1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <[email protected]> 13 * Corey Minyard <[email protected]> 14 * Donald J. Becker, <[email protected]> 15 * Alan Cox, <[email protected]> 16 * Bjorn Ekwall. <[email protected]> 17 * Pekka Riikonen <[email protected]> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 #include <asm/local.h> 32 33 #include <linux/percpu.h> 34 #include <linux/rculist.h> 35 #include <linux/workqueue.h> 36 #include <linux/dynamic_queue_limits.h> 37 38 #include <net/net_namespace.h> 39 #ifdef CONFIG_DCB 40 #include <net/dcbnl.h> 41 #endif 42 #include <net/netprio_cgroup.h> 43 44 #include <linux/netdev_features.h> 45 #include <linux/neighbour.h> 46 #include <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; 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 /* DPLL specific */ 83 struct dpll_pin; 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 #ifdef CONFIG_RPS 231 #include <linux/static_key.h> 232 extern struct static_key_false rps_needed; 233 extern struct static_key_false rfs_needed; 234 #endif 235 236 struct neighbour; 237 struct neigh_parms; 238 struct sk_buff; 239 240 struct netdev_hw_addr { 241 struct list_head list; 242 struct rb_node node; 243 unsigned char addr[MAX_ADDR_LEN]; 244 unsigned char type; 245 #define NETDEV_HW_ADDR_T_LAN 1 246 #define NETDEV_HW_ADDR_T_SAN 2 247 #define NETDEV_HW_ADDR_T_UNICAST 3 248 #define NETDEV_HW_ADDR_T_MULTICAST 4 249 bool global_use; 250 int sync_cnt; 251 int refcount; 252 int synced; 253 struct rcu_head rcu_head; 254 }; 255 256 struct netdev_hw_addr_list { 257 struct list_head list; 258 int count; 259 260 /* Auxiliary tree for faster lookup on addition and deletion */ 261 struct rb_root tree; 262 }; 263 264 #define netdev_hw_addr_list_count(l) ((l)->count) 265 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 266 #define netdev_hw_addr_list_for_each(ha, l) \ 267 list_for_each_entry(ha, &(l)->list, list) 268 269 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 270 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 271 #define netdev_for_each_uc_addr(ha, dev) \ 272 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 273 #define netdev_for_each_synced_uc_addr(_ha, _dev) \ 274 netdev_for_each_uc_addr((_ha), (_dev)) \ 275 if ((_ha)->sync_cnt) 276 277 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 278 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 279 #define netdev_for_each_mc_addr(ha, dev) \ 280 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 281 #define netdev_for_each_synced_mc_addr(_ha, _dev) \ 282 netdev_for_each_mc_addr((_ha), (_dev)) \ 283 if ((_ha)->sync_cnt) 284 285 struct hh_cache { 286 unsigned int hh_len; 287 seqlock_t hh_lock; 288 289 /* cached hardware header; allow for machine alignment needs. */ 290 #define HH_DATA_MOD 16 291 #define HH_DATA_OFF(__len) \ 292 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 293 #define HH_DATA_ALIGN(__len) \ 294 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 295 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 296 }; 297 298 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 299 * Alternative is: 300 * dev->hard_header_len ? (dev->hard_header_len + 301 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 302 * 303 * We could use other alignment values, but we must maintain the 304 * relationship HH alignment <= LL alignment. 305 */ 306 #define LL_RESERVED_SPACE(dev) \ 307 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \ 308 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 309 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 310 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \ 311 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 312 313 struct header_ops { 314 int (*create) (struct sk_buff *skb, struct net_device *dev, 315 unsigned short type, const void *daddr, 316 const void *saddr, unsigned int len); 317 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 318 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 319 void (*cache_update)(struct hh_cache *hh, 320 const struct net_device *dev, 321 const unsigned char *haddr); 322 bool (*validate)(const char *ll_header, unsigned int len); 323 __be16 (*parse_protocol)(const struct sk_buff *skb); 324 }; 325 326 /* These flag bits are private to the generic network queueing 327 * layer; they may not be explicitly referenced by any other 328 * code. 329 */ 330 331 enum netdev_state_t { 332 __LINK_STATE_START, 333 __LINK_STATE_PRESENT, 334 __LINK_STATE_NOCARRIER, 335 __LINK_STATE_LINKWATCH_PENDING, 336 __LINK_STATE_DORMANT, 337 __LINK_STATE_TESTING, 338 }; 339 340 struct gro_list { 341 struct list_head list; 342 int count; 343 }; 344 345 /* 346 * size of gro hash buckets, must less than bit number of 347 * napi_struct::gro_bitmask 348 */ 349 #define GRO_HASH_BUCKETS 8 350 351 /* 352 * Structure for NAPI scheduling similar to tasklet but with weighting 353 */ 354 struct napi_struct { 355 /* The poll_list must only be managed by the entity which 356 * changes the state of the NAPI_STATE_SCHED bit. This means 357 * whoever atomically sets that bit can add this napi_struct 358 * to the per-CPU poll_list, and whoever clears that bit 359 * can remove from the list right before clearing the bit. 360 */ 361 struct list_head poll_list; 362 363 unsigned long state; 364 int weight; 365 int defer_hard_irqs_count; 366 unsigned long gro_bitmask; 367 int (*poll)(struct napi_struct *, int); 368 #ifdef CONFIG_NETPOLL 369 /* CPU actively polling if netpoll is configured */ 370 int poll_owner; 371 #endif 372 /* CPU on which NAPI has been scheduled for processing */ 373 int list_owner; 374 struct net_device *dev; 375 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 376 struct sk_buff *skb; 377 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 378 int rx_count; /* length of rx_list */ 379 unsigned int napi_id; 380 struct hrtimer timer; 381 struct task_struct *thread; 382 /* control-path-only fields follow */ 383 struct list_head dev_list; 384 struct hlist_node napi_hash_node; 385 int irq; 386 }; 387 388 enum { 389 NAPI_STATE_SCHED, /* Poll is scheduled */ 390 NAPI_STATE_MISSED, /* reschedule a napi */ 391 NAPI_STATE_DISABLE, /* Disable pending */ 392 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 393 NAPI_STATE_LISTED, /* NAPI added to system lists */ 394 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 395 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 396 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 397 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 398 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 399 }; 400 401 enum { 402 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 403 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 404 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 405 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 406 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 407 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 408 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 409 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 410 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 411 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 412 }; 413 414 enum gro_result { 415 GRO_MERGED, 416 GRO_MERGED_FREE, 417 GRO_HELD, 418 GRO_NORMAL, 419 GRO_CONSUMED, 420 }; 421 typedef enum gro_result gro_result_t; 422 423 /* 424 * enum rx_handler_result - Possible return values for rx_handlers. 425 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 426 * further. 427 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 428 * case skb->dev was changed by rx_handler. 429 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 430 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 431 * 432 * rx_handlers are functions called from inside __netif_receive_skb(), to do 433 * special processing of the skb, prior to delivery to protocol handlers. 434 * 435 * Currently, a net_device can only have a single rx_handler registered. Trying 436 * to register a second rx_handler will return -EBUSY. 437 * 438 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 439 * To unregister a rx_handler on a net_device, use 440 * netdev_rx_handler_unregister(). 441 * 442 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 443 * do with the skb. 444 * 445 * If the rx_handler consumed the skb in some way, it should return 446 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 447 * the skb to be delivered in some other way. 448 * 449 * If the rx_handler changed skb->dev, to divert the skb to another 450 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 451 * new device will be called if it exists. 452 * 453 * If the rx_handler decides the skb should be ignored, it should return 454 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 455 * are registered on exact device (ptype->dev == skb->dev). 456 * 457 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 458 * delivered, it should return RX_HANDLER_PASS. 459 * 460 * A device without a registered rx_handler will behave as if rx_handler 461 * returned RX_HANDLER_PASS. 462 */ 463 464 enum rx_handler_result { 465 RX_HANDLER_CONSUMED, 466 RX_HANDLER_ANOTHER, 467 RX_HANDLER_EXACT, 468 RX_HANDLER_PASS, 469 }; 470 typedef enum rx_handler_result rx_handler_result_t; 471 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 472 473 void __napi_schedule(struct napi_struct *n); 474 void __napi_schedule_irqoff(struct napi_struct *n); 475 476 static inline bool napi_disable_pending(struct napi_struct *n) 477 { 478 return test_bit(NAPI_STATE_DISABLE, &n->state); 479 } 480 481 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 482 { 483 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 484 } 485 486 /** 487 * napi_is_scheduled - test if NAPI is scheduled 488 * @n: NAPI context 489 * 490 * This check is "best-effort". With no locking implemented, 491 * a NAPI can be scheduled or terminate right after this check 492 * and produce not precise results. 493 * 494 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled 495 * should not be used normally and napi_schedule should be 496 * used instead. 497 * 498 * Use only if the driver really needs to check if a NAPI 499 * is scheduled for example in the context of delayed timer 500 * that can be skipped if a NAPI is already scheduled. 501 * 502 * Return True if NAPI is scheduled, False otherwise. 503 */ 504 static inline bool napi_is_scheduled(struct napi_struct *n) 505 { 506 return test_bit(NAPI_STATE_SCHED, &n->state); 507 } 508 509 bool napi_schedule_prep(struct napi_struct *n); 510 511 /** 512 * napi_schedule - schedule NAPI poll 513 * @n: NAPI context 514 * 515 * Schedule NAPI poll routine to be called if it is not already 516 * running. 517 * Return true if we schedule a NAPI or false if not. 518 * Refer to napi_schedule_prep() for additional reason on why 519 * a NAPI might not be scheduled. 520 */ 521 static inline bool napi_schedule(struct napi_struct *n) 522 { 523 if (napi_schedule_prep(n)) { 524 __napi_schedule(n); 525 return true; 526 } 527 528 return false; 529 } 530 531 /** 532 * napi_schedule_irqoff - schedule NAPI poll 533 * @n: NAPI context 534 * 535 * Variant of napi_schedule(), assuming hard irqs are masked. 536 */ 537 static inline void napi_schedule_irqoff(struct napi_struct *n) 538 { 539 if (napi_schedule_prep(n)) 540 __napi_schedule_irqoff(n); 541 } 542 543 /** 544 * napi_complete_done - NAPI processing complete 545 * @n: NAPI context 546 * @work_done: number of packets processed 547 * 548 * Mark NAPI processing as complete. Should only be called if poll budget 549 * has not been completely consumed. 550 * Prefer over napi_complete(). 551 * Return false if device should avoid rearming interrupts. 552 */ 553 bool napi_complete_done(struct napi_struct *n, int work_done); 554 555 static inline bool napi_complete(struct napi_struct *n) 556 { 557 return napi_complete_done(n, 0); 558 } 559 560 int dev_set_threaded(struct net_device *dev, bool threaded); 561 562 /** 563 * napi_disable - prevent NAPI from scheduling 564 * @n: NAPI context 565 * 566 * Stop NAPI from being scheduled on this context. 567 * Waits till any outstanding processing completes. 568 */ 569 void napi_disable(struct napi_struct *n); 570 571 void napi_enable(struct napi_struct *n); 572 573 /** 574 * napi_synchronize - wait until NAPI is not running 575 * @n: NAPI context 576 * 577 * Wait until NAPI is done being scheduled on this context. 578 * Waits till any outstanding processing completes but 579 * does not disable future activations. 580 */ 581 static inline void napi_synchronize(const struct napi_struct *n) 582 { 583 if (IS_ENABLED(CONFIG_SMP)) 584 while (test_bit(NAPI_STATE_SCHED, &n->state)) 585 msleep(1); 586 else 587 barrier(); 588 } 589 590 /** 591 * napi_if_scheduled_mark_missed - if napi is running, set the 592 * NAPIF_STATE_MISSED 593 * @n: NAPI context 594 * 595 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 596 * NAPI is scheduled. 597 **/ 598 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 599 { 600 unsigned long val, new; 601 602 val = READ_ONCE(n->state); 603 do { 604 if (val & NAPIF_STATE_DISABLE) 605 return true; 606 607 if (!(val & NAPIF_STATE_SCHED)) 608 return false; 609 610 new = val | NAPIF_STATE_MISSED; 611 } while (!try_cmpxchg(&n->state, &val, new)); 612 613 return true; 614 } 615 616 enum netdev_queue_state_t { 617 __QUEUE_STATE_DRV_XOFF, 618 __QUEUE_STATE_STACK_XOFF, 619 __QUEUE_STATE_FROZEN, 620 }; 621 622 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 623 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 624 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 625 626 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 627 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 628 QUEUE_STATE_FROZEN) 629 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 630 QUEUE_STATE_FROZEN) 631 632 /* 633 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 634 * netif_tx_* functions below are used to manipulate this flag. The 635 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 636 * queue independently. The netif_xmit_*stopped functions below are called 637 * to check if the queue has been stopped by the driver or stack (either 638 * of the XOFF bits are set in the state). Drivers should not need to call 639 * netif_xmit*stopped functions, they should only be using netif_tx_*. 640 */ 641 642 struct netdev_queue { 643 /* 644 * read-mostly part 645 */ 646 struct net_device *dev; 647 netdevice_tracker dev_tracker; 648 649 struct Qdisc __rcu *qdisc; 650 struct Qdisc __rcu *qdisc_sleeping; 651 #ifdef CONFIG_SYSFS 652 struct kobject kobj; 653 #endif 654 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 655 int numa_node; 656 #endif 657 unsigned long tx_maxrate; 658 /* 659 * Number of TX timeouts for this queue 660 * (/sys/class/net/DEV/Q/trans_timeout) 661 */ 662 atomic_long_t trans_timeout; 663 664 /* Subordinate device that the queue has been assigned to */ 665 struct net_device *sb_dev; 666 #ifdef CONFIG_XDP_SOCKETS 667 struct xsk_buff_pool *pool; 668 #endif 669 /* NAPI instance for the queue 670 * Readers and writers must hold RTNL 671 */ 672 struct napi_struct *napi; 673 /* 674 * write-mostly part 675 */ 676 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 677 int xmit_lock_owner; 678 /* 679 * Time (in jiffies) of last Tx 680 */ 681 unsigned long trans_start; 682 683 unsigned long state; 684 685 #ifdef CONFIG_BQL 686 struct dql dql; 687 #endif 688 } ____cacheline_aligned_in_smp; 689 690 extern int sysctl_fb_tunnels_only_for_init_net; 691 extern int sysctl_devconf_inherit_init_net; 692 693 /* 694 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 695 * == 1 : For initns only 696 * == 2 : For none. 697 */ 698 static inline bool net_has_fallback_tunnels(const struct net *net) 699 { 700 #if IS_ENABLED(CONFIG_SYSCTL) 701 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net); 702 703 return !fb_tunnels_only_for_init_net || 704 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1); 705 #else 706 return true; 707 #endif 708 } 709 710 static inline int net_inherit_devconf(void) 711 { 712 #if IS_ENABLED(CONFIG_SYSCTL) 713 return READ_ONCE(sysctl_devconf_inherit_init_net); 714 #else 715 return 0; 716 #endif 717 } 718 719 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 720 { 721 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 722 return q->numa_node; 723 #else 724 return NUMA_NO_NODE; 725 #endif 726 } 727 728 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 729 { 730 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 731 q->numa_node = node; 732 #endif 733 } 734 735 #ifdef CONFIG_RPS 736 /* 737 * This structure holds an RPS map which can be of variable length. The 738 * map is an array of CPUs. 739 */ 740 struct rps_map { 741 unsigned int len; 742 struct rcu_head rcu; 743 u16 cpus[]; 744 }; 745 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 746 747 /* 748 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 749 * tail pointer for that CPU's input queue at the time of last enqueue, and 750 * a hardware filter index. 751 */ 752 struct rps_dev_flow { 753 u16 cpu; 754 u16 filter; 755 unsigned int last_qtail; 756 }; 757 #define RPS_NO_FILTER 0xffff 758 759 /* 760 * The rps_dev_flow_table structure contains a table of flow mappings. 761 */ 762 struct rps_dev_flow_table { 763 unsigned int mask; 764 struct rcu_head rcu; 765 struct rps_dev_flow flows[]; 766 }; 767 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 768 ((_num) * sizeof(struct rps_dev_flow))) 769 770 /* 771 * The rps_sock_flow_table contains mappings of flows to the last CPU 772 * on which they were processed by the application (set in recvmsg). 773 * Each entry is a 32bit value. Upper part is the high-order bits 774 * of flow hash, lower part is CPU number. 775 * rps_cpu_mask is used to partition the space, depending on number of 776 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 777 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 778 * meaning we use 32-6=26 bits for the hash. 779 */ 780 struct rps_sock_flow_table { 781 u32 mask; 782 783 u32 ents[] ____cacheline_aligned_in_smp; 784 }; 785 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 786 787 #define RPS_NO_CPU 0xffff 788 789 extern u32 rps_cpu_mask; 790 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 791 792 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 793 u32 hash) 794 { 795 if (table && hash) { 796 unsigned int index = hash & table->mask; 797 u32 val = hash & ~rps_cpu_mask; 798 799 /* We only give a hint, preemption can change CPU under us */ 800 val |= raw_smp_processor_id(); 801 802 /* The following WRITE_ONCE() is paired with the READ_ONCE() 803 * here, and another one in get_rps_cpu(). 804 */ 805 if (READ_ONCE(table->ents[index]) != val) 806 WRITE_ONCE(table->ents[index], val); 807 } 808 } 809 810 #ifdef CONFIG_RFS_ACCEL 811 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 812 u16 filter_id); 813 #endif 814 #endif /* CONFIG_RPS */ 815 816 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 817 enum xps_map_type { 818 XPS_CPUS = 0, 819 XPS_RXQS, 820 XPS_MAPS_MAX, 821 }; 822 823 #ifdef CONFIG_XPS 824 /* 825 * This structure holds an XPS map which can be of variable length. The 826 * map is an array of queues. 827 */ 828 struct xps_map { 829 unsigned int len; 830 unsigned int alloc_len; 831 struct rcu_head rcu; 832 u16 queues[]; 833 }; 834 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 835 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 836 - sizeof(struct xps_map)) / sizeof(u16)) 837 838 /* 839 * This structure holds all XPS maps for device. Maps are indexed by CPU. 840 * 841 * We keep track of the number of cpus/rxqs used when the struct is allocated, 842 * in nr_ids. This will help not accessing out-of-bound memory. 843 * 844 * We keep track of the number of traffic classes used when the struct is 845 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 846 * not crossing its upper bound, as the original dev->num_tc can be updated in 847 * the meantime. 848 */ 849 struct xps_dev_maps { 850 struct rcu_head rcu; 851 unsigned int nr_ids; 852 s16 num_tc; 853 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 854 }; 855 856 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 857 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 858 859 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 860 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 861 862 #endif /* CONFIG_XPS */ 863 864 #define TC_MAX_QUEUE 16 865 #define TC_BITMASK 15 866 /* HW offloaded queuing disciplines txq count and offset maps */ 867 struct netdev_tc_txq { 868 u16 count; 869 u16 offset; 870 }; 871 872 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 873 /* 874 * This structure is to hold information about the device 875 * configured to run FCoE protocol stack. 876 */ 877 struct netdev_fcoe_hbainfo { 878 char manufacturer[64]; 879 char serial_number[64]; 880 char hardware_version[64]; 881 char driver_version[64]; 882 char optionrom_version[64]; 883 char firmware_version[64]; 884 char model[256]; 885 char model_description[256]; 886 }; 887 #endif 888 889 #define MAX_PHYS_ITEM_ID_LEN 32 890 891 /* This structure holds a unique identifier to identify some 892 * physical item (port for example) used by a netdevice. 893 */ 894 struct netdev_phys_item_id { 895 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 896 unsigned char id_len; 897 }; 898 899 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 900 struct netdev_phys_item_id *b) 901 { 902 return a->id_len == b->id_len && 903 memcmp(a->id, b->id, a->id_len) == 0; 904 } 905 906 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 907 struct sk_buff *skb, 908 struct net_device *sb_dev); 909 910 enum net_device_path_type { 911 DEV_PATH_ETHERNET = 0, 912 DEV_PATH_VLAN, 913 DEV_PATH_BRIDGE, 914 DEV_PATH_PPPOE, 915 DEV_PATH_DSA, 916 DEV_PATH_MTK_WDMA, 917 }; 918 919 struct net_device_path { 920 enum net_device_path_type type; 921 const struct net_device *dev; 922 union { 923 struct { 924 u16 id; 925 __be16 proto; 926 u8 h_dest[ETH_ALEN]; 927 } encap; 928 struct { 929 enum { 930 DEV_PATH_BR_VLAN_KEEP, 931 DEV_PATH_BR_VLAN_TAG, 932 DEV_PATH_BR_VLAN_UNTAG, 933 DEV_PATH_BR_VLAN_UNTAG_HW, 934 } vlan_mode; 935 u16 vlan_id; 936 __be16 vlan_proto; 937 } bridge; 938 struct { 939 int port; 940 u16 proto; 941 } dsa; 942 struct { 943 u8 wdma_idx; 944 u8 queue; 945 u16 wcid; 946 u8 bss; 947 u8 amsdu; 948 } mtk_wdma; 949 }; 950 }; 951 952 #define NET_DEVICE_PATH_STACK_MAX 5 953 #define NET_DEVICE_PATH_VLAN_MAX 2 954 955 struct net_device_path_stack { 956 int num_paths; 957 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 958 }; 959 960 struct net_device_path_ctx { 961 const struct net_device *dev; 962 u8 daddr[ETH_ALEN]; 963 964 int num_vlans; 965 struct { 966 u16 id; 967 __be16 proto; 968 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 969 }; 970 971 enum tc_setup_type { 972 TC_QUERY_CAPS, 973 TC_SETUP_QDISC_MQPRIO, 974 TC_SETUP_CLSU32, 975 TC_SETUP_CLSFLOWER, 976 TC_SETUP_CLSMATCHALL, 977 TC_SETUP_CLSBPF, 978 TC_SETUP_BLOCK, 979 TC_SETUP_QDISC_CBS, 980 TC_SETUP_QDISC_RED, 981 TC_SETUP_QDISC_PRIO, 982 TC_SETUP_QDISC_MQ, 983 TC_SETUP_QDISC_ETF, 984 TC_SETUP_ROOT_QDISC, 985 TC_SETUP_QDISC_GRED, 986 TC_SETUP_QDISC_TAPRIO, 987 TC_SETUP_FT, 988 TC_SETUP_QDISC_ETS, 989 TC_SETUP_QDISC_TBF, 990 TC_SETUP_QDISC_FIFO, 991 TC_SETUP_QDISC_HTB, 992 TC_SETUP_ACT, 993 }; 994 995 /* These structures hold the attributes of bpf state that are being passed 996 * to the netdevice through the bpf op. 997 */ 998 enum bpf_netdev_command { 999 /* Set or clear a bpf program used in the earliest stages of packet 1000 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 1001 * is responsible for calling bpf_prog_put on any old progs that are 1002 * stored. In case of error, the callee need not release the new prog 1003 * reference, but on success it takes ownership and must bpf_prog_put 1004 * when it is no longer used. 1005 */ 1006 XDP_SETUP_PROG, 1007 XDP_SETUP_PROG_HW, 1008 /* BPF program for offload callbacks, invoked at program load time. */ 1009 BPF_OFFLOAD_MAP_ALLOC, 1010 BPF_OFFLOAD_MAP_FREE, 1011 XDP_SETUP_XSK_POOL, 1012 }; 1013 1014 struct bpf_prog_offload_ops; 1015 struct netlink_ext_ack; 1016 struct xdp_umem; 1017 struct xdp_dev_bulk_queue; 1018 struct bpf_xdp_link; 1019 1020 enum bpf_xdp_mode { 1021 XDP_MODE_SKB = 0, 1022 XDP_MODE_DRV = 1, 1023 XDP_MODE_HW = 2, 1024 __MAX_XDP_MODE 1025 }; 1026 1027 struct bpf_xdp_entity { 1028 struct bpf_prog *prog; 1029 struct bpf_xdp_link *link; 1030 }; 1031 1032 struct netdev_bpf { 1033 enum bpf_netdev_command command; 1034 union { 1035 /* XDP_SETUP_PROG */ 1036 struct { 1037 u32 flags; 1038 struct bpf_prog *prog; 1039 struct netlink_ext_ack *extack; 1040 }; 1041 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 1042 struct { 1043 struct bpf_offloaded_map *offmap; 1044 }; 1045 /* XDP_SETUP_XSK_POOL */ 1046 struct { 1047 struct xsk_buff_pool *pool; 1048 u16 queue_id; 1049 } xsk; 1050 }; 1051 }; 1052 1053 /* Flags for ndo_xsk_wakeup. */ 1054 #define XDP_WAKEUP_RX (1 << 0) 1055 #define XDP_WAKEUP_TX (1 << 1) 1056 1057 #ifdef CONFIG_XFRM_OFFLOAD 1058 struct xfrmdev_ops { 1059 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack); 1060 void (*xdo_dev_state_delete) (struct xfrm_state *x); 1061 void (*xdo_dev_state_free) (struct xfrm_state *x); 1062 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 1063 struct xfrm_state *x); 1064 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1065 void (*xdo_dev_state_update_curlft) (struct xfrm_state *x); 1066 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack); 1067 void (*xdo_dev_policy_delete) (struct xfrm_policy *x); 1068 void (*xdo_dev_policy_free) (struct xfrm_policy *x); 1069 }; 1070 #endif 1071 1072 struct dev_ifalias { 1073 struct rcu_head rcuhead; 1074 char ifalias[]; 1075 }; 1076 1077 struct devlink; 1078 struct tlsdev_ops; 1079 1080 struct netdev_net_notifier { 1081 struct list_head list; 1082 struct notifier_block *nb; 1083 }; 1084 1085 /* 1086 * This structure defines the management hooks for network devices. 1087 * The following hooks can be defined; unless noted otherwise, they are 1088 * optional and can be filled with a null pointer. 1089 * 1090 * int (*ndo_init)(struct net_device *dev); 1091 * This function is called once when a network device is registered. 1092 * The network device can use this for any late stage initialization 1093 * or semantic validation. It can fail with an error code which will 1094 * be propagated back to register_netdev. 1095 * 1096 * void (*ndo_uninit)(struct net_device *dev); 1097 * This function is called when device is unregistered or when registration 1098 * fails. It is not called if init fails. 1099 * 1100 * int (*ndo_open)(struct net_device *dev); 1101 * This function is called when a network device transitions to the up 1102 * state. 1103 * 1104 * int (*ndo_stop)(struct net_device *dev); 1105 * This function is called when a network device transitions to the down 1106 * state. 1107 * 1108 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1109 * struct net_device *dev); 1110 * Called when a packet needs to be transmitted. 1111 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1112 * the queue before that can happen; it's for obsolete devices and weird 1113 * corner cases, but the stack really does a non-trivial amount 1114 * of useless work if you return NETDEV_TX_BUSY. 1115 * Required; cannot be NULL. 1116 * 1117 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1118 * struct net_device *dev 1119 * netdev_features_t features); 1120 * Called by core transmit path to determine if device is capable of 1121 * performing offload operations on a given packet. This is to give 1122 * the device an opportunity to implement any restrictions that cannot 1123 * be otherwise expressed by feature flags. The check is called with 1124 * the set of features that the stack has calculated and it returns 1125 * those the driver believes to be appropriate. 1126 * 1127 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1128 * struct net_device *sb_dev); 1129 * Called to decide which queue to use when device supports multiple 1130 * transmit queues. 1131 * 1132 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1133 * This function is called to allow device receiver to make 1134 * changes to configuration when multicast or promiscuous is enabled. 1135 * 1136 * void (*ndo_set_rx_mode)(struct net_device *dev); 1137 * This function is called device changes address list filtering. 1138 * If driver handles unicast address filtering, it should set 1139 * IFF_UNICAST_FLT in its priv_flags. 1140 * 1141 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1142 * This function is called when the Media Access Control address 1143 * needs to be changed. If this interface is not defined, the 1144 * MAC address can not be changed. 1145 * 1146 * int (*ndo_validate_addr)(struct net_device *dev); 1147 * Test if Media Access Control address is valid for the device. 1148 * 1149 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1150 * Old-style ioctl entry point. This is used internally by the 1151 * appletalk and ieee802154 subsystems but is no longer called by 1152 * the device ioctl handler. 1153 * 1154 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1155 * Used by the bonding driver for its device specific ioctls: 1156 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1157 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1158 * 1159 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1160 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1161 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1162 * 1163 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1164 * Used to set network devices bus interface parameters. This interface 1165 * is retained for legacy reasons; new devices should use the bus 1166 * interface (PCI) for low level management. 1167 * 1168 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1169 * Called when a user wants to change the Maximum Transfer Unit 1170 * of a device. 1171 * 1172 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1173 * Callback used when the transmitter has not made any progress 1174 * for dev->watchdog ticks. 1175 * 1176 * void (*ndo_get_stats64)(struct net_device *dev, 1177 * struct rtnl_link_stats64 *storage); 1178 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1179 * Called when a user wants to get the network device usage 1180 * statistics. Drivers must do one of the following: 1181 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1182 * rtnl_link_stats64 structure passed by the caller. 1183 * 2. Define @ndo_get_stats to update a net_device_stats structure 1184 * (which should normally be dev->stats) and return a pointer to 1185 * it. The structure may be changed asynchronously only if each 1186 * field is written atomically. 1187 * 3. Update dev->stats asynchronously and atomically, and define 1188 * neither operation. 1189 * 1190 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1191 * Return true if this device supports offload stats of this attr_id. 1192 * 1193 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1194 * void *attr_data) 1195 * Get statistics for offload operations by attr_id. Write it into the 1196 * attr_data pointer. 1197 * 1198 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1199 * If device supports VLAN filtering this function is called when a 1200 * VLAN id is registered. 1201 * 1202 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1203 * If device supports VLAN filtering this function is called when a 1204 * VLAN id is unregistered. 1205 * 1206 * void (*ndo_poll_controller)(struct net_device *dev); 1207 * 1208 * SR-IOV management functions. 1209 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1210 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1211 * u8 qos, __be16 proto); 1212 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1213 * int max_tx_rate); 1214 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1215 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1216 * int (*ndo_get_vf_config)(struct net_device *dev, 1217 * int vf, struct ifla_vf_info *ivf); 1218 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1219 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1220 * struct nlattr *port[]); 1221 * 1222 * Enable or disable the VF ability to query its RSS Redirection Table and 1223 * Hash Key. This is needed since on some devices VF share this information 1224 * with PF and querying it may introduce a theoretical security risk. 1225 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1226 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1227 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1228 * void *type_data); 1229 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1230 * This is always called from the stack with the rtnl lock held and netif 1231 * tx queues stopped. This allows the netdevice to perform queue 1232 * management safely. 1233 * 1234 * Fiber Channel over Ethernet (FCoE) offload functions. 1235 * int (*ndo_fcoe_enable)(struct net_device *dev); 1236 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1237 * so the underlying device can perform whatever needed configuration or 1238 * initialization to support acceleration of FCoE traffic. 1239 * 1240 * int (*ndo_fcoe_disable)(struct net_device *dev); 1241 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1242 * so the underlying device can perform whatever needed clean-ups to 1243 * stop supporting acceleration of FCoE traffic. 1244 * 1245 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1246 * struct scatterlist *sgl, unsigned int sgc); 1247 * Called when the FCoE Initiator wants to initialize an I/O that 1248 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1249 * perform necessary setup and returns 1 to indicate the device is set up 1250 * successfully to perform DDP on this I/O, otherwise this returns 0. 1251 * 1252 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1253 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1254 * indicated by the FC exchange id 'xid', so the underlying device can 1255 * clean up and reuse resources for later DDP requests. 1256 * 1257 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1258 * struct scatterlist *sgl, unsigned int sgc); 1259 * Called when the FCoE Target wants to initialize an I/O that 1260 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1261 * perform necessary setup and returns 1 to indicate the device is set up 1262 * successfully to perform DDP on this I/O, otherwise this returns 0. 1263 * 1264 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1265 * struct netdev_fcoe_hbainfo *hbainfo); 1266 * Called when the FCoE Protocol stack wants information on the underlying 1267 * device. This information is utilized by the FCoE protocol stack to 1268 * register attributes with Fiber Channel management service as per the 1269 * FC-GS Fabric Device Management Information(FDMI) specification. 1270 * 1271 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1272 * Called when the underlying device wants to override default World Wide 1273 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1274 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1275 * protocol stack to use. 1276 * 1277 * RFS acceleration. 1278 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1279 * u16 rxq_index, u32 flow_id); 1280 * Set hardware filter for RFS. rxq_index is the target queue index; 1281 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1282 * Return the filter ID on success, or a negative error code. 1283 * 1284 * Slave management functions (for bridge, bonding, etc). 1285 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1286 * Called to make another netdev an underling. 1287 * 1288 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1289 * Called to release previously enslaved netdev. 1290 * 1291 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1292 * struct sk_buff *skb, 1293 * bool all_slaves); 1294 * Get the xmit slave of master device. If all_slaves is true, function 1295 * assume all the slaves can transmit. 1296 * 1297 * Feature/offload setting functions. 1298 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1299 * netdev_features_t features); 1300 * Adjusts the requested feature flags according to device-specific 1301 * constraints, and returns the resulting flags. Must not modify 1302 * the device state. 1303 * 1304 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1305 * Called to update device configuration to new features. Passed 1306 * feature set might be less than what was returned by ndo_fix_features()). 1307 * Must return >0 or -errno if it changed dev->features itself. 1308 * 1309 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1310 * struct net_device *dev, 1311 * const unsigned char *addr, u16 vid, u16 flags, 1312 * struct netlink_ext_ack *extack); 1313 * Adds an FDB entry to dev for addr. 1314 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1315 * struct net_device *dev, 1316 * const unsigned char *addr, u16 vid) 1317 * Deletes the FDB entry from dev coresponding to addr. 1318 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev, 1319 * struct netlink_ext_ack *extack); 1320 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1321 * struct net_device *dev, struct net_device *filter_dev, 1322 * int *idx) 1323 * Used to add FDB entries to dump requests. Implementers should add 1324 * entries to skb and update idx with the number of entries. 1325 * 1326 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[], 1327 * u16 nlmsg_flags, struct netlink_ext_ack *extack); 1328 * Adds an MDB entry to dev. 1329 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[], 1330 * struct netlink_ext_ack *extack); 1331 * Deletes the MDB entry from dev. 1332 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb, 1333 * struct netlink_callback *cb); 1334 * Dumps MDB entries from dev. The first argument (marker) in the netlink 1335 * callback is used by core rtnetlink code. 1336 * 1337 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1338 * u16 flags, struct netlink_ext_ack *extack) 1339 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1340 * struct net_device *dev, u32 filter_mask, 1341 * int nlflags) 1342 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1343 * u16 flags); 1344 * 1345 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1346 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1347 * which do not represent real hardware may define this to allow their 1348 * userspace components to manage their virtual carrier state. Devices 1349 * that determine carrier state from physical hardware properties (eg 1350 * network cables) or protocol-dependent mechanisms (eg 1351 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1352 * 1353 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1354 * struct netdev_phys_item_id *ppid); 1355 * Called to get ID of physical port of this device. If driver does 1356 * not implement this, it is assumed that the hw is not able to have 1357 * multiple net devices on single physical port. 1358 * 1359 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1360 * struct netdev_phys_item_id *ppid) 1361 * Called to get the parent ID of the physical port of this device. 1362 * 1363 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1364 * struct net_device *dev) 1365 * Called by upper layer devices to accelerate switching or other 1366 * station functionality into hardware. 'pdev is the lowerdev 1367 * to use for the offload and 'dev' is the net device that will 1368 * back the offload. Returns a pointer to the private structure 1369 * the upper layer will maintain. 1370 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1371 * Called by upper layer device to delete the station created 1372 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1373 * the station and priv is the structure returned by the add 1374 * operation. 1375 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1376 * int queue_index, u32 maxrate); 1377 * Called when a user wants to set a max-rate limitation of specific 1378 * TX queue. 1379 * int (*ndo_get_iflink)(const struct net_device *dev); 1380 * Called to get the iflink value of this device. 1381 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1382 * This function is used to get egress tunnel information for given skb. 1383 * This is useful for retrieving outer tunnel header parameters while 1384 * sampling packet. 1385 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1386 * This function is used to specify the headroom that the skb must 1387 * consider when allocation skb during packet reception. Setting 1388 * appropriate rx headroom value allows avoiding skb head copy on 1389 * forward. Setting a negative value resets the rx headroom to the 1390 * default value. 1391 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1392 * This function is used to set or query state related to XDP on the 1393 * netdevice and manage BPF offload. See definition of 1394 * enum bpf_netdev_command for details. 1395 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1396 * u32 flags); 1397 * This function is used to submit @n XDP packets for transmit on a 1398 * netdevice. Returns number of frames successfully transmitted, frames 1399 * that got dropped are freed/returned via xdp_return_frame(). 1400 * Returns negative number, means general error invoking ndo, meaning 1401 * no frames were xmit'ed and core-caller will free all frames. 1402 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1403 * struct xdp_buff *xdp); 1404 * Get the xmit slave of master device based on the xdp_buff. 1405 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1406 * This function is used to wake up the softirq, ksoftirqd or kthread 1407 * responsible for sending and/or receiving packets on a specific 1408 * queue id bound to an AF_XDP socket. The flags field specifies if 1409 * only RX, only Tx, or both should be woken up using the flags 1410 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1411 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1412 * int cmd); 1413 * Add, change, delete or get information on an IPv4 tunnel. 1414 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1415 * If a device is paired with a peer device, return the peer instance. 1416 * The caller must be under RCU read context. 1417 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1418 * Get the forwarding path to reach the real device from the HW destination address 1419 * ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1420 * const struct skb_shared_hwtstamps *hwtstamps, 1421 * bool cycles); 1422 * Get hardware timestamp based on normal/adjustable time or free running 1423 * cycle counter. This function is required if physical clock supports a 1424 * free running cycle counter. 1425 * 1426 * int (*ndo_hwtstamp_get)(struct net_device *dev, 1427 * struct kernel_hwtstamp_config *kernel_config); 1428 * Get the currently configured hardware timestamping parameters for the 1429 * NIC device. 1430 * 1431 * int (*ndo_hwtstamp_set)(struct net_device *dev, 1432 * struct kernel_hwtstamp_config *kernel_config, 1433 * struct netlink_ext_ack *extack); 1434 * Change the hardware timestamping parameters for NIC device. 1435 */ 1436 struct net_device_ops { 1437 int (*ndo_init)(struct net_device *dev); 1438 void (*ndo_uninit)(struct net_device *dev); 1439 int (*ndo_open)(struct net_device *dev); 1440 int (*ndo_stop)(struct net_device *dev); 1441 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1442 struct net_device *dev); 1443 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1444 struct net_device *dev, 1445 netdev_features_t features); 1446 u16 (*ndo_select_queue)(struct net_device *dev, 1447 struct sk_buff *skb, 1448 struct net_device *sb_dev); 1449 void (*ndo_change_rx_flags)(struct net_device *dev, 1450 int flags); 1451 void (*ndo_set_rx_mode)(struct net_device *dev); 1452 int (*ndo_set_mac_address)(struct net_device *dev, 1453 void *addr); 1454 int (*ndo_validate_addr)(struct net_device *dev); 1455 int (*ndo_do_ioctl)(struct net_device *dev, 1456 struct ifreq *ifr, int cmd); 1457 int (*ndo_eth_ioctl)(struct net_device *dev, 1458 struct ifreq *ifr, int cmd); 1459 int (*ndo_siocbond)(struct net_device *dev, 1460 struct ifreq *ifr, int cmd); 1461 int (*ndo_siocwandev)(struct net_device *dev, 1462 struct if_settings *ifs); 1463 int (*ndo_siocdevprivate)(struct net_device *dev, 1464 struct ifreq *ifr, 1465 void __user *data, int cmd); 1466 int (*ndo_set_config)(struct net_device *dev, 1467 struct ifmap *map); 1468 int (*ndo_change_mtu)(struct net_device *dev, 1469 int new_mtu); 1470 int (*ndo_neigh_setup)(struct net_device *dev, 1471 struct neigh_parms *); 1472 void (*ndo_tx_timeout) (struct net_device *dev, 1473 unsigned int txqueue); 1474 1475 void (*ndo_get_stats64)(struct net_device *dev, 1476 struct rtnl_link_stats64 *storage); 1477 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1478 int (*ndo_get_offload_stats)(int attr_id, 1479 const struct net_device *dev, 1480 void *attr_data); 1481 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1482 1483 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1484 __be16 proto, u16 vid); 1485 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1486 __be16 proto, u16 vid); 1487 #ifdef CONFIG_NET_POLL_CONTROLLER 1488 void (*ndo_poll_controller)(struct net_device *dev); 1489 int (*ndo_netpoll_setup)(struct net_device *dev, 1490 struct netpoll_info *info); 1491 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1492 #endif 1493 int (*ndo_set_vf_mac)(struct net_device *dev, 1494 int queue, u8 *mac); 1495 int (*ndo_set_vf_vlan)(struct net_device *dev, 1496 int queue, u16 vlan, 1497 u8 qos, __be16 proto); 1498 int (*ndo_set_vf_rate)(struct net_device *dev, 1499 int vf, int min_tx_rate, 1500 int max_tx_rate); 1501 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1502 int vf, bool setting); 1503 int (*ndo_set_vf_trust)(struct net_device *dev, 1504 int vf, bool setting); 1505 int (*ndo_get_vf_config)(struct net_device *dev, 1506 int vf, 1507 struct ifla_vf_info *ivf); 1508 int (*ndo_set_vf_link_state)(struct net_device *dev, 1509 int vf, int link_state); 1510 int (*ndo_get_vf_stats)(struct net_device *dev, 1511 int vf, 1512 struct ifla_vf_stats 1513 *vf_stats); 1514 int (*ndo_set_vf_port)(struct net_device *dev, 1515 int vf, 1516 struct nlattr *port[]); 1517 int (*ndo_get_vf_port)(struct net_device *dev, 1518 int vf, struct sk_buff *skb); 1519 int (*ndo_get_vf_guid)(struct net_device *dev, 1520 int vf, 1521 struct ifla_vf_guid *node_guid, 1522 struct ifla_vf_guid *port_guid); 1523 int (*ndo_set_vf_guid)(struct net_device *dev, 1524 int vf, u64 guid, 1525 int guid_type); 1526 int (*ndo_set_vf_rss_query_en)( 1527 struct net_device *dev, 1528 int vf, bool setting); 1529 int (*ndo_setup_tc)(struct net_device *dev, 1530 enum tc_setup_type type, 1531 void *type_data); 1532 #if IS_ENABLED(CONFIG_FCOE) 1533 int (*ndo_fcoe_enable)(struct net_device *dev); 1534 int (*ndo_fcoe_disable)(struct net_device *dev); 1535 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1536 u16 xid, 1537 struct scatterlist *sgl, 1538 unsigned int sgc); 1539 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1540 u16 xid); 1541 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1542 u16 xid, 1543 struct scatterlist *sgl, 1544 unsigned int sgc); 1545 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1546 struct netdev_fcoe_hbainfo *hbainfo); 1547 #endif 1548 1549 #if IS_ENABLED(CONFIG_LIBFCOE) 1550 #define NETDEV_FCOE_WWNN 0 1551 #define NETDEV_FCOE_WWPN 1 1552 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1553 u64 *wwn, int type); 1554 #endif 1555 1556 #ifdef CONFIG_RFS_ACCEL 1557 int (*ndo_rx_flow_steer)(struct net_device *dev, 1558 const struct sk_buff *skb, 1559 u16 rxq_index, 1560 u32 flow_id); 1561 #endif 1562 int (*ndo_add_slave)(struct net_device *dev, 1563 struct net_device *slave_dev, 1564 struct netlink_ext_ack *extack); 1565 int (*ndo_del_slave)(struct net_device *dev, 1566 struct net_device *slave_dev); 1567 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1568 struct sk_buff *skb, 1569 bool all_slaves); 1570 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1571 struct sock *sk); 1572 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1573 netdev_features_t features); 1574 int (*ndo_set_features)(struct net_device *dev, 1575 netdev_features_t features); 1576 int (*ndo_neigh_construct)(struct net_device *dev, 1577 struct neighbour *n); 1578 void (*ndo_neigh_destroy)(struct net_device *dev, 1579 struct neighbour *n); 1580 1581 int (*ndo_fdb_add)(struct ndmsg *ndm, 1582 struct nlattr *tb[], 1583 struct net_device *dev, 1584 const unsigned char *addr, 1585 u16 vid, 1586 u16 flags, 1587 struct netlink_ext_ack *extack); 1588 int (*ndo_fdb_del)(struct ndmsg *ndm, 1589 struct nlattr *tb[], 1590 struct net_device *dev, 1591 const unsigned char *addr, 1592 u16 vid, struct netlink_ext_ack *extack); 1593 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, 1594 struct net_device *dev, 1595 struct netlink_ext_ack *extack); 1596 int (*ndo_fdb_dump)(struct sk_buff *skb, 1597 struct netlink_callback *cb, 1598 struct net_device *dev, 1599 struct net_device *filter_dev, 1600 int *idx); 1601 int (*ndo_fdb_get)(struct sk_buff *skb, 1602 struct nlattr *tb[], 1603 struct net_device *dev, 1604 const unsigned char *addr, 1605 u16 vid, u32 portid, u32 seq, 1606 struct netlink_ext_ack *extack); 1607 int (*ndo_mdb_add)(struct net_device *dev, 1608 struct nlattr *tb[], 1609 u16 nlmsg_flags, 1610 struct netlink_ext_ack *extack); 1611 int (*ndo_mdb_del)(struct net_device *dev, 1612 struct nlattr *tb[], 1613 struct netlink_ext_ack *extack); 1614 int (*ndo_mdb_dump)(struct net_device *dev, 1615 struct sk_buff *skb, 1616 struct netlink_callback *cb); 1617 int (*ndo_mdb_get)(struct net_device *dev, 1618 struct nlattr *tb[], u32 portid, 1619 u32 seq, 1620 struct netlink_ext_ack *extack); 1621 int (*ndo_bridge_setlink)(struct net_device *dev, 1622 struct nlmsghdr *nlh, 1623 u16 flags, 1624 struct netlink_ext_ack *extack); 1625 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1626 u32 pid, u32 seq, 1627 struct net_device *dev, 1628 u32 filter_mask, 1629 int nlflags); 1630 int (*ndo_bridge_dellink)(struct net_device *dev, 1631 struct nlmsghdr *nlh, 1632 u16 flags); 1633 int (*ndo_change_carrier)(struct net_device *dev, 1634 bool new_carrier); 1635 int (*ndo_get_phys_port_id)(struct net_device *dev, 1636 struct netdev_phys_item_id *ppid); 1637 int (*ndo_get_port_parent_id)(struct net_device *dev, 1638 struct netdev_phys_item_id *ppid); 1639 int (*ndo_get_phys_port_name)(struct net_device *dev, 1640 char *name, size_t len); 1641 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1642 struct net_device *dev); 1643 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1644 void *priv); 1645 1646 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1647 int queue_index, 1648 u32 maxrate); 1649 int (*ndo_get_iflink)(const struct net_device *dev); 1650 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1651 struct sk_buff *skb); 1652 void (*ndo_set_rx_headroom)(struct net_device *dev, 1653 int needed_headroom); 1654 int (*ndo_bpf)(struct net_device *dev, 1655 struct netdev_bpf *bpf); 1656 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1657 struct xdp_frame **xdp, 1658 u32 flags); 1659 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1660 struct xdp_buff *xdp); 1661 int (*ndo_xsk_wakeup)(struct net_device *dev, 1662 u32 queue_id, u32 flags); 1663 int (*ndo_tunnel_ctl)(struct net_device *dev, 1664 struct ip_tunnel_parm *p, int cmd); 1665 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1666 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1667 struct net_device_path *path); 1668 ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1669 const struct skb_shared_hwtstamps *hwtstamps, 1670 bool cycles); 1671 int (*ndo_hwtstamp_get)(struct net_device *dev, 1672 struct kernel_hwtstamp_config *kernel_config); 1673 int (*ndo_hwtstamp_set)(struct net_device *dev, 1674 struct kernel_hwtstamp_config *kernel_config, 1675 struct netlink_ext_ack *extack); 1676 }; 1677 1678 /** 1679 * enum netdev_priv_flags - &struct net_device priv_flags 1680 * 1681 * These are the &struct net_device, they are only set internally 1682 * by drivers and used in the kernel. These flags are invisible to 1683 * userspace; this means that the order of these flags can change 1684 * during any kernel release. 1685 * 1686 * You should have a pretty good reason to be extending these flags. 1687 * 1688 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1689 * @IFF_EBRIDGE: Ethernet bridging device 1690 * @IFF_BONDING: bonding master or slave 1691 * @IFF_ISATAP: ISATAP interface (RFC4214) 1692 * @IFF_WAN_HDLC: WAN HDLC device 1693 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1694 * release skb->dst 1695 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1696 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1697 * @IFF_MACVLAN_PORT: device used as macvlan port 1698 * @IFF_BRIDGE_PORT: device used as bridge port 1699 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1700 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1701 * @IFF_UNICAST_FLT: Supports unicast filtering 1702 * @IFF_TEAM_PORT: device used as team port 1703 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1704 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1705 * change when it's running 1706 * @IFF_MACVLAN: Macvlan device 1707 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1708 * underlying stacked devices 1709 * @IFF_L3MDEV_MASTER: device is an L3 master device 1710 * @IFF_NO_QUEUE: device can run without qdisc attached 1711 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1712 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1713 * @IFF_TEAM: device is a team device 1714 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1715 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1716 * entity (i.e. the master device for bridged veth) 1717 * @IFF_MACSEC: device is a MACsec device 1718 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1719 * @IFF_FAILOVER: device is a failover master device 1720 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1721 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1722 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf 1723 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1724 * skb_headlen(skb) == 0 (data starts from frag0) 1725 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN 1726 * @IFF_SEE_ALL_HWTSTAMP_REQUESTS: device wants to see calls to 1727 * ndo_hwtstamp_set() for all timestamp requests regardless of source, 1728 * even if those aren't HWTSTAMP_SOURCE_NETDEV. 1729 */ 1730 enum netdev_priv_flags { 1731 IFF_802_1Q_VLAN = 1<<0, 1732 IFF_EBRIDGE = 1<<1, 1733 IFF_BONDING = 1<<2, 1734 IFF_ISATAP = 1<<3, 1735 IFF_WAN_HDLC = 1<<4, 1736 IFF_XMIT_DST_RELEASE = 1<<5, 1737 IFF_DONT_BRIDGE = 1<<6, 1738 IFF_DISABLE_NETPOLL = 1<<7, 1739 IFF_MACVLAN_PORT = 1<<8, 1740 IFF_BRIDGE_PORT = 1<<9, 1741 IFF_OVS_DATAPATH = 1<<10, 1742 IFF_TX_SKB_SHARING = 1<<11, 1743 IFF_UNICAST_FLT = 1<<12, 1744 IFF_TEAM_PORT = 1<<13, 1745 IFF_SUPP_NOFCS = 1<<14, 1746 IFF_LIVE_ADDR_CHANGE = 1<<15, 1747 IFF_MACVLAN = 1<<16, 1748 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1749 IFF_L3MDEV_MASTER = 1<<18, 1750 IFF_NO_QUEUE = 1<<19, 1751 IFF_OPENVSWITCH = 1<<20, 1752 IFF_L3MDEV_SLAVE = 1<<21, 1753 IFF_TEAM = 1<<22, 1754 IFF_RXFH_CONFIGURED = 1<<23, 1755 IFF_PHONY_HEADROOM = 1<<24, 1756 IFF_MACSEC = 1<<25, 1757 IFF_NO_RX_HANDLER = 1<<26, 1758 IFF_FAILOVER = 1<<27, 1759 IFF_FAILOVER_SLAVE = 1<<28, 1760 IFF_L3MDEV_RX_HANDLER = 1<<29, 1761 IFF_NO_ADDRCONF = BIT_ULL(30), 1762 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31), 1763 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32), 1764 IFF_SEE_ALL_HWTSTAMP_REQUESTS = BIT_ULL(33), 1765 }; 1766 1767 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1768 #define IFF_EBRIDGE IFF_EBRIDGE 1769 #define IFF_BONDING IFF_BONDING 1770 #define IFF_ISATAP IFF_ISATAP 1771 #define IFF_WAN_HDLC IFF_WAN_HDLC 1772 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1773 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1774 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1775 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1776 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1777 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1778 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1779 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1780 #define IFF_TEAM_PORT IFF_TEAM_PORT 1781 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1782 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1783 #define IFF_MACVLAN IFF_MACVLAN 1784 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1785 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1786 #define IFF_NO_QUEUE IFF_NO_QUEUE 1787 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1788 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1789 #define IFF_TEAM IFF_TEAM 1790 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1791 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM 1792 #define IFF_MACSEC IFF_MACSEC 1793 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1794 #define IFF_FAILOVER IFF_FAILOVER 1795 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1796 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1797 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR 1798 1799 /* Specifies the type of the struct net_device::ml_priv pointer */ 1800 enum netdev_ml_priv_type { 1801 ML_PRIV_NONE, 1802 ML_PRIV_CAN, 1803 }; 1804 1805 enum netdev_stat_type { 1806 NETDEV_PCPU_STAT_NONE, 1807 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */ 1808 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */ 1809 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */ 1810 }; 1811 1812 /** 1813 * struct net_device - The DEVICE structure. 1814 * 1815 * Actually, this whole structure is a big mistake. It mixes I/O 1816 * data with strictly "high-level" data, and it has to know about 1817 * almost every data structure used in the INET module. 1818 * 1819 * @name: This is the first field of the "visible" part of this structure 1820 * (i.e. as seen by users in the "Space.c" file). It is the name 1821 * of the interface. 1822 * 1823 * @name_node: Name hashlist node 1824 * @ifalias: SNMP alias 1825 * @mem_end: Shared memory end 1826 * @mem_start: Shared memory start 1827 * @base_addr: Device I/O address 1828 * @irq: Device IRQ number 1829 * 1830 * @state: Generic network queuing layer state, see netdev_state_t 1831 * @dev_list: The global list of network devices 1832 * @napi_list: List entry used for polling NAPI devices 1833 * @unreg_list: List entry when we are unregistering the 1834 * device; see the function unregister_netdev 1835 * @close_list: List entry used when we are closing the device 1836 * @ptype_all: Device-specific packet handlers for all protocols 1837 * @ptype_specific: Device-specific, protocol-specific packet handlers 1838 * 1839 * @adj_list: Directly linked devices, like slaves for bonding 1840 * @features: Currently active device features 1841 * @hw_features: User-changeable features 1842 * 1843 * @wanted_features: User-requested features 1844 * @vlan_features: Mask of features inheritable by VLAN devices 1845 * 1846 * @hw_enc_features: Mask of features inherited by encapsulating devices 1847 * This field indicates what encapsulation 1848 * offloads the hardware is capable of doing, 1849 * and drivers will need to set them appropriately. 1850 * 1851 * @mpls_features: Mask of features inheritable by MPLS 1852 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1853 * 1854 * @ifindex: interface index 1855 * @group: The group the device belongs to 1856 * 1857 * @stats: Statistics struct, which was left as a legacy, use 1858 * rtnl_link_stats64 instead 1859 * 1860 * @core_stats: core networking counters, 1861 * do not use this in drivers 1862 * @carrier_up_count: Number of times the carrier has been up 1863 * @carrier_down_count: Number of times the carrier has been down 1864 * 1865 * @wireless_handlers: List of functions to handle Wireless Extensions, 1866 * instead of ioctl, 1867 * see <net/iw_handler.h> for details. 1868 * @wireless_data: Instance data managed by the core of wireless extensions 1869 * 1870 * @netdev_ops: Includes several pointers to callbacks, 1871 * if one wants to override the ndo_*() functions 1872 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks. 1873 * @xsk_tx_metadata_ops: Includes pointers to AF_XDP TX metadata callbacks. 1874 * @ethtool_ops: Management operations 1875 * @l3mdev_ops: Layer 3 master device operations 1876 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1877 * discovery handling. Necessary for e.g. 6LoWPAN. 1878 * @xfrmdev_ops: Transformation offload operations 1879 * @tlsdev_ops: Transport Layer Security offload operations 1880 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1881 * of Layer 2 headers. 1882 * 1883 * @flags: Interface flags (a la BSD) 1884 * @xdp_features: XDP capability supported by the device 1885 * @priv_flags: Like 'flags' but invisible to userspace, 1886 * see if.h for the definitions 1887 * @gflags: Global flags ( kept as legacy ) 1888 * @padded: How much padding added by alloc_netdev() 1889 * @operstate: RFC2863 operstate 1890 * @link_mode: Mapping policy to operstate 1891 * @if_port: Selectable AUI, TP, ... 1892 * @dma: DMA channel 1893 * @mtu: Interface MTU value 1894 * @min_mtu: Interface Minimum MTU value 1895 * @max_mtu: Interface Maximum MTU value 1896 * @type: Interface hardware type 1897 * @hard_header_len: Maximum hardware header length. 1898 * @min_header_len: Minimum hardware header length 1899 * 1900 * @needed_headroom: Extra headroom the hardware may need, but not in all 1901 * cases can this be guaranteed 1902 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1903 * cases can this be guaranteed. Some cases also use 1904 * LL_MAX_HEADER instead to allocate the skb 1905 * 1906 * interface address info: 1907 * 1908 * @perm_addr: Permanent hw address 1909 * @addr_assign_type: Hw address assignment type 1910 * @addr_len: Hardware address length 1911 * @upper_level: Maximum depth level of upper devices. 1912 * @lower_level: Maximum depth level of lower devices. 1913 * @neigh_priv_len: Used in neigh_alloc() 1914 * @dev_id: Used to differentiate devices that share 1915 * the same link layer address 1916 * @dev_port: Used to differentiate devices that share 1917 * the same function 1918 * @addr_list_lock: XXX: need comments on this one 1919 * @name_assign_type: network interface name assignment type 1920 * @uc_promisc: Counter that indicates promiscuous mode 1921 * has been enabled due to the need to listen to 1922 * additional unicast addresses in a device that 1923 * does not implement ndo_set_rx_mode() 1924 * @uc: unicast mac addresses 1925 * @mc: multicast mac addresses 1926 * @dev_addrs: list of device hw addresses 1927 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1928 * @promiscuity: Number of times the NIC is told to work in 1929 * promiscuous mode; if it becomes 0 the NIC will 1930 * exit promiscuous mode 1931 * @allmulti: Counter, enables or disables allmulticast mode 1932 * 1933 * @vlan_info: VLAN info 1934 * @dsa_ptr: dsa specific data 1935 * @tipc_ptr: TIPC specific data 1936 * @atalk_ptr: AppleTalk link 1937 * @ip_ptr: IPv4 specific data 1938 * @ip6_ptr: IPv6 specific data 1939 * @ax25_ptr: AX.25 specific data 1940 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1941 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1942 * device struct 1943 * @mpls_ptr: mpls_dev struct pointer 1944 * @mctp_ptr: MCTP specific data 1945 * 1946 * @dev_addr: Hw address (before bcast, 1947 * because most packets are unicast) 1948 * 1949 * @_rx: Array of RX queues 1950 * @num_rx_queues: Number of RX queues 1951 * allocated at register_netdev() time 1952 * @real_num_rx_queues: Number of RX queues currently active in device 1953 * @xdp_prog: XDP sockets filter program pointer 1954 * @gro_flush_timeout: timeout for GRO layer in NAPI 1955 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1956 * allow to avoid NIC hard IRQ, on busy queues. 1957 * 1958 * @rx_handler: handler for received packets 1959 * @rx_handler_data: XXX: need comments on this one 1960 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing 1961 * @ingress_queue: XXX: need comments on this one 1962 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1963 * @broadcast: hw bcast address 1964 * 1965 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1966 * indexed by RX queue number. Assigned by driver. 1967 * This must only be set if the ndo_rx_flow_steer 1968 * operation is defined 1969 * @index_hlist: Device index hash chain 1970 * 1971 * @_tx: Array of TX queues 1972 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1973 * @real_num_tx_queues: Number of TX queues currently active in device 1974 * @qdisc: Root qdisc from userspace point of view 1975 * @tx_queue_len: Max frames per queue allowed 1976 * @tx_global_lock: XXX: need comments on this one 1977 * @xdp_bulkq: XDP device bulk queue 1978 * @xps_maps: all CPUs/RXQs maps for XPS device 1979 * 1980 * @xps_maps: XXX: need comments on this one 1981 * @tcx_egress: BPF & clsact qdisc specific data for egress processing 1982 * @nf_hooks_egress: netfilter hooks executed for egress packets 1983 * @qdisc_hash: qdisc hash table 1984 * @watchdog_timeo: Represents the timeout that is used by 1985 * the watchdog (see dev_watchdog()) 1986 * @watchdog_timer: List of timers 1987 * 1988 * @proto_down_reason: reason a netdev interface is held down 1989 * @pcpu_refcnt: Number of references to this device 1990 * @dev_refcnt: Number of references to this device 1991 * @refcnt_tracker: Tracker directory for tracked references to this device 1992 * @todo_list: Delayed register/unregister 1993 * @link_watch_list: XXX: need comments on this one 1994 * 1995 * @reg_state: Register/unregister state machine 1996 * @dismantle: Device is going to be freed 1997 * @rtnl_link_state: This enum represents the phases of creating 1998 * a new link 1999 * 2000 * @needs_free_netdev: Should unregister perform free_netdev? 2001 * @priv_destructor: Called from unregister 2002 * @npinfo: XXX: need comments on this one 2003 * @nd_net: Network namespace this network device is inside 2004 * 2005 * @ml_priv: Mid-layer private 2006 * @ml_priv_type: Mid-layer private type 2007 * 2008 * @pcpu_stat_type: Type of device statistics which the core should 2009 * allocate/free: none, lstats, tstats, dstats. none 2010 * means the driver is handling statistics allocation/ 2011 * freeing internally. 2012 * @lstats: Loopback statistics: packets, bytes 2013 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes 2014 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes 2015 * 2016 * @garp_port: GARP 2017 * @mrp_port: MRP 2018 * 2019 * @dm_private: Drop monitor private 2020 * 2021 * @dev: Class/net/name entry 2022 * @sysfs_groups: Space for optional device, statistics and wireless 2023 * sysfs groups 2024 * 2025 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 2026 * @rtnl_link_ops: Rtnl_link_ops 2027 * 2028 * @gso_max_size: Maximum size of generic segmentation offload 2029 * @tso_max_size: Device (as in HW) limit on the max TSO request size 2030 * @gso_max_segs: Maximum number of segments that can be passed to the 2031 * NIC for GSO 2032 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count 2033 * @gso_ipv4_max_size: Maximum size of generic segmentation offload, 2034 * for IPv4. 2035 * 2036 * @dcbnl_ops: Data Center Bridging netlink ops 2037 * @num_tc: Number of traffic classes in the net device 2038 * @tc_to_txq: XXX: need comments on this one 2039 * @prio_tc_map: XXX: need comments on this one 2040 * 2041 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 2042 * 2043 * @priomap: XXX: need comments on this one 2044 * @phydev: Physical device may attach itself 2045 * for hardware timestamping 2046 * @sfp_bus: attached &struct sfp_bus structure. 2047 * 2048 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 2049 * 2050 * @proto_down: protocol port state information can be sent to the 2051 * switch driver and used to set the phys state of the 2052 * switch port. 2053 * 2054 * @wol_enabled: Wake-on-LAN is enabled 2055 * 2056 * @threaded: napi threaded mode is enabled 2057 * 2058 * @net_notifier_list: List of per-net netdev notifier block 2059 * that follow this device when it is moved 2060 * to another network namespace. 2061 * 2062 * @macsec_ops: MACsec offloading ops 2063 * 2064 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 2065 * offload capabilities of the device 2066 * @udp_tunnel_nic: UDP tunnel offload state 2067 * @xdp_state: stores info on attached XDP BPF programs 2068 * 2069 * @nested_level: Used as a parameter of spin_lock_nested() of 2070 * dev->addr_list_lock. 2071 * @unlink_list: As netif_addr_lock() can be called recursively, 2072 * keep a list of interfaces to be deleted. 2073 * @gro_max_size: Maximum size of aggregated packet in generic 2074 * receive offload (GRO) 2075 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic 2076 * receive offload (GRO), for IPv4. 2077 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP 2078 * zero copy driver 2079 * 2080 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 2081 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 2082 * @watchdog_dev_tracker: refcount tracker used by watchdog. 2083 * @dev_registered_tracker: tracker for reference held while 2084 * registered 2085 * @offload_xstats_l3: L3 HW stats for this netdevice. 2086 * 2087 * @devlink_port: Pointer to related devlink port structure. 2088 * Assigned by a driver before netdev registration using 2089 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static 2090 * during the time netdevice is registered. 2091 * 2092 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem, 2093 * where the clock is recovered. 2094 * 2095 * FIXME: cleanup struct net_device such that network protocol info 2096 * moves out. 2097 */ 2098 2099 struct net_device { 2100 char name[IFNAMSIZ]; 2101 struct netdev_name_node *name_node; 2102 struct dev_ifalias __rcu *ifalias; 2103 /* 2104 * I/O specific fields 2105 * FIXME: Merge these and struct ifmap into one 2106 */ 2107 unsigned long mem_end; 2108 unsigned long mem_start; 2109 unsigned long base_addr; 2110 2111 /* 2112 * Some hardware also needs these fields (state,dev_list, 2113 * napi_list,unreg_list,close_list) but they are not 2114 * part of the usual set specified in Space.c. 2115 */ 2116 2117 unsigned long state; 2118 2119 struct list_head dev_list; 2120 struct list_head napi_list; 2121 struct list_head unreg_list; 2122 struct list_head close_list; 2123 struct list_head ptype_all; 2124 struct list_head ptype_specific; 2125 2126 struct { 2127 struct list_head upper; 2128 struct list_head lower; 2129 } adj_list; 2130 2131 /* Read-mostly cache-line for fast-path access */ 2132 unsigned int flags; 2133 xdp_features_t xdp_features; 2134 unsigned long long priv_flags; 2135 const struct net_device_ops *netdev_ops; 2136 const struct xdp_metadata_ops *xdp_metadata_ops; 2137 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops; 2138 int ifindex; 2139 unsigned short gflags; 2140 unsigned short hard_header_len; 2141 2142 /* Note : dev->mtu is often read without holding a lock. 2143 * Writers usually hold RTNL. 2144 * It is recommended to use READ_ONCE() to annotate the reads, 2145 * and to use WRITE_ONCE() to annotate the writes. 2146 */ 2147 unsigned int mtu; 2148 unsigned short needed_headroom; 2149 unsigned short needed_tailroom; 2150 2151 netdev_features_t features; 2152 netdev_features_t hw_features; 2153 netdev_features_t wanted_features; 2154 netdev_features_t vlan_features; 2155 netdev_features_t hw_enc_features; 2156 netdev_features_t mpls_features; 2157 netdev_features_t gso_partial_features; 2158 2159 unsigned int min_mtu; 2160 unsigned int max_mtu; 2161 unsigned short type; 2162 unsigned char min_header_len; 2163 unsigned char name_assign_type; 2164 2165 int group; 2166 2167 struct net_device_stats stats; /* not used by modern drivers */ 2168 2169 struct net_device_core_stats __percpu *core_stats; 2170 2171 /* Stats to monitor link on/off, flapping */ 2172 atomic_t carrier_up_count; 2173 atomic_t carrier_down_count; 2174 2175 #ifdef CONFIG_WIRELESS_EXT 2176 const struct iw_handler_def *wireless_handlers; 2177 struct iw_public_data *wireless_data; 2178 #endif 2179 const struct ethtool_ops *ethtool_ops; 2180 #ifdef CONFIG_NET_L3_MASTER_DEV 2181 const struct l3mdev_ops *l3mdev_ops; 2182 #endif 2183 #if IS_ENABLED(CONFIG_IPV6) 2184 const struct ndisc_ops *ndisc_ops; 2185 #endif 2186 2187 #ifdef CONFIG_XFRM_OFFLOAD 2188 const struct xfrmdev_ops *xfrmdev_ops; 2189 #endif 2190 2191 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2192 const struct tlsdev_ops *tlsdev_ops; 2193 #endif 2194 2195 const struct header_ops *header_ops; 2196 2197 unsigned char operstate; 2198 unsigned char link_mode; 2199 2200 unsigned char if_port; 2201 unsigned char dma; 2202 2203 /* Interface address info. */ 2204 unsigned char perm_addr[MAX_ADDR_LEN]; 2205 unsigned char addr_assign_type; 2206 unsigned char addr_len; 2207 unsigned char upper_level; 2208 unsigned char lower_level; 2209 2210 unsigned short neigh_priv_len; 2211 unsigned short dev_id; 2212 unsigned short dev_port; 2213 unsigned short padded; 2214 2215 spinlock_t addr_list_lock; 2216 int irq; 2217 2218 struct netdev_hw_addr_list uc; 2219 struct netdev_hw_addr_list mc; 2220 struct netdev_hw_addr_list dev_addrs; 2221 2222 #ifdef CONFIG_SYSFS 2223 struct kset *queues_kset; 2224 #endif 2225 #ifdef CONFIG_LOCKDEP 2226 struct list_head unlink_list; 2227 #endif 2228 unsigned int promiscuity; 2229 unsigned int allmulti; 2230 bool uc_promisc; 2231 #ifdef CONFIG_LOCKDEP 2232 unsigned char nested_level; 2233 #endif 2234 2235 2236 /* Protocol-specific pointers */ 2237 2238 struct in_device __rcu *ip_ptr; 2239 struct inet6_dev __rcu *ip6_ptr; 2240 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2241 struct vlan_info __rcu *vlan_info; 2242 #endif 2243 #if IS_ENABLED(CONFIG_NET_DSA) 2244 struct dsa_port *dsa_ptr; 2245 #endif 2246 #if IS_ENABLED(CONFIG_TIPC) 2247 struct tipc_bearer __rcu *tipc_ptr; 2248 #endif 2249 #if IS_ENABLED(CONFIG_ATALK) 2250 void *atalk_ptr; 2251 #endif 2252 #if IS_ENABLED(CONFIG_AX25) 2253 void *ax25_ptr; 2254 #endif 2255 #if IS_ENABLED(CONFIG_CFG80211) 2256 struct wireless_dev *ieee80211_ptr; 2257 #endif 2258 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2259 struct wpan_dev *ieee802154_ptr; 2260 #endif 2261 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2262 struct mpls_dev __rcu *mpls_ptr; 2263 #endif 2264 #if IS_ENABLED(CONFIG_MCTP) 2265 struct mctp_dev __rcu *mctp_ptr; 2266 #endif 2267 2268 /* 2269 * Cache lines mostly used on receive path (including eth_type_trans()) 2270 */ 2271 /* Interface address info used in eth_type_trans() */ 2272 const unsigned char *dev_addr; 2273 2274 struct netdev_rx_queue *_rx; 2275 unsigned int num_rx_queues; 2276 unsigned int real_num_rx_queues; 2277 2278 struct bpf_prog __rcu *xdp_prog; 2279 unsigned long gro_flush_timeout; 2280 int napi_defer_hard_irqs; 2281 #define GRO_LEGACY_MAX_SIZE 65536u 2282 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2283 * and shinfo->gso_segs is a 16bit field. 2284 */ 2285 #define GRO_MAX_SIZE (8 * 65535u) 2286 unsigned int gro_max_size; 2287 unsigned int gro_ipv4_max_size; 2288 unsigned int xdp_zc_max_segs; 2289 rx_handler_func_t __rcu *rx_handler; 2290 void __rcu *rx_handler_data; 2291 #ifdef CONFIG_NET_XGRESS 2292 struct bpf_mprog_entry __rcu *tcx_ingress; 2293 #endif 2294 struct netdev_queue __rcu *ingress_queue; 2295 #ifdef CONFIG_NETFILTER_INGRESS 2296 struct nf_hook_entries __rcu *nf_hooks_ingress; 2297 #endif 2298 2299 unsigned char broadcast[MAX_ADDR_LEN]; 2300 #ifdef CONFIG_RFS_ACCEL 2301 struct cpu_rmap *rx_cpu_rmap; 2302 #endif 2303 struct hlist_node index_hlist; 2304 2305 /* 2306 * Cache lines mostly used on transmit path 2307 */ 2308 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2309 unsigned int num_tx_queues; 2310 unsigned int real_num_tx_queues; 2311 struct Qdisc __rcu *qdisc; 2312 unsigned int tx_queue_len; 2313 spinlock_t tx_global_lock; 2314 2315 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2316 2317 #ifdef CONFIG_XPS 2318 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2319 #endif 2320 #ifdef CONFIG_NET_XGRESS 2321 struct bpf_mprog_entry __rcu *tcx_egress; 2322 #endif 2323 #ifdef CONFIG_NETFILTER_EGRESS 2324 struct nf_hook_entries __rcu *nf_hooks_egress; 2325 #endif 2326 2327 #ifdef CONFIG_NET_SCHED 2328 DECLARE_HASHTABLE (qdisc_hash, 4); 2329 #endif 2330 /* These may be needed for future network-power-down code. */ 2331 struct timer_list watchdog_timer; 2332 int watchdog_timeo; 2333 2334 u32 proto_down_reason; 2335 2336 struct list_head todo_list; 2337 2338 #ifdef CONFIG_PCPU_DEV_REFCNT 2339 int __percpu *pcpu_refcnt; 2340 #else 2341 refcount_t dev_refcnt; 2342 #endif 2343 struct ref_tracker_dir refcnt_tracker; 2344 2345 struct list_head link_watch_list; 2346 2347 enum { NETREG_UNINITIALIZED=0, 2348 NETREG_REGISTERED, /* completed register_netdevice */ 2349 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2350 NETREG_UNREGISTERED, /* completed unregister todo */ 2351 NETREG_RELEASED, /* called free_netdev */ 2352 NETREG_DUMMY, /* dummy device for NAPI poll */ 2353 } reg_state:8; 2354 2355 bool dismantle; 2356 2357 enum { 2358 RTNL_LINK_INITIALIZED, 2359 RTNL_LINK_INITIALIZING, 2360 } rtnl_link_state:16; 2361 2362 bool needs_free_netdev; 2363 void (*priv_destructor)(struct net_device *dev); 2364 2365 #ifdef CONFIG_NETPOLL 2366 struct netpoll_info __rcu *npinfo; 2367 #endif 2368 2369 possible_net_t nd_net; 2370 2371 /* mid-layer private */ 2372 void *ml_priv; 2373 enum netdev_ml_priv_type ml_priv_type; 2374 2375 enum netdev_stat_type pcpu_stat_type:8; 2376 union { 2377 struct pcpu_lstats __percpu *lstats; 2378 struct pcpu_sw_netstats __percpu *tstats; 2379 struct pcpu_dstats __percpu *dstats; 2380 }; 2381 2382 #if IS_ENABLED(CONFIG_GARP) 2383 struct garp_port __rcu *garp_port; 2384 #endif 2385 #if IS_ENABLED(CONFIG_MRP) 2386 struct mrp_port __rcu *mrp_port; 2387 #endif 2388 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2389 struct dm_hw_stat_delta __rcu *dm_private; 2390 #endif 2391 struct device dev; 2392 const struct attribute_group *sysfs_groups[4]; 2393 const struct attribute_group *sysfs_rx_queue_group; 2394 2395 const struct rtnl_link_ops *rtnl_link_ops; 2396 2397 /* for setting kernel sock attribute on TCP connection setup */ 2398 #define GSO_MAX_SEGS 65535u 2399 #define GSO_LEGACY_MAX_SIZE 65536u 2400 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2401 * and shinfo->gso_segs is a 16bit field. 2402 */ 2403 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2404 2405 unsigned int gso_max_size; 2406 #define TSO_LEGACY_MAX_SIZE 65536 2407 #define TSO_MAX_SIZE UINT_MAX 2408 unsigned int tso_max_size; 2409 u16 gso_max_segs; 2410 #define TSO_MAX_SEGS U16_MAX 2411 u16 tso_max_segs; 2412 unsigned int gso_ipv4_max_size; 2413 2414 #ifdef CONFIG_DCB 2415 const struct dcbnl_rtnl_ops *dcbnl_ops; 2416 #endif 2417 s16 num_tc; 2418 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2419 u8 prio_tc_map[TC_BITMASK + 1]; 2420 2421 #if IS_ENABLED(CONFIG_FCOE) 2422 unsigned int fcoe_ddp_xid; 2423 #endif 2424 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2425 struct netprio_map __rcu *priomap; 2426 #endif 2427 struct phy_device *phydev; 2428 struct sfp_bus *sfp_bus; 2429 struct lock_class_key *qdisc_tx_busylock; 2430 bool proto_down; 2431 unsigned wol_enabled:1; 2432 unsigned threaded:1; 2433 2434 struct list_head net_notifier_list; 2435 2436 #if IS_ENABLED(CONFIG_MACSEC) 2437 /* MACsec management functions */ 2438 const struct macsec_ops *macsec_ops; 2439 #endif 2440 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2441 struct udp_tunnel_nic *udp_tunnel_nic; 2442 2443 /* protected by rtnl_lock */ 2444 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2445 2446 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2447 netdevice_tracker linkwatch_dev_tracker; 2448 netdevice_tracker watchdog_dev_tracker; 2449 netdevice_tracker dev_registered_tracker; 2450 struct rtnl_hw_stats64 *offload_xstats_l3; 2451 2452 struct devlink_port *devlink_port; 2453 2454 #if IS_ENABLED(CONFIG_DPLL) 2455 struct dpll_pin *dpll_pin; 2456 #endif 2457 #if IS_ENABLED(CONFIG_PAGE_POOL) 2458 /** @page_pools: page pools created for this netdevice */ 2459 struct hlist_head page_pools; 2460 #endif 2461 }; 2462 #define to_net_dev(d) container_of(d, struct net_device, dev) 2463 2464 /* 2465 * Driver should use this to assign devlink port instance to a netdevice 2466 * before it registers the netdevice. Therefore devlink_port is static 2467 * during the netdev lifetime after it is registered. 2468 */ 2469 #define SET_NETDEV_DEVLINK_PORT(dev, port) \ 2470 ({ \ 2471 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \ 2472 ((dev)->devlink_port = (port)); \ 2473 }) 2474 2475 static inline bool netif_elide_gro(const struct net_device *dev) 2476 { 2477 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2478 return true; 2479 return false; 2480 } 2481 2482 #define NETDEV_ALIGN 32 2483 2484 static inline 2485 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2486 { 2487 return dev->prio_tc_map[prio & TC_BITMASK]; 2488 } 2489 2490 static inline 2491 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2492 { 2493 if (tc >= dev->num_tc) 2494 return -EINVAL; 2495 2496 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2497 return 0; 2498 } 2499 2500 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2501 void netdev_reset_tc(struct net_device *dev); 2502 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2503 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2504 2505 static inline 2506 int netdev_get_num_tc(struct net_device *dev) 2507 { 2508 return dev->num_tc; 2509 } 2510 2511 static inline void net_prefetch(void *p) 2512 { 2513 prefetch(p); 2514 #if L1_CACHE_BYTES < 128 2515 prefetch((u8 *)p + L1_CACHE_BYTES); 2516 #endif 2517 } 2518 2519 static inline void net_prefetchw(void *p) 2520 { 2521 prefetchw(p); 2522 #if L1_CACHE_BYTES < 128 2523 prefetchw((u8 *)p + L1_CACHE_BYTES); 2524 #endif 2525 } 2526 2527 void netdev_unbind_sb_channel(struct net_device *dev, 2528 struct net_device *sb_dev); 2529 int netdev_bind_sb_channel_queue(struct net_device *dev, 2530 struct net_device *sb_dev, 2531 u8 tc, u16 count, u16 offset); 2532 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2533 static inline int netdev_get_sb_channel(struct net_device *dev) 2534 { 2535 return max_t(int, -dev->num_tc, 0); 2536 } 2537 2538 static inline 2539 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2540 unsigned int index) 2541 { 2542 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues); 2543 return &dev->_tx[index]; 2544 } 2545 2546 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2547 const struct sk_buff *skb) 2548 { 2549 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2550 } 2551 2552 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2553 void (*f)(struct net_device *, 2554 struct netdev_queue *, 2555 void *), 2556 void *arg) 2557 { 2558 unsigned int i; 2559 2560 for (i = 0; i < dev->num_tx_queues; i++) 2561 f(dev, &dev->_tx[i], arg); 2562 } 2563 2564 #define netdev_lockdep_set_classes(dev) \ 2565 { \ 2566 static struct lock_class_key qdisc_tx_busylock_key; \ 2567 static struct lock_class_key qdisc_xmit_lock_key; \ 2568 static struct lock_class_key dev_addr_list_lock_key; \ 2569 unsigned int i; \ 2570 \ 2571 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2572 lockdep_set_class(&(dev)->addr_list_lock, \ 2573 &dev_addr_list_lock_key); \ 2574 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2575 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2576 &qdisc_xmit_lock_key); \ 2577 } 2578 2579 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2580 struct net_device *sb_dev); 2581 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2582 struct sk_buff *skb, 2583 struct net_device *sb_dev); 2584 2585 /* returns the headroom that the master device needs to take in account 2586 * when forwarding to this dev 2587 */ 2588 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2589 { 2590 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2591 } 2592 2593 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2594 { 2595 if (dev->netdev_ops->ndo_set_rx_headroom) 2596 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2597 } 2598 2599 /* set the device rx headroom to the dev's default */ 2600 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2601 { 2602 netdev_set_rx_headroom(dev, -1); 2603 } 2604 2605 static inline void *netdev_get_ml_priv(struct net_device *dev, 2606 enum netdev_ml_priv_type type) 2607 { 2608 if (dev->ml_priv_type != type) 2609 return NULL; 2610 2611 return dev->ml_priv; 2612 } 2613 2614 static inline void netdev_set_ml_priv(struct net_device *dev, 2615 void *ml_priv, 2616 enum netdev_ml_priv_type type) 2617 { 2618 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2619 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2620 dev->ml_priv_type, type); 2621 WARN(!dev->ml_priv_type && dev->ml_priv, 2622 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2623 2624 dev->ml_priv = ml_priv; 2625 dev->ml_priv_type = type; 2626 } 2627 2628 /* 2629 * Net namespace inlines 2630 */ 2631 static inline 2632 struct net *dev_net(const struct net_device *dev) 2633 { 2634 return read_pnet(&dev->nd_net); 2635 } 2636 2637 static inline 2638 void dev_net_set(struct net_device *dev, struct net *net) 2639 { 2640 write_pnet(&dev->nd_net, net); 2641 } 2642 2643 /** 2644 * netdev_priv - access network device private data 2645 * @dev: network device 2646 * 2647 * Get network device private data 2648 */ 2649 static inline void *netdev_priv(const struct net_device *dev) 2650 { 2651 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2652 } 2653 2654 /* Set the sysfs physical device reference for the network logical device 2655 * if set prior to registration will cause a symlink during initialization. 2656 */ 2657 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2658 2659 /* Set the sysfs device type for the network logical device to allow 2660 * fine-grained identification of different network device types. For 2661 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2662 */ 2663 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2664 2665 void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index, 2666 enum netdev_queue_type type, 2667 struct napi_struct *napi); 2668 2669 static inline void netif_napi_set_irq(struct napi_struct *napi, int irq) 2670 { 2671 napi->irq = irq; 2672 } 2673 2674 /* Default NAPI poll() weight 2675 * Device drivers are strongly advised to not use bigger value 2676 */ 2677 #define NAPI_POLL_WEIGHT 64 2678 2679 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2680 int (*poll)(struct napi_struct *, int), int weight); 2681 2682 /** 2683 * netif_napi_add() - initialize a NAPI context 2684 * @dev: network device 2685 * @napi: NAPI context 2686 * @poll: polling function 2687 * 2688 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2689 * *any* of the other NAPI-related functions. 2690 */ 2691 static inline void 2692 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2693 int (*poll)(struct napi_struct *, int)) 2694 { 2695 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2696 } 2697 2698 static inline void 2699 netif_napi_add_tx_weight(struct net_device *dev, 2700 struct napi_struct *napi, 2701 int (*poll)(struct napi_struct *, int), 2702 int weight) 2703 { 2704 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2705 netif_napi_add_weight(dev, napi, poll, weight); 2706 } 2707 2708 /** 2709 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2710 * @dev: network device 2711 * @napi: NAPI context 2712 * @poll: polling function 2713 * 2714 * This variant of netif_napi_add() should be used from drivers using NAPI 2715 * to exclusively poll a TX queue. 2716 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2717 */ 2718 static inline void netif_napi_add_tx(struct net_device *dev, 2719 struct napi_struct *napi, 2720 int (*poll)(struct napi_struct *, int)) 2721 { 2722 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2723 } 2724 2725 /** 2726 * __netif_napi_del - remove a NAPI context 2727 * @napi: NAPI context 2728 * 2729 * Warning: caller must observe RCU grace period before freeing memory 2730 * containing @napi. Drivers might want to call this helper to combine 2731 * all the needed RCU grace periods into a single one. 2732 */ 2733 void __netif_napi_del(struct napi_struct *napi); 2734 2735 /** 2736 * netif_napi_del - remove a NAPI context 2737 * @napi: NAPI context 2738 * 2739 * netif_napi_del() removes a NAPI context from the network device NAPI list 2740 */ 2741 static inline void netif_napi_del(struct napi_struct *napi) 2742 { 2743 __netif_napi_del(napi); 2744 synchronize_net(); 2745 } 2746 2747 struct packet_type { 2748 __be16 type; /* This is really htons(ether_type). */ 2749 bool ignore_outgoing; 2750 struct net_device *dev; /* NULL is wildcarded here */ 2751 netdevice_tracker dev_tracker; 2752 int (*func) (struct sk_buff *, 2753 struct net_device *, 2754 struct packet_type *, 2755 struct net_device *); 2756 void (*list_func) (struct list_head *, 2757 struct packet_type *, 2758 struct net_device *); 2759 bool (*id_match)(struct packet_type *ptype, 2760 struct sock *sk); 2761 struct net *af_packet_net; 2762 void *af_packet_priv; 2763 struct list_head list; 2764 }; 2765 2766 struct offload_callbacks { 2767 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2768 netdev_features_t features); 2769 struct sk_buff *(*gro_receive)(struct list_head *head, 2770 struct sk_buff *skb); 2771 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2772 }; 2773 2774 struct packet_offload { 2775 __be16 type; /* This is really htons(ether_type). */ 2776 u16 priority; 2777 struct offload_callbacks callbacks; 2778 struct list_head list; 2779 }; 2780 2781 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2782 struct pcpu_sw_netstats { 2783 u64_stats_t rx_packets; 2784 u64_stats_t rx_bytes; 2785 u64_stats_t tx_packets; 2786 u64_stats_t tx_bytes; 2787 struct u64_stats_sync syncp; 2788 } __aligned(4 * sizeof(u64)); 2789 2790 struct pcpu_dstats { 2791 u64 rx_packets; 2792 u64 rx_bytes; 2793 u64 rx_drops; 2794 u64 tx_packets; 2795 u64 tx_bytes; 2796 u64 tx_drops; 2797 struct u64_stats_sync syncp; 2798 } __aligned(8 * sizeof(u64)); 2799 2800 struct pcpu_lstats { 2801 u64_stats_t packets; 2802 u64_stats_t bytes; 2803 struct u64_stats_sync syncp; 2804 } __aligned(2 * sizeof(u64)); 2805 2806 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2807 2808 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2809 { 2810 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2811 2812 u64_stats_update_begin(&tstats->syncp); 2813 u64_stats_add(&tstats->rx_bytes, len); 2814 u64_stats_inc(&tstats->rx_packets); 2815 u64_stats_update_end(&tstats->syncp); 2816 } 2817 2818 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2819 unsigned int packets, 2820 unsigned int len) 2821 { 2822 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2823 2824 u64_stats_update_begin(&tstats->syncp); 2825 u64_stats_add(&tstats->tx_bytes, len); 2826 u64_stats_add(&tstats->tx_packets, packets); 2827 u64_stats_update_end(&tstats->syncp); 2828 } 2829 2830 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2831 { 2832 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2833 2834 u64_stats_update_begin(&lstats->syncp); 2835 u64_stats_add(&lstats->bytes, len); 2836 u64_stats_inc(&lstats->packets); 2837 u64_stats_update_end(&lstats->syncp); 2838 } 2839 2840 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2841 ({ \ 2842 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2843 if (pcpu_stats) { \ 2844 int __cpu; \ 2845 for_each_possible_cpu(__cpu) { \ 2846 typeof(type) *stat; \ 2847 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2848 u64_stats_init(&stat->syncp); \ 2849 } \ 2850 } \ 2851 pcpu_stats; \ 2852 }) 2853 2854 #define netdev_alloc_pcpu_stats(type) \ 2855 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2856 2857 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2858 ({ \ 2859 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2860 if (pcpu_stats) { \ 2861 int __cpu; \ 2862 for_each_possible_cpu(__cpu) { \ 2863 typeof(type) *stat; \ 2864 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2865 u64_stats_init(&stat->syncp); \ 2866 } \ 2867 } \ 2868 pcpu_stats; \ 2869 }) 2870 2871 enum netdev_lag_tx_type { 2872 NETDEV_LAG_TX_TYPE_UNKNOWN, 2873 NETDEV_LAG_TX_TYPE_RANDOM, 2874 NETDEV_LAG_TX_TYPE_BROADCAST, 2875 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2876 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2877 NETDEV_LAG_TX_TYPE_HASH, 2878 }; 2879 2880 enum netdev_lag_hash { 2881 NETDEV_LAG_HASH_NONE, 2882 NETDEV_LAG_HASH_L2, 2883 NETDEV_LAG_HASH_L34, 2884 NETDEV_LAG_HASH_L23, 2885 NETDEV_LAG_HASH_E23, 2886 NETDEV_LAG_HASH_E34, 2887 NETDEV_LAG_HASH_VLAN_SRCMAC, 2888 NETDEV_LAG_HASH_UNKNOWN, 2889 }; 2890 2891 struct netdev_lag_upper_info { 2892 enum netdev_lag_tx_type tx_type; 2893 enum netdev_lag_hash hash_type; 2894 }; 2895 2896 struct netdev_lag_lower_state_info { 2897 u8 link_up : 1, 2898 tx_enabled : 1; 2899 }; 2900 2901 #include <linux/notifier.h> 2902 2903 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2904 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2905 * adding new types. 2906 */ 2907 enum netdev_cmd { 2908 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2909 NETDEV_DOWN, 2910 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2911 detected a hardware crash and restarted 2912 - we can use this eg to kick tcp sessions 2913 once done */ 2914 NETDEV_CHANGE, /* Notify device state change */ 2915 NETDEV_REGISTER, 2916 NETDEV_UNREGISTER, 2917 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2918 NETDEV_CHANGEADDR, /* notify after the address change */ 2919 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2920 NETDEV_GOING_DOWN, 2921 NETDEV_CHANGENAME, 2922 NETDEV_FEAT_CHANGE, 2923 NETDEV_BONDING_FAILOVER, 2924 NETDEV_PRE_UP, 2925 NETDEV_PRE_TYPE_CHANGE, 2926 NETDEV_POST_TYPE_CHANGE, 2927 NETDEV_POST_INIT, 2928 NETDEV_PRE_UNINIT, 2929 NETDEV_RELEASE, 2930 NETDEV_NOTIFY_PEERS, 2931 NETDEV_JOIN, 2932 NETDEV_CHANGEUPPER, 2933 NETDEV_RESEND_IGMP, 2934 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2935 NETDEV_CHANGEINFODATA, 2936 NETDEV_BONDING_INFO, 2937 NETDEV_PRECHANGEUPPER, 2938 NETDEV_CHANGELOWERSTATE, 2939 NETDEV_UDP_TUNNEL_PUSH_INFO, 2940 NETDEV_UDP_TUNNEL_DROP_INFO, 2941 NETDEV_CHANGE_TX_QUEUE_LEN, 2942 NETDEV_CVLAN_FILTER_PUSH_INFO, 2943 NETDEV_CVLAN_FILTER_DROP_INFO, 2944 NETDEV_SVLAN_FILTER_PUSH_INFO, 2945 NETDEV_SVLAN_FILTER_DROP_INFO, 2946 NETDEV_OFFLOAD_XSTATS_ENABLE, 2947 NETDEV_OFFLOAD_XSTATS_DISABLE, 2948 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 2949 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 2950 NETDEV_XDP_FEAT_CHANGE, 2951 }; 2952 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2953 2954 int register_netdevice_notifier(struct notifier_block *nb); 2955 int unregister_netdevice_notifier(struct notifier_block *nb); 2956 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2957 int unregister_netdevice_notifier_net(struct net *net, 2958 struct notifier_block *nb); 2959 int register_netdevice_notifier_dev_net(struct net_device *dev, 2960 struct notifier_block *nb, 2961 struct netdev_net_notifier *nn); 2962 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2963 struct notifier_block *nb, 2964 struct netdev_net_notifier *nn); 2965 2966 struct netdev_notifier_info { 2967 struct net_device *dev; 2968 struct netlink_ext_ack *extack; 2969 }; 2970 2971 struct netdev_notifier_info_ext { 2972 struct netdev_notifier_info info; /* must be first */ 2973 union { 2974 u32 mtu; 2975 } ext; 2976 }; 2977 2978 struct netdev_notifier_change_info { 2979 struct netdev_notifier_info info; /* must be first */ 2980 unsigned int flags_changed; 2981 }; 2982 2983 struct netdev_notifier_changeupper_info { 2984 struct netdev_notifier_info info; /* must be first */ 2985 struct net_device *upper_dev; /* new upper dev */ 2986 bool master; /* is upper dev master */ 2987 bool linking; /* is the notification for link or unlink */ 2988 void *upper_info; /* upper dev info */ 2989 }; 2990 2991 struct netdev_notifier_changelowerstate_info { 2992 struct netdev_notifier_info info; /* must be first */ 2993 void *lower_state_info; /* is lower dev state */ 2994 }; 2995 2996 struct netdev_notifier_pre_changeaddr_info { 2997 struct netdev_notifier_info info; /* must be first */ 2998 const unsigned char *dev_addr; 2999 }; 3000 3001 enum netdev_offload_xstats_type { 3002 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 3003 }; 3004 3005 struct netdev_notifier_offload_xstats_info { 3006 struct netdev_notifier_info info; /* must be first */ 3007 enum netdev_offload_xstats_type type; 3008 3009 union { 3010 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 3011 struct netdev_notifier_offload_xstats_rd *report_delta; 3012 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 3013 struct netdev_notifier_offload_xstats_ru *report_used; 3014 }; 3015 }; 3016 3017 int netdev_offload_xstats_enable(struct net_device *dev, 3018 enum netdev_offload_xstats_type type, 3019 struct netlink_ext_ack *extack); 3020 int netdev_offload_xstats_disable(struct net_device *dev, 3021 enum netdev_offload_xstats_type type); 3022 bool netdev_offload_xstats_enabled(const struct net_device *dev, 3023 enum netdev_offload_xstats_type type); 3024 int netdev_offload_xstats_get(struct net_device *dev, 3025 enum netdev_offload_xstats_type type, 3026 struct rtnl_hw_stats64 *stats, bool *used, 3027 struct netlink_ext_ack *extack); 3028 void 3029 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 3030 const struct rtnl_hw_stats64 *stats); 3031 void 3032 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 3033 void netdev_offload_xstats_push_delta(struct net_device *dev, 3034 enum netdev_offload_xstats_type type, 3035 const struct rtnl_hw_stats64 *stats); 3036 3037 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 3038 struct net_device *dev) 3039 { 3040 info->dev = dev; 3041 info->extack = NULL; 3042 } 3043 3044 static inline struct net_device * 3045 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 3046 { 3047 return info->dev; 3048 } 3049 3050 static inline struct netlink_ext_ack * 3051 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 3052 { 3053 return info->extack; 3054 } 3055 3056 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 3057 int call_netdevice_notifiers_info(unsigned long val, 3058 struct netdev_notifier_info *info); 3059 3060 extern rwlock_t dev_base_lock; /* Device list lock */ 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 xa_for_each_start(&(net)->dev_by_index, (ifindex), (d), (ifindex)) 3084 3085 static inline struct net_device *next_net_device(struct net_device *dev) 3086 { 3087 struct list_head *lh; 3088 struct net *net; 3089 3090 net = dev_net(dev); 3091 lh = dev->dev_list.next; 3092 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3093 } 3094 3095 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 3096 { 3097 struct list_head *lh; 3098 struct net *net; 3099 3100 net = dev_net(dev); 3101 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 3102 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3103 } 3104 3105 static inline struct net_device *first_net_device(struct net *net) 3106 { 3107 return list_empty(&net->dev_base_head) ? NULL : 3108 net_device_entry(net->dev_base_head.next); 3109 } 3110 3111 static inline struct net_device *first_net_device_rcu(struct net *net) 3112 { 3113 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 3114 3115 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3116 } 3117 3118 int netdev_boot_setup_check(struct net_device *dev); 3119 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 3120 const char *hwaddr); 3121 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 3122 void dev_add_pack(struct packet_type *pt); 3123 void dev_remove_pack(struct packet_type *pt); 3124 void __dev_remove_pack(struct packet_type *pt); 3125 void dev_add_offload(struct packet_offload *po); 3126 void dev_remove_offload(struct packet_offload *po); 3127 3128 int dev_get_iflink(const struct net_device *dev); 3129 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 3130 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 3131 struct net_device_path_stack *stack); 3132 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 3133 unsigned short mask); 3134 struct net_device *dev_get_by_name(struct net *net, const char *name); 3135 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 3136 struct net_device *__dev_get_by_name(struct net *net, const char *name); 3137 bool netdev_name_in_use(struct net *net, const char *name); 3138 int dev_alloc_name(struct net_device *dev, const char *name); 3139 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 3140 void dev_close(struct net_device *dev); 3141 void dev_close_many(struct list_head *head, bool unlink); 3142 void dev_disable_lro(struct net_device *dev); 3143 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 3144 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 3145 struct net_device *sb_dev); 3146 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 3147 struct net_device *sb_dev); 3148 3149 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3150 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3151 3152 static inline int dev_queue_xmit(struct sk_buff *skb) 3153 { 3154 return __dev_queue_xmit(skb, NULL); 3155 } 3156 3157 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3158 struct net_device *sb_dev) 3159 { 3160 return __dev_queue_xmit(skb, sb_dev); 3161 } 3162 3163 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3164 { 3165 int ret; 3166 3167 ret = __dev_direct_xmit(skb, queue_id); 3168 if (!dev_xmit_complete(ret)) 3169 kfree_skb(skb); 3170 return ret; 3171 } 3172 3173 int register_netdevice(struct net_device *dev); 3174 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3175 void unregister_netdevice_many(struct list_head *head); 3176 static inline void unregister_netdevice(struct net_device *dev) 3177 { 3178 unregister_netdevice_queue(dev, NULL); 3179 } 3180 3181 int netdev_refcnt_read(const struct net_device *dev); 3182 void free_netdev(struct net_device *dev); 3183 void netdev_freemem(struct net_device *dev); 3184 int init_dummy_netdev(struct net_device *dev); 3185 3186 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3187 struct sk_buff *skb, 3188 bool all_slaves); 3189 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3190 struct sock *sk); 3191 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3192 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3193 struct net_device *netdev_get_by_index(struct net *net, int ifindex, 3194 netdevice_tracker *tracker, gfp_t gfp); 3195 struct net_device *netdev_get_by_name(struct net *net, const char *name, 3196 netdevice_tracker *tracker, gfp_t gfp); 3197 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3198 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 3199 3200 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3201 unsigned short type, 3202 const void *daddr, const void *saddr, 3203 unsigned int len) 3204 { 3205 if (!dev->header_ops || !dev->header_ops->create) 3206 return 0; 3207 3208 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3209 } 3210 3211 static inline int dev_parse_header(const struct sk_buff *skb, 3212 unsigned char *haddr) 3213 { 3214 const struct net_device *dev = skb->dev; 3215 3216 if (!dev->header_ops || !dev->header_ops->parse) 3217 return 0; 3218 return dev->header_ops->parse(skb, haddr); 3219 } 3220 3221 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3222 { 3223 const struct net_device *dev = skb->dev; 3224 3225 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3226 return 0; 3227 return dev->header_ops->parse_protocol(skb); 3228 } 3229 3230 /* ll_header must have at least hard_header_len allocated */ 3231 static inline bool dev_validate_header(const struct net_device *dev, 3232 char *ll_header, int len) 3233 { 3234 if (likely(len >= dev->hard_header_len)) 3235 return true; 3236 if (len < dev->min_header_len) 3237 return false; 3238 3239 if (capable(CAP_SYS_RAWIO)) { 3240 memset(ll_header + len, 0, dev->hard_header_len - len); 3241 return true; 3242 } 3243 3244 if (dev->header_ops && dev->header_ops->validate) 3245 return dev->header_ops->validate(ll_header, len); 3246 3247 return false; 3248 } 3249 3250 static inline bool dev_has_header(const struct net_device *dev) 3251 { 3252 return dev->header_ops && dev->header_ops->create; 3253 } 3254 3255 /* 3256 * Incoming packets are placed on per-CPU queues 3257 */ 3258 struct softnet_data { 3259 struct list_head poll_list; 3260 struct sk_buff_head process_queue; 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 bool in_net_rx_action; 3270 bool in_napi_threaded_poll; 3271 3272 #ifdef CONFIG_NET_FLOW_LIMIT 3273 struct sd_flow_limit __rcu *flow_limit; 3274 #endif 3275 struct Qdisc *output_queue; 3276 struct Qdisc **output_queue_tailp; 3277 struct sk_buff *completion_queue; 3278 #ifdef CONFIG_XFRM_OFFLOAD 3279 struct sk_buff_head xfrm_backlog; 3280 #endif 3281 /* written and read only by owning cpu: */ 3282 struct { 3283 u16 recursion; 3284 u8 more; 3285 #ifdef CONFIG_NET_EGRESS 3286 u8 skip_txqueue; 3287 #endif 3288 } xmit; 3289 #ifdef CONFIG_RPS 3290 /* input_queue_head should be written by cpu owning this struct, 3291 * and only read by other cpus. Worth using a cache line. 3292 */ 3293 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3294 3295 /* Elements below can be accessed between CPUs for RPS/RFS */ 3296 call_single_data_t csd ____cacheline_aligned_in_smp; 3297 struct softnet_data *rps_ipi_next; 3298 unsigned int cpu; 3299 unsigned int input_queue_tail; 3300 #endif 3301 unsigned int received_rps; 3302 unsigned int dropped; 3303 struct sk_buff_head input_pkt_queue; 3304 struct napi_struct backlog; 3305 3306 /* Another possibly contended cache line */ 3307 spinlock_t defer_lock ____cacheline_aligned_in_smp; 3308 int defer_count; 3309 int defer_ipi_scheduled; 3310 struct sk_buff *defer_list; 3311 call_single_data_t defer_csd; 3312 }; 3313 3314 static inline void input_queue_head_incr(struct softnet_data *sd) 3315 { 3316 #ifdef CONFIG_RPS 3317 sd->input_queue_head++; 3318 #endif 3319 } 3320 3321 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3322 unsigned int *qtail) 3323 { 3324 #ifdef CONFIG_RPS 3325 *qtail = ++sd->input_queue_tail; 3326 #endif 3327 } 3328 3329 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3330 3331 static inline int dev_recursion_level(void) 3332 { 3333 return this_cpu_read(softnet_data.xmit.recursion); 3334 } 3335 3336 #define XMIT_RECURSION_LIMIT 8 3337 static inline bool dev_xmit_recursion(void) 3338 { 3339 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3340 XMIT_RECURSION_LIMIT); 3341 } 3342 3343 static inline void dev_xmit_recursion_inc(void) 3344 { 3345 __this_cpu_inc(softnet_data.xmit.recursion); 3346 } 3347 3348 static inline void dev_xmit_recursion_dec(void) 3349 { 3350 __this_cpu_dec(softnet_data.xmit.recursion); 3351 } 3352 3353 void __netif_schedule(struct Qdisc *q); 3354 void netif_schedule_queue(struct netdev_queue *txq); 3355 3356 static inline void netif_tx_schedule_all(struct net_device *dev) 3357 { 3358 unsigned int i; 3359 3360 for (i = 0; i < dev->num_tx_queues; i++) 3361 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3362 } 3363 3364 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3365 { 3366 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3367 } 3368 3369 /** 3370 * netif_start_queue - allow transmit 3371 * @dev: network device 3372 * 3373 * Allow upper layers to call the device hard_start_xmit routine. 3374 */ 3375 static inline void netif_start_queue(struct net_device *dev) 3376 { 3377 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3378 } 3379 3380 static inline void netif_tx_start_all_queues(struct net_device *dev) 3381 { 3382 unsigned int i; 3383 3384 for (i = 0; i < dev->num_tx_queues; i++) { 3385 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3386 netif_tx_start_queue(txq); 3387 } 3388 } 3389 3390 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3391 3392 /** 3393 * netif_wake_queue - restart transmit 3394 * @dev: network device 3395 * 3396 * Allow upper layers to call the device hard_start_xmit routine. 3397 * Used for flow control when transmit resources are available. 3398 */ 3399 static inline void netif_wake_queue(struct net_device *dev) 3400 { 3401 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3402 } 3403 3404 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3405 { 3406 unsigned int i; 3407 3408 for (i = 0; i < dev->num_tx_queues; i++) { 3409 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3410 netif_tx_wake_queue(txq); 3411 } 3412 } 3413 3414 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3415 { 3416 /* Must be an atomic op see netif_txq_try_stop() */ 3417 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3418 } 3419 3420 /** 3421 * netif_stop_queue - stop transmitted packets 3422 * @dev: network device 3423 * 3424 * Stop upper layers calling the device hard_start_xmit routine. 3425 * Used for flow control when transmit resources are unavailable. 3426 */ 3427 static inline void netif_stop_queue(struct net_device *dev) 3428 { 3429 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3430 } 3431 3432 void netif_tx_stop_all_queues(struct net_device *dev); 3433 3434 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3435 { 3436 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3437 } 3438 3439 /** 3440 * netif_queue_stopped - test if transmit queue is flowblocked 3441 * @dev: network device 3442 * 3443 * Test if transmit queue on device is currently unable to send. 3444 */ 3445 static inline bool netif_queue_stopped(const struct net_device *dev) 3446 { 3447 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3448 } 3449 3450 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3451 { 3452 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3453 } 3454 3455 static inline bool 3456 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3457 { 3458 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3459 } 3460 3461 static inline bool 3462 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3463 { 3464 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3465 } 3466 3467 /** 3468 * netdev_queue_set_dql_min_limit - set dql minimum limit 3469 * @dev_queue: pointer to transmit queue 3470 * @min_limit: dql minimum limit 3471 * 3472 * Forces xmit_more() to return true until the minimum threshold 3473 * defined by @min_limit is reached (or until the tx queue is 3474 * empty). Warning: to be use with care, misuse will impact the 3475 * latency. 3476 */ 3477 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3478 unsigned int min_limit) 3479 { 3480 #ifdef CONFIG_BQL 3481 dev_queue->dql.min_limit = min_limit; 3482 #endif 3483 } 3484 3485 /** 3486 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3487 * @dev_queue: pointer to transmit queue 3488 * 3489 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3490 * to give appropriate hint to the CPU. 3491 */ 3492 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3493 { 3494 #ifdef CONFIG_BQL 3495 prefetchw(&dev_queue->dql.num_queued); 3496 #endif 3497 } 3498 3499 /** 3500 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3501 * @dev_queue: pointer to transmit queue 3502 * 3503 * BQL enabled drivers might use this helper in their TX completion path, 3504 * to give appropriate hint to the CPU. 3505 */ 3506 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3507 { 3508 #ifdef CONFIG_BQL 3509 prefetchw(&dev_queue->dql.limit); 3510 #endif 3511 } 3512 3513 /** 3514 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3515 * @dev_queue: network device queue 3516 * @bytes: number of bytes queued to the device queue 3517 * 3518 * Report the number of bytes queued for sending/completion to the network 3519 * device hardware queue. @bytes should be a good approximation and should 3520 * exactly match netdev_completed_queue() @bytes. 3521 * This is typically called once per packet, from ndo_start_xmit(). 3522 */ 3523 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3524 unsigned int bytes) 3525 { 3526 #ifdef CONFIG_BQL 3527 dql_queued(&dev_queue->dql, bytes); 3528 3529 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3530 return; 3531 3532 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3533 3534 /* 3535 * The XOFF flag must be set before checking the dql_avail below, 3536 * because in netdev_tx_completed_queue we update the dql_completed 3537 * before checking the XOFF flag. 3538 */ 3539 smp_mb(); 3540 3541 /* check again in case another CPU has just made room avail */ 3542 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3543 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3544 #endif 3545 } 3546 3547 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3548 * that they should not test BQL status themselves. 3549 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3550 * skb of a batch. 3551 * Returns true if the doorbell must be used to kick the NIC. 3552 */ 3553 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3554 unsigned int bytes, 3555 bool xmit_more) 3556 { 3557 if (xmit_more) { 3558 #ifdef CONFIG_BQL 3559 dql_queued(&dev_queue->dql, bytes); 3560 #endif 3561 return netif_tx_queue_stopped(dev_queue); 3562 } 3563 netdev_tx_sent_queue(dev_queue, bytes); 3564 return true; 3565 } 3566 3567 /** 3568 * netdev_sent_queue - report the number of bytes queued to hardware 3569 * @dev: network device 3570 * @bytes: number of bytes queued to the hardware device queue 3571 * 3572 * Report the number of bytes queued for sending/completion to the network 3573 * device hardware queue#0. @bytes should be a good approximation and should 3574 * exactly match netdev_completed_queue() @bytes. 3575 * This is typically called once per packet, from ndo_start_xmit(). 3576 */ 3577 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3578 { 3579 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3580 } 3581 3582 static inline bool __netdev_sent_queue(struct net_device *dev, 3583 unsigned int bytes, 3584 bool xmit_more) 3585 { 3586 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3587 xmit_more); 3588 } 3589 3590 /** 3591 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3592 * @dev_queue: network device queue 3593 * @pkts: number of packets (currently ignored) 3594 * @bytes: number of bytes dequeued from the device queue 3595 * 3596 * Must be called at most once per TX completion round (and not per 3597 * individual packet), so that BQL can adjust its limits appropriately. 3598 */ 3599 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3600 unsigned int pkts, unsigned int bytes) 3601 { 3602 #ifdef CONFIG_BQL 3603 if (unlikely(!bytes)) 3604 return; 3605 3606 dql_completed(&dev_queue->dql, bytes); 3607 3608 /* 3609 * Without the memory barrier there is a small possiblity that 3610 * netdev_tx_sent_queue will miss the update and cause the queue to 3611 * be stopped forever 3612 */ 3613 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */ 3614 3615 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3616 return; 3617 3618 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3619 netif_schedule_queue(dev_queue); 3620 #endif 3621 } 3622 3623 /** 3624 * netdev_completed_queue - report bytes and packets completed by device 3625 * @dev: network device 3626 * @pkts: actual number of packets sent over the medium 3627 * @bytes: actual number of bytes sent over the medium 3628 * 3629 * Report the number of bytes and packets transmitted by the network device 3630 * hardware queue over the physical medium, @bytes must exactly match the 3631 * @bytes amount passed to netdev_sent_queue() 3632 */ 3633 static inline void netdev_completed_queue(struct net_device *dev, 3634 unsigned int pkts, unsigned int bytes) 3635 { 3636 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3637 } 3638 3639 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3640 { 3641 #ifdef CONFIG_BQL 3642 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3643 dql_reset(&q->dql); 3644 #endif 3645 } 3646 3647 /** 3648 * netdev_reset_queue - reset the packets and bytes count of a network device 3649 * @dev_queue: network device 3650 * 3651 * Reset the bytes and packet count of a network device and clear the 3652 * software flow control OFF bit for this network device 3653 */ 3654 static inline void netdev_reset_queue(struct net_device *dev_queue) 3655 { 3656 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3657 } 3658 3659 /** 3660 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3661 * @dev: network device 3662 * @queue_index: given tx queue index 3663 * 3664 * Returns 0 if given tx queue index >= number of device tx queues, 3665 * otherwise returns the originally passed tx queue index. 3666 */ 3667 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3668 { 3669 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3670 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3671 dev->name, queue_index, 3672 dev->real_num_tx_queues); 3673 return 0; 3674 } 3675 3676 return queue_index; 3677 } 3678 3679 /** 3680 * netif_running - test if up 3681 * @dev: network device 3682 * 3683 * Test if the device has been brought up. 3684 */ 3685 static inline bool netif_running(const struct net_device *dev) 3686 { 3687 return test_bit(__LINK_STATE_START, &dev->state); 3688 } 3689 3690 /* 3691 * Routines to manage the subqueues on a device. We only need start, 3692 * stop, and a check if it's stopped. All other device management is 3693 * done at the overall netdevice level. 3694 * Also test the device if we're multiqueue. 3695 */ 3696 3697 /** 3698 * netif_start_subqueue - allow sending packets on subqueue 3699 * @dev: network device 3700 * @queue_index: sub queue index 3701 * 3702 * Start individual transmit queue of a device with multiple transmit queues. 3703 */ 3704 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3705 { 3706 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3707 3708 netif_tx_start_queue(txq); 3709 } 3710 3711 /** 3712 * netif_stop_subqueue - stop sending packets on subqueue 3713 * @dev: network device 3714 * @queue_index: sub queue index 3715 * 3716 * Stop individual transmit queue of a device with multiple transmit queues. 3717 */ 3718 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3719 { 3720 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3721 netif_tx_stop_queue(txq); 3722 } 3723 3724 /** 3725 * __netif_subqueue_stopped - test status of subqueue 3726 * @dev: network device 3727 * @queue_index: sub queue index 3728 * 3729 * Check individual transmit queue of a device with multiple transmit queues. 3730 */ 3731 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3732 u16 queue_index) 3733 { 3734 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3735 3736 return netif_tx_queue_stopped(txq); 3737 } 3738 3739 /** 3740 * netif_subqueue_stopped - test status of subqueue 3741 * @dev: network device 3742 * @skb: sub queue buffer pointer 3743 * 3744 * Check individual transmit queue of a device with multiple transmit queues. 3745 */ 3746 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3747 struct sk_buff *skb) 3748 { 3749 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3750 } 3751 3752 /** 3753 * netif_wake_subqueue - allow sending packets on subqueue 3754 * @dev: network device 3755 * @queue_index: sub queue index 3756 * 3757 * Resume individual transmit queue of a device with multiple transmit queues. 3758 */ 3759 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3760 { 3761 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3762 3763 netif_tx_wake_queue(txq); 3764 } 3765 3766 #ifdef CONFIG_XPS 3767 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3768 u16 index); 3769 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3770 u16 index, enum xps_map_type type); 3771 3772 /** 3773 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3774 * @j: CPU/Rx queue index 3775 * @mask: bitmask of all cpus/rx queues 3776 * @nr_bits: number of bits in the bitmask 3777 * 3778 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3779 */ 3780 static inline bool netif_attr_test_mask(unsigned long j, 3781 const unsigned long *mask, 3782 unsigned int nr_bits) 3783 { 3784 cpu_max_bits_warn(j, nr_bits); 3785 return test_bit(j, mask); 3786 } 3787 3788 /** 3789 * netif_attr_test_online - Test for online CPU/Rx queue 3790 * @j: CPU/Rx queue index 3791 * @online_mask: bitmask for CPUs/Rx queues that are online 3792 * @nr_bits: number of bits in the bitmask 3793 * 3794 * Returns true if a CPU/Rx queue is online. 3795 */ 3796 static inline bool netif_attr_test_online(unsigned long j, 3797 const unsigned long *online_mask, 3798 unsigned int nr_bits) 3799 { 3800 cpu_max_bits_warn(j, nr_bits); 3801 3802 if (online_mask) 3803 return test_bit(j, online_mask); 3804 3805 return (j < nr_bits); 3806 } 3807 3808 /** 3809 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3810 * @n: CPU/Rx queue index 3811 * @srcp: the cpumask/Rx queue mask pointer 3812 * @nr_bits: number of bits in the bitmask 3813 * 3814 * Returns >= nr_bits if no further CPUs/Rx queues set. 3815 */ 3816 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3817 unsigned int nr_bits) 3818 { 3819 /* -1 is a legal arg here. */ 3820 if (n != -1) 3821 cpu_max_bits_warn(n, nr_bits); 3822 3823 if (srcp) 3824 return find_next_bit(srcp, nr_bits, n + 1); 3825 3826 return n + 1; 3827 } 3828 3829 /** 3830 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3831 * @n: CPU/Rx queue index 3832 * @src1p: the first CPUs/Rx queues mask pointer 3833 * @src2p: the second CPUs/Rx queues mask pointer 3834 * @nr_bits: number of bits in the bitmask 3835 * 3836 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3837 */ 3838 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3839 const unsigned long *src2p, 3840 unsigned int nr_bits) 3841 { 3842 /* -1 is a legal arg here. */ 3843 if (n != -1) 3844 cpu_max_bits_warn(n, nr_bits); 3845 3846 if (src1p && src2p) 3847 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3848 else if (src1p) 3849 return find_next_bit(src1p, nr_bits, n + 1); 3850 else if (src2p) 3851 return find_next_bit(src2p, nr_bits, n + 1); 3852 3853 return n + 1; 3854 } 3855 #else 3856 static inline int netif_set_xps_queue(struct net_device *dev, 3857 const struct cpumask *mask, 3858 u16 index) 3859 { 3860 return 0; 3861 } 3862 3863 static inline int __netif_set_xps_queue(struct net_device *dev, 3864 const unsigned long *mask, 3865 u16 index, enum xps_map_type type) 3866 { 3867 return 0; 3868 } 3869 #endif 3870 3871 /** 3872 * netif_is_multiqueue - test if device has multiple transmit queues 3873 * @dev: network device 3874 * 3875 * Check if device has multiple transmit queues 3876 */ 3877 static inline bool netif_is_multiqueue(const struct net_device *dev) 3878 { 3879 return dev->num_tx_queues > 1; 3880 } 3881 3882 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3883 3884 #ifdef CONFIG_SYSFS 3885 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3886 #else 3887 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3888 unsigned int rxqs) 3889 { 3890 dev->real_num_rx_queues = rxqs; 3891 return 0; 3892 } 3893 #endif 3894 int netif_set_real_num_queues(struct net_device *dev, 3895 unsigned int txq, unsigned int rxq); 3896 3897 int netif_get_num_default_rss_queues(void); 3898 3899 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason); 3900 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason); 3901 3902 /* 3903 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3904 * interrupt context or with hardware interrupts being disabled. 3905 * (in_hardirq() || irqs_disabled()) 3906 * 3907 * We provide four helpers that can be used in following contexts : 3908 * 3909 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3910 * replacing kfree_skb(skb) 3911 * 3912 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3913 * Typically used in place of consume_skb(skb) in TX completion path 3914 * 3915 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3916 * replacing kfree_skb(skb) 3917 * 3918 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3919 * and consumed a packet. Used in place of consume_skb(skb) 3920 */ 3921 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3922 { 3923 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 3924 } 3925 3926 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3927 { 3928 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED); 3929 } 3930 3931 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3932 { 3933 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 3934 } 3935 3936 static inline void dev_consume_skb_any(struct sk_buff *skb) 3937 { 3938 dev_kfree_skb_any_reason(skb, SKB_CONSUMED); 3939 } 3940 3941 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3942 struct bpf_prog *xdp_prog); 3943 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3944 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3945 int netif_rx(struct sk_buff *skb); 3946 int __netif_rx(struct sk_buff *skb); 3947 3948 int netif_receive_skb(struct sk_buff *skb); 3949 int netif_receive_skb_core(struct sk_buff *skb); 3950 void netif_receive_skb_list_internal(struct list_head *head); 3951 void netif_receive_skb_list(struct list_head *head); 3952 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3953 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3954 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3955 void napi_get_frags_check(struct napi_struct *napi); 3956 gro_result_t napi_gro_frags(struct napi_struct *napi); 3957 struct packet_offload *gro_find_receive_by_type(__be16 type); 3958 struct packet_offload *gro_find_complete_by_type(__be16 type); 3959 3960 static inline void napi_free_frags(struct napi_struct *napi) 3961 { 3962 kfree_skb(napi->skb); 3963 napi->skb = NULL; 3964 } 3965 3966 bool netdev_is_rx_handler_busy(struct net_device *dev); 3967 int netdev_rx_handler_register(struct net_device *dev, 3968 rx_handler_func_t *rx_handler, 3969 void *rx_handler_data); 3970 void netdev_rx_handler_unregister(struct net_device *dev); 3971 3972 bool dev_valid_name(const char *name); 3973 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 3974 { 3975 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 3976 } 3977 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 3978 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 3979 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3980 void __user *data, bool *need_copyout); 3981 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 3982 int generic_hwtstamp_get_lower(struct net_device *dev, 3983 struct kernel_hwtstamp_config *kernel_cfg); 3984 int generic_hwtstamp_set_lower(struct net_device *dev, 3985 struct kernel_hwtstamp_config *kernel_cfg, 3986 struct netlink_ext_ack *extack); 3987 int dev_set_hwtstamp_phylib(struct net_device *dev, 3988 struct kernel_hwtstamp_config *cfg, 3989 struct netlink_ext_ack *extack); 3990 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 3991 unsigned int dev_get_flags(const struct net_device *); 3992 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3993 struct netlink_ext_ack *extack); 3994 int dev_change_flags(struct net_device *dev, unsigned int flags, 3995 struct netlink_ext_ack *extack); 3996 int dev_set_alias(struct net_device *, const char *, size_t); 3997 int dev_get_alias(const struct net_device *, char *, size_t); 3998 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 3999 const char *pat, int new_ifindex); 4000 static inline 4001 int dev_change_net_namespace(struct net_device *dev, struct net *net, 4002 const char *pat) 4003 { 4004 return __dev_change_net_namespace(dev, net, pat, 0); 4005 } 4006 int __dev_set_mtu(struct net_device *, int); 4007 int dev_set_mtu(struct net_device *, int); 4008 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 4009 struct netlink_ext_ack *extack); 4010 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 4011 struct netlink_ext_ack *extack); 4012 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 4013 struct netlink_ext_ack *extack); 4014 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 4015 int dev_get_port_parent_id(struct net_device *dev, 4016 struct netdev_phys_item_id *ppid, bool recurse); 4017 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 4018 void netdev_dpll_pin_set(struct net_device *dev, struct dpll_pin *dpll_pin); 4019 void netdev_dpll_pin_clear(struct net_device *dev); 4020 4021 static inline struct dpll_pin *netdev_dpll_pin(const struct net_device *dev) 4022 { 4023 #if IS_ENABLED(CONFIG_DPLL) 4024 return dev->dpll_pin; 4025 #else 4026 return NULL; 4027 #endif 4028 } 4029 4030 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 4031 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 4032 struct netdev_queue *txq, int *ret); 4033 4034 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 4035 u8 dev_xdp_prog_count(struct net_device *dev); 4036 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 4037 4038 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4039 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4040 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 4041 bool is_skb_forwardable(const struct net_device *dev, 4042 const struct sk_buff *skb); 4043 4044 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 4045 const struct sk_buff *skb, 4046 const bool check_mtu) 4047 { 4048 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 4049 unsigned int len; 4050 4051 if (!(dev->flags & IFF_UP)) 4052 return false; 4053 4054 if (!check_mtu) 4055 return true; 4056 4057 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 4058 if (skb->len <= len) 4059 return true; 4060 4061 /* if TSO is enabled, we don't care about the length as the packet 4062 * could be forwarded without being segmented before 4063 */ 4064 if (skb_is_gso(skb)) 4065 return true; 4066 4067 return false; 4068 } 4069 4070 void netdev_core_stats_inc(struct net_device *dev, u32 offset); 4071 4072 #define DEV_CORE_STATS_INC(FIELD) \ 4073 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 4074 { \ 4075 netdev_core_stats_inc(dev, \ 4076 offsetof(struct net_device_core_stats, FIELD)); \ 4077 } 4078 DEV_CORE_STATS_INC(rx_dropped) 4079 DEV_CORE_STATS_INC(tx_dropped) 4080 DEV_CORE_STATS_INC(rx_nohandler) 4081 DEV_CORE_STATS_INC(rx_otherhost_dropped) 4082 #undef DEV_CORE_STATS_INC 4083 4084 static __always_inline int ____dev_forward_skb(struct net_device *dev, 4085 struct sk_buff *skb, 4086 const bool check_mtu) 4087 { 4088 if (skb_orphan_frags(skb, GFP_ATOMIC) || 4089 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 4090 dev_core_stats_rx_dropped_inc(dev); 4091 kfree_skb(skb); 4092 return NET_RX_DROP; 4093 } 4094 4095 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 4096 skb->priority = 0; 4097 return 0; 4098 } 4099 4100 bool dev_nit_active(struct net_device *dev); 4101 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 4102 4103 static inline void __dev_put(struct net_device *dev) 4104 { 4105 if (dev) { 4106 #ifdef CONFIG_PCPU_DEV_REFCNT 4107 this_cpu_dec(*dev->pcpu_refcnt); 4108 #else 4109 refcount_dec(&dev->dev_refcnt); 4110 #endif 4111 } 4112 } 4113 4114 static inline void __dev_hold(struct net_device *dev) 4115 { 4116 if (dev) { 4117 #ifdef CONFIG_PCPU_DEV_REFCNT 4118 this_cpu_inc(*dev->pcpu_refcnt); 4119 #else 4120 refcount_inc(&dev->dev_refcnt); 4121 #endif 4122 } 4123 } 4124 4125 static inline void __netdev_tracker_alloc(struct net_device *dev, 4126 netdevice_tracker *tracker, 4127 gfp_t gfp) 4128 { 4129 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4130 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 4131 #endif 4132 } 4133 4134 /* netdev_tracker_alloc() can upgrade a prior untracked reference 4135 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4136 */ 4137 static inline void netdev_tracker_alloc(struct net_device *dev, 4138 netdevice_tracker *tracker, gfp_t gfp) 4139 { 4140 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4141 refcount_dec(&dev->refcnt_tracker.no_tracker); 4142 __netdev_tracker_alloc(dev, tracker, gfp); 4143 #endif 4144 } 4145 4146 static inline void netdev_tracker_free(struct net_device *dev, 4147 netdevice_tracker *tracker) 4148 { 4149 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4150 ref_tracker_free(&dev->refcnt_tracker, tracker); 4151 #endif 4152 } 4153 4154 static inline void netdev_hold(struct net_device *dev, 4155 netdevice_tracker *tracker, gfp_t gfp) 4156 { 4157 if (dev) { 4158 __dev_hold(dev); 4159 __netdev_tracker_alloc(dev, tracker, gfp); 4160 } 4161 } 4162 4163 static inline void netdev_put(struct net_device *dev, 4164 netdevice_tracker *tracker) 4165 { 4166 if (dev) { 4167 netdev_tracker_free(dev, tracker); 4168 __dev_put(dev); 4169 } 4170 } 4171 4172 /** 4173 * dev_hold - get reference to device 4174 * @dev: network device 4175 * 4176 * Hold reference to device to keep it from being freed. 4177 * Try using netdev_hold() instead. 4178 */ 4179 static inline void dev_hold(struct net_device *dev) 4180 { 4181 netdev_hold(dev, NULL, GFP_ATOMIC); 4182 } 4183 4184 /** 4185 * dev_put - release reference to device 4186 * @dev: network device 4187 * 4188 * Release reference to device to allow it to be freed. 4189 * Try using netdev_put() instead. 4190 */ 4191 static inline void dev_put(struct net_device *dev) 4192 { 4193 netdev_put(dev, NULL); 4194 } 4195 4196 static inline void netdev_ref_replace(struct net_device *odev, 4197 struct net_device *ndev, 4198 netdevice_tracker *tracker, 4199 gfp_t gfp) 4200 { 4201 if (odev) 4202 netdev_tracker_free(odev, tracker); 4203 4204 __dev_hold(ndev); 4205 __dev_put(odev); 4206 4207 if (ndev) 4208 __netdev_tracker_alloc(ndev, tracker, gfp); 4209 } 4210 4211 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4212 * and _off may be called from IRQ context, but it is caller 4213 * who is responsible for serialization of these calls. 4214 * 4215 * The name carrier is inappropriate, these functions should really be 4216 * called netif_lowerlayer_*() because they represent the state of any 4217 * kind of lower layer not just hardware media. 4218 */ 4219 void linkwatch_fire_event(struct net_device *dev); 4220 4221 /** 4222 * netif_carrier_ok - test if carrier present 4223 * @dev: network device 4224 * 4225 * Check if carrier is present on device 4226 */ 4227 static inline bool netif_carrier_ok(const struct net_device *dev) 4228 { 4229 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4230 } 4231 4232 unsigned long dev_trans_start(struct net_device *dev); 4233 4234 void __netdev_watchdog_up(struct net_device *dev); 4235 4236 void netif_carrier_on(struct net_device *dev); 4237 void netif_carrier_off(struct net_device *dev); 4238 void netif_carrier_event(struct net_device *dev); 4239 4240 /** 4241 * netif_dormant_on - mark device as dormant. 4242 * @dev: network device 4243 * 4244 * Mark device as dormant (as per RFC2863). 4245 * 4246 * The dormant state indicates that the relevant interface is not 4247 * actually in a condition to pass packets (i.e., it is not 'up') but is 4248 * in a "pending" state, waiting for some external event. For "on- 4249 * demand" interfaces, this new state identifies the situation where the 4250 * interface is waiting for events to place it in the up state. 4251 */ 4252 static inline void netif_dormant_on(struct net_device *dev) 4253 { 4254 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4255 linkwatch_fire_event(dev); 4256 } 4257 4258 /** 4259 * netif_dormant_off - set device as not dormant. 4260 * @dev: network device 4261 * 4262 * Device is not in dormant state. 4263 */ 4264 static inline void netif_dormant_off(struct net_device *dev) 4265 { 4266 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4267 linkwatch_fire_event(dev); 4268 } 4269 4270 /** 4271 * netif_dormant - test if device is dormant 4272 * @dev: network device 4273 * 4274 * Check if device is dormant. 4275 */ 4276 static inline bool netif_dormant(const struct net_device *dev) 4277 { 4278 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4279 } 4280 4281 4282 /** 4283 * netif_testing_on - mark device as under test. 4284 * @dev: network device 4285 * 4286 * Mark device as under test (as per RFC2863). 4287 * 4288 * The testing state indicates that some test(s) must be performed on 4289 * the interface. After completion, of the test, the interface state 4290 * will change to up, dormant, or down, as appropriate. 4291 */ 4292 static inline void netif_testing_on(struct net_device *dev) 4293 { 4294 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4295 linkwatch_fire_event(dev); 4296 } 4297 4298 /** 4299 * netif_testing_off - set device as not under test. 4300 * @dev: network device 4301 * 4302 * Device is not in testing state. 4303 */ 4304 static inline void netif_testing_off(struct net_device *dev) 4305 { 4306 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4307 linkwatch_fire_event(dev); 4308 } 4309 4310 /** 4311 * netif_testing - test if device is under test 4312 * @dev: network device 4313 * 4314 * Check if device is under test 4315 */ 4316 static inline bool netif_testing(const struct net_device *dev) 4317 { 4318 return test_bit(__LINK_STATE_TESTING, &dev->state); 4319 } 4320 4321 4322 /** 4323 * netif_oper_up - test if device is operational 4324 * @dev: network device 4325 * 4326 * Check if carrier is operational 4327 */ 4328 static inline bool netif_oper_up(const struct net_device *dev) 4329 { 4330 return (dev->operstate == IF_OPER_UP || 4331 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4332 } 4333 4334 /** 4335 * netif_device_present - is device available or removed 4336 * @dev: network device 4337 * 4338 * Check if device has not been removed from system. 4339 */ 4340 static inline bool netif_device_present(const struct net_device *dev) 4341 { 4342 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4343 } 4344 4345 void netif_device_detach(struct net_device *dev); 4346 4347 void netif_device_attach(struct net_device *dev); 4348 4349 /* 4350 * Network interface message level settings 4351 */ 4352 4353 enum { 4354 NETIF_MSG_DRV_BIT, 4355 NETIF_MSG_PROBE_BIT, 4356 NETIF_MSG_LINK_BIT, 4357 NETIF_MSG_TIMER_BIT, 4358 NETIF_MSG_IFDOWN_BIT, 4359 NETIF_MSG_IFUP_BIT, 4360 NETIF_MSG_RX_ERR_BIT, 4361 NETIF_MSG_TX_ERR_BIT, 4362 NETIF_MSG_TX_QUEUED_BIT, 4363 NETIF_MSG_INTR_BIT, 4364 NETIF_MSG_TX_DONE_BIT, 4365 NETIF_MSG_RX_STATUS_BIT, 4366 NETIF_MSG_PKTDATA_BIT, 4367 NETIF_MSG_HW_BIT, 4368 NETIF_MSG_WOL_BIT, 4369 4370 /* When you add a new bit above, update netif_msg_class_names array 4371 * in net/ethtool/common.c 4372 */ 4373 NETIF_MSG_CLASS_COUNT, 4374 }; 4375 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4376 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4377 4378 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4379 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4380 4381 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4382 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4383 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4384 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4385 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4386 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4387 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4388 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4389 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4390 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4391 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4392 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4393 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4394 #define NETIF_MSG_HW __NETIF_MSG(HW) 4395 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4396 4397 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4398 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4399 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4400 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4401 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4402 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4403 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4404 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4405 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4406 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4407 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4408 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4409 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4410 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4411 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4412 4413 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4414 { 4415 /* use default */ 4416 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4417 return default_msg_enable_bits; 4418 if (debug_value == 0) /* no output */ 4419 return 0; 4420 /* set low N bits */ 4421 return (1U << debug_value) - 1; 4422 } 4423 4424 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4425 { 4426 spin_lock(&txq->_xmit_lock); 4427 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4428 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4429 } 4430 4431 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4432 { 4433 __acquire(&txq->_xmit_lock); 4434 return true; 4435 } 4436 4437 static inline void __netif_tx_release(struct netdev_queue *txq) 4438 { 4439 __release(&txq->_xmit_lock); 4440 } 4441 4442 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4443 { 4444 spin_lock_bh(&txq->_xmit_lock); 4445 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4446 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4447 } 4448 4449 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4450 { 4451 bool ok = spin_trylock(&txq->_xmit_lock); 4452 4453 if (likely(ok)) { 4454 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4455 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4456 } 4457 return ok; 4458 } 4459 4460 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4461 { 4462 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4463 WRITE_ONCE(txq->xmit_lock_owner, -1); 4464 spin_unlock(&txq->_xmit_lock); 4465 } 4466 4467 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4468 { 4469 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4470 WRITE_ONCE(txq->xmit_lock_owner, -1); 4471 spin_unlock_bh(&txq->_xmit_lock); 4472 } 4473 4474 /* 4475 * txq->trans_start can be read locklessly from dev_watchdog() 4476 */ 4477 static inline void txq_trans_update(struct netdev_queue *txq) 4478 { 4479 if (txq->xmit_lock_owner != -1) 4480 WRITE_ONCE(txq->trans_start, jiffies); 4481 } 4482 4483 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4484 { 4485 unsigned long now = jiffies; 4486 4487 if (READ_ONCE(txq->trans_start) != now) 4488 WRITE_ONCE(txq->trans_start, now); 4489 } 4490 4491 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4492 static inline void netif_trans_update(struct net_device *dev) 4493 { 4494 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4495 4496 txq_trans_cond_update(txq); 4497 } 4498 4499 /** 4500 * netif_tx_lock - grab network device transmit lock 4501 * @dev: network device 4502 * 4503 * Get network device transmit lock 4504 */ 4505 void netif_tx_lock(struct net_device *dev); 4506 4507 static inline void netif_tx_lock_bh(struct net_device *dev) 4508 { 4509 local_bh_disable(); 4510 netif_tx_lock(dev); 4511 } 4512 4513 void netif_tx_unlock(struct net_device *dev); 4514 4515 static inline void netif_tx_unlock_bh(struct net_device *dev) 4516 { 4517 netif_tx_unlock(dev); 4518 local_bh_enable(); 4519 } 4520 4521 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4522 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4523 __netif_tx_lock(txq, cpu); \ 4524 } else { \ 4525 __netif_tx_acquire(txq); \ 4526 } \ 4527 } 4528 4529 #define HARD_TX_TRYLOCK(dev, txq) \ 4530 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4531 __netif_tx_trylock(txq) : \ 4532 __netif_tx_acquire(txq)) 4533 4534 #define HARD_TX_UNLOCK(dev, txq) { \ 4535 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4536 __netif_tx_unlock(txq); \ 4537 } else { \ 4538 __netif_tx_release(txq); \ 4539 } \ 4540 } 4541 4542 static inline void netif_tx_disable(struct net_device *dev) 4543 { 4544 unsigned int i; 4545 int cpu; 4546 4547 local_bh_disable(); 4548 cpu = smp_processor_id(); 4549 spin_lock(&dev->tx_global_lock); 4550 for (i = 0; i < dev->num_tx_queues; i++) { 4551 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4552 4553 __netif_tx_lock(txq, cpu); 4554 netif_tx_stop_queue(txq); 4555 __netif_tx_unlock(txq); 4556 } 4557 spin_unlock(&dev->tx_global_lock); 4558 local_bh_enable(); 4559 } 4560 4561 static inline void netif_addr_lock(struct net_device *dev) 4562 { 4563 unsigned char nest_level = 0; 4564 4565 #ifdef CONFIG_LOCKDEP 4566 nest_level = dev->nested_level; 4567 #endif 4568 spin_lock_nested(&dev->addr_list_lock, nest_level); 4569 } 4570 4571 static inline void netif_addr_lock_bh(struct net_device *dev) 4572 { 4573 unsigned char nest_level = 0; 4574 4575 #ifdef CONFIG_LOCKDEP 4576 nest_level = dev->nested_level; 4577 #endif 4578 local_bh_disable(); 4579 spin_lock_nested(&dev->addr_list_lock, nest_level); 4580 } 4581 4582 static inline void netif_addr_unlock(struct net_device *dev) 4583 { 4584 spin_unlock(&dev->addr_list_lock); 4585 } 4586 4587 static inline void netif_addr_unlock_bh(struct net_device *dev) 4588 { 4589 spin_unlock_bh(&dev->addr_list_lock); 4590 } 4591 4592 /* 4593 * dev_addrs walker. Should be used only for read access. Call with 4594 * rcu_read_lock held. 4595 */ 4596 #define for_each_dev_addr(dev, ha) \ 4597 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4598 4599 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4600 4601 void ether_setup(struct net_device *dev); 4602 4603 /* Support for loadable net-drivers */ 4604 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4605 unsigned char name_assign_type, 4606 void (*setup)(struct net_device *), 4607 unsigned int txqs, unsigned int rxqs); 4608 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4609 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4610 4611 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4612 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4613 count) 4614 4615 int register_netdev(struct net_device *dev); 4616 void unregister_netdev(struct net_device *dev); 4617 4618 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4619 4620 /* General hardware address lists handling functions */ 4621 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4622 struct netdev_hw_addr_list *from_list, int addr_len); 4623 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4624 struct netdev_hw_addr_list *from_list, int addr_len); 4625 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4626 struct net_device *dev, 4627 int (*sync)(struct net_device *, const unsigned char *), 4628 int (*unsync)(struct net_device *, 4629 const unsigned char *)); 4630 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4631 struct net_device *dev, 4632 int (*sync)(struct net_device *, 4633 const unsigned char *, int), 4634 int (*unsync)(struct net_device *, 4635 const unsigned char *, int)); 4636 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4637 struct net_device *dev, 4638 int (*unsync)(struct net_device *, 4639 const unsigned char *, int)); 4640 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4641 struct net_device *dev, 4642 int (*unsync)(struct net_device *, 4643 const unsigned char *)); 4644 void __hw_addr_init(struct netdev_hw_addr_list *list); 4645 4646 /* Functions used for device addresses handling */ 4647 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4648 const void *addr, size_t len); 4649 4650 static inline void 4651 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4652 { 4653 dev_addr_mod(dev, 0, addr, len); 4654 } 4655 4656 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4657 { 4658 __dev_addr_set(dev, addr, dev->addr_len); 4659 } 4660 4661 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4662 unsigned char addr_type); 4663 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4664 unsigned char addr_type); 4665 4666 /* Functions used for unicast addresses handling */ 4667 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4668 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4669 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4670 int dev_uc_sync(struct net_device *to, struct net_device *from); 4671 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4672 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4673 void dev_uc_flush(struct net_device *dev); 4674 void dev_uc_init(struct net_device *dev); 4675 4676 /** 4677 * __dev_uc_sync - Synchonize device's unicast list 4678 * @dev: device to sync 4679 * @sync: function to call if address should be added 4680 * @unsync: function to call if address should be removed 4681 * 4682 * Add newly added addresses to the interface, and release 4683 * addresses that have been deleted. 4684 */ 4685 static inline int __dev_uc_sync(struct net_device *dev, 4686 int (*sync)(struct net_device *, 4687 const unsigned char *), 4688 int (*unsync)(struct net_device *, 4689 const unsigned char *)) 4690 { 4691 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4692 } 4693 4694 /** 4695 * __dev_uc_unsync - Remove synchronized addresses from device 4696 * @dev: device to sync 4697 * @unsync: function to call if address should be removed 4698 * 4699 * Remove all addresses that were added to the device by dev_uc_sync(). 4700 */ 4701 static inline void __dev_uc_unsync(struct net_device *dev, 4702 int (*unsync)(struct net_device *, 4703 const unsigned char *)) 4704 { 4705 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4706 } 4707 4708 /* Functions used for multicast addresses handling */ 4709 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4710 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4711 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4712 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4713 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4714 int dev_mc_sync(struct net_device *to, struct net_device *from); 4715 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4716 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4717 void dev_mc_flush(struct net_device *dev); 4718 void dev_mc_init(struct net_device *dev); 4719 4720 /** 4721 * __dev_mc_sync - Synchonize device's multicast list 4722 * @dev: device to sync 4723 * @sync: function to call if address should be added 4724 * @unsync: function to call if address should be removed 4725 * 4726 * Add newly added addresses to the interface, and release 4727 * addresses that have been deleted. 4728 */ 4729 static inline int __dev_mc_sync(struct net_device *dev, 4730 int (*sync)(struct net_device *, 4731 const unsigned char *), 4732 int (*unsync)(struct net_device *, 4733 const unsigned char *)) 4734 { 4735 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4736 } 4737 4738 /** 4739 * __dev_mc_unsync - Remove synchronized addresses from device 4740 * @dev: device to sync 4741 * @unsync: function to call if address should be removed 4742 * 4743 * Remove all addresses that were added to the device by dev_mc_sync(). 4744 */ 4745 static inline void __dev_mc_unsync(struct net_device *dev, 4746 int (*unsync)(struct net_device *, 4747 const unsigned char *)) 4748 { 4749 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4750 } 4751 4752 /* Functions used for secondary unicast and multicast support */ 4753 void dev_set_rx_mode(struct net_device *dev); 4754 int dev_set_promiscuity(struct net_device *dev, int inc); 4755 int dev_set_allmulti(struct net_device *dev, int inc); 4756 void netdev_state_change(struct net_device *dev); 4757 void __netdev_notify_peers(struct net_device *dev); 4758 void netdev_notify_peers(struct net_device *dev); 4759 void netdev_features_change(struct net_device *dev); 4760 /* Load a device via the kmod */ 4761 void dev_load(struct net *net, const char *name); 4762 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4763 struct rtnl_link_stats64 *storage); 4764 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4765 const struct net_device_stats *netdev_stats); 4766 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4767 const struct pcpu_sw_netstats __percpu *netstats); 4768 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4769 4770 extern int netdev_max_backlog; 4771 extern int dev_rx_weight; 4772 extern int dev_tx_weight; 4773 extern int gro_normal_batch; 4774 4775 enum { 4776 NESTED_SYNC_IMM_BIT, 4777 NESTED_SYNC_TODO_BIT, 4778 }; 4779 4780 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4781 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4782 4783 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4784 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4785 4786 struct netdev_nested_priv { 4787 unsigned char flags; 4788 void *data; 4789 }; 4790 4791 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4792 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4793 struct list_head **iter); 4794 4795 /* iterate through upper list, must be called under RCU read lock */ 4796 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4797 for (iter = &(dev)->adj_list.upper, \ 4798 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4799 updev; \ 4800 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4801 4802 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4803 int (*fn)(struct net_device *upper_dev, 4804 struct netdev_nested_priv *priv), 4805 struct netdev_nested_priv *priv); 4806 4807 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4808 struct net_device *upper_dev); 4809 4810 bool netdev_has_any_upper_dev(struct net_device *dev); 4811 4812 void *netdev_lower_get_next_private(struct net_device *dev, 4813 struct list_head **iter); 4814 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4815 struct list_head **iter); 4816 4817 #define netdev_for_each_lower_private(dev, priv, iter) \ 4818 for (iter = (dev)->adj_list.lower.next, \ 4819 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4820 priv; \ 4821 priv = netdev_lower_get_next_private(dev, &(iter))) 4822 4823 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4824 for (iter = &(dev)->adj_list.lower, \ 4825 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4826 priv; \ 4827 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4828 4829 void *netdev_lower_get_next(struct net_device *dev, 4830 struct list_head **iter); 4831 4832 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4833 for (iter = (dev)->adj_list.lower.next, \ 4834 ldev = netdev_lower_get_next(dev, &(iter)); \ 4835 ldev; \ 4836 ldev = netdev_lower_get_next(dev, &(iter))) 4837 4838 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4839 struct list_head **iter); 4840 int netdev_walk_all_lower_dev(struct net_device *dev, 4841 int (*fn)(struct net_device *lower_dev, 4842 struct netdev_nested_priv *priv), 4843 struct netdev_nested_priv *priv); 4844 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4845 int (*fn)(struct net_device *lower_dev, 4846 struct netdev_nested_priv *priv), 4847 struct netdev_nested_priv *priv); 4848 4849 void *netdev_adjacent_get_private(struct list_head *adj_list); 4850 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4851 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4852 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4853 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4854 struct netlink_ext_ack *extack); 4855 int netdev_master_upper_dev_link(struct net_device *dev, 4856 struct net_device *upper_dev, 4857 void *upper_priv, void *upper_info, 4858 struct netlink_ext_ack *extack); 4859 void netdev_upper_dev_unlink(struct net_device *dev, 4860 struct net_device *upper_dev); 4861 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4862 struct net_device *new_dev, 4863 struct net_device *dev, 4864 struct netlink_ext_ack *extack); 4865 void netdev_adjacent_change_commit(struct net_device *old_dev, 4866 struct net_device *new_dev, 4867 struct net_device *dev); 4868 void netdev_adjacent_change_abort(struct net_device *old_dev, 4869 struct net_device *new_dev, 4870 struct net_device *dev); 4871 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4872 void *netdev_lower_dev_get_private(struct net_device *dev, 4873 struct net_device *lower_dev); 4874 void netdev_lower_state_changed(struct net_device *lower_dev, 4875 void *lower_state_info); 4876 4877 /* RSS keys are 40 or 52 bytes long */ 4878 #define NETDEV_RSS_KEY_LEN 52 4879 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4880 void netdev_rss_key_fill(void *buffer, size_t len); 4881 4882 int skb_checksum_help(struct sk_buff *skb); 4883 int skb_crc32c_csum_help(struct sk_buff *skb); 4884 int skb_csum_hwoffload_help(struct sk_buff *skb, 4885 const netdev_features_t features); 4886 4887 struct netdev_bonding_info { 4888 ifslave slave; 4889 ifbond master; 4890 }; 4891 4892 struct netdev_notifier_bonding_info { 4893 struct netdev_notifier_info info; /* must be first */ 4894 struct netdev_bonding_info bonding_info; 4895 }; 4896 4897 void netdev_bonding_info_change(struct net_device *dev, 4898 struct netdev_bonding_info *bonding_info); 4899 4900 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4901 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4902 #else 4903 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4904 const void *data) 4905 { 4906 } 4907 #endif 4908 4909 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4910 4911 static inline bool can_checksum_protocol(netdev_features_t features, 4912 __be16 protocol) 4913 { 4914 if (protocol == htons(ETH_P_FCOE)) 4915 return !!(features & NETIF_F_FCOE_CRC); 4916 4917 /* Assume this is an IP checksum (not SCTP CRC) */ 4918 4919 if (features & NETIF_F_HW_CSUM) { 4920 /* Can checksum everything */ 4921 return true; 4922 } 4923 4924 switch (protocol) { 4925 case htons(ETH_P_IP): 4926 return !!(features & NETIF_F_IP_CSUM); 4927 case htons(ETH_P_IPV6): 4928 return !!(features & NETIF_F_IPV6_CSUM); 4929 default: 4930 return false; 4931 } 4932 } 4933 4934 #ifdef CONFIG_BUG 4935 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4936 #else 4937 static inline void netdev_rx_csum_fault(struct net_device *dev, 4938 struct sk_buff *skb) 4939 { 4940 } 4941 #endif 4942 /* rx skb timestamps */ 4943 void net_enable_timestamp(void); 4944 void net_disable_timestamp(void); 4945 4946 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 4947 const struct skb_shared_hwtstamps *hwtstamps, 4948 bool cycles) 4949 { 4950 const struct net_device_ops *ops = dev->netdev_ops; 4951 4952 if (ops->ndo_get_tstamp) 4953 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 4954 4955 return hwtstamps->hwtstamp; 4956 } 4957 4958 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4959 struct sk_buff *skb, struct net_device *dev, 4960 bool more) 4961 { 4962 __this_cpu_write(softnet_data.xmit.more, more); 4963 return ops->ndo_start_xmit(skb, dev); 4964 } 4965 4966 static inline bool netdev_xmit_more(void) 4967 { 4968 return __this_cpu_read(softnet_data.xmit.more); 4969 } 4970 4971 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4972 struct netdev_queue *txq, bool more) 4973 { 4974 const struct net_device_ops *ops = dev->netdev_ops; 4975 netdev_tx_t rc; 4976 4977 rc = __netdev_start_xmit(ops, skb, dev, more); 4978 if (rc == NETDEV_TX_OK) 4979 txq_trans_update(txq); 4980 4981 return rc; 4982 } 4983 4984 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4985 const void *ns); 4986 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4987 const void *ns); 4988 4989 extern const struct kobj_ns_type_operations net_ns_type_operations; 4990 4991 const char *netdev_drivername(const struct net_device *dev); 4992 4993 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4994 netdev_features_t f2) 4995 { 4996 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4997 if (f1 & NETIF_F_HW_CSUM) 4998 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4999 else 5000 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5001 } 5002 5003 return f1 & f2; 5004 } 5005 5006 static inline netdev_features_t netdev_get_wanted_features( 5007 struct net_device *dev) 5008 { 5009 return (dev->features & ~dev->hw_features) | dev->wanted_features; 5010 } 5011 netdev_features_t netdev_increment_features(netdev_features_t all, 5012 netdev_features_t one, netdev_features_t mask); 5013 5014 /* Allow TSO being used on stacked device : 5015 * Performing the GSO segmentation before last device 5016 * is a performance improvement. 5017 */ 5018 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 5019 netdev_features_t mask) 5020 { 5021 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 5022 } 5023 5024 int __netdev_update_features(struct net_device *dev); 5025 void netdev_update_features(struct net_device *dev); 5026 void netdev_change_features(struct net_device *dev); 5027 5028 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 5029 struct net_device *dev); 5030 5031 netdev_features_t passthru_features_check(struct sk_buff *skb, 5032 struct net_device *dev, 5033 netdev_features_t features); 5034 netdev_features_t netif_skb_features(struct sk_buff *skb); 5035 void skb_warn_bad_offload(const struct sk_buff *skb); 5036 5037 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 5038 { 5039 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 5040 5041 /* check flags correspondence */ 5042 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 5043 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 5044 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 5045 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 5046 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 5047 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 5048 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 5049 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 5050 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 5051 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 5052 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 5053 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 5054 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 5055 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 5056 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 5057 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 5058 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 5059 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 5060 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 5061 5062 return (features & feature) == feature; 5063 } 5064 5065 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 5066 { 5067 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 5068 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 5069 } 5070 5071 static inline bool netif_needs_gso(struct sk_buff *skb, 5072 netdev_features_t features) 5073 { 5074 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 5075 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 5076 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 5077 } 5078 5079 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 5080 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 5081 void netif_inherit_tso_max(struct net_device *to, 5082 const struct net_device *from); 5083 5084 static inline bool netif_is_macsec(const struct net_device *dev) 5085 { 5086 return dev->priv_flags & IFF_MACSEC; 5087 } 5088 5089 static inline bool netif_is_macvlan(const struct net_device *dev) 5090 { 5091 return dev->priv_flags & IFF_MACVLAN; 5092 } 5093 5094 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5095 { 5096 return dev->priv_flags & IFF_MACVLAN_PORT; 5097 } 5098 5099 static inline bool netif_is_bond_master(const struct net_device *dev) 5100 { 5101 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5102 } 5103 5104 static inline bool netif_is_bond_slave(const struct net_device *dev) 5105 { 5106 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5107 } 5108 5109 static inline bool netif_supports_nofcs(struct net_device *dev) 5110 { 5111 return dev->priv_flags & IFF_SUPP_NOFCS; 5112 } 5113 5114 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5115 { 5116 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5117 } 5118 5119 static inline bool netif_is_l3_master(const struct net_device *dev) 5120 { 5121 return dev->priv_flags & IFF_L3MDEV_MASTER; 5122 } 5123 5124 static inline bool netif_is_l3_slave(const struct net_device *dev) 5125 { 5126 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5127 } 5128 5129 static inline int dev_sdif(const struct net_device *dev) 5130 { 5131 #ifdef CONFIG_NET_L3_MASTER_DEV 5132 if (netif_is_l3_slave(dev)) 5133 return dev->ifindex; 5134 #endif 5135 return 0; 5136 } 5137 5138 static inline bool netif_is_bridge_master(const struct net_device *dev) 5139 { 5140 return dev->priv_flags & IFF_EBRIDGE; 5141 } 5142 5143 static inline bool netif_is_bridge_port(const struct net_device *dev) 5144 { 5145 return dev->priv_flags & IFF_BRIDGE_PORT; 5146 } 5147 5148 static inline bool netif_is_ovs_master(const struct net_device *dev) 5149 { 5150 return dev->priv_flags & IFF_OPENVSWITCH; 5151 } 5152 5153 static inline bool netif_is_ovs_port(const struct net_device *dev) 5154 { 5155 return dev->priv_flags & IFF_OVS_DATAPATH; 5156 } 5157 5158 static inline bool netif_is_any_bridge_master(const struct net_device *dev) 5159 { 5160 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev); 5161 } 5162 5163 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5164 { 5165 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5166 } 5167 5168 static inline bool netif_is_team_master(const struct net_device *dev) 5169 { 5170 return dev->priv_flags & IFF_TEAM; 5171 } 5172 5173 static inline bool netif_is_team_port(const struct net_device *dev) 5174 { 5175 return dev->priv_flags & IFF_TEAM_PORT; 5176 } 5177 5178 static inline bool netif_is_lag_master(const struct net_device *dev) 5179 { 5180 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5181 } 5182 5183 static inline bool netif_is_lag_port(const struct net_device *dev) 5184 { 5185 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5186 } 5187 5188 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5189 { 5190 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5191 } 5192 5193 static inline bool netif_is_failover(const struct net_device *dev) 5194 { 5195 return dev->priv_flags & IFF_FAILOVER; 5196 } 5197 5198 static inline bool netif_is_failover_slave(const struct net_device *dev) 5199 { 5200 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5201 } 5202 5203 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5204 static inline void netif_keep_dst(struct net_device *dev) 5205 { 5206 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5207 } 5208 5209 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5210 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5211 { 5212 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5213 return netif_is_macsec(dev); 5214 } 5215 5216 extern struct pernet_operations __net_initdata loopback_net_ops; 5217 5218 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5219 5220 /* netdev_printk helpers, similar to dev_printk */ 5221 5222 static inline const char *netdev_name(const struct net_device *dev) 5223 { 5224 if (!dev->name[0] || strchr(dev->name, '%')) 5225 return "(unnamed net_device)"; 5226 return dev->name; 5227 } 5228 5229 static inline const char *netdev_reg_state(const struct net_device *dev) 5230 { 5231 switch (dev->reg_state) { 5232 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5233 case NETREG_REGISTERED: return ""; 5234 case NETREG_UNREGISTERING: return " (unregistering)"; 5235 case NETREG_UNREGISTERED: return " (unregistered)"; 5236 case NETREG_RELEASED: return " (released)"; 5237 case NETREG_DUMMY: return " (dummy)"; 5238 } 5239 5240 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5241 return " (unknown)"; 5242 } 5243 5244 #define MODULE_ALIAS_NETDEV(device) \ 5245 MODULE_ALIAS("netdev-" device) 5246 5247 /* 5248 * netdev_WARN() acts like dev_printk(), but with the key difference 5249 * of using a WARN/WARN_ON to get the message out, including the 5250 * file/line information and a backtrace. 5251 */ 5252 #define netdev_WARN(dev, format, args...) \ 5253 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5254 netdev_reg_state(dev), ##args) 5255 5256 #define netdev_WARN_ONCE(dev, format, args...) \ 5257 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5258 netdev_reg_state(dev), ##args) 5259 5260 /* 5261 * The list of packet types we will receive (as opposed to discard) 5262 * and the routines to invoke. 5263 * 5264 * Why 16. Because with 16 the only overlap we get on a hash of the 5265 * low nibble of the protocol value is RARP/SNAP/X.25. 5266 * 5267 * 0800 IP 5268 * 0001 802.3 5269 * 0002 AX.25 5270 * 0004 802.2 5271 * 8035 RARP 5272 * 0005 SNAP 5273 * 0805 X.25 5274 * 0806 ARP 5275 * 8137 IPX 5276 * 0009 Localtalk 5277 * 86DD IPv6 5278 */ 5279 #define PTYPE_HASH_SIZE (16) 5280 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5281 5282 extern struct list_head ptype_all __read_mostly; 5283 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5284 5285 extern struct net_device *blackhole_netdev; 5286 5287 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */ 5288 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD) 5289 #define DEV_STATS_ADD(DEV, FIELD, VAL) \ 5290 atomic_long_add((VAL), &(DEV)->stats.__##FIELD) 5291 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD) 5292 5293 #endif /* _LINUX_NETDEVICE_H */ 5294