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 63 struct sfp_bus; 64 /* 802.11 specific */ 65 struct wireless_dev; 66 /* 802.15.4 specific */ 67 struct wpan_dev; 68 struct mpls_dev; 69 /* UDP Tunnel offloads */ 70 struct udp_tunnel_info; 71 struct udp_tunnel_nic_info; 72 struct udp_tunnel_nic; 73 struct bpf_prog; 74 struct xdp_buff; 75 76 void synchronize_net(void); 77 void netdev_set_default_ethtool_ops(struct net_device *dev, 78 const struct ethtool_ops *ops); 79 80 /* Backlog congestion levels */ 81 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 82 #define NET_RX_DROP 1 /* packet dropped */ 83 84 #define MAX_NEST_DEV 8 85 86 /* 87 * Transmit return codes: transmit return codes originate from three different 88 * namespaces: 89 * 90 * - qdisc return codes 91 * - driver transmit return codes 92 * - errno values 93 * 94 * Drivers are allowed to return any one of those in their hard_start_xmit() 95 * function. Real network devices commonly used with qdiscs should only return 96 * the driver transmit return codes though - when qdiscs are used, the actual 97 * transmission happens asynchronously, so the value is not propagated to 98 * higher layers. Virtual network devices transmit synchronously; in this case 99 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 100 * others are propagated to higher layers. 101 */ 102 103 /* qdisc ->enqueue() return codes. */ 104 #define NET_XMIT_SUCCESS 0x00 105 #define NET_XMIT_DROP 0x01 /* skb dropped */ 106 #define NET_XMIT_CN 0x02 /* congestion notification */ 107 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 108 109 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 110 * indicates that the device will soon be dropping packets, or already drops 111 * some packets of the same priority; prompting us to send less aggressively. */ 112 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 113 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 114 115 /* Driver transmit return codes */ 116 #define NETDEV_TX_MASK 0xf0 117 118 enum netdev_tx { 119 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 120 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 121 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 122 }; 123 typedef enum netdev_tx netdev_tx_t; 124 125 /* 126 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 127 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 128 */ 129 static inline bool dev_xmit_complete(int rc) 130 { 131 /* 132 * Positive cases with an skb consumed by a driver: 133 * - successful transmission (rc == NETDEV_TX_OK) 134 * - error while transmitting (rc < 0) 135 * - error while queueing to a different device (rc & NET_XMIT_MASK) 136 */ 137 if (likely(rc < NET_XMIT_MASK)) 138 return true; 139 140 return false; 141 } 142 143 /* 144 * Compute the worst-case header length according to the protocols 145 * used. 146 */ 147 148 #if defined(CONFIG_HYPERV_NET) 149 # define LL_MAX_HEADER 128 150 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 151 # if defined(CONFIG_MAC80211_MESH) 152 # define LL_MAX_HEADER 128 153 # else 154 # define LL_MAX_HEADER 96 155 # endif 156 #else 157 # define LL_MAX_HEADER 32 158 #endif 159 160 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 161 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 162 #define MAX_HEADER LL_MAX_HEADER 163 #else 164 #define MAX_HEADER (LL_MAX_HEADER + 48) 165 #endif 166 167 /* 168 * Old network device statistics. Fields are native words 169 * (unsigned long) so they can be read and written atomically. 170 */ 171 172 struct net_device_stats { 173 unsigned long rx_packets; 174 unsigned long tx_packets; 175 unsigned long rx_bytes; 176 unsigned long tx_bytes; 177 unsigned long rx_errors; 178 unsigned long tx_errors; 179 unsigned long rx_dropped; 180 unsigned long tx_dropped; 181 unsigned long multicast; 182 unsigned long collisions; 183 unsigned long rx_length_errors; 184 unsigned long rx_over_errors; 185 unsigned long rx_crc_errors; 186 unsigned long rx_frame_errors; 187 unsigned long rx_fifo_errors; 188 unsigned long rx_missed_errors; 189 unsigned long tx_aborted_errors; 190 unsigned long tx_carrier_errors; 191 unsigned long tx_fifo_errors; 192 unsigned long tx_heartbeat_errors; 193 unsigned long tx_window_errors; 194 unsigned long rx_compressed; 195 unsigned long tx_compressed; 196 }; 197 198 /* per-cpu stats, allocated on demand. 199 * Try to fit them in a single cache line, for dev_get_stats() sake. 200 */ 201 struct net_device_core_stats { 202 local_t rx_dropped; 203 local_t tx_dropped; 204 local_t rx_nohandler; 205 } __aligned(4 * sizeof(local_t)); 206 207 #include <linux/cache.h> 208 #include <linux/skbuff.h> 209 210 #ifdef CONFIG_RPS 211 #include <linux/static_key.h> 212 extern struct static_key_false rps_needed; 213 extern struct static_key_false rfs_needed; 214 #endif 215 216 struct neighbour; 217 struct neigh_parms; 218 struct sk_buff; 219 220 struct netdev_hw_addr { 221 struct list_head list; 222 struct rb_node node; 223 unsigned char addr[MAX_ADDR_LEN]; 224 unsigned char type; 225 #define NETDEV_HW_ADDR_T_LAN 1 226 #define NETDEV_HW_ADDR_T_SAN 2 227 #define NETDEV_HW_ADDR_T_UNICAST 3 228 #define NETDEV_HW_ADDR_T_MULTICAST 4 229 bool global_use; 230 int sync_cnt; 231 int refcount; 232 int synced; 233 struct rcu_head rcu_head; 234 }; 235 236 struct netdev_hw_addr_list { 237 struct list_head list; 238 int count; 239 240 /* Auxiliary tree for faster lookup on addition and deletion */ 241 struct rb_root tree; 242 }; 243 244 #define netdev_hw_addr_list_count(l) ((l)->count) 245 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 246 #define netdev_hw_addr_list_for_each(ha, l) \ 247 list_for_each_entry(ha, &(l)->list, list) 248 249 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 250 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 251 #define netdev_for_each_uc_addr(ha, dev) \ 252 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 253 254 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 255 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 256 #define netdev_for_each_mc_addr(ha, dev) \ 257 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 258 259 struct hh_cache { 260 unsigned int hh_len; 261 seqlock_t hh_lock; 262 263 /* cached hardware header; allow for machine alignment needs. */ 264 #define HH_DATA_MOD 16 265 #define HH_DATA_OFF(__len) \ 266 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 267 #define HH_DATA_ALIGN(__len) \ 268 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 269 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 270 }; 271 272 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 273 * Alternative is: 274 * dev->hard_header_len ? (dev->hard_header_len + 275 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 276 * 277 * We could use other alignment values, but we must maintain the 278 * relationship HH alignment <= LL alignment. 279 */ 280 #define LL_RESERVED_SPACE(dev) \ 281 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 282 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 283 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 284 285 struct header_ops { 286 int (*create) (struct sk_buff *skb, struct net_device *dev, 287 unsigned short type, const void *daddr, 288 const void *saddr, unsigned int len); 289 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 290 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 291 void (*cache_update)(struct hh_cache *hh, 292 const struct net_device *dev, 293 const unsigned char *haddr); 294 bool (*validate)(const char *ll_header, unsigned int len); 295 __be16 (*parse_protocol)(const struct sk_buff *skb); 296 }; 297 298 /* These flag bits are private to the generic network queueing 299 * layer; they may not be explicitly referenced by any other 300 * code. 301 */ 302 303 enum netdev_state_t { 304 __LINK_STATE_START, 305 __LINK_STATE_PRESENT, 306 __LINK_STATE_NOCARRIER, 307 __LINK_STATE_LINKWATCH_PENDING, 308 __LINK_STATE_DORMANT, 309 __LINK_STATE_TESTING, 310 }; 311 312 struct gro_list { 313 struct list_head list; 314 int count; 315 }; 316 317 /* 318 * size of gro hash buckets, must less than bit number of 319 * napi_struct::gro_bitmask 320 */ 321 #define GRO_HASH_BUCKETS 8 322 323 /* 324 * Structure for NAPI scheduling similar to tasklet but with weighting 325 */ 326 struct napi_struct { 327 /* The poll_list must only be managed by the entity which 328 * changes the state of the NAPI_STATE_SCHED bit. This means 329 * whoever atomically sets that bit can add this napi_struct 330 * to the per-CPU poll_list, and whoever clears that bit 331 * can remove from the list right before clearing the bit. 332 */ 333 struct list_head poll_list; 334 335 unsigned long state; 336 int weight; 337 int defer_hard_irqs_count; 338 unsigned long gro_bitmask; 339 int (*poll)(struct napi_struct *, int); 340 #ifdef CONFIG_NETPOLL 341 int poll_owner; 342 #endif 343 struct net_device *dev; 344 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 345 struct sk_buff *skb; 346 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 347 int rx_count; /* length of rx_list */ 348 struct hrtimer timer; 349 struct list_head dev_list; 350 struct hlist_node napi_hash_node; 351 unsigned int napi_id; 352 struct task_struct *thread; 353 }; 354 355 enum { 356 NAPI_STATE_SCHED, /* Poll is scheduled */ 357 NAPI_STATE_MISSED, /* reschedule a napi */ 358 NAPI_STATE_DISABLE, /* Disable pending */ 359 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 360 NAPI_STATE_LISTED, /* NAPI added to system lists */ 361 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 362 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 363 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 364 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 365 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 366 }; 367 368 enum { 369 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 370 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 371 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 372 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 373 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 374 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 375 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 376 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 377 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 378 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 379 }; 380 381 enum gro_result { 382 GRO_MERGED, 383 GRO_MERGED_FREE, 384 GRO_HELD, 385 GRO_NORMAL, 386 GRO_CONSUMED, 387 }; 388 typedef enum gro_result gro_result_t; 389 390 /* 391 * enum rx_handler_result - Possible return values for rx_handlers. 392 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 393 * further. 394 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 395 * case skb->dev was changed by rx_handler. 396 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 397 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 398 * 399 * rx_handlers are functions called from inside __netif_receive_skb(), to do 400 * special processing of the skb, prior to delivery to protocol handlers. 401 * 402 * Currently, a net_device can only have a single rx_handler registered. Trying 403 * to register a second rx_handler will return -EBUSY. 404 * 405 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 406 * To unregister a rx_handler on a net_device, use 407 * netdev_rx_handler_unregister(). 408 * 409 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 410 * do with the skb. 411 * 412 * If the rx_handler consumed the skb in some way, it should return 413 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 414 * the skb to be delivered in some other way. 415 * 416 * If the rx_handler changed skb->dev, to divert the skb to another 417 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 418 * new device will be called if it exists. 419 * 420 * If the rx_handler decides the skb should be ignored, it should return 421 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 422 * are registered on exact device (ptype->dev == skb->dev). 423 * 424 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 425 * delivered, it should return RX_HANDLER_PASS. 426 * 427 * A device without a registered rx_handler will behave as if rx_handler 428 * returned RX_HANDLER_PASS. 429 */ 430 431 enum rx_handler_result { 432 RX_HANDLER_CONSUMED, 433 RX_HANDLER_ANOTHER, 434 RX_HANDLER_EXACT, 435 RX_HANDLER_PASS, 436 }; 437 typedef enum rx_handler_result rx_handler_result_t; 438 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 439 440 void __napi_schedule(struct napi_struct *n); 441 void __napi_schedule_irqoff(struct napi_struct *n); 442 443 static inline bool napi_disable_pending(struct napi_struct *n) 444 { 445 return test_bit(NAPI_STATE_DISABLE, &n->state); 446 } 447 448 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 449 { 450 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 451 } 452 453 bool napi_schedule_prep(struct napi_struct *n); 454 455 /** 456 * napi_schedule - schedule NAPI poll 457 * @n: NAPI context 458 * 459 * Schedule NAPI poll routine to be called if it is not already 460 * running. 461 */ 462 static inline void napi_schedule(struct napi_struct *n) 463 { 464 if (napi_schedule_prep(n)) 465 __napi_schedule(n); 466 } 467 468 /** 469 * napi_schedule_irqoff - schedule NAPI poll 470 * @n: NAPI context 471 * 472 * Variant of napi_schedule(), assuming hard irqs are masked. 473 */ 474 static inline void napi_schedule_irqoff(struct napi_struct *n) 475 { 476 if (napi_schedule_prep(n)) 477 __napi_schedule_irqoff(n); 478 } 479 480 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 481 static inline bool napi_reschedule(struct napi_struct *napi) 482 { 483 if (napi_schedule_prep(napi)) { 484 __napi_schedule(napi); 485 return true; 486 } 487 return false; 488 } 489 490 bool napi_complete_done(struct napi_struct *n, int work_done); 491 /** 492 * napi_complete - NAPI processing complete 493 * @n: NAPI context 494 * 495 * Mark NAPI processing as complete. 496 * Consider using napi_complete_done() instead. 497 * Return false if device should avoid rearming interrupts. 498 */ 499 static inline bool napi_complete(struct napi_struct *n) 500 { 501 return napi_complete_done(n, 0); 502 } 503 504 int dev_set_threaded(struct net_device *dev, bool threaded); 505 506 /** 507 * napi_disable - prevent NAPI from scheduling 508 * @n: NAPI context 509 * 510 * Stop NAPI from being scheduled on this context. 511 * Waits till any outstanding processing completes. 512 */ 513 void napi_disable(struct napi_struct *n); 514 515 void napi_enable(struct napi_struct *n); 516 517 /** 518 * napi_synchronize - wait until NAPI is not running 519 * @n: NAPI context 520 * 521 * Wait until NAPI is done being scheduled on this context. 522 * Waits till any outstanding processing completes but 523 * does not disable future activations. 524 */ 525 static inline void napi_synchronize(const struct napi_struct *n) 526 { 527 if (IS_ENABLED(CONFIG_SMP)) 528 while (test_bit(NAPI_STATE_SCHED, &n->state)) 529 msleep(1); 530 else 531 barrier(); 532 } 533 534 /** 535 * napi_if_scheduled_mark_missed - if napi is running, set the 536 * NAPIF_STATE_MISSED 537 * @n: NAPI context 538 * 539 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 540 * NAPI is scheduled. 541 **/ 542 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 543 { 544 unsigned long val, new; 545 546 do { 547 val = READ_ONCE(n->state); 548 if (val & NAPIF_STATE_DISABLE) 549 return true; 550 551 if (!(val & NAPIF_STATE_SCHED)) 552 return false; 553 554 new = val | NAPIF_STATE_MISSED; 555 } while (cmpxchg(&n->state, val, new) != val); 556 557 return true; 558 } 559 560 enum netdev_queue_state_t { 561 __QUEUE_STATE_DRV_XOFF, 562 __QUEUE_STATE_STACK_XOFF, 563 __QUEUE_STATE_FROZEN, 564 }; 565 566 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 567 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 568 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 569 570 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 571 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 572 QUEUE_STATE_FROZEN) 573 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 574 QUEUE_STATE_FROZEN) 575 576 /* 577 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 578 * netif_tx_* functions below are used to manipulate this flag. The 579 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 580 * queue independently. The netif_xmit_*stopped functions below are called 581 * to check if the queue has been stopped by the driver or stack (either 582 * of the XOFF bits are set in the state). Drivers should not need to call 583 * netif_xmit*stopped functions, they should only be using netif_tx_*. 584 */ 585 586 struct netdev_queue { 587 /* 588 * read-mostly part 589 */ 590 struct net_device *dev; 591 netdevice_tracker dev_tracker; 592 593 struct Qdisc __rcu *qdisc; 594 struct Qdisc *qdisc_sleeping; 595 #ifdef CONFIG_SYSFS 596 struct kobject kobj; 597 #endif 598 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 599 int numa_node; 600 #endif 601 unsigned long tx_maxrate; 602 /* 603 * Number of TX timeouts for this queue 604 * (/sys/class/net/DEV/Q/trans_timeout) 605 */ 606 atomic_long_t trans_timeout; 607 608 /* Subordinate device that the queue has been assigned to */ 609 struct net_device *sb_dev; 610 #ifdef CONFIG_XDP_SOCKETS 611 struct xsk_buff_pool *pool; 612 #endif 613 /* 614 * write-mostly part 615 */ 616 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 617 int xmit_lock_owner; 618 /* 619 * Time (in jiffies) of last Tx 620 */ 621 unsigned long trans_start; 622 623 unsigned long state; 624 625 #ifdef CONFIG_BQL 626 struct dql dql; 627 #endif 628 } ____cacheline_aligned_in_smp; 629 630 extern int sysctl_fb_tunnels_only_for_init_net; 631 extern int sysctl_devconf_inherit_init_net; 632 633 /* 634 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 635 * == 1 : For initns only 636 * == 2 : For none. 637 */ 638 static inline bool net_has_fallback_tunnels(const struct net *net) 639 { 640 return !IS_ENABLED(CONFIG_SYSCTL) || 641 !sysctl_fb_tunnels_only_for_init_net || 642 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1); 643 } 644 645 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 646 { 647 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 648 return q->numa_node; 649 #else 650 return NUMA_NO_NODE; 651 #endif 652 } 653 654 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 655 { 656 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 657 q->numa_node = node; 658 #endif 659 } 660 661 #ifdef CONFIG_RPS 662 /* 663 * This structure holds an RPS map which can be of variable length. The 664 * map is an array of CPUs. 665 */ 666 struct rps_map { 667 unsigned int len; 668 struct rcu_head rcu; 669 u16 cpus[]; 670 }; 671 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 672 673 /* 674 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 675 * tail pointer for that CPU's input queue at the time of last enqueue, and 676 * a hardware filter index. 677 */ 678 struct rps_dev_flow { 679 u16 cpu; 680 u16 filter; 681 unsigned int last_qtail; 682 }; 683 #define RPS_NO_FILTER 0xffff 684 685 /* 686 * The rps_dev_flow_table structure contains a table of flow mappings. 687 */ 688 struct rps_dev_flow_table { 689 unsigned int mask; 690 struct rcu_head rcu; 691 struct rps_dev_flow flows[]; 692 }; 693 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 694 ((_num) * sizeof(struct rps_dev_flow))) 695 696 /* 697 * The rps_sock_flow_table contains mappings of flows to the last CPU 698 * on which they were processed by the application (set in recvmsg). 699 * Each entry is a 32bit value. Upper part is the high-order bits 700 * of flow hash, lower part is CPU number. 701 * rps_cpu_mask is used to partition the space, depending on number of 702 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 703 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 704 * meaning we use 32-6=26 bits for the hash. 705 */ 706 struct rps_sock_flow_table { 707 u32 mask; 708 709 u32 ents[] ____cacheline_aligned_in_smp; 710 }; 711 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 712 713 #define RPS_NO_CPU 0xffff 714 715 extern u32 rps_cpu_mask; 716 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 717 718 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 719 u32 hash) 720 { 721 if (table && hash) { 722 unsigned int index = hash & table->mask; 723 u32 val = hash & ~rps_cpu_mask; 724 725 /* We only give a hint, preemption can change CPU under us */ 726 val |= raw_smp_processor_id(); 727 728 if (table->ents[index] != val) 729 table->ents[index] = val; 730 } 731 } 732 733 #ifdef CONFIG_RFS_ACCEL 734 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 735 u16 filter_id); 736 #endif 737 #endif /* CONFIG_RPS */ 738 739 /* This structure contains an instance of an RX queue. */ 740 struct netdev_rx_queue { 741 struct xdp_rxq_info xdp_rxq; 742 #ifdef CONFIG_RPS 743 struct rps_map __rcu *rps_map; 744 struct rps_dev_flow_table __rcu *rps_flow_table; 745 #endif 746 struct kobject kobj; 747 struct net_device *dev; 748 netdevice_tracker dev_tracker; 749 750 #ifdef CONFIG_XDP_SOCKETS 751 struct xsk_buff_pool *pool; 752 #endif 753 } ____cacheline_aligned_in_smp; 754 755 /* 756 * RX queue sysfs structures and functions. 757 */ 758 struct rx_queue_attribute { 759 struct attribute attr; 760 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 761 ssize_t (*store)(struct netdev_rx_queue *queue, 762 const char *buf, size_t len); 763 }; 764 765 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 766 enum xps_map_type { 767 XPS_CPUS = 0, 768 XPS_RXQS, 769 XPS_MAPS_MAX, 770 }; 771 772 #ifdef CONFIG_XPS 773 /* 774 * This structure holds an XPS map which can be of variable length. The 775 * map is an array of queues. 776 */ 777 struct xps_map { 778 unsigned int len; 779 unsigned int alloc_len; 780 struct rcu_head rcu; 781 u16 queues[]; 782 }; 783 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 784 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 785 - sizeof(struct xps_map)) / sizeof(u16)) 786 787 /* 788 * This structure holds all XPS maps for device. Maps are indexed by CPU. 789 * 790 * We keep track of the number of cpus/rxqs used when the struct is allocated, 791 * in nr_ids. This will help not accessing out-of-bound memory. 792 * 793 * We keep track of the number of traffic classes used when the struct is 794 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 795 * not crossing its upper bound, as the original dev->num_tc can be updated in 796 * the meantime. 797 */ 798 struct xps_dev_maps { 799 struct rcu_head rcu; 800 unsigned int nr_ids; 801 s16 num_tc; 802 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 803 }; 804 805 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 806 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 807 808 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 809 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 810 811 #endif /* CONFIG_XPS */ 812 813 #define TC_MAX_QUEUE 16 814 #define TC_BITMASK 15 815 /* HW offloaded queuing disciplines txq count and offset maps */ 816 struct netdev_tc_txq { 817 u16 count; 818 u16 offset; 819 }; 820 821 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 822 /* 823 * This structure is to hold information about the device 824 * configured to run FCoE protocol stack. 825 */ 826 struct netdev_fcoe_hbainfo { 827 char manufacturer[64]; 828 char serial_number[64]; 829 char hardware_version[64]; 830 char driver_version[64]; 831 char optionrom_version[64]; 832 char firmware_version[64]; 833 char model[256]; 834 char model_description[256]; 835 }; 836 #endif 837 838 #define MAX_PHYS_ITEM_ID_LEN 32 839 840 /* This structure holds a unique identifier to identify some 841 * physical item (port for example) used by a netdevice. 842 */ 843 struct netdev_phys_item_id { 844 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 845 unsigned char id_len; 846 }; 847 848 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 849 struct netdev_phys_item_id *b) 850 { 851 return a->id_len == b->id_len && 852 memcmp(a->id, b->id, a->id_len) == 0; 853 } 854 855 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 856 struct sk_buff *skb, 857 struct net_device *sb_dev); 858 859 enum net_device_path_type { 860 DEV_PATH_ETHERNET = 0, 861 DEV_PATH_VLAN, 862 DEV_PATH_BRIDGE, 863 DEV_PATH_PPPOE, 864 DEV_PATH_DSA, 865 }; 866 867 struct net_device_path { 868 enum net_device_path_type type; 869 const struct net_device *dev; 870 union { 871 struct { 872 u16 id; 873 __be16 proto; 874 u8 h_dest[ETH_ALEN]; 875 } encap; 876 struct { 877 enum { 878 DEV_PATH_BR_VLAN_KEEP, 879 DEV_PATH_BR_VLAN_TAG, 880 DEV_PATH_BR_VLAN_UNTAG, 881 DEV_PATH_BR_VLAN_UNTAG_HW, 882 } vlan_mode; 883 u16 vlan_id; 884 __be16 vlan_proto; 885 } bridge; 886 struct { 887 int port; 888 u16 proto; 889 } dsa; 890 }; 891 }; 892 893 #define NET_DEVICE_PATH_STACK_MAX 5 894 #define NET_DEVICE_PATH_VLAN_MAX 2 895 896 struct net_device_path_stack { 897 int num_paths; 898 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 899 }; 900 901 struct net_device_path_ctx { 902 const struct net_device *dev; 903 const u8 *daddr; 904 905 int num_vlans; 906 struct { 907 u16 id; 908 __be16 proto; 909 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 910 }; 911 912 enum tc_setup_type { 913 TC_SETUP_QDISC_MQPRIO, 914 TC_SETUP_CLSU32, 915 TC_SETUP_CLSFLOWER, 916 TC_SETUP_CLSMATCHALL, 917 TC_SETUP_CLSBPF, 918 TC_SETUP_BLOCK, 919 TC_SETUP_QDISC_CBS, 920 TC_SETUP_QDISC_RED, 921 TC_SETUP_QDISC_PRIO, 922 TC_SETUP_QDISC_MQ, 923 TC_SETUP_QDISC_ETF, 924 TC_SETUP_ROOT_QDISC, 925 TC_SETUP_QDISC_GRED, 926 TC_SETUP_QDISC_TAPRIO, 927 TC_SETUP_FT, 928 TC_SETUP_QDISC_ETS, 929 TC_SETUP_QDISC_TBF, 930 TC_SETUP_QDISC_FIFO, 931 TC_SETUP_QDISC_HTB, 932 TC_SETUP_ACT, 933 }; 934 935 /* These structures hold the attributes of bpf state that are being passed 936 * to the netdevice through the bpf op. 937 */ 938 enum bpf_netdev_command { 939 /* Set or clear a bpf program used in the earliest stages of packet 940 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 941 * is responsible for calling bpf_prog_put on any old progs that are 942 * stored. In case of error, the callee need not release the new prog 943 * reference, but on success it takes ownership and must bpf_prog_put 944 * when it is no longer used. 945 */ 946 XDP_SETUP_PROG, 947 XDP_SETUP_PROG_HW, 948 /* BPF program for offload callbacks, invoked at program load time. */ 949 BPF_OFFLOAD_MAP_ALLOC, 950 BPF_OFFLOAD_MAP_FREE, 951 XDP_SETUP_XSK_POOL, 952 }; 953 954 struct bpf_prog_offload_ops; 955 struct netlink_ext_ack; 956 struct xdp_umem; 957 struct xdp_dev_bulk_queue; 958 struct bpf_xdp_link; 959 960 enum bpf_xdp_mode { 961 XDP_MODE_SKB = 0, 962 XDP_MODE_DRV = 1, 963 XDP_MODE_HW = 2, 964 __MAX_XDP_MODE 965 }; 966 967 struct bpf_xdp_entity { 968 struct bpf_prog *prog; 969 struct bpf_xdp_link *link; 970 }; 971 972 struct netdev_bpf { 973 enum bpf_netdev_command command; 974 union { 975 /* XDP_SETUP_PROG */ 976 struct { 977 u32 flags; 978 struct bpf_prog *prog; 979 struct netlink_ext_ack *extack; 980 }; 981 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 982 struct { 983 struct bpf_offloaded_map *offmap; 984 }; 985 /* XDP_SETUP_XSK_POOL */ 986 struct { 987 struct xsk_buff_pool *pool; 988 u16 queue_id; 989 } xsk; 990 }; 991 }; 992 993 /* Flags for ndo_xsk_wakeup. */ 994 #define XDP_WAKEUP_RX (1 << 0) 995 #define XDP_WAKEUP_TX (1 << 1) 996 997 #ifdef CONFIG_XFRM_OFFLOAD 998 struct xfrmdev_ops { 999 int (*xdo_dev_state_add) (struct xfrm_state *x); 1000 void (*xdo_dev_state_delete) (struct xfrm_state *x); 1001 void (*xdo_dev_state_free) (struct xfrm_state *x); 1002 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 1003 struct xfrm_state *x); 1004 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1005 }; 1006 #endif 1007 1008 struct dev_ifalias { 1009 struct rcu_head rcuhead; 1010 char ifalias[]; 1011 }; 1012 1013 struct devlink; 1014 struct tlsdev_ops; 1015 1016 struct netdev_name_node { 1017 struct hlist_node hlist; 1018 struct list_head list; 1019 struct net_device *dev; 1020 const char *name; 1021 }; 1022 1023 int netdev_name_node_alt_create(struct net_device *dev, const char *name); 1024 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name); 1025 1026 struct netdev_net_notifier { 1027 struct list_head list; 1028 struct notifier_block *nb; 1029 }; 1030 1031 /* 1032 * This structure defines the management hooks for network devices. 1033 * The following hooks can be defined; unless noted otherwise, they are 1034 * optional and can be filled with a null pointer. 1035 * 1036 * int (*ndo_init)(struct net_device *dev); 1037 * This function is called once when a network device is registered. 1038 * The network device can use this for any late stage initialization 1039 * or semantic validation. It can fail with an error code which will 1040 * be propagated back to register_netdev. 1041 * 1042 * void (*ndo_uninit)(struct net_device *dev); 1043 * This function is called when device is unregistered or when registration 1044 * fails. It is not called if init fails. 1045 * 1046 * int (*ndo_open)(struct net_device *dev); 1047 * This function is called when a network device transitions to the up 1048 * state. 1049 * 1050 * int (*ndo_stop)(struct net_device *dev); 1051 * This function is called when a network device transitions to the down 1052 * state. 1053 * 1054 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1055 * struct net_device *dev); 1056 * Called when a packet needs to be transmitted. 1057 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1058 * the queue before that can happen; it's for obsolete devices and weird 1059 * corner cases, but the stack really does a non-trivial amount 1060 * of useless work if you return NETDEV_TX_BUSY. 1061 * Required; cannot be NULL. 1062 * 1063 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1064 * struct net_device *dev 1065 * netdev_features_t features); 1066 * Called by core transmit path to determine if device is capable of 1067 * performing offload operations on a given packet. This is to give 1068 * the device an opportunity to implement any restrictions that cannot 1069 * be otherwise expressed by feature flags. The check is called with 1070 * the set of features that the stack has calculated and it returns 1071 * those the driver believes to be appropriate. 1072 * 1073 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1074 * struct net_device *sb_dev); 1075 * Called to decide which queue to use when device supports multiple 1076 * transmit queues. 1077 * 1078 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1079 * This function is called to allow device receiver to make 1080 * changes to configuration when multicast or promiscuous is enabled. 1081 * 1082 * void (*ndo_set_rx_mode)(struct net_device *dev); 1083 * This function is called device changes address list filtering. 1084 * If driver handles unicast address filtering, it should set 1085 * IFF_UNICAST_FLT in its priv_flags. 1086 * 1087 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1088 * This function is called when the Media Access Control address 1089 * needs to be changed. If this interface is not defined, the 1090 * MAC address can not be changed. 1091 * 1092 * int (*ndo_validate_addr)(struct net_device *dev); 1093 * Test if Media Access Control address is valid for the device. 1094 * 1095 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1096 * Old-style ioctl entry point. This is used internally by the 1097 * appletalk and ieee802154 subsystems but is no longer called by 1098 * the device ioctl handler. 1099 * 1100 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1101 * Used by the bonding driver for its device specific ioctls: 1102 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1103 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1104 * 1105 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1106 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1107 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1108 * 1109 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1110 * Used to set network devices bus interface parameters. This interface 1111 * is retained for legacy reasons; new devices should use the bus 1112 * interface (PCI) for low level management. 1113 * 1114 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1115 * Called when a user wants to change the Maximum Transfer Unit 1116 * of a device. 1117 * 1118 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1119 * Callback used when the transmitter has not made any progress 1120 * for dev->watchdog ticks. 1121 * 1122 * void (*ndo_get_stats64)(struct net_device *dev, 1123 * struct rtnl_link_stats64 *storage); 1124 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1125 * Called when a user wants to get the network device usage 1126 * statistics. Drivers must do one of the following: 1127 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1128 * rtnl_link_stats64 structure passed by the caller. 1129 * 2. Define @ndo_get_stats to update a net_device_stats structure 1130 * (which should normally be dev->stats) and return a pointer to 1131 * it. The structure may be changed asynchronously only if each 1132 * field is written atomically. 1133 * 3. Update dev->stats asynchronously and atomically, and define 1134 * neither operation. 1135 * 1136 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1137 * Return true if this device supports offload stats of this attr_id. 1138 * 1139 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1140 * void *attr_data) 1141 * Get statistics for offload operations by attr_id. Write it into the 1142 * attr_data pointer. 1143 * 1144 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1145 * If device supports VLAN filtering this function is called when a 1146 * VLAN id is registered. 1147 * 1148 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1149 * If device supports VLAN filtering this function is called when a 1150 * VLAN id is unregistered. 1151 * 1152 * void (*ndo_poll_controller)(struct net_device *dev); 1153 * 1154 * SR-IOV management functions. 1155 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1156 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1157 * u8 qos, __be16 proto); 1158 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1159 * int max_tx_rate); 1160 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1161 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1162 * int (*ndo_get_vf_config)(struct net_device *dev, 1163 * int vf, struct ifla_vf_info *ivf); 1164 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1165 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1166 * struct nlattr *port[]); 1167 * 1168 * Enable or disable the VF ability to query its RSS Redirection Table and 1169 * Hash Key. This is needed since on some devices VF share this information 1170 * with PF and querying it may introduce a theoretical security risk. 1171 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1172 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1173 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1174 * void *type_data); 1175 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1176 * This is always called from the stack with the rtnl lock held and netif 1177 * tx queues stopped. This allows the netdevice to perform queue 1178 * management safely. 1179 * 1180 * Fiber Channel over Ethernet (FCoE) offload functions. 1181 * int (*ndo_fcoe_enable)(struct net_device *dev); 1182 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1183 * so the underlying device can perform whatever needed configuration or 1184 * initialization to support acceleration of FCoE traffic. 1185 * 1186 * int (*ndo_fcoe_disable)(struct net_device *dev); 1187 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1188 * so the underlying device can perform whatever needed clean-ups to 1189 * stop supporting acceleration of FCoE traffic. 1190 * 1191 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1192 * struct scatterlist *sgl, unsigned int sgc); 1193 * Called when the FCoE Initiator wants to initialize an I/O that 1194 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1195 * perform necessary setup and returns 1 to indicate the device is set up 1196 * successfully to perform DDP on this I/O, otherwise this returns 0. 1197 * 1198 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1199 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1200 * indicated by the FC exchange id 'xid', so the underlying device can 1201 * clean up and reuse resources for later DDP requests. 1202 * 1203 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1204 * struct scatterlist *sgl, unsigned int sgc); 1205 * Called when the FCoE Target wants to initialize an I/O that 1206 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1207 * perform necessary setup and returns 1 to indicate the device is set up 1208 * successfully to perform DDP on this I/O, otherwise this returns 0. 1209 * 1210 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1211 * struct netdev_fcoe_hbainfo *hbainfo); 1212 * Called when the FCoE Protocol stack wants information on the underlying 1213 * device. This information is utilized by the FCoE protocol stack to 1214 * register attributes with Fiber Channel management service as per the 1215 * FC-GS Fabric Device Management Information(FDMI) specification. 1216 * 1217 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1218 * Called when the underlying device wants to override default World Wide 1219 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1220 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1221 * protocol stack to use. 1222 * 1223 * RFS acceleration. 1224 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1225 * u16 rxq_index, u32 flow_id); 1226 * Set hardware filter for RFS. rxq_index is the target queue index; 1227 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1228 * Return the filter ID on success, or a negative error code. 1229 * 1230 * Slave management functions (for bridge, bonding, etc). 1231 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1232 * Called to make another netdev an underling. 1233 * 1234 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1235 * Called to release previously enslaved netdev. 1236 * 1237 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1238 * struct sk_buff *skb, 1239 * bool all_slaves); 1240 * Get the xmit slave of master device. If all_slaves is true, function 1241 * assume all the slaves can transmit. 1242 * 1243 * Feature/offload setting functions. 1244 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1245 * netdev_features_t features); 1246 * Adjusts the requested feature flags according to device-specific 1247 * constraints, and returns the resulting flags. Must not modify 1248 * the device state. 1249 * 1250 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1251 * Called to update device configuration to new features. Passed 1252 * feature set might be less than what was returned by ndo_fix_features()). 1253 * Must return >0 or -errno if it changed dev->features itself. 1254 * 1255 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1256 * struct net_device *dev, 1257 * const unsigned char *addr, u16 vid, u16 flags, 1258 * struct netlink_ext_ack *extack); 1259 * Adds an FDB entry to dev for addr. 1260 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1261 * struct net_device *dev, 1262 * const unsigned char *addr, u16 vid) 1263 * Deletes the FDB entry from dev coresponding to addr. 1264 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1265 * struct net_device *dev, struct net_device *filter_dev, 1266 * int *idx) 1267 * Used to add FDB entries to dump requests. Implementers should add 1268 * entries to skb and update idx with the number of entries. 1269 * 1270 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1271 * u16 flags, struct netlink_ext_ack *extack) 1272 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1273 * struct net_device *dev, u32 filter_mask, 1274 * int nlflags) 1275 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1276 * u16 flags); 1277 * 1278 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1279 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1280 * which do not represent real hardware may define this to allow their 1281 * userspace components to manage their virtual carrier state. Devices 1282 * that determine carrier state from physical hardware properties (eg 1283 * network cables) or protocol-dependent mechanisms (eg 1284 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1285 * 1286 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1287 * struct netdev_phys_item_id *ppid); 1288 * Called to get ID of physical port of this device. If driver does 1289 * not implement this, it is assumed that the hw is not able to have 1290 * multiple net devices on single physical port. 1291 * 1292 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1293 * struct netdev_phys_item_id *ppid) 1294 * Called to get the parent ID of the physical port of this device. 1295 * 1296 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1297 * struct net_device *dev) 1298 * Called by upper layer devices to accelerate switching or other 1299 * station functionality into hardware. 'pdev is the lowerdev 1300 * to use for the offload and 'dev' is the net device that will 1301 * back the offload. Returns a pointer to the private structure 1302 * the upper layer will maintain. 1303 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1304 * Called by upper layer device to delete the station created 1305 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1306 * the station and priv is the structure returned by the add 1307 * operation. 1308 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1309 * int queue_index, u32 maxrate); 1310 * Called when a user wants to set a max-rate limitation of specific 1311 * TX queue. 1312 * int (*ndo_get_iflink)(const struct net_device *dev); 1313 * Called to get the iflink value of this device. 1314 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1315 * This function is used to get egress tunnel information for given skb. 1316 * This is useful for retrieving outer tunnel header parameters while 1317 * sampling packet. 1318 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1319 * This function is used to specify the headroom that the skb must 1320 * consider when allocation skb during packet reception. Setting 1321 * appropriate rx headroom value allows avoiding skb head copy on 1322 * forward. Setting a negative value resets the rx headroom to the 1323 * default value. 1324 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1325 * This function is used to set or query state related to XDP on the 1326 * netdevice and manage BPF offload. See definition of 1327 * enum bpf_netdev_command for details. 1328 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1329 * u32 flags); 1330 * This function is used to submit @n XDP packets for transmit on a 1331 * netdevice. Returns number of frames successfully transmitted, frames 1332 * that got dropped are freed/returned via xdp_return_frame(). 1333 * Returns negative number, means general error invoking ndo, meaning 1334 * no frames were xmit'ed and core-caller will free all frames. 1335 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1336 * struct xdp_buff *xdp); 1337 * Get the xmit slave of master device based on the xdp_buff. 1338 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1339 * This function is used to wake up the softirq, ksoftirqd or kthread 1340 * responsible for sending and/or receiving packets on a specific 1341 * queue id bound to an AF_XDP socket. The flags field specifies if 1342 * only RX, only Tx, or both should be woken up using the flags 1343 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1344 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1345 * Get devlink port instance associated with a given netdev. 1346 * Called with a reference on the netdevice and devlink locks only, 1347 * rtnl_lock is not held. 1348 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1349 * int cmd); 1350 * Add, change, delete or get information on an IPv4 tunnel. 1351 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1352 * If a device is paired with a peer device, return the peer instance. 1353 * The caller must be under RCU read context. 1354 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1355 * Get the forwarding path to reach the real device from the HW destination address 1356 */ 1357 struct net_device_ops { 1358 int (*ndo_init)(struct net_device *dev); 1359 void (*ndo_uninit)(struct net_device *dev); 1360 int (*ndo_open)(struct net_device *dev); 1361 int (*ndo_stop)(struct net_device *dev); 1362 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1363 struct net_device *dev); 1364 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1365 struct net_device *dev, 1366 netdev_features_t features); 1367 u16 (*ndo_select_queue)(struct net_device *dev, 1368 struct sk_buff *skb, 1369 struct net_device *sb_dev); 1370 void (*ndo_change_rx_flags)(struct net_device *dev, 1371 int flags); 1372 void (*ndo_set_rx_mode)(struct net_device *dev); 1373 int (*ndo_set_mac_address)(struct net_device *dev, 1374 void *addr); 1375 int (*ndo_validate_addr)(struct net_device *dev); 1376 int (*ndo_do_ioctl)(struct net_device *dev, 1377 struct ifreq *ifr, int cmd); 1378 int (*ndo_eth_ioctl)(struct net_device *dev, 1379 struct ifreq *ifr, int cmd); 1380 int (*ndo_siocbond)(struct net_device *dev, 1381 struct ifreq *ifr, int cmd); 1382 int (*ndo_siocwandev)(struct net_device *dev, 1383 struct if_settings *ifs); 1384 int (*ndo_siocdevprivate)(struct net_device *dev, 1385 struct ifreq *ifr, 1386 void __user *data, int cmd); 1387 int (*ndo_set_config)(struct net_device *dev, 1388 struct ifmap *map); 1389 int (*ndo_change_mtu)(struct net_device *dev, 1390 int new_mtu); 1391 int (*ndo_neigh_setup)(struct net_device *dev, 1392 struct neigh_parms *); 1393 void (*ndo_tx_timeout) (struct net_device *dev, 1394 unsigned int txqueue); 1395 1396 void (*ndo_get_stats64)(struct net_device *dev, 1397 struct rtnl_link_stats64 *storage); 1398 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1399 int (*ndo_get_offload_stats)(int attr_id, 1400 const struct net_device *dev, 1401 void *attr_data); 1402 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1403 1404 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1405 __be16 proto, u16 vid); 1406 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1407 __be16 proto, u16 vid); 1408 #ifdef CONFIG_NET_POLL_CONTROLLER 1409 void (*ndo_poll_controller)(struct net_device *dev); 1410 int (*ndo_netpoll_setup)(struct net_device *dev, 1411 struct netpoll_info *info); 1412 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1413 #endif 1414 int (*ndo_set_vf_mac)(struct net_device *dev, 1415 int queue, u8 *mac); 1416 int (*ndo_set_vf_vlan)(struct net_device *dev, 1417 int queue, u16 vlan, 1418 u8 qos, __be16 proto); 1419 int (*ndo_set_vf_rate)(struct net_device *dev, 1420 int vf, int min_tx_rate, 1421 int max_tx_rate); 1422 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1423 int vf, bool setting); 1424 int (*ndo_set_vf_trust)(struct net_device *dev, 1425 int vf, bool setting); 1426 int (*ndo_get_vf_config)(struct net_device *dev, 1427 int vf, 1428 struct ifla_vf_info *ivf); 1429 int (*ndo_set_vf_link_state)(struct net_device *dev, 1430 int vf, int link_state); 1431 int (*ndo_get_vf_stats)(struct net_device *dev, 1432 int vf, 1433 struct ifla_vf_stats 1434 *vf_stats); 1435 int (*ndo_set_vf_port)(struct net_device *dev, 1436 int vf, 1437 struct nlattr *port[]); 1438 int (*ndo_get_vf_port)(struct net_device *dev, 1439 int vf, struct sk_buff *skb); 1440 int (*ndo_get_vf_guid)(struct net_device *dev, 1441 int vf, 1442 struct ifla_vf_guid *node_guid, 1443 struct ifla_vf_guid *port_guid); 1444 int (*ndo_set_vf_guid)(struct net_device *dev, 1445 int vf, u64 guid, 1446 int guid_type); 1447 int (*ndo_set_vf_rss_query_en)( 1448 struct net_device *dev, 1449 int vf, bool setting); 1450 int (*ndo_setup_tc)(struct net_device *dev, 1451 enum tc_setup_type type, 1452 void *type_data); 1453 #if IS_ENABLED(CONFIG_FCOE) 1454 int (*ndo_fcoe_enable)(struct net_device *dev); 1455 int (*ndo_fcoe_disable)(struct net_device *dev); 1456 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1457 u16 xid, 1458 struct scatterlist *sgl, 1459 unsigned int sgc); 1460 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1461 u16 xid); 1462 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1463 u16 xid, 1464 struct scatterlist *sgl, 1465 unsigned int sgc); 1466 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1467 struct netdev_fcoe_hbainfo *hbainfo); 1468 #endif 1469 1470 #if IS_ENABLED(CONFIG_LIBFCOE) 1471 #define NETDEV_FCOE_WWNN 0 1472 #define NETDEV_FCOE_WWPN 1 1473 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1474 u64 *wwn, int type); 1475 #endif 1476 1477 #ifdef CONFIG_RFS_ACCEL 1478 int (*ndo_rx_flow_steer)(struct net_device *dev, 1479 const struct sk_buff *skb, 1480 u16 rxq_index, 1481 u32 flow_id); 1482 #endif 1483 int (*ndo_add_slave)(struct net_device *dev, 1484 struct net_device *slave_dev, 1485 struct netlink_ext_ack *extack); 1486 int (*ndo_del_slave)(struct net_device *dev, 1487 struct net_device *slave_dev); 1488 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1489 struct sk_buff *skb, 1490 bool all_slaves); 1491 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1492 struct sock *sk); 1493 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1494 netdev_features_t features); 1495 int (*ndo_set_features)(struct net_device *dev, 1496 netdev_features_t features); 1497 int (*ndo_neigh_construct)(struct net_device *dev, 1498 struct neighbour *n); 1499 void (*ndo_neigh_destroy)(struct net_device *dev, 1500 struct neighbour *n); 1501 1502 int (*ndo_fdb_add)(struct ndmsg *ndm, 1503 struct nlattr *tb[], 1504 struct net_device *dev, 1505 const unsigned char *addr, 1506 u16 vid, 1507 u16 flags, 1508 struct netlink_ext_ack *extack); 1509 int (*ndo_fdb_del)(struct ndmsg *ndm, 1510 struct nlattr *tb[], 1511 struct net_device *dev, 1512 const unsigned char *addr, 1513 u16 vid); 1514 int (*ndo_fdb_dump)(struct sk_buff *skb, 1515 struct netlink_callback *cb, 1516 struct net_device *dev, 1517 struct net_device *filter_dev, 1518 int *idx); 1519 int (*ndo_fdb_get)(struct sk_buff *skb, 1520 struct nlattr *tb[], 1521 struct net_device *dev, 1522 const unsigned char *addr, 1523 u16 vid, u32 portid, u32 seq, 1524 struct netlink_ext_ack *extack); 1525 int (*ndo_bridge_setlink)(struct net_device *dev, 1526 struct nlmsghdr *nlh, 1527 u16 flags, 1528 struct netlink_ext_ack *extack); 1529 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1530 u32 pid, u32 seq, 1531 struct net_device *dev, 1532 u32 filter_mask, 1533 int nlflags); 1534 int (*ndo_bridge_dellink)(struct net_device *dev, 1535 struct nlmsghdr *nlh, 1536 u16 flags); 1537 int (*ndo_change_carrier)(struct net_device *dev, 1538 bool new_carrier); 1539 int (*ndo_get_phys_port_id)(struct net_device *dev, 1540 struct netdev_phys_item_id *ppid); 1541 int (*ndo_get_port_parent_id)(struct net_device *dev, 1542 struct netdev_phys_item_id *ppid); 1543 int (*ndo_get_phys_port_name)(struct net_device *dev, 1544 char *name, size_t len); 1545 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1546 struct net_device *dev); 1547 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1548 void *priv); 1549 1550 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1551 int queue_index, 1552 u32 maxrate); 1553 int (*ndo_get_iflink)(const struct net_device *dev); 1554 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1555 struct sk_buff *skb); 1556 void (*ndo_set_rx_headroom)(struct net_device *dev, 1557 int needed_headroom); 1558 int (*ndo_bpf)(struct net_device *dev, 1559 struct netdev_bpf *bpf); 1560 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1561 struct xdp_frame **xdp, 1562 u32 flags); 1563 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1564 struct xdp_buff *xdp); 1565 int (*ndo_xsk_wakeup)(struct net_device *dev, 1566 u32 queue_id, u32 flags); 1567 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1568 int (*ndo_tunnel_ctl)(struct net_device *dev, 1569 struct ip_tunnel_parm *p, int cmd); 1570 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1571 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1572 struct net_device_path *path); 1573 }; 1574 1575 /** 1576 * enum netdev_priv_flags - &struct net_device priv_flags 1577 * 1578 * These are the &struct net_device, they are only set internally 1579 * by drivers and used in the kernel. These flags are invisible to 1580 * userspace; this means that the order of these flags can change 1581 * during any kernel release. 1582 * 1583 * You should have a pretty good reason to be extending these flags. 1584 * 1585 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1586 * @IFF_EBRIDGE: Ethernet bridging device 1587 * @IFF_BONDING: bonding master or slave 1588 * @IFF_ISATAP: ISATAP interface (RFC4214) 1589 * @IFF_WAN_HDLC: WAN HDLC device 1590 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1591 * release skb->dst 1592 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1593 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1594 * @IFF_MACVLAN_PORT: device used as macvlan port 1595 * @IFF_BRIDGE_PORT: device used as bridge port 1596 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1597 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1598 * @IFF_UNICAST_FLT: Supports unicast filtering 1599 * @IFF_TEAM_PORT: device used as team port 1600 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1601 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1602 * change when it's running 1603 * @IFF_MACVLAN: Macvlan device 1604 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1605 * underlying stacked devices 1606 * @IFF_L3MDEV_MASTER: device is an L3 master device 1607 * @IFF_NO_QUEUE: device can run without qdisc attached 1608 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1609 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1610 * @IFF_TEAM: device is a team device 1611 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1612 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1613 * entity (i.e. the master device for bridged veth) 1614 * @IFF_MACSEC: device is a MACsec device 1615 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1616 * @IFF_FAILOVER: device is a failover master device 1617 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1618 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1619 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1620 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1621 * skb_headlen(skb) == 0 (data starts from frag0) 1622 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN 1623 */ 1624 enum netdev_priv_flags { 1625 IFF_802_1Q_VLAN = 1<<0, 1626 IFF_EBRIDGE = 1<<1, 1627 IFF_BONDING = 1<<2, 1628 IFF_ISATAP = 1<<3, 1629 IFF_WAN_HDLC = 1<<4, 1630 IFF_XMIT_DST_RELEASE = 1<<5, 1631 IFF_DONT_BRIDGE = 1<<6, 1632 IFF_DISABLE_NETPOLL = 1<<7, 1633 IFF_MACVLAN_PORT = 1<<8, 1634 IFF_BRIDGE_PORT = 1<<9, 1635 IFF_OVS_DATAPATH = 1<<10, 1636 IFF_TX_SKB_SHARING = 1<<11, 1637 IFF_UNICAST_FLT = 1<<12, 1638 IFF_TEAM_PORT = 1<<13, 1639 IFF_SUPP_NOFCS = 1<<14, 1640 IFF_LIVE_ADDR_CHANGE = 1<<15, 1641 IFF_MACVLAN = 1<<16, 1642 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1643 IFF_L3MDEV_MASTER = 1<<18, 1644 IFF_NO_QUEUE = 1<<19, 1645 IFF_OPENVSWITCH = 1<<20, 1646 IFF_L3MDEV_SLAVE = 1<<21, 1647 IFF_TEAM = 1<<22, 1648 IFF_RXFH_CONFIGURED = 1<<23, 1649 IFF_PHONY_HEADROOM = 1<<24, 1650 IFF_MACSEC = 1<<25, 1651 IFF_NO_RX_HANDLER = 1<<26, 1652 IFF_FAILOVER = 1<<27, 1653 IFF_FAILOVER_SLAVE = 1<<28, 1654 IFF_L3MDEV_RX_HANDLER = 1<<29, 1655 IFF_LIVE_RENAME_OK = 1<<30, 1656 IFF_TX_SKB_NO_LINEAR = 1<<31, 1657 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32), 1658 }; 1659 1660 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1661 #define IFF_EBRIDGE IFF_EBRIDGE 1662 #define IFF_BONDING IFF_BONDING 1663 #define IFF_ISATAP IFF_ISATAP 1664 #define IFF_WAN_HDLC IFF_WAN_HDLC 1665 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1666 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1667 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1668 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1669 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1670 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1671 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1672 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1673 #define IFF_TEAM_PORT IFF_TEAM_PORT 1674 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1675 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1676 #define IFF_MACVLAN IFF_MACVLAN 1677 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1678 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1679 #define IFF_NO_QUEUE IFF_NO_QUEUE 1680 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1681 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1682 #define IFF_TEAM IFF_TEAM 1683 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1684 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM 1685 #define IFF_MACSEC IFF_MACSEC 1686 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1687 #define IFF_FAILOVER IFF_FAILOVER 1688 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1689 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1690 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1691 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR 1692 1693 /* Specifies the type of the struct net_device::ml_priv pointer */ 1694 enum netdev_ml_priv_type { 1695 ML_PRIV_NONE, 1696 ML_PRIV_CAN, 1697 }; 1698 1699 /** 1700 * struct net_device - The DEVICE structure. 1701 * 1702 * Actually, this whole structure is a big mistake. It mixes I/O 1703 * data with strictly "high-level" data, and it has to know about 1704 * almost every data structure used in the INET module. 1705 * 1706 * @name: This is the first field of the "visible" part of this structure 1707 * (i.e. as seen by users in the "Space.c" file). It is the name 1708 * of the interface. 1709 * 1710 * @name_node: Name hashlist node 1711 * @ifalias: SNMP alias 1712 * @mem_end: Shared memory end 1713 * @mem_start: Shared memory start 1714 * @base_addr: Device I/O address 1715 * @irq: Device IRQ number 1716 * 1717 * @state: Generic network queuing layer state, see netdev_state_t 1718 * @dev_list: The global list of network devices 1719 * @napi_list: List entry used for polling NAPI devices 1720 * @unreg_list: List entry when we are unregistering the 1721 * device; see the function unregister_netdev 1722 * @close_list: List entry used when we are closing the device 1723 * @ptype_all: Device-specific packet handlers for all protocols 1724 * @ptype_specific: Device-specific, protocol-specific packet handlers 1725 * 1726 * @adj_list: Directly linked devices, like slaves for bonding 1727 * @features: Currently active device features 1728 * @hw_features: User-changeable features 1729 * 1730 * @wanted_features: User-requested features 1731 * @vlan_features: Mask of features inheritable by VLAN devices 1732 * 1733 * @hw_enc_features: Mask of features inherited by encapsulating devices 1734 * This field indicates what encapsulation 1735 * offloads the hardware is capable of doing, 1736 * and drivers will need to set them appropriately. 1737 * 1738 * @mpls_features: Mask of features inheritable by MPLS 1739 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1740 * 1741 * @ifindex: interface index 1742 * @group: The group the device belongs to 1743 * 1744 * @stats: Statistics struct, which was left as a legacy, use 1745 * rtnl_link_stats64 instead 1746 * 1747 * @core_stats: core networking counters, 1748 * do not use this in drivers 1749 * @carrier_up_count: Number of times the carrier has been up 1750 * @carrier_down_count: Number of times the carrier has been down 1751 * 1752 * @wireless_handlers: List of functions to handle Wireless Extensions, 1753 * instead of ioctl, 1754 * see <net/iw_handler.h> for details. 1755 * @wireless_data: Instance data managed by the core of wireless extensions 1756 * 1757 * @netdev_ops: Includes several pointers to callbacks, 1758 * if one wants to override the ndo_*() functions 1759 * @ethtool_ops: Management operations 1760 * @l3mdev_ops: Layer 3 master device operations 1761 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1762 * discovery handling. Necessary for e.g. 6LoWPAN. 1763 * @xfrmdev_ops: Transformation offload operations 1764 * @tlsdev_ops: Transport Layer Security offload operations 1765 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1766 * of Layer 2 headers. 1767 * 1768 * @flags: Interface flags (a la BSD) 1769 * @priv_flags: Like 'flags' but invisible to userspace, 1770 * see if.h for the definitions 1771 * @gflags: Global flags ( kept as legacy ) 1772 * @padded: How much padding added by alloc_netdev() 1773 * @operstate: RFC2863 operstate 1774 * @link_mode: Mapping policy to operstate 1775 * @if_port: Selectable AUI, TP, ... 1776 * @dma: DMA channel 1777 * @mtu: Interface MTU value 1778 * @min_mtu: Interface Minimum MTU value 1779 * @max_mtu: Interface Maximum MTU value 1780 * @type: Interface hardware type 1781 * @hard_header_len: Maximum hardware header length. 1782 * @min_header_len: Minimum hardware header length 1783 * 1784 * @needed_headroom: Extra headroom the hardware may need, but not in all 1785 * cases can this be guaranteed 1786 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1787 * cases can this be guaranteed. Some cases also use 1788 * LL_MAX_HEADER instead to allocate the skb 1789 * 1790 * interface address info: 1791 * 1792 * @perm_addr: Permanent hw address 1793 * @addr_assign_type: Hw address assignment type 1794 * @addr_len: Hardware address length 1795 * @upper_level: Maximum depth level of upper devices. 1796 * @lower_level: Maximum depth level of lower devices. 1797 * @neigh_priv_len: Used in neigh_alloc() 1798 * @dev_id: Used to differentiate devices that share 1799 * the same link layer address 1800 * @dev_port: Used to differentiate devices that share 1801 * the same function 1802 * @addr_list_lock: XXX: need comments on this one 1803 * @name_assign_type: network interface name assignment type 1804 * @uc_promisc: Counter that indicates promiscuous mode 1805 * has been enabled due to the need to listen to 1806 * additional unicast addresses in a device that 1807 * does not implement ndo_set_rx_mode() 1808 * @uc: unicast mac addresses 1809 * @mc: multicast mac addresses 1810 * @dev_addrs: list of device hw addresses 1811 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1812 * @promiscuity: Number of times the NIC is told to work in 1813 * promiscuous mode; if it becomes 0 the NIC will 1814 * exit promiscuous mode 1815 * @allmulti: Counter, enables or disables allmulticast mode 1816 * 1817 * @vlan_info: VLAN info 1818 * @dsa_ptr: dsa specific data 1819 * @tipc_ptr: TIPC specific data 1820 * @atalk_ptr: AppleTalk link 1821 * @ip_ptr: IPv4 specific data 1822 * @dn_ptr: DECnet specific data 1823 * @ip6_ptr: IPv6 specific data 1824 * @ax25_ptr: AX.25 specific data 1825 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1826 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1827 * device struct 1828 * @mpls_ptr: mpls_dev struct pointer 1829 * @mctp_ptr: MCTP specific data 1830 * 1831 * @dev_addr: Hw address (before bcast, 1832 * because most packets are unicast) 1833 * 1834 * @_rx: Array of RX queues 1835 * @num_rx_queues: Number of RX queues 1836 * allocated at register_netdev() time 1837 * @real_num_rx_queues: Number of RX queues currently active in device 1838 * @xdp_prog: XDP sockets filter program pointer 1839 * @gro_flush_timeout: timeout for GRO layer in NAPI 1840 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1841 * allow to avoid NIC hard IRQ, on busy queues. 1842 * 1843 * @rx_handler: handler for received packets 1844 * @rx_handler_data: XXX: need comments on this one 1845 * @miniq_ingress: ingress/clsact qdisc specific data for 1846 * ingress processing 1847 * @ingress_queue: XXX: need comments on this one 1848 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1849 * @broadcast: hw bcast address 1850 * 1851 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1852 * indexed by RX queue number. Assigned by driver. 1853 * This must only be set if the ndo_rx_flow_steer 1854 * operation is defined 1855 * @index_hlist: Device index hash chain 1856 * 1857 * @_tx: Array of TX queues 1858 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1859 * @real_num_tx_queues: Number of TX queues currently active in device 1860 * @qdisc: Root qdisc from userspace point of view 1861 * @tx_queue_len: Max frames per queue allowed 1862 * @tx_global_lock: XXX: need comments on this one 1863 * @xdp_bulkq: XDP device bulk queue 1864 * @xps_maps: all CPUs/RXQs maps for XPS device 1865 * 1866 * @xps_maps: XXX: need comments on this one 1867 * @miniq_egress: clsact qdisc specific data for 1868 * egress processing 1869 * @nf_hooks_egress: netfilter hooks executed for egress packets 1870 * @qdisc_hash: qdisc hash table 1871 * @watchdog_timeo: Represents the timeout that is used by 1872 * the watchdog (see dev_watchdog()) 1873 * @watchdog_timer: List of timers 1874 * 1875 * @proto_down_reason: reason a netdev interface is held down 1876 * @pcpu_refcnt: Number of references to this device 1877 * @dev_refcnt: Number of references to this device 1878 * @refcnt_tracker: Tracker directory for tracked references to this device 1879 * @todo_list: Delayed register/unregister 1880 * @link_watch_list: XXX: need comments on this one 1881 * 1882 * @reg_state: Register/unregister state machine 1883 * @dismantle: Device is going to be freed 1884 * @rtnl_link_state: This enum represents the phases of creating 1885 * a new link 1886 * 1887 * @needs_free_netdev: Should unregister perform free_netdev? 1888 * @priv_destructor: Called from unregister 1889 * @npinfo: XXX: need comments on this one 1890 * @nd_net: Network namespace this network device is inside 1891 * 1892 * @ml_priv: Mid-layer private 1893 * @ml_priv_type: Mid-layer private type 1894 * @lstats: Loopback statistics 1895 * @tstats: Tunnel statistics 1896 * @dstats: Dummy statistics 1897 * @vstats: Virtual ethernet statistics 1898 * 1899 * @garp_port: GARP 1900 * @mrp_port: MRP 1901 * 1902 * @dev: Class/net/name entry 1903 * @sysfs_groups: Space for optional device, statistics and wireless 1904 * sysfs groups 1905 * 1906 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1907 * @rtnl_link_ops: Rtnl_link_ops 1908 * 1909 * @gso_max_size: Maximum size of generic segmentation offload 1910 * @gso_max_segs: Maximum number of segments that can be passed to the 1911 * NIC for GSO 1912 * 1913 * @dcbnl_ops: Data Center Bridging netlink ops 1914 * @num_tc: Number of traffic classes in the net device 1915 * @tc_to_txq: XXX: need comments on this one 1916 * @prio_tc_map: XXX: need comments on this one 1917 * 1918 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1919 * 1920 * @priomap: XXX: need comments on this one 1921 * @phydev: Physical device may attach itself 1922 * for hardware timestamping 1923 * @sfp_bus: attached &struct sfp_bus structure. 1924 * 1925 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1926 * 1927 * @proto_down: protocol port state information can be sent to the 1928 * switch driver and used to set the phys state of the 1929 * switch port. 1930 * 1931 * @wol_enabled: Wake-on-LAN is enabled 1932 * 1933 * @threaded: napi threaded mode is enabled 1934 * 1935 * @net_notifier_list: List of per-net netdev notifier block 1936 * that follow this device when it is moved 1937 * to another network namespace. 1938 * 1939 * @macsec_ops: MACsec offloading ops 1940 * 1941 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1942 * offload capabilities of the device 1943 * @udp_tunnel_nic: UDP tunnel offload state 1944 * @xdp_state: stores info on attached XDP BPF programs 1945 * 1946 * @nested_level: Used as a parameter of spin_lock_nested() of 1947 * dev->addr_list_lock. 1948 * @unlink_list: As netif_addr_lock() can be called recursively, 1949 * keep a list of interfaces to be deleted. 1950 * @gro_max_size: Maximum size of aggregated packet in generic 1951 * receive offload (GRO) 1952 * 1953 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 1954 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 1955 * @watchdog_dev_tracker: refcount tracker used by watchdog. 1956 * @dev_registered_tracker: tracker for reference held while 1957 * registered 1958 * @offload_xstats_l3: L3 HW stats for this netdevice. 1959 * 1960 * FIXME: cleanup struct net_device such that network protocol info 1961 * moves out. 1962 */ 1963 1964 struct net_device { 1965 char name[IFNAMSIZ]; 1966 struct netdev_name_node *name_node; 1967 struct dev_ifalias __rcu *ifalias; 1968 /* 1969 * I/O specific fields 1970 * FIXME: Merge these and struct ifmap into one 1971 */ 1972 unsigned long mem_end; 1973 unsigned long mem_start; 1974 unsigned long base_addr; 1975 1976 /* 1977 * Some hardware also needs these fields (state,dev_list, 1978 * napi_list,unreg_list,close_list) but they are not 1979 * part of the usual set specified in Space.c. 1980 */ 1981 1982 unsigned long state; 1983 1984 struct list_head dev_list; 1985 struct list_head napi_list; 1986 struct list_head unreg_list; 1987 struct list_head close_list; 1988 struct list_head ptype_all; 1989 struct list_head ptype_specific; 1990 1991 struct { 1992 struct list_head upper; 1993 struct list_head lower; 1994 } adj_list; 1995 1996 /* Read-mostly cache-line for fast-path access */ 1997 unsigned int flags; 1998 unsigned long long priv_flags; 1999 const struct net_device_ops *netdev_ops; 2000 int ifindex; 2001 unsigned short gflags; 2002 unsigned short hard_header_len; 2003 2004 /* Note : dev->mtu is often read without holding a lock. 2005 * Writers usually hold RTNL. 2006 * It is recommended to use READ_ONCE() to annotate the reads, 2007 * and to use WRITE_ONCE() to annotate the writes. 2008 */ 2009 unsigned int mtu; 2010 unsigned short needed_headroom; 2011 unsigned short needed_tailroom; 2012 2013 netdev_features_t features; 2014 netdev_features_t hw_features; 2015 netdev_features_t wanted_features; 2016 netdev_features_t vlan_features; 2017 netdev_features_t hw_enc_features; 2018 netdev_features_t mpls_features; 2019 netdev_features_t gso_partial_features; 2020 2021 unsigned int min_mtu; 2022 unsigned int max_mtu; 2023 unsigned short type; 2024 unsigned char min_header_len; 2025 unsigned char name_assign_type; 2026 2027 int group; 2028 2029 struct net_device_stats stats; /* not used by modern drivers */ 2030 2031 struct net_device_core_stats __percpu *core_stats; 2032 2033 /* Stats to monitor link on/off, flapping */ 2034 atomic_t carrier_up_count; 2035 atomic_t carrier_down_count; 2036 2037 #ifdef CONFIG_WIRELESS_EXT 2038 const struct iw_handler_def *wireless_handlers; 2039 struct iw_public_data *wireless_data; 2040 #endif 2041 const struct ethtool_ops *ethtool_ops; 2042 #ifdef CONFIG_NET_L3_MASTER_DEV 2043 const struct l3mdev_ops *l3mdev_ops; 2044 #endif 2045 #if IS_ENABLED(CONFIG_IPV6) 2046 const struct ndisc_ops *ndisc_ops; 2047 #endif 2048 2049 #ifdef CONFIG_XFRM_OFFLOAD 2050 const struct xfrmdev_ops *xfrmdev_ops; 2051 #endif 2052 2053 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2054 const struct tlsdev_ops *tlsdev_ops; 2055 #endif 2056 2057 const struct header_ops *header_ops; 2058 2059 unsigned char operstate; 2060 unsigned char link_mode; 2061 2062 unsigned char if_port; 2063 unsigned char dma; 2064 2065 /* Interface address info. */ 2066 unsigned char perm_addr[MAX_ADDR_LEN]; 2067 unsigned char addr_assign_type; 2068 unsigned char addr_len; 2069 unsigned char upper_level; 2070 unsigned char lower_level; 2071 2072 unsigned short neigh_priv_len; 2073 unsigned short dev_id; 2074 unsigned short dev_port; 2075 unsigned short padded; 2076 2077 spinlock_t addr_list_lock; 2078 int irq; 2079 2080 struct netdev_hw_addr_list uc; 2081 struct netdev_hw_addr_list mc; 2082 struct netdev_hw_addr_list dev_addrs; 2083 2084 #ifdef CONFIG_SYSFS 2085 struct kset *queues_kset; 2086 #endif 2087 #ifdef CONFIG_LOCKDEP 2088 struct list_head unlink_list; 2089 #endif 2090 unsigned int promiscuity; 2091 unsigned int allmulti; 2092 bool uc_promisc; 2093 #ifdef CONFIG_LOCKDEP 2094 unsigned char nested_level; 2095 #endif 2096 2097 2098 /* Protocol-specific pointers */ 2099 2100 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2101 struct vlan_info __rcu *vlan_info; 2102 #endif 2103 #if IS_ENABLED(CONFIG_NET_DSA) 2104 struct dsa_port *dsa_ptr; 2105 #endif 2106 #if IS_ENABLED(CONFIG_TIPC) 2107 struct tipc_bearer __rcu *tipc_ptr; 2108 #endif 2109 #if IS_ENABLED(CONFIG_ATALK) 2110 void *atalk_ptr; 2111 #endif 2112 struct in_device __rcu *ip_ptr; 2113 #if IS_ENABLED(CONFIG_DECNET) 2114 struct dn_dev __rcu *dn_ptr; 2115 #endif 2116 struct inet6_dev __rcu *ip6_ptr; 2117 #if IS_ENABLED(CONFIG_AX25) 2118 void *ax25_ptr; 2119 #endif 2120 struct wireless_dev *ieee80211_ptr; 2121 struct wpan_dev *ieee802154_ptr; 2122 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2123 struct mpls_dev __rcu *mpls_ptr; 2124 #endif 2125 #if IS_ENABLED(CONFIG_MCTP) 2126 struct mctp_dev __rcu *mctp_ptr; 2127 #endif 2128 2129 /* 2130 * Cache lines mostly used on receive path (including eth_type_trans()) 2131 */ 2132 /* Interface address info used in eth_type_trans() */ 2133 const unsigned char *dev_addr; 2134 2135 struct netdev_rx_queue *_rx; 2136 unsigned int num_rx_queues; 2137 unsigned int real_num_rx_queues; 2138 2139 struct bpf_prog __rcu *xdp_prog; 2140 unsigned long gro_flush_timeout; 2141 int napi_defer_hard_irqs; 2142 #define GRO_MAX_SIZE 65536 2143 unsigned int gro_max_size; 2144 rx_handler_func_t __rcu *rx_handler; 2145 void __rcu *rx_handler_data; 2146 2147 #ifdef CONFIG_NET_CLS_ACT 2148 struct mini_Qdisc __rcu *miniq_ingress; 2149 #endif 2150 struct netdev_queue __rcu *ingress_queue; 2151 #ifdef CONFIG_NETFILTER_INGRESS 2152 struct nf_hook_entries __rcu *nf_hooks_ingress; 2153 #endif 2154 2155 unsigned char broadcast[MAX_ADDR_LEN]; 2156 #ifdef CONFIG_RFS_ACCEL 2157 struct cpu_rmap *rx_cpu_rmap; 2158 #endif 2159 struct hlist_node index_hlist; 2160 2161 /* 2162 * Cache lines mostly used on transmit path 2163 */ 2164 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2165 unsigned int num_tx_queues; 2166 unsigned int real_num_tx_queues; 2167 struct Qdisc __rcu *qdisc; 2168 unsigned int tx_queue_len; 2169 spinlock_t tx_global_lock; 2170 2171 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2172 2173 #ifdef CONFIG_XPS 2174 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2175 #endif 2176 #ifdef CONFIG_NET_CLS_ACT 2177 struct mini_Qdisc __rcu *miniq_egress; 2178 #endif 2179 #ifdef CONFIG_NETFILTER_EGRESS 2180 struct nf_hook_entries __rcu *nf_hooks_egress; 2181 #endif 2182 2183 #ifdef CONFIG_NET_SCHED 2184 DECLARE_HASHTABLE (qdisc_hash, 4); 2185 #endif 2186 /* These may be needed for future network-power-down code. */ 2187 struct timer_list watchdog_timer; 2188 int watchdog_timeo; 2189 2190 u32 proto_down_reason; 2191 2192 struct list_head todo_list; 2193 2194 #ifdef CONFIG_PCPU_DEV_REFCNT 2195 int __percpu *pcpu_refcnt; 2196 #else 2197 refcount_t dev_refcnt; 2198 #endif 2199 struct ref_tracker_dir refcnt_tracker; 2200 2201 struct list_head link_watch_list; 2202 2203 enum { NETREG_UNINITIALIZED=0, 2204 NETREG_REGISTERED, /* completed register_netdevice */ 2205 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2206 NETREG_UNREGISTERED, /* completed unregister todo */ 2207 NETREG_RELEASED, /* called free_netdev */ 2208 NETREG_DUMMY, /* dummy device for NAPI poll */ 2209 } reg_state:8; 2210 2211 bool dismantle; 2212 2213 enum { 2214 RTNL_LINK_INITIALIZED, 2215 RTNL_LINK_INITIALIZING, 2216 } rtnl_link_state:16; 2217 2218 bool needs_free_netdev; 2219 void (*priv_destructor)(struct net_device *dev); 2220 2221 #ifdef CONFIG_NETPOLL 2222 struct netpoll_info __rcu *npinfo; 2223 #endif 2224 2225 possible_net_t nd_net; 2226 2227 /* mid-layer private */ 2228 void *ml_priv; 2229 enum netdev_ml_priv_type ml_priv_type; 2230 2231 union { 2232 struct pcpu_lstats __percpu *lstats; 2233 struct pcpu_sw_netstats __percpu *tstats; 2234 struct pcpu_dstats __percpu *dstats; 2235 }; 2236 2237 #if IS_ENABLED(CONFIG_GARP) 2238 struct garp_port __rcu *garp_port; 2239 #endif 2240 #if IS_ENABLED(CONFIG_MRP) 2241 struct mrp_port __rcu *mrp_port; 2242 #endif 2243 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2244 struct dm_hw_stat_delta __rcu *dm_private; 2245 #endif 2246 struct device dev; 2247 const struct attribute_group *sysfs_groups[4]; 2248 const struct attribute_group *sysfs_rx_queue_group; 2249 2250 const struct rtnl_link_ops *rtnl_link_ops; 2251 2252 /* for setting kernel sock attribute on TCP connection setup */ 2253 #define GSO_MAX_SIZE 65536 2254 unsigned int gso_max_size; 2255 #define GSO_MAX_SEGS 65535 2256 u16 gso_max_segs; 2257 2258 #ifdef CONFIG_DCB 2259 const struct dcbnl_rtnl_ops *dcbnl_ops; 2260 #endif 2261 s16 num_tc; 2262 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2263 u8 prio_tc_map[TC_BITMASK + 1]; 2264 2265 #if IS_ENABLED(CONFIG_FCOE) 2266 unsigned int fcoe_ddp_xid; 2267 #endif 2268 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2269 struct netprio_map __rcu *priomap; 2270 #endif 2271 struct phy_device *phydev; 2272 struct sfp_bus *sfp_bus; 2273 struct lock_class_key *qdisc_tx_busylock; 2274 bool proto_down; 2275 unsigned wol_enabled:1; 2276 unsigned threaded:1; 2277 2278 struct list_head net_notifier_list; 2279 2280 #if IS_ENABLED(CONFIG_MACSEC) 2281 /* MACsec management functions */ 2282 const struct macsec_ops *macsec_ops; 2283 #endif 2284 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2285 struct udp_tunnel_nic *udp_tunnel_nic; 2286 2287 /* protected by rtnl_lock */ 2288 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2289 2290 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2291 netdevice_tracker linkwatch_dev_tracker; 2292 netdevice_tracker watchdog_dev_tracker; 2293 netdevice_tracker dev_registered_tracker; 2294 struct rtnl_hw_stats64 *offload_xstats_l3; 2295 }; 2296 #define to_net_dev(d) container_of(d, struct net_device, dev) 2297 2298 static inline bool netif_elide_gro(const struct net_device *dev) 2299 { 2300 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2301 return true; 2302 return false; 2303 } 2304 2305 #define NETDEV_ALIGN 32 2306 2307 static inline 2308 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2309 { 2310 return dev->prio_tc_map[prio & TC_BITMASK]; 2311 } 2312 2313 static inline 2314 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2315 { 2316 if (tc >= dev->num_tc) 2317 return -EINVAL; 2318 2319 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2320 return 0; 2321 } 2322 2323 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2324 void netdev_reset_tc(struct net_device *dev); 2325 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2326 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2327 2328 static inline 2329 int netdev_get_num_tc(struct net_device *dev) 2330 { 2331 return dev->num_tc; 2332 } 2333 2334 static inline void net_prefetch(void *p) 2335 { 2336 prefetch(p); 2337 #if L1_CACHE_BYTES < 128 2338 prefetch((u8 *)p + L1_CACHE_BYTES); 2339 #endif 2340 } 2341 2342 static inline void net_prefetchw(void *p) 2343 { 2344 prefetchw(p); 2345 #if L1_CACHE_BYTES < 128 2346 prefetchw((u8 *)p + L1_CACHE_BYTES); 2347 #endif 2348 } 2349 2350 void netdev_unbind_sb_channel(struct net_device *dev, 2351 struct net_device *sb_dev); 2352 int netdev_bind_sb_channel_queue(struct net_device *dev, 2353 struct net_device *sb_dev, 2354 u8 tc, u16 count, u16 offset); 2355 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2356 static inline int netdev_get_sb_channel(struct net_device *dev) 2357 { 2358 return max_t(int, -dev->num_tc, 0); 2359 } 2360 2361 static inline 2362 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2363 unsigned int index) 2364 { 2365 return &dev->_tx[index]; 2366 } 2367 2368 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2369 const struct sk_buff *skb) 2370 { 2371 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2372 } 2373 2374 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2375 void (*f)(struct net_device *, 2376 struct netdev_queue *, 2377 void *), 2378 void *arg) 2379 { 2380 unsigned int i; 2381 2382 for (i = 0; i < dev->num_tx_queues; i++) 2383 f(dev, &dev->_tx[i], arg); 2384 } 2385 2386 #define netdev_lockdep_set_classes(dev) \ 2387 { \ 2388 static struct lock_class_key qdisc_tx_busylock_key; \ 2389 static struct lock_class_key qdisc_xmit_lock_key; \ 2390 static struct lock_class_key dev_addr_list_lock_key; \ 2391 unsigned int i; \ 2392 \ 2393 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2394 lockdep_set_class(&(dev)->addr_list_lock, \ 2395 &dev_addr_list_lock_key); \ 2396 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2397 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2398 &qdisc_xmit_lock_key); \ 2399 } 2400 2401 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2402 struct net_device *sb_dev); 2403 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2404 struct sk_buff *skb, 2405 struct net_device *sb_dev); 2406 2407 /* returns the headroom that the master device needs to take in account 2408 * when forwarding to this dev 2409 */ 2410 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2411 { 2412 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2413 } 2414 2415 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2416 { 2417 if (dev->netdev_ops->ndo_set_rx_headroom) 2418 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2419 } 2420 2421 /* set the device rx headroom to the dev's default */ 2422 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2423 { 2424 netdev_set_rx_headroom(dev, -1); 2425 } 2426 2427 static inline void *netdev_get_ml_priv(struct net_device *dev, 2428 enum netdev_ml_priv_type type) 2429 { 2430 if (dev->ml_priv_type != type) 2431 return NULL; 2432 2433 return dev->ml_priv; 2434 } 2435 2436 static inline void netdev_set_ml_priv(struct net_device *dev, 2437 void *ml_priv, 2438 enum netdev_ml_priv_type type) 2439 { 2440 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2441 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2442 dev->ml_priv_type, type); 2443 WARN(!dev->ml_priv_type && dev->ml_priv, 2444 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2445 2446 dev->ml_priv = ml_priv; 2447 dev->ml_priv_type = type; 2448 } 2449 2450 /* 2451 * Net namespace inlines 2452 */ 2453 static inline 2454 struct net *dev_net(const struct net_device *dev) 2455 { 2456 return read_pnet(&dev->nd_net); 2457 } 2458 2459 static inline 2460 void dev_net_set(struct net_device *dev, struct net *net) 2461 { 2462 write_pnet(&dev->nd_net, net); 2463 } 2464 2465 /** 2466 * netdev_priv - access network device private data 2467 * @dev: network device 2468 * 2469 * Get network device private data 2470 */ 2471 static inline void *netdev_priv(const struct net_device *dev) 2472 { 2473 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2474 } 2475 2476 /* Set the sysfs physical device reference for the network logical device 2477 * if set prior to registration will cause a symlink during initialization. 2478 */ 2479 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2480 2481 /* Set the sysfs device type for the network logical device to allow 2482 * fine-grained identification of different network device types. For 2483 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2484 */ 2485 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2486 2487 /* Default NAPI poll() weight 2488 * Device drivers are strongly advised to not use bigger value 2489 */ 2490 #define NAPI_POLL_WEIGHT 64 2491 2492 /** 2493 * netif_napi_add - initialize a NAPI context 2494 * @dev: network device 2495 * @napi: NAPI context 2496 * @poll: polling function 2497 * @weight: default weight 2498 * 2499 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2500 * *any* of the other NAPI-related functions. 2501 */ 2502 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2503 int (*poll)(struct napi_struct *, int), int weight); 2504 2505 /** 2506 * netif_tx_napi_add - initialize a NAPI context 2507 * @dev: network device 2508 * @napi: NAPI context 2509 * @poll: polling function 2510 * @weight: default weight 2511 * 2512 * This variant of netif_napi_add() should be used from drivers using NAPI 2513 * to exclusively poll a TX queue. 2514 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2515 */ 2516 static inline void netif_tx_napi_add(struct net_device *dev, 2517 struct napi_struct *napi, 2518 int (*poll)(struct napi_struct *, int), 2519 int weight) 2520 { 2521 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2522 netif_napi_add(dev, napi, poll, weight); 2523 } 2524 2525 /** 2526 * __netif_napi_del - remove a NAPI context 2527 * @napi: NAPI context 2528 * 2529 * Warning: caller must observe RCU grace period before freeing memory 2530 * containing @napi. Drivers might want to call this helper to combine 2531 * all the needed RCU grace periods into a single one. 2532 */ 2533 void __netif_napi_del(struct napi_struct *napi); 2534 2535 /** 2536 * netif_napi_del - remove a NAPI context 2537 * @napi: NAPI context 2538 * 2539 * netif_napi_del() removes a NAPI context from the network device NAPI list 2540 */ 2541 static inline void netif_napi_del(struct napi_struct *napi) 2542 { 2543 __netif_napi_del(napi); 2544 synchronize_net(); 2545 } 2546 2547 struct packet_type { 2548 __be16 type; /* This is really htons(ether_type). */ 2549 bool ignore_outgoing; 2550 struct net_device *dev; /* NULL is wildcarded here */ 2551 netdevice_tracker dev_tracker; 2552 int (*func) (struct sk_buff *, 2553 struct net_device *, 2554 struct packet_type *, 2555 struct net_device *); 2556 void (*list_func) (struct list_head *, 2557 struct packet_type *, 2558 struct net_device *); 2559 bool (*id_match)(struct packet_type *ptype, 2560 struct sock *sk); 2561 struct net *af_packet_net; 2562 void *af_packet_priv; 2563 struct list_head list; 2564 }; 2565 2566 struct offload_callbacks { 2567 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2568 netdev_features_t features); 2569 struct sk_buff *(*gro_receive)(struct list_head *head, 2570 struct sk_buff *skb); 2571 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2572 }; 2573 2574 struct packet_offload { 2575 __be16 type; /* This is really htons(ether_type). */ 2576 u16 priority; 2577 struct offload_callbacks callbacks; 2578 struct list_head list; 2579 }; 2580 2581 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2582 struct pcpu_sw_netstats { 2583 u64 rx_packets; 2584 u64 rx_bytes; 2585 u64 tx_packets; 2586 u64 tx_bytes; 2587 struct u64_stats_sync syncp; 2588 } __aligned(4 * sizeof(u64)); 2589 2590 struct pcpu_lstats { 2591 u64_stats_t packets; 2592 u64_stats_t bytes; 2593 struct u64_stats_sync syncp; 2594 } __aligned(2 * sizeof(u64)); 2595 2596 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2597 2598 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2599 { 2600 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2601 2602 u64_stats_update_begin(&tstats->syncp); 2603 tstats->rx_bytes += len; 2604 tstats->rx_packets++; 2605 u64_stats_update_end(&tstats->syncp); 2606 } 2607 2608 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2609 unsigned int packets, 2610 unsigned int len) 2611 { 2612 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2613 2614 u64_stats_update_begin(&tstats->syncp); 2615 tstats->tx_bytes += len; 2616 tstats->tx_packets += packets; 2617 u64_stats_update_end(&tstats->syncp); 2618 } 2619 2620 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2621 { 2622 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2623 2624 u64_stats_update_begin(&lstats->syncp); 2625 u64_stats_add(&lstats->bytes, len); 2626 u64_stats_inc(&lstats->packets); 2627 u64_stats_update_end(&lstats->syncp); 2628 } 2629 2630 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2631 ({ \ 2632 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2633 if (pcpu_stats) { \ 2634 int __cpu; \ 2635 for_each_possible_cpu(__cpu) { \ 2636 typeof(type) *stat; \ 2637 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2638 u64_stats_init(&stat->syncp); \ 2639 } \ 2640 } \ 2641 pcpu_stats; \ 2642 }) 2643 2644 #define netdev_alloc_pcpu_stats(type) \ 2645 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2646 2647 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2648 ({ \ 2649 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2650 if (pcpu_stats) { \ 2651 int __cpu; \ 2652 for_each_possible_cpu(__cpu) { \ 2653 typeof(type) *stat; \ 2654 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2655 u64_stats_init(&stat->syncp); \ 2656 } \ 2657 } \ 2658 pcpu_stats; \ 2659 }) 2660 2661 enum netdev_lag_tx_type { 2662 NETDEV_LAG_TX_TYPE_UNKNOWN, 2663 NETDEV_LAG_TX_TYPE_RANDOM, 2664 NETDEV_LAG_TX_TYPE_BROADCAST, 2665 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2666 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2667 NETDEV_LAG_TX_TYPE_HASH, 2668 }; 2669 2670 enum netdev_lag_hash { 2671 NETDEV_LAG_HASH_NONE, 2672 NETDEV_LAG_HASH_L2, 2673 NETDEV_LAG_HASH_L34, 2674 NETDEV_LAG_HASH_L23, 2675 NETDEV_LAG_HASH_E23, 2676 NETDEV_LAG_HASH_E34, 2677 NETDEV_LAG_HASH_VLAN_SRCMAC, 2678 NETDEV_LAG_HASH_UNKNOWN, 2679 }; 2680 2681 struct netdev_lag_upper_info { 2682 enum netdev_lag_tx_type tx_type; 2683 enum netdev_lag_hash hash_type; 2684 }; 2685 2686 struct netdev_lag_lower_state_info { 2687 u8 link_up : 1, 2688 tx_enabled : 1; 2689 }; 2690 2691 #include <linux/notifier.h> 2692 2693 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2694 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2695 * adding new types. 2696 */ 2697 enum netdev_cmd { 2698 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2699 NETDEV_DOWN, 2700 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2701 detected a hardware crash and restarted 2702 - we can use this eg to kick tcp sessions 2703 once done */ 2704 NETDEV_CHANGE, /* Notify device state change */ 2705 NETDEV_REGISTER, 2706 NETDEV_UNREGISTER, 2707 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2708 NETDEV_CHANGEADDR, /* notify after the address change */ 2709 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2710 NETDEV_GOING_DOWN, 2711 NETDEV_CHANGENAME, 2712 NETDEV_FEAT_CHANGE, 2713 NETDEV_BONDING_FAILOVER, 2714 NETDEV_PRE_UP, 2715 NETDEV_PRE_TYPE_CHANGE, 2716 NETDEV_POST_TYPE_CHANGE, 2717 NETDEV_POST_INIT, 2718 NETDEV_RELEASE, 2719 NETDEV_NOTIFY_PEERS, 2720 NETDEV_JOIN, 2721 NETDEV_CHANGEUPPER, 2722 NETDEV_RESEND_IGMP, 2723 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2724 NETDEV_CHANGEINFODATA, 2725 NETDEV_BONDING_INFO, 2726 NETDEV_PRECHANGEUPPER, 2727 NETDEV_CHANGELOWERSTATE, 2728 NETDEV_UDP_TUNNEL_PUSH_INFO, 2729 NETDEV_UDP_TUNNEL_DROP_INFO, 2730 NETDEV_CHANGE_TX_QUEUE_LEN, 2731 NETDEV_CVLAN_FILTER_PUSH_INFO, 2732 NETDEV_CVLAN_FILTER_DROP_INFO, 2733 NETDEV_SVLAN_FILTER_PUSH_INFO, 2734 NETDEV_SVLAN_FILTER_DROP_INFO, 2735 NETDEV_OFFLOAD_XSTATS_ENABLE, 2736 NETDEV_OFFLOAD_XSTATS_DISABLE, 2737 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 2738 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 2739 }; 2740 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2741 2742 int register_netdevice_notifier(struct notifier_block *nb); 2743 int unregister_netdevice_notifier(struct notifier_block *nb); 2744 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2745 int unregister_netdevice_notifier_net(struct net *net, 2746 struct notifier_block *nb); 2747 int register_netdevice_notifier_dev_net(struct net_device *dev, 2748 struct notifier_block *nb, 2749 struct netdev_net_notifier *nn); 2750 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2751 struct notifier_block *nb, 2752 struct netdev_net_notifier *nn); 2753 2754 struct netdev_notifier_info { 2755 struct net_device *dev; 2756 struct netlink_ext_ack *extack; 2757 }; 2758 2759 struct netdev_notifier_info_ext { 2760 struct netdev_notifier_info info; /* must be first */ 2761 union { 2762 u32 mtu; 2763 } ext; 2764 }; 2765 2766 struct netdev_notifier_change_info { 2767 struct netdev_notifier_info info; /* must be first */ 2768 unsigned int flags_changed; 2769 }; 2770 2771 struct netdev_notifier_changeupper_info { 2772 struct netdev_notifier_info info; /* must be first */ 2773 struct net_device *upper_dev; /* new upper dev */ 2774 bool master; /* is upper dev master */ 2775 bool linking; /* is the notification for link or unlink */ 2776 void *upper_info; /* upper dev info */ 2777 }; 2778 2779 struct netdev_notifier_changelowerstate_info { 2780 struct netdev_notifier_info info; /* must be first */ 2781 void *lower_state_info; /* is lower dev state */ 2782 }; 2783 2784 struct netdev_notifier_pre_changeaddr_info { 2785 struct netdev_notifier_info info; /* must be first */ 2786 const unsigned char *dev_addr; 2787 }; 2788 2789 enum netdev_offload_xstats_type { 2790 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 2791 }; 2792 2793 struct netdev_notifier_offload_xstats_info { 2794 struct netdev_notifier_info info; /* must be first */ 2795 enum netdev_offload_xstats_type type; 2796 2797 union { 2798 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 2799 struct netdev_notifier_offload_xstats_rd *report_delta; 2800 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 2801 struct netdev_notifier_offload_xstats_ru *report_used; 2802 }; 2803 }; 2804 2805 int netdev_offload_xstats_enable(struct net_device *dev, 2806 enum netdev_offload_xstats_type type, 2807 struct netlink_ext_ack *extack); 2808 int netdev_offload_xstats_disable(struct net_device *dev, 2809 enum netdev_offload_xstats_type type); 2810 bool netdev_offload_xstats_enabled(const struct net_device *dev, 2811 enum netdev_offload_xstats_type type); 2812 int netdev_offload_xstats_get(struct net_device *dev, 2813 enum netdev_offload_xstats_type type, 2814 struct rtnl_hw_stats64 *stats, bool *used, 2815 struct netlink_ext_ack *extack); 2816 void 2817 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 2818 const struct rtnl_hw_stats64 *stats); 2819 void 2820 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 2821 void netdev_offload_xstats_push_delta(struct net_device *dev, 2822 enum netdev_offload_xstats_type type, 2823 const struct rtnl_hw_stats64 *stats); 2824 2825 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2826 struct net_device *dev) 2827 { 2828 info->dev = dev; 2829 info->extack = NULL; 2830 } 2831 2832 static inline struct net_device * 2833 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2834 { 2835 return info->dev; 2836 } 2837 2838 static inline struct netlink_ext_ack * 2839 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2840 { 2841 return info->extack; 2842 } 2843 2844 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2845 2846 2847 extern rwlock_t dev_base_lock; /* Device list lock */ 2848 2849 #define for_each_netdev(net, d) \ 2850 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2851 #define for_each_netdev_reverse(net, d) \ 2852 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2853 #define for_each_netdev_rcu(net, d) \ 2854 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2855 #define for_each_netdev_safe(net, d, n) \ 2856 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2857 #define for_each_netdev_continue(net, d) \ 2858 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2859 #define for_each_netdev_continue_reverse(net, d) \ 2860 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2861 dev_list) 2862 #define for_each_netdev_continue_rcu(net, d) \ 2863 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2864 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2865 for_each_netdev_rcu(&init_net, slave) \ 2866 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2867 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2868 2869 static inline struct net_device *next_net_device(struct net_device *dev) 2870 { 2871 struct list_head *lh; 2872 struct net *net; 2873 2874 net = dev_net(dev); 2875 lh = dev->dev_list.next; 2876 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2877 } 2878 2879 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2880 { 2881 struct list_head *lh; 2882 struct net *net; 2883 2884 net = dev_net(dev); 2885 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2886 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2887 } 2888 2889 static inline struct net_device *first_net_device(struct net *net) 2890 { 2891 return list_empty(&net->dev_base_head) ? NULL : 2892 net_device_entry(net->dev_base_head.next); 2893 } 2894 2895 static inline struct net_device *first_net_device_rcu(struct net *net) 2896 { 2897 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2898 2899 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2900 } 2901 2902 int netdev_boot_setup_check(struct net_device *dev); 2903 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2904 const char *hwaddr); 2905 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2906 void dev_add_pack(struct packet_type *pt); 2907 void dev_remove_pack(struct packet_type *pt); 2908 void __dev_remove_pack(struct packet_type *pt); 2909 void dev_add_offload(struct packet_offload *po); 2910 void dev_remove_offload(struct packet_offload *po); 2911 2912 int dev_get_iflink(const struct net_device *dev); 2913 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2914 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 2915 struct net_device_path_stack *stack); 2916 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2917 unsigned short mask); 2918 struct net_device *dev_get_by_name(struct net *net, const char *name); 2919 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2920 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2921 bool netdev_name_in_use(struct net *net, const char *name); 2922 int dev_alloc_name(struct net_device *dev, const char *name); 2923 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2924 void dev_close(struct net_device *dev); 2925 void dev_close_many(struct list_head *head, bool unlink); 2926 void dev_disable_lro(struct net_device *dev); 2927 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2928 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2929 struct net_device *sb_dev); 2930 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2931 struct net_device *sb_dev); 2932 2933 int dev_queue_xmit(struct sk_buff *skb); 2934 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2935 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2936 2937 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 2938 { 2939 int ret; 2940 2941 ret = __dev_direct_xmit(skb, queue_id); 2942 if (!dev_xmit_complete(ret)) 2943 kfree_skb(skb); 2944 return ret; 2945 } 2946 2947 int register_netdevice(struct net_device *dev); 2948 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2949 void unregister_netdevice_many(struct list_head *head); 2950 static inline void unregister_netdevice(struct net_device *dev) 2951 { 2952 unregister_netdevice_queue(dev, NULL); 2953 } 2954 2955 int netdev_refcnt_read(const struct net_device *dev); 2956 void free_netdev(struct net_device *dev); 2957 void netdev_freemem(struct net_device *dev); 2958 int init_dummy_netdev(struct net_device *dev); 2959 2960 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 2961 struct sk_buff *skb, 2962 bool all_slaves); 2963 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 2964 struct sock *sk); 2965 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2966 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2967 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2968 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2969 int netdev_get_name(struct net *net, char *name, int ifindex); 2970 int dev_restart(struct net_device *dev); 2971 2972 2973 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2974 unsigned short type, 2975 const void *daddr, const void *saddr, 2976 unsigned int len) 2977 { 2978 if (!dev->header_ops || !dev->header_ops->create) 2979 return 0; 2980 2981 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2982 } 2983 2984 static inline int dev_parse_header(const struct sk_buff *skb, 2985 unsigned char *haddr) 2986 { 2987 const struct net_device *dev = skb->dev; 2988 2989 if (!dev->header_ops || !dev->header_ops->parse) 2990 return 0; 2991 return dev->header_ops->parse(skb, haddr); 2992 } 2993 2994 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 2995 { 2996 const struct net_device *dev = skb->dev; 2997 2998 if (!dev->header_ops || !dev->header_ops->parse_protocol) 2999 return 0; 3000 return dev->header_ops->parse_protocol(skb); 3001 } 3002 3003 /* ll_header must have at least hard_header_len allocated */ 3004 static inline bool dev_validate_header(const struct net_device *dev, 3005 char *ll_header, int len) 3006 { 3007 if (likely(len >= dev->hard_header_len)) 3008 return true; 3009 if (len < dev->min_header_len) 3010 return false; 3011 3012 if (capable(CAP_SYS_RAWIO)) { 3013 memset(ll_header + len, 0, dev->hard_header_len - len); 3014 return true; 3015 } 3016 3017 if (dev->header_ops && dev->header_ops->validate) 3018 return dev->header_ops->validate(ll_header, len); 3019 3020 return false; 3021 } 3022 3023 static inline bool dev_has_header(const struct net_device *dev) 3024 { 3025 return dev->header_ops && dev->header_ops->create; 3026 } 3027 3028 #ifdef CONFIG_NET_FLOW_LIMIT 3029 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 3030 struct sd_flow_limit { 3031 u64 count; 3032 unsigned int num_buckets; 3033 unsigned int history_head; 3034 u16 history[FLOW_LIMIT_HISTORY]; 3035 u8 buckets[]; 3036 }; 3037 3038 extern int netdev_flow_limit_table_len; 3039 #endif /* CONFIG_NET_FLOW_LIMIT */ 3040 3041 /* 3042 * Incoming packets are placed on per-CPU queues 3043 */ 3044 struct softnet_data { 3045 struct list_head poll_list; 3046 struct sk_buff_head process_queue; 3047 3048 /* stats */ 3049 unsigned int processed; 3050 unsigned int time_squeeze; 3051 unsigned int received_rps; 3052 #ifdef CONFIG_RPS 3053 struct softnet_data *rps_ipi_list; 3054 #endif 3055 #ifdef CONFIG_NET_FLOW_LIMIT 3056 struct sd_flow_limit __rcu *flow_limit; 3057 #endif 3058 struct Qdisc *output_queue; 3059 struct Qdisc **output_queue_tailp; 3060 struct sk_buff *completion_queue; 3061 #ifdef CONFIG_XFRM_OFFLOAD 3062 struct sk_buff_head xfrm_backlog; 3063 #endif 3064 /* written and read only by owning cpu: */ 3065 struct { 3066 u16 recursion; 3067 u8 more; 3068 } xmit; 3069 #ifdef CONFIG_RPS 3070 /* input_queue_head should be written by cpu owning this struct, 3071 * and only read by other cpus. Worth using a cache line. 3072 */ 3073 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3074 3075 /* Elements below can be accessed between CPUs for RPS/RFS */ 3076 call_single_data_t csd ____cacheline_aligned_in_smp; 3077 struct softnet_data *rps_ipi_next; 3078 unsigned int cpu; 3079 unsigned int input_queue_tail; 3080 #endif 3081 unsigned int dropped; 3082 struct sk_buff_head input_pkt_queue; 3083 struct napi_struct backlog; 3084 3085 }; 3086 3087 static inline void input_queue_head_incr(struct softnet_data *sd) 3088 { 3089 #ifdef CONFIG_RPS 3090 sd->input_queue_head++; 3091 #endif 3092 } 3093 3094 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3095 unsigned int *qtail) 3096 { 3097 #ifdef CONFIG_RPS 3098 *qtail = ++sd->input_queue_tail; 3099 #endif 3100 } 3101 3102 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3103 3104 static inline int dev_recursion_level(void) 3105 { 3106 return this_cpu_read(softnet_data.xmit.recursion); 3107 } 3108 3109 #define XMIT_RECURSION_LIMIT 8 3110 static inline bool dev_xmit_recursion(void) 3111 { 3112 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3113 XMIT_RECURSION_LIMIT); 3114 } 3115 3116 static inline void dev_xmit_recursion_inc(void) 3117 { 3118 __this_cpu_inc(softnet_data.xmit.recursion); 3119 } 3120 3121 static inline void dev_xmit_recursion_dec(void) 3122 { 3123 __this_cpu_dec(softnet_data.xmit.recursion); 3124 } 3125 3126 void __netif_schedule(struct Qdisc *q); 3127 void netif_schedule_queue(struct netdev_queue *txq); 3128 3129 static inline void netif_tx_schedule_all(struct net_device *dev) 3130 { 3131 unsigned int i; 3132 3133 for (i = 0; i < dev->num_tx_queues; i++) 3134 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3135 } 3136 3137 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3138 { 3139 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3140 } 3141 3142 /** 3143 * netif_start_queue - allow transmit 3144 * @dev: network device 3145 * 3146 * Allow upper layers to call the device hard_start_xmit routine. 3147 */ 3148 static inline void netif_start_queue(struct net_device *dev) 3149 { 3150 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3151 } 3152 3153 static inline void netif_tx_start_all_queues(struct net_device *dev) 3154 { 3155 unsigned int i; 3156 3157 for (i = 0; i < dev->num_tx_queues; i++) { 3158 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3159 netif_tx_start_queue(txq); 3160 } 3161 } 3162 3163 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3164 3165 /** 3166 * netif_wake_queue - restart transmit 3167 * @dev: network device 3168 * 3169 * Allow upper layers to call the device hard_start_xmit routine. 3170 * Used for flow control when transmit resources are available. 3171 */ 3172 static inline void netif_wake_queue(struct net_device *dev) 3173 { 3174 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3175 } 3176 3177 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3178 { 3179 unsigned int i; 3180 3181 for (i = 0; i < dev->num_tx_queues; i++) { 3182 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3183 netif_tx_wake_queue(txq); 3184 } 3185 } 3186 3187 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3188 { 3189 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3190 } 3191 3192 /** 3193 * netif_stop_queue - stop transmitted packets 3194 * @dev: network device 3195 * 3196 * Stop upper layers calling the device hard_start_xmit routine. 3197 * Used for flow control when transmit resources are unavailable. 3198 */ 3199 static inline void netif_stop_queue(struct net_device *dev) 3200 { 3201 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3202 } 3203 3204 void netif_tx_stop_all_queues(struct net_device *dev); 3205 3206 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3207 { 3208 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3209 } 3210 3211 /** 3212 * netif_queue_stopped - test if transmit queue is flowblocked 3213 * @dev: network device 3214 * 3215 * Test if transmit queue on device is currently unable to send. 3216 */ 3217 static inline bool netif_queue_stopped(const struct net_device *dev) 3218 { 3219 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3220 } 3221 3222 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3223 { 3224 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3225 } 3226 3227 static inline bool 3228 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3229 { 3230 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3231 } 3232 3233 static inline bool 3234 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3235 { 3236 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3237 } 3238 3239 /** 3240 * netdev_queue_set_dql_min_limit - set dql minimum limit 3241 * @dev_queue: pointer to transmit queue 3242 * @min_limit: dql minimum limit 3243 * 3244 * Forces xmit_more() to return true until the minimum threshold 3245 * defined by @min_limit is reached (or until the tx queue is 3246 * empty). Warning: to be use with care, misuse will impact the 3247 * latency. 3248 */ 3249 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3250 unsigned int min_limit) 3251 { 3252 #ifdef CONFIG_BQL 3253 dev_queue->dql.min_limit = min_limit; 3254 #endif 3255 } 3256 3257 /** 3258 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3259 * @dev_queue: pointer to transmit queue 3260 * 3261 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3262 * to give appropriate hint to the CPU. 3263 */ 3264 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3265 { 3266 #ifdef CONFIG_BQL 3267 prefetchw(&dev_queue->dql.num_queued); 3268 #endif 3269 } 3270 3271 /** 3272 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3273 * @dev_queue: pointer to transmit queue 3274 * 3275 * BQL enabled drivers might use this helper in their TX completion path, 3276 * to give appropriate hint to the CPU. 3277 */ 3278 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3279 { 3280 #ifdef CONFIG_BQL 3281 prefetchw(&dev_queue->dql.limit); 3282 #endif 3283 } 3284 3285 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3286 unsigned int bytes) 3287 { 3288 #ifdef CONFIG_BQL 3289 dql_queued(&dev_queue->dql, bytes); 3290 3291 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3292 return; 3293 3294 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3295 3296 /* 3297 * The XOFF flag must be set before checking the dql_avail below, 3298 * because in netdev_tx_completed_queue we update the dql_completed 3299 * before checking the XOFF flag. 3300 */ 3301 smp_mb(); 3302 3303 /* check again in case another CPU has just made room avail */ 3304 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3305 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3306 #endif 3307 } 3308 3309 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3310 * that they should not test BQL status themselves. 3311 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3312 * skb of a batch. 3313 * Returns true if the doorbell must be used to kick the NIC. 3314 */ 3315 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3316 unsigned int bytes, 3317 bool xmit_more) 3318 { 3319 if (xmit_more) { 3320 #ifdef CONFIG_BQL 3321 dql_queued(&dev_queue->dql, bytes); 3322 #endif 3323 return netif_tx_queue_stopped(dev_queue); 3324 } 3325 netdev_tx_sent_queue(dev_queue, bytes); 3326 return true; 3327 } 3328 3329 /** 3330 * netdev_sent_queue - report the number of bytes queued to hardware 3331 * @dev: network device 3332 * @bytes: number of bytes queued to the hardware device queue 3333 * 3334 * Report the number of bytes queued for sending/completion to the network 3335 * device hardware queue. @bytes should be a good approximation and should 3336 * exactly match netdev_completed_queue() @bytes 3337 */ 3338 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3339 { 3340 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3341 } 3342 3343 static inline bool __netdev_sent_queue(struct net_device *dev, 3344 unsigned int bytes, 3345 bool xmit_more) 3346 { 3347 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3348 xmit_more); 3349 } 3350 3351 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3352 unsigned int pkts, unsigned int bytes) 3353 { 3354 #ifdef CONFIG_BQL 3355 if (unlikely(!bytes)) 3356 return; 3357 3358 dql_completed(&dev_queue->dql, bytes); 3359 3360 /* 3361 * Without the memory barrier there is a small possiblity that 3362 * netdev_tx_sent_queue will miss the update and cause the queue to 3363 * be stopped forever 3364 */ 3365 smp_mb(); 3366 3367 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3368 return; 3369 3370 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3371 netif_schedule_queue(dev_queue); 3372 #endif 3373 } 3374 3375 /** 3376 * netdev_completed_queue - report bytes and packets completed by device 3377 * @dev: network device 3378 * @pkts: actual number of packets sent over the medium 3379 * @bytes: actual number of bytes sent over the medium 3380 * 3381 * Report the number of bytes and packets transmitted by the network device 3382 * hardware queue over the physical medium, @bytes must exactly match the 3383 * @bytes amount passed to netdev_sent_queue() 3384 */ 3385 static inline void netdev_completed_queue(struct net_device *dev, 3386 unsigned int pkts, unsigned int bytes) 3387 { 3388 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3389 } 3390 3391 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3392 { 3393 #ifdef CONFIG_BQL 3394 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3395 dql_reset(&q->dql); 3396 #endif 3397 } 3398 3399 /** 3400 * netdev_reset_queue - reset the packets and bytes count of a network device 3401 * @dev_queue: network device 3402 * 3403 * Reset the bytes and packet count of a network device and clear the 3404 * software flow control OFF bit for this network device 3405 */ 3406 static inline void netdev_reset_queue(struct net_device *dev_queue) 3407 { 3408 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3409 } 3410 3411 /** 3412 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3413 * @dev: network device 3414 * @queue_index: given tx queue index 3415 * 3416 * Returns 0 if given tx queue index >= number of device tx queues, 3417 * otherwise returns the originally passed tx queue index. 3418 */ 3419 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3420 { 3421 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3422 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3423 dev->name, queue_index, 3424 dev->real_num_tx_queues); 3425 return 0; 3426 } 3427 3428 return queue_index; 3429 } 3430 3431 /** 3432 * netif_running - test if up 3433 * @dev: network device 3434 * 3435 * Test if the device has been brought up. 3436 */ 3437 static inline bool netif_running(const struct net_device *dev) 3438 { 3439 return test_bit(__LINK_STATE_START, &dev->state); 3440 } 3441 3442 /* 3443 * Routines to manage the subqueues on a device. We only need start, 3444 * stop, and a check if it's stopped. All other device management is 3445 * done at the overall netdevice level. 3446 * Also test the device if we're multiqueue. 3447 */ 3448 3449 /** 3450 * netif_start_subqueue - allow sending packets on subqueue 3451 * @dev: network device 3452 * @queue_index: sub queue index 3453 * 3454 * Start individual transmit queue of a device with multiple transmit queues. 3455 */ 3456 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3457 { 3458 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3459 3460 netif_tx_start_queue(txq); 3461 } 3462 3463 /** 3464 * netif_stop_subqueue - stop sending packets on subqueue 3465 * @dev: network device 3466 * @queue_index: sub queue index 3467 * 3468 * Stop individual transmit queue of a device with multiple transmit queues. 3469 */ 3470 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3471 { 3472 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3473 netif_tx_stop_queue(txq); 3474 } 3475 3476 /** 3477 * __netif_subqueue_stopped - test status of subqueue 3478 * @dev: network device 3479 * @queue_index: sub queue index 3480 * 3481 * Check individual transmit queue of a device with multiple transmit queues. 3482 */ 3483 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3484 u16 queue_index) 3485 { 3486 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3487 3488 return netif_tx_queue_stopped(txq); 3489 } 3490 3491 /** 3492 * netif_subqueue_stopped - test status of subqueue 3493 * @dev: network device 3494 * @skb: sub queue buffer pointer 3495 * 3496 * Check individual transmit queue of a device with multiple transmit queues. 3497 */ 3498 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3499 struct sk_buff *skb) 3500 { 3501 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3502 } 3503 3504 /** 3505 * netif_wake_subqueue - allow sending packets on subqueue 3506 * @dev: network device 3507 * @queue_index: sub queue index 3508 * 3509 * Resume individual transmit queue of a device with multiple transmit queues. 3510 */ 3511 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3512 { 3513 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3514 3515 netif_tx_wake_queue(txq); 3516 } 3517 3518 #ifdef CONFIG_XPS 3519 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3520 u16 index); 3521 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3522 u16 index, enum xps_map_type type); 3523 3524 /** 3525 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3526 * @j: CPU/Rx queue index 3527 * @mask: bitmask of all cpus/rx queues 3528 * @nr_bits: number of bits in the bitmask 3529 * 3530 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3531 */ 3532 static inline bool netif_attr_test_mask(unsigned long j, 3533 const unsigned long *mask, 3534 unsigned int nr_bits) 3535 { 3536 cpu_max_bits_warn(j, nr_bits); 3537 return test_bit(j, mask); 3538 } 3539 3540 /** 3541 * netif_attr_test_online - Test for online CPU/Rx queue 3542 * @j: CPU/Rx queue index 3543 * @online_mask: bitmask for CPUs/Rx queues that are online 3544 * @nr_bits: number of bits in the bitmask 3545 * 3546 * Returns true if a CPU/Rx queue is online. 3547 */ 3548 static inline bool netif_attr_test_online(unsigned long j, 3549 const unsigned long *online_mask, 3550 unsigned int nr_bits) 3551 { 3552 cpu_max_bits_warn(j, nr_bits); 3553 3554 if (online_mask) 3555 return test_bit(j, online_mask); 3556 3557 return (j < nr_bits); 3558 } 3559 3560 /** 3561 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3562 * @n: CPU/Rx queue index 3563 * @srcp: the cpumask/Rx queue mask pointer 3564 * @nr_bits: number of bits in the bitmask 3565 * 3566 * Returns >= nr_bits if no further CPUs/Rx queues set. 3567 */ 3568 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3569 unsigned int nr_bits) 3570 { 3571 /* -1 is a legal arg here. */ 3572 if (n != -1) 3573 cpu_max_bits_warn(n, nr_bits); 3574 3575 if (srcp) 3576 return find_next_bit(srcp, nr_bits, n + 1); 3577 3578 return n + 1; 3579 } 3580 3581 /** 3582 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3583 * @n: CPU/Rx queue index 3584 * @src1p: the first CPUs/Rx queues mask pointer 3585 * @src2p: the second CPUs/Rx queues mask pointer 3586 * @nr_bits: number of bits in the bitmask 3587 * 3588 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3589 */ 3590 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3591 const unsigned long *src2p, 3592 unsigned int nr_bits) 3593 { 3594 /* -1 is a legal arg here. */ 3595 if (n != -1) 3596 cpu_max_bits_warn(n, nr_bits); 3597 3598 if (src1p && src2p) 3599 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3600 else if (src1p) 3601 return find_next_bit(src1p, nr_bits, n + 1); 3602 else if (src2p) 3603 return find_next_bit(src2p, nr_bits, n + 1); 3604 3605 return n + 1; 3606 } 3607 #else 3608 static inline int netif_set_xps_queue(struct net_device *dev, 3609 const struct cpumask *mask, 3610 u16 index) 3611 { 3612 return 0; 3613 } 3614 3615 static inline int __netif_set_xps_queue(struct net_device *dev, 3616 const unsigned long *mask, 3617 u16 index, enum xps_map_type type) 3618 { 3619 return 0; 3620 } 3621 #endif 3622 3623 /** 3624 * netif_is_multiqueue - test if device has multiple transmit queues 3625 * @dev: network device 3626 * 3627 * Check if device has multiple transmit queues 3628 */ 3629 static inline bool netif_is_multiqueue(const struct net_device *dev) 3630 { 3631 return dev->num_tx_queues > 1; 3632 } 3633 3634 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3635 3636 #ifdef CONFIG_SYSFS 3637 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3638 #else 3639 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3640 unsigned int rxqs) 3641 { 3642 dev->real_num_rx_queues = rxqs; 3643 return 0; 3644 } 3645 #endif 3646 int netif_set_real_num_queues(struct net_device *dev, 3647 unsigned int txq, unsigned int rxq); 3648 3649 static inline struct netdev_rx_queue * 3650 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3651 { 3652 return dev->_rx + rxq; 3653 } 3654 3655 #ifdef CONFIG_SYSFS 3656 static inline unsigned int get_netdev_rx_queue_index( 3657 struct netdev_rx_queue *queue) 3658 { 3659 struct net_device *dev = queue->dev; 3660 int index = queue - dev->_rx; 3661 3662 BUG_ON(index >= dev->num_rx_queues); 3663 return index; 3664 } 3665 #endif 3666 3667 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3668 int netif_get_num_default_rss_queues(void); 3669 3670 enum skb_free_reason { 3671 SKB_REASON_CONSUMED, 3672 SKB_REASON_DROPPED, 3673 }; 3674 3675 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3676 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3677 3678 /* 3679 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3680 * interrupt context or with hardware interrupts being disabled. 3681 * (in_hardirq() || irqs_disabled()) 3682 * 3683 * We provide four helpers that can be used in following contexts : 3684 * 3685 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3686 * replacing kfree_skb(skb) 3687 * 3688 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3689 * Typically used in place of consume_skb(skb) in TX completion path 3690 * 3691 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3692 * replacing kfree_skb(skb) 3693 * 3694 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3695 * and consumed a packet. Used in place of consume_skb(skb) 3696 */ 3697 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3698 { 3699 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3700 } 3701 3702 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3703 { 3704 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3705 } 3706 3707 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3708 { 3709 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3710 } 3711 3712 static inline void dev_consume_skb_any(struct sk_buff *skb) 3713 { 3714 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3715 } 3716 3717 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3718 struct bpf_prog *xdp_prog); 3719 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3720 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3721 int netif_rx(struct sk_buff *skb); 3722 int __netif_rx(struct sk_buff *skb); 3723 3724 int netif_receive_skb(struct sk_buff *skb); 3725 int netif_receive_skb_core(struct sk_buff *skb); 3726 void netif_receive_skb_list_internal(struct list_head *head); 3727 void netif_receive_skb_list(struct list_head *head); 3728 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3729 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3730 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3731 gro_result_t napi_gro_frags(struct napi_struct *napi); 3732 struct packet_offload *gro_find_receive_by_type(__be16 type); 3733 struct packet_offload *gro_find_complete_by_type(__be16 type); 3734 3735 static inline void napi_free_frags(struct napi_struct *napi) 3736 { 3737 kfree_skb(napi->skb); 3738 napi->skb = NULL; 3739 } 3740 3741 bool netdev_is_rx_handler_busy(struct net_device *dev); 3742 int netdev_rx_handler_register(struct net_device *dev, 3743 rx_handler_func_t *rx_handler, 3744 void *rx_handler_data); 3745 void netdev_rx_handler_unregister(struct net_device *dev); 3746 3747 bool dev_valid_name(const char *name); 3748 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 3749 { 3750 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 3751 } 3752 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 3753 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 3754 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3755 void __user *data, bool *need_copyout); 3756 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 3757 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 3758 unsigned int dev_get_flags(const struct net_device *); 3759 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3760 struct netlink_ext_ack *extack); 3761 int dev_change_flags(struct net_device *dev, unsigned int flags, 3762 struct netlink_ext_ack *extack); 3763 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3764 unsigned int gchanges); 3765 int dev_change_name(struct net_device *, const char *); 3766 int dev_set_alias(struct net_device *, const char *, size_t); 3767 int dev_get_alias(const struct net_device *, char *, size_t); 3768 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 3769 const char *pat, int new_ifindex); 3770 static inline 3771 int dev_change_net_namespace(struct net_device *dev, struct net *net, 3772 const char *pat) 3773 { 3774 return __dev_change_net_namespace(dev, net, pat, 0); 3775 } 3776 int __dev_set_mtu(struct net_device *, int); 3777 int dev_validate_mtu(struct net_device *dev, int mtu, 3778 struct netlink_ext_ack *extack); 3779 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3780 struct netlink_ext_ack *extack); 3781 int dev_set_mtu(struct net_device *, int); 3782 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3783 void dev_set_group(struct net_device *, int); 3784 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3785 struct netlink_ext_ack *extack); 3786 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3787 struct netlink_ext_ack *extack); 3788 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 3789 struct netlink_ext_ack *extack); 3790 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 3791 int dev_change_carrier(struct net_device *, bool new_carrier); 3792 int dev_get_phys_port_id(struct net_device *dev, 3793 struct netdev_phys_item_id *ppid); 3794 int dev_get_phys_port_name(struct net_device *dev, 3795 char *name, size_t len); 3796 int dev_get_port_parent_id(struct net_device *dev, 3797 struct netdev_phys_item_id *ppid, bool recurse); 3798 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3799 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3800 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask, 3801 u32 value); 3802 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3803 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3804 struct netdev_queue *txq, int *ret); 3805 3806 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3807 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3808 int fd, int expected_fd, u32 flags); 3809 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3810 u8 dev_xdp_prog_count(struct net_device *dev); 3811 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3812 3813 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3814 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3815 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 3816 bool is_skb_forwardable(const struct net_device *dev, 3817 const struct sk_buff *skb); 3818 3819 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 3820 const struct sk_buff *skb, 3821 const bool check_mtu) 3822 { 3823 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 3824 unsigned int len; 3825 3826 if (!(dev->flags & IFF_UP)) 3827 return false; 3828 3829 if (!check_mtu) 3830 return true; 3831 3832 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 3833 if (skb->len <= len) 3834 return true; 3835 3836 /* if TSO is enabled, we don't care about the length as the packet 3837 * could be forwarded without being segmented before 3838 */ 3839 if (skb_is_gso(skb)) 3840 return true; 3841 3842 return false; 3843 } 3844 3845 struct net_device_core_stats *netdev_core_stats_alloc(struct net_device *dev); 3846 3847 static inline struct net_device_core_stats *dev_core_stats(struct net_device *dev) 3848 { 3849 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */ 3850 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats); 3851 3852 if (likely(p)) 3853 return this_cpu_ptr(p); 3854 3855 return netdev_core_stats_alloc(dev); 3856 } 3857 3858 #define DEV_CORE_STATS_INC(FIELD) \ 3859 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 3860 { \ 3861 struct net_device_core_stats *p; \ 3862 \ 3863 preempt_disable(); \ 3864 p = dev_core_stats(dev); \ 3865 \ 3866 if (p) \ 3867 local_inc(&p->FIELD); \ 3868 preempt_enable(); \ 3869 } 3870 DEV_CORE_STATS_INC(rx_dropped) 3871 DEV_CORE_STATS_INC(tx_dropped) 3872 DEV_CORE_STATS_INC(rx_nohandler) 3873 3874 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3875 struct sk_buff *skb, 3876 const bool check_mtu) 3877 { 3878 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3879 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 3880 dev_core_stats_rx_dropped_inc(dev); 3881 kfree_skb(skb); 3882 return NET_RX_DROP; 3883 } 3884 3885 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 3886 skb->priority = 0; 3887 return 0; 3888 } 3889 3890 bool dev_nit_active(struct net_device *dev); 3891 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3892 3893 extern int netdev_budget; 3894 extern unsigned int netdev_budget_usecs; 3895 3896 /* Called by rtnetlink.c:rtnl_unlock() */ 3897 void netdev_run_todo(void); 3898 3899 static inline void __dev_put(struct net_device *dev) 3900 { 3901 if (dev) { 3902 #ifdef CONFIG_PCPU_DEV_REFCNT 3903 this_cpu_dec(*dev->pcpu_refcnt); 3904 #else 3905 refcount_dec(&dev->dev_refcnt); 3906 #endif 3907 } 3908 } 3909 3910 static inline void __dev_hold(struct net_device *dev) 3911 { 3912 if (dev) { 3913 #ifdef CONFIG_PCPU_DEV_REFCNT 3914 this_cpu_inc(*dev->pcpu_refcnt); 3915 #else 3916 refcount_inc(&dev->dev_refcnt); 3917 #endif 3918 } 3919 } 3920 3921 static inline void __netdev_tracker_alloc(struct net_device *dev, 3922 netdevice_tracker *tracker, 3923 gfp_t gfp) 3924 { 3925 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 3926 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 3927 #endif 3928 } 3929 3930 /* netdev_tracker_alloc() can upgrade a prior untracked reference 3931 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 3932 */ 3933 static inline void netdev_tracker_alloc(struct net_device *dev, 3934 netdevice_tracker *tracker, gfp_t gfp) 3935 { 3936 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 3937 refcount_dec(&dev->refcnt_tracker.no_tracker); 3938 __netdev_tracker_alloc(dev, tracker, gfp); 3939 #endif 3940 } 3941 3942 static inline void netdev_tracker_free(struct net_device *dev, 3943 netdevice_tracker *tracker) 3944 { 3945 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 3946 ref_tracker_free(&dev->refcnt_tracker, tracker); 3947 #endif 3948 } 3949 3950 static inline void dev_hold_track(struct net_device *dev, 3951 netdevice_tracker *tracker, gfp_t gfp) 3952 { 3953 if (dev) { 3954 __dev_hold(dev); 3955 __netdev_tracker_alloc(dev, tracker, gfp); 3956 } 3957 } 3958 3959 static inline void dev_put_track(struct net_device *dev, 3960 netdevice_tracker *tracker) 3961 { 3962 if (dev) { 3963 netdev_tracker_free(dev, tracker); 3964 __dev_put(dev); 3965 } 3966 } 3967 3968 /** 3969 * dev_hold - get reference to device 3970 * @dev: network device 3971 * 3972 * Hold reference to device to keep it from being freed. 3973 * Try using dev_hold_track() instead. 3974 */ 3975 static inline void dev_hold(struct net_device *dev) 3976 { 3977 dev_hold_track(dev, NULL, GFP_ATOMIC); 3978 } 3979 3980 /** 3981 * dev_put - release reference to device 3982 * @dev: network device 3983 * 3984 * Release reference to device to allow it to be freed. 3985 * Try using dev_put_track() instead. 3986 */ 3987 static inline void dev_put(struct net_device *dev) 3988 { 3989 dev_put_track(dev, NULL); 3990 } 3991 3992 static inline void dev_replace_track(struct net_device *odev, 3993 struct net_device *ndev, 3994 netdevice_tracker *tracker, 3995 gfp_t gfp) 3996 { 3997 if (odev) 3998 netdev_tracker_free(odev, tracker); 3999 4000 __dev_hold(ndev); 4001 __dev_put(odev); 4002 4003 if (ndev) 4004 __netdev_tracker_alloc(ndev, tracker, gfp); 4005 } 4006 4007 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4008 * and _off may be called from IRQ context, but it is caller 4009 * who is responsible for serialization of these calls. 4010 * 4011 * The name carrier is inappropriate, these functions should really be 4012 * called netif_lowerlayer_*() because they represent the state of any 4013 * kind of lower layer not just hardware media. 4014 */ 4015 4016 void linkwatch_init_dev(struct net_device *dev); 4017 void linkwatch_fire_event(struct net_device *dev); 4018 void linkwatch_forget_dev(struct net_device *dev); 4019 4020 /** 4021 * netif_carrier_ok - test if carrier present 4022 * @dev: network device 4023 * 4024 * Check if carrier is present on device 4025 */ 4026 static inline bool netif_carrier_ok(const struct net_device *dev) 4027 { 4028 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4029 } 4030 4031 unsigned long dev_trans_start(struct net_device *dev); 4032 4033 void __netdev_watchdog_up(struct net_device *dev); 4034 4035 void netif_carrier_on(struct net_device *dev); 4036 void netif_carrier_off(struct net_device *dev); 4037 void netif_carrier_event(struct net_device *dev); 4038 4039 /** 4040 * netif_dormant_on - mark device as dormant. 4041 * @dev: network device 4042 * 4043 * Mark device as dormant (as per RFC2863). 4044 * 4045 * The dormant state indicates that the relevant interface is not 4046 * actually in a condition to pass packets (i.e., it is not 'up') but is 4047 * in a "pending" state, waiting for some external event. For "on- 4048 * demand" interfaces, this new state identifies the situation where the 4049 * interface is waiting for events to place it in the up state. 4050 */ 4051 static inline void netif_dormant_on(struct net_device *dev) 4052 { 4053 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4054 linkwatch_fire_event(dev); 4055 } 4056 4057 /** 4058 * netif_dormant_off - set device as not dormant. 4059 * @dev: network device 4060 * 4061 * Device is not in dormant state. 4062 */ 4063 static inline void netif_dormant_off(struct net_device *dev) 4064 { 4065 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4066 linkwatch_fire_event(dev); 4067 } 4068 4069 /** 4070 * netif_dormant - test if device is dormant 4071 * @dev: network device 4072 * 4073 * Check if device is dormant. 4074 */ 4075 static inline bool netif_dormant(const struct net_device *dev) 4076 { 4077 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4078 } 4079 4080 4081 /** 4082 * netif_testing_on - mark device as under test. 4083 * @dev: network device 4084 * 4085 * Mark device as under test (as per RFC2863). 4086 * 4087 * The testing state indicates that some test(s) must be performed on 4088 * the interface. After completion, of the test, the interface state 4089 * will change to up, dormant, or down, as appropriate. 4090 */ 4091 static inline void netif_testing_on(struct net_device *dev) 4092 { 4093 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4094 linkwatch_fire_event(dev); 4095 } 4096 4097 /** 4098 * netif_testing_off - set device as not under test. 4099 * @dev: network device 4100 * 4101 * Device is not in testing state. 4102 */ 4103 static inline void netif_testing_off(struct net_device *dev) 4104 { 4105 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4106 linkwatch_fire_event(dev); 4107 } 4108 4109 /** 4110 * netif_testing - test if device is under test 4111 * @dev: network device 4112 * 4113 * Check if device is under test 4114 */ 4115 static inline bool netif_testing(const struct net_device *dev) 4116 { 4117 return test_bit(__LINK_STATE_TESTING, &dev->state); 4118 } 4119 4120 4121 /** 4122 * netif_oper_up - test if device is operational 4123 * @dev: network device 4124 * 4125 * Check if carrier is operational 4126 */ 4127 static inline bool netif_oper_up(const struct net_device *dev) 4128 { 4129 return (dev->operstate == IF_OPER_UP || 4130 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4131 } 4132 4133 /** 4134 * netif_device_present - is device available or removed 4135 * @dev: network device 4136 * 4137 * Check if device has not been removed from system. 4138 */ 4139 static inline bool netif_device_present(const struct net_device *dev) 4140 { 4141 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4142 } 4143 4144 void netif_device_detach(struct net_device *dev); 4145 4146 void netif_device_attach(struct net_device *dev); 4147 4148 /* 4149 * Network interface message level settings 4150 */ 4151 4152 enum { 4153 NETIF_MSG_DRV_BIT, 4154 NETIF_MSG_PROBE_BIT, 4155 NETIF_MSG_LINK_BIT, 4156 NETIF_MSG_TIMER_BIT, 4157 NETIF_MSG_IFDOWN_BIT, 4158 NETIF_MSG_IFUP_BIT, 4159 NETIF_MSG_RX_ERR_BIT, 4160 NETIF_MSG_TX_ERR_BIT, 4161 NETIF_MSG_TX_QUEUED_BIT, 4162 NETIF_MSG_INTR_BIT, 4163 NETIF_MSG_TX_DONE_BIT, 4164 NETIF_MSG_RX_STATUS_BIT, 4165 NETIF_MSG_PKTDATA_BIT, 4166 NETIF_MSG_HW_BIT, 4167 NETIF_MSG_WOL_BIT, 4168 4169 /* When you add a new bit above, update netif_msg_class_names array 4170 * in net/ethtool/common.c 4171 */ 4172 NETIF_MSG_CLASS_COUNT, 4173 }; 4174 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4175 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4176 4177 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4178 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4179 4180 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4181 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4182 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4183 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4184 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4185 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4186 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4187 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4188 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4189 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4190 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4191 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4192 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4193 #define NETIF_MSG_HW __NETIF_MSG(HW) 4194 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4195 4196 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4197 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4198 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4199 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4200 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4201 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4202 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4203 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4204 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4205 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4206 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4207 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4208 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4209 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4210 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4211 4212 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4213 { 4214 /* use default */ 4215 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4216 return default_msg_enable_bits; 4217 if (debug_value == 0) /* no output */ 4218 return 0; 4219 /* set low N bits */ 4220 return (1U << debug_value) - 1; 4221 } 4222 4223 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4224 { 4225 spin_lock(&txq->_xmit_lock); 4226 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4227 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4228 } 4229 4230 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4231 { 4232 __acquire(&txq->_xmit_lock); 4233 return true; 4234 } 4235 4236 static inline void __netif_tx_release(struct netdev_queue *txq) 4237 { 4238 __release(&txq->_xmit_lock); 4239 } 4240 4241 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4242 { 4243 spin_lock_bh(&txq->_xmit_lock); 4244 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4245 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4246 } 4247 4248 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4249 { 4250 bool ok = spin_trylock(&txq->_xmit_lock); 4251 4252 if (likely(ok)) { 4253 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4254 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4255 } 4256 return ok; 4257 } 4258 4259 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4260 { 4261 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4262 WRITE_ONCE(txq->xmit_lock_owner, -1); 4263 spin_unlock(&txq->_xmit_lock); 4264 } 4265 4266 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4267 { 4268 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4269 WRITE_ONCE(txq->xmit_lock_owner, -1); 4270 spin_unlock_bh(&txq->_xmit_lock); 4271 } 4272 4273 /* 4274 * txq->trans_start can be read locklessly from dev_watchdog() 4275 */ 4276 static inline void txq_trans_update(struct netdev_queue *txq) 4277 { 4278 if (txq->xmit_lock_owner != -1) 4279 WRITE_ONCE(txq->trans_start, jiffies); 4280 } 4281 4282 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4283 { 4284 unsigned long now = jiffies; 4285 4286 if (READ_ONCE(txq->trans_start) != now) 4287 WRITE_ONCE(txq->trans_start, now); 4288 } 4289 4290 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4291 static inline void netif_trans_update(struct net_device *dev) 4292 { 4293 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4294 4295 txq_trans_cond_update(txq); 4296 } 4297 4298 /** 4299 * netif_tx_lock - grab network device transmit lock 4300 * @dev: network device 4301 * 4302 * Get network device transmit lock 4303 */ 4304 void netif_tx_lock(struct net_device *dev); 4305 4306 static inline void netif_tx_lock_bh(struct net_device *dev) 4307 { 4308 local_bh_disable(); 4309 netif_tx_lock(dev); 4310 } 4311 4312 void netif_tx_unlock(struct net_device *dev); 4313 4314 static inline void netif_tx_unlock_bh(struct net_device *dev) 4315 { 4316 netif_tx_unlock(dev); 4317 local_bh_enable(); 4318 } 4319 4320 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4321 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4322 __netif_tx_lock(txq, cpu); \ 4323 } else { \ 4324 __netif_tx_acquire(txq); \ 4325 } \ 4326 } 4327 4328 #define HARD_TX_TRYLOCK(dev, txq) \ 4329 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4330 __netif_tx_trylock(txq) : \ 4331 __netif_tx_acquire(txq)) 4332 4333 #define HARD_TX_UNLOCK(dev, txq) { \ 4334 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4335 __netif_tx_unlock(txq); \ 4336 } else { \ 4337 __netif_tx_release(txq); \ 4338 } \ 4339 } 4340 4341 static inline void netif_tx_disable(struct net_device *dev) 4342 { 4343 unsigned int i; 4344 int cpu; 4345 4346 local_bh_disable(); 4347 cpu = smp_processor_id(); 4348 spin_lock(&dev->tx_global_lock); 4349 for (i = 0; i < dev->num_tx_queues; i++) { 4350 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4351 4352 __netif_tx_lock(txq, cpu); 4353 netif_tx_stop_queue(txq); 4354 __netif_tx_unlock(txq); 4355 } 4356 spin_unlock(&dev->tx_global_lock); 4357 local_bh_enable(); 4358 } 4359 4360 static inline void netif_addr_lock(struct net_device *dev) 4361 { 4362 unsigned char nest_level = 0; 4363 4364 #ifdef CONFIG_LOCKDEP 4365 nest_level = dev->nested_level; 4366 #endif 4367 spin_lock_nested(&dev->addr_list_lock, nest_level); 4368 } 4369 4370 static inline void netif_addr_lock_bh(struct net_device *dev) 4371 { 4372 unsigned char nest_level = 0; 4373 4374 #ifdef CONFIG_LOCKDEP 4375 nest_level = dev->nested_level; 4376 #endif 4377 local_bh_disable(); 4378 spin_lock_nested(&dev->addr_list_lock, nest_level); 4379 } 4380 4381 static inline void netif_addr_unlock(struct net_device *dev) 4382 { 4383 spin_unlock(&dev->addr_list_lock); 4384 } 4385 4386 static inline void netif_addr_unlock_bh(struct net_device *dev) 4387 { 4388 spin_unlock_bh(&dev->addr_list_lock); 4389 } 4390 4391 /* 4392 * dev_addrs walker. Should be used only for read access. Call with 4393 * rcu_read_lock held. 4394 */ 4395 #define for_each_dev_addr(dev, ha) \ 4396 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4397 4398 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4399 4400 void ether_setup(struct net_device *dev); 4401 4402 /* Support for loadable net-drivers */ 4403 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4404 unsigned char name_assign_type, 4405 void (*setup)(struct net_device *), 4406 unsigned int txqs, unsigned int rxqs); 4407 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4408 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4409 4410 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4411 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4412 count) 4413 4414 int register_netdev(struct net_device *dev); 4415 void unregister_netdev(struct net_device *dev); 4416 4417 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4418 4419 /* General hardware address lists handling functions */ 4420 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4421 struct netdev_hw_addr_list *from_list, int addr_len); 4422 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4423 struct netdev_hw_addr_list *from_list, int addr_len); 4424 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4425 struct net_device *dev, 4426 int (*sync)(struct net_device *, const unsigned char *), 4427 int (*unsync)(struct net_device *, 4428 const unsigned char *)); 4429 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4430 struct net_device *dev, 4431 int (*sync)(struct net_device *, 4432 const unsigned char *, int), 4433 int (*unsync)(struct net_device *, 4434 const unsigned char *, int)); 4435 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4436 struct net_device *dev, 4437 int (*unsync)(struct net_device *, 4438 const unsigned char *, int)); 4439 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4440 struct net_device *dev, 4441 int (*unsync)(struct net_device *, 4442 const unsigned char *)); 4443 void __hw_addr_init(struct netdev_hw_addr_list *list); 4444 4445 /* Functions used for device addresses handling */ 4446 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4447 const void *addr, size_t len); 4448 4449 static inline void 4450 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4451 { 4452 dev_addr_mod(dev, 0, addr, len); 4453 } 4454 4455 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4456 { 4457 __dev_addr_set(dev, addr, dev->addr_len); 4458 } 4459 4460 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4461 unsigned char addr_type); 4462 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4463 unsigned char addr_type); 4464 void dev_addr_flush(struct net_device *dev); 4465 int dev_addr_init(struct net_device *dev); 4466 void dev_addr_check(struct net_device *dev); 4467 4468 /* Functions used for unicast addresses handling */ 4469 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4470 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4471 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4472 int dev_uc_sync(struct net_device *to, struct net_device *from); 4473 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4474 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4475 void dev_uc_flush(struct net_device *dev); 4476 void dev_uc_init(struct net_device *dev); 4477 4478 /** 4479 * __dev_uc_sync - Synchonize device's unicast list 4480 * @dev: device to sync 4481 * @sync: function to call if address should be added 4482 * @unsync: function to call if address should be removed 4483 * 4484 * Add newly added addresses to the interface, and release 4485 * addresses that have been deleted. 4486 */ 4487 static inline int __dev_uc_sync(struct net_device *dev, 4488 int (*sync)(struct net_device *, 4489 const unsigned char *), 4490 int (*unsync)(struct net_device *, 4491 const unsigned char *)) 4492 { 4493 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4494 } 4495 4496 /** 4497 * __dev_uc_unsync - Remove synchronized addresses from device 4498 * @dev: device to sync 4499 * @unsync: function to call if address should be removed 4500 * 4501 * Remove all addresses that were added to the device by dev_uc_sync(). 4502 */ 4503 static inline void __dev_uc_unsync(struct net_device *dev, 4504 int (*unsync)(struct net_device *, 4505 const unsigned char *)) 4506 { 4507 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4508 } 4509 4510 /* Functions used for multicast addresses handling */ 4511 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4512 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4513 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4514 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4515 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4516 int dev_mc_sync(struct net_device *to, struct net_device *from); 4517 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4518 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4519 void dev_mc_flush(struct net_device *dev); 4520 void dev_mc_init(struct net_device *dev); 4521 4522 /** 4523 * __dev_mc_sync - Synchonize device's multicast list 4524 * @dev: device to sync 4525 * @sync: function to call if address should be added 4526 * @unsync: function to call if address should be removed 4527 * 4528 * Add newly added addresses to the interface, and release 4529 * addresses that have been deleted. 4530 */ 4531 static inline int __dev_mc_sync(struct net_device *dev, 4532 int (*sync)(struct net_device *, 4533 const unsigned char *), 4534 int (*unsync)(struct net_device *, 4535 const unsigned char *)) 4536 { 4537 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4538 } 4539 4540 /** 4541 * __dev_mc_unsync - Remove synchronized addresses from device 4542 * @dev: device to sync 4543 * @unsync: function to call if address should be removed 4544 * 4545 * Remove all addresses that were added to the device by dev_mc_sync(). 4546 */ 4547 static inline void __dev_mc_unsync(struct net_device *dev, 4548 int (*unsync)(struct net_device *, 4549 const unsigned char *)) 4550 { 4551 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4552 } 4553 4554 /* Functions used for secondary unicast and multicast support */ 4555 void dev_set_rx_mode(struct net_device *dev); 4556 void __dev_set_rx_mode(struct net_device *dev); 4557 int dev_set_promiscuity(struct net_device *dev, int inc); 4558 int dev_set_allmulti(struct net_device *dev, int inc); 4559 void netdev_state_change(struct net_device *dev); 4560 void __netdev_notify_peers(struct net_device *dev); 4561 void netdev_notify_peers(struct net_device *dev); 4562 void netdev_features_change(struct net_device *dev); 4563 /* Load a device via the kmod */ 4564 void dev_load(struct net *net, const char *name); 4565 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4566 struct rtnl_link_stats64 *storage); 4567 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4568 const struct net_device_stats *netdev_stats); 4569 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4570 const struct pcpu_sw_netstats __percpu *netstats); 4571 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4572 4573 extern int netdev_max_backlog; 4574 extern int netdev_tstamp_prequeue; 4575 extern int netdev_unregister_timeout_secs; 4576 extern int weight_p; 4577 extern int dev_weight_rx_bias; 4578 extern int dev_weight_tx_bias; 4579 extern int dev_rx_weight; 4580 extern int dev_tx_weight; 4581 extern int gro_normal_batch; 4582 4583 enum { 4584 NESTED_SYNC_IMM_BIT, 4585 NESTED_SYNC_TODO_BIT, 4586 }; 4587 4588 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4589 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4590 4591 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4592 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4593 4594 struct netdev_nested_priv { 4595 unsigned char flags; 4596 void *data; 4597 }; 4598 4599 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4600 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4601 struct list_head **iter); 4602 4603 #ifdef CONFIG_LOCKDEP 4604 static LIST_HEAD(net_unlink_list); 4605 4606 static inline void net_unlink_todo(struct net_device *dev) 4607 { 4608 if (list_empty(&dev->unlink_list)) 4609 list_add_tail(&dev->unlink_list, &net_unlink_list); 4610 } 4611 #endif 4612 4613 /* iterate through upper list, must be called under RCU read lock */ 4614 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4615 for (iter = &(dev)->adj_list.upper, \ 4616 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4617 updev; \ 4618 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4619 4620 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4621 int (*fn)(struct net_device *upper_dev, 4622 struct netdev_nested_priv *priv), 4623 struct netdev_nested_priv *priv); 4624 4625 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4626 struct net_device *upper_dev); 4627 4628 bool netdev_has_any_upper_dev(struct net_device *dev); 4629 4630 void *netdev_lower_get_next_private(struct net_device *dev, 4631 struct list_head **iter); 4632 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4633 struct list_head **iter); 4634 4635 #define netdev_for_each_lower_private(dev, priv, iter) \ 4636 for (iter = (dev)->adj_list.lower.next, \ 4637 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4638 priv; \ 4639 priv = netdev_lower_get_next_private(dev, &(iter))) 4640 4641 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4642 for (iter = &(dev)->adj_list.lower, \ 4643 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4644 priv; \ 4645 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4646 4647 void *netdev_lower_get_next(struct net_device *dev, 4648 struct list_head **iter); 4649 4650 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4651 for (iter = (dev)->adj_list.lower.next, \ 4652 ldev = netdev_lower_get_next(dev, &(iter)); \ 4653 ldev; \ 4654 ldev = netdev_lower_get_next(dev, &(iter))) 4655 4656 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4657 struct list_head **iter); 4658 int netdev_walk_all_lower_dev(struct net_device *dev, 4659 int (*fn)(struct net_device *lower_dev, 4660 struct netdev_nested_priv *priv), 4661 struct netdev_nested_priv *priv); 4662 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4663 int (*fn)(struct net_device *lower_dev, 4664 struct netdev_nested_priv *priv), 4665 struct netdev_nested_priv *priv); 4666 4667 void *netdev_adjacent_get_private(struct list_head *adj_list); 4668 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4669 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4670 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4671 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4672 struct netlink_ext_ack *extack); 4673 int netdev_master_upper_dev_link(struct net_device *dev, 4674 struct net_device *upper_dev, 4675 void *upper_priv, void *upper_info, 4676 struct netlink_ext_ack *extack); 4677 void netdev_upper_dev_unlink(struct net_device *dev, 4678 struct net_device *upper_dev); 4679 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4680 struct net_device *new_dev, 4681 struct net_device *dev, 4682 struct netlink_ext_ack *extack); 4683 void netdev_adjacent_change_commit(struct net_device *old_dev, 4684 struct net_device *new_dev, 4685 struct net_device *dev); 4686 void netdev_adjacent_change_abort(struct net_device *old_dev, 4687 struct net_device *new_dev, 4688 struct net_device *dev); 4689 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4690 void *netdev_lower_dev_get_private(struct net_device *dev, 4691 struct net_device *lower_dev); 4692 void netdev_lower_state_changed(struct net_device *lower_dev, 4693 void *lower_state_info); 4694 4695 /* RSS keys are 40 or 52 bytes long */ 4696 #define NETDEV_RSS_KEY_LEN 52 4697 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4698 void netdev_rss_key_fill(void *buffer, size_t len); 4699 4700 int skb_checksum_help(struct sk_buff *skb); 4701 int skb_crc32c_csum_help(struct sk_buff *skb); 4702 int skb_csum_hwoffload_help(struct sk_buff *skb, 4703 const netdev_features_t features); 4704 4705 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4706 netdev_features_t features, bool tx_path); 4707 struct sk_buff *skb_eth_gso_segment(struct sk_buff *skb, 4708 netdev_features_t features, __be16 type); 4709 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4710 netdev_features_t features); 4711 4712 struct netdev_bonding_info { 4713 ifslave slave; 4714 ifbond master; 4715 }; 4716 4717 struct netdev_notifier_bonding_info { 4718 struct netdev_notifier_info info; /* must be first */ 4719 struct netdev_bonding_info bonding_info; 4720 }; 4721 4722 void netdev_bonding_info_change(struct net_device *dev, 4723 struct netdev_bonding_info *bonding_info); 4724 4725 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4726 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4727 #else 4728 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4729 const void *data) 4730 { 4731 } 4732 #endif 4733 4734 static inline 4735 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4736 { 4737 return __skb_gso_segment(skb, features, true); 4738 } 4739 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4740 4741 static inline bool can_checksum_protocol(netdev_features_t features, 4742 __be16 protocol) 4743 { 4744 if (protocol == htons(ETH_P_FCOE)) 4745 return !!(features & NETIF_F_FCOE_CRC); 4746 4747 /* Assume this is an IP checksum (not SCTP CRC) */ 4748 4749 if (features & NETIF_F_HW_CSUM) { 4750 /* Can checksum everything */ 4751 return true; 4752 } 4753 4754 switch (protocol) { 4755 case htons(ETH_P_IP): 4756 return !!(features & NETIF_F_IP_CSUM); 4757 case htons(ETH_P_IPV6): 4758 return !!(features & NETIF_F_IPV6_CSUM); 4759 default: 4760 return false; 4761 } 4762 } 4763 4764 #ifdef CONFIG_BUG 4765 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4766 #else 4767 static inline void netdev_rx_csum_fault(struct net_device *dev, 4768 struct sk_buff *skb) 4769 { 4770 } 4771 #endif 4772 /* rx skb timestamps */ 4773 void net_enable_timestamp(void); 4774 void net_disable_timestamp(void); 4775 4776 #ifdef CONFIG_PROC_FS 4777 int __init dev_proc_init(void); 4778 #else 4779 #define dev_proc_init() 0 4780 #endif 4781 4782 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4783 struct sk_buff *skb, struct net_device *dev, 4784 bool more) 4785 { 4786 __this_cpu_write(softnet_data.xmit.more, more); 4787 return ops->ndo_start_xmit(skb, dev); 4788 } 4789 4790 static inline bool netdev_xmit_more(void) 4791 { 4792 return __this_cpu_read(softnet_data.xmit.more); 4793 } 4794 4795 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4796 struct netdev_queue *txq, bool more) 4797 { 4798 const struct net_device_ops *ops = dev->netdev_ops; 4799 netdev_tx_t rc; 4800 4801 rc = __netdev_start_xmit(ops, skb, dev, more); 4802 if (rc == NETDEV_TX_OK) 4803 txq_trans_update(txq); 4804 4805 return rc; 4806 } 4807 4808 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4809 const void *ns); 4810 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4811 const void *ns); 4812 4813 extern const struct kobj_ns_type_operations net_ns_type_operations; 4814 4815 const char *netdev_drivername(const struct net_device *dev); 4816 4817 void linkwatch_run_queue(void); 4818 4819 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4820 netdev_features_t f2) 4821 { 4822 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4823 if (f1 & NETIF_F_HW_CSUM) 4824 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4825 else 4826 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4827 } 4828 4829 return f1 & f2; 4830 } 4831 4832 static inline netdev_features_t netdev_get_wanted_features( 4833 struct net_device *dev) 4834 { 4835 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4836 } 4837 netdev_features_t netdev_increment_features(netdev_features_t all, 4838 netdev_features_t one, netdev_features_t mask); 4839 4840 /* Allow TSO being used on stacked device : 4841 * Performing the GSO segmentation before last device 4842 * is a performance improvement. 4843 */ 4844 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4845 netdev_features_t mask) 4846 { 4847 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4848 } 4849 4850 int __netdev_update_features(struct net_device *dev); 4851 void netdev_update_features(struct net_device *dev); 4852 void netdev_change_features(struct net_device *dev); 4853 4854 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4855 struct net_device *dev); 4856 4857 netdev_features_t passthru_features_check(struct sk_buff *skb, 4858 struct net_device *dev, 4859 netdev_features_t features); 4860 netdev_features_t netif_skb_features(struct sk_buff *skb); 4861 4862 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4863 { 4864 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4865 4866 /* check flags correspondence */ 4867 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4868 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4869 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4870 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4871 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4872 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4873 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4874 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4875 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4876 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4877 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4878 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4879 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4880 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4881 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4882 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4883 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4884 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4885 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4886 4887 return (features & feature) == feature; 4888 } 4889 4890 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4891 { 4892 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4893 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4894 } 4895 4896 static inline bool netif_needs_gso(struct sk_buff *skb, 4897 netdev_features_t features) 4898 { 4899 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4900 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4901 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4902 } 4903 4904 static inline void netif_set_gso_max_size(struct net_device *dev, 4905 unsigned int size) 4906 { 4907 /* dev->gso_max_size is read locklessly from sk_setup_caps() */ 4908 WRITE_ONCE(dev->gso_max_size, size); 4909 } 4910 4911 static inline void netif_set_gso_max_segs(struct net_device *dev, 4912 unsigned int segs) 4913 { 4914 /* dev->gso_max_segs is read locklessly from sk_setup_caps() */ 4915 WRITE_ONCE(dev->gso_max_segs, segs); 4916 } 4917 4918 static inline void netif_set_gro_max_size(struct net_device *dev, 4919 unsigned int size) 4920 { 4921 /* This pairs with the READ_ONCE() in skb_gro_receive() */ 4922 WRITE_ONCE(dev->gro_max_size, size); 4923 } 4924 4925 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4926 int pulled_hlen, u16 mac_offset, 4927 int mac_len) 4928 { 4929 skb->protocol = protocol; 4930 skb->encapsulation = 1; 4931 skb_push(skb, pulled_hlen); 4932 skb_reset_transport_header(skb); 4933 skb->mac_header = mac_offset; 4934 skb->network_header = skb->mac_header + mac_len; 4935 skb->mac_len = mac_len; 4936 } 4937 4938 static inline bool netif_is_macsec(const struct net_device *dev) 4939 { 4940 return dev->priv_flags & IFF_MACSEC; 4941 } 4942 4943 static inline bool netif_is_macvlan(const struct net_device *dev) 4944 { 4945 return dev->priv_flags & IFF_MACVLAN; 4946 } 4947 4948 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4949 { 4950 return dev->priv_flags & IFF_MACVLAN_PORT; 4951 } 4952 4953 static inline bool netif_is_bond_master(const struct net_device *dev) 4954 { 4955 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4956 } 4957 4958 static inline bool netif_is_bond_slave(const struct net_device *dev) 4959 { 4960 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4961 } 4962 4963 static inline bool netif_supports_nofcs(struct net_device *dev) 4964 { 4965 return dev->priv_flags & IFF_SUPP_NOFCS; 4966 } 4967 4968 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4969 { 4970 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4971 } 4972 4973 static inline bool netif_is_l3_master(const struct net_device *dev) 4974 { 4975 return dev->priv_flags & IFF_L3MDEV_MASTER; 4976 } 4977 4978 static inline bool netif_is_l3_slave(const struct net_device *dev) 4979 { 4980 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4981 } 4982 4983 static inline bool netif_is_bridge_master(const struct net_device *dev) 4984 { 4985 return dev->priv_flags & IFF_EBRIDGE; 4986 } 4987 4988 static inline bool netif_is_bridge_port(const struct net_device *dev) 4989 { 4990 return dev->priv_flags & IFF_BRIDGE_PORT; 4991 } 4992 4993 static inline bool netif_is_ovs_master(const struct net_device *dev) 4994 { 4995 return dev->priv_flags & IFF_OPENVSWITCH; 4996 } 4997 4998 static inline bool netif_is_ovs_port(const struct net_device *dev) 4999 { 5000 return dev->priv_flags & IFF_OVS_DATAPATH; 5001 } 5002 5003 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5004 { 5005 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5006 } 5007 5008 static inline bool netif_is_team_master(const struct net_device *dev) 5009 { 5010 return dev->priv_flags & IFF_TEAM; 5011 } 5012 5013 static inline bool netif_is_team_port(const struct net_device *dev) 5014 { 5015 return dev->priv_flags & IFF_TEAM_PORT; 5016 } 5017 5018 static inline bool netif_is_lag_master(const struct net_device *dev) 5019 { 5020 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5021 } 5022 5023 static inline bool netif_is_lag_port(const struct net_device *dev) 5024 { 5025 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5026 } 5027 5028 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5029 { 5030 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5031 } 5032 5033 static inline bool netif_is_failover(const struct net_device *dev) 5034 { 5035 return dev->priv_flags & IFF_FAILOVER; 5036 } 5037 5038 static inline bool netif_is_failover_slave(const struct net_device *dev) 5039 { 5040 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5041 } 5042 5043 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5044 static inline void netif_keep_dst(struct net_device *dev) 5045 { 5046 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5047 } 5048 5049 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5050 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5051 { 5052 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5053 return netif_is_macsec(dev); 5054 } 5055 5056 extern struct pernet_operations __net_initdata loopback_net_ops; 5057 5058 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5059 5060 /* netdev_printk helpers, similar to dev_printk */ 5061 5062 static inline const char *netdev_name(const struct net_device *dev) 5063 { 5064 if (!dev->name[0] || strchr(dev->name, '%')) 5065 return "(unnamed net_device)"; 5066 return dev->name; 5067 } 5068 5069 static inline bool netdev_unregistering(const struct net_device *dev) 5070 { 5071 return dev->reg_state == NETREG_UNREGISTERING; 5072 } 5073 5074 static inline const char *netdev_reg_state(const struct net_device *dev) 5075 { 5076 switch (dev->reg_state) { 5077 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5078 case NETREG_REGISTERED: return ""; 5079 case NETREG_UNREGISTERING: return " (unregistering)"; 5080 case NETREG_UNREGISTERED: return " (unregistered)"; 5081 case NETREG_RELEASED: return " (released)"; 5082 case NETREG_DUMMY: return " (dummy)"; 5083 } 5084 5085 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5086 return " (unknown)"; 5087 } 5088 5089 __printf(3, 4) __cold 5090 void netdev_printk(const char *level, const struct net_device *dev, 5091 const char *format, ...); 5092 __printf(2, 3) __cold 5093 void netdev_emerg(const struct net_device *dev, const char *format, ...); 5094 __printf(2, 3) __cold 5095 void netdev_alert(const struct net_device *dev, const char *format, ...); 5096 __printf(2, 3) __cold 5097 void netdev_crit(const struct net_device *dev, const char *format, ...); 5098 __printf(2, 3) __cold 5099 void netdev_err(const struct net_device *dev, const char *format, ...); 5100 __printf(2, 3) __cold 5101 void netdev_warn(const struct net_device *dev, const char *format, ...); 5102 __printf(2, 3) __cold 5103 void netdev_notice(const struct net_device *dev, const char *format, ...); 5104 __printf(2, 3) __cold 5105 void netdev_info(const struct net_device *dev, const char *format, ...); 5106 5107 #define netdev_level_once(level, dev, fmt, ...) \ 5108 do { \ 5109 static bool __section(".data.once") __print_once; \ 5110 \ 5111 if (!__print_once) { \ 5112 __print_once = true; \ 5113 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 5114 } \ 5115 } while (0) 5116 5117 #define netdev_emerg_once(dev, fmt, ...) \ 5118 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 5119 #define netdev_alert_once(dev, fmt, ...) \ 5120 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 5121 #define netdev_crit_once(dev, fmt, ...) \ 5122 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 5123 #define netdev_err_once(dev, fmt, ...) \ 5124 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 5125 #define netdev_warn_once(dev, fmt, ...) \ 5126 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 5127 #define netdev_notice_once(dev, fmt, ...) \ 5128 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 5129 #define netdev_info_once(dev, fmt, ...) \ 5130 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 5131 5132 #define MODULE_ALIAS_NETDEV(device) \ 5133 MODULE_ALIAS("netdev-" device) 5134 5135 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5136 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5137 #define netdev_dbg(__dev, format, args...) \ 5138 do { \ 5139 dynamic_netdev_dbg(__dev, format, ##args); \ 5140 } while (0) 5141 #elif defined(DEBUG) 5142 #define netdev_dbg(__dev, format, args...) \ 5143 netdev_printk(KERN_DEBUG, __dev, format, ##args) 5144 #else 5145 #define netdev_dbg(__dev, format, args...) \ 5146 ({ \ 5147 if (0) \ 5148 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 5149 }) 5150 #endif 5151 5152 #if defined(VERBOSE_DEBUG) 5153 #define netdev_vdbg netdev_dbg 5154 #else 5155 5156 #define netdev_vdbg(dev, format, args...) \ 5157 ({ \ 5158 if (0) \ 5159 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 5160 0; \ 5161 }) 5162 #endif 5163 5164 /* 5165 * netdev_WARN() acts like dev_printk(), but with the key difference 5166 * of using a WARN/WARN_ON to get the message out, including the 5167 * file/line information and a backtrace. 5168 */ 5169 #define netdev_WARN(dev, format, args...) \ 5170 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5171 netdev_reg_state(dev), ##args) 5172 5173 #define netdev_WARN_ONCE(dev, format, args...) \ 5174 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5175 netdev_reg_state(dev), ##args) 5176 5177 /* netif printk helpers, similar to netdev_printk */ 5178 5179 #define netif_printk(priv, type, level, dev, fmt, args...) \ 5180 do { \ 5181 if (netif_msg_##type(priv)) \ 5182 netdev_printk(level, (dev), fmt, ##args); \ 5183 } while (0) 5184 5185 #define netif_level(level, priv, type, dev, fmt, args...) \ 5186 do { \ 5187 if (netif_msg_##type(priv)) \ 5188 netdev_##level(dev, fmt, ##args); \ 5189 } while (0) 5190 5191 #define netif_emerg(priv, type, dev, fmt, args...) \ 5192 netif_level(emerg, priv, type, dev, fmt, ##args) 5193 #define netif_alert(priv, type, dev, fmt, args...) \ 5194 netif_level(alert, priv, type, dev, fmt, ##args) 5195 #define netif_crit(priv, type, dev, fmt, args...) \ 5196 netif_level(crit, priv, type, dev, fmt, ##args) 5197 #define netif_err(priv, type, dev, fmt, args...) \ 5198 netif_level(err, priv, type, dev, fmt, ##args) 5199 #define netif_warn(priv, type, dev, fmt, args...) \ 5200 netif_level(warn, priv, type, dev, fmt, ##args) 5201 #define netif_notice(priv, type, dev, fmt, args...) \ 5202 netif_level(notice, priv, type, dev, fmt, ##args) 5203 #define netif_info(priv, type, dev, fmt, args...) \ 5204 netif_level(info, priv, type, dev, fmt, ##args) 5205 5206 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5207 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5208 #define netif_dbg(priv, type, netdev, format, args...) \ 5209 do { \ 5210 if (netif_msg_##type(priv)) \ 5211 dynamic_netdev_dbg(netdev, format, ##args); \ 5212 } while (0) 5213 #elif defined(DEBUG) 5214 #define netif_dbg(priv, type, dev, format, args...) \ 5215 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 5216 #else 5217 #define netif_dbg(priv, type, dev, format, args...) \ 5218 ({ \ 5219 if (0) \ 5220 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5221 0; \ 5222 }) 5223 #endif 5224 5225 /* if @cond then downgrade to debug, else print at @level */ 5226 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 5227 do { \ 5228 if (cond) \ 5229 netif_dbg(priv, type, netdev, fmt, ##args); \ 5230 else \ 5231 netif_ ## level(priv, type, netdev, fmt, ##args); \ 5232 } while (0) 5233 5234 #if defined(VERBOSE_DEBUG) 5235 #define netif_vdbg netif_dbg 5236 #else 5237 #define netif_vdbg(priv, type, dev, format, args...) \ 5238 ({ \ 5239 if (0) \ 5240 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5241 0; \ 5242 }) 5243 #endif 5244 5245 /* 5246 * The list of packet types we will receive (as opposed to discard) 5247 * and the routines to invoke. 5248 * 5249 * Why 16. Because with 16 the only overlap we get on a hash of the 5250 * low nibble of the protocol value is RARP/SNAP/X.25. 5251 * 5252 * 0800 IP 5253 * 0001 802.3 5254 * 0002 AX.25 5255 * 0004 802.2 5256 * 8035 RARP 5257 * 0005 SNAP 5258 * 0805 X.25 5259 * 0806 ARP 5260 * 8137 IPX 5261 * 0009 Localtalk 5262 * 86DD IPv6 5263 */ 5264 #define PTYPE_HASH_SIZE (16) 5265 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5266 5267 extern struct list_head ptype_all __read_mostly; 5268 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5269 5270 extern struct net_device *blackhole_netdev; 5271 5272 #endif /* _LINUX_NETDEVICE_H */ 5273