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1206 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1207 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1208 * void *type_data); 1209 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1210 * This is always called from the stack with the rtnl lock held and netif 1211 * tx queues stopped. This allows the netdevice to perform queue 1212 * management safely. 1213 * 1214 * Fiber Channel over Ethernet (FCoE) offload functions. 1215 * int (*ndo_fcoe_enable)(struct net_device *dev); 1216 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1217 * so the underlying device can perform whatever needed configuration or 1218 * initialization to support acceleration of FCoE traffic. 1219 * 1220 * int (*ndo_fcoe_disable)(struct net_device *dev); 1221 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1222 * so the underlying device can perform whatever needed clean-ups to 1223 * stop supporting acceleration of FCoE traffic. 1224 * 1225 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1226 * struct scatterlist *sgl, unsigned int sgc); 1227 * Called when the FCoE Initiator wants to initialize an I/O that 1228 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1229 * perform necessary setup and returns 1 to indicate the device is set up 1230 * successfully to perform DDP on this I/O, otherwise this returns 0. 1231 * 1232 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1233 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1234 * indicated by the FC exchange id 'xid', so the underlying device can 1235 * clean up and reuse resources for later DDP requests. 1236 * 1237 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1238 * struct scatterlist *sgl, unsigned int sgc); 1239 * Called when the FCoE Target wants to initialize an I/O that 1240 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1241 * perform necessary setup and returns 1 to indicate the device is set up 1242 * successfully to perform DDP on this I/O, otherwise this returns 0. 1243 * 1244 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1245 * struct netdev_fcoe_hbainfo *hbainfo); 1246 * Called when the FCoE Protocol stack wants information on the underlying 1247 * device. This information is utilized by the FCoE protocol stack to 1248 * register attributes with Fiber Channel management service as per the 1249 * FC-GS Fabric Device Management Information(FDMI) specification. 1250 * 1251 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1252 * Called when the underlying device wants to override default World Wide 1253 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1254 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1255 * protocol stack to use. 1256 * 1257 * RFS acceleration. 1258 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1259 * u16 rxq_index, u32 flow_id); 1260 * Set hardware filter for RFS. rxq_index is the target queue index; 1261 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1262 * Return the filter ID on success, or a negative error code. 1263 * 1264 * Slave management functions (for bridge, bonding, etc). 1265 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1266 * Called to make another netdev an underling. 1267 * 1268 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1269 * Called to release previously enslaved netdev. 1270 * 1271 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1272 * struct sk_buff *skb, 1273 * bool all_slaves); 1274 * Get the xmit slave of master device. If all_slaves is true, function 1275 * assume all the slaves can transmit. 1276 * 1277 * Feature/offload setting functions. 1278 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1279 * netdev_features_t features); 1280 * Adjusts the requested feature flags according to device-specific 1281 * constraints, and returns the resulting flags. Must not modify 1282 * the device state. 1283 * 1284 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1285 * Called to update device configuration to new features. Passed 1286 * feature set might be less than what was returned by ndo_fix_features()). 1287 * Must return >0 or -errno if it changed dev->features itself. 1288 * 1289 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1290 * struct net_device *dev, 1291 * const unsigned char *addr, u16 vid, u16 flags, 1292 * struct netlink_ext_ack *extack); 1293 * Adds an FDB entry to dev for addr. 1294 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1295 * struct net_device *dev, 1296 * const unsigned char *addr, u16 vid) 1297 * Deletes the FDB entry from dev coresponding to addr. 1298 * int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, struct nlattr *tb[], 1299 * struct net_device *dev, 1300 * u16 vid, 1301 * struct netlink_ext_ack *extack); 1302 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1303 * struct net_device *dev, struct net_device *filter_dev, 1304 * int *idx) 1305 * Used to add FDB entries to dump requests. Implementers should add 1306 * entries to skb and update idx with the number of entries. 1307 * 1308 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1309 * u16 flags, struct netlink_ext_ack *extack) 1310 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1311 * struct net_device *dev, u32 filter_mask, 1312 * int nlflags) 1313 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1314 * u16 flags); 1315 * 1316 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1317 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1318 * which do not represent real hardware may define this to allow their 1319 * userspace components to manage their virtual carrier state. Devices 1320 * that determine carrier state from physical hardware properties (eg 1321 * network cables) or protocol-dependent mechanisms (eg 1322 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1323 * 1324 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1325 * struct netdev_phys_item_id *ppid); 1326 * Called to get ID of physical port of this device. If driver does 1327 * not implement this, it is assumed that the hw is not able to have 1328 * multiple net devices on single physical port. 1329 * 1330 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1331 * struct netdev_phys_item_id *ppid) 1332 * Called to get the parent ID of the physical port of this device. 1333 * 1334 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1335 * struct net_device *dev) 1336 * Called by upper layer devices to accelerate switching or other 1337 * station functionality into hardware. 'pdev is the lowerdev 1338 * to use for the offload and 'dev' is the net device that will 1339 * back the offload. Returns a pointer to the private structure 1340 * the upper layer will maintain. 1341 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1342 * Called by upper layer device to delete the station created 1343 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1344 * the station and priv is the structure returned by the add 1345 * operation. 1346 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1347 * int queue_index, u32 maxrate); 1348 * Called when a user wants to set a max-rate limitation of specific 1349 * TX queue. 1350 * int (*ndo_get_iflink)(const struct net_device *dev); 1351 * Called to get the iflink value of this device. 1352 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1353 * This function is used to get egress tunnel information for given skb. 1354 * This is useful for retrieving outer tunnel header parameters while 1355 * sampling packet. 1356 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1357 * This function is used to specify the headroom that the skb must 1358 * consider when allocation skb during packet reception. Setting 1359 * appropriate rx headroom value allows avoiding skb head copy on 1360 * forward. Setting a negative value resets the rx headroom to the 1361 * default value. 1362 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1363 * This function is used to set or query state related to XDP on the 1364 * netdevice and manage BPF offload. See definition of 1365 * enum bpf_netdev_command for details. 1366 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1367 * u32 flags); 1368 * This function is used to submit @n XDP packets for transmit on a 1369 * netdevice. Returns number of frames successfully transmitted, frames 1370 * that got dropped are freed/returned via xdp_return_frame(). 1371 * Returns negative number, means general error invoking ndo, meaning 1372 * no frames were xmit'ed and core-caller will free all frames. 1373 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1374 * struct xdp_buff *xdp); 1375 * Get the xmit slave of master device based on the xdp_buff. 1376 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1377 * This function is used to wake up the softirq, ksoftirqd or kthread 1378 * responsible for sending and/or receiving packets on a specific 1379 * queue id bound to an AF_XDP socket. The flags field specifies if 1380 * only RX, only Tx, or both should be woken up using the flags 1381 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1382 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1383 * int cmd); 1384 * Add, change, delete or get information on an IPv4 tunnel. 1385 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1386 * If a device is paired with a peer device, return the peer instance. 1387 * The caller must be under RCU read context. 1388 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1389 * Get the forwarding path to reach the real device from the HW destination address 1390 * ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1391 * const struct skb_shared_hwtstamps *hwtstamps, 1392 * bool cycles); 1393 * Get hardware timestamp based on normal/adjustable time or free running 1394 * cycle counter. This function is required if physical clock supports a 1395 * free running cycle counter. 1396 */ 1397 struct net_device_ops { 1398 int (*ndo_init)(struct net_device *dev); 1399 void (*ndo_uninit)(struct net_device *dev); 1400 int (*ndo_open)(struct net_device *dev); 1401 int (*ndo_stop)(struct net_device *dev); 1402 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1403 struct net_device *dev); 1404 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1405 struct net_device *dev, 1406 netdev_features_t features); 1407 u16 (*ndo_select_queue)(struct net_device *dev, 1408 struct sk_buff *skb, 1409 struct net_device *sb_dev); 1410 void (*ndo_change_rx_flags)(struct net_device *dev, 1411 int flags); 1412 void (*ndo_set_rx_mode)(struct net_device *dev); 1413 int (*ndo_set_mac_address)(struct net_device *dev, 1414 void *addr); 1415 int (*ndo_validate_addr)(struct net_device *dev); 1416 int (*ndo_do_ioctl)(struct net_device *dev, 1417 struct ifreq *ifr, int cmd); 1418 int (*ndo_eth_ioctl)(struct net_device *dev, 1419 struct ifreq *ifr, int cmd); 1420 int (*ndo_siocbond)(struct net_device *dev, 1421 struct ifreq *ifr, int cmd); 1422 int (*ndo_siocwandev)(struct net_device *dev, 1423 struct if_settings *ifs); 1424 int (*ndo_siocdevprivate)(struct net_device *dev, 1425 struct ifreq *ifr, 1426 void __user *data, int cmd); 1427 int (*ndo_set_config)(struct net_device *dev, 1428 struct ifmap *map); 1429 int (*ndo_change_mtu)(struct net_device *dev, 1430 int new_mtu); 1431 int (*ndo_neigh_setup)(struct net_device *dev, 1432 struct neigh_parms *); 1433 void (*ndo_tx_timeout) (struct net_device *dev, 1434 unsigned int txqueue); 1435 1436 void (*ndo_get_stats64)(struct net_device *dev, 1437 struct rtnl_link_stats64 *storage); 1438 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1439 int (*ndo_get_offload_stats)(int attr_id, 1440 const struct net_device *dev, 1441 void *attr_data); 1442 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1443 1444 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1445 __be16 proto, u16 vid); 1446 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1447 __be16 proto, u16 vid); 1448 #ifdef CONFIG_NET_POLL_CONTROLLER 1449 void (*ndo_poll_controller)(struct net_device *dev); 1450 int (*ndo_netpoll_setup)(struct net_device *dev, 1451 struct netpoll_info *info); 1452 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1453 #endif 1454 int (*ndo_set_vf_mac)(struct net_device *dev, 1455 int queue, u8 *mac); 1456 int (*ndo_set_vf_vlan)(struct net_device *dev, 1457 int queue, u16 vlan, 1458 u8 qos, __be16 proto); 1459 int (*ndo_set_vf_rate)(struct net_device *dev, 1460 int vf, int min_tx_rate, 1461 int max_tx_rate); 1462 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1463 int vf, bool setting); 1464 int (*ndo_set_vf_trust)(struct net_device *dev, 1465 int vf, bool setting); 1466 int (*ndo_get_vf_config)(struct net_device *dev, 1467 int vf, 1468 struct ifla_vf_info *ivf); 1469 int (*ndo_set_vf_link_state)(struct net_device *dev, 1470 int vf, int link_state); 1471 int (*ndo_get_vf_stats)(struct net_device *dev, 1472 int vf, 1473 struct ifla_vf_stats 1474 *vf_stats); 1475 int (*ndo_set_vf_port)(struct net_device *dev, 1476 int vf, 1477 struct nlattr *port[]); 1478 int (*ndo_get_vf_port)(struct net_device *dev, 1479 int vf, struct sk_buff *skb); 1480 int (*ndo_get_vf_guid)(struct net_device *dev, 1481 int vf, 1482 struct ifla_vf_guid *node_guid, 1483 struct ifla_vf_guid *port_guid); 1484 int (*ndo_set_vf_guid)(struct net_device *dev, 1485 int vf, u64 guid, 1486 int guid_type); 1487 int (*ndo_set_vf_rss_query_en)( 1488 struct net_device *dev, 1489 int vf, bool setting); 1490 int (*ndo_setup_tc)(struct net_device *dev, 1491 enum tc_setup_type type, 1492 void *type_data); 1493 #if IS_ENABLED(CONFIG_FCOE) 1494 int (*ndo_fcoe_enable)(struct net_device *dev); 1495 int (*ndo_fcoe_disable)(struct net_device *dev); 1496 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1497 u16 xid, 1498 struct scatterlist *sgl, 1499 unsigned int sgc); 1500 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1501 u16 xid); 1502 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1503 u16 xid, 1504 struct scatterlist *sgl, 1505 unsigned int sgc); 1506 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1507 struct netdev_fcoe_hbainfo *hbainfo); 1508 #endif 1509 1510 #if IS_ENABLED(CONFIG_LIBFCOE) 1511 #define NETDEV_FCOE_WWNN 0 1512 #define NETDEV_FCOE_WWPN 1 1513 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1514 u64 *wwn, int type); 1515 #endif 1516 1517 #ifdef CONFIG_RFS_ACCEL 1518 int (*ndo_rx_flow_steer)(struct net_device *dev, 1519 const struct sk_buff *skb, 1520 u16 rxq_index, 1521 u32 flow_id); 1522 #endif 1523 int (*ndo_add_slave)(struct net_device *dev, 1524 struct net_device *slave_dev, 1525 struct netlink_ext_ack *extack); 1526 int (*ndo_del_slave)(struct net_device *dev, 1527 struct net_device *slave_dev); 1528 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1529 struct sk_buff *skb, 1530 bool all_slaves); 1531 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1532 struct sock *sk); 1533 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1534 netdev_features_t features); 1535 int (*ndo_set_features)(struct net_device *dev, 1536 netdev_features_t features); 1537 int (*ndo_neigh_construct)(struct net_device *dev, 1538 struct neighbour *n); 1539 void (*ndo_neigh_destroy)(struct net_device *dev, 1540 struct neighbour *n); 1541 1542 int (*ndo_fdb_add)(struct ndmsg *ndm, 1543 struct nlattr *tb[], 1544 struct net_device *dev, 1545 const unsigned char *addr, 1546 u16 vid, 1547 u16 flags, 1548 struct netlink_ext_ack *extack); 1549 int (*ndo_fdb_del)(struct ndmsg *ndm, 1550 struct nlattr *tb[], 1551 struct net_device *dev, 1552 const unsigned char *addr, 1553 u16 vid, struct netlink_ext_ack *extack); 1554 int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, 1555 struct nlattr *tb[], 1556 struct net_device *dev, 1557 u16 vid, 1558 struct netlink_ext_ack *extack); 1559 int (*ndo_fdb_dump)(struct sk_buff *skb, 1560 struct netlink_callback *cb, 1561 struct net_device *dev, 1562 struct net_device *filter_dev, 1563 int *idx); 1564 int (*ndo_fdb_get)(struct sk_buff *skb, 1565 struct nlattr *tb[], 1566 struct net_device *dev, 1567 const unsigned char *addr, 1568 u16 vid, u32 portid, u32 seq, 1569 struct netlink_ext_ack *extack); 1570 int (*ndo_bridge_setlink)(struct net_device *dev, 1571 struct nlmsghdr *nlh, 1572 u16 flags, 1573 struct netlink_ext_ack *extack); 1574 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1575 u32 pid, u32 seq, 1576 struct net_device *dev, 1577 u32 filter_mask, 1578 int nlflags); 1579 int (*ndo_bridge_dellink)(struct net_device *dev, 1580 struct nlmsghdr *nlh, 1581 u16 flags); 1582 int (*ndo_change_carrier)(struct net_device *dev, 1583 bool new_carrier); 1584 int (*ndo_get_phys_port_id)(struct net_device *dev, 1585 struct netdev_phys_item_id *ppid); 1586 int (*ndo_get_port_parent_id)(struct net_device *dev, 1587 struct netdev_phys_item_id *ppid); 1588 int (*ndo_get_phys_port_name)(struct net_device *dev, 1589 char *name, size_t len); 1590 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1591 struct net_device *dev); 1592 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1593 void *priv); 1594 1595 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1596 int queue_index, 1597 u32 maxrate); 1598 int (*ndo_get_iflink)(const struct net_device *dev); 1599 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1600 struct sk_buff *skb); 1601 void (*ndo_set_rx_headroom)(struct net_device *dev, 1602 int needed_headroom); 1603 int (*ndo_bpf)(struct net_device *dev, 1604 struct netdev_bpf *bpf); 1605 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1606 struct xdp_frame **xdp, 1607 u32 flags); 1608 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1609 struct xdp_buff *xdp); 1610 int (*ndo_xsk_wakeup)(struct net_device *dev, 1611 u32 queue_id, u32 flags); 1612 int (*ndo_tunnel_ctl)(struct net_device *dev, 1613 struct ip_tunnel_parm *p, int cmd); 1614 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1615 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1616 struct net_device_path *path); 1617 ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1618 const struct skb_shared_hwtstamps *hwtstamps, 1619 bool cycles); 1620 }; 1621 1622 struct xdp_metadata_ops { 1623 int (*xmo_rx_timestamp)(const struct xdp_md *ctx, u64 *timestamp); 1624 int (*xmo_rx_hash)(const struct xdp_md *ctx, u32 *hash); 1625 }; 1626 1627 /** 1628 * enum netdev_priv_flags - &struct net_device priv_flags 1629 * 1630 * These are the &struct net_device, they are only set internally 1631 * by drivers and used in the kernel. These flags are invisible to 1632 * userspace; this means that the order of these flags can change 1633 * during any kernel release. 1634 * 1635 * You should have a pretty good reason to be extending these flags. 1636 * 1637 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1638 * @IFF_EBRIDGE: Ethernet bridging device 1639 * @IFF_BONDING: bonding master or slave 1640 * @IFF_ISATAP: ISATAP interface (RFC4214) 1641 * @IFF_WAN_HDLC: WAN HDLC device 1642 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1643 * release skb->dst 1644 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1645 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1646 * @IFF_MACVLAN_PORT: device used as macvlan port 1647 * @IFF_BRIDGE_PORT: device used as bridge port 1648 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1649 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1650 * @IFF_UNICAST_FLT: Supports unicast filtering 1651 * @IFF_TEAM_PORT: device used as team port 1652 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1653 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1654 * change when it's running 1655 * @IFF_MACVLAN: Macvlan device 1656 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1657 * underlying stacked devices 1658 * @IFF_L3MDEV_MASTER: device is an L3 master device 1659 * @IFF_NO_QUEUE: device can run without qdisc attached 1660 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1661 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1662 * @IFF_TEAM: device is a team device 1663 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1664 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1665 * entity (i.e. the master device for bridged veth) 1666 * @IFF_MACSEC: device is a MACsec device 1667 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1668 * @IFF_FAILOVER: device is a failover master device 1669 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1670 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1671 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf 1672 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1673 * skb_headlen(skb) == 0 (data starts from frag0) 1674 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN 1675 */ 1676 enum netdev_priv_flags { 1677 IFF_802_1Q_VLAN = 1<<0, 1678 IFF_EBRIDGE = 1<<1, 1679 IFF_BONDING = 1<<2, 1680 IFF_ISATAP = 1<<3, 1681 IFF_WAN_HDLC = 1<<4, 1682 IFF_XMIT_DST_RELEASE = 1<<5, 1683 IFF_DONT_BRIDGE = 1<<6, 1684 IFF_DISABLE_NETPOLL = 1<<7, 1685 IFF_MACVLAN_PORT = 1<<8, 1686 IFF_BRIDGE_PORT = 1<<9, 1687 IFF_OVS_DATAPATH = 1<<10, 1688 IFF_TX_SKB_SHARING = 1<<11, 1689 IFF_UNICAST_FLT = 1<<12, 1690 IFF_TEAM_PORT = 1<<13, 1691 IFF_SUPP_NOFCS = 1<<14, 1692 IFF_LIVE_ADDR_CHANGE = 1<<15, 1693 IFF_MACVLAN = 1<<16, 1694 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1695 IFF_L3MDEV_MASTER = 1<<18, 1696 IFF_NO_QUEUE = 1<<19, 1697 IFF_OPENVSWITCH = 1<<20, 1698 IFF_L3MDEV_SLAVE = 1<<21, 1699 IFF_TEAM = 1<<22, 1700 IFF_RXFH_CONFIGURED = 1<<23, 1701 IFF_PHONY_HEADROOM = 1<<24, 1702 IFF_MACSEC = 1<<25, 1703 IFF_NO_RX_HANDLER = 1<<26, 1704 IFF_FAILOVER = 1<<27, 1705 IFF_FAILOVER_SLAVE = 1<<28, 1706 IFF_L3MDEV_RX_HANDLER = 1<<29, 1707 IFF_NO_ADDRCONF = BIT_ULL(30), 1708 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31), 1709 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32), 1710 }; 1711 1712 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1713 #define IFF_EBRIDGE IFF_EBRIDGE 1714 #define IFF_BONDING IFF_BONDING 1715 #define IFF_ISATAP IFF_ISATAP 1716 #define IFF_WAN_HDLC IFF_WAN_HDLC 1717 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1718 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1719 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1720 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1721 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1722 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1723 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1724 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1725 #define IFF_TEAM_PORT IFF_TEAM_PORT 1726 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1727 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1728 #define IFF_MACVLAN IFF_MACVLAN 1729 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1730 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1731 #define IFF_NO_QUEUE IFF_NO_QUEUE 1732 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1733 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1734 #define IFF_TEAM IFF_TEAM 1735 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1736 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM 1737 #define IFF_MACSEC IFF_MACSEC 1738 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1739 #define IFF_FAILOVER IFF_FAILOVER 1740 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1741 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1742 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR 1743 1744 /* Specifies the type of the struct net_device::ml_priv pointer */ 1745 enum netdev_ml_priv_type { 1746 ML_PRIV_NONE, 1747 ML_PRIV_CAN, 1748 }; 1749 1750 /** 1751 * struct net_device - The DEVICE structure. 1752 * 1753 * Actually, this whole structure is a big mistake. It mixes I/O 1754 * data with strictly "high-level" data, and it has to know about 1755 * almost every data structure used in the INET module. 1756 * 1757 * @name: This is the first field of the "visible" part of this structure 1758 * (i.e. as seen by users in the "Space.c" file). It is the name 1759 * of the interface. 1760 * 1761 * @name_node: Name hashlist node 1762 * @ifalias: SNMP alias 1763 * @mem_end: Shared memory end 1764 * @mem_start: Shared memory start 1765 * @base_addr: Device I/O address 1766 * @irq: Device IRQ number 1767 * 1768 * @state: Generic network queuing layer state, see netdev_state_t 1769 * @dev_list: The global list of network devices 1770 * @napi_list: List entry used for polling NAPI devices 1771 * @unreg_list: List entry when we are unregistering the 1772 * device; see the function unregister_netdev 1773 * @close_list: List entry used when we are closing the device 1774 * @ptype_all: Device-specific packet handlers for all protocols 1775 * @ptype_specific: Device-specific, protocol-specific packet handlers 1776 * 1777 * @adj_list: Directly linked devices, like slaves for bonding 1778 * @features: Currently active device features 1779 * @hw_features: User-changeable features 1780 * 1781 * @wanted_features: User-requested features 1782 * @vlan_features: Mask of features inheritable by VLAN devices 1783 * 1784 * @hw_enc_features: Mask of features inherited by encapsulating devices 1785 * This field indicates what encapsulation 1786 * offloads the hardware is capable of doing, 1787 * and drivers will need to set them appropriately. 1788 * 1789 * @mpls_features: Mask of features inheritable by MPLS 1790 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1791 * 1792 * @ifindex: interface index 1793 * @group: The group the device belongs to 1794 * 1795 * @stats: Statistics struct, which was left as a legacy, use 1796 * rtnl_link_stats64 instead 1797 * 1798 * @core_stats: core networking counters, 1799 * do not use this in drivers 1800 * @carrier_up_count: Number of times the carrier has been up 1801 * @carrier_down_count: Number of times the carrier has been down 1802 * 1803 * @wireless_handlers: List of functions to handle Wireless Extensions, 1804 * instead of ioctl, 1805 * see <net/iw_handler.h> for details. 1806 * @wireless_data: Instance data managed by the core of wireless extensions 1807 * 1808 * @netdev_ops: Includes several pointers to callbacks, 1809 * if one wants to override the ndo_*() functions 1810 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks. 1811 * @ethtool_ops: Management operations 1812 * @l3mdev_ops: Layer 3 master device operations 1813 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1814 * discovery handling. Necessary for e.g. 6LoWPAN. 1815 * @xfrmdev_ops: Transformation offload operations 1816 * @tlsdev_ops: Transport Layer Security offload operations 1817 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1818 * of Layer 2 headers. 1819 * 1820 * @flags: Interface flags (a la BSD) 1821 * @priv_flags: Like 'flags' but invisible to userspace, 1822 * see if.h for the definitions 1823 * @gflags: Global flags ( kept as legacy ) 1824 * @padded: How much padding added by alloc_netdev() 1825 * @operstate: RFC2863 operstate 1826 * @link_mode: Mapping policy to operstate 1827 * @if_port: Selectable AUI, TP, ... 1828 * @dma: DMA channel 1829 * @mtu: Interface MTU value 1830 * @min_mtu: Interface Minimum MTU value 1831 * @max_mtu: Interface Maximum MTU value 1832 * @type: Interface hardware type 1833 * @hard_header_len: Maximum hardware header length. 1834 * @min_header_len: Minimum hardware header length 1835 * 1836 * @needed_headroom: Extra headroom the hardware may need, but not in all 1837 * cases can this be guaranteed 1838 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1839 * cases can this be guaranteed. Some cases also use 1840 * LL_MAX_HEADER instead to allocate the skb 1841 * 1842 * interface address info: 1843 * 1844 * @perm_addr: Permanent hw address 1845 * @addr_assign_type: Hw address assignment type 1846 * @addr_len: Hardware address length 1847 * @upper_level: Maximum depth level of upper devices. 1848 * @lower_level: Maximum depth level of lower devices. 1849 * @neigh_priv_len: Used in neigh_alloc() 1850 * @dev_id: Used to differentiate devices that share 1851 * the same link layer address 1852 * @dev_port: Used to differentiate devices that share 1853 * the same function 1854 * @addr_list_lock: XXX: need comments on this one 1855 * @name_assign_type: network interface name assignment type 1856 * @uc_promisc: Counter that indicates promiscuous mode 1857 * has been enabled due to the need to listen to 1858 * additional unicast addresses in a device that 1859 * does not implement ndo_set_rx_mode() 1860 * @uc: unicast mac addresses 1861 * @mc: multicast mac addresses 1862 * @dev_addrs: list of device hw addresses 1863 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1864 * @promiscuity: Number of times the NIC is told to work in 1865 * promiscuous mode; if it becomes 0 the NIC will 1866 * exit promiscuous mode 1867 * @allmulti: Counter, enables or disables allmulticast mode 1868 * 1869 * @vlan_info: VLAN info 1870 * @dsa_ptr: dsa specific data 1871 * @tipc_ptr: TIPC specific data 1872 * @atalk_ptr: AppleTalk link 1873 * @ip_ptr: IPv4 specific data 1874 * @ip6_ptr: IPv6 specific data 1875 * @ax25_ptr: AX.25 specific data 1876 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1877 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1878 * device struct 1879 * @mpls_ptr: mpls_dev struct pointer 1880 * @mctp_ptr: MCTP specific data 1881 * 1882 * @dev_addr: Hw address (before bcast, 1883 * because most packets are unicast) 1884 * 1885 * @_rx: Array of RX queues 1886 * @num_rx_queues: Number of RX queues 1887 * allocated at register_netdev() time 1888 * @real_num_rx_queues: Number of RX queues currently active in device 1889 * @xdp_prog: XDP sockets filter program pointer 1890 * @gro_flush_timeout: timeout for GRO layer in NAPI 1891 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1892 * allow to avoid NIC hard IRQ, on busy queues. 1893 * 1894 * @rx_handler: handler for received packets 1895 * @rx_handler_data: XXX: need comments on this one 1896 * @miniq_ingress: ingress/clsact qdisc specific data for 1897 * ingress processing 1898 * @ingress_queue: XXX: need comments on this one 1899 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1900 * @broadcast: hw bcast address 1901 * 1902 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1903 * indexed by RX queue number. Assigned by driver. 1904 * This must only be set if the ndo_rx_flow_steer 1905 * operation is defined 1906 * @index_hlist: Device index hash chain 1907 * 1908 * @_tx: Array of TX queues 1909 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1910 * @real_num_tx_queues: Number of TX queues currently active in device 1911 * @qdisc: Root qdisc from userspace point of view 1912 * @tx_queue_len: Max frames per queue allowed 1913 * @tx_global_lock: XXX: need comments on this one 1914 * @xdp_bulkq: XDP device bulk queue 1915 * @xps_maps: all CPUs/RXQs maps for XPS device 1916 * 1917 * @xps_maps: XXX: need comments on this one 1918 * @miniq_egress: clsact qdisc specific data for 1919 * egress processing 1920 * @nf_hooks_egress: netfilter hooks executed for egress packets 1921 * @qdisc_hash: qdisc hash table 1922 * @watchdog_timeo: Represents the timeout that is used by 1923 * the watchdog (see dev_watchdog()) 1924 * @watchdog_timer: List of timers 1925 * 1926 * @proto_down_reason: reason a netdev interface is held down 1927 * @pcpu_refcnt: Number of references to this device 1928 * @dev_refcnt: Number of references to this device 1929 * @refcnt_tracker: Tracker directory for tracked references to this device 1930 * @todo_list: Delayed register/unregister 1931 * @link_watch_list: XXX: need comments on this one 1932 * 1933 * @reg_state: Register/unregister state machine 1934 * @dismantle: Device is going to be freed 1935 * @rtnl_link_state: This enum represents the phases of creating 1936 * a new link 1937 * 1938 * @needs_free_netdev: Should unregister perform free_netdev? 1939 * @priv_destructor: Called from unregister 1940 * @npinfo: XXX: need comments on this one 1941 * @nd_net: Network namespace this network device is inside 1942 * 1943 * @ml_priv: Mid-layer private 1944 * @ml_priv_type: Mid-layer private type 1945 * @lstats: Loopback statistics 1946 * @tstats: Tunnel statistics 1947 * @dstats: Dummy statistics 1948 * @vstats: Virtual ethernet statistics 1949 * 1950 * @garp_port: GARP 1951 * @mrp_port: MRP 1952 * 1953 * @dm_private: Drop monitor private 1954 * 1955 * @dev: Class/net/name entry 1956 * @sysfs_groups: Space for optional device, statistics and wireless 1957 * sysfs groups 1958 * 1959 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1960 * @rtnl_link_ops: Rtnl_link_ops 1961 * 1962 * @gso_max_size: Maximum size of generic segmentation offload 1963 * @tso_max_size: Device (as in HW) limit on the max TSO request size 1964 * @gso_max_segs: Maximum number of segments that can be passed to the 1965 * NIC for GSO 1966 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count 1967 * 1968 * @dcbnl_ops: Data Center Bridging netlink ops 1969 * @num_tc: Number of traffic classes in the net device 1970 * @tc_to_txq: XXX: need comments on this one 1971 * @prio_tc_map: XXX: need comments on this one 1972 * 1973 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1974 * 1975 * @priomap: XXX: need comments on this one 1976 * @phydev: Physical device may attach itself 1977 * for hardware timestamping 1978 * @sfp_bus: attached &struct sfp_bus structure. 1979 * 1980 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1981 * 1982 * @proto_down: protocol port state information can be sent to the 1983 * switch driver and used to set the phys state of the 1984 * switch port. 1985 * 1986 * @wol_enabled: Wake-on-LAN is enabled 1987 * 1988 * @threaded: napi threaded mode is enabled 1989 * 1990 * @net_notifier_list: List of per-net netdev notifier block 1991 * that follow this device when it is moved 1992 * to another network namespace. 1993 * 1994 * @macsec_ops: MACsec offloading ops 1995 * 1996 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1997 * offload capabilities of the device 1998 * @udp_tunnel_nic: UDP tunnel offload state 1999 * @xdp_state: stores info on attached XDP BPF programs 2000 * 2001 * @nested_level: Used as a parameter of spin_lock_nested() of 2002 * dev->addr_list_lock. 2003 * @unlink_list: As netif_addr_lock() can be called recursively, 2004 * keep a list of interfaces to be deleted. 2005 * @gro_max_size: Maximum size of aggregated packet in generic 2006 * receive offload (GRO) 2007 * 2008 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 2009 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 2010 * @watchdog_dev_tracker: refcount tracker used by watchdog. 2011 * @dev_registered_tracker: tracker for reference held while 2012 * registered 2013 * @offload_xstats_l3: L3 HW stats for this netdevice. 2014 * 2015 * @devlink_port: Pointer to related devlink port structure. 2016 * Assigned by a driver before netdev registration using 2017 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static 2018 * during the time netdevice is registered. 2019 * 2020 * FIXME: cleanup struct net_device such that network protocol info 2021 * moves out. 2022 */ 2023 2024 struct net_device { 2025 char name[IFNAMSIZ]; 2026 struct netdev_name_node *name_node; 2027 struct dev_ifalias __rcu *ifalias; 2028 /* 2029 * I/O specific fields 2030 * FIXME: Merge these and struct ifmap into one 2031 */ 2032 unsigned long mem_end; 2033 unsigned long mem_start; 2034 unsigned long base_addr; 2035 2036 /* 2037 * Some hardware also needs these fields (state,dev_list, 2038 * napi_list,unreg_list,close_list) but they are not 2039 * part of the usual set specified in Space.c. 2040 */ 2041 2042 unsigned long state; 2043 2044 struct list_head dev_list; 2045 struct list_head napi_list; 2046 struct list_head unreg_list; 2047 struct list_head close_list; 2048 struct list_head ptype_all; 2049 struct list_head ptype_specific; 2050 2051 struct { 2052 struct list_head upper; 2053 struct list_head lower; 2054 } adj_list; 2055 2056 /* Read-mostly cache-line for fast-path access */ 2057 unsigned int flags; 2058 unsigned long long priv_flags; 2059 const struct net_device_ops *netdev_ops; 2060 const struct xdp_metadata_ops *xdp_metadata_ops; 2061 int ifindex; 2062 unsigned short gflags; 2063 unsigned short hard_header_len; 2064 2065 /* Note : dev->mtu is often read without holding a lock. 2066 * Writers usually hold RTNL. 2067 * It is recommended to use READ_ONCE() to annotate the reads, 2068 * and to use WRITE_ONCE() to annotate the writes. 2069 */ 2070 unsigned int mtu; 2071 unsigned short needed_headroom; 2072 unsigned short needed_tailroom; 2073 2074 netdev_features_t features; 2075 netdev_features_t hw_features; 2076 netdev_features_t wanted_features; 2077 netdev_features_t vlan_features; 2078 netdev_features_t hw_enc_features; 2079 netdev_features_t mpls_features; 2080 netdev_features_t gso_partial_features; 2081 2082 unsigned int min_mtu; 2083 unsigned int max_mtu; 2084 unsigned short type; 2085 unsigned char min_header_len; 2086 unsigned char name_assign_type; 2087 2088 int group; 2089 2090 struct net_device_stats stats; /* not used by modern drivers */ 2091 2092 struct net_device_core_stats __percpu *core_stats; 2093 2094 /* Stats to monitor link on/off, flapping */ 2095 atomic_t carrier_up_count; 2096 atomic_t carrier_down_count; 2097 2098 #ifdef CONFIG_WIRELESS_EXT 2099 const struct iw_handler_def *wireless_handlers; 2100 struct iw_public_data *wireless_data; 2101 #endif 2102 const struct ethtool_ops *ethtool_ops; 2103 #ifdef CONFIG_NET_L3_MASTER_DEV 2104 const struct l3mdev_ops *l3mdev_ops; 2105 #endif 2106 #if IS_ENABLED(CONFIG_IPV6) 2107 const struct ndisc_ops *ndisc_ops; 2108 #endif 2109 2110 #ifdef CONFIG_XFRM_OFFLOAD 2111 const struct xfrmdev_ops *xfrmdev_ops; 2112 #endif 2113 2114 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2115 const struct tlsdev_ops *tlsdev_ops; 2116 #endif 2117 2118 const struct header_ops *header_ops; 2119 2120 unsigned char operstate; 2121 unsigned char link_mode; 2122 2123 unsigned char if_port; 2124 unsigned char dma; 2125 2126 /* Interface address info. */ 2127 unsigned char perm_addr[MAX_ADDR_LEN]; 2128 unsigned char addr_assign_type; 2129 unsigned char addr_len; 2130 unsigned char upper_level; 2131 unsigned char lower_level; 2132 2133 unsigned short neigh_priv_len; 2134 unsigned short dev_id; 2135 unsigned short dev_port; 2136 unsigned short padded; 2137 2138 spinlock_t addr_list_lock; 2139 int irq; 2140 2141 struct netdev_hw_addr_list uc; 2142 struct netdev_hw_addr_list mc; 2143 struct netdev_hw_addr_list dev_addrs; 2144 2145 #ifdef CONFIG_SYSFS 2146 struct kset *queues_kset; 2147 #endif 2148 #ifdef CONFIG_LOCKDEP 2149 struct list_head unlink_list; 2150 #endif 2151 unsigned int promiscuity; 2152 unsigned int allmulti; 2153 bool uc_promisc; 2154 #ifdef CONFIG_LOCKDEP 2155 unsigned char nested_level; 2156 #endif 2157 2158 2159 /* Protocol-specific pointers */ 2160 2161 struct in_device __rcu *ip_ptr; 2162 struct inet6_dev __rcu *ip6_ptr; 2163 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2164 struct vlan_info __rcu *vlan_info; 2165 #endif 2166 #if IS_ENABLED(CONFIG_NET_DSA) 2167 struct dsa_port *dsa_ptr; 2168 #endif 2169 #if IS_ENABLED(CONFIG_TIPC) 2170 struct tipc_bearer __rcu *tipc_ptr; 2171 #endif 2172 #if IS_ENABLED(CONFIG_ATALK) 2173 void *atalk_ptr; 2174 #endif 2175 #if IS_ENABLED(CONFIG_AX25) 2176 void *ax25_ptr; 2177 #endif 2178 #if IS_ENABLED(CONFIG_CFG80211) 2179 struct wireless_dev *ieee80211_ptr; 2180 #endif 2181 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2182 struct wpan_dev *ieee802154_ptr; 2183 #endif 2184 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2185 struct mpls_dev __rcu *mpls_ptr; 2186 #endif 2187 #if IS_ENABLED(CONFIG_MCTP) 2188 struct mctp_dev __rcu *mctp_ptr; 2189 #endif 2190 2191 /* 2192 * Cache lines mostly used on receive path (including eth_type_trans()) 2193 */ 2194 /* Interface address info used in eth_type_trans() */ 2195 const unsigned char *dev_addr; 2196 2197 struct netdev_rx_queue *_rx; 2198 unsigned int num_rx_queues; 2199 unsigned int real_num_rx_queues; 2200 2201 struct bpf_prog __rcu *xdp_prog; 2202 unsigned long gro_flush_timeout; 2203 int napi_defer_hard_irqs; 2204 #define GRO_LEGACY_MAX_SIZE 65536u 2205 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2206 * and shinfo->gso_segs is a 16bit field. 2207 */ 2208 #define GRO_MAX_SIZE (8 * 65535u) 2209 unsigned int gro_max_size; 2210 rx_handler_func_t __rcu *rx_handler; 2211 void __rcu *rx_handler_data; 2212 2213 #ifdef CONFIG_NET_CLS_ACT 2214 struct mini_Qdisc __rcu *miniq_ingress; 2215 #endif 2216 struct netdev_queue __rcu *ingress_queue; 2217 #ifdef CONFIG_NETFILTER_INGRESS 2218 struct nf_hook_entries __rcu *nf_hooks_ingress; 2219 #endif 2220 2221 unsigned char broadcast[MAX_ADDR_LEN]; 2222 #ifdef CONFIG_RFS_ACCEL 2223 struct cpu_rmap *rx_cpu_rmap; 2224 #endif 2225 struct hlist_node index_hlist; 2226 2227 /* 2228 * Cache lines mostly used on transmit path 2229 */ 2230 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2231 unsigned int num_tx_queues; 2232 unsigned int real_num_tx_queues; 2233 struct Qdisc __rcu *qdisc; 2234 unsigned int tx_queue_len; 2235 spinlock_t tx_global_lock; 2236 2237 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2238 2239 #ifdef CONFIG_XPS 2240 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2241 #endif 2242 #ifdef CONFIG_NET_CLS_ACT 2243 struct mini_Qdisc __rcu *miniq_egress; 2244 #endif 2245 #ifdef CONFIG_NETFILTER_EGRESS 2246 struct nf_hook_entries __rcu *nf_hooks_egress; 2247 #endif 2248 2249 #ifdef CONFIG_NET_SCHED 2250 DECLARE_HASHTABLE (qdisc_hash, 4); 2251 #endif 2252 /* These may be needed for future network-power-down code. */ 2253 struct timer_list watchdog_timer; 2254 int watchdog_timeo; 2255 2256 u32 proto_down_reason; 2257 2258 struct list_head todo_list; 2259 2260 #ifdef CONFIG_PCPU_DEV_REFCNT 2261 int __percpu *pcpu_refcnt; 2262 #else 2263 refcount_t dev_refcnt; 2264 #endif 2265 struct ref_tracker_dir refcnt_tracker; 2266 2267 struct list_head link_watch_list; 2268 2269 enum { NETREG_UNINITIALIZED=0, 2270 NETREG_REGISTERED, /* completed register_netdevice */ 2271 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2272 NETREG_UNREGISTERED, /* completed unregister todo */ 2273 NETREG_RELEASED, /* called free_netdev */ 2274 NETREG_DUMMY, /* dummy device for NAPI poll */ 2275 } reg_state:8; 2276 2277 bool dismantle; 2278 2279 enum { 2280 RTNL_LINK_INITIALIZED, 2281 RTNL_LINK_INITIALIZING, 2282 } rtnl_link_state:16; 2283 2284 bool needs_free_netdev; 2285 void (*priv_destructor)(struct net_device *dev); 2286 2287 #ifdef CONFIG_NETPOLL 2288 struct netpoll_info __rcu *npinfo; 2289 #endif 2290 2291 possible_net_t nd_net; 2292 2293 /* mid-layer private */ 2294 void *ml_priv; 2295 enum netdev_ml_priv_type ml_priv_type; 2296 2297 union { 2298 struct pcpu_lstats __percpu *lstats; 2299 struct pcpu_sw_netstats __percpu *tstats; 2300 struct pcpu_dstats __percpu *dstats; 2301 }; 2302 2303 #if IS_ENABLED(CONFIG_GARP) 2304 struct garp_port __rcu *garp_port; 2305 #endif 2306 #if IS_ENABLED(CONFIG_MRP) 2307 struct mrp_port __rcu *mrp_port; 2308 #endif 2309 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2310 struct dm_hw_stat_delta __rcu *dm_private; 2311 #endif 2312 struct device dev; 2313 const struct attribute_group *sysfs_groups[4]; 2314 const struct attribute_group *sysfs_rx_queue_group; 2315 2316 const struct rtnl_link_ops *rtnl_link_ops; 2317 2318 /* for setting kernel sock attribute on TCP connection setup */ 2319 #define GSO_MAX_SEGS 65535u 2320 #define GSO_LEGACY_MAX_SIZE 65536u 2321 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2322 * and shinfo->gso_segs is a 16bit field. 2323 */ 2324 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2325 2326 unsigned int gso_max_size; 2327 #define TSO_LEGACY_MAX_SIZE 65536 2328 #define TSO_MAX_SIZE UINT_MAX 2329 unsigned int tso_max_size; 2330 u16 gso_max_segs; 2331 #define TSO_MAX_SEGS U16_MAX 2332 u16 tso_max_segs; 2333 2334 #ifdef CONFIG_DCB 2335 const struct dcbnl_rtnl_ops *dcbnl_ops; 2336 #endif 2337 s16 num_tc; 2338 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2339 u8 prio_tc_map[TC_BITMASK + 1]; 2340 2341 #if IS_ENABLED(CONFIG_FCOE) 2342 unsigned int fcoe_ddp_xid; 2343 #endif 2344 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2345 struct netprio_map __rcu *priomap; 2346 #endif 2347 struct phy_device *phydev; 2348 struct sfp_bus *sfp_bus; 2349 struct lock_class_key *qdisc_tx_busylock; 2350 bool proto_down; 2351 unsigned wol_enabled:1; 2352 unsigned threaded:1; 2353 2354 struct list_head net_notifier_list; 2355 2356 #if IS_ENABLED(CONFIG_MACSEC) 2357 /* MACsec management functions */ 2358 const struct macsec_ops *macsec_ops; 2359 #endif 2360 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2361 struct udp_tunnel_nic *udp_tunnel_nic; 2362 2363 /* protected by rtnl_lock */ 2364 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2365 2366 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2367 netdevice_tracker linkwatch_dev_tracker; 2368 netdevice_tracker watchdog_dev_tracker; 2369 netdevice_tracker dev_registered_tracker; 2370 struct rtnl_hw_stats64 *offload_xstats_l3; 2371 2372 struct devlink_port *devlink_port; 2373 }; 2374 #define to_net_dev(d) container_of(d, struct net_device, dev) 2375 2376 /* 2377 * Driver should use this to assign devlink port instance to a netdevice 2378 * before it registers the netdevice. Therefore devlink_port is static 2379 * during the netdev lifetime after it is registered. 2380 */ 2381 #define SET_NETDEV_DEVLINK_PORT(dev, port) \ 2382 ({ \ 2383 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \ 2384 ((dev)->devlink_port = (port)); \ 2385 }) 2386 2387 static inline bool netif_elide_gro(const struct net_device *dev) 2388 { 2389 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2390 return true; 2391 return false; 2392 } 2393 2394 #define NETDEV_ALIGN 32 2395 2396 static inline 2397 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2398 { 2399 return dev->prio_tc_map[prio & TC_BITMASK]; 2400 } 2401 2402 static inline 2403 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2404 { 2405 if (tc >= dev->num_tc) 2406 return -EINVAL; 2407 2408 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2409 return 0; 2410 } 2411 2412 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2413 void netdev_reset_tc(struct net_device *dev); 2414 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2415 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2416 2417 static inline 2418 int netdev_get_num_tc(struct net_device *dev) 2419 { 2420 return dev->num_tc; 2421 } 2422 2423 static inline void net_prefetch(void *p) 2424 { 2425 prefetch(p); 2426 #if L1_CACHE_BYTES < 128 2427 prefetch((u8 *)p + L1_CACHE_BYTES); 2428 #endif 2429 } 2430 2431 static inline void net_prefetchw(void *p) 2432 { 2433 prefetchw(p); 2434 #if L1_CACHE_BYTES < 128 2435 prefetchw((u8 *)p + L1_CACHE_BYTES); 2436 #endif 2437 } 2438 2439 void netdev_unbind_sb_channel(struct net_device *dev, 2440 struct net_device *sb_dev); 2441 int netdev_bind_sb_channel_queue(struct net_device *dev, 2442 struct net_device *sb_dev, 2443 u8 tc, u16 count, u16 offset); 2444 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2445 static inline int netdev_get_sb_channel(struct net_device *dev) 2446 { 2447 return max_t(int, -dev->num_tc, 0); 2448 } 2449 2450 static inline 2451 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2452 unsigned int index) 2453 { 2454 return &dev->_tx[index]; 2455 } 2456 2457 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2458 const struct sk_buff *skb) 2459 { 2460 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2461 } 2462 2463 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2464 void (*f)(struct net_device *, 2465 struct netdev_queue *, 2466 void *), 2467 void *arg) 2468 { 2469 unsigned int i; 2470 2471 for (i = 0; i < dev->num_tx_queues; i++) 2472 f(dev, &dev->_tx[i], arg); 2473 } 2474 2475 #define netdev_lockdep_set_classes(dev) \ 2476 { \ 2477 static struct lock_class_key qdisc_tx_busylock_key; \ 2478 static struct lock_class_key qdisc_xmit_lock_key; \ 2479 static struct lock_class_key dev_addr_list_lock_key; \ 2480 unsigned int i; \ 2481 \ 2482 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2483 lockdep_set_class(&(dev)->addr_list_lock, \ 2484 &dev_addr_list_lock_key); \ 2485 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2486 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2487 &qdisc_xmit_lock_key); \ 2488 } 2489 2490 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2491 struct net_device *sb_dev); 2492 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2493 struct sk_buff *skb, 2494 struct net_device *sb_dev); 2495 2496 /* returns the headroom that the master device needs to take in account 2497 * when forwarding to this dev 2498 */ 2499 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2500 { 2501 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2502 } 2503 2504 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2505 { 2506 if (dev->netdev_ops->ndo_set_rx_headroom) 2507 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2508 } 2509 2510 /* set the device rx headroom to the dev's default */ 2511 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2512 { 2513 netdev_set_rx_headroom(dev, -1); 2514 } 2515 2516 static inline void *netdev_get_ml_priv(struct net_device *dev, 2517 enum netdev_ml_priv_type type) 2518 { 2519 if (dev->ml_priv_type != type) 2520 return NULL; 2521 2522 return dev->ml_priv; 2523 } 2524 2525 static inline void netdev_set_ml_priv(struct net_device *dev, 2526 void *ml_priv, 2527 enum netdev_ml_priv_type type) 2528 { 2529 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2530 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2531 dev->ml_priv_type, type); 2532 WARN(!dev->ml_priv_type && dev->ml_priv, 2533 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2534 2535 dev->ml_priv = ml_priv; 2536 dev->ml_priv_type = type; 2537 } 2538 2539 /* 2540 * Net namespace inlines 2541 */ 2542 static inline 2543 struct net *dev_net(const struct net_device *dev) 2544 { 2545 return read_pnet(&dev->nd_net); 2546 } 2547 2548 static inline 2549 void dev_net_set(struct net_device *dev, struct net *net) 2550 { 2551 write_pnet(&dev->nd_net, net); 2552 } 2553 2554 /** 2555 * netdev_priv - access network device private data 2556 * @dev: network device 2557 * 2558 * Get network device private data 2559 */ 2560 static inline void *netdev_priv(const struct net_device *dev) 2561 { 2562 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2563 } 2564 2565 /* Set the sysfs physical device reference for the network logical device 2566 * if set prior to registration will cause a symlink during initialization. 2567 */ 2568 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2569 2570 /* Set the sysfs device type for the network logical device to allow 2571 * fine-grained identification of different network device types. For 2572 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2573 */ 2574 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2575 2576 /* Default NAPI poll() weight 2577 * Device drivers are strongly advised to not use bigger value 2578 */ 2579 #define NAPI_POLL_WEIGHT 64 2580 2581 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2582 int (*poll)(struct napi_struct *, int), int weight); 2583 2584 /** 2585 * netif_napi_add() - initialize a NAPI context 2586 * @dev: network device 2587 * @napi: NAPI context 2588 * @poll: polling function 2589 * 2590 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2591 * *any* of the other NAPI-related functions. 2592 */ 2593 static inline void 2594 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2595 int (*poll)(struct napi_struct *, int)) 2596 { 2597 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2598 } 2599 2600 static inline void 2601 netif_napi_add_tx_weight(struct net_device *dev, 2602 struct napi_struct *napi, 2603 int (*poll)(struct napi_struct *, int), 2604 int weight) 2605 { 2606 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2607 netif_napi_add_weight(dev, napi, poll, weight); 2608 } 2609 2610 /** 2611 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2612 * @dev: network device 2613 * @napi: NAPI context 2614 * @poll: polling function 2615 * 2616 * This variant of netif_napi_add() should be used from drivers using NAPI 2617 * to exclusively poll a TX queue. 2618 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2619 */ 2620 static inline void netif_napi_add_tx(struct net_device *dev, 2621 struct napi_struct *napi, 2622 int (*poll)(struct napi_struct *, int)) 2623 { 2624 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2625 } 2626 2627 /** 2628 * __netif_napi_del - remove a NAPI context 2629 * @napi: NAPI context 2630 * 2631 * Warning: caller must observe RCU grace period before freeing memory 2632 * containing @napi. Drivers might want to call this helper to combine 2633 * all the needed RCU grace periods into a single one. 2634 */ 2635 void __netif_napi_del(struct napi_struct *napi); 2636 2637 /** 2638 * netif_napi_del - remove a NAPI context 2639 * @napi: NAPI context 2640 * 2641 * netif_napi_del() removes a NAPI context from the network device NAPI list 2642 */ 2643 static inline void netif_napi_del(struct napi_struct *napi) 2644 { 2645 __netif_napi_del(napi); 2646 synchronize_net(); 2647 } 2648 2649 struct packet_type { 2650 __be16 type; /* This is really htons(ether_type). */ 2651 bool ignore_outgoing; 2652 struct net_device *dev; /* NULL is wildcarded here */ 2653 netdevice_tracker dev_tracker; 2654 int (*func) (struct sk_buff *, 2655 struct net_device *, 2656 struct packet_type *, 2657 struct net_device *); 2658 void (*list_func) (struct list_head *, 2659 struct packet_type *, 2660 struct net_device *); 2661 bool (*id_match)(struct packet_type *ptype, 2662 struct sock *sk); 2663 struct net *af_packet_net; 2664 void *af_packet_priv; 2665 struct list_head list; 2666 }; 2667 2668 struct offload_callbacks { 2669 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2670 netdev_features_t features); 2671 struct sk_buff *(*gro_receive)(struct list_head *head, 2672 struct sk_buff *skb); 2673 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2674 }; 2675 2676 struct packet_offload { 2677 __be16 type; /* This is really htons(ether_type). */ 2678 u16 priority; 2679 struct offload_callbacks callbacks; 2680 struct list_head list; 2681 }; 2682 2683 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2684 struct pcpu_sw_netstats { 2685 u64_stats_t rx_packets; 2686 u64_stats_t rx_bytes; 2687 u64_stats_t tx_packets; 2688 u64_stats_t tx_bytes; 2689 struct u64_stats_sync syncp; 2690 } __aligned(4 * sizeof(u64)); 2691 2692 struct pcpu_lstats { 2693 u64_stats_t packets; 2694 u64_stats_t bytes; 2695 struct u64_stats_sync syncp; 2696 } __aligned(2 * sizeof(u64)); 2697 2698 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2699 2700 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2701 { 2702 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2703 2704 u64_stats_update_begin(&tstats->syncp); 2705 u64_stats_add(&tstats->rx_bytes, len); 2706 u64_stats_inc(&tstats->rx_packets); 2707 u64_stats_update_end(&tstats->syncp); 2708 } 2709 2710 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2711 unsigned int packets, 2712 unsigned int len) 2713 { 2714 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2715 2716 u64_stats_update_begin(&tstats->syncp); 2717 u64_stats_add(&tstats->tx_bytes, len); 2718 u64_stats_add(&tstats->tx_packets, packets); 2719 u64_stats_update_end(&tstats->syncp); 2720 } 2721 2722 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2723 { 2724 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2725 2726 u64_stats_update_begin(&lstats->syncp); 2727 u64_stats_add(&lstats->bytes, len); 2728 u64_stats_inc(&lstats->packets); 2729 u64_stats_update_end(&lstats->syncp); 2730 } 2731 2732 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2733 ({ \ 2734 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2735 if (pcpu_stats) { \ 2736 int __cpu; \ 2737 for_each_possible_cpu(__cpu) { \ 2738 typeof(type) *stat; \ 2739 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2740 u64_stats_init(&stat->syncp); \ 2741 } \ 2742 } \ 2743 pcpu_stats; \ 2744 }) 2745 2746 #define netdev_alloc_pcpu_stats(type) \ 2747 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2748 2749 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2750 ({ \ 2751 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2752 if (pcpu_stats) { \ 2753 int __cpu; \ 2754 for_each_possible_cpu(__cpu) { \ 2755 typeof(type) *stat; \ 2756 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2757 u64_stats_init(&stat->syncp); \ 2758 } \ 2759 } \ 2760 pcpu_stats; \ 2761 }) 2762 2763 enum netdev_lag_tx_type { 2764 NETDEV_LAG_TX_TYPE_UNKNOWN, 2765 NETDEV_LAG_TX_TYPE_RANDOM, 2766 NETDEV_LAG_TX_TYPE_BROADCAST, 2767 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2768 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2769 NETDEV_LAG_TX_TYPE_HASH, 2770 }; 2771 2772 enum netdev_lag_hash { 2773 NETDEV_LAG_HASH_NONE, 2774 NETDEV_LAG_HASH_L2, 2775 NETDEV_LAG_HASH_L34, 2776 NETDEV_LAG_HASH_L23, 2777 NETDEV_LAG_HASH_E23, 2778 NETDEV_LAG_HASH_E34, 2779 NETDEV_LAG_HASH_VLAN_SRCMAC, 2780 NETDEV_LAG_HASH_UNKNOWN, 2781 }; 2782 2783 struct netdev_lag_upper_info { 2784 enum netdev_lag_tx_type tx_type; 2785 enum netdev_lag_hash hash_type; 2786 }; 2787 2788 struct netdev_lag_lower_state_info { 2789 u8 link_up : 1, 2790 tx_enabled : 1; 2791 }; 2792 2793 #include <linux/notifier.h> 2794 2795 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2796 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2797 * adding new types. 2798 */ 2799 enum netdev_cmd { 2800 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2801 NETDEV_DOWN, 2802 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2803 detected a hardware crash and restarted 2804 - we can use this eg to kick tcp sessions 2805 once done */ 2806 NETDEV_CHANGE, /* Notify device state change */ 2807 NETDEV_REGISTER, 2808 NETDEV_UNREGISTER, 2809 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2810 NETDEV_CHANGEADDR, /* notify after the address change */ 2811 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2812 NETDEV_GOING_DOWN, 2813 NETDEV_CHANGENAME, 2814 NETDEV_FEAT_CHANGE, 2815 NETDEV_BONDING_FAILOVER, 2816 NETDEV_PRE_UP, 2817 NETDEV_PRE_TYPE_CHANGE, 2818 NETDEV_POST_TYPE_CHANGE, 2819 NETDEV_POST_INIT, 2820 NETDEV_PRE_UNINIT, 2821 NETDEV_RELEASE, 2822 NETDEV_NOTIFY_PEERS, 2823 NETDEV_JOIN, 2824 NETDEV_CHANGEUPPER, 2825 NETDEV_RESEND_IGMP, 2826 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2827 NETDEV_CHANGEINFODATA, 2828 NETDEV_BONDING_INFO, 2829 NETDEV_PRECHANGEUPPER, 2830 NETDEV_CHANGELOWERSTATE, 2831 NETDEV_UDP_TUNNEL_PUSH_INFO, 2832 NETDEV_UDP_TUNNEL_DROP_INFO, 2833 NETDEV_CHANGE_TX_QUEUE_LEN, 2834 NETDEV_CVLAN_FILTER_PUSH_INFO, 2835 NETDEV_CVLAN_FILTER_DROP_INFO, 2836 NETDEV_SVLAN_FILTER_PUSH_INFO, 2837 NETDEV_SVLAN_FILTER_DROP_INFO, 2838 NETDEV_OFFLOAD_XSTATS_ENABLE, 2839 NETDEV_OFFLOAD_XSTATS_DISABLE, 2840 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 2841 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 2842 }; 2843 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2844 2845 int register_netdevice_notifier(struct notifier_block *nb); 2846 int unregister_netdevice_notifier(struct notifier_block *nb); 2847 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2848 int unregister_netdevice_notifier_net(struct net *net, 2849 struct notifier_block *nb); 2850 void move_netdevice_notifier_net(struct net *src_net, struct net *dst_net, 2851 struct notifier_block *nb); 2852 int register_netdevice_notifier_dev_net(struct net_device *dev, 2853 struct notifier_block *nb, 2854 struct netdev_net_notifier *nn); 2855 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2856 struct notifier_block *nb, 2857 struct netdev_net_notifier *nn); 2858 2859 struct netdev_notifier_info { 2860 struct net_device *dev; 2861 struct netlink_ext_ack *extack; 2862 }; 2863 2864 struct netdev_notifier_info_ext { 2865 struct netdev_notifier_info info; /* must be first */ 2866 union { 2867 u32 mtu; 2868 } ext; 2869 }; 2870 2871 struct netdev_notifier_change_info { 2872 struct netdev_notifier_info info; /* must be first */ 2873 unsigned int flags_changed; 2874 }; 2875 2876 struct netdev_notifier_changeupper_info { 2877 struct netdev_notifier_info info; /* must be first */ 2878 struct net_device *upper_dev; /* new upper dev */ 2879 bool master; /* is upper dev master */ 2880 bool linking; /* is the notification for link or unlink */ 2881 void *upper_info; /* upper dev info */ 2882 }; 2883 2884 struct netdev_notifier_changelowerstate_info { 2885 struct netdev_notifier_info info; /* must be first */ 2886 void *lower_state_info; /* is lower dev state */ 2887 }; 2888 2889 struct netdev_notifier_pre_changeaddr_info { 2890 struct netdev_notifier_info info; /* must be first */ 2891 const unsigned char *dev_addr; 2892 }; 2893 2894 enum netdev_offload_xstats_type { 2895 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 2896 }; 2897 2898 struct netdev_notifier_offload_xstats_info { 2899 struct netdev_notifier_info info; /* must be first */ 2900 enum netdev_offload_xstats_type type; 2901 2902 union { 2903 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 2904 struct netdev_notifier_offload_xstats_rd *report_delta; 2905 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 2906 struct netdev_notifier_offload_xstats_ru *report_used; 2907 }; 2908 }; 2909 2910 int netdev_offload_xstats_enable(struct net_device *dev, 2911 enum netdev_offload_xstats_type type, 2912 struct netlink_ext_ack *extack); 2913 int netdev_offload_xstats_disable(struct net_device *dev, 2914 enum netdev_offload_xstats_type type); 2915 bool netdev_offload_xstats_enabled(const struct net_device *dev, 2916 enum netdev_offload_xstats_type type); 2917 int netdev_offload_xstats_get(struct net_device *dev, 2918 enum netdev_offload_xstats_type type, 2919 struct rtnl_hw_stats64 *stats, bool *used, 2920 struct netlink_ext_ack *extack); 2921 void 2922 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 2923 const struct rtnl_hw_stats64 *stats); 2924 void 2925 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 2926 void netdev_offload_xstats_push_delta(struct net_device *dev, 2927 enum netdev_offload_xstats_type type, 2928 const struct rtnl_hw_stats64 *stats); 2929 2930 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2931 struct net_device *dev) 2932 { 2933 info->dev = dev; 2934 info->extack = NULL; 2935 } 2936 2937 static inline struct net_device * 2938 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2939 { 2940 return info->dev; 2941 } 2942 2943 static inline struct netlink_ext_ack * 2944 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2945 { 2946 return info->extack; 2947 } 2948 2949 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2950 2951 2952 extern rwlock_t dev_base_lock; /* Device list lock */ 2953 2954 #define for_each_netdev(net, d) \ 2955 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2956 #define for_each_netdev_reverse(net, d) \ 2957 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2958 #define for_each_netdev_rcu(net, d) \ 2959 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2960 #define for_each_netdev_safe(net, d, n) \ 2961 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2962 #define for_each_netdev_continue(net, d) \ 2963 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2964 #define for_each_netdev_continue_reverse(net, d) \ 2965 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2966 dev_list) 2967 #define for_each_netdev_continue_rcu(net, d) \ 2968 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2969 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2970 for_each_netdev_rcu(&init_net, slave) \ 2971 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2972 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2973 2974 static inline struct net_device *next_net_device(struct net_device *dev) 2975 { 2976 struct list_head *lh; 2977 struct net *net; 2978 2979 net = dev_net(dev); 2980 lh = dev->dev_list.next; 2981 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2982 } 2983 2984 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2985 { 2986 struct list_head *lh; 2987 struct net *net; 2988 2989 net = dev_net(dev); 2990 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2991 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2992 } 2993 2994 static inline struct net_device *first_net_device(struct net *net) 2995 { 2996 return list_empty(&net->dev_base_head) ? NULL : 2997 net_device_entry(net->dev_base_head.next); 2998 } 2999 3000 static inline struct net_device *first_net_device_rcu(struct net *net) 3001 { 3002 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 3003 3004 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3005 } 3006 3007 int netdev_boot_setup_check(struct net_device *dev); 3008 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 3009 const char *hwaddr); 3010 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 3011 void dev_add_pack(struct packet_type *pt); 3012 void dev_remove_pack(struct packet_type *pt); 3013 void __dev_remove_pack(struct packet_type *pt); 3014 void dev_add_offload(struct packet_offload *po); 3015 void dev_remove_offload(struct packet_offload *po); 3016 3017 int dev_get_iflink(const struct net_device *dev); 3018 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 3019 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 3020 struct net_device_path_stack *stack); 3021 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 3022 unsigned short mask); 3023 struct net_device *dev_get_by_name(struct net *net, const char *name); 3024 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 3025 struct net_device *__dev_get_by_name(struct net *net, const char *name); 3026 bool netdev_name_in_use(struct net *net, const char *name); 3027 int dev_alloc_name(struct net_device *dev, const char *name); 3028 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 3029 void dev_close(struct net_device *dev); 3030 void dev_close_many(struct list_head *head, bool unlink); 3031 void dev_disable_lro(struct net_device *dev); 3032 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 3033 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 3034 struct net_device *sb_dev); 3035 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 3036 struct net_device *sb_dev); 3037 3038 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3039 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3040 3041 static inline int dev_queue_xmit(struct sk_buff *skb) 3042 { 3043 return __dev_queue_xmit(skb, NULL); 3044 } 3045 3046 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3047 struct net_device *sb_dev) 3048 { 3049 return __dev_queue_xmit(skb, sb_dev); 3050 } 3051 3052 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3053 { 3054 int ret; 3055 3056 ret = __dev_direct_xmit(skb, queue_id); 3057 if (!dev_xmit_complete(ret)) 3058 kfree_skb(skb); 3059 return ret; 3060 } 3061 3062 int register_netdevice(struct net_device *dev); 3063 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3064 void unregister_netdevice_many(struct list_head *head); 3065 static inline void unregister_netdevice(struct net_device *dev) 3066 { 3067 unregister_netdevice_queue(dev, NULL); 3068 } 3069 3070 int netdev_refcnt_read(const struct net_device *dev); 3071 void free_netdev(struct net_device *dev); 3072 void netdev_freemem(struct net_device *dev); 3073 int init_dummy_netdev(struct net_device *dev); 3074 3075 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3076 struct sk_buff *skb, 3077 bool all_slaves); 3078 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3079 struct sock *sk); 3080 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3081 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3082 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3083 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 3084 int dev_restart(struct net_device *dev); 3085 3086 3087 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3088 unsigned short type, 3089 const void *daddr, const void *saddr, 3090 unsigned int len) 3091 { 3092 if (!dev->header_ops || !dev->header_ops->create) 3093 return 0; 3094 3095 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3096 } 3097 3098 static inline int dev_parse_header(const struct sk_buff *skb, 3099 unsigned char *haddr) 3100 { 3101 const struct net_device *dev = skb->dev; 3102 3103 if (!dev->header_ops || !dev->header_ops->parse) 3104 return 0; 3105 return dev->header_ops->parse(skb, haddr); 3106 } 3107 3108 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3109 { 3110 const struct net_device *dev = skb->dev; 3111 3112 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3113 return 0; 3114 return dev->header_ops->parse_protocol(skb); 3115 } 3116 3117 /* ll_header must have at least hard_header_len allocated */ 3118 static inline bool dev_validate_header(const struct net_device *dev, 3119 char *ll_header, int len) 3120 { 3121 if (likely(len >= dev->hard_header_len)) 3122 return true; 3123 if (len < dev->min_header_len) 3124 return false; 3125 3126 if (capable(CAP_SYS_RAWIO)) { 3127 memset(ll_header + len, 0, dev->hard_header_len - len); 3128 return true; 3129 } 3130 3131 if (dev->header_ops && dev->header_ops->validate) 3132 return dev->header_ops->validate(ll_header, len); 3133 3134 return false; 3135 } 3136 3137 static inline bool dev_has_header(const struct net_device *dev) 3138 { 3139 return dev->header_ops && dev->header_ops->create; 3140 } 3141 3142 /* 3143 * Incoming packets are placed on per-CPU queues 3144 */ 3145 struct softnet_data { 3146 struct list_head poll_list; 3147 struct sk_buff_head process_queue; 3148 3149 /* stats */ 3150 unsigned int processed; 3151 unsigned int time_squeeze; 3152 #ifdef CONFIG_RPS 3153 struct softnet_data *rps_ipi_list; 3154 #endif 3155 #ifdef CONFIG_NET_FLOW_LIMIT 3156 struct sd_flow_limit __rcu *flow_limit; 3157 #endif 3158 struct Qdisc *output_queue; 3159 struct Qdisc **output_queue_tailp; 3160 struct sk_buff *completion_queue; 3161 #ifdef CONFIG_XFRM_OFFLOAD 3162 struct sk_buff_head xfrm_backlog; 3163 #endif 3164 /* written and read only by owning cpu: */ 3165 struct { 3166 u16 recursion; 3167 u8 more; 3168 #ifdef CONFIG_NET_EGRESS 3169 u8 skip_txqueue; 3170 #endif 3171 } xmit; 3172 #ifdef CONFIG_RPS 3173 /* input_queue_head should be written by cpu owning this struct, 3174 * and only read by other cpus. Worth using a cache line. 3175 */ 3176 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3177 3178 /* Elements below can be accessed between CPUs for RPS/RFS */ 3179 call_single_data_t csd ____cacheline_aligned_in_smp; 3180 struct softnet_data *rps_ipi_next; 3181 unsigned int cpu; 3182 unsigned int input_queue_tail; 3183 #endif 3184 unsigned int received_rps; 3185 unsigned int dropped; 3186 struct sk_buff_head input_pkt_queue; 3187 struct napi_struct backlog; 3188 3189 /* Another possibly contended cache line */ 3190 spinlock_t defer_lock ____cacheline_aligned_in_smp; 3191 int defer_count; 3192 int defer_ipi_scheduled; 3193 struct sk_buff *defer_list; 3194 call_single_data_t defer_csd; 3195 }; 3196 3197 static inline void input_queue_head_incr(struct softnet_data *sd) 3198 { 3199 #ifdef CONFIG_RPS 3200 sd->input_queue_head++; 3201 #endif 3202 } 3203 3204 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3205 unsigned int *qtail) 3206 { 3207 #ifdef CONFIG_RPS 3208 *qtail = ++sd->input_queue_tail; 3209 #endif 3210 } 3211 3212 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3213 3214 static inline int dev_recursion_level(void) 3215 { 3216 return this_cpu_read(softnet_data.xmit.recursion); 3217 } 3218 3219 #define XMIT_RECURSION_LIMIT 8 3220 static inline bool dev_xmit_recursion(void) 3221 { 3222 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3223 XMIT_RECURSION_LIMIT); 3224 } 3225 3226 static inline void dev_xmit_recursion_inc(void) 3227 { 3228 __this_cpu_inc(softnet_data.xmit.recursion); 3229 } 3230 3231 static inline void dev_xmit_recursion_dec(void) 3232 { 3233 __this_cpu_dec(softnet_data.xmit.recursion); 3234 } 3235 3236 void __netif_schedule(struct Qdisc *q); 3237 void netif_schedule_queue(struct netdev_queue *txq); 3238 3239 static inline void netif_tx_schedule_all(struct net_device *dev) 3240 { 3241 unsigned int i; 3242 3243 for (i = 0; i < dev->num_tx_queues; i++) 3244 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3245 } 3246 3247 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3248 { 3249 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3250 } 3251 3252 /** 3253 * netif_start_queue - allow transmit 3254 * @dev: network device 3255 * 3256 * Allow upper layers to call the device hard_start_xmit routine. 3257 */ 3258 static inline void netif_start_queue(struct net_device *dev) 3259 { 3260 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3261 } 3262 3263 static inline void netif_tx_start_all_queues(struct net_device *dev) 3264 { 3265 unsigned int i; 3266 3267 for (i = 0; i < dev->num_tx_queues; i++) { 3268 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3269 netif_tx_start_queue(txq); 3270 } 3271 } 3272 3273 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3274 3275 /** 3276 * netif_wake_queue - restart transmit 3277 * @dev: network device 3278 * 3279 * Allow upper layers to call the device hard_start_xmit routine. 3280 * Used for flow control when transmit resources are available. 3281 */ 3282 static inline void netif_wake_queue(struct net_device *dev) 3283 { 3284 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3285 } 3286 3287 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3288 { 3289 unsigned int i; 3290 3291 for (i = 0; i < dev->num_tx_queues; i++) { 3292 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3293 netif_tx_wake_queue(txq); 3294 } 3295 } 3296 3297 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3298 { 3299 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3300 } 3301 3302 /** 3303 * netif_stop_queue - stop transmitted packets 3304 * @dev: network device 3305 * 3306 * Stop upper layers calling the device hard_start_xmit routine. 3307 * Used for flow control when transmit resources are unavailable. 3308 */ 3309 static inline void netif_stop_queue(struct net_device *dev) 3310 { 3311 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3312 } 3313 3314 void netif_tx_stop_all_queues(struct net_device *dev); 3315 3316 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3317 { 3318 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3319 } 3320 3321 /** 3322 * netif_queue_stopped - test if transmit queue is flowblocked 3323 * @dev: network device 3324 * 3325 * Test if transmit queue on device is currently unable to send. 3326 */ 3327 static inline bool netif_queue_stopped(const struct net_device *dev) 3328 { 3329 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3330 } 3331 3332 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3333 { 3334 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3335 } 3336 3337 static inline bool 3338 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3339 { 3340 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3341 } 3342 3343 static inline bool 3344 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3345 { 3346 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3347 } 3348 3349 /** 3350 * netdev_queue_set_dql_min_limit - set dql minimum limit 3351 * @dev_queue: pointer to transmit queue 3352 * @min_limit: dql minimum limit 3353 * 3354 * Forces xmit_more() to return true until the minimum threshold 3355 * defined by @min_limit is reached (or until the tx queue is 3356 * empty). Warning: to be use with care, misuse will impact the 3357 * latency. 3358 */ 3359 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3360 unsigned int min_limit) 3361 { 3362 #ifdef CONFIG_BQL 3363 dev_queue->dql.min_limit = min_limit; 3364 #endif 3365 } 3366 3367 /** 3368 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3369 * @dev_queue: pointer to transmit queue 3370 * 3371 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3372 * to give appropriate hint to the CPU. 3373 */ 3374 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3375 { 3376 #ifdef CONFIG_BQL 3377 prefetchw(&dev_queue->dql.num_queued); 3378 #endif 3379 } 3380 3381 /** 3382 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3383 * @dev_queue: pointer to transmit queue 3384 * 3385 * BQL enabled drivers might use this helper in their TX completion path, 3386 * to give appropriate hint to the CPU. 3387 */ 3388 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3389 { 3390 #ifdef CONFIG_BQL 3391 prefetchw(&dev_queue->dql.limit); 3392 #endif 3393 } 3394 3395 /** 3396 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3397 * @dev_queue: network device queue 3398 * @bytes: number of bytes queued to the device queue 3399 * 3400 * Report the number of bytes queued for sending/completion to the network 3401 * device hardware queue. @bytes should be a good approximation and should 3402 * exactly match netdev_completed_queue() @bytes. 3403 * This is typically called once per packet, from ndo_start_xmit(). 3404 */ 3405 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3406 unsigned int bytes) 3407 { 3408 #ifdef CONFIG_BQL 3409 dql_queued(&dev_queue->dql, bytes); 3410 3411 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3412 return; 3413 3414 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3415 3416 /* 3417 * The XOFF flag must be set before checking the dql_avail below, 3418 * because in netdev_tx_completed_queue we update the dql_completed 3419 * before checking the XOFF flag. 3420 */ 3421 smp_mb(); 3422 3423 /* check again in case another CPU has just made room avail */ 3424 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3425 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3426 #endif 3427 } 3428 3429 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3430 * that they should not test BQL status themselves. 3431 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3432 * skb of a batch. 3433 * Returns true if the doorbell must be used to kick the NIC. 3434 */ 3435 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3436 unsigned int bytes, 3437 bool xmit_more) 3438 { 3439 if (xmit_more) { 3440 #ifdef CONFIG_BQL 3441 dql_queued(&dev_queue->dql, bytes); 3442 #endif 3443 return netif_tx_queue_stopped(dev_queue); 3444 } 3445 netdev_tx_sent_queue(dev_queue, bytes); 3446 return true; 3447 } 3448 3449 /** 3450 * netdev_sent_queue - report the number of bytes queued to hardware 3451 * @dev: network device 3452 * @bytes: number of bytes queued to the hardware device queue 3453 * 3454 * Report the number of bytes queued for sending/completion to the network 3455 * device hardware queue#0. @bytes should be a good approximation and should 3456 * exactly match netdev_completed_queue() @bytes. 3457 * This is typically called once per packet, from ndo_start_xmit(). 3458 */ 3459 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3460 { 3461 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3462 } 3463 3464 static inline bool __netdev_sent_queue(struct net_device *dev, 3465 unsigned int bytes, 3466 bool xmit_more) 3467 { 3468 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3469 xmit_more); 3470 } 3471 3472 /** 3473 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3474 * @dev_queue: network device queue 3475 * @pkts: number of packets (currently ignored) 3476 * @bytes: number of bytes dequeued from the device queue 3477 * 3478 * Must be called at most once per TX completion round (and not per 3479 * individual packet), so that BQL can adjust its limits appropriately. 3480 */ 3481 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3482 unsigned int pkts, unsigned int bytes) 3483 { 3484 #ifdef CONFIG_BQL 3485 if (unlikely(!bytes)) 3486 return; 3487 3488 dql_completed(&dev_queue->dql, bytes); 3489 3490 /* 3491 * Without the memory barrier there is a small possiblity that 3492 * netdev_tx_sent_queue will miss the update and cause the queue to 3493 * be stopped forever 3494 */ 3495 smp_mb(); 3496 3497 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3498 return; 3499 3500 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3501 netif_schedule_queue(dev_queue); 3502 #endif 3503 } 3504 3505 /** 3506 * netdev_completed_queue - report bytes and packets completed by device 3507 * @dev: network device 3508 * @pkts: actual number of packets sent over the medium 3509 * @bytes: actual number of bytes sent over the medium 3510 * 3511 * Report the number of bytes and packets transmitted by the network device 3512 * hardware queue over the physical medium, @bytes must exactly match the 3513 * @bytes amount passed to netdev_sent_queue() 3514 */ 3515 static inline void netdev_completed_queue(struct net_device *dev, 3516 unsigned int pkts, unsigned int bytes) 3517 { 3518 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3519 } 3520 3521 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3522 { 3523 #ifdef CONFIG_BQL 3524 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3525 dql_reset(&q->dql); 3526 #endif 3527 } 3528 3529 /** 3530 * netdev_reset_queue - reset the packets and bytes count of a network device 3531 * @dev_queue: network device 3532 * 3533 * Reset the bytes and packet count of a network device and clear the 3534 * software flow control OFF bit for this network device 3535 */ 3536 static inline void netdev_reset_queue(struct net_device *dev_queue) 3537 { 3538 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3539 } 3540 3541 /** 3542 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3543 * @dev: network device 3544 * @queue_index: given tx queue index 3545 * 3546 * Returns 0 if given tx queue index >= number of device tx queues, 3547 * otherwise returns the originally passed tx queue index. 3548 */ 3549 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3550 { 3551 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3552 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3553 dev->name, queue_index, 3554 dev->real_num_tx_queues); 3555 return 0; 3556 } 3557 3558 return queue_index; 3559 } 3560 3561 /** 3562 * netif_running - test if up 3563 * @dev: network device 3564 * 3565 * Test if the device has been brought up. 3566 */ 3567 static inline bool netif_running(const struct net_device *dev) 3568 { 3569 return test_bit(__LINK_STATE_START, &dev->state); 3570 } 3571 3572 /* 3573 * Routines to manage the subqueues on a device. We only need start, 3574 * stop, and a check if it's stopped. All other device management is 3575 * done at the overall netdevice level. 3576 * Also test the device if we're multiqueue. 3577 */ 3578 3579 /** 3580 * netif_start_subqueue - allow sending packets on subqueue 3581 * @dev: network device 3582 * @queue_index: sub queue index 3583 * 3584 * Start individual transmit queue of a device with multiple transmit queues. 3585 */ 3586 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3587 { 3588 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3589 3590 netif_tx_start_queue(txq); 3591 } 3592 3593 /** 3594 * netif_stop_subqueue - stop sending packets on subqueue 3595 * @dev: network device 3596 * @queue_index: sub queue index 3597 * 3598 * Stop individual transmit queue of a device with multiple transmit queues. 3599 */ 3600 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3601 { 3602 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3603 netif_tx_stop_queue(txq); 3604 } 3605 3606 /** 3607 * __netif_subqueue_stopped - test status of subqueue 3608 * @dev: network device 3609 * @queue_index: sub queue index 3610 * 3611 * Check individual transmit queue of a device with multiple transmit queues. 3612 */ 3613 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3614 u16 queue_index) 3615 { 3616 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3617 3618 return netif_tx_queue_stopped(txq); 3619 } 3620 3621 /** 3622 * netif_subqueue_stopped - test status of subqueue 3623 * @dev: network device 3624 * @skb: sub queue buffer pointer 3625 * 3626 * Check individual transmit queue of a device with multiple transmit queues. 3627 */ 3628 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3629 struct sk_buff *skb) 3630 { 3631 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3632 } 3633 3634 /** 3635 * netif_wake_subqueue - allow sending packets on subqueue 3636 * @dev: network device 3637 * @queue_index: sub queue index 3638 * 3639 * Resume individual transmit queue of a device with multiple transmit queues. 3640 */ 3641 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3642 { 3643 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3644 3645 netif_tx_wake_queue(txq); 3646 } 3647 3648 #ifdef CONFIG_XPS 3649 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3650 u16 index); 3651 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3652 u16 index, enum xps_map_type type); 3653 3654 /** 3655 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3656 * @j: CPU/Rx queue index 3657 * @mask: bitmask of all cpus/rx queues 3658 * @nr_bits: number of bits in the bitmask 3659 * 3660 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3661 */ 3662 static inline bool netif_attr_test_mask(unsigned long j, 3663 const unsigned long *mask, 3664 unsigned int nr_bits) 3665 { 3666 cpu_max_bits_warn(j, nr_bits); 3667 return test_bit(j, mask); 3668 } 3669 3670 /** 3671 * netif_attr_test_online - Test for online CPU/Rx queue 3672 * @j: CPU/Rx queue index 3673 * @online_mask: bitmask for CPUs/Rx queues that are online 3674 * @nr_bits: number of bits in the bitmask 3675 * 3676 * Returns true if a CPU/Rx queue is online. 3677 */ 3678 static inline bool netif_attr_test_online(unsigned long j, 3679 const unsigned long *online_mask, 3680 unsigned int nr_bits) 3681 { 3682 cpu_max_bits_warn(j, nr_bits); 3683 3684 if (online_mask) 3685 return test_bit(j, online_mask); 3686 3687 return (j < nr_bits); 3688 } 3689 3690 /** 3691 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3692 * @n: CPU/Rx queue index 3693 * @srcp: the cpumask/Rx queue mask pointer 3694 * @nr_bits: number of bits in the bitmask 3695 * 3696 * Returns >= nr_bits if no further CPUs/Rx queues set. 3697 */ 3698 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3699 unsigned int nr_bits) 3700 { 3701 /* -1 is a legal arg here. */ 3702 if (n != -1) 3703 cpu_max_bits_warn(n, nr_bits); 3704 3705 if (srcp) 3706 return find_next_bit(srcp, nr_bits, n + 1); 3707 3708 return n + 1; 3709 } 3710 3711 /** 3712 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3713 * @n: CPU/Rx queue index 3714 * @src1p: the first CPUs/Rx queues mask pointer 3715 * @src2p: the second CPUs/Rx queues mask pointer 3716 * @nr_bits: number of bits in the bitmask 3717 * 3718 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3719 */ 3720 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3721 const unsigned long *src2p, 3722 unsigned int nr_bits) 3723 { 3724 /* -1 is a legal arg here. */ 3725 if (n != -1) 3726 cpu_max_bits_warn(n, nr_bits); 3727 3728 if (src1p && src2p) 3729 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3730 else if (src1p) 3731 return find_next_bit(src1p, nr_bits, n + 1); 3732 else if (src2p) 3733 return find_next_bit(src2p, nr_bits, n + 1); 3734 3735 return n + 1; 3736 } 3737 #else 3738 static inline int netif_set_xps_queue(struct net_device *dev, 3739 const struct cpumask *mask, 3740 u16 index) 3741 { 3742 return 0; 3743 } 3744 3745 static inline int __netif_set_xps_queue(struct net_device *dev, 3746 const unsigned long *mask, 3747 u16 index, enum xps_map_type type) 3748 { 3749 return 0; 3750 } 3751 #endif 3752 3753 /** 3754 * netif_is_multiqueue - test if device has multiple transmit queues 3755 * @dev: network device 3756 * 3757 * Check if device has multiple transmit queues 3758 */ 3759 static inline bool netif_is_multiqueue(const struct net_device *dev) 3760 { 3761 return dev->num_tx_queues > 1; 3762 } 3763 3764 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3765 3766 #ifdef CONFIG_SYSFS 3767 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3768 #else 3769 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3770 unsigned int rxqs) 3771 { 3772 dev->real_num_rx_queues = rxqs; 3773 return 0; 3774 } 3775 #endif 3776 int netif_set_real_num_queues(struct net_device *dev, 3777 unsigned int txq, unsigned int rxq); 3778 3779 static inline struct netdev_rx_queue * 3780 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3781 { 3782 return dev->_rx + rxq; 3783 } 3784 3785 #ifdef CONFIG_SYSFS 3786 static inline unsigned int get_netdev_rx_queue_index( 3787 struct netdev_rx_queue *queue) 3788 { 3789 struct net_device *dev = queue->dev; 3790 int index = queue - dev->_rx; 3791 3792 BUG_ON(index >= dev->num_rx_queues); 3793 return index; 3794 } 3795 #endif 3796 3797 int netif_get_num_default_rss_queues(void); 3798 3799 enum skb_free_reason { 3800 SKB_REASON_CONSUMED, 3801 SKB_REASON_DROPPED, 3802 }; 3803 3804 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3805 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3806 3807 /* 3808 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3809 * interrupt context or with hardware interrupts being disabled. 3810 * (in_hardirq() || irqs_disabled()) 3811 * 3812 * We provide four helpers that can be used in following contexts : 3813 * 3814 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3815 * replacing kfree_skb(skb) 3816 * 3817 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3818 * Typically used in place of consume_skb(skb) in TX completion path 3819 * 3820 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3821 * replacing kfree_skb(skb) 3822 * 3823 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3824 * and consumed a packet. Used in place of consume_skb(skb) 3825 */ 3826 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3827 { 3828 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3829 } 3830 3831 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3832 { 3833 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3834 } 3835 3836 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3837 { 3838 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3839 } 3840 3841 static inline void dev_consume_skb_any(struct sk_buff *skb) 3842 { 3843 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3844 } 3845 3846 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3847 struct bpf_prog *xdp_prog); 3848 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3849 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3850 int netif_rx(struct sk_buff *skb); 3851 int __netif_rx(struct sk_buff *skb); 3852 3853 int netif_receive_skb(struct sk_buff *skb); 3854 int netif_receive_skb_core(struct sk_buff *skb); 3855 void netif_receive_skb_list_internal(struct list_head *head); 3856 void netif_receive_skb_list(struct list_head *head); 3857 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3858 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3859 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3860 void napi_get_frags_check(struct napi_struct *napi); 3861 gro_result_t napi_gro_frags(struct napi_struct *napi); 3862 struct packet_offload *gro_find_receive_by_type(__be16 type); 3863 struct packet_offload *gro_find_complete_by_type(__be16 type); 3864 3865 static inline void napi_free_frags(struct napi_struct *napi) 3866 { 3867 kfree_skb(napi->skb); 3868 napi->skb = NULL; 3869 } 3870 3871 bool netdev_is_rx_handler_busy(struct net_device *dev); 3872 int netdev_rx_handler_register(struct net_device *dev, 3873 rx_handler_func_t *rx_handler, 3874 void *rx_handler_data); 3875 void netdev_rx_handler_unregister(struct net_device *dev); 3876 3877 bool dev_valid_name(const char *name); 3878 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 3879 { 3880 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 3881 } 3882 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 3883 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 3884 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3885 void __user *data, bool *need_copyout); 3886 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 3887 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 3888 unsigned int dev_get_flags(const struct net_device *); 3889 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3890 struct netlink_ext_ack *extack); 3891 int dev_change_flags(struct net_device *dev, unsigned int flags, 3892 struct netlink_ext_ack *extack); 3893 int dev_set_alias(struct net_device *, const char *, size_t); 3894 int dev_get_alias(const struct net_device *, char *, size_t); 3895 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 3896 const char *pat, int new_ifindex); 3897 static inline 3898 int dev_change_net_namespace(struct net_device *dev, struct net *net, 3899 const char *pat) 3900 { 3901 return __dev_change_net_namespace(dev, net, pat, 0); 3902 } 3903 int __dev_set_mtu(struct net_device *, int); 3904 int dev_set_mtu(struct net_device *, int); 3905 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3906 struct netlink_ext_ack *extack); 3907 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3908 struct netlink_ext_ack *extack); 3909 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 3910 struct netlink_ext_ack *extack); 3911 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 3912 int dev_get_port_parent_id(struct net_device *dev, 3913 struct netdev_phys_item_id *ppid, bool recurse); 3914 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 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 struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev); 3956 3957 static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev) 3958 { 3959 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */ 3960 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats); 3961 3962 if (likely(p)) 3963 return p; 3964 3965 return netdev_core_stats_alloc(dev); 3966 } 3967 3968 #define DEV_CORE_STATS_INC(FIELD) \ 3969 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 3970 { \ 3971 struct net_device_core_stats __percpu *p; \ 3972 \ 3973 p = dev_core_stats(dev); \ 3974 if (p) \ 3975 this_cpu_inc(p->FIELD); \ 3976 } 3977 DEV_CORE_STATS_INC(rx_dropped) 3978 DEV_CORE_STATS_INC(tx_dropped) 3979 DEV_CORE_STATS_INC(rx_nohandler) 3980 DEV_CORE_STATS_INC(rx_otherhost_dropped) 3981 3982 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3983 struct sk_buff *skb, 3984 const bool check_mtu) 3985 { 3986 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3987 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 3988 dev_core_stats_rx_dropped_inc(dev); 3989 kfree_skb(skb); 3990 return NET_RX_DROP; 3991 } 3992 3993 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 3994 skb->priority = 0; 3995 return 0; 3996 } 3997 3998 bool dev_nit_active(struct net_device *dev); 3999 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 4000 4001 static inline void __dev_put(struct net_device *dev) 4002 { 4003 if (dev) { 4004 #ifdef CONFIG_PCPU_DEV_REFCNT 4005 this_cpu_dec(*dev->pcpu_refcnt); 4006 #else 4007 refcount_dec(&dev->dev_refcnt); 4008 #endif 4009 } 4010 } 4011 4012 static inline void __dev_hold(struct net_device *dev) 4013 { 4014 if (dev) { 4015 #ifdef CONFIG_PCPU_DEV_REFCNT 4016 this_cpu_inc(*dev->pcpu_refcnt); 4017 #else 4018 refcount_inc(&dev->dev_refcnt); 4019 #endif 4020 } 4021 } 4022 4023 static inline void __netdev_tracker_alloc(struct net_device *dev, 4024 netdevice_tracker *tracker, 4025 gfp_t gfp) 4026 { 4027 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4028 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 4029 #endif 4030 } 4031 4032 /* netdev_tracker_alloc() can upgrade a prior untracked reference 4033 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4034 */ 4035 static inline void netdev_tracker_alloc(struct net_device *dev, 4036 netdevice_tracker *tracker, gfp_t gfp) 4037 { 4038 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4039 refcount_dec(&dev->refcnt_tracker.no_tracker); 4040 __netdev_tracker_alloc(dev, tracker, gfp); 4041 #endif 4042 } 4043 4044 static inline void netdev_tracker_free(struct net_device *dev, 4045 netdevice_tracker *tracker) 4046 { 4047 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4048 ref_tracker_free(&dev->refcnt_tracker, tracker); 4049 #endif 4050 } 4051 4052 static inline void netdev_hold(struct net_device *dev, 4053 netdevice_tracker *tracker, gfp_t gfp) 4054 { 4055 if (dev) { 4056 __dev_hold(dev); 4057 __netdev_tracker_alloc(dev, tracker, gfp); 4058 } 4059 } 4060 4061 static inline void netdev_put(struct net_device *dev, 4062 netdevice_tracker *tracker) 4063 { 4064 if (dev) { 4065 netdev_tracker_free(dev, tracker); 4066 __dev_put(dev); 4067 } 4068 } 4069 4070 /** 4071 * dev_hold - get reference to device 4072 * @dev: network device 4073 * 4074 * Hold reference to device to keep it from being freed. 4075 * Try using netdev_hold() instead. 4076 */ 4077 static inline void dev_hold(struct net_device *dev) 4078 { 4079 netdev_hold(dev, NULL, GFP_ATOMIC); 4080 } 4081 4082 /** 4083 * dev_put - release reference to device 4084 * @dev: network device 4085 * 4086 * Release reference to device to allow it to be freed. 4087 * Try using netdev_put() instead. 4088 */ 4089 static inline void dev_put(struct net_device *dev) 4090 { 4091 netdev_put(dev, NULL); 4092 } 4093 4094 static inline void netdev_ref_replace(struct net_device *odev, 4095 struct net_device *ndev, 4096 netdevice_tracker *tracker, 4097 gfp_t gfp) 4098 { 4099 if (odev) 4100 netdev_tracker_free(odev, tracker); 4101 4102 __dev_hold(ndev); 4103 __dev_put(odev); 4104 4105 if (ndev) 4106 __netdev_tracker_alloc(ndev, tracker, gfp); 4107 } 4108 4109 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4110 * and _off may be called from IRQ context, but it is caller 4111 * who is responsible for serialization of these calls. 4112 * 4113 * The name carrier is inappropriate, these functions should really be 4114 * called netif_lowerlayer_*() because they represent the state of any 4115 * kind of lower layer not just hardware media. 4116 */ 4117 void linkwatch_fire_event(struct net_device *dev); 4118 4119 /** 4120 * netif_carrier_ok - test if carrier present 4121 * @dev: network device 4122 * 4123 * Check if carrier is present on device 4124 */ 4125 static inline bool netif_carrier_ok(const struct net_device *dev) 4126 { 4127 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4128 } 4129 4130 unsigned long dev_trans_start(struct net_device *dev); 4131 4132 void __netdev_watchdog_up(struct net_device *dev); 4133 4134 void netif_carrier_on(struct net_device *dev); 4135 void netif_carrier_off(struct net_device *dev); 4136 void netif_carrier_event(struct net_device *dev); 4137 4138 /** 4139 * netif_dormant_on - mark device as dormant. 4140 * @dev: network device 4141 * 4142 * Mark device as dormant (as per RFC2863). 4143 * 4144 * The dormant state indicates that the relevant interface is not 4145 * actually in a condition to pass packets (i.e., it is not 'up') but is 4146 * in a "pending" state, waiting for some external event. For "on- 4147 * demand" interfaces, this new state identifies the situation where the 4148 * interface is waiting for events to place it in the up state. 4149 */ 4150 static inline void netif_dormant_on(struct net_device *dev) 4151 { 4152 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4153 linkwatch_fire_event(dev); 4154 } 4155 4156 /** 4157 * netif_dormant_off - set device as not dormant. 4158 * @dev: network device 4159 * 4160 * Device is not in dormant state. 4161 */ 4162 static inline void netif_dormant_off(struct net_device *dev) 4163 { 4164 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4165 linkwatch_fire_event(dev); 4166 } 4167 4168 /** 4169 * netif_dormant - test if device is dormant 4170 * @dev: network device 4171 * 4172 * Check if device is dormant. 4173 */ 4174 static inline bool netif_dormant(const struct net_device *dev) 4175 { 4176 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4177 } 4178 4179 4180 /** 4181 * netif_testing_on - mark device as under test. 4182 * @dev: network device 4183 * 4184 * Mark device as under test (as per RFC2863). 4185 * 4186 * The testing state indicates that some test(s) must be performed on 4187 * the interface. After completion, of the test, the interface state 4188 * will change to up, dormant, or down, as appropriate. 4189 */ 4190 static inline void netif_testing_on(struct net_device *dev) 4191 { 4192 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4193 linkwatch_fire_event(dev); 4194 } 4195 4196 /** 4197 * netif_testing_off - set device as not under test. 4198 * @dev: network device 4199 * 4200 * Device is not in testing state. 4201 */ 4202 static inline void netif_testing_off(struct net_device *dev) 4203 { 4204 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4205 linkwatch_fire_event(dev); 4206 } 4207 4208 /** 4209 * netif_testing - test if device is under test 4210 * @dev: network device 4211 * 4212 * Check if device is under test 4213 */ 4214 static inline bool netif_testing(const struct net_device *dev) 4215 { 4216 return test_bit(__LINK_STATE_TESTING, &dev->state); 4217 } 4218 4219 4220 /** 4221 * netif_oper_up - test if device is operational 4222 * @dev: network device 4223 * 4224 * Check if carrier is operational 4225 */ 4226 static inline bool netif_oper_up(const struct net_device *dev) 4227 { 4228 return (dev->operstate == IF_OPER_UP || 4229 dev->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 /* Support for loadable net-drivers */ 4502 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4503 unsigned char name_assign_type, 4504 void (*setup)(struct net_device *), 4505 unsigned int txqs, unsigned int rxqs); 4506 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4507 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4508 4509 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4510 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4511 count) 4512 4513 int register_netdev(struct net_device *dev); 4514 void unregister_netdev(struct net_device *dev); 4515 4516 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4517 4518 /* General hardware address lists handling functions */ 4519 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4520 struct netdev_hw_addr_list *from_list, int addr_len); 4521 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4522 struct netdev_hw_addr_list *from_list, int addr_len); 4523 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4524 struct net_device *dev, 4525 int (*sync)(struct net_device *, const unsigned char *), 4526 int (*unsync)(struct net_device *, 4527 const unsigned char *)); 4528 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4529 struct net_device *dev, 4530 int (*sync)(struct net_device *, 4531 const unsigned char *, int), 4532 int (*unsync)(struct net_device *, 4533 const unsigned char *, int)); 4534 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4535 struct net_device *dev, 4536 int (*unsync)(struct net_device *, 4537 const unsigned char *, int)); 4538 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4539 struct net_device *dev, 4540 int (*unsync)(struct net_device *, 4541 const unsigned char *)); 4542 void __hw_addr_init(struct netdev_hw_addr_list *list); 4543 4544 /* Functions used for device addresses handling */ 4545 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4546 const void *addr, size_t len); 4547 4548 static inline void 4549 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4550 { 4551 dev_addr_mod(dev, 0, addr, len); 4552 } 4553 4554 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4555 { 4556 __dev_addr_set(dev, addr, dev->addr_len); 4557 } 4558 4559 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4560 unsigned char addr_type); 4561 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4562 unsigned char addr_type); 4563 4564 /* Functions used for unicast addresses handling */ 4565 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4566 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4567 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4568 int dev_uc_sync(struct net_device *to, struct net_device *from); 4569 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4570 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4571 void dev_uc_flush(struct net_device *dev); 4572 void dev_uc_init(struct net_device *dev); 4573 4574 /** 4575 * __dev_uc_sync - Synchonize device's unicast list 4576 * @dev: device to sync 4577 * @sync: function to call if address should be added 4578 * @unsync: function to call if address should be removed 4579 * 4580 * Add newly added addresses to the interface, and release 4581 * addresses that have been deleted. 4582 */ 4583 static inline int __dev_uc_sync(struct net_device *dev, 4584 int (*sync)(struct net_device *, 4585 const unsigned char *), 4586 int (*unsync)(struct net_device *, 4587 const unsigned char *)) 4588 { 4589 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4590 } 4591 4592 /** 4593 * __dev_uc_unsync - Remove synchronized addresses from device 4594 * @dev: device to sync 4595 * @unsync: function to call if address should be removed 4596 * 4597 * Remove all addresses that were added to the device by dev_uc_sync(). 4598 */ 4599 static inline void __dev_uc_unsync(struct net_device *dev, 4600 int (*unsync)(struct net_device *, 4601 const unsigned char *)) 4602 { 4603 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4604 } 4605 4606 /* Functions used for multicast addresses handling */ 4607 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4608 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4609 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4610 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4611 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4612 int dev_mc_sync(struct net_device *to, struct net_device *from); 4613 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4614 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4615 void dev_mc_flush(struct net_device *dev); 4616 void dev_mc_init(struct net_device *dev); 4617 4618 /** 4619 * __dev_mc_sync - Synchonize device's multicast list 4620 * @dev: device to sync 4621 * @sync: function to call if address should be added 4622 * @unsync: function to call if address should be removed 4623 * 4624 * Add newly added addresses to the interface, and release 4625 * addresses that have been deleted. 4626 */ 4627 static inline int __dev_mc_sync(struct net_device *dev, 4628 int (*sync)(struct net_device *, 4629 const unsigned char *), 4630 int (*unsync)(struct net_device *, 4631 const unsigned char *)) 4632 { 4633 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4634 } 4635 4636 /** 4637 * __dev_mc_unsync - Remove synchronized addresses from device 4638 * @dev: device to sync 4639 * @unsync: function to call if address should be removed 4640 * 4641 * Remove all addresses that were added to the device by dev_mc_sync(). 4642 */ 4643 static inline void __dev_mc_unsync(struct net_device *dev, 4644 int (*unsync)(struct net_device *, 4645 const unsigned char *)) 4646 { 4647 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4648 } 4649 4650 /* Functions used for secondary unicast and multicast support */ 4651 void dev_set_rx_mode(struct net_device *dev); 4652 int dev_set_promiscuity(struct net_device *dev, int inc); 4653 int dev_set_allmulti(struct net_device *dev, int inc); 4654 void netdev_state_change(struct net_device *dev); 4655 void __netdev_notify_peers(struct net_device *dev); 4656 void netdev_notify_peers(struct net_device *dev); 4657 void netdev_features_change(struct net_device *dev); 4658 /* Load a device via the kmod */ 4659 void dev_load(struct net *net, const char *name); 4660 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4661 struct rtnl_link_stats64 *storage); 4662 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4663 const struct net_device_stats *netdev_stats); 4664 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4665 const struct pcpu_sw_netstats __percpu *netstats); 4666 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4667 4668 extern int netdev_max_backlog; 4669 extern int dev_rx_weight; 4670 extern int dev_tx_weight; 4671 extern int gro_normal_batch; 4672 4673 enum { 4674 NESTED_SYNC_IMM_BIT, 4675 NESTED_SYNC_TODO_BIT, 4676 }; 4677 4678 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4679 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4680 4681 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4682 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4683 4684 struct netdev_nested_priv { 4685 unsigned char flags; 4686 void *data; 4687 }; 4688 4689 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4690 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4691 struct list_head **iter); 4692 4693 /* iterate through upper list, must be called under RCU read lock */ 4694 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4695 for (iter = &(dev)->adj_list.upper, \ 4696 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4697 updev; \ 4698 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4699 4700 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4701 int (*fn)(struct net_device *upper_dev, 4702 struct netdev_nested_priv *priv), 4703 struct netdev_nested_priv *priv); 4704 4705 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4706 struct net_device *upper_dev); 4707 4708 bool netdev_has_any_upper_dev(struct net_device *dev); 4709 4710 void *netdev_lower_get_next_private(struct net_device *dev, 4711 struct list_head **iter); 4712 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4713 struct list_head **iter); 4714 4715 #define netdev_for_each_lower_private(dev, priv, iter) \ 4716 for (iter = (dev)->adj_list.lower.next, \ 4717 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4718 priv; \ 4719 priv = netdev_lower_get_next_private(dev, &(iter))) 4720 4721 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4722 for (iter = &(dev)->adj_list.lower, \ 4723 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4724 priv; \ 4725 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4726 4727 void *netdev_lower_get_next(struct net_device *dev, 4728 struct list_head **iter); 4729 4730 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4731 for (iter = (dev)->adj_list.lower.next, \ 4732 ldev = netdev_lower_get_next(dev, &(iter)); \ 4733 ldev; \ 4734 ldev = netdev_lower_get_next(dev, &(iter))) 4735 4736 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4737 struct list_head **iter); 4738 int netdev_walk_all_lower_dev(struct net_device *dev, 4739 int (*fn)(struct net_device *lower_dev, 4740 struct netdev_nested_priv *priv), 4741 struct netdev_nested_priv *priv); 4742 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4743 int (*fn)(struct net_device *lower_dev, 4744 struct netdev_nested_priv *priv), 4745 struct netdev_nested_priv *priv); 4746 4747 void *netdev_adjacent_get_private(struct list_head *adj_list); 4748 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4749 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4750 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4751 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4752 struct netlink_ext_ack *extack); 4753 int netdev_master_upper_dev_link(struct net_device *dev, 4754 struct net_device *upper_dev, 4755 void *upper_priv, void *upper_info, 4756 struct netlink_ext_ack *extack); 4757 void netdev_upper_dev_unlink(struct net_device *dev, 4758 struct net_device *upper_dev); 4759 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4760 struct net_device *new_dev, 4761 struct net_device *dev, 4762 struct netlink_ext_ack *extack); 4763 void netdev_adjacent_change_commit(struct net_device *old_dev, 4764 struct net_device *new_dev, 4765 struct net_device *dev); 4766 void netdev_adjacent_change_abort(struct net_device *old_dev, 4767 struct net_device *new_dev, 4768 struct net_device *dev); 4769 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4770 void *netdev_lower_dev_get_private(struct net_device *dev, 4771 struct net_device *lower_dev); 4772 void netdev_lower_state_changed(struct net_device *lower_dev, 4773 void *lower_state_info); 4774 4775 /* RSS keys are 40 or 52 bytes long */ 4776 #define NETDEV_RSS_KEY_LEN 52 4777 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4778 void netdev_rss_key_fill(void *buffer, size_t len); 4779 4780 int skb_checksum_help(struct sk_buff *skb); 4781 int skb_crc32c_csum_help(struct sk_buff *skb); 4782 int skb_csum_hwoffload_help(struct sk_buff *skb, 4783 const netdev_features_t features); 4784 4785 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4786 netdev_features_t features, bool tx_path); 4787 struct sk_buff *skb_eth_gso_segment(struct sk_buff *skb, 4788 netdev_features_t features, __be16 type); 4789 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4790 netdev_features_t features); 4791 4792 struct netdev_bonding_info { 4793 ifslave slave; 4794 ifbond master; 4795 }; 4796 4797 struct netdev_notifier_bonding_info { 4798 struct netdev_notifier_info info; /* must be first */ 4799 struct netdev_bonding_info bonding_info; 4800 }; 4801 4802 void netdev_bonding_info_change(struct net_device *dev, 4803 struct netdev_bonding_info *bonding_info); 4804 4805 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4806 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4807 #else 4808 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4809 const void *data) 4810 { 4811 } 4812 #endif 4813 4814 static inline 4815 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4816 { 4817 return __skb_gso_segment(skb, features, true); 4818 } 4819 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4820 4821 static inline bool can_checksum_protocol(netdev_features_t features, 4822 __be16 protocol) 4823 { 4824 if (protocol == htons(ETH_P_FCOE)) 4825 return !!(features & NETIF_F_FCOE_CRC); 4826 4827 /* Assume this is an IP checksum (not SCTP CRC) */ 4828 4829 if (features & NETIF_F_HW_CSUM) { 4830 /* Can checksum everything */ 4831 return true; 4832 } 4833 4834 switch (protocol) { 4835 case htons(ETH_P_IP): 4836 return !!(features & NETIF_F_IP_CSUM); 4837 case htons(ETH_P_IPV6): 4838 return !!(features & NETIF_F_IPV6_CSUM); 4839 default: 4840 return false; 4841 } 4842 } 4843 4844 #ifdef CONFIG_BUG 4845 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4846 #else 4847 static inline void netdev_rx_csum_fault(struct net_device *dev, 4848 struct sk_buff *skb) 4849 { 4850 } 4851 #endif 4852 /* rx skb timestamps */ 4853 void net_enable_timestamp(void); 4854 void net_disable_timestamp(void); 4855 4856 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 4857 const struct skb_shared_hwtstamps *hwtstamps, 4858 bool cycles) 4859 { 4860 const struct net_device_ops *ops = dev->netdev_ops; 4861 4862 if (ops->ndo_get_tstamp) 4863 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 4864 4865 return hwtstamps->hwtstamp; 4866 } 4867 4868 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4869 struct sk_buff *skb, struct net_device *dev, 4870 bool more) 4871 { 4872 __this_cpu_write(softnet_data.xmit.more, more); 4873 return ops->ndo_start_xmit(skb, dev); 4874 } 4875 4876 static inline bool netdev_xmit_more(void) 4877 { 4878 return __this_cpu_read(softnet_data.xmit.more); 4879 } 4880 4881 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4882 struct netdev_queue *txq, bool more) 4883 { 4884 const struct net_device_ops *ops = dev->netdev_ops; 4885 netdev_tx_t rc; 4886 4887 rc = __netdev_start_xmit(ops, skb, dev, more); 4888 if (rc == NETDEV_TX_OK) 4889 txq_trans_update(txq); 4890 4891 return rc; 4892 } 4893 4894 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4895 const void *ns); 4896 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4897 const void *ns); 4898 4899 extern const struct kobj_ns_type_operations net_ns_type_operations; 4900 4901 const char *netdev_drivername(const struct net_device *dev); 4902 4903 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4904 netdev_features_t f2) 4905 { 4906 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4907 if (f1 & NETIF_F_HW_CSUM) 4908 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4909 else 4910 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4911 } 4912 4913 return f1 & f2; 4914 } 4915 4916 static inline netdev_features_t netdev_get_wanted_features( 4917 struct net_device *dev) 4918 { 4919 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4920 } 4921 netdev_features_t netdev_increment_features(netdev_features_t all, 4922 netdev_features_t one, netdev_features_t mask); 4923 4924 /* Allow TSO being used on stacked device : 4925 * Performing the GSO segmentation before last device 4926 * is a performance improvement. 4927 */ 4928 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4929 netdev_features_t mask) 4930 { 4931 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4932 } 4933 4934 int __netdev_update_features(struct net_device *dev); 4935 void netdev_update_features(struct net_device *dev); 4936 void netdev_change_features(struct net_device *dev); 4937 4938 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4939 struct net_device *dev); 4940 4941 netdev_features_t passthru_features_check(struct sk_buff *skb, 4942 struct net_device *dev, 4943 netdev_features_t features); 4944 netdev_features_t netif_skb_features(struct sk_buff *skb); 4945 4946 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4947 { 4948 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4949 4950 /* check flags correspondence */ 4951 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4952 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4953 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4954 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4955 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4956 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4957 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4958 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4959 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4960 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4961 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4962 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4963 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4964 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4965 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4966 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4967 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4968 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4969 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4970 4971 return (features & feature) == feature; 4972 } 4973 4974 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4975 { 4976 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4977 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4978 } 4979 4980 static inline bool netif_needs_gso(struct sk_buff *skb, 4981 netdev_features_t features) 4982 { 4983 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4984 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4985 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4986 } 4987 4988 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 4989 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 4990 void netif_inherit_tso_max(struct net_device *to, 4991 const struct net_device *from); 4992 4993 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4994 int pulled_hlen, u16 mac_offset, 4995 int mac_len) 4996 { 4997 skb->protocol = protocol; 4998 skb->encapsulation = 1; 4999 skb_push(skb, pulled_hlen); 5000 skb_reset_transport_header(skb); 5001 skb->mac_header = mac_offset; 5002 skb->network_header = skb->mac_header + mac_len; 5003 skb->mac_len = mac_len; 5004 } 5005 5006 static inline bool netif_is_macsec(const struct net_device *dev) 5007 { 5008 return dev->priv_flags & IFF_MACSEC; 5009 } 5010 5011 static inline bool netif_is_macvlan(const struct net_device *dev) 5012 { 5013 return dev->priv_flags & IFF_MACVLAN; 5014 } 5015 5016 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5017 { 5018 return dev->priv_flags & IFF_MACVLAN_PORT; 5019 } 5020 5021 static inline bool netif_is_bond_master(const struct net_device *dev) 5022 { 5023 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5024 } 5025 5026 static inline bool netif_is_bond_slave(const struct net_device *dev) 5027 { 5028 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5029 } 5030 5031 static inline bool netif_supports_nofcs(struct net_device *dev) 5032 { 5033 return dev->priv_flags & IFF_SUPP_NOFCS; 5034 } 5035 5036 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5037 { 5038 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5039 } 5040 5041 static inline bool netif_is_l3_master(const struct net_device *dev) 5042 { 5043 return dev->priv_flags & IFF_L3MDEV_MASTER; 5044 } 5045 5046 static inline bool netif_is_l3_slave(const struct net_device *dev) 5047 { 5048 return dev->priv_flags & IFF_L3MDEV_SLAVE; 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_port(const struct net_device *dev) 5072 { 5073 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5074 } 5075 5076 static inline bool netif_is_team_master(const struct net_device *dev) 5077 { 5078 return dev->priv_flags & IFF_TEAM; 5079 } 5080 5081 static inline bool netif_is_team_port(const struct net_device *dev) 5082 { 5083 return dev->priv_flags & IFF_TEAM_PORT; 5084 } 5085 5086 static inline bool netif_is_lag_master(const struct net_device *dev) 5087 { 5088 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5089 } 5090 5091 static inline bool netif_is_lag_port(const struct net_device *dev) 5092 { 5093 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5094 } 5095 5096 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5097 { 5098 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5099 } 5100 5101 static inline bool netif_is_failover(const struct net_device *dev) 5102 { 5103 return dev->priv_flags & IFF_FAILOVER; 5104 } 5105 5106 static inline bool netif_is_failover_slave(const struct net_device *dev) 5107 { 5108 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5109 } 5110 5111 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5112 static inline void netif_keep_dst(struct net_device *dev) 5113 { 5114 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5115 } 5116 5117 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5118 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5119 { 5120 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5121 return netif_is_macsec(dev); 5122 } 5123 5124 extern struct pernet_operations __net_initdata loopback_net_ops; 5125 5126 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5127 5128 /* netdev_printk helpers, similar to dev_printk */ 5129 5130 static inline const char *netdev_name(const struct net_device *dev) 5131 { 5132 if (!dev->name[0] || strchr(dev->name, '%')) 5133 return "(unnamed net_device)"; 5134 return dev->name; 5135 } 5136 5137 static inline const char *netdev_reg_state(const struct net_device *dev) 5138 { 5139 switch (dev->reg_state) { 5140 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5141 case NETREG_REGISTERED: return ""; 5142 case NETREG_UNREGISTERING: return " (unregistering)"; 5143 case NETREG_UNREGISTERED: return " (unregistered)"; 5144 case NETREG_RELEASED: return " (released)"; 5145 case NETREG_DUMMY: return " (dummy)"; 5146 } 5147 5148 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5149 return " (unknown)"; 5150 } 5151 5152 #define MODULE_ALIAS_NETDEV(device) \ 5153 MODULE_ALIAS("netdev-" device) 5154 5155 /* 5156 * netdev_WARN() acts like dev_printk(), but with the key difference 5157 * of using a WARN/WARN_ON to get the message out, including the 5158 * file/line information and a backtrace. 5159 */ 5160 #define netdev_WARN(dev, format, args...) \ 5161 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5162 netdev_reg_state(dev), ##args) 5163 5164 #define netdev_WARN_ONCE(dev, format, args...) \ 5165 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5166 netdev_reg_state(dev), ##args) 5167 5168 /* 5169 * The list of packet types we will receive (as opposed to discard) 5170 * and the routines to invoke. 5171 * 5172 * Why 16. Because with 16 the only overlap we get on a hash of the 5173 * low nibble of the protocol value is RARP/SNAP/X.25. 5174 * 5175 * 0800 IP 5176 * 0001 802.3 5177 * 0002 AX.25 5178 * 0004 802.2 5179 * 8035 RARP 5180 * 0005 SNAP 5181 * 0805 X.25 5182 * 0806 ARP 5183 * 8137 IPX 5184 * 0009 Localtalk 5185 * 86DD IPv6 5186 */ 5187 #define PTYPE_HASH_SIZE (16) 5188 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5189 5190 extern struct list_head ptype_all __read_mostly; 5191 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5192 5193 extern struct net_device *blackhole_netdev; 5194 5195 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */ 5196 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD) 5197 #define DEV_STATS_ADD(DEV, FIELD, VAL) \ 5198 atomic_long_add((VAL), &(DEV)->stats.__##FIELD) 5199 5200 #endif /* _LINUX_NETDEVICE_H */ 5201