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 56 struct netpoll_info; 57 struct device; 58 struct ethtool_ops; 59 struct kernel_hwtstamp_config; 60 struct phy_device; 61 struct dsa_port; 62 struct ip_tunnel_parm_kern; 63 struct macsec_context; 64 struct macsec_ops; 65 struct netdev_name_node; 66 struct sd_flow_limit; 67 struct sfp_bus; 68 /* 802.11 specific */ 69 struct wireless_dev; 70 /* 802.15.4 specific */ 71 struct wpan_dev; 72 struct mpls_dev; 73 /* UDP Tunnel offloads */ 74 struct udp_tunnel_info; 75 struct udp_tunnel_nic_info; 76 struct udp_tunnel_nic; 77 struct bpf_prog; 78 struct xdp_buff; 79 struct xdp_frame; 80 struct xdp_metadata_ops; 81 struct xdp_md; 82 struct ethtool_netdev_state; 83 struct phy_link_topology; 84 85 typedef u32 xdp_features_t; 86 87 void synchronize_net(void); 88 void netdev_set_default_ethtool_ops(struct net_device *dev, 89 const struct ethtool_ops *ops); 90 void netdev_sw_irq_coalesce_default_on(struct net_device *dev); 91 92 /* Backlog congestion levels */ 93 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 94 #define NET_RX_DROP 1 /* packet dropped */ 95 96 #define MAX_NEST_DEV 8 97 98 /* 99 * Transmit return codes: transmit return codes originate from three different 100 * namespaces: 101 * 102 * - qdisc return codes 103 * - driver transmit return codes 104 * - errno values 105 * 106 * Drivers are allowed to return any one of those in their hard_start_xmit() 107 * function. Real network devices commonly used with qdiscs should only return 108 * the driver transmit return codes though - when qdiscs are used, the actual 109 * transmission happens asynchronously, so the value is not propagated to 110 * higher layers. Virtual network devices transmit synchronously; in this case 111 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 112 * others are propagated to higher layers. 113 */ 114 115 /* qdisc ->enqueue() return codes. */ 116 #define NET_XMIT_SUCCESS 0x00 117 #define NET_XMIT_DROP 0x01 /* skb dropped */ 118 #define NET_XMIT_CN 0x02 /* congestion notification */ 119 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 120 121 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 122 * indicates that the device will soon be dropping packets, or already drops 123 * some packets of the same priority; prompting us to send less aggressively. */ 124 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 125 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 126 127 /* Driver transmit return codes */ 128 #define NETDEV_TX_MASK 0xf0 129 130 enum netdev_tx { 131 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 132 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 133 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 134 }; 135 typedef enum netdev_tx netdev_tx_t; 136 137 /* 138 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 139 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 140 */ 141 static inline bool dev_xmit_complete(int rc) 142 { 143 /* 144 * Positive cases with an skb consumed by a driver: 145 * - successful transmission (rc == NETDEV_TX_OK) 146 * - error while transmitting (rc < 0) 147 * - error while queueing to a different device (rc & NET_XMIT_MASK) 148 */ 149 if (likely(rc < NET_XMIT_MASK)) 150 return true; 151 152 return false; 153 } 154 155 /* 156 * Compute the worst-case header length according to the protocols 157 * used. 158 */ 159 160 #if defined(CONFIG_HYPERV_NET) 161 # define LL_MAX_HEADER 128 162 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 163 # if defined(CONFIG_MAC80211_MESH) 164 # define LL_MAX_HEADER 128 165 # else 166 # define LL_MAX_HEADER 96 167 # endif 168 #else 169 # define LL_MAX_HEADER 32 170 #endif 171 172 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 173 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 174 #define MAX_HEADER LL_MAX_HEADER 175 #else 176 #define MAX_HEADER (LL_MAX_HEADER + 48) 177 #endif 178 179 /* 180 * Old network device statistics. Fields are native words 181 * (unsigned long) so they can be read and written atomically. 182 */ 183 184 #define NET_DEV_STAT(FIELD) \ 185 union { \ 186 unsigned long FIELD; \ 187 atomic_long_t __##FIELD; \ 188 } 189 190 struct net_device_stats { 191 NET_DEV_STAT(rx_packets); 192 NET_DEV_STAT(tx_packets); 193 NET_DEV_STAT(rx_bytes); 194 NET_DEV_STAT(tx_bytes); 195 NET_DEV_STAT(rx_errors); 196 NET_DEV_STAT(tx_errors); 197 NET_DEV_STAT(rx_dropped); 198 NET_DEV_STAT(tx_dropped); 199 NET_DEV_STAT(multicast); 200 NET_DEV_STAT(collisions); 201 NET_DEV_STAT(rx_length_errors); 202 NET_DEV_STAT(rx_over_errors); 203 NET_DEV_STAT(rx_crc_errors); 204 NET_DEV_STAT(rx_frame_errors); 205 NET_DEV_STAT(rx_fifo_errors); 206 NET_DEV_STAT(rx_missed_errors); 207 NET_DEV_STAT(tx_aborted_errors); 208 NET_DEV_STAT(tx_carrier_errors); 209 NET_DEV_STAT(tx_fifo_errors); 210 NET_DEV_STAT(tx_heartbeat_errors); 211 NET_DEV_STAT(tx_window_errors); 212 NET_DEV_STAT(rx_compressed); 213 NET_DEV_STAT(tx_compressed); 214 }; 215 #undef NET_DEV_STAT 216 217 /* per-cpu stats, allocated on demand. 218 * Try to fit them in a single cache line, for dev_get_stats() sake. 219 */ 220 struct net_device_core_stats { 221 unsigned long rx_dropped; 222 unsigned long tx_dropped; 223 unsigned long rx_nohandler; 224 unsigned long rx_otherhost_dropped; 225 } __aligned(4 * sizeof(unsigned long)); 226 227 #include <linux/cache.h> 228 #include <linux/skbuff.h> 229 230 struct neighbour; 231 struct neigh_parms; 232 struct sk_buff; 233 234 struct netdev_hw_addr { 235 struct list_head list; 236 struct rb_node node; 237 unsigned char addr[MAX_ADDR_LEN]; 238 unsigned char type; 239 #define NETDEV_HW_ADDR_T_LAN 1 240 #define NETDEV_HW_ADDR_T_SAN 2 241 #define NETDEV_HW_ADDR_T_UNICAST 3 242 #define NETDEV_HW_ADDR_T_MULTICAST 4 243 bool global_use; 244 int sync_cnt; 245 int refcount; 246 int synced; 247 struct rcu_head rcu_head; 248 }; 249 250 struct netdev_hw_addr_list { 251 struct list_head list; 252 int count; 253 254 /* Auxiliary tree for faster lookup on addition and deletion */ 255 struct rb_root tree; 256 }; 257 258 #define netdev_hw_addr_list_count(l) ((l)->count) 259 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 260 #define netdev_hw_addr_list_for_each(ha, l) \ 261 list_for_each_entry(ha, &(l)->list, list) 262 263 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 264 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 265 #define netdev_for_each_uc_addr(ha, dev) \ 266 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 267 #define netdev_for_each_synced_uc_addr(_ha, _dev) \ 268 netdev_for_each_uc_addr((_ha), (_dev)) \ 269 if ((_ha)->sync_cnt) 270 271 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 272 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 273 #define netdev_for_each_mc_addr(ha, dev) \ 274 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 275 #define netdev_for_each_synced_mc_addr(_ha, _dev) \ 276 netdev_for_each_mc_addr((_ha), (_dev)) \ 277 if ((_ha)->sync_cnt) 278 279 struct hh_cache { 280 unsigned int hh_len; 281 seqlock_t hh_lock; 282 283 /* cached hardware header; allow for machine alignment needs. */ 284 #define HH_DATA_MOD 16 285 #define HH_DATA_OFF(__len) \ 286 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 287 #define HH_DATA_ALIGN(__len) \ 288 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 289 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 290 }; 291 292 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 293 * Alternative is: 294 * dev->hard_header_len ? (dev->hard_header_len + 295 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 296 * 297 * We could use other alignment values, but we must maintain the 298 * relationship HH alignment <= LL alignment. 299 */ 300 #define LL_RESERVED_SPACE(dev) \ 301 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \ 302 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 303 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 304 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \ 305 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 306 307 struct header_ops { 308 int (*create) (struct sk_buff *skb, struct net_device *dev, 309 unsigned short type, const void *daddr, 310 const void *saddr, unsigned int len); 311 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 312 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 313 void (*cache_update)(struct hh_cache *hh, 314 const struct net_device *dev, 315 const unsigned char *haddr); 316 bool (*validate)(const char *ll_header, unsigned int len); 317 __be16 (*parse_protocol)(const struct sk_buff *skb); 318 }; 319 320 /* These flag bits are private to the generic network queueing 321 * layer; they may not be explicitly referenced by any other 322 * code. 323 */ 324 325 enum netdev_state_t { 326 __LINK_STATE_START, 327 __LINK_STATE_PRESENT, 328 __LINK_STATE_NOCARRIER, 329 __LINK_STATE_LINKWATCH_PENDING, 330 __LINK_STATE_DORMANT, 331 __LINK_STATE_TESTING, 332 }; 333 334 struct gro_list { 335 struct list_head list; 336 int count; 337 }; 338 339 /* 340 * size of gro hash buckets, must less than bit number of 341 * napi_struct::gro_bitmask 342 */ 343 #define GRO_HASH_BUCKETS 8 344 345 /* 346 * Structure for per-NAPI config 347 */ 348 struct napi_config { 349 u64 gro_flush_timeout; 350 u32 defer_hard_irqs; 351 unsigned int napi_id; 352 }; 353 354 /* 355 * Structure for NAPI scheduling similar to tasklet but with weighting 356 */ 357 struct napi_struct { 358 /* The poll_list must only be managed by the entity which 359 * changes the state of the NAPI_STATE_SCHED bit. This means 360 * whoever atomically sets that bit can add this napi_struct 361 * to the per-CPU poll_list, and whoever clears that bit 362 * can remove from the list right before clearing the bit. 363 */ 364 struct list_head poll_list; 365 366 unsigned long state; 367 int weight; 368 u32 defer_hard_irqs_count; 369 unsigned long gro_bitmask; 370 int (*poll)(struct napi_struct *, int); 371 #ifdef CONFIG_NETPOLL 372 /* CPU actively polling if netpoll is configured */ 373 int poll_owner; 374 #endif 375 /* CPU on which NAPI has been scheduled for processing */ 376 int list_owner; 377 struct net_device *dev; 378 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 379 struct sk_buff *skb; 380 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 381 int rx_count; /* length of rx_list */ 382 unsigned int napi_id; 383 struct hrtimer timer; 384 struct task_struct *thread; 385 unsigned long gro_flush_timeout; 386 u32 defer_hard_irqs; 387 /* control-path-only fields follow */ 388 struct list_head dev_list; 389 struct hlist_node napi_hash_node; 390 int irq; 391 int index; 392 struct napi_config *config; 393 }; 394 395 enum { 396 NAPI_STATE_SCHED, /* Poll is scheduled */ 397 NAPI_STATE_MISSED, /* reschedule a napi */ 398 NAPI_STATE_DISABLE, /* Disable pending */ 399 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 400 NAPI_STATE_LISTED, /* NAPI added to system lists */ 401 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 402 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 403 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 404 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 405 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 406 }; 407 408 enum { 409 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 410 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 411 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 412 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 413 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 414 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 415 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 416 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 417 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 418 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 419 }; 420 421 enum gro_result { 422 GRO_MERGED, 423 GRO_MERGED_FREE, 424 GRO_HELD, 425 GRO_NORMAL, 426 GRO_CONSUMED, 427 }; 428 typedef enum gro_result gro_result_t; 429 430 /* 431 * enum rx_handler_result - Possible return values for rx_handlers. 432 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 433 * further. 434 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 435 * case skb->dev was changed by rx_handler. 436 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 437 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 438 * 439 * rx_handlers are functions called from inside __netif_receive_skb(), to do 440 * special processing of the skb, prior to delivery to protocol handlers. 441 * 442 * Currently, a net_device can only have a single rx_handler registered. Trying 443 * to register a second rx_handler will return -EBUSY. 444 * 445 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 446 * To unregister a rx_handler on a net_device, use 447 * netdev_rx_handler_unregister(). 448 * 449 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 450 * do with the skb. 451 * 452 * If the rx_handler consumed the skb in some way, it should return 453 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 454 * the skb to be delivered in some other way. 455 * 456 * If the rx_handler changed skb->dev, to divert the skb to another 457 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 458 * new device will be called if it exists. 459 * 460 * If the rx_handler decides the skb should be ignored, it should return 461 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 462 * are registered on exact device (ptype->dev == skb->dev). 463 * 464 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 465 * delivered, it should return RX_HANDLER_PASS. 466 * 467 * A device without a registered rx_handler will behave as if rx_handler 468 * returned RX_HANDLER_PASS. 469 */ 470 471 enum rx_handler_result { 472 RX_HANDLER_CONSUMED, 473 RX_HANDLER_ANOTHER, 474 RX_HANDLER_EXACT, 475 RX_HANDLER_PASS, 476 }; 477 typedef enum rx_handler_result rx_handler_result_t; 478 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 479 480 void __napi_schedule(struct napi_struct *n); 481 void __napi_schedule_irqoff(struct napi_struct *n); 482 483 static inline bool napi_disable_pending(struct napi_struct *n) 484 { 485 return test_bit(NAPI_STATE_DISABLE, &n->state); 486 } 487 488 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 489 { 490 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 491 } 492 493 /** 494 * napi_is_scheduled - test if NAPI is scheduled 495 * @n: NAPI context 496 * 497 * This check is "best-effort". With no locking implemented, 498 * a NAPI can be scheduled or terminate right after this check 499 * and produce not precise results. 500 * 501 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled 502 * should not be used normally and napi_schedule should be 503 * used instead. 504 * 505 * Use only if the driver really needs to check if a NAPI 506 * is scheduled for example in the context of delayed timer 507 * that can be skipped if a NAPI is already scheduled. 508 * 509 * Return True if NAPI is scheduled, False otherwise. 510 */ 511 static inline bool napi_is_scheduled(struct napi_struct *n) 512 { 513 return test_bit(NAPI_STATE_SCHED, &n->state); 514 } 515 516 bool napi_schedule_prep(struct napi_struct *n); 517 518 /** 519 * napi_schedule - schedule NAPI poll 520 * @n: NAPI context 521 * 522 * Schedule NAPI poll routine to be called if it is not already 523 * running. 524 * Return true if we schedule a NAPI or false if not. 525 * Refer to napi_schedule_prep() for additional reason on why 526 * a NAPI might not be scheduled. 527 */ 528 static inline bool napi_schedule(struct napi_struct *n) 529 { 530 if (napi_schedule_prep(n)) { 531 __napi_schedule(n); 532 return true; 533 } 534 535 return false; 536 } 537 538 /** 539 * napi_schedule_irqoff - schedule NAPI poll 540 * @n: NAPI context 541 * 542 * Variant of napi_schedule(), assuming hard irqs are masked. 543 */ 544 static inline void napi_schedule_irqoff(struct napi_struct *n) 545 { 546 if (napi_schedule_prep(n)) 547 __napi_schedule_irqoff(n); 548 } 549 550 /** 551 * napi_complete_done - NAPI processing complete 552 * @n: NAPI context 553 * @work_done: number of packets processed 554 * 555 * Mark NAPI processing as complete. Should only be called if poll budget 556 * has not been completely consumed. 557 * Prefer over napi_complete(). 558 * Return false if device should avoid rearming interrupts. 559 */ 560 bool napi_complete_done(struct napi_struct *n, int work_done); 561 562 static inline bool napi_complete(struct napi_struct *n) 563 { 564 return napi_complete_done(n, 0); 565 } 566 567 int dev_set_threaded(struct net_device *dev, bool threaded); 568 569 /** 570 * napi_disable - prevent NAPI from scheduling 571 * @n: NAPI context 572 * 573 * Stop NAPI from being scheduled on this context. 574 * Waits till any outstanding processing completes. 575 */ 576 void napi_disable(struct napi_struct *n); 577 578 void napi_enable(struct napi_struct *n); 579 580 /** 581 * napi_synchronize - wait until NAPI is not running 582 * @n: NAPI context 583 * 584 * Wait until NAPI is done being scheduled on this context. 585 * Waits till any outstanding processing completes but 586 * does not disable future activations. 587 */ 588 static inline void napi_synchronize(const struct napi_struct *n) 589 { 590 if (IS_ENABLED(CONFIG_SMP)) 591 while (test_bit(NAPI_STATE_SCHED, &n->state)) 592 msleep(1); 593 else 594 barrier(); 595 } 596 597 /** 598 * napi_if_scheduled_mark_missed - if napi is running, set the 599 * NAPIF_STATE_MISSED 600 * @n: NAPI context 601 * 602 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 603 * NAPI is scheduled. 604 **/ 605 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 606 { 607 unsigned long val, new; 608 609 val = READ_ONCE(n->state); 610 do { 611 if (val & NAPIF_STATE_DISABLE) 612 return true; 613 614 if (!(val & NAPIF_STATE_SCHED)) 615 return false; 616 617 new = val | NAPIF_STATE_MISSED; 618 } while (!try_cmpxchg(&n->state, &val, new)); 619 620 return true; 621 } 622 623 enum netdev_queue_state_t { 624 __QUEUE_STATE_DRV_XOFF, 625 __QUEUE_STATE_STACK_XOFF, 626 __QUEUE_STATE_FROZEN, 627 }; 628 629 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 630 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 631 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 632 633 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 634 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 635 QUEUE_STATE_FROZEN) 636 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 637 QUEUE_STATE_FROZEN) 638 639 /* 640 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 641 * netif_tx_* functions below are used to manipulate this flag. The 642 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 643 * queue independently. The netif_xmit_*stopped functions below are called 644 * to check if the queue has been stopped by the driver or stack (either 645 * of the XOFF bits are set in the state). Drivers should not need to call 646 * netif_xmit*stopped functions, they should only be using netif_tx_*. 647 */ 648 649 struct netdev_queue { 650 /* 651 * read-mostly part 652 */ 653 struct net_device *dev; 654 netdevice_tracker dev_tracker; 655 656 struct Qdisc __rcu *qdisc; 657 struct Qdisc __rcu *qdisc_sleeping; 658 #ifdef CONFIG_SYSFS 659 struct kobject kobj; 660 #endif 661 unsigned long tx_maxrate; 662 /* 663 * Number of TX timeouts for this queue 664 * (/sys/class/net/DEV/Q/trans_timeout) 665 */ 666 atomic_long_t trans_timeout; 667 668 /* Subordinate device that the queue has been assigned to */ 669 struct net_device *sb_dev; 670 #ifdef CONFIG_XDP_SOCKETS 671 struct xsk_buff_pool *pool; 672 #endif 673 674 /* 675 * write-mostly part 676 */ 677 #ifdef CONFIG_BQL 678 struct dql dql; 679 #endif 680 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 681 int xmit_lock_owner; 682 /* 683 * Time (in jiffies) of last Tx 684 */ 685 unsigned long trans_start; 686 687 unsigned long state; 688 689 /* 690 * slow- / control-path part 691 */ 692 /* NAPI instance for the queue 693 * Readers and writers must hold RTNL 694 */ 695 struct napi_struct *napi; 696 697 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 698 int numa_node; 699 #endif 700 } ____cacheline_aligned_in_smp; 701 702 extern int sysctl_fb_tunnels_only_for_init_net; 703 extern int sysctl_devconf_inherit_init_net; 704 705 /* 706 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 707 * == 1 : For initns only 708 * == 2 : For none. 709 */ 710 static inline bool net_has_fallback_tunnels(const struct net *net) 711 { 712 #if IS_ENABLED(CONFIG_SYSCTL) 713 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net); 714 715 return !fb_tunnels_only_for_init_net || 716 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1); 717 #else 718 return true; 719 #endif 720 } 721 722 static inline int net_inherit_devconf(void) 723 { 724 #if IS_ENABLED(CONFIG_SYSCTL) 725 return READ_ONCE(sysctl_devconf_inherit_init_net); 726 #else 727 return 0; 728 #endif 729 } 730 731 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 732 { 733 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 734 return q->numa_node; 735 #else 736 return NUMA_NO_NODE; 737 #endif 738 } 739 740 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 741 { 742 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 743 q->numa_node = node; 744 #endif 745 } 746 747 #ifdef CONFIG_RFS_ACCEL 748 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 749 u16 filter_id); 750 #endif 751 752 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 753 enum xps_map_type { 754 XPS_CPUS = 0, 755 XPS_RXQS, 756 XPS_MAPS_MAX, 757 }; 758 759 #ifdef CONFIG_XPS 760 /* 761 * This structure holds an XPS map which can be of variable length. The 762 * map is an array of queues. 763 */ 764 struct xps_map { 765 unsigned int len; 766 unsigned int alloc_len; 767 struct rcu_head rcu; 768 u16 queues[]; 769 }; 770 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 771 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 772 - sizeof(struct xps_map)) / sizeof(u16)) 773 774 /* 775 * This structure holds all XPS maps for device. Maps are indexed by CPU. 776 * 777 * We keep track of the number of cpus/rxqs used when the struct is allocated, 778 * in nr_ids. This will help not accessing out-of-bound memory. 779 * 780 * We keep track of the number of traffic classes used when the struct is 781 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 782 * not crossing its upper bound, as the original dev->num_tc can be updated in 783 * the meantime. 784 */ 785 struct xps_dev_maps { 786 struct rcu_head rcu; 787 unsigned int nr_ids; 788 s16 num_tc; 789 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 790 }; 791 792 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 793 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 794 795 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 796 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 797 798 #endif /* CONFIG_XPS */ 799 800 #define TC_MAX_QUEUE 16 801 #define TC_BITMASK 15 802 /* HW offloaded queuing disciplines txq count and offset maps */ 803 struct netdev_tc_txq { 804 u16 count; 805 u16 offset; 806 }; 807 808 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 809 /* 810 * This structure is to hold information about the device 811 * configured to run FCoE protocol stack. 812 */ 813 struct netdev_fcoe_hbainfo { 814 char manufacturer[64]; 815 char serial_number[64]; 816 char hardware_version[64]; 817 char driver_version[64]; 818 char optionrom_version[64]; 819 char firmware_version[64]; 820 char model[256]; 821 char model_description[256]; 822 }; 823 #endif 824 825 #define MAX_PHYS_ITEM_ID_LEN 32 826 827 /* This structure holds a unique identifier to identify some 828 * physical item (port for example) used by a netdevice. 829 */ 830 struct netdev_phys_item_id { 831 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 832 unsigned char id_len; 833 }; 834 835 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 836 struct netdev_phys_item_id *b) 837 { 838 return a->id_len == b->id_len && 839 memcmp(a->id, b->id, a->id_len) == 0; 840 } 841 842 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 843 struct sk_buff *skb, 844 struct net_device *sb_dev); 845 846 enum net_device_path_type { 847 DEV_PATH_ETHERNET = 0, 848 DEV_PATH_VLAN, 849 DEV_PATH_BRIDGE, 850 DEV_PATH_PPPOE, 851 DEV_PATH_DSA, 852 DEV_PATH_MTK_WDMA, 853 }; 854 855 struct net_device_path { 856 enum net_device_path_type type; 857 const struct net_device *dev; 858 union { 859 struct { 860 u16 id; 861 __be16 proto; 862 u8 h_dest[ETH_ALEN]; 863 } encap; 864 struct { 865 enum { 866 DEV_PATH_BR_VLAN_KEEP, 867 DEV_PATH_BR_VLAN_TAG, 868 DEV_PATH_BR_VLAN_UNTAG, 869 DEV_PATH_BR_VLAN_UNTAG_HW, 870 } vlan_mode; 871 u16 vlan_id; 872 __be16 vlan_proto; 873 } bridge; 874 struct { 875 int port; 876 u16 proto; 877 } dsa; 878 struct { 879 u8 wdma_idx; 880 u8 queue; 881 u16 wcid; 882 u8 bss; 883 u8 amsdu; 884 } mtk_wdma; 885 }; 886 }; 887 888 #define NET_DEVICE_PATH_STACK_MAX 5 889 #define NET_DEVICE_PATH_VLAN_MAX 2 890 891 struct net_device_path_stack { 892 int num_paths; 893 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 894 }; 895 896 struct net_device_path_ctx { 897 const struct net_device *dev; 898 u8 daddr[ETH_ALEN]; 899 900 int num_vlans; 901 struct { 902 u16 id; 903 __be16 proto; 904 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 905 }; 906 907 enum tc_setup_type { 908 TC_QUERY_CAPS, 909 TC_SETUP_QDISC_MQPRIO, 910 TC_SETUP_CLSU32, 911 TC_SETUP_CLSFLOWER, 912 TC_SETUP_CLSMATCHALL, 913 TC_SETUP_CLSBPF, 914 TC_SETUP_BLOCK, 915 TC_SETUP_QDISC_CBS, 916 TC_SETUP_QDISC_RED, 917 TC_SETUP_QDISC_PRIO, 918 TC_SETUP_QDISC_MQ, 919 TC_SETUP_QDISC_ETF, 920 TC_SETUP_ROOT_QDISC, 921 TC_SETUP_QDISC_GRED, 922 TC_SETUP_QDISC_TAPRIO, 923 TC_SETUP_FT, 924 TC_SETUP_QDISC_ETS, 925 TC_SETUP_QDISC_TBF, 926 TC_SETUP_QDISC_FIFO, 927 TC_SETUP_QDISC_HTB, 928 TC_SETUP_ACT, 929 }; 930 931 /* These structures hold the attributes of bpf state that are being passed 932 * to the netdevice through the bpf op. 933 */ 934 enum bpf_netdev_command { 935 /* Set or clear a bpf program used in the earliest stages of packet 936 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 937 * is responsible for calling bpf_prog_put on any old progs that are 938 * stored. In case of error, the callee need not release the new prog 939 * reference, but on success it takes ownership and must bpf_prog_put 940 * when it is no longer used. 941 */ 942 XDP_SETUP_PROG, 943 XDP_SETUP_PROG_HW, 944 /* BPF program for offload callbacks, invoked at program load time. */ 945 BPF_OFFLOAD_MAP_ALLOC, 946 BPF_OFFLOAD_MAP_FREE, 947 XDP_SETUP_XSK_POOL, 948 }; 949 950 struct bpf_prog_offload_ops; 951 struct netlink_ext_ack; 952 struct xdp_umem; 953 struct xdp_dev_bulk_queue; 954 struct bpf_xdp_link; 955 956 enum bpf_xdp_mode { 957 XDP_MODE_SKB = 0, 958 XDP_MODE_DRV = 1, 959 XDP_MODE_HW = 2, 960 __MAX_XDP_MODE 961 }; 962 963 struct bpf_xdp_entity { 964 struct bpf_prog *prog; 965 struct bpf_xdp_link *link; 966 }; 967 968 struct netdev_bpf { 969 enum bpf_netdev_command command; 970 union { 971 /* XDP_SETUP_PROG */ 972 struct { 973 u32 flags; 974 struct bpf_prog *prog; 975 struct netlink_ext_ack *extack; 976 }; 977 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 978 struct { 979 struct bpf_offloaded_map *offmap; 980 }; 981 /* XDP_SETUP_XSK_POOL */ 982 struct { 983 struct xsk_buff_pool *pool; 984 u16 queue_id; 985 } xsk; 986 }; 987 }; 988 989 /* Flags for ndo_xsk_wakeup. */ 990 #define XDP_WAKEUP_RX (1 << 0) 991 #define XDP_WAKEUP_TX (1 << 1) 992 993 #ifdef CONFIG_XFRM_OFFLOAD 994 struct xfrmdev_ops { 995 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack); 996 void (*xdo_dev_state_delete) (struct xfrm_state *x); 997 void (*xdo_dev_state_free) (struct xfrm_state *x); 998 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 999 struct xfrm_state *x); 1000 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1001 void (*xdo_dev_state_update_stats) (struct xfrm_state *x); 1002 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack); 1003 void (*xdo_dev_policy_delete) (struct xfrm_policy *x); 1004 void (*xdo_dev_policy_free) (struct xfrm_policy *x); 1005 }; 1006 #endif 1007 1008 struct dev_ifalias { 1009 struct rcu_head rcuhead; 1010 char ifalias[]; 1011 }; 1012 1013 struct devlink; 1014 struct tlsdev_ops; 1015 1016 struct netdev_net_notifier { 1017 struct list_head list; 1018 struct notifier_block *nb; 1019 }; 1020 1021 /* 1022 * This structure defines the management hooks for network devices. 1023 * The following hooks can be defined; unless noted otherwise, they are 1024 * optional and can be filled with a null pointer. 1025 * 1026 * int (*ndo_init)(struct net_device *dev); 1027 * This function is called once when a network device is registered. 1028 * The network device can use this for any late stage initialization 1029 * or semantic validation. It can fail with an error code which will 1030 * be propagated back to register_netdev. 1031 * 1032 * void (*ndo_uninit)(struct net_device *dev); 1033 * This function is called when device is unregistered or when registration 1034 * fails. It is not called if init fails. 1035 * 1036 * int (*ndo_open)(struct net_device *dev); 1037 * This function is called when a network device transitions to the up 1038 * state. 1039 * 1040 * int (*ndo_stop)(struct net_device *dev); 1041 * This function is called when a network device transitions to the down 1042 * state. 1043 * 1044 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1045 * struct net_device *dev); 1046 * Called when a packet needs to be transmitted. 1047 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1048 * the queue before that can happen; it's for obsolete devices and weird 1049 * corner cases, but the stack really does a non-trivial amount 1050 * of useless work if you return NETDEV_TX_BUSY. 1051 * Required; cannot be NULL. 1052 * 1053 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1054 * struct net_device *dev 1055 * netdev_features_t features); 1056 * Called by core transmit path to determine if device is capable of 1057 * performing offload operations on a given packet. This is to give 1058 * the device an opportunity to implement any restrictions that cannot 1059 * be otherwise expressed by feature flags. The check is called with 1060 * the set of features that the stack has calculated and it returns 1061 * those the driver believes to be appropriate. 1062 * 1063 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1064 * struct net_device *sb_dev); 1065 * Called to decide which queue to use when device supports multiple 1066 * transmit queues. 1067 * 1068 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1069 * This function is called to allow device receiver to make 1070 * changes to configuration when multicast or promiscuous is enabled. 1071 * 1072 * void (*ndo_set_rx_mode)(struct net_device *dev); 1073 * This function is called device changes address list filtering. 1074 * If driver handles unicast address filtering, it should set 1075 * IFF_UNICAST_FLT in its priv_flags. 1076 * 1077 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1078 * This function is called when the Media Access Control address 1079 * needs to be changed. If this interface is not defined, the 1080 * MAC address can not be changed. 1081 * 1082 * int (*ndo_validate_addr)(struct net_device *dev); 1083 * Test if Media Access Control address is valid for the device. 1084 * 1085 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1086 * Old-style ioctl entry point. This is used internally by the 1087 * appletalk and ieee802154 subsystems but is no longer called by 1088 * the device ioctl handler. 1089 * 1090 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1091 * Used by the bonding driver for its device specific ioctls: 1092 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1093 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1094 * 1095 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1096 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1097 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1098 * 1099 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1100 * Used to set network devices bus interface parameters. This interface 1101 * is retained for legacy reasons; new devices should use the bus 1102 * interface (PCI) for low level management. 1103 * 1104 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1105 * Called when a user wants to change the Maximum Transfer Unit 1106 * of a device. 1107 * 1108 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1109 * Callback used when the transmitter has not made any progress 1110 * for dev->watchdog ticks. 1111 * 1112 * void (*ndo_get_stats64)(struct net_device *dev, 1113 * struct rtnl_link_stats64 *storage); 1114 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1115 * Called when a user wants to get the network device usage 1116 * statistics. Drivers must do one of the following: 1117 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1118 * rtnl_link_stats64 structure passed by the caller. 1119 * 2. Define @ndo_get_stats to update a net_device_stats structure 1120 * (which should normally be dev->stats) and return a pointer to 1121 * it. The structure may be changed asynchronously only if each 1122 * field is written atomically. 1123 * 3. Update dev->stats asynchronously and atomically, and define 1124 * neither operation. 1125 * 1126 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1127 * Return true if this device supports offload stats of this attr_id. 1128 * 1129 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1130 * void *attr_data) 1131 * Get statistics for offload operations by attr_id. Write it into the 1132 * attr_data pointer. 1133 * 1134 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1135 * If device supports VLAN filtering this function is called when a 1136 * VLAN id is registered. 1137 * 1138 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1139 * If device supports VLAN filtering this function is called when a 1140 * VLAN id is unregistered. 1141 * 1142 * void (*ndo_poll_controller)(struct net_device *dev); 1143 * 1144 * SR-IOV management functions. 1145 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1146 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1147 * u8 qos, __be16 proto); 1148 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1149 * int max_tx_rate); 1150 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1151 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1152 * int (*ndo_get_vf_config)(struct net_device *dev, 1153 * int vf, struct ifla_vf_info *ivf); 1154 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1155 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1156 * struct nlattr *port[]); 1157 * 1158 * Enable or disable the VF ability to query its RSS Redirection Table and 1159 * Hash Key. This is needed since on some devices VF share this information 1160 * with PF and querying it may introduce a theoretical security risk. 1161 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1162 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1163 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1164 * void *type_data); 1165 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1166 * This is always called from the stack with the rtnl lock held and netif 1167 * tx queues stopped. This allows the netdevice to perform queue 1168 * management safely. 1169 * 1170 * Fiber Channel over Ethernet (FCoE) offload functions. 1171 * int (*ndo_fcoe_enable)(struct net_device *dev); 1172 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1173 * so the underlying device can perform whatever needed configuration or 1174 * initialization to support acceleration of FCoE traffic. 1175 * 1176 * int (*ndo_fcoe_disable)(struct net_device *dev); 1177 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1178 * so the underlying device can perform whatever needed clean-ups to 1179 * stop supporting acceleration of FCoE traffic. 1180 * 1181 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1182 * struct scatterlist *sgl, unsigned int sgc); 1183 * Called when the FCoE Initiator wants to initialize an I/O that 1184 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1185 * perform necessary setup and returns 1 to indicate the device is set up 1186 * successfully to perform DDP on this I/O, otherwise this returns 0. 1187 * 1188 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1189 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1190 * indicated by the FC exchange id 'xid', so the underlying device can 1191 * clean up and reuse resources for later DDP requests. 1192 * 1193 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1194 * struct scatterlist *sgl, unsigned int sgc); 1195 * Called when the FCoE Target wants to initialize an I/O that 1196 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1197 * perform necessary setup and returns 1 to indicate the device is set up 1198 * successfully to perform DDP on this I/O, otherwise this returns 0. 1199 * 1200 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1201 * struct netdev_fcoe_hbainfo *hbainfo); 1202 * Called when the FCoE Protocol stack wants information on the underlying 1203 * device. This information is utilized by the FCoE protocol stack to 1204 * register attributes with Fiber Channel management service as per the 1205 * FC-GS Fabric Device Management Information(FDMI) specification. 1206 * 1207 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1208 * Called when the underlying device wants to override default World Wide 1209 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1210 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1211 * protocol stack to use. 1212 * 1213 * RFS acceleration. 1214 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1215 * u16 rxq_index, u32 flow_id); 1216 * Set hardware filter for RFS. rxq_index is the target queue index; 1217 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1218 * Return the filter ID on success, or a negative error code. 1219 * 1220 * Slave management functions (for bridge, bonding, etc). 1221 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1222 * Called to make another netdev an underling. 1223 * 1224 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1225 * Called to release previously enslaved netdev. 1226 * 1227 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1228 * struct sk_buff *skb, 1229 * bool all_slaves); 1230 * Get the xmit slave of master device. If all_slaves is true, function 1231 * assume all the slaves can transmit. 1232 * 1233 * Feature/offload setting functions. 1234 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1235 * netdev_features_t features); 1236 * Adjusts the requested feature flags according to device-specific 1237 * constraints, and returns the resulting flags. Must not modify 1238 * the device state. 1239 * 1240 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1241 * Called to update device configuration to new features. Passed 1242 * feature set might be less than what was returned by ndo_fix_features()). 1243 * Must return >0 or -errno if it changed dev->features itself. 1244 * 1245 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1246 * struct net_device *dev, 1247 * const unsigned char *addr, u16 vid, u16 flags, 1248 * struct netlink_ext_ack *extack); 1249 * Adds an FDB entry to dev for addr. 1250 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1251 * struct net_device *dev, 1252 * const unsigned char *addr, u16 vid) 1253 * Deletes the FDB entry from dev corresponding to addr. 1254 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev, 1255 * struct netlink_ext_ack *extack); 1256 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1257 * struct net_device *dev, struct net_device *filter_dev, 1258 * int *idx) 1259 * Used to add FDB entries to dump requests. Implementers should add 1260 * entries to skb and update idx with the number of entries. 1261 * 1262 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[], 1263 * u16 nlmsg_flags, struct netlink_ext_ack *extack); 1264 * Adds an MDB entry to dev. 1265 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[], 1266 * struct netlink_ext_ack *extack); 1267 * Deletes the MDB entry from dev. 1268 * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[], 1269 * struct netlink_ext_ack *extack); 1270 * Bulk deletes MDB entries from dev. 1271 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb, 1272 * struct netlink_callback *cb); 1273 * Dumps MDB entries from dev. The first argument (marker) in the netlink 1274 * callback is used by core rtnetlink code. 1275 * 1276 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1277 * u16 flags, struct netlink_ext_ack *extack) 1278 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1279 * struct net_device *dev, u32 filter_mask, 1280 * int nlflags) 1281 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1282 * u16 flags); 1283 * 1284 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1285 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1286 * which do not represent real hardware may define this to allow their 1287 * userspace components to manage their virtual carrier state. Devices 1288 * that determine carrier state from physical hardware properties (eg 1289 * network cables) or protocol-dependent mechanisms (eg 1290 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1291 * 1292 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1293 * struct netdev_phys_item_id *ppid); 1294 * Called to get ID of physical port of this device. If driver does 1295 * not implement this, it is assumed that the hw is not able to have 1296 * multiple net devices on single physical port. 1297 * 1298 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1299 * struct netdev_phys_item_id *ppid) 1300 * Called to get the parent ID of the physical port of this device. 1301 * 1302 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1303 * struct net_device *dev) 1304 * Called by upper layer devices to accelerate switching or other 1305 * station functionality into hardware. 'pdev is the lowerdev 1306 * to use for the offload and 'dev' is the net device that will 1307 * back the offload. Returns a pointer to the private structure 1308 * the upper layer will maintain. 1309 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1310 * Called by upper layer device to delete the station created 1311 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1312 * the station and priv is the structure returned by the add 1313 * operation. 1314 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1315 * int queue_index, u32 maxrate); 1316 * Called when a user wants to set a max-rate limitation of specific 1317 * TX queue. 1318 * int (*ndo_get_iflink)(const struct net_device *dev); 1319 * Called to get the iflink value of this device. 1320 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1321 * This function is used to get egress tunnel information for given skb. 1322 * This is useful for retrieving outer tunnel header parameters while 1323 * sampling packet. 1324 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1325 * This function is used to specify the headroom that the skb must 1326 * consider when allocation skb during packet reception. Setting 1327 * appropriate rx headroom value allows avoiding skb head copy on 1328 * forward. Setting a negative value resets the rx headroom to the 1329 * default value. 1330 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1331 * This function is used to set or query state related to XDP on the 1332 * netdevice and manage BPF offload. See definition of 1333 * enum bpf_netdev_command for details. 1334 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1335 * u32 flags); 1336 * This function is used to submit @n XDP packets for transmit on a 1337 * netdevice. Returns number of frames successfully transmitted, frames 1338 * that got dropped are freed/returned via xdp_return_frame(). 1339 * Returns negative number, means general error invoking ndo, meaning 1340 * no frames were xmit'ed and core-caller will free all frames. 1341 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1342 * struct xdp_buff *xdp); 1343 * Get the xmit slave of master device based on the xdp_buff. 1344 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1345 * This function is used to wake up the softirq, ksoftirqd or kthread 1346 * responsible for sending and/or receiving packets on a specific 1347 * queue id bound to an AF_XDP socket. The flags field specifies if 1348 * only RX, only Tx, or both should be woken up using the flags 1349 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1350 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm_kern *p, 1351 * int cmd); 1352 * Add, change, delete or get information on an IPv4 tunnel. 1353 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1354 * If a device is paired with a peer device, return the peer instance. 1355 * The caller must be under RCU read context. 1356 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1357 * Get the forwarding path to reach the real device from the HW destination address 1358 * ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1359 * const struct skb_shared_hwtstamps *hwtstamps, 1360 * bool cycles); 1361 * Get hardware timestamp based on normal/adjustable time or free running 1362 * cycle counter. This function is required if physical clock supports a 1363 * free running cycle counter. 1364 * 1365 * int (*ndo_hwtstamp_get)(struct net_device *dev, 1366 * struct kernel_hwtstamp_config *kernel_config); 1367 * Get the currently configured hardware timestamping parameters for the 1368 * NIC device. 1369 * 1370 * int (*ndo_hwtstamp_set)(struct net_device *dev, 1371 * struct kernel_hwtstamp_config *kernel_config, 1372 * struct netlink_ext_ack *extack); 1373 * Change the hardware timestamping parameters for NIC device. 1374 */ 1375 struct net_device_ops { 1376 int (*ndo_init)(struct net_device *dev); 1377 void (*ndo_uninit)(struct net_device *dev); 1378 int (*ndo_open)(struct net_device *dev); 1379 int (*ndo_stop)(struct net_device *dev); 1380 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1381 struct net_device *dev); 1382 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1383 struct net_device *dev, 1384 netdev_features_t features); 1385 u16 (*ndo_select_queue)(struct net_device *dev, 1386 struct sk_buff *skb, 1387 struct net_device *sb_dev); 1388 void (*ndo_change_rx_flags)(struct net_device *dev, 1389 int flags); 1390 void (*ndo_set_rx_mode)(struct net_device *dev); 1391 int (*ndo_set_mac_address)(struct net_device *dev, 1392 void *addr); 1393 int (*ndo_validate_addr)(struct net_device *dev); 1394 int (*ndo_do_ioctl)(struct net_device *dev, 1395 struct ifreq *ifr, int cmd); 1396 int (*ndo_eth_ioctl)(struct net_device *dev, 1397 struct ifreq *ifr, int cmd); 1398 int (*ndo_siocbond)(struct net_device *dev, 1399 struct ifreq *ifr, int cmd); 1400 int (*ndo_siocwandev)(struct net_device *dev, 1401 struct if_settings *ifs); 1402 int (*ndo_siocdevprivate)(struct net_device *dev, 1403 struct ifreq *ifr, 1404 void __user *data, int cmd); 1405 int (*ndo_set_config)(struct net_device *dev, 1406 struct ifmap *map); 1407 int (*ndo_change_mtu)(struct net_device *dev, 1408 int new_mtu); 1409 int (*ndo_neigh_setup)(struct net_device *dev, 1410 struct neigh_parms *); 1411 void (*ndo_tx_timeout) (struct net_device *dev, 1412 unsigned int txqueue); 1413 1414 void (*ndo_get_stats64)(struct net_device *dev, 1415 struct rtnl_link_stats64 *storage); 1416 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1417 int (*ndo_get_offload_stats)(int attr_id, 1418 const struct net_device *dev, 1419 void *attr_data); 1420 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1421 1422 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1423 __be16 proto, u16 vid); 1424 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1425 __be16 proto, u16 vid); 1426 #ifdef CONFIG_NET_POLL_CONTROLLER 1427 void (*ndo_poll_controller)(struct net_device *dev); 1428 int (*ndo_netpoll_setup)(struct net_device *dev); 1429 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1430 #endif 1431 int (*ndo_set_vf_mac)(struct net_device *dev, 1432 int queue, u8 *mac); 1433 int (*ndo_set_vf_vlan)(struct net_device *dev, 1434 int queue, u16 vlan, 1435 u8 qos, __be16 proto); 1436 int (*ndo_set_vf_rate)(struct net_device *dev, 1437 int vf, int min_tx_rate, 1438 int max_tx_rate); 1439 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1440 int vf, bool setting); 1441 int (*ndo_set_vf_trust)(struct net_device *dev, 1442 int vf, bool setting); 1443 int (*ndo_get_vf_config)(struct net_device *dev, 1444 int vf, 1445 struct ifla_vf_info *ivf); 1446 int (*ndo_set_vf_link_state)(struct net_device *dev, 1447 int vf, int link_state); 1448 int (*ndo_get_vf_stats)(struct net_device *dev, 1449 int vf, 1450 struct ifla_vf_stats 1451 *vf_stats); 1452 int (*ndo_set_vf_port)(struct net_device *dev, 1453 int vf, 1454 struct nlattr *port[]); 1455 int (*ndo_get_vf_port)(struct net_device *dev, 1456 int vf, struct sk_buff *skb); 1457 int (*ndo_get_vf_guid)(struct net_device *dev, 1458 int vf, 1459 struct ifla_vf_guid *node_guid, 1460 struct ifla_vf_guid *port_guid); 1461 int (*ndo_set_vf_guid)(struct net_device *dev, 1462 int vf, u64 guid, 1463 int guid_type); 1464 int (*ndo_set_vf_rss_query_en)( 1465 struct net_device *dev, 1466 int vf, bool setting); 1467 int (*ndo_setup_tc)(struct net_device *dev, 1468 enum tc_setup_type type, 1469 void *type_data); 1470 #if IS_ENABLED(CONFIG_FCOE) 1471 int (*ndo_fcoe_enable)(struct net_device *dev); 1472 int (*ndo_fcoe_disable)(struct net_device *dev); 1473 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1474 u16 xid, 1475 struct scatterlist *sgl, 1476 unsigned int sgc); 1477 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1478 u16 xid); 1479 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1480 u16 xid, 1481 struct scatterlist *sgl, 1482 unsigned int sgc); 1483 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1484 struct netdev_fcoe_hbainfo *hbainfo); 1485 #endif 1486 1487 #if IS_ENABLED(CONFIG_LIBFCOE) 1488 #define NETDEV_FCOE_WWNN 0 1489 #define NETDEV_FCOE_WWPN 1 1490 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1491 u64 *wwn, int type); 1492 #endif 1493 1494 #ifdef CONFIG_RFS_ACCEL 1495 int (*ndo_rx_flow_steer)(struct net_device *dev, 1496 const struct sk_buff *skb, 1497 u16 rxq_index, 1498 u32 flow_id); 1499 #endif 1500 int (*ndo_add_slave)(struct net_device *dev, 1501 struct net_device *slave_dev, 1502 struct netlink_ext_ack *extack); 1503 int (*ndo_del_slave)(struct net_device *dev, 1504 struct net_device *slave_dev); 1505 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1506 struct sk_buff *skb, 1507 bool all_slaves); 1508 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1509 struct sock *sk); 1510 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1511 netdev_features_t features); 1512 int (*ndo_set_features)(struct net_device *dev, 1513 netdev_features_t features); 1514 int (*ndo_neigh_construct)(struct net_device *dev, 1515 struct neighbour *n); 1516 void (*ndo_neigh_destroy)(struct net_device *dev, 1517 struct neighbour *n); 1518 1519 int (*ndo_fdb_add)(struct ndmsg *ndm, 1520 struct nlattr *tb[], 1521 struct net_device *dev, 1522 const unsigned char *addr, 1523 u16 vid, 1524 u16 flags, 1525 struct netlink_ext_ack *extack); 1526 int (*ndo_fdb_del)(struct ndmsg *ndm, 1527 struct nlattr *tb[], 1528 struct net_device *dev, 1529 const unsigned char *addr, 1530 u16 vid, struct netlink_ext_ack *extack); 1531 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, 1532 struct net_device *dev, 1533 struct netlink_ext_ack *extack); 1534 int (*ndo_fdb_dump)(struct sk_buff *skb, 1535 struct netlink_callback *cb, 1536 struct net_device *dev, 1537 struct net_device *filter_dev, 1538 int *idx); 1539 int (*ndo_fdb_get)(struct sk_buff *skb, 1540 struct nlattr *tb[], 1541 struct net_device *dev, 1542 const unsigned char *addr, 1543 u16 vid, u32 portid, u32 seq, 1544 struct netlink_ext_ack *extack); 1545 int (*ndo_mdb_add)(struct net_device *dev, 1546 struct nlattr *tb[], 1547 u16 nlmsg_flags, 1548 struct netlink_ext_ack *extack); 1549 int (*ndo_mdb_del)(struct net_device *dev, 1550 struct nlattr *tb[], 1551 struct netlink_ext_ack *extack); 1552 int (*ndo_mdb_del_bulk)(struct net_device *dev, 1553 struct nlattr *tb[], 1554 struct netlink_ext_ack *extack); 1555 int (*ndo_mdb_dump)(struct net_device *dev, 1556 struct sk_buff *skb, 1557 struct netlink_callback *cb); 1558 int (*ndo_mdb_get)(struct net_device *dev, 1559 struct nlattr *tb[], u32 portid, 1560 u32 seq, 1561 struct netlink_ext_ack *extack); 1562 int (*ndo_bridge_setlink)(struct net_device *dev, 1563 struct nlmsghdr *nlh, 1564 u16 flags, 1565 struct netlink_ext_ack *extack); 1566 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1567 u32 pid, u32 seq, 1568 struct net_device *dev, 1569 u32 filter_mask, 1570 int nlflags); 1571 int (*ndo_bridge_dellink)(struct net_device *dev, 1572 struct nlmsghdr *nlh, 1573 u16 flags); 1574 int (*ndo_change_carrier)(struct net_device *dev, 1575 bool new_carrier); 1576 int (*ndo_get_phys_port_id)(struct net_device *dev, 1577 struct netdev_phys_item_id *ppid); 1578 int (*ndo_get_port_parent_id)(struct net_device *dev, 1579 struct netdev_phys_item_id *ppid); 1580 int (*ndo_get_phys_port_name)(struct net_device *dev, 1581 char *name, size_t len); 1582 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1583 struct net_device *dev); 1584 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1585 void *priv); 1586 1587 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1588 int queue_index, 1589 u32 maxrate); 1590 int (*ndo_get_iflink)(const struct net_device *dev); 1591 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1592 struct sk_buff *skb); 1593 void (*ndo_set_rx_headroom)(struct net_device *dev, 1594 int needed_headroom); 1595 int (*ndo_bpf)(struct net_device *dev, 1596 struct netdev_bpf *bpf); 1597 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1598 struct xdp_frame **xdp, 1599 u32 flags); 1600 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1601 struct xdp_buff *xdp); 1602 int (*ndo_xsk_wakeup)(struct net_device *dev, 1603 u32 queue_id, u32 flags); 1604 int (*ndo_tunnel_ctl)(struct net_device *dev, 1605 struct ip_tunnel_parm_kern *p, 1606 int cmd); 1607 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1608 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1609 struct net_device_path *path); 1610 ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1611 const struct skb_shared_hwtstamps *hwtstamps, 1612 bool cycles); 1613 int (*ndo_hwtstamp_get)(struct net_device *dev, 1614 struct kernel_hwtstamp_config *kernel_config); 1615 int (*ndo_hwtstamp_set)(struct net_device *dev, 1616 struct kernel_hwtstamp_config *kernel_config, 1617 struct netlink_ext_ack *extack); 1618 1619 #if IS_ENABLED(CONFIG_NET_SHAPER) 1620 /** 1621 * @net_shaper_ops: Device shaping offload operations 1622 * see include/net/net_shapers.h 1623 */ 1624 const struct net_shaper_ops *net_shaper_ops; 1625 #endif 1626 }; 1627 1628 /** 1629 * enum netdev_priv_flags - &struct net_device priv_flags 1630 * 1631 * These are the &struct net_device, they are only set internally 1632 * by drivers and used in the kernel. These flags are invisible to 1633 * userspace; this means that the order of these flags can change 1634 * during any kernel release. 1635 * 1636 * You should add bitfield booleans after either net_device::priv_flags 1637 * (hotpath) or ::threaded (slowpath) instead of extending these flags. 1638 * 1639 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1640 * @IFF_EBRIDGE: Ethernet bridging device 1641 * @IFF_BONDING: bonding master or slave 1642 * @IFF_ISATAP: ISATAP interface (RFC4214) 1643 * @IFF_WAN_HDLC: WAN HDLC device 1644 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1645 * release skb->dst 1646 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1647 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1648 * @IFF_MACVLAN_PORT: device used as macvlan port 1649 * @IFF_BRIDGE_PORT: device used as bridge port 1650 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1651 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1652 * @IFF_UNICAST_FLT: Supports unicast filtering 1653 * @IFF_TEAM_PORT: device used as team port 1654 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1655 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1656 * change when it's running 1657 * @IFF_MACVLAN: Macvlan device 1658 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1659 * underlying stacked devices 1660 * @IFF_L3MDEV_MASTER: device is an L3 master device 1661 * @IFF_NO_QUEUE: device can run without qdisc attached 1662 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1663 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1664 * @IFF_TEAM: device is a team device 1665 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1666 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1667 * entity (i.e. the master device for bridged veth) 1668 * @IFF_MACSEC: device is a MACsec device 1669 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1670 * @IFF_FAILOVER: device is a failover master device 1671 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1672 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1673 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf 1674 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1675 * skb_headlen(skb) == 0 (data starts from frag0) 1676 */ 1677 enum netdev_priv_flags { 1678 IFF_802_1Q_VLAN = 1<<0, 1679 IFF_EBRIDGE = 1<<1, 1680 IFF_BONDING = 1<<2, 1681 IFF_ISATAP = 1<<3, 1682 IFF_WAN_HDLC = 1<<4, 1683 IFF_XMIT_DST_RELEASE = 1<<5, 1684 IFF_DONT_BRIDGE = 1<<6, 1685 IFF_DISABLE_NETPOLL = 1<<7, 1686 IFF_MACVLAN_PORT = 1<<8, 1687 IFF_BRIDGE_PORT = 1<<9, 1688 IFF_OVS_DATAPATH = 1<<10, 1689 IFF_TX_SKB_SHARING = 1<<11, 1690 IFF_UNICAST_FLT = 1<<12, 1691 IFF_TEAM_PORT = 1<<13, 1692 IFF_SUPP_NOFCS = 1<<14, 1693 IFF_LIVE_ADDR_CHANGE = 1<<15, 1694 IFF_MACVLAN = 1<<16, 1695 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1696 IFF_L3MDEV_MASTER = 1<<18, 1697 IFF_NO_QUEUE = 1<<19, 1698 IFF_OPENVSWITCH = 1<<20, 1699 IFF_L3MDEV_SLAVE = 1<<21, 1700 IFF_TEAM = 1<<22, 1701 IFF_RXFH_CONFIGURED = 1<<23, 1702 IFF_PHONY_HEADROOM = 1<<24, 1703 IFF_MACSEC = 1<<25, 1704 IFF_NO_RX_HANDLER = 1<<26, 1705 IFF_FAILOVER = 1<<27, 1706 IFF_FAILOVER_SLAVE = 1<<28, 1707 IFF_L3MDEV_RX_HANDLER = 1<<29, 1708 IFF_NO_ADDRCONF = BIT_ULL(30), 1709 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31), 1710 }; 1711 1712 /* Specifies the type of the struct net_device::ml_priv pointer */ 1713 enum netdev_ml_priv_type { 1714 ML_PRIV_NONE, 1715 ML_PRIV_CAN, 1716 }; 1717 1718 enum netdev_stat_type { 1719 NETDEV_PCPU_STAT_NONE, 1720 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */ 1721 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */ 1722 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */ 1723 }; 1724 1725 enum netdev_reg_state { 1726 NETREG_UNINITIALIZED = 0, 1727 NETREG_REGISTERED, /* completed register_netdevice */ 1728 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1729 NETREG_UNREGISTERED, /* completed unregister todo */ 1730 NETREG_RELEASED, /* called free_netdev */ 1731 NETREG_DUMMY, /* dummy device for NAPI poll */ 1732 }; 1733 1734 /** 1735 * struct net_device - The DEVICE structure. 1736 * 1737 * Actually, this whole structure is a big mistake. It mixes I/O 1738 * data with strictly "high-level" data, and it has to know about 1739 * almost every data structure used in the INET module. 1740 * 1741 * @priv_flags: flags invisible to userspace defined as bits, see 1742 * enum netdev_priv_flags for the definitions 1743 * @lltx: device supports lockless Tx. Deprecated for real HW 1744 * drivers. Mainly used by logical interfaces, such as 1745 * bonding and tunnels 1746 * 1747 * @name: This is the first field of the "visible" part of this structure 1748 * (i.e. as seen by users in the "Space.c" file). It is the name 1749 * of the interface. 1750 * 1751 * @name_node: Name hashlist node 1752 * @ifalias: SNMP alias 1753 * @mem_end: Shared memory end 1754 * @mem_start: Shared memory start 1755 * @base_addr: Device I/O address 1756 * @irq: Device IRQ number 1757 * 1758 * @state: Generic network queuing layer state, see netdev_state_t 1759 * @dev_list: The global list of network devices 1760 * @napi_list: List entry used for polling NAPI devices 1761 * @unreg_list: List entry when we are unregistering the 1762 * device; see the function unregister_netdev 1763 * @close_list: List entry used when we are closing the device 1764 * @ptype_all: Device-specific packet handlers for all protocols 1765 * @ptype_specific: Device-specific, protocol-specific packet handlers 1766 * 1767 * @adj_list: Directly linked devices, like slaves for bonding 1768 * @features: Currently active device features 1769 * @hw_features: User-changeable features 1770 * 1771 * @wanted_features: User-requested features 1772 * @vlan_features: Mask of features inheritable by VLAN devices 1773 * 1774 * @hw_enc_features: Mask of features inherited by encapsulating devices 1775 * This field indicates what encapsulation 1776 * offloads the hardware is capable of doing, 1777 * and drivers will need to set them appropriately. 1778 * 1779 * @mpls_features: Mask of features inheritable by MPLS 1780 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1781 * 1782 * @ifindex: interface index 1783 * @group: The group the device belongs to 1784 * 1785 * @stats: Statistics struct, which was left as a legacy, use 1786 * rtnl_link_stats64 instead 1787 * 1788 * @core_stats: core networking counters, 1789 * do not use this in drivers 1790 * @carrier_up_count: Number of times the carrier has been up 1791 * @carrier_down_count: Number of times the carrier has been down 1792 * 1793 * @wireless_handlers: List of functions to handle Wireless Extensions, 1794 * instead of ioctl, 1795 * see <net/iw_handler.h> for details. 1796 * @wireless_data: Instance data managed by the core of wireless extensions 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 * FIXME: cleanup struct net_device such that network protocol info 2037 * moves out. 2038 */ 2039 2040 struct net_device { 2041 /* Cacheline organization can be found documented in 2042 * Documentation/networking/net_cachelines/net_device.rst. 2043 * Please update the document when adding new fields. 2044 */ 2045 2046 /* TX read-mostly hotpath */ 2047 __cacheline_group_begin(net_device_read_tx); 2048 struct_group(priv_flags_fast, 2049 unsigned long priv_flags:32; 2050 unsigned long lltx:1; 2051 ); 2052 const struct net_device_ops *netdev_ops; 2053 const struct header_ops *header_ops; 2054 struct netdev_queue *_tx; 2055 netdev_features_t gso_partial_features; 2056 unsigned int real_num_tx_queues; 2057 unsigned int gso_max_size; 2058 unsigned int gso_ipv4_max_size; 2059 u16 gso_max_segs; 2060 s16 num_tc; 2061 /* Note : dev->mtu is often read without holding a lock. 2062 * Writers usually hold RTNL. 2063 * It is recommended to use READ_ONCE() to annotate the reads, 2064 * and to use WRITE_ONCE() to annotate the writes. 2065 */ 2066 unsigned int mtu; 2067 unsigned short needed_headroom; 2068 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2069 #ifdef CONFIG_XPS 2070 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2071 #endif 2072 #ifdef CONFIG_NETFILTER_EGRESS 2073 struct nf_hook_entries __rcu *nf_hooks_egress; 2074 #endif 2075 #ifdef CONFIG_NET_XGRESS 2076 struct bpf_mprog_entry __rcu *tcx_egress; 2077 #endif 2078 __cacheline_group_end(net_device_read_tx); 2079 2080 /* TXRX read-mostly hotpath */ 2081 __cacheline_group_begin(net_device_read_txrx); 2082 union { 2083 struct pcpu_lstats __percpu *lstats; 2084 struct pcpu_sw_netstats __percpu *tstats; 2085 struct pcpu_dstats __percpu *dstats; 2086 }; 2087 unsigned long state; 2088 unsigned int flags; 2089 unsigned short hard_header_len; 2090 netdev_features_t features; 2091 struct inet6_dev __rcu *ip6_ptr; 2092 __cacheline_group_end(net_device_read_txrx); 2093 2094 /* RX read-mostly hotpath */ 2095 __cacheline_group_begin(net_device_read_rx); 2096 struct bpf_prog __rcu *xdp_prog; 2097 struct list_head ptype_specific; 2098 int ifindex; 2099 unsigned int real_num_rx_queues; 2100 struct netdev_rx_queue *_rx; 2101 unsigned int gro_max_size; 2102 unsigned int gro_ipv4_max_size; 2103 rx_handler_func_t __rcu *rx_handler; 2104 void __rcu *rx_handler_data; 2105 possible_net_t nd_net; 2106 #ifdef CONFIG_NETPOLL 2107 struct netpoll_info __rcu *npinfo; 2108 #endif 2109 #ifdef CONFIG_NET_XGRESS 2110 struct bpf_mprog_entry __rcu *tcx_ingress; 2111 #endif 2112 __cacheline_group_end(net_device_read_rx); 2113 2114 char name[IFNAMSIZ]; 2115 struct netdev_name_node *name_node; 2116 struct dev_ifalias __rcu *ifalias; 2117 /* 2118 * I/O specific fields 2119 * FIXME: Merge these and struct ifmap into one 2120 */ 2121 unsigned long mem_end; 2122 unsigned long mem_start; 2123 unsigned long base_addr; 2124 2125 /* 2126 * Some hardware also needs these fields (state,dev_list, 2127 * napi_list,unreg_list,close_list) but they are not 2128 * part of the usual set specified in Space.c. 2129 */ 2130 2131 2132 struct list_head dev_list; 2133 struct list_head napi_list; 2134 struct list_head unreg_list; 2135 struct list_head close_list; 2136 struct list_head ptype_all; 2137 2138 struct { 2139 struct list_head upper; 2140 struct list_head lower; 2141 } adj_list; 2142 2143 /* Read-mostly cache-line for fast-path access */ 2144 xdp_features_t xdp_features; 2145 const struct xdp_metadata_ops *xdp_metadata_ops; 2146 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops; 2147 unsigned short gflags; 2148 2149 unsigned short needed_tailroom; 2150 2151 netdev_features_t hw_features; 2152 netdev_features_t wanted_features; 2153 netdev_features_t vlan_features; 2154 netdev_features_t hw_enc_features; 2155 netdev_features_t mpls_features; 2156 2157 unsigned int min_mtu; 2158 unsigned int max_mtu; 2159 unsigned short type; 2160 unsigned char min_header_len; 2161 unsigned char name_assign_type; 2162 2163 int group; 2164 2165 struct net_device_stats stats; /* not used by modern drivers */ 2166 2167 struct net_device_core_stats __percpu *core_stats; 2168 2169 /* Stats to monitor link on/off, flapping */ 2170 atomic_t carrier_up_count; 2171 atomic_t carrier_down_count; 2172 2173 #ifdef CONFIG_WIRELESS_EXT 2174 const struct iw_handler_def *wireless_handlers; 2175 struct iw_public_data *wireless_data; 2176 #endif 2177 const struct ethtool_ops *ethtool_ops; 2178 #ifdef CONFIG_NET_L3_MASTER_DEV 2179 const struct l3mdev_ops *l3mdev_ops; 2180 #endif 2181 #if IS_ENABLED(CONFIG_IPV6) 2182 const struct ndisc_ops *ndisc_ops; 2183 #endif 2184 2185 #ifdef CONFIG_XFRM_OFFLOAD 2186 const struct xfrmdev_ops *xfrmdev_ops; 2187 #endif 2188 2189 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2190 const struct tlsdev_ops *tlsdev_ops; 2191 #endif 2192 2193 unsigned int operstate; 2194 unsigned char link_mode; 2195 2196 unsigned char if_port; 2197 unsigned char dma; 2198 2199 /* Interface address info. */ 2200 unsigned char perm_addr[MAX_ADDR_LEN]; 2201 unsigned char addr_assign_type; 2202 unsigned char addr_len; 2203 unsigned char upper_level; 2204 unsigned char lower_level; 2205 2206 unsigned short neigh_priv_len; 2207 unsigned short dev_id; 2208 unsigned short dev_port; 2209 int irq; 2210 u32 priv_len; 2211 2212 spinlock_t addr_list_lock; 2213 2214 struct netdev_hw_addr_list uc; 2215 struct netdev_hw_addr_list mc; 2216 struct netdev_hw_addr_list dev_addrs; 2217 2218 #ifdef CONFIG_SYSFS 2219 struct kset *queues_kset; 2220 #endif 2221 #ifdef CONFIG_LOCKDEP 2222 struct list_head unlink_list; 2223 #endif 2224 unsigned int promiscuity; 2225 unsigned int allmulti; 2226 bool uc_promisc; 2227 #ifdef CONFIG_LOCKDEP 2228 unsigned char nested_level; 2229 #endif 2230 2231 2232 /* Protocol-specific pointers */ 2233 struct in_device __rcu *ip_ptr; 2234 /** @fib_nh_head: nexthops associated with this netdev */ 2235 struct hlist_head fib_nh_head; 2236 2237 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2238 struct vlan_info __rcu *vlan_info; 2239 #endif 2240 #if IS_ENABLED(CONFIG_NET_DSA) 2241 struct dsa_port *dsa_ptr; 2242 #endif 2243 #if IS_ENABLED(CONFIG_TIPC) 2244 struct tipc_bearer __rcu *tipc_ptr; 2245 #endif 2246 #if IS_ENABLED(CONFIG_ATALK) 2247 void *atalk_ptr; 2248 #endif 2249 #if IS_ENABLED(CONFIG_AX25) 2250 void *ax25_ptr; 2251 #endif 2252 #if IS_ENABLED(CONFIG_CFG80211) 2253 struct wireless_dev *ieee80211_ptr; 2254 #endif 2255 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2256 struct wpan_dev *ieee802154_ptr; 2257 #endif 2258 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2259 struct mpls_dev __rcu *mpls_ptr; 2260 #endif 2261 #if IS_ENABLED(CONFIG_MCTP) 2262 struct mctp_dev __rcu *mctp_ptr; 2263 #endif 2264 2265 /* 2266 * Cache lines mostly used on receive path (including eth_type_trans()) 2267 */ 2268 /* Interface address info used in eth_type_trans() */ 2269 const unsigned char *dev_addr; 2270 2271 unsigned int num_rx_queues; 2272 #define GRO_LEGACY_MAX_SIZE 65536u 2273 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2274 * and shinfo->gso_segs is a 16bit field. 2275 */ 2276 #define GRO_MAX_SIZE (8 * 65535u) 2277 unsigned int xdp_zc_max_segs; 2278 struct netdev_queue __rcu *ingress_queue; 2279 #ifdef CONFIG_NETFILTER_INGRESS 2280 struct nf_hook_entries __rcu *nf_hooks_ingress; 2281 #endif 2282 2283 unsigned char broadcast[MAX_ADDR_LEN]; 2284 #ifdef CONFIG_RFS_ACCEL 2285 struct cpu_rmap *rx_cpu_rmap; 2286 #endif 2287 struct hlist_node index_hlist; 2288 2289 /* 2290 * Cache lines mostly used on transmit path 2291 */ 2292 unsigned int num_tx_queues; 2293 struct Qdisc __rcu *qdisc; 2294 unsigned int tx_queue_len; 2295 spinlock_t tx_global_lock; 2296 2297 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2298 2299 #ifdef CONFIG_NET_SCHED 2300 DECLARE_HASHTABLE (qdisc_hash, 4); 2301 #endif 2302 /* These may be needed for future network-power-down code. */ 2303 struct timer_list watchdog_timer; 2304 int watchdog_timeo; 2305 2306 u32 proto_down_reason; 2307 2308 struct list_head todo_list; 2309 2310 #ifdef CONFIG_PCPU_DEV_REFCNT 2311 int __percpu *pcpu_refcnt; 2312 #else 2313 refcount_t dev_refcnt; 2314 #endif 2315 struct ref_tracker_dir refcnt_tracker; 2316 2317 struct list_head link_watch_list; 2318 2319 u8 reg_state; 2320 2321 bool dismantle; 2322 2323 enum { 2324 RTNL_LINK_INITIALIZED, 2325 RTNL_LINK_INITIALIZING, 2326 } rtnl_link_state:16; 2327 2328 bool needs_free_netdev; 2329 void (*priv_destructor)(struct net_device *dev); 2330 2331 /* mid-layer private */ 2332 void *ml_priv; 2333 enum netdev_ml_priv_type ml_priv_type; 2334 2335 enum netdev_stat_type pcpu_stat_type:8; 2336 2337 #if IS_ENABLED(CONFIG_GARP) 2338 struct garp_port __rcu *garp_port; 2339 #endif 2340 #if IS_ENABLED(CONFIG_MRP) 2341 struct mrp_port __rcu *mrp_port; 2342 #endif 2343 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2344 struct dm_hw_stat_delta __rcu *dm_private; 2345 #endif 2346 struct device dev; 2347 const struct attribute_group *sysfs_groups[4]; 2348 const struct attribute_group *sysfs_rx_queue_group; 2349 2350 const struct rtnl_link_ops *rtnl_link_ops; 2351 2352 const struct netdev_stat_ops *stat_ops; 2353 2354 const struct netdev_queue_mgmt_ops *queue_mgmt_ops; 2355 2356 /* for setting kernel sock attribute on TCP connection setup */ 2357 #define GSO_MAX_SEGS 65535u 2358 #define GSO_LEGACY_MAX_SIZE 65536u 2359 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2360 * and shinfo->gso_segs is a 16bit field. 2361 */ 2362 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2363 2364 #define TSO_LEGACY_MAX_SIZE 65536 2365 #define TSO_MAX_SIZE UINT_MAX 2366 unsigned int tso_max_size; 2367 #define TSO_MAX_SEGS U16_MAX 2368 u16 tso_max_segs; 2369 2370 #ifdef CONFIG_DCB 2371 const struct dcbnl_rtnl_ops *dcbnl_ops; 2372 #endif 2373 u8 prio_tc_map[TC_BITMASK + 1]; 2374 2375 #if IS_ENABLED(CONFIG_FCOE) 2376 unsigned int fcoe_ddp_xid; 2377 #endif 2378 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2379 struct netprio_map __rcu *priomap; 2380 #endif 2381 struct phy_link_topology *link_topo; 2382 struct phy_device *phydev; 2383 struct sfp_bus *sfp_bus; 2384 struct lock_class_key *qdisc_tx_busylock; 2385 bool proto_down; 2386 bool threaded; 2387 2388 /* priv_flags_slow, ungrouped to save space */ 2389 unsigned long see_all_hwtstamp_requests:1; 2390 unsigned long change_proto_down:1; 2391 unsigned long netns_local:1; 2392 unsigned long fcoe_mtu:1; 2393 2394 struct list_head net_notifier_list; 2395 2396 #if IS_ENABLED(CONFIG_MACSEC) 2397 /* MACsec management functions */ 2398 const struct macsec_ops *macsec_ops; 2399 #endif 2400 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2401 struct udp_tunnel_nic *udp_tunnel_nic; 2402 2403 struct ethtool_netdev_state *ethtool; 2404 2405 /* protected by rtnl_lock */ 2406 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2407 2408 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2409 netdevice_tracker linkwatch_dev_tracker; 2410 netdevice_tracker watchdog_dev_tracker; 2411 netdevice_tracker dev_registered_tracker; 2412 struct rtnl_hw_stats64 *offload_xstats_l3; 2413 2414 struct devlink_port *devlink_port; 2415 2416 #if IS_ENABLED(CONFIG_DPLL) 2417 struct dpll_pin __rcu *dpll_pin; 2418 #endif 2419 #if IS_ENABLED(CONFIG_PAGE_POOL) 2420 /** @page_pools: page pools created for this netdevice */ 2421 struct hlist_head page_pools; 2422 #endif 2423 2424 /** @irq_moder: dim parameters used if IS_ENABLED(CONFIG_DIMLIB). */ 2425 struct dim_irq_moder *irq_moder; 2426 2427 u64 max_pacing_offload_horizon; 2428 struct napi_config *napi_config; 2429 unsigned long gro_flush_timeout; 2430 u32 napi_defer_hard_irqs; 2431 2432 /** 2433 * @lock: protects @net_shaper_hierarchy, feel free to use for other 2434 * netdev-scope protection. Ordering: take after rtnl_lock. 2435 */ 2436 struct mutex lock; 2437 2438 #if IS_ENABLED(CONFIG_NET_SHAPER) 2439 /** 2440 * @net_shaper_hierarchy: data tracking the current shaper status 2441 * see include/net/net_shapers.h 2442 */ 2443 struct net_shaper_hierarchy *net_shaper_hierarchy; 2444 #endif 2445 u8 priv[] ____cacheline_aligned 2446 __counted_by(priv_len); 2447 } ____cacheline_aligned; 2448 #define to_net_dev(d) container_of(d, struct net_device, dev) 2449 2450 /* 2451 * Driver should use this to assign devlink port instance to a netdevice 2452 * before it registers the netdevice. Therefore devlink_port is static 2453 * during the netdev lifetime after it is registered. 2454 */ 2455 #define SET_NETDEV_DEVLINK_PORT(dev, port) \ 2456 ({ \ 2457 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \ 2458 ((dev)->devlink_port = (port)); \ 2459 }) 2460 2461 static inline bool netif_elide_gro(const struct net_device *dev) 2462 { 2463 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2464 return true; 2465 return false; 2466 } 2467 2468 #define NETDEV_ALIGN 32 2469 2470 static inline 2471 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2472 { 2473 return dev->prio_tc_map[prio & TC_BITMASK]; 2474 } 2475 2476 static inline 2477 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2478 { 2479 if (tc >= dev->num_tc) 2480 return -EINVAL; 2481 2482 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2483 return 0; 2484 } 2485 2486 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2487 void netdev_reset_tc(struct net_device *dev); 2488 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2489 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2490 2491 static inline 2492 int netdev_get_num_tc(struct net_device *dev) 2493 { 2494 return dev->num_tc; 2495 } 2496 2497 static inline void net_prefetch(void *p) 2498 { 2499 prefetch(p); 2500 #if L1_CACHE_BYTES < 128 2501 prefetch((u8 *)p + L1_CACHE_BYTES); 2502 #endif 2503 } 2504 2505 static inline void net_prefetchw(void *p) 2506 { 2507 prefetchw(p); 2508 #if L1_CACHE_BYTES < 128 2509 prefetchw((u8 *)p + L1_CACHE_BYTES); 2510 #endif 2511 } 2512 2513 void netdev_unbind_sb_channel(struct net_device *dev, 2514 struct net_device *sb_dev); 2515 int netdev_bind_sb_channel_queue(struct net_device *dev, 2516 struct net_device *sb_dev, 2517 u8 tc, u16 count, u16 offset); 2518 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2519 static inline int netdev_get_sb_channel(struct net_device *dev) 2520 { 2521 return max_t(int, -dev->num_tc, 0); 2522 } 2523 2524 static inline 2525 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2526 unsigned int index) 2527 { 2528 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues); 2529 return &dev->_tx[index]; 2530 } 2531 2532 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2533 const struct sk_buff *skb) 2534 { 2535 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2536 } 2537 2538 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2539 void (*f)(struct net_device *, 2540 struct netdev_queue *, 2541 void *), 2542 void *arg) 2543 { 2544 unsigned int i; 2545 2546 for (i = 0; i < dev->num_tx_queues; i++) 2547 f(dev, &dev->_tx[i], arg); 2548 } 2549 2550 #define netdev_lockdep_set_classes(dev) \ 2551 { \ 2552 static struct lock_class_key qdisc_tx_busylock_key; \ 2553 static struct lock_class_key qdisc_xmit_lock_key; \ 2554 static struct lock_class_key dev_addr_list_lock_key; \ 2555 unsigned int i; \ 2556 \ 2557 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2558 lockdep_set_class(&(dev)->addr_list_lock, \ 2559 &dev_addr_list_lock_key); \ 2560 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2561 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2562 &qdisc_xmit_lock_key); \ 2563 } 2564 2565 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2566 struct net_device *sb_dev); 2567 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2568 struct sk_buff *skb, 2569 struct net_device *sb_dev); 2570 2571 /* returns the headroom that the master device needs to take in account 2572 * when forwarding to this dev 2573 */ 2574 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2575 { 2576 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2577 } 2578 2579 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2580 { 2581 if (dev->netdev_ops->ndo_set_rx_headroom) 2582 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2583 } 2584 2585 /* set the device rx headroom to the dev's default */ 2586 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2587 { 2588 netdev_set_rx_headroom(dev, -1); 2589 } 2590 2591 static inline void *netdev_get_ml_priv(struct net_device *dev, 2592 enum netdev_ml_priv_type type) 2593 { 2594 if (dev->ml_priv_type != type) 2595 return NULL; 2596 2597 return dev->ml_priv; 2598 } 2599 2600 static inline void netdev_set_ml_priv(struct net_device *dev, 2601 void *ml_priv, 2602 enum netdev_ml_priv_type type) 2603 { 2604 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2605 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2606 dev->ml_priv_type, type); 2607 WARN(!dev->ml_priv_type && dev->ml_priv, 2608 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2609 2610 dev->ml_priv = ml_priv; 2611 dev->ml_priv_type = type; 2612 } 2613 2614 /* 2615 * Net namespace inlines 2616 */ 2617 static inline 2618 struct net *dev_net(const struct net_device *dev) 2619 { 2620 return read_pnet(&dev->nd_net); 2621 } 2622 2623 static inline 2624 void dev_net_set(struct net_device *dev, struct net *net) 2625 { 2626 write_pnet(&dev->nd_net, net); 2627 } 2628 2629 /** 2630 * netdev_priv - access network device private data 2631 * @dev: network device 2632 * 2633 * Get network device private data 2634 */ 2635 static inline void *netdev_priv(const struct net_device *dev) 2636 { 2637 return (void *)dev->priv; 2638 } 2639 2640 /* Set the sysfs physical device reference for the network logical device 2641 * if set prior to registration will cause a symlink during initialization. 2642 */ 2643 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2644 2645 /* Set the sysfs device type for the network logical device to allow 2646 * fine-grained identification of different network device types. For 2647 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2648 */ 2649 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2650 2651 void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index, 2652 enum netdev_queue_type type, 2653 struct napi_struct *napi); 2654 2655 static inline void netif_napi_set_irq(struct napi_struct *napi, int irq) 2656 { 2657 napi->irq = irq; 2658 } 2659 2660 /* Default NAPI poll() weight 2661 * Device drivers are strongly advised to not use bigger value 2662 */ 2663 #define NAPI_POLL_WEIGHT 64 2664 2665 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2666 int (*poll)(struct napi_struct *, int), int weight); 2667 2668 /** 2669 * netif_napi_add() - initialize a NAPI context 2670 * @dev: network device 2671 * @napi: NAPI context 2672 * @poll: polling function 2673 * 2674 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2675 * *any* of the other NAPI-related functions. 2676 */ 2677 static inline void 2678 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2679 int (*poll)(struct napi_struct *, int)) 2680 { 2681 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2682 } 2683 2684 static inline void 2685 netif_napi_add_tx_weight(struct net_device *dev, 2686 struct napi_struct *napi, 2687 int (*poll)(struct napi_struct *, int), 2688 int weight) 2689 { 2690 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2691 netif_napi_add_weight(dev, napi, poll, weight); 2692 } 2693 2694 /** 2695 * netif_napi_add_config - initialize a NAPI context with persistent config 2696 * @dev: network device 2697 * @napi: NAPI context 2698 * @poll: polling function 2699 * @index: the NAPI index 2700 */ 2701 static inline void 2702 netif_napi_add_config(struct net_device *dev, struct napi_struct *napi, 2703 int (*poll)(struct napi_struct *, int), int index) 2704 { 2705 napi->index = index; 2706 napi->config = &dev->napi_config[index]; 2707 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2708 } 2709 2710 /** 2711 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2712 * @dev: network device 2713 * @napi: NAPI context 2714 * @poll: polling function 2715 * 2716 * This variant of netif_napi_add() should be used from drivers using NAPI 2717 * to exclusively poll a TX queue. 2718 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2719 */ 2720 static inline void netif_napi_add_tx(struct net_device *dev, 2721 struct napi_struct *napi, 2722 int (*poll)(struct napi_struct *, int)) 2723 { 2724 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2725 } 2726 2727 /** 2728 * __netif_napi_del - remove a NAPI context 2729 * @napi: NAPI context 2730 * 2731 * Warning: caller must observe RCU grace period before freeing memory 2732 * containing @napi. Drivers might want to call this helper to combine 2733 * all the needed RCU grace periods into a single one. 2734 */ 2735 void __netif_napi_del(struct napi_struct *napi); 2736 2737 /** 2738 * netif_napi_del - remove a NAPI context 2739 * @napi: NAPI context 2740 * 2741 * netif_napi_del() removes a NAPI context from the network device NAPI list 2742 */ 2743 static inline void netif_napi_del(struct napi_struct *napi) 2744 { 2745 __netif_napi_del(napi); 2746 synchronize_net(); 2747 } 2748 2749 struct packet_type { 2750 __be16 type; /* This is really htons(ether_type). */ 2751 bool ignore_outgoing; 2752 struct net_device *dev; /* NULL is wildcarded here */ 2753 netdevice_tracker dev_tracker; 2754 int (*func) (struct sk_buff *, 2755 struct net_device *, 2756 struct packet_type *, 2757 struct net_device *); 2758 void (*list_func) (struct list_head *, 2759 struct packet_type *, 2760 struct net_device *); 2761 bool (*id_match)(struct packet_type *ptype, 2762 struct sock *sk); 2763 struct net *af_packet_net; 2764 void *af_packet_priv; 2765 struct list_head list; 2766 }; 2767 2768 struct offload_callbacks { 2769 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2770 netdev_features_t features); 2771 struct sk_buff *(*gro_receive)(struct list_head *head, 2772 struct sk_buff *skb); 2773 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2774 }; 2775 2776 struct packet_offload { 2777 __be16 type; /* This is really htons(ether_type). */ 2778 u16 priority; 2779 struct offload_callbacks callbacks; 2780 struct list_head list; 2781 }; 2782 2783 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2784 struct pcpu_sw_netstats { 2785 u64_stats_t rx_packets; 2786 u64_stats_t rx_bytes; 2787 u64_stats_t tx_packets; 2788 u64_stats_t tx_bytes; 2789 struct u64_stats_sync syncp; 2790 } __aligned(4 * sizeof(u64)); 2791 2792 struct pcpu_dstats { 2793 u64_stats_t rx_packets; 2794 u64_stats_t rx_bytes; 2795 u64_stats_t rx_drops; 2796 u64_stats_t tx_packets; 2797 u64_stats_t tx_bytes; 2798 u64_stats_t tx_drops; 2799 struct u64_stats_sync syncp; 2800 } __aligned(8 * sizeof(u64)); 2801 2802 struct pcpu_lstats { 2803 u64_stats_t packets; 2804 u64_stats_t bytes; 2805 struct u64_stats_sync syncp; 2806 } __aligned(2 * sizeof(u64)); 2807 2808 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2809 2810 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2811 { 2812 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2813 2814 u64_stats_update_begin(&tstats->syncp); 2815 u64_stats_add(&tstats->rx_bytes, len); 2816 u64_stats_inc(&tstats->rx_packets); 2817 u64_stats_update_end(&tstats->syncp); 2818 } 2819 2820 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2821 unsigned int packets, 2822 unsigned int len) 2823 { 2824 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2825 2826 u64_stats_update_begin(&tstats->syncp); 2827 u64_stats_add(&tstats->tx_bytes, len); 2828 u64_stats_add(&tstats->tx_packets, packets); 2829 u64_stats_update_end(&tstats->syncp); 2830 } 2831 2832 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2833 { 2834 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2835 2836 u64_stats_update_begin(&lstats->syncp); 2837 u64_stats_add(&lstats->bytes, len); 2838 u64_stats_inc(&lstats->packets); 2839 u64_stats_update_end(&lstats->syncp); 2840 } 2841 2842 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2843 ({ \ 2844 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2845 if (pcpu_stats) { \ 2846 int __cpu; \ 2847 for_each_possible_cpu(__cpu) { \ 2848 typeof(type) *stat; \ 2849 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2850 u64_stats_init(&stat->syncp); \ 2851 } \ 2852 } \ 2853 pcpu_stats; \ 2854 }) 2855 2856 #define netdev_alloc_pcpu_stats(type) \ 2857 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2858 2859 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2860 ({ \ 2861 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2862 if (pcpu_stats) { \ 2863 int __cpu; \ 2864 for_each_possible_cpu(__cpu) { \ 2865 typeof(type) *stat; \ 2866 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2867 u64_stats_init(&stat->syncp); \ 2868 } \ 2869 } \ 2870 pcpu_stats; \ 2871 }) 2872 2873 enum netdev_lag_tx_type { 2874 NETDEV_LAG_TX_TYPE_UNKNOWN, 2875 NETDEV_LAG_TX_TYPE_RANDOM, 2876 NETDEV_LAG_TX_TYPE_BROADCAST, 2877 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2878 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2879 NETDEV_LAG_TX_TYPE_HASH, 2880 }; 2881 2882 enum netdev_lag_hash { 2883 NETDEV_LAG_HASH_NONE, 2884 NETDEV_LAG_HASH_L2, 2885 NETDEV_LAG_HASH_L34, 2886 NETDEV_LAG_HASH_L23, 2887 NETDEV_LAG_HASH_E23, 2888 NETDEV_LAG_HASH_E34, 2889 NETDEV_LAG_HASH_VLAN_SRCMAC, 2890 NETDEV_LAG_HASH_UNKNOWN, 2891 }; 2892 2893 struct netdev_lag_upper_info { 2894 enum netdev_lag_tx_type tx_type; 2895 enum netdev_lag_hash hash_type; 2896 }; 2897 2898 struct netdev_lag_lower_state_info { 2899 u8 link_up : 1, 2900 tx_enabled : 1; 2901 }; 2902 2903 #include <linux/notifier.h> 2904 2905 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2906 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2907 * adding new types. 2908 */ 2909 enum netdev_cmd { 2910 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2911 NETDEV_DOWN, 2912 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2913 detected a hardware crash and restarted 2914 - we can use this eg to kick tcp sessions 2915 once done */ 2916 NETDEV_CHANGE, /* Notify device state change */ 2917 NETDEV_REGISTER, 2918 NETDEV_UNREGISTER, 2919 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2920 NETDEV_CHANGEADDR, /* notify after the address change */ 2921 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2922 NETDEV_GOING_DOWN, 2923 NETDEV_CHANGENAME, 2924 NETDEV_FEAT_CHANGE, 2925 NETDEV_BONDING_FAILOVER, 2926 NETDEV_PRE_UP, 2927 NETDEV_PRE_TYPE_CHANGE, 2928 NETDEV_POST_TYPE_CHANGE, 2929 NETDEV_POST_INIT, 2930 NETDEV_PRE_UNINIT, 2931 NETDEV_RELEASE, 2932 NETDEV_NOTIFY_PEERS, 2933 NETDEV_JOIN, 2934 NETDEV_CHANGEUPPER, 2935 NETDEV_RESEND_IGMP, 2936 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2937 NETDEV_CHANGEINFODATA, 2938 NETDEV_BONDING_INFO, 2939 NETDEV_PRECHANGEUPPER, 2940 NETDEV_CHANGELOWERSTATE, 2941 NETDEV_UDP_TUNNEL_PUSH_INFO, 2942 NETDEV_UDP_TUNNEL_DROP_INFO, 2943 NETDEV_CHANGE_TX_QUEUE_LEN, 2944 NETDEV_CVLAN_FILTER_PUSH_INFO, 2945 NETDEV_CVLAN_FILTER_DROP_INFO, 2946 NETDEV_SVLAN_FILTER_PUSH_INFO, 2947 NETDEV_SVLAN_FILTER_DROP_INFO, 2948 NETDEV_OFFLOAD_XSTATS_ENABLE, 2949 NETDEV_OFFLOAD_XSTATS_DISABLE, 2950 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 2951 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 2952 NETDEV_XDP_FEAT_CHANGE, 2953 }; 2954 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2955 2956 int register_netdevice_notifier(struct notifier_block *nb); 2957 int unregister_netdevice_notifier(struct notifier_block *nb); 2958 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2959 int unregister_netdevice_notifier_net(struct net *net, 2960 struct notifier_block *nb); 2961 int register_netdevice_notifier_dev_net(struct net_device *dev, 2962 struct notifier_block *nb, 2963 struct netdev_net_notifier *nn); 2964 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2965 struct notifier_block *nb, 2966 struct netdev_net_notifier *nn); 2967 2968 struct netdev_notifier_info { 2969 struct net_device *dev; 2970 struct netlink_ext_ack *extack; 2971 }; 2972 2973 struct netdev_notifier_info_ext { 2974 struct netdev_notifier_info info; /* must be first */ 2975 union { 2976 u32 mtu; 2977 } ext; 2978 }; 2979 2980 struct netdev_notifier_change_info { 2981 struct netdev_notifier_info info; /* must be first */ 2982 unsigned int flags_changed; 2983 }; 2984 2985 struct netdev_notifier_changeupper_info { 2986 struct netdev_notifier_info info; /* must be first */ 2987 struct net_device *upper_dev; /* new upper dev */ 2988 bool master; /* is upper dev master */ 2989 bool linking; /* is the notification for link or unlink */ 2990 void *upper_info; /* upper dev info */ 2991 }; 2992 2993 struct netdev_notifier_changelowerstate_info { 2994 struct netdev_notifier_info info; /* must be first */ 2995 void *lower_state_info; /* is lower dev state */ 2996 }; 2997 2998 struct netdev_notifier_pre_changeaddr_info { 2999 struct netdev_notifier_info info; /* must be first */ 3000 const unsigned char *dev_addr; 3001 }; 3002 3003 enum netdev_offload_xstats_type { 3004 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 3005 }; 3006 3007 struct netdev_notifier_offload_xstats_info { 3008 struct netdev_notifier_info info; /* must be first */ 3009 enum netdev_offload_xstats_type type; 3010 3011 union { 3012 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 3013 struct netdev_notifier_offload_xstats_rd *report_delta; 3014 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 3015 struct netdev_notifier_offload_xstats_ru *report_used; 3016 }; 3017 }; 3018 3019 int netdev_offload_xstats_enable(struct net_device *dev, 3020 enum netdev_offload_xstats_type type, 3021 struct netlink_ext_ack *extack); 3022 int netdev_offload_xstats_disable(struct net_device *dev, 3023 enum netdev_offload_xstats_type type); 3024 bool netdev_offload_xstats_enabled(const struct net_device *dev, 3025 enum netdev_offload_xstats_type type); 3026 int netdev_offload_xstats_get(struct net_device *dev, 3027 enum netdev_offload_xstats_type type, 3028 struct rtnl_hw_stats64 *stats, bool *used, 3029 struct netlink_ext_ack *extack); 3030 void 3031 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 3032 const struct rtnl_hw_stats64 *stats); 3033 void 3034 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 3035 void netdev_offload_xstats_push_delta(struct net_device *dev, 3036 enum netdev_offload_xstats_type type, 3037 const struct rtnl_hw_stats64 *stats); 3038 3039 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 3040 struct net_device *dev) 3041 { 3042 info->dev = dev; 3043 info->extack = NULL; 3044 } 3045 3046 static inline struct net_device * 3047 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 3048 { 3049 return info->dev; 3050 } 3051 3052 static inline struct netlink_ext_ack * 3053 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 3054 { 3055 return info->extack; 3056 } 3057 3058 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 3059 int call_netdevice_notifiers_info(unsigned long val, 3060 struct netdev_notifier_info *info); 3061 3062 #define for_each_netdev(net, d) \ 3063 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 3064 #define for_each_netdev_reverse(net, d) \ 3065 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 3066 #define for_each_netdev_rcu(net, d) \ 3067 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 3068 #define for_each_netdev_safe(net, d, n) \ 3069 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 3070 #define for_each_netdev_continue(net, d) \ 3071 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 3072 #define for_each_netdev_continue_reverse(net, d) \ 3073 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 3074 dev_list) 3075 #define for_each_netdev_continue_rcu(net, d) \ 3076 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 3077 #define for_each_netdev_in_bond_rcu(bond, slave) \ 3078 for_each_netdev_rcu(&init_net, slave) \ 3079 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 3080 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 3081 3082 #define for_each_netdev_dump(net, d, ifindex) \ 3083 for (; (d = xa_find(&(net)->dev_by_index, &ifindex, \ 3084 ULONG_MAX, XA_PRESENT)); ifindex++) 3085 3086 static inline struct net_device *next_net_device(struct net_device *dev) 3087 { 3088 struct list_head *lh; 3089 struct net *net; 3090 3091 net = dev_net(dev); 3092 lh = dev->dev_list.next; 3093 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3094 } 3095 3096 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 3097 { 3098 struct list_head *lh; 3099 struct net *net; 3100 3101 net = dev_net(dev); 3102 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 3103 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3104 } 3105 3106 static inline struct net_device *first_net_device(struct net *net) 3107 { 3108 return list_empty(&net->dev_base_head) ? NULL : 3109 net_device_entry(net->dev_base_head.next); 3110 } 3111 3112 static inline struct net_device *first_net_device_rcu(struct net *net) 3113 { 3114 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 3115 3116 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3117 } 3118 3119 int netdev_boot_setup_check(struct net_device *dev); 3120 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 3121 const char *hwaddr); 3122 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 3123 void dev_add_pack(struct packet_type *pt); 3124 void dev_remove_pack(struct packet_type *pt); 3125 void __dev_remove_pack(struct packet_type *pt); 3126 void dev_add_offload(struct packet_offload *po); 3127 void dev_remove_offload(struct packet_offload *po); 3128 3129 int dev_get_iflink(const struct net_device *dev); 3130 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 3131 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 3132 struct net_device_path_stack *stack); 3133 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 3134 unsigned short mask); 3135 struct net_device *dev_get_by_name(struct net *net, const char *name); 3136 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 3137 struct net_device *__dev_get_by_name(struct net *net, const char *name); 3138 bool netdev_name_in_use(struct net *net, const char *name); 3139 int dev_alloc_name(struct net_device *dev, const char *name); 3140 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 3141 void dev_close(struct net_device *dev); 3142 void dev_close_many(struct list_head *head, bool unlink); 3143 void dev_disable_lro(struct net_device *dev); 3144 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 3145 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 3146 struct net_device *sb_dev); 3147 3148 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3149 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3150 3151 static inline int dev_queue_xmit(struct sk_buff *skb) 3152 { 3153 return __dev_queue_xmit(skb, NULL); 3154 } 3155 3156 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3157 struct net_device *sb_dev) 3158 { 3159 return __dev_queue_xmit(skb, sb_dev); 3160 } 3161 3162 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3163 { 3164 int ret; 3165 3166 ret = __dev_direct_xmit(skb, queue_id); 3167 if (!dev_xmit_complete(ret)) 3168 kfree_skb(skb); 3169 return ret; 3170 } 3171 3172 int register_netdevice(struct net_device *dev); 3173 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3174 void unregister_netdevice_many(struct list_head *head); 3175 static inline void unregister_netdevice(struct net_device *dev) 3176 { 3177 unregister_netdevice_queue(dev, NULL); 3178 } 3179 3180 int netdev_refcnt_read(const struct net_device *dev); 3181 void free_netdev(struct net_device *dev); 3182 void init_dummy_netdev(struct net_device *dev); 3183 3184 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3185 struct sk_buff *skb, 3186 bool all_slaves); 3187 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3188 struct sock *sk); 3189 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3190 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3191 struct net_device *netdev_get_by_index(struct net *net, int ifindex, 3192 netdevice_tracker *tracker, gfp_t gfp); 3193 struct net_device *netdev_get_by_name(struct net *net, const char *name, 3194 netdevice_tracker *tracker, gfp_t gfp); 3195 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3196 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 3197 void netdev_copy_name(struct net_device *dev, char *name); 3198 3199 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3200 unsigned short type, 3201 const void *daddr, const void *saddr, 3202 unsigned int len) 3203 { 3204 if (!dev->header_ops || !dev->header_ops->create) 3205 return 0; 3206 3207 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3208 } 3209 3210 static inline int dev_parse_header(const struct sk_buff *skb, 3211 unsigned char *haddr) 3212 { 3213 const struct net_device *dev = skb->dev; 3214 3215 if (!dev->header_ops || !dev->header_ops->parse) 3216 return 0; 3217 return dev->header_ops->parse(skb, haddr); 3218 } 3219 3220 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3221 { 3222 const struct net_device *dev = skb->dev; 3223 3224 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3225 return 0; 3226 return dev->header_ops->parse_protocol(skb); 3227 } 3228 3229 /* ll_header must have at least hard_header_len allocated */ 3230 static inline bool dev_validate_header(const struct net_device *dev, 3231 char *ll_header, int len) 3232 { 3233 if (likely(len >= dev->hard_header_len)) 3234 return true; 3235 if (len < dev->min_header_len) 3236 return false; 3237 3238 if (capable(CAP_SYS_RAWIO)) { 3239 memset(ll_header + len, 0, dev->hard_header_len - len); 3240 return true; 3241 } 3242 3243 if (dev->header_ops && dev->header_ops->validate) 3244 return dev->header_ops->validate(ll_header, len); 3245 3246 return false; 3247 } 3248 3249 static inline bool dev_has_header(const struct net_device *dev) 3250 { 3251 return dev->header_ops && dev->header_ops->create; 3252 } 3253 3254 /* 3255 * Incoming packets are placed on per-CPU queues 3256 */ 3257 struct softnet_data { 3258 struct list_head poll_list; 3259 struct sk_buff_head process_queue; 3260 local_lock_t process_queue_bh_lock; 3261 3262 /* stats */ 3263 unsigned int processed; 3264 unsigned int time_squeeze; 3265 #ifdef CONFIG_RPS 3266 struct softnet_data *rps_ipi_list; 3267 #endif 3268 3269 unsigned int received_rps; 3270 bool in_net_rx_action; 3271 bool in_napi_threaded_poll; 3272 3273 #ifdef CONFIG_NET_FLOW_LIMIT 3274 struct sd_flow_limit __rcu *flow_limit; 3275 #endif 3276 struct Qdisc *output_queue; 3277 struct Qdisc **output_queue_tailp; 3278 struct sk_buff *completion_queue; 3279 #ifdef CONFIG_XFRM_OFFLOAD 3280 struct sk_buff_head xfrm_backlog; 3281 #endif 3282 /* written and read only by owning cpu: */ 3283 struct netdev_xmit xmit; 3284 #ifdef CONFIG_RPS 3285 /* input_queue_head should be written by cpu owning this struct, 3286 * and only read by other cpus. Worth using a cache line. 3287 */ 3288 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3289 3290 /* Elements below can be accessed between CPUs for RPS/RFS */ 3291 call_single_data_t csd ____cacheline_aligned_in_smp; 3292 struct softnet_data *rps_ipi_next; 3293 unsigned int cpu; 3294 unsigned int input_queue_tail; 3295 #endif 3296 struct sk_buff_head input_pkt_queue; 3297 struct napi_struct backlog; 3298 3299 atomic_t dropped ____cacheline_aligned_in_smp; 3300 3301 /* Another possibly contended cache line */ 3302 spinlock_t defer_lock ____cacheline_aligned_in_smp; 3303 int defer_count; 3304 int defer_ipi_scheduled; 3305 struct sk_buff *defer_list; 3306 call_single_data_t defer_csd; 3307 }; 3308 3309 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3310 3311 #ifndef CONFIG_PREEMPT_RT 3312 static inline int dev_recursion_level(void) 3313 { 3314 return this_cpu_read(softnet_data.xmit.recursion); 3315 } 3316 #else 3317 static inline int dev_recursion_level(void) 3318 { 3319 return current->net_xmit.recursion; 3320 } 3321 3322 #endif 3323 3324 void __netif_schedule(struct Qdisc *q); 3325 void netif_schedule_queue(struct netdev_queue *txq); 3326 3327 static inline void netif_tx_schedule_all(struct net_device *dev) 3328 { 3329 unsigned int i; 3330 3331 for (i = 0; i < dev->num_tx_queues; i++) 3332 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3333 } 3334 3335 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3336 { 3337 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3338 } 3339 3340 /** 3341 * netif_start_queue - allow transmit 3342 * @dev: network device 3343 * 3344 * Allow upper layers to call the device hard_start_xmit routine. 3345 */ 3346 static inline void netif_start_queue(struct net_device *dev) 3347 { 3348 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3349 } 3350 3351 static inline void netif_tx_start_all_queues(struct net_device *dev) 3352 { 3353 unsigned int i; 3354 3355 for (i = 0; i < dev->num_tx_queues; i++) { 3356 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3357 netif_tx_start_queue(txq); 3358 } 3359 } 3360 3361 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3362 3363 /** 3364 * netif_wake_queue - restart transmit 3365 * @dev: network device 3366 * 3367 * Allow upper layers to call the device hard_start_xmit routine. 3368 * Used for flow control when transmit resources are available. 3369 */ 3370 static inline void netif_wake_queue(struct net_device *dev) 3371 { 3372 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3373 } 3374 3375 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3376 { 3377 unsigned int i; 3378 3379 for (i = 0; i < dev->num_tx_queues; i++) { 3380 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3381 netif_tx_wake_queue(txq); 3382 } 3383 } 3384 3385 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3386 { 3387 /* Paired with READ_ONCE() from dev_watchdog() */ 3388 WRITE_ONCE(dev_queue->trans_start, jiffies); 3389 3390 /* This barrier is paired with smp_mb() from dev_watchdog() */ 3391 smp_mb__before_atomic(); 3392 3393 /* Must be an atomic op see netif_txq_try_stop() */ 3394 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3395 } 3396 3397 /** 3398 * netif_stop_queue - stop transmitted packets 3399 * @dev: network device 3400 * 3401 * Stop upper layers calling the device hard_start_xmit routine. 3402 * Used for flow control when transmit resources are unavailable. 3403 */ 3404 static inline void netif_stop_queue(struct net_device *dev) 3405 { 3406 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3407 } 3408 3409 void netif_tx_stop_all_queues(struct net_device *dev); 3410 3411 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3412 { 3413 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3414 } 3415 3416 /** 3417 * netif_queue_stopped - test if transmit queue is flowblocked 3418 * @dev: network device 3419 * 3420 * Test if transmit queue on device is currently unable to send. 3421 */ 3422 static inline bool netif_queue_stopped(const struct net_device *dev) 3423 { 3424 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3425 } 3426 3427 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3428 { 3429 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3430 } 3431 3432 static inline bool 3433 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3434 { 3435 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3436 } 3437 3438 static inline bool 3439 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3440 { 3441 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3442 } 3443 3444 /** 3445 * netdev_queue_set_dql_min_limit - set dql minimum limit 3446 * @dev_queue: pointer to transmit queue 3447 * @min_limit: dql minimum limit 3448 * 3449 * Forces xmit_more() to return true until the minimum threshold 3450 * defined by @min_limit is reached (or until the tx queue is 3451 * empty). Warning: to be use with care, misuse will impact the 3452 * latency. 3453 */ 3454 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3455 unsigned int min_limit) 3456 { 3457 #ifdef CONFIG_BQL 3458 dev_queue->dql.min_limit = min_limit; 3459 #endif 3460 } 3461 3462 static inline int netdev_queue_dql_avail(const struct netdev_queue *txq) 3463 { 3464 #ifdef CONFIG_BQL 3465 /* Non-BQL migrated drivers will return 0, too. */ 3466 return dql_avail(&txq->dql); 3467 #else 3468 return 0; 3469 #endif 3470 } 3471 3472 /** 3473 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3474 * @dev_queue: pointer to transmit queue 3475 * 3476 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3477 * to give appropriate hint to the CPU. 3478 */ 3479 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3480 { 3481 #ifdef CONFIG_BQL 3482 prefetchw(&dev_queue->dql.num_queued); 3483 #endif 3484 } 3485 3486 /** 3487 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3488 * @dev_queue: pointer to transmit queue 3489 * 3490 * BQL enabled drivers might use this helper in their TX completion path, 3491 * to give appropriate hint to the CPU. 3492 */ 3493 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3494 { 3495 #ifdef CONFIG_BQL 3496 prefetchw(&dev_queue->dql.limit); 3497 #endif 3498 } 3499 3500 /** 3501 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3502 * @dev_queue: network device queue 3503 * @bytes: number of bytes queued to the device queue 3504 * 3505 * Report the number of bytes queued for sending/completion to the network 3506 * device hardware queue. @bytes should be a good approximation and should 3507 * exactly match netdev_completed_queue() @bytes. 3508 * This is typically called once per packet, from ndo_start_xmit(). 3509 */ 3510 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3511 unsigned int bytes) 3512 { 3513 #ifdef CONFIG_BQL 3514 dql_queued(&dev_queue->dql, bytes); 3515 3516 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3517 return; 3518 3519 /* Paired with READ_ONCE() from dev_watchdog() */ 3520 WRITE_ONCE(dev_queue->trans_start, jiffies); 3521 3522 /* This barrier is paired with smp_mb() from dev_watchdog() */ 3523 smp_mb__before_atomic(); 3524 3525 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3526 3527 /* 3528 * The XOFF flag must be set before checking the dql_avail below, 3529 * because in netdev_tx_completed_queue we update the dql_completed 3530 * before checking the XOFF flag. 3531 */ 3532 smp_mb__after_atomic(); 3533 3534 /* check again in case another CPU has just made room avail */ 3535 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3536 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3537 #endif 3538 } 3539 3540 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3541 * that they should not test BQL status themselves. 3542 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3543 * skb of a batch. 3544 * Returns true if the doorbell must be used to kick the NIC. 3545 */ 3546 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3547 unsigned int bytes, 3548 bool xmit_more) 3549 { 3550 if (xmit_more) { 3551 #ifdef CONFIG_BQL 3552 dql_queued(&dev_queue->dql, bytes); 3553 #endif 3554 return netif_tx_queue_stopped(dev_queue); 3555 } 3556 netdev_tx_sent_queue(dev_queue, bytes); 3557 return true; 3558 } 3559 3560 /** 3561 * netdev_sent_queue - report the number of bytes queued to hardware 3562 * @dev: network device 3563 * @bytes: number of bytes queued to the hardware device queue 3564 * 3565 * Report the number of bytes queued for sending/completion to the network 3566 * device hardware queue#0. @bytes should be a good approximation and should 3567 * exactly match netdev_completed_queue() @bytes. 3568 * This is typically called once per packet, from ndo_start_xmit(). 3569 */ 3570 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3571 { 3572 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3573 } 3574 3575 static inline bool __netdev_sent_queue(struct net_device *dev, 3576 unsigned int bytes, 3577 bool xmit_more) 3578 { 3579 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3580 xmit_more); 3581 } 3582 3583 /** 3584 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3585 * @dev_queue: network device queue 3586 * @pkts: number of packets (currently ignored) 3587 * @bytes: number of bytes dequeued from the device queue 3588 * 3589 * Must be called at most once per TX completion round (and not per 3590 * individual packet), so that BQL can adjust its limits appropriately. 3591 */ 3592 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3593 unsigned int pkts, unsigned int bytes) 3594 { 3595 #ifdef CONFIG_BQL 3596 if (unlikely(!bytes)) 3597 return; 3598 3599 dql_completed(&dev_queue->dql, bytes); 3600 3601 /* 3602 * Without the memory barrier there is a small possibility that 3603 * netdev_tx_sent_queue will miss the update and cause the queue to 3604 * be stopped forever 3605 */ 3606 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */ 3607 3608 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3609 return; 3610 3611 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3612 netif_schedule_queue(dev_queue); 3613 #endif 3614 } 3615 3616 /** 3617 * netdev_completed_queue - report bytes and packets completed by device 3618 * @dev: network device 3619 * @pkts: actual number of packets sent over the medium 3620 * @bytes: actual number of bytes sent over the medium 3621 * 3622 * Report the number of bytes and packets transmitted by the network device 3623 * hardware queue over the physical medium, @bytes must exactly match the 3624 * @bytes amount passed to netdev_sent_queue() 3625 */ 3626 static inline void netdev_completed_queue(struct net_device *dev, 3627 unsigned int pkts, unsigned int bytes) 3628 { 3629 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3630 } 3631 3632 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3633 { 3634 #ifdef CONFIG_BQL 3635 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3636 dql_reset(&q->dql); 3637 #endif 3638 } 3639 3640 /** 3641 * netdev_tx_reset_subqueue - reset the BQL stats and state of a netdev queue 3642 * @dev: network device 3643 * @qid: stack index of the queue to reset 3644 */ 3645 static inline void netdev_tx_reset_subqueue(const struct net_device *dev, 3646 u32 qid) 3647 { 3648 netdev_tx_reset_queue(netdev_get_tx_queue(dev, qid)); 3649 } 3650 3651 /** 3652 * netdev_reset_queue - reset the packets and bytes count of a network device 3653 * @dev_queue: network device 3654 * 3655 * Reset the bytes and packet count of a network device and clear the 3656 * software flow control OFF bit for this network device 3657 */ 3658 static inline void netdev_reset_queue(struct net_device *dev_queue) 3659 { 3660 netdev_tx_reset_subqueue(dev_queue, 0); 3661 } 3662 3663 /** 3664 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3665 * @dev: network device 3666 * @queue_index: given tx queue index 3667 * 3668 * Returns 0 if given tx queue index >= number of device tx queues, 3669 * otherwise returns the originally passed tx queue index. 3670 */ 3671 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3672 { 3673 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3674 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3675 dev->name, queue_index, 3676 dev->real_num_tx_queues); 3677 return 0; 3678 } 3679 3680 return queue_index; 3681 } 3682 3683 /** 3684 * netif_running - test if up 3685 * @dev: network device 3686 * 3687 * Test if the device has been brought up. 3688 */ 3689 static inline bool netif_running(const struct net_device *dev) 3690 { 3691 return test_bit(__LINK_STATE_START, &dev->state); 3692 } 3693 3694 /* 3695 * Routines to manage the subqueues on a device. We only need start, 3696 * stop, and a check if it's stopped. All other device management is 3697 * done at the overall netdevice level. 3698 * Also test the device if we're multiqueue. 3699 */ 3700 3701 /** 3702 * netif_start_subqueue - allow sending packets on subqueue 3703 * @dev: network device 3704 * @queue_index: sub queue index 3705 * 3706 * Start individual transmit queue of a device with multiple transmit queues. 3707 */ 3708 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3709 { 3710 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3711 3712 netif_tx_start_queue(txq); 3713 } 3714 3715 /** 3716 * netif_stop_subqueue - stop sending packets on subqueue 3717 * @dev: network device 3718 * @queue_index: sub queue index 3719 * 3720 * Stop individual transmit queue of a device with multiple transmit queues. 3721 */ 3722 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3723 { 3724 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3725 netif_tx_stop_queue(txq); 3726 } 3727 3728 /** 3729 * __netif_subqueue_stopped - test status of subqueue 3730 * @dev: network device 3731 * @queue_index: sub queue index 3732 * 3733 * Check individual transmit queue of a device with multiple transmit queues. 3734 */ 3735 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3736 u16 queue_index) 3737 { 3738 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3739 3740 return netif_tx_queue_stopped(txq); 3741 } 3742 3743 /** 3744 * netif_subqueue_stopped - test status of subqueue 3745 * @dev: network device 3746 * @skb: sub queue buffer pointer 3747 * 3748 * Check individual transmit queue of a device with multiple transmit queues. 3749 */ 3750 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3751 struct sk_buff *skb) 3752 { 3753 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3754 } 3755 3756 /** 3757 * netif_wake_subqueue - allow sending packets on subqueue 3758 * @dev: network device 3759 * @queue_index: sub queue index 3760 * 3761 * Resume individual transmit queue of a device with multiple transmit queues. 3762 */ 3763 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3764 { 3765 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3766 3767 netif_tx_wake_queue(txq); 3768 } 3769 3770 #ifdef CONFIG_XPS 3771 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3772 u16 index); 3773 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3774 u16 index, enum xps_map_type type); 3775 3776 /** 3777 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3778 * @j: CPU/Rx queue index 3779 * @mask: bitmask of all cpus/rx queues 3780 * @nr_bits: number of bits in the bitmask 3781 * 3782 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3783 */ 3784 static inline bool netif_attr_test_mask(unsigned long j, 3785 const unsigned long *mask, 3786 unsigned int nr_bits) 3787 { 3788 cpu_max_bits_warn(j, nr_bits); 3789 return test_bit(j, mask); 3790 } 3791 3792 /** 3793 * netif_attr_test_online - Test for online CPU/Rx queue 3794 * @j: CPU/Rx queue index 3795 * @online_mask: bitmask for CPUs/Rx queues that are online 3796 * @nr_bits: number of bits in the bitmask 3797 * 3798 * Returns true if a CPU/Rx queue is online. 3799 */ 3800 static inline bool netif_attr_test_online(unsigned long j, 3801 const unsigned long *online_mask, 3802 unsigned int nr_bits) 3803 { 3804 cpu_max_bits_warn(j, nr_bits); 3805 3806 if (online_mask) 3807 return test_bit(j, online_mask); 3808 3809 return (j < nr_bits); 3810 } 3811 3812 /** 3813 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3814 * @n: CPU/Rx queue index 3815 * @srcp: the cpumask/Rx queue mask pointer 3816 * @nr_bits: number of bits in the bitmask 3817 * 3818 * Returns >= nr_bits if no further CPUs/Rx queues set. 3819 */ 3820 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3821 unsigned int nr_bits) 3822 { 3823 /* -1 is a legal arg here. */ 3824 if (n != -1) 3825 cpu_max_bits_warn(n, nr_bits); 3826 3827 if (srcp) 3828 return find_next_bit(srcp, nr_bits, n + 1); 3829 3830 return n + 1; 3831 } 3832 3833 /** 3834 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3835 * @n: CPU/Rx queue index 3836 * @src1p: the first CPUs/Rx queues mask pointer 3837 * @src2p: the second CPUs/Rx queues mask pointer 3838 * @nr_bits: number of bits in the bitmask 3839 * 3840 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3841 */ 3842 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3843 const unsigned long *src2p, 3844 unsigned int nr_bits) 3845 { 3846 /* -1 is a legal arg here. */ 3847 if (n != -1) 3848 cpu_max_bits_warn(n, nr_bits); 3849 3850 if (src1p && src2p) 3851 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3852 else if (src1p) 3853 return find_next_bit(src1p, nr_bits, n + 1); 3854 else if (src2p) 3855 return find_next_bit(src2p, nr_bits, n + 1); 3856 3857 return n + 1; 3858 } 3859 #else 3860 static inline int netif_set_xps_queue(struct net_device *dev, 3861 const struct cpumask *mask, 3862 u16 index) 3863 { 3864 return 0; 3865 } 3866 3867 static inline int __netif_set_xps_queue(struct net_device *dev, 3868 const unsigned long *mask, 3869 u16 index, enum xps_map_type type) 3870 { 3871 return 0; 3872 } 3873 #endif 3874 3875 /** 3876 * netif_is_multiqueue - test if device has multiple transmit queues 3877 * @dev: network device 3878 * 3879 * Check if device has multiple transmit queues 3880 */ 3881 static inline bool netif_is_multiqueue(const struct net_device *dev) 3882 { 3883 return dev->num_tx_queues > 1; 3884 } 3885 3886 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3887 3888 #ifdef CONFIG_SYSFS 3889 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3890 #else 3891 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3892 unsigned int rxqs) 3893 { 3894 dev->real_num_rx_queues = rxqs; 3895 return 0; 3896 } 3897 #endif 3898 int netif_set_real_num_queues(struct net_device *dev, 3899 unsigned int txq, unsigned int rxq); 3900 3901 int netif_get_num_default_rss_queues(void); 3902 3903 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason); 3904 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason); 3905 3906 /* 3907 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3908 * interrupt context or with hardware interrupts being disabled. 3909 * (in_hardirq() || irqs_disabled()) 3910 * 3911 * We provide four helpers that can be used in following contexts : 3912 * 3913 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3914 * replacing kfree_skb(skb) 3915 * 3916 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3917 * Typically used in place of consume_skb(skb) in TX completion path 3918 * 3919 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3920 * replacing kfree_skb(skb) 3921 * 3922 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3923 * and consumed a packet. Used in place of consume_skb(skb) 3924 */ 3925 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3926 { 3927 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 3928 } 3929 3930 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3931 { 3932 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED); 3933 } 3934 3935 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3936 { 3937 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 3938 } 3939 3940 static inline void dev_consume_skb_any(struct sk_buff *skb) 3941 { 3942 dev_kfree_skb_any_reason(skb, SKB_CONSUMED); 3943 } 3944 3945 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3946 struct bpf_prog *xdp_prog); 3947 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3948 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff **pskb); 3949 int netif_rx(struct sk_buff *skb); 3950 int __netif_rx(struct sk_buff *skb); 3951 3952 int netif_receive_skb(struct sk_buff *skb); 3953 int netif_receive_skb_core(struct sk_buff *skb); 3954 void netif_receive_skb_list_internal(struct list_head *head); 3955 void netif_receive_skb_list(struct list_head *head); 3956 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3957 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3958 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3959 void napi_get_frags_check(struct napi_struct *napi); 3960 gro_result_t napi_gro_frags(struct napi_struct *napi); 3961 3962 static inline void napi_free_frags(struct napi_struct *napi) 3963 { 3964 kfree_skb(napi->skb); 3965 napi->skb = NULL; 3966 } 3967 3968 bool netdev_is_rx_handler_busy(struct net_device *dev); 3969 int netdev_rx_handler_register(struct net_device *dev, 3970 rx_handler_func_t *rx_handler, 3971 void *rx_handler_data); 3972 void netdev_rx_handler_unregister(struct net_device *dev); 3973 3974 bool dev_valid_name(const char *name); 3975 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 3976 { 3977 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 3978 } 3979 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 3980 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 3981 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3982 void __user *data, bool *need_copyout); 3983 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 3984 int generic_hwtstamp_get_lower(struct net_device *dev, 3985 struct kernel_hwtstamp_config *kernel_cfg); 3986 int generic_hwtstamp_set_lower(struct net_device *dev, 3987 struct kernel_hwtstamp_config *kernel_cfg, 3988 struct netlink_ext_ack *extack); 3989 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 3990 unsigned int dev_get_flags(const struct net_device *); 3991 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3992 struct netlink_ext_ack *extack); 3993 int dev_change_flags(struct net_device *dev, unsigned int flags, 3994 struct netlink_ext_ack *extack); 3995 int dev_set_alias(struct net_device *, const char *, size_t); 3996 int dev_get_alias(const struct net_device *, char *, size_t); 3997 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 3998 const char *pat, int new_ifindex); 3999 static inline 4000 int dev_change_net_namespace(struct net_device *dev, struct net *net, 4001 const char *pat) 4002 { 4003 return __dev_change_net_namespace(dev, net, pat, 0); 4004 } 4005 int __dev_set_mtu(struct net_device *, int); 4006 int dev_set_mtu(struct net_device *, int); 4007 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 4008 struct netlink_ext_ack *extack); 4009 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 4010 struct netlink_ext_ack *extack); 4011 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 4012 struct netlink_ext_ack *extack); 4013 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 4014 int dev_get_port_parent_id(struct net_device *dev, 4015 struct netdev_phys_item_id *ppid, bool recurse); 4016 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 4017 4018 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 4019 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 4020 struct netdev_queue *txq, int *ret); 4021 4022 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 4023 u8 dev_xdp_prog_count(struct net_device *dev); 4024 int dev_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf); 4025 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 4026 4027 u32 dev_get_min_mp_channel_count(const struct net_device *dev); 4028 4029 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4030 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4031 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 4032 bool is_skb_forwardable(const struct net_device *dev, 4033 const struct sk_buff *skb); 4034 4035 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 4036 const struct sk_buff *skb, 4037 const bool check_mtu) 4038 { 4039 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 4040 unsigned int len; 4041 4042 if (!(dev->flags & IFF_UP)) 4043 return false; 4044 4045 if (!check_mtu) 4046 return true; 4047 4048 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 4049 if (skb->len <= len) 4050 return true; 4051 4052 /* if TSO is enabled, we don't care about the length as the packet 4053 * could be forwarded without being segmented before 4054 */ 4055 if (skb_is_gso(skb)) 4056 return true; 4057 4058 return false; 4059 } 4060 4061 void netdev_core_stats_inc(struct net_device *dev, u32 offset); 4062 4063 #define DEV_CORE_STATS_INC(FIELD) \ 4064 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 4065 { \ 4066 netdev_core_stats_inc(dev, \ 4067 offsetof(struct net_device_core_stats, FIELD)); \ 4068 } 4069 DEV_CORE_STATS_INC(rx_dropped) 4070 DEV_CORE_STATS_INC(tx_dropped) 4071 DEV_CORE_STATS_INC(rx_nohandler) 4072 DEV_CORE_STATS_INC(rx_otherhost_dropped) 4073 #undef DEV_CORE_STATS_INC 4074 4075 static __always_inline int ____dev_forward_skb(struct net_device *dev, 4076 struct sk_buff *skb, 4077 const bool check_mtu) 4078 { 4079 if (skb_orphan_frags(skb, GFP_ATOMIC) || 4080 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 4081 dev_core_stats_rx_dropped_inc(dev); 4082 kfree_skb(skb); 4083 return NET_RX_DROP; 4084 } 4085 4086 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 4087 skb->priority = 0; 4088 return 0; 4089 } 4090 4091 bool dev_nit_active(struct net_device *dev); 4092 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 4093 4094 static inline void __dev_put(struct net_device *dev) 4095 { 4096 if (dev) { 4097 #ifdef CONFIG_PCPU_DEV_REFCNT 4098 this_cpu_dec(*dev->pcpu_refcnt); 4099 #else 4100 refcount_dec(&dev->dev_refcnt); 4101 #endif 4102 } 4103 } 4104 4105 static inline void __dev_hold(struct net_device *dev) 4106 { 4107 if (dev) { 4108 #ifdef CONFIG_PCPU_DEV_REFCNT 4109 this_cpu_inc(*dev->pcpu_refcnt); 4110 #else 4111 refcount_inc(&dev->dev_refcnt); 4112 #endif 4113 } 4114 } 4115 4116 static inline void __netdev_tracker_alloc(struct net_device *dev, 4117 netdevice_tracker *tracker, 4118 gfp_t gfp) 4119 { 4120 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4121 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 4122 #endif 4123 } 4124 4125 /* netdev_tracker_alloc() can upgrade a prior untracked reference 4126 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4127 */ 4128 static inline void netdev_tracker_alloc(struct net_device *dev, 4129 netdevice_tracker *tracker, gfp_t gfp) 4130 { 4131 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4132 refcount_dec(&dev->refcnt_tracker.no_tracker); 4133 __netdev_tracker_alloc(dev, tracker, gfp); 4134 #endif 4135 } 4136 4137 static inline void netdev_tracker_free(struct net_device *dev, 4138 netdevice_tracker *tracker) 4139 { 4140 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4141 ref_tracker_free(&dev->refcnt_tracker, tracker); 4142 #endif 4143 } 4144 4145 static inline void netdev_hold(struct net_device *dev, 4146 netdevice_tracker *tracker, gfp_t gfp) 4147 { 4148 if (dev) { 4149 __dev_hold(dev); 4150 __netdev_tracker_alloc(dev, tracker, gfp); 4151 } 4152 } 4153 4154 static inline void netdev_put(struct net_device *dev, 4155 netdevice_tracker *tracker) 4156 { 4157 if (dev) { 4158 netdev_tracker_free(dev, tracker); 4159 __dev_put(dev); 4160 } 4161 } 4162 4163 /** 4164 * dev_hold - get reference to device 4165 * @dev: network device 4166 * 4167 * Hold reference to device to keep it from being freed. 4168 * Try using netdev_hold() instead. 4169 */ 4170 static inline void dev_hold(struct net_device *dev) 4171 { 4172 netdev_hold(dev, NULL, GFP_ATOMIC); 4173 } 4174 4175 /** 4176 * dev_put - release reference to device 4177 * @dev: network device 4178 * 4179 * Release reference to device to allow it to be freed. 4180 * Try using netdev_put() instead. 4181 */ 4182 static inline void dev_put(struct net_device *dev) 4183 { 4184 netdev_put(dev, NULL); 4185 } 4186 4187 DEFINE_FREE(dev_put, struct net_device *, if (_T) dev_put(_T)) 4188 4189 static inline void netdev_ref_replace(struct net_device *odev, 4190 struct net_device *ndev, 4191 netdevice_tracker *tracker, 4192 gfp_t gfp) 4193 { 4194 if (odev) 4195 netdev_tracker_free(odev, tracker); 4196 4197 __dev_hold(ndev); 4198 __dev_put(odev); 4199 4200 if (ndev) 4201 __netdev_tracker_alloc(ndev, tracker, gfp); 4202 } 4203 4204 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4205 * and _off may be called from IRQ context, but it is caller 4206 * who is responsible for serialization of these calls. 4207 * 4208 * The name carrier is inappropriate, these functions should really be 4209 * called netif_lowerlayer_*() because they represent the state of any 4210 * kind of lower layer not just hardware media. 4211 */ 4212 void linkwatch_fire_event(struct net_device *dev); 4213 4214 /** 4215 * linkwatch_sync_dev - sync linkwatch for the given device 4216 * @dev: network device to sync linkwatch for 4217 * 4218 * Sync linkwatch for the given device, removing it from the 4219 * pending work list (if queued). 4220 */ 4221 void linkwatch_sync_dev(struct net_device *dev); 4222 4223 /** 4224 * netif_carrier_ok - test if carrier present 4225 * @dev: network device 4226 * 4227 * Check if carrier is present on device 4228 */ 4229 static inline bool netif_carrier_ok(const struct net_device *dev) 4230 { 4231 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4232 } 4233 4234 unsigned long dev_trans_start(struct net_device *dev); 4235 4236 void __netdev_watchdog_up(struct net_device *dev); 4237 4238 void netif_carrier_on(struct net_device *dev); 4239 void netif_carrier_off(struct net_device *dev); 4240 void netif_carrier_event(struct net_device *dev); 4241 4242 /** 4243 * netif_dormant_on - mark device as dormant. 4244 * @dev: network device 4245 * 4246 * Mark device as dormant (as per RFC2863). 4247 * 4248 * The dormant state indicates that the relevant interface is not 4249 * actually in a condition to pass packets (i.e., it is not 'up') but is 4250 * in a "pending" state, waiting for some external event. For "on- 4251 * demand" interfaces, this new state identifies the situation where the 4252 * interface is waiting for events to place it in the up state. 4253 */ 4254 static inline void netif_dormant_on(struct net_device *dev) 4255 { 4256 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4257 linkwatch_fire_event(dev); 4258 } 4259 4260 /** 4261 * netif_dormant_off - set device as not dormant. 4262 * @dev: network device 4263 * 4264 * Device is not in dormant state. 4265 */ 4266 static inline void netif_dormant_off(struct net_device *dev) 4267 { 4268 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4269 linkwatch_fire_event(dev); 4270 } 4271 4272 /** 4273 * netif_dormant - test if device is dormant 4274 * @dev: network device 4275 * 4276 * Check if device is dormant. 4277 */ 4278 static inline bool netif_dormant(const struct net_device *dev) 4279 { 4280 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4281 } 4282 4283 4284 /** 4285 * netif_testing_on - mark device as under test. 4286 * @dev: network device 4287 * 4288 * Mark device as under test (as per RFC2863). 4289 * 4290 * The testing state indicates that some test(s) must be performed on 4291 * the interface. After completion, of the test, the interface state 4292 * will change to up, dormant, or down, as appropriate. 4293 */ 4294 static inline void netif_testing_on(struct net_device *dev) 4295 { 4296 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4297 linkwatch_fire_event(dev); 4298 } 4299 4300 /** 4301 * netif_testing_off - set device as not under test. 4302 * @dev: network device 4303 * 4304 * Device is not in testing state. 4305 */ 4306 static inline void netif_testing_off(struct net_device *dev) 4307 { 4308 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4309 linkwatch_fire_event(dev); 4310 } 4311 4312 /** 4313 * netif_testing - test if device is under test 4314 * @dev: network device 4315 * 4316 * Check if device is under test 4317 */ 4318 static inline bool netif_testing(const struct net_device *dev) 4319 { 4320 return test_bit(__LINK_STATE_TESTING, &dev->state); 4321 } 4322 4323 4324 /** 4325 * netif_oper_up - test if device is operational 4326 * @dev: network device 4327 * 4328 * Check if carrier is operational 4329 */ 4330 static inline bool netif_oper_up(const struct net_device *dev) 4331 { 4332 unsigned int operstate = READ_ONCE(dev->operstate); 4333 4334 return operstate == IF_OPER_UP || 4335 operstate == IF_OPER_UNKNOWN /* backward compat */; 4336 } 4337 4338 /** 4339 * netif_device_present - is device available or removed 4340 * @dev: network device 4341 * 4342 * Check if device has not been removed from system. 4343 */ 4344 static inline bool netif_device_present(const struct net_device *dev) 4345 { 4346 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4347 } 4348 4349 void netif_device_detach(struct net_device *dev); 4350 4351 void netif_device_attach(struct net_device *dev); 4352 4353 /* 4354 * Network interface message level settings 4355 */ 4356 4357 enum { 4358 NETIF_MSG_DRV_BIT, 4359 NETIF_MSG_PROBE_BIT, 4360 NETIF_MSG_LINK_BIT, 4361 NETIF_MSG_TIMER_BIT, 4362 NETIF_MSG_IFDOWN_BIT, 4363 NETIF_MSG_IFUP_BIT, 4364 NETIF_MSG_RX_ERR_BIT, 4365 NETIF_MSG_TX_ERR_BIT, 4366 NETIF_MSG_TX_QUEUED_BIT, 4367 NETIF_MSG_INTR_BIT, 4368 NETIF_MSG_TX_DONE_BIT, 4369 NETIF_MSG_RX_STATUS_BIT, 4370 NETIF_MSG_PKTDATA_BIT, 4371 NETIF_MSG_HW_BIT, 4372 NETIF_MSG_WOL_BIT, 4373 4374 /* When you add a new bit above, update netif_msg_class_names array 4375 * in net/ethtool/common.c 4376 */ 4377 NETIF_MSG_CLASS_COUNT, 4378 }; 4379 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4380 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4381 4382 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4383 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4384 4385 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4386 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4387 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4388 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4389 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4390 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4391 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4392 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4393 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4394 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4395 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4396 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4397 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4398 #define NETIF_MSG_HW __NETIF_MSG(HW) 4399 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4400 4401 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4402 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4403 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4404 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4405 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4406 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4407 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4408 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4409 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4410 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4411 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4412 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4413 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4414 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4415 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4416 4417 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4418 { 4419 /* use default */ 4420 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4421 return default_msg_enable_bits; 4422 if (debug_value == 0) /* no output */ 4423 return 0; 4424 /* set low N bits */ 4425 return (1U << debug_value) - 1; 4426 } 4427 4428 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4429 { 4430 spin_lock(&txq->_xmit_lock); 4431 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4432 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4433 } 4434 4435 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4436 { 4437 __acquire(&txq->_xmit_lock); 4438 return true; 4439 } 4440 4441 static inline void __netif_tx_release(struct netdev_queue *txq) 4442 { 4443 __release(&txq->_xmit_lock); 4444 } 4445 4446 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4447 { 4448 spin_lock_bh(&txq->_xmit_lock); 4449 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4450 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4451 } 4452 4453 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4454 { 4455 bool ok = spin_trylock(&txq->_xmit_lock); 4456 4457 if (likely(ok)) { 4458 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4459 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4460 } 4461 return ok; 4462 } 4463 4464 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4465 { 4466 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4467 WRITE_ONCE(txq->xmit_lock_owner, -1); 4468 spin_unlock(&txq->_xmit_lock); 4469 } 4470 4471 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4472 { 4473 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4474 WRITE_ONCE(txq->xmit_lock_owner, -1); 4475 spin_unlock_bh(&txq->_xmit_lock); 4476 } 4477 4478 /* 4479 * txq->trans_start can be read locklessly from dev_watchdog() 4480 */ 4481 static inline void txq_trans_update(struct netdev_queue *txq) 4482 { 4483 if (txq->xmit_lock_owner != -1) 4484 WRITE_ONCE(txq->trans_start, jiffies); 4485 } 4486 4487 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4488 { 4489 unsigned long now = jiffies; 4490 4491 if (READ_ONCE(txq->trans_start) != now) 4492 WRITE_ONCE(txq->trans_start, now); 4493 } 4494 4495 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4496 static inline void netif_trans_update(struct net_device *dev) 4497 { 4498 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4499 4500 txq_trans_cond_update(txq); 4501 } 4502 4503 /** 4504 * netif_tx_lock - grab network device transmit lock 4505 * @dev: network device 4506 * 4507 * Get network device transmit lock 4508 */ 4509 void netif_tx_lock(struct net_device *dev); 4510 4511 static inline void netif_tx_lock_bh(struct net_device *dev) 4512 { 4513 local_bh_disable(); 4514 netif_tx_lock(dev); 4515 } 4516 4517 void netif_tx_unlock(struct net_device *dev); 4518 4519 static inline void netif_tx_unlock_bh(struct net_device *dev) 4520 { 4521 netif_tx_unlock(dev); 4522 local_bh_enable(); 4523 } 4524 4525 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4526 if (!(dev)->lltx) { \ 4527 __netif_tx_lock(txq, cpu); \ 4528 } else { \ 4529 __netif_tx_acquire(txq); \ 4530 } \ 4531 } 4532 4533 #define HARD_TX_TRYLOCK(dev, txq) \ 4534 (!(dev)->lltx ? \ 4535 __netif_tx_trylock(txq) : \ 4536 __netif_tx_acquire(txq)) 4537 4538 #define HARD_TX_UNLOCK(dev, txq) { \ 4539 if (!(dev)->lltx) { \ 4540 __netif_tx_unlock(txq); \ 4541 } else { \ 4542 __netif_tx_release(txq); \ 4543 } \ 4544 } 4545 4546 static inline void netif_tx_disable(struct net_device *dev) 4547 { 4548 unsigned int i; 4549 int cpu; 4550 4551 local_bh_disable(); 4552 cpu = smp_processor_id(); 4553 spin_lock(&dev->tx_global_lock); 4554 for (i = 0; i < dev->num_tx_queues; i++) { 4555 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4556 4557 __netif_tx_lock(txq, cpu); 4558 netif_tx_stop_queue(txq); 4559 __netif_tx_unlock(txq); 4560 } 4561 spin_unlock(&dev->tx_global_lock); 4562 local_bh_enable(); 4563 } 4564 4565 static inline void netif_addr_lock(struct net_device *dev) 4566 { 4567 unsigned char nest_level = 0; 4568 4569 #ifdef CONFIG_LOCKDEP 4570 nest_level = dev->nested_level; 4571 #endif 4572 spin_lock_nested(&dev->addr_list_lock, nest_level); 4573 } 4574 4575 static inline void netif_addr_lock_bh(struct net_device *dev) 4576 { 4577 unsigned char nest_level = 0; 4578 4579 #ifdef CONFIG_LOCKDEP 4580 nest_level = dev->nested_level; 4581 #endif 4582 local_bh_disable(); 4583 spin_lock_nested(&dev->addr_list_lock, nest_level); 4584 } 4585 4586 static inline void netif_addr_unlock(struct net_device *dev) 4587 { 4588 spin_unlock(&dev->addr_list_lock); 4589 } 4590 4591 static inline void netif_addr_unlock_bh(struct net_device *dev) 4592 { 4593 spin_unlock_bh(&dev->addr_list_lock); 4594 } 4595 4596 /* 4597 * dev_addrs walker. Should be used only for read access. Call with 4598 * rcu_read_lock held. 4599 */ 4600 #define for_each_dev_addr(dev, ha) \ 4601 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4602 4603 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4604 4605 void ether_setup(struct net_device *dev); 4606 4607 /* Allocate dummy net_device */ 4608 struct net_device *alloc_netdev_dummy(int sizeof_priv); 4609 4610 /* Support for loadable net-drivers */ 4611 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4612 unsigned char name_assign_type, 4613 void (*setup)(struct net_device *), 4614 unsigned int txqs, unsigned int rxqs); 4615 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4616 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4617 4618 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4619 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4620 count) 4621 4622 int register_netdev(struct net_device *dev); 4623 void unregister_netdev(struct net_device *dev); 4624 4625 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4626 4627 /* General hardware address lists handling functions */ 4628 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4629 struct netdev_hw_addr_list *from_list, int addr_len); 4630 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4631 struct netdev_hw_addr_list *from_list, int addr_len); 4632 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4633 struct net_device *dev, 4634 int (*sync)(struct net_device *, const unsigned char *), 4635 int (*unsync)(struct net_device *, 4636 const unsigned char *)); 4637 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4638 struct net_device *dev, 4639 int (*sync)(struct net_device *, 4640 const unsigned char *, int), 4641 int (*unsync)(struct net_device *, 4642 const unsigned char *, int)); 4643 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4644 struct net_device *dev, 4645 int (*unsync)(struct net_device *, 4646 const unsigned char *, int)); 4647 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4648 struct net_device *dev, 4649 int (*unsync)(struct net_device *, 4650 const unsigned char *)); 4651 void __hw_addr_init(struct netdev_hw_addr_list *list); 4652 4653 /* Functions used for device addresses handling */ 4654 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4655 const void *addr, size_t len); 4656 4657 static inline void 4658 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4659 { 4660 dev_addr_mod(dev, 0, addr, len); 4661 } 4662 4663 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4664 { 4665 __dev_addr_set(dev, addr, dev->addr_len); 4666 } 4667 4668 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4669 unsigned char addr_type); 4670 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4671 unsigned char addr_type); 4672 4673 /* Functions used for unicast addresses handling */ 4674 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4675 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4676 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4677 int dev_uc_sync(struct net_device *to, struct net_device *from); 4678 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4679 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4680 void dev_uc_flush(struct net_device *dev); 4681 void dev_uc_init(struct net_device *dev); 4682 4683 /** 4684 * __dev_uc_sync - Synchronize device's unicast list 4685 * @dev: device to sync 4686 * @sync: function to call if address should be added 4687 * @unsync: function to call if address should be removed 4688 * 4689 * Add newly added addresses to the interface, and release 4690 * addresses that have been deleted. 4691 */ 4692 static inline int __dev_uc_sync(struct net_device *dev, 4693 int (*sync)(struct net_device *, 4694 const unsigned char *), 4695 int (*unsync)(struct net_device *, 4696 const unsigned char *)) 4697 { 4698 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4699 } 4700 4701 /** 4702 * __dev_uc_unsync - Remove synchronized addresses from device 4703 * @dev: device to sync 4704 * @unsync: function to call if address should be removed 4705 * 4706 * Remove all addresses that were added to the device by dev_uc_sync(). 4707 */ 4708 static inline void __dev_uc_unsync(struct net_device *dev, 4709 int (*unsync)(struct net_device *, 4710 const unsigned char *)) 4711 { 4712 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4713 } 4714 4715 /* Functions used for multicast addresses handling */ 4716 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4717 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4718 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4719 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4720 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4721 int dev_mc_sync(struct net_device *to, struct net_device *from); 4722 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4723 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4724 void dev_mc_flush(struct net_device *dev); 4725 void dev_mc_init(struct net_device *dev); 4726 4727 /** 4728 * __dev_mc_sync - Synchronize device's multicast list 4729 * @dev: device to sync 4730 * @sync: function to call if address should be added 4731 * @unsync: function to call if address should be removed 4732 * 4733 * Add newly added addresses to the interface, and release 4734 * addresses that have been deleted. 4735 */ 4736 static inline int __dev_mc_sync(struct net_device *dev, 4737 int (*sync)(struct net_device *, 4738 const unsigned char *), 4739 int (*unsync)(struct net_device *, 4740 const unsigned char *)) 4741 { 4742 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4743 } 4744 4745 /** 4746 * __dev_mc_unsync - Remove synchronized addresses from device 4747 * @dev: device to sync 4748 * @unsync: function to call if address should be removed 4749 * 4750 * Remove all addresses that were added to the device by dev_mc_sync(). 4751 */ 4752 static inline void __dev_mc_unsync(struct net_device *dev, 4753 int (*unsync)(struct net_device *, 4754 const unsigned char *)) 4755 { 4756 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4757 } 4758 4759 /* Functions used for secondary unicast and multicast support */ 4760 void dev_set_rx_mode(struct net_device *dev); 4761 int dev_set_promiscuity(struct net_device *dev, int inc); 4762 int dev_set_allmulti(struct net_device *dev, int inc); 4763 void netdev_state_change(struct net_device *dev); 4764 void __netdev_notify_peers(struct net_device *dev); 4765 void netdev_notify_peers(struct net_device *dev); 4766 void netdev_features_change(struct net_device *dev); 4767 /* Load a device via the kmod */ 4768 void dev_load(struct net *net, const char *name); 4769 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4770 struct rtnl_link_stats64 *storage); 4771 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4772 const struct net_device_stats *netdev_stats); 4773 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4774 const struct pcpu_sw_netstats __percpu *netstats); 4775 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4776 4777 enum { 4778 NESTED_SYNC_IMM_BIT, 4779 NESTED_SYNC_TODO_BIT, 4780 }; 4781 4782 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4783 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4784 4785 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4786 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4787 4788 struct netdev_nested_priv { 4789 unsigned char flags; 4790 void *data; 4791 }; 4792 4793 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4794 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4795 struct list_head **iter); 4796 4797 /* iterate through upper list, must be called under RCU read lock */ 4798 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4799 for (iter = &(dev)->adj_list.upper, \ 4800 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4801 updev; \ 4802 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4803 4804 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4805 int (*fn)(struct net_device *upper_dev, 4806 struct netdev_nested_priv *priv), 4807 struct netdev_nested_priv *priv); 4808 4809 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4810 struct net_device *upper_dev); 4811 4812 bool netdev_has_any_upper_dev(struct net_device *dev); 4813 4814 void *netdev_lower_get_next_private(struct net_device *dev, 4815 struct list_head **iter); 4816 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4817 struct list_head **iter); 4818 4819 #define netdev_for_each_lower_private(dev, priv, iter) \ 4820 for (iter = (dev)->adj_list.lower.next, \ 4821 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4822 priv; \ 4823 priv = netdev_lower_get_next_private(dev, &(iter))) 4824 4825 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4826 for (iter = &(dev)->adj_list.lower, \ 4827 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4828 priv; \ 4829 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4830 4831 void *netdev_lower_get_next(struct net_device *dev, 4832 struct list_head **iter); 4833 4834 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4835 for (iter = (dev)->adj_list.lower.next, \ 4836 ldev = netdev_lower_get_next(dev, &(iter)); \ 4837 ldev; \ 4838 ldev = netdev_lower_get_next(dev, &(iter))) 4839 4840 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4841 struct list_head **iter); 4842 int netdev_walk_all_lower_dev(struct net_device *dev, 4843 int (*fn)(struct net_device *lower_dev, 4844 struct netdev_nested_priv *priv), 4845 struct netdev_nested_priv *priv); 4846 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4847 int (*fn)(struct net_device *lower_dev, 4848 struct netdev_nested_priv *priv), 4849 struct netdev_nested_priv *priv); 4850 4851 void *netdev_adjacent_get_private(struct list_head *adj_list); 4852 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4853 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4854 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4855 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4856 struct netlink_ext_ack *extack); 4857 int netdev_master_upper_dev_link(struct net_device *dev, 4858 struct net_device *upper_dev, 4859 void *upper_priv, void *upper_info, 4860 struct netlink_ext_ack *extack); 4861 void netdev_upper_dev_unlink(struct net_device *dev, 4862 struct net_device *upper_dev); 4863 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4864 struct net_device *new_dev, 4865 struct net_device *dev, 4866 struct netlink_ext_ack *extack); 4867 void netdev_adjacent_change_commit(struct net_device *old_dev, 4868 struct net_device *new_dev, 4869 struct net_device *dev); 4870 void netdev_adjacent_change_abort(struct net_device *old_dev, 4871 struct net_device *new_dev, 4872 struct net_device *dev); 4873 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4874 void *netdev_lower_dev_get_private(struct net_device *dev, 4875 struct net_device *lower_dev); 4876 void netdev_lower_state_changed(struct net_device *lower_dev, 4877 void *lower_state_info); 4878 4879 /* RSS keys are 40 or 52 bytes long */ 4880 #define NETDEV_RSS_KEY_LEN 52 4881 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4882 void netdev_rss_key_fill(void *buffer, size_t len); 4883 4884 int skb_checksum_help(struct sk_buff *skb); 4885 int skb_crc32c_csum_help(struct sk_buff *skb); 4886 int skb_csum_hwoffload_help(struct sk_buff *skb, 4887 const netdev_features_t features); 4888 4889 struct netdev_bonding_info { 4890 ifslave slave; 4891 ifbond master; 4892 }; 4893 4894 struct netdev_notifier_bonding_info { 4895 struct netdev_notifier_info info; /* must be first */ 4896 struct netdev_bonding_info bonding_info; 4897 }; 4898 4899 void netdev_bonding_info_change(struct net_device *dev, 4900 struct netdev_bonding_info *bonding_info); 4901 4902 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4903 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4904 #else 4905 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4906 const void *data) 4907 { 4908 } 4909 #endif 4910 4911 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4912 4913 static inline bool can_checksum_protocol(netdev_features_t features, 4914 __be16 protocol) 4915 { 4916 if (protocol == htons(ETH_P_FCOE)) 4917 return !!(features & NETIF_F_FCOE_CRC); 4918 4919 /* Assume this is an IP checksum (not SCTP CRC) */ 4920 4921 if (features & NETIF_F_HW_CSUM) { 4922 /* Can checksum everything */ 4923 return true; 4924 } 4925 4926 switch (protocol) { 4927 case htons(ETH_P_IP): 4928 return !!(features & NETIF_F_IP_CSUM); 4929 case htons(ETH_P_IPV6): 4930 return !!(features & NETIF_F_IPV6_CSUM); 4931 default: 4932 return false; 4933 } 4934 } 4935 4936 #ifdef CONFIG_BUG 4937 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4938 #else 4939 static inline void netdev_rx_csum_fault(struct net_device *dev, 4940 struct sk_buff *skb) 4941 { 4942 } 4943 #endif 4944 /* rx skb timestamps */ 4945 void net_enable_timestamp(void); 4946 void net_disable_timestamp(void); 4947 4948 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 4949 const struct skb_shared_hwtstamps *hwtstamps, 4950 bool cycles) 4951 { 4952 const struct net_device_ops *ops = dev->netdev_ops; 4953 4954 if (ops->ndo_get_tstamp) 4955 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 4956 4957 return hwtstamps->hwtstamp; 4958 } 4959 4960 #ifndef CONFIG_PREEMPT_RT 4961 static inline void netdev_xmit_set_more(bool more) 4962 { 4963 __this_cpu_write(softnet_data.xmit.more, more); 4964 } 4965 4966 static inline bool netdev_xmit_more(void) 4967 { 4968 return __this_cpu_read(softnet_data.xmit.more); 4969 } 4970 #else 4971 static inline void netdev_xmit_set_more(bool more) 4972 { 4973 current->net_xmit.more = more; 4974 } 4975 4976 static inline bool netdev_xmit_more(void) 4977 { 4978 return current->net_xmit.more; 4979 } 4980 #endif 4981 4982 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4983 struct sk_buff *skb, struct net_device *dev, 4984 bool more) 4985 { 4986 netdev_xmit_set_more(more); 4987 return ops->ndo_start_xmit(skb, dev); 4988 } 4989 4990 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4991 struct netdev_queue *txq, bool more) 4992 { 4993 const struct net_device_ops *ops = dev->netdev_ops; 4994 netdev_tx_t rc; 4995 4996 rc = __netdev_start_xmit(ops, skb, dev, more); 4997 if (rc == NETDEV_TX_OK) 4998 txq_trans_update(txq); 4999 5000 return rc; 5001 } 5002 5003 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 5004 const void *ns); 5005 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 5006 const void *ns); 5007 5008 extern const struct kobj_ns_type_operations net_ns_type_operations; 5009 5010 const char *netdev_drivername(const struct net_device *dev); 5011 5012 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 5013 netdev_features_t f2) 5014 { 5015 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 5016 if (f1 & NETIF_F_HW_CSUM) 5017 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5018 else 5019 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5020 } 5021 5022 return f1 & f2; 5023 } 5024 5025 static inline netdev_features_t netdev_get_wanted_features( 5026 struct net_device *dev) 5027 { 5028 return (dev->features & ~dev->hw_features) | dev->wanted_features; 5029 } 5030 netdev_features_t netdev_increment_features(netdev_features_t all, 5031 netdev_features_t one, netdev_features_t mask); 5032 5033 /* Allow TSO being used on stacked device : 5034 * Performing the GSO segmentation before last device 5035 * is a performance improvement. 5036 */ 5037 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 5038 netdev_features_t mask) 5039 { 5040 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 5041 } 5042 5043 int __netdev_update_features(struct net_device *dev); 5044 void netdev_update_features(struct net_device *dev); 5045 void netdev_change_features(struct net_device *dev); 5046 5047 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 5048 struct net_device *dev); 5049 5050 netdev_features_t passthru_features_check(struct sk_buff *skb, 5051 struct net_device *dev, 5052 netdev_features_t features); 5053 netdev_features_t netif_skb_features(struct sk_buff *skb); 5054 void skb_warn_bad_offload(const struct sk_buff *skb); 5055 5056 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 5057 { 5058 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 5059 5060 /* check flags correspondence */ 5061 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 5062 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 5063 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 5064 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 5065 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 5066 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 5067 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 5068 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 5069 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 5070 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 5071 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 5072 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 5073 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 5074 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 5075 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 5076 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 5077 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 5078 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 5079 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 5080 5081 return (features & feature) == feature; 5082 } 5083 5084 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 5085 { 5086 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 5087 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 5088 } 5089 5090 static inline bool netif_needs_gso(struct sk_buff *skb, 5091 netdev_features_t features) 5092 { 5093 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 5094 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 5095 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 5096 } 5097 5098 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 5099 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 5100 void netif_inherit_tso_max(struct net_device *to, 5101 const struct net_device *from); 5102 5103 static inline unsigned int 5104 netif_get_gro_max_size(const struct net_device *dev, const struct sk_buff *skb) 5105 { 5106 /* pairs with WRITE_ONCE() in netif_set_gro(_ipv4)_max_size() */ 5107 return skb->protocol == htons(ETH_P_IPV6) ? 5108 READ_ONCE(dev->gro_max_size) : 5109 READ_ONCE(dev->gro_ipv4_max_size); 5110 } 5111 5112 static inline unsigned int 5113 netif_get_gso_max_size(const struct net_device *dev, const struct sk_buff *skb) 5114 { 5115 /* pairs with WRITE_ONCE() in netif_set_gso(_ipv4)_max_size() */ 5116 return skb->protocol == htons(ETH_P_IPV6) ? 5117 READ_ONCE(dev->gso_max_size) : 5118 READ_ONCE(dev->gso_ipv4_max_size); 5119 } 5120 5121 static inline bool netif_is_macsec(const struct net_device *dev) 5122 { 5123 return dev->priv_flags & IFF_MACSEC; 5124 } 5125 5126 static inline bool netif_is_macvlan(const struct net_device *dev) 5127 { 5128 return dev->priv_flags & IFF_MACVLAN; 5129 } 5130 5131 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5132 { 5133 return dev->priv_flags & IFF_MACVLAN_PORT; 5134 } 5135 5136 static inline bool netif_is_bond_master(const struct net_device *dev) 5137 { 5138 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5139 } 5140 5141 static inline bool netif_is_bond_slave(const struct net_device *dev) 5142 { 5143 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5144 } 5145 5146 static inline bool netif_supports_nofcs(struct net_device *dev) 5147 { 5148 return dev->priv_flags & IFF_SUPP_NOFCS; 5149 } 5150 5151 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5152 { 5153 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5154 } 5155 5156 static inline bool netif_is_l3_master(const struct net_device *dev) 5157 { 5158 return dev->priv_flags & IFF_L3MDEV_MASTER; 5159 } 5160 5161 static inline bool netif_is_l3_slave(const struct net_device *dev) 5162 { 5163 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5164 } 5165 5166 static inline int dev_sdif(const struct net_device *dev) 5167 { 5168 #ifdef CONFIG_NET_L3_MASTER_DEV 5169 if (netif_is_l3_slave(dev)) 5170 return dev->ifindex; 5171 #endif 5172 return 0; 5173 } 5174 5175 static inline bool netif_is_bridge_master(const struct net_device *dev) 5176 { 5177 return dev->priv_flags & IFF_EBRIDGE; 5178 } 5179 5180 static inline bool netif_is_bridge_port(const struct net_device *dev) 5181 { 5182 return dev->priv_flags & IFF_BRIDGE_PORT; 5183 } 5184 5185 static inline bool netif_is_ovs_master(const struct net_device *dev) 5186 { 5187 return dev->priv_flags & IFF_OPENVSWITCH; 5188 } 5189 5190 static inline bool netif_is_ovs_port(const struct net_device *dev) 5191 { 5192 return dev->priv_flags & IFF_OVS_DATAPATH; 5193 } 5194 5195 static inline bool netif_is_any_bridge_master(const struct net_device *dev) 5196 { 5197 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev); 5198 } 5199 5200 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5201 { 5202 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5203 } 5204 5205 static inline bool netif_is_team_master(const struct net_device *dev) 5206 { 5207 return dev->priv_flags & IFF_TEAM; 5208 } 5209 5210 static inline bool netif_is_team_port(const struct net_device *dev) 5211 { 5212 return dev->priv_flags & IFF_TEAM_PORT; 5213 } 5214 5215 static inline bool netif_is_lag_master(const struct net_device *dev) 5216 { 5217 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5218 } 5219 5220 static inline bool netif_is_lag_port(const struct net_device *dev) 5221 { 5222 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5223 } 5224 5225 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5226 { 5227 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5228 } 5229 5230 static inline bool netif_is_failover(const struct net_device *dev) 5231 { 5232 return dev->priv_flags & IFF_FAILOVER; 5233 } 5234 5235 static inline bool netif_is_failover_slave(const struct net_device *dev) 5236 { 5237 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5238 } 5239 5240 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5241 static inline void netif_keep_dst(struct net_device *dev) 5242 { 5243 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5244 } 5245 5246 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5247 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5248 { 5249 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5250 return netif_is_macsec(dev); 5251 } 5252 5253 extern struct pernet_operations __net_initdata loopback_net_ops; 5254 5255 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5256 5257 /* netdev_printk helpers, similar to dev_printk */ 5258 5259 static inline const char *netdev_name(const struct net_device *dev) 5260 { 5261 if (!dev->name[0] || strchr(dev->name, '%')) 5262 return "(unnamed net_device)"; 5263 return dev->name; 5264 } 5265 5266 static inline const char *netdev_reg_state(const struct net_device *dev) 5267 { 5268 u8 reg_state = READ_ONCE(dev->reg_state); 5269 5270 switch (reg_state) { 5271 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5272 case NETREG_REGISTERED: return ""; 5273 case NETREG_UNREGISTERING: return " (unregistering)"; 5274 case NETREG_UNREGISTERED: return " (unregistered)"; 5275 case NETREG_RELEASED: return " (released)"; 5276 case NETREG_DUMMY: return " (dummy)"; 5277 } 5278 5279 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state); 5280 return " (unknown)"; 5281 } 5282 5283 #define MODULE_ALIAS_NETDEV(device) \ 5284 MODULE_ALIAS("netdev-" device) 5285 5286 /* 5287 * netdev_WARN() acts like dev_printk(), but with the key difference 5288 * of using a WARN/WARN_ON to get the message out, including the 5289 * file/line information and a backtrace. 5290 */ 5291 #define netdev_WARN(dev, format, args...) \ 5292 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5293 netdev_reg_state(dev), ##args) 5294 5295 #define netdev_WARN_ONCE(dev, format, args...) \ 5296 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5297 netdev_reg_state(dev), ##args) 5298 5299 /* 5300 * The list of packet types we will receive (as opposed to discard) 5301 * and the routines to invoke. 5302 * 5303 * Why 16. Because with 16 the only overlap we get on a hash of the 5304 * low nibble of the protocol value is RARP/SNAP/X.25. 5305 * 5306 * 0800 IP 5307 * 0001 802.3 5308 * 0002 AX.25 5309 * 0004 802.2 5310 * 8035 RARP 5311 * 0005 SNAP 5312 * 0805 X.25 5313 * 0806 ARP 5314 * 8137 IPX 5315 * 0009 Localtalk 5316 * 86DD IPv6 5317 */ 5318 #define PTYPE_HASH_SIZE (16) 5319 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5320 5321 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5322 5323 extern struct net_device *blackhole_netdev; 5324 5325 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */ 5326 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD) 5327 #define DEV_STATS_ADD(DEV, FIELD, VAL) \ 5328 atomic_long_add((VAL), &(DEV)->stats.__##FIELD) 5329 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD) 5330 5331 #endif /* _LINUX_NETDEVICE_H */ 5332