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/pm_qos.h> 29 #include <linux/timer.h> 30 #include <linux/bug.h> 31 #include <linux/delay.h> 32 #include <linux/atomic.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 54 struct netpoll_info; 55 struct device; 56 struct phy_device; 57 /* 802.11 specific */ 58 struct wireless_dev; 59 /* source back-compat hooks */ 60 #define SET_ETHTOOL_OPS(netdev,ops) \ 61 ( (netdev)->ethtool_ops = (ops) ) 62 63 void netdev_set_default_ethtool_ops(struct net_device *dev, 64 const struct ethtool_ops *ops); 65 66 /* Backlog congestion levels */ 67 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 68 #define NET_RX_DROP 1 /* packet dropped */ 69 70 /* 71 * Transmit return codes: transmit return codes originate from three different 72 * namespaces: 73 * 74 * - qdisc return codes 75 * - driver transmit return codes 76 * - errno values 77 * 78 * Drivers are allowed to return any one of those in their hard_start_xmit() 79 * function. Real network devices commonly used with qdiscs should only return 80 * the driver transmit return codes though - when qdiscs are used, the actual 81 * transmission happens asynchronously, so the value is not propagated to 82 * higher layers. Virtual network devices transmit synchronously, in this case 83 * the driver transmit return codes are consumed by dev_queue_xmit(), all 84 * others are propagated to higher layers. 85 */ 86 87 /* qdisc ->enqueue() return codes. */ 88 #define NET_XMIT_SUCCESS 0x00 89 #define NET_XMIT_DROP 0x01 /* skb dropped */ 90 #define NET_XMIT_CN 0x02 /* congestion notification */ 91 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */ 92 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 93 94 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 95 * indicates that the device will soon be dropping packets, or already drops 96 * some packets of the same priority; prompting us to send less aggressively. */ 97 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 98 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 99 100 /* Driver transmit return codes */ 101 #define NETDEV_TX_MASK 0xf0 102 103 enum netdev_tx { 104 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 105 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 106 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 107 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */ 108 }; 109 typedef enum netdev_tx netdev_tx_t; 110 111 /* 112 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 113 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 114 */ 115 static inline bool dev_xmit_complete(int rc) 116 { 117 /* 118 * Positive cases with an skb consumed by a driver: 119 * - successful transmission (rc == NETDEV_TX_OK) 120 * - error while transmitting (rc < 0) 121 * - error while queueing to a different device (rc & NET_XMIT_MASK) 122 */ 123 if (likely(rc < NET_XMIT_MASK)) 124 return true; 125 126 return false; 127 } 128 129 /* 130 * Compute the worst case header length according to the protocols 131 * used. 132 */ 133 134 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 135 # if defined(CONFIG_MAC80211_MESH) 136 # define LL_MAX_HEADER 128 137 # else 138 # define LL_MAX_HEADER 96 139 # endif 140 #else 141 # define LL_MAX_HEADER 32 142 #endif 143 144 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 145 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 146 #define MAX_HEADER LL_MAX_HEADER 147 #else 148 #define MAX_HEADER (LL_MAX_HEADER + 48) 149 #endif 150 151 /* 152 * Old network device statistics. Fields are native words 153 * (unsigned long) so they can be read and written atomically. 154 */ 155 156 struct net_device_stats { 157 unsigned long rx_packets; 158 unsigned long tx_packets; 159 unsigned long rx_bytes; 160 unsigned long tx_bytes; 161 unsigned long rx_errors; 162 unsigned long tx_errors; 163 unsigned long rx_dropped; 164 unsigned long tx_dropped; 165 unsigned long multicast; 166 unsigned long collisions; 167 unsigned long rx_length_errors; 168 unsigned long rx_over_errors; 169 unsigned long rx_crc_errors; 170 unsigned long rx_frame_errors; 171 unsigned long rx_fifo_errors; 172 unsigned long rx_missed_errors; 173 unsigned long tx_aborted_errors; 174 unsigned long tx_carrier_errors; 175 unsigned long tx_fifo_errors; 176 unsigned long tx_heartbeat_errors; 177 unsigned long tx_window_errors; 178 unsigned long rx_compressed; 179 unsigned long tx_compressed; 180 }; 181 182 183 #include <linux/cache.h> 184 #include <linux/skbuff.h> 185 186 #ifdef CONFIG_RPS 187 #include <linux/static_key.h> 188 extern struct static_key rps_needed; 189 #endif 190 191 struct neighbour; 192 struct neigh_parms; 193 struct sk_buff; 194 195 struct netdev_hw_addr { 196 struct list_head list; 197 unsigned char addr[MAX_ADDR_LEN]; 198 unsigned char type; 199 #define NETDEV_HW_ADDR_T_LAN 1 200 #define NETDEV_HW_ADDR_T_SAN 2 201 #define NETDEV_HW_ADDR_T_SLAVE 3 202 #define NETDEV_HW_ADDR_T_UNICAST 4 203 #define NETDEV_HW_ADDR_T_MULTICAST 5 204 bool global_use; 205 int sync_cnt; 206 int refcount; 207 int synced; 208 struct rcu_head rcu_head; 209 }; 210 211 struct netdev_hw_addr_list { 212 struct list_head list; 213 int count; 214 }; 215 216 #define netdev_hw_addr_list_count(l) ((l)->count) 217 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 218 #define netdev_hw_addr_list_for_each(ha, l) \ 219 list_for_each_entry(ha, &(l)->list, list) 220 221 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 222 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 223 #define netdev_for_each_uc_addr(ha, dev) \ 224 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 225 226 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 227 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 228 #define netdev_for_each_mc_addr(ha, dev) \ 229 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 230 231 struct hh_cache { 232 u16 hh_len; 233 u16 __pad; 234 seqlock_t hh_lock; 235 236 /* cached hardware header; allow for machine alignment needs. */ 237 #define HH_DATA_MOD 16 238 #define HH_DATA_OFF(__len) \ 239 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 240 #define HH_DATA_ALIGN(__len) \ 241 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 242 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 243 }; 244 245 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much. 246 * Alternative is: 247 * dev->hard_header_len ? (dev->hard_header_len + 248 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 249 * 250 * We could use other alignment values, but we must maintain the 251 * relationship HH alignment <= LL alignment. 252 */ 253 #define LL_RESERVED_SPACE(dev) \ 254 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 255 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 256 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 257 258 struct header_ops { 259 int (*create) (struct sk_buff *skb, struct net_device *dev, 260 unsigned short type, const void *daddr, 261 const void *saddr, unsigned int len); 262 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 263 int (*rebuild)(struct sk_buff *skb); 264 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 265 void (*cache_update)(struct hh_cache *hh, 266 const struct net_device *dev, 267 const unsigned char *haddr); 268 }; 269 270 /* These flag bits are private to the generic network queueing 271 * layer, they may not be explicitly referenced by any other 272 * code. 273 */ 274 275 enum netdev_state_t { 276 __LINK_STATE_START, 277 __LINK_STATE_PRESENT, 278 __LINK_STATE_NOCARRIER, 279 __LINK_STATE_LINKWATCH_PENDING, 280 __LINK_STATE_DORMANT, 281 }; 282 283 284 /* 285 * This structure holds at boot time configured netdevice settings. They 286 * are then used in the device probing. 287 */ 288 struct netdev_boot_setup { 289 char name[IFNAMSIZ]; 290 struct ifmap map; 291 }; 292 #define NETDEV_BOOT_SETUP_MAX 8 293 294 int __init netdev_boot_setup(char *str); 295 296 /* 297 * Structure for NAPI scheduling similar to tasklet but with weighting 298 */ 299 struct napi_struct { 300 /* The poll_list must only be managed by the entity which 301 * changes the state of the NAPI_STATE_SCHED bit. This means 302 * whoever atomically sets that bit can add this napi_struct 303 * to the per-cpu poll_list, and whoever clears that bit 304 * can remove from the list right before clearing the bit. 305 */ 306 struct list_head poll_list; 307 308 unsigned long state; 309 int weight; 310 unsigned int gro_count; 311 int (*poll)(struct napi_struct *, int); 312 #ifdef CONFIG_NETPOLL 313 spinlock_t poll_lock; 314 int poll_owner; 315 #endif 316 struct net_device *dev; 317 struct sk_buff *gro_list; 318 struct sk_buff *skb; 319 struct list_head dev_list; 320 struct hlist_node napi_hash_node; 321 unsigned int napi_id; 322 }; 323 324 enum { 325 NAPI_STATE_SCHED, /* Poll is scheduled */ 326 NAPI_STATE_DISABLE, /* Disable pending */ 327 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 328 NAPI_STATE_HASHED, /* In NAPI hash */ 329 }; 330 331 enum gro_result { 332 GRO_MERGED, 333 GRO_MERGED_FREE, 334 GRO_HELD, 335 GRO_NORMAL, 336 GRO_DROP, 337 }; 338 typedef enum gro_result gro_result_t; 339 340 /* 341 * enum rx_handler_result - Possible return values for rx_handlers. 342 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 343 * further. 344 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 345 * case skb->dev was changed by rx_handler. 346 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 347 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called. 348 * 349 * rx_handlers are functions called from inside __netif_receive_skb(), to do 350 * special processing of the skb, prior to delivery to protocol handlers. 351 * 352 * Currently, a net_device can only have a single rx_handler registered. Trying 353 * to register a second rx_handler will return -EBUSY. 354 * 355 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 356 * To unregister a rx_handler on a net_device, use 357 * netdev_rx_handler_unregister(). 358 * 359 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 360 * do with the skb. 361 * 362 * If the rx_handler consumed to skb in some way, it should return 363 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 364 * the skb to be delivered in some other ways. 365 * 366 * If the rx_handler changed skb->dev, to divert the skb to another 367 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 368 * new device will be called if it exists. 369 * 370 * If the rx_handler consider the skb should be ignored, it should return 371 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 372 * are registered on exact device (ptype->dev == skb->dev). 373 * 374 * If the rx_handler didn't changed skb->dev, but want the skb to be normally 375 * delivered, it should return RX_HANDLER_PASS. 376 * 377 * A device without a registered rx_handler will behave as if rx_handler 378 * returned RX_HANDLER_PASS. 379 */ 380 381 enum rx_handler_result { 382 RX_HANDLER_CONSUMED, 383 RX_HANDLER_ANOTHER, 384 RX_HANDLER_EXACT, 385 RX_HANDLER_PASS, 386 }; 387 typedef enum rx_handler_result rx_handler_result_t; 388 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 389 390 void __napi_schedule(struct napi_struct *n); 391 392 static inline bool napi_disable_pending(struct napi_struct *n) 393 { 394 return test_bit(NAPI_STATE_DISABLE, &n->state); 395 } 396 397 /** 398 * napi_schedule_prep - check if napi can be scheduled 399 * @n: napi context 400 * 401 * Test if NAPI routine is already running, and if not mark 402 * it as running. This is used as a condition variable 403 * insure only one NAPI poll instance runs. We also make 404 * sure there is no pending NAPI disable. 405 */ 406 static inline bool napi_schedule_prep(struct napi_struct *n) 407 { 408 return !napi_disable_pending(n) && 409 !test_and_set_bit(NAPI_STATE_SCHED, &n->state); 410 } 411 412 /** 413 * napi_schedule - schedule NAPI poll 414 * @n: napi context 415 * 416 * Schedule NAPI poll routine to be called if it is not already 417 * running. 418 */ 419 static inline void napi_schedule(struct napi_struct *n) 420 { 421 if (napi_schedule_prep(n)) 422 __napi_schedule(n); 423 } 424 425 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 426 static inline bool napi_reschedule(struct napi_struct *napi) 427 { 428 if (napi_schedule_prep(napi)) { 429 __napi_schedule(napi); 430 return true; 431 } 432 return false; 433 } 434 435 /** 436 * napi_complete - NAPI processing complete 437 * @n: napi context 438 * 439 * Mark NAPI processing as complete. 440 */ 441 void __napi_complete(struct napi_struct *n); 442 void napi_complete(struct napi_struct *n); 443 444 /** 445 * napi_by_id - lookup a NAPI by napi_id 446 * @napi_id: hashed napi_id 447 * 448 * lookup @napi_id in napi_hash table 449 * must be called under rcu_read_lock() 450 */ 451 struct napi_struct *napi_by_id(unsigned int napi_id); 452 453 /** 454 * napi_hash_add - add a NAPI to global hashtable 455 * @napi: napi context 456 * 457 * generate a new napi_id and store a @napi under it in napi_hash 458 */ 459 void napi_hash_add(struct napi_struct *napi); 460 461 /** 462 * napi_hash_del - remove a NAPI from global table 463 * @napi: napi context 464 * 465 * Warning: caller must observe rcu grace period 466 * before freeing memory containing @napi 467 */ 468 void napi_hash_del(struct napi_struct *napi); 469 470 /** 471 * napi_disable - prevent NAPI from scheduling 472 * @n: napi context 473 * 474 * Stop NAPI from being scheduled on this context. 475 * Waits till any outstanding processing completes. 476 */ 477 static inline void napi_disable(struct napi_struct *n) 478 { 479 might_sleep(); 480 set_bit(NAPI_STATE_DISABLE, &n->state); 481 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state)) 482 msleep(1); 483 clear_bit(NAPI_STATE_DISABLE, &n->state); 484 } 485 486 /** 487 * napi_enable - enable NAPI scheduling 488 * @n: napi context 489 * 490 * Resume NAPI from being scheduled on this context. 491 * Must be paired with napi_disable. 492 */ 493 static inline void napi_enable(struct napi_struct *n) 494 { 495 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 496 smp_mb__before_clear_bit(); 497 clear_bit(NAPI_STATE_SCHED, &n->state); 498 } 499 500 #ifdef CONFIG_SMP 501 /** 502 * napi_synchronize - wait until NAPI is not running 503 * @n: napi context 504 * 505 * Wait until NAPI is done being scheduled on this context. 506 * Waits till any outstanding processing completes but 507 * does not disable future activations. 508 */ 509 static inline void napi_synchronize(const struct napi_struct *n) 510 { 511 while (test_bit(NAPI_STATE_SCHED, &n->state)) 512 msleep(1); 513 } 514 #else 515 # define napi_synchronize(n) barrier() 516 #endif 517 518 enum netdev_queue_state_t { 519 __QUEUE_STATE_DRV_XOFF, 520 __QUEUE_STATE_STACK_XOFF, 521 __QUEUE_STATE_FROZEN, 522 }; 523 524 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 525 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 526 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 527 528 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 529 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 530 QUEUE_STATE_FROZEN) 531 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 532 QUEUE_STATE_FROZEN) 533 534 /* 535 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 536 * netif_tx_* functions below are used to manipulate this flag. The 537 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 538 * queue independently. The netif_xmit_*stopped functions below are called 539 * to check if the queue has been stopped by the driver or stack (either 540 * of the XOFF bits are set in the state). Drivers should not need to call 541 * netif_xmit*stopped functions, they should only be using netif_tx_*. 542 */ 543 544 struct netdev_queue { 545 /* 546 * read mostly part 547 */ 548 struct net_device *dev; 549 struct Qdisc *qdisc; 550 struct Qdisc *qdisc_sleeping; 551 #ifdef CONFIG_SYSFS 552 struct kobject kobj; 553 #endif 554 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 555 int numa_node; 556 #endif 557 /* 558 * write mostly part 559 */ 560 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 561 int xmit_lock_owner; 562 /* 563 * please use this field instead of dev->trans_start 564 */ 565 unsigned long trans_start; 566 567 /* 568 * Number of TX timeouts for this queue 569 * (/sys/class/net/DEV/Q/trans_timeout) 570 */ 571 unsigned long trans_timeout; 572 573 unsigned long state; 574 575 #ifdef CONFIG_BQL 576 struct dql dql; 577 #endif 578 } ____cacheline_aligned_in_smp; 579 580 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 581 { 582 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 583 return q->numa_node; 584 #else 585 return NUMA_NO_NODE; 586 #endif 587 } 588 589 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 590 { 591 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 592 q->numa_node = node; 593 #endif 594 } 595 596 #ifdef CONFIG_RPS 597 /* 598 * This structure holds an RPS map which can be of variable length. The 599 * map is an array of CPUs. 600 */ 601 struct rps_map { 602 unsigned int len; 603 struct rcu_head rcu; 604 u16 cpus[0]; 605 }; 606 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 607 608 /* 609 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 610 * tail pointer for that CPU's input queue at the time of last enqueue, and 611 * a hardware filter index. 612 */ 613 struct rps_dev_flow { 614 u16 cpu; 615 u16 filter; 616 unsigned int last_qtail; 617 }; 618 #define RPS_NO_FILTER 0xffff 619 620 /* 621 * The rps_dev_flow_table structure contains a table of flow mappings. 622 */ 623 struct rps_dev_flow_table { 624 unsigned int mask; 625 struct rcu_head rcu; 626 struct rps_dev_flow flows[0]; 627 }; 628 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 629 ((_num) * sizeof(struct rps_dev_flow))) 630 631 /* 632 * The rps_sock_flow_table contains mappings of flows to the last CPU 633 * on which they were processed by the application (set in recvmsg). 634 */ 635 struct rps_sock_flow_table { 636 unsigned int mask; 637 u16 ents[0]; 638 }; 639 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \ 640 ((_num) * sizeof(u16))) 641 642 #define RPS_NO_CPU 0xffff 643 644 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 645 u32 hash) 646 { 647 if (table && hash) { 648 unsigned int cpu, index = hash & table->mask; 649 650 /* We only give a hint, preemption can change cpu under us */ 651 cpu = raw_smp_processor_id(); 652 653 if (table->ents[index] != cpu) 654 table->ents[index] = cpu; 655 } 656 } 657 658 static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table, 659 u32 hash) 660 { 661 if (table && hash) 662 table->ents[hash & table->mask] = RPS_NO_CPU; 663 } 664 665 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 666 667 #ifdef CONFIG_RFS_ACCEL 668 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 669 u16 filter_id); 670 #endif 671 #endif /* CONFIG_RPS */ 672 673 /* This structure contains an instance of an RX queue. */ 674 struct netdev_rx_queue { 675 #ifdef CONFIG_RPS 676 struct rps_map __rcu *rps_map; 677 struct rps_dev_flow_table __rcu *rps_flow_table; 678 #endif 679 struct kobject kobj; 680 struct net_device *dev; 681 } ____cacheline_aligned_in_smp; 682 683 /* 684 * RX queue sysfs structures and functions. 685 */ 686 struct rx_queue_attribute { 687 struct attribute attr; 688 ssize_t (*show)(struct netdev_rx_queue *queue, 689 struct rx_queue_attribute *attr, char *buf); 690 ssize_t (*store)(struct netdev_rx_queue *queue, 691 struct rx_queue_attribute *attr, const char *buf, size_t len); 692 }; 693 694 #ifdef CONFIG_XPS 695 /* 696 * This structure holds an XPS map which can be of variable length. The 697 * map is an array of queues. 698 */ 699 struct xps_map { 700 unsigned int len; 701 unsigned int alloc_len; 702 struct rcu_head rcu; 703 u16 queues[0]; 704 }; 705 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 706 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \ 707 / sizeof(u16)) 708 709 /* 710 * This structure holds all XPS maps for device. Maps are indexed by CPU. 711 */ 712 struct xps_dev_maps { 713 struct rcu_head rcu; 714 struct xps_map __rcu *cpu_map[0]; 715 }; 716 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \ 717 (nr_cpu_ids * sizeof(struct xps_map *))) 718 #endif /* CONFIG_XPS */ 719 720 #define TC_MAX_QUEUE 16 721 #define TC_BITMASK 15 722 /* HW offloaded queuing disciplines txq count and offset maps */ 723 struct netdev_tc_txq { 724 u16 count; 725 u16 offset; 726 }; 727 728 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 729 /* 730 * This structure is to hold information about the device 731 * configured to run FCoE protocol stack. 732 */ 733 struct netdev_fcoe_hbainfo { 734 char manufacturer[64]; 735 char serial_number[64]; 736 char hardware_version[64]; 737 char driver_version[64]; 738 char optionrom_version[64]; 739 char firmware_version[64]; 740 char model[256]; 741 char model_description[256]; 742 }; 743 #endif 744 745 #define MAX_PHYS_PORT_ID_LEN 32 746 747 /* This structure holds a unique identifier to identify the 748 * physical port used by a netdevice. 749 */ 750 struct netdev_phys_port_id { 751 unsigned char id[MAX_PHYS_PORT_ID_LEN]; 752 unsigned char id_len; 753 }; 754 755 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 756 struct sk_buff *skb); 757 758 /* 759 * This structure defines the management hooks for network devices. 760 * The following hooks can be defined; unless noted otherwise, they are 761 * optional and can be filled with a null pointer. 762 * 763 * int (*ndo_init)(struct net_device *dev); 764 * This function is called once when network device is registered. 765 * The network device can use this to any late stage initializaton 766 * or semantic validattion. It can fail with an error code which will 767 * be propogated back to register_netdev 768 * 769 * void (*ndo_uninit)(struct net_device *dev); 770 * This function is called when device is unregistered or when registration 771 * fails. It is not called if init fails. 772 * 773 * int (*ndo_open)(struct net_device *dev); 774 * This function is called when network device transistions to the up 775 * state. 776 * 777 * int (*ndo_stop)(struct net_device *dev); 778 * This function is called when network device transistions to the down 779 * state. 780 * 781 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 782 * struct net_device *dev); 783 * Called when a packet needs to be transmitted. 784 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY. 785 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX) 786 * Required can not be NULL. 787 * 788 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 789 * void *accel_priv, select_queue_fallback_t fallback); 790 * Called to decide which queue to when device supports multiple 791 * transmit queues. 792 * 793 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 794 * This function is called to allow device receiver to make 795 * changes to configuration when multicast or promiscious is enabled. 796 * 797 * void (*ndo_set_rx_mode)(struct net_device *dev); 798 * This function is called device changes address list filtering. 799 * If driver handles unicast address filtering, it should set 800 * IFF_UNICAST_FLT to its priv_flags. 801 * 802 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 803 * This function is called when the Media Access Control address 804 * needs to be changed. If this interface is not defined, the 805 * mac address can not be changed. 806 * 807 * int (*ndo_validate_addr)(struct net_device *dev); 808 * Test if Media Access Control address is valid for the device. 809 * 810 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 811 * Called when a user request an ioctl which can't be handled by 812 * the generic interface code. If not defined ioctl's return 813 * not supported error code. 814 * 815 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 816 * Used to set network devices bus interface parameters. This interface 817 * is retained for legacy reason, new devices should use the bus 818 * interface (PCI) for low level management. 819 * 820 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 821 * Called when a user wants to change the Maximum Transfer Unit 822 * of a device. If not defined, any request to change MTU will 823 * will return an error. 824 * 825 * void (*ndo_tx_timeout)(struct net_device *dev); 826 * Callback uses when the transmitter has not made any progress 827 * for dev->watchdog ticks. 828 * 829 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev, 830 * struct rtnl_link_stats64 *storage); 831 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 832 * Called when a user wants to get the network device usage 833 * statistics. Drivers must do one of the following: 834 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 835 * rtnl_link_stats64 structure passed by the caller. 836 * 2. Define @ndo_get_stats to update a net_device_stats structure 837 * (which should normally be dev->stats) and return a pointer to 838 * it. The structure may be changed asynchronously only if each 839 * field is written atomically. 840 * 3. Update dev->stats asynchronously and atomically, and define 841 * neither operation. 842 * 843 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid); 844 * If device support VLAN filtering this function is called when a 845 * VLAN id is registered. 846 * 847 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid); 848 * If device support VLAN filtering this function is called when a 849 * VLAN id is unregistered. 850 * 851 * void (*ndo_poll_controller)(struct net_device *dev); 852 * 853 * SR-IOV management functions. 854 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 855 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos); 856 * int (*ndo_set_vf_tx_rate)(struct net_device *dev, int vf, int rate); 857 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 858 * int (*ndo_get_vf_config)(struct net_device *dev, 859 * int vf, struct ifla_vf_info *ivf); 860 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 861 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 862 * struct nlattr *port[]); 863 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 864 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc) 865 * Called to setup 'tc' number of traffic classes in the net device. This 866 * is always called from the stack with the rtnl lock held and netif tx 867 * queues stopped. This allows the netdevice to perform queue management 868 * safely. 869 * 870 * Fiber Channel over Ethernet (FCoE) offload functions. 871 * int (*ndo_fcoe_enable)(struct net_device *dev); 872 * Called when the FCoE protocol stack wants to start using LLD for FCoE 873 * so the underlying device can perform whatever needed configuration or 874 * initialization to support acceleration of FCoE traffic. 875 * 876 * int (*ndo_fcoe_disable)(struct net_device *dev); 877 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 878 * so the underlying device can perform whatever needed clean-ups to 879 * stop supporting acceleration of FCoE traffic. 880 * 881 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 882 * struct scatterlist *sgl, unsigned int sgc); 883 * Called when the FCoE Initiator wants to initialize an I/O that 884 * is a possible candidate for Direct Data Placement (DDP). The LLD can 885 * perform necessary setup and returns 1 to indicate the device is set up 886 * successfully to perform DDP on this I/O, otherwise this returns 0. 887 * 888 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 889 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 890 * indicated by the FC exchange id 'xid', so the underlying device can 891 * clean up and reuse resources for later DDP requests. 892 * 893 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 894 * struct scatterlist *sgl, unsigned int sgc); 895 * Called when the FCoE Target wants to initialize an I/O that 896 * is a possible candidate for Direct Data Placement (DDP). The LLD can 897 * perform necessary setup and returns 1 to indicate the device is set up 898 * successfully to perform DDP on this I/O, otherwise this returns 0. 899 * 900 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 901 * struct netdev_fcoe_hbainfo *hbainfo); 902 * Called when the FCoE Protocol stack wants information on the underlying 903 * device. This information is utilized by the FCoE protocol stack to 904 * register attributes with Fiber Channel management service as per the 905 * FC-GS Fabric Device Management Information(FDMI) specification. 906 * 907 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 908 * Called when the underlying device wants to override default World Wide 909 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 910 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 911 * protocol stack to use. 912 * 913 * RFS acceleration. 914 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 915 * u16 rxq_index, u32 flow_id); 916 * Set hardware filter for RFS. rxq_index is the target queue index; 917 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 918 * Return the filter ID on success, or a negative error code. 919 * 920 * Slave management functions (for bridge, bonding, etc). 921 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 922 * Called to make another netdev an underling. 923 * 924 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 925 * Called to release previously enslaved netdev. 926 * 927 * Feature/offload setting functions. 928 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 929 * netdev_features_t features); 930 * Adjusts the requested feature flags according to device-specific 931 * constraints, and returns the resulting flags. Must not modify 932 * the device state. 933 * 934 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 935 * Called to update device configuration to new features. Passed 936 * feature set might be less than what was returned by ndo_fix_features()). 937 * Must return >0 or -errno if it changed dev->features itself. 938 * 939 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 940 * struct net_device *dev, 941 * const unsigned char *addr, u16 flags) 942 * Adds an FDB entry to dev for addr. 943 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 944 * struct net_device *dev, 945 * const unsigned char *addr) 946 * Deletes the FDB entry from dev coresponding to addr. 947 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 948 * struct net_device *dev, int idx) 949 * Used to add FDB entries to dump requests. Implementers should add 950 * entries to skb and update idx with the number of entries. 951 * 952 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh) 953 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 954 * struct net_device *dev, u32 filter_mask) 955 * 956 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 957 * Called to change device carrier. Soft-devices (like dummy, team, etc) 958 * which do not represent real hardware may define this to allow their 959 * userspace components to manage their virtual carrier state. Devices 960 * that determine carrier state from physical hardware properties (eg 961 * network cables) or protocol-dependent mechanisms (eg 962 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 963 * 964 * int (*ndo_get_phys_port_id)(struct net_device *dev, 965 * struct netdev_phys_port_id *ppid); 966 * Called to get ID of physical port of this device. If driver does 967 * not implement this, it is assumed that the hw is not able to have 968 * multiple net devices on single physical port. 969 * 970 * void (*ndo_add_vxlan_port)(struct net_device *dev, 971 * sa_family_t sa_family, __be16 port); 972 * Called by vxlan to notiy a driver about the UDP port and socket 973 * address family that vxlan is listnening to. It is called only when 974 * a new port starts listening. The operation is protected by the 975 * vxlan_net->sock_lock. 976 * 977 * void (*ndo_del_vxlan_port)(struct net_device *dev, 978 * sa_family_t sa_family, __be16 port); 979 * Called by vxlan to notify the driver about a UDP port and socket 980 * address family that vxlan is not listening to anymore. The operation 981 * is protected by the vxlan_net->sock_lock. 982 * 983 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 984 * struct net_device *dev) 985 * Called by upper layer devices to accelerate switching or other 986 * station functionality into hardware. 'pdev is the lowerdev 987 * to use for the offload and 'dev' is the net device that will 988 * back the offload. Returns a pointer to the private structure 989 * the upper layer will maintain. 990 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 991 * Called by upper layer device to delete the station created 992 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 993 * the station and priv is the structure returned by the add 994 * operation. 995 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb, 996 * struct net_device *dev, 997 * void *priv); 998 * Callback to use for xmit over the accelerated station. This 999 * is used in place of ndo_start_xmit on accelerated net 1000 * devices. 1001 */ 1002 struct net_device_ops { 1003 int (*ndo_init)(struct net_device *dev); 1004 void (*ndo_uninit)(struct net_device *dev); 1005 int (*ndo_open)(struct net_device *dev); 1006 int (*ndo_stop)(struct net_device *dev); 1007 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb, 1008 struct net_device *dev); 1009 u16 (*ndo_select_queue)(struct net_device *dev, 1010 struct sk_buff *skb, 1011 void *accel_priv, 1012 select_queue_fallback_t fallback); 1013 void (*ndo_change_rx_flags)(struct net_device *dev, 1014 int flags); 1015 void (*ndo_set_rx_mode)(struct net_device *dev); 1016 int (*ndo_set_mac_address)(struct net_device *dev, 1017 void *addr); 1018 int (*ndo_validate_addr)(struct net_device *dev); 1019 int (*ndo_do_ioctl)(struct net_device *dev, 1020 struct ifreq *ifr, int cmd); 1021 int (*ndo_set_config)(struct net_device *dev, 1022 struct ifmap *map); 1023 int (*ndo_change_mtu)(struct net_device *dev, 1024 int new_mtu); 1025 int (*ndo_neigh_setup)(struct net_device *dev, 1026 struct neigh_parms *); 1027 void (*ndo_tx_timeout) (struct net_device *dev); 1028 1029 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev, 1030 struct rtnl_link_stats64 *storage); 1031 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1032 1033 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1034 __be16 proto, u16 vid); 1035 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1036 __be16 proto, u16 vid); 1037 #ifdef CONFIG_NET_POLL_CONTROLLER 1038 void (*ndo_poll_controller)(struct net_device *dev); 1039 int (*ndo_netpoll_setup)(struct net_device *dev, 1040 struct netpoll_info *info); 1041 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1042 #endif 1043 #ifdef CONFIG_NET_RX_BUSY_POLL 1044 int (*ndo_busy_poll)(struct napi_struct *dev); 1045 #endif 1046 int (*ndo_set_vf_mac)(struct net_device *dev, 1047 int queue, u8 *mac); 1048 int (*ndo_set_vf_vlan)(struct net_device *dev, 1049 int queue, u16 vlan, u8 qos); 1050 int (*ndo_set_vf_tx_rate)(struct net_device *dev, 1051 int vf, int rate); 1052 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1053 int vf, bool setting); 1054 int (*ndo_get_vf_config)(struct net_device *dev, 1055 int vf, 1056 struct ifla_vf_info *ivf); 1057 int (*ndo_set_vf_link_state)(struct net_device *dev, 1058 int vf, int link_state); 1059 int (*ndo_set_vf_port)(struct net_device *dev, 1060 int vf, 1061 struct nlattr *port[]); 1062 int (*ndo_get_vf_port)(struct net_device *dev, 1063 int vf, struct sk_buff *skb); 1064 int (*ndo_setup_tc)(struct net_device *dev, u8 tc); 1065 #if IS_ENABLED(CONFIG_FCOE) 1066 int (*ndo_fcoe_enable)(struct net_device *dev); 1067 int (*ndo_fcoe_disable)(struct net_device *dev); 1068 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1069 u16 xid, 1070 struct scatterlist *sgl, 1071 unsigned int sgc); 1072 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1073 u16 xid); 1074 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1075 u16 xid, 1076 struct scatterlist *sgl, 1077 unsigned int sgc); 1078 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1079 struct netdev_fcoe_hbainfo *hbainfo); 1080 #endif 1081 1082 #if IS_ENABLED(CONFIG_LIBFCOE) 1083 #define NETDEV_FCOE_WWNN 0 1084 #define NETDEV_FCOE_WWPN 1 1085 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1086 u64 *wwn, int type); 1087 #endif 1088 1089 #ifdef CONFIG_RFS_ACCEL 1090 int (*ndo_rx_flow_steer)(struct net_device *dev, 1091 const struct sk_buff *skb, 1092 u16 rxq_index, 1093 u32 flow_id); 1094 #endif 1095 int (*ndo_add_slave)(struct net_device *dev, 1096 struct net_device *slave_dev); 1097 int (*ndo_del_slave)(struct net_device *dev, 1098 struct net_device *slave_dev); 1099 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1100 netdev_features_t features); 1101 int (*ndo_set_features)(struct net_device *dev, 1102 netdev_features_t features); 1103 int (*ndo_neigh_construct)(struct neighbour *n); 1104 void (*ndo_neigh_destroy)(struct neighbour *n); 1105 1106 int (*ndo_fdb_add)(struct ndmsg *ndm, 1107 struct nlattr *tb[], 1108 struct net_device *dev, 1109 const unsigned char *addr, 1110 u16 flags); 1111 int (*ndo_fdb_del)(struct ndmsg *ndm, 1112 struct nlattr *tb[], 1113 struct net_device *dev, 1114 const unsigned char *addr); 1115 int (*ndo_fdb_dump)(struct sk_buff *skb, 1116 struct netlink_callback *cb, 1117 struct net_device *dev, 1118 int idx); 1119 1120 int (*ndo_bridge_setlink)(struct net_device *dev, 1121 struct nlmsghdr *nlh); 1122 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1123 u32 pid, u32 seq, 1124 struct net_device *dev, 1125 u32 filter_mask); 1126 int (*ndo_bridge_dellink)(struct net_device *dev, 1127 struct nlmsghdr *nlh); 1128 int (*ndo_change_carrier)(struct net_device *dev, 1129 bool new_carrier); 1130 int (*ndo_get_phys_port_id)(struct net_device *dev, 1131 struct netdev_phys_port_id *ppid); 1132 void (*ndo_add_vxlan_port)(struct net_device *dev, 1133 sa_family_t sa_family, 1134 __be16 port); 1135 void (*ndo_del_vxlan_port)(struct net_device *dev, 1136 sa_family_t sa_family, 1137 __be16 port); 1138 1139 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1140 struct net_device *dev); 1141 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1142 void *priv); 1143 1144 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb, 1145 struct net_device *dev, 1146 void *priv); 1147 int (*ndo_get_lock_subclass)(struct net_device *dev); 1148 }; 1149 1150 /** 1151 * enum net_device_priv_flags - &struct net_device priv_flags 1152 * 1153 * These are the &struct net_device, they are only set internally 1154 * by drivers and used in the kernel. These flags are invisible to 1155 * userspace, this means that the order of these flags can change 1156 * during any kernel release. 1157 * 1158 * You should have a pretty good reason to be extending these flags. 1159 * 1160 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1161 * @IFF_EBRIDGE: Ethernet bridging device 1162 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active 1163 * @IFF_MASTER_8023AD: bonding master, 802.3ad 1164 * @IFF_MASTER_ALB: bonding master, balance-alb 1165 * @IFF_BONDING: bonding master or slave 1166 * @IFF_SLAVE_NEEDARP: need ARPs for validation 1167 * @IFF_ISATAP: ISATAP interface (RFC4214) 1168 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use 1169 * @IFF_WAN_HDLC: WAN HDLC device 1170 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1171 * release skb->dst 1172 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1173 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1174 * @IFF_MACVLAN_PORT: device used as macvlan port 1175 * @IFF_BRIDGE_PORT: device used as bridge port 1176 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1177 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1178 * @IFF_UNICAST_FLT: Supports unicast filtering 1179 * @IFF_TEAM_PORT: device used as team port 1180 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1181 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1182 * change when it's running 1183 * @IFF_MACVLAN: Macvlan device 1184 */ 1185 enum netdev_priv_flags { 1186 IFF_802_1Q_VLAN = 1<<0, 1187 IFF_EBRIDGE = 1<<1, 1188 IFF_SLAVE_INACTIVE = 1<<2, 1189 IFF_MASTER_8023AD = 1<<3, 1190 IFF_MASTER_ALB = 1<<4, 1191 IFF_BONDING = 1<<5, 1192 IFF_SLAVE_NEEDARP = 1<<6, 1193 IFF_ISATAP = 1<<7, 1194 IFF_MASTER_ARPMON = 1<<8, 1195 IFF_WAN_HDLC = 1<<9, 1196 IFF_XMIT_DST_RELEASE = 1<<10, 1197 IFF_DONT_BRIDGE = 1<<11, 1198 IFF_DISABLE_NETPOLL = 1<<12, 1199 IFF_MACVLAN_PORT = 1<<13, 1200 IFF_BRIDGE_PORT = 1<<14, 1201 IFF_OVS_DATAPATH = 1<<15, 1202 IFF_TX_SKB_SHARING = 1<<16, 1203 IFF_UNICAST_FLT = 1<<17, 1204 IFF_TEAM_PORT = 1<<18, 1205 IFF_SUPP_NOFCS = 1<<19, 1206 IFF_LIVE_ADDR_CHANGE = 1<<20, 1207 IFF_MACVLAN = 1<<21, 1208 }; 1209 1210 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1211 #define IFF_EBRIDGE IFF_EBRIDGE 1212 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE 1213 #define IFF_MASTER_8023AD IFF_MASTER_8023AD 1214 #define IFF_MASTER_ALB IFF_MASTER_ALB 1215 #define IFF_BONDING IFF_BONDING 1216 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP 1217 #define IFF_ISATAP IFF_ISATAP 1218 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON 1219 #define IFF_WAN_HDLC IFF_WAN_HDLC 1220 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1221 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1222 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1223 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1224 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1225 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1226 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1227 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1228 #define IFF_TEAM_PORT IFF_TEAM_PORT 1229 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1230 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1231 #define IFF_MACVLAN IFF_MACVLAN 1232 1233 /* 1234 * The DEVICE structure. 1235 * Actually, this whole structure is a big mistake. It mixes I/O 1236 * data with strictly "high-level" data, and it has to know about 1237 * almost every data structure used in the INET module. 1238 * 1239 * FIXME: cleanup struct net_device such that network protocol info 1240 * moves out. 1241 */ 1242 1243 struct net_device { 1244 1245 /* 1246 * This is the first field of the "visible" part of this structure 1247 * (i.e. as seen by users in the "Space.c" file). It is the name 1248 * of the interface. 1249 */ 1250 char name[IFNAMSIZ]; 1251 1252 /* device name hash chain, please keep it close to name[] */ 1253 struct hlist_node name_hlist; 1254 1255 /* snmp alias */ 1256 char *ifalias; 1257 1258 /* 1259 * I/O specific fields 1260 * FIXME: Merge these and struct ifmap into one 1261 */ 1262 unsigned long mem_end; /* shared mem end */ 1263 unsigned long mem_start; /* shared mem start */ 1264 unsigned long base_addr; /* device I/O address */ 1265 int irq; /* device IRQ number */ 1266 1267 /* 1268 * Some hardware also needs these fields, but they are not 1269 * part of the usual set specified in Space.c. 1270 */ 1271 1272 unsigned long state; 1273 1274 struct list_head dev_list; 1275 struct list_head napi_list; 1276 struct list_head unreg_list; 1277 struct list_head close_list; 1278 1279 /* directly linked devices, like slaves for bonding */ 1280 struct { 1281 struct list_head upper; 1282 struct list_head lower; 1283 } adj_list; 1284 1285 /* all linked devices, *including* neighbours */ 1286 struct { 1287 struct list_head upper; 1288 struct list_head lower; 1289 } all_adj_list; 1290 1291 1292 /* currently active device features */ 1293 netdev_features_t features; 1294 /* user-changeable features */ 1295 netdev_features_t hw_features; 1296 /* user-requested features */ 1297 netdev_features_t wanted_features; 1298 /* mask of features inheritable by VLAN devices */ 1299 netdev_features_t vlan_features; 1300 /* mask of features inherited by encapsulating devices 1301 * This field indicates what encapsulation offloads 1302 * the hardware is capable of doing, and drivers will 1303 * need to set them appropriately. 1304 */ 1305 netdev_features_t hw_enc_features; 1306 /* mask of fetures inheritable by MPLS */ 1307 netdev_features_t mpls_features; 1308 1309 /* Interface index. Unique device identifier */ 1310 int ifindex; 1311 int iflink; 1312 1313 struct net_device_stats stats; 1314 1315 /* dropped packets by core network, Do not use this in drivers */ 1316 atomic_long_t rx_dropped; 1317 atomic_long_t tx_dropped; 1318 1319 /* Stats to monitor carrier on<->off transitions */ 1320 atomic_t carrier_changes; 1321 1322 #ifdef CONFIG_WIRELESS_EXT 1323 /* List of functions to handle Wireless Extensions (instead of ioctl). 1324 * See <net/iw_handler.h> for details. Jean II */ 1325 const struct iw_handler_def * wireless_handlers; 1326 /* Instance data managed by the core of Wireless Extensions. */ 1327 struct iw_public_data * wireless_data; 1328 #endif 1329 /* Management operations */ 1330 const struct net_device_ops *netdev_ops; 1331 const struct ethtool_ops *ethtool_ops; 1332 const struct forwarding_accel_ops *fwd_ops; 1333 1334 /* Hardware header description */ 1335 const struct header_ops *header_ops; 1336 1337 unsigned int flags; /* interface flags (a la BSD) */ 1338 unsigned int priv_flags; /* Like 'flags' but invisible to userspace. 1339 * See if.h for definitions. */ 1340 unsigned short gflags; 1341 unsigned short padded; /* How much padding added by alloc_netdev() */ 1342 1343 unsigned char operstate; /* RFC2863 operstate */ 1344 unsigned char link_mode; /* mapping policy to operstate */ 1345 1346 unsigned char if_port; /* Selectable AUI, TP,..*/ 1347 unsigned char dma; /* DMA channel */ 1348 1349 unsigned int mtu; /* interface MTU value */ 1350 unsigned short type; /* interface hardware type */ 1351 unsigned short hard_header_len; /* hardware hdr length */ 1352 1353 /* extra head- and tailroom the hardware may need, but not in all cases 1354 * can this be guaranteed, especially tailroom. Some cases also use 1355 * LL_MAX_HEADER instead to allocate the skb. 1356 */ 1357 unsigned short needed_headroom; 1358 unsigned short needed_tailroom; 1359 1360 /* Interface address info. */ 1361 unsigned char perm_addr[MAX_ADDR_LEN]; /* permanent hw address */ 1362 unsigned char addr_assign_type; /* hw address assignment type */ 1363 unsigned char addr_len; /* hardware address length */ 1364 unsigned short neigh_priv_len; 1365 unsigned short dev_id; /* Used to differentiate devices 1366 * that share the same link 1367 * layer address 1368 */ 1369 unsigned short dev_port; /* Used to differentiate 1370 * devices that share the same 1371 * function 1372 */ 1373 spinlock_t addr_list_lock; 1374 struct netdev_hw_addr_list uc; /* Unicast mac addresses */ 1375 struct netdev_hw_addr_list mc; /* Multicast mac addresses */ 1376 struct netdev_hw_addr_list dev_addrs; /* list of device 1377 * hw addresses 1378 */ 1379 #ifdef CONFIG_SYSFS 1380 struct kset *queues_kset; 1381 #endif 1382 1383 bool uc_promisc; 1384 unsigned int promiscuity; 1385 unsigned int allmulti; 1386 1387 1388 /* Protocol specific pointers */ 1389 1390 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1391 struct vlan_info __rcu *vlan_info; /* VLAN info */ 1392 #endif 1393 #if IS_ENABLED(CONFIG_NET_DSA) 1394 struct dsa_switch_tree *dsa_ptr; /* dsa specific data */ 1395 #endif 1396 #if IS_ENABLED(CONFIG_TIPC) 1397 struct tipc_bearer __rcu *tipc_ptr; /* TIPC specific data */ 1398 #endif 1399 void *atalk_ptr; /* AppleTalk link */ 1400 struct in_device __rcu *ip_ptr; /* IPv4 specific data */ 1401 struct dn_dev __rcu *dn_ptr; /* DECnet specific data */ 1402 struct inet6_dev __rcu *ip6_ptr; /* IPv6 specific data */ 1403 void *ax25_ptr; /* AX.25 specific data */ 1404 struct wireless_dev *ieee80211_ptr; /* IEEE 802.11 specific data, 1405 assign before registering */ 1406 1407 /* 1408 * Cache lines mostly used on receive path (including eth_type_trans()) 1409 */ 1410 unsigned long last_rx; /* Time of last Rx */ 1411 1412 /* Interface address info used in eth_type_trans() */ 1413 unsigned char *dev_addr; /* hw address, (before bcast 1414 because most packets are 1415 unicast) */ 1416 1417 1418 #ifdef CONFIG_SYSFS 1419 struct netdev_rx_queue *_rx; 1420 1421 /* Number of RX queues allocated at register_netdev() time */ 1422 unsigned int num_rx_queues; 1423 1424 /* Number of RX queues currently active in device */ 1425 unsigned int real_num_rx_queues; 1426 1427 #endif 1428 1429 rx_handler_func_t __rcu *rx_handler; 1430 void __rcu *rx_handler_data; 1431 1432 struct netdev_queue __rcu *ingress_queue; 1433 unsigned char broadcast[MAX_ADDR_LEN]; /* hw bcast add */ 1434 1435 1436 /* 1437 * Cache lines mostly used on transmit path 1438 */ 1439 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1440 1441 /* Number of TX queues allocated at alloc_netdev_mq() time */ 1442 unsigned int num_tx_queues; 1443 1444 /* Number of TX queues currently active in device */ 1445 unsigned int real_num_tx_queues; 1446 1447 /* root qdisc from userspace point of view */ 1448 struct Qdisc *qdisc; 1449 1450 unsigned long tx_queue_len; /* Max frames per queue allowed */ 1451 spinlock_t tx_global_lock; 1452 1453 #ifdef CONFIG_XPS 1454 struct xps_dev_maps __rcu *xps_maps; 1455 #endif 1456 #ifdef CONFIG_RFS_ACCEL 1457 /* CPU reverse-mapping for RX completion interrupts, indexed 1458 * by RX queue number. Assigned by driver. This must only be 1459 * set if the ndo_rx_flow_steer operation is defined. */ 1460 struct cpu_rmap *rx_cpu_rmap; 1461 #endif 1462 1463 /* These may be needed for future network-power-down code. */ 1464 1465 /* 1466 * trans_start here is expensive for high speed devices on SMP, 1467 * please use netdev_queue->trans_start instead. 1468 */ 1469 unsigned long trans_start; /* Time (in jiffies) of last Tx */ 1470 1471 int watchdog_timeo; /* used by dev_watchdog() */ 1472 struct timer_list watchdog_timer; 1473 1474 /* Number of references to this device */ 1475 int __percpu *pcpu_refcnt; 1476 1477 /* delayed register/unregister */ 1478 struct list_head todo_list; 1479 /* device index hash chain */ 1480 struct hlist_node index_hlist; 1481 1482 struct list_head link_watch_list; 1483 1484 /* register/unregister state machine */ 1485 enum { NETREG_UNINITIALIZED=0, 1486 NETREG_REGISTERED, /* completed register_netdevice */ 1487 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1488 NETREG_UNREGISTERED, /* completed unregister todo */ 1489 NETREG_RELEASED, /* called free_netdev */ 1490 NETREG_DUMMY, /* dummy device for NAPI poll */ 1491 } reg_state:8; 1492 1493 bool dismantle; /* device is going do be freed */ 1494 1495 enum { 1496 RTNL_LINK_INITIALIZED, 1497 RTNL_LINK_INITIALIZING, 1498 } rtnl_link_state:16; 1499 1500 /* Called from unregister, can be used to call free_netdev */ 1501 void (*destructor)(struct net_device *dev); 1502 1503 #ifdef CONFIG_NETPOLL 1504 struct netpoll_info __rcu *npinfo; 1505 #endif 1506 1507 #ifdef CONFIG_NET_NS 1508 /* Network namespace this network device is inside */ 1509 struct net *nd_net; 1510 #endif 1511 1512 /* mid-layer private */ 1513 union { 1514 void *ml_priv; 1515 struct pcpu_lstats __percpu *lstats; /* loopback stats */ 1516 struct pcpu_sw_netstats __percpu *tstats; 1517 struct pcpu_dstats __percpu *dstats; /* dummy stats */ 1518 struct pcpu_vstats __percpu *vstats; /* veth stats */ 1519 }; 1520 /* GARP */ 1521 struct garp_port __rcu *garp_port; 1522 /* MRP */ 1523 struct mrp_port __rcu *mrp_port; 1524 1525 /* class/net/name entry */ 1526 struct device dev; 1527 /* space for optional device, statistics, and wireless sysfs groups */ 1528 const struct attribute_group *sysfs_groups[4]; 1529 /* space for optional per-rx queue attributes */ 1530 const struct attribute_group *sysfs_rx_queue_group; 1531 1532 /* rtnetlink link ops */ 1533 const struct rtnl_link_ops *rtnl_link_ops; 1534 1535 /* for setting kernel sock attribute on TCP connection setup */ 1536 #define GSO_MAX_SIZE 65536 1537 unsigned int gso_max_size; 1538 #define GSO_MAX_SEGS 65535 1539 u16 gso_max_segs; 1540 1541 #ifdef CONFIG_DCB 1542 /* Data Center Bridging netlink ops */ 1543 const struct dcbnl_rtnl_ops *dcbnl_ops; 1544 #endif 1545 u8 num_tc; 1546 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 1547 u8 prio_tc_map[TC_BITMASK + 1]; 1548 1549 #if IS_ENABLED(CONFIG_FCOE) 1550 /* max exchange id for FCoE LRO by ddp */ 1551 unsigned int fcoe_ddp_xid; 1552 #endif 1553 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 1554 struct netprio_map __rcu *priomap; 1555 #endif 1556 /* phy device may attach itself for hardware timestamping */ 1557 struct phy_device *phydev; 1558 1559 struct lock_class_key *qdisc_tx_busylock; 1560 1561 /* group the device belongs to */ 1562 int group; 1563 1564 struct pm_qos_request pm_qos_req; 1565 }; 1566 #define to_net_dev(d) container_of(d, struct net_device, dev) 1567 1568 #define NETDEV_ALIGN 32 1569 1570 static inline 1571 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 1572 { 1573 return dev->prio_tc_map[prio & TC_BITMASK]; 1574 } 1575 1576 static inline 1577 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 1578 { 1579 if (tc >= dev->num_tc) 1580 return -EINVAL; 1581 1582 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 1583 return 0; 1584 } 1585 1586 static inline 1587 void netdev_reset_tc(struct net_device *dev) 1588 { 1589 dev->num_tc = 0; 1590 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq)); 1591 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map)); 1592 } 1593 1594 static inline 1595 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset) 1596 { 1597 if (tc >= dev->num_tc) 1598 return -EINVAL; 1599 1600 dev->tc_to_txq[tc].count = count; 1601 dev->tc_to_txq[tc].offset = offset; 1602 return 0; 1603 } 1604 1605 static inline 1606 int netdev_set_num_tc(struct net_device *dev, u8 num_tc) 1607 { 1608 if (num_tc > TC_MAX_QUEUE) 1609 return -EINVAL; 1610 1611 dev->num_tc = num_tc; 1612 return 0; 1613 } 1614 1615 static inline 1616 int netdev_get_num_tc(struct net_device *dev) 1617 { 1618 return dev->num_tc; 1619 } 1620 1621 static inline 1622 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 1623 unsigned int index) 1624 { 1625 return &dev->_tx[index]; 1626 } 1627 1628 static inline void netdev_for_each_tx_queue(struct net_device *dev, 1629 void (*f)(struct net_device *, 1630 struct netdev_queue *, 1631 void *), 1632 void *arg) 1633 { 1634 unsigned int i; 1635 1636 for (i = 0; i < dev->num_tx_queues; i++) 1637 f(dev, &dev->_tx[i], arg); 1638 } 1639 1640 struct netdev_queue *netdev_pick_tx(struct net_device *dev, 1641 struct sk_buff *skb, 1642 void *accel_priv); 1643 1644 /* 1645 * Net namespace inlines 1646 */ 1647 static inline 1648 struct net *dev_net(const struct net_device *dev) 1649 { 1650 return read_pnet(&dev->nd_net); 1651 } 1652 1653 static inline 1654 void dev_net_set(struct net_device *dev, struct net *net) 1655 { 1656 #ifdef CONFIG_NET_NS 1657 release_net(dev->nd_net); 1658 dev->nd_net = hold_net(net); 1659 #endif 1660 } 1661 1662 static inline bool netdev_uses_dsa_tags(struct net_device *dev) 1663 { 1664 #ifdef CONFIG_NET_DSA_TAG_DSA 1665 if (dev->dsa_ptr != NULL) 1666 return dsa_uses_dsa_tags(dev->dsa_ptr); 1667 #endif 1668 1669 return 0; 1670 } 1671 1672 static inline bool netdev_uses_trailer_tags(struct net_device *dev) 1673 { 1674 #ifdef CONFIG_NET_DSA_TAG_TRAILER 1675 if (dev->dsa_ptr != NULL) 1676 return dsa_uses_trailer_tags(dev->dsa_ptr); 1677 #endif 1678 1679 return 0; 1680 } 1681 1682 /** 1683 * netdev_priv - access network device private data 1684 * @dev: network device 1685 * 1686 * Get network device private data 1687 */ 1688 static inline void *netdev_priv(const struct net_device *dev) 1689 { 1690 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 1691 } 1692 1693 /* Set the sysfs physical device reference for the network logical device 1694 * if set prior to registration will cause a symlink during initialization. 1695 */ 1696 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 1697 1698 /* Set the sysfs device type for the network logical device to allow 1699 * fine-grained identification of different network device types. For 1700 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc. 1701 */ 1702 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 1703 1704 /* Default NAPI poll() weight 1705 * Device drivers are strongly advised to not use bigger value 1706 */ 1707 #define NAPI_POLL_WEIGHT 64 1708 1709 /** 1710 * netif_napi_add - initialize a napi context 1711 * @dev: network device 1712 * @napi: napi context 1713 * @poll: polling function 1714 * @weight: default weight 1715 * 1716 * netif_napi_add() must be used to initialize a napi context prior to calling 1717 * *any* of the other napi related functions. 1718 */ 1719 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 1720 int (*poll)(struct napi_struct *, int), int weight); 1721 1722 /** 1723 * netif_napi_del - remove a napi context 1724 * @napi: napi context 1725 * 1726 * netif_napi_del() removes a napi context from the network device napi list 1727 */ 1728 void netif_napi_del(struct napi_struct *napi); 1729 1730 struct napi_gro_cb { 1731 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 1732 void *frag0; 1733 1734 /* Length of frag0. */ 1735 unsigned int frag0_len; 1736 1737 /* This indicates where we are processing relative to skb->data. */ 1738 int data_offset; 1739 1740 /* This is non-zero if the packet cannot be merged with the new skb. */ 1741 u16 flush; 1742 1743 /* Save the IP ID here and check when we get to the transport layer */ 1744 u16 flush_id; 1745 1746 /* Number of segments aggregated. */ 1747 u16 count; 1748 1749 /* This is non-zero if the packet may be of the same flow. */ 1750 u8 same_flow; 1751 1752 /* Free the skb? */ 1753 u8 free; 1754 #define NAPI_GRO_FREE 1 1755 #define NAPI_GRO_FREE_STOLEN_HEAD 2 1756 1757 /* jiffies when first packet was created/queued */ 1758 unsigned long age; 1759 1760 /* Used in ipv6_gro_receive() */ 1761 u16 proto; 1762 1763 /* Used in udp_gro_receive */ 1764 u16 udp_mark; 1765 1766 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 1767 __wsum csum; 1768 1769 /* used in skb_gro_receive() slow path */ 1770 struct sk_buff *last; 1771 }; 1772 1773 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 1774 1775 struct packet_type { 1776 __be16 type; /* This is really htons(ether_type). */ 1777 struct net_device *dev; /* NULL is wildcarded here */ 1778 int (*func) (struct sk_buff *, 1779 struct net_device *, 1780 struct packet_type *, 1781 struct net_device *); 1782 bool (*id_match)(struct packet_type *ptype, 1783 struct sock *sk); 1784 void *af_packet_priv; 1785 struct list_head list; 1786 }; 1787 1788 struct offload_callbacks { 1789 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 1790 netdev_features_t features); 1791 int (*gso_send_check)(struct sk_buff *skb); 1792 struct sk_buff **(*gro_receive)(struct sk_buff **head, 1793 struct sk_buff *skb); 1794 int (*gro_complete)(struct sk_buff *skb, int nhoff); 1795 }; 1796 1797 struct packet_offload { 1798 __be16 type; /* This is really htons(ether_type). */ 1799 struct offload_callbacks callbacks; 1800 struct list_head list; 1801 }; 1802 1803 struct udp_offload { 1804 __be16 port; 1805 struct offload_callbacks callbacks; 1806 }; 1807 1808 /* often modified stats are per cpu, other are shared (netdev->stats) */ 1809 struct pcpu_sw_netstats { 1810 u64 rx_packets; 1811 u64 rx_bytes; 1812 u64 tx_packets; 1813 u64 tx_bytes; 1814 struct u64_stats_sync syncp; 1815 }; 1816 1817 #define netdev_alloc_pcpu_stats(type) \ 1818 ({ \ 1819 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \ 1820 if (pcpu_stats) { \ 1821 int i; \ 1822 for_each_possible_cpu(i) { \ 1823 typeof(type) *stat; \ 1824 stat = per_cpu_ptr(pcpu_stats, i); \ 1825 u64_stats_init(&stat->syncp); \ 1826 } \ 1827 } \ 1828 pcpu_stats; \ 1829 }) 1830 1831 #include <linux/notifier.h> 1832 1833 /* netdevice notifier chain. Please remember to update the rtnetlink 1834 * notification exclusion list in rtnetlink_event() when adding new 1835 * types. 1836 */ 1837 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */ 1838 #define NETDEV_DOWN 0x0002 1839 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface 1840 detected a hardware crash and restarted 1841 - we can use this eg to kick tcp sessions 1842 once done */ 1843 #define NETDEV_CHANGE 0x0004 /* Notify device state change */ 1844 #define NETDEV_REGISTER 0x0005 1845 #define NETDEV_UNREGISTER 0x0006 1846 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */ 1847 #define NETDEV_CHANGEADDR 0x0008 1848 #define NETDEV_GOING_DOWN 0x0009 1849 #define NETDEV_CHANGENAME 0x000A 1850 #define NETDEV_FEAT_CHANGE 0x000B 1851 #define NETDEV_BONDING_FAILOVER 0x000C 1852 #define NETDEV_PRE_UP 0x000D 1853 #define NETDEV_PRE_TYPE_CHANGE 0x000E 1854 #define NETDEV_POST_TYPE_CHANGE 0x000F 1855 #define NETDEV_POST_INIT 0x0010 1856 #define NETDEV_UNREGISTER_FINAL 0x0011 1857 #define NETDEV_RELEASE 0x0012 1858 #define NETDEV_NOTIFY_PEERS 0x0013 1859 #define NETDEV_JOIN 0x0014 1860 #define NETDEV_CHANGEUPPER 0x0015 1861 #define NETDEV_RESEND_IGMP 0x0016 1862 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */ 1863 1864 int register_netdevice_notifier(struct notifier_block *nb); 1865 int unregister_netdevice_notifier(struct notifier_block *nb); 1866 1867 struct netdev_notifier_info { 1868 struct net_device *dev; 1869 }; 1870 1871 struct netdev_notifier_change_info { 1872 struct netdev_notifier_info info; /* must be first */ 1873 unsigned int flags_changed; 1874 }; 1875 1876 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 1877 struct net_device *dev) 1878 { 1879 info->dev = dev; 1880 } 1881 1882 static inline struct net_device * 1883 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 1884 { 1885 return info->dev; 1886 } 1887 1888 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 1889 1890 1891 extern rwlock_t dev_base_lock; /* Device list lock */ 1892 1893 #define for_each_netdev(net, d) \ 1894 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 1895 #define for_each_netdev_reverse(net, d) \ 1896 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 1897 #define for_each_netdev_rcu(net, d) \ 1898 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 1899 #define for_each_netdev_safe(net, d, n) \ 1900 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 1901 #define for_each_netdev_continue(net, d) \ 1902 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 1903 #define for_each_netdev_continue_rcu(net, d) \ 1904 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 1905 #define for_each_netdev_in_bond_rcu(bond, slave) \ 1906 for_each_netdev_rcu(&init_net, slave) \ 1907 if (netdev_master_upper_dev_get_rcu(slave) == bond) 1908 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 1909 1910 static inline struct net_device *next_net_device(struct net_device *dev) 1911 { 1912 struct list_head *lh; 1913 struct net *net; 1914 1915 net = dev_net(dev); 1916 lh = dev->dev_list.next; 1917 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 1918 } 1919 1920 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 1921 { 1922 struct list_head *lh; 1923 struct net *net; 1924 1925 net = dev_net(dev); 1926 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 1927 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 1928 } 1929 1930 static inline struct net_device *first_net_device(struct net *net) 1931 { 1932 return list_empty(&net->dev_base_head) ? NULL : 1933 net_device_entry(net->dev_base_head.next); 1934 } 1935 1936 static inline struct net_device *first_net_device_rcu(struct net *net) 1937 { 1938 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 1939 1940 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 1941 } 1942 1943 int netdev_boot_setup_check(struct net_device *dev); 1944 unsigned long netdev_boot_base(const char *prefix, int unit); 1945 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 1946 const char *hwaddr); 1947 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 1948 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 1949 void dev_add_pack(struct packet_type *pt); 1950 void dev_remove_pack(struct packet_type *pt); 1951 void __dev_remove_pack(struct packet_type *pt); 1952 void dev_add_offload(struct packet_offload *po); 1953 void dev_remove_offload(struct packet_offload *po); 1954 1955 struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short flags, 1956 unsigned short mask); 1957 struct net_device *dev_get_by_name(struct net *net, const char *name); 1958 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 1959 struct net_device *__dev_get_by_name(struct net *net, const char *name); 1960 int dev_alloc_name(struct net_device *dev, const char *name); 1961 int dev_open(struct net_device *dev); 1962 int dev_close(struct net_device *dev); 1963 void dev_disable_lro(struct net_device *dev); 1964 int dev_loopback_xmit(struct sk_buff *newskb); 1965 int dev_queue_xmit(struct sk_buff *skb); 1966 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv); 1967 int register_netdevice(struct net_device *dev); 1968 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 1969 void unregister_netdevice_many(struct list_head *head); 1970 static inline void unregister_netdevice(struct net_device *dev) 1971 { 1972 unregister_netdevice_queue(dev, NULL); 1973 } 1974 1975 int netdev_refcnt_read(const struct net_device *dev); 1976 void free_netdev(struct net_device *dev); 1977 void netdev_freemem(struct net_device *dev); 1978 void synchronize_net(void); 1979 int init_dummy_netdev(struct net_device *dev); 1980 1981 struct net_device *dev_get_by_index(struct net *net, int ifindex); 1982 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 1983 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 1984 int netdev_get_name(struct net *net, char *name, int ifindex); 1985 int dev_restart(struct net_device *dev); 1986 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb); 1987 1988 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 1989 { 1990 return NAPI_GRO_CB(skb)->data_offset; 1991 } 1992 1993 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 1994 { 1995 return skb->len - NAPI_GRO_CB(skb)->data_offset; 1996 } 1997 1998 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 1999 { 2000 NAPI_GRO_CB(skb)->data_offset += len; 2001 } 2002 2003 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2004 unsigned int offset) 2005 { 2006 return NAPI_GRO_CB(skb)->frag0 + offset; 2007 } 2008 2009 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2010 { 2011 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2012 } 2013 2014 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2015 unsigned int offset) 2016 { 2017 if (!pskb_may_pull(skb, hlen)) 2018 return NULL; 2019 2020 NAPI_GRO_CB(skb)->frag0 = NULL; 2021 NAPI_GRO_CB(skb)->frag0_len = 0; 2022 return skb->data + offset; 2023 } 2024 2025 static inline void *skb_gro_network_header(struct sk_buff *skb) 2026 { 2027 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2028 skb_network_offset(skb); 2029 } 2030 2031 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2032 const void *start, unsigned int len) 2033 { 2034 if (skb->ip_summed == CHECKSUM_COMPLETE) 2035 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2036 csum_partial(start, len, 0)); 2037 } 2038 2039 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2040 unsigned short type, 2041 const void *daddr, const void *saddr, 2042 unsigned int len) 2043 { 2044 if (!dev->header_ops || !dev->header_ops->create) 2045 return 0; 2046 2047 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2048 } 2049 2050 static inline int dev_parse_header(const struct sk_buff *skb, 2051 unsigned char *haddr) 2052 { 2053 const struct net_device *dev = skb->dev; 2054 2055 if (!dev->header_ops || !dev->header_ops->parse) 2056 return 0; 2057 return dev->header_ops->parse(skb, haddr); 2058 } 2059 2060 static inline int dev_rebuild_header(struct sk_buff *skb) 2061 { 2062 const struct net_device *dev = skb->dev; 2063 2064 if (!dev->header_ops || !dev->header_ops->rebuild) 2065 return 0; 2066 return dev->header_ops->rebuild(skb); 2067 } 2068 2069 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len); 2070 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2071 static inline int unregister_gifconf(unsigned int family) 2072 { 2073 return register_gifconf(family, NULL); 2074 } 2075 2076 #ifdef CONFIG_NET_FLOW_LIMIT 2077 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2078 struct sd_flow_limit { 2079 u64 count; 2080 unsigned int num_buckets; 2081 unsigned int history_head; 2082 u16 history[FLOW_LIMIT_HISTORY]; 2083 u8 buckets[]; 2084 }; 2085 2086 extern int netdev_flow_limit_table_len; 2087 #endif /* CONFIG_NET_FLOW_LIMIT */ 2088 2089 /* 2090 * Incoming packets are placed on per-cpu queues 2091 */ 2092 struct softnet_data { 2093 struct Qdisc *output_queue; 2094 struct Qdisc **output_queue_tailp; 2095 struct list_head poll_list; 2096 struct sk_buff *completion_queue; 2097 struct sk_buff_head process_queue; 2098 2099 /* stats */ 2100 unsigned int processed; 2101 unsigned int time_squeeze; 2102 unsigned int cpu_collision; 2103 unsigned int received_rps; 2104 2105 #ifdef CONFIG_RPS 2106 struct softnet_data *rps_ipi_list; 2107 2108 /* Elements below can be accessed between CPUs for RPS */ 2109 struct call_single_data csd ____cacheline_aligned_in_smp; 2110 struct softnet_data *rps_ipi_next; 2111 unsigned int cpu; 2112 unsigned int input_queue_head; 2113 unsigned int input_queue_tail; 2114 #endif 2115 unsigned int dropped; 2116 struct sk_buff_head input_pkt_queue; 2117 struct napi_struct backlog; 2118 2119 #ifdef CONFIG_NET_FLOW_LIMIT 2120 struct sd_flow_limit __rcu *flow_limit; 2121 #endif 2122 }; 2123 2124 static inline void input_queue_head_incr(struct softnet_data *sd) 2125 { 2126 #ifdef CONFIG_RPS 2127 sd->input_queue_head++; 2128 #endif 2129 } 2130 2131 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 2132 unsigned int *qtail) 2133 { 2134 #ifdef CONFIG_RPS 2135 *qtail = ++sd->input_queue_tail; 2136 #endif 2137 } 2138 2139 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 2140 2141 void __netif_schedule(struct Qdisc *q); 2142 2143 static inline void netif_schedule_queue(struct netdev_queue *txq) 2144 { 2145 if (!(txq->state & QUEUE_STATE_ANY_XOFF)) 2146 __netif_schedule(txq->qdisc); 2147 } 2148 2149 static inline void netif_tx_schedule_all(struct net_device *dev) 2150 { 2151 unsigned int i; 2152 2153 for (i = 0; i < dev->num_tx_queues; i++) 2154 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 2155 } 2156 2157 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 2158 { 2159 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2160 } 2161 2162 /** 2163 * netif_start_queue - allow transmit 2164 * @dev: network device 2165 * 2166 * Allow upper layers to call the device hard_start_xmit routine. 2167 */ 2168 static inline void netif_start_queue(struct net_device *dev) 2169 { 2170 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 2171 } 2172 2173 static inline void netif_tx_start_all_queues(struct net_device *dev) 2174 { 2175 unsigned int i; 2176 2177 for (i = 0; i < dev->num_tx_queues; i++) { 2178 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2179 netif_tx_start_queue(txq); 2180 } 2181 } 2182 2183 static inline void netif_tx_wake_queue(struct netdev_queue *dev_queue) 2184 { 2185 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state)) 2186 __netif_schedule(dev_queue->qdisc); 2187 } 2188 2189 /** 2190 * netif_wake_queue - restart transmit 2191 * @dev: network device 2192 * 2193 * Allow upper layers to call the device hard_start_xmit routine. 2194 * Used for flow control when transmit resources are available. 2195 */ 2196 static inline void netif_wake_queue(struct net_device *dev) 2197 { 2198 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 2199 } 2200 2201 static inline void netif_tx_wake_all_queues(struct net_device *dev) 2202 { 2203 unsigned int i; 2204 2205 for (i = 0; i < dev->num_tx_queues; i++) { 2206 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2207 netif_tx_wake_queue(txq); 2208 } 2209 } 2210 2211 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 2212 { 2213 if (WARN_ON(!dev_queue)) { 2214 pr_info("netif_stop_queue() cannot be called before register_netdev()\n"); 2215 return; 2216 } 2217 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2218 } 2219 2220 /** 2221 * netif_stop_queue - stop transmitted packets 2222 * @dev: network device 2223 * 2224 * Stop upper layers calling the device hard_start_xmit routine. 2225 * Used for flow control when transmit resources are unavailable. 2226 */ 2227 static inline void netif_stop_queue(struct net_device *dev) 2228 { 2229 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 2230 } 2231 2232 static inline void netif_tx_stop_all_queues(struct net_device *dev) 2233 { 2234 unsigned int i; 2235 2236 for (i = 0; i < dev->num_tx_queues; i++) { 2237 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2238 netif_tx_stop_queue(txq); 2239 } 2240 } 2241 2242 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 2243 { 2244 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2245 } 2246 2247 /** 2248 * netif_queue_stopped - test if transmit queue is flowblocked 2249 * @dev: network device 2250 * 2251 * Test if transmit queue on device is currently unable to send. 2252 */ 2253 static inline bool netif_queue_stopped(const struct net_device *dev) 2254 { 2255 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 2256 } 2257 2258 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 2259 { 2260 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 2261 } 2262 2263 static inline bool 2264 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 2265 { 2266 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 2267 } 2268 2269 static inline bool 2270 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 2271 { 2272 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 2273 } 2274 2275 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 2276 unsigned int bytes) 2277 { 2278 #ifdef CONFIG_BQL 2279 dql_queued(&dev_queue->dql, bytes); 2280 2281 if (likely(dql_avail(&dev_queue->dql) >= 0)) 2282 return; 2283 2284 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2285 2286 /* 2287 * The XOFF flag must be set before checking the dql_avail below, 2288 * because in netdev_tx_completed_queue we update the dql_completed 2289 * before checking the XOFF flag. 2290 */ 2291 smp_mb(); 2292 2293 /* check again in case another CPU has just made room avail */ 2294 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 2295 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2296 #endif 2297 } 2298 2299 /** 2300 * netdev_sent_queue - report the number of bytes queued to hardware 2301 * @dev: network device 2302 * @bytes: number of bytes queued to the hardware device queue 2303 * 2304 * Report the number of bytes queued for sending/completion to the network 2305 * device hardware queue. @bytes should be a good approximation and should 2306 * exactly match netdev_completed_queue() @bytes 2307 */ 2308 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 2309 { 2310 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 2311 } 2312 2313 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 2314 unsigned int pkts, unsigned int bytes) 2315 { 2316 #ifdef CONFIG_BQL 2317 if (unlikely(!bytes)) 2318 return; 2319 2320 dql_completed(&dev_queue->dql, bytes); 2321 2322 /* 2323 * Without the memory barrier there is a small possiblity that 2324 * netdev_tx_sent_queue will miss the update and cause the queue to 2325 * be stopped forever 2326 */ 2327 smp_mb(); 2328 2329 if (dql_avail(&dev_queue->dql) < 0) 2330 return; 2331 2332 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 2333 netif_schedule_queue(dev_queue); 2334 #endif 2335 } 2336 2337 /** 2338 * netdev_completed_queue - report bytes and packets completed by device 2339 * @dev: network device 2340 * @pkts: actual number of packets sent over the medium 2341 * @bytes: actual number of bytes sent over the medium 2342 * 2343 * Report the number of bytes and packets transmitted by the network device 2344 * hardware queue over the physical medium, @bytes must exactly match the 2345 * @bytes amount passed to netdev_sent_queue() 2346 */ 2347 static inline void netdev_completed_queue(struct net_device *dev, 2348 unsigned int pkts, unsigned int bytes) 2349 { 2350 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 2351 } 2352 2353 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 2354 { 2355 #ifdef CONFIG_BQL 2356 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 2357 dql_reset(&q->dql); 2358 #endif 2359 } 2360 2361 /** 2362 * netdev_reset_queue - reset the packets and bytes count of a network device 2363 * @dev_queue: network device 2364 * 2365 * Reset the bytes and packet count of a network device and clear the 2366 * software flow control OFF bit for this network device 2367 */ 2368 static inline void netdev_reset_queue(struct net_device *dev_queue) 2369 { 2370 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 2371 } 2372 2373 /** 2374 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 2375 * @dev: network device 2376 * @queue_index: given tx queue index 2377 * 2378 * Returns 0 if given tx queue index >= number of device tx queues, 2379 * otherwise returns the originally passed tx queue index. 2380 */ 2381 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 2382 { 2383 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 2384 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 2385 dev->name, queue_index, 2386 dev->real_num_tx_queues); 2387 return 0; 2388 } 2389 2390 return queue_index; 2391 } 2392 2393 /** 2394 * netif_running - test if up 2395 * @dev: network device 2396 * 2397 * Test if the device has been brought up. 2398 */ 2399 static inline bool netif_running(const struct net_device *dev) 2400 { 2401 return test_bit(__LINK_STATE_START, &dev->state); 2402 } 2403 2404 /* 2405 * Routines to manage the subqueues on a device. We only need start 2406 * stop, and a check if it's stopped. All other device management is 2407 * done at the overall netdevice level. 2408 * Also test the device if we're multiqueue. 2409 */ 2410 2411 /** 2412 * netif_start_subqueue - allow sending packets on subqueue 2413 * @dev: network device 2414 * @queue_index: sub queue index 2415 * 2416 * Start individual transmit queue of a device with multiple transmit queues. 2417 */ 2418 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 2419 { 2420 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2421 2422 netif_tx_start_queue(txq); 2423 } 2424 2425 /** 2426 * netif_stop_subqueue - stop sending packets on subqueue 2427 * @dev: network device 2428 * @queue_index: sub queue index 2429 * 2430 * Stop individual transmit queue of a device with multiple transmit queues. 2431 */ 2432 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 2433 { 2434 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2435 netif_tx_stop_queue(txq); 2436 } 2437 2438 /** 2439 * netif_subqueue_stopped - test status of subqueue 2440 * @dev: network device 2441 * @queue_index: sub queue index 2442 * 2443 * Check individual transmit queue of a device with multiple transmit queues. 2444 */ 2445 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 2446 u16 queue_index) 2447 { 2448 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2449 2450 return netif_tx_queue_stopped(txq); 2451 } 2452 2453 static inline bool netif_subqueue_stopped(const struct net_device *dev, 2454 struct sk_buff *skb) 2455 { 2456 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 2457 } 2458 2459 /** 2460 * netif_wake_subqueue - allow sending packets on subqueue 2461 * @dev: network device 2462 * @queue_index: sub queue index 2463 * 2464 * Resume individual transmit queue of a device with multiple transmit queues. 2465 */ 2466 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 2467 { 2468 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2469 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state)) 2470 __netif_schedule(txq->qdisc); 2471 } 2472 2473 #ifdef CONFIG_XPS 2474 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 2475 u16 index); 2476 #else 2477 static inline int netif_set_xps_queue(struct net_device *dev, 2478 const struct cpumask *mask, 2479 u16 index) 2480 { 2481 return 0; 2482 } 2483 #endif 2484 2485 /* 2486 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used 2487 * as a distribution range limit for the returned value. 2488 */ 2489 static inline u16 skb_tx_hash(const struct net_device *dev, 2490 const struct sk_buff *skb) 2491 { 2492 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues); 2493 } 2494 2495 /** 2496 * netif_is_multiqueue - test if device has multiple transmit queues 2497 * @dev: network device 2498 * 2499 * Check if device has multiple transmit queues 2500 */ 2501 static inline bool netif_is_multiqueue(const struct net_device *dev) 2502 { 2503 return dev->num_tx_queues > 1; 2504 } 2505 2506 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 2507 2508 #ifdef CONFIG_SYSFS 2509 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 2510 #else 2511 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 2512 unsigned int rxq) 2513 { 2514 return 0; 2515 } 2516 #endif 2517 2518 static inline int netif_copy_real_num_queues(struct net_device *to_dev, 2519 const struct net_device *from_dev) 2520 { 2521 int err; 2522 2523 err = netif_set_real_num_tx_queues(to_dev, 2524 from_dev->real_num_tx_queues); 2525 if (err) 2526 return err; 2527 #ifdef CONFIG_SYSFS 2528 return netif_set_real_num_rx_queues(to_dev, 2529 from_dev->real_num_rx_queues); 2530 #else 2531 return 0; 2532 #endif 2533 } 2534 2535 #ifdef CONFIG_SYSFS 2536 static inline unsigned int get_netdev_rx_queue_index( 2537 struct netdev_rx_queue *queue) 2538 { 2539 struct net_device *dev = queue->dev; 2540 int index = queue - dev->_rx; 2541 2542 BUG_ON(index >= dev->num_rx_queues); 2543 return index; 2544 } 2545 #endif 2546 2547 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 2548 int netif_get_num_default_rss_queues(void); 2549 2550 enum skb_free_reason { 2551 SKB_REASON_CONSUMED, 2552 SKB_REASON_DROPPED, 2553 }; 2554 2555 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 2556 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 2557 2558 /* 2559 * It is not allowed to call kfree_skb() or consume_skb() from hardware 2560 * interrupt context or with hardware interrupts being disabled. 2561 * (in_irq() || irqs_disabled()) 2562 * 2563 * We provide four helpers that can be used in following contexts : 2564 * 2565 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 2566 * replacing kfree_skb(skb) 2567 * 2568 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 2569 * Typically used in place of consume_skb(skb) in TX completion path 2570 * 2571 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 2572 * replacing kfree_skb(skb) 2573 * 2574 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 2575 * and consumed a packet. Used in place of consume_skb(skb) 2576 */ 2577 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 2578 { 2579 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 2580 } 2581 2582 static inline void dev_consume_skb_irq(struct sk_buff *skb) 2583 { 2584 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 2585 } 2586 2587 static inline void dev_kfree_skb_any(struct sk_buff *skb) 2588 { 2589 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 2590 } 2591 2592 static inline void dev_consume_skb_any(struct sk_buff *skb) 2593 { 2594 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 2595 } 2596 2597 int netif_rx(struct sk_buff *skb); 2598 int netif_rx_ni(struct sk_buff *skb); 2599 int netif_receive_skb(struct sk_buff *skb); 2600 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 2601 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 2602 struct sk_buff *napi_get_frags(struct napi_struct *napi); 2603 gro_result_t napi_gro_frags(struct napi_struct *napi); 2604 struct packet_offload *gro_find_receive_by_type(__be16 type); 2605 struct packet_offload *gro_find_complete_by_type(__be16 type); 2606 2607 static inline void napi_free_frags(struct napi_struct *napi) 2608 { 2609 kfree_skb(napi->skb); 2610 napi->skb = NULL; 2611 } 2612 2613 int netdev_rx_handler_register(struct net_device *dev, 2614 rx_handler_func_t *rx_handler, 2615 void *rx_handler_data); 2616 void netdev_rx_handler_unregister(struct net_device *dev); 2617 2618 bool dev_valid_name(const char *name); 2619 int dev_ioctl(struct net *net, unsigned int cmd, void __user *); 2620 int dev_ethtool(struct net *net, struct ifreq *); 2621 unsigned int dev_get_flags(const struct net_device *); 2622 int __dev_change_flags(struct net_device *, unsigned int flags); 2623 int dev_change_flags(struct net_device *, unsigned int); 2624 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 2625 unsigned int gchanges); 2626 int dev_change_name(struct net_device *, const char *); 2627 int dev_set_alias(struct net_device *, const char *, size_t); 2628 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 2629 int dev_set_mtu(struct net_device *, int); 2630 void dev_set_group(struct net_device *, int); 2631 int dev_set_mac_address(struct net_device *, struct sockaddr *); 2632 int dev_change_carrier(struct net_device *, bool new_carrier); 2633 int dev_get_phys_port_id(struct net_device *dev, 2634 struct netdev_phys_port_id *ppid); 2635 int dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 2636 struct netdev_queue *txq); 2637 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 2638 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb); 2639 2640 extern int netdev_budget; 2641 2642 /* Called by rtnetlink.c:rtnl_unlock() */ 2643 void netdev_run_todo(void); 2644 2645 /** 2646 * dev_put - release reference to device 2647 * @dev: network device 2648 * 2649 * Release reference to device to allow it to be freed. 2650 */ 2651 static inline void dev_put(struct net_device *dev) 2652 { 2653 this_cpu_dec(*dev->pcpu_refcnt); 2654 } 2655 2656 /** 2657 * dev_hold - get reference to device 2658 * @dev: network device 2659 * 2660 * Hold reference to device to keep it from being freed. 2661 */ 2662 static inline void dev_hold(struct net_device *dev) 2663 { 2664 this_cpu_inc(*dev->pcpu_refcnt); 2665 } 2666 2667 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 2668 * and _off may be called from IRQ context, but it is caller 2669 * who is responsible for serialization of these calls. 2670 * 2671 * The name carrier is inappropriate, these functions should really be 2672 * called netif_lowerlayer_*() because they represent the state of any 2673 * kind of lower layer not just hardware media. 2674 */ 2675 2676 void linkwatch_init_dev(struct net_device *dev); 2677 void linkwatch_fire_event(struct net_device *dev); 2678 void linkwatch_forget_dev(struct net_device *dev); 2679 2680 /** 2681 * netif_carrier_ok - test if carrier present 2682 * @dev: network device 2683 * 2684 * Check if carrier is present on device 2685 */ 2686 static inline bool netif_carrier_ok(const struct net_device *dev) 2687 { 2688 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 2689 } 2690 2691 unsigned long dev_trans_start(struct net_device *dev); 2692 2693 void __netdev_watchdog_up(struct net_device *dev); 2694 2695 void netif_carrier_on(struct net_device *dev); 2696 2697 void netif_carrier_off(struct net_device *dev); 2698 2699 /** 2700 * netif_dormant_on - mark device as dormant. 2701 * @dev: network device 2702 * 2703 * Mark device as dormant (as per RFC2863). 2704 * 2705 * The dormant state indicates that the relevant interface is not 2706 * actually in a condition to pass packets (i.e., it is not 'up') but is 2707 * in a "pending" state, waiting for some external event. For "on- 2708 * demand" interfaces, this new state identifies the situation where the 2709 * interface is waiting for events to place it in the up state. 2710 * 2711 */ 2712 static inline void netif_dormant_on(struct net_device *dev) 2713 { 2714 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 2715 linkwatch_fire_event(dev); 2716 } 2717 2718 /** 2719 * netif_dormant_off - set device as not dormant. 2720 * @dev: network device 2721 * 2722 * Device is not in dormant state. 2723 */ 2724 static inline void netif_dormant_off(struct net_device *dev) 2725 { 2726 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 2727 linkwatch_fire_event(dev); 2728 } 2729 2730 /** 2731 * netif_dormant - test if carrier present 2732 * @dev: network device 2733 * 2734 * Check if carrier is present on device 2735 */ 2736 static inline bool netif_dormant(const struct net_device *dev) 2737 { 2738 return test_bit(__LINK_STATE_DORMANT, &dev->state); 2739 } 2740 2741 2742 /** 2743 * netif_oper_up - test if device is operational 2744 * @dev: network device 2745 * 2746 * Check if carrier is operational 2747 */ 2748 static inline bool netif_oper_up(const struct net_device *dev) 2749 { 2750 return (dev->operstate == IF_OPER_UP || 2751 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 2752 } 2753 2754 /** 2755 * netif_device_present - is device available or removed 2756 * @dev: network device 2757 * 2758 * Check if device has not been removed from system. 2759 */ 2760 static inline bool netif_device_present(struct net_device *dev) 2761 { 2762 return test_bit(__LINK_STATE_PRESENT, &dev->state); 2763 } 2764 2765 void netif_device_detach(struct net_device *dev); 2766 2767 void netif_device_attach(struct net_device *dev); 2768 2769 /* 2770 * Network interface message level settings 2771 */ 2772 2773 enum { 2774 NETIF_MSG_DRV = 0x0001, 2775 NETIF_MSG_PROBE = 0x0002, 2776 NETIF_MSG_LINK = 0x0004, 2777 NETIF_MSG_TIMER = 0x0008, 2778 NETIF_MSG_IFDOWN = 0x0010, 2779 NETIF_MSG_IFUP = 0x0020, 2780 NETIF_MSG_RX_ERR = 0x0040, 2781 NETIF_MSG_TX_ERR = 0x0080, 2782 NETIF_MSG_TX_QUEUED = 0x0100, 2783 NETIF_MSG_INTR = 0x0200, 2784 NETIF_MSG_TX_DONE = 0x0400, 2785 NETIF_MSG_RX_STATUS = 0x0800, 2786 NETIF_MSG_PKTDATA = 0x1000, 2787 NETIF_MSG_HW = 0x2000, 2788 NETIF_MSG_WOL = 0x4000, 2789 }; 2790 2791 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 2792 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 2793 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 2794 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 2795 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 2796 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 2797 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 2798 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 2799 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 2800 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 2801 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 2802 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 2803 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 2804 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 2805 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 2806 2807 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 2808 { 2809 /* use default */ 2810 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 2811 return default_msg_enable_bits; 2812 if (debug_value == 0) /* no output */ 2813 return 0; 2814 /* set low N bits */ 2815 return (1 << debug_value) - 1; 2816 } 2817 2818 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 2819 { 2820 spin_lock(&txq->_xmit_lock); 2821 txq->xmit_lock_owner = cpu; 2822 } 2823 2824 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 2825 { 2826 spin_lock_bh(&txq->_xmit_lock); 2827 txq->xmit_lock_owner = smp_processor_id(); 2828 } 2829 2830 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 2831 { 2832 bool ok = spin_trylock(&txq->_xmit_lock); 2833 if (likely(ok)) 2834 txq->xmit_lock_owner = smp_processor_id(); 2835 return ok; 2836 } 2837 2838 static inline void __netif_tx_unlock(struct netdev_queue *txq) 2839 { 2840 txq->xmit_lock_owner = -1; 2841 spin_unlock(&txq->_xmit_lock); 2842 } 2843 2844 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 2845 { 2846 txq->xmit_lock_owner = -1; 2847 spin_unlock_bh(&txq->_xmit_lock); 2848 } 2849 2850 static inline void txq_trans_update(struct netdev_queue *txq) 2851 { 2852 if (txq->xmit_lock_owner != -1) 2853 txq->trans_start = jiffies; 2854 } 2855 2856 /** 2857 * netif_tx_lock - grab network device transmit lock 2858 * @dev: network device 2859 * 2860 * Get network device transmit lock 2861 */ 2862 static inline void netif_tx_lock(struct net_device *dev) 2863 { 2864 unsigned int i; 2865 int cpu; 2866 2867 spin_lock(&dev->tx_global_lock); 2868 cpu = smp_processor_id(); 2869 for (i = 0; i < dev->num_tx_queues; i++) { 2870 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2871 2872 /* We are the only thread of execution doing a 2873 * freeze, but we have to grab the _xmit_lock in 2874 * order to synchronize with threads which are in 2875 * the ->hard_start_xmit() handler and already 2876 * checked the frozen bit. 2877 */ 2878 __netif_tx_lock(txq, cpu); 2879 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 2880 __netif_tx_unlock(txq); 2881 } 2882 } 2883 2884 static inline void netif_tx_lock_bh(struct net_device *dev) 2885 { 2886 local_bh_disable(); 2887 netif_tx_lock(dev); 2888 } 2889 2890 static inline void netif_tx_unlock(struct net_device *dev) 2891 { 2892 unsigned int i; 2893 2894 for (i = 0; i < dev->num_tx_queues; i++) { 2895 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2896 2897 /* No need to grab the _xmit_lock here. If the 2898 * queue is not stopped for another reason, we 2899 * force a schedule. 2900 */ 2901 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 2902 netif_schedule_queue(txq); 2903 } 2904 spin_unlock(&dev->tx_global_lock); 2905 } 2906 2907 static inline void netif_tx_unlock_bh(struct net_device *dev) 2908 { 2909 netif_tx_unlock(dev); 2910 local_bh_enable(); 2911 } 2912 2913 #define HARD_TX_LOCK(dev, txq, cpu) { \ 2914 if ((dev->features & NETIF_F_LLTX) == 0) { \ 2915 __netif_tx_lock(txq, cpu); \ 2916 } \ 2917 } 2918 2919 #define HARD_TX_TRYLOCK(dev, txq) \ 2920 (((dev->features & NETIF_F_LLTX) == 0) ? \ 2921 __netif_tx_trylock(txq) : \ 2922 true ) 2923 2924 #define HARD_TX_UNLOCK(dev, txq) { \ 2925 if ((dev->features & NETIF_F_LLTX) == 0) { \ 2926 __netif_tx_unlock(txq); \ 2927 } \ 2928 } 2929 2930 static inline void netif_tx_disable(struct net_device *dev) 2931 { 2932 unsigned int i; 2933 int cpu; 2934 2935 local_bh_disable(); 2936 cpu = smp_processor_id(); 2937 for (i = 0; i < dev->num_tx_queues; i++) { 2938 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2939 2940 __netif_tx_lock(txq, cpu); 2941 netif_tx_stop_queue(txq); 2942 __netif_tx_unlock(txq); 2943 } 2944 local_bh_enable(); 2945 } 2946 2947 static inline void netif_addr_lock(struct net_device *dev) 2948 { 2949 spin_lock(&dev->addr_list_lock); 2950 } 2951 2952 static inline void netif_addr_lock_nested(struct net_device *dev) 2953 { 2954 int subclass = SINGLE_DEPTH_NESTING; 2955 2956 if (dev->netdev_ops->ndo_get_lock_subclass) 2957 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev); 2958 2959 spin_lock_nested(&dev->addr_list_lock, subclass); 2960 } 2961 2962 static inline void netif_addr_lock_bh(struct net_device *dev) 2963 { 2964 spin_lock_bh(&dev->addr_list_lock); 2965 } 2966 2967 static inline void netif_addr_unlock(struct net_device *dev) 2968 { 2969 spin_unlock(&dev->addr_list_lock); 2970 } 2971 2972 static inline void netif_addr_unlock_bh(struct net_device *dev) 2973 { 2974 spin_unlock_bh(&dev->addr_list_lock); 2975 } 2976 2977 /* 2978 * dev_addrs walker. Should be used only for read access. Call with 2979 * rcu_read_lock held. 2980 */ 2981 #define for_each_dev_addr(dev, ha) \ 2982 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 2983 2984 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 2985 2986 void ether_setup(struct net_device *dev); 2987 2988 /* Support for loadable net-drivers */ 2989 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 2990 void (*setup)(struct net_device *), 2991 unsigned int txqs, unsigned int rxqs); 2992 #define alloc_netdev(sizeof_priv, name, setup) \ 2993 alloc_netdev_mqs(sizeof_priv, name, setup, 1, 1) 2994 2995 #define alloc_netdev_mq(sizeof_priv, name, setup, count) \ 2996 alloc_netdev_mqs(sizeof_priv, name, setup, count, count) 2997 2998 int register_netdev(struct net_device *dev); 2999 void unregister_netdev(struct net_device *dev); 3000 3001 /* General hardware address lists handling functions */ 3002 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 3003 struct netdev_hw_addr_list *from_list, int addr_len); 3004 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 3005 struct netdev_hw_addr_list *from_list, int addr_len); 3006 void __hw_addr_init(struct netdev_hw_addr_list *list); 3007 3008 /* Functions used for device addresses handling */ 3009 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 3010 unsigned char addr_type); 3011 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 3012 unsigned char addr_type); 3013 void dev_addr_flush(struct net_device *dev); 3014 int dev_addr_init(struct net_device *dev); 3015 3016 /* Functions used for unicast addresses handling */ 3017 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 3018 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 3019 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 3020 int dev_uc_sync(struct net_device *to, struct net_device *from); 3021 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 3022 void dev_uc_unsync(struct net_device *to, struct net_device *from); 3023 void dev_uc_flush(struct net_device *dev); 3024 void dev_uc_init(struct net_device *dev); 3025 3026 /* Functions used for multicast addresses handling */ 3027 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 3028 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 3029 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 3030 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 3031 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 3032 int dev_mc_sync(struct net_device *to, struct net_device *from); 3033 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 3034 void dev_mc_unsync(struct net_device *to, struct net_device *from); 3035 void dev_mc_flush(struct net_device *dev); 3036 void dev_mc_init(struct net_device *dev); 3037 3038 /* Functions used for secondary unicast and multicast support */ 3039 void dev_set_rx_mode(struct net_device *dev); 3040 void __dev_set_rx_mode(struct net_device *dev); 3041 int dev_set_promiscuity(struct net_device *dev, int inc); 3042 int dev_set_allmulti(struct net_device *dev, int inc); 3043 void netdev_state_change(struct net_device *dev); 3044 void netdev_notify_peers(struct net_device *dev); 3045 void netdev_features_change(struct net_device *dev); 3046 /* Load a device via the kmod */ 3047 void dev_load(struct net *net, const char *name); 3048 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 3049 struct rtnl_link_stats64 *storage); 3050 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 3051 const struct net_device_stats *netdev_stats); 3052 3053 extern int netdev_max_backlog; 3054 extern int netdev_tstamp_prequeue; 3055 extern int weight_p; 3056 extern int bpf_jit_enable; 3057 3058 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 3059 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 3060 struct list_head **iter); 3061 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 3062 struct list_head **iter); 3063 3064 /* iterate through upper list, must be called under RCU read lock */ 3065 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 3066 for (iter = &(dev)->adj_list.upper, \ 3067 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 3068 updev; \ 3069 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 3070 3071 /* iterate through upper list, must be called under RCU read lock */ 3072 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \ 3073 for (iter = &(dev)->all_adj_list.upper, \ 3074 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \ 3075 updev; \ 3076 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter))) 3077 3078 void *netdev_lower_get_next_private(struct net_device *dev, 3079 struct list_head **iter); 3080 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 3081 struct list_head **iter); 3082 3083 #define netdev_for_each_lower_private(dev, priv, iter) \ 3084 for (iter = (dev)->adj_list.lower.next, \ 3085 priv = netdev_lower_get_next_private(dev, &(iter)); \ 3086 priv; \ 3087 priv = netdev_lower_get_next_private(dev, &(iter))) 3088 3089 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 3090 for (iter = &(dev)->adj_list.lower, \ 3091 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 3092 priv; \ 3093 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 3094 3095 void *netdev_lower_get_next(struct net_device *dev, 3096 struct list_head **iter); 3097 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 3098 for (iter = &(dev)->adj_list.lower, \ 3099 ldev = netdev_lower_get_next(dev, &(iter)); \ 3100 ldev; \ 3101 ldev = netdev_lower_get_next(dev, &(iter))) 3102 3103 void *netdev_adjacent_get_private(struct list_head *adj_list); 3104 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 3105 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 3106 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 3107 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev); 3108 int netdev_master_upper_dev_link(struct net_device *dev, 3109 struct net_device *upper_dev); 3110 int netdev_master_upper_dev_link_private(struct net_device *dev, 3111 struct net_device *upper_dev, 3112 void *private); 3113 void netdev_upper_dev_unlink(struct net_device *dev, 3114 struct net_device *upper_dev); 3115 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 3116 void *netdev_lower_dev_get_private(struct net_device *dev, 3117 struct net_device *lower_dev); 3118 int dev_get_nest_level(struct net_device *dev, 3119 bool (*type_check)(struct net_device *dev)); 3120 int skb_checksum_help(struct sk_buff *skb); 3121 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 3122 netdev_features_t features, bool tx_path); 3123 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 3124 netdev_features_t features); 3125 3126 static inline 3127 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 3128 { 3129 return __skb_gso_segment(skb, features, true); 3130 } 3131 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 3132 3133 static inline bool can_checksum_protocol(netdev_features_t features, 3134 __be16 protocol) 3135 { 3136 return ((features & NETIF_F_GEN_CSUM) || 3137 ((features & NETIF_F_V4_CSUM) && 3138 protocol == htons(ETH_P_IP)) || 3139 ((features & NETIF_F_V6_CSUM) && 3140 protocol == htons(ETH_P_IPV6)) || 3141 ((features & NETIF_F_FCOE_CRC) && 3142 protocol == htons(ETH_P_FCOE))); 3143 } 3144 3145 #ifdef CONFIG_BUG 3146 void netdev_rx_csum_fault(struct net_device *dev); 3147 #else 3148 static inline void netdev_rx_csum_fault(struct net_device *dev) 3149 { 3150 } 3151 #endif 3152 /* rx skb timestamps */ 3153 void net_enable_timestamp(void); 3154 void net_disable_timestamp(void); 3155 3156 #ifdef CONFIG_PROC_FS 3157 int __init dev_proc_init(void); 3158 #else 3159 #define dev_proc_init() 0 3160 #endif 3161 3162 int netdev_class_create_file_ns(struct class_attribute *class_attr, 3163 const void *ns); 3164 void netdev_class_remove_file_ns(struct class_attribute *class_attr, 3165 const void *ns); 3166 3167 static inline int netdev_class_create_file(struct class_attribute *class_attr) 3168 { 3169 return netdev_class_create_file_ns(class_attr, NULL); 3170 } 3171 3172 static inline void netdev_class_remove_file(struct class_attribute *class_attr) 3173 { 3174 netdev_class_remove_file_ns(class_attr, NULL); 3175 } 3176 3177 extern struct kobj_ns_type_operations net_ns_type_operations; 3178 3179 const char *netdev_drivername(const struct net_device *dev); 3180 3181 void linkwatch_run_queue(void); 3182 3183 static inline netdev_features_t netdev_get_wanted_features( 3184 struct net_device *dev) 3185 { 3186 return (dev->features & ~dev->hw_features) | dev->wanted_features; 3187 } 3188 netdev_features_t netdev_increment_features(netdev_features_t all, 3189 netdev_features_t one, netdev_features_t mask); 3190 3191 /* Allow TSO being used on stacked device : 3192 * Performing the GSO segmentation before last device 3193 * is a performance improvement. 3194 */ 3195 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 3196 netdev_features_t mask) 3197 { 3198 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 3199 } 3200 3201 int __netdev_update_features(struct net_device *dev); 3202 void netdev_update_features(struct net_device *dev); 3203 void netdev_change_features(struct net_device *dev); 3204 3205 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 3206 struct net_device *dev); 3207 3208 netdev_features_t netif_skb_features(struct sk_buff *skb); 3209 3210 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 3211 { 3212 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT; 3213 3214 /* check flags correspondence */ 3215 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 3216 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT)); 3217 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 3218 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 3219 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 3220 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 3221 3222 return (features & feature) == feature; 3223 } 3224 3225 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 3226 { 3227 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 3228 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 3229 } 3230 3231 static inline bool netif_needs_gso(struct sk_buff *skb, 3232 netdev_features_t features) 3233 { 3234 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 3235 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 3236 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 3237 } 3238 3239 static inline void netif_set_gso_max_size(struct net_device *dev, 3240 unsigned int size) 3241 { 3242 dev->gso_max_size = size; 3243 } 3244 3245 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 3246 int pulled_hlen, u16 mac_offset, 3247 int mac_len) 3248 { 3249 skb->protocol = protocol; 3250 skb->encapsulation = 1; 3251 skb_push(skb, pulled_hlen); 3252 skb_reset_transport_header(skb); 3253 skb->mac_header = mac_offset; 3254 skb->network_header = skb->mac_header + mac_len; 3255 skb->mac_len = mac_len; 3256 } 3257 3258 static inline bool netif_is_macvlan(struct net_device *dev) 3259 { 3260 return dev->priv_flags & IFF_MACVLAN; 3261 } 3262 3263 static inline bool netif_is_bond_master(struct net_device *dev) 3264 { 3265 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 3266 } 3267 3268 static inline bool netif_is_bond_slave(struct net_device *dev) 3269 { 3270 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 3271 } 3272 3273 static inline bool netif_supports_nofcs(struct net_device *dev) 3274 { 3275 return dev->priv_flags & IFF_SUPP_NOFCS; 3276 } 3277 3278 extern struct pernet_operations __net_initdata loopback_net_ops; 3279 3280 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 3281 3282 /* netdev_printk helpers, similar to dev_printk */ 3283 3284 static inline const char *netdev_name(const struct net_device *dev) 3285 { 3286 if (dev->reg_state != NETREG_REGISTERED) 3287 return "(unregistered net_device)"; 3288 return dev->name; 3289 } 3290 3291 __printf(3, 4) 3292 int netdev_printk(const char *level, const struct net_device *dev, 3293 const char *format, ...); 3294 __printf(2, 3) 3295 int netdev_emerg(const struct net_device *dev, const char *format, ...); 3296 __printf(2, 3) 3297 int netdev_alert(const struct net_device *dev, const char *format, ...); 3298 __printf(2, 3) 3299 int netdev_crit(const struct net_device *dev, const char *format, ...); 3300 __printf(2, 3) 3301 int netdev_err(const struct net_device *dev, const char *format, ...); 3302 __printf(2, 3) 3303 int netdev_warn(const struct net_device *dev, const char *format, ...); 3304 __printf(2, 3) 3305 int netdev_notice(const struct net_device *dev, const char *format, ...); 3306 __printf(2, 3) 3307 int netdev_info(const struct net_device *dev, const char *format, ...); 3308 3309 #define MODULE_ALIAS_NETDEV(device) \ 3310 MODULE_ALIAS("netdev-" device) 3311 3312 #if defined(CONFIG_DYNAMIC_DEBUG) 3313 #define netdev_dbg(__dev, format, args...) \ 3314 do { \ 3315 dynamic_netdev_dbg(__dev, format, ##args); \ 3316 } while (0) 3317 #elif defined(DEBUG) 3318 #define netdev_dbg(__dev, format, args...) \ 3319 netdev_printk(KERN_DEBUG, __dev, format, ##args) 3320 #else 3321 #define netdev_dbg(__dev, format, args...) \ 3322 ({ \ 3323 if (0) \ 3324 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 3325 0; \ 3326 }) 3327 #endif 3328 3329 #if defined(VERBOSE_DEBUG) 3330 #define netdev_vdbg netdev_dbg 3331 #else 3332 3333 #define netdev_vdbg(dev, format, args...) \ 3334 ({ \ 3335 if (0) \ 3336 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 3337 0; \ 3338 }) 3339 #endif 3340 3341 /* 3342 * netdev_WARN() acts like dev_printk(), but with the key difference 3343 * of using a WARN/WARN_ON to get the message out, including the 3344 * file/line information and a backtrace. 3345 */ 3346 #define netdev_WARN(dev, format, args...) \ 3347 WARN(1, "netdevice: %s\n" format, netdev_name(dev), ##args) 3348 3349 /* netif printk helpers, similar to netdev_printk */ 3350 3351 #define netif_printk(priv, type, level, dev, fmt, args...) \ 3352 do { \ 3353 if (netif_msg_##type(priv)) \ 3354 netdev_printk(level, (dev), fmt, ##args); \ 3355 } while (0) 3356 3357 #define netif_level(level, priv, type, dev, fmt, args...) \ 3358 do { \ 3359 if (netif_msg_##type(priv)) \ 3360 netdev_##level(dev, fmt, ##args); \ 3361 } while (0) 3362 3363 #define netif_emerg(priv, type, dev, fmt, args...) \ 3364 netif_level(emerg, priv, type, dev, fmt, ##args) 3365 #define netif_alert(priv, type, dev, fmt, args...) \ 3366 netif_level(alert, priv, type, dev, fmt, ##args) 3367 #define netif_crit(priv, type, dev, fmt, args...) \ 3368 netif_level(crit, priv, type, dev, fmt, ##args) 3369 #define netif_err(priv, type, dev, fmt, args...) \ 3370 netif_level(err, priv, type, dev, fmt, ##args) 3371 #define netif_warn(priv, type, dev, fmt, args...) \ 3372 netif_level(warn, priv, type, dev, fmt, ##args) 3373 #define netif_notice(priv, type, dev, fmt, args...) \ 3374 netif_level(notice, priv, type, dev, fmt, ##args) 3375 #define netif_info(priv, type, dev, fmt, args...) \ 3376 netif_level(info, priv, type, dev, fmt, ##args) 3377 3378 #if defined(CONFIG_DYNAMIC_DEBUG) 3379 #define netif_dbg(priv, type, netdev, format, args...) \ 3380 do { \ 3381 if (netif_msg_##type(priv)) \ 3382 dynamic_netdev_dbg(netdev, format, ##args); \ 3383 } while (0) 3384 #elif defined(DEBUG) 3385 #define netif_dbg(priv, type, dev, format, args...) \ 3386 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 3387 #else 3388 #define netif_dbg(priv, type, dev, format, args...) \ 3389 ({ \ 3390 if (0) \ 3391 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 3392 0; \ 3393 }) 3394 #endif 3395 3396 #if defined(VERBOSE_DEBUG) 3397 #define netif_vdbg netif_dbg 3398 #else 3399 #define netif_vdbg(priv, type, dev, format, args...) \ 3400 ({ \ 3401 if (0) \ 3402 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 3403 0; \ 3404 }) 3405 #endif 3406 3407 /* 3408 * The list of packet types we will receive (as opposed to discard) 3409 * and the routines to invoke. 3410 * 3411 * Why 16. Because with 16 the only overlap we get on a hash of the 3412 * low nibble of the protocol value is RARP/SNAP/X.25. 3413 * 3414 * NOTE: That is no longer true with the addition of VLAN tags. Not 3415 * sure which should go first, but I bet it won't make much 3416 * difference if we are running VLANs. The good news is that 3417 * this protocol won't be in the list unless compiled in, so 3418 * the average user (w/out VLANs) will not be adversely affected. 3419 * --BLG 3420 * 3421 * 0800 IP 3422 * 8100 802.1Q VLAN 3423 * 0001 802.3 3424 * 0002 AX.25 3425 * 0004 802.2 3426 * 8035 RARP 3427 * 0005 SNAP 3428 * 0805 X.25 3429 * 0806 ARP 3430 * 8137 IPX 3431 * 0009 Localtalk 3432 * 86DD IPv6 3433 */ 3434 #define PTYPE_HASH_SIZE (16) 3435 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 3436 3437 #endif /* _LINUX_NETDEVICE_H */ 3438