1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <[email protected]> 13 * Corey Minyard <[email protected]> 14 * Donald J. Becker, <[email protected]> 15 * Alan Cox, <[email protected]> 16 * Bjorn Ekwall. <[email protected]> 17 * Pekka Riikonen <[email protected]> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 #include <asm/local.h> 32 33 #include <linux/percpu.h> 34 #include <linux/rculist.h> 35 #include <linux/workqueue.h> 36 #include <linux/dynamic_queue_limits.h> 37 38 #include <net/net_namespace.h> 39 #ifdef CONFIG_DCB 40 #include <net/dcbnl.h> 41 #endif 42 #include <net/netprio_cgroup.h> 43 #include <linux/netdev_features.h> 44 #include <linux/neighbour.h> 45 #include <linux/netdevice_xmit.h> 46 #include <uapi/linux/netdevice.h> 47 #include <uapi/linux/if_bonding.h> 48 #include <uapi/linux/pkt_cls.h> 49 #include <uapi/linux/netdev.h> 50 #include <linux/hashtable.h> 51 #include <linux/rbtree.h> 52 #include <net/net_trackers.h> 53 #include <net/net_debug.h> 54 #include <net/dropreason-core.h> 55 #include <net/neighbour_tables.h> 56 57 struct netpoll_info; 58 struct device; 59 struct ethtool_ops; 60 struct kernel_hwtstamp_config; 61 struct phy_device; 62 struct dsa_port; 63 struct ip_tunnel_parm_kern; 64 struct macsec_context; 65 struct macsec_ops; 66 struct netdev_name_node; 67 struct sd_flow_limit; 68 struct sfp_bus; 69 /* 802.11 specific */ 70 struct wireless_dev; 71 /* 802.15.4 specific */ 72 struct wpan_dev; 73 struct mpls_dev; 74 /* UDP Tunnel offloads */ 75 struct udp_tunnel_info; 76 struct udp_tunnel_nic_info; 77 struct udp_tunnel_nic; 78 struct bpf_prog; 79 struct xdp_buff; 80 struct xdp_frame; 81 struct xdp_metadata_ops; 82 struct xdp_md; 83 struct ethtool_netdev_state; 84 struct phy_link_topology; 85 86 typedef u32 xdp_features_t; 87 88 void synchronize_net(void); 89 void netdev_set_default_ethtool_ops(struct net_device *dev, 90 const struct ethtool_ops *ops); 91 void netdev_sw_irq_coalesce_default_on(struct net_device *dev); 92 93 /* Backlog congestion levels */ 94 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 95 #define NET_RX_DROP 1 /* packet dropped */ 96 97 #define MAX_NEST_DEV 8 98 99 /* 100 * Transmit return codes: transmit return codes originate from three different 101 * namespaces: 102 * 103 * - qdisc return codes 104 * - driver transmit return codes 105 * - errno values 106 * 107 * Drivers are allowed to return any one of those in their hard_start_xmit() 108 * function. Real network devices commonly used with qdiscs should only return 109 * the driver transmit return codes though - when qdiscs are used, the actual 110 * transmission happens asynchronously, so the value is not propagated to 111 * higher layers. Virtual network devices transmit synchronously; in this case 112 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 113 * others are propagated to higher layers. 114 */ 115 116 /* qdisc ->enqueue() return codes. */ 117 #define NET_XMIT_SUCCESS 0x00 118 #define NET_XMIT_DROP 0x01 /* skb dropped */ 119 #define NET_XMIT_CN 0x02 /* congestion notification */ 120 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 121 122 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 123 * indicates that the device will soon be dropping packets, or already drops 124 * some packets of the same priority; prompting us to send less aggressively. */ 125 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 126 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 127 128 /* Driver transmit return codes */ 129 #define NETDEV_TX_MASK 0xf0 130 131 enum netdev_tx { 132 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 133 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 134 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 135 }; 136 typedef enum netdev_tx netdev_tx_t; 137 138 /* 139 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 140 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 141 */ 142 static inline bool dev_xmit_complete(int rc) 143 { 144 /* 145 * Positive cases with an skb consumed by a driver: 146 * - successful transmission (rc == NETDEV_TX_OK) 147 * - error while transmitting (rc < 0) 148 * - error while queueing to a different device (rc & NET_XMIT_MASK) 149 */ 150 if (likely(rc < NET_XMIT_MASK)) 151 return true; 152 153 return false; 154 } 155 156 /* 157 * Compute the worst-case header length according to the protocols 158 * used. 159 */ 160 161 #if defined(CONFIG_HYPERV_NET) 162 # define LL_MAX_HEADER 128 163 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 164 # if defined(CONFIG_MAC80211_MESH) 165 # define LL_MAX_HEADER 128 166 # else 167 # define LL_MAX_HEADER 96 168 # endif 169 #else 170 # define LL_MAX_HEADER 32 171 #endif 172 173 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 174 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 175 #define MAX_HEADER LL_MAX_HEADER 176 #else 177 #define MAX_HEADER (LL_MAX_HEADER + 48) 178 #endif 179 180 /* 181 * Old network device statistics. Fields are native words 182 * (unsigned long) so they can be read and written atomically. 183 */ 184 185 #define NET_DEV_STAT(FIELD) \ 186 union { \ 187 unsigned long FIELD; \ 188 atomic_long_t __##FIELD; \ 189 } 190 191 struct net_device_stats { 192 NET_DEV_STAT(rx_packets); 193 NET_DEV_STAT(tx_packets); 194 NET_DEV_STAT(rx_bytes); 195 NET_DEV_STAT(tx_bytes); 196 NET_DEV_STAT(rx_errors); 197 NET_DEV_STAT(tx_errors); 198 NET_DEV_STAT(rx_dropped); 199 NET_DEV_STAT(tx_dropped); 200 NET_DEV_STAT(multicast); 201 NET_DEV_STAT(collisions); 202 NET_DEV_STAT(rx_length_errors); 203 NET_DEV_STAT(rx_over_errors); 204 NET_DEV_STAT(rx_crc_errors); 205 NET_DEV_STAT(rx_frame_errors); 206 NET_DEV_STAT(rx_fifo_errors); 207 NET_DEV_STAT(rx_missed_errors); 208 NET_DEV_STAT(tx_aborted_errors); 209 NET_DEV_STAT(tx_carrier_errors); 210 NET_DEV_STAT(tx_fifo_errors); 211 NET_DEV_STAT(tx_heartbeat_errors); 212 NET_DEV_STAT(tx_window_errors); 213 NET_DEV_STAT(rx_compressed); 214 NET_DEV_STAT(tx_compressed); 215 }; 216 #undef NET_DEV_STAT 217 218 /* per-cpu stats, allocated on demand. 219 * Try to fit them in a single cache line, for dev_get_stats() sake. 220 */ 221 struct net_device_core_stats { 222 unsigned long rx_dropped; 223 unsigned long tx_dropped; 224 unsigned long rx_nohandler; 225 unsigned long rx_otherhost_dropped; 226 } __aligned(4 * sizeof(unsigned long)); 227 228 #include <linux/cache.h> 229 #include <linux/skbuff.h> 230 231 struct neighbour; 232 struct neigh_parms; 233 struct sk_buff; 234 235 struct netdev_hw_addr { 236 struct list_head list; 237 struct rb_node node; 238 unsigned char addr[MAX_ADDR_LEN]; 239 unsigned char type; 240 #define NETDEV_HW_ADDR_T_LAN 1 241 #define NETDEV_HW_ADDR_T_SAN 2 242 #define NETDEV_HW_ADDR_T_UNICAST 3 243 #define NETDEV_HW_ADDR_T_MULTICAST 4 244 bool global_use; 245 int sync_cnt; 246 int refcount; 247 int synced; 248 struct rcu_head rcu_head; 249 }; 250 251 struct netdev_hw_addr_list { 252 struct list_head list; 253 int count; 254 255 /* Auxiliary tree for faster lookup on addition and deletion */ 256 struct rb_root tree; 257 }; 258 259 #define netdev_hw_addr_list_count(l) ((l)->count) 260 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 261 #define netdev_hw_addr_list_for_each(ha, l) \ 262 list_for_each_entry(ha, &(l)->list, list) 263 264 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 265 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 266 #define netdev_for_each_uc_addr(ha, dev) \ 267 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 268 #define netdev_for_each_synced_uc_addr(_ha, _dev) \ 269 netdev_for_each_uc_addr((_ha), (_dev)) \ 270 if ((_ha)->sync_cnt) 271 272 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 273 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 274 #define netdev_for_each_mc_addr(ha, dev) \ 275 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 276 #define netdev_for_each_synced_mc_addr(_ha, _dev) \ 277 netdev_for_each_mc_addr((_ha), (_dev)) \ 278 if ((_ha)->sync_cnt) 279 280 struct hh_cache { 281 unsigned int hh_len; 282 seqlock_t hh_lock; 283 284 /* cached hardware header; allow for machine alignment needs. */ 285 #define HH_DATA_MOD 16 286 #define HH_DATA_OFF(__len) \ 287 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 288 #define HH_DATA_ALIGN(__len) \ 289 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 290 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 291 }; 292 293 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 294 * Alternative is: 295 * dev->hard_header_len ? (dev->hard_header_len + 296 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 297 * 298 * We could use other alignment values, but we must maintain the 299 * relationship HH alignment <= LL alignment. 300 */ 301 #define LL_RESERVED_SPACE(dev) \ 302 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \ 303 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 304 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 305 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \ 306 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 307 308 struct header_ops { 309 int (*create) (struct sk_buff *skb, struct net_device *dev, 310 unsigned short type, const void *daddr, 311 const void *saddr, unsigned int len); 312 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 313 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 314 void (*cache_update)(struct hh_cache *hh, 315 const struct net_device *dev, 316 const unsigned char *haddr); 317 bool (*validate)(const char *ll_header, unsigned int len); 318 __be16 (*parse_protocol)(const struct sk_buff *skb); 319 }; 320 321 /* These flag bits are private to the generic network queueing 322 * layer; they may not be explicitly referenced by any other 323 * code. 324 */ 325 326 enum netdev_state_t { 327 __LINK_STATE_START, 328 __LINK_STATE_PRESENT, 329 __LINK_STATE_NOCARRIER, 330 __LINK_STATE_LINKWATCH_PENDING, 331 __LINK_STATE_DORMANT, 332 __LINK_STATE_TESTING, 333 }; 334 335 struct gro_list { 336 struct list_head list; 337 int count; 338 }; 339 340 /* 341 * size of gro hash buckets, must less than bit number of 342 * napi_struct::gro_bitmask 343 */ 344 #define GRO_HASH_BUCKETS 8 345 346 /* 347 * Structure for per-NAPI config 348 */ 349 struct napi_config { 350 u64 gro_flush_timeout; 351 u64 irq_suspend_timeout; 352 u32 defer_hard_irqs; 353 unsigned int napi_id; 354 }; 355 356 /* 357 * Structure for NAPI scheduling similar to tasklet but with weighting 358 */ 359 struct napi_struct { 360 /* The poll_list must only be managed by the entity which 361 * changes the state of the NAPI_STATE_SCHED bit. This means 362 * whoever atomically sets that bit can add this napi_struct 363 * to the per-CPU poll_list, and whoever clears that bit 364 * can remove from the list right before clearing the bit. 365 */ 366 struct list_head poll_list; 367 368 unsigned long state; 369 int weight; 370 u32 defer_hard_irqs_count; 371 unsigned long gro_bitmask; 372 int (*poll)(struct napi_struct *, int); 373 #ifdef CONFIG_NETPOLL 374 /* CPU actively polling if netpoll is configured */ 375 int poll_owner; 376 #endif 377 /* CPU on which NAPI has been scheduled for processing */ 378 int list_owner; 379 struct net_device *dev; 380 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 381 struct sk_buff *skb; 382 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 383 int rx_count; /* length of rx_list */ 384 unsigned int napi_id; 385 struct hrtimer timer; 386 struct task_struct *thread; 387 unsigned long gro_flush_timeout; 388 unsigned long irq_suspend_timeout; 389 u32 defer_hard_irqs; 390 /* control-path-only fields follow */ 391 struct list_head dev_list; 392 struct hlist_node napi_hash_node; 393 int irq; 394 int index; 395 struct napi_config *config; 396 }; 397 398 enum { 399 NAPI_STATE_SCHED, /* Poll is scheduled */ 400 NAPI_STATE_MISSED, /* reschedule a napi */ 401 NAPI_STATE_DISABLE, /* Disable pending */ 402 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 403 NAPI_STATE_LISTED, /* NAPI added to system lists */ 404 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 405 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 406 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 407 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 408 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 409 }; 410 411 enum { 412 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 413 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 414 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 415 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 416 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 417 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 418 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 419 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 420 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 421 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 422 }; 423 424 enum gro_result { 425 GRO_MERGED, 426 GRO_MERGED_FREE, 427 GRO_HELD, 428 GRO_NORMAL, 429 GRO_CONSUMED, 430 }; 431 typedef enum gro_result gro_result_t; 432 433 /* 434 * enum rx_handler_result - Possible return values for rx_handlers. 435 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 436 * further. 437 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 438 * case skb->dev was changed by rx_handler. 439 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 440 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 441 * 442 * rx_handlers are functions called from inside __netif_receive_skb(), to do 443 * special processing of the skb, prior to delivery to protocol handlers. 444 * 445 * Currently, a net_device can only have a single rx_handler registered. Trying 446 * to register a second rx_handler will return -EBUSY. 447 * 448 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 449 * To unregister a rx_handler on a net_device, use 450 * netdev_rx_handler_unregister(). 451 * 452 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 453 * do with the skb. 454 * 455 * If the rx_handler consumed the skb in some way, it should return 456 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 457 * the skb to be delivered in some other way. 458 * 459 * If the rx_handler changed skb->dev, to divert the skb to another 460 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 461 * new device will be called if it exists. 462 * 463 * If the rx_handler decides the skb should be ignored, it should return 464 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 465 * are registered on exact device (ptype->dev == skb->dev). 466 * 467 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 468 * delivered, it should return RX_HANDLER_PASS. 469 * 470 * A device without a registered rx_handler will behave as if rx_handler 471 * returned RX_HANDLER_PASS. 472 */ 473 474 enum rx_handler_result { 475 RX_HANDLER_CONSUMED, 476 RX_HANDLER_ANOTHER, 477 RX_HANDLER_EXACT, 478 RX_HANDLER_PASS, 479 }; 480 typedef enum rx_handler_result rx_handler_result_t; 481 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 482 483 void __napi_schedule(struct napi_struct *n); 484 void __napi_schedule_irqoff(struct napi_struct *n); 485 486 static inline bool napi_disable_pending(struct napi_struct *n) 487 { 488 return test_bit(NAPI_STATE_DISABLE, &n->state); 489 } 490 491 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 492 { 493 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 494 } 495 496 /** 497 * napi_is_scheduled - test if NAPI is scheduled 498 * @n: NAPI context 499 * 500 * This check is "best-effort". With no locking implemented, 501 * a NAPI can be scheduled or terminate right after this check 502 * and produce not precise results. 503 * 504 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled 505 * should not be used normally and napi_schedule should be 506 * used instead. 507 * 508 * Use only if the driver really needs to check if a NAPI 509 * is scheduled for example in the context of delayed timer 510 * that can be skipped if a NAPI is already scheduled. 511 * 512 * Return True if NAPI is scheduled, False otherwise. 513 */ 514 static inline bool napi_is_scheduled(struct napi_struct *n) 515 { 516 return test_bit(NAPI_STATE_SCHED, &n->state); 517 } 518 519 bool napi_schedule_prep(struct napi_struct *n); 520 521 /** 522 * napi_schedule - schedule NAPI poll 523 * @n: NAPI context 524 * 525 * Schedule NAPI poll routine to be called if it is not already 526 * running. 527 * Return true if we schedule a NAPI or false if not. 528 * Refer to napi_schedule_prep() for additional reason on why 529 * a NAPI might not be scheduled. 530 */ 531 static inline bool napi_schedule(struct napi_struct *n) 532 { 533 if (napi_schedule_prep(n)) { 534 __napi_schedule(n); 535 return true; 536 } 537 538 return false; 539 } 540 541 /** 542 * napi_schedule_irqoff - schedule NAPI poll 543 * @n: NAPI context 544 * 545 * Variant of napi_schedule(), assuming hard irqs are masked. 546 */ 547 static inline void napi_schedule_irqoff(struct napi_struct *n) 548 { 549 if (napi_schedule_prep(n)) 550 __napi_schedule_irqoff(n); 551 } 552 553 /** 554 * napi_complete_done - NAPI processing complete 555 * @n: NAPI context 556 * @work_done: number of packets processed 557 * 558 * Mark NAPI processing as complete. Should only be called if poll budget 559 * has not been completely consumed. 560 * Prefer over napi_complete(). 561 * Return false if device should avoid rearming interrupts. 562 */ 563 bool napi_complete_done(struct napi_struct *n, int work_done); 564 565 static inline bool napi_complete(struct napi_struct *n) 566 { 567 return napi_complete_done(n, 0); 568 } 569 570 int dev_set_threaded(struct net_device *dev, bool threaded); 571 572 /** 573 * napi_disable - prevent NAPI from scheduling 574 * @n: NAPI context 575 * 576 * Stop NAPI from being scheduled on this context. 577 * Waits till any outstanding processing completes. 578 */ 579 void napi_disable(struct napi_struct *n); 580 581 void napi_enable(struct napi_struct *n); 582 583 /** 584 * napi_synchronize - wait until NAPI is not running 585 * @n: NAPI context 586 * 587 * Wait until NAPI is done being scheduled on this context. 588 * Waits till any outstanding processing completes but 589 * does not disable future activations. 590 */ 591 static inline void napi_synchronize(const struct napi_struct *n) 592 { 593 if (IS_ENABLED(CONFIG_SMP)) 594 while (test_bit(NAPI_STATE_SCHED, &n->state)) 595 msleep(1); 596 else 597 barrier(); 598 } 599 600 /** 601 * napi_if_scheduled_mark_missed - if napi is running, set the 602 * NAPIF_STATE_MISSED 603 * @n: NAPI context 604 * 605 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 606 * NAPI is scheduled. 607 **/ 608 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 609 { 610 unsigned long val, new; 611 612 val = READ_ONCE(n->state); 613 do { 614 if (val & NAPIF_STATE_DISABLE) 615 return true; 616 617 if (!(val & NAPIF_STATE_SCHED)) 618 return false; 619 620 new = val | NAPIF_STATE_MISSED; 621 } while (!try_cmpxchg(&n->state, &val, new)); 622 623 return true; 624 } 625 626 enum netdev_queue_state_t { 627 __QUEUE_STATE_DRV_XOFF, 628 __QUEUE_STATE_STACK_XOFF, 629 __QUEUE_STATE_FROZEN, 630 }; 631 632 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 633 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 634 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 635 636 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 637 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 638 QUEUE_STATE_FROZEN) 639 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 640 QUEUE_STATE_FROZEN) 641 642 /* 643 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 644 * netif_tx_* functions below are used to manipulate this flag. The 645 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 646 * queue independently. The netif_xmit_*stopped functions below are called 647 * to check if the queue has been stopped by the driver or stack (either 648 * of the XOFF bits are set in the state). Drivers should not need to call 649 * netif_xmit*stopped functions, they should only be using netif_tx_*. 650 */ 651 652 struct netdev_queue { 653 /* 654 * read-mostly part 655 */ 656 struct net_device *dev; 657 netdevice_tracker dev_tracker; 658 659 struct Qdisc __rcu *qdisc; 660 struct Qdisc __rcu *qdisc_sleeping; 661 #ifdef CONFIG_SYSFS 662 struct kobject kobj; 663 #endif 664 unsigned long tx_maxrate; 665 /* 666 * Number of TX timeouts for this queue 667 * (/sys/class/net/DEV/Q/trans_timeout) 668 */ 669 atomic_long_t trans_timeout; 670 671 /* Subordinate device that the queue has been assigned to */ 672 struct net_device *sb_dev; 673 #ifdef CONFIG_XDP_SOCKETS 674 struct xsk_buff_pool *pool; 675 #endif 676 677 /* 678 * write-mostly part 679 */ 680 #ifdef CONFIG_BQL 681 struct dql dql; 682 #endif 683 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 684 int xmit_lock_owner; 685 /* 686 * Time (in jiffies) of last Tx 687 */ 688 unsigned long trans_start; 689 690 unsigned long state; 691 692 /* 693 * slow- / control-path part 694 */ 695 /* NAPI instance for the queue 696 * Readers and writers must hold RTNL 697 */ 698 struct napi_struct *napi; 699 700 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 701 int numa_node; 702 #endif 703 } ____cacheline_aligned_in_smp; 704 705 extern int sysctl_fb_tunnels_only_for_init_net; 706 extern int sysctl_devconf_inherit_init_net; 707 708 /* 709 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 710 * == 1 : For initns only 711 * == 2 : For none. 712 */ 713 static inline bool net_has_fallback_tunnels(const struct net *net) 714 { 715 #if IS_ENABLED(CONFIG_SYSCTL) 716 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net); 717 718 return !fb_tunnels_only_for_init_net || 719 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1); 720 #else 721 return true; 722 #endif 723 } 724 725 static inline int net_inherit_devconf(void) 726 { 727 #if IS_ENABLED(CONFIG_SYSCTL) 728 return READ_ONCE(sysctl_devconf_inherit_init_net); 729 #else 730 return 0; 731 #endif 732 } 733 734 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 735 { 736 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 737 return q->numa_node; 738 #else 739 return NUMA_NO_NODE; 740 #endif 741 } 742 743 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 744 { 745 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 746 q->numa_node = node; 747 #endif 748 } 749 750 #ifdef CONFIG_RFS_ACCEL 751 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 752 u16 filter_id); 753 #endif 754 755 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 756 enum xps_map_type { 757 XPS_CPUS = 0, 758 XPS_RXQS, 759 XPS_MAPS_MAX, 760 }; 761 762 #ifdef CONFIG_XPS 763 /* 764 * This structure holds an XPS map which can be of variable length. The 765 * map is an array of queues. 766 */ 767 struct xps_map { 768 unsigned int len; 769 unsigned int alloc_len; 770 struct rcu_head rcu; 771 u16 queues[]; 772 }; 773 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 774 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 775 - sizeof(struct xps_map)) / sizeof(u16)) 776 777 /* 778 * This structure holds all XPS maps for device. Maps are indexed by CPU. 779 * 780 * We keep track of the number of cpus/rxqs used when the struct is allocated, 781 * in nr_ids. This will help not accessing out-of-bound memory. 782 * 783 * We keep track of the number of traffic classes used when the struct is 784 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 785 * not crossing its upper bound, as the original dev->num_tc can be updated in 786 * the meantime. 787 */ 788 struct xps_dev_maps { 789 struct rcu_head rcu; 790 unsigned int nr_ids; 791 s16 num_tc; 792 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 793 }; 794 795 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 796 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 797 798 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 799 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 800 801 #endif /* CONFIG_XPS */ 802 803 #define TC_MAX_QUEUE 16 804 #define TC_BITMASK 15 805 /* HW offloaded queuing disciplines txq count and offset maps */ 806 struct netdev_tc_txq { 807 u16 count; 808 u16 offset; 809 }; 810 811 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 812 /* 813 * This structure is to hold information about the device 814 * configured to run FCoE protocol stack. 815 */ 816 struct netdev_fcoe_hbainfo { 817 char manufacturer[64]; 818 char serial_number[64]; 819 char hardware_version[64]; 820 char driver_version[64]; 821 char optionrom_version[64]; 822 char firmware_version[64]; 823 char model[256]; 824 char model_description[256]; 825 }; 826 #endif 827 828 #define MAX_PHYS_ITEM_ID_LEN 32 829 830 /* This structure holds a unique identifier to identify some 831 * physical item (port for example) used by a netdevice. 832 */ 833 struct netdev_phys_item_id { 834 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 835 unsigned char id_len; 836 }; 837 838 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 839 struct netdev_phys_item_id *b) 840 { 841 return a->id_len == b->id_len && 842 memcmp(a->id, b->id, a->id_len) == 0; 843 } 844 845 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 846 struct sk_buff *skb, 847 struct net_device *sb_dev); 848 849 enum net_device_path_type { 850 DEV_PATH_ETHERNET = 0, 851 DEV_PATH_VLAN, 852 DEV_PATH_BRIDGE, 853 DEV_PATH_PPPOE, 854 DEV_PATH_DSA, 855 DEV_PATH_MTK_WDMA, 856 }; 857 858 struct net_device_path { 859 enum net_device_path_type type; 860 const struct net_device *dev; 861 union { 862 struct { 863 u16 id; 864 __be16 proto; 865 u8 h_dest[ETH_ALEN]; 866 } encap; 867 struct { 868 enum { 869 DEV_PATH_BR_VLAN_KEEP, 870 DEV_PATH_BR_VLAN_TAG, 871 DEV_PATH_BR_VLAN_UNTAG, 872 DEV_PATH_BR_VLAN_UNTAG_HW, 873 } vlan_mode; 874 u16 vlan_id; 875 __be16 vlan_proto; 876 } bridge; 877 struct { 878 int port; 879 u16 proto; 880 } dsa; 881 struct { 882 u8 wdma_idx; 883 u8 queue; 884 u16 wcid; 885 u8 bss; 886 u8 amsdu; 887 } mtk_wdma; 888 }; 889 }; 890 891 #define NET_DEVICE_PATH_STACK_MAX 5 892 #define NET_DEVICE_PATH_VLAN_MAX 2 893 894 struct net_device_path_stack { 895 int num_paths; 896 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 897 }; 898 899 struct net_device_path_ctx { 900 const struct net_device *dev; 901 u8 daddr[ETH_ALEN]; 902 903 int num_vlans; 904 struct { 905 u16 id; 906 __be16 proto; 907 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 908 }; 909 910 enum tc_setup_type { 911 TC_QUERY_CAPS, 912 TC_SETUP_QDISC_MQPRIO, 913 TC_SETUP_CLSU32, 914 TC_SETUP_CLSFLOWER, 915 TC_SETUP_CLSMATCHALL, 916 TC_SETUP_CLSBPF, 917 TC_SETUP_BLOCK, 918 TC_SETUP_QDISC_CBS, 919 TC_SETUP_QDISC_RED, 920 TC_SETUP_QDISC_PRIO, 921 TC_SETUP_QDISC_MQ, 922 TC_SETUP_QDISC_ETF, 923 TC_SETUP_ROOT_QDISC, 924 TC_SETUP_QDISC_GRED, 925 TC_SETUP_QDISC_TAPRIO, 926 TC_SETUP_FT, 927 TC_SETUP_QDISC_ETS, 928 TC_SETUP_QDISC_TBF, 929 TC_SETUP_QDISC_FIFO, 930 TC_SETUP_QDISC_HTB, 931 TC_SETUP_ACT, 932 }; 933 934 /* These structures hold the attributes of bpf state that are being passed 935 * to the netdevice through the bpf op. 936 */ 937 enum bpf_netdev_command { 938 /* Set or clear a bpf program used in the earliest stages of packet 939 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 940 * is responsible for calling bpf_prog_put on any old progs that are 941 * stored. In case of error, the callee need not release the new prog 942 * reference, but on success it takes ownership and must bpf_prog_put 943 * when it is no longer used. 944 */ 945 XDP_SETUP_PROG, 946 XDP_SETUP_PROG_HW, 947 /* BPF program for offload callbacks, invoked at program load time. */ 948 BPF_OFFLOAD_MAP_ALLOC, 949 BPF_OFFLOAD_MAP_FREE, 950 XDP_SETUP_XSK_POOL, 951 }; 952 953 struct bpf_prog_offload_ops; 954 struct netlink_ext_ack; 955 struct xdp_umem; 956 struct xdp_dev_bulk_queue; 957 struct bpf_xdp_link; 958 959 enum bpf_xdp_mode { 960 XDP_MODE_SKB = 0, 961 XDP_MODE_DRV = 1, 962 XDP_MODE_HW = 2, 963 __MAX_XDP_MODE 964 }; 965 966 struct bpf_xdp_entity { 967 struct bpf_prog *prog; 968 struct bpf_xdp_link *link; 969 }; 970 971 struct netdev_bpf { 972 enum bpf_netdev_command command; 973 union { 974 /* XDP_SETUP_PROG */ 975 struct { 976 u32 flags; 977 struct bpf_prog *prog; 978 struct netlink_ext_ack *extack; 979 }; 980 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 981 struct { 982 struct bpf_offloaded_map *offmap; 983 }; 984 /* XDP_SETUP_XSK_POOL */ 985 struct { 986 struct xsk_buff_pool *pool; 987 u16 queue_id; 988 } xsk; 989 }; 990 }; 991 992 /* Flags for ndo_xsk_wakeup. */ 993 #define XDP_WAKEUP_RX (1 << 0) 994 #define XDP_WAKEUP_TX (1 << 1) 995 996 #ifdef CONFIG_XFRM_OFFLOAD 997 struct xfrmdev_ops { 998 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack); 999 void (*xdo_dev_state_delete) (struct xfrm_state *x); 1000 void (*xdo_dev_state_free) (struct xfrm_state *x); 1001 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 1002 struct xfrm_state *x); 1003 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1004 void (*xdo_dev_state_update_stats) (struct xfrm_state *x); 1005 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack); 1006 void (*xdo_dev_policy_delete) (struct xfrm_policy *x); 1007 void (*xdo_dev_policy_free) (struct xfrm_policy *x); 1008 }; 1009 #endif 1010 1011 struct dev_ifalias { 1012 struct rcu_head rcuhead; 1013 char ifalias[]; 1014 }; 1015 1016 struct devlink; 1017 struct tlsdev_ops; 1018 1019 struct netdev_net_notifier { 1020 struct list_head list; 1021 struct notifier_block *nb; 1022 }; 1023 1024 /* 1025 * This structure defines the management hooks for network devices. 1026 * The following hooks can be defined; unless noted otherwise, they are 1027 * optional and can be filled with a null pointer. 1028 * 1029 * int (*ndo_init)(struct net_device *dev); 1030 * This function is called once when a network device is registered. 1031 * The network device can use this for any late stage initialization 1032 * or semantic validation. It can fail with an error code which will 1033 * be propagated back to register_netdev. 1034 * 1035 * void (*ndo_uninit)(struct net_device *dev); 1036 * This function is called when device is unregistered or when registration 1037 * fails. It is not called if init fails. 1038 * 1039 * int (*ndo_open)(struct net_device *dev); 1040 * This function is called when a network device transitions to the up 1041 * state. 1042 * 1043 * int (*ndo_stop)(struct net_device *dev); 1044 * This function is called when a network device transitions to the down 1045 * state. 1046 * 1047 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1048 * struct net_device *dev); 1049 * Called when a packet needs to be transmitted. 1050 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1051 * the queue before that can happen; it's for obsolete devices and weird 1052 * corner cases, but the stack really does a non-trivial amount 1053 * of useless work if you return NETDEV_TX_BUSY. 1054 * Required; cannot be NULL. 1055 * 1056 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1057 * struct net_device *dev 1058 * netdev_features_t features); 1059 * Called by core transmit path to determine if device is capable of 1060 * performing offload operations on a given packet. This is to give 1061 * the device an opportunity to implement any restrictions that cannot 1062 * be otherwise expressed by feature flags. The check is called with 1063 * the set of features that the stack has calculated and it returns 1064 * those the driver believes to be appropriate. 1065 * 1066 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1067 * struct net_device *sb_dev); 1068 * Called to decide which queue to use when device supports multiple 1069 * transmit queues. 1070 * 1071 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1072 * This function is called to allow device receiver to make 1073 * changes to configuration when multicast or promiscuous is enabled. 1074 * 1075 * void (*ndo_set_rx_mode)(struct net_device *dev); 1076 * This function is called device changes address list filtering. 1077 * If driver handles unicast address filtering, it should set 1078 * IFF_UNICAST_FLT in its priv_flags. 1079 * 1080 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1081 * This function is called when the Media Access Control address 1082 * needs to be changed. If this interface is not defined, the 1083 * MAC address can not be changed. 1084 * 1085 * int (*ndo_validate_addr)(struct net_device *dev); 1086 * Test if Media Access Control address is valid for the device. 1087 * 1088 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1089 * Old-style ioctl entry point. This is used internally by the 1090 * appletalk and ieee802154 subsystems but is no longer called by 1091 * the device ioctl handler. 1092 * 1093 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1094 * Used by the bonding driver for its device specific ioctls: 1095 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1096 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1097 * 1098 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1099 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1100 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1101 * 1102 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1103 * Used to set network devices bus interface parameters. This interface 1104 * is retained for legacy reasons; new devices should use the bus 1105 * interface (PCI) for low level management. 1106 * 1107 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1108 * Called when a user wants to change the Maximum Transfer Unit 1109 * of a device. 1110 * 1111 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1112 * Callback used when the transmitter has not made any progress 1113 * for dev->watchdog ticks. 1114 * 1115 * void (*ndo_get_stats64)(struct net_device *dev, 1116 * struct rtnl_link_stats64 *storage); 1117 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1118 * Called when a user wants to get the network device usage 1119 * statistics. Drivers must do one of the following: 1120 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1121 * rtnl_link_stats64 structure passed by the caller. 1122 * 2. Define @ndo_get_stats to update a net_device_stats structure 1123 * (which should normally be dev->stats) and return a pointer to 1124 * it. The structure may be changed asynchronously only if each 1125 * field is written atomically. 1126 * 3. Update dev->stats asynchronously and atomically, and define 1127 * neither operation. 1128 * 1129 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1130 * Return true if this device supports offload stats of this attr_id. 1131 * 1132 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1133 * void *attr_data) 1134 * Get statistics for offload operations by attr_id. Write it into the 1135 * attr_data pointer. 1136 * 1137 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1138 * If device supports VLAN filtering this function is called when a 1139 * VLAN id is registered. 1140 * 1141 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1142 * If device supports VLAN filtering this function is called when a 1143 * VLAN id is unregistered. 1144 * 1145 * void (*ndo_poll_controller)(struct net_device *dev); 1146 * 1147 * SR-IOV management functions. 1148 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1149 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1150 * u8 qos, __be16 proto); 1151 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1152 * int max_tx_rate); 1153 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1154 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1155 * int (*ndo_get_vf_config)(struct net_device *dev, 1156 * int vf, struct ifla_vf_info *ivf); 1157 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1158 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1159 * struct nlattr *port[]); 1160 * 1161 * Enable or disable the VF ability to query its RSS Redirection Table and 1162 * Hash Key. This is needed since on some devices VF share this information 1163 * with PF and querying it may introduce a theoretical security risk. 1164 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1165 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1166 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1167 * void *type_data); 1168 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1169 * This is always called from the stack with the rtnl lock held and netif 1170 * tx queues stopped. This allows the netdevice to perform queue 1171 * management safely. 1172 * 1173 * Fiber Channel over Ethernet (FCoE) offload functions. 1174 * int (*ndo_fcoe_enable)(struct net_device *dev); 1175 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1176 * so the underlying device can perform whatever needed configuration or 1177 * initialization to support acceleration of FCoE traffic. 1178 * 1179 * int (*ndo_fcoe_disable)(struct net_device *dev); 1180 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1181 * so the underlying device can perform whatever needed clean-ups to 1182 * stop supporting acceleration of FCoE traffic. 1183 * 1184 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1185 * struct scatterlist *sgl, unsigned int sgc); 1186 * Called when the FCoE Initiator wants to initialize an I/O that 1187 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1188 * perform necessary setup and returns 1 to indicate the device is set up 1189 * successfully to perform DDP on this I/O, otherwise this returns 0. 1190 * 1191 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1192 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1193 * indicated by the FC exchange id 'xid', so the underlying device can 1194 * clean up and reuse resources for later DDP requests. 1195 * 1196 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1197 * struct scatterlist *sgl, unsigned int sgc); 1198 * Called when the FCoE Target wants to initialize an I/O that 1199 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1200 * perform necessary setup and returns 1 to indicate the device is set up 1201 * successfully to perform DDP on this I/O, otherwise this returns 0. 1202 * 1203 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1204 * struct netdev_fcoe_hbainfo *hbainfo); 1205 * Called when the FCoE Protocol stack wants information on the underlying 1206 * device. This information is utilized by the FCoE protocol stack to 1207 * register attributes with Fiber Channel management service as per the 1208 * FC-GS Fabric Device Management Information(FDMI) specification. 1209 * 1210 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1211 * Called when the underlying device wants to override default World Wide 1212 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1213 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1214 * protocol stack to use. 1215 * 1216 * RFS acceleration. 1217 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1218 * u16 rxq_index, u32 flow_id); 1219 * Set hardware filter for RFS. rxq_index is the target queue index; 1220 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1221 * Return the filter ID on success, or a negative error code. 1222 * 1223 * Slave management functions (for bridge, bonding, etc). 1224 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1225 * Called to make another netdev an underling. 1226 * 1227 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1228 * Called to release previously enslaved netdev. 1229 * 1230 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1231 * struct sk_buff *skb, 1232 * bool all_slaves); 1233 * Get the xmit slave of master device. If all_slaves is true, function 1234 * assume all the slaves can transmit. 1235 * 1236 * Feature/offload setting functions. 1237 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1238 * netdev_features_t features); 1239 * Adjusts the requested feature flags according to device-specific 1240 * constraints, and returns the resulting flags. Must not modify 1241 * the device state. 1242 * 1243 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1244 * Called to update device configuration to new features. Passed 1245 * feature set might be less than what was returned by ndo_fix_features()). 1246 * Must return >0 or -errno if it changed dev->features itself. 1247 * 1248 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1249 * struct net_device *dev, 1250 * const unsigned char *addr, u16 vid, u16 flags, 1251 * struct netlink_ext_ack *extack); 1252 * Adds an FDB entry to dev for addr. 1253 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1254 * struct net_device *dev, 1255 * const unsigned char *addr, u16 vid) 1256 * Deletes the FDB entry from dev corresponding to addr. 1257 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev, 1258 * struct netlink_ext_ack *extack); 1259 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1260 * struct net_device *dev, struct net_device *filter_dev, 1261 * int *idx) 1262 * Used to add FDB entries to dump requests. Implementers should add 1263 * entries to skb and update idx with the number of entries. 1264 * 1265 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[], 1266 * u16 nlmsg_flags, struct netlink_ext_ack *extack); 1267 * Adds an MDB entry to dev. 1268 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[], 1269 * struct netlink_ext_ack *extack); 1270 * Deletes the MDB entry from dev. 1271 * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[], 1272 * struct netlink_ext_ack *extack); 1273 * Bulk deletes MDB entries from dev. 1274 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb, 1275 * struct netlink_callback *cb); 1276 * Dumps MDB entries from dev. The first argument (marker) in the netlink 1277 * callback is used by core rtnetlink code. 1278 * 1279 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1280 * u16 flags, struct netlink_ext_ack *extack) 1281 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1282 * struct net_device *dev, u32 filter_mask, 1283 * int nlflags) 1284 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1285 * u16 flags); 1286 * 1287 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1288 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1289 * which do not represent real hardware may define this to allow their 1290 * userspace components to manage their virtual carrier state. Devices 1291 * that determine carrier state from physical hardware properties (eg 1292 * network cables) or protocol-dependent mechanisms (eg 1293 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1294 * 1295 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1296 * struct netdev_phys_item_id *ppid); 1297 * Called to get ID of physical port of this device. If driver does 1298 * not implement this, it is assumed that the hw is not able to have 1299 * multiple net devices on single physical port. 1300 * 1301 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1302 * struct netdev_phys_item_id *ppid) 1303 * Called to get the parent ID of the physical port of this device. 1304 * 1305 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1306 * struct net_device *dev) 1307 * Called by upper layer devices to accelerate switching or other 1308 * station functionality into hardware. 'pdev is the lowerdev 1309 * to use for the offload and 'dev' is the net device that will 1310 * back the offload. Returns a pointer to the private structure 1311 * the upper layer will maintain. 1312 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1313 * Called by upper layer device to delete the station created 1314 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1315 * the station and priv is the structure returned by the add 1316 * operation. 1317 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1318 * int queue_index, u32 maxrate); 1319 * Called when a user wants to set a max-rate limitation of specific 1320 * TX queue. 1321 * int (*ndo_get_iflink)(const struct net_device *dev); 1322 * Called to get the iflink value of this device. 1323 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1324 * This function is used to get egress tunnel information for given skb. 1325 * This is useful for retrieving outer tunnel header parameters while 1326 * sampling packet. 1327 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1328 * This function is used to specify the headroom that the skb must 1329 * consider when allocation skb during packet reception. Setting 1330 * appropriate rx headroom value allows avoiding skb head copy on 1331 * forward. Setting a negative value resets the rx headroom to the 1332 * default value. 1333 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1334 * This function is used to set or query state related to XDP on the 1335 * netdevice and manage BPF offload. See definition of 1336 * enum bpf_netdev_command for details. 1337 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1338 * u32 flags); 1339 * This function is used to submit @n XDP packets for transmit on a 1340 * netdevice. Returns number of frames successfully transmitted, frames 1341 * that got dropped are freed/returned via xdp_return_frame(). 1342 * Returns negative number, means general error invoking ndo, meaning 1343 * no frames were xmit'ed and core-caller will free all frames. 1344 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1345 * struct xdp_buff *xdp); 1346 * Get the xmit slave of master device based on the xdp_buff. 1347 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1348 * This function is used to wake up the softirq, ksoftirqd or kthread 1349 * responsible for sending and/or receiving packets on a specific 1350 * queue id bound to an AF_XDP socket. The flags field specifies if 1351 * only RX, only Tx, or both should be woken up using the flags 1352 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1353 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm_kern *p, 1354 * int cmd); 1355 * Add, change, delete or get information on an IPv4 tunnel. 1356 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1357 * If a device is paired with a peer device, return the peer instance. 1358 * The caller must be under RCU read context. 1359 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1360 * Get the forwarding path to reach the real device from the HW destination address 1361 * ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1362 * const struct skb_shared_hwtstamps *hwtstamps, 1363 * bool cycles); 1364 * Get hardware timestamp based on normal/adjustable time or free running 1365 * cycle counter. This function is required if physical clock supports a 1366 * free running cycle counter. 1367 * 1368 * int (*ndo_hwtstamp_get)(struct net_device *dev, 1369 * struct kernel_hwtstamp_config *kernel_config); 1370 * Get the currently configured hardware timestamping parameters for the 1371 * NIC device. 1372 * 1373 * int (*ndo_hwtstamp_set)(struct net_device *dev, 1374 * struct kernel_hwtstamp_config *kernel_config, 1375 * struct netlink_ext_ack *extack); 1376 * Change the hardware timestamping parameters for NIC device. 1377 */ 1378 struct net_device_ops { 1379 int (*ndo_init)(struct net_device *dev); 1380 void (*ndo_uninit)(struct net_device *dev); 1381 int (*ndo_open)(struct net_device *dev); 1382 int (*ndo_stop)(struct net_device *dev); 1383 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1384 struct net_device *dev); 1385 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1386 struct net_device *dev, 1387 netdev_features_t features); 1388 u16 (*ndo_select_queue)(struct net_device *dev, 1389 struct sk_buff *skb, 1390 struct net_device *sb_dev); 1391 void (*ndo_change_rx_flags)(struct net_device *dev, 1392 int flags); 1393 void (*ndo_set_rx_mode)(struct net_device *dev); 1394 int (*ndo_set_mac_address)(struct net_device *dev, 1395 void *addr); 1396 int (*ndo_validate_addr)(struct net_device *dev); 1397 int (*ndo_do_ioctl)(struct net_device *dev, 1398 struct ifreq *ifr, int cmd); 1399 int (*ndo_eth_ioctl)(struct net_device *dev, 1400 struct ifreq *ifr, int cmd); 1401 int (*ndo_siocbond)(struct net_device *dev, 1402 struct ifreq *ifr, int cmd); 1403 int (*ndo_siocwandev)(struct net_device *dev, 1404 struct if_settings *ifs); 1405 int (*ndo_siocdevprivate)(struct net_device *dev, 1406 struct ifreq *ifr, 1407 void __user *data, int cmd); 1408 int (*ndo_set_config)(struct net_device *dev, 1409 struct ifmap *map); 1410 int (*ndo_change_mtu)(struct net_device *dev, 1411 int new_mtu); 1412 int (*ndo_neigh_setup)(struct net_device *dev, 1413 struct neigh_parms *); 1414 void (*ndo_tx_timeout) (struct net_device *dev, 1415 unsigned int txqueue); 1416 1417 void (*ndo_get_stats64)(struct net_device *dev, 1418 struct rtnl_link_stats64 *storage); 1419 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1420 int (*ndo_get_offload_stats)(int attr_id, 1421 const struct net_device *dev, 1422 void *attr_data); 1423 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1424 1425 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1426 __be16 proto, u16 vid); 1427 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1428 __be16 proto, u16 vid); 1429 #ifdef CONFIG_NET_POLL_CONTROLLER 1430 void (*ndo_poll_controller)(struct net_device *dev); 1431 int (*ndo_netpoll_setup)(struct net_device *dev); 1432 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1433 #endif 1434 int (*ndo_set_vf_mac)(struct net_device *dev, 1435 int queue, u8 *mac); 1436 int (*ndo_set_vf_vlan)(struct net_device *dev, 1437 int queue, u16 vlan, 1438 u8 qos, __be16 proto); 1439 int (*ndo_set_vf_rate)(struct net_device *dev, 1440 int vf, int min_tx_rate, 1441 int max_tx_rate); 1442 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1443 int vf, bool setting); 1444 int (*ndo_set_vf_trust)(struct net_device *dev, 1445 int vf, bool setting); 1446 int (*ndo_get_vf_config)(struct net_device *dev, 1447 int vf, 1448 struct ifla_vf_info *ivf); 1449 int (*ndo_set_vf_link_state)(struct net_device *dev, 1450 int vf, int link_state); 1451 int (*ndo_get_vf_stats)(struct net_device *dev, 1452 int vf, 1453 struct ifla_vf_stats 1454 *vf_stats); 1455 int (*ndo_set_vf_port)(struct net_device *dev, 1456 int vf, 1457 struct nlattr *port[]); 1458 int (*ndo_get_vf_port)(struct net_device *dev, 1459 int vf, struct sk_buff *skb); 1460 int (*ndo_get_vf_guid)(struct net_device *dev, 1461 int vf, 1462 struct ifla_vf_guid *node_guid, 1463 struct ifla_vf_guid *port_guid); 1464 int (*ndo_set_vf_guid)(struct net_device *dev, 1465 int vf, u64 guid, 1466 int guid_type); 1467 int (*ndo_set_vf_rss_query_en)( 1468 struct net_device *dev, 1469 int vf, bool setting); 1470 int (*ndo_setup_tc)(struct net_device *dev, 1471 enum tc_setup_type type, 1472 void *type_data); 1473 #if IS_ENABLED(CONFIG_FCOE) 1474 int (*ndo_fcoe_enable)(struct net_device *dev); 1475 int (*ndo_fcoe_disable)(struct net_device *dev); 1476 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1477 u16 xid, 1478 struct scatterlist *sgl, 1479 unsigned int sgc); 1480 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1481 u16 xid); 1482 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1483 u16 xid, 1484 struct scatterlist *sgl, 1485 unsigned int sgc); 1486 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1487 struct netdev_fcoe_hbainfo *hbainfo); 1488 #endif 1489 1490 #if IS_ENABLED(CONFIG_LIBFCOE) 1491 #define NETDEV_FCOE_WWNN 0 1492 #define NETDEV_FCOE_WWPN 1 1493 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1494 u64 *wwn, int type); 1495 #endif 1496 1497 #ifdef CONFIG_RFS_ACCEL 1498 int (*ndo_rx_flow_steer)(struct net_device *dev, 1499 const struct sk_buff *skb, 1500 u16 rxq_index, 1501 u32 flow_id); 1502 #endif 1503 int (*ndo_add_slave)(struct net_device *dev, 1504 struct net_device *slave_dev, 1505 struct netlink_ext_ack *extack); 1506 int (*ndo_del_slave)(struct net_device *dev, 1507 struct net_device *slave_dev); 1508 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1509 struct sk_buff *skb, 1510 bool all_slaves); 1511 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1512 struct sock *sk); 1513 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1514 netdev_features_t features); 1515 int (*ndo_set_features)(struct net_device *dev, 1516 netdev_features_t features); 1517 int (*ndo_neigh_construct)(struct net_device *dev, 1518 struct neighbour *n); 1519 void (*ndo_neigh_destroy)(struct net_device *dev, 1520 struct neighbour *n); 1521 1522 int (*ndo_fdb_add)(struct ndmsg *ndm, 1523 struct nlattr *tb[], 1524 struct net_device *dev, 1525 const unsigned char *addr, 1526 u16 vid, 1527 u16 flags, 1528 struct netlink_ext_ack *extack); 1529 int (*ndo_fdb_del)(struct ndmsg *ndm, 1530 struct nlattr *tb[], 1531 struct net_device *dev, 1532 const unsigned char *addr, 1533 u16 vid, struct netlink_ext_ack *extack); 1534 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, 1535 struct net_device *dev, 1536 struct netlink_ext_ack *extack); 1537 int (*ndo_fdb_dump)(struct sk_buff *skb, 1538 struct netlink_callback *cb, 1539 struct net_device *dev, 1540 struct net_device *filter_dev, 1541 int *idx); 1542 int (*ndo_fdb_get)(struct sk_buff *skb, 1543 struct nlattr *tb[], 1544 struct net_device *dev, 1545 const unsigned char *addr, 1546 u16 vid, u32 portid, u32 seq, 1547 struct netlink_ext_ack *extack); 1548 int (*ndo_mdb_add)(struct net_device *dev, 1549 struct nlattr *tb[], 1550 u16 nlmsg_flags, 1551 struct netlink_ext_ack *extack); 1552 int (*ndo_mdb_del)(struct net_device *dev, 1553 struct nlattr *tb[], 1554 struct netlink_ext_ack *extack); 1555 int (*ndo_mdb_del_bulk)(struct net_device *dev, 1556 struct nlattr *tb[], 1557 struct netlink_ext_ack *extack); 1558 int (*ndo_mdb_dump)(struct net_device *dev, 1559 struct sk_buff *skb, 1560 struct netlink_callback *cb); 1561 int (*ndo_mdb_get)(struct net_device *dev, 1562 struct nlattr *tb[], u32 portid, 1563 u32 seq, 1564 struct netlink_ext_ack *extack); 1565 int (*ndo_bridge_setlink)(struct net_device *dev, 1566 struct nlmsghdr *nlh, 1567 u16 flags, 1568 struct netlink_ext_ack *extack); 1569 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1570 u32 pid, u32 seq, 1571 struct net_device *dev, 1572 u32 filter_mask, 1573 int nlflags); 1574 int (*ndo_bridge_dellink)(struct net_device *dev, 1575 struct nlmsghdr *nlh, 1576 u16 flags); 1577 int (*ndo_change_carrier)(struct net_device *dev, 1578 bool new_carrier); 1579 int (*ndo_get_phys_port_id)(struct net_device *dev, 1580 struct netdev_phys_item_id *ppid); 1581 int (*ndo_get_port_parent_id)(struct net_device *dev, 1582 struct netdev_phys_item_id *ppid); 1583 int (*ndo_get_phys_port_name)(struct net_device *dev, 1584 char *name, size_t len); 1585 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1586 struct net_device *dev); 1587 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1588 void *priv); 1589 1590 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1591 int queue_index, 1592 u32 maxrate); 1593 int (*ndo_get_iflink)(const struct net_device *dev); 1594 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1595 struct sk_buff *skb); 1596 void (*ndo_set_rx_headroom)(struct net_device *dev, 1597 int needed_headroom); 1598 int (*ndo_bpf)(struct net_device *dev, 1599 struct netdev_bpf *bpf); 1600 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1601 struct xdp_frame **xdp, 1602 u32 flags); 1603 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1604 struct xdp_buff *xdp); 1605 int (*ndo_xsk_wakeup)(struct net_device *dev, 1606 u32 queue_id, u32 flags); 1607 int (*ndo_tunnel_ctl)(struct net_device *dev, 1608 struct ip_tunnel_parm_kern *p, 1609 int cmd); 1610 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1611 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1612 struct net_device_path *path); 1613 ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1614 const struct skb_shared_hwtstamps *hwtstamps, 1615 bool cycles); 1616 int (*ndo_hwtstamp_get)(struct net_device *dev, 1617 struct kernel_hwtstamp_config *kernel_config); 1618 int (*ndo_hwtstamp_set)(struct net_device *dev, 1619 struct kernel_hwtstamp_config *kernel_config, 1620 struct netlink_ext_ack *extack); 1621 1622 #if IS_ENABLED(CONFIG_NET_SHAPER) 1623 /** 1624 * @net_shaper_ops: Device shaping offload operations 1625 * see include/net/net_shapers.h 1626 */ 1627 const struct net_shaper_ops *net_shaper_ops; 1628 #endif 1629 }; 1630 1631 /** 1632 * enum netdev_priv_flags - &struct net_device priv_flags 1633 * 1634 * These are the &struct net_device, they are only set internally 1635 * by drivers and used in the kernel. These flags are invisible to 1636 * userspace; this means that the order of these flags can change 1637 * during any kernel release. 1638 * 1639 * You should add bitfield booleans after either net_device::priv_flags 1640 * (hotpath) or ::threaded (slowpath) instead of extending these flags. 1641 * 1642 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1643 * @IFF_EBRIDGE: Ethernet bridging device 1644 * @IFF_BONDING: bonding master or slave 1645 * @IFF_ISATAP: ISATAP interface (RFC4214) 1646 * @IFF_WAN_HDLC: WAN HDLC device 1647 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1648 * release skb->dst 1649 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1650 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1651 * @IFF_MACVLAN_PORT: device used as macvlan port 1652 * @IFF_BRIDGE_PORT: device used as bridge port 1653 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1654 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1655 * @IFF_UNICAST_FLT: Supports unicast filtering 1656 * @IFF_TEAM_PORT: device used as team port 1657 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1658 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1659 * change when it's running 1660 * @IFF_MACVLAN: Macvlan device 1661 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1662 * underlying stacked devices 1663 * @IFF_L3MDEV_MASTER: device is an L3 master device 1664 * @IFF_NO_QUEUE: device can run without qdisc attached 1665 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1666 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1667 * @IFF_TEAM: device is a team device 1668 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1669 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1670 * entity (i.e. the master device for bridged veth) 1671 * @IFF_MACSEC: device is a MACsec device 1672 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1673 * @IFF_FAILOVER: device is a failover master device 1674 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1675 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1676 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf 1677 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1678 * skb_headlen(skb) == 0 (data starts from frag0) 1679 */ 1680 enum netdev_priv_flags { 1681 IFF_802_1Q_VLAN = 1<<0, 1682 IFF_EBRIDGE = 1<<1, 1683 IFF_BONDING = 1<<2, 1684 IFF_ISATAP = 1<<3, 1685 IFF_WAN_HDLC = 1<<4, 1686 IFF_XMIT_DST_RELEASE = 1<<5, 1687 IFF_DONT_BRIDGE = 1<<6, 1688 IFF_DISABLE_NETPOLL = 1<<7, 1689 IFF_MACVLAN_PORT = 1<<8, 1690 IFF_BRIDGE_PORT = 1<<9, 1691 IFF_OVS_DATAPATH = 1<<10, 1692 IFF_TX_SKB_SHARING = 1<<11, 1693 IFF_UNICAST_FLT = 1<<12, 1694 IFF_TEAM_PORT = 1<<13, 1695 IFF_SUPP_NOFCS = 1<<14, 1696 IFF_LIVE_ADDR_CHANGE = 1<<15, 1697 IFF_MACVLAN = 1<<16, 1698 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1699 IFF_L3MDEV_MASTER = 1<<18, 1700 IFF_NO_QUEUE = 1<<19, 1701 IFF_OPENVSWITCH = 1<<20, 1702 IFF_L3MDEV_SLAVE = 1<<21, 1703 IFF_TEAM = 1<<22, 1704 IFF_RXFH_CONFIGURED = 1<<23, 1705 IFF_PHONY_HEADROOM = 1<<24, 1706 IFF_MACSEC = 1<<25, 1707 IFF_NO_RX_HANDLER = 1<<26, 1708 IFF_FAILOVER = 1<<27, 1709 IFF_FAILOVER_SLAVE = 1<<28, 1710 IFF_L3MDEV_RX_HANDLER = 1<<29, 1711 IFF_NO_ADDRCONF = BIT_ULL(30), 1712 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31), 1713 }; 1714 1715 /* Specifies the type of the struct net_device::ml_priv pointer */ 1716 enum netdev_ml_priv_type { 1717 ML_PRIV_NONE, 1718 ML_PRIV_CAN, 1719 }; 1720 1721 enum netdev_stat_type { 1722 NETDEV_PCPU_STAT_NONE, 1723 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */ 1724 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */ 1725 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */ 1726 }; 1727 1728 enum netdev_reg_state { 1729 NETREG_UNINITIALIZED = 0, 1730 NETREG_REGISTERED, /* completed register_netdevice */ 1731 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1732 NETREG_UNREGISTERED, /* completed unregister todo */ 1733 NETREG_RELEASED, /* called free_netdev */ 1734 NETREG_DUMMY, /* dummy device for NAPI poll */ 1735 }; 1736 1737 /** 1738 * struct net_device - The DEVICE structure. 1739 * 1740 * Actually, this whole structure is a big mistake. It mixes I/O 1741 * data with strictly "high-level" data, and it has to know about 1742 * almost every data structure used in the INET module. 1743 * 1744 * @priv_flags: flags invisible to userspace defined as bits, see 1745 * enum netdev_priv_flags for the definitions 1746 * @lltx: device supports lockless Tx. Deprecated for real HW 1747 * drivers. Mainly used by logical interfaces, such as 1748 * bonding and tunnels 1749 * 1750 * @name: This is the first field of the "visible" part of this structure 1751 * (i.e. as seen by users in the "Space.c" file). It is the name 1752 * of the interface. 1753 * 1754 * @name_node: Name hashlist node 1755 * @ifalias: SNMP alias 1756 * @mem_end: Shared memory end 1757 * @mem_start: Shared memory start 1758 * @base_addr: Device I/O address 1759 * @irq: Device IRQ number 1760 * 1761 * @state: Generic network queuing layer state, see netdev_state_t 1762 * @dev_list: The global list of network devices 1763 * @napi_list: List entry used for polling NAPI devices 1764 * @unreg_list: List entry when we are unregistering the 1765 * device; see the function unregister_netdev 1766 * @close_list: List entry used when we are closing the device 1767 * @ptype_all: Device-specific packet handlers for all protocols 1768 * @ptype_specific: Device-specific, protocol-specific packet handlers 1769 * 1770 * @adj_list: Directly linked devices, like slaves for bonding 1771 * @features: Currently active device features 1772 * @hw_features: User-changeable features 1773 * 1774 * @wanted_features: User-requested features 1775 * @vlan_features: Mask of features inheritable by VLAN devices 1776 * 1777 * @hw_enc_features: Mask of features inherited by encapsulating devices 1778 * This field indicates what encapsulation 1779 * offloads the hardware is capable of doing, 1780 * and drivers will need to set them appropriately. 1781 * 1782 * @mpls_features: Mask of features inheritable by MPLS 1783 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1784 * 1785 * @ifindex: interface index 1786 * @group: The group the device belongs to 1787 * 1788 * @stats: Statistics struct, which was left as a legacy, use 1789 * rtnl_link_stats64 instead 1790 * 1791 * @core_stats: core networking counters, 1792 * do not use this in drivers 1793 * @carrier_up_count: Number of times the carrier has been up 1794 * @carrier_down_count: Number of times the carrier has been down 1795 * 1796 * @wireless_handlers: List of functions to handle Wireless Extensions, 1797 * instead of ioctl, 1798 * see <net/iw_handler.h> for details. 1799 * 1800 * @netdev_ops: Includes several pointers to callbacks, 1801 * if one wants to override the ndo_*() functions 1802 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks. 1803 * @xsk_tx_metadata_ops: Includes pointers to AF_XDP TX metadata callbacks. 1804 * @ethtool_ops: Management operations 1805 * @l3mdev_ops: Layer 3 master device operations 1806 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1807 * discovery handling. Necessary for e.g. 6LoWPAN. 1808 * @xfrmdev_ops: Transformation offload operations 1809 * @tlsdev_ops: Transport Layer Security offload operations 1810 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1811 * of Layer 2 headers. 1812 * 1813 * @flags: Interface flags (a la BSD) 1814 * @xdp_features: XDP capability supported by the device 1815 * @gflags: Global flags ( kept as legacy ) 1816 * @priv_len: Size of the ->priv flexible array 1817 * @priv: Flexible array containing private data 1818 * @operstate: RFC2863 operstate 1819 * @link_mode: Mapping policy to operstate 1820 * @if_port: Selectable AUI, TP, ... 1821 * @dma: DMA channel 1822 * @mtu: Interface MTU value 1823 * @min_mtu: Interface Minimum MTU value 1824 * @max_mtu: Interface Maximum MTU value 1825 * @type: Interface hardware type 1826 * @hard_header_len: Maximum hardware header length. 1827 * @min_header_len: Minimum hardware header length 1828 * 1829 * @needed_headroom: Extra headroom the hardware may need, but not in all 1830 * cases can this be guaranteed 1831 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1832 * cases can this be guaranteed. Some cases also use 1833 * LL_MAX_HEADER instead to allocate the skb 1834 * 1835 * interface address info: 1836 * 1837 * @perm_addr: Permanent hw address 1838 * @addr_assign_type: Hw address assignment type 1839 * @addr_len: Hardware address length 1840 * @upper_level: Maximum depth level of upper devices. 1841 * @lower_level: Maximum depth level of lower devices. 1842 * @neigh_priv_len: Used in neigh_alloc() 1843 * @dev_id: Used to differentiate devices that share 1844 * the same link layer address 1845 * @dev_port: Used to differentiate devices that share 1846 * the same function 1847 * @addr_list_lock: XXX: need comments on this one 1848 * @name_assign_type: network interface name assignment type 1849 * @uc_promisc: Counter that indicates promiscuous mode 1850 * has been enabled due to the need to listen to 1851 * additional unicast addresses in a device that 1852 * does not implement ndo_set_rx_mode() 1853 * @uc: unicast mac addresses 1854 * @mc: multicast mac addresses 1855 * @dev_addrs: list of device hw addresses 1856 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1857 * @promiscuity: Number of times the NIC is told to work in 1858 * promiscuous mode; if it becomes 0 the NIC will 1859 * exit promiscuous mode 1860 * @allmulti: Counter, enables or disables allmulticast mode 1861 * 1862 * @vlan_info: VLAN info 1863 * @dsa_ptr: dsa specific data 1864 * @tipc_ptr: TIPC specific data 1865 * @atalk_ptr: AppleTalk link 1866 * @ip_ptr: IPv4 specific data 1867 * @ip6_ptr: IPv6 specific data 1868 * @ax25_ptr: AX.25 specific data 1869 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1870 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1871 * device struct 1872 * @mpls_ptr: mpls_dev struct pointer 1873 * @mctp_ptr: MCTP specific data 1874 * 1875 * @dev_addr: Hw address (before bcast, 1876 * because most packets are unicast) 1877 * 1878 * @_rx: Array of RX queues 1879 * @num_rx_queues: Number of RX queues 1880 * allocated at register_netdev() time 1881 * @real_num_rx_queues: Number of RX queues currently active in device 1882 * @xdp_prog: XDP sockets filter program pointer 1883 * 1884 * @rx_handler: handler for received packets 1885 * @rx_handler_data: XXX: need comments on this one 1886 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing 1887 * @ingress_queue: XXX: need comments on this one 1888 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1889 * @broadcast: hw bcast address 1890 * 1891 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1892 * indexed by RX queue number. Assigned by driver. 1893 * This must only be set if the ndo_rx_flow_steer 1894 * operation is defined 1895 * @index_hlist: Device index hash chain 1896 * 1897 * @_tx: Array of TX queues 1898 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1899 * @real_num_tx_queues: Number of TX queues currently active in device 1900 * @qdisc: Root qdisc from userspace point of view 1901 * @tx_queue_len: Max frames per queue allowed 1902 * @tx_global_lock: XXX: need comments on this one 1903 * @xdp_bulkq: XDP device bulk queue 1904 * @xps_maps: all CPUs/RXQs maps for XPS device 1905 * 1906 * @xps_maps: XXX: need comments on this one 1907 * @tcx_egress: BPF & clsact qdisc specific data for egress processing 1908 * @nf_hooks_egress: netfilter hooks executed for egress packets 1909 * @qdisc_hash: qdisc hash table 1910 * @watchdog_timeo: Represents the timeout that is used by 1911 * the watchdog (see dev_watchdog()) 1912 * @watchdog_timer: List of timers 1913 * 1914 * @proto_down_reason: reason a netdev interface is held down 1915 * @pcpu_refcnt: Number of references to this device 1916 * @dev_refcnt: Number of references to this device 1917 * @refcnt_tracker: Tracker directory for tracked references to this device 1918 * @todo_list: Delayed register/unregister 1919 * @link_watch_list: XXX: need comments on this one 1920 * 1921 * @reg_state: Register/unregister state machine 1922 * @dismantle: Device is going to be freed 1923 * @rtnl_link_state: This enum represents the phases of creating 1924 * a new link 1925 * 1926 * @needs_free_netdev: Should unregister perform free_netdev? 1927 * @priv_destructor: Called from unregister 1928 * @npinfo: XXX: need comments on this one 1929 * @nd_net: Network namespace this network device is inside 1930 * 1931 * @ml_priv: Mid-layer private 1932 * @ml_priv_type: Mid-layer private type 1933 * 1934 * @pcpu_stat_type: Type of device statistics which the core should 1935 * allocate/free: none, lstats, tstats, dstats. none 1936 * means the driver is handling statistics allocation/ 1937 * freeing internally. 1938 * @lstats: Loopback statistics: packets, bytes 1939 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes 1940 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes 1941 * 1942 * @garp_port: GARP 1943 * @mrp_port: MRP 1944 * 1945 * @dm_private: Drop monitor private 1946 * 1947 * @dev: Class/net/name entry 1948 * @sysfs_groups: Space for optional device, statistics and wireless 1949 * sysfs groups 1950 * 1951 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1952 * @rtnl_link_ops: Rtnl_link_ops 1953 * @stat_ops: Optional ops for queue-aware statistics 1954 * @queue_mgmt_ops: Optional ops for queue management 1955 * 1956 * @gso_max_size: Maximum size of generic segmentation offload 1957 * @tso_max_size: Device (as in HW) limit on the max TSO request size 1958 * @gso_max_segs: Maximum number of segments that can be passed to the 1959 * NIC for GSO 1960 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count 1961 * @gso_ipv4_max_size: Maximum size of generic segmentation offload, 1962 * for IPv4. 1963 * 1964 * @dcbnl_ops: Data Center Bridging netlink ops 1965 * @num_tc: Number of traffic classes in the net device 1966 * @tc_to_txq: XXX: need comments on this one 1967 * @prio_tc_map: XXX: need comments on this one 1968 * 1969 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1970 * 1971 * @priomap: XXX: need comments on this one 1972 * @link_topo: Physical link topology tracking attached PHYs 1973 * @phydev: Physical device may attach itself 1974 * for hardware timestamping 1975 * @sfp_bus: attached &struct sfp_bus structure. 1976 * 1977 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1978 * 1979 * @proto_down: protocol port state information can be sent to the 1980 * switch driver and used to set the phys state of the 1981 * switch port. 1982 * 1983 * @threaded: napi threaded mode is enabled 1984 * 1985 * @see_all_hwtstamp_requests: device wants to see calls to 1986 * ndo_hwtstamp_set() for all timestamp requests 1987 * regardless of source, even if those aren't 1988 * HWTSTAMP_SOURCE_NETDEV 1989 * @change_proto_down: device supports setting carrier via IFLA_PROTO_DOWN 1990 * @netns_local: interface can't change network namespaces 1991 * @fcoe_mtu: device supports maximum FCoE MTU, 2158 bytes 1992 * 1993 * @net_notifier_list: List of per-net netdev notifier block 1994 * that follow this device when it is moved 1995 * to another network namespace. 1996 * 1997 * @macsec_ops: MACsec offloading ops 1998 * 1999 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 2000 * offload capabilities of the device 2001 * @udp_tunnel_nic: UDP tunnel offload state 2002 * @ethtool: ethtool related state 2003 * @xdp_state: stores info on attached XDP BPF programs 2004 * 2005 * @nested_level: Used as a parameter of spin_lock_nested() of 2006 * dev->addr_list_lock. 2007 * @unlink_list: As netif_addr_lock() can be called recursively, 2008 * keep a list of interfaces to be deleted. 2009 * @gro_max_size: Maximum size of aggregated packet in generic 2010 * receive offload (GRO) 2011 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic 2012 * receive offload (GRO), for IPv4. 2013 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP 2014 * zero copy driver 2015 * 2016 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 2017 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 2018 * @watchdog_dev_tracker: refcount tracker used by watchdog. 2019 * @dev_registered_tracker: tracker for reference held while 2020 * registered 2021 * @offload_xstats_l3: L3 HW stats for this netdevice. 2022 * 2023 * @devlink_port: Pointer to related devlink port structure. 2024 * Assigned by a driver before netdev registration using 2025 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static 2026 * during the time netdevice is registered. 2027 * 2028 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem, 2029 * where the clock is recovered. 2030 * 2031 * @max_pacing_offload_horizon: max EDT offload horizon in nsec. 2032 * @napi_config: An array of napi_config structures containing per-NAPI 2033 * settings. 2034 * @gro_flush_timeout: timeout for GRO layer in NAPI 2035 * @napi_defer_hard_irqs: If not zero, provides a counter that would 2036 * allow to avoid NIC hard IRQ, on busy queues. 2037 * 2038 * @neighbours: List heads pointing to this device's neighbours' 2039 * dev_list, one per address-family. 2040 * 2041 * FIXME: cleanup struct net_device such that network protocol info 2042 * moves out. 2043 */ 2044 2045 struct net_device { 2046 /* Cacheline organization can be found documented in 2047 * Documentation/networking/net_cachelines/net_device.rst. 2048 * Please update the document when adding new fields. 2049 */ 2050 2051 /* TX read-mostly hotpath */ 2052 __cacheline_group_begin(net_device_read_tx); 2053 struct_group(priv_flags_fast, 2054 unsigned long priv_flags:32; 2055 unsigned long lltx:1; 2056 ); 2057 const struct net_device_ops *netdev_ops; 2058 const struct header_ops *header_ops; 2059 struct netdev_queue *_tx; 2060 netdev_features_t gso_partial_features; 2061 unsigned int real_num_tx_queues; 2062 unsigned int gso_max_size; 2063 unsigned int gso_ipv4_max_size; 2064 u16 gso_max_segs; 2065 s16 num_tc; 2066 /* Note : dev->mtu is often read without holding a lock. 2067 * Writers usually hold RTNL. 2068 * It is recommended to use READ_ONCE() to annotate the reads, 2069 * and to use WRITE_ONCE() to annotate the writes. 2070 */ 2071 unsigned int mtu; 2072 unsigned short needed_headroom; 2073 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2074 #ifdef CONFIG_XPS 2075 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2076 #endif 2077 #ifdef CONFIG_NETFILTER_EGRESS 2078 struct nf_hook_entries __rcu *nf_hooks_egress; 2079 #endif 2080 #ifdef CONFIG_NET_XGRESS 2081 struct bpf_mprog_entry __rcu *tcx_egress; 2082 #endif 2083 __cacheline_group_end(net_device_read_tx); 2084 2085 /* TXRX read-mostly hotpath */ 2086 __cacheline_group_begin(net_device_read_txrx); 2087 union { 2088 struct pcpu_lstats __percpu *lstats; 2089 struct pcpu_sw_netstats __percpu *tstats; 2090 struct pcpu_dstats __percpu *dstats; 2091 }; 2092 unsigned long state; 2093 unsigned int flags; 2094 unsigned short hard_header_len; 2095 netdev_features_t features; 2096 struct inet6_dev __rcu *ip6_ptr; 2097 __cacheline_group_end(net_device_read_txrx); 2098 2099 /* RX read-mostly hotpath */ 2100 __cacheline_group_begin(net_device_read_rx); 2101 struct bpf_prog __rcu *xdp_prog; 2102 struct list_head ptype_specific; 2103 int ifindex; 2104 unsigned int real_num_rx_queues; 2105 struct netdev_rx_queue *_rx; 2106 unsigned int gro_max_size; 2107 unsigned int gro_ipv4_max_size; 2108 rx_handler_func_t __rcu *rx_handler; 2109 void __rcu *rx_handler_data; 2110 possible_net_t nd_net; 2111 #ifdef CONFIG_NETPOLL 2112 struct netpoll_info __rcu *npinfo; 2113 #endif 2114 #ifdef CONFIG_NET_XGRESS 2115 struct bpf_mprog_entry __rcu *tcx_ingress; 2116 #endif 2117 __cacheline_group_end(net_device_read_rx); 2118 2119 char name[IFNAMSIZ]; 2120 struct netdev_name_node *name_node; 2121 struct dev_ifalias __rcu *ifalias; 2122 /* 2123 * I/O specific fields 2124 * FIXME: Merge these and struct ifmap into one 2125 */ 2126 unsigned long mem_end; 2127 unsigned long mem_start; 2128 unsigned long base_addr; 2129 2130 /* 2131 * Some hardware also needs these fields (state,dev_list, 2132 * napi_list,unreg_list,close_list) but they are not 2133 * part of the usual set specified in Space.c. 2134 */ 2135 2136 2137 struct list_head dev_list; 2138 struct list_head napi_list; 2139 struct list_head unreg_list; 2140 struct list_head close_list; 2141 struct list_head ptype_all; 2142 2143 struct { 2144 struct list_head upper; 2145 struct list_head lower; 2146 } adj_list; 2147 2148 /* Read-mostly cache-line for fast-path access */ 2149 xdp_features_t xdp_features; 2150 const struct xdp_metadata_ops *xdp_metadata_ops; 2151 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops; 2152 unsigned short gflags; 2153 2154 unsigned short needed_tailroom; 2155 2156 netdev_features_t hw_features; 2157 netdev_features_t wanted_features; 2158 netdev_features_t vlan_features; 2159 netdev_features_t hw_enc_features; 2160 netdev_features_t mpls_features; 2161 2162 unsigned int min_mtu; 2163 unsigned int max_mtu; 2164 unsigned short type; 2165 unsigned char min_header_len; 2166 unsigned char name_assign_type; 2167 2168 int group; 2169 2170 struct net_device_stats stats; /* not used by modern drivers */ 2171 2172 struct net_device_core_stats __percpu *core_stats; 2173 2174 /* Stats to monitor link on/off, flapping */ 2175 atomic_t carrier_up_count; 2176 atomic_t carrier_down_count; 2177 2178 #ifdef CONFIG_WIRELESS_EXT 2179 const struct iw_handler_def *wireless_handlers; 2180 #endif 2181 const struct ethtool_ops *ethtool_ops; 2182 #ifdef CONFIG_NET_L3_MASTER_DEV 2183 const struct l3mdev_ops *l3mdev_ops; 2184 #endif 2185 #if IS_ENABLED(CONFIG_IPV6) 2186 const struct ndisc_ops *ndisc_ops; 2187 #endif 2188 2189 #ifdef CONFIG_XFRM_OFFLOAD 2190 const struct xfrmdev_ops *xfrmdev_ops; 2191 #endif 2192 2193 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2194 const struct tlsdev_ops *tlsdev_ops; 2195 #endif 2196 2197 unsigned int operstate; 2198 unsigned char link_mode; 2199 2200 unsigned char if_port; 2201 unsigned char dma; 2202 2203 /* Interface address info. */ 2204 unsigned char perm_addr[MAX_ADDR_LEN]; 2205 unsigned char addr_assign_type; 2206 unsigned char addr_len; 2207 unsigned char upper_level; 2208 unsigned char lower_level; 2209 2210 unsigned short neigh_priv_len; 2211 unsigned short dev_id; 2212 unsigned short dev_port; 2213 int irq; 2214 u32 priv_len; 2215 2216 spinlock_t addr_list_lock; 2217 2218 struct netdev_hw_addr_list uc; 2219 struct netdev_hw_addr_list mc; 2220 struct netdev_hw_addr_list dev_addrs; 2221 2222 #ifdef CONFIG_SYSFS 2223 struct kset *queues_kset; 2224 #endif 2225 #ifdef CONFIG_LOCKDEP 2226 struct list_head unlink_list; 2227 #endif 2228 unsigned int promiscuity; 2229 unsigned int allmulti; 2230 bool uc_promisc; 2231 #ifdef CONFIG_LOCKDEP 2232 unsigned char nested_level; 2233 #endif 2234 2235 2236 /* Protocol-specific pointers */ 2237 struct in_device __rcu *ip_ptr; 2238 /** @fib_nh_head: nexthops associated with this netdev */ 2239 struct hlist_head fib_nh_head; 2240 2241 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2242 struct vlan_info __rcu *vlan_info; 2243 #endif 2244 #if IS_ENABLED(CONFIG_NET_DSA) 2245 struct dsa_port *dsa_ptr; 2246 #endif 2247 #if IS_ENABLED(CONFIG_TIPC) 2248 struct tipc_bearer __rcu *tipc_ptr; 2249 #endif 2250 #if IS_ENABLED(CONFIG_ATALK) 2251 void *atalk_ptr; 2252 #endif 2253 #if IS_ENABLED(CONFIG_AX25) 2254 void *ax25_ptr; 2255 #endif 2256 #if IS_ENABLED(CONFIG_CFG80211) 2257 struct wireless_dev *ieee80211_ptr; 2258 #endif 2259 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2260 struct wpan_dev *ieee802154_ptr; 2261 #endif 2262 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2263 struct mpls_dev __rcu *mpls_ptr; 2264 #endif 2265 #if IS_ENABLED(CONFIG_MCTP) 2266 struct mctp_dev __rcu *mctp_ptr; 2267 #endif 2268 2269 /* 2270 * Cache lines mostly used on receive path (including eth_type_trans()) 2271 */ 2272 /* Interface address info used in eth_type_trans() */ 2273 const unsigned char *dev_addr; 2274 2275 unsigned int num_rx_queues; 2276 #define GRO_LEGACY_MAX_SIZE 65536u 2277 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2278 * and shinfo->gso_segs is a 16bit field. 2279 */ 2280 #define GRO_MAX_SIZE (8 * 65535u) 2281 unsigned int xdp_zc_max_segs; 2282 struct netdev_queue __rcu *ingress_queue; 2283 #ifdef CONFIG_NETFILTER_INGRESS 2284 struct nf_hook_entries __rcu *nf_hooks_ingress; 2285 #endif 2286 2287 unsigned char broadcast[MAX_ADDR_LEN]; 2288 #ifdef CONFIG_RFS_ACCEL 2289 struct cpu_rmap *rx_cpu_rmap; 2290 #endif 2291 struct hlist_node index_hlist; 2292 2293 /* 2294 * Cache lines mostly used on transmit path 2295 */ 2296 unsigned int num_tx_queues; 2297 struct Qdisc __rcu *qdisc; 2298 unsigned int tx_queue_len; 2299 spinlock_t tx_global_lock; 2300 2301 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2302 2303 #ifdef CONFIG_NET_SCHED 2304 DECLARE_HASHTABLE (qdisc_hash, 4); 2305 #endif 2306 /* These may be needed for future network-power-down code. */ 2307 struct timer_list watchdog_timer; 2308 int watchdog_timeo; 2309 2310 u32 proto_down_reason; 2311 2312 struct list_head todo_list; 2313 2314 #ifdef CONFIG_PCPU_DEV_REFCNT 2315 int __percpu *pcpu_refcnt; 2316 #else 2317 refcount_t dev_refcnt; 2318 #endif 2319 struct ref_tracker_dir refcnt_tracker; 2320 2321 struct list_head link_watch_list; 2322 2323 u8 reg_state; 2324 2325 bool dismantle; 2326 2327 enum { 2328 RTNL_LINK_INITIALIZED, 2329 RTNL_LINK_INITIALIZING, 2330 } rtnl_link_state:16; 2331 2332 bool needs_free_netdev; 2333 void (*priv_destructor)(struct net_device *dev); 2334 2335 /* mid-layer private */ 2336 void *ml_priv; 2337 enum netdev_ml_priv_type ml_priv_type; 2338 2339 enum netdev_stat_type pcpu_stat_type:8; 2340 2341 #if IS_ENABLED(CONFIG_GARP) 2342 struct garp_port __rcu *garp_port; 2343 #endif 2344 #if IS_ENABLED(CONFIG_MRP) 2345 struct mrp_port __rcu *mrp_port; 2346 #endif 2347 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2348 struct dm_hw_stat_delta __rcu *dm_private; 2349 #endif 2350 struct device dev; 2351 const struct attribute_group *sysfs_groups[4]; 2352 const struct attribute_group *sysfs_rx_queue_group; 2353 2354 const struct rtnl_link_ops *rtnl_link_ops; 2355 2356 const struct netdev_stat_ops *stat_ops; 2357 2358 const struct netdev_queue_mgmt_ops *queue_mgmt_ops; 2359 2360 /* for setting kernel sock attribute on TCP connection setup */ 2361 #define GSO_MAX_SEGS 65535u 2362 #define GSO_LEGACY_MAX_SIZE 65536u 2363 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2364 * and shinfo->gso_segs is a 16bit field. 2365 */ 2366 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2367 2368 #define TSO_LEGACY_MAX_SIZE 65536 2369 #define TSO_MAX_SIZE UINT_MAX 2370 unsigned int tso_max_size; 2371 #define TSO_MAX_SEGS U16_MAX 2372 u16 tso_max_segs; 2373 2374 #ifdef CONFIG_DCB 2375 const struct dcbnl_rtnl_ops *dcbnl_ops; 2376 #endif 2377 u8 prio_tc_map[TC_BITMASK + 1]; 2378 2379 #if IS_ENABLED(CONFIG_FCOE) 2380 unsigned int fcoe_ddp_xid; 2381 #endif 2382 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2383 struct netprio_map __rcu *priomap; 2384 #endif 2385 struct phy_link_topology *link_topo; 2386 struct phy_device *phydev; 2387 struct sfp_bus *sfp_bus; 2388 struct lock_class_key *qdisc_tx_busylock; 2389 bool proto_down; 2390 bool threaded; 2391 2392 /* priv_flags_slow, ungrouped to save space */ 2393 unsigned long see_all_hwtstamp_requests:1; 2394 unsigned long change_proto_down:1; 2395 unsigned long netns_local:1; 2396 unsigned long fcoe_mtu:1; 2397 2398 struct list_head net_notifier_list; 2399 2400 #if IS_ENABLED(CONFIG_MACSEC) 2401 /* MACsec management functions */ 2402 const struct macsec_ops *macsec_ops; 2403 #endif 2404 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2405 struct udp_tunnel_nic *udp_tunnel_nic; 2406 2407 struct ethtool_netdev_state *ethtool; 2408 2409 /* protected by rtnl_lock */ 2410 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2411 2412 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2413 netdevice_tracker linkwatch_dev_tracker; 2414 netdevice_tracker watchdog_dev_tracker; 2415 netdevice_tracker dev_registered_tracker; 2416 struct rtnl_hw_stats64 *offload_xstats_l3; 2417 2418 struct devlink_port *devlink_port; 2419 2420 #if IS_ENABLED(CONFIG_DPLL) 2421 struct dpll_pin __rcu *dpll_pin; 2422 #endif 2423 #if IS_ENABLED(CONFIG_PAGE_POOL) 2424 /** @page_pools: page pools created for this netdevice */ 2425 struct hlist_head page_pools; 2426 #endif 2427 2428 /** @irq_moder: dim parameters used if IS_ENABLED(CONFIG_DIMLIB). */ 2429 struct dim_irq_moder *irq_moder; 2430 2431 u64 max_pacing_offload_horizon; 2432 struct napi_config *napi_config; 2433 unsigned long gro_flush_timeout; 2434 u32 napi_defer_hard_irqs; 2435 2436 /** 2437 * @lock: protects @net_shaper_hierarchy, feel free to use for other 2438 * netdev-scope protection. Ordering: take after rtnl_lock. 2439 */ 2440 struct mutex lock; 2441 2442 #if IS_ENABLED(CONFIG_NET_SHAPER) 2443 /** 2444 * @net_shaper_hierarchy: data tracking the current shaper status 2445 * see include/net/net_shapers.h 2446 */ 2447 struct net_shaper_hierarchy *net_shaper_hierarchy; 2448 #endif 2449 2450 struct hlist_head neighbours[NEIGH_NR_TABLES]; 2451 2452 u8 priv[] ____cacheline_aligned 2453 __counted_by(priv_len); 2454 } ____cacheline_aligned; 2455 #define to_net_dev(d) container_of(d, struct net_device, dev) 2456 2457 /* 2458 * Driver should use this to assign devlink port instance to a netdevice 2459 * before it registers the netdevice. Therefore devlink_port is static 2460 * during the netdev lifetime after it is registered. 2461 */ 2462 #define SET_NETDEV_DEVLINK_PORT(dev, port) \ 2463 ({ \ 2464 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \ 2465 ((dev)->devlink_port = (port)); \ 2466 }) 2467 2468 static inline bool netif_elide_gro(const struct net_device *dev) 2469 { 2470 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2471 return true; 2472 return false; 2473 } 2474 2475 #define NETDEV_ALIGN 32 2476 2477 static inline 2478 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2479 { 2480 return dev->prio_tc_map[prio & TC_BITMASK]; 2481 } 2482 2483 static inline 2484 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2485 { 2486 if (tc >= dev->num_tc) 2487 return -EINVAL; 2488 2489 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2490 return 0; 2491 } 2492 2493 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2494 void netdev_reset_tc(struct net_device *dev); 2495 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2496 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2497 2498 static inline 2499 int netdev_get_num_tc(struct net_device *dev) 2500 { 2501 return dev->num_tc; 2502 } 2503 2504 static inline void net_prefetch(void *p) 2505 { 2506 prefetch(p); 2507 #if L1_CACHE_BYTES < 128 2508 prefetch((u8 *)p + L1_CACHE_BYTES); 2509 #endif 2510 } 2511 2512 static inline void net_prefetchw(void *p) 2513 { 2514 prefetchw(p); 2515 #if L1_CACHE_BYTES < 128 2516 prefetchw((u8 *)p + L1_CACHE_BYTES); 2517 #endif 2518 } 2519 2520 void netdev_unbind_sb_channel(struct net_device *dev, 2521 struct net_device *sb_dev); 2522 int netdev_bind_sb_channel_queue(struct net_device *dev, 2523 struct net_device *sb_dev, 2524 u8 tc, u16 count, u16 offset); 2525 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2526 static inline int netdev_get_sb_channel(struct net_device *dev) 2527 { 2528 return max_t(int, -dev->num_tc, 0); 2529 } 2530 2531 static inline 2532 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2533 unsigned int index) 2534 { 2535 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues); 2536 return &dev->_tx[index]; 2537 } 2538 2539 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2540 const struct sk_buff *skb) 2541 { 2542 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2543 } 2544 2545 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2546 void (*f)(struct net_device *, 2547 struct netdev_queue *, 2548 void *), 2549 void *arg) 2550 { 2551 unsigned int i; 2552 2553 for (i = 0; i < dev->num_tx_queues; i++) 2554 f(dev, &dev->_tx[i], arg); 2555 } 2556 2557 #define netdev_lockdep_set_classes(dev) \ 2558 { \ 2559 static struct lock_class_key qdisc_tx_busylock_key; \ 2560 static struct lock_class_key qdisc_xmit_lock_key; \ 2561 static struct lock_class_key dev_addr_list_lock_key; \ 2562 unsigned int i; \ 2563 \ 2564 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2565 lockdep_set_class(&(dev)->addr_list_lock, \ 2566 &dev_addr_list_lock_key); \ 2567 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2568 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2569 &qdisc_xmit_lock_key); \ 2570 } 2571 2572 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2573 struct net_device *sb_dev); 2574 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2575 struct sk_buff *skb, 2576 struct net_device *sb_dev); 2577 2578 /* returns the headroom that the master device needs to take in account 2579 * when forwarding to this dev 2580 */ 2581 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2582 { 2583 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2584 } 2585 2586 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2587 { 2588 if (dev->netdev_ops->ndo_set_rx_headroom) 2589 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2590 } 2591 2592 /* set the device rx headroom to the dev's default */ 2593 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2594 { 2595 netdev_set_rx_headroom(dev, -1); 2596 } 2597 2598 static inline void *netdev_get_ml_priv(struct net_device *dev, 2599 enum netdev_ml_priv_type type) 2600 { 2601 if (dev->ml_priv_type != type) 2602 return NULL; 2603 2604 return dev->ml_priv; 2605 } 2606 2607 static inline void netdev_set_ml_priv(struct net_device *dev, 2608 void *ml_priv, 2609 enum netdev_ml_priv_type type) 2610 { 2611 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2612 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2613 dev->ml_priv_type, type); 2614 WARN(!dev->ml_priv_type && dev->ml_priv, 2615 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2616 2617 dev->ml_priv = ml_priv; 2618 dev->ml_priv_type = type; 2619 } 2620 2621 /* 2622 * Net namespace inlines 2623 */ 2624 static inline 2625 struct net *dev_net(const struct net_device *dev) 2626 { 2627 return read_pnet(&dev->nd_net); 2628 } 2629 2630 static inline 2631 void dev_net_set(struct net_device *dev, struct net *net) 2632 { 2633 write_pnet(&dev->nd_net, net); 2634 } 2635 2636 /** 2637 * netdev_priv - access network device private data 2638 * @dev: network device 2639 * 2640 * Get network device private data 2641 */ 2642 static inline void *netdev_priv(const struct net_device *dev) 2643 { 2644 return (void *)dev->priv; 2645 } 2646 2647 /* Set the sysfs physical device reference for the network logical device 2648 * if set prior to registration will cause a symlink during initialization. 2649 */ 2650 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2651 2652 /* Set the sysfs device type for the network logical device to allow 2653 * fine-grained identification of different network device types. For 2654 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2655 */ 2656 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2657 2658 void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index, 2659 enum netdev_queue_type type, 2660 struct napi_struct *napi); 2661 2662 static inline void netif_napi_set_irq(struct napi_struct *napi, int irq) 2663 { 2664 napi->irq = irq; 2665 } 2666 2667 /* Default NAPI poll() weight 2668 * Device drivers are strongly advised to not use bigger value 2669 */ 2670 #define NAPI_POLL_WEIGHT 64 2671 2672 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2673 int (*poll)(struct napi_struct *, int), int weight); 2674 2675 /** 2676 * netif_napi_add() - initialize a NAPI context 2677 * @dev: network device 2678 * @napi: NAPI context 2679 * @poll: polling function 2680 * 2681 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2682 * *any* of the other NAPI-related functions. 2683 */ 2684 static inline void 2685 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2686 int (*poll)(struct napi_struct *, int)) 2687 { 2688 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2689 } 2690 2691 static inline void 2692 netif_napi_add_tx_weight(struct net_device *dev, 2693 struct napi_struct *napi, 2694 int (*poll)(struct napi_struct *, int), 2695 int weight) 2696 { 2697 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2698 netif_napi_add_weight(dev, napi, poll, weight); 2699 } 2700 2701 /** 2702 * netif_napi_add_config - initialize a NAPI context with persistent config 2703 * @dev: network device 2704 * @napi: NAPI context 2705 * @poll: polling function 2706 * @index: the NAPI index 2707 */ 2708 static inline void 2709 netif_napi_add_config(struct net_device *dev, struct napi_struct *napi, 2710 int (*poll)(struct napi_struct *, int), int index) 2711 { 2712 napi->index = index; 2713 napi->config = &dev->napi_config[index]; 2714 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2715 } 2716 2717 /** 2718 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2719 * @dev: network device 2720 * @napi: NAPI context 2721 * @poll: polling function 2722 * 2723 * This variant of netif_napi_add() should be used from drivers using NAPI 2724 * to exclusively poll a TX queue. 2725 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2726 */ 2727 static inline void netif_napi_add_tx(struct net_device *dev, 2728 struct napi_struct *napi, 2729 int (*poll)(struct napi_struct *, int)) 2730 { 2731 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2732 } 2733 2734 /** 2735 * __netif_napi_del - remove a NAPI context 2736 * @napi: NAPI context 2737 * 2738 * Warning: caller must observe RCU grace period before freeing memory 2739 * containing @napi. Drivers might want to call this helper to combine 2740 * all the needed RCU grace periods into a single one. 2741 */ 2742 void __netif_napi_del(struct napi_struct *napi); 2743 2744 /** 2745 * netif_napi_del - remove a NAPI context 2746 * @napi: NAPI context 2747 * 2748 * netif_napi_del() removes a NAPI context from the network device NAPI list 2749 */ 2750 static inline void netif_napi_del(struct napi_struct *napi) 2751 { 2752 __netif_napi_del(napi); 2753 synchronize_net(); 2754 } 2755 2756 struct packet_type { 2757 __be16 type; /* This is really htons(ether_type). */ 2758 bool ignore_outgoing; 2759 struct net_device *dev; /* NULL is wildcarded here */ 2760 netdevice_tracker dev_tracker; 2761 int (*func) (struct sk_buff *, 2762 struct net_device *, 2763 struct packet_type *, 2764 struct net_device *); 2765 void (*list_func) (struct list_head *, 2766 struct packet_type *, 2767 struct net_device *); 2768 bool (*id_match)(struct packet_type *ptype, 2769 struct sock *sk); 2770 struct net *af_packet_net; 2771 void *af_packet_priv; 2772 struct list_head list; 2773 }; 2774 2775 struct offload_callbacks { 2776 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2777 netdev_features_t features); 2778 struct sk_buff *(*gro_receive)(struct list_head *head, 2779 struct sk_buff *skb); 2780 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2781 }; 2782 2783 struct packet_offload { 2784 __be16 type; /* This is really htons(ether_type). */ 2785 u16 priority; 2786 struct offload_callbacks callbacks; 2787 struct list_head list; 2788 }; 2789 2790 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2791 struct pcpu_sw_netstats { 2792 u64_stats_t rx_packets; 2793 u64_stats_t rx_bytes; 2794 u64_stats_t tx_packets; 2795 u64_stats_t tx_bytes; 2796 struct u64_stats_sync syncp; 2797 } __aligned(4 * sizeof(u64)); 2798 2799 struct pcpu_dstats { 2800 u64_stats_t rx_packets; 2801 u64_stats_t rx_bytes; 2802 u64_stats_t rx_drops; 2803 u64_stats_t tx_packets; 2804 u64_stats_t tx_bytes; 2805 u64_stats_t tx_drops; 2806 struct u64_stats_sync syncp; 2807 } __aligned(8 * sizeof(u64)); 2808 2809 struct pcpu_lstats { 2810 u64_stats_t packets; 2811 u64_stats_t bytes; 2812 struct u64_stats_sync syncp; 2813 } __aligned(2 * sizeof(u64)); 2814 2815 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2816 2817 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2818 { 2819 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2820 2821 u64_stats_update_begin(&tstats->syncp); 2822 u64_stats_add(&tstats->rx_bytes, len); 2823 u64_stats_inc(&tstats->rx_packets); 2824 u64_stats_update_end(&tstats->syncp); 2825 } 2826 2827 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2828 unsigned int packets, 2829 unsigned int len) 2830 { 2831 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2832 2833 u64_stats_update_begin(&tstats->syncp); 2834 u64_stats_add(&tstats->tx_bytes, len); 2835 u64_stats_add(&tstats->tx_packets, packets); 2836 u64_stats_update_end(&tstats->syncp); 2837 } 2838 2839 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2840 { 2841 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2842 2843 u64_stats_update_begin(&lstats->syncp); 2844 u64_stats_add(&lstats->bytes, len); 2845 u64_stats_inc(&lstats->packets); 2846 u64_stats_update_end(&lstats->syncp); 2847 } 2848 2849 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2850 ({ \ 2851 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2852 if (pcpu_stats) { \ 2853 int __cpu; \ 2854 for_each_possible_cpu(__cpu) { \ 2855 typeof(type) *stat; \ 2856 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2857 u64_stats_init(&stat->syncp); \ 2858 } \ 2859 } \ 2860 pcpu_stats; \ 2861 }) 2862 2863 #define netdev_alloc_pcpu_stats(type) \ 2864 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2865 2866 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2867 ({ \ 2868 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2869 if (pcpu_stats) { \ 2870 int __cpu; \ 2871 for_each_possible_cpu(__cpu) { \ 2872 typeof(type) *stat; \ 2873 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2874 u64_stats_init(&stat->syncp); \ 2875 } \ 2876 } \ 2877 pcpu_stats; \ 2878 }) 2879 2880 enum netdev_lag_tx_type { 2881 NETDEV_LAG_TX_TYPE_UNKNOWN, 2882 NETDEV_LAG_TX_TYPE_RANDOM, 2883 NETDEV_LAG_TX_TYPE_BROADCAST, 2884 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2885 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2886 NETDEV_LAG_TX_TYPE_HASH, 2887 }; 2888 2889 enum netdev_lag_hash { 2890 NETDEV_LAG_HASH_NONE, 2891 NETDEV_LAG_HASH_L2, 2892 NETDEV_LAG_HASH_L34, 2893 NETDEV_LAG_HASH_L23, 2894 NETDEV_LAG_HASH_E23, 2895 NETDEV_LAG_HASH_E34, 2896 NETDEV_LAG_HASH_VLAN_SRCMAC, 2897 NETDEV_LAG_HASH_UNKNOWN, 2898 }; 2899 2900 struct netdev_lag_upper_info { 2901 enum netdev_lag_tx_type tx_type; 2902 enum netdev_lag_hash hash_type; 2903 }; 2904 2905 struct netdev_lag_lower_state_info { 2906 u8 link_up : 1, 2907 tx_enabled : 1; 2908 }; 2909 2910 #include <linux/notifier.h> 2911 2912 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2913 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2914 * adding new types. 2915 */ 2916 enum netdev_cmd { 2917 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2918 NETDEV_DOWN, 2919 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2920 detected a hardware crash and restarted 2921 - we can use this eg to kick tcp sessions 2922 once done */ 2923 NETDEV_CHANGE, /* Notify device state change */ 2924 NETDEV_REGISTER, 2925 NETDEV_UNREGISTER, 2926 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2927 NETDEV_CHANGEADDR, /* notify after the address change */ 2928 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2929 NETDEV_GOING_DOWN, 2930 NETDEV_CHANGENAME, 2931 NETDEV_FEAT_CHANGE, 2932 NETDEV_BONDING_FAILOVER, 2933 NETDEV_PRE_UP, 2934 NETDEV_PRE_TYPE_CHANGE, 2935 NETDEV_POST_TYPE_CHANGE, 2936 NETDEV_POST_INIT, 2937 NETDEV_PRE_UNINIT, 2938 NETDEV_RELEASE, 2939 NETDEV_NOTIFY_PEERS, 2940 NETDEV_JOIN, 2941 NETDEV_CHANGEUPPER, 2942 NETDEV_RESEND_IGMP, 2943 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2944 NETDEV_CHANGEINFODATA, 2945 NETDEV_BONDING_INFO, 2946 NETDEV_PRECHANGEUPPER, 2947 NETDEV_CHANGELOWERSTATE, 2948 NETDEV_UDP_TUNNEL_PUSH_INFO, 2949 NETDEV_UDP_TUNNEL_DROP_INFO, 2950 NETDEV_CHANGE_TX_QUEUE_LEN, 2951 NETDEV_CVLAN_FILTER_PUSH_INFO, 2952 NETDEV_CVLAN_FILTER_DROP_INFO, 2953 NETDEV_SVLAN_FILTER_PUSH_INFO, 2954 NETDEV_SVLAN_FILTER_DROP_INFO, 2955 NETDEV_OFFLOAD_XSTATS_ENABLE, 2956 NETDEV_OFFLOAD_XSTATS_DISABLE, 2957 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 2958 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 2959 NETDEV_XDP_FEAT_CHANGE, 2960 }; 2961 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2962 2963 int register_netdevice_notifier(struct notifier_block *nb); 2964 int unregister_netdevice_notifier(struct notifier_block *nb); 2965 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2966 int unregister_netdevice_notifier_net(struct net *net, 2967 struct notifier_block *nb); 2968 int register_netdevice_notifier_dev_net(struct net_device *dev, 2969 struct notifier_block *nb, 2970 struct netdev_net_notifier *nn); 2971 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2972 struct notifier_block *nb, 2973 struct netdev_net_notifier *nn); 2974 2975 struct netdev_notifier_info { 2976 struct net_device *dev; 2977 struct netlink_ext_ack *extack; 2978 }; 2979 2980 struct netdev_notifier_info_ext { 2981 struct netdev_notifier_info info; /* must be first */ 2982 union { 2983 u32 mtu; 2984 } ext; 2985 }; 2986 2987 struct netdev_notifier_change_info { 2988 struct netdev_notifier_info info; /* must be first */ 2989 unsigned int flags_changed; 2990 }; 2991 2992 struct netdev_notifier_changeupper_info { 2993 struct netdev_notifier_info info; /* must be first */ 2994 struct net_device *upper_dev; /* new upper dev */ 2995 bool master; /* is upper dev master */ 2996 bool linking; /* is the notification for link or unlink */ 2997 void *upper_info; /* upper dev info */ 2998 }; 2999 3000 struct netdev_notifier_changelowerstate_info { 3001 struct netdev_notifier_info info; /* must be first */ 3002 void *lower_state_info; /* is lower dev state */ 3003 }; 3004 3005 struct netdev_notifier_pre_changeaddr_info { 3006 struct netdev_notifier_info info; /* must be first */ 3007 const unsigned char *dev_addr; 3008 }; 3009 3010 enum netdev_offload_xstats_type { 3011 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 3012 }; 3013 3014 struct netdev_notifier_offload_xstats_info { 3015 struct netdev_notifier_info info; /* must be first */ 3016 enum netdev_offload_xstats_type type; 3017 3018 union { 3019 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 3020 struct netdev_notifier_offload_xstats_rd *report_delta; 3021 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 3022 struct netdev_notifier_offload_xstats_ru *report_used; 3023 }; 3024 }; 3025 3026 int netdev_offload_xstats_enable(struct net_device *dev, 3027 enum netdev_offload_xstats_type type, 3028 struct netlink_ext_ack *extack); 3029 int netdev_offload_xstats_disable(struct net_device *dev, 3030 enum netdev_offload_xstats_type type); 3031 bool netdev_offload_xstats_enabled(const struct net_device *dev, 3032 enum netdev_offload_xstats_type type); 3033 int netdev_offload_xstats_get(struct net_device *dev, 3034 enum netdev_offload_xstats_type type, 3035 struct rtnl_hw_stats64 *stats, bool *used, 3036 struct netlink_ext_ack *extack); 3037 void 3038 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 3039 const struct rtnl_hw_stats64 *stats); 3040 void 3041 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 3042 void netdev_offload_xstats_push_delta(struct net_device *dev, 3043 enum netdev_offload_xstats_type type, 3044 const struct rtnl_hw_stats64 *stats); 3045 3046 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 3047 struct net_device *dev) 3048 { 3049 info->dev = dev; 3050 info->extack = NULL; 3051 } 3052 3053 static inline struct net_device * 3054 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 3055 { 3056 return info->dev; 3057 } 3058 3059 static inline struct netlink_ext_ack * 3060 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 3061 { 3062 return info->extack; 3063 } 3064 3065 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 3066 int call_netdevice_notifiers_info(unsigned long val, 3067 struct netdev_notifier_info *info); 3068 3069 #define for_each_netdev(net, d) \ 3070 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 3071 #define for_each_netdev_reverse(net, d) \ 3072 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 3073 #define for_each_netdev_rcu(net, d) \ 3074 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 3075 #define for_each_netdev_safe(net, d, n) \ 3076 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 3077 #define for_each_netdev_continue(net, d) \ 3078 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 3079 #define for_each_netdev_continue_reverse(net, d) \ 3080 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 3081 dev_list) 3082 #define for_each_netdev_continue_rcu(net, d) \ 3083 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 3084 #define for_each_netdev_in_bond_rcu(bond, slave) \ 3085 for_each_netdev_rcu(&init_net, slave) \ 3086 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 3087 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 3088 3089 #define for_each_netdev_dump(net, d, ifindex) \ 3090 for (; (d = xa_find(&(net)->dev_by_index, &ifindex, \ 3091 ULONG_MAX, XA_PRESENT)); ifindex++) 3092 3093 static inline struct net_device *next_net_device(struct net_device *dev) 3094 { 3095 struct list_head *lh; 3096 struct net *net; 3097 3098 net = dev_net(dev); 3099 lh = dev->dev_list.next; 3100 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3101 } 3102 3103 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 3104 { 3105 struct list_head *lh; 3106 struct net *net; 3107 3108 net = dev_net(dev); 3109 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 3110 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3111 } 3112 3113 static inline struct net_device *first_net_device(struct net *net) 3114 { 3115 return list_empty(&net->dev_base_head) ? NULL : 3116 net_device_entry(net->dev_base_head.next); 3117 } 3118 3119 static inline struct net_device *first_net_device_rcu(struct net *net) 3120 { 3121 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 3122 3123 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3124 } 3125 3126 int netdev_boot_setup_check(struct net_device *dev); 3127 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 3128 const char *hwaddr); 3129 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 3130 void dev_add_pack(struct packet_type *pt); 3131 void dev_remove_pack(struct packet_type *pt); 3132 void __dev_remove_pack(struct packet_type *pt); 3133 void dev_add_offload(struct packet_offload *po); 3134 void dev_remove_offload(struct packet_offload *po); 3135 3136 int dev_get_iflink(const struct net_device *dev); 3137 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 3138 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 3139 struct net_device_path_stack *stack); 3140 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 3141 unsigned short mask); 3142 struct net_device *dev_get_by_name(struct net *net, const char *name); 3143 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 3144 struct net_device *__dev_get_by_name(struct net *net, const char *name); 3145 bool netdev_name_in_use(struct net *net, const char *name); 3146 int dev_alloc_name(struct net_device *dev, const char *name); 3147 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 3148 void dev_close(struct net_device *dev); 3149 void dev_close_many(struct list_head *head, bool unlink); 3150 void dev_disable_lro(struct net_device *dev); 3151 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 3152 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 3153 struct net_device *sb_dev); 3154 3155 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3156 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3157 3158 static inline int dev_queue_xmit(struct sk_buff *skb) 3159 { 3160 return __dev_queue_xmit(skb, NULL); 3161 } 3162 3163 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3164 struct net_device *sb_dev) 3165 { 3166 return __dev_queue_xmit(skb, sb_dev); 3167 } 3168 3169 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3170 { 3171 int ret; 3172 3173 ret = __dev_direct_xmit(skb, queue_id); 3174 if (!dev_xmit_complete(ret)) 3175 kfree_skb(skb); 3176 return ret; 3177 } 3178 3179 int register_netdevice(struct net_device *dev); 3180 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3181 void unregister_netdevice_many(struct list_head *head); 3182 static inline void unregister_netdevice(struct net_device *dev) 3183 { 3184 unregister_netdevice_queue(dev, NULL); 3185 } 3186 3187 int netdev_refcnt_read(const struct net_device *dev); 3188 void free_netdev(struct net_device *dev); 3189 void init_dummy_netdev(struct net_device *dev); 3190 3191 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3192 struct sk_buff *skb, 3193 bool all_slaves); 3194 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3195 struct sock *sk); 3196 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3197 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3198 struct net_device *netdev_get_by_index(struct net *net, int ifindex, 3199 netdevice_tracker *tracker, gfp_t gfp); 3200 struct net_device *netdev_get_by_name(struct net *net, const char *name, 3201 netdevice_tracker *tracker, gfp_t gfp); 3202 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3203 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 3204 void netdev_copy_name(struct net_device *dev, char *name); 3205 3206 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3207 unsigned short type, 3208 const void *daddr, const void *saddr, 3209 unsigned int len) 3210 { 3211 if (!dev->header_ops || !dev->header_ops->create) 3212 return 0; 3213 3214 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3215 } 3216 3217 static inline int dev_parse_header(const struct sk_buff *skb, 3218 unsigned char *haddr) 3219 { 3220 const struct net_device *dev = skb->dev; 3221 3222 if (!dev->header_ops || !dev->header_ops->parse) 3223 return 0; 3224 return dev->header_ops->parse(skb, haddr); 3225 } 3226 3227 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3228 { 3229 const struct net_device *dev = skb->dev; 3230 3231 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3232 return 0; 3233 return dev->header_ops->parse_protocol(skb); 3234 } 3235 3236 /* ll_header must have at least hard_header_len allocated */ 3237 static inline bool dev_validate_header(const struct net_device *dev, 3238 char *ll_header, int len) 3239 { 3240 if (likely(len >= dev->hard_header_len)) 3241 return true; 3242 if (len < dev->min_header_len) 3243 return false; 3244 3245 if (capable(CAP_SYS_RAWIO)) { 3246 memset(ll_header + len, 0, dev->hard_header_len - len); 3247 return true; 3248 } 3249 3250 if (dev->header_ops && dev->header_ops->validate) 3251 return dev->header_ops->validate(ll_header, len); 3252 3253 return false; 3254 } 3255 3256 static inline bool dev_has_header(const struct net_device *dev) 3257 { 3258 return dev->header_ops && dev->header_ops->create; 3259 } 3260 3261 /* 3262 * Incoming packets are placed on per-CPU queues 3263 */ 3264 struct softnet_data { 3265 struct list_head poll_list; 3266 struct sk_buff_head process_queue; 3267 local_lock_t process_queue_bh_lock; 3268 3269 /* stats */ 3270 unsigned int processed; 3271 unsigned int time_squeeze; 3272 #ifdef CONFIG_RPS 3273 struct softnet_data *rps_ipi_list; 3274 #endif 3275 3276 unsigned int received_rps; 3277 bool in_net_rx_action; 3278 bool in_napi_threaded_poll; 3279 3280 #ifdef CONFIG_NET_FLOW_LIMIT 3281 struct sd_flow_limit __rcu *flow_limit; 3282 #endif 3283 struct Qdisc *output_queue; 3284 struct Qdisc **output_queue_tailp; 3285 struct sk_buff *completion_queue; 3286 #ifdef CONFIG_XFRM_OFFLOAD 3287 struct sk_buff_head xfrm_backlog; 3288 #endif 3289 /* written and read only by owning cpu: */ 3290 struct netdev_xmit xmit; 3291 #ifdef CONFIG_RPS 3292 /* input_queue_head should be written by cpu owning this struct, 3293 * and only read by other cpus. Worth using a cache line. 3294 */ 3295 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3296 3297 /* Elements below can be accessed between CPUs for RPS/RFS */ 3298 call_single_data_t csd ____cacheline_aligned_in_smp; 3299 struct softnet_data *rps_ipi_next; 3300 unsigned int cpu; 3301 unsigned int input_queue_tail; 3302 #endif 3303 struct sk_buff_head input_pkt_queue; 3304 struct napi_struct backlog; 3305 3306 atomic_t dropped ____cacheline_aligned_in_smp; 3307 3308 /* Another possibly contended cache line */ 3309 spinlock_t defer_lock ____cacheline_aligned_in_smp; 3310 int defer_count; 3311 int defer_ipi_scheduled; 3312 struct sk_buff *defer_list; 3313 call_single_data_t defer_csd; 3314 }; 3315 3316 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3317 3318 #ifndef CONFIG_PREEMPT_RT 3319 static inline int dev_recursion_level(void) 3320 { 3321 return this_cpu_read(softnet_data.xmit.recursion); 3322 } 3323 #else 3324 static inline int dev_recursion_level(void) 3325 { 3326 return current->net_xmit.recursion; 3327 } 3328 3329 #endif 3330 3331 void __netif_schedule(struct Qdisc *q); 3332 void netif_schedule_queue(struct netdev_queue *txq); 3333 3334 static inline void netif_tx_schedule_all(struct net_device *dev) 3335 { 3336 unsigned int i; 3337 3338 for (i = 0; i < dev->num_tx_queues; i++) 3339 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3340 } 3341 3342 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3343 { 3344 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3345 } 3346 3347 /** 3348 * netif_start_queue - allow transmit 3349 * @dev: network device 3350 * 3351 * Allow upper layers to call the device hard_start_xmit routine. 3352 */ 3353 static inline void netif_start_queue(struct net_device *dev) 3354 { 3355 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3356 } 3357 3358 static inline void netif_tx_start_all_queues(struct net_device *dev) 3359 { 3360 unsigned int i; 3361 3362 for (i = 0; i < dev->num_tx_queues; i++) { 3363 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3364 netif_tx_start_queue(txq); 3365 } 3366 } 3367 3368 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3369 3370 /** 3371 * netif_wake_queue - restart transmit 3372 * @dev: network device 3373 * 3374 * Allow upper layers to call the device hard_start_xmit routine. 3375 * Used for flow control when transmit resources are available. 3376 */ 3377 static inline void netif_wake_queue(struct net_device *dev) 3378 { 3379 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3380 } 3381 3382 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3383 { 3384 unsigned int i; 3385 3386 for (i = 0; i < dev->num_tx_queues; i++) { 3387 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3388 netif_tx_wake_queue(txq); 3389 } 3390 } 3391 3392 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3393 { 3394 /* Paired with READ_ONCE() from dev_watchdog() */ 3395 WRITE_ONCE(dev_queue->trans_start, jiffies); 3396 3397 /* This barrier is paired with smp_mb() from dev_watchdog() */ 3398 smp_mb__before_atomic(); 3399 3400 /* Must be an atomic op see netif_txq_try_stop() */ 3401 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3402 } 3403 3404 /** 3405 * netif_stop_queue - stop transmitted packets 3406 * @dev: network device 3407 * 3408 * Stop upper layers calling the device hard_start_xmit routine. 3409 * Used for flow control when transmit resources are unavailable. 3410 */ 3411 static inline void netif_stop_queue(struct net_device *dev) 3412 { 3413 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3414 } 3415 3416 void netif_tx_stop_all_queues(struct net_device *dev); 3417 3418 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3419 { 3420 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3421 } 3422 3423 /** 3424 * netif_queue_stopped - test if transmit queue is flowblocked 3425 * @dev: network device 3426 * 3427 * Test if transmit queue on device is currently unable to send. 3428 */ 3429 static inline bool netif_queue_stopped(const struct net_device *dev) 3430 { 3431 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3432 } 3433 3434 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3435 { 3436 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3437 } 3438 3439 static inline bool 3440 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3441 { 3442 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3443 } 3444 3445 static inline bool 3446 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3447 { 3448 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3449 } 3450 3451 /** 3452 * netdev_queue_set_dql_min_limit - set dql minimum limit 3453 * @dev_queue: pointer to transmit queue 3454 * @min_limit: dql minimum limit 3455 * 3456 * Forces xmit_more() to return true until the minimum threshold 3457 * defined by @min_limit is reached (or until the tx queue is 3458 * empty). Warning: to be use with care, misuse will impact the 3459 * latency. 3460 */ 3461 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3462 unsigned int min_limit) 3463 { 3464 #ifdef CONFIG_BQL 3465 dev_queue->dql.min_limit = min_limit; 3466 #endif 3467 } 3468 3469 static inline int netdev_queue_dql_avail(const struct netdev_queue *txq) 3470 { 3471 #ifdef CONFIG_BQL 3472 /* Non-BQL migrated drivers will return 0, too. */ 3473 return dql_avail(&txq->dql); 3474 #else 3475 return 0; 3476 #endif 3477 } 3478 3479 /** 3480 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3481 * @dev_queue: pointer to transmit queue 3482 * 3483 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3484 * to give appropriate hint to the CPU. 3485 */ 3486 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3487 { 3488 #ifdef CONFIG_BQL 3489 prefetchw(&dev_queue->dql.num_queued); 3490 #endif 3491 } 3492 3493 /** 3494 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3495 * @dev_queue: pointer to transmit queue 3496 * 3497 * BQL enabled drivers might use this helper in their TX completion path, 3498 * to give appropriate hint to the CPU. 3499 */ 3500 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3501 { 3502 #ifdef CONFIG_BQL 3503 prefetchw(&dev_queue->dql.limit); 3504 #endif 3505 } 3506 3507 /** 3508 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3509 * @dev_queue: network device queue 3510 * @bytes: number of bytes queued to the device queue 3511 * 3512 * Report the number of bytes queued for sending/completion to the network 3513 * device hardware queue. @bytes should be a good approximation and should 3514 * exactly match netdev_completed_queue() @bytes. 3515 * This is typically called once per packet, from ndo_start_xmit(). 3516 */ 3517 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3518 unsigned int bytes) 3519 { 3520 #ifdef CONFIG_BQL 3521 dql_queued(&dev_queue->dql, bytes); 3522 3523 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3524 return; 3525 3526 /* Paired with READ_ONCE() from dev_watchdog() */ 3527 WRITE_ONCE(dev_queue->trans_start, jiffies); 3528 3529 /* This barrier is paired with smp_mb() from dev_watchdog() */ 3530 smp_mb__before_atomic(); 3531 3532 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3533 3534 /* 3535 * The XOFF flag must be set before checking the dql_avail below, 3536 * because in netdev_tx_completed_queue we update the dql_completed 3537 * before checking the XOFF flag. 3538 */ 3539 smp_mb__after_atomic(); 3540 3541 /* check again in case another CPU has just made room avail */ 3542 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3543 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3544 #endif 3545 } 3546 3547 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3548 * that they should not test BQL status themselves. 3549 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3550 * skb of a batch. 3551 * Returns true if the doorbell must be used to kick the NIC. 3552 */ 3553 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3554 unsigned int bytes, 3555 bool xmit_more) 3556 { 3557 if (xmit_more) { 3558 #ifdef CONFIG_BQL 3559 dql_queued(&dev_queue->dql, bytes); 3560 #endif 3561 return netif_tx_queue_stopped(dev_queue); 3562 } 3563 netdev_tx_sent_queue(dev_queue, bytes); 3564 return true; 3565 } 3566 3567 /** 3568 * netdev_sent_queue - report the number of bytes queued to hardware 3569 * @dev: network device 3570 * @bytes: number of bytes queued to the hardware device queue 3571 * 3572 * Report the number of bytes queued for sending/completion to the network 3573 * device hardware queue#0. @bytes should be a good approximation and should 3574 * exactly match netdev_completed_queue() @bytes. 3575 * This is typically called once per packet, from ndo_start_xmit(). 3576 */ 3577 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3578 { 3579 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3580 } 3581 3582 static inline bool __netdev_sent_queue(struct net_device *dev, 3583 unsigned int bytes, 3584 bool xmit_more) 3585 { 3586 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3587 xmit_more); 3588 } 3589 3590 /** 3591 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3592 * @dev_queue: network device queue 3593 * @pkts: number of packets (currently ignored) 3594 * @bytes: number of bytes dequeued from the device queue 3595 * 3596 * Must be called at most once per TX completion round (and not per 3597 * individual packet), so that BQL can adjust its limits appropriately. 3598 */ 3599 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3600 unsigned int pkts, unsigned int bytes) 3601 { 3602 #ifdef CONFIG_BQL 3603 if (unlikely(!bytes)) 3604 return; 3605 3606 dql_completed(&dev_queue->dql, bytes); 3607 3608 /* 3609 * Without the memory barrier there is a small possibility that 3610 * netdev_tx_sent_queue will miss the update and cause the queue to 3611 * be stopped forever 3612 */ 3613 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */ 3614 3615 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3616 return; 3617 3618 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3619 netif_schedule_queue(dev_queue); 3620 #endif 3621 } 3622 3623 /** 3624 * netdev_completed_queue - report bytes and packets completed by device 3625 * @dev: network device 3626 * @pkts: actual number of packets sent over the medium 3627 * @bytes: actual number of bytes sent over the medium 3628 * 3629 * Report the number of bytes and packets transmitted by the network device 3630 * hardware queue over the physical medium, @bytes must exactly match the 3631 * @bytes amount passed to netdev_sent_queue() 3632 */ 3633 static inline void netdev_completed_queue(struct net_device *dev, 3634 unsigned int pkts, unsigned int bytes) 3635 { 3636 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3637 } 3638 3639 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3640 { 3641 #ifdef CONFIG_BQL 3642 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3643 dql_reset(&q->dql); 3644 #endif 3645 } 3646 3647 /** 3648 * netdev_tx_reset_subqueue - reset the BQL stats and state of a netdev queue 3649 * @dev: network device 3650 * @qid: stack index of the queue to reset 3651 */ 3652 static inline void netdev_tx_reset_subqueue(const struct net_device *dev, 3653 u32 qid) 3654 { 3655 netdev_tx_reset_queue(netdev_get_tx_queue(dev, qid)); 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_subqueue(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 **pskb); 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 3969 static inline void napi_free_frags(struct napi_struct *napi) 3970 { 3971 kfree_skb(napi->skb); 3972 napi->skb = NULL; 3973 } 3974 3975 bool netdev_is_rx_handler_busy(struct net_device *dev); 3976 int netdev_rx_handler_register(struct net_device *dev, 3977 rx_handler_func_t *rx_handler, 3978 void *rx_handler_data); 3979 void netdev_rx_handler_unregister(struct net_device *dev); 3980 3981 bool dev_valid_name(const char *name); 3982 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 3983 { 3984 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 3985 } 3986 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 3987 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 3988 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3989 void __user *data, bool *need_copyout); 3990 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 3991 int generic_hwtstamp_get_lower(struct net_device *dev, 3992 struct kernel_hwtstamp_config *kernel_cfg); 3993 int generic_hwtstamp_set_lower(struct net_device *dev, 3994 struct kernel_hwtstamp_config *kernel_cfg, 3995 struct netlink_ext_ack *extack); 3996 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 3997 unsigned int dev_get_flags(const struct net_device *); 3998 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3999 struct netlink_ext_ack *extack); 4000 int dev_change_flags(struct net_device *dev, unsigned int flags, 4001 struct netlink_ext_ack *extack); 4002 int dev_set_alias(struct net_device *, const char *, size_t); 4003 int dev_get_alias(const struct net_device *, char *, size_t); 4004 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 4005 const char *pat, int new_ifindex); 4006 static inline 4007 int dev_change_net_namespace(struct net_device *dev, struct net *net, 4008 const char *pat) 4009 { 4010 return __dev_change_net_namespace(dev, net, pat, 0); 4011 } 4012 int __dev_set_mtu(struct net_device *, int); 4013 int dev_set_mtu(struct net_device *, int); 4014 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 4015 struct netlink_ext_ack *extack); 4016 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 4017 struct netlink_ext_ack *extack); 4018 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 4019 struct netlink_ext_ack *extack); 4020 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 4021 int dev_get_port_parent_id(struct net_device *dev, 4022 struct netdev_phys_item_id *ppid, bool recurse); 4023 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 4024 4025 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 4026 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 4027 struct netdev_queue *txq, int *ret); 4028 4029 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 4030 u8 dev_xdp_prog_count(struct net_device *dev); 4031 int dev_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf); 4032 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 4033 4034 u32 dev_get_min_mp_channel_count(const struct net_device *dev); 4035 4036 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4037 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4038 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 4039 bool is_skb_forwardable(const struct net_device *dev, 4040 const struct sk_buff *skb); 4041 4042 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 4043 const struct sk_buff *skb, 4044 const bool check_mtu) 4045 { 4046 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 4047 unsigned int len; 4048 4049 if (!(dev->flags & IFF_UP)) 4050 return false; 4051 4052 if (!check_mtu) 4053 return true; 4054 4055 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 4056 if (skb->len <= len) 4057 return true; 4058 4059 /* if TSO is enabled, we don't care about the length as the packet 4060 * could be forwarded without being segmented before 4061 */ 4062 if (skb_is_gso(skb)) 4063 return true; 4064 4065 return false; 4066 } 4067 4068 void netdev_core_stats_inc(struct net_device *dev, u32 offset); 4069 4070 #define DEV_CORE_STATS_INC(FIELD) \ 4071 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 4072 { \ 4073 netdev_core_stats_inc(dev, \ 4074 offsetof(struct net_device_core_stats, FIELD)); \ 4075 } 4076 DEV_CORE_STATS_INC(rx_dropped) 4077 DEV_CORE_STATS_INC(tx_dropped) 4078 DEV_CORE_STATS_INC(rx_nohandler) 4079 DEV_CORE_STATS_INC(rx_otherhost_dropped) 4080 #undef DEV_CORE_STATS_INC 4081 4082 static __always_inline int ____dev_forward_skb(struct net_device *dev, 4083 struct sk_buff *skb, 4084 const bool check_mtu) 4085 { 4086 if (skb_orphan_frags(skb, GFP_ATOMIC) || 4087 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 4088 dev_core_stats_rx_dropped_inc(dev); 4089 kfree_skb(skb); 4090 return NET_RX_DROP; 4091 } 4092 4093 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 4094 skb->priority = 0; 4095 return 0; 4096 } 4097 4098 bool dev_nit_active(struct net_device *dev); 4099 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 4100 4101 static inline void __dev_put(struct net_device *dev) 4102 { 4103 if (dev) { 4104 #ifdef CONFIG_PCPU_DEV_REFCNT 4105 this_cpu_dec(*dev->pcpu_refcnt); 4106 #else 4107 refcount_dec(&dev->dev_refcnt); 4108 #endif 4109 } 4110 } 4111 4112 static inline void __dev_hold(struct net_device *dev) 4113 { 4114 if (dev) { 4115 #ifdef CONFIG_PCPU_DEV_REFCNT 4116 this_cpu_inc(*dev->pcpu_refcnt); 4117 #else 4118 refcount_inc(&dev->dev_refcnt); 4119 #endif 4120 } 4121 } 4122 4123 static inline void __netdev_tracker_alloc(struct net_device *dev, 4124 netdevice_tracker *tracker, 4125 gfp_t gfp) 4126 { 4127 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4128 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 4129 #endif 4130 } 4131 4132 /* netdev_tracker_alloc() can upgrade a prior untracked reference 4133 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4134 */ 4135 static inline void netdev_tracker_alloc(struct net_device *dev, 4136 netdevice_tracker *tracker, gfp_t gfp) 4137 { 4138 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4139 refcount_dec(&dev->refcnt_tracker.no_tracker); 4140 __netdev_tracker_alloc(dev, tracker, gfp); 4141 #endif 4142 } 4143 4144 static inline void netdev_tracker_free(struct net_device *dev, 4145 netdevice_tracker *tracker) 4146 { 4147 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4148 ref_tracker_free(&dev->refcnt_tracker, tracker); 4149 #endif 4150 } 4151 4152 static inline void netdev_hold(struct net_device *dev, 4153 netdevice_tracker *tracker, gfp_t gfp) 4154 { 4155 if (dev) { 4156 __dev_hold(dev); 4157 __netdev_tracker_alloc(dev, tracker, gfp); 4158 } 4159 } 4160 4161 static inline void netdev_put(struct net_device *dev, 4162 netdevice_tracker *tracker) 4163 { 4164 if (dev) { 4165 netdev_tracker_free(dev, tracker); 4166 __dev_put(dev); 4167 } 4168 } 4169 4170 /** 4171 * dev_hold - get reference to device 4172 * @dev: network device 4173 * 4174 * Hold reference to device to keep it from being freed. 4175 * Try using netdev_hold() instead. 4176 */ 4177 static inline void dev_hold(struct net_device *dev) 4178 { 4179 netdev_hold(dev, NULL, GFP_ATOMIC); 4180 } 4181 4182 /** 4183 * dev_put - release reference to device 4184 * @dev: network device 4185 * 4186 * Release reference to device to allow it to be freed. 4187 * Try using netdev_put() instead. 4188 */ 4189 static inline void dev_put(struct net_device *dev) 4190 { 4191 netdev_put(dev, NULL); 4192 } 4193 4194 DEFINE_FREE(dev_put, struct net_device *, if (_T) dev_put(_T)) 4195 4196 static inline void netdev_ref_replace(struct net_device *odev, 4197 struct net_device *ndev, 4198 netdevice_tracker *tracker, 4199 gfp_t gfp) 4200 { 4201 if (odev) 4202 netdev_tracker_free(odev, tracker); 4203 4204 __dev_hold(ndev); 4205 __dev_put(odev); 4206 4207 if (ndev) 4208 __netdev_tracker_alloc(ndev, tracker, gfp); 4209 } 4210 4211 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4212 * and _off may be called from IRQ context, but it is caller 4213 * who is responsible for serialization of these calls. 4214 * 4215 * The name carrier is inappropriate, these functions should really be 4216 * called netif_lowerlayer_*() because they represent the state of any 4217 * kind of lower layer not just hardware media. 4218 */ 4219 void linkwatch_fire_event(struct net_device *dev); 4220 4221 /** 4222 * linkwatch_sync_dev - sync linkwatch for the given device 4223 * @dev: network device to sync linkwatch for 4224 * 4225 * Sync linkwatch for the given device, removing it from the 4226 * pending work list (if queued). 4227 */ 4228 void linkwatch_sync_dev(struct net_device *dev); 4229 4230 /** 4231 * netif_carrier_ok - test if carrier present 4232 * @dev: network device 4233 * 4234 * Check if carrier is present on device 4235 */ 4236 static inline bool netif_carrier_ok(const struct net_device *dev) 4237 { 4238 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4239 } 4240 4241 unsigned long dev_trans_start(struct net_device *dev); 4242 4243 void __netdev_watchdog_up(struct net_device *dev); 4244 4245 void netif_carrier_on(struct net_device *dev); 4246 void netif_carrier_off(struct net_device *dev); 4247 void netif_carrier_event(struct net_device *dev); 4248 4249 /** 4250 * netif_dormant_on - mark device as dormant. 4251 * @dev: network device 4252 * 4253 * Mark device as dormant (as per RFC2863). 4254 * 4255 * The dormant state indicates that the relevant interface is not 4256 * actually in a condition to pass packets (i.e., it is not 'up') but is 4257 * in a "pending" state, waiting for some external event. For "on- 4258 * demand" interfaces, this new state identifies the situation where the 4259 * interface is waiting for events to place it in the up state. 4260 */ 4261 static inline void netif_dormant_on(struct net_device *dev) 4262 { 4263 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4264 linkwatch_fire_event(dev); 4265 } 4266 4267 /** 4268 * netif_dormant_off - set device as not dormant. 4269 * @dev: network device 4270 * 4271 * Device is not in dormant state. 4272 */ 4273 static inline void netif_dormant_off(struct net_device *dev) 4274 { 4275 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4276 linkwatch_fire_event(dev); 4277 } 4278 4279 /** 4280 * netif_dormant - test if device is dormant 4281 * @dev: network device 4282 * 4283 * Check if device is dormant. 4284 */ 4285 static inline bool netif_dormant(const struct net_device *dev) 4286 { 4287 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4288 } 4289 4290 4291 /** 4292 * netif_testing_on - mark device as under test. 4293 * @dev: network device 4294 * 4295 * Mark device as under test (as per RFC2863). 4296 * 4297 * The testing state indicates that some test(s) must be performed on 4298 * the interface. After completion, of the test, the interface state 4299 * will change to up, dormant, or down, as appropriate. 4300 */ 4301 static inline void netif_testing_on(struct net_device *dev) 4302 { 4303 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4304 linkwatch_fire_event(dev); 4305 } 4306 4307 /** 4308 * netif_testing_off - set device as not under test. 4309 * @dev: network device 4310 * 4311 * Device is not in testing state. 4312 */ 4313 static inline void netif_testing_off(struct net_device *dev) 4314 { 4315 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4316 linkwatch_fire_event(dev); 4317 } 4318 4319 /** 4320 * netif_testing - test if device is under test 4321 * @dev: network device 4322 * 4323 * Check if device is under test 4324 */ 4325 static inline bool netif_testing(const struct net_device *dev) 4326 { 4327 return test_bit(__LINK_STATE_TESTING, &dev->state); 4328 } 4329 4330 4331 /** 4332 * netif_oper_up - test if device is operational 4333 * @dev: network device 4334 * 4335 * Check if carrier is operational 4336 */ 4337 static inline bool netif_oper_up(const struct net_device *dev) 4338 { 4339 unsigned int operstate = READ_ONCE(dev->operstate); 4340 4341 return operstate == IF_OPER_UP || 4342 operstate == IF_OPER_UNKNOWN /* backward compat */; 4343 } 4344 4345 /** 4346 * netif_device_present - is device available or removed 4347 * @dev: network device 4348 * 4349 * Check if device has not been removed from system. 4350 */ 4351 static inline bool netif_device_present(const struct net_device *dev) 4352 { 4353 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4354 } 4355 4356 void netif_device_detach(struct net_device *dev); 4357 4358 void netif_device_attach(struct net_device *dev); 4359 4360 /* 4361 * Network interface message level settings 4362 */ 4363 4364 enum { 4365 NETIF_MSG_DRV_BIT, 4366 NETIF_MSG_PROBE_BIT, 4367 NETIF_MSG_LINK_BIT, 4368 NETIF_MSG_TIMER_BIT, 4369 NETIF_MSG_IFDOWN_BIT, 4370 NETIF_MSG_IFUP_BIT, 4371 NETIF_MSG_RX_ERR_BIT, 4372 NETIF_MSG_TX_ERR_BIT, 4373 NETIF_MSG_TX_QUEUED_BIT, 4374 NETIF_MSG_INTR_BIT, 4375 NETIF_MSG_TX_DONE_BIT, 4376 NETIF_MSG_RX_STATUS_BIT, 4377 NETIF_MSG_PKTDATA_BIT, 4378 NETIF_MSG_HW_BIT, 4379 NETIF_MSG_WOL_BIT, 4380 4381 /* When you add a new bit above, update netif_msg_class_names array 4382 * in net/ethtool/common.c 4383 */ 4384 NETIF_MSG_CLASS_COUNT, 4385 }; 4386 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4387 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4388 4389 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4390 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4391 4392 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4393 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4394 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4395 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4396 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4397 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4398 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4399 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4400 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4401 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4402 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4403 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4404 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4405 #define NETIF_MSG_HW __NETIF_MSG(HW) 4406 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4407 4408 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4409 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4410 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4411 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4412 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4413 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4414 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4415 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4416 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4417 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4418 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4419 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4420 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4421 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4422 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4423 4424 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4425 { 4426 /* use default */ 4427 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4428 return default_msg_enable_bits; 4429 if (debug_value == 0) /* no output */ 4430 return 0; 4431 /* set low N bits */ 4432 return (1U << debug_value) - 1; 4433 } 4434 4435 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4436 { 4437 spin_lock(&txq->_xmit_lock); 4438 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4439 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4440 } 4441 4442 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4443 { 4444 __acquire(&txq->_xmit_lock); 4445 return true; 4446 } 4447 4448 static inline void __netif_tx_release(struct netdev_queue *txq) 4449 { 4450 __release(&txq->_xmit_lock); 4451 } 4452 4453 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4454 { 4455 spin_lock_bh(&txq->_xmit_lock); 4456 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4457 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4458 } 4459 4460 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4461 { 4462 bool ok = spin_trylock(&txq->_xmit_lock); 4463 4464 if (likely(ok)) { 4465 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4466 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4467 } 4468 return ok; 4469 } 4470 4471 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4472 { 4473 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4474 WRITE_ONCE(txq->xmit_lock_owner, -1); 4475 spin_unlock(&txq->_xmit_lock); 4476 } 4477 4478 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4479 { 4480 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4481 WRITE_ONCE(txq->xmit_lock_owner, -1); 4482 spin_unlock_bh(&txq->_xmit_lock); 4483 } 4484 4485 /* 4486 * txq->trans_start can be read locklessly from dev_watchdog() 4487 */ 4488 static inline void txq_trans_update(struct netdev_queue *txq) 4489 { 4490 if (txq->xmit_lock_owner != -1) 4491 WRITE_ONCE(txq->trans_start, jiffies); 4492 } 4493 4494 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4495 { 4496 unsigned long now = jiffies; 4497 4498 if (READ_ONCE(txq->trans_start) != now) 4499 WRITE_ONCE(txq->trans_start, now); 4500 } 4501 4502 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4503 static inline void netif_trans_update(struct net_device *dev) 4504 { 4505 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4506 4507 txq_trans_cond_update(txq); 4508 } 4509 4510 /** 4511 * netif_tx_lock - grab network device transmit lock 4512 * @dev: network device 4513 * 4514 * Get network device transmit lock 4515 */ 4516 void netif_tx_lock(struct net_device *dev); 4517 4518 static inline void netif_tx_lock_bh(struct net_device *dev) 4519 { 4520 local_bh_disable(); 4521 netif_tx_lock(dev); 4522 } 4523 4524 void netif_tx_unlock(struct net_device *dev); 4525 4526 static inline void netif_tx_unlock_bh(struct net_device *dev) 4527 { 4528 netif_tx_unlock(dev); 4529 local_bh_enable(); 4530 } 4531 4532 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4533 if (!(dev)->lltx) { \ 4534 __netif_tx_lock(txq, cpu); \ 4535 } else { \ 4536 __netif_tx_acquire(txq); \ 4537 } \ 4538 } 4539 4540 #define HARD_TX_TRYLOCK(dev, txq) \ 4541 (!(dev)->lltx ? \ 4542 __netif_tx_trylock(txq) : \ 4543 __netif_tx_acquire(txq)) 4544 4545 #define HARD_TX_UNLOCK(dev, txq) { \ 4546 if (!(dev)->lltx) { \ 4547 __netif_tx_unlock(txq); \ 4548 } else { \ 4549 __netif_tx_release(txq); \ 4550 } \ 4551 } 4552 4553 static inline void netif_tx_disable(struct net_device *dev) 4554 { 4555 unsigned int i; 4556 int cpu; 4557 4558 local_bh_disable(); 4559 cpu = smp_processor_id(); 4560 spin_lock(&dev->tx_global_lock); 4561 for (i = 0; i < dev->num_tx_queues; i++) { 4562 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4563 4564 __netif_tx_lock(txq, cpu); 4565 netif_tx_stop_queue(txq); 4566 __netif_tx_unlock(txq); 4567 } 4568 spin_unlock(&dev->tx_global_lock); 4569 local_bh_enable(); 4570 } 4571 4572 static inline void netif_addr_lock(struct net_device *dev) 4573 { 4574 unsigned char nest_level = 0; 4575 4576 #ifdef CONFIG_LOCKDEP 4577 nest_level = dev->nested_level; 4578 #endif 4579 spin_lock_nested(&dev->addr_list_lock, nest_level); 4580 } 4581 4582 static inline void netif_addr_lock_bh(struct net_device *dev) 4583 { 4584 unsigned char nest_level = 0; 4585 4586 #ifdef CONFIG_LOCKDEP 4587 nest_level = dev->nested_level; 4588 #endif 4589 local_bh_disable(); 4590 spin_lock_nested(&dev->addr_list_lock, nest_level); 4591 } 4592 4593 static inline void netif_addr_unlock(struct net_device *dev) 4594 { 4595 spin_unlock(&dev->addr_list_lock); 4596 } 4597 4598 static inline void netif_addr_unlock_bh(struct net_device *dev) 4599 { 4600 spin_unlock_bh(&dev->addr_list_lock); 4601 } 4602 4603 /* 4604 * dev_addrs walker. Should be used only for read access. Call with 4605 * rcu_read_lock held. 4606 */ 4607 #define for_each_dev_addr(dev, ha) \ 4608 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4609 4610 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4611 4612 void ether_setup(struct net_device *dev); 4613 4614 /* Allocate dummy net_device */ 4615 struct net_device *alloc_netdev_dummy(int sizeof_priv); 4616 4617 /* Support for loadable net-drivers */ 4618 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4619 unsigned char name_assign_type, 4620 void (*setup)(struct net_device *), 4621 unsigned int txqs, unsigned int rxqs); 4622 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4623 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4624 4625 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4626 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4627 count) 4628 4629 int register_netdev(struct net_device *dev); 4630 void unregister_netdev(struct net_device *dev); 4631 4632 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4633 4634 /* General hardware address lists handling functions */ 4635 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4636 struct netdev_hw_addr_list *from_list, int addr_len); 4637 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4638 struct netdev_hw_addr_list *from_list, int addr_len); 4639 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4640 struct net_device *dev, 4641 int (*sync)(struct net_device *, const unsigned char *), 4642 int (*unsync)(struct net_device *, 4643 const unsigned char *)); 4644 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4645 struct net_device *dev, 4646 int (*sync)(struct net_device *, 4647 const unsigned char *, int), 4648 int (*unsync)(struct net_device *, 4649 const unsigned char *, int)); 4650 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4651 struct net_device *dev, 4652 int (*unsync)(struct net_device *, 4653 const unsigned char *, int)); 4654 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4655 struct net_device *dev, 4656 int (*unsync)(struct net_device *, 4657 const unsigned char *)); 4658 void __hw_addr_init(struct netdev_hw_addr_list *list); 4659 4660 /* Functions used for device addresses handling */ 4661 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4662 const void *addr, size_t len); 4663 4664 static inline void 4665 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4666 { 4667 dev_addr_mod(dev, 0, addr, len); 4668 } 4669 4670 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4671 { 4672 __dev_addr_set(dev, addr, dev->addr_len); 4673 } 4674 4675 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4676 unsigned char addr_type); 4677 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4678 unsigned char addr_type); 4679 4680 /* Functions used for unicast addresses handling */ 4681 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4682 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4683 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4684 int dev_uc_sync(struct net_device *to, struct net_device *from); 4685 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4686 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4687 void dev_uc_flush(struct net_device *dev); 4688 void dev_uc_init(struct net_device *dev); 4689 4690 /** 4691 * __dev_uc_sync - Synchronize device's unicast list 4692 * @dev: device to sync 4693 * @sync: function to call if address should be added 4694 * @unsync: function to call if address should be removed 4695 * 4696 * Add newly added addresses to the interface, and release 4697 * addresses that have been deleted. 4698 */ 4699 static inline int __dev_uc_sync(struct net_device *dev, 4700 int (*sync)(struct net_device *, 4701 const unsigned char *), 4702 int (*unsync)(struct net_device *, 4703 const unsigned char *)) 4704 { 4705 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4706 } 4707 4708 /** 4709 * __dev_uc_unsync - Remove synchronized addresses from device 4710 * @dev: device to sync 4711 * @unsync: function to call if address should be removed 4712 * 4713 * Remove all addresses that were added to the device by dev_uc_sync(). 4714 */ 4715 static inline void __dev_uc_unsync(struct net_device *dev, 4716 int (*unsync)(struct net_device *, 4717 const unsigned char *)) 4718 { 4719 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4720 } 4721 4722 /* Functions used for multicast addresses handling */ 4723 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4724 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4725 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4726 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4727 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4728 int dev_mc_sync(struct net_device *to, struct net_device *from); 4729 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4730 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4731 void dev_mc_flush(struct net_device *dev); 4732 void dev_mc_init(struct net_device *dev); 4733 4734 /** 4735 * __dev_mc_sync - Synchronize device's multicast list 4736 * @dev: device to sync 4737 * @sync: function to call if address should be added 4738 * @unsync: function to call if address should be removed 4739 * 4740 * Add newly added addresses to the interface, and release 4741 * addresses that have been deleted. 4742 */ 4743 static inline int __dev_mc_sync(struct net_device *dev, 4744 int (*sync)(struct net_device *, 4745 const unsigned char *), 4746 int (*unsync)(struct net_device *, 4747 const unsigned char *)) 4748 { 4749 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4750 } 4751 4752 /** 4753 * __dev_mc_unsync - Remove synchronized addresses from device 4754 * @dev: device to sync 4755 * @unsync: function to call if address should be removed 4756 * 4757 * Remove all addresses that were added to the device by dev_mc_sync(). 4758 */ 4759 static inline void __dev_mc_unsync(struct net_device *dev, 4760 int (*unsync)(struct net_device *, 4761 const unsigned char *)) 4762 { 4763 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4764 } 4765 4766 /* Functions used for secondary unicast and multicast support */ 4767 void dev_set_rx_mode(struct net_device *dev); 4768 int dev_set_promiscuity(struct net_device *dev, int inc); 4769 int dev_set_allmulti(struct net_device *dev, int inc); 4770 void netdev_state_change(struct net_device *dev); 4771 void __netdev_notify_peers(struct net_device *dev); 4772 void netdev_notify_peers(struct net_device *dev); 4773 void netdev_features_change(struct net_device *dev); 4774 /* Load a device via the kmod */ 4775 void dev_load(struct net *net, const char *name); 4776 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4777 struct rtnl_link_stats64 *storage); 4778 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4779 const struct net_device_stats *netdev_stats); 4780 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4781 const struct pcpu_sw_netstats __percpu *netstats); 4782 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4783 4784 enum { 4785 NESTED_SYNC_IMM_BIT, 4786 NESTED_SYNC_TODO_BIT, 4787 }; 4788 4789 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4790 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4791 4792 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4793 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4794 4795 struct netdev_nested_priv { 4796 unsigned char flags; 4797 void *data; 4798 }; 4799 4800 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4801 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4802 struct list_head **iter); 4803 4804 /* iterate through upper list, must be called under RCU read lock */ 4805 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4806 for (iter = &(dev)->adj_list.upper, \ 4807 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4808 updev; \ 4809 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4810 4811 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4812 int (*fn)(struct net_device *upper_dev, 4813 struct netdev_nested_priv *priv), 4814 struct netdev_nested_priv *priv); 4815 4816 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4817 struct net_device *upper_dev); 4818 4819 bool netdev_has_any_upper_dev(struct net_device *dev); 4820 4821 void *netdev_lower_get_next_private(struct net_device *dev, 4822 struct list_head **iter); 4823 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4824 struct list_head **iter); 4825 4826 #define netdev_for_each_lower_private(dev, priv, iter) \ 4827 for (iter = (dev)->adj_list.lower.next, \ 4828 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4829 priv; \ 4830 priv = netdev_lower_get_next_private(dev, &(iter))) 4831 4832 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4833 for (iter = &(dev)->adj_list.lower, \ 4834 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4835 priv; \ 4836 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4837 4838 void *netdev_lower_get_next(struct net_device *dev, 4839 struct list_head **iter); 4840 4841 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4842 for (iter = (dev)->adj_list.lower.next, \ 4843 ldev = netdev_lower_get_next(dev, &(iter)); \ 4844 ldev; \ 4845 ldev = netdev_lower_get_next(dev, &(iter))) 4846 4847 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4848 struct list_head **iter); 4849 int netdev_walk_all_lower_dev(struct net_device *dev, 4850 int (*fn)(struct net_device *lower_dev, 4851 struct netdev_nested_priv *priv), 4852 struct netdev_nested_priv *priv); 4853 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4854 int (*fn)(struct net_device *lower_dev, 4855 struct netdev_nested_priv *priv), 4856 struct netdev_nested_priv *priv); 4857 4858 void *netdev_adjacent_get_private(struct list_head *adj_list); 4859 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4860 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4861 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4862 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4863 struct netlink_ext_ack *extack); 4864 int netdev_master_upper_dev_link(struct net_device *dev, 4865 struct net_device *upper_dev, 4866 void *upper_priv, void *upper_info, 4867 struct netlink_ext_ack *extack); 4868 void netdev_upper_dev_unlink(struct net_device *dev, 4869 struct net_device *upper_dev); 4870 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4871 struct net_device *new_dev, 4872 struct net_device *dev, 4873 struct netlink_ext_ack *extack); 4874 void netdev_adjacent_change_commit(struct net_device *old_dev, 4875 struct net_device *new_dev, 4876 struct net_device *dev); 4877 void netdev_adjacent_change_abort(struct net_device *old_dev, 4878 struct net_device *new_dev, 4879 struct net_device *dev); 4880 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4881 void *netdev_lower_dev_get_private(struct net_device *dev, 4882 struct net_device *lower_dev); 4883 void netdev_lower_state_changed(struct net_device *lower_dev, 4884 void *lower_state_info); 4885 4886 /* RSS keys are 40 or 52 bytes long */ 4887 #define NETDEV_RSS_KEY_LEN 52 4888 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4889 void netdev_rss_key_fill(void *buffer, size_t len); 4890 4891 int skb_checksum_help(struct sk_buff *skb); 4892 int skb_crc32c_csum_help(struct sk_buff *skb); 4893 int skb_csum_hwoffload_help(struct sk_buff *skb, 4894 const netdev_features_t features); 4895 4896 struct netdev_bonding_info { 4897 ifslave slave; 4898 ifbond master; 4899 }; 4900 4901 struct netdev_notifier_bonding_info { 4902 struct netdev_notifier_info info; /* must be first */ 4903 struct netdev_bonding_info bonding_info; 4904 }; 4905 4906 void netdev_bonding_info_change(struct net_device *dev, 4907 struct netdev_bonding_info *bonding_info); 4908 4909 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4910 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4911 #else 4912 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4913 const void *data) 4914 { 4915 } 4916 #endif 4917 4918 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4919 4920 static inline bool can_checksum_protocol(netdev_features_t features, 4921 __be16 protocol) 4922 { 4923 if (protocol == htons(ETH_P_FCOE)) 4924 return !!(features & NETIF_F_FCOE_CRC); 4925 4926 /* Assume this is an IP checksum (not SCTP CRC) */ 4927 4928 if (features & NETIF_F_HW_CSUM) { 4929 /* Can checksum everything */ 4930 return true; 4931 } 4932 4933 switch (protocol) { 4934 case htons(ETH_P_IP): 4935 return !!(features & NETIF_F_IP_CSUM); 4936 case htons(ETH_P_IPV6): 4937 return !!(features & NETIF_F_IPV6_CSUM); 4938 default: 4939 return false; 4940 } 4941 } 4942 4943 #ifdef CONFIG_BUG 4944 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4945 #else 4946 static inline void netdev_rx_csum_fault(struct net_device *dev, 4947 struct sk_buff *skb) 4948 { 4949 } 4950 #endif 4951 /* rx skb timestamps */ 4952 void net_enable_timestamp(void); 4953 void net_disable_timestamp(void); 4954 4955 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 4956 const struct skb_shared_hwtstamps *hwtstamps, 4957 bool cycles) 4958 { 4959 const struct net_device_ops *ops = dev->netdev_ops; 4960 4961 if (ops->ndo_get_tstamp) 4962 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 4963 4964 return hwtstamps->hwtstamp; 4965 } 4966 4967 #ifndef CONFIG_PREEMPT_RT 4968 static inline void netdev_xmit_set_more(bool more) 4969 { 4970 __this_cpu_write(softnet_data.xmit.more, more); 4971 } 4972 4973 static inline bool netdev_xmit_more(void) 4974 { 4975 return __this_cpu_read(softnet_data.xmit.more); 4976 } 4977 #else 4978 static inline void netdev_xmit_set_more(bool more) 4979 { 4980 current->net_xmit.more = more; 4981 } 4982 4983 static inline bool netdev_xmit_more(void) 4984 { 4985 return current->net_xmit.more; 4986 } 4987 #endif 4988 4989 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4990 struct sk_buff *skb, struct net_device *dev, 4991 bool more) 4992 { 4993 netdev_xmit_set_more(more); 4994 return ops->ndo_start_xmit(skb, dev); 4995 } 4996 4997 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4998 struct netdev_queue *txq, bool more) 4999 { 5000 const struct net_device_ops *ops = dev->netdev_ops; 5001 netdev_tx_t rc; 5002 5003 rc = __netdev_start_xmit(ops, skb, dev, more); 5004 if (rc == NETDEV_TX_OK) 5005 txq_trans_update(txq); 5006 5007 return rc; 5008 } 5009 5010 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 5011 const void *ns); 5012 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 5013 const void *ns); 5014 5015 extern const struct kobj_ns_type_operations net_ns_type_operations; 5016 5017 const char *netdev_drivername(const struct net_device *dev); 5018 5019 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 5020 netdev_features_t f2) 5021 { 5022 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 5023 if (f1 & NETIF_F_HW_CSUM) 5024 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5025 else 5026 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5027 } 5028 5029 return f1 & f2; 5030 } 5031 5032 static inline netdev_features_t netdev_get_wanted_features( 5033 struct net_device *dev) 5034 { 5035 return (dev->features & ~dev->hw_features) | dev->wanted_features; 5036 } 5037 netdev_features_t netdev_increment_features(netdev_features_t all, 5038 netdev_features_t one, netdev_features_t mask); 5039 5040 /* Allow TSO being used on stacked device : 5041 * Performing the GSO segmentation before last device 5042 * is a performance improvement. 5043 */ 5044 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 5045 netdev_features_t mask) 5046 { 5047 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 5048 } 5049 5050 int __netdev_update_features(struct net_device *dev); 5051 void netdev_update_features(struct net_device *dev); 5052 void netdev_change_features(struct net_device *dev); 5053 5054 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 5055 struct net_device *dev); 5056 5057 netdev_features_t passthru_features_check(struct sk_buff *skb, 5058 struct net_device *dev, 5059 netdev_features_t features); 5060 netdev_features_t netif_skb_features(struct sk_buff *skb); 5061 void skb_warn_bad_offload(const struct sk_buff *skb); 5062 5063 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 5064 { 5065 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 5066 5067 /* check flags correspondence */ 5068 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 5069 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 5070 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 5071 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 5072 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 5073 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 5074 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 5075 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 5076 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 5077 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 5078 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 5079 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 5080 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 5081 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 5082 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 5083 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 5084 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 5085 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 5086 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 5087 5088 return (features & feature) == feature; 5089 } 5090 5091 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 5092 { 5093 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 5094 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 5095 } 5096 5097 static inline bool netif_needs_gso(struct sk_buff *skb, 5098 netdev_features_t features) 5099 { 5100 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 5101 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 5102 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 5103 } 5104 5105 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 5106 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 5107 void netif_inherit_tso_max(struct net_device *to, 5108 const struct net_device *from); 5109 5110 static inline unsigned int 5111 netif_get_gro_max_size(const struct net_device *dev, const struct sk_buff *skb) 5112 { 5113 /* pairs with WRITE_ONCE() in netif_set_gro(_ipv4)_max_size() */ 5114 return skb->protocol == htons(ETH_P_IPV6) ? 5115 READ_ONCE(dev->gro_max_size) : 5116 READ_ONCE(dev->gro_ipv4_max_size); 5117 } 5118 5119 static inline unsigned int 5120 netif_get_gso_max_size(const struct net_device *dev, const struct sk_buff *skb) 5121 { 5122 /* pairs with WRITE_ONCE() in netif_set_gso(_ipv4)_max_size() */ 5123 return skb->protocol == htons(ETH_P_IPV6) ? 5124 READ_ONCE(dev->gso_max_size) : 5125 READ_ONCE(dev->gso_ipv4_max_size); 5126 } 5127 5128 static inline bool netif_is_macsec(const struct net_device *dev) 5129 { 5130 return dev->priv_flags & IFF_MACSEC; 5131 } 5132 5133 static inline bool netif_is_macvlan(const struct net_device *dev) 5134 { 5135 return dev->priv_flags & IFF_MACVLAN; 5136 } 5137 5138 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5139 { 5140 return dev->priv_flags & IFF_MACVLAN_PORT; 5141 } 5142 5143 static inline bool netif_is_bond_master(const struct net_device *dev) 5144 { 5145 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5146 } 5147 5148 static inline bool netif_is_bond_slave(const struct net_device *dev) 5149 { 5150 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5151 } 5152 5153 static inline bool netif_supports_nofcs(struct net_device *dev) 5154 { 5155 return dev->priv_flags & IFF_SUPP_NOFCS; 5156 } 5157 5158 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5159 { 5160 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5161 } 5162 5163 static inline bool netif_is_l3_master(const struct net_device *dev) 5164 { 5165 return dev->priv_flags & IFF_L3MDEV_MASTER; 5166 } 5167 5168 static inline bool netif_is_l3_slave(const struct net_device *dev) 5169 { 5170 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5171 } 5172 5173 static inline int dev_sdif(const struct net_device *dev) 5174 { 5175 #ifdef CONFIG_NET_L3_MASTER_DEV 5176 if (netif_is_l3_slave(dev)) 5177 return dev->ifindex; 5178 #endif 5179 return 0; 5180 } 5181 5182 static inline bool netif_is_bridge_master(const struct net_device *dev) 5183 { 5184 return dev->priv_flags & IFF_EBRIDGE; 5185 } 5186 5187 static inline bool netif_is_bridge_port(const struct net_device *dev) 5188 { 5189 return dev->priv_flags & IFF_BRIDGE_PORT; 5190 } 5191 5192 static inline bool netif_is_ovs_master(const struct net_device *dev) 5193 { 5194 return dev->priv_flags & IFF_OPENVSWITCH; 5195 } 5196 5197 static inline bool netif_is_ovs_port(const struct net_device *dev) 5198 { 5199 return dev->priv_flags & IFF_OVS_DATAPATH; 5200 } 5201 5202 static inline bool netif_is_any_bridge_master(const struct net_device *dev) 5203 { 5204 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev); 5205 } 5206 5207 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5208 { 5209 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5210 } 5211 5212 static inline bool netif_is_team_master(const struct net_device *dev) 5213 { 5214 return dev->priv_flags & IFF_TEAM; 5215 } 5216 5217 static inline bool netif_is_team_port(const struct net_device *dev) 5218 { 5219 return dev->priv_flags & IFF_TEAM_PORT; 5220 } 5221 5222 static inline bool netif_is_lag_master(const struct net_device *dev) 5223 { 5224 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5225 } 5226 5227 static inline bool netif_is_lag_port(const struct net_device *dev) 5228 { 5229 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5230 } 5231 5232 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5233 { 5234 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5235 } 5236 5237 static inline bool netif_is_failover(const struct net_device *dev) 5238 { 5239 return dev->priv_flags & IFF_FAILOVER; 5240 } 5241 5242 static inline bool netif_is_failover_slave(const struct net_device *dev) 5243 { 5244 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5245 } 5246 5247 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5248 static inline void netif_keep_dst(struct net_device *dev) 5249 { 5250 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5251 } 5252 5253 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5254 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5255 { 5256 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5257 return netif_is_macsec(dev); 5258 } 5259 5260 extern struct pernet_operations __net_initdata loopback_net_ops; 5261 5262 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5263 5264 /* netdev_printk helpers, similar to dev_printk */ 5265 5266 static inline const char *netdev_name(const struct net_device *dev) 5267 { 5268 if (!dev->name[0] || strchr(dev->name, '%')) 5269 return "(unnamed net_device)"; 5270 return dev->name; 5271 } 5272 5273 static inline const char *netdev_reg_state(const struct net_device *dev) 5274 { 5275 u8 reg_state = READ_ONCE(dev->reg_state); 5276 5277 switch (reg_state) { 5278 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5279 case NETREG_REGISTERED: return ""; 5280 case NETREG_UNREGISTERING: return " (unregistering)"; 5281 case NETREG_UNREGISTERED: return " (unregistered)"; 5282 case NETREG_RELEASED: return " (released)"; 5283 case NETREG_DUMMY: return " (dummy)"; 5284 } 5285 5286 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state); 5287 return " (unknown)"; 5288 } 5289 5290 #define MODULE_ALIAS_NETDEV(device) \ 5291 MODULE_ALIAS("netdev-" device) 5292 5293 /* 5294 * netdev_WARN() acts like dev_printk(), but with the key difference 5295 * of using a WARN/WARN_ON to get the message out, including the 5296 * file/line information and a backtrace. 5297 */ 5298 #define netdev_WARN(dev, format, args...) \ 5299 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5300 netdev_reg_state(dev), ##args) 5301 5302 #define netdev_WARN_ONCE(dev, format, args...) \ 5303 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5304 netdev_reg_state(dev), ##args) 5305 5306 /* 5307 * The list of packet types we will receive (as opposed to discard) 5308 * and the routines to invoke. 5309 * 5310 * Why 16. Because with 16 the only overlap we get on a hash of the 5311 * low nibble of the protocol value is RARP/SNAP/X.25. 5312 * 5313 * 0800 IP 5314 * 0001 802.3 5315 * 0002 AX.25 5316 * 0004 802.2 5317 * 8035 RARP 5318 * 0005 SNAP 5319 * 0805 X.25 5320 * 0806 ARP 5321 * 8137 IPX 5322 * 0009 Localtalk 5323 * 86DD IPv6 5324 */ 5325 #define PTYPE_HASH_SIZE (16) 5326 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5327 5328 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5329 5330 extern struct net_device *blackhole_netdev; 5331 5332 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */ 5333 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD) 5334 #define DEV_STATS_ADD(DEV, FIELD, VAL) \ 5335 atomic_long_add((VAL), &(DEV)->stats.__##FIELD) 5336 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD) 5337 5338 #endif /* _LINUX_NETDEVICE_H */ 5339