1 /* 2 * INET An implementation of the TCP/IP protocol suite for the LINUX 3 * operating system. INET is implemented using the BSD Socket 4 * interface as the means of communication with the user level. 5 * 6 * Definitions for the Interfaces handler. 7 * 8 * Version: @(#)dev.h 1.0.10 08/12/93 9 * 10 * Authors: Ross Biro 11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 12 * Corey Minyard <wf-rch!minyard@relay.EU.net> 13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov> 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk> 15 * Bjorn Ekwall. <bj0rn@blox.se> 16 * Pekka Riikonen <priikone@poseidon.pspt.fi> 17 * 18 * This program is free software; you can redistribute it and/or 19 * modify it under the terms of the GNU General Public License 20 * as published by the Free Software Foundation; either version 21 * 2 of the License, or (at your option) any later version. 22 * 23 * Moved to /usr/include/linux for NET3 24 */ 25 #ifndef _LINUX_NETDEVICE_H 26 #define _LINUX_NETDEVICE_H 27 28 #include <linux/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 <linux/prefetch.h> 34 #include <asm/cache.h> 35 #include <asm/byteorder.h> 36 37 #include <linux/percpu.h> 38 #include <linux/rculist.h> 39 #include <linux/dmaengine.h> 40 #include <linux/workqueue.h> 41 #include <linux/dynamic_queue_limits.h> 42 43 #include <linux/ethtool.h> 44 #include <net/net_namespace.h> 45 #include <net/dsa.h> 46 #ifdef CONFIG_DCB 47 #include <net/dcbnl.h> 48 #endif 49 #include <net/netprio_cgroup.h> 50 51 #include <linux/netdev_features.h> 52 #include <linux/neighbour.h> 53 #include <uapi/linux/netdevice.h> 54 55 struct netpoll_info; 56 struct device; 57 struct phy_device; 58 /* 802.11 specific */ 59 struct wireless_dev; 60 61 void netdev_set_default_ethtool_ops(struct net_device *dev, 62 const struct ethtool_ops *ops); 63 64 /* Backlog congestion levels */ 65 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 66 #define NET_RX_DROP 1 /* packet dropped */ 67 68 /* 69 * Transmit return codes: transmit return codes originate from three different 70 * namespaces: 71 * 72 * - qdisc return codes 73 * - driver transmit return codes 74 * - errno values 75 * 76 * Drivers are allowed to return any one of those in their hard_start_xmit() 77 * function. Real network devices commonly used with qdiscs should only return 78 * the driver transmit return codes though - when qdiscs are used, the actual 79 * transmission happens asynchronously, so the value is not propagated to 80 * higher layers. Virtual network devices transmit synchronously, in this case 81 * the driver transmit return codes are consumed by dev_queue_xmit(), all 82 * others are propagated to higher layers. 83 */ 84 85 /* qdisc ->enqueue() return codes. */ 86 #define NET_XMIT_SUCCESS 0x00 87 #define NET_XMIT_DROP 0x01 /* skb dropped */ 88 #define NET_XMIT_CN 0x02 /* congestion notification */ 89 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */ 90 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 91 92 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 93 * indicates that the device will soon be dropping packets, or already drops 94 * some packets of the same priority; prompting us to send less aggressively. */ 95 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 96 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 97 98 /* Driver transmit return codes */ 99 #define NETDEV_TX_MASK 0xf0 100 101 enum netdev_tx { 102 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 103 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 104 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 105 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */ 106 }; 107 typedef enum netdev_tx netdev_tx_t; 108 109 /* 110 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 111 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 112 */ 113 static inline bool dev_xmit_complete(int rc) 114 { 115 /* 116 * Positive cases with an skb consumed by a driver: 117 * - successful transmission (rc == NETDEV_TX_OK) 118 * - error while transmitting (rc < 0) 119 * - error while queueing to a different device (rc & NET_XMIT_MASK) 120 */ 121 if (likely(rc < NET_XMIT_MASK)) 122 return true; 123 124 return false; 125 } 126 127 /* 128 * Compute the worst case header length according to the protocols 129 * used. 130 */ 131 132 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 133 # if defined(CONFIG_MAC80211_MESH) 134 # define LL_MAX_HEADER 128 135 # else 136 # define LL_MAX_HEADER 96 137 # endif 138 #else 139 # define LL_MAX_HEADER 32 140 #endif 141 142 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 143 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 144 #define MAX_HEADER LL_MAX_HEADER 145 #else 146 #define MAX_HEADER (LL_MAX_HEADER + 48) 147 #endif 148 149 /* 150 * Old network device statistics. Fields are native words 151 * (unsigned long) so they can be read and written atomically. 152 */ 153 154 struct net_device_stats { 155 unsigned long rx_packets; 156 unsigned long tx_packets; 157 unsigned long rx_bytes; 158 unsigned long tx_bytes; 159 unsigned long rx_errors; 160 unsigned long tx_errors; 161 unsigned long rx_dropped; 162 unsigned long tx_dropped; 163 unsigned long multicast; 164 unsigned long collisions; 165 unsigned long rx_length_errors; 166 unsigned long rx_over_errors; 167 unsigned long rx_crc_errors; 168 unsigned long rx_frame_errors; 169 unsigned long rx_fifo_errors; 170 unsigned long rx_missed_errors; 171 unsigned long tx_aborted_errors; 172 unsigned long tx_carrier_errors; 173 unsigned long tx_fifo_errors; 174 unsigned long tx_heartbeat_errors; 175 unsigned long tx_window_errors; 176 unsigned long rx_compressed; 177 unsigned long tx_compressed; 178 }; 179 180 181 #include <linux/cache.h> 182 #include <linux/skbuff.h> 183 184 #ifdef CONFIG_RPS 185 #include <linux/static_key.h> 186 extern struct static_key rps_needed; 187 #endif 188 189 struct neighbour; 190 struct neigh_parms; 191 struct sk_buff; 192 193 struct netdev_hw_addr { 194 struct list_head list; 195 unsigned char addr[MAX_ADDR_LEN]; 196 unsigned char type; 197 #define NETDEV_HW_ADDR_T_LAN 1 198 #define NETDEV_HW_ADDR_T_SAN 2 199 #define NETDEV_HW_ADDR_T_SLAVE 3 200 #define NETDEV_HW_ADDR_T_UNICAST 4 201 #define NETDEV_HW_ADDR_T_MULTICAST 5 202 bool global_use; 203 int sync_cnt; 204 int refcount; 205 int synced; 206 struct rcu_head rcu_head; 207 }; 208 209 struct netdev_hw_addr_list { 210 struct list_head list; 211 int count; 212 }; 213 214 #define netdev_hw_addr_list_count(l) ((l)->count) 215 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 216 #define netdev_hw_addr_list_for_each(ha, l) \ 217 list_for_each_entry(ha, &(l)->list, list) 218 219 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 220 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 221 #define netdev_for_each_uc_addr(ha, dev) \ 222 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 223 224 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 225 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 226 #define netdev_for_each_mc_addr(ha, dev) \ 227 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 228 229 struct hh_cache { 230 u16 hh_len; 231 u16 __pad; 232 seqlock_t hh_lock; 233 234 /* cached hardware header; allow for machine alignment needs. */ 235 #define HH_DATA_MOD 16 236 #define HH_DATA_OFF(__len) \ 237 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 238 #define HH_DATA_ALIGN(__len) \ 239 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 240 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 241 }; 242 243 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much. 244 * Alternative is: 245 * dev->hard_header_len ? (dev->hard_header_len + 246 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 247 * 248 * We could use other alignment values, but we must maintain the 249 * relationship HH alignment <= LL alignment. 250 */ 251 #define LL_RESERVED_SPACE(dev) \ 252 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 253 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 254 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 255 256 struct header_ops { 257 int (*create) (struct sk_buff *skb, struct net_device *dev, 258 unsigned short type, const void *daddr, 259 const void *saddr, unsigned int len); 260 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 261 int (*rebuild)(struct sk_buff *skb); 262 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 263 void (*cache_update)(struct hh_cache *hh, 264 const struct net_device *dev, 265 const unsigned char *haddr); 266 }; 267 268 /* These flag bits are private to the generic network queueing 269 * layer, they may not be explicitly referenced by any other 270 * code. 271 */ 272 273 enum netdev_state_t { 274 __LINK_STATE_START, 275 __LINK_STATE_PRESENT, 276 __LINK_STATE_NOCARRIER, 277 __LINK_STATE_LINKWATCH_PENDING, 278 __LINK_STATE_DORMANT, 279 }; 280 281 282 /* 283 * This structure holds at boot time configured netdevice settings. They 284 * are then used in the device probing. 285 */ 286 struct netdev_boot_setup { 287 char name[IFNAMSIZ]; 288 struct ifmap map; 289 }; 290 #define NETDEV_BOOT_SETUP_MAX 8 291 292 int __init netdev_boot_setup(char *str); 293 294 /* 295 * Structure for NAPI scheduling similar to tasklet but with weighting 296 */ 297 struct napi_struct { 298 /* The poll_list must only be managed by the entity which 299 * changes the state of the NAPI_STATE_SCHED bit. This means 300 * whoever atomically sets that bit can add this napi_struct 301 * to the per-cpu poll_list, and whoever clears that bit 302 * can remove from the list right before clearing the bit. 303 */ 304 struct list_head poll_list; 305 306 unsigned long state; 307 int weight; 308 unsigned int gro_count; 309 int (*poll)(struct napi_struct *, int); 310 #ifdef CONFIG_NETPOLL 311 spinlock_t poll_lock; 312 int poll_owner; 313 #endif 314 struct net_device *dev; 315 struct sk_buff *gro_list; 316 struct sk_buff *skb; 317 struct list_head dev_list; 318 struct hlist_node napi_hash_node; 319 unsigned int napi_id; 320 }; 321 322 enum { 323 NAPI_STATE_SCHED, /* Poll is scheduled */ 324 NAPI_STATE_DISABLE, /* Disable pending */ 325 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 326 NAPI_STATE_HASHED, /* In NAPI hash */ 327 }; 328 329 enum gro_result { 330 GRO_MERGED, 331 GRO_MERGED_FREE, 332 GRO_HELD, 333 GRO_NORMAL, 334 GRO_DROP, 335 }; 336 typedef enum gro_result gro_result_t; 337 338 /* 339 * enum rx_handler_result - Possible return values for rx_handlers. 340 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 341 * further. 342 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 343 * case skb->dev was changed by rx_handler. 344 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 345 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called. 346 * 347 * rx_handlers are functions called from inside __netif_receive_skb(), to do 348 * special processing of the skb, prior to delivery to protocol handlers. 349 * 350 * Currently, a net_device can only have a single rx_handler registered. Trying 351 * to register a second rx_handler will return -EBUSY. 352 * 353 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 354 * To unregister a rx_handler on a net_device, use 355 * netdev_rx_handler_unregister(). 356 * 357 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 358 * do with the skb. 359 * 360 * If the rx_handler consumed to skb in some way, it should return 361 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 362 * the skb to be delivered in some other ways. 363 * 364 * If the rx_handler changed skb->dev, to divert the skb to another 365 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 366 * new device will be called if it exists. 367 * 368 * If the rx_handler consider the skb should be ignored, it should return 369 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 370 * are registered on exact device (ptype->dev == skb->dev). 371 * 372 * If the rx_handler didn't changed skb->dev, but want the skb to be normally 373 * delivered, it should return RX_HANDLER_PASS. 374 * 375 * A device without a registered rx_handler will behave as if rx_handler 376 * returned RX_HANDLER_PASS. 377 */ 378 379 enum rx_handler_result { 380 RX_HANDLER_CONSUMED, 381 RX_HANDLER_ANOTHER, 382 RX_HANDLER_EXACT, 383 RX_HANDLER_PASS, 384 }; 385 typedef enum rx_handler_result rx_handler_result_t; 386 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 387 388 void __napi_schedule(struct napi_struct *n); 389 390 static inline bool napi_disable_pending(struct napi_struct *n) 391 { 392 return test_bit(NAPI_STATE_DISABLE, &n->state); 393 } 394 395 /** 396 * napi_schedule_prep - check if napi can be scheduled 397 * @n: napi context 398 * 399 * Test if NAPI routine is already running, and if not mark 400 * it as running. This is used as a condition variable 401 * insure only one NAPI poll instance runs. We also make 402 * sure there is no pending NAPI disable. 403 */ 404 static inline bool napi_schedule_prep(struct napi_struct *n) 405 { 406 return !napi_disable_pending(n) && 407 !test_and_set_bit(NAPI_STATE_SCHED, &n->state); 408 } 409 410 /** 411 * napi_schedule - schedule NAPI poll 412 * @n: napi context 413 * 414 * Schedule NAPI poll routine to be called if it is not already 415 * running. 416 */ 417 static inline void napi_schedule(struct napi_struct *n) 418 { 419 if (napi_schedule_prep(n)) 420 __napi_schedule(n); 421 } 422 423 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 424 static inline bool napi_reschedule(struct napi_struct *napi) 425 { 426 if (napi_schedule_prep(napi)) { 427 __napi_schedule(napi); 428 return true; 429 } 430 return false; 431 } 432 433 /** 434 * napi_complete - NAPI processing complete 435 * @n: napi context 436 * 437 * Mark NAPI processing as complete. 438 */ 439 void __napi_complete(struct napi_struct *n); 440 void napi_complete(struct napi_struct *n); 441 442 /** 443 * napi_by_id - lookup a NAPI by napi_id 444 * @napi_id: hashed napi_id 445 * 446 * lookup @napi_id in napi_hash table 447 * must be called under rcu_read_lock() 448 */ 449 struct napi_struct *napi_by_id(unsigned int napi_id); 450 451 /** 452 * napi_hash_add - add a NAPI to global hashtable 453 * @napi: napi context 454 * 455 * generate a new napi_id and store a @napi under it in napi_hash 456 */ 457 void napi_hash_add(struct napi_struct *napi); 458 459 /** 460 * napi_hash_del - remove a NAPI from global table 461 * @napi: napi context 462 * 463 * Warning: caller must observe rcu grace period 464 * before freeing memory containing @napi 465 */ 466 void napi_hash_del(struct napi_struct *napi); 467 468 /** 469 * napi_disable - prevent NAPI from scheduling 470 * @n: napi context 471 * 472 * Stop NAPI from being scheduled on this context. 473 * Waits till any outstanding processing completes. 474 */ 475 static inline void napi_disable(struct napi_struct *n) 476 { 477 might_sleep(); 478 set_bit(NAPI_STATE_DISABLE, &n->state); 479 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state)) 480 msleep(1); 481 clear_bit(NAPI_STATE_DISABLE, &n->state); 482 } 483 484 /** 485 * napi_enable - enable NAPI scheduling 486 * @n: napi context 487 * 488 * Resume NAPI from being scheduled on this context. 489 * Must be paired with napi_disable. 490 */ 491 static inline void napi_enable(struct napi_struct *n) 492 { 493 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 494 smp_mb__before_atomic(); 495 clear_bit(NAPI_STATE_SCHED, &n->state); 496 } 497 498 #ifdef CONFIG_SMP 499 /** 500 * napi_synchronize - wait until NAPI is not running 501 * @n: napi context 502 * 503 * Wait until NAPI is done being scheduled on this context. 504 * Waits till any outstanding processing completes but 505 * does not disable future activations. 506 */ 507 static inline void napi_synchronize(const struct napi_struct *n) 508 { 509 while (test_bit(NAPI_STATE_SCHED, &n->state)) 510 msleep(1); 511 } 512 #else 513 # define napi_synchronize(n) barrier() 514 #endif 515 516 enum netdev_queue_state_t { 517 __QUEUE_STATE_DRV_XOFF, 518 __QUEUE_STATE_STACK_XOFF, 519 __QUEUE_STATE_FROZEN, 520 }; 521 522 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 523 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 524 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 525 526 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 527 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 528 QUEUE_STATE_FROZEN) 529 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 530 QUEUE_STATE_FROZEN) 531 532 /* 533 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 534 * netif_tx_* functions below are used to manipulate this flag. The 535 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 536 * queue independently. The netif_xmit_*stopped functions below are called 537 * to check if the queue has been stopped by the driver or stack (either 538 * of the XOFF bits are set in the state). Drivers should not need to call 539 * netif_xmit*stopped functions, they should only be using netif_tx_*. 540 */ 541 542 struct netdev_queue { 543 /* 544 * read mostly part 545 */ 546 struct net_device *dev; 547 struct Qdisc __rcu *qdisc; 548 struct Qdisc *qdisc_sleeping; 549 #ifdef CONFIG_SYSFS 550 struct kobject kobj; 551 #endif 552 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 553 int numa_node; 554 #endif 555 /* 556 * write mostly part 557 */ 558 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 559 int xmit_lock_owner; 560 /* 561 * please use this field instead of dev->trans_start 562 */ 563 unsigned long trans_start; 564 565 /* 566 * Number of TX timeouts for this queue 567 * (/sys/class/net/DEV/Q/trans_timeout) 568 */ 569 unsigned long trans_timeout; 570 571 unsigned long state; 572 573 #ifdef CONFIG_BQL 574 struct dql dql; 575 #endif 576 } ____cacheline_aligned_in_smp; 577 578 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 579 { 580 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 581 return q->numa_node; 582 #else 583 return NUMA_NO_NODE; 584 #endif 585 } 586 587 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 588 { 589 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 590 q->numa_node = node; 591 #endif 592 } 593 594 #ifdef CONFIG_RPS 595 /* 596 * This structure holds an RPS map which can be of variable length. The 597 * map is an array of CPUs. 598 */ 599 struct rps_map { 600 unsigned int len; 601 struct rcu_head rcu; 602 u16 cpus[0]; 603 }; 604 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 605 606 /* 607 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 608 * tail pointer for that CPU's input queue at the time of last enqueue, and 609 * a hardware filter index. 610 */ 611 struct rps_dev_flow { 612 u16 cpu; 613 u16 filter; 614 unsigned int last_qtail; 615 }; 616 #define RPS_NO_FILTER 0xffff 617 618 /* 619 * The rps_dev_flow_table structure contains a table of flow mappings. 620 */ 621 struct rps_dev_flow_table { 622 unsigned int mask; 623 struct rcu_head rcu; 624 struct rps_dev_flow flows[0]; 625 }; 626 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 627 ((_num) * sizeof(struct rps_dev_flow))) 628 629 /* 630 * The rps_sock_flow_table contains mappings of flows to the last CPU 631 * on which they were processed by the application (set in recvmsg). 632 */ 633 struct rps_sock_flow_table { 634 unsigned int mask; 635 u16 ents[0]; 636 }; 637 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \ 638 ((_num) * sizeof(u16))) 639 640 #define RPS_NO_CPU 0xffff 641 642 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 643 u32 hash) 644 { 645 if (table && hash) { 646 unsigned int cpu, index = hash & table->mask; 647 648 /* We only give a hint, preemption can change cpu under us */ 649 cpu = raw_smp_processor_id(); 650 651 if (table->ents[index] != cpu) 652 table->ents[index] = cpu; 653 } 654 } 655 656 static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table, 657 u32 hash) 658 { 659 if (table && hash) 660 table->ents[hash & table->mask] = RPS_NO_CPU; 661 } 662 663 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 664 665 #ifdef CONFIG_RFS_ACCEL 666 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 667 u16 filter_id); 668 #endif 669 #endif /* CONFIG_RPS */ 670 671 /* This structure contains an instance of an RX queue. */ 672 struct netdev_rx_queue { 673 #ifdef CONFIG_RPS 674 struct rps_map __rcu *rps_map; 675 struct rps_dev_flow_table __rcu *rps_flow_table; 676 #endif 677 struct kobject kobj; 678 struct net_device *dev; 679 } ____cacheline_aligned_in_smp; 680 681 /* 682 * RX queue sysfs structures and functions. 683 */ 684 struct rx_queue_attribute { 685 struct attribute attr; 686 ssize_t (*show)(struct netdev_rx_queue *queue, 687 struct rx_queue_attribute *attr, char *buf); 688 ssize_t (*store)(struct netdev_rx_queue *queue, 689 struct rx_queue_attribute *attr, const char *buf, size_t len); 690 }; 691 692 #ifdef CONFIG_XPS 693 /* 694 * This structure holds an XPS map which can be of variable length. The 695 * map is an array of queues. 696 */ 697 struct xps_map { 698 unsigned int len; 699 unsigned int alloc_len; 700 struct rcu_head rcu; 701 u16 queues[0]; 702 }; 703 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 704 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \ 705 / sizeof(u16)) 706 707 /* 708 * This structure holds all XPS maps for device. Maps are indexed by CPU. 709 */ 710 struct xps_dev_maps { 711 struct rcu_head rcu; 712 struct xps_map __rcu *cpu_map[0]; 713 }; 714 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \ 715 (nr_cpu_ids * sizeof(struct xps_map *))) 716 #endif /* CONFIG_XPS */ 717 718 #define TC_MAX_QUEUE 16 719 #define TC_BITMASK 15 720 /* HW offloaded queuing disciplines txq count and offset maps */ 721 struct netdev_tc_txq { 722 u16 count; 723 u16 offset; 724 }; 725 726 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 727 /* 728 * This structure is to hold information about the device 729 * configured to run FCoE protocol stack. 730 */ 731 struct netdev_fcoe_hbainfo { 732 char manufacturer[64]; 733 char serial_number[64]; 734 char hardware_version[64]; 735 char driver_version[64]; 736 char optionrom_version[64]; 737 char firmware_version[64]; 738 char model[256]; 739 char model_description[256]; 740 }; 741 #endif 742 743 #define MAX_PHYS_PORT_ID_LEN 32 744 745 /* This structure holds a unique identifier to identify the 746 * physical port used by a netdevice. 747 */ 748 struct netdev_phys_port_id { 749 unsigned char id[MAX_PHYS_PORT_ID_LEN]; 750 unsigned char id_len; 751 }; 752 753 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 754 struct sk_buff *skb); 755 756 /* 757 * This structure defines the management hooks for network devices. 758 * The following hooks can be defined; unless noted otherwise, they are 759 * optional and can be filled with a null pointer. 760 * 761 * int (*ndo_init)(struct net_device *dev); 762 * This function is called once when network device is registered. 763 * The network device can use this to any late stage initializaton 764 * or semantic validattion. It can fail with an error code which will 765 * be propogated back to register_netdev 766 * 767 * void (*ndo_uninit)(struct net_device *dev); 768 * This function is called when device is unregistered or when registration 769 * fails. It is not called if init fails. 770 * 771 * int (*ndo_open)(struct net_device *dev); 772 * This function is called when network device transistions to the up 773 * state. 774 * 775 * int (*ndo_stop)(struct net_device *dev); 776 * This function is called when network device transistions to the down 777 * state. 778 * 779 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 780 * struct net_device *dev); 781 * Called when a packet needs to be transmitted. 782 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY. 783 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX) 784 * Required can not be NULL. 785 * 786 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 787 * void *accel_priv, select_queue_fallback_t fallback); 788 * Called to decide which queue to when device supports multiple 789 * transmit queues. 790 * 791 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 792 * This function is called to allow device receiver to make 793 * changes to configuration when multicast or promiscious is enabled. 794 * 795 * void (*ndo_set_rx_mode)(struct net_device *dev); 796 * This function is called device changes address list filtering. 797 * If driver handles unicast address filtering, it should set 798 * IFF_UNICAST_FLT to its priv_flags. 799 * 800 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 801 * This function is called when the Media Access Control address 802 * needs to be changed. If this interface is not defined, the 803 * mac address can not be changed. 804 * 805 * int (*ndo_validate_addr)(struct net_device *dev); 806 * Test if Media Access Control address is valid for the device. 807 * 808 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 809 * Called when a user request an ioctl which can't be handled by 810 * the generic interface code. If not defined ioctl's return 811 * not supported error code. 812 * 813 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 814 * Used to set network devices bus interface parameters. This interface 815 * is retained for legacy reason, new devices should use the bus 816 * interface (PCI) for low level management. 817 * 818 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 819 * Called when a user wants to change the Maximum Transfer Unit 820 * of a device. If not defined, any request to change MTU will 821 * will return an error. 822 * 823 * void (*ndo_tx_timeout)(struct net_device *dev); 824 * Callback uses when the transmitter has not made any progress 825 * for dev->watchdog ticks. 826 * 827 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev, 828 * struct rtnl_link_stats64 *storage); 829 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 830 * Called when a user wants to get the network device usage 831 * statistics. Drivers must do one of the following: 832 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 833 * rtnl_link_stats64 structure passed by the caller. 834 * 2. Define @ndo_get_stats to update a net_device_stats structure 835 * (which should normally be dev->stats) and return a pointer to 836 * it. The structure may be changed asynchronously only if each 837 * field is written atomically. 838 * 3. Update dev->stats asynchronously and atomically, and define 839 * neither operation. 840 * 841 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid); 842 * If device support VLAN filtering this function is called when a 843 * VLAN id is registered. 844 * 845 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid); 846 * If device support VLAN filtering this function is called when a 847 * VLAN id is unregistered. 848 * 849 * void (*ndo_poll_controller)(struct net_device *dev); 850 * 851 * SR-IOV management functions. 852 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 853 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos); 854 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 855 * int max_tx_rate); 856 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 857 * int (*ndo_get_vf_config)(struct net_device *dev, 858 * int vf, struct ifla_vf_info *ivf); 859 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 860 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 861 * struct nlattr *port[]); 862 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 863 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc) 864 * Called to setup 'tc' number of traffic classes in the net device. This 865 * is always called from the stack with the rtnl lock held and netif tx 866 * queues stopped. This allows the netdevice to perform queue management 867 * safely. 868 * 869 * Fiber Channel over Ethernet (FCoE) offload functions. 870 * int (*ndo_fcoe_enable)(struct net_device *dev); 871 * Called when the FCoE protocol stack wants to start using LLD for FCoE 872 * so the underlying device can perform whatever needed configuration or 873 * initialization to support acceleration of FCoE traffic. 874 * 875 * int (*ndo_fcoe_disable)(struct net_device *dev); 876 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 877 * so the underlying device can perform whatever needed clean-ups to 878 * stop supporting acceleration of FCoE traffic. 879 * 880 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 881 * struct scatterlist *sgl, unsigned int sgc); 882 * Called when the FCoE Initiator wants to initialize an I/O that 883 * is a possible candidate for Direct Data Placement (DDP). The LLD can 884 * perform necessary setup and returns 1 to indicate the device is set up 885 * successfully to perform DDP on this I/O, otherwise this returns 0. 886 * 887 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 888 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 889 * indicated by the FC exchange id 'xid', so the underlying device can 890 * clean up and reuse resources for later DDP requests. 891 * 892 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 893 * struct scatterlist *sgl, unsigned int sgc); 894 * Called when the FCoE Target wants to initialize an I/O that 895 * is a possible candidate for Direct Data Placement (DDP). The LLD can 896 * perform necessary setup and returns 1 to indicate the device is set up 897 * successfully to perform DDP on this I/O, otherwise this returns 0. 898 * 899 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 900 * struct netdev_fcoe_hbainfo *hbainfo); 901 * Called when the FCoE Protocol stack wants information on the underlying 902 * device. This information is utilized by the FCoE protocol stack to 903 * register attributes with Fiber Channel management service as per the 904 * FC-GS Fabric Device Management Information(FDMI) specification. 905 * 906 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 907 * Called when the underlying device wants to override default World Wide 908 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 909 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 910 * protocol stack to use. 911 * 912 * RFS acceleration. 913 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 914 * u16 rxq_index, u32 flow_id); 915 * Set hardware filter for RFS. rxq_index is the target queue index; 916 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 917 * Return the filter ID on success, or a negative error code. 918 * 919 * Slave management functions (for bridge, bonding, etc). 920 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 921 * Called to make another netdev an underling. 922 * 923 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 924 * Called to release previously enslaved netdev. 925 * 926 * Feature/offload setting functions. 927 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 928 * netdev_features_t features); 929 * Adjusts the requested feature flags according to device-specific 930 * constraints, and returns the resulting flags. Must not modify 931 * the device state. 932 * 933 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 934 * Called to update device configuration to new features. Passed 935 * feature set might be less than what was returned by ndo_fix_features()). 936 * Must return >0 or -errno if it changed dev->features itself. 937 * 938 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 939 * struct net_device *dev, 940 * const unsigned char *addr, u16 flags) 941 * Adds an FDB entry to dev for addr. 942 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 943 * struct net_device *dev, 944 * const unsigned char *addr) 945 * Deletes the FDB entry from dev coresponding to addr. 946 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 947 * struct net_device *dev, struct net_device *filter_dev, 948 * 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 * bool (*ndo_gso_check) (struct sk_buff *skb, 1002 * struct net_device *dev); 1003 * Called by core transmit path to determine if device is capable of 1004 * performing GSO on a packet. The device returns true if it is 1005 * able to GSO the packet, false otherwise. If the return value is 1006 * false the stack will do software GSO. 1007 */ 1008 struct net_device_ops { 1009 int (*ndo_init)(struct net_device *dev); 1010 void (*ndo_uninit)(struct net_device *dev); 1011 int (*ndo_open)(struct net_device *dev); 1012 int (*ndo_stop)(struct net_device *dev); 1013 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb, 1014 struct net_device *dev); 1015 u16 (*ndo_select_queue)(struct net_device *dev, 1016 struct sk_buff *skb, 1017 void *accel_priv, 1018 select_queue_fallback_t fallback); 1019 void (*ndo_change_rx_flags)(struct net_device *dev, 1020 int flags); 1021 void (*ndo_set_rx_mode)(struct net_device *dev); 1022 int (*ndo_set_mac_address)(struct net_device *dev, 1023 void *addr); 1024 int (*ndo_validate_addr)(struct net_device *dev); 1025 int (*ndo_do_ioctl)(struct net_device *dev, 1026 struct ifreq *ifr, int cmd); 1027 int (*ndo_set_config)(struct net_device *dev, 1028 struct ifmap *map); 1029 int (*ndo_change_mtu)(struct net_device *dev, 1030 int new_mtu); 1031 int (*ndo_neigh_setup)(struct net_device *dev, 1032 struct neigh_parms *); 1033 void (*ndo_tx_timeout) (struct net_device *dev); 1034 1035 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev, 1036 struct rtnl_link_stats64 *storage); 1037 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1038 1039 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1040 __be16 proto, u16 vid); 1041 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1042 __be16 proto, u16 vid); 1043 #ifdef CONFIG_NET_POLL_CONTROLLER 1044 void (*ndo_poll_controller)(struct net_device *dev); 1045 int (*ndo_netpoll_setup)(struct net_device *dev, 1046 struct netpoll_info *info); 1047 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1048 #endif 1049 #ifdef CONFIG_NET_RX_BUSY_POLL 1050 int (*ndo_busy_poll)(struct napi_struct *dev); 1051 #endif 1052 int (*ndo_set_vf_mac)(struct net_device *dev, 1053 int queue, u8 *mac); 1054 int (*ndo_set_vf_vlan)(struct net_device *dev, 1055 int queue, u16 vlan, u8 qos); 1056 int (*ndo_set_vf_rate)(struct net_device *dev, 1057 int vf, int min_tx_rate, 1058 int max_tx_rate); 1059 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1060 int vf, bool setting); 1061 int (*ndo_get_vf_config)(struct net_device *dev, 1062 int vf, 1063 struct ifla_vf_info *ivf); 1064 int (*ndo_set_vf_link_state)(struct net_device *dev, 1065 int vf, int link_state); 1066 int (*ndo_set_vf_port)(struct net_device *dev, 1067 int vf, 1068 struct nlattr *port[]); 1069 int (*ndo_get_vf_port)(struct net_device *dev, 1070 int vf, struct sk_buff *skb); 1071 int (*ndo_setup_tc)(struct net_device *dev, u8 tc); 1072 #if IS_ENABLED(CONFIG_FCOE) 1073 int (*ndo_fcoe_enable)(struct net_device *dev); 1074 int (*ndo_fcoe_disable)(struct net_device *dev); 1075 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1076 u16 xid, 1077 struct scatterlist *sgl, 1078 unsigned int sgc); 1079 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1080 u16 xid); 1081 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1082 u16 xid, 1083 struct scatterlist *sgl, 1084 unsigned int sgc); 1085 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1086 struct netdev_fcoe_hbainfo *hbainfo); 1087 #endif 1088 1089 #if IS_ENABLED(CONFIG_LIBFCOE) 1090 #define NETDEV_FCOE_WWNN 0 1091 #define NETDEV_FCOE_WWPN 1 1092 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1093 u64 *wwn, int type); 1094 #endif 1095 1096 #ifdef CONFIG_RFS_ACCEL 1097 int (*ndo_rx_flow_steer)(struct net_device *dev, 1098 const struct sk_buff *skb, 1099 u16 rxq_index, 1100 u32 flow_id); 1101 #endif 1102 int (*ndo_add_slave)(struct net_device *dev, 1103 struct net_device *slave_dev); 1104 int (*ndo_del_slave)(struct net_device *dev, 1105 struct net_device *slave_dev); 1106 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1107 netdev_features_t features); 1108 int (*ndo_set_features)(struct net_device *dev, 1109 netdev_features_t features); 1110 int (*ndo_neigh_construct)(struct neighbour *n); 1111 void (*ndo_neigh_destroy)(struct neighbour *n); 1112 1113 int (*ndo_fdb_add)(struct ndmsg *ndm, 1114 struct nlattr *tb[], 1115 struct net_device *dev, 1116 const unsigned char *addr, 1117 u16 flags); 1118 int (*ndo_fdb_del)(struct ndmsg *ndm, 1119 struct nlattr *tb[], 1120 struct net_device *dev, 1121 const unsigned char *addr); 1122 int (*ndo_fdb_dump)(struct sk_buff *skb, 1123 struct netlink_callback *cb, 1124 struct net_device *dev, 1125 struct net_device *filter_dev, 1126 int idx); 1127 1128 int (*ndo_bridge_setlink)(struct net_device *dev, 1129 struct nlmsghdr *nlh); 1130 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1131 u32 pid, u32 seq, 1132 struct net_device *dev, 1133 u32 filter_mask); 1134 int (*ndo_bridge_dellink)(struct net_device *dev, 1135 struct nlmsghdr *nlh); 1136 int (*ndo_change_carrier)(struct net_device *dev, 1137 bool new_carrier); 1138 int (*ndo_get_phys_port_id)(struct net_device *dev, 1139 struct netdev_phys_port_id *ppid); 1140 void (*ndo_add_vxlan_port)(struct net_device *dev, 1141 sa_family_t sa_family, 1142 __be16 port); 1143 void (*ndo_del_vxlan_port)(struct net_device *dev, 1144 sa_family_t sa_family, 1145 __be16 port); 1146 1147 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1148 struct net_device *dev); 1149 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1150 void *priv); 1151 1152 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb, 1153 struct net_device *dev, 1154 void *priv); 1155 int (*ndo_get_lock_subclass)(struct net_device *dev); 1156 bool (*ndo_gso_check) (struct sk_buff *skb, 1157 struct net_device *dev); 1158 }; 1159 1160 /** 1161 * enum net_device_priv_flags - &struct net_device priv_flags 1162 * 1163 * These are the &struct net_device, they are only set internally 1164 * by drivers and used in the kernel. These flags are invisible to 1165 * userspace, this means that the order of these flags can change 1166 * during any kernel release. 1167 * 1168 * You should have a pretty good reason to be extending these flags. 1169 * 1170 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1171 * @IFF_EBRIDGE: Ethernet bridging device 1172 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active 1173 * @IFF_MASTER_8023AD: bonding master, 802.3ad 1174 * @IFF_MASTER_ALB: bonding master, balance-alb 1175 * @IFF_BONDING: bonding master or slave 1176 * @IFF_SLAVE_NEEDARP: need ARPs for validation 1177 * @IFF_ISATAP: ISATAP interface (RFC4214) 1178 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use 1179 * @IFF_WAN_HDLC: WAN HDLC device 1180 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1181 * release skb->dst 1182 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1183 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1184 * @IFF_MACVLAN_PORT: device used as macvlan port 1185 * @IFF_BRIDGE_PORT: device used as bridge port 1186 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1187 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1188 * @IFF_UNICAST_FLT: Supports unicast filtering 1189 * @IFF_TEAM_PORT: device used as team port 1190 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1191 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1192 * change when it's running 1193 * @IFF_MACVLAN: Macvlan device 1194 */ 1195 enum netdev_priv_flags { 1196 IFF_802_1Q_VLAN = 1<<0, 1197 IFF_EBRIDGE = 1<<1, 1198 IFF_SLAVE_INACTIVE = 1<<2, 1199 IFF_MASTER_8023AD = 1<<3, 1200 IFF_MASTER_ALB = 1<<4, 1201 IFF_BONDING = 1<<5, 1202 IFF_SLAVE_NEEDARP = 1<<6, 1203 IFF_ISATAP = 1<<7, 1204 IFF_MASTER_ARPMON = 1<<8, 1205 IFF_WAN_HDLC = 1<<9, 1206 IFF_XMIT_DST_RELEASE = 1<<10, 1207 IFF_DONT_BRIDGE = 1<<11, 1208 IFF_DISABLE_NETPOLL = 1<<12, 1209 IFF_MACVLAN_PORT = 1<<13, 1210 IFF_BRIDGE_PORT = 1<<14, 1211 IFF_OVS_DATAPATH = 1<<15, 1212 IFF_TX_SKB_SHARING = 1<<16, 1213 IFF_UNICAST_FLT = 1<<17, 1214 IFF_TEAM_PORT = 1<<18, 1215 IFF_SUPP_NOFCS = 1<<19, 1216 IFF_LIVE_ADDR_CHANGE = 1<<20, 1217 IFF_MACVLAN = 1<<21, 1218 IFF_XMIT_DST_RELEASE_PERM = 1<<22, 1219 }; 1220 1221 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1222 #define IFF_EBRIDGE IFF_EBRIDGE 1223 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE 1224 #define IFF_MASTER_8023AD IFF_MASTER_8023AD 1225 #define IFF_MASTER_ALB IFF_MASTER_ALB 1226 #define IFF_BONDING IFF_BONDING 1227 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP 1228 #define IFF_ISATAP IFF_ISATAP 1229 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON 1230 #define IFF_WAN_HDLC IFF_WAN_HDLC 1231 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1232 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1233 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1234 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1235 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1236 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1237 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1238 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1239 #define IFF_TEAM_PORT IFF_TEAM_PORT 1240 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1241 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1242 #define IFF_MACVLAN IFF_MACVLAN 1243 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1244 1245 /** 1246 * struct net_device - The DEVICE structure. 1247 * Actually, this whole structure is a big mistake. It mixes I/O 1248 * data with strictly "high-level" data, and it has to know about 1249 * almost every data structure used in the INET module. 1250 * 1251 * @name: This is the first field of the "visible" part of this structure 1252 * (i.e. as seen by users in the "Space.c" file). It is the name 1253 * of the interface. 1254 * 1255 * @name_hlist: Device name hash chain, please keep it close to name[] 1256 * @ifalias: SNMP alias 1257 * @mem_end: Shared memory end 1258 * @mem_start: Shared memory start 1259 * @base_addr: Device I/O address 1260 * @irq: Device IRQ number 1261 * 1262 * @state: Generic network queuing layer state, see netdev_state_t 1263 * @dev_list: The global list of network devices 1264 * @napi_list: List entry, that is used for polling napi devices 1265 * @unreg_list: List entry, that is used, when we are unregistering the 1266 * device, see the function unregister_netdev 1267 * @close_list: List entry, that is used, when we are closing the device 1268 * 1269 * @adj_list: Directly linked devices, like slaves for bonding 1270 * @all_adj_list: All linked devices, *including* neighbours 1271 * @features: Currently active device features 1272 * @hw_features: User-changeable features 1273 * 1274 * @wanted_features: User-requested features 1275 * @vlan_features: Mask of features inheritable by VLAN devices 1276 * 1277 * @hw_enc_features: Mask of features inherited by encapsulating devices 1278 * This field indicates what encapsulation 1279 * offloads the hardware is capable of doing, 1280 * and drivers will need to set them appropriately. 1281 * 1282 * @mpls_features: Mask of features inheritable by MPLS 1283 * 1284 * @ifindex: interface index 1285 * @iflink: unique device identifier 1286 * 1287 * @stats: Statistics struct, which was left as a legacy, use 1288 * rtnl_link_stats64 instead 1289 * 1290 * @rx_dropped: Dropped packets by core network, 1291 * do not use this in drivers 1292 * @tx_dropped: Dropped packets by core network, 1293 * do not use this in drivers 1294 * 1295 * @carrier_changes: Stats to monitor carrier on<->off transitions 1296 * 1297 * @wireless_handlers: List of functions to handle Wireless Extensions, 1298 * instead of ioctl, 1299 * see <net/iw_handler.h> for details. 1300 * @wireless_data: Instance data managed by the core of wireless extensions 1301 * 1302 * @netdev_ops: Includes several pointers to callbacks, 1303 * if one wants to override the ndo_*() functions 1304 * @ethtool_ops: Management operations 1305 * @fwd_ops: Management operations 1306 * @header_ops: Includes callbacks for creating,parsing,rebuilding,etc 1307 * of Layer 2 headers. 1308 * 1309 * @flags: Interface flags (a la BSD) 1310 * @priv_flags: Like 'flags' but invisible to userspace, 1311 * see if.h for the definitions 1312 * @gflags: Global flags ( kept as legacy ) 1313 * @padded: How much padding added by alloc_netdev() 1314 * @operstate: RFC2863 operstate 1315 * @link_mode: Mapping policy to operstate 1316 * @if_port: Selectable AUI, TP, ... 1317 * @dma: DMA channel 1318 * @mtu: Interface MTU value 1319 * @type: Interface hardware type 1320 * @hard_header_len: Hardware header length 1321 * 1322 * @needed_headroom: Extra headroom the hardware may need, but not in all 1323 * cases can this be guaranteed 1324 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1325 * cases can this be guaranteed. Some cases also use 1326 * LL_MAX_HEADER instead to allocate the skb 1327 * 1328 * interface address info: 1329 * 1330 * @perm_addr: Permanent hw address 1331 * @addr_assign_type: Hw address assignment type 1332 * @addr_len: Hardware address length 1333 * @neigh_priv_len; Used in neigh_alloc(), 1334 * initialized only in atm/clip.c 1335 * @dev_id: Used to differentiate devices that share 1336 * the same link layer address 1337 * @dev_port: Used to differentiate devices that share 1338 * the same function 1339 * @addr_list_lock: XXX: need comments on this one 1340 * @uc: unicast mac addresses 1341 * @mc: multicast mac addresses 1342 * @dev_addrs: list of device hw addresses 1343 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1344 * @uc_promisc: Counter, that indicates, that promiscuous mode 1345 * has been enabled due to the need to listen to 1346 * additional unicast addresses in a device that 1347 * does not implement ndo_set_rx_mode() 1348 * @promiscuity: Number of times, the NIC is told to work in 1349 * Promiscuous mode, if it becomes 0 the NIC will 1350 * exit from working in Promiscuous mode 1351 * @allmulti: Counter, enables or disables allmulticast mode 1352 * 1353 * @vlan_info: VLAN info 1354 * @dsa_ptr: dsa specific data 1355 * @tipc_ptr: TIPC specific data 1356 * @atalk_ptr: AppleTalk link 1357 * @ip_ptr: IPv4 specific data 1358 * @dn_ptr: DECnet specific data 1359 * @ip6_ptr: IPv6 specific data 1360 * @ax25_ptr: AX.25 specific data 1361 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1362 * 1363 * @last_rx: Time of last Rx 1364 * @dev_addr: Hw address (before bcast, 1365 * because most packets are unicast) 1366 * 1367 * @_rx: Array of RX queues 1368 * @num_rx_queues: Number of RX queues 1369 * allocated at register_netdev() time 1370 * @real_num_rx_queues: Number of RX queues currently active in device 1371 * 1372 * @rx_handler: handler for received packets 1373 * @rx_handler_data: XXX: need comments on this one 1374 * @ingress_queue: XXX: need comments on this one 1375 * @broadcast: hw bcast address 1376 * 1377 * @_tx: Array of TX queues 1378 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1379 * @real_num_tx_queues: Number of TX queues currently active in device 1380 * @qdisc: Root qdisc from userspace point of view 1381 * @tx_queue_len: Max frames per queue allowed 1382 * @tx_global_lock: XXX: need comments on this one 1383 * 1384 * @xps_maps: XXX: need comments on this one 1385 * 1386 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1387 * indexed by RX queue number. Assigned by driver. 1388 * This must only be set if the ndo_rx_flow_steer 1389 * operation is defined 1390 * 1391 * @trans_start: Time (in jiffies) of last Tx 1392 * @watchdog_timeo: Represents the timeout that is used by 1393 * the watchdog ( see dev_watchdog() ) 1394 * @watchdog_timer: List of timers 1395 * 1396 * @pcpu_refcnt: Number of references to this device 1397 * @todo_list: Delayed register/unregister 1398 * @index_hlist: Device index hash chain 1399 * @link_watch_list: XXX: need comments on this one 1400 * 1401 * @reg_state: Register/unregister state machine 1402 * @dismantle: Device is going to be freed 1403 * @rtnl_link_state: This enum represents the phases of creating 1404 * a new link 1405 * 1406 * @destructor: Called from unregister, 1407 * can be used to call free_netdev 1408 * @npinfo: XXX: need comments on this one 1409 * @nd_net: Network namespace this network device is inside 1410 * 1411 * @ml_priv: Mid-layer private 1412 * @lstats: Loopback statistics 1413 * @tstats: Tunnel statistics 1414 * @dstats: Dummy statistics 1415 * @vstats: Virtual ethernet statistics 1416 * 1417 * @garp_port: GARP 1418 * @mrp_port: MRP 1419 * 1420 * @dev: Class/net/name entry 1421 * @sysfs_groups: Space for optional device, statistics and wireless 1422 * sysfs groups 1423 * 1424 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1425 * @rtnl_link_ops: Rtnl_link_ops 1426 * 1427 * @gso_max_size: Maximum size of generic segmentation offload 1428 * @gso_max_segs: Maximum number of segments that can be passed to the 1429 * NIC for GSO 1430 * @gso_min_segs: Minimum number of segments that can be passed to the 1431 * NIC for GSO 1432 * 1433 * @dcbnl_ops: Data Center Bridging netlink ops 1434 * @num_tc: Number of traffic classes in the net device 1435 * @tc_to_txq: XXX: need comments on this one 1436 * @prio_tc_map XXX: need comments on this one 1437 * 1438 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1439 * 1440 * @priomap: XXX: need comments on this one 1441 * @phydev: Physical device may attach itself 1442 * for hardware timestamping 1443 * 1444 * @qdisc_tx_busylock: XXX: need comments on this one 1445 * 1446 * @group: The group, that the device belongs to 1447 * @pm_qos_req: Power Management QoS object 1448 * 1449 * FIXME: cleanup struct net_device such that network protocol info 1450 * moves out. 1451 */ 1452 1453 struct net_device { 1454 char name[IFNAMSIZ]; 1455 struct hlist_node name_hlist; 1456 char *ifalias; 1457 /* 1458 * I/O specific fields 1459 * FIXME: Merge these and struct ifmap into one 1460 */ 1461 unsigned long mem_end; 1462 unsigned long mem_start; 1463 unsigned long base_addr; 1464 int irq; 1465 1466 /* 1467 * Some hardware also needs these fields (state,dev_list, 1468 * napi_list,unreg_list,close_list) but they are not 1469 * part of the usual set specified in Space.c. 1470 */ 1471 1472 unsigned long state; 1473 1474 struct list_head dev_list; 1475 struct list_head napi_list; 1476 struct list_head unreg_list; 1477 struct list_head close_list; 1478 1479 struct { 1480 struct list_head upper; 1481 struct list_head lower; 1482 } adj_list; 1483 1484 struct { 1485 struct list_head upper; 1486 struct list_head lower; 1487 } all_adj_list; 1488 1489 netdev_features_t features; 1490 netdev_features_t hw_features; 1491 netdev_features_t wanted_features; 1492 netdev_features_t vlan_features; 1493 netdev_features_t hw_enc_features; 1494 netdev_features_t mpls_features; 1495 1496 int ifindex; 1497 int iflink; 1498 1499 struct net_device_stats stats; 1500 1501 atomic_long_t rx_dropped; 1502 atomic_long_t tx_dropped; 1503 1504 atomic_t carrier_changes; 1505 1506 #ifdef CONFIG_WIRELESS_EXT 1507 const struct iw_handler_def * wireless_handlers; 1508 struct iw_public_data * wireless_data; 1509 #endif 1510 const struct net_device_ops *netdev_ops; 1511 const struct ethtool_ops *ethtool_ops; 1512 const struct forwarding_accel_ops *fwd_ops; 1513 1514 const struct header_ops *header_ops; 1515 1516 unsigned int flags; 1517 unsigned int priv_flags; 1518 1519 unsigned short gflags; 1520 unsigned short padded; 1521 1522 unsigned char operstate; 1523 unsigned char link_mode; 1524 1525 unsigned char if_port; 1526 unsigned char dma; 1527 1528 unsigned int mtu; 1529 unsigned short type; 1530 unsigned short hard_header_len; 1531 1532 unsigned short needed_headroom; 1533 unsigned short needed_tailroom; 1534 1535 /* Interface address info. */ 1536 unsigned char perm_addr[MAX_ADDR_LEN]; 1537 unsigned char addr_assign_type; 1538 unsigned char addr_len; 1539 unsigned short neigh_priv_len; 1540 unsigned short dev_id; 1541 unsigned short dev_port; 1542 spinlock_t addr_list_lock; 1543 struct netdev_hw_addr_list uc; 1544 struct netdev_hw_addr_list mc; 1545 struct netdev_hw_addr_list dev_addrs; 1546 1547 #ifdef CONFIG_SYSFS 1548 struct kset *queues_kset; 1549 #endif 1550 1551 unsigned char name_assign_type; 1552 1553 bool uc_promisc; 1554 unsigned int promiscuity; 1555 unsigned int allmulti; 1556 1557 1558 /* Protocol specific pointers */ 1559 1560 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1561 struct vlan_info __rcu *vlan_info; 1562 #endif 1563 #if IS_ENABLED(CONFIG_NET_DSA) 1564 struct dsa_switch_tree *dsa_ptr; 1565 #endif 1566 #if IS_ENABLED(CONFIG_TIPC) 1567 struct tipc_bearer __rcu *tipc_ptr; 1568 #endif 1569 void *atalk_ptr; 1570 struct in_device __rcu *ip_ptr; 1571 struct dn_dev __rcu *dn_ptr; 1572 struct inet6_dev __rcu *ip6_ptr; 1573 void *ax25_ptr; 1574 struct wireless_dev *ieee80211_ptr; 1575 1576 /* 1577 * Cache lines mostly used on receive path (including eth_type_trans()) 1578 */ 1579 unsigned long last_rx; 1580 1581 /* Interface address info used in eth_type_trans() */ 1582 unsigned char *dev_addr; 1583 1584 1585 #ifdef CONFIG_SYSFS 1586 struct netdev_rx_queue *_rx; 1587 1588 unsigned int num_rx_queues; 1589 unsigned int real_num_rx_queues; 1590 1591 #endif 1592 1593 rx_handler_func_t __rcu *rx_handler; 1594 void __rcu *rx_handler_data; 1595 1596 struct netdev_queue __rcu *ingress_queue; 1597 unsigned char broadcast[MAX_ADDR_LEN]; 1598 1599 1600 /* 1601 * Cache lines mostly used on transmit path 1602 */ 1603 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1604 unsigned int num_tx_queues; 1605 unsigned int real_num_tx_queues; 1606 struct Qdisc *qdisc; 1607 unsigned long tx_queue_len; 1608 spinlock_t tx_global_lock; 1609 1610 #ifdef CONFIG_XPS 1611 struct xps_dev_maps __rcu *xps_maps; 1612 #endif 1613 #ifdef CONFIG_RFS_ACCEL 1614 struct cpu_rmap *rx_cpu_rmap; 1615 #endif 1616 1617 /* These may be needed for future network-power-down code. */ 1618 1619 /* 1620 * trans_start here is expensive for high speed devices on SMP, 1621 * please use netdev_queue->trans_start instead. 1622 */ 1623 unsigned long trans_start; 1624 1625 int watchdog_timeo; 1626 struct timer_list watchdog_timer; 1627 1628 int __percpu *pcpu_refcnt; 1629 struct list_head todo_list; 1630 1631 struct hlist_node index_hlist; 1632 struct list_head link_watch_list; 1633 1634 enum { NETREG_UNINITIALIZED=0, 1635 NETREG_REGISTERED, /* completed register_netdevice */ 1636 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1637 NETREG_UNREGISTERED, /* completed unregister todo */ 1638 NETREG_RELEASED, /* called free_netdev */ 1639 NETREG_DUMMY, /* dummy device for NAPI poll */ 1640 } reg_state:8; 1641 1642 bool dismantle; 1643 1644 enum { 1645 RTNL_LINK_INITIALIZED, 1646 RTNL_LINK_INITIALIZING, 1647 } rtnl_link_state:16; 1648 1649 void (*destructor)(struct net_device *dev); 1650 1651 #ifdef CONFIG_NETPOLL 1652 struct netpoll_info __rcu *npinfo; 1653 #endif 1654 1655 #ifdef CONFIG_NET_NS 1656 struct net *nd_net; 1657 #endif 1658 1659 /* mid-layer private */ 1660 union { 1661 void *ml_priv; 1662 struct pcpu_lstats __percpu *lstats; 1663 struct pcpu_sw_netstats __percpu *tstats; 1664 struct pcpu_dstats __percpu *dstats; 1665 struct pcpu_vstats __percpu *vstats; 1666 }; 1667 1668 struct garp_port __rcu *garp_port; 1669 struct mrp_port __rcu *mrp_port; 1670 1671 struct device dev; 1672 const struct attribute_group *sysfs_groups[4]; 1673 const struct attribute_group *sysfs_rx_queue_group; 1674 1675 const struct rtnl_link_ops *rtnl_link_ops; 1676 1677 /* for setting kernel sock attribute on TCP connection setup */ 1678 #define GSO_MAX_SIZE 65536 1679 unsigned int gso_max_size; 1680 #define GSO_MAX_SEGS 65535 1681 u16 gso_max_segs; 1682 u16 gso_min_segs; 1683 #ifdef CONFIG_DCB 1684 const struct dcbnl_rtnl_ops *dcbnl_ops; 1685 #endif 1686 u8 num_tc; 1687 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 1688 u8 prio_tc_map[TC_BITMASK + 1]; 1689 1690 #if IS_ENABLED(CONFIG_FCOE) 1691 unsigned int fcoe_ddp_xid; 1692 #endif 1693 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 1694 struct netprio_map __rcu *priomap; 1695 #endif 1696 struct phy_device *phydev; 1697 struct lock_class_key *qdisc_tx_busylock; 1698 int group; 1699 struct pm_qos_request pm_qos_req; 1700 }; 1701 #define to_net_dev(d) container_of(d, struct net_device, dev) 1702 1703 #define NETDEV_ALIGN 32 1704 1705 static inline 1706 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 1707 { 1708 return dev->prio_tc_map[prio & TC_BITMASK]; 1709 } 1710 1711 static inline 1712 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 1713 { 1714 if (tc >= dev->num_tc) 1715 return -EINVAL; 1716 1717 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 1718 return 0; 1719 } 1720 1721 static inline 1722 void netdev_reset_tc(struct net_device *dev) 1723 { 1724 dev->num_tc = 0; 1725 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq)); 1726 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map)); 1727 } 1728 1729 static inline 1730 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset) 1731 { 1732 if (tc >= dev->num_tc) 1733 return -EINVAL; 1734 1735 dev->tc_to_txq[tc].count = count; 1736 dev->tc_to_txq[tc].offset = offset; 1737 return 0; 1738 } 1739 1740 static inline 1741 int netdev_set_num_tc(struct net_device *dev, u8 num_tc) 1742 { 1743 if (num_tc > TC_MAX_QUEUE) 1744 return -EINVAL; 1745 1746 dev->num_tc = num_tc; 1747 return 0; 1748 } 1749 1750 static inline 1751 int netdev_get_num_tc(struct net_device *dev) 1752 { 1753 return dev->num_tc; 1754 } 1755 1756 static inline 1757 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 1758 unsigned int index) 1759 { 1760 return &dev->_tx[index]; 1761 } 1762 1763 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 1764 const struct sk_buff *skb) 1765 { 1766 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 1767 } 1768 1769 static inline void netdev_for_each_tx_queue(struct net_device *dev, 1770 void (*f)(struct net_device *, 1771 struct netdev_queue *, 1772 void *), 1773 void *arg) 1774 { 1775 unsigned int i; 1776 1777 for (i = 0; i < dev->num_tx_queues; i++) 1778 f(dev, &dev->_tx[i], arg); 1779 } 1780 1781 struct netdev_queue *netdev_pick_tx(struct net_device *dev, 1782 struct sk_buff *skb, 1783 void *accel_priv); 1784 1785 /* 1786 * Net namespace inlines 1787 */ 1788 static inline 1789 struct net *dev_net(const struct net_device *dev) 1790 { 1791 return read_pnet(&dev->nd_net); 1792 } 1793 1794 static inline 1795 void dev_net_set(struct net_device *dev, struct net *net) 1796 { 1797 #ifdef CONFIG_NET_NS 1798 release_net(dev->nd_net); 1799 dev->nd_net = hold_net(net); 1800 #endif 1801 } 1802 1803 static inline bool netdev_uses_dsa(struct net_device *dev) 1804 { 1805 #if IS_ENABLED(CONFIG_NET_DSA) 1806 if (dev->dsa_ptr != NULL) 1807 return dsa_uses_tagged_protocol(dev->dsa_ptr); 1808 #endif 1809 return false; 1810 } 1811 1812 /** 1813 * netdev_priv - access network device private data 1814 * @dev: network device 1815 * 1816 * Get network device private data 1817 */ 1818 static inline void *netdev_priv(const struct net_device *dev) 1819 { 1820 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 1821 } 1822 1823 /* Set the sysfs physical device reference for the network logical device 1824 * if set prior to registration will cause a symlink during initialization. 1825 */ 1826 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 1827 1828 /* Set the sysfs device type for the network logical device to allow 1829 * fine-grained identification of different network device types. For 1830 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc. 1831 */ 1832 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 1833 1834 /* Default NAPI poll() weight 1835 * Device drivers are strongly advised to not use bigger value 1836 */ 1837 #define NAPI_POLL_WEIGHT 64 1838 1839 /** 1840 * netif_napi_add - initialize a napi context 1841 * @dev: network device 1842 * @napi: napi context 1843 * @poll: polling function 1844 * @weight: default weight 1845 * 1846 * netif_napi_add() must be used to initialize a napi context prior to calling 1847 * *any* of the other napi related functions. 1848 */ 1849 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 1850 int (*poll)(struct napi_struct *, int), int weight); 1851 1852 /** 1853 * netif_napi_del - remove a napi context 1854 * @napi: napi context 1855 * 1856 * netif_napi_del() removes a napi context from the network device napi list 1857 */ 1858 void netif_napi_del(struct napi_struct *napi); 1859 1860 struct napi_gro_cb { 1861 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 1862 void *frag0; 1863 1864 /* Length of frag0. */ 1865 unsigned int frag0_len; 1866 1867 /* This indicates where we are processing relative to skb->data. */ 1868 int data_offset; 1869 1870 /* This is non-zero if the packet cannot be merged with the new skb. */ 1871 u16 flush; 1872 1873 /* Save the IP ID here and check when we get to the transport layer */ 1874 u16 flush_id; 1875 1876 /* Number of segments aggregated. */ 1877 u16 count; 1878 1879 /* This is non-zero if the packet may be of the same flow. */ 1880 u8 same_flow; 1881 1882 /* Free the skb? */ 1883 u8 free; 1884 #define NAPI_GRO_FREE 1 1885 #define NAPI_GRO_FREE_STOLEN_HEAD 2 1886 1887 /* jiffies when first packet was created/queued */ 1888 unsigned long age; 1889 1890 /* Used in ipv6_gro_receive() and foo-over-udp */ 1891 u16 proto; 1892 1893 /* Used in udp_gro_receive */ 1894 u8 udp_mark:1; 1895 1896 /* GRO checksum is valid */ 1897 u8 csum_valid:1; 1898 1899 /* Number of checksums via CHECKSUM_UNNECESSARY */ 1900 u8 csum_cnt:3; 1901 1902 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 1903 u8 is_ipv6:1; 1904 1905 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 1906 __wsum csum; 1907 1908 /* used in skb_gro_receive() slow path */ 1909 struct sk_buff *last; 1910 }; 1911 1912 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 1913 1914 struct packet_type { 1915 __be16 type; /* This is really htons(ether_type). */ 1916 struct net_device *dev; /* NULL is wildcarded here */ 1917 int (*func) (struct sk_buff *, 1918 struct net_device *, 1919 struct packet_type *, 1920 struct net_device *); 1921 bool (*id_match)(struct packet_type *ptype, 1922 struct sock *sk); 1923 void *af_packet_priv; 1924 struct list_head list; 1925 }; 1926 1927 struct offload_callbacks { 1928 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 1929 netdev_features_t features); 1930 struct sk_buff **(*gro_receive)(struct sk_buff **head, 1931 struct sk_buff *skb); 1932 int (*gro_complete)(struct sk_buff *skb, int nhoff); 1933 }; 1934 1935 struct packet_offload { 1936 __be16 type; /* This is really htons(ether_type). */ 1937 struct offload_callbacks callbacks; 1938 struct list_head list; 1939 }; 1940 1941 struct udp_offload { 1942 __be16 port; 1943 u8 ipproto; 1944 struct offload_callbacks callbacks; 1945 }; 1946 1947 /* often modified stats are per cpu, other are shared (netdev->stats) */ 1948 struct pcpu_sw_netstats { 1949 u64 rx_packets; 1950 u64 rx_bytes; 1951 u64 tx_packets; 1952 u64 tx_bytes; 1953 struct u64_stats_sync syncp; 1954 }; 1955 1956 #define netdev_alloc_pcpu_stats(type) \ 1957 ({ \ 1958 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \ 1959 if (pcpu_stats) { \ 1960 int i; \ 1961 for_each_possible_cpu(i) { \ 1962 typeof(type) *stat; \ 1963 stat = per_cpu_ptr(pcpu_stats, i); \ 1964 u64_stats_init(&stat->syncp); \ 1965 } \ 1966 } \ 1967 pcpu_stats; \ 1968 }) 1969 1970 #include <linux/notifier.h> 1971 1972 /* netdevice notifier chain. Please remember to update the rtnetlink 1973 * notification exclusion list in rtnetlink_event() when adding new 1974 * types. 1975 */ 1976 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */ 1977 #define NETDEV_DOWN 0x0002 1978 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface 1979 detected a hardware crash and restarted 1980 - we can use this eg to kick tcp sessions 1981 once done */ 1982 #define NETDEV_CHANGE 0x0004 /* Notify device state change */ 1983 #define NETDEV_REGISTER 0x0005 1984 #define NETDEV_UNREGISTER 0x0006 1985 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */ 1986 #define NETDEV_CHANGEADDR 0x0008 1987 #define NETDEV_GOING_DOWN 0x0009 1988 #define NETDEV_CHANGENAME 0x000A 1989 #define NETDEV_FEAT_CHANGE 0x000B 1990 #define NETDEV_BONDING_FAILOVER 0x000C 1991 #define NETDEV_PRE_UP 0x000D 1992 #define NETDEV_PRE_TYPE_CHANGE 0x000E 1993 #define NETDEV_POST_TYPE_CHANGE 0x000F 1994 #define NETDEV_POST_INIT 0x0010 1995 #define NETDEV_UNREGISTER_FINAL 0x0011 1996 #define NETDEV_RELEASE 0x0012 1997 #define NETDEV_NOTIFY_PEERS 0x0013 1998 #define NETDEV_JOIN 0x0014 1999 #define NETDEV_CHANGEUPPER 0x0015 2000 #define NETDEV_RESEND_IGMP 0x0016 2001 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */ 2002 #define NETDEV_CHANGEINFODATA 0x0018 2003 2004 int register_netdevice_notifier(struct notifier_block *nb); 2005 int unregister_netdevice_notifier(struct notifier_block *nb); 2006 2007 struct netdev_notifier_info { 2008 struct net_device *dev; 2009 }; 2010 2011 struct netdev_notifier_change_info { 2012 struct netdev_notifier_info info; /* must be first */ 2013 unsigned int flags_changed; 2014 }; 2015 2016 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2017 struct net_device *dev) 2018 { 2019 info->dev = dev; 2020 } 2021 2022 static inline struct net_device * 2023 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2024 { 2025 return info->dev; 2026 } 2027 2028 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2029 2030 2031 extern rwlock_t dev_base_lock; /* Device list lock */ 2032 2033 #define for_each_netdev(net, d) \ 2034 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2035 #define for_each_netdev_reverse(net, d) \ 2036 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2037 #define for_each_netdev_rcu(net, d) \ 2038 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2039 #define for_each_netdev_safe(net, d, n) \ 2040 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2041 #define for_each_netdev_continue(net, d) \ 2042 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2043 #define for_each_netdev_continue_rcu(net, d) \ 2044 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2045 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2046 for_each_netdev_rcu(&init_net, slave) \ 2047 if (netdev_master_upper_dev_get_rcu(slave) == bond) 2048 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2049 2050 static inline struct net_device *next_net_device(struct net_device *dev) 2051 { 2052 struct list_head *lh; 2053 struct net *net; 2054 2055 net = dev_net(dev); 2056 lh = dev->dev_list.next; 2057 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2058 } 2059 2060 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2061 { 2062 struct list_head *lh; 2063 struct net *net; 2064 2065 net = dev_net(dev); 2066 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2067 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2068 } 2069 2070 static inline struct net_device *first_net_device(struct net *net) 2071 { 2072 return list_empty(&net->dev_base_head) ? NULL : 2073 net_device_entry(net->dev_base_head.next); 2074 } 2075 2076 static inline struct net_device *first_net_device_rcu(struct net *net) 2077 { 2078 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2079 2080 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2081 } 2082 2083 int netdev_boot_setup_check(struct net_device *dev); 2084 unsigned long netdev_boot_base(const char *prefix, int unit); 2085 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2086 const char *hwaddr); 2087 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2088 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2089 void dev_add_pack(struct packet_type *pt); 2090 void dev_remove_pack(struct packet_type *pt); 2091 void __dev_remove_pack(struct packet_type *pt); 2092 void dev_add_offload(struct packet_offload *po); 2093 void dev_remove_offload(struct packet_offload *po); 2094 2095 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2096 unsigned short mask); 2097 struct net_device *dev_get_by_name(struct net *net, const char *name); 2098 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2099 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2100 int dev_alloc_name(struct net_device *dev, const char *name); 2101 int dev_open(struct net_device *dev); 2102 int dev_close(struct net_device *dev); 2103 void dev_disable_lro(struct net_device *dev); 2104 int dev_loopback_xmit(struct sk_buff *newskb); 2105 int dev_queue_xmit(struct sk_buff *skb); 2106 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv); 2107 int register_netdevice(struct net_device *dev); 2108 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2109 void unregister_netdevice_many(struct list_head *head); 2110 static inline void unregister_netdevice(struct net_device *dev) 2111 { 2112 unregister_netdevice_queue(dev, NULL); 2113 } 2114 2115 int netdev_refcnt_read(const struct net_device *dev); 2116 void free_netdev(struct net_device *dev); 2117 void netdev_freemem(struct net_device *dev); 2118 void synchronize_net(void); 2119 int init_dummy_netdev(struct net_device *dev); 2120 2121 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2122 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2123 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2124 int netdev_get_name(struct net *net, char *name, int ifindex); 2125 int dev_restart(struct net_device *dev); 2126 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb); 2127 2128 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2129 { 2130 return NAPI_GRO_CB(skb)->data_offset; 2131 } 2132 2133 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2134 { 2135 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2136 } 2137 2138 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2139 { 2140 NAPI_GRO_CB(skb)->data_offset += len; 2141 } 2142 2143 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2144 unsigned int offset) 2145 { 2146 return NAPI_GRO_CB(skb)->frag0 + offset; 2147 } 2148 2149 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2150 { 2151 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2152 } 2153 2154 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2155 unsigned int offset) 2156 { 2157 if (!pskb_may_pull(skb, hlen)) 2158 return NULL; 2159 2160 NAPI_GRO_CB(skb)->frag0 = NULL; 2161 NAPI_GRO_CB(skb)->frag0_len = 0; 2162 return skb->data + offset; 2163 } 2164 2165 static inline void *skb_gro_network_header(struct sk_buff *skb) 2166 { 2167 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2168 skb_network_offset(skb); 2169 } 2170 2171 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2172 const void *start, unsigned int len) 2173 { 2174 if (NAPI_GRO_CB(skb)->csum_valid) 2175 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2176 csum_partial(start, len, 0)); 2177 } 2178 2179 /* GRO checksum functions. These are logical equivalents of the normal 2180 * checksum functions (in skbuff.h) except that they operate on the GRO 2181 * offsets and fields in sk_buff. 2182 */ 2183 2184 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2185 2186 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2187 bool zero_okay, 2188 __sum16 check) 2189 { 2190 return (skb->ip_summed != CHECKSUM_PARTIAL && 2191 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2192 (!zero_okay || check)); 2193 } 2194 2195 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2196 __wsum psum) 2197 { 2198 if (NAPI_GRO_CB(skb)->csum_valid && 2199 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2200 return 0; 2201 2202 NAPI_GRO_CB(skb)->csum = psum; 2203 2204 return __skb_gro_checksum_complete(skb); 2205 } 2206 2207 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2208 { 2209 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2210 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2211 NAPI_GRO_CB(skb)->csum_cnt--; 2212 } else { 2213 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2214 * verified a new top level checksum or an encapsulated one 2215 * during GRO. This saves work if we fallback to normal path. 2216 */ 2217 __skb_incr_checksum_unnecessary(skb); 2218 } 2219 } 2220 2221 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2222 compute_pseudo) \ 2223 ({ \ 2224 __sum16 __ret = 0; \ 2225 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2226 __ret = __skb_gro_checksum_validate_complete(skb, \ 2227 compute_pseudo(skb, proto)); \ 2228 if (__ret) \ 2229 __skb_mark_checksum_bad(skb); \ 2230 else \ 2231 skb_gro_incr_csum_unnecessary(skb); \ 2232 __ret; \ 2233 }) 2234 2235 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2236 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2237 2238 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2239 compute_pseudo) \ 2240 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2241 2242 #define skb_gro_checksum_simple_validate(skb) \ 2243 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2244 2245 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2246 { 2247 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2248 !NAPI_GRO_CB(skb)->csum_valid); 2249 } 2250 2251 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2252 __sum16 check, __wsum pseudo) 2253 { 2254 NAPI_GRO_CB(skb)->csum = ~pseudo; 2255 NAPI_GRO_CB(skb)->csum_valid = 1; 2256 } 2257 2258 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \ 2259 do { \ 2260 if (__skb_gro_checksum_convert_check(skb)) \ 2261 __skb_gro_checksum_convert(skb, check, \ 2262 compute_pseudo(skb, proto)); \ 2263 } while (0) 2264 2265 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2266 unsigned short type, 2267 const void *daddr, const void *saddr, 2268 unsigned int len) 2269 { 2270 if (!dev->header_ops || !dev->header_ops->create) 2271 return 0; 2272 2273 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2274 } 2275 2276 static inline int dev_parse_header(const struct sk_buff *skb, 2277 unsigned char *haddr) 2278 { 2279 const struct net_device *dev = skb->dev; 2280 2281 if (!dev->header_ops || !dev->header_ops->parse) 2282 return 0; 2283 return dev->header_ops->parse(skb, haddr); 2284 } 2285 2286 static inline int dev_rebuild_header(struct sk_buff *skb) 2287 { 2288 const struct net_device *dev = skb->dev; 2289 2290 if (!dev->header_ops || !dev->header_ops->rebuild) 2291 return 0; 2292 return dev->header_ops->rebuild(skb); 2293 } 2294 2295 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len); 2296 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2297 static inline int unregister_gifconf(unsigned int family) 2298 { 2299 return register_gifconf(family, NULL); 2300 } 2301 2302 #ifdef CONFIG_NET_FLOW_LIMIT 2303 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2304 struct sd_flow_limit { 2305 u64 count; 2306 unsigned int num_buckets; 2307 unsigned int history_head; 2308 u16 history[FLOW_LIMIT_HISTORY]; 2309 u8 buckets[]; 2310 }; 2311 2312 extern int netdev_flow_limit_table_len; 2313 #endif /* CONFIG_NET_FLOW_LIMIT */ 2314 2315 /* 2316 * Incoming packets are placed on per-cpu queues 2317 */ 2318 struct softnet_data { 2319 struct Qdisc *output_queue; 2320 struct Qdisc **output_queue_tailp; 2321 struct list_head poll_list; 2322 struct sk_buff *completion_queue; 2323 struct sk_buff_head process_queue; 2324 2325 /* stats */ 2326 unsigned int processed; 2327 unsigned int time_squeeze; 2328 unsigned int cpu_collision; 2329 unsigned int received_rps; 2330 2331 #ifdef CONFIG_RPS 2332 struct softnet_data *rps_ipi_list; 2333 2334 /* Elements below can be accessed between CPUs for RPS */ 2335 struct call_single_data csd ____cacheline_aligned_in_smp; 2336 struct softnet_data *rps_ipi_next; 2337 unsigned int cpu; 2338 unsigned int input_queue_head; 2339 unsigned int input_queue_tail; 2340 #endif 2341 unsigned int dropped; 2342 struct sk_buff_head input_pkt_queue; 2343 struct napi_struct backlog; 2344 2345 #ifdef CONFIG_NET_FLOW_LIMIT 2346 struct sd_flow_limit __rcu *flow_limit; 2347 #endif 2348 }; 2349 2350 static inline void input_queue_head_incr(struct softnet_data *sd) 2351 { 2352 #ifdef CONFIG_RPS 2353 sd->input_queue_head++; 2354 #endif 2355 } 2356 2357 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 2358 unsigned int *qtail) 2359 { 2360 #ifdef CONFIG_RPS 2361 *qtail = ++sd->input_queue_tail; 2362 #endif 2363 } 2364 2365 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 2366 2367 void __netif_schedule(struct Qdisc *q); 2368 void netif_schedule_queue(struct netdev_queue *txq); 2369 2370 static inline void netif_tx_schedule_all(struct net_device *dev) 2371 { 2372 unsigned int i; 2373 2374 for (i = 0; i < dev->num_tx_queues; i++) 2375 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 2376 } 2377 2378 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 2379 { 2380 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2381 } 2382 2383 /** 2384 * netif_start_queue - allow transmit 2385 * @dev: network device 2386 * 2387 * Allow upper layers to call the device hard_start_xmit routine. 2388 */ 2389 static inline void netif_start_queue(struct net_device *dev) 2390 { 2391 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 2392 } 2393 2394 static inline void netif_tx_start_all_queues(struct net_device *dev) 2395 { 2396 unsigned int i; 2397 2398 for (i = 0; i < dev->num_tx_queues; i++) { 2399 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2400 netif_tx_start_queue(txq); 2401 } 2402 } 2403 2404 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 2405 2406 /** 2407 * netif_wake_queue - restart transmit 2408 * @dev: network device 2409 * 2410 * Allow upper layers to call the device hard_start_xmit routine. 2411 * Used for flow control when transmit resources are available. 2412 */ 2413 static inline void netif_wake_queue(struct net_device *dev) 2414 { 2415 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 2416 } 2417 2418 static inline void netif_tx_wake_all_queues(struct net_device *dev) 2419 { 2420 unsigned int i; 2421 2422 for (i = 0; i < dev->num_tx_queues; i++) { 2423 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2424 netif_tx_wake_queue(txq); 2425 } 2426 } 2427 2428 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 2429 { 2430 if (WARN_ON(!dev_queue)) { 2431 pr_info("netif_stop_queue() cannot be called before register_netdev()\n"); 2432 return; 2433 } 2434 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2435 } 2436 2437 /** 2438 * netif_stop_queue - stop transmitted packets 2439 * @dev: network device 2440 * 2441 * Stop upper layers calling the device hard_start_xmit routine. 2442 * Used for flow control when transmit resources are unavailable. 2443 */ 2444 static inline void netif_stop_queue(struct net_device *dev) 2445 { 2446 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 2447 } 2448 2449 static inline void netif_tx_stop_all_queues(struct net_device *dev) 2450 { 2451 unsigned int i; 2452 2453 for (i = 0; i < dev->num_tx_queues; i++) { 2454 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2455 netif_tx_stop_queue(txq); 2456 } 2457 } 2458 2459 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 2460 { 2461 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2462 } 2463 2464 /** 2465 * netif_queue_stopped - test if transmit queue is flowblocked 2466 * @dev: network device 2467 * 2468 * Test if transmit queue on device is currently unable to send. 2469 */ 2470 static inline bool netif_queue_stopped(const struct net_device *dev) 2471 { 2472 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 2473 } 2474 2475 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 2476 { 2477 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 2478 } 2479 2480 static inline bool 2481 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 2482 { 2483 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 2484 } 2485 2486 static inline bool 2487 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 2488 { 2489 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 2490 } 2491 2492 /** 2493 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 2494 * @dev_queue: pointer to transmit queue 2495 * 2496 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 2497 * to give appropriate hint to the cpu. 2498 */ 2499 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 2500 { 2501 #ifdef CONFIG_BQL 2502 prefetchw(&dev_queue->dql.num_queued); 2503 #endif 2504 } 2505 2506 /** 2507 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 2508 * @dev_queue: pointer to transmit queue 2509 * 2510 * BQL enabled drivers might use this helper in their TX completion path, 2511 * to give appropriate hint to the cpu. 2512 */ 2513 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 2514 { 2515 #ifdef CONFIG_BQL 2516 prefetchw(&dev_queue->dql.limit); 2517 #endif 2518 } 2519 2520 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 2521 unsigned int bytes) 2522 { 2523 #ifdef CONFIG_BQL 2524 dql_queued(&dev_queue->dql, bytes); 2525 2526 if (likely(dql_avail(&dev_queue->dql) >= 0)) 2527 return; 2528 2529 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2530 2531 /* 2532 * The XOFF flag must be set before checking the dql_avail below, 2533 * because in netdev_tx_completed_queue we update the dql_completed 2534 * before checking the XOFF flag. 2535 */ 2536 smp_mb(); 2537 2538 /* check again in case another CPU has just made room avail */ 2539 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 2540 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2541 #endif 2542 } 2543 2544 /** 2545 * netdev_sent_queue - report the number of bytes queued to hardware 2546 * @dev: network device 2547 * @bytes: number of bytes queued to the hardware device queue 2548 * 2549 * Report the number of bytes queued for sending/completion to the network 2550 * device hardware queue. @bytes should be a good approximation and should 2551 * exactly match netdev_completed_queue() @bytes 2552 */ 2553 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 2554 { 2555 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 2556 } 2557 2558 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 2559 unsigned int pkts, unsigned int bytes) 2560 { 2561 #ifdef CONFIG_BQL 2562 if (unlikely(!bytes)) 2563 return; 2564 2565 dql_completed(&dev_queue->dql, bytes); 2566 2567 /* 2568 * Without the memory barrier there is a small possiblity that 2569 * netdev_tx_sent_queue will miss the update and cause the queue to 2570 * be stopped forever 2571 */ 2572 smp_mb(); 2573 2574 if (dql_avail(&dev_queue->dql) < 0) 2575 return; 2576 2577 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 2578 netif_schedule_queue(dev_queue); 2579 #endif 2580 } 2581 2582 /** 2583 * netdev_completed_queue - report bytes and packets completed by device 2584 * @dev: network device 2585 * @pkts: actual number of packets sent over the medium 2586 * @bytes: actual number of bytes sent over the medium 2587 * 2588 * Report the number of bytes and packets transmitted by the network device 2589 * hardware queue over the physical medium, @bytes must exactly match the 2590 * @bytes amount passed to netdev_sent_queue() 2591 */ 2592 static inline void netdev_completed_queue(struct net_device *dev, 2593 unsigned int pkts, unsigned int bytes) 2594 { 2595 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 2596 } 2597 2598 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 2599 { 2600 #ifdef CONFIG_BQL 2601 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 2602 dql_reset(&q->dql); 2603 #endif 2604 } 2605 2606 /** 2607 * netdev_reset_queue - reset the packets and bytes count of a network device 2608 * @dev_queue: network device 2609 * 2610 * Reset the bytes and packet count of a network device and clear the 2611 * software flow control OFF bit for this network device 2612 */ 2613 static inline void netdev_reset_queue(struct net_device *dev_queue) 2614 { 2615 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 2616 } 2617 2618 /** 2619 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 2620 * @dev: network device 2621 * @queue_index: given tx queue index 2622 * 2623 * Returns 0 if given tx queue index >= number of device tx queues, 2624 * otherwise returns the originally passed tx queue index. 2625 */ 2626 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 2627 { 2628 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 2629 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 2630 dev->name, queue_index, 2631 dev->real_num_tx_queues); 2632 return 0; 2633 } 2634 2635 return queue_index; 2636 } 2637 2638 /** 2639 * netif_running - test if up 2640 * @dev: network device 2641 * 2642 * Test if the device has been brought up. 2643 */ 2644 static inline bool netif_running(const struct net_device *dev) 2645 { 2646 return test_bit(__LINK_STATE_START, &dev->state); 2647 } 2648 2649 /* 2650 * Routines to manage the subqueues on a device. We only need start 2651 * stop, and a check if it's stopped. All other device management is 2652 * done at the overall netdevice level. 2653 * Also test the device if we're multiqueue. 2654 */ 2655 2656 /** 2657 * netif_start_subqueue - allow sending packets on subqueue 2658 * @dev: network device 2659 * @queue_index: sub queue index 2660 * 2661 * Start individual transmit queue of a device with multiple transmit queues. 2662 */ 2663 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 2664 { 2665 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2666 2667 netif_tx_start_queue(txq); 2668 } 2669 2670 /** 2671 * netif_stop_subqueue - stop sending packets on subqueue 2672 * @dev: network device 2673 * @queue_index: sub queue index 2674 * 2675 * Stop individual transmit queue of a device with multiple transmit queues. 2676 */ 2677 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 2678 { 2679 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2680 netif_tx_stop_queue(txq); 2681 } 2682 2683 /** 2684 * netif_subqueue_stopped - test status of subqueue 2685 * @dev: network device 2686 * @queue_index: sub queue index 2687 * 2688 * Check individual transmit queue of a device with multiple transmit queues. 2689 */ 2690 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 2691 u16 queue_index) 2692 { 2693 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2694 2695 return netif_tx_queue_stopped(txq); 2696 } 2697 2698 static inline bool netif_subqueue_stopped(const struct net_device *dev, 2699 struct sk_buff *skb) 2700 { 2701 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 2702 } 2703 2704 void netif_wake_subqueue(struct net_device *dev, u16 queue_index); 2705 2706 #ifdef CONFIG_XPS 2707 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 2708 u16 index); 2709 #else 2710 static inline int netif_set_xps_queue(struct net_device *dev, 2711 const struct cpumask *mask, 2712 u16 index) 2713 { 2714 return 0; 2715 } 2716 #endif 2717 2718 /* 2719 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used 2720 * as a distribution range limit for the returned value. 2721 */ 2722 static inline u16 skb_tx_hash(const struct net_device *dev, 2723 struct sk_buff *skb) 2724 { 2725 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues); 2726 } 2727 2728 /** 2729 * netif_is_multiqueue - test if device has multiple transmit queues 2730 * @dev: network device 2731 * 2732 * Check if device has multiple transmit queues 2733 */ 2734 static inline bool netif_is_multiqueue(const struct net_device *dev) 2735 { 2736 return dev->num_tx_queues > 1; 2737 } 2738 2739 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 2740 2741 #ifdef CONFIG_SYSFS 2742 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 2743 #else 2744 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 2745 unsigned int rxq) 2746 { 2747 return 0; 2748 } 2749 #endif 2750 2751 static inline int netif_copy_real_num_queues(struct net_device *to_dev, 2752 const struct net_device *from_dev) 2753 { 2754 int err; 2755 2756 err = netif_set_real_num_tx_queues(to_dev, 2757 from_dev->real_num_tx_queues); 2758 if (err) 2759 return err; 2760 #ifdef CONFIG_SYSFS 2761 return netif_set_real_num_rx_queues(to_dev, 2762 from_dev->real_num_rx_queues); 2763 #else 2764 return 0; 2765 #endif 2766 } 2767 2768 #ifdef CONFIG_SYSFS 2769 static inline unsigned int get_netdev_rx_queue_index( 2770 struct netdev_rx_queue *queue) 2771 { 2772 struct net_device *dev = queue->dev; 2773 int index = queue - dev->_rx; 2774 2775 BUG_ON(index >= dev->num_rx_queues); 2776 return index; 2777 } 2778 #endif 2779 2780 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 2781 int netif_get_num_default_rss_queues(void); 2782 2783 enum skb_free_reason { 2784 SKB_REASON_CONSUMED, 2785 SKB_REASON_DROPPED, 2786 }; 2787 2788 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 2789 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 2790 2791 /* 2792 * It is not allowed to call kfree_skb() or consume_skb() from hardware 2793 * interrupt context or with hardware interrupts being disabled. 2794 * (in_irq() || irqs_disabled()) 2795 * 2796 * We provide four helpers that can be used in following contexts : 2797 * 2798 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 2799 * replacing kfree_skb(skb) 2800 * 2801 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 2802 * Typically used in place of consume_skb(skb) in TX completion path 2803 * 2804 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 2805 * replacing kfree_skb(skb) 2806 * 2807 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 2808 * and consumed a packet. Used in place of consume_skb(skb) 2809 */ 2810 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 2811 { 2812 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 2813 } 2814 2815 static inline void dev_consume_skb_irq(struct sk_buff *skb) 2816 { 2817 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 2818 } 2819 2820 static inline void dev_kfree_skb_any(struct sk_buff *skb) 2821 { 2822 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 2823 } 2824 2825 static inline void dev_consume_skb_any(struct sk_buff *skb) 2826 { 2827 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 2828 } 2829 2830 int netif_rx(struct sk_buff *skb); 2831 int netif_rx_ni(struct sk_buff *skb); 2832 int netif_receive_skb(struct sk_buff *skb); 2833 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 2834 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 2835 struct sk_buff *napi_get_frags(struct napi_struct *napi); 2836 gro_result_t napi_gro_frags(struct napi_struct *napi); 2837 struct packet_offload *gro_find_receive_by_type(__be16 type); 2838 struct packet_offload *gro_find_complete_by_type(__be16 type); 2839 2840 static inline void napi_free_frags(struct napi_struct *napi) 2841 { 2842 kfree_skb(napi->skb); 2843 napi->skb = NULL; 2844 } 2845 2846 int netdev_rx_handler_register(struct net_device *dev, 2847 rx_handler_func_t *rx_handler, 2848 void *rx_handler_data); 2849 void netdev_rx_handler_unregister(struct net_device *dev); 2850 2851 bool dev_valid_name(const char *name); 2852 int dev_ioctl(struct net *net, unsigned int cmd, void __user *); 2853 int dev_ethtool(struct net *net, struct ifreq *); 2854 unsigned int dev_get_flags(const struct net_device *); 2855 int __dev_change_flags(struct net_device *, unsigned int flags); 2856 int dev_change_flags(struct net_device *, unsigned int); 2857 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 2858 unsigned int gchanges); 2859 int dev_change_name(struct net_device *, const char *); 2860 int dev_set_alias(struct net_device *, const char *, size_t); 2861 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 2862 int dev_set_mtu(struct net_device *, int); 2863 void dev_set_group(struct net_device *, int); 2864 int dev_set_mac_address(struct net_device *, struct sockaddr *); 2865 int dev_change_carrier(struct net_device *, bool new_carrier); 2866 int dev_get_phys_port_id(struct net_device *dev, 2867 struct netdev_phys_port_id *ppid); 2868 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev); 2869 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 2870 struct netdev_queue *txq, int *ret); 2871 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 2872 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 2873 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb); 2874 2875 extern int netdev_budget; 2876 2877 /* Called by rtnetlink.c:rtnl_unlock() */ 2878 void netdev_run_todo(void); 2879 2880 /** 2881 * dev_put - release reference to device 2882 * @dev: network device 2883 * 2884 * Release reference to device to allow it to be freed. 2885 */ 2886 static inline void dev_put(struct net_device *dev) 2887 { 2888 this_cpu_dec(*dev->pcpu_refcnt); 2889 } 2890 2891 /** 2892 * dev_hold - get reference to device 2893 * @dev: network device 2894 * 2895 * Hold reference to device to keep it from being freed. 2896 */ 2897 static inline void dev_hold(struct net_device *dev) 2898 { 2899 this_cpu_inc(*dev->pcpu_refcnt); 2900 } 2901 2902 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 2903 * and _off may be called from IRQ context, but it is caller 2904 * who is responsible for serialization of these calls. 2905 * 2906 * The name carrier is inappropriate, these functions should really be 2907 * called netif_lowerlayer_*() because they represent the state of any 2908 * kind of lower layer not just hardware media. 2909 */ 2910 2911 void linkwatch_init_dev(struct net_device *dev); 2912 void linkwatch_fire_event(struct net_device *dev); 2913 void linkwatch_forget_dev(struct net_device *dev); 2914 2915 /** 2916 * netif_carrier_ok - test if carrier present 2917 * @dev: network device 2918 * 2919 * Check if carrier is present on device 2920 */ 2921 static inline bool netif_carrier_ok(const struct net_device *dev) 2922 { 2923 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 2924 } 2925 2926 unsigned long dev_trans_start(struct net_device *dev); 2927 2928 void __netdev_watchdog_up(struct net_device *dev); 2929 2930 void netif_carrier_on(struct net_device *dev); 2931 2932 void netif_carrier_off(struct net_device *dev); 2933 2934 /** 2935 * netif_dormant_on - mark device as dormant. 2936 * @dev: network device 2937 * 2938 * Mark device as dormant (as per RFC2863). 2939 * 2940 * The dormant state indicates that the relevant interface is not 2941 * actually in a condition to pass packets (i.e., it is not 'up') but is 2942 * in a "pending" state, waiting for some external event. For "on- 2943 * demand" interfaces, this new state identifies the situation where the 2944 * interface is waiting for events to place it in the up state. 2945 * 2946 */ 2947 static inline void netif_dormant_on(struct net_device *dev) 2948 { 2949 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 2950 linkwatch_fire_event(dev); 2951 } 2952 2953 /** 2954 * netif_dormant_off - set device as not dormant. 2955 * @dev: network device 2956 * 2957 * Device is not in dormant state. 2958 */ 2959 static inline void netif_dormant_off(struct net_device *dev) 2960 { 2961 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 2962 linkwatch_fire_event(dev); 2963 } 2964 2965 /** 2966 * netif_dormant - test if carrier present 2967 * @dev: network device 2968 * 2969 * Check if carrier is present on device 2970 */ 2971 static inline bool netif_dormant(const struct net_device *dev) 2972 { 2973 return test_bit(__LINK_STATE_DORMANT, &dev->state); 2974 } 2975 2976 2977 /** 2978 * netif_oper_up - test if device is operational 2979 * @dev: network device 2980 * 2981 * Check if carrier is operational 2982 */ 2983 static inline bool netif_oper_up(const struct net_device *dev) 2984 { 2985 return (dev->operstate == IF_OPER_UP || 2986 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 2987 } 2988 2989 /** 2990 * netif_device_present - is device available or removed 2991 * @dev: network device 2992 * 2993 * Check if device has not been removed from system. 2994 */ 2995 static inline bool netif_device_present(struct net_device *dev) 2996 { 2997 return test_bit(__LINK_STATE_PRESENT, &dev->state); 2998 } 2999 3000 void netif_device_detach(struct net_device *dev); 3001 3002 void netif_device_attach(struct net_device *dev); 3003 3004 /* 3005 * Network interface message level settings 3006 */ 3007 3008 enum { 3009 NETIF_MSG_DRV = 0x0001, 3010 NETIF_MSG_PROBE = 0x0002, 3011 NETIF_MSG_LINK = 0x0004, 3012 NETIF_MSG_TIMER = 0x0008, 3013 NETIF_MSG_IFDOWN = 0x0010, 3014 NETIF_MSG_IFUP = 0x0020, 3015 NETIF_MSG_RX_ERR = 0x0040, 3016 NETIF_MSG_TX_ERR = 0x0080, 3017 NETIF_MSG_TX_QUEUED = 0x0100, 3018 NETIF_MSG_INTR = 0x0200, 3019 NETIF_MSG_TX_DONE = 0x0400, 3020 NETIF_MSG_RX_STATUS = 0x0800, 3021 NETIF_MSG_PKTDATA = 0x1000, 3022 NETIF_MSG_HW = 0x2000, 3023 NETIF_MSG_WOL = 0x4000, 3024 }; 3025 3026 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3027 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3028 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3029 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3030 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3031 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3032 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3033 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3034 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3035 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3036 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3037 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3038 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3039 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3040 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3041 3042 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3043 { 3044 /* use default */ 3045 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3046 return default_msg_enable_bits; 3047 if (debug_value == 0) /* no output */ 3048 return 0; 3049 /* set low N bits */ 3050 return (1 << debug_value) - 1; 3051 } 3052 3053 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 3054 { 3055 spin_lock(&txq->_xmit_lock); 3056 txq->xmit_lock_owner = cpu; 3057 } 3058 3059 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 3060 { 3061 spin_lock_bh(&txq->_xmit_lock); 3062 txq->xmit_lock_owner = smp_processor_id(); 3063 } 3064 3065 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 3066 { 3067 bool ok = spin_trylock(&txq->_xmit_lock); 3068 if (likely(ok)) 3069 txq->xmit_lock_owner = smp_processor_id(); 3070 return ok; 3071 } 3072 3073 static inline void __netif_tx_unlock(struct netdev_queue *txq) 3074 { 3075 txq->xmit_lock_owner = -1; 3076 spin_unlock(&txq->_xmit_lock); 3077 } 3078 3079 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 3080 { 3081 txq->xmit_lock_owner = -1; 3082 spin_unlock_bh(&txq->_xmit_lock); 3083 } 3084 3085 static inline void txq_trans_update(struct netdev_queue *txq) 3086 { 3087 if (txq->xmit_lock_owner != -1) 3088 txq->trans_start = jiffies; 3089 } 3090 3091 /** 3092 * netif_tx_lock - grab network device transmit lock 3093 * @dev: network device 3094 * 3095 * Get network device transmit lock 3096 */ 3097 static inline void netif_tx_lock(struct net_device *dev) 3098 { 3099 unsigned int i; 3100 int cpu; 3101 3102 spin_lock(&dev->tx_global_lock); 3103 cpu = smp_processor_id(); 3104 for (i = 0; i < dev->num_tx_queues; i++) { 3105 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3106 3107 /* We are the only thread of execution doing a 3108 * freeze, but we have to grab the _xmit_lock in 3109 * order to synchronize with threads which are in 3110 * the ->hard_start_xmit() handler and already 3111 * checked the frozen bit. 3112 */ 3113 __netif_tx_lock(txq, cpu); 3114 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 3115 __netif_tx_unlock(txq); 3116 } 3117 } 3118 3119 static inline void netif_tx_lock_bh(struct net_device *dev) 3120 { 3121 local_bh_disable(); 3122 netif_tx_lock(dev); 3123 } 3124 3125 static inline void netif_tx_unlock(struct net_device *dev) 3126 { 3127 unsigned int i; 3128 3129 for (i = 0; i < dev->num_tx_queues; i++) { 3130 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3131 3132 /* No need to grab the _xmit_lock here. If the 3133 * queue is not stopped for another reason, we 3134 * force a schedule. 3135 */ 3136 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 3137 netif_schedule_queue(txq); 3138 } 3139 spin_unlock(&dev->tx_global_lock); 3140 } 3141 3142 static inline void netif_tx_unlock_bh(struct net_device *dev) 3143 { 3144 netif_tx_unlock(dev); 3145 local_bh_enable(); 3146 } 3147 3148 #define HARD_TX_LOCK(dev, txq, cpu) { \ 3149 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3150 __netif_tx_lock(txq, cpu); \ 3151 } \ 3152 } 3153 3154 #define HARD_TX_TRYLOCK(dev, txq) \ 3155 (((dev->features & NETIF_F_LLTX) == 0) ? \ 3156 __netif_tx_trylock(txq) : \ 3157 true ) 3158 3159 #define HARD_TX_UNLOCK(dev, txq) { \ 3160 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3161 __netif_tx_unlock(txq); \ 3162 } \ 3163 } 3164 3165 static inline void netif_tx_disable(struct net_device *dev) 3166 { 3167 unsigned int i; 3168 int cpu; 3169 3170 local_bh_disable(); 3171 cpu = smp_processor_id(); 3172 for (i = 0; i < dev->num_tx_queues; i++) { 3173 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3174 3175 __netif_tx_lock(txq, cpu); 3176 netif_tx_stop_queue(txq); 3177 __netif_tx_unlock(txq); 3178 } 3179 local_bh_enable(); 3180 } 3181 3182 static inline void netif_addr_lock(struct net_device *dev) 3183 { 3184 spin_lock(&dev->addr_list_lock); 3185 } 3186 3187 static inline void netif_addr_lock_nested(struct net_device *dev) 3188 { 3189 int subclass = SINGLE_DEPTH_NESTING; 3190 3191 if (dev->netdev_ops->ndo_get_lock_subclass) 3192 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev); 3193 3194 spin_lock_nested(&dev->addr_list_lock, subclass); 3195 } 3196 3197 static inline void netif_addr_lock_bh(struct net_device *dev) 3198 { 3199 spin_lock_bh(&dev->addr_list_lock); 3200 } 3201 3202 static inline void netif_addr_unlock(struct net_device *dev) 3203 { 3204 spin_unlock(&dev->addr_list_lock); 3205 } 3206 3207 static inline void netif_addr_unlock_bh(struct net_device *dev) 3208 { 3209 spin_unlock_bh(&dev->addr_list_lock); 3210 } 3211 3212 /* 3213 * dev_addrs walker. Should be used only for read access. Call with 3214 * rcu_read_lock held. 3215 */ 3216 #define for_each_dev_addr(dev, ha) \ 3217 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 3218 3219 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 3220 3221 void ether_setup(struct net_device *dev); 3222 3223 /* Support for loadable net-drivers */ 3224 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 3225 unsigned char name_assign_type, 3226 void (*setup)(struct net_device *), 3227 unsigned int txqs, unsigned int rxqs); 3228 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 3229 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 3230 3231 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 3232 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 3233 count) 3234 3235 int register_netdev(struct net_device *dev); 3236 void unregister_netdev(struct net_device *dev); 3237 3238 /* General hardware address lists handling functions */ 3239 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 3240 struct netdev_hw_addr_list *from_list, int addr_len); 3241 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 3242 struct netdev_hw_addr_list *from_list, int addr_len); 3243 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 3244 struct net_device *dev, 3245 int (*sync)(struct net_device *, const unsigned char *), 3246 int (*unsync)(struct net_device *, 3247 const unsigned char *)); 3248 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 3249 struct net_device *dev, 3250 int (*unsync)(struct net_device *, 3251 const unsigned char *)); 3252 void __hw_addr_init(struct netdev_hw_addr_list *list); 3253 3254 /* Functions used for device addresses handling */ 3255 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 3256 unsigned char addr_type); 3257 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 3258 unsigned char addr_type); 3259 void dev_addr_flush(struct net_device *dev); 3260 int dev_addr_init(struct net_device *dev); 3261 3262 /* Functions used for unicast addresses handling */ 3263 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 3264 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 3265 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 3266 int dev_uc_sync(struct net_device *to, struct net_device *from); 3267 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 3268 void dev_uc_unsync(struct net_device *to, struct net_device *from); 3269 void dev_uc_flush(struct net_device *dev); 3270 void dev_uc_init(struct net_device *dev); 3271 3272 /** 3273 * __dev_uc_sync - Synchonize device's unicast list 3274 * @dev: device to sync 3275 * @sync: function to call if address should be added 3276 * @unsync: function to call if address should be removed 3277 * 3278 * Add newly added addresses to the interface, and release 3279 * addresses that have been deleted. 3280 **/ 3281 static inline int __dev_uc_sync(struct net_device *dev, 3282 int (*sync)(struct net_device *, 3283 const unsigned char *), 3284 int (*unsync)(struct net_device *, 3285 const unsigned char *)) 3286 { 3287 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 3288 } 3289 3290 /** 3291 * __dev_uc_unsync - Remove synchronized addresses from device 3292 * @dev: device to sync 3293 * @unsync: function to call if address should be removed 3294 * 3295 * Remove all addresses that were added to the device by dev_uc_sync(). 3296 **/ 3297 static inline void __dev_uc_unsync(struct net_device *dev, 3298 int (*unsync)(struct net_device *, 3299 const unsigned char *)) 3300 { 3301 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 3302 } 3303 3304 /* Functions used for multicast addresses handling */ 3305 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 3306 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 3307 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 3308 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 3309 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 3310 int dev_mc_sync(struct net_device *to, struct net_device *from); 3311 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 3312 void dev_mc_unsync(struct net_device *to, struct net_device *from); 3313 void dev_mc_flush(struct net_device *dev); 3314 void dev_mc_init(struct net_device *dev); 3315 3316 /** 3317 * __dev_mc_sync - Synchonize device's multicast list 3318 * @dev: device to sync 3319 * @sync: function to call if address should be added 3320 * @unsync: function to call if address should be removed 3321 * 3322 * Add newly added addresses to the interface, and release 3323 * addresses that have been deleted. 3324 **/ 3325 static inline int __dev_mc_sync(struct net_device *dev, 3326 int (*sync)(struct net_device *, 3327 const unsigned char *), 3328 int (*unsync)(struct net_device *, 3329 const unsigned char *)) 3330 { 3331 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 3332 } 3333 3334 /** 3335 * __dev_mc_unsync - Remove synchronized addresses from device 3336 * @dev: device to sync 3337 * @unsync: function to call if address should be removed 3338 * 3339 * Remove all addresses that were added to the device by dev_mc_sync(). 3340 **/ 3341 static inline void __dev_mc_unsync(struct net_device *dev, 3342 int (*unsync)(struct net_device *, 3343 const unsigned char *)) 3344 { 3345 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 3346 } 3347 3348 /* Functions used for secondary unicast and multicast support */ 3349 void dev_set_rx_mode(struct net_device *dev); 3350 void __dev_set_rx_mode(struct net_device *dev); 3351 int dev_set_promiscuity(struct net_device *dev, int inc); 3352 int dev_set_allmulti(struct net_device *dev, int inc); 3353 void netdev_state_change(struct net_device *dev); 3354 void netdev_notify_peers(struct net_device *dev); 3355 void netdev_features_change(struct net_device *dev); 3356 /* Load a device via the kmod */ 3357 void dev_load(struct net *net, const char *name); 3358 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 3359 struct rtnl_link_stats64 *storage); 3360 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 3361 const struct net_device_stats *netdev_stats); 3362 3363 extern int netdev_max_backlog; 3364 extern int netdev_tstamp_prequeue; 3365 extern int weight_p; 3366 extern int bpf_jit_enable; 3367 3368 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 3369 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 3370 struct list_head **iter); 3371 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 3372 struct list_head **iter); 3373 3374 /* iterate through upper list, must be called under RCU read lock */ 3375 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 3376 for (iter = &(dev)->adj_list.upper, \ 3377 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 3378 updev; \ 3379 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 3380 3381 /* iterate through upper list, must be called under RCU read lock */ 3382 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \ 3383 for (iter = &(dev)->all_adj_list.upper, \ 3384 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \ 3385 updev; \ 3386 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter))) 3387 3388 void *netdev_lower_get_next_private(struct net_device *dev, 3389 struct list_head **iter); 3390 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 3391 struct list_head **iter); 3392 3393 #define netdev_for_each_lower_private(dev, priv, iter) \ 3394 for (iter = (dev)->adj_list.lower.next, \ 3395 priv = netdev_lower_get_next_private(dev, &(iter)); \ 3396 priv; \ 3397 priv = netdev_lower_get_next_private(dev, &(iter))) 3398 3399 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 3400 for (iter = &(dev)->adj_list.lower, \ 3401 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 3402 priv; \ 3403 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 3404 3405 void *netdev_lower_get_next(struct net_device *dev, 3406 struct list_head **iter); 3407 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 3408 for (iter = &(dev)->adj_list.lower, \ 3409 ldev = netdev_lower_get_next(dev, &(iter)); \ 3410 ldev; \ 3411 ldev = netdev_lower_get_next(dev, &(iter))) 3412 3413 void *netdev_adjacent_get_private(struct list_head *adj_list); 3414 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 3415 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 3416 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 3417 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev); 3418 int netdev_master_upper_dev_link(struct net_device *dev, 3419 struct net_device *upper_dev); 3420 int netdev_master_upper_dev_link_private(struct net_device *dev, 3421 struct net_device *upper_dev, 3422 void *private); 3423 void netdev_upper_dev_unlink(struct net_device *dev, 3424 struct net_device *upper_dev); 3425 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 3426 void *netdev_lower_dev_get_private(struct net_device *dev, 3427 struct net_device *lower_dev); 3428 int dev_get_nest_level(struct net_device *dev, 3429 bool (*type_check)(struct net_device *dev)); 3430 int skb_checksum_help(struct sk_buff *skb); 3431 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 3432 netdev_features_t features, bool tx_path); 3433 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 3434 netdev_features_t features); 3435 3436 static inline 3437 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 3438 { 3439 return __skb_gso_segment(skb, features, true); 3440 } 3441 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 3442 3443 static inline bool can_checksum_protocol(netdev_features_t features, 3444 __be16 protocol) 3445 { 3446 return ((features & NETIF_F_GEN_CSUM) || 3447 ((features & NETIF_F_V4_CSUM) && 3448 protocol == htons(ETH_P_IP)) || 3449 ((features & NETIF_F_V6_CSUM) && 3450 protocol == htons(ETH_P_IPV6)) || 3451 ((features & NETIF_F_FCOE_CRC) && 3452 protocol == htons(ETH_P_FCOE))); 3453 } 3454 3455 #ifdef CONFIG_BUG 3456 void netdev_rx_csum_fault(struct net_device *dev); 3457 #else 3458 static inline void netdev_rx_csum_fault(struct net_device *dev) 3459 { 3460 } 3461 #endif 3462 /* rx skb timestamps */ 3463 void net_enable_timestamp(void); 3464 void net_disable_timestamp(void); 3465 3466 #ifdef CONFIG_PROC_FS 3467 int __init dev_proc_init(void); 3468 #else 3469 #define dev_proc_init() 0 3470 #endif 3471 3472 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 3473 struct sk_buff *skb, struct net_device *dev, 3474 bool more) 3475 { 3476 skb->xmit_more = more ? 1 : 0; 3477 return ops->ndo_start_xmit(skb, dev); 3478 } 3479 3480 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 3481 struct netdev_queue *txq, bool more) 3482 { 3483 const struct net_device_ops *ops = dev->netdev_ops; 3484 int rc; 3485 3486 rc = __netdev_start_xmit(ops, skb, dev, more); 3487 if (rc == NETDEV_TX_OK) 3488 txq_trans_update(txq); 3489 3490 return rc; 3491 } 3492 3493 int netdev_class_create_file_ns(struct class_attribute *class_attr, 3494 const void *ns); 3495 void netdev_class_remove_file_ns(struct class_attribute *class_attr, 3496 const void *ns); 3497 3498 static inline int netdev_class_create_file(struct class_attribute *class_attr) 3499 { 3500 return netdev_class_create_file_ns(class_attr, NULL); 3501 } 3502 3503 static inline void netdev_class_remove_file(struct class_attribute *class_attr) 3504 { 3505 netdev_class_remove_file_ns(class_attr, NULL); 3506 } 3507 3508 extern struct kobj_ns_type_operations net_ns_type_operations; 3509 3510 const char *netdev_drivername(const struct net_device *dev); 3511 3512 void linkwatch_run_queue(void); 3513 3514 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 3515 netdev_features_t f2) 3516 { 3517 if (f1 & NETIF_F_GEN_CSUM) 3518 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM); 3519 if (f2 & NETIF_F_GEN_CSUM) 3520 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM); 3521 f1 &= f2; 3522 if (f1 & NETIF_F_GEN_CSUM) 3523 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM); 3524 3525 return f1; 3526 } 3527 3528 static inline netdev_features_t netdev_get_wanted_features( 3529 struct net_device *dev) 3530 { 3531 return (dev->features & ~dev->hw_features) | dev->wanted_features; 3532 } 3533 netdev_features_t netdev_increment_features(netdev_features_t all, 3534 netdev_features_t one, netdev_features_t mask); 3535 3536 /* Allow TSO being used on stacked device : 3537 * Performing the GSO segmentation before last device 3538 * is a performance improvement. 3539 */ 3540 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 3541 netdev_features_t mask) 3542 { 3543 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 3544 } 3545 3546 int __netdev_update_features(struct net_device *dev); 3547 void netdev_update_features(struct net_device *dev); 3548 void netdev_change_features(struct net_device *dev); 3549 3550 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 3551 struct net_device *dev); 3552 3553 netdev_features_t netif_skb_features(struct sk_buff *skb); 3554 3555 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 3556 { 3557 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT; 3558 3559 /* check flags correspondence */ 3560 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 3561 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT)); 3562 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 3563 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 3564 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 3565 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 3566 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 3567 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 3568 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT)); 3569 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT)); 3570 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 3571 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 3572 BUILD_BUG_ON(SKB_GSO_MPLS != (NETIF_F_GSO_MPLS >> NETIF_F_GSO_SHIFT)); 3573 3574 return (features & feature) == feature; 3575 } 3576 3577 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 3578 { 3579 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 3580 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 3581 } 3582 3583 static inline bool netif_needs_gso(struct net_device *dev, struct sk_buff *skb, 3584 netdev_features_t features) 3585 { 3586 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 3587 (dev->netdev_ops->ndo_gso_check && 3588 !dev->netdev_ops->ndo_gso_check(skb, dev)) || 3589 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 3590 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 3591 } 3592 3593 static inline void netif_set_gso_max_size(struct net_device *dev, 3594 unsigned int size) 3595 { 3596 dev->gso_max_size = size; 3597 } 3598 3599 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 3600 int pulled_hlen, u16 mac_offset, 3601 int mac_len) 3602 { 3603 skb->protocol = protocol; 3604 skb->encapsulation = 1; 3605 skb_push(skb, pulled_hlen); 3606 skb_reset_transport_header(skb); 3607 skb->mac_header = mac_offset; 3608 skb->network_header = skb->mac_header + mac_len; 3609 skb->mac_len = mac_len; 3610 } 3611 3612 static inline bool netif_is_macvlan(struct net_device *dev) 3613 { 3614 return dev->priv_flags & IFF_MACVLAN; 3615 } 3616 3617 static inline bool netif_is_bond_master(struct net_device *dev) 3618 { 3619 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 3620 } 3621 3622 static inline bool netif_is_bond_slave(struct net_device *dev) 3623 { 3624 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 3625 } 3626 3627 static inline bool netif_supports_nofcs(struct net_device *dev) 3628 { 3629 return dev->priv_flags & IFF_SUPP_NOFCS; 3630 } 3631 3632 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 3633 static inline void netif_keep_dst(struct net_device *dev) 3634 { 3635 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 3636 } 3637 3638 extern struct pernet_operations __net_initdata loopback_net_ops; 3639 3640 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 3641 3642 /* netdev_printk helpers, similar to dev_printk */ 3643 3644 static inline const char *netdev_name(const struct net_device *dev) 3645 { 3646 if (!dev->name[0] || strchr(dev->name, '%')) 3647 return "(unnamed net_device)"; 3648 return dev->name; 3649 } 3650 3651 static inline const char *netdev_reg_state(const struct net_device *dev) 3652 { 3653 switch (dev->reg_state) { 3654 case NETREG_UNINITIALIZED: return " (uninitialized)"; 3655 case NETREG_REGISTERED: return ""; 3656 case NETREG_UNREGISTERING: return " (unregistering)"; 3657 case NETREG_UNREGISTERED: return " (unregistered)"; 3658 case NETREG_RELEASED: return " (released)"; 3659 case NETREG_DUMMY: return " (dummy)"; 3660 } 3661 3662 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 3663 return " (unknown)"; 3664 } 3665 3666 __printf(3, 4) 3667 void netdev_printk(const char *level, const struct net_device *dev, 3668 const char *format, ...); 3669 __printf(2, 3) 3670 void netdev_emerg(const struct net_device *dev, const char *format, ...); 3671 __printf(2, 3) 3672 void netdev_alert(const struct net_device *dev, const char *format, ...); 3673 __printf(2, 3) 3674 void netdev_crit(const struct net_device *dev, const char *format, ...); 3675 __printf(2, 3) 3676 void netdev_err(const struct net_device *dev, const char *format, ...); 3677 __printf(2, 3) 3678 void netdev_warn(const struct net_device *dev, const char *format, ...); 3679 __printf(2, 3) 3680 void netdev_notice(const struct net_device *dev, const char *format, ...); 3681 __printf(2, 3) 3682 void netdev_info(const struct net_device *dev, const char *format, ...); 3683 3684 #define MODULE_ALIAS_NETDEV(device) \ 3685 MODULE_ALIAS("netdev-" device) 3686 3687 #if defined(CONFIG_DYNAMIC_DEBUG) 3688 #define netdev_dbg(__dev, format, args...) \ 3689 do { \ 3690 dynamic_netdev_dbg(__dev, format, ##args); \ 3691 } while (0) 3692 #elif defined(DEBUG) 3693 #define netdev_dbg(__dev, format, args...) \ 3694 netdev_printk(KERN_DEBUG, __dev, format, ##args) 3695 #else 3696 #define netdev_dbg(__dev, format, args...) \ 3697 ({ \ 3698 if (0) \ 3699 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 3700 }) 3701 #endif 3702 3703 #if defined(VERBOSE_DEBUG) 3704 #define netdev_vdbg netdev_dbg 3705 #else 3706 3707 #define netdev_vdbg(dev, format, args...) \ 3708 ({ \ 3709 if (0) \ 3710 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 3711 0; \ 3712 }) 3713 #endif 3714 3715 /* 3716 * netdev_WARN() acts like dev_printk(), but with the key difference 3717 * of using a WARN/WARN_ON to get the message out, including the 3718 * file/line information and a backtrace. 3719 */ 3720 #define netdev_WARN(dev, format, args...) \ 3721 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \ 3722 netdev_reg_state(dev), ##args) 3723 3724 /* netif printk helpers, similar to netdev_printk */ 3725 3726 #define netif_printk(priv, type, level, dev, fmt, args...) \ 3727 do { \ 3728 if (netif_msg_##type(priv)) \ 3729 netdev_printk(level, (dev), fmt, ##args); \ 3730 } while (0) 3731 3732 #define netif_level(level, priv, type, dev, fmt, args...) \ 3733 do { \ 3734 if (netif_msg_##type(priv)) \ 3735 netdev_##level(dev, fmt, ##args); \ 3736 } while (0) 3737 3738 #define netif_emerg(priv, type, dev, fmt, args...) \ 3739 netif_level(emerg, priv, type, dev, fmt, ##args) 3740 #define netif_alert(priv, type, dev, fmt, args...) \ 3741 netif_level(alert, priv, type, dev, fmt, ##args) 3742 #define netif_crit(priv, type, dev, fmt, args...) \ 3743 netif_level(crit, priv, type, dev, fmt, ##args) 3744 #define netif_err(priv, type, dev, fmt, args...) \ 3745 netif_level(err, priv, type, dev, fmt, ##args) 3746 #define netif_warn(priv, type, dev, fmt, args...) \ 3747 netif_level(warn, priv, type, dev, fmt, ##args) 3748 #define netif_notice(priv, type, dev, fmt, args...) \ 3749 netif_level(notice, priv, type, dev, fmt, ##args) 3750 #define netif_info(priv, type, dev, fmt, args...) \ 3751 netif_level(info, priv, type, dev, fmt, ##args) 3752 3753 #if defined(CONFIG_DYNAMIC_DEBUG) 3754 #define netif_dbg(priv, type, netdev, format, args...) \ 3755 do { \ 3756 if (netif_msg_##type(priv)) \ 3757 dynamic_netdev_dbg(netdev, format, ##args); \ 3758 } while (0) 3759 #elif defined(DEBUG) 3760 #define netif_dbg(priv, type, dev, format, args...) \ 3761 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 3762 #else 3763 #define netif_dbg(priv, type, dev, format, args...) \ 3764 ({ \ 3765 if (0) \ 3766 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 3767 0; \ 3768 }) 3769 #endif 3770 3771 #if defined(VERBOSE_DEBUG) 3772 #define netif_vdbg netif_dbg 3773 #else 3774 #define netif_vdbg(priv, type, dev, format, args...) \ 3775 ({ \ 3776 if (0) \ 3777 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 3778 0; \ 3779 }) 3780 #endif 3781 3782 /* 3783 * The list of packet types we will receive (as opposed to discard) 3784 * and the routines to invoke. 3785 * 3786 * Why 16. Because with 16 the only overlap we get on a hash of the 3787 * low nibble of the protocol value is RARP/SNAP/X.25. 3788 * 3789 * NOTE: That is no longer true with the addition of VLAN tags. Not 3790 * sure which should go first, but I bet it won't make much 3791 * difference if we are running VLANs. The good news is that 3792 * this protocol won't be in the list unless compiled in, so 3793 * the average user (w/out VLANs) will not be adversely affected. 3794 * --BLG 3795 * 3796 * 0800 IP 3797 * 8100 802.1Q VLAN 3798 * 0001 802.3 3799 * 0002 AX.25 3800 * 0004 802.2 3801 * 8035 RARP 3802 * 0005 SNAP 3803 * 0805 X.25 3804 * 0806 ARP 3805 * 8137 IPX 3806 * 0009 Localtalk 3807 * 86DD IPv6 3808 */ 3809 #define PTYPE_HASH_SIZE (16) 3810 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 3811 3812 #endif /* _LINUX_NETDEVICE_H */ 3813