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 * @gso_ipv4_max_size: Maximum size of generic segmentation offload, 1968 * for IPv4. 1969 * 1970 * @dcbnl_ops: Data Center Bridging netlink ops 1971 * @num_tc: Number of traffic classes in the net device 1972 * @tc_to_txq: XXX: need comments on this one 1973 * @prio_tc_map: XXX: need comments on this one 1974 * 1975 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1976 * 1977 * @priomap: XXX: need comments on this one 1978 * @phydev: Physical device may attach itself 1979 * for hardware timestamping 1980 * @sfp_bus: attached &struct sfp_bus structure. 1981 * 1982 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1983 * 1984 * @proto_down: protocol port state information can be sent to the 1985 * switch driver and used to set the phys state of the 1986 * switch port. 1987 * 1988 * @wol_enabled: Wake-on-LAN is enabled 1989 * 1990 * @threaded: napi threaded mode is enabled 1991 * 1992 * @net_notifier_list: List of per-net netdev notifier block 1993 * that follow this device when it is moved 1994 * to another network namespace. 1995 * 1996 * @macsec_ops: MACsec offloading ops 1997 * 1998 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1999 * offload capabilities of the device 2000 * @udp_tunnel_nic: UDP tunnel offload state 2001 * @xdp_state: stores info on attached XDP BPF programs 2002 * 2003 * @nested_level: Used as a parameter of spin_lock_nested() of 2004 * dev->addr_list_lock. 2005 * @unlink_list: As netif_addr_lock() can be called recursively, 2006 * keep a list of interfaces to be deleted. 2007 * @gro_max_size: Maximum size of aggregated packet in generic 2008 * receive offload (GRO) 2009 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic 2010 * receive offload (GRO), for IPv4. 2011 * 2012 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 2013 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 2014 * @watchdog_dev_tracker: refcount tracker used by watchdog. 2015 * @dev_registered_tracker: tracker for reference held while 2016 * registered 2017 * @offload_xstats_l3: L3 HW stats for this netdevice. 2018 * 2019 * @devlink_port: Pointer to related devlink port structure. 2020 * Assigned by a driver before netdev registration using 2021 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static 2022 * during the time netdevice is registered. 2023 * 2024 * FIXME: cleanup struct net_device such that network protocol info 2025 * moves out. 2026 */ 2027 2028 struct net_device { 2029 char name[IFNAMSIZ]; 2030 struct netdev_name_node *name_node; 2031 struct dev_ifalias __rcu *ifalias; 2032 /* 2033 * I/O specific fields 2034 * FIXME: Merge these and struct ifmap into one 2035 */ 2036 unsigned long mem_end; 2037 unsigned long mem_start; 2038 unsigned long base_addr; 2039 2040 /* 2041 * Some hardware also needs these fields (state,dev_list, 2042 * napi_list,unreg_list,close_list) but they are not 2043 * part of the usual set specified in Space.c. 2044 */ 2045 2046 unsigned long state; 2047 2048 struct list_head dev_list; 2049 struct list_head napi_list; 2050 struct list_head unreg_list; 2051 struct list_head close_list; 2052 struct list_head ptype_all; 2053 struct list_head ptype_specific; 2054 2055 struct { 2056 struct list_head upper; 2057 struct list_head lower; 2058 } adj_list; 2059 2060 /* Read-mostly cache-line for fast-path access */ 2061 unsigned int flags; 2062 unsigned long long priv_flags; 2063 const struct net_device_ops *netdev_ops; 2064 const struct xdp_metadata_ops *xdp_metadata_ops; 2065 int ifindex; 2066 unsigned short gflags; 2067 unsigned short hard_header_len; 2068 2069 /* Note : dev->mtu is often read without holding a lock. 2070 * Writers usually hold RTNL. 2071 * It is recommended to use READ_ONCE() to annotate the reads, 2072 * and to use WRITE_ONCE() to annotate the writes. 2073 */ 2074 unsigned int mtu; 2075 unsigned short needed_headroom; 2076 unsigned short needed_tailroom; 2077 2078 netdev_features_t features; 2079 netdev_features_t hw_features; 2080 netdev_features_t wanted_features; 2081 netdev_features_t vlan_features; 2082 netdev_features_t hw_enc_features; 2083 netdev_features_t mpls_features; 2084 netdev_features_t gso_partial_features; 2085 2086 unsigned int min_mtu; 2087 unsigned int max_mtu; 2088 unsigned short type; 2089 unsigned char min_header_len; 2090 unsigned char name_assign_type; 2091 2092 int group; 2093 2094 struct net_device_stats stats; /* not used by modern drivers */ 2095 2096 struct net_device_core_stats __percpu *core_stats; 2097 2098 /* Stats to monitor link on/off, flapping */ 2099 atomic_t carrier_up_count; 2100 atomic_t carrier_down_count; 2101 2102 #ifdef CONFIG_WIRELESS_EXT 2103 const struct iw_handler_def *wireless_handlers; 2104 struct iw_public_data *wireless_data; 2105 #endif 2106 const struct ethtool_ops *ethtool_ops; 2107 #ifdef CONFIG_NET_L3_MASTER_DEV 2108 const struct l3mdev_ops *l3mdev_ops; 2109 #endif 2110 #if IS_ENABLED(CONFIG_IPV6) 2111 const struct ndisc_ops *ndisc_ops; 2112 #endif 2113 2114 #ifdef CONFIG_XFRM_OFFLOAD 2115 const struct xfrmdev_ops *xfrmdev_ops; 2116 #endif 2117 2118 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2119 const struct tlsdev_ops *tlsdev_ops; 2120 #endif 2121 2122 const struct header_ops *header_ops; 2123 2124 unsigned char operstate; 2125 unsigned char link_mode; 2126 2127 unsigned char if_port; 2128 unsigned char dma; 2129 2130 /* Interface address info. */ 2131 unsigned char perm_addr[MAX_ADDR_LEN]; 2132 unsigned char addr_assign_type; 2133 unsigned char addr_len; 2134 unsigned char upper_level; 2135 unsigned char lower_level; 2136 2137 unsigned short neigh_priv_len; 2138 unsigned short dev_id; 2139 unsigned short dev_port; 2140 unsigned short padded; 2141 2142 spinlock_t addr_list_lock; 2143 int irq; 2144 2145 struct netdev_hw_addr_list uc; 2146 struct netdev_hw_addr_list mc; 2147 struct netdev_hw_addr_list dev_addrs; 2148 2149 #ifdef CONFIG_SYSFS 2150 struct kset *queues_kset; 2151 #endif 2152 #ifdef CONFIG_LOCKDEP 2153 struct list_head unlink_list; 2154 #endif 2155 unsigned int promiscuity; 2156 unsigned int allmulti; 2157 bool uc_promisc; 2158 #ifdef CONFIG_LOCKDEP 2159 unsigned char nested_level; 2160 #endif 2161 2162 2163 /* Protocol-specific pointers */ 2164 2165 struct in_device __rcu *ip_ptr; 2166 struct inet6_dev __rcu *ip6_ptr; 2167 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2168 struct vlan_info __rcu *vlan_info; 2169 #endif 2170 #if IS_ENABLED(CONFIG_NET_DSA) 2171 struct dsa_port *dsa_ptr; 2172 #endif 2173 #if IS_ENABLED(CONFIG_TIPC) 2174 struct tipc_bearer __rcu *tipc_ptr; 2175 #endif 2176 #if IS_ENABLED(CONFIG_ATALK) 2177 void *atalk_ptr; 2178 #endif 2179 #if IS_ENABLED(CONFIG_AX25) 2180 void *ax25_ptr; 2181 #endif 2182 #if IS_ENABLED(CONFIG_CFG80211) 2183 struct wireless_dev *ieee80211_ptr; 2184 #endif 2185 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2186 struct wpan_dev *ieee802154_ptr; 2187 #endif 2188 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2189 struct mpls_dev __rcu *mpls_ptr; 2190 #endif 2191 #if IS_ENABLED(CONFIG_MCTP) 2192 struct mctp_dev __rcu *mctp_ptr; 2193 #endif 2194 2195 /* 2196 * Cache lines mostly used on receive path (including eth_type_trans()) 2197 */ 2198 /* Interface address info used in eth_type_trans() */ 2199 const unsigned char *dev_addr; 2200 2201 struct netdev_rx_queue *_rx; 2202 unsigned int num_rx_queues; 2203 unsigned int real_num_rx_queues; 2204 2205 struct bpf_prog __rcu *xdp_prog; 2206 unsigned long gro_flush_timeout; 2207 int napi_defer_hard_irqs; 2208 #define GRO_LEGACY_MAX_SIZE 65536u 2209 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2210 * and shinfo->gso_segs is a 16bit field. 2211 */ 2212 #define GRO_MAX_SIZE (8 * 65535u) 2213 unsigned int gro_max_size; 2214 unsigned int gro_ipv4_max_size; 2215 rx_handler_func_t __rcu *rx_handler; 2216 void __rcu *rx_handler_data; 2217 2218 #ifdef CONFIG_NET_CLS_ACT 2219 struct mini_Qdisc __rcu *miniq_ingress; 2220 #endif 2221 struct netdev_queue __rcu *ingress_queue; 2222 #ifdef CONFIG_NETFILTER_INGRESS 2223 struct nf_hook_entries __rcu *nf_hooks_ingress; 2224 #endif 2225 2226 unsigned char broadcast[MAX_ADDR_LEN]; 2227 #ifdef CONFIG_RFS_ACCEL 2228 struct cpu_rmap *rx_cpu_rmap; 2229 #endif 2230 struct hlist_node index_hlist; 2231 2232 /* 2233 * Cache lines mostly used on transmit path 2234 */ 2235 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2236 unsigned int num_tx_queues; 2237 unsigned int real_num_tx_queues; 2238 struct Qdisc __rcu *qdisc; 2239 unsigned int tx_queue_len; 2240 spinlock_t tx_global_lock; 2241 2242 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2243 2244 #ifdef CONFIG_XPS 2245 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2246 #endif 2247 #ifdef CONFIG_NET_CLS_ACT 2248 struct mini_Qdisc __rcu *miniq_egress; 2249 #endif 2250 #ifdef CONFIG_NETFILTER_EGRESS 2251 struct nf_hook_entries __rcu *nf_hooks_egress; 2252 #endif 2253 2254 #ifdef CONFIG_NET_SCHED 2255 DECLARE_HASHTABLE (qdisc_hash, 4); 2256 #endif 2257 /* These may be needed for future network-power-down code. */ 2258 struct timer_list watchdog_timer; 2259 int watchdog_timeo; 2260 2261 u32 proto_down_reason; 2262 2263 struct list_head todo_list; 2264 2265 #ifdef CONFIG_PCPU_DEV_REFCNT 2266 int __percpu *pcpu_refcnt; 2267 #else 2268 refcount_t dev_refcnt; 2269 #endif 2270 struct ref_tracker_dir refcnt_tracker; 2271 2272 struct list_head link_watch_list; 2273 2274 enum { NETREG_UNINITIALIZED=0, 2275 NETREG_REGISTERED, /* completed register_netdevice */ 2276 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2277 NETREG_UNREGISTERED, /* completed unregister todo */ 2278 NETREG_RELEASED, /* called free_netdev */ 2279 NETREG_DUMMY, /* dummy device for NAPI poll */ 2280 } reg_state:8; 2281 2282 bool dismantle; 2283 2284 enum { 2285 RTNL_LINK_INITIALIZED, 2286 RTNL_LINK_INITIALIZING, 2287 } rtnl_link_state:16; 2288 2289 bool needs_free_netdev; 2290 void (*priv_destructor)(struct net_device *dev); 2291 2292 #ifdef CONFIG_NETPOLL 2293 struct netpoll_info __rcu *npinfo; 2294 #endif 2295 2296 possible_net_t nd_net; 2297 2298 /* mid-layer private */ 2299 void *ml_priv; 2300 enum netdev_ml_priv_type ml_priv_type; 2301 2302 union { 2303 struct pcpu_lstats __percpu *lstats; 2304 struct pcpu_sw_netstats __percpu *tstats; 2305 struct pcpu_dstats __percpu *dstats; 2306 }; 2307 2308 #if IS_ENABLED(CONFIG_GARP) 2309 struct garp_port __rcu *garp_port; 2310 #endif 2311 #if IS_ENABLED(CONFIG_MRP) 2312 struct mrp_port __rcu *mrp_port; 2313 #endif 2314 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2315 struct dm_hw_stat_delta __rcu *dm_private; 2316 #endif 2317 struct device dev; 2318 const struct attribute_group *sysfs_groups[4]; 2319 const struct attribute_group *sysfs_rx_queue_group; 2320 2321 const struct rtnl_link_ops *rtnl_link_ops; 2322 2323 /* for setting kernel sock attribute on TCP connection setup */ 2324 #define GSO_MAX_SEGS 65535u 2325 #define GSO_LEGACY_MAX_SIZE 65536u 2326 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2327 * and shinfo->gso_segs is a 16bit field. 2328 */ 2329 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2330 2331 unsigned int gso_max_size; 2332 #define TSO_LEGACY_MAX_SIZE 65536 2333 #define TSO_MAX_SIZE UINT_MAX 2334 unsigned int tso_max_size; 2335 u16 gso_max_segs; 2336 #define TSO_MAX_SEGS U16_MAX 2337 u16 tso_max_segs; 2338 unsigned int gso_ipv4_max_size; 2339 2340 #ifdef CONFIG_DCB 2341 const struct dcbnl_rtnl_ops *dcbnl_ops; 2342 #endif 2343 s16 num_tc; 2344 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2345 u8 prio_tc_map[TC_BITMASK + 1]; 2346 2347 #if IS_ENABLED(CONFIG_FCOE) 2348 unsigned int fcoe_ddp_xid; 2349 #endif 2350 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2351 struct netprio_map __rcu *priomap; 2352 #endif 2353 struct phy_device *phydev; 2354 struct sfp_bus *sfp_bus; 2355 struct lock_class_key *qdisc_tx_busylock; 2356 bool proto_down; 2357 unsigned wol_enabled:1; 2358 unsigned threaded:1; 2359 2360 struct list_head net_notifier_list; 2361 2362 #if IS_ENABLED(CONFIG_MACSEC) 2363 /* MACsec management functions */ 2364 const struct macsec_ops *macsec_ops; 2365 #endif 2366 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2367 struct udp_tunnel_nic *udp_tunnel_nic; 2368 2369 /* protected by rtnl_lock */ 2370 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2371 2372 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2373 netdevice_tracker linkwatch_dev_tracker; 2374 netdevice_tracker watchdog_dev_tracker; 2375 netdevice_tracker dev_registered_tracker; 2376 struct rtnl_hw_stats64 *offload_xstats_l3; 2377 2378 struct devlink_port *devlink_port; 2379 }; 2380 #define to_net_dev(d) container_of(d, struct net_device, dev) 2381 2382 /* 2383 * Driver should use this to assign devlink port instance to a netdevice 2384 * before it registers the netdevice. Therefore devlink_port is static 2385 * during the netdev lifetime after it is registered. 2386 */ 2387 #define SET_NETDEV_DEVLINK_PORT(dev, port) \ 2388 ({ \ 2389 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \ 2390 ((dev)->devlink_port = (port)); \ 2391 }) 2392 2393 static inline bool netif_elide_gro(const struct net_device *dev) 2394 { 2395 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2396 return true; 2397 return false; 2398 } 2399 2400 #define NETDEV_ALIGN 32 2401 2402 static inline 2403 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2404 { 2405 return dev->prio_tc_map[prio & TC_BITMASK]; 2406 } 2407 2408 static inline 2409 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2410 { 2411 if (tc >= dev->num_tc) 2412 return -EINVAL; 2413 2414 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2415 return 0; 2416 } 2417 2418 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2419 void netdev_reset_tc(struct net_device *dev); 2420 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2421 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2422 2423 static inline 2424 int netdev_get_num_tc(struct net_device *dev) 2425 { 2426 return dev->num_tc; 2427 } 2428 2429 static inline void net_prefetch(void *p) 2430 { 2431 prefetch(p); 2432 #if L1_CACHE_BYTES < 128 2433 prefetch((u8 *)p + L1_CACHE_BYTES); 2434 #endif 2435 } 2436 2437 static inline void net_prefetchw(void *p) 2438 { 2439 prefetchw(p); 2440 #if L1_CACHE_BYTES < 128 2441 prefetchw((u8 *)p + L1_CACHE_BYTES); 2442 #endif 2443 } 2444 2445 void netdev_unbind_sb_channel(struct net_device *dev, 2446 struct net_device *sb_dev); 2447 int netdev_bind_sb_channel_queue(struct net_device *dev, 2448 struct net_device *sb_dev, 2449 u8 tc, u16 count, u16 offset); 2450 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2451 static inline int netdev_get_sb_channel(struct net_device *dev) 2452 { 2453 return max_t(int, -dev->num_tc, 0); 2454 } 2455 2456 static inline 2457 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2458 unsigned int index) 2459 { 2460 return &dev->_tx[index]; 2461 } 2462 2463 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2464 const struct sk_buff *skb) 2465 { 2466 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2467 } 2468 2469 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2470 void (*f)(struct net_device *, 2471 struct netdev_queue *, 2472 void *), 2473 void *arg) 2474 { 2475 unsigned int i; 2476 2477 for (i = 0; i < dev->num_tx_queues; i++) 2478 f(dev, &dev->_tx[i], arg); 2479 } 2480 2481 #define netdev_lockdep_set_classes(dev) \ 2482 { \ 2483 static struct lock_class_key qdisc_tx_busylock_key; \ 2484 static struct lock_class_key qdisc_xmit_lock_key; \ 2485 static struct lock_class_key dev_addr_list_lock_key; \ 2486 unsigned int i; \ 2487 \ 2488 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2489 lockdep_set_class(&(dev)->addr_list_lock, \ 2490 &dev_addr_list_lock_key); \ 2491 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2492 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2493 &qdisc_xmit_lock_key); \ 2494 } 2495 2496 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2497 struct net_device *sb_dev); 2498 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2499 struct sk_buff *skb, 2500 struct net_device *sb_dev); 2501 2502 /* returns the headroom that the master device needs to take in account 2503 * when forwarding to this dev 2504 */ 2505 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2506 { 2507 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2508 } 2509 2510 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2511 { 2512 if (dev->netdev_ops->ndo_set_rx_headroom) 2513 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2514 } 2515 2516 /* set the device rx headroom to the dev's default */ 2517 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2518 { 2519 netdev_set_rx_headroom(dev, -1); 2520 } 2521 2522 static inline void *netdev_get_ml_priv(struct net_device *dev, 2523 enum netdev_ml_priv_type type) 2524 { 2525 if (dev->ml_priv_type != type) 2526 return NULL; 2527 2528 return dev->ml_priv; 2529 } 2530 2531 static inline void netdev_set_ml_priv(struct net_device *dev, 2532 void *ml_priv, 2533 enum netdev_ml_priv_type type) 2534 { 2535 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2536 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2537 dev->ml_priv_type, type); 2538 WARN(!dev->ml_priv_type && dev->ml_priv, 2539 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2540 2541 dev->ml_priv = ml_priv; 2542 dev->ml_priv_type = type; 2543 } 2544 2545 /* 2546 * Net namespace inlines 2547 */ 2548 static inline 2549 struct net *dev_net(const struct net_device *dev) 2550 { 2551 return read_pnet(&dev->nd_net); 2552 } 2553 2554 static inline 2555 void dev_net_set(struct net_device *dev, struct net *net) 2556 { 2557 write_pnet(&dev->nd_net, net); 2558 } 2559 2560 /** 2561 * netdev_priv - access network device private data 2562 * @dev: network device 2563 * 2564 * Get network device private data 2565 */ 2566 static inline void *netdev_priv(const struct net_device *dev) 2567 { 2568 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2569 } 2570 2571 /* Set the sysfs physical device reference for the network logical device 2572 * if set prior to registration will cause a symlink during initialization. 2573 */ 2574 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2575 2576 /* Set the sysfs device type for the network logical device to allow 2577 * fine-grained identification of different network device types. For 2578 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2579 */ 2580 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2581 2582 /* Default NAPI poll() weight 2583 * Device drivers are strongly advised to not use bigger value 2584 */ 2585 #define NAPI_POLL_WEIGHT 64 2586 2587 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2588 int (*poll)(struct napi_struct *, int), int weight); 2589 2590 /** 2591 * netif_napi_add() - initialize a NAPI context 2592 * @dev: network device 2593 * @napi: NAPI context 2594 * @poll: polling function 2595 * 2596 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2597 * *any* of the other NAPI-related functions. 2598 */ 2599 static inline void 2600 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2601 int (*poll)(struct napi_struct *, int)) 2602 { 2603 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2604 } 2605 2606 static inline void 2607 netif_napi_add_tx_weight(struct net_device *dev, 2608 struct napi_struct *napi, 2609 int (*poll)(struct napi_struct *, int), 2610 int weight) 2611 { 2612 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2613 netif_napi_add_weight(dev, napi, poll, weight); 2614 } 2615 2616 /** 2617 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2618 * @dev: network device 2619 * @napi: NAPI context 2620 * @poll: polling function 2621 * 2622 * This variant of netif_napi_add() should be used from drivers using NAPI 2623 * to exclusively poll a TX queue. 2624 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2625 */ 2626 static inline void netif_napi_add_tx(struct net_device *dev, 2627 struct napi_struct *napi, 2628 int (*poll)(struct napi_struct *, int)) 2629 { 2630 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2631 } 2632 2633 /** 2634 * __netif_napi_del - remove a NAPI context 2635 * @napi: NAPI context 2636 * 2637 * Warning: caller must observe RCU grace period before freeing memory 2638 * containing @napi. Drivers might want to call this helper to combine 2639 * all the needed RCU grace periods into a single one. 2640 */ 2641 void __netif_napi_del(struct napi_struct *napi); 2642 2643 /** 2644 * netif_napi_del - remove a NAPI context 2645 * @napi: NAPI context 2646 * 2647 * netif_napi_del() removes a NAPI context from the network device NAPI list 2648 */ 2649 static inline void netif_napi_del(struct napi_struct *napi) 2650 { 2651 __netif_napi_del(napi); 2652 synchronize_net(); 2653 } 2654 2655 struct packet_type { 2656 __be16 type; /* This is really htons(ether_type). */ 2657 bool ignore_outgoing; 2658 struct net_device *dev; /* NULL is wildcarded here */ 2659 netdevice_tracker dev_tracker; 2660 int (*func) (struct sk_buff *, 2661 struct net_device *, 2662 struct packet_type *, 2663 struct net_device *); 2664 void (*list_func) (struct list_head *, 2665 struct packet_type *, 2666 struct net_device *); 2667 bool (*id_match)(struct packet_type *ptype, 2668 struct sock *sk); 2669 struct net *af_packet_net; 2670 void *af_packet_priv; 2671 struct list_head list; 2672 }; 2673 2674 struct offload_callbacks { 2675 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2676 netdev_features_t features); 2677 struct sk_buff *(*gro_receive)(struct list_head *head, 2678 struct sk_buff *skb); 2679 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2680 }; 2681 2682 struct packet_offload { 2683 __be16 type; /* This is really htons(ether_type). */ 2684 u16 priority; 2685 struct offload_callbacks callbacks; 2686 struct list_head list; 2687 }; 2688 2689 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2690 struct pcpu_sw_netstats { 2691 u64_stats_t rx_packets; 2692 u64_stats_t rx_bytes; 2693 u64_stats_t tx_packets; 2694 u64_stats_t tx_bytes; 2695 struct u64_stats_sync syncp; 2696 } __aligned(4 * sizeof(u64)); 2697 2698 struct pcpu_lstats { 2699 u64_stats_t packets; 2700 u64_stats_t bytes; 2701 struct u64_stats_sync syncp; 2702 } __aligned(2 * sizeof(u64)); 2703 2704 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2705 2706 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2707 { 2708 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2709 2710 u64_stats_update_begin(&tstats->syncp); 2711 u64_stats_add(&tstats->rx_bytes, len); 2712 u64_stats_inc(&tstats->rx_packets); 2713 u64_stats_update_end(&tstats->syncp); 2714 } 2715 2716 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2717 unsigned int packets, 2718 unsigned int len) 2719 { 2720 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2721 2722 u64_stats_update_begin(&tstats->syncp); 2723 u64_stats_add(&tstats->tx_bytes, len); 2724 u64_stats_add(&tstats->tx_packets, packets); 2725 u64_stats_update_end(&tstats->syncp); 2726 } 2727 2728 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2729 { 2730 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2731 2732 u64_stats_update_begin(&lstats->syncp); 2733 u64_stats_add(&lstats->bytes, len); 2734 u64_stats_inc(&lstats->packets); 2735 u64_stats_update_end(&lstats->syncp); 2736 } 2737 2738 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2739 ({ \ 2740 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2741 if (pcpu_stats) { \ 2742 int __cpu; \ 2743 for_each_possible_cpu(__cpu) { \ 2744 typeof(type) *stat; \ 2745 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2746 u64_stats_init(&stat->syncp); \ 2747 } \ 2748 } \ 2749 pcpu_stats; \ 2750 }) 2751 2752 #define netdev_alloc_pcpu_stats(type) \ 2753 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2754 2755 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2756 ({ \ 2757 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2758 if (pcpu_stats) { \ 2759 int __cpu; \ 2760 for_each_possible_cpu(__cpu) { \ 2761 typeof(type) *stat; \ 2762 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2763 u64_stats_init(&stat->syncp); \ 2764 } \ 2765 } \ 2766 pcpu_stats; \ 2767 }) 2768 2769 enum netdev_lag_tx_type { 2770 NETDEV_LAG_TX_TYPE_UNKNOWN, 2771 NETDEV_LAG_TX_TYPE_RANDOM, 2772 NETDEV_LAG_TX_TYPE_BROADCAST, 2773 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2774 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2775 NETDEV_LAG_TX_TYPE_HASH, 2776 }; 2777 2778 enum netdev_lag_hash { 2779 NETDEV_LAG_HASH_NONE, 2780 NETDEV_LAG_HASH_L2, 2781 NETDEV_LAG_HASH_L34, 2782 NETDEV_LAG_HASH_L23, 2783 NETDEV_LAG_HASH_E23, 2784 NETDEV_LAG_HASH_E34, 2785 NETDEV_LAG_HASH_VLAN_SRCMAC, 2786 NETDEV_LAG_HASH_UNKNOWN, 2787 }; 2788 2789 struct netdev_lag_upper_info { 2790 enum netdev_lag_tx_type tx_type; 2791 enum netdev_lag_hash hash_type; 2792 }; 2793 2794 struct netdev_lag_lower_state_info { 2795 u8 link_up : 1, 2796 tx_enabled : 1; 2797 }; 2798 2799 #include <linux/notifier.h> 2800 2801 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2802 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2803 * adding new types. 2804 */ 2805 enum netdev_cmd { 2806 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2807 NETDEV_DOWN, 2808 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2809 detected a hardware crash and restarted 2810 - we can use this eg to kick tcp sessions 2811 once done */ 2812 NETDEV_CHANGE, /* Notify device state change */ 2813 NETDEV_REGISTER, 2814 NETDEV_UNREGISTER, 2815 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2816 NETDEV_CHANGEADDR, /* notify after the address change */ 2817 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2818 NETDEV_GOING_DOWN, 2819 NETDEV_CHANGENAME, 2820 NETDEV_FEAT_CHANGE, 2821 NETDEV_BONDING_FAILOVER, 2822 NETDEV_PRE_UP, 2823 NETDEV_PRE_TYPE_CHANGE, 2824 NETDEV_POST_TYPE_CHANGE, 2825 NETDEV_POST_INIT, 2826 NETDEV_PRE_UNINIT, 2827 NETDEV_RELEASE, 2828 NETDEV_NOTIFY_PEERS, 2829 NETDEV_JOIN, 2830 NETDEV_CHANGEUPPER, 2831 NETDEV_RESEND_IGMP, 2832 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2833 NETDEV_CHANGEINFODATA, 2834 NETDEV_BONDING_INFO, 2835 NETDEV_PRECHANGEUPPER, 2836 NETDEV_CHANGELOWERSTATE, 2837 NETDEV_UDP_TUNNEL_PUSH_INFO, 2838 NETDEV_UDP_TUNNEL_DROP_INFO, 2839 NETDEV_CHANGE_TX_QUEUE_LEN, 2840 NETDEV_CVLAN_FILTER_PUSH_INFO, 2841 NETDEV_CVLAN_FILTER_DROP_INFO, 2842 NETDEV_SVLAN_FILTER_PUSH_INFO, 2843 NETDEV_SVLAN_FILTER_DROP_INFO, 2844 NETDEV_OFFLOAD_XSTATS_ENABLE, 2845 NETDEV_OFFLOAD_XSTATS_DISABLE, 2846 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 2847 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 2848 }; 2849 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2850 2851 int register_netdevice_notifier(struct notifier_block *nb); 2852 int unregister_netdevice_notifier(struct notifier_block *nb); 2853 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2854 int unregister_netdevice_notifier_net(struct net *net, 2855 struct notifier_block *nb); 2856 void move_netdevice_notifier_net(struct net *src_net, struct net *dst_net, 2857 struct notifier_block *nb); 2858 int register_netdevice_notifier_dev_net(struct net_device *dev, 2859 struct notifier_block *nb, 2860 struct netdev_net_notifier *nn); 2861 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2862 struct notifier_block *nb, 2863 struct netdev_net_notifier *nn); 2864 2865 struct netdev_notifier_info { 2866 struct net_device *dev; 2867 struct netlink_ext_ack *extack; 2868 }; 2869 2870 struct netdev_notifier_info_ext { 2871 struct netdev_notifier_info info; /* must be first */ 2872 union { 2873 u32 mtu; 2874 } ext; 2875 }; 2876 2877 struct netdev_notifier_change_info { 2878 struct netdev_notifier_info info; /* must be first */ 2879 unsigned int flags_changed; 2880 }; 2881 2882 struct netdev_notifier_changeupper_info { 2883 struct netdev_notifier_info info; /* must be first */ 2884 struct net_device *upper_dev; /* new upper dev */ 2885 bool master; /* is upper dev master */ 2886 bool linking; /* is the notification for link or unlink */ 2887 void *upper_info; /* upper dev info */ 2888 }; 2889 2890 struct netdev_notifier_changelowerstate_info { 2891 struct netdev_notifier_info info; /* must be first */ 2892 void *lower_state_info; /* is lower dev state */ 2893 }; 2894 2895 struct netdev_notifier_pre_changeaddr_info { 2896 struct netdev_notifier_info info; /* must be first */ 2897 const unsigned char *dev_addr; 2898 }; 2899 2900 enum netdev_offload_xstats_type { 2901 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 2902 }; 2903 2904 struct netdev_notifier_offload_xstats_info { 2905 struct netdev_notifier_info info; /* must be first */ 2906 enum netdev_offload_xstats_type type; 2907 2908 union { 2909 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 2910 struct netdev_notifier_offload_xstats_rd *report_delta; 2911 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 2912 struct netdev_notifier_offload_xstats_ru *report_used; 2913 }; 2914 }; 2915 2916 int netdev_offload_xstats_enable(struct net_device *dev, 2917 enum netdev_offload_xstats_type type, 2918 struct netlink_ext_ack *extack); 2919 int netdev_offload_xstats_disable(struct net_device *dev, 2920 enum netdev_offload_xstats_type type); 2921 bool netdev_offload_xstats_enabled(const struct net_device *dev, 2922 enum netdev_offload_xstats_type type); 2923 int netdev_offload_xstats_get(struct net_device *dev, 2924 enum netdev_offload_xstats_type type, 2925 struct rtnl_hw_stats64 *stats, bool *used, 2926 struct netlink_ext_ack *extack); 2927 void 2928 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 2929 const struct rtnl_hw_stats64 *stats); 2930 void 2931 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 2932 void netdev_offload_xstats_push_delta(struct net_device *dev, 2933 enum netdev_offload_xstats_type type, 2934 const struct rtnl_hw_stats64 *stats); 2935 2936 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2937 struct net_device *dev) 2938 { 2939 info->dev = dev; 2940 info->extack = NULL; 2941 } 2942 2943 static inline struct net_device * 2944 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2945 { 2946 return info->dev; 2947 } 2948 2949 static inline struct netlink_ext_ack * 2950 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2951 { 2952 return info->extack; 2953 } 2954 2955 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2956 2957 2958 extern rwlock_t dev_base_lock; /* Device list lock */ 2959 2960 #define for_each_netdev(net, d) \ 2961 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2962 #define for_each_netdev_reverse(net, d) \ 2963 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2964 #define for_each_netdev_rcu(net, d) \ 2965 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2966 #define for_each_netdev_safe(net, d, n) \ 2967 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2968 #define for_each_netdev_continue(net, d) \ 2969 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2970 #define for_each_netdev_continue_reverse(net, d) \ 2971 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2972 dev_list) 2973 #define for_each_netdev_continue_rcu(net, d) \ 2974 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2975 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2976 for_each_netdev_rcu(&init_net, slave) \ 2977 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2978 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2979 2980 static inline struct net_device *next_net_device(struct net_device *dev) 2981 { 2982 struct list_head *lh; 2983 struct net *net; 2984 2985 net = dev_net(dev); 2986 lh = dev->dev_list.next; 2987 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2988 } 2989 2990 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2991 { 2992 struct list_head *lh; 2993 struct net *net; 2994 2995 net = dev_net(dev); 2996 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2997 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2998 } 2999 3000 static inline struct net_device *first_net_device(struct net *net) 3001 { 3002 return list_empty(&net->dev_base_head) ? NULL : 3003 net_device_entry(net->dev_base_head.next); 3004 } 3005 3006 static inline struct net_device *first_net_device_rcu(struct net *net) 3007 { 3008 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 3009 3010 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3011 } 3012 3013 int netdev_boot_setup_check(struct net_device *dev); 3014 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 3015 const char *hwaddr); 3016 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 3017 void dev_add_pack(struct packet_type *pt); 3018 void dev_remove_pack(struct packet_type *pt); 3019 void __dev_remove_pack(struct packet_type *pt); 3020 void dev_add_offload(struct packet_offload *po); 3021 void dev_remove_offload(struct packet_offload *po); 3022 3023 int dev_get_iflink(const struct net_device *dev); 3024 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 3025 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 3026 struct net_device_path_stack *stack); 3027 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 3028 unsigned short mask); 3029 struct net_device *dev_get_by_name(struct net *net, const char *name); 3030 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 3031 struct net_device *__dev_get_by_name(struct net *net, const char *name); 3032 bool netdev_name_in_use(struct net *net, const char *name); 3033 int dev_alloc_name(struct net_device *dev, const char *name); 3034 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 3035 void dev_close(struct net_device *dev); 3036 void dev_close_many(struct list_head *head, bool unlink); 3037 void dev_disable_lro(struct net_device *dev); 3038 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 3039 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 3040 struct net_device *sb_dev); 3041 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 3042 struct net_device *sb_dev); 3043 3044 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3045 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3046 3047 static inline int dev_queue_xmit(struct sk_buff *skb) 3048 { 3049 return __dev_queue_xmit(skb, NULL); 3050 } 3051 3052 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3053 struct net_device *sb_dev) 3054 { 3055 return __dev_queue_xmit(skb, sb_dev); 3056 } 3057 3058 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3059 { 3060 int ret; 3061 3062 ret = __dev_direct_xmit(skb, queue_id); 3063 if (!dev_xmit_complete(ret)) 3064 kfree_skb(skb); 3065 return ret; 3066 } 3067 3068 int register_netdevice(struct net_device *dev); 3069 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3070 void unregister_netdevice_many(struct list_head *head); 3071 static inline void unregister_netdevice(struct net_device *dev) 3072 { 3073 unregister_netdevice_queue(dev, NULL); 3074 } 3075 3076 int netdev_refcnt_read(const struct net_device *dev); 3077 void free_netdev(struct net_device *dev); 3078 void netdev_freemem(struct net_device *dev); 3079 int init_dummy_netdev(struct net_device *dev); 3080 3081 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3082 struct sk_buff *skb, 3083 bool all_slaves); 3084 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3085 struct sock *sk); 3086 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3087 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3088 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3089 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 3090 int dev_restart(struct net_device *dev); 3091 3092 3093 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3094 unsigned short type, 3095 const void *daddr, const void *saddr, 3096 unsigned int len) 3097 { 3098 if (!dev->header_ops || !dev->header_ops->create) 3099 return 0; 3100 3101 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3102 } 3103 3104 static inline int dev_parse_header(const struct sk_buff *skb, 3105 unsigned char *haddr) 3106 { 3107 const struct net_device *dev = skb->dev; 3108 3109 if (!dev->header_ops || !dev->header_ops->parse) 3110 return 0; 3111 return dev->header_ops->parse(skb, haddr); 3112 } 3113 3114 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3115 { 3116 const struct net_device *dev = skb->dev; 3117 3118 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3119 return 0; 3120 return dev->header_ops->parse_protocol(skb); 3121 } 3122 3123 /* ll_header must have at least hard_header_len allocated */ 3124 static inline bool dev_validate_header(const struct net_device *dev, 3125 char *ll_header, int len) 3126 { 3127 if (likely(len >= dev->hard_header_len)) 3128 return true; 3129 if (len < dev->min_header_len) 3130 return false; 3131 3132 if (capable(CAP_SYS_RAWIO)) { 3133 memset(ll_header + len, 0, dev->hard_header_len - len); 3134 return true; 3135 } 3136 3137 if (dev->header_ops && dev->header_ops->validate) 3138 return dev->header_ops->validate(ll_header, len); 3139 3140 return false; 3141 } 3142 3143 static inline bool dev_has_header(const struct net_device *dev) 3144 { 3145 return dev->header_ops && dev->header_ops->create; 3146 } 3147 3148 /* 3149 * Incoming packets are placed on per-CPU queues 3150 */ 3151 struct softnet_data { 3152 struct list_head poll_list; 3153 struct sk_buff_head process_queue; 3154 3155 /* stats */ 3156 unsigned int processed; 3157 unsigned int time_squeeze; 3158 #ifdef CONFIG_RPS 3159 struct softnet_data *rps_ipi_list; 3160 #endif 3161 #ifdef CONFIG_NET_FLOW_LIMIT 3162 struct sd_flow_limit __rcu *flow_limit; 3163 #endif 3164 struct Qdisc *output_queue; 3165 struct Qdisc **output_queue_tailp; 3166 struct sk_buff *completion_queue; 3167 #ifdef CONFIG_XFRM_OFFLOAD 3168 struct sk_buff_head xfrm_backlog; 3169 #endif 3170 /* written and read only by owning cpu: */ 3171 struct { 3172 u16 recursion; 3173 u8 more; 3174 #ifdef CONFIG_NET_EGRESS 3175 u8 skip_txqueue; 3176 #endif 3177 } xmit; 3178 #ifdef CONFIG_RPS 3179 /* input_queue_head should be written by cpu owning this struct, 3180 * and only read by other cpus. Worth using a cache line. 3181 */ 3182 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3183 3184 /* Elements below can be accessed between CPUs for RPS/RFS */ 3185 call_single_data_t csd ____cacheline_aligned_in_smp; 3186 struct softnet_data *rps_ipi_next; 3187 unsigned int cpu; 3188 unsigned int input_queue_tail; 3189 #endif 3190 unsigned int received_rps; 3191 unsigned int dropped; 3192 struct sk_buff_head input_pkt_queue; 3193 struct napi_struct backlog; 3194 3195 /* Another possibly contended cache line */ 3196 spinlock_t defer_lock ____cacheline_aligned_in_smp; 3197 int defer_count; 3198 int defer_ipi_scheduled; 3199 struct sk_buff *defer_list; 3200 call_single_data_t defer_csd; 3201 }; 3202 3203 static inline void input_queue_head_incr(struct softnet_data *sd) 3204 { 3205 #ifdef CONFIG_RPS 3206 sd->input_queue_head++; 3207 #endif 3208 } 3209 3210 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3211 unsigned int *qtail) 3212 { 3213 #ifdef CONFIG_RPS 3214 *qtail = ++sd->input_queue_tail; 3215 #endif 3216 } 3217 3218 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3219 3220 static inline int dev_recursion_level(void) 3221 { 3222 return this_cpu_read(softnet_data.xmit.recursion); 3223 } 3224 3225 #define XMIT_RECURSION_LIMIT 8 3226 static inline bool dev_xmit_recursion(void) 3227 { 3228 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3229 XMIT_RECURSION_LIMIT); 3230 } 3231 3232 static inline void dev_xmit_recursion_inc(void) 3233 { 3234 __this_cpu_inc(softnet_data.xmit.recursion); 3235 } 3236 3237 static inline void dev_xmit_recursion_dec(void) 3238 { 3239 __this_cpu_dec(softnet_data.xmit.recursion); 3240 } 3241 3242 void __netif_schedule(struct Qdisc *q); 3243 void netif_schedule_queue(struct netdev_queue *txq); 3244 3245 static inline void netif_tx_schedule_all(struct net_device *dev) 3246 { 3247 unsigned int i; 3248 3249 for (i = 0; i < dev->num_tx_queues; i++) 3250 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3251 } 3252 3253 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3254 { 3255 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3256 } 3257 3258 /** 3259 * netif_start_queue - allow transmit 3260 * @dev: network device 3261 * 3262 * Allow upper layers to call the device hard_start_xmit routine. 3263 */ 3264 static inline void netif_start_queue(struct net_device *dev) 3265 { 3266 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3267 } 3268 3269 static inline void netif_tx_start_all_queues(struct net_device *dev) 3270 { 3271 unsigned int i; 3272 3273 for (i = 0; i < dev->num_tx_queues; i++) { 3274 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3275 netif_tx_start_queue(txq); 3276 } 3277 } 3278 3279 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3280 3281 /** 3282 * netif_wake_queue - restart transmit 3283 * @dev: network device 3284 * 3285 * Allow upper layers to call the device hard_start_xmit routine. 3286 * Used for flow control when transmit resources are available. 3287 */ 3288 static inline void netif_wake_queue(struct net_device *dev) 3289 { 3290 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3291 } 3292 3293 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3294 { 3295 unsigned int i; 3296 3297 for (i = 0; i < dev->num_tx_queues; i++) { 3298 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3299 netif_tx_wake_queue(txq); 3300 } 3301 } 3302 3303 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3304 { 3305 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3306 } 3307 3308 /** 3309 * netif_stop_queue - stop transmitted packets 3310 * @dev: network device 3311 * 3312 * Stop upper layers calling the device hard_start_xmit routine. 3313 * Used for flow control when transmit resources are unavailable. 3314 */ 3315 static inline void netif_stop_queue(struct net_device *dev) 3316 { 3317 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3318 } 3319 3320 void netif_tx_stop_all_queues(struct net_device *dev); 3321 3322 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3323 { 3324 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3325 } 3326 3327 /** 3328 * netif_queue_stopped - test if transmit queue is flowblocked 3329 * @dev: network device 3330 * 3331 * Test if transmit queue on device is currently unable to send. 3332 */ 3333 static inline bool netif_queue_stopped(const struct net_device *dev) 3334 { 3335 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3336 } 3337 3338 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3339 { 3340 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3341 } 3342 3343 static inline bool 3344 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3345 { 3346 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3347 } 3348 3349 static inline bool 3350 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3351 { 3352 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3353 } 3354 3355 /** 3356 * netdev_queue_set_dql_min_limit - set dql minimum limit 3357 * @dev_queue: pointer to transmit queue 3358 * @min_limit: dql minimum limit 3359 * 3360 * Forces xmit_more() to return true until the minimum threshold 3361 * defined by @min_limit is reached (or until the tx queue is 3362 * empty). Warning: to be use with care, misuse will impact the 3363 * latency. 3364 */ 3365 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3366 unsigned int min_limit) 3367 { 3368 #ifdef CONFIG_BQL 3369 dev_queue->dql.min_limit = min_limit; 3370 #endif 3371 } 3372 3373 /** 3374 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3375 * @dev_queue: pointer to transmit queue 3376 * 3377 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3378 * to give appropriate hint to the CPU. 3379 */ 3380 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3381 { 3382 #ifdef CONFIG_BQL 3383 prefetchw(&dev_queue->dql.num_queued); 3384 #endif 3385 } 3386 3387 /** 3388 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3389 * @dev_queue: pointer to transmit queue 3390 * 3391 * BQL enabled drivers might use this helper in their TX completion path, 3392 * to give appropriate hint to the CPU. 3393 */ 3394 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3395 { 3396 #ifdef CONFIG_BQL 3397 prefetchw(&dev_queue->dql.limit); 3398 #endif 3399 } 3400 3401 /** 3402 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3403 * @dev_queue: network device queue 3404 * @bytes: number of bytes queued to the device queue 3405 * 3406 * Report the number of bytes queued for sending/completion to the network 3407 * device hardware queue. @bytes should be a good approximation and should 3408 * exactly match netdev_completed_queue() @bytes. 3409 * This is typically called once per packet, from ndo_start_xmit(). 3410 */ 3411 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3412 unsigned int bytes) 3413 { 3414 #ifdef CONFIG_BQL 3415 dql_queued(&dev_queue->dql, bytes); 3416 3417 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3418 return; 3419 3420 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3421 3422 /* 3423 * The XOFF flag must be set before checking the dql_avail below, 3424 * because in netdev_tx_completed_queue we update the dql_completed 3425 * before checking the XOFF flag. 3426 */ 3427 smp_mb(); 3428 3429 /* check again in case another CPU has just made room avail */ 3430 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3431 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3432 #endif 3433 } 3434 3435 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3436 * that they should not test BQL status themselves. 3437 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3438 * skb of a batch. 3439 * Returns true if the doorbell must be used to kick the NIC. 3440 */ 3441 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3442 unsigned int bytes, 3443 bool xmit_more) 3444 { 3445 if (xmit_more) { 3446 #ifdef CONFIG_BQL 3447 dql_queued(&dev_queue->dql, bytes); 3448 #endif 3449 return netif_tx_queue_stopped(dev_queue); 3450 } 3451 netdev_tx_sent_queue(dev_queue, bytes); 3452 return true; 3453 } 3454 3455 /** 3456 * netdev_sent_queue - report the number of bytes queued to hardware 3457 * @dev: network device 3458 * @bytes: number of bytes queued to the hardware device queue 3459 * 3460 * Report the number of bytes queued for sending/completion to the network 3461 * device hardware queue#0. @bytes should be a good approximation and should 3462 * exactly match netdev_completed_queue() @bytes. 3463 * This is typically called once per packet, from ndo_start_xmit(). 3464 */ 3465 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3466 { 3467 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3468 } 3469 3470 static inline bool __netdev_sent_queue(struct net_device *dev, 3471 unsigned int bytes, 3472 bool xmit_more) 3473 { 3474 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3475 xmit_more); 3476 } 3477 3478 /** 3479 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3480 * @dev_queue: network device queue 3481 * @pkts: number of packets (currently ignored) 3482 * @bytes: number of bytes dequeued from the device queue 3483 * 3484 * Must be called at most once per TX completion round (and not per 3485 * individual packet), so that BQL can adjust its limits appropriately. 3486 */ 3487 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3488 unsigned int pkts, unsigned int bytes) 3489 { 3490 #ifdef CONFIG_BQL 3491 if (unlikely(!bytes)) 3492 return; 3493 3494 dql_completed(&dev_queue->dql, bytes); 3495 3496 /* 3497 * Without the memory barrier there is a small possiblity that 3498 * netdev_tx_sent_queue will miss the update and cause the queue to 3499 * be stopped forever 3500 */ 3501 smp_mb(); 3502 3503 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3504 return; 3505 3506 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3507 netif_schedule_queue(dev_queue); 3508 #endif 3509 } 3510 3511 /** 3512 * netdev_completed_queue - report bytes and packets completed by device 3513 * @dev: network device 3514 * @pkts: actual number of packets sent over the medium 3515 * @bytes: actual number of bytes sent over the medium 3516 * 3517 * Report the number of bytes and packets transmitted by the network device 3518 * hardware queue over the physical medium, @bytes must exactly match the 3519 * @bytes amount passed to netdev_sent_queue() 3520 */ 3521 static inline void netdev_completed_queue(struct net_device *dev, 3522 unsigned int pkts, unsigned int bytes) 3523 { 3524 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3525 } 3526 3527 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3528 { 3529 #ifdef CONFIG_BQL 3530 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3531 dql_reset(&q->dql); 3532 #endif 3533 } 3534 3535 /** 3536 * netdev_reset_queue - reset the packets and bytes count of a network device 3537 * @dev_queue: network device 3538 * 3539 * Reset the bytes and packet count of a network device and clear the 3540 * software flow control OFF bit for this network device 3541 */ 3542 static inline void netdev_reset_queue(struct net_device *dev_queue) 3543 { 3544 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3545 } 3546 3547 /** 3548 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3549 * @dev: network device 3550 * @queue_index: given tx queue index 3551 * 3552 * Returns 0 if given tx queue index >= number of device tx queues, 3553 * otherwise returns the originally passed tx queue index. 3554 */ 3555 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3556 { 3557 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3558 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3559 dev->name, queue_index, 3560 dev->real_num_tx_queues); 3561 return 0; 3562 } 3563 3564 return queue_index; 3565 } 3566 3567 /** 3568 * netif_running - test if up 3569 * @dev: network device 3570 * 3571 * Test if the device has been brought up. 3572 */ 3573 static inline bool netif_running(const struct net_device *dev) 3574 { 3575 return test_bit(__LINK_STATE_START, &dev->state); 3576 } 3577 3578 /* 3579 * Routines to manage the subqueues on a device. We only need start, 3580 * stop, and a check if it's stopped. All other device management is 3581 * done at the overall netdevice level. 3582 * Also test the device if we're multiqueue. 3583 */ 3584 3585 /** 3586 * netif_start_subqueue - allow sending packets on subqueue 3587 * @dev: network device 3588 * @queue_index: sub queue index 3589 * 3590 * Start individual transmit queue of a device with multiple transmit queues. 3591 */ 3592 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3593 { 3594 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3595 3596 netif_tx_start_queue(txq); 3597 } 3598 3599 /** 3600 * netif_stop_subqueue - stop sending packets on subqueue 3601 * @dev: network device 3602 * @queue_index: sub queue index 3603 * 3604 * Stop individual transmit queue of a device with multiple transmit queues. 3605 */ 3606 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3607 { 3608 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3609 netif_tx_stop_queue(txq); 3610 } 3611 3612 /** 3613 * __netif_subqueue_stopped - test status of subqueue 3614 * @dev: network device 3615 * @queue_index: sub queue index 3616 * 3617 * Check individual transmit queue of a device with multiple transmit queues. 3618 */ 3619 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3620 u16 queue_index) 3621 { 3622 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3623 3624 return netif_tx_queue_stopped(txq); 3625 } 3626 3627 /** 3628 * netif_subqueue_stopped - test status of subqueue 3629 * @dev: network device 3630 * @skb: sub queue buffer pointer 3631 * 3632 * Check individual transmit queue of a device with multiple transmit queues. 3633 */ 3634 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3635 struct sk_buff *skb) 3636 { 3637 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3638 } 3639 3640 /** 3641 * netif_wake_subqueue - allow sending packets on subqueue 3642 * @dev: network device 3643 * @queue_index: sub queue index 3644 * 3645 * Resume individual transmit queue of a device with multiple transmit queues. 3646 */ 3647 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3648 { 3649 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3650 3651 netif_tx_wake_queue(txq); 3652 } 3653 3654 #ifdef CONFIG_XPS 3655 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3656 u16 index); 3657 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3658 u16 index, enum xps_map_type type); 3659 3660 /** 3661 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3662 * @j: CPU/Rx queue index 3663 * @mask: bitmask of all cpus/rx queues 3664 * @nr_bits: number of bits in the bitmask 3665 * 3666 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3667 */ 3668 static inline bool netif_attr_test_mask(unsigned long j, 3669 const unsigned long *mask, 3670 unsigned int nr_bits) 3671 { 3672 cpu_max_bits_warn(j, nr_bits); 3673 return test_bit(j, mask); 3674 } 3675 3676 /** 3677 * netif_attr_test_online - Test for online CPU/Rx queue 3678 * @j: CPU/Rx queue index 3679 * @online_mask: bitmask for CPUs/Rx queues that are online 3680 * @nr_bits: number of bits in the bitmask 3681 * 3682 * Returns true if a CPU/Rx queue is online. 3683 */ 3684 static inline bool netif_attr_test_online(unsigned long j, 3685 const unsigned long *online_mask, 3686 unsigned int nr_bits) 3687 { 3688 cpu_max_bits_warn(j, nr_bits); 3689 3690 if (online_mask) 3691 return test_bit(j, online_mask); 3692 3693 return (j < nr_bits); 3694 } 3695 3696 /** 3697 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3698 * @n: CPU/Rx queue index 3699 * @srcp: the cpumask/Rx queue mask pointer 3700 * @nr_bits: number of bits in the bitmask 3701 * 3702 * Returns >= nr_bits if no further CPUs/Rx queues set. 3703 */ 3704 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3705 unsigned int nr_bits) 3706 { 3707 /* -1 is a legal arg here. */ 3708 if (n != -1) 3709 cpu_max_bits_warn(n, nr_bits); 3710 3711 if (srcp) 3712 return find_next_bit(srcp, nr_bits, n + 1); 3713 3714 return n + 1; 3715 } 3716 3717 /** 3718 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3719 * @n: CPU/Rx queue index 3720 * @src1p: the first CPUs/Rx queues mask pointer 3721 * @src2p: the second CPUs/Rx queues mask pointer 3722 * @nr_bits: number of bits in the bitmask 3723 * 3724 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3725 */ 3726 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3727 const unsigned long *src2p, 3728 unsigned int nr_bits) 3729 { 3730 /* -1 is a legal arg here. */ 3731 if (n != -1) 3732 cpu_max_bits_warn(n, nr_bits); 3733 3734 if (src1p && src2p) 3735 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3736 else if (src1p) 3737 return find_next_bit(src1p, nr_bits, n + 1); 3738 else if (src2p) 3739 return find_next_bit(src2p, nr_bits, n + 1); 3740 3741 return n + 1; 3742 } 3743 #else 3744 static inline int netif_set_xps_queue(struct net_device *dev, 3745 const struct cpumask *mask, 3746 u16 index) 3747 { 3748 return 0; 3749 } 3750 3751 static inline int __netif_set_xps_queue(struct net_device *dev, 3752 const unsigned long *mask, 3753 u16 index, enum xps_map_type type) 3754 { 3755 return 0; 3756 } 3757 #endif 3758 3759 /** 3760 * netif_is_multiqueue - test if device has multiple transmit queues 3761 * @dev: network device 3762 * 3763 * Check if device has multiple transmit queues 3764 */ 3765 static inline bool netif_is_multiqueue(const struct net_device *dev) 3766 { 3767 return dev->num_tx_queues > 1; 3768 } 3769 3770 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3771 3772 #ifdef CONFIG_SYSFS 3773 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3774 #else 3775 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3776 unsigned int rxqs) 3777 { 3778 dev->real_num_rx_queues = rxqs; 3779 return 0; 3780 } 3781 #endif 3782 int netif_set_real_num_queues(struct net_device *dev, 3783 unsigned int txq, unsigned int rxq); 3784 3785 static inline struct netdev_rx_queue * 3786 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3787 { 3788 return dev->_rx + rxq; 3789 } 3790 3791 #ifdef CONFIG_SYSFS 3792 static inline unsigned int get_netdev_rx_queue_index( 3793 struct netdev_rx_queue *queue) 3794 { 3795 struct net_device *dev = queue->dev; 3796 int index = queue - dev->_rx; 3797 3798 BUG_ON(index >= dev->num_rx_queues); 3799 return index; 3800 } 3801 #endif 3802 3803 int netif_get_num_default_rss_queues(void); 3804 3805 enum skb_free_reason { 3806 SKB_REASON_CONSUMED, 3807 SKB_REASON_DROPPED, 3808 }; 3809 3810 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3811 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3812 3813 /* 3814 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3815 * interrupt context or with hardware interrupts being disabled. 3816 * (in_hardirq() || irqs_disabled()) 3817 * 3818 * We provide four helpers that can be used in following contexts : 3819 * 3820 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3821 * replacing kfree_skb(skb) 3822 * 3823 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3824 * Typically used in place of consume_skb(skb) in TX completion path 3825 * 3826 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3827 * replacing kfree_skb(skb) 3828 * 3829 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3830 * and consumed a packet. Used in place of consume_skb(skb) 3831 */ 3832 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3833 { 3834 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3835 } 3836 3837 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3838 { 3839 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3840 } 3841 3842 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3843 { 3844 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3845 } 3846 3847 static inline void dev_consume_skb_any(struct sk_buff *skb) 3848 { 3849 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3850 } 3851 3852 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3853 struct bpf_prog *xdp_prog); 3854 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3855 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3856 int netif_rx(struct sk_buff *skb); 3857 int __netif_rx(struct sk_buff *skb); 3858 3859 int netif_receive_skb(struct sk_buff *skb); 3860 int netif_receive_skb_core(struct sk_buff *skb); 3861 void netif_receive_skb_list_internal(struct list_head *head); 3862 void netif_receive_skb_list(struct list_head *head); 3863 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3864 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3865 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3866 void napi_get_frags_check(struct napi_struct *napi); 3867 gro_result_t napi_gro_frags(struct napi_struct *napi); 3868 struct packet_offload *gro_find_receive_by_type(__be16 type); 3869 struct packet_offload *gro_find_complete_by_type(__be16 type); 3870 3871 static inline void napi_free_frags(struct napi_struct *napi) 3872 { 3873 kfree_skb(napi->skb); 3874 napi->skb = NULL; 3875 } 3876 3877 bool netdev_is_rx_handler_busy(struct net_device *dev); 3878 int netdev_rx_handler_register(struct net_device *dev, 3879 rx_handler_func_t *rx_handler, 3880 void *rx_handler_data); 3881 void netdev_rx_handler_unregister(struct net_device *dev); 3882 3883 bool dev_valid_name(const char *name); 3884 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 3885 { 3886 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 3887 } 3888 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 3889 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 3890 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3891 void __user *data, bool *need_copyout); 3892 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 3893 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 3894 unsigned int dev_get_flags(const struct net_device *); 3895 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3896 struct netlink_ext_ack *extack); 3897 int dev_change_flags(struct net_device *dev, unsigned int flags, 3898 struct netlink_ext_ack *extack); 3899 int dev_set_alias(struct net_device *, const char *, size_t); 3900 int dev_get_alias(const struct net_device *, char *, size_t); 3901 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 3902 const char *pat, int new_ifindex); 3903 static inline 3904 int dev_change_net_namespace(struct net_device *dev, struct net *net, 3905 const char *pat) 3906 { 3907 return __dev_change_net_namespace(dev, net, pat, 0); 3908 } 3909 int __dev_set_mtu(struct net_device *, int); 3910 int dev_set_mtu(struct net_device *, int); 3911 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3912 struct netlink_ext_ack *extack); 3913 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3914 struct netlink_ext_ack *extack); 3915 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 3916 struct netlink_ext_ack *extack); 3917 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 3918 int dev_get_port_parent_id(struct net_device *dev, 3919 struct netdev_phys_item_id *ppid, bool recurse); 3920 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3921 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3922 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3923 struct netdev_queue *txq, int *ret); 3924 3925 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3926 u8 dev_xdp_prog_count(struct net_device *dev); 3927 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3928 3929 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3930 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3931 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 3932 bool is_skb_forwardable(const struct net_device *dev, 3933 const struct sk_buff *skb); 3934 3935 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 3936 const struct sk_buff *skb, 3937 const bool check_mtu) 3938 { 3939 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 3940 unsigned int len; 3941 3942 if (!(dev->flags & IFF_UP)) 3943 return false; 3944 3945 if (!check_mtu) 3946 return true; 3947 3948 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 3949 if (skb->len <= len) 3950 return true; 3951 3952 /* if TSO is enabled, we don't care about the length as the packet 3953 * could be forwarded without being segmented before 3954 */ 3955 if (skb_is_gso(skb)) 3956 return true; 3957 3958 return false; 3959 } 3960 3961 struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev); 3962 3963 static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev) 3964 { 3965 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */ 3966 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats); 3967 3968 if (likely(p)) 3969 return p; 3970 3971 return netdev_core_stats_alloc(dev); 3972 } 3973 3974 #define DEV_CORE_STATS_INC(FIELD) \ 3975 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 3976 { \ 3977 struct net_device_core_stats __percpu *p; \ 3978 \ 3979 p = dev_core_stats(dev); \ 3980 if (p) \ 3981 this_cpu_inc(p->FIELD); \ 3982 } 3983 DEV_CORE_STATS_INC(rx_dropped) 3984 DEV_CORE_STATS_INC(tx_dropped) 3985 DEV_CORE_STATS_INC(rx_nohandler) 3986 DEV_CORE_STATS_INC(rx_otherhost_dropped) 3987 3988 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3989 struct sk_buff *skb, 3990 const bool check_mtu) 3991 { 3992 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3993 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 3994 dev_core_stats_rx_dropped_inc(dev); 3995 kfree_skb(skb); 3996 return NET_RX_DROP; 3997 } 3998 3999 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 4000 skb->priority = 0; 4001 return 0; 4002 } 4003 4004 bool dev_nit_active(struct net_device *dev); 4005 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 4006 4007 static inline void __dev_put(struct net_device *dev) 4008 { 4009 if (dev) { 4010 #ifdef CONFIG_PCPU_DEV_REFCNT 4011 this_cpu_dec(*dev->pcpu_refcnt); 4012 #else 4013 refcount_dec(&dev->dev_refcnt); 4014 #endif 4015 } 4016 } 4017 4018 static inline void __dev_hold(struct net_device *dev) 4019 { 4020 if (dev) { 4021 #ifdef CONFIG_PCPU_DEV_REFCNT 4022 this_cpu_inc(*dev->pcpu_refcnt); 4023 #else 4024 refcount_inc(&dev->dev_refcnt); 4025 #endif 4026 } 4027 } 4028 4029 static inline void __netdev_tracker_alloc(struct net_device *dev, 4030 netdevice_tracker *tracker, 4031 gfp_t gfp) 4032 { 4033 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4034 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 4035 #endif 4036 } 4037 4038 /* netdev_tracker_alloc() can upgrade a prior untracked reference 4039 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4040 */ 4041 static inline void netdev_tracker_alloc(struct net_device *dev, 4042 netdevice_tracker *tracker, gfp_t gfp) 4043 { 4044 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4045 refcount_dec(&dev->refcnt_tracker.no_tracker); 4046 __netdev_tracker_alloc(dev, tracker, gfp); 4047 #endif 4048 } 4049 4050 static inline void netdev_tracker_free(struct net_device *dev, 4051 netdevice_tracker *tracker) 4052 { 4053 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4054 ref_tracker_free(&dev->refcnt_tracker, tracker); 4055 #endif 4056 } 4057 4058 static inline void netdev_hold(struct net_device *dev, 4059 netdevice_tracker *tracker, gfp_t gfp) 4060 { 4061 if (dev) { 4062 __dev_hold(dev); 4063 __netdev_tracker_alloc(dev, tracker, gfp); 4064 } 4065 } 4066 4067 static inline void netdev_put(struct net_device *dev, 4068 netdevice_tracker *tracker) 4069 { 4070 if (dev) { 4071 netdev_tracker_free(dev, tracker); 4072 __dev_put(dev); 4073 } 4074 } 4075 4076 /** 4077 * dev_hold - get reference to device 4078 * @dev: network device 4079 * 4080 * Hold reference to device to keep it from being freed. 4081 * Try using netdev_hold() instead. 4082 */ 4083 static inline void dev_hold(struct net_device *dev) 4084 { 4085 netdev_hold(dev, NULL, GFP_ATOMIC); 4086 } 4087 4088 /** 4089 * dev_put - release reference to device 4090 * @dev: network device 4091 * 4092 * Release reference to device to allow it to be freed. 4093 * Try using netdev_put() instead. 4094 */ 4095 static inline void dev_put(struct net_device *dev) 4096 { 4097 netdev_put(dev, NULL); 4098 } 4099 4100 static inline void netdev_ref_replace(struct net_device *odev, 4101 struct net_device *ndev, 4102 netdevice_tracker *tracker, 4103 gfp_t gfp) 4104 { 4105 if (odev) 4106 netdev_tracker_free(odev, tracker); 4107 4108 __dev_hold(ndev); 4109 __dev_put(odev); 4110 4111 if (ndev) 4112 __netdev_tracker_alloc(ndev, tracker, gfp); 4113 } 4114 4115 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4116 * and _off may be called from IRQ context, but it is caller 4117 * who is responsible for serialization of these calls. 4118 * 4119 * The name carrier is inappropriate, these functions should really be 4120 * called netif_lowerlayer_*() because they represent the state of any 4121 * kind of lower layer not just hardware media. 4122 */ 4123 void linkwatch_fire_event(struct net_device *dev); 4124 4125 /** 4126 * netif_carrier_ok - test if carrier present 4127 * @dev: network device 4128 * 4129 * Check if carrier is present on device 4130 */ 4131 static inline bool netif_carrier_ok(const struct net_device *dev) 4132 { 4133 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4134 } 4135 4136 unsigned long dev_trans_start(struct net_device *dev); 4137 4138 void __netdev_watchdog_up(struct net_device *dev); 4139 4140 void netif_carrier_on(struct net_device *dev); 4141 void netif_carrier_off(struct net_device *dev); 4142 void netif_carrier_event(struct net_device *dev); 4143 4144 /** 4145 * netif_dormant_on - mark device as dormant. 4146 * @dev: network device 4147 * 4148 * Mark device as dormant (as per RFC2863). 4149 * 4150 * The dormant state indicates that the relevant interface is not 4151 * actually in a condition to pass packets (i.e., it is not 'up') but is 4152 * in a "pending" state, waiting for some external event. For "on- 4153 * demand" interfaces, this new state identifies the situation where the 4154 * interface is waiting for events to place it in the up state. 4155 */ 4156 static inline void netif_dormant_on(struct net_device *dev) 4157 { 4158 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4159 linkwatch_fire_event(dev); 4160 } 4161 4162 /** 4163 * netif_dormant_off - set device as not dormant. 4164 * @dev: network device 4165 * 4166 * Device is not in dormant state. 4167 */ 4168 static inline void netif_dormant_off(struct net_device *dev) 4169 { 4170 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4171 linkwatch_fire_event(dev); 4172 } 4173 4174 /** 4175 * netif_dormant - test if device is dormant 4176 * @dev: network device 4177 * 4178 * Check if device is dormant. 4179 */ 4180 static inline bool netif_dormant(const struct net_device *dev) 4181 { 4182 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4183 } 4184 4185 4186 /** 4187 * netif_testing_on - mark device as under test. 4188 * @dev: network device 4189 * 4190 * Mark device as under test (as per RFC2863). 4191 * 4192 * The testing state indicates that some test(s) must be performed on 4193 * the interface. After completion, of the test, the interface state 4194 * will change to up, dormant, or down, as appropriate. 4195 */ 4196 static inline void netif_testing_on(struct net_device *dev) 4197 { 4198 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4199 linkwatch_fire_event(dev); 4200 } 4201 4202 /** 4203 * netif_testing_off - set device as not under test. 4204 * @dev: network device 4205 * 4206 * Device is not in testing state. 4207 */ 4208 static inline void netif_testing_off(struct net_device *dev) 4209 { 4210 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4211 linkwatch_fire_event(dev); 4212 } 4213 4214 /** 4215 * netif_testing - test if device is under test 4216 * @dev: network device 4217 * 4218 * Check if device is under test 4219 */ 4220 static inline bool netif_testing(const struct net_device *dev) 4221 { 4222 return test_bit(__LINK_STATE_TESTING, &dev->state); 4223 } 4224 4225 4226 /** 4227 * netif_oper_up - test if device is operational 4228 * @dev: network device 4229 * 4230 * Check if carrier is operational 4231 */ 4232 static inline bool netif_oper_up(const struct net_device *dev) 4233 { 4234 return (dev->operstate == IF_OPER_UP || 4235 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4236 } 4237 4238 /** 4239 * netif_device_present - is device available or removed 4240 * @dev: network device 4241 * 4242 * Check if device has not been removed from system. 4243 */ 4244 static inline bool netif_device_present(const struct net_device *dev) 4245 { 4246 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4247 } 4248 4249 void netif_device_detach(struct net_device *dev); 4250 4251 void netif_device_attach(struct net_device *dev); 4252 4253 /* 4254 * Network interface message level settings 4255 */ 4256 4257 enum { 4258 NETIF_MSG_DRV_BIT, 4259 NETIF_MSG_PROBE_BIT, 4260 NETIF_MSG_LINK_BIT, 4261 NETIF_MSG_TIMER_BIT, 4262 NETIF_MSG_IFDOWN_BIT, 4263 NETIF_MSG_IFUP_BIT, 4264 NETIF_MSG_RX_ERR_BIT, 4265 NETIF_MSG_TX_ERR_BIT, 4266 NETIF_MSG_TX_QUEUED_BIT, 4267 NETIF_MSG_INTR_BIT, 4268 NETIF_MSG_TX_DONE_BIT, 4269 NETIF_MSG_RX_STATUS_BIT, 4270 NETIF_MSG_PKTDATA_BIT, 4271 NETIF_MSG_HW_BIT, 4272 NETIF_MSG_WOL_BIT, 4273 4274 /* When you add a new bit above, update netif_msg_class_names array 4275 * in net/ethtool/common.c 4276 */ 4277 NETIF_MSG_CLASS_COUNT, 4278 }; 4279 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4280 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4281 4282 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4283 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4284 4285 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4286 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4287 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4288 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4289 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4290 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4291 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4292 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4293 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4294 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4295 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4296 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4297 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4298 #define NETIF_MSG_HW __NETIF_MSG(HW) 4299 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4300 4301 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4302 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4303 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4304 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4305 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4306 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4307 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4308 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4309 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4310 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4311 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4312 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4313 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4314 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4315 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4316 4317 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4318 { 4319 /* use default */ 4320 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4321 return default_msg_enable_bits; 4322 if (debug_value == 0) /* no output */ 4323 return 0; 4324 /* set low N bits */ 4325 return (1U << debug_value) - 1; 4326 } 4327 4328 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4329 { 4330 spin_lock(&txq->_xmit_lock); 4331 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4332 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4333 } 4334 4335 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4336 { 4337 __acquire(&txq->_xmit_lock); 4338 return true; 4339 } 4340 4341 static inline void __netif_tx_release(struct netdev_queue *txq) 4342 { 4343 __release(&txq->_xmit_lock); 4344 } 4345 4346 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4347 { 4348 spin_lock_bh(&txq->_xmit_lock); 4349 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4350 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4351 } 4352 4353 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4354 { 4355 bool ok = spin_trylock(&txq->_xmit_lock); 4356 4357 if (likely(ok)) { 4358 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4359 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4360 } 4361 return ok; 4362 } 4363 4364 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4365 { 4366 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4367 WRITE_ONCE(txq->xmit_lock_owner, -1); 4368 spin_unlock(&txq->_xmit_lock); 4369 } 4370 4371 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4372 { 4373 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4374 WRITE_ONCE(txq->xmit_lock_owner, -1); 4375 spin_unlock_bh(&txq->_xmit_lock); 4376 } 4377 4378 /* 4379 * txq->trans_start can be read locklessly from dev_watchdog() 4380 */ 4381 static inline void txq_trans_update(struct netdev_queue *txq) 4382 { 4383 if (txq->xmit_lock_owner != -1) 4384 WRITE_ONCE(txq->trans_start, jiffies); 4385 } 4386 4387 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4388 { 4389 unsigned long now = jiffies; 4390 4391 if (READ_ONCE(txq->trans_start) != now) 4392 WRITE_ONCE(txq->trans_start, now); 4393 } 4394 4395 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4396 static inline void netif_trans_update(struct net_device *dev) 4397 { 4398 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4399 4400 txq_trans_cond_update(txq); 4401 } 4402 4403 /** 4404 * netif_tx_lock - grab network device transmit lock 4405 * @dev: network device 4406 * 4407 * Get network device transmit lock 4408 */ 4409 void netif_tx_lock(struct net_device *dev); 4410 4411 static inline void netif_tx_lock_bh(struct net_device *dev) 4412 { 4413 local_bh_disable(); 4414 netif_tx_lock(dev); 4415 } 4416 4417 void netif_tx_unlock(struct net_device *dev); 4418 4419 static inline void netif_tx_unlock_bh(struct net_device *dev) 4420 { 4421 netif_tx_unlock(dev); 4422 local_bh_enable(); 4423 } 4424 4425 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4426 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4427 __netif_tx_lock(txq, cpu); \ 4428 } else { \ 4429 __netif_tx_acquire(txq); \ 4430 } \ 4431 } 4432 4433 #define HARD_TX_TRYLOCK(dev, txq) \ 4434 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4435 __netif_tx_trylock(txq) : \ 4436 __netif_tx_acquire(txq)) 4437 4438 #define HARD_TX_UNLOCK(dev, txq) { \ 4439 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4440 __netif_tx_unlock(txq); \ 4441 } else { \ 4442 __netif_tx_release(txq); \ 4443 } \ 4444 } 4445 4446 static inline void netif_tx_disable(struct net_device *dev) 4447 { 4448 unsigned int i; 4449 int cpu; 4450 4451 local_bh_disable(); 4452 cpu = smp_processor_id(); 4453 spin_lock(&dev->tx_global_lock); 4454 for (i = 0; i < dev->num_tx_queues; i++) { 4455 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4456 4457 __netif_tx_lock(txq, cpu); 4458 netif_tx_stop_queue(txq); 4459 __netif_tx_unlock(txq); 4460 } 4461 spin_unlock(&dev->tx_global_lock); 4462 local_bh_enable(); 4463 } 4464 4465 static inline void netif_addr_lock(struct net_device *dev) 4466 { 4467 unsigned char nest_level = 0; 4468 4469 #ifdef CONFIG_LOCKDEP 4470 nest_level = dev->nested_level; 4471 #endif 4472 spin_lock_nested(&dev->addr_list_lock, nest_level); 4473 } 4474 4475 static inline void netif_addr_lock_bh(struct net_device *dev) 4476 { 4477 unsigned char nest_level = 0; 4478 4479 #ifdef CONFIG_LOCKDEP 4480 nest_level = dev->nested_level; 4481 #endif 4482 local_bh_disable(); 4483 spin_lock_nested(&dev->addr_list_lock, nest_level); 4484 } 4485 4486 static inline void netif_addr_unlock(struct net_device *dev) 4487 { 4488 spin_unlock(&dev->addr_list_lock); 4489 } 4490 4491 static inline void netif_addr_unlock_bh(struct net_device *dev) 4492 { 4493 spin_unlock_bh(&dev->addr_list_lock); 4494 } 4495 4496 /* 4497 * dev_addrs walker. Should be used only for read access. Call with 4498 * rcu_read_lock held. 4499 */ 4500 #define for_each_dev_addr(dev, ha) \ 4501 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4502 4503 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4504 4505 void ether_setup(struct net_device *dev); 4506 4507 /* Support for loadable net-drivers */ 4508 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4509 unsigned char name_assign_type, 4510 void (*setup)(struct net_device *), 4511 unsigned int txqs, unsigned int rxqs); 4512 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4513 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4514 4515 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4516 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4517 count) 4518 4519 int register_netdev(struct net_device *dev); 4520 void unregister_netdev(struct net_device *dev); 4521 4522 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4523 4524 /* General hardware address lists handling functions */ 4525 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4526 struct netdev_hw_addr_list *from_list, int addr_len); 4527 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4528 struct netdev_hw_addr_list *from_list, int addr_len); 4529 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4530 struct net_device *dev, 4531 int (*sync)(struct net_device *, const unsigned char *), 4532 int (*unsync)(struct net_device *, 4533 const unsigned char *)); 4534 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4535 struct net_device *dev, 4536 int (*sync)(struct net_device *, 4537 const unsigned char *, int), 4538 int (*unsync)(struct net_device *, 4539 const unsigned char *, int)); 4540 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4541 struct net_device *dev, 4542 int (*unsync)(struct net_device *, 4543 const unsigned char *, int)); 4544 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4545 struct net_device *dev, 4546 int (*unsync)(struct net_device *, 4547 const unsigned char *)); 4548 void __hw_addr_init(struct netdev_hw_addr_list *list); 4549 4550 /* Functions used for device addresses handling */ 4551 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4552 const void *addr, size_t len); 4553 4554 static inline void 4555 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4556 { 4557 dev_addr_mod(dev, 0, addr, len); 4558 } 4559 4560 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4561 { 4562 __dev_addr_set(dev, addr, dev->addr_len); 4563 } 4564 4565 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4566 unsigned char addr_type); 4567 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4568 unsigned char addr_type); 4569 4570 /* Functions used for unicast addresses handling */ 4571 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4572 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4573 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4574 int dev_uc_sync(struct net_device *to, struct net_device *from); 4575 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4576 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4577 void dev_uc_flush(struct net_device *dev); 4578 void dev_uc_init(struct net_device *dev); 4579 4580 /** 4581 * __dev_uc_sync - Synchonize device's unicast list 4582 * @dev: device to sync 4583 * @sync: function to call if address should be added 4584 * @unsync: function to call if address should be removed 4585 * 4586 * Add newly added addresses to the interface, and release 4587 * addresses that have been deleted. 4588 */ 4589 static inline int __dev_uc_sync(struct net_device *dev, 4590 int (*sync)(struct net_device *, 4591 const unsigned char *), 4592 int (*unsync)(struct net_device *, 4593 const unsigned char *)) 4594 { 4595 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4596 } 4597 4598 /** 4599 * __dev_uc_unsync - Remove synchronized addresses from device 4600 * @dev: device to sync 4601 * @unsync: function to call if address should be removed 4602 * 4603 * Remove all addresses that were added to the device by dev_uc_sync(). 4604 */ 4605 static inline void __dev_uc_unsync(struct net_device *dev, 4606 int (*unsync)(struct net_device *, 4607 const unsigned char *)) 4608 { 4609 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4610 } 4611 4612 /* Functions used for multicast addresses handling */ 4613 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4614 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4615 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4616 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4617 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4618 int dev_mc_sync(struct net_device *to, struct net_device *from); 4619 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4620 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4621 void dev_mc_flush(struct net_device *dev); 4622 void dev_mc_init(struct net_device *dev); 4623 4624 /** 4625 * __dev_mc_sync - Synchonize device's multicast list 4626 * @dev: device to sync 4627 * @sync: function to call if address should be added 4628 * @unsync: function to call if address should be removed 4629 * 4630 * Add newly added addresses to the interface, and release 4631 * addresses that have been deleted. 4632 */ 4633 static inline int __dev_mc_sync(struct net_device *dev, 4634 int (*sync)(struct net_device *, 4635 const unsigned char *), 4636 int (*unsync)(struct net_device *, 4637 const unsigned char *)) 4638 { 4639 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4640 } 4641 4642 /** 4643 * __dev_mc_unsync - Remove synchronized addresses from device 4644 * @dev: device to sync 4645 * @unsync: function to call if address should be removed 4646 * 4647 * Remove all addresses that were added to the device by dev_mc_sync(). 4648 */ 4649 static inline void __dev_mc_unsync(struct net_device *dev, 4650 int (*unsync)(struct net_device *, 4651 const unsigned char *)) 4652 { 4653 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4654 } 4655 4656 /* Functions used for secondary unicast and multicast support */ 4657 void dev_set_rx_mode(struct net_device *dev); 4658 int dev_set_promiscuity(struct net_device *dev, int inc); 4659 int dev_set_allmulti(struct net_device *dev, int inc); 4660 void netdev_state_change(struct net_device *dev); 4661 void __netdev_notify_peers(struct net_device *dev); 4662 void netdev_notify_peers(struct net_device *dev); 4663 void netdev_features_change(struct net_device *dev); 4664 /* Load a device via the kmod */ 4665 void dev_load(struct net *net, const char *name); 4666 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4667 struct rtnl_link_stats64 *storage); 4668 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4669 const struct net_device_stats *netdev_stats); 4670 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4671 const struct pcpu_sw_netstats __percpu *netstats); 4672 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4673 4674 extern int netdev_max_backlog; 4675 extern int dev_rx_weight; 4676 extern int dev_tx_weight; 4677 extern int gro_normal_batch; 4678 4679 enum { 4680 NESTED_SYNC_IMM_BIT, 4681 NESTED_SYNC_TODO_BIT, 4682 }; 4683 4684 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4685 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4686 4687 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4688 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4689 4690 struct netdev_nested_priv { 4691 unsigned char flags; 4692 void *data; 4693 }; 4694 4695 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4696 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4697 struct list_head **iter); 4698 4699 /* iterate through upper list, must be called under RCU read lock */ 4700 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4701 for (iter = &(dev)->adj_list.upper, \ 4702 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4703 updev; \ 4704 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4705 4706 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4707 int (*fn)(struct net_device *upper_dev, 4708 struct netdev_nested_priv *priv), 4709 struct netdev_nested_priv *priv); 4710 4711 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4712 struct net_device *upper_dev); 4713 4714 bool netdev_has_any_upper_dev(struct net_device *dev); 4715 4716 void *netdev_lower_get_next_private(struct net_device *dev, 4717 struct list_head **iter); 4718 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4719 struct list_head **iter); 4720 4721 #define netdev_for_each_lower_private(dev, priv, iter) \ 4722 for (iter = (dev)->adj_list.lower.next, \ 4723 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4724 priv; \ 4725 priv = netdev_lower_get_next_private(dev, &(iter))) 4726 4727 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4728 for (iter = &(dev)->adj_list.lower, \ 4729 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4730 priv; \ 4731 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4732 4733 void *netdev_lower_get_next(struct net_device *dev, 4734 struct list_head **iter); 4735 4736 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4737 for (iter = (dev)->adj_list.lower.next, \ 4738 ldev = netdev_lower_get_next(dev, &(iter)); \ 4739 ldev; \ 4740 ldev = netdev_lower_get_next(dev, &(iter))) 4741 4742 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4743 struct list_head **iter); 4744 int netdev_walk_all_lower_dev(struct net_device *dev, 4745 int (*fn)(struct net_device *lower_dev, 4746 struct netdev_nested_priv *priv), 4747 struct netdev_nested_priv *priv); 4748 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4749 int (*fn)(struct net_device *lower_dev, 4750 struct netdev_nested_priv *priv), 4751 struct netdev_nested_priv *priv); 4752 4753 void *netdev_adjacent_get_private(struct list_head *adj_list); 4754 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4755 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4756 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4757 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4758 struct netlink_ext_ack *extack); 4759 int netdev_master_upper_dev_link(struct net_device *dev, 4760 struct net_device *upper_dev, 4761 void *upper_priv, void *upper_info, 4762 struct netlink_ext_ack *extack); 4763 void netdev_upper_dev_unlink(struct net_device *dev, 4764 struct net_device *upper_dev); 4765 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4766 struct net_device *new_dev, 4767 struct net_device *dev, 4768 struct netlink_ext_ack *extack); 4769 void netdev_adjacent_change_commit(struct net_device *old_dev, 4770 struct net_device *new_dev, 4771 struct net_device *dev); 4772 void netdev_adjacent_change_abort(struct net_device *old_dev, 4773 struct net_device *new_dev, 4774 struct net_device *dev); 4775 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4776 void *netdev_lower_dev_get_private(struct net_device *dev, 4777 struct net_device *lower_dev); 4778 void netdev_lower_state_changed(struct net_device *lower_dev, 4779 void *lower_state_info); 4780 4781 /* RSS keys are 40 or 52 bytes long */ 4782 #define NETDEV_RSS_KEY_LEN 52 4783 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4784 void netdev_rss_key_fill(void *buffer, size_t len); 4785 4786 int skb_checksum_help(struct sk_buff *skb); 4787 int skb_crc32c_csum_help(struct sk_buff *skb); 4788 int skb_csum_hwoffload_help(struct sk_buff *skb, 4789 const netdev_features_t features); 4790 4791 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4792 netdev_features_t features, bool tx_path); 4793 struct sk_buff *skb_eth_gso_segment(struct sk_buff *skb, 4794 netdev_features_t features, __be16 type); 4795 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4796 netdev_features_t features); 4797 4798 struct netdev_bonding_info { 4799 ifslave slave; 4800 ifbond master; 4801 }; 4802 4803 struct netdev_notifier_bonding_info { 4804 struct netdev_notifier_info info; /* must be first */ 4805 struct netdev_bonding_info bonding_info; 4806 }; 4807 4808 void netdev_bonding_info_change(struct net_device *dev, 4809 struct netdev_bonding_info *bonding_info); 4810 4811 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4812 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4813 #else 4814 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4815 const void *data) 4816 { 4817 } 4818 #endif 4819 4820 static inline 4821 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4822 { 4823 return __skb_gso_segment(skb, features, true); 4824 } 4825 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4826 4827 static inline bool can_checksum_protocol(netdev_features_t features, 4828 __be16 protocol) 4829 { 4830 if (protocol == htons(ETH_P_FCOE)) 4831 return !!(features & NETIF_F_FCOE_CRC); 4832 4833 /* Assume this is an IP checksum (not SCTP CRC) */ 4834 4835 if (features & NETIF_F_HW_CSUM) { 4836 /* Can checksum everything */ 4837 return true; 4838 } 4839 4840 switch (protocol) { 4841 case htons(ETH_P_IP): 4842 return !!(features & NETIF_F_IP_CSUM); 4843 case htons(ETH_P_IPV6): 4844 return !!(features & NETIF_F_IPV6_CSUM); 4845 default: 4846 return false; 4847 } 4848 } 4849 4850 #ifdef CONFIG_BUG 4851 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4852 #else 4853 static inline void netdev_rx_csum_fault(struct net_device *dev, 4854 struct sk_buff *skb) 4855 { 4856 } 4857 #endif 4858 /* rx skb timestamps */ 4859 void net_enable_timestamp(void); 4860 void net_disable_timestamp(void); 4861 4862 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 4863 const struct skb_shared_hwtstamps *hwtstamps, 4864 bool cycles) 4865 { 4866 const struct net_device_ops *ops = dev->netdev_ops; 4867 4868 if (ops->ndo_get_tstamp) 4869 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 4870 4871 return hwtstamps->hwtstamp; 4872 } 4873 4874 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4875 struct sk_buff *skb, struct net_device *dev, 4876 bool more) 4877 { 4878 __this_cpu_write(softnet_data.xmit.more, more); 4879 return ops->ndo_start_xmit(skb, dev); 4880 } 4881 4882 static inline bool netdev_xmit_more(void) 4883 { 4884 return __this_cpu_read(softnet_data.xmit.more); 4885 } 4886 4887 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4888 struct netdev_queue *txq, bool more) 4889 { 4890 const struct net_device_ops *ops = dev->netdev_ops; 4891 netdev_tx_t rc; 4892 4893 rc = __netdev_start_xmit(ops, skb, dev, more); 4894 if (rc == NETDEV_TX_OK) 4895 txq_trans_update(txq); 4896 4897 return rc; 4898 } 4899 4900 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4901 const void *ns); 4902 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4903 const void *ns); 4904 4905 extern const struct kobj_ns_type_operations net_ns_type_operations; 4906 4907 const char *netdev_drivername(const struct net_device *dev); 4908 4909 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4910 netdev_features_t f2) 4911 { 4912 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4913 if (f1 & NETIF_F_HW_CSUM) 4914 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4915 else 4916 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4917 } 4918 4919 return f1 & f2; 4920 } 4921 4922 static inline netdev_features_t netdev_get_wanted_features( 4923 struct net_device *dev) 4924 { 4925 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4926 } 4927 netdev_features_t netdev_increment_features(netdev_features_t all, 4928 netdev_features_t one, netdev_features_t mask); 4929 4930 /* Allow TSO being used on stacked device : 4931 * Performing the GSO segmentation before last device 4932 * is a performance improvement. 4933 */ 4934 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4935 netdev_features_t mask) 4936 { 4937 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4938 } 4939 4940 int __netdev_update_features(struct net_device *dev); 4941 void netdev_update_features(struct net_device *dev); 4942 void netdev_change_features(struct net_device *dev); 4943 4944 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4945 struct net_device *dev); 4946 4947 netdev_features_t passthru_features_check(struct sk_buff *skb, 4948 struct net_device *dev, 4949 netdev_features_t features); 4950 netdev_features_t netif_skb_features(struct sk_buff *skb); 4951 4952 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4953 { 4954 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4955 4956 /* check flags correspondence */ 4957 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4958 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4959 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4960 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4961 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4962 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4963 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4964 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4965 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4966 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4967 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4968 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4969 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4970 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4971 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4972 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4973 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4974 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4975 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4976 4977 return (features & feature) == feature; 4978 } 4979 4980 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4981 { 4982 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4983 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4984 } 4985 4986 static inline bool netif_needs_gso(struct sk_buff *skb, 4987 netdev_features_t features) 4988 { 4989 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4990 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4991 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4992 } 4993 4994 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 4995 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 4996 void netif_inherit_tso_max(struct net_device *to, 4997 const struct net_device *from); 4998 4999 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 5000 int pulled_hlen, u16 mac_offset, 5001 int mac_len) 5002 { 5003 skb->protocol = protocol; 5004 skb->encapsulation = 1; 5005 skb_push(skb, pulled_hlen); 5006 skb_reset_transport_header(skb); 5007 skb->mac_header = mac_offset; 5008 skb->network_header = skb->mac_header + mac_len; 5009 skb->mac_len = mac_len; 5010 } 5011 5012 static inline bool netif_is_macsec(const struct net_device *dev) 5013 { 5014 return dev->priv_flags & IFF_MACSEC; 5015 } 5016 5017 static inline bool netif_is_macvlan(const struct net_device *dev) 5018 { 5019 return dev->priv_flags & IFF_MACVLAN; 5020 } 5021 5022 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5023 { 5024 return dev->priv_flags & IFF_MACVLAN_PORT; 5025 } 5026 5027 static inline bool netif_is_bond_master(const struct net_device *dev) 5028 { 5029 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5030 } 5031 5032 static inline bool netif_is_bond_slave(const struct net_device *dev) 5033 { 5034 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5035 } 5036 5037 static inline bool netif_supports_nofcs(struct net_device *dev) 5038 { 5039 return dev->priv_flags & IFF_SUPP_NOFCS; 5040 } 5041 5042 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5043 { 5044 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5045 } 5046 5047 static inline bool netif_is_l3_master(const struct net_device *dev) 5048 { 5049 return dev->priv_flags & IFF_L3MDEV_MASTER; 5050 } 5051 5052 static inline bool netif_is_l3_slave(const struct net_device *dev) 5053 { 5054 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5055 } 5056 5057 static inline bool netif_is_bridge_master(const struct net_device *dev) 5058 { 5059 return dev->priv_flags & IFF_EBRIDGE; 5060 } 5061 5062 static inline bool netif_is_bridge_port(const struct net_device *dev) 5063 { 5064 return dev->priv_flags & IFF_BRIDGE_PORT; 5065 } 5066 5067 static inline bool netif_is_ovs_master(const struct net_device *dev) 5068 { 5069 return dev->priv_flags & IFF_OPENVSWITCH; 5070 } 5071 5072 static inline bool netif_is_ovs_port(const struct net_device *dev) 5073 { 5074 return dev->priv_flags & IFF_OVS_DATAPATH; 5075 } 5076 5077 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5078 { 5079 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5080 } 5081 5082 static inline bool netif_is_team_master(const struct net_device *dev) 5083 { 5084 return dev->priv_flags & IFF_TEAM; 5085 } 5086 5087 static inline bool netif_is_team_port(const struct net_device *dev) 5088 { 5089 return dev->priv_flags & IFF_TEAM_PORT; 5090 } 5091 5092 static inline bool netif_is_lag_master(const struct net_device *dev) 5093 { 5094 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5095 } 5096 5097 static inline bool netif_is_lag_port(const struct net_device *dev) 5098 { 5099 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5100 } 5101 5102 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5103 { 5104 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5105 } 5106 5107 static inline bool netif_is_failover(const struct net_device *dev) 5108 { 5109 return dev->priv_flags & IFF_FAILOVER; 5110 } 5111 5112 static inline bool netif_is_failover_slave(const struct net_device *dev) 5113 { 5114 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5115 } 5116 5117 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5118 static inline void netif_keep_dst(struct net_device *dev) 5119 { 5120 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5121 } 5122 5123 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5124 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5125 { 5126 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5127 return netif_is_macsec(dev); 5128 } 5129 5130 extern struct pernet_operations __net_initdata loopback_net_ops; 5131 5132 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5133 5134 /* netdev_printk helpers, similar to dev_printk */ 5135 5136 static inline const char *netdev_name(const struct net_device *dev) 5137 { 5138 if (!dev->name[0] || strchr(dev->name, '%')) 5139 return "(unnamed net_device)"; 5140 return dev->name; 5141 } 5142 5143 static inline const char *netdev_reg_state(const struct net_device *dev) 5144 { 5145 switch (dev->reg_state) { 5146 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5147 case NETREG_REGISTERED: return ""; 5148 case NETREG_UNREGISTERING: return " (unregistering)"; 5149 case NETREG_UNREGISTERED: return " (unregistered)"; 5150 case NETREG_RELEASED: return " (released)"; 5151 case NETREG_DUMMY: return " (dummy)"; 5152 } 5153 5154 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5155 return " (unknown)"; 5156 } 5157 5158 #define MODULE_ALIAS_NETDEV(device) \ 5159 MODULE_ALIAS("netdev-" device) 5160 5161 /* 5162 * netdev_WARN() acts like dev_printk(), but with the key difference 5163 * of using a WARN/WARN_ON to get the message out, including the 5164 * file/line information and a backtrace. 5165 */ 5166 #define netdev_WARN(dev, format, args...) \ 5167 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5168 netdev_reg_state(dev), ##args) 5169 5170 #define netdev_WARN_ONCE(dev, format, args...) \ 5171 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5172 netdev_reg_state(dev), ##args) 5173 5174 /* 5175 * The list of packet types we will receive (as opposed to discard) 5176 * and the routines to invoke. 5177 * 5178 * Why 16. Because with 16 the only overlap we get on a hash of the 5179 * low nibble of the protocol value is RARP/SNAP/X.25. 5180 * 5181 * 0800 IP 5182 * 0001 802.3 5183 * 0002 AX.25 5184 * 0004 802.2 5185 * 8035 RARP 5186 * 0005 SNAP 5187 * 0805 X.25 5188 * 0806 ARP 5189 * 8137 IPX 5190 * 0009 Localtalk 5191 * 86DD IPv6 5192 */ 5193 #define PTYPE_HASH_SIZE (16) 5194 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5195 5196 extern struct list_head ptype_all __read_mostly; 5197 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5198 5199 extern struct net_device *blackhole_netdev; 5200 5201 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */ 5202 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD) 5203 #define DEV_STATS_ADD(DEV, FIELD, VAL) \ 5204 atomic_long_add((VAL), &(DEV)->stats.__##FIELD) 5205 5206 #endif /* _LINUX_NETDEVICE_H */ 5207