1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <[email protected]> 13 * Corey Minyard <[email protected]> 14 * Donald J. Becker, <[email protected]> 15 * Alan Cox, <[email protected]> 16 * Bjorn Ekwall. <[email protected]> 17 * Pekka Riikonen <[email protected]> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 #include <asm/local.h> 32 33 #include <linux/percpu.h> 34 #include <linux/rculist.h> 35 #include <linux/workqueue.h> 36 #include <linux/dynamic_queue_limits.h> 37 38 #include <net/net_namespace.h> 39 #ifdef CONFIG_DCB 40 #include <net/dcbnl.h> 41 #endif 42 #include <net/netprio_cgroup.h> 43 #include <linux/netdev_features.h> 44 #include <linux/neighbour.h> 45 #include <linux/netdevice_xmit.h> 46 #include <uapi/linux/netdevice.h> 47 #include <uapi/linux/if_bonding.h> 48 #include <uapi/linux/pkt_cls.h> 49 #include <uapi/linux/netdev.h> 50 #include <linux/hashtable.h> 51 #include <linux/rbtree.h> 52 #include <net/net_trackers.h> 53 #include <net/net_debug.h> 54 #include <net/dropreason-core.h> 55 #include <net/neighbour_tables.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 struct phy_link_topology; 85 86 typedef u32 xdp_features_t; 87 88 void synchronize_net(void); 89 void netdev_set_default_ethtool_ops(struct net_device *dev, 90 const struct ethtool_ops *ops); 91 void netdev_sw_irq_coalesce_default_on(struct net_device *dev); 92 93 /* Backlog congestion levels */ 94 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 95 #define NET_RX_DROP 1 /* packet dropped */ 96 97 #define MAX_NEST_DEV 8 98 99 /* 100 * Transmit return codes: transmit return codes originate from three different 101 * namespaces: 102 * 103 * - qdisc return codes 104 * - driver transmit return codes 105 * - errno values 106 * 107 * Drivers are allowed to return any one of those in their hard_start_xmit() 108 * function. Real network devices commonly used with qdiscs should only return 109 * the driver transmit return codes though - when qdiscs are used, the actual 110 * transmission happens asynchronously, so the value is not propagated to 111 * higher layers. Virtual network devices transmit synchronously; in this case 112 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 113 * others are propagated to higher layers. 114 */ 115 116 /* qdisc ->enqueue() return codes. */ 117 #define NET_XMIT_SUCCESS 0x00 118 #define NET_XMIT_DROP 0x01 /* skb dropped */ 119 #define NET_XMIT_CN 0x02 /* congestion notification */ 120 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 121 122 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 123 * indicates that the device will soon be dropping packets, or already drops 124 * some packets of the same priority; prompting us to send less aggressively. */ 125 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 126 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 127 128 /* Driver transmit return codes */ 129 #define NETDEV_TX_MASK 0xf0 130 131 enum netdev_tx { 132 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 133 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 134 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 135 }; 136 typedef enum netdev_tx netdev_tx_t; 137 138 /* 139 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 140 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 141 */ 142 static inline bool dev_xmit_complete(int rc) 143 { 144 /* 145 * Positive cases with an skb consumed by a driver: 146 * - successful transmission (rc == NETDEV_TX_OK) 147 * - error while transmitting (rc < 0) 148 * - error while queueing to a different device (rc & NET_XMIT_MASK) 149 */ 150 if (likely(rc < NET_XMIT_MASK)) 151 return true; 152 153 return false; 154 } 155 156 /* 157 * Compute the worst-case header length according to the protocols 158 * used. 159 */ 160 161 #if defined(CONFIG_HYPERV_NET) 162 # define LL_MAX_HEADER 128 163 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 164 # if defined(CONFIG_MAC80211_MESH) 165 # define LL_MAX_HEADER 128 166 # else 167 # define LL_MAX_HEADER 96 168 # endif 169 #else 170 # define LL_MAX_HEADER 32 171 #endif 172 173 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 174 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 175 #define MAX_HEADER LL_MAX_HEADER 176 #else 177 #define MAX_HEADER (LL_MAX_HEADER + 48) 178 #endif 179 180 /* 181 * Old network device statistics. Fields are native words 182 * (unsigned long) so they can be read and written atomically. 183 */ 184 185 #define NET_DEV_STAT(FIELD) \ 186 union { \ 187 unsigned long FIELD; \ 188 atomic_long_t __##FIELD; \ 189 } 190 191 struct net_device_stats { 192 NET_DEV_STAT(rx_packets); 193 NET_DEV_STAT(tx_packets); 194 NET_DEV_STAT(rx_bytes); 195 NET_DEV_STAT(tx_bytes); 196 NET_DEV_STAT(rx_errors); 197 NET_DEV_STAT(tx_errors); 198 NET_DEV_STAT(rx_dropped); 199 NET_DEV_STAT(tx_dropped); 200 NET_DEV_STAT(multicast); 201 NET_DEV_STAT(collisions); 202 NET_DEV_STAT(rx_length_errors); 203 NET_DEV_STAT(rx_over_errors); 204 NET_DEV_STAT(rx_crc_errors); 205 NET_DEV_STAT(rx_frame_errors); 206 NET_DEV_STAT(rx_fifo_errors); 207 NET_DEV_STAT(rx_missed_errors); 208 NET_DEV_STAT(tx_aborted_errors); 209 NET_DEV_STAT(tx_carrier_errors); 210 NET_DEV_STAT(tx_fifo_errors); 211 NET_DEV_STAT(tx_heartbeat_errors); 212 NET_DEV_STAT(tx_window_errors); 213 NET_DEV_STAT(rx_compressed); 214 NET_DEV_STAT(tx_compressed); 215 }; 216 #undef NET_DEV_STAT 217 218 /* per-cpu stats, allocated on demand. 219 * Try to fit them in a single cache line, for dev_get_stats() sake. 220 */ 221 struct net_device_core_stats { 222 unsigned long rx_dropped; 223 unsigned long tx_dropped; 224 unsigned long rx_nohandler; 225 unsigned long rx_otherhost_dropped; 226 } __aligned(4 * sizeof(unsigned long)); 227 228 #include <linux/cache.h> 229 #include <linux/skbuff.h> 230 231 struct neighbour; 232 struct neigh_parms; 233 struct sk_buff; 234 235 struct netdev_hw_addr { 236 struct list_head list; 237 struct rb_node node; 238 unsigned char addr[MAX_ADDR_LEN]; 239 unsigned char type; 240 #define NETDEV_HW_ADDR_T_LAN 1 241 #define NETDEV_HW_ADDR_T_SAN 2 242 #define NETDEV_HW_ADDR_T_UNICAST 3 243 #define NETDEV_HW_ADDR_T_MULTICAST 4 244 bool global_use; 245 int sync_cnt; 246 int refcount; 247 int synced; 248 struct rcu_head rcu_head; 249 }; 250 251 struct netdev_hw_addr_list { 252 struct list_head list; 253 int count; 254 255 /* Auxiliary tree for faster lookup on addition and deletion */ 256 struct rb_root tree; 257 }; 258 259 #define netdev_hw_addr_list_count(l) ((l)->count) 260 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 261 #define netdev_hw_addr_list_for_each(ha, l) \ 262 list_for_each_entry(ha, &(l)->list, list) 263 264 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 265 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 266 #define netdev_for_each_uc_addr(ha, dev) \ 267 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 268 #define netdev_for_each_synced_uc_addr(_ha, _dev) \ 269 netdev_for_each_uc_addr((_ha), (_dev)) \ 270 if ((_ha)->sync_cnt) 271 272 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 273 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 274 #define netdev_for_each_mc_addr(ha, dev) \ 275 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 276 #define netdev_for_each_synced_mc_addr(_ha, _dev) \ 277 netdev_for_each_mc_addr((_ha), (_dev)) \ 278 if ((_ha)->sync_cnt) 279 280 struct hh_cache { 281 unsigned int hh_len; 282 seqlock_t hh_lock; 283 284 /* cached hardware header; allow for machine alignment needs. */ 285 #define HH_DATA_MOD 16 286 #define HH_DATA_OFF(__len) \ 287 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 288 #define HH_DATA_ALIGN(__len) \ 289 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 290 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 291 }; 292 293 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 294 * Alternative is: 295 * dev->hard_header_len ? (dev->hard_header_len + 296 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 297 * 298 * We could use other alignment values, but we must maintain the 299 * relationship HH alignment <= LL alignment. 300 */ 301 #define LL_RESERVED_SPACE(dev) \ 302 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \ 303 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 304 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 305 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \ 306 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 307 308 struct header_ops { 309 int (*create) (struct sk_buff *skb, struct net_device *dev, 310 unsigned short type, const void *daddr, 311 const void *saddr, unsigned int len); 312 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 313 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 314 void (*cache_update)(struct hh_cache *hh, 315 const struct net_device *dev, 316 const unsigned char *haddr); 317 bool (*validate)(const char *ll_header, unsigned int len); 318 __be16 (*parse_protocol)(const struct sk_buff *skb); 319 }; 320 321 /* These flag bits are private to the generic network queueing 322 * layer; they may not be explicitly referenced by any other 323 * code. 324 */ 325 326 enum netdev_state_t { 327 __LINK_STATE_START, 328 __LINK_STATE_PRESENT, 329 __LINK_STATE_NOCARRIER, 330 __LINK_STATE_LINKWATCH_PENDING, 331 __LINK_STATE_DORMANT, 332 __LINK_STATE_TESTING, 333 }; 334 335 struct gro_list { 336 struct list_head list; 337 int count; 338 }; 339 340 /* 341 * size of gro hash buckets, must less than bit number of 342 * napi_struct::gro_bitmask 343 */ 344 #define GRO_HASH_BUCKETS 8 345 346 /* 347 * Structure for per-NAPI config 348 */ 349 struct napi_config { 350 u64 gro_flush_timeout; 351 u32 defer_hard_irqs; 352 unsigned int napi_id; 353 }; 354 355 /* 356 * Structure for NAPI scheduling similar to tasklet but with weighting 357 */ 358 struct napi_struct { 359 /* The poll_list must only be managed by the entity which 360 * changes the state of the NAPI_STATE_SCHED bit. This means 361 * whoever atomically sets that bit can add this napi_struct 362 * to the per-CPU poll_list, and whoever clears that bit 363 * can remove from the list right before clearing the bit. 364 */ 365 struct list_head poll_list; 366 367 unsigned long state; 368 int weight; 369 u32 defer_hard_irqs_count; 370 unsigned long gro_bitmask; 371 int (*poll)(struct napi_struct *, int); 372 #ifdef CONFIG_NETPOLL 373 /* CPU actively polling if netpoll is configured */ 374 int poll_owner; 375 #endif 376 /* CPU on which NAPI has been scheduled for processing */ 377 int list_owner; 378 struct net_device *dev; 379 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 380 struct sk_buff *skb; 381 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 382 int rx_count; /* length of rx_list */ 383 unsigned int napi_id; 384 struct hrtimer timer; 385 struct task_struct *thread; 386 unsigned long gro_flush_timeout; 387 u32 defer_hard_irqs; 388 /* control-path-only fields follow */ 389 struct list_head dev_list; 390 struct hlist_node napi_hash_node; 391 int irq; 392 int index; 393 struct napi_config *config; 394 }; 395 396 enum { 397 NAPI_STATE_SCHED, /* Poll is scheduled */ 398 NAPI_STATE_MISSED, /* reschedule a napi */ 399 NAPI_STATE_DISABLE, /* Disable pending */ 400 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 401 NAPI_STATE_LISTED, /* NAPI added to system lists */ 402 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 403 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 404 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 405 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 406 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 407 }; 408 409 enum { 410 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 411 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 412 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 413 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 414 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 415 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 416 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 417 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 418 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 419 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 420 }; 421 422 enum gro_result { 423 GRO_MERGED, 424 GRO_MERGED_FREE, 425 GRO_HELD, 426 GRO_NORMAL, 427 GRO_CONSUMED, 428 }; 429 typedef enum gro_result gro_result_t; 430 431 /* 432 * enum rx_handler_result - Possible return values for rx_handlers. 433 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 434 * further. 435 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 436 * case skb->dev was changed by rx_handler. 437 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 438 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 439 * 440 * rx_handlers are functions called from inside __netif_receive_skb(), to do 441 * special processing of the skb, prior to delivery to protocol handlers. 442 * 443 * Currently, a net_device can only have a single rx_handler registered. Trying 444 * to register a second rx_handler will return -EBUSY. 445 * 446 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 447 * To unregister a rx_handler on a net_device, use 448 * netdev_rx_handler_unregister(). 449 * 450 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 451 * do with the skb. 452 * 453 * If the rx_handler consumed the skb in some way, it should return 454 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 455 * the skb to be delivered in some other way. 456 * 457 * If the rx_handler changed skb->dev, to divert the skb to another 458 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 459 * new device will be called if it exists. 460 * 461 * If the rx_handler decides the skb should be ignored, it should return 462 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 463 * are registered on exact device (ptype->dev == skb->dev). 464 * 465 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 466 * delivered, it should return RX_HANDLER_PASS. 467 * 468 * A device without a registered rx_handler will behave as if rx_handler 469 * returned RX_HANDLER_PASS. 470 */ 471 472 enum rx_handler_result { 473 RX_HANDLER_CONSUMED, 474 RX_HANDLER_ANOTHER, 475 RX_HANDLER_EXACT, 476 RX_HANDLER_PASS, 477 }; 478 typedef enum rx_handler_result rx_handler_result_t; 479 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 480 481 void __napi_schedule(struct napi_struct *n); 482 void __napi_schedule_irqoff(struct napi_struct *n); 483 484 static inline bool napi_disable_pending(struct napi_struct *n) 485 { 486 return test_bit(NAPI_STATE_DISABLE, &n->state); 487 } 488 489 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 490 { 491 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 492 } 493 494 /** 495 * napi_is_scheduled - test if NAPI is scheduled 496 * @n: NAPI context 497 * 498 * This check is "best-effort". With no locking implemented, 499 * a NAPI can be scheduled or terminate right after this check 500 * and produce not precise results. 501 * 502 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled 503 * should not be used normally and napi_schedule should be 504 * used instead. 505 * 506 * Use only if the driver really needs to check if a NAPI 507 * is scheduled for example in the context of delayed timer 508 * that can be skipped if a NAPI is already scheduled. 509 * 510 * Return True if NAPI is scheduled, False otherwise. 511 */ 512 static inline bool napi_is_scheduled(struct napi_struct *n) 513 { 514 return test_bit(NAPI_STATE_SCHED, &n->state); 515 } 516 517 bool napi_schedule_prep(struct napi_struct *n); 518 519 /** 520 * napi_schedule - schedule NAPI poll 521 * @n: NAPI context 522 * 523 * Schedule NAPI poll routine to be called if it is not already 524 * running. 525 * Return true if we schedule a NAPI or false if not. 526 * Refer to napi_schedule_prep() for additional reason on why 527 * a NAPI might not be scheduled. 528 */ 529 static inline bool napi_schedule(struct napi_struct *n) 530 { 531 if (napi_schedule_prep(n)) { 532 __napi_schedule(n); 533 return true; 534 } 535 536 return false; 537 } 538 539 /** 540 * napi_schedule_irqoff - schedule NAPI poll 541 * @n: NAPI context 542 * 543 * Variant of napi_schedule(), assuming hard irqs are masked. 544 */ 545 static inline void napi_schedule_irqoff(struct napi_struct *n) 546 { 547 if (napi_schedule_prep(n)) 548 __napi_schedule_irqoff(n); 549 } 550 551 /** 552 * napi_complete_done - NAPI processing complete 553 * @n: NAPI context 554 * @work_done: number of packets processed 555 * 556 * Mark NAPI processing as complete. Should only be called if poll budget 557 * has not been completely consumed. 558 * Prefer over napi_complete(). 559 * Return false if device should avoid rearming interrupts. 560 */ 561 bool napi_complete_done(struct napi_struct *n, int work_done); 562 563 static inline bool napi_complete(struct napi_struct *n) 564 { 565 return napi_complete_done(n, 0); 566 } 567 568 int dev_set_threaded(struct net_device *dev, bool threaded); 569 570 /** 571 * napi_disable - prevent NAPI from scheduling 572 * @n: NAPI context 573 * 574 * Stop NAPI from being scheduled on this context. 575 * Waits till any outstanding processing completes. 576 */ 577 void napi_disable(struct napi_struct *n); 578 579 void napi_enable(struct napi_struct *n); 580 581 /** 582 * napi_synchronize - wait until NAPI is not running 583 * @n: NAPI context 584 * 585 * Wait until NAPI is done being scheduled on this context. 586 * Waits till any outstanding processing completes but 587 * does not disable future activations. 588 */ 589 static inline void napi_synchronize(const struct napi_struct *n) 590 { 591 if (IS_ENABLED(CONFIG_SMP)) 592 while (test_bit(NAPI_STATE_SCHED, &n->state)) 593 msleep(1); 594 else 595 barrier(); 596 } 597 598 /** 599 * napi_if_scheduled_mark_missed - if napi is running, set the 600 * NAPIF_STATE_MISSED 601 * @n: NAPI context 602 * 603 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 604 * NAPI is scheduled. 605 **/ 606 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 607 { 608 unsigned long val, new; 609 610 val = READ_ONCE(n->state); 611 do { 612 if (val & NAPIF_STATE_DISABLE) 613 return true; 614 615 if (!(val & NAPIF_STATE_SCHED)) 616 return false; 617 618 new = val | NAPIF_STATE_MISSED; 619 } while (!try_cmpxchg(&n->state, &val, new)); 620 621 return true; 622 } 623 624 enum netdev_queue_state_t { 625 __QUEUE_STATE_DRV_XOFF, 626 __QUEUE_STATE_STACK_XOFF, 627 __QUEUE_STATE_FROZEN, 628 }; 629 630 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 631 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 632 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 633 634 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 635 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 636 QUEUE_STATE_FROZEN) 637 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 638 QUEUE_STATE_FROZEN) 639 640 /* 641 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 642 * netif_tx_* functions below are used to manipulate this flag. The 643 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 644 * queue independently. The netif_xmit_*stopped functions below are called 645 * to check if the queue has been stopped by the driver or stack (either 646 * of the XOFF bits are set in the state). Drivers should not need to call 647 * netif_xmit*stopped functions, they should only be using netif_tx_*. 648 */ 649 650 struct netdev_queue { 651 /* 652 * read-mostly part 653 */ 654 struct net_device *dev; 655 netdevice_tracker dev_tracker; 656 657 struct Qdisc __rcu *qdisc; 658 struct Qdisc __rcu *qdisc_sleeping; 659 #ifdef CONFIG_SYSFS 660 struct kobject kobj; 661 #endif 662 unsigned long tx_maxrate; 663 /* 664 * Number of TX timeouts for this queue 665 * (/sys/class/net/DEV/Q/trans_timeout) 666 */ 667 atomic_long_t trans_timeout; 668 669 /* Subordinate device that the queue has been assigned to */ 670 struct net_device *sb_dev; 671 #ifdef CONFIG_XDP_SOCKETS 672 struct xsk_buff_pool *pool; 673 #endif 674 675 /* 676 * write-mostly part 677 */ 678 #ifdef CONFIG_BQL 679 struct dql dql; 680 #endif 681 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 682 int xmit_lock_owner; 683 /* 684 * Time (in jiffies) of last Tx 685 */ 686 unsigned long trans_start; 687 688 unsigned long state; 689 690 /* 691 * slow- / control-path part 692 */ 693 /* NAPI instance for the queue 694 * Readers and writers must hold RTNL 695 */ 696 struct napi_struct *napi; 697 698 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 699 int numa_node; 700 #endif 701 } ____cacheline_aligned_in_smp; 702 703 extern int sysctl_fb_tunnels_only_for_init_net; 704 extern int sysctl_devconf_inherit_init_net; 705 706 /* 707 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 708 * == 1 : For initns only 709 * == 2 : For none. 710 */ 711 static inline bool net_has_fallback_tunnels(const struct net *net) 712 { 713 #if IS_ENABLED(CONFIG_SYSCTL) 714 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net); 715 716 return !fb_tunnels_only_for_init_net || 717 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1); 718 #else 719 return true; 720 #endif 721 } 722 723 static inline int net_inherit_devconf(void) 724 { 725 #if IS_ENABLED(CONFIG_SYSCTL) 726 return READ_ONCE(sysctl_devconf_inherit_init_net); 727 #else 728 return 0; 729 #endif 730 } 731 732 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 733 { 734 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 735 return q->numa_node; 736 #else 737 return NUMA_NO_NODE; 738 #endif 739 } 740 741 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 742 { 743 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 744 q->numa_node = node; 745 #endif 746 } 747 748 #ifdef CONFIG_RFS_ACCEL 749 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 750 u16 filter_id); 751 #endif 752 753 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 754 enum xps_map_type { 755 XPS_CPUS = 0, 756 XPS_RXQS, 757 XPS_MAPS_MAX, 758 }; 759 760 #ifdef CONFIG_XPS 761 /* 762 * This structure holds an XPS map which can be of variable length. The 763 * map is an array of queues. 764 */ 765 struct xps_map { 766 unsigned int len; 767 unsigned int alloc_len; 768 struct rcu_head rcu; 769 u16 queues[]; 770 }; 771 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 772 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 773 - sizeof(struct xps_map)) / sizeof(u16)) 774 775 /* 776 * This structure holds all XPS maps for device. Maps are indexed by CPU. 777 * 778 * We keep track of the number of cpus/rxqs used when the struct is allocated, 779 * in nr_ids. This will help not accessing out-of-bound memory. 780 * 781 * We keep track of the number of traffic classes used when the struct is 782 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 783 * not crossing its upper bound, as the original dev->num_tc can be updated in 784 * the meantime. 785 */ 786 struct xps_dev_maps { 787 struct rcu_head rcu; 788 unsigned int nr_ids; 789 s16 num_tc; 790 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 791 }; 792 793 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 794 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 795 796 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 797 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 798 799 #endif /* CONFIG_XPS */ 800 801 #define TC_MAX_QUEUE 16 802 #define TC_BITMASK 15 803 /* HW offloaded queuing disciplines txq count and offset maps */ 804 struct netdev_tc_txq { 805 u16 count; 806 u16 offset; 807 }; 808 809 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 810 /* 811 * This structure is to hold information about the device 812 * configured to run FCoE protocol stack. 813 */ 814 struct netdev_fcoe_hbainfo { 815 char manufacturer[64]; 816 char serial_number[64]; 817 char hardware_version[64]; 818 char driver_version[64]; 819 char optionrom_version[64]; 820 char firmware_version[64]; 821 char model[256]; 822 char model_description[256]; 823 }; 824 #endif 825 826 #define MAX_PHYS_ITEM_ID_LEN 32 827 828 /* This structure holds a unique identifier to identify some 829 * physical item (port for example) used by a netdevice. 830 */ 831 struct netdev_phys_item_id { 832 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 833 unsigned char id_len; 834 }; 835 836 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 837 struct netdev_phys_item_id *b) 838 { 839 return a->id_len == b->id_len && 840 memcmp(a->id, b->id, a->id_len) == 0; 841 } 842 843 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 844 struct sk_buff *skb, 845 struct net_device *sb_dev); 846 847 enum net_device_path_type { 848 DEV_PATH_ETHERNET = 0, 849 DEV_PATH_VLAN, 850 DEV_PATH_BRIDGE, 851 DEV_PATH_PPPOE, 852 DEV_PATH_DSA, 853 DEV_PATH_MTK_WDMA, 854 }; 855 856 struct net_device_path { 857 enum net_device_path_type type; 858 const struct net_device *dev; 859 union { 860 struct { 861 u16 id; 862 __be16 proto; 863 u8 h_dest[ETH_ALEN]; 864 } encap; 865 struct { 866 enum { 867 DEV_PATH_BR_VLAN_KEEP, 868 DEV_PATH_BR_VLAN_TAG, 869 DEV_PATH_BR_VLAN_UNTAG, 870 DEV_PATH_BR_VLAN_UNTAG_HW, 871 } vlan_mode; 872 u16 vlan_id; 873 __be16 vlan_proto; 874 } bridge; 875 struct { 876 int port; 877 u16 proto; 878 } dsa; 879 struct { 880 u8 wdma_idx; 881 u8 queue; 882 u16 wcid; 883 u8 bss; 884 u8 amsdu; 885 } mtk_wdma; 886 }; 887 }; 888 889 #define NET_DEVICE_PATH_STACK_MAX 5 890 #define NET_DEVICE_PATH_VLAN_MAX 2 891 892 struct net_device_path_stack { 893 int num_paths; 894 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 895 }; 896 897 struct net_device_path_ctx { 898 const struct net_device *dev; 899 u8 daddr[ETH_ALEN]; 900 901 int num_vlans; 902 struct { 903 u16 id; 904 __be16 proto; 905 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 906 }; 907 908 enum tc_setup_type { 909 TC_QUERY_CAPS, 910 TC_SETUP_QDISC_MQPRIO, 911 TC_SETUP_CLSU32, 912 TC_SETUP_CLSFLOWER, 913 TC_SETUP_CLSMATCHALL, 914 TC_SETUP_CLSBPF, 915 TC_SETUP_BLOCK, 916 TC_SETUP_QDISC_CBS, 917 TC_SETUP_QDISC_RED, 918 TC_SETUP_QDISC_PRIO, 919 TC_SETUP_QDISC_MQ, 920 TC_SETUP_QDISC_ETF, 921 TC_SETUP_ROOT_QDISC, 922 TC_SETUP_QDISC_GRED, 923 TC_SETUP_QDISC_TAPRIO, 924 TC_SETUP_FT, 925 TC_SETUP_QDISC_ETS, 926 TC_SETUP_QDISC_TBF, 927 TC_SETUP_QDISC_FIFO, 928 TC_SETUP_QDISC_HTB, 929 TC_SETUP_ACT, 930 }; 931 932 /* These structures hold the attributes of bpf state that are being passed 933 * to the netdevice through the bpf op. 934 */ 935 enum bpf_netdev_command { 936 /* Set or clear a bpf program used in the earliest stages of packet 937 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 938 * is responsible for calling bpf_prog_put on any old progs that are 939 * stored. In case of error, the callee need not release the new prog 940 * reference, but on success it takes ownership and must bpf_prog_put 941 * when it is no longer used. 942 */ 943 XDP_SETUP_PROG, 944 XDP_SETUP_PROG_HW, 945 /* BPF program for offload callbacks, invoked at program load time. */ 946 BPF_OFFLOAD_MAP_ALLOC, 947 BPF_OFFLOAD_MAP_FREE, 948 XDP_SETUP_XSK_POOL, 949 }; 950 951 struct bpf_prog_offload_ops; 952 struct netlink_ext_ack; 953 struct xdp_umem; 954 struct xdp_dev_bulk_queue; 955 struct bpf_xdp_link; 956 957 enum bpf_xdp_mode { 958 XDP_MODE_SKB = 0, 959 XDP_MODE_DRV = 1, 960 XDP_MODE_HW = 2, 961 __MAX_XDP_MODE 962 }; 963 964 struct bpf_xdp_entity { 965 struct bpf_prog *prog; 966 struct bpf_xdp_link *link; 967 }; 968 969 struct netdev_bpf { 970 enum bpf_netdev_command command; 971 union { 972 /* XDP_SETUP_PROG */ 973 struct { 974 u32 flags; 975 struct bpf_prog *prog; 976 struct netlink_ext_ack *extack; 977 }; 978 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 979 struct { 980 struct bpf_offloaded_map *offmap; 981 }; 982 /* XDP_SETUP_XSK_POOL */ 983 struct { 984 struct xsk_buff_pool *pool; 985 u16 queue_id; 986 } xsk; 987 }; 988 }; 989 990 /* Flags for ndo_xsk_wakeup. */ 991 #define XDP_WAKEUP_RX (1 << 0) 992 #define XDP_WAKEUP_TX (1 << 1) 993 994 #ifdef CONFIG_XFRM_OFFLOAD 995 struct xfrmdev_ops { 996 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack); 997 void (*xdo_dev_state_delete) (struct xfrm_state *x); 998 void (*xdo_dev_state_free) (struct xfrm_state *x); 999 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 1000 struct xfrm_state *x); 1001 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1002 void (*xdo_dev_state_update_stats) (struct xfrm_state *x); 1003 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack); 1004 void (*xdo_dev_policy_delete) (struct xfrm_policy *x); 1005 void (*xdo_dev_policy_free) (struct xfrm_policy *x); 1006 }; 1007 #endif 1008 1009 struct dev_ifalias { 1010 struct rcu_head rcuhead; 1011 char ifalias[]; 1012 }; 1013 1014 struct devlink; 1015 struct tlsdev_ops; 1016 1017 struct netdev_net_notifier { 1018 struct list_head list; 1019 struct notifier_block *nb; 1020 }; 1021 1022 /* 1023 * This structure defines the management hooks for network devices. 1024 * The following hooks can be defined; unless noted otherwise, they are 1025 * optional and can be filled with a null pointer. 1026 * 1027 * int (*ndo_init)(struct net_device *dev); 1028 * This function is called once when a network device is registered. 1029 * The network device can use this for any late stage initialization 1030 * or semantic validation. It can fail with an error code which will 1031 * be propagated back to register_netdev. 1032 * 1033 * void (*ndo_uninit)(struct net_device *dev); 1034 * This function is called when device is unregistered or when registration 1035 * fails. It is not called if init fails. 1036 * 1037 * int (*ndo_open)(struct net_device *dev); 1038 * This function is called when a network device transitions to the up 1039 * state. 1040 * 1041 * int (*ndo_stop)(struct net_device *dev); 1042 * This function is called when a network device transitions to the down 1043 * state. 1044 * 1045 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1046 * struct net_device *dev); 1047 * Called when a packet needs to be transmitted. 1048 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1049 * the queue before that can happen; it's for obsolete devices and weird 1050 * corner cases, but the stack really does a non-trivial amount 1051 * of useless work if you return NETDEV_TX_BUSY. 1052 * Required; cannot be NULL. 1053 * 1054 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1055 * struct net_device *dev 1056 * netdev_features_t features); 1057 * Called by core transmit path to determine if device is capable of 1058 * performing offload operations on a given packet. This is to give 1059 * the device an opportunity to implement any restrictions that cannot 1060 * be otherwise expressed by feature flags. The check is called with 1061 * the set of features that the stack has calculated and it returns 1062 * those the driver believes to be appropriate. 1063 * 1064 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1065 * struct net_device *sb_dev); 1066 * Called to decide which queue to use when device supports multiple 1067 * transmit queues. 1068 * 1069 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1070 * This function is called to allow device receiver to make 1071 * changes to configuration when multicast or promiscuous is enabled. 1072 * 1073 * void (*ndo_set_rx_mode)(struct net_device *dev); 1074 * This function is called device changes address list filtering. 1075 * If driver handles unicast address filtering, it should set 1076 * IFF_UNICAST_FLT in its priv_flags. 1077 * 1078 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1079 * This function is called when the Media Access Control address 1080 * needs to be changed. If this interface is not defined, the 1081 * MAC address can not be changed. 1082 * 1083 * int (*ndo_validate_addr)(struct net_device *dev); 1084 * Test if Media Access Control address is valid for the device. 1085 * 1086 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1087 * Old-style ioctl entry point. This is used internally by the 1088 * appletalk and ieee802154 subsystems but is no longer called by 1089 * the device ioctl handler. 1090 * 1091 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1092 * Used by the bonding driver for its device specific ioctls: 1093 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1094 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1095 * 1096 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1097 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1098 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1099 * 1100 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1101 * Used to set network devices bus interface parameters. This interface 1102 * is retained for legacy reasons; new devices should use the bus 1103 * interface (PCI) for low level management. 1104 * 1105 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1106 * Called when a user wants to change the Maximum Transfer Unit 1107 * of a device. 1108 * 1109 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1110 * Callback used when the transmitter has not made any progress 1111 * for dev->watchdog ticks. 1112 * 1113 * void (*ndo_get_stats64)(struct net_device *dev, 1114 * struct rtnl_link_stats64 *storage); 1115 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1116 * Called when a user wants to get the network device usage 1117 * statistics. Drivers must do one of the following: 1118 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1119 * rtnl_link_stats64 structure passed by the caller. 1120 * 2. Define @ndo_get_stats to update a net_device_stats structure 1121 * (which should normally be dev->stats) and return a pointer to 1122 * it. The structure may be changed asynchronously only if each 1123 * field is written atomically. 1124 * 3. Update dev->stats asynchronously and atomically, and define 1125 * neither operation. 1126 * 1127 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1128 * Return true if this device supports offload stats of this attr_id. 1129 * 1130 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1131 * void *attr_data) 1132 * Get statistics for offload operations by attr_id. Write it into the 1133 * attr_data pointer. 1134 * 1135 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1136 * If device supports VLAN filtering this function is called when a 1137 * VLAN id is registered. 1138 * 1139 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1140 * If device supports VLAN filtering this function is called when a 1141 * VLAN id is unregistered. 1142 * 1143 * void (*ndo_poll_controller)(struct net_device *dev); 1144 * 1145 * SR-IOV management functions. 1146 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1147 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1148 * u8 qos, __be16 proto); 1149 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1150 * int max_tx_rate); 1151 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1152 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1153 * int (*ndo_get_vf_config)(struct net_device *dev, 1154 * int vf, struct ifla_vf_info *ivf); 1155 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1156 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1157 * struct nlattr *port[]); 1158 * 1159 * Enable or disable the VF ability to query its RSS Redirection Table and 1160 * Hash Key. This is needed since on some devices VF share this information 1161 * with PF and querying it may introduce a theoretical security risk. 1162 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1163 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1164 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1165 * void *type_data); 1166 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1167 * This is always called from the stack with the rtnl lock held and netif 1168 * tx queues stopped. This allows the netdevice to perform queue 1169 * management safely. 1170 * 1171 * Fiber Channel over Ethernet (FCoE) offload functions. 1172 * int (*ndo_fcoe_enable)(struct net_device *dev); 1173 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1174 * so the underlying device can perform whatever needed configuration or 1175 * initialization to support acceleration of FCoE traffic. 1176 * 1177 * int (*ndo_fcoe_disable)(struct net_device *dev); 1178 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1179 * so the underlying device can perform whatever needed clean-ups to 1180 * stop supporting acceleration of FCoE traffic. 1181 * 1182 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1183 * struct scatterlist *sgl, unsigned int sgc); 1184 * Called when the FCoE Initiator wants to initialize an I/O that 1185 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1186 * perform necessary setup and returns 1 to indicate the device is set up 1187 * successfully to perform DDP on this I/O, otherwise this returns 0. 1188 * 1189 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1190 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1191 * indicated by the FC exchange id 'xid', so the underlying device can 1192 * clean up and reuse resources for later DDP requests. 1193 * 1194 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1195 * struct scatterlist *sgl, unsigned int sgc); 1196 * Called when the FCoE Target wants to initialize an I/O that 1197 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1198 * perform necessary setup and returns 1 to indicate the device is set up 1199 * successfully to perform DDP on this I/O, otherwise this returns 0. 1200 * 1201 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1202 * struct netdev_fcoe_hbainfo *hbainfo); 1203 * Called when the FCoE Protocol stack wants information on the underlying 1204 * device. This information is utilized by the FCoE protocol stack to 1205 * register attributes with Fiber Channel management service as per the 1206 * FC-GS Fabric Device Management Information(FDMI) specification. 1207 * 1208 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1209 * Called when the underlying device wants to override default World Wide 1210 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1211 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1212 * protocol stack to use. 1213 * 1214 * RFS acceleration. 1215 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1216 * u16 rxq_index, u32 flow_id); 1217 * Set hardware filter for RFS. rxq_index is the target queue index; 1218 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1219 * Return the filter ID on success, or a negative error code. 1220 * 1221 * Slave management functions (for bridge, bonding, etc). 1222 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1223 * Called to make another netdev an underling. 1224 * 1225 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1226 * Called to release previously enslaved netdev. 1227 * 1228 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1229 * struct sk_buff *skb, 1230 * bool all_slaves); 1231 * Get the xmit slave of master device. If all_slaves is true, function 1232 * assume all the slaves can transmit. 1233 * 1234 * Feature/offload setting functions. 1235 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1236 * netdev_features_t features); 1237 * Adjusts the requested feature flags according to device-specific 1238 * constraints, and returns the resulting flags. Must not modify 1239 * the device state. 1240 * 1241 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1242 * Called to update device configuration to new features. Passed 1243 * feature set might be less than what was returned by ndo_fix_features()). 1244 * Must return >0 or -errno if it changed dev->features itself. 1245 * 1246 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1247 * struct net_device *dev, 1248 * const unsigned char *addr, u16 vid, u16 flags, 1249 * struct netlink_ext_ack *extack); 1250 * Adds an FDB entry to dev for addr. 1251 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1252 * struct net_device *dev, 1253 * const unsigned char *addr, u16 vid) 1254 * Deletes the FDB entry from dev corresponding to addr. 1255 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev, 1256 * struct netlink_ext_ack *extack); 1257 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1258 * struct net_device *dev, struct net_device *filter_dev, 1259 * int *idx) 1260 * Used to add FDB entries to dump requests. Implementers should add 1261 * entries to skb and update idx with the number of entries. 1262 * 1263 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[], 1264 * u16 nlmsg_flags, struct netlink_ext_ack *extack); 1265 * Adds an MDB entry to dev. 1266 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[], 1267 * struct netlink_ext_ack *extack); 1268 * Deletes the MDB entry from dev. 1269 * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[], 1270 * struct netlink_ext_ack *extack); 1271 * Bulk deletes MDB entries from dev. 1272 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb, 1273 * struct netlink_callback *cb); 1274 * Dumps MDB entries from dev. The first argument (marker) in the netlink 1275 * callback is used by core rtnetlink code. 1276 * 1277 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1278 * u16 flags, struct netlink_ext_ack *extack) 1279 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1280 * struct net_device *dev, u32 filter_mask, 1281 * int nlflags) 1282 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1283 * u16 flags); 1284 * 1285 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1286 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1287 * which do not represent real hardware may define this to allow their 1288 * userspace components to manage their virtual carrier state. Devices 1289 * that determine carrier state from physical hardware properties (eg 1290 * network cables) or protocol-dependent mechanisms (eg 1291 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1292 * 1293 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1294 * struct netdev_phys_item_id *ppid); 1295 * Called to get ID of physical port of this device. If driver does 1296 * not implement this, it is assumed that the hw is not able to have 1297 * multiple net devices on single physical port. 1298 * 1299 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1300 * struct netdev_phys_item_id *ppid) 1301 * Called to get the parent ID of the physical port of this device. 1302 * 1303 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1304 * struct net_device *dev) 1305 * Called by upper layer devices to accelerate switching or other 1306 * station functionality into hardware. 'pdev is the lowerdev 1307 * to use for the offload and 'dev' is the net device that will 1308 * back the offload. Returns a pointer to the private structure 1309 * the upper layer will maintain. 1310 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1311 * Called by upper layer device to delete the station created 1312 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1313 * the station and priv is the structure returned by the add 1314 * operation. 1315 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1316 * int queue_index, u32 maxrate); 1317 * Called when a user wants to set a max-rate limitation of specific 1318 * TX queue. 1319 * int (*ndo_get_iflink)(const struct net_device *dev); 1320 * Called to get the iflink value of this device. 1321 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1322 * This function is used to get egress tunnel information for given skb. 1323 * This is useful for retrieving outer tunnel header parameters while 1324 * sampling packet. 1325 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1326 * This function is used to specify the headroom that the skb must 1327 * consider when allocation skb during packet reception. Setting 1328 * appropriate rx headroom value allows avoiding skb head copy on 1329 * forward. Setting a negative value resets the rx headroom to the 1330 * default value. 1331 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1332 * This function is used to set or query state related to XDP on the 1333 * netdevice and manage BPF offload. See definition of 1334 * enum bpf_netdev_command for details. 1335 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1336 * u32 flags); 1337 * This function is used to submit @n XDP packets for transmit on a 1338 * netdevice. Returns number of frames successfully transmitted, frames 1339 * that got dropped are freed/returned via xdp_return_frame(). 1340 * Returns negative number, means general error invoking ndo, meaning 1341 * no frames were xmit'ed and core-caller will free all frames. 1342 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1343 * struct xdp_buff *xdp); 1344 * Get the xmit slave of master device based on the xdp_buff. 1345 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1346 * This function is used to wake up the softirq, ksoftirqd or kthread 1347 * responsible for sending and/or receiving packets on a specific 1348 * queue id bound to an AF_XDP socket. The flags field specifies if 1349 * only RX, only Tx, or both should be woken up using the flags 1350 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1351 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm_kern *p, 1352 * int cmd); 1353 * Add, change, delete or get information on an IPv4 tunnel. 1354 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1355 * If a device is paired with a peer device, return the peer instance. 1356 * The caller must be under RCU read context. 1357 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1358 * Get the forwarding path to reach the real device from the HW destination address 1359 * ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1360 * const struct skb_shared_hwtstamps *hwtstamps, 1361 * bool cycles); 1362 * Get hardware timestamp based on normal/adjustable time or free running 1363 * cycle counter. This function is required if physical clock supports a 1364 * free running cycle counter. 1365 * 1366 * int (*ndo_hwtstamp_get)(struct net_device *dev, 1367 * struct kernel_hwtstamp_config *kernel_config); 1368 * Get the currently configured hardware timestamping parameters for the 1369 * NIC device. 1370 * 1371 * int (*ndo_hwtstamp_set)(struct net_device *dev, 1372 * struct kernel_hwtstamp_config *kernel_config, 1373 * struct netlink_ext_ack *extack); 1374 * Change the hardware timestamping parameters for NIC device. 1375 */ 1376 struct net_device_ops { 1377 int (*ndo_init)(struct net_device *dev); 1378 void (*ndo_uninit)(struct net_device *dev); 1379 int (*ndo_open)(struct net_device *dev); 1380 int (*ndo_stop)(struct net_device *dev); 1381 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1382 struct net_device *dev); 1383 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1384 struct net_device *dev, 1385 netdev_features_t features); 1386 u16 (*ndo_select_queue)(struct net_device *dev, 1387 struct sk_buff *skb, 1388 struct net_device *sb_dev); 1389 void (*ndo_change_rx_flags)(struct net_device *dev, 1390 int flags); 1391 void (*ndo_set_rx_mode)(struct net_device *dev); 1392 int (*ndo_set_mac_address)(struct net_device *dev, 1393 void *addr); 1394 int (*ndo_validate_addr)(struct net_device *dev); 1395 int (*ndo_do_ioctl)(struct net_device *dev, 1396 struct ifreq *ifr, int cmd); 1397 int (*ndo_eth_ioctl)(struct net_device *dev, 1398 struct ifreq *ifr, int cmd); 1399 int (*ndo_siocbond)(struct net_device *dev, 1400 struct ifreq *ifr, int cmd); 1401 int (*ndo_siocwandev)(struct net_device *dev, 1402 struct if_settings *ifs); 1403 int (*ndo_siocdevprivate)(struct net_device *dev, 1404 struct ifreq *ifr, 1405 void __user *data, int cmd); 1406 int (*ndo_set_config)(struct net_device *dev, 1407 struct ifmap *map); 1408 int (*ndo_change_mtu)(struct net_device *dev, 1409 int new_mtu); 1410 int (*ndo_neigh_setup)(struct net_device *dev, 1411 struct neigh_parms *); 1412 void (*ndo_tx_timeout) (struct net_device *dev, 1413 unsigned int txqueue); 1414 1415 void (*ndo_get_stats64)(struct net_device *dev, 1416 struct rtnl_link_stats64 *storage); 1417 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1418 int (*ndo_get_offload_stats)(int attr_id, 1419 const struct net_device *dev, 1420 void *attr_data); 1421 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1422 1423 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1424 __be16 proto, u16 vid); 1425 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1426 __be16 proto, u16 vid); 1427 #ifdef CONFIG_NET_POLL_CONTROLLER 1428 void (*ndo_poll_controller)(struct net_device *dev); 1429 int (*ndo_netpoll_setup)(struct net_device *dev); 1430 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1431 #endif 1432 int (*ndo_set_vf_mac)(struct net_device *dev, 1433 int queue, u8 *mac); 1434 int (*ndo_set_vf_vlan)(struct net_device *dev, 1435 int queue, u16 vlan, 1436 u8 qos, __be16 proto); 1437 int (*ndo_set_vf_rate)(struct net_device *dev, 1438 int vf, int min_tx_rate, 1439 int max_tx_rate); 1440 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1441 int vf, bool setting); 1442 int (*ndo_set_vf_trust)(struct net_device *dev, 1443 int vf, bool setting); 1444 int (*ndo_get_vf_config)(struct net_device *dev, 1445 int vf, 1446 struct ifla_vf_info *ivf); 1447 int (*ndo_set_vf_link_state)(struct net_device *dev, 1448 int vf, int link_state); 1449 int (*ndo_get_vf_stats)(struct net_device *dev, 1450 int vf, 1451 struct ifla_vf_stats 1452 *vf_stats); 1453 int (*ndo_set_vf_port)(struct net_device *dev, 1454 int vf, 1455 struct nlattr *port[]); 1456 int (*ndo_get_vf_port)(struct net_device *dev, 1457 int vf, struct sk_buff *skb); 1458 int (*ndo_get_vf_guid)(struct net_device *dev, 1459 int vf, 1460 struct ifla_vf_guid *node_guid, 1461 struct ifla_vf_guid *port_guid); 1462 int (*ndo_set_vf_guid)(struct net_device *dev, 1463 int vf, u64 guid, 1464 int guid_type); 1465 int (*ndo_set_vf_rss_query_en)( 1466 struct net_device *dev, 1467 int vf, bool setting); 1468 int (*ndo_setup_tc)(struct net_device *dev, 1469 enum tc_setup_type type, 1470 void *type_data); 1471 #if IS_ENABLED(CONFIG_FCOE) 1472 int (*ndo_fcoe_enable)(struct net_device *dev); 1473 int (*ndo_fcoe_disable)(struct net_device *dev); 1474 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1475 u16 xid, 1476 struct scatterlist *sgl, 1477 unsigned int sgc); 1478 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1479 u16 xid); 1480 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1481 u16 xid, 1482 struct scatterlist *sgl, 1483 unsigned int sgc); 1484 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1485 struct netdev_fcoe_hbainfo *hbainfo); 1486 #endif 1487 1488 #if IS_ENABLED(CONFIG_LIBFCOE) 1489 #define NETDEV_FCOE_WWNN 0 1490 #define NETDEV_FCOE_WWPN 1 1491 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1492 u64 *wwn, int type); 1493 #endif 1494 1495 #ifdef CONFIG_RFS_ACCEL 1496 int (*ndo_rx_flow_steer)(struct net_device *dev, 1497 const struct sk_buff *skb, 1498 u16 rxq_index, 1499 u32 flow_id); 1500 #endif 1501 int (*ndo_add_slave)(struct net_device *dev, 1502 struct net_device *slave_dev, 1503 struct netlink_ext_ack *extack); 1504 int (*ndo_del_slave)(struct net_device *dev, 1505 struct net_device *slave_dev); 1506 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1507 struct sk_buff *skb, 1508 bool all_slaves); 1509 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1510 struct sock *sk); 1511 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1512 netdev_features_t features); 1513 int (*ndo_set_features)(struct net_device *dev, 1514 netdev_features_t features); 1515 int (*ndo_neigh_construct)(struct net_device *dev, 1516 struct neighbour *n); 1517 void (*ndo_neigh_destroy)(struct net_device *dev, 1518 struct neighbour *n); 1519 1520 int (*ndo_fdb_add)(struct ndmsg *ndm, 1521 struct nlattr *tb[], 1522 struct net_device *dev, 1523 const unsigned char *addr, 1524 u16 vid, 1525 u16 flags, 1526 struct netlink_ext_ack *extack); 1527 int (*ndo_fdb_del)(struct ndmsg *ndm, 1528 struct nlattr *tb[], 1529 struct net_device *dev, 1530 const unsigned char *addr, 1531 u16 vid, struct netlink_ext_ack *extack); 1532 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, 1533 struct net_device *dev, 1534 struct netlink_ext_ack *extack); 1535 int (*ndo_fdb_dump)(struct sk_buff *skb, 1536 struct netlink_callback *cb, 1537 struct net_device *dev, 1538 struct net_device *filter_dev, 1539 int *idx); 1540 int (*ndo_fdb_get)(struct sk_buff *skb, 1541 struct nlattr *tb[], 1542 struct net_device *dev, 1543 const unsigned char *addr, 1544 u16 vid, u32 portid, u32 seq, 1545 struct netlink_ext_ack *extack); 1546 int (*ndo_mdb_add)(struct net_device *dev, 1547 struct nlattr *tb[], 1548 u16 nlmsg_flags, 1549 struct netlink_ext_ack *extack); 1550 int (*ndo_mdb_del)(struct net_device *dev, 1551 struct nlattr *tb[], 1552 struct netlink_ext_ack *extack); 1553 int (*ndo_mdb_del_bulk)(struct net_device *dev, 1554 struct nlattr *tb[], 1555 struct netlink_ext_ack *extack); 1556 int (*ndo_mdb_dump)(struct net_device *dev, 1557 struct sk_buff *skb, 1558 struct netlink_callback *cb); 1559 int (*ndo_mdb_get)(struct net_device *dev, 1560 struct nlattr *tb[], u32 portid, 1561 u32 seq, 1562 struct netlink_ext_ack *extack); 1563 int (*ndo_bridge_setlink)(struct net_device *dev, 1564 struct nlmsghdr *nlh, 1565 u16 flags, 1566 struct netlink_ext_ack *extack); 1567 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1568 u32 pid, u32 seq, 1569 struct net_device *dev, 1570 u32 filter_mask, 1571 int nlflags); 1572 int (*ndo_bridge_dellink)(struct net_device *dev, 1573 struct nlmsghdr *nlh, 1574 u16 flags); 1575 int (*ndo_change_carrier)(struct net_device *dev, 1576 bool new_carrier); 1577 int (*ndo_get_phys_port_id)(struct net_device *dev, 1578 struct netdev_phys_item_id *ppid); 1579 int (*ndo_get_port_parent_id)(struct net_device *dev, 1580 struct netdev_phys_item_id *ppid); 1581 int (*ndo_get_phys_port_name)(struct net_device *dev, 1582 char *name, size_t len); 1583 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1584 struct net_device *dev); 1585 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1586 void *priv); 1587 1588 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1589 int queue_index, 1590 u32 maxrate); 1591 int (*ndo_get_iflink)(const struct net_device *dev); 1592 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1593 struct sk_buff *skb); 1594 void (*ndo_set_rx_headroom)(struct net_device *dev, 1595 int needed_headroom); 1596 int (*ndo_bpf)(struct net_device *dev, 1597 struct netdev_bpf *bpf); 1598 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1599 struct xdp_frame **xdp, 1600 u32 flags); 1601 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1602 struct xdp_buff *xdp); 1603 int (*ndo_xsk_wakeup)(struct net_device *dev, 1604 u32 queue_id, u32 flags); 1605 int (*ndo_tunnel_ctl)(struct net_device *dev, 1606 struct ip_tunnel_parm_kern *p, 1607 int cmd); 1608 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1609 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1610 struct net_device_path *path); 1611 ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1612 const struct skb_shared_hwtstamps *hwtstamps, 1613 bool cycles); 1614 int (*ndo_hwtstamp_get)(struct net_device *dev, 1615 struct kernel_hwtstamp_config *kernel_config); 1616 int (*ndo_hwtstamp_set)(struct net_device *dev, 1617 struct kernel_hwtstamp_config *kernel_config, 1618 struct netlink_ext_ack *extack); 1619 1620 #if IS_ENABLED(CONFIG_NET_SHAPER) 1621 /** 1622 * @net_shaper_ops: Device shaping offload operations 1623 * see include/net/net_shapers.h 1624 */ 1625 const struct net_shaper_ops *net_shaper_ops; 1626 #endif 1627 }; 1628 1629 /** 1630 * enum netdev_priv_flags - &struct net_device priv_flags 1631 * 1632 * These are the &struct net_device, they are only set internally 1633 * by drivers and used in the kernel. These flags are invisible to 1634 * userspace; this means that the order of these flags can change 1635 * during any kernel release. 1636 * 1637 * You should add bitfield booleans after either net_device::priv_flags 1638 * (hotpath) or ::threaded (slowpath) instead of extending these flags. 1639 * 1640 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1641 * @IFF_EBRIDGE: Ethernet bridging device 1642 * @IFF_BONDING: bonding master or slave 1643 * @IFF_ISATAP: ISATAP interface (RFC4214) 1644 * @IFF_WAN_HDLC: WAN HDLC device 1645 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1646 * release skb->dst 1647 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1648 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1649 * @IFF_MACVLAN_PORT: device used as macvlan port 1650 * @IFF_BRIDGE_PORT: device used as bridge port 1651 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1652 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1653 * @IFF_UNICAST_FLT: Supports unicast filtering 1654 * @IFF_TEAM_PORT: device used as team port 1655 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1656 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1657 * change when it's running 1658 * @IFF_MACVLAN: Macvlan device 1659 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1660 * underlying stacked devices 1661 * @IFF_L3MDEV_MASTER: device is an L3 master device 1662 * @IFF_NO_QUEUE: device can run without qdisc attached 1663 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1664 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1665 * @IFF_TEAM: device is a team device 1666 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1667 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1668 * entity (i.e. the master device for bridged veth) 1669 * @IFF_MACSEC: device is a MACsec device 1670 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1671 * @IFF_FAILOVER: device is a failover master device 1672 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1673 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1674 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf 1675 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1676 * skb_headlen(skb) == 0 (data starts from frag0) 1677 */ 1678 enum netdev_priv_flags { 1679 IFF_802_1Q_VLAN = 1<<0, 1680 IFF_EBRIDGE = 1<<1, 1681 IFF_BONDING = 1<<2, 1682 IFF_ISATAP = 1<<3, 1683 IFF_WAN_HDLC = 1<<4, 1684 IFF_XMIT_DST_RELEASE = 1<<5, 1685 IFF_DONT_BRIDGE = 1<<6, 1686 IFF_DISABLE_NETPOLL = 1<<7, 1687 IFF_MACVLAN_PORT = 1<<8, 1688 IFF_BRIDGE_PORT = 1<<9, 1689 IFF_OVS_DATAPATH = 1<<10, 1690 IFF_TX_SKB_SHARING = 1<<11, 1691 IFF_UNICAST_FLT = 1<<12, 1692 IFF_TEAM_PORT = 1<<13, 1693 IFF_SUPP_NOFCS = 1<<14, 1694 IFF_LIVE_ADDR_CHANGE = 1<<15, 1695 IFF_MACVLAN = 1<<16, 1696 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1697 IFF_L3MDEV_MASTER = 1<<18, 1698 IFF_NO_QUEUE = 1<<19, 1699 IFF_OPENVSWITCH = 1<<20, 1700 IFF_L3MDEV_SLAVE = 1<<21, 1701 IFF_TEAM = 1<<22, 1702 IFF_RXFH_CONFIGURED = 1<<23, 1703 IFF_PHONY_HEADROOM = 1<<24, 1704 IFF_MACSEC = 1<<25, 1705 IFF_NO_RX_HANDLER = 1<<26, 1706 IFF_FAILOVER = 1<<27, 1707 IFF_FAILOVER_SLAVE = 1<<28, 1708 IFF_L3MDEV_RX_HANDLER = 1<<29, 1709 IFF_NO_ADDRCONF = BIT_ULL(30), 1710 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31), 1711 }; 1712 1713 /* Specifies the type of the struct net_device::ml_priv pointer */ 1714 enum netdev_ml_priv_type { 1715 ML_PRIV_NONE, 1716 ML_PRIV_CAN, 1717 }; 1718 1719 enum netdev_stat_type { 1720 NETDEV_PCPU_STAT_NONE, 1721 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */ 1722 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */ 1723 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */ 1724 }; 1725 1726 enum netdev_reg_state { 1727 NETREG_UNINITIALIZED = 0, 1728 NETREG_REGISTERED, /* completed register_netdevice */ 1729 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1730 NETREG_UNREGISTERED, /* completed unregister todo */ 1731 NETREG_RELEASED, /* called free_netdev */ 1732 NETREG_DUMMY, /* dummy device for NAPI poll */ 1733 }; 1734 1735 /** 1736 * struct net_device - The DEVICE structure. 1737 * 1738 * Actually, this whole structure is a big mistake. It mixes I/O 1739 * data with strictly "high-level" data, and it has to know about 1740 * almost every data structure used in the INET module. 1741 * 1742 * @priv_flags: flags invisible to userspace defined as bits, see 1743 * enum netdev_priv_flags for the definitions 1744 * @lltx: device supports lockless Tx. Deprecated for real HW 1745 * drivers. Mainly used by logical interfaces, such as 1746 * bonding and tunnels 1747 * 1748 * @name: This is the first field of the "visible" part of this structure 1749 * (i.e. as seen by users in the "Space.c" file). It is the name 1750 * of the interface. 1751 * 1752 * @name_node: Name hashlist node 1753 * @ifalias: SNMP alias 1754 * @mem_end: Shared memory end 1755 * @mem_start: Shared memory start 1756 * @base_addr: Device I/O address 1757 * @irq: Device IRQ number 1758 * 1759 * @state: Generic network queuing layer state, see netdev_state_t 1760 * @dev_list: The global list of network devices 1761 * @napi_list: List entry used for polling NAPI devices 1762 * @unreg_list: List entry when we are unregistering the 1763 * device; see the function unregister_netdev 1764 * @close_list: List entry used when we are closing the device 1765 * @ptype_all: Device-specific packet handlers for all protocols 1766 * @ptype_specific: Device-specific, protocol-specific packet handlers 1767 * 1768 * @adj_list: Directly linked devices, like slaves for bonding 1769 * @features: Currently active device features 1770 * @hw_features: User-changeable features 1771 * 1772 * @wanted_features: User-requested features 1773 * @vlan_features: Mask of features inheritable by VLAN devices 1774 * 1775 * @hw_enc_features: Mask of features inherited by encapsulating devices 1776 * This field indicates what encapsulation 1777 * offloads the hardware is capable of doing, 1778 * and drivers will need to set them appropriately. 1779 * 1780 * @mpls_features: Mask of features inheritable by MPLS 1781 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1782 * 1783 * @ifindex: interface index 1784 * @group: The group the device belongs to 1785 * 1786 * @stats: Statistics struct, which was left as a legacy, use 1787 * rtnl_link_stats64 instead 1788 * 1789 * @core_stats: core networking counters, 1790 * do not use this in drivers 1791 * @carrier_up_count: Number of times the carrier has been up 1792 * @carrier_down_count: Number of times the carrier has been down 1793 * 1794 * @wireless_handlers: List of functions to handle Wireless Extensions, 1795 * instead of ioctl, 1796 * see <net/iw_handler.h> for details. 1797 * 1798 * @netdev_ops: Includes several pointers to callbacks, 1799 * if one wants to override the ndo_*() functions 1800 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks. 1801 * @xsk_tx_metadata_ops: Includes pointers to AF_XDP TX metadata callbacks. 1802 * @ethtool_ops: Management operations 1803 * @l3mdev_ops: Layer 3 master device operations 1804 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1805 * discovery handling. Necessary for e.g. 6LoWPAN. 1806 * @xfrmdev_ops: Transformation offload operations 1807 * @tlsdev_ops: Transport Layer Security offload operations 1808 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1809 * of Layer 2 headers. 1810 * 1811 * @flags: Interface flags (a la BSD) 1812 * @xdp_features: XDP capability supported by the device 1813 * @gflags: Global flags ( kept as legacy ) 1814 * @priv_len: Size of the ->priv flexible array 1815 * @priv: Flexible array containing private data 1816 * @operstate: RFC2863 operstate 1817 * @link_mode: Mapping policy to operstate 1818 * @if_port: Selectable AUI, TP, ... 1819 * @dma: DMA channel 1820 * @mtu: Interface MTU value 1821 * @min_mtu: Interface Minimum MTU value 1822 * @max_mtu: Interface Maximum MTU value 1823 * @type: Interface hardware type 1824 * @hard_header_len: Maximum hardware header length. 1825 * @min_header_len: Minimum hardware header length 1826 * 1827 * @needed_headroom: Extra headroom the hardware may need, but not in all 1828 * cases can this be guaranteed 1829 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1830 * cases can this be guaranteed. Some cases also use 1831 * LL_MAX_HEADER instead to allocate the skb 1832 * 1833 * interface address info: 1834 * 1835 * @perm_addr: Permanent hw address 1836 * @addr_assign_type: Hw address assignment type 1837 * @addr_len: Hardware address length 1838 * @upper_level: Maximum depth level of upper devices. 1839 * @lower_level: Maximum depth level of lower devices. 1840 * @neigh_priv_len: Used in neigh_alloc() 1841 * @dev_id: Used to differentiate devices that share 1842 * the same link layer address 1843 * @dev_port: Used to differentiate devices that share 1844 * the same function 1845 * @addr_list_lock: XXX: need comments on this one 1846 * @name_assign_type: network interface name assignment type 1847 * @uc_promisc: Counter that indicates promiscuous mode 1848 * has been enabled due to the need to listen to 1849 * additional unicast addresses in a device that 1850 * does not implement ndo_set_rx_mode() 1851 * @uc: unicast mac addresses 1852 * @mc: multicast mac addresses 1853 * @dev_addrs: list of device hw addresses 1854 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1855 * @promiscuity: Number of times the NIC is told to work in 1856 * promiscuous mode; if it becomes 0 the NIC will 1857 * exit promiscuous mode 1858 * @allmulti: Counter, enables or disables allmulticast mode 1859 * 1860 * @vlan_info: VLAN info 1861 * @dsa_ptr: dsa specific data 1862 * @tipc_ptr: TIPC specific data 1863 * @atalk_ptr: AppleTalk link 1864 * @ip_ptr: IPv4 specific data 1865 * @ip6_ptr: IPv6 specific data 1866 * @ax25_ptr: AX.25 specific data 1867 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1868 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1869 * device struct 1870 * @mpls_ptr: mpls_dev struct pointer 1871 * @mctp_ptr: MCTP specific data 1872 * 1873 * @dev_addr: Hw address (before bcast, 1874 * because most packets are unicast) 1875 * 1876 * @_rx: Array of RX queues 1877 * @num_rx_queues: Number of RX queues 1878 * allocated at register_netdev() time 1879 * @real_num_rx_queues: Number of RX queues currently active in device 1880 * @xdp_prog: XDP sockets filter program pointer 1881 * 1882 * @rx_handler: handler for received packets 1883 * @rx_handler_data: XXX: need comments on this one 1884 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing 1885 * @ingress_queue: XXX: need comments on this one 1886 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1887 * @broadcast: hw bcast address 1888 * 1889 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1890 * indexed by RX queue number. Assigned by driver. 1891 * This must only be set if the ndo_rx_flow_steer 1892 * operation is defined 1893 * @index_hlist: Device index hash chain 1894 * 1895 * @_tx: Array of TX queues 1896 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1897 * @real_num_tx_queues: Number of TX queues currently active in device 1898 * @qdisc: Root qdisc from userspace point of view 1899 * @tx_queue_len: Max frames per queue allowed 1900 * @tx_global_lock: XXX: need comments on this one 1901 * @xdp_bulkq: XDP device bulk queue 1902 * @xps_maps: all CPUs/RXQs maps for XPS device 1903 * 1904 * @xps_maps: XXX: need comments on this one 1905 * @tcx_egress: BPF & clsact qdisc specific data for egress processing 1906 * @nf_hooks_egress: netfilter hooks executed for egress packets 1907 * @qdisc_hash: qdisc hash table 1908 * @watchdog_timeo: Represents the timeout that is used by 1909 * the watchdog (see dev_watchdog()) 1910 * @watchdog_timer: List of timers 1911 * 1912 * @proto_down_reason: reason a netdev interface is held down 1913 * @pcpu_refcnt: Number of references to this device 1914 * @dev_refcnt: Number of references to this device 1915 * @refcnt_tracker: Tracker directory for tracked references to this device 1916 * @todo_list: Delayed register/unregister 1917 * @link_watch_list: XXX: need comments on this one 1918 * 1919 * @reg_state: Register/unregister state machine 1920 * @dismantle: Device is going to be freed 1921 * @rtnl_link_state: This enum represents the phases of creating 1922 * a new link 1923 * 1924 * @needs_free_netdev: Should unregister perform free_netdev? 1925 * @priv_destructor: Called from unregister 1926 * @npinfo: XXX: need comments on this one 1927 * @nd_net: Network namespace this network device is inside 1928 * 1929 * @ml_priv: Mid-layer private 1930 * @ml_priv_type: Mid-layer private type 1931 * 1932 * @pcpu_stat_type: Type of device statistics which the core should 1933 * allocate/free: none, lstats, tstats, dstats. none 1934 * means the driver is handling statistics allocation/ 1935 * freeing internally. 1936 * @lstats: Loopback statistics: packets, bytes 1937 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes 1938 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes 1939 * 1940 * @garp_port: GARP 1941 * @mrp_port: MRP 1942 * 1943 * @dm_private: Drop monitor private 1944 * 1945 * @dev: Class/net/name entry 1946 * @sysfs_groups: Space for optional device, statistics and wireless 1947 * sysfs groups 1948 * 1949 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1950 * @rtnl_link_ops: Rtnl_link_ops 1951 * @stat_ops: Optional ops for queue-aware statistics 1952 * @queue_mgmt_ops: Optional ops for queue management 1953 * 1954 * @gso_max_size: Maximum size of generic segmentation offload 1955 * @tso_max_size: Device (as in HW) limit on the max TSO request size 1956 * @gso_max_segs: Maximum number of segments that can be passed to the 1957 * NIC for GSO 1958 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count 1959 * @gso_ipv4_max_size: Maximum size of generic segmentation offload, 1960 * for IPv4. 1961 * 1962 * @dcbnl_ops: Data Center Bridging netlink ops 1963 * @num_tc: Number of traffic classes in the net device 1964 * @tc_to_txq: XXX: need comments on this one 1965 * @prio_tc_map: XXX: need comments on this one 1966 * 1967 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1968 * 1969 * @priomap: XXX: need comments on this one 1970 * @link_topo: Physical link topology tracking attached PHYs 1971 * @phydev: Physical device may attach itself 1972 * for hardware timestamping 1973 * @sfp_bus: attached &struct sfp_bus structure. 1974 * 1975 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1976 * 1977 * @proto_down: protocol port state information can be sent to the 1978 * switch driver and used to set the phys state of the 1979 * switch port. 1980 * 1981 * @threaded: napi threaded mode is enabled 1982 * 1983 * @see_all_hwtstamp_requests: device wants to see calls to 1984 * ndo_hwtstamp_set() for all timestamp requests 1985 * regardless of source, even if those aren't 1986 * HWTSTAMP_SOURCE_NETDEV 1987 * @change_proto_down: device supports setting carrier via IFLA_PROTO_DOWN 1988 * @netns_local: interface can't change network namespaces 1989 * @fcoe_mtu: device supports maximum FCoE MTU, 2158 bytes 1990 * 1991 * @net_notifier_list: List of per-net netdev notifier block 1992 * that follow this device when it is moved 1993 * to another network namespace. 1994 * 1995 * @macsec_ops: MACsec offloading ops 1996 * 1997 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1998 * offload capabilities of the device 1999 * @udp_tunnel_nic: UDP tunnel offload state 2000 * @ethtool: ethtool related state 2001 * @xdp_state: stores info on attached XDP BPF programs 2002 * 2003 * @nested_level: Used as a parameter of spin_lock_nested() of 2004 * dev->addr_list_lock. 2005 * @unlink_list: As netif_addr_lock() can be called recursively, 2006 * keep a list of interfaces to be deleted. 2007 * @gro_max_size: Maximum size of aggregated packet in generic 2008 * receive offload (GRO) 2009 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic 2010 * receive offload (GRO), for IPv4. 2011 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP 2012 * zero copy driver 2013 * 2014 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 2015 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 2016 * @watchdog_dev_tracker: refcount tracker used by watchdog. 2017 * @dev_registered_tracker: tracker for reference held while 2018 * registered 2019 * @offload_xstats_l3: L3 HW stats for this netdevice. 2020 * 2021 * @devlink_port: Pointer to related devlink port structure. 2022 * Assigned by a driver before netdev registration using 2023 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static 2024 * during the time netdevice is registered. 2025 * 2026 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem, 2027 * where the clock is recovered. 2028 * 2029 * @max_pacing_offload_horizon: max EDT offload horizon in nsec. 2030 * @napi_config: An array of napi_config structures containing per-NAPI 2031 * settings. 2032 * @gro_flush_timeout: timeout for GRO layer in NAPI 2033 * @napi_defer_hard_irqs: If not zero, provides a counter that would 2034 * allow to avoid NIC hard IRQ, on busy queues. 2035 * 2036 * @neighbours: List heads pointing to this device's neighbours' 2037 * dev_list, one per address-family. 2038 * 2039 * FIXME: cleanup struct net_device such that network protocol info 2040 * moves out. 2041 */ 2042 2043 struct net_device { 2044 /* Cacheline organization can be found documented in 2045 * Documentation/networking/net_cachelines/net_device.rst. 2046 * Please update the document when adding new fields. 2047 */ 2048 2049 /* TX read-mostly hotpath */ 2050 __cacheline_group_begin(net_device_read_tx); 2051 struct_group(priv_flags_fast, 2052 unsigned long priv_flags:32; 2053 unsigned long lltx:1; 2054 ); 2055 const struct net_device_ops *netdev_ops; 2056 const struct header_ops *header_ops; 2057 struct netdev_queue *_tx; 2058 netdev_features_t gso_partial_features; 2059 unsigned int real_num_tx_queues; 2060 unsigned int gso_max_size; 2061 unsigned int gso_ipv4_max_size; 2062 u16 gso_max_segs; 2063 s16 num_tc; 2064 /* Note : dev->mtu is often read without holding a lock. 2065 * Writers usually hold RTNL. 2066 * It is recommended to use READ_ONCE() to annotate the reads, 2067 * and to use WRITE_ONCE() to annotate the writes. 2068 */ 2069 unsigned int mtu; 2070 unsigned short needed_headroom; 2071 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2072 #ifdef CONFIG_XPS 2073 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2074 #endif 2075 #ifdef CONFIG_NETFILTER_EGRESS 2076 struct nf_hook_entries __rcu *nf_hooks_egress; 2077 #endif 2078 #ifdef CONFIG_NET_XGRESS 2079 struct bpf_mprog_entry __rcu *tcx_egress; 2080 #endif 2081 __cacheline_group_end(net_device_read_tx); 2082 2083 /* TXRX read-mostly hotpath */ 2084 __cacheline_group_begin(net_device_read_txrx); 2085 union { 2086 struct pcpu_lstats __percpu *lstats; 2087 struct pcpu_sw_netstats __percpu *tstats; 2088 struct pcpu_dstats __percpu *dstats; 2089 }; 2090 unsigned long state; 2091 unsigned int flags; 2092 unsigned short hard_header_len; 2093 netdev_features_t features; 2094 struct inet6_dev __rcu *ip6_ptr; 2095 __cacheline_group_end(net_device_read_txrx); 2096 2097 /* RX read-mostly hotpath */ 2098 __cacheline_group_begin(net_device_read_rx); 2099 struct bpf_prog __rcu *xdp_prog; 2100 struct list_head ptype_specific; 2101 int ifindex; 2102 unsigned int real_num_rx_queues; 2103 struct netdev_rx_queue *_rx; 2104 unsigned int gro_max_size; 2105 unsigned int gro_ipv4_max_size; 2106 rx_handler_func_t __rcu *rx_handler; 2107 void __rcu *rx_handler_data; 2108 possible_net_t nd_net; 2109 #ifdef CONFIG_NETPOLL 2110 struct netpoll_info __rcu *npinfo; 2111 #endif 2112 #ifdef CONFIG_NET_XGRESS 2113 struct bpf_mprog_entry __rcu *tcx_ingress; 2114 #endif 2115 __cacheline_group_end(net_device_read_rx); 2116 2117 char name[IFNAMSIZ]; 2118 struct netdev_name_node *name_node; 2119 struct dev_ifalias __rcu *ifalias; 2120 /* 2121 * I/O specific fields 2122 * FIXME: Merge these and struct ifmap into one 2123 */ 2124 unsigned long mem_end; 2125 unsigned long mem_start; 2126 unsigned long base_addr; 2127 2128 /* 2129 * Some hardware also needs these fields (state,dev_list, 2130 * napi_list,unreg_list,close_list) but they are not 2131 * part of the usual set specified in Space.c. 2132 */ 2133 2134 2135 struct list_head dev_list; 2136 struct list_head napi_list; 2137 struct list_head unreg_list; 2138 struct list_head close_list; 2139 struct list_head ptype_all; 2140 2141 struct { 2142 struct list_head upper; 2143 struct list_head lower; 2144 } adj_list; 2145 2146 /* Read-mostly cache-line for fast-path access */ 2147 xdp_features_t xdp_features; 2148 const struct xdp_metadata_ops *xdp_metadata_ops; 2149 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops; 2150 unsigned short gflags; 2151 2152 unsigned short needed_tailroom; 2153 2154 netdev_features_t hw_features; 2155 netdev_features_t wanted_features; 2156 netdev_features_t vlan_features; 2157 netdev_features_t hw_enc_features; 2158 netdev_features_t mpls_features; 2159 2160 unsigned int min_mtu; 2161 unsigned int max_mtu; 2162 unsigned short type; 2163 unsigned char min_header_len; 2164 unsigned char name_assign_type; 2165 2166 int group; 2167 2168 struct net_device_stats stats; /* not used by modern drivers */ 2169 2170 struct net_device_core_stats __percpu *core_stats; 2171 2172 /* Stats to monitor link on/off, flapping */ 2173 atomic_t carrier_up_count; 2174 atomic_t carrier_down_count; 2175 2176 #ifdef CONFIG_WIRELESS_EXT 2177 const struct iw_handler_def *wireless_handlers; 2178 #endif 2179 const struct ethtool_ops *ethtool_ops; 2180 #ifdef CONFIG_NET_L3_MASTER_DEV 2181 const struct l3mdev_ops *l3mdev_ops; 2182 #endif 2183 #if IS_ENABLED(CONFIG_IPV6) 2184 const struct ndisc_ops *ndisc_ops; 2185 #endif 2186 2187 #ifdef CONFIG_XFRM_OFFLOAD 2188 const struct xfrmdev_ops *xfrmdev_ops; 2189 #endif 2190 2191 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2192 const struct tlsdev_ops *tlsdev_ops; 2193 #endif 2194 2195 unsigned int operstate; 2196 unsigned char link_mode; 2197 2198 unsigned char if_port; 2199 unsigned char dma; 2200 2201 /* Interface address info. */ 2202 unsigned char perm_addr[MAX_ADDR_LEN]; 2203 unsigned char addr_assign_type; 2204 unsigned char addr_len; 2205 unsigned char upper_level; 2206 unsigned char lower_level; 2207 2208 unsigned short neigh_priv_len; 2209 unsigned short dev_id; 2210 unsigned short dev_port; 2211 int irq; 2212 u32 priv_len; 2213 2214 spinlock_t addr_list_lock; 2215 2216 struct netdev_hw_addr_list uc; 2217 struct netdev_hw_addr_list mc; 2218 struct netdev_hw_addr_list dev_addrs; 2219 2220 #ifdef CONFIG_SYSFS 2221 struct kset *queues_kset; 2222 #endif 2223 #ifdef CONFIG_LOCKDEP 2224 struct list_head unlink_list; 2225 #endif 2226 unsigned int promiscuity; 2227 unsigned int allmulti; 2228 bool uc_promisc; 2229 #ifdef CONFIG_LOCKDEP 2230 unsigned char nested_level; 2231 #endif 2232 2233 2234 /* Protocol-specific pointers */ 2235 struct in_device __rcu *ip_ptr; 2236 /** @fib_nh_head: nexthops associated with this netdev */ 2237 struct hlist_head fib_nh_head; 2238 2239 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2240 struct vlan_info __rcu *vlan_info; 2241 #endif 2242 #if IS_ENABLED(CONFIG_NET_DSA) 2243 struct dsa_port *dsa_ptr; 2244 #endif 2245 #if IS_ENABLED(CONFIG_TIPC) 2246 struct tipc_bearer __rcu *tipc_ptr; 2247 #endif 2248 #if IS_ENABLED(CONFIG_ATALK) 2249 void *atalk_ptr; 2250 #endif 2251 #if IS_ENABLED(CONFIG_AX25) 2252 void *ax25_ptr; 2253 #endif 2254 #if IS_ENABLED(CONFIG_CFG80211) 2255 struct wireless_dev *ieee80211_ptr; 2256 #endif 2257 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2258 struct wpan_dev *ieee802154_ptr; 2259 #endif 2260 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2261 struct mpls_dev __rcu *mpls_ptr; 2262 #endif 2263 #if IS_ENABLED(CONFIG_MCTP) 2264 struct mctp_dev __rcu *mctp_ptr; 2265 #endif 2266 2267 /* 2268 * Cache lines mostly used on receive path (including eth_type_trans()) 2269 */ 2270 /* Interface address info used in eth_type_trans() */ 2271 const unsigned char *dev_addr; 2272 2273 unsigned int num_rx_queues; 2274 #define GRO_LEGACY_MAX_SIZE 65536u 2275 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2276 * and shinfo->gso_segs is a 16bit field. 2277 */ 2278 #define GRO_MAX_SIZE (8 * 65535u) 2279 unsigned int xdp_zc_max_segs; 2280 struct netdev_queue __rcu *ingress_queue; 2281 #ifdef CONFIG_NETFILTER_INGRESS 2282 struct nf_hook_entries __rcu *nf_hooks_ingress; 2283 #endif 2284 2285 unsigned char broadcast[MAX_ADDR_LEN]; 2286 #ifdef CONFIG_RFS_ACCEL 2287 struct cpu_rmap *rx_cpu_rmap; 2288 #endif 2289 struct hlist_node index_hlist; 2290 2291 /* 2292 * Cache lines mostly used on transmit path 2293 */ 2294 unsigned int num_tx_queues; 2295 struct Qdisc __rcu *qdisc; 2296 unsigned int tx_queue_len; 2297 spinlock_t tx_global_lock; 2298 2299 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2300 2301 #ifdef CONFIG_NET_SCHED 2302 DECLARE_HASHTABLE (qdisc_hash, 4); 2303 #endif 2304 /* These may be needed for future network-power-down code. */ 2305 struct timer_list watchdog_timer; 2306 int watchdog_timeo; 2307 2308 u32 proto_down_reason; 2309 2310 struct list_head todo_list; 2311 2312 #ifdef CONFIG_PCPU_DEV_REFCNT 2313 int __percpu *pcpu_refcnt; 2314 #else 2315 refcount_t dev_refcnt; 2316 #endif 2317 struct ref_tracker_dir refcnt_tracker; 2318 2319 struct list_head link_watch_list; 2320 2321 u8 reg_state; 2322 2323 bool dismantle; 2324 2325 enum { 2326 RTNL_LINK_INITIALIZED, 2327 RTNL_LINK_INITIALIZING, 2328 } rtnl_link_state:16; 2329 2330 bool needs_free_netdev; 2331 void (*priv_destructor)(struct net_device *dev); 2332 2333 /* mid-layer private */ 2334 void *ml_priv; 2335 enum netdev_ml_priv_type ml_priv_type; 2336 2337 enum netdev_stat_type pcpu_stat_type:8; 2338 2339 #if IS_ENABLED(CONFIG_GARP) 2340 struct garp_port __rcu *garp_port; 2341 #endif 2342 #if IS_ENABLED(CONFIG_MRP) 2343 struct mrp_port __rcu *mrp_port; 2344 #endif 2345 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2346 struct dm_hw_stat_delta __rcu *dm_private; 2347 #endif 2348 struct device dev; 2349 const struct attribute_group *sysfs_groups[4]; 2350 const struct attribute_group *sysfs_rx_queue_group; 2351 2352 const struct rtnl_link_ops *rtnl_link_ops; 2353 2354 const struct netdev_stat_ops *stat_ops; 2355 2356 const struct netdev_queue_mgmt_ops *queue_mgmt_ops; 2357 2358 /* for setting kernel sock attribute on TCP connection setup */ 2359 #define GSO_MAX_SEGS 65535u 2360 #define GSO_LEGACY_MAX_SIZE 65536u 2361 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2362 * and shinfo->gso_segs is a 16bit field. 2363 */ 2364 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2365 2366 #define TSO_LEGACY_MAX_SIZE 65536 2367 #define TSO_MAX_SIZE UINT_MAX 2368 unsigned int tso_max_size; 2369 #define TSO_MAX_SEGS U16_MAX 2370 u16 tso_max_segs; 2371 2372 #ifdef CONFIG_DCB 2373 const struct dcbnl_rtnl_ops *dcbnl_ops; 2374 #endif 2375 u8 prio_tc_map[TC_BITMASK + 1]; 2376 2377 #if IS_ENABLED(CONFIG_FCOE) 2378 unsigned int fcoe_ddp_xid; 2379 #endif 2380 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2381 struct netprio_map __rcu *priomap; 2382 #endif 2383 struct phy_link_topology *link_topo; 2384 struct phy_device *phydev; 2385 struct sfp_bus *sfp_bus; 2386 struct lock_class_key *qdisc_tx_busylock; 2387 bool proto_down; 2388 bool threaded; 2389 2390 /* priv_flags_slow, ungrouped to save space */ 2391 unsigned long see_all_hwtstamp_requests:1; 2392 unsigned long change_proto_down:1; 2393 unsigned long netns_local:1; 2394 unsigned long fcoe_mtu:1; 2395 2396 struct list_head net_notifier_list; 2397 2398 #if IS_ENABLED(CONFIG_MACSEC) 2399 /* MACsec management functions */ 2400 const struct macsec_ops *macsec_ops; 2401 #endif 2402 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2403 struct udp_tunnel_nic *udp_tunnel_nic; 2404 2405 struct ethtool_netdev_state *ethtool; 2406 2407 /* protected by rtnl_lock */ 2408 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2409 2410 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2411 netdevice_tracker linkwatch_dev_tracker; 2412 netdevice_tracker watchdog_dev_tracker; 2413 netdevice_tracker dev_registered_tracker; 2414 struct rtnl_hw_stats64 *offload_xstats_l3; 2415 2416 struct devlink_port *devlink_port; 2417 2418 #if IS_ENABLED(CONFIG_DPLL) 2419 struct dpll_pin __rcu *dpll_pin; 2420 #endif 2421 #if IS_ENABLED(CONFIG_PAGE_POOL) 2422 /** @page_pools: page pools created for this netdevice */ 2423 struct hlist_head page_pools; 2424 #endif 2425 2426 /** @irq_moder: dim parameters used if IS_ENABLED(CONFIG_DIMLIB). */ 2427 struct dim_irq_moder *irq_moder; 2428 2429 u64 max_pacing_offload_horizon; 2430 struct napi_config *napi_config; 2431 unsigned long gro_flush_timeout; 2432 u32 napi_defer_hard_irqs; 2433 2434 /** 2435 * @lock: protects @net_shaper_hierarchy, feel free to use for other 2436 * netdev-scope protection. Ordering: take after rtnl_lock. 2437 */ 2438 struct mutex lock; 2439 2440 #if IS_ENABLED(CONFIG_NET_SHAPER) 2441 /** 2442 * @net_shaper_hierarchy: data tracking the current shaper status 2443 * see include/net/net_shapers.h 2444 */ 2445 struct net_shaper_hierarchy *net_shaper_hierarchy; 2446 #endif 2447 2448 struct hlist_head neighbours[NEIGH_NR_TABLES]; 2449 2450 u8 priv[] ____cacheline_aligned 2451 __counted_by(priv_len); 2452 } ____cacheline_aligned; 2453 #define to_net_dev(d) container_of(d, struct net_device, dev) 2454 2455 /* 2456 * Driver should use this to assign devlink port instance to a netdevice 2457 * before it registers the netdevice. Therefore devlink_port is static 2458 * during the netdev lifetime after it is registered. 2459 */ 2460 #define SET_NETDEV_DEVLINK_PORT(dev, port) \ 2461 ({ \ 2462 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \ 2463 ((dev)->devlink_port = (port)); \ 2464 }) 2465 2466 static inline bool netif_elide_gro(const struct net_device *dev) 2467 { 2468 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2469 return true; 2470 return false; 2471 } 2472 2473 #define NETDEV_ALIGN 32 2474 2475 static inline 2476 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2477 { 2478 return dev->prio_tc_map[prio & TC_BITMASK]; 2479 } 2480 2481 static inline 2482 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2483 { 2484 if (tc >= dev->num_tc) 2485 return -EINVAL; 2486 2487 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2488 return 0; 2489 } 2490 2491 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2492 void netdev_reset_tc(struct net_device *dev); 2493 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2494 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2495 2496 static inline 2497 int netdev_get_num_tc(struct net_device *dev) 2498 { 2499 return dev->num_tc; 2500 } 2501 2502 static inline void net_prefetch(void *p) 2503 { 2504 prefetch(p); 2505 #if L1_CACHE_BYTES < 128 2506 prefetch((u8 *)p + L1_CACHE_BYTES); 2507 #endif 2508 } 2509 2510 static inline void net_prefetchw(void *p) 2511 { 2512 prefetchw(p); 2513 #if L1_CACHE_BYTES < 128 2514 prefetchw((u8 *)p + L1_CACHE_BYTES); 2515 #endif 2516 } 2517 2518 void netdev_unbind_sb_channel(struct net_device *dev, 2519 struct net_device *sb_dev); 2520 int netdev_bind_sb_channel_queue(struct net_device *dev, 2521 struct net_device *sb_dev, 2522 u8 tc, u16 count, u16 offset); 2523 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2524 static inline int netdev_get_sb_channel(struct net_device *dev) 2525 { 2526 return max_t(int, -dev->num_tc, 0); 2527 } 2528 2529 static inline 2530 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2531 unsigned int index) 2532 { 2533 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues); 2534 return &dev->_tx[index]; 2535 } 2536 2537 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2538 const struct sk_buff *skb) 2539 { 2540 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2541 } 2542 2543 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2544 void (*f)(struct net_device *, 2545 struct netdev_queue *, 2546 void *), 2547 void *arg) 2548 { 2549 unsigned int i; 2550 2551 for (i = 0; i < dev->num_tx_queues; i++) 2552 f(dev, &dev->_tx[i], arg); 2553 } 2554 2555 #define netdev_lockdep_set_classes(dev) \ 2556 { \ 2557 static struct lock_class_key qdisc_tx_busylock_key; \ 2558 static struct lock_class_key qdisc_xmit_lock_key; \ 2559 static struct lock_class_key dev_addr_list_lock_key; \ 2560 unsigned int i; \ 2561 \ 2562 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2563 lockdep_set_class(&(dev)->addr_list_lock, \ 2564 &dev_addr_list_lock_key); \ 2565 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2566 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2567 &qdisc_xmit_lock_key); \ 2568 } 2569 2570 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2571 struct net_device *sb_dev); 2572 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2573 struct sk_buff *skb, 2574 struct net_device *sb_dev); 2575 2576 /* returns the headroom that the master device needs to take in account 2577 * when forwarding to this dev 2578 */ 2579 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2580 { 2581 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2582 } 2583 2584 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2585 { 2586 if (dev->netdev_ops->ndo_set_rx_headroom) 2587 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2588 } 2589 2590 /* set the device rx headroom to the dev's default */ 2591 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2592 { 2593 netdev_set_rx_headroom(dev, -1); 2594 } 2595 2596 static inline void *netdev_get_ml_priv(struct net_device *dev, 2597 enum netdev_ml_priv_type type) 2598 { 2599 if (dev->ml_priv_type != type) 2600 return NULL; 2601 2602 return dev->ml_priv; 2603 } 2604 2605 static inline void netdev_set_ml_priv(struct net_device *dev, 2606 void *ml_priv, 2607 enum netdev_ml_priv_type type) 2608 { 2609 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2610 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2611 dev->ml_priv_type, type); 2612 WARN(!dev->ml_priv_type && dev->ml_priv, 2613 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2614 2615 dev->ml_priv = ml_priv; 2616 dev->ml_priv_type = type; 2617 } 2618 2619 /* 2620 * Net namespace inlines 2621 */ 2622 static inline 2623 struct net *dev_net(const struct net_device *dev) 2624 { 2625 return read_pnet(&dev->nd_net); 2626 } 2627 2628 static inline 2629 void dev_net_set(struct net_device *dev, struct net *net) 2630 { 2631 write_pnet(&dev->nd_net, net); 2632 } 2633 2634 /** 2635 * netdev_priv - access network device private data 2636 * @dev: network device 2637 * 2638 * Get network device private data 2639 */ 2640 static inline void *netdev_priv(const struct net_device *dev) 2641 { 2642 return (void *)dev->priv; 2643 } 2644 2645 /* Set the sysfs physical device reference for the network logical device 2646 * if set prior to registration will cause a symlink during initialization. 2647 */ 2648 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2649 2650 /* Set the sysfs device type for the network logical device to allow 2651 * fine-grained identification of different network device types. For 2652 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2653 */ 2654 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2655 2656 void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index, 2657 enum netdev_queue_type type, 2658 struct napi_struct *napi); 2659 2660 static inline void netif_napi_set_irq(struct napi_struct *napi, int irq) 2661 { 2662 napi->irq = irq; 2663 } 2664 2665 /* Default NAPI poll() weight 2666 * Device drivers are strongly advised to not use bigger value 2667 */ 2668 #define NAPI_POLL_WEIGHT 64 2669 2670 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2671 int (*poll)(struct napi_struct *, int), int weight); 2672 2673 /** 2674 * netif_napi_add() - initialize a NAPI context 2675 * @dev: network device 2676 * @napi: NAPI context 2677 * @poll: polling function 2678 * 2679 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2680 * *any* of the other NAPI-related functions. 2681 */ 2682 static inline void 2683 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2684 int (*poll)(struct napi_struct *, int)) 2685 { 2686 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2687 } 2688 2689 static inline void 2690 netif_napi_add_tx_weight(struct net_device *dev, 2691 struct napi_struct *napi, 2692 int (*poll)(struct napi_struct *, int), 2693 int weight) 2694 { 2695 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2696 netif_napi_add_weight(dev, napi, poll, weight); 2697 } 2698 2699 /** 2700 * netif_napi_add_config - initialize a NAPI context with persistent config 2701 * @dev: network device 2702 * @napi: NAPI context 2703 * @poll: polling function 2704 * @index: the NAPI index 2705 */ 2706 static inline void 2707 netif_napi_add_config(struct net_device *dev, struct napi_struct *napi, 2708 int (*poll)(struct napi_struct *, int), int index) 2709 { 2710 napi->index = index; 2711 napi->config = &dev->napi_config[index]; 2712 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2713 } 2714 2715 /** 2716 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2717 * @dev: network device 2718 * @napi: NAPI context 2719 * @poll: polling function 2720 * 2721 * This variant of netif_napi_add() should be used from drivers using NAPI 2722 * to exclusively poll a TX queue. 2723 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2724 */ 2725 static inline void netif_napi_add_tx(struct net_device *dev, 2726 struct napi_struct *napi, 2727 int (*poll)(struct napi_struct *, int)) 2728 { 2729 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2730 } 2731 2732 /** 2733 * __netif_napi_del - remove a NAPI context 2734 * @napi: NAPI context 2735 * 2736 * Warning: caller must observe RCU grace period before freeing memory 2737 * containing @napi. Drivers might want to call this helper to combine 2738 * all the needed RCU grace periods into a single one. 2739 */ 2740 void __netif_napi_del(struct napi_struct *napi); 2741 2742 /** 2743 * netif_napi_del - remove a NAPI context 2744 * @napi: NAPI context 2745 * 2746 * netif_napi_del() removes a NAPI context from the network device NAPI list 2747 */ 2748 static inline void netif_napi_del(struct napi_struct *napi) 2749 { 2750 __netif_napi_del(napi); 2751 synchronize_net(); 2752 } 2753 2754 struct packet_type { 2755 __be16 type; /* This is really htons(ether_type). */ 2756 bool ignore_outgoing; 2757 struct net_device *dev; /* NULL is wildcarded here */ 2758 netdevice_tracker dev_tracker; 2759 int (*func) (struct sk_buff *, 2760 struct net_device *, 2761 struct packet_type *, 2762 struct net_device *); 2763 void (*list_func) (struct list_head *, 2764 struct packet_type *, 2765 struct net_device *); 2766 bool (*id_match)(struct packet_type *ptype, 2767 struct sock *sk); 2768 struct net *af_packet_net; 2769 void *af_packet_priv; 2770 struct list_head list; 2771 }; 2772 2773 struct offload_callbacks { 2774 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2775 netdev_features_t features); 2776 struct sk_buff *(*gro_receive)(struct list_head *head, 2777 struct sk_buff *skb); 2778 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2779 }; 2780 2781 struct packet_offload { 2782 __be16 type; /* This is really htons(ether_type). */ 2783 u16 priority; 2784 struct offload_callbacks callbacks; 2785 struct list_head list; 2786 }; 2787 2788 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2789 struct pcpu_sw_netstats { 2790 u64_stats_t rx_packets; 2791 u64_stats_t rx_bytes; 2792 u64_stats_t tx_packets; 2793 u64_stats_t tx_bytes; 2794 struct u64_stats_sync syncp; 2795 } __aligned(4 * sizeof(u64)); 2796 2797 struct pcpu_dstats { 2798 u64_stats_t rx_packets; 2799 u64_stats_t rx_bytes; 2800 u64_stats_t rx_drops; 2801 u64_stats_t tx_packets; 2802 u64_stats_t tx_bytes; 2803 u64_stats_t tx_drops; 2804 struct u64_stats_sync syncp; 2805 } __aligned(8 * sizeof(u64)); 2806 2807 struct pcpu_lstats { 2808 u64_stats_t packets; 2809 u64_stats_t bytes; 2810 struct u64_stats_sync syncp; 2811 } __aligned(2 * sizeof(u64)); 2812 2813 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2814 2815 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2816 { 2817 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2818 2819 u64_stats_update_begin(&tstats->syncp); 2820 u64_stats_add(&tstats->rx_bytes, len); 2821 u64_stats_inc(&tstats->rx_packets); 2822 u64_stats_update_end(&tstats->syncp); 2823 } 2824 2825 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2826 unsigned int packets, 2827 unsigned int len) 2828 { 2829 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2830 2831 u64_stats_update_begin(&tstats->syncp); 2832 u64_stats_add(&tstats->tx_bytes, len); 2833 u64_stats_add(&tstats->tx_packets, packets); 2834 u64_stats_update_end(&tstats->syncp); 2835 } 2836 2837 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2838 { 2839 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2840 2841 u64_stats_update_begin(&lstats->syncp); 2842 u64_stats_add(&lstats->bytes, len); 2843 u64_stats_inc(&lstats->packets); 2844 u64_stats_update_end(&lstats->syncp); 2845 } 2846 2847 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2848 ({ \ 2849 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2850 if (pcpu_stats) { \ 2851 int __cpu; \ 2852 for_each_possible_cpu(__cpu) { \ 2853 typeof(type) *stat; \ 2854 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2855 u64_stats_init(&stat->syncp); \ 2856 } \ 2857 } \ 2858 pcpu_stats; \ 2859 }) 2860 2861 #define netdev_alloc_pcpu_stats(type) \ 2862 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2863 2864 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2865 ({ \ 2866 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2867 if (pcpu_stats) { \ 2868 int __cpu; \ 2869 for_each_possible_cpu(__cpu) { \ 2870 typeof(type) *stat; \ 2871 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2872 u64_stats_init(&stat->syncp); \ 2873 } \ 2874 } \ 2875 pcpu_stats; \ 2876 }) 2877 2878 enum netdev_lag_tx_type { 2879 NETDEV_LAG_TX_TYPE_UNKNOWN, 2880 NETDEV_LAG_TX_TYPE_RANDOM, 2881 NETDEV_LAG_TX_TYPE_BROADCAST, 2882 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2883 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2884 NETDEV_LAG_TX_TYPE_HASH, 2885 }; 2886 2887 enum netdev_lag_hash { 2888 NETDEV_LAG_HASH_NONE, 2889 NETDEV_LAG_HASH_L2, 2890 NETDEV_LAG_HASH_L34, 2891 NETDEV_LAG_HASH_L23, 2892 NETDEV_LAG_HASH_E23, 2893 NETDEV_LAG_HASH_E34, 2894 NETDEV_LAG_HASH_VLAN_SRCMAC, 2895 NETDEV_LAG_HASH_UNKNOWN, 2896 }; 2897 2898 struct netdev_lag_upper_info { 2899 enum netdev_lag_tx_type tx_type; 2900 enum netdev_lag_hash hash_type; 2901 }; 2902 2903 struct netdev_lag_lower_state_info { 2904 u8 link_up : 1, 2905 tx_enabled : 1; 2906 }; 2907 2908 #include <linux/notifier.h> 2909 2910 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2911 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2912 * adding new types. 2913 */ 2914 enum netdev_cmd { 2915 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2916 NETDEV_DOWN, 2917 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2918 detected a hardware crash and restarted 2919 - we can use this eg to kick tcp sessions 2920 once done */ 2921 NETDEV_CHANGE, /* Notify device state change */ 2922 NETDEV_REGISTER, 2923 NETDEV_UNREGISTER, 2924 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2925 NETDEV_CHANGEADDR, /* notify after the address change */ 2926 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2927 NETDEV_GOING_DOWN, 2928 NETDEV_CHANGENAME, 2929 NETDEV_FEAT_CHANGE, 2930 NETDEV_BONDING_FAILOVER, 2931 NETDEV_PRE_UP, 2932 NETDEV_PRE_TYPE_CHANGE, 2933 NETDEV_POST_TYPE_CHANGE, 2934 NETDEV_POST_INIT, 2935 NETDEV_PRE_UNINIT, 2936 NETDEV_RELEASE, 2937 NETDEV_NOTIFY_PEERS, 2938 NETDEV_JOIN, 2939 NETDEV_CHANGEUPPER, 2940 NETDEV_RESEND_IGMP, 2941 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2942 NETDEV_CHANGEINFODATA, 2943 NETDEV_BONDING_INFO, 2944 NETDEV_PRECHANGEUPPER, 2945 NETDEV_CHANGELOWERSTATE, 2946 NETDEV_UDP_TUNNEL_PUSH_INFO, 2947 NETDEV_UDP_TUNNEL_DROP_INFO, 2948 NETDEV_CHANGE_TX_QUEUE_LEN, 2949 NETDEV_CVLAN_FILTER_PUSH_INFO, 2950 NETDEV_CVLAN_FILTER_DROP_INFO, 2951 NETDEV_SVLAN_FILTER_PUSH_INFO, 2952 NETDEV_SVLAN_FILTER_DROP_INFO, 2953 NETDEV_OFFLOAD_XSTATS_ENABLE, 2954 NETDEV_OFFLOAD_XSTATS_DISABLE, 2955 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 2956 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 2957 NETDEV_XDP_FEAT_CHANGE, 2958 }; 2959 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2960 2961 int register_netdevice_notifier(struct notifier_block *nb); 2962 int unregister_netdevice_notifier(struct notifier_block *nb); 2963 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2964 int unregister_netdevice_notifier_net(struct net *net, 2965 struct notifier_block *nb); 2966 int register_netdevice_notifier_dev_net(struct net_device *dev, 2967 struct notifier_block *nb, 2968 struct netdev_net_notifier *nn); 2969 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2970 struct notifier_block *nb, 2971 struct netdev_net_notifier *nn); 2972 2973 struct netdev_notifier_info { 2974 struct net_device *dev; 2975 struct netlink_ext_ack *extack; 2976 }; 2977 2978 struct netdev_notifier_info_ext { 2979 struct netdev_notifier_info info; /* must be first */ 2980 union { 2981 u32 mtu; 2982 } ext; 2983 }; 2984 2985 struct netdev_notifier_change_info { 2986 struct netdev_notifier_info info; /* must be first */ 2987 unsigned int flags_changed; 2988 }; 2989 2990 struct netdev_notifier_changeupper_info { 2991 struct netdev_notifier_info info; /* must be first */ 2992 struct net_device *upper_dev; /* new upper dev */ 2993 bool master; /* is upper dev master */ 2994 bool linking; /* is the notification for link or unlink */ 2995 void *upper_info; /* upper dev info */ 2996 }; 2997 2998 struct netdev_notifier_changelowerstate_info { 2999 struct netdev_notifier_info info; /* must be first */ 3000 void *lower_state_info; /* is lower dev state */ 3001 }; 3002 3003 struct netdev_notifier_pre_changeaddr_info { 3004 struct netdev_notifier_info info; /* must be first */ 3005 const unsigned char *dev_addr; 3006 }; 3007 3008 enum netdev_offload_xstats_type { 3009 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 3010 }; 3011 3012 struct netdev_notifier_offload_xstats_info { 3013 struct netdev_notifier_info info; /* must be first */ 3014 enum netdev_offload_xstats_type type; 3015 3016 union { 3017 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 3018 struct netdev_notifier_offload_xstats_rd *report_delta; 3019 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 3020 struct netdev_notifier_offload_xstats_ru *report_used; 3021 }; 3022 }; 3023 3024 int netdev_offload_xstats_enable(struct net_device *dev, 3025 enum netdev_offload_xstats_type type, 3026 struct netlink_ext_ack *extack); 3027 int netdev_offload_xstats_disable(struct net_device *dev, 3028 enum netdev_offload_xstats_type type); 3029 bool netdev_offload_xstats_enabled(const struct net_device *dev, 3030 enum netdev_offload_xstats_type type); 3031 int netdev_offload_xstats_get(struct net_device *dev, 3032 enum netdev_offload_xstats_type type, 3033 struct rtnl_hw_stats64 *stats, bool *used, 3034 struct netlink_ext_ack *extack); 3035 void 3036 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 3037 const struct rtnl_hw_stats64 *stats); 3038 void 3039 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 3040 void netdev_offload_xstats_push_delta(struct net_device *dev, 3041 enum netdev_offload_xstats_type type, 3042 const struct rtnl_hw_stats64 *stats); 3043 3044 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 3045 struct net_device *dev) 3046 { 3047 info->dev = dev; 3048 info->extack = NULL; 3049 } 3050 3051 static inline struct net_device * 3052 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 3053 { 3054 return info->dev; 3055 } 3056 3057 static inline struct netlink_ext_ack * 3058 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 3059 { 3060 return info->extack; 3061 } 3062 3063 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 3064 int call_netdevice_notifiers_info(unsigned long val, 3065 struct netdev_notifier_info *info); 3066 3067 #define for_each_netdev(net, d) \ 3068 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 3069 #define for_each_netdev_reverse(net, d) \ 3070 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 3071 #define for_each_netdev_rcu(net, d) \ 3072 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 3073 #define for_each_netdev_safe(net, d, n) \ 3074 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 3075 #define for_each_netdev_continue(net, d) \ 3076 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 3077 #define for_each_netdev_continue_reverse(net, d) \ 3078 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 3079 dev_list) 3080 #define for_each_netdev_continue_rcu(net, d) \ 3081 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 3082 #define for_each_netdev_in_bond_rcu(bond, slave) \ 3083 for_each_netdev_rcu(&init_net, slave) \ 3084 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 3085 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 3086 3087 #define for_each_netdev_dump(net, d, ifindex) \ 3088 for (; (d = xa_find(&(net)->dev_by_index, &ifindex, \ 3089 ULONG_MAX, XA_PRESENT)); ifindex++) 3090 3091 static inline struct net_device *next_net_device(struct net_device *dev) 3092 { 3093 struct list_head *lh; 3094 struct net *net; 3095 3096 net = dev_net(dev); 3097 lh = dev->dev_list.next; 3098 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3099 } 3100 3101 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 3102 { 3103 struct list_head *lh; 3104 struct net *net; 3105 3106 net = dev_net(dev); 3107 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 3108 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3109 } 3110 3111 static inline struct net_device *first_net_device(struct net *net) 3112 { 3113 return list_empty(&net->dev_base_head) ? NULL : 3114 net_device_entry(net->dev_base_head.next); 3115 } 3116 3117 static inline struct net_device *first_net_device_rcu(struct net *net) 3118 { 3119 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 3120 3121 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3122 } 3123 3124 int netdev_boot_setup_check(struct net_device *dev); 3125 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 3126 const char *hwaddr); 3127 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 3128 void dev_add_pack(struct packet_type *pt); 3129 void dev_remove_pack(struct packet_type *pt); 3130 void __dev_remove_pack(struct packet_type *pt); 3131 void dev_add_offload(struct packet_offload *po); 3132 void dev_remove_offload(struct packet_offload *po); 3133 3134 int dev_get_iflink(const struct net_device *dev); 3135 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 3136 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 3137 struct net_device_path_stack *stack); 3138 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 3139 unsigned short mask); 3140 struct net_device *dev_get_by_name(struct net *net, const char *name); 3141 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 3142 struct net_device *__dev_get_by_name(struct net *net, const char *name); 3143 bool netdev_name_in_use(struct net *net, const char *name); 3144 int dev_alloc_name(struct net_device *dev, const char *name); 3145 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 3146 void dev_close(struct net_device *dev); 3147 void dev_close_many(struct list_head *head, bool unlink); 3148 void dev_disable_lro(struct net_device *dev); 3149 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 3150 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 3151 struct net_device *sb_dev); 3152 3153 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3154 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3155 3156 static inline int dev_queue_xmit(struct sk_buff *skb) 3157 { 3158 return __dev_queue_xmit(skb, NULL); 3159 } 3160 3161 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3162 struct net_device *sb_dev) 3163 { 3164 return __dev_queue_xmit(skb, sb_dev); 3165 } 3166 3167 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3168 { 3169 int ret; 3170 3171 ret = __dev_direct_xmit(skb, queue_id); 3172 if (!dev_xmit_complete(ret)) 3173 kfree_skb(skb); 3174 return ret; 3175 } 3176 3177 int register_netdevice(struct net_device *dev); 3178 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3179 void unregister_netdevice_many(struct list_head *head); 3180 static inline void unregister_netdevice(struct net_device *dev) 3181 { 3182 unregister_netdevice_queue(dev, NULL); 3183 } 3184 3185 int netdev_refcnt_read(const struct net_device *dev); 3186 void free_netdev(struct net_device *dev); 3187 void init_dummy_netdev(struct net_device *dev); 3188 3189 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3190 struct sk_buff *skb, 3191 bool all_slaves); 3192 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3193 struct sock *sk); 3194 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3195 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3196 struct net_device *netdev_get_by_index(struct net *net, int ifindex, 3197 netdevice_tracker *tracker, gfp_t gfp); 3198 struct net_device *netdev_get_by_name(struct net *net, const char *name, 3199 netdevice_tracker *tracker, gfp_t gfp); 3200 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3201 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 3202 void netdev_copy_name(struct net_device *dev, char *name); 3203 3204 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3205 unsigned short type, 3206 const void *daddr, const void *saddr, 3207 unsigned int len) 3208 { 3209 if (!dev->header_ops || !dev->header_ops->create) 3210 return 0; 3211 3212 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3213 } 3214 3215 static inline int dev_parse_header(const struct sk_buff *skb, 3216 unsigned char *haddr) 3217 { 3218 const struct net_device *dev = skb->dev; 3219 3220 if (!dev->header_ops || !dev->header_ops->parse) 3221 return 0; 3222 return dev->header_ops->parse(skb, haddr); 3223 } 3224 3225 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3226 { 3227 const struct net_device *dev = skb->dev; 3228 3229 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3230 return 0; 3231 return dev->header_ops->parse_protocol(skb); 3232 } 3233 3234 /* ll_header must have at least hard_header_len allocated */ 3235 static inline bool dev_validate_header(const struct net_device *dev, 3236 char *ll_header, int len) 3237 { 3238 if (likely(len >= dev->hard_header_len)) 3239 return true; 3240 if (len < dev->min_header_len) 3241 return false; 3242 3243 if (capable(CAP_SYS_RAWIO)) { 3244 memset(ll_header + len, 0, dev->hard_header_len - len); 3245 return true; 3246 } 3247 3248 if (dev->header_ops && dev->header_ops->validate) 3249 return dev->header_ops->validate(ll_header, len); 3250 3251 return false; 3252 } 3253 3254 static inline bool dev_has_header(const struct net_device *dev) 3255 { 3256 return dev->header_ops && dev->header_ops->create; 3257 } 3258 3259 /* 3260 * Incoming packets are placed on per-CPU queues 3261 */ 3262 struct softnet_data { 3263 struct list_head poll_list; 3264 struct sk_buff_head process_queue; 3265 local_lock_t process_queue_bh_lock; 3266 3267 /* stats */ 3268 unsigned int processed; 3269 unsigned int time_squeeze; 3270 #ifdef CONFIG_RPS 3271 struct softnet_data *rps_ipi_list; 3272 #endif 3273 3274 unsigned int received_rps; 3275 bool in_net_rx_action; 3276 bool in_napi_threaded_poll; 3277 3278 #ifdef CONFIG_NET_FLOW_LIMIT 3279 struct sd_flow_limit __rcu *flow_limit; 3280 #endif 3281 struct Qdisc *output_queue; 3282 struct Qdisc **output_queue_tailp; 3283 struct sk_buff *completion_queue; 3284 #ifdef CONFIG_XFRM_OFFLOAD 3285 struct sk_buff_head xfrm_backlog; 3286 #endif 3287 /* written and read only by owning cpu: */ 3288 struct netdev_xmit xmit; 3289 #ifdef CONFIG_RPS 3290 /* input_queue_head should be written by cpu owning this struct, 3291 * and only read by other cpus. Worth using a cache line. 3292 */ 3293 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3294 3295 /* Elements below can be accessed between CPUs for RPS/RFS */ 3296 call_single_data_t csd ____cacheline_aligned_in_smp; 3297 struct softnet_data *rps_ipi_next; 3298 unsigned int cpu; 3299 unsigned int input_queue_tail; 3300 #endif 3301 struct sk_buff_head input_pkt_queue; 3302 struct napi_struct backlog; 3303 3304 atomic_t dropped ____cacheline_aligned_in_smp; 3305 3306 /* Another possibly contended cache line */ 3307 spinlock_t defer_lock ____cacheline_aligned_in_smp; 3308 int defer_count; 3309 int defer_ipi_scheduled; 3310 struct sk_buff *defer_list; 3311 call_single_data_t defer_csd; 3312 }; 3313 3314 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3315 3316 #ifndef CONFIG_PREEMPT_RT 3317 static inline int dev_recursion_level(void) 3318 { 3319 return this_cpu_read(softnet_data.xmit.recursion); 3320 } 3321 #else 3322 static inline int dev_recursion_level(void) 3323 { 3324 return current->net_xmit.recursion; 3325 } 3326 3327 #endif 3328 3329 void __netif_schedule(struct Qdisc *q); 3330 void netif_schedule_queue(struct netdev_queue *txq); 3331 3332 static inline void netif_tx_schedule_all(struct net_device *dev) 3333 { 3334 unsigned int i; 3335 3336 for (i = 0; i < dev->num_tx_queues; i++) 3337 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3338 } 3339 3340 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3341 { 3342 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3343 } 3344 3345 /** 3346 * netif_start_queue - allow transmit 3347 * @dev: network device 3348 * 3349 * Allow upper layers to call the device hard_start_xmit routine. 3350 */ 3351 static inline void netif_start_queue(struct net_device *dev) 3352 { 3353 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3354 } 3355 3356 static inline void netif_tx_start_all_queues(struct net_device *dev) 3357 { 3358 unsigned int i; 3359 3360 for (i = 0; i < dev->num_tx_queues; i++) { 3361 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3362 netif_tx_start_queue(txq); 3363 } 3364 } 3365 3366 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3367 3368 /** 3369 * netif_wake_queue - restart transmit 3370 * @dev: network device 3371 * 3372 * Allow upper layers to call the device hard_start_xmit routine. 3373 * Used for flow control when transmit resources are available. 3374 */ 3375 static inline void netif_wake_queue(struct net_device *dev) 3376 { 3377 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3378 } 3379 3380 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3381 { 3382 unsigned int i; 3383 3384 for (i = 0; i < dev->num_tx_queues; i++) { 3385 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3386 netif_tx_wake_queue(txq); 3387 } 3388 } 3389 3390 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3391 { 3392 /* Paired with READ_ONCE() from dev_watchdog() */ 3393 WRITE_ONCE(dev_queue->trans_start, jiffies); 3394 3395 /* This barrier is paired with smp_mb() from dev_watchdog() */ 3396 smp_mb__before_atomic(); 3397 3398 /* Must be an atomic op see netif_txq_try_stop() */ 3399 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3400 } 3401 3402 /** 3403 * netif_stop_queue - stop transmitted packets 3404 * @dev: network device 3405 * 3406 * Stop upper layers calling the device hard_start_xmit routine. 3407 * Used for flow control when transmit resources are unavailable. 3408 */ 3409 static inline void netif_stop_queue(struct net_device *dev) 3410 { 3411 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3412 } 3413 3414 void netif_tx_stop_all_queues(struct net_device *dev); 3415 3416 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3417 { 3418 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3419 } 3420 3421 /** 3422 * netif_queue_stopped - test if transmit queue is flowblocked 3423 * @dev: network device 3424 * 3425 * Test if transmit queue on device is currently unable to send. 3426 */ 3427 static inline bool netif_queue_stopped(const struct net_device *dev) 3428 { 3429 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3430 } 3431 3432 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3433 { 3434 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3435 } 3436 3437 static inline bool 3438 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3439 { 3440 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3441 } 3442 3443 static inline bool 3444 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3445 { 3446 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3447 } 3448 3449 /** 3450 * netdev_queue_set_dql_min_limit - set dql minimum limit 3451 * @dev_queue: pointer to transmit queue 3452 * @min_limit: dql minimum limit 3453 * 3454 * Forces xmit_more() to return true until the minimum threshold 3455 * defined by @min_limit is reached (or until the tx queue is 3456 * empty). Warning: to be use with care, misuse will impact the 3457 * latency. 3458 */ 3459 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3460 unsigned int min_limit) 3461 { 3462 #ifdef CONFIG_BQL 3463 dev_queue->dql.min_limit = min_limit; 3464 #endif 3465 } 3466 3467 static inline int netdev_queue_dql_avail(const struct netdev_queue *txq) 3468 { 3469 #ifdef CONFIG_BQL 3470 /* Non-BQL migrated drivers will return 0, too. */ 3471 return dql_avail(&txq->dql); 3472 #else 3473 return 0; 3474 #endif 3475 } 3476 3477 /** 3478 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3479 * @dev_queue: pointer to transmit queue 3480 * 3481 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3482 * to give appropriate hint to the CPU. 3483 */ 3484 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3485 { 3486 #ifdef CONFIG_BQL 3487 prefetchw(&dev_queue->dql.num_queued); 3488 #endif 3489 } 3490 3491 /** 3492 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3493 * @dev_queue: pointer to transmit queue 3494 * 3495 * BQL enabled drivers might use this helper in their TX completion path, 3496 * to give appropriate hint to the CPU. 3497 */ 3498 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3499 { 3500 #ifdef CONFIG_BQL 3501 prefetchw(&dev_queue->dql.limit); 3502 #endif 3503 } 3504 3505 /** 3506 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3507 * @dev_queue: network device queue 3508 * @bytes: number of bytes queued to the device queue 3509 * 3510 * Report the number of bytes queued for sending/completion to the network 3511 * device hardware queue. @bytes should be a good approximation and should 3512 * exactly match netdev_completed_queue() @bytes. 3513 * This is typically called once per packet, from ndo_start_xmit(). 3514 */ 3515 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3516 unsigned int bytes) 3517 { 3518 #ifdef CONFIG_BQL 3519 dql_queued(&dev_queue->dql, bytes); 3520 3521 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3522 return; 3523 3524 /* Paired with READ_ONCE() from dev_watchdog() */ 3525 WRITE_ONCE(dev_queue->trans_start, jiffies); 3526 3527 /* This barrier is paired with smp_mb() from dev_watchdog() */ 3528 smp_mb__before_atomic(); 3529 3530 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3531 3532 /* 3533 * The XOFF flag must be set before checking the dql_avail below, 3534 * because in netdev_tx_completed_queue we update the dql_completed 3535 * before checking the XOFF flag. 3536 */ 3537 smp_mb__after_atomic(); 3538 3539 /* check again in case another CPU has just made room avail */ 3540 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3541 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3542 #endif 3543 } 3544 3545 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3546 * that they should not test BQL status themselves. 3547 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3548 * skb of a batch. 3549 * Returns true if the doorbell must be used to kick the NIC. 3550 */ 3551 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3552 unsigned int bytes, 3553 bool xmit_more) 3554 { 3555 if (xmit_more) { 3556 #ifdef CONFIG_BQL 3557 dql_queued(&dev_queue->dql, bytes); 3558 #endif 3559 return netif_tx_queue_stopped(dev_queue); 3560 } 3561 netdev_tx_sent_queue(dev_queue, bytes); 3562 return true; 3563 } 3564 3565 /** 3566 * netdev_sent_queue - report the number of bytes queued to hardware 3567 * @dev: network device 3568 * @bytes: number of bytes queued to the hardware device queue 3569 * 3570 * Report the number of bytes queued for sending/completion to the network 3571 * device hardware queue#0. @bytes should be a good approximation and should 3572 * exactly match netdev_completed_queue() @bytes. 3573 * This is typically called once per packet, from ndo_start_xmit(). 3574 */ 3575 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3576 { 3577 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3578 } 3579 3580 static inline bool __netdev_sent_queue(struct net_device *dev, 3581 unsigned int bytes, 3582 bool xmit_more) 3583 { 3584 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3585 xmit_more); 3586 } 3587 3588 /** 3589 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3590 * @dev_queue: network device queue 3591 * @pkts: number of packets (currently ignored) 3592 * @bytes: number of bytes dequeued from the device queue 3593 * 3594 * Must be called at most once per TX completion round (and not per 3595 * individual packet), so that BQL can adjust its limits appropriately. 3596 */ 3597 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3598 unsigned int pkts, unsigned int bytes) 3599 { 3600 #ifdef CONFIG_BQL 3601 if (unlikely(!bytes)) 3602 return; 3603 3604 dql_completed(&dev_queue->dql, bytes); 3605 3606 /* 3607 * Without the memory barrier there is a small possibility that 3608 * netdev_tx_sent_queue will miss the update and cause the queue to 3609 * be stopped forever 3610 */ 3611 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */ 3612 3613 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3614 return; 3615 3616 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3617 netif_schedule_queue(dev_queue); 3618 #endif 3619 } 3620 3621 /** 3622 * netdev_completed_queue - report bytes and packets completed by device 3623 * @dev: network device 3624 * @pkts: actual number of packets sent over the medium 3625 * @bytes: actual number of bytes sent over the medium 3626 * 3627 * Report the number of bytes and packets transmitted by the network device 3628 * hardware queue over the physical medium, @bytes must exactly match the 3629 * @bytes amount passed to netdev_sent_queue() 3630 */ 3631 static inline void netdev_completed_queue(struct net_device *dev, 3632 unsigned int pkts, unsigned int bytes) 3633 { 3634 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3635 } 3636 3637 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3638 { 3639 #ifdef CONFIG_BQL 3640 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3641 dql_reset(&q->dql); 3642 #endif 3643 } 3644 3645 /** 3646 * netdev_tx_reset_subqueue - reset the BQL stats and state of a netdev queue 3647 * @dev: network device 3648 * @qid: stack index of the queue to reset 3649 */ 3650 static inline void netdev_tx_reset_subqueue(const struct net_device *dev, 3651 u32 qid) 3652 { 3653 netdev_tx_reset_queue(netdev_get_tx_queue(dev, qid)); 3654 } 3655 3656 /** 3657 * netdev_reset_queue - reset the packets and bytes count of a network device 3658 * @dev_queue: network device 3659 * 3660 * Reset the bytes and packet count of a network device and clear the 3661 * software flow control OFF bit for this network device 3662 */ 3663 static inline void netdev_reset_queue(struct net_device *dev_queue) 3664 { 3665 netdev_tx_reset_subqueue(dev_queue, 0); 3666 } 3667 3668 /** 3669 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3670 * @dev: network device 3671 * @queue_index: given tx queue index 3672 * 3673 * Returns 0 if given tx queue index >= number of device tx queues, 3674 * otherwise returns the originally passed tx queue index. 3675 */ 3676 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3677 { 3678 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3679 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3680 dev->name, queue_index, 3681 dev->real_num_tx_queues); 3682 return 0; 3683 } 3684 3685 return queue_index; 3686 } 3687 3688 /** 3689 * netif_running - test if up 3690 * @dev: network device 3691 * 3692 * Test if the device has been brought up. 3693 */ 3694 static inline bool netif_running(const struct net_device *dev) 3695 { 3696 return test_bit(__LINK_STATE_START, &dev->state); 3697 } 3698 3699 /* 3700 * Routines to manage the subqueues on a device. We only need start, 3701 * stop, and a check if it's stopped. All other device management is 3702 * done at the overall netdevice level. 3703 * Also test the device if we're multiqueue. 3704 */ 3705 3706 /** 3707 * netif_start_subqueue - allow sending packets on subqueue 3708 * @dev: network device 3709 * @queue_index: sub queue index 3710 * 3711 * Start individual transmit queue of a device with multiple transmit queues. 3712 */ 3713 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3714 { 3715 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3716 3717 netif_tx_start_queue(txq); 3718 } 3719 3720 /** 3721 * netif_stop_subqueue - stop sending packets on subqueue 3722 * @dev: network device 3723 * @queue_index: sub queue index 3724 * 3725 * Stop individual transmit queue of a device with multiple transmit queues. 3726 */ 3727 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3728 { 3729 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3730 netif_tx_stop_queue(txq); 3731 } 3732 3733 /** 3734 * __netif_subqueue_stopped - test status of subqueue 3735 * @dev: network device 3736 * @queue_index: sub queue index 3737 * 3738 * Check individual transmit queue of a device with multiple transmit queues. 3739 */ 3740 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3741 u16 queue_index) 3742 { 3743 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3744 3745 return netif_tx_queue_stopped(txq); 3746 } 3747 3748 /** 3749 * netif_subqueue_stopped - test status of subqueue 3750 * @dev: network device 3751 * @skb: sub queue buffer pointer 3752 * 3753 * Check individual transmit queue of a device with multiple transmit queues. 3754 */ 3755 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3756 struct sk_buff *skb) 3757 { 3758 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3759 } 3760 3761 /** 3762 * netif_wake_subqueue - allow sending packets on subqueue 3763 * @dev: network device 3764 * @queue_index: sub queue index 3765 * 3766 * Resume individual transmit queue of a device with multiple transmit queues. 3767 */ 3768 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3769 { 3770 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3771 3772 netif_tx_wake_queue(txq); 3773 } 3774 3775 #ifdef CONFIG_XPS 3776 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3777 u16 index); 3778 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3779 u16 index, enum xps_map_type type); 3780 3781 /** 3782 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3783 * @j: CPU/Rx queue index 3784 * @mask: bitmask of all cpus/rx queues 3785 * @nr_bits: number of bits in the bitmask 3786 * 3787 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3788 */ 3789 static inline bool netif_attr_test_mask(unsigned long j, 3790 const unsigned long *mask, 3791 unsigned int nr_bits) 3792 { 3793 cpu_max_bits_warn(j, nr_bits); 3794 return test_bit(j, mask); 3795 } 3796 3797 /** 3798 * netif_attr_test_online - Test for online CPU/Rx queue 3799 * @j: CPU/Rx queue index 3800 * @online_mask: bitmask for CPUs/Rx queues that are online 3801 * @nr_bits: number of bits in the bitmask 3802 * 3803 * Returns true if a CPU/Rx queue is online. 3804 */ 3805 static inline bool netif_attr_test_online(unsigned long j, 3806 const unsigned long *online_mask, 3807 unsigned int nr_bits) 3808 { 3809 cpu_max_bits_warn(j, nr_bits); 3810 3811 if (online_mask) 3812 return test_bit(j, online_mask); 3813 3814 return (j < nr_bits); 3815 } 3816 3817 /** 3818 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3819 * @n: CPU/Rx queue index 3820 * @srcp: the cpumask/Rx queue mask pointer 3821 * @nr_bits: number of bits in the bitmask 3822 * 3823 * Returns >= nr_bits if no further CPUs/Rx queues set. 3824 */ 3825 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3826 unsigned int nr_bits) 3827 { 3828 /* -1 is a legal arg here. */ 3829 if (n != -1) 3830 cpu_max_bits_warn(n, nr_bits); 3831 3832 if (srcp) 3833 return find_next_bit(srcp, nr_bits, n + 1); 3834 3835 return n + 1; 3836 } 3837 3838 /** 3839 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3840 * @n: CPU/Rx queue index 3841 * @src1p: the first CPUs/Rx queues mask pointer 3842 * @src2p: the second CPUs/Rx queues mask pointer 3843 * @nr_bits: number of bits in the bitmask 3844 * 3845 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3846 */ 3847 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3848 const unsigned long *src2p, 3849 unsigned int nr_bits) 3850 { 3851 /* -1 is a legal arg here. */ 3852 if (n != -1) 3853 cpu_max_bits_warn(n, nr_bits); 3854 3855 if (src1p && src2p) 3856 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3857 else if (src1p) 3858 return find_next_bit(src1p, nr_bits, n + 1); 3859 else if (src2p) 3860 return find_next_bit(src2p, nr_bits, n + 1); 3861 3862 return n + 1; 3863 } 3864 #else 3865 static inline int netif_set_xps_queue(struct net_device *dev, 3866 const struct cpumask *mask, 3867 u16 index) 3868 { 3869 return 0; 3870 } 3871 3872 static inline int __netif_set_xps_queue(struct net_device *dev, 3873 const unsigned long *mask, 3874 u16 index, enum xps_map_type type) 3875 { 3876 return 0; 3877 } 3878 #endif 3879 3880 /** 3881 * netif_is_multiqueue - test if device has multiple transmit queues 3882 * @dev: network device 3883 * 3884 * Check if device has multiple transmit queues 3885 */ 3886 static inline bool netif_is_multiqueue(const struct net_device *dev) 3887 { 3888 return dev->num_tx_queues > 1; 3889 } 3890 3891 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3892 3893 #ifdef CONFIG_SYSFS 3894 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3895 #else 3896 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3897 unsigned int rxqs) 3898 { 3899 dev->real_num_rx_queues = rxqs; 3900 return 0; 3901 } 3902 #endif 3903 int netif_set_real_num_queues(struct net_device *dev, 3904 unsigned int txq, unsigned int rxq); 3905 3906 int netif_get_num_default_rss_queues(void); 3907 3908 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason); 3909 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason); 3910 3911 /* 3912 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3913 * interrupt context or with hardware interrupts being disabled. 3914 * (in_hardirq() || irqs_disabled()) 3915 * 3916 * We provide four helpers that can be used in following contexts : 3917 * 3918 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3919 * replacing kfree_skb(skb) 3920 * 3921 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3922 * Typically used in place of consume_skb(skb) in TX completion path 3923 * 3924 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3925 * replacing kfree_skb(skb) 3926 * 3927 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3928 * and consumed a packet. Used in place of consume_skb(skb) 3929 */ 3930 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3931 { 3932 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 3933 } 3934 3935 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3936 { 3937 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED); 3938 } 3939 3940 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3941 { 3942 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 3943 } 3944 3945 static inline void dev_consume_skb_any(struct sk_buff *skb) 3946 { 3947 dev_kfree_skb_any_reason(skb, SKB_CONSUMED); 3948 } 3949 3950 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3951 struct bpf_prog *xdp_prog); 3952 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3953 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff **pskb); 3954 int netif_rx(struct sk_buff *skb); 3955 int __netif_rx(struct sk_buff *skb); 3956 3957 int netif_receive_skb(struct sk_buff *skb); 3958 int netif_receive_skb_core(struct sk_buff *skb); 3959 void netif_receive_skb_list_internal(struct list_head *head); 3960 void netif_receive_skb_list(struct list_head *head); 3961 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3962 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3963 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3964 void napi_get_frags_check(struct napi_struct *napi); 3965 gro_result_t napi_gro_frags(struct napi_struct *napi); 3966 3967 static inline void napi_free_frags(struct napi_struct *napi) 3968 { 3969 kfree_skb(napi->skb); 3970 napi->skb = NULL; 3971 } 3972 3973 bool netdev_is_rx_handler_busy(struct net_device *dev); 3974 int netdev_rx_handler_register(struct net_device *dev, 3975 rx_handler_func_t *rx_handler, 3976 void *rx_handler_data); 3977 void netdev_rx_handler_unregister(struct net_device *dev); 3978 3979 bool dev_valid_name(const char *name); 3980 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 3981 { 3982 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 3983 } 3984 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 3985 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 3986 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3987 void __user *data, bool *need_copyout); 3988 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 3989 int generic_hwtstamp_get_lower(struct net_device *dev, 3990 struct kernel_hwtstamp_config *kernel_cfg); 3991 int generic_hwtstamp_set_lower(struct net_device *dev, 3992 struct kernel_hwtstamp_config *kernel_cfg, 3993 struct netlink_ext_ack *extack); 3994 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 3995 unsigned int dev_get_flags(const struct net_device *); 3996 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3997 struct netlink_ext_ack *extack); 3998 int dev_change_flags(struct net_device *dev, unsigned int flags, 3999 struct netlink_ext_ack *extack); 4000 int dev_set_alias(struct net_device *, const char *, size_t); 4001 int dev_get_alias(const struct net_device *, char *, size_t); 4002 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 4003 const char *pat, int new_ifindex); 4004 static inline 4005 int dev_change_net_namespace(struct net_device *dev, struct net *net, 4006 const char *pat) 4007 { 4008 return __dev_change_net_namespace(dev, net, pat, 0); 4009 } 4010 int __dev_set_mtu(struct net_device *, int); 4011 int dev_set_mtu(struct net_device *, int); 4012 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 4013 struct netlink_ext_ack *extack); 4014 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 4015 struct netlink_ext_ack *extack); 4016 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 4017 struct netlink_ext_ack *extack); 4018 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 4019 int dev_get_port_parent_id(struct net_device *dev, 4020 struct netdev_phys_item_id *ppid, bool recurse); 4021 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 4022 4023 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 4024 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 4025 struct netdev_queue *txq, int *ret); 4026 4027 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 4028 u8 dev_xdp_prog_count(struct net_device *dev); 4029 int dev_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf); 4030 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 4031 4032 u32 dev_get_min_mp_channel_count(const struct net_device *dev); 4033 4034 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4035 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4036 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 4037 bool is_skb_forwardable(const struct net_device *dev, 4038 const struct sk_buff *skb); 4039 4040 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 4041 const struct sk_buff *skb, 4042 const bool check_mtu) 4043 { 4044 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 4045 unsigned int len; 4046 4047 if (!(dev->flags & IFF_UP)) 4048 return false; 4049 4050 if (!check_mtu) 4051 return true; 4052 4053 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 4054 if (skb->len <= len) 4055 return true; 4056 4057 /* if TSO is enabled, we don't care about the length as the packet 4058 * could be forwarded without being segmented before 4059 */ 4060 if (skb_is_gso(skb)) 4061 return true; 4062 4063 return false; 4064 } 4065 4066 void netdev_core_stats_inc(struct net_device *dev, u32 offset); 4067 4068 #define DEV_CORE_STATS_INC(FIELD) \ 4069 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 4070 { \ 4071 netdev_core_stats_inc(dev, \ 4072 offsetof(struct net_device_core_stats, FIELD)); \ 4073 } 4074 DEV_CORE_STATS_INC(rx_dropped) 4075 DEV_CORE_STATS_INC(tx_dropped) 4076 DEV_CORE_STATS_INC(rx_nohandler) 4077 DEV_CORE_STATS_INC(rx_otherhost_dropped) 4078 #undef DEV_CORE_STATS_INC 4079 4080 static __always_inline int ____dev_forward_skb(struct net_device *dev, 4081 struct sk_buff *skb, 4082 const bool check_mtu) 4083 { 4084 if (skb_orphan_frags(skb, GFP_ATOMIC) || 4085 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 4086 dev_core_stats_rx_dropped_inc(dev); 4087 kfree_skb(skb); 4088 return NET_RX_DROP; 4089 } 4090 4091 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 4092 skb->priority = 0; 4093 return 0; 4094 } 4095 4096 bool dev_nit_active(struct net_device *dev); 4097 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 4098 4099 static inline void __dev_put(struct net_device *dev) 4100 { 4101 if (dev) { 4102 #ifdef CONFIG_PCPU_DEV_REFCNT 4103 this_cpu_dec(*dev->pcpu_refcnt); 4104 #else 4105 refcount_dec(&dev->dev_refcnt); 4106 #endif 4107 } 4108 } 4109 4110 static inline void __dev_hold(struct net_device *dev) 4111 { 4112 if (dev) { 4113 #ifdef CONFIG_PCPU_DEV_REFCNT 4114 this_cpu_inc(*dev->pcpu_refcnt); 4115 #else 4116 refcount_inc(&dev->dev_refcnt); 4117 #endif 4118 } 4119 } 4120 4121 static inline void __netdev_tracker_alloc(struct net_device *dev, 4122 netdevice_tracker *tracker, 4123 gfp_t gfp) 4124 { 4125 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4126 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 4127 #endif 4128 } 4129 4130 /* netdev_tracker_alloc() can upgrade a prior untracked reference 4131 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4132 */ 4133 static inline void netdev_tracker_alloc(struct net_device *dev, 4134 netdevice_tracker *tracker, gfp_t gfp) 4135 { 4136 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4137 refcount_dec(&dev->refcnt_tracker.no_tracker); 4138 __netdev_tracker_alloc(dev, tracker, gfp); 4139 #endif 4140 } 4141 4142 static inline void netdev_tracker_free(struct net_device *dev, 4143 netdevice_tracker *tracker) 4144 { 4145 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4146 ref_tracker_free(&dev->refcnt_tracker, tracker); 4147 #endif 4148 } 4149 4150 static inline void netdev_hold(struct net_device *dev, 4151 netdevice_tracker *tracker, gfp_t gfp) 4152 { 4153 if (dev) { 4154 __dev_hold(dev); 4155 __netdev_tracker_alloc(dev, tracker, gfp); 4156 } 4157 } 4158 4159 static inline void netdev_put(struct net_device *dev, 4160 netdevice_tracker *tracker) 4161 { 4162 if (dev) { 4163 netdev_tracker_free(dev, tracker); 4164 __dev_put(dev); 4165 } 4166 } 4167 4168 /** 4169 * dev_hold - get reference to device 4170 * @dev: network device 4171 * 4172 * Hold reference to device to keep it from being freed. 4173 * Try using netdev_hold() instead. 4174 */ 4175 static inline void dev_hold(struct net_device *dev) 4176 { 4177 netdev_hold(dev, NULL, GFP_ATOMIC); 4178 } 4179 4180 /** 4181 * dev_put - release reference to device 4182 * @dev: network device 4183 * 4184 * Release reference to device to allow it to be freed. 4185 * Try using netdev_put() instead. 4186 */ 4187 static inline void dev_put(struct net_device *dev) 4188 { 4189 netdev_put(dev, NULL); 4190 } 4191 4192 DEFINE_FREE(dev_put, struct net_device *, if (_T) dev_put(_T)) 4193 4194 static inline void netdev_ref_replace(struct net_device *odev, 4195 struct net_device *ndev, 4196 netdevice_tracker *tracker, 4197 gfp_t gfp) 4198 { 4199 if (odev) 4200 netdev_tracker_free(odev, tracker); 4201 4202 __dev_hold(ndev); 4203 __dev_put(odev); 4204 4205 if (ndev) 4206 __netdev_tracker_alloc(ndev, tracker, gfp); 4207 } 4208 4209 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4210 * and _off may be called from IRQ context, but it is caller 4211 * who is responsible for serialization of these calls. 4212 * 4213 * The name carrier is inappropriate, these functions should really be 4214 * called netif_lowerlayer_*() because they represent the state of any 4215 * kind of lower layer not just hardware media. 4216 */ 4217 void linkwatch_fire_event(struct net_device *dev); 4218 4219 /** 4220 * linkwatch_sync_dev - sync linkwatch for the given device 4221 * @dev: network device to sync linkwatch for 4222 * 4223 * Sync linkwatch for the given device, removing it from the 4224 * pending work list (if queued). 4225 */ 4226 void linkwatch_sync_dev(struct net_device *dev); 4227 4228 /** 4229 * netif_carrier_ok - test if carrier present 4230 * @dev: network device 4231 * 4232 * Check if carrier is present on device 4233 */ 4234 static inline bool netif_carrier_ok(const struct net_device *dev) 4235 { 4236 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4237 } 4238 4239 unsigned long dev_trans_start(struct net_device *dev); 4240 4241 void __netdev_watchdog_up(struct net_device *dev); 4242 4243 void netif_carrier_on(struct net_device *dev); 4244 void netif_carrier_off(struct net_device *dev); 4245 void netif_carrier_event(struct net_device *dev); 4246 4247 /** 4248 * netif_dormant_on - mark device as dormant. 4249 * @dev: network device 4250 * 4251 * Mark device as dormant (as per RFC2863). 4252 * 4253 * The dormant state indicates that the relevant interface is not 4254 * actually in a condition to pass packets (i.e., it is not 'up') but is 4255 * in a "pending" state, waiting for some external event. For "on- 4256 * demand" interfaces, this new state identifies the situation where the 4257 * interface is waiting for events to place it in the up state. 4258 */ 4259 static inline void netif_dormant_on(struct net_device *dev) 4260 { 4261 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4262 linkwatch_fire_event(dev); 4263 } 4264 4265 /** 4266 * netif_dormant_off - set device as not dormant. 4267 * @dev: network device 4268 * 4269 * Device is not in dormant state. 4270 */ 4271 static inline void netif_dormant_off(struct net_device *dev) 4272 { 4273 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4274 linkwatch_fire_event(dev); 4275 } 4276 4277 /** 4278 * netif_dormant - test if device is dormant 4279 * @dev: network device 4280 * 4281 * Check if device is dormant. 4282 */ 4283 static inline bool netif_dormant(const struct net_device *dev) 4284 { 4285 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4286 } 4287 4288 4289 /** 4290 * netif_testing_on - mark device as under test. 4291 * @dev: network device 4292 * 4293 * Mark device as under test (as per RFC2863). 4294 * 4295 * The testing state indicates that some test(s) must be performed on 4296 * the interface. After completion, of the test, the interface state 4297 * will change to up, dormant, or down, as appropriate. 4298 */ 4299 static inline void netif_testing_on(struct net_device *dev) 4300 { 4301 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4302 linkwatch_fire_event(dev); 4303 } 4304 4305 /** 4306 * netif_testing_off - set device as not under test. 4307 * @dev: network device 4308 * 4309 * Device is not in testing state. 4310 */ 4311 static inline void netif_testing_off(struct net_device *dev) 4312 { 4313 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4314 linkwatch_fire_event(dev); 4315 } 4316 4317 /** 4318 * netif_testing - test if device is under test 4319 * @dev: network device 4320 * 4321 * Check if device is under test 4322 */ 4323 static inline bool netif_testing(const struct net_device *dev) 4324 { 4325 return test_bit(__LINK_STATE_TESTING, &dev->state); 4326 } 4327 4328 4329 /** 4330 * netif_oper_up - test if device is operational 4331 * @dev: network device 4332 * 4333 * Check if carrier is operational 4334 */ 4335 static inline bool netif_oper_up(const struct net_device *dev) 4336 { 4337 unsigned int operstate = READ_ONCE(dev->operstate); 4338 4339 return operstate == IF_OPER_UP || 4340 operstate == IF_OPER_UNKNOWN /* backward compat */; 4341 } 4342 4343 /** 4344 * netif_device_present - is device available or removed 4345 * @dev: network device 4346 * 4347 * Check if device has not been removed from system. 4348 */ 4349 static inline bool netif_device_present(const struct net_device *dev) 4350 { 4351 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4352 } 4353 4354 void netif_device_detach(struct net_device *dev); 4355 4356 void netif_device_attach(struct net_device *dev); 4357 4358 /* 4359 * Network interface message level settings 4360 */ 4361 4362 enum { 4363 NETIF_MSG_DRV_BIT, 4364 NETIF_MSG_PROBE_BIT, 4365 NETIF_MSG_LINK_BIT, 4366 NETIF_MSG_TIMER_BIT, 4367 NETIF_MSG_IFDOWN_BIT, 4368 NETIF_MSG_IFUP_BIT, 4369 NETIF_MSG_RX_ERR_BIT, 4370 NETIF_MSG_TX_ERR_BIT, 4371 NETIF_MSG_TX_QUEUED_BIT, 4372 NETIF_MSG_INTR_BIT, 4373 NETIF_MSG_TX_DONE_BIT, 4374 NETIF_MSG_RX_STATUS_BIT, 4375 NETIF_MSG_PKTDATA_BIT, 4376 NETIF_MSG_HW_BIT, 4377 NETIF_MSG_WOL_BIT, 4378 4379 /* When you add a new bit above, update netif_msg_class_names array 4380 * in net/ethtool/common.c 4381 */ 4382 NETIF_MSG_CLASS_COUNT, 4383 }; 4384 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4385 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4386 4387 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4388 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4389 4390 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4391 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4392 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4393 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4394 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4395 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4396 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4397 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4398 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4399 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4400 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4401 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4402 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4403 #define NETIF_MSG_HW __NETIF_MSG(HW) 4404 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4405 4406 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4407 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4408 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4409 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4410 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4411 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4412 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4413 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4414 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4415 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4416 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4417 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4418 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4419 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4420 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4421 4422 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4423 { 4424 /* use default */ 4425 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4426 return default_msg_enable_bits; 4427 if (debug_value == 0) /* no output */ 4428 return 0; 4429 /* set low N bits */ 4430 return (1U << debug_value) - 1; 4431 } 4432 4433 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4434 { 4435 spin_lock(&txq->_xmit_lock); 4436 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4437 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4438 } 4439 4440 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4441 { 4442 __acquire(&txq->_xmit_lock); 4443 return true; 4444 } 4445 4446 static inline void __netif_tx_release(struct netdev_queue *txq) 4447 { 4448 __release(&txq->_xmit_lock); 4449 } 4450 4451 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4452 { 4453 spin_lock_bh(&txq->_xmit_lock); 4454 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4455 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4456 } 4457 4458 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4459 { 4460 bool ok = spin_trylock(&txq->_xmit_lock); 4461 4462 if (likely(ok)) { 4463 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4464 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4465 } 4466 return ok; 4467 } 4468 4469 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4470 { 4471 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4472 WRITE_ONCE(txq->xmit_lock_owner, -1); 4473 spin_unlock(&txq->_xmit_lock); 4474 } 4475 4476 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4477 { 4478 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4479 WRITE_ONCE(txq->xmit_lock_owner, -1); 4480 spin_unlock_bh(&txq->_xmit_lock); 4481 } 4482 4483 /* 4484 * txq->trans_start can be read locklessly from dev_watchdog() 4485 */ 4486 static inline void txq_trans_update(struct netdev_queue *txq) 4487 { 4488 if (txq->xmit_lock_owner != -1) 4489 WRITE_ONCE(txq->trans_start, jiffies); 4490 } 4491 4492 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4493 { 4494 unsigned long now = jiffies; 4495 4496 if (READ_ONCE(txq->trans_start) != now) 4497 WRITE_ONCE(txq->trans_start, now); 4498 } 4499 4500 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4501 static inline void netif_trans_update(struct net_device *dev) 4502 { 4503 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4504 4505 txq_trans_cond_update(txq); 4506 } 4507 4508 /** 4509 * netif_tx_lock - grab network device transmit lock 4510 * @dev: network device 4511 * 4512 * Get network device transmit lock 4513 */ 4514 void netif_tx_lock(struct net_device *dev); 4515 4516 static inline void netif_tx_lock_bh(struct net_device *dev) 4517 { 4518 local_bh_disable(); 4519 netif_tx_lock(dev); 4520 } 4521 4522 void netif_tx_unlock(struct net_device *dev); 4523 4524 static inline void netif_tx_unlock_bh(struct net_device *dev) 4525 { 4526 netif_tx_unlock(dev); 4527 local_bh_enable(); 4528 } 4529 4530 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4531 if (!(dev)->lltx) { \ 4532 __netif_tx_lock(txq, cpu); \ 4533 } else { \ 4534 __netif_tx_acquire(txq); \ 4535 } \ 4536 } 4537 4538 #define HARD_TX_TRYLOCK(dev, txq) \ 4539 (!(dev)->lltx ? \ 4540 __netif_tx_trylock(txq) : \ 4541 __netif_tx_acquire(txq)) 4542 4543 #define HARD_TX_UNLOCK(dev, txq) { \ 4544 if (!(dev)->lltx) { \ 4545 __netif_tx_unlock(txq); \ 4546 } else { \ 4547 __netif_tx_release(txq); \ 4548 } \ 4549 } 4550 4551 static inline void netif_tx_disable(struct net_device *dev) 4552 { 4553 unsigned int i; 4554 int cpu; 4555 4556 local_bh_disable(); 4557 cpu = smp_processor_id(); 4558 spin_lock(&dev->tx_global_lock); 4559 for (i = 0; i < dev->num_tx_queues; i++) { 4560 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4561 4562 __netif_tx_lock(txq, cpu); 4563 netif_tx_stop_queue(txq); 4564 __netif_tx_unlock(txq); 4565 } 4566 spin_unlock(&dev->tx_global_lock); 4567 local_bh_enable(); 4568 } 4569 4570 static inline void netif_addr_lock(struct net_device *dev) 4571 { 4572 unsigned char nest_level = 0; 4573 4574 #ifdef CONFIG_LOCKDEP 4575 nest_level = dev->nested_level; 4576 #endif 4577 spin_lock_nested(&dev->addr_list_lock, nest_level); 4578 } 4579 4580 static inline void netif_addr_lock_bh(struct net_device *dev) 4581 { 4582 unsigned char nest_level = 0; 4583 4584 #ifdef CONFIG_LOCKDEP 4585 nest_level = dev->nested_level; 4586 #endif 4587 local_bh_disable(); 4588 spin_lock_nested(&dev->addr_list_lock, nest_level); 4589 } 4590 4591 static inline void netif_addr_unlock(struct net_device *dev) 4592 { 4593 spin_unlock(&dev->addr_list_lock); 4594 } 4595 4596 static inline void netif_addr_unlock_bh(struct net_device *dev) 4597 { 4598 spin_unlock_bh(&dev->addr_list_lock); 4599 } 4600 4601 /* 4602 * dev_addrs walker. Should be used only for read access. Call with 4603 * rcu_read_lock held. 4604 */ 4605 #define for_each_dev_addr(dev, ha) \ 4606 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4607 4608 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4609 4610 void ether_setup(struct net_device *dev); 4611 4612 /* Allocate dummy net_device */ 4613 struct net_device *alloc_netdev_dummy(int sizeof_priv); 4614 4615 /* Support for loadable net-drivers */ 4616 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4617 unsigned char name_assign_type, 4618 void (*setup)(struct net_device *), 4619 unsigned int txqs, unsigned int rxqs); 4620 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4621 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4622 4623 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4624 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4625 count) 4626 4627 int register_netdev(struct net_device *dev); 4628 void unregister_netdev(struct net_device *dev); 4629 4630 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4631 4632 /* General hardware address lists handling functions */ 4633 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4634 struct netdev_hw_addr_list *from_list, int addr_len); 4635 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4636 struct netdev_hw_addr_list *from_list, int addr_len); 4637 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4638 struct net_device *dev, 4639 int (*sync)(struct net_device *, const unsigned char *), 4640 int (*unsync)(struct net_device *, 4641 const unsigned char *)); 4642 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4643 struct net_device *dev, 4644 int (*sync)(struct net_device *, 4645 const unsigned char *, int), 4646 int (*unsync)(struct net_device *, 4647 const unsigned char *, int)); 4648 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4649 struct net_device *dev, 4650 int (*unsync)(struct net_device *, 4651 const unsigned char *, int)); 4652 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4653 struct net_device *dev, 4654 int (*unsync)(struct net_device *, 4655 const unsigned char *)); 4656 void __hw_addr_init(struct netdev_hw_addr_list *list); 4657 4658 /* Functions used for device addresses handling */ 4659 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4660 const void *addr, size_t len); 4661 4662 static inline void 4663 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4664 { 4665 dev_addr_mod(dev, 0, addr, len); 4666 } 4667 4668 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4669 { 4670 __dev_addr_set(dev, addr, dev->addr_len); 4671 } 4672 4673 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4674 unsigned char addr_type); 4675 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4676 unsigned char addr_type); 4677 4678 /* Functions used for unicast addresses handling */ 4679 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4680 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4681 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4682 int dev_uc_sync(struct net_device *to, struct net_device *from); 4683 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4684 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4685 void dev_uc_flush(struct net_device *dev); 4686 void dev_uc_init(struct net_device *dev); 4687 4688 /** 4689 * __dev_uc_sync - Synchronize device's unicast list 4690 * @dev: device to sync 4691 * @sync: function to call if address should be added 4692 * @unsync: function to call if address should be removed 4693 * 4694 * Add newly added addresses to the interface, and release 4695 * addresses that have been deleted. 4696 */ 4697 static inline int __dev_uc_sync(struct net_device *dev, 4698 int (*sync)(struct net_device *, 4699 const unsigned char *), 4700 int (*unsync)(struct net_device *, 4701 const unsigned char *)) 4702 { 4703 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4704 } 4705 4706 /** 4707 * __dev_uc_unsync - Remove synchronized addresses from device 4708 * @dev: device to sync 4709 * @unsync: function to call if address should be removed 4710 * 4711 * Remove all addresses that were added to the device by dev_uc_sync(). 4712 */ 4713 static inline void __dev_uc_unsync(struct net_device *dev, 4714 int (*unsync)(struct net_device *, 4715 const unsigned char *)) 4716 { 4717 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4718 } 4719 4720 /* Functions used for multicast addresses handling */ 4721 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4722 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4723 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4724 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4725 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4726 int dev_mc_sync(struct net_device *to, struct net_device *from); 4727 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4728 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4729 void dev_mc_flush(struct net_device *dev); 4730 void dev_mc_init(struct net_device *dev); 4731 4732 /** 4733 * __dev_mc_sync - Synchronize device's multicast list 4734 * @dev: device to sync 4735 * @sync: function to call if address should be added 4736 * @unsync: function to call if address should be removed 4737 * 4738 * Add newly added addresses to the interface, and release 4739 * addresses that have been deleted. 4740 */ 4741 static inline int __dev_mc_sync(struct net_device *dev, 4742 int (*sync)(struct net_device *, 4743 const unsigned char *), 4744 int (*unsync)(struct net_device *, 4745 const unsigned char *)) 4746 { 4747 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4748 } 4749 4750 /** 4751 * __dev_mc_unsync - Remove synchronized addresses from device 4752 * @dev: device to sync 4753 * @unsync: function to call if address should be removed 4754 * 4755 * Remove all addresses that were added to the device by dev_mc_sync(). 4756 */ 4757 static inline void __dev_mc_unsync(struct net_device *dev, 4758 int (*unsync)(struct net_device *, 4759 const unsigned char *)) 4760 { 4761 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4762 } 4763 4764 /* Functions used for secondary unicast and multicast support */ 4765 void dev_set_rx_mode(struct net_device *dev); 4766 int dev_set_promiscuity(struct net_device *dev, int inc); 4767 int dev_set_allmulti(struct net_device *dev, int inc); 4768 void netdev_state_change(struct net_device *dev); 4769 void __netdev_notify_peers(struct net_device *dev); 4770 void netdev_notify_peers(struct net_device *dev); 4771 void netdev_features_change(struct net_device *dev); 4772 /* Load a device via the kmod */ 4773 void dev_load(struct net *net, const char *name); 4774 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4775 struct rtnl_link_stats64 *storage); 4776 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4777 const struct net_device_stats *netdev_stats); 4778 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4779 const struct pcpu_sw_netstats __percpu *netstats); 4780 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4781 4782 enum { 4783 NESTED_SYNC_IMM_BIT, 4784 NESTED_SYNC_TODO_BIT, 4785 }; 4786 4787 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4788 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4789 4790 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4791 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4792 4793 struct netdev_nested_priv { 4794 unsigned char flags; 4795 void *data; 4796 }; 4797 4798 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4799 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4800 struct list_head **iter); 4801 4802 /* iterate through upper list, must be called under RCU read lock */ 4803 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4804 for (iter = &(dev)->adj_list.upper, \ 4805 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4806 updev; \ 4807 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4808 4809 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4810 int (*fn)(struct net_device *upper_dev, 4811 struct netdev_nested_priv *priv), 4812 struct netdev_nested_priv *priv); 4813 4814 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4815 struct net_device *upper_dev); 4816 4817 bool netdev_has_any_upper_dev(struct net_device *dev); 4818 4819 void *netdev_lower_get_next_private(struct net_device *dev, 4820 struct list_head **iter); 4821 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4822 struct list_head **iter); 4823 4824 #define netdev_for_each_lower_private(dev, priv, iter) \ 4825 for (iter = (dev)->adj_list.lower.next, \ 4826 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4827 priv; \ 4828 priv = netdev_lower_get_next_private(dev, &(iter))) 4829 4830 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4831 for (iter = &(dev)->adj_list.lower, \ 4832 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4833 priv; \ 4834 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4835 4836 void *netdev_lower_get_next(struct net_device *dev, 4837 struct list_head **iter); 4838 4839 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4840 for (iter = (dev)->adj_list.lower.next, \ 4841 ldev = netdev_lower_get_next(dev, &(iter)); \ 4842 ldev; \ 4843 ldev = netdev_lower_get_next(dev, &(iter))) 4844 4845 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4846 struct list_head **iter); 4847 int netdev_walk_all_lower_dev(struct net_device *dev, 4848 int (*fn)(struct net_device *lower_dev, 4849 struct netdev_nested_priv *priv), 4850 struct netdev_nested_priv *priv); 4851 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4852 int (*fn)(struct net_device *lower_dev, 4853 struct netdev_nested_priv *priv), 4854 struct netdev_nested_priv *priv); 4855 4856 void *netdev_adjacent_get_private(struct list_head *adj_list); 4857 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4858 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4859 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4860 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4861 struct netlink_ext_ack *extack); 4862 int netdev_master_upper_dev_link(struct net_device *dev, 4863 struct net_device *upper_dev, 4864 void *upper_priv, void *upper_info, 4865 struct netlink_ext_ack *extack); 4866 void netdev_upper_dev_unlink(struct net_device *dev, 4867 struct net_device *upper_dev); 4868 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4869 struct net_device *new_dev, 4870 struct net_device *dev, 4871 struct netlink_ext_ack *extack); 4872 void netdev_adjacent_change_commit(struct net_device *old_dev, 4873 struct net_device *new_dev, 4874 struct net_device *dev); 4875 void netdev_adjacent_change_abort(struct net_device *old_dev, 4876 struct net_device *new_dev, 4877 struct net_device *dev); 4878 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4879 void *netdev_lower_dev_get_private(struct net_device *dev, 4880 struct net_device *lower_dev); 4881 void netdev_lower_state_changed(struct net_device *lower_dev, 4882 void *lower_state_info); 4883 4884 /* RSS keys are 40 or 52 bytes long */ 4885 #define NETDEV_RSS_KEY_LEN 52 4886 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4887 void netdev_rss_key_fill(void *buffer, size_t len); 4888 4889 int skb_checksum_help(struct sk_buff *skb); 4890 int skb_crc32c_csum_help(struct sk_buff *skb); 4891 int skb_csum_hwoffload_help(struct sk_buff *skb, 4892 const netdev_features_t features); 4893 4894 struct netdev_bonding_info { 4895 ifslave slave; 4896 ifbond master; 4897 }; 4898 4899 struct netdev_notifier_bonding_info { 4900 struct netdev_notifier_info info; /* must be first */ 4901 struct netdev_bonding_info bonding_info; 4902 }; 4903 4904 void netdev_bonding_info_change(struct net_device *dev, 4905 struct netdev_bonding_info *bonding_info); 4906 4907 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4908 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4909 #else 4910 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4911 const void *data) 4912 { 4913 } 4914 #endif 4915 4916 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4917 4918 static inline bool can_checksum_protocol(netdev_features_t features, 4919 __be16 protocol) 4920 { 4921 if (protocol == htons(ETH_P_FCOE)) 4922 return !!(features & NETIF_F_FCOE_CRC); 4923 4924 /* Assume this is an IP checksum (not SCTP CRC) */ 4925 4926 if (features & NETIF_F_HW_CSUM) { 4927 /* Can checksum everything */ 4928 return true; 4929 } 4930 4931 switch (protocol) { 4932 case htons(ETH_P_IP): 4933 return !!(features & NETIF_F_IP_CSUM); 4934 case htons(ETH_P_IPV6): 4935 return !!(features & NETIF_F_IPV6_CSUM); 4936 default: 4937 return false; 4938 } 4939 } 4940 4941 #ifdef CONFIG_BUG 4942 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4943 #else 4944 static inline void netdev_rx_csum_fault(struct net_device *dev, 4945 struct sk_buff *skb) 4946 { 4947 } 4948 #endif 4949 /* rx skb timestamps */ 4950 void net_enable_timestamp(void); 4951 void net_disable_timestamp(void); 4952 4953 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 4954 const struct skb_shared_hwtstamps *hwtstamps, 4955 bool cycles) 4956 { 4957 const struct net_device_ops *ops = dev->netdev_ops; 4958 4959 if (ops->ndo_get_tstamp) 4960 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 4961 4962 return hwtstamps->hwtstamp; 4963 } 4964 4965 #ifndef CONFIG_PREEMPT_RT 4966 static inline void netdev_xmit_set_more(bool more) 4967 { 4968 __this_cpu_write(softnet_data.xmit.more, more); 4969 } 4970 4971 static inline bool netdev_xmit_more(void) 4972 { 4973 return __this_cpu_read(softnet_data.xmit.more); 4974 } 4975 #else 4976 static inline void netdev_xmit_set_more(bool more) 4977 { 4978 current->net_xmit.more = more; 4979 } 4980 4981 static inline bool netdev_xmit_more(void) 4982 { 4983 return current->net_xmit.more; 4984 } 4985 #endif 4986 4987 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4988 struct sk_buff *skb, struct net_device *dev, 4989 bool more) 4990 { 4991 netdev_xmit_set_more(more); 4992 return ops->ndo_start_xmit(skb, dev); 4993 } 4994 4995 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4996 struct netdev_queue *txq, bool more) 4997 { 4998 const struct net_device_ops *ops = dev->netdev_ops; 4999 netdev_tx_t rc; 5000 5001 rc = __netdev_start_xmit(ops, skb, dev, more); 5002 if (rc == NETDEV_TX_OK) 5003 txq_trans_update(txq); 5004 5005 return rc; 5006 } 5007 5008 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 5009 const void *ns); 5010 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 5011 const void *ns); 5012 5013 extern const struct kobj_ns_type_operations net_ns_type_operations; 5014 5015 const char *netdev_drivername(const struct net_device *dev); 5016 5017 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 5018 netdev_features_t f2) 5019 { 5020 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 5021 if (f1 & NETIF_F_HW_CSUM) 5022 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5023 else 5024 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5025 } 5026 5027 return f1 & f2; 5028 } 5029 5030 static inline netdev_features_t netdev_get_wanted_features( 5031 struct net_device *dev) 5032 { 5033 return (dev->features & ~dev->hw_features) | dev->wanted_features; 5034 } 5035 netdev_features_t netdev_increment_features(netdev_features_t all, 5036 netdev_features_t one, netdev_features_t mask); 5037 5038 /* Allow TSO being used on stacked device : 5039 * Performing the GSO segmentation before last device 5040 * is a performance improvement. 5041 */ 5042 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 5043 netdev_features_t mask) 5044 { 5045 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 5046 } 5047 5048 int __netdev_update_features(struct net_device *dev); 5049 void netdev_update_features(struct net_device *dev); 5050 void netdev_change_features(struct net_device *dev); 5051 5052 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 5053 struct net_device *dev); 5054 5055 netdev_features_t passthru_features_check(struct sk_buff *skb, 5056 struct net_device *dev, 5057 netdev_features_t features); 5058 netdev_features_t netif_skb_features(struct sk_buff *skb); 5059 void skb_warn_bad_offload(const struct sk_buff *skb); 5060 5061 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 5062 { 5063 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 5064 5065 /* check flags correspondence */ 5066 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 5067 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 5068 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 5069 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 5070 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 5071 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 5072 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 5073 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 5074 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 5075 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 5076 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 5077 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 5078 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 5079 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 5080 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 5081 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 5082 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 5083 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 5084 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 5085 5086 return (features & feature) == feature; 5087 } 5088 5089 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 5090 { 5091 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 5092 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 5093 } 5094 5095 static inline bool netif_needs_gso(struct sk_buff *skb, 5096 netdev_features_t features) 5097 { 5098 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 5099 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 5100 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 5101 } 5102 5103 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 5104 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 5105 void netif_inherit_tso_max(struct net_device *to, 5106 const struct net_device *from); 5107 5108 static inline unsigned int 5109 netif_get_gro_max_size(const struct net_device *dev, const struct sk_buff *skb) 5110 { 5111 /* pairs with WRITE_ONCE() in netif_set_gro(_ipv4)_max_size() */ 5112 return skb->protocol == htons(ETH_P_IPV6) ? 5113 READ_ONCE(dev->gro_max_size) : 5114 READ_ONCE(dev->gro_ipv4_max_size); 5115 } 5116 5117 static inline unsigned int 5118 netif_get_gso_max_size(const struct net_device *dev, const struct sk_buff *skb) 5119 { 5120 /* pairs with WRITE_ONCE() in netif_set_gso(_ipv4)_max_size() */ 5121 return skb->protocol == htons(ETH_P_IPV6) ? 5122 READ_ONCE(dev->gso_max_size) : 5123 READ_ONCE(dev->gso_ipv4_max_size); 5124 } 5125 5126 static inline bool netif_is_macsec(const struct net_device *dev) 5127 { 5128 return dev->priv_flags & IFF_MACSEC; 5129 } 5130 5131 static inline bool netif_is_macvlan(const struct net_device *dev) 5132 { 5133 return dev->priv_flags & IFF_MACVLAN; 5134 } 5135 5136 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5137 { 5138 return dev->priv_flags & IFF_MACVLAN_PORT; 5139 } 5140 5141 static inline bool netif_is_bond_master(const struct net_device *dev) 5142 { 5143 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5144 } 5145 5146 static inline bool netif_is_bond_slave(const struct net_device *dev) 5147 { 5148 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5149 } 5150 5151 static inline bool netif_supports_nofcs(struct net_device *dev) 5152 { 5153 return dev->priv_flags & IFF_SUPP_NOFCS; 5154 } 5155 5156 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5157 { 5158 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5159 } 5160 5161 static inline bool netif_is_l3_master(const struct net_device *dev) 5162 { 5163 return dev->priv_flags & IFF_L3MDEV_MASTER; 5164 } 5165 5166 static inline bool netif_is_l3_slave(const struct net_device *dev) 5167 { 5168 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5169 } 5170 5171 static inline int dev_sdif(const struct net_device *dev) 5172 { 5173 #ifdef CONFIG_NET_L3_MASTER_DEV 5174 if (netif_is_l3_slave(dev)) 5175 return dev->ifindex; 5176 #endif 5177 return 0; 5178 } 5179 5180 static inline bool netif_is_bridge_master(const struct net_device *dev) 5181 { 5182 return dev->priv_flags & IFF_EBRIDGE; 5183 } 5184 5185 static inline bool netif_is_bridge_port(const struct net_device *dev) 5186 { 5187 return dev->priv_flags & IFF_BRIDGE_PORT; 5188 } 5189 5190 static inline bool netif_is_ovs_master(const struct net_device *dev) 5191 { 5192 return dev->priv_flags & IFF_OPENVSWITCH; 5193 } 5194 5195 static inline bool netif_is_ovs_port(const struct net_device *dev) 5196 { 5197 return dev->priv_flags & IFF_OVS_DATAPATH; 5198 } 5199 5200 static inline bool netif_is_any_bridge_master(const struct net_device *dev) 5201 { 5202 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev); 5203 } 5204 5205 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5206 { 5207 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5208 } 5209 5210 static inline bool netif_is_team_master(const struct net_device *dev) 5211 { 5212 return dev->priv_flags & IFF_TEAM; 5213 } 5214 5215 static inline bool netif_is_team_port(const struct net_device *dev) 5216 { 5217 return dev->priv_flags & IFF_TEAM_PORT; 5218 } 5219 5220 static inline bool netif_is_lag_master(const struct net_device *dev) 5221 { 5222 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5223 } 5224 5225 static inline bool netif_is_lag_port(const struct net_device *dev) 5226 { 5227 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5228 } 5229 5230 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5231 { 5232 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5233 } 5234 5235 static inline bool netif_is_failover(const struct net_device *dev) 5236 { 5237 return dev->priv_flags & IFF_FAILOVER; 5238 } 5239 5240 static inline bool netif_is_failover_slave(const struct net_device *dev) 5241 { 5242 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5243 } 5244 5245 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5246 static inline void netif_keep_dst(struct net_device *dev) 5247 { 5248 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5249 } 5250 5251 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5252 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5253 { 5254 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5255 return netif_is_macsec(dev); 5256 } 5257 5258 extern struct pernet_operations __net_initdata loopback_net_ops; 5259 5260 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5261 5262 /* netdev_printk helpers, similar to dev_printk */ 5263 5264 static inline const char *netdev_name(const struct net_device *dev) 5265 { 5266 if (!dev->name[0] || strchr(dev->name, '%')) 5267 return "(unnamed net_device)"; 5268 return dev->name; 5269 } 5270 5271 static inline const char *netdev_reg_state(const struct net_device *dev) 5272 { 5273 u8 reg_state = READ_ONCE(dev->reg_state); 5274 5275 switch (reg_state) { 5276 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5277 case NETREG_REGISTERED: return ""; 5278 case NETREG_UNREGISTERING: return " (unregistering)"; 5279 case NETREG_UNREGISTERED: return " (unregistered)"; 5280 case NETREG_RELEASED: return " (released)"; 5281 case NETREG_DUMMY: return " (dummy)"; 5282 } 5283 5284 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state); 5285 return " (unknown)"; 5286 } 5287 5288 #define MODULE_ALIAS_NETDEV(device) \ 5289 MODULE_ALIAS("netdev-" device) 5290 5291 /* 5292 * netdev_WARN() acts like dev_printk(), but with the key difference 5293 * of using a WARN/WARN_ON to get the message out, including the 5294 * file/line information and a backtrace. 5295 */ 5296 #define netdev_WARN(dev, format, args...) \ 5297 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5298 netdev_reg_state(dev), ##args) 5299 5300 #define netdev_WARN_ONCE(dev, format, args...) \ 5301 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5302 netdev_reg_state(dev), ##args) 5303 5304 /* 5305 * The list of packet types we will receive (as opposed to discard) 5306 * and the routines to invoke. 5307 * 5308 * Why 16. Because with 16 the only overlap we get on a hash of the 5309 * low nibble of the protocol value is RARP/SNAP/X.25. 5310 * 5311 * 0800 IP 5312 * 0001 802.3 5313 * 0002 AX.25 5314 * 0004 802.2 5315 * 8035 RARP 5316 * 0005 SNAP 5317 * 0805 X.25 5318 * 0806 ARP 5319 * 8137 IPX 5320 * 0009 Localtalk 5321 * 86DD IPv6 5322 */ 5323 #define PTYPE_HASH_SIZE (16) 5324 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5325 5326 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5327 5328 extern struct net_device *blackhole_netdev; 5329 5330 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */ 5331 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD) 5332 #define DEV_STATS_ADD(DEV, FIELD, VAL) \ 5333 atomic_long_add((VAL), &(DEV)->stats.__##FIELD) 5334 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD) 5335 5336 #endif /* _LINUX_NETDEVICE_H */ 5337