1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <[email protected]> 13 * Corey Minyard <[email protected]> 14 * Donald J. Becker, <[email protected]> 15 * Alan Cox, <[email protected]> 16 * Bjorn Ekwall. <[email protected]> 17 * Pekka Riikonen <[email protected]> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 #include <asm/local.h> 32 33 #include <linux/percpu.h> 34 #include <linux/rculist.h> 35 #include <linux/workqueue.h> 36 #include <linux/dynamic_queue_limits.h> 37 38 #include <net/net_namespace.h> 39 #ifdef CONFIG_DCB 40 #include <net/dcbnl.h> 41 #endif 42 #include <net/netprio_cgroup.h> 43 #include <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 <linux/hashtable.h> 51 #include <linux/rbtree.h> 52 #include <net/net_trackers.h> 53 54 struct netpoll_info; 55 struct device; 56 struct ethtool_ops; 57 struct phy_device; 58 struct dsa_port; 59 struct ip_tunnel_parm; 60 struct macsec_context; 61 struct macsec_ops; 62 struct netdev_name_node; 63 struct sd_flow_limit; 64 struct sfp_bus; 65 /* 802.11 specific */ 66 struct wireless_dev; 67 /* 802.15.4 specific */ 68 struct wpan_dev; 69 struct mpls_dev; 70 /* UDP Tunnel offloads */ 71 struct udp_tunnel_info; 72 struct udp_tunnel_nic_info; 73 struct udp_tunnel_nic; 74 struct bpf_prog; 75 struct xdp_buff; 76 77 void synchronize_net(void); 78 void netdev_set_default_ethtool_ops(struct net_device *dev, 79 const struct ethtool_ops *ops); 80 81 /* Backlog congestion levels */ 82 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 83 #define NET_RX_DROP 1 /* packet dropped */ 84 85 #define MAX_NEST_DEV 8 86 87 /* 88 * Transmit return codes: transmit return codes originate from three different 89 * namespaces: 90 * 91 * - qdisc return codes 92 * - driver transmit return codes 93 * - errno values 94 * 95 * Drivers are allowed to return any one of those in their hard_start_xmit() 96 * function. Real network devices commonly used with qdiscs should only return 97 * the driver transmit return codes though - when qdiscs are used, the actual 98 * transmission happens asynchronously, so the value is not propagated to 99 * higher layers. Virtual network devices transmit synchronously; in this case 100 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 101 * others are propagated to higher layers. 102 */ 103 104 /* qdisc ->enqueue() return codes. */ 105 #define NET_XMIT_SUCCESS 0x00 106 #define NET_XMIT_DROP 0x01 /* skb dropped */ 107 #define NET_XMIT_CN 0x02 /* congestion notification */ 108 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 109 110 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 111 * indicates that the device will soon be dropping packets, or already drops 112 * some packets of the same priority; prompting us to send less aggressively. */ 113 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 114 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 115 116 /* Driver transmit return codes */ 117 #define NETDEV_TX_MASK 0xf0 118 119 enum netdev_tx { 120 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 121 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 122 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 123 }; 124 typedef enum netdev_tx netdev_tx_t; 125 126 /* 127 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 128 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 129 */ 130 static inline bool dev_xmit_complete(int rc) 131 { 132 /* 133 * Positive cases with an skb consumed by a driver: 134 * - successful transmission (rc == NETDEV_TX_OK) 135 * - error while transmitting (rc < 0) 136 * - error while queueing to a different device (rc & NET_XMIT_MASK) 137 */ 138 if (likely(rc < NET_XMIT_MASK)) 139 return true; 140 141 return false; 142 } 143 144 /* 145 * Compute the worst-case header length according to the protocols 146 * used. 147 */ 148 149 #if defined(CONFIG_HYPERV_NET) 150 # define LL_MAX_HEADER 128 151 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 152 # if defined(CONFIG_MAC80211_MESH) 153 # define LL_MAX_HEADER 128 154 # else 155 # define LL_MAX_HEADER 96 156 # endif 157 #else 158 # define LL_MAX_HEADER 32 159 #endif 160 161 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 162 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 163 #define MAX_HEADER LL_MAX_HEADER 164 #else 165 #define MAX_HEADER (LL_MAX_HEADER + 48) 166 #endif 167 168 /* 169 * Old network device statistics. Fields are native words 170 * (unsigned long) so they can be read and written atomically. 171 */ 172 173 struct net_device_stats { 174 unsigned long rx_packets; 175 unsigned long tx_packets; 176 unsigned long rx_bytes; 177 unsigned long tx_bytes; 178 unsigned long rx_errors; 179 unsigned long tx_errors; 180 unsigned long rx_dropped; 181 unsigned long tx_dropped; 182 unsigned long multicast; 183 unsigned long collisions; 184 unsigned long rx_length_errors; 185 unsigned long rx_over_errors; 186 unsigned long rx_crc_errors; 187 unsigned long rx_frame_errors; 188 unsigned long rx_fifo_errors; 189 unsigned long rx_missed_errors; 190 unsigned long tx_aborted_errors; 191 unsigned long tx_carrier_errors; 192 unsigned long tx_fifo_errors; 193 unsigned long tx_heartbeat_errors; 194 unsigned long tx_window_errors; 195 unsigned long rx_compressed; 196 unsigned long tx_compressed; 197 }; 198 199 /* per-cpu stats, allocated on demand. 200 * Try to fit them in a single cache line, for dev_get_stats() sake. 201 */ 202 struct net_device_core_stats { 203 unsigned long rx_dropped; 204 unsigned long tx_dropped; 205 unsigned long rx_nohandler; 206 unsigned long rx_otherhost_dropped; 207 } __aligned(4 * sizeof(unsigned long)); 208 209 #include <linux/cache.h> 210 #include <linux/skbuff.h> 211 212 #ifdef CONFIG_RPS 213 #include <linux/static_key.h> 214 extern struct static_key_false rps_needed; 215 extern struct static_key_false rfs_needed; 216 #endif 217 218 struct neighbour; 219 struct neigh_parms; 220 struct sk_buff; 221 222 struct netdev_hw_addr { 223 struct list_head list; 224 struct rb_node node; 225 unsigned char addr[MAX_ADDR_LEN]; 226 unsigned char type; 227 #define NETDEV_HW_ADDR_T_LAN 1 228 #define NETDEV_HW_ADDR_T_SAN 2 229 #define NETDEV_HW_ADDR_T_UNICAST 3 230 #define NETDEV_HW_ADDR_T_MULTICAST 4 231 bool global_use; 232 int sync_cnt; 233 int refcount; 234 int synced; 235 struct rcu_head rcu_head; 236 }; 237 238 struct netdev_hw_addr_list { 239 struct list_head list; 240 int count; 241 242 /* Auxiliary tree for faster lookup on addition and deletion */ 243 struct rb_root tree; 244 }; 245 246 #define netdev_hw_addr_list_count(l) ((l)->count) 247 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 248 #define netdev_hw_addr_list_for_each(ha, l) \ 249 list_for_each_entry(ha, &(l)->list, list) 250 251 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 252 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 253 #define netdev_for_each_uc_addr(ha, dev) \ 254 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 255 256 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 257 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 258 #define netdev_for_each_mc_addr(ha, dev) \ 259 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 260 261 struct hh_cache { 262 unsigned int hh_len; 263 seqlock_t hh_lock; 264 265 /* cached hardware header; allow for machine alignment needs. */ 266 #define HH_DATA_MOD 16 267 #define HH_DATA_OFF(__len) \ 268 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 269 #define HH_DATA_ALIGN(__len) \ 270 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 271 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 272 }; 273 274 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 275 * Alternative is: 276 * dev->hard_header_len ? (dev->hard_header_len + 277 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 278 * 279 * We could use other alignment values, but we must maintain the 280 * relationship HH alignment <= LL alignment. 281 */ 282 #define LL_RESERVED_SPACE(dev) \ 283 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 284 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 285 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 286 287 struct header_ops { 288 int (*create) (struct sk_buff *skb, struct net_device *dev, 289 unsigned short type, const void *daddr, 290 const void *saddr, unsigned int len); 291 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 292 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 293 void (*cache_update)(struct hh_cache *hh, 294 const struct net_device *dev, 295 const unsigned char *haddr); 296 bool (*validate)(const char *ll_header, unsigned int len); 297 __be16 (*parse_protocol)(const struct sk_buff *skb); 298 }; 299 300 /* These flag bits are private to the generic network queueing 301 * layer; they may not be explicitly referenced by any other 302 * code. 303 */ 304 305 enum netdev_state_t { 306 __LINK_STATE_START, 307 __LINK_STATE_PRESENT, 308 __LINK_STATE_NOCARRIER, 309 __LINK_STATE_LINKWATCH_PENDING, 310 __LINK_STATE_DORMANT, 311 __LINK_STATE_TESTING, 312 }; 313 314 struct gro_list { 315 struct list_head list; 316 int count; 317 }; 318 319 /* 320 * size of gro hash buckets, must less than bit number of 321 * napi_struct::gro_bitmask 322 */ 323 #define GRO_HASH_BUCKETS 8 324 325 /* 326 * Structure for NAPI scheduling similar to tasklet but with weighting 327 */ 328 struct napi_struct { 329 /* The poll_list must only be managed by the entity which 330 * changes the state of the NAPI_STATE_SCHED bit. This means 331 * whoever atomically sets that bit can add this napi_struct 332 * to the per-CPU poll_list, and whoever clears that bit 333 * can remove from the list right before clearing the bit. 334 */ 335 struct list_head poll_list; 336 337 unsigned long state; 338 int weight; 339 int defer_hard_irqs_count; 340 unsigned long gro_bitmask; 341 int (*poll)(struct napi_struct *, int); 342 #ifdef CONFIG_NETPOLL 343 int poll_owner; 344 #endif 345 struct net_device *dev; 346 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 347 struct sk_buff *skb; 348 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 349 int rx_count; /* length of rx_list */ 350 struct hrtimer timer; 351 struct list_head dev_list; 352 struct hlist_node napi_hash_node; 353 unsigned int napi_id; 354 struct task_struct *thread; 355 }; 356 357 enum { 358 NAPI_STATE_SCHED, /* Poll is scheduled */ 359 NAPI_STATE_MISSED, /* reschedule a napi */ 360 NAPI_STATE_DISABLE, /* Disable pending */ 361 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 362 NAPI_STATE_LISTED, /* NAPI added to system lists */ 363 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 364 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 365 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 366 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 367 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 368 }; 369 370 enum { 371 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 372 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 373 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 374 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 375 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 376 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 377 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 378 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 379 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 380 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 381 }; 382 383 enum gro_result { 384 GRO_MERGED, 385 GRO_MERGED_FREE, 386 GRO_HELD, 387 GRO_NORMAL, 388 GRO_CONSUMED, 389 }; 390 typedef enum gro_result gro_result_t; 391 392 /* 393 * enum rx_handler_result - Possible return values for rx_handlers. 394 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 395 * further. 396 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 397 * case skb->dev was changed by rx_handler. 398 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 399 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 400 * 401 * rx_handlers are functions called from inside __netif_receive_skb(), to do 402 * special processing of the skb, prior to delivery to protocol handlers. 403 * 404 * Currently, a net_device can only have a single rx_handler registered. Trying 405 * to register a second rx_handler will return -EBUSY. 406 * 407 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 408 * To unregister a rx_handler on a net_device, use 409 * netdev_rx_handler_unregister(). 410 * 411 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 412 * do with the skb. 413 * 414 * If the rx_handler consumed the skb in some way, it should return 415 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 416 * the skb to be delivered in some other way. 417 * 418 * If the rx_handler changed skb->dev, to divert the skb to another 419 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 420 * new device will be called if it exists. 421 * 422 * If the rx_handler decides the skb should be ignored, it should return 423 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 424 * are registered on exact device (ptype->dev == skb->dev). 425 * 426 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 427 * delivered, it should return RX_HANDLER_PASS. 428 * 429 * A device without a registered rx_handler will behave as if rx_handler 430 * returned RX_HANDLER_PASS. 431 */ 432 433 enum rx_handler_result { 434 RX_HANDLER_CONSUMED, 435 RX_HANDLER_ANOTHER, 436 RX_HANDLER_EXACT, 437 RX_HANDLER_PASS, 438 }; 439 typedef enum rx_handler_result rx_handler_result_t; 440 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 441 442 void __napi_schedule(struct napi_struct *n); 443 void __napi_schedule_irqoff(struct napi_struct *n); 444 445 static inline bool napi_disable_pending(struct napi_struct *n) 446 { 447 return test_bit(NAPI_STATE_DISABLE, &n->state); 448 } 449 450 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 451 { 452 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 453 } 454 455 bool napi_schedule_prep(struct napi_struct *n); 456 457 /** 458 * napi_schedule - schedule NAPI poll 459 * @n: NAPI context 460 * 461 * Schedule NAPI poll routine to be called if it is not already 462 * running. 463 */ 464 static inline void napi_schedule(struct napi_struct *n) 465 { 466 if (napi_schedule_prep(n)) 467 __napi_schedule(n); 468 } 469 470 /** 471 * napi_schedule_irqoff - schedule NAPI poll 472 * @n: NAPI context 473 * 474 * Variant of napi_schedule(), assuming hard irqs are masked. 475 */ 476 static inline void napi_schedule_irqoff(struct napi_struct *n) 477 { 478 if (napi_schedule_prep(n)) 479 __napi_schedule_irqoff(n); 480 } 481 482 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 483 static inline bool napi_reschedule(struct napi_struct *napi) 484 { 485 if (napi_schedule_prep(napi)) { 486 __napi_schedule(napi); 487 return true; 488 } 489 return false; 490 } 491 492 bool napi_complete_done(struct napi_struct *n, int work_done); 493 /** 494 * napi_complete - NAPI processing complete 495 * @n: NAPI context 496 * 497 * Mark NAPI processing as complete. 498 * Consider using napi_complete_done() instead. 499 * Return false if device should avoid rearming interrupts. 500 */ 501 static inline bool napi_complete(struct napi_struct *n) 502 { 503 return napi_complete_done(n, 0); 504 } 505 506 int dev_set_threaded(struct net_device *dev, bool threaded); 507 508 /** 509 * napi_disable - prevent NAPI from scheduling 510 * @n: NAPI context 511 * 512 * Stop NAPI from being scheduled on this context. 513 * Waits till any outstanding processing completes. 514 */ 515 void napi_disable(struct napi_struct *n); 516 517 void napi_enable(struct napi_struct *n); 518 519 /** 520 * napi_synchronize - wait until NAPI is not running 521 * @n: NAPI context 522 * 523 * Wait until NAPI is done being scheduled on this context. 524 * Waits till any outstanding processing completes but 525 * does not disable future activations. 526 */ 527 static inline void napi_synchronize(const struct napi_struct *n) 528 { 529 if (IS_ENABLED(CONFIG_SMP)) 530 while (test_bit(NAPI_STATE_SCHED, &n->state)) 531 msleep(1); 532 else 533 barrier(); 534 } 535 536 /** 537 * napi_if_scheduled_mark_missed - if napi is running, set the 538 * NAPIF_STATE_MISSED 539 * @n: NAPI context 540 * 541 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 542 * NAPI is scheduled. 543 **/ 544 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 545 { 546 unsigned long val, new; 547 548 do { 549 val = READ_ONCE(n->state); 550 if (val & NAPIF_STATE_DISABLE) 551 return true; 552 553 if (!(val & NAPIF_STATE_SCHED)) 554 return false; 555 556 new = val | NAPIF_STATE_MISSED; 557 } while (cmpxchg(&n->state, val, new) != val); 558 559 return true; 560 } 561 562 enum netdev_queue_state_t { 563 __QUEUE_STATE_DRV_XOFF, 564 __QUEUE_STATE_STACK_XOFF, 565 __QUEUE_STATE_FROZEN, 566 }; 567 568 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 569 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 570 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 571 572 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 573 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 574 QUEUE_STATE_FROZEN) 575 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 576 QUEUE_STATE_FROZEN) 577 578 /* 579 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 580 * netif_tx_* functions below are used to manipulate this flag. The 581 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 582 * queue independently. The netif_xmit_*stopped functions below are called 583 * to check if the queue has been stopped by the driver or stack (either 584 * of the XOFF bits are set in the state). Drivers should not need to call 585 * netif_xmit*stopped functions, they should only be using netif_tx_*. 586 */ 587 588 struct netdev_queue { 589 /* 590 * read-mostly part 591 */ 592 struct net_device *dev; 593 netdevice_tracker dev_tracker; 594 595 struct Qdisc __rcu *qdisc; 596 struct Qdisc *qdisc_sleeping; 597 #ifdef CONFIG_SYSFS 598 struct kobject kobj; 599 #endif 600 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 601 int numa_node; 602 #endif 603 unsigned long tx_maxrate; 604 /* 605 * Number of TX timeouts for this queue 606 * (/sys/class/net/DEV/Q/trans_timeout) 607 */ 608 atomic_long_t trans_timeout; 609 610 /* Subordinate device that the queue has been assigned to */ 611 struct net_device *sb_dev; 612 #ifdef CONFIG_XDP_SOCKETS 613 struct xsk_buff_pool *pool; 614 #endif 615 /* 616 * write-mostly part 617 */ 618 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 619 int xmit_lock_owner; 620 /* 621 * Time (in jiffies) of last Tx 622 */ 623 unsigned long trans_start; 624 625 unsigned long state; 626 627 #ifdef CONFIG_BQL 628 struct dql dql; 629 #endif 630 } ____cacheline_aligned_in_smp; 631 632 extern int sysctl_fb_tunnels_only_for_init_net; 633 extern int sysctl_devconf_inherit_init_net; 634 635 /* 636 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 637 * == 1 : For initns only 638 * == 2 : For none. 639 */ 640 static inline bool net_has_fallback_tunnels(const struct net *net) 641 { 642 return !IS_ENABLED(CONFIG_SYSCTL) || 643 !sysctl_fb_tunnels_only_for_init_net || 644 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1); 645 } 646 647 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 648 { 649 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 650 return q->numa_node; 651 #else 652 return NUMA_NO_NODE; 653 #endif 654 } 655 656 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 657 { 658 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 659 q->numa_node = node; 660 #endif 661 } 662 663 #ifdef CONFIG_RPS 664 /* 665 * This structure holds an RPS map which can be of variable length. The 666 * map is an array of CPUs. 667 */ 668 struct rps_map { 669 unsigned int len; 670 struct rcu_head rcu; 671 u16 cpus[]; 672 }; 673 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 674 675 /* 676 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 677 * tail pointer for that CPU's input queue at the time of last enqueue, and 678 * a hardware filter index. 679 */ 680 struct rps_dev_flow { 681 u16 cpu; 682 u16 filter; 683 unsigned int last_qtail; 684 }; 685 #define RPS_NO_FILTER 0xffff 686 687 /* 688 * The rps_dev_flow_table structure contains a table of flow mappings. 689 */ 690 struct rps_dev_flow_table { 691 unsigned int mask; 692 struct rcu_head rcu; 693 struct rps_dev_flow flows[]; 694 }; 695 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 696 ((_num) * sizeof(struct rps_dev_flow))) 697 698 /* 699 * The rps_sock_flow_table contains mappings of flows to the last CPU 700 * on which they were processed by the application (set in recvmsg). 701 * Each entry is a 32bit value. Upper part is the high-order bits 702 * of flow hash, lower part is CPU number. 703 * rps_cpu_mask is used to partition the space, depending on number of 704 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 705 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 706 * meaning we use 32-6=26 bits for the hash. 707 */ 708 struct rps_sock_flow_table { 709 u32 mask; 710 711 u32 ents[] ____cacheline_aligned_in_smp; 712 }; 713 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 714 715 #define RPS_NO_CPU 0xffff 716 717 extern u32 rps_cpu_mask; 718 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 719 720 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 721 u32 hash) 722 { 723 if (table && hash) { 724 unsigned int index = hash & table->mask; 725 u32 val = hash & ~rps_cpu_mask; 726 727 /* We only give a hint, preemption can change CPU under us */ 728 val |= raw_smp_processor_id(); 729 730 if (table->ents[index] != val) 731 table->ents[index] = val; 732 } 733 } 734 735 #ifdef CONFIG_RFS_ACCEL 736 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 737 u16 filter_id); 738 #endif 739 #endif /* CONFIG_RPS */ 740 741 /* This structure contains an instance of an RX queue. */ 742 struct netdev_rx_queue { 743 struct xdp_rxq_info xdp_rxq; 744 #ifdef CONFIG_RPS 745 struct rps_map __rcu *rps_map; 746 struct rps_dev_flow_table __rcu *rps_flow_table; 747 #endif 748 struct kobject kobj; 749 struct net_device *dev; 750 netdevice_tracker dev_tracker; 751 752 #ifdef CONFIG_XDP_SOCKETS 753 struct xsk_buff_pool *pool; 754 #endif 755 } ____cacheline_aligned_in_smp; 756 757 /* 758 * RX queue sysfs structures and functions. 759 */ 760 struct rx_queue_attribute { 761 struct attribute attr; 762 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 763 ssize_t (*store)(struct netdev_rx_queue *queue, 764 const char *buf, size_t len); 765 }; 766 767 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 768 enum xps_map_type { 769 XPS_CPUS = 0, 770 XPS_RXQS, 771 XPS_MAPS_MAX, 772 }; 773 774 #ifdef CONFIG_XPS 775 /* 776 * This structure holds an XPS map which can be of variable length. The 777 * map is an array of queues. 778 */ 779 struct xps_map { 780 unsigned int len; 781 unsigned int alloc_len; 782 struct rcu_head rcu; 783 u16 queues[]; 784 }; 785 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 786 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 787 - sizeof(struct xps_map)) / sizeof(u16)) 788 789 /* 790 * This structure holds all XPS maps for device. Maps are indexed by CPU. 791 * 792 * We keep track of the number of cpus/rxqs used when the struct is allocated, 793 * in nr_ids. This will help not accessing out-of-bound memory. 794 * 795 * We keep track of the number of traffic classes used when the struct is 796 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 797 * not crossing its upper bound, as the original dev->num_tc can be updated in 798 * the meantime. 799 */ 800 struct xps_dev_maps { 801 struct rcu_head rcu; 802 unsigned int nr_ids; 803 s16 num_tc; 804 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 805 }; 806 807 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 808 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 809 810 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 811 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 812 813 #endif /* CONFIG_XPS */ 814 815 #define TC_MAX_QUEUE 16 816 #define TC_BITMASK 15 817 /* HW offloaded queuing disciplines txq count and offset maps */ 818 struct netdev_tc_txq { 819 u16 count; 820 u16 offset; 821 }; 822 823 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 824 /* 825 * This structure is to hold information about the device 826 * configured to run FCoE protocol stack. 827 */ 828 struct netdev_fcoe_hbainfo { 829 char manufacturer[64]; 830 char serial_number[64]; 831 char hardware_version[64]; 832 char driver_version[64]; 833 char optionrom_version[64]; 834 char firmware_version[64]; 835 char model[256]; 836 char model_description[256]; 837 }; 838 #endif 839 840 #define MAX_PHYS_ITEM_ID_LEN 32 841 842 /* This structure holds a unique identifier to identify some 843 * physical item (port for example) used by a netdevice. 844 */ 845 struct netdev_phys_item_id { 846 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 847 unsigned char id_len; 848 }; 849 850 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 851 struct netdev_phys_item_id *b) 852 { 853 return a->id_len == b->id_len && 854 memcmp(a->id, b->id, a->id_len) == 0; 855 } 856 857 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 858 struct sk_buff *skb, 859 struct net_device *sb_dev); 860 861 enum net_device_path_type { 862 DEV_PATH_ETHERNET = 0, 863 DEV_PATH_VLAN, 864 DEV_PATH_BRIDGE, 865 DEV_PATH_PPPOE, 866 DEV_PATH_DSA, 867 DEV_PATH_MTK_WDMA, 868 }; 869 870 struct net_device_path { 871 enum net_device_path_type type; 872 const struct net_device *dev; 873 union { 874 struct { 875 u16 id; 876 __be16 proto; 877 u8 h_dest[ETH_ALEN]; 878 } encap; 879 struct { 880 enum { 881 DEV_PATH_BR_VLAN_KEEP, 882 DEV_PATH_BR_VLAN_TAG, 883 DEV_PATH_BR_VLAN_UNTAG, 884 DEV_PATH_BR_VLAN_UNTAG_HW, 885 } vlan_mode; 886 u16 vlan_id; 887 __be16 vlan_proto; 888 } bridge; 889 struct { 890 int port; 891 u16 proto; 892 } dsa; 893 struct { 894 u8 wdma_idx; 895 u8 queue; 896 u16 wcid; 897 u8 bss; 898 } mtk_wdma; 899 }; 900 }; 901 902 #define NET_DEVICE_PATH_STACK_MAX 5 903 #define NET_DEVICE_PATH_VLAN_MAX 2 904 905 struct net_device_path_stack { 906 int num_paths; 907 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 908 }; 909 910 struct net_device_path_ctx { 911 const struct net_device *dev; 912 const u8 *daddr; 913 914 int num_vlans; 915 struct { 916 u16 id; 917 __be16 proto; 918 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 919 }; 920 921 enum tc_setup_type { 922 TC_SETUP_QDISC_MQPRIO, 923 TC_SETUP_CLSU32, 924 TC_SETUP_CLSFLOWER, 925 TC_SETUP_CLSMATCHALL, 926 TC_SETUP_CLSBPF, 927 TC_SETUP_BLOCK, 928 TC_SETUP_QDISC_CBS, 929 TC_SETUP_QDISC_RED, 930 TC_SETUP_QDISC_PRIO, 931 TC_SETUP_QDISC_MQ, 932 TC_SETUP_QDISC_ETF, 933 TC_SETUP_ROOT_QDISC, 934 TC_SETUP_QDISC_GRED, 935 TC_SETUP_QDISC_TAPRIO, 936 TC_SETUP_FT, 937 TC_SETUP_QDISC_ETS, 938 TC_SETUP_QDISC_TBF, 939 TC_SETUP_QDISC_FIFO, 940 TC_SETUP_QDISC_HTB, 941 TC_SETUP_ACT, 942 }; 943 944 /* These structures hold the attributes of bpf state that are being passed 945 * to the netdevice through the bpf op. 946 */ 947 enum bpf_netdev_command { 948 /* Set or clear a bpf program used in the earliest stages of packet 949 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 950 * is responsible for calling bpf_prog_put on any old progs that are 951 * stored. In case of error, the callee need not release the new prog 952 * reference, but on success it takes ownership and must bpf_prog_put 953 * when it is no longer used. 954 */ 955 XDP_SETUP_PROG, 956 XDP_SETUP_PROG_HW, 957 /* BPF program for offload callbacks, invoked at program load time. */ 958 BPF_OFFLOAD_MAP_ALLOC, 959 BPF_OFFLOAD_MAP_FREE, 960 XDP_SETUP_XSK_POOL, 961 }; 962 963 struct bpf_prog_offload_ops; 964 struct netlink_ext_ack; 965 struct xdp_umem; 966 struct xdp_dev_bulk_queue; 967 struct bpf_xdp_link; 968 969 enum bpf_xdp_mode { 970 XDP_MODE_SKB = 0, 971 XDP_MODE_DRV = 1, 972 XDP_MODE_HW = 2, 973 __MAX_XDP_MODE 974 }; 975 976 struct bpf_xdp_entity { 977 struct bpf_prog *prog; 978 struct bpf_xdp_link *link; 979 }; 980 981 struct netdev_bpf { 982 enum bpf_netdev_command command; 983 union { 984 /* XDP_SETUP_PROG */ 985 struct { 986 u32 flags; 987 struct bpf_prog *prog; 988 struct netlink_ext_ack *extack; 989 }; 990 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 991 struct { 992 struct bpf_offloaded_map *offmap; 993 }; 994 /* XDP_SETUP_XSK_POOL */ 995 struct { 996 struct xsk_buff_pool *pool; 997 u16 queue_id; 998 } xsk; 999 }; 1000 }; 1001 1002 /* Flags for ndo_xsk_wakeup. */ 1003 #define XDP_WAKEUP_RX (1 << 0) 1004 #define XDP_WAKEUP_TX (1 << 1) 1005 1006 #ifdef CONFIG_XFRM_OFFLOAD 1007 struct xfrmdev_ops { 1008 int (*xdo_dev_state_add) (struct xfrm_state *x); 1009 void (*xdo_dev_state_delete) (struct xfrm_state *x); 1010 void (*xdo_dev_state_free) (struct xfrm_state *x); 1011 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 1012 struct xfrm_state *x); 1013 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1014 }; 1015 #endif 1016 1017 struct dev_ifalias { 1018 struct rcu_head rcuhead; 1019 char ifalias[]; 1020 }; 1021 1022 struct devlink; 1023 struct tlsdev_ops; 1024 1025 struct netdev_net_notifier { 1026 struct list_head list; 1027 struct notifier_block *nb; 1028 }; 1029 1030 /* 1031 * This structure defines the management hooks for network devices. 1032 * The following hooks can be defined; unless noted otherwise, they are 1033 * optional and can be filled with a null pointer. 1034 * 1035 * int (*ndo_init)(struct net_device *dev); 1036 * This function is called once when a network device is registered. 1037 * The network device can use this for any late stage initialization 1038 * or semantic validation. It can fail with an error code which will 1039 * be propagated back to register_netdev. 1040 * 1041 * void (*ndo_uninit)(struct net_device *dev); 1042 * This function is called when device is unregistered or when registration 1043 * fails. It is not called if init fails. 1044 * 1045 * int (*ndo_open)(struct net_device *dev); 1046 * This function is called when a network device transitions to the up 1047 * state. 1048 * 1049 * int (*ndo_stop)(struct net_device *dev); 1050 * This function is called when a network device transitions to the down 1051 * state. 1052 * 1053 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1054 * struct net_device *dev); 1055 * Called when a packet needs to be transmitted. 1056 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1057 * the queue before that can happen; it's for obsolete devices and weird 1058 * corner cases, but the stack really does a non-trivial amount 1059 * of useless work if you return NETDEV_TX_BUSY. 1060 * Required; cannot be NULL. 1061 * 1062 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1063 * struct net_device *dev 1064 * netdev_features_t features); 1065 * Called by core transmit path to determine if device is capable of 1066 * performing offload operations on a given packet. This is to give 1067 * the device an opportunity to implement any restrictions that cannot 1068 * be otherwise expressed by feature flags. The check is called with 1069 * the set of features that the stack has calculated and it returns 1070 * those the driver believes to be appropriate. 1071 * 1072 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1073 * struct net_device *sb_dev); 1074 * Called to decide which queue to use when device supports multiple 1075 * transmit queues. 1076 * 1077 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1078 * This function is called to allow device receiver to make 1079 * changes to configuration when multicast or promiscuous is enabled. 1080 * 1081 * void (*ndo_set_rx_mode)(struct net_device *dev); 1082 * This function is called device changes address list filtering. 1083 * If driver handles unicast address filtering, it should set 1084 * IFF_UNICAST_FLT in its priv_flags. 1085 * 1086 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1087 * This function is called when the Media Access Control address 1088 * needs to be changed. If this interface is not defined, the 1089 * MAC address can not be changed. 1090 * 1091 * int (*ndo_validate_addr)(struct net_device *dev); 1092 * Test if Media Access Control address is valid for the device. 1093 * 1094 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1095 * Old-style ioctl entry point. This is used internally by the 1096 * appletalk and ieee802154 subsystems but is no longer called by 1097 * the device ioctl handler. 1098 * 1099 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1100 * Used by the bonding driver for its device specific ioctls: 1101 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1102 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1103 * 1104 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1105 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1106 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1107 * 1108 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1109 * Used to set network devices bus interface parameters. This interface 1110 * is retained for legacy reasons; new devices should use the bus 1111 * interface (PCI) for low level management. 1112 * 1113 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1114 * Called when a user wants to change the Maximum Transfer Unit 1115 * of a device. 1116 * 1117 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1118 * Callback used when the transmitter has not made any progress 1119 * for dev->watchdog ticks. 1120 * 1121 * void (*ndo_get_stats64)(struct net_device *dev, 1122 * struct rtnl_link_stats64 *storage); 1123 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1124 * Called when a user wants to get the network device usage 1125 * statistics. Drivers must do one of the following: 1126 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1127 * rtnl_link_stats64 structure passed by the caller. 1128 * 2. Define @ndo_get_stats to update a net_device_stats structure 1129 * (which should normally be dev->stats) and return a pointer to 1130 * it. The structure may be changed asynchronously only if each 1131 * field is written atomically. 1132 * 3. Update dev->stats asynchronously and atomically, and define 1133 * neither operation. 1134 * 1135 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1136 * Return true if this device supports offload stats of this attr_id. 1137 * 1138 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1139 * void *attr_data) 1140 * Get statistics for offload operations by attr_id. Write it into the 1141 * attr_data pointer. 1142 * 1143 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1144 * If device supports VLAN filtering this function is called when a 1145 * VLAN id is registered. 1146 * 1147 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1148 * If device supports VLAN filtering this function is called when a 1149 * VLAN id is unregistered. 1150 * 1151 * void (*ndo_poll_controller)(struct net_device *dev); 1152 * 1153 * SR-IOV management functions. 1154 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1155 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1156 * u8 qos, __be16 proto); 1157 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1158 * int max_tx_rate); 1159 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1160 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1161 * int (*ndo_get_vf_config)(struct net_device *dev, 1162 * int vf, struct ifla_vf_info *ivf); 1163 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1164 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1165 * struct nlattr *port[]); 1166 * 1167 * Enable or disable the VF ability to query its RSS Redirection Table and 1168 * Hash Key. This is needed since on some devices VF share this information 1169 * with PF and querying it may introduce a theoretical security risk. 1170 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1171 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1172 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1173 * void *type_data); 1174 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1175 * This is always called from the stack with the rtnl lock held and netif 1176 * tx queues stopped. This allows the netdevice to perform queue 1177 * management safely. 1178 * 1179 * Fiber Channel over Ethernet (FCoE) offload functions. 1180 * int (*ndo_fcoe_enable)(struct net_device *dev); 1181 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1182 * so the underlying device can perform whatever needed configuration or 1183 * initialization to support acceleration of FCoE traffic. 1184 * 1185 * int (*ndo_fcoe_disable)(struct net_device *dev); 1186 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1187 * so the underlying device can perform whatever needed clean-ups to 1188 * stop supporting acceleration of FCoE traffic. 1189 * 1190 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1191 * struct scatterlist *sgl, unsigned int sgc); 1192 * Called when the FCoE Initiator wants to initialize an I/O that 1193 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1194 * perform necessary setup and returns 1 to indicate the device is set up 1195 * successfully to perform DDP on this I/O, otherwise this returns 0. 1196 * 1197 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1198 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1199 * indicated by the FC exchange id 'xid', so the underlying device can 1200 * clean up and reuse resources for later DDP requests. 1201 * 1202 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1203 * struct scatterlist *sgl, unsigned int sgc); 1204 * Called when the FCoE Target wants to initialize an I/O that 1205 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1206 * perform necessary setup and returns 1 to indicate the device is set up 1207 * successfully to perform DDP on this I/O, otherwise this returns 0. 1208 * 1209 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1210 * struct netdev_fcoe_hbainfo *hbainfo); 1211 * Called when the FCoE Protocol stack wants information on the underlying 1212 * device. This information is utilized by the FCoE protocol stack to 1213 * register attributes with Fiber Channel management service as per the 1214 * FC-GS Fabric Device Management Information(FDMI) specification. 1215 * 1216 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1217 * Called when the underlying device wants to override default World Wide 1218 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1219 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1220 * protocol stack to use. 1221 * 1222 * RFS acceleration. 1223 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1224 * u16 rxq_index, u32 flow_id); 1225 * Set hardware filter for RFS. rxq_index is the target queue index; 1226 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1227 * Return the filter ID on success, or a negative error code. 1228 * 1229 * Slave management functions (for bridge, bonding, etc). 1230 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1231 * Called to make another netdev an underling. 1232 * 1233 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1234 * Called to release previously enslaved netdev. 1235 * 1236 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1237 * struct sk_buff *skb, 1238 * bool all_slaves); 1239 * Get the xmit slave of master device. If all_slaves is true, function 1240 * assume all the slaves can transmit. 1241 * 1242 * Feature/offload setting functions. 1243 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1244 * netdev_features_t features); 1245 * Adjusts the requested feature flags according to device-specific 1246 * constraints, and returns the resulting flags. Must not modify 1247 * the device state. 1248 * 1249 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1250 * Called to update device configuration to new features. Passed 1251 * feature set might be less than what was returned by ndo_fix_features()). 1252 * Must return >0 or -errno if it changed dev->features itself. 1253 * 1254 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1255 * struct net_device *dev, 1256 * const unsigned char *addr, u16 vid, u16 flags, 1257 * struct netlink_ext_ack *extack); 1258 * Adds an FDB entry to dev for addr. 1259 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1260 * struct net_device *dev, 1261 * const unsigned char *addr, u16 vid) 1262 * Deletes the FDB entry from dev coresponding to addr. 1263 * int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, struct nlattr *tb[], 1264 * struct net_device *dev, 1265 * u16 vid, 1266 * struct netlink_ext_ack *extack); 1267 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1268 * struct net_device *dev, struct net_device *filter_dev, 1269 * int *idx) 1270 * Used to add FDB entries to dump requests. Implementers should add 1271 * entries to skb and update idx with the number of entries. 1272 * 1273 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1274 * u16 flags, struct netlink_ext_ack *extack) 1275 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1276 * struct net_device *dev, u32 filter_mask, 1277 * int nlflags) 1278 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1279 * u16 flags); 1280 * 1281 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1282 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1283 * which do not represent real hardware may define this to allow their 1284 * userspace components to manage their virtual carrier state. Devices 1285 * that determine carrier state from physical hardware properties (eg 1286 * network cables) or protocol-dependent mechanisms (eg 1287 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1288 * 1289 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1290 * struct netdev_phys_item_id *ppid); 1291 * Called to get ID of physical port of this device. If driver does 1292 * not implement this, it is assumed that the hw is not able to have 1293 * multiple net devices on single physical port. 1294 * 1295 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1296 * struct netdev_phys_item_id *ppid) 1297 * Called to get the parent ID of the physical port of this device. 1298 * 1299 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1300 * struct net_device *dev) 1301 * Called by upper layer devices to accelerate switching or other 1302 * station functionality into hardware. 'pdev is the lowerdev 1303 * to use for the offload and 'dev' is the net device that will 1304 * back the offload. Returns a pointer to the private structure 1305 * the upper layer will maintain. 1306 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1307 * Called by upper layer device to delete the station created 1308 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1309 * the station and priv is the structure returned by the add 1310 * operation. 1311 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1312 * int queue_index, u32 maxrate); 1313 * Called when a user wants to set a max-rate limitation of specific 1314 * TX queue. 1315 * int (*ndo_get_iflink)(const struct net_device *dev); 1316 * Called to get the iflink value of this device. 1317 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1318 * This function is used to get egress tunnel information for given skb. 1319 * This is useful for retrieving outer tunnel header parameters while 1320 * sampling packet. 1321 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1322 * This function is used to specify the headroom that the skb must 1323 * consider when allocation skb during packet reception. Setting 1324 * appropriate rx headroom value allows avoiding skb head copy on 1325 * forward. Setting a negative value resets the rx headroom to the 1326 * default value. 1327 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1328 * This function is used to set or query state related to XDP on the 1329 * netdevice and manage BPF offload. See definition of 1330 * enum bpf_netdev_command for details. 1331 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1332 * u32 flags); 1333 * This function is used to submit @n XDP packets for transmit on a 1334 * netdevice. Returns number of frames successfully transmitted, frames 1335 * that got dropped are freed/returned via xdp_return_frame(). 1336 * Returns negative number, means general error invoking ndo, meaning 1337 * no frames were xmit'ed and core-caller will free all frames. 1338 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1339 * struct xdp_buff *xdp); 1340 * Get the xmit slave of master device based on the xdp_buff. 1341 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1342 * This function is used to wake up the softirq, ksoftirqd or kthread 1343 * responsible for sending and/or receiving packets on a specific 1344 * queue id bound to an AF_XDP socket. The flags field specifies if 1345 * only RX, only Tx, or both should be woken up using the flags 1346 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1347 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1348 * Get devlink port instance associated with a given netdev. 1349 * Called with a reference on the netdevice and devlink locks only, 1350 * rtnl_lock is not held. 1351 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1352 * int cmd); 1353 * Add, change, delete or get information on an IPv4 tunnel. 1354 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1355 * If a device is paired with a peer device, return the peer instance. 1356 * The caller must be under RCU read context. 1357 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1358 * Get the forwarding path to reach the real device from the HW destination address 1359 * ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1360 * const struct skb_shared_hwtstamps *hwtstamps, 1361 * bool cycles); 1362 * Get hardware timestamp based on normal/adjustable time or free running 1363 * cycle counter. This function is required if physical clock supports a 1364 * free running cycle counter. 1365 */ 1366 struct net_device_ops { 1367 int (*ndo_init)(struct net_device *dev); 1368 void (*ndo_uninit)(struct net_device *dev); 1369 int (*ndo_open)(struct net_device *dev); 1370 int (*ndo_stop)(struct net_device *dev); 1371 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1372 struct net_device *dev); 1373 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1374 struct net_device *dev, 1375 netdev_features_t features); 1376 u16 (*ndo_select_queue)(struct net_device *dev, 1377 struct sk_buff *skb, 1378 struct net_device *sb_dev); 1379 void (*ndo_change_rx_flags)(struct net_device *dev, 1380 int flags); 1381 void (*ndo_set_rx_mode)(struct net_device *dev); 1382 int (*ndo_set_mac_address)(struct net_device *dev, 1383 void *addr); 1384 int (*ndo_validate_addr)(struct net_device *dev); 1385 int (*ndo_do_ioctl)(struct net_device *dev, 1386 struct ifreq *ifr, int cmd); 1387 int (*ndo_eth_ioctl)(struct net_device *dev, 1388 struct ifreq *ifr, int cmd); 1389 int (*ndo_siocbond)(struct net_device *dev, 1390 struct ifreq *ifr, int cmd); 1391 int (*ndo_siocwandev)(struct net_device *dev, 1392 struct if_settings *ifs); 1393 int (*ndo_siocdevprivate)(struct net_device *dev, 1394 struct ifreq *ifr, 1395 void __user *data, int cmd); 1396 int (*ndo_set_config)(struct net_device *dev, 1397 struct ifmap *map); 1398 int (*ndo_change_mtu)(struct net_device *dev, 1399 int new_mtu); 1400 int (*ndo_neigh_setup)(struct net_device *dev, 1401 struct neigh_parms *); 1402 void (*ndo_tx_timeout) (struct net_device *dev, 1403 unsigned int txqueue); 1404 1405 void (*ndo_get_stats64)(struct net_device *dev, 1406 struct rtnl_link_stats64 *storage); 1407 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1408 int (*ndo_get_offload_stats)(int attr_id, 1409 const struct net_device *dev, 1410 void *attr_data); 1411 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1412 1413 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1414 __be16 proto, u16 vid); 1415 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1416 __be16 proto, u16 vid); 1417 #ifdef CONFIG_NET_POLL_CONTROLLER 1418 void (*ndo_poll_controller)(struct net_device *dev); 1419 int (*ndo_netpoll_setup)(struct net_device *dev, 1420 struct netpoll_info *info); 1421 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1422 #endif 1423 int (*ndo_set_vf_mac)(struct net_device *dev, 1424 int queue, u8 *mac); 1425 int (*ndo_set_vf_vlan)(struct net_device *dev, 1426 int queue, u16 vlan, 1427 u8 qos, __be16 proto); 1428 int (*ndo_set_vf_rate)(struct net_device *dev, 1429 int vf, int min_tx_rate, 1430 int max_tx_rate); 1431 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1432 int vf, bool setting); 1433 int (*ndo_set_vf_trust)(struct net_device *dev, 1434 int vf, bool setting); 1435 int (*ndo_get_vf_config)(struct net_device *dev, 1436 int vf, 1437 struct ifla_vf_info *ivf); 1438 int (*ndo_set_vf_link_state)(struct net_device *dev, 1439 int vf, int link_state); 1440 int (*ndo_get_vf_stats)(struct net_device *dev, 1441 int vf, 1442 struct ifla_vf_stats 1443 *vf_stats); 1444 int (*ndo_set_vf_port)(struct net_device *dev, 1445 int vf, 1446 struct nlattr *port[]); 1447 int (*ndo_get_vf_port)(struct net_device *dev, 1448 int vf, struct sk_buff *skb); 1449 int (*ndo_get_vf_guid)(struct net_device *dev, 1450 int vf, 1451 struct ifla_vf_guid *node_guid, 1452 struct ifla_vf_guid *port_guid); 1453 int (*ndo_set_vf_guid)(struct net_device *dev, 1454 int vf, u64 guid, 1455 int guid_type); 1456 int (*ndo_set_vf_rss_query_en)( 1457 struct net_device *dev, 1458 int vf, bool setting); 1459 int (*ndo_setup_tc)(struct net_device *dev, 1460 enum tc_setup_type type, 1461 void *type_data); 1462 #if IS_ENABLED(CONFIG_FCOE) 1463 int (*ndo_fcoe_enable)(struct net_device *dev); 1464 int (*ndo_fcoe_disable)(struct net_device *dev); 1465 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1466 u16 xid, 1467 struct scatterlist *sgl, 1468 unsigned int sgc); 1469 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1470 u16 xid); 1471 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1472 u16 xid, 1473 struct scatterlist *sgl, 1474 unsigned int sgc); 1475 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1476 struct netdev_fcoe_hbainfo *hbainfo); 1477 #endif 1478 1479 #if IS_ENABLED(CONFIG_LIBFCOE) 1480 #define NETDEV_FCOE_WWNN 0 1481 #define NETDEV_FCOE_WWPN 1 1482 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1483 u64 *wwn, int type); 1484 #endif 1485 1486 #ifdef CONFIG_RFS_ACCEL 1487 int (*ndo_rx_flow_steer)(struct net_device *dev, 1488 const struct sk_buff *skb, 1489 u16 rxq_index, 1490 u32 flow_id); 1491 #endif 1492 int (*ndo_add_slave)(struct net_device *dev, 1493 struct net_device *slave_dev, 1494 struct netlink_ext_ack *extack); 1495 int (*ndo_del_slave)(struct net_device *dev, 1496 struct net_device *slave_dev); 1497 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1498 struct sk_buff *skb, 1499 bool all_slaves); 1500 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1501 struct sock *sk); 1502 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1503 netdev_features_t features); 1504 int (*ndo_set_features)(struct net_device *dev, 1505 netdev_features_t features); 1506 int (*ndo_neigh_construct)(struct net_device *dev, 1507 struct neighbour *n); 1508 void (*ndo_neigh_destroy)(struct net_device *dev, 1509 struct neighbour *n); 1510 1511 int (*ndo_fdb_add)(struct ndmsg *ndm, 1512 struct nlattr *tb[], 1513 struct net_device *dev, 1514 const unsigned char *addr, 1515 u16 vid, 1516 u16 flags, 1517 struct netlink_ext_ack *extack); 1518 int (*ndo_fdb_del)(struct ndmsg *ndm, 1519 struct nlattr *tb[], 1520 struct net_device *dev, 1521 const unsigned char *addr, 1522 u16 vid, struct netlink_ext_ack *extack); 1523 int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, 1524 struct nlattr *tb[], 1525 struct net_device *dev, 1526 u16 vid, 1527 struct netlink_ext_ack *extack); 1528 int (*ndo_fdb_dump)(struct sk_buff *skb, 1529 struct netlink_callback *cb, 1530 struct net_device *dev, 1531 struct net_device *filter_dev, 1532 int *idx); 1533 int (*ndo_fdb_get)(struct sk_buff *skb, 1534 struct nlattr *tb[], 1535 struct net_device *dev, 1536 const unsigned char *addr, 1537 u16 vid, u32 portid, u32 seq, 1538 struct netlink_ext_ack *extack); 1539 int (*ndo_bridge_setlink)(struct net_device *dev, 1540 struct nlmsghdr *nlh, 1541 u16 flags, 1542 struct netlink_ext_ack *extack); 1543 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1544 u32 pid, u32 seq, 1545 struct net_device *dev, 1546 u32 filter_mask, 1547 int nlflags); 1548 int (*ndo_bridge_dellink)(struct net_device *dev, 1549 struct nlmsghdr *nlh, 1550 u16 flags); 1551 int (*ndo_change_carrier)(struct net_device *dev, 1552 bool new_carrier); 1553 int (*ndo_get_phys_port_id)(struct net_device *dev, 1554 struct netdev_phys_item_id *ppid); 1555 int (*ndo_get_port_parent_id)(struct net_device *dev, 1556 struct netdev_phys_item_id *ppid); 1557 int (*ndo_get_phys_port_name)(struct net_device *dev, 1558 char *name, size_t len); 1559 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1560 struct net_device *dev); 1561 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1562 void *priv); 1563 1564 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1565 int queue_index, 1566 u32 maxrate); 1567 int (*ndo_get_iflink)(const struct net_device *dev); 1568 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1569 struct sk_buff *skb); 1570 void (*ndo_set_rx_headroom)(struct net_device *dev, 1571 int needed_headroom); 1572 int (*ndo_bpf)(struct net_device *dev, 1573 struct netdev_bpf *bpf); 1574 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1575 struct xdp_frame **xdp, 1576 u32 flags); 1577 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1578 struct xdp_buff *xdp); 1579 int (*ndo_xsk_wakeup)(struct net_device *dev, 1580 u32 queue_id, u32 flags); 1581 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1582 int (*ndo_tunnel_ctl)(struct net_device *dev, 1583 struct ip_tunnel_parm *p, int cmd); 1584 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1585 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1586 struct net_device_path *path); 1587 ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1588 const struct skb_shared_hwtstamps *hwtstamps, 1589 bool cycles); 1590 }; 1591 1592 /** 1593 * enum netdev_priv_flags - &struct net_device priv_flags 1594 * 1595 * These are the &struct net_device, they are only set internally 1596 * by drivers and used in the kernel. These flags are invisible to 1597 * userspace; this means that the order of these flags can change 1598 * during any kernel release. 1599 * 1600 * You should have a pretty good reason to be extending these flags. 1601 * 1602 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1603 * @IFF_EBRIDGE: Ethernet bridging device 1604 * @IFF_BONDING: bonding master or slave 1605 * @IFF_ISATAP: ISATAP interface (RFC4214) 1606 * @IFF_WAN_HDLC: WAN HDLC device 1607 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1608 * release skb->dst 1609 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1610 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1611 * @IFF_MACVLAN_PORT: device used as macvlan port 1612 * @IFF_BRIDGE_PORT: device used as bridge port 1613 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1614 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1615 * @IFF_UNICAST_FLT: Supports unicast filtering 1616 * @IFF_TEAM_PORT: device used as team port 1617 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1618 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1619 * change when it's running 1620 * @IFF_MACVLAN: Macvlan device 1621 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1622 * underlying stacked devices 1623 * @IFF_L3MDEV_MASTER: device is an L3 master device 1624 * @IFF_NO_QUEUE: device can run without qdisc attached 1625 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1626 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1627 * @IFF_TEAM: device is a team device 1628 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1629 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1630 * entity (i.e. the master device for bridged veth) 1631 * @IFF_MACSEC: device is a MACsec device 1632 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1633 * @IFF_FAILOVER: device is a failover master device 1634 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1635 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1636 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1637 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1638 * skb_headlen(skb) == 0 (data starts from frag0) 1639 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN 1640 */ 1641 enum netdev_priv_flags { 1642 IFF_802_1Q_VLAN = 1<<0, 1643 IFF_EBRIDGE = 1<<1, 1644 IFF_BONDING = 1<<2, 1645 IFF_ISATAP = 1<<3, 1646 IFF_WAN_HDLC = 1<<4, 1647 IFF_XMIT_DST_RELEASE = 1<<5, 1648 IFF_DONT_BRIDGE = 1<<6, 1649 IFF_DISABLE_NETPOLL = 1<<7, 1650 IFF_MACVLAN_PORT = 1<<8, 1651 IFF_BRIDGE_PORT = 1<<9, 1652 IFF_OVS_DATAPATH = 1<<10, 1653 IFF_TX_SKB_SHARING = 1<<11, 1654 IFF_UNICAST_FLT = 1<<12, 1655 IFF_TEAM_PORT = 1<<13, 1656 IFF_SUPP_NOFCS = 1<<14, 1657 IFF_LIVE_ADDR_CHANGE = 1<<15, 1658 IFF_MACVLAN = 1<<16, 1659 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1660 IFF_L3MDEV_MASTER = 1<<18, 1661 IFF_NO_QUEUE = 1<<19, 1662 IFF_OPENVSWITCH = 1<<20, 1663 IFF_L3MDEV_SLAVE = 1<<21, 1664 IFF_TEAM = 1<<22, 1665 IFF_RXFH_CONFIGURED = 1<<23, 1666 IFF_PHONY_HEADROOM = 1<<24, 1667 IFF_MACSEC = 1<<25, 1668 IFF_NO_RX_HANDLER = 1<<26, 1669 IFF_FAILOVER = 1<<27, 1670 IFF_FAILOVER_SLAVE = 1<<28, 1671 IFF_L3MDEV_RX_HANDLER = 1<<29, 1672 IFF_LIVE_RENAME_OK = 1<<30, 1673 IFF_TX_SKB_NO_LINEAR = 1<<31, 1674 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32), 1675 }; 1676 1677 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1678 #define IFF_EBRIDGE IFF_EBRIDGE 1679 #define IFF_BONDING IFF_BONDING 1680 #define IFF_ISATAP IFF_ISATAP 1681 #define IFF_WAN_HDLC IFF_WAN_HDLC 1682 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1683 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1684 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1685 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1686 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1687 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1688 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1689 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1690 #define IFF_TEAM_PORT IFF_TEAM_PORT 1691 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1692 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1693 #define IFF_MACVLAN IFF_MACVLAN 1694 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1695 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1696 #define IFF_NO_QUEUE IFF_NO_QUEUE 1697 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1698 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1699 #define IFF_TEAM IFF_TEAM 1700 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1701 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM 1702 #define IFF_MACSEC IFF_MACSEC 1703 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1704 #define IFF_FAILOVER IFF_FAILOVER 1705 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1706 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1707 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1708 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR 1709 1710 /* Specifies the type of the struct net_device::ml_priv pointer */ 1711 enum netdev_ml_priv_type { 1712 ML_PRIV_NONE, 1713 ML_PRIV_CAN, 1714 }; 1715 1716 /** 1717 * struct net_device - The DEVICE structure. 1718 * 1719 * Actually, this whole structure is a big mistake. It mixes I/O 1720 * data with strictly "high-level" data, and it has to know about 1721 * almost every data structure used in the INET module. 1722 * 1723 * @name: This is the first field of the "visible" part of this structure 1724 * (i.e. as seen by users in the "Space.c" file). It is the name 1725 * of the interface. 1726 * 1727 * @name_node: Name hashlist node 1728 * @ifalias: SNMP alias 1729 * @mem_end: Shared memory end 1730 * @mem_start: Shared memory start 1731 * @base_addr: Device I/O address 1732 * @irq: Device IRQ number 1733 * 1734 * @state: Generic network queuing layer state, see netdev_state_t 1735 * @dev_list: The global list of network devices 1736 * @napi_list: List entry used for polling NAPI devices 1737 * @unreg_list: List entry when we are unregistering the 1738 * device; see the function unregister_netdev 1739 * @close_list: List entry used when we are closing the device 1740 * @ptype_all: Device-specific packet handlers for all protocols 1741 * @ptype_specific: Device-specific, protocol-specific packet handlers 1742 * 1743 * @adj_list: Directly linked devices, like slaves for bonding 1744 * @features: Currently active device features 1745 * @hw_features: User-changeable features 1746 * 1747 * @wanted_features: User-requested features 1748 * @vlan_features: Mask of features inheritable by VLAN devices 1749 * 1750 * @hw_enc_features: Mask of features inherited by encapsulating devices 1751 * This field indicates what encapsulation 1752 * offloads the hardware is capable of doing, 1753 * and drivers will need to set them appropriately. 1754 * 1755 * @mpls_features: Mask of features inheritable by MPLS 1756 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1757 * 1758 * @ifindex: interface index 1759 * @group: The group the device belongs to 1760 * 1761 * @stats: Statistics struct, which was left as a legacy, use 1762 * rtnl_link_stats64 instead 1763 * 1764 * @core_stats: core networking counters, 1765 * do not use this in drivers 1766 * @carrier_up_count: Number of times the carrier has been up 1767 * @carrier_down_count: Number of times the carrier has been down 1768 * 1769 * @wireless_handlers: List of functions to handle Wireless Extensions, 1770 * instead of ioctl, 1771 * see <net/iw_handler.h> for details. 1772 * @wireless_data: Instance data managed by the core of wireless extensions 1773 * 1774 * @netdev_ops: Includes several pointers to callbacks, 1775 * if one wants to override the ndo_*() functions 1776 * @ethtool_ops: Management operations 1777 * @l3mdev_ops: Layer 3 master device operations 1778 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1779 * discovery handling. Necessary for e.g. 6LoWPAN. 1780 * @xfrmdev_ops: Transformation offload operations 1781 * @tlsdev_ops: Transport Layer Security offload operations 1782 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1783 * of Layer 2 headers. 1784 * 1785 * @flags: Interface flags (a la BSD) 1786 * @priv_flags: Like 'flags' but invisible to userspace, 1787 * see if.h for the definitions 1788 * @gflags: Global flags ( kept as legacy ) 1789 * @padded: How much padding added by alloc_netdev() 1790 * @operstate: RFC2863 operstate 1791 * @link_mode: Mapping policy to operstate 1792 * @if_port: Selectable AUI, TP, ... 1793 * @dma: DMA channel 1794 * @mtu: Interface MTU value 1795 * @min_mtu: Interface Minimum MTU value 1796 * @max_mtu: Interface Maximum MTU value 1797 * @type: Interface hardware type 1798 * @hard_header_len: Maximum hardware header length. 1799 * @min_header_len: Minimum hardware header length 1800 * 1801 * @needed_headroom: Extra headroom the hardware may need, but not in all 1802 * cases can this be guaranteed 1803 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1804 * cases can this be guaranteed. Some cases also use 1805 * LL_MAX_HEADER instead to allocate the skb 1806 * 1807 * interface address info: 1808 * 1809 * @perm_addr: Permanent hw address 1810 * @addr_assign_type: Hw address assignment type 1811 * @addr_len: Hardware address length 1812 * @upper_level: Maximum depth level of upper devices. 1813 * @lower_level: Maximum depth level of lower devices. 1814 * @neigh_priv_len: Used in neigh_alloc() 1815 * @dev_id: Used to differentiate devices that share 1816 * the same link layer address 1817 * @dev_port: Used to differentiate devices that share 1818 * the same function 1819 * @addr_list_lock: XXX: need comments on this one 1820 * @name_assign_type: network interface name assignment type 1821 * @uc_promisc: Counter that indicates promiscuous mode 1822 * has been enabled due to the need to listen to 1823 * additional unicast addresses in a device that 1824 * does not implement ndo_set_rx_mode() 1825 * @uc: unicast mac addresses 1826 * @mc: multicast mac addresses 1827 * @dev_addrs: list of device hw addresses 1828 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1829 * @promiscuity: Number of times the NIC is told to work in 1830 * promiscuous mode; if it becomes 0 the NIC will 1831 * exit promiscuous mode 1832 * @allmulti: Counter, enables or disables allmulticast mode 1833 * 1834 * @vlan_info: VLAN info 1835 * @dsa_ptr: dsa specific data 1836 * @tipc_ptr: TIPC specific data 1837 * @atalk_ptr: AppleTalk link 1838 * @ip_ptr: IPv4 specific data 1839 * @dn_ptr: DECnet specific data 1840 * @ip6_ptr: IPv6 specific data 1841 * @ax25_ptr: AX.25 specific data 1842 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1843 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1844 * device struct 1845 * @mpls_ptr: mpls_dev struct pointer 1846 * @mctp_ptr: MCTP specific data 1847 * 1848 * @dev_addr: Hw address (before bcast, 1849 * because most packets are unicast) 1850 * 1851 * @_rx: Array of RX queues 1852 * @num_rx_queues: Number of RX queues 1853 * allocated at register_netdev() time 1854 * @real_num_rx_queues: Number of RX queues currently active in device 1855 * @xdp_prog: XDP sockets filter program pointer 1856 * @gro_flush_timeout: timeout for GRO layer in NAPI 1857 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1858 * allow to avoid NIC hard IRQ, on busy queues. 1859 * 1860 * @rx_handler: handler for received packets 1861 * @rx_handler_data: XXX: need comments on this one 1862 * @miniq_ingress: ingress/clsact qdisc specific data for 1863 * ingress processing 1864 * @ingress_queue: XXX: need comments on this one 1865 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1866 * @broadcast: hw bcast address 1867 * 1868 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1869 * indexed by RX queue number. Assigned by driver. 1870 * This must only be set if the ndo_rx_flow_steer 1871 * operation is defined 1872 * @index_hlist: Device index hash chain 1873 * 1874 * @_tx: Array of TX queues 1875 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1876 * @real_num_tx_queues: Number of TX queues currently active in device 1877 * @qdisc: Root qdisc from userspace point of view 1878 * @tx_queue_len: Max frames per queue allowed 1879 * @tx_global_lock: XXX: need comments on this one 1880 * @xdp_bulkq: XDP device bulk queue 1881 * @xps_maps: all CPUs/RXQs maps for XPS device 1882 * 1883 * @xps_maps: XXX: need comments on this one 1884 * @miniq_egress: clsact qdisc specific data for 1885 * egress processing 1886 * @nf_hooks_egress: netfilter hooks executed for egress packets 1887 * @qdisc_hash: qdisc hash table 1888 * @watchdog_timeo: Represents the timeout that is used by 1889 * the watchdog (see dev_watchdog()) 1890 * @watchdog_timer: List of timers 1891 * 1892 * @proto_down_reason: reason a netdev interface is held down 1893 * @pcpu_refcnt: Number of references to this device 1894 * @dev_refcnt: Number of references to this device 1895 * @refcnt_tracker: Tracker directory for tracked references to this device 1896 * @todo_list: Delayed register/unregister 1897 * @link_watch_list: XXX: need comments on this one 1898 * 1899 * @reg_state: Register/unregister state machine 1900 * @dismantle: Device is going to be freed 1901 * @rtnl_link_state: This enum represents the phases of creating 1902 * a new link 1903 * 1904 * @needs_free_netdev: Should unregister perform free_netdev? 1905 * @priv_destructor: Called from unregister 1906 * @npinfo: XXX: need comments on this one 1907 * @nd_net: Network namespace this network device is inside 1908 * 1909 * @ml_priv: Mid-layer private 1910 * @ml_priv_type: Mid-layer private type 1911 * @lstats: Loopback statistics 1912 * @tstats: Tunnel statistics 1913 * @dstats: Dummy statistics 1914 * @vstats: Virtual ethernet statistics 1915 * 1916 * @garp_port: GARP 1917 * @mrp_port: MRP 1918 * 1919 * @dm_private: Drop monitor private 1920 * 1921 * @dev: Class/net/name entry 1922 * @sysfs_groups: Space for optional device, statistics and wireless 1923 * sysfs groups 1924 * 1925 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1926 * @rtnl_link_ops: Rtnl_link_ops 1927 * 1928 * @gso_max_size: Maximum size of generic segmentation offload 1929 * @tso_max_size: Device (as in HW) limit on the max TSO request size 1930 * @gso_max_segs: Maximum number of segments that can be passed to the 1931 * NIC for GSO 1932 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count 1933 * 1934 * @dcbnl_ops: Data Center Bridging netlink ops 1935 * @num_tc: Number of traffic classes in the net device 1936 * @tc_to_txq: XXX: need comments on this one 1937 * @prio_tc_map: XXX: need comments on this one 1938 * 1939 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1940 * 1941 * @priomap: XXX: need comments on this one 1942 * @phydev: Physical device may attach itself 1943 * for hardware timestamping 1944 * @sfp_bus: attached &struct sfp_bus structure. 1945 * 1946 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1947 * 1948 * @proto_down: protocol port state information can be sent to the 1949 * switch driver and used to set the phys state of the 1950 * switch port. 1951 * 1952 * @wol_enabled: Wake-on-LAN is enabled 1953 * 1954 * @threaded: napi threaded mode is enabled 1955 * 1956 * @net_notifier_list: List of per-net netdev notifier block 1957 * that follow this device when it is moved 1958 * to another network namespace. 1959 * 1960 * @macsec_ops: MACsec offloading ops 1961 * 1962 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1963 * offload capabilities of the device 1964 * @udp_tunnel_nic: UDP tunnel offload state 1965 * @xdp_state: stores info on attached XDP BPF programs 1966 * 1967 * @nested_level: Used as a parameter of spin_lock_nested() of 1968 * dev->addr_list_lock. 1969 * @unlink_list: As netif_addr_lock() can be called recursively, 1970 * keep a list of interfaces to be deleted. 1971 * @gro_max_size: Maximum size of aggregated packet in generic 1972 * receive offload (GRO) 1973 * 1974 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 1975 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 1976 * @watchdog_dev_tracker: refcount tracker used by watchdog. 1977 * @dev_registered_tracker: tracker for reference held while 1978 * registered 1979 * @offload_xstats_l3: L3 HW stats for this netdevice. 1980 * 1981 * FIXME: cleanup struct net_device such that network protocol info 1982 * moves out. 1983 */ 1984 1985 struct net_device { 1986 char name[IFNAMSIZ]; 1987 struct netdev_name_node *name_node; 1988 struct dev_ifalias __rcu *ifalias; 1989 /* 1990 * I/O specific fields 1991 * FIXME: Merge these and struct ifmap into one 1992 */ 1993 unsigned long mem_end; 1994 unsigned long mem_start; 1995 unsigned long base_addr; 1996 1997 /* 1998 * Some hardware also needs these fields (state,dev_list, 1999 * napi_list,unreg_list,close_list) but they are not 2000 * part of the usual set specified in Space.c. 2001 */ 2002 2003 unsigned long state; 2004 2005 struct list_head dev_list; 2006 struct list_head napi_list; 2007 struct list_head unreg_list; 2008 struct list_head close_list; 2009 struct list_head ptype_all; 2010 struct list_head ptype_specific; 2011 2012 struct { 2013 struct list_head upper; 2014 struct list_head lower; 2015 } adj_list; 2016 2017 /* Read-mostly cache-line for fast-path access */ 2018 unsigned int flags; 2019 unsigned long long priv_flags; 2020 const struct net_device_ops *netdev_ops; 2021 int ifindex; 2022 unsigned short gflags; 2023 unsigned short hard_header_len; 2024 2025 /* Note : dev->mtu is often read without holding a lock. 2026 * Writers usually hold RTNL. 2027 * It is recommended to use READ_ONCE() to annotate the reads, 2028 * and to use WRITE_ONCE() to annotate the writes. 2029 */ 2030 unsigned int mtu; 2031 unsigned short needed_headroom; 2032 unsigned short needed_tailroom; 2033 2034 netdev_features_t features; 2035 netdev_features_t hw_features; 2036 netdev_features_t wanted_features; 2037 netdev_features_t vlan_features; 2038 netdev_features_t hw_enc_features; 2039 netdev_features_t mpls_features; 2040 netdev_features_t gso_partial_features; 2041 2042 unsigned int min_mtu; 2043 unsigned int max_mtu; 2044 unsigned short type; 2045 unsigned char min_header_len; 2046 unsigned char name_assign_type; 2047 2048 int group; 2049 2050 struct net_device_stats stats; /* not used by modern drivers */ 2051 2052 struct net_device_core_stats __percpu *core_stats; 2053 2054 /* Stats to monitor link on/off, flapping */ 2055 atomic_t carrier_up_count; 2056 atomic_t carrier_down_count; 2057 2058 #ifdef CONFIG_WIRELESS_EXT 2059 const struct iw_handler_def *wireless_handlers; 2060 struct iw_public_data *wireless_data; 2061 #endif 2062 const struct ethtool_ops *ethtool_ops; 2063 #ifdef CONFIG_NET_L3_MASTER_DEV 2064 const struct l3mdev_ops *l3mdev_ops; 2065 #endif 2066 #if IS_ENABLED(CONFIG_IPV6) 2067 const struct ndisc_ops *ndisc_ops; 2068 #endif 2069 2070 #ifdef CONFIG_XFRM_OFFLOAD 2071 const struct xfrmdev_ops *xfrmdev_ops; 2072 #endif 2073 2074 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2075 const struct tlsdev_ops *tlsdev_ops; 2076 #endif 2077 2078 const struct header_ops *header_ops; 2079 2080 unsigned char operstate; 2081 unsigned char link_mode; 2082 2083 unsigned char if_port; 2084 unsigned char dma; 2085 2086 /* Interface address info. */ 2087 unsigned char perm_addr[MAX_ADDR_LEN]; 2088 unsigned char addr_assign_type; 2089 unsigned char addr_len; 2090 unsigned char upper_level; 2091 unsigned char lower_level; 2092 2093 unsigned short neigh_priv_len; 2094 unsigned short dev_id; 2095 unsigned short dev_port; 2096 unsigned short padded; 2097 2098 spinlock_t addr_list_lock; 2099 int irq; 2100 2101 struct netdev_hw_addr_list uc; 2102 struct netdev_hw_addr_list mc; 2103 struct netdev_hw_addr_list dev_addrs; 2104 2105 #ifdef CONFIG_SYSFS 2106 struct kset *queues_kset; 2107 #endif 2108 #ifdef CONFIG_LOCKDEP 2109 struct list_head unlink_list; 2110 #endif 2111 unsigned int promiscuity; 2112 unsigned int allmulti; 2113 bool uc_promisc; 2114 #ifdef CONFIG_LOCKDEP 2115 unsigned char nested_level; 2116 #endif 2117 2118 2119 /* Protocol-specific pointers */ 2120 2121 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2122 struct vlan_info __rcu *vlan_info; 2123 #endif 2124 #if IS_ENABLED(CONFIG_NET_DSA) 2125 struct dsa_port *dsa_ptr; 2126 #endif 2127 #if IS_ENABLED(CONFIG_TIPC) 2128 struct tipc_bearer __rcu *tipc_ptr; 2129 #endif 2130 #if IS_ENABLED(CONFIG_ATALK) 2131 void *atalk_ptr; 2132 #endif 2133 struct in_device __rcu *ip_ptr; 2134 #if IS_ENABLED(CONFIG_DECNET) 2135 struct dn_dev __rcu *dn_ptr; 2136 #endif 2137 struct inet6_dev __rcu *ip6_ptr; 2138 #if IS_ENABLED(CONFIG_AX25) 2139 void *ax25_ptr; 2140 #endif 2141 struct wireless_dev *ieee80211_ptr; 2142 struct wpan_dev *ieee802154_ptr; 2143 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2144 struct mpls_dev __rcu *mpls_ptr; 2145 #endif 2146 #if IS_ENABLED(CONFIG_MCTP) 2147 struct mctp_dev __rcu *mctp_ptr; 2148 #endif 2149 2150 /* 2151 * Cache lines mostly used on receive path (including eth_type_trans()) 2152 */ 2153 /* Interface address info used in eth_type_trans() */ 2154 const unsigned char *dev_addr; 2155 2156 struct netdev_rx_queue *_rx; 2157 unsigned int num_rx_queues; 2158 unsigned int real_num_rx_queues; 2159 2160 struct bpf_prog __rcu *xdp_prog; 2161 unsigned long gro_flush_timeout; 2162 int napi_defer_hard_irqs; 2163 #define GRO_MAX_SIZE 65536 2164 unsigned int gro_max_size; 2165 rx_handler_func_t __rcu *rx_handler; 2166 void __rcu *rx_handler_data; 2167 2168 #ifdef CONFIG_NET_CLS_ACT 2169 struct mini_Qdisc __rcu *miniq_ingress; 2170 #endif 2171 struct netdev_queue __rcu *ingress_queue; 2172 #ifdef CONFIG_NETFILTER_INGRESS 2173 struct nf_hook_entries __rcu *nf_hooks_ingress; 2174 #endif 2175 2176 unsigned char broadcast[MAX_ADDR_LEN]; 2177 #ifdef CONFIG_RFS_ACCEL 2178 struct cpu_rmap *rx_cpu_rmap; 2179 #endif 2180 struct hlist_node index_hlist; 2181 2182 /* 2183 * Cache lines mostly used on transmit path 2184 */ 2185 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2186 unsigned int num_tx_queues; 2187 unsigned int real_num_tx_queues; 2188 struct Qdisc __rcu *qdisc; 2189 unsigned int tx_queue_len; 2190 spinlock_t tx_global_lock; 2191 2192 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2193 2194 #ifdef CONFIG_XPS 2195 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2196 #endif 2197 #ifdef CONFIG_NET_CLS_ACT 2198 struct mini_Qdisc __rcu *miniq_egress; 2199 #endif 2200 #ifdef CONFIG_NETFILTER_EGRESS 2201 struct nf_hook_entries __rcu *nf_hooks_egress; 2202 #endif 2203 2204 #ifdef CONFIG_NET_SCHED 2205 DECLARE_HASHTABLE (qdisc_hash, 4); 2206 #endif 2207 /* These may be needed for future network-power-down code. */ 2208 struct timer_list watchdog_timer; 2209 int watchdog_timeo; 2210 2211 u32 proto_down_reason; 2212 2213 struct list_head todo_list; 2214 2215 #ifdef CONFIG_PCPU_DEV_REFCNT 2216 int __percpu *pcpu_refcnt; 2217 #else 2218 refcount_t dev_refcnt; 2219 #endif 2220 struct ref_tracker_dir refcnt_tracker; 2221 2222 struct list_head link_watch_list; 2223 2224 enum { NETREG_UNINITIALIZED=0, 2225 NETREG_REGISTERED, /* completed register_netdevice */ 2226 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2227 NETREG_UNREGISTERED, /* completed unregister todo */ 2228 NETREG_RELEASED, /* called free_netdev */ 2229 NETREG_DUMMY, /* dummy device for NAPI poll */ 2230 } reg_state:8; 2231 2232 bool dismantle; 2233 2234 enum { 2235 RTNL_LINK_INITIALIZED, 2236 RTNL_LINK_INITIALIZING, 2237 } rtnl_link_state:16; 2238 2239 bool needs_free_netdev; 2240 void (*priv_destructor)(struct net_device *dev); 2241 2242 #ifdef CONFIG_NETPOLL 2243 struct netpoll_info __rcu *npinfo; 2244 #endif 2245 2246 possible_net_t nd_net; 2247 2248 /* mid-layer private */ 2249 void *ml_priv; 2250 enum netdev_ml_priv_type ml_priv_type; 2251 2252 union { 2253 struct pcpu_lstats __percpu *lstats; 2254 struct pcpu_sw_netstats __percpu *tstats; 2255 struct pcpu_dstats __percpu *dstats; 2256 }; 2257 2258 #if IS_ENABLED(CONFIG_GARP) 2259 struct garp_port __rcu *garp_port; 2260 #endif 2261 #if IS_ENABLED(CONFIG_MRP) 2262 struct mrp_port __rcu *mrp_port; 2263 #endif 2264 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2265 struct dm_hw_stat_delta __rcu *dm_private; 2266 #endif 2267 struct device dev; 2268 const struct attribute_group *sysfs_groups[4]; 2269 const struct attribute_group *sysfs_rx_queue_group; 2270 2271 const struct rtnl_link_ops *rtnl_link_ops; 2272 2273 /* for setting kernel sock attribute on TCP connection setup */ 2274 #define GSO_MAX_SIZE 65536 2275 unsigned int gso_max_size; 2276 #define TSO_LEGACY_MAX_SIZE 65536 2277 #define TSO_MAX_SIZE UINT_MAX 2278 unsigned int tso_max_size; 2279 #define GSO_MAX_SEGS 65535 2280 u16 gso_max_segs; 2281 #define TSO_MAX_SEGS U16_MAX 2282 u16 tso_max_segs; 2283 2284 #ifdef CONFIG_DCB 2285 const struct dcbnl_rtnl_ops *dcbnl_ops; 2286 #endif 2287 s16 num_tc; 2288 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2289 u8 prio_tc_map[TC_BITMASK + 1]; 2290 2291 #if IS_ENABLED(CONFIG_FCOE) 2292 unsigned int fcoe_ddp_xid; 2293 #endif 2294 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2295 struct netprio_map __rcu *priomap; 2296 #endif 2297 struct phy_device *phydev; 2298 struct sfp_bus *sfp_bus; 2299 struct lock_class_key *qdisc_tx_busylock; 2300 bool proto_down; 2301 unsigned wol_enabled:1; 2302 unsigned threaded:1; 2303 2304 struct list_head net_notifier_list; 2305 2306 #if IS_ENABLED(CONFIG_MACSEC) 2307 /* MACsec management functions */ 2308 const struct macsec_ops *macsec_ops; 2309 #endif 2310 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2311 struct udp_tunnel_nic *udp_tunnel_nic; 2312 2313 /* protected by rtnl_lock */ 2314 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2315 2316 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2317 netdevice_tracker linkwatch_dev_tracker; 2318 netdevice_tracker watchdog_dev_tracker; 2319 netdevice_tracker dev_registered_tracker; 2320 struct rtnl_hw_stats64 *offload_xstats_l3; 2321 }; 2322 #define to_net_dev(d) container_of(d, struct net_device, dev) 2323 2324 static inline bool netif_elide_gro(const struct net_device *dev) 2325 { 2326 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2327 return true; 2328 return false; 2329 } 2330 2331 #define NETDEV_ALIGN 32 2332 2333 static inline 2334 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2335 { 2336 return dev->prio_tc_map[prio & TC_BITMASK]; 2337 } 2338 2339 static inline 2340 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2341 { 2342 if (tc >= dev->num_tc) 2343 return -EINVAL; 2344 2345 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2346 return 0; 2347 } 2348 2349 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2350 void netdev_reset_tc(struct net_device *dev); 2351 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2352 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2353 2354 static inline 2355 int netdev_get_num_tc(struct net_device *dev) 2356 { 2357 return dev->num_tc; 2358 } 2359 2360 static inline void net_prefetch(void *p) 2361 { 2362 prefetch(p); 2363 #if L1_CACHE_BYTES < 128 2364 prefetch((u8 *)p + L1_CACHE_BYTES); 2365 #endif 2366 } 2367 2368 static inline void net_prefetchw(void *p) 2369 { 2370 prefetchw(p); 2371 #if L1_CACHE_BYTES < 128 2372 prefetchw((u8 *)p + L1_CACHE_BYTES); 2373 #endif 2374 } 2375 2376 void netdev_unbind_sb_channel(struct net_device *dev, 2377 struct net_device *sb_dev); 2378 int netdev_bind_sb_channel_queue(struct net_device *dev, 2379 struct net_device *sb_dev, 2380 u8 tc, u16 count, u16 offset); 2381 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2382 static inline int netdev_get_sb_channel(struct net_device *dev) 2383 { 2384 return max_t(int, -dev->num_tc, 0); 2385 } 2386 2387 static inline 2388 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2389 unsigned int index) 2390 { 2391 return &dev->_tx[index]; 2392 } 2393 2394 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2395 const struct sk_buff *skb) 2396 { 2397 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2398 } 2399 2400 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2401 void (*f)(struct net_device *, 2402 struct netdev_queue *, 2403 void *), 2404 void *arg) 2405 { 2406 unsigned int i; 2407 2408 for (i = 0; i < dev->num_tx_queues; i++) 2409 f(dev, &dev->_tx[i], arg); 2410 } 2411 2412 #define netdev_lockdep_set_classes(dev) \ 2413 { \ 2414 static struct lock_class_key qdisc_tx_busylock_key; \ 2415 static struct lock_class_key qdisc_xmit_lock_key; \ 2416 static struct lock_class_key dev_addr_list_lock_key; \ 2417 unsigned int i; \ 2418 \ 2419 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2420 lockdep_set_class(&(dev)->addr_list_lock, \ 2421 &dev_addr_list_lock_key); \ 2422 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2423 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2424 &qdisc_xmit_lock_key); \ 2425 } 2426 2427 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2428 struct net_device *sb_dev); 2429 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2430 struct sk_buff *skb, 2431 struct net_device *sb_dev); 2432 2433 /* returns the headroom that the master device needs to take in account 2434 * when forwarding to this dev 2435 */ 2436 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2437 { 2438 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2439 } 2440 2441 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2442 { 2443 if (dev->netdev_ops->ndo_set_rx_headroom) 2444 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2445 } 2446 2447 /* set the device rx headroom to the dev's default */ 2448 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2449 { 2450 netdev_set_rx_headroom(dev, -1); 2451 } 2452 2453 static inline void *netdev_get_ml_priv(struct net_device *dev, 2454 enum netdev_ml_priv_type type) 2455 { 2456 if (dev->ml_priv_type != type) 2457 return NULL; 2458 2459 return dev->ml_priv; 2460 } 2461 2462 static inline void netdev_set_ml_priv(struct net_device *dev, 2463 void *ml_priv, 2464 enum netdev_ml_priv_type type) 2465 { 2466 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2467 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2468 dev->ml_priv_type, type); 2469 WARN(!dev->ml_priv_type && dev->ml_priv, 2470 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2471 2472 dev->ml_priv = ml_priv; 2473 dev->ml_priv_type = type; 2474 } 2475 2476 /* 2477 * Net namespace inlines 2478 */ 2479 static inline 2480 struct net *dev_net(const struct net_device *dev) 2481 { 2482 return read_pnet(&dev->nd_net); 2483 } 2484 2485 static inline 2486 void dev_net_set(struct net_device *dev, struct net *net) 2487 { 2488 write_pnet(&dev->nd_net, net); 2489 } 2490 2491 /** 2492 * netdev_priv - access network device private data 2493 * @dev: network device 2494 * 2495 * Get network device private data 2496 */ 2497 static inline void *netdev_priv(const struct net_device *dev) 2498 { 2499 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2500 } 2501 2502 /* Set the sysfs physical device reference for the network logical device 2503 * if set prior to registration will cause a symlink during initialization. 2504 */ 2505 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2506 2507 /* Set the sysfs device type for the network logical device to allow 2508 * fine-grained identification of different network device types. For 2509 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2510 */ 2511 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2512 2513 /* Default NAPI poll() weight 2514 * Device drivers are strongly advised to not use bigger value 2515 */ 2516 #define NAPI_POLL_WEIGHT 64 2517 2518 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2519 int (*poll)(struct napi_struct *, int), int weight); 2520 2521 /** 2522 * netif_napi_add() - initialize a NAPI context 2523 * @dev: network device 2524 * @napi: NAPI context 2525 * @poll: polling function 2526 * @weight: default weight 2527 * 2528 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2529 * *any* of the other NAPI-related functions. 2530 */ 2531 static inline void 2532 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2533 int (*poll)(struct napi_struct *, int), int weight) 2534 { 2535 netif_napi_add_weight(dev, napi, poll, weight); 2536 } 2537 2538 static inline void 2539 netif_napi_add_tx_weight(struct net_device *dev, 2540 struct napi_struct *napi, 2541 int (*poll)(struct napi_struct *, int), 2542 int weight) 2543 { 2544 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2545 netif_napi_add_weight(dev, napi, poll, weight); 2546 } 2547 2548 #define netif_tx_napi_add netif_napi_add_tx_weight 2549 2550 /** 2551 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2552 * @dev: network device 2553 * @napi: NAPI context 2554 * @poll: polling function 2555 * 2556 * This variant of netif_napi_add() should be used from drivers using NAPI 2557 * to exclusively poll a TX queue. 2558 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2559 */ 2560 static inline void netif_napi_add_tx(struct net_device *dev, 2561 struct napi_struct *napi, 2562 int (*poll)(struct napi_struct *, int)) 2563 { 2564 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2565 } 2566 2567 /** 2568 * __netif_napi_del - remove a NAPI context 2569 * @napi: NAPI context 2570 * 2571 * Warning: caller must observe RCU grace period before freeing memory 2572 * containing @napi. Drivers might want to call this helper to combine 2573 * all the needed RCU grace periods into a single one. 2574 */ 2575 void __netif_napi_del(struct napi_struct *napi); 2576 2577 /** 2578 * netif_napi_del - remove a NAPI context 2579 * @napi: NAPI context 2580 * 2581 * netif_napi_del() removes a NAPI context from the network device NAPI list 2582 */ 2583 static inline void netif_napi_del(struct napi_struct *napi) 2584 { 2585 __netif_napi_del(napi); 2586 synchronize_net(); 2587 } 2588 2589 struct packet_type { 2590 __be16 type; /* This is really htons(ether_type). */ 2591 bool ignore_outgoing; 2592 struct net_device *dev; /* NULL is wildcarded here */ 2593 netdevice_tracker dev_tracker; 2594 int (*func) (struct sk_buff *, 2595 struct net_device *, 2596 struct packet_type *, 2597 struct net_device *); 2598 void (*list_func) (struct list_head *, 2599 struct packet_type *, 2600 struct net_device *); 2601 bool (*id_match)(struct packet_type *ptype, 2602 struct sock *sk); 2603 struct net *af_packet_net; 2604 void *af_packet_priv; 2605 struct list_head list; 2606 }; 2607 2608 struct offload_callbacks { 2609 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2610 netdev_features_t features); 2611 struct sk_buff *(*gro_receive)(struct list_head *head, 2612 struct sk_buff *skb); 2613 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2614 }; 2615 2616 struct packet_offload { 2617 __be16 type; /* This is really htons(ether_type). */ 2618 u16 priority; 2619 struct offload_callbacks callbacks; 2620 struct list_head list; 2621 }; 2622 2623 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2624 struct pcpu_sw_netstats { 2625 u64 rx_packets; 2626 u64 rx_bytes; 2627 u64 tx_packets; 2628 u64 tx_bytes; 2629 struct u64_stats_sync syncp; 2630 } __aligned(4 * sizeof(u64)); 2631 2632 struct pcpu_lstats { 2633 u64_stats_t packets; 2634 u64_stats_t bytes; 2635 struct u64_stats_sync syncp; 2636 } __aligned(2 * sizeof(u64)); 2637 2638 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2639 2640 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2641 { 2642 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2643 2644 u64_stats_update_begin(&tstats->syncp); 2645 tstats->rx_bytes += len; 2646 tstats->rx_packets++; 2647 u64_stats_update_end(&tstats->syncp); 2648 } 2649 2650 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2651 unsigned int packets, 2652 unsigned int len) 2653 { 2654 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2655 2656 u64_stats_update_begin(&tstats->syncp); 2657 tstats->tx_bytes += len; 2658 tstats->tx_packets += packets; 2659 u64_stats_update_end(&tstats->syncp); 2660 } 2661 2662 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2663 { 2664 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2665 2666 u64_stats_update_begin(&lstats->syncp); 2667 u64_stats_add(&lstats->bytes, len); 2668 u64_stats_inc(&lstats->packets); 2669 u64_stats_update_end(&lstats->syncp); 2670 } 2671 2672 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2673 ({ \ 2674 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2675 if (pcpu_stats) { \ 2676 int __cpu; \ 2677 for_each_possible_cpu(__cpu) { \ 2678 typeof(type) *stat; \ 2679 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2680 u64_stats_init(&stat->syncp); \ 2681 } \ 2682 } \ 2683 pcpu_stats; \ 2684 }) 2685 2686 #define netdev_alloc_pcpu_stats(type) \ 2687 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2688 2689 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2690 ({ \ 2691 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2692 if (pcpu_stats) { \ 2693 int __cpu; \ 2694 for_each_possible_cpu(__cpu) { \ 2695 typeof(type) *stat; \ 2696 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2697 u64_stats_init(&stat->syncp); \ 2698 } \ 2699 } \ 2700 pcpu_stats; \ 2701 }) 2702 2703 enum netdev_lag_tx_type { 2704 NETDEV_LAG_TX_TYPE_UNKNOWN, 2705 NETDEV_LAG_TX_TYPE_RANDOM, 2706 NETDEV_LAG_TX_TYPE_BROADCAST, 2707 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2708 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2709 NETDEV_LAG_TX_TYPE_HASH, 2710 }; 2711 2712 enum netdev_lag_hash { 2713 NETDEV_LAG_HASH_NONE, 2714 NETDEV_LAG_HASH_L2, 2715 NETDEV_LAG_HASH_L34, 2716 NETDEV_LAG_HASH_L23, 2717 NETDEV_LAG_HASH_E23, 2718 NETDEV_LAG_HASH_E34, 2719 NETDEV_LAG_HASH_VLAN_SRCMAC, 2720 NETDEV_LAG_HASH_UNKNOWN, 2721 }; 2722 2723 struct netdev_lag_upper_info { 2724 enum netdev_lag_tx_type tx_type; 2725 enum netdev_lag_hash hash_type; 2726 }; 2727 2728 struct netdev_lag_lower_state_info { 2729 u8 link_up : 1, 2730 tx_enabled : 1; 2731 }; 2732 2733 #include <linux/notifier.h> 2734 2735 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2736 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2737 * adding new types. 2738 */ 2739 enum netdev_cmd { 2740 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2741 NETDEV_DOWN, 2742 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2743 detected a hardware crash and restarted 2744 - we can use this eg to kick tcp sessions 2745 once done */ 2746 NETDEV_CHANGE, /* Notify device state change */ 2747 NETDEV_REGISTER, 2748 NETDEV_UNREGISTER, 2749 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2750 NETDEV_CHANGEADDR, /* notify after the address change */ 2751 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2752 NETDEV_GOING_DOWN, 2753 NETDEV_CHANGENAME, 2754 NETDEV_FEAT_CHANGE, 2755 NETDEV_BONDING_FAILOVER, 2756 NETDEV_PRE_UP, 2757 NETDEV_PRE_TYPE_CHANGE, 2758 NETDEV_POST_TYPE_CHANGE, 2759 NETDEV_POST_INIT, 2760 NETDEV_RELEASE, 2761 NETDEV_NOTIFY_PEERS, 2762 NETDEV_JOIN, 2763 NETDEV_CHANGEUPPER, 2764 NETDEV_RESEND_IGMP, 2765 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2766 NETDEV_CHANGEINFODATA, 2767 NETDEV_BONDING_INFO, 2768 NETDEV_PRECHANGEUPPER, 2769 NETDEV_CHANGELOWERSTATE, 2770 NETDEV_UDP_TUNNEL_PUSH_INFO, 2771 NETDEV_UDP_TUNNEL_DROP_INFO, 2772 NETDEV_CHANGE_TX_QUEUE_LEN, 2773 NETDEV_CVLAN_FILTER_PUSH_INFO, 2774 NETDEV_CVLAN_FILTER_DROP_INFO, 2775 NETDEV_SVLAN_FILTER_PUSH_INFO, 2776 NETDEV_SVLAN_FILTER_DROP_INFO, 2777 NETDEV_OFFLOAD_XSTATS_ENABLE, 2778 NETDEV_OFFLOAD_XSTATS_DISABLE, 2779 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 2780 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 2781 }; 2782 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2783 2784 int register_netdevice_notifier(struct notifier_block *nb); 2785 int unregister_netdevice_notifier(struct notifier_block *nb); 2786 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2787 int unregister_netdevice_notifier_net(struct net *net, 2788 struct notifier_block *nb); 2789 int register_netdevice_notifier_dev_net(struct net_device *dev, 2790 struct notifier_block *nb, 2791 struct netdev_net_notifier *nn); 2792 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2793 struct notifier_block *nb, 2794 struct netdev_net_notifier *nn); 2795 2796 struct netdev_notifier_info { 2797 struct net_device *dev; 2798 struct netlink_ext_ack *extack; 2799 }; 2800 2801 struct netdev_notifier_info_ext { 2802 struct netdev_notifier_info info; /* must be first */ 2803 union { 2804 u32 mtu; 2805 } ext; 2806 }; 2807 2808 struct netdev_notifier_change_info { 2809 struct netdev_notifier_info info; /* must be first */ 2810 unsigned int flags_changed; 2811 }; 2812 2813 struct netdev_notifier_changeupper_info { 2814 struct netdev_notifier_info info; /* must be first */ 2815 struct net_device *upper_dev; /* new upper dev */ 2816 bool master; /* is upper dev master */ 2817 bool linking; /* is the notification for link or unlink */ 2818 void *upper_info; /* upper dev info */ 2819 }; 2820 2821 struct netdev_notifier_changelowerstate_info { 2822 struct netdev_notifier_info info; /* must be first */ 2823 void *lower_state_info; /* is lower dev state */ 2824 }; 2825 2826 struct netdev_notifier_pre_changeaddr_info { 2827 struct netdev_notifier_info info; /* must be first */ 2828 const unsigned char *dev_addr; 2829 }; 2830 2831 enum netdev_offload_xstats_type { 2832 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 2833 }; 2834 2835 struct netdev_notifier_offload_xstats_info { 2836 struct netdev_notifier_info info; /* must be first */ 2837 enum netdev_offload_xstats_type type; 2838 2839 union { 2840 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 2841 struct netdev_notifier_offload_xstats_rd *report_delta; 2842 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 2843 struct netdev_notifier_offload_xstats_ru *report_used; 2844 }; 2845 }; 2846 2847 int netdev_offload_xstats_enable(struct net_device *dev, 2848 enum netdev_offload_xstats_type type, 2849 struct netlink_ext_ack *extack); 2850 int netdev_offload_xstats_disable(struct net_device *dev, 2851 enum netdev_offload_xstats_type type); 2852 bool netdev_offload_xstats_enabled(const struct net_device *dev, 2853 enum netdev_offload_xstats_type type); 2854 int netdev_offload_xstats_get(struct net_device *dev, 2855 enum netdev_offload_xstats_type type, 2856 struct rtnl_hw_stats64 *stats, bool *used, 2857 struct netlink_ext_ack *extack); 2858 void 2859 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 2860 const struct rtnl_hw_stats64 *stats); 2861 void 2862 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 2863 void netdev_offload_xstats_push_delta(struct net_device *dev, 2864 enum netdev_offload_xstats_type type, 2865 const struct rtnl_hw_stats64 *stats); 2866 2867 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2868 struct net_device *dev) 2869 { 2870 info->dev = dev; 2871 info->extack = NULL; 2872 } 2873 2874 static inline struct net_device * 2875 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2876 { 2877 return info->dev; 2878 } 2879 2880 static inline struct netlink_ext_ack * 2881 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2882 { 2883 return info->extack; 2884 } 2885 2886 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2887 2888 2889 extern rwlock_t dev_base_lock; /* Device list lock */ 2890 2891 #define for_each_netdev(net, d) \ 2892 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2893 #define for_each_netdev_reverse(net, d) \ 2894 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2895 #define for_each_netdev_rcu(net, d) \ 2896 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2897 #define for_each_netdev_safe(net, d, n) \ 2898 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2899 #define for_each_netdev_continue(net, d) \ 2900 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2901 #define for_each_netdev_continue_reverse(net, d) \ 2902 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2903 dev_list) 2904 #define for_each_netdev_continue_rcu(net, d) \ 2905 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2906 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2907 for_each_netdev_rcu(&init_net, slave) \ 2908 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2909 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2910 2911 static inline struct net_device *next_net_device(struct net_device *dev) 2912 { 2913 struct list_head *lh; 2914 struct net *net; 2915 2916 net = dev_net(dev); 2917 lh = dev->dev_list.next; 2918 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2919 } 2920 2921 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2922 { 2923 struct list_head *lh; 2924 struct net *net; 2925 2926 net = dev_net(dev); 2927 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2928 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2929 } 2930 2931 static inline struct net_device *first_net_device(struct net *net) 2932 { 2933 return list_empty(&net->dev_base_head) ? NULL : 2934 net_device_entry(net->dev_base_head.next); 2935 } 2936 2937 static inline struct net_device *first_net_device_rcu(struct net *net) 2938 { 2939 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2940 2941 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2942 } 2943 2944 int netdev_boot_setup_check(struct net_device *dev); 2945 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2946 const char *hwaddr); 2947 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2948 void dev_add_pack(struct packet_type *pt); 2949 void dev_remove_pack(struct packet_type *pt); 2950 void __dev_remove_pack(struct packet_type *pt); 2951 void dev_add_offload(struct packet_offload *po); 2952 void dev_remove_offload(struct packet_offload *po); 2953 2954 int dev_get_iflink(const struct net_device *dev); 2955 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2956 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 2957 struct net_device_path_stack *stack); 2958 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2959 unsigned short mask); 2960 struct net_device *dev_get_by_name(struct net *net, const char *name); 2961 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2962 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2963 bool netdev_name_in_use(struct net *net, const char *name); 2964 int dev_alloc_name(struct net_device *dev, const char *name); 2965 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2966 void dev_close(struct net_device *dev); 2967 void dev_close_many(struct list_head *head, bool unlink); 2968 void dev_disable_lro(struct net_device *dev); 2969 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2970 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2971 struct net_device *sb_dev); 2972 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2973 struct net_device *sb_dev); 2974 2975 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 2976 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2977 2978 static inline int dev_queue_xmit(struct sk_buff *skb) 2979 { 2980 return __dev_queue_xmit(skb, NULL); 2981 } 2982 2983 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 2984 struct net_device *sb_dev) 2985 { 2986 return __dev_queue_xmit(skb, sb_dev); 2987 } 2988 2989 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 2990 { 2991 int ret; 2992 2993 ret = __dev_direct_xmit(skb, queue_id); 2994 if (!dev_xmit_complete(ret)) 2995 kfree_skb(skb); 2996 return ret; 2997 } 2998 2999 int register_netdevice(struct net_device *dev); 3000 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3001 void unregister_netdevice_many(struct list_head *head); 3002 static inline void unregister_netdevice(struct net_device *dev) 3003 { 3004 unregister_netdevice_queue(dev, NULL); 3005 } 3006 3007 int netdev_refcnt_read(const struct net_device *dev); 3008 void free_netdev(struct net_device *dev); 3009 void netdev_freemem(struct net_device *dev); 3010 int init_dummy_netdev(struct net_device *dev); 3011 3012 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3013 struct sk_buff *skb, 3014 bool all_slaves); 3015 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3016 struct sock *sk); 3017 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3018 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3019 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3020 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 3021 int dev_restart(struct net_device *dev); 3022 3023 3024 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3025 unsigned short type, 3026 const void *daddr, const void *saddr, 3027 unsigned int len) 3028 { 3029 if (!dev->header_ops || !dev->header_ops->create) 3030 return 0; 3031 3032 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3033 } 3034 3035 static inline int dev_parse_header(const struct sk_buff *skb, 3036 unsigned char *haddr) 3037 { 3038 const struct net_device *dev = skb->dev; 3039 3040 if (!dev->header_ops || !dev->header_ops->parse) 3041 return 0; 3042 return dev->header_ops->parse(skb, haddr); 3043 } 3044 3045 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3046 { 3047 const struct net_device *dev = skb->dev; 3048 3049 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3050 return 0; 3051 return dev->header_ops->parse_protocol(skb); 3052 } 3053 3054 /* ll_header must have at least hard_header_len allocated */ 3055 static inline bool dev_validate_header(const struct net_device *dev, 3056 char *ll_header, int len) 3057 { 3058 if (likely(len >= dev->hard_header_len)) 3059 return true; 3060 if (len < dev->min_header_len) 3061 return false; 3062 3063 if (capable(CAP_SYS_RAWIO)) { 3064 memset(ll_header + len, 0, dev->hard_header_len - len); 3065 return true; 3066 } 3067 3068 if (dev->header_ops && dev->header_ops->validate) 3069 return dev->header_ops->validate(ll_header, len); 3070 3071 return false; 3072 } 3073 3074 static inline bool dev_has_header(const struct net_device *dev) 3075 { 3076 return dev->header_ops && dev->header_ops->create; 3077 } 3078 3079 /* 3080 * Incoming packets are placed on per-CPU queues 3081 */ 3082 struct softnet_data { 3083 struct list_head poll_list; 3084 struct sk_buff_head process_queue; 3085 3086 /* stats */ 3087 unsigned int processed; 3088 unsigned int time_squeeze; 3089 unsigned int received_rps; 3090 #ifdef CONFIG_RPS 3091 struct softnet_data *rps_ipi_list; 3092 #endif 3093 #ifdef CONFIG_NET_FLOW_LIMIT 3094 struct sd_flow_limit __rcu *flow_limit; 3095 #endif 3096 struct Qdisc *output_queue; 3097 struct Qdisc **output_queue_tailp; 3098 struct sk_buff *completion_queue; 3099 #ifdef CONFIG_XFRM_OFFLOAD 3100 struct sk_buff_head xfrm_backlog; 3101 #endif 3102 /* written and read only by owning cpu: */ 3103 struct { 3104 u16 recursion; 3105 u8 more; 3106 #ifdef CONFIG_NET_EGRESS 3107 u8 skip_txqueue; 3108 #endif 3109 } xmit; 3110 #ifdef CONFIG_RPS 3111 /* input_queue_head should be written by cpu owning this struct, 3112 * and only read by other cpus. Worth using a cache line. 3113 */ 3114 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3115 3116 /* Elements below can be accessed between CPUs for RPS/RFS */ 3117 call_single_data_t csd ____cacheline_aligned_in_smp; 3118 struct softnet_data *rps_ipi_next; 3119 unsigned int cpu; 3120 unsigned int input_queue_tail; 3121 #endif 3122 unsigned int dropped; 3123 struct sk_buff_head input_pkt_queue; 3124 struct napi_struct backlog; 3125 3126 /* Another possibly contended cache line */ 3127 spinlock_t defer_lock ____cacheline_aligned_in_smp; 3128 int defer_count; 3129 struct sk_buff *defer_list; 3130 call_single_data_t defer_csd; 3131 }; 3132 3133 static inline void input_queue_head_incr(struct softnet_data *sd) 3134 { 3135 #ifdef CONFIG_RPS 3136 sd->input_queue_head++; 3137 #endif 3138 } 3139 3140 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3141 unsigned int *qtail) 3142 { 3143 #ifdef CONFIG_RPS 3144 *qtail = ++sd->input_queue_tail; 3145 #endif 3146 } 3147 3148 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3149 3150 static inline int dev_recursion_level(void) 3151 { 3152 return this_cpu_read(softnet_data.xmit.recursion); 3153 } 3154 3155 #define XMIT_RECURSION_LIMIT 8 3156 static inline bool dev_xmit_recursion(void) 3157 { 3158 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3159 XMIT_RECURSION_LIMIT); 3160 } 3161 3162 static inline void dev_xmit_recursion_inc(void) 3163 { 3164 __this_cpu_inc(softnet_data.xmit.recursion); 3165 } 3166 3167 static inline void dev_xmit_recursion_dec(void) 3168 { 3169 __this_cpu_dec(softnet_data.xmit.recursion); 3170 } 3171 3172 void __netif_schedule(struct Qdisc *q); 3173 void netif_schedule_queue(struct netdev_queue *txq); 3174 3175 static inline void netif_tx_schedule_all(struct net_device *dev) 3176 { 3177 unsigned int i; 3178 3179 for (i = 0; i < dev->num_tx_queues; i++) 3180 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3181 } 3182 3183 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3184 { 3185 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3186 } 3187 3188 /** 3189 * netif_start_queue - allow transmit 3190 * @dev: network device 3191 * 3192 * Allow upper layers to call the device hard_start_xmit routine. 3193 */ 3194 static inline void netif_start_queue(struct net_device *dev) 3195 { 3196 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3197 } 3198 3199 static inline void netif_tx_start_all_queues(struct net_device *dev) 3200 { 3201 unsigned int i; 3202 3203 for (i = 0; i < dev->num_tx_queues; i++) { 3204 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3205 netif_tx_start_queue(txq); 3206 } 3207 } 3208 3209 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3210 3211 /** 3212 * netif_wake_queue - restart transmit 3213 * @dev: network device 3214 * 3215 * Allow upper layers to call the device hard_start_xmit routine. 3216 * Used for flow control when transmit resources are available. 3217 */ 3218 static inline void netif_wake_queue(struct net_device *dev) 3219 { 3220 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3221 } 3222 3223 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3224 { 3225 unsigned int i; 3226 3227 for (i = 0; i < dev->num_tx_queues; i++) { 3228 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3229 netif_tx_wake_queue(txq); 3230 } 3231 } 3232 3233 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3234 { 3235 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3236 } 3237 3238 /** 3239 * netif_stop_queue - stop transmitted packets 3240 * @dev: network device 3241 * 3242 * Stop upper layers calling the device hard_start_xmit routine. 3243 * Used for flow control when transmit resources are unavailable. 3244 */ 3245 static inline void netif_stop_queue(struct net_device *dev) 3246 { 3247 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3248 } 3249 3250 void netif_tx_stop_all_queues(struct net_device *dev); 3251 3252 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3253 { 3254 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3255 } 3256 3257 /** 3258 * netif_queue_stopped - test if transmit queue is flowblocked 3259 * @dev: network device 3260 * 3261 * Test if transmit queue on device is currently unable to send. 3262 */ 3263 static inline bool netif_queue_stopped(const struct net_device *dev) 3264 { 3265 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3266 } 3267 3268 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3269 { 3270 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3271 } 3272 3273 static inline bool 3274 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3275 { 3276 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3277 } 3278 3279 static inline bool 3280 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3281 { 3282 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3283 } 3284 3285 /** 3286 * netdev_queue_set_dql_min_limit - set dql minimum limit 3287 * @dev_queue: pointer to transmit queue 3288 * @min_limit: dql minimum limit 3289 * 3290 * Forces xmit_more() to return true until the minimum threshold 3291 * defined by @min_limit is reached (or until the tx queue is 3292 * empty). Warning: to be use with care, misuse will impact the 3293 * latency. 3294 */ 3295 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3296 unsigned int min_limit) 3297 { 3298 #ifdef CONFIG_BQL 3299 dev_queue->dql.min_limit = min_limit; 3300 #endif 3301 } 3302 3303 /** 3304 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3305 * @dev_queue: pointer to transmit queue 3306 * 3307 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3308 * to give appropriate hint to the CPU. 3309 */ 3310 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3311 { 3312 #ifdef CONFIG_BQL 3313 prefetchw(&dev_queue->dql.num_queued); 3314 #endif 3315 } 3316 3317 /** 3318 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3319 * @dev_queue: pointer to transmit queue 3320 * 3321 * BQL enabled drivers might use this helper in their TX completion path, 3322 * to give appropriate hint to the CPU. 3323 */ 3324 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3325 { 3326 #ifdef CONFIG_BQL 3327 prefetchw(&dev_queue->dql.limit); 3328 #endif 3329 } 3330 3331 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3332 unsigned int bytes) 3333 { 3334 #ifdef CONFIG_BQL 3335 dql_queued(&dev_queue->dql, bytes); 3336 3337 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3338 return; 3339 3340 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3341 3342 /* 3343 * The XOFF flag must be set before checking the dql_avail below, 3344 * because in netdev_tx_completed_queue we update the dql_completed 3345 * before checking the XOFF flag. 3346 */ 3347 smp_mb(); 3348 3349 /* check again in case another CPU has just made room avail */ 3350 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3351 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3352 #endif 3353 } 3354 3355 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3356 * that they should not test BQL status themselves. 3357 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3358 * skb of a batch. 3359 * Returns true if the doorbell must be used to kick the NIC. 3360 */ 3361 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3362 unsigned int bytes, 3363 bool xmit_more) 3364 { 3365 if (xmit_more) { 3366 #ifdef CONFIG_BQL 3367 dql_queued(&dev_queue->dql, bytes); 3368 #endif 3369 return netif_tx_queue_stopped(dev_queue); 3370 } 3371 netdev_tx_sent_queue(dev_queue, bytes); 3372 return true; 3373 } 3374 3375 /** 3376 * netdev_sent_queue - report the number of bytes queued to hardware 3377 * @dev: network device 3378 * @bytes: number of bytes queued to the hardware device queue 3379 * 3380 * Report the number of bytes queued for sending/completion to the network 3381 * device hardware queue. @bytes should be a good approximation and should 3382 * exactly match netdev_completed_queue() @bytes 3383 */ 3384 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3385 { 3386 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3387 } 3388 3389 static inline bool __netdev_sent_queue(struct net_device *dev, 3390 unsigned int bytes, 3391 bool xmit_more) 3392 { 3393 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3394 xmit_more); 3395 } 3396 3397 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3398 unsigned int pkts, unsigned int bytes) 3399 { 3400 #ifdef CONFIG_BQL 3401 if (unlikely(!bytes)) 3402 return; 3403 3404 dql_completed(&dev_queue->dql, bytes); 3405 3406 /* 3407 * Without the memory barrier there is a small possiblity that 3408 * netdev_tx_sent_queue will miss the update and cause the queue to 3409 * be stopped forever 3410 */ 3411 smp_mb(); 3412 3413 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3414 return; 3415 3416 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3417 netif_schedule_queue(dev_queue); 3418 #endif 3419 } 3420 3421 /** 3422 * netdev_completed_queue - report bytes and packets completed by device 3423 * @dev: network device 3424 * @pkts: actual number of packets sent over the medium 3425 * @bytes: actual number of bytes sent over the medium 3426 * 3427 * Report the number of bytes and packets transmitted by the network device 3428 * hardware queue over the physical medium, @bytes must exactly match the 3429 * @bytes amount passed to netdev_sent_queue() 3430 */ 3431 static inline void netdev_completed_queue(struct net_device *dev, 3432 unsigned int pkts, unsigned int bytes) 3433 { 3434 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3435 } 3436 3437 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3438 { 3439 #ifdef CONFIG_BQL 3440 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3441 dql_reset(&q->dql); 3442 #endif 3443 } 3444 3445 /** 3446 * netdev_reset_queue - reset the packets and bytes count of a network device 3447 * @dev_queue: network device 3448 * 3449 * Reset the bytes and packet count of a network device and clear the 3450 * software flow control OFF bit for this network device 3451 */ 3452 static inline void netdev_reset_queue(struct net_device *dev_queue) 3453 { 3454 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3455 } 3456 3457 /** 3458 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3459 * @dev: network device 3460 * @queue_index: given tx queue index 3461 * 3462 * Returns 0 if given tx queue index >= number of device tx queues, 3463 * otherwise returns the originally passed tx queue index. 3464 */ 3465 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3466 { 3467 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3468 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3469 dev->name, queue_index, 3470 dev->real_num_tx_queues); 3471 return 0; 3472 } 3473 3474 return queue_index; 3475 } 3476 3477 /** 3478 * netif_running - test if up 3479 * @dev: network device 3480 * 3481 * Test if the device has been brought up. 3482 */ 3483 static inline bool netif_running(const struct net_device *dev) 3484 { 3485 return test_bit(__LINK_STATE_START, &dev->state); 3486 } 3487 3488 /* 3489 * Routines to manage the subqueues on a device. We only need start, 3490 * stop, and a check if it's stopped. All other device management is 3491 * done at the overall netdevice level. 3492 * Also test the device if we're multiqueue. 3493 */ 3494 3495 /** 3496 * netif_start_subqueue - allow sending packets on subqueue 3497 * @dev: network device 3498 * @queue_index: sub queue index 3499 * 3500 * Start individual transmit queue of a device with multiple transmit queues. 3501 */ 3502 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3503 { 3504 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3505 3506 netif_tx_start_queue(txq); 3507 } 3508 3509 /** 3510 * netif_stop_subqueue - stop sending packets on subqueue 3511 * @dev: network device 3512 * @queue_index: sub queue index 3513 * 3514 * Stop individual transmit queue of a device with multiple transmit queues. 3515 */ 3516 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3517 { 3518 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3519 netif_tx_stop_queue(txq); 3520 } 3521 3522 /** 3523 * __netif_subqueue_stopped - test status of subqueue 3524 * @dev: network device 3525 * @queue_index: sub queue index 3526 * 3527 * Check individual transmit queue of a device with multiple transmit queues. 3528 */ 3529 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3530 u16 queue_index) 3531 { 3532 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3533 3534 return netif_tx_queue_stopped(txq); 3535 } 3536 3537 /** 3538 * netif_subqueue_stopped - test status of subqueue 3539 * @dev: network device 3540 * @skb: sub queue buffer pointer 3541 * 3542 * Check individual transmit queue of a device with multiple transmit queues. 3543 */ 3544 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3545 struct sk_buff *skb) 3546 { 3547 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3548 } 3549 3550 /** 3551 * netif_wake_subqueue - allow sending packets on subqueue 3552 * @dev: network device 3553 * @queue_index: sub queue index 3554 * 3555 * Resume individual transmit queue of a device with multiple transmit queues. 3556 */ 3557 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3558 { 3559 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3560 3561 netif_tx_wake_queue(txq); 3562 } 3563 3564 #ifdef CONFIG_XPS 3565 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3566 u16 index); 3567 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3568 u16 index, enum xps_map_type type); 3569 3570 /** 3571 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3572 * @j: CPU/Rx queue index 3573 * @mask: bitmask of all cpus/rx queues 3574 * @nr_bits: number of bits in the bitmask 3575 * 3576 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3577 */ 3578 static inline bool netif_attr_test_mask(unsigned long j, 3579 const unsigned long *mask, 3580 unsigned int nr_bits) 3581 { 3582 cpu_max_bits_warn(j, nr_bits); 3583 return test_bit(j, mask); 3584 } 3585 3586 /** 3587 * netif_attr_test_online - Test for online CPU/Rx queue 3588 * @j: CPU/Rx queue index 3589 * @online_mask: bitmask for CPUs/Rx queues that are online 3590 * @nr_bits: number of bits in the bitmask 3591 * 3592 * Returns true if a CPU/Rx queue is online. 3593 */ 3594 static inline bool netif_attr_test_online(unsigned long j, 3595 const unsigned long *online_mask, 3596 unsigned int nr_bits) 3597 { 3598 cpu_max_bits_warn(j, nr_bits); 3599 3600 if (online_mask) 3601 return test_bit(j, online_mask); 3602 3603 return (j < nr_bits); 3604 } 3605 3606 /** 3607 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3608 * @n: CPU/Rx queue index 3609 * @srcp: the cpumask/Rx queue mask pointer 3610 * @nr_bits: number of bits in the bitmask 3611 * 3612 * Returns >= nr_bits if no further CPUs/Rx queues set. 3613 */ 3614 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3615 unsigned int nr_bits) 3616 { 3617 /* -1 is a legal arg here. */ 3618 if (n != -1) 3619 cpu_max_bits_warn(n, nr_bits); 3620 3621 if (srcp) 3622 return find_next_bit(srcp, nr_bits, n + 1); 3623 3624 return n + 1; 3625 } 3626 3627 /** 3628 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3629 * @n: CPU/Rx queue index 3630 * @src1p: the first CPUs/Rx queues mask pointer 3631 * @src2p: the second CPUs/Rx queues mask pointer 3632 * @nr_bits: number of bits in the bitmask 3633 * 3634 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3635 */ 3636 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3637 const unsigned long *src2p, 3638 unsigned int nr_bits) 3639 { 3640 /* -1 is a legal arg here. */ 3641 if (n != -1) 3642 cpu_max_bits_warn(n, nr_bits); 3643 3644 if (src1p && src2p) 3645 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3646 else if (src1p) 3647 return find_next_bit(src1p, nr_bits, n + 1); 3648 else if (src2p) 3649 return find_next_bit(src2p, nr_bits, n + 1); 3650 3651 return n + 1; 3652 } 3653 #else 3654 static inline int netif_set_xps_queue(struct net_device *dev, 3655 const struct cpumask *mask, 3656 u16 index) 3657 { 3658 return 0; 3659 } 3660 3661 static inline int __netif_set_xps_queue(struct net_device *dev, 3662 const unsigned long *mask, 3663 u16 index, enum xps_map_type type) 3664 { 3665 return 0; 3666 } 3667 #endif 3668 3669 /** 3670 * netif_is_multiqueue - test if device has multiple transmit queues 3671 * @dev: network device 3672 * 3673 * Check if device has multiple transmit queues 3674 */ 3675 static inline bool netif_is_multiqueue(const struct net_device *dev) 3676 { 3677 return dev->num_tx_queues > 1; 3678 } 3679 3680 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3681 3682 #ifdef CONFIG_SYSFS 3683 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3684 #else 3685 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3686 unsigned int rxqs) 3687 { 3688 dev->real_num_rx_queues = rxqs; 3689 return 0; 3690 } 3691 #endif 3692 int netif_set_real_num_queues(struct net_device *dev, 3693 unsigned int txq, unsigned int rxq); 3694 3695 static inline struct netdev_rx_queue * 3696 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3697 { 3698 return dev->_rx + rxq; 3699 } 3700 3701 #ifdef CONFIG_SYSFS 3702 static inline unsigned int get_netdev_rx_queue_index( 3703 struct netdev_rx_queue *queue) 3704 { 3705 struct net_device *dev = queue->dev; 3706 int index = queue - dev->_rx; 3707 3708 BUG_ON(index >= dev->num_rx_queues); 3709 return index; 3710 } 3711 #endif 3712 3713 int netif_get_num_default_rss_queues(void); 3714 3715 enum skb_free_reason { 3716 SKB_REASON_CONSUMED, 3717 SKB_REASON_DROPPED, 3718 }; 3719 3720 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3721 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3722 3723 /* 3724 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3725 * interrupt context or with hardware interrupts being disabled. 3726 * (in_hardirq() || irqs_disabled()) 3727 * 3728 * We provide four helpers that can be used in following contexts : 3729 * 3730 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3731 * replacing kfree_skb(skb) 3732 * 3733 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3734 * Typically used in place of consume_skb(skb) in TX completion path 3735 * 3736 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3737 * replacing kfree_skb(skb) 3738 * 3739 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3740 * and consumed a packet. Used in place of consume_skb(skb) 3741 */ 3742 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3743 { 3744 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3745 } 3746 3747 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3748 { 3749 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3750 } 3751 3752 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3753 { 3754 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3755 } 3756 3757 static inline void dev_consume_skb_any(struct sk_buff *skb) 3758 { 3759 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3760 } 3761 3762 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3763 struct bpf_prog *xdp_prog); 3764 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3765 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3766 int netif_rx(struct sk_buff *skb); 3767 int __netif_rx(struct sk_buff *skb); 3768 3769 int netif_receive_skb(struct sk_buff *skb); 3770 int netif_receive_skb_core(struct sk_buff *skb); 3771 void netif_receive_skb_list_internal(struct list_head *head); 3772 void netif_receive_skb_list(struct list_head *head); 3773 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3774 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3775 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3776 gro_result_t napi_gro_frags(struct napi_struct *napi); 3777 struct packet_offload *gro_find_receive_by_type(__be16 type); 3778 struct packet_offload *gro_find_complete_by_type(__be16 type); 3779 3780 static inline void napi_free_frags(struct napi_struct *napi) 3781 { 3782 kfree_skb(napi->skb); 3783 napi->skb = NULL; 3784 } 3785 3786 bool netdev_is_rx_handler_busy(struct net_device *dev); 3787 int netdev_rx_handler_register(struct net_device *dev, 3788 rx_handler_func_t *rx_handler, 3789 void *rx_handler_data); 3790 void netdev_rx_handler_unregister(struct net_device *dev); 3791 3792 bool dev_valid_name(const char *name); 3793 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 3794 { 3795 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 3796 } 3797 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 3798 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 3799 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3800 void __user *data, bool *need_copyout); 3801 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 3802 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 3803 unsigned int dev_get_flags(const struct net_device *); 3804 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3805 struct netlink_ext_ack *extack); 3806 int dev_change_flags(struct net_device *dev, unsigned int flags, 3807 struct netlink_ext_ack *extack); 3808 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3809 unsigned int gchanges); 3810 int dev_set_alias(struct net_device *, const char *, size_t); 3811 int dev_get_alias(const struct net_device *, char *, size_t); 3812 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 3813 const char *pat, int new_ifindex); 3814 static inline 3815 int dev_change_net_namespace(struct net_device *dev, struct net *net, 3816 const char *pat) 3817 { 3818 return __dev_change_net_namespace(dev, net, pat, 0); 3819 } 3820 int __dev_set_mtu(struct net_device *, int); 3821 int dev_set_mtu(struct net_device *, int); 3822 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3823 struct netlink_ext_ack *extack); 3824 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3825 struct netlink_ext_ack *extack); 3826 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 3827 struct netlink_ext_ack *extack); 3828 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 3829 int dev_get_port_parent_id(struct net_device *dev, 3830 struct netdev_phys_item_id *ppid, bool recurse); 3831 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3832 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3833 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3834 struct netdev_queue *txq, int *ret); 3835 3836 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3837 u8 dev_xdp_prog_count(struct net_device *dev); 3838 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3839 3840 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3841 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3842 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 3843 bool is_skb_forwardable(const struct net_device *dev, 3844 const struct sk_buff *skb); 3845 3846 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 3847 const struct sk_buff *skb, 3848 const bool check_mtu) 3849 { 3850 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 3851 unsigned int len; 3852 3853 if (!(dev->flags & IFF_UP)) 3854 return false; 3855 3856 if (!check_mtu) 3857 return true; 3858 3859 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 3860 if (skb->len <= len) 3861 return true; 3862 3863 /* if TSO is enabled, we don't care about the length as the packet 3864 * could be forwarded without being segmented before 3865 */ 3866 if (skb_is_gso(skb)) 3867 return true; 3868 3869 return false; 3870 } 3871 3872 struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev); 3873 3874 static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev) 3875 { 3876 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */ 3877 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats); 3878 3879 if (likely(p)) 3880 return p; 3881 3882 return netdev_core_stats_alloc(dev); 3883 } 3884 3885 #define DEV_CORE_STATS_INC(FIELD) \ 3886 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 3887 { \ 3888 struct net_device_core_stats __percpu *p; \ 3889 \ 3890 p = dev_core_stats(dev); \ 3891 if (p) \ 3892 this_cpu_inc(p->FIELD); \ 3893 } 3894 DEV_CORE_STATS_INC(rx_dropped) 3895 DEV_CORE_STATS_INC(tx_dropped) 3896 DEV_CORE_STATS_INC(rx_nohandler) 3897 DEV_CORE_STATS_INC(rx_otherhost_dropped) 3898 3899 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3900 struct sk_buff *skb, 3901 const bool check_mtu) 3902 { 3903 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3904 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 3905 dev_core_stats_rx_dropped_inc(dev); 3906 kfree_skb(skb); 3907 return NET_RX_DROP; 3908 } 3909 3910 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 3911 skb->priority = 0; 3912 return 0; 3913 } 3914 3915 bool dev_nit_active(struct net_device *dev); 3916 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3917 3918 static inline void __dev_put(struct net_device *dev) 3919 { 3920 if (dev) { 3921 #ifdef CONFIG_PCPU_DEV_REFCNT 3922 this_cpu_dec(*dev->pcpu_refcnt); 3923 #else 3924 refcount_dec(&dev->dev_refcnt); 3925 #endif 3926 } 3927 } 3928 3929 static inline void __dev_hold(struct net_device *dev) 3930 { 3931 if (dev) { 3932 #ifdef CONFIG_PCPU_DEV_REFCNT 3933 this_cpu_inc(*dev->pcpu_refcnt); 3934 #else 3935 refcount_inc(&dev->dev_refcnt); 3936 #endif 3937 } 3938 } 3939 3940 static inline void __netdev_tracker_alloc(struct net_device *dev, 3941 netdevice_tracker *tracker, 3942 gfp_t gfp) 3943 { 3944 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 3945 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 3946 #endif 3947 } 3948 3949 /* netdev_tracker_alloc() can upgrade a prior untracked reference 3950 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 3951 */ 3952 static inline void netdev_tracker_alloc(struct net_device *dev, 3953 netdevice_tracker *tracker, gfp_t gfp) 3954 { 3955 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 3956 refcount_dec(&dev->refcnt_tracker.no_tracker); 3957 __netdev_tracker_alloc(dev, tracker, gfp); 3958 #endif 3959 } 3960 3961 static inline void netdev_tracker_free(struct net_device *dev, 3962 netdevice_tracker *tracker) 3963 { 3964 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 3965 ref_tracker_free(&dev->refcnt_tracker, tracker); 3966 #endif 3967 } 3968 3969 static inline void dev_hold_track(struct net_device *dev, 3970 netdevice_tracker *tracker, gfp_t gfp) 3971 { 3972 if (dev) { 3973 __dev_hold(dev); 3974 __netdev_tracker_alloc(dev, tracker, gfp); 3975 } 3976 } 3977 3978 static inline void dev_put_track(struct net_device *dev, 3979 netdevice_tracker *tracker) 3980 { 3981 if (dev) { 3982 netdev_tracker_free(dev, tracker); 3983 __dev_put(dev); 3984 } 3985 } 3986 3987 /** 3988 * dev_hold - get reference to device 3989 * @dev: network device 3990 * 3991 * Hold reference to device to keep it from being freed. 3992 * Try using dev_hold_track() instead. 3993 */ 3994 static inline void dev_hold(struct net_device *dev) 3995 { 3996 dev_hold_track(dev, NULL, GFP_ATOMIC); 3997 } 3998 3999 /** 4000 * dev_put - release reference to device 4001 * @dev: network device 4002 * 4003 * Release reference to device to allow it to be freed. 4004 * Try using dev_put_track() instead. 4005 */ 4006 static inline void dev_put(struct net_device *dev) 4007 { 4008 dev_put_track(dev, NULL); 4009 } 4010 4011 static inline void dev_replace_track(struct net_device *odev, 4012 struct net_device *ndev, 4013 netdevice_tracker *tracker, 4014 gfp_t gfp) 4015 { 4016 if (odev) 4017 netdev_tracker_free(odev, tracker); 4018 4019 __dev_hold(ndev); 4020 __dev_put(odev); 4021 4022 if (ndev) 4023 __netdev_tracker_alloc(ndev, tracker, gfp); 4024 } 4025 4026 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4027 * and _off may be called from IRQ context, but it is caller 4028 * who is responsible for serialization of these calls. 4029 * 4030 * The name carrier is inappropriate, these functions should really be 4031 * called netif_lowerlayer_*() because they represent the state of any 4032 * kind of lower layer not just hardware media. 4033 */ 4034 void linkwatch_fire_event(struct net_device *dev); 4035 4036 /** 4037 * netif_carrier_ok - test if carrier present 4038 * @dev: network device 4039 * 4040 * Check if carrier is present on device 4041 */ 4042 static inline bool netif_carrier_ok(const struct net_device *dev) 4043 { 4044 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4045 } 4046 4047 unsigned long dev_trans_start(struct net_device *dev); 4048 4049 void __netdev_watchdog_up(struct net_device *dev); 4050 4051 void netif_carrier_on(struct net_device *dev); 4052 void netif_carrier_off(struct net_device *dev); 4053 void netif_carrier_event(struct net_device *dev); 4054 4055 /** 4056 * netif_dormant_on - mark device as dormant. 4057 * @dev: network device 4058 * 4059 * Mark device as dormant (as per RFC2863). 4060 * 4061 * The dormant state indicates that the relevant interface is not 4062 * actually in a condition to pass packets (i.e., it is not 'up') but is 4063 * in a "pending" state, waiting for some external event. For "on- 4064 * demand" interfaces, this new state identifies the situation where the 4065 * interface is waiting for events to place it in the up state. 4066 */ 4067 static inline void netif_dormant_on(struct net_device *dev) 4068 { 4069 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4070 linkwatch_fire_event(dev); 4071 } 4072 4073 /** 4074 * netif_dormant_off - set device as not dormant. 4075 * @dev: network device 4076 * 4077 * Device is not in dormant state. 4078 */ 4079 static inline void netif_dormant_off(struct net_device *dev) 4080 { 4081 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4082 linkwatch_fire_event(dev); 4083 } 4084 4085 /** 4086 * netif_dormant - test if device is dormant 4087 * @dev: network device 4088 * 4089 * Check if device is dormant. 4090 */ 4091 static inline bool netif_dormant(const struct net_device *dev) 4092 { 4093 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4094 } 4095 4096 4097 /** 4098 * netif_testing_on - mark device as under test. 4099 * @dev: network device 4100 * 4101 * Mark device as under test (as per RFC2863). 4102 * 4103 * The testing state indicates that some test(s) must be performed on 4104 * the interface. After completion, of the test, the interface state 4105 * will change to up, dormant, or down, as appropriate. 4106 */ 4107 static inline void netif_testing_on(struct net_device *dev) 4108 { 4109 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4110 linkwatch_fire_event(dev); 4111 } 4112 4113 /** 4114 * netif_testing_off - set device as not under test. 4115 * @dev: network device 4116 * 4117 * Device is not in testing state. 4118 */ 4119 static inline void netif_testing_off(struct net_device *dev) 4120 { 4121 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4122 linkwatch_fire_event(dev); 4123 } 4124 4125 /** 4126 * netif_testing - test if device is under test 4127 * @dev: network device 4128 * 4129 * Check if device is under test 4130 */ 4131 static inline bool netif_testing(const struct net_device *dev) 4132 { 4133 return test_bit(__LINK_STATE_TESTING, &dev->state); 4134 } 4135 4136 4137 /** 4138 * netif_oper_up - test if device is operational 4139 * @dev: network device 4140 * 4141 * Check if carrier is operational 4142 */ 4143 static inline bool netif_oper_up(const struct net_device *dev) 4144 { 4145 return (dev->operstate == IF_OPER_UP || 4146 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4147 } 4148 4149 /** 4150 * netif_device_present - is device available or removed 4151 * @dev: network device 4152 * 4153 * Check if device has not been removed from system. 4154 */ 4155 static inline bool netif_device_present(const struct net_device *dev) 4156 { 4157 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4158 } 4159 4160 void netif_device_detach(struct net_device *dev); 4161 4162 void netif_device_attach(struct net_device *dev); 4163 4164 /* 4165 * Network interface message level settings 4166 */ 4167 4168 enum { 4169 NETIF_MSG_DRV_BIT, 4170 NETIF_MSG_PROBE_BIT, 4171 NETIF_MSG_LINK_BIT, 4172 NETIF_MSG_TIMER_BIT, 4173 NETIF_MSG_IFDOWN_BIT, 4174 NETIF_MSG_IFUP_BIT, 4175 NETIF_MSG_RX_ERR_BIT, 4176 NETIF_MSG_TX_ERR_BIT, 4177 NETIF_MSG_TX_QUEUED_BIT, 4178 NETIF_MSG_INTR_BIT, 4179 NETIF_MSG_TX_DONE_BIT, 4180 NETIF_MSG_RX_STATUS_BIT, 4181 NETIF_MSG_PKTDATA_BIT, 4182 NETIF_MSG_HW_BIT, 4183 NETIF_MSG_WOL_BIT, 4184 4185 /* When you add a new bit above, update netif_msg_class_names array 4186 * in net/ethtool/common.c 4187 */ 4188 NETIF_MSG_CLASS_COUNT, 4189 }; 4190 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4191 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4192 4193 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4194 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4195 4196 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4197 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4198 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4199 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4200 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4201 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4202 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4203 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4204 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4205 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4206 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4207 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4208 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4209 #define NETIF_MSG_HW __NETIF_MSG(HW) 4210 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4211 4212 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4213 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4214 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4215 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4216 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4217 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4218 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4219 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4220 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4221 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4222 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4223 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4224 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4225 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4226 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4227 4228 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4229 { 4230 /* use default */ 4231 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4232 return default_msg_enable_bits; 4233 if (debug_value == 0) /* no output */ 4234 return 0; 4235 /* set low N bits */ 4236 return (1U << debug_value) - 1; 4237 } 4238 4239 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4240 { 4241 spin_lock(&txq->_xmit_lock); 4242 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4243 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4244 } 4245 4246 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4247 { 4248 __acquire(&txq->_xmit_lock); 4249 return true; 4250 } 4251 4252 static inline void __netif_tx_release(struct netdev_queue *txq) 4253 { 4254 __release(&txq->_xmit_lock); 4255 } 4256 4257 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4258 { 4259 spin_lock_bh(&txq->_xmit_lock); 4260 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4261 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4262 } 4263 4264 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4265 { 4266 bool ok = spin_trylock(&txq->_xmit_lock); 4267 4268 if (likely(ok)) { 4269 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4270 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4271 } 4272 return ok; 4273 } 4274 4275 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4276 { 4277 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4278 WRITE_ONCE(txq->xmit_lock_owner, -1); 4279 spin_unlock(&txq->_xmit_lock); 4280 } 4281 4282 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4283 { 4284 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4285 WRITE_ONCE(txq->xmit_lock_owner, -1); 4286 spin_unlock_bh(&txq->_xmit_lock); 4287 } 4288 4289 /* 4290 * txq->trans_start can be read locklessly from dev_watchdog() 4291 */ 4292 static inline void txq_trans_update(struct netdev_queue *txq) 4293 { 4294 if (txq->xmit_lock_owner != -1) 4295 WRITE_ONCE(txq->trans_start, jiffies); 4296 } 4297 4298 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4299 { 4300 unsigned long now = jiffies; 4301 4302 if (READ_ONCE(txq->trans_start) != now) 4303 WRITE_ONCE(txq->trans_start, now); 4304 } 4305 4306 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4307 static inline void netif_trans_update(struct net_device *dev) 4308 { 4309 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4310 4311 txq_trans_cond_update(txq); 4312 } 4313 4314 /** 4315 * netif_tx_lock - grab network device transmit lock 4316 * @dev: network device 4317 * 4318 * Get network device transmit lock 4319 */ 4320 void netif_tx_lock(struct net_device *dev); 4321 4322 static inline void netif_tx_lock_bh(struct net_device *dev) 4323 { 4324 local_bh_disable(); 4325 netif_tx_lock(dev); 4326 } 4327 4328 void netif_tx_unlock(struct net_device *dev); 4329 4330 static inline void netif_tx_unlock_bh(struct net_device *dev) 4331 { 4332 netif_tx_unlock(dev); 4333 local_bh_enable(); 4334 } 4335 4336 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4337 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4338 __netif_tx_lock(txq, cpu); \ 4339 } else { \ 4340 __netif_tx_acquire(txq); \ 4341 } \ 4342 } 4343 4344 #define HARD_TX_TRYLOCK(dev, txq) \ 4345 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4346 __netif_tx_trylock(txq) : \ 4347 __netif_tx_acquire(txq)) 4348 4349 #define HARD_TX_UNLOCK(dev, txq) { \ 4350 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4351 __netif_tx_unlock(txq); \ 4352 } else { \ 4353 __netif_tx_release(txq); \ 4354 } \ 4355 } 4356 4357 static inline void netif_tx_disable(struct net_device *dev) 4358 { 4359 unsigned int i; 4360 int cpu; 4361 4362 local_bh_disable(); 4363 cpu = smp_processor_id(); 4364 spin_lock(&dev->tx_global_lock); 4365 for (i = 0; i < dev->num_tx_queues; i++) { 4366 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4367 4368 __netif_tx_lock(txq, cpu); 4369 netif_tx_stop_queue(txq); 4370 __netif_tx_unlock(txq); 4371 } 4372 spin_unlock(&dev->tx_global_lock); 4373 local_bh_enable(); 4374 } 4375 4376 static inline void netif_addr_lock(struct net_device *dev) 4377 { 4378 unsigned char nest_level = 0; 4379 4380 #ifdef CONFIG_LOCKDEP 4381 nest_level = dev->nested_level; 4382 #endif 4383 spin_lock_nested(&dev->addr_list_lock, nest_level); 4384 } 4385 4386 static inline void netif_addr_lock_bh(struct net_device *dev) 4387 { 4388 unsigned char nest_level = 0; 4389 4390 #ifdef CONFIG_LOCKDEP 4391 nest_level = dev->nested_level; 4392 #endif 4393 local_bh_disable(); 4394 spin_lock_nested(&dev->addr_list_lock, nest_level); 4395 } 4396 4397 static inline void netif_addr_unlock(struct net_device *dev) 4398 { 4399 spin_unlock(&dev->addr_list_lock); 4400 } 4401 4402 static inline void netif_addr_unlock_bh(struct net_device *dev) 4403 { 4404 spin_unlock_bh(&dev->addr_list_lock); 4405 } 4406 4407 /* 4408 * dev_addrs walker. Should be used only for read access. Call with 4409 * rcu_read_lock held. 4410 */ 4411 #define for_each_dev_addr(dev, ha) \ 4412 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4413 4414 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4415 4416 void ether_setup(struct net_device *dev); 4417 4418 /* Support for loadable net-drivers */ 4419 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4420 unsigned char name_assign_type, 4421 void (*setup)(struct net_device *), 4422 unsigned int txqs, unsigned int rxqs); 4423 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4424 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4425 4426 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4427 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4428 count) 4429 4430 int register_netdev(struct net_device *dev); 4431 void unregister_netdev(struct net_device *dev); 4432 4433 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4434 4435 /* General hardware address lists handling functions */ 4436 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4437 struct netdev_hw_addr_list *from_list, int addr_len); 4438 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4439 struct netdev_hw_addr_list *from_list, int addr_len); 4440 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4441 struct net_device *dev, 4442 int (*sync)(struct net_device *, const unsigned char *), 4443 int (*unsync)(struct net_device *, 4444 const unsigned char *)); 4445 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4446 struct net_device *dev, 4447 int (*sync)(struct net_device *, 4448 const unsigned char *, int), 4449 int (*unsync)(struct net_device *, 4450 const unsigned char *, int)); 4451 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4452 struct net_device *dev, 4453 int (*unsync)(struct net_device *, 4454 const unsigned char *, int)); 4455 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4456 struct net_device *dev, 4457 int (*unsync)(struct net_device *, 4458 const unsigned char *)); 4459 void __hw_addr_init(struct netdev_hw_addr_list *list); 4460 4461 /* Functions used for device addresses handling */ 4462 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4463 const void *addr, size_t len); 4464 4465 static inline void 4466 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4467 { 4468 dev_addr_mod(dev, 0, addr, len); 4469 } 4470 4471 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4472 { 4473 __dev_addr_set(dev, addr, dev->addr_len); 4474 } 4475 4476 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4477 unsigned char addr_type); 4478 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4479 unsigned char addr_type); 4480 4481 /* Functions used for unicast addresses handling */ 4482 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4483 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4484 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4485 int dev_uc_sync(struct net_device *to, struct net_device *from); 4486 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4487 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4488 void dev_uc_flush(struct net_device *dev); 4489 void dev_uc_init(struct net_device *dev); 4490 4491 /** 4492 * __dev_uc_sync - Synchonize device's unicast list 4493 * @dev: device to sync 4494 * @sync: function to call if address should be added 4495 * @unsync: function to call if address should be removed 4496 * 4497 * Add newly added addresses to the interface, and release 4498 * addresses that have been deleted. 4499 */ 4500 static inline int __dev_uc_sync(struct net_device *dev, 4501 int (*sync)(struct net_device *, 4502 const unsigned char *), 4503 int (*unsync)(struct net_device *, 4504 const unsigned char *)) 4505 { 4506 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4507 } 4508 4509 /** 4510 * __dev_uc_unsync - Remove synchronized addresses from device 4511 * @dev: device to sync 4512 * @unsync: function to call if address should be removed 4513 * 4514 * Remove all addresses that were added to the device by dev_uc_sync(). 4515 */ 4516 static inline void __dev_uc_unsync(struct net_device *dev, 4517 int (*unsync)(struct net_device *, 4518 const unsigned char *)) 4519 { 4520 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4521 } 4522 4523 /* Functions used for multicast addresses handling */ 4524 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4525 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4526 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4527 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4528 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4529 int dev_mc_sync(struct net_device *to, struct net_device *from); 4530 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4531 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4532 void dev_mc_flush(struct net_device *dev); 4533 void dev_mc_init(struct net_device *dev); 4534 4535 /** 4536 * __dev_mc_sync - Synchonize device's multicast list 4537 * @dev: device to sync 4538 * @sync: function to call if address should be added 4539 * @unsync: function to call if address should be removed 4540 * 4541 * Add newly added addresses to the interface, and release 4542 * addresses that have been deleted. 4543 */ 4544 static inline int __dev_mc_sync(struct net_device *dev, 4545 int (*sync)(struct net_device *, 4546 const unsigned char *), 4547 int (*unsync)(struct net_device *, 4548 const unsigned char *)) 4549 { 4550 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4551 } 4552 4553 /** 4554 * __dev_mc_unsync - Remove synchronized addresses from device 4555 * @dev: device to sync 4556 * @unsync: function to call if address should be removed 4557 * 4558 * Remove all addresses that were added to the device by dev_mc_sync(). 4559 */ 4560 static inline void __dev_mc_unsync(struct net_device *dev, 4561 int (*unsync)(struct net_device *, 4562 const unsigned char *)) 4563 { 4564 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4565 } 4566 4567 /* Functions used for secondary unicast and multicast support */ 4568 void dev_set_rx_mode(struct net_device *dev); 4569 int dev_set_promiscuity(struct net_device *dev, int inc); 4570 int dev_set_allmulti(struct net_device *dev, int inc); 4571 void netdev_state_change(struct net_device *dev); 4572 void __netdev_notify_peers(struct net_device *dev); 4573 void netdev_notify_peers(struct net_device *dev); 4574 void netdev_features_change(struct net_device *dev); 4575 /* Load a device via the kmod */ 4576 void dev_load(struct net *net, const char *name); 4577 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4578 struct rtnl_link_stats64 *storage); 4579 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4580 const struct net_device_stats *netdev_stats); 4581 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4582 const struct pcpu_sw_netstats __percpu *netstats); 4583 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4584 4585 extern int netdev_max_backlog; 4586 extern int dev_rx_weight; 4587 extern int dev_tx_weight; 4588 extern int gro_normal_batch; 4589 4590 enum { 4591 NESTED_SYNC_IMM_BIT, 4592 NESTED_SYNC_TODO_BIT, 4593 }; 4594 4595 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4596 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4597 4598 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4599 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4600 4601 struct netdev_nested_priv { 4602 unsigned char flags; 4603 void *data; 4604 }; 4605 4606 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4607 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4608 struct list_head **iter); 4609 4610 /* iterate through upper list, must be called under RCU read lock */ 4611 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4612 for (iter = &(dev)->adj_list.upper, \ 4613 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4614 updev; \ 4615 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4616 4617 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4618 int (*fn)(struct net_device *upper_dev, 4619 struct netdev_nested_priv *priv), 4620 struct netdev_nested_priv *priv); 4621 4622 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4623 struct net_device *upper_dev); 4624 4625 bool netdev_has_any_upper_dev(struct net_device *dev); 4626 4627 void *netdev_lower_get_next_private(struct net_device *dev, 4628 struct list_head **iter); 4629 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4630 struct list_head **iter); 4631 4632 #define netdev_for_each_lower_private(dev, priv, iter) \ 4633 for (iter = (dev)->adj_list.lower.next, \ 4634 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4635 priv; \ 4636 priv = netdev_lower_get_next_private(dev, &(iter))) 4637 4638 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4639 for (iter = &(dev)->adj_list.lower, \ 4640 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4641 priv; \ 4642 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4643 4644 void *netdev_lower_get_next(struct net_device *dev, 4645 struct list_head **iter); 4646 4647 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4648 for (iter = (dev)->adj_list.lower.next, \ 4649 ldev = netdev_lower_get_next(dev, &(iter)); \ 4650 ldev; \ 4651 ldev = netdev_lower_get_next(dev, &(iter))) 4652 4653 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4654 struct list_head **iter); 4655 int netdev_walk_all_lower_dev(struct net_device *dev, 4656 int (*fn)(struct net_device *lower_dev, 4657 struct netdev_nested_priv *priv), 4658 struct netdev_nested_priv *priv); 4659 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4660 int (*fn)(struct net_device *lower_dev, 4661 struct netdev_nested_priv *priv), 4662 struct netdev_nested_priv *priv); 4663 4664 void *netdev_adjacent_get_private(struct list_head *adj_list); 4665 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4666 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4667 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4668 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4669 struct netlink_ext_ack *extack); 4670 int netdev_master_upper_dev_link(struct net_device *dev, 4671 struct net_device *upper_dev, 4672 void *upper_priv, void *upper_info, 4673 struct netlink_ext_ack *extack); 4674 void netdev_upper_dev_unlink(struct net_device *dev, 4675 struct net_device *upper_dev); 4676 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4677 struct net_device *new_dev, 4678 struct net_device *dev, 4679 struct netlink_ext_ack *extack); 4680 void netdev_adjacent_change_commit(struct net_device *old_dev, 4681 struct net_device *new_dev, 4682 struct net_device *dev); 4683 void netdev_adjacent_change_abort(struct net_device *old_dev, 4684 struct net_device *new_dev, 4685 struct net_device *dev); 4686 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4687 void *netdev_lower_dev_get_private(struct net_device *dev, 4688 struct net_device *lower_dev); 4689 void netdev_lower_state_changed(struct net_device *lower_dev, 4690 void *lower_state_info); 4691 4692 /* RSS keys are 40 or 52 bytes long */ 4693 #define NETDEV_RSS_KEY_LEN 52 4694 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4695 void netdev_rss_key_fill(void *buffer, size_t len); 4696 4697 int skb_checksum_help(struct sk_buff *skb); 4698 int skb_crc32c_csum_help(struct sk_buff *skb); 4699 int skb_csum_hwoffload_help(struct sk_buff *skb, 4700 const netdev_features_t features); 4701 4702 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4703 netdev_features_t features, bool tx_path); 4704 struct sk_buff *skb_eth_gso_segment(struct sk_buff *skb, 4705 netdev_features_t features, __be16 type); 4706 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4707 netdev_features_t features); 4708 4709 struct netdev_bonding_info { 4710 ifslave slave; 4711 ifbond master; 4712 }; 4713 4714 struct netdev_notifier_bonding_info { 4715 struct netdev_notifier_info info; /* must be first */ 4716 struct netdev_bonding_info bonding_info; 4717 }; 4718 4719 void netdev_bonding_info_change(struct net_device *dev, 4720 struct netdev_bonding_info *bonding_info); 4721 4722 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4723 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4724 #else 4725 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4726 const void *data) 4727 { 4728 } 4729 #endif 4730 4731 static inline 4732 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4733 { 4734 return __skb_gso_segment(skb, features, true); 4735 } 4736 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4737 4738 static inline bool can_checksum_protocol(netdev_features_t features, 4739 __be16 protocol) 4740 { 4741 if (protocol == htons(ETH_P_FCOE)) 4742 return !!(features & NETIF_F_FCOE_CRC); 4743 4744 /* Assume this is an IP checksum (not SCTP CRC) */ 4745 4746 if (features & NETIF_F_HW_CSUM) { 4747 /* Can checksum everything */ 4748 return true; 4749 } 4750 4751 switch (protocol) { 4752 case htons(ETH_P_IP): 4753 return !!(features & NETIF_F_IP_CSUM); 4754 case htons(ETH_P_IPV6): 4755 return !!(features & NETIF_F_IPV6_CSUM); 4756 default: 4757 return false; 4758 } 4759 } 4760 4761 #ifdef CONFIG_BUG 4762 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4763 #else 4764 static inline void netdev_rx_csum_fault(struct net_device *dev, 4765 struct sk_buff *skb) 4766 { 4767 } 4768 #endif 4769 /* rx skb timestamps */ 4770 void net_enable_timestamp(void); 4771 void net_disable_timestamp(void); 4772 4773 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 4774 const struct skb_shared_hwtstamps *hwtstamps, 4775 bool cycles) 4776 { 4777 const struct net_device_ops *ops = dev->netdev_ops; 4778 4779 if (ops->ndo_get_tstamp) 4780 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 4781 4782 return hwtstamps->hwtstamp; 4783 } 4784 4785 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4786 struct sk_buff *skb, struct net_device *dev, 4787 bool more) 4788 { 4789 __this_cpu_write(softnet_data.xmit.more, more); 4790 return ops->ndo_start_xmit(skb, dev); 4791 } 4792 4793 static inline bool netdev_xmit_more(void) 4794 { 4795 return __this_cpu_read(softnet_data.xmit.more); 4796 } 4797 4798 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4799 struct netdev_queue *txq, bool more) 4800 { 4801 const struct net_device_ops *ops = dev->netdev_ops; 4802 netdev_tx_t rc; 4803 4804 rc = __netdev_start_xmit(ops, skb, dev, more); 4805 if (rc == NETDEV_TX_OK) 4806 txq_trans_update(txq); 4807 4808 return rc; 4809 } 4810 4811 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4812 const void *ns); 4813 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4814 const void *ns); 4815 4816 extern const struct kobj_ns_type_operations net_ns_type_operations; 4817 4818 const char *netdev_drivername(const struct net_device *dev); 4819 4820 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4821 netdev_features_t f2) 4822 { 4823 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4824 if (f1 & NETIF_F_HW_CSUM) 4825 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4826 else 4827 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4828 } 4829 4830 return f1 & f2; 4831 } 4832 4833 static inline netdev_features_t netdev_get_wanted_features( 4834 struct net_device *dev) 4835 { 4836 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4837 } 4838 netdev_features_t netdev_increment_features(netdev_features_t all, 4839 netdev_features_t one, netdev_features_t mask); 4840 4841 /* Allow TSO being used on stacked device : 4842 * Performing the GSO segmentation before last device 4843 * is a performance improvement. 4844 */ 4845 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4846 netdev_features_t mask) 4847 { 4848 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4849 } 4850 4851 int __netdev_update_features(struct net_device *dev); 4852 void netdev_update_features(struct net_device *dev); 4853 void netdev_change_features(struct net_device *dev); 4854 4855 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4856 struct net_device *dev); 4857 4858 netdev_features_t passthru_features_check(struct sk_buff *skb, 4859 struct net_device *dev, 4860 netdev_features_t features); 4861 netdev_features_t netif_skb_features(struct sk_buff *skb); 4862 4863 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4864 { 4865 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4866 4867 /* check flags correspondence */ 4868 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4869 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4870 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4871 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4872 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4873 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4874 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4875 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4876 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4877 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4878 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4879 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4880 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4881 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4882 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4883 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4884 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4885 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4886 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4887 4888 return (features & feature) == feature; 4889 } 4890 4891 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4892 { 4893 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4894 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4895 } 4896 4897 static inline bool netif_needs_gso(struct sk_buff *skb, 4898 netdev_features_t features) 4899 { 4900 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4901 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4902 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4903 } 4904 4905 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 4906 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 4907 void netif_inherit_tso_max(struct net_device *to, 4908 const struct net_device *from); 4909 4910 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4911 int pulled_hlen, u16 mac_offset, 4912 int mac_len) 4913 { 4914 skb->protocol = protocol; 4915 skb->encapsulation = 1; 4916 skb_push(skb, pulled_hlen); 4917 skb_reset_transport_header(skb); 4918 skb->mac_header = mac_offset; 4919 skb->network_header = skb->mac_header + mac_len; 4920 skb->mac_len = mac_len; 4921 } 4922 4923 static inline bool netif_is_macsec(const struct net_device *dev) 4924 { 4925 return dev->priv_flags & IFF_MACSEC; 4926 } 4927 4928 static inline bool netif_is_macvlan(const struct net_device *dev) 4929 { 4930 return dev->priv_flags & IFF_MACVLAN; 4931 } 4932 4933 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4934 { 4935 return dev->priv_flags & IFF_MACVLAN_PORT; 4936 } 4937 4938 static inline bool netif_is_bond_master(const struct net_device *dev) 4939 { 4940 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4941 } 4942 4943 static inline bool netif_is_bond_slave(const struct net_device *dev) 4944 { 4945 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4946 } 4947 4948 static inline bool netif_supports_nofcs(struct net_device *dev) 4949 { 4950 return dev->priv_flags & IFF_SUPP_NOFCS; 4951 } 4952 4953 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4954 { 4955 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4956 } 4957 4958 static inline bool netif_is_l3_master(const struct net_device *dev) 4959 { 4960 return dev->priv_flags & IFF_L3MDEV_MASTER; 4961 } 4962 4963 static inline bool netif_is_l3_slave(const struct net_device *dev) 4964 { 4965 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4966 } 4967 4968 static inline bool netif_is_bridge_master(const struct net_device *dev) 4969 { 4970 return dev->priv_flags & IFF_EBRIDGE; 4971 } 4972 4973 static inline bool netif_is_bridge_port(const struct net_device *dev) 4974 { 4975 return dev->priv_flags & IFF_BRIDGE_PORT; 4976 } 4977 4978 static inline bool netif_is_ovs_master(const struct net_device *dev) 4979 { 4980 return dev->priv_flags & IFF_OPENVSWITCH; 4981 } 4982 4983 static inline bool netif_is_ovs_port(const struct net_device *dev) 4984 { 4985 return dev->priv_flags & IFF_OVS_DATAPATH; 4986 } 4987 4988 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 4989 { 4990 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 4991 } 4992 4993 static inline bool netif_is_team_master(const struct net_device *dev) 4994 { 4995 return dev->priv_flags & IFF_TEAM; 4996 } 4997 4998 static inline bool netif_is_team_port(const struct net_device *dev) 4999 { 5000 return dev->priv_flags & IFF_TEAM_PORT; 5001 } 5002 5003 static inline bool netif_is_lag_master(const struct net_device *dev) 5004 { 5005 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5006 } 5007 5008 static inline bool netif_is_lag_port(const struct net_device *dev) 5009 { 5010 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5011 } 5012 5013 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5014 { 5015 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5016 } 5017 5018 static inline bool netif_is_failover(const struct net_device *dev) 5019 { 5020 return dev->priv_flags & IFF_FAILOVER; 5021 } 5022 5023 static inline bool netif_is_failover_slave(const struct net_device *dev) 5024 { 5025 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5026 } 5027 5028 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5029 static inline void netif_keep_dst(struct net_device *dev) 5030 { 5031 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5032 } 5033 5034 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5035 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5036 { 5037 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5038 return netif_is_macsec(dev); 5039 } 5040 5041 extern struct pernet_operations __net_initdata loopback_net_ops; 5042 5043 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5044 5045 /* netdev_printk helpers, similar to dev_printk */ 5046 5047 static inline const char *netdev_name(const struct net_device *dev) 5048 { 5049 if (!dev->name[0] || strchr(dev->name, '%')) 5050 return "(unnamed net_device)"; 5051 return dev->name; 5052 } 5053 5054 static inline bool netdev_unregistering(const struct net_device *dev) 5055 { 5056 return dev->reg_state == NETREG_UNREGISTERING; 5057 } 5058 5059 static inline const char *netdev_reg_state(const struct net_device *dev) 5060 { 5061 switch (dev->reg_state) { 5062 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5063 case NETREG_REGISTERED: return ""; 5064 case NETREG_UNREGISTERING: return " (unregistering)"; 5065 case NETREG_UNREGISTERED: return " (unregistered)"; 5066 case NETREG_RELEASED: return " (released)"; 5067 case NETREG_DUMMY: return " (dummy)"; 5068 } 5069 5070 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5071 return " (unknown)"; 5072 } 5073 5074 __printf(3, 4) __cold 5075 void netdev_printk(const char *level, const struct net_device *dev, 5076 const char *format, ...); 5077 __printf(2, 3) __cold 5078 void netdev_emerg(const struct net_device *dev, const char *format, ...); 5079 __printf(2, 3) __cold 5080 void netdev_alert(const struct net_device *dev, const char *format, ...); 5081 __printf(2, 3) __cold 5082 void netdev_crit(const struct net_device *dev, const char *format, ...); 5083 __printf(2, 3) __cold 5084 void netdev_err(const struct net_device *dev, const char *format, ...); 5085 __printf(2, 3) __cold 5086 void netdev_warn(const struct net_device *dev, const char *format, ...); 5087 __printf(2, 3) __cold 5088 void netdev_notice(const struct net_device *dev, const char *format, ...); 5089 __printf(2, 3) __cold 5090 void netdev_info(const struct net_device *dev, const char *format, ...); 5091 5092 #define netdev_level_once(level, dev, fmt, ...) \ 5093 do { \ 5094 static bool __section(".data.once") __print_once; \ 5095 \ 5096 if (!__print_once) { \ 5097 __print_once = true; \ 5098 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 5099 } \ 5100 } while (0) 5101 5102 #define netdev_emerg_once(dev, fmt, ...) \ 5103 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 5104 #define netdev_alert_once(dev, fmt, ...) \ 5105 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 5106 #define netdev_crit_once(dev, fmt, ...) \ 5107 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 5108 #define netdev_err_once(dev, fmt, ...) \ 5109 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 5110 #define netdev_warn_once(dev, fmt, ...) \ 5111 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 5112 #define netdev_notice_once(dev, fmt, ...) \ 5113 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 5114 #define netdev_info_once(dev, fmt, ...) \ 5115 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 5116 5117 #define MODULE_ALIAS_NETDEV(device) \ 5118 MODULE_ALIAS("netdev-" device) 5119 5120 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5121 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5122 #define netdev_dbg(__dev, format, args...) \ 5123 do { \ 5124 dynamic_netdev_dbg(__dev, format, ##args); \ 5125 } while (0) 5126 #elif defined(DEBUG) 5127 #define netdev_dbg(__dev, format, args...) \ 5128 netdev_printk(KERN_DEBUG, __dev, format, ##args) 5129 #else 5130 #define netdev_dbg(__dev, format, args...) \ 5131 ({ \ 5132 if (0) \ 5133 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 5134 }) 5135 #endif 5136 5137 #if defined(VERBOSE_DEBUG) 5138 #define netdev_vdbg netdev_dbg 5139 #else 5140 5141 #define netdev_vdbg(dev, format, args...) \ 5142 ({ \ 5143 if (0) \ 5144 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 5145 0; \ 5146 }) 5147 #endif 5148 5149 /* 5150 * netdev_WARN() acts like dev_printk(), but with the key difference 5151 * of using a WARN/WARN_ON to get the message out, including the 5152 * file/line information and a backtrace. 5153 */ 5154 #define netdev_WARN(dev, format, args...) \ 5155 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5156 netdev_reg_state(dev), ##args) 5157 5158 #define netdev_WARN_ONCE(dev, format, args...) \ 5159 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5160 netdev_reg_state(dev), ##args) 5161 5162 /* netif printk helpers, similar to netdev_printk */ 5163 5164 #define netif_printk(priv, type, level, dev, fmt, args...) \ 5165 do { \ 5166 if (netif_msg_##type(priv)) \ 5167 netdev_printk(level, (dev), fmt, ##args); \ 5168 } while (0) 5169 5170 #define netif_level(level, priv, type, dev, fmt, args...) \ 5171 do { \ 5172 if (netif_msg_##type(priv)) \ 5173 netdev_##level(dev, fmt, ##args); \ 5174 } while (0) 5175 5176 #define netif_emerg(priv, type, dev, fmt, args...) \ 5177 netif_level(emerg, priv, type, dev, fmt, ##args) 5178 #define netif_alert(priv, type, dev, fmt, args...) \ 5179 netif_level(alert, priv, type, dev, fmt, ##args) 5180 #define netif_crit(priv, type, dev, fmt, args...) \ 5181 netif_level(crit, priv, type, dev, fmt, ##args) 5182 #define netif_err(priv, type, dev, fmt, args...) \ 5183 netif_level(err, priv, type, dev, fmt, ##args) 5184 #define netif_warn(priv, type, dev, fmt, args...) \ 5185 netif_level(warn, priv, type, dev, fmt, ##args) 5186 #define netif_notice(priv, type, dev, fmt, args...) \ 5187 netif_level(notice, priv, type, dev, fmt, ##args) 5188 #define netif_info(priv, type, dev, fmt, args...) \ 5189 netif_level(info, priv, type, dev, fmt, ##args) 5190 5191 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5192 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5193 #define netif_dbg(priv, type, netdev, format, args...) \ 5194 do { \ 5195 if (netif_msg_##type(priv)) \ 5196 dynamic_netdev_dbg(netdev, format, ##args); \ 5197 } while (0) 5198 #elif defined(DEBUG) 5199 #define netif_dbg(priv, type, dev, format, args...) \ 5200 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 5201 #else 5202 #define netif_dbg(priv, type, dev, format, args...) \ 5203 ({ \ 5204 if (0) \ 5205 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5206 0; \ 5207 }) 5208 #endif 5209 5210 /* if @cond then downgrade to debug, else print at @level */ 5211 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 5212 do { \ 5213 if (cond) \ 5214 netif_dbg(priv, type, netdev, fmt, ##args); \ 5215 else \ 5216 netif_ ## level(priv, type, netdev, fmt, ##args); \ 5217 } while (0) 5218 5219 #if defined(VERBOSE_DEBUG) 5220 #define netif_vdbg netif_dbg 5221 #else 5222 #define netif_vdbg(priv, type, dev, format, args...) \ 5223 ({ \ 5224 if (0) \ 5225 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5226 0; \ 5227 }) 5228 #endif 5229 5230 /* 5231 * The list of packet types we will receive (as opposed to discard) 5232 * and the routines to invoke. 5233 * 5234 * Why 16. Because with 16 the only overlap we get on a hash of the 5235 * low nibble of the protocol value is RARP/SNAP/X.25. 5236 * 5237 * 0800 IP 5238 * 0001 802.3 5239 * 0002 AX.25 5240 * 0004 802.2 5241 * 8035 RARP 5242 * 0005 SNAP 5243 * 0805 X.25 5244 * 0806 ARP 5245 * 8137 IPX 5246 * 0009 Localtalk 5247 * 86DD IPv6 5248 */ 5249 #define PTYPE_HASH_SIZE (16) 5250 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5251 5252 extern struct list_head ptype_all __read_mostly; 5253 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5254 5255 extern struct net_device *blackhole_netdev; 5256 5257 #endif /* _LINUX_NETDEVICE_H */ 5258