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 * @module_fw_flash_in_progress: Module firmware flashing is in progress. 1994 * 1995 * @net_notifier_list: List of per-net netdev notifier block 1996 * that follow this device when it is moved 1997 * to another network namespace. 1998 * 1999 * @macsec_ops: MACsec offloading ops 2000 * 2001 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 2002 * offload capabilities of the device 2003 * @udp_tunnel_nic: UDP tunnel offload state 2004 * @xdp_state: stores info on attached XDP BPF programs 2005 * 2006 * @nested_level: Used as a parameter of spin_lock_nested() of 2007 * dev->addr_list_lock. 2008 * @unlink_list: As netif_addr_lock() can be called recursively, 2009 * keep a list of interfaces to be deleted. 2010 * @gro_max_size: Maximum size of aggregated packet in generic 2011 * receive offload (GRO) 2012 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic 2013 * receive offload (GRO), for IPv4. 2014 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP 2015 * zero copy driver 2016 * 2017 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 2018 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 2019 * @watchdog_dev_tracker: refcount tracker used by watchdog. 2020 * @dev_registered_tracker: tracker for reference held while 2021 * registered 2022 * @offload_xstats_l3: L3 HW stats for this netdevice. 2023 * 2024 * @devlink_port: Pointer to related devlink port structure. 2025 * Assigned by a driver before netdev registration using 2026 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static 2027 * during the time netdevice is registered. 2028 * 2029 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem, 2030 * where the clock is recovered. 2031 * 2032 * FIXME: cleanup struct net_device such that network protocol info 2033 * moves out. 2034 */ 2035 2036 struct net_device { 2037 /* Cacheline organization can be found documented in 2038 * Documentation/networking/net_cachelines/net_device.rst. 2039 * Please update the document when adding new fields. 2040 */ 2041 2042 /* TX read-mostly hotpath */ 2043 __cacheline_group_begin(net_device_read_tx); 2044 unsigned long long priv_flags; 2045 const struct net_device_ops *netdev_ops; 2046 const struct header_ops *header_ops; 2047 struct netdev_queue *_tx; 2048 netdev_features_t gso_partial_features; 2049 unsigned int real_num_tx_queues; 2050 unsigned int gso_max_size; 2051 unsigned int gso_ipv4_max_size; 2052 u16 gso_max_segs; 2053 s16 num_tc; 2054 /* Note : dev->mtu is often read without holding a lock. 2055 * Writers usually hold RTNL. 2056 * It is recommended to use READ_ONCE() to annotate the reads, 2057 * and to use WRITE_ONCE() to annotate the writes. 2058 */ 2059 unsigned int mtu; 2060 unsigned short needed_headroom; 2061 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2062 #ifdef CONFIG_XPS 2063 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2064 #endif 2065 #ifdef CONFIG_NETFILTER_EGRESS 2066 struct nf_hook_entries __rcu *nf_hooks_egress; 2067 #endif 2068 #ifdef CONFIG_NET_XGRESS 2069 struct bpf_mprog_entry __rcu *tcx_egress; 2070 #endif 2071 __cacheline_group_end(net_device_read_tx); 2072 2073 /* TXRX read-mostly hotpath */ 2074 __cacheline_group_begin(net_device_read_txrx); 2075 union { 2076 struct pcpu_lstats __percpu *lstats; 2077 struct pcpu_sw_netstats __percpu *tstats; 2078 struct pcpu_dstats __percpu *dstats; 2079 }; 2080 unsigned long state; 2081 unsigned int flags; 2082 unsigned short hard_header_len; 2083 netdev_features_t features; 2084 struct inet6_dev __rcu *ip6_ptr; 2085 __cacheline_group_end(net_device_read_txrx); 2086 2087 /* RX read-mostly hotpath */ 2088 __cacheline_group_begin(net_device_read_rx); 2089 struct bpf_prog __rcu *xdp_prog; 2090 struct list_head ptype_specific; 2091 int ifindex; 2092 unsigned int real_num_rx_queues; 2093 struct netdev_rx_queue *_rx; 2094 unsigned long gro_flush_timeout; 2095 int napi_defer_hard_irqs; 2096 unsigned int gro_max_size; 2097 unsigned int gro_ipv4_max_size; 2098 rx_handler_func_t __rcu *rx_handler; 2099 void __rcu *rx_handler_data; 2100 possible_net_t nd_net; 2101 #ifdef CONFIG_NETPOLL 2102 struct netpoll_info __rcu *npinfo; 2103 #endif 2104 #ifdef CONFIG_NET_XGRESS 2105 struct bpf_mprog_entry __rcu *tcx_ingress; 2106 #endif 2107 __cacheline_group_end(net_device_read_rx); 2108 2109 char name[IFNAMSIZ]; 2110 struct netdev_name_node *name_node; 2111 struct dev_ifalias __rcu *ifalias; 2112 /* 2113 * I/O specific fields 2114 * FIXME: Merge these and struct ifmap into one 2115 */ 2116 unsigned long mem_end; 2117 unsigned long mem_start; 2118 unsigned long base_addr; 2119 2120 /* 2121 * Some hardware also needs these fields (state,dev_list, 2122 * napi_list,unreg_list,close_list) but they are not 2123 * part of the usual set specified in Space.c. 2124 */ 2125 2126 2127 struct list_head dev_list; 2128 struct list_head napi_list; 2129 struct list_head unreg_list; 2130 struct list_head close_list; 2131 struct list_head ptype_all; 2132 2133 struct { 2134 struct list_head upper; 2135 struct list_head lower; 2136 } adj_list; 2137 2138 /* Read-mostly cache-line for fast-path access */ 2139 xdp_features_t xdp_features; 2140 const struct xdp_metadata_ops *xdp_metadata_ops; 2141 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops; 2142 unsigned short gflags; 2143 2144 unsigned short needed_tailroom; 2145 2146 netdev_features_t hw_features; 2147 netdev_features_t wanted_features; 2148 netdev_features_t vlan_features; 2149 netdev_features_t hw_enc_features; 2150 netdev_features_t mpls_features; 2151 2152 unsigned int min_mtu; 2153 unsigned int max_mtu; 2154 unsigned short type; 2155 unsigned char min_header_len; 2156 unsigned char name_assign_type; 2157 2158 int group; 2159 2160 struct net_device_stats stats; /* not used by modern drivers */ 2161 2162 struct net_device_core_stats __percpu *core_stats; 2163 2164 /* Stats to monitor link on/off, flapping */ 2165 atomic_t carrier_up_count; 2166 atomic_t carrier_down_count; 2167 2168 #ifdef CONFIG_WIRELESS_EXT 2169 const struct iw_handler_def *wireless_handlers; 2170 struct iw_public_data *wireless_data; 2171 #endif 2172 const struct ethtool_ops *ethtool_ops; 2173 #ifdef CONFIG_NET_L3_MASTER_DEV 2174 const struct l3mdev_ops *l3mdev_ops; 2175 #endif 2176 #if IS_ENABLED(CONFIG_IPV6) 2177 const struct ndisc_ops *ndisc_ops; 2178 #endif 2179 2180 #ifdef CONFIG_XFRM_OFFLOAD 2181 const struct xfrmdev_ops *xfrmdev_ops; 2182 #endif 2183 2184 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2185 const struct tlsdev_ops *tlsdev_ops; 2186 #endif 2187 2188 unsigned int operstate; 2189 unsigned char link_mode; 2190 2191 unsigned char if_port; 2192 unsigned char dma; 2193 2194 /* Interface address info. */ 2195 unsigned char perm_addr[MAX_ADDR_LEN]; 2196 unsigned char addr_assign_type; 2197 unsigned char addr_len; 2198 unsigned char upper_level; 2199 unsigned char lower_level; 2200 2201 unsigned short neigh_priv_len; 2202 unsigned short dev_id; 2203 unsigned short dev_port; 2204 unsigned short padded; 2205 2206 spinlock_t addr_list_lock; 2207 int irq; 2208 2209 struct netdev_hw_addr_list uc; 2210 struct netdev_hw_addr_list mc; 2211 struct netdev_hw_addr_list dev_addrs; 2212 2213 #ifdef CONFIG_SYSFS 2214 struct kset *queues_kset; 2215 #endif 2216 #ifdef CONFIG_LOCKDEP 2217 struct list_head unlink_list; 2218 #endif 2219 unsigned int promiscuity; 2220 unsigned int allmulti; 2221 bool uc_promisc; 2222 #ifdef CONFIG_LOCKDEP 2223 unsigned char nested_level; 2224 #endif 2225 2226 2227 /* Protocol-specific pointers */ 2228 struct in_device __rcu *ip_ptr; 2229 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2230 struct vlan_info __rcu *vlan_info; 2231 #endif 2232 #if IS_ENABLED(CONFIG_NET_DSA) 2233 struct dsa_port *dsa_ptr; 2234 #endif 2235 #if IS_ENABLED(CONFIG_TIPC) 2236 struct tipc_bearer __rcu *tipc_ptr; 2237 #endif 2238 #if IS_ENABLED(CONFIG_ATALK) 2239 void *atalk_ptr; 2240 #endif 2241 #if IS_ENABLED(CONFIG_AX25) 2242 void *ax25_ptr; 2243 #endif 2244 #if IS_ENABLED(CONFIG_CFG80211) 2245 struct wireless_dev *ieee80211_ptr; 2246 #endif 2247 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2248 struct wpan_dev *ieee802154_ptr; 2249 #endif 2250 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2251 struct mpls_dev __rcu *mpls_ptr; 2252 #endif 2253 #if IS_ENABLED(CONFIG_MCTP) 2254 struct mctp_dev __rcu *mctp_ptr; 2255 #endif 2256 2257 /* 2258 * Cache lines mostly used on receive path (including eth_type_trans()) 2259 */ 2260 /* Interface address info used in eth_type_trans() */ 2261 const unsigned char *dev_addr; 2262 2263 unsigned int num_rx_queues; 2264 #define GRO_LEGACY_MAX_SIZE 65536u 2265 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2266 * and shinfo->gso_segs is a 16bit field. 2267 */ 2268 #define GRO_MAX_SIZE (8 * 65535u) 2269 unsigned int xdp_zc_max_segs; 2270 struct netdev_queue __rcu *ingress_queue; 2271 #ifdef CONFIG_NETFILTER_INGRESS 2272 struct nf_hook_entries __rcu *nf_hooks_ingress; 2273 #endif 2274 2275 unsigned char broadcast[MAX_ADDR_LEN]; 2276 #ifdef CONFIG_RFS_ACCEL 2277 struct cpu_rmap *rx_cpu_rmap; 2278 #endif 2279 struct hlist_node index_hlist; 2280 2281 /* 2282 * Cache lines mostly used on transmit path 2283 */ 2284 unsigned int num_tx_queues; 2285 struct Qdisc __rcu *qdisc; 2286 unsigned int tx_queue_len; 2287 spinlock_t tx_global_lock; 2288 2289 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2290 2291 #ifdef CONFIG_NET_SCHED 2292 DECLARE_HASHTABLE (qdisc_hash, 4); 2293 #endif 2294 /* These may be needed for future network-power-down code. */ 2295 struct timer_list watchdog_timer; 2296 int watchdog_timeo; 2297 2298 u32 proto_down_reason; 2299 2300 struct list_head todo_list; 2301 2302 #ifdef CONFIG_PCPU_DEV_REFCNT 2303 int __percpu *pcpu_refcnt; 2304 #else 2305 refcount_t dev_refcnt; 2306 #endif 2307 struct ref_tracker_dir refcnt_tracker; 2308 2309 struct list_head link_watch_list; 2310 2311 u8 reg_state; 2312 2313 bool dismantle; 2314 2315 enum { 2316 RTNL_LINK_INITIALIZED, 2317 RTNL_LINK_INITIALIZING, 2318 } rtnl_link_state:16; 2319 2320 bool needs_free_netdev; 2321 void (*priv_destructor)(struct net_device *dev); 2322 2323 /* mid-layer private */ 2324 void *ml_priv; 2325 enum netdev_ml_priv_type ml_priv_type; 2326 2327 enum netdev_stat_type pcpu_stat_type:8; 2328 2329 #if IS_ENABLED(CONFIG_GARP) 2330 struct garp_port __rcu *garp_port; 2331 #endif 2332 #if IS_ENABLED(CONFIG_MRP) 2333 struct mrp_port __rcu *mrp_port; 2334 #endif 2335 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2336 struct dm_hw_stat_delta __rcu *dm_private; 2337 #endif 2338 struct device dev; 2339 const struct attribute_group *sysfs_groups[4]; 2340 const struct attribute_group *sysfs_rx_queue_group; 2341 2342 const struct rtnl_link_ops *rtnl_link_ops; 2343 2344 const struct netdev_stat_ops *stat_ops; 2345 2346 const struct netdev_queue_mgmt_ops *queue_mgmt_ops; 2347 2348 /* for setting kernel sock attribute on TCP connection setup */ 2349 #define GSO_MAX_SEGS 65535u 2350 #define GSO_LEGACY_MAX_SIZE 65536u 2351 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2352 * and shinfo->gso_segs is a 16bit field. 2353 */ 2354 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2355 2356 #define TSO_LEGACY_MAX_SIZE 65536 2357 #define TSO_MAX_SIZE UINT_MAX 2358 unsigned int tso_max_size; 2359 #define TSO_MAX_SEGS U16_MAX 2360 u16 tso_max_segs; 2361 2362 #ifdef CONFIG_DCB 2363 const struct dcbnl_rtnl_ops *dcbnl_ops; 2364 #endif 2365 u8 prio_tc_map[TC_BITMASK + 1]; 2366 2367 #if IS_ENABLED(CONFIG_FCOE) 2368 unsigned int fcoe_ddp_xid; 2369 #endif 2370 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2371 struct netprio_map __rcu *priomap; 2372 #endif 2373 struct phy_device *phydev; 2374 struct sfp_bus *sfp_bus; 2375 struct lock_class_key *qdisc_tx_busylock; 2376 bool proto_down; 2377 bool threaded; 2378 unsigned wol_enabled:1; 2379 unsigned module_fw_flash_in_progress:1; 2380 struct list_head net_notifier_list; 2381 2382 #if IS_ENABLED(CONFIG_MACSEC) 2383 /* MACsec management functions */ 2384 const struct macsec_ops *macsec_ops; 2385 #endif 2386 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2387 struct udp_tunnel_nic *udp_tunnel_nic; 2388 2389 /* protected by rtnl_lock */ 2390 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2391 2392 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2393 netdevice_tracker linkwatch_dev_tracker; 2394 netdevice_tracker watchdog_dev_tracker; 2395 netdevice_tracker dev_registered_tracker; 2396 struct rtnl_hw_stats64 *offload_xstats_l3; 2397 2398 struct devlink_port *devlink_port; 2399 2400 #if IS_ENABLED(CONFIG_DPLL) 2401 struct dpll_pin __rcu *dpll_pin; 2402 #endif 2403 #if IS_ENABLED(CONFIG_PAGE_POOL) 2404 /** @page_pools: page pools created for this netdevice */ 2405 struct hlist_head page_pools; 2406 #endif 2407 2408 /** @irq_moder: dim parameters used if IS_ENABLED(CONFIG_DIMLIB). */ 2409 struct dim_irq_moder *irq_moder; 2410 }; 2411 #define to_net_dev(d) container_of(d, struct net_device, dev) 2412 2413 /* 2414 * Driver should use this to assign devlink port instance to a netdevice 2415 * before it registers the netdevice. Therefore devlink_port is static 2416 * during the netdev lifetime after it is registered. 2417 */ 2418 #define SET_NETDEV_DEVLINK_PORT(dev, port) \ 2419 ({ \ 2420 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \ 2421 ((dev)->devlink_port = (port)); \ 2422 }) 2423 2424 static inline bool netif_elide_gro(const struct net_device *dev) 2425 { 2426 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2427 return true; 2428 return false; 2429 } 2430 2431 #define NETDEV_ALIGN 32 2432 2433 static inline 2434 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2435 { 2436 return dev->prio_tc_map[prio & TC_BITMASK]; 2437 } 2438 2439 static inline 2440 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2441 { 2442 if (tc >= dev->num_tc) 2443 return -EINVAL; 2444 2445 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2446 return 0; 2447 } 2448 2449 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2450 void netdev_reset_tc(struct net_device *dev); 2451 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2452 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2453 2454 static inline 2455 int netdev_get_num_tc(struct net_device *dev) 2456 { 2457 return dev->num_tc; 2458 } 2459 2460 static inline void net_prefetch(void *p) 2461 { 2462 prefetch(p); 2463 #if L1_CACHE_BYTES < 128 2464 prefetch((u8 *)p + L1_CACHE_BYTES); 2465 #endif 2466 } 2467 2468 static inline void net_prefetchw(void *p) 2469 { 2470 prefetchw(p); 2471 #if L1_CACHE_BYTES < 128 2472 prefetchw((u8 *)p + L1_CACHE_BYTES); 2473 #endif 2474 } 2475 2476 void netdev_unbind_sb_channel(struct net_device *dev, 2477 struct net_device *sb_dev); 2478 int netdev_bind_sb_channel_queue(struct net_device *dev, 2479 struct net_device *sb_dev, 2480 u8 tc, u16 count, u16 offset); 2481 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2482 static inline int netdev_get_sb_channel(struct net_device *dev) 2483 { 2484 return max_t(int, -dev->num_tc, 0); 2485 } 2486 2487 static inline 2488 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2489 unsigned int index) 2490 { 2491 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues); 2492 return &dev->_tx[index]; 2493 } 2494 2495 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2496 const struct sk_buff *skb) 2497 { 2498 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2499 } 2500 2501 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2502 void (*f)(struct net_device *, 2503 struct netdev_queue *, 2504 void *), 2505 void *arg) 2506 { 2507 unsigned int i; 2508 2509 for (i = 0; i < dev->num_tx_queues; i++) 2510 f(dev, &dev->_tx[i], arg); 2511 } 2512 2513 #define netdev_lockdep_set_classes(dev) \ 2514 { \ 2515 static struct lock_class_key qdisc_tx_busylock_key; \ 2516 static struct lock_class_key qdisc_xmit_lock_key; \ 2517 static struct lock_class_key dev_addr_list_lock_key; \ 2518 unsigned int i; \ 2519 \ 2520 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2521 lockdep_set_class(&(dev)->addr_list_lock, \ 2522 &dev_addr_list_lock_key); \ 2523 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2524 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2525 &qdisc_xmit_lock_key); \ 2526 } 2527 2528 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2529 struct net_device *sb_dev); 2530 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2531 struct sk_buff *skb, 2532 struct net_device *sb_dev); 2533 2534 /* returns the headroom that the master device needs to take in account 2535 * when forwarding to this dev 2536 */ 2537 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2538 { 2539 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2540 } 2541 2542 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2543 { 2544 if (dev->netdev_ops->ndo_set_rx_headroom) 2545 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2546 } 2547 2548 /* set the device rx headroom to the dev's default */ 2549 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2550 { 2551 netdev_set_rx_headroom(dev, -1); 2552 } 2553 2554 static inline void *netdev_get_ml_priv(struct net_device *dev, 2555 enum netdev_ml_priv_type type) 2556 { 2557 if (dev->ml_priv_type != type) 2558 return NULL; 2559 2560 return dev->ml_priv; 2561 } 2562 2563 static inline void netdev_set_ml_priv(struct net_device *dev, 2564 void *ml_priv, 2565 enum netdev_ml_priv_type type) 2566 { 2567 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2568 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2569 dev->ml_priv_type, type); 2570 WARN(!dev->ml_priv_type && dev->ml_priv, 2571 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2572 2573 dev->ml_priv = ml_priv; 2574 dev->ml_priv_type = type; 2575 } 2576 2577 /* 2578 * Net namespace inlines 2579 */ 2580 static inline 2581 struct net *dev_net(const struct net_device *dev) 2582 { 2583 return read_pnet(&dev->nd_net); 2584 } 2585 2586 static inline 2587 void dev_net_set(struct net_device *dev, struct net *net) 2588 { 2589 write_pnet(&dev->nd_net, net); 2590 } 2591 2592 /** 2593 * netdev_priv - access network device private data 2594 * @dev: network device 2595 * 2596 * Get network device private data 2597 */ 2598 static inline void *netdev_priv(const struct net_device *dev) 2599 { 2600 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2601 } 2602 2603 /* Set the sysfs physical device reference for the network logical device 2604 * if set prior to registration will cause a symlink during initialization. 2605 */ 2606 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2607 2608 /* Set the sysfs device type for the network logical device to allow 2609 * fine-grained identification of different network device types. For 2610 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2611 */ 2612 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2613 2614 void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index, 2615 enum netdev_queue_type type, 2616 struct napi_struct *napi); 2617 2618 static inline void netif_napi_set_irq(struct napi_struct *napi, int irq) 2619 { 2620 napi->irq = irq; 2621 } 2622 2623 /* Default NAPI poll() weight 2624 * Device drivers are strongly advised to not use bigger value 2625 */ 2626 #define NAPI_POLL_WEIGHT 64 2627 2628 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2629 int (*poll)(struct napi_struct *, int), int weight); 2630 2631 /** 2632 * netif_napi_add() - initialize a NAPI context 2633 * @dev: network device 2634 * @napi: NAPI context 2635 * @poll: polling function 2636 * 2637 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2638 * *any* of the other NAPI-related functions. 2639 */ 2640 static inline void 2641 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2642 int (*poll)(struct napi_struct *, int)) 2643 { 2644 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2645 } 2646 2647 static inline void 2648 netif_napi_add_tx_weight(struct net_device *dev, 2649 struct napi_struct *napi, 2650 int (*poll)(struct napi_struct *, int), 2651 int weight) 2652 { 2653 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2654 netif_napi_add_weight(dev, napi, poll, weight); 2655 } 2656 2657 /** 2658 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2659 * @dev: network device 2660 * @napi: NAPI context 2661 * @poll: polling function 2662 * 2663 * This variant of netif_napi_add() should be used from drivers using NAPI 2664 * to exclusively poll a TX queue. 2665 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2666 */ 2667 static inline void netif_napi_add_tx(struct net_device *dev, 2668 struct napi_struct *napi, 2669 int (*poll)(struct napi_struct *, int)) 2670 { 2671 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2672 } 2673 2674 /** 2675 * __netif_napi_del - remove a NAPI context 2676 * @napi: NAPI context 2677 * 2678 * Warning: caller must observe RCU grace period before freeing memory 2679 * containing @napi. Drivers might want to call this helper to combine 2680 * all the needed RCU grace periods into a single one. 2681 */ 2682 void __netif_napi_del(struct napi_struct *napi); 2683 2684 /** 2685 * netif_napi_del - remove a NAPI context 2686 * @napi: NAPI context 2687 * 2688 * netif_napi_del() removes a NAPI context from the network device NAPI list 2689 */ 2690 static inline void netif_napi_del(struct napi_struct *napi) 2691 { 2692 __netif_napi_del(napi); 2693 synchronize_net(); 2694 } 2695 2696 struct packet_type { 2697 __be16 type; /* This is really htons(ether_type). */ 2698 bool ignore_outgoing; 2699 struct net_device *dev; /* NULL is wildcarded here */ 2700 netdevice_tracker dev_tracker; 2701 int (*func) (struct sk_buff *, 2702 struct net_device *, 2703 struct packet_type *, 2704 struct net_device *); 2705 void (*list_func) (struct list_head *, 2706 struct packet_type *, 2707 struct net_device *); 2708 bool (*id_match)(struct packet_type *ptype, 2709 struct sock *sk); 2710 struct net *af_packet_net; 2711 void *af_packet_priv; 2712 struct list_head list; 2713 }; 2714 2715 struct offload_callbacks { 2716 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2717 netdev_features_t features); 2718 struct sk_buff *(*gro_receive)(struct list_head *head, 2719 struct sk_buff *skb); 2720 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2721 }; 2722 2723 struct packet_offload { 2724 __be16 type; /* This is really htons(ether_type). */ 2725 u16 priority; 2726 struct offload_callbacks callbacks; 2727 struct list_head list; 2728 }; 2729 2730 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2731 struct pcpu_sw_netstats { 2732 u64_stats_t rx_packets; 2733 u64_stats_t rx_bytes; 2734 u64_stats_t tx_packets; 2735 u64_stats_t tx_bytes; 2736 struct u64_stats_sync syncp; 2737 } __aligned(4 * sizeof(u64)); 2738 2739 struct pcpu_dstats { 2740 u64_stats_t rx_packets; 2741 u64_stats_t rx_bytes; 2742 u64_stats_t rx_drops; 2743 u64_stats_t tx_packets; 2744 u64_stats_t tx_bytes; 2745 u64_stats_t tx_drops; 2746 struct u64_stats_sync syncp; 2747 } __aligned(8 * sizeof(u64)); 2748 2749 struct pcpu_lstats { 2750 u64_stats_t packets; 2751 u64_stats_t bytes; 2752 struct u64_stats_sync syncp; 2753 } __aligned(2 * sizeof(u64)); 2754 2755 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2756 2757 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2758 { 2759 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2760 2761 u64_stats_update_begin(&tstats->syncp); 2762 u64_stats_add(&tstats->rx_bytes, len); 2763 u64_stats_inc(&tstats->rx_packets); 2764 u64_stats_update_end(&tstats->syncp); 2765 } 2766 2767 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2768 unsigned int packets, 2769 unsigned int len) 2770 { 2771 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2772 2773 u64_stats_update_begin(&tstats->syncp); 2774 u64_stats_add(&tstats->tx_bytes, len); 2775 u64_stats_add(&tstats->tx_packets, packets); 2776 u64_stats_update_end(&tstats->syncp); 2777 } 2778 2779 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2780 { 2781 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2782 2783 u64_stats_update_begin(&lstats->syncp); 2784 u64_stats_add(&lstats->bytes, len); 2785 u64_stats_inc(&lstats->packets); 2786 u64_stats_update_end(&lstats->syncp); 2787 } 2788 2789 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2790 ({ \ 2791 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2792 if (pcpu_stats) { \ 2793 int __cpu; \ 2794 for_each_possible_cpu(__cpu) { \ 2795 typeof(type) *stat; \ 2796 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2797 u64_stats_init(&stat->syncp); \ 2798 } \ 2799 } \ 2800 pcpu_stats; \ 2801 }) 2802 2803 #define netdev_alloc_pcpu_stats(type) \ 2804 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2805 2806 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2807 ({ \ 2808 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2809 if (pcpu_stats) { \ 2810 int __cpu; \ 2811 for_each_possible_cpu(__cpu) { \ 2812 typeof(type) *stat; \ 2813 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2814 u64_stats_init(&stat->syncp); \ 2815 } \ 2816 } \ 2817 pcpu_stats; \ 2818 }) 2819 2820 enum netdev_lag_tx_type { 2821 NETDEV_LAG_TX_TYPE_UNKNOWN, 2822 NETDEV_LAG_TX_TYPE_RANDOM, 2823 NETDEV_LAG_TX_TYPE_BROADCAST, 2824 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2825 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2826 NETDEV_LAG_TX_TYPE_HASH, 2827 }; 2828 2829 enum netdev_lag_hash { 2830 NETDEV_LAG_HASH_NONE, 2831 NETDEV_LAG_HASH_L2, 2832 NETDEV_LAG_HASH_L34, 2833 NETDEV_LAG_HASH_L23, 2834 NETDEV_LAG_HASH_E23, 2835 NETDEV_LAG_HASH_E34, 2836 NETDEV_LAG_HASH_VLAN_SRCMAC, 2837 NETDEV_LAG_HASH_UNKNOWN, 2838 }; 2839 2840 struct netdev_lag_upper_info { 2841 enum netdev_lag_tx_type tx_type; 2842 enum netdev_lag_hash hash_type; 2843 }; 2844 2845 struct netdev_lag_lower_state_info { 2846 u8 link_up : 1, 2847 tx_enabled : 1; 2848 }; 2849 2850 #include <linux/notifier.h> 2851 2852 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2853 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2854 * adding new types. 2855 */ 2856 enum netdev_cmd { 2857 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2858 NETDEV_DOWN, 2859 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2860 detected a hardware crash and restarted 2861 - we can use this eg to kick tcp sessions 2862 once done */ 2863 NETDEV_CHANGE, /* Notify device state change */ 2864 NETDEV_REGISTER, 2865 NETDEV_UNREGISTER, 2866 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2867 NETDEV_CHANGEADDR, /* notify after the address change */ 2868 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2869 NETDEV_GOING_DOWN, 2870 NETDEV_CHANGENAME, 2871 NETDEV_FEAT_CHANGE, 2872 NETDEV_BONDING_FAILOVER, 2873 NETDEV_PRE_UP, 2874 NETDEV_PRE_TYPE_CHANGE, 2875 NETDEV_POST_TYPE_CHANGE, 2876 NETDEV_POST_INIT, 2877 NETDEV_PRE_UNINIT, 2878 NETDEV_RELEASE, 2879 NETDEV_NOTIFY_PEERS, 2880 NETDEV_JOIN, 2881 NETDEV_CHANGEUPPER, 2882 NETDEV_RESEND_IGMP, 2883 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2884 NETDEV_CHANGEINFODATA, 2885 NETDEV_BONDING_INFO, 2886 NETDEV_PRECHANGEUPPER, 2887 NETDEV_CHANGELOWERSTATE, 2888 NETDEV_UDP_TUNNEL_PUSH_INFO, 2889 NETDEV_UDP_TUNNEL_DROP_INFO, 2890 NETDEV_CHANGE_TX_QUEUE_LEN, 2891 NETDEV_CVLAN_FILTER_PUSH_INFO, 2892 NETDEV_CVLAN_FILTER_DROP_INFO, 2893 NETDEV_SVLAN_FILTER_PUSH_INFO, 2894 NETDEV_SVLAN_FILTER_DROP_INFO, 2895 NETDEV_OFFLOAD_XSTATS_ENABLE, 2896 NETDEV_OFFLOAD_XSTATS_DISABLE, 2897 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 2898 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 2899 NETDEV_XDP_FEAT_CHANGE, 2900 }; 2901 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2902 2903 int register_netdevice_notifier(struct notifier_block *nb); 2904 int unregister_netdevice_notifier(struct notifier_block *nb); 2905 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2906 int unregister_netdevice_notifier_net(struct net *net, 2907 struct notifier_block *nb); 2908 int register_netdevice_notifier_dev_net(struct net_device *dev, 2909 struct notifier_block *nb, 2910 struct netdev_net_notifier *nn); 2911 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2912 struct notifier_block *nb, 2913 struct netdev_net_notifier *nn); 2914 2915 struct netdev_notifier_info { 2916 struct net_device *dev; 2917 struct netlink_ext_ack *extack; 2918 }; 2919 2920 struct netdev_notifier_info_ext { 2921 struct netdev_notifier_info info; /* must be first */ 2922 union { 2923 u32 mtu; 2924 } ext; 2925 }; 2926 2927 struct netdev_notifier_change_info { 2928 struct netdev_notifier_info info; /* must be first */ 2929 unsigned int flags_changed; 2930 }; 2931 2932 struct netdev_notifier_changeupper_info { 2933 struct netdev_notifier_info info; /* must be first */ 2934 struct net_device *upper_dev; /* new upper dev */ 2935 bool master; /* is upper dev master */ 2936 bool linking; /* is the notification for link or unlink */ 2937 void *upper_info; /* upper dev info */ 2938 }; 2939 2940 struct netdev_notifier_changelowerstate_info { 2941 struct netdev_notifier_info info; /* must be first */ 2942 void *lower_state_info; /* is lower dev state */ 2943 }; 2944 2945 struct netdev_notifier_pre_changeaddr_info { 2946 struct netdev_notifier_info info; /* must be first */ 2947 const unsigned char *dev_addr; 2948 }; 2949 2950 enum netdev_offload_xstats_type { 2951 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 2952 }; 2953 2954 struct netdev_notifier_offload_xstats_info { 2955 struct netdev_notifier_info info; /* must be first */ 2956 enum netdev_offload_xstats_type type; 2957 2958 union { 2959 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 2960 struct netdev_notifier_offload_xstats_rd *report_delta; 2961 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 2962 struct netdev_notifier_offload_xstats_ru *report_used; 2963 }; 2964 }; 2965 2966 int netdev_offload_xstats_enable(struct net_device *dev, 2967 enum netdev_offload_xstats_type type, 2968 struct netlink_ext_ack *extack); 2969 int netdev_offload_xstats_disable(struct net_device *dev, 2970 enum netdev_offload_xstats_type type); 2971 bool netdev_offload_xstats_enabled(const struct net_device *dev, 2972 enum netdev_offload_xstats_type type); 2973 int netdev_offload_xstats_get(struct net_device *dev, 2974 enum netdev_offload_xstats_type type, 2975 struct rtnl_hw_stats64 *stats, bool *used, 2976 struct netlink_ext_ack *extack); 2977 void 2978 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 2979 const struct rtnl_hw_stats64 *stats); 2980 void 2981 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 2982 void netdev_offload_xstats_push_delta(struct net_device *dev, 2983 enum netdev_offload_xstats_type type, 2984 const struct rtnl_hw_stats64 *stats); 2985 2986 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2987 struct net_device *dev) 2988 { 2989 info->dev = dev; 2990 info->extack = NULL; 2991 } 2992 2993 static inline struct net_device * 2994 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2995 { 2996 return info->dev; 2997 } 2998 2999 static inline struct netlink_ext_ack * 3000 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 3001 { 3002 return info->extack; 3003 } 3004 3005 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 3006 int call_netdevice_notifiers_info(unsigned long val, 3007 struct netdev_notifier_info *info); 3008 3009 #define for_each_netdev(net, d) \ 3010 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 3011 #define for_each_netdev_reverse(net, d) \ 3012 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 3013 #define for_each_netdev_rcu(net, d) \ 3014 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 3015 #define for_each_netdev_safe(net, d, n) \ 3016 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 3017 #define for_each_netdev_continue(net, d) \ 3018 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 3019 #define for_each_netdev_continue_reverse(net, d) \ 3020 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 3021 dev_list) 3022 #define for_each_netdev_continue_rcu(net, d) \ 3023 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 3024 #define for_each_netdev_in_bond_rcu(bond, slave) \ 3025 for_each_netdev_rcu(&init_net, slave) \ 3026 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 3027 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 3028 3029 #define for_each_netdev_dump(net, d, ifindex) \ 3030 for (; (d = xa_find(&(net)->dev_by_index, &ifindex, \ 3031 ULONG_MAX, XA_PRESENT)); ifindex++) 3032 3033 static inline struct net_device *next_net_device(struct net_device *dev) 3034 { 3035 struct list_head *lh; 3036 struct net *net; 3037 3038 net = dev_net(dev); 3039 lh = dev->dev_list.next; 3040 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3041 } 3042 3043 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 3044 { 3045 struct list_head *lh; 3046 struct net *net; 3047 3048 net = dev_net(dev); 3049 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 3050 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3051 } 3052 3053 static inline struct net_device *first_net_device(struct net *net) 3054 { 3055 return list_empty(&net->dev_base_head) ? NULL : 3056 net_device_entry(net->dev_base_head.next); 3057 } 3058 3059 static inline struct net_device *first_net_device_rcu(struct net *net) 3060 { 3061 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 3062 3063 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3064 } 3065 3066 int netdev_boot_setup_check(struct net_device *dev); 3067 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 3068 const char *hwaddr); 3069 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 3070 void dev_add_pack(struct packet_type *pt); 3071 void dev_remove_pack(struct packet_type *pt); 3072 void __dev_remove_pack(struct packet_type *pt); 3073 void dev_add_offload(struct packet_offload *po); 3074 void dev_remove_offload(struct packet_offload *po); 3075 3076 int dev_get_iflink(const struct net_device *dev); 3077 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 3078 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 3079 struct net_device_path_stack *stack); 3080 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 3081 unsigned short mask); 3082 struct net_device *dev_get_by_name(struct net *net, const char *name); 3083 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 3084 struct net_device *__dev_get_by_name(struct net *net, const char *name); 3085 bool netdev_name_in_use(struct net *net, const char *name); 3086 int dev_alloc_name(struct net_device *dev, const char *name); 3087 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 3088 void dev_close(struct net_device *dev); 3089 void dev_close_many(struct list_head *head, bool unlink); 3090 void dev_disable_lro(struct net_device *dev); 3091 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 3092 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 3093 struct net_device *sb_dev); 3094 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 3095 struct net_device *sb_dev); 3096 3097 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3098 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3099 3100 static inline int dev_queue_xmit(struct sk_buff *skb) 3101 { 3102 return __dev_queue_xmit(skb, NULL); 3103 } 3104 3105 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3106 struct net_device *sb_dev) 3107 { 3108 return __dev_queue_xmit(skb, sb_dev); 3109 } 3110 3111 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3112 { 3113 int ret; 3114 3115 ret = __dev_direct_xmit(skb, queue_id); 3116 if (!dev_xmit_complete(ret)) 3117 kfree_skb(skb); 3118 return ret; 3119 } 3120 3121 int register_netdevice(struct net_device *dev); 3122 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3123 void unregister_netdevice_many(struct list_head *head); 3124 static inline void unregister_netdevice(struct net_device *dev) 3125 { 3126 unregister_netdevice_queue(dev, NULL); 3127 } 3128 3129 int netdev_refcnt_read(const struct net_device *dev); 3130 void free_netdev(struct net_device *dev); 3131 void netdev_freemem(struct net_device *dev); 3132 void init_dummy_netdev(struct net_device *dev); 3133 3134 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3135 struct sk_buff *skb, 3136 bool all_slaves); 3137 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3138 struct sock *sk); 3139 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3140 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3141 struct net_device *netdev_get_by_index(struct net *net, int ifindex, 3142 netdevice_tracker *tracker, gfp_t gfp); 3143 struct net_device *netdev_get_by_name(struct net *net, const char *name, 3144 netdevice_tracker *tracker, gfp_t gfp); 3145 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3146 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 3147 void netdev_copy_name(struct net_device *dev, char *name); 3148 3149 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3150 unsigned short type, 3151 const void *daddr, const void *saddr, 3152 unsigned int len) 3153 { 3154 if (!dev->header_ops || !dev->header_ops->create) 3155 return 0; 3156 3157 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3158 } 3159 3160 static inline int dev_parse_header(const struct sk_buff *skb, 3161 unsigned char *haddr) 3162 { 3163 const struct net_device *dev = skb->dev; 3164 3165 if (!dev->header_ops || !dev->header_ops->parse) 3166 return 0; 3167 return dev->header_ops->parse(skb, haddr); 3168 } 3169 3170 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3171 { 3172 const struct net_device *dev = skb->dev; 3173 3174 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3175 return 0; 3176 return dev->header_ops->parse_protocol(skb); 3177 } 3178 3179 /* ll_header must have at least hard_header_len allocated */ 3180 static inline bool dev_validate_header(const struct net_device *dev, 3181 char *ll_header, int len) 3182 { 3183 if (likely(len >= dev->hard_header_len)) 3184 return true; 3185 if (len < dev->min_header_len) 3186 return false; 3187 3188 if (capable(CAP_SYS_RAWIO)) { 3189 memset(ll_header + len, 0, dev->hard_header_len - len); 3190 return true; 3191 } 3192 3193 if (dev->header_ops && dev->header_ops->validate) 3194 return dev->header_ops->validate(ll_header, len); 3195 3196 return false; 3197 } 3198 3199 static inline bool dev_has_header(const struct net_device *dev) 3200 { 3201 return dev->header_ops && dev->header_ops->create; 3202 } 3203 3204 /* 3205 * Incoming packets are placed on per-CPU queues 3206 */ 3207 struct softnet_data { 3208 struct list_head poll_list; 3209 struct sk_buff_head process_queue; 3210 local_lock_t process_queue_bh_lock; 3211 3212 /* stats */ 3213 unsigned int processed; 3214 unsigned int time_squeeze; 3215 #ifdef CONFIG_RPS 3216 struct softnet_data *rps_ipi_list; 3217 #endif 3218 3219 unsigned int received_rps; 3220 bool in_net_rx_action; 3221 bool in_napi_threaded_poll; 3222 3223 #ifdef CONFIG_NET_FLOW_LIMIT 3224 struct sd_flow_limit __rcu *flow_limit; 3225 #endif 3226 struct Qdisc *output_queue; 3227 struct Qdisc **output_queue_tailp; 3228 struct sk_buff *completion_queue; 3229 #ifdef CONFIG_XFRM_OFFLOAD 3230 struct sk_buff_head xfrm_backlog; 3231 #endif 3232 /* written and read only by owning cpu: */ 3233 struct netdev_xmit xmit; 3234 #ifdef CONFIG_RPS 3235 /* input_queue_head should be written by cpu owning this struct, 3236 * and only read by other cpus. Worth using a cache line. 3237 */ 3238 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3239 3240 /* Elements below can be accessed between CPUs for RPS/RFS */ 3241 call_single_data_t csd ____cacheline_aligned_in_smp; 3242 struct softnet_data *rps_ipi_next; 3243 unsigned int cpu; 3244 unsigned int input_queue_tail; 3245 #endif 3246 struct sk_buff_head input_pkt_queue; 3247 struct napi_struct backlog; 3248 3249 atomic_t dropped ____cacheline_aligned_in_smp; 3250 3251 /* Another possibly contended cache line */ 3252 spinlock_t defer_lock ____cacheline_aligned_in_smp; 3253 int defer_count; 3254 int defer_ipi_scheduled; 3255 struct sk_buff *defer_list; 3256 call_single_data_t defer_csd; 3257 }; 3258 3259 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3260 3261 #ifndef CONFIG_PREEMPT_RT 3262 static inline int dev_recursion_level(void) 3263 { 3264 return this_cpu_read(softnet_data.xmit.recursion); 3265 } 3266 #else 3267 static inline int dev_recursion_level(void) 3268 { 3269 return current->net_xmit.recursion; 3270 } 3271 3272 #endif 3273 3274 void __netif_schedule(struct Qdisc *q); 3275 void netif_schedule_queue(struct netdev_queue *txq); 3276 3277 static inline void netif_tx_schedule_all(struct net_device *dev) 3278 { 3279 unsigned int i; 3280 3281 for (i = 0; i < dev->num_tx_queues; i++) 3282 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3283 } 3284 3285 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3286 { 3287 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3288 } 3289 3290 /** 3291 * netif_start_queue - allow transmit 3292 * @dev: network device 3293 * 3294 * Allow upper layers to call the device hard_start_xmit routine. 3295 */ 3296 static inline void netif_start_queue(struct net_device *dev) 3297 { 3298 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3299 } 3300 3301 static inline void netif_tx_start_all_queues(struct net_device *dev) 3302 { 3303 unsigned int i; 3304 3305 for (i = 0; i < dev->num_tx_queues; i++) { 3306 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3307 netif_tx_start_queue(txq); 3308 } 3309 } 3310 3311 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3312 3313 /** 3314 * netif_wake_queue - restart transmit 3315 * @dev: network device 3316 * 3317 * Allow upper layers to call the device hard_start_xmit routine. 3318 * Used for flow control when transmit resources are available. 3319 */ 3320 static inline void netif_wake_queue(struct net_device *dev) 3321 { 3322 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3323 } 3324 3325 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3326 { 3327 unsigned int i; 3328 3329 for (i = 0; i < dev->num_tx_queues; i++) { 3330 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3331 netif_tx_wake_queue(txq); 3332 } 3333 } 3334 3335 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3336 { 3337 /* Must be an atomic op see netif_txq_try_stop() */ 3338 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3339 } 3340 3341 /** 3342 * netif_stop_queue - stop transmitted packets 3343 * @dev: network device 3344 * 3345 * Stop upper layers calling the device hard_start_xmit routine. 3346 * Used for flow control when transmit resources are unavailable. 3347 */ 3348 static inline void netif_stop_queue(struct net_device *dev) 3349 { 3350 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3351 } 3352 3353 void netif_tx_stop_all_queues(struct net_device *dev); 3354 3355 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3356 { 3357 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3358 } 3359 3360 /** 3361 * netif_queue_stopped - test if transmit queue is flowblocked 3362 * @dev: network device 3363 * 3364 * Test if transmit queue on device is currently unable to send. 3365 */ 3366 static inline bool netif_queue_stopped(const struct net_device *dev) 3367 { 3368 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3369 } 3370 3371 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3372 { 3373 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3374 } 3375 3376 static inline bool 3377 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3378 { 3379 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3380 } 3381 3382 static inline bool 3383 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3384 { 3385 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3386 } 3387 3388 /** 3389 * netdev_queue_set_dql_min_limit - set dql minimum limit 3390 * @dev_queue: pointer to transmit queue 3391 * @min_limit: dql minimum limit 3392 * 3393 * Forces xmit_more() to return true until the minimum threshold 3394 * defined by @min_limit is reached (or until the tx queue is 3395 * empty). Warning: to be use with care, misuse will impact the 3396 * latency. 3397 */ 3398 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3399 unsigned int min_limit) 3400 { 3401 #ifdef CONFIG_BQL 3402 dev_queue->dql.min_limit = min_limit; 3403 #endif 3404 } 3405 3406 static inline int netdev_queue_dql_avail(const struct netdev_queue *txq) 3407 { 3408 #ifdef CONFIG_BQL 3409 /* Non-BQL migrated drivers will return 0, too. */ 3410 return dql_avail(&txq->dql); 3411 #else 3412 return 0; 3413 #endif 3414 } 3415 3416 /** 3417 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3418 * @dev_queue: pointer to transmit queue 3419 * 3420 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3421 * to give appropriate hint to the CPU. 3422 */ 3423 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3424 { 3425 #ifdef CONFIG_BQL 3426 prefetchw(&dev_queue->dql.num_queued); 3427 #endif 3428 } 3429 3430 /** 3431 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3432 * @dev_queue: pointer to transmit queue 3433 * 3434 * BQL enabled drivers might use this helper in their TX completion path, 3435 * to give appropriate hint to the CPU. 3436 */ 3437 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3438 { 3439 #ifdef CONFIG_BQL 3440 prefetchw(&dev_queue->dql.limit); 3441 #endif 3442 } 3443 3444 /** 3445 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3446 * @dev_queue: network device queue 3447 * @bytes: number of bytes queued to the device queue 3448 * 3449 * Report the number of bytes queued for sending/completion to the network 3450 * device hardware queue. @bytes should be a good approximation and should 3451 * exactly match netdev_completed_queue() @bytes. 3452 * This is typically called once per packet, from ndo_start_xmit(). 3453 */ 3454 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3455 unsigned int bytes) 3456 { 3457 #ifdef CONFIG_BQL 3458 dql_queued(&dev_queue->dql, bytes); 3459 3460 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3461 return; 3462 3463 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3464 3465 /* 3466 * The XOFF flag must be set before checking the dql_avail below, 3467 * because in netdev_tx_completed_queue we update the dql_completed 3468 * before checking the XOFF flag. 3469 */ 3470 smp_mb(); 3471 3472 /* check again in case another CPU has just made room avail */ 3473 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3474 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3475 #endif 3476 } 3477 3478 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3479 * that they should not test BQL status themselves. 3480 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3481 * skb of a batch. 3482 * Returns true if the doorbell must be used to kick the NIC. 3483 */ 3484 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3485 unsigned int bytes, 3486 bool xmit_more) 3487 { 3488 if (xmit_more) { 3489 #ifdef CONFIG_BQL 3490 dql_queued(&dev_queue->dql, bytes); 3491 #endif 3492 return netif_tx_queue_stopped(dev_queue); 3493 } 3494 netdev_tx_sent_queue(dev_queue, bytes); 3495 return true; 3496 } 3497 3498 /** 3499 * netdev_sent_queue - report the number of bytes queued to hardware 3500 * @dev: network device 3501 * @bytes: number of bytes queued to the hardware device queue 3502 * 3503 * Report the number of bytes queued for sending/completion to the network 3504 * device hardware queue#0. @bytes should be a good approximation and should 3505 * exactly match netdev_completed_queue() @bytes. 3506 * This is typically called once per packet, from ndo_start_xmit(). 3507 */ 3508 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3509 { 3510 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3511 } 3512 3513 static inline bool __netdev_sent_queue(struct net_device *dev, 3514 unsigned int bytes, 3515 bool xmit_more) 3516 { 3517 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3518 xmit_more); 3519 } 3520 3521 /** 3522 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3523 * @dev_queue: network device queue 3524 * @pkts: number of packets (currently ignored) 3525 * @bytes: number of bytes dequeued from the device queue 3526 * 3527 * Must be called at most once per TX completion round (and not per 3528 * individual packet), so that BQL can adjust its limits appropriately. 3529 */ 3530 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3531 unsigned int pkts, unsigned int bytes) 3532 { 3533 #ifdef CONFIG_BQL 3534 if (unlikely(!bytes)) 3535 return; 3536 3537 dql_completed(&dev_queue->dql, bytes); 3538 3539 /* 3540 * Without the memory barrier there is a small possiblity that 3541 * netdev_tx_sent_queue will miss the update and cause the queue to 3542 * be stopped forever 3543 */ 3544 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */ 3545 3546 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3547 return; 3548 3549 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3550 netif_schedule_queue(dev_queue); 3551 #endif 3552 } 3553 3554 /** 3555 * netdev_completed_queue - report bytes and packets completed by device 3556 * @dev: network device 3557 * @pkts: actual number of packets sent over the medium 3558 * @bytes: actual number of bytes sent over the medium 3559 * 3560 * Report the number of bytes and packets transmitted by the network device 3561 * hardware queue over the physical medium, @bytes must exactly match the 3562 * @bytes amount passed to netdev_sent_queue() 3563 */ 3564 static inline void netdev_completed_queue(struct net_device *dev, 3565 unsigned int pkts, unsigned int bytes) 3566 { 3567 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3568 } 3569 3570 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3571 { 3572 #ifdef CONFIG_BQL 3573 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3574 dql_reset(&q->dql); 3575 #endif 3576 } 3577 3578 /** 3579 * netdev_reset_queue - reset the packets and bytes count of a network device 3580 * @dev_queue: network device 3581 * 3582 * Reset the bytes and packet count of a network device and clear the 3583 * software flow control OFF bit for this network device 3584 */ 3585 static inline void netdev_reset_queue(struct net_device *dev_queue) 3586 { 3587 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3588 } 3589 3590 /** 3591 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3592 * @dev: network device 3593 * @queue_index: given tx queue index 3594 * 3595 * Returns 0 if given tx queue index >= number of device tx queues, 3596 * otherwise returns the originally passed tx queue index. 3597 */ 3598 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3599 { 3600 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3601 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3602 dev->name, queue_index, 3603 dev->real_num_tx_queues); 3604 return 0; 3605 } 3606 3607 return queue_index; 3608 } 3609 3610 /** 3611 * netif_running - test if up 3612 * @dev: network device 3613 * 3614 * Test if the device has been brought up. 3615 */ 3616 static inline bool netif_running(const struct net_device *dev) 3617 { 3618 return test_bit(__LINK_STATE_START, &dev->state); 3619 } 3620 3621 /* 3622 * Routines to manage the subqueues on a device. We only need start, 3623 * stop, and a check if it's stopped. All other device management is 3624 * done at the overall netdevice level. 3625 * Also test the device if we're multiqueue. 3626 */ 3627 3628 /** 3629 * netif_start_subqueue - allow sending packets on subqueue 3630 * @dev: network device 3631 * @queue_index: sub queue index 3632 * 3633 * Start individual transmit queue of a device with multiple transmit queues. 3634 */ 3635 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3636 { 3637 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3638 3639 netif_tx_start_queue(txq); 3640 } 3641 3642 /** 3643 * netif_stop_subqueue - stop sending packets on subqueue 3644 * @dev: network device 3645 * @queue_index: sub queue index 3646 * 3647 * Stop individual transmit queue of a device with multiple transmit queues. 3648 */ 3649 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3650 { 3651 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3652 netif_tx_stop_queue(txq); 3653 } 3654 3655 /** 3656 * __netif_subqueue_stopped - test status of subqueue 3657 * @dev: network device 3658 * @queue_index: sub queue index 3659 * 3660 * Check individual transmit queue of a device with multiple transmit queues. 3661 */ 3662 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3663 u16 queue_index) 3664 { 3665 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3666 3667 return netif_tx_queue_stopped(txq); 3668 } 3669 3670 /** 3671 * netif_subqueue_stopped - test status of subqueue 3672 * @dev: network device 3673 * @skb: sub queue buffer pointer 3674 * 3675 * Check individual transmit queue of a device with multiple transmit queues. 3676 */ 3677 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3678 struct sk_buff *skb) 3679 { 3680 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3681 } 3682 3683 /** 3684 * netif_wake_subqueue - allow sending packets on subqueue 3685 * @dev: network device 3686 * @queue_index: sub queue index 3687 * 3688 * Resume individual transmit queue of a device with multiple transmit queues. 3689 */ 3690 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3691 { 3692 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3693 3694 netif_tx_wake_queue(txq); 3695 } 3696 3697 #ifdef CONFIG_XPS 3698 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3699 u16 index); 3700 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3701 u16 index, enum xps_map_type type); 3702 3703 /** 3704 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3705 * @j: CPU/Rx queue index 3706 * @mask: bitmask of all cpus/rx queues 3707 * @nr_bits: number of bits in the bitmask 3708 * 3709 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3710 */ 3711 static inline bool netif_attr_test_mask(unsigned long j, 3712 const unsigned long *mask, 3713 unsigned int nr_bits) 3714 { 3715 cpu_max_bits_warn(j, nr_bits); 3716 return test_bit(j, mask); 3717 } 3718 3719 /** 3720 * netif_attr_test_online - Test for online CPU/Rx queue 3721 * @j: CPU/Rx queue index 3722 * @online_mask: bitmask for CPUs/Rx queues that are online 3723 * @nr_bits: number of bits in the bitmask 3724 * 3725 * Returns true if a CPU/Rx queue is online. 3726 */ 3727 static inline bool netif_attr_test_online(unsigned long j, 3728 const unsigned long *online_mask, 3729 unsigned int nr_bits) 3730 { 3731 cpu_max_bits_warn(j, nr_bits); 3732 3733 if (online_mask) 3734 return test_bit(j, online_mask); 3735 3736 return (j < nr_bits); 3737 } 3738 3739 /** 3740 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3741 * @n: CPU/Rx queue index 3742 * @srcp: the cpumask/Rx queue mask pointer 3743 * @nr_bits: number of bits in the bitmask 3744 * 3745 * Returns >= nr_bits if no further CPUs/Rx queues set. 3746 */ 3747 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3748 unsigned int nr_bits) 3749 { 3750 /* -1 is a legal arg here. */ 3751 if (n != -1) 3752 cpu_max_bits_warn(n, nr_bits); 3753 3754 if (srcp) 3755 return find_next_bit(srcp, nr_bits, n + 1); 3756 3757 return n + 1; 3758 } 3759 3760 /** 3761 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3762 * @n: CPU/Rx queue index 3763 * @src1p: the first CPUs/Rx queues mask pointer 3764 * @src2p: the second CPUs/Rx queues mask pointer 3765 * @nr_bits: number of bits in the bitmask 3766 * 3767 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3768 */ 3769 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3770 const unsigned long *src2p, 3771 unsigned int nr_bits) 3772 { 3773 /* -1 is a legal arg here. */ 3774 if (n != -1) 3775 cpu_max_bits_warn(n, nr_bits); 3776 3777 if (src1p && src2p) 3778 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3779 else if (src1p) 3780 return find_next_bit(src1p, nr_bits, n + 1); 3781 else if (src2p) 3782 return find_next_bit(src2p, nr_bits, n + 1); 3783 3784 return n + 1; 3785 } 3786 #else 3787 static inline int netif_set_xps_queue(struct net_device *dev, 3788 const struct cpumask *mask, 3789 u16 index) 3790 { 3791 return 0; 3792 } 3793 3794 static inline int __netif_set_xps_queue(struct net_device *dev, 3795 const unsigned long *mask, 3796 u16 index, enum xps_map_type type) 3797 { 3798 return 0; 3799 } 3800 #endif 3801 3802 /** 3803 * netif_is_multiqueue - test if device has multiple transmit queues 3804 * @dev: network device 3805 * 3806 * Check if device has multiple transmit queues 3807 */ 3808 static inline bool netif_is_multiqueue(const struct net_device *dev) 3809 { 3810 return dev->num_tx_queues > 1; 3811 } 3812 3813 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3814 3815 #ifdef CONFIG_SYSFS 3816 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3817 #else 3818 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3819 unsigned int rxqs) 3820 { 3821 dev->real_num_rx_queues = rxqs; 3822 return 0; 3823 } 3824 #endif 3825 int netif_set_real_num_queues(struct net_device *dev, 3826 unsigned int txq, unsigned int rxq); 3827 3828 int netif_get_num_default_rss_queues(void); 3829 3830 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason); 3831 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason); 3832 3833 /* 3834 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3835 * interrupt context or with hardware interrupts being disabled. 3836 * (in_hardirq() || irqs_disabled()) 3837 * 3838 * We provide four helpers that can be used in following contexts : 3839 * 3840 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3841 * replacing kfree_skb(skb) 3842 * 3843 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3844 * Typically used in place of consume_skb(skb) in TX completion path 3845 * 3846 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3847 * replacing kfree_skb(skb) 3848 * 3849 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3850 * and consumed a packet. Used in place of consume_skb(skb) 3851 */ 3852 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3853 { 3854 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 3855 } 3856 3857 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3858 { 3859 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED); 3860 } 3861 3862 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3863 { 3864 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 3865 } 3866 3867 static inline void dev_consume_skb_any(struct sk_buff *skb) 3868 { 3869 dev_kfree_skb_any_reason(skb, SKB_CONSUMED); 3870 } 3871 3872 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3873 struct bpf_prog *xdp_prog); 3874 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3875 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff **pskb); 3876 int netif_rx(struct sk_buff *skb); 3877 int __netif_rx(struct sk_buff *skb); 3878 3879 int netif_receive_skb(struct sk_buff *skb); 3880 int netif_receive_skb_core(struct sk_buff *skb); 3881 void netif_receive_skb_list_internal(struct list_head *head); 3882 void netif_receive_skb_list(struct list_head *head); 3883 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3884 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3885 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3886 void napi_get_frags_check(struct napi_struct *napi); 3887 gro_result_t napi_gro_frags(struct napi_struct *napi); 3888 3889 static inline void napi_free_frags(struct napi_struct *napi) 3890 { 3891 kfree_skb(napi->skb); 3892 napi->skb = NULL; 3893 } 3894 3895 bool netdev_is_rx_handler_busy(struct net_device *dev); 3896 int netdev_rx_handler_register(struct net_device *dev, 3897 rx_handler_func_t *rx_handler, 3898 void *rx_handler_data); 3899 void netdev_rx_handler_unregister(struct net_device *dev); 3900 3901 bool dev_valid_name(const char *name); 3902 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 3903 { 3904 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 3905 } 3906 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 3907 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 3908 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3909 void __user *data, bool *need_copyout); 3910 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 3911 int generic_hwtstamp_get_lower(struct net_device *dev, 3912 struct kernel_hwtstamp_config *kernel_cfg); 3913 int generic_hwtstamp_set_lower(struct net_device *dev, 3914 struct kernel_hwtstamp_config *kernel_cfg, 3915 struct netlink_ext_ack *extack); 3916 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 3917 unsigned int dev_get_flags(const struct net_device *); 3918 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3919 struct netlink_ext_ack *extack); 3920 int dev_change_flags(struct net_device *dev, unsigned int flags, 3921 struct netlink_ext_ack *extack); 3922 int dev_set_alias(struct net_device *, const char *, size_t); 3923 int dev_get_alias(const struct net_device *, char *, size_t); 3924 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 3925 const char *pat, int new_ifindex); 3926 static inline 3927 int dev_change_net_namespace(struct net_device *dev, struct net *net, 3928 const char *pat) 3929 { 3930 return __dev_change_net_namespace(dev, net, pat, 0); 3931 } 3932 int __dev_set_mtu(struct net_device *, int); 3933 int dev_set_mtu(struct net_device *, int); 3934 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3935 struct netlink_ext_ack *extack); 3936 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3937 struct netlink_ext_ack *extack); 3938 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 3939 struct netlink_ext_ack *extack); 3940 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 3941 int dev_get_port_parent_id(struct net_device *dev, 3942 struct netdev_phys_item_id *ppid, bool recurse); 3943 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3944 3945 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3946 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3947 struct netdev_queue *txq, int *ret); 3948 3949 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3950 u8 dev_xdp_prog_count(struct net_device *dev); 3951 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3952 3953 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3954 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3955 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 3956 bool is_skb_forwardable(const struct net_device *dev, 3957 const struct sk_buff *skb); 3958 3959 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 3960 const struct sk_buff *skb, 3961 const bool check_mtu) 3962 { 3963 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 3964 unsigned int len; 3965 3966 if (!(dev->flags & IFF_UP)) 3967 return false; 3968 3969 if (!check_mtu) 3970 return true; 3971 3972 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 3973 if (skb->len <= len) 3974 return true; 3975 3976 /* if TSO is enabled, we don't care about the length as the packet 3977 * could be forwarded without being segmented before 3978 */ 3979 if (skb_is_gso(skb)) 3980 return true; 3981 3982 return false; 3983 } 3984 3985 void netdev_core_stats_inc(struct net_device *dev, u32 offset); 3986 3987 #define DEV_CORE_STATS_INC(FIELD) \ 3988 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 3989 { \ 3990 netdev_core_stats_inc(dev, \ 3991 offsetof(struct net_device_core_stats, FIELD)); \ 3992 } 3993 DEV_CORE_STATS_INC(rx_dropped) 3994 DEV_CORE_STATS_INC(tx_dropped) 3995 DEV_CORE_STATS_INC(rx_nohandler) 3996 DEV_CORE_STATS_INC(rx_otherhost_dropped) 3997 #undef DEV_CORE_STATS_INC 3998 3999 static __always_inline int ____dev_forward_skb(struct net_device *dev, 4000 struct sk_buff *skb, 4001 const bool check_mtu) 4002 { 4003 if (skb_orphan_frags(skb, GFP_ATOMIC) || 4004 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 4005 dev_core_stats_rx_dropped_inc(dev); 4006 kfree_skb(skb); 4007 return NET_RX_DROP; 4008 } 4009 4010 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 4011 skb->priority = 0; 4012 return 0; 4013 } 4014 4015 bool dev_nit_active(struct net_device *dev); 4016 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 4017 4018 static inline void __dev_put(struct net_device *dev) 4019 { 4020 if (dev) { 4021 #ifdef CONFIG_PCPU_DEV_REFCNT 4022 this_cpu_dec(*dev->pcpu_refcnt); 4023 #else 4024 refcount_dec(&dev->dev_refcnt); 4025 #endif 4026 } 4027 } 4028 4029 static inline void __dev_hold(struct net_device *dev) 4030 { 4031 if (dev) { 4032 #ifdef CONFIG_PCPU_DEV_REFCNT 4033 this_cpu_inc(*dev->pcpu_refcnt); 4034 #else 4035 refcount_inc(&dev->dev_refcnt); 4036 #endif 4037 } 4038 } 4039 4040 static inline void __netdev_tracker_alloc(struct net_device *dev, 4041 netdevice_tracker *tracker, 4042 gfp_t gfp) 4043 { 4044 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4045 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 4046 #endif 4047 } 4048 4049 /* netdev_tracker_alloc() can upgrade a prior untracked reference 4050 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4051 */ 4052 static inline void netdev_tracker_alloc(struct net_device *dev, 4053 netdevice_tracker *tracker, gfp_t gfp) 4054 { 4055 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4056 refcount_dec(&dev->refcnt_tracker.no_tracker); 4057 __netdev_tracker_alloc(dev, tracker, gfp); 4058 #endif 4059 } 4060 4061 static inline void netdev_tracker_free(struct net_device *dev, 4062 netdevice_tracker *tracker) 4063 { 4064 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4065 ref_tracker_free(&dev->refcnt_tracker, tracker); 4066 #endif 4067 } 4068 4069 static inline void netdev_hold(struct net_device *dev, 4070 netdevice_tracker *tracker, gfp_t gfp) 4071 { 4072 if (dev) { 4073 __dev_hold(dev); 4074 __netdev_tracker_alloc(dev, tracker, gfp); 4075 } 4076 } 4077 4078 static inline void netdev_put(struct net_device *dev, 4079 netdevice_tracker *tracker) 4080 { 4081 if (dev) { 4082 netdev_tracker_free(dev, tracker); 4083 __dev_put(dev); 4084 } 4085 } 4086 4087 /** 4088 * dev_hold - get reference to device 4089 * @dev: network device 4090 * 4091 * Hold reference to device to keep it from being freed. 4092 * Try using netdev_hold() instead. 4093 */ 4094 static inline void dev_hold(struct net_device *dev) 4095 { 4096 netdev_hold(dev, NULL, GFP_ATOMIC); 4097 } 4098 4099 /** 4100 * dev_put - release reference to device 4101 * @dev: network device 4102 * 4103 * Release reference to device to allow it to be freed. 4104 * Try using netdev_put() instead. 4105 */ 4106 static inline void dev_put(struct net_device *dev) 4107 { 4108 netdev_put(dev, NULL); 4109 } 4110 4111 DEFINE_FREE(dev_put, struct net_device *, if (_T) dev_put(_T)) 4112 4113 static inline void netdev_ref_replace(struct net_device *odev, 4114 struct net_device *ndev, 4115 netdevice_tracker *tracker, 4116 gfp_t gfp) 4117 { 4118 if (odev) 4119 netdev_tracker_free(odev, tracker); 4120 4121 __dev_hold(ndev); 4122 __dev_put(odev); 4123 4124 if (ndev) 4125 __netdev_tracker_alloc(ndev, tracker, gfp); 4126 } 4127 4128 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4129 * and _off may be called from IRQ context, but it is caller 4130 * who is responsible for serialization of these calls. 4131 * 4132 * The name carrier is inappropriate, these functions should really be 4133 * called netif_lowerlayer_*() because they represent the state of any 4134 * kind of lower layer not just hardware media. 4135 */ 4136 void linkwatch_fire_event(struct net_device *dev); 4137 4138 /** 4139 * linkwatch_sync_dev - sync linkwatch for the given device 4140 * @dev: network device to sync linkwatch for 4141 * 4142 * Sync linkwatch for the given device, removing it from the 4143 * pending work list (if queued). 4144 */ 4145 void linkwatch_sync_dev(struct net_device *dev); 4146 4147 /** 4148 * netif_carrier_ok - test if carrier present 4149 * @dev: network device 4150 * 4151 * Check if carrier is present on device 4152 */ 4153 static inline bool netif_carrier_ok(const struct net_device *dev) 4154 { 4155 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4156 } 4157 4158 unsigned long dev_trans_start(struct net_device *dev); 4159 4160 void __netdev_watchdog_up(struct net_device *dev); 4161 4162 void netif_carrier_on(struct net_device *dev); 4163 void netif_carrier_off(struct net_device *dev); 4164 void netif_carrier_event(struct net_device *dev); 4165 4166 /** 4167 * netif_dormant_on - mark device as dormant. 4168 * @dev: network device 4169 * 4170 * Mark device as dormant (as per RFC2863). 4171 * 4172 * The dormant state indicates that the relevant interface is not 4173 * actually in a condition to pass packets (i.e., it is not 'up') but is 4174 * in a "pending" state, waiting for some external event. For "on- 4175 * demand" interfaces, this new state identifies the situation where the 4176 * interface is waiting for events to place it in the up state. 4177 */ 4178 static inline void netif_dormant_on(struct net_device *dev) 4179 { 4180 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4181 linkwatch_fire_event(dev); 4182 } 4183 4184 /** 4185 * netif_dormant_off - set device as not dormant. 4186 * @dev: network device 4187 * 4188 * Device is not in dormant state. 4189 */ 4190 static inline void netif_dormant_off(struct net_device *dev) 4191 { 4192 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4193 linkwatch_fire_event(dev); 4194 } 4195 4196 /** 4197 * netif_dormant - test if device is dormant 4198 * @dev: network device 4199 * 4200 * Check if device is dormant. 4201 */ 4202 static inline bool netif_dormant(const struct net_device *dev) 4203 { 4204 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4205 } 4206 4207 4208 /** 4209 * netif_testing_on - mark device as under test. 4210 * @dev: network device 4211 * 4212 * Mark device as under test (as per RFC2863). 4213 * 4214 * The testing state indicates that some test(s) must be performed on 4215 * the interface. After completion, of the test, the interface state 4216 * will change to up, dormant, or down, as appropriate. 4217 */ 4218 static inline void netif_testing_on(struct net_device *dev) 4219 { 4220 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4221 linkwatch_fire_event(dev); 4222 } 4223 4224 /** 4225 * netif_testing_off - set device as not under test. 4226 * @dev: network device 4227 * 4228 * Device is not in testing state. 4229 */ 4230 static inline void netif_testing_off(struct net_device *dev) 4231 { 4232 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4233 linkwatch_fire_event(dev); 4234 } 4235 4236 /** 4237 * netif_testing - test if device is under test 4238 * @dev: network device 4239 * 4240 * Check if device is under test 4241 */ 4242 static inline bool netif_testing(const struct net_device *dev) 4243 { 4244 return test_bit(__LINK_STATE_TESTING, &dev->state); 4245 } 4246 4247 4248 /** 4249 * netif_oper_up - test if device is operational 4250 * @dev: network device 4251 * 4252 * Check if carrier is operational 4253 */ 4254 static inline bool netif_oper_up(const struct net_device *dev) 4255 { 4256 unsigned int operstate = READ_ONCE(dev->operstate); 4257 4258 return operstate == IF_OPER_UP || 4259 operstate == IF_OPER_UNKNOWN /* backward compat */; 4260 } 4261 4262 /** 4263 * netif_device_present - is device available or removed 4264 * @dev: network device 4265 * 4266 * Check if device has not been removed from system. 4267 */ 4268 static inline bool netif_device_present(const struct net_device *dev) 4269 { 4270 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4271 } 4272 4273 void netif_device_detach(struct net_device *dev); 4274 4275 void netif_device_attach(struct net_device *dev); 4276 4277 /* 4278 * Network interface message level settings 4279 */ 4280 4281 enum { 4282 NETIF_MSG_DRV_BIT, 4283 NETIF_MSG_PROBE_BIT, 4284 NETIF_MSG_LINK_BIT, 4285 NETIF_MSG_TIMER_BIT, 4286 NETIF_MSG_IFDOWN_BIT, 4287 NETIF_MSG_IFUP_BIT, 4288 NETIF_MSG_RX_ERR_BIT, 4289 NETIF_MSG_TX_ERR_BIT, 4290 NETIF_MSG_TX_QUEUED_BIT, 4291 NETIF_MSG_INTR_BIT, 4292 NETIF_MSG_TX_DONE_BIT, 4293 NETIF_MSG_RX_STATUS_BIT, 4294 NETIF_MSG_PKTDATA_BIT, 4295 NETIF_MSG_HW_BIT, 4296 NETIF_MSG_WOL_BIT, 4297 4298 /* When you add a new bit above, update netif_msg_class_names array 4299 * in net/ethtool/common.c 4300 */ 4301 NETIF_MSG_CLASS_COUNT, 4302 }; 4303 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4304 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4305 4306 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4307 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4308 4309 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4310 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4311 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4312 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4313 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4314 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4315 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4316 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4317 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4318 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4319 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4320 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4321 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4322 #define NETIF_MSG_HW __NETIF_MSG(HW) 4323 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4324 4325 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4326 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4327 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4328 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4329 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4330 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4331 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4332 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4333 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4334 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4335 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4336 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4337 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4338 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4339 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4340 4341 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4342 { 4343 /* use default */ 4344 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4345 return default_msg_enable_bits; 4346 if (debug_value == 0) /* no output */ 4347 return 0; 4348 /* set low N bits */ 4349 return (1U << debug_value) - 1; 4350 } 4351 4352 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4353 { 4354 spin_lock(&txq->_xmit_lock); 4355 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4356 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4357 } 4358 4359 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4360 { 4361 __acquire(&txq->_xmit_lock); 4362 return true; 4363 } 4364 4365 static inline void __netif_tx_release(struct netdev_queue *txq) 4366 { 4367 __release(&txq->_xmit_lock); 4368 } 4369 4370 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4371 { 4372 spin_lock_bh(&txq->_xmit_lock); 4373 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4374 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4375 } 4376 4377 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4378 { 4379 bool ok = spin_trylock(&txq->_xmit_lock); 4380 4381 if (likely(ok)) { 4382 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4383 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4384 } 4385 return ok; 4386 } 4387 4388 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4389 { 4390 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4391 WRITE_ONCE(txq->xmit_lock_owner, -1); 4392 spin_unlock(&txq->_xmit_lock); 4393 } 4394 4395 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4396 { 4397 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4398 WRITE_ONCE(txq->xmit_lock_owner, -1); 4399 spin_unlock_bh(&txq->_xmit_lock); 4400 } 4401 4402 /* 4403 * txq->trans_start can be read locklessly from dev_watchdog() 4404 */ 4405 static inline void txq_trans_update(struct netdev_queue *txq) 4406 { 4407 if (txq->xmit_lock_owner != -1) 4408 WRITE_ONCE(txq->trans_start, jiffies); 4409 } 4410 4411 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4412 { 4413 unsigned long now = jiffies; 4414 4415 if (READ_ONCE(txq->trans_start) != now) 4416 WRITE_ONCE(txq->trans_start, now); 4417 } 4418 4419 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4420 static inline void netif_trans_update(struct net_device *dev) 4421 { 4422 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4423 4424 txq_trans_cond_update(txq); 4425 } 4426 4427 /** 4428 * netif_tx_lock - grab network device transmit lock 4429 * @dev: network device 4430 * 4431 * Get network device transmit lock 4432 */ 4433 void netif_tx_lock(struct net_device *dev); 4434 4435 static inline void netif_tx_lock_bh(struct net_device *dev) 4436 { 4437 local_bh_disable(); 4438 netif_tx_lock(dev); 4439 } 4440 4441 void netif_tx_unlock(struct net_device *dev); 4442 4443 static inline void netif_tx_unlock_bh(struct net_device *dev) 4444 { 4445 netif_tx_unlock(dev); 4446 local_bh_enable(); 4447 } 4448 4449 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4450 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4451 __netif_tx_lock(txq, cpu); \ 4452 } else { \ 4453 __netif_tx_acquire(txq); \ 4454 } \ 4455 } 4456 4457 #define HARD_TX_TRYLOCK(dev, txq) \ 4458 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4459 __netif_tx_trylock(txq) : \ 4460 __netif_tx_acquire(txq)) 4461 4462 #define HARD_TX_UNLOCK(dev, txq) { \ 4463 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4464 __netif_tx_unlock(txq); \ 4465 } else { \ 4466 __netif_tx_release(txq); \ 4467 } \ 4468 } 4469 4470 static inline void netif_tx_disable(struct net_device *dev) 4471 { 4472 unsigned int i; 4473 int cpu; 4474 4475 local_bh_disable(); 4476 cpu = smp_processor_id(); 4477 spin_lock(&dev->tx_global_lock); 4478 for (i = 0; i < dev->num_tx_queues; i++) { 4479 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4480 4481 __netif_tx_lock(txq, cpu); 4482 netif_tx_stop_queue(txq); 4483 __netif_tx_unlock(txq); 4484 } 4485 spin_unlock(&dev->tx_global_lock); 4486 local_bh_enable(); 4487 } 4488 4489 static inline void netif_addr_lock(struct net_device *dev) 4490 { 4491 unsigned char nest_level = 0; 4492 4493 #ifdef CONFIG_LOCKDEP 4494 nest_level = dev->nested_level; 4495 #endif 4496 spin_lock_nested(&dev->addr_list_lock, nest_level); 4497 } 4498 4499 static inline void netif_addr_lock_bh(struct net_device *dev) 4500 { 4501 unsigned char nest_level = 0; 4502 4503 #ifdef CONFIG_LOCKDEP 4504 nest_level = dev->nested_level; 4505 #endif 4506 local_bh_disable(); 4507 spin_lock_nested(&dev->addr_list_lock, nest_level); 4508 } 4509 4510 static inline void netif_addr_unlock(struct net_device *dev) 4511 { 4512 spin_unlock(&dev->addr_list_lock); 4513 } 4514 4515 static inline void netif_addr_unlock_bh(struct net_device *dev) 4516 { 4517 spin_unlock_bh(&dev->addr_list_lock); 4518 } 4519 4520 /* 4521 * dev_addrs walker. Should be used only for read access. Call with 4522 * rcu_read_lock held. 4523 */ 4524 #define for_each_dev_addr(dev, ha) \ 4525 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4526 4527 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4528 4529 void ether_setup(struct net_device *dev); 4530 4531 /* Allocate dummy net_device */ 4532 struct net_device *alloc_netdev_dummy(int sizeof_priv); 4533 4534 /* Support for loadable net-drivers */ 4535 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4536 unsigned char name_assign_type, 4537 void (*setup)(struct net_device *), 4538 unsigned int txqs, unsigned int rxqs); 4539 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4540 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4541 4542 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4543 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4544 count) 4545 4546 int register_netdev(struct net_device *dev); 4547 void unregister_netdev(struct net_device *dev); 4548 4549 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4550 4551 /* General hardware address lists handling functions */ 4552 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4553 struct netdev_hw_addr_list *from_list, int addr_len); 4554 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4555 struct netdev_hw_addr_list *from_list, int addr_len); 4556 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4557 struct net_device *dev, 4558 int (*sync)(struct net_device *, const unsigned char *), 4559 int (*unsync)(struct net_device *, 4560 const unsigned char *)); 4561 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4562 struct net_device *dev, 4563 int (*sync)(struct net_device *, 4564 const unsigned char *, int), 4565 int (*unsync)(struct net_device *, 4566 const unsigned char *, int)); 4567 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4568 struct net_device *dev, 4569 int (*unsync)(struct net_device *, 4570 const unsigned char *, int)); 4571 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4572 struct net_device *dev, 4573 int (*unsync)(struct net_device *, 4574 const unsigned char *)); 4575 void __hw_addr_init(struct netdev_hw_addr_list *list); 4576 4577 /* Functions used for device addresses handling */ 4578 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4579 const void *addr, size_t len); 4580 4581 static inline void 4582 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4583 { 4584 dev_addr_mod(dev, 0, addr, len); 4585 } 4586 4587 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4588 { 4589 __dev_addr_set(dev, addr, dev->addr_len); 4590 } 4591 4592 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4593 unsigned char addr_type); 4594 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4595 unsigned char addr_type); 4596 4597 /* Functions used for unicast addresses handling */ 4598 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4599 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4600 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4601 int dev_uc_sync(struct net_device *to, struct net_device *from); 4602 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4603 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4604 void dev_uc_flush(struct net_device *dev); 4605 void dev_uc_init(struct net_device *dev); 4606 4607 /** 4608 * __dev_uc_sync - Synchonize device's unicast list 4609 * @dev: device to sync 4610 * @sync: function to call if address should be added 4611 * @unsync: function to call if address should be removed 4612 * 4613 * Add newly added addresses to the interface, and release 4614 * addresses that have been deleted. 4615 */ 4616 static inline int __dev_uc_sync(struct net_device *dev, 4617 int (*sync)(struct net_device *, 4618 const unsigned char *), 4619 int (*unsync)(struct net_device *, 4620 const unsigned char *)) 4621 { 4622 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4623 } 4624 4625 /** 4626 * __dev_uc_unsync - Remove synchronized addresses from device 4627 * @dev: device to sync 4628 * @unsync: function to call if address should be removed 4629 * 4630 * Remove all addresses that were added to the device by dev_uc_sync(). 4631 */ 4632 static inline void __dev_uc_unsync(struct net_device *dev, 4633 int (*unsync)(struct net_device *, 4634 const unsigned char *)) 4635 { 4636 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4637 } 4638 4639 /* Functions used for multicast addresses handling */ 4640 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4641 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4642 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4643 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4644 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4645 int dev_mc_sync(struct net_device *to, struct net_device *from); 4646 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4647 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4648 void dev_mc_flush(struct net_device *dev); 4649 void dev_mc_init(struct net_device *dev); 4650 4651 /** 4652 * __dev_mc_sync - Synchonize device's multicast list 4653 * @dev: device to sync 4654 * @sync: function to call if address should be added 4655 * @unsync: function to call if address should be removed 4656 * 4657 * Add newly added addresses to the interface, and release 4658 * addresses that have been deleted. 4659 */ 4660 static inline int __dev_mc_sync(struct net_device *dev, 4661 int (*sync)(struct net_device *, 4662 const unsigned char *), 4663 int (*unsync)(struct net_device *, 4664 const unsigned char *)) 4665 { 4666 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4667 } 4668 4669 /** 4670 * __dev_mc_unsync - Remove synchronized addresses from device 4671 * @dev: device to sync 4672 * @unsync: function to call if address should be removed 4673 * 4674 * Remove all addresses that were added to the device by dev_mc_sync(). 4675 */ 4676 static inline void __dev_mc_unsync(struct net_device *dev, 4677 int (*unsync)(struct net_device *, 4678 const unsigned char *)) 4679 { 4680 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4681 } 4682 4683 /* Functions used for secondary unicast and multicast support */ 4684 void dev_set_rx_mode(struct net_device *dev); 4685 int dev_set_promiscuity(struct net_device *dev, int inc); 4686 int dev_set_allmulti(struct net_device *dev, int inc); 4687 void netdev_state_change(struct net_device *dev); 4688 void __netdev_notify_peers(struct net_device *dev); 4689 void netdev_notify_peers(struct net_device *dev); 4690 void netdev_features_change(struct net_device *dev); 4691 /* Load a device via the kmod */ 4692 void dev_load(struct net *net, const char *name); 4693 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4694 struct rtnl_link_stats64 *storage); 4695 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4696 const struct net_device_stats *netdev_stats); 4697 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4698 const struct pcpu_sw_netstats __percpu *netstats); 4699 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4700 4701 enum { 4702 NESTED_SYNC_IMM_BIT, 4703 NESTED_SYNC_TODO_BIT, 4704 }; 4705 4706 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4707 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4708 4709 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4710 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4711 4712 struct netdev_nested_priv { 4713 unsigned char flags; 4714 void *data; 4715 }; 4716 4717 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4718 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4719 struct list_head **iter); 4720 4721 /* iterate through upper list, must be called under RCU read lock */ 4722 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4723 for (iter = &(dev)->adj_list.upper, \ 4724 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4725 updev; \ 4726 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4727 4728 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4729 int (*fn)(struct net_device *upper_dev, 4730 struct netdev_nested_priv *priv), 4731 struct netdev_nested_priv *priv); 4732 4733 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4734 struct net_device *upper_dev); 4735 4736 bool netdev_has_any_upper_dev(struct net_device *dev); 4737 4738 void *netdev_lower_get_next_private(struct net_device *dev, 4739 struct list_head **iter); 4740 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4741 struct list_head **iter); 4742 4743 #define netdev_for_each_lower_private(dev, priv, iter) \ 4744 for (iter = (dev)->adj_list.lower.next, \ 4745 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4746 priv; \ 4747 priv = netdev_lower_get_next_private(dev, &(iter))) 4748 4749 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4750 for (iter = &(dev)->adj_list.lower, \ 4751 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4752 priv; \ 4753 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4754 4755 void *netdev_lower_get_next(struct net_device *dev, 4756 struct list_head **iter); 4757 4758 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4759 for (iter = (dev)->adj_list.lower.next, \ 4760 ldev = netdev_lower_get_next(dev, &(iter)); \ 4761 ldev; \ 4762 ldev = netdev_lower_get_next(dev, &(iter))) 4763 4764 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4765 struct list_head **iter); 4766 int netdev_walk_all_lower_dev(struct net_device *dev, 4767 int (*fn)(struct net_device *lower_dev, 4768 struct netdev_nested_priv *priv), 4769 struct netdev_nested_priv *priv); 4770 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4771 int (*fn)(struct net_device *lower_dev, 4772 struct netdev_nested_priv *priv), 4773 struct netdev_nested_priv *priv); 4774 4775 void *netdev_adjacent_get_private(struct list_head *adj_list); 4776 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4777 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4778 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4779 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4780 struct netlink_ext_ack *extack); 4781 int netdev_master_upper_dev_link(struct net_device *dev, 4782 struct net_device *upper_dev, 4783 void *upper_priv, void *upper_info, 4784 struct netlink_ext_ack *extack); 4785 void netdev_upper_dev_unlink(struct net_device *dev, 4786 struct net_device *upper_dev); 4787 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4788 struct net_device *new_dev, 4789 struct net_device *dev, 4790 struct netlink_ext_ack *extack); 4791 void netdev_adjacent_change_commit(struct net_device *old_dev, 4792 struct net_device *new_dev, 4793 struct net_device *dev); 4794 void netdev_adjacent_change_abort(struct net_device *old_dev, 4795 struct net_device *new_dev, 4796 struct net_device *dev); 4797 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4798 void *netdev_lower_dev_get_private(struct net_device *dev, 4799 struct net_device *lower_dev); 4800 void netdev_lower_state_changed(struct net_device *lower_dev, 4801 void *lower_state_info); 4802 4803 /* RSS keys are 40 or 52 bytes long */ 4804 #define NETDEV_RSS_KEY_LEN 52 4805 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4806 void netdev_rss_key_fill(void *buffer, size_t len); 4807 4808 int skb_checksum_help(struct sk_buff *skb); 4809 int skb_crc32c_csum_help(struct sk_buff *skb); 4810 int skb_csum_hwoffload_help(struct sk_buff *skb, 4811 const netdev_features_t features); 4812 4813 struct netdev_bonding_info { 4814 ifslave slave; 4815 ifbond master; 4816 }; 4817 4818 struct netdev_notifier_bonding_info { 4819 struct netdev_notifier_info info; /* must be first */ 4820 struct netdev_bonding_info bonding_info; 4821 }; 4822 4823 void netdev_bonding_info_change(struct net_device *dev, 4824 struct netdev_bonding_info *bonding_info); 4825 4826 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4827 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4828 #else 4829 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4830 const void *data) 4831 { 4832 } 4833 #endif 4834 4835 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4836 4837 static inline bool can_checksum_protocol(netdev_features_t features, 4838 __be16 protocol) 4839 { 4840 if (protocol == htons(ETH_P_FCOE)) 4841 return !!(features & NETIF_F_FCOE_CRC); 4842 4843 /* Assume this is an IP checksum (not SCTP CRC) */ 4844 4845 if (features & NETIF_F_HW_CSUM) { 4846 /* Can checksum everything */ 4847 return true; 4848 } 4849 4850 switch (protocol) { 4851 case htons(ETH_P_IP): 4852 return !!(features & NETIF_F_IP_CSUM); 4853 case htons(ETH_P_IPV6): 4854 return !!(features & NETIF_F_IPV6_CSUM); 4855 default: 4856 return false; 4857 } 4858 } 4859 4860 #ifdef CONFIG_BUG 4861 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4862 #else 4863 static inline void netdev_rx_csum_fault(struct net_device *dev, 4864 struct sk_buff *skb) 4865 { 4866 } 4867 #endif 4868 /* rx skb timestamps */ 4869 void net_enable_timestamp(void); 4870 void net_disable_timestamp(void); 4871 4872 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 4873 const struct skb_shared_hwtstamps *hwtstamps, 4874 bool cycles) 4875 { 4876 const struct net_device_ops *ops = dev->netdev_ops; 4877 4878 if (ops->ndo_get_tstamp) 4879 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 4880 4881 return hwtstamps->hwtstamp; 4882 } 4883 4884 #ifndef CONFIG_PREEMPT_RT 4885 static inline void netdev_xmit_set_more(bool more) 4886 { 4887 __this_cpu_write(softnet_data.xmit.more, more); 4888 } 4889 4890 static inline bool netdev_xmit_more(void) 4891 { 4892 return __this_cpu_read(softnet_data.xmit.more); 4893 } 4894 #else 4895 static inline void netdev_xmit_set_more(bool more) 4896 { 4897 current->net_xmit.more = more; 4898 } 4899 4900 static inline bool netdev_xmit_more(void) 4901 { 4902 return current->net_xmit.more; 4903 } 4904 #endif 4905 4906 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4907 struct sk_buff *skb, struct net_device *dev, 4908 bool more) 4909 { 4910 netdev_xmit_set_more(more); 4911 return ops->ndo_start_xmit(skb, dev); 4912 } 4913 4914 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4915 struct netdev_queue *txq, bool more) 4916 { 4917 const struct net_device_ops *ops = dev->netdev_ops; 4918 netdev_tx_t rc; 4919 4920 rc = __netdev_start_xmit(ops, skb, dev, more); 4921 if (rc == NETDEV_TX_OK) 4922 txq_trans_update(txq); 4923 4924 return rc; 4925 } 4926 4927 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4928 const void *ns); 4929 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4930 const void *ns); 4931 4932 extern const struct kobj_ns_type_operations net_ns_type_operations; 4933 4934 const char *netdev_drivername(const struct net_device *dev); 4935 4936 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4937 netdev_features_t f2) 4938 { 4939 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4940 if (f1 & NETIF_F_HW_CSUM) 4941 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4942 else 4943 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4944 } 4945 4946 return f1 & f2; 4947 } 4948 4949 static inline netdev_features_t netdev_get_wanted_features( 4950 struct net_device *dev) 4951 { 4952 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4953 } 4954 netdev_features_t netdev_increment_features(netdev_features_t all, 4955 netdev_features_t one, netdev_features_t mask); 4956 4957 /* Allow TSO being used on stacked device : 4958 * Performing the GSO segmentation before last device 4959 * is a performance improvement. 4960 */ 4961 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4962 netdev_features_t mask) 4963 { 4964 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4965 } 4966 4967 int __netdev_update_features(struct net_device *dev); 4968 void netdev_update_features(struct net_device *dev); 4969 void netdev_change_features(struct net_device *dev); 4970 4971 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4972 struct net_device *dev); 4973 4974 netdev_features_t passthru_features_check(struct sk_buff *skb, 4975 struct net_device *dev, 4976 netdev_features_t features); 4977 netdev_features_t netif_skb_features(struct sk_buff *skb); 4978 void skb_warn_bad_offload(const struct sk_buff *skb); 4979 4980 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4981 { 4982 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4983 4984 /* check flags correspondence */ 4985 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4986 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4987 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4988 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4989 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4990 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4991 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4992 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4993 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4994 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4995 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4996 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4997 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4998 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4999 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 5000 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 5001 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 5002 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 5003 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 5004 5005 return (features & feature) == feature; 5006 } 5007 5008 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 5009 { 5010 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 5011 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 5012 } 5013 5014 static inline bool netif_needs_gso(struct sk_buff *skb, 5015 netdev_features_t features) 5016 { 5017 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 5018 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 5019 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 5020 } 5021 5022 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 5023 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 5024 void netif_inherit_tso_max(struct net_device *to, 5025 const struct net_device *from); 5026 5027 static inline bool netif_is_macsec(const struct net_device *dev) 5028 { 5029 return dev->priv_flags & IFF_MACSEC; 5030 } 5031 5032 static inline bool netif_is_macvlan(const struct net_device *dev) 5033 { 5034 return dev->priv_flags & IFF_MACVLAN; 5035 } 5036 5037 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5038 { 5039 return dev->priv_flags & IFF_MACVLAN_PORT; 5040 } 5041 5042 static inline bool netif_is_bond_master(const struct net_device *dev) 5043 { 5044 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5045 } 5046 5047 static inline bool netif_is_bond_slave(const struct net_device *dev) 5048 { 5049 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5050 } 5051 5052 static inline bool netif_supports_nofcs(struct net_device *dev) 5053 { 5054 return dev->priv_flags & IFF_SUPP_NOFCS; 5055 } 5056 5057 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5058 { 5059 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5060 } 5061 5062 static inline bool netif_is_l3_master(const struct net_device *dev) 5063 { 5064 return dev->priv_flags & IFF_L3MDEV_MASTER; 5065 } 5066 5067 static inline bool netif_is_l3_slave(const struct net_device *dev) 5068 { 5069 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5070 } 5071 5072 static inline int dev_sdif(const struct net_device *dev) 5073 { 5074 #ifdef CONFIG_NET_L3_MASTER_DEV 5075 if (netif_is_l3_slave(dev)) 5076 return dev->ifindex; 5077 #endif 5078 return 0; 5079 } 5080 5081 static inline bool netif_is_bridge_master(const struct net_device *dev) 5082 { 5083 return dev->priv_flags & IFF_EBRIDGE; 5084 } 5085 5086 static inline bool netif_is_bridge_port(const struct net_device *dev) 5087 { 5088 return dev->priv_flags & IFF_BRIDGE_PORT; 5089 } 5090 5091 static inline bool netif_is_ovs_master(const struct net_device *dev) 5092 { 5093 return dev->priv_flags & IFF_OPENVSWITCH; 5094 } 5095 5096 static inline bool netif_is_ovs_port(const struct net_device *dev) 5097 { 5098 return dev->priv_flags & IFF_OVS_DATAPATH; 5099 } 5100 5101 static inline bool netif_is_any_bridge_master(const struct net_device *dev) 5102 { 5103 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev); 5104 } 5105 5106 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5107 { 5108 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5109 } 5110 5111 static inline bool netif_is_team_master(const struct net_device *dev) 5112 { 5113 return dev->priv_flags & IFF_TEAM; 5114 } 5115 5116 static inline bool netif_is_team_port(const struct net_device *dev) 5117 { 5118 return dev->priv_flags & IFF_TEAM_PORT; 5119 } 5120 5121 static inline bool netif_is_lag_master(const struct net_device *dev) 5122 { 5123 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5124 } 5125 5126 static inline bool netif_is_lag_port(const struct net_device *dev) 5127 { 5128 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5129 } 5130 5131 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5132 { 5133 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5134 } 5135 5136 static inline bool netif_is_failover(const struct net_device *dev) 5137 { 5138 return dev->priv_flags & IFF_FAILOVER; 5139 } 5140 5141 static inline bool netif_is_failover_slave(const struct net_device *dev) 5142 { 5143 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5144 } 5145 5146 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5147 static inline void netif_keep_dst(struct net_device *dev) 5148 { 5149 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5150 } 5151 5152 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5153 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5154 { 5155 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5156 return netif_is_macsec(dev); 5157 } 5158 5159 extern struct pernet_operations __net_initdata loopback_net_ops; 5160 5161 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5162 5163 /* netdev_printk helpers, similar to dev_printk */ 5164 5165 static inline const char *netdev_name(const struct net_device *dev) 5166 { 5167 if (!dev->name[0] || strchr(dev->name, '%')) 5168 return "(unnamed net_device)"; 5169 return dev->name; 5170 } 5171 5172 static inline const char *netdev_reg_state(const struct net_device *dev) 5173 { 5174 u8 reg_state = READ_ONCE(dev->reg_state); 5175 5176 switch (reg_state) { 5177 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5178 case NETREG_REGISTERED: return ""; 5179 case NETREG_UNREGISTERING: return " (unregistering)"; 5180 case NETREG_UNREGISTERED: return " (unregistered)"; 5181 case NETREG_RELEASED: return " (released)"; 5182 case NETREG_DUMMY: return " (dummy)"; 5183 } 5184 5185 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state); 5186 return " (unknown)"; 5187 } 5188 5189 #define MODULE_ALIAS_NETDEV(device) \ 5190 MODULE_ALIAS("netdev-" device) 5191 5192 /* 5193 * netdev_WARN() acts like dev_printk(), but with the key difference 5194 * of using a WARN/WARN_ON to get the message out, including the 5195 * file/line information and a backtrace. 5196 */ 5197 #define netdev_WARN(dev, format, args...) \ 5198 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5199 netdev_reg_state(dev), ##args) 5200 5201 #define netdev_WARN_ONCE(dev, format, args...) \ 5202 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5203 netdev_reg_state(dev), ##args) 5204 5205 /* 5206 * The list of packet types we will receive (as opposed to discard) 5207 * and the routines to invoke. 5208 * 5209 * Why 16. Because with 16 the only overlap we get on a hash of the 5210 * low nibble of the protocol value is RARP/SNAP/X.25. 5211 * 5212 * 0800 IP 5213 * 0001 802.3 5214 * 0002 AX.25 5215 * 0004 802.2 5216 * 8035 RARP 5217 * 0005 SNAP 5218 * 0805 X.25 5219 * 0806 ARP 5220 * 8137 IPX 5221 * 0009 Localtalk 5222 * 86DD IPv6 5223 */ 5224 #define PTYPE_HASH_SIZE (16) 5225 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5226 5227 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5228 5229 extern struct net_device *blackhole_netdev; 5230 5231 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */ 5232 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD) 5233 #define DEV_STATS_ADD(DEV, FIELD, VAL) \ 5234 atomic_long_add((VAL), &(DEV)->stats.__##FIELD) 5235 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD) 5236 5237 #endif /* _LINUX_NETDEVICE_H */ 5238