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