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 /* Cacheline organization can be found documented in 2101 * Documentation/networking/net_cachelines/net_device.rst. 2102 * Please update the document when adding new fields. 2103 */ 2104 2105 /* TX read-mostly hotpath */ 2106 __cacheline_group_begin(net_device_read_tx); 2107 unsigned long long priv_flags; 2108 const struct net_device_ops *netdev_ops; 2109 const struct header_ops *header_ops; 2110 struct netdev_queue *_tx; 2111 unsigned int real_num_tx_queues; 2112 unsigned int gso_max_size; 2113 unsigned int gso_ipv4_max_size; 2114 u16 gso_max_segs; 2115 s16 num_tc; 2116 /* Note : dev->mtu is often read without holding a lock. 2117 * Writers usually hold RTNL. 2118 * It is recommended to use READ_ONCE() to annotate the reads, 2119 * and to use WRITE_ONCE() to annotate the writes. 2120 */ 2121 unsigned int mtu; 2122 unsigned short needed_headroom; 2123 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2124 #ifdef CONFIG_XPS 2125 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2126 #endif 2127 #ifdef CONFIG_NETFILTER_EGRESS 2128 struct nf_hook_entries __rcu *nf_hooks_egress; 2129 #endif 2130 #ifdef CONFIG_NET_XGRESS 2131 struct bpf_mprog_entry __rcu *tcx_egress; 2132 #endif 2133 __cacheline_group_end(net_device_read_tx); 2134 2135 /* TXRX read-mostly hotpath */ 2136 __cacheline_group_begin(net_device_read_txrx); 2137 unsigned int flags; 2138 unsigned short hard_header_len; 2139 netdev_features_t features; 2140 struct inet6_dev __rcu *ip6_ptr; 2141 __cacheline_group_end(net_device_read_txrx); 2142 2143 /* RX read-mostly hotpath */ 2144 __cacheline_group_begin(net_device_read_rx); 2145 struct list_head ptype_specific; 2146 int ifindex; 2147 unsigned int real_num_rx_queues; 2148 struct netdev_rx_queue *_rx; 2149 unsigned long gro_flush_timeout; 2150 int napi_defer_hard_irqs; 2151 unsigned int gro_max_size; 2152 unsigned int gro_ipv4_max_size; 2153 rx_handler_func_t __rcu *rx_handler; 2154 void __rcu *rx_handler_data; 2155 possible_net_t nd_net; 2156 #ifdef CONFIG_NETPOLL 2157 struct netpoll_info __rcu *npinfo; 2158 #endif 2159 #ifdef CONFIG_NET_XGRESS 2160 struct bpf_mprog_entry __rcu *tcx_ingress; 2161 #endif 2162 __cacheline_group_end(net_device_read_rx); 2163 2164 char name[IFNAMSIZ]; 2165 struct netdev_name_node *name_node; 2166 struct dev_ifalias __rcu *ifalias; 2167 /* 2168 * I/O specific fields 2169 * FIXME: Merge these and struct ifmap into one 2170 */ 2171 unsigned long mem_end; 2172 unsigned long mem_start; 2173 unsigned long base_addr; 2174 2175 /* 2176 * Some hardware also needs these fields (state,dev_list, 2177 * napi_list,unreg_list,close_list) but they are not 2178 * part of the usual set specified in Space.c. 2179 */ 2180 2181 unsigned long state; 2182 2183 struct list_head dev_list; 2184 struct list_head napi_list; 2185 struct list_head unreg_list; 2186 struct list_head close_list; 2187 struct list_head ptype_all; 2188 2189 struct { 2190 struct list_head upper; 2191 struct list_head lower; 2192 } adj_list; 2193 2194 /* Read-mostly cache-line for fast-path access */ 2195 xdp_features_t xdp_features; 2196 const struct xdp_metadata_ops *xdp_metadata_ops; 2197 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops; 2198 unsigned short gflags; 2199 2200 unsigned short needed_tailroom; 2201 2202 netdev_features_t hw_features; 2203 netdev_features_t wanted_features; 2204 netdev_features_t vlan_features; 2205 netdev_features_t hw_enc_features; 2206 netdev_features_t mpls_features; 2207 netdev_features_t gso_partial_features; 2208 2209 unsigned int min_mtu; 2210 unsigned int max_mtu; 2211 unsigned short type; 2212 unsigned char min_header_len; 2213 unsigned char name_assign_type; 2214 2215 int group; 2216 2217 struct net_device_stats stats; /* not used by modern drivers */ 2218 2219 struct net_device_core_stats __percpu *core_stats; 2220 2221 /* Stats to monitor link on/off, flapping */ 2222 atomic_t carrier_up_count; 2223 atomic_t carrier_down_count; 2224 2225 #ifdef CONFIG_WIRELESS_EXT 2226 const struct iw_handler_def *wireless_handlers; 2227 struct iw_public_data *wireless_data; 2228 #endif 2229 const struct ethtool_ops *ethtool_ops; 2230 #ifdef CONFIG_NET_L3_MASTER_DEV 2231 const struct l3mdev_ops *l3mdev_ops; 2232 #endif 2233 #if IS_ENABLED(CONFIG_IPV6) 2234 const struct ndisc_ops *ndisc_ops; 2235 #endif 2236 2237 #ifdef CONFIG_XFRM_OFFLOAD 2238 const struct xfrmdev_ops *xfrmdev_ops; 2239 #endif 2240 2241 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2242 const struct tlsdev_ops *tlsdev_ops; 2243 #endif 2244 2245 unsigned char operstate; 2246 unsigned char link_mode; 2247 2248 unsigned char if_port; 2249 unsigned char dma; 2250 2251 /* Interface address info. */ 2252 unsigned char perm_addr[MAX_ADDR_LEN]; 2253 unsigned char addr_assign_type; 2254 unsigned char addr_len; 2255 unsigned char upper_level; 2256 unsigned char lower_level; 2257 2258 unsigned short neigh_priv_len; 2259 unsigned short dev_id; 2260 unsigned short dev_port; 2261 unsigned short padded; 2262 2263 spinlock_t addr_list_lock; 2264 int irq; 2265 2266 struct netdev_hw_addr_list uc; 2267 struct netdev_hw_addr_list mc; 2268 struct netdev_hw_addr_list dev_addrs; 2269 2270 #ifdef CONFIG_SYSFS 2271 struct kset *queues_kset; 2272 #endif 2273 #ifdef CONFIG_LOCKDEP 2274 struct list_head unlink_list; 2275 #endif 2276 unsigned int promiscuity; 2277 unsigned int allmulti; 2278 bool uc_promisc; 2279 #ifdef CONFIG_LOCKDEP 2280 unsigned char nested_level; 2281 #endif 2282 2283 2284 /* Protocol-specific pointers */ 2285 struct in_device __rcu *ip_ptr; 2286 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2287 struct vlan_info __rcu *vlan_info; 2288 #endif 2289 #if IS_ENABLED(CONFIG_NET_DSA) 2290 struct dsa_port *dsa_ptr; 2291 #endif 2292 #if IS_ENABLED(CONFIG_TIPC) 2293 struct tipc_bearer __rcu *tipc_ptr; 2294 #endif 2295 #if IS_ENABLED(CONFIG_ATALK) 2296 void *atalk_ptr; 2297 #endif 2298 #if IS_ENABLED(CONFIG_AX25) 2299 void *ax25_ptr; 2300 #endif 2301 #if IS_ENABLED(CONFIG_CFG80211) 2302 struct wireless_dev *ieee80211_ptr; 2303 #endif 2304 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2305 struct wpan_dev *ieee802154_ptr; 2306 #endif 2307 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2308 struct mpls_dev __rcu *mpls_ptr; 2309 #endif 2310 #if IS_ENABLED(CONFIG_MCTP) 2311 struct mctp_dev __rcu *mctp_ptr; 2312 #endif 2313 2314 /* 2315 * Cache lines mostly used on receive path (including eth_type_trans()) 2316 */ 2317 /* Interface address info used in eth_type_trans() */ 2318 const unsigned char *dev_addr; 2319 2320 unsigned int num_rx_queues; 2321 struct bpf_prog __rcu *xdp_prog; 2322 #define GRO_LEGACY_MAX_SIZE 65536u 2323 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2324 * and shinfo->gso_segs is a 16bit field. 2325 */ 2326 #define GRO_MAX_SIZE (8 * 65535u) 2327 unsigned int xdp_zc_max_segs; 2328 struct netdev_queue __rcu *ingress_queue; 2329 #ifdef CONFIG_NETFILTER_INGRESS 2330 struct nf_hook_entries __rcu *nf_hooks_ingress; 2331 #endif 2332 2333 unsigned char broadcast[MAX_ADDR_LEN]; 2334 #ifdef CONFIG_RFS_ACCEL 2335 struct cpu_rmap *rx_cpu_rmap; 2336 #endif 2337 struct hlist_node index_hlist; 2338 2339 /* 2340 * Cache lines mostly used on transmit path 2341 */ 2342 unsigned int num_tx_queues; 2343 struct Qdisc __rcu *qdisc; 2344 unsigned int tx_queue_len; 2345 spinlock_t tx_global_lock; 2346 2347 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2348 2349 #ifdef CONFIG_NET_SCHED 2350 DECLARE_HASHTABLE (qdisc_hash, 4); 2351 #endif 2352 /* These may be needed for future network-power-down code. */ 2353 struct timer_list watchdog_timer; 2354 int watchdog_timeo; 2355 2356 u32 proto_down_reason; 2357 2358 struct list_head todo_list; 2359 2360 #ifdef CONFIG_PCPU_DEV_REFCNT 2361 int __percpu *pcpu_refcnt; 2362 #else 2363 refcount_t dev_refcnt; 2364 #endif 2365 struct ref_tracker_dir refcnt_tracker; 2366 2367 struct list_head link_watch_list; 2368 2369 enum { NETREG_UNINITIALIZED=0, 2370 NETREG_REGISTERED, /* completed register_netdevice */ 2371 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2372 NETREG_UNREGISTERED, /* completed unregister todo */ 2373 NETREG_RELEASED, /* called free_netdev */ 2374 NETREG_DUMMY, /* dummy device for NAPI poll */ 2375 } reg_state:8; 2376 2377 bool dismantle; 2378 2379 enum { 2380 RTNL_LINK_INITIALIZED, 2381 RTNL_LINK_INITIALIZING, 2382 } rtnl_link_state:16; 2383 2384 bool needs_free_netdev; 2385 void (*priv_destructor)(struct net_device *dev); 2386 2387 /* mid-layer private */ 2388 void *ml_priv; 2389 enum netdev_ml_priv_type ml_priv_type; 2390 2391 enum netdev_stat_type pcpu_stat_type:8; 2392 union { 2393 struct pcpu_lstats __percpu *lstats; 2394 struct pcpu_sw_netstats __percpu *tstats; 2395 struct pcpu_dstats __percpu *dstats; 2396 }; 2397 2398 #if IS_ENABLED(CONFIG_GARP) 2399 struct garp_port __rcu *garp_port; 2400 #endif 2401 #if IS_ENABLED(CONFIG_MRP) 2402 struct mrp_port __rcu *mrp_port; 2403 #endif 2404 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2405 struct dm_hw_stat_delta __rcu *dm_private; 2406 #endif 2407 struct device dev; 2408 const struct attribute_group *sysfs_groups[4]; 2409 const struct attribute_group *sysfs_rx_queue_group; 2410 2411 const struct rtnl_link_ops *rtnl_link_ops; 2412 2413 /* for setting kernel sock attribute on TCP connection setup */ 2414 #define GSO_MAX_SEGS 65535u 2415 #define GSO_LEGACY_MAX_SIZE 65536u 2416 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2417 * and shinfo->gso_segs is a 16bit field. 2418 */ 2419 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2420 2421 #define TSO_LEGACY_MAX_SIZE 65536 2422 #define TSO_MAX_SIZE UINT_MAX 2423 unsigned int tso_max_size; 2424 #define TSO_MAX_SEGS U16_MAX 2425 u16 tso_max_segs; 2426 2427 #ifdef CONFIG_DCB 2428 const struct dcbnl_rtnl_ops *dcbnl_ops; 2429 #endif 2430 u8 prio_tc_map[TC_BITMASK + 1]; 2431 2432 #if IS_ENABLED(CONFIG_FCOE) 2433 unsigned int fcoe_ddp_xid; 2434 #endif 2435 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2436 struct netprio_map __rcu *priomap; 2437 #endif 2438 struct phy_device *phydev; 2439 struct sfp_bus *sfp_bus; 2440 struct lock_class_key *qdisc_tx_busylock; 2441 bool proto_down; 2442 unsigned wol_enabled:1; 2443 unsigned threaded:1; 2444 2445 struct list_head net_notifier_list; 2446 2447 #if IS_ENABLED(CONFIG_MACSEC) 2448 /* MACsec management functions */ 2449 const struct macsec_ops *macsec_ops; 2450 #endif 2451 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2452 struct udp_tunnel_nic *udp_tunnel_nic; 2453 2454 /* protected by rtnl_lock */ 2455 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2456 2457 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2458 netdevice_tracker linkwatch_dev_tracker; 2459 netdevice_tracker watchdog_dev_tracker; 2460 netdevice_tracker dev_registered_tracker; 2461 struct rtnl_hw_stats64 *offload_xstats_l3; 2462 2463 struct devlink_port *devlink_port; 2464 2465 #if IS_ENABLED(CONFIG_DPLL) 2466 struct dpll_pin *dpll_pin; 2467 #endif 2468 #if IS_ENABLED(CONFIG_PAGE_POOL) 2469 /** @page_pools: page pools created for this netdevice */ 2470 struct hlist_head page_pools; 2471 #endif 2472 }; 2473 #define to_net_dev(d) container_of(d, struct net_device, dev) 2474 2475 /* 2476 * Driver should use this to assign devlink port instance to a netdevice 2477 * before it registers the netdevice. Therefore devlink_port is static 2478 * during the netdev lifetime after it is registered. 2479 */ 2480 #define SET_NETDEV_DEVLINK_PORT(dev, port) \ 2481 ({ \ 2482 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \ 2483 ((dev)->devlink_port = (port)); \ 2484 }) 2485 2486 static inline bool netif_elide_gro(const struct net_device *dev) 2487 { 2488 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2489 return true; 2490 return false; 2491 } 2492 2493 #define NETDEV_ALIGN 32 2494 2495 static inline 2496 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2497 { 2498 return dev->prio_tc_map[prio & TC_BITMASK]; 2499 } 2500 2501 static inline 2502 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2503 { 2504 if (tc >= dev->num_tc) 2505 return -EINVAL; 2506 2507 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2508 return 0; 2509 } 2510 2511 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2512 void netdev_reset_tc(struct net_device *dev); 2513 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2514 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2515 2516 static inline 2517 int netdev_get_num_tc(struct net_device *dev) 2518 { 2519 return dev->num_tc; 2520 } 2521 2522 static inline void net_prefetch(void *p) 2523 { 2524 prefetch(p); 2525 #if L1_CACHE_BYTES < 128 2526 prefetch((u8 *)p + L1_CACHE_BYTES); 2527 #endif 2528 } 2529 2530 static inline void net_prefetchw(void *p) 2531 { 2532 prefetchw(p); 2533 #if L1_CACHE_BYTES < 128 2534 prefetchw((u8 *)p + L1_CACHE_BYTES); 2535 #endif 2536 } 2537 2538 void netdev_unbind_sb_channel(struct net_device *dev, 2539 struct net_device *sb_dev); 2540 int netdev_bind_sb_channel_queue(struct net_device *dev, 2541 struct net_device *sb_dev, 2542 u8 tc, u16 count, u16 offset); 2543 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2544 static inline int netdev_get_sb_channel(struct net_device *dev) 2545 { 2546 return max_t(int, -dev->num_tc, 0); 2547 } 2548 2549 static inline 2550 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2551 unsigned int index) 2552 { 2553 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues); 2554 return &dev->_tx[index]; 2555 } 2556 2557 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2558 const struct sk_buff *skb) 2559 { 2560 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2561 } 2562 2563 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2564 void (*f)(struct net_device *, 2565 struct netdev_queue *, 2566 void *), 2567 void *arg) 2568 { 2569 unsigned int i; 2570 2571 for (i = 0; i < dev->num_tx_queues; i++) 2572 f(dev, &dev->_tx[i], arg); 2573 } 2574 2575 #define netdev_lockdep_set_classes(dev) \ 2576 { \ 2577 static struct lock_class_key qdisc_tx_busylock_key; \ 2578 static struct lock_class_key qdisc_xmit_lock_key; \ 2579 static struct lock_class_key dev_addr_list_lock_key; \ 2580 unsigned int i; \ 2581 \ 2582 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2583 lockdep_set_class(&(dev)->addr_list_lock, \ 2584 &dev_addr_list_lock_key); \ 2585 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2586 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2587 &qdisc_xmit_lock_key); \ 2588 } 2589 2590 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2591 struct net_device *sb_dev); 2592 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2593 struct sk_buff *skb, 2594 struct net_device *sb_dev); 2595 2596 /* returns the headroom that the master device needs to take in account 2597 * when forwarding to this dev 2598 */ 2599 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2600 { 2601 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2602 } 2603 2604 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2605 { 2606 if (dev->netdev_ops->ndo_set_rx_headroom) 2607 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2608 } 2609 2610 /* set the device rx headroom to the dev's default */ 2611 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2612 { 2613 netdev_set_rx_headroom(dev, -1); 2614 } 2615 2616 static inline void *netdev_get_ml_priv(struct net_device *dev, 2617 enum netdev_ml_priv_type type) 2618 { 2619 if (dev->ml_priv_type != type) 2620 return NULL; 2621 2622 return dev->ml_priv; 2623 } 2624 2625 static inline void netdev_set_ml_priv(struct net_device *dev, 2626 void *ml_priv, 2627 enum netdev_ml_priv_type type) 2628 { 2629 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2630 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2631 dev->ml_priv_type, type); 2632 WARN(!dev->ml_priv_type && dev->ml_priv, 2633 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2634 2635 dev->ml_priv = ml_priv; 2636 dev->ml_priv_type = type; 2637 } 2638 2639 /* 2640 * Net namespace inlines 2641 */ 2642 static inline 2643 struct net *dev_net(const struct net_device *dev) 2644 { 2645 return read_pnet(&dev->nd_net); 2646 } 2647 2648 static inline 2649 void dev_net_set(struct net_device *dev, struct net *net) 2650 { 2651 write_pnet(&dev->nd_net, net); 2652 } 2653 2654 /** 2655 * netdev_priv - access network device private data 2656 * @dev: network device 2657 * 2658 * Get network device private data 2659 */ 2660 static inline void *netdev_priv(const struct net_device *dev) 2661 { 2662 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2663 } 2664 2665 /* Set the sysfs physical device reference for the network logical device 2666 * if set prior to registration will cause a symlink during initialization. 2667 */ 2668 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2669 2670 /* Set the sysfs device type for the network logical device to allow 2671 * fine-grained identification of different network device types. For 2672 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2673 */ 2674 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2675 2676 void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index, 2677 enum netdev_queue_type type, 2678 struct napi_struct *napi); 2679 2680 static inline void netif_napi_set_irq(struct napi_struct *napi, int irq) 2681 { 2682 napi->irq = irq; 2683 } 2684 2685 /* Default NAPI poll() weight 2686 * Device drivers are strongly advised to not use bigger value 2687 */ 2688 #define NAPI_POLL_WEIGHT 64 2689 2690 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2691 int (*poll)(struct napi_struct *, int), int weight); 2692 2693 /** 2694 * netif_napi_add() - initialize a NAPI context 2695 * @dev: network device 2696 * @napi: NAPI context 2697 * @poll: polling function 2698 * 2699 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2700 * *any* of the other NAPI-related functions. 2701 */ 2702 static inline void 2703 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2704 int (*poll)(struct napi_struct *, int)) 2705 { 2706 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2707 } 2708 2709 static inline void 2710 netif_napi_add_tx_weight(struct net_device *dev, 2711 struct napi_struct *napi, 2712 int (*poll)(struct napi_struct *, int), 2713 int weight) 2714 { 2715 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2716 netif_napi_add_weight(dev, napi, poll, weight); 2717 } 2718 2719 /** 2720 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2721 * @dev: network device 2722 * @napi: NAPI context 2723 * @poll: polling function 2724 * 2725 * This variant of netif_napi_add() should be used from drivers using NAPI 2726 * to exclusively poll a TX queue. 2727 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2728 */ 2729 static inline void netif_napi_add_tx(struct net_device *dev, 2730 struct napi_struct *napi, 2731 int (*poll)(struct napi_struct *, int)) 2732 { 2733 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2734 } 2735 2736 /** 2737 * __netif_napi_del - remove a NAPI context 2738 * @napi: NAPI context 2739 * 2740 * Warning: caller must observe RCU grace period before freeing memory 2741 * containing @napi. Drivers might want to call this helper to combine 2742 * all the needed RCU grace periods into a single one. 2743 */ 2744 void __netif_napi_del(struct napi_struct *napi); 2745 2746 /** 2747 * netif_napi_del - remove a NAPI context 2748 * @napi: NAPI context 2749 * 2750 * netif_napi_del() removes a NAPI context from the network device NAPI list 2751 */ 2752 static inline void netif_napi_del(struct napi_struct *napi) 2753 { 2754 __netif_napi_del(napi); 2755 synchronize_net(); 2756 } 2757 2758 struct packet_type { 2759 __be16 type; /* This is really htons(ether_type). */ 2760 bool ignore_outgoing; 2761 struct net_device *dev; /* NULL is wildcarded here */ 2762 netdevice_tracker dev_tracker; 2763 int (*func) (struct sk_buff *, 2764 struct net_device *, 2765 struct packet_type *, 2766 struct net_device *); 2767 void (*list_func) (struct list_head *, 2768 struct packet_type *, 2769 struct net_device *); 2770 bool (*id_match)(struct packet_type *ptype, 2771 struct sock *sk); 2772 struct net *af_packet_net; 2773 void *af_packet_priv; 2774 struct list_head list; 2775 }; 2776 2777 struct offload_callbacks { 2778 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2779 netdev_features_t features); 2780 struct sk_buff *(*gro_receive)(struct list_head *head, 2781 struct sk_buff *skb); 2782 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2783 }; 2784 2785 struct packet_offload { 2786 __be16 type; /* This is really htons(ether_type). */ 2787 u16 priority; 2788 struct offload_callbacks callbacks; 2789 struct list_head list; 2790 }; 2791 2792 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2793 struct pcpu_sw_netstats { 2794 u64_stats_t rx_packets; 2795 u64_stats_t rx_bytes; 2796 u64_stats_t tx_packets; 2797 u64_stats_t tx_bytes; 2798 struct u64_stats_sync syncp; 2799 } __aligned(4 * sizeof(u64)); 2800 2801 struct pcpu_dstats { 2802 u64 rx_packets; 2803 u64 rx_bytes; 2804 u64 rx_drops; 2805 u64 tx_packets; 2806 u64 tx_bytes; 2807 u64 tx_drops; 2808 struct u64_stats_sync syncp; 2809 } __aligned(8 * sizeof(u64)); 2810 2811 struct pcpu_lstats { 2812 u64_stats_t packets; 2813 u64_stats_t bytes; 2814 struct u64_stats_sync syncp; 2815 } __aligned(2 * sizeof(u64)); 2816 2817 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2818 2819 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2820 { 2821 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2822 2823 u64_stats_update_begin(&tstats->syncp); 2824 u64_stats_add(&tstats->rx_bytes, len); 2825 u64_stats_inc(&tstats->rx_packets); 2826 u64_stats_update_end(&tstats->syncp); 2827 } 2828 2829 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2830 unsigned int packets, 2831 unsigned int len) 2832 { 2833 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2834 2835 u64_stats_update_begin(&tstats->syncp); 2836 u64_stats_add(&tstats->tx_bytes, len); 2837 u64_stats_add(&tstats->tx_packets, packets); 2838 u64_stats_update_end(&tstats->syncp); 2839 } 2840 2841 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2842 { 2843 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2844 2845 u64_stats_update_begin(&lstats->syncp); 2846 u64_stats_add(&lstats->bytes, len); 2847 u64_stats_inc(&lstats->packets); 2848 u64_stats_update_end(&lstats->syncp); 2849 } 2850 2851 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2852 ({ \ 2853 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2854 if (pcpu_stats) { \ 2855 int __cpu; \ 2856 for_each_possible_cpu(__cpu) { \ 2857 typeof(type) *stat; \ 2858 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2859 u64_stats_init(&stat->syncp); \ 2860 } \ 2861 } \ 2862 pcpu_stats; \ 2863 }) 2864 2865 #define netdev_alloc_pcpu_stats(type) \ 2866 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2867 2868 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2869 ({ \ 2870 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2871 if (pcpu_stats) { \ 2872 int __cpu; \ 2873 for_each_possible_cpu(__cpu) { \ 2874 typeof(type) *stat; \ 2875 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2876 u64_stats_init(&stat->syncp); \ 2877 } \ 2878 } \ 2879 pcpu_stats; \ 2880 }) 2881 2882 enum netdev_lag_tx_type { 2883 NETDEV_LAG_TX_TYPE_UNKNOWN, 2884 NETDEV_LAG_TX_TYPE_RANDOM, 2885 NETDEV_LAG_TX_TYPE_BROADCAST, 2886 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2887 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2888 NETDEV_LAG_TX_TYPE_HASH, 2889 }; 2890 2891 enum netdev_lag_hash { 2892 NETDEV_LAG_HASH_NONE, 2893 NETDEV_LAG_HASH_L2, 2894 NETDEV_LAG_HASH_L34, 2895 NETDEV_LAG_HASH_L23, 2896 NETDEV_LAG_HASH_E23, 2897 NETDEV_LAG_HASH_E34, 2898 NETDEV_LAG_HASH_VLAN_SRCMAC, 2899 NETDEV_LAG_HASH_UNKNOWN, 2900 }; 2901 2902 struct netdev_lag_upper_info { 2903 enum netdev_lag_tx_type tx_type; 2904 enum netdev_lag_hash hash_type; 2905 }; 2906 2907 struct netdev_lag_lower_state_info { 2908 u8 link_up : 1, 2909 tx_enabled : 1; 2910 }; 2911 2912 #include <linux/notifier.h> 2913 2914 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2915 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2916 * adding new types. 2917 */ 2918 enum netdev_cmd { 2919 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2920 NETDEV_DOWN, 2921 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2922 detected a hardware crash and restarted 2923 - we can use this eg to kick tcp sessions 2924 once done */ 2925 NETDEV_CHANGE, /* Notify device state change */ 2926 NETDEV_REGISTER, 2927 NETDEV_UNREGISTER, 2928 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2929 NETDEV_CHANGEADDR, /* notify after the address change */ 2930 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2931 NETDEV_GOING_DOWN, 2932 NETDEV_CHANGENAME, 2933 NETDEV_FEAT_CHANGE, 2934 NETDEV_BONDING_FAILOVER, 2935 NETDEV_PRE_UP, 2936 NETDEV_PRE_TYPE_CHANGE, 2937 NETDEV_POST_TYPE_CHANGE, 2938 NETDEV_POST_INIT, 2939 NETDEV_PRE_UNINIT, 2940 NETDEV_RELEASE, 2941 NETDEV_NOTIFY_PEERS, 2942 NETDEV_JOIN, 2943 NETDEV_CHANGEUPPER, 2944 NETDEV_RESEND_IGMP, 2945 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2946 NETDEV_CHANGEINFODATA, 2947 NETDEV_BONDING_INFO, 2948 NETDEV_PRECHANGEUPPER, 2949 NETDEV_CHANGELOWERSTATE, 2950 NETDEV_UDP_TUNNEL_PUSH_INFO, 2951 NETDEV_UDP_TUNNEL_DROP_INFO, 2952 NETDEV_CHANGE_TX_QUEUE_LEN, 2953 NETDEV_CVLAN_FILTER_PUSH_INFO, 2954 NETDEV_CVLAN_FILTER_DROP_INFO, 2955 NETDEV_SVLAN_FILTER_PUSH_INFO, 2956 NETDEV_SVLAN_FILTER_DROP_INFO, 2957 NETDEV_OFFLOAD_XSTATS_ENABLE, 2958 NETDEV_OFFLOAD_XSTATS_DISABLE, 2959 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 2960 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 2961 NETDEV_XDP_FEAT_CHANGE, 2962 }; 2963 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2964 2965 int register_netdevice_notifier(struct notifier_block *nb); 2966 int unregister_netdevice_notifier(struct notifier_block *nb); 2967 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2968 int unregister_netdevice_notifier_net(struct net *net, 2969 struct notifier_block *nb); 2970 int register_netdevice_notifier_dev_net(struct net_device *dev, 2971 struct notifier_block *nb, 2972 struct netdev_net_notifier *nn); 2973 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2974 struct notifier_block *nb, 2975 struct netdev_net_notifier *nn); 2976 2977 struct netdev_notifier_info { 2978 struct net_device *dev; 2979 struct netlink_ext_ack *extack; 2980 }; 2981 2982 struct netdev_notifier_info_ext { 2983 struct netdev_notifier_info info; /* must be first */ 2984 union { 2985 u32 mtu; 2986 } ext; 2987 }; 2988 2989 struct netdev_notifier_change_info { 2990 struct netdev_notifier_info info; /* must be first */ 2991 unsigned int flags_changed; 2992 }; 2993 2994 struct netdev_notifier_changeupper_info { 2995 struct netdev_notifier_info info; /* must be first */ 2996 struct net_device *upper_dev; /* new upper dev */ 2997 bool master; /* is upper dev master */ 2998 bool linking; /* is the notification for link or unlink */ 2999 void *upper_info; /* upper dev info */ 3000 }; 3001 3002 struct netdev_notifier_changelowerstate_info { 3003 struct netdev_notifier_info info; /* must be first */ 3004 void *lower_state_info; /* is lower dev state */ 3005 }; 3006 3007 struct netdev_notifier_pre_changeaddr_info { 3008 struct netdev_notifier_info info; /* must be first */ 3009 const unsigned char *dev_addr; 3010 }; 3011 3012 enum netdev_offload_xstats_type { 3013 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 3014 }; 3015 3016 struct netdev_notifier_offload_xstats_info { 3017 struct netdev_notifier_info info; /* must be first */ 3018 enum netdev_offload_xstats_type type; 3019 3020 union { 3021 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 3022 struct netdev_notifier_offload_xstats_rd *report_delta; 3023 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 3024 struct netdev_notifier_offload_xstats_ru *report_used; 3025 }; 3026 }; 3027 3028 int netdev_offload_xstats_enable(struct net_device *dev, 3029 enum netdev_offload_xstats_type type, 3030 struct netlink_ext_ack *extack); 3031 int netdev_offload_xstats_disable(struct net_device *dev, 3032 enum netdev_offload_xstats_type type); 3033 bool netdev_offload_xstats_enabled(const struct net_device *dev, 3034 enum netdev_offload_xstats_type type); 3035 int netdev_offload_xstats_get(struct net_device *dev, 3036 enum netdev_offload_xstats_type type, 3037 struct rtnl_hw_stats64 *stats, bool *used, 3038 struct netlink_ext_ack *extack); 3039 void 3040 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 3041 const struct rtnl_hw_stats64 *stats); 3042 void 3043 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 3044 void netdev_offload_xstats_push_delta(struct net_device *dev, 3045 enum netdev_offload_xstats_type type, 3046 const struct rtnl_hw_stats64 *stats); 3047 3048 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 3049 struct net_device *dev) 3050 { 3051 info->dev = dev; 3052 info->extack = NULL; 3053 } 3054 3055 static inline struct net_device * 3056 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 3057 { 3058 return info->dev; 3059 } 3060 3061 static inline struct netlink_ext_ack * 3062 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 3063 { 3064 return info->extack; 3065 } 3066 3067 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 3068 int call_netdevice_notifiers_info(unsigned long val, 3069 struct netdev_notifier_info *info); 3070 3071 extern rwlock_t dev_base_lock; /* Device list lock */ 3072 3073 #define for_each_netdev(net, d) \ 3074 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 3075 #define for_each_netdev_reverse(net, d) \ 3076 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 3077 #define for_each_netdev_rcu(net, d) \ 3078 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 3079 #define for_each_netdev_safe(net, d, n) \ 3080 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 3081 #define for_each_netdev_continue(net, d) \ 3082 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 3083 #define for_each_netdev_continue_reverse(net, d) \ 3084 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 3085 dev_list) 3086 #define for_each_netdev_continue_rcu(net, d) \ 3087 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 3088 #define for_each_netdev_in_bond_rcu(bond, slave) \ 3089 for_each_netdev_rcu(&init_net, slave) \ 3090 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 3091 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 3092 3093 #define for_each_netdev_dump(net, d, ifindex) \ 3094 xa_for_each_start(&(net)->dev_by_index, (ifindex), (d), (ifindex)) 3095 3096 static inline struct net_device *next_net_device(struct net_device *dev) 3097 { 3098 struct list_head *lh; 3099 struct net *net; 3100 3101 net = dev_net(dev); 3102 lh = dev->dev_list.next; 3103 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3104 } 3105 3106 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 3107 { 3108 struct list_head *lh; 3109 struct net *net; 3110 3111 net = dev_net(dev); 3112 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 3113 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3114 } 3115 3116 static inline struct net_device *first_net_device(struct net *net) 3117 { 3118 return list_empty(&net->dev_base_head) ? NULL : 3119 net_device_entry(net->dev_base_head.next); 3120 } 3121 3122 static inline struct net_device *first_net_device_rcu(struct net *net) 3123 { 3124 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 3125 3126 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3127 } 3128 3129 int netdev_boot_setup_check(struct net_device *dev); 3130 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 3131 const char *hwaddr); 3132 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 3133 void dev_add_pack(struct packet_type *pt); 3134 void dev_remove_pack(struct packet_type *pt); 3135 void __dev_remove_pack(struct packet_type *pt); 3136 void dev_add_offload(struct packet_offload *po); 3137 void dev_remove_offload(struct packet_offload *po); 3138 3139 int dev_get_iflink(const struct net_device *dev); 3140 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 3141 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 3142 struct net_device_path_stack *stack); 3143 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 3144 unsigned short mask); 3145 struct net_device *dev_get_by_name(struct net *net, const char *name); 3146 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 3147 struct net_device *__dev_get_by_name(struct net *net, const char *name); 3148 bool netdev_name_in_use(struct net *net, const char *name); 3149 int dev_alloc_name(struct net_device *dev, const char *name); 3150 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 3151 void dev_close(struct net_device *dev); 3152 void dev_close_many(struct list_head *head, bool unlink); 3153 void dev_disable_lro(struct net_device *dev); 3154 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 3155 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 3156 struct net_device *sb_dev); 3157 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 3158 struct net_device *sb_dev); 3159 3160 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3161 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3162 3163 static inline int dev_queue_xmit(struct sk_buff *skb) 3164 { 3165 return __dev_queue_xmit(skb, NULL); 3166 } 3167 3168 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3169 struct net_device *sb_dev) 3170 { 3171 return __dev_queue_xmit(skb, sb_dev); 3172 } 3173 3174 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3175 { 3176 int ret; 3177 3178 ret = __dev_direct_xmit(skb, queue_id); 3179 if (!dev_xmit_complete(ret)) 3180 kfree_skb(skb); 3181 return ret; 3182 } 3183 3184 int register_netdevice(struct net_device *dev); 3185 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3186 void unregister_netdevice_many(struct list_head *head); 3187 static inline void unregister_netdevice(struct net_device *dev) 3188 { 3189 unregister_netdevice_queue(dev, NULL); 3190 } 3191 3192 int netdev_refcnt_read(const struct net_device *dev); 3193 void free_netdev(struct net_device *dev); 3194 void netdev_freemem(struct net_device *dev); 3195 int init_dummy_netdev(struct net_device *dev); 3196 3197 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3198 struct sk_buff *skb, 3199 bool all_slaves); 3200 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3201 struct sock *sk); 3202 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3203 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3204 struct net_device *netdev_get_by_index(struct net *net, int ifindex, 3205 netdevice_tracker *tracker, gfp_t gfp); 3206 struct net_device *netdev_get_by_name(struct net *net, const char *name, 3207 netdevice_tracker *tracker, gfp_t gfp); 3208 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3209 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 3210 3211 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3212 unsigned short type, 3213 const void *daddr, const void *saddr, 3214 unsigned int len) 3215 { 3216 if (!dev->header_ops || !dev->header_ops->create) 3217 return 0; 3218 3219 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3220 } 3221 3222 static inline int dev_parse_header(const struct sk_buff *skb, 3223 unsigned char *haddr) 3224 { 3225 const struct net_device *dev = skb->dev; 3226 3227 if (!dev->header_ops || !dev->header_ops->parse) 3228 return 0; 3229 return dev->header_ops->parse(skb, haddr); 3230 } 3231 3232 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3233 { 3234 const struct net_device *dev = skb->dev; 3235 3236 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3237 return 0; 3238 return dev->header_ops->parse_protocol(skb); 3239 } 3240 3241 /* ll_header must have at least hard_header_len allocated */ 3242 static inline bool dev_validate_header(const struct net_device *dev, 3243 char *ll_header, int len) 3244 { 3245 if (likely(len >= dev->hard_header_len)) 3246 return true; 3247 if (len < dev->min_header_len) 3248 return false; 3249 3250 if (capable(CAP_SYS_RAWIO)) { 3251 memset(ll_header + len, 0, dev->hard_header_len - len); 3252 return true; 3253 } 3254 3255 if (dev->header_ops && dev->header_ops->validate) 3256 return dev->header_ops->validate(ll_header, len); 3257 3258 return false; 3259 } 3260 3261 static inline bool dev_has_header(const struct net_device *dev) 3262 { 3263 return dev->header_ops && dev->header_ops->create; 3264 } 3265 3266 /* 3267 * Incoming packets are placed on per-CPU queues 3268 */ 3269 struct softnet_data { 3270 struct list_head poll_list; 3271 struct sk_buff_head process_queue; 3272 3273 /* stats */ 3274 unsigned int processed; 3275 unsigned int time_squeeze; 3276 #ifdef CONFIG_RPS 3277 struct softnet_data *rps_ipi_list; 3278 #endif 3279 3280 bool in_net_rx_action; 3281 bool in_napi_threaded_poll; 3282 3283 #ifdef CONFIG_NET_FLOW_LIMIT 3284 struct sd_flow_limit __rcu *flow_limit; 3285 #endif 3286 struct Qdisc *output_queue; 3287 struct Qdisc **output_queue_tailp; 3288 struct sk_buff *completion_queue; 3289 #ifdef CONFIG_XFRM_OFFLOAD 3290 struct sk_buff_head xfrm_backlog; 3291 #endif 3292 /* written and read only by owning cpu: */ 3293 struct { 3294 u16 recursion; 3295 u8 more; 3296 #ifdef CONFIG_NET_EGRESS 3297 u8 skip_txqueue; 3298 #endif 3299 } xmit; 3300 #ifdef CONFIG_RPS 3301 /* input_queue_head should be written by cpu owning this struct, 3302 * and only read by other cpus. Worth using a cache line. 3303 */ 3304 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3305 3306 /* Elements below can be accessed between CPUs for RPS/RFS */ 3307 call_single_data_t csd ____cacheline_aligned_in_smp; 3308 struct softnet_data *rps_ipi_next; 3309 unsigned int cpu; 3310 unsigned int input_queue_tail; 3311 #endif 3312 unsigned int received_rps; 3313 unsigned int dropped; 3314 struct sk_buff_head input_pkt_queue; 3315 struct napi_struct backlog; 3316 3317 /* Another possibly contended cache line */ 3318 spinlock_t defer_lock ____cacheline_aligned_in_smp; 3319 int defer_count; 3320 int defer_ipi_scheduled; 3321 struct sk_buff *defer_list; 3322 call_single_data_t defer_csd; 3323 }; 3324 3325 static inline void input_queue_head_incr(struct softnet_data *sd) 3326 { 3327 #ifdef CONFIG_RPS 3328 sd->input_queue_head++; 3329 #endif 3330 } 3331 3332 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3333 unsigned int *qtail) 3334 { 3335 #ifdef CONFIG_RPS 3336 *qtail = ++sd->input_queue_tail; 3337 #endif 3338 } 3339 3340 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3341 3342 static inline int dev_recursion_level(void) 3343 { 3344 return this_cpu_read(softnet_data.xmit.recursion); 3345 } 3346 3347 #define XMIT_RECURSION_LIMIT 8 3348 static inline bool dev_xmit_recursion(void) 3349 { 3350 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3351 XMIT_RECURSION_LIMIT); 3352 } 3353 3354 static inline void dev_xmit_recursion_inc(void) 3355 { 3356 __this_cpu_inc(softnet_data.xmit.recursion); 3357 } 3358 3359 static inline void dev_xmit_recursion_dec(void) 3360 { 3361 __this_cpu_dec(softnet_data.xmit.recursion); 3362 } 3363 3364 void __netif_schedule(struct Qdisc *q); 3365 void netif_schedule_queue(struct netdev_queue *txq); 3366 3367 static inline void netif_tx_schedule_all(struct net_device *dev) 3368 { 3369 unsigned int i; 3370 3371 for (i = 0; i < dev->num_tx_queues; i++) 3372 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3373 } 3374 3375 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3376 { 3377 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3378 } 3379 3380 /** 3381 * netif_start_queue - allow transmit 3382 * @dev: network device 3383 * 3384 * Allow upper layers to call the device hard_start_xmit routine. 3385 */ 3386 static inline void netif_start_queue(struct net_device *dev) 3387 { 3388 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3389 } 3390 3391 static inline void netif_tx_start_all_queues(struct net_device *dev) 3392 { 3393 unsigned int i; 3394 3395 for (i = 0; i < dev->num_tx_queues; i++) { 3396 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3397 netif_tx_start_queue(txq); 3398 } 3399 } 3400 3401 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3402 3403 /** 3404 * netif_wake_queue - restart transmit 3405 * @dev: network device 3406 * 3407 * Allow upper layers to call the device hard_start_xmit routine. 3408 * Used for flow control when transmit resources are available. 3409 */ 3410 static inline void netif_wake_queue(struct net_device *dev) 3411 { 3412 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3413 } 3414 3415 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3416 { 3417 unsigned int i; 3418 3419 for (i = 0; i < dev->num_tx_queues; i++) { 3420 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3421 netif_tx_wake_queue(txq); 3422 } 3423 } 3424 3425 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3426 { 3427 /* Must be an atomic op see netif_txq_try_stop() */ 3428 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3429 } 3430 3431 /** 3432 * netif_stop_queue - stop transmitted packets 3433 * @dev: network device 3434 * 3435 * Stop upper layers calling the device hard_start_xmit routine. 3436 * Used for flow control when transmit resources are unavailable. 3437 */ 3438 static inline void netif_stop_queue(struct net_device *dev) 3439 { 3440 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3441 } 3442 3443 void netif_tx_stop_all_queues(struct net_device *dev); 3444 3445 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3446 { 3447 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3448 } 3449 3450 /** 3451 * netif_queue_stopped - test if transmit queue is flowblocked 3452 * @dev: network device 3453 * 3454 * Test if transmit queue on device is currently unable to send. 3455 */ 3456 static inline bool netif_queue_stopped(const struct net_device *dev) 3457 { 3458 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3459 } 3460 3461 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3462 { 3463 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3464 } 3465 3466 static inline bool 3467 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3468 { 3469 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3470 } 3471 3472 static inline bool 3473 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3474 { 3475 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3476 } 3477 3478 /** 3479 * netdev_queue_set_dql_min_limit - set dql minimum limit 3480 * @dev_queue: pointer to transmit queue 3481 * @min_limit: dql minimum limit 3482 * 3483 * Forces xmit_more() to return true until the minimum threshold 3484 * defined by @min_limit is reached (or until the tx queue is 3485 * empty). Warning: to be use with care, misuse will impact the 3486 * latency. 3487 */ 3488 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3489 unsigned int min_limit) 3490 { 3491 #ifdef CONFIG_BQL 3492 dev_queue->dql.min_limit = min_limit; 3493 #endif 3494 } 3495 3496 /** 3497 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3498 * @dev_queue: pointer to transmit queue 3499 * 3500 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3501 * to give appropriate hint to the CPU. 3502 */ 3503 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3504 { 3505 #ifdef CONFIG_BQL 3506 prefetchw(&dev_queue->dql.num_queued); 3507 #endif 3508 } 3509 3510 /** 3511 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3512 * @dev_queue: pointer to transmit queue 3513 * 3514 * BQL enabled drivers might use this helper in their TX completion path, 3515 * to give appropriate hint to the CPU. 3516 */ 3517 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3518 { 3519 #ifdef CONFIG_BQL 3520 prefetchw(&dev_queue->dql.limit); 3521 #endif 3522 } 3523 3524 /** 3525 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3526 * @dev_queue: network device queue 3527 * @bytes: number of bytes queued to the device queue 3528 * 3529 * Report the number of bytes queued for sending/completion to the network 3530 * device hardware queue. @bytes should be a good approximation and should 3531 * exactly match netdev_completed_queue() @bytes. 3532 * This is typically called once per packet, from ndo_start_xmit(). 3533 */ 3534 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3535 unsigned int bytes) 3536 { 3537 #ifdef CONFIG_BQL 3538 dql_queued(&dev_queue->dql, bytes); 3539 3540 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3541 return; 3542 3543 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3544 3545 /* 3546 * The XOFF flag must be set before checking the dql_avail below, 3547 * because in netdev_tx_completed_queue we update the dql_completed 3548 * before checking the XOFF flag. 3549 */ 3550 smp_mb(); 3551 3552 /* check again in case another CPU has just made room avail */ 3553 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3554 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3555 #endif 3556 } 3557 3558 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3559 * that they should not test BQL status themselves. 3560 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3561 * skb of a batch. 3562 * Returns true if the doorbell must be used to kick the NIC. 3563 */ 3564 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3565 unsigned int bytes, 3566 bool xmit_more) 3567 { 3568 if (xmit_more) { 3569 #ifdef CONFIG_BQL 3570 dql_queued(&dev_queue->dql, bytes); 3571 #endif 3572 return netif_tx_queue_stopped(dev_queue); 3573 } 3574 netdev_tx_sent_queue(dev_queue, bytes); 3575 return true; 3576 } 3577 3578 /** 3579 * netdev_sent_queue - report the number of bytes queued to hardware 3580 * @dev: network device 3581 * @bytes: number of bytes queued to the hardware device queue 3582 * 3583 * Report the number of bytes queued for sending/completion to the network 3584 * device hardware queue#0. @bytes should be a good approximation and should 3585 * exactly match netdev_completed_queue() @bytes. 3586 * This is typically called once per packet, from ndo_start_xmit(). 3587 */ 3588 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3589 { 3590 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3591 } 3592 3593 static inline bool __netdev_sent_queue(struct net_device *dev, 3594 unsigned int bytes, 3595 bool xmit_more) 3596 { 3597 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3598 xmit_more); 3599 } 3600 3601 /** 3602 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3603 * @dev_queue: network device queue 3604 * @pkts: number of packets (currently ignored) 3605 * @bytes: number of bytes dequeued from the device queue 3606 * 3607 * Must be called at most once per TX completion round (and not per 3608 * individual packet), so that BQL can adjust its limits appropriately. 3609 */ 3610 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3611 unsigned int pkts, unsigned int bytes) 3612 { 3613 #ifdef CONFIG_BQL 3614 if (unlikely(!bytes)) 3615 return; 3616 3617 dql_completed(&dev_queue->dql, bytes); 3618 3619 /* 3620 * Without the memory barrier there is a small possiblity that 3621 * netdev_tx_sent_queue will miss the update and cause the queue to 3622 * be stopped forever 3623 */ 3624 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */ 3625 3626 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3627 return; 3628 3629 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3630 netif_schedule_queue(dev_queue); 3631 #endif 3632 } 3633 3634 /** 3635 * netdev_completed_queue - report bytes and packets completed by device 3636 * @dev: network device 3637 * @pkts: actual number of packets sent over the medium 3638 * @bytes: actual number of bytes sent over the medium 3639 * 3640 * Report the number of bytes and packets transmitted by the network device 3641 * hardware queue over the physical medium, @bytes must exactly match the 3642 * @bytes amount passed to netdev_sent_queue() 3643 */ 3644 static inline void netdev_completed_queue(struct net_device *dev, 3645 unsigned int pkts, unsigned int bytes) 3646 { 3647 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3648 } 3649 3650 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3651 { 3652 #ifdef CONFIG_BQL 3653 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3654 dql_reset(&q->dql); 3655 #endif 3656 } 3657 3658 /** 3659 * netdev_reset_queue - reset the packets and bytes count of a network device 3660 * @dev_queue: network device 3661 * 3662 * Reset the bytes and packet count of a network device and clear the 3663 * software flow control OFF bit for this network device 3664 */ 3665 static inline void netdev_reset_queue(struct net_device *dev_queue) 3666 { 3667 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3668 } 3669 3670 /** 3671 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3672 * @dev: network device 3673 * @queue_index: given tx queue index 3674 * 3675 * Returns 0 if given tx queue index >= number of device tx queues, 3676 * otherwise returns the originally passed tx queue index. 3677 */ 3678 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3679 { 3680 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3681 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3682 dev->name, queue_index, 3683 dev->real_num_tx_queues); 3684 return 0; 3685 } 3686 3687 return queue_index; 3688 } 3689 3690 /** 3691 * netif_running - test if up 3692 * @dev: network device 3693 * 3694 * Test if the device has been brought up. 3695 */ 3696 static inline bool netif_running(const struct net_device *dev) 3697 { 3698 return test_bit(__LINK_STATE_START, &dev->state); 3699 } 3700 3701 /* 3702 * Routines to manage the subqueues on a device. We only need start, 3703 * stop, and a check if it's stopped. All other device management is 3704 * done at the overall netdevice level. 3705 * Also test the device if we're multiqueue. 3706 */ 3707 3708 /** 3709 * netif_start_subqueue - allow sending packets on subqueue 3710 * @dev: network device 3711 * @queue_index: sub queue index 3712 * 3713 * Start individual transmit queue of a device with multiple transmit queues. 3714 */ 3715 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3716 { 3717 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3718 3719 netif_tx_start_queue(txq); 3720 } 3721 3722 /** 3723 * netif_stop_subqueue - stop sending packets on subqueue 3724 * @dev: network device 3725 * @queue_index: sub queue index 3726 * 3727 * Stop individual transmit queue of a device with multiple transmit queues. 3728 */ 3729 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3730 { 3731 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3732 netif_tx_stop_queue(txq); 3733 } 3734 3735 /** 3736 * __netif_subqueue_stopped - test status of subqueue 3737 * @dev: network device 3738 * @queue_index: sub queue index 3739 * 3740 * Check individual transmit queue of a device with multiple transmit queues. 3741 */ 3742 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3743 u16 queue_index) 3744 { 3745 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3746 3747 return netif_tx_queue_stopped(txq); 3748 } 3749 3750 /** 3751 * netif_subqueue_stopped - test status of subqueue 3752 * @dev: network device 3753 * @skb: sub queue buffer pointer 3754 * 3755 * Check individual transmit queue of a device with multiple transmit queues. 3756 */ 3757 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3758 struct sk_buff *skb) 3759 { 3760 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3761 } 3762 3763 /** 3764 * netif_wake_subqueue - allow sending packets on subqueue 3765 * @dev: network device 3766 * @queue_index: sub queue index 3767 * 3768 * Resume individual transmit queue of a device with multiple transmit queues. 3769 */ 3770 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3771 { 3772 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3773 3774 netif_tx_wake_queue(txq); 3775 } 3776 3777 #ifdef CONFIG_XPS 3778 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3779 u16 index); 3780 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3781 u16 index, enum xps_map_type type); 3782 3783 /** 3784 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3785 * @j: CPU/Rx queue index 3786 * @mask: bitmask of all cpus/rx queues 3787 * @nr_bits: number of bits in the bitmask 3788 * 3789 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3790 */ 3791 static inline bool netif_attr_test_mask(unsigned long j, 3792 const unsigned long *mask, 3793 unsigned int nr_bits) 3794 { 3795 cpu_max_bits_warn(j, nr_bits); 3796 return test_bit(j, mask); 3797 } 3798 3799 /** 3800 * netif_attr_test_online - Test for online CPU/Rx queue 3801 * @j: CPU/Rx queue index 3802 * @online_mask: bitmask for CPUs/Rx queues that are online 3803 * @nr_bits: number of bits in the bitmask 3804 * 3805 * Returns true if a CPU/Rx queue is online. 3806 */ 3807 static inline bool netif_attr_test_online(unsigned long j, 3808 const unsigned long *online_mask, 3809 unsigned int nr_bits) 3810 { 3811 cpu_max_bits_warn(j, nr_bits); 3812 3813 if (online_mask) 3814 return test_bit(j, online_mask); 3815 3816 return (j < nr_bits); 3817 } 3818 3819 /** 3820 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3821 * @n: CPU/Rx queue index 3822 * @srcp: the cpumask/Rx queue mask pointer 3823 * @nr_bits: number of bits in the bitmask 3824 * 3825 * Returns >= nr_bits if no further CPUs/Rx queues set. 3826 */ 3827 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3828 unsigned int nr_bits) 3829 { 3830 /* -1 is a legal arg here. */ 3831 if (n != -1) 3832 cpu_max_bits_warn(n, nr_bits); 3833 3834 if (srcp) 3835 return find_next_bit(srcp, nr_bits, n + 1); 3836 3837 return n + 1; 3838 } 3839 3840 /** 3841 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3842 * @n: CPU/Rx queue index 3843 * @src1p: the first CPUs/Rx queues mask pointer 3844 * @src2p: the second CPUs/Rx queues mask pointer 3845 * @nr_bits: number of bits in the bitmask 3846 * 3847 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3848 */ 3849 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3850 const unsigned long *src2p, 3851 unsigned int nr_bits) 3852 { 3853 /* -1 is a legal arg here. */ 3854 if (n != -1) 3855 cpu_max_bits_warn(n, nr_bits); 3856 3857 if (src1p && src2p) 3858 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3859 else if (src1p) 3860 return find_next_bit(src1p, nr_bits, n + 1); 3861 else if (src2p) 3862 return find_next_bit(src2p, nr_bits, n + 1); 3863 3864 return n + 1; 3865 } 3866 #else 3867 static inline int netif_set_xps_queue(struct net_device *dev, 3868 const struct cpumask *mask, 3869 u16 index) 3870 { 3871 return 0; 3872 } 3873 3874 static inline int __netif_set_xps_queue(struct net_device *dev, 3875 const unsigned long *mask, 3876 u16 index, enum xps_map_type type) 3877 { 3878 return 0; 3879 } 3880 #endif 3881 3882 /** 3883 * netif_is_multiqueue - test if device has multiple transmit queues 3884 * @dev: network device 3885 * 3886 * Check if device has multiple transmit queues 3887 */ 3888 static inline bool netif_is_multiqueue(const struct net_device *dev) 3889 { 3890 return dev->num_tx_queues > 1; 3891 } 3892 3893 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3894 3895 #ifdef CONFIG_SYSFS 3896 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3897 #else 3898 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3899 unsigned int rxqs) 3900 { 3901 dev->real_num_rx_queues = rxqs; 3902 return 0; 3903 } 3904 #endif 3905 int netif_set_real_num_queues(struct net_device *dev, 3906 unsigned int txq, unsigned int rxq); 3907 3908 int netif_get_num_default_rss_queues(void); 3909 3910 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason); 3911 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason); 3912 3913 /* 3914 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3915 * interrupt context or with hardware interrupts being disabled. 3916 * (in_hardirq() || irqs_disabled()) 3917 * 3918 * We provide four helpers that can be used in following contexts : 3919 * 3920 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3921 * replacing kfree_skb(skb) 3922 * 3923 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3924 * Typically used in place of consume_skb(skb) in TX completion path 3925 * 3926 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3927 * replacing kfree_skb(skb) 3928 * 3929 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3930 * and consumed a packet. Used in place of consume_skb(skb) 3931 */ 3932 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3933 { 3934 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 3935 } 3936 3937 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3938 { 3939 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED); 3940 } 3941 3942 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3943 { 3944 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 3945 } 3946 3947 static inline void dev_consume_skb_any(struct sk_buff *skb) 3948 { 3949 dev_kfree_skb_any_reason(skb, SKB_CONSUMED); 3950 } 3951 3952 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3953 struct bpf_prog *xdp_prog); 3954 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3955 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3956 int netif_rx(struct sk_buff *skb); 3957 int __netif_rx(struct sk_buff *skb); 3958 3959 int netif_receive_skb(struct sk_buff *skb); 3960 int netif_receive_skb_core(struct sk_buff *skb); 3961 void netif_receive_skb_list_internal(struct list_head *head); 3962 void netif_receive_skb_list(struct list_head *head); 3963 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3964 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3965 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3966 void napi_get_frags_check(struct napi_struct *napi); 3967 gro_result_t napi_gro_frags(struct napi_struct *napi); 3968 struct packet_offload *gro_find_receive_by_type(__be16 type); 3969 struct packet_offload *gro_find_complete_by_type(__be16 type); 3970 3971 static inline void napi_free_frags(struct napi_struct *napi) 3972 { 3973 kfree_skb(napi->skb); 3974 napi->skb = NULL; 3975 } 3976 3977 bool netdev_is_rx_handler_busy(struct net_device *dev); 3978 int netdev_rx_handler_register(struct net_device *dev, 3979 rx_handler_func_t *rx_handler, 3980 void *rx_handler_data); 3981 void netdev_rx_handler_unregister(struct net_device *dev); 3982 3983 bool dev_valid_name(const char *name); 3984 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 3985 { 3986 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 3987 } 3988 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 3989 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 3990 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3991 void __user *data, bool *need_copyout); 3992 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 3993 int generic_hwtstamp_get_lower(struct net_device *dev, 3994 struct kernel_hwtstamp_config *kernel_cfg); 3995 int generic_hwtstamp_set_lower(struct net_device *dev, 3996 struct kernel_hwtstamp_config *kernel_cfg, 3997 struct netlink_ext_ack *extack); 3998 int dev_set_hwtstamp_phylib(struct net_device *dev, 3999 struct kernel_hwtstamp_config *cfg, 4000 struct netlink_ext_ack *extack); 4001 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 4002 unsigned int dev_get_flags(const struct net_device *); 4003 int __dev_change_flags(struct net_device *dev, unsigned int flags, 4004 struct netlink_ext_ack *extack); 4005 int dev_change_flags(struct net_device *dev, unsigned int flags, 4006 struct netlink_ext_ack *extack); 4007 int dev_set_alias(struct net_device *, const char *, size_t); 4008 int dev_get_alias(const struct net_device *, char *, size_t); 4009 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 4010 const char *pat, int new_ifindex); 4011 static inline 4012 int dev_change_net_namespace(struct net_device *dev, struct net *net, 4013 const char *pat) 4014 { 4015 return __dev_change_net_namespace(dev, net, pat, 0); 4016 } 4017 int __dev_set_mtu(struct net_device *, int); 4018 int dev_set_mtu(struct net_device *, int); 4019 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 4020 struct netlink_ext_ack *extack); 4021 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 4022 struct netlink_ext_ack *extack); 4023 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 4024 struct netlink_ext_ack *extack); 4025 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 4026 int dev_get_port_parent_id(struct net_device *dev, 4027 struct netdev_phys_item_id *ppid, bool recurse); 4028 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 4029 void netdev_dpll_pin_set(struct net_device *dev, struct dpll_pin *dpll_pin); 4030 void netdev_dpll_pin_clear(struct net_device *dev); 4031 4032 static inline struct dpll_pin *netdev_dpll_pin(const struct net_device *dev) 4033 { 4034 #if IS_ENABLED(CONFIG_DPLL) 4035 return dev->dpll_pin; 4036 #else 4037 return NULL; 4038 #endif 4039 } 4040 4041 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 4042 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 4043 struct netdev_queue *txq, int *ret); 4044 4045 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 4046 u8 dev_xdp_prog_count(struct net_device *dev); 4047 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 4048 4049 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4050 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4051 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 4052 bool is_skb_forwardable(const struct net_device *dev, 4053 const struct sk_buff *skb); 4054 4055 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 4056 const struct sk_buff *skb, 4057 const bool check_mtu) 4058 { 4059 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 4060 unsigned int len; 4061 4062 if (!(dev->flags & IFF_UP)) 4063 return false; 4064 4065 if (!check_mtu) 4066 return true; 4067 4068 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 4069 if (skb->len <= len) 4070 return true; 4071 4072 /* if TSO is enabled, we don't care about the length as the packet 4073 * could be forwarded without being segmented before 4074 */ 4075 if (skb_is_gso(skb)) 4076 return true; 4077 4078 return false; 4079 } 4080 4081 void netdev_core_stats_inc(struct net_device *dev, u32 offset); 4082 4083 #define DEV_CORE_STATS_INC(FIELD) \ 4084 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 4085 { \ 4086 netdev_core_stats_inc(dev, \ 4087 offsetof(struct net_device_core_stats, FIELD)); \ 4088 } 4089 DEV_CORE_STATS_INC(rx_dropped) 4090 DEV_CORE_STATS_INC(tx_dropped) 4091 DEV_CORE_STATS_INC(rx_nohandler) 4092 DEV_CORE_STATS_INC(rx_otherhost_dropped) 4093 #undef DEV_CORE_STATS_INC 4094 4095 static __always_inline int ____dev_forward_skb(struct net_device *dev, 4096 struct sk_buff *skb, 4097 const bool check_mtu) 4098 { 4099 if (skb_orphan_frags(skb, GFP_ATOMIC) || 4100 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 4101 dev_core_stats_rx_dropped_inc(dev); 4102 kfree_skb(skb); 4103 return NET_RX_DROP; 4104 } 4105 4106 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 4107 skb->priority = 0; 4108 return 0; 4109 } 4110 4111 bool dev_nit_active(struct net_device *dev); 4112 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 4113 4114 static inline void __dev_put(struct net_device *dev) 4115 { 4116 if (dev) { 4117 #ifdef CONFIG_PCPU_DEV_REFCNT 4118 this_cpu_dec(*dev->pcpu_refcnt); 4119 #else 4120 refcount_dec(&dev->dev_refcnt); 4121 #endif 4122 } 4123 } 4124 4125 static inline void __dev_hold(struct net_device *dev) 4126 { 4127 if (dev) { 4128 #ifdef CONFIG_PCPU_DEV_REFCNT 4129 this_cpu_inc(*dev->pcpu_refcnt); 4130 #else 4131 refcount_inc(&dev->dev_refcnt); 4132 #endif 4133 } 4134 } 4135 4136 static inline void __netdev_tracker_alloc(struct net_device *dev, 4137 netdevice_tracker *tracker, 4138 gfp_t gfp) 4139 { 4140 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4141 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 4142 #endif 4143 } 4144 4145 /* netdev_tracker_alloc() can upgrade a prior untracked reference 4146 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4147 */ 4148 static inline void netdev_tracker_alloc(struct net_device *dev, 4149 netdevice_tracker *tracker, gfp_t gfp) 4150 { 4151 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4152 refcount_dec(&dev->refcnt_tracker.no_tracker); 4153 __netdev_tracker_alloc(dev, tracker, gfp); 4154 #endif 4155 } 4156 4157 static inline void netdev_tracker_free(struct net_device *dev, 4158 netdevice_tracker *tracker) 4159 { 4160 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4161 ref_tracker_free(&dev->refcnt_tracker, tracker); 4162 #endif 4163 } 4164 4165 static inline void netdev_hold(struct net_device *dev, 4166 netdevice_tracker *tracker, gfp_t gfp) 4167 { 4168 if (dev) { 4169 __dev_hold(dev); 4170 __netdev_tracker_alloc(dev, tracker, gfp); 4171 } 4172 } 4173 4174 static inline void netdev_put(struct net_device *dev, 4175 netdevice_tracker *tracker) 4176 { 4177 if (dev) { 4178 netdev_tracker_free(dev, tracker); 4179 __dev_put(dev); 4180 } 4181 } 4182 4183 /** 4184 * dev_hold - get reference to device 4185 * @dev: network device 4186 * 4187 * Hold reference to device to keep it from being freed. 4188 * Try using netdev_hold() instead. 4189 */ 4190 static inline void dev_hold(struct net_device *dev) 4191 { 4192 netdev_hold(dev, NULL, GFP_ATOMIC); 4193 } 4194 4195 /** 4196 * dev_put - release reference to device 4197 * @dev: network device 4198 * 4199 * Release reference to device to allow it to be freed. 4200 * Try using netdev_put() instead. 4201 */ 4202 static inline void dev_put(struct net_device *dev) 4203 { 4204 netdev_put(dev, NULL); 4205 } 4206 4207 static inline void netdev_ref_replace(struct net_device *odev, 4208 struct net_device *ndev, 4209 netdevice_tracker *tracker, 4210 gfp_t gfp) 4211 { 4212 if (odev) 4213 netdev_tracker_free(odev, tracker); 4214 4215 __dev_hold(ndev); 4216 __dev_put(odev); 4217 4218 if (ndev) 4219 __netdev_tracker_alloc(ndev, tracker, gfp); 4220 } 4221 4222 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4223 * and _off may be called from IRQ context, but it is caller 4224 * who is responsible for serialization of these calls. 4225 * 4226 * The name carrier is inappropriate, these functions should really be 4227 * called netif_lowerlayer_*() because they represent the state of any 4228 * kind of lower layer not just hardware media. 4229 */ 4230 void linkwatch_fire_event(struct net_device *dev); 4231 4232 /** 4233 * linkwatch_sync_dev - sync linkwatch for the given device 4234 * @dev: network device to sync linkwatch for 4235 * 4236 * Sync linkwatch for the given device, removing it from the 4237 * pending work list (if queued). 4238 */ 4239 void linkwatch_sync_dev(struct net_device *dev); 4240 4241 /** 4242 * netif_carrier_ok - test if carrier present 4243 * @dev: network device 4244 * 4245 * Check if carrier is present on device 4246 */ 4247 static inline bool netif_carrier_ok(const struct net_device *dev) 4248 { 4249 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4250 } 4251 4252 unsigned long dev_trans_start(struct net_device *dev); 4253 4254 void __netdev_watchdog_up(struct net_device *dev); 4255 4256 void netif_carrier_on(struct net_device *dev); 4257 void netif_carrier_off(struct net_device *dev); 4258 void netif_carrier_event(struct net_device *dev); 4259 4260 /** 4261 * netif_dormant_on - mark device as dormant. 4262 * @dev: network device 4263 * 4264 * Mark device as dormant (as per RFC2863). 4265 * 4266 * The dormant state indicates that the relevant interface is not 4267 * actually in a condition to pass packets (i.e., it is not 'up') but is 4268 * in a "pending" state, waiting for some external event. For "on- 4269 * demand" interfaces, this new state identifies the situation where the 4270 * interface is waiting for events to place it in the up state. 4271 */ 4272 static inline void netif_dormant_on(struct net_device *dev) 4273 { 4274 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4275 linkwatch_fire_event(dev); 4276 } 4277 4278 /** 4279 * netif_dormant_off - set device as not dormant. 4280 * @dev: network device 4281 * 4282 * Device is not in dormant state. 4283 */ 4284 static inline void netif_dormant_off(struct net_device *dev) 4285 { 4286 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4287 linkwatch_fire_event(dev); 4288 } 4289 4290 /** 4291 * netif_dormant - test if device is dormant 4292 * @dev: network device 4293 * 4294 * Check if device is dormant. 4295 */ 4296 static inline bool netif_dormant(const struct net_device *dev) 4297 { 4298 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4299 } 4300 4301 4302 /** 4303 * netif_testing_on - mark device as under test. 4304 * @dev: network device 4305 * 4306 * Mark device as under test (as per RFC2863). 4307 * 4308 * The testing state indicates that some test(s) must be performed on 4309 * the interface. After completion, of the test, the interface state 4310 * will change to up, dormant, or down, as appropriate. 4311 */ 4312 static inline void netif_testing_on(struct net_device *dev) 4313 { 4314 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4315 linkwatch_fire_event(dev); 4316 } 4317 4318 /** 4319 * netif_testing_off - set device as not under test. 4320 * @dev: network device 4321 * 4322 * Device is not in testing state. 4323 */ 4324 static inline void netif_testing_off(struct net_device *dev) 4325 { 4326 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4327 linkwatch_fire_event(dev); 4328 } 4329 4330 /** 4331 * netif_testing - test if device is under test 4332 * @dev: network device 4333 * 4334 * Check if device is under test 4335 */ 4336 static inline bool netif_testing(const struct net_device *dev) 4337 { 4338 return test_bit(__LINK_STATE_TESTING, &dev->state); 4339 } 4340 4341 4342 /** 4343 * netif_oper_up - test if device is operational 4344 * @dev: network device 4345 * 4346 * Check if carrier is operational 4347 */ 4348 static inline bool netif_oper_up(const struct net_device *dev) 4349 { 4350 return (dev->operstate == IF_OPER_UP || 4351 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4352 } 4353 4354 /** 4355 * netif_device_present - is device available or removed 4356 * @dev: network device 4357 * 4358 * Check if device has not been removed from system. 4359 */ 4360 static inline bool netif_device_present(const struct net_device *dev) 4361 { 4362 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4363 } 4364 4365 void netif_device_detach(struct net_device *dev); 4366 4367 void netif_device_attach(struct net_device *dev); 4368 4369 /* 4370 * Network interface message level settings 4371 */ 4372 4373 enum { 4374 NETIF_MSG_DRV_BIT, 4375 NETIF_MSG_PROBE_BIT, 4376 NETIF_MSG_LINK_BIT, 4377 NETIF_MSG_TIMER_BIT, 4378 NETIF_MSG_IFDOWN_BIT, 4379 NETIF_MSG_IFUP_BIT, 4380 NETIF_MSG_RX_ERR_BIT, 4381 NETIF_MSG_TX_ERR_BIT, 4382 NETIF_MSG_TX_QUEUED_BIT, 4383 NETIF_MSG_INTR_BIT, 4384 NETIF_MSG_TX_DONE_BIT, 4385 NETIF_MSG_RX_STATUS_BIT, 4386 NETIF_MSG_PKTDATA_BIT, 4387 NETIF_MSG_HW_BIT, 4388 NETIF_MSG_WOL_BIT, 4389 4390 /* When you add a new bit above, update netif_msg_class_names array 4391 * in net/ethtool/common.c 4392 */ 4393 NETIF_MSG_CLASS_COUNT, 4394 }; 4395 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4396 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4397 4398 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4399 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4400 4401 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4402 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4403 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4404 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4405 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4406 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4407 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4408 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4409 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4410 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4411 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4412 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4413 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4414 #define NETIF_MSG_HW __NETIF_MSG(HW) 4415 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4416 4417 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4418 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4419 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4420 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4421 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4422 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4423 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4424 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4425 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4426 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4427 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4428 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4429 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4430 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4431 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4432 4433 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4434 { 4435 /* use default */ 4436 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4437 return default_msg_enable_bits; 4438 if (debug_value == 0) /* no output */ 4439 return 0; 4440 /* set low N bits */ 4441 return (1U << debug_value) - 1; 4442 } 4443 4444 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4445 { 4446 spin_lock(&txq->_xmit_lock); 4447 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4448 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4449 } 4450 4451 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4452 { 4453 __acquire(&txq->_xmit_lock); 4454 return true; 4455 } 4456 4457 static inline void __netif_tx_release(struct netdev_queue *txq) 4458 { 4459 __release(&txq->_xmit_lock); 4460 } 4461 4462 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4463 { 4464 spin_lock_bh(&txq->_xmit_lock); 4465 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4466 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4467 } 4468 4469 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4470 { 4471 bool ok = spin_trylock(&txq->_xmit_lock); 4472 4473 if (likely(ok)) { 4474 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4475 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4476 } 4477 return ok; 4478 } 4479 4480 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4481 { 4482 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4483 WRITE_ONCE(txq->xmit_lock_owner, -1); 4484 spin_unlock(&txq->_xmit_lock); 4485 } 4486 4487 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4488 { 4489 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4490 WRITE_ONCE(txq->xmit_lock_owner, -1); 4491 spin_unlock_bh(&txq->_xmit_lock); 4492 } 4493 4494 /* 4495 * txq->trans_start can be read locklessly from dev_watchdog() 4496 */ 4497 static inline void txq_trans_update(struct netdev_queue *txq) 4498 { 4499 if (txq->xmit_lock_owner != -1) 4500 WRITE_ONCE(txq->trans_start, jiffies); 4501 } 4502 4503 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4504 { 4505 unsigned long now = jiffies; 4506 4507 if (READ_ONCE(txq->trans_start) != now) 4508 WRITE_ONCE(txq->trans_start, now); 4509 } 4510 4511 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4512 static inline void netif_trans_update(struct net_device *dev) 4513 { 4514 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4515 4516 txq_trans_cond_update(txq); 4517 } 4518 4519 /** 4520 * netif_tx_lock - grab network device transmit lock 4521 * @dev: network device 4522 * 4523 * Get network device transmit lock 4524 */ 4525 void netif_tx_lock(struct net_device *dev); 4526 4527 static inline void netif_tx_lock_bh(struct net_device *dev) 4528 { 4529 local_bh_disable(); 4530 netif_tx_lock(dev); 4531 } 4532 4533 void netif_tx_unlock(struct net_device *dev); 4534 4535 static inline void netif_tx_unlock_bh(struct net_device *dev) 4536 { 4537 netif_tx_unlock(dev); 4538 local_bh_enable(); 4539 } 4540 4541 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4542 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4543 __netif_tx_lock(txq, cpu); \ 4544 } else { \ 4545 __netif_tx_acquire(txq); \ 4546 } \ 4547 } 4548 4549 #define HARD_TX_TRYLOCK(dev, txq) \ 4550 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4551 __netif_tx_trylock(txq) : \ 4552 __netif_tx_acquire(txq)) 4553 4554 #define HARD_TX_UNLOCK(dev, txq) { \ 4555 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4556 __netif_tx_unlock(txq); \ 4557 } else { \ 4558 __netif_tx_release(txq); \ 4559 } \ 4560 } 4561 4562 static inline void netif_tx_disable(struct net_device *dev) 4563 { 4564 unsigned int i; 4565 int cpu; 4566 4567 local_bh_disable(); 4568 cpu = smp_processor_id(); 4569 spin_lock(&dev->tx_global_lock); 4570 for (i = 0; i < dev->num_tx_queues; i++) { 4571 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4572 4573 __netif_tx_lock(txq, cpu); 4574 netif_tx_stop_queue(txq); 4575 __netif_tx_unlock(txq); 4576 } 4577 spin_unlock(&dev->tx_global_lock); 4578 local_bh_enable(); 4579 } 4580 4581 static inline void netif_addr_lock(struct net_device *dev) 4582 { 4583 unsigned char nest_level = 0; 4584 4585 #ifdef CONFIG_LOCKDEP 4586 nest_level = dev->nested_level; 4587 #endif 4588 spin_lock_nested(&dev->addr_list_lock, nest_level); 4589 } 4590 4591 static inline void netif_addr_lock_bh(struct net_device *dev) 4592 { 4593 unsigned char nest_level = 0; 4594 4595 #ifdef CONFIG_LOCKDEP 4596 nest_level = dev->nested_level; 4597 #endif 4598 local_bh_disable(); 4599 spin_lock_nested(&dev->addr_list_lock, nest_level); 4600 } 4601 4602 static inline void netif_addr_unlock(struct net_device *dev) 4603 { 4604 spin_unlock(&dev->addr_list_lock); 4605 } 4606 4607 static inline void netif_addr_unlock_bh(struct net_device *dev) 4608 { 4609 spin_unlock_bh(&dev->addr_list_lock); 4610 } 4611 4612 /* 4613 * dev_addrs walker. Should be used only for read access. Call with 4614 * rcu_read_lock held. 4615 */ 4616 #define for_each_dev_addr(dev, ha) \ 4617 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4618 4619 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4620 4621 void ether_setup(struct net_device *dev); 4622 4623 /* Support for loadable net-drivers */ 4624 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4625 unsigned char name_assign_type, 4626 void (*setup)(struct net_device *), 4627 unsigned int txqs, unsigned int rxqs); 4628 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4629 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4630 4631 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4632 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4633 count) 4634 4635 int register_netdev(struct net_device *dev); 4636 void unregister_netdev(struct net_device *dev); 4637 4638 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4639 4640 /* General hardware address lists handling functions */ 4641 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4642 struct netdev_hw_addr_list *from_list, int addr_len); 4643 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4644 struct netdev_hw_addr_list *from_list, int addr_len); 4645 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4646 struct net_device *dev, 4647 int (*sync)(struct net_device *, const unsigned char *), 4648 int (*unsync)(struct net_device *, 4649 const unsigned char *)); 4650 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4651 struct net_device *dev, 4652 int (*sync)(struct net_device *, 4653 const unsigned char *, int), 4654 int (*unsync)(struct net_device *, 4655 const unsigned char *, int)); 4656 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4657 struct net_device *dev, 4658 int (*unsync)(struct net_device *, 4659 const unsigned char *, int)); 4660 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4661 struct net_device *dev, 4662 int (*unsync)(struct net_device *, 4663 const unsigned char *)); 4664 void __hw_addr_init(struct netdev_hw_addr_list *list); 4665 4666 /* Functions used for device addresses handling */ 4667 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4668 const void *addr, size_t len); 4669 4670 static inline void 4671 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4672 { 4673 dev_addr_mod(dev, 0, addr, len); 4674 } 4675 4676 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4677 { 4678 __dev_addr_set(dev, addr, dev->addr_len); 4679 } 4680 4681 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4682 unsigned char addr_type); 4683 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4684 unsigned char addr_type); 4685 4686 /* Functions used for unicast addresses handling */ 4687 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4688 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4689 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4690 int dev_uc_sync(struct net_device *to, struct net_device *from); 4691 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4692 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4693 void dev_uc_flush(struct net_device *dev); 4694 void dev_uc_init(struct net_device *dev); 4695 4696 /** 4697 * __dev_uc_sync - Synchonize device's unicast list 4698 * @dev: device to sync 4699 * @sync: function to call if address should be added 4700 * @unsync: function to call if address should be removed 4701 * 4702 * Add newly added addresses to the interface, and release 4703 * addresses that have been deleted. 4704 */ 4705 static inline int __dev_uc_sync(struct net_device *dev, 4706 int (*sync)(struct net_device *, 4707 const unsigned char *), 4708 int (*unsync)(struct net_device *, 4709 const unsigned char *)) 4710 { 4711 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4712 } 4713 4714 /** 4715 * __dev_uc_unsync - Remove synchronized addresses from device 4716 * @dev: device to sync 4717 * @unsync: function to call if address should be removed 4718 * 4719 * Remove all addresses that were added to the device by dev_uc_sync(). 4720 */ 4721 static inline void __dev_uc_unsync(struct net_device *dev, 4722 int (*unsync)(struct net_device *, 4723 const unsigned char *)) 4724 { 4725 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4726 } 4727 4728 /* Functions used for multicast addresses handling */ 4729 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4730 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4731 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4732 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4733 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4734 int dev_mc_sync(struct net_device *to, struct net_device *from); 4735 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4736 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4737 void dev_mc_flush(struct net_device *dev); 4738 void dev_mc_init(struct net_device *dev); 4739 4740 /** 4741 * __dev_mc_sync - Synchonize device's multicast list 4742 * @dev: device to sync 4743 * @sync: function to call if address should be added 4744 * @unsync: function to call if address should be removed 4745 * 4746 * Add newly added addresses to the interface, and release 4747 * addresses that have been deleted. 4748 */ 4749 static inline int __dev_mc_sync(struct net_device *dev, 4750 int (*sync)(struct net_device *, 4751 const unsigned char *), 4752 int (*unsync)(struct net_device *, 4753 const unsigned char *)) 4754 { 4755 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4756 } 4757 4758 /** 4759 * __dev_mc_unsync - Remove synchronized addresses from device 4760 * @dev: device to sync 4761 * @unsync: function to call if address should be removed 4762 * 4763 * Remove all addresses that were added to the device by dev_mc_sync(). 4764 */ 4765 static inline void __dev_mc_unsync(struct net_device *dev, 4766 int (*unsync)(struct net_device *, 4767 const unsigned char *)) 4768 { 4769 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4770 } 4771 4772 /* Functions used for secondary unicast and multicast support */ 4773 void dev_set_rx_mode(struct net_device *dev); 4774 int dev_set_promiscuity(struct net_device *dev, int inc); 4775 int dev_set_allmulti(struct net_device *dev, int inc); 4776 void netdev_state_change(struct net_device *dev); 4777 void __netdev_notify_peers(struct net_device *dev); 4778 void netdev_notify_peers(struct net_device *dev); 4779 void netdev_features_change(struct net_device *dev); 4780 /* Load a device via the kmod */ 4781 void dev_load(struct net *net, const char *name); 4782 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4783 struct rtnl_link_stats64 *storage); 4784 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4785 const struct net_device_stats *netdev_stats); 4786 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4787 const struct pcpu_sw_netstats __percpu *netstats); 4788 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4789 4790 extern int netdev_max_backlog; 4791 extern int dev_rx_weight; 4792 extern int dev_tx_weight; 4793 extern int gro_normal_batch; 4794 4795 enum { 4796 NESTED_SYNC_IMM_BIT, 4797 NESTED_SYNC_TODO_BIT, 4798 }; 4799 4800 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4801 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4802 4803 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4804 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4805 4806 struct netdev_nested_priv { 4807 unsigned char flags; 4808 void *data; 4809 }; 4810 4811 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4812 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4813 struct list_head **iter); 4814 4815 /* iterate through upper list, must be called under RCU read lock */ 4816 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4817 for (iter = &(dev)->adj_list.upper, \ 4818 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4819 updev; \ 4820 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4821 4822 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4823 int (*fn)(struct net_device *upper_dev, 4824 struct netdev_nested_priv *priv), 4825 struct netdev_nested_priv *priv); 4826 4827 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4828 struct net_device *upper_dev); 4829 4830 bool netdev_has_any_upper_dev(struct net_device *dev); 4831 4832 void *netdev_lower_get_next_private(struct net_device *dev, 4833 struct list_head **iter); 4834 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4835 struct list_head **iter); 4836 4837 #define netdev_for_each_lower_private(dev, priv, iter) \ 4838 for (iter = (dev)->adj_list.lower.next, \ 4839 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4840 priv; \ 4841 priv = netdev_lower_get_next_private(dev, &(iter))) 4842 4843 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4844 for (iter = &(dev)->adj_list.lower, \ 4845 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4846 priv; \ 4847 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4848 4849 void *netdev_lower_get_next(struct net_device *dev, 4850 struct list_head **iter); 4851 4852 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4853 for (iter = (dev)->adj_list.lower.next, \ 4854 ldev = netdev_lower_get_next(dev, &(iter)); \ 4855 ldev; \ 4856 ldev = netdev_lower_get_next(dev, &(iter))) 4857 4858 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4859 struct list_head **iter); 4860 int netdev_walk_all_lower_dev(struct net_device *dev, 4861 int (*fn)(struct net_device *lower_dev, 4862 struct netdev_nested_priv *priv), 4863 struct netdev_nested_priv *priv); 4864 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4865 int (*fn)(struct net_device *lower_dev, 4866 struct netdev_nested_priv *priv), 4867 struct netdev_nested_priv *priv); 4868 4869 void *netdev_adjacent_get_private(struct list_head *adj_list); 4870 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4871 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4872 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4873 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4874 struct netlink_ext_ack *extack); 4875 int netdev_master_upper_dev_link(struct net_device *dev, 4876 struct net_device *upper_dev, 4877 void *upper_priv, void *upper_info, 4878 struct netlink_ext_ack *extack); 4879 void netdev_upper_dev_unlink(struct net_device *dev, 4880 struct net_device *upper_dev); 4881 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4882 struct net_device *new_dev, 4883 struct net_device *dev, 4884 struct netlink_ext_ack *extack); 4885 void netdev_adjacent_change_commit(struct net_device *old_dev, 4886 struct net_device *new_dev, 4887 struct net_device *dev); 4888 void netdev_adjacent_change_abort(struct net_device *old_dev, 4889 struct net_device *new_dev, 4890 struct net_device *dev); 4891 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4892 void *netdev_lower_dev_get_private(struct net_device *dev, 4893 struct net_device *lower_dev); 4894 void netdev_lower_state_changed(struct net_device *lower_dev, 4895 void *lower_state_info); 4896 4897 /* RSS keys are 40 or 52 bytes long */ 4898 #define NETDEV_RSS_KEY_LEN 52 4899 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4900 void netdev_rss_key_fill(void *buffer, size_t len); 4901 4902 int skb_checksum_help(struct sk_buff *skb); 4903 int skb_crc32c_csum_help(struct sk_buff *skb); 4904 int skb_csum_hwoffload_help(struct sk_buff *skb, 4905 const netdev_features_t features); 4906 4907 struct netdev_bonding_info { 4908 ifslave slave; 4909 ifbond master; 4910 }; 4911 4912 struct netdev_notifier_bonding_info { 4913 struct netdev_notifier_info info; /* must be first */ 4914 struct netdev_bonding_info bonding_info; 4915 }; 4916 4917 void netdev_bonding_info_change(struct net_device *dev, 4918 struct netdev_bonding_info *bonding_info); 4919 4920 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4921 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4922 #else 4923 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4924 const void *data) 4925 { 4926 } 4927 #endif 4928 4929 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4930 4931 static inline bool can_checksum_protocol(netdev_features_t features, 4932 __be16 protocol) 4933 { 4934 if (protocol == htons(ETH_P_FCOE)) 4935 return !!(features & NETIF_F_FCOE_CRC); 4936 4937 /* Assume this is an IP checksum (not SCTP CRC) */ 4938 4939 if (features & NETIF_F_HW_CSUM) { 4940 /* Can checksum everything */ 4941 return true; 4942 } 4943 4944 switch (protocol) { 4945 case htons(ETH_P_IP): 4946 return !!(features & NETIF_F_IP_CSUM); 4947 case htons(ETH_P_IPV6): 4948 return !!(features & NETIF_F_IPV6_CSUM); 4949 default: 4950 return false; 4951 } 4952 } 4953 4954 #ifdef CONFIG_BUG 4955 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4956 #else 4957 static inline void netdev_rx_csum_fault(struct net_device *dev, 4958 struct sk_buff *skb) 4959 { 4960 } 4961 #endif 4962 /* rx skb timestamps */ 4963 void net_enable_timestamp(void); 4964 void net_disable_timestamp(void); 4965 4966 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 4967 const struct skb_shared_hwtstamps *hwtstamps, 4968 bool cycles) 4969 { 4970 const struct net_device_ops *ops = dev->netdev_ops; 4971 4972 if (ops->ndo_get_tstamp) 4973 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 4974 4975 return hwtstamps->hwtstamp; 4976 } 4977 4978 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4979 struct sk_buff *skb, struct net_device *dev, 4980 bool more) 4981 { 4982 __this_cpu_write(softnet_data.xmit.more, more); 4983 return ops->ndo_start_xmit(skb, dev); 4984 } 4985 4986 static inline bool netdev_xmit_more(void) 4987 { 4988 return __this_cpu_read(softnet_data.xmit.more); 4989 } 4990 4991 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4992 struct netdev_queue *txq, bool more) 4993 { 4994 const struct net_device_ops *ops = dev->netdev_ops; 4995 netdev_tx_t rc; 4996 4997 rc = __netdev_start_xmit(ops, skb, dev, more); 4998 if (rc == NETDEV_TX_OK) 4999 txq_trans_update(txq); 5000 5001 return rc; 5002 } 5003 5004 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 5005 const void *ns); 5006 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 5007 const void *ns); 5008 5009 extern const struct kobj_ns_type_operations net_ns_type_operations; 5010 5011 const char *netdev_drivername(const struct net_device *dev); 5012 5013 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 5014 netdev_features_t f2) 5015 { 5016 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 5017 if (f1 & NETIF_F_HW_CSUM) 5018 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5019 else 5020 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5021 } 5022 5023 return f1 & f2; 5024 } 5025 5026 static inline netdev_features_t netdev_get_wanted_features( 5027 struct net_device *dev) 5028 { 5029 return (dev->features & ~dev->hw_features) | dev->wanted_features; 5030 } 5031 netdev_features_t netdev_increment_features(netdev_features_t all, 5032 netdev_features_t one, netdev_features_t mask); 5033 5034 /* Allow TSO being used on stacked device : 5035 * Performing the GSO segmentation before last device 5036 * is a performance improvement. 5037 */ 5038 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 5039 netdev_features_t mask) 5040 { 5041 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 5042 } 5043 5044 int __netdev_update_features(struct net_device *dev); 5045 void netdev_update_features(struct net_device *dev); 5046 void netdev_change_features(struct net_device *dev); 5047 5048 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 5049 struct net_device *dev); 5050 5051 netdev_features_t passthru_features_check(struct sk_buff *skb, 5052 struct net_device *dev, 5053 netdev_features_t features); 5054 netdev_features_t netif_skb_features(struct sk_buff *skb); 5055 void skb_warn_bad_offload(const struct sk_buff *skb); 5056 5057 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 5058 { 5059 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 5060 5061 /* check flags correspondence */ 5062 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 5063 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 5064 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 5065 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 5066 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 5067 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 5068 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 5069 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 5070 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 5071 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 5072 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 5073 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 5074 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 5075 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 5076 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 5077 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 5078 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 5079 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 5080 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 5081 5082 return (features & feature) == feature; 5083 } 5084 5085 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 5086 { 5087 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 5088 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 5089 } 5090 5091 static inline bool netif_needs_gso(struct sk_buff *skb, 5092 netdev_features_t features) 5093 { 5094 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 5095 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 5096 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 5097 } 5098 5099 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 5100 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 5101 void netif_inherit_tso_max(struct net_device *to, 5102 const struct net_device *from); 5103 5104 static inline bool netif_is_macsec(const struct net_device *dev) 5105 { 5106 return dev->priv_flags & IFF_MACSEC; 5107 } 5108 5109 static inline bool netif_is_macvlan(const struct net_device *dev) 5110 { 5111 return dev->priv_flags & IFF_MACVLAN; 5112 } 5113 5114 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5115 { 5116 return dev->priv_flags & IFF_MACVLAN_PORT; 5117 } 5118 5119 static inline bool netif_is_bond_master(const struct net_device *dev) 5120 { 5121 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5122 } 5123 5124 static inline bool netif_is_bond_slave(const struct net_device *dev) 5125 { 5126 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5127 } 5128 5129 static inline bool netif_supports_nofcs(struct net_device *dev) 5130 { 5131 return dev->priv_flags & IFF_SUPP_NOFCS; 5132 } 5133 5134 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5135 { 5136 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5137 } 5138 5139 static inline bool netif_is_l3_master(const struct net_device *dev) 5140 { 5141 return dev->priv_flags & IFF_L3MDEV_MASTER; 5142 } 5143 5144 static inline bool netif_is_l3_slave(const struct net_device *dev) 5145 { 5146 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5147 } 5148 5149 static inline int dev_sdif(const struct net_device *dev) 5150 { 5151 #ifdef CONFIG_NET_L3_MASTER_DEV 5152 if (netif_is_l3_slave(dev)) 5153 return dev->ifindex; 5154 #endif 5155 return 0; 5156 } 5157 5158 static inline bool netif_is_bridge_master(const struct net_device *dev) 5159 { 5160 return dev->priv_flags & IFF_EBRIDGE; 5161 } 5162 5163 static inline bool netif_is_bridge_port(const struct net_device *dev) 5164 { 5165 return dev->priv_flags & IFF_BRIDGE_PORT; 5166 } 5167 5168 static inline bool netif_is_ovs_master(const struct net_device *dev) 5169 { 5170 return dev->priv_flags & IFF_OPENVSWITCH; 5171 } 5172 5173 static inline bool netif_is_ovs_port(const struct net_device *dev) 5174 { 5175 return dev->priv_flags & IFF_OVS_DATAPATH; 5176 } 5177 5178 static inline bool netif_is_any_bridge_master(const struct net_device *dev) 5179 { 5180 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev); 5181 } 5182 5183 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5184 { 5185 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5186 } 5187 5188 static inline bool netif_is_team_master(const struct net_device *dev) 5189 { 5190 return dev->priv_flags & IFF_TEAM; 5191 } 5192 5193 static inline bool netif_is_team_port(const struct net_device *dev) 5194 { 5195 return dev->priv_flags & IFF_TEAM_PORT; 5196 } 5197 5198 static inline bool netif_is_lag_master(const struct net_device *dev) 5199 { 5200 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5201 } 5202 5203 static inline bool netif_is_lag_port(const struct net_device *dev) 5204 { 5205 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5206 } 5207 5208 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5209 { 5210 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5211 } 5212 5213 static inline bool netif_is_failover(const struct net_device *dev) 5214 { 5215 return dev->priv_flags & IFF_FAILOVER; 5216 } 5217 5218 static inline bool netif_is_failover_slave(const struct net_device *dev) 5219 { 5220 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5221 } 5222 5223 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5224 static inline void netif_keep_dst(struct net_device *dev) 5225 { 5226 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5227 } 5228 5229 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5230 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5231 { 5232 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5233 return netif_is_macsec(dev); 5234 } 5235 5236 extern struct pernet_operations __net_initdata loopback_net_ops; 5237 5238 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5239 5240 /* netdev_printk helpers, similar to dev_printk */ 5241 5242 static inline const char *netdev_name(const struct net_device *dev) 5243 { 5244 if (!dev->name[0] || strchr(dev->name, '%')) 5245 return "(unnamed net_device)"; 5246 return dev->name; 5247 } 5248 5249 static inline const char *netdev_reg_state(const struct net_device *dev) 5250 { 5251 switch (dev->reg_state) { 5252 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5253 case NETREG_REGISTERED: return ""; 5254 case NETREG_UNREGISTERING: return " (unregistering)"; 5255 case NETREG_UNREGISTERED: return " (unregistered)"; 5256 case NETREG_RELEASED: return " (released)"; 5257 case NETREG_DUMMY: return " (dummy)"; 5258 } 5259 5260 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5261 return " (unknown)"; 5262 } 5263 5264 #define MODULE_ALIAS_NETDEV(device) \ 5265 MODULE_ALIAS("netdev-" device) 5266 5267 /* 5268 * netdev_WARN() acts like dev_printk(), but with the key difference 5269 * of using a WARN/WARN_ON to get the message out, including the 5270 * file/line information and a backtrace. 5271 */ 5272 #define netdev_WARN(dev, format, args...) \ 5273 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5274 netdev_reg_state(dev), ##args) 5275 5276 #define netdev_WARN_ONCE(dev, format, args...) \ 5277 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5278 netdev_reg_state(dev), ##args) 5279 5280 /* 5281 * The list of packet types we will receive (as opposed to discard) 5282 * and the routines to invoke. 5283 * 5284 * Why 16. Because with 16 the only overlap we get on a hash of the 5285 * low nibble of the protocol value is RARP/SNAP/X.25. 5286 * 5287 * 0800 IP 5288 * 0001 802.3 5289 * 0002 AX.25 5290 * 0004 802.2 5291 * 8035 RARP 5292 * 0005 SNAP 5293 * 0805 X.25 5294 * 0806 ARP 5295 * 8137 IPX 5296 * 0009 Localtalk 5297 * 86DD IPv6 5298 */ 5299 #define PTYPE_HASH_SIZE (16) 5300 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5301 5302 extern struct list_head ptype_all __read_mostly; 5303 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5304 5305 extern struct net_device *blackhole_netdev; 5306 5307 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */ 5308 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD) 5309 #define DEV_STATS_ADD(DEV, FIELD, VAL) \ 5310 atomic_long_add((VAL), &(DEV)->stats.__##FIELD) 5311 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD) 5312 5313 #endif /* _LINUX_NETDEVICE_H */ 5314