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