1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 13 * Corey Minyard <wf-rch!minyard@relay.EU.net> 14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov> 15 * Alan Cox, <alan@lxorguk.ukuu.org.uk> 16 * Bjorn Ekwall. <bj0rn@blox.se> 17 * Pekka Riikonen <priikone@poseidon.pspt.fi> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 32 #include <linux/percpu.h> 33 #include <linux/rculist.h> 34 #include <linux/workqueue.h> 35 #include <linux/dynamic_queue_limits.h> 36 37 #include <net/net_namespace.h> 38 #ifdef CONFIG_DCB 39 #include <net/dcbnl.h> 40 #endif 41 #include <net/netprio_cgroup.h> 42 #include <net/xdp.h> 43 44 #include <linux/netdev_features.h> 45 #include <linux/neighbour.h> 46 #include <uapi/linux/netdevice.h> 47 #include <uapi/linux/if_bonding.h> 48 #include <uapi/linux/pkt_cls.h> 49 #include <linux/hashtable.h> 50 #include <linux/rbtree.h> 51 #include <net/net_trackers.h> 52 53 struct netpoll_info; 54 struct device; 55 struct ethtool_ops; 56 struct phy_device; 57 struct dsa_port; 58 struct ip_tunnel_parm; 59 struct macsec_context; 60 struct macsec_ops; 61 62 struct sfp_bus; 63 /* 802.11 specific */ 64 struct wireless_dev; 65 /* 802.15.4 specific */ 66 struct wpan_dev; 67 struct mpls_dev; 68 /* UDP Tunnel offloads */ 69 struct udp_tunnel_info; 70 struct udp_tunnel_nic_info; 71 struct udp_tunnel_nic; 72 struct bpf_prog; 73 struct xdp_buff; 74 75 void synchronize_net(void); 76 void netdev_set_default_ethtool_ops(struct net_device *dev, 77 const struct ethtool_ops *ops); 78 79 /* Backlog congestion levels */ 80 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 81 #define NET_RX_DROP 1 /* packet dropped */ 82 83 #define MAX_NEST_DEV 8 84 85 /* 86 * Transmit return codes: transmit return codes originate from three different 87 * namespaces: 88 * 89 * - qdisc return codes 90 * - driver transmit return codes 91 * - errno values 92 * 93 * Drivers are allowed to return any one of those in their hard_start_xmit() 94 * function. Real network devices commonly used with qdiscs should only return 95 * the driver transmit return codes though - when qdiscs are used, the actual 96 * transmission happens asynchronously, so the value is not propagated to 97 * higher layers. Virtual network devices transmit synchronously; in this case 98 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 99 * others are propagated to higher layers. 100 */ 101 102 /* qdisc ->enqueue() return codes. */ 103 #define NET_XMIT_SUCCESS 0x00 104 #define NET_XMIT_DROP 0x01 /* skb dropped */ 105 #define NET_XMIT_CN 0x02 /* congestion notification */ 106 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 107 108 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 109 * indicates that the device will soon be dropping packets, or already drops 110 * some packets of the same priority; prompting us to send less aggressively. */ 111 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 112 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 113 114 /* Driver transmit return codes */ 115 #define NETDEV_TX_MASK 0xf0 116 117 enum netdev_tx { 118 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 119 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 120 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 121 }; 122 typedef enum netdev_tx netdev_tx_t; 123 124 /* 125 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 126 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 127 */ 128 static inline bool dev_xmit_complete(int rc) 129 { 130 /* 131 * Positive cases with an skb consumed by a driver: 132 * - successful transmission (rc == NETDEV_TX_OK) 133 * - error while transmitting (rc < 0) 134 * - error while queueing to a different device (rc & NET_XMIT_MASK) 135 */ 136 if (likely(rc < NET_XMIT_MASK)) 137 return true; 138 139 return false; 140 } 141 142 /* 143 * Compute the worst-case header length according to the protocols 144 * used. 145 */ 146 147 #if defined(CONFIG_HYPERV_NET) 148 # define LL_MAX_HEADER 128 149 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 150 # if defined(CONFIG_MAC80211_MESH) 151 # define LL_MAX_HEADER 128 152 # else 153 # define LL_MAX_HEADER 96 154 # endif 155 #else 156 # define LL_MAX_HEADER 32 157 #endif 158 159 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 160 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 161 #define MAX_HEADER LL_MAX_HEADER 162 #else 163 #define MAX_HEADER (LL_MAX_HEADER + 48) 164 #endif 165 166 /* 167 * Old network device statistics. Fields are native words 168 * (unsigned long) so they can be read and written atomically. 169 */ 170 171 struct net_device_stats { 172 unsigned long rx_packets; 173 unsigned long tx_packets; 174 unsigned long rx_bytes; 175 unsigned long tx_bytes; 176 unsigned long rx_errors; 177 unsigned long tx_errors; 178 unsigned long rx_dropped; 179 unsigned long tx_dropped; 180 unsigned long multicast; 181 unsigned long collisions; 182 unsigned long rx_length_errors; 183 unsigned long rx_over_errors; 184 unsigned long rx_crc_errors; 185 unsigned long rx_frame_errors; 186 unsigned long rx_fifo_errors; 187 unsigned long rx_missed_errors; 188 unsigned long tx_aborted_errors; 189 unsigned long tx_carrier_errors; 190 unsigned long tx_fifo_errors; 191 unsigned long tx_heartbeat_errors; 192 unsigned long tx_window_errors; 193 unsigned long rx_compressed; 194 unsigned long tx_compressed; 195 }; 196 197 198 #include <linux/cache.h> 199 #include <linux/skbuff.h> 200 201 #ifdef CONFIG_RPS 202 #include <linux/static_key.h> 203 extern struct static_key_false rps_needed; 204 extern struct static_key_false rfs_needed; 205 #endif 206 207 struct neighbour; 208 struct neigh_parms; 209 struct sk_buff; 210 211 struct netdev_hw_addr { 212 struct list_head list; 213 struct rb_node node; 214 unsigned char addr[MAX_ADDR_LEN]; 215 unsigned char type; 216 #define NETDEV_HW_ADDR_T_LAN 1 217 #define NETDEV_HW_ADDR_T_SAN 2 218 #define NETDEV_HW_ADDR_T_UNICAST 3 219 #define NETDEV_HW_ADDR_T_MULTICAST 4 220 bool global_use; 221 int sync_cnt; 222 int refcount; 223 int synced; 224 struct rcu_head rcu_head; 225 }; 226 227 struct netdev_hw_addr_list { 228 struct list_head list; 229 int count; 230 231 /* Auxiliary tree for faster lookup on addition and deletion */ 232 struct rb_root tree; 233 }; 234 235 #define netdev_hw_addr_list_count(l) ((l)->count) 236 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 237 #define netdev_hw_addr_list_for_each(ha, l) \ 238 list_for_each_entry(ha, &(l)->list, list) 239 240 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 241 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 242 #define netdev_for_each_uc_addr(ha, dev) \ 243 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 244 245 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 246 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 247 #define netdev_for_each_mc_addr(ha, dev) \ 248 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 249 250 struct hh_cache { 251 unsigned int hh_len; 252 seqlock_t hh_lock; 253 254 /* cached hardware header; allow for machine alignment needs. */ 255 #define HH_DATA_MOD 16 256 #define HH_DATA_OFF(__len) \ 257 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 258 #define HH_DATA_ALIGN(__len) \ 259 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 260 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 261 }; 262 263 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 264 * Alternative is: 265 * dev->hard_header_len ? (dev->hard_header_len + 266 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 267 * 268 * We could use other alignment values, but we must maintain the 269 * relationship HH alignment <= LL alignment. 270 */ 271 #define LL_RESERVED_SPACE(dev) \ 272 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 273 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 274 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 275 276 struct header_ops { 277 int (*create) (struct sk_buff *skb, struct net_device *dev, 278 unsigned short type, const void *daddr, 279 const void *saddr, unsigned int len); 280 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 281 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 282 void (*cache_update)(struct hh_cache *hh, 283 const struct net_device *dev, 284 const unsigned char *haddr); 285 bool (*validate)(const char *ll_header, unsigned int len); 286 __be16 (*parse_protocol)(const struct sk_buff *skb); 287 }; 288 289 /* These flag bits are private to the generic network queueing 290 * layer; they may not be explicitly referenced by any other 291 * code. 292 */ 293 294 enum netdev_state_t { 295 __LINK_STATE_START, 296 __LINK_STATE_PRESENT, 297 __LINK_STATE_NOCARRIER, 298 __LINK_STATE_LINKWATCH_PENDING, 299 __LINK_STATE_DORMANT, 300 __LINK_STATE_TESTING, 301 }; 302 303 struct gro_list { 304 struct list_head list; 305 int count; 306 }; 307 308 /* 309 * size of gro hash buckets, must less than bit number of 310 * napi_struct::gro_bitmask 311 */ 312 #define GRO_HASH_BUCKETS 8 313 314 /* 315 * Structure for NAPI scheduling similar to tasklet but with weighting 316 */ 317 struct napi_struct { 318 /* The poll_list must only be managed by the entity which 319 * changes the state of the NAPI_STATE_SCHED bit. This means 320 * whoever atomically sets that bit can add this napi_struct 321 * to the per-CPU poll_list, and whoever clears that bit 322 * can remove from the list right before clearing the bit. 323 */ 324 struct list_head poll_list; 325 326 unsigned long state; 327 int weight; 328 int defer_hard_irqs_count; 329 unsigned long gro_bitmask; 330 int (*poll)(struct napi_struct *, int); 331 #ifdef CONFIG_NETPOLL 332 int poll_owner; 333 #endif 334 struct net_device *dev; 335 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 336 struct sk_buff *skb; 337 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 338 int rx_count; /* length of rx_list */ 339 struct hrtimer timer; 340 struct list_head dev_list; 341 struct hlist_node napi_hash_node; 342 unsigned int napi_id; 343 struct task_struct *thread; 344 }; 345 346 enum { 347 NAPI_STATE_SCHED, /* Poll is scheduled */ 348 NAPI_STATE_MISSED, /* reschedule a napi */ 349 NAPI_STATE_DISABLE, /* Disable pending */ 350 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 351 NAPI_STATE_LISTED, /* NAPI added to system lists */ 352 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 353 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 354 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 355 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 356 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 357 }; 358 359 enum { 360 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 361 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 362 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 363 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 364 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 365 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 366 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 367 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 368 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 369 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 370 }; 371 372 enum gro_result { 373 GRO_MERGED, 374 GRO_MERGED_FREE, 375 GRO_HELD, 376 GRO_NORMAL, 377 GRO_CONSUMED, 378 }; 379 typedef enum gro_result gro_result_t; 380 381 /* 382 * enum rx_handler_result - Possible return values for rx_handlers. 383 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 384 * further. 385 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 386 * case skb->dev was changed by rx_handler. 387 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 388 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 389 * 390 * rx_handlers are functions called from inside __netif_receive_skb(), to do 391 * special processing of the skb, prior to delivery to protocol handlers. 392 * 393 * Currently, a net_device can only have a single rx_handler registered. Trying 394 * to register a second rx_handler will return -EBUSY. 395 * 396 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 397 * To unregister a rx_handler on a net_device, use 398 * netdev_rx_handler_unregister(). 399 * 400 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 401 * do with the skb. 402 * 403 * If the rx_handler consumed the skb in some way, it should return 404 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 405 * the skb to be delivered in some other way. 406 * 407 * If the rx_handler changed skb->dev, to divert the skb to another 408 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 409 * new device will be called if it exists. 410 * 411 * If the rx_handler decides the skb should be ignored, it should return 412 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 413 * are registered on exact device (ptype->dev == skb->dev). 414 * 415 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 416 * delivered, it should return RX_HANDLER_PASS. 417 * 418 * A device without a registered rx_handler will behave as if rx_handler 419 * returned RX_HANDLER_PASS. 420 */ 421 422 enum rx_handler_result { 423 RX_HANDLER_CONSUMED, 424 RX_HANDLER_ANOTHER, 425 RX_HANDLER_EXACT, 426 RX_HANDLER_PASS, 427 }; 428 typedef enum rx_handler_result rx_handler_result_t; 429 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 430 431 void __napi_schedule(struct napi_struct *n); 432 void __napi_schedule_irqoff(struct napi_struct *n); 433 434 static inline bool napi_disable_pending(struct napi_struct *n) 435 { 436 return test_bit(NAPI_STATE_DISABLE, &n->state); 437 } 438 439 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 440 { 441 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 442 } 443 444 bool napi_schedule_prep(struct napi_struct *n); 445 446 /** 447 * napi_schedule - schedule NAPI poll 448 * @n: NAPI context 449 * 450 * Schedule NAPI poll routine to be called if it is not already 451 * running. 452 */ 453 static inline void napi_schedule(struct napi_struct *n) 454 { 455 if (napi_schedule_prep(n)) 456 __napi_schedule(n); 457 } 458 459 /** 460 * napi_schedule_irqoff - schedule NAPI poll 461 * @n: NAPI context 462 * 463 * Variant of napi_schedule(), assuming hard irqs are masked. 464 */ 465 static inline void napi_schedule_irqoff(struct napi_struct *n) 466 { 467 if (napi_schedule_prep(n)) 468 __napi_schedule_irqoff(n); 469 } 470 471 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 472 static inline bool napi_reschedule(struct napi_struct *napi) 473 { 474 if (napi_schedule_prep(napi)) { 475 __napi_schedule(napi); 476 return true; 477 } 478 return false; 479 } 480 481 bool napi_complete_done(struct napi_struct *n, int work_done); 482 /** 483 * napi_complete - NAPI processing complete 484 * @n: NAPI context 485 * 486 * Mark NAPI processing as complete. 487 * Consider using napi_complete_done() instead. 488 * Return false if device should avoid rearming interrupts. 489 */ 490 static inline bool napi_complete(struct napi_struct *n) 491 { 492 return napi_complete_done(n, 0); 493 } 494 495 int dev_set_threaded(struct net_device *dev, bool threaded); 496 497 /** 498 * napi_disable - prevent NAPI from scheduling 499 * @n: NAPI context 500 * 501 * Stop NAPI from being scheduled on this context. 502 * Waits till any outstanding processing completes. 503 */ 504 void napi_disable(struct napi_struct *n); 505 506 void napi_enable(struct napi_struct *n); 507 508 /** 509 * napi_synchronize - wait until NAPI is not running 510 * @n: NAPI context 511 * 512 * Wait until NAPI is done being scheduled on this context. 513 * Waits till any outstanding processing completes but 514 * does not disable future activations. 515 */ 516 static inline void napi_synchronize(const struct napi_struct *n) 517 { 518 if (IS_ENABLED(CONFIG_SMP)) 519 while (test_bit(NAPI_STATE_SCHED, &n->state)) 520 msleep(1); 521 else 522 barrier(); 523 } 524 525 /** 526 * napi_if_scheduled_mark_missed - if napi is running, set the 527 * NAPIF_STATE_MISSED 528 * @n: NAPI context 529 * 530 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 531 * NAPI is scheduled. 532 **/ 533 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 534 { 535 unsigned long val, new; 536 537 do { 538 val = READ_ONCE(n->state); 539 if (val & NAPIF_STATE_DISABLE) 540 return true; 541 542 if (!(val & NAPIF_STATE_SCHED)) 543 return false; 544 545 new = val | NAPIF_STATE_MISSED; 546 } while (cmpxchg(&n->state, val, new) != val); 547 548 return true; 549 } 550 551 enum netdev_queue_state_t { 552 __QUEUE_STATE_DRV_XOFF, 553 __QUEUE_STATE_STACK_XOFF, 554 __QUEUE_STATE_FROZEN, 555 }; 556 557 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 558 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 559 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 560 561 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 562 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 563 QUEUE_STATE_FROZEN) 564 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 565 QUEUE_STATE_FROZEN) 566 567 /* 568 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 569 * netif_tx_* functions below are used to manipulate this flag. The 570 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 571 * queue independently. The netif_xmit_*stopped functions below are called 572 * to check if the queue has been stopped by the driver or stack (either 573 * of the XOFF bits are set in the state). Drivers should not need to call 574 * netif_xmit*stopped functions, they should only be using netif_tx_*. 575 */ 576 577 struct netdev_queue { 578 /* 579 * read-mostly part 580 */ 581 struct net_device *dev; 582 netdevice_tracker dev_tracker; 583 584 struct Qdisc __rcu *qdisc; 585 struct Qdisc *qdisc_sleeping; 586 #ifdef CONFIG_SYSFS 587 struct kobject kobj; 588 #endif 589 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 590 int numa_node; 591 #endif 592 unsigned long tx_maxrate; 593 /* 594 * Number of TX timeouts for this queue 595 * (/sys/class/net/DEV/Q/trans_timeout) 596 */ 597 atomic_long_t trans_timeout; 598 599 /* Subordinate device that the queue has been assigned to */ 600 struct net_device *sb_dev; 601 #ifdef CONFIG_XDP_SOCKETS 602 struct xsk_buff_pool *pool; 603 #endif 604 /* 605 * write-mostly part 606 */ 607 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 608 int xmit_lock_owner; 609 /* 610 * Time (in jiffies) of last Tx 611 */ 612 unsigned long trans_start; 613 614 unsigned long state; 615 616 #ifdef CONFIG_BQL 617 struct dql dql; 618 #endif 619 } ____cacheline_aligned_in_smp; 620 621 extern int sysctl_fb_tunnels_only_for_init_net; 622 extern int sysctl_devconf_inherit_init_net; 623 624 /* 625 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 626 * == 1 : For initns only 627 * == 2 : For none. 628 */ 629 static inline bool net_has_fallback_tunnels(const struct net *net) 630 { 631 return !IS_ENABLED(CONFIG_SYSCTL) || 632 !sysctl_fb_tunnels_only_for_init_net || 633 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1); 634 } 635 636 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 637 { 638 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 639 return q->numa_node; 640 #else 641 return NUMA_NO_NODE; 642 #endif 643 } 644 645 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 646 { 647 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 648 q->numa_node = node; 649 #endif 650 } 651 652 #ifdef CONFIG_RPS 653 /* 654 * This structure holds an RPS map which can be of variable length. The 655 * map is an array of CPUs. 656 */ 657 struct rps_map { 658 unsigned int len; 659 struct rcu_head rcu; 660 u16 cpus[]; 661 }; 662 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 663 664 /* 665 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 666 * tail pointer for that CPU's input queue at the time of last enqueue, and 667 * a hardware filter index. 668 */ 669 struct rps_dev_flow { 670 u16 cpu; 671 u16 filter; 672 unsigned int last_qtail; 673 }; 674 #define RPS_NO_FILTER 0xffff 675 676 /* 677 * The rps_dev_flow_table structure contains a table of flow mappings. 678 */ 679 struct rps_dev_flow_table { 680 unsigned int mask; 681 struct rcu_head rcu; 682 struct rps_dev_flow flows[]; 683 }; 684 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 685 ((_num) * sizeof(struct rps_dev_flow))) 686 687 /* 688 * The rps_sock_flow_table contains mappings of flows to the last CPU 689 * on which they were processed by the application (set in recvmsg). 690 * Each entry is a 32bit value. Upper part is the high-order bits 691 * of flow hash, lower part is CPU number. 692 * rps_cpu_mask is used to partition the space, depending on number of 693 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 694 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 695 * meaning we use 32-6=26 bits for the hash. 696 */ 697 struct rps_sock_flow_table { 698 u32 mask; 699 700 u32 ents[] ____cacheline_aligned_in_smp; 701 }; 702 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 703 704 #define RPS_NO_CPU 0xffff 705 706 extern u32 rps_cpu_mask; 707 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 708 709 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 710 u32 hash) 711 { 712 if (table && hash) { 713 unsigned int index = hash & table->mask; 714 u32 val = hash & ~rps_cpu_mask; 715 716 /* We only give a hint, preemption can change CPU under us */ 717 val |= raw_smp_processor_id(); 718 719 if (table->ents[index] != val) 720 table->ents[index] = val; 721 } 722 } 723 724 #ifdef CONFIG_RFS_ACCEL 725 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 726 u16 filter_id); 727 #endif 728 #endif /* CONFIG_RPS */ 729 730 /* This structure contains an instance of an RX queue. */ 731 struct netdev_rx_queue { 732 struct xdp_rxq_info xdp_rxq; 733 #ifdef CONFIG_RPS 734 struct rps_map __rcu *rps_map; 735 struct rps_dev_flow_table __rcu *rps_flow_table; 736 #endif 737 struct kobject kobj; 738 struct net_device *dev; 739 netdevice_tracker dev_tracker; 740 741 #ifdef CONFIG_XDP_SOCKETS 742 struct xsk_buff_pool *pool; 743 #endif 744 } ____cacheline_aligned_in_smp; 745 746 /* 747 * RX queue sysfs structures and functions. 748 */ 749 struct rx_queue_attribute { 750 struct attribute attr; 751 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 752 ssize_t (*store)(struct netdev_rx_queue *queue, 753 const char *buf, size_t len); 754 }; 755 756 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 757 enum xps_map_type { 758 XPS_CPUS = 0, 759 XPS_RXQS, 760 XPS_MAPS_MAX, 761 }; 762 763 #ifdef CONFIG_XPS 764 /* 765 * This structure holds an XPS map which can be of variable length. The 766 * map is an array of queues. 767 */ 768 struct xps_map { 769 unsigned int len; 770 unsigned int alloc_len; 771 struct rcu_head rcu; 772 u16 queues[]; 773 }; 774 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 775 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 776 - sizeof(struct xps_map)) / sizeof(u16)) 777 778 /* 779 * This structure holds all XPS maps for device. Maps are indexed by CPU. 780 * 781 * We keep track of the number of cpus/rxqs used when the struct is allocated, 782 * in nr_ids. This will help not accessing out-of-bound memory. 783 * 784 * We keep track of the number of traffic classes used when the struct is 785 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 786 * not crossing its upper bound, as the original dev->num_tc can be updated in 787 * the meantime. 788 */ 789 struct xps_dev_maps { 790 struct rcu_head rcu; 791 unsigned int nr_ids; 792 s16 num_tc; 793 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 794 }; 795 796 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 797 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 798 799 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 800 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 801 802 #endif /* CONFIG_XPS */ 803 804 #define TC_MAX_QUEUE 16 805 #define TC_BITMASK 15 806 /* HW offloaded queuing disciplines txq count and offset maps */ 807 struct netdev_tc_txq { 808 u16 count; 809 u16 offset; 810 }; 811 812 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 813 /* 814 * This structure is to hold information about the device 815 * configured to run FCoE protocol stack. 816 */ 817 struct netdev_fcoe_hbainfo { 818 char manufacturer[64]; 819 char serial_number[64]; 820 char hardware_version[64]; 821 char driver_version[64]; 822 char optionrom_version[64]; 823 char firmware_version[64]; 824 char model[256]; 825 char model_description[256]; 826 }; 827 #endif 828 829 #define MAX_PHYS_ITEM_ID_LEN 32 830 831 /* This structure holds a unique identifier to identify some 832 * physical item (port for example) used by a netdevice. 833 */ 834 struct netdev_phys_item_id { 835 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 836 unsigned char id_len; 837 }; 838 839 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 840 struct netdev_phys_item_id *b) 841 { 842 return a->id_len == b->id_len && 843 memcmp(a->id, b->id, a->id_len) == 0; 844 } 845 846 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 847 struct sk_buff *skb, 848 struct net_device *sb_dev); 849 850 enum net_device_path_type { 851 DEV_PATH_ETHERNET = 0, 852 DEV_PATH_VLAN, 853 DEV_PATH_BRIDGE, 854 DEV_PATH_PPPOE, 855 DEV_PATH_DSA, 856 }; 857 858 struct net_device_path { 859 enum net_device_path_type type; 860 const struct net_device *dev; 861 union { 862 struct { 863 u16 id; 864 __be16 proto; 865 u8 h_dest[ETH_ALEN]; 866 } encap; 867 struct { 868 enum { 869 DEV_PATH_BR_VLAN_KEEP, 870 DEV_PATH_BR_VLAN_TAG, 871 DEV_PATH_BR_VLAN_UNTAG, 872 DEV_PATH_BR_VLAN_UNTAG_HW, 873 } vlan_mode; 874 u16 vlan_id; 875 __be16 vlan_proto; 876 } bridge; 877 struct { 878 int port; 879 u16 proto; 880 } dsa; 881 }; 882 }; 883 884 #define NET_DEVICE_PATH_STACK_MAX 5 885 #define NET_DEVICE_PATH_VLAN_MAX 2 886 887 struct net_device_path_stack { 888 int num_paths; 889 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 890 }; 891 892 struct net_device_path_ctx { 893 const struct net_device *dev; 894 const u8 *daddr; 895 896 int num_vlans; 897 struct { 898 u16 id; 899 __be16 proto; 900 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 901 }; 902 903 enum tc_setup_type { 904 TC_SETUP_QDISC_MQPRIO, 905 TC_SETUP_CLSU32, 906 TC_SETUP_CLSFLOWER, 907 TC_SETUP_CLSMATCHALL, 908 TC_SETUP_CLSBPF, 909 TC_SETUP_BLOCK, 910 TC_SETUP_QDISC_CBS, 911 TC_SETUP_QDISC_RED, 912 TC_SETUP_QDISC_PRIO, 913 TC_SETUP_QDISC_MQ, 914 TC_SETUP_QDISC_ETF, 915 TC_SETUP_ROOT_QDISC, 916 TC_SETUP_QDISC_GRED, 917 TC_SETUP_QDISC_TAPRIO, 918 TC_SETUP_FT, 919 TC_SETUP_QDISC_ETS, 920 TC_SETUP_QDISC_TBF, 921 TC_SETUP_QDISC_FIFO, 922 TC_SETUP_QDISC_HTB, 923 TC_SETUP_ACT, 924 }; 925 926 /* These structures hold the attributes of bpf state that are being passed 927 * to the netdevice through the bpf op. 928 */ 929 enum bpf_netdev_command { 930 /* Set or clear a bpf program used in the earliest stages of packet 931 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 932 * is responsible for calling bpf_prog_put on any old progs that are 933 * stored. In case of error, the callee need not release the new prog 934 * reference, but on success it takes ownership and must bpf_prog_put 935 * when it is no longer used. 936 */ 937 XDP_SETUP_PROG, 938 XDP_SETUP_PROG_HW, 939 /* BPF program for offload callbacks, invoked at program load time. */ 940 BPF_OFFLOAD_MAP_ALLOC, 941 BPF_OFFLOAD_MAP_FREE, 942 XDP_SETUP_XSK_POOL, 943 }; 944 945 struct bpf_prog_offload_ops; 946 struct netlink_ext_ack; 947 struct xdp_umem; 948 struct xdp_dev_bulk_queue; 949 struct bpf_xdp_link; 950 951 enum bpf_xdp_mode { 952 XDP_MODE_SKB = 0, 953 XDP_MODE_DRV = 1, 954 XDP_MODE_HW = 2, 955 __MAX_XDP_MODE 956 }; 957 958 struct bpf_xdp_entity { 959 struct bpf_prog *prog; 960 struct bpf_xdp_link *link; 961 }; 962 963 struct netdev_bpf { 964 enum bpf_netdev_command command; 965 union { 966 /* XDP_SETUP_PROG */ 967 struct { 968 u32 flags; 969 struct bpf_prog *prog; 970 struct netlink_ext_ack *extack; 971 }; 972 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 973 struct { 974 struct bpf_offloaded_map *offmap; 975 }; 976 /* XDP_SETUP_XSK_POOL */ 977 struct { 978 struct xsk_buff_pool *pool; 979 u16 queue_id; 980 } xsk; 981 }; 982 }; 983 984 /* Flags for ndo_xsk_wakeup. */ 985 #define XDP_WAKEUP_RX (1 << 0) 986 #define XDP_WAKEUP_TX (1 << 1) 987 988 #ifdef CONFIG_XFRM_OFFLOAD 989 struct xfrmdev_ops { 990 int (*xdo_dev_state_add) (struct xfrm_state *x); 991 void (*xdo_dev_state_delete) (struct xfrm_state *x); 992 void (*xdo_dev_state_free) (struct xfrm_state *x); 993 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 994 struct xfrm_state *x); 995 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 996 }; 997 #endif 998 999 struct dev_ifalias { 1000 struct rcu_head rcuhead; 1001 char ifalias[]; 1002 }; 1003 1004 struct devlink; 1005 struct tlsdev_ops; 1006 1007 struct netdev_name_node { 1008 struct hlist_node hlist; 1009 struct list_head list; 1010 struct net_device *dev; 1011 const char *name; 1012 }; 1013 1014 int netdev_name_node_alt_create(struct net_device *dev, const char *name); 1015 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name); 1016 1017 struct netdev_net_notifier { 1018 struct list_head list; 1019 struct notifier_block *nb; 1020 }; 1021 1022 /* 1023 * This structure defines the management hooks for network devices. 1024 * The following hooks can be defined; unless noted otherwise, they are 1025 * optional and can be filled with a null pointer. 1026 * 1027 * int (*ndo_init)(struct net_device *dev); 1028 * This function is called once when a network device is registered. 1029 * The network device can use this for any late stage initialization 1030 * or semantic validation. It can fail with an error code which will 1031 * be propagated back to register_netdev. 1032 * 1033 * void (*ndo_uninit)(struct net_device *dev); 1034 * This function is called when device is unregistered or when registration 1035 * fails. It is not called if init fails. 1036 * 1037 * int (*ndo_open)(struct net_device *dev); 1038 * This function is called when a network device transitions to the up 1039 * state. 1040 * 1041 * int (*ndo_stop)(struct net_device *dev); 1042 * This function is called when a network device transitions to the down 1043 * state. 1044 * 1045 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1046 * struct net_device *dev); 1047 * Called when a packet needs to be transmitted. 1048 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1049 * the queue before that can happen; it's for obsolete devices and weird 1050 * corner cases, but the stack really does a non-trivial amount 1051 * of useless work if you return NETDEV_TX_BUSY. 1052 * Required; cannot be NULL. 1053 * 1054 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1055 * struct net_device *dev 1056 * netdev_features_t features); 1057 * Called by core transmit path to determine if device is capable of 1058 * performing offload operations on a given packet. This is to give 1059 * the device an opportunity to implement any restrictions that cannot 1060 * be otherwise expressed by feature flags. The check is called with 1061 * the set of features that the stack has calculated and it returns 1062 * those the driver believes to be appropriate. 1063 * 1064 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1065 * struct net_device *sb_dev); 1066 * Called to decide which queue to use when device supports multiple 1067 * transmit queues. 1068 * 1069 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1070 * This function is called to allow device receiver to make 1071 * changes to configuration when multicast or promiscuous is enabled. 1072 * 1073 * void (*ndo_set_rx_mode)(struct net_device *dev); 1074 * This function is called device changes address list filtering. 1075 * If driver handles unicast address filtering, it should set 1076 * IFF_UNICAST_FLT in its priv_flags. 1077 * 1078 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1079 * This function is called when the Media Access Control address 1080 * needs to be changed. If this interface is not defined, the 1081 * MAC address can not be changed. 1082 * 1083 * int (*ndo_validate_addr)(struct net_device *dev); 1084 * Test if Media Access Control address is valid for the device. 1085 * 1086 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1087 * Old-style ioctl entry point. This is used internally by the 1088 * appletalk and ieee802154 subsystems but is no longer called by 1089 * the device ioctl handler. 1090 * 1091 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1092 * Used by the bonding driver for its device specific ioctls: 1093 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1094 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1095 * 1096 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1097 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1098 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1099 * 1100 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1101 * Used to set network devices bus interface parameters. This interface 1102 * is retained for legacy reasons; new devices should use the bus 1103 * interface (PCI) for low level management. 1104 * 1105 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1106 * Called when a user wants to change the Maximum Transfer Unit 1107 * of a device. 1108 * 1109 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1110 * Callback used when the transmitter has not made any progress 1111 * for dev->watchdog ticks. 1112 * 1113 * void (*ndo_get_stats64)(struct net_device *dev, 1114 * struct rtnl_link_stats64 *storage); 1115 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1116 * Called when a user wants to get the network device usage 1117 * statistics. Drivers must do one of the following: 1118 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1119 * rtnl_link_stats64 structure passed by the caller. 1120 * 2. Define @ndo_get_stats to update a net_device_stats structure 1121 * (which should normally be dev->stats) and return a pointer to 1122 * it. The structure may be changed asynchronously only if each 1123 * field is written atomically. 1124 * 3. Update dev->stats asynchronously and atomically, and define 1125 * neither operation. 1126 * 1127 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1128 * Return true if this device supports offload stats of this attr_id. 1129 * 1130 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1131 * void *attr_data) 1132 * Get statistics for offload operations by attr_id. Write it into the 1133 * attr_data pointer. 1134 * 1135 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1136 * If device supports VLAN filtering this function is called when a 1137 * VLAN id is registered. 1138 * 1139 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1140 * If device supports VLAN filtering this function is called when a 1141 * VLAN id is unregistered. 1142 * 1143 * void (*ndo_poll_controller)(struct net_device *dev); 1144 * 1145 * SR-IOV management functions. 1146 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1147 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1148 * u8 qos, __be16 proto); 1149 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1150 * int max_tx_rate); 1151 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1152 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1153 * int (*ndo_get_vf_config)(struct net_device *dev, 1154 * int vf, struct ifla_vf_info *ivf); 1155 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1156 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1157 * struct nlattr *port[]); 1158 * 1159 * Enable or disable the VF ability to query its RSS Redirection Table and 1160 * Hash Key. This is needed since on some devices VF share this information 1161 * with PF and querying it may introduce a theoretical security risk. 1162 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1163 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1164 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1165 * void *type_data); 1166 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1167 * This is always called from the stack with the rtnl lock held and netif 1168 * tx queues stopped. This allows the netdevice to perform queue 1169 * management safely. 1170 * 1171 * Fiber Channel over Ethernet (FCoE) offload functions. 1172 * int (*ndo_fcoe_enable)(struct net_device *dev); 1173 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1174 * so the underlying device can perform whatever needed configuration or 1175 * initialization to support acceleration of FCoE traffic. 1176 * 1177 * int (*ndo_fcoe_disable)(struct net_device *dev); 1178 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1179 * so the underlying device can perform whatever needed clean-ups to 1180 * stop supporting acceleration of FCoE traffic. 1181 * 1182 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1183 * struct scatterlist *sgl, unsigned int sgc); 1184 * Called when the FCoE Initiator wants to initialize an I/O that 1185 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1186 * perform necessary setup and returns 1 to indicate the device is set up 1187 * successfully to perform DDP on this I/O, otherwise this returns 0. 1188 * 1189 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1190 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1191 * indicated by the FC exchange id 'xid', so the underlying device can 1192 * clean up and reuse resources for later DDP requests. 1193 * 1194 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1195 * struct scatterlist *sgl, unsigned int sgc); 1196 * Called when the FCoE Target wants to initialize an I/O that 1197 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1198 * perform necessary setup and returns 1 to indicate the device is set up 1199 * successfully to perform DDP on this I/O, otherwise this returns 0. 1200 * 1201 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1202 * struct netdev_fcoe_hbainfo *hbainfo); 1203 * Called when the FCoE Protocol stack wants information on the underlying 1204 * device. This information is utilized by the FCoE protocol stack to 1205 * register attributes with Fiber Channel management service as per the 1206 * FC-GS Fabric Device Management Information(FDMI) specification. 1207 * 1208 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1209 * Called when the underlying device wants to override default World Wide 1210 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1211 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1212 * protocol stack to use. 1213 * 1214 * RFS acceleration. 1215 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1216 * u16 rxq_index, u32 flow_id); 1217 * Set hardware filter for RFS. rxq_index is the target queue index; 1218 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1219 * Return the filter ID on success, or a negative error code. 1220 * 1221 * Slave management functions (for bridge, bonding, etc). 1222 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1223 * Called to make another netdev an underling. 1224 * 1225 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1226 * Called to release previously enslaved netdev. 1227 * 1228 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1229 * struct sk_buff *skb, 1230 * bool all_slaves); 1231 * Get the xmit slave of master device. If all_slaves is true, function 1232 * assume all the slaves can transmit. 1233 * 1234 * Feature/offload setting functions. 1235 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1236 * netdev_features_t features); 1237 * Adjusts the requested feature flags according to device-specific 1238 * constraints, and returns the resulting flags. Must not modify 1239 * the device state. 1240 * 1241 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1242 * Called to update device configuration to new features. Passed 1243 * feature set might be less than what was returned by ndo_fix_features()). 1244 * Must return >0 or -errno if it changed dev->features itself. 1245 * 1246 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1247 * struct net_device *dev, 1248 * const unsigned char *addr, u16 vid, u16 flags, 1249 * struct netlink_ext_ack *extack); 1250 * Adds an FDB entry to dev for addr. 1251 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1252 * struct net_device *dev, 1253 * const unsigned char *addr, u16 vid) 1254 * Deletes the FDB entry from dev coresponding to addr. 1255 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1256 * struct net_device *dev, struct net_device *filter_dev, 1257 * int *idx) 1258 * Used to add FDB entries to dump requests. Implementers should add 1259 * entries to skb and update idx with the number of entries. 1260 * 1261 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1262 * u16 flags, struct netlink_ext_ack *extack) 1263 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1264 * struct net_device *dev, u32 filter_mask, 1265 * int nlflags) 1266 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1267 * u16 flags); 1268 * 1269 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1270 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1271 * which do not represent real hardware may define this to allow their 1272 * userspace components to manage their virtual carrier state. Devices 1273 * that determine carrier state from physical hardware properties (eg 1274 * network cables) or protocol-dependent mechanisms (eg 1275 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1276 * 1277 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1278 * struct netdev_phys_item_id *ppid); 1279 * Called to get ID of physical port of this device. If driver does 1280 * not implement this, it is assumed that the hw is not able to have 1281 * multiple net devices on single physical port. 1282 * 1283 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1284 * struct netdev_phys_item_id *ppid) 1285 * Called to get the parent ID of the physical port of this device. 1286 * 1287 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1288 * struct net_device *dev) 1289 * Called by upper layer devices to accelerate switching or other 1290 * station functionality into hardware. 'pdev is the lowerdev 1291 * to use for the offload and 'dev' is the net device that will 1292 * back the offload. Returns a pointer to the private structure 1293 * the upper layer will maintain. 1294 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1295 * Called by upper layer device to delete the station created 1296 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1297 * the station and priv is the structure returned by the add 1298 * operation. 1299 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1300 * int queue_index, u32 maxrate); 1301 * Called when a user wants to set a max-rate limitation of specific 1302 * TX queue. 1303 * int (*ndo_get_iflink)(const struct net_device *dev); 1304 * Called to get the iflink value of this device. 1305 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1306 * This function is used to get egress tunnel information for given skb. 1307 * This is useful for retrieving outer tunnel header parameters while 1308 * sampling packet. 1309 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1310 * This function is used to specify the headroom that the skb must 1311 * consider when allocation skb during packet reception. Setting 1312 * appropriate rx headroom value allows avoiding skb head copy on 1313 * forward. Setting a negative value resets the rx headroom to the 1314 * default value. 1315 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1316 * This function is used to set or query state related to XDP on the 1317 * netdevice and manage BPF offload. See definition of 1318 * enum bpf_netdev_command for details. 1319 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1320 * u32 flags); 1321 * This function is used to submit @n XDP packets for transmit on a 1322 * netdevice. Returns number of frames successfully transmitted, frames 1323 * that got dropped are freed/returned via xdp_return_frame(). 1324 * Returns negative number, means general error invoking ndo, meaning 1325 * no frames were xmit'ed and core-caller will free all frames. 1326 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1327 * struct xdp_buff *xdp); 1328 * Get the xmit slave of master device based on the xdp_buff. 1329 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1330 * This function is used to wake up the softirq, ksoftirqd or kthread 1331 * responsible for sending and/or receiving packets on a specific 1332 * queue id bound to an AF_XDP socket. The flags field specifies if 1333 * only RX, only Tx, or both should be woken up using the flags 1334 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1335 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1336 * Get devlink port instance associated with a given netdev. 1337 * Called with a reference on the netdevice and devlink locks only, 1338 * rtnl_lock is not held. 1339 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1340 * int cmd); 1341 * Add, change, delete or get information on an IPv4 tunnel. 1342 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1343 * If a device is paired with a peer device, return the peer instance. 1344 * The caller must be under RCU read context. 1345 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1346 * Get the forwarding path to reach the real device from the HW destination address 1347 */ 1348 struct net_device_ops { 1349 int (*ndo_init)(struct net_device *dev); 1350 void (*ndo_uninit)(struct net_device *dev); 1351 int (*ndo_open)(struct net_device *dev); 1352 int (*ndo_stop)(struct net_device *dev); 1353 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1354 struct net_device *dev); 1355 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1356 struct net_device *dev, 1357 netdev_features_t features); 1358 u16 (*ndo_select_queue)(struct net_device *dev, 1359 struct sk_buff *skb, 1360 struct net_device *sb_dev); 1361 void (*ndo_change_rx_flags)(struct net_device *dev, 1362 int flags); 1363 void (*ndo_set_rx_mode)(struct net_device *dev); 1364 int (*ndo_set_mac_address)(struct net_device *dev, 1365 void *addr); 1366 int (*ndo_validate_addr)(struct net_device *dev); 1367 int (*ndo_do_ioctl)(struct net_device *dev, 1368 struct ifreq *ifr, int cmd); 1369 int (*ndo_eth_ioctl)(struct net_device *dev, 1370 struct ifreq *ifr, int cmd); 1371 int (*ndo_siocbond)(struct net_device *dev, 1372 struct ifreq *ifr, int cmd); 1373 int (*ndo_siocwandev)(struct net_device *dev, 1374 struct if_settings *ifs); 1375 int (*ndo_siocdevprivate)(struct net_device *dev, 1376 struct ifreq *ifr, 1377 void __user *data, int cmd); 1378 int (*ndo_set_config)(struct net_device *dev, 1379 struct ifmap *map); 1380 int (*ndo_change_mtu)(struct net_device *dev, 1381 int new_mtu); 1382 int (*ndo_neigh_setup)(struct net_device *dev, 1383 struct neigh_parms *); 1384 void (*ndo_tx_timeout) (struct net_device *dev, 1385 unsigned int txqueue); 1386 1387 void (*ndo_get_stats64)(struct net_device *dev, 1388 struct rtnl_link_stats64 *storage); 1389 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1390 int (*ndo_get_offload_stats)(int attr_id, 1391 const struct net_device *dev, 1392 void *attr_data); 1393 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1394 1395 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1396 __be16 proto, u16 vid); 1397 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1398 __be16 proto, u16 vid); 1399 #ifdef CONFIG_NET_POLL_CONTROLLER 1400 void (*ndo_poll_controller)(struct net_device *dev); 1401 int (*ndo_netpoll_setup)(struct net_device *dev, 1402 struct netpoll_info *info); 1403 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1404 #endif 1405 int (*ndo_set_vf_mac)(struct net_device *dev, 1406 int queue, u8 *mac); 1407 int (*ndo_set_vf_vlan)(struct net_device *dev, 1408 int queue, u16 vlan, 1409 u8 qos, __be16 proto); 1410 int (*ndo_set_vf_rate)(struct net_device *dev, 1411 int vf, int min_tx_rate, 1412 int max_tx_rate); 1413 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1414 int vf, bool setting); 1415 int (*ndo_set_vf_trust)(struct net_device *dev, 1416 int vf, bool setting); 1417 int (*ndo_get_vf_config)(struct net_device *dev, 1418 int vf, 1419 struct ifla_vf_info *ivf); 1420 int (*ndo_set_vf_link_state)(struct net_device *dev, 1421 int vf, int link_state); 1422 int (*ndo_get_vf_stats)(struct net_device *dev, 1423 int vf, 1424 struct ifla_vf_stats 1425 *vf_stats); 1426 int (*ndo_set_vf_port)(struct net_device *dev, 1427 int vf, 1428 struct nlattr *port[]); 1429 int (*ndo_get_vf_port)(struct net_device *dev, 1430 int vf, struct sk_buff *skb); 1431 int (*ndo_get_vf_guid)(struct net_device *dev, 1432 int vf, 1433 struct ifla_vf_guid *node_guid, 1434 struct ifla_vf_guid *port_guid); 1435 int (*ndo_set_vf_guid)(struct net_device *dev, 1436 int vf, u64 guid, 1437 int guid_type); 1438 int (*ndo_set_vf_rss_query_en)( 1439 struct net_device *dev, 1440 int vf, bool setting); 1441 int (*ndo_setup_tc)(struct net_device *dev, 1442 enum tc_setup_type type, 1443 void *type_data); 1444 #if IS_ENABLED(CONFIG_FCOE) 1445 int (*ndo_fcoe_enable)(struct net_device *dev); 1446 int (*ndo_fcoe_disable)(struct net_device *dev); 1447 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1448 u16 xid, 1449 struct scatterlist *sgl, 1450 unsigned int sgc); 1451 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1452 u16 xid); 1453 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1454 u16 xid, 1455 struct scatterlist *sgl, 1456 unsigned int sgc); 1457 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1458 struct netdev_fcoe_hbainfo *hbainfo); 1459 #endif 1460 1461 #if IS_ENABLED(CONFIG_LIBFCOE) 1462 #define NETDEV_FCOE_WWNN 0 1463 #define NETDEV_FCOE_WWPN 1 1464 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1465 u64 *wwn, int type); 1466 #endif 1467 1468 #ifdef CONFIG_RFS_ACCEL 1469 int (*ndo_rx_flow_steer)(struct net_device *dev, 1470 const struct sk_buff *skb, 1471 u16 rxq_index, 1472 u32 flow_id); 1473 #endif 1474 int (*ndo_add_slave)(struct net_device *dev, 1475 struct net_device *slave_dev, 1476 struct netlink_ext_ack *extack); 1477 int (*ndo_del_slave)(struct net_device *dev, 1478 struct net_device *slave_dev); 1479 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1480 struct sk_buff *skb, 1481 bool all_slaves); 1482 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1483 struct sock *sk); 1484 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1485 netdev_features_t features); 1486 int (*ndo_set_features)(struct net_device *dev, 1487 netdev_features_t features); 1488 int (*ndo_neigh_construct)(struct net_device *dev, 1489 struct neighbour *n); 1490 void (*ndo_neigh_destroy)(struct net_device *dev, 1491 struct neighbour *n); 1492 1493 int (*ndo_fdb_add)(struct ndmsg *ndm, 1494 struct nlattr *tb[], 1495 struct net_device *dev, 1496 const unsigned char *addr, 1497 u16 vid, 1498 u16 flags, 1499 struct netlink_ext_ack *extack); 1500 int (*ndo_fdb_del)(struct ndmsg *ndm, 1501 struct nlattr *tb[], 1502 struct net_device *dev, 1503 const unsigned char *addr, 1504 u16 vid); 1505 int (*ndo_fdb_dump)(struct sk_buff *skb, 1506 struct netlink_callback *cb, 1507 struct net_device *dev, 1508 struct net_device *filter_dev, 1509 int *idx); 1510 int (*ndo_fdb_get)(struct sk_buff *skb, 1511 struct nlattr *tb[], 1512 struct net_device *dev, 1513 const unsigned char *addr, 1514 u16 vid, u32 portid, u32 seq, 1515 struct netlink_ext_ack *extack); 1516 int (*ndo_bridge_setlink)(struct net_device *dev, 1517 struct nlmsghdr *nlh, 1518 u16 flags, 1519 struct netlink_ext_ack *extack); 1520 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1521 u32 pid, u32 seq, 1522 struct net_device *dev, 1523 u32 filter_mask, 1524 int nlflags); 1525 int (*ndo_bridge_dellink)(struct net_device *dev, 1526 struct nlmsghdr *nlh, 1527 u16 flags); 1528 int (*ndo_change_carrier)(struct net_device *dev, 1529 bool new_carrier); 1530 int (*ndo_get_phys_port_id)(struct net_device *dev, 1531 struct netdev_phys_item_id *ppid); 1532 int (*ndo_get_port_parent_id)(struct net_device *dev, 1533 struct netdev_phys_item_id *ppid); 1534 int (*ndo_get_phys_port_name)(struct net_device *dev, 1535 char *name, size_t len); 1536 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1537 struct net_device *dev); 1538 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1539 void *priv); 1540 1541 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1542 int queue_index, 1543 u32 maxrate); 1544 int (*ndo_get_iflink)(const struct net_device *dev); 1545 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1546 struct sk_buff *skb); 1547 void (*ndo_set_rx_headroom)(struct net_device *dev, 1548 int needed_headroom); 1549 int (*ndo_bpf)(struct net_device *dev, 1550 struct netdev_bpf *bpf); 1551 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1552 struct xdp_frame **xdp, 1553 u32 flags); 1554 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1555 struct xdp_buff *xdp); 1556 int (*ndo_xsk_wakeup)(struct net_device *dev, 1557 u32 queue_id, u32 flags); 1558 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1559 int (*ndo_tunnel_ctl)(struct net_device *dev, 1560 struct ip_tunnel_parm *p, int cmd); 1561 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1562 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1563 struct net_device_path *path); 1564 }; 1565 1566 /** 1567 * enum netdev_priv_flags - &struct net_device priv_flags 1568 * 1569 * These are the &struct net_device, they are only set internally 1570 * by drivers and used in the kernel. These flags are invisible to 1571 * userspace; this means that the order of these flags can change 1572 * during any kernel release. 1573 * 1574 * You should have a pretty good reason to be extending these flags. 1575 * 1576 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1577 * @IFF_EBRIDGE: Ethernet bridging device 1578 * @IFF_BONDING: bonding master or slave 1579 * @IFF_ISATAP: ISATAP interface (RFC4214) 1580 * @IFF_WAN_HDLC: WAN HDLC device 1581 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1582 * release skb->dst 1583 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1584 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1585 * @IFF_MACVLAN_PORT: device used as macvlan port 1586 * @IFF_BRIDGE_PORT: device used as bridge port 1587 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1588 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1589 * @IFF_UNICAST_FLT: Supports unicast filtering 1590 * @IFF_TEAM_PORT: device used as team port 1591 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1592 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1593 * change when it's running 1594 * @IFF_MACVLAN: Macvlan device 1595 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1596 * underlying stacked devices 1597 * @IFF_L3MDEV_MASTER: device is an L3 master device 1598 * @IFF_NO_QUEUE: device can run without qdisc attached 1599 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1600 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1601 * @IFF_TEAM: device is a team device 1602 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1603 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1604 * entity (i.e. the master device for bridged veth) 1605 * @IFF_MACSEC: device is a MACsec device 1606 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1607 * @IFF_FAILOVER: device is a failover master device 1608 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1609 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1610 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1611 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1612 * skb_headlen(skb) == 0 (data starts from frag0) 1613 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN 1614 */ 1615 enum netdev_priv_flags { 1616 IFF_802_1Q_VLAN = 1<<0, 1617 IFF_EBRIDGE = 1<<1, 1618 IFF_BONDING = 1<<2, 1619 IFF_ISATAP = 1<<3, 1620 IFF_WAN_HDLC = 1<<4, 1621 IFF_XMIT_DST_RELEASE = 1<<5, 1622 IFF_DONT_BRIDGE = 1<<6, 1623 IFF_DISABLE_NETPOLL = 1<<7, 1624 IFF_MACVLAN_PORT = 1<<8, 1625 IFF_BRIDGE_PORT = 1<<9, 1626 IFF_OVS_DATAPATH = 1<<10, 1627 IFF_TX_SKB_SHARING = 1<<11, 1628 IFF_UNICAST_FLT = 1<<12, 1629 IFF_TEAM_PORT = 1<<13, 1630 IFF_SUPP_NOFCS = 1<<14, 1631 IFF_LIVE_ADDR_CHANGE = 1<<15, 1632 IFF_MACVLAN = 1<<16, 1633 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1634 IFF_L3MDEV_MASTER = 1<<18, 1635 IFF_NO_QUEUE = 1<<19, 1636 IFF_OPENVSWITCH = 1<<20, 1637 IFF_L3MDEV_SLAVE = 1<<21, 1638 IFF_TEAM = 1<<22, 1639 IFF_RXFH_CONFIGURED = 1<<23, 1640 IFF_PHONY_HEADROOM = 1<<24, 1641 IFF_MACSEC = 1<<25, 1642 IFF_NO_RX_HANDLER = 1<<26, 1643 IFF_FAILOVER = 1<<27, 1644 IFF_FAILOVER_SLAVE = 1<<28, 1645 IFF_L3MDEV_RX_HANDLER = 1<<29, 1646 IFF_LIVE_RENAME_OK = 1<<30, 1647 IFF_TX_SKB_NO_LINEAR = 1<<31, 1648 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32), 1649 }; 1650 1651 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1652 #define IFF_EBRIDGE IFF_EBRIDGE 1653 #define IFF_BONDING IFF_BONDING 1654 #define IFF_ISATAP IFF_ISATAP 1655 #define IFF_WAN_HDLC IFF_WAN_HDLC 1656 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1657 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1658 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1659 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1660 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1661 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1662 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1663 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1664 #define IFF_TEAM_PORT IFF_TEAM_PORT 1665 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1666 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1667 #define IFF_MACVLAN IFF_MACVLAN 1668 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1669 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1670 #define IFF_NO_QUEUE IFF_NO_QUEUE 1671 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1672 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1673 #define IFF_TEAM IFF_TEAM 1674 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1675 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM 1676 #define IFF_MACSEC IFF_MACSEC 1677 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1678 #define IFF_FAILOVER IFF_FAILOVER 1679 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1680 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1681 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1682 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR 1683 1684 /* Specifies the type of the struct net_device::ml_priv pointer */ 1685 enum netdev_ml_priv_type { 1686 ML_PRIV_NONE, 1687 ML_PRIV_CAN, 1688 }; 1689 1690 /** 1691 * struct net_device - The DEVICE structure. 1692 * 1693 * Actually, this whole structure is a big mistake. It mixes I/O 1694 * data with strictly "high-level" data, and it has to know about 1695 * almost every data structure used in the INET module. 1696 * 1697 * @name: This is the first field of the "visible" part of this structure 1698 * (i.e. as seen by users in the "Space.c" file). It is the name 1699 * of the interface. 1700 * 1701 * @name_node: Name hashlist node 1702 * @ifalias: SNMP alias 1703 * @mem_end: Shared memory end 1704 * @mem_start: Shared memory start 1705 * @base_addr: Device I/O address 1706 * @irq: Device IRQ number 1707 * 1708 * @state: Generic network queuing layer state, see netdev_state_t 1709 * @dev_list: The global list of network devices 1710 * @napi_list: List entry used for polling NAPI devices 1711 * @unreg_list: List entry when we are unregistering the 1712 * device; see the function unregister_netdev 1713 * @close_list: List entry used when we are closing the device 1714 * @ptype_all: Device-specific packet handlers for all protocols 1715 * @ptype_specific: Device-specific, protocol-specific packet handlers 1716 * 1717 * @adj_list: Directly linked devices, like slaves for bonding 1718 * @features: Currently active device features 1719 * @hw_features: User-changeable features 1720 * 1721 * @wanted_features: User-requested features 1722 * @vlan_features: Mask of features inheritable by VLAN devices 1723 * 1724 * @hw_enc_features: Mask of features inherited by encapsulating devices 1725 * This field indicates what encapsulation 1726 * offloads the hardware is capable of doing, 1727 * and drivers will need to set them appropriately. 1728 * 1729 * @mpls_features: Mask of features inheritable by MPLS 1730 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1731 * 1732 * @ifindex: interface index 1733 * @group: The group the device belongs to 1734 * 1735 * @stats: Statistics struct, which was left as a legacy, use 1736 * rtnl_link_stats64 instead 1737 * 1738 * @rx_dropped: Dropped packets by core network, 1739 * do not use this in drivers 1740 * @tx_dropped: Dropped packets by core network, 1741 * do not use this in drivers 1742 * @rx_nohandler: nohandler dropped packets by core network on 1743 * inactive devices, do not use this in drivers 1744 * @carrier_up_count: Number of times the carrier has been up 1745 * @carrier_down_count: Number of times the carrier has been down 1746 * 1747 * @wireless_handlers: List of functions to handle Wireless Extensions, 1748 * instead of ioctl, 1749 * see <net/iw_handler.h> for details. 1750 * @wireless_data: Instance data managed by the core of wireless extensions 1751 * 1752 * @netdev_ops: Includes several pointers to callbacks, 1753 * if one wants to override the ndo_*() functions 1754 * @ethtool_ops: Management operations 1755 * @l3mdev_ops: Layer 3 master device operations 1756 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1757 * discovery handling. Necessary for e.g. 6LoWPAN. 1758 * @xfrmdev_ops: Transformation offload operations 1759 * @tlsdev_ops: Transport Layer Security offload operations 1760 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1761 * of Layer 2 headers. 1762 * 1763 * @flags: Interface flags (a la BSD) 1764 * @priv_flags: Like 'flags' but invisible to userspace, 1765 * see if.h for the definitions 1766 * @gflags: Global flags ( kept as legacy ) 1767 * @padded: How much padding added by alloc_netdev() 1768 * @operstate: RFC2863 operstate 1769 * @link_mode: Mapping policy to operstate 1770 * @if_port: Selectable AUI, TP, ... 1771 * @dma: DMA channel 1772 * @mtu: Interface MTU value 1773 * @min_mtu: Interface Minimum MTU value 1774 * @max_mtu: Interface Maximum MTU value 1775 * @type: Interface hardware type 1776 * @hard_header_len: Maximum hardware header length. 1777 * @min_header_len: Minimum hardware header length 1778 * 1779 * @needed_headroom: Extra headroom the hardware may need, but not in all 1780 * cases can this be guaranteed 1781 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1782 * cases can this be guaranteed. Some cases also use 1783 * LL_MAX_HEADER instead to allocate the skb 1784 * 1785 * interface address info: 1786 * 1787 * @perm_addr: Permanent hw address 1788 * @addr_assign_type: Hw address assignment type 1789 * @addr_len: Hardware address length 1790 * @upper_level: Maximum depth level of upper devices. 1791 * @lower_level: Maximum depth level of lower devices. 1792 * @neigh_priv_len: Used in neigh_alloc() 1793 * @dev_id: Used to differentiate devices that share 1794 * the same link layer address 1795 * @dev_port: Used to differentiate devices that share 1796 * the same function 1797 * @addr_list_lock: XXX: need comments on this one 1798 * @name_assign_type: network interface name assignment type 1799 * @uc_promisc: Counter that indicates promiscuous mode 1800 * has been enabled due to the need to listen to 1801 * additional unicast addresses in a device that 1802 * does not implement ndo_set_rx_mode() 1803 * @uc: unicast mac addresses 1804 * @mc: multicast mac addresses 1805 * @dev_addrs: list of device hw addresses 1806 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1807 * @promiscuity: Number of times the NIC is told to work in 1808 * promiscuous mode; if it becomes 0 the NIC will 1809 * exit promiscuous mode 1810 * @allmulti: Counter, enables or disables allmulticast mode 1811 * 1812 * @vlan_info: VLAN info 1813 * @dsa_ptr: dsa specific data 1814 * @tipc_ptr: TIPC specific data 1815 * @atalk_ptr: AppleTalk link 1816 * @ip_ptr: IPv4 specific data 1817 * @dn_ptr: DECnet specific data 1818 * @ip6_ptr: IPv6 specific data 1819 * @ax25_ptr: AX.25 specific data 1820 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1821 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1822 * device struct 1823 * @mpls_ptr: mpls_dev struct pointer 1824 * @mctp_ptr: MCTP specific data 1825 * 1826 * @dev_addr: Hw address (before bcast, 1827 * because most packets are unicast) 1828 * 1829 * @_rx: Array of RX queues 1830 * @num_rx_queues: Number of RX queues 1831 * allocated at register_netdev() time 1832 * @real_num_rx_queues: Number of RX queues currently active in device 1833 * @xdp_prog: XDP sockets filter program pointer 1834 * @gro_flush_timeout: timeout for GRO layer in NAPI 1835 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1836 * allow to avoid NIC hard IRQ, on busy queues. 1837 * 1838 * @rx_handler: handler for received packets 1839 * @rx_handler_data: XXX: need comments on this one 1840 * @miniq_ingress: ingress/clsact qdisc specific data for 1841 * ingress processing 1842 * @ingress_queue: XXX: need comments on this one 1843 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1844 * @broadcast: hw bcast address 1845 * 1846 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1847 * indexed by RX queue number. Assigned by driver. 1848 * This must only be set if the ndo_rx_flow_steer 1849 * operation is defined 1850 * @index_hlist: Device index hash chain 1851 * 1852 * @_tx: Array of TX queues 1853 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1854 * @real_num_tx_queues: Number of TX queues currently active in device 1855 * @qdisc: Root qdisc from userspace point of view 1856 * @tx_queue_len: Max frames per queue allowed 1857 * @tx_global_lock: XXX: need comments on this one 1858 * @xdp_bulkq: XDP device bulk queue 1859 * @xps_maps: all CPUs/RXQs maps for XPS device 1860 * 1861 * @xps_maps: XXX: need comments on this one 1862 * @miniq_egress: clsact qdisc specific data for 1863 * egress processing 1864 * @nf_hooks_egress: netfilter hooks executed for egress packets 1865 * @qdisc_hash: qdisc hash table 1866 * @watchdog_timeo: Represents the timeout that is used by 1867 * the watchdog (see dev_watchdog()) 1868 * @watchdog_timer: List of timers 1869 * 1870 * @proto_down_reason: reason a netdev interface is held down 1871 * @pcpu_refcnt: Number of references to this device 1872 * @dev_refcnt: Number of references to this device 1873 * @refcnt_tracker: Tracker directory for tracked references to this device 1874 * @todo_list: Delayed register/unregister 1875 * @link_watch_list: XXX: need comments on this one 1876 * 1877 * @reg_state: Register/unregister state machine 1878 * @dismantle: Device is going to be freed 1879 * @rtnl_link_state: This enum represents the phases of creating 1880 * a new link 1881 * 1882 * @needs_free_netdev: Should unregister perform free_netdev? 1883 * @priv_destructor: Called from unregister 1884 * @npinfo: XXX: need comments on this one 1885 * @nd_net: Network namespace this network device is inside 1886 * 1887 * @ml_priv: Mid-layer private 1888 * @ml_priv_type: Mid-layer private type 1889 * @lstats: Loopback statistics 1890 * @tstats: Tunnel statistics 1891 * @dstats: Dummy statistics 1892 * @vstats: Virtual ethernet statistics 1893 * 1894 * @garp_port: GARP 1895 * @mrp_port: MRP 1896 * 1897 * @dev: Class/net/name entry 1898 * @sysfs_groups: Space for optional device, statistics and wireless 1899 * sysfs groups 1900 * 1901 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1902 * @rtnl_link_ops: Rtnl_link_ops 1903 * 1904 * @gso_max_size: Maximum size of generic segmentation offload 1905 * @gso_max_segs: Maximum number of segments that can be passed to the 1906 * NIC for GSO 1907 * 1908 * @dcbnl_ops: Data Center Bridging netlink ops 1909 * @num_tc: Number of traffic classes in the net device 1910 * @tc_to_txq: XXX: need comments on this one 1911 * @prio_tc_map: XXX: need comments on this one 1912 * 1913 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1914 * 1915 * @priomap: XXX: need comments on this one 1916 * @phydev: Physical device may attach itself 1917 * for hardware timestamping 1918 * @sfp_bus: attached &struct sfp_bus structure. 1919 * 1920 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1921 * 1922 * @proto_down: protocol port state information can be sent to the 1923 * switch driver and used to set the phys state of the 1924 * switch port. 1925 * 1926 * @wol_enabled: Wake-on-LAN is enabled 1927 * 1928 * @threaded: napi threaded mode is enabled 1929 * 1930 * @net_notifier_list: List of per-net netdev notifier block 1931 * that follow this device when it is moved 1932 * to another network namespace. 1933 * 1934 * @macsec_ops: MACsec offloading ops 1935 * 1936 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1937 * offload capabilities of the device 1938 * @udp_tunnel_nic: UDP tunnel offload state 1939 * @xdp_state: stores info on attached XDP BPF programs 1940 * 1941 * @nested_level: Used as a parameter of spin_lock_nested() of 1942 * dev->addr_list_lock. 1943 * @unlink_list: As netif_addr_lock() can be called recursively, 1944 * keep a list of interfaces to be deleted. 1945 * @gro_max_size: Maximum size of aggregated packet in generic 1946 * receive offload (GRO) 1947 * 1948 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 1949 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 1950 * @watchdog_dev_tracker: refcount tracker used by watchdog. 1951 * @dev_registered_tracker: tracker for reference held while 1952 * registered 1953 * @offload_xstats_l3: L3 HW stats for this netdevice. 1954 * 1955 * FIXME: cleanup struct net_device such that network protocol info 1956 * moves out. 1957 */ 1958 1959 struct net_device { 1960 char name[IFNAMSIZ]; 1961 struct netdev_name_node *name_node; 1962 struct dev_ifalias __rcu *ifalias; 1963 /* 1964 * I/O specific fields 1965 * FIXME: Merge these and struct ifmap into one 1966 */ 1967 unsigned long mem_end; 1968 unsigned long mem_start; 1969 unsigned long base_addr; 1970 1971 /* 1972 * Some hardware also needs these fields (state,dev_list, 1973 * napi_list,unreg_list,close_list) but they are not 1974 * part of the usual set specified in Space.c. 1975 */ 1976 1977 unsigned long state; 1978 1979 struct list_head dev_list; 1980 struct list_head napi_list; 1981 struct list_head unreg_list; 1982 struct list_head close_list; 1983 struct list_head ptype_all; 1984 struct list_head ptype_specific; 1985 1986 struct { 1987 struct list_head upper; 1988 struct list_head lower; 1989 } adj_list; 1990 1991 /* Read-mostly cache-line for fast-path access */ 1992 unsigned int flags; 1993 unsigned long long priv_flags; 1994 const struct net_device_ops *netdev_ops; 1995 int ifindex; 1996 unsigned short gflags; 1997 unsigned short hard_header_len; 1998 1999 /* Note : dev->mtu is often read without holding a lock. 2000 * Writers usually hold RTNL. 2001 * It is recommended to use READ_ONCE() to annotate the reads, 2002 * and to use WRITE_ONCE() to annotate the writes. 2003 */ 2004 unsigned int mtu; 2005 unsigned short needed_headroom; 2006 unsigned short needed_tailroom; 2007 2008 netdev_features_t features; 2009 netdev_features_t hw_features; 2010 netdev_features_t wanted_features; 2011 netdev_features_t vlan_features; 2012 netdev_features_t hw_enc_features; 2013 netdev_features_t mpls_features; 2014 netdev_features_t gso_partial_features; 2015 2016 unsigned int min_mtu; 2017 unsigned int max_mtu; 2018 unsigned short type; 2019 unsigned char min_header_len; 2020 unsigned char name_assign_type; 2021 2022 int group; 2023 2024 struct net_device_stats stats; /* not used by modern drivers */ 2025 2026 atomic_long_t rx_dropped; 2027 atomic_long_t tx_dropped; 2028 atomic_long_t rx_nohandler; 2029 2030 /* Stats to monitor link on/off, flapping */ 2031 atomic_t carrier_up_count; 2032 atomic_t carrier_down_count; 2033 2034 #ifdef CONFIG_WIRELESS_EXT 2035 const struct iw_handler_def *wireless_handlers; 2036 struct iw_public_data *wireless_data; 2037 #endif 2038 const struct ethtool_ops *ethtool_ops; 2039 #ifdef CONFIG_NET_L3_MASTER_DEV 2040 const struct l3mdev_ops *l3mdev_ops; 2041 #endif 2042 #if IS_ENABLED(CONFIG_IPV6) 2043 const struct ndisc_ops *ndisc_ops; 2044 #endif 2045 2046 #ifdef CONFIG_XFRM_OFFLOAD 2047 const struct xfrmdev_ops *xfrmdev_ops; 2048 #endif 2049 2050 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2051 const struct tlsdev_ops *tlsdev_ops; 2052 #endif 2053 2054 const struct header_ops *header_ops; 2055 2056 unsigned char operstate; 2057 unsigned char link_mode; 2058 2059 unsigned char if_port; 2060 unsigned char dma; 2061 2062 /* Interface address info. */ 2063 unsigned char perm_addr[MAX_ADDR_LEN]; 2064 unsigned char addr_assign_type; 2065 unsigned char addr_len; 2066 unsigned char upper_level; 2067 unsigned char lower_level; 2068 2069 unsigned short neigh_priv_len; 2070 unsigned short dev_id; 2071 unsigned short dev_port; 2072 unsigned short padded; 2073 2074 spinlock_t addr_list_lock; 2075 int irq; 2076 2077 struct netdev_hw_addr_list uc; 2078 struct netdev_hw_addr_list mc; 2079 struct netdev_hw_addr_list dev_addrs; 2080 2081 #ifdef CONFIG_SYSFS 2082 struct kset *queues_kset; 2083 #endif 2084 #ifdef CONFIG_LOCKDEP 2085 struct list_head unlink_list; 2086 #endif 2087 unsigned int promiscuity; 2088 unsigned int allmulti; 2089 bool uc_promisc; 2090 #ifdef CONFIG_LOCKDEP 2091 unsigned char nested_level; 2092 #endif 2093 2094 2095 /* Protocol-specific pointers */ 2096 2097 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2098 struct vlan_info __rcu *vlan_info; 2099 #endif 2100 #if IS_ENABLED(CONFIG_NET_DSA) 2101 struct dsa_port *dsa_ptr; 2102 #endif 2103 #if IS_ENABLED(CONFIG_TIPC) 2104 struct tipc_bearer __rcu *tipc_ptr; 2105 #endif 2106 #if IS_ENABLED(CONFIG_ATALK) 2107 void *atalk_ptr; 2108 #endif 2109 struct in_device __rcu *ip_ptr; 2110 #if IS_ENABLED(CONFIG_DECNET) 2111 struct dn_dev __rcu *dn_ptr; 2112 #endif 2113 struct inet6_dev __rcu *ip6_ptr; 2114 #if IS_ENABLED(CONFIG_AX25) 2115 void *ax25_ptr; 2116 #endif 2117 struct wireless_dev *ieee80211_ptr; 2118 struct wpan_dev *ieee802154_ptr; 2119 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2120 struct mpls_dev __rcu *mpls_ptr; 2121 #endif 2122 #if IS_ENABLED(CONFIG_MCTP) 2123 struct mctp_dev __rcu *mctp_ptr; 2124 #endif 2125 2126 /* 2127 * Cache lines mostly used on receive path (including eth_type_trans()) 2128 */ 2129 /* Interface address info used in eth_type_trans() */ 2130 const unsigned char *dev_addr; 2131 2132 struct netdev_rx_queue *_rx; 2133 unsigned int num_rx_queues; 2134 unsigned int real_num_rx_queues; 2135 2136 struct bpf_prog __rcu *xdp_prog; 2137 unsigned long gro_flush_timeout; 2138 int napi_defer_hard_irqs; 2139 #define GRO_MAX_SIZE 65536 2140 unsigned int gro_max_size; 2141 rx_handler_func_t __rcu *rx_handler; 2142 void __rcu *rx_handler_data; 2143 2144 #ifdef CONFIG_NET_CLS_ACT 2145 struct mini_Qdisc __rcu *miniq_ingress; 2146 #endif 2147 struct netdev_queue __rcu *ingress_queue; 2148 #ifdef CONFIG_NETFILTER_INGRESS 2149 struct nf_hook_entries __rcu *nf_hooks_ingress; 2150 #endif 2151 2152 unsigned char broadcast[MAX_ADDR_LEN]; 2153 #ifdef CONFIG_RFS_ACCEL 2154 struct cpu_rmap *rx_cpu_rmap; 2155 #endif 2156 struct hlist_node index_hlist; 2157 2158 /* 2159 * Cache lines mostly used on transmit path 2160 */ 2161 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2162 unsigned int num_tx_queues; 2163 unsigned int real_num_tx_queues; 2164 struct Qdisc __rcu *qdisc; 2165 unsigned int tx_queue_len; 2166 spinlock_t tx_global_lock; 2167 2168 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2169 2170 #ifdef CONFIG_XPS 2171 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2172 #endif 2173 #ifdef CONFIG_NET_CLS_ACT 2174 struct mini_Qdisc __rcu *miniq_egress; 2175 #endif 2176 #ifdef CONFIG_NETFILTER_EGRESS 2177 struct nf_hook_entries __rcu *nf_hooks_egress; 2178 #endif 2179 2180 #ifdef CONFIG_NET_SCHED 2181 DECLARE_HASHTABLE (qdisc_hash, 4); 2182 #endif 2183 /* These may be needed for future network-power-down code. */ 2184 struct timer_list watchdog_timer; 2185 int watchdog_timeo; 2186 2187 u32 proto_down_reason; 2188 2189 struct list_head todo_list; 2190 2191 #ifdef CONFIG_PCPU_DEV_REFCNT 2192 int __percpu *pcpu_refcnt; 2193 #else 2194 refcount_t dev_refcnt; 2195 #endif 2196 struct ref_tracker_dir refcnt_tracker; 2197 2198 struct list_head link_watch_list; 2199 2200 enum { NETREG_UNINITIALIZED=0, 2201 NETREG_REGISTERED, /* completed register_netdevice */ 2202 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2203 NETREG_UNREGISTERED, /* completed unregister todo */ 2204 NETREG_RELEASED, /* called free_netdev */ 2205 NETREG_DUMMY, /* dummy device for NAPI poll */ 2206 } reg_state:8; 2207 2208 bool dismantle; 2209 2210 enum { 2211 RTNL_LINK_INITIALIZED, 2212 RTNL_LINK_INITIALIZING, 2213 } rtnl_link_state:16; 2214 2215 bool needs_free_netdev; 2216 void (*priv_destructor)(struct net_device *dev); 2217 2218 #ifdef CONFIG_NETPOLL 2219 struct netpoll_info __rcu *npinfo; 2220 #endif 2221 2222 possible_net_t nd_net; 2223 2224 /* mid-layer private */ 2225 void *ml_priv; 2226 enum netdev_ml_priv_type ml_priv_type; 2227 2228 union { 2229 struct pcpu_lstats __percpu *lstats; 2230 struct pcpu_sw_netstats __percpu *tstats; 2231 struct pcpu_dstats __percpu *dstats; 2232 }; 2233 2234 #if IS_ENABLED(CONFIG_GARP) 2235 struct garp_port __rcu *garp_port; 2236 #endif 2237 #if IS_ENABLED(CONFIG_MRP) 2238 struct mrp_port __rcu *mrp_port; 2239 #endif 2240 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2241 struct dm_hw_stat_delta __rcu *dm_private; 2242 #endif 2243 struct device dev; 2244 const struct attribute_group *sysfs_groups[4]; 2245 const struct attribute_group *sysfs_rx_queue_group; 2246 2247 const struct rtnl_link_ops *rtnl_link_ops; 2248 2249 /* for setting kernel sock attribute on TCP connection setup */ 2250 #define GSO_MAX_SIZE 65536 2251 unsigned int gso_max_size; 2252 #define GSO_MAX_SEGS 65535 2253 u16 gso_max_segs; 2254 2255 #ifdef CONFIG_DCB 2256 const struct dcbnl_rtnl_ops *dcbnl_ops; 2257 #endif 2258 s16 num_tc; 2259 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2260 u8 prio_tc_map[TC_BITMASK + 1]; 2261 2262 #if IS_ENABLED(CONFIG_FCOE) 2263 unsigned int fcoe_ddp_xid; 2264 #endif 2265 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2266 struct netprio_map __rcu *priomap; 2267 #endif 2268 struct phy_device *phydev; 2269 struct sfp_bus *sfp_bus; 2270 struct lock_class_key *qdisc_tx_busylock; 2271 bool proto_down; 2272 unsigned wol_enabled:1; 2273 unsigned threaded:1; 2274 2275 struct list_head net_notifier_list; 2276 2277 #if IS_ENABLED(CONFIG_MACSEC) 2278 /* MACsec management functions */ 2279 const struct macsec_ops *macsec_ops; 2280 #endif 2281 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2282 struct udp_tunnel_nic *udp_tunnel_nic; 2283 2284 /* protected by rtnl_lock */ 2285 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2286 2287 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2288 netdevice_tracker linkwatch_dev_tracker; 2289 netdevice_tracker watchdog_dev_tracker; 2290 netdevice_tracker dev_registered_tracker; 2291 struct rtnl_hw_stats64 *offload_xstats_l3; 2292 }; 2293 #define to_net_dev(d) container_of(d, struct net_device, dev) 2294 2295 static inline bool netif_elide_gro(const struct net_device *dev) 2296 { 2297 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2298 return true; 2299 return false; 2300 } 2301 2302 #define NETDEV_ALIGN 32 2303 2304 static inline 2305 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2306 { 2307 return dev->prio_tc_map[prio & TC_BITMASK]; 2308 } 2309 2310 static inline 2311 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2312 { 2313 if (tc >= dev->num_tc) 2314 return -EINVAL; 2315 2316 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2317 return 0; 2318 } 2319 2320 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2321 void netdev_reset_tc(struct net_device *dev); 2322 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2323 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2324 2325 static inline 2326 int netdev_get_num_tc(struct net_device *dev) 2327 { 2328 return dev->num_tc; 2329 } 2330 2331 static inline void net_prefetch(void *p) 2332 { 2333 prefetch(p); 2334 #if L1_CACHE_BYTES < 128 2335 prefetch((u8 *)p + L1_CACHE_BYTES); 2336 #endif 2337 } 2338 2339 static inline void net_prefetchw(void *p) 2340 { 2341 prefetchw(p); 2342 #if L1_CACHE_BYTES < 128 2343 prefetchw((u8 *)p + L1_CACHE_BYTES); 2344 #endif 2345 } 2346 2347 void netdev_unbind_sb_channel(struct net_device *dev, 2348 struct net_device *sb_dev); 2349 int netdev_bind_sb_channel_queue(struct net_device *dev, 2350 struct net_device *sb_dev, 2351 u8 tc, u16 count, u16 offset); 2352 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2353 static inline int netdev_get_sb_channel(struct net_device *dev) 2354 { 2355 return max_t(int, -dev->num_tc, 0); 2356 } 2357 2358 static inline 2359 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2360 unsigned int index) 2361 { 2362 return &dev->_tx[index]; 2363 } 2364 2365 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2366 const struct sk_buff *skb) 2367 { 2368 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2369 } 2370 2371 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2372 void (*f)(struct net_device *, 2373 struct netdev_queue *, 2374 void *), 2375 void *arg) 2376 { 2377 unsigned int i; 2378 2379 for (i = 0; i < dev->num_tx_queues; i++) 2380 f(dev, &dev->_tx[i], arg); 2381 } 2382 2383 #define netdev_lockdep_set_classes(dev) \ 2384 { \ 2385 static struct lock_class_key qdisc_tx_busylock_key; \ 2386 static struct lock_class_key qdisc_xmit_lock_key; \ 2387 static struct lock_class_key dev_addr_list_lock_key; \ 2388 unsigned int i; \ 2389 \ 2390 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2391 lockdep_set_class(&(dev)->addr_list_lock, \ 2392 &dev_addr_list_lock_key); \ 2393 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2394 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2395 &qdisc_xmit_lock_key); \ 2396 } 2397 2398 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2399 struct net_device *sb_dev); 2400 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2401 struct sk_buff *skb, 2402 struct net_device *sb_dev); 2403 2404 /* returns the headroom that the master device needs to take in account 2405 * when forwarding to this dev 2406 */ 2407 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2408 { 2409 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2410 } 2411 2412 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2413 { 2414 if (dev->netdev_ops->ndo_set_rx_headroom) 2415 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2416 } 2417 2418 /* set the device rx headroom to the dev's default */ 2419 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2420 { 2421 netdev_set_rx_headroom(dev, -1); 2422 } 2423 2424 static inline void *netdev_get_ml_priv(struct net_device *dev, 2425 enum netdev_ml_priv_type type) 2426 { 2427 if (dev->ml_priv_type != type) 2428 return NULL; 2429 2430 return dev->ml_priv; 2431 } 2432 2433 static inline void netdev_set_ml_priv(struct net_device *dev, 2434 void *ml_priv, 2435 enum netdev_ml_priv_type type) 2436 { 2437 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2438 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2439 dev->ml_priv_type, type); 2440 WARN(!dev->ml_priv_type && dev->ml_priv, 2441 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2442 2443 dev->ml_priv = ml_priv; 2444 dev->ml_priv_type = type; 2445 } 2446 2447 /* 2448 * Net namespace inlines 2449 */ 2450 static inline 2451 struct net *dev_net(const struct net_device *dev) 2452 { 2453 return read_pnet(&dev->nd_net); 2454 } 2455 2456 static inline 2457 void dev_net_set(struct net_device *dev, struct net *net) 2458 { 2459 write_pnet(&dev->nd_net, net); 2460 } 2461 2462 /** 2463 * netdev_priv - access network device private data 2464 * @dev: network device 2465 * 2466 * Get network device private data 2467 */ 2468 static inline void *netdev_priv(const struct net_device *dev) 2469 { 2470 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2471 } 2472 2473 /* Set the sysfs physical device reference for the network logical device 2474 * if set prior to registration will cause a symlink during initialization. 2475 */ 2476 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2477 2478 /* Set the sysfs device type for the network logical device to allow 2479 * fine-grained identification of different network device types. For 2480 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2481 */ 2482 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2483 2484 /* Default NAPI poll() weight 2485 * Device drivers are strongly advised to not use bigger value 2486 */ 2487 #define NAPI_POLL_WEIGHT 64 2488 2489 /** 2490 * netif_napi_add - initialize a NAPI context 2491 * @dev: network device 2492 * @napi: NAPI context 2493 * @poll: polling function 2494 * @weight: default weight 2495 * 2496 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2497 * *any* of the other NAPI-related functions. 2498 */ 2499 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2500 int (*poll)(struct napi_struct *, int), int weight); 2501 2502 /** 2503 * netif_tx_napi_add - initialize a NAPI context 2504 * @dev: network device 2505 * @napi: NAPI context 2506 * @poll: polling function 2507 * @weight: default weight 2508 * 2509 * This variant of netif_napi_add() should be used from drivers using NAPI 2510 * to exclusively poll a TX queue. 2511 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2512 */ 2513 static inline void netif_tx_napi_add(struct net_device *dev, 2514 struct napi_struct *napi, 2515 int (*poll)(struct napi_struct *, int), 2516 int weight) 2517 { 2518 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2519 netif_napi_add(dev, napi, poll, weight); 2520 } 2521 2522 /** 2523 * __netif_napi_del - remove a NAPI context 2524 * @napi: NAPI context 2525 * 2526 * Warning: caller must observe RCU grace period before freeing memory 2527 * containing @napi. Drivers might want to call this helper to combine 2528 * all the needed RCU grace periods into a single one. 2529 */ 2530 void __netif_napi_del(struct napi_struct *napi); 2531 2532 /** 2533 * netif_napi_del - remove a NAPI context 2534 * @napi: NAPI context 2535 * 2536 * netif_napi_del() removes a NAPI context from the network device NAPI list 2537 */ 2538 static inline void netif_napi_del(struct napi_struct *napi) 2539 { 2540 __netif_napi_del(napi); 2541 synchronize_net(); 2542 } 2543 2544 struct packet_type { 2545 __be16 type; /* This is really htons(ether_type). */ 2546 bool ignore_outgoing; 2547 struct net_device *dev; /* NULL is wildcarded here */ 2548 netdevice_tracker dev_tracker; 2549 int (*func) (struct sk_buff *, 2550 struct net_device *, 2551 struct packet_type *, 2552 struct net_device *); 2553 void (*list_func) (struct list_head *, 2554 struct packet_type *, 2555 struct net_device *); 2556 bool (*id_match)(struct packet_type *ptype, 2557 struct sock *sk); 2558 struct net *af_packet_net; 2559 void *af_packet_priv; 2560 struct list_head list; 2561 }; 2562 2563 struct offload_callbacks { 2564 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2565 netdev_features_t features); 2566 struct sk_buff *(*gro_receive)(struct list_head *head, 2567 struct sk_buff *skb); 2568 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2569 }; 2570 2571 struct packet_offload { 2572 __be16 type; /* This is really htons(ether_type). */ 2573 u16 priority; 2574 struct offload_callbacks callbacks; 2575 struct list_head list; 2576 }; 2577 2578 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2579 struct pcpu_sw_netstats { 2580 u64 rx_packets; 2581 u64 rx_bytes; 2582 u64 tx_packets; 2583 u64 tx_bytes; 2584 struct u64_stats_sync syncp; 2585 } __aligned(4 * sizeof(u64)); 2586 2587 struct pcpu_lstats { 2588 u64_stats_t packets; 2589 u64_stats_t bytes; 2590 struct u64_stats_sync syncp; 2591 } __aligned(2 * sizeof(u64)); 2592 2593 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2594 2595 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2596 { 2597 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2598 2599 u64_stats_update_begin(&tstats->syncp); 2600 tstats->rx_bytes += len; 2601 tstats->rx_packets++; 2602 u64_stats_update_end(&tstats->syncp); 2603 } 2604 2605 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2606 unsigned int packets, 2607 unsigned int len) 2608 { 2609 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2610 2611 u64_stats_update_begin(&tstats->syncp); 2612 tstats->tx_bytes += len; 2613 tstats->tx_packets += packets; 2614 u64_stats_update_end(&tstats->syncp); 2615 } 2616 2617 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2618 { 2619 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2620 2621 u64_stats_update_begin(&lstats->syncp); 2622 u64_stats_add(&lstats->bytes, len); 2623 u64_stats_inc(&lstats->packets); 2624 u64_stats_update_end(&lstats->syncp); 2625 } 2626 2627 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2628 ({ \ 2629 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2630 if (pcpu_stats) { \ 2631 int __cpu; \ 2632 for_each_possible_cpu(__cpu) { \ 2633 typeof(type) *stat; \ 2634 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2635 u64_stats_init(&stat->syncp); \ 2636 } \ 2637 } \ 2638 pcpu_stats; \ 2639 }) 2640 2641 #define netdev_alloc_pcpu_stats(type) \ 2642 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2643 2644 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2645 ({ \ 2646 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2647 if (pcpu_stats) { \ 2648 int __cpu; \ 2649 for_each_possible_cpu(__cpu) { \ 2650 typeof(type) *stat; \ 2651 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2652 u64_stats_init(&stat->syncp); \ 2653 } \ 2654 } \ 2655 pcpu_stats; \ 2656 }) 2657 2658 enum netdev_lag_tx_type { 2659 NETDEV_LAG_TX_TYPE_UNKNOWN, 2660 NETDEV_LAG_TX_TYPE_RANDOM, 2661 NETDEV_LAG_TX_TYPE_BROADCAST, 2662 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2663 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2664 NETDEV_LAG_TX_TYPE_HASH, 2665 }; 2666 2667 enum netdev_lag_hash { 2668 NETDEV_LAG_HASH_NONE, 2669 NETDEV_LAG_HASH_L2, 2670 NETDEV_LAG_HASH_L34, 2671 NETDEV_LAG_HASH_L23, 2672 NETDEV_LAG_HASH_E23, 2673 NETDEV_LAG_HASH_E34, 2674 NETDEV_LAG_HASH_VLAN_SRCMAC, 2675 NETDEV_LAG_HASH_UNKNOWN, 2676 }; 2677 2678 struct netdev_lag_upper_info { 2679 enum netdev_lag_tx_type tx_type; 2680 enum netdev_lag_hash hash_type; 2681 }; 2682 2683 struct netdev_lag_lower_state_info { 2684 u8 link_up : 1, 2685 tx_enabled : 1; 2686 }; 2687 2688 #include <linux/notifier.h> 2689 2690 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2691 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2692 * adding new types. 2693 */ 2694 enum netdev_cmd { 2695 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2696 NETDEV_DOWN, 2697 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2698 detected a hardware crash and restarted 2699 - we can use this eg to kick tcp sessions 2700 once done */ 2701 NETDEV_CHANGE, /* Notify device state change */ 2702 NETDEV_REGISTER, 2703 NETDEV_UNREGISTER, 2704 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2705 NETDEV_CHANGEADDR, /* notify after the address change */ 2706 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2707 NETDEV_GOING_DOWN, 2708 NETDEV_CHANGENAME, 2709 NETDEV_FEAT_CHANGE, 2710 NETDEV_BONDING_FAILOVER, 2711 NETDEV_PRE_UP, 2712 NETDEV_PRE_TYPE_CHANGE, 2713 NETDEV_POST_TYPE_CHANGE, 2714 NETDEV_POST_INIT, 2715 NETDEV_RELEASE, 2716 NETDEV_NOTIFY_PEERS, 2717 NETDEV_JOIN, 2718 NETDEV_CHANGEUPPER, 2719 NETDEV_RESEND_IGMP, 2720 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2721 NETDEV_CHANGEINFODATA, 2722 NETDEV_BONDING_INFO, 2723 NETDEV_PRECHANGEUPPER, 2724 NETDEV_CHANGELOWERSTATE, 2725 NETDEV_UDP_TUNNEL_PUSH_INFO, 2726 NETDEV_UDP_TUNNEL_DROP_INFO, 2727 NETDEV_CHANGE_TX_QUEUE_LEN, 2728 NETDEV_CVLAN_FILTER_PUSH_INFO, 2729 NETDEV_CVLAN_FILTER_DROP_INFO, 2730 NETDEV_SVLAN_FILTER_PUSH_INFO, 2731 NETDEV_SVLAN_FILTER_DROP_INFO, 2732 NETDEV_OFFLOAD_XSTATS_ENABLE, 2733 NETDEV_OFFLOAD_XSTATS_DISABLE, 2734 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 2735 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 2736 }; 2737 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2738 2739 int register_netdevice_notifier(struct notifier_block *nb); 2740 int unregister_netdevice_notifier(struct notifier_block *nb); 2741 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2742 int unregister_netdevice_notifier_net(struct net *net, 2743 struct notifier_block *nb); 2744 int register_netdevice_notifier_dev_net(struct net_device *dev, 2745 struct notifier_block *nb, 2746 struct netdev_net_notifier *nn); 2747 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2748 struct notifier_block *nb, 2749 struct netdev_net_notifier *nn); 2750 2751 struct netdev_notifier_info { 2752 struct net_device *dev; 2753 struct netlink_ext_ack *extack; 2754 }; 2755 2756 struct netdev_notifier_info_ext { 2757 struct netdev_notifier_info info; /* must be first */ 2758 union { 2759 u32 mtu; 2760 } ext; 2761 }; 2762 2763 struct netdev_notifier_change_info { 2764 struct netdev_notifier_info info; /* must be first */ 2765 unsigned int flags_changed; 2766 }; 2767 2768 struct netdev_notifier_changeupper_info { 2769 struct netdev_notifier_info info; /* must be first */ 2770 struct net_device *upper_dev; /* new upper dev */ 2771 bool master; /* is upper dev master */ 2772 bool linking; /* is the notification for link or unlink */ 2773 void *upper_info; /* upper dev info */ 2774 }; 2775 2776 struct netdev_notifier_changelowerstate_info { 2777 struct netdev_notifier_info info; /* must be first */ 2778 void *lower_state_info; /* is lower dev state */ 2779 }; 2780 2781 struct netdev_notifier_pre_changeaddr_info { 2782 struct netdev_notifier_info info; /* must be first */ 2783 const unsigned char *dev_addr; 2784 }; 2785 2786 enum netdev_offload_xstats_type { 2787 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 2788 }; 2789 2790 struct netdev_notifier_offload_xstats_info { 2791 struct netdev_notifier_info info; /* must be first */ 2792 enum netdev_offload_xstats_type type; 2793 2794 union { 2795 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 2796 struct netdev_notifier_offload_xstats_rd *report_delta; 2797 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 2798 struct netdev_notifier_offload_xstats_ru *report_used; 2799 }; 2800 }; 2801 2802 int netdev_offload_xstats_enable(struct net_device *dev, 2803 enum netdev_offload_xstats_type type, 2804 struct netlink_ext_ack *extack); 2805 int netdev_offload_xstats_disable(struct net_device *dev, 2806 enum netdev_offload_xstats_type type); 2807 bool netdev_offload_xstats_enabled(const struct net_device *dev, 2808 enum netdev_offload_xstats_type type); 2809 int netdev_offload_xstats_get(struct net_device *dev, 2810 enum netdev_offload_xstats_type type, 2811 struct rtnl_hw_stats64 *stats, bool *used, 2812 struct netlink_ext_ack *extack); 2813 void 2814 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 2815 const struct rtnl_hw_stats64 *stats); 2816 void 2817 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 2818 void netdev_offload_xstats_push_delta(struct net_device *dev, 2819 enum netdev_offload_xstats_type type, 2820 const struct rtnl_hw_stats64 *stats); 2821 2822 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2823 struct net_device *dev) 2824 { 2825 info->dev = dev; 2826 info->extack = NULL; 2827 } 2828 2829 static inline struct net_device * 2830 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2831 { 2832 return info->dev; 2833 } 2834 2835 static inline struct netlink_ext_ack * 2836 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2837 { 2838 return info->extack; 2839 } 2840 2841 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2842 2843 2844 extern rwlock_t dev_base_lock; /* Device list lock */ 2845 2846 #define for_each_netdev(net, d) \ 2847 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2848 #define for_each_netdev_reverse(net, d) \ 2849 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2850 #define for_each_netdev_rcu(net, d) \ 2851 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2852 #define for_each_netdev_safe(net, d, n) \ 2853 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2854 #define for_each_netdev_continue(net, d) \ 2855 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2856 #define for_each_netdev_continue_reverse(net, d) \ 2857 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2858 dev_list) 2859 #define for_each_netdev_continue_rcu(net, d) \ 2860 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2861 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2862 for_each_netdev_rcu(&init_net, slave) \ 2863 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2864 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2865 2866 static inline struct net_device *next_net_device(struct net_device *dev) 2867 { 2868 struct list_head *lh; 2869 struct net *net; 2870 2871 net = dev_net(dev); 2872 lh = dev->dev_list.next; 2873 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2874 } 2875 2876 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2877 { 2878 struct list_head *lh; 2879 struct net *net; 2880 2881 net = dev_net(dev); 2882 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2883 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2884 } 2885 2886 static inline struct net_device *first_net_device(struct net *net) 2887 { 2888 return list_empty(&net->dev_base_head) ? NULL : 2889 net_device_entry(net->dev_base_head.next); 2890 } 2891 2892 static inline struct net_device *first_net_device_rcu(struct net *net) 2893 { 2894 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2895 2896 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2897 } 2898 2899 int netdev_boot_setup_check(struct net_device *dev); 2900 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2901 const char *hwaddr); 2902 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2903 void dev_add_pack(struct packet_type *pt); 2904 void dev_remove_pack(struct packet_type *pt); 2905 void __dev_remove_pack(struct packet_type *pt); 2906 void dev_add_offload(struct packet_offload *po); 2907 void dev_remove_offload(struct packet_offload *po); 2908 2909 int dev_get_iflink(const struct net_device *dev); 2910 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2911 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 2912 struct net_device_path_stack *stack); 2913 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2914 unsigned short mask); 2915 struct net_device *dev_get_by_name(struct net *net, const char *name); 2916 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2917 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2918 bool netdev_name_in_use(struct net *net, const char *name); 2919 int dev_alloc_name(struct net_device *dev, const char *name); 2920 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2921 void dev_close(struct net_device *dev); 2922 void dev_close_many(struct list_head *head, bool unlink); 2923 void dev_disable_lro(struct net_device *dev); 2924 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2925 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2926 struct net_device *sb_dev); 2927 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2928 struct net_device *sb_dev); 2929 2930 int dev_queue_xmit(struct sk_buff *skb); 2931 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2932 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2933 2934 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 2935 { 2936 int ret; 2937 2938 ret = __dev_direct_xmit(skb, queue_id); 2939 if (!dev_xmit_complete(ret)) 2940 kfree_skb(skb); 2941 return ret; 2942 } 2943 2944 int register_netdevice(struct net_device *dev); 2945 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2946 void unregister_netdevice_many(struct list_head *head); 2947 static inline void unregister_netdevice(struct net_device *dev) 2948 { 2949 unregister_netdevice_queue(dev, NULL); 2950 } 2951 2952 int netdev_refcnt_read(const struct net_device *dev); 2953 void free_netdev(struct net_device *dev); 2954 void netdev_freemem(struct net_device *dev); 2955 int init_dummy_netdev(struct net_device *dev); 2956 2957 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 2958 struct sk_buff *skb, 2959 bool all_slaves); 2960 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 2961 struct sock *sk); 2962 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2963 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2964 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2965 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2966 int netdev_get_name(struct net *net, char *name, int ifindex); 2967 int dev_restart(struct net_device *dev); 2968 2969 2970 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2971 unsigned short type, 2972 const void *daddr, const void *saddr, 2973 unsigned int len) 2974 { 2975 if (!dev->header_ops || !dev->header_ops->create) 2976 return 0; 2977 2978 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2979 } 2980 2981 static inline int dev_parse_header(const struct sk_buff *skb, 2982 unsigned char *haddr) 2983 { 2984 const struct net_device *dev = skb->dev; 2985 2986 if (!dev->header_ops || !dev->header_ops->parse) 2987 return 0; 2988 return dev->header_ops->parse(skb, haddr); 2989 } 2990 2991 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 2992 { 2993 const struct net_device *dev = skb->dev; 2994 2995 if (!dev->header_ops || !dev->header_ops->parse_protocol) 2996 return 0; 2997 return dev->header_ops->parse_protocol(skb); 2998 } 2999 3000 /* ll_header must have at least hard_header_len allocated */ 3001 static inline bool dev_validate_header(const struct net_device *dev, 3002 char *ll_header, int len) 3003 { 3004 if (likely(len >= dev->hard_header_len)) 3005 return true; 3006 if (len < dev->min_header_len) 3007 return false; 3008 3009 if (capable(CAP_SYS_RAWIO)) { 3010 memset(ll_header + len, 0, dev->hard_header_len - len); 3011 return true; 3012 } 3013 3014 if (dev->header_ops && dev->header_ops->validate) 3015 return dev->header_ops->validate(ll_header, len); 3016 3017 return false; 3018 } 3019 3020 static inline bool dev_has_header(const struct net_device *dev) 3021 { 3022 return dev->header_ops && dev->header_ops->create; 3023 } 3024 3025 #ifdef CONFIG_NET_FLOW_LIMIT 3026 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 3027 struct sd_flow_limit { 3028 u64 count; 3029 unsigned int num_buckets; 3030 unsigned int history_head; 3031 u16 history[FLOW_LIMIT_HISTORY]; 3032 u8 buckets[]; 3033 }; 3034 3035 extern int netdev_flow_limit_table_len; 3036 #endif /* CONFIG_NET_FLOW_LIMIT */ 3037 3038 /* 3039 * Incoming packets are placed on per-CPU queues 3040 */ 3041 struct softnet_data { 3042 struct list_head poll_list; 3043 struct sk_buff_head process_queue; 3044 3045 /* stats */ 3046 unsigned int processed; 3047 unsigned int time_squeeze; 3048 unsigned int received_rps; 3049 #ifdef CONFIG_RPS 3050 struct softnet_data *rps_ipi_list; 3051 #endif 3052 #ifdef CONFIG_NET_FLOW_LIMIT 3053 struct sd_flow_limit __rcu *flow_limit; 3054 #endif 3055 struct Qdisc *output_queue; 3056 struct Qdisc **output_queue_tailp; 3057 struct sk_buff *completion_queue; 3058 #ifdef CONFIG_XFRM_OFFLOAD 3059 struct sk_buff_head xfrm_backlog; 3060 #endif 3061 /* written and read only by owning cpu: */ 3062 struct { 3063 u16 recursion; 3064 u8 more; 3065 } xmit; 3066 #ifdef CONFIG_RPS 3067 /* input_queue_head should be written by cpu owning this struct, 3068 * and only read by other cpus. Worth using a cache line. 3069 */ 3070 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3071 3072 /* Elements below can be accessed between CPUs for RPS/RFS */ 3073 call_single_data_t csd ____cacheline_aligned_in_smp; 3074 struct softnet_data *rps_ipi_next; 3075 unsigned int cpu; 3076 unsigned int input_queue_tail; 3077 #endif 3078 unsigned int dropped; 3079 struct sk_buff_head input_pkt_queue; 3080 struct napi_struct backlog; 3081 3082 }; 3083 3084 static inline void input_queue_head_incr(struct softnet_data *sd) 3085 { 3086 #ifdef CONFIG_RPS 3087 sd->input_queue_head++; 3088 #endif 3089 } 3090 3091 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3092 unsigned int *qtail) 3093 { 3094 #ifdef CONFIG_RPS 3095 *qtail = ++sd->input_queue_tail; 3096 #endif 3097 } 3098 3099 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3100 3101 static inline int dev_recursion_level(void) 3102 { 3103 return this_cpu_read(softnet_data.xmit.recursion); 3104 } 3105 3106 #define XMIT_RECURSION_LIMIT 8 3107 static inline bool dev_xmit_recursion(void) 3108 { 3109 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3110 XMIT_RECURSION_LIMIT); 3111 } 3112 3113 static inline void dev_xmit_recursion_inc(void) 3114 { 3115 __this_cpu_inc(softnet_data.xmit.recursion); 3116 } 3117 3118 static inline void dev_xmit_recursion_dec(void) 3119 { 3120 __this_cpu_dec(softnet_data.xmit.recursion); 3121 } 3122 3123 void __netif_schedule(struct Qdisc *q); 3124 void netif_schedule_queue(struct netdev_queue *txq); 3125 3126 static inline void netif_tx_schedule_all(struct net_device *dev) 3127 { 3128 unsigned int i; 3129 3130 for (i = 0; i < dev->num_tx_queues; i++) 3131 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3132 } 3133 3134 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3135 { 3136 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3137 } 3138 3139 /** 3140 * netif_start_queue - allow transmit 3141 * @dev: network device 3142 * 3143 * Allow upper layers to call the device hard_start_xmit routine. 3144 */ 3145 static inline void netif_start_queue(struct net_device *dev) 3146 { 3147 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3148 } 3149 3150 static inline void netif_tx_start_all_queues(struct net_device *dev) 3151 { 3152 unsigned int i; 3153 3154 for (i = 0; i < dev->num_tx_queues; i++) { 3155 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3156 netif_tx_start_queue(txq); 3157 } 3158 } 3159 3160 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3161 3162 /** 3163 * netif_wake_queue - restart transmit 3164 * @dev: network device 3165 * 3166 * Allow upper layers to call the device hard_start_xmit routine. 3167 * Used for flow control when transmit resources are available. 3168 */ 3169 static inline void netif_wake_queue(struct net_device *dev) 3170 { 3171 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3172 } 3173 3174 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3175 { 3176 unsigned int i; 3177 3178 for (i = 0; i < dev->num_tx_queues; i++) { 3179 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3180 netif_tx_wake_queue(txq); 3181 } 3182 } 3183 3184 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3185 { 3186 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3187 } 3188 3189 /** 3190 * netif_stop_queue - stop transmitted packets 3191 * @dev: network device 3192 * 3193 * Stop upper layers calling the device hard_start_xmit routine. 3194 * Used for flow control when transmit resources are unavailable. 3195 */ 3196 static inline void netif_stop_queue(struct net_device *dev) 3197 { 3198 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3199 } 3200 3201 void netif_tx_stop_all_queues(struct net_device *dev); 3202 3203 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3204 { 3205 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3206 } 3207 3208 /** 3209 * netif_queue_stopped - test if transmit queue is flowblocked 3210 * @dev: network device 3211 * 3212 * Test if transmit queue on device is currently unable to send. 3213 */ 3214 static inline bool netif_queue_stopped(const struct net_device *dev) 3215 { 3216 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3217 } 3218 3219 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3220 { 3221 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3222 } 3223 3224 static inline bool 3225 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3226 { 3227 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3228 } 3229 3230 static inline bool 3231 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3232 { 3233 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3234 } 3235 3236 /** 3237 * netdev_queue_set_dql_min_limit - set dql minimum limit 3238 * @dev_queue: pointer to transmit queue 3239 * @min_limit: dql minimum limit 3240 * 3241 * Forces xmit_more() to return true until the minimum threshold 3242 * defined by @min_limit is reached (or until the tx queue is 3243 * empty). Warning: to be use with care, misuse will impact the 3244 * latency. 3245 */ 3246 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3247 unsigned int min_limit) 3248 { 3249 #ifdef CONFIG_BQL 3250 dev_queue->dql.min_limit = min_limit; 3251 #endif 3252 } 3253 3254 /** 3255 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3256 * @dev_queue: pointer to transmit queue 3257 * 3258 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3259 * to give appropriate hint to the CPU. 3260 */ 3261 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3262 { 3263 #ifdef CONFIG_BQL 3264 prefetchw(&dev_queue->dql.num_queued); 3265 #endif 3266 } 3267 3268 /** 3269 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3270 * @dev_queue: pointer to transmit queue 3271 * 3272 * BQL enabled drivers might use this helper in their TX completion path, 3273 * to give appropriate hint to the CPU. 3274 */ 3275 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3276 { 3277 #ifdef CONFIG_BQL 3278 prefetchw(&dev_queue->dql.limit); 3279 #endif 3280 } 3281 3282 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3283 unsigned int bytes) 3284 { 3285 #ifdef CONFIG_BQL 3286 dql_queued(&dev_queue->dql, bytes); 3287 3288 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3289 return; 3290 3291 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3292 3293 /* 3294 * The XOFF flag must be set before checking the dql_avail below, 3295 * because in netdev_tx_completed_queue we update the dql_completed 3296 * before checking the XOFF flag. 3297 */ 3298 smp_mb(); 3299 3300 /* check again in case another CPU has just made room avail */ 3301 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3302 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3303 #endif 3304 } 3305 3306 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3307 * that they should not test BQL status themselves. 3308 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3309 * skb of a batch. 3310 * Returns true if the doorbell must be used to kick the NIC. 3311 */ 3312 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3313 unsigned int bytes, 3314 bool xmit_more) 3315 { 3316 if (xmit_more) { 3317 #ifdef CONFIG_BQL 3318 dql_queued(&dev_queue->dql, bytes); 3319 #endif 3320 return netif_tx_queue_stopped(dev_queue); 3321 } 3322 netdev_tx_sent_queue(dev_queue, bytes); 3323 return true; 3324 } 3325 3326 /** 3327 * netdev_sent_queue - report the number of bytes queued to hardware 3328 * @dev: network device 3329 * @bytes: number of bytes queued to the hardware device queue 3330 * 3331 * Report the number of bytes queued for sending/completion to the network 3332 * device hardware queue. @bytes should be a good approximation and should 3333 * exactly match netdev_completed_queue() @bytes 3334 */ 3335 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3336 { 3337 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3338 } 3339 3340 static inline bool __netdev_sent_queue(struct net_device *dev, 3341 unsigned int bytes, 3342 bool xmit_more) 3343 { 3344 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3345 xmit_more); 3346 } 3347 3348 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3349 unsigned int pkts, unsigned int bytes) 3350 { 3351 #ifdef CONFIG_BQL 3352 if (unlikely(!bytes)) 3353 return; 3354 3355 dql_completed(&dev_queue->dql, bytes); 3356 3357 /* 3358 * Without the memory barrier there is a small possiblity that 3359 * netdev_tx_sent_queue will miss the update and cause the queue to 3360 * be stopped forever 3361 */ 3362 smp_mb(); 3363 3364 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3365 return; 3366 3367 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3368 netif_schedule_queue(dev_queue); 3369 #endif 3370 } 3371 3372 /** 3373 * netdev_completed_queue - report bytes and packets completed by device 3374 * @dev: network device 3375 * @pkts: actual number of packets sent over the medium 3376 * @bytes: actual number of bytes sent over the medium 3377 * 3378 * Report the number of bytes and packets transmitted by the network device 3379 * hardware queue over the physical medium, @bytes must exactly match the 3380 * @bytes amount passed to netdev_sent_queue() 3381 */ 3382 static inline void netdev_completed_queue(struct net_device *dev, 3383 unsigned int pkts, unsigned int bytes) 3384 { 3385 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3386 } 3387 3388 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3389 { 3390 #ifdef CONFIG_BQL 3391 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3392 dql_reset(&q->dql); 3393 #endif 3394 } 3395 3396 /** 3397 * netdev_reset_queue - reset the packets and bytes count of a network device 3398 * @dev_queue: network device 3399 * 3400 * Reset the bytes and packet count of a network device and clear the 3401 * software flow control OFF bit for this network device 3402 */ 3403 static inline void netdev_reset_queue(struct net_device *dev_queue) 3404 { 3405 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3406 } 3407 3408 /** 3409 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3410 * @dev: network device 3411 * @queue_index: given tx queue index 3412 * 3413 * Returns 0 if given tx queue index >= number of device tx queues, 3414 * otherwise returns the originally passed tx queue index. 3415 */ 3416 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3417 { 3418 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3419 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3420 dev->name, queue_index, 3421 dev->real_num_tx_queues); 3422 return 0; 3423 } 3424 3425 return queue_index; 3426 } 3427 3428 /** 3429 * netif_running - test if up 3430 * @dev: network device 3431 * 3432 * Test if the device has been brought up. 3433 */ 3434 static inline bool netif_running(const struct net_device *dev) 3435 { 3436 return test_bit(__LINK_STATE_START, &dev->state); 3437 } 3438 3439 /* 3440 * Routines to manage the subqueues on a device. We only need start, 3441 * stop, and a check if it's stopped. All other device management is 3442 * done at the overall netdevice level. 3443 * Also test the device if we're multiqueue. 3444 */ 3445 3446 /** 3447 * netif_start_subqueue - allow sending packets on subqueue 3448 * @dev: network device 3449 * @queue_index: sub queue index 3450 * 3451 * Start individual transmit queue of a device with multiple transmit queues. 3452 */ 3453 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3454 { 3455 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3456 3457 netif_tx_start_queue(txq); 3458 } 3459 3460 /** 3461 * netif_stop_subqueue - stop sending packets on subqueue 3462 * @dev: network device 3463 * @queue_index: sub queue index 3464 * 3465 * Stop individual transmit queue of a device with multiple transmit queues. 3466 */ 3467 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3468 { 3469 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3470 netif_tx_stop_queue(txq); 3471 } 3472 3473 /** 3474 * __netif_subqueue_stopped - test status of subqueue 3475 * @dev: network device 3476 * @queue_index: sub queue index 3477 * 3478 * Check individual transmit queue of a device with multiple transmit queues. 3479 */ 3480 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3481 u16 queue_index) 3482 { 3483 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3484 3485 return netif_tx_queue_stopped(txq); 3486 } 3487 3488 /** 3489 * netif_subqueue_stopped - test status of subqueue 3490 * @dev: network device 3491 * @skb: sub queue buffer pointer 3492 * 3493 * Check individual transmit queue of a device with multiple transmit queues. 3494 */ 3495 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3496 struct sk_buff *skb) 3497 { 3498 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3499 } 3500 3501 /** 3502 * netif_wake_subqueue - allow sending packets on subqueue 3503 * @dev: network device 3504 * @queue_index: sub queue index 3505 * 3506 * Resume individual transmit queue of a device with multiple transmit queues. 3507 */ 3508 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3509 { 3510 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3511 3512 netif_tx_wake_queue(txq); 3513 } 3514 3515 #ifdef CONFIG_XPS 3516 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3517 u16 index); 3518 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3519 u16 index, enum xps_map_type type); 3520 3521 /** 3522 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3523 * @j: CPU/Rx queue index 3524 * @mask: bitmask of all cpus/rx queues 3525 * @nr_bits: number of bits in the bitmask 3526 * 3527 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3528 */ 3529 static inline bool netif_attr_test_mask(unsigned long j, 3530 const unsigned long *mask, 3531 unsigned int nr_bits) 3532 { 3533 cpu_max_bits_warn(j, nr_bits); 3534 return test_bit(j, mask); 3535 } 3536 3537 /** 3538 * netif_attr_test_online - Test for online CPU/Rx queue 3539 * @j: CPU/Rx queue index 3540 * @online_mask: bitmask for CPUs/Rx queues that are online 3541 * @nr_bits: number of bits in the bitmask 3542 * 3543 * Returns true if a CPU/Rx queue is online. 3544 */ 3545 static inline bool netif_attr_test_online(unsigned long j, 3546 const unsigned long *online_mask, 3547 unsigned int nr_bits) 3548 { 3549 cpu_max_bits_warn(j, nr_bits); 3550 3551 if (online_mask) 3552 return test_bit(j, online_mask); 3553 3554 return (j < nr_bits); 3555 } 3556 3557 /** 3558 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3559 * @n: CPU/Rx queue index 3560 * @srcp: the cpumask/Rx queue mask pointer 3561 * @nr_bits: number of bits in the bitmask 3562 * 3563 * Returns >= nr_bits if no further CPUs/Rx queues set. 3564 */ 3565 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3566 unsigned int nr_bits) 3567 { 3568 /* -1 is a legal arg here. */ 3569 if (n != -1) 3570 cpu_max_bits_warn(n, nr_bits); 3571 3572 if (srcp) 3573 return find_next_bit(srcp, nr_bits, n + 1); 3574 3575 return n + 1; 3576 } 3577 3578 /** 3579 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3580 * @n: CPU/Rx queue index 3581 * @src1p: the first CPUs/Rx queues mask pointer 3582 * @src2p: the second CPUs/Rx queues mask pointer 3583 * @nr_bits: number of bits in the bitmask 3584 * 3585 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3586 */ 3587 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3588 const unsigned long *src2p, 3589 unsigned int nr_bits) 3590 { 3591 /* -1 is a legal arg here. */ 3592 if (n != -1) 3593 cpu_max_bits_warn(n, nr_bits); 3594 3595 if (src1p && src2p) 3596 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3597 else if (src1p) 3598 return find_next_bit(src1p, nr_bits, n + 1); 3599 else if (src2p) 3600 return find_next_bit(src2p, nr_bits, n + 1); 3601 3602 return n + 1; 3603 } 3604 #else 3605 static inline int netif_set_xps_queue(struct net_device *dev, 3606 const struct cpumask *mask, 3607 u16 index) 3608 { 3609 return 0; 3610 } 3611 3612 static inline int __netif_set_xps_queue(struct net_device *dev, 3613 const unsigned long *mask, 3614 u16 index, enum xps_map_type type) 3615 { 3616 return 0; 3617 } 3618 #endif 3619 3620 /** 3621 * netif_is_multiqueue - test if device has multiple transmit queues 3622 * @dev: network device 3623 * 3624 * Check if device has multiple transmit queues 3625 */ 3626 static inline bool netif_is_multiqueue(const struct net_device *dev) 3627 { 3628 return dev->num_tx_queues > 1; 3629 } 3630 3631 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3632 3633 #ifdef CONFIG_SYSFS 3634 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3635 #else 3636 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3637 unsigned int rxqs) 3638 { 3639 dev->real_num_rx_queues = rxqs; 3640 return 0; 3641 } 3642 #endif 3643 int netif_set_real_num_queues(struct net_device *dev, 3644 unsigned int txq, unsigned int rxq); 3645 3646 static inline struct netdev_rx_queue * 3647 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3648 { 3649 return dev->_rx + rxq; 3650 } 3651 3652 #ifdef CONFIG_SYSFS 3653 static inline unsigned int get_netdev_rx_queue_index( 3654 struct netdev_rx_queue *queue) 3655 { 3656 struct net_device *dev = queue->dev; 3657 int index = queue - dev->_rx; 3658 3659 BUG_ON(index >= dev->num_rx_queues); 3660 return index; 3661 } 3662 #endif 3663 3664 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3665 int netif_get_num_default_rss_queues(void); 3666 3667 enum skb_free_reason { 3668 SKB_REASON_CONSUMED, 3669 SKB_REASON_DROPPED, 3670 }; 3671 3672 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3673 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3674 3675 /* 3676 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3677 * interrupt context or with hardware interrupts being disabled. 3678 * (in_hardirq() || irqs_disabled()) 3679 * 3680 * We provide four helpers that can be used in following contexts : 3681 * 3682 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3683 * replacing kfree_skb(skb) 3684 * 3685 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3686 * Typically used in place of consume_skb(skb) in TX completion path 3687 * 3688 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3689 * replacing kfree_skb(skb) 3690 * 3691 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3692 * and consumed a packet. Used in place of consume_skb(skb) 3693 */ 3694 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3695 { 3696 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3697 } 3698 3699 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3700 { 3701 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3702 } 3703 3704 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3705 { 3706 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3707 } 3708 3709 static inline void dev_consume_skb_any(struct sk_buff *skb) 3710 { 3711 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3712 } 3713 3714 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3715 struct bpf_prog *xdp_prog); 3716 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3717 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3718 int netif_rx(struct sk_buff *skb); 3719 int __netif_rx(struct sk_buff *skb); 3720 3721 int netif_receive_skb(struct sk_buff *skb); 3722 int netif_receive_skb_core(struct sk_buff *skb); 3723 void netif_receive_skb_list_internal(struct list_head *head); 3724 void netif_receive_skb_list(struct list_head *head); 3725 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3726 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3727 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3728 gro_result_t napi_gro_frags(struct napi_struct *napi); 3729 struct packet_offload *gro_find_receive_by_type(__be16 type); 3730 struct packet_offload *gro_find_complete_by_type(__be16 type); 3731 3732 static inline void napi_free_frags(struct napi_struct *napi) 3733 { 3734 kfree_skb(napi->skb); 3735 napi->skb = NULL; 3736 } 3737 3738 bool netdev_is_rx_handler_busy(struct net_device *dev); 3739 int netdev_rx_handler_register(struct net_device *dev, 3740 rx_handler_func_t *rx_handler, 3741 void *rx_handler_data); 3742 void netdev_rx_handler_unregister(struct net_device *dev); 3743 3744 bool dev_valid_name(const char *name); 3745 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 3746 { 3747 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 3748 } 3749 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 3750 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 3751 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3752 void __user *data, bool *need_copyout); 3753 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 3754 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 3755 unsigned int dev_get_flags(const struct net_device *); 3756 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3757 struct netlink_ext_ack *extack); 3758 int dev_change_flags(struct net_device *dev, unsigned int flags, 3759 struct netlink_ext_ack *extack); 3760 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3761 unsigned int gchanges); 3762 int dev_change_name(struct net_device *, const char *); 3763 int dev_set_alias(struct net_device *, const char *, size_t); 3764 int dev_get_alias(const struct net_device *, char *, size_t); 3765 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 3766 const char *pat, int new_ifindex); 3767 static inline 3768 int dev_change_net_namespace(struct net_device *dev, struct net *net, 3769 const char *pat) 3770 { 3771 return __dev_change_net_namespace(dev, net, pat, 0); 3772 } 3773 int __dev_set_mtu(struct net_device *, int); 3774 int dev_validate_mtu(struct net_device *dev, int mtu, 3775 struct netlink_ext_ack *extack); 3776 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3777 struct netlink_ext_ack *extack); 3778 int dev_set_mtu(struct net_device *, int); 3779 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3780 void dev_set_group(struct net_device *, int); 3781 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3782 struct netlink_ext_ack *extack); 3783 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3784 struct netlink_ext_ack *extack); 3785 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 3786 struct netlink_ext_ack *extack); 3787 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 3788 int dev_change_carrier(struct net_device *, bool new_carrier); 3789 int dev_get_phys_port_id(struct net_device *dev, 3790 struct netdev_phys_item_id *ppid); 3791 int dev_get_phys_port_name(struct net_device *dev, 3792 char *name, size_t len); 3793 int dev_get_port_parent_id(struct net_device *dev, 3794 struct netdev_phys_item_id *ppid, bool recurse); 3795 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3796 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3797 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask, 3798 u32 value); 3799 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3800 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3801 struct netdev_queue *txq, int *ret); 3802 3803 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3804 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3805 int fd, int expected_fd, u32 flags); 3806 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3807 u8 dev_xdp_prog_count(struct net_device *dev); 3808 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3809 3810 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3811 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3812 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 3813 bool is_skb_forwardable(const struct net_device *dev, 3814 const struct sk_buff *skb); 3815 3816 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 3817 const struct sk_buff *skb, 3818 const bool check_mtu) 3819 { 3820 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 3821 unsigned int len; 3822 3823 if (!(dev->flags & IFF_UP)) 3824 return false; 3825 3826 if (!check_mtu) 3827 return true; 3828 3829 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 3830 if (skb->len <= len) 3831 return true; 3832 3833 /* if TSO is enabled, we don't care about the length as the packet 3834 * could be forwarded without being segmented before 3835 */ 3836 if (skb_is_gso(skb)) 3837 return true; 3838 3839 return false; 3840 } 3841 3842 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3843 struct sk_buff *skb, 3844 const bool check_mtu) 3845 { 3846 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3847 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 3848 atomic_long_inc(&dev->rx_dropped); 3849 kfree_skb(skb); 3850 return NET_RX_DROP; 3851 } 3852 3853 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 3854 skb->priority = 0; 3855 return 0; 3856 } 3857 3858 bool dev_nit_active(struct net_device *dev); 3859 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3860 3861 extern int netdev_budget; 3862 extern unsigned int netdev_budget_usecs; 3863 3864 /* Called by rtnetlink.c:rtnl_unlock() */ 3865 void netdev_run_todo(void); 3866 3867 static inline void __dev_put(struct net_device *dev) 3868 { 3869 if (dev) { 3870 #ifdef CONFIG_PCPU_DEV_REFCNT 3871 this_cpu_dec(*dev->pcpu_refcnt); 3872 #else 3873 refcount_dec(&dev->dev_refcnt); 3874 #endif 3875 } 3876 } 3877 3878 static inline void __dev_hold(struct net_device *dev) 3879 { 3880 if (dev) { 3881 #ifdef CONFIG_PCPU_DEV_REFCNT 3882 this_cpu_inc(*dev->pcpu_refcnt); 3883 #else 3884 refcount_inc(&dev->dev_refcnt); 3885 #endif 3886 } 3887 } 3888 3889 static inline void __netdev_tracker_alloc(struct net_device *dev, 3890 netdevice_tracker *tracker, 3891 gfp_t gfp) 3892 { 3893 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 3894 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 3895 #endif 3896 } 3897 3898 /* netdev_tracker_alloc() can upgrade a prior untracked reference 3899 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 3900 */ 3901 static inline void netdev_tracker_alloc(struct net_device *dev, 3902 netdevice_tracker *tracker, gfp_t gfp) 3903 { 3904 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 3905 refcount_dec(&dev->refcnt_tracker.no_tracker); 3906 __netdev_tracker_alloc(dev, tracker, gfp); 3907 #endif 3908 } 3909 3910 static inline void netdev_tracker_free(struct net_device *dev, 3911 netdevice_tracker *tracker) 3912 { 3913 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 3914 ref_tracker_free(&dev->refcnt_tracker, tracker); 3915 #endif 3916 } 3917 3918 static inline void dev_hold_track(struct net_device *dev, 3919 netdevice_tracker *tracker, gfp_t gfp) 3920 { 3921 if (dev) { 3922 __dev_hold(dev); 3923 __netdev_tracker_alloc(dev, tracker, gfp); 3924 } 3925 } 3926 3927 static inline void dev_put_track(struct net_device *dev, 3928 netdevice_tracker *tracker) 3929 { 3930 if (dev) { 3931 netdev_tracker_free(dev, tracker); 3932 __dev_put(dev); 3933 } 3934 } 3935 3936 /** 3937 * dev_hold - get reference to device 3938 * @dev: network device 3939 * 3940 * Hold reference to device to keep it from being freed. 3941 * Try using dev_hold_track() instead. 3942 */ 3943 static inline void dev_hold(struct net_device *dev) 3944 { 3945 dev_hold_track(dev, NULL, GFP_ATOMIC); 3946 } 3947 3948 /** 3949 * dev_put - release reference to device 3950 * @dev: network device 3951 * 3952 * Release reference to device to allow it to be freed. 3953 * Try using dev_put_track() instead. 3954 */ 3955 static inline void dev_put(struct net_device *dev) 3956 { 3957 dev_put_track(dev, NULL); 3958 } 3959 3960 static inline void dev_replace_track(struct net_device *odev, 3961 struct net_device *ndev, 3962 netdevice_tracker *tracker, 3963 gfp_t gfp) 3964 { 3965 if (odev) 3966 netdev_tracker_free(odev, tracker); 3967 3968 __dev_hold(ndev); 3969 __dev_put(odev); 3970 3971 if (ndev) 3972 __netdev_tracker_alloc(ndev, tracker, gfp); 3973 } 3974 3975 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3976 * and _off may be called from IRQ context, but it is caller 3977 * who is responsible for serialization of these calls. 3978 * 3979 * The name carrier is inappropriate, these functions should really be 3980 * called netif_lowerlayer_*() because they represent the state of any 3981 * kind of lower layer not just hardware media. 3982 */ 3983 3984 void linkwatch_init_dev(struct net_device *dev); 3985 void linkwatch_fire_event(struct net_device *dev); 3986 void linkwatch_forget_dev(struct net_device *dev); 3987 3988 /** 3989 * netif_carrier_ok - test if carrier present 3990 * @dev: network device 3991 * 3992 * Check if carrier is present on device 3993 */ 3994 static inline bool netif_carrier_ok(const struct net_device *dev) 3995 { 3996 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3997 } 3998 3999 unsigned long dev_trans_start(struct net_device *dev); 4000 4001 void __netdev_watchdog_up(struct net_device *dev); 4002 4003 void netif_carrier_on(struct net_device *dev); 4004 void netif_carrier_off(struct net_device *dev); 4005 void netif_carrier_event(struct net_device *dev); 4006 4007 /** 4008 * netif_dormant_on - mark device as dormant. 4009 * @dev: network device 4010 * 4011 * Mark device as dormant (as per RFC2863). 4012 * 4013 * The dormant state indicates that the relevant interface is not 4014 * actually in a condition to pass packets (i.e., it is not 'up') but is 4015 * in a "pending" state, waiting for some external event. For "on- 4016 * demand" interfaces, this new state identifies the situation where the 4017 * interface is waiting for events to place it in the up state. 4018 */ 4019 static inline void netif_dormant_on(struct net_device *dev) 4020 { 4021 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4022 linkwatch_fire_event(dev); 4023 } 4024 4025 /** 4026 * netif_dormant_off - set device as not dormant. 4027 * @dev: network device 4028 * 4029 * Device is not in dormant state. 4030 */ 4031 static inline void netif_dormant_off(struct net_device *dev) 4032 { 4033 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4034 linkwatch_fire_event(dev); 4035 } 4036 4037 /** 4038 * netif_dormant - test if device is dormant 4039 * @dev: network device 4040 * 4041 * Check if device is dormant. 4042 */ 4043 static inline bool netif_dormant(const struct net_device *dev) 4044 { 4045 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4046 } 4047 4048 4049 /** 4050 * netif_testing_on - mark device as under test. 4051 * @dev: network device 4052 * 4053 * Mark device as under test (as per RFC2863). 4054 * 4055 * The testing state indicates that some test(s) must be performed on 4056 * the interface. After completion, of the test, the interface state 4057 * will change to up, dormant, or down, as appropriate. 4058 */ 4059 static inline void netif_testing_on(struct net_device *dev) 4060 { 4061 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4062 linkwatch_fire_event(dev); 4063 } 4064 4065 /** 4066 * netif_testing_off - set device as not under test. 4067 * @dev: network device 4068 * 4069 * Device is not in testing state. 4070 */ 4071 static inline void netif_testing_off(struct net_device *dev) 4072 { 4073 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4074 linkwatch_fire_event(dev); 4075 } 4076 4077 /** 4078 * netif_testing - test if device is under test 4079 * @dev: network device 4080 * 4081 * Check if device is under test 4082 */ 4083 static inline bool netif_testing(const struct net_device *dev) 4084 { 4085 return test_bit(__LINK_STATE_TESTING, &dev->state); 4086 } 4087 4088 4089 /** 4090 * netif_oper_up - test if device is operational 4091 * @dev: network device 4092 * 4093 * Check if carrier is operational 4094 */ 4095 static inline bool netif_oper_up(const struct net_device *dev) 4096 { 4097 return (dev->operstate == IF_OPER_UP || 4098 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4099 } 4100 4101 /** 4102 * netif_device_present - is device available or removed 4103 * @dev: network device 4104 * 4105 * Check if device has not been removed from system. 4106 */ 4107 static inline bool netif_device_present(const struct net_device *dev) 4108 { 4109 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4110 } 4111 4112 void netif_device_detach(struct net_device *dev); 4113 4114 void netif_device_attach(struct net_device *dev); 4115 4116 /* 4117 * Network interface message level settings 4118 */ 4119 4120 enum { 4121 NETIF_MSG_DRV_BIT, 4122 NETIF_MSG_PROBE_BIT, 4123 NETIF_MSG_LINK_BIT, 4124 NETIF_MSG_TIMER_BIT, 4125 NETIF_MSG_IFDOWN_BIT, 4126 NETIF_MSG_IFUP_BIT, 4127 NETIF_MSG_RX_ERR_BIT, 4128 NETIF_MSG_TX_ERR_BIT, 4129 NETIF_MSG_TX_QUEUED_BIT, 4130 NETIF_MSG_INTR_BIT, 4131 NETIF_MSG_TX_DONE_BIT, 4132 NETIF_MSG_RX_STATUS_BIT, 4133 NETIF_MSG_PKTDATA_BIT, 4134 NETIF_MSG_HW_BIT, 4135 NETIF_MSG_WOL_BIT, 4136 4137 /* When you add a new bit above, update netif_msg_class_names array 4138 * in net/ethtool/common.c 4139 */ 4140 NETIF_MSG_CLASS_COUNT, 4141 }; 4142 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4143 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4144 4145 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4146 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4147 4148 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4149 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4150 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4151 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4152 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4153 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4154 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4155 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4156 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4157 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4158 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4159 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4160 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4161 #define NETIF_MSG_HW __NETIF_MSG(HW) 4162 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4163 4164 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4165 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4166 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4167 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4168 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4169 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4170 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4171 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4172 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4173 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4174 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4175 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4176 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4177 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4178 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4179 4180 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4181 { 4182 /* use default */ 4183 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4184 return default_msg_enable_bits; 4185 if (debug_value == 0) /* no output */ 4186 return 0; 4187 /* set low N bits */ 4188 return (1U << debug_value) - 1; 4189 } 4190 4191 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4192 { 4193 spin_lock(&txq->_xmit_lock); 4194 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4195 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4196 } 4197 4198 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4199 { 4200 __acquire(&txq->_xmit_lock); 4201 return true; 4202 } 4203 4204 static inline void __netif_tx_release(struct netdev_queue *txq) 4205 { 4206 __release(&txq->_xmit_lock); 4207 } 4208 4209 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4210 { 4211 spin_lock_bh(&txq->_xmit_lock); 4212 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4213 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4214 } 4215 4216 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4217 { 4218 bool ok = spin_trylock(&txq->_xmit_lock); 4219 4220 if (likely(ok)) { 4221 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4222 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4223 } 4224 return ok; 4225 } 4226 4227 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4228 { 4229 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4230 WRITE_ONCE(txq->xmit_lock_owner, -1); 4231 spin_unlock(&txq->_xmit_lock); 4232 } 4233 4234 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4235 { 4236 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4237 WRITE_ONCE(txq->xmit_lock_owner, -1); 4238 spin_unlock_bh(&txq->_xmit_lock); 4239 } 4240 4241 /* 4242 * txq->trans_start can be read locklessly from dev_watchdog() 4243 */ 4244 static inline void txq_trans_update(struct netdev_queue *txq) 4245 { 4246 if (txq->xmit_lock_owner != -1) 4247 WRITE_ONCE(txq->trans_start, jiffies); 4248 } 4249 4250 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4251 { 4252 unsigned long now = jiffies; 4253 4254 if (READ_ONCE(txq->trans_start) != now) 4255 WRITE_ONCE(txq->trans_start, now); 4256 } 4257 4258 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4259 static inline void netif_trans_update(struct net_device *dev) 4260 { 4261 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4262 4263 txq_trans_cond_update(txq); 4264 } 4265 4266 /** 4267 * netif_tx_lock - grab network device transmit lock 4268 * @dev: network device 4269 * 4270 * Get network device transmit lock 4271 */ 4272 void netif_tx_lock(struct net_device *dev); 4273 4274 static inline void netif_tx_lock_bh(struct net_device *dev) 4275 { 4276 local_bh_disable(); 4277 netif_tx_lock(dev); 4278 } 4279 4280 void netif_tx_unlock(struct net_device *dev); 4281 4282 static inline void netif_tx_unlock_bh(struct net_device *dev) 4283 { 4284 netif_tx_unlock(dev); 4285 local_bh_enable(); 4286 } 4287 4288 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4289 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4290 __netif_tx_lock(txq, cpu); \ 4291 } else { \ 4292 __netif_tx_acquire(txq); \ 4293 } \ 4294 } 4295 4296 #define HARD_TX_TRYLOCK(dev, txq) \ 4297 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4298 __netif_tx_trylock(txq) : \ 4299 __netif_tx_acquire(txq)) 4300 4301 #define HARD_TX_UNLOCK(dev, txq) { \ 4302 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4303 __netif_tx_unlock(txq); \ 4304 } else { \ 4305 __netif_tx_release(txq); \ 4306 } \ 4307 } 4308 4309 static inline void netif_tx_disable(struct net_device *dev) 4310 { 4311 unsigned int i; 4312 int cpu; 4313 4314 local_bh_disable(); 4315 cpu = smp_processor_id(); 4316 spin_lock(&dev->tx_global_lock); 4317 for (i = 0; i < dev->num_tx_queues; i++) { 4318 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4319 4320 __netif_tx_lock(txq, cpu); 4321 netif_tx_stop_queue(txq); 4322 __netif_tx_unlock(txq); 4323 } 4324 spin_unlock(&dev->tx_global_lock); 4325 local_bh_enable(); 4326 } 4327 4328 static inline void netif_addr_lock(struct net_device *dev) 4329 { 4330 unsigned char nest_level = 0; 4331 4332 #ifdef CONFIG_LOCKDEP 4333 nest_level = dev->nested_level; 4334 #endif 4335 spin_lock_nested(&dev->addr_list_lock, nest_level); 4336 } 4337 4338 static inline void netif_addr_lock_bh(struct net_device *dev) 4339 { 4340 unsigned char nest_level = 0; 4341 4342 #ifdef CONFIG_LOCKDEP 4343 nest_level = dev->nested_level; 4344 #endif 4345 local_bh_disable(); 4346 spin_lock_nested(&dev->addr_list_lock, nest_level); 4347 } 4348 4349 static inline void netif_addr_unlock(struct net_device *dev) 4350 { 4351 spin_unlock(&dev->addr_list_lock); 4352 } 4353 4354 static inline void netif_addr_unlock_bh(struct net_device *dev) 4355 { 4356 spin_unlock_bh(&dev->addr_list_lock); 4357 } 4358 4359 /* 4360 * dev_addrs walker. Should be used only for read access. Call with 4361 * rcu_read_lock held. 4362 */ 4363 #define for_each_dev_addr(dev, ha) \ 4364 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4365 4366 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4367 4368 void ether_setup(struct net_device *dev); 4369 4370 /* Support for loadable net-drivers */ 4371 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4372 unsigned char name_assign_type, 4373 void (*setup)(struct net_device *), 4374 unsigned int txqs, unsigned int rxqs); 4375 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4376 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4377 4378 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4379 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4380 count) 4381 4382 int register_netdev(struct net_device *dev); 4383 void unregister_netdev(struct net_device *dev); 4384 4385 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4386 4387 /* General hardware address lists handling functions */ 4388 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4389 struct netdev_hw_addr_list *from_list, int addr_len); 4390 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4391 struct netdev_hw_addr_list *from_list, int addr_len); 4392 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4393 struct net_device *dev, 4394 int (*sync)(struct net_device *, const unsigned char *), 4395 int (*unsync)(struct net_device *, 4396 const unsigned char *)); 4397 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4398 struct net_device *dev, 4399 int (*sync)(struct net_device *, 4400 const unsigned char *, int), 4401 int (*unsync)(struct net_device *, 4402 const unsigned char *, int)); 4403 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4404 struct net_device *dev, 4405 int (*unsync)(struct net_device *, 4406 const unsigned char *, int)); 4407 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4408 struct net_device *dev, 4409 int (*unsync)(struct net_device *, 4410 const unsigned char *)); 4411 void __hw_addr_init(struct netdev_hw_addr_list *list); 4412 4413 /* Functions used for device addresses handling */ 4414 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4415 const void *addr, size_t len); 4416 4417 static inline void 4418 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4419 { 4420 dev_addr_mod(dev, 0, addr, len); 4421 } 4422 4423 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4424 { 4425 __dev_addr_set(dev, addr, dev->addr_len); 4426 } 4427 4428 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4429 unsigned char addr_type); 4430 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4431 unsigned char addr_type); 4432 void dev_addr_flush(struct net_device *dev); 4433 int dev_addr_init(struct net_device *dev); 4434 void dev_addr_check(struct net_device *dev); 4435 4436 /* Functions used for unicast addresses handling */ 4437 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4438 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4439 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4440 int dev_uc_sync(struct net_device *to, struct net_device *from); 4441 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4442 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4443 void dev_uc_flush(struct net_device *dev); 4444 void dev_uc_init(struct net_device *dev); 4445 4446 /** 4447 * __dev_uc_sync - Synchonize device's unicast list 4448 * @dev: device to sync 4449 * @sync: function to call if address should be added 4450 * @unsync: function to call if address should be removed 4451 * 4452 * Add newly added addresses to the interface, and release 4453 * addresses that have been deleted. 4454 */ 4455 static inline int __dev_uc_sync(struct net_device *dev, 4456 int (*sync)(struct net_device *, 4457 const unsigned char *), 4458 int (*unsync)(struct net_device *, 4459 const unsigned char *)) 4460 { 4461 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4462 } 4463 4464 /** 4465 * __dev_uc_unsync - Remove synchronized addresses from device 4466 * @dev: device to sync 4467 * @unsync: function to call if address should be removed 4468 * 4469 * Remove all addresses that were added to the device by dev_uc_sync(). 4470 */ 4471 static inline void __dev_uc_unsync(struct net_device *dev, 4472 int (*unsync)(struct net_device *, 4473 const unsigned char *)) 4474 { 4475 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4476 } 4477 4478 /* Functions used for multicast addresses handling */ 4479 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4480 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4481 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4482 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4483 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4484 int dev_mc_sync(struct net_device *to, struct net_device *from); 4485 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4486 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4487 void dev_mc_flush(struct net_device *dev); 4488 void dev_mc_init(struct net_device *dev); 4489 4490 /** 4491 * __dev_mc_sync - Synchonize device's multicast list 4492 * @dev: device to sync 4493 * @sync: function to call if address should be added 4494 * @unsync: function to call if address should be removed 4495 * 4496 * Add newly added addresses to the interface, and release 4497 * addresses that have been deleted. 4498 */ 4499 static inline int __dev_mc_sync(struct net_device *dev, 4500 int (*sync)(struct net_device *, 4501 const unsigned char *), 4502 int (*unsync)(struct net_device *, 4503 const unsigned char *)) 4504 { 4505 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4506 } 4507 4508 /** 4509 * __dev_mc_unsync - Remove synchronized addresses from device 4510 * @dev: device to sync 4511 * @unsync: function to call if address should be removed 4512 * 4513 * Remove all addresses that were added to the device by dev_mc_sync(). 4514 */ 4515 static inline void __dev_mc_unsync(struct net_device *dev, 4516 int (*unsync)(struct net_device *, 4517 const unsigned char *)) 4518 { 4519 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4520 } 4521 4522 /* Functions used for secondary unicast and multicast support */ 4523 void dev_set_rx_mode(struct net_device *dev); 4524 void __dev_set_rx_mode(struct net_device *dev); 4525 int dev_set_promiscuity(struct net_device *dev, int inc); 4526 int dev_set_allmulti(struct net_device *dev, int inc); 4527 void netdev_state_change(struct net_device *dev); 4528 void __netdev_notify_peers(struct net_device *dev); 4529 void netdev_notify_peers(struct net_device *dev); 4530 void netdev_features_change(struct net_device *dev); 4531 /* Load a device via the kmod */ 4532 void dev_load(struct net *net, const char *name); 4533 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4534 struct rtnl_link_stats64 *storage); 4535 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4536 const struct net_device_stats *netdev_stats); 4537 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4538 const struct pcpu_sw_netstats __percpu *netstats); 4539 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4540 4541 extern int netdev_max_backlog; 4542 extern int netdev_tstamp_prequeue; 4543 extern int netdev_unregister_timeout_secs; 4544 extern int weight_p; 4545 extern int dev_weight_rx_bias; 4546 extern int dev_weight_tx_bias; 4547 extern int dev_rx_weight; 4548 extern int dev_tx_weight; 4549 extern int gro_normal_batch; 4550 4551 enum { 4552 NESTED_SYNC_IMM_BIT, 4553 NESTED_SYNC_TODO_BIT, 4554 }; 4555 4556 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4557 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4558 4559 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4560 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4561 4562 struct netdev_nested_priv { 4563 unsigned char flags; 4564 void *data; 4565 }; 4566 4567 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4568 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4569 struct list_head **iter); 4570 4571 #ifdef CONFIG_LOCKDEP 4572 static LIST_HEAD(net_unlink_list); 4573 4574 static inline void net_unlink_todo(struct net_device *dev) 4575 { 4576 if (list_empty(&dev->unlink_list)) 4577 list_add_tail(&dev->unlink_list, &net_unlink_list); 4578 } 4579 #endif 4580 4581 /* iterate through upper list, must be called under RCU read lock */ 4582 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4583 for (iter = &(dev)->adj_list.upper, \ 4584 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4585 updev; \ 4586 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4587 4588 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4589 int (*fn)(struct net_device *upper_dev, 4590 struct netdev_nested_priv *priv), 4591 struct netdev_nested_priv *priv); 4592 4593 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4594 struct net_device *upper_dev); 4595 4596 bool netdev_has_any_upper_dev(struct net_device *dev); 4597 4598 void *netdev_lower_get_next_private(struct net_device *dev, 4599 struct list_head **iter); 4600 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4601 struct list_head **iter); 4602 4603 #define netdev_for_each_lower_private(dev, priv, iter) \ 4604 for (iter = (dev)->adj_list.lower.next, \ 4605 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4606 priv; \ 4607 priv = netdev_lower_get_next_private(dev, &(iter))) 4608 4609 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4610 for (iter = &(dev)->adj_list.lower, \ 4611 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4612 priv; \ 4613 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4614 4615 void *netdev_lower_get_next(struct net_device *dev, 4616 struct list_head **iter); 4617 4618 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4619 for (iter = (dev)->adj_list.lower.next, \ 4620 ldev = netdev_lower_get_next(dev, &(iter)); \ 4621 ldev; \ 4622 ldev = netdev_lower_get_next(dev, &(iter))) 4623 4624 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4625 struct list_head **iter); 4626 int netdev_walk_all_lower_dev(struct net_device *dev, 4627 int (*fn)(struct net_device *lower_dev, 4628 struct netdev_nested_priv *priv), 4629 struct netdev_nested_priv *priv); 4630 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4631 int (*fn)(struct net_device *lower_dev, 4632 struct netdev_nested_priv *priv), 4633 struct netdev_nested_priv *priv); 4634 4635 void *netdev_adjacent_get_private(struct list_head *adj_list); 4636 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4637 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4638 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4639 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4640 struct netlink_ext_ack *extack); 4641 int netdev_master_upper_dev_link(struct net_device *dev, 4642 struct net_device *upper_dev, 4643 void *upper_priv, void *upper_info, 4644 struct netlink_ext_ack *extack); 4645 void netdev_upper_dev_unlink(struct net_device *dev, 4646 struct net_device *upper_dev); 4647 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4648 struct net_device *new_dev, 4649 struct net_device *dev, 4650 struct netlink_ext_ack *extack); 4651 void netdev_adjacent_change_commit(struct net_device *old_dev, 4652 struct net_device *new_dev, 4653 struct net_device *dev); 4654 void netdev_adjacent_change_abort(struct net_device *old_dev, 4655 struct net_device *new_dev, 4656 struct net_device *dev); 4657 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4658 void *netdev_lower_dev_get_private(struct net_device *dev, 4659 struct net_device *lower_dev); 4660 void netdev_lower_state_changed(struct net_device *lower_dev, 4661 void *lower_state_info); 4662 4663 /* RSS keys are 40 or 52 bytes long */ 4664 #define NETDEV_RSS_KEY_LEN 52 4665 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4666 void netdev_rss_key_fill(void *buffer, size_t len); 4667 4668 int skb_checksum_help(struct sk_buff *skb); 4669 int skb_crc32c_csum_help(struct sk_buff *skb); 4670 int skb_csum_hwoffload_help(struct sk_buff *skb, 4671 const netdev_features_t features); 4672 4673 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4674 netdev_features_t features, bool tx_path); 4675 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4676 netdev_features_t features); 4677 4678 struct netdev_bonding_info { 4679 ifslave slave; 4680 ifbond master; 4681 }; 4682 4683 struct netdev_notifier_bonding_info { 4684 struct netdev_notifier_info info; /* must be first */ 4685 struct netdev_bonding_info bonding_info; 4686 }; 4687 4688 void netdev_bonding_info_change(struct net_device *dev, 4689 struct netdev_bonding_info *bonding_info); 4690 4691 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4692 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4693 #else 4694 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4695 const void *data) 4696 { 4697 } 4698 #endif 4699 4700 static inline 4701 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4702 { 4703 return __skb_gso_segment(skb, features, true); 4704 } 4705 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4706 4707 static inline bool can_checksum_protocol(netdev_features_t features, 4708 __be16 protocol) 4709 { 4710 if (protocol == htons(ETH_P_FCOE)) 4711 return !!(features & NETIF_F_FCOE_CRC); 4712 4713 /* Assume this is an IP checksum (not SCTP CRC) */ 4714 4715 if (features & NETIF_F_HW_CSUM) { 4716 /* Can checksum everything */ 4717 return true; 4718 } 4719 4720 switch (protocol) { 4721 case htons(ETH_P_IP): 4722 return !!(features & NETIF_F_IP_CSUM); 4723 case htons(ETH_P_IPV6): 4724 return !!(features & NETIF_F_IPV6_CSUM); 4725 default: 4726 return false; 4727 } 4728 } 4729 4730 #ifdef CONFIG_BUG 4731 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4732 #else 4733 static inline void netdev_rx_csum_fault(struct net_device *dev, 4734 struct sk_buff *skb) 4735 { 4736 } 4737 #endif 4738 /* rx skb timestamps */ 4739 void net_enable_timestamp(void); 4740 void net_disable_timestamp(void); 4741 4742 #ifdef CONFIG_PROC_FS 4743 int __init dev_proc_init(void); 4744 #else 4745 #define dev_proc_init() 0 4746 #endif 4747 4748 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4749 struct sk_buff *skb, struct net_device *dev, 4750 bool more) 4751 { 4752 __this_cpu_write(softnet_data.xmit.more, more); 4753 return ops->ndo_start_xmit(skb, dev); 4754 } 4755 4756 static inline bool netdev_xmit_more(void) 4757 { 4758 return __this_cpu_read(softnet_data.xmit.more); 4759 } 4760 4761 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4762 struct netdev_queue *txq, bool more) 4763 { 4764 const struct net_device_ops *ops = dev->netdev_ops; 4765 netdev_tx_t rc; 4766 4767 rc = __netdev_start_xmit(ops, skb, dev, more); 4768 if (rc == NETDEV_TX_OK) 4769 txq_trans_update(txq); 4770 4771 return rc; 4772 } 4773 4774 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4775 const void *ns); 4776 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4777 const void *ns); 4778 4779 extern const struct kobj_ns_type_operations net_ns_type_operations; 4780 4781 const char *netdev_drivername(const struct net_device *dev); 4782 4783 void linkwatch_run_queue(void); 4784 4785 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4786 netdev_features_t f2) 4787 { 4788 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4789 if (f1 & NETIF_F_HW_CSUM) 4790 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4791 else 4792 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4793 } 4794 4795 return f1 & f2; 4796 } 4797 4798 static inline netdev_features_t netdev_get_wanted_features( 4799 struct net_device *dev) 4800 { 4801 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4802 } 4803 netdev_features_t netdev_increment_features(netdev_features_t all, 4804 netdev_features_t one, netdev_features_t mask); 4805 4806 /* Allow TSO being used on stacked device : 4807 * Performing the GSO segmentation before last device 4808 * is a performance improvement. 4809 */ 4810 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4811 netdev_features_t mask) 4812 { 4813 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4814 } 4815 4816 int __netdev_update_features(struct net_device *dev); 4817 void netdev_update_features(struct net_device *dev); 4818 void netdev_change_features(struct net_device *dev); 4819 4820 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4821 struct net_device *dev); 4822 4823 netdev_features_t passthru_features_check(struct sk_buff *skb, 4824 struct net_device *dev, 4825 netdev_features_t features); 4826 netdev_features_t netif_skb_features(struct sk_buff *skb); 4827 4828 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4829 { 4830 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4831 4832 /* check flags correspondence */ 4833 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4834 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4835 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4836 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4837 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4838 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4839 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4840 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4841 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4842 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4843 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4844 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4845 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4846 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4847 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4848 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4849 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4850 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4851 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4852 4853 return (features & feature) == feature; 4854 } 4855 4856 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4857 { 4858 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4859 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4860 } 4861 4862 static inline bool netif_needs_gso(struct sk_buff *skb, 4863 netdev_features_t features) 4864 { 4865 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4866 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4867 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4868 } 4869 4870 static inline void netif_set_gso_max_size(struct net_device *dev, 4871 unsigned int size) 4872 { 4873 /* dev->gso_max_size is read locklessly from sk_setup_caps() */ 4874 WRITE_ONCE(dev->gso_max_size, size); 4875 } 4876 4877 static inline void netif_set_gso_max_segs(struct net_device *dev, 4878 unsigned int segs) 4879 { 4880 /* dev->gso_max_segs is read locklessly from sk_setup_caps() */ 4881 WRITE_ONCE(dev->gso_max_segs, segs); 4882 } 4883 4884 static inline void netif_set_gro_max_size(struct net_device *dev, 4885 unsigned int size) 4886 { 4887 /* This pairs with the READ_ONCE() in skb_gro_receive() */ 4888 WRITE_ONCE(dev->gro_max_size, size); 4889 } 4890 4891 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4892 int pulled_hlen, u16 mac_offset, 4893 int mac_len) 4894 { 4895 skb->protocol = protocol; 4896 skb->encapsulation = 1; 4897 skb_push(skb, pulled_hlen); 4898 skb_reset_transport_header(skb); 4899 skb->mac_header = mac_offset; 4900 skb->network_header = skb->mac_header + mac_len; 4901 skb->mac_len = mac_len; 4902 } 4903 4904 static inline bool netif_is_macsec(const struct net_device *dev) 4905 { 4906 return dev->priv_flags & IFF_MACSEC; 4907 } 4908 4909 static inline bool netif_is_macvlan(const struct net_device *dev) 4910 { 4911 return dev->priv_flags & IFF_MACVLAN; 4912 } 4913 4914 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4915 { 4916 return dev->priv_flags & IFF_MACVLAN_PORT; 4917 } 4918 4919 static inline bool netif_is_bond_master(const struct net_device *dev) 4920 { 4921 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4922 } 4923 4924 static inline bool netif_is_bond_slave(const struct net_device *dev) 4925 { 4926 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4927 } 4928 4929 static inline bool netif_supports_nofcs(struct net_device *dev) 4930 { 4931 return dev->priv_flags & IFF_SUPP_NOFCS; 4932 } 4933 4934 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4935 { 4936 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4937 } 4938 4939 static inline bool netif_is_l3_master(const struct net_device *dev) 4940 { 4941 return dev->priv_flags & IFF_L3MDEV_MASTER; 4942 } 4943 4944 static inline bool netif_is_l3_slave(const struct net_device *dev) 4945 { 4946 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4947 } 4948 4949 static inline bool netif_is_bridge_master(const struct net_device *dev) 4950 { 4951 return dev->priv_flags & IFF_EBRIDGE; 4952 } 4953 4954 static inline bool netif_is_bridge_port(const struct net_device *dev) 4955 { 4956 return dev->priv_flags & IFF_BRIDGE_PORT; 4957 } 4958 4959 static inline bool netif_is_ovs_master(const struct net_device *dev) 4960 { 4961 return dev->priv_flags & IFF_OPENVSWITCH; 4962 } 4963 4964 static inline bool netif_is_ovs_port(const struct net_device *dev) 4965 { 4966 return dev->priv_flags & IFF_OVS_DATAPATH; 4967 } 4968 4969 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 4970 { 4971 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 4972 } 4973 4974 static inline bool netif_is_team_master(const struct net_device *dev) 4975 { 4976 return dev->priv_flags & IFF_TEAM; 4977 } 4978 4979 static inline bool netif_is_team_port(const struct net_device *dev) 4980 { 4981 return dev->priv_flags & IFF_TEAM_PORT; 4982 } 4983 4984 static inline bool netif_is_lag_master(const struct net_device *dev) 4985 { 4986 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4987 } 4988 4989 static inline bool netif_is_lag_port(const struct net_device *dev) 4990 { 4991 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4992 } 4993 4994 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4995 { 4996 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4997 } 4998 4999 static inline bool netif_is_failover(const struct net_device *dev) 5000 { 5001 return dev->priv_flags & IFF_FAILOVER; 5002 } 5003 5004 static inline bool netif_is_failover_slave(const struct net_device *dev) 5005 { 5006 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5007 } 5008 5009 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5010 static inline void netif_keep_dst(struct net_device *dev) 5011 { 5012 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5013 } 5014 5015 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5016 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5017 { 5018 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5019 return netif_is_macsec(dev); 5020 } 5021 5022 extern struct pernet_operations __net_initdata loopback_net_ops; 5023 5024 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5025 5026 /* netdev_printk helpers, similar to dev_printk */ 5027 5028 static inline const char *netdev_name(const struct net_device *dev) 5029 { 5030 if (!dev->name[0] || strchr(dev->name, '%')) 5031 return "(unnamed net_device)"; 5032 return dev->name; 5033 } 5034 5035 static inline bool netdev_unregistering(const struct net_device *dev) 5036 { 5037 return dev->reg_state == NETREG_UNREGISTERING; 5038 } 5039 5040 static inline const char *netdev_reg_state(const struct net_device *dev) 5041 { 5042 switch (dev->reg_state) { 5043 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5044 case NETREG_REGISTERED: return ""; 5045 case NETREG_UNREGISTERING: return " (unregistering)"; 5046 case NETREG_UNREGISTERED: return " (unregistered)"; 5047 case NETREG_RELEASED: return " (released)"; 5048 case NETREG_DUMMY: return " (dummy)"; 5049 } 5050 5051 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5052 return " (unknown)"; 5053 } 5054 5055 __printf(3, 4) __cold 5056 void netdev_printk(const char *level, const struct net_device *dev, 5057 const char *format, ...); 5058 __printf(2, 3) __cold 5059 void netdev_emerg(const struct net_device *dev, const char *format, ...); 5060 __printf(2, 3) __cold 5061 void netdev_alert(const struct net_device *dev, const char *format, ...); 5062 __printf(2, 3) __cold 5063 void netdev_crit(const struct net_device *dev, const char *format, ...); 5064 __printf(2, 3) __cold 5065 void netdev_err(const struct net_device *dev, const char *format, ...); 5066 __printf(2, 3) __cold 5067 void netdev_warn(const struct net_device *dev, const char *format, ...); 5068 __printf(2, 3) __cold 5069 void netdev_notice(const struct net_device *dev, const char *format, ...); 5070 __printf(2, 3) __cold 5071 void netdev_info(const struct net_device *dev, const char *format, ...); 5072 5073 #define netdev_level_once(level, dev, fmt, ...) \ 5074 do { \ 5075 static bool __section(".data.once") __print_once; \ 5076 \ 5077 if (!__print_once) { \ 5078 __print_once = true; \ 5079 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 5080 } \ 5081 } while (0) 5082 5083 #define netdev_emerg_once(dev, fmt, ...) \ 5084 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 5085 #define netdev_alert_once(dev, fmt, ...) \ 5086 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 5087 #define netdev_crit_once(dev, fmt, ...) \ 5088 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 5089 #define netdev_err_once(dev, fmt, ...) \ 5090 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 5091 #define netdev_warn_once(dev, fmt, ...) \ 5092 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 5093 #define netdev_notice_once(dev, fmt, ...) \ 5094 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 5095 #define netdev_info_once(dev, fmt, ...) \ 5096 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 5097 5098 #define MODULE_ALIAS_NETDEV(device) \ 5099 MODULE_ALIAS("netdev-" device) 5100 5101 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5102 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5103 #define netdev_dbg(__dev, format, args...) \ 5104 do { \ 5105 dynamic_netdev_dbg(__dev, format, ##args); \ 5106 } while (0) 5107 #elif defined(DEBUG) 5108 #define netdev_dbg(__dev, format, args...) \ 5109 netdev_printk(KERN_DEBUG, __dev, format, ##args) 5110 #else 5111 #define netdev_dbg(__dev, format, args...) \ 5112 ({ \ 5113 if (0) \ 5114 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 5115 }) 5116 #endif 5117 5118 #if defined(VERBOSE_DEBUG) 5119 #define netdev_vdbg netdev_dbg 5120 #else 5121 5122 #define netdev_vdbg(dev, format, args...) \ 5123 ({ \ 5124 if (0) \ 5125 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 5126 0; \ 5127 }) 5128 #endif 5129 5130 /* 5131 * netdev_WARN() acts like dev_printk(), but with the key difference 5132 * of using a WARN/WARN_ON to get the message out, including the 5133 * file/line information and a backtrace. 5134 */ 5135 #define netdev_WARN(dev, format, args...) \ 5136 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5137 netdev_reg_state(dev), ##args) 5138 5139 #define netdev_WARN_ONCE(dev, format, args...) \ 5140 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5141 netdev_reg_state(dev), ##args) 5142 5143 /* netif printk helpers, similar to netdev_printk */ 5144 5145 #define netif_printk(priv, type, level, dev, fmt, args...) \ 5146 do { \ 5147 if (netif_msg_##type(priv)) \ 5148 netdev_printk(level, (dev), fmt, ##args); \ 5149 } while (0) 5150 5151 #define netif_level(level, priv, type, dev, fmt, args...) \ 5152 do { \ 5153 if (netif_msg_##type(priv)) \ 5154 netdev_##level(dev, fmt, ##args); \ 5155 } while (0) 5156 5157 #define netif_emerg(priv, type, dev, fmt, args...) \ 5158 netif_level(emerg, priv, type, dev, fmt, ##args) 5159 #define netif_alert(priv, type, dev, fmt, args...) \ 5160 netif_level(alert, priv, type, dev, fmt, ##args) 5161 #define netif_crit(priv, type, dev, fmt, args...) \ 5162 netif_level(crit, priv, type, dev, fmt, ##args) 5163 #define netif_err(priv, type, dev, fmt, args...) \ 5164 netif_level(err, priv, type, dev, fmt, ##args) 5165 #define netif_warn(priv, type, dev, fmt, args...) \ 5166 netif_level(warn, priv, type, dev, fmt, ##args) 5167 #define netif_notice(priv, type, dev, fmt, args...) \ 5168 netif_level(notice, priv, type, dev, fmt, ##args) 5169 #define netif_info(priv, type, dev, fmt, args...) \ 5170 netif_level(info, priv, type, dev, fmt, ##args) 5171 5172 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5173 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5174 #define netif_dbg(priv, type, netdev, format, args...) \ 5175 do { \ 5176 if (netif_msg_##type(priv)) \ 5177 dynamic_netdev_dbg(netdev, format, ##args); \ 5178 } while (0) 5179 #elif defined(DEBUG) 5180 #define netif_dbg(priv, type, dev, format, args...) \ 5181 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 5182 #else 5183 #define netif_dbg(priv, type, dev, format, args...) \ 5184 ({ \ 5185 if (0) \ 5186 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5187 0; \ 5188 }) 5189 #endif 5190 5191 /* if @cond then downgrade to debug, else print at @level */ 5192 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 5193 do { \ 5194 if (cond) \ 5195 netif_dbg(priv, type, netdev, fmt, ##args); \ 5196 else \ 5197 netif_ ## level(priv, type, netdev, fmt, ##args); \ 5198 } while (0) 5199 5200 #if defined(VERBOSE_DEBUG) 5201 #define netif_vdbg netif_dbg 5202 #else 5203 #define netif_vdbg(priv, type, dev, format, args...) \ 5204 ({ \ 5205 if (0) \ 5206 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5207 0; \ 5208 }) 5209 #endif 5210 5211 /* 5212 * The list of packet types we will receive (as opposed to discard) 5213 * and the routines to invoke. 5214 * 5215 * Why 16. Because with 16 the only overlap we get on a hash of the 5216 * low nibble of the protocol value is RARP/SNAP/X.25. 5217 * 5218 * 0800 IP 5219 * 0001 802.3 5220 * 0002 AX.25 5221 * 0004 802.2 5222 * 8035 RARP 5223 * 0005 SNAP 5224 * 0805 X.25 5225 * 0806 ARP 5226 * 8137 IPX 5227 * 0009 Localtalk 5228 * 86DD IPv6 5229 */ 5230 #define PTYPE_HASH_SIZE (16) 5231 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5232 5233 extern struct list_head ptype_all __read_mostly; 5234 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5235 5236 extern struct net_device *blackhole_netdev; 5237 5238 #endif /* _LINUX_NETDEVICE_H */ 5239