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, <waltje@uWalt.NL.Mugnet.ORG> 13 * Corey Minyard <wf-rch!minyard@relay.EU.net> 14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov> 15 * Alan Cox, <alan@lxorguk.ukuu.org.uk> 16 * Bjorn Ekwall. <bj0rn@blox.se> 17 * Pekka Riikonen <priikone@poseidon.pspt.fi> 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 <net/xdp.h> 44 45 #include <linux/netdev_features.h> 46 #include <linux/neighbour.h> 47 #include <uapi/linux/netdevice.h> 48 #include <uapi/linux/if_bonding.h> 49 #include <uapi/linux/pkt_cls.h> 50 #include <uapi/linux/netdev.h> 51 #include <linux/hashtable.h> 52 #include <linux/rbtree.h> 53 #include <net/net_trackers.h> 54 #include <net/net_debug.h> 55 #include <net/dropreason.h> 56 57 struct netpoll_info; 58 struct device; 59 struct ethtool_ops; 60 struct phy_device; 61 struct dsa_port; 62 struct ip_tunnel_parm; 63 struct macsec_context; 64 struct macsec_ops; 65 struct netdev_name_node; 66 struct sd_flow_limit; 67 struct sfp_bus; 68 /* 802.11 specific */ 69 struct wireless_dev; 70 /* 802.15.4 specific */ 71 struct wpan_dev; 72 struct mpls_dev; 73 /* UDP Tunnel offloads */ 74 struct udp_tunnel_info; 75 struct udp_tunnel_nic_info; 76 struct udp_tunnel_nic; 77 struct bpf_prog; 78 struct xdp_buff; 79 struct xdp_md; 80 81 void synchronize_net(void); 82 void netdev_set_default_ethtool_ops(struct net_device *dev, 83 const struct ethtool_ops *ops); 84 void netdev_sw_irq_coalesce_default_on(struct net_device *dev); 85 86 /* Backlog congestion levels */ 87 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 88 #define NET_RX_DROP 1 /* packet dropped */ 89 90 #define MAX_NEST_DEV 8 91 92 /* 93 * Transmit return codes: transmit return codes originate from three different 94 * namespaces: 95 * 96 * - qdisc return codes 97 * - driver transmit return codes 98 * - errno values 99 * 100 * Drivers are allowed to return any one of those in their hard_start_xmit() 101 * function. Real network devices commonly used with qdiscs should only return 102 * the driver transmit return codes though - when qdiscs are used, the actual 103 * transmission happens asynchronously, so the value is not propagated to 104 * higher layers. Virtual network devices transmit synchronously; in this case 105 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 106 * others are propagated to higher layers. 107 */ 108 109 /* qdisc ->enqueue() return codes. */ 110 #define NET_XMIT_SUCCESS 0x00 111 #define NET_XMIT_DROP 0x01 /* skb dropped */ 112 #define NET_XMIT_CN 0x02 /* congestion notification */ 113 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 114 115 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 116 * indicates that the device will soon be dropping packets, or already drops 117 * some packets of the same priority; prompting us to send less aggressively. */ 118 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 119 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 120 121 /* Driver transmit return codes */ 122 #define NETDEV_TX_MASK 0xf0 123 124 enum netdev_tx { 125 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 126 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 127 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 128 }; 129 typedef enum netdev_tx netdev_tx_t; 130 131 /* 132 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 133 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 134 */ 135 static inline bool dev_xmit_complete(int rc) 136 { 137 /* 138 * Positive cases with an skb consumed by a driver: 139 * - successful transmission (rc == NETDEV_TX_OK) 140 * - error while transmitting (rc < 0) 141 * - error while queueing to a different device (rc & NET_XMIT_MASK) 142 */ 143 if (likely(rc < NET_XMIT_MASK)) 144 return true; 145 146 return false; 147 } 148 149 /* 150 * Compute the worst-case header length according to the protocols 151 * used. 152 */ 153 154 #if defined(CONFIG_HYPERV_NET) 155 # define LL_MAX_HEADER 128 156 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 157 # if defined(CONFIG_MAC80211_MESH) 158 # define LL_MAX_HEADER 128 159 # else 160 # define LL_MAX_HEADER 96 161 # endif 162 #else 163 # define LL_MAX_HEADER 32 164 #endif 165 166 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 167 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 168 #define MAX_HEADER LL_MAX_HEADER 169 #else 170 #define MAX_HEADER (LL_MAX_HEADER + 48) 171 #endif 172 173 /* 174 * Old network device statistics. Fields are native words 175 * (unsigned long) so they can be read and written atomically. 176 */ 177 178 #define NET_DEV_STAT(FIELD) \ 179 union { \ 180 unsigned long FIELD; \ 181 atomic_long_t __##FIELD; \ 182 } 183 184 struct net_device_stats { 185 NET_DEV_STAT(rx_packets); 186 NET_DEV_STAT(tx_packets); 187 NET_DEV_STAT(rx_bytes); 188 NET_DEV_STAT(tx_bytes); 189 NET_DEV_STAT(rx_errors); 190 NET_DEV_STAT(tx_errors); 191 NET_DEV_STAT(rx_dropped); 192 NET_DEV_STAT(tx_dropped); 193 NET_DEV_STAT(multicast); 194 NET_DEV_STAT(collisions); 195 NET_DEV_STAT(rx_length_errors); 196 NET_DEV_STAT(rx_over_errors); 197 NET_DEV_STAT(rx_crc_errors); 198 NET_DEV_STAT(rx_frame_errors); 199 NET_DEV_STAT(rx_fifo_errors); 200 NET_DEV_STAT(rx_missed_errors); 201 NET_DEV_STAT(tx_aborted_errors); 202 NET_DEV_STAT(tx_carrier_errors); 203 NET_DEV_STAT(tx_fifo_errors); 204 NET_DEV_STAT(tx_heartbeat_errors); 205 NET_DEV_STAT(tx_window_errors); 206 NET_DEV_STAT(rx_compressed); 207 NET_DEV_STAT(tx_compressed); 208 }; 209 #undef NET_DEV_STAT 210 211 /* per-cpu stats, allocated on demand. 212 * Try to fit them in a single cache line, for dev_get_stats() sake. 213 */ 214 struct net_device_core_stats { 215 unsigned long rx_dropped; 216 unsigned long tx_dropped; 217 unsigned long rx_nohandler; 218 unsigned long rx_otherhost_dropped; 219 } __aligned(4 * sizeof(unsigned long)); 220 221 #include <linux/cache.h> 222 #include <linux/skbuff.h> 223 224 #ifdef CONFIG_RPS 225 #include <linux/static_key.h> 226 extern struct static_key_false rps_needed; 227 extern struct static_key_false rfs_needed; 228 #endif 229 230 struct neighbour; 231 struct neigh_parms; 232 struct sk_buff; 233 234 struct netdev_hw_addr { 235 struct list_head list; 236 struct rb_node node; 237 unsigned char addr[MAX_ADDR_LEN]; 238 unsigned char type; 239 #define NETDEV_HW_ADDR_T_LAN 1 240 #define NETDEV_HW_ADDR_T_SAN 2 241 #define NETDEV_HW_ADDR_T_UNICAST 3 242 #define NETDEV_HW_ADDR_T_MULTICAST 4 243 bool global_use; 244 int sync_cnt; 245 int refcount; 246 int synced; 247 struct rcu_head rcu_head; 248 }; 249 250 struct netdev_hw_addr_list { 251 struct list_head list; 252 int count; 253 254 /* Auxiliary tree for faster lookup on addition and deletion */ 255 struct rb_root tree; 256 }; 257 258 #define netdev_hw_addr_list_count(l) ((l)->count) 259 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 260 #define netdev_hw_addr_list_for_each(ha, l) \ 261 list_for_each_entry(ha, &(l)->list, list) 262 263 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 264 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 265 #define netdev_for_each_uc_addr(ha, dev) \ 266 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 267 #define netdev_for_each_synced_uc_addr(_ha, _dev) \ 268 netdev_for_each_uc_addr((_ha), (_dev)) \ 269 if ((_ha)->sync_cnt) 270 271 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 272 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 273 #define netdev_for_each_mc_addr(ha, dev) \ 274 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 275 #define netdev_for_each_synced_mc_addr(_ha, _dev) \ 276 netdev_for_each_mc_addr((_ha), (_dev)) \ 277 if ((_ha)->sync_cnt) 278 279 struct hh_cache { 280 unsigned int hh_len; 281 seqlock_t hh_lock; 282 283 /* cached hardware header; allow for machine alignment needs. */ 284 #define HH_DATA_MOD 16 285 #define HH_DATA_OFF(__len) \ 286 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 287 #define HH_DATA_ALIGN(__len) \ 288 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 289 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 290 }; 291 292 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 293 * Alternative is: 294 * dev->hard_header_len ? (dev->hard_header_len + 295 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 296 * 297 * We could use other alignment values, but we must maintain the 298 * relationship HH alignment <= LL alignment. 299 */ 300 #define LL_RESERVED_SPACE(dev) \ 301 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \ 302 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 303 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 304 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \ 305 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 306 307 struct header_ops { 308 int (*create) (struct sk_buff *skb, struct net_device *dev, 309 unsigned short type, const void *daddr, 310 const void *saddr, unsigned int len); 311 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 312 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 313 void (*cache_update)(struct hh_cache *hh, 314 const struct net_device *dev, 315 const unsigned char *haddr); 316 bool (*validate)(const char *ll_header, unsigned int len); 317 __be16 (*parse_protocol)(const struct sk_buff *skb); 318 }; 319 320 /* These flag bits are private to the generic network queueing 321 * layer; they may not be explicitly referenced by any other 322 * code. 323 */ 324 325 enum netdev_state_t { 326 __LINK_STATE_START, 327 __LINK_STATE_PRESENT, 328 __LINK_STATE_NOCARRIER, 329 __LINK_STATE_LINKWATCH_PENDING, 330 __LINK_STATE_DORMANT, 331 __LINK_STATE_TESTING, 332 }; 333 334 struct gro_list { 335 struct list_head list; 336 int count; 337 }; 338 339 /* 340 * size of gro hash buckets, must less than bit number of 341 * napi_struct::gro_bitmask 342 */ 343 #define GRO_HASH_BUCKETS 8 344 345 /* 346 * Structure for NAPI scheduling similar to tasklet but with weighting 347 */ 348 struct napi_struct { 349 /* The poll_list must only be managed by the entity which 350 * changes the state of the NAPI_STATE_SCHED bit. This means 351 * whoever atomically sets that bit can add this napi_struct 352 * to the per-CPU poll_list, and whoever clears that bit 353 * can remove from the list right before clearing the bit. 354 */ 355 struct list_head poll_list; 356 357 unsigned long state; 358 int weight; 359 int defer_hard_irqs_count; 360 unsigned long gro_bitmask; 361 int (*poll)(struct napi_struct *, int); 362 #ifdef CONFIG_NETPOLL 363 int poll_owner; 364 #endif 365 struct net_device *dev; 366 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 367 struct sk_buff *skb; 368 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 369 int rx_count; /* length of rx_list */ 370 struct hrtimer timer; 371 struct list_head dev_list; 372 struct hlist_node napi_hash_node; 373 unsigned int napi_id; 374 struct task_struct *thread; 375 }; 376 377 enum { 378 NAPI_STATE_SCHED, /* Poll is scheduled */ 379 NAPI_STATE_MISSED, /* reschedule a napi */ 380 NAPI_STATE_DISABLE, /* Disable pending */ 381 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 382 NAPI_STATE_LISTED, /* NAPI added to system lists */ 383 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 384 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 385 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 386 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 387 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 388 }; 389 390 enum { 391 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 392 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 393 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 394 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 395 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 396 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 397 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 398 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 399 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 400 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 401 }; 402 403 enum gro_result { 404 GRO_MERGED, 405 GRO_MERGED_FREE, 406 GRO_HELD, 407 GRO_NORMAL, 408 GRO_CONSUMED, 409 }; 410 typedef enum gro_result gro_result_t; 411 412 /* 413 * enum rx_handler_result - Possible return values for rx_handlers. 414 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 415 * further. 416 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 417 * case skb->dev was changed by rx_handler. 418 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 419 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 420 * 421 * rx_handlers are functions called from inside __netif_receive_skb(), to do 422 * special processing of the skb, prior to delivery to protocol handlers. 423 * 424 * Currently, a net_device can only have a single rx_handler registered. Trying 425 * to register a second rx_handler will return -EBUSY. 426 * 427 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 428 * To unregister a rx_handler on a net_device, use 429 * netdev_rx_handler_unregister(). 430 * 431 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 432 * do with the skb. 433 * 434 * If the rx_handler consumed the skb in some way, it should return 435 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 436 * the skb to be delivered in some other way. 437 * 438 * If the rx_handler changed skb->dev, to divert the skb to another 439 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 440 * new device will be called if it exists. 441 * 442 * If the rx_handler decides the skb should be ignored, it should return 443 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 444 * are registered on exact device (ptype->dev == skb->dev). 445 * 446 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 447 * delivered, it should return RX_HANDLER_PASS. 448 * 449 * A device without a registered rx_handler will behave as if rx_handler 450 * returned RX_HANDLER_PASS. 451 */ 452 453 enum rx_handler_result { 454 RX_HANDLER_CONSUMED, 455 RX_HANDLER_ANOTHER, 456 RX_HANDLER_EXACT, 457 RX_HANDLER_PASS, 458 }; 459 typedef enum rx_handler_result rx_handler_result_t; 460 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 461 462 void __napi_schedule(struct napi_struct *n); 463 void __napi_schedule_irqoff(struct napi_struct *n); 464 465 static inline bool napi_disable_pending(struct napi_struct *n) 466 { 467 return test_bit(NAPI_STATE_DISABLE, &n->state); 468 } 469 470 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 471 { 472 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 473 } 474 475 bool napi_schedule_prep(struct napi_struct *n); 476 477 /** 478 * napi_schedule - schedule NAPI poll 479 * @n: NAPI context 480 * 481 * Schedule NAPI poll routine to be called if it is not already 482 * running. 483 */ 484 static inline void napi_schedule(struct napi_struct *n) 485 { 486 if (napi_schedule_prep(n)) 487 __napi_schedule(n); 488 } 489 490 /** 491 * napi_schedule_irqoff - schedule NAPI poll 492 * @n: NAPI context 493 * 494 * Variant of napi_schedule(), assuming hard irqs are masked. 495 */ 496 static inline void napi_schedule_irqoff(struct napi_struct *n) 497 { 498 if (napi_schedule_prep(n)) 499 __napi_schedule_irqoff(n); 500 } 501 502 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 503 static inline bool napi_reschedule(struct napi_struct *napi) 504 { 505 if (napi_schedule_prep(napi)) { 506 __napi_schedule(napi); 507 return true; 508 } 509 return false; 510 } 511 512 /** 513 * napi_complete_done - NAPI processing complete 514 * @n: NAPI context 515 * @work_done: number of packets processed 516 * 517 * Mark NAPI processing as complete. Should only be called if poll budget 518 * has not been completely consumed. 519 * Prefer over napi_complete(). 520 * Return false if device should avoid rearming interrupts. 521 */ 522 bool napi_complete_done(struct napi_struct *n, int work_done); 523 524 static inline bool napi_complete(struct napi_struct *n) 525 { 526 return napi_complete_done(n, 0); 527 } 528 529 int dev_set_threaded(struct net_device *dev, bool threaded); 530 531 /** 532 * napi_disable - prevent NAPI from scheduling 533 * @n: NAPI context 534 * 535 * Stop NAPI from being scheduled on this context. 536 * Waits till any outstanding processing completes. 537 */ 538 void napi_disable(struct napi_struct *n); 539 540 void napi_enable(struct napi_struct *n); 541 542 /** 543 * napi_synchronize - wait until NAPI is not running 544 * @n: NAPI context 545 * 546 * Wait until NAPI is done being scheduled on this context. 547 * Waits till any outstanding processing completes but 548 * does not disable future activations. 549 */ 550 static inline void napi_synchronize(const struct napi_struct *n) 551 { 552 if (IS_ENABLED(CONFIG_SMP)) 553 while (test_bit(NAPI_STATE_SCHED, &n->state)) 554 msleep(1); 555 else 556 barrier(); 557 } 558 559 /** 560 * napi_if_scheduled_mark_missed - if napi is running, set the 561 * NAPIF_STATE_MISSED 562 * @n: NAPI context 563 * 564 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 565 * NAPI is scheduled. 566 **/ 567 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 568 { 569 unsigned long val, new; 570 571 val = READ_ONCE(n->state); 572 do { 573 if (val & NAPIF_STATE_DISABLE) 574 return true; 575 576 if (!(val & NAPIF_STATE_SCHED)) 577 return false; 578 579 new = val | NAPIF_STATE_MISSED; 580 } while (!try_cmpxchg(&n->state, &val, new)); 581 582 return true; 583 } 584 585 enum netdev_queue_state_t { 586 __QUEUE_STATE_DRV_XOFF, 587 __QUEUE_STATE_STACK_XOFF, 588 __QUEUE_STATE_FROZEN, 589 }; 590 591 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 592 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 593 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 594 595 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 596 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 597 QUEUE_STATE_FROZEN) 598 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 599 QUEUE_STATE_FROZEN) 600 601 /* 602 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 603 * netif_tx_* functions below are used to manipulate this flag. The 604 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 605 * queue independently. The netif_xmit_*stopped functions below are called 606 * to check if the queue has been stopped by the driver or stack (either 607 * of the XOFF bits are set in the state). Drivers should not need to call 608 * netif_xmit*stopped functions, they should only be using netif_tx_*. 609 */ 610 611 struct netdev_queue { 612 /* 613 * read-mostly part 614 */ 615 struct net_device *dev; 616 netdevice_tracker dev_tracker; 617 618 struct Qdisc __rcu *qdisc; 619 struct Qdisc *qdisc_sleeping; 620 #ifdef CONFIG_SYSFS 621 struct kobject kobj; 622 #endif 623 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 624 int numa_node; 625 #endif 626 unsigned long tx_maxrate; 627 /* 628 * Number of TX timeouts for this queue 629 * (/sys/class/net/DEV/Q/trans_timeout) 630 */ 631 atomic_long_t trans_timeout; 632 633 /* Subordinate device that the queue has been assigned to */ 634 struct net_device *sb_dev; 635 #ifdef CONFIG_XDP_SOCKETS 636 struct xsk_buff_pool *pool; 637 #endif 638 /* 639 * write-mostly part 640 */ 641 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 642 int xmit_lock_owner; 643 /* 644 * Time (in jiffies) of last Tx 645 */ 646 unsigned long trans_start; 647 648 unsigned long state; 649 650 #ifdef CONFIG_BQL 651 struct dql dql; 652 #endif 653 } ____cacheline_aligned_in_smp; 654 655 extern int sysctl_fb_tunnels_only_for_init_net; 656 extern int sysctl_devconf_inherit_init_net; 657 658 /* 659 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 660 * == 1 : For initns only 661 * == 2 : For none. 662 */ 663 static inline bool net_has_fallback_tunnels(const struct net *net) 664 { 665 #if IS_ENABLED(CONFIG_SYSCTL) 666 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net); 667 668 return !fb_tunnels_only_for_init_net || 669 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1); 670 #else 671 return true; 672 #endif 673 } 674 675 static inline int net_inherit_devconf(void) 676 { 677 #if IS_ENABLED(CONFIG_SYSCTL) 678 return READ_ONCE(sysctl_devconf_inherit_init_net); 679 #else 680 return 0; 681 #endif 682 } 683 684 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 685 { 686 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 687 return q->numa_node; 688 #else 689 return NUMA_NO_NODE; 690 #endif 691 } 692 693 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 694 { 695 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 696 q->numa_node = node; 697 #endif 698 } 699 700 #ifdef CONFIG_RPS 701 /* 702 * This structure holds an RPS map which can be of variable length. The 703 * map is an array of CPUs. 704 */ 705 struct rps_map { 706 unsigned int len; 707 struct rcu_head rcu; 708 u16 cpus[]; 709 }; 710 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 711 712 /* 713 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 714 * tail pointer for that CPU's input queue at the time of last enqueue, and 715 * a hardware filter index. 716 */ 717 struct rps_dev_flow { 718 u16 cpu; 719 u16 filter; 720 unsigned int last_qtail; 721 }; 722 #define RPS_NO_FILTER 0xffff 723 724 /* 725 * The rps_dev_flow_table structure contains a table of flow mappings. 726 */ 727 struct rps_dev_flow_table { 728 unsigned int mask; 729 struct rcu_head rcu; 730 struct rps_dev_flow flows[]; 731 }; 732 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 733 ((_num) * sizeof(struct rps_dev_flow))) 734 735 /* 736 * The rps_sock_flow_table contains mappings of flows to the last CPU 737 * on which they were processed by the application (set in recvmsg). 738 * Each entry is a 32bit value. Upper part is the high-order bits 739 * of flow hash, lower part is CPU number. 740 * rps_cpu_mask is used to partition the space, depending on number of 741 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 742 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 743 * meaning we use 32-6=26 bits for the hash. 744 */ 745 struct rps_sock_flow_table { 746 u32 mask; 747 748 u32 ents[] ____cacheline_aligned_in_smp; 749 }; 750 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 751 752 #define RPS_NO_CPU 0xffff 753 754 extern u32 rps_cpu_mask; 755 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 756 757 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 758 u32 hash) 759 { 760 if (table && hash) { 761 unsigned int index = hash & table->mask; 762 u32 val = hash & ~rps_cpu_mask; 763 764 /* We only give a hint, preemption can change CPU under us */ 765 val |= raw_smp_processor_id(); 766 767 if (table->ents[index] != val) 768 table->ents[index] = val; 769 } 770 } 771 772 #ifdef CONFIG_RFS_ACCEL 773 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 774 u16 filter_id); 775 #endif 776 #endif /* CONFIG_RPS */ 777 778 /* This structure contains an instance of an RX queue. */ 779 struct netdev_rx_queue { 780 struct xdp_rxq_info xdp_rxq; 781 #ifdef CONFIG_RPS 782 struct rps_map __rcu *rps_map; 783 struct rps_dev_flow_table __rcu *rps_flow_table; 784 #endif 785 struct kobject kobj; 786 struct net_device *dev; 787 netdevice_tracker dev_tracker; 788 789 #ifdef CONFIG_XDP_SOCKETS 790 struct xsk_buff_pool *pool; 791 #endif 792 } ____cacheline_aligned_in_smp; 793 794 /* 795 * RX queue sysfs structures and functions. 796 */ 797 struct rx_queue_attribute { 798 struct attribute attr; 799 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 800 ssize_t (*store)(struct netdev_rx_queue *queue, 801 const char *buf, size_t len); 802 }; 803 804 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 805 enum xps_map_type { 806 XPS_CPUS = 0, 807 XPS_RXQS, 808 XPS_MAPS_MAX, 809 }; 810 811 #ifdef CONFIG_XPS 812 /* 813 * This structure holds an XPS map which can be of variable length. The 814 * map is an array of queues. 815 */ 816 struct xps_map { 817 unsigned int len; 818 unsigned int alloc_len; 819 struct rcu_head rcu; 820 u16 queues[]; 821 }; 822 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 823 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 824 - sizeof(struct xps_map)) / sizeof(u16)) 825 826 /* 827 * This structure holds all XPS maps for device. Maps are indexed by CPU. 828 * 829 * We keep track of the number of cpus/rxqs used when the struct is allocated, 830 * in nr_ids. This will help not accessing out-of-bound memory. 831 * 832 * We keep track of the number of traffic classes used when the struct is 833 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 834 * not crossing its upper bound, as the original dev->num_tc can be updated in 835 * the meantime. 836 */ 837 struct xps_dev_maps { 838 struct rcu_head rcu; 839 unsigned int nr_ids; 840 s16 num_tc; 841 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 842 }; 843 844 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 845 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 846 847 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 848 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 849 850 #endif /* CONFIG_XPS */ 851 852 #define TC_MAX_QUEUE 16 853 #define TC_BITMASK 15 854 /* HW offloaded queuing disciplines txq count and offset maps */ 855 struct netdev_tc_txq { 856 u16 count; 857 u16 offset; 858 }; 859 860 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 861 /* 862 * This structure is to hold information about the device 863 * configured to run FCoE protocol stack. 864 */ 865 struct netdev_fcoe_hbainfo { 866 char manufacturer[64]; 867 char serial_number[64]; 868 char hardware_version[64]; 869 char driver_version[64]; 870 char optionrom_version[64]; 871 char firmware_version[64]; 872 char model[256]; 873 char model_description[256]; 874 }; 875 #endif 876 877 #define MAX_PHYS_ITEM_ID_LEN 32 878 879 /* This structure holds a unique identifier to identify some 880 * physical item (port for example) used by a netdevice. 881 */ 882 struct netdev_phys_item_id { 883 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 884 unsigned char id_len; 885 }; 886 887 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 888 struct netdev_phys_item_id *b) 889 { 890 return a->id_len == b->id_len && 891 memcmp(a->id, b->id, a->id_len) == 0; 892 } 893 894 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 895 struct sk_buff *skb, 896 struct net_device *sb_dev); 897 898 enum net_device_path_type { 899 DEV_PATH_ETHERNET = 0, 900 DEV_PATH_VLAN, 901 DEV_PATH_BRIDGE, 902 DEV_PATH_PPPOE, 903 DEV_PATH_DSA, 904 DEV_PATH_MTK_WDMA, 905 }; 906 907 struct net_device_path { 908 enum net_device_path_type type; 909 const struct net_device *dev; 910 union { 911 struct { 912 u16 id; 913 __be16 proto; 914 u8 h_dest[ETH_ALEN]; 915 } encap; 916 struct { 917 enum { 918 DEV_PATH_BR_VLAN_KEEP, 919 DEV_PATH_BR_VLAN_TAG, 920 DEV_PATH_BR_VLAN_UNTAG, 921 DEV_PATH_BR_VLAN_UNTAG_HW, 922 } vlan_mode; 923 u16 vlan_id; 924 __be16 vlan_proto; 925 } bridge; 926 struct { 927 int port; 928 u16 proto; 929 } dsa; 930 struct { 931 u8 wdma_idx; 932 u8 queue; 933 u16 wcid; 934 u8 bss; 935 } mtk_wdma; 936 }; 937 }; 938 939 #define NET_DEVICE_PATH_STACK_MAX 5 940 #define NET_DEVICE_PATH_VLAN_MAX 2 941 942 struct net_device_path_stack { 943 int num_paths; 944 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 945 }; 946 947 struct net_device_path_ctx { 948 const struct net_device *dev; 949 u8 daddr[ETH_ALEN]; 950 951 int num_vlans; 952 struct { 953 u16 id; 954 __be16 proto; 955 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 956 }; 957 958 enum tc_setup_type { 959 TC_QUERY_CAPS, 960 TC_SETUP_QDISC_MQPRIO, 961 TC_SETUP_CLSU32, 962 TC_SETUP_CLSFLOWER, 963 TC_SETUP_CLSMATCHALL, 964 TC_SETUP_CLSBPF, 965 TC_SETUP_BLOCK, 966 TC_SETUP_QDISC_CBS, 967 TC_SETUP_QDISC_RED, 968 TC_SETUP_QDISC_PRIO, 969 TC_SETUP_QDISC_MQ, 970 TC_SETUP_QDISC_ETF, 971 TC_SETUP_ROOT_QDISC, 972 TC_SETUP_QDISC_GRED, 973 TC_SETUP_QDISC_TAPRIO, 974 TC_SETUP_FT, 975 TC_SETUP_QDISC_ETS, 976 TC_SETUP_QDISC_TBF, 977 TC_SETUP_QDISC_FIFO, 978 TC_SETUP_QDISC_HTB, 979 TC_SETUP_ACT, 980 }; 981 982 /* These structures hold the attributes of bpf state that are being passed 983 * to the netdevice through the bpf op. 984 */ 985 enum bpf_netdev_command { 986 /* Set or clear a bpf program used in the earliest stages of packet 987 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 988 * is responsible for calling bpf_prog_put on any old progs that are 989 * stored. In case of error, the callee need not release the new prog 990 * reference, but on success it takes ownership and must bpf_prog_put 991 * when it is no longer used. 992 */ 993 XDP_SETUP_PROG, 994 XDP_SETUP_PROG_HW, 995 /* BPF program for offload callbacks, invoked at program load time. */ 996 BPF_OFFLOAD_MAP_ALLOC, 997 BPF_OFFLOAD_MAP_FREE, 998 XDP_SETUP_XSK_POOL, 999 }; 1000 1001 struct bpf_prog_offload_ops; 1002 struct netlink_ext_ack; 1003 struct xdp_umem; 1004 struct xdp_dev_bulk_queue; 1005 struct bpf_xdp_link; 1006 1007 enum bpf_xdp_mode { 1008 XDP_MODE_SKB = 0, 1009 XDP_MODE_DRV = 1, 1010 XDP_MODE_HW = 2, 1011 __MAX_XDP_MODE 1012 }; 1013 1014 struct bpf_xdp_entity { 1015 struct bpf_prog *prog; 1016 struct bpf_xdp_link *link; 1017 }; 1018 1019 struct netdev_bpf { 1020 enum bpf_netdev_command command; 1021 union { 1022 /* XDP_SETUP_PROG */ 1023 struct { 1024 u32 flags; 1025 struct bpf_prog *prog; 1026 struct netlink_ext_ack *extack; 1027 }; 1028 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 1029 struct { 1030 struct bpf_offloaded_map *offmap; 1031 }; 1032 /* XDP_SETUP_XSK_POOL */ 1033 struct { 1034 struct xsk_buff_pool *pool; 1035 u16 queue_id; 1036 } xsk; 1037 }; 1038 }; 1039 1040 /* Flags for ndo_xsk_wakeup. */ 1041 #define XDP_WAKEUP_RX (1 << 0) 1042 #define XDP_WAKEUP_TX (1 << 1) 1043 1044 #ifdef CONFIG_XFRM_OFFLOAD 1045 struct xfrmdev_ops { 1046 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack); 1047 void (*xdo_dev_state_delete) (struct xfrm_state *x); 1048 void (*xdo_dev_state_free) (struct xfrm_state *x); 1049 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 1050 struct xfrm_state *x); 1051 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1052 void (*xdo_dev_state_update_curlft) (struct xfrm_state *x); 1053 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack); 1054 void (*xdo_dev_policy_delete) (struct xfrm_policy *x); 1055 void (*xdo_dev_policy_free) (struct xfrm_policy *x); 1056 }; 1057 #endif 1058 1059 struct dev_ifalias { 1060 struct rcu_head rcuhead; 1061 char ifalias[]; 1062 }; 1063 1064 struct devlink; 1065 struct tlsdev_ops; 1066 1067 struct netdev_net_notifier { 1068 struct list_head list; 1069 struct notifier_block *nb; 1070 }; 1071 1072 /* 1073 * This structure defines the management hooks for network devices. 1074 * The following hooks can be defined; unless noted otherwise, they are 1075 * optional and can be filled with a null pointer. 1076 * 1077 * int (*ndo_init)(struct net_device *dev); 1078 * This function is called once when a network device is registered. 1079 * The network device can use this for any late stage initialization 1080 * or semantic validation. It can fail with an error code which will 1081 * be propagated back to register_netdev. 1082 * 1083 * void (*ndo_uninit)(struct net_device *dev); 1084 * This function is called when device is unregistered or when registration 1085 * fails. It is not called if init fails. 1086 * 1087 * int (*ndo_open)(struct net_device *dev); 1088 * This function is called when a network device transitions to the up 1089 * state. 1090 * 1091 * int (*ndo_stop)(struct net_device *dev); 1092 * This function is called when a network device transitions to the down 1093 * state. 1094 * 1095 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1096 * struct net_device *dev); 1097 * Called when a packet needs to be transmitted. 1098 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1099 * the queue before that can happen; it's for obsolete devices and weird 1100 * corner cases, but the stack really does a non-trivial amount 1101 * of useless work if you return NETDEV_TX_BUSY. 1102 * Required; cannot be NULL. 1103 * 1104 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1105 * struct net_device *dev 1106 * netdev_features_t features); 1107 * Called by core transmit path to determine if device is capable of 1108 * performing offload operations on a given packet. This is to give 1109 * the device an opportunity to implement any restrictions that cannot 1110 * be otherwise expressed by feature flags. The check is called with 1111 * the set of features that the stack has calculated and it returns 1112 * those the driver believes to be appropriate. 1113 * 1114 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1115 * struct net_device *sb_dev); 1116 * Called to decide which queue to use when device supports multiple 1117 * transmit queues. 1118 * 1119 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1120 * This function is called to allow device receiver to make 1121 * changes to configuration when multicast or promiscuous is enabled. 1122 * 1123 * void (*ndo_set_rx_mode)(struct net_device *dev); 1124 * This function is called device changes address list filtering. 1125 * If driver handles unicast address filtering, it should set 1126 * IFF_UNICAST_FLT in its priv_flags. 1127 * 1128 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1129 * This function is called when the Media Access Control address 1130 * needs to be changed. If this interface is not defined, the 1131 * MAC address can not be changed. 1132 * 1133 * int (*ndo_validate_addr)(struct net_device *dev); 1134 * Test if Media Access Control address is valid for the device. 1135 * 1136 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1137 * Old-style ioctl entry point. This is used internally by the 1138 * appletalk and ieee802154 subsystems but is no longer called by 1139 * the device ioctl handler. 1140 * 1141 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1142 * Used by the bonding driver for its device specific ioctls: 1143 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1144 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1145 * 1146 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1147 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1148 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1149 * 1150 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1151 * Used to set network devices bus interface parameters. This interface 1152 * is retained for legacy reasons; new devices should use the bus 1153 * interface (PCI) for low level management. 1154 * 1155 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1156 * Called when a user wants to change the Maximum Transfer Unit 1157 * of a device. 1158 * 1159 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1160 * Callback used when the transmitter has not made any progress 1161 * for dev->watchdog ticks. 1162 * 1163 * void (*ndo_get_stats64)(struct net_device *dev, 1164 * struct rtnl_link_stats64 *storage); 1165 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1166 * Called when a user wants to get the network device usage 1167 * statistics. Drivers must do one of the following: 1168 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1169 * rtnl_link_stats64 structure passed by the caller. 1170 * 2. Define @ndo_get_stats to update a net_device_stats structure 1171 * (which should normally be dev->stats) and return a pointer to 1172 * it. The structure may be changed asynchronously only if each 1173 * field is written atomically. 1174 * 3. Update dev->stats asynchronously and atomically, and define 1175 * neither operation. 1176 * 1177 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1178 * Return true if this device supports offload stats of this attr_id. 1179 * 1180 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1181 * void *attr_data) 1182 * Get statistics for offload operations by attr_id. Write it into the 1183 * attr_data pointer. 1184 * 1185 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1186 * If device supports VLAN filtering this function is called when a 1187 * VLAN id is registered. 1188 * 1189 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1190 * If device supports VLAN filtering this function is called when a 1191 * VLAN id is unregistered. 1192 * 1193 * void (*ndo_poll_controller)(struct net_device *dev); 1194 * 1195 * SR-IOV management functions. 1196 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1197 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1198 * u8 qos, __be16 proto); 1199 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1200 * int max_tx_rate); 1201 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1202 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1203 * int (*ndo_get_vf_config)(struct net_device *dev, 1204 * int vf, struct ifla_vf_info *ivf); 1205 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1206 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1207 * struct nlattr *port[]); 1208 * 1209 * Enable or disable the VF ability to query its RSS Redirection Table and 1210 * Hash Key. This is needed since on some devices VF share this information 1211 * with PF and querying it may introduce a theoretical security risk. 1212 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1213 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1214 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1215 * void *type_data); 1216 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1217 * This is always called from the stack with the rtnl lock held and netif 1218 * tx queues stopped. This allows the netdevice to perform queue 1219 * management safely. 1220 * 1221 * Fiber Channel over Ethernet (FCoE) offload functions. 1222 * int (*ndo_fcoe_enable)(struct net_device *dev); 1223 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1224 * so the underlying device can perform whatever needed configuration or 1225 * initialization to support acceleration of FCoE traffic. 1226 * 1227 * int (*ndo_fcoe_disable)(struct net_device *dev); 1228 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1229 * so the underlying device can perform whatever needed clean-ups to 1230 * stop supporting acceleration of FCoE traffic. 1231 * 1232 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1233 * struct scatterlist *sgl, unsigned int sgc); 1234 * Called when the FCoE Initiator wants to initialize an I/O that 1235 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1236 * perform necessary setup and returns 1 to indicate the device is set up 1237 * successfully to perform DDP on this I/O, otherwise this returns 0. 1238 * 1239 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1240 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1241 * indicated by the FC exchange id 'xid', so the underlying device can 1242 * clean up and reuse resources for later DDP requests. 1243 * 1244 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1245 * struct scatterlist *sgl, unsigned int sgc); 1246 * Called when the FCoE Target wants to initialize an I/O that 1247 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1248 * perform necessary setup and returns 1 to indicate the device is set up 1249 * successfully to perform DDP on this I/O, otherwise this returns 0. 1250 * 1251 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1252 * struct netdev_fcoe_hbainfo *hbainfo); 1253 * Called when the FCoE Protocol stack wants information on the underlying 1254 * device. This information is utilized by the FCoE protocol stack to 1255 * register attributes with Fiber Channel management service as per the 1256 * FC-GS Fabric Device Management Information(FDMI) specification. 1257 * 1258 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1259 * Called when the underlying device wants to override default World Wide 1260 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1261 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1262 * protocol stack to use. 1263 * 1264 * RFS acceleration. 1265 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1266 * u16 rxq_index, u32 flow_id); 1267 * Set hardware filter for RFS. rxq_index is the target queue index; 1268 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1269 * Return the filter ID on success, or a negative error code. 1270 * 1271 * Slave management functions (for bridge, bonding, etc). 1272 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1273 * Called to make another netdev an underling. 1274 * 1275 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1276 * Called to release previously enslaved netdev. 1277 * 1278 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1279 * struct sk_buff *skb, 1280 * bool all_slaves); 1281 * Get the xmit slave of master device. If all_slaves is true, function 1282 * assume all the slaves can transmit. 1283 * 1284 * Feature/offload setting functions. 1285 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1286 * netdev_features_t features); 1287 * Adjusts the requested feature flags according to device-specific 1288 * constraints, and returns the resulting flags. Must not modify 1289 * the device state. 1290 * 1291 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1292 * Called to update device configuration to new features. Passed 1293 * feature set might be less than what was returned by ndo_fix_features()). 1294 * Must return >0 or -errno if it changed dev->features itself. 1295 * 1296 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1297 * struct net_device *dev, 1298 * const unsigned char *addr, u16 vid, u16 flags, 1299 * struct netlink_ext_ack *extack); 1300 * Adds an FDB entry to dev for addr. 1301 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1302 * struct net_device *dev, 1303 * const unsigned char *addr, u16 vid) 1304 * Deletes the FDB entry from dev coresponding to addr. 1305 * int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, struct nlattr *tb[], 1306 * struct net_device *dev, 1307 * u16 vid, 1308 * struct netlink_ext_ack *extack); 1309 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1310 * struct net_device *dev, struct net_device *filter_dev, 1311 * int *idx) 1312 * Used to add FDB entries to dump requests. Implementers should add 1313 * entries to skb and update idx with the number of entries. 1314 * 1315 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[], 1316 * u16 nlmsg_flags, struct netlink_ext_ack *extack); 1317 * Adds an MDB entry to dev. 1318 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[], 1319 * struct netlink_ext_ack *extack); 1320 * Deletes the MDB entry from dev. 1321 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb, 1322 * struct netlink_callback *cb); 1323 * Dumps MDB entries from dev. The first argument (marker) in the netlink 1324 * callback is used by core rtnetlink code. 1325 * 1326 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1327 * u16 flags, struct netlink_ext_ack *extack) 1328 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1329 * struct net_device *dev, u32 filter_mask, 1330 * int nlflags) 1331 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1332 * u16 flags); 1333 * 1334 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1335 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1336 * which do not represent real hardware may define this to allow their 1337 * userspace components to manage their virtual carrier state. Devices 1338 * that determine carrier state from physical hardware properties (eg 1339 * network cables) or protocol-dependent mechanisms (eg 1340 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1341 * 1342 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1343 * struct netdev_phys_item_id *ppid); 1344 * Called to get ID of physical port of this device. If driver does 1345 * not implement this, it is assumed that the hw is not able to have 1346 * multiple net devices on single physical port. 1347 * 1348 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1349 * struct netdev_phys_item_id *ppid) 1350 * Called to get the parent ID of the physical port of this device. 1351 * 1352 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1353 * struct net_device *dev) 1354 * Called by upper layer devices to accelerate switching or other 1355 * station functionality into hardware. 'pdev is the lowerdev 1356 * to use for the offload and 'dev' is the net device that will 1357 * back the offload. Returns a pointer to the private structure 1358 * the upper layer will maintain. 1359 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1360 * Called by upper layer device to delete the station created 1361 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1362 * the station and priv is the structure returned by the add 1363 * operation. 1364 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1365 * int queue_index, u32 maxrate); 1366 * Called when a user wants to set a max-rate limitation of specific 1367 * TX queue. 1368 * int (*ndo_get_iflink)(const struct net_device *dev); 1369 * Called to get the iflink value of this device. 1370 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1371 * This function is used to get egress tunnel information for given skb. 1372 * This is useful for retrieving outer tunnel header parameters while 1373 * sampling packet. 1374 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1375 * This function is used to specify the headroom that the skb must 1376 * consider when allocation skb during packet reception. Setting 1377 * appropriate rx headroom value allows avoiding skb head copy on 1378 * forward. Setting a negative value resets the rx headroom to the 1379 * default value. 1380 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1381 * This function is used to set or query state related to XDP on the 1382 * netdevice and manage BPF offload. See definition of 1383 * enum bpf_netdev_command for details. 1384 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1385 * u32 flags); 1386 * This function is used to submit @n XDP packets for transmit on a 1387 * netdevice. Returns number of frames successfully transmitted, frames 1388 * that got dropped are freed/returned via xdp_return_frame(). 1389 * Returns negative number, means general error invoking ndo, meaning 1390 * no frames were xmit'ed and core-caller will free all frames. 1391 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1392 * struct xdp_buff *xdp); 1393 * Get the xmit slave of master device based on the xdp_buff. 1394 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1395 * This function is used to wake up the softirq, ksoftirqd or kthread 1396 * responsible for sending and/or receiving packets on a specific 1397 * queue id bound to an AF_XDP socket. The flags field specifies if 1398 * only RX, only Tx, or both should be woken up using the flags 1399 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1400 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1401 * int cmd); 1402 * Add, change, delete or get information on an IPv4 tunnel. 1403 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1404 * If a device is paired with a peer device, return the peer instance. 1405 * The caller must be under RCU read context. 1406 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1407 * Get the forwarding path to reach the real device from the HW destination address 1408 * ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1409 * const struct skb_shared_hwtstamps *hwtstamps, 1410 * bool cycles); 1411 * Get hardware timestamp based on normal/adjustable time or free running 1412 * cycle counter. This function is required if physical clock supports a 1413 * free running cycle counter. 1414 */ 1415 struct net_device_ops { 1416 int (*ndo_init)(struct net_device *dev); 1417 void (*ndo_uninit)(struct net_device *dev); 1418 int (*ndo_open)(struct net_device *dev); 1419 int (*ndo_stop)(struct net_device *dev); 1420 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1421 struct net_device *dev); 1422 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1423 struct net_device *dev, 1424 netdev_features_t features); 1425 u16 (*ndo_select_queue)(struct net_device *dev, 1426 struct sk_buff *skb, 1427 struct net_device *sb_dev); 1428 void (*ndo_change_rx_flags)(struct net_device *dev, 1429 int flags); 1430 void (*ndo_set_rx_mode)(struct net_device *dev); 1431 int (*ndo_set_mac_address)(struct net_device *dev, 1432 void *addr); 1433 int (*ndo_validate_addr)(struct net_device *dev); 1434 int (*ndo_do_ioctl)(struct net_device *dev, 1435 struct ifreq *ifr, int cmd); 1436 int (*ndo_eth_ioctl)(struct net_device *dev, 1437 struct ifreq *ifr, int cmd); 1438 int (*ndo_siocbond)(struct net_device *dev, 1439 struct ifreq *ifr, int cmd); 1440 int (*ndo_siocwandev)(struct net_device *dev, 1441 struct if_settings *ifs); 1442 int (*ndo_siocdevprivate)(struct net_device *dev, 1443 struct ifreq *ifr, 1444 void __user *data, int cmd); 1445 int (*ndo_set_config)(struct net_device *dev, 1446 struct ifmap *map); 1447 int (*ndo_change_mtu)(struct net_device *dev, 1448 int new_mtu); 1449 int (*ndo_neigh_setup)(struct net_device *dev, 1450 struct neigh_parms *); 1451 void (*ndo_tx_timeout) (struct net_device *dev, 1452 unsigned int txqueue); 1453 1454 void (*ndo_get_stats64)(struct net_device *dev, 1455 struct rtnl_link_stats64 *storage); 1456 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1457 int (*ndo_get_offload_stats)(int attr_id, 1458 const struct net_device *dev, 1459 void *attr_data); 1460 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1461 1462 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1463 __be16 proto, u16 vid); 1464 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1465 __be16 proto, u16 vid); 1466 #ifdef CONFIG_NET_POLL_CONTROLLER 1467 void (*ndo_poll_controller)(struct net_device *dev); 1468 int (*ndo_netpoll_setup)(struct net_device *dev, 1469 struct netpoll_info *info); 1470 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1471 #endif 1472 int (*ndo_set_vf_mac)(struct net_device *dev, 1473 int queue, u8 *mac); 1474 int (*ndo_set_vf_vlan)(struct net_device *dev, 1475 int queue, u16 vlan, 1476 u8 qos, __be16 proto); 1477 int (*ndo_set_vf_rate)(struct net_device *dev, 1478 int vf, int min_tx_rate, 1479 int max_tx_rate); 1480 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1481 int vf, bool setting); 1482 int (*ndo_set_vf_trust)(struct net_device *dev, 1483 int vf, bool setting); 1484 int (*ndo_get_vf_config)(struct net_device *dev, 1485 int vf, 1486 struct ifla_vf_info *ivf); 1487 int (*ndo_set_vf_link_state)(struct net_device *dev, 1488 int vf, int link_state); 1489 int (*ndo_get_vf_stats)(struct net_device *dev, 1490 int vf, 1491 struct ifla_vf_stats 1492 *vf_stats); 1493 int (*ndo_set_vf_port)(struct net_device *dev, 1494 int vf, 1495 struct nlattr *port[]); 1496 int (*ndo_get_vf_port)(struct net_device *dev, 1497 int vf, struct sk_buff *skb); 1498 int (*ndo_get_vf_guid)(struct net_device *dev, 1499 int vf, 1500 struct ifla_vf_guid *node_guid, 1501 struct ifla_vf_guid *port_guid); 1502 int (*ndo_set_vf_guid)(struct net_device *dev, 1503 int vf, u64 guid, 1504 int guid_type); 1505 int (*ndo_set_vf_rss_query_en)( 1506 struct net_device *dev, 1507 int vf, bool setting); 1508 int (*ndo_setup_tc)(struct net_device *dev, 1509 enum tc_setup_type type, 1510 void *type_data); 1511 #if IS_ENABLED(CONFIG_FCOE) 1512 int (*ndo_fcoe_enable)(struct net_device *dev); 1513 int (*ndo_fcoe_disable)(struct net_device *dev); 1514 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1515 u16 xid, 1516 struct scatterlist *sgl, 1517 unsigned int sgc); 1518 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1519 u16 xid); 1520 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1521 u16 xid, 1522 struct scatterlist *sgl, 1523 unsigned int sgc); 1524 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1525 struct netdev_fcoe_hbainfo *hbainfo); 1526 #endif 1527 1528 #if IS_ENABLED(CONFIG_LIBFCOE) 1529 #define NETDEV_FCOE_WWNN 0 1530 #define NETDEV_FCOE_WWPN 1 1531 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1532 u64 *wwn, int type); 1533 #endif 1534 1535 #ifdef CONFIG_RFS_ACCEL 1536 int (*ndo_rx_flow_steer)(struct net_device *dev, 1537 const struct sk_buff *skb, 1538 u16 rxq_index, 1539 u32 flow_id); 1540 #endif 1541 int (*ndo_add_slave)(struct net_device *dev, 1542 struct net_device *slave_dev, 1543 struct netlink_ext_ack *extack); 1544 int (*ndo_del_slave)(struct net_device *dev, 1545 struct net_device *slave_dev); 1546 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1547 struct sk_buff *skb, 1548 bool all_slaves); 1549 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1550 struct sock *sk); 1551 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1552 netdev_features_t features); 1553 int (*ndo_set_features)(struct net_device *dev, 1554 netdev_features_t features); 1555 int (*ndo_neigh_construct)(struct net_device *dev, 1556 struct neighbour *n); 1557 void (*ndo_neigh_destroy)(struct net_device *dev, 1558 struct neighbour *n); 1559 1560 int (*ndo_fdb_add)(struct ndmsg *ndm, 1561 struct nlattr *tb[], 1562 struct net_device *dev, 1563 const unsigned char *addr, 1564 u16 vid, 1565 u16 flags, 1566 struct netlink_ext_ack *extack); 1567 int (*ndo_fdb_del)(struct ndmsg *ndm, 1568 struct nlattr *tb[], 1569 struct net_device *dev, 1570 const unsigned char *addr, 1571 u16 vid, struct netlink_ext_ack *extack); 1572 int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, 1573 struct nlattr *tb[], 1574 struct net_device *dev, 1575 u16 vid, 1576 struct netlink_ext_ack *extack); 1577 int (*ndo_fdb_dump)(struct sk_buff *skb, 1578 struct netlink_callback *cb, 1579 struct net_device *dev, 1580 struct net_device *filter_dev, 1581 int *idx); 1582 int (*ndo_fdb_get)(struct sk_buff *skb, 1583 struct nlattr *tb[], 1584 struct net_device *dev, 1585 const unsigned char *addr, 1586 u16 vid, u32 portid, u32 seq, 1587 struct netlink_ext_ack *extack); 1588 int (*ndo_mdb_add)(struct net_device *dev, 1589 struct nlattr *tb[], 1590 u16 nlmsg_flags, 1591 struct netlink_ext_ack *extack); 1592 int (*ndo_mdb_del)(struct net_device *dev, 1593 struct nlattr *tb[], 1594 struct netlink_ext_ack *extack); 1595 int (*ndo_mdb_dump)(struct net_device *dev, 1596 struct sk_buff *skb, 1597 struct netlink_callback *cb); 1598 int (*ndo_bridge_setlink)(struct net_device *dev, 1599 struct nlmsghdr *nlh, 1600 u16 flags, 1601 struct netlink_ext_ack *extack); 1602 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1603 u32 pid, u32 seq, 1604 struct net_device *dev, 1605 u32 filter_mask, 1606 int nlflags); 1607 int (*ndo_bridge_dellink)(struct net_device *dev, 1608 struct nlmsghdr *nlh, 1609 u16 flags); 1610 int (*ndo_change_carrier)(struct net_device *dev, 1611 bool new_carrier); 1612 int (*ndo_get_phys_port_id)(struct net_device *dev, 1613 struct netdev_phys_item_id *ppid); 1614 int (*ndo_get_port_parent_id)(struct net_device *dev, 1615 struct netdev_phys_item_id *ppid); 1616 int (*ndo_get_phys_port_name)(struct net_device *dev, 1617 char *name, size_t len); 1618 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1619 struct net_device *dev); 1620 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1621 void *priv); 1622 1623 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1624 int queue_index, 1625 u32 maxrate); 1626 int (*ndo_get_iflink)(const struct net_device *dev); 1627 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1628 struct sk_buff *skb); 1629 void (*ndo_set_rx_headroom)(struct net_device *dev, 1630 int needed_headroom); 1631 int (*ndo_bpf)(struct net_device *dev, 1632 struct netdev_bpf *bpf); 1633 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1634 struct xdp_frame **xdp, 1635 u32 flags); 1636 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1637 struct xdp_buff *xdp); 1638 int (*ndo_xsk_wakeup)(struct net_device *dev, 1639 u32 queue_id, u32 flags); 1640 int (*ndo_tunnel_ctl)(struct net_device *dev, 1641 struct ip_tunnel_parm *p, int cmd); 1642 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1643 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1644 struct net_device_path *path); 1645 ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1646 const struct skb_shared_hwtstamps *hwtstamps, 1647 bool cycles); 1648 }; 1649 1650 struct xdp_metadata_ops { 1651 int (*xmo_rx_timestamp)(const struct xdp_md *ctx, u64 *timestamp); 1652 int (*xmo_rx_hash)(const struct xdp_md *ctx, u32 *hash); 1653 }; 1654 1655 /** 1656 * enum netdev_priv_flags - &struct net_device priv_flags 1657 * 1658 * These are the &struct net_device, they are only set internally 1659 * by drivers and used in the kernel. These flags are invisible to 1660 * userspace; this means that the order of these flags can change 1661 * during any kernel release. 1662 * 1663 * You should have a pretty good reason to be extending these flags. 1664 * 1665 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1666 * @IFF_EBRIDGE: Ethernet bridging device 1667 * @IFF_BONDING: bonding master or slave 1668 * @IFF_ISATAP: ISATAP interface (RFC4214) 1669 * @IFF_WAN_HDLC: WAN HDLC device 1670 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1671 * release skb->dst 1672 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1673 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1674 * @IFF_MACVLAN_PORT: device used as macvlan port 1675 * @IFF_BRIDGE_PORT: device used as bridge port 1676 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1677 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1678 * @IFF_UNICAST_FLT: Supports unicast filtering 1679 * @IFF_TEAM_PORT: device used as team port 1680 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1681 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1682 * change when it's running 1683 * @IFF_MACVLAN: Macvlan device 1684 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1685 * underlying stacked devices 1686 * @IFF_L3MDEV_MASTER: device is an L3 master device 1687 * @IFF_NO_QUEUE: device can run without qdisc attached 1688 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1689 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1690 * @IFF_TEAM: device is a team device 1691 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1692 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1693 * entity (i.e. the master device for bridged veth) 1694 * @IFF_MACSEC: device is a MACsec device 1695 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1696 * @IFF_FAILOVER: device is a failover master device 1697 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1698 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1699 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf 1700 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1701 * skb_headlen(skb) == 0 (data starts from frag0) 1702 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN 1703 */ 1704 enum netdev_priv_flags { 1705 IFF_802_1Q_VLAN = 1<<0, 1706 IFF_EBRIDGE = 1<<1, 1707 IFF_BONDING = 1<<2, 1708 IFF_ISATAP = 1<<3, 1709 IFF_WAN_HDLC = 1<<4, 1710 IFF_XMIT_DST_RELEASE = 1<<5, 1711 IFF_DONT_BRIDGE = 1<<6, 1712 IFF_DISABLE_NETPOLL = 1<<7, 1713 IFF_MACVLAN_PORT = 1<<8, 1714 IFF_BRIDGE_PORT = 1<<9, 1715 IFF_OVS_DATAPATH = 1<<10, 1716 IFF_TX_SKB_SHARING = 1<<11, 1717 IFF_UNICAST_FLT = 1<<12, 1718 IFF_TEAM_PORT = 1<<13, 1719 IFF_SUPP_NOFCS = 1<<14, 1720 IFF_LIVE_ADDR_CHANGE = 1<<15, 1721 IFF_MACVLAN = 1<<16, 1722 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1723 IFF_L3MDEV_MASTER = 1<<18, 1724 IFF_NO_QUEUE = 1<<19, 1725 IFF_OPENVSWITCH = 1<<20, 1726 IFF_L3MDEV_SLAVE = 1<<21, 1727 IFF_TEAM = 1<<22, 1728 IFF_RXFH_CONFIGURED = 1<<23, 1729 IFF_PHONY_HEADROOM = 1<<24, 1730 IFF_MACSEC = 1<<25, 1731 IFF_NO_RX_HANDLER = 1<<26, 1732 IFF_FAILOVER = 1<<27, 1733 IFF_FAILOVER_SLAVE = 1<<28, 1734 IFF_L3MDEV_RX_HANDLER = 1<<29, 1735 IFF_NO_ADDRCONF = BIT_ULL(30), 1736 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31), 1737 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32), 1738 }; 1739 1740 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1741 #define IFF_EBRIDGE IFF_EBRIDGE 1742 #define IFF_BONDING IFF_BONDING 1743 #define IFF_ISATAP IFF_ISATAP 1744 #define IFF_WAN_HDLC IFF_WAN_HDLC 1745 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1746 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1747 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1748 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1749 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1750 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1751 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1752 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1753 #define IFF_TEAM_PORT IFF_TEAM_PORT 1754 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1755 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1756 #define IFF_MACVLAN IFF_MACVLAN 1757 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1758 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1759 #define IFF_NO_QUEUE IFF_NO_QUEUE 1760 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1761 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1762 #define IFF_TEAM IFF_TEAM 1763 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1764 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM 1765 #define IFF_MACSEC IFF_MACSEC 1766 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1767 #define IFF_FAILOVER IFF_FAILOVER 1768 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1769 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1770 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR 1771 1772 /* Specifies the type of the struct net_device::ml_priv pointer */ 1773 enum netdev_ml_priv_type { 1774 ML_PRIV_NONE, 1775 ML_PRIV_CAN, 1776 }; 1777 1778 /** 1779 * struct net_device - The DEVICE structure. 1780 * 1781 * Actually, this whole structure is a big mistake. It mixes I/O 1782 * data with strictly "high-level" data, and it has to know about 1783 * almost every data structure used in the INET module. 1784 * 1785 * @name: This is the first field of the "visible" part of this structure 1786 * (i.e. as seen by users in the "Space.c" file). It is the name 1787 * of the interface. 1788 * 1789 * @name_node: Name hashlist node 1790 * @ifalias: SNMP alias 1791 * @mem_end: Shared memory end 1792 * @mem_start: Shared memory start 1793 * @base_addr: Device I/O address 1794 * @irq: Device IRQ number 1795 * 1796 * @state: Generic network queuing layer state, see netdev_state_t 1797 * @dev_list: The global list of network devices 1798 * @napi_list: List entry used for polling NAPI devices 1799 * @unreg_list: List entry when we are unregistering the 1800 * device; see the function unregister_netdev 1801 * @close_list: List entry used when we are closing the device 1802 * @ptype_all: Device-specific packet handlers for all protocols 1803 * @ptype_specific: Device-specific, protocol-specific packet handlers 1804 * 1805 * @adj_list: Directly linked devices, like slaves for bonding 1806 * @features: Currently active device features 1807 * @hw_features: User-changeable features 1808 * 1809 * @wanted_features: User-requested features 1810 * @vlan_features: Mask of features inheritable by VLAN devices 1811 * 1812 * @hw_enc_features: Mask of features inherited by encapsulating devices 1813 * This field indicates what encapsulation 1814 * offloads the hardware is capable of doing, 1815 * and drivers will need to set them appropriately. 1816 * 1817 * @mpls_features: Mask of features inheritable by MPLS 1818 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1819 * 1820 * @ifindex: interface index 1821 * @group: The group the device belongs to 1822 * 1823 * @stats: Statistics struct, which was left as a legacy, use 1824 * rtnl_link_stats64 instead 1825 * 1826 * @core_stats: core networking counters, 1827 * do not use this in drivers 1828 * @carrier_up_count: Number of times the carrier has been up 1829 * @carrier_down_count: Number of times the carrier has been down 1830 * 1831 * @wireless_handlers: List of functions to handle Wireless Extensions, 1832 * instead of ioctl, 1833 * see <net/iw_handler.h> for details. 1834 * @wireless_data: Instance data managed by the core of wireless extensions 1835 * 1836 * @netdev_ops: Includes several pointers to callbacks, 1837 * if one wants to override the ndo_*() functions 1838 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks. 1839 * @ethtool_ops: Management operations 1840 * @l3mdev_ops: Layer 3 master device operations 1841 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1842 * discovery handling. Necessary for e.g. 6LoWPAN. 1843 * @xfrmdev_ops: Transformation offload operations 1844 * @tlsdev_ops: Transport Layer Security offload operations 1845 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1846 * of Layer 2 headers. 1847 * 1848 * @flags: Interface flags (a la BSD) 1849 * @xdp_features: XDP capability supported by the device 1850 * @priv_flags: Like 'flags' but invisible to userspace, 1851 * see if.h for the definitions 1852 * @gflags: Global flags ( kept as legacy ) 1853 * @padded: How much padding added by alloc_netdev() 1854 * @operstate: RFC2863 operstate 1855 * @link_mode: Mapping policy to operstate 1856 * @if_port: Selectable AUI, TP, ... 1857 * @dma: DMA channel 1858 * @mtu: Interface MTU value 1859 * @min_mtu: Interface Minimum MTU value 1860 * @max_mtu: Interface Maximum MTU value 1861 * @type: Interface hardware type 1862 * @hard_header_len: Maximum hardware header length. 1863 * @min_header_len: Minimum hardware header length 1864 * 1865 * @needed_headroom: Extra headroom the hardware may need, but not in all 1866 * cases can this be guaranteed 1867 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1868 * cases can this be guaranteed. Some cases also use 1869 * LL_MAX_HEADER instead to allocate the skb 1870 * 1871 * interface address info: 1872 * 1873 * @perm_addr: Permanent hw address 1874 * @addr_assign_type: Hw address assignment type 1875 * @addr_len: Hardware address length 1876 * @upper_level: Maximum depth level of upper devices. 1877 * @lower_level: Maximum depth level of lower devices. 1878 * @neigh_priv_len: Used in neigh_alloc() 1879 * @dev_id: Used to differentiate devices that share 1880 * the same link layer address 1881 * @dev_port: Used to differentiate devices that share 1882 * the same function 1883 * @addr_list_lock: XXX: need comments on this one 1884 * @name_assign_type: network interface name assignment type 1885 * @uc_promisc: Counter that indicates promiscuous mode 1886 * has been enabled due to the need to listen to 1887 * additional unicast addresses in a device that 1888 * does not implement ndo_set_rx_mode() 1889 * @uc: unicast mac addresses 1890 * @mc: multicast mac addresses 1891 * @dev_addrs: list of device hw addresses 1892 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1893 * @promiscuity: Number of times the NIC is told to work in 1894 * promiscuous mode; if it becomes 0 the NIC will 1895 * exit promiscuous mode 1896 * @allmulti: Counter, enables or disables allmulticast mode 1897 * 1898 * @vlan_info: VLAN info 1899 * @dsa_ptr: dsa specific data 1900 * @tipc_ptr: TIPC specific data 1901 * @atalk_ptr: AppleTalk link 1902 * @ip_ptr: IPv4 specific data 1903 * @ip6_ptr: IPv6 specific data 1904 * @ax25_ptr: AX.25 specific data 1905 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1906 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1907 * device struct 1908 * @mpls_ptr: mpls_dev struct pointer 1909 * @mctp_ptr: MCTP specific data 1910 * 1911 * @dev_addr: Hw address (before bcast, 1912 * because most packets are unicast) 1913 * 1914 * @_rx: Array of RX queues 1915 * @num_rx_queues: Number of RX queues 1916 * allocated at register_netdev() time 1917 * @real_num_rx_queues: Number of RX queues currently active in device 1918 * @xdp_prog: XDP sockets filter program pointer 1919 * @gro_flush_timeout: timeout for GRO layer in NAPI 1920 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1921 * allow to avoid NIC hard IRQ, on busy queues. 1922 * 1923 * @rx_handler: handler for received packets 1924 * @rx_handler_data: XXX: need comments on this one 1925 * @miniq_ingress: ingress/clsact qdisc specific data for 1926 * ingress processing 1927 * @ingress_queue: XXX: need comments on this one 1928 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1929 * @broadcast: hw bcast address 1930 * 1931 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1932 * indexed by RX queue number. Assigned by driver. 1933 * This must only be set if the ndo_rx_flow_steer 1934 * operation is defined 1935 * @index_hlist: Device index hash chain 1936 * 1937 * @_tx: Array of TX queues 1938 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1939 * @real_num_tx_queues: Number of TX queues currently active in device 1940 * @qdisc: Root qdisc from userspace point of view 1941 * @tx_queue_len: Max frames per queue allowed 1942 * @tx_global_lock: XXX: need comments on this one 1943 * @xdp_bulkq: XDP device bulk queue 1944 * @xps_maps: all CPUs/RXQs maps for XPS device 1945 * 1946 * @xps_maps: XXX: need comments on this one 1947 * @miniq_egress: clsact qdisc specific data for 1948 * egress processing 1949 * @nf_hooks_egress: netfilter hooks executed for egress packets 1950 * @qdisc_hash: qdisc hash table 1951 * @watchdog_timeo: Represents the timeout that is used by 1952 * the watchdog (see dev_watchdog()) 1953 * @watchdog_timer: List of timers 1954 * 1955 * @proto_down_reason: reason a netdev interface is held down 1956 * @pcpu_refcnt: Number of references to this device 1957 * @dev_refcnt: Number of references to this device 1958 * @refcnt_tracker: Tracker directory for tracked references to this device 1959 * @todo_list: Delayed register/unregister 1960 * @link_watch_list: XXX: need comments on this one 1961 * 1962 * @reg_state: Register/unregister state machine 1963 * @dismantle: Device is going to be freed 1964 * @rtnl_link_state: This enum represents the phases of creating 1965 * a new link 1966 * 1967 * @needs_free_netdev: Should unregister perform free_netdev? 1968 * @priv_destructor: Called from unregister 1969 * @npinfo: XXX: need comments on this one 1970 * @nd_net: Network namespace this network device is inside 1971 * 1972 * @ml_priv: Mid-layer private 1973 * @ml_priv_type: Mid-layer private type 1974 * @lstats: Loopback statistics 1975 * @tstats: Tunnel statistics 1976 * @dstats: Dummy statistics 1977 * @vstats: Virtual ethernet statistics 1978 * 1979 * @garp_port: GARP 1980 * @mrp_port: MRP 1981 * 1982 * @dm_private: Drop monitor private 1983 * 1984 * @dev: Class/net/name entry 1985 * @sysfs_groups: Space for optional device, statistics and wireless 1986 * sysfs groups 1987 * 1988 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1989 * @rtnl_link_ops: Rtnl_link_ops 1990 * 1991 * @gso_max_size: Maximum size of generic segmentation offload 1992 * @tso_max_size: Device (as in HW) limit on the max TSO request size 1993 * @gso_max_segs: Maximum number of segments that can be passed to the 1994 * NIC for GSO 1995 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count 1996 * @gso_ipv4_max_size: Maximum size of generic segmentation offload, 1997 * for IPv4. 1998 * 1999 * @dcbnl_ops: Data Center Bridging netlink ops 2000 * @num_tc: Number of traffic classes in the net device 2001 * @tc_to_txq: XXX: need comments on this one 2002 * @prio_tc_map: XXX: need comments on this one 2003 * 2004 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 2005 * 2006 * @priomap: XXX: need comments on this one 2007 * @phydev: Physical device may attach itself 2008 * for hardware timestamping 2009 * @sfp_bus: attached &struct sfp_bus structure. 2010 * 2011 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 2012 * 2013 * @proto_down: protocol port state information can be sent to the 2014 * switch driver and used to set the phys state of the 2015 * switch port. 2016 * 2017 * @wol_enabled: Wake-on-LAN is enabled 2018 * 2019 * @threaded: napi threaded mode is enabled 2020 * 2021 * @net_notifier_list: List of per-net netdev notifier block 2022 * that follow this device when it is moved 2023 * to another network namespace. 2024 * 2025 * @macsec_ops: MACsec offloading ops 2026 * 2027 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 2028 * offload capabilities of the device 2029 * @udp_tunnel_nic: UDP tunnel offload state 2030 * @xdp_state: stores info on attached XDP BPF programs 2031 * 2032 * @nested_level: Used as a parameter of spin_lock_nested() of 2033 * dev->addr_list_lock. 2034 * @unlink_list: As netif_addr_lock() can be called recursively, 2035 * keep a list of interfaces to be deleted. 2036 * @gro_max_size: Maximum size of aggregated packet in generic 2037 * receive offload (GRO) 2038 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic 2039 * receive offload (GRO), for IPv4. 2040 * 2041 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 2042 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 2043 * @watchdog_dev_tracker: refcount tracker used by watchdog. 2044 * @dev_registered_tracker: tracker for reference held while 2045 * registered 2046 * @offload_xstats_l3: L3 HW stats for this netdevice. 2047 * 2048 * @devlink_port: Pointer to related devlink port structure. 2049 * Assigned by a driver before netdev registration using 2050 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static 2051 * during the time netdevice is registered. 2052 * 2053 * FIXME: cleanup struct net_device such that network protocol info 2054 * moves out. 2055 */ 2056 2057 struct net_device { 2058 char name[IFNAMSIZ]; 2059 struct netdev_name_node *name_node; 2060 struct dev_ifalias __rcu *ifalias; 2061 /* 2062 * I/O specific fields 2063 * FIXME: Merge these and struct ifmap into one 2064 */ 2065 unsigned long mem_end; 2066 unsigned long mem_start; 2067 unsigned long base_addr; 2068 2069 /* 2070 * Some hardware also needs these fields (state,dev_list, 2071 * napi_list,unreg_list,close_list) but they are not 2072 * part of the usual set specified in Space.c. 2073 */ 2074 2075 unsigned long state; 2076 2077 struct list_head dev_list; 2078 struct list_head napi_list; 2079 struct list_head unreg_list; 2080 struct list_head close_list; 2081 struct list_head ptype_all; 2082 struct list_head ptype_specific; 2083 2084 struct { 2085 struct list_head upper; 2086 struct list_head lower; 2087 } adj_list; 2088 2089 /* Read-mostly cache-line for fast-path access */ 2090 unsigned int flags; 2091 xdp_features_t xdp_features; 2092 unsigned long long priv_flags; 2093 const struct net_device_ops *netdev_ops; 2094 const struct xdp_metadata_ops *xdp_metadata_ops; 2095 int ifindex; 2096 unsigned short gflags; 2097 unsigned short hard_header_len; 2098 2099 /* Note : dev->mtu is often read without holding a lock. 2100 * Writers usually hold RTNL. 2101 * It is recommended to use READ_ONCE() to annotate the reads, 2102 * and to use WRITE_ONCE() to annotate the writes. 2103 */ 2104 unsigned int mtu; 2105 unsigned short needed_headroom; 2106 unsigned short needed_tailroom; 2107 2108 netdev_features_t features; 2109 netdev_features_t hw_features; 2110 netdev_features_t wanted_features; 2111 netdev_features_t vlan_features; 2112 netdev_features_t hw_enc_features; 2113 netdev_features_t mpls_features; 2114 netdev_features_t gso_partial_features; 2115 2116 unsigned int min_mtu; 2117 unsigned int max_mtu; 2118 unsigned short type; 2119 unsigned char min_header_len; 2120 unsigned char name_assign_type; 2121 2122 int group; 2123 2124 struct net_device_stats stats; /* not used by modern drivers */ 2125 2126 struct net_device_core_stats __percpu *core_stats; 2127 2128 /* Stats to monitor link on/off, flapping */ 2129 atomic_t carrier_up_count; 2130 atomic_t carrier_down_count; 2131 2132 #ifdef CONFIG_WIRELESS_EXT 2133 const struct iw_handler_def *wireless_handlers; 2134 struct iw_public_data *wireless_data; 2135 #endif 2136 const struct ethtool_ops *ethtool_ops; 2137 #ifdef CONFIG_NET_L3_MASTER_DEV 2138 const struct l3mdev_ops *l3mdev_ops; 2139 #endif 2140 #if IS_ENABLED(CONFIG_IPV6) 2141 const struct ndisc_ops *ndisc_ops; 2142 #endif 2143 2144 #ifdef CONFIG_XFRM_OFFLOAD 2145 const struct xfrmdev_ops *xfrmdev_ops; 2146 #endif 2147 2148 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2149 const struct tlsdev_ops *tlsdev_ops; 2150 #endif 2151 2152 const struct header_ops *header_ops; 2153 2154 unsigned char operstate; 2155 unsigned char link_mode; 2156 2157 unsigned char if_port; 2158 unsigned char dma; 2159 2160 /* Interface address info. */ 2161 unsigned char perm_addr[MAX_ADDR_LEN]; 2162 unsigned char addr_assign_type; 2163 unsigned char addr_len; 2164 unsigned char upper_level; 2165 unsigned char lower_level; 2166 2167 unsigned short neigh_priv_len; 2168 unsigned short dev_id; 2169 unsigned short dev_port; 2170 unsigned short padded; 2171 2172 spinlock_t addr_list_lock; 2173 int irq; 2174 2175 struct netdev_hw_addr_list uc; 2176 struct netdev_hw_addr_list mc; 2177 struct netdev_hw_addr_list dev_addrs; 2178 2179 #ifdef CONFIG_SYSFS 2180 struct kset *queues_kset; 2181 #endif 2182 #ifdef CONFIG_LOCKDEP 2183 struct list_head unlink_list; 2184 #endif 2185 unsigned int promiscuity; 2186 unsigned int allmulti; 2187 bool uc_promisc; 2188 #ifdef CONFIG_LOCKDEP 2189 unsigned char nested_level; 2190 #endif 2191 2192 2193 /* Protocol-specific pointers */ 2194 2195 struct in_device __rcu *ip_ptr; 2196 struct inet6_dev __rcu *ip6_ptr; 2197 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2198 struct vlan_info __rcu *vlan_info; 2199 #endif 2200 #if IS_ENABLED(CONFIG_NET_DSA) 2201 struct dsa_port *dsa_ptr; 2202 #endif 2203 #if IS_ENABLED(CONFIG_TIPC) 2204 struct tipc_bearer __rcu *tipc_ptr; 2205 #endif 2206 #if IS_ENABLED(CONFIG_ATALK) 2207 void *atalk_ptr; 2208 #endif 2209 #if IS_ENABLED(CONFIG_AX25) 2210 void *ax25_ptr; 2211 #endif 2212 #if IS_ENABLED(CONFIG_CFG80211) 2213 struct wireless_dev *ieee80211_ptr; 2214 #endif 2215 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2216 struct wpan_dev *ieee802154_ptr; 2217 #endif 2218 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2219 struct mpls_dev __rcu *mpls_ptr; 2220 #endif 2221 #if IS_ENABLED(CONFIG_MCTP) 2222 struct mctp_dev __rcu *mctp_ptr; 2223 #endif 2224 2225 /* 2226 * Cache lines mostly used on receive path (including eth_type_trans()) 2227 */ 2228 /* Interface address info used in eth_type_trans() */ 2229 const unsigned char *dev_addr; 2230 2231 struct netdev_rx_queue *_rx; 2232 unsigned int num_rx_queues; 2233 unsigned int real_num_rx_queues; 2234 2235 struct bpf_prog __rcu *xdp_prog; 2236 unsigned long gro_flush_timeout; 2237 int napi_defer_hard_irqs; 2238 #define GRO_LEGACY_MAX_SIZE 65536u 2239 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2240 * and shinfo->gso_segs is a 16bit field. 2241 */ 2242 #define GRO_MAX_SIZE (8 * 65535u) 2243 unsigned int gro_max_size; 2244 unsigned int gro_ipv4_max_size; 2245 rx_handler_func_t __rcu *rx_handler; 2246 void __rcu *rx_handler_data; 2247 2248 #ifdef CONFIG_NET_CLS_ACT 2249 struct mini_Qdisc __rcu *miniq_ingress; 2250 #endif 2251 struct netdev_queue __rcu *ingress_queue; 2252 #ifdef CONFIG_NETFILTER_INGRESS 2253 struct nf_hook_entries __rcu *nf_hooks_ingress; 2254 #endif 2255 2256 unsigned char broadcast[MAX_ADDR_LEN]; 2257 #ifdef CONFIG_RFS_ACCEL 2258 struct cpu_rmap *rx_cpu_rmap; 2259 #endif 2260 struct hlist_node index_hlist; 2261 2262 /* 2263 * Cache lines mostly used on transmit path 2264 */ 2265 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2266 unsigned int num_tx_queues; 2267 unsigned int real_num_tx_queues; 2268 struct Qdisc __rcu *qdisc; 2269 unsigned int tx_queue_len; 2270 spinlock_t tx_global_lock; 2271 2272 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2273 2274 #ifdef CONFIG_XPS 2275 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2276 #endif 2277 #ifdef CONFIG_NET_CLS_ACT 2278 struct mini_Qdisc __rcu *miniq_egress; 2279 #endif 2280 #ifdef CONFIG_NETFILTER_EGRESS 2281 struct nf_hook_entries __rcu *nf_hooks_egress; 2282 #endif 2283 2284 #ifdef CONFIG_NET_SCHED 2285 DECLARE_HASHTABLE (qdisc_hash, 4); 2286 #endif 2287 /* These may be needed for future network-power-down code. */ 2288 struct timer_list watchdog_timer; 2289 int watchdog_timeo; 2290 2291 u32 proto_down_reason; 2292 2293 struct list_head todo_list; 2294 2295 #ifdef CONFIG_PCPU_DEV_REFCNT 2296 int __percpu *pcpu_refcnt; 2297 #else 2298 refcount_t dev_refcnt; 2299 #endif 2300 struct ref_tracker_dir refcnt_tracker; 2301 2302 struct list_head link_watch_list; 2303 2304 enum { NETREG_UNINITIALIZED=0, 2305 NETREG_REGISTERED, /* completed register_netdevice */ 2306 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2307 NETREG_UNREGISTERED, /* completed unregister todo */ 2308 NETREG_RELEASED, /* called free_netdev */ 2309 NETREG_DUMMY, /* dummy device for NAPI poll */ 2310 } reg_state:8; 2311 2312 bool dismantle; 2313 2314 enum { 2315 RTNL_LINK_INITIALIZED, 2316 RTNL_LINK_INITIALIZING, 2317 } rtnl_link_state:16; 2318 2319 bool needs_free_netdev; 2320 void (*priv_destructor)(struct net_device *dev); 2321 2322 #ifdef CONFIG_NETPOLL 2323 struct netpoll_info __rcu *npinfo; 2324 #endif 2325 2326 possible_net_t nd_net; 2327 2328 /* mid-layer private */ 2329 void *ml_priv; 2330 enum netdev_ml_priv_type ml_priv_type; 2331 2332 union { 2333 struct pcpu_lstats __percpu *lstats; 2334 struct pcpu_sw_netstats __percpu *tstats; 2335 struct pcpu_dstats __percpu *dstats; 2336 }; 2337 2338 #if IS_ENABLED(CONFIG_GARP) 2339 struct garp_port __rcu *garp_port; 2340 #endif 2341 #if IS_ENABLED(CONFIG_MRP) 2342 struct mrp_port __rcu *mrp_port; 2343 #endif 2344 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2345 struct dm_hw_stat_delta __rcu *dm_private; 2346 #endif 2347 struct device dev; 2348 const struct attribute_group *sysfs_groups[4]; 2349 const struct attribute_group *sysfs_rx_queue_group; 2350 2351 const struct rtnl_link_ops *rtnl_link_ops; 2352 2353 /* for setting kernel sock attribute on TCP connection setup */ 2354 #define GSO_MAX_SEGS 65535u 2355 #define GSO_LEGACY_MAX_SIZE 65536u 2356 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2357 * and shinfo->gso_segs is a 16bit field. 2358 */ 2359 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2360 2361 unsigned int gso_max_size; 2362 #define TSO_LEGACY_MAX_SIZE 65536 2363 #define TSO_MAX_SIZE UINT_MAX 2364 unsigned int tso_max_size; 2365 u16 gso_max_segs; 2366 #define TSO_MAX_SEGS U16_MAX 2367 u16 tso_max_segs; 2368 unsigned int gso_ipv4_max_size; 2369 2370 #ifdef CONFIG_DCB 2371 const struct dcbnl_rtnl_ops *dcbnl_ops; 2372 #endif 2373 s16 num_tc; 2374 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2375 u8 prio_tc_map[TC_BITMASK + 1]; 2376 2377 #if IS_ENABLED(CONFIG_FCOE) 2378 unsigned int fcoe_ddp_xid; 2379 #endif 2380 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2381 struct netprio_map __rcu *priomap; 2382 #endif 2383 struct phy_device *phydev; 2384 struct sfp_bus *sfp_bus; 2385 struct lock_class_key *qdisc_tx_busylock; 2386 bool proto_down; 2387 unsigned wol_enabled:1; 2388 unsigned threaded:1; 2389 2390 struct list_head net_notifier_list; 2391 2392 #if IS_ENABLED(CONFIG_MACSEC) 2393 /* MACsec management functions */ 2394 const struct macsec_ops *macsec_ops; 2395 #endif 2396 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2397 struct udp_tunnel_nic *udp_tunnel_nic; 2398 2399 /* protected by rtnl_lock */ 2400 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2401 2402 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2403 netdevice_tracker linkwatch_dev_tracker; 2404 netdevice_tracker watchdog_dev_tracker; 2405 netdevice_tracker dev_registered_tracker; 2406 struct rtnl_hw_stats64 *offload_xstats_l3; 2407 2408 struct devlink_port *devlink_port; 2409 }; 2410 #define to_net_dev(d) container_of(d, struct net_device, dev) 2411 2412 /* 2413 * Driver should use this to assign devlink port instance to a netdevice 2414 * before it registers the netdevice. Therefore devlink_port is static 2415 * during the netdev lifetime after it is registered. 2416 */ 2417 #define SET_NETDEV_DEVLINK_PORT(dev, port) \ 2418 ({ \ 2419 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \ 2420 ((dev)->devlink_port = (port)); \ 2421 }) 2422 2423 static inline bool netif_elide_gro(const struct net_device *dev) 2424 { 2425 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2426 return true; 2427 return false; 2428 } 2429 2430 #define NETDEV_ALIGN 32 2431 2432 static inline 2433 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2434 { 2435 return dev->prio_tc_map[prio & TC_BITMASK]; 2436 } 2437 2438 static inline 2439 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2440 { 2441 if (tc >= dev->num_tc) 2442 return -EINVAL; 2443 2444 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2445 return 0; 2446 } 2447 2448 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2449 void netdev_reset_tc(struct net_device *dev); 2450 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2451 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2452 2453 static inline 2454 int netdev_get_num_tc(struct net_device *dev) 2455 { 2456 return dev->num_tc; 2457 } 2458 2459 static inline void net_prefetch(void *p) 2460 { 2461 prefetch(p); 2462 #if L1_CACHE_BYTES < 128 2463 prefetch((u8 *)p + L1_CACHE_BYTES); 2464 #endif 2465 } 2466 2467 static inline void net_prefetchw(void *p) 2468 { 2469 prefetchw(p); 2470 #if L1_CACHE_BYTES < 128 2471 prefetchw((u8 *)p + L1_CACHE_BYTES); 2472 #endif 2473 } 2474 2475 void netdev_unbind_sb_channel(struct net_device *dev, 2476 struct net_device *sb_dev); 2477 int netdev_bind_sb_channel_queue(struct net_device *dev, 2478 struct net_device *sb_dev, 2479 u8 tc, u16 count, u16 offset); 2480 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2481 static inline int netdev_get_sb_channel(struct net_device *dev) 2482 { 2483 return max_t(int, -dev->num_tc, 0); 2484 } 2485 2486 static inline 2487 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2488 unsigned int index) 2489 { 2490 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues); 2491 return &dev->_tx[index]; 2492 } 2493 2494 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2495 const struct sk_buff *skb) 2496 { 2497 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2498 } 2499 2500 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2501 void (*f)(struct net_device *, 2502 struct netdev_queue *, 2503 void *), 2504 void *arg) 2505 { 2506 unsigned int i; 2507 2508 for (i = 0; i < dev->num_tx_queues; i++) 2509 f(dev, &dev->_tx[i], arg); 2510 } 2511 2512 #define netdev_lockdep_set_classes(dev) \ 2513 { \ 2514 static struct lock_class_key qdisc_tx_busylock_key; \ 2515 static struct lock_class_key qdisc_xmit_lock_key; \ 2516 static struct lock_class_key dev_addr_list_lock_key; \ 2517 unsigned int i; \ 2518 \ 2519 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2520 lockdep_set_class(&(dev)->addr_list_lock, \ 2521 &dev_addr_list_lock_key); \ 2522 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2523 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2524 &qdisc_xmit_lock_key); \ 2525 } 2526 2527 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2528 struct net_device *sb_dev); 2529 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2530 struct sk_buff *skb, 2531 struct net_device *sb_dev); 2532 2533 /* returns the headroom that the master device needs to take in account 2534 * when forwarding to this dev 2535 */ 2536 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2537 { 2538 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2539 } 2540 2541 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2542 { 2543 if (dev->netdev_ops->ndo_set_rx_headroom) 2544 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2545 } 2546 2547 /* set the device rx headroom to the dev's default */ 2548 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2549 { 2550 netdev_set_rx_headroom(dev, -1); 2551 } 2552 2553 static inline void *netdev_get_ml_priv(struct net_device *dev, 2554 enum netdev_ml_priv_type type) 2555 { 2556 if (dev->ml_priv_type != type) 2557 return NULL; 2558 2559 return dev->ml_priv; 2560 } 2561 2562 static inline void netdev_set_ml_priv(struct net_device *dev, 2563 void *ml_priv, 2564 enum netdev_ml_priv_type type) 2565 { 2566 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2567 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2568 dev->ml_priv_type, type); 2569 WARN(!dev->ml_priv_type && dev->ml_priv, 2570 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2571 2572 dev->ml_priv = ml_priv; 2573 dev->ml_priv_type = type; 2574 } 2575 2576 /* 2577 * Net namespace inlines 2578 */ 2579 static inline 2580 struct net *dev_net(const struct net_device *dev) 2581 { 2582 return read_pnet(&dev->nd_net); 2583 } 2584 2585 static inline 2586 void dev_net_set(struct net_device *dev, struct net *net) 2587 { 2588 write_pnet(&dev->nd_net, net); 2589 } 2590 2591 /** 2592 * netdev_priv - access network device private data 2593 * @dev: network device 2594 * 2595 * Get network device private data 2596 */ 2597 static inline void *netdev_priv(const struct net_device *dev) 2598 { 2599 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2600 } 2601 2602 /* Set the sysfs physical device reference for the network logical device 2603 * if set prior to registration will cause a symlink during initialization. 2604 */ 2605 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2606 2607 /* Set the sysfs device type for the network logical device to allow 2608 * fine-grained identification of different network device types. For 2609 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2610 */ 2611 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2612 2613 /* Default NAPI poll() weight 2614 * Device drivers are strongly advised to not use bigger value 2615 */ 2616 #define NAPI_POLL_WEIGHT 64 2617 2618 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2619 int (*poll)(struct napi_struct *, int), int weight); 2620 2621 /** 2622 * netif_napi_add() - initialize a NAPI context 2623 * @dev: network device 2624 * @napi: NAPI context 2625 * @poll: polling function 2626 * 2627 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2628 * *any* of the other NAPI-related functions. 2629 */ 2630 static inline void 2631 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2632 int (*poll)(struct napi_struct *, int)) 2633 { 2634 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2635 } 2636 2637 static inline void 2638 netif_napi_add_tx_weight(struct net_device *dev, 2639 struct napi_struct *napi, 2640 int (*poll)(struct napi_struct *, int), 2641 int weight) 2642 { 2643 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2644 netif_napi_add_weight(dev, napi, poll, weight); 2645 } 2646 2647 /** 2648 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2649 * @dev: network device 2650 * @napi: NAPI context 2651 * @poll: polling function 2652 * 2653 * This variant of netif_napi_add() should be used from drivers using NAPI 2654 * to exclusively poll a TX queue. 2655 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2656 */ 2657 static inline void netif_napi_add_tx(struct net_device *dev, 2658 struct napi_struct *napi, 2659 int (*poll)(struct napi_struct *, int)) 2660 { 2661 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2662 } 2663 2664 /** 2665 * __netif_napi_del - remove a NAPI context 2666 * @napi: NAPI context 2667 * 2668 * Warning: caller must observe RCU grace period before freeing memory 2669 * containing @napi. Drivers might want to call this helper to combine 2670 * all the needed RCU grace periods into a single one. 2671 */ 2672 void __netif_napi_del(struct napi_struct *napi); 2673 2674 /** 2675 * netif_napi_del - remove a NAPI context 2676 * @napi: NAPI context 2677 * 2678 * netif_napi_del() removes a NAPI context from the network device NAPI list 2679 */ 2680 static inline void netif_napi_del(struct napi_struct *napi) 2681 { 2682 __netif_napi_del(napi); 2683 synchronize_net(); 2684 } 2685 2686 struct packet_type { 2687 __be16 type; /* This is really htons(ether_type). */ 2688 bool ignore_outgoing; 2689 struct net_device *dev; /* NULL is wildcarded here */ 2690 netdevice_tracker dev_tracker; 2691 int (*func) (struct sk_buff *, 2692 struct net_device *, 2693 struct packet_type *, 2694 struct net_device *); 2695 void (*list_func) (struct list_head *, 2696 struct packet_type *, 2697 struct net_device *); 2698 bool (*id_match)(struct packet_type *ptype, 2699 struct sock *sk); 2700 struct net *af_packet_net; 2701 void *af_packet_priv; 2702 struct list_head list; 2703 }; 2704 2705 struct offload_callbacks { 2706 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2707 netdev_features_t features); 2708 struct sk_buff *(*gro_receive)(struct list_head *head, 2709 struct sk_buff *skb); 2710 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2711 }; 2712 2713 struct packet_offload { 2714 __be16 type; /* This is really htons(ether_type). */ 2715 u16 priority; 2716 struct offload_callbacks callbacks; 2717 struct list_head list; 2718 }; 2719 2720 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2721 struct pcpu_sw_netstats { 2722 u64_stats_t rx_packets; 2723 u64_stats_t rx_bytes; 2724 u64_stats_t tx_packets; 2725 u64_stats_t tx_bytes; 2726 struct u64_stats_sync syncp; 2727 } __aligned(4 * sizeof(u64)); 2728 2729 struct pcpu_lstats { 2730 u64_stats_t packets; 2731 u64_stats_t bytes; 2732 struct u64_stats_sync syncp; 2733 } __aligned(2 * sizeof(u64)); 2734 2735 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2736 2737 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2738 { 2739 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2740 2741 u64_stats_update_begin(&tstats->syncp); 2742 u64_stats_add(&tstats->rx_bytes, len); 2743 u64_stats_inc(&tstats->rx_packets); 2744 u64_stats_update_end(&tstats->syncp); 2745 } 2746 2747 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2748 unsigned int packets, 2749 unsigned int len) 2750 { 2751 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2752 2753 u64_stats_update_begin(&tstats->syncp); 2754 u64_stats_add(&tstats->tx_bytes, len); 2755 u64_stats_add(&tstats->tx_packets, packets); 2756 u64_stats_update_end(&tstats->syncp); 2757 } 2758 2759 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2760 { 2761 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2762 2763 u64_stats_update_begin(&lstats->syncp); 2764 u64_stats_add(&lstats->bytes, len); 2765 u64_stats_inc(&lstats->packets); 2766 u64_stats_update_end(&lstats->syncp); 2767 } 2768 2769 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2770 ({ \ 2771 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2772 if (pcpu_stats) { \ 2773 int __cpu; \ 2774 for_each_possible_cpu(__cpu) { \ 2775 typeof(type) *stat; \ 2776 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2777 u64_stats_init(&stat->syncp); \ 2778 } \ 2779 } \ 2780 pcpu_stats; \ 2781 }) 2782 2783 #define netdev_alloc_pcpu_stats(type) \ 2784 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2785 2786 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2787 ({ \ 2788 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2789 if (pcpu_stats) { \ 2790 int __cpu; \ 2791 for_each_possible_cpu(__cpu) { \ 2792 typeof(type) *stat; \ 2793 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2794 u64_stats_init(&stat->syncp); \ 2795 } \ 2796 } \ 2797 pcpu_stats; \ 2798 }) 2799 2800 enum netdev_lag_tx_type { 2801 NETDEV_LAG_TX_TYPE_UNKNOWN, 2802 NETDEV_LAG_TX_TYPE_RANDOM, 2803 NETDEV_LAG_TX_TYPE_BROADCAST, 2804 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2805 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2806 NETDEV_LAG_TX_TYPE_HASH, 2807 }; 2808 2809 enum netdev_lag_hash { 2810 NETDEV_LAG_HASH_NONE, 2811 NETDEV_LAG_HASH_L2, 2812 NETDEV_LAG_HASH_L34, 2813 NETDEV_LAG_HASH_L23, 2814 NETDEV_LAG_HASH_E23, 2815 NETDEV_LAG_HASH_E34, 2816 NETDEV_LAG_HASH_VLAN_SRCMAC, 2817 NETDEV_LAG_HASH_UNKNOWN, 2818 }; 2819 2820 struct netdev_lag_upper_info { 2821 enum netdev_lag_tx_type tx_type; 2822 enum netdev_lag_hash hash_type; 2823 }; 2824 2825 struct netdev_lag_lower_state_info { 2826 u8 link_up : 1, 2827 tx_enabled : 1; 2828 }; 2829 2830 #include <linux/notifier.h> 2831 2832 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2833 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2834 * adding new types. 2835 */ 2836 enum netdev_cmd { 2837 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2838 NETDEV_DOWN, 2839 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2840 detected a hardware crash and restarted 2841 - we can use this eg to kick tcp sessions 2842 once done */ 2843 NETDEV_CHANGE, /* Notify device state change */ 2844 NETDEV_REGISTER, 2845 NETDEV_UNREGISTER, 2846 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2847 NETDEV_CHANGEADDR, /* notify after the address change */ 2848 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2849 NETDEV_GOING_DOWN, 2850 NETDEV_CHANGENAME, 2851 NETDEV_FEAT_CHANGE, 2852 NETDEV_BONDING_FAILOVER, 2853 NETDEV_PRE_UP, 2854 NETDEV_PRE_TYPE_CHANGE, 2855 NETDEV_POST_TYPE_CHANGE, 2856 NETDEV_POST_INIT, 2857 NETDEV_PRE_UNINIT, 2858 NETDEV_RELEASE, 2859 NETDEV_NOTIFY_PEERS, 2860 NETDEV_JOIN, 2861 NETDEV_CHANGEUPPER, 2862 NETDEV_RESEND_IGMP, 2863 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2864 NETDEV_CHANGEINFODATA, 2865 NETDEV_BONDING_INFO, 2866 NETDEV_PRECHANGEUPPER, 2867 NETDEV_CHANGELOWERSTATE, 2868 NETDEV_UDP_TUNNEL_PUSH_INFO, 2869 NETDEV_UDP_TUNNEL_DROP_INFO, 2870 NETDEV_CHANGE_TX_QUEUE_LEN, 2871 NETDEV_CVLAN_FILTER_PUSH_INFO, 2872 NETDEV_CVLAN_FILTER_DROP_INFO, 2873 NETDEV_SVLAN_FILTER_PUSH_INFO, 2874 NETDEV_SVLAN_FILTER_DROP_INFO, 2875 NETDEV_OFFLOAD_XSTATS_ENABLE, 2876 NETDEV_OFFLOAD_XSTATS_DISABLE, 2877 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 2878 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 2879 NETDEV_XDP_FEAT_CHANGE, 2880 }; 2881 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2882 2883 int register_netdevice_notifier(struct notifier_block *nb); 2884 int unregister_netdevice_notifier(struct notifier_block *nb); 2885 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2886 int unregister_netdevice_notifier_net(struct net *net, 2887 struct notifier_block *nb); 2888 int register_netdevice_notifier_dev_net(struct net_device *dev, 2889 struct notifier_block *nb, 2890 struct netdev_net_notifier *nn); 2891 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2892 struct notifier_block *nb, 2893 struct netdev_net_notifier *nn); 2894 2895 struct netdev_notifier_info { 2896 struct net_device *dev; 2897 struct netlink_ext_ack *extack; 2898 }; 2899 2900 struct netdev_notifier_info_ext { 2901 struct netdev_notifier_info info; /* must be first */ 2902 union { 2903 u32 mtu; 2904 } ext; 2905 }; 2906 2907 struct netdev_notifier_change_info { 2908 struct netdev_notifier_info info; /* must be first */ 2909 unsigned int flags_changed; 2910 }; 2911 2912 struct netdev_notifier_changeupper_info { 2913 struct netdev_notifier_info info; /* must be first */ 2914 struct net_device *upper_dev; /* new upper dev */ 2915 bool master; /* is upper dev master */ 2916 bool linking; /* is the notification for link or unlink */ 2917 void *upper_info; /* upper dev info */ 2918 }; 2919 2920 struct netdev_notifier_changelowerstate_info { 2921 struct netdev_notifier_info info; /* must be first */ 2922 void *lower_state_info; /* is lower dev state */ 2923 }; 2924 2925 struct netdev_notifier_pre_changeaddr_info { 2926 struct netdev_notifier_info info; /* must be first */ 2927 const unsigned char *dev_addr; 2928 }; 2929 2930 enum netdev_offload_xstats_type { 2931 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 2932 }; 2933 2934 struct netdev_notifier_offload_xstats_info { 2935 struct netdev_notifier_info info; /* must be first */ 2936 enum netdev_offload_xstats_type type; 2937 2938 union { 2939 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 2940 struct netdev_notifier_offload_xstats_rd *report_delta; 2941 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 2942 struct netdev_notifier_offload_xstats_ru *report_used; 2943 }; 2944 }; 2945 2946 int netdev_offload_xstats_enable(struct net_device *dev, 2947 enum netdev_offload_xstats_type type, 2948 struct netlink_ext_ack *extack); 2949 int netdev_offload_xstats_disable(struct net_device *dev, 2950 enum netdev_offload_xstats_type type); 2951 bool netdev_offload_xstats_enabled(const struct net_device *dev, 2952 enum netdev_offload_xstats_type type); 2953 int netdev_offload_xstats_get(struct net_device *dev, 2954 enum netdev_offload_xstats_type type, 2955 struct rtnl_hw_stats64 *stats, bool *used, 2956 struct netlink_ext_ack *extack); 2957 void 2958 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 2959 const struct rtnl_hw_stats64 *stats); 2960 void 2961 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 2962 void netdev_offload_xstats_push_delta(struct net_device *dev, 2963 enum netdev_offload_xstats_type type, 2964 const struct rtnl_hw_stats64 *stats); 2965 2966 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2967 struct net_device *dev) 2968 { 2969 info->dev = dev; 2970 info->extack = NULL; 2971 } 2972 2973 static inline struct net_device * 2974 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2975 { 2976 return info->dev; 2977 } 2978 2979 static inline struct netlink_ext_ack * 2980 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2981 { 2982 return info->extack; 2983 } 2984 2985 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2986 2987 2988 extern rwlock_t dev_base_lock; /* Device list lock */ 2989 2990 #define for_each_netdev(net, d) \ 2991 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2992 #define for_each_netdev_reverse(net, d) \ 2993 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2994 #define for_each_netdev_rcu(net, d) \ 2995 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2996 #define for_each_netdev_safe(net, d, n) \ 2997 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2998 #define for_each_netdev_continue(net, d) \ 2999 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 3000 #define for_each_netdev_continue_reverse(net, d) \ 3001 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 3002 dev_list) 3003 #define for_each_netdev_continue_rcu(net, d) \ 3004 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 3005 #define for_each_netdev_in_bond_rcu(bond, slave) \ 3006 for_each_netdev_rcu(&init_net, slave) \ 3007 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 3008 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 3009 3010 static inline struct net_device *next_net_device(struct net_device *dev) 3011 { 3012 struct list_head *lh; 3013 struct net *net; 3014 3015 net = dev_net(dev); 3016 lh = dev->dev_list.next; 3017 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3018 } 3019 3020 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 3021 { 3022 struct list_head *lh; 3023 struct net *net; 3024 3025 net = dev_net(dev); 3026 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 3027 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3028 } 3029 3030 static inline struct net_device *first_net_device(struct net *net) 3031 { 3032 return list_empty(&net->dev_base_head) ? NULL : 3033 net_device_entry(net->dev_base_head.next); 3034 } 3035 3036 static inline struct net_device *first_net_device_rcu(struct net *net) 3037 { 3038 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 3039 3040 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3041 } 3042 3043 int netdev_boot_setup_check(struct net_device *dev); 3044 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 3045 const char *hwaddr); 3046 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 3047 void dev_add_pack(struct packet_type *pt); 3048 void dev_remove_pack(struct packet_type *pt); 3049 void __dev_remove_pack(struct packet_type *pt); 3050 void dev_add_offload(struct packet_offload *po); 3051 void dev_remove_offload(struct packet_offload *po); 3052 3053 int dev_get_iflink(const struct net_device *dev); 3054 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 3055 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 3056 struct net_device_path_stack *stack); 3057 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 3058 unsigned short mask); 3059 struct net_device *dev_get_by_name(struct net *net, const char *name); 3060 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 3061 struct net_device *__dev_get_by_name(struct net *net, const char *name); 3062 bool netdev_name_in_use(struct net *net, const char *name); 3063 int dev_alloc_name(struct net_device *dev, const char *name); 3064 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 3065 void dev_close(struct net_device *dev); 3066 void dev_close_many(struct list_head *head, bool unlink); 3067 void dev_disable_lro(struct net_device *dev); 3068 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 3069 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 3070 struct net_device *sb_dev); 3071 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 3072 struct net_device *sb_dev); 3073 3074 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3075 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3076 3077 static inline int dev_queue_xmit(struct sk_buff *skb) 3078 { 3079 return __dev_queue_xmit(skb, NULL); 3080 } 3081 3082 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3083 struct net_device *sb_dev) 3084 { 3085 return __dev_queue_xmit(skb, sb_dev); 3086 } 3087 3088 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3089 { 3090 int ret; 3091 3092 ret = __dev_direct_xmit(skb, queue_id); 3093 if (!dev_xmit_complete(ret)) 3094 kfree_skb(skb); 3095 return ret; 3096 } 3097 3098 int register_netdevice(struct net_device *dev); 3099 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3100 void unregister_netdevice_many(struct list_head *head); 3101 static inline void unregister_netdevice(struct net_device *dev) 3102 { 3103 unregister_netdevice_queue(dev, NULL); 3104 } 3105 3106 int netdev_refcnt_read(const struct net_device *dev); 3107 void free_netdev(struct net_device *dev); 3108 void netdev_freemem(struct net_device *dev); 3109 int init_dummy_netdev(struct net_device *dev); 3110 3111 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3112 struct sk_buff *skb, 3113 bool all_slaves); 3114 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3115 struct sock *sk); 3116 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3117 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3118 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3119 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 3120 int dev_restart(struct net_device *dev); 3121 3122 3123 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3124 unsigned short type, 3125 const void *daddr, const void *saddr, 3126 unsigned int len) 3127 { 3128 if (!dev->header_ops || !dev->header_ops->create) 3129 return 0; 3130 3131 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3132 } 3133 3134 static inline int dev_parse_header(const struct sk_buff *skb, 3135 unsigned char *haddr) 3136 { 3137 const struct net_device *dev = skb->dev; 3138 3139 if (!dev->header_ops || !dev->header_ops->parse) 3140 return 0; 3141 return dev->header_ops->parse(skb, haddr); 3142 } 3143 3144 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3145 { 3146 const struct net_device *dev = skb->dev; 3147 3148 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3149 return 0; 3150 return dev->header_ops->parse_protocol(skb); 3151 } 3152 3153 /* ll_header must have at least hard_header_len allocated */ 3154 static inline bool dev_validate_header(const struct net_device *dev, 3155 char *ll_header, int len) 3156 { 3157 if (likely(len >= dev->hard_header_len)) 3158 return true; 3159 if (len < dev->min_header_len) 3160 return false; 3161 3162 if (capable(CAP_SYS_RAWIO)) { 3163 memset(ll_header + len, 0, dev->hard_header_len - len); 3164 return true; 3165 } 3166 3167 if (dev->header_ops && dev->header_ops->validate) 3168 return dev->header_ops->validate(ll_header, len); 3169 3170 return false; 3171 } 3172 3173 static inline bool dev_has_header(const struct net_device *dev) 3174 { 3175 return dev->header_ops && dev->header_ops->create; 3176 } 3177 3178 /* 3179 * Incoming packets are placed on per-CPU queues 3180 */ 3181 struct softnet_data { 3182 struct list_head poll_list; 3183 struct sk_buff_head process_queue; 3184 3185 /* stats */ 3186 unsigned int processed; 3187 unsigned int time_squeeze; 3188 #ifdef CONFIG_RPS 3189 struct softnet_data *rps_ipi_list; 3190 #endif 3191 #ifdef CONFIG_NET_FLOW_LIMIT 3192 struct sd_flow_limit __rcu *flow_limit; 3193 #endif 3194 struct Qdisc *output_queue; 3195 struct Qdisc **output_queue_tailp; 3196 struct sk_buff *completion_queue; 3197 #ifdef CONFIG_XFRM_OFFLOAD 3198 struct sk_buff_head xfrm_backlog; 3199 #endif 3200 /* written and read only by owning cpu: */ 3201 struct { 3202 u16 recursion; 3203 u8 more; 3204 #ifdef CONFIG_NET_EGRESS 3205 u8 skip_txqueue; 3206 #endif 3207 } xmit; 3208 #ifdef CONFIG_RPS 3209 /* input_queue_head should be written by cpu owning this struct, 3210 * and only read by other cpus. Worth using a cache line. 3211 */ 3212 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3213 3214 /* Elements below can be accessed between CPUs for RPS/RFS */ 3215 call_single_data_t csd ____cacheline_aligned_in_smp; 3216 struct softnet_data *rps_ipi_next; 3217 unsigned int cpu; 3218 unsigned int input_queue_tail; 3219 #endif 3220 unsigned int received_rps; 3221 unsigned int dropped; 3222 struct sk_buff_head input_pkt_queue; 3223 struct napi_struct backlog; 3224 3225 /* Another possibly contended cache line */ 3226 spinlock_t defer_lock ____cacheline_aligned_in_smp; 3227 int defer_count; 3228 int defer_ipi_scheduled; 3229 struct sk_buff *defer_list; 3230 call_single_data_t defer_csd; 3231 }; 3232 3233 static inline void input_queue_head_incr(struct softnet_data *sd) 3234 { 3235 #ifdef CONFIG_RPS 3236 sd->input_queue_head++; 3237 #endif 3238 } 3239 3240 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3241 unsigned int *qtail) 3242 { 3243 #ifdef CONFIG_RPS 3244 *qtail = ++sd->input_queue_tail; 3245 #endif 3246 } 3247 3248 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3249 3250 static inline int dev_recursion_level(void) 3251 { 3252 return this_cpu_read(softnet_data.xmit.recursion); 3253 } 3254 3255 #define XMIT_RECURSION_LIMIT 8 3256 static inline bool dev_xmit_recursion(void) 3257 { 3258 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3259 XMIT_RECURSION_LIMIT); 3260 } 3261 3262 static inline void dev_xmit_recursion_inc(void) 3263 { 3264 __this_cpu_inc(softnet_data.xmit.recursion); 3265 } 3266 3267 static inline void dev_xmit_recursion_dec(void) 3268 { 3269 __this_cpu_dec(softnet_data.xmit.recursion); 3270 } 3271 3272 void __netif_schedule(struct Qdisc *q); 3273 void netif_schedule_queue(struct netdev_queue *txq); 3274 3275 static inline void netif_tx_schedule_all(struct net_device *dev) 3276 { 3277 unsigned int i; 3278 3279 for (i = 0; i < dev->num_tx_queues; i++) 3280 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3281 } 3282 3283 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3284 { 3285 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3286 } 3287 3288 /** 3289 * netif_start_queue - allow transmit 3290 * @dev: network device 3291 * 3292 * Allow upper layers to call the device hard_start_xmit routine. 3293 */ 3294 static inline void netif_start_queue(struct net_device *dev) 3295 { 3296 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3297 } 3298 3299 static inline void netif_tx_start_all_queues(struct net_device *dev) 3300 { 3301 unsigned int i; 3302 3303 for (i = 0; i < dev->num_tx_queues; i++) { 3304 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3305 netif_tx_start_queue(txq); 3306 } 3307 } 3308 3309 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3310 3311 /** 3312 * netif_wake_queue - restart transmit 3313 * @dev: network device 3314 * 3315 * Allow upper layers to call the device hard_start_xmit routine. 3316 * Used for flow control when transmit resources are available. 3317 */ 3318 static inline void netif_wake_queue(struct net_device *dev) 3319 { 3320 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3321 } 3322 3323 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3324 { 3325 unsigned int i; 3326 3327 for (i = 0; i < dev->num_tx_queues; i++) { 3328 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3329 netif_tx_wake_queue(txq); 3330 } 3331 } 3332 3333 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3334 { 3335 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3336 } 3337 3338 /** 3339 * netif_stop_queue - stop transmitted packets 3340 * @dev: network device 3341 * 3342 * Stop upper layers calling the device hard_start_xmit routine. 3343 * Used for flow control when transmit resources are unavailable. 3344 */ 3345 static inline void netif_stop_queue(struct net_device *dev) 3346 { 3347 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3348 } 3349 3350 void netif_tx_stop_all_queues(struct net_device *dev); 3351 3352 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3353 { 3354 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3355 } 3356 3357 /** 3358 * netif_queue_stopped - test if transmit queue is flowblocked 3359 * @dev: network device 3360 * 3361 * Test if transmit queue on device is currently unable to send. 3362 */ 3363 static inline bool netif_queue_stopped(const struct net_device *dev) 3364 { 3365 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3366 } 3367 3368 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3369 { 3370 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3371 } 3372 3373 static inline bool 3374 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3375 { 3376 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3377 } 3378 3379 static inline bool 3380 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3381 { 3382 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3383 } 3384 3385 /** 3386 * netdev_queue_set_dql_min_limit - set dql minimum limit 3387 * @dev_queue: pointer to transmit queue 3388 * @min_limit: dql minimum limit 3389 * 3390 * Forces xmit_more() to return true until the minimum threshold 3391 * defined by @min_limit is reached (or until the tx queue is 3392 * empty). Warning: to be use with care, misuse will impact the 3393 * latency. 3394 */ 3395 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3396 unsigned int min_limit) 3397 { 3398 #ifdef CONFIG_BQL 3399 dev_queue->dql.min_limit = min_limit; 3400 #endif 3401 } 3402 3403 /** 3404 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3405 * @dev_queue: pointer to transmit queue 3406 * 3407 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3408 * to give appropriate hint to the CPU. 3409 */ 3410 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3411 { 3412 #ifdef CONFIG_BQL 3413 prefetchw(&dev_queue->dql.num_queued); 3414 #endif 3415 } 3416 3417 /** 3418 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3419 * @dev_queue: pointer to transmit queue 3420 * 3421 * BQL enabled drivers might use this helper in their TX completion path, 3422 * to give appropriate hint to the CPU. 3423 */ 3424 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3425 { 3426 #ifdef CONFIG_BQL 3427 prefetchw(&dev_queue->dql.limit); 3428 #endif 3429 } 3430 3431 /** 3432 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3433 * @dev_queue: network device queue 3434 * @bytes: number of bytes queued to the device queue 3435 * 3436 * Report the number of bytes queued for sending/completion to the network 3437 * device hardware queue. @bytes should be a good approximation and should 3438 * exactly match netdev_completed_queue() @bytes. 3439 * This is typically called once per packet, from ndo_start_xmit(). 3440 */ 3441 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3442 unsigned int bytes) 3443 { 3444 #ifdef CONFIG_BQL 3445 dql_queued(&dev_queue->dql, bytes); 3446 3447 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3448 return; 3449 3450 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3451 3452 /* 3453 * The XOFF flag must be set before checking the dql_avail below, 3454 * because in netdev_tx_completed_queue we update the dql_completed 3455 * before checking the XOFF flag. 3456 */ 3457 smp_mb(); 3458 3459 /* check again in case another CPU has just made room avail */ 3460 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3461 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3462 #endif 3463 } 3464 3465 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3466 * that they should not test BQL status themselves. 3467 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3468 * skb of a batch. 3469 * Returns true if the doorbell must be used to kick the NIC. 3470 */ 3471 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3472 unsigned int bytes, 3473 bool xmit_more) 3474 { 3475 if (xmit_more) { 3476 #ifdef CONFIG_BQL 3477 dql_queued(&dev_queue->dql, bytes); 3478 #endif 3479 return netif_tx_queue_stopped(dev_queue); 3480 } 3481 netdev_tx_sent_queue(dev_queue, bytes); 3482 return true; 3483 } 3484 3485 /** 3486 * netdev_sent_queue - report the number of bytes queued to hardware 3487 * @dev: network device 3488 * @bytes: number of bytes queued to the hardware device queue 3489 * 3490 * Report the number of bytes queued for sending/completion to the network 3491 * device hardware queue#0. @bytes should be a good approximation and should 3492 * exactly match netdev_completed_queue() @bytes. 3493 * This is typically called once per packet, from ndo_start_xmit(). 3494 */ 3495 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3496 { 3497 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3498 } 3499 3500 static inline bool __netdev_sent_queue(struct net_device *dev, 3501 unsigned int bytes, 3502 bool xmit_more) 3503 { 3504 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3505 xmit_more); 3506 } 3507 3508 /** 3509 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3510 * @dev_queue: network device queue 3511 * @pkts: number of packets (currently ignored) 3512 * @bytes: number of bytes dequeued from the device queue 3513 * 3514 * Must be called at most once per TX completion round (and not per 3515 * individual packet), so that BQL can adjust its limits appropriately. 3516 */ 3517 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3518 unsigned int pkts, unsigned int bytes) 3519 { 3520 #ifdef CONFIG_BQL 3521 if (unlikely(!bytes)) 3522 return; 3523 3524 dql_completed(&dev_queue->dql, bytes); 3525 3526 /* 3527 * Without the memory barrier there is a small possiblity that 3528 * netdev_tx_sent_queue will miss the update and cause the queue to 3529 * be stopped forever 3530 */ 3531 smp_mb(); 3532 3533 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3534 return; 3535 3536 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3537 netif_schedule_queue(dev_queue); 3538 #endif 3539 } 3540 3541 /** 3542 * netdev_completed_queue - report bytes and packets completed by device 3543 * @dev: network device 3544 * @pkts: actual number of packets sent over the medium 3545 * @bytes: actual number of bytes sent over the medium 3546 * 3547 * Report the number of bytes and packets transmitted by the network device 3548 * hardware queue over the physical medium, @bytes must exactly match the 3549 * @bytes amount passed to netdev_sent_queue() 3550 */ 3551 static inline void netdev_completed_queue(struct net_device *dev, 3552 unsigned int pkts, unsigned int bytes) 3553 { 3554 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3555 } 3556 3557 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3558 { 3559 #ifdef CONFIG_BQL 3560 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3561 dql_reset(&q->dql); 3562 #endif 3563 } 3564 3565 /** 3566 * netdev_reset_queue - reset the packets and bytes count of a network device 3567 * @dev_queue: network device 3568 * 3569 * Reset the bytes and packet count of a network device and clear the 3570 * software flow control OFF bit for this network device 3571 */ 3572 static inline void netdev_reset_queue(struct net_device *dev_queue) 3573 { 3574 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3575 } 3576 3577 /** 3578 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3579 * @dev: network device 3580 * @queue_index: given tx queue index 3581 * 3582 * Returns 0 if given tx queue index >= number of device tx queues, 3583 * otherwise returns the originally passed tx queue index. 3584 */ 3585 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3586 { 3587 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3588 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3589 dev->name, queue_index, 3590 dev->real_num_tx_queues); 3591 return 0; 3592 } 3593 3594 return queue_index; 3595 } 3596 3597 /** 3598 * netif_running - test if up 3599 * @dev: network device 3600 * 3601 * Test if the device has been brought up. 3602 */ 3603 static inline bool netif_running(const struct net_device *dev) 3604 { 3605 return test_bit(__LINK_STATE_START, &dev->state); 3606 } 3607 3608 /* 3609 * Routines to manage the subqueues on a device. We only need start, 3610 * stop, and a check if it's stopped. All other device management is 3611 * done at the overall netdevice level. 3612 * Also test the device if we're multiqueue. 3613 */ 3614 3615 /** 3616 * netif_start_subqueue - allow sending packets on subqueue 3617 * @dev: network device 3618 * @queue_index: sub queue index 3619 * 3620 * Start individual transmit queue of a device with multiple transmit queues. 3621 */ 3622 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3623 { 3624 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3625 3626 netif_tx_start_queue(txq); 3627 } 3628 3629 /** 3630 * netif_stop_subqueue - stop sending packets on subqueue 3631 * @dev: network device 3632 * @queue_index: sub queue index 3633 * 3634 * Stop individual transmit queue of a device with multiple transmit queues. 3635 */ 3636 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3637 { 3638 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3639 netif_tx_stop_queue(txq); 3640 } 3641 3642 /** 3643 * __netif_subqueue_stopped - test status of subqueue 3644 * @dev: network device 3645 * @queue_index: sub queue index 3646 * 3647 * Check individual transmit queue of a device with multiple transmit queues. 3648 */ 3649 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3650 u16 queue_index) 3651 { 3652 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3653 3654 return netif_tx_queue_stopped(txq); 3655 } 3656 3657 /** 3658 * netif_subqueue_stopped - test status of subqueue 3659 * @dev: network device 3660 * @skb: sub queue buffer pointer 3661 * 3662 * Check individual transmit queue of a device with multiple transmit queues. 3663 */ 3664 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3665 struct sk_buff *skb) 3666 { 3667 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3668 } 3669 3670 /** 3671 * netif_wake_subqueue - allow sending packets on subqueue 3672 * @dev: network device 3673 * @queue_index: sub queue index 3674 * 3675 * Resume individual transmit queue of a device with multiple transmit queues. 3676 */ 3677 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3678 { 3679 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3680 3681 netif_tx_wake_queue(txq); 3682 } 3683 3684 #ifdef CONFIG_XPS 3685 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3686 u16 index); 3687 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3688 u16 index, enum xps_map_type type); 3689 3690 /** 3691 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3692 * @j: CPU/Rx queue index 3693 * @mask: bitmask of all cpus/rx queues 3694 * @nr_bits: number of bits in the bitmask 3695 * 3696 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3697 */ 3698 static inline bool netif_attr_test_mask(unsigned long j, 3699 const unsigned long *mask, 3700 unsigned int nr_bits) 3701 { 3702 cpu_max_bits_warn(j, nr_bits); 3703 return test_bit(j, mask); 3704 } 3705 3706 /** 3707 * netif_attr_test_online - Test for online CPU/Rx queue 3708 * @j: CPU/Rx queue index 3709 * @online_mask: bitmask for CPUs/Rx queues that are online 3710 * @nr_bits: number of bits in the bitmask 3711 * 3712 * Returns true if a CPU/Rx queue is online. 3713 */ 3714 static inline bool netif_attr_test_online(unsigned long j, 3715 const unsigned long *online_mask, 3716 unsigned int nr_bits) 3717 { 3718 cpu_max_bits_warn(j, nr_bits); 3719 3720 if (online_mask) 3721 return test_bit(j, online_mask); 3722 3723 return (j < nr_bits); 3724 } 3725 3726 /** 3727 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3728 * @n: CPU/Rx queue index 3729 * @srcp: the cpumask/Rx queue mask pointer 3730 * @nr_bits: number of bits in the bitmask 3731 * 3732 * Returns >= nr_bits if no further CPUs/Rx queues set. 3733 */ 3734 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3735 unsigned int nr_bits) 3736 { 3737 /* -1 is a legal arg here. */ 3738 if (n != -1) 3739 cpu_max_bits_warn(n, nr_bits); 3740 3741 if (srcp) 3742 return find_next_bit(srcp, nr_bits, n + 1); 3743 3744 return n + 1; 3745 } 3746 3747 /** 3748 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3749 * @n: CPU/Rx queue index 3750 * @src1p: the first CPUs/Rx queues mask pointer 3751 * @src2p: the second CPUs/Rx queues mask pointer 3752 * @nr_bits: number of bits in the bitmask 3753 * 3754 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3755 */ 3756 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3757 const unsigned long *src2p, 3758 unsigned int nr_bits) 3759 { 3760 /* -1 is a legal arg here. */ 3761 if (n != -1) 3762 cpu_max_bits_warn(n, nr_bits); 3763 3764 if (src1p && src2p) 3765 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3766 else if (src1p) 3767 return find_next_bit(src1p, nr_bits, n + 1); 3768 else if (src2p) 3769 return find_next_bit(src2p, nr_bits, n + 1); 3770 3771 return n + 1; 3772 } 3773 #else 3774 static inline int netif_set_xps_queue(struct net_device *dev, 3775 const struct cpumask *mask, 3776 u16 index) 3777 { 3778 return 0; 3779 } 3780 3781 static inline int __netif_set_xps_queue(struct net_device *dev, 3782 const unsigned long *mask, 3783 u16 index, enum xps_map_type type) 3784 { 3785 return 0; 3786 } 3787 #endif 3788 3789 /** 3790 * netif_is_multiqueue - test if device has multiple transmit queues 3791 * @dev: network device 3792 * 3793 * Check if device has multiple transmit queues 3794 */ 3795 static inline bool netif_is_multiqueue(const struct net_device *dev) 3796 { 3797 return dev->num_tx_queues > 1; 3798 } 3799 3800 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3801 3802 #ifdef CONFIG_SYSFS 3803 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3804 #else 3805 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3806 unsigned int rxqs) 3807 { 3808 dev->real_num_rx_queues = rxqs; 3809 return 0; 3810 } 3811 #endif 3812 int netif_set_real_num_queues(struct net_device *dev, 3813 unsigned int txq, unsigned int rxq); 3814 3815 static inline struct netdev_rx_queue * 3816 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3817 { 3818 return dev->_rx + rxq; 3819 } 3820 3821 #ifdef CONFIG_SYSFS 3822 static inline unsigned int get_netdev_rx_queue_index( 3823 struct netdev_rx_queue *queue) 3824 { 3825 struct net_device *dev = queue->dev; 3826 int index = queue - dev->_rx; 3827 3828 BUG_ON(index >= dev->num_rx_queues); 3829 return index; 3830 } 3831 #endif 3832 3833 int netif_get_num_default_rss_queues(void); 3834 3835 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason); 3836 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason); 3837 3838 /* 3839 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3840 * interrupt context or with hardware interrupts being disabled. 3841 * (in_hardirq() || irqs_disabled()) 3842 * 3843 * We provide four helpers that can be used in following contexts : 3844 * 3845 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3846 * replacing kfree_skb(skb) 3847 * 3848 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3849 * Typically used in place of consume_skb(skb) in TX completion path 3850 * 3851 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3852 * replacing kfree_skb(skb) 3853 * 3854 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3855 * and consumed a packet. Used in place of consume_skb(skb) 3856 */ 3857 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3858 { 3859 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 3860 } 3861 3862 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3863 { 3864 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED); 3865 } 3866 3867 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3868 { 3869 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 3870 } 3871 3872 static inline void dev_consume_skb_any(struct sk_buff *skb) 3873 { 3874 dev_kfree_skb_any_reason(skb, SKB_CONSUMED); 3875 } 3876 3877 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3878 struct bpf_prog *xdp_prog); 3879 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3880 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3881 int netif_rx(struct sk_buff *skb); 3882 int __netif_rx(struct sk_buff *skb); 3883 3884 int netif_receive_skb(struct sk_buff *skb); 3885 int netif_receive_skb_core(struct sk_buff *skb); 3886 void netif_receive_skb_list_internal(struct list_head *head); 3887 void netif_receive_skb_list(struct list_head *head); 3888 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3889 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3890 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3891 void napi_get_frags_check(struct napi_struct *napi); 3892 gro_result_t napi_gro_frags(struct napi_struct *napi); 3893 struct packet_offload *gro_find_receive_by_type(__be16 type); 3894 struct packet_offload *gro_find_complete_by_type(__be16 type); 3895 3896 static inline void napi_free_frags(struct napi_struct *napi) 3897 { 3898 kfree_skb(napi->skb); 3899 napi->skb = NULL; 3900 } 3901 3902 bool netdev_is_rx_handler_busy(struct net_device *dev); 3903 int netdev_rx_handler_register(struct net_device *dev, 3904 rx_handler_func_t *rx_handler, 3905 void *rx_handler_data); 3906 void netdev_rx_handler_unregister(struct net_device *dev); 3907 3908 bool dev_valid_name(const char *name); 3909 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 3910 { 3911 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 3912 } 3913 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 3914 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 3915 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3916 void __user *data, bool *need_copyout); 3917 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 3918 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 3919 unsigned int dev_get_flags(const struct net_device *); 3920 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3921 struct netlink_ext_ack *extack); 3922 int dev_change_flags(struct net_device *dev, unsigned int flags, 3923 struct netlink_ext_ack *extack); 3924 int dev_set_alias(struct net_device *, const char *, size_t); 3925 int dev_get_alias(const struct net_device *, char *, size_t); 3926 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 3927 const char *pat, int new_ifindex); 3928 static inline 3929 int dev_change_net_namespace(struct net_device *dev, struct net *net, 3930 const char *pat) 3931 { 3932 return __dev_change_net_namespace(dev, net, pat, 0); 3933 } 3934 int __dev_set_mtu(struct net_device *, int); 3935 int dev_set_mtu(struct net_device *, int); 3936 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3937 struct netlink_ext_ack *extack); 3938 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3939 struct netlink_ext_ack *extack); 3940 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 3941 struct netlink_ext_ack *extack); 3942 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 3943 int dev_get_port_parent_id(struct net_device *dev, 3944 struct netdev_phys_item_id *ppid, bool recurse); 3945 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3946 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3947 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3948 struct netdev_queue *txq, int *ret); 3949 3950 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3951 u8 dev_xdp_prog_count(struct net_device *dev); 3952 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3953 3954 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3955 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3956 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 3957 bool is_skb_forwardable(const struct net_device *dev, 3958 const struct sk_buff *skb); 3959 3960 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 3961 const struct sk_buff *skb, 3962 const bool check_mtu) 3963 { 3964 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 3965 unsigned int len; 3966 3967 if (!(dev->flags & IFF_UP)) 3968 return false; 3969 3970 if (!check_mtu) 3971 return true; 3972 3973 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 3974 if (skb->len <= len) 3975 return true; 3976 3977 /* if TSO is enabled, we don't care about the length as the packet 3978 * could be forwarded without being segmented before 3979 */ 3980 if (skb_is_gso(skb)) 3981 return true; 3982 3983 return false; 3984 } 3985 3986 struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev); 3987 3988 static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev) 3989 { 3990 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */ 3991 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats); 3992 3993 if (likely(p)) 3994 return p; 3995 3996 return netdev_core_stats_alloc(dev); 3997 } 3998 3999 #define DEV_CORE_STATS_INC(FIELD) \ 4000 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 4001 { \ 4002 struct net_device_core_stats __percpu *p; \ 4003 \ 4004 p = dev_core_stats(dev); \ 4005 if (p) \ 4006 this_cpu_inc(p->FIELD); \ 4007 } 4008 DEV_CORE_STATS_INC(rx_dropped) 4009 DEV_CORE_STATS_INC(tx_dropped) 4010 DEV_CORE_STATS_INC(rx_nohandler) 4011 DEV_CORE_STATS_INC(rx_otherhost_dropped) 4012 4013 static __always_inline int ____dev_forward_skb(struct net_device *dev, 4014 struct sk_buff *skb, 4015 const bool check_mtu) 4016 { 4017 if (skb_orphan_frags(skb, GFP_ATOMIC) || 4018 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 4019 dev_core_stats_rx_dropped_inc(dev); 4020 kfree_skb(skb); 4021 return NET_RX_DROP; 4022 } 4023 4024 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 4025 skb->priority = 0; 4026 return 0; 4027 } 4028 4029 bool dev_nit_active(struct net_device *dev); 4030 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 4031 4032 static inline void __dev_put(struct net_device *dev) 4033 { 4034 if (dev) { 4035 #ifdef CONFIG_PCPU_DEV_REFCNT 4036 this_cpu_dec(*dev->pcpu_refcnt); 4037 #else 4038 refcount_dec(&dev->dev_refcnt); 4039 #endif 4040 } 4041 } 4042 4043 static inline void __dev_hold(struct net_device *dev) 4044 { 4045 if (dev) { 4046 #ifdef CONFIG_PCPU_DEV_REFCNT 4047 this_cpu_inc(*dev->pcpu_refcnt); 4048 #else 4049 refcount_inc(&dev->dev_refcnt); 4050 #endif 4051 } 4052 } 4053 4054 static inline void __netdev_tracker_alloc(struct net_device *dev, 4055 netdevice_tracker *tracker, 4056 gfp_t gfp) 4057 { 4058 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4059 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 4060 #endif 4061 } 4062 4063 /* netdev_tracker_alloc() can upgrade a prior untracked reference 4064 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4065 */ 4066 static inline void netdev_tracker_alloc(struct net_device *dev, 4067 netdevice_tracker *tracker, gfp_t gfp) 4068 { 4069 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4070 refcount_dec(&dev->refcnt_tracker.no_tracker); 4071 __netdev_tracker_alloc(dev, tracker, gfp); 4072 #endif 4073 } 4074 4075 static inline void netdev_tracker_free(struct net_device *dev, 4076 netdevice_tracker *tracker) 4077 { 4078 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4079 ref_tracker_free(&dev->refcnt_tracker, tracker); 4080 #endif 4081 } 4082 4083 static inline void netdev_hold(struct net_device *dev, 4084 netdevice_tracker *tracker, gfp_t gfp) 4085 { 4086 if (dev) { 4087 __dev_hold(dev); 4088 __netdev_tracker_alloc(dev, tracker, gfp); 4089 } 4090 } 4091 4092 static inline void netdev_put(struct net_device *dev, 4093 netdevice_tracker *tracker) 4094 { 4095 if (dev) { 4096 netdev_tracker_free(dev, tracker); 4097 __dev_put(dev); 4098 } 4099 } 4100 4101 /** 4102 * dev_hold - get reference to device 4103 * @dev: network device 4104 * 4105 * Hold reference to device to keep it from being freed. 4106 * Try using netdev_hold() instead. 4107 */ 4108 static inline void dev_hold(struct net_device *dev) 4109 { 4110 netdev_hold(dev, NULL, GFP_ATOMIC); 4111 } 4112 4113 /** 4114 * dev_put - release reference to device 4115 * @dev: network device 4116 * 4117 * Release reference to device to allow it to be freed. 4118 * Try using netdev_put() instead. 4119 */ 4120 static inline void dev_put(struct net_device *dev) 4121 { 4122 netdev_put(dev, NULL); 4123 } 4124 4125 static inline void netdev_ref_replace(struct net_device *odev, 4126 struct net_device *ndev, 4127 netdevice_tracker *tracker, 4128 gfp_t gfp) 4129 { 4130 if (odev) 4131 netdev_tracker_free(odev, tracker); 4132 4133 __dev_hold(ndev); 4134 __dev_put(odev); 4135 4136 if (ndev) 4137 __netdev_tracker_alloc(ndev, tracker, gfp); 4138 } 4139 4140 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4141 * and _off may be called from IRQ context, but it is caller 4142 * who is responsible for serialization of these calls. 4143 * 4144 * The name carrier is inappropriate, these functions should really be 4145 * called netif_lowerlayer_*() because they represent the state of any 4146 * kind of lower layer not just hardware media. 4147 */ 4148 void linkwatch_fire_event(struct net_device *dev); 4149 4150 /** 4151 * netif_carrier_ok - test if carrier present 4152 * @dev: network device 4153 * 4154 * Check if carrier is present on device 4155 */ 4156 static inline bool netif_carrier_ok(const struct net_device *dev) 4157 { 4158 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4159 } 4160 4161 unsigned long dev_trans_start(struct net_device *dev); 4162 4163 void __netdev_watchdog_up(struct net_device *dev); 4164 4165 void netif_carrier_on(struct net_device *dev); 4166 void netif_carrier_off(struct net_device *dev); 4167 void netif_carrier_event(struct net_device *dev); 4168 4169 /** 4170 * netif_dormant_on - mark device as dormant. 4171 * @dev: network device 4172 * 4173 * Mark device as dormant (as per RFC2863). 4174 * 4175 * The dormant state indicates that the relevant interface is not 4176 * actually in a condition to pass packets (i.e., it is not 'up') but is 4177 * in a "pending" state, waiting for some external event. For "on- 4178 * demand" interfaces, this new state identifies the situation where the 4179 * interface is waiting for events to place it in the up state. 4180 */ 4181 static inline void netif_dormant_on(struct net_device *dev) 4182 { 4183 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4184 linkwatch_fire_event(dev); 4185 } 4186 4187 /** 4188 * netif_dormant_off - set device as not dormant. 4189 * @dev: network device 4190 * 4191 * Device is not in dormant state. 4192 */ 4193 static inline void netif_dormant_off(struct net_device *dev) 4194 { 4195 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4196 linkwatch_fire_event(dev); 4197 } 4198 4199 /** 4200 * netif_dormant - test if device is dormant 4201 * @dev: network device 4202 * 4203 * Check if device is dormant. 4204 */ 4205 static inline bool netif_dormant(const struct net_device *dev) 4206 { 4207 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4208 } 4209 4210 4211 /** 4212 * netif_testing_on - mark device as under test. 4213 * @dev: network device 4214 * 4215 * Mark device as under test (as per RFC2863). 4216 * 4217 * The testing state indicates that some test(s) must be performed on 4218 * the interface. After completion, of the test, the interface state 4219 * will change to up, dormant, or down, as appropriate. 4220 */ 4221 static inline void netif_testing_on(struct net_device *dev) 4222 { 4223 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4224 linkwatch_fire_event(dev); 4225 } 4226 4227 /** 4228 * netif_testing_off - set device as not under test. 4229 * @dev: network device 4230 * 4231 * Device is not in testing state. 4232 */ 4233 static inline void netif_testing_off(struct net_device *dev) 4234 { 4235 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4236 linkwatch_fire_event(dev); 4237 } 4238 4239 /** 4240 * netif_testing - test if device is under test 4241 * @dev: network device 4242 * 4243 * Check if device is under test 4244 */ 4245 static inline bool netif_testing(const struct net_device *dev) 4246 { 4247 return test_bit(__LINK_STATE_TESTING, &dev->state); 4248 } 4249 4250 4251 /** 4252 * netif_oper_up - test if device is operational 4253 * @dev: network device 4254 * 4255 * Check if carrier is operational 4256 */ 4257 static inline bool netif_oper_up(const struct net_device *dev) 4258 { 4259 return (dev->operstate == IF_OPER_UP || 4260 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4261 } 4262 4263 /** 4264 * netif_device_present - is device available or removed 4265 * @dev: network device 4266 * 4267 * Check if device has not been removed from system. 4268 */ 4269 static inline bool netif_device_present(const struct net_device *dev) 4270 { 4271 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4272 } 4273 4274 void netif_device_detach(struct net_device *dev); 4275 4276 void netif_device_attach(struct net_device *dev); 4277 4278 /* 4279 * Network interface message level settings 4280 */ 4281 4282 enum { 4283 NETIF_MSG_DRV_BIT, 4284 NETIF_MSG_PROBE_BIT, 4285 NETIF_MSG_LINK_BIT, 4286 NETIF_MSG_TIMER_BIT, 4287 NETIF_MSG_IFDOWN_BIT, 4288 NETIF_MSG_IFUP_BIT, 4289 NETIF_MSG_RX_ERR_BIT, 4290 NETIF_MSG_TX_ERR_BIT, 4291 NETIF_MSG_TX_QUEUED_BIT, 4292 NETIF_MSG_INTR_BIT, 4293 NETIF_MSG_TX_DONE_BIT, 4294 NETIF_MSG_RX_STATUS_BIT, 4295 NETIF_MSG_PKTDATA_BIT, 4296 NETIF_MSG_HW_BIT, 4297 NETIF_MSG_WOL_BIT, 4298 4299 /* When you add a new bit above, update netif_msg_class_names array 4300 * in net/ethtool/common.c 4301 */ 4302 NETIF_MSG_CLASS_COUNT, 4303 }; 4304 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4305 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4306 4307 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4308 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4309 4310 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4311 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4312 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4313 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4314 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4315 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4316 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4317 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4318 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4319 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4320 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4321 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4322 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4323 #define NETIF_MSG_HW __NETIF_MSG(HW) 4324 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4325 4326 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4327 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4328 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4329 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4330 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4331 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4332 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4333 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4334 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4335 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4336 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4337 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4338 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4339 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4340 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4341 4342 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4343 { 4344 /* use default */ 4345 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4346 return default_msg_enable_bits; 4347 if (debug_value == 0) /* no output */ 4348 return 0; 4349 /* set low N bits */ 4350 return (1U << debug_value) - 1; 4351 } 4352 4353 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4354 { 4355 spin_lock(&txq->_xmit_lock); 4356 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4357 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4358 } 4359 4360 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4361 { 4362 __acquire(&txq->_xmit_lock); 4363 return true; 4364 } 4365 4366 static inline void __netif_tx_release(struct netdev_queue *txq) 4367 { 4368 __release(&txq->_xmit_lock); 4369 } 4370 4371 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4372 { 4373 spin_lock_bh(&txq->_xmit_lock); 4374 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4375 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4376 } 4377 4378 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4379 { 4380 bool ok = spin_trylock(&txq->_xmit_lock); 4381 4382 if (likely(ok)) { 4383 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4384 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4385 } 4386 return ok; 4387 } 4388 4389 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4390 { 4391 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4392 WRITE_ONCE(txq->xmit_lock_owner, -1); 4393 spin_unlock(&txq->_xmit_lock); 4394 } 4395 4396 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4397 { 4398 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4399 WRITE_ONCE(txq->xmit_lock_owner, -1); 4400 spin_unlock_bh(&txq->_xmit_lock); 4401 } 4402 4403 /* 4404 * txq->trans_start can be read locklessly from dev_watchdog() 4405 */ 4406 static inline void txq_trans_update(struct netdev_queue *txq) 4407 { 4408 if (txq->xmit_lock_owner != -1) 4409 WRITE_ONCE(txq->trans_start, jiffies); 4410 } 4411 4412 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4413 { 4414 unsigned long now = jiffies; 4415 4416 if (READ_ONCE(txq->trans_start) != now) 4417 WRITE_ONCE(txq->trans_start, now); 4418 } 4419 4420 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4421 static inline void netif_trans_update(struct net_device *dev) 4422 { 4423 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4424 4425 txq_trans_cond_update(txq); 4426 } 4427 4428 /** 4429 * netif_tx_lock - grab network device transmit lock 4430 * @dev: network device 4431 * 4432 * Get network device transmit lock 4433 */ 4434 void netif_tx_lock(struct net_device *dev); 4435 4436 static inline void netif_tx_lock_bh(struct net_device *dev) 4437 { 4438 local_bh_disable(); 4439 netif_tx_lock(dev); 4440 } 4441 4442 void netif_tx_unlock(struct net_device *dev); 4443 4444 static inline void netif_tx_unlock_bh(struct net_device *dev) 4445 { 4446 netif_tx_unlock(dev); 4447 local_bh_enable(); 4448 } 4449 4450 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4451 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4452 __netif_tx_lock(txq, cpu); \ 4453 } else { \ 4454 __netif_tx_acquire(txq); \ 4455 } \ 4456 } 4457 4458 #define HARD_TX_TRYLOCK(dev, txq) \ 4459 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4460 __netif_tx_trylock(txq) : \ 4461 __netif_tx_acquire(txq)) 4462 4463 #define HARD_TX_UNLOCK(dev, txq) { \ 4464 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4465 __netif_tx_unlock(txq); \ 4466 } else { \ 4467 __netif_tx_release(txq); \ 4468 } \ 4469 } 4470 4471 static inline void netif_tx_disable(struct net_device *dev) 4472 { 4473 unsigned int i; 4474 int cpu; 4475 4476 local_bh_disable(); 4477 cpu = smp_processor_id(); 4478 spin_lock(&dev->tx_global_lock); 4479 for (i = 0; i < dev->num_tx_queues; i++) { 4480 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4481 4482 __netif_tx_lock(txq, cpu); 4483 netif_tx_stop_queue(txq); 4484 __netif_tx_unlock(txq); 4485 } 4486 spin_unlock(&dev->tx_global_lock); 4487 local_bh_enable(); 4488 } 4489 4490 static inline void netif_addr_lock(struct net_device *dev) 4491 { 4492 unsigned char nest_level = 0; 4493 4494 #ifdef CONFIG_LOCKDEP 4495 nest_level = dev->nested_level; 4496 #endif 4497 spin_lock_nested(&dev->addr_list_lock, nest_level); 4498 } 4499 4500 static inline void netif_addr_lock_bh(struct net_device *dev) 4501 { 4502 unsigned char nest_level = 0; 4503 4504 #ifdef CONFIG_LOCKDEP 4505 nest_level = dev->nested_level; 4506 #endif 4507 local_bh_disable(); 4508 spin_lock_nested(&dev->addr_list_lock, nest_level); 4509 } 4510 4511 static inline void netif_addr_unlock(struct net_device *dev) 4512 { 4513 spin_unlock(&dev->addr_list_lock); 4514 } 4515 4516 static inline void netif_addr_unlock_bh(struct net_device *dev) 4517 { 4518 spin_unlock_bh(&dev->addr_list_lock); 4519 } 4520 4521 /* 4522 * dev_addrs walker. Should be used only for read access. Call with 4523 * rcu_read_lock held. 4524 */ 4525 #define for_each_dev_addr(dev, ha) \ 4526 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4527 4528 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4529 4530 void ether_setup(struct net_device *dev); 4531 4532 /* Support for loadable net-drivers */ 4533 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4534 unsigned char name_assign_type, 4535 void (*setup)(struct net_device *), 4536 unsigned int txqs, unsigned int rxqs); 4537 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4538 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4539 4540 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4541 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4542 count) 4543 4544 int register_netdev(struct net_device *dev); 4545 void unregister_netdev(struct net_device *dev); 4546 4547 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4548 4549 /* General hardware address lists handling functions */ 4550 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4551 struct netdev_hw_addr_list *from_list, int addr_len); 4552 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4553 struct netdev_hw_addr_list *from_list, int addr_len); 4554 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4555 struct net_device *dev, 4556 int (*sync)(struct net_device *, const unsigned char *), 4557 int (*unsync)(struct net_device *, 4558 const unsigned char *)); 4559 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4560 struct net_device *dev, 4561 int (*sync)(struct net_device *, 4562 const unsigned char *, int), 4563 int (*unsync)(struct net_device *, 4564 const unsigned char *, int)); 4565 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4566 struct net_device *dev, 4567 int (*unsync)(struct net_device *, 4568 const unsigned char *, int)); 4569 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4570 struct net_device *dev, 4571 int (*unsync)(struct net_device *, 4572 const unsigned char *)); 4573 void __hw_addr_init(struct netdev_hw_addr_list *list); 4574 4575 /* Functions used for device addresses handling */ 4576 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4577 const void *addr, size_t len); 4578 4579 static inline void 4580 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4581 { 4582 dev_addr_mod(dev, 0, addr, len); 4583 } 4584 4585 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4586 { 4587 __dev_addr_set(dev, addr, dev->addr_len); 4588 } 4589 4590 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4591 unsigned char addr_type); 4592 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4593 unsigned char addr_type); 4594 4595 /* Functions used for unicast addresses handling */ 4596 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4597 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4598 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4599 int dev_uc_sync(struct net_device *to, struct net_device *from); 4600 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4601 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4602 void dev_uc_flush(struct net_device *dev); 4603 void dev_uc_init(struct net_device *dev); 4604 4605 /** 4606 * __dev_uc_sync - Synchonize device's unicast list 4607 * @dev: device to sync 4608 * @sync: function to call if address should be added 4609 * @unsync: function to call if address should be removed 4610 * 4611 * Add newly added addresses to the interface, and release 4612 * addresses that have been deleted. 4613 */ 4614 static inline int __dev_uc_sync(struct net_device *dev, 4615 int (*sync)(struct net_device *, 4616 const unsigned char *), 4617 int (*unsync)(struct net_device *, 4618 const unsigned char *)) 4619 { 4620 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4621 } 4622 4623 /** 4624 * __dev_uc_unsync - Remove synchronized addresses from device 4625 * @dev: device to sync 4626 * @unsync: function to call if address should be removed 4627 * 4628 * Remove all addresses that were added to the device by dev_uc_sync(). 4629 */ 4630 static inline void __dev_uc_unsync(struct net_device *dev, 4631 int (*unsync)(struct net_device *, 4632 const unsigned char *)) 4633 { 4634 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4635 } 4636 4637 /* Functions used for multicast addresses handling */ 4638 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4639 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4640 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4641 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4642 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4643 int dev_mc_sync(struct net_device *to, struct net_device *from); 4644 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4645 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4646 void dev_mc_flush(struct net_device *dev); 4647 void dev_mc_init(struct net_device *dev); 4648 4649 /** 4650 * __dev_mc_sync - Synchonize device's multicast list 4651 * @dev: device to sync 4652 * @sync: function to call if address should be added 4653 * @unsync: function to call if address should be removed 4654 * 4655 * Add newly added addresses to the interface, and release 4656 * addresses that have been deleted. 4657 */ 4658 static inline int __dev_mc_sync(struct net_device *dev, 4659 int (*sync)(struct net_device *, 4660 const unsigned char *), 4661 int (*unsync)(struct net_device *, 4662 const unsigned char *)) 4663 { 4664 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4665 } 4666 4667 /** 4668 * __dev_mc_unsync - Remove synchronized addresses from device 4669 * @dev: device to sync 4670 * @unsync: function to call if address should be removed 4671 * 4672 * Remove all addresses that were added to the device by dev_mc_sync(). 4673 */ 4674 static inline void __dev_mc_unsync(struct net_device *dev, 4675 int (*unsync)(struct net_device *, 4676 const unsigned char *)) 4677 { 4678 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4679 } 4680 4681 /* Functions used for secondary unicast and multicast support */ 4682 void dev_set_rx_mode(struct net_device *dev); 4683 int dev_set_promiscuity(struct net_device *dev, int inc); 4684 int dev_set_allmulti(struct net_device *dev, int inc); 4685 void netdev_state_change(struct net_device *dev); 4686 void __netdev_notify_peers(struct net_device *dev); 4687 void netdev_notify_peers(struct net_device *dev); 4688 void netdev_features_change(struct net_device *dev); 4689 /* Load a device via the kmod */ 4690 void dev_load(struct net *net, const char *name); 4691 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4692 struct rtnl_link_stats64 *storage); 4693 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4694 const struct net_device_stats *netdev_stats); 4695 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4696 const struct pcpu_sw_netstats __percpu *netstats); 4697 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4698 4699 extern int netdev_max_backlog; 4700 extern int dev_rx_weight; 4701 extern int dev_tx_weight; 4702 extern int gro_normal_batch; 4703 4704 enum { 4705 NESTED_SYNC_IMM_BIT, 4706 NESTED_SYNC_TODO_BIT, 4707 }; 4708 4709 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4710 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4711 4712 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4713 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4714 4715 struct netdev_nested_priv { 4716 unsigned char flags; 4717 void *data; 4718 }; 4719 4720 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4721 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4722 struct list_head **iter); 4723 4724 /* iterate through upper list, must be called under RCU read lock */ 4725 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4726 for (iter = &(dev)->adj_list.upper, \ 4727 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4728 updev; \ 4729 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4730 4731 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4732 int (*fn)(struct net_device *upper_dev, 4733 struct netdev_nested_priv *priv), 4734 struct netdev_nested_priv *priv); 4735 4736 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4737 struct net_device *upper_dev); 4738 4739 bool netdev_has_any_upper_dev(struct net_device *dev); 4740 4741 void *netdev_lower_get_next_private(struct net_device *dev, 4742 struct list_head **iter); 4743 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4744 struct list_head **iter); 4745 4746 #define netdev_for_each_lower_private(dev, priv, iter) \ 4747 for (iter = (dev)->adj_list.lower.next, \ 4748 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4749 priv; \ 4750 priv = netdev_lower_get_next_private(dev, &(iter))) 4751 4752 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4753 for (iter = &(dev)->adj_list.lower, \ 4754 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4755 priv; \ 4756 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4757 4758 void *netdev_lower_get_next(struct net_device *dev, 4759 struct list_head **iter); 4760 4761 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4762 for (iter = (dev)->adj_list.lower.next, \ 4763 ldev = netdev_lower_get_next(dev, &(iter)); \ 4764 ldev; \ 4765 ldev = netdev_lower_get_next(dev, &(iter))) 4766 4767 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4768 struct list_head **iter); 4769 int netdev_walk_all_lower_dev(struct net_device *dev, 4770 int (*fn)(struct net_device *lower_dev, 4771 struct netdev_nested_priv *priv), 4772 struct netdev_nested_priv *priv); 4773 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4774 int (*fn)(struct net_device *lower_dev, 4775 struct netdev_nested_priv *priv), 4776 struct netdev_nested_priv *priv); 4777 4778 void *netdev_adjacent_get_private(struct list_head *adj_list); 4779 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4780 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4781 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4782 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4783 struct netlink_ext_ack *extack); 4784 int netdev_master_upper_dev_link(struct net_device *dev, 4785 struct net_device *upper_dev, 4786 void *upper_priv, void *upper_info, 4787 struct netlink_ext_ack *extack); 4788 void netdev_upper_dev_unlink(struct net_device *dev, 4789 struct net_device *upper_dev); 4790 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4791 struct net_device *new_dev, 4792 struct net_device *dev, 4793 struct netlink_ext_ack *extack); 4794 void netdev_adjacent_change_commit(struct net_device *old_dev, 4795 struct net_device *new_dev, 4796 struct net_device *dev); 4797 void netdev_adjacent_change_abort(struct net_device *old_dev, 4798 struct net_device *new_dev, 4799 struct net_device *dev); 4800 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4801 void *netdev_lower_dev_get_private(struct net_device *dev, 4802 struct net_device *lower_dev); 4803 void netdev_lower_state_changed(struct net_device *lower_dev, 4804 void *lower_state_info); 4805 4806 /* RSS keys are 40 or 52 bytes long */ 4807 #define NETDEV_RSS_KEY_LEN 52 4808 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4809 void netdev_rss_key_fill(void *buffer, size_t len); 4810 4811 int skb_checksum_help(struct sk_buff *skb); 4812 int skb_crc32c_csum_help(struct sk_buff *skb); 4813 int skb_csum_hwoffload_help(struct sk_buff *skb, 4814 const netdev_features_t features); 4815 4816 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4817 netdev_features_t features, bool tx_path); 4818 struct sk_buff *skb_eth_gso_segment(struct sk_buff *skb, 4819 netdev_features_t features, __be16 type); 4820 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4821 netdev_features_t features); 4822 4823 struct netdev_bonding_info { 4824 ifslave slave; 4825 ifbond master; 4826 }; 4827 4828 struct netdev_notifier_bonding_info { 4829 struct netdev_notifier_info info; /* must be first */ 4830 struct netdev_bonding_info bonding_info; 4831 }; 4832 4833 void netdev_bonding_info_change(struct net_device *dev, 4834 struct netdev_bonding_info *bonding_info); 4835 4836 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4837 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4838 #else 4839 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4840 const void *data) 4841 { 4842 } 4843 #endif 4844 4845 static inline 4846 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4847 { 4848 return __skb_gso_segment(skb, features, true); 4849 } 4850 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4851 4852 static inline bool can_checksum_protocol(netdev_features_t features, 4853 __be16 protocol) 4854 { 4855 if (protocol == htons(ETH_P_FCOE)) 4856 return !!(features & NETIF_F_FCOE_CRC); 4857 4858 /* Assume this is an IP checksum (not SCTP CRC) */ 4859 4860 if (features & NETIF_F_HW_CSUM) { 4861 /* Can checksum everything */ 4862 return true; 4863 } 4864 4865 switch (protocol) { 4866 case htons(ETH_P_IP): 4867 return !!(features & NETIF_F_IP_CSUM); 4868 case htons(ETH_P_IPV6): 4869 return !!(features & NETIF_F_IPV6_CSUM); 4870 default: 4871 return false; 4872 } 4873 } 4874 4875 #ifdef CONFIG_BUG 4876 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4877 #else 4878 static inline void netdev_rx_csum_fault(struct net_device *dev, 4879 struct sk_buff *skb) 4880 { 4881 } 4882 #endif 4883 /* rx skb timestamps */ 4884 void net_enable_timestamp(void); 4885 void net_disable_timestamp(void); 4886 4887 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 4888 const struct skb_shared_hwtstamps *hwtstamps, 4889 bool cycles) 4890 { 4891 const struct net_device_ops *ops = dev->netdev_ops; 4892 4893 if (ops->ndo_get_tstamp) 4894 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 4895 4896 return hwtstamps->hwtstamp; 4897 } 4898 4899 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4900 struct sk_buff *skb, struct net_device *dev, 4901 bool more) 4902 { 4903 __this_cpu_write(softnet_data.xmit.more, more); 4904 return ops->ndo_start_xmit(skb, dev); 4905 } 4906 4907 static inline bool netdev_xmit_more(void) 4908 { 4909 return __this_cpu_read(softnet_data.xmit.more); 4910 } 4911 4912 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4913 struct netdev_queue *txq, bool more) 4914 { 4915 const struct net_device_ops *ops = dev->netdev_ops; 4916 netdev_tx_t rc; 4917 4918 rc = __netdev_start_xmit(ops, skb, dev, more); 4919 if (rc == NETDEV_TX_OK) 4920 txq_trans_update(txq); 4921 4922 return rc; 4923 } 4924 4925 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4926 const void *ns); 4927 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4928 const void *ns); 4929 4930 extern const struct kobj_ns_type_operations net_ns_type_operations; 4931 4932 const char *netdev_drivername(const struct net_device *dev); 4933 4934 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4935 netdev_features_t f2) 4936 { 4937 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4938 if (f1 & NETIF_F_HW_CSUM) 4939 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4940 else 4941 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4942 } 4943 4944 return f1 & f2; 4945 } 4946 4947 static inline netdev_features_t netdev_get_wanted_features( 4948 struct net_device *dev) 4949 { 4950 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4951 } 4952 netdev_features_t netdev_increment_features(netdev_features_t all, 4953 netdev_features_t one, netdev_features_t mask); 4954 4955 /* Allow TSO being used on stacked device : 4956 * Performing the GSO segmentation before last device 4957 * is a performance improvement. 4958 */ 4959 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4960 netdev_features_t mask) 4961 { 4962 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4963 } 4964 4965 int __netdev_update_features(struct net_device *dev); 4966 void netdev_update_features(struct net_device *dev); 4967 void netdev_change_features(struct net_device *dev); 4968 4969 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4970 struct net_device *dev); 4971 4972 netdev_features_t passthru_features_check(struct sk_buff *skb, 4973 struct net_device *dev, 4974 netdev_features_t features); 4975 netdev_features_t netif_skb_features(struct sk_buff *skb); 4976 4977 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4978 { 4979 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4980 4981 /* check flags correspondence */ 4982 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4983 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4984 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4985 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4986 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4987 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4988 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4989 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4990 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4991 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4992 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4993 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4994 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4995 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4996 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4997 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4998 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4999 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 5000 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 5001 5002 return (features & feature) == feature; 5003 } 5004 5005 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 5006 { 5007 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 5008 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 5009 } 5010 5011 static inline bool netif_needs_gso(struct sk_buff *skb, 5012 netdev_features_t features) 5013 { 5014 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 5015 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 5016 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 5017 } 5018 5019 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 5020 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 5021 void netif_inherit_tso_max(struct net_device *to, 5022 const struct net_device *from); 5023 5024 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 5025 int pulled_hlen, u16 mac_offset, 5026 int mac_len) 5027 { 5028 skb->protocol = protocol; 5029 skb->encapsulation = 1; 5030 skb_push(skb, pulled_hlen); 5031 skb_reset_transport_header(skb); 5032 skb->mac_header = mac_offset; 5033 skb->network_header = skb->mac_header + mac_len; 5034 skb->mac_len = mac_len; 5035 } 5036 5037 static inline bool netif_is_macsec(const struct net_device *dev) 5038 { 5039 return dev->priv_flags & IFF_MACSEC; 5040 } 5041 5042 static inline bool netif_is_macvlan(const struct net_device *dev) 5043 { 5044 return dev->priv_flags & IFF_MACVLAN; 5045 } 5046 5047 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5048 { 5049 return dev->priv_flags & IFF_MACVLAN_PORT; 5050 } 5051 5052 static inline bool netif_is_bond_master(const struct net_device *dev) 5053 { 5054 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5055 } 5056 5057 static inline bool netif_is_bond_slave(const struct net_device *dev) 5058 { 5059 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5060 } 5061 5062 static inline bool netif_supports_nofcs(struct net_device *dev) 5063 { 5064 return dev->priv_flags & IFF_SUPP_NOFCS; 5065 } 5066 5067 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5068 { 5069 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5070 } 5071 5072 static inline bool netif_is_l3_master(const struct net_device *dev) 5073 { 5074 return dev->priv_flags & IFF_L3MDEV_MASTER; 5075 } 5076 5077 static inline bool netif_is_l3_slave(const struct net_device *dev) 5078 { 5079 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5080 } 5081 5082 static inline bool netif_is_bridge_master(const struct net_device *dev) 5083 { 5084 return dev->priv_flags & IFF_EBRIDGE; 5085 } 5086 5087 static inline bool netif_is_bridge_port(const struct net_device *dev) 5088 { 5089 return dev->priv_flags & IFF_BRIDGE_PORT; 5090 } 5091 5092 static inline bool netif_is_ovs_master(const struct net_device *dev) 5093 { 5094 return dev->priv_flags & IFF_OPENVSWITCH; 5095 } 5096 5097 static inline bool netif_is_ovs_port(const struct net_device *dev) 5098 { 5099 return dev->priv_flags & IFF_OVS_DATAPATH; 5100 } 5101 5102 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5103 { 5104 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5105 } 5106 5107 static inline bool netif_is_team_master(const struct net_device *dev) 5108 { 5109 return dev->priv_flags & IFF_TEAM; 5110 } 5111 5112 static inline bool netif_is_team_port(const struct net_device *dev) 5113 { 5114 return dev->priv_flags & IFF_TEAM_PORT; 5115 } 5116 5117 static inline bool netif_is_lag_master(const struct net_device *dev) 5118 { 5119 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5120 } 5121 5122 static inline bool netif_is_lag_port(const struct net_device *dev) 5123 { 5124 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5125 } 5126 5127 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5128 { 5129 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5130 } 5131 5132 static inline bool netif_is_failover(const struct net_device *dev) 5133 { 5134 return dev->priv_flags & IFF_FAILOVER; 5135 } 5136 5137 static inline bool netif_is_failover_slave(const struct net_device *dev) 5138 { 5139 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5140 } 5141 5142 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5143 static inline void netif_keep_dst(struct net_device *dev) 5144 { 5145 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5146 } 5147 5148 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5149 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5150 { 5151 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5152 return netif_is_macsec(dev); 5153 } 5154 5155 extern struct pernet_operations __net_initdata loopback_net_ops; 5156 5157 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5158 5159 /* netdev_printk helpers, similar to dev_printk */ 5160 5161 static inline const char *netdev_name(const struct net_device *dev) 5162 { 5163 if (!dev->name[0] || strchr(dev->name, '%')) 5164 return "(unnamed net_device)"; 5165 return dev->name; 5166 } 5167 5168 static inline const char *netdev_reg_state(const struct net_device *dev) 5169 { 5170 switch (dev->reg_state) { 5171 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5172 case NETREG_REGISTERED: return ""; 5173 case NETREG_UNREGISTERING: return " (unregistering)"; 5174 case NETREG_UNREGISTERED: return " (unregistered)"; 5175 case NETREG_RELEASED: return " (released)"; 5176 case NETREG_DUMMY: return " (dummy)"; 5177 } 5178 5179 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5180 return " (unknown)"; 5181 } 5182 5183 #define MODULE_ALIAS_NETDEV(device) \ 5184 MODULE_ALIAS("netdev-" device) 5185 5186 /* 5187 * netdev_WARN() acts like dev_printk(), but with the key difference 5188 * of using a WARN/WARN_ON to get the message out, including the 5189 * file/line information and a backtrace. 5190 */ 5191 #define netdev_WARN(dev, format, args...) \ 5192 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5193 netdev_reg_state(dev), ##args) 5194 5195 #define netdev_WARN_ONCE(dev, format, args...) \ 5196 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5197 netdev_reg_state(dev), ##args) 5198 5199 /* 5200 * The list of packet types we will receive (as opposed to discard) 5201 * and the routines to invoke. 5202 * 5203 * Why 16. Because with 16 the only overlap we get on a hash of the 5204 * low nibble of the protocol value is RARP/SNAP/X.25. 5205 * 5206 * 0800 IP 5207 * 0001 802.3 5208 * 0002 AX.25 5209 * 0004 802.2 5210 * 8035 RARP 5211 * 0005 SNAP 5212 * 0805 X.25 5213 * 0806 ARP 5214 * 8137 IPX 5215 * 0009 Localtalk 5216 * 86DD IPv6 5217 */ 5218 #define PTYPE_HASH_SIZE (16) 5219 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5220 5221 extern struct list_head ptype_all __read_mostly; 5222 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5223 5224 extern struct net_device *blackhole_netdev; 5225 5226 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */ 5227 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD) 5228 #define DEV_STATS_ADD(DEV, FIELD, VAL) \ 5229 atomic_long_add((VAL), &(DEV)->stats.__##FIELD) 5230 5231 #endif /* _LINUX_NETDEVICE_H */ 5232