1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <[email protected]> 13 * Corey Minyard <[email protected]> 14 * Donald J. Becker, <[email protected]> 15 * Alan Cox, <[email protected]> 16 * Bjorn Ekwall. <[email protected]> 17 * Pekka Riikonen <[email protected]> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 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 51 struct netpoll_info; 52 struct device; 53 struct ethtool_ops; 54 struct phy_device; 55 struct dsa_port; 56 struct ip_tunnel_parm; 57 struct macsec_context; 58 struct macsec_ops; 59 60 struct sfp_bus; 61 /* 802.11 specific */ 62 struct wireless_dev; 63 /* 802.15.4 specific */ 64 struct wpan_dev; 65 struct mpls_dev; 66 /* UDP Tunnel offloads */ 67 struct udp_tunnel_info; 68 struct udp_tunnel_nic_info; 69 struct udp_tunnel_nic; 70 struct bpf_prog; 71 struct xdp_buff; 72 73 void synchronize_net(void); 74 void netdev_set_default_ethtool_ops(struct net_device *dev, 75 const struct ethtool_ops *ops); 76 77 /* Backlog congestion levels */ 78 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 79 #define NET_RX_DROP 1 /* packet dropped */ 80 81 #define MAX_NEST_DEV 8 82 83 /* 84 * Transmit return codes: transmit return codes originate from three different 85 * namespaces: 86 * 87 * - qdisc return codes 88 * - driver transmit return codes 89 * - errno values 90 * 91 * Drivers are allowed to return any one of those in their hard_start_xmit() 92 * function. Real network devices commonly used with qdiscs should only return 93 * the driver transmit return codes though - when qdiscs are used, the actual 94 * transmission happens asynchronously, so the value is not propagated to 95 * higher layers. Virtual network devices transmit synchronously; in this case 96 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 97 * others are propagated to higher layers. 98 */ 99 100 /* qdisc ->enqueue() return codes. */ 101 #define NET_XMIT_SUCCESS 0x00 102 #define NET_XMIT_DROP 0x01 /* skb dropped */ 103 #define NET_XMIT_CN 0x02 /* congestion notification */ 104 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 105 106 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 107 * indicates that the device will soon be dropping packets, or already drops 108 * some packets of the same priority; prompting us to send less aggressively. */ 109 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 110 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 111 112 /* Driver transmit return codes */ 113 #define NETDEV_TX_MASK 0xf0 114 115 enum netdev_tx { 116 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 117 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 118 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 119 }; 120 typedef enum netdev_tx netdev_tx_t; 121 122 /* 123 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 124 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 125 */ 126 static inline bool dev_xmit_complete(int rc) 127 { 128 /* 129 * Positive cases with an skb consumed by a driver: 130 * - successful transmission (rc == NETDEV_TX_OK) 131 * - error while transmitting (rc < 0) 132 * - error while queueing to a different device (rc & NET_XMIT_MASK) 133 */ 134 if (likely(rc < NET_XMIT_MASK)) 135 return true; 136 137 return false; 138 } 139 140 /* 141 * Compute the worst-case header length according to the protocols 142 * used. 143 */ 144 145 #if defined(CONFIG_HYPERV_NET) 146 # define LL_MAX_HEADER 128 147 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 148 # if defined(CONFIG_MAC80211_MESH) 149 # define LL_MAX_HEADER 128 150 # else 151 # define LL_MAX_HEADER 96 152 # endif 153 #else 154 # define LL_MAX_HEADER 32 155 #endif 156 157 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 158 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 159 #define MAX_HEADER LL_MAX_HEADER 160 #else 161 #define MAX_HEADER (LL_MAX_HEADER + 48) 162 #endif 163 164 /* 165 * Old network device statistics. Fields are native words 166 * (unsigned long) so they can be read and written atomically. 167 */ 168 169 struct net_device_stats { 170 unsigned long rx_packets; 171 unsigned long tx_packets; 172 unsigned long rx_bytes; 173 unsigned long tx_bytes; 174 unsigned long rx_errors; 175 unsigned long tx_errors; 176 unsigned long rx_dropped; 177 unsigned long tx_dropped; 178 unsigned long multicast; 179 unsigned long collisions; 180 unsigned long rx_length_errors; 181 unsigned long rx_over_errors; 182 unsigned long rx_crc_errors; 183 unsigned long rx_frame_errors; 184 unsigned long rx_fifo_errors; 185 unsigned long rx_missed_errors; 186 unsigned long tx_aborted_errors; 187 unsigned long tx_carrier_errors; 188 unsigned long tx_fifo_errors; 189 unsigned long tx_heartbeat_errors; 190 unsigned long tx_window_errors; 191 unsigned long rx_compressed; 192 unsigned long tx_compressed; 193 }; 194 195 196 #include <linux/cache.h> 197 #include <linux/skbuff.h> 198 199 #ifdef CONFIG_RPS 200 #include <linux/static_key.h> 201 extern struct static_key_false rps_needed; 202 extern struct static_key_false rfs_needed; 203 #endif 204 205 struct neighbour; 206 struct neigh_parms; 207 struct sk_buff; 208 209 struct netdev_hw_addr { 210 struct list_head list; 211 unsigned char addr[MAX_ADDR_LEN]; 212 unsigned char type; 213 #define NETDEV_HW_ADDR_T_LAN 1 214 #define NETDEV_HW_ADDR_T_SAN 2 215 #define NETDEV_HW_ADDR_T_UNICAST 3 216 #define NETDEV_HW_ADDR_T_MULTICAST 4 217 bool global_use; 218 int sync_cnt; 219 int refcount; 220 int synced; 221 struct rcu_head rcu_head; 222 }; 223 224 struct netdev_hw_addr_list { 225 struct list_head list; 226 int count; 227 }; 228 229 #define netdev_hw_addr_list_count(l) ((l)->count) 230 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 231 #define netdev_hw_addr_list_for_each(ha, l) \ 232 list_for_each_entry(ha, &(l)->list, list) 233 234 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 235 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 236 #define netdev_for_each_uc_addr(ha, dev) \ 237 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 238 239 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 240 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 241 #define netdev_for_each_mc_addr(ha, dev) \ 242 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 243 244 struct hh_cache { 245 unsigned int hh_len; 246 seqlock_t hh_lock; 247 248 /* cached hardware header; allow for machine alignment needs. */ 249 #define HH_DATA_MOD 16 250 #define HH_DATA_OFF(__len) \ 251 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 252 #define HH_DATA_ALIGN(__len) \ 253 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 254 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 255 }; 256 257 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 258 * Alternative is: 259 * dev->hard_header_len ? (dev->hard_header_len + 260 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 261 * 262 * We could use other alignment values, but we must maintain the 263 * relationship HH alignment <= LL alignment. 264 */ 265 #define LL_RESERVED_SPACE(dev) \ 266 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 267 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 268 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 269 270 struct header_ops { 271 int (*create) (struct sk_buff *skb, struct net_device *dev, 272 unsigned short type, const void *daddr, 273 const void *saddr, unsigned int len); 274 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 275 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 276 void (*cache_update)(struct hh_cache *hh, 277 const struct net_device *dev, 278 const unsigned char *haddr); 279 bool (*validate)(const char *ll_header, unsigned int len); 280 __be16 (*parse_protocol)(const struct sk_buff *skb); 281 }; 282 283 /* These flag bits are private to the generic network queueing 284 * layer; they may not be explicitly referenced by any other 285 * code. 286 */ 287 288 enum netdev_state_t { 289 __LINK_STATE_START, 290 __LINK_STATE_PRESENT, 291 __LINK_STATE_NOCARRIER, 292 __LINK_STATE_LINKWATCH_PENDING, 293 __LINK_STATE_DORMANT, 294 __LINK_STATE_TESTING, 295 }; 296 297 298 /* 299 * This structure holds boot-time configured netdevice settings. They 300 * are then used in the device probing. 301 */ 302 struct netdev_boot_setup { 303 char name[IFNAMSIZ]; 304 struct ifmap map; 305 }; 306 #define NETDEV_BOOT_SETUP_MAX 8 307 308 int __init netdev_boot_setup(char *str); 309 310 struct gro_list { 311 struct list_head list; 312 int count; 313 }; 314 315 /* 316 * size of gro hash buckets, must less than bit number of 317 * napi_struct::gro_bitmask 318 */ 319 #define GRO_HASH_BUCKETS 8 320 321 /* 322 * Structure for NAPI scheduling similar to tasklet but with weighting 323 */ 324 struct napi_struct { 325 /* The poll_list must only be managed by the entity which 326 * changes the state of the NAPI_STATE_SCHED bit. This means 327 * whoever atomically sets that bit can add this napi_struct 328 * to the per-CPU poll_list, and whoever clears that bit 329 * can remove from the list right before clearing the bit. 330 */ 331 struct list_head poll_list; 332 333 unsigned long state; 334 int weight; 335 int defer_hard_irqs_count; 336 unsigned long gro_bitmask; 337 int (*poll)(struct napi_struct *, int); 338 #ifdef CONFIG_NETPOLL 339 int poll_owner; 340 #endif 341 struct net_device *dev; 342 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 343 struct sk_buff *skb; 344 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 345 int rx_count; /* length of rx_list */ 346 struct hrtimer timer; 347 struct list_head dev_list; 348 struct hlist_node napi_hash_node; 349 unsigned int napi_id; 350 struct task_struct *thread; 351 }; 352 353 enum { 354 NAPI_STATE_SCHED, /* Poll is scheduled */ 355 NAPI_STATE_MISSED, /* reschedule a napi */ 356 NAPI_STATE_DISABLE, /* Disable pending */ 357 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 358 NAPI_STATE_LISTED, /* NAPI added to system lists */ 359 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 360 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 361 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 362 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 363 }; 364 365 enum { 366 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 367 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 368 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 369 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 370 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 371 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 372 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 373 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 374 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 375 }; 376 377 enum gro_result { 378 GRO_MERGED, 379 GRO_MERGED_FREE, 380 GRO_HELD, 381 GRO_NORMAL, 382 GRO_CONSUMED, 383 }; 384 typedef enum gro_result gro_result_t; 385 386 /* 387 * enum rx_handler_result - Possible return values for rx_handlers. 388 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 389 * further. 390 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 391 * case skb->dev was changed by rx_handler. 392 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 393 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 394 * 395 * rx_handlers are functions called from inside __netif_receive_skb(), to do 396 * special processing of the skb, prior to delivery to protocol handlers. 397 * 398 * Currently, a net_device can only have a single rx_handler registered. Trying 399 * to register a second rx_handler will return -EBUSY. 400 * 401 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 402 * To unregister a rx_handler on a net_device, use 403 * netdev_rx_handler_unregister(). 404 * 405 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 406 * do with the skb. 407 * 408 * If the rx_handler consumed the skb in some way, it should return 409 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 410 * the skb to be delivered in some other way. 411 * 412 * If the rx_handler changed skb->dev, to divert the skb to another 413 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 414 * new device will be called if it exists. 415 * 416 * If the rx_handler decides the skb should be ignored, it should return 417 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 418 * are registered on exact device (ptype->dev == skb->dev). 419 * 420 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 421 * delivered, it should return RX_HANDLER_PASS. 422 * 423 * A device without a registered rx_handler will behave as if rx_handler 424 * returned RX_HANDLER_PASS. 425 */ 426 427 enum rx_handler_result { 428 RX_HANDLER_CONSUMED, 429 RX_HANDLER_ANOTHER, 430 RX_HANDLER_EXACT, 431 RX_HANDLER_PASS, 432 }; 433 typedef enum rx_handler_result rx_handler_result_t; 434 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 435 436 void __napi_schedule(struct napi_struct *n); 437 void __napi_schedule_irqoff(struct napi_struct *n); 438 439 static inline bool napi_disable_pending(struct napi_struct *n) 440 { 441 return test_bit(NAPI_STATE_DISABLE, &n->state); 442 } 443 444 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 445 { 446 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 447 } 448 449 bool napi_schedule_prep(struct napi_struct *n); 450 451 /** 452 * napi_schedule - schedule NAPI poll 453 * @n: NAPI context 454 * 455 * Schedule NAPI poll routine to be called if it is not already 456 * running. 457 */ 458 static inline void napi_schedule(struct napi_struct *n) 459 { 460 if (napi_schedule_prep(n)) 461 __napi_schedule(n); 462 } 463 464 /** 465 * napi_schedule_irqoff - schedule NAPI poll 466 * @n: NAPI context 467 * 468 * Variant of napi_schedule(), assuming hard irqs are masked. 469 */ 470 static inline void napi_schedule_irqoff(struct napi_struct *n) 471 { 472 if (napi_schedule_prep(n)) 473 __napi_schedule_irqoff(n); 474 } 475 476 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 477 static inline bool napi_reschedule(struct napi_struct *napi) 478 { 479 if (napi_schedule_prep(napi)) { 480 __napi_schedule(napi); 481 return true; 482 } 483 return false; 484 } 485 486 bool napi_complete_done(struct napi_struct *n, int work_done); 487 /** 488 * napi_complete - NAPI processing complete 489 * @n: NAPI context 490 * 491 * Mark NAPI processing as complete. 492 * Consider using napi_complete_done() instead. 493 * Return false if device should avoid rearming interrupts. 494 */ 495 static inline bool napi_complete(struct napi_struct *n) 496 { 497 return napi_complete_done(n, 0); 498 } 499 500 int dev_set_threaded(struct net_device *dev, bool threaded); 501 502 /** 503 * napi_disable - prevent NAPI from scheduling 504 * @n: NAPI context 505 * 506 * Stop NAPI from being scheduled on this context. 507 * Waits till any outstanding processing completes. 508 */ 509 void napi_disable(struct napi_struct *n); 510 511 void napi_enable(struct napi_struct *n); 512 513 /** 514 * napi_synchronize - wait until NAPI is not running 515 * @n: NAPI context 516 * 517 * Wait until NAPI is done being scheduled on this context. 518 * Waits till any outstanding processing completes but 519 * does not disable future activations. 520 */ 521 static inline void napi_synchronize(const struct napi_struct *n) 522 { 523 if (IS_ENABLED(CONFIG_SMP)) 524 while (test_bit(NAPI_STATE_SCHED, &n->state)) 525 msleep(1); 526 else 527 barrier(); 528 } 529 530 /** 531 * napi_if_scheduled_mark_missed - if napi is running, set the 532 * NAPIF_STATE_MISSED 533 * @n: NAPI context 534 * 535 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 536 * NAPI is scheduled. 537 **/ 538 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 539 { 540 unsigned long val, new; 541 542 do { 543 val = READ_ONCE(n->state); 544 if (val & NAPIF_STATE_DISABLE) 545 return true; 546 547 if (!(val & NAPIF_STATE_SCHED)) 548 return false; 549 550 new = val | NAPIF_STATE_MISSED; 551 } while (cmpxchg(&n->state, val, new) != val); 552 553 return true; 554 } 555 556 enum netdev_queue_state_t { 557 __QUEUE_STATE_DRV_XOFF, 558 __QUEUE_STATE_STACK_XOFF, 559 __QUEUE_STATE_FROZEN, 560 }; 561 562 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 563 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 564 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 565 566 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 567 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 568 QUEUE_STATE_FROZEN) 569 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 570 QUEUE_STATE_FROZEN) 571 572 /* 573 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 574 * netif_tx_* functions below are used to manipulate this flag. The 575 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 576 * queue independently. The netif_xmit_*stopped functions below are called 577 * to check if the queue has been stopped by the driver or stack (either 578 * of the XOFF bits are set in the state). Drivers should not need to call 579 * netif_xmit*stopped functions, they should only be using netif_tx_*. 580 */ 581 582 struct netdev_queue { 583 /* 584 * read-mostly part 585 */ 586 struct net_device *dev; 587 struct Qdisc __rcu *qdisc; 588 struct Qdisc *qdisc_sleeping; 589 #ifdef CONFIG_SYSFS 590 struct kobject kobj; 591 #endif 592 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 593 int numa_node; 594 #endif 595 unsigned long tx_maxrate; 596 /* 597 * Number of TX timeouts for this queue 598 * (/sys/class/net/DEV/Q/trans_timeout) 599 */ 600 unsigned long trans_timeout; 601 602 /* Subordinate device that the queue has been assigned to */ 603 struct net_device *sb_dev; 604 #ifdef CONFIG_XDP_SOCKETS 605 struct xsk_buff_pool *pool; 606 #endif 607 /* 608 * write-mostly part 609 */ 610 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 611 int xmit_lock_owner; 612 /* 613 * Time (in jiffies) of last Tx 614 */ 615 unsigned long trans_start; 616 617 unsigned long state; 618 619 #ifdef CONFIG_BQL 620 struct dql dql; 621 #endif 622 } ____cacheline_aligned_in_smp; 623 624 extern int sysctl_fb_tunnels_only_for_init_net; 625 extern int sysctl_devconf_inherit_init_net; 626 627 /* 628 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 629 * == 1 : For initns only 630 * == 2 : For none. 631 */ 632 static inline bool net_has_fallback_tunnels(const struct net *net) 633 { 634 return !IS_ENABLED(CONFIG_SYSCTL) || 635 !sysctl_fb_tunnels_only_for_init_net || 636 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1); 637 } 638 639 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 640 { 641 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 642 return q->numa_node; 643 #else 644 return NUMA_NO_NODE; 645 #endif 646 } 647 648 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 649 { 650 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 651 q->numa_node = node; 652 #endif 653 } 654 655 #ifdef CONFIG_RPS 656 /* 657 * This structure holds an RPS map which can be of variable length. The 658 * map is an array of CPUs. 659 */ 660 struct rps_map { 661 unsigned int len; 662 struct rcu_head rcu; 663 u16 cpus[]; 664 }; 665 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 666 667 /* 668 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 669 * tail pointer for that CPU's input queue at the time of last enqueue, and 670 * a hardware filter index. 671 */ 672 struct rps_dev_flow { 673 u16 cpu; 674 u16 filter; 675 unsigned int last_qtail; 676 }; 677 #define RPS_NO_FILTER 0xffff 678 679 /* 680 * The rps_dev_flow_table structure contains a table of flow mappings. 681 */ 682 struct rps_dev_flow_table { 683 unsigned int mask; 684 struct rcu_head rcu; 685 struct rps_dev_flow flows[]; 686 }; 687 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 688 ((_num) * sizeof(struct rps_dev_flow))) 689 690 /* 691 * The rps_sock_flow_table contains mappings of flows to the last CPU 692 * on which they were processed by the application (set in recvmsg). 693 * Each entry is a 32bit value. Upper part is the high-order bits 694 * of flow hash, lower part is CPU number. 695 * rps_cpu_mask is used to partition the space, depending on number of 696 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 697 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 698 * meaning we use 32-6=26 bits for the hash. 699 */ 700 struct rps_sock_flow_table { 701 u32 mask; 702 703 u32 ents[] ____cacheline_aligned_in_smp; 704 }; 705 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 706 707 #define RPS_NO_CPU 0xffff 708 709 extern u32 rps_cpu_mask; 710 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 711 712 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 713 u32 hash) 714 { 715 if (table && hash) { 716 unsigned int index = hash & table->mask; 717 u32 val = hash & ~rps_cpu_mask; 718 719 /* We only give a hint, preemption can change CPU under us */ 720 val |= raw_smp_processor_id(); 721 722 if (table->ents[index] != val) 723 table->ents[index] = val; 724 } 725 } 726 727 #ifdef CONFIG_RFS_ACCEL 728 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 729 u16 filter_id); 730 #endif 731 #endif /* CONFIG_RPS */ 732 733 /* This structure contains an instance of an RX queue. */ 734 struct netdev_rx_queue { 735 #ifdef CONFIG_RPS 736 struct rps_map __rcu *rps_map; 737 struct rps_dev_flow_table __rcu *rps_flow_table; 738 #endif 739 struct kobject kobj; 740 struct net_device *dev; 741 struct xdp_rxq_info xdp_rxq; 742 #ifdef CONFIG_XDP_SOCKETS 743 struct xsk_buff_pool *pool; 744 #endif 745 } ____cacheline_aligned_in_smp; 746 747 /* 748 * RX queue sysfs structures and functions. 749 */ 750 struct rx_queue_attribute { 751 struct attribute attr; 752 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 753 ssize_t (*store)(struct netdev_rx_queue *queue, 754 const char *buf, size_t len); 755 }; 756 757 #ifdef CONFIG_XPS 758 /* 759 * This structure holds an XPS map which can be of variable length. The 760 * map is an array of queues. 761 */ 762 struct xps_map { 763 unsigned int len; 764 unsigned int alloc_len; 765 struct rcu_head rcu; 766 u16 queues[]; 767 }; 768 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 769 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 770 - sizeof(struct xps_map)) / sizeof(u16)) 771 772 /* 773 * This structure holds all XPS maps for device. Maps are indexed by CPU. 774 */ 775 struct xps_dev_maps { 776 struct rcu_head rcu; 777 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 778 }; 779 780 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 781 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 782 783 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 784 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 785 786 #endif /* CONFIG_XPS */ 787 788 #define TC_MAX_QUEUE 16 789 #define TC_BITMASK 15 790 /* HW offloaded queuing disciplines txq count and offset maps */ 791 struct netdev_tc_txq { 792 u16 count; 793 u16 offset; 794 }; 795 796 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 797 /* 798 * This structure is to hold information about the device 799 * configured to run FCoE protocol stack. 800 */ 801 struct netdev_fcoe_hbainfo { 802 char manufacturer[64]; 803 char serial_number[64]; 804 char hardware_version[64]; 805 char driver_version[64]; 806 char optionrom_version[64]; 807 char firmware_version[64]; 808 char model[256]; 809 char model_description[256]; 810 }; 811 #endif 812 813 #define MAX_PHYS_ITEM_ID_LEN 32 814 815 /* This structure holds a unique identifier to identify some 816 * physical item (port for example) used by a netdevice. 817 */ 818 struct netdev_phys_item_id { 819 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 820 unsigned char id_len; 821 }; 822 823 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 824 struct netdev_phys_item_id *b) 825 { 826 return a->id_len == b->id_len && 827 memcmp(a->id, b->id, a->id_len) == 0; 828 } 829 830 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 831 struct sk_buff *skb, 832 struct net_device *sb_dev); 833 834 enum tc_setup_type { 835 TC_SETUP_QDISC_MQPRIO, 836 TC_SETUP_CLSU32, 837 TC_SETUP_CLSFLOWER, 838 TC_SETUP_CLSMATCHALL, 839 TC_SETUP_CLSBPF, 840 TC_SETUP_BLOCK, 841 TC_SETUP_QDISC_CBS, 842 TC_SETUP_QDISC_RED, 843 TC_SETUP_QDISC_PRIO, 844 TC_SETUP_QDISC_MQ, 845 TC_SETUP_QDISC_ETF, 846 TC_SETUP_ROOT_QDISC, 847 TC_SETUP_QDISC_GRED, 848 TC_SETUP_QDISC_TAPRIO, 849 TC_SETUP_FT, 850 TC_SETUP_QDISC_ETS, 851 TC_SETUP_QDISC_TBF, 852 TC_SETUP_QDISC_FIFO, 853 TC_SETUP_QDISC_HTB, 854 }; 855 856 /* These structures hold the attributes of bpf state that are being passed 857 * to the netdevice through the bpf op. 858 */ 859 enum bpf_netdev_command { 860 /* Set or clear a bpf program used in the earliest stages of packet 861 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 862 * is responsible for calling bpf_prog_put on any old progs that are 863 * stored. In case of error, the callee need not release the new prog 864 * reference, but on success it takes ownership and must bpf_prog_put 865 * when it is no longer used. 866 */ 867 XDP_SETUP_PROG, 868 XDP_SETUP_PROG_HW, 869 /* BPF program for offload callbacks, invoked at program load time. */ 870 BPF_OFFLOAD_MAP_ALLOC, 871 BPF_OFFLOAD_MAP_FREE, 872 XDP_SETUP_XSK_POOL, 873 }; 874 875 struct bpf_prog_offload_ops; 876 struct netlink_ext_ack; 877 struct xdp_umem; 878 struct xdp_dev_bulk_queue; 879 struct bpf_xdp_link; 880 881 enum bpf_xdp_mode { 882 XDP_MODE_SKB = 0, 883 XDP_MODE_DRV = 1, 884 XDP_MODE_HW = 2, 885 __MAX_XDP_MODE 886 }; 887 888 struct bpf_xdp_entity { 889 struct bpf_prog *prog; 890 struct bpf_xdp_link *link; 891 }; 892 893 struct netdev_bpf { 894 enum bpf_netdev_command command; 895 union { 896 /* XDP_SETUP_PROG */ 897 struct { 898 u32 flags; 899 struct bpf_prog *prog; 900 struct netlink_ext_ack *extack; 901 }; 902 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 903 struct { 904 struct bpf_offloaded_map *offmap; 905 }; 906 /* XDP_SETUP_XSK_POOL */ 907 struct { 908 struct xsk_buff_pool *pool; 909 u16 queue_id; 910 } xsk; 911 }; 912 }; 913 914 /* Flags for ndo_xsk_wakeup. */ 915 #define XDP_WAKEUP_RX (1 << 0) 916 #define XDP_WAKEUP_TX (1 << 1) 917 918 #ifdef CONFIG_XFRM_OFFLOAD 919 struct xfrmdev_ops { 920 int (*xdo_dev_state_add) (struct xfrm_state *x); 921 void (*xdo_dev_state_delete) (struct xfrm_state *x); 922 void (*xdo_dev_state_free) (struct xfrm_state *x); 923 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 924 struct xfrm_state *x); 925 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 926 }; 927 #endif 928 929 struct dev_ifalias { 930 struct rcu_head rcuhead; 931 char ifalias[]; 932 }; 933 934 struct devlink; 935 struct tlsdev_ops; 936 937 struct netdev_name_node { 938 struct hlist_node hlist; 939 struct list_head list; 940 struct net_device *dev; 941 const char *name; 942 }; 943 944 int netdev_name_node_alt_create(struct net_device *dev, const char *name); 945 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name); 946 947 struct netdev_net_notifier { 948 struct list_head list; 949 struct notifier_block *nb; 950 }; 951 952 /* 953 * This structure defines the management hooks for network devices. 954 * The following hooks can be defined; unless noted otherwise, they are 955 * optional and can be filled with a null pointer. 956 * 957 * int (*ndo_init)(struct net_device *dev); 958 * This function is called once when a network device is registered. 959 * The network device can use this for any late stage initialization 960 * or semantic validation. It can fail with an error code which will 961 * be propagated back to register_netdev. 962 * 963 * void (*ndo_uninit)(struct net_device *dev); 964 * This function is called when device is unregistered or when registration 965 * fails. It is not called if init fails. 966 * 967 * int (*ndo_open)(struct net_device *dev); 968 * This function is called when a network device transitions to the up 969 * state. 970 * 971 * int (*ndo_stop)(struct net_device *dev); 972 * This function is called when a network device transitions to the down 973 * state. 974 * 975 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 976 * struct net_device *dev); 977 * Called when a packet needs to be transmitted. 978 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 979 * the queue before that can happen; it's for obsolete devices and weird 980 * corner cases, but the stack really does a non-trivial amount 981 * of useless work if you return NETDEV_TX_BUSY. 982 * Required; cannot be NULL. 983 * 984 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 985 * struct net_device *dev 986 * netdev_features_t features); 987 * Called by core transmit path to determine if device is capable of 988 * performing offload operations on a given packet. This is to give 989 * the device an opportunity to implement any restrictions that cannot 990 * be otherwise expressed by feature flags. The check is called with 991 * the set of features that the stack has calculated and it returns 992 * those the driver believes to be appropriate. 993 * 994 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 995 * struct net_device *sb_dev); 996 * Called to decide which queue to use when device supports multiple 997 * transmit queues. 998 * 999 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1000 * This function is called to allow device receiver to make 1001 * changes to configuration when multicast or promiscuous is enabled. 1002 * 1003 * void (*ndo_set_rx_mode)(struct net_device *dev); 1004 * This function is called device changes address list filtering. 1005 * If driver handles unicast address filtering, it should set 1006 * IFF_UNICAST_FLT in its priv_flags. 1007 * 1008 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1009 * This function is called when the Media Access Control address 1010 * needs to be changed. If this interface is not defined, the 1011 * MAC address can not be changed. 1012 * 1013 * int (*ndo_validate_addr)(struct net_device *dev); 1014 * Test if Media Access Control address is valid for the device. 1015 * 1016 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1017 * Called when a user requests an ioctl which can't be handled by 1018 * the generic interface code. If not defined ioctls return 1019 * not supported error code. 1020 * 1021 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1022 * Used to set network devices bus interface parameters. This interface 1023 * is retained for legacy reasons; new devices should use the bus 1024 * interface (PCI) for low level management. 1025 * 1026 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1027 * Called when a user wants to change the Maximum Transfer Unit 1028 * of a device. 1029 * 1030 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1031 * Callback used when the transmitter has not made any progress 1032 * for dev->watchdog ticks. 1033 * 1034 * void (*ndo_get_stats64)(struct net_device *dev, 1035 * struct rtnl_link_stats64 *storage); 1036 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1037 * Called when a user wants to get the network device usage 1038 * statistics. Drivers must do one of the following: 1039 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1040 * rtnl_link_stats64 structure passed by the caller. 1041 * 2. Define @ndo_get_stats to update a net_device_stats structure 1042 * (which should normally be dev->stats) and return a pointer to 1043 * it. The structure may be changed asynchronously only if each 1044 * field is written atomically. 1045 * 3. Update dev->stats asynchronously and atomically, and define 1046 * neither operation. 1047 * 1048 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1049 * Return true if this device supports offload stats of this attr_id. 1050 * 1051 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1052 * void *attr_data) 1053 * Get statistics for offload operations by attr_id. Write it into the 1054 * attr_data pointer. 1055 * 1056 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1057 * If device supports VLAN filtering this function is called when a 1058 * VLAN id is registered. 1059 * 1060 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1061 * If device supports VLAN filtering this function is called when a 1062 * VLAN id is unregistered. 1063 * 1064 * void (*ndo_poll_controller)(struct net_device *dev); 1065 * 1066 * SR-IOV management functions. 1067 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1068 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1069 * u8 qos, __be16 proto); 1070 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1071 * int max_tx_rate); 1072 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1073 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1074 * int (*ndo_get_vf_config)(struct net_device *dev, 1075 * int vf, struct ifla_vf_info *ivf); 1076 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1077 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1078 * struct nlattr *port[]); 1079 * 1080 * Enable or disable the VF ability to query its RSS Redirection Table and 1081 * Hash Key. This is needed since on some devices VF share this information 1082 * with PF and querying it may introduce a theoretical security risk. 1083 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1084 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1085 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1086 * void *type_data); 1087 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1088 * This is always called from the stack with the rtnl lock held and netif 1089 * tx queues stopped. This allows the netdevice to perform queue 1090 * management safely. 1091 * 1092 * Fiber Channel over Ethernet (FCoE) offload functions. 1093 * int (*ndo_fcoe_enable)(struct net_device *dev); 1094 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1095 * so the underlying device can perform whatever needed configuration or 1096 * initialization to support acceleration of FCoE traffic. 1097 * 1098 * int (*ndo_fcoe_disable)(struct net_device *dev); 1099 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1100 * so the underlying device can perform whatever needed clean-ups to 1101 * stop supporting acceleration of FCoE traffic. 1102 * 1103 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1104 * struct scatterlist *sgl, unsigned int sgc); 1105 * Called when the FCoE Initiator wants to initialize an I/O that 1106 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1107 * perform necessary setup and returns 1 to indicate the device is set up 1108 * successfully to perform DDP on this I/O, otherwise this returns 0. 1109 * 1110 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1111 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1112 * indicated by the FC exchange id 'xid', so the underlying device can 1113 * clean up and reuse resources for later DDP requests. 1114 * 1115 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1116 * struct scatterlist *sgl, unsigned int sgc); 1117 * Called when the FCoE Target wants to initialize an I/O that 1118 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1119 * perform necessary setup and returns 1 to indicate the device is set up 1120 * successfully to perform DDP on this I/O, otherwise this returns 0. 1121 * 1122 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1123 * struct netdev_fcoe_hbainfo *hbainfo); 1124 * Called when the FCoE Protocol stack wants information on the underlying 1125 * device. This information is utilized by the FCoE protocol stack to 1126 * register attributes with Fiber Channel management service as per the 1127 * FC-GS Fabric Device Management Information(FDMI) specification. 1128 * 1129 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1130 * Called when the underlying device wants to override default World Wide 1131 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1132 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1133 * protocol stack to use. 1134 * 1135 * RFS acceleration. 1136 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1137 * u16 rxq_index, u32 flow_id); 1138 * Set hardware filter for RFS. rxq_index is the target queue index; 1139 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1140 * Return the filter ID on success, or a negative error code. 1141 * 1142 * Slave management functions (for bridge, bonding, etc). 1143 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1144 * Called to make another netdev an underling. 1145 * 1146 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1147 * Called to release previously enslaved netdev. 1148 * 1149 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1150 * struct sk_buff *skb, 1151 * bool all_slaves); 1152 * Get the xmit slave of master device. If all_slaves is true, function 1153 * assume all the slaves can transmit. 1154 * 1155 * Feature/offload setting functions. 1156 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1157 * netdev_features_t features); 1158 * Adjusts the requested feature flags according to device-specific 1159 * constraints, and returns the resulting flags. Must not modify 1160 * the device state. 1161 * 1162 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1163 * Called to update device configuration to new features. Passed 1164 * feature set might be less than what was returned by ndo_fix_features()). 1165 * Must return >0 or -errno if it changed dev->features itself. 1166 * 1167 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1168 * struct net_device *dev, 1169 * const unsigned char *addr, u16 vid, u16 flags, 1170 * struct netlink_ext_ack *extack); 1171 * Adds an FDB entry to dev for addr. 1172 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1173 * struct net_device *dev, 1174 * const unsigned char *addr, u16 vid) 1175 * Deletes the FDB entry from dev coresponding to addr. 1176 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1177 * struct net_device *dev, struct net_device *filter_dev, 1178 * int *idx) 1179 * Used to add FDB entries to dump requests. Implementers should add 1180 * entries to skb and update idx with the number of entries. 1181 * 1182 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1183 * u16 flags, struct netlink_ext_ack *extack) 1184 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1185 * struct net_device *dev, u32 filter_mask, 1186 * int nlflags) 1187 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1188 * u16 flags); 1189 * 1190 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1191 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1192 * which do not represent real hardware may define this to allow their 1193 * userspace components to manage their virtual carrier state. Devices 1194 * that determine carrier state from physical hardware properties (eg 1195 * network cables) or protocol-dependent mechanisms (eg 1196 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1197 * 1198 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1199 * struct netdev_phys_item_id *ppid); 1200 * Called to get ID of physical port of this device. If driver does 1201 * not implement this, it is assumed that the hw is not able to have 1202 * multiple net devices on single physical port. 1203 * 1204 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1205 * struct netdev_phys_item_id *ppid) 1206 * Called to get the parent ID of the physical port of this device. 1207 * 1208 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1209 * struct net_device *dev) 1210 * Called by upper layer devices to accelerate switching or other 1211 * station functionality into hardware. 'pdev is the lowerdev 1212 * to use for the offload and 'dev' is the net device that will 1213 * back the offload. Returns a pointer to the private structure 1214 * the upper layer will maintain. 1215 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1216 * Called by upper layer device to delete the station created 1217 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1218 * the station and priv is the structure returned by the add 1219 * operation. 1220 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1221 * int queue_index, u32 maxrate); 1222 * Called when a user wants to set a max-rate limitation of specific 1223 * TX queue. 1224 * int (*ndo_get_iflink)(const struct net_device *dev); 1225 * Called to get the iflink value of this device. 1226 * void (*ndo_change_proto_down)(struct net_device *dev, 1227 * bool proto_down); 1228 * This function is used to pass protocol port error state information 1229 * to the switch driver. The switch driver can react to the proto_down 1230 * by doing a phys down on the associated switch port. 1231 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1232 * This function is used to get egress tunnel information for given skb. 1233 * This is useful for retrieving outer tunnel header parameters while 1234 * sampling packet. 1235 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1236 * This function is used to specify the headroom that the skb must 1237 * consider when allocation skb during packet reception. Setting 1238 * appropriate rx headroom value allows avoiding skb head copy on 1239 * forward. Setting a negative value resets the rx headroom to the 1240 * default value. 1241 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1242 * This function is used to set or query state related to XDP on the 1243 * netdevice and manage BPF offload. See definition of 1244 * enum bpf_netdev_command for details. 1245 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1246 * u32 flags); 1247 * This function is used to submit @n XDP packets for transmit on a 1248 * netdevice. Returns number of frames successfully transmitted, frames 1249 * that got dropped are freed/returned via xdp_return_frame(). 1250 * Returns negative number, means general error invoking ndo, meaning 1251 * no frames were xmit'ed and core-caller will free all frames. 1252 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1253 * This function is used to wake up the softirq, ksoftirqd or kthread 1254 * responsible for sending and/or receiving packets on a specific 1255 * queue id bound to an AF_XDP socket. The flags field specifies if 1256 * only RX, only Tx, or both should be woken up using the flags 1257 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1258 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1259 * Get devlink port instance associated with a given netdev. 1260 * Called with a reference on the netdevice and devlink locks only, 1261 * rtnl_lock is not held. 1262 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1263 * int cmd); 1264 * Add, change, delete or get information on an IPv4 tunnel. 1265 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1266 * If a device is paired with a peer device, return the peer instance. 1267 * The caller must be under RCU read context. 1268 */ 1269 struct net_device_ops { 1270 int (*ndo_init)(struct net_device *dev); 1271 void (*ndo_uninit)(struct net_device *dev); 1272 int (*ndo_open)(struct net_device *dev); 1273 int (*ndo_stop)(struct net_device *dev); 1274 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1275 struct net_device *dev); 1276 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1277 struct net_device *dev, 1278 netdev_features_t features); 1279 u16 (*ndo_select_queue)(struct net_device *dev, 1280 struct sk_buff *skb, 1281 struct net_device *sb_dev); 1282 void (*ndo_change_rx_flags)(struct net_device *dev, 1283 int flags); 1284 void (*ndo_set_rx_mode)(struct net_device *dev); 1285 int (*ndo_set_mac_address)(struct net_device *dev, 1286 void *addr); 1287 int (*ndo_validate_addr)(struct net_device *dev); 1288 int (*ndo_do_ioctl)(struct net_device *dev, 1289 struct ifreq *ifr, int cmd); 1290 int (*ndo_set_config)(struct net_device *dev, 1291 struct ifmap *map); 1292 int (*ndo_change_mtu)(struct net_device *dev, 1293 int new_mtu); 1294 int (*ndo_neigh_setup)(struct net_device *dev, 1295 struct neigh_parms *); 1296 void (*ndo_tx_timeout) (struct net_device *dev, 1297 unsigned int txqueue); 1298 1299 void (*ndo_get_stats64)(struct net_device *dev, 1300 struct rtnl_link_stats64 *storage); 1301 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1302 int (*ndo_get_offload_stats)(int attr_id, 1303 const struct net_device *dev, 1304 void *attr_data); 1305 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1306 1307 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1308 __be16 proto, u16 vid); 1309 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1310 __be16 proto, u16 vid); 1311 #ifdef CONFIG_NET_POLL_CONTROLLER 1312 void (*ndo_poll_controller)(struct net_device *dev); 1313 int (*ndo_netpoll_setup)(struct net_device *dev, 1314 struct netpoll_info *info); 1315 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1316 #endif 1317 int (*ndo_set_vf_mac)(struct net_device *dev, 1318 int queue, u8 *mac); 1319 int (*ndo_set_vf_vlan)(struct net_device *dev, 1320 int queue, u16 vlan, 1321 u8 qos, __be16 proto); 1322 int (*ndo_set_vf_rate)(struct net_device *dev, 1323 int vf, int min_tx_rate, 1324 int max_tx_rate); 1325 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1326 int vf, bool setting); 1327 int (*ndo_set_vf_trust)(struct net_device *dev, 1328 int vf, bool setting); 1329 int (*ndo_get_vf_config)(struct net_device *dev, 1330 int vf, 1331 struct ifla_vf_info *ivf); 1332 int (*ndo_set_vf_link_state)(struct net_device *dev, 1333 int vf, int link_state); 1334 int (*ndo_get_vf_stats)(struct net_device *dev, 1335 int vf, 1336 struct ifla_vf_stats 1337 *vf_stats); 1338 int (*ndo_set_vf_port)(struct net_device *dev, 1339 int vf, 1340 struct nlattr *port[]); 1341 int (*ndo_get_vf_port)(struct net_device *dev, 1342 int vf, struct sk_buff *skb); 1343 int (*ndo_get_vf_guid)(struct net_device *dev, 1344 int vf, 1345 struct ifla_vf_guid *node_guid, 1346 struct ifla_vf_guid *port_guid); 1347 int (*ndo_set_vf_guid)(struct net_device *dev, 1348 int vf, u64 guid, 1349 int guid_type); 1350 int (*ndo_set_vf_rss_query_en)( 1351 struct net_device *dev, 1352 int vf, bool setting); 1353 int (*ndo_setup_tc)(struct net_device *dev, 1354 enum tc_setup_type type, 1355 void *type_data); 1356 #if IS_ENABLED(CONFIG_FCOE) 1357 int (*ndo_fcoe_enable)(struct net_device *dev); 1358 int (*ndo_fcoe_disable)(struct net_device *dev); 1359 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1360 u16 xid, 1361 struct scatterlist *sgl, 1362 unsigned int sgc); 1363 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1364 u16 xid); 1365 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1366 u16 xid, 1367 struct scatterlist *sgl, 1368 unsigned int sgc); 1369 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1370 struct netdev_fcoe_hbainfo *hbainfo); 1371 #endif 1372 1373 #if IS_ENABLED(CONFIG_LIBFCOE) 1374 #define NETDEV_FCOE_WWNN 0 1375 #define NETDEV_FCOE_WWPN 1 1376 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1377 u64 *wwn, int type); 1378 #endif 1379 1380 #ifdef CONFIG_RFS_ACCEL 1381 int (*ndo_rx_flow_steer)(struct net_device *dev, 1382 const struct sk_buff *skb, 1383 u16 rxq_index, 1384 u32 flow_id); 1385 #endif 1386 int (*ndo_add_slave)(struct net_device *dev, 1387 struct net_device *slave_dev, 1388 struct netlink_ext_ack *extack); 1389 int (*ndo_del_slave)(struct net_device *dev, 1390 struct net_device *slave_dev); 1391 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1392 struct sk_buff *skb, 1393 bool all_slaves); 1394 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1395 struct sock *sk); 1396 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1397 netdev_features_t features); 1398 int (*ndo_set_features)(struct net_device *dev, 1399 netdev_features_t features); 1400 int (*ndo_neigh_construct)(struct net_device *dev, 1401 struct neighbour *n); 1402 void (*ndo_neigh_destroy)(struct net_device *dev, 1403 struct neighbour *n); 1404 1405 int (*ndo_fdb_add)(struct ndmsg *ndm, 1406 struct nlattr *tb[], 1407 struct net_device *dev, 1408 const unsigned char *addr, 1409 u16 vid, 1410 u16 flags, 1411 struct netlink_ext_ack *extack); 1412 int (*ndo_fdb_del)(struct ndmsg *ndm, 1413 struct nlattr *tb[], 1414 struct net_device *dev, 1415 const unsigned char *addr, 1416 u16 vid); 1417 int (*ndo_fdb_dump)(struct sk_buff *skb, 1418 struct netlink_callback *cb, 1419 struct net_device *dev, 1420 struct net_device *filter_dev, 1421 int *idx); 1422 int (*ndo_fdb_get)(struct sk_buff *skb, 1423 struct nlattr *tb[], 1424 struct net_device *dev, 1425 const unsigned char *addr, 1426 u16 vid, u32 portid, u32 seq, 1427 struct netlink_ext_ack *extack); 1428 int (*ndo_bridge_setlink)(struct net_device *dev, 1429 struct nlmsghdr *nlh, 1430 u16 flags, 1431 struct netlink_ext_ack *extack); 1432 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1433 u32 pid, u32 seq, 1434 struct net_device *dev, 1435 u32 filter_mask, 1436 int nlflags); 1437 int (*ndo_bridge_dellink)(struct net_device *dev, 1438 struct nlmsghdr *nlh, 1439 u16 flags); 1440 int (*ndo_change_carrier)(struct net_device *dev, 1441 bool new_carrier); 1442 int (*ndo_get_phys_port_id)(struct net_device *dev, 1443 struct netdev_phys_item_id *ppid); 1444 int (*ndo_get_port_parent_id)(struct net_device *dev, 1445 struct netdev_phys_item_id *ppid); 1446 int (*ndo_get_phys_port_name)(struct net_device *dev, 1447 char *name, size_t len); 1448 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1449 struct net_device *dev); 1450 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1451 void *priv); 1452 1453 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1454 int queue_index, 1455 u32 maxrate); 1456 int (*ndo_get_iflink)(const struct net_device *dev); 1457 int (*ndo_change_proto_down)(struct net_device *dev, 1458 bool proto_down); 1459 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1460 struct sk_buff *skb); 1461 void (*ndo_set_rx_headroom)(struct net_device *dev, 1462 int needed_headroom); 1463 int (*ndo_bpf)(struct net_device *dev, 1464 struct netdev_bpf *bpf); 1465 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1466 struct xdp_frame **xdp, 1467 u32 flags); 1468 int (*ndo_xsk_wakeup)(struct net_device *dev, 1469 u32 queue_id, u32 flags); 1470 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1471 int (*ndo_tunnel_ctl)(struct net_device *dev, 1472 struct ip_tunnel_parm *p, int cmd); 1473 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1474 }; 1475 1476 /** 1477 * enum netdev_priv_flags - &struct net_device priv_flags 1478 * 1479 * These are the &struct net_device, they are only set internally 1480 * by drivers and used in the kernel. These flags are invisible to 1481 * userspace; this means that the order of these flags can change 1482 * during any kernel release. 1483 * 1484 * You should have a pretty good reason to be extending these flags. 1485 * 1486 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1487 * @IFF_EBRIDGE: Ethernet bridging device 1488 * @IFF_BONDING: bonding master or slave 1489 * @IFF_ISATAP: ISATAP interface (RFC4214) 1490 * @IFF_WAN_HDLC: WAN HDLC device 1491 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1492 * release skb->dst 1493 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1494 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1495 * @IFF_MACVLAN_PORT: device used as macvlan port 1496 * @IFF_BRIDGE_PORT: device used as bridge port 1497 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1498 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1499 * @IFF_UNICAST_FLT: Supports unicast filtering 1500 * @IFF_TEAM_PORT: device used as team port 1501 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1502 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1503 * change when it's running 1504 * @IFF_MACVLAN: Macvlan device 1505 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1506 * underlying stacked devices 1507 * @IFF_L3MDEV_MASTER: device is an L3 master device 1508 * @IFF_NO_QUEUE: device can run without qdisc attached 1509 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1510 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1511 * @IFF_TEAM: device is a team device 1512 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1513 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1514 * entity (i.e. the master device for bridged veth) 1515 * @IFF_MACSEC: device is a MACsec device 1516 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1517 * @IFF_FAILOVER: device is a failover master device 1518 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1519 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1520 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1521 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1522 * skb_headlen(skb) == 0 (data starts from frag0) 1523 */ 1524 enum netdev_priv_flags { 1525 IFF_802_1Q_VLAN = 1<<0, 1526 IFF_EBRIDGE = 1<<1, 1527 IFF_BONDING = 1<<2, 1528 IFF_ISATAP = 1<<3, 1529 IFF_WAN_HDLC = 1<<4, 1530 IFF_XMIT_DST_RELEASE = 1<<5, 1531 IFF_DONT_BRIDGE = 1<<6, 1532 IFF_DISABLE_NETPOLL = 1<<7, 1533 IFF_MACVLAN_PORT = 1<<8, 1534 IFF_BRIDGE_PORT = 1<<9, 1535 IFF_OVS_DATAPATH = 1<<10, 1536 IFF_TX_SKB_SHARING = 1<<11, 1537 IFF_UNICAST_FLT = 1<<12, 1538 IFF_TEAM_PORT = 1<<13, 1539 IFF_SUPP_NOFCS = 1<<14, 1540 IFF_LIVE_ADDR_CHANGE = 1<<15, 1541 IFF_MACVLAN = 1<<16, 1542 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1543 IFF_L3MDEV_MASTER = 1<<18, 1544 IFF_NO_QUEUE = 1<<19, 1545 IFF_OPENVSWITCH = 1<<20, 1546 IFF_L3MDEV_SLAVE = 1<<21, 1547 IFF_TEAM = 1<<22, 1548 IFF_RXFH_CONFIGURED = 1<<23, 1549 IFF_PHONY_HEADROOM = 1<<24, 1550 IFF_MACSEC = 1<<25, 1551 IFF_NO_RX_HANDLER = 1<<26, 1552 IFF_FAILOVER = 1<<27, 1553 IFF_FAILOVER_SLAVE = 1<<28, 1554 IFF_L3MDEV_RX_HANDLER = 1<<29, 1555 IFF_LIVE_RENAME_OK = 1<<30, 1556 IFF_TX_SKB_NO_LINEAR = 1<<31, 1557 }; 1558 1559 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1560 #define IFF_EBRIDGE IFF_EBRIDGE 1561 #define IFF_BONDING IFF_BONDING 1562 #define IFF_ISATAP IFF_ISATAP 1563 #define IFF_WAN_HDLC IFF_WAN_HDLC 1564 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1565 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1566 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1567 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1568 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1569 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1570 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1571 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1572 #define IFF_TEAM_PORT IFF_TEAM_PORT 1573 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1574 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1575 #define IFF_MACVLAN IFF_MACVLAN 1576 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1577 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1578 #define IFF_NO_QUEUE IFF_NO_QUEUE 1579 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1580 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1581 #define IFF_TEAM IFF_TEAM 1582 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1583 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM 1584 #define IFF_MACSEC IFF_MACSEC 1585 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1586 #define IFF_FAILOVER IFF_FAILOVER 1587 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1588 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1589 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1590 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR 1591 1592 /** 1593 * struct net_device - The DEVICE structure. 1594 * 1595 * Actually, this whole structure is a big mistake. It mixes I/O 1596 * data with strictly "high-level" data, and it has to know about 1597 * almost every data structure used in the INET module. 1598 * 1599 * @name: This is the first field of the "visible" part of this structure 1600 * (i.e. as seen by users in the "Space.c" file). It is the name 1601 * of the interface. 1602 * 1603 * @name_node: Name hashlist node 1604 * @ifalias: SNMP alias 1605 * @mem_end: Shared memory end 1606 * @mem_start: Shared memory start 1607 * @base_addr: Device I/O address 1608 * @irq: Device IRQ number 1609 * 1610 * @state: Generic network queuing layer state, see netdev_state_t 1611 * @dev_list: The global list of network devices 1612 * @napi_list: List entry used for polling NAPI devices 1613 * @unreg_list: List entry when we are unregistering the 1614 * device; see the function unregister_netdev 1615 * @close_list: List entry used when we are closing the device 1616 * @ptype_all: Device-specific packet handlers for all protocols 1617 * @ptype_specific: Device-specific, protocol-specific packet handlers 1618 * 1619 * @adj_list: Directly linked devices, like slaves for bonding 1620 * @features: Currently active device features 1621 * @hw_features: User-changeable features 1622 * 1623 * @wanted_features: User-requested features 1624 * @vlan_features: Mask of features inheritable by VLAN devices 1625 * 1626 * @hw_enc_features: Mask of features inherited by encapsulating devices 1627 * This field indicates what encapsulation 1628 * offloads the hardware is capable of doing, 1629 * and drivers will need to set them appropriately. 1630 * 1631 * @mpls_features: Mask of features inheritable by MPLS 1632 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1633 * 1634 * @ifindex: interface index 1635 * @group: The group the device belongs to 1636 * 1637 * @stats: Statistics struct, which was left as a legacy, use 1638 * rtnl_link_stats64 instead 1639 * 1640 * @rx_dropped: Dropped packets by core network, 1641 * do not use this in drivers 1642 * @tx_dropped: Dropped packets by core network, 1643 * do not use this in drivers 1644 * @rx_nohandler: nohandler dropped packets by core network on 1645 * inactive devices, do not use this in drivers 1646 * @carrier_up_count: Number of times the carrier has been up 1647 * @carrier_down_count: Number of times the carrier has been down 1648 * 1649 * @wireless_handlers: List of functions to handle Wireless Extensions, 1650 * instead of ioctl, 1651 * see <net/iw_handler.h> for details. 1652 * @wireless_data: Instance data managed by the core of wireless extensions 1653 * 1654 * @netdev_ops: Includes several pointers to callbacks, 1655 * if one wants to override the ndo_*() functions 1656 * @ethtool_ops: Management operations 1657 * @l3mdev_ops: Layer 3 master device operations 1658 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1659 * discovery handling. Necessary for e.g. 6LoWPAN. 1660 * @xfrmdev_ops: Transformation offload operations 1661 * @tlsdev_ops: Transport Layer Security offload operations 1662 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1663 * of Layer 2 headers. 1664 * 1665 * @flags: Interface flags (a la BSD) 1666 * @priv_flags: Like 'flags' but invisible to userspace, 1667 * see if.h for the definitions 1668 * @gflags: Global flags ( kept as legacy ) 1669 * @padded: How much padding added by alloc_netdev() 1670 * @operstate: RFC2863 operstate 1671 * @link_mode: Mapping policy to operstate 1672 * @if_port: Selectable AUI, TP, ... 1673 * @dma: DMA channel 1674 * @mtu: Interface MTU value 1675 * @min_mtu: Interface Minimum MTU value 1676 * @max_mtu: Interface Maximum MTU value 1677 * @type: Interface hardware type 1678 * @hard_header_len: Maximum hardware header length. 1679 * @min_header_len: Minimum hardware header length 1680 * 1681 * @needed_headroom: Extra headroom the hardware may need, but not in all 1682 * cases can this be guaranteed 1683 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1684 * cases can this be guaranteed. Some cases also use 1685 * LL_MAX_HEADER instead to allocate the skb 1686 * 1687 * interface address info: 1688 * 1689 * @perm_addr: Permanent hw address 1690 * @addr_assign_type: Hw address assignment type 1691 * @addr_len: Hardware address length 1692 * @upper_level: Maximum depth level of upper devices. 1693 * @lower_level: Maximum depth level of lower devices. 1694 * @neigh_priv_len: Used in neigh_alloc() 1695 * @dev_id: Used to differentiate devices that share 1696 * the same link layer address 1697 * @dev_port: Used to differentiate devices that share 1698 * the same function 1699 * @addr_list_lock: XXX: need comments on this one 1700 * @name_assign_type: network interface name assignment type 1701 * @uc_promisc: Counter that indicates promiscuous mode 1702 * has been enabled due to the need to listen to 1703 * additional unicast addresses in a device that 1704 * does not implement ndo_set_rx_mode() 1705 * @uc: unicast mac addresses 1706 * @mc: multicast mac addresses 1707 * @dev_addrs: list of device hw addresses 1708 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1709 * @promiscuity: Number of times the NIC is told to work in 1710 * promiscuous mode; if it becomes 0 the NIC will 1711 * exit promiscuous mode 1712 * @allmulti: Counter, enables or disables allmulticast mode 1713 * 1714 * @vlan_info: VLAN info 1715 * @dsa_ptr: dsa specific data 1716 * @tipc_ptr: TIPC specific data 1717 * @atalk_ptr: AppleTalk link 1718 * @ip_ptr: IPv4 specific data 1719 * @dn_ptr: DECnet specific data 1720 * @ip6_ptr: IPv6 specific data 1721 * @ax25_ptr: AX.25 specific data 1722 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1723 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1724 * device struct 1725 * @mpls_ptr: mpls_dev struct pointer 1726 * 1727 * @dev_addr: Hw address (before bcast, 1728 * because most packets are unicast) 1729 * 1730 * @_rx: Array of RX queues 1731 * @num_rx_queues: Number of RX queues 1732 * allocated at register_netdev() time 1733 * @real_num_rx_queues: Number of RX queues currently active in device 1734 * @xdp_prog: XDP sockets filter program pointer 1735 * @gro_flush_timeout: timeout for GRO layer in NAPI 1736 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1737 * allow to avoid NIC hard IRQ, on busy queues. 1738 * 1739 * @rx_handler: handler for received packets 1740 * @rx_handler_data: XXX: need comments on this one 1741 * @miniq_ingress: ingress/clsact qdisc specific data for 1742 * ingress processing 1743 * @ingress_queue: XXX: need comments on this one 1744 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1745 * @broadcast: hw bcast address 1746 * 1747 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1748 * indexed by RX queue number. Assigned by driver. 1749 * This must only be set if the ndo_rx_flow_steer 1750 * operation is defined 1751 * @index_hlist: Device index hash chain 1752 * 1753 * @_tx: Array of TX queues 1754 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1755 * @real_num_tx_queues: Number of TX queues currently active in device 1756 * @qdisc: Root qdisc from userspace point of view 1757 * @tx_queue_len: Max frames per queue allowed 1758 * @tx_global_lock: XXX: need comments on this one 1759 * @xdp_bulkq: XDP device bulk queue 1760 * @xps_cpus_map: all CPUs map for XPS device 1761 * @xps_rxqs_map: all RXQs map for XPS device 1762 * 1763 * @xps_maps: XXX: need comments on this one 1764 * @miniq_egress: clsact qdisc specific data for 1765 * egress processing 1766 * @qdisc_hash: qdisc hash table 1767 * @watchdog_timeo: Represents the timeout that is used by 1768 * the watchdog (see dev_watchdog()) 1769 * @watchdog_timer: List of timers 1770 * 1771 * @proto_down_reason: reason a netdev interface is held down 1772 * @pcpu_refcnt: Number of references to this device 1773 * @todo_list: Delayed register/unregister 1774 * @link_watch_list: XXX: need comments on this one 1775 * 1776 * @reg_state: Register/unregister state machine 1777 * @dismantle: Device is going to be freed 1778 * @rtnl_link_state: This enum represents the phases of creating 1779 * a new link 1780 * 1781 * @needs_free_netdev: Should unregister perform free_netdev? 1782 * @priv_destructor: Called from unregister 1783 * @npinfo: XXX: need comments on this one 1784 * @nd_net: Network namespace this network device is inside 1785 * 1786 * @ml_priv: Mid-layer private 1787 * @lstats: Loopback statistics 1788 * @tstats: Tunnel statistics 1789 * @dstats: Dummy statistics 1790 * @vstats: Virtual ethernet statistics 1791 * 1792 * @garp_port: GARP 1793 * @mrp_port: MRP 1794 * 1795 * @dev: Class/net/name entry 1796 * @sysfs_groups: Space for optional device, statistics and wireless 1797 * sysfs groups 1798 * 1799 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1800 * @rtnl_link_ops: Rtnl_link_ops 1801 * 1802 * @gso_max_size: Maximum size of generic segmentation offload 1803 * @gso_max_segs: Maximum number of segments that can be passed to the 1804 * NIC for GSO 1805 * 1806 * @dcbnl_ops: Data Center Bridging netlink ops 1807 * @num_tc: Number of traffic classes in the net device 1808 * @tc_to_txq: XXX: need comments on this one 1809 * @prio_tc_map: XXX: need comments on this one 1810 * 1811 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1812 * 1813 * @priomap: XXX: need comments on this one 1814 * @phydev: Physical device may attach itself 1815 * for hardware timestamping 1816 * @sfp_bus: attached &struct sfp_bus structure. 1817 * 1818 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1819 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1820 * 1821 * @proto_down: protocol port state information can be sent to the 1822 * switch driver and used to set the phys state of the 1823 * switch port. 1824 * 1825 * @wol_enabled: Wake-on-LAN is enabled 1826 * 1827 * @threaded: napi threaded mode is enabled 1828 * 1829 * @net_notifier_list: List of per-net netdev notifier block 1830 * that follow this device when it is moved 1831 * to another network namespace. 1832 * 1833 * @macsec_ops: MACsec offloading ops 1834 * 1835 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1836 * offload capabilities of the device 1837 * @udp_tunnel_nic: UDP tunnel offload state 1838 * @xdp_state: stores info on attached XDP BPF programs 1839 * 1840 * @nested_level: Used as as a parameter of spin_lock_nested() of 1841 * dev->addr_list_lock. 1842 * @unlink_list: As netif_addr_lock() can be called recursively, 1843 * keep a list of interfaces to be deleted. 1844 * 1845 * FIXME: cleanup struct net_device such that network protocol info 1846 * moves out. 1847 */ 1848 1849 struct net_device { 1850 char name[IFNAMSIZ]; 1851 struct netdev_name_node *name_node; 1852 struct dev_ifalias __rcu *ifalias; 1853 /* 1854 * I/O specific fields 1855 * FIXME: Merge these and struct ifmap into one 1856 */ 1857 unsigned long mem_end; 1858 unsigned long mem_start; 1859 unsigned long base_addr; 1860 1861 /* 1862 * Some hardware also needs these fields (state,dev_list, 1863 * napi_list,unreg_list,close_list) but they are not 1864 * part of the usual set specified in Space.c. 1865 */ 1866 1867 unsigned long state; 1868 1869 struct list_head dev_list; 1870 struct list_head napi_list; 1871 struct list_head unreg_list; 1872 struct list_head close_list; 1873 struct list_head ptype_all; 1874 struct list_head ptype_specific; 1875 1876 struct { 1877 struct list_head upper; 1878 struct list_head lower; 1879 } adj_list; 1880 1881 /* Read-mostly cache-line for fast-path access */ 1882 unsigned int flags; 1883 unsigned int priv_flags; 1884 const struct net_device_ops *netdev_ops; 1885 int ifindex; 1886 unsigned short gflags; 1887 unsigned short hard_header_len; 1888 1889 /* Note : dev->mtu is often read without holding a lock. 1890 * Writers usually hold RTNL. 1891 * It is recommended to use READ_ONCE() to annotate the reads, 1892 * and to use WRITE_ONCE() to annotate the writes. 1893 */ 1894 unsigned int mtu; 1895 unsigned short needed_headroom; 1896 unsigned short needed_tailroom; 1897 1898 netdev_features_t features; 1899 netdev_features_t hw_features; 1900 netdev_features_t wanted_features; 1901 netdev_features_t vlan_features; 1902 netdev_features_t hw_enc_features; 1903 netdev_features_t mpls_features; 1904 netdev_features_t gso_partial_features; 1905 1906 unsigned int min_mtu; 1907 unsigned int max_mtu; 1908 unsigned short type; 1909 unsigned char min_header_len; 1910 unsigned char name_assign_type; 1911 1912 int group; 1913 1914 struct net_device_stats stats; /* not used by modern drivers */ 1915 1916 atomic_long_t rx_dropped; 1917 atomic_long_t tx_dropped; 1918 atomic_long_t rx_nohandler; 1919 1920 /* Stats to monitor link on/off, flapping */ 1921 atomic_t carrier_up_count; 1922 atomic_t carrier_down_count; 1923 1924 #ifdef CONFIG_WIRELESS_EXT 1925 const struct iw_handler_def *wireless_handlers; 1926 struct iw_public_data *wireless_data; 1927 #endif 1928 const struct ethtool_ops *ethtool_ops; 1929 #ifdef CONFIG_NET_L3_MASTER_DEV 1930 const struct l3mdev_ops *l3mdev_ops; 1931 #endif 1932 #if IS_ENABLED(CONFIG_IPV6) 1933 const struct ndisc_ops *ndisc_ops; 1934 #endif 1935 1936 #ifdef CONFIG_XFRM_OFFLOAD 1937 const struct xfrmdev_ops *xfrmdev_ops; 1938 #endif 1939 1940 #if IS_ENABLED(CONFIG_TLS_DEVICE) 1941 const struct tlsdev_ops *tlsdev_ops; 1942 #endif 1943 1944 const struct header_ops *header_ops; 1945 1946 unsigned char operstate; 1947 unsigned char link_mode; 1948 1949 unsigned char if_port; 1950 unsigned char dma; 1951 1952 /* Interface address info. */ 1953 unsigned char perm_addr[MAX_ADDR_LEN]; 1954 unsigned char addr_assign_type; 1955 unsigned char addr_len; 1956 unsigned char upper_level; 1957 unsigned char lower_level; 1958 1959 unsigned short neigh_priv_len; 1960 unsigned short dev_id; 1961 unsigned short dev_port; 1962 unsigned short padded; 1963 1964 spinlock_t addr_list_lock; 1965 int irq; 1966 1967 struct netdev_hw_addr_list uc; 1968 struct netdev_hw_addr_list mc; 1969 struct netdev_hw_addr_list dev_addrs; 1970 1971 #ifdef CONFIG_SYSFS 1972 struct kset *queues_kset; 1973 #endif 1974 #ifdef CONFIG_LOCKDEP 1975 struct list_head unlink_list; 1976 #endif 1977 unsigned int promiscuity; 1978 unsigned int allmulti; 1979 bool uc_promisc; 1980 #ifdef CONFIG_LOCKDEP 1981 unsigned char nested_level; 1982 #endif 1983 1984 1985 /* Protocol-specific pointers */ 1986 1987 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1988 struct vlan_info __rcu *vlan_info; 1989 #endif 1990 #if IS_ENABLED(CONFIG_NET_DSA) 1991 struct dsa_port *dsa_ptr; 1992 #endif 1993 #if IS_ENABLED(CONFIG_TIPC) 1994 struct tipc_bearer __rcu *tipc_ptr; 1995 #endif 1996 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 1997 void *atalk_ptr; 1998 #endif 1999 struct in_device __rcu *ip_ptr; 2000 #if IS_ENABLED(CONFIG_DECNET) 2001 struct dn_dev __rcu *dn_ptr; 2002 #endif 2003 struct inet6_dev __rcu *ip6_ptr; 2004 #if IS_ENABLED(CONFIG_AX25) 2005 void *ax25_ptr; 2006 #endif 2007 struct wireless_dev *ieee80211_ptr; 2008 struct wpan_dev *ieee802154_ptr; 2009 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2010 struct mpls_dev __rcu *mpls_ptr; 2011 #endif 2012 2013 /* 2014 * Cache lines mostly used on receive path (including eth_type_trans()) 2015 */ 2016 /* Interface address info used in eth_type_trans() */ 2017 unsigned char *dev_addr; 2018 2019 struct netdev_rx_queue *_rx; 2020 unsigned int num_rx_queues; 2021 unsigned int real_num_rx_queues; 2022 2023 struct bpf_prog __rcu *xdp_prog; 2024 unsigned long gro_flush_timeout; 2025 int napi_defer_hard_irqs; 2026 rx_handler_func_t __rcu *rx_handler; 2027 void __rcu *rx_handler_data; 2028 2029 #ifdef CONFIG_NET_CLS_ACT 2030 struct mini_Qdisc __rcu *miniq_ingress; 2031 #endif 2032 struct netdev_queue __rcu *ingress_queue; 2033 #ifdef CONFIG_NETFILTER_INGRESS 2034 struct nf_hook_entries __rcu *nf_hooks_ingress; 2035 #endif 2036 2037 unsigned char broadcast[MAX_ADDR_LEN]; 2038 #ifdef CONFIG_RFS_ACCEL 2039 struct cpu_rmap *rx_cpu_rmap; 2040 #endif 2041 struct hlist_node index_hlist; 2042 2043 /* 2044 * Cache lines mostly used on transmit path 2045 */ 2046 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2047 unsigned int num_tx_queues; 2048 unsigned int real_num_tx_queues; 2049 struct Qdisc *qdisc; 2050 unsigned int tx_queue_len; 2051 spinlock_t tx_global_lock; 2052 2053 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2054 2055 #ifdef CONFIG_XPS 2056 struct xps_dev_maps __rcu *xps_cpus_map; 2057 struct xps_dev_maps __rcu *xps_rxqs_map; 2058 #endif 2059 #ifdef CONFIG_NET_CLS_ACT 2060 struct mini_Qdisc __rcu *miniq_egress; 2061 #endif 2062 2063 #ifdef CONFIG_NET_SCHED 2064 DECLARE_HASHTABLE (qdisc_hash, 4); 2065 #endif 2066 /* These may be needed for future network-power-down code. */ 2067 struct timer_list watchdog_timer; 2068 int watchdog_timeo; 2069 2070 u32 proto_down_reason; 2071 2072 struct list_head todo_list; 2073 int __percpu *pcpu_refcnt; 2074 2075 struct list_head link_watch_list; 2076 2077 enum { NETREG_UNINITIALIZED=0, 2078 NETREG_REGISTERED, /* completed register_netdevice */ 2079 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2080 NETREG_UNREGISTERED, /* completed unregister todo */ 2081 NETREG_RELEASED, /* called free_netdev */ 2082 NETREG_DUMMY, /* dummy device for NAPI poll */ 2083 } reg_state:8; 2084 2085 bool dismantle; 2086 2087 enum { 2088 RTNL_LINK_INITIALIZED, 2089 RTNL_LINK_INITIALIZING, 2090 } rtnl_link_state:16; 2091 2092 bool needs_free_netdev; 2093 void (*priv_destructor)(struct net_device *dev); 2094 2095 #ifdef CONFIG_NETPOLL 2096 struct netpoll_info __rcu *npinfo; 2097 #endif 2098 2099 possible_net_t nd_net; 2100 2101 /* mid-layer private */ 2102 union { 2103 void *ml_priv; 2104 struct pcpu_lstats __percpu *lstats; 2105 struct pcpu_sw_netstats __percpu *tstats; 2106 struct pcpu_dstats __percpu *dstats; 2107 }; 2108 2109 #if IS_ENABLED(CONFIG_GARP) 2110 struct garp_port __rcu *garp_port; 2111 #endif 2112 #if IS_ENABLED(CONFIG_MRP) 2113 struct mrp_port __rcu *mrp_port; 2114 #endif 2115 2116 struct device dev; 2117 const struct attribute_group *sysfs_groups[4]; 2118 const struct attribute_group *sysfs_rx_queue_group; 2119 2120 const struct rtnl_link_ops *rtnl_link_ops; 2121 2122 /* for setting kernel sock attribute on TCP connection setup */ 2123 #define GSO_MAX_SIZE 65536 2124 unsigned int gso_max_size; 2125 #define GSO_MAX_SEGS 65535 2126 u16 gso_max_segs; 2127 2128 #ifdef CONFIG_DCB 2129 const struct dcbnl_rtnl_ops *dcbnl_ops; 2130 #endif 2131 s16 num_tc; 2132 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2133 u8 prio_tc_map[TC_BITMASK + 1]; 2134 2135 #if IS_ENABLED(CONFIG_FCOE) 2136 unsigned int fcoe_ddp_xid; 2137 #endif 2138 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2139 struct netprio_map __rcu *priomap; 2140 #endif 2141 struct phy_device *phydev; 2142 struct sfp_bus *sfp_bus; 2143 struct lock_class_key *qdisc_tx_busylock; 2144 struct lock_class_key *qdisc_running_key; 2145 bool proto_down; 2146 unsigned wol_enabled:1; 2147 unsigned threaded:1; 2148 2149 struct list_head net_notifier_list; 2150 2151 #if IS_ENABLED(CONFIG_MACSEC) 2152 /* MACsec management functions */ 2153 const struct macsec_ops *macsec_ops; 2154 #endif 2155 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2156 struct udp_tunnel_nic *udp_tunnel_nic; 2157 2158 /* protected by rtnl_lock */ 2159 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2160 }; 2161 #define to_net_dev(d) container_of(d, struct net_device, dev) 2162 2163 static inline bool netif_elide_gro(const struct net_device *dev) 2164 { 2165 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2166 return true; 2167 return false; 2168 } 2169 2170 #define NETDEV_ALIGN 32 2171 2172 static inline 2173 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2174 { 2175 return dev->prio_tc_map[prio & TC_BITMASK]; 2176 } 2177 2178 static inline 2179 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2180 { 2181 if (tc >= dev->num_tc) 2182 return -EINVAL; 2183 2184 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2185 return 0; 2186 } 2187 2188 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2189 void netdev_reset_tc(struct net_device *dev); 2190 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2191 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2192 2193 static inline 2194 int netdev_get_num_tc(struct net_device *dev) 2195 { 2196 return dev->num_tc; 2197 } 2198 2199 static inline void net_prefetch(void *p) 2200 { 2201 prefetch(p); 2202 #if L1_CACHE_BYTES < 128 2203 prefetch((u8 *)p + L1_CACHE_BYTES); 2204 #endif 2205 } 2206 2207 static inline void net_prefetchw(void *p) 2208 { 2209 prefetchw(p); 2210 #if L1_CACHE_BYTES < 128 2211 prefetchw((u8 *)p + L1_CACHE_BYTES); 2212 #endif 2213 } 2214 2215 void netdev_unbind_sb_channel(struct net_device *dev, 2216 struct net_device *sb_dev); 2217 int netdev_bind_sb_channel_queue(struct net_device *dev, 2218 struct net_device *sb_dev, 2219 u8 tc, u16 count, u16 offset); 2220 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2221 static inline int netdev_get_sb_channel(struct net_device *dev) 2222 { 2223 return max_t(int, -dev->num_tc, 0); 2224 } 2225 2226 static inline 2227 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2228 unsigned int index) 2229 { 2230 return &dev->_tx[index]; 2231 } 2232 2233 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2234 const struct sk_buff *skb) 2235 { 2236 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2237 } 2238 2239 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2240 void (*f)(struct net_device *, 2241 struct netdev_queue *, 2242 void *), 2243 void *arg) 2244 { 2245 unsigned int i; 2246 2247 for (i = 0; i < dev->num_tx_queues; i++) 2248 f(dev, &dev->_tx[i], arg); 2249 } 2250 2251 #define netdev_lockdep_set_classes(dev) \ 2252 { \ 2253 static struct lock_class_key qdisc_tx_busylock_key; \ 2254 static struct lock_class_key qdisc_running_key; \ 2255 static struct lock_class_key qdisc_xmit_lock_key; \ 2256 static struct lock_class_key dev_addr_list_lock_key; \ 2257 unsigned int i; \ 2258 \ 2259 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2260 (dev)->qdisc_running_key = &qdisc_running_key; \ 2261 lockdep_set_class(&(dev)->addr_list_lock, \ 2262 &dev_addr_list_lock_key); \ 2263 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2264 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2265 &qdisc_xmit_lock_key); \ 2266 } 2267 2268 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2269 struct net_device *sb_dev); 2270 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2271 struct sk_buff *skb, 2272 struct net_device *sb_dev); 2273 2274 /* returns the headroom that the master device needs to take in account 2275 * when forwarding to this dev 2276 */ 2277 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2278 { 2279 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2280 } 2281 2282 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2283 { 2284 if (dev->netdev_ops->ndo_set_rx_headroom) 2285 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2286 } 2287 2288 /* set the device rx headroom to the dev's default */ 2289 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2290 { 2291 netdev_set_rx_headroom(dev, -1); 2292 } 2293 2294 /* 2295 * Net namespace inlines 2296 */ 2297 static inline 2298 struct net *dev_net(const struct net_device *dev) 2299 { 2300 return read_pnet(&dev->nd_net); 2301 } 2302 2303 static inline 2304 void dev_net_set(struct net_device *dev, struct net *net) 2305 { 2306 write_pnet(&dev->nd_net, net); 2307 } 2308 2309 /** 2310 * netdev_priv - access network device private data 2311 * @dev: network device 2312 * 2313 * Get network device private data 2314 */ 2315 static inline void *netdev_priv(const struct net_device *dev) 2316 { 2317 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2318 } 2319 2320 /* Set the sysfs physical device reference for the network logical device 2321 * if set prior to registration will cause a symlink during initialization. 2322 */ 2323 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2324 2325 /* Set the sysfs device type for the network logical device to allow 2326 * fine-grained identification of different network device types. For 2327 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2328 */ 2329 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2330 2331 /* Default NAPI poll() weight 2332 * Device drivers are strongly advised to not use bigger value 2333 */ 2334 #define NAPI_POLL_WEIGHT 64 2335 2336 /** 2337 * netif_napi_add - initialize a NAPI context 2338 * @dev: network device 2339 * @napi: NAPI context 2340 * @poll: polling function 2341 * @weight: default weight 2342 * 2343 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2344 * *any* of the other NAPI-related functions. 2345 */ 2346 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2347 int (*poll)(struct napi_struct *, int), int weight); 2348 2349 /** 2350 * netif_tx_napi_add - initialize a NAPI context 2351 * @dev: network device 2352 * @napi: NAPI context 2353 * @poll: polling function 2354 * @weight: default weight 2355 * 2356 * This variant of netif_napi_add() should be used from drivers using NAPI 2357 * to exclusively poll a TX queue. 2358 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2359 */ 2360 static inline void netif_tx_napi_add(struct net_device *dev, 2361 struct napi_struct *napi, 2362 int (*poll)(struct napi_struct *, int), 2363 int weight) 2364 { 2365 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2366 netif_napi_add(dev, napi, poll, weight); 2367 } 2368 2369 /** 2370 * __netif_napi_del - remove a NAPI context 2371 * @napi: NAPI context 2372 * 2373 * Warning: caller must observe RCU grace period before freeing memory 2374 * containing @napi. Drivers might want to call this helper to combine 2375 * all the needed RCU grace periods into a single one. 2376 */ 2377 void __netif_napi_del(struct napi_struct *napi); 2378 2379 /** 2380 * netif_napi_del - remove a NAPI context 2381 * @napi: NAPI context 2382 * 2383 * netif_napi_del() removes a NAPI context from the network device NAPI list 2384 */ 2385 static inline void netif_napi_del(struct napi_struct *napi) 2386 { 2387 __netif_napi_del(napi); 2388 synchronize_net(); 2389 } 2390 2391 struct napi_gro_cb { 2392 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2393 void *frag0; 2394 2395 /* Length of frag0. */ 2396 unsigned int frag0_len; 2397 2398 /* This indicates where we are processing relative to skb->data. */ 2399 int data_offset; 2400 2401 /* This is non-zero if the packet cannot be merged with the new skb. */ 2402 u16 flush; 2403 2404 /* Save the IP ID here and check when we get to the transport layer */ 2405 u16 flush_id; 2406 2407 /* Number of segments aggregated. */ 2408 u16 count; 2409 2410 /* Start offset for remote checksum offload */ 2411 u16 gro_remcsum_start; 2412 2413 /* jiffies when first packet was created/queued */ 2414 unsigned long age; 2415 2416 /* Used in ipv6_gro_receive() and foo-over-udp */ 2417 u16 proto; 2418 2419 /* This is non-zero if the packet may be of the same flow. */ 2420 u8 same_flow:1; 2421 2422 /* Used in tunnel GRO receive */ 2423 u8 encap_mark:1; 2424 2425 /* GRO checksum is valid */ 2426 u8 csum_valid:1; 2427 2428 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2429 u8 csum_cnt:3; 2430 2431 /* Free the skb? */ 2432 u8 free:2; 2433 #define NAPI_GRO_FREE 1 2434 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2435 2436 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2437 u8 is_ipv6:1; 2438 2439 /* Used in GRE, set in fou/gue_gro_receive */ 2440 u8 is_fou:1; 2441 2442 /* Used to determine if flush_id can be ignored */ 2443 u8 is_atomic:1; 2444 2445 /* Number of gro_receive callbacks this packet already went through */ 2446 u8 recursion_counter:4; 2447 2448 /* GRO is done by frag_list pointer chaining. */ 2449 u8 is_flist:1; 2450 2451 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2452 __wsum csum; 2453 2454 /* used in skb_gro_receive() slow path */ 2455 struct sk_buff *last; 2456 }; 2457 2458 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2459 2460 #define GRO_RECURSION_LIMIT 15 2461 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2462 { 2463 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2464 } 2465 2466 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2467 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2468 struct list_head *head, 2469 struct sk_buff *skb) 2470 { 2471 if (unlikely(gro_recursion_inc_test(skb))) { 2472 NAPI_GRO_CB(skb)->flush |= 1; 2473 return NULL; 2474 } 2475 2476 return cb(head, skb); 2477 } 2478 2479 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2480 struct sk_buff *); 2481 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2482 struct sock *sk, 2483 struct list_head *head, 2484 struct sk_buff *skb) 2485 { 2486 if (unlikely(gro_recursion_inc_test(skb))) { 2487 NAPI_GRO_CB(skb)->flush |= 1; 2488 return NULL; 2489 } 2490 2491 return cb(sk, head, skb); 2492 } 2493 2494 struct packet_type { 2495 __be16 type; /* This is really htons(ether_type). */ 2496 bool ignore_outgoing; 2497 struct net_device *dev; /* NULL is wildcarded here */ 2498 int (*func) (struct sk_buff *, 2499 struct net_device *, 2500 struct packet_type *, 2501 struct net_device *); 2502 void (*list_func) (struct list_head *, 2503 struct packet_type *, 2504 struct net_device *); 2505 bool (*id_match)(struct packet_type *ptype, 2506 struct sock *sk); 2507 void *af_packet_priv; 2508 struct list_head list; 2509 }; 2510 2511 struct offload_callbacks { 2512 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2513 netdev_features_t features); 2514 struct sk_buff *(*gro_receive)(struct list_head *head, 2515 struct sk_buff *skb); 2516 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2517 }; 2518 2519 struct packet_offload { 2520 __be16 type; /* This is really htons(ether_type). */ 2521 u16 priority; 2522 struct offload_callbacks callbacks; 2523 struct list_head list; 2524 }; 2525 2526 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2527 struct pcpu_sw_netstats { 2528 u64 rx_packets; 2529 u64 rx_bytes; 2530 u64 tx_packets; 2531 u64 tx_bytes; 2532 struct u64_stats_sync syncp; 2533 } __aligned(4 * sizeof(u64)); 2534 2535 struct pcpu_lstats { 2536 u64_stats_t packets; 2537 u64_stats_t bytes; 2538 struct u64_stats_sync syncp; 2539 } __aligned(2 * sizeof(u64)); 2540 2541 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2542 2543 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2544 { 2545 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2546 2547 u64_stats_update_begin(&tstats->syncp); 2548 tstats->rx_bytes += len; 2549 tstats->rx_packets++; 2550 u64_stats_update_end(&tstats->syncp); 2551 } 2552 2553 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2554 unsigned int packets, 2555 unsigned int len) 2556 { 2557 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2558 2559 u64_stats_update_begin(&tstats->syncp); 2560 tstats->tx_bytes += len; 2561 tstats->tx_packets += packets; 2562 u64_stats_update_end(&tstats->syncp); 2563 } 2564 2565 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2566 { 2567 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2568 2569 u64_stats_update_begin(&lstats->syncp); 2570 u64_stats_add(&lstats->bytes, len); 2571 u64_stats_inc(&lstats->packets); 2572 u64_stats_update_end(&lstats->syncp); 2573 } 2574 2575 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2576 ({ \ 2577 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2578 if (pcpu_stats) { \ 2579 int __cpu; \ 2580 for_each_possible_cpu(__cpu) { \ 2581 typeof(type) *stat; \ 2582 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2583 u64_stats_init(&stat->syncp); \ 2584 } \ 2585 } \ 2586 pcpu_stats; \ 2587 }) 2588 2589 #define netdev_alloc_pcpu_stats(type) \ 2590 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2591 2592 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2593 ({ \ 2594 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2595 if (pcpu_stats) { \ 2596 int __cpu; \ 2597 for_each_possible_cpu(__cpu) { \ 2598 typeof(type) *stat; \ 2599 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2600 u64_stats_init(&stat->syncp); \ 2601 } \ 2602 } \ 2603 pcpu_stats; \ 2604 }) 2605 2606 enum netdev_lag_tx_type { 2607 NETDEV_LAG_TX_TYPE_UNKNOWN, 2608 NETDEV_LAG_TX_TYPE_RANDOM, 2609 NETDEV_LAG_TX_TYPE_BROADCAST, 2610 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2611 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2612 NETDEV_LAG_TX_TYPE_HASH, 2613 }; 2614 2615 enum netdev_lag_hash { 2616 NETDEV_LAG_HASH_NONE, 2617 NETDEV_LAG_HASH_L2, 2618 NETDEV_LAG_HASH_L34, 2619 NETDEV_LAG_HASH_L23, 2620 NETDEV_LAG_HASH_E23, 2621 NETDEV_LAG_HASH_E34, 2622 NETDEV_LAG_HASH_VLAN_SRCMAC, 2623 NETDEV_LAG_HASH_UNKNOWN, 2624 }; 2625 2626 struct netdev_lag_upper_info { 2627 enum netdev_lag_tx_type tx_type; 2628 enum netdev_lag_hash hash_type; 2629 }; 2630 2631 struct netdev_lag_lower_state_info { 2632 u8 link_up : 1, 2633 tx_enabled : 1; 2634 }; 2635 2636 #include <linux/notifier.h> 2637 2638 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2639 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2640 * adding new types. 2641 */ 2642 enum netdev_cmd { 2643 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2644 NETDEV_DOWN, 2645 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2646 detected a hardware crash and restarted 2647 - we can use this eg to kick tcp sessions 2648 once done */ 2649 NETDEV_CHANGE, /* Notify device state change */ 2650 NETDEV_REGISTER, 2651 NETDEV_UNREGISTER, 2652 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2653 NETDEV_CHANGEADDR, /* notify after the address change */ 2654 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2655 NETDEV_GOING_DOWN, 2656 NETDEV_CHANGENAME, 2657 NETDEV_FEAT_CHANGE, 2658 NETDEV_BONDING_FAILOVER, 2659 NETDEV_PRE_UP, 2660 NETDEV_PRE_TYPE_CHANGE, 2661 NETDEV_POST_TYPE_CHANGE, 2662 NETDEV_POST_INIT, 2663 NETDEV_RELEASE, 2664 NETDEV_NOTIFY_PEERS, 2665 NETDEV_JOIN, 2666 NETDEV_CHANGEUPPER, 2667 NETDEV_RESEND_IGMP, 2668 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2669 NETDEV_CHANGEINFODATA, 2670 NETDEV_BONDING_INFO, 2671 NETDEV_PRECHANGEUPPER, 2672 NETDEV_CHANGELOWERSTATE, 2673 NETDEV_UDP_TUNNEL_PUSH_INFO, 2674 NETDEV_UDP_TUNNEL_DROP_INFO, 2675 NETDEV_CHANGE_TX_QUEUE_LEN, 2676 NETDEV_CVLAN_FILTER_PUSH_INFO, 2677 NETDEV_CVLAN_FILTER_DROP_INFO, 2678 NETDEV_SVLAN_FILTER_PUSH_INFO, 2679 NETDEV_SVLAN_FILTER_DROP_INFO, 2680 }; 2681 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2682 2683 int register_netdevice_notifier(struct notifier_block *nb); 2684 int unregister_netdevice_notifier(struct notifier_block *nb); 2685 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2686 int unregister_netdevice_notifier_net(struct net *net, 2687 struct notifier_block *nb); 2688 int register_netdevice_notifier_dev_net(struct net_device *dev, 2689 struct notifier_block *nb, 2690 struct netdev_net_notifier *nn); 2691 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2692 struct notifier_block *nb, 2693 struct netdev_net_notifier *nn); 2694 2695 struct netdev_notifier_info { 2696 struct net_device *dev; 2697 struct netlink_ext_ack *extack; 2698 }; 2699 2700 struct netdev_notifier_info_ext { 2701 struct netdev_notifier_info info; /* must be first */ 2702 union { 2703 u32 mtu; 2704 } ext; 2705 }; 2706 2707 struct netdev_notifier_change_info { 2708 struct netdev_notifier_info info; /* must be first */ 2709 unsigned int flags_changed; 2710 }; 2711 2712 struct netdev_notifier_changeupper_info { 2713 struct netdev_notifier_info info; /* must be first */ 2714 struct net_device *upper_dev; /* new upper dev */ 2715 bool master; /* is upper dev master */ 2716 bool linking; /* is the notification for link or unlink */ 2717 void *upper_info; /* upper dev info */ 2718 }; 2719 2720 struct netdev_notifier_changelowerstate_info { 2721 struct netdev_notifier_info info; /* must be first */ 2722 void *lower_state_info; /* is lower dev state */ 2723 }; 2724 2725 struct netdev_notifier_pre_changeaddr_info { 2726 struct netdev_notifier_info info; /* must be first */ 2727 const unsigned char *dev_addr; 2728 }; 2729 2730 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2731 struct net_device *dev) 2732 { 2733 info->dev = dev; 2734 info->extack = NULL; 2735 } 2736 2737 static inline struct net_device * 2738 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2739 { 2740 return info->dev; 2741 } 2742 2743 static inline struct netlink_ext_ack * 2744 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2745 { 2746 return info->extack; 2747 } 2748 2749 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2750 2751 2752 extern rwlock_t dev_base_lock; /* Device list lock */ 2753 2754 #define for_each_netdev(net, d) \ 2755 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2756 #define for_each_netdev_reverse(net, d) \ 2757 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2758 #define for_each_netdev_rcu(net, d) \ 2759 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2760 #define for_each_netdev_safe(net, d, n) \ 2761 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2762 #define for_each_netdev_continue(net, d) \ 2763 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2764 #define for_each_netdev_continue_reverse(net, d) \ 2765 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2766 dev_list) 2767 #define for_each_netdev_continue_rcu(net, d) \ 2768 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2769 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2770 for_each_netdev_rcu(&init_net, slave) \ 2771 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2772 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2773 2774 static inline struct net_device *next_net_device(struct net_device *dev) 2775 { 2776 struct list_head *lh; 2777 struct net *net; 2778 2779 net = dev_net(dev); 2780 lh = dev->dev_list.next; 2781 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2782 } 2783 2784 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2785 { 2786 struct list_head *lh; 2787 struct net *net; 2788 2789 net = dev_net(dev); 2790 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2791 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2792 } 2793 2794 static inline struct net_device *first_net_device(struct net *net) 2795 { 2796 return list_empty(&net->dev_base_head) ? NULL : 2797 net_device_entry(net->dev_base_head.next); 2798 } 2799 2800 static inline struct net_device *first_net_device_rcu(struct net *net) 2801 { 2802 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2803 2804 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2805 } 2806 2807 int netdev_boot_setup_check(struct net_device *dev); 2808 unsigned long netdev_boot_base(const char *prefix, int unit); 2809 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2810 const char *hwaddr); 2811 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2812 void dev_add_pack(struct packet_type *pt); 2813 void dev_remove_pack(struct packet_type *pt); 2814 void __dev_remove_pack(struct packet_type *pt); 2815 void dev_add_offload(struct packet_offload *po); 2816 void dev_remove_offload(struct packet_offload *po); 2817 2818 int dev_get_iflink(const struct net_device *dev); 2819 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2820 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2821 unsigned short mask); 2822 struct net_device *dev_get_by_name(struct net *net, const char *name); 2823 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2824 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2825 int dev_alloc_name(struct net_device *dev, const char *name); 2826 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2827 void dev_close(struct net_device *dev); 2828 void dev_close_many(struct list_head *head, bool unlink); 2829 void dev_disable_lro(struct net_device *dev); 2830 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2831 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2832 struct net_device *sb_dev); 2833 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2834 struct net_device *sb_dev); 2835 2836 int dev_queue_xmit(struct sk_buff *skb); 2837 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2838 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2839 2840 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 2841 { 2842 int ret; 2843 2844 ret = __dev_direct_xmit(skb, queue_id); 2845 if (!dev_xmit_complete(ret)) 2846 kfree_skb(skb); 2847 return ret; 2848 } 2849 2850 int register_netdevice(struct net_device *dev); 2851 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2852 void unregister_netdevice_many(struct list_head *head); 2853 static inline void unregister_netdevice(struct net_device *dev) 2854 { 2855 unregister_netdevice_queue(dev, NULL); 2856 } 2857 2858 int netdev_refcnt_read(const struct net_device *dev); 2859 void free_netdev(struct net_device *dev); 2860 void netdev_freemem(struct net_device *dev); 2861 int init_dummy_netdev(struct net_device *dev); 2862 2863 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 2864 struct sk_buff *skb, 2865 bool all_slaves); 2866 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 2867 struct sock *sk); 2868 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2869 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2870 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2871 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2872 int netdev_get_name(struct net *net, char *name, int ifindex); 2873 int dev_restart(struct net_device *dev); 2874 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 2875 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb); 2876 2877 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2878 { 2879 return NAPI_GRO_CB(skb)->data_offset; 2880 } 2881 2882 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2883 { 2884 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2885 } 2886 2887 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2888 { 2889 NAPI_GRO_CB(skb)->data_offset += len; 2890 } 2891 2892 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2893 unsigned int offset) 2894 { 2895 return NAPI_GRO_CB(skb)->frag0 + offset; 2896 } 2897 2898 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2899 { 2900 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2901 } 2902 2903 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2904 { 2905 NAPI_GRO_CB(skb)->frag0 = NULL; 2906 NAPI_GRO_CB(skb)->frag0_len = 0; 2907 } 2908 2909 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2910 unsigned int offset) 2911 { 2912 if (!pskb_may_pull(skb, hlen)) 2913 return NULL; 2914 2915 skb_gro_frag0_invalidate(skb); 2916 return skb->data + offset; 2917 } 2918 2919 static inline void *skb_gro_network_header(struct sk_buff *skb) 2920 { 2921 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2922 skb_network_offset(skb); 2923 } 2924 2925 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2926 const void *start, unsigned int len) 2927 { 2928 if (NAPI_GRO_CB(skb)->csum_valid) 2929 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2930 csum_partial(start, len, 0)); 2931 } 2932 2933 /* GRO checksum functions. These are logical equivalents of the normal 2934 * checksum functions (in skbuff.h) except that they operate on the GRO 2935 * offsets and fields in sk_buff. 2936 */ 2937 2938 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2939 2940 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2941 { 2942 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2943 } 2944 2945 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2946 bool zero_okay, 2947 __sum16 check) 2948 { 2949 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2950 skb_checksum_start_offset(skb) < 2951 skb_gro_offset(skb)) && 2952 !skb_at_gro_remcsum_start(skb) && 2953 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2954 (!zero_okay || check)); 2955 } 2956 2957 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2958 __wsum psum) 2959 { 2960 if (NAPI_GRO_CB(skb)->csum_valid && 2961 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2962 return 0; 2963 2964 NAPI_GRO_CB(skb)->csum = psum; 2965 2966 return __skb_gro_checksum_complete(skb); 2967 } 2968 2969 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2970 { 2971 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2972 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2973 NAPI_GRO_CB(skb)->csum_cnt--; 2974 } else { 2975 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2976 * verified a new top level checksum or an encapsulated one 2977 * during GRO. This saves work if we fallback to normal path. 2978 */ 2979 __skb_incr_checksum_unnecessary(skb); 2980 } 2981 } 2982 2983 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2984 compute_pseudo) \ 2985 ({ \ 2986 __sum16 __ret = 0; \ 2987 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2988 __ret = __skb_gro_checksum_validate_complete(skb, \ 2989 compute_pseudo(skb, proto)); \ 2990 if (!__ret) \ 2991 skb_gro_incr_csum_unnecessary(skb); \ 2992 __ret; \ 2993 }) 2994 2995 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2996 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2997 2998 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2999 compute_pseudo) \ 3000 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 3001 3002 #define skb_gro_checksum_simple_validate(skb) \ 3003 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 3004 3005 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 3006 { 3007 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 3008 !NAPI_GRO_CB(skb)->csum_valid); 3009 } 3010 3011 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 3012 __wsum pseudo) 3013 { 3014 NAPI_GRO_CB(skb)->csum = ~pseudo; 3015 NAPI_GRO_CB(skb)->csum_valid = 1; 3016 } 3017 3018 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \ 3019 do { \ 3020 if (__skb_gro_checksum_convert_check(skb)) \ 3021 __skb_gro_checksum_convert(skb, \ 3022 compute_pseudo(skb, proto)); \ 3023 } while (0) 3024 3025 struct gro_remcsum { 3026 int offset; 3027 __wsum delta; 3028 }; 3029 3030 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 3031 { 3032 grc->offset = 0; 3033 grc->delta = 0; 3034 } 3035 3036 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 3037 unsigned int off, size_t hdrlen, 3038 int start, int offset, 3039 struct gro_remcsum *grc, 3040 bool nopartial) 3041 { 3042 __wsum delta; 3043 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 3044 3045 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 3046 3047 if (!nopartial) { 3048 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 3049 return ptr; 3050 } 3051 3052 ptr = skb_gro_header_fast(skb, off); 3053 if (skb_gro_header_hard(skb, off + plen)) { 3054 ptr = skb_gro_header_slow(skb, off + plen, off); 3055 if (!ptr) 3056 return NULL; 3057 } 3058 3059 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 3060 start, offset); 3061 3062 /* Adjust skb->csum since we changed the packet */ 3063 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 3064 3065 grc->offset = off + hdrlen + offset; 3066 grc->delta = delta; 3067 3068 return ptr; 3069 } 3070 3071 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 3072 struct gro_remcsum *grc) 3073 { 3074 void *ptr; 3075 size_t plen = grc->offset + sizeof(u16); 3076 3077 if (!grc->delta) 3078 return; 3079 3080 ptr = skb_gro_header_fast(skb, grc->offset); 3081 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 3082 ptr = skb_gro_header_slow(skb, plen, grc->offset); 3083 if (!ptr) 3084 return; 3085 } 3086 3087 remcsum_unadjust((__sum16 *)ptr, grc->delta); 3088 } 3089 3090 #ifdef CONFIG_XFRM_OFFLOAD 3091 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 3092 { 3093 if (PTR_ERR(pp) != -EINPROGRESS) 3094 NAPI_GRO_CB(skb)->flush |= flush; 3095 } 3096 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 3097 struct sk_buff *pp, 3098 int flush, 3099 struct gro_remcsum *grc) 3100 { 3101 if (PTR_ERR(pp) != -EINPROGRESS) { 3102 NAPI_GRO_CB(skb)->flush |= flush; 3103 skb_gro_remcsum_cleanup(skb, grc); 3104 skb->remcsum_offload = 0; 3105 } 3106 } 3107 #else 3108 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 3109 { 3110 NAPI_GRO_CB(skb)->flush |= flush; 3111 } 3112 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 3113 struct sk_buff *pp, 3114 int flush, 3115 struct gro_remcsum *grc) 3116 { 3117 NAPI_GRO_CB(skb)->flush |= flush; 3118 skb_gro_remcsum_cleanup(skb, grc); 3119 skb->remcsum_offload = 0; 3120 } 3121 #endif 3122 3123 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3124 unsigned short type, 3125 const void *daddr, const void *saddr, 3126 unsigned int len) 3127 { 3128 if (!dev->header_ops || !dev->header_ops->create) 3129 return 0; 3130 3131 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3132 } 3133 3134 static inline int dev_parse_header(const struct sk_buff *skb, 3135 unsigned char *haddr) 3136 { 3137 const struct net_device *dev = skb->dev; 3138 3139 if (!dev->header_ops || !dev->header_ops->parse) 3140 return 0; 3141 return dev->header_ops->parse(skb, haddr); 3142 } 3143 3144 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3145 { 3146 const struct net_device *dev = skb->dev; 3147 3148 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3149 return 0; 3150 return dev->header_ops->parse_protocol(skb); 3151 } 3152 3153 /* ll_header must have at least hard_header_len allocated */ 3154 static inline bool dev_validate_header(const struct net_device *dev, 3155 char *ll_header, int len) 3156 { 3157 if (likely(len >= dev->hard_header_len)) 3158 return true; 3159 if (len < dev->min_header_len) 3160 return false; 3161 3162 if (capable(CAP_SYS_RAWIO)) { 3163 memset(ll_header + len, 0, dev->hard_header_len - len); 3164 return true; 3165 } 3166 3167 if (dev->header_ops && dev->header_ops->validate) 3168 return dev->header_ops->validate(ll_header, len); 3169 3170 return false; 3171 } 3172 3173 static inline bool dev_has_header(const struct net_device *dev) 3174 { 3175 return dev->header_ops && dev->header_ops->create; 3176 } 3177 3178 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, 3179 int len, int size); 3180 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 3181 static inline int unregister_gifconf(unsigned int family) 3182 { 3183 return register_gifconf(family, NULL); 3184 } 3185 3186 #ifdef CONFIG_NET_FLOW_LIMIT 3187 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 3188 struct sd_flow_limit { 3189 u64 count; 3190 unsigned int num_buckets; 3191 unsigned int history_head; 3192 u16 history[FLOW_LIMIT_HISTORY]; 3193 u8 buckets[]; 3194 }; 3195 3196 extern int netdev_flow_limit_table_len; 3197 #endif /* CONFIG_NET_FLOW_LIMIT */ 3198 3199 /* 3200 * Incoming packets are placed on per-CPU queues 3201 */ 3202 struct softnet_data { 3203 struct list_head poll_list; 3204 struct sk_buff_head process_queue; 3205 3206 /* stats */ 3207 unsigned int processed; 3208 unsigned int time_squeeze; 3209 unsigned int received_rps; 3210 #ifdef CONFIG_RPS 3211 struct softnet_data *rps_ipi_list; 3212 #endif 3213 #ifdef CONFIG_NET_FLOW_LIMIT 3214 struct sd_flow_limit __rcu *flow_limit; 3215 #endif 3216 struct Qdisc *output_queue; 3217 struct Qdisc **output_queue_tailp; 3218 struct sk_buff *completion_queue; 3219 #ifdef CONFIG_XFRM_OFFLOAD 3220 struct sk_buff_head xfrm_backlog; 3221 #endif 3222 /* written and read only by owning cpu: */ 3223 struct { 3224 u16 recursion; 3225 u8 more; 3226 } xmit; 3227 #ifdef CONFIG_RPS 3228 /* input_queue_head should be written by cpu owning this struct, 3229 * and only read by other cpus. Worth using a cache line. 3230 */ 3231 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3232 3233 /* Elements below can be accessed between CPUs for RPS/RFS */ 3234 call_single_data_t csd ____cacheline_aligned_in_smp; 3235 struct softnet_data *rps_ipi_next; 3236 unsigned int cpu; 3237 unsigned int input_queue_tail; 3238 #endif 3239 unsigned int dropped; 3240 struct sk_buff_head input_pkt_queue; 3241 struct napi_struct backlog; 3242 3243 }; 3244 3245 static inline void input_queue_head_incr(struct softnet_data *sd) 3246 { 3247 #ifdef CONFIG_RPS 3248 sd->input_queue_head++; 3249 #endif 3250 } 3251 3252 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3253 unsigned int *qtail) 3254 { 3255 #ifdef CONFIG_RPS 3256 *qtail = ++sd->input_queue_tail; 3257 #endif 3258 } 3259 3260 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3261 3262 static inline int dev_recursion_level(void) 3263 { 3264 return this_cpu_read(softnet_data.xmit.recursion); 3265 } 3266 3267 #define XMIT_RECURSION_LIMIT 8 3268 static inline bool dev_xmit_recursion(void) 3269 { 3270 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3271 XMIT_RECURSION_LIMIT); 3272 } 3273 3274 static inline void dev_xmit_recursion_inc(void) 3275 { 3276 __this_cpu_inc(softnet_data.xmit.recursion); 3277 } 3278 3279 static inline void dev_xmit_recursion_dec(void) 3280 { 3281 __this_cpu_dec(softnet_data.xmit.recursion); 3282 } 3283 3284 void __netif_schedule(struct Qdisc *q); 3285 void netif_schedule_queue(struct netdev_queue *txq); 3286 3287 static inline void netif_tx_schedule_all(struct net_device *dev) 3288 { 3289 unsigned int i; 3290 3291 for (i = 0; i < dev->num_tx_queues; i++) 3292 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3293 } 3294 3295 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3296 { 3297 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3298 } 3299 3300 /** 3301 * netif_start_queue - allow transmit 3302 * @dev: network device 3303 * 3304 * Allow upper layers to call the device hard_start_xmit routine. 3305 */ 3306 static inline void netif_start_queue(struct net_device *dev) 3307 { 3308 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3309 } 3310 3311 static inline void netif_tx_start_all_queues(struct net_device *dev) 3312 { 3313 unsigned int i; 3314 3315 for (i = 0; i < dev->num_tx_queues; i++) { 3316 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3317 netif_tx_start_queue(txq); 3318 } 3319 } 3320 3321 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3322 3323 /** 3324 * netif_wake_queue - restart transmit 3325 * @dev: network device 3326 * 3327 * Allow upper layers to call the device hard_start_xmit routine. 3328 * Used for flow control when transmit resources are available. 3329 */ 3330 static inline void netif_wake_queue(struct net_device *dev) 3331 { 3332 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3333 } 3334 3335 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3336 { 3337 unsigned int i; 3338 3339 for (i = 0; i < dev->num_tx_queues; i++) { 3340 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3341 netif_tx_wake_queue(txq); 3342 } 3343 } 3344 3345 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3346 { 3347 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3348 } 3349 3350 /** 3351 * netif_stop_queue - stop transmitted packets 3352 * @dev: network device 3353 * 3354 * Stop upper layers calling the device hard_start_xmit routine. 3355 * Used for flow control when transmit resources are unavailable. 3356 */ 3357 static inline void netif_stop_queue(struct net_device *dev) 3358 { 3359 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3360 } 3361 3362 void netif_tx_stop_all_queues(struct net_device *dev); 3363 3364 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3365 { 3366 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3367 } 3368 3369 /** 3370 * netif_queue_stopped - test if transmit queue is flowblocked 3371 * @dev: network device 3372 * 3373 * Test if transmit queue on device is currently unable to send. 3374 */ 3375 static inline bool netif_queue_stopped(const struct net_device *dev) 3376 { 3377 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3378 } 3379 3380 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3381 { 3382 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3383 } 3384 3385 static inline bool 3386 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3387 { 3388 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3389 } 3390 3391 static inline bool 3392 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3393 { 3394 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3395 } 3396 3397 /** 3398 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3399 * @dev_queue: pointer to transmit queue 3400 * 3401 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3402 * to give appropriate hint to the CPU. 3403 */ 3404 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3405 { 3406 #ifdef CONFIG_BQL 3407 prefetchw(&dev_queue->dql.num_queued); 3408 #endif 3409 } 3410 3411 /** 3412 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3413 * @dev_queue: pointer to transmit queue 3414 * 3415 * BQL enabled drivers might use this helper in their TX completion path, 3416 * to give appropriate hint to the CPU. 3417 */ 3418 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3419 { 3420 #ifdef CONFIG_BQL 3421 prefetchw(&dev_queue->dql.limit); 3422 #endif 3423 } 3424 3425 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3426 unsigned int bytes) 3427 { 3428 #ifdef CONFIG_BQL 3429 dql_queued(&dev_queue->dql, bytes); 3430 3431 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3432 return; 3433 3434 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3435 3436 /* 3437 * The XOFF flag must be set before checking the dql_avail below, 3438 * because in netdev_tx_completed_queue we update the dql_completed 3439 * before checking the XOFF flag. 3440 */ 3441 smp_mb(); 3442 3443 /* check again in case another CPU has just made room avail */ 3444 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3445 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3446 #endif 3447 } 3448 3449 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3450 * that they should not test BQL status themselves. 3451 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3452 * skb of a batch. 3453 * Returns true if the doorbell must be used to kick the NIC. 3454 */ 3455 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3456 unsigned int bytes, 3457 bool xmit_more) 3458 { 3459 if (xmit_more) { 3460 #ifdef CONFIG_BQL 3461 dql_queued(&dev_queue->dql, bytes); 3462 #endif 3463 return netif_tx_queue_stopped(dev_queue); 3464 } 3465 netdev_tx_sent_queue(dev_queue, bytes); 3466 return true; 3467 } 3468 3469 /** 3470 * netdev_sent_queue - report the number of bytes queued to hardware 3471 * @dev: network device 3472 * @bytes: number of bytes queued to the hardware device queue 3473 * 3474 * Report the number of bytes queued for sending/completion to the network 3475 * device hardware queue. @bytes should be a good approximation and should 3476 * exactly match netdev_completed_queue() @bytes 3477 */ 3478 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3479 { 3480 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3481 } 3482 3483 static inline bool __netdev_sent_queue(struct net_device *dev, 3484 unsigned int bytes, 3485 bool xmit_more) 3486 { 3487 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3488 xmit_more); 3489 } 3490 3491 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3492 unsigned int pkts, unsigned int bytes) 3493 { 3494 #ifdef CONFIG_BQL 3495 if (unlikely(!bytes)) 3496 return; 3497 3498 dql_completed(&dev_queue->dql, bytes); 3499 3500 /* 3501 * Without the memory barrier there is a small possiblity that 3502 * netdev_tx_sent_queue will miss the update and cause the queue to 3503 * be stopped forever 3504 */ 3505 smp_mb(); 3506 3507 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3508 return; 3509 3510 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3511 netif_schedule_queue(dev_queue); 3512 #endif 3513 } 3514 3515 /** 3516 * netdev_completed_queue - report bytes and packets completed by device 3517 * @dev: network device 3518 * @pkts: actual number of packets sent over the medium 3519 * @bytes: actual number of bytes sent over the medium 3520 * 3521 * Report the number of bytes and packets transmitted by the network device 3522 * hardware queue over the physical medium, @bytes must exactly match the 3523 * @bytes amount passed to netdev_sent_queue() 3524 */ 3525 static inline void netdev_completed_queue(struct net_device *dev, 3526 unsigned int pkts, unsigned int bytes) 3527 { 3528 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3529 } 3530 3531 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3532 { 3533 #ifdef CONFIG_BQL 3534 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3535 dql_reset(&q->dql); 3536 #endif 3537 } 3538 3539 /** 3540 * netdev_reset_queue - reset the packets and bytes count of a network device 3541 * @dev_queue: network device 3542 * 3543 * Reset the bytes and packet count of a network device and clear the 3544 * software flow control OFF bit for this network device 3545 */ 3546 static inline void netdev_reset_queue(struct net_device *dev_queue) 3547 { 3548 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3549 } 3550 3551 /** 3552 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3553 * @dev: network device 3554 * @queue_index: given tx queue index 3555 * 3556 * Returns 0 if given tx queue index >= number of device tx queues, 3557 * otherwise returns the originally passed tx queue index. 3558 */ 3559 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3560 { 3561 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3562 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3563 dev->name, queue_index, 3564 dev->real_num_tx_queues); 3565 return 0; 3566 } 3567 3568 return queue_index; 3569 } 3570 3571 /** 3572 * netif_running - test if up 3573 * @dev: network device 3574 * 3575 * Test if the device has been brought up. 3576 */ 3577 static inline bool netif_running(const struct net_device *dev) 3578 { 3579 return test_bit(__LINK_STATE_START, &dev->state); 3580 } 3581 3582 /* 3583 * Routines to manage the subqueues on a device. We only need start, 3584 * stop, and a check if it's stopped. All other device management is 3585 * done at the overall netdevice level. 3586 * Also test the device if we're multiqueue. 3587 */ 3588 3589 /** 3590 * netif_start_subqueue - allow sending packets on subqueue 3591 * @dev: network device 3592 * @queue_index: sub queue index 3593 * 3594 * Start individual transmit queue of a device with multiple transmit queues. 3595 */ 3596 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3597 { 3598 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3599 3600 netif_tx_start_queue(txq); 3601 } 3602 3603 /** 3604 * netif_stop_subqueue - stop sending packets on subqueue 3605 * @dev: network device 3606 * @queue_index: sub queue index 3607 * 3608 * Stop individual transmit queue of a device with multiple transmit queues. 3609 */ 3610 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3611 { 3612 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3613 netif_tx_stop_queue(txq); 3614 } 3615 3616 /** 3617 * __netif_subqueue_stopped - test status of subqueue 3618 * @dev: network device 3619 * @queue_index: sub queue index 3620 * 3621 * Check individual transmit queue of a device with multiple transmit queues. 3622 */ 3623 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3624 u16 queue_index) 3625 { 3626 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3627 3628 return netif_tx_queue_stopped(txq); 3629 } 3630 3631 /** 3632 * netif_subqueue_stopped - test status of subqueue 3633 * @dev: network device 3634 * @skb: sub queue buffer pointer 3635 * 3636 * Check individual transmit queue of a device with multiple transmit queues. 3637 */ 3638 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3639 struct sk_buff *skb) 3640 { 3641 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3642 } 3643 3644 /** 3645 * netif_wake_subqueue - allow sending packets on subqueue 3646 * @dev: network device 3647 * @queue_index: sub queue index 3648 * 3649 * Resume individual transmit queue of a device with multiple transmit queues. 3650 */ 3651 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3652 { 3653 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3654 3655 netif_tx_wake_queue(txq); 3656 } 3657 3658 #ifdef CONFIG_XPS 3659 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3660 u16 index); 3661 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3662 u16 index, bool is_rxqs_map); 3663 3664 /** 3665 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3666 * @j: CPU/Rx queue index 3667 * @mask: bitmask of all cpus/rx queues 3668 * @nr_bits: number of bits in the bitmask 3669 * 3670 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3671 */ 3672 static inline bool netif_attr_test_mask(unsigned long j, 3673 const unsigned long *mask, 3674 unsigned int nr_bits) 3675 { 3676 cpu_max_bits_warn(j, nr_bits); 3677 return test_bit(j, mask); 3678 } 3679 3680 /** 3681 * netif_attr_test_online - Test for online CPU/Rx queue 3682 * @j: CPU/Rx queue index 3683 * @online_mask: bitmask for CPUs/Rx queues that are online 3684 * @nr_bits: number of bits in the bitmask 3685 * 3686 * Returns true if a CPU/Rx queue is online. 3687 */ 3688 static inline bool netif_attr_test_online(unsigned long j, 3689 const unsigned long *online_mask, 3690 unsigned int nr_bits) 3691 { 3692 cpu_max_bits_warn(j, nr_bits); 3693 3694 if (online_mask) 3695 return test_bit(j, online_mask); 3696 3697 return (j < nr_bits); 3698 } 3699 3700 /** 3701 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3702 * @n: CPU/Rx queue index 3703 * @srcp: the cpumask/Rx queue mask pointer 3704 * @nr_bits: number of bits in the bitmask 3705 * 3706 * Returns >= nr_bits if no further CPUs/Rx queues set. 3707 */ 3708 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3709 unsigned int nr_bits) 3710 { 3711 /* -1 is a legal arg here. */ 3712 if (n != -1) 3713 cpu_max_bits_warn(n, nr_bits); 3714 3715 if (srcp) 3716 return find_next_bit(srcp, nr_bits, n + 1); 3717 3718 return n + 1; 3719 } 3720 3721 /** 3722 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3723 * @n: CPU/Rx queue index 3724 * @src1p: the first CPUs/Rx queues mask pointer 3725 * @src2p: the second CPUs/Rx queues mask pointer 3726 * @nr_bits: number of bits in the bitmask 3727 * 3728 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3729 */ 3730 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3731 const unsigned long *src2p, 3732 unsigned int nr_bits) 3733 { 3734 /* -1 is a legal arg here. */ 3735 if (n != -1) 3736 cpu_max_bits_warn(n, nr_bits); 3737 3738 if (src1p && src2p) 3739 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3740 else if (src1p) 3741 return find_next_bit(src1p, nr_bits, n + 1); 3742 else if (src2p) 3743 return find_next_bit(src2p, nr_bits, n + 1); 3744 3745 return n + 1; 3746 } 3747 #else 3748 static inline int netif_set_xps_queue(struct net_device *dev, 3749 const struct cpumask *mask, 3750 u16 index) 3751 { 3752 return 0; 3753 } 3754 3755 static inline int __netif_set_xps_queue(struct net_device *dev, 3756 const unsigned long *mask, 3757 u16 index, bool is_rxqs_map) 3758 { 3759 return 0; 3760 } 3761 #endif 3762 3763 /** 3764 * netif_is_multiqueue - test if device has multiple transmit queues 3765 * @dev: network device 3766 * 3767 * Check if device has multiple transmit queues 3768 */ 3769 static inline bool netif_is_multiqueue(const struct net_device *dev) 3770 { 3771 return dev->num_tx_queues > 1; 3772 } 3773 3774 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3775 3776 #ifdef CONFIG_SYSFS 3777 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3778 #else 3779 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3780 unsigned int rxqs) 3781 { 3782 dev->real_num_rx_queues = rxqs; 3783 return 0; 3784 } 3785 #endif 3786 3787 static inline struct netdev_rx_queue * 3788 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3789 { 3790 return dev->_rx + rxq; 3791 } 3792 3793 #ifdef CONFIG_SYSFS 3794 static inline unsigned int get_netdev_rx_queue_index( 3795 struct netdev_rx_queue *queue) 3796 { 3797 struct net_device *dev = queue->dev; 3798 int index = queue - dev->_rx; 3799 3800 BUG_ON(index >= dev->num_rx_queues); 3801 return index; 3802 } 3803 #endif 3804 3805 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3806 int netif_get_num_default_rss_queues(void); 3807 3808 enum skb_free_reason { 3809 SKB_REASON_CONSUMED, 3810 SKB_REASON_DROPPED, 3811 }; 3812 3813 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3814 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3815 3816 /* 3817 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3818 * interrupt context or with hardware interrupts being disabled. 3819 * (in_irq() || irqs_disabled()) 3820 * 3821 * We provide four helpers that can be used in following contexts : 3822 * 3823 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3824 * replacing kfree_skb(skb) 3825 * 3826 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3827 * Typically used in place of consume_skb(skb) in TX completion path 3828 * 3829 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3830 * replacing kfree_skb(skb) 3831 * 3832 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3833 * and consumed a packet. Used in place of consume_skb(skb) 3834 */ 3835 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3836 { 3837 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3838 } 3839 3840 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3841 { 3842 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3843 } 3844 3845 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3846 { 3847 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3848 } 3849 3850 static inline void dev_consume_skb_any(struct sk_buff *skb) 3851 { 3852 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3853 } 3854 3855 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3856 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3857 int netif_rx(struct sk_buff *skb); 3858 int netif_rx_ni(struct sk_buff *skb); 3859 int netif_rx_any_context(struct sk_buff *skb); 3860 int netif_receive_skb(struct sk_buff *skb); 3861 int netif_receive_skb_core(struct sk_buff *skb); 3862 void netif_receive_skb_list(struct list_head *head); 3863 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3864 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3865 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3866 gro_result_t napi_gro_frags(struct napi_struct *napi); 3867 struct packet_offload *gro_find_receive_by_type(__be16 type); 3868 struct packet_offload *gro_find_complete_by_type(__be16 type); 3869 3870 static inline void napi_free_frags(struct napi_struct *napi) 3871 { 3872 kfree_skb(napi->skb); 3873 napi->skb = NULL; 3874 } 3875 3876 bool netdev_is_rx_handler_busy(struct net_device *dev); 3877 int netdev_rx_handler_register(struct net_device *dev, 3878 rx_handler_func_t *rx_handler, 3879 void *rx_handler_data); 3880 void netdev_rx_handler_unregister(struct net_device *dev); 3881 3882 bool dev_valid_name(const char *name); 3883 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3884 bool *need_copyout); 3885 int dev_ifconf(struct net *net, struct ifconf *, int); 3886 int dev_ethtool(struct net *net, struct ifreq *); 3887 unsigned int dev_get_flags(const struct net_device *); 3888 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3889 struct netlink_ext_ack *extack); 3890 int dev_change_flags(struct net_device *dev, unsigned int flags, 3891 struct netlink_ext_ack *extack); 3892 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3893 unsigned int gchanges); 3894 int dev_change_name(struct net_device *, const char *); 3895 int dev_set_alias(struct net_device *, const char *, size_t); 3896 int dev_get_alias(const struct net_device *, char *, size_t); 3897 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3898 int __dev_set_mtu(struct net_device *, int); 3899 int dev_validate_mtu(struct net_device *dev, int mtu, 3900 struct netlink_ext_ack *extack); 3901 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3902 struct netlink_ext_ack *extack); 3903 int dev_set_mtu(struct net_device *, int); 3904 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3905 void dev_set_group(struct net_device *, int); 3906 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3907 struct netlink_ext_ack *extack); 3908 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3909 struct netlink_ext_ack *extack); 3910 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 3911 struct netlink_ext_ack *extack); 3912 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 3913 int dev_change_carrier(struct net_device *, bool new_carrier); 3914 int dev_get_phys_port_id(struct net_device *dev, 3915 struct netdev_phys_item_id *ppid); 3916 int dev_get_phys_port_name(struct net_device *dev, 3917 char *name, size_t len); 3918 int dev_get_port_parent_id(struct net_device *dev, 3919 struct netdev_phys_item_id *ppid, bool recurse); 3920 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3921 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3922 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down); 3923 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask, 3924 u32 value); 3925 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3926 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3927 struct netdev_queue *txq, int *ret); 3928 3929 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3930 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3931 int fd, int expected_fd, u32 flags); 3932 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3933 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3934 3935 int xdp_umem_query(struct net_device *dev, u16 queue_id); 3936 3937 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3938 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3939 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 3940 bool is_skb_forwardable(const struct net_device *dev, 3941 const struct sk_buff *skb); 3942 3943 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 3944 const struct sk_buff *skb, 3945 const bool check_mtu) 3946 { 3947 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 3948 unsigned int len; 3949 3950 if (!(dev->flags & IFF_UP)) 3951 return false; 3952 3953 if (!check_mtu) 3954 return true; 3955 3956 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 3957 if (skb->len <= len) 3958 return true; 3959 3960 /* if TSO is enabled, we don't care about the length as the packet 3961 * could be forwarded without being segmented before 3962 */ 3963 if (skb_is_gso(skb)) 3964 return true; 3965 3966 return false; 3967 } 3968 3969 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3970 struct sk_buff *skb, 3971 const bool check_mtu) 3972 { 3973 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3974 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 3975 atomic_long_inc(&dev->rx_dropped); 3976 kfree_skb(skb); 3977 return NET_RX_DROP; 3978 } 3979 3980 skb_scrub_packet(skb, true); 3981 skb->priority = 0; 3982 return 0; 3983 } 3984 3985 bool dev_nit_active(struct net_device *dev); 3986 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3987 3988 extern int netdev_budget; 3989 extern unsigned int netdev_budget_usecs; 3990 3991 /* Called by rtnetlink.c:rtnl_unlock() */ 3992 void netdev_run_todo(void); 3993 3994 /** 3995 * dev_put - release reference to device 3996 * @dev: network device 3997 * 3998 * Release reference to device to allow it to be freed. 3999 */ 4000 static inline void dev_put(struct net_device *dev) 4001 { 4002 this_cpu_dec(*dev->pcpu_refcnt); 4003 } 4004 4005 /** 4006 * dev_hold - get reference to device 4007 * @dev: network device 4008 * 4009 * Hold reference to device to keep it from being freed. 4010 */ 4011 static inline void dev_hold(struct net_device *dev) 4012 { 4013 this_cpu_inc(*dev->pcpu_refcnt); 4014 } 4015 4016 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4017 * and _off may be called from IRQ context, but it is caller 4018 * who is responsible for serialization of these calls. 4019 * 4020 * The name carrier is inappropriate, these functions should really be 4021 * called netif_lowerlayer_*() because they represent the state of any 4022 * kind of lower layer not just hardware media. 4023 */ 4024 4025 void linkwatch_init_dev(struct net_device *dev); 4026 void linkwatch_fire_event(struct net_device *dev); 4027 void linkwatch_forget_dev(struct net_device *dev); 4028 4029 /** 4030 * netif_carrier_ok - test if carrier present 4031 * @dev: network device 4032 * 4033 * Check if carrier is present on device 4034 */ 4035 static inline bool netif_carrier_ok(const struct net_device *dev) 4036 { 4037 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4038 } 4039 4040 unsigned long dev_trans_start(struct net_device *dev); 4041 4042 void __netdev_watchdog_up(struct net_device *dev); 4043 4044 void netif_carrier_on(struct net_device *dev); 4045 4046 void netif_carrier_off(struct net_device *dev); 4047 4048 /** 4049 * netif_dormant_on - mark device as dormant. 4050 * @dev: network device 4051 * 4052 * Mark device as dormant (as per RFC2863). 4053 * 4054 * The dormant state indicates that the relevant interface is not 4055 * actually in a condition to pass packets (i.e., it is not 'up') but is 4056 * in a "pending" state, waiting for some external event. For "on- 4057 * demand" interfaces, this new state identifies the situation where the 4058 * interface is waiting for events to place it in the up state. 4059 */ 4060 static inline void netif_dormant_on(struct net_device *dev) 4061 { 4062 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4063 linkwatch_fire_event(dev); 4064 } 4065 4066 /** 4067 * netif_dormant_off - set device as not dormant. 4068 * @dev: network device 4069 * 4070 * Device is not in dormant state. 4071 */ 4072 static inline void netif_dormant_off(struct net_device *dev) 4073 { 4074 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4075 linkwatch_fire_event(dev); 4076 } 4077 4078 /** 4079 * netif_dormant - test if device is dormant 4080 * @dev: network device 4081 * 4082 * Check if device is dormant. 4083 */ 4084 static inline bool netif_dormant(const struct net_device *dev) 4085 { 4086 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4087 } 4088 4089 4090 /** 4091 * netif_testing_on - mark device as under test. 4092 * @dev: network device 4093 * 4094 * Mark device as under test (as per RFC2863). 4095 * 4096 * The testing state indicates that some test(s) must be performed on 4097 * the interface. After completion, of the test, the interface state 4098 * will change to up, dormant, or down, as appropriate. 4099 */ 4100 static inline void netif_testing_on(struct net_device *dev) 4101 { 4102 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4103 linkwatch_fire_event(dev); 4104 } 4105 4106 /** 4107 * netif_testing_off - set device as not under test. 4108 * @dev: network device 4109 * 4110 * Device is not in testing state. 4111 */ 4112 static inline void netif_testing_off(struct net_device *dev) 4113 { 4114 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4115 linkwatch_fire_event(dev); 4116 } 4117 4118 /** 4119 * netif_testing - test if device is under test 4120 * @dev: network device 4121 * 4122 * Check if device is under test 4123 */ 4124 static inline bool netif_testing(const struct net_device *dev) 4125 { 4126 return test_bit(__LINK_STATE_TESTING, &dev->state); 4127 } 4128 4129 4130 /** 4131 * netif_oper_up - test if device is operational 4132 * @dev: network device 4133 * 4134 * Check if carrier is operational 4135 */ 4136 static inline bool netif_oper_up(const struct net_device *dev) 4137 { 4138 return (dev->operstate == IF_OPER_UP || 4139 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4140 } 4141 4142 /** 4143 * netif_device_present - is device available or removed 4144 * @dev: network device 4145 * 4146 * Check if device has not been removed from system. 4147 */ 4148 static inline bool netif_device_present(struct net_device *dev) 4149 { 4150 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4151 } 4152 4153 void netif_device_detach(struct net_device *dev); 4154 4155 void netif_device_attach(struct net_device *dev); 4156 4157 /* 4158 * Network interface message level settings 4159 */ 4160 4161 enum { 4162 NETIF_MSG_DRV_BIT, 4163 NETIF_MSG_PROBE_BIT, 4164 NETIF_MSG_LINK_BIT, 4165 NETIF_MSG_TIMER_BIT, 4166 NETIF_MSG_IFDOWN_BIT, 4167 NETIF_MSG_IFUP_BIT, 4168 NETIF_MSG_RX_ERR_BIT, 4169 NETIF_MSG_TX_ERR_BIT, 4170 NETIF_MSG_TX_QUEUED_BIT, 4171 NETIF_MSG_INTR_BIT, 4172 NETIF_MSG_TX_DONE_BIT, 4173 NETIF_MSG_RX_STATUS_BIT, 4174 NETIF_MSG_PKTDATA_BIT, 4175 NETIF_MSG_HW_BIT, 4176 NETIF_MSG_WOL_BIT, 4177 4178 /* When you add a new bit above, update netif_msg_class_names array 4179 * in net/ethtool/common.c 4180 */ 4181 NETIF_MSG_CLASS_COUNT, 4182 }; 4183 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4184 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4185 4186 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4187 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4188 4189 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4190 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4191 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4192 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4193 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4194 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4195 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4196 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4197 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4198 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4199 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4200 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4201 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4202 #define NETIF_MSG_HW __NETIF_MSG(HW) 4203 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4204 4205 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4206 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4207 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4208 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4209 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4210 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4211 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4212 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4213 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4214 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4215 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4216 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4217 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4218 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4219 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4220 4221 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4222 { 4223 /* use default */ 4224 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4225 return default_msg_enable_bits; 4226 if (debug_value == 0) /* no output */ 4227 return 0; 4228 /* set low N bits */ 4229 return (1U << debug_value) - 1; 4230 } 4231 4232 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4233 { 4234 spin_lock(&txq->_xmit_lock); 4235 txq->xmit_lock_owner = cpu; 4236 } 4237 4238 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4239 { 4240 __acquire(&txq->_xmit_lock); 4241 return true; 4242 } 4243 4244 static inline void __netif_tx_release(struct netdev_queue *txq) 4245 { 4246 __release(&txq->_xmit_lock); 4247 } 4248 4249 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4250 { 4251 spin_lock_bh(&txq->_xmit_lock); 4252 txq->xmit_lock_owner = smp_processor_id(); 4253 } 4254 4255 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4256 { 4257 bool ok = spin_trylock(&txq->_xmit_lock); 4258 if (likely(ok)) 4259 txq->xmit_lock_owner = smp_processor_id(); 4260 return ok; 4261 } 4262 4263 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4264 { 4265 txq->xmit_lock_owner = -1; 4266 spin_unlock(&txq->_xmit_lock); 4267 } 4268 4269 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4270 { 4271 txq->xmit_lock_owner = -1; 4272 spin_unlock_bh(&txq->_xmit_lock); 4273 } 4274 4275 static inline void txq_trans_update(struct netdev_queue *txq) 4276 { 4277 if (txq->xmit_lock_owner != -1) 4278 txq->trans_start = jiffies; 4279 } 4280 4281 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4282 static inline void netif_trans_update(struct net_device *dev) 4283 { 4284 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4285 4286 if (txq->trans_start != jiffies) 4287 txq->trans_start = jiffies; 4288 } 4289 4290 /** 4291 * netif_tx_lock - grab network device transmit lock 4292 * @dev: network device 4293 * 4294 * Get network device transmit lock 4295 */ 4296 static inline void netif_tx_lock(struct net_device *dev) 4297 { 4298 unsigned int i; 4299 int cpu; 4300 4301 spin_lock(&dev->tx_global_lock); 4302 cpu = smp_processor_id(); 4303 for (i = 0; i < dev->num_tx_queues; i++) { 4304 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4305 4306 /* We are the only thread of execution doing a 4307 * freeze, but we have to grab the _xmit_lock in 4308 * order to synchronize with threads which are in 4309 * the ->hard_start_xmit() handler and already 4310 * checked the frozen bit. 4311 */ 4312 __netif_tx_lock(txq, cpu); 4313 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 4314 __netif_tx_unlock(txq); 4315 } 4316 } 4317 4318 static inline void netif_tx_lock_bh(struct net_device *dev) 4319 { 4320 local_bh_disable(); 4321 netif_tx_lock(dev); 4322 } 4323 4324 static inline void netif_tx_unlock(struct net_device *dev) 4325 { 4326 unsigned int i; 4327 4328 for (i = 0; i < dev->num_tx_queues; i++) { 4329 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4330 4331 /* No need to grab the _xmit_lock here. If the 4332 * queue is not stopped for another reason, we 4333 * force a schedule. 4334 */ 4335 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 4336 netif_schedule_queue(txq); 4337 } 4338 spin_unlock(&dev->tx_global_lock); 4339 } 4340 4341 static inline void netif_tx_unlock_bh(struct net_device *dev) 4342 { 4343 netif_tx_unlock(dev); 4344 local_bh_enable(); 4345 } 4346 4347 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4348 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4349 __netif_tx_lock(txq, cpu); \ 4350 } else { \ 4351 __netif_tx_acquire(txq); \ 4352 } \ 4353 } 4354 4355 #define HARD_TX_TRYLOCK(dev, txq) \ 4356 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4357 __netif_tx_trylock(txq) : \ 4358 __netif_tx_acquire(txq)) 4359 4360 #define HARD_TX_UNLOCK(dev, txq) { \ 4361 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4362 __netif_tx_unlock(txq); \ 4363 } else { \ 4364 __netif_tx_release(txq); \ 4365 } \ 4366 } 4367 4368 static inline void netif_tx_disable(struct net_device *dev) 4369 { 4370 unsigned int i; 4371 int cpu; 4372 4373 local_bh_disable(); 4374 cpu = smp_processor_id(); 4375 spin_lock(&dev->tx_global_lock); 4376 for (i = 0; i < dev->num_tx_queues; i++) { 4377 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4378 4379 __netif_tx_lock(txq, cpu); 4380 netif_tx_stop_queue(txq); 4381 __netif_tx_unlock(txq); 4382 } 4383 spin_unlock(&dev->tx_global_lock); 4384 local_bh_enable(); 4385 } 4386 4387 static inline void netif_addr_lock(struct net_device *dev) 4388 { 4389 unsigned char nest_level = 0; 4390 4391 #ifdef CONFIG_LOCKDEP 4392 nest_level = dev->nested_level; 4393 #endif 4394 spin_lock_nested(&dev->addr_list_lock, nest_level); 4395 } 4396 4397 static inline void netif_addr_lock_bh(struct net_device *dev) 4398 { 4399 unsigned char nest_level = 0; 4400 4401 #ifdef CONFIG_LOCKDEP 4402 nest_level = dev->nested_level; 4403 #endif 4404 local_bh_disable(); 4405 spin_lock_nested(&dev->addr_list_lock, nest_level); 4406 } 4407 4408 static inline void netif_addr_unlock(struct net_device *dev) 4409 { 4410 spin_unlock(&dev->addr_list_lock); 4411 } 4412 4413 static inline void netif_addr_unlock_bh(struct net_device *dev) 4414 { 4415 spin_unlock_bh(&dev->addr_list_lock); 4416 } 4417 4418 /* 4419 * dev_addrs walker. Should be used only for read access. Call with 4420 * rcu_read_lock held. 4421 */ 4422 #define for_each_dev_addr(dev, ha) \ 4423 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4424 4425 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4426 4427 void ether_setup(struct net_device *dev); 4428 4429 /* Support for loadable net-drivers */ 4430 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4431 unsigned char name_assign_type, 4432 void (*setup)(struct net_device *), 4433 unsigned int txqs, unsigned int rxqs); 4434 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4435 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4436 4437 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4438 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4439 count) 4440 4441 int register_netdev(struct net_device *dev); 4442 void unregister_netdev(struct net_device *dev); 4443 4444 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4445 4446 /* General hardware address lists handling functions */ 4447 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4448 struct netdev_hw_addr_list *from_list, int addr_len); 4449 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4450 struct netdev_hw_addr_list *from_list, int addr_len); 4451 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4452 struct net_device *dev, 4453 int (*sync)(struct net_device *, const unsigned char *), 4454 int (*unsync)(struct net_device *, 4455 const unsigned char *)); 4456 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4457 struct net_device *dev, 4458 int (*sync)(struct net_device *, 4459 const unsigned char *, int), 4460 int (*unsync)(struct net_device *, 4461 const unsigned char *, int)); 4462 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4463 struct net_device *dev, 4464 int (*unsync)(struct net_device *, 4465 const unsigned char *, int)); 4466 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4467 struct net_device *dev, 4468 int (*unsync)(struct net_device *, 4469 const unsigned char *)); 4470 void __hw_addr_init(struct netdev_hw_addr_list *list); 4471 4472 /* Functions used for device addresses handling */ 4473 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4474 unsigned char addr_type); 4475 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4476 unsigned char addr_type); 4477 void dev_addr_flush(struct net_device *dev); 4478 int dev_addr_init(struct net_device *dev); 4479 4480 /* Functions used for unicast addresses handling */ 4481 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4482 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4483 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4484 int dev_uc_sync(struct net_device *to, struct net_device *from); 4485 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4486 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4487 void dev_uc_flush(struct net_device *dev); 4488 void dev_uc_init(struct net_device *dev); 4489 4490 /** 4491 * __dev_uc_sync - Synchonize device's unicast 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_uc_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->uc, dev, sync, unsync); 4506 } 4507 4508 /** 4509 * __dev_uc_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_uc_sync(). 4514 */ 4515 static inline void __dev_uc_unsync(struct net_device *dev, 4516 int (*unsync)(struct net_device *, 4517 const unsigned char *)) 4518 { 4519 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4520 } 4521 4522 /* Functions used for multicast addresses handling */ 4523 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4524 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4525 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4526 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4527 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4528 int dev_mc_sync(struct net_device *to, struct net_device *from); 4529 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4530 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4531 void dev_mc_flush(struct net_device *dev); 4532 void dev_mc_init(struct net_device *dev); 4533 4534 /** 4535 * __dev_mc_sync - Synchonize device's multicast list 4536 * @dev: device to sync 4537 * @sync: function to call if address should be added 4538 * @unsync: function to call if address should be removed 4539 * 4540 * Add newly added addresses to the interface, and release 4541 * addresses that have been deleted. 4542 */ 4543 static inline int __dev_mc_sync(struct net_device *dev, 4544 int (*sync)(struct net_device *, 4545 const unsigned char *), 4546 int (*unsync)(struct net_device *, 4547 const unsigned char *)) 4548 { 4549 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4550 } 4551 4552 /** 4553 * __dev_mc_unsync - Remove synchronized addresses from device 4554 * @dev: device to sync 4555 * @unsync: function to call if address should be removed 4556 * 4557 * Remove all addresses that were added to the device by dev_mc_sync(). 4558 */ 4559 static inline void __dev_mc_unsync(struct net_device *dev, 4560 int (*unsync)(struct net_device *, 4561 const unsigned char *)) 4562 { 4563 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4564 } 4565 4566 /* Functions used for secondary unicast and multicast support */ 4567 void dev_set_rx_mode(struct net_device *dev); 4568 void __dev_set_rx_mode(struct net_device *dev); 4569 int dev_set_promiscuity(struct net_device *dev, int inc); 4570 int dev_set_allmulti(struct net_device *dev, int inc); 4571 void netdev_state_change(struct net_device *dev); 4572 void __netdev_notify_peers(struct net_device *dev); 4573 void netdev_notify_peers(struct net_device *dev); 4574 void netdev_features_change(struct net_device *dev); 4575 /* Load a device via the kmod */ 4576 void dev_load(struct net *net, const char *name); 4577 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4578 struct rtnl_link_stats64 *storage); 4579 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4580 const struct net_device_stats *netdev_stats); 4581 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4582 const struct pcpu_sw_netstats __percpu *netstats); 4583 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4584 4585 extern int netdev_max_backlog; 4586 extern int netdev_tstamp_prequeue; 4587 extern int weight_p; 4588 extern int dev_weight_rx_bias; 4589 extern int dev_weight_tx_bias; 4590 extern int dev_rx_weight; 4591 extern int dev_tx_weight; 4592 extern int gro_normal_batch; 4593 4594 enum { 4595 NESTED_SYNC_IMM_BIT, 4596 NESTED_SYNC_TODO_BIT, 4597 }; 4598 4599 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4600 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4601 4602 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4603 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4604 4605 struct netdev_nested_priv { 4606 unsigned char flags; 4607 void *data; 4608 }; 4609 4610 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4611 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4612 struct list_head **iter); 4613 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 4614 struct list_head **iter); 4615 4616 #ifdef CONFIG_LOCKDEP 4617 static LIST_HEAD(net_unlink_list); 4618 4619 static inline void net_unlink_todo(struct net_device *dev) 4620 { 4621 if (list_empty(&dev->unlink_list)) 4622 list_add_tail(&dev->unlink_list, &net_unlink_list); 4623 } 4624 #endif 4625 4626 /* iterate through upper list, must be called under RCU read lock */ 4627 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4628 for (iter = &(dev)->adj_list.upper, \ 4629 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4630 updev; \ 4631 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4632 4633 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4634 int (*fn)(struct net_device *upper_dev, 4635 struct netdev_nested_priv *priv), 4636 struct netdev_nested_priv *priv); 4637 4638 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4639 struct net_device *upper_dev); 4640 4641 bool netdev_has_any_upper_dev(struct net_device *dev); 4642 4643 void *netdev_lower_get_next_private(struct net_device *dev, 4644 struct list_head **iter); 4645 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4646 struct list_head **iter); 4647 4648 #define netdev_for_each_lower_private(dev, priv, iter) \ 4649 for (iter = (dev)->adj_list.lower.next, \ 4650 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4651 priv; \ 4652 priv = netdev_lower_get_next_private(dev, &(iter))) 4653 4654 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4655 for (iter = &(dev)->adj_list.lower, \ 4656 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4657 priv; \ 4658 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4659 4660 void *netdev_lower_get_next(struct net_device *dev, 4661 struct list_head **iter); 4662 4663 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4664 for (iter = (dev)->adj_list.lower.next, \ 4665 ldev = netdev_lower_get_next(dev, &(iter)); \ 4666 ldev; \ 4667 ldev = netdev_lower_get_next(dev, &(iter))) 4668 4669 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4670 struct list_head **iter); 4671 int netdev_walk_all_lower_dev(struct net_device *dev, 4672 int (*fn)(struct net_device *lower_dev, 4673 struct netdev_nested_priv *priv), 4674 struct netdev_nested_priv *priv); 4675 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4676 int (*fn)(struct net_device *lower_dev, 4677 struct netdev_nested_priv *priv), 4678 struct netdev_nested_priv *priv); 4679 4680 void *netdev_adjacent_get_private(struct list_head *adj_list); 4681 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4682 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4683 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4684 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4685 struct netlink_ext_ack *extack); 4686 int netdev_master_upper_dev_link(struct net_device *dev, 4687 struct net_device *upper_dev, 4688 void *upper_priv, void *upper_info, 4689 struct netlink_ext_ack *extack); 4690 void netdev_upper_dev_unlink(struct net_device *dev, 4691 struct net_device *upper_dev); 4692 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4693 struct net_device *new_dev, 4694 struct net_device *dev, 4695 struct netlink_ext_ack *extack); 4696 void netdev_adjacent_change_commit(struct net_device *old_dev, 4697 struct net_device *new_dev, 4698 struct net_device *dev); 4699 void netdev_adjacent_change_abort(struct net_device *old_dev, 4700 struct net_device *new_dev, 4701 struct net_device *dev); 4702 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4703 void *netdev_lower_dev_get_private(struct net_device *dev, 4704 struct net_device *lower_dev); 4705 void netdev_lower_state_changed(struct net_device *lower_dev, 4706 void *lower_state_info); 4707 4708 /* RSS keys are 40 or 52 bytes long */ 4709 #define NETDEV_RSS_KEY_LEN 52 4710 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4711 void netdev_rss_key_fill(void *buffer, size_t len); 4712 4713 int skb_checksum_help(struct sk_buff *skb); 4714 int skb_crc32c_csum_help(struct sk_buff *skb); 4715 int skb_csum_hwoffload_help(struct sk_buff *skb, 4716 const netdev_features_t features); 4717 4718 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4719 netdev_features_t features, bool tx_path); 4720 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4721 netdev_features_t features); 4722 4723 struct netdev_bonding_info { 4724 ifslave slave; 4725 ifbond master; 4726 }; 4727 4728 struct netdev_notifier_bonding_info { 4729 struct netdev_notifier_info info; /* must be first */ 4730 struct netdev_bonding_info bonding_info; 4731 }; 4732 4733 void netdev_bonding_info_change(struct net_device *dev, 4734 struct netdev_bonding_info *bonding_info); 4735 4736 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4737 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4738 #else 4739 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4740 const void *data) 4741 { 4742 } 4743 #endif 4744 4745 static inline 4746 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4747 { 4748 return __skb_gso_segment(skb, features, true); 4749 } 4750 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4751 4752 static inline bool can_checksum_protocol(netdev_features_t features, 4753 __be16 protocol) 4754 { 4755 if (protocol == htons(ETH_P_FCOE)) 4756 return !!(features & NETIF_F_FCOE_CRC); 4757 4758 /* Assume this is an IP checksum (not SCTP CRC) */ 4759 4760 if (features & NETIF_F_HW_CSUM) { 4761 /* Can checksum everything */ 4762 return true; 4763 } 4764 4765 switch (protocol) { 4766 case htons(ETH_P_IP): 4767 return !!(features & NETIF_F_IP_CSUM); 4768 case htons(ETH_P_IPV6): 4769 return !!(features & NETIF_F_IPV6_CSUM); 4770 default: 4771 return false; 4772 } 4773 } 4774 4775 #ifdef CONFIG_BUG 4776 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4777 #else 4778 static inline void netdev_rx_csum_fault(struct net_device *dev, 4779 struct sk_buff *skb) 4780 { 4781 } 4782 #endif 4783 /* rx skb timestamps */ 4784 void net_enable_timestamp(void); 4785 void net_disable_timestamp(void); 4786 4787 #ifdef CONFIG_PROC_FS 4788 int __init dev_proc_init(void); 4789 #else 4790 #define dev_proc_init() 0 4791 #endif 4792 4793 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4794 struct sk_buff *skb, struct net_device *dev, 4795 bool more) 4796 { 4797 __this_cpu_write(softnet_data.xmit.more, more); 4798 return ops->ndo_start_xmit(skb, dev); 4799 } 4800 4801 static inline bool netdev_xmit_more(void) 4802 { 4803 return __this_cpu_read(softnet_data.xmit.more); 4804 } 4805 4806 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4807 struct netdev_queue *txq, bool more) 4808 { 4809 const struct net_device_ops *ops = dev->netdev_ops; 4810 netdev_tx_t rc; 4811 4812 rc = __netdev_start_xmit(ops, skb, dev, more); 4813 if (rc == NETDEV_TX_OK) 4814 txq_trans_update(txq); 4815 4816 return rc; 4817 } 4818 4819 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4820 const void *ns); 4821 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4822 const void *ns); 4823 4824 extern const struct kobj_ns_type_operations net_ns_type_operations; 4825 4826 const char *netdev_drivername(const struct net_device *dev); 4827 4828 void linkwatch_run_queue(void); 4829 4830 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4831 netdev_features_t f2) 4832 { 4833 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4834 if (f1 & NETIF_F_HW_CSUM) 4835 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4836 else 4837 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4838 } 4839 4840 return f1 & f2; 4841 } 4842 4843 static inline netdev_features_t netdev_get_wanted_features( 4844 struct net_device *dev) 4845 { 4846 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4847 } 4848 netdev_features_t netdev_increment_features(netdev_features_t all, 4849 netdev_features_t one, netdev_features_t mask); 4850 4851 /* Allow TSO being used on stacked device : 4852 * Performing the GSO segmentation before last device 4853 * is a performance improvement. 4854 */ 4855 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4856 netdev_features_t mask) 4857 { 4858 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4859 } 4860 4861 int __netdev_update_features(struct net_device *dev); 4862 void netdev_update_features(struct net_device *dev); 4863 void netdev_change_features(struct net_device *dev); 4864 4865 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4866 struct net_device *dev); 4867 4868 netdev_features_t passthru_features_check(struct sk_buff *skb, 4869 struct net_device *dev, 4870 netdev_features_t features); 4871 netdev_features_t netif_skb_features(struct sk_buff *skb); 4872 4873 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4874 { 4875 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4876 4877 /* check flags correspondence */ 4878 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4879 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4880 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4881 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4882 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4883 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4884 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4885 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4886 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4887 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4888 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4889 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4890 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4891 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4892 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4893 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4894 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4895 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4896 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4897 4898 return (features & feature) == feature; 4899 } 4900 4901 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4902 { 4903 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4904 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4905 } 4906 4907 static inline bool netif_needs_gso(struct sk_buff *skb, 4908 netdev_features_t features) 4909 { 4910 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4911 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4912 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4913 } 4914 4915 static inline void netif_set_gso_max_size(struct net_device *dev, 4916 unsigned int size) 4917 { 4918 dev->gso_max_size = size; 4919 } 4920 4921 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4922 int pulled_hlen, u16 mac_offset, 4923 int mac_len) 4924 { 4925 skb->protocol = protocol; 4926 skb->encapsulation = 1; 4927 skb_push(skb, pulled_hlen); 4928 skb_reset_transport_header(skb); 4929 skb->mac_header = mac_offset; 4930 skb->network_header = skb->mac_header + mac_len; 4931 skb->mac_len = mac_len; 4932 } 4933 4934 static inline bool netif_is_macsec(const struct net_device *dev) 4935 { 4936 return dev->priv_flags & IFF_MACSEC; 4937 } 4938 4939 static inline bool netif_is_macvlan(const struct net_device *dev) 4940 { 4941 return dev->priv_flags & IFF_MACVLAN; 4942 } 4943 4944 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4945 { 4946 return dev->priv_flags & IFF_MACVLAN_PORT; 4947 } 4948 4949 static inline bool netif_is_bond_master(const struct net_device *dev) 4950 { 4951 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4952 } 4953 4954 static inline bool netif_is_bond_slave(const struct net_device *dev) 4955 { 4956 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4957 } 4958 4959 static inline bool netif_supports_nofcs(struct net_device *dev) 4960 { 4961 return dev->priv_flags & IFF_SUPP_NOFCS; 4962 } 4963 4964 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4965 { 4966 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4967 } 4968 4969 static inline bool netif_is_l3_master(const struct net_device *dev) 4970 { 4971 return dev->priv_flags & IFF_L3MDEV_MASTER; 4972 } 4973 4974 static inline bool netif_is_l3_slave(const struct net_device *dev) 4975 { 4976 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4977 } 4978 4979 static inline bool netif_is_bridge_master(const struct net_device *dev) 4980 { 4981 return dev->priv_flags & IFF_EBRIDGE; 4982 } 4983 4984 static inline bool netif_is_bridge_port(const struct net_device *dev) 4985 { 4986 return dev->priv_flags & IFF_BRIDGE_PORT; 4987 } 4988 4989 static inline bool netif_is_ovs_master(const struct net_device *dev) 4990 { 4991 return dev->priv_flags & IFF_OPENVSWITCH; 4992 } 4993 4994 static inline bool netif_is_ovs_port(const struct net_device *dev) 4995 { 4996 return dev->priv_flags & IFF_OVS_DATAPATH; 4997 } 4998 4999 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5000 { 5001 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5002 } 5003 5004 static inline bool netif_is_team_master(const struct net_device *dev) 5005 { 5006 return dev->priv_flags & IFF_TEAM; 5007 } 5008 5009 static inline bool netif_is_team_port(const struct net_device *dev) 5010 { 5011 return dev->priv_flags & IFF_TEAM_PORT; 5012 } 5013 5014 static inline bool netif_is_lag_master(const struct net_device *dev) 5015 { 5016 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5017 } 5018 5019 static inline bool netif_is_lag_port(const struct net_device *dev) 5020 { 5021 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5022 } 5023 5024 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5025 { 5026 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5027 } 5028 5029 static inline bool netif_is_failover(const struct net_device *dev) 5030 { 5031 return dev->priv_flags & IFF_FAILOVER; 5032 } 5033 5034 static inline bool netif_is_failover_slave(const struct net_device *dev) 5035 { 5036 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5037 } 5038 5039 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5040 static inline void netif_keep_dst(struct net_device *dev) 5041 { 5042 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5043 } 5044 5045 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5046 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5047 { 5048 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5049 return dev->priv_flags & IFF_MACSEC; 5050 } 5051 5052 extern struct pernet_operations __net_initdata loopback_net_ops; 5053 5054 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5055 5056 /* netdev_printk helpers, similar to dev_printk */ 5057 5058 static inline const char *netdev_name(const struct net_device *dev) 5059 { 5060 if (!dev->name[0] || strchr(dev->name, '%')) 5061 return "(unnamed net_device)"; 5062 return dev->name; 5063 } 5064 5065 static inline bool netdev_unregistering(const struct net_device *dev) 5066 { 5067 return dev->reg_state == NETREG_UNREGISTERING; 5068 } 5069 5070 static inline const char *netdev_reg_state(const struct net_device *dev) 5071 { 5072 switch (dev->reg_state) { 5073 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5074 case NETREG_REGISTERED: return ""; 5075 case NETREG_UNREGISTERING: return " (unregistering)"; 5076 case NETREG_UNREGISTERED: return " (unregistered)"; 5077 case NETREG_RELEASED: return " (released)"; 5078 case NETREG_DUMMY: return " (dummy)"; 5079 } 5080 5081 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5082 return " (unknown)"; 5083 } 5084 5085 __printf(3, 4) __cold 5086 void netdev_printk(const char *level, const struct net_device *dev, 5087 const char *format, ...); 5088 __printf(2, 3) __cold 5089 void netdev_emerg(const struct net_device *dev, const char *format, ...); 5090 __printf(2, 3) __cold 5091 void netdev_alert(const struct net_device *dev, const char *format, ...); 5092 __printf(2, 3) __cold 5093 void netdev_crit(const struct net_device *dev, const char *format, ...); 5094 __printf(2, 3) __cold 5095 void netdev_err(const struct net_device *dev, const char *format, ...); 5096 __printf(2, 3) __cold 5097 void netdev_warn(const struct net_device *dev, const char *format, ...); 5098 __printf(2, 3) __cold 5099 void netdev_notice(const struct net_device *dev, const char *format, ...); 5100 __printf(2, 3) __cold 5101 void netdev_info(const struct net_device *dev, const char *format, ...); 5102 5103 #define netdev_level_once(level, dev, fmt, ...) \ 5104 do { \ 5105 static bool __print_once __read_mostly; \ 5106 \ 5107 if (!__print_once) { \ 5108 __print_once = true; \ 5109 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 5110 } \ 5111 } while (0) 5112 5113 #define netdev_emerg_once(dev, fmt, ...) \ 5114 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 5115 #define netdev_alert_once(dev, fmt, ...) \ 5116 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 5117 #define netdev_crit_once(dev, fmt, ...) \ 5118 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 5119 #define netdev_err_once(dev, fmt, ...) \ 5120 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 5121 #define netdev_warn_once(dev, fmt, ...) \ 5122 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 5123 #define netdev_notice_once(dev, fmt, ...) \ 5124 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 5125 #define netdev_info_once(dev, fmt, ...) \ 5126 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 5127 5128 #define MODULE_ALIAS_NETDEV(device) \ 5129 MODULE_ALIAS("netdev-" device) 5130 5131 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5132 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5133 #define netdev_dbg(__dev, format, args...) \ 5134 do { \ 5135 dynamic_netdev_dbg(__dev, format, ##args); \ 5136 } while (0) 5137 #elif defined(DEBUG) 5138 #define netdev_dbg(__dev, format, args...) \ 5139 netdev_printk(KERN_DEBUG, __dev, format, ##args) 5140 #else 5141 #define netdev_dbg(__dev, format, args...) \ 5142 ({ \ 5143 if (0) \ 5144 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 5145 }) 5146 #endif 5147 5148 #if defined(VERBOSE_DEBUG) 5149 #define netdev_vdbg netdev_dbg 5150 #else 5151 5152 #define netdev_vdbg(dev, format, args...) \ 5153 ({ \ 5154 if (0) \ 5155 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 5156 0; \ 5157 }) 5158 #endif 5159 5160 /* 5161 * netdev_WARN() acts like dev_printk(), but with the key difference 5162 * of using a WARN/WARN_ON to get the message out, including the 5163 * file/line information and a backtrace. 5164 */ 5165 #define netdev_WARN(dev, format, args...) \ 5166 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5167 netdev_reg_state(dev), ##args) 5168 5169 #define netdev_WARN_ONCE(dev, format, args...) \ 5170 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5171 netdev_reg_state(dev), ##args) 5172 5173 /* netif printk helpers, similar to netdev_printk */ 5174 5175 #define netif_printk(priv, type, level, dev, fmt, args...) \ 5176 do { \ 5177 if (netif_msg_##type(priv)) \ 5178 netdev_printk(level, (dev), fmt, ##args); \ 5179 } while (0) 5180 5181 #define netif_level(level, priv, type, dev, fmt, args...) \ 5182 do { \ 5183 if (netif_msg_##type(priv)) \ 5184 netdev_##level(dev, fmt, ##args); \ 5185 } while (0) 5186 5187 #define netif_emerg(priv, type, dev, fmt, args...) \ 5188 netif_level(emerg, priv, type, dev, fmt, ##args) 5189 #define netif_alert(priv, type, dev, fmt, args...) \ 5190 netif_level(alert, priv, type, dev, fmt, ##args) 5191 #define netif_crit(priv, type, dev, fmt, args...) \ 5192 netif_level(crit, priv, type, dev, fmt, ##args) 5193 #define netif_err(priv, type, dev, fmt, args...) \ 5194 netif_level(err, priv, type, dev, fmt, ##args) 5195 #define netif_warn(priv, type, dev, fmt, args...) \ 5196 netif_level(warn, priv, type, dev, fmt, ##args) 5197 #define netif_notice(priv, type, dev, fmt, args...) \ 5198 netif_level(notice, priv, type, dev, fmt, ##args) 5199 #define netif_info(priv, type, dev, fmt, args...) \ 5200 netif_level(info, priv, type, dev, fmt, ##args) 5201 5202 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5203 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5204 #define netif_dbg(priv, type, netdev, format, args...) \ 5205 do { \ 5206 if (netif_msg_##type(priv)) \ 5207 dynamic_netdev_dbg(netdev, format, ##args); \ 5208 } while (0) 5209 #elif defined(DEBUG) 5210 #define netif_dbg(priv, type, dev, format, args...) \ 5211 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 5212 #else 5213 #define netif_dbg(priv, type, dev, format, args...) \ 5214 ({ \ 5215 if (0) \ 5216 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5217 0; \ 5218 }) 5219 #endif 5220 5221 /* if @cond then downgrade to debug, else print at @level */ 5222 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 5223 do { \ 5224 if (cond) \ 5225 netif_dbg(priv, type, netdev, fmt, ##args); \ 5226 else \ 5227 netif_ ## level(priv, type, netdev, fmt, ##args); \ 5228 } while (0) 5229 5230 #if defined(VERBOSE_DEBUG) 5231 #define netif_vdbg netif_dbg 5232 #else 5233 #define netif_vdbg(priv, type, dev, format, args...) \ 5234 ({ \ 5235 if (0) \ 5236 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5237 0; \ 5238 }) 5239 #endif 5240 5241 /* 5242 * The list of packet types we will receive (as opposed to discard) 5243 * and the routines to invoke. 5244 * 5245 * Why 16. Because with 16 the only overlap we get on a hash of the 5246 * low nibble of the protocol value is RARP/SNAP/X.25. 5247 * 5248 * 0800 IP 5249 * 0001 802.3 5250 * 0002 AX.25 5251 * 0004 802.2 5252 * 8035 RARP 5253 * 0005 SNAP 5254 * 0805 X.25 5255 * 0806 ARP 5256 * 8137 IPX 5257 * 0009 Localtalk 5258 * 86DD IPv6 5259 */ 5260 #define PTYPE_HASH_SIZE (16) 5261 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5262 5263 extern struct net_device *blackhole_netdev; 5264 5265 #endif /* _LINUX_NETDEVICE_H */ 5266