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, <[email protected]> 12 * Corey Minyard <[email protected]> 13 * Donald J. Becker, <[email protected]> 14 * Alan Cox, <[email protected]> 15 * Bjorn Ekwall. <[email protected]> 16 * Pekka Riikonen <[email protected]> 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 56 struct netpoll_info; 57 struct device; 58 struct phy_device; 59 /* 802.11 specific */ 60 struct wireless_dev; 61 /* 802.15.4 specific */ 62 struct wpan_dev; 63 struct mpls_dev; 64 65 void netdev_set_default_ethtool_ops(struct net_device *dev, 66 const struct ethtool_ops *ops); 67 68 /* Backlog congestion levels */ 69 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 70 #define NET_RX_DROP 1 /* packet dropped */ 71 72 /* 73 * Transmit return codes: transmit return codes originate from three different 74 * namespaces: 75 * 76 * - qdisc return codes 77 * - driver transmit return codes 78 * - errno values 79 * 80 * Drivers are allowed to return any one of those in their hard_start_xmit() 81 * function. Real network devices commonly used with qdiscs should only return 82 * the driver transmit return codes though - when qdiscs are used, the actual 83 * transmission happens asynchronously, so the value is not propagated to 84 * higher layers. Virtual network devices transmit synchronously; in this case 85 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 86 * others are propagated to higher layers. 87 */ 88 89 /* qdisc ->enqueue() return codes. */ 90 #define NET_XMIT_SUCCESS 0x00 91 #define NET_XMIT_DROP 0x01 /* skb dropped */ 92 #define NET_XMIT_CN 0x02 /* congestion notification */ 93 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 94 95 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 96 * indicates that the device will soon be dropping packets, or already drops 97 * some packets of the same priority; prompting us to send less aggressively. */ 98 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 99 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 100 101 /* Driver transmit return codes */ 102 #define NETDEV_TX_MASK 0xf0 103 104 enum netdev_tx { 105 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 106 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 107 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 108 }; 109 typedef enum netdev_tx netdev_tx_t; 110 111 /* 112 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 113 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 114 */ 115 static inline bool dev_xmit_complete(int rc) 116 { 117 /* 118 * Positive cases with an skb consumed by a driver: 119 * - successful transmission (rc == NETDEV_TX_OK) 120 * - error while transmitting (rc < 0) 121 * - error while queueing to a different device (rc & NET_XMIT_MASK) 122 */ 123 if (likely(rc < NET_XMIT_MASK)) 124 return true; 125 126 return false; 127 } 128 129 /* 130 * Compute the worst-case header length according to the protocols 131 * used. 132 */ 133 134 #if defined(CONFIG_HYPERV_NET) 135 # define LL_MAX_HEADER 128 136 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 137 # if defined(CONFIG_MAC80211_MESH) 138 # define LL_MAX_HEADER 128 139 # else 140 # define LL_MAX_HEADER 96 141 # endif 142 #else 143 # define LL_MAX_HEADER 32 144 #endif 145 146 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 147 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 148 #define MAX_HEADER LL_MAX_HEADER 149 #else 150 #define MAX_HEADER (LL_MAX_HEADER + 48) 151 #endif 152 153 /* 154 * Old network device statistics. Fields are native words 155 * (unsigned long) so they can be read and written atomically. 156 */ 157 158 struct net_device_stats { 159 unsigned long rx_packets; 160 unsigned long tx_packets; 161 unsigned long rx_bytes; 162 unsigned long tx_bytes; 163 unsigned long rx_errors; 164 unsigned long tx_errors; 165 unsigned long rx_dropped; 166 unsigned long tx_dropped; 167 unsigned long multicast; 168 unsigned long collisions; 169 unsigned long rx_length_errors; 170 unsigned long rx_over_errors; 171 unsigned long rx_crc_errors; 172 unsigned long rx_frame_errors; 173 unsigned long rx_fifo_errors; 174 unsigned long rx_missed_errors; 175 unsigned long tx_aborted_errors; 176 unsigned long tx_carrier_errors; 177 unsigned long tx_fifo_errors; 178 unsigned long tx_heartbeat_errors; 179 unsigned long tx_window_errors; 180 unsigned long rx_compressed; 181 unsigned long tx_compressed; 182 }; 183 184 185 #include <linux/cache.h> 186 #include <linux/skbuff.h> 187 188 #ifdef CONFIG_RPS 189 #include <linux/static_key.h> 190 extern struct static_key rps_needed; 191 #endif 192 193 struct neighbour; 194 struct neigh_parms; 195 struct sk_buff; 196 197 struct netdev_hw_addr { 198 struct list_head list; 199 unsigned char addr[MAX_ADDR_LEN]; 200 unsigned char type; 201 #define NETDEV_HW_ADDR_T_LAN 1 202 #define NETDEV_HW_ADDR_T_SAN 2 203 #define NETDEV_HW_ADDR_T_SLAVE 3 204 #define NETDEV_HW_ADDR_T_UNICAST 4 205 #define NETDEV_HW_ADDR_T_MULTICAST 5 206 bool global_use; 207 int sync_cnt; 208 int refcount; 209 int synced; 210 struct rcu_head rcu_head; 211 }; 212 213 struct netdev_hw_addr_list { 214 struct list_head list; 215 int count; 216 }; 217 218 #define netdev_hw_addr_list_count(l) ((l)->count) 219 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 220 #define netdev_hw_addr_list_for_each(ha, l) \ 221 list_for_each_entry(ha, &(l)->list, list) 222 223 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 224 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 225 #define netdev_for_each_uc_addr(ha, dev) \ 226 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 227 228 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 229 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 230 #define netdev_for_each_mc_addr(ha, dev) \ 231 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 232 233 struct hh_cache { 234 u16 hh_len; 235 u16 __pad; 236 seqlock_t hh_lock; 237 238 /* cached hardware header; allow for machine alignment needs. */ 239 #define HH_DATA_MOD 16 240 #define HH_DATA_OFF(__len) \ 241 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 242 #define HH_DATA_ALIGN(__len) \ 243 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 244 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 245 }; 246 247 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 248 * Alternative is: 249 * dev->hard_header_len ? (dev->hard_header_len + 250 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 251 * 252 * We could use other alignment values, but we must maintain the 253 * relationship HH alignment <= LL alignment. 254 */ 255 #define LL_RESERVED_SPACE(dev) \ 256 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 257 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 258 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 259 260 struct header_ops { 261 int (*create) (struct sk_buff *skb, struct net_device *dev, 262 unsigned short type, const void *daddr, 263 const void *saddr, unsigned int len); 264 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 265 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 266 void (*cache_update)(struct hh_cache *hh, 267 const struct net_device *dev, 268 const unsigned char *haddr); 269 bool (*validate)(const char *ll_header, unsigned int len); 270 }; 271 272 /* These flag bits are private to the generic network queueing 273 * layer; they may not be explicitly referenced by any other 274 * code. 275 */ 276 277 enum netdev_state_t { 278 __LINK_STATE_START, 279 __LINK_STATE_PRESENT, 280 __LINK_STATE_NOCARRIER, 281 __LINK_STATE_LINKWATCH_PENDING, 282 __LINK_STATE_DORMANT, 283 }; 284 285 286 /* 287 * This structure holds boot-time configured netdevice settings. They 288 * are then used in the device probing. 289 */ 290 struct netdev_boot_setup { 291 char name[IFNAMSIZ]; 292 struct ifmap map; 293 }; 294 #define NETDEV_BOOT_SETUP_MAX 8 295 296 int __init netdev_boot_setup(char *str); 297 298 /* 299 * Structure for NAPI scheduling similar to tasklet but with weighting 300 */ 301 struct napi_struct { 302 /* The poll_list must only be managed by the entity which 303 * changes the state of the NAPI_STATE_SCHED bit. This means 304 * whoever atomically sets that bit can add this napi_struct 305 * to the per-CPU poll_list, and whoever clears that bit 306 * can remove from the list right before clearing the bit. 307 */ 308 struct list_head poll_list; 309 310 unsigned long state; 311 int weight; 312 unsigned int gro_count; 313 int (*poll)(struct napi_struct *, int); 314 #ifdef CONFIG_NETPOLL 315 spinlock_t poll_lock; 316 int poll_owner; 317 #endif 318 struct net_device *dev; 319 struct sk_buff *gro_list; 320 struct sk_buff *skb; 321 struct hrtimer timer; 322 struct list_head dev_list; 323 struct hlist_node napi_hash_node; 324 unsigned int napi_id; 325 }; 326 327 enum { 328 NAPI_STATE_SCHED, /* Poll is scheduled */ 329 NAPI_STATE_DISABLE, /* Disable pending */ 330 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 331 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */ 332 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */ 333 }; 334 335 enum gro_result { 336 GRO_MERGED, 337 GRO_MERGED_FREE, 338 GRO_HELD, 339 GRO_NORMAL, 340 GRO_DROP, 341 }; 342 typedef enum gro_result gro_result_t; 343 344 /* 345 * enum rx_handler_result - Possible return values for rx_handlers. 346 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 347 * further. 348 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 349 * case skb->dev was changed by rx_handler. 350 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 351 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 352 * 353 * rx_handlers are functions called from inside __netif_receive_skb(), to do 354 * special processing of the skb, prior to delivery to protocol handlers. 355 * 356 * Currently, a net_device can only have a single rx_handler registered. Trying 357 * to register a second rx_handler will return -EBUSY. 358 * 359 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 360 * To unregister a rx_handler on a net_device, use 361 * netdev_rx_handler_unregister(). 362 * 363 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 364 * do with the skb. 365 * 366 * If the rx_handler consumed the skb in some way, it should return 367 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 368 * the skb to be delivered in some other way. 369 * 370 * If the rx_handler changed skb->dev, to divert the skb to another 371 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 372 * new device will be called if it exists. 373 * 374 * If the rx_handler decides the skb should be ignored, it should return 375 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 376 * are registered on exact device (ptype->dev == skb->dev). 377 * 378 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 379 * delivered, it should return RX_HANDLER_PASS. 380 * 381 * A device without a registered rx_handler will behave as if rx_handler 382 * returned RX_HANDLER_PASS. 383 */ 384 385 enum rx_handler_result { 386 RX_HANDLER_CONSUMED, 387 RX_HANDLER_ANOTHER, 388 RX_HANDLER_EXACT, 389 RX_HANDLER_PASS, 390 }; 391 typedef enum rx_handler_result rx_handler_result_t; 392 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 393 394 void __napi_schedule(struct napi_struct *n); 395 void __napi_schedule_irqoff(struct napi_struct *n); 396 397 static inline bool napi_disable_pending(struct napi_struct *n) 398 { 399 return test_bit(NAPI_STATE_DISABLE, &n->state); 400 } 401 402 /** 403 * napi_schedule_prep - check if NAPI can be scheduled 404 * @n: NAPI context 405 * 406 * Test if NAPI routine is already running, and if not mark 407 * it as running. This is used as a condition variable to 408 * insure only one NAPI poll instance runs. We also make 409 * sure there is no pending NAPI disable. 410 */ 411 static inline bool napi_schedule_prep(struct napi_struct *n) 412 { 413 return !napi_disable_pending(n) && 414 !test_and_set_bit(NAPI_STATE_SCHED, &n->state); 415 } 416 417 /** 418 * napi_schedule - schedule NAPI poll 419 * @n: NAPI context 420 * 421 * Schedule NAPI poll routine to be called if it is not already 422 * running. 423 */ 424 static inline void napi_schedule(struct napi_struct *n) 425 { 426 if (napi_schedule_prep(n)) 427 __napi_schedule(n); 428 } 429 430 /** 431 * napi_schedule_irqoff - schedule NAPI poll 432 * @n: NAPI context 433 * 434 * Variant of napi_schedule(), assuming hard irqs are masked. 435 */ 436 static inline void napi_schedule_irqoff(struct napi_struct *n) 437 { 438 if (napi_schedule_prep(n)) 439 __napi_schedule_irqoff(n); 440 } 441 442 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 443 static inline bool napi_reschedule(struct napi_struct *napi) 444 { 445 if (napi_schedule_prep(napi)) { 446 __napi_schedule(napi); 447 return true; 448 } 449 return false; 450 } 451 452 void __napi_complete(struct napi_struct *n); 453 void napi_complete_done(struct napi_struct *n, int work_done); 454 /** 455 * napi_complete - NAPI processing complete 456 * @n: NAPI context 457 * 458 * Mark NAPI processing as complete. 459 * Consider using napi_complete_done() instead. 460 */ 461 static inline void napi_complete(struct napi_struct *n) 462 { 463 return napi_complete_done(n, 0); 464 } 465 466 /** 467 * napi_hash_add - add a NAPI to global hashtable 468 * @napi: NAPI context 469 * 470 * Generate a new napi_id and store a @napi under it in napi_hash. 471 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL). 472 * Note: This is normally automatically done from netif_napi_add(), 473 * so might disappear in a future Linux version. 474 */ 475 void napi_hash_add(struct napi_struct *napi); 476 477 /** 478 * napi_hash_del - remove a NAPI from global table 479 * @napi: NAPI context 480 * 481 * Warning: caller must observe RCU grace period 482 * before freeing memory containing @napi, if 483 * this function returns true. 484 * Note: core networking stack automatically calls it 485 * from netif_napi_del(). 486 * Drivers might want to call this helper to combine all 487 * the needed RCU grace periods into a single one. 488 */ 489 bool napi_hash_del(struct napi_struct *napi); 490 491 /** 492 * napi_disable - prevent NAPI from scheduling 493 * @n: NAPI context 494 * 495 * Stop NAPI from being scheduled on this context. 496 * Waits till any outstanding processing completes. 497 */ 498 void napi_disable(struct napi_struct *n); 499 500 /** 501 * napi_enable - enable NAPI scheduling 502 * @n: NAPI context 503 * 504 * Resume NAPI from being scheduled on this context. 505 * Must be paired with napi_disable. 506 */ 507 static inline void napi_enable(struct napi_struct *n) 508 { 509 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 510 smp_mb__before_atomic(); 511 clear_bit(NAPI_STATE_SCHED, &n->state); 512 clear_bit(NAPI_STATE_NPSVC, &n->state); 513 } 514 515 /** 516 * napi_synchronize - wait until NAPI is not running 517 * @n: NAPI context 518 * 519 * Wait until NAPI is done being scheduled on this context. 520 * Waits till any outstanding processing completes but 521 * does not disable future activations. 522 */ 523 static inline void napi_synchronize(const struct napi_struct *n) 524 { 525 if (IS_ENABLED(CONFIG_SMP)) 526 while (test_bit(NAPI_STATE_SCHED, &n->state)) 527 msleep(1); 528 else 529 barrier(); 530 } 531 532 enum netdev_queue_state_t { 533 __QUEUE_STATE_DRV_XOFF, 534 __QUEUE_STATE_STACK_XOFF, 535 __QUEUE_STATE_FROZEN, 536 }; 537 538 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 539 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 540 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 541 542 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 543 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 544 QUEUE_STATE_FROZEN) 545 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 546 QUEUE_STATE_FROZEN) 547 548 /* 549 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 550 * netif_tx_* functions below are used to manipulate this flag. The 551 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 552 * queue independently. The netif_xmit_*stopped functions below are called 553 * to check if the queue has been stopped by the driver or stack (either 554 * of the XOFF bits are set in the state). Drivers should not need to call 555 * netif_xmit*stopped functions, they should only be using netif_tx_*. 556 */ 557 558 struct netdev_queue { 559 /* 560 * read-mostly part 561 */ 562 struct net_device *dev; 563 struct Qdisc __rcu *qdisc; 564 struct Qdisc *qdisc_sleeping; 565 #ifdef CONFIG_SYSFS 566 struct kobject kobj; 567 #endif 568 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 569 int numa_node; 570 #endif 571 unsigned long tx_maxrate; 572 /* 573 * Number of TX timeouts for this queue 574 * (/sys/class/net/DEV/Q/trans_timeout) 575 */ 576 unsigned long trans_timeout; 577 /* 578 * write-mostly part 579 */ 580 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 581 int xmit_lock_owner; 582 /* 583 * Time (in jiffies) of last Tx 584 */ 585 unsigned long trans_start; 586 587 unsigned long state; 588 589 #ifdef CONFIG_BQL 590 struct dql dql; 591 #endif 592 } ____cacheline_aligned_in_smp; 593 594 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 595 { 596 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 597 return q->numa_node; 598 #else 599 return NUMA_NO_NODE; 600 #endif 601 } 602 603 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 604 { 605 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 606 q->numa_node = node; 607 #endif 608 } 609 610 #ifdef CONFIG_RPS 611 /* 612 * This structure holds an RPS map which can be of variable length. The 613 * map is an array of CPUs. 614 */ 615 struct rps_map { 616 unsigned int len; 617 struct rcu_head rcu; 618 u16 cpus[0]; 619 }; 620 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 621 622 /* 623 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 624 * tail pointer for that CPU's input queue at the time of last enqueue, and 625 * a hardware filter index. 626 */ 627 struct rps_dev_flow { 628 u16 cpu; 629 u16 filter; 630 unsigned int last_qtail; 631 }; 632 #define RPS_NO_FILTER 0xffff 633 634 /* 635 * The rps_dev_flow_table structure contains a table of flow mappings. 636 */ 637 struct rps_dev_flow_table { 638 unsigned int mask; 639 struct rcu_head rcu; 640 struct rps_dev_flow flows[0]; 641 }; 642 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 643 ((_num) * sizeof(struct rps_dev_flow))) 644 645 /* 646 * The rps_sock_flow_table contains mappings of flows to the last CPU 647 * on which they were processed by the application (set in recvmsg). 648 * Each entry is a 32bit value. Upper part is the high-order bits 649 * of flow hash, lower part is CPU number. 650 * rps_cpu_mask is used to partition the space, depending on number of 651 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 652 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 653 * meaning we use 32-6=26 bits for the hash. 654 */ 655 struct rps_sock_flow_table { 656 u32 mask; 657 658 u32 ents[0] ____cacheline_aligned_in_smp; 659 }; 660 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 661 662 #define RPS_NO_CPU 0xffff 663 664 extern u32 rps_cpu_mask; 665 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 666 667 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 668 u32 hash) 669 { 670 if (table && hash) { 671 unsigned int index = hash & table->mask; 672 u32 val = hash & ~rps_cpu_mask; 673 674 /* We only give a hint, preemption can change CPU under us */ 675 val |= raw_smp_processor_id(); 676 677 if (table->ents[index] != val) 678 table->ents[index] = val; 679 } 680 } 681 682 #ifdef CONFIG_RFS_ACCEL 683 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 684 u16 filter_id); 685 #endif 686 #endif /* CONFIG_RPS */ 687 688 /* This structure contains an instance of an RX queue. */ 689 struct netdev_rx_queue { 690 #ifdef CONFIG_RPS 691 struct rps_map __rcu *rps_map; 692 struct rps_dev_flow_table __rcu *rps_flow_table; 693 #endif 694 struct kobject kobj; 695 struct net_device *dev; 696 } ____cacheline_aligned_in_smp; 697 698 /* 699 * RX queue sysfs structures and functions. 700 */ 701 struct rx_queue_attribute { 702 struct attribute attr; 703 ssize_t (*show)(struct netdev_rx_queue *queue, 704 struct rx_queue_attribute *attr, char *buf); 705 ssize_t (*store)(struct netdev_rx_queue *queue, 706 struct rx_queue_attribute *attr, const char *buf, size_t len); 707 }; 708 709 #ifdef CONFIG_XPS 710 /* 711 * This structure holds an XPS map which can be of variable length. The 712 * map is an array of queues. 713 */ 714 struct xps_map { 715 unsigned int len; 716 unsigned int alloc_len; 717 struct rcu_head rcu; 718 u16 queues[0]; 719 }; 720 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 721 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 722 - sizeof(struct xps_map)) / sizeof(u16)) 723 724 /* 725 * This structure holds all XPS maps for device. Maps are indexed by CPU. 726 */ 727 struct xps_dev_maps { 728 struct rcu_head rcu; 729 struct xps_map __rcu *cpu_map[0]; 730 }; 731 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \ 732 (nr_cpu_ids * sizeof(struct xps_map *))) 733 #endif /* CONFIG_XPS */ 734 735 #define TC_MAX_QUEUE 16 736 #define TC_BITMASK 15 737 /* HW offloaded queuing disciplines txq count and offset maps */ 738 struct netdev_tc_txq { 739 u16 count; 740 u16 offset; 741 }; 742 743 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 744 /* 745 * This structure is to hold information about the device 746 * configured to run FCoE protocol stack. 747 */ 748 struct netdev_fcoe_hbainfo { 749 char manufacturer[64]; 750 char serial_number[64]; 751 char hardware_version[64]; 752 char driver_version[64]; 753 char optionrom_version[64]; 754 char firmware_version[64]; 755 char model[256]; 756 char model_description[256]; 757 }; 758 #endif 759 760 #define MAX_PHYS_ITEM_ID_LEN 32 761 762 /* This structure holds a unique identifier to identify some 763 * physical item (port for example) used by a netdevice. 764 */ 765 struct netdev_phys_item_id { 766 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 767 unsigned char id_len; 768 }; 769 770 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 771 struct netdev_phys_item_id *b) 772 { 773 return a->id_len == b->id_len && 774 memcmp(a->id, b->id, a->id_len) == 0; 775 } 776 777 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 778 struct sk_buff *skb); 779 780 /* These structures hold the attributes of qdisc and classifiers 781 * that are being passed to the netdevice through the setup_tc op. 782 */ 783 enum { 784 TC_SETUP_MQPRIO, 785 TC_SETUP_CLSU32, 786 TC_SETUP_CLSFLOWER, 787 }; 788 789 struct tc_cls_u32_offload; 790 791 struct tc_to_netdev { 792 unsigned int type; 793 union { 794 u8 tc; 795 struct tc_cls_u32_offload *cls_u32; 796 struct tc_cls_flower_offload *cls_flower; 797 }; 798 }; 799 800 801 /* 802 * This structure defines the management hooks for network devices. 803 * The following hooks can be defined; unless noted otherwise, they are 804 * optional and can be filled with a null pointer. 805 * 806 * int (*ndo_init)(struct net_device *dev); 807 * This function is called once when a network device is registered. 808 * The network device can use this for any late stage initialization 809 * or semantic validation. It can fail with an error code which will 810 * be propagated back to register_netdev. 811 * 812 * void (*ndo_uninit)(struct net_device *dev); 813 * This function is called when device is unregistered or when registration 814 * fails. It is not called if init fails. 815 * 816 * int (*ndo_open)(struct net_device *dev); 817 * This function is called when a network device transitions to the up 818 * state. 819 * 820 * int (*ndo_stop)(struct net_device *dev); 821 * This function is called when a network device transitions to the down 822 * state. 823 * 824 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 825 * struct net_device *dev); 826 * Called when a packet needs to be transmitted. 827 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 828 * the queue before that can happen; it's for obsolete devices and weird 829 * corner cases, but the stack really does a non-trivial amount 830 * of useless work if you return NETDEV_TX_BUSY. 831 * Required; cannot be NULL. 832 * 833 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 834 * netdev_features_t features); 835 * Adjusts the requested feature flags according to device-specific 836 * constraints, and returns the resulting flags. Must not modify 837 * the device state. 838 * 839 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 840 * void *accel_priv, select_queue_fallback_t fallback); 841 * Called to decide which queue to use when device supports multiple 842 * transmit queues. 843 * 844 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 845 * This function is called to allow device receiver to make 846 * changes to configuration when multicast or promiscuous is enabled. 847 * 848 * void (*ndo_set_rx_mode)(struct net_device *dev); 849 * This function is called device changes address list filtering. 850 * If driver handles unicast address filtering, it should set 851 * IFF_UNICAST_FLT in its priv_flags. 852 * 853 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 854 * This function is called when the Media Access Control address 855 * needs to be changed. If this interface is not defined, the 856 * MAC address can not be changed. 857 * 858 * int (*ndo_validate_addr)(struct net_device *dev); 859 * Test if Media Access Control address is valid for the device. 860 * 861 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 862 * Called when a user requests an ioctl which can't be handled by 863 * the generic interface code. If not defined ioctls return 864 * not supported error code. 865 * 866 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 867 * Used to set network devices bus interface parameters. This interface 868 * is retained for legacy reasons; new devices should use the bus 869 * interface (PCI) for low level management. 870 * 871 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 872 * Called when a user wants to change the Maximum Transfer Unit 873 * of a device. If not defined, any request to change MTU will 874 * will return an error. 875 * 876 * void (*ndo_tx_timeout)(struct net_device *dev); 877 * Callback used when the transmitter has not made any progress 878 * for dev->watchdog ticks. 879 * 880 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev, 881 * struct rtnl_link_stats64 *storage); 882 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 883 * Called when a user wants to get the network device usage 884 * statistics. Drivers must do one of the following: 885 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 886 * rtnl_link_stats64 structure passed by the caller. 887 * 2. Define @ndo_get_stats to update a net_device_stats structure 888 * (which should normally be dev->stats) and return a pointer to 889 * it. The structure may be changed asynchronously only if each 890 * field is written atomically. 891 * 3. Update dev->stats asynchronously and atomically, and define 892 * neither operation. 893 * 894 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 895 * If device supports VLAN filtering this function is called when a 896 * VLAN id is registered. 897 * 898 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 899 * If device supports VLAN filtering this function is called when a 900 * VLAN id is unregistered. 901 * 902 * void (*ndo_poll_controller)(struct net_device *dev); 903 * 904 * SR-IOV management functions. 905 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 906 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos); 907 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 908 * int max_tx_rate); 909 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 910 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 911 * int (*ndo_get_vf_config)(struct net_device *dev, 912 * int vf, struct ifla_vf_info *ivf); 913 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 914 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 915 * struct nlattr *port[]); 916 * 917 * Enable or disable the VF ability to query its RSS Redirection Table and 918 * Hash Key. This is needed since on some devices VF share this information 919 * with PF and querying it may introduce a theoretical security risk. 920 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 921 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 922 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc) 923 * Called to setup 'tc' number of traffic classes in the net device. This 924 * is always called from the stack with the rtnl lock held and netif tx 925 * queues stopped. This allows the netdevice to perform queue management 926 * safely. 927 * 928 * Fiber Channel over Ethernet (FCoE) offload functions. 929 * int (*ndo_fcoe_enable)(struct net_device *dev); 930 * Called when the FCoE protocol stack wants to start using LLD for FCoE 931 * so the underlying device can perform whatever needed configuration or 932 * initialization to support acceleration of FCoE traffic. 933 * 934 * int (*ndo_fcoe_disable)(struct net_device *dev); 935 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 936 * so the underlying device can perform whatever needed clean-ups to 937 * stop supporting acceleration of FCoE traffic. 938 * 939 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 940 * struct scatterlist *sgl, unsigned int sgc); 941 * Called when the FCoE Initiator wants to initialize an I/O that 942 * is a possible candidate for Direct Data Placement (DDP). The LLD can 943 * perform necessary setup and returns 1 to indicate the device is set up 944 * successfully to perform DDP on this I/O, otherwise this returns 0. 945 * 946 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 947 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 948 * indicated by the FC exchange id 'xid', so the underlying device can 949 * clean up and reuse resources for later DDP requests. 950 * 951 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 952 * struct scatterlist *sgl, unsigned int sgc); 953 * Called when the FCoE Target wants to initialize an I/O that 954 * is a possible candidate for Direct Data Placement (DDP). The LLD can 955 * perform necessary setup and returns 1 to indicate the device is set up 956 * successfully to perform DDP on this I/O, otherwise this returns 0. 957 * 958 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 959 * struct netdev_fcoe_hbainfo *hbainfo); 960 * Called when the FCoE Protocol stack wants information on the underlying 961 * device. This information is utilized by the FCoE protocol stack to 962 * register attributes with Fiber Channel management service as per the 963 * FC-GS Fabric Device Management Information(FDMI) specification. 964 * 965 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 966 * Called when the underlying device wants to override default World Wide 967 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 968 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 969 * protocol stack to use. 970 * 971 * RFS acceleration. 972 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 973 * u16 rxq_index, u32 flow_id); 974 * Set hardware filter for RFS. rxq_index is the target queue index; 975 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 976 * Return the filter ID on success, or a negative error code. 977 * 978 * Slave management functions (for bridge, bonding, etc). 979 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 980 * Called to make another netdev an underling. 981 * 982 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 983 * Called to release previously enslaved netdev. 984 * 985 * Feature/offload setting functions. 986 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 987 * Called to update device configuration to new features. Passed 988 * feature set might be less than what was returned by ndo_fix_features()). 989 * Must return >0 or -errno if it changed dev->features itself. 990 * 991 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 992 * struct net_device *dev, 993 * const unsigned char *addr, u16 vid, u16 flags) 994 * Adds an FDB entry to dev for addr. 995 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 996 * struct net_device *dev, 997 * const unsigned char *addr, u16 vid) 998 * Deletes the FDB entry from dev coresponding to addr. 999 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1000 * struct net_device *dev, struct net_device *filter_dev, 1001 * int idx) 1002 * Used to add FDB entries to dump requests. Implementers should add 1003 * entries to skb and update idx with the number of entries. 1004 * 1005 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1006 * u16 flags) 1007 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1008 * struct net_device *dev, u32 filter_mask, 1009 * int nlflags) 1010 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1011 * u16 flags); 1012 * 1013 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1014 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1015 * which do not represent real hardware may define this to allow their 1016 * userspace components to manage their virtual carrier state. Devices 1017 * that determine carrier state from physical hardware properties (eg 1018 * network cables) or protocol-dependent mechanisms (eg 1019 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1020 * 1021 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1022 * struct netdev_phys_item_id *ppid); 1023 * Called to get ID of physical port of this device. If driver does 1024 * not implement this, it is assumed that the hw is not able to have 1025 * multiple net devices on single physical port. 1026 * 1027 * void (*ndo_add_vxlan_port)(struct net_device *dev, 1028 * sa_family_t sa_family, __be16 port); 1029 * Called by vxlan to notify a driver about the UDP port and socket 1030 * address family that vxlan is listening to. It is called only when 1031 * a new port starts listening. The operation is protected by the 1032 * vxlan_net->sock_lock. 1033 * 1034 * void (*ndo_add_geneve_port)(struct net_device *dev, 1035 * sa_family_t sa_family, __be16 port); 1036 * Called by geneve to notify a driver about the UDP port and socket 1037 * address family that geneve is listnening to. It is called only when 1038 * a new port starts listening. The operation is protected by the 1039 * geneve_net->sock_lock. 1040 * 1041 * void (*ndo_del_geneve_port)(struct net_device *dev, 1042 * sa_family_t sa_family, __be16 port); 1043 * Called by geneve to notify the driver about a UDP port and socket 1044 * address family that geneve is not listening to anymore. The operation 1045 * is protected by the geneve_net->sock_lock. 1046 * 1047 * void (*ndo_del_vxlan_port)(struct net_device *dev, 1048 * sa_family_t sa_family, __be16 port); 1049 * Called by vxlan to notify the driver about a UDP port and socket 1050 * address family that vxlan is not listening to anymore. The operation 1051 * is protected by the vxlan_net->sock_lock. 1052 * 1053 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1054 * struct net_device *dev) 1055 * Called by upper layer devices to accelerate switching or other 1056 * station functionality into hardware. 'pdev is the lowerdev 1057 * to use for the offload and 'dev' is the net device that will 1058 * back the offload. Returns a pointer to the private structure 1059 * the upper layer will maintain. 1060 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1061 * Called by upper layer device to delete the station created 1062 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1063 * the station and priv is the structure returned by the add 1064 * operation. 1065 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb, 1066 * struct net_device *dev, 1067 * void *priv); 1068 * Callback to use for xmit over the accelerated station. This 1069 * is used in place of ndo_start_xmit on accelerated net 1070 * devices. 1071 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1072 * struct net_device *dev 1073 * netdev_features_t features); 1074 * Called by core transmit path to determine if device is capable of 1075 * performing offload operations on a given packet. This is to give 1076 * the device an opportunity to implement any restrictions that cannot 1077 * be otherwise expressed by feature flags. The check is called with 1078 * the set of features that the stack has calculated and it returns 1079 * those the driver believes to be appropriate. 1080 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1081 * int queue_index, u32 maxrate); 1082 * Called when a user wants to set a max-rate limitation of specific 1083 * TX queue. 1084 * int (*ndo_get_iflink)(const struct net_device *dev); 1085 * Called to get the iflink value of this device. 1086 * void (*ndo_change_proto_down)(struct net_device *dev, 1087 * bool proto_down); 1088 * This function is used to pass protocol port error state information 1089 * to the switch driver. The switch driver can react to the proto_down 1090 * by doing a phys down on the associated switch port. 1091 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1092 * This function is used to get egress tunnel information for given skb. 1093 * This is useful for retrieving outer tunnel header parameters while 1094 * sampling packet. 1095 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1096 * This function is used to specify the headroom that the skb must 1097 * consider when allocation skb during packet reception. Setting 1098 * appropriate rx headroom value allows avoiding skb head copy on 1099 * forward. Setting a negative value resets the rx headroom to the 1100 * default value. 1101 * 1102 */ 1103 struct net_device_ops { 1104 int (*ndo_init)(struct net_device *dev); 1105 void (*ndo_uninit)(struct net_device *dev); 1106 int (*ndo_open)(struct net_device *dev); 1107 int (*ndo_stop)(struct net_device *dev); 1108 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1109 struct net_device *dev); 1110 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1111 struct net_device *dev, 1112 netdev_features_t features); 1113 u16 (*ndo_select_queue)(struct net_device *dev, 1114 struct sk_buff *skb, 1115 void *accel_priv, 1116 select_queue_fallback_t fallback); 1117 void (*ndo_change_rx_flags)(struct net_device *dev, 1118 int flags); 1119 void (*ndo_set_rx_mode)(struct net_device *dev); 1120 int (*ndo_set_mac_address)(struct net_device *dev, 1121 void *addr); 1122 int (*ndo_validate_addr)(struct net_device *dev); 1123 int (*ndo_do_ioctl)(struct net_device *dev, 1124 struct ifreq *ifr, int cmd); 1125 int (*ndo_set_config)(struct net_device *dev, 1126 struct ifmap *map); 1127 int (*ndo_change_mtu)(struct net_device *dev, 1128 int new_mtu); 1129 int (*ndo_neigh_setup)(struct net_device *dev, 1130 struct neigh_parms *); 1131 void (*ndo_tx_timeout) (struct net_device *dev); 1132 1133 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev, 1134 struct rtnl_link_stats64 *storage); 1135 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1136 1137 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1138 __be16 proto, u16 vid); 1139 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1140 __be16 proto, u16 vid); 1141 #ifdef CONFIG_NET_POLL_CONTROLLER 1142 void (*ndo_poll_controller)(struct net_device *dev); 1143 int (*ndo_netpoll_setup)(struct net_device *dev, 1144 struct netpoll_info *info); 1145 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1146 #endif 1147 #ifdef CONFIG_NET_RX_BUSY_POLL 1148 int (*ndo_busy_poll)(struct napi_struct *dev); 1149 #endif 1150 int (*ndo_set_vf_mac)(struct net_device *dev, 1151 int queue, u8 *mac); 1152 int (*ndo_set_vf_vlan)(struct net_device *dev, 1153 int queue, u16 vlan, u8 qos); 1154 int (*ndo_set_vf_rate)(struct net_device *dev, 1155 int vf, int min_tx_rate, 1156 int max_tx_rate); 1157 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1158 int vf, bool setting); 1159 int (*ndo_set_vf_trust)(struct net_device *dev, 1160 int vf, bool setting); 1161 int (*ndo_get_vf_config)(struct net_device *dev, 1162 int vf, 1163 struct ifla_vf_info *ivf); 1164 int (*ndo_set_vf_link_state)(struct net_device *dev, 1165 int vf, int link_state); 1166 int (*ndo_get_vf_stats)(struct net_device *dev, 1167 int vf, 1168 struct ifla_vf_stats 1169 *vf_stats); 1170 int (*ndo_set_vf_port)(struct net_device *dev, 1171 int vf, 1172 struct nlattr *port[]); 1173 int (*ndo_get_vf_port)(struct net_device *dev, 1174 int vf, struct sk_buff *skb); 1175 int (*ndo_set_vf_guid)(struct net_device *dev, 1176 int vf, u64 guid, 1177 int guid_type); 1178 int (*ndo_set_vf_rss_query_en)( 1179 struct net_device *dev, 1180 int vf, bool setting); 1181 int (*ndo_setup_tc)(struct net_device *dev, 1182 u32 handle, 1183 __be16 protocol, 1184 struct tc_to_netdev *tc); 1185 #if IS_ENABLED(CONFIG_FCOE) 1186 int (*ndo_fcoe_enable)(struct net_device *dev); 1187 int (*ndo_fcoe_disable)(struct net_device *dev); 1188 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1189 u16 xid, 1190 struct scatterlist *sgl, 1191 unsigned int sgc); 1192 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1193 u16 xid); 1194 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1195 u16 xid, 1196 struct scatterlist *sgl, 1197 unsigned int sgc); 1198 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1199 struct netdev_fcoe_hbainfo *hbainfo); 1200 #endif 1201 1202 #if IS_ENABLED(CONFIG_LIBFCOE) 1203 #define NETDEV_FCOE_WWNN 0 1204 #define NETDEV_FCOE_WWPN 1 1205 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1206 u64 *wwn, int type); 1207 #endif 1208 1209 #ifdef CONFIG_RFS_ACCEL 1210 int (*ndo_rx_flow_steer)(struct net_device *dev, 1211 const struct sk_buff *skb, 1212 u16 rxq_index, 1213 u32 flow_id); 1214 #endif 1215 int (*ndo_add_slave)(struct net_device *dev, 1216 struct net_device *slave_dev); 1217 int (*ndo_del_slave)(struct net_device *dev, 1218 struct net_device *slave_dev); 1219 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1220 netdev_features_t features); 1221 int (*ndo_set_features)(struct net_device *dev, 1222 netdev_features_t features); 1223 int (*ndo_neigh_construct)(struct neighbour *n); 1224 void (*ndo_neigh_destroy)(struct neighbour *n); 1225 1226 int (*ndo_fdb_add)(struct ndmsg *ndm, 1227 struct nlattr *tb[], 1228 struct net_device *dev, 1229 const unsigned char *addr, 1230 u16 vid, 1231 u16 flags); 1232 int (*ndo_fdb_del)(struct ndmsg *ndm, 1233 struct nlattr *tb[], 1234 struct net_device *dev, 1235 const unsigned char *addr, 1236 u16 vid); 1237 int (*ndo_fdb_dump)(struct sk_buff *skb, 1238 struct netlink_callback *cb, 1239 struct net_device *dev, 1240 struct net_device *filter_dev, 1241 int idx); 1242 1243 int (*ndo_bridge_setlink)(struct net_device *dev, 1244 struct nlmsghdr *nlh, 1245 u16 flags); 1246 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1247 u32 pid, u32 seq, 1248 struct net_device *dev, 1249 u32 filter_mask, 1250 int nlflags); 1251 int (*ndo_bridge_dellink)(struct net_device *dev, 1252 struct nlmsghdr *nlh, 1253 u16 flags); 1254 int (*ndo_change_carrier)(struct net_device *dev, 1255 bool new_carrier); 1256 int (*ndo_get_phys_port_id)(struct net_device *dev, 1257 struct netdev_phys_item_id *ppid); 1258 int (*ndo_get_phys_port_name)(struct net_device *dev, 1259 char *name, size_t len); 1260 void (*ndo_add_vxlan_port)(struct net_device *dev, 1261 sa_family_t sa_family, 1262 __be16 port); 1263 void (*ndo_del_vxlan_port)(struct net_device *dev, 1264 sa_family_t sa_family, 1265 __be16 port); 1266 void (*ndo_add_geneve_port)(struct net_device *dev, 1267 sa_family_t sa_family, 1268 __be16 port); 1269 void (*ndo_del_geneve_port)(struct net_device *dev, 1270 sa_family_t sa_family, 1271 __be16 port); 1272 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1273 struct net_device *dev); 1274 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1275 void *priv); 1276 1277 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb, 1278 struct net_device *dev, 1279 void *priv); 1280 int (*ndo_get_lock_subclass)(struct net_device *dev); 1281 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1282 int queue_index, 1283 u32 maxrate); 1284 int (*ndo_get_iflink)(const struct net_device *dev); 1285 int (*ndo_change_proto_down)(struct net_device *dev, 1286 bool proto_down); 1287 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1288 struct sk_buff *skb); 1289 void (*ndo_set_rx_headroom)(struct net_device *dev, 1290 int needed_headroom); 1291 }; 1292 1293 /** 1294 * enum net_device_priv_flags - &struct net_device priv_flags 1295 * 1296 * These are the &struct net_device, they are only set internally 1297 * by drivers and used in the kernel. These flags are invisible to 1298 * userspace; this means that the order of these flags can change 1299 * during any kernel release. 1300 * 1301 * You should have a pretty good reason to be extending these flags. 1302 * 1303 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1304 * @IFF_EBRIDGE: Ethernet bridging device 1305 * @IFF_BONDING: bonding master or slave 1306 * @IFF_ISATAP: ISATAP interface (RFC4214) 1307 * @IFF_WAN_HDLC: WAN HDLC device 1308 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1309 * release skb->dst 1310 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1311 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1312 * @IFF_MACVLAN_PORT: device used as macvlan port 1313 * @IFF_BRIDGE_PORT: device used as bridge port 1314 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1315 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1316 * @IFF_UNICAST_FLT: Supports unicast filtering 1317 * @IFF_TEAM_PORT: device used as team port 1318 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1319 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1320 * change when it's running 1321 * @IFF_MACVLAN: Macvlan device 1322 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1323 * underlying stacked devices 1324 * @IFF_IPVLAN_MASTER: IPvlan master device 1325 * @IFF_IPVLAN_SLAVE: IPvlan slave device 1326 * @IFF_L3MDEV_MASTER: device is an L3 master device 1327 * @IFF_NO_QUEUE: device can run without qdisc attached 1328 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1329 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1330 * @IFF_TEAM: device is a team device 1331 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1332 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1333 * entity (i.e. the master device for bridged veth) 1334 * @IFF_MACSEC: device is a MACsec device 1335 */ 1336 enum netdev_priv_flags { 1337 IFF_802_1Q_VLAN = 1<<0, 1338 IFF_EBRIDGE = 1<<1, 1339 IFF_BONDING = 1<<2, 1340 IFF_ISATAP = 1<<3, 1341 IFF_WAN_HDLC = 1<<4, 1342 IFF_XMIT_DST_RELEASE = 1<<5, 1343 IFF_DONT_BRIDGE = 1<<6, 1344 IFF_DISABLE_NETPOLL = 1<<7, 1345 IFF_MACVLAN_PORT = 1<<8, 1346 IFF_BRIDGE_PORT = 1<<9, 1347 IFF_OVS_DATAPATH = 1<<10, 1348 IFF_TX_SKB_SHARING = 1<<11, 1349 IFF_UNICAST_FLT = 1<<12, 1350 IFF_TEAM_PORT = 1<<13, 1351 IFF_SUPP_NOFCS = 1<<14, 1352 IFF_LIVE_ADDR_CHANGE = 1<<15, 1353 IFF_MACVLAN = 1<<16, 1354 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1355 IFF_IPVLAN_MASTER = 1<<18, 1356 IFF_IPVLAN_SLAVE = 1<<19, 1357 IFF_L3MDEV_MASTER = 1<<20, 1358 IFF_NO_QUEUE = 1<<21, 1359 IFF_OPENVSWITCH = 1<<22, 1360 IFF_L3MDEV_SLAVE = 1<<23, 1361 IFF_TEAM = 1<<24, 1362 IFF_RXFH_CONFIGURED = 1<<25, 1363 IFF_PHONY_HEADROOM = 1<<26, 1364 IFF_MACSEC = 1<<27, 1365 }; 1366 1367 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1368 #define IFF_EBRIDGE IFF_EBRIDGE 1369 #define IFF_BONDING IFF_BONDING 1370 #define IFF_ISATAP IFF_ISATAP 1371 #define IFF_WAN_HDLC IFF_WAN_HDLC 1372 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1373 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1374 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1375 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1376 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1377 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1378 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1379 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1380 #define IFF_TEAM_PORT IFF_TEAM_PORT 1381 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1382 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1383 #define IFF_MACVLAN IFF_MACVLAN 1384 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1385 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER 1386 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE 1387 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1388 #define IFF_NO_QUEUE IFF_NO_QUEUE 1389 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1390 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1391 #define IFF_TEAM IFF_TEAM 1392 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1393 #define IFF_MACSEC IFF_MACSEC 1394 1395 /** 1396 * struct net_device - The DEVICE structure. 1397 * Actually, this whole structure is a big mistake. It mixes I/O 1398 * data with strictly "high-level" data, and it has to know about 1399 * almost every data structure used in the INET module. 1400 * 1401 * @name: This is the first field of the "visible" part of this structure 1402 * (i.e. as seen by users in the "Space.c" file). It is the name 1403 * of the interface. 1404 * 1405 * @name_hlist: Device name hash chain, please keep it close to name[] 1406 * @ifalias: SNMP alias 1407 * @mem_end: Shared memory end 1408 * @mem_start: Shared memory start 1409 * @base_addr: Device I/O address 1410 * @irq: Device IRQ number 1411 * 1412 * @carrier_changes: Stats to monitor carrier on<->off transitions 1413 * 1414 * @state: Generic network queuing layer state, see netdev_state_t 1415 * @dev_list: The global list of network devices 1416 * @napi_list: List entry used for polling NAPI devices 1417 * @unreg_list: List entry when we are unregistering the 1418 * device; see the function unregister_netdev 1419 * @close_list: List entry used when we are closing the device 1420 * @ptype_all: Device-specific packet handlers for all protocols 1421 * @ptype_specific: Device-specific, protocol-specific packet handlers 1422 * 1423 * @adj_list: Directly linked devices, like slaves for bonding 1424 * @all_adj_list: All linked devices, *including* neighbours 1425 * @features: Currently active device features 1426 * @hw_features: User-changeable features 1427 * 1428 * @wanted_features: User-requested features 1429 * @vlan_features: Mask of features inheritable by VLAN devices 1430 * 1431 * @hw_enc_features: Mask of features inherited by encapsulating devices 1432 * This field indicates what encapsulation 1433 * offloads the hardware is capable of doing, 1434 * and drivers will need to set them appropriately. 1435 * 1436 * @mpls_features: Mask of features inheritable by MPLS 1437 * 1438 * @ifindex: interface index 1439 * @group: The group the device belongs to 1440 * 1441 * @stats: Statistics struct, which was left as a legacy, use 1442 * rtnl_link_stats64 instead 1443 * 1444 * @rx_dropped: Dropped packets by core network, 1445 * do not use this in drivers 1446 * @tx_dropped: Dropped packets by core network, 1447 * do not use this in drivers 1448 * @rx_nohandler: nohandler dropped packets by core network on 1449 * inactive devices, do not use this in drivers 1450 * 1451 * @wireless_handlers: List of functions to handle Wireless Extensions, 1452 * instead of ioctl, 1453 * see <net/iw_handler.h> for details. 1454 * @wireless_data: Instance data managed by the core of wireless extensions 1455 * 1456 * @netdev_ops: Includes several pointers to callbacks, 1457 * if one wants to override the ndo_*() functions 1458 * @ethtool_ops: Management operations 1459 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1460 * of Layer 2 headers. 1461 * 1462 * @flags: Interface flags (a la BSD) 1463 * @priv_flags: Like 'flags' but invisible to userspace, 1464 * see if.h for the definitions 1465 * @gflags: Global flags ( kept as legacy ) 1466 * @padded: How much padding added by alloc_netdev() 1467 * @operstate: RFC2863 operstate 1468 * @link_mode: Mapping policy to operstate 1469 * @if_port: Selectable AUI, TP, ... 1470 * @dma: DMA channel 1471 * @mtu: Interface MTU value 1472 * @type: Interface hardware type 1473 * @hard_header_len: Maximum hardware header length. 1474 * 1475 * @needed_headroom: Extra headroom the hardware may need, but not in all 1476 * cases can this be guaranteed 1477 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1478 * cases can this be guaranteed. Some cases also use 1479 * LL_MAX_HEADER instead to allocate the skb 1480 * 1481 * interface address info: 1482 * 1483 * @perm_addr: Permanent hw address 1484 * @addr_assign_type: Hw address assignment type 1485 * @addr_len: Hardware address length 1486 * @neigh_priv_len; Used in neigh_alloc(), 1487 * initialized only in atm/clip.c 1488 * @dev_id: Used to differentiate devices that share 1489 * the same link layer address 1490 * @dev_port: Used to differentiate devices that share 1491 * the same function 1492 * @addr_list_lock: XXX: need comments on this one 1493 * @uc_promisc: Counter that indicates promiscuous mode 1494 * has been enabled due to the need to listen to 1495 * additional unicast addresses in a device that 1496 * does not implement ndo_set_rx_mode() 1497 * @uc: unicast mac addresses 1498 * @mc: multicast mac addresses 1499 * @dev_addrs: list of device hw addresses 1500 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1501 * @promiscuity: Number of times the NIC is told to work in 1502 * promiscuous mode; if it becomes 0 the NIC will 1503 * exit promiscuous mode 1504 * @allmulti: Counter, enables or disables allmulticast mode 1505 * 1506 * @vlan_info: VLAN info 1507 * @dsa_ptr: dsa specific data 1508 * @tipc_ptr: TIPC specific data 1509 * @atalk_ptr: AppleTalk link 1510 * @ip_ptr: IPv4 specific data 1511 * @dn_ptr: DECnet specific data 1512 * @ip6_ptr: IPv6 specific data 1513 * @ax25_ptr: AX.25 specific data 1514 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1515 * 1516 * @last_rx: Time of last Rx 1517 * @dev_addr: Hw address (before bcast, 1518 * because most packets are unicast) 1519 * 1520 * @_rx: Array of RX queues 1521 * @num_rx_queues: Number of RX queues 1522 * allocated at register_netdev() time 1523 * @real_num_rx_queues: Number of RX queues currently active in device 1524 * 1525 * @rx_handler: handler for received packets 1526 * @rx_handler_data: XXX: need comments on this one 1527 * @ingress_queue: XXX: need comments on this one 1528 * @broadcast: hw bcast address 1529 * 1530 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1531 * indexed by RX queue number. Assigned by driver. 1532 * This must only be set if the ndo_rx_flow_steer 1533 * operation is defined 1534 * @index_hlist: Device index hash chain 1535 * 1536 * @_tx: Array of TX queues 1537 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1538 * @real_num_tx_queues: Number of TX queues currently active in device 1539 * @qdisc: Root qdisc from userspace point of view 1540 * @tx_queue_len: Max frames per queue allowed 1541 * @tx_global_lock: XXX: need comments on this one 1542 * 1543 * @xps_maps: XXX: need comments on this one 1544 * 1545 * @offload_fwd_mark: Offload device fwding mark 1546 * 1547 * @watchdog_timeo: Represents the timeout that is used by 1548 * the watchdog (see dev_watchdog()) 1549 * @watchdog_timer: List of timers 1550 * 1551 * @pcpu_refcnt: Number of references to this device 1552 * @todo_list: Delayed register/unregister 1553 * @link_watch_list: XXX: need comments on this one 1554 * 1555 * @reg_state: Register/unregister state machine 1556 * @dismantle: Device is going to be freed 1557 * @rtnl_link_state: This enum represents the phases of creating 1558 * a new link 1559 * 1560 * @destructor: Called from unregister, 1561 * can be used to call free_netdev 1562 * @npinfo: XXX: need comments on this one 1563 * @nd_net: Network namespace this network device is inside 1564 * 1565 * @ml_priv: Mid-layer private 1566 * @lstats: Loopback statistics 1567 * @tstats: Tunnel statistics 1568 * @dstats: Dummy statistics 1569 * @vstats: Virtual ethernet statistics 1570 * 1571 * @garp_port: GARP 1572 * @mrp_port: MRP 1573 * 1574 * @dev: Class/net/name entry 1575 * @sysfs_groups: Space for optional device, statistics and wireless 1576 * sysfs groups 1577 * 1578 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1579 * @rtnl_link_ops: Rtnl_link_ops 1580 * 1581 * @gso_max_size: Maximum size of generic segmentation offload 1582 * @gso_max_segs: Maximum number of segments that can be passed to the 1583 * NIC for GSO 1584 * 1585 * @dcbnl_ops: Data Center Bridging netlink ops 1586 * @num_tc: Number of traffic classes in the net device 1587 * @tc_to_txq: XXX: need comments on this one 1588 * @prio_tc_map XXX: need comments on this one 1589 * 1590 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1591 * 1592 * @priomap: XXX: need comments on this one 1593 * @phydev: Physical device may attach itself 1594 * for hardware timestamping 1595 * 1596 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1597 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1598 * 1599 * @proto_down: protocol port state information can be sent to the 1600 * switch driver and used to set the phys state of the 1601 * switch port. 1602 * 1603 * FIXME: cleanup struct net_device such that network protocol info 1604 * moves out. 1605 */ 1606 1607 struct net_device { 1608 char name[IFNAMSIZ]; 1609 struct hlist_node name_hlist; 1610 char *ifalias; 1611 /* 1612 * I/O specific fields 1613 * FIXME: Merge these and struct ifmap into one 1614 */ 1615 unsigned long mem_end; 1616 unsigned long mem_start; 1617 unsigned long base_addr; 1618 int irq; 1619 1620 atomic_t carrier_changes; 1621 1622 /* 1623 * Some hardware also needs these fields (state,dev_list, 1624 * napi_list,unreg_list,close_list) but they are not 1625 * part of the usual set specified in Space.c. 1626 */ 1627 1628 unsigned long state; 1629 1630 struct list_head dev_list; 1631 struct list_head napi_list; 1632 struct list_head unreg_list; 1633 struct list_head close_list; 1634 struct list_head ptype_all; 1635 struct list_head ptype_specific; 1636 1637 struct { 1638 struct list_head upper; 1639 struct list_head lower; 1640 } adj_list; 1641 1642 struct { 1643 struct list_head upper; 1644 struct list_head lower; 1645 } all_adj_list; 1646 1647 netdev_features_t features; 1648 netdev_features_t hw_features; 1649 netdev_features_t wanted_features; 1650 netdev_features_t vlan_features; 1651 netdev_features_t hw_enc_features; 1652 netdev_features_t mpls_features; 1653 netdev_features_t gso_partial_features; 1654 1655 int ifindex; 1656 int group; 1657 1658 struct net_device_stats stats; 1659 1660 atomic_long_t rx_dropped; 1661 atomic_long_t tx_dropped; 1662 atomic_long_t rx_nohandler; 1663 1664 #ifdef CONFIG_WIRELESS_EXT 1665 const struct iw_handler_def *wireless_handlers; 1666 struct iw_public_data *wireless_data; 1667 #endif 1668 const struct net_device_ops *netdev_ops; 1669 const struct ethtool_ops *ethtool_ops; 1670 #ifdef CONFIG_NET_SWITCHDEV 1671 const struct switchdev_ops *switchdev_ops; 1672 #endif 1673 #ifdef CONFIG_NET_L3_MASTER_DEV 1674 const struct l3mdev_ops *l3mdev_ops; 1675 #endif 1676 1677 const struct header_ops *header_ops; 1678 1679 unsigned int flags; 1680 unsigned int priv_flags; 1681 1682 unsigned short gflags; 1683 unsigned short padded; 1684 1685 unsigned char operstate; 1686 unsigned char link_mode; 1687 1688 unsigned char if_port; 1689 unsigned char dma; 1690 1691 unsigned int mtu; 1692 unsigned short type; 1693 unsigned short hard_header_len; 1694 1695 unsigned short needed_headroom; 1696 unsigned short needed_tailroom; 1697 1698 /* Interface address info. */ 1699 unsigned char perm_addr[MAX_ADDR_LEN]; 1700 unsigned char addr_assign_type; 1701 unsigned char addr_len; 1702 unsigned short neigh_priv_len; 1703 unsigned short dev_id; 1704 unsigned short dev_port; 1705 spinlock_t addr_list_lock; 1706 unsigned char name_assign_type; 1707 bool uc_promisc; 1708 struct netdev_hw_addr_list uc; 1709 struct netdev_hw_addr_list mc; 1710 struct netdev_hw_addr_list dev_addrs; 1711 1712 #ifdef CONFIG_SYSFS 1713 struct kset *queues_kset; 1714 #endif 1715 unsigned int promiscuity; 1716 unsigned int allmulti; 1717 1718 1719 /* Protocol-specific pointers */ 1720 1721 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1722 struct vlan_info __rcu *vlan_info; 1723 #endif 1724 #if IS_ENABLED(CONFIG_NET_DSA) 1725 struct dsa_switch_tree *dsa_ptr; 1726 #endif 1727 #if IS_ENABLED(CONFIG_TIPC) 1728 struct tipc_bearer __rcu *tipc_ptr; 1729 #endif 1730 void *atalk_ptr; 1731 struct in_device __rcu *ip_ptr; 1732 struct dn_dev __rcu *dn_ptr; 1733 struct inet6_dev __rcu *ip6_ptr; 1734 void *ax25_ptr; 1735 struct wireless_dev *ieee80211_ptr; 1736 struct wpan_dev *ieee802154_ptr; 1737 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 1738 struct mpls_dev __rcu *mpls_ptr; 1739 #endif 1740 1741 /* 1742 * Cache lines mostly used on receive path (including eth_type_trans()) 1743 */ 1744 unsigned long last_rx; 1745 1746 /* Interface address info used in eth_type_trans() */ 1747 unsigned char *dev_addr; 1748 1749 #ifdef CONFIG_SYSFS 1750 struct netdev_rx_queue *_rx; 1751 1752 unsigned int num_rx_queues; 1753 unsigned int real_num_rx_queues; 1754 #endif 1755 1756 unsigned long gro_flush_timeout; 1757 rx_handler_func_t __rcu *rx_handler; 1758 void __rcu *rx_handler_data; 1759 1760 #ifdef CONFIG_NET_CLS_ACT 1761 struct tcf_proto __rcu *ingress_cl_list; 1762 #endif 1763 struct netdev_queue __rcu *ingress_queue; 1764 #ifdef CONFIG_NETFILTER_INGRESS 1765 struct list_head nf_hooks_ingress; 1766 #endif 1767 1768 unsigned char broadcast[MAX_ADDR_LEN]; 1769 #ifdef CONFIG_RFS_ACCEL 1770 struct cpu_rmap *rx_cpu_rmap; 1771 #endif 1772 struct hlist_node index_hlist; 1773 1774 /* 1775 * Cache lines mostly used on transmit path 1776 */ 1777 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1778 unsigned int num_tx_queues; 1779 unsigned int real_num_tx_queues; 1780 struct Qdisc *qdisc; 1781 unsigned long tx_queue_len; 1782 spinlock_t tx_global_lock; 1783 int watchdog_timeo; 1784 1785 #ifdef CONFIG_XPS 1786 struct xps_dev_maps __rcu *xps_maps; 1787 #endif 1788 #ifdef CONFIG_NET_CLS_ACT 1789 struct tcf_proto __rcu *egress_cl_list; 1790 #endif 1791 #ifdef CONFIG_NET_SWITCHDEV 1792 u32 offload_fwd_mark; 1793 #endif 1794 1795 /* These may be needed for future network-power-down code. */ 1796 struct timer_list watchdog_timer; 1797 1798 int __percpu *pcpu_refcnt; 1799 struct list_head todo_list; 1800 1801 struct list_head link_watch_list; 1802 1803 enum { NETREG_UNINITIALIZED=0, 1804 NETREG_REGISTERED, /* completed register_netdevice */ 1805 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1806 NETREG_UNREGISTERED, /* completed unregister todo */ 1807 NETREG_RELEASED, /* called free_netdev */ 1808 NETREG_DUMMY, /* dummy device for NAPI poll */ 1809 } reg_state:8; 1810 1811 bool dismantle; 1812 1813 enum { 1814 RTNL_LINK_INITIALIZED, 1815 RTNL_LINK_INITIALIZING, 1816 } rtnl_link_state:16; 1817 1818 void (*destructor)(struct net_device *dev); 1819 1820 #ifdef CONFIG_NETPOLL 1821 struct netpoll_info __rcu *npinfo; 1822 #endif 1823 1824 possible_net_t nd_net; 1825 1826 /* mid-layer private */ 1827 union { 1828 void *ml_priv; 1829 struct pcpu_lstats __percpu *lstats; 1830 struct pcpu_sw_netstats __percpu *tstats; 1831 struct pcpu_dstats __percpu *dstats; 1832 struct pcpu_vstats __percpu *vstats; 1833 }; 1834 1835 struct garp_port __rcu *garp_port; 1836 struct mrp_port __rcu *mrp_port; 1837 1838 struct device dev; 1839 const struct attribute_group *sysfs_groups[4]; 1840 const struct attribute_group *sysfs_rx_queue_group; 1841 1842 const struct rtnl_link_ops *rtnl_link_ops; 1843 1844 /* for setting kernel sock attribute on TCP connection setup */ 1845 #define GSO_MAX_SIZE 65536 1846 unsigned int gso_max_size; 1847 #define GSO_MAX_SEGS 65535 1848 u16 gso_max_segs; 1849 1850 #ifdef CONFIG_DCB 1851 const struct dcbnl_rtnl_ops *dcbnl_ops; 1852 #endif 1853 u8 num_tc; 1854 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 1855 u8 prio_tc_map[TC_BITMASK + 1]; 1856 1857 #if IS_ENABLED(CONFIG_FCOE) 1858 unsigned int fcoe_ddp_xid; 1859 #endif 1860 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 1861 struct netprio_map __rcu *priomap; 1862 #endif 1863 struct phy_device *phydev; 1864 struct lock_class_key *qdisc_tx_busylock; 1865 struct lock_class_key *qdisc_running_key; 1866 bool proto_down; 1867 }; 1868 #define to_net_dev(d) container_of(d, struct net_device, dev) 1869 1870 #define NETDEV_ALIGN 32 1871 1872 static inline 1873 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 1874 { 1875 return dev->prio_tc_map[prio & TC_BITMASK]; 1876 } 1877 1878 static inline 1879 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 1880 { 1881 if (tc >= dev->num_tc) 1882 return -EINVAL; 1883 1884 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 1885 return 0; 1886 } 1887 1888 static inline 1889 void netdev_reset_tc(struct net_device *dev) 1890 { 1891 dev->num_tc = 0; 1892 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq)); 1893 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map)); 1894 } 1895 1896 static inline 1897 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset) 1898 { 1899 if (tc >= dev->num_tc) 1900 return -EINVAL; 1901 1902 dev->tc_to_txq[tc].count = count; 1903 dev->tc_to_txq[tc].offset = offset; 1904 return 0; 1905 } 1906 1907 static inline 1908 int netdev_set_num_tc(struct net_device *dev, u8 num_tc) 1909 { 1910 if (num_tc > TC_MAX_QUEUE) 1911 return -EINVAL; 1912 1913 dev->num_tc = num_tc; 1914 return 0; 1915 } 1916 1917 static inline 1918 int netdev_get_num_tc(struct net_device *dev) 1919 { 1920 return dev->num_tc; 1921 } 1922 1923 static inline 1924 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 1925 unsigned int index) 1926 { 1927 return &dev->_tx[index]; 1928 } 1929 1930 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 1931 const struct sk_buff *skb) 1932 { 1933 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 1934 } 1935 1936 static inline void netdev_for_each_tx_queue(struct net_device *dev, 1937 void (*f)(struct net_device *, 1938 struct netdev_queue *, 1939 void *), 1940 void *arg) 1941 { 1942 unsigned int i; 1943 1944 for (i = 0; i < dev->num_tx_queues; i++) 1945 f(dev, &dev->_tx[i], arg); 1946 } 1947 1948 #define netdev_lockdep_set_classes(dev) \ 1949 { \ 1950 static struct lock_class_key qdisc_tx_busylock_key; \ 1951 static struct lock_class_key qdisc_running_key; \ 1952 static struct lock_class_key qdisc_xmit_lock_key; \ 1953 static struct lock_class_key dev_addr_list_lock_key; \ 1954 unsigned int i; \ 1955 \ 1956 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 1957 (dev)->qdisc_running_key = &qdisc_running_key; \ 1958 lockdep_set_class(&(dev)->addr_list_lock, \ 1959 &dev_addr_list_lock_key); \ 1960 for (i = 0; i < (dev)->num_tx_queues; i++) \ 1961 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 1962 &qdisc_xmit_lock_key); \ 1963 } 1964 1965 struct netdev_queue *netdev_pick_tx(struct net_device *dev, 1966 struct sk_buff *skb, 1967 void *accel_priv); 1968 1969 /* returns the headroom that the master device needs to take in account 1970 * when forwarding to this dev 1971 */ 1972 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 1973 { 1974 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 1975 } 1976 1977 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 1978 { 1979 if (dev->netdev_ops->ndo_set_rx_headroom) 1980 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 1981 } 1982 1983 /* set the device rx headroom to the dev's default */ 1984 static inline void netdev_reset_rx_headroom(struct net_device *dev) 1985 { 1986 netdev_set_rx_headroom(dev, -1); 1987 } 1988 1989 /* 1990 * Net namespace inlines 1991 */ 1992 static inline 1993 struct net *dev_net(const struct net_device *dev) 1994 { 1995 return read_pnet(&dev->nd_net); 1996 } 1997 1998 static inline 1999 void dev_net_set(struct net_device *dev, struct net *net) 2000 { 2001 write_pnet(&dev->nd_net, net); 2002 } 2003 2004 static inline bool netdev_uses_dsa(struct net_device *dev) 2005 { 2006 #if IS_ENABLED(CONFIG_NET_DSA) 2007 if (dev->dsa_ptr != NULL) 2008 return dsa_uses_tagged_protocol(dev->dsa_ptr); 2009 #endif 2010 return false; 2011 } 2012 2013 /** 2014 * netdev_priv - access network device private data 2015 * @dev: network device 2016 * 2017 * Get network device private data 2018 */ 2019 static inline void *netdev_priv(const struct net_device *dev) 2020 { 2021 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2022 } 2023 2024 /* Set the sysfs physical device reference for the network logical device 2025 * if set prior to registration will cause a symlink during initialization. 2026 */ 2027 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2028 2029 /* Set the sysfs device type for the network logical device to allow 2030 * fine-grained identification of different network device types. For 2031 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2032 */ 2033 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2034 2035 /* Default NAPI poll() weight 2036 * Device drivers are strongly advised to not use bigger value 2037 */ 2038 #define NAPI_POLL_WEIGHT 64 2039 2040 /** 2041 * netif_napi_add - initialize a NAPI context 2042 * @dev: network device 2043 * @napi: NAPI context 2044 * @poll: polling function 2045 * @weight: default weight 2046 * 2047 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2048 * *any* of the other NAPI-related functions. 2049 */ 2050 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2051 int (*poll)(struct napi_struct *, int), int weight); 2052 2053 /** 2054 * netif_tx_napi_add - initialize a NAPI context 2055 * @dev: network device 2056 * @napi: NAPI context 2057 * @poll: polling function 2058 * @weight: default weight 2059 * 2060 * This variant of netif_napi_add() should be used from drivers using NAPI 2061 * to exclusively poll a TX queue. 2062 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2063 */ 2064 static inline void netif_tx_napi_add(struct net_device *dev, 2065 struct napi_struct *napi, 2066 int (*poll)(struct napi_struct *, int), 2067 int weight) 2068 { 2069 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2070 netif_napi_add(dev, napi, poll, weight); 2071 } 2072 2073 /** 2074 * netif_napi_del - remove a NAPI context 2075 * @napi: NAPI context 2076 * 2077 * netif_napi_del() removes a NAPI context from the network device NAPI list 2078 */ 2079 void netif_napi_del(struct napi_struct *napi); 2080 2081 struct napi_gro_cb { 2082 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2083 void *frag0; 2084 2085 /* Length of frag0. */ 2086 unsigned int frag0_len; 2087 2088 /* This indicates where we are processing relative to skb->data. */ 2089 int data_offset; 2090 2091 /* This is non-zero if the packet cannot be merged with the new skb. */ 2092 u16 flush; 2093 2094 /* Save the IP ID here and check when we get to the transport layer */ 2095 u16 flush_id; 2096 2097 /* Number of segments aggregated. */ 2098 u16 count; 2099 2100 /* Start offset for remote checksum offload */ 2101 u16 gro_remcsum_start; 2102 2103 /* jiffies when first packet was created/queued */ 2104 unsigned long age; 2105 2106 /* Used in ipv6_gro_receive() and foo-over-udp */ 2107 u16 proto; 2108 2109 /* This is non-zero if the packet may be of the same flow. */ 2110 u8 same_flow:1; 2111 2112 /* Used in tunnel GRO receive */ 2113 u8 encap_mark:1; 2114 2115 /* GRO checksum is valid */ 2116 u8 csum_valid:1; 2117 2118 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2119 u8 csum_cnt:3; 2120 2121 /* Free the skb? */ 2122 u8 free:2; 2123 #define NAPI_GRO_FREE 1 2124 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2125 2126 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2127 u8 is_ipv6:1; 2128 2129 /* Used in GRE, set in fou/gue_gro_receive */ 2130 u8 is_fou:1; 2131 2132 /* Used to determine if flush_id can be ignored */ 2133 u8 is_atomic:1; 2134 2135 /* 5 bit hole */ 2136 2137 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2138 __wsum csum; 2139 2140 /* used in skb_gro_receive() slow path */ 2141 struct sk_buff *last; 2142 }; 2143 2144 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2145 2146 struct packet_type { 2147 __be16 type; /* This is really htons(ether_type). */ 2148 struct net_device *dev; /* NULL is wildcarded here */ 2149 int (*func) (struct sk_buff *, 2150 struct net_device *, 2151 struct packet_type *, 2152 struct net_device *); 2153 bool (*id_match)(struct packet_type *ptype, 2154 struct sock *sk); 2155 void *af_packet_priv; 2156 struct list_head list; 2157 }; 2158 2159 struct offload_callbacks { 2160 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2161 netdev_features_t features); 2162 struct sk_buff **(*gro_receive)(struct sk_buff **head, 2163 struct sk_buff *skb); 2164 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2165 }; 2166 2167 struct packet_offload { 2168 __be16 type; /* This is really htons(ether_type). */ 2169 u16 priority; 2170 struct offload_callbacks callbacks; 2171 struct list_head list; 2172 }; 2173 2174 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2175 struct pcpu_sw_netstats { 2176 u64 rx_packets; 2177 u64 rx_bytes; 2178 u64 tx_packets; 2179 u64 tx_bytes; 2180 struct u64_stats_sync syncp; 2181 }; 2182 2183 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2184 ({ \ 2185 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2186 if (pcpu_stats) { \ 2187 int __cpu; \ 2188 for_each_possible_cpu(__cpu) { \ 2189 typeof(type) *stat; \ 2190 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2191 u64_stats_init(&stat->syncp); \ 2192 } \ 2193 } \ 2194 pcpu_stats; \ 2195 }) 2196 2197 #define netdev_alloc_pcpu_stats(type) \ 2198 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2199 2200 enum netdev_lag_tx_type { 2201 NETDEV_LAG_TX_TYPE_UNKNOWN, 2202 NETDEV_LAG_TX_TYPE_RANDOM, 2203 NETDEV_LAG_TX_TYPE_BROADCAST, 2204 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2205 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2206 NETDEV_LAG_TX_TYPE_HASH, 2207 }; 2208 2209 struct netdev_lag_upper_info { 2210 enum netdev_lag_tx_type tx_type; 2211 }; 2212 2213 struct netdev_lag_lower_state_info { 2214 u8 link_up : 1, 2215 tx_enabled : 1; 2216 }; 2217 2218 #include <linux/notifier.h> 2219 2220 /* netdevice notifier chain. Please remember to update the rtnetlink 2221 * notification exclusion list in rtnetlink_event() when adding new 2222 * types. 2223 */ 2224 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */ 2225 #define NETDEV_DOWN 0x0002 2226 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface 2227 detected a hardware crash and restarted 2228 - we can use this eg to kick tcp sessions 2229 once done */ 2230 #define NETDEV_CHANGE 0x0004 /* Notify device state change */ 2231 #define NETDEV_REGISTER 0x0005 2232 #define NETDEV_UNREGISTER 0x0006 2233 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */ 2234 #define NETDEV_CHANGEADDR 0x0008 2235 #define NETDEV_GOING_DOWN 0x0009 2236 #define NETDEV_CHANGENAME 0x000A 2237 #define NETDEV_FEAT_CHANGE 0x000B 2238 #define NETDEV_BONDING_FAILOVER 0x000C 2239 #define NETDEV_PRE_UP 0x000D 2240 #define NETDEV_PRE_TYPE_CHANGE 0x000E 2241 #define NETDEV_POST_TYPE_CHANGE 0x000F 2242 #define NETDEV_POST_INIT 0x0010 2243 #define NETDEV_UNREGISTER_FINAL 0x0011 2244 #define NETDEV_RELEASE 0x0012 2245 #define NETDEV_NOTIFY_PEERS 0x0013 2246 #define NETDEV_JOIN 0x0014 2247 #define NETDEV_CHANGEUPPER 0x0015 2248 #define NETDEV_RESEND_IGMP 0x0016 2249 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */ 2250 #define NETDEV_CHANGEINFODATA 0x0018 2251 #define NETDEV_BONDING_INFO 0x0019 2252 #define NETDEV_PRECHANGEUPPER 0x001A 2253 #define NETDEV_CHANGELOWERSTATE 0x001B 2254 #define NETDEV_OFFLOAD_PUSH_VXLAN 0x001C 2255 #define NETDEV_OFFLOAD_PUSH_GENEVE 0x001D 2256 2257 int register_netdevice_notifier(struct notifier_block *nb); 2258 int unregister_netdevice_notifier(struct notifier_block *nb); 2259 2260 struct netdev_notifier_info { 2261 struct net_device *dev; 2262 }; 2263 2264 struct netdev_notifier_change_info { 2265 struct netdev_notifier_info info; /* must be first */ 2266 unsigned int flags_changed; 2267 }; 2268 2269 struct netdev_notifier_changeupper_info { 2270 struct netdev_notifier_info info; /* must be first */ 2271 struct net_device *upper_dev; /* new upper dev */ 2272 bool master; /* is upper dev master */ 2273 bool linking; /* is the notification for link or unlink */ 2274 void *upper_info; /* upper dev info */ 2275 }; 2276 2277 struct netdev_notifier_changelowerstate_info { 2278 struct netdev_notifier_info info; /* must be first */ 2279 void *lower_state_info; /* is lower dev state */ 2280 }; 2281 2282 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2283 struct net_device *dev) 2284 { 2285 info->dev = dev; 2286 } 2287 2288 static inline struct net_device * 2289 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2290 { 2291 return info->dev; 2292 } 2293 2294 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2295 2296 2297 extern rwlock_t dev_base_lock; /* Device list lock */ 2298 2299 #define for_each_netdev(net, d) \ 2300 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2301 #define for_each_netdev_reverse(net, d) \ 2302 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2303 #define for_each_netdev_rcu(net, d) \ 2304 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2305 #define for_each_netdev_safe(net, d, n) \ 2306 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2307 #define for_each_netdev_continue(net, d) \ 2308 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2309 #define for_each_netdev_continue_rcu(net, d) \ 2310 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2311 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2312 for_each_netdev_rcu(&init_net, slave) \ 2313 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2314 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2315 2316 static inline struct net_device *next_net_device(struct net_device *dev) 2317 { 2318 struct list_head *lh; 2319 struct net *net; 2320 2321 net = dev_net(dev); 2322 lh = dev->dev_list.next; 2323 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2324 } 2325 2326 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2327 { 2328 struct list_head *lh; 2329 struct net *net; 2330 2331 net = dev_net(dev); 2332 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2333 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2334 } 2335 2336 static inline struct net_device *first_net_device(struct net *net) 2337 { 2338 return list_empty(&net->dev_base_head) ? NULL : 2339 net_device_entry(net->dev_base_head.next); 2340 } 2341 2342 static inline struct net_device *first_net_device_rcu(struct net *net) 2343 { 2344 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2345 2346 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2347 } 2348 2349 int netdev_boot_setup_check(struct net_device *dev); 2350 unsigned long netdev_boot_base(const char *prefix, int unit); 2351 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2352 const char *hwaddr); 2353 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2354 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2355 void dev_add_pack(struct packet_type *pt); 2356 void dev_remove_pack(struct packet_type *pt); 2357 void __dev_remove_pack(struct packet_type *pt); 2358 void dev_add_offload(struct packet_offload *po); 2359 void dev_remove_offload(struct packet_offload *po); 2360 2361 int dev_get_iflink(const struct net_device *dev); 2362 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2363 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2364 unsigned short mask); 2365 struct net_device *dev_get_by_name(struct net *net, const char *name); 2366 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2367 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2368 int dev_alloc_name(struct net_device *dev, const char *name); 2369 int dev_open(struct net_device *dev); 2370 int dev_close(struct net_device *dev); 2371 int dev_close_many(struct list_head *head, bool unlink); 2372 void dev_disable_lro(struct net_device *dev); 2373 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2374 int dev_queue_xmit(struct sk_buff *skb); 2375 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv); 2376 int register_netdevice(struct net_device *dev); 2377 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2378 void unregister_netdevice_many(struct list_head *head); 2379 static inline void unregister_netdevice(struct net_device *dev) 2380 { 2381 unregister_netdevice_queue(dev, NULL); 2382 } 2383 2384 int netdev_refcnt_read(const struct net_device *dev); 2385 void free_netdev(struct net_device *dev); 2386 void netdev_freemem(struct net_device *dev); 2387 void synchronize_net(void); 2388 int init_dummy_netdev(struct net_device *dev); 2389 2390 DECLARE_PER_CPU(int, xmit_recursion); 2391 #define XMIT_RECURSION_LIMIT 10 2392 2393 static inline int dev_recursion_level(void) 2394 { 2395 return this_cpu_read(xmit_recursion); 2396 } 2397 2398 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2399 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2400 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2401 int netdev_get_name(struct net *net, char *name, int ifindex); 2402 int dev_restart(struct net_device *dev); 2403 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb); 2404 2405 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2406 { 2407 return NAPI_GRO_CB(skb)->data_offset; 2408 } 2409 2410 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2411 { 2412 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2413 } 2414 2415 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2416 { 2417 NAPI_GRO_CB(skb)->data_offset += len; 2418 } 2419 2420 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2421 unsigned int offset) 2422 { 2423 return NAPI_GRO_CB(skb)->frag0 + offset; 2424 } 2425 2426 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2427 { 2428 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2429 } 2430 2431 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2432 unsigned int offset) 2433 { 2434 if (!pskb_may_pull(skb, hlen)) 2435 return NULL; 2436 2437 NAPI_GRO_CB(skb)->frag0 = NULL; 2438 NAPI_GRO_CB(skb)->frag0_len = 0; 2439 return skb->data + offset; 2440 } 2441 2442 static inline void *skb_gro_network_header(struct sk_buff *skb) 2443 { 2444 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2445 skb_network_offset(skb); 2446 } 2447 2448 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2449 const void *start, unsigned int len) 2450 { 2451 if (NAPI_GRO_CB(skb)->csum_valid) 2452 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2453 csum_partial(start, len, 0)); 2454 } 2455 2456 /* GRO checksum functions. These are logical equivalents of the normal 2457 * checksum functions (in skbuff.h) except that they operate on the GRO 2458 * offsets and fields in sk_buff. 2459 */ 2460 2461 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2462 2463 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2464 { 2465 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2466 } 2467 2468 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2469 bool zero_okay, 2470 __sum16 check) 2471 { 2472 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2473 skb_checksum_start_offset(skb) < 2474 skb_gro_offset(skb)) && 2475 !skb_at_gro_remcsum_start(skb) && 2476 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2477 (!zero_okay || check)); 2478 } 2479 2480 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2481 __wsum psum) 2482 { 2483 if (NAPI_GRO_CB(skb)->csum_valid && 2484 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2485 return 0; 2486 2487 NAPI_GRO_CB(skb)->csum = psum; 2488 2489 return __skb_gro_checksum_complete(skb); 2490 } 2491 2492 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2493 { 2494 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2495 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2496 NAPI_GRO_CB(skb)->csum_cnt--; 2497 } else { 2498 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2499 * verified a new top level checksum or an encapsulated one 2500 * during GRO. This saves work if we fallback to normal path. 2501 */ 2502 __skb_incr_checksum_unnecessary(skb); 2503 } 2504 } 2505 2506 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2507 compute_pseudo) \ 2508 ({ \ 2509 __sum16 __ret = 0; \ 2510 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2511 __ret = __skb_gro_checksum_validate_complete(skb, \ 2512 compute_pseudo(skb, proto)); \ 2513 if (__ret) \ 2514 __skb_mark_checksum_bad(skb); \ 2515 else \ 2516 skb_gro_incr_csum_unnecessary(skb); \ 2517 __ret; \ 2518 }) 2519 2520 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2521 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2522 2523 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2524 compute_pseudo) \ 2525 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2526 2527 #define skb_gro_checksum_simple_validate(skb) \ 2528 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2529 2530 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2531 { 2532 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2533 !NAPI_GRO_CB(skb)->csum_valid); 2534 } 2535 2536 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2537 __sum16 check, __wsum pseudo) 2538 { 2539 NAPI_GRO_CB(skb)->csum = ~pseudo; 2540 NAPI_GRO_CB(skb)->csum_valid = 1; 2541 } 2542 2543 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \ 2544 do { \ 2545 if (__skb_gro_checksum_convert_check(skb)) \ 2546 __skb_gro_checksum_convert(skb, check, \ 2547 compute_pseudo(skb, proto)); \ 2548 } while (0) 2549 2550 struct gro_remcsum { 2551 int offset; 2552 __wsum delta; 2553 }; 2554 2555 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2556 { 2557 grc->offset = 0; 2558 grc->delta = 0; 2559 } 2560 2561 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2562 unsigned int off, size_t hdrlen, 2563 int start, int offset, 2564 struct gro_remcsum *grc, 2565 bool nopartial) 2566 { 2567 __wsum delta; 2568 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2569 2570 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2571 2572 if (!nopartial) { 2573 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2574 return ptr; 2575 } 2576 2577 ptr = skb_gro_header_fast(skb, off); 2578 if (skb_gro_header_hard(skb, off + plen)) { 2579 ptr = skb_gro_header_slow(skb, off + plen, off); 2580 if (!ptr) 2581 return NULL; 2582 } 2583 2584 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2585 start, offset); 2586 2587 /* Adjust skb->csum since we changed the packet */ 2588 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2589 2590 grc->offset = off + hdrlen + offset; 2591 grc->delta = delta; 2592 2593 return ptr; 2594 } 2595 2596 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2597 struct gro_remcsum *grc) 2598 { 2599 void *ptr; 2600 size_t plen = grc->offset + sizeof(u16); 2601 2602 if (!grc->delta) 2603 return; 2604 2605 ptr = skb_gro_header_fast(skb, grc->offset); 2606 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 2607 ptr = skb_gro_header_slow(skb, plen, grc->offset); 2608 if (!ptr) 2609 return; 2610 } 2611 2612 remcsum_unadjust((__sum16 *)ptr, grc->delta); 2613 } 2614 2615 struct skb_csum_offl_spec { 2616 __u16 ipv4_okay:1, 2617 ipv6_okay:1, 2618 encap_okay:1, 2619 ip_options_okay:1, 2620 ext_hdrs_okay:1, 2621 tcp_okay:1, 2622 udp_okay:1, 2623 sctp_okay:1, 2624 vlan_okay:1, 2625 no_encapped_ipv6:1, 2626 no_not_encapped:1; 2627 }; 2628 2629 bool __skb_csum_offload_chk(struct sk_buff *skb, 2630 const struct skb_csum_offl_spec *spec, 2631 bool *csum_encapped, 2632 bool csum_help); 2633 2634 static inline bool skb_csum_offload_chk(struct sk_buff *skb, 2635 const struct skb_csum_offl_spec *spec, 2636 bool *csum_encapped, 2637 bool csum_help) 2638 { 2639 if (skb->ip_summed != CHECKSUM_PARTIAL) 2640 return false; 2641 2642 return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help); 2643 } 2644 2645 static inline bool skb_csum_offload_chk_help(struct sk_buff *skb, 2646 const struct skb_csum_offl_spec *spec) 2647 { 2648 bool csum_encapped; 2649 2650 return skb_csum_offload_chk(skb, spec, &csum_encapped, true); 2651 } 2652 2653 static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb) 2654 { 2655 static const struct skb_csum_offl_spec csum_offl_spec = { 2656 .ipv4_okay = 1, 2657 .ip_options_okay = 1, 2658 .ipv6_okay = 1, 2659 .vlan_okay = 1, 2660 .tcp_okay = 1, 2661 .udp_okay = 1, 2662 }; 2663 2664 return skb_csum_offload_chk_help(skb, &csum_offl_spec); 2665 } 2666 2667 static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb) 2668 { 2669 static const struct skb_csum_offl_spec csum_offl_spec = { 2670 .ipv4_okay = 1, 2671 .ip_options_okay = 1, 2672 .tcp_okay = 1, 2673 .udp_okay = 1, 2674 .vlan_okay = 1, 2675 }; 2676 2677 return skb_csum_offload_chk_help(skb, &csum_offl_spec); 2678 } 2679 2680 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2681 unsigned short type, 2682 const void *daddr, const void *saddr, 2683 unsigned int len) 2684 { 2685 if (!dev->header_ops || !dev->header_ops->create) 2686 return 0; 2687 2688 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2689 } 2690 2691 static inline int dev_parse_header(const struct sk_buff *skb, 2692 unsigned char *haddr) 2693 { 2694 const struct net_device *dev = skb->dev; 2695 2696 if (!dev->header_ops || !dev->header_ops->parse) 2697 return 0; 2698 return dev->header_ops->parse(skb, haddr); 2699 } 2700 2701 /* ll_header must have at least hard_header_len allocated */ 2702 static inline bool dev_validate_header(const struct net_device *dev, 2703 char *ll_header, int len) 2704 { 2705 if (likely(len >= dev->hard_header_len)) 2706 return true; 2707 2708 if (capable(CAP_SYS_RAWIO)) { 2709 memset(ll_header + len, 0, dev->hard_header_len - len); 2710 return true; 2711 } 2712 2713 if (dev->header_ops && dev->header_ops->validate) 2714 return dev->header_ops->validate(ll_header, len); 2715 2716 return false; 2717 } 2718 2719 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len); 2720 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2721 static inline int unregister_gifconf(unsigned int family) 2722 { 2723 return register_gifconf(family, NULL); 2724 } 2725 2726 #ifdef CONFIG_NET_FLOW_LIMIT 2727 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2728 struct sd_flow_limit { 2729 u64 count; 2730 unsigned int num_buckets; 2731 unsigned int history_head; 2732 u16 history[FLOW_LIMIT_HISTORY]; 2733 u8 buckets[]; 2734 }; 2735 2736 extern int netdev_flow_limit_table_len; 2737 #endif /* CONFIG_NET_FLOW_LIMIT */ 2738 2739 /* 2740 * Incoming packets are placed on per-CPU queues 2741 */ 2742 struct softnet_data { 2743 struct list_head poll_list; 2744 struct sk_buff_head process_queue; 2745 2746 /* stats */ 2747 unsigned int processed; 2748 unsigned int time_squeeze; 2749 unsigned int received_rps; 2750 #ifdef CONFIG_RPS 2751 struct softnet_data *rps_ipi_list; 2752 #endif 2753 #ifdef CONFIG_NET_FLOW_LIMIT 2754 struct sd_flow_limit __rcu *flow_limit; 2755 #endif 2756 struct Qdisc *output_queue; 2757 struct Qdisc **output_queue_tailp; 2758 struct sk_buff *completion_queue; 2759 2760 #ifdef CONFIG_RPS 2761 /* input_queue_head should be written by cpu owning this struct, 2762 * and only read by other cpus. Worth using a cache line. 2763 */ 2764 unsigned int input_queue_head ____cacheline_aligned_in_smp; 2765 2766 /* Elements below can be accessed between CPUs for RPS/RFS */ 2767 struct call_single_data csd ____cacheline_aligned_in_smp; 2768 struct softnet_data *rps_ipi_next; 2769 unsigned int cpu; 2770 unsigned int input_queue_tail; 2771 #endif 2772 unsigned int dropped; 2773 struct sk_buff_head input_pkt_queue; 2774 struct napi_struct backlog; 2775 2776 }; 2777 2778 static inline void input_queue_head_incr(struct softnet_data *sd) 2779 { 2780 #ifdef CONFIG_RPS 2781 sd->input_queue_head++; 2782 #endif 2783 } 2784 2785 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 2786 unsigned int *qtail) 2787 { 2788 #ifdef CONFIG_RPS 2789 *qtail = ++sd->input_queue_tail; 2790 #endif 2791 } 2792 2793 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 2794 2795 void __netif_schedule(struct Qdisc *q); 2796 void netif_schedule_queue(struct netdev_queue *txq); 2797 2798 static inline void netif_tx_schedule_all(struct net_device *dev) 2799 { 2800 unsigned int i; 2801 2802 for (i = 0; i < dev->num_tx_queues; i++) 2803 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 2804 } 2805 2806 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 2807 { 2808 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2809 } 2810 2811 /** 2812 * netif_start_queue - allow transmit 2813 * @dev: network device 2814 * 2815 * Allow upper layers to call the device hard_start_xmit routine. 2816 */ 2817 static inline void netif_start_queue(struct net_device *dev) 2818 { 2819 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 2820 } 2821 2822 static inline void netif_tx_start_all_queues(struct net_device *dev) 2823 { 2824 unsigned int i; 2825 2826 for (i = 0; i < dev->num_tx_queues; i++) { 2827 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2828 netif_tx_start_queue(txq); 2829 } 2830 } 2831 2832 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 2833 2834 /** 2835 * netif_wake_queue - restart transmit 2836 * @dev: network device 2837 * 2838 * Allow upper layers to call the device hard_start_xmit routine. 2839 * Used for flow control when transmit resources are available. 2840 */ 2841 static inline void netif_wake_queue(struct net_device *dev) 2842 { 2843 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 2844 } 2845 2846 static inline void netif_tx_wake_all_queues(struct net_device *dev) 2847 { 2848 unsigned int i; 2849 2850 for (i = 0; i < dev->num_tx_queues; i++) { 2851 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2852 netif_tx_wake_queue(txq); 2853 } 2854 } 2855 2856 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 2857 { 2858 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2859 } 2860 2861 /** 2862 * netif_stop_queue - stop transmitted packets 2863 * @dev: network device 2864 * 2865 * Stop upper layers calling the device hard_start_xmit routine. 2866 * Used for flow control when transmit resources are unavailable. 2867 */ 2868 static inline void netif_stop_queue(struct net_device *dev) 2869 { 2870 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 2871 } 2872 2873 void netif_tx_stop_all_queues(struct net_device *dev); 2874 2875 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 2876 { 2877 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2878 } 2879 2880 /** 2881 * netif_queue_stopped - test if transmit queue is flowblocked 2882 * @dev: network device 2883 * 2884 * Test if transmit queue on device is currently unable to send. 2885 */ 2886 static inline bool netif_queue_stopped(const struct net_device *dev) 2887 { 2888 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 2889 } 2890 2891 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 2892 { 2893 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 2894 } 2895 2896 static inline bool 2897 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 2898 { 2899 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 2900 } 2901 2902 static inline bool 2903 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 2904 { 2905 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 2906 } 2907 2908 /** 2909 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 2910 * @dev_queue: pointer to transmit queue 2911 * 2912 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 2913 * to give appropriate hint to the CPU. 2914 */ 2915 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 2916 { 2917 #ifdef CONFIG_BQL 2918 prefetchw(&dev_queue->dql.num_queued); 2919 #endif 2920 } 2921 2922 /** 2923 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 2924 * @dev_queue: pointer to transmit queue 2925 * 2926 * BQL enabled drivers might use this helper in their TX completion path, 2927 * to give appropriate hint to the CPU. 2928 */ 2929 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 2930 { 2931 #ifdef CONFIG_BQL 2932 prefetchw(&dev_queue->dql.limit); 2933 #endif 2934 } 2935 2936 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 2937 unsigned int bytes) 2938 { 2939 #ifdef CONFIG_BQL 2940 dql_queued(&dev_queue->dql, bytes); 2941 2942 if (likely(dql_avail(&dev_queue->dql) >= 0)) 2943 return; 2944 2945 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2946 2947 /* 2948 * The XOFF flag must be set before checking the dql_avail below, 2949 * because in netdev_tx_completed_queue we update the dql_completed 2950 * before checking the XOFF flag. 2951 */ 2952 smp_mb(); 2953 2954 /* check again in case another CPU has just made room avail */ 2955 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 2956 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2957 #endif 2958 } 2959 2960 /** 2961 * netdev_sent_queue - report the number of bytes queued to hardware 2962 * @dev: network device 2963 * @bytes: number of bytes queued to the hardware device queue 2964 * 2965 * Report the number of bytes queued for sending/completion to the network 2966 * device hardware queue. @bytes should be a good approximation and should 2967 * exactly match netdev_completed_queue() @bytes 2968 */ 2969 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 2970 { 2971 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 2972 } 2973 2974 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 2975 unsigned int pkts, unsigned int bytes) 2976 { 2977 #ifdef CONFIG_BQL 2978 if (unlikely(!bytes)) 2979 return; 2980 2981 dql_completed(&dev_queue->dql, bytes); 2982 2983 /* 2984 * Without the memory barrier there is a small possiblity that 2985 * netdev_tx_sent_queue will miss the update and cause the queue to 2986 * be stopped forever 2987 */ 2988 smp_mb(); 2989 2990 if (dql_avail(&dev_queue->dql) < 0) 2991 return; 2992 2993 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 2994 netif_schedule_queue(dev_queue); 2995 #endif 2996 } 2997 2998 /** 2999 * netdev_completed_queue - report bytes and packets completed by device 3000 * @dev: network device 3001 * @pkts: actual number of packets sent over the medium 3002 * @bytes: actual number of bytes sent over the medium 3003 * 3004 * Report the number of bytes and packets transmitted by the network device 3005 * hardware queue over the physical medium, @bytes must exactly match the 3006 * @bytes amount passed to netdev_sent_queue() 3007 */ 3008 static inline void netdev_completed_queue(struct net_device *dev, 3009 unsigned int pkts, unsigned int bytes) 3010 { 3011 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3012 } 3013 3014 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3015 { 3016 #ifdef CONFIG_BQL 3017 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3018 dql_reset(&q->dql); 3019 #endif 3020 } 3021 3022 /** 3023 * netdev_reset_queue - reset the packets and bytes count of a network device 3024 * @dev_queue: network device 3025 * 3026 * Reset the bytes and packet count of a network device and clear the 3027 * software flow control OFF bit for this network device 3028 */ 3029 static inline void netdev_reset_queue(struct net_device *dev_queue) 3030 { 3031 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3032 } 3033 3034 /** 3035 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3036 * @dev: network device 3037 * @queue_index: given tx queue index 3038 * 3039 * Returns 0 if given tx queue index >= number of device tx queues, 3040 * otherwise returns the originally passed tx queue index. 3041 */ 3042 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3043 { 3044 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3045 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3046 dev->name, queue_index, 3047 dev->real_num_tx_queues); 3048 return 0; 3049 } 3050 3051 return queue_index; 3052 } 3053 3054 /** 3055 * netif_running - test if up 3056 * @dev: network device 3057 * 3058 * Test if the device has been brought up. 3059 */ 3060 static inline bool netif_running(const struct net_device *dev) 3061 { 3062 return test_bit(__LINK_STATE_START, &dev->state); 3063 } 3064 3065 /* 3066 * Routines to manage the subqueues on a device. We only need start, 3067 * stop, and a check if it's stopped. All other device management is 3068 * done at the overall netdevice level. 3069 * Also test the device if we're multiqueue. 3070 */ 3071 3072 /** 3073 * netif_start_subqueue - allow sending packets on subqueue 3074 * @dev: network device 3075 * @queue_index: sub queue index 3076 * 3077 * Start individual transmit queue of a device with multiple transmit queues. 3078 */ 3079 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3080 { 3081 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3082 3083 netif_tx_start_queue(txq); 3084 } 3085 3086 /** 3087 * netif_stop_subqueue - stop sending packets on subqueue 3088 * @dev: network device 3089 * @queue_index: sub queue index 3090 * 3091 * Stop individual transmit queue of a device with multiple transmit queues. 3092 */ 3093 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3094 { 3095 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3096 netif_tx_stop_queue(txq); 3097 } 3098 3099 /** 3100 * netif_subqueue_stopped - test status of subqueue 3101 * @dev: network device 3102 * @queue_index: sub queue index 3103 * 3104 * Check individual transmit queue of a device with multiple transmit queues. 3105 */ 3106 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3107 u16 queue_index) 3108 { 3109 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3110 3111 return netif_tx_queue_stopped(txq); 3112 } 3113 3114 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3115 struct sk_buff *skb) 3116 { 3117 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3118 } 3119 3120 void netif_wake_subqueue(struct net_device *dev, u16 queue_index); 3121 3122 #ifdef CONFIG_XPS 3123 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3124 u16 index); 3125 #else 3126 static inline int netif_set_xps_queue(struct net_device *dev, 3127 const struct cpumask *mask, 3128 u16 index) 3129 { 3130 return 0; 3131 } 3132 #endif 3133 3134 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb, 3135 unsigned int num_tx_queues); 3136 3137 /* 3138 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used 3139 * as a distribution range limit for the returned value. 3140 */ 3141 static inline u16 skb_tx_hash(const struct net_device *dev, 3142 struct sk_buff *skb) 3143 { 3144 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues); 3145 } 3146 3147 /** 3148 * netif_is_multiqueue - test if device has multiple transmit queues 3149 * @dev: network device 3150 * 3151 * Check if device has multiple transmit queues 3152 */ 3153 static inline bool netif_is_multiqueue(const struct net_device *dev) 3154 { 3155 return dev->num_tx_queues > 1; 3156 } 3157 3158 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3159 3160 #ifdef CONFIG_SYSFS 3161 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3162 #else 3163 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3164 unsigned int rxq) 3165 { 3166 return 0; 3167 } 3168 #endif 3169 3170 #ifdef CONFIG_SYSFS 3171 static inline unsigned int get_netdev_rx_queue_index( 3172 struct netdev_rx_queue *queue) 3173 { 3174 struct net_device *dev = queue->dev; 3175 int index = queue - dev->_rx; 3176 3177 BUG_ON(index >= dev->num_rx_queues); 3178 return index; 3179 } 3180 #endif 3181 3182 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3183 int netif_get_num_default_rss_queues(void); 3184 3185 enum skb_free_reason { 3186 SKB_REASON_CONSUMED, 3187 SKB_REASON_DROPPED, 3188 }; 3189 3190 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3191 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3192 3193 /* 3194 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3195 * interrupt context or with hardware interrupts being disabled. 3196 * (in_irq() || irqs_disabled()) 3197 * 3198 * We provide four helpers that can be used in following contexts : 3199 * 3200 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3201 * replacing kfree_skb(skb) 3202 * 3203 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3204 * Typically used in place of consume_skb(skb) in TX completion path 3205 * 3206 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3207 * replacing kfree_skb(skb) 3208 * 3209 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3210 * and consumed a packet. Used in place of consume_skb(skb) 3211 */ 3212 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3213 { 3214 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3215 } 3216 3217 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3218 { 3219 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3220 } 3221 3222 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3223 { 3224 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3225 } 3226 3227 static inline void dev_consume_skb_any(struct sk_buff *skb) 3228 { 3229 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3230 } 3231 3232 int netif_rx(struct sk_buff *skb); 3233 int netif_rx_ni(struct sk_buff *skb); 3234 int netif_receive_skb(struct sk_buff *skb); 3235 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3236 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3237 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3238 gro_result_t napi_gro_frags(struct napi_struct *napi); 3239 struct packet_offload *gro_find_receive_by_type(__be16 type); 3240 struct packet_offload *gro_find_complete_by_type(__be16 type); 3241 3242 static inline void napi_free_frags(struct napi_struct *napi) 3243 { 3244 kfree_skb(napi->skb); 3245 napi->skb = NULL; 3246 } 3247 3248 int netdev_rx_handler_register(struct net_device *dev, 3249 rx_handler_func_t *rx_handler, 3250 void *rx_handler_data); 3251 void netdev_rx_handler_unregister(struct net_device *dev); 3252 3253 bool dev_valid_name(const char *name); 3254 int dev_ioctl(struct net *net, unsigned int cmd, void __user *); 3255 int dev_ethtool(struct net *net, struct ifreq *); 3256 unsigned int dev_get_flags(const struct net_device *); 3257 int __dev_change_flags(struct net_device *, unsigned int flags); 3258 int dev_change_flags(struct net_device *, unsigned int); 3259 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3260 unsigned int gchanges); 3261 int dev_change_name(struct net_device *, const char *); 3262 int dev_set_alias(struct net_device *, const char *, size_t); 3263 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3264 int dev_set_mtu(struct net_device *, int); 3265 void dev_set_group(struct net_device *, int); 3266 int dev_set_mac_address(struct net_device *, struct sockaddr *); 3267 int dev_change_carrier(struct net_device *, bool new_carrier); 3268 int dev_get_phys_port_id(struct net_device *dev, 3269 struct netdev_phys_item_id *ppid); 3270 int dev_get_phys_port_name(struct net_device *dev, 3271 char *name, size_t len); 3272 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3273 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev); 3274 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3275 struct netdev_queue *txq, int *ret); 3276 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3277 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3278 bool is_skb_forwardable(const struct net_device *dev, 3279 const struct sk_buff *skb); 3280 3281 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3282 3283 extern int netdev_budget; 3284 3285 /* Called by rtnetlink.c:rtnl_unlock() */ 3286 void netdev_run_todo(void); 3287 3288 /** 3289 * dev_put - release reference to device 3290 * @dev: network device 3291 * 3292 * Release reference to device to allow it to be freed. 3293 */ 3294 static inline void dev_put(struct net_device *dev) 3295 { 3296 this_cpu_dec(*dev->pcpu_refcnt); 3297 } 3298 3299 /** 3300 * dev_hold - get reference to device 3301 * @dev: network device 3302 * 3303 * Hold reference to device to keep it from being freed. 3304 */ 3305 static inline void dev_hold(struct net_device *dev) 3306 { 3307 this_cpu_inc(*dev->pcpu_refcnt); 3308 } 3309 3310 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3311 * and _off may be called from IRQ context, but it is caller 3312 * who is responsible for serialization of these calls. 3313 * 3314 * The name carrier is inappropriate, these functions should really be 3315 * called netif_lowerlayer_*() because they represent the state of any 3316 * kind of lower layer not just hardware media. 3317 */ 3318 3319 void linkwatch_init_dev(struct net_device *dev); 3320 void linkwatch_fire_event(struct net_device *dev); 3321 void linkwatch_forget_dev(struct net_device *dev); 3322 3323 /** 3324 * netif_carrier_ok - test if carrier present 3325 * @dev: network device 3326 * 3327 * Check if carrier is present on device 3328 */ 3329 static inline bool netif_carrier_ok(const struct net_device *dev) 3330 { 3331 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3332 } 3333 3334 unsigned long dev_trans_start(struct net_device *dev); 3335 3336 void __netdev_watchdog_up(struct net_device *dev); 3337 3338 void netif_carrier_on(struct net_device *dev); 3339 3340 void netif_carrier_off(struct net_device *dev); 3341 3342 /** 3343 * netif_dormant_on - mark device as dormant. 3344 * @dev: network device 3345 * 3346 * Mark device as dormant (as per RFC2863). 3347 * 3348 * The dormant state indicates that the relevant interface is not 3349 * actually in a condition to pass packets (i.e., it is not 'up') but is 3350 * in a "pending" state, waiting for some external event. For "on- 3351 * demand" interfaces, this new state identifies the situation where the 3352 * interface is waiting for events to place it in the up state. 3353 */ 3354 static inline void netif_dormant_on(struct net_device *dev) 3355 { 3356 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3357 linkwatch_fire_event(dev); 3358 } 3359 3360 /** 3361 * netif_dormant_off - set device as not dormant. 3362 * @dev: network device 3363 * 3364 * Device is not in dormant state. 3365 */ 3366 static inline void netif_dormant_off(struct net_device *dev) 3367 { 3368 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3369 linkwatch_fire_event(dev); 3370 } 3371 3372 /** 3373 * netif_dormant - test if carrier present 3374 * @dev: network device 3375 * 3376 * Check if carrier is present on device 3377 */ 3378 static inline bool netif_dormant(const struct net_device *dev) 3379 { 3380 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3381 } 3382 3383 3384 /** 3385 * netif_oper_up - test if device is operational 3386 * @dev: network device 3387 * 3388 * Check if carrier is operational 3389 */ 3390 static inline bool netif_oper_up(const struct net_device *dev) 3391 { 3392 return (dev->operstate == IF_OPER_UP || 3393 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3394 } 3395 3396 /** 3397 * netif_device_present - is device available or removed 3398 * @dev: network device 3399 * 3400 * Check if device has not been removed from system. 3401 */ 3402 static inline bool netif_device_present(struct net_device *dev) 3403 { 3404 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3405 } 3406 3407 void netif_device_detach(struct net_device *dev); 3408 3409 void netif_device_attach(struct net_device *dev); 3410 3411 /* 3412 * Network interface message level settings 3413 */ 3414 3415 enum { 3416 NETIF_MSG_DRV = 0x0001, 3417 NETIF_MSG_PROBE = 0x0002, 3418 NETIF_MSG_LINK = 0x0004, 3419 NETIF_MSG_TIMER = 0x0008, 3420 NETIF_MSG_IFDOWN = 0x0010, 3421 NETIF_MSG_IFUP = 0x0020, 3422 NETIF_MSG_RX_ERR = 0x0040, 3423 NETIF_MSG_TX_ERR = 0x0080, 3424 NETIF_MSG_TX_QUEUED = 0x0100, 3425 NETIF_MSG_INTR = 0x0200, 3426 NETIF_MSG_TX_DONE = 0x0400, 3427 NETIF_MSG_RX_STATUS = 0x0800, 3428 NETIF_MSG_PKTDATA = 0x1000, 3429 NETIF_MSG_HW = 0x2000, 3430 NETIF_MSG_WOL = 0x4000, 3431 }; 3432 3433 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3434 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3435 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3436 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3437 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3438 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3439 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3440 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3441 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3442 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3443 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3444 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3445 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3446 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3447 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3448 3449 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3450 { 3451 /* use default */ 3452 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3453 return default_msg_enable_bits; 3454 if (debug_value == 0) /* no output */ 3455 return 0; 3456 /* set low N bits */ 3457 return (1 << debug_value) - 1; 3458 } 3459 3460 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 3461 { 3462 spin_lock(&txq->_xmit_lock); 3463 txq->xmit_lock_owner = cpu; 3464 } 3465 3466 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 3467 { 3468 spin_lock_bh(&txq->_xmit_lock); 3469 txq->xmit_lock_owner = smp_processor_id(); 3470 } 3471 3472 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 3473 { 3474 bool ok = spin_trylock(&txq->_xmit_lock); 3475 if (likely(ok)) 3476 txq->xmit_lock_owner = smp_processor_id(); 3477 return ok; 3478 } 3479 3480 static inline void __netif_tx_unlock(struct netdev_queue *txq) 3481 { 3482 txq->xmit_lock_owner = -1; 3483 spin_unlock(&txq->_xmit_lock); 3484 } 3485 3486 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 3487 { 3488 txq->xmit_lock_owner = -1; 3489 spin_unlock_bh(&txq->_xmit_lock); 3490 } 3491 3492 static inline void txq_trans_update(struct netdev_queue *txq) 3493 { 3494 if (txq->xmit_lock_owner != -1) 3495 txq->trans_start = jiffies; 3496 } 3497 3498 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 3499 static inline void netif_trans_update(struct net_device *dev) 3500 { 3501 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 3502 3503 if (txq->trans_start != jiffies) 3504 txq->trans_start = jiffies; 3505 } 3506 3507 /** 3508 * netif_tx_lock - grab network device transmit lock 3509 * @dev: network device 3510 * 3511 * Get network device transmit lock 3512 */ 3513 static inline void netif_tx_lock(struct net_device *dev) 3514 { 3515 unsigned int i; 3516 int cpu; 3517 3518 spin_lock(&dev->tx_global_lock); 3519 cpu = smp_processor_id(); 3520 for (i = 0; i < dev->num_tx_queues; i++) { 3521 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3522 3523 /* We are the only thread of execution doing a 3524 * freeze, but we have to grab the _xmit_lock in 3525 * order to synchronize with threads which are in 3526 * the ->hard_start_xmit() handler and already 3527 * checked the frozen bit. 3528 */ 3529 __netif_tx_lock(txq, cpu); 3530 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 3531 __netif_tx_unlock(txq); 3532 } 3533 } 3534 3535 static inline void netif_tx_lock_bh(struct net_device *dev) 3536 { 3537 local_bh_disable(); 3538 netif_tx_lock(dev); 3539 } 3540 3541 static inline void netif_tx_unlock(struct net_device *dev) 3542 { 3543 unsigned int i; 3544 3545 for (i = 0; i < dev->num_tx_queues; i++) { 3546 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3547 3548 /* No need to grab the _xmit_lock here. If the 3549 * queue is not stopped for another reason, we 3550 * force a schedule. 3551 */ 3552 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 3553 netif_schedule_queue(txq); 3554 } 3555 spin_unlock(&dev->tx_global_lock); 3556 } 3557 3558 static inline void netif_tx_unlock_bh(struct net_device *dev) 3559 { 3560 netif_tx_unlock(dev); 3561 local_bh_enable(); 3562 } 3563 3564 #define HARD_TX_LOCK(dev, txq, cpu) { \ 3565 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3566 __netif_tx_lock(txq, cpu); \ 3567 } \ 3568 } 3569 3570 #define HARD_TX_TRYLOCK(dev, txq) \ 3571 (((dev->features & NETIF_F_LLTX) == 0) ? \ 3572 __netif_tx_trylock(txq) : \ 3573 true ) 3574 3575 #define HARD_TX_UNLOCK(dev, txq) { \ 3576 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3577 __netif_tx_unlock(txq); \ 3578 } \ 3579 } 3580 3581 static inline void netif_tx_disable(struct net_device *dev) 3582 { 3583 unsigned int i; 3584 int cpu; 3585 3586 local_bh_disable(); 3587 cpu = smp_processor_id(); 3588 for (i = 0; i < dev->num_tx_queues; i++) { 3589 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3590 3591 __netif_tx_lock(txq, cpu); 3592 netif_tx_stop_queue(txq); 3593 __netif_tx_unlock(txq); 3594 } 3595 local_bh_enable(); 3596 } 3597 3598 static inline void netif_addr_lock(struct net_device *dev) 3599 { 3600 spin_lock(&dev->addr_list_lock); 3601 } 3602 3603 static inline void netif_addr_lock_nested(struct net_device *dev) 3604 { 3605 int subclass = SINGLE_DEPTH_NESTING; 3606 3607 if (dev->netdev_ops->ndo_get_lock_subclass) 3608 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev); 3609 3610 spin_lock_nested(&dev->addr_list_lock, subclass); 3611 } 3612 3613 static inline void netif_addr_lock_bh(struct net_device *dev) 3614 { 3615 spin_lock_bh(&dev->addr_list_lock); 3616 } 3617 3618 static inline void netif_addr_unlock(struct net_device *dev) 3619 { 3620 spin_unlock(&dev->addr_list_lock); 3621 } 3622 3623 static inline void netif_addr_unlock_bh(struct net_device *dev) 3624 { 3625 spin_unlock_bh(&dev->addr_list_lock); 3626 } 3627 3628 /* 3629 * dev_addrs walker. Should be used only for read access. Call with 3630 * rcu_read_lock held. 3631 */ 3632 #define for_each_dev_addr(dev, ha) \ 3633 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 3634 3635 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 3636 3637 void ether_setup(struct net_device *dev); 3638 3639 /* Support for loadable net-drivers */ 3640 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 3641 unsigned char name_assign_type, 3642 void (*setup)(struct net_device *), 3643 unsigned int txqs, unsigned int rxqs); 3644 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 3645 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 3646 3647 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 3648 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 3649 count) 3650 3651 int register_netdev(struct net_device *dev); 3652 void unregister_netdev(struct net_device *dev); 3653 3654 /* General hardware address lists handling functions */ 3655 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 3656 struct netdev_hw_addr_list *from_list, int addr_len); 3657 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 3658 struct netdev_hw_addr_list *from_list, int addr_len); 3659 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 3660 struct net_device *dev, 3661 int (*sync)(struct net_device *, const unsigned char *), 3662 int (*unsync)(struct net_device *, 3663 const unsigned char *)); 3664 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 3665 struct net_device *dev, 3666 int (*unsync)(struct net_device *, 3667 const unsigned char *)); 3668 void __hw_addr_init(struct netdev_hw_addr_list *list); 3669 3670 /* Functions used for device addresses handling */ 3671 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 3672 unsigned char addr_type); 3673 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 3674 unsigned char addr_type); 3675 void dev_addr_flush(struct net_device *dev); 3676 int dev_addr_init(struct net_device *dev); 3677 3678 /* Functions used for unicast addresses handling */ 3679 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 3680 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 3681 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 3682 int dev_uc_sync(struct net_device *to, struct net_device *from); 3683 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 3684 void dev_uc_unsync(struct net_device *to, struct net_device *from); 3685 void dev_uc_flush(struct net_device *dev); 3686 void dev_uc_init(struct net_device *dev); 3687 3688 /** 3689 * __dev_uc_sync - Synchonize device's unicast list 3690 * @dev: device to sync 3691 * @sync: function to call if address should be added 3692 * @unsync: function to call if address should be removed 3693 * 3694 * Add newly added addresses to the interface, and release 3695 * addresses that have been deleted. 3696 */ 3697 static inline int __dev_uc_sync(struct net_device *dev, 3698 int (*sync)(struct net_device *, 3699 const unsigned char *), 3700 int (*unsync)(struct net_device *, 3701 const unsigned char *)) 3702 { 3703 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 3704 } 3705 3706 /** 3707 * __dev_uc_unsync - Remove synchronized addresses from device 3708 * @dev: device to sync 3709 * @unsync: function to call if address should be removed 3710 * 3711 * Remove all addresses that were added to the device by dev_uc_sync(). 3712 */ 3713 static inline void __dev_uc_unsync(struct net_device *dev, 3714 int (*unsync)(struct net_device *, 3715 const unsigned char *)) 3716 { 3717 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 3718 } 3719 3720 /* Functions used for multicast addresses handling */ 3721 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 3722 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 3723 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 3724 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 3725 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 3726 int dev_mc_sync(struct net_device *to, struct net_device *from); 3727 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 3728 void dev_mc_unsync(struct net_device *to, struct net_device *from); 3729 void dev_mc_flush(struct net_device *dev); 3730 void dev_mc_init(struct net_device *dev); 3731 3732 /** 3733 * __dev_mc_sync - Synchonize device's multicast list 3734 * @dev: device to sync 3735 * @sync: function to call if address should be added 3736 * @unsync: function to call if address should be removed 3737 * 3738 * Add newly added addresses to the interface, and release 3739 * addresses that have been deleted. 3740 */ 3741 static inline int __dev_mc_sync(struct net_device *dev, 3742 int (*sync)(struct net_device *, 3743 const unsigned char *), 3744 int (*unsync)(struct net_device *, 3745 const unsigned char *)) 3746 { 3747 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 3748 } 3749 3750 /** 3751 * __dev_mc_unsync - Remove synchronized addresses from device 3752 * @dev: device to sync 3753 * @unsync: function to call if address should be removed 3754 * 3755 * Remove all addresses that were added to the device by dev_mc_sync(). 3756 */ 3757 static inline void __dev_mc_unsync(struct net_device *dev, 3758 int (*unsync)(struct net_device *, 3759 const unsigned char *)) 3760 { 3761 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 3762 } 3763 3764 /* Functions used for secondary unicast and multicast support */ 3765 void dev_set_rx_mode(struct net_device *dev); 3766 void __dev_set_rx_mode(struct net_device *dev); 3767 int dev_set_promiscuity(struct net_device *dev, int inc); 3768 int dev_set_allmulti(struct net_device *dev, int inc); 3769 void netdev_state_change(struct net_device *dev); 3770 void netdev_notify_peers(struct net_device *dev); 3771 void netdev_features_change(struct net_device *dev); 3772 /* Load a device via the kmod */ 3773 void dev_load(struct net *net, const char *name); 3774 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 3775 struct rtnl_link_stats64 *storage); 3776 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 3777 const struct net_device_stats *netdev_stats); 3778 3779 extern int netdev_max_backlog; 3780 extern int netdev_tstamp_prequeue; 3781 extern int weight_p; 3782 3783 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 3784 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 3785 struct list_head **iter); 3786 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 3787 struct list_head **iter); 3788 3789 /* iterate through upper list, must be called under RCU read lock */ 3790 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 3791 for (iter = &(dev)->adj_list.upper, \ 3792 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 3793 updev; \ 3794 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 3795 3796 /* iterate through upper list, must be called under RCU read lock */ 3797 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \ 3798 for (iter = &(dev)->all_adj_list.upper, \ 3799 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \ 3800 updev; \ 3801 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter))) 3802 3803 void *netdev_lower_get_next_private(struct net_device *dev, 3804 struct list_head **iter); 3805 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 3806 struct list_head **iter); 3807 3808 #define netdev_for_each_lower_private(dev, priv, iter) \ 3809 for (iter = (dev)->adj_list.lower.next, \ 3810 priv = netdev_lower_get_next_private(dev, &(iter)); \ 3811 priv; \ 3812 priv = netdev_lower_get_next_private(dev, &(iter))) 3813 3814 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 3815 for (iter = &(dev)->adj_list.lower, \ 3816 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 3817 priv; \ 3818 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 3819 3820 void *netdev_lower_get_next(struct net_device *dev, 3821 struct list_head **iter); 3822 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 3823 for (iter = (dev)->adj_list.lower.next, \ 3824 ldev = netdev_lower_get_next(dev, &(iter)); \ 3825 ldev; \ 3826 ldev = netdev_lower_get_next(dev, &(iter))) 3827 3828 void *netdev_adjacent_get_private(struct list_head *adj_list); 3829 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 3830 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 3831 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 3832 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev); 3833 int netdev_master_upper_dev_link(struct net_device *dev, 3834 struct net_device *upper_dev, 3835 void *upper_priv, void *upper_info); 3836 void netdev_upper_dev_unlink(struct net_device *dev, 3837 struct net_device *upper_dev); 3838 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 3839 void *netdev_lower_dev_get_private(struct net_device *dev, 3840 struct net_device *lower_dev); 3841 void netdev_lower_state_changed(struct net_device *lower_dev, 3842 void *lower_state_info); 3843 3844 /* RSS keys are 40 or 52 bytes long */ 3845 #define NETDEV_RSS_KEY_LEN 52 3846 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 3847 void netdev_rss_key_fill(void *buffer, size_t len); 3848 3849 int dev_get_nest_level(struct net_device *dev, 3850 bool (*type_check)(const struct net_device *dev)); 3851 int skb_checksum_help(struct sk_buff *skb); 3852 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 3853 netdev_features_t features, bool tx_path); 3854 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 3855 netdev_features_t features); 3856 3857 struct netdev_bonding_info { 3858 ifslave slave; 3859 ifbond master; 3860 }; 3861 3862 struct netdev_notifier_bonding_info { 3863 struct netdev_notifier_info info; /* must be first */ 3864 struct netdev_bonding_info bonding_info; 3865 }; 3866 3867 void netdev_bonding_info_change(struct net_device *dev, 3868 struct netdev_bonding_info *bonding_info); 3869 3870 static inline 3871 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 3872 { 3873 return __skb_gso_segment(skb, features, true); 3874 } 3875 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 3876 3877 static inline bool can_checksum_protocol(netdev_features_t features, 3878 __be16 protocol) 3879 { 3880 if (protocol == htons(ETH_P_FCOE)) 3881 return !!(features & NETIF_F_FCOE_CRC); 3882 3883 /* Assume this is an IP checksum (not SCTP CRC) */ 3884 3885 if (features & NETIF_F_HW_CSUM) { 3886 /* Can checksum everything */ 3887 return true; 3888 } 3889 3890 switch (protocol) { 3891 case htons(ETH_P_IP): 3892 return !!(features & NETIF_F_IP_CSUM); 3893 case htons(ETH_P_IPV6): 3894 return !!(features & NETIF_F_IPV6_CSUM); 3895 default: 3896 return false; 3897 } 3898 } 3899 3900 /* Map an ethertype into IP protocol if possible */ 3901 static inline int eproto_to_ipproto(int eproto) 3902 { 3903 switch (eproto) { 3904 case htons(ETH_P_IP): 3905 return IPPROTO_IP; 3906 case htons(ETH_P_IPV6): 3907 return IPPROTO_IPV6; 3908 default: 3909 return -1; 3910 } 3911 } 3912 3913 #ifdef CONFIG_BUG 3914 void netdev_rx_csum_fault(struct net_device *dev); 3915 #else 3916 static inline void netdev_rx_csum_fault(struct net_device *dev) 3917 { 3918 } 3919 #endif 3920 /* rx skb timestamps */ 3921 void net_enable_timestamp(void); 3922 void net_disable_timestamp(void); 3923 3924 #ifdef CONFIG_PROC_FS 3925 int __init dev_proc_init(void); 3926 #else 3927 #define dev_proc_init() 0 3928 #endif 3929 3930 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 3931 struct sk_buff *skb, struct net_device *dev, 3932 bool more) 3933 { 3934 skb->xmit_more = more ? 1 : 0; 3935 return ops->ndo_start_xmit(skb, dev); 3936 } 3937 3938 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 3939 struct netdev_queue *txq, bool more) 3940 { 3941 const struct net_device_ops *ops = dev->netdev_ops; 3942 int rc; 3943 3944 rc = __netdev_start_xmit(ops, skb, dev, more); 3945 if (rc == NETDEV_TX_OK) 3946 txq_trans_update(txq); 3947 3948 return rc; 3949 } 3950 3951 int netdev_class_create_file_ns(struct class_attribute *class_attr, 3952 const void *ns); 3953 void netdev_class_remove_file_ns(struct class_attribute *class_attr, 3954 const void *ns); 3955 3956 static inline int netdev_class_create_file(struct class_attribute *class_attr) 3957 { 3958 return netdev_class_create_file_ns(class_attr, NULL); 3959 } 3960 3961 static inline void netdev_class_remove_file(struct class_attribute *class_attr) 3962 { 3963 netdev_class_remove_file_ns(class_attr, NULL); 3964 } 3965 3966 extern struct kobj_ns_type_operations net_ns_type_operations; 3967 3968 const char *netdev_drivername(const struct net_device *dev); 3969 3970 void linkwatch_run_queue(void); 3971 3972 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 3973 netdev_features_t f2) 3974 { 3975 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 3976 if (f1 & NETIF_F_HW_CSUM) 3977 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 3978 else 3979 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 3980 } 3981 3982 return f1 & f2; 3983 } 3984 3985 static inline netdev_features_t netdev_get_wanted_features( 3986 struct net_device *dev) 3987 { 3988 return (dev->features & ~dev->hw_features) | dev->wanted_features; 3989 } 3990 netdev_features_t netdev_increment_features(netdev_features_t all, 3991 netdev_features_t one, netdev_features_t mask); 3992 3993 /* Allow TSO being used on stacked device : 3994 * Performing the GSO segmentation before last device 3995 * is a performance improvement. 3996 */ 3997 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 3998 netdev_features_t mask) 3999 { 4000 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4001 } 4002 4003 int __netdev_update_features(struct net_device *dev); 4004 void netdev_update_features(struct net_device *dev); 4005 void netdev_change_features(struct net_device *dev); 4006 4007 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4008 struct net_device *dev); 4009 4010 netdev_features_t passthru_features_check(struct sk_buff *skb, 4011 struct net_device *dev, 4012 netdev_features_t features); 4013 netdev_features_t netif_skb_features(struct sk_buff *skb); 4014 4015 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4016 { 4017 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4018 4019 /* check flags correspondence */ 4020 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4021 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT)); 4022 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4023 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4024 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4025 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4026 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4027 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4028 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4029 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4030 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4031 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4032 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4033 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4034 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4035 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4036 4037 return (features & feature) == feature; 4038 } 4039 4040 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4041 { 4042 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4043 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4044 } 4045 4046 static inline bool netif_needs_gso(struct sk_buff *skb, 4047 netdev_features_t features) 4048 { 4049 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4050 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4051 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4052 } 4053 4054 static inline void netif_set_gso_max_size(struct net_device *dev, 4055 unsigned int size) 4056 { 4057 dev->gso_max_size = size; 4058 } 4059 4060 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4061 int pulled_hlen, u16 mac_offset, 4062 int mac_len) 4063 { 4064 skb->protocol = protocol; 4065 skb->encapsulation = 1; 4066 skb_push(skb, pulled_hlen); 4067 skb_reset_transport_header(skb); 4068 skb->mac_header = mac_offset; 4069 skb->network_header = skb->mac_header + mac_len; 4070 skb->mac_len = mac_len; 4071 } 4072 4073 static inline bool netif_is_macsec(const struct net_device *dev) 4074 { 4075 return dev->priv_flags & IFF_MACSEC; 4076 } 4077 4078 static inline bool netif_is_macvlan(const struct net_device *dev) 4079 { 4080 return dev->priv_flags & IFF_MACVLAN; 4081 } 4082 4083 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4084 { 4085 return dev->priv_flags & IFF_MACVLAN_PORT; 4086 } 4087 4088 static inline bool netif_is_ipvlan(const struct net_device *dev) 4089 { 4090 return dev->priv_flags & IFF_IPVLAN_SLAVE; 4091 } 4092 4093 static inline bool netif_is_ipvlan_port(const struct net_device *dev) 4094 { 4095 return dev->priv_flags & IFF_IPVLAN_MASTER; 4096 } 4097 4098 static inline bool netif_is_bond_master(const struct net_device *dev) 4099 { 4100 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4101 } 4102 4103 static inline bool netif_is_bond_slave(const struct net_device *dev) 4104 { 4105 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4106 } 4107 4108 static inline bool netif_supports_nofcs(struct net_device *dev) 4109 { 4110 return dev->priv_flags & IFF_SUPP_NOFCS; 4111 } 4112 4113 static inline bool netif_is_l3_master(const struct net_device *dev) 4114 { 4115 return dev->priv_flags & IFF_L3MDEV_MASTER; 4116 } 4117 4118 static inline bool netif_is_l3_slave(const struct net_device *dev) 4119 { 4120 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4121 } 4122 4123 static inline bool netif_is_bridge_master(const struct net_device *dev) 4124 { 4125 return dev->priv_flags & IFF_EBRIDGE; 4126 } 4127 4128 static inline bool netif_is_bridge_port(const struct net_device *dev) 4129 { 4130 return dev->priv_flags & IFF_BRIDGE_PORT; 4131 } 4132 4133 static inline bool netif_is_ovs_master(const struct net_device *dev) 4134 { 4135 return dev->priv_flags & IFF_OPENVSWITCH; 4136 } 4137 4138 static inline bool netif_is_team_master(const struct net_device *dev) 4139 { 4140 return dev->priv_flags & IFF_TEAM; 4141 } 4142 4143 static inline bool netif_is_team_port(const struct net_device *dev) 4144 { 4145 return dev->priv_flags & IFF_TEAM_PORT; 4146 } 4147 4148 static inline bool netif_is_lag_master(const struct net_device *dev) 4149 { 4150 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4151 } 4152 4153 static inline bool netif_is_lag_port(const struct net_device *dev) 4154 { 4155 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4156 } 4157 4158 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4159 { 4160 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4161 } 4162 4163 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4164 static inline void netif_keep_dst(struct net_device *dev) 4165 { 4166 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4167 } 4168 4169 extern struct pernet_operations __net_initdata loopback_net_ops; 4170 4171 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4172 4173 /* netdev_printk helpers, similar to dev_printk */ 4174 4175 static inline const char *netdev_name(const struct net_device *dev) 4176 { 4177 if (!dev->name[0] || strchr(dev->name, '%')) 4178 return "(unnamed net_device)"; 4179 return dev->name; 4180 } 4181 4182 static inline const char *netdev_reg_state(const struct net_device *dev) 4183 { 4184 switch (dev->reg_state) { 4185 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4186 case NETREG_REGISTERED: return ""; 4187 case NETREG_UNREGISTERING: return " (unregistering)"; 4188 case NETREG_UNREGISTERED: return " (unregistered)"; 4189 case NETREG_RELEASED: return " (released)"; 4190 case NETREG_DUMMY: return " (dummy)"; 4191 } 4192 4193 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4194 return " (unknown)"; 4195 } 4196 4197 __printf(3, 4) 4198 void netdev_printk(const char *level, const struct net_device *dev, 4199 const char *format, ...); 4200 __printf(2, 3) 4201 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4202 __printf(2, 3) 4203 void netdev_alert(const struct net_device *dev, const char *format, ...); 4204 __printf(2, 3) 4205 void netdev_crit(const struct net_device *dev, const char *format, ...); 4206 __printf(2, 3) 4207 void netdev_err(const struct net_device *dev, const char *format, ...); 4208 __printf(2, 3) 4209 void netdev_warn(const struct net_device *dev, const char *format, ...); 4210 __printf(2, 3) 4211 void netdev_notice(const struct net_device *dev, const char *format, ...); 4212 __printf(2, 3) 4213 void netdev_info(const struct net_device *dev, const char *format, ...); 4214 4215 #define MODULE_ALIAS_NETDEV(device) \ 4216 MODULE_ALIAS("netdev-" device) 4217 4218 #if defined(CONFIG_DYNAMIC_DEBUG) 4219 #define netdev_dbg(__dev, format, args...) \ 4220 do { \ 4221 dynamic_netdev_dbg(__dev, format, ##args); \ 4222 } while (0) 4223 #elif defined(DEBUG) 4224 #define netdev_dbg(__dev, format, args...) \ 4225 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4226 #else 4227 #define netdev_dbg(__dev, format, args...) \ 4228 ({ \ 4229 if (0) \ 4230 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4231 }) 4232 #endif 4233 4234 #if defined(VERBOSE_DEBUG) 4235 #define netdev_vdbg netdev_dbg 4236 #else 4237 4238 #define netdev_vdbg(dev, format, args...) \ 4239 ({ \ 4240 if (0) \ 4241 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 4242 0; \ 4243 }) 4244 #endif 4245 4246 /* 4247 * netdev_WARN() acts like dev_printk(), but with the key difference 4248 * of using a WARN/WARN_ON to get the message out, including the 4249 * file/line information and a backtrace. 4250 */ 4251 #define netdev_WARN(dev, format, args...) \ 4252 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \ 4253 netdev_reg_state(dev), ##args) 4254 4255 /* netif printk helpers, similar to netdev_printk */ 4256 4257 #define netif_printk(priv, type, level, dev, fmt, args...) \ 4258 do { \ 4259 if (netif_msg_##type(priv)) \ 4260 netdev_printk(level, (dev), fmt, ##args); \ 4261 } while (0) 4262 4263 #define netif_level(level, priv, type, dev, fmt, args...) \ 4264 do { \ 4265 if (netif_msg_##type(priv)) \ 4266 netdev_##level(dev, fmt, ##args); \ 4267 } while (0) 4268 4269 #define netif_emerg(priv, type, dev, fmt, args...) \ 4270 netif_level(emerg, priv, type, dev, fmt, ##args) 4271 #define netif_alert(priv, type, dev, fmt, args...) \ 4272 netif_level(alert, priv, type, dev, fmt, ##args) 4273 #define netif_crit(priv, type, dev, fmt, args...) \ 4274 netif_level(crit, priv, type, dev, fmt, ##args) 4275 #define netif_err(priv, type, dev, fmt, args...) \ 4276 netif_level(err, priv, type, dev, fmt, ##args) 4277 #define netif_warn(priv, type, dev, fmt, args...) \ 4278 netif_level(warn, priv, type, dev, fmt, ##args) 4279 #define netif_notice(priv, type, dev, fmt, args...) \ 4280 netif_level(notice, priv, type, dev, fmt, ##args) 4281 #define netif_info(priv, type, dev, fmt, args...) \ 4282 netif_level(info, priv, type, dev, fmt, ##args) 4283 4284 #if defined(CONFIG_DYNAMIC_DEBUG) 4285 #define netif_dbg(priv, type, netdev, format, args...) \ 4286 do { \ 4287 if (netif_msg_##type(priv)) \ 4288 dynamic_netdev_dbg(netdev, format, ##args); \ 4289 } while (0) 4290 #elif defined(DEBUG) 4291 #define netif_dbg(priv, type, dev, format, args...) \ 4292 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 4293 #else 4294 #define netif_dbg(priv, type, dev, format, args...) \ 4295 ({ \ 4296 if (0) \ 4297 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4298 0; \ 4299 }) 4300 #endif 4301 4302 #if defined(VERBOSE_DEBUG) 4303 #define netif_vdbg netif_dbg 4304 #else 4305 #define netif_vdbg(priv, type, dev, format, args...) \ 4306 ({ \ 4307 if (0) \ 4308 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4309 0; \ 4310 }) 4311 #endif 4312 4313 /* 4314 * The list of packet types we will receive (as opposed to discard) 4315 * and the routines to invoke. 4316 * 4317 * Why 16. Because with 16 the only overlap we get on a hash of the 4318 * low nibble of the protocol value is RARP/SNAP/X.25. 4319 * 4320 * NOTE: That is no longer true with the addition of VLAN tags. Not 4321 * sure which should go first, but I bet it won't make much 4322 * difference if we are running VLANs. The good news is that 4323 * this protocol won't be in the list unless compiled in, so 4324 * the average user (w/out VLANs) will not be adversely affected. 4325 * --BLG 4326 * 4327 * 0800 IP 4328 * 8100 802.1Q VLAN 4329 * 0001 802.3 4330 * 0002 AX.25 4331 * 0004 802.2 4332 * 8035 RARP 4333 * 0005 SNAP 4334 * 0805 X.25 4335 * 0806 ARP 4336 * 8137 IPX 4337 * 0009 Localtalk 4338 * 86DD IPv6 4339 */ 4340 #define PTYPE_HASH_SIZE (16) 4341 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 4342 4343 #endif /* _LINUX_NETDEVICE_H */ 4344