1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <[email protected]> 13 * Corey Minyard <[email protected]> 14 * Donald J. Becker, <[email protected]> 15 * Alan Cox, <[email protected]> 16 * Bjorn Ekwall. <[email protected]> 17 * Pekka Riikonen <[email protected]> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 #include <asm/local.h> 32 33 #include <linux/percpu.h> 34 #include <linux/rculist.h> 35 #include <linux/workqueue.h> 36 #include <linux/dynamic_queue_limits.h> 37 38 #include <net/net_namespace.h> 39 #ifdef CONFIG_DCB 40 #include <net/dcbnl.h> 41 #endif 42 #include <net/netprio_cgroup.h> 43 44 #include <linux/netdev_features.h> 45 #include <linux/neighbour.h> 46 #include <linux/netdevice_xmit.h> 47 #include <uapi/linux/netdevice.h> 48 #include <uapi/linux/if_bonding.h> 49 #include <uapi/linux/pkt_cls.h> 50 #include <uapi/linux/netdev.h> 51 #include <linux/hashtable.h> 52 #include <linux/rbtree.h> 53 #include <net/net_trackers.h> 54 #include <net/net_debug.h> 55 #include <net/dropreason-core.h> 56 57 struct netpoll_info; 58 struct device; 59 struct ethtool_ops; 60 struct kernel_hwtstamp_config; 61 struct phy_device; 62 struct dsa_port; 63 struct ip_tunnel_parm_kern; 64 struct macsec_context; 65 struct macsec_ops; 66 struct netdev_name_node; 67 struct sd_flow_limit; 68 struct sfp_bus; 69 /* 802.11 specific */ 70 struct wireless_dev; 71 /* 802.15.4 specific */ 72 struct wpan_dev; 73 struct mpls_dev; 74 /* UDP Tunnel offloads */ 75 struct udp_tunnel_info; 76 struct udp_tunnel_nic_info; 77 struct udp_tunnel_nic; 78 struct bpf_prog; 79 struct xdp_buff; 80 struct xdp_frame; 81 struct xdp_metadata_ops; 82 struct xdp_md; 83 struct ethtool_netdev_state; 84 85 typedef u32 xdp_features_t; 86 87 void synchronize_net(void); 88 void netdev_set_default_ethtool_ops(struct net_device *dev, 89 const struct ethtool_ops *ops); 90 void netdev_sw_irq_coalesce_default_on(struct net_device *dev); 91 92 /* Backlog congestion levels */ 93 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 94 #define NET_RX_DROP 1 /* packet dropped */ 95 96 #define MAX_NEST_DEV 8 97 98 /* 99 * Transmit return codes: transmit return codes originate from three different 100 * namespaces: 101 * 102 * - qdisc return codes 103 * - driver transmit return codes 104 * - errno values 105 * 106 * Drivers are allowed to return any one of those in their hard_start_xmit() 107 * function. Real network devices commonly used with qdiscs should only return 108 * the driver transmit return codes though - when qdiscs are used, the actual 109 * transmission happens asynchronously, so the value is not propagated to 110 * higher layers. Virtual network devices transmit synchronously; in this case 111 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 112 * others are propagated to higher layers. 113 */ 114 115 /* qdisc ->enqueue() return codes. */ 116 #define NET_XMIT_SUCCESS 0x00 117 #define NET_XMIT_DROP 0x01 /* skb dropped */ 118 #define NET_XMIT_CN 0x02 /* congestion notification */ 119 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 120 121 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 122 * indicates that the device will soon be dropping packets, or already drops 123 * some packets of the same priority; prompting us to send less aggressively. */ 124 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 125 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 126 127 /* Driver transmit return codes */ 128 #define NETDEV_TX_MASK 0xf0 129 130 enum netdev_tx { 131 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 132 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 133 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 134 }; 135 typedef enum netdev_tx netdev_tx_t; 136 137 /* 138 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 139 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 140 */ 141 static inline bool dev_xmit_complete(int rc) 142 { 143 /* 144 * Positive cases with an skb consumed by a driver: 145 * - successful transmission (rc == NETDEV_TX_OK) 146 * - error while transmitting (rc < 0) 147 * - error while queueing to a different device (rc & NET_XMIT_MASK) 148 */ 149 if (likely(rc < NET_XMIT_MASK)) 150 return true; 151 152 return false; 153 } 154 155 /* 156 * Compute the worst-case header length according to the protocols 157 * used. 158 */ 159 160 #if defined(CONFIG_HYPERV_NET) 161 # define LL_MAX_HEADER 128 162 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 163 # if defined(CONFIG_MAC80211_MESH) 164 # define LL_MAX_HEADER 128 165 # else 166 # define LL_MAX_HEADER 96 167 # endif 168 #else 169 # define LL_MAX_HEADER 32 170 #endif 171 172 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 173 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 174 #define MAX_HEADER LL_MAX_HEADER 175 #else 176 #define MAX_HEADER (LL_MAX_HEADER + 48) 177 #endif 178 179 /* 180 * Old network device statistics. Fields are native words 181 * (unsigned long) so they can be read and written atomically. 182 */ 183 184 #define NET_DEV_STAT(FIELD) \ 185 union { \ 186 unsigned long FIELD; \ 187 atomic_long_t __##FIELD; \ 188 } 189 190 struct net_device_stats { 191 NET_DEV_STAT(rx_packets); 192 NET_DEV_STAT(tx_packets); 193 NET_DEV_STAT(rx_bytes); 194 NET_DEV_STAT(tx_bytes); 195 NET_DEV_STAT(rx_errors); 196 NET_DEV_STAT(tx_errors); 197 NET_DEV_STAT(rx_dropped); 198 NET_DEV_STAT(tx_dropped); 199 NET_DEV_STAT(multicast); 200 NET_DEV_STAT(collisions); 201 NET_DEV_STAT(rx_length_errors); 202 NET_DEV_STAT(rx_over_errors); 203 NET_DEV_STAT(rx_crc_errors); 204 NET_DEV_STAT(rx_frame_errors); 205 NET_DEV_STAT(rx_fifo_errors); 206 NET_DEV_STAT(rx_missed_errors); 207 NET_DEV_STAT(tx_aborted_errors); 208 NET_DEV_STAT(tx_carrier_errors); 209 NET_DEV_STAT(tx_fifo_errors); 210 NET_DEV_STAT(tx_heartbeat_errors); 211 NET_DEV_STAT(tx_window_errors); 212 NET_DEV_STAT(rx_compressed); 213 NET_DEV_STAT(tx_compressed); 214 }; 215 #undef NET_DEV_STAT 216 217 /* per-cpu stats, allocated on demand. 218 * Try to fit them in a single cache line, for dev_get_stats() sake. 219 */ 220 struct net_device_core_stats { 221 unsigned long rx_dropped; 222 unsigned long tx_dropped; 223 unsigned long rx_nohandler; 224 unsigned long rx_otherhost_dropped; 225 } __aligned(4 * sizeof(unsigned long)); 226 227 #include <linux/cache.h> 228 #include <linux/skbuff.h> 229 230 struct neighbour; 231 struct neigh_parms; 232 struct sk_buff; 233 234 struct netdev_hw_addr { 235 struct list_head list; 236 struct rb_node node; 237 unsigned char addr[MAX_ADDR_LEN]; 238 unsigned char type; 239 #define NETDEV_HW_ADDR_T_LAN 1 240 #define NETDEV_HW_ADDR_T_SAN 2 241 #define NETDEV_HW_ADDR_T_UNICAST 3 242 #define NETDEV_HW_ADDR_T_MULTICAST 4 243 bool global_use; 244 int sync_cnt; 245 int refcount; 246 int synced; 247 struct rcu_head rcu_head; 248 }; 249 250 struct netdev_hw_addr_list { 251 struct list_head list; 252 int count; 253 254 /* Auxiliary tree for faster lookup on addition and deletion */ 255 struct rb_root tree; 256 }; 257 258 #define netdev_hw_addr_list_count(l) ((l)->count) 259 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 260 #define netdev_hw_addr_list_for_each(ha, l) \ 261 list_for_each_entry(ha, &(l)->list, list) 262 263 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 264 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 265 #define netdev_for_each_uc_addr(ha, dev) \ 266 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 267 #define netdev_for_each_synced_uc_addr(_ha, _dev) \ 268 netdev_for_each_uc_addr((_ha), (_dev)) \ 269 if ((_ha)->sync_cnt) 270 271 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 272 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 273 #define netdev_for_each_mc_addr(ha, dev) \ 274 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 275 #define netdev_for_each_synced_mc_addr(_ha, _dev) \ 276 netdev_for_each_mc_addr((_ha), (_dev)) \ 277 if ((_ha)->sync_cnt) 278 279 struct hh_cache { 280 unsigned int hh_len; 281 seqlock_t hh_lock; 282 283 /* cached hardware header; allow for machine alignment needs. */ 284 #define HH_DATA_MOD 16 285 #define HH_DATA_OFF(__len) \ 286 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 287 #define HH_DATA_ALIGN(__len) \ 288 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 289 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 290 }; 291 292 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 293 * Alternative is: 294 * dev->hard_header_len ? (dev->hard_header_len + 295 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 296 * 297 * We could use other alignment values, but we must maintain the 298 * relationship HH alignment <= LL alignment. 299 */ 300 #define LL_RESERVED_SPACE(dev) \ 301 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \ 302 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 303 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 304 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \ 305 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 306 307 struct header_ops { 308 int (*create) (struct sk_buff *skb, struct net_device *dev, 309 unsigned short type, const void *daddr, 310 const void *saddr, unsigned int len); 311 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 312 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 313 void (*cache_update)(struct hh_cache *hh, 314 const struct net_device *dev, 315 const unsigned char *haddr); 316 bool (*validate)(const char *ll_header, unsigned int len); 317 __be16 (*parse_protocol)(const struct sk_buff *skb); 318 }; 319 320 /* These flag bits are private to the generic network queueing 321 * layer; they may not be explicitly referenced by any other 322 * code. 323 */ 324 325 enum netdev_state_t { 326 __LINK_STATE_START, 327 __LINK_STATE_PRESENT, 328 __LINK_STATE_NOCARRIER, 329 __LINK_STATE_LINKWATCH_PENDING, 330 __LINK_STATE_DORMANT, 331 __LINK_STATE_TESTING, 332 }; 333 334 struct gro_list { 335 struct list_head list; 336 int count; 337 }; 338 339 /* 340 * size of gro hash buckets, must less than bit number of 341 * napi_struct::gro_bitmask 342 */ 343 #define GRO_HASH_BUCKETS 8 344 345 /* 346 * Structure for NAPI scheduling similar to tasklet but with weighting 347 */ 348 struct napi_struct { 349 /* The poll_list must only be managed by the entity which 350 * changes the state of the NAPI_STATE_SCHED bit. This means 351 * whoever atomically sets that bit can add this napi_struct 352 * to the per-CPU poll_list, and whoever clears that bit 353 * can remove from the list right before clearing the bit. 354 */ 355 struct list_head poll_list; 356 357 unsigned long state; 358 int weight; 359 int defer_hard_irqs_count; 360 unsigned long gro_bitmask; 361 int (*poll)(struct napi_struct *, int); 362 #ifdef CONFIG_NETPOLL 363 /* CPU actively polling if netpoll is configured */ 364 int poll_owner; 365 #endif 366 /* CPU on which NAPI has been scheduled for processing */ 367 int list_owner; 368 struct net_device *dev; 369 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 370 struct sk_buff *skb; 371 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 372 int rx_count; /* length of rx_list */ 373 unsigned int napi_id; 374 struct hrtimer timer; 375 struct task_struct *thread; 376 /* control-path-only fields follow */ 377 struct list_head dev_list; 378 struct hlist_node napi_hash_node; 379 int irq; 380 }; 381 382 enum { 383 NAPI_STATE_SCHED, /* Poll is scheduled */ 384 NAPI_STATE_MISSED, /* reschedule a napi */ 385 NAPI_STATE_DISABLE, /* Disable pending */ 386 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 387 NAPI_STATE_LISTED, /* NAPI added to system lists */ 388 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 389 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 390 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 391 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 392 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 393 }; 394 395 enum { 396 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 397 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 398 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 399 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 400 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 401 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 402 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 403 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 404 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 405 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 406 }; 407 408 enum gro_result { 409 GRO_MERGED, 410 GRO_MERGED_FREE, 411 GRO_HELD, 412 GRO_NORMAL, 413 GRO_CONSUMED, 414 }; 415 typedef enum gro_result gro_result_t; 416 417 /* 418 * enum rx_handler_result - Possible return values for rx_handlers. 419 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 420 * further. 421 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 422 * case skb->dev was changed by rx_handler. 423 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 424 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 425 * 426 * rx_handlers are functions called from inside __netif_receive_skb(), to do 427 * special processing of the skb, prior to delivery to protocol handlers. 428 * 429 * Currently, a net_device can only have a single rx_handler registered. Trying 430 * to register a second rx_handler will return -EBUSY. 431 * 432 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 433 * To unregister a rx_handler on a net_device, use 434 * netdev_rx_handler_unregister(). 435 * 436 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 437 * do with the skb. 438 * 439 * If the rx_handler consumed the skb in some way, it should return 440 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 441 * the skb to be delivered in some other way. 442 * 443 * If the rx_handler changed skb->dev, to divert the skb to another 444 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 445 * new device will be called if it exists. 446 * 447 * If the rx_handler decides the skb should be ignored, it should return 448 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 449 * are registered on exact device (ptype->dev == skb->dev). 450 * 451 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 452 * delivered, it should return RX_HANDLER_PASS. 453 * 454 * A device without a registered rx_handler will behave as if rx_handler 455 * returned RX_HANDLER_PASS. 456 */ 457 458 enum rx_handler_result { 459 RX_HANDLER_CONSUMED, 460 RX_HANDLER_ANOTHER, 461 RX_HANDLER_EXACT, 462 RX_HANDLER_PASS, 463 }; 464 typedef enum rx_handler_result rx_handler_result_t; 465 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 466 467 void __napi_schedule(struct napi_struct *n); 468 void __napi_schedule_irqoff(struct napi_struct *n); 469 470 static inline bool napi_disable_pending(struct napi_struct *n) 471 { 472 return test_bit(NAPI_STATE_DISABLE, &n->state); 473 } 474 475 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 476 { 477 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 478 } 479 480 /** 481 * napi_is_scheduled - test if NAPI is scheduled 482 * @n: NAPI context 483 * 484 * This check is "best-effort". With no locking implemented, 485 * a NAPI can be scheduled or terminate right after this check 486 * and produce not precise results. 487 * 488 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled 489 * should not be used normally and napi_schedule should be 490 * used instead. 491 * 492 * Use only if the driver really needs to check if a NAPI 493 * is scheduled for example in the context of delayed timer 494 * that can be skipped if a NAPI is already scheduled. 495 * 496 * Return True if NAPI is scheduled, False otherwise. 497 */ 498 static inline bool napi_is_scheduled(struct napi_struct *n) 499 { 500 return test_bit(NAPI_STATE_SCHED, &n->state); 501 } 502 503 bool napi_schedule_prep(struct napi_struct *n); 504 505 /** 506 * napi_schedule - schedule NAPI poll 507 * @n: NAPI context 508 * 509 * Schedule NAPI poll routine to be called if it is not already 510 * running. 511 * Return true if we schedule a NAPI or false if not. 512 * Refer to napi_schedule_prep() for additional reason on why 513 * a NAPI might not be scheduled. 514 */ 515 static inline bool napi_schedule(struct napi_struct *n) 516 { 517 if (napi_schedule_prep(n)) { 518 __napi_schedule(n); 519 return true; 520 } 521 522 return false; 523 } 524 525 /** 526 * napi_schedule_irqoff - schedule NAPI poll 527 * @n: NAPI context 528 * 529 * Variant of napi_schedule(), assuming hard irqs are masked. 530 */ 531 static inline void napi_schedule_irqoff(struct napi_struct *n) 532 { 533 if (napi_schedule_prep(n)) 534 __napi_schedule_irqoff(n); 535 } 536 537 /** 538 * napi_complete_done - NAPI processing complete 539 * @n: NAPI context 540 * @work_done: number of packets processed 541 * 542 * Mark NAPI processing as complete. Should only be called if poll budget 543 * has not been completely consumed. 544 * Prefer over napi_complete(). 545 * Return false if device should avoid rearming interrupts. 546 */ 547 bool napi_complete_done(struct napi_struct *n, int work_done); 548 549 static inline bool napi_complete(struct napi_struct *n) 550 { 551 return napi_complete_done(n, 0); 552 } 553 554 int dev_set_threaded(struct net_device *dev, bool threaded); 555 556 /** 557 * napi_disable - prevent NAPI from scheduling 558 * @n: NAPI context 559 * 560 * Stop NAPI from being scheduled on this context. 561 * Waits till any outstanding processing completes. 562 */ 563 void napi_disable(struct napi_struct *n); 564 565 void napi_enable(struct napi_struct *n); 566 567 /** 568 * napi_synchronize - wait until NAPI is not running 569 * @n: NAPI context 570 * 571 * Wait until NAPI is done being scheduled on this context. 572 * Waits till any outstanding processing completes but 573 * does not disable future activations. 574 */ 575 static inline void napi_synchronize(const struct napi_struct *n) 576 { 577 if (IS_ENABLED(CONFIG_SMP)) 578 while (test_bit(NAPI_STATE_SCHED, &n->state)) 579 msleep(1); 580 else 581 barrier(); 582 } 583 584 /** 585 * napi_if_scheduled_mark_missed - if napi is running, set the 586 * NAPIF_STATE_MISSED 587 * @n: NAPI context 588 * 589 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 590 * NAPI is scheduled. 591 **/ 592 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 593 { 594 unsigned long val, new; 595 596 val = READ_ONCE(n->state); 597 do { 598 if (val & NAPIF_STATE_DISABLE) 599 return true; 600 601 if (!(val & NAPIF_STATE_SCHED)) 602 return false; 603 604 new = val | NAPIF_STATE_MISSED; 605 } while (!try_cmpxchg(&n->state, &val, new)); 606 607 return true; 608 } 609 610 enum netdev_queue_state_t { 611 __QUEUE_STATE_DRV_XOFF, 612 __QUEUE_STATE_STACK_XOFF, 613 __QUEUE_STATE_FROZEN, 614 }; 615 616 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 617 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 618 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 619 620 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 621 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 622 QUEUE_STATE_FROZEN) 623 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 624 QUEUE_STATE_FROZEN) 625 626 /* 627 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 628 * netif_tx_* functions below are used to manipulate this flag. The 629 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 630 * queue independently. The netif_xmit_*stopped functions below are called 631 * to check if the queue has been stopped by the driver or stack (either 632 * of the XOFF bits are set in the state). Drivers should not need to call 633 * netif_xmit*stopped functions, they should only be using netif_tx_*. 634 */ 635 636 struct netdev_queue { 637 /* 638 * read-mostly part 639 */ 640 struct net_device *dev; 641 netdevice_tracker dev_tracker; 642 643 struct Qdisc __rcu *qdisc; 644 struct Qdisc __rcu *qdisc_sleeping; 645 #ifdef CONFIG_SYSFS 646 struct kobject kobj; 647 #endif 648 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 649 int numa_node; 650 #endif 651 unsigned long tx_maxrate; 652 /* 653 * Number of TX timeouts for this queue 654 * (/sys/class/net/DEV/Q/trans_timeout) 655 */ 656 atomic_long_t trans_timeout; 657 658 /* Subordinate device that the queue has been assigned to */ 659 struct net_device *sb_dev; 660 #ifdef CONFIG_XDP_SOCKETS 661 struct xsk_buff_pool *pool; 662 #endif 663 /* NAPI instance for the queue 664 * Readers and writers must hold RTNL 665 */ 666 struct napi_struct *napi; 667 /* 668 * write-mostly part 669 */ 670 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 671 int xmit_lock_owner; 672 /* 673 * Time (in jiffies) of last Tx 674 */ 675 unsigned long trans_start; 676 677 unsigned long state; 678 679 #ifdef CONFIG_BQL 680 struct dql dql; 681 #endif 682 } ____cacheline_aligned_in_smp; 683 684 extern int sysctl_fb_tunnels_only_for_init_net; 685 extern int sysctl_devconf_inherit_init_net; 686 687 /* 688 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 689 * == 1 : For initns only 690 * == 2 : For none. 691 */ 692 static inline bool net_has_fallback_tunnels(const struct net *net) 693 { 694 #if IS_ENABLED(CONFIG_SYSCTL) 695 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net); 696 697 return !fb_tunnels_only_for_init_net || 698 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1); 699 #else 700 return true; 701 #endif 702 } 703 704 static inline int net_inherit_devconf(void) 705 { 706 #if IS_ENABLED(CONFIG_SYSCTL) 707 return READ_ONCE(sysctl_devconf_inherit_init_net); 708 #else 709 return 0; 710 #endif 711 } 712 713 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 714 { 715 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 716 return q->numa_node; 717 #else 718 return NUMA_NO_NODE; 719 #endif 720 } 721 722 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 723 { 724 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 725 q->numa_node = node; 726 #endif 727 } 728 729 #ifdef CONFIG_RFS_ACCEL 730 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 731 u16 filter_id); 732 #endif 733 734 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 735 enum xps_map_type { 736 XPS_CPUS = 0, 737 XPS_RXQS, 738 XPS_MAPS_MAX, 739 }; 740 741 #ifdef CONFIG_XPS 742 /* 743 * This structure holds an XPS map which can be of variable length. The 744 * map is an array of queues. 745 */ 746 struct xps_map { 747 unsigned int len; 748 unsigned int alloc_len; 749 struct rcu_head rcu; 750 u16 queues[]; 751 }; 752 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 753 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 754 - sizeof(struct xps_map)) / sizeof(u16)) 755 756 /* 757 * This structure holds all XPS maps for device. Maps are indexed by CPU. 758 * 759 * We keep track of the number of cpus/rxqs used when the struct is allocated, 760 * in nr_ids. This will help not accessing out-of-bound memory. 761 * 762 * We keep track of the number of traffic classes used when the struct is 763 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 764 * not crossing its upper bound, as the original dev->num_tc can be updated in 765 * the meantime. 766 */ 767 struct xps_dev_maps { 768 struct rcu_head rcu; 769 unsigned int nr_ids; 770 s16 num_tc; 771 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 772 }; 773 774 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 775 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 776 777 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 778 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 779 780 #endif /* CONFIG_XPS */ 781 782 #define TC_MAX_QUEUE 16 783 #define TC_BITMASK 15 784 /* HW offloaded queuing disciplines txq count and offset maps */ 785 struct netdev_tc_txq { 786 u16 count; 787 u16 offset; 788 }; 789 790 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 791 /* 792 * This structure is to hold information about the device 793 * configured to run FCoE protocol stack. 794 */ 795 struct netdev_fcoe_hbainfo { 796 char manufacturer[64]; 797 char serial_number[64]; 798 char hardware_version[64]; 799 char driver_version[64]; 800 char optionrom_version[64]; 801 char firmware_version[64]; 802 char model[256]; 803 char model_description[256]; 804 }; 805 #endif 806 807 #define MAX_PHYS_ITEM_ID_LEN 32 808 809 /* This structure holds a unique identifier to identify some 810 * physical item (port for example) used by a netdevice. 811 */ 812 struct netdev_phys_item_id { 813 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 814 unsigned char id_len; 815 }; 816 817 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 818 struct netdev_phys_item_id *b) 819 { 820 return a->id_len == b->id_len && 821 memcmp(a->id, b->id, a->id_len) == 0; 822 } 823 824 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 825 struct sk_buff *skb, 826 struct net_device *sb_dev); 827 828 enum net_device_path_type { 829 DEV_PATH_ETHERNET = 0, 830 DEV_PATH_VLAN, 831 DEV_PATH_BRIDGE, 832 DEV_PATH_PPPOE, 833 DEV_PATH_DSA, 834 DEV_PATH_MTK_WDMA, 835 }; 836 837 struct net_device_path { 838 enum net_device_path_type type; 839 const struct net_device *dev; 840 union { 841 struct { 842 u16 id; 843 __be16 proto; 844 u8 h_dest[ETH_ALEN]; 845 } encap; 846 struct { 847 enum { 848 DEV_PATH_BR_VLAN_KEEP, 849 DEV_PATH_BR_VLAN_TAG, 850 DEV_PATH_BR_VLAN_UNTAG, 851 DEV_PATH_BR_VLAN_UNTAG_HW, 852 } vlan_mode; 853 u16 vlan_id; 854 __be16 vlan_proto; 855 } bridge; 856 struct { 857 int port; 858 u16 proto; 859 } dsa; 860 struct { 861 u8 wdma_idx; 862 u8 queue; 863 u16 wcid; 864 u8 bss; 865 u8 amsdu; 866 } mtk_wdma; 867 }; 868 }; 869 870 #define NET_DEVICE_PATH_STACK_MAX 5 871 #define NET_DEVICE_PATH_VLAN_MAX 2 872 873 struct net_device_path_stack { 874 int num_paths; 875 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 876 }; 877 878 struct net_device_path_ctx { 879 const struct net_device *dev; 880 u8 daddr[ETH_ALEN]; 881 882 int num_vlans; 883 struct { 884 u16 id; 885 __be16 proto; 886 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 887 }; 888 889 enum tc_setup_type { 890 TC_QUERY_CAPS, 891 TC_SETUP_QDISC_MQPRIO, 892 TC_SETUP_CLSU32, 893 TC_SETUP_CLSFLOWER, 894 TC_SETUP_CLSMATCHALL, 895 TC_SETUP_CLSBPF, 896 TC_SETUP_BLOCK, 897 TC_SETUP_QDISC_CBS, 898 TC_SETUP_QDISC_RED, 899 TC_SETUP_QDISC_PRIO, 900 TC_SETUP_QDISC_MQ, 901 TC_SETUP_QDISC_ETF, 902 TC_SETUP_ROOT_QDISC, 903 TC_SETUP_QDISC_GRED, 904 TC_SETUP_QDISC_TAPRIO, 905 TC_SETUP_FT, 906 TC_SETUP_QDISC_ETS, 907 TC_SETUP_QDISC_TBF, 908 TC_SETUP_QDISC_FIFO, 909 TC_SETUP_QDISC_HTB, 910 TC_SETUP_ACT, 911 }; 912 913 /* These structures hold the attributes of bpf state that are being passed 914 * to the netdevice through the bpf op. 915 */ 916 enum bpf_netdev_command { 917 /* Set or clear a bpf program used in the earliest stages of packet 918 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 919 * is responsible for calling bpf_prog_put on any old progs that are 920 * stored. In case of error, the callee need not release the new prog 921 * reference, but on success it takes ownership and must bpf_prog_put 922 * when it is no longer used. 923 */ 924 XDP_SETUP_PROG, 925 XDP_SETUP_PROG_HW, 926 /* BPF program for offload callbacks, invoked at program load time. */ 927 BPF_OFFLOAD_MAP_ALLOC, 928 BPF_OFFLOAD_MAP_FREE, 929 XDP_SETUP_XSK_POOL, 930 }; 931 932 struct bpf_prog_offload_ops; 933 struct netlink_ext_ack; 934 struct xdp_umem; 935 struct xdp_dev_bulk_queue; 936 struct bpf_xdp_link; 937 938 enum bpf_xdp_mode { 939 XDP_MODE_SKB = 0, 940 XDP_MODE_DRV = 1, 941 XDP_MODE_HW = 2, 942 __MAX_XDP_MODE 943 }; 944 945 struct bpf_xdp_entity { 946 struct bpf_prog *prog; 947 struct bpf_xdp_link *link; 948 }; 949 950 struct netdev_bpf { 951 enum bpf_netdev_command command; 952 union { 953 /* XDP_SETUP_PROG */ 954 struct { 955 u32 flags; 956 struct bpf_prog *prog; 957 struct netlink_ext_ack *extack; 958 }; 959 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 960 struct { 961 struct bpf_offloaded_map *offmap; 962 }; 963 /* XDP_SETUP_XSK_POOL */ 964 struct { 965 struct xsk_buff_pool *pool; 966 u16 queue_id; 967 } xsk; 968 }; 969 }; 970 971 /* Flags for ndo_xsk_wakeup. */ 972 #define XDP_WAKEUP_RX (1 << 0) 973 #define XDP_WAKEUP_TX (1 << 1) 974 975 #ifdef CONFIG_XFRM_OFFLOAD 976 struct xfrmdev_ops { 977 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack); 978 void (*xdo_dev_state_delete) (struct xfrm_state *x); 979 void (*xdo_dev_state_free) (struct xfrm_state *x); 980 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 981 struct xfrm_state *x); 982 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 983 void (*xdo_dev_state_update_stats) (struct xfrm_state *x); 984 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack); 985 void (*xdo_dev_policy_delete) (struct xfrm_policy *x); 986 void (*xdo_dev_policy_free) (struct xfrm_policy *x); 987 }; 988 #endif 989 990 struct dev_ifalias { 991 struct rcu_head rcuhead; 992 char ifalias[]; 993 }; 994 995 struct devlink; 996 struct tlsdev_ops; 997 998 struct netdev_net_notifier { 999 struct list_head list; 1000 struct notifier_block *nb; 1001 }; 1002 1003 /* 1004 * This structure defines the management hooks for network devices. 1005 * The following hooks can be defined; unless noted otherwise, they are 1006 * optional and can be filled with a null pointer. 1007 * 1008 * int (*ndo_init)(struct net_device *dev); 1009 * This function is called once when a network device is registered. 1010 * The network device can use this for any late stage initialization 1011 * or semantic validation. It can fail with an error code which will 1012 * be propagated back to register_netdev. 1013 * 1014 * void (*ndo_uninit)(struct net_device *dev); 1015 * This function is called when device is unregistered or when registration 1016 * fails. It is not called if init fails. 1017 * 1018 * int (*ndo_open)(struct net_device *dev); 1019 * This function is called when a network device transitions to the up 1020 * state. 1021 * 1022 * int (*ndo_stop)(struct net_device *dev); 1023 * This function is called when a network device transitions to the down 1024 * state. 1025 * 1026 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1027 * struct net_device *dev); 1028 * Called when a packet needs to be transmitted. 1029 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1030 * the queue before that can happen; it's for obsolete devices and weird 1031 * corner cases, but the stack really does a non-trivial amount 1032 * of useless work if you return NETDEV_TX_BUSY. 1033 * Required; cannot be NULL. 1034 * 1035 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1036 * struct net_device *dev 1037 * netdev_features_t features); 1038 * Called by core transmit path to determine if device is capable of 1039 * performing offload operations on a given packet. This is to give 1040 * the device an opportunity to implement any restrictions that cannot 1041 * be otherwise expressed by feature flags. The check is called with 1042 * the set of features that the stack has calculated and it returns 1043 * those the driver believes to be appropriate. 1044 * 1045 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1046 * struct net_device *sb_dev); 1047 * Called to decide which queue to use when device supports multiple 1048 * transmit queues. 1049 * 1050 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1051 * This function is called to allow device receiver to make 1052 * changes to configuration when multicast or promiscuous is enabled. 1053 * 1054 * void (*ndo_set_rx_mode)(struct net_device *dev); 1055 * This function is called device changes address list filtering. 1056 * If driver handles unicast address filtering, it should set 1057 * IFF_UNICAST_FLT in its priv_flags. 1058 * 1059 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1060 * This function is called when the Media Access Control address 1061 * needs to be changed. If this interface is not defined, the 1062 * MAC address can not be changed. 1063 * 1064 * int (*ndo_validate_addr)(struct net_device *dev); 1065 * Test if Media Access Control address is valid for the device. 1066 * 1067 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1068 * Old-style ioctl entry point. This is used internally by the 1069 * appletalk and ieee802154 subsystems but is no longer called by 1070 * the device ioctl handler. 1071 * 1072 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1073 * Used by the bonding driver for its device specific ioctls: 1074 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1075 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1076 * 1077 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1078 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1079 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1080 * 1081 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1082 * Used to set network devices bus interface parameters. This interface 1083 * is retained for legacy reasons; new devices should use the bus 1084 * interface (PCI) for low level management. 1085 * 1086 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1087 * Called when a user wants to change the Maximum Transfer Unit 1088 * of a device. 1089 * 1090 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1091 * Callback used when the transmitter has not made any progress 1092 * for dev->watchdog ticks. 1093 * 1094 * void (*ndo_get_stats64)(struct net_device *dev, 1095 * struct rtnl_link_stats64 *storage); 1096 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1097 * Called when a user wants to get the network device usage 1098 * statistics. Drivers must do one of the following: 1099 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1100 * rtnl_link_stats64 structure passed by the caller. 1101 * 2. Define @ndo_get_stats to update a net_device_stats structure 1102 * (which should normally be dev->stats) and return a pointer to 1103 * it. The structure may be changed asynchronously only if each 1104 * field is written atomically. 1105 * 3. Update dev->stats asynchronously and atomically, and define 1106 * neither operation. 1107 * 1108 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1109 * Return true if this device supports offload stats of this attr_id. 1110 * 1111 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1112 * void *attr_data) 1113 * Get statistics for offload operations by attr_id. Write it into the 1114 * attr_data pointer. 1115 * 1116 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1117 * If device supports VLAN filtering this function is called when a 1118 * VLAN id is registered. 1119 * 1120 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1121 * If device supports VLAN filtering this function is called when a 1122 * VLAN id is unregistered. 1123 * 1124 * void (*ndo_poll_controller)(struct net_device *dev); 1125 * 1126 * SR-IOV management functions. 1127 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1128 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1129 * u8 qos, __be16 proto); 1130 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1131 * int max_tx_rate); 1132 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1133 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1134 * int (*ndo_get_vf_config)(struct net_device *dev, 1135 * int vf, struct ifla_vf_info *ivf); 1136 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1137 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1138 * struct nlattr *port[]); 1139 * 1140 * Enable or disable the VF ability to query its RSS Redirection Table and 1141 * Hash Key. This is needed since on some devices VF share this information 1142 * with PF and querying it may introduce a theoretical security risk. 1143 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1144 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1145 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1146 * void *type_data); 1147 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1148 * This is always called from the stack with the rtnl lock held and netif 1149 * tx queues stopped. This allows the netdevice to perform queue 1150 * management safely. 1151 * 1152 * Fiber Channel over Ethernet (FCoE) offload functions. 1153 * int (*ndo_fcoe_enable)(struct net_device *dev); 1154 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1155 * so the underlying device can perform whatever needed configuration or 1156 * initialization to support acceleration of FCoE traffic. 1157 * 1158 * int (*ndo_fcoe_disable)(struct net_device *dev); 1159 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1160 * so the underlying device can perform whatever needed clean-ups to 1161 * stop supporting acceleration of FCoE traffic. 1162 * 1163 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1164 * struct scatterlist *sgl, unsigned int sgc); 1165 * Called when the FCoE Initiator wants to initialize an I/O that 1166 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1167 * perform necessary setup and returns 1 to indicate the device is set up 1168 * successfully to perform DDP on this I/O, otherwise this returns 0. 1169 * 1170 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1171 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1172 * indicated by the FC exchange id 'xid', so the underlying device can 1173 * clean up and reuse resources for later DDP requests. 1174 * 1175 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1176 * struct scatterlist *sgl, unsigned int sgc); 1177 * Called when the FCoE Target wants to initialize an I/O that 1178 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1179 * perform necessary setup and returns 1 to indicate the device is set up 1180 * successfully to perform DDP on this I/O, otherwise this returns 0. 1181 * 1182 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1183 * struct netdev_fcoe_hbainfo *hbainfo); 1184 * Called when the FCoE Protocol stack wants information on the underlying 1185 * device. This information is utilized by the FCoE protocol stack to 1186 * register attributes with Fiber Channel management service as per the 1187 * FC-GS Fabric Device Management Information(FDMI) specification. 1188 * 1189 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1190 * Called when the underlying device wants to override default World Wide 1191 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1192 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1193 * protocol stack to use. 1194 * 1195 * RFS acceleration. 1196 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1197 * u16 rxq_index, u32 flow_id); 1198 * Set hardware filter for RFS. rxq_index is the target queue index; 1199 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1200 * Return the filter ID on success, or a negative error code. 1201 * 1202 * Slave management functions (for bridge, bonding, etc). 1203 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1204 * Called to make another netdev an underling. 1205 * 1206 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1207 * Called to release previously enslaved netdev. 1208 * 1209 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1210 * struct sk_buff *skb, 1211 * bool all_slaves); 1212 * Get the xmit slave of master device. If all_slaves is true, function 1213 * assume all the slaves can transmit. 1214 * 1215 * Feature/offload setting functions. 1216 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1217 * netdev_features_t features); 1218 * Adjusts the requested feature flags according to device-specific 1219 * constraints, and returns the resulting flags. Must not modify 1220 * the device state. 1221 * 1222 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1223 * Called to update device configuration to new features. Passed 1224 * feature set might be less than what was returned by ndo_fix_features()). 1225 * Must return >0 or -errno if it changed dev->features itself. 1226 * 1227 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1228 * struct net_device *dev, 1229 * const unsigned char *addr, u16 vid, u16 flags, 1230 * struct netlink_ext_ack *extack); 1231 * Adds an FDB entry to dev for addr. 1232 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1233 * struct net_device *dev, 1234 * const unsigned char *addr, u16 vid) 1235 * Deletes the FDB entry from dev coresponding to addr. 1236 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev, 1237 * struct netlink_ext_ack *extack); 1238 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1239 * struct net_device *dev, struct net_device *filter_dev, 1240 * int *idx) 1241 * Used to add FDB entries to dump requests. Implementers should add 1242 * entries to skb and update idx with the number of entries. 1243 * 1244 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[], 1245 * u16 nlmsg_flags, struct netlink_ext_ack *extack); 1246 * Adds an MDB entry to dev. 1247 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[], 1248 * struct netlink_ext_ack *extack); 1249 * Deletes the MDB entry from dev. 1250 * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[], 1251 * struct netlink_ext_ack *extack); 1252 * Bulk deletes MDB entries from dev. 1253 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb, 1254 * struct netlink_callback *cb); 1255 * Dumps MDB entries from dev. The first argument (marker) in the netlink 1256 * callback is used by core rtnetlink code. 1257 * 1258 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1259 * u16 flags, struct netlink_ext_ack *extack) 1260 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1261 * struct net_device *dev, u32 filter_mask, 1262 * int nlflags) 1263 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1264 * u16 flags); 1265 * 1266 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1267 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1268 * which do not represent real hardware may define this to allow their 1269 * userspace components to manage their virtual carrier state. Devices 1270 * that determine carrier state from physical hardware properties (eg 1271 * network cables) or protocol-dependent mechanisms (eg 1272 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1273 * 1274 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1275 * struct netdev_phys_item_id *ppid); 1276 * Called to get ID of physical port of this device. If driver does 1277 * not implement this, it is assumed that the hw is not able to have 1278 * multiple net devices on single physical port. 1279 * 1280 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1281 * struct netdev_phys_item_id *ppid) 1282 * Called to get the parent ID of the physical port of this device. 1283 * 1284 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1285 * struct net_device *dev) 1286 * Called by upper layer devices to accelerate switching or other 1287 * station functionality into hardware. 'pdev is the lowerdev 1288 * to use for the offload and 'dev' is the net device that will 1289 * back the offload. Returns a pointer to the private structure 1290 * the upper layer will maintain. 1291 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1292 * Called by upper layer device to delete the station created 1293 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1294 * the station and priv is the structure returned by the add 1295 * operation. 1296 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1297 * int queue_index, u32 maxrate); 1298 * Called when a user wants to set a max-rate limitation of specific 1299 * TX queue. 1300 * int (*ndo_get_iflink)(const struct net_device *dev); 1301 * Called to get the iflink value of this device. 1302 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1303 * This function is used to get egress tunnel information for given skb. 1304 * This is useful for retrieving outer tunnel header parameters while 1305 * sampling packet. 1306 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1307 * This function is used to specify the headroom that the skb must 1308 * consider when allocation skb during packet reception. Setting 1309 * appropriate rx headroom value allows avoiding skb head copy on 1310 * forward. Setting a negative value resets the rx headroom to the 1311 * default value. 1312 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1313 * This function is used to set or query state related to XDP on the 1314 * netdevice and manage BPF offload. See definition of 1315 * enum bpf_netdev_command for details. 1316 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1317 * u32 flags); 1318 * This function is used to submit @n XDP packets for transmit on a 1319 * netdevice. Returns number of frames successfully transmitted, frames 1320 * that got dropped are freed/returned via xdp_return_frame(). 1321 * Returns negative number, means general error invoking ndo, meaning 1322 * no frames were xmit'ed and core-caller will free all frames. 1323 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1324 * struct xdp_buff *xdp); 1325 * Get the xmit slave of master device based on the xdp_buff. 1326 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1327 * This function is used to wake up the softirq, ksoftirqd or kthread 1328 * responsible for sending and/or receiving packets on a specific 1329 * queue id bound to an AF_XDP socket. The flags field specifies if 1330 * only RX, only Tx, or both should be woken up using the flags 1331 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1332 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm_kern *p, 1333 * int cmd); 1334 * Add, change, delete or get information on an IPv4 tunnel. 1335 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1336 * If a device is paired with a peer device, return the peer instance. 1337 * The caller must be under RCU read context. 1338 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1339 * Get the forwarding path to reach the real device from the HW destination address 1340 * ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1341 * const struct skb_shared_hwtstamps *hwtstamps, 1342 * bool cycles); 1343 * Get hardware timestamp based on normal/adjustable time or free running 1344 * cycle counter. This function is required if physical clock supports a 1345 * free running cycle counter. 1346 * 1347 * int (*ndo_hwtstamp_get)(struct net_device *dev, 1348 * struct kernel_hwtstamp_config *kernel_config); 1349 * Get the currently configured hardware timestamping parameters for the 1350 * NIC device. 1351 * 1352 * int (*ndo_hwtstamp_set)(struct net_device *dev, 1353 * struct kernel_hwtstamp_config *kernel_config, 1354 * struct netlink_ext_ack *extack); 1355 * Change the hardware timestamping parameters for NIC device. 1356 */ 1357 struct net_device_ops { 1358 int (*ndo_init)(struct net_device *dev); 1359 void (*ndo_uninit)(struct net_device *dev); 1360 int (*ndo_open)(struct net_device *dev); 1361 int (*ndo_stop)(struct net_device *dev); 1362 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1363 struct net_device *dev); 1364 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1365 struct net_device *dev, 1366 netdev_features_t features); 1367 u16 (*ndo_select_queue)(struct net_device *dev, 1368 struct sk_buff *skb, 1369 struct net_device *sb_dev); 1370 void (*ndo_change_rx_flags)(struct net_device *dev, 1371 int flags); 1372 void (*ndo_set_rx_mode)(struct net_device *dev); 1373 int (*ndo_set_mac_address)(struct net_device *dev, 1374 void *addr); 1375 int (*ndo_validate_addr)(struct net_device *dev); 1376 int (*ndo_do_ioctl)(struct net_device *dev, 1377 struct ifreq *ifr, int cmd); 1378 int (*ndo_eth_ioctl)(struct net_device *dev, 1379 struct ifreq *ifr, int cmd); 1380 int (*ndo_siocbond)(struct net_device *dev, 1381 struct ifreq *ifr, int cmd); 1382 int (*ndo_siocwandev)(struct net_device *dev, 1383 struct if_settings *ifs); 1384 int (*ndo_siocdevprivate)(struct net_device *dev, 1385 struct ifreq *ifr, 1386 void __user *data, int cmd); 1387 int (*ndo_set_config)(struct net_device *dev, 1388 struct ifmap *map); 1389 int (*ndo_change_mtu)(struct net_device *dev, 1390 int new_mtu); 1391 int (*ndo_neigh_setup)(struct net_device *dev, 1392 struct neigh_parms *); 1393 void (*ndo_tx_timeout) (struct net_device *dev, 1394 unsigned int txqueue); 1395 1396 void (*ndo_get_stats64)(struct net_device *dev, 1397 struct rtnl_link_stats64 *storage); 1398 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1399 int (*ndo_get_offload_stats)(int attr_id, 1400 const struct net_device *dev, 1401 void *attr_data); 1402 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1403 1404 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1405 __be16 proto, u16 vid); 1406 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1407 __be16 proto, u16 vid); 1408 #ifdef CONFIG_NET_POLL_CONTROLLER 1409 void (*ndo_poll_controller)(struct net_device *dev); 1410 int (*ndo_netpoll_setup)(struct net_device *dev, 1411 struct netpoll_info *info); 1412 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1413 #endif 1414 int (*ndo_set_vf_mac)(struct net_device *dev, 1415 int queue, u8 *mac); 1416 int (*ndo_set_vf_vlan)(struct net_device *dev, 1417 int queue, u16 vlan, 1418 u8 qos, __be16 proto); 1419 int (*ndo_set_vf_rate)(struct net_device *dev, 1420 int vf, int min_tx_rate, 1421 int max_tx_rate); 1422 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1423 int vf, bool setting); 1424 int (*ndo_set_vf_trust)(struct net_device *dev, 1425 int vf, bool setting); 1426 int (*ndo_get_vf_config)(struct net_device *dev, 1427 int vf, 1428 struct ifla_vf_info *ivf); 1429 int (*ndo_set_vf_link_state)(struct net_device *dev, 1430 int vf, int link_state); 1431 int (*ndo_get_vf_stats)(struct net_device *dev, 1432 int vf, 1433 struct ifla_vf_stats 1434 *vf_stats); 1435 int (*ndo_set_vf_port)(struct net_device *dev, 1436 int vf, 1437 struct nlattr *port[]); 1438 int (*ndo_get_vf_port)(struct net_device *dev, 1439 int vf, struct sk_buff *skb); 1440 int (*ndo_get_vf_guid)(struct net_device *dev, 1441 int vf, 1442 struct ifla_vf_guid *node_guid, 1443 struct ifla_vf_guid *port_guid); 1444 int (*ndo_set_vf_guid)(struct net_device *dev, 1445 int vf, u64 guid, 1446 int guid_type); 1447 int (*ndo_set_vf_rss_query_en)( 1448 struct net_device *dev, 1449 int vf, bool setting); 1450 int (*ndo_setup_tc)(struct net_device *dev, 1451 enum tc_setup_type type, 1452 void *type_data); 1453 #if IS_ENABLED(CONFIG_FCOE) 1454 int (*ndo_fcoe_enable)(struct net_device *dev); 1455 int (*ndo_fcoe_disable)(struct net_device *dev); 1456 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1457 u16 xid, 1458 struct scatterlist *sgl, 1459 unsigned int sgc); 1460 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1461 u16 xid); 1462 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1463 u16 xid, 1464 struct scatterlist *sgl, 1465 unsigned int sgc); 1466 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1467 struct netdev_fcoe_hbainfo *hbainfo); 1468 #endif 1469 1470 #if IS_ENABLED(CONFIG_LIBFCOE) 1471 #define NETDEV_FCOE_WWNN 0 1472 #define NETDEV_FCOE_WWPN 1 1473 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1474 u64 *wwn, int type); 1475 #endif 1476 1477 #ifdef CONFIG_RFS_ACCEL 1478 int (*ndo_rx_flow_steer)(struct net_device *dev, 1479 const struct sk_buff *skb, 1480 u16 rxq_index, 1481 u32 flow_id); 1482 #endif 1483 int (*ndo_add_slave)(struct net_device *dev, 1484 struct net_device *slave_dev, 1485 struct netlink_ext_ack *extack); 1486 int (*ndo_del_slave)(struct net_device *dev, 1487 struct net_device *slave_dev); 1488 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1489 struct sk_buff *skb, 1490 bool all_slaves); 1491 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1492 struct sock *sk); 1493 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1494 netdev_features_t features); 1495 int (*ndo_set_features)(struct net_device *dev, 1496 netdev_features_t features); 1497 int (*ndo_neigh_construct)(struct net_device *dev, 1498 struct neighbour *n); 1499 void (*ndo_neigh_destroy)(struct net_device *dev, 1500 struct neighbour *n); 1501 1502 int (*ndo_fdb_add)(struct ndmsg *ndm, 1503 struct nlattr *tb[], 1504 struct net_device *dev, 1505 const unsigned char *addr, 1506 u16 vid, 1507 u16 flags, 1508 struct netlink_ext_ack *extack); 1509 int (*ndo_fdb_del)(struct ndmsg *ndm, 1510 struct nlattr *tb[], 1511 struct net_device *dev, 1512 const unsigned char *addr, 1513 u16 vid, struct netlink_ext_ack *extack); 1514 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, 1515 struct net_device *dev, 1516 struct netlink_ext_ack *extack); 1517 int (*ndo_fdb_dump)(struct sk_buff *skb, 1518 struct netlink_callback *cb, 1519 struct net_device *dev, 1520 struct net_device *filter_dev, 1521 int *idx); 1522 int (*ndo_fdb_get)(struct sk_buff *skb, 1523 struct nlattr *tb[], 1524 struct net_device *dev, 1525 const unsigned char *addr, 1526 u16 vid, u32 portid, u32 seq, 1527 struct netlink_ext_ack *extack); 1528 int (*ndo_mdb_add)(struct net_device *dev, 1529 struct nlattr *tb[], 1530 u16 nlmsg_flags, 1531 struct netlink_ext_ack *extack); 1532 int (*ndo_mdb_del)(struct net_device *dev, 1533 struct nlattr *tb[], 1534 struct netlink_ext_ack *extack); 1535 int (*ndo_mdb_del_bulk)(struct net_device *dev, 1536 struct nlattr *tb[], 1537 struct netlink_ext_ack *extack); 1538 int (*ndo_mdb_dump)(struct net_device *dev, 1539 struct sk_buff *skb, 1540 struct netlink_callback *cb); 1541 int (*ndo_mdb_get)(struct net_device *dev, 1542 struct nlattr *tb[], u32 portid, 1543 u32 seq, 1544 struct netlink_ext_ack *extack); 1545 int (*ndo_bridge_setlink)(struct net_device *dev, 1546 struct nlmsghdr *nlh, 1547 u16 flags, 1548 struct netlink_ext_ack *extack); 1549 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1550 u32 pid, u32 seq, 1551 struct net_device *dev, 1552 u32 filter_mask, 1553 int nlflags); 1554 int (*ndo_bridge_dellink)(struct net_device *dev, 1555 struct nlmsghdr *nlh, 1556 u16 flags); 1557 int (*ndo_change_carrier)(struct net_device *dev, 1558 bool new_carrier); 1559 int (*ndo_get_phys_port_id)(struct net_device *dev, 1560 struct netdev_phys_item_id *ppid); 1561 int (*ndo_get_port_parent_id)(struct net_device *dev, 1562 struct netdev_phys_item_id *ppid); 1563 int (*ndo_get_phys_port_name)(struct net_device *dev, 1564 char *name, size_t len); 1565 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1566 struct net_device *dev); 1567 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1568 void *priv); 1569 1570 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1571 int queue_index, 1572 u32 maxrate); 1573 int (*ndo_get_iflink)(const struct net_device *dev); 1574 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1575 struct sk_buff *skb); 1576 void (*ndo_set_rx_headroom)(struct net_device *dev, 1577 int needed_headroom); 1578 int (*ndo_bpf)(struct net_device *dev, 1579 struct netdev_bpf *bpf); 1580 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1581 struct xdp_frame **xdp, 1582 u32 flags); 1583 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1584 struct xdp_buff *xdp); 1585 int (*ndo_xsk_wakeup)(struct net_device *dev, 1586 u32 queue_id, u32 flags); 1587 int (*ndo_tunnel_ctl)(struct net_device *dev, 1588 struct ip_tunnel_parm_kern *p, 1589 int cmd); 1590 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1591 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1592 struct net_device_path *path); 1593 ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1594 const struct skb_shared_hwtstamps *hwtstamps, 1595 bool cycles); 1596 int (*ndo_hwtstamp_get)(struct net_device *dev, 1597 struct kernel_hwtstamp_config *kernel_config); 1598 int (*ndo_hwtstamp_set)(struct net_device *dev, 1599 struct kernel_hwtstamp_config *kernel_config, 1600 struct netlink_ext_ack *extack); 1601 }; 1602 1603 /** 1604 * enum netdev_priv_flags - &struct net_device priv_flags 1605 * 1606 * These are the &struct net_device, they are only set internally 1607 * by drivers and used in the kernel. These flags are invisible to 1608 * userspace; this means that the order of these flags can change 1609 * during any kernel release. 1610 * 1611 * You should have a pretty good reason to be extending these flags. 1612 * 1613 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1614 * @IFF_EBRIDGE: Ethernet bridging device 1615 * @IFF_BONDING: bonding master or slave 1616 * @IFF_ISATAP: ISATAP interface (RFC4214) 1617 * @IFF_WAN_HDLC: WAN HDLC device 1618 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1619 * release skb->dst 1620 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1621 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1622 * @IFF_MACVLAN_PORT: device used as macvlan port 1623 * @IFF_BRIDGE_PORT: device used as bridge port 1624 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1625 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1626 * @IFF_UNICAST_FLT: Supports unicast filtering 1627 * @IFF_TEAM_PORT: device used as team port 1628 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1629 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1630 * change when it's running 1631 * @IFF_MACVLAN: Macvlan device 1632 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1633 * underlying stacked devices 1634 * @IFF_L3MDEV_MASTER: device is an L3 master device 1635 * @IFF_NO_QUEUE: device can run without qdisc attached 1636 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1637 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1638 * @IFF_TEAM: device is a team device 1639 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1640 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1641 * entity (i.e. the master device for bridged veth) 1642 * @IFF_MACSEC: device is a MACsec device 1643 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1644 * @IFF_FAILOVER: device is a failover master device 1645 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1646 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1647 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf 1648 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1649 * skb_headlen(skb) == 0 (data starts from frag0) 1650 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN 1651 * @IFF_SEE_ALL_HWTSTAMP_REQUESTS: device wants to see calls to 1652 * ndo_hwtstamp_set() for all timestamp requests regardless of source, 1653 * even if those aren't HWTSTAMP_SOURCE_NETDEV. 1654 */ 1655 enum netdev_priv_flags { 1656 IFF_802_1Q_VLAN = 1<<0, 1657 IFF_EBRIDGE = 1<<1, 1658 IFF_BONDING = 1<<2, 1659 IFF_ISATAP = 1<<3, 1660 IFF_WAN_HDLC = 1<<4, 1661 IFF_XMIT_DST_RELEASE = 1<<5, 1662 IFF_DONT_BRIDGE = 1<<6, 1663 IFF_DISABLE_NETPOLL = 1<<7, 1664 IFF_MACVLAN_PORT = 1<<8, 1665 IFF_BRIDGE_PORT = 1<<9, 1666 IFF_OVS_DATAPATH = 1<<10, 1667 IFF_TX_SKB_SHARING = 1<<11, 1668 IFF_UNICAST_FLT = 1<<12, 1669 IFF_TEAM_PORT = 1<<13, 1670 IFF_SUPP_NOFCS = 1<<14, 1671 IFF_LIVE_ADDR_CHANGE = 1<<15, 1672 IFF_MACVLAN = 1<<16, 1673 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1674 IFF_L3MDEV_MASTER = 1<<18, 1675 IFF_NO_QUEUE = 1<<19, 1676 IFF_OPENVSWITCH = 1<<20, 1677 IFF_L3MDEV_SLAVE = 1<<21, 1678 IFF_TEAM = 1<<22, 1679 IFF_RXFH_CONFIGURED = 1<<23, 1680 IFF_PHONY_HEADROOM = 1<<24, 1681 IFF_MACSEC = 1<<25, 1682 IFF_NO_RX_HANDLER = 1<<26, 1683 IFF_FAILOVER = 1<<27, 1684 IFF_FAILOVER_SLAVE = 1<<28, 1685 IFF_L3MDEV_RX_HANDLER = 1<<29, 1686 IFF_NO_ADDRCONF = BIT_ULL(30), 1687 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31), 1688 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32), 1689 IFF_SEE_ALL_HWTSTAMP_REQUESTS = BIT_ULL(33), 1690 }; 1691 1692 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1693 #define IFF_EBRIDGE IFF_EBRIDGE 1694 #define IFF_BONDING IFF_BONDING 1695 #define IFF_ISATAP IFF_ISATAP 1696 #define IFF_WAN_HDLC IFF_WAN_HDLC 1697 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1698 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1699 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1700 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1701 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1702 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1703 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1704 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1705 #define IFF_TEAM_PORT IFF_TEAM_PORT 1706 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1707 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1708 #define IFF_MACVLAN IFF_MACVLAN 1709 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1710 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1711 #define IFF_NO_QUEUE IFF_NO_QUEUE 1712 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1713 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1714 #define IFF_TEAM IFF_TEAM 1715 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1716 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM 1717 #define IFF_MACSEC IFF_MACSEC 1718 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1719 #define IFF_FAILOVER IFF_FAILOVER 1720 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1721 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1722 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR 1723 1724 /* Specifies the type of the struct net_device::ml_priv pointer */ 1725 enum netdev_ml_priv_type { 1726 ML_PRIV_NONE, 1727 ML_PRIV_CAN, 1728 }; 1729 1730 enum netdev_stat_type { 1731 NETDEV_PCPU_STAT_NONE, 1732 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */ 1733 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */ 1734 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */ 1735 }; 1736 1737 enum netdev_reg_state { 1738 NETREG_UNINITIALIZED = 0, 1739 NETREG_REGISTERED, /* completed register_netdevice */ 1740 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1741 NETREG_UNREGISTERED, /* completed unregister todo */ 1742 NETREG_RELEASED, /* called free_netdev */ 1743 NETREG_DUMMY, /* dummy device for NAPI poll */ 1744 }; 1745 1746 /** 1747 * struct net_device - The DEVICE structure. 1748 * 1749 * Actually, this whole structure is a big mistake. It mixes I/O 1750 * data with strictly "high-level" data, and it has to know about 1751 * almost every data structure used in the INET module. 1752 * 1753 * @name: This is the first field of the "visible" part of this structure 1754 * (i.e. as seen by users in the "Space.c" file). It is the name 1755 * of the interface. 1756 * 1757 * @name_node: Name hashlist node 1758 * @ifalias: SNMP alias 1759 * @mem_end: Shared memory end 1760 * @mem_start: Shared memory start 1761 * @base_addr: Device I/O address 1762 * @irq: Device IRQ number 1763 * 1764 * @state: Generic network queuing layer state, see netdev_state_t 1765 * @dev_list: The global list of network devices 1766 * @napi_list: List entry used for polling NAPI devices 1767 * @unreg_list: List entry when we are unregistering the 1768 * device; see the function unregister_netdev 1769 * @close_list: List entry used when we are closing the device 1770 * @ptype_all: Device-specific packet handlers for all protocols 1771 * @ptype_specific: Device-specific, protocol-specific packet handlers 1772 * 1773 * @adj_list: Directly linked devices, like slaves for bonding 1774 * @features: Currently active device features 1775 * @hw_features: User-changeable features 1776 * 1777 * @wanted_features: User-requested features 1778 * @vlan_features: Mask of features inheritable by VLAN devices 1779 * 1780 * @hw_enc_features: Mask of features inherited by encapsulating devices 1781 * This field indicates what encapsulation 1782 * offloads the hardware is capable of doing, 1783 * and drivers will need to set them appropriately. 1784 * 1785 * @mpls_features: Mask of features inheritable by MPLS 1786 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1787 * 1788 * @ifindex: interface index 1789 * @group: The group the device belongs to 1790 * 1791 * @stats: Statistics struct, which was left as a legacy, use 1792 * rtnl_link_stats64 instead 1793 * 1794 * @core_stats: core networking counters, 1795 * do not use this in drivers 1796 * @carrier_up_count: Number of times the carrier has been up 1797 * @carrier_down_count: Number of times the carrier has been down 1798 * 1799 * @wireless_handlers: List of functions to handle Wireless Extensions, 1800 * instead of ioctl, 1801 * see <net/iw_handler.h> for details. 1802 * @wireless_data: Instance data managed by the core of wireless extensions 1803 * 1804 * @netdev_ops: Includes several pointers to callbacks, 1805 * if one wants to override the ndo_*() functions 1806 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks. 1807 * @xsk_tx_metadata_ops: Includes pointers to AF_XDP TX metadata callbacks. 1808 * @ethtool_ops: Management operations 1809 * @l3mdev_ops: Layer 3 master device operations 1810 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1811 * discovery handling. Necessary for e.g. 6LoWPAN. 1812 * @xfrmdev_ops: Transformation offload operations 1813 * @tlsdev_ops: Transport Layer Security offload operations 1814 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1815 * of Layer 2 headers. 1816 * 1817 * @flags: Interface flags (a la BSD) 1818 * @xdp_features: XDP capability supported by the device 1819 * @priv_flags: Like 'flags' but invisible to userspace, 1820 * see if.h for the definitions 1821 * @gflags: Global flags ( kept as legacy ) 1822 * @padded: How much padding added by alloc_netdev() 1823 * @operstate: RFC2863 operstate 1824 * @link_mode: Mapping policy to operstate 1825 * @if_port: Selectable AUI, TP, ... 1826 * @dma: DMA channel 1827 * @mtu: Interface MTU value 1828 * @min_mtu: Interface Minimum MTU value 1829 * @max_mtu: Interface Maximum MTU value 1830 * @type: Interface hardware type 1831 * @hard_header_len: Maximum hardware header length. 1832 * @min_header_len: Minimum hardware header length 1833 * 1834 * @needed_headroom: Extra headroom the hardware may need, but not in all 1835 * cases can this be guaranteed 1836 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1837 * cases can this be guaranteed. Some cases also use 1838 * LL_MAX_HEADER instead to allocate the skb 1839 * 1840 * interface address info: 1841 * 1842 * @perm_addr: Permanent hw address 1843 * @addr_assign_type: Hw address assignment type 1844 * @addr_len: Hardware address length 1845 * @upper_level: Maximum depth level of upper devices. 1846 * @lower_level: Maximum depth level of lower devices. 1847 * @neigh_priv_len: Used in neigh_alloc() 1848 * @dev_id: Used to differentiate devices that share 1849 * the same link layer address 1850 * @dev_port: Used to differentiate devices that share 1851 * the same function 1852 * @addr_list_lock: XXX: need comments on this one 1853 * @name_assign_type: network interface name assignment type 1854 * @uc_promisc: Counter that indicates promiscuous mode 1855 * has been enabled due to the need to listen to 1856 * additional unicast addresses in a device that 1857 * does not implement ndo_set_rx_mode() 1858 * @uc: unicast mac addresses 1859 * @mc: multicast mac addresses 1860 * @dev_addrs: list of device hw addresses 1861 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1862 * @promiscuity: Number of times the NIC is told to work in 1863 * promiscuous mode; if it becomes 0 the NIC will 1864 * exit promiscuous mode 1865 * @allmulti: Counter, enables or disables allmulticast mode 1866 * 1867 * @vlan_info: VLAN info 1868 * @dsa_ptr: dsa specific data 1869 * @tipc_ptr: TIPC specific data 1870 * @atalk_ptr: AppleTalk link 1871 * @ip_ptr: IPv4 specific data 1872 * @ip6_ptr: IPv6 specific data 1873 * @ax25_ptr: AX.25 specific data 1874 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1875 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1876 * device struct 1877 * @mpls_ptr: mpls_dev struct pointer 1878 * @mctp_ptr: MCTP specific data 1879 * 1880 * @dev_addr: Hw address (before bcast, 1881 * because most packets are unicast) 1882 * 1883 * @_rx: Array of RX queues 1884 * @num_rx_queues: Number of RX queues 1885 * allocated at register_netdev() time 1886 * @real_num_rx_queues: Number of RX queues currently active in device 1887 * @xdp_prog: XDP sockets filter program pointer 1888 * @gro_flush_timeout: timeout for GRO layer in NAPI 1889 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1890 * allow to avoid NIC hard IRQ, on busy queues. 1891 * 1892 * @rx_handler: handler for received packets 1893 * @rx_handler_data: XXX: need comments on this one 1894 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing 1895 * @ingress_queue: XXX: need comments on this one 1896 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1897 * @broadcast: hw bcast address 1898 * 1899 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1900 * indexed by RX queue number. Assigned by driver. 1901 * This must only be set if the ndo_rx_flow_steer 1902 * operation is defined 1903 * @index_hlist: Device index hash chain 1904 * 1905 * @_tx: Array of TX queues 1906 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1907 * @real_num_tx_queues: Number of TX queues currently active in device 1908 * @qdisc: Root qdisc from userspace point of view 1909 * @tx_queue_len: Max frames per queue allowed 1910 * @tx_global_lock: XXX: need comments on this one 1911 * @xdp_bulkq: XDP device bulk queue 1912 * @xps_maps: all CPUs/RXQs maps for XPS device 1913 * 1914 * @xps_maps: XXX: need comments on this one 1915 * @tcx_egress: BPF & clsact qdisc specific data for egress processing 1916 * @nf_hooks_egress: netfilter hooks executed for egress packets 1917 * @qdisc_hash: qdisc hash table 1918 * @watchdog_timeo: Represents the timeout that is used by 1919 * the watchdog (see dev_watchdog()) 1920 * @watchdog_timer: List of timers 1921 * 1922 * @proto_down_reason: reason a netdev interface is held down 1923 * @pcpu_refcnt: Number of references to this device 1924 * @dev_refcnt: Number of references to this device 1925 * @refcnt_tracker: Tracker directory for tracked references to this device 1926 * @todo_list: Delayed register/unregister 1927 * @link_watch_list: XXX: need comments on this one 1928 * 1929 * @reg_state: Register/unregister state machine 1930 * @dismantle: Device is going to be freed 1931 * @rtnl_link_state: This enum represents the phases of creating 1932 * a new link 1933 * 1934 * @needs_free_netdev: Should unregister perform free_netdev? 1935 * @priv_destructor: Called from unregister 1936 * @npinfo: XXX: need comments on this one 1937 * @nd_net: Network namespace this network device is inside 1938 * 1939 * @ml_priv: Mid-layer private 1940 * @ml_priv_type: Mid-layer private type 1941 * 1942 * @pcpu_stat_type: Type of device statistics which the core should 1943 * allocate/free: none, lstats, tstats, dstats. none 1944 * means the driver is handling statistics allocation/ 1945 * freeing internally. 1946 * @lstats: Loopback statistics: packets, bytes 1947 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes 1948 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes 1949 * 1950 * @garp_port: GARP 1951 * @mrp_port: MRP 1952 * 1953 * @dm_private: Drop monitor private 1954 * 1955 * @dev: Class/net/name entry 1956 * @sysfs_groups: Space for optional device, statistics and wireless 1957 * sysfs groups 1958 * 1959 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1960 * @rtnl_link_ops: Rtnl_link_ops 1961 * @stat_ops: Optional ops for queue-aware statistics 1962 * @queue_mgmt_ops: Optional ops for queue management 1963 * 1964 * @gso_max_size: Maximum size of generic segmentation offload 1965 * @tso_max_size: Device (as in HW) limit on the max TSO request size 1966 * @gso_max_segs: Maximum number of segments that can be passed to the 1967 * NIC for GSO 1968 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count 1969 * @gso_ipv4_max_size: Maximum size of generic segmentation offload, 1970 * for IPv4. 1971 * 1972 * @dcbnl_ops: Data Center Bridging netlink ops 1973 * @num_tc: Number of traffic classes in the net device 1974 * @tc_to_txq: XXX: need comments on this one 1975 * @prio_tc_map: XXX: need comments on this one 1976 * 1977 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1978 * 1979 * @priomap: XXX: need comments on this one 1980 * @phydev: Physical device may attach itself 1981 * for hardware timestamping 1982 * @sfp_bus: attached &struct sfp_bus structure. 1983 * 1984 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1985 * 1986 * @proto_down: protocol port state information can be sent to the 1987 * switch driver and used to set the phys state of the 1988 * switch port. 1989 * 1990 * @threaded: napi threaded mode is enabled 1991 * 1992 * @net_notifier_list: List of per-net netdev notifier block 1993 * that follow this device when it is moved 1994 * to another network namespace. 1995 * 1996 * @macsec_ops: MACsec offloading ops 1997 * 1998 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1999 * offload capabilities of the device 2000 * @udp_tunnel_nic: UDP tunnel offload state 2001 * @ethtool: ethtool related state 2002 * @xdp_state: stores info on attached XDP BPF programs 2003 * 2004 * @nested_level: Used as a parameter of spin_lock_nested() of 2005 * dev->addr_list_lock. 2006 * @unlink_list: As netif_addr_lock() can be called recursively, 2007 * keep a list of interfaces to be deleted. 2008 * @gro_max_size: Maximum size of aggregated packet in generic 2009 * receive offload (GRO) 2010 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic 2011 * receive offload (GRO), for IPv4. 2012 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP 2013 * zero copy driver 2014 * 2015 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 2016 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 2017 * @watchdog_dev_tracker: refcount tracker used by watchdog. 2018 * @dev_registered_tracker: tracker for reference held while 2019 * registered 2020 * @offload_xstats_l3: L3 HW stats for this netdevice. 2021 * 2022 * @devlink_port: Pointer to related devlink port structure. 2023 * Assigned by a driver before netdev registration using 2024 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static 2025 * during the time netdevice is registered. 2026 * 2027 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem, 2028 * where the clock is recovered. 2029 * 2030 * FIXME: cleanup struct net_device such that network protocol info 2031 * moves out. 2032 */ 2033 2034 struct net_device { 2035 /* Cacheline organization can be found documented in 2036 * Documentation/networking/net_cachelines/net_device.rst. 2037 * Please update the document when adding new fields. 2038 */ 2039 2040 /* TX read-mostly hotpath */ 2041 __cacheline_group_begin(net_device_read_tx); 2042 unsigned long long priv_flags; 2043 const struct net_device_ops *netdev_ops; 2044 const struct header_ops *header_ops; 2045 struct netdev_queue *_tx; 2046 netdev_features_t gso_partial_features; 2047 unsigned int real_num_tx_queues; 2048 unsigned int gso_max_size; 2049 unsigned int gso_ipv4_max_size; 2050 u16 gso_max_segs; 2051 s16 num_tc; 2052 /* Note : dev->mtu is often read without holding a lock. 2053 * Writers usually hold RTNL. 2054 * It is recommended to use READ_ONCE() to annotate the reads, 2055 * and to use WRITE_ONCE() to annotate the writes. 2056 */ 2057 unsigned int mtu; 2058 unsigned short needed_headroom; 2059 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2060 #ifdef CONFIG_XPS 2061 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2062 #endif 2063 #ifdef CONFIG_NETFILTER_EGRESS 2064 struct nf_hook_entries __rcu *nf_hooks_egress; 2065 #endif 2066 #ifdef CONFIG_NET_XGRESS 2067 struct bpf_mprog_entry __rcu *tcx_egress; 2068 #endif 2069 __cacheline_group_end(net_device_read_tx); 2070 2071 /* TXRX read-mostly hotpath */ 2072 __cacheline_group_begin(net_device_read_txrx); 2073 union { 2074 struct pcpu_lstats __percpu *lstats; 2075 struct pcpu_sw_netstats __percpu *tstats; 2076 struct pcpu_dstats __percpu *dstats; 2077 }; 2078 unsigned long state; 2079 unsigned int flags; 2080 unsigned short hard_header_len; 2081 netdev_features_t features; 2082 struct inet6_dev __rcu *ip6_ptr; 2083 __cacheline_group_end(net_device_read_txrx); 2084 2085 /* RX read-mostly hotpath */ 2086 __cacheline_group_begin(net_device_read_rx); 2087 struct bpf_prog __rcu *xdp_prog; 2088 struct list_head ptype_specific; 2089 int ifindex; 2090 unsigned int real_num_rx_queues; 2091 struct netdev_rx_queue *_rx; 2092 unsigned long gro_flush_timeout; 2093 int napi_defer_hard_irqs; 2094 unsigned int gro_max_size; 2095 unsigned int gro_ipv4_max_size; 2096 rx_handler_func_t __rcu *rx_handler; 2097 void __rcu *rx_handler_data; 2098 possible_net_t nd_net; 2099 #ifdef CONFIG_NETPOLL 2100 struct netpoll_info __rcu *npinfo; 2101 #endif 2102 #ifdef CONFIG_NET_XGRESS 2103 struct bpf_mprog_entry __rcu *tcx_ingress; 2104 #endif 2105 __cacheline_group_end(net_device_read_rx); 2106 2107 char name[IFNAMSIZ]; 2108 struct netdev_name_node *name_node; 2109 struct dev_ifalias __rcu *ifalias; 2110 /* 2111 * I/O specific fields 2112 * FIXME: Merge these and struct ifmap into one 2113 */ 2114 unsigned long mem_end; 2115 unsigned long mem_start; 2116 unsigned long base_addr; 2117 2118 /* 2119 * Some hardware also needs these fields (state,dev_list, 2120 * napi_list,unreg_list,close_list) but they are not 2121 * part of the usual set specified in Space.c. 2122 */ 2123 2124 2125 struct list_head dev_list; 2126 struct list_head napi_list; 2127 struct list_head unreg_list; 2128 struct list_head close_list; 2129 struct list_head ptype_all; 2130 2131 struct { 2132 struct list_head upper; 2133 struct list_head lower; 2134 } adj_list; 2135 2136 /* Read-mostly cache-line for fast-path access */ 2137 xdp_features_t xdp_features; 2138 const struct xdp_metadata_ops *xdp_metadata_ops; 2139 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops; 2140 unsigned short gflags; 2141 2142 unsigned short needed_tailroom; 2143 2144 netdev_features_t hw_features; 2145 netdev_features_t wanted_features; 2146 netdev_features_t vlan_features; 2147 netdev_features_t hw_enc_features; 2148 netdev_features_t mpls_features; 2149 2150 unsigned int min_mtu; 2151 unsigned int max_mtu; 2152 unsigned short type; 2153 unsigned char min_header_len; 2154 unsigned char name_assign_type; 2155 2156 int group; 2157 2158 struct net_device_stats stats; /* not used by modern drivers */ 2159 2160 struct net_device_core_stats __percpu *core_stats; 2161 2162 /* Stats to monitor link on/off, flapping */ 2163 atomic_t carrier_up_count; 2164 atomic_t carrier_down_count; 2165 2166 #ifdef CONFIG_WIRELESS_EXT 2167 const struct iw_handler_def *wireless_handlers; 2168 struct iw_public_data *wireless_data; 2169 #endif 2170 const struct ethtool_ops *ethtool_ops; 2171 #ifdef CONFIG_NET_L3_MASTER_DEV 2172 const struct l3mdev_ops *l3mdev_ops; 2173 #endif 2174 #if IS_ENABLED(CONFIG_IPV6) 2175 const struct ndisc_ops *ndisc_ops; 2176 #endif 2177 2178 #ifdef CONFIG_XFRM_OFFLOAD 2179 const struct xfrmdev_ops *xfrmdev_ops; 2180 #endif 2181 2182 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2183 const struct tlsdev_ops *tlsdev_ops; 2184 #endif 2185 2186 unsigned int operstate; 2187 unsigned char link_mode; 2188 2189 unsigned char if_port; 2190 unsigned char dma; 2191 2192 /* Interface address info. */ 2193 unsigned char perm_addr[MAX_ADDR_LEN]; 2194 unsigned char addr_assign_type; 2195 unsigned char addr_len; 2196 unsigned char upper_level; 2197 unsigned char lower_level; 2198 2199 unsigned short neigh_priv_len; 2200 unsigned short dev_id; 2201 unsigned short dev_port; 2202 unsigned short padded; 2203 2204 spinlock_t addr_list_lock; 2205 int irq; 2206 2207 struct netdev_hw_addr_list uc; 2208 struct netdev_hw_addr_list mc; 2209 struct netdev_hw_addr_list dev_addrs; 2210 2211 #ifdef CONFIG_SYSFS 2212 struct kset *queues_kset; 2213 #endif 2214 #ifdef CONFIG_LOCKDEP 2215 struct list_head unlink_list; 2216 #endif 2217 unsigned int promiscuity; 2218 unsigned int allmulti; 2219 bool uc_promisc; 2220 #ifdef CONFIG_LOCKDEP 2221 unsigned char nested_level; 2222 #endif 2223 2224 2225 /* Protocol-specific pointers */ 2226 struct in_device __rcu *ip_ptr; 2227 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2228 struct vlan_info __rcu *vlan_info; 2229 #endif 2230 #if IS_ENABLED(CONFIG_NET_DSA) 2231 struct dsa_port *dsa_ptr; 2232 #endif 2233 #if IS_ENABLED(CONFIG_TIPC) 2234 struct tipc_bearer __rcu *tipc_ptr; 2235 #endif 2236 #if IS_ENABLED(CONFIG_ATALK) 2237 void *atalk_ptr; 2238 #endif 2239 #if IS_ENABLED(CONFIG_AX25) 2240 void *ax25_ptr; 2241 #endif 2242 #if IS_ENABLED(CONFIG_CFG80211) 2243 struct wireless_dev *ieee80211_ptr; 2244 #endif 2245 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2246 struct wpan_dev *ieee802154_ptr; 2247 #endif 2248 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2249 struct mpls_dev __rcu *mpls_ptr; 2250 #endif 2251 #if IS_ENABLED(CONFIG_MCTP) 2252 struct mctp_dev __rcu *mctp_ptr; 2253 #endif 2254 2255 /* 2256 * Cache lines mostly used on receive path (including eth_type_trans()) 2257 */ 2258 /* Interface address info used in eth_type_trans() */ 2259 const unsigned char *dev_addr; 2260 2261 unsigned int num_rx_queues; 2262 #define GRO_LEGACY_MAX_SIZE 65536u 2263 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2264 * and shinfo->gso_segs is a 16bit field. 2265 */ 2266 #define GRO_MAX_SIZE (8 * 65535u) 2267 unsigned int xdp_zc_max_segs; 2268 struct netdev_queue __rcu *ingress_queue; 2269 #ifdef CONFIG_NETFILTER_INGRESS 2270 struct nf_hook_entries __rcu *nf_hooks_ingress; 2271 #endif 2272 2273 unsigned char broadcast[MAX_ADDR_LEN]; 2274 #ifdef CONFIG_RFS_ACCEL 2275 struct cpu_rmap *rx_cpu_rmap; 2276 #endif 2277 struct hlist_node index_hlist; 2278 2279 /* 2280 * Cache lines mostly used on transmit path 2281 */ 2282 unsigned int num_tx_queues; 2283 struct Qdisc __rcu *qdisc; 2284 unsigned int tx_queue_len; 2285 spinlock_t tx_global_lock; 2286 2287 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2288 2289 #ifdef CONFIG_NET_SCHED 2290 DECLARE_HASHTABLE (qdisc_hash, 4); 2291 #endif 2292 /* These may be needed for future network-power-down code. */ 2293 struct timer_list watchdog_timer; 2294 int watchdog_timeo; 2295 2296 u32 proto_down_reason; 2297 2298 struct list_head todo_list; 2299 2300 #ifdef CONFIG_PCPU_DEV_REFCNT 2301 int __percpu *pcpu_refcnt; 2302 #else 2303 refcount_t dev_refcnt; 2304 #endif 2305 struct ref_tracker_dir refcnt_tracker; 2306 2307 struct list_head link_watch_list; 2308 2309 u8 reg_state; 2310 2311 bool dismantle; 2312 2313 enum { 2314 RTNL_LINK_INITIALIZED, 2315 RTNL_LINK_INITIALIZING, 2316 } rtnl_link_state:16; 2317 2318 bool needs_free_netdev; 2319 void (*priv_destructor)(struct net_device *dev); 2320 2321 /* mid-layer private */ 2322 void *ml_priv; 2323 enum netdev_ml_priv_type ml_priv_type; 2324 2325 enum netdev_stat_type pcpu_stat_type:8; 2326 2327 #if IS_ENABLED(CONFIG_GARP) 2328 struct garp_port __rcu *garp_port; 2329 #endif 2330 #if IS_ENABLED(CONFIG_MRP) 2331 struct mrp_port __rcu *mrp_port; 2332 #endif 2333 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2334 struct dm_hw_stat_delta __rcu *dm_private; 2335 #endif 2336 struct device dev; 2337 const struct attribute_group *sysfs_groups[4]; 2338 const struct attribute_group *sysfs_rx_queue_group; 2339 2340 const struct rtnl_link_ops *rtnl_link_ops; 2341 2342 const struct netdev_stat_ops *stat_ops; 2343 2344 const struct netdev_queue_mgmt_ops *queue_mgmt_ops; 2345 2346 /* for setting kernel sock attribute on TCP connection setup */ 2347 #define GSO_MAX_SEGS 65535u 2348 #define GSO_LEGACY_MAX_SIZE 65536u 2349 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2350 * and shinfo->gso_segs is a 16bit field. 2351 */ 2352 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2353 2354 #define TSO_LEGACY_MAX_SIZE 65536 2355 #define TSO_MAX_SIZE UINT_MAX 2356 unsigned int tso_max_size; 2357 #define TSO_MAX_SEGS U16_MAX 2358 u16 tso_max_segs; 2359 2360 #ifdef CONFIG_DCB 2361 const struct dcbnl_rtnl_ops *dcbnl_ops; 2362 #endif 2363 u8 prio_tc_map[TC_BITMASK + 1]; 2364 2365 #if IS_ENABLED(CONFIG_FCOE) 2366 unsigned int fcoe_ddp_xid; 2367 #endif 2368 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2369 struct netprio_map __rcu *priomap; 2370 #endif 2371 struct phy_device *phydev; 2372 struct sfp_bus *sfp_bus; 2373 struct lock_class_key *qdisc_tx_busylock; 2374 bool proto_down; 2375 bool threaded; 2376 2377 struct list_head net_notifier_list; 2378 2379 #if IS_ENABLED(CONFIG_MACSEC) 2380 /* MACsec management functions */ 2381 const struct macsec_ops *macsec_ops; 2382 #endif 2383 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2384 struct udp_tunnel_nic *udp_tunnel_nic; 2385 2386 struct ethtool_netdev_state *ethtool; 2387 2388 /* protected by rtnl_lock */ 2389 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2390 2391 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2392 netdevice_tracker linkwatch_dev_tracker; 2393 netdevice_tracker watchdog_dev_tracker; 2394 netdevice_tracker dev_registered_tracker; 2395 struct rtnl_hw_stats64 *offload_xstats_l3; 2396 2397 struct devlink_port *devlink_port; 2398 2399 #if IS_ENABLED(CONFIG_DPLL) 2400 struct dpll_pin __rcu *dpll_pin; 2401 #endif 2402 #if IS_ENABLED(CONFIG_PAGE_POOL) 2403 /** @page_pools: page pools created for this netdevice */ 2404 struct hlist_head page_pools; 2405 #endif 2406 2407 /** @irq_moder: dim parameters used if IS_ENABLED(CONFIG_DIMLIB). */ 2408 struct dim_irq_moder *irq_moder; 2409 }; 2410 #define to_net_dev(d) container_of(d, struct net_device, dev) 2411 2412 /* 2413 * Driver should use this to assign devlink port instance to a netdevice 2414 * before it registers the netdevice. Therefore devlink_port is static 2415 * during the netdev lifetime after it is registered. 2416 */ 2417 #define SET_NETDEV_DEVLINK_PORT(dev, port) \ 2418 ({ \ 2419 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \ 2420 ((dev)->devlink_port = (port)); \ 2421 }) 2422 2423 static inline bool netif_elide_gro(const struct net_device *dev) 2424 { 2425 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2426 return true; 2427 return false; 2428 } 2429 2430 #define NETDEV_ALIGN 32 2431 2432 static inline 2433 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2434 { 2435 return dev->prio_tc_map[prio & TC_BITMASK]; 2436 } 2437 2438 static inline 2439 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2440 { 2441 if (tc >= dev->num_tc) 2442 return -EINVAL; 2443 2444 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2445 return 0; 2446 } 2447 2448 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2449 void netdev_reset_tc(struct net_device *dev); 2450 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2451 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2452 2453 static inline 2454 int netdev_get_num_tc(struct net_device *dev) 2455 { 2456 return dev->num_tc; 2457 } 2458 2459 static inline void net_prefetch(void *p) 2460 { 2461 prefetch(p); 2462 #if L1_CACHE_BYTES < 128 2463 prefetch((u8 *)p + L1_CACHE_BYTES); 2464 #endif 2465 } 2466 2467 static inline void net_prefetchw(void *p) 2468 { 2469 prefetchw(p); 2470 #if L1_CACHE_BYTES < 128 2471 prefetchw((u8 *)p + L1_CACHE_BYTES); 2472 #endif 2473 } 2474 2475 void netdev_unbind_sb_channel(struct net_device *dev, 2476 struct net_device *sb_dev); 2477 int netdev_bind_sb_channel_queue(struct net_device *dev, 2478 struct net_device *sb_dev, 2479 u8 tc, u16 count, u16 offset); 2480 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2481 static inline int netdev_get_sb_channel(struct net_device *dev) 2482 { 2483 return max_t(int, -dev->num_tc, 0); 2484 } 2485 2486 static inline 2487 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2488 unsigned int index) 2489 { 2490 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues); 2491 return &dev->_tx[index]; 2492 } 2493 2494 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2495 const struct sk_buff *skb) 2496 { 2497 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2498 } 2499 2500 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2501 void (*f)(struct net_device *, 2502 struct netdev_queue *, 2503 void *), 2504 void *arg) 2505 { 2506 unsigned int i; 2507 2508 for (i = 0; i < dev->num_tx_queues; i++) 2509 f(dev, &dev->_tx[i], arg); 2510 } 2511 2512 #define netdev_lockdep_set_classes(dev) \ 2513 { \ 2514 static struct lock_class_key qdisc_tx_busylock_key; \ 2515 static struct lock_class_key qdisc_xmit_lock_key; \ 2516 static struct lock_class_key dev_addr_list_lock_key; \ 2517 unsigned int i; \ 2518 \ 2519 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2520 lockdep_set_class(&(dev)->addr_list_lock, \ 2521 &dev_addr_list_lock_key); \ 2522 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2523 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2524 &qdisc_xmit_lock_key); \ 2525 } 2526 2527 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2528 struct net_device *sb_dev); 2529 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2530 struct sk_buff *skb, 2531 struct net_device *sb_dev); 2532 2533 /* returns the headroom that the master device needs to take in account 2534 * when forwarding to this dev 2535 */ 2536 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2537 { 2538 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2539 } 2540 2541 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2542 { 2543 if (dev->netdev_ops->ndo_set_rx_headroom) 2544 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2545 } 2546 2547 /* set the device rx headroom to the dev's default */ 2548 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2549 { 2550 netdev_set_rx_headroom(dev, -1); 2551 } 2552 2553 static inline void *netdev_get_ml_priv(struct net_device *dev, 2554 enum netdev_ml_priv_type type) 2555 { 2556 if (dev->ml_priv_type != type) 2557 return NULL; 2558 2559 return dev->ml_priv; 2560 } 2561 2562 static inline void netdev_set_ml_priv(struct net_device *dev, 2563 void *ml_priv, 2564 enum netdev_ml_priv_type type) 2565 { 2566 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2567 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2568 dev->ml_priv_type, type); 2569 WARN(!dev->ml_priv_type && dev->ml_priv, 2570 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2571 2572 dev->ml_priv = ml_priv; 2573 dev->ml_priv_type = type; 2574 } 2575 2576 /* 2577 * Net namespace inlines 2578 */ 2579 static inline 2580 struct net *dev_net(const struct net_device *dev) 2581 { 2582 return read_pnet(&dev->nd_net); 2583 } 2584 2585 static inline 2586 void dev_net_set(struct net_device *dev, struct net *net) 2587 { 2588 write_pnet(&dev->nd_net, net); 2589 } 2590 2591 /** 2592 * netdev_priv - access network device private data 2593 * @dev: network device 2594 * 2595 * Get network device private data 2596 */ 2597 static inline void *netdev_priv(const struct net_device *dev) 2598 { 2599 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2600 } 2601 2602 /* Set the sysfs physical device reference for the network logical device 2603 * if set prior to registration will cause a symlink during initialization. 2604 */ 2605 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2606 2607 /* Set the sysfs device type for the network logical device to allow 2608 * fine-grained identification of different network device types. For 2609 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2610 */ 2611 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2612 2613 void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index, 2614 enum netdev_queue_type type, 2615 struct napi_struct *napi); 2616 2617 static inline void netif_napi_set_irq(struct napi_struct *napi, int irq) 2618 { 2619 napi->irq = irq; 2620 } 2621 2622 /* Default NAPI poll() weight 2623 * Device drivers are strongly advised to not use bigger value 2624 */ 2625 #define NAPI_POLL_WEIGHT 64 2626 2627 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2628 int (*poll)(struct napi_struct *, int), int weight); 2629 2630 /** 2631 * netif_napi_add() - initialize a NAPI context 2632 * @dev: network device 2633 * @napi: NAPI context 2634 * @poll: polling function 2635 * 2636 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2637 * *any* of the other NAPI-related functions. 2638 */ 2639 static inline void 2640 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2641 int (*poll)(struct napi_struct *, int)) 2642 { 2643 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2644 } 2645 2646 static inline void 2647 netif_napi_add_tx_weight(struct net_device *dev, 2648 struct napi_struct *napi, 2649 int (*poll)(struct napi_struct *, int), 2650 int weight) 2651 { 2652 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2653 netif_napi_add_weight(dev, napi, poll, weight); 2654 } 2655 2656 /** 2657 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2658 * @dev: network device 2659 * @napi: NAPI context 2660 * @poll: polling function 2661 * 2662 * This variant of netif_napi_add() should be used from drivers using NAPI 2663 * to exclusively poll a TX queue. 2664 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2665 */ 2666 static inline void netif_napi_add_tx(struct net_device *dev, 2667 struct napi_struct *napi, 2668 int (*poll)(struct napi_struct *, int)) 2669 { 2670 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2671 } 2672 2673 /** 2674 * __netif_napi_del - remove a NAPI context 2675 * @napi: NAPI context 2676 * 2677 * Warning: caller must observe RCU grace period before freeing memory 2678 * containing @napi. Drivers might want to call this helper to combine 2679 * all the needed RCU grace periods into a single one. 2680 */ 2681 void __netif_napi_del(struct napi_struct *napi); 2682 2683 /** 2684 * netif_napi_del - remove a NAPI context 2685 * @napi: NAPI context 2686 * 2687 * netif_napi_del() removes a NAPI context from the network device NAPI list 2688 */ 2689 static inline void netif_napi_del(struct napi_struct *napi) 2690 { 2691 __netif_napi_del(napi); 2692 synchronize_net(); 2693 } 2694 2695 struct packet_type { 2696 __be16 type; /* This is really htons(ether_type). */ 2697 bool ignore_outgoing; 2698 struct net_device *dev; /* NULL is wildcarded here */ 2699 netdevice_tracker dev_tracker; 2700 int (*func) (struct sk_buff *, 2701 struct net_device *, 2702 struct packet_type *, 2703 struct net_device *); 2704 void (*list_func) (struct list_head *, 2705 struct packet_type *, 2706 struct net_device *); 2707 bool (*id_match)(struct packet_type *ptype, 2708 struct sock *sk); 2709 struct net *af_packet_net; 2710 void *af_packet_priv; 2711 struct list_head list; 2712 }; 2713 2714 struct offload_callbacks { 2715 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2716 netdev_features_t features); 2717 struct sk_buff *(*gro_receive)(struct list_head *head, 2718 struct sk_buff *skb); 2719 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2720 }; 2721 2722 struct packet_offload { 2723 __be16 type; /* This is really htons(ether_type). */ 2724 u16 priority; 2725 struct offload_callbacks callbacks; 2726 struct list_head list; 2727 }; 2728 2729 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2730 struct pcpu_sw_netstats { 2731 u64_stats_t rx_packets; 2732 u64_stats_t rx_bytes; 2733 u64_stats_t tx_packets; 2734 u64_stats_t tx_bytes; 2735 struct u64_stats_sync syncp; 2736 } __aligned(4 * sizeof(u64)); 2737 2738 struct pcpu_dstats { 2739 u64_stats_t rx_packets; 2740 u64_stats_t rx_bytes; 2741 u64_stats_t rx_drops; 2742 u64_stats_t tx_packets; 2743 u64_stats_t tx_bytes; 2744 u64_stats_t tx_drops; 2745 struct u64_stats_sync syncp; 2746 } __aligned(8 * sizeof(u64)); 2747 2748 struct pcpu_lstats { 2749 u64_stats_t packets; 2750 u64_stats_t bytes; 2751 struct u64_stats_sync syncp; 2752 } __aligned(2 * sizeof(u64)); 2753 2754 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2755 2756 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2757 { 2758 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2759 2760 u64_stats_update_begin(&tstats->syncp); 2761 u64_stats_add(&tstats->rx_bytes, len); 2762 u64_stats_inc(&tstats->rx_packets); 2763 u64_stats_update_end(&tstats->syncp); 2764 } 2765 2766 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2767 unsigned int packets, 2768 unsigned int len) 2769 { 2770 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2771 2772 u64_stats_update_begin(&tstats->syncp); 2773 u64_stats_add(&tstats->tx_bytes, len); 2774 u64_stats_add(&tstats->tx_packets, packets); 2775 u64_stats_update_end(&tstats->syncp); 2776 } 2777 2778 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2779 { 2780 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2781 2782 u64_stats_update_begin(&lstats->syncp); 2783 u64_stats_add(&lstats->bytes, len); 2784 u64_stats_inc(&lstats->packets); 2785 u64_stats_update_end(&lstats->syncp); 2786 } 2787 2788 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2789 ({ \ 2790 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2791 if (pcpu_stats) { \ 2792 int __cpu; \ 2793 for_each_possible_cpu(__cpu) { \ 2794 typeof(type) *stat; \ 2795 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2796 u64_stats_init(&stat->syncp); \ 2797 } \ 2798 } \ 2799 pcpu_stats; \ 2800 }) 2801 2802 #define netdev_alloc_pcpu_stats(type) \ 2803 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2804 2805 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2806 ({ \ 2807 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2808 if (pcpu_stats) { \ 2809 int __cpu; \ 2810 for_each_possible_cpu(__cpu) { \ 2811 typeof(type) *stat; \ 2812 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2813 u64_stats_init(&stat->syncp); \ 2814 } \ 2815 } \ 2816 pcpu_stats; \ 2817 }) 2818 2819 enum netdev_lag_tx_type { 2820 NETDEV_LAG_TX_TYPE_UNKNOWN, 2821 NETDEV_LAG_TX_TYPE_RANDOM, 2822 NETDEV_LAG_TX_TYPE_BROADCAST, 2823 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2824 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2825 NETDEV_LAG_TX_TYPE_HASH, 2826 }; 2827 2828 enum netdev_lag_hash { 2829 NETDEV_LAG_HASH_NONE, 2830 NETDEV_LAG_HASH_L2, 2831 NETDEV_LAG_HASH_L34, 2832 NETDEV_LAG_HASH_L23, 2833 NETDEV_LAG_HASH_E23, 2834 NETDEV_LAG_HASH_E34, 2835 NETDEV_LAG_HASH_VLAN_SRCMAC, 2836 NETDEV_LAG_HASH_UNKNOWN, 2837 }; 2838 2839 struct netdev_lag_upper_info { 2840 enum netdev_lag_tx_type tx_type; 2841 enum netdev_lag_hash hash_type; 2842 }; 2843 2844 struct netdev_lag_lower_state_info { 2845 u8 link_up : 1, 2846 tx_enabled : 1; 2847 }; 2848 2849 #include <linux/notifier.h> 2850 2851 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2852 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2853 * adding new types. 2854 */ 2855 enum netdev_cmd { 2856 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2857 NETDEV_DOWN, 2858 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2859 detected a hardware crash and restarted 2860 - we can use this eg to kick tcp sessions 2861 once done */ 2862 NETDEV_CHANGE, /* Notify device state change */ 2863 NETDEV_REGISTER, 2864 NETDEV_UNREGISTER, 2865 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2866 NETDEV_CHANGEADDR, /* notify after the address change */ 2867 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2868 NETDEV_GOING_DOWN, 2869 NETDEV_CHANGENAME, 2870 NETDEV_FEAT_CHANGE, 2871 NETDEV_BONDING_FAILOVER, 2872 NETDEV_PRE_UP, 2873 NETDEV_PRE_TYPE_CHANGE, 2874 NETDEV_POST_TYPE_CHANGE, 2875 NETDEV_POST_INIT, 2876 NETDEV_PRE_UNINIT, 2877 NETDEV_RELEASE, 2878 NETDEV_NOTIFY_PEERS, 2879 NETDEV_JOIN, 2880 NETDEV_CHANGEUPPER, 2881 NETDEV_RESEND_IGMP, 2882 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2883 NETDEV_CHANGEINFODATA, 2884 NETDEV_BONDING_INFO, 2885 NETDEV_PRECHANGEUPPER, 2886 NETDEV_CHANGELOWERSTATE, 2887 NETDEV_UDP_TUNNEL_PUSH_INFO, 2888 NETDEV_UDP_TUNNEL_DROP_INFO, 2889 NETDEV_CHANGE_TX_QUEUE_LEN, 2890 NETDEV_CVLAN_FILTER_PUSH_INFO, 2891 NETDEV_CVLAN_FILTER_DROP_INFO, 2892 NETDEV_SVLAN_FILTER_PUSH_INFO, 2893 NETDEV_SVLAN_FILTER_DROP_INFO, 2894 NETDEV_OFFLOAD_XSTATS_ENABLE, 2895 NETDEV_OFFLOAD_XSTATS_DISABLE, 2896 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 2897 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 2898 NETDEV_XDP_FEAT_CHANGE, 2899 }; 2900 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2901 2902 int register_netdevice_notifier(struct notifier_block *nb); 2903 int unregister_netdevice_notifier(struct notifier_block *nb); 2904 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2905 int unregister_netdevice_notifier_net(struct net *net, 2906 struct notifier_block *nb); 2907 int register_netdevice_notifier_dev_net(struct net_device *dev, 2908 struct notifier_block *nb, 2909 struct netdev_net_notifier *nn); 2910 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2911 struct notifier_block *nb, 2912 struct netdev_net_notifier *nn); 2913 2914 struct netdev_notifier_info { 2915 struct net_device *dev; 2916 struct netlink_ext_ack *extack; 2917 }; 2918 2919 struct netdev_notifier_info_ext { 2920 struct netdev_notifier_info info; /* must be first */ 2921 union { 2922 u32 mtu; 2923 } ext; 2924 }; 2925 2926 struct netdev_notifier_change_info { 2927 struct netdev_notifier_info info; /* must be first */ 2928 unsigned int flags_changed; 2929 }; 2930 2931 struct netdev_notifier_changeupper_info { 2932 struct netdev_notifier_info info; /* must be first */ 2933 struct net_device *upper_dev; /* new upper dev */ 2934 bool master; /* is upper dev master */ 2935 bool linking; /* is the notification for link or unlink */ 2936 void *upper_info; /* upper dev info */ 2937 }; 2938 2939 struct netdev_notifier_changelowerstate_info { 2940 struct netdev_notifier_info info; /* must be first */ 2941 void *lower_state_info; /* is lower dev state */ 2942 }; 2943 2944 struct netdev_notifier_pre_changeaddr_info { 2945 struct netdev_notifier_info info; /* must be first */ 2946 const unsigned char *dev_addr; 2947 }; 2948 2949 enum netdev_offload_xstats_type { 2950 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 2951 }; 2952 2953 struct netdev_notifier_offload_xstats_info { 2954 struct netdev_notifier_info info; /* must be first */ 2955 enum netdev_offload_xstats_type type; 2956 2957 union { 2958 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 2959 struct netdev_notifier_offload_xstats_rd *report_delta; 2960 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 2961 struct netdev_notifier_offload_xstats_ru *report_used; 2962 }; 2963 }; 2964 2965 int netdev_offload_xstats_enable(struct net_device *dev, 2966 enum netdev_offload_xstats_type type, 2967 struct netlink_ext_ack *extack); 2968 int netdev_offload_xstats_disable(struct net_device *dev, 2969 enum netdev_offload_xstats_type type); 2970 bool netdev_offload_xstats_enabled(const struct net_device *dev, 2971 enum netdev_offload_xstats_type type); 2972 int netdev_offload_xstats_get(struct net_device *dev, 2973 enum netdev_offload_xstats_type type, 2974 struct rtnl_hw_stats64 *stats, bool *used, 2975 struct netlink_ext_ack *extack); 2976 void 2977 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 2978 const struct rtnl_hw_stats64 *stats); 2979 void 2980 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 2981 void netdev_offload_xstats_push_delta(struct net_device *dev, 2982 enum netdev_offload_xstats_type type, 2983 const struct rtnl_hw_stats64 *stats); 2984 2985 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2986 struct net_device *dev) 2987 { 2988 info->dev = dev; 2989 info->extack = NULL; 2990 } 2991 2992 static inline struct net_device * 2993 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2994 { 2995 return info->dev; 2996 } 2997 2998 static inline struct netlink_ext_ack * 2999 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 3000 { 3001 return info->extack; 3002 } 3003 3004 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 3005 int call_netdevice_notifiers_info(unsigned long val, 3006 struct netdev_notifier_info *info); 3007 3008 #define for_each_netdev(net, d) \ 3009 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 3010 #define for_each_netdev_reverse(net, d) \ 3011 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 3012 #define for_each_netdev_rcu(net, d) \ 3013 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 3014 #define for_each_netdev_safe(net, d, n) \ 3015 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 3016 #define for_each_netdev_continue(net, d) \ 3017 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 3018 #define for_each_netdev_continue_reverse(net, d) \ 3019 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 3020 dev_list) 3021 #define for_each_netdev_continue_rcu(net, d) \ 3022 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 3023 #define for_each_netdev_in_bond_rcu(bond, slave) \ 3024 for_each_netdev_rcu(&init_net, slave) \ 3025 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 3026 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 3027 3028 #define for_each_netdev_dump(net, d, ifindex) \ 3029 for (; (d = xa_find(&(net)->dev_by_index, &ifindex, \ 3030 ULONG_MAX, XA_PRESENT)); ifindex++) 3031 3032 static inline struct net_device *next_net_device(struct net_device *dev) 3033 { 3034 struct list_head *lh; 3035 struct net *net; 3036 3037 net = dev_net(dev); 3038 lh = dev->dev_list.next; 3039 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3040 } 3041 3042 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 3043 { 3044 struct list_head *lh; 3045 struct net *net; 3046 3047 net = dev_net(dev); 3048 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 3049 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3050 } 3051 3052 static inline struct net_device *first_net_device(struct net *net) 3053 { 3054 return list_empty(&net->dev_base_head) ? NULL : 3055 net_device_entry(net->dev_base_head.next); 3056 } 3057 3058 static inline struct net_device *first_net_device_rcu(struct net *net) 3059 { 3060 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 3061 3062 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3063 } 3064 3065 int netdev_boot_setup_check(struct net_device *dev); 3066 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 3067 const char *hwaddr); 3068 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 3069 void dev_add_pack(struct packet_type *pt); 3070 void dev_remove_pack(struct packet_type *pt); 3071 void __dev_remove_pack(struct packet_type *pt); 3072 void dev_add_offload(struct packet_offload *po); 3073 void dev_remove_offload(struct packet_offload *po); 3074 3075 int dev_get_iflink(const struct net_device *dev); 3076 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 3077 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 3078 struct net_device_path_stack *stack); 3079 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 3080 unsigned short mask); 3081 struct net_device *dev_get_by_name(struct net *net, const char *name); 3082 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 3083 struct net_device *__dev_get_by_name(struct net *net, const char *name); 3084 bool netdev_name_in_use(struct net *net, const char *name); 3085 int dev_alloc_name(struct net_device *dev, const char *name); 3086 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 3087 void dev_close(struct net_device *dev); 3088 void dev_close_many(struct list_head *head, bool unlink); 3089 void dev_disable_lro(struct net_device *dev); 3090 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 3091 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 3092 struct net_device *sb_dev); 3093 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 3094 struct net_device *sb_dev); 3095 3096 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3097 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3098 3099 static inline int dev_queue_xmit(struct sk_buff *skb) 3100 { 3101 return __dev_queue_xmit(skb, NULL); 3102 } 3103 3104 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3105 struct net_device *sb_dev) 3106 { 3107 return __dev_queue_xmit(skb, sb_dev); 3108 } 3109 3110 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3111 { 3112 int ret; 3113 3114 ret = __dev_direct_xmit(skb, queue_id); 3115 if (!dev_xmit_complete(ret)) 3116 kfree_skb(skb); 3117 return ret; 3118 } 3119 3120 int register_netdevice(struct net_device *dev); 3121 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3122 void unregister_netdevice_many(struct list_head *head); 3123 static inline void unregister_netdevice(struct net_device *dev) 3124 { 3125 unregister_netdevice_queue(dev, NULL); 3126 } 3127 3128 int netdev_refcnt_read(const struct net_device *dev); 3129 void free_netdev(struct net_device *dev); 3130 void netdev_freemem(struct net_device *dev); 3131 void init_dummy_netdev(struct net_device *dev); 3132 3133 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3134 struct sk_buff *skb, 3135 bool all_slaves); 3136 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3137 struct sock *sk); 3138 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3139 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3140 struct net_device *netdev_get_by_index(struct net *net, int ifindex, 3141 netdevice_tracker *tracker, gfp_t gfp); 3142 struct net_device *netdev_get_by_name(struct net *net, const char *name, 3143 netdevice_tracker *tracker, gfp_t gfp); 3144 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3145 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 3146 void netdev_copy_name(struct net_device *dev, char *name); 3147 3148 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3149 unsigned short type, 3150 const void *daddr, const void *saddr, 3151 unsigned int len) 3152 { 3153 if (!dev->header_ops || !dev->header_ops->create) 3154 return 0; 3155 3156 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3157 } 3158 3159 static inline int dev_parse_header(const struct sk_buff *skb, 3160 unsigned char *haddr) 3161 { 3162 const struct net_device *dev = skb->dev; 3163 3164 if (!dev->header_ops || !dev->header_ops->parse) 3165 return 0; 3166 return dev->header_ops->parse(skb, haddr); 3167 } 3168 3169 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3170 { 3171 const struct net_device *dev = skb->dev; 3172 3173 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3174 return 0; 3175 return dev->header_ops->parse_protocol(skb); 3176 } 3177 3178 /* ll_header must have at least hard_header_len allocated */ 3179 static inline bool dev_validate_header(const struct net_device *dev, 3180 char *ll_header, int len) 3181 { 3182 if (likely(len >= dev->hard_header_len)) 3183 return true; 3184 if (len < dev->min_header_len) 3185 return false; 3186 3187 if (capable(CAP_SYS_RAWIO)) { 3188 memset(ll_header + len, 0, dev->hard_header_len - len); 3189 return true; 3190 } 3191 3192 if (dev->header_ops && dev->header_ops->validate) 3193 return dev->header_ops->validate(ll_header, len); 3194 3195 return false; 3196 } 3197 3198 static inline bool dev_has_header(const struct net_device *dev) 3199 { 3200 return dev->header_ops && dev->header_ops->create; 3201 } 3202 3203 /* 3204 * Incoming packets are placed on per-CPU queues 3205 */ 3206 struct softnet_data { 3207 struct list_head poll_list; 3208 struct sk_buff_head process_queue; 3209 local_lock_t process_queue_bh_lock; 3210 3211 /* stats */ 3212 unsigned int processed; 3213 unsigned int time_squeeze; 3214 #ifdef CONFIG_RPS 3215 struct softnet_data *rps_ipi_list; 3216 #endif 3217 3218 unsigned int received_rps; 3219 bool in_net_rx_action; 3220 bool in_napi_threaded_poll; 3221 3222 #ifdef CONFIG_NET_FLOW_LIMIT 3223 struct sd_flow_limit __rcu *flow_limit; 3224 #endif 3225 struct Qdisc *output_queue; 3226 struct Qdisc **output_queue_tailp; 3227 struct sk_buff *completion_queue; 3228 #ifdef CONFIG_XFRM_OFFLOAD 3229 struct sk_buff_head xfrm_backlog; 3230 #endif 3231 /* written and read only by owning cpu: */ 3232 struct netdev_xmit xmit; 3233 #ifdef CONFIG_RPS 3234 /* input_queue_head should be written by cpu owning this struct, 3235 * and only read by other cpus. Worth using a cache line. 3236 */ 3237 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3238 3239 /* Elements below can be accessed between CPUs for RPS/RFS */ 3240 call_single_data_t csd ____cacheline_aligned_in_smp; 3241 struct softnet_data *rps_ipi_next; 3242 unsigned int cpu; 3243 unsigned int input_queue_tail; 3244 #endif 3245 struct sk_buff_head input_pkt_queue; 3246 struct napi_struct backlog; 3247 3248 atomic_t dropped ____cacheline_aligned_in_smp; 3249 3250 /* Another possibly contended cache line */ 3251 spinlock_t defer_lock ____cacheline_aligned_in_smp; 3252 int defer_count; 3253 int defer_ipi_scheduled; 3254 struct sk_buff *defer_list; 3255 call_single_data_t defer_csd; 3256 }; 3257 3258 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3259 3260 #ifndef CONFIG_PREEMPT_RT 3261 static inline int dev_recursion_level(void) 3262 { 3263 return this_cpu_read(softnet_data.xmit.recursion); 3264 } 3265 #else 3266 static inline int dev_recursion_level(void) 3267 { 3268 return current->net_xmit.recursion; 3269 } 3270 3271 #endif 3272 3273 void __netif_schedule(struct Qdisc *q); 3274 void netif_schedule_queue(struct netdev_queue *txq); 3275 3276 static inline void netif_tx_schedule_all(struct net_device *dev) 3277 { 3278 unsigned int i; 3279 3280 for (i = 0; i < dev->num_tx_queues; i++) 3281 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3282 } 3283 3284 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3285 { 3286 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3287 } 3288 3289 /** 3290 * netif_start_queue - allow transmit 3291 * @dev: network device 3292 * 3293 * Allow upper layers to call the device hard_start_xmit routine. 3294 */ 3295 static inline void netif_start_queue(struct net_device *dev) 3296 { 3297 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3298 } 3299 3300 static inline void netif_tx_start_all_queues(struct net_device *dev) 3301 { 3302 unsigned int i; 3303 3304 for (i = 0; i < dev->num_tx_queues; i++) { 3305 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3306 netif_tx_start_queue(txq); 3307 } 3308 } 3309 3310 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3311 3312 /** 3313 * netif_wake_queue - restart transmit 3314 * @dev: network device 3315 * 3316 * Allow upper layers to call the device hard_start_xmit routine. 3317 * Used for flow control when transmit resources are available. 3318 */ 3319 static inline void netif_wake_queue(struct net_device *dev) 3320 { 3321 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3322 } 3323 3324 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3325 { 3326 unsigned int i; 3327 3328 for (i = 0; i < dev->num_tx_queues; i++) { 3329 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3330 netif_tx_wake_queue(txq); 3331 } 3332 } 3333 3334 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3335 { 3336 /* Must be an atomic op see netif_txq_try_stop() */ 3337 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3338 } 3339 3340 /** 3341 * netif_stop_queue - stop transmitted packets 3342 * @dev: network device 3343 * 3344 * Stop upper layers calling the device hard_start_xmit routine. 3345 * Used for flow control when transmit resources are unavailable. 3346 */ 3347 static inline void netif_stop_queue(struct net_device *dev) 3348 { 3349 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3350 } 3351 3352 void netif_tx_stop_all_queues(struct net_device *dev); 3353 3354 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3355 { 3356 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3357 } 3358 3359 /** 3360 * netif_queue_stopped - test if transmit queue is flowblocked 3361 * @dev: network device 3362 * 3363 * Test if transmit queue on device is currently unable to send. 3364 */ 3365 static inline bool netif_queue_stopped(const struct net_device *dev) 3366 { 3367 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3368 } 3369 3370 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3371 { 3372 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3373 } 3374 3375 static inline bool 3376 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3377 { 3378 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3379 } 3380 3381 static inline bool 3382 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3383 { 3384 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3385 } 3386 3387 /** 3388 * netdev_queue_set_dql_min_limit - set dql minimum limit 3389 * @dev_queue: pointer to transmit queue 3390 * @min_limit: dql minimum limit 3391 * 3392 * Forces xmit_more() to return true until the minimum threshold 3393 * defined by @min_limit is reached (or until the tx queue is 3394 * empty). Warning: to be use with care, misuse will impact the 3395 * latency. 3396 */ 3397 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3398 unsigned int min_limit) 3399 { 3400 #ifdef CONFIG_BQL 3401 dev_queue->dql.min_limit = min_limit; 3402 #endif 3403 } 3404 3405 static inline int netdev_queue_dql_avail(const struct netdev_queue *txq) 3406 { 3407 #ifdef CONFIG_BQL 3408 /* Non-BQL migrated drivers will return 0, too. */ 3409 return dql_avail(&txq->dql); 3410 #else 3411 return 0; 3412 #endif 3413 } 3414 3415 /** 3416 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3417 * @dev_queue: pointer to transmit queue 3418 * 3419 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3420 * to give appropriate hint to the CPU. 3421 */ 3422 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3423 { 3424 #ifdef CONFIG_BQL 3425 prefetchw(&dev_queue->dql.num_queued); 3426 #endif 3427 } 3428 3429 /** 3430 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3431 * @dev_queue: pointer to transmit queue 3432 * 3433 * BQL enabled drivers might use this helper in their TX completion path, 3434 * to give appropriate hint to the CPU. 3435 */ 3436 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3437 { 3438 #ifdef CONFIG_BQL 3439 prefetchw(&dev_queue->dql.limit); 3440 #endif 3441 } 3442 3443 /** 3444 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3445 * @dev_queue: network device queue 3446 * @bytes: number of bytes queued to the device queue 3447 * 3448 * Report the number of bytes queued for sending/completion to the network 3449 * device hardware queue. @bytes should be a good approximation and should 3450 * exactly match netdev_completed_queue() @bytes. 3451 * This is typically called once per packet, from ndo_start_xmit(). 3452 */ 3453 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3454 unsigned int bytes) 3455 { 3456 #ifdef CONFIG_BQL 3457 dql_queued(&dev_queue->dql, bytes); 3458 3459 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3460 return; 3461 3462 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3463 3464 /* 3465 * The XOFF flag must be set before checking the dql_avail below, 3466 * because in netdev_tx_completed_queue we update the dql_completed 3467 * before checking the XOFF flag. 3468 */ 3469 smp_mb(); 3470 3471 /* check again in case another CPU has just made room avail */ 3472 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3473 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3474 #endif 3475 } 3476 3477 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3478 * that they should not test BQL status themselves. 3479 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3480 * skb of a batch. 3481 * Returns true if the doorbell must be used to kick the NIC. 3482 */ 3483 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3484 unsigned int bytes, 3485 bool xmit_more) 3486 { 3487 if (xmit_more) { 3488 #ifdef CONFIG_BQL 3489 dql_queued(&dev_queue->dql, bytes); 3490 #endif 3491 return netif_tx_queue_stopped(dev_queue); 3492 } 3493 netdev_tx_sent_queue(dev_queue, bytes); 3494 return true; 3495 } 3496 3497 /** 3498 * netdev_sent_queue - report the number of bytes queued to hardware 3499 * @dev: network device 3500 * @bytes: number of bytes queued to the hardware device queue 3501 * 3502 * Report the number of bytes queued for sending/completion to the network 3503 * device hardware queue#0. @bytes should be a good approximation and should 3504 * exactly match netdev_completed_queue() @bytes. 3505 * This is typically called once per packet, from ndo_start_xmit(). 3506 */ 3507 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3508 { 3509 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3510 } 3511 3512 static inline bool __netdev_sent_queue(struct net_device *dev, 3513 unsigned int bytes, 3514 bool xmit_more) 3515 { 3516 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3517 xmit_more); 3518 } 3519 3520 /** 3521 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3522 * @dev_queue: network device queue 3523 * @pkts: number of packets (currently ignored) 3524 * @bytes: number of bytes dequeued from the device queue 3525 * 3526 * Must be called at most once per TX completion round (and not per 3527 * individual packet), so that BQL can adjust its limits appropriately. 3528 */ 3529 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3530 unsigned int pkts, unsigned int bytes) 3531 { 3532 #ifdef CONFIG_BQL 3533 if (unlikely(!bytes)) 3534 return; 3535 3536 dql_completed(&dev_queue->dql, bytes); 3537 3538 /* 3539 * Without the memory barrier there is a small possiblity that 3540 * netdev_tx_sent_queue will miss the update and cause the queue to 3541 * be stopped forever 3542 */ 3543 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */ 3544 3545 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3546 return; 3547 3548 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3549 netif_schedule_queue(dev_queue); 3550 #endif 3551 } 3552 3553 /** 3554 * netdev_completed_queue - report bytes and packets completed by device 3555 * @dev: network device 3556 * @pkts: actual number of packets sent over the medium 3557 * @bytes: actual number of bytes sent over the medium 3558 * 3559 * Report the number of bytes and packets transmitted by the network device 3560 * hardware queue over the physical medium, @bytes must exactly match the 3561 * @bytes amount passed to netdev_sent_queue() 3562 */ 3563 static inline void netdev_completed_queue(struct net_device *dev, 3564 unsigned int pkts, unsigned int bytes) 3565 { 3566 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3567 } 3568 3569 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3570 { 3571 #ifdef CONFIG_BQL 3572 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3573 dql_reset(&q->dql); 3574 #endif 3575 } 3576 3577 /** 3578 * netdev_reset_queue - reset the packets and bytes count of a network device 3579 * @dev_queue: network device 3580 * 3581 * Reset the bytes and packet count of a network device and clear the 3582 * software flow control OFF bit for this network device 3583 */ 3584 static inline void netdev_reset_queue(struct net_device *dev_queue) 3585 { 3586 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3587 } 3588 3589 /** 3590 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3591 * @dev: network device 3592 * @queue_index: given tx queue index 3593 * 3594 * Returns 0 if given tx queue index >= number of device tx queues, 3595 * otherwise returns the originally passed tx queue index. 3596 */ 3597 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3598 { 3599 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3600 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3601 dev->name, queue_index, 3602 dev->real_num_tx_queues); 3603 return 0; 3604 } 3605 3606 return queue_index; 3607 } 3608 3609 /** 3610 * netif_running - test if up 3611 * @dev: network device 3612 * 3613 * Test if the device has been brought up. 3614 */ 3615 static inline bool netif_running(const struct net_device *dev) 3616 { 3617 return test_bit(__LINK_STATE_START, &dev->state); 3618 } 3619 3620 /* 3621 * Routines to manage the subqueues on a device. We only need start, 3622 * stop, and a check if it's stopped. All other device management is 3623 * done at the overall netdevice level. 3624 * Also test the device if we're multiqueue. 3625 */ 3626 3627 /** 3628 * netif_start_subqueue - allow sending packets on subqueue 3629 * @dev: network device 3630 * @queue_index: sub queue index 3631 * 3632 * Start individual transmit queue of a device with multiple transmit queues. 3633 */ 3634 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3635 { 3636 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3637 3638 netif_tx_start_queue(txq); 3639 } 3640 3641 /** 3642 * netif_stop_subqueue - stop sending packets on subqueue 3643 * @dev: network device 3644 * @queue_index: sub queue index 3645 * 3646 * Stop individual transmit queue of a device with multiple transmit queues. 3647 */ 3648 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3649 { 3650 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3651 netif_tx_stop_queue(txq); 3652 } 3653 3654 /** 3655 * __netif_subqueue_stopped - test status of subqueue 3656 * @dev: network device 3657 * @queue_index: sub queue index 3658 * 3659 * Check individual transmit queue of a device with multiple transmit queues. 3660 */ 3661 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3662 u16 queue_index) 3663 { 3664 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3665 3666 return netif_tx_queue_stopped(txq); 3667 } 3668 3669 /** 3670 * netif_subqueue_stopped - test status of subqueue 3671 * @dev: network device 3672 * @skb: sub queue buffer pointer 3673 * 3674 * Check individual transmit queue of a device with multiple transmit queues. 3675 */ 3676 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3677 struct sk_buff *skb) 3678 { 3679 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3680 } 3681 3682 /** 3683 * netif_wake_subqueue - allow sending packets on subqueue 3684 * @dev: network device 3685 * @queue_index: sub queue index 3686 * 3687 * Resume individual transmit queue of a device with multiple transmit queues. 3688 */ 3689 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3690 { 3691 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3692 3693 netif_tx_wake_queue(txq); 3694 } 3695 3696 #ifdef CONFIG_XPS 3697 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3698 u16 index); 3699 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3700 u16 index, enum xps_map_type type); 3701 3702 /** 3703 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3704 * @j: CPU/Rx queue index 3705 * @mask: bitmask of all cpus/rx queues 3706 * @nr_bits: number of bits in the bitmask 3707 * 3708 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3709 */ 3710 static inline bool netif_attr_test_mask(unsigned long j, 3711 const unsigned long *mask, 3712 unsigned int nr_bits) 3713 { 3714 cpu_max_bits_warn(j, nr_bits); 3715 return test_bit(j, mask); 3716 } 3717 3718 /** 3719 * netif_attr_test_online - Test for online CPU/Rx queue 3720 * @j: CPU/Rx queue index 3721 * @online_mask: bitmask for CPUs/Rx queues that are online 3722 * @nr_bits: number of bits in the bitmask 3723 * 3724 * Returns true if a CPU/Rx queue is online. 3725 */ 3726 static inline bool netif_attr_test_online(unsigned long j, 3727 const unsigned long *online_mask, 3728 unsigned int nr_bits) 3729 { 3730 cpu_max_bits_warn(j, nr_bits); 3731 3732 if (online_mask) 3733 return test_bit(j, online_mask); 3734 3735 return (j < nr_bits); 3736 } 3737 3738 /** 3739 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3740 * @n: CPU/Rx queue index 3741 * @srcp: the cpumask/Rx queue mask pointer 3742 * @nr_bits: number of bits in the bitmask 3743 * 3744 * Returns >= nr_bits if no further CPUs/Rx queues set. 3745 */ 3746 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3747 unsigned int nr_bits) 3748 { 3749 /* -1 is a legal arg here. */ 3750 if (n != -1) 3751 cpu_max_bits_warn(n, nr_bits); 3752 3753 if (srcp) 3754 return find_next_bit(srcp, nr_bits, n + 1); 3755 3756 return n + 1; 3757 } 3758 3759 /** 3760 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3761 * @n: CPU/Rx queue index 3762 * @src1p: the first CPUs/Rx queues mask pointer 3763 * @src2p: the second CPUs/Rx queues mask pointer 3764 * @nr_bits: number of bits in the bitmask 3765 * 3766 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3767 */ 3768 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3769 const unsigned long *src2p, 3770 unsigned int nr_bits) 3771 { 3772 /* -1 is a legal arg here. */ 3773 if (n != -1) 3774 cpu_max_bits_warn(n, nr_bits); 3775 3776 if (src1p && src2p) 3777 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3778 else if (src1p) 3779 return find_next_bit(src1p, nr_bits, n + 1); 3780 else if (src2p) 3781 return find_next_bit(src2p, nr_bits, n + 1); 3782 3783 return n + 1; 3784 } 3785 #else 3786 static inline int netif_set_xps_queue(struct net_device *dev, 3787 const struct cpumask *mask, 3788 u16 index) 3789 { 3790 return 0; 3791 } 3792 3793 static inline int __netif_set_xps_queue(struct net_device *dev, 3794 const unsigned long *mask, 3795 u16 index, enum xps_map_type type) 3796 { 3797 return 0; 3798 } 3799 #endif 3800 3801 /** 3802 * netif_is_multiqueue - test if device has multiple transmit queues 3803 * @dev: network device 3804 * 3805 * Check if device has multiple transmit queues 3806 */ 3807 static inline bool netif_is_multiqueue(const struct net_device *dev) 3808 { 3809 return dev->num_tx_queues > 1; 3810 } 3811 3812 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3813 3814 #ifdef CONFIG_SYSFS 3815 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3816 #else 3817 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3818 unsigned int rxqs) 3819 { 3820 dev->real_num_rx_queues = rxqs; 3821 return 0; 3822 } 3823 #endif 3824 int netif_set_real_num_queues(struct net_device *dev, 3825 unsigned int txq, unsigned int rxq); 3826 3827 int netif_get_num_default_rss_queues(void); 3828 3829 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason); 3830 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason); 3831 3832 /* 3833 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3834 * interrupt context or with hardware interrupts being disabled. 3835 * (in_hardirq() || irqs_disabled()) 3836 * 3837 * We provide four helpers that can be used in following contexts : 3838 * 3839 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3840 * replacing kfree_skb(skb) 3841 * 3842 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3843 * Typically used in place of consume_skb(skb) in TX completion path 3844 * 3845 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3846 * replacing kfree_skb(skb) 3847 * 3848 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3849 * and consumed a packet. Used in place of consume_skb(skb) 3850 */ 3851 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3852 { 3853 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 3854 } 3855 3856 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3857 { 3858 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED); 3859 } 3860 3861 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3862 { 3863 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 3864 } 3865 3866 static inline void dev_consume_skb_any(struct sk_buff *skb) 3867 { 3868 dev_kfree_skb_any_reason(skb, SKB_CONSUMED); 3869 } 3870 3871 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3872 struct bpf_prog *xdp_prog); 3873 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3874 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff **pskb); 3875 int netif_rx(struct sk_buff *skb); 3876 int __netif_rx(struct sk_buff *skb); 3877 3878 int netif_receive_skb(struct sk_buff *skb); 3879 int netif_receive_skb_core(struct sk_buff *skb); 3880 void netif_receive_skb_list_internal(struct list_head *head); 3881 void netif_receive_skb_list(struct list_head *head); 3882 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3883 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3884 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3885 void napi_get_frags_check(struct napi_struct *napi); 3886 gro_result_t napi_gro_frags(struct napi_struct *napi); 3887 3888 static inline void napi_free_frags(struct napi_struct *napi) 3889 { 3890 kfree_skb(napi->skb); 3891 napi->skb = NULL; 3892 } 3893 3894 bool netdev_is_rx_handler_busy(struct net_device *dev); 3895 int netdev_rx_handler_register(struct net_device *dev, 3896 rx_handler_func_t *rx_handler, 3897 void *rx_handler_data); 3898 void netdev_rx_handler_unregister(struct net_device *dev); 3899 3900 bool dev_valid_name(const char *name); 3901 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 3902 { 3903 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 3904 } 3905 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 3906 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 3907 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3908 void __user *data, bool *need_copyout); 3909 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 3910 int generic_hwtstamp_get_lower(struct net_device *dev, 3911 struct kernel_hwtstamp_config *kernel_cfg); 3912 int generic_hwtstamp_set_lower(struct net_device *dev, 3913 struct kernel_hwtstamp_config *kernel_cfg, 3914 struct netlink_ext_ack *extack); 3915 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 3916 unsigned int dev_get_flags(const struct net_device *); 3917 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3918 struct netlink_ext_ack *extack); 3919 int dev_change_flags(struct net_device *dev, unsigned int flags, 3920 struct netlink_ext_ack *extack); 3921 int dev_set_alias(struct net_device *, const char *, size_t); 3922 int dev_get_alias(const struct net_device *, char *, size_t); 3923 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 3924 const char *pat, int new_ifindex); 3925 static inline 3926 int dev_change_net_namespace(struct net_device *dev, struct net *net, 3927 const char *pat) 3928 { 3929 return __dev_change_net_namespace(dev, net, pat, 0); 3930 } 3931 int __dev_set_mtu(struct net_device *, int); 3932 int dev_set_mtu(struct net_device *, int); 3933 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3934 struct netlink_ext_ack *extack); 3935 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3936 struct netlink_ext_ack *extack); 3937 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 3938 struct netlink_ext_ack *extack); 3939 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 3940 int dev_get_port_parent_id(struct net_device *dev, 3941 struct netdev_phys_item_id *ppid, bool recurse); 3942 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3943 3944 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3945 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3946 struct netdev_queue *txq, int *ret); 3947 3948 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3949 u8 dev_xdp_prog_count(struct net_device *dev); 3950 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3951 3952 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3953 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3954 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 3955 bool is_skb_forwardable(const struct net_device *dev, 3956 const struct sk_buff *skb); 3957 3958 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 3959 const struct sk_buff *skb, 3960 const bool check_mtu) 3961 { 3962 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 3963 unsigned int len; 3964 3965 if (!(dev->flags & IFF_UP)) 3966 return false; 3967 3968 if (!check_mtu) 3969 return true; 3970 3971 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 3972 if (skb->len <= len) 3973 return true; 3974 3975 /* if TSO is enabled, we don't care about the length as the packet 3976 * could be forwarded without being segmented before 3977 */ 3978 if (skb_is_gso(skb)) 3979 return true; 3980 3981 return false; 3982 } 3983 3984 void netdev_core_stats_inc(struct net_device *dev, u32 offset); 3985 3986 #define DEV_CORE_STATS_INC(FIELD) \ 3987 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 3988 { \ 3989 netdev_core_stats_inc(dev, \ 3990 offsetof(struct net_device_core_stats, FIELD)); \ 3991 } 3992 DEV_CORE_STATS_INC(rx_dropped) 3993 DEV_CORE_STATS_INC(tx_dropped) 3994 DEV_CORE_STATS_INC(rx_nohandler) 3995 DEV_CORE_STATS_INC(rx_otherhost_dropped) 3996 #undef DEV_CORE_STATS_INC 3997 3998 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3999 struct sk_buff *skb, 4000 const bool check_mtu) 4001 { 4002 if (skb_orphan_frags(skb, GFP_ATOMIC) || 4003 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 4004 dev_core_stats_rx_dropped_inc(dev); 4005 kfree_skb(skb); 4006 return NET_RX_DROP; 4007 } 4008 4009 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 4010 skb->priority = 0; 4011 return 0; 4012 } 4013 4014 bool dev_nit_active(struct net_device *dev); 4015 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 4016 4017 static inline void __dev_put(struct net_device *dev) 4018 { 4019 if (dev) { 4020 #ifdef CONFIG_PCPU_DEV_REFCNT 4021 this_cpu_dec(*dev->pcpu_refcnt); 4022 #else 4023 refcount_dec(&dev->dev_refcnt); 4024 #endif 4025 } 4026 } 4027 4028 static inline void __dev_hold(struct net_device *dev) 4029 { 4030 if (dev) { 4031 #ifdef CONFIG_PCPU_DEV_REFCNT 4032 this_cpu_inc(*dev->pcpu_refcnt); 4033 #else 4034 refcount_inc(&dev->dev_refcnt); 4035 #endif 4036 } 4037 } 4038 4039 static inline void __netdev_tracker_alloc(struct net_device *dev, 4040 netdevice_tracker *tracker, 4041 gfp_t gfp) 4042 { 4043 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4044 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 4045 #endif 4046 } 4047 4048 /* netdev_tracker_alloc() can upgrade a prior untracked reference 4049 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4050 */ 4051 static inline void netdev_tracker_alloc(struct net_device *dev, 4052 netdevice_tracker *tracker, gfp_t gfp) 4053 { 4054 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4055 refcount_dec(&dev->refcnt_tracker.no_tracker); 4056 __netdev_tracker_alloc(dev, tracker, gfp); 4057 #endif 4058 } 4059 4060 static inline void netdev_tracker_free(struct net_device *dev, 4061 netdevice_tracker *tracker) 4062 { 4063 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4064 ref_tracker_free(&dev->refcnt_tracker, tracker); 4065 #endif 4066 } 4067 4068 static inline void netdev_hold(struct net_device *dev, 4069 netdevice_tracker *tracker, gfp_t gfp) 4070 { 4071 if (dev) { 4072 __dev_hold(dev); 4073 __netdev_tracker_alloc(dev, tracker, gfp); 4074 } 4075 } 4076 4077 static inline void netdev_put(struct net_device *dev, 4078 netdevice_tracker *tracker) 4079 { 4080 if (dev) { 4081 netdev_tracker_free(dev, tracker); 4082 __dev_put(dev); 4083 } 4084 } 4085 4086 /** 4087 * dev_hold - get reference to device 4088 * @dev: network device 4089 * 4090 * Hold reference to device to keep it from being freed. 4091 * Try using netdev_hold() instead. 4092 */ 4093 static inline void dev_hold(struct net_device *dev) 4094 { 4095 netdev_hold(dev, NULL, GFP_ATOMIC); 4096 } 4097 4098 /** 4099 * dev_put - release reference to device 4100 * @dev: network device 4101 * 4102 * Release reference to device to allow it to be freed. 4103 * Try using netdev_put() instead. 4104 */ 4105 static inline void dev_put(struct net_device *dev) 4106 { 4107 netdev_put(dev, NULL); 4108 } 4109 4110 DEFINE_FREE(dev_put, struct net_device *, if (_T) dev_put(_T)) 4111 4112 static inline void netdev_ref_replace(struct net_device *odev, 4113 struct net_device *ndev, 4114 netdevice_tracker *tracker, 4115 gfp_t gfp) 4116 { 4117 if (odev) 4118 netdev_tracker_free(odev, tracker); 4119 4120 __dev_hold(ndev); 4121 __dev_put(odev); 4122 4123 if (ndev) 4124 __netdev_tracker_alloc(ndev, tracker, gfp); 4125 } 4126 4127 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4128 * and _off may be called from IRQ context, but it is caller 4129 * who is responsible for serialization of these calls. 4130 * 4131 * The name carrier is inappropriate, these functions should really be 4132 * called netif_lowerlayer_*() because they represent the state of any 4133 * kind of lower layer not just hardware media. 4134 */ 4135 void linkwatch_fire_event(struct net_device *dev); 4136 4137 /** 4138 * linkwatch_sync_dev - sync linkwatch for the given device 4139 * @dev: network device to sync linkwatch for 4140 * 4141 * Sync linkwatch for the given device, removing it from the 4142 * pending work list (if queued). 4143 */ 4144 void linkwatch_sync_dev(struct net_device *dev); 4145 4146 /** 4147 * netif_carrier_ok - test if carrier present 4148 * @dev: network device 4149 * 4150 * Check if carrier is present on device 4151 */ 4152 static inline bool netif_carrier_ok(const struct net_device *dev) 4153 { 4154 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4155 } 4156 4157 unsigned long dev_trans_start(struct net_device *dev); 4158 4159 void __netdev_watchdog_up(struct net_device *dev); 4160 4161 void netif_carrier_on(struct net_device *dev); 4162 void netif_carrier_off(struct net_device *dev); 4163 void netif_carrier_event(struct net_device *dev); 4164 4165 /** 4166 * netif_dormant_on - mark device as dormant. 4167 * @dev: network device 4168 * 4169 * Mark device as dormant (as per RFC2863). 4170 * 4171 * The dormant state indicates that the relevant interface is not 4172 * actually in a condition to pass packets (i.e., it is not 'up') but is 4173 * in a "pending" state, waiting for some external event. For "on- 4174 * demand" interfaces, this new state identifies the situation where the 4175 * interface is waiting for events to place it in the up state. 4176 */ 4177 static inline void netif_dormant_on(struct net_device *dev) 4178 { 4179 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4180 linkwatch_fire_event(dev); 4181 } 4182 4183 /** 4184 * netif_dormant_off - set device as not dormant. 4185 * @dev: network device 4186 * 4187 * Device is not in dormant state. 4188 */ 4189 static inline void netif_dormant_off(struct net_device *dev) 4190 { 4191 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4192 linkwatch_fire_event(dev); 4193 } 4194 4195 /** 4196 * netif_dormant - test if device is dormant 4197 * @dev: network device 4198 * 4199 * Check if device is dormant. 4200 */ 4201 static inline bool netif_dormant(const struct net_device *dev) 4202 { 4203 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4204 } 4205 4206 4207 /** 4208 * netif_testing_on - mark device as under test. 4209 * @dev: network device 4210 * 4211 * Mark device as under test (as per RFC2863). 4212 * 4213 * The testing state indicates that some test(s) must be performed on 4214 * the interface. After completion, of the test, the interface state 4215 * will change to up, dormant, or down, as appropriate. 4216 */ 4217 static inline void netif_testing_on(struct net_device *dev) 4218 { 4219 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4220 linkwatch_fire_event(dev); 4221 } 4222 4223 /** 4224 * netif_testing_off - set device as not under test. 4225 * @dev: network device 4226 * 4227 * Device is not in testing state. 4228 */ 4229 static inline void netif_testing_off(struct net_device *dev) 4230 { 4231 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4232 linkwatch_fire_event(dev); 4233 } 4234 4235 /** 4236 * netif_testing - test if device is under test 4237 * @dev: network device 4238 * 4239 * Check if device is under test 4240 */ 4241 static inline bool netif_testing(const struct net_device *dev) 4242 { 4243 return test_bit(__LINK_STATE_TESTING, &dev->state); 4244 } 4245 4246 4247 /** 4248 * netif_oper_up - test if device is operational 4249 * @dev: network device 4250 * 4251 * Check if carrier is operational 4252 */ 4253 static inline bool netif_oper_up(const struct net_device *dev) 4254 { 4255 unsigned int operstate = READ_ONCE(dev->operstate); 4256 4257 return operstate == IF_OPER_UP || 4258 operstate == IF_OPER_UNKNOWN /* backward compat */; 4259 } 4260 4261 /** 4262 * netif_device_present - is device available or removed 4263 * @dev: network device 4264 * 4265 * Check if device has not been removed from system. 4266 */ 4267 static inline bool netif_device_present(const struct net_device *dev) 4268 { 4269 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4270 } 4271 4272 void netif_device_detach(struct net_device *dev); 4273 4274 void netif_device_attach(struct net_device *dev); 4275 4276 /* 4277 * Network interface message level settings 4278 */ 4279 4280 enum { 4281 NETIF_MSG_DRV_BIT, 4282 NETIF_MSG_PROBE_BIT, 4283 NETIF_MSG_LINK_BIT, 4284 NETIF_MSG_TIMER_BIT, 4285 NETIF_MSG_IFDOWN_BIT, 4286 NETIF_MSG_IFUP_BIT, 4287 NETIF_MSG_RX_ERR_BIT, 4288 NETIF_MSG_TX_ERR_BIT, 4289 NETIF_MSG_TX_QUEUED_BIT, 4290 NETIF_MSG_INTR_BIT, 4291 NETIF_MSG_TX_DONE_BIT, 4292 NETIF_MSG_RX_STATUS_BIT, 4293 NETIF_MSG_PKTDATA_BIT, 4294 NETIF_MSG_HW_BIT, 4295 NETIF_MSG_WOL_BIT, 4296 4297 /* When you add a new bit above, update netif_msg_class_names array 4298 * in net/ethtool/common.c 4299 */ 4300 NETIF_MSG_CLASS_COUNT, 4301 }; 4302 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4303 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4304 4305 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4306 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4307 4308 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4309 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4310 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4311 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4312 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4313 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4314 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4315 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4316 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4317 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4318 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4319 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4320 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4321 #define NETIF_MSG_HW __NETIF_MSG(HW) 4322 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4323 4324 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4325 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4326 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4327 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4328 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4329 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4330 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4331 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4332 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4333 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4334 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4335 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4336 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4337 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4338 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4339 4340 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4341 { 4342 /* use default */ 4343 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4344 return default_msg_enable_bits; 4345 if (debug_value == 0) /* no output */ 4346 return 0; 4347 /* set low N bits */ 4348 return (1U << debug_value) - 1; 4349 } 4350 4351 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4352 { 4353 spin_lock(&txq->_xmit_lock); 4354 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4355 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4356 } 4357 4358 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4359 { 4360 __acquire(&txq->_xmit_lock); 4361 return true; 4362 } 4363 4364 static inline void __netif_tx_release(struct netdev_queue *txq) 4365 { 4366 __release(&txq->_xmit_lock); 4367 } 4368 4369 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4370 { 4371 spin_lock_bh(&txq->_xmit_lock); 4372 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4373 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4374 } 4375 4376 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4377 { 4378 bool ok = spin_trylock(&txq->_xmit_lock); 4379 4380 if (likely(ok)) { 4381 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4382 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4383 } 4384 return ok; 4385 } 4386 4387 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4388 { 4389 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4390 WRITE_ONCE(txq->xmit_lock_owner, -1); 4391 spin_unlock(&txq->_xmit_lock); 4392 } 4393 4394 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4395 { 4396 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4397 WRITE_ONCE(txq->xmit_lock_owner, -1); 4398 spin_unlock_bh(&txq->_xmit_lock); 4399 } 4400 4401 /* 4402 * txq->trans_start can be read locklessly from dev_watchdog() 4403 */ 4404 static inline void txq_trans_update(struct netdev_queue *txq) 4405 { 4406 if (txq->xmit_lock_owner != -1) 4407 WRITE_ONCE(txq->trans_start, jiffies); 4408 } 4409 4410 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4411 { 4412 unsigned long now = jiffies; 4413 4414 if (READ_ONCE(txq->trans_start) != now) 4415 WRITE_ONCE(txq->trans_start, now); 4416 } 4417 4418 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4419 static inline void netif_trans_update(struct net_device *dev) 4420 { 4421 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4422 4423 txq_trans_cond_update(txq); 4424 } 4425 4426 /** 4427 * netif_tx_lock - grab network device transmit lock 4428 * @dev: network device 4429 * 4430 * Get network device transmit lock 4431 */ 4432 void netif_tx_lock(struct net_device *dev); 4433 4434 static inline void netif_tx_lock_bh(struct net_device *dev) 4435 { 4436 local_bh_disable(); 4437 netif_tx_lock(dev); 4438 } 4439 4440 void netif_tx_unlock(struct net_device *dev); 4441 4442 static inline void netif_tx_unlock_bh(struct net_device *dev) 4443 { 4444 netif_tx_unlock(dev); 4445 local_bh_enable(); 4446 } 4447 4448 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4449 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4450 __netif_tx_lock(txq, cpu); \ 4451 } else { \ 4452 __netif_tx_acquire(txq); \ 4453 } \ 4454 } 4455 4456 #define HARD_TX_TRYLOCK(dev, txq) \ 4457 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4458 __netif_tx_trylock(txq) : \ 4459 __netif_tx_acquire(txq)) 4460 4461 #define HARD_TX_UNLOCK(dev, txq) { \ 4462 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4463 __netif_tx_unlock(txq); \ 4464 } else { \ 4465 __netif_tx_release(txq); \ 4466 } \ 4467 } 4468 4469 static inline void netif_tx_disable(struct net_device *dev) 4470 { 4471 unsigned int i; 4472 int cpu; 4473 4474 local_bh_disable(); 4475 cpu = smp_processor_id(); 4476 spin_lock(&dev->tx_global_lock); 4477 for (i = 0; i < dev->num_tx_queues; i++) { 4478 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4479 4480 __netif_tx_lock(txq, cpu); 4481 netif_tx_stop_queue(txq); 4482 __netif_tx_unlock(txq); 4483 } 4484 spin_unlock(&dev->tx_global_lock); 4485 local_bh_enable(); 4486 } 4487 4488 static inline void netif_addr_lock(struct net_device *dev) 4489 { 4490 unsigned char nest_level = 0; 4491 4492 #ifdef CONFIG_LOCKDEP 4493 nest_level = dev->nested_level; 4494 #endif 4495 spin_lock_nested(&dev->addr_list_lock, nest_level); 4496 } 4497 4498 static inline void netif_addr_lock_bh(struct net_device *dev) 4499 { 4500 unsigned char nest_level = 0; 4501 4502 #ifdef CONFIG_LOCKDEP 4503 nest_level = dev->nested_level; 4504 #endif 4505 local_bh_disable(); 4506 spin_lock_nested(&dev->addr_list_lock, nest_level); 4507 } 4508 4509 static inline void netif_addr_unlock(struct net_device *dev) 4510 { 4511 spin_unlock(&dev->addr_list_lock); 4512 } 4513 4514 static inline void netif_addr_unlock_bh(struct net_device *dev) 4515 { 4516 spin_unlock_bh(&dev->addr_list_lock); 4517 } 4518 4519 /* 4520 * dev_addrs walker. Should be used only for read access. Call with 4521 * rcu_read_lock held. 4522 */ 4523 #define for_each_dev_addr(dev, ha) \ 4524 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4525 4526 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4527 4528 void ether_setup(struct net_device *dev); 4529 4530 /* Allocate dummy net_device */ 4531 struct net_device *alloc_netdev_dummy(int sizeof_priv); 4532 4533 /* Support for loadable net-drivers */ 4534 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4535 unsigned char name_assign_type, 4536 void (*setup)(struct net_device *), 4537 unsigned int txqs, unsigned int rxqs); 4538 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4539 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4540 4541 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4542 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4543 count) 4544 4545 int register_netdev(struct net_device *dev); 4546 void unregister_netdev(struct net_device *dev); 4547 4548 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4549 4550 /* General hardware address lists handling functions */ 4551 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4552 struct netdev_hw_addr_list *from_list, int addr_len); 4553 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4554 struct netdev_hw_addr_list *from_list, int addr_len); 4555 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4556 struct net_device *dev, 4557 int (*sync)(struct net_device *, const unsigned char *), 4558 int (*unsync)(struct net_device *, 4559 const unsigned char *)); 4560 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4561 struct net_device *dev, 4562 int (*sync)(struct net_device *, 4563 const unsigned char *, int), 4564 int (*unsync)(struct net_device *, 4565 const unsigned char *, int)); 4566 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4567 struct net_device *dev, 4568 int (*unsync)(struct net_device *, 4569 const unsigned char *, int)); 4570 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4571 struct net_device *dev, 4572 int (*unsync)(struct net_device *, 4573 const unsigned char *)); 4574 void __hw_addr_init(struct netdev_hw_addr_list *list); 4575 4576 /* Functions used for device addresses handling */ 4577 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4578 const void *addr, size_t len); 4579 4580 static inline void 4581 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4582 { 4583 dev_addr_mod(dev, 0, addr, len); 4584 } 4585 4586 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4587 { 4588 __dev_addr_set(dev, addr, dev->addr_len); 4589 } 4590 4591 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4592 unsigned char addr_type); 4593 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4594 unsigned char addr_type); 4595 4596 /* Functions used for unicast addresses handling */ 4597 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4598 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4599 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4600 int dev_uc_sync(struct net_device *to, struct net_device *from); 4601 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4602 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4603 void dev_uc_flush(struct net_device *dev); 4604 void dev_uc_init(struct net_device *dev); 4605 4606 /** 4607 * __dev_uc_sync - Synchonize device's unicast list 4608 * @dev: device to sync 4609 * @sync: function to call if address should be added 4610 * @unsync: function to call if address should be removed 4611 * 4612 * Add newly added addresses to the interface, and release 4613 * addresses that have been deleted. 4614 */ 4615 static inline int __dev_uc_sync(struct net_device *dev, 4616 int (*sync)(struct net_device *, 4617 const unsigned char *), 4618 int (*unsync)(struct net_device *, 4619 const unsigned char *)) 4620 { 4621 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4622 } 4623 4624 /** 4625 * __dev_uc_unsync - Remove synchronized addresses from device 4626 * @dev: device to sync 4627 * @unsync: function to call if address should be removed 4628 * 4629 * Remove all addresses that were added to the device by dev_uc_sync(). 4630 */ 4631 static inline void __dev_uc_unsync(struct net_device *dev, 4632 int (*unsync)(struct net_device *, 4633 const unsigned char *)) 4634 { 4635 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4636 } 4637 4638 /* Functions used for multicast addresses handling */ 4639 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4640 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4641 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4642 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4643 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4644 int dev_mc_sync(struct net_device *to, struct net_device *from); 4645 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4646 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4647 void dev_mc_flush(struct net_device *dev); 4648 void dev_mc_init(struct net_device *dev); 4649 4650 /** 4651 * __dev_mc_sync - Synchonize device's multicast list 4652 * @dev: device to sync 4653 * @sync: function to call if address should be added 4654 * @unsync: function to call if address should be removed 4655 * 4656 * Add newly added addresses to the interface, and release 4657 * addresses that have been deleted. 4658 */ 4659 static inline int __dev_mc_sync(struct net_device *dev, 4660 int (*sync)(struct net_device *, 4661 const unsigned char *), 4662 int (*unsync)(struct net_device *, 4663 const unsigned char *)) 4664 { 4665 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4666 } 4667 4668 /** 4669 * __dev_mc_unsync - Remove synchronized addresses from device 4670 * @dev: device to sync 4671 * @unsync: function to call if address should be removed 4672 * 4673 * Remove all addresses that were added to the device by dev_mc_sync(). 4674 */ 4675 static inline void __dev_mc_unsync(struct net_device *dev, 4676 int (*unsync)(struct net_device *, 4677 const unsigned char *)) 4678 { 4679 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4680 } 4681 4682 /* Functions used for secondary unicast and multicast support */ 4683 void dev_set_rx_mode(struct net_device *dev); 4684 int dev_set_promiscuity(struct net_device *dev, int inc); 4685 int dev_set_allmulti(struct net_device *dev, int inc); 4686 void netdev_state_change(struct net_device *dev); 4687 void __netdev_notify_peers(struct net_device *dev); 4688 void netdev_notify_peers(struct net_device *dev); 4689 void netdev_features_change(struct net_device *dev); 4690 /* Load a device via the kmod */ 4691 void dev_load(struct net *net, const char *name); 4692 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4693 struct rtnl_link_stats64 *storage); 4694 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4695 const struct net_device_stats *netdev_stats); 4696 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4697 const struct pcpu_sw_netstats __percpu *netstats); 4698 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4699 4700 enum { 4701 NESTED_SYNC_IMM_BIT, 4702 NESTED_SYNC_TODO_BIT, 4703 }; 4704 4705 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4706 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4707 4708 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4709 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4710 4711 struct netdev_nested_priv { 4712 unsigned char flags; 4713 void *data; 4714 }; 4715 4716 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4717 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4718 struct list_head **iter); 4719 4720 /* iterate through upper list, must be called under RCU read lock */ 4721 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4722 for (iter = &(dev)->adj_list.upper, \ 4723 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4724 updev; \ 4725 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4726 4727 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4728 int (*fn)(struct net_device *upper_dev, 4729 struct netdev_nested_priv *priv), 4730 struct netdev_nested_priv *priv); 4731 4732 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4733 struct net_device *upper_dev); 4734 4735 bool netdev_has_any_upper_dev(struct net_device *dev); 4736 4737 void *netdev_lower_get_next_private(struct net_device *dev, 4738 struct list_head **iter); 4739 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4740 struct list_head **iter); 4741 4742 #define netdev_for_each_lower_private(dev, priv, iter) \ 4743 for (iter = (dev)->adj_list.lower.next, \ 4744 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4745 priv; \ 4746 priv = netdev_lower_get_next_private(dev, &(iter))) 4747 4748 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4749 for (iter = &(dev)->adj_list.lower, \ 4750 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4751 priv; \ 4752 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4753 4754 void *netdev_lower_get_next(struct net_device *dev, 4755 struct list_head **iter); 4756 4757 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4758 for (iter = (dev)->adj_list.lower.next, \ 4759 ldev = netdev_lower_get_next(dev, &(iter)); \ 4760 ldev; \ 4761 ldev = netdev_lower_get_next(dev, &(iter))) 4762 4763 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4764 struct list_head **iter); 4765 int netdev_walk_all_lower_dev(struct net_device *dev, 4766 int (*fn)(struct net_device *lower_dev, 4767 struct netdev_nested_priv *priv), 4768 struct netdev_nested_priv *priv); 4769 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4770 int (*fn)(struct net_device *lower_dev, 4771 struct netdev_nested_priv *priv), 4772 struct netdev_nested_priv *priv); 4773 4774 void *netdev_adjacent_get_private(struct list_head *adj_list); 4775 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4776 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4777 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4778 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4779 struct netlink_ext_ack *extack); 4780 int netdev_master_upper_dev_link(struct net_device *dev, 4781 struct net_device *upper_dev, 4782 void *upper_priv, void *upper_info, 4783 struct netlink_ext_ack *extack); 4784 void netdev_upper_dev_unlink(struct net_device *dev, 4785 struct net_device *upper_dev); 4786 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4787 struct net_device *new_dev, 4788 struct net_device *dev, 4789 struct netlink_ext_ack *extack); 4790 void netdev_adjacent_change_commit(struct net_device *old_dev, 4791 struct net_device *new_dev, 4792 struct net_device *dev); 4793 void netdev_adjacent_change_abort(struct net_device *old_dev, 4794 struct net_device *new_dev, 4795 struct net_device *dev); 4796 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4797 void *netdev_lower_dev_get_private(struct net_device *dev, 4798 struct net_device *lower_dev); 4799 void netdev_lower_state_changed(struct net_device *lower_dev, 4800 void *lower_state_info); 4801 4802 /* RSS keys are 40 or 52 bytes long */ 4803 #define NETDEV_RSS_KEY_LEN 52 4804 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4805 void netdev_rss_key_fill(void *buffer, size_t len); 4806 4807 int skb_checksum_help(struct sk_buff *skb); 4808 int skb_crc32c_csum_help(struct sk_buff *skb); 4809 int skb_csum_hwoffload_help(struct sk_buff *skb, 4810 const netdev_features_t features); 4811 4812 struct netdev_bonding_info { 4813 ifslave slave; 4814 ifbond master; 4815 }; 4816 4817 struct netdev_notifier_bonding_info { 4818 struct netdev_notifier_info info; /* must be first */ 4819 struct netdev_bonding_info bonding_info; 4820 }; 4821 4822 void netdev_bonding_info_change(struct net_device *dev, 4823 struct netdev_bonding_info *bonding_info); 4824 4825 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4826 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4827 #else 4828 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4829 const void *data) 4830 { 4831 } 4832 #endif 4833 4834 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4835 4836 static inline bool can_checksum_protocol(netdev_features_t features, 4837 __be16 protocol) 4838 { 4839 if (protocol == htons(ETH_P_FCOE)) 4840 return !!(features & NETIF_F_FCOE_CRC); 4841 4842 /* Assume this is an IP checksum (not SCTP CRC) */ 4843 4844 if (features & NETIF_F_HW_CSUM) { 4845 /* Can checksum everything */ 4846 return true; 4847 } 4848 4849 switch (protocol) { 4850 case htons(ETH_P_IP): 4851 return !!(features & NETIF_F_IP_CSUM); 4852 case htons(ETH_P_IPV6): 4853 return !!(features & NETIF_F_IPV6_CSUM); 4854 default: 4855 return false; 4856 } 4857 } 4858 4859 #ifdef CONFIG_BUG 4860 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4861 #else 4862 static inline void netdev_rx_csum_fault(struct net_device *dev, 4863 struct sk_buff *skb) 4864 { 4865 } 4866 #endif 4867 /* rx skb timestamps */ 4868 void net_enable_timestamp(void); 4869 void net_disable_timestamp(void); 4870 4871 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 4872 const struct skb_shared_hwtstamps *hwtstamps, 4873 bool cycles) 4874 { 4875 const struct net_device_ops *ops = dev->netdev_ops; 4876 4877 if (ops->ndo_get_tstamp) 4878 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 4879 4880 return hwtstamps->hwtstamp; 4881 } 4882 4883 #ifndef CONFIG_PREEMPT_RT 4884 static inline void netdev_xmit_set_more(bool more) 4885 { 4886 __this_cpu_write(softnet_data.xmit.more, more); 4887 } 4888 4889 static inline bool netdev_xmit_more(void) 4890 { 4891 return __this_cpu_read(softnet_data.xmit.more); 4892 } 4893 #else 4894 static inline void netdev_xmit_set_more(bool more) 4895 { 4896 current->net_xmit.more = more; 4897 } 4898 4899 static inline bool netdev_xmit_more(void) 4900 { 4901 return current->net_xmit.more; 4902 } 4903 #endif 4904 4905 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4906 struct sk_buff *skb, struct net_device *dev, 4907 bool more) 4908 { 4909 netdev_xmit_set_more(more); 4910 return ops->ndo_start_xmit(skb, dev); 4911 } 4912 4913 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4914 struct netdev_queue *txq, bool more) 4915 { 4916 const struct net_device_ops *ops = dev->netdev_ops; 4917 netdev_tx_t rc; 4918 4919 rc = __netdev_start_xmit(ops, skb, dev, more); 4920 if (rc == NETDEV_TX_OK) 4921 txq_trans_update(txq); 4922 4923 return rc; 4924 } 4925 4926 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4927 const void *ns); 4928 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4929 const void *ns); 4930 4931 extern const struct kobj_ns_type_operations net_ns_type_operations; 4932 4933 const char *netdev_drivername(const struct net_device *dev); 4934 4935 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4936 netdev_features_t f2) 4937 { 4938 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4939 if (f1 & NETIF_F_HW_CSUM) 4940 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4941 else 4942 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4943 } 4944 4945 return f1 & f2; 4946 } 4947 4948 static inline netdev_features_t netdev_get_wanted_features( 4949 struct net_device *dev) 4950 { 4951 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4952 } 4953 netdev_features_t netdev_increment_features(netdev_features_t all, 4954 netdev_features_t one, netdev_features_t mask); 4955 4956 /* Allow TSO being used on stacked device : 4957 * Performing the GSO segmentation before last device 4958 * is a performance improvement. 4959 */ 4960 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4961 netdev_features_t mask) 4962 { 4963 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4964 } 4965 4966 int __netdev_update_features(struct net_device *dev); 4967 void netdev_update_features(struct net_device *dev); 4968 void netdev_change_features(struct net_device *dev); 4969 4970 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4971 struct net_device *dev); 4972 4973 netdev_features_t passthru_features_check(struct sk_buff *skb, 4974 struct net_device *dev, 4975 netdev_features_t features); 4976 netdev_features_t netif_skb_features(struct sk_buff *skb); 4977 void skb_warn_bad_offload(const struct sk_buff *skb); 4978 4979 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4980 { 4981 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4982 4983 /* check flags correspondence */ 4984 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4985 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4986 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4987 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4988 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4989 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4990 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4991 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4992 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4993 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4994 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4995 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4996 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4997 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4998 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4999 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 5000 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 5001 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 5002 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 5003 5004 return (features & feature) == feature; 5005 } 5006 5007 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 5008 { 5009 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 5010 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 5011 } 5012 5013 static inline bool netif_needs_gso(struct sk_buff *skb, 5014 netdev_features_t features) 5015 { 5016 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 5017 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 5018 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 5019 } 5020 5021 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 5022 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 5023 void netif_inherit_tso_max(struct net_device *to, 5024 const struct net_device *from); 5025 5026 static inline bool netif_is_macsec(const struct net_device *dev) 5027 { 5028 return dev->priv_flags & IFF_MACSEC; 5029 } 5030 5031 static inline bool netif_is_macvlan(const struct net_device *dev) 5032 { 5033 return dev->priv_flags & IFF_MACVLAN; 5034 } 5035 5036 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5037 { 5038 return dev->priv_flags & IFF_MACVLAN_PORT; 5039 } 5040 5041 static inline bool netif_is_bond_master(const struct net_device *dev) 5042 { 5043 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5044 } 5045 5046 static inline bool netif_is_bond_slave(const struct net_device *dev) 5047 { 5048 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5049 } 5050 5051 static inline bool netif_supports_nofcs(struct net_device *dev) 5052 { 5053 return dev->priv_flags & IFF_SUPP_NOFCS; 5054 } 5055 5056 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5057 { 5058 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5059 } 5060 5061 static inline bool netif_is_l3_master(const struct net_device *dev) 5062 { 5063 return dev->priv_flags & IFF_L3MDEV_MASTER; 5064 } 5065 5066 static inline bool netif_is_l3_slave(const struct net_device *dev) 5067 { 5068 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5069 } 5070 5071 static inline int dev_sdif(const struct net_device *dev) 5072 { 5073 #ifdef CONFIG_NET_L3_MASTER_DEV 5074 if (netif_is_l3_slave(dev)) 5075 return dev->ifindex; 5076 #endif 5077 return 0; 5078 } 5079 5080 static inline bool netif_is_bridge_master(const struct net_device *dev) 5081 { 5082 return dev->priv_flags & IFF_EBRIDGE; 5083 } 5084 5085 static inline bool netif_is_bridge_port(const struct net_device *dev) 5086 { 5087 return dev->priv_flags & IFF_BRIDGE_PORT; 5088 } 5089 5090 static inline bool netif_is_ovs_master(const struct net_device *dev) 5091 { 5092 return dev->priv_flags & IFF_OPENVSWITCH; 5093 } 5094 5095 static inline bool netif_is_ovs_port(const struct net_device *dev) 5096 { 5097 return dev->priv_flags & IFF_OVS_DATAPATH; 5098 } 5099 5100 static inline bool netif_is_any_bridge_master(const struct net_device *dev) 5101 { 5102 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev); 5103 } 5104 5105 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5106 { 5107 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5108 } 5109 5110 static inline bool netif_is_team_master(const struct net_device *dev) 5111 { 5112 return dev->priv_flags & IFF_TEAM; 5113 } 5114 5115 static inline bool netif_is_team_port(const struct net_device *dev) 5116 { 5117 return dev->priv_flags & IFF_TEAM_PORT; 5118 } 5119 5120 static inline bool netif_is_lag_master(const struct net_device *dev) 5121 { 5122 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5123 } 5124 5125 static inline bool netif_is_lag_port(const struct net_device *dev) 5126 { 5127 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5128 } 5129 5130 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5131 { 5132 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5133 } 5134 5135 static inline bool netif_is_failover(const struct net_device *dev) 5136 { 5137 return dev->priv_flags & IFF_FAILOVER; 5138 } 5139 5140 static inline bool netif_is_failover_slave(const struct net_device *dev) 5141 { 5142 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5143 } 5144 5145 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5146 static inline void netif_keep_dst(struct net_device *dev) 5147 { 5148 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5149 } 5150 5151 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5152 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5153 { 5154 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5155 return netif_is_macsec(dev); 5156 } 5157 5158 extern struct pernet_operations __net_initdata loopback_net_ops; 5159 5160 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5161 5162 /* netdev_printk helpers, similar to dev_printk */ 5163 5164 static inline const char *netdev_name(const struct net_device *dev) 5165 { 5166 if (!dev->name[0] || strchr(dev->name, '%')) 5167 return "(unnamed net_device)"; 5168 return dev->name; 5169 } 5170 5171 static inline const char *netdev_reg_state(const struct net_device *dev) 5172 { 5173 u8 reg_state = READ_ONCE(dev->reg_state); 5174 5175 switch (reg_state) { 5176 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5177 case NETREG_REGISTERED: return ""; 5178 case NETREG_UNREGISTERING: return " (unregistering)"; 5179 case NETREG_UNREGISTERED: return " (unregistered)"; 5180 case NETREG_RELEASED: return " (released)"; 5181 case NETREG_DUMMY: return " (dummy)"; 5182 } 5183 5184 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state); 5185 return " (unknown)"; 5186 } 5187 5188 #define MODULE_ALIAS_NETDEV(device) \ 5189 MODULE_ALIAS("netdev-" device) 5190 5191 /* 5192 * netdev_WARN() acts like dev_printk(), but with the key difference 5193 * of using a WARN/WARN_ON to get the message out, including the 5194 * file/line information and a backtrace. 5195 */ 5196 #define netdev_WARN(dev, format, args...) \ 5197 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5198 netdev_reg_state(dev), ##args) 5199 5200 #define netdev_WARN_ONCE(dev, format, args...) \ 5201 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5202 netdev_reg_state(dev), ##args) 5203 5204 /* 5205 * The list of packet types we will receive (as opposed to discard) 5206 * and the routines to invoke. 5207 * 5208 * Why 16. Because with 16 the only overlap we get on a hash of the 5209 * low nibble of the protocol value is RARP/SNAP/X.25. 5210 * 5211 * 0800 IP 5212 * 0001 802.3 5213 * 0002 AX.25 5214 * 0004 802.2 5215 * 8035 RARP 5216 * 0005 SNAP 5217 * 0805 X.25 5218 * 0806 ARP 5219 * 8137 IPX 5220 * 0009 Localtalk 5221 * 86DD IPv6 5222 */ 5223 #define PTYPE_HASH_SIZE (16) 5224 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5225 5226 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5227 5228 extern struct net_device *blackhole_netdev; 5229 5230 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */ 5231 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD) 5232 #define DEV_STATS_ADD(DEV, FIELD, VAL) \ 5233 atomic_long_add((VAL), &(DEV)->stats.__##FIELD) 5234 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD) 5235 5236 #endif /* _LINUX_NETDEVICE_H */ 5237