1 /* 2 * INET An implementation of the TCP/IP protocol suite for the LINUX 3 * operating system. INET is implemented using the BSD Socket 4 * interface as the means of communication with the user level. 5 * 6 * Definitions for the Interfaces handler. 7 * 8 * Version: @(#)dev.h 1.0.10 08/12/93 9 * 10 * Authors: Ross Biro 11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 12 * Corey Minyard <wf-rch!minyard@relay.EU.net> 13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov> 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk> 15 * Bjorn Ekwall. <bj0rn@blox.se> 16 * Pekka Riikonen <priikone@poseidon.pspt.fi> 17 * 18 * This program is free software; you can redistribute it and/or 19 * modify it under the terms of the GNU General Public License 20 * as published by the Free Software Foundation; either version 21 * 2 of the License, or (at your option) any later version. 22 * 23 * Moved to /usr/include/linux for NET3 24 */ 25 #ifndef _LINUX_NETDEVICE_H 26 #define _LINUX_NETDEVICE_H 27 28 #include <linux/timer.h> 29 #include <linux/bug.h> 30 #include <linux/delay.h> 31 #include <linux/atomic.h> 32 #include <linux/prefetch.h> 33 #include <asm/cache.h> 34 #include <asm/byteorder.h> 35 36 #include <linux/percpu.h> 37 #include <linux/rculist.h> 38 #include <linux/dmaengine.h> 39 #include <linux/workqueue.h> 40 #include <linux/dynamic_queue_limits.h> 41 42 #include <linux/ethtool.h> 43 #include <net/net_namespace.h> 44 #include <net/dsa.h> 45 #ifdef CONFIG_DCB 46 #include <net/dcbnl.h> 47 #endif 48 #include <net/netprio_cgroup.h> 49 50 #include <linux/netdev_features.h> 51 #include <linux/neighbour.h> 52 #include <uapi/linux/netdevice.h> 53 #include <uapi/linux/if_bonding.h> 54 #include <uapi/linux/pkt_cls.h> 55 #include <linux/hashtable.h> 56 57 struct netpoll_info; 58 struct device; 59 struct phy_device; 60 /* 802.11 specific */ 61 struct wireless_dev; 62 /* 802.15.4 specific */ 63 struct wpan_dev; 64 struct mpls_dev; 65 /* UDP Tunnel offloads */ 66 struct udp_tunnel_info; 67 struct bpf_prog; 68 69 void netdev_set_default_ethtool_ops(struct net_device *dev, 70 const struct ethtool_ops *ops); 71 72 /* Backlog congestion levels */ 73 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 74 #define NET_RX_DROP 1 /* packet dropped */ 75 76 /* 77 * Transmit return codes: transmit return codes originate from three different 78 * namespaces: 79 * 80 * - qdisc return codes 81 * - driver transmit return codes 82 * - errno values 83 * 84 * Drivers are allowed to return any one of those in their hard_start_xmit() 85 * function. Real network devices commonly used with qdiscs should only return 86 * the driver transmit return codes though - when qdiscs are used, the actual 87 * transmission happens asynchronously, so the value is not propagated to 88 * higher layers. Virtual network devices transmit synchronously; in this case 89 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 90 * others are propagated to higher layers. 91 */ 92 93 /* qdisc ->enqueue() return codes. */ 94 #define NET_XMIT_SUCCESS 0x00 95 #define NET_XMIT_DROP 0x01 /* skb dropped */ 96 #define NET_XMIT_CN 0x02 /* congestion notification */ 97 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 98 99 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 100 * indicates that the device will soon be dropping packets, or already drops 101 * some packets of the same priority; prompting us to send less aggressively. */ 102 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 103 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 104 105 /* Driver transmit return codes */ 106 #define NETDEV_TX_MASK 0xf0 107 108 enum netdev_tx { 109 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 110 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 111 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 112 }; 113 typedef enum netdev_tx netdev_tx_t; 114 115 /* 116 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 117 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 118 */ 119 static inline bool dev_xmit_complete(int rc) 120 { 121 /* 122 * Positive cases with an skb consumed by a driver: 123 * - successful transmission (rc == NETDEV_TX_OK) 124 * - error while transmitting (rc < 0) 125 * - error while queueing to a different device (rc & NET_XMIT_MASK) 126 */ 127 if (likely(rc < NET_XMIT_MASK)) 128 return true; 129 130 return false; 131 } 132 133 /* 134 * Compute the worst-case header length according to the protocols 135 * used. 136 */ 137 138 #if defined(CONFIG_HYPERV_NET) 139 # define LL_MAX_HEADER 128 140 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 141 # if defined(CONFIG_MAC80211_MESH) 142 # define LL_MAX_HEADER 128 143 # else 144 # define LL_MAX_HEADER 96 145 # endif 146 #else 147 # define LL_MAX_HEADER 32 148 #endif 149 150 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 151 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 152 #define MAX_HEADER LL_MAX_HEADER 153 #else 154 #define MAX_HEADER (LL_MAX_HEADER + 48) 155 #endif 156 157 /* 158 * Old network device statistics. Fields are native words 159 * (unsigned long) so they can be read and written atomically. 160 */ 161 162 struct net_device_stats { 163 unsigned long rx_packets; 164 unsigned long tx_packets; 165 unsigned long rx_bytes; 166 unsigned long tx_bytes; 167 unsigned long rx_errors; 168 unsigned long tx_errors; 169 unsigned long rx_dropped; 170 unsigned long tx_dropped; 171 unsigned long multicast; 172 unsigned long collisions; 173 unsigned long rx_length_errors; 174 unsigned long rx_over_errors; 175 unsigned long rx_crc_errors; 176 unsigned long rx_frame_errors; 177 unsigned long rx_fifo_errors; 178 unsigned long rx_missed_errors; 179 unsigned long tx_aborted_errors; 180 unsigned long tx_carrier_errors; 181 unsigned long tx_fifo_errors; 182 unsigned long tx_heartbeat_errors; 183 unsigned long tx_window_errors; 184 unsigned long rx_compressed; 185 unsigned long tx_compressed; 186 }; 187 188 189 #include <linux/cache.h> 190 #include <linux/skbuff.h> 191 192 #ifdef CONFIG_RPS 193 #include <linux/static_key.h> 194 extern struct static_key rps_needed; 195 #endif 196 197 struct neighbour; 198 struct neigh_parms; 199 struct sk_buff; 200 201 struct netdev_hw_addr { 202 struct list_head list; 203 unsigned char addr[MAX_ADDR_LEN]; 204 unsigned char type; 205 #define NETDEV_HW_ADDR_T_LAN 1 206 #define NETDEV_HW_ADDR_T_SAN 2 207 #define NETDEV_HW_ADDR_T_SLAVE 3 208 #define NETDEV_HW_ADDR_T_UNICAST 4 209 #define NETDEV_HW_ADDR_T_MULTICAST 5 210 bool global_use; 211 int sync_cnt; 212 int refcount; 213 int synced; 214 struct rcu_head rcu_head; 215 }; 216 217 struct netdev_hw_addr_list { 218 struct list_head list; 219 int count; 220 }; 221 222 #define netdev_hw_addr_list_count(l) ((l)->count) 223 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 224 #define netdev_hw_addr_list_for_each(ha, l) \ 225 list_for_each_entry(ha, &(l)->list, list) 226 227 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 228 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 229 #define netdev_for_each_uc_addr(ha, dev) \ 230 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 231 232 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 233 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 234 #define netdev_for_each_mc_addr(ha, dev) \ 235 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 236 237 struct hh_cache { 238 u16 hh_len; 239 u16 __pad; 240 seqlock_t hh_lock; 241 242 /* cached hardware header; allow for machine alignment needs. */ 243 #define HH_DATA_MOD 16 244 #define HH_DATA_OFF(__len) \ 245 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 246 #define HH_DATA_ALIGN(__len) \ 247 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 248 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 249 }; 250 251 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 252 * Alternative is: 253 * dev->hard_header_len ? (dev->hard_header_len + 254 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 255 * 256 * We could use other alignment values, but we must maintain the 257 * relationship HH alignment <= LL alignment. 258 */ 259 #define LL_RESERVED_SPACE(dev) \ 260 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 261 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 262 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 263 264 struct header_ops { 265 int (*create) (struct sk_buff *skb, struct net_device *dev, 266 unsigned short type, const void *daddr, 267 const void *saddr, unsigned int len); 268 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 269 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 270 void (*cache_update)(struct hh_cache *hh, 271 const struct net_device *dev, 272 const unsigned char *haddr); 273 bool (*validate)(const char *ll_header, unsigned int len); 274 }; 275 276 /* These flag bits are private to the generic network queueing 277 * layer; they may not be explicitly referenced by any other 278 * code. 279 */ 280 281 enum netdev_state_t { 282 __LINK_STATE_START, 283 __LINK_STATE_PRESENT, 284 __LINK_STATE_NOCARRIER, 285 __LINK_STATE_LINKWATCH_PENDING, 286 __LINK_STATE_DORMANT, 287 }; 288 289 290 /* 291 * This structure holds boot-time configured netdevice settings. They 292 * are then used in the device probing. 293 */ 294 struct netdev_boot_setup { 295 char name[IFNAMSIZ]; 296 struct ifmap map; 297 }; 298 #define NETDEV_BOOT_SETUP_MAX 8 299 300 int __init netdev_boot_setup(char *str); 301 302 /* 303 * Structure for NAPI scheduling similar to tasklet but with weighting 304 */ 305 struct napi_struct { 306 /* The poll_list must only be managed by the entity which 307 * changes the state of the NAPI_STATE_SCHED bit. This means 308 * whoever atomically sets that bit can add this napi_struct 309 * to the per-CPU poll_list, and whoever clears that bit 310 * can remove from the list right before clearing the bit. 311 */ 312 struct list_head poll_list; 313 314 unsigned long state; 315 int weight; 316 unsigned int gro_count; 317 int (*poll)(struct napi_struct *, int); 318 #ifdef CONFIG_NETPOLL 319 spinlock_t poll_lock; 320 int poll_owner; 321 #endif 322 struct net_device *dev; 323 struct sk_buff *gro_list; 324 struct sk_buff *skb; 325 struct hrtimer timer; 326 struct list_head dev_list; 327 struct hlist_node napi_hash_node; 328 unsigned int napi_id; 329 }; 330 331 enum { 332 NAPI_STATE_SCHED, /* Poll is scheduled */ 333 NAPI_STATE_DISABLE, /* Disable pending */ 334 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 335 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */ 336 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */ 337 }; 338 339 enum gro_result { 340 GRO_MERGED, 341 GRO_MERGED_FREE, 342 GRO_HELD, 343 GRO_NORMAL, 344 GRO_DROP, 345 }; 346 typedef enum gro_result gro_result_t; 347 348 /* 349 * enum rx_handler_result - Possible return values for rx_handlers. 350 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 351 * further. 352 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 353 * case skb->dev was changed by rx_handler. 354 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 355 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 356 * 357 * rx_handlers are functions called from inside __netif_receive_skb(), to do 358 * special processing of the skb, prior to delivery to protocol handlers. 359 * 360 * Currently, a net_device can only have a single rx_handler registered. Trying 361 * to register a second rx_handler will return -EBUSY. 362 * 363 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 364 * To unregister a rx_handler on a net_device, use 365 * netdev_rx_handler_unregister(). 366 * 367 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 368 * do with the skb. 369 * 370 * If the rx_handler consumed the skb in some way, it should return 371 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 372 * the skb to be delivered in some other way. 373 * 374 * If the rx_handler changed skb->dev, to divert the skb to another 375 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 376 * new device will be called if it exists. 377 * 378 * If the rx_handler decides the skb should be ignored, it should return 379 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 380 * are registered on exact device (ptype->dev == skb->dev). 381 * 382 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 383 * delivered, it should return RX_HANDLER_PASS. 384 * 385 * A device without a registered rx_handler will behave as if rx_handler 386 * returned RX_HANDLER_PASS. 387 */ 388 389 enum rx_handler_result { 390 RX_HANDLER_CONSUMED, 391 RX_HANDLER_ANOTHER, 392 RX_HANDLER_EXACT, 393 RX_HANDLER_PASS, 394 }; 395 typedef enum rx_handler_result rx_handler_result_t; 396 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 397 398 void __napi_schedule(struct napi_struct *n); 399 void __napi_schedule_irqoff(struct napi_struct *n); 400 401 static inline bool napi_disable_pending(struct napi_struct *n) 402 { 403 return test_bit(NAPI_STATE_DISABLE, &n->state); 404 } 405 406 /** 407 * napi_schedule_prep - check if NAPI can be scheduled 408 * @n: NAPI context 409 * 410 * Test if NAPI routine is already running, and if not mark 411 * it as running. This is used as a condition variable to 412 * insure only one NAPI poll instance runs. We also make 413 * sure there is no pending NAPI disable. 414 */ 415 static inline bool napi_schedule_prep(struct napi_struct *n) 416 { 417 return !napi_disable_pending(n) && 418 !test_and_set_bit(NAPI_STATE_SCHED, &n->state); 419 } 420 421 /** 422 * napi_schedule - schedule NAPI poll 423 * @n: NAPI context 424 * 425 * Schedule NAPI poll routine to be called if it is not already 426 * running. 427 */ 428 static inline void napi_schedule(struct napi_struct *n) 429 { 430 if (napi_schedule_prep(n)) 431 __napi_schedule(n); 432 } 433 434 /** 435 * napi_schedule_irqoff - schedule NAPI poll 436 * @n: NAPI context 437 * 438 * Variant of napi_schedule(), assuming hard irqs are masked. 439 */ 440 static inline void napi_schedule_irqoff(struct napi_struct *n) 441 { 442 if (napi_schedule_prep(n)) 443 __napi_schedule_irqoff(n); 444 } 445 446 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 447 static inline bool napi_reschedule(struct napi_struct *napi) 448 { 449 if (napi_schedule_prep(napi)) { 450 __napi_schedule(napi); 451 return true; 452 } 453 return false; 454 } 455 456 void __napi_complete(struct napi_struct *n); 457 void napi_complete_done(struct napi_struct *n, int work_done); 458 /** 459 * napi_complete - NAPI processing complete 460 * @n: NAPI context 461 * 462 * Mark NAPI processing as complete. 463 * Consider using napi_complete_done() instead. 464 */ 465 static inline void napi_complete(struct napi_struct *n) 466 { 467 return napi_complete_done(n, 0); 468 } 469 470 /** 471 * napi_hash_add - add a NAPI to global hashtable 472 * @napi: NAPI context 473 * 474 * Generate a new napi_id and store a @napi under it in napi_hash. 475 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL). 476 * Note: This is normally automatically done from netif_napi_add(), 477 * so might disappear in a future Linux version. 478 */ 479 void napi_hash_add(struct napi_struct *napi); 480 481 /** 482 * napi_hash_del - remove a NAPI from global table 483 * @napi: NAPI context 484 * 485 * Warning: caller must observe RCU grace period 486 * before freeing memory containing @napi, if 487 * this function returns true. 488 * Note: core networking stack automatically calls it 489 * from netif_napi_del(). 490 * Drivers might want to call this helper to combine all 491 * the needed RCU grace periods into a single one. 492 */ 493 bool napi_hash_del(struct napi_struct *napi); 494 495 /** 496 * napi_disable - prevent NAPI from scheduling 497 * @n: NAPI context 498 * 499 * Stop NAPI from being scheduled on this context. 500 * Waits till any outstanding processing completes. 501 */ 502 void napi_disable(struct napi_struct *n); 503 504 /** 505 * napi_enable - enable NAPI scheduling 506 * @n: NAPI context 507 * 508 * Resume NAPI from being scheduled on this context. 509 * Must be paired with napi_disable. 510 */ 511 static inline void napi_enable(struct napi_struct *n) 512 { 513 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 514 smp_mb__before_atomic(); 515 clear_bit(NAPI_STATE_SCHED, &n->state); 516 clear_bit(NAPI_STATE_NPSVC, &n->state); 517 } 518 519 /** 520 * napi_synchronize - wait until NAPI is not running 521 * @n: NAPI context 522 * 523 * Wait until NAPI is done being scheduled on this context. 524 * Waits till any outstanding processing completes but 525 * does not disable future activations. 526 */ 527 static inline void napi_synchronize(const struct napi_struct *n) 528 { 529 if (IS_ENABLED(CONFIG_SMP)) 530 while (test_bit(NAPI_STATE_SCHED, &n->state)) 531 msleep(1); 532 else 533 barrier(); 534 } 535 536 enum netdev_queue_state_t { 537 __QUEUE_STATE_DRV_XOFF, 538 __QUEUE_STATE_STACK_XOFF, 539 __QUEUE_STATE_FROZEN, 540 }; 541 542 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 543 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 544 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 545 546 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 547 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 548 QUEUE_STATE_FROZEN) 549 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 550 QUEUE_STATE_FROZEN) 551 552 /* 553 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 554 * netif_tx_* functions below are used to manipulate this flag. The 555 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 556 * queue independently. The netif_xmit_*stopped functions below are called 557 * to check if the queue has been stopped by the driver or stack (either 558 * of the XOFF bits are set in the state). Drivers should not need to call 559 * netif_xmit*stopped functions, they should only be using netif_tx_*. 560 */ 561 562 struct netdev_queue { 563 /* 564 * read-mostly part 565 */ 566 struct net_device *dev; 567 struct Qdisc __rcu *qdisc; 568 struct Qdisc *qdisc_sleeping; 569 #ifdef CONFIG_SYSFS 570 struct kobject kobj; 571 #endif 572 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 573 int numa_node; 574 #endif 575 unsigned long tx_maxrate; 576 /* 577 * Number of TX timeouts for this queue 578 * (/sys/class/net/DEV/Q/trans_timeout) 579 */ 580 unsigned long trans_timeout; 581 /* 582 * write-mostly part 583 */ 584 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 585 int xmit_lock_owner; 586 /* 587 * Time (in jiffies) of last Tx 588 */ 589 unsigned long trans_start; 590 591 unsigned long state; 592 593 #ifdef CONFIG_BQL 594 struct dql dql; 595 #endif 596 } ____cacheline_aligned_in_smp; 597 598 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 599 { 600 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 601 return q->numa_node; 602 #else 603 return NUMA_NO_NODE; 604 #endif 605 } 606 607 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 608 { 609 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 610 q->numa_node = node; 611 #endif 612 } 613 614 #ifdef CONFIG_RPS 615 /* 616 * This structure holds an RPS map which can be of variable length. The 617 * map is an array of CPUs. 618 */ 619 struct rps_map { 620 unsigned int len; 621 struct rcu_head rcu; 622 u16 cpus[0]; 623 }; 624 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 625 626 /* 627 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 628 * tail pointer for that CPU's input queue at the time of last enqueue, and 629 * a hardware filter index. 630 */ 631 struct rps_dev_flow { 632 u16 cpu; 633 u16 filter; 634 unsigned int last_qtail; 635 }; 636 #define RPS_NO_FILTER 0xffff 637 638 /* 639 * The rps_dev_flow_table structure contains a table of flow mappings. 640 */ 641 struct rps_dev_flow_table { 642 unsigned int mask; 643 struct rcu_head rcu; 644 struct rps_dev_flow flows[0]; 645 }; 646 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 647 ((_num) * sizeof(struct rps_dev_flow))) 648 649 /* 650 * The rps_sock_flow_table contains mappings of flows to the last CPU 651 * on which they were processed by the application (set in recvmsg). 652 * Each entry is a 32bit value. Upper part is the high-order bits 653 * of flow hash, lower part is CPU number. 654 * rps_cpu_mask is used to partition the space, depending on number of 655 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 656 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 657 * meaning we use 32-6=26 bits for the hash. 658 */ 659 struct rps_sock_flow_table { 660 u32 mask; 661 662 u32 ents[0] ____cacheline_aligned_in_smp; 663 }; 664 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 665 666 #define RPS_NO_CPU 0xffff 667 668 extern u32 rps_cpu_mask; 669 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 670 671 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 672 u32 hash) 673 { 674 if (table && hash) { 675 unsigned int index = hash & table->mask; 676 u32 val = hash & ~rps_cpu_mask; 677 678 /* We only give a hint, preemption can change CPU under us */ 679 val |= raw_smp_processor_id(); 680 681 if (table->ents[index] != val) 682 table->ents[index] = val; 683 } 684 } 685 686 #ifdef CONFIG_RFS_ACCEL 687 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 688 u16 filter_id); 689 #endif 690 #endif /* CONFIG_RPS */ 691 692 /* This structure contains an instance of an RX queue. */ 693 struct netdev_rx_queue { 694 #ifdef CONFIG_RPS 695 struct rps_map __rcu *rps_map; 696 struct rps_dev_flow_table __rcu *rps_flow_table; 697 #endif 698 struct kobject kobj; 699 struct net_device *dev; 700 } ____cacheline_aligned_in_smp; 701 702 /* 703 * RX queue sysfs structures and functions. 704 */ 705 struct rx_queue_attribute { 706 struct attribute attr; 707 ssize_t (*show)(struct netdev_rx_queue *queue, 708 struct rx_queue_attribute *attr, char *buf); 709 ssize_t (*store)(struct netdev_rx_queue *queue, 710 struct rx_queue_attribute *attr, const char *buf, size_t len); 711 }; 712 713 #ifdef CONFIG_XPS 714 /* 715 * This structure holds an XPS map which can be of variable length. The 716 * map is an array of queues. 717 */ 718 struct xps_map { 719 unsigned int len; 720 unsigned int alloc_len; 721 struct rcu_head rcu; 722 u16 queues[0]; 723 }; 724 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 725 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 726 - sizeof(struct xps_map)) / sizeof(u16)) 727 728 /* 729 * This structure holds all XPS maps for device. Maps are indexed by CPU. 730 */ 731 struct xps_dev_maps { 732 struct rcu_head rcu; 733 struct xps_map __rcu *cpu_map[0]; 734 }; 735 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \ 736 (nr_cpu_ids * sizeof(struct xps_map *))) 737 #endif /* CONFIG_XPS */ 738 739 #define TC_MAX_QUEUE 16 740 #define TC_BITMASK 15 741 /* HW offloaded queuing disciplines txq count and offset maps */ 742 struct netdev_tc_txq { 743 u16 count; 744 u16 offset; 745 }; 746 747 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 748 /* 749 * This structure is to hold information about the device 750 * configured to run FCoE protocol stack. 751 */ 752 struct netdev_fcoe_hbainfo { 753 char manufacturer[64]; 754 char serial_number[64]; 755 char hardware_version[64]; 756 char driver_version[64]; 757 char optionrom_version[64]; 758 char firmware_version[64]; 759 char model[256]; 760 char model_description[256]; 761 }; 762 #endif 763 764 #define MAX_PHYS_ITEM_ID_LEN 32 765 766 /* This structure holds a unique identifier to identify some 767 * physical item (port for example) used by a netdevice. 768 */ 769 struct netdev_phys_item_id { 770 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 771 unsigned char id_len; 772 }; 773 774 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 775 struct netdev_phys_item_id *b) 776 { 777 return a->id_len == b->id_len && 778 memcmp(a->id, b->id, a->id_len) == 0; 779 } 780 781 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 782 struct sk_buff *skb); 783 784 /* These structures hold the attributes of qdisc and classifiers 785 * that are being passed to the netdevice through the setup_tc op. 786 */ 787 enum { 788 TC_SETUP_MQPRIO, 789 TC_SETUP_CLSU32, 790 TC_SETUP_CLSFLOWER, 791 TC_SETUP_MATCHALL, 792 TC_SETUP_CLSBPF, 793 }; 794 795 struct tc_cls_u32_offload; 796 797 struct tc_to_netdev { 798 unsigned int type; 799 union { 800 u8 tc; 801 struct tc_cls_u32_offload *cls_u32; 802 struct tc_cls_flower_offload *cls_flower; 803 struct tc_cls_matchall_offload *cls_mall; 804 struct tc_cls_bpf_offload *cls_bpf; 805 }; 806 }; 807 808 /* These structures hold the attributes of xdp state that are being passed 809 * to the netdevice through the xdp op. 810 */ 811 enum xdp_netdev_command { 812 /* Set or clear a bpf program used in the earliest stages of packet 813 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 814 * is responsible for calling bpf_prog_put on any old progs that are 815 * stored. In case of error, the callee need not release the new prog 816 * reference, but on success it takes ownership and must bpf_prog_put 817 * when it is no longer used. 818 */ 819 XDP_SETUP_PROG, 820 /* Check if a bpf program is set on the device. The callee should 821 * return true if a program is currently attached and running. 822 */ 823 XDP_QUERY_PROG, 824 }; 825 826 struct netdev_xdp { 827 enum xdp_netdev_command command; 828 union { 829 /* XDP_SETUP_PROG */ 830 struct bpf_prog *prog; 831 /* XDP_QUERY_PROG */ 832 bool prog_attached; 833 }; 834 }; 835 836 /* 837 * This structure defines the management hooks for network devices. 838 * The following hooks can be defined; unless noted otherwise, they are 839 * optional and can be filled with a null pointer. 840 * 841 * int (*ndo_init)(struct net_device *dev); 842 * This function is called once when a network device is registered. 843 * The network device can use this for any late stage initialization 844 * or semantic validation. It can fail with an error code which will 845 * be propagated back to register_netdev. 846 * 847 * void (*ndo_uninit)(struct net_device *dev); 848 * This function is called when device is unregistered or when registration 849 * fails. It is not called if init fails. 850 * 851 * int (*ndo_open)(struct net_device *dev); 852 * This function is called when a network device transitions to the up 853 * state. 854 * 855 * int (*ndo_stop)(struct net_device *dev); 856 * This function is called when a network device transitions to the down 857 * state. 858 * 859 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 860 * struct net_device *dev); 861 * Called when a packet needs to be transmitted. 862 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 863 * the queue before that can happen; it's for obsolete devices and weird 864 * corner cases, but the stack really does a non-trivial amount 865 * of useless work if you return NETDEV_TX_BUSY. 866 * Required; cannot be NULL. 867 * 868 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 869 * netdev_features_t features); 870 * Adjusts the requested feature flags according to device-specific 871 * constraints, and returns the resulting flags. Must not modify 872 * the device state. 873 * 874 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 875 * void *accel_priv, select_queue_fallback_t fallback); 876 * Called to decide which queue to use when device supports multiple 877 * transmit queues. 878 * 879 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 880 * This function is called to allow device receiver to make 881 * changes to configuration when multicast or promiscuous is enabled. 882 * 883 * void (*ndo_set_rx_mode)(struct net_device *dev); 884 * This function is called device changes address list filtering. 885 * If driver handles unicast address filtering, it should set 886 * IFF_UNICAST_FLT in its priv_flags. 887 * 888 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 889 * This function is called when the Media Access Control address 890 * needs to be changed. If this interface is not defined, the 891 * MAC address can not be changed. 892 * 893 * int (*ndo_validate_addr)(struct net_device *dev); 894 * Test if Media Access Control address is valid for the device. 895 * 896 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 897 * Called when a user requests an ioctl which can't be handled by 898 * the generic interface code. If not defined ioctls return 899 * not supported error code. 900 * 901 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 902 * Used to set network devices bus interface parameters. This interface 903 * is retained for legacy reasons; new devices should use the bus 904 * interface (PCI) for low level management. 905 * 906 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 907 * Called when a user wants to change the Maximum Transfer Unit 908 * of a device. If not defined, any request to change MTU will 909 * will return an error. 910 * 911 * void (*ndo_tx_timeout)(struct net_device *dev); 912 * Callback used when the transmitter has not made any progress 913 * for dev->watchdog ticks. 914 * 915 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev, 916 * struct rtnl_link_stats64 *storage); 917 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 918 * Called when a user wants to get the network device usage 919 * statistics. Drivers must do one of the following: 920 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 921 * rtnl_link_stats64 structure passed by the caller. 922 * 2. Define @ndo_get_stats to update a net_device_stats structure 923 * (which should normally be dev->stats) and return a pointer to 924 * it. The structure may be changed asynchronously only if each 925 * field is written atomically. 926 * 3. Update dev->stats asynchronously and atomically, and define 927 * neither operation. 928 * 929 * bool (*ndo_has_offload_stats)(int attr_id) 930 * Return true if this device supports offload stats of this attr_id. 931 * 932 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 933 * void *attr_data) 934 * Get statistics for offload operations by attr_id. Write it into the 935 * attr_data pointer. 936 * 937 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 938 * If device supports VLAN filtering this function is called when a 939 * VLAN id is registered. 940 * 941 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 942 * If device supports VLAN filtering this function is called when a 943 * VLAN id is unregistered. 944 * 945 * void (*ndo_poll_controller)(struct net_device *dev); 946 * 947 * SR-IOV management functions. 948 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 949 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 950 * u8 qos, __be16 proto); 951 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 952 * int max_tx_rate); 953 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 954 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 955 * int (*ndo_get_vf_config)(struct net_device *dev, 956 * int vf, struct ifla_vf_info *ivf); 957 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 958 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 959 * struct nlattr *port[]); 960 * 961 * Enable or disable the VF ability to query its RSS Redirection Table and 962 * Hash Key. This is needed since on some devices VF share this information 963 * with PF and querying it may introduce a theoretical security risk. 964 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 965 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 966 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc) 967 * Called to setup 'tc' number of traffic classes in the net device. This 968 * is always called from the stack with the rtnl lock held and netif tx 969 * queues stopped. This allows the netdevice to perform queue management 970 * safely. 971 * 972 * Fiber Channel over Ethernet (FCoE) offload functions. 973 * int (*ndo_fcoe_enable)(struct net_device *dev); 974 * Called when the FCoE protocol stack wants to start using LLD for FCoE 975 * so the underlying device can perform whatever needed configuration or 976 * initialization to support acceleration of FCoE traffic. 977 * 978 * int (*ndo_fcoe_disable)(struct net_device *dev); 979 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 980 * so the underlying device can perform whatever needed clean-ups to 981 * stop supporting acceleration of FCoE traffic. 982 * 983 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 984 * struct scatterlist *sgl, unsigned int sgc); 985 * Called when the FCoE Initiator wants to initialize an I/O that 986 * is a possible candidate for Direct Data Placement (DDP). The LLD can 987 * perform necessary setup and returns 1 to indicate the device is set up 988 * successfully to perform DDP on this I/O, otherwise this returns 0. 989 * 990 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 991 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 992 * indicated by the FC exchange id 'xid', so the underlying device can 993 * clean up and reuse resources for later DDP requests. 994 * 995 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 996 * struct scatterlist *sgl, unsigned int sgc); 997 * Called when the FCoE Target wants to initialize an I/O that 998 * is a possible candidate for Direct Data Placement (DDP). The LLD can 999 * perform necessary setup and returns 1 to indicate the device is set up 1000 * successfully to perform DDP on this I/O, otherwise this returns 0. 1001 * 1002 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1003 * struct netdev_fcoe_hbainfo *hbainfo); 1004 * Called when the FCoE Protocol stack wants information on the underlying 1005 * device. This information is utilized by the FCoE protocol stack to 1006 * register attributes with Fiber Channel management service as per the 1007 * FC-GS Fabric Device Management Information(FDMI) specification. 1008 * 1009 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1010 * Called when the underlying device wants to override default World Wide 1011 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1012 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1013 * protocol stack to use. 1014 * 1015 * RFS acceleration. 1016 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1017 * u16 rxq_index, u32 flow_id); 1018 * Set hardware filter for RFS. rxq_index is the target queue index; 1019 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1020 * Return the filter ID on success, or a negative error code. 1021 * 1022 * Slave management functions (for bridge, bonding, etc). 1023 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1024 * Called to make another netdev an underling. 1025 * 1026 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1027 * Called to release previously enslaved netdev. 1028 * 1029 * Feature/offload setting functions. 1030 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1031 * Called to update device configuration to new features. Passed 1032 * feature set might be less than what was returned by ndo_fix_features()). 1033 * Must return >0 or -errno if it changed dev->features itself. 1034 * 1035 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1036 * struct net_device *dev, 1037 * const unsigned char *addr, u16 vid, u16 flags) 1038 * Adds an FDB entry to dev for addr. 1039 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1040 * struct net_device *dev, 1041 * const unsigned char *addr, u16 vid) 1042 * Deletes the FDB entry from dev coresponding to addr. 1043 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1044 * struct net_device *dev, struct net_device *filter_dev, 1045 * int *idx) 1046 * Used to add FDB entries to dump requests. Implementers should add 1047 * entries to skb and update idx with the number of entries. 1048 * 1049 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1050 * u16 flags) 1051 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1052 * struct net_device *dev, u32 filter_mask, 1053 * int nlflags) 1054 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1055 * u16 flags); 1056 * 1057 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1058 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1059 * which do not represent real hardware may define this to allow their 1060 * userspace components to manage their virtual carrier state. Devices 1061 * that determine carrier state from physical hardware properties (eg 1062 * network cables) or protocol-dependent mechanisms (eg 1063 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1064 * 1065 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1066 * struct netdev_phys_item_id *ppid); 1067 * Called to get ID of physical port of this device. If driver does 1068 * not implement this, it is assumed that the hw is not able to have 1069 * multiple net devices on single physical port. 1070 * 1071 * void (*ndo_udp_tunnel_add)(struct net_device *dev, 1072 * struct udp_tunnel_info *ti); 1073 * Called by UDP tunnel to notify a driver about the UDP port and socket 1074 * address family that a UDP tunnel is listnening to. It is called only 1075 * when a new port starts listening. The operation is protected by the 1076 * RTNL. 1077 * 1078 * void (*ndo_udp_tunnel_del)(struct net_device *dev, 1079 * struct udp_tunnel_info *ti); 1080 * Called by UDP tunnel to notify the driver about a UDP port and socket 1081 * address family that the UDP tunnel is not listening to anymore. The 1082 * operation is protected by the RTNL. 1083 * 1084 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1085 * struct net_device *dev) 1086 * Called by upper layer devices to accelerate switching or other 1087 * station functionality into hardware. 'pdev is the lowerdev 1088 * to use for the offload and 'dev' is the net device that will 1089 * back the offload. Returns a pointer to the private structure 1090 * the upper layer will maintain. 1091 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1092 * Called by upper layer device to delete the station created 1093 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1094 * the station and priv is the structure returned by the add 1095 * operation. 1096 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb, 1097 * struct net_device *dev, 1098 * void *priv); 1099 * Callback to use for xmit over the accelerated station. This 1100 * is used in place of ndo_start_xmit on accelerated net 1101 * devices. 1102 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1103 * struct net_device *dev 1104 * netdev_features_t features); 1105 * Called by core transmit path to determine if device is capable of 1106 * performing offload operations on a given packet. This is to give 1107 * the device an opportunity to implement any restrictions that cannot 1108 * be otherwise expressed by feature flags. The check is called with 1109 * the set of features that the stack has calculated and it returns 1110 * those the driver believes to be appropriate. 1111 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1112 * int queue_index, u32 maxrate); 1113 * Called when a user wants to set a max-rate limitation of specific 1114 * TX queue. 1115 * int (*ndo_get_iflink)(const struct net_device *dev); 1116 * Called to get the iflink value of this device. 1117 * void (*ndo_change_proto_down)(struct net_device *dev, 1118 * bool proto_down); 1119 * This function is used to pass protocol port error state information 1120 * to the switch driver. The switch driver can react to the proto_down 1121 * by doing a phys down on the associated switch port. 1122 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1123 * This function is used to get egress tunnel information for given skb. 1124 * This is useful for retrieving outer tunnel header parameters while 1125 * sampling packet. 1126 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1127 * This function is used to specify the headroom that the skb must 1128 * consider when allocation skb during packet reception. Setting 1129 * appropriate rx headroom value allows avoiding skb head copy on 1130 * forward. Setting a negative value resets the rx headroom to the 1131 * default value. 1132 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp); 1133 * This function is used to set or query state related to XDP on the 1134 * netdevice. See definition of enum xdp_netdev_command for details. 1135 * 1136 */ 1137 struct net_device_ops { 1138 int (*ndo_init)(struct net_device *dev); 1139 void (*ndo_uninit)(struct net_device *dev); 1140 int (*ndo_open)(struct net_device *dev); 1141 int (*ndo_stop)(struct net_device *dev); 1142 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1143 struct net_device *dev); 1144 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1145 struct net_device *dev, 1146 netdev_features_t features); 1147 u16 (*ndo_select_queue)(struct net_device *dev, 1148 struct sk_buff *skb, 1149 void *accel_priv, 1150 select_queue_fallback_t fallback); 1151 void (*ndo_change_rx_flags)(struct net_device *dev, 1152 int flags); 1153 void (*ndo_set_rx_mode)(struct net_device *dev); 1154 int (*ndo_set_mac_address)(struct net_device *dev, 1155 void *addr); 1156 int (*ndo_validate_addr)(struct net_device *dev); 1157 int (*ndo_do_ioctl)(struct net_device *dev, 1158 struct ifreq *ifr, int cmd); 1159 int (*ndo_set_config)(struct net_device *dev, 1160 struct ifmap *map); 1161 int (*ndo_change_mtu)(struct net_device *dev, 1162 int new_mtu); 1163 int (*ndo_neigh_setup)(struct net_device *dev, 1164 struct neigh_parms *); 1165 void (*ndo_tx_timeout) (struct net_device *dev); 1166 1167 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev, 1168 struct rtnl_link_stats64 *storage); 1169 bool (*ndo_has_offload_stats)(int attr_id); 1170 int (*ndo_get_offload_stats)(int attr_id, 1171 const struct net_device *dev, 1172 void *attr_data); 1173 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1174 1175 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1176 __be16 proto, u16 vid); 1177 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1178 __be16 proto, u16 vid); 1179 #ifdef CONFIG_NET_POLL_CONTROLLER 1180 void (*ndo_poll_controller)(struct net_device *dev); 1181 int (*ndo_netpoll_setup)(struct net_device *dev, 1182 struct netpoll_info *info); 1183 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1184 #endif 1185 #ifdef CONFIG_NET_RX_BUSY_POLL 1186 int (*ndo_busy_poll)(struct napi_struct *dev); 1187 #endif 1188 int (*ndo_set_vf_mac)(struct net_device *dev, 1189 int queue, u8 *mac); 1190 int (*ndo_set_vf_vlan)(struct net_device *dev, 1191 int queue, u16 vlan, 1192 u8 qos, __be16 proto); 1193 int (*ndo_set_vf_rate)(struct net_device *dev, 1194 int vf, int min_tx_rate, 1195 int max_tx_rate); 1196 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1197 int vf, bool setting); 1198 int (*ndo_set_vf_trust)(struct net_device *dev, 1199 int vf, bool setting); 1200 int (*ndo_get_vf_config)(struct net_device *dev, 1201 int vf, 1202 struct ifla_vf_info *ivf); 1203 int (*ndo_set_vf_link_state)(struct net_device *dev, 1204 int vf, int link_state); 1205 int (*ndo_get_vf_stats)(struct net_device *dev, 1206 int vf, 1207 struct ifla_vf_stats 1208 *vf_stats); 1209 int (*ndo_set_vf_port)(struct net_device *dev, 1210 int vf, 1211 struct nlattr *port[]); 1212 int (*ndo_get_vf_port)(struct net_device *dev, 1213 int vf, struct sk_buff *skb); 1214 int (*ndo_set_vf_guid)(struct net_device *dev, 1215 int vf, u64 guid, 1216 int guid_type); 1217 int (*ndo_set_vf_rss_query_en)( 1218 struct net_device *dev, 1219 int vf, bool setting); 1220 int (*ndo_setup_tc)(struct net_device *dev, 1221 u32 handle, 1222 __be16 protocol, 1223 struct tc_to_netdev *tc); 1224 #if IS_ENABLED(CONFIG_FCOE) 1225 int (*ndo_fcoe_enable)(struct net_device *dev); 1226 int (*ndo_fcoe_disable)(struct net_device *dev); 1227 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1228 u16 xid, 1229 struct scatterlist *sgl, 1230 unsigned int sgc); 1231 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1232 u16 xid); 1233 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1234 u16 xid, 1235 struct scatterlist *sgl, 1236 unsigned int sgc); 1237 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1238 struct netdev_fcoe_hbainfo *hbainfo); 1239 #endif 1240 1241 #if IS_ENABLED(CONFIG_LIBFCOE) 1242 #define NETDEV_FCOE_WWNN 0 1243 #define NETDEV_FCOE_WWPN 1 1244 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1245 u64 *wwn, int type); 1246 #endif 1247 1248 #ifdef CONFIG_RFS_ACCEL 1249 int (*ndo_rx_flow_steer)(struct net_device *dev, 1250 const struct sk_buff *skb, 1251 u16 rxq_index, 1252 u32 flow_id); 1253 #endif 1254 int (*ndo_add_slave)(struct net_device *dev, 1255 struct net_device *slave_dev); 1256 int (*ndo_del_slave)(struct net_device *dev, 1257 struct net_device *slave_dev); 1258 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1259 netdev_features_t features); 1260 int (*ndo_set_features)(struct net_device *dev, 1261 netdev_features_t features); 1262 int (*ndo_neigh_construct)(struct net_device *dev, 1263 struct neighbour *n); 1264 void (*ndo_neigh_destroy)(struct net_device *dev, 1265 struct neighbour *n); 1266 1267 int (*ndo_fdb_add)(struct ndmsg *ndm, 1268 struct nlattr *tb[], 1269 struct net_device *dev, 1270 const unsigned char *addr, 1271 u16 vid, 1272 u16 flags); 1273 int (*ndo_fdb_del)(struct ndmsg *ndm, 1274 struct nlattr *tb[], 1275 struct net_device *dev, 1276 const unsigned char *addr, 1277 u16 vid); 1278 int (*ndo_fdb_dump)(struct sk_buff *skb, 1279 struct netlink_callback *cb, 1280 struct net_device *dev, 1281 struct net_device *filter_dev, 1282 int *idx); 1283 1284 int (*ndo_bridge_setlink)(struct net_device *dev, 1285 struct nlmsghdr *nlh, 1286 u16 flags); 1287 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1288 u32 pid, u32 seq, 1289 struct net_device *dev, 1290 u32 filter_mask, 1291 int nlflags); 1292 int (*ndo_bridge_dellink)(struct net_device *dev, 1293 struct nlmsghdr *nlh, 1294 u16 flags); 1295 int (*ndo_change_carrier)(struct net_device *dev, 1296 bool new_carrier); 1297 int (*ndo_get_phys_port_id)(struct net_device *dev, 1298 struct netdev_phys_item_id *ppid); 1299 int (*ndo_get_phys_port_name)(struct net_device *dev, 1300 char *name, size_t len); 1301 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1302 struct udp_tunnel_info *ti); 1303 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1304 struct udp_tunnel_info *ti); 1305 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1306 struct net_device *dev); 1307 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1308 void *priv); 1309 1310 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb, 1311 struct net_device *dev, 1312 void *priv); 1313 int (*ndo_get_lock_subclass)(struct net_device *dev); 1314 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1315 int queue_index, 1316 u32 maxrate); 1317 int (*ndo_get_iflink)(const struct net_device *dev); 1318 int (*ndo_change_proto_down)(struct net_device *dev, 1319 bool proto_down); 1320 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1321 struct sk_buff *skb); 1322 void (*ndo_set_rx_headroom)(struct net_device *dev, 1323 int needed_headroom); 1324 int (*ndo_xdp)(struct net_device *dev, 1325 struct netdev_xdp *xdp); 1326 }; 1327 1328 /** 1329 * enum net_device_priv_flags - &struct net_device priv_flags 1330 * 1331 * These are the &struct net_device, they are only set internally 1332 * by drivers and used in the kernel. These flags are invisible to 1333 * userspace; this means that the order of these flags can change 1334 * during any kernel release. 1335 * 1336 * You should have a pretty good reason to be extending these flags. 1337 * 1338 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1339 * @IFF_EBRIDGE: Ethernet bridging device 1340 * @IFF_BONDING: bonding master or slave 1341 * @IFF_ISATAP: ISATAP interface (RFC4214) 1342 * @IFF_WAN_HDLC: WAN HDLC device 1343 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1344 * release skb->dst 1345 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1346 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1347 * @IFF_MACVLAN_PORT: device used as macvlan port 1348 * @IFF_BRIDGE_PORT: device used as bridge port 1349 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1350 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1351 * @IFF_UNICAST_FLT: Supports unicast filtering 1352 * @IFF_TEAM_PORT: device used as team port 1353 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1354 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1355 * change when it's running 1356 * @IFF_MACVLAN: Macvlan device 1357 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1358 * underlying stacked devices 1359 * @IFF_IPVLAN_MASTER: IPvlan master device 1360 * @IFF_IPVLAN_SLAVE: IPvlan slave device 1361 * @IFF_L3MDEV_MASTER: device is an L3 master device 1362 * @IFF_NO_QUEUE: device can run without qdisc attached 1363 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1364 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1365 * @IFF_TEAM: device is a team device 1366 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1367 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1368 * entity (i.e. the master device for bridged veth) 1369 * @IFF_MACSEC: device is a MACsec device 1370 */ 1371 enum netdev_priv_flags { 1372 IFF_802_1Q_VLAN = 1<<0, 1373 IFF_EBRIDGE = 1<<1, 1374 IFF_BONDING = 1<<2, 1375 IFF_ISATAP = 1<<3, 1376 IFF_WAN_HDLC = 1<<4, 1377 IFF_XMIT_DST_RELEASE = 1<<5, 1378 IFF_DONT_BRIDGE = 1<<6, 1379 IFF_DISABLE_NETPOLL = 1<<7, 1380 IFF_MACVLAN_PORT = 1<<8, 1381 IFF_BRIDGE_PORT = 1<<9, 1382 IFF_OVS_DATAPATH = 1<<10, 1383 IFF_TX_SKB_SHARING = 1<<11, 1384 IFF_UNICAST_FLT = 1<<12, 1385 IFF_TEAM_PORT = 1<<13, 1386 IFF_SUPP_NOFCS = 1<<14, 1387 IFF_LIVE_ADDR_CHANGE = 1<<15, 1388 IFF_MACVLAN = 1<<16, 1389 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1390 IFF_IPVLAN_MASTER = 1<<18, 1391 IFF_IPVLAN_SLAVE = 1<<19, 1392 IFF_L3MDEV_MASTER = 1<<20, 1393 IFF_NO_QUEUE = 1<<21, 1394 IFF_OPENVSWITCH = 1<<22, 1395 IFF_L3MDEV_SLAVE = 1<<23, 1396 IFF_TEAM = 1<<24, 1397 IFF_RXFH_CONFIGURED = 1<<25, 1398 IFF_PHONY_HEADROOM = 1<<26, 1399 IFF_MACSEC = 1<<27, 1400 }; 1401 1402 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1403 #define IFF_EBRIDGE IFF_EBRIDGE 1404 #define IFF_BONDING IFF_BONDING 1405 #define IFF_ISATAP IFF_ISATAP 1406 #define IFF_WAN_HDLC IFF_WAN_HDLC 1407 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1408 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1409 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1410 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1411 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1412 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1413 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1414 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1415 #define IFF_TEAM_PORT IFF_TEAM_PORT 1416 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1417 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1418 #define IFF_MACVLAN IFF_MACVLAN 1419 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1420 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER 1421 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE 1422 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1423 #define IFF_NO_QUEUE IFF_NO_QUEUE 1424 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1425 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1426 #define IFF_TEAM IFF_TEAM 1427 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1428 #define IFF_MACSEC IFF_MACSEC 1429 1430 /** 1431 * struct net_device - The DEVICE structure. 1432 * Actually, this whole structure is a big mistake. It mixes I/O 1433 * data with strictly "high-level" data, and it has to know about 1434 * almost every data structure used in the INET module. 1435 * 1436 * @name: This is the first field of the "visible" part of this structure 1437 * (i.e. as seen by users in the "Space.c" file). It is the name 1438 * of the interface. 1439 * 1440 * @name_hlist: Device name hash chain, please keep it close to name[] 1441 * @ifalias: SNMP alias 1442 * @mem_end: Shared memory end 1443 * @mem_start: Shared memory start 1444 * @base_addr: Device I/O address 1445 * @irq: Device IRQ number 1446 * 1447 * @carrier_changes: Stats to monitor carrier on<->off transitions 1448 * 1449 * @state: Generic network queuing layer state, see netdev_state_t 1450 * @dev_list: The global list of network devices 1451 * @napi_list: List entry used for polling NAPI devices 1452 * @unreg_list: List entry when we are unregistering the 1453 * device; see the function unregister_netdev 1454 * @close_list: List entry used when we are closing the device 1455 * @ptype_all: Device-specific packet handlers for all protocols 1456 * @ptype_specific: Device-specific, protocol-specific packet handlers 1457 * 1458 * @adj_list: Directly linked devices, like slaves for bonding 1459 * @all_adj_list: All linked devices, *including* neighbours 1460 * @features: Currently active device features 1461 * @hw_features: User-changeable features 1462 * 1463 * @wanted_features: User-requested features 1464 * @vlan_features: Mask of features inheritable by VLAN devices 1465 * 1466 * @hw_enc_features: Mask of features inherited by encapsulating devices 1467 * This field indicates what encapsulation 1468 * offloads the hardware is capable of doing, 1469 * and drivers will need to set them appropriately. 1470 * 1471 * @mpls_features: Mask of features inheritable by MPLS 1472 * 1473 * @ifindex: interface index 1474 * @group: The group the device belongs to 1475 * 1476 * @stats: Statistics struct, which was left as a legacy, use 1477 * rtnl_link_stats64 instead 1478 * 1479 * @rx_dropped: Dropped packets by core network, 1480 * do not use this in drivers 1481 * @tx_dropped: Dropped packets by core network, 1482 * do not use this in drivers 1483 * @rx_nohandler: nohandler dropped packets by core network on 1484 * inactive devices, do not use this in drivers 1485 * 1486 * @wireless_handlers: List of functions to handle Wireless Extensions, 1487 * instead of ioctl, 1488 * see <net/iw_handler.h> for details. 1489 * @wireless_data: Instance data managed by the core of wireless extensions 1490 * 1491 * @netdev_ops: Includes several pointers to callbacks, 1492 * if one wants to override the ndo_*() functions 1493 * @ethtool_ops: Management operations 1494 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1495 * discovery handling. Necessary for e.g. 6LoWPAN. 1496 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1497 * of Layer 2 headers. 1498 * 1499 * @flags: Interface flags (a la BSD) 1500 * @priv_flags: Like 'flags' but invisible to userspace, 1501 * see if.h for the definitions 1502 * @gflags: Global flags ( kept as legacy ) 1503 * @padded: How much padding added by alloc_netdev() 1504 * @operstate: RFC2863 operstate 1505 * @link_mode: Mapping policy to operstate 1506 * @if_port: Selectable AUI, TP, ... 1507 * @dma: DMA channel 1508 * @mtu: Interface MTU value 1509 * @type: Interface hardware type 1510 * @hard_header_len: Maximum hardware header length. 1511 * 1512 * @needed_headroom: Extra headroom the hardware may need, but not in all 1513 * cases can this be guaranteed 1514 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1515 * cases can this be guaranteed. Some cases also use 1516 * LL_MAX_HEADER instead to allocate the skb 1517 * 1518 * interface address info: 1519 * 1520 * @perm_addr: Permanent hw address 1521 * @addr_assign_type: Hw address assignment type 1522 * @addr_len: Hardware address length 1523 * @neigh_priv_len: Used in neigh_alloc() 1524 * @dev_id: Used to differentiate devices that share 1525 * the same link layer address 1526 * @dev_port: Used to differentiate devices that share 1527 * the same function 1528 * @addr_list_lock: XXX: need comments on this one 1529 * @uc_promisc: Counter that indicates promiscuous mode 1530 * has been enabled due to the need to listen to 1531 * additional unicast addresses in a device that 1532 * does not implement ndo_set_rx_mode() 1533 * @uc: unicast mac addresses 1534 * @mc: multicast mac addresses 1535 * @dev_addrs: list of device hw addresses 1536 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1537 * @promiscuity: Number of times the NIC is told to work in 1538 * promiscuous mode; if it becomes 0 the NIC will 1539 * exit promiscuous mode 1540 * @allmulti: Counter, enables or disables allmulticast mode 1541 * 1542 * @vlan_info: VLAN info 1543 * @dsa_ptr: dsa specific data 1544 * @tipc_ptr: TIPC specific data 1545 * @atalk_ptr: AppleTalk link 1546 * @ip_ptr: IPv4 specific data 1547 * @dn_ptr: DECnet specific data 1548 * @ip6_ptr: IPv6 specific data 1549 * @ax25_ptr: AX.25 specific data 1550 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1551 * 1552 * @last_rx: Time of last Rx 1553 * @dev_addr: Hw address (before bcast, 1554 * because most packets are unicast) 1555 * 1556 * @_rx: Array of RX queues 1557 * @num_rx_queues: Number of RX queues 1558 * allocated at register_netdev() time 1559 * @real_num_rx_queues: Number of RX queues currently active in device 1560 * 1561 * @rx_handler: handler for received packets 1562 * @rx_handler_data: XXX: need comments on this one 1563 * @ingress_queue: XXX: need comments on this one 1564 * @broadcast: hw bcast address 1565 * 1566 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1567 * indexed by RX queue number. Assigned by driver. 1568 * This must only be set if the ndo_rx_flow_steer 1569 * operation is defined 1570 * @index_hlist: Device index hash chain 1571 * 1572 * @_tx: Array of TX queues 1573 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1574 * @real_num_tx_queues: Number of TX queues currently active in device 1575 * @qdisc: Root qdisc from userspace point of view 1576 * @tx_queue_len: Max frames per queue allowed 1577 * @tx_global_lock: XXX: need comments on this one 1578 * 1579 * @xps_maps: XXX: need comments on this one 1580 * 1581 * @watchdog_timeo: Represents the timeout that is used by 1582 * the watchdog (see dev_watchdog()) 1583 * @watchdog_timer: List of timers 1584 * 1585 * @pcpu_refcnt: Number of references to this device 1586 * @todo_list: Delayed register/unregister 1587 * @link_watch_list: XXX: need comments on this one 1588 * 1589 * @reg_state: Register/unregister state machine 1590 * @dismantle: Device is going to be freed 1591 * @rtnl_link_state: This enum represents the phases of creating 1592 * a new link 1593 * 1594 * @destructor: Called from unregister, 1595 * can be used to call free_netdev 1596 * @npinfo: XXX: need comments on this one 1597 * @nd_net: Network namespace this network device is inside 1598 * 1599 * @ml_priv: Mid-layer private 1600 * @lstats: Loopback statistics 1601 * @tstats: Tunnel statistics 1602 * @dstats: Dummy statistics 1603 * @vstats: Virtual ethernet statistics 1604 * 1605 * @garp_port: GARP 1606 * @mrp_port: MRP 1607 * 1608 * @dev: Class/net/name entry 1609 * @sysfs_groups: Space for optional device, statistics and wireless 1610 * sysfs groups 1611 * 1612 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1613 * @rtnl_link_ops: Rtnl_link_ops 1614 * 1615 * @gso_max_size: Maximum size of generic segmentation offload 1616 * @gso_max_segs: Maximum number of segments that can be passed to the 1617 * NIC for GSO 1618 * 1619 * @dcbnl_ops: Data Center Bridging netlink ops 1620 * @num_tc: Number of traffic classes in the net device 1621 * @tc_to_txq: XXX: need comments on this one 1622 * @prio_tc_map XXX: need comments on this one 1623 * 1624 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1625 * 1626 * @priomap: XXX: need comments on this one 1627 * @phydev: Physical device may attach itself 1628 * for hardware timestamping 1629 * 1630 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1631 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1632 * 1633 * @proto_down: protocol port state information can be sent to the 1634 * switch driver and used to set the phys state of the 1635 * switch port. 1636 * 1637 * FIXME: cleanup struct net_device such that network protocol info 1638 * moves out. 1639 */ 1640 1641 struct net_device { 1642 char name[IFNAMSIZ]; 1643 struct hlist_node name_hlist; 1644 char *ifalias; 1645 /* 1646 * I/O specific fields 1647 * FIXME: Merge these and struct ifmap into one 1648 */ 1649 unsigned long mem_end; 1650 unsigned long mem_start; 1651 unsigned long base_addr; 1652 int irq; 1653 1654 atomic_t carrier_changes; 1655 1656 /* 1657 * Some hardware also needs these fields (state,dev_list, 1658 * napi_list,unreg_list,close_list) but they are not 1659 * part of the usual set specified in Space.c. 1660 */ 1661 1662 unsigned long state; 1663 1664 struct list_head dev_list; 1665 struct list_head napi_list; 1666 struct list_head unreg_list; 1667 struct list_head close_list; 1668 struct list_head ptype_all; 1669 struct list_head ptype_specific; 1670 1671 struct { 1672 struct list_head upper; 1673 struct list_head lower; 1674 } adj_list; 1675 1676 struct { 1677 struct list_head upper; 1678 struct list_head lower; 1679 } all_adj_list; 1680 1681 netdev_features_t features; 1682 netdev_features_t hw_features; 1683 netdev_features_t wanted_features; 1684 netdev_features_t vlan_features; 1685 netdev_features_t hw_enc_features; 1686 netdev_features_t mpls_features; 1687 netdev_features_t gso_partial_features; 1688 1689 int ifindex; 1690 int group; 1691 1692 struct net_device_stats stats; 1693 1694 atomic_long_t rx_dropped; 1695 atomic_long_t tx_dropped; 1696 atomic_long_t rx_nohandler; 1697 1698 #ifdef CONFIG_WIRELESS_EXT 1699 const struct iw_handler_def *wireless_handlers; 1700 struct iw_public_data *wireless_data; 1701 #endif 1702 const struct net_device_ops *netdev_ops; 1703 const struct ethtool_ops *ethtool_ops; 1704 #ifdef CONFIG_NET_SWITCHDEV 1705 const struct switchdev_ops *switchdev_ops; 1706 #endif 1707 #ifdef CONFIG_NET_L3_MASTER_DEV 1708 const struct l3mdev_ops *l3mdev_ops; 1709 #endif 1710 #if IS_ENABLED(CONFIG_IPV6) 1711 const struct ndisc_ops *ndisc_ops; 1712 #endif 1713 1714 const struct header_ops *header_ops; 1715 1716 unsigned int flags; 1717 unsigned int priv_flags; 1718 1719 unsigned short gflags; 1720 unsigned short padded; 1721 1722 unsigned char operstate; 1723 unsigned char link_mode; 1724 1725 unsigned char if_port; 1726 unsigned char dma; 1727 1728 unsigned int mtu; 1729 unsigned short type; 1730 unsigned short hard_header_len; 1731 1732 unsigned short needed_headroom; 1733 unsigned short needed_tailroom; 1734 1735 /* Interface address info. */ 1736 unsigned char perm_addr[MAX_ADDR_LEN]; 1737 unsigned char addr_assign_type; 1738 unsigned char addr_len; 1739 unsigned short neigh_priv_len; 1740 unsigned short dev_id; 1741 unsigned short dev_port; 1742 spinlock_t addr_list_lock; 1743 unsigned char name_assign_type; 1744 bool uc_promisc; 1745 struct netdev_hw_addr_list uc; 1746 struct netdev_hw_addr_list mc; 1747 struct netdev_hw_addr_list dev_addrs; 1748 1749 #ifdef CONFIG_SYSFS 1750 struct kset *queues_kset; 1751 #endif 1752 unsigned int promiscuity; 1753 unsigned int allmulti; 1754 1755 1756 /* Protocol-specific pointers */ 1757 1758 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1759 struct vlan_info __rcu *vlan_info; 1760 #endif 1761 #if IS_ENABLED(CONFIG_NET_DSA) 1762 struct dsa_switch_tree *dsa_ptr; 1763 #endif 1764 #if IS_ENABLED(CONFIG_TIPC) 1765 struct tipc_bearer __rcu *tipc_ptr; 1766 #endif 1767 void *atalk_ptr; 1768 struct in_device __rcu *ip_ptr; 1769 struct dn_dev __rcu *dn_ptr; 1770 struct inet6_dev __rcu *ip6_ptr; 1771 void *ax25_ptr; 1772 struct wireless_dev *ieee80211_ptr; 1773 struct wpan_dev *ieee802154_ptr; 1774 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 1775 struct mpls_dev __rcu *mpls_ptr; 1776 #endif 1777 1778 /* 1779 * Cache lines mostly used on receive path (including eth_type_trans()) 1780 */ 1781 unsigned long last_rx; 1782 1783 /* Interface address info used in eth_type_trans() */ 1784 unsigned char *dev_addr; 1785 1786 #ifdef CONFIG_SYSFS 1787 struct netdev_rx_queue *_rx; 1788 1789 unsigned int num_rx_queues; 1790 unsigned int real_num_rx_queues; 1791 #endif 1792 1793 unsigned long gro_flush_timeout; 1794 rx_handler_func_t __rcu *rx_handler; 1795 void __rcu *rx_handler_data; 1796 1797 #ifdef CONFIG_NET_CLS_ACT 1798 struct tcf_proto __rcu *ingress_cl_list; 1799 #endif 1800 struct netdev_queue __rcu *ingress_queue; 1801 #ifdef CONFIG_NETFILTER_INGRESS 1802 struct nf_hook_entry __rcu *nf_hooks_ingress; 1803 #endif 1804 1805 unsigned char broadcast[MAX_ADDR_LEN]; 1806 #ifdef CONFIG_RFS_ACCEL 1807 struct cpu_rmap *rx_cpu_rmap; 1808 #endif 1809 struct hlist_node index_hlist; 1810 1811 /* 1812 * Cache lines mostly used on transmit path 1813 */ 1814 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1815 unsigned int num_tx_queues; 1816 unsigned int real_num_tx_queues; 1817 struct Qdisc *qdisc; 1818 #ifdef CONFIG_NET_SCHED 1819 DECLARE_HASHTABLE (qdisc_hash, 4); 1820 #endif 1821 unsigned long tx_queue_len; 1822 spinlock_t tx_global_lock; 1823 int watchdog_timeo; 1824 1825 #ifdef CONFIG_XPS 1826 struct xps_dev_maps __rcu *xps_maps; 1827 #endif 1828 #ifdef CONFIG_NET_CLS_ACT 1829 struct tcf_proto __rcu *egress_cl_list; 1830 #endif 1831 1832 /* These may be needed for future network-power-down code. */ 1833 struct timer_list watchdog_timer; 1834 1835 int __percpu *pcpu_refcnt; 1836 struct list_head todo_list; 1837 1838 struct list_head link_watch_list; 1839 1840 enum { NETREG_UNINITIALIZED=0, 1841 NETREG_REGISTERED, /* completed register_netdevice */ 1842 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1843 NETREG_UNREGISTERED, /* completed unregister todo */ 1844 NETREG_RELEASED, /* called free_netdev */ 1845 NETREG_DUMMY, /* dummy device for NAPI poll */ 1846 } reg_state:8; 1847 1848 bool dismantle; 1849 1850 enum { 1851 RTNL_LINK_INITIALIZED, 1852 RTNL_LINK_INITIALIZING, 1853 } rtnl_link_state:16; 1854 1855 void (*destructor)(struct net_device *dev); 1856 1857 #ifdef CONFIG_NETPOLL 1858 struct netpoll_info __rcu *npinfo; 1859 #endif 1860 1861 possible_net_t nd_net; 1862 1863 /* mid-layer private */ 1864 union { 1865 void *ml_priv; 1866 struct pcpu_lstats __percpu *lstats; 1867 struct pcpu_sw_netstats __percpu *tstats; 1868 struct pcpu_dstats __percpu *dstats; 1869 struct pcpu_vstats __percpu *vstats; 1870 }; 1871 1872 struct garp_port __rcu *garp_port; 1873 struct mrp_port __rcu *mrp_port; 1874 1875 struct device dev; 1876 const struct attribute_group *sysfs_groups[4]; 1877 const struct attribute_group *sysfs_rx_queue_group; 1878 1879 const struct rtnl_link_ops *rtnl_link_ops; 1880 1881 /* for setting kernel sock attribute on TCP connection setup */ 1882 #define GSO_MAX_SIZE 65536 1883 unsigned int gso_max_size; 1884 #define GSO_MAX_SEGS 65535 1885 u16 gso_max_segs; 1886 1887 #ifdef CONFIG_DCB 1888 const struct dcbnl_rtnl_ops *dcbnl_ops; 1889 #endif 1890 u8 num_tc; 1891 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 1892 u8 prio_tc_map[TC_BITMASK + 1]; 1893 1894 #if IS_ENABLED(CONFIG_FCOE) 1895 unsigned int fcoe_ddp_xid; 1896 #endif 1897 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 1898 struct netprio_map __rcu *priomap; 1899 #endif 1900 struct phy_device *phydev; 1901 struct lock_class_key *qdisc_tx_busylock; 1902 struct lock_class_key *qdisc_running_key; 1903 bool proto_down; 1904 }; 1905 #define to_net_dev(d) container_of(d, struct net_device, dev) 1906 1907 #define NETDEV_ALIGN 32 1908 1909 static inline 1910 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 1911 { 1912 return dev->prio_tc_map[prio & TC_BITMASK]; 1913 } 1914 1915 static inline 1916 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 1917 { 1918 if (tc >= dev->num_tc) 1919 return -EINVAL; 1920 1921 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 1922 return 0; 1923 } 1924 1925 static inline 1926 void netdev_reset_tc(struct net_device *dev) 1927 { 1928 dev->num_tc = 0; 1929 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq)); 1930 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map)); 1931 } 1932 1933 static inline 1934 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset) 1935 { 1936 if (tc >= dev->num_tc) 1937 return -EINVAL; 1938 1939 dev->tc_to_txq[tc].count = count; 1940 dev->tc_to_txq[tc].offset = offset; 1941 return 0; 1942 } 1943 1944 static inline 1945 int netdev_set_num_tc(struct net_device *dev, u8 num_tc) 1946 { 1947 if (num_tc > TC_MAX_QUEUE) 1948 return -EINVAL; 1949 1950 dev->num_tc = num_tc; 1951 return 0; 1952 } 1953 1954 static inline 1955 int netdev_get_num_tc(struct net_device *dev) 1956 { 1957 return dev->num_tc; 1958 } 1959 1960 static inline 1961 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 1962 unsigned int index) 1963 { 1964 return &dev->_tx[index]; 1965 } 1966 1967 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 1968 const struct sk_buff *skb) 1969 { 1970 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 1971 } 1972 1973 static inline void netdev_for_each_tx_queue(struct net_device *dev, 1974 void (*f)(struct net_device *, 1975 struct netdev_queue *, 1976 void *), 1977 void *arg) 1978 { 1979 unsigned int i; 1980 1981 for (i = 0; i < dev->num_tx_queues; i++) 1982 f(dev, &dev->_tx[i], arg); 1983 } 1984 1985 #define netdev_lockdep_set_classes(dev) \ 1986 { \ 1987 static struct lock_class_key qdisc_tx_busylock_key; \ 1988 static struct lock_class_key qdisc_running_key; \ 1989 static struct lock_class_key qdisc_xmit_lock_key; \ 1990 static struct lock_class_key dev_addr_list_lock_key; \ 1991 unsigned int i; \ 1992 \ 1993 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 1994 (dev)->qdisc_running_key = &qdisc_running_key; \ 1995 lockdep_set_class(&(dev)->addr_list_lock, \ 1996 &dev_addr_list_lock_key); \ 1997 for (i = 0; i < (dev)->num_tx_queues; i++) \ 1998 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 1999 &qdisc_xmit_lock_key); \ 2000 } 2001 2002 struct netdev_queue *netdev_pick_tx(struct net_device *dev, 2003 struct sk_buff *skb, 2004 void *accel_priv); 2005 2006 /* returns the headroom that the master device needs to take in account 2007 * when forwarding to this dev 2008 */ 2009 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2010 { 2011 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2012 } 2013 2014 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2015 { 2016 if (dev->netdev_ops->ndo_set_rx_headroom) 2017 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2018 } 2019 2020 /* set the device rx headroom to the dev's default */ 2021 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2022 { 2023 netdev_set_rx_headroom(dev, -1); 2024 } 2025 2026 /* 2027 * Net namespace inlines 2028 */ 2029 static inline 2030 struct net *dev_net(const struct net_device *dev) 2031 { 2032 return read_pnet(&dev->nd_net); 2033 } 2034 2035 static inline 2036 void dev_net_set(struct net_device *dev, struct net *net) 2037 { 2038 write_pnet(&dev->nd_net, net); 2039 } 2040 2041 static inline bool netdev_uses_dsa(struct net_device *dev) 2042 { 2043 #if IS_ENABLED(CONFIG_NET_DSA) 2044 if (dev->dsa_ptr != NULL) 2045 return dsa_uses_tagged_protocol(dev->dsa_ptr); 2046 #endif 2047 return false; 2048 } 2049 2050 /** 2051 * netdev_priv - access network device private data 2052 * @dev: network device 2053 * 2054 * Get network device private data 2055 */ 2056 static inline void *netdev_priv(const struct net_device *dev) 2057 { 2058 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2059 } 2060 2061 /* Set the sysfs physical device reference for the network logical device 2062 * if set prior to registration will cause a symlink during initialization. 2063 */ 2064 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2065 2066 /* Set the sysfs device type for the network logical device to allow 2067 * fine-grained identification of different network device types. For 2068 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2069 */ 2070 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2071 2072 /* Default NAPI poll() weight 2073 * Device drivers are strongly advised to not use bigger value 2074 */ 2075 #define NAPI_POLL_WEIGHT 64 2076 2077 /** 2078 * netif_napi_add - initialize a NAPI context 2079 * @dev: network device 2080 * @napi: NAPI context 2081 * @poll: polling function 2082 * @weight: default weight 2083 * 2084 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2085 * *any* of the other NAPI-related functions. 2086 */ 2087 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2088 int (*poll)(struct napi_struct *, int), int weight); 2089 2090 /** 2091 * netif_tx_napi_add - initialize a NAPI context 2092 * @dev: network device 2093 * @napi: NAPI context 2094 * @poll: polling function 2095 * @weight: default weight 2096 * 2097 * This variant of netif_napi_add() should be used from drivers using NAPI 2098 * to exclusively poll a TX queue. 2099 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2100 */ 2101 static inline void netif_tx_napi_add(struct net_device *dev, 2102 struct napi_struct *napi, 2103 int (*poll)(struct napi_struct *, int), 2104 int weight) 2105 { 2106 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2107 netif_napi_add(dev, napi, poll, weight); 2108 } 2109 2110 /** 2111 * netif_napi_del - remove a NAPI context 2112 * @napi: NAPI context 2113 * 2114 * netif_napi_del() removes a NAPI context from the network device NAPI list 2115 */ 2116 void netif_napi_del(struct napi_struct *napi); 2117 2118 struct napi_gro_cb { 2119 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2120 void *frag0; 2121 2122 /* Length of frag0. */ 2123 unsigned int frag0_len; 2124 2125 /* This indicates where we are processing relative to skb->data. */ 2126 int data_offset; 2127 2128 /* This is non-zero if the packet cannot be merged with the new skb. */ 2129 u16 flush; 2130 2131 /* Save the IP ID here and check when we get to the transport layer */ 2132 u16 flush_id; 2133 2134 /* Number of segments aggregated. */ 2135 u16 count; 2136 2137 /* Start offset for remote checksum offload */ 2138 u16 gro_remcsum_start; 2139 2140 /* jiffies when first packet was created/queued */ 2141 unsigned long age; 2142 2143 /* Used in ipv6_gro_receive() and foo-over-udp */ 2144 u16 proto; 2145 2146 /* This is non-zero if the packet may be of the same flow. */ 2147 u8 same_flow:1; 2148 2149 /* Used in tunnel GRO receive */ 2150 u8 encap_mark:1; 2151 2152 /* GRO checksum is valid */ 2153 u8 csum_valid:1; 2154 2155 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2156 u8 csum_cnt:3; 2157 2158 /* Free the skb? */ 2159 u8 free:2; 2160 #define NAPI_GRO_FREE 1 2161 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2162 2163 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2164 u8 is_ipv6:1; 2165 2166 /* Used in GRE, set in fou/gue_gro_receive */ 2167 u8 is_fou:1; 2168 2169 /* Used to determine if flush_id can be ignored */ 2170 u8 is_atomic:1; 2171 2172 /* 5 bit hole */ 2173 2174 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2175 __wsum csum; 2176 2177 /* used in skb_gro_receive() slow path */ 2178 struct sk_buff *last; 2179 }; 2180 2181 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2182 2183 struct packet_type { 2184 __be16 type; /* This is really htons(ether_type). */ 2185 struct net_device *dev; /* NULL is wildcarded here */ 2186 int (*func) (struct sk_buff *, 2187 struct net_device *, 2188 struct packet_type *, 2189 struct net_device *); 2190 bool (*id_match)(struct packet_type *ptype, 2191 struct sock *sk); 2192 void *af_packet_priv; 2193 struct list_head list; 2194 }; 2195 2196 struct offload_callbacks { 2197 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2198 netdev_features_t features); 2199 struct sk_buff **(*gro_receive)(struct sk_buff **head, 2200 struct sk_buff *skb); 2201 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2202 }; 2203 2204 struct packet_offload { 2205 __be16 type; /* This is really htons(ether_type). */ 2206 u16 priority; 2207 struct offload_callbacks callbacks; 2208 struct list_head list; 2209 }; 2210 2211 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2212 struct pcpu_sw_netstats { 2213 u64 rx_packets; 2214 u64 rx_bytes; 2215 u64 tx_packets; 2216 u64 tx_bytes; 2217 struct u64_stats_sync syncp; 2218 }; 2219 2220 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2221 ({ \ 2222 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2223 if (pcpu_stats) { \ 2224 int __cpu; \ 2225 for_each_possible_cpu(__cpu) { \ 2226 typeof(type) *stat; \ 2227 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2228 u64_stats_init(&stat->syncp); \ 2229 } \ 2230 } \ 2231 pcpu_stats; \ 2232 }) 2233 2234 #define netdev_alloc_pcpu_stats(type) \ 2235 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2236 2237 enum netdev_lag_tx_type { 2238 NETDEV_LAG_TX_TYPE_UNKNOWN, 2239 NETDEV_LAG_TX_TYPE_RANDOM, 2240 NETDEV_LAG_TX_TYPE_BROADCAST, 2241 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2242 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2243 NETDEV_LAG_TX_TYPE_HASH, 2244 }; 2245 2246 struct netdev_lag_upper_info { 2247 enum netdev_lag_tx_type tx_type; 2248 }; 2249 2250 struct netdev_lag_lower_state_info { 2251 u8 link_up : 1, 2252 tx_enabled : 1; 2253 }; 2254 2255 #include <linux/notifier.h> 2256 2257 /* netdevice notifier chain. Please remember to update the rtnetlink 2258 * notification exclusion list in rtnetlink_event() when adding new 2259 * types. 2260 */ 2261 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */ 2262 #define NETDEV_DOWN 0x0002 2263 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface 2264 detected a hardware crash and restarted 2265 - we can use this eg to kick tcp sessions 2266 once done */ 2267 #define NETDEV_CHANGE 0x0004 /* Notify device state change */ 2268 #define NETDEV_REGISTER 0x0005 2269 #define NETDEV_UNREGISTER 0x0006 2270 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */ 2271 #define NETDEV_CHANGEADDR 0x0008 2272 #define NETDEV_GOING_DOWN 0x0009 2273 #define NETDEV_CHANGENAME 0x000A 2274 #define NETDEV_FEAT_CHANGE 0x000B 2275 #define NETDEV_BONDING_FAILOVER 0x000C 2276 #define NETDEV_PRE_UP 0x000D 2277 #define NETDEV_PRE_TYPE_CHANGE 0x000E 2278 #define NETDEV_POST_TYPE_CHANGE 0x000F 2279 #define NETDEV_POST_INIT 0x0010 2280 #define NETDEV_UNREGISTER_FINAL 0x0011 2281 #define NETDEV_RELEASE 0x0012 2282 #define NETDEV_NOTIFY_PEERS 0x0013 2283 #define NETDEV_JOIN 0x0014 2284 #define NETDEV_CHANGEUPPER 0x0015 2285 #define NETDEV_RESEND_IGMP 0x0016 2286 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */ 2287 #define NETDEV_CHANGEINFODATA 0x0018 2288 #define NETDEV_BONDING_INFO 0x0019 2289 #define NETDEV_PRECHANGEUPPER 0x001A 2290 #define NETDEV_CHANGELOWERSTATE 0x001B 2291 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C 2292 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E 2293 2294 int register_netdevice_notifier(struct notifier_block *nb); 2295 int unregister_netdevice_notifier(struct notifier_block *nb); 2296 2297 struct netdev_notifier_info { 2298 struct net_device *dev; 2299 }; 2300 2301 struct netdev_notifier_change_info { 2302 struct netdev_notifier_info info; /* must be first */ 2303 unsigned int flags_changed; 2304 }; 2305 2306 struct netdev_notifier_changeupper_info { 2307 struct netdev_notifier_info info; /* must be first */ 2308 struct net_device *upper_dev; /* new upper dev */ 2309 bool master; /* is upper dev master */ 2310 bool linking; /* is the notification for link or unlink */ 2311 void *upper_info; /* upper dev info */ 2312 }; 2313 2314 struct netdev_notifier_changelowerstate_info { 2315 struct netdev_notifier_info info; /* must be first */ 2316 void *lower_state_info; /* is lower dev state */ 2317 }; 2318 2319 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2320 struct net_device *dev) 2321 { 2322 info->dev = dev; 2323 } 2324 2325 static inline struct net_device * 2326 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2327 { 2328 return info->dev; 2329 } 2330 2331 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2332 2333 2334 extern rwlock_t dev_base_lock; /* Device list lock */ 2335 2336 #define for_each_netdev(net, d) \ 2337 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2338 #define for_each_netdev_reverse(net, d) \ 2339 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2340 #define for_each_netdev_rcu(net, d) \ 2341 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2342 #define for_each_netdev_safe(net, d, n) \ 2343 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2344 #define for_each_netdev_continue(net, d) \ 2345 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2346 #define for_each_netdev_continue_rcu(net, d) \ 2347 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2348 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2349 for_each_netdev_rcu(&init_net, slave) \ 2350 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2351 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2352 2353 static inline struct net_device *next_net_device(struct net_device *dev) 2354 { 2355 struct list_head *lh; 2356 struct net *net; 2357 2358 net = dev_net(dev); 2359 lh = dev->dev_list.next; 2360 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2361 } 2362 2363 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2364 { 2365 struct list_head *lh; 2366 struct net *net; 2367 2368 net = dev_net(dev); 2369 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2370 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2371 } 2372 2373 static inline struct net_device *first_net_device(struct net *net) 2374 { 2375 return list_empty(&net->dev_base_head) ? NULL : 2376 net_device_entry(net->dev_base_head.next); 2377 } 2378 2379 static inline struct net_device *first_net_device_rcu(struct net *net) 2380 { 2381 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2382 2383 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2384 } 2385 2386 int netdev_boot_setup_check(struct net_device *dev); 2387 unsigned long netdev_boot_base(const char *prefix, int unit); 2388 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2389 const char *hwaddr); 2390 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2391 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2392 void dev_add_pack(struct packet_type *pt); 2393 void dev_remove_pack(struct packet_type *pt); 2394 void __dev_remove_pack(struct packet_type *pt); 2395 void dev_add_offload(struct packet_offload *po); 2396 void dev_remove_offload(struct packet_offload *po); 2397 2398 int dev_get_iflink(const struct net_device *dev); 2399 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2400 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2401 unsigned short mask); 2402 struct net_device *dev_get_by_name(struct net *net, const char *name); 2403 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2404 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2405 int dev_alloc_name(struct net_device *dev, const char *name); 2406 int dev_open(struct net_device *dev); 2407 int dev_close(struct net_device *dev); 2408 int dev_close_many(struct list_head *head, bool unlink); 2409 void dev_disable_lro(struct net_device *dev); 2410 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2411 int dev_queue_xmit(struct sk_buff *skb); 2412 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv); 2413 int register_netdevice(struct net_device *dev); 2414 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2415 void unregister_netdevice_many(struct list_head *head); 2416 static inline void unregister_netdevice(struct net_device *dev) 2417 { 2418 unregister_netdevice_queue(dev, NULL); 2419 } 2420 2421 int netdev_refcnt_read(const struct net_device *dev); 2422 void free_netdev(struct net_device *dev); 2423 void netdev_freemem(struct net_device *dev); 2424 void synchronize_net(void); 2425 int init_dummy_netdev(struct net_device *dev); 2426 2427 DECLARE_PER_CPU(int, xmit_recursion); 2428 #define XMIT_RECURSION_LIMIT 10 2429 2430 static inline int dev_recursion_level(void) 2431 { 2432 return this_cpu_read(xmit_recursion); 2433 } 2434 2435 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2436 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2437 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2438 int netdev_get_name(struct net *net, char *name, int ifindex); 2439 int dev_restart(struct net_device *dev); 2440 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb); 2441 2442 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2443 { 2444 return NAPI_GRO_CB(skb)->data_offset; 2445 } 2446 2447 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2448 { 2449 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2450 } 2451 2452 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2453 { 2454 NAPI_GRO_CB(skb)->data_offset += len; 2455 } 2456 2457 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2458 unsigned int offset) 2459 { 2460 return NAPI_GRO_CB(skb)->frag0 + offset; 2461 } 2462 2463 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2464 { 2465 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2466 } 2467 2468 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2469 unsigned int offset) 2470 { 2471 if (!pskb_may_pull(skb, hlen)) 2472 return NULL; 2473 2474 NAPI_GRO_CB(skb)->frag0 = NULL; 2475 NAPI_GRO_CB(skb)->frag0_len = 0; 2476 return skb->data + offset; 2477 } 2478 2479 static inline void *skb_gro_network_header(struct sk_buff *skb) 2480 { 2481 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2482 skb_network_offset(skb); 2483 } 2484 2485 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2486 const void *start, unsigned int len) 2487 { 2488 if (NAPI_GRO_CB(skb)->csum_valid) 2489 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2490 csum_partial(start, len, 0)); 2491 } 2492 2493 /* GRO checksum functions. These are logical equivalents of the normal 2494 * checksum functions (in skbuff.h) except that they operate on the GRO 2495 * offsets and fields in sk_buff. 2496 */ 2497 2498 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2499 2500 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2501 { 2502 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2503 } 2504 2505 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2506 bool zero_okay, 2507 __sum16 check) 2508 { 2509 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2510 skb_checksum_start_offset(skb) < 2511 skb_gro_offset(skb)) && 2512 !skb_at_gro_remcsum_start(skb) && 2513 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2514 (!zero_okay || check)); 2515 } 2516 2517 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2518 __wsum psum) 2519 { 2520 if (NAPI_GRO_CB(skb)->csum_valid && 2521 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2522 return 0; 2523 2524 NAPI_GRO_CB(skb)->csum = psum; 2525 2526 return __skb_gro_checksum_complete(skb); 2527 } 2528 2529 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2530 { 2531 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2532 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2533 NAPI_GRO_CB(skb)->csum_cnt--; 2534 } else { 2535 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2536 * verified a new top level checksum or an encapsulated one 2537 * during GRO. This saves work if we fallback to normal path. 2538 */ 2539 __skb_incr_checksum_unnecessary(skb); 2540 } 2541 } 2542 2543 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2544 compute_pseudo) \ 2545 ({ \ 2546 __sum16 __ret = 0; \ 2547 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2548 __ret = __skb_gro_checksum_validate_complete(skb, \ 2549 compute_pseudo(skb, proto)); \ 2550 if (__ret) \ 2551 __skb_mark_checksum_bad(skb); \ 2552 else \ 2553 skb_gro_incr_csum_unnecessary(skb); \ 2554 __ret; \ 2555 }) 2556 2557 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2558 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2559 2560 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2561 compute_pseudo) \ 2562 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2563 2564 #define skb_gro_checksum_simple_validate(skb) \ 2565 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2566 2567 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2568 { 2569 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2570 !NAPI_GRO_CB(skb)->csum_valid); 2571 } 2572 2573 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2574 __sum16 check, __wsum pseudo) 2575 { 2576 NAPI_GRO_CB(skb)->csum = ~pseudo; 2577 NAPI_GRO_CB(skb)->csum_valid = 1; 2578 } 2579 2580 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \ 2581 do { \ 2582 if (__skb_gro_checksum_convert_check(skb)) \ 2583 __skb_gro_checksum_convert(skb, check, \ 2584 compute_pseudo(skb, proto)); \ 2585 } while (0) 2586 2587 struct gro_remcsum { 2588 int offset; 2589 __wsum delta; 2590 }; 2591 2592 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2593 { 2594 grc->offset = 0; 2595 grc->delta = 0; 2596 } 2597 2598 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2599 unsigned int off, size_t hdrlen, 2600 int start, int offset, 2601 struct gro_remcsum *grc, 2602 bool nopartial) 2603 { 2604 __wsum delta; 2605 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2606 2607 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2608 2609 if (!nopartial) { 2610 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2611 return ptr; 2612 } 2613 2614 ptr = skb_gro_header_fast(skb, off); 2615 if (skb_gro_header_hard(skb, off + plen)) { 2616 ptr = skb_gro_header_slow(skb, off + plen, off); 2617 if (!ptr) 2618 return NULL; 2619 } 2620 2621 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2622 start, offset); 2623 2624 /* Adjust skb->csum since we changed the packet */ 2625 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2626 2627 grc->offset = off + hdrlen + offset; 2628 grc->delta = delta; 2629 2630 return ptr; 2631 } 2632 2633 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2634 struct gro_remcsum *grc) 2635 { 2636 void *ptr; 2637 size_t plen = grc->offset + sizeof(u16); 2638 2639 if (!grc->delta) 2640 return; 2641 2642 ptr = skb_gro_header_fast(skb, grc->offset); 2643 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 2644 ptr = skb_gro_header_slow(skb, plen, grc->offset); 2645 if (!ptr) 2646 return; 2647 } 2648 2649 remcsum_unadjust((__sum16 *)ptr, grc->delta); 2650 } 2651 2652 struct skb_csum_offl_spec { 2653 __u16 ipv4_okay:1, 2654 ipv6_okay:1, 2655 encap_okay:1, 2656 ip_options_okay:1, 2657 ext_hdrs_okay:1, 2658 tcp_okay:1, 2659 udp_okay:1, 2660 sctp_okay:1, 2661 vlan_okay:1, 2662 no_encapped_ipv6:1, 2663 no_not_encapped:1; 2664 }; 2665 2666 bool __skb_csum_offload_chk(struct sk_buff *skb, 2667 const struct skb_csum_offl_spec *spec, 2668 bool *csum_encapped, 2669 bool csum_help); 2670 2671 static inline bool skb_csum_offload_chk(struct sk_buff *skb, 2672 const struct skb_csum_offl_spec *spec, 2673 bool *csum_encapped, 2674 bool csum_help) 2675 { 2676 if (skb->ip_summed != CHECKSUM_PARTIAL) 2677 return false; 2678 2679 return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help); 2680 } 2681 2682 static inline bool skb_csum_offload_chk_help(struct sk_buff *skb, 2683 const struct skb_csum_offl_spec *spec) 2684 { 2685 bool csum_encapped; 2686 2687 return skb_csum_offload_chk(skb, spec, &csum_encapped, true); 2688 } 2689 2690 static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb) 2691 { 2692 static const struct skb_csum_offl_spec csum_offl_spec = { 2693 .ipv4_okay = 1, 2694 .ip_options_okay = 1, 2695 .ipv6_okay = 1, 2696 .vlan_okay = 1, 2697 .tcp_okay = 1, 2698 .udp_okay = 1, 2699 }; 2700 2701 return skb_csum_offload_chk_help(skb, &csum_offl_spec); 2702 } 2703 2704 static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb) 2705 { 2706 static const struct skb_csum_offl_spec csum_offl_spec = { 2707 .ipv4_okay = 1, 2708 .ip_options_okay = 1, 2709 .tcp_okay = 1, 2710 .udp_okay = 1, 2711 .vlan_okay = 1, 2712 }; 2713 2714 return skb_csum_offload_chk_help(skb, &csum_offl_spec); 2715 } 2716 2717 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2718 unsigned short type, 2719 const void *daddr, const void *saddr, 2720 unsigned int len) 2721 { 2722 if (!dev->header_ops || !dev->header_ops->create) 2723 return 0; 2724 2725 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2726 } 2727 2728 static inline int dev_parse_header(const struct sk_buff *skb, 2729 unsigned char *haddr) 2730 { 2731 const struct net_device *dev = skb->dev; 2732 2733 if (!dev->header_ops || !dev->header_ops->parse) 2734 return 0; 2735 return dev->header_ops->parse(skb, haddr); 2736 } 2737 2738 /* ll_header must have at least hard_header_len allocated */ 2739 static inline bool dev_validate_header(const struct net_device *dev, 2740 char *ll_header, int len) 2741 { 2742 if (likely(len >= dev->hard_header_len)) 2743 return true; 2744 2745 if (capable(CAP_SYS_RAWIO)) { 2746 memset(ll_header + len, 0, dev->hard_header_len - len); 2747 return true; 2748 } 2749 2750 if (dev->header_ops && dev->header_ops->validate) 2751 return dev->header_ops->validate(ll_header, len); 2752 2753 return false; 2754 } 2755 2756 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len); 2757 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2758 static inline int unregister_gifconf(unsigned int family) 2759 { 2760 return register_gifconf(family, NULL); 2761 } 2762 2763 #ifdef CONFIG_NET_FLOW_LIMIT 2764 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2765 struct sd_flow_limit { 2766 u64 count; 2767 unsigned int num_buckets; 2768 unsigned int history_head; 2769 u16 history[FLOW_LIMIT_HISTORY]; 2770 u8 buckets[]; 2771 }; 2772 2773 extern int netdev_flow_limit_table_len; 2774 #endif /* CONFIG_NET_FLOW_LIMIT */ 2775 2776 /* 2777 * Incoming packets are placed on per-CPU queues 2778 */ 2779 struct softnet_data { 2780 struct list_head poll_list; 2781 struct sk_buff_head process_queue; 2782 2783 /* stats */ 2784 unsigned int processed; 2785 unsigned int time_squeeze; 2786 unsigned int received_rps; 2787 #ifdef CONFIG_RPS 2788 struct softnet_data *rps_ipi_list; 2789 #endif 2790 #ifdef CONFIG_NET_FLOW_LIMIT 2791 struct sd_flow_limit __rcu *flow_limit; 2792 #endif 2793 struct Qdisc *output_queue; 2794 struct Qdisc **output_queue_tailp; 2795 struct sk_buff *completion_queue; 2796 2797 #ifdef CONFIG_RPS 2798 /* input_queue_head should be written by cpu owning this struct, 2799 * and only read by other cpus. Worth using a cache line. 2800 */ 2801 unsigned int input_queue_head ____cacheline_aligned_in_smp; 2802 2803 /* Elements below can be accessed between CPUs for RPS/RFS */ 2804 struct call_single_data csd ____cacheline_aligned_in_smp; 2805 struct softnet_data *rps_ipi_next; 2806 unsigned int cpu; 2807 unsigned int input_queue_tail; 2808 #endif 2809 unsigned int dropped; 2810 struct sk_buff_head input_pkt_queue; 2811 struct napi_struct backlog; 2812 2813 }; 2814 2815 static inline void input_queue_head_incr(struct softnet_data *sd) 2816 { 2817 #ifdef CONFIG_RPS 2818 sd->input_queue_head++; 2819 #endif 2820 } 2821 2822 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 2823 unsigned int *qtail) 2824 { 2825 #ifdef CONFIG_RPS 2826 *qtail = ++sd->input_queue_tail; 2827 #endif 2828 } 2829 2830 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 2831 2832 void __netif_schedule(struct Qdisc *q); 2833 void netif_schedule_queue(struct netdev_queue *txq); 2834 2835 static inline void netif_tx_schedule_all(struct net_device *dev) 2836 { 2837 unsigned int i; 2838 2839 for (i = 0; i < dev->num_tx_queues; i++) 2840 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 2841 } 2842 2843 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 2844 { 2845 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2846 } 2847 2848 /** 2849 * netif_start_queue - allow transmit 2850 * @dev: network device 2851 * 2852 * Allow upper layers to call the device hard_start_xmit routine. 2853 */ 2854 static inline void netif_start_queue(struct net_device *dev) 2855 { 2856 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 2857 } 2858 2859 static inline void netif_tx_start_all_queues(struct net_device *dev) 2860 { 2861 unsigned int i; 2862 2863 for (i = 0; i < dev->num_tx_queues; i++) { 2864 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2865 netif_tx_start_queue(txq); 2866 } 2867 } 2868 2869 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 2870 2871 /** 2872 * netif_wake_queue - restart transmit 2873 * @dev: network device 2874 * 2875 * Allow upper layers to call the device hard_start_xmit routine. 2876 * Used for flow control when transmit resources are available. 2877 */ 2878 static inline void netif_wake_queue(struct net_device *dev) 2879 { 2880 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 2881 } 2882 2883 static inline void netif_tx_wake_all_queues(struct net_device *dev) 2884 { 2885 unsigned int i; 2886 2887 for (i = 0; i < dev->num_tx_queues; i++) { 2888 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2889 netif_tx_wake_queue(txq); 2890 } 2891 } 2892 2893 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 2894 { 2895 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2896 } 2897 2898 /** 2899 * netif_stop_queue - stop transmitted packets 2900 * @dev: network device 2901 * 2902 * Stop upper layers calling the device hard_start_xmit routine. 2903 * Used for flow control when transmit resources are unavailable. 2904 */ 2905 static inline void netif_stop_queue(struct net_device *dev) 2906 { 2907 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 2908 } 2909 2910 void netif_tx_stop_all_queues(struct net_device *dev); 2911 2912 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 2913 { 2914 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2915 } 2916 2917 /** 2918 * netif_queue_stopped - test if transmit queue is flowblocked 2919 * @dev: network device 2920 * 2921 * Test if transmit queue on device is currently unable to send. 2922 */ 2923 static inline bool netif_queue_stopped(const struct net_device *dev) 2924 { 2925 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 2926 } 2927 2928 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 2929 { 2930 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 2931 } 2932 2933 static inline bool 2934 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 2935 { 2936 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 2937 } 2938 2939 static inline bool 2940 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 2941 { 2942 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 2943 } 2944 2945 /** 2946 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 2947 * @dev_queue: pointer to transmit queue 2948 * 2949 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 2950 * to give appropriate hint to the CPU. 2951 */ 2952 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 2953 { 2954 #ifdef CONFIG_BQL 2955 prefetchw(&dev_queue->dql.num_queued); 2956 #endif 2957 } 2958 2959 /** 2960 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 2961 * @dev_queue: pointer to transmit queue 2962 * 2963 * BQL enabled drivers might use this helper in their TX completion path, 2964 * to give appropriate hint to the CPU. 2965 */ 2966 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 2967 { 2968 #ifdef CONFIG_BQL 2969 prefetchw(&dev_queue->dql.limit); 2970 #endif 2971 } 2972 2973 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 2974 unsigned int bytes) 2975 { 2976 #ifdef CONFIG_BQL 2977 dql_queued(&dev_queue->dql, bytes); 2978 2979 if (likely(dql_avail(&dev_queue->dql) >= 0)) 2980 return; 2981 2982 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2983 2984 /* 2985 * The XOFF flag must be set before checking the dql_avail below, 2986 * because in netdev_tx_completed_queue we update the dql_completed 2987 * before checking the XOFF flag. 2988 */ 2989 smp_mb(); 2990 2991 /* check again in case another CPU has just made room avail */ 2992 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 2993 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2994 #endif 2995 } 2996 2997 /** 2998 * netdev_sent_queue - report the number of bytes queued to hardware 2999 * @dev: network device 3000 * @bytes: number of bytes queued to the hardware device queue 3001 * 3002 * Report the number of bytes queued for sending/completion to the network 3003 * device hardware queue. @bytes should be a good approximation and should 3004 * exactly match netdev_completed_queue() @bytes 3005 */ 3006 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3007 { 3008 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3009 } 3010 3011 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3012 unsigned int pkts, unsigned int bytes) 3013 { 3014 #ifdef CONFIG_BQL 3015 if (unlikely(!bytes)) 3016 return; 3017 3018 dql_completed(&dev_queue->dql, bytes); 3019 3020 /* 3021 * Without the memory barrier there is a small possiblity that 3022 * netdev_tx_sent_queue will miss the update and cause the queue to 3023 * be stopped forever 3024 */ 3025 smp_mb(); 3026 3027 if (dql_avail(&dev_queue->dql) < 0) 3028 return; 3029 3030 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3031 netif_schedule_queue(dev_queue); 3032 #endif 3033 } 3034 3035 /** 3036 * netdev_completed_queue - report bytes and packets completed by device 3037 * @dev: network device 3038 * @pkts: actual number of packets sent over the medium 3039 * @bytes: actual number of bytes sent over the medium 3040 * 3041 * Report the number of bytes and packets transmitted by the network device 3042 * hardware queue over the physical medium, @bytes must exactly match the 3043 * @bytes amount passed to netdev_sent_queue() 3044 */ 3045 static inline void netdev_completed_queue(struct net_device *dev, 3046 unsigned int pkts, unsigned int bytes) 3047 { 3048 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3049 } 3050 3051 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3052 { 3053 #ifdef CONFIG_BQL 3054 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3055 dql_reset(&q->dql); 3056 #endif 3057 } 3058 3059 /** 3060 * netdev_reset_queue - reset the packets and bytes count of a network device 3061 * @dev_queue: network device 3062 * 3063 * Reset the bytes and packet count of a network device and clear the 3064 * software flow control OFF bit for this network device 3065 */ 3066 static inline void netdev_reset_queue(struct net_device *dev_queue) 3067 { 3068 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3069 } 3070 3071 /** 3072 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3073 * @dev: network device 3074 * @queue_index: given tx queue index 3075 * 3076 * Returns 0 if given tx queue index >= number of device tx queues, 3077 * otherwise returns the originally passed tx queue index. 3078 */ 3079 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3080 { 3081 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3082 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3083 dev->name, queue_index, 3084 dev->real_num_tx_queues); 3085 return 0; 3086 } 3087 3088 return queue_index; 3089 } 3090 3091 /** 3092 * netif_running - test if up 3093 * @dev: network device 3094 * 3095 * Test if the device has been brought up. 3096 */ 3097 static inline bool netif_running(const struct net_device *dev) 3098 { 3099 return test_bit(__LINK_STATE_START, &dev->state); 3100 } 3101 3102 /* 3103 * Routines to manage the subqueues on a device. We only need start, 3104 * stop, and a check if it's stopped. All other device management is 3105 * done at the overall netdevice level. 3106 * Also test the device if we're multiqueue. 3107 */ 3108 3109 /** 3110 * netif_start_subqueue - allow sending packets on subqueue 3111 * @dev: network device 3112 * @queue_index: sub queue index 3113 * 3114 * Start individual transmit queue of a device with multiple transmit queues. 3115 */ 3116 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3117 { 3118 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3119 3120 netif_tx_start_queue(txq); 3121 } 3122 3123 /** 3124 * netif_stop_subqueue - stop sending packets on subqueue 3125 * @dev: network device 3126 * @queue_index: sub queue index 3127 * 3128 * Stop individual transmit queue of a device with multiple transmit queues. 3129 */ 3130 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3131 { 3132 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3133 netif_tx_stop_queue(txq); 3134 } 3135 3136 /** 3137 * netif_subqueue_stopped - test status of subqueue 3138 * @dev: network device 3139 * @queue_index: sub queue index 3140 * 3141 * Check individual transmit queue of a device with multiple transmit queues. 3142 */ 3143 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3144 u16 queue_index) 3145 { 3146 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3147 3148 return netif_tx_queue_stopped(txq); 3149 } 3150 3151 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3152 struct sk_buff *skb) 3153 { 3154 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3155 } 3156 3157 void netif_wake_subqueue(struct net_device *dev, u16 queue_index); 3158 3159 #ifdef CONFIG_XPS 3160 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3161 u16 index); 3162 #else 3163 static inline int netif_set_xps_queue(struct net_device *dev, 3164 const struct cpumask *mask, 3165 u16 index) 3166 { 3167 return 0; 3168 } 3169 #endif 3170 3171 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb, 3172 unsigned int num_tx_queues); 3173 3174 /* 3175 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used 3176 * as a distribution range limit for the returned value. 3177 */ 3178 static inline u16 skb_tx_hash(const struct net_device *dev, 3179 struct sk_buff *skb) 3180 { 3181 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues); 3182 } 3183 3184 /** 3185 * netif_is_multiqueue - test if device has multiple transmit queues 3186 * @dev: network device 3187 * 3188 * Check if device has multiple transmit queues 3189 */ 3190 static inline bool netif_is_multiqueue(const struct net_device *dev) 3191 { 3192 return dev->num_tx_queues > 1; 3193 } 3194 3195 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3196 3197 #ifdef CONFIG_SYSFS 3198 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3199 #else 3200 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3201 unsigned int rxq) 3202 { 3203 return 0; 3204 } 3205 #endif 3206 3207 #ifdef CONFIG_SYSFS 3208 static inline unsigned int get_netdev_rx_queue_index( 3209 struct netdev_rx_queue *queue) 3210 { 3211 struct net_device *dev = queue->dev; 3212 int index = queue - dev->_rx; 3213 3214 BUG_ON(index >= dev->num_rx_queues); 3215 return index; 3216 } 3217 #endif 3218 3219 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3220 int netif_get_num_default_rss_queues(void); 3221 3222 enum skb_free_reason { 3223 SKB_REASON_CONSUMED, 3224 SKB_REASON_DROPPED, 3225 }; 3226 3227 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3228 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3229 3230 /* 3231 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3232 * interrupt context or with hardware interrupts being disabled. 3233 * (in_irq() || irqs_disabled()) 3234 * 3235 * We provide four helpers that can be used in following contexts : 3236 * 3237 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3238 * replacing kfree_skb(skb) 3239 * 3240 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3241 * Typically used in place of consume_skb(skb) in TX completion path 3242 * 3243 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3244 * replacing kfree_skb(skb) 3245 * 3246 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3247 * and consumed a packet. Used in place of consume_skb(skb) 3248 */ 3249 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3250 { 3251 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3252 } 3253 3254 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3255 { 3256 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3257 } 3258 3259 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3260 { 3261 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3262 } 3263 3264 static inline void dev_consume_skb_any(struct sk_buff *skb) 3265 { 3266 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3267 } 3268 3269 int netif_rx(struct sk_buff *skb); 3270 int netif_rx_ni(struct sk_buff *skb); 3271 int netif_receive_skb(struct sk_buff *skb); 3272 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3273 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3274 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3275 gro_result_t napi_gro_frags(struct napi_struct *napi); 3276 struct packet_offload *gro_find_receive_by_type(__be16 type); 3277 struct packet_offload *gro_find_complete_by_type(__be16 type); 3278 3279 static inline void napi_free_frags(struct napi_struct *napi) 3280 { 3281 kfree_skb(napi->skb); 3282 napi->skb = NULL; 3283 } 3284 3285 bool netdev_is_rx_handler_busy(struct net_device *dev); 3286 int netdev_rx_handler_register(struct net_device *dev, 3287 rx_handler_func_t *rx_handler, 3288 void *rx_handler_data); 3289 void netdev_rx_handler_unregister(struct net_device *dev); 3290 3291 bool dev_valid_name(const char *name); 3292 int dev_ioctl(struct net *net, unsigned int cmd, void __user *); 3293 int dev_ethtool(struct net *net, struct ifreq *); 3294 unsigned int dev_get_flags(const struct net_device *); 3295 int __dev_change_flags(struct net_device *, unsigned int flags); 3296 int dev_change_flags(struct net_device *, unsigned int); 3297 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3298 unsigned int gchanges); 3299 int dev_change_name(struct net_device *, const char *); 3300 int dev_set_alias(struct net_device *, const char *, size_t); 3301 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3302 int dev_set_mtu(struct net_device *, int); 3303 void dev_set_group(struct net_device *, int); 3304 int dev_set_mac_address(struct net_device *, struct sockaddr *); 3305 int dev_change_carrier(struct net_device *, bool new_carrier); 3306 int dev_get_phys_port_id(struct net_device *dev, 3307 struct netdev_phys_item_id *ppid); 3308 int dev_get_phys_port_name(struct net_device *dev, 3309 char *name, size_t len); 3310 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3311 int dev_change_xdp_fd(struct net_device *dev, int fd); 3312 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev); 3313 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3314 struct netdev_queue *txq, int *ret); 3315 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3316 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3317 bool is_skb_forwardable(const struct net_device *dev, 3318 const struct sk_buff *skb); 3319 3320 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3321 3322 extern int netdev_budget; 3323 3324 /* Called by rtnetlink.c:rtnl_unlock() */ 3325 void netdev_run_todo(void); 3326 3327 /** 3328 * dev_put - release reference to device 3329 * @dev: network device 3330 * 3331 * Release reference to device to allow it to be freed. 3332 */ 3333 static inline void dev_put(struct net_device *dev) 3334 { 3335 this_cpu_dec(*dev->pcpu_refcnt); 3336 } 3337 3338 /** 3339 * dev_hold - get reference to device 3340 * @dev: network device 3341 * 3342 * Hold reference to device to keep it from being freed. 3343 */ 3344 static inline void dev_hold(struct net_device *dev) 3345 { 3346 this_cpu_inc(*dev->pcpu_refcnt); 3347 } 3348 3349 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3350 * and _off may be called from IRQ context, but it is caller 3351 * who is responsible for serialization of these calls. 3352 * 3353 * The name carrier is inappropriate, these functions should really be 3354 * called netif_lowerlayer_*() because they represent the state of any 3355 * kind of lower layer not just hardware media. 3356 */ 3357 3358 void linkwatch_init_dev(struct net_device *dev); 3359 void linkwatch_fire_event(struct net_device *dev); 3360 void linkwatch_forget_dev(struct net_device *dev); 3361 3362 /** 3363 * netif_carrier_ok - test if carrier present 3364 * @dev: network device 3365 * 3366 * Check if carrier is present on device 3367 */ 3368 static inline bool netif_carrier_ok(const struct net_device *dev) 3369 { 3370 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3371 } 3372 3373 unsigned long dev_trans_start(struct net_device *dev); 3374 3375 void __netdev_watchdog_up(struct net_device *dev); 3376 3377 void netif_carrier_on(struct net_device *dev); 3378 3379 void netif_carrier_off(struct net_device *dev); 3380 3381 /** 3382 * netif_dormant_on - mark device as dormant. 3383 * @dev: network device 3384 * 3385 * Mark device as dormant (as per RFC2863). 3386 * 3387 * The dormant state indicates that the relevant interface is not 3388 * actually in a condition to pass packets (i.e., it is not 'up') but is 3389 * in a "pending" state, waiting for some external event. For "on- 3390 * demand" interfaces, this new state identifies the situation where the 3391 * interface is waiting for events to place it in the up state. 3392 */ 3393 static inline void netif_dormant_on(struct net_device *dev) 3394 { 3395 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3396 linkwatch_fire_event(dev); 3397 } 3398 3399 /** 3400 * netif_dormant_off - set device as not dormant. 3401 * @dev: network device 3402 * 3403 * Device is not in dormant state. 3404 */ 3405 static inline void netif_dormant_off(struct net_device *dev) 3406 { 3407 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3408 linkwatch_fire_event(dev); 3409 } 3410 3411 /** 3412 * netif_dormant - test if carrier present 3413 * @dev: network device 3414 * 3415 * Check if carrier is present on device 3416 */ 3417 static inline bool netif_dormant(const struct net_device *dev) 3418 { 3419 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3420 } 3421 3422 3423 /** 3424 * netif_oper_up - test if device is operational 3425 * @dev: network device 3426 * 3427 * Check if carrier is operational 3428 */ 3429 static inline bool netif_oper_up(const struct net_device *dev) 3430 { 3431 return (dev->operstate == IF_OPER_UP || 3432 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3433 } 3434 3435 /** 3436 * netif_device_present - is device available or removed 3437 * @dev: network device 3438 * 3439 * Check if device has not been removed from system. 3440 */ 3441 static inline bool netif_device_present(struct net_device *dev) 3442 { 3443 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3444 } 3445 3446 void netif_device_detach(struct net_device *dev); 3447 3448 void netif_device_attach(struct net_device *dev); 3449 3450 /* 3451 * Network interface message level settings 3452 */ 3453 3454 enum { 3455 NETIF_MSG_DRV = 0x0001, 3456 NETIF_MSG_PROBE = 0x0002, 3457 NETIF_MSG_LINK = 0x0004, 3458 NETIF_MSG_TIMER = 0x0008, 3459 NETIF_MSG_IFDOWN = 0x0010, 3460 NETIF_MSG_IFUP = 0x0020, 3461 NETIF_MSG_RX_ERR = 0x0040, 3462 NETIF_MSG_TX_ERR = 0x0080, 3463 NETIF_MSG_TX_QUEUED = 0x0100, 3464 NETIF_MSG_INTR = 0x0200, 3465 NETIF_MSG_TX_DONE = 0x0400, 3466 NETIF_MSG_RX_STATUS = 0x0800, 3467 NETIF_MSG_PKTDATA = 0x1000, 3468 NETIF_MSG_HW = 0x2000, 3469 NETIF_MSG_WOL = 0x4000, 3470 }; 3471 3472 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3473 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3474 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3475 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3476 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3477 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3478 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3479 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3480 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3481 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3482 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3483 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3484 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3485 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3486 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3487 3488 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3489 { 3490 /* use default */ 3491 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3492 return default_msg_enable_bits; 3493 if (debug_value == 0) /* no output */ 3494 return 0; 3495 /* set low N bits */ 3496 return (1 << debug_value) - 1; 3497 } 3498 3499 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 3500 { 3501 spin_lock(&txq->_xmit_lock); 3502 txq->xmit_lock_owner = cpu; 3503 } 3504 3505 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 3506 { 3507 spin_lock_bh(&txq->_xmit_lock); 3508 txq->xmit_lock_owner = smp_processor_id(); 3509 } 3510 3511 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 3512 { 3513 bool ok = spin_trylock(&txq->_xmit_lock); 3514 if (likely(ok)) 3515 txq->xmit_lock_owner = smp_processor_id(); 3516 return ok; 3517 } 3518 3519 static inline void __netif_tx_unlock(struct netdev_queue *txq) 3520 { 3521 txq->xmit_lock_owner = -1; 3522 spin_unlock(&txq->_xmit_lock); 3523 } 3524 3525 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 3526 { 3527 txq->xmit_lock_owner = -1; 3528 spin_unlock_bh(&txq->_xmit_lock); 3529 } 3530 3531 static inline void txq_trans_update(struct netdev_queue *txq) 3532 { 3533 if (txq->xmit_lock_owner != -1) 3534 txq->trans_start = jiffies; 3535 } 3536 3537 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 3538 static inline void netif_trans_update(struct net_device *dev) 3539 { 3540 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 3541 3542 if (txq->trans_start != jiffies) 3543 txq->trans_start = jiffies; 3544 } 3545 3546 /** 3547 * netif_tx_lock - grab network device transmit lock 3548 * @dev: network device 3549 * 3550 * Get network device transmit lock 3551 */ 3552 static inline void netif_tx_lock(struct net_device *dev) 3553 { 3554 unsigned int i; 3555 int cpu; 3556 3557 spin_lock(&dev->tx_global_lock); 3558 cpu = smp_processor_id(); 3559 for (i = 0; i < dev->num_tx_queues; i++) { 3560 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3561 3562 /* We are the only thread of execution doing a 3563 * freeze, but we have to grab the _xmit_lock in 3564 * order to synchronize with threads which are in 3565 * the ->hard_start_xmit() handler and already 3566 * checked the frozen bit. 3567 */ 3568 __netif_tx_lock(txq, cpu); 3569 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 3570 __netif_tx_unlock(txq); 3571 } 3572 } 3573 3574 static inline void netif_tx_lock_bh(struct net_device *dev) 3575 { 3576 local_bh_disable(); 3577 netif_tx_lock(dev); 3578 } 3579 3580 static inline void netif_tx_unlock(struct net_device *dev) 3581 { 3582 unsigned int i; 3583 3584 for (i = 0; i < dev->num_tx_queues; i++) { 3585 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3586 3587 /* No need to grab the _xmit_lock here. If the 3588 * queue is not stopped for another reason, we 3589 * force a schedule. 3590 */ 3591 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 3592 netif_schedule_queue(txq); 3593 } 3594 spin_unlock(&dev->tx_global_lock); 3595 } 3596 3597 static inline void netif_tx_unlock_bh(struct net_device *dev) 3598 { 3599 netif_tx_unlock(dev); 3600 local_bh_enable(); 3601 } 3602 3603 #define HARD_TX_LOCK(dev, txq, cpu) { \ 3604 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3605 __netif_tx_lock(txq, cpu); \ 3606 } \ 3607 } 3608 3609 #define HARD_TX_TRYLOCK(dev, txq) \ 3610 (((dev->features & NETIF_F_LLTX) == 0) ? \ 3611 __netif_tx_trylock(txq) : \ 3612 true ) 3613 3614 #define HARD_TX_UNLOCK(dev, txq) { \ 3615 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3616 __netif_tx_unlock(txq); \ 3617 } \ 3618 } 3619 3620 static inline void netif_tx_disable(struct net_device *dev) 3621 { 3622 unsigned int i; 3623 int cpu; 3624 3625 local_bh_disable(); 3626 cpu = smp_processor_id(); 3627 for (i = 0; i < dev->num_tx_queues; i++) { 3628 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3629 3630 __netif_tx_lock(txq, cpu); 3631 netif_tx_stop_queue(txq); 3632 __netif_tx_unlock(txq); 3633 } 3634 local_bh_enable(); 3635 } 3636 3637 static inline void netif_addr_lock(struct net_device *dev) 3638 { 3639 spin_lock(&dev->addr_list_lock); 3640 } 3641 3642 static inline void netif_addr_lock_nested(struct net_device *dev) 3643 { 3644 int subclass = SINGLE_DEPTH_NESTING; 3645 3646 if (dev->netdev_ops->ndo_get_lock_subclass) 3647 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev); 3648 3649 spin_lock_nested(&dev->addr_list_lock, subclass); 3650 } 3651 3652 static inline void netif_addr_lock_bh(struct net_device *dev) 3653 { 3654 spin_lock_bh(&dev->addr_list_lock); 3655 } 3656 3657 static inline void netif_addr_unlock(struct net_device *dev) 3658 { 3659 spin_unlock(&dev->addr_list_lock); 3660 } 3661 3662 static inline void netif_addr_unlock_bh(struct net_device *dev) 3663 { 3664 spin_unlock_bh(&dev->addr_list_lock); 3665 } 3666 3667 /* 3668 * dev_addrs walker. Should be used only for read access. Call with 3669 * rcu_read_lock held. 3670 */ 3671 #define for_each_dev_addr(dev, ha) \ 3672 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 3673 3674 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 3675 3676 void ether_setup(struct net_device *dev); 3677 3678 /* Support for loadable net-drivers */ 3679 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 3680 unsigned char name_assign_type, 3681 void (*setup)(struct net_device *), 3682 unsigned int txqs, unsigned int rxqs); 3683 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 3684 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 3685 3686 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 3687 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 3688 count) 3689 3690 int register_netdev(struct net_device *dev); 3691 void unregister_netdev(struct net_device *dev); 3692 3693 /* General hardware address lists handling functions */ 3694 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 3695 struct netdev_hw_addr_list *from_list, int addr_len); 3696 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 3697 struct netdev_hw_addr_list *from_list, int addr_len); 3698 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 3699 struct net_device *dev, 3700 int (*sync)(struct net_device *, const unsigned char *), 3701 int (*unsync)(struct net_device *, 3702 const unsigned char *)); 3703 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 3704 struct net_device *dev, 3705 int (*unsync)(struct net_device *, 3706 const unsigned char *)); 3707 void __hw_addr_init(struct netdev_hw_addr_list *list); 3708 3709 /* Functions used for device addresses handling */ 3710 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 3711 unsigned char addr_type); 3712 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 3713 unsigned char addr_type); 3714 void dev_addr_flush(struct net_device *dev); 3715 int dev_addr_init(struct net_device *dev); 3716 3717 /* Functions used for unicast addresses handling */ 3718 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 3719 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 3720 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 3721 int dev_uc_sync(struct net_device *to, struct net_device *from); 3722 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 3723 void dev_uc_unsync(struct net_device *to, struct net_device *from); 3724 void dev_uc_flush(struct net_device *dev); 3725 void dev_uc_init(struct net_device *dev); 3726 3727 /** 3728 * __dev_uc_sync - Synchonize device's unicast list 3729 * @dev: device to sync 3730 * @sync: function to call if address should be added 3731 * @unsync: function to call if address should be removed 3732 * 3733 * Add newly added addresses to the interface, and release 3734 * addresses that have been deleted. 3735 */ 3736 static inline int __dev_uc_sync(struct net_device *dev, 3737 int (*sync)(struct net_device *, 3738 const unsigned char *), 3739 int (*unsync)(struct net_device *, 3740 const unsigned char *)) 3741 { 3742 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 3743 } 3744 3745 /** 3746 * __dev_uc_unsync - Remove synchronized addresses from device 3747 * @dev: device to sync 3748 * @unsync: function to call if address should be removed 3749 * 3750 * Remove all addresses that were added to the device by dev_uc_sync(). 3751 */ 3752 static inline void __dev_uc_unsync(struct net_device *dev, 3753 int (*unsync)(struct net_device *, 3754 const unsigned char *)) 3755 { 3756 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 3757 } 3758 3759 /* Functions used for multicast addresses handling */ 3760 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 3761 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 3762 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 3763 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 3764 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 3765 int dev_mc_sync(struct net_device *to, struct net_device *from); 3766 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 3767 void dev_mc_unsync(struct net_device *to, struct net_device *from); 3768 void dev_mc_flush(struct net_device *dev); 3769 void dev_mc_init(struct net_device *dev); 3770 3771 /** 3772 * __dev_mc_sync - Synchonize device's multicast list 3773 * @dev: device to sync 3774 * @sync: function to call if address should be added 3775 * @unsync: function to call if address should be removed 3776 * 3777 * Add newly added addresses to the interface, and release 3778 * addresses that have been deleted. 3779 */ 3780 static inline int __dev_mc_sync(struct net_device *dev, 3781 int (*sync)(struct net_device *, 3782 const unsigned char *), 3783 int (*unsync)(struct net_device *, 3784 const unsigned char *)) 3785 { 3786 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 3787 } 3788 3789 /** 3790 * __dev_mc_unsync - Remove synchronized addresses from device 3791 * @dev: device to sync 3792 * @unsync: function to call if address should be removed 3793 * 3794 * Remove all addresses that were added to the device by dev_mc_sync(). 3795 */ 3796 static inline void __dev_mc_unsync(struct net_device *dev, 3797 int (*unsync)(struct net_device *, 3798 const unsigned char *)) 3799 { 3800 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 3801 } 3802 3803 /* Functions used for secondary unicast and multicast support */ 3804 void dev_set_rx_mode(struct net_device *dev); 3805 void __dev_set_rx_mode(struct net_device *dev); 3806 int dev_set_promiscuity(struct net_device *dev, int inc); 3807 int dev_set_allmulti(struct net_device *dev, int inc); 3808 void netdev_state_change(struct net_device *dev); 3809 void netdev_notify_peers(struct net_device *dev); 3810 void netdev_features_change(struct net_device *dev); 3811 /* Load a device via the kmod */ 3812 void dev_load(struct net *net, const char *name); 3813 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 3814 struct rtnl_link_stats64 *storage); 3815 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 3816 const struct net_device_stats *netdev_stats); 3817 3818 extern int netdev_max_backlog; 3819 extern int netdev_tstamp_prequeue; 3820 extern int weight_p; 3821 3822 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 3823 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 3824 struct list_head **iter); 3825 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 3826 struct list_head **iter); 3827 3828 /* iterate through upper list, must be called under RCU read lock */ 3829 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 3830 for (iter = &(dev)->adj_list.upper, \ 3831 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 3832 updev; \ 3833 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 3834 3835 /* iterate through upper list, must be called under RCU read lock */ 3836 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \ 3837 for (iter = &(dev)->all_adj_list.upper, \ 3838 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \ 3839 updev; \ 3840 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter))) 3841 3842 void *netdev_lower_get_next_private(struct net_device *dev, 3843 struct list_head **iter); 3844 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 3845 struct list_head **iter); 3846 3847 #define netdev_for_each_lower_private(dev, priv, iter) \ 3848 for (iter = (dev)->adj_list.lower.next, \ 3849 priv = netdev_lower_get_next_private(dev, &(iter)); \ 3850 priv; \ 3851 priv = netdev_lower_get_next_private(dev, &(iter))) 3852 3853 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 3854 for (iter = &(dev)->adj_list.lower, \ 3855 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 3856 priv; \ 3857 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 3858 3859 void *netdev_lower_get_next(struct net_device *dev, 3860 struct list_head **iter); 3861 3862 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 3863 for (iter = (dev)->adj_list.lower.next, \ 3864 ldev = netdev_lower_get_next(dev, &(iter)); \ 3865 ldev; \ 3866 ldev = netdev_lower_get_next(dev, &(iter))) 3867 3868 struct net_device *netdev_all_lower_get_next(struct net_device *dev, 3869 struct list_head **iter); 3870 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev, 3871 struct list_head **iter); 3872 3873 #define netdev_for_each_all_lower_dev(dev, ldev, iter) \ 3874 for (iter = (dev)->all_adj_list.lower.next, \ 3875 ldev = netdev_all_lower_get_next(dev, &(iter)); \ 3876 ldev; \ 3877 ldev = netdev_all_lower_get_next(dev, &(iter))) 3878 3879 #define netdev_for_each_all_lower_dev_rcu(dev, ldev, iter) \ 3880 for (iter = (dev)->all_adj_list.lower.next, \ 3881 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)); \ 3882 ldev; \ 3883 ldev = netdev_all_lower_get_next_rcu(dev, &(iter))) 3884 3885 void *netdev_adjacent_get_private(struct list_head *adj_list); 3886 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 3887 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 3888 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 3889 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev); 3890 int netdev_master_upper_dev_link(struct net_device *dev, 3891 struct net_device *upper_dev, 3892 void *upper_priv, void *upper_info); 3893 void netdev_upper_dev_unlink(struct net_device *dev, 3894 struct net_device *upper_dev); 3895 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 3896 void *netdev_lower_dev_get_private(struct net_device *dev, 3897 struct net_device *lower_dev); 3898 void netdev_lower_state_changed(struct net_device *lower_dev, 3899 void *lower_state_info); 3900 int netdev_default_l2upper_neigh_construct(struct net_device *dev, 3901 struct neighbour *n); 3902 void netdev_default_l2upper_neigh_destroy(struct net_device *dev, 3903 struct neighbour *n); 3904 3905 /* RSS keys are 40 or 52 bytes long */ 3906 #define NETDEV_RSS_KEY_LEN 52 3907 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 3908 void netdev_rss_key_fill(void *buffer, size_t len); 3909 3910 int dev_get_nest_level(struct net_device *dev); 3911 int skb_checksum_help(struct sk_buff *skb); 3912 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 3913 netdev_features_t features, bool tx_path); 3914 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 3915 netdev_features_t features); 3916 3917 struct netdev_bonding_info { 3918 ifslave slave; 3919 ifbond master; 3920 }; 3921 3922 struct netdev_notifier_bonding_info { 3923 struct netdev_notifier_info info; /* must be first */ 3924 struct netdev_bonding_info bonding_info; 3925 }; 3926 3927 void netdev_bonding_info_change(struct net_device *dev, 3928 struct netdev_bonding_info *bonding_info); 3929 3930 static inline 3931 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 3932 { 3933 return __skb_gso_segment(skb, features, true); 3934 } 3935 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 3936 3937 static inline bool can_checksum_protocol(netdev_features_t features, 3938 __be16 protocol) 3939 { 3940 if (protocol == htons(ETH_P_FCOE)) 3941 return !!(features & NETIF_F_FCOE_CRC); 3942 3943 /* Assume this is an IP checksum (not SCTP CRC) */ 3944 3945 if (features & NETIF_F_HW_CSUM) { 3946 /* Can checksum everything */ 3947 return true; 3948 } 3949 3950 switch (protocol) { 3951 case htons(ETH_P_IP): 3952 return !!(features & NETIF_F_IP_CSUM); 3953 case htons(ETH_P_IPV6): 3954 return !!(features & NETIF_F_IPV6_CSUM); 3955 default: 3956 return false; 3957 } 3958 } 3959 3960 /* Map an ethertype into IP protocol if possible */ 3961 static inline int eproto_to_ipproto(int eproto) 3962 { 3963 switch (eproto) { 3964 case htons(ETH_P_IP): 3965 return IPPROTO_IP; 3966 case htons(ETH_P_IPV6): 3967 return IPPROTO_IPV6; 3968 default: 3969 return -1; 3970 } 3971 } 3972 3973 #ifdef CONFIG_BUG 3974 void netdev_rx_csum_fault(struct net_device *dev); 3975 #else 3976 static inline void netdev_rx_csum_fault(struct net_device *dev) 3977 { 3978 } 3979 #endif 3980 /* rx skb timestamps */ 3981 void net_enable_timestamp(void); 3982 void net_disable_timestamp(void); 3983 3984 #ifdef CONFIG_PROC_FS 3985 int __init dev_proc_init(void); 3986 #else 3987 #define dev_proc_init() 0 3988 #endif 3989 3990 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 3991 struct sk_buff *skb, struct net_device *dev, 3992 bool more) 3993 { 3994 skb->xmit_more = more ? 1 : 0; 3995 return ops->ndo_start_xmit(skb, dev); 3996 } 3997 3998 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 3999 struct netdev_queue *txq, bool more) 4000 { 4001 const struct net_device_ops *ops = dev->netdev_ops; 4002 int rc; 4003 4004 rc = __netdev_start_xmit(ops, skb, dev, more); 4005 if (rc == NETDEV_TX_OK) 4006 txq_trans_update(txq); 4007 4008 return rc; 4009 } 4010 4011 int netdev_class_create_file_ns(struct class_attribute *class_attr, 4012 const void *ns); 4013 void netdev_class_remove_file_ns(struct class_attribute *class_attr, 4014 const void *ns); 4015 4016 static inline int netdev_class_create_file(struct class_attribute *class_attr) 4017 { 4018 return netdev_class_create_file_ns(class_attr, NULL); 4019 } 4020 4021 static inline void netdev_class_remove_file(struct class_attribute *class_attr) 4022 { 4023 netdev_class_remove_file_ns(class_attr, NULL); 4024 } 4025 4026 extern struct kobj_ns_type_operations net_ns_type_operations; 4027 4028 const char *netdev_drivername(const struct net_device *dev); 4029 4030 void linkwatch_run_queue(void); 4031 4032 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4033 netdev_features_t f2) 4034 { 4035 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4036 if (f1 & NETIF_F_HW_CSUM) 4037 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4038 else 4039 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4040 } 4041 4042 return f1 & f2; 4043 } 4044 4045 static inline netdev_features_t netdev_get_wanted_features( 4046 struct net_device *dev) 4047 { 4048 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4049 } 4050 netdev_features_t netdev_increment_features(netdev_features_t all, 4051 netdev_features_t one, netdev_features_t mask); 4052 4053 /* Allow TSO being used on stacked device : 4054 * Performing the GSO segmentation before last device 4055 * is a performance improvement. 4056 */ 4057 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4058 netdev_features_t mask) 4059 { 4060 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4061 } 4062 4063 int __netdev_update_features(struct net_device *dev); 4064 void netdev_update_features(struct net_device *dev); 4065 void netdev_change_features(struct net_device *dev); 4066 4067 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4068 struct net_device *dev); 4069 4070 netdev_features_t passthru_features_check(struct sk_buff *skb, 4071 struct net_device *dev, 4072 netdev_features_t features); 4073 netdev_features_t netif_skb_features(struct sk_buff *skb); 4074 4075 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4076 { 4077 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4078 4079 /* check flags correspondence */ 4080 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4081 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT)); 4082 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4083 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4084 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4085 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4086 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4087 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4088 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4089 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4090 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4091 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4092 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4093 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4094 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4095 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4096 4097 return (features & feature) == feature; 4098 } 4099 4100 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4101 { 4102 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4103 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4104 } 4105 4106 static inline bool netif_needs_gso(struct sk_buff *skb, 4107 netdev_features_t features) 4108 { 4109 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4110 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4111 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4112 } 4113 4114 static inline void netif_set_gso_max_size(struct net_device *dev, 4115 unsigned int size) 4116 { 4117 dev->gso_max_size = size; 4118 } 4119 4120 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4121 int pulled_hlen, u16 mac_offset, 4122 int mac_len) 4123 { 4124 skb->protocol = protocol; 4125 skb->encapsulation = 1; 4126 skb_push(skb, pulled_hlen); 4127 skb_reset_transport_header(skb); 4128 skb->mac_header = mac_offset; 4129 skb->network_header = skb->mac_header + mac_len; 4130 skb->mac_len = mac_len; 4131 } 4132 4133 static inline bool netif_is_macsec(const struct net_device *dev) 4134 { 4135 return dev->priv_flags & IFF_MACSEC; 4136 } 4137 4138 static inline bool netif_is_macvlan(const struct net_device *dev) 4139 { 4140 return dev->priv_flags & IFF_MACVLAN; 4141 } 4142 4143 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4144 { 4145 return dev->priv_flags & IFF_MACVLAN_PORT; 4146 } 4147 4148 static inline bool netif_is_ipvlan(const struct net_device *dev) 4149 { 4150 return dev->priv_flags & IFF_IPVLAN_SLAVE; 4151 } 4152 4153 static inline bool netif_is_ipvlan_port(const struct net_device *dev) 4154 { 4155 return dev->priv_flags & IFF_IPVLAN_MASTER; 4156 } 4157 4158 static inline bool netif_is_bond_master(const struct net_device *dev) 4159 { 4160 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4161 } 4162 4163 static inline bool netif_is_bond_slave(const struct net_device *dev) 4164 { 4165 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4166 } 4167 4168 static inline bool netif_supports_nofcs(struct net_device *dev) 4169 { 4170 return dev->priv_flags & IFF_SUPP_NOFCS; 4171 } 4172 4173 static inline bool netif_is_l3_master(const struct net_device *dev) 4174 { 4175 return dev->priv_flags & IFF_L3MDEV_MASTER; 4176 } 4177 4178 static inline bool netif_is_l3_slave(const struct net_device *dev) 4179 { 4180 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4181 } 4182 4183 static inline bool netif_is_bridge_master(const struct net_device *dev) 4184 { 4185 return dev->priv_flags & IFF_EBRIDGE; 4186 } 4187 4188 static inline bool netif_is_bridge_port(const struct net_device *dev) 4189 { 4190 return dev->priv_flags & IFF_BRIDGE_PORT; 4191 } 4192 4193 static inline bool netif_is_ovs_master(const struct net_device *dev) 4194 { 4195 return dev->priv_flags & IFF_OPENVSWITCH; 4196 } 4197 4198 static inline bool netif_is_team_master(const struct net_device *dev) 4199 { 4200 return dev->priv_flags & IFF_TEAM; 4201 } 4202 4203 static inline bool netif_is_team_port(const struct net_device *dev) 4204 { 4205 return dev->priv_flags & IFF_TEAM_PORT; 4206 } 4207 4208 static inline bool netif_is_lag_master(const struct net_device *dev) 4209 { 4210 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4211 } 4212 4213 static inline bool netif_is_lag_port(const struct net_device *dev) 4214 { 4215 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4216 } 4217 4218 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4219 { 4220 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4221 } 4222 4223 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4224 static inline void netif_keep_dst(struct net_device *dev) 4225 { 4226 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4227 } 4228 4229 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4230 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4231 { 4232 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4233 return dev->priv_flags & IFF_MACSEC; 4234 } 4235 4236 extern struct pernet_operations __net_initdata loopback_net_ops; 4237 4238 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4239 4240 /* netdev_printk helpers, similar to dev_printk */ 4241 4242 static inline const char *netdev_name(const struct net_device *dev) 4243 { 4244 if (!dev->name[0] || strchr(dev->name, '%')) 4245 return "(unnamed net_device)"; 4246 return dev->name; 4247 } 4248 4249 static inline const char *netdev_reg_state(const struct net_device *dev) 4250 { 4251 switch (dev->reg_state) { 4252 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4253 case NETREG_REGISTERED: return ""; 4254 case NETREG_UNREGISTERING: return " (unregistering)"; 4255 case NETREG_UNREGISTERED: return " (unregistered)"; 4256 case NETREG_RELEASED: return " (released)"; 4257 case NETREG_DUMMY: return " (dummy)"; 4258 } 4259 4260 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4261 return " (unknown)"; 4262 } 4263 4264 __printf(3, 4) 4265 void netdev_printk(const char *level, const struct net_device *dev, 4266 const char *format, ...); 4267 __printf(2, 3) 4268 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4269 __printf(2, 3) 4270 void netdev_alert(const struct net_device *dev, const char *format, ...); 4271 __printf(2, 3) 4272 void netdev_crit(const struct net_device *dev, const char *format, ...); 4273 __printf(2, 3) 4274 void netdev_err(const struct net_device *dev, const char *format, ...); 4275 __printf(2, 3) 4276 void netdev_warn(const struct net_device *dev, const char *format, ...); 4277 __printf(2, 3) 4278 void netdev_notice(const struct net_device *dev, const char *format, ...); 4279 __printf(2, 3) 4280 void netdev_info(const struct net_device *dev, const char *format, ...); 4281 4282 #define MODULE_ALIAS_NETDEV(device) \ 4283 MODULE_ALIAS("netdev-" device) 4284 4285 #if defined(CONFIG_DYNAMIC_DEBUG) 4286 #define netdev_dbg(__dev, format, args...) \ 4287 do { \ 4288 dynamic_netdev_dbg(__dev, format, ##args); \ 4289 } while (0) 4290 #elif defined(DEBUG) 4291 #define netdev_dbg(__dev, format, args...) \ 4292 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4293 #else 4294 #define netdev_dbg(__dev, format, args...) \ 4295 ({ \ 4296 if (0) \ 4297 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4298 }) 4299 #endif 4300 4301 #if defined(VERBOSE_DEBUG) 4302 #define netdev_vdbg netdev_dbg 4303 #else 4304 4305 #define netdev_vdbg(dev, format, args...) \ 4306 ({ \ 4307 if (0) \ 4308 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 4309 0; \ 4310 }) 4311 #endif 4312 4313 /* 4314 * netdev_WARN() acts like dev_printk(), but with the key difference 4315 * of using a WARN/WARN_ON to get the message out, including the 4316 * file/line information and a backtrace. 4317 */ 4318 #define netdev_WARN(dev, format, args...) \ 4319 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \ 4320 netdev_reg_state(dev), ##args) 4321 4322 /* netif printk helpers, similar to netdev_printk */ 4323 4324 #define netif_printk(priv, type, level, dev, fmt, args...) \ 4325 do { \ 4326 if (netif_msg_##type(priv)) \ 4327 netdev_printk(level, (dev), fmt, ##args); \ 4328 } while (0) 4329 4330 #define netif_level(level, priv, type, dev, fmt, args...) \ 4331 do { \ 4332 if (netif_msg_##type(priv)) \ 4333 netdev_##level(dev, fmt, ##args); \ 4334 } while (0) 4335 4336 #define netif_emerg(priv, type, dev, fmt, args...) \ 4337 netif_level(emerg, priv, type, dev, fmt, ##args) 4338 #define netif_alert(priv, type, dev, fmt, args...) \ 4339 netif_level(alert, priv, type, dev, fmt, ##args) 4340 #define netif_crit(priv, type, dev, fmt, args...) \ 4341 netif_level(crit, priv, type, dev, fmt, ##args) 4342 #define netif_err(priv, type, dev, fmt, args...) \ 4343 netif_level(err, priv, type, dev, fmt, ##args) 4344 #define netif_warn(priv, type, dev, fmt, args...) \ 4345 netif_level(warn, priv, type, dev, fmt, ##args) 4346 #define netif_notice(priv, type, dev, fmt, args...) \ 4347 netif_level(notice, priv, type, dev, fmt, ##args) 4348 #define netif_info(priv, type, dev, fmt, args...) \ 4349 netif_level(info, priv, type, dev, fmt, ##args) 4350 4351 #if defined(CONFIG_DYNAMIC_DEBUG) 4352 #define netif_dbg(priv, type, netdev, format, args...) \ 4353 do { \ 4354 if (netif_msg_##type(priv)) \ 4355 dynamic_netdev_dbg(netdev, format, ##args); \ 4356 } while (0) 4357 #elif defined(DEBUG) 4358 #define netif_dbg(priv, type, dev, format, args...) \ 4359 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 4360 #else 4361 #define netif_dbg(priv, type, dev, format, args...) \ 4362 ({ \ 4363 if (0) \ 4364 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4365 0; \ 4366 }) 4367 #endif 4368 4369 #if defined(VERBOSE_DEBUG) 4370 #define netif_vdbg netif_dbg 4371 #else 4372 #define netif_vdbg(priv, type, dev, format, args...) \ 4373 ({ \ 4374 if (0) \ 4375 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4376 0; \ 4377 }) 4378 #endif 4379 4380 /* 4381 * The list of packet types we will receive (as opposed to discard) 4382 * and the routines to invoke. 4383 * 4384 * Why 16. Because with 16 the only overlap we get on a hash of the 4385 * low nibble of the protocol value is RARP/SNAP/X.25. 4386 * 4387 * NOTE: That is no longer true with the addition of VLAN tags. Not 4388 * sure which should go first, but I bet it won't make much 4389 * difference if we are running VLANs. The good news is that 4390 * this protocol won't be in the list unless compiled in, so 4391 * the average user (w/out VLANs) will not be adversely affected. 4392 * --BLG 4393 * 4394 * 0800 IP 4395 * 8100 802.1Q VLAN 4396 * 0001 802.3 4397 * 0002 AX.25 4398 * 0004 802.2 4399 * 8035 RARP 4400 * 0005 SNAP 4401 * 0805 X.25 4402 * 0806 ARP 4403 * 8137 IPX 4404 * 0009 Localtalk 4405 * 86DD IPv6 4406 */ 4407 #define PTYPE_HASH_SIZE (16) 4408 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 4409 4410 #endif /* _LINUX_NETDEVICE_H */ 4411