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 #include <uapi/linux/if_bonding.h> 55 56 struct netpoll_info; 57 struct device; 58 struct phy_device; 59 /* 802.11 specific */ 60 struct wireless_dev; 61 /* 802.15.4 specific */ 62 struct wpan_dev; 63 64 void netdev_set_default_ethtool_ops(struct net_device *dev, 65 const struct ethtool_ops *ops); 66 67 /* Backlog congestion levels */ 68 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 69 #define NET_RX_DROP 1 /* packet dropped */ 70 71 /* 72 * Transmit return codes: transmit return codes originate from three different 73 * namespaces: 74 * 75 * - qdisc return codes 76 * - driver transmit return codes 77 * - errno values 78 * 79 * Drivers are allowed to return any one of those in their hard_start_xmit() 80 * function. Real network devices commonly used with qdiscs should only return 81 * the driver transmit return codes though - when qdiscs are used, the actual 82 * transmission happens asynchronously, so the value is not propagated to 83 * higher layers. Virtual network devices transmit synchronously, in this case 84 * the driver transmit return codes are consumed by dev_queue_xmit(), all 85 * others are propagated to higher layers. 86 */ 87 88 /* qdisc ->enqueue() return codes. */ 89 #define NET_XMIT_SUCCESS 0x00 90 #define NET_XMIT_DROP 0x01 /* skb dropped */ 91 #define NET_XMIT_CN 0x02 /* congestion notification */ 92 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */ 93 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 94 95 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 96 * indicates that the device will soon be dropping packets, or already drops 97 * some packets of the same priority; prompting us to send less aggressively. */ 98 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 99 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 100 101 /* Driver transmit return codes */ 102 #define NETDEV_TX_MASK 0xf0 103 104 enum netdev_tx { 105 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 106 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 107 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 108 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */ 109 }; 110 typedef enum netdev_tx netdev_tx_t; 111 112 /* 113 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 114 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 115 */ 116 static inline bool dev_xmit_complete(int rc) 117 { 118 /* 119 * Positive cases with an skb consumed by a driver: 120 * - successful transmission (rc == NETDEV_TX_OK) 121 * - error while transmitting (rc < 0) 122 * - error while queueing to a different device (rc & NET_XMIT_MASK) 123 */ 124 if (likely(rc < NET_XMIT_MASK)) 125 return true; 126 127 return false; 128 } 129 130 /* 131 * Compute the worst case header length according to the protocols 132 * used. 133 */ 134 135 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 136 # if defined(CONFIG_MAC80211_MESH) 137 # define LL_MAX_HEADER 128 138 # else 139 # define LL_MAX_HEADER 96 140 # endif 141 #else 142 # define LL_MAX_HEADER 32 143 #endif 144 145 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 146 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 147 #define MAX_HEADER LL_MAX_HEADER 148 #else 149 #define MAX_HEADER (LL_MAX_HEADER + 48) 150 #endif 151 152 /* 153 * Old network device statistics. Fields are native words 154 * (unsigned long) so they can be read and written atomically. 155 */ 156 157 struct net_device_stats { 158 unsigned long rx_packets; 159 unsigned long tx_packets; 160 unsigned long rx_bytes; 161 unsigned long tx_bytes; 162 unsigned long rx_errors; 163 unsigned long tx_errors; 164 unsigned long rx_dropped; 165 unsigned long tx_dropped; 166 unsigned long multicast; 167 unsigned long collisions; 168 unsigned long rx_length_errors; 169 unsigned long rx_over_errors; 170 unsigned long rx_crc_errors; 171 unsigned long rx_frame_errors; 172 unsigned long rx_fifo_errors; 173 unsigned long rx_missed_errors; 174 unsigned long tx_aborted_errors; 175 unsigned long tx_carrier_errors; 176 unsigned long tx_fifo_errors; 177 unsigned long tx_heartbeat_errors; 178 unsigned long tx_window_errors; 179 unsigned long rx_compressed; 180 unsigned long tx_compressed; 181 }; 182 183 184 #include <linux/cache.h> 185 #include <linux/skbuff.h> 186 187 #ifdef CONFIG_RPS 188 #include <linux/static_key.h> 189 extern struct static_key rps_needed; 190 #endif 191 192 struct neighbour; 193 struct neigh_parms; 194 struct sk_buff; 195 196 struct netdev_hw_addr { 197 struct list_head list; 198 unsigned char addr[MAX_ADDR_LEN]; 199 unsigned char type; 200 #define NETDEV_HW_ADDR_T_LAN 1 201 #define NETDEV_HW_ADDR_T_SAN 2 202 #define NETDEV_HW_ADDR_T_SLAVE 3 203 #define NETDEV_HW_ADDR_T_UNICAST 4 204 #define NETDEV_HW_ADDR_T_MULTICAST 5 205 bool global_use; 206 int sync_cnt; 207 int refcount; 208 int synced; 209 struct rcu_head rcu_head; 210 }; 211 212 struct netdev_hw_addr_list { 213 struct list_head list; 214 int count; 215 }; 216 217 #define netdev_hw_addr_list_count(l) ((l)->count) 218 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 219 #define netdev_hw_addr_list_for_each(ha, l) \ 220 list_for_each_entry(ha, &(l)->list, list) 221 222 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 223 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 224 #define netdev_for_each_uc_addr(ha, dev) \ 225 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 226 227 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 228 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 229 #define netdev_for_each_mc_addr(ha, dev) \ 230 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 231 232 struct hh_cache { 233 u16 hh_len; 234 u16 __pad; 235 seqlock_t hh_lock; 236 237 /* cached hardware header; allow for machine alignment needs. */ 238 #define HH_DATA_MOD 16 239 #define HH_DATA_OFF(__len) \ 240 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 241 #define HH_DATA_ALIGN(__len) \ 242 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 243 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 244 }; 245 246 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much. 247 * Alternative is: 248 * dev->hard_header_len ? (dev->hard_header_len + 249 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 250 * 251 * We could use other alignment values, but we must maintain the 252 * relationship HH alignment <= LL alignment. 253 */ 254 #define LL_RESERVED_SPACE(dev) \ 255 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 256 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 257 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 258 259 struct header_ops { 260 int (*create) (struct sk_buff *skb, struct net_device *dev, 261 unsigned short type, const void *daddr, 262 const void *saddr, unsigned int len); 263 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 264 int (*rebuild)(struct sk_buff *skb); 265 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 266 void (*cache_update)(struct hh_cache *hh, 267 const struct net_device *dev, 268 const unsigned char *haddr); 269 }; 270 271 /* These flag bits are private to the generic network queueing 272 * layer, they may not be explicitly referenced by any other 273 * code. 274 */ 275 276 enum netdev_state_t { 277 __LINK_STATE_START, 278 __LINK_STATE_PRESENT, 279 __LINK_STATE_NOCARRIER, 280 __LINK_STATE_LINKWATCH_PENDING, 281 __LINK_STATE_DORMANT, 282 }; 283 284 285 /* 286 * This structure holds at boot time configured netdevice settings. They 287 * are then used in the device probing. 288 */ 289 struct netdev_boot_setup { 290 char name[IFNAMSIZ]; 291 struct ifmap map; 292 }; 293 #define NETDEV_BOOT_SETUP_MAX 8 294 295 int __init netdev_boot_setup(char *str); 296 297 /* 298 * Structure for NAPI scheduling similar to tasklet but with weighting 299 */ 300 struct napi_struct { 301 /* The poll_list must only be managed by the entity which 302 * changes the state of the NAPI_STATE_SCHED bit. This means 303 * whoever atomically sets that bit can add this napi_struct 304 * to the per-cpu poll_list, and whoever clears that bit 305 * can remove from the list right before clearing the bit. 306 */ 307 struct list_head poll_list; 308 309 unsigned long state; 310 int weight; 311 unsigned int gro_count; 312 int (*poll)(struct napi_struct *, int); 313 #ifdef CONFIG_NETPOLL 314 spinlock_t poll_lock; 315 int poll_owner; 316 #endif 317 struct net_device *dev; 318 struct sk_buff *gro_list; 319 struct sk_buff *skb; 320 struct hrtimer timer; 321 struct list_head dev_list; 322 struct hlist_node napi_hash_node; 323 unsigned int napi_id; 324 }; 325 326 enum { 327 NAPI_STATE_SCHED, /* Poll is scheduled */ 328 NAPI_STATE_DISABLE, /* Disable pending */ 329 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 330 NAPI_STATE_HASHED, /* In NAPI hash */ 331 }; 332 333 enum gro_result { 334 GRO_MERGED, 335 GRO_MERGED_FREE, 336 GRO_HELD, 337 GRO_NORMAL, 338 GRO_DROP, 339 }; 340 typedef enum gro_result gro_result_t; 341 342 /* 343 * enum rx_handler_result - Possible return values for rx_handlers. 344 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 345 * further. 346 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 347 * case skb->dev was changed by rx_handler. 348 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 349 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called. 350 * 351 * rx_handlers are functions called from inside __netif_receive_skb(), to do 352 * special processing of the skb, prior to delivery to protocol handlers. 353 * 354 * Currently, a net_device can only have a single rx_handler registered. Trying 355 * to register a second rx_handler will return -EBUSY. 356 * 357 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 358 * To unregister a rx_handler on a net_device, use 359 * netdev_rx_handler_unregister(). 360 * 361 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 362 * do with the skb. 363 * 364 * If the rx_handler consumed to skb in some way, it should return 365 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 366 * the skb to be delivered in some other ways. 367 * 368 * If the rx_handler changed skb->dev, to divert the skb to another 369 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 370 * new device will be called if it exists. 371 * 372 * If the rx_handler consider the skb should be ignored, it should return 373 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 374 * are registered on exact device (ptype->dev == skb->dev). 375 * 376 * If the rx_handler didn't changed skb->dev, but want the skb to be normally 377 * delivered, it should return RX_HANDLER_PASS. 378 * 379 * A device without a registered rx_handler will behave as if rx_handler 380 * returned RX_HANDLER_PASS. 381 */ 382 383 enum rx_handler_result { 384 RX_HANDLER_CONSUMED, 385 RX_HANDLER_ANOTHER, 386 RX_HANDLER_EXACT, 387 RX_HANDLER_PASS, 388 }; 389 typedef enum rx_handler_result rx_handler_result_t; 390 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 391 392 void __napi_schedule(struct napi_struct *n); 393 void __napi_schedule_irqoff(struct napi_struct *n); 394 395 static inline bool napi_disable_pending(struct napi_struct *n) 396 { 397 return test_bit(NAPI_STATE_DISABLE, &n->state); 398 } 399 400 /** 401 * napi_schedule_prep - check if napi can be scheduled 402 * @n: napi context 403 * 404 * Test if NAPI routine is already running, and if not mark 405 * it as running. This is used as a condition variable 406 * insure only one NAPI poll instance runs. We also make 407 * sure there is no pending NAPI disable. 408 */ 409 static inline bool napi_schedule_prep(struct napi_struct *n) 410 { 411 return !napi_disable_pending(n) && 412 !test_and_set_bit(NAPI_STATE_SCHED, &n->state); 413 } 414 415 /** 416 * napi_schedule - schedule NAPI poll 417 * @n: napi context 418 * 419 * Schedule NAPI poll routine to be called if it is not already 420 * running. 421 */ 422 static inline void napi_schedule(struct napi_struct *n) 423 { 424 if (napi_schedule_prep(n)) 425 __napi_schedule(n); 426 } 427 428 /** 429 * napi_schedule_irqoff - schedule NAPI poll 430 * @n: napi context 431 * 432 * Variant of napi_schedule(), assuming hard irqs are masked. 433 */ 434 static inline void napi_schedule_irqoff(struct napi_struct *n) 435 { 436 if (napi_schedule_prep(n)) 437 __napi_schedule_irqoff(n); 438 } 439 440 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 441 static inline bool napi_reschedule(struct napi_struct *napi) 442 { 443 if (napi_schedule_prep(napi)) { 444 __napi_schedule(napi); 445 return true; 446 } 447 return false; 448 } 449 450 void __napi_complete(struct napi_struct *n); 451 void napi_complete_done(struct napi_struct *n, int work_done); 452 /** 453 * napi_complete - NAPI processing complete 454 * @n: napi context 455 * 456 * Mark NAPI processing as complete. 457 * Consider using napi_complete_done() instead. 458 */ 459 static inline void napi_complete(struct napi_struct *n) 460 { 461 return napi_complete_done(n, 0); 462 } 463 464 /** 465 * napi_by_id - lookup a NAPI by napi_id 466 * @napi_id: hashed napi_id 467 * 468 * lookup @napi_id in napi_hash table 469 * must be called under rcu_read_lock() 470 */ 471 struct napi_struct *napi_by_id(unsigned int napi_id); 472 473 /** 474 * napi_hash_add - add a NAPI to global hashtable 475 * @napi: napi context 476 * 477 * generate a new napi_id and store a @napi under it in napi_hash 478 */ 479 void napi_hash_add(struct napi_struct *napi); 480 481 /** 482 * napi_hash_del - remove a NAPI from global table 483 * @napi: napi context 484 * 485 * Warning: caller must observe rcu grace period 486 * before freeing memory containing @napi 487 */ 488 void napi_hash_del(struct napi_struct *napi); 489 490 /** 491 * napi_disable - prevent NAPI from scheduling 492 * @n: napi context 493 * 494 * Stop NAPI from being scheduled on this context. 495 * Waits till any outstanding processing completes. 496 */ 497 void napi_disable(struct napi_struct *n); 498 499 /** 500 * napi_enable - enable NAPI scheduling 501 * @n: napi context 502 * 503 * Resume NAPI from being scheduled on this context. 504 * Must be paired with napi_disable. 505 */ 506 static inline void napi_enable(struct napi_struct *n) 507 { 508 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 509 smp_mb__before_atomic(); 510 clear_bit(NAPI_STATE_SCHED, &n->state); 511 } 512 513 #ifdef CONFIG_SMP 514 /** 515 * napi_synchronize - wait until NAPI is not running 516 * @n: napi context 517 * 518 * Wait until NAPI is done being scheduled on this context. 519 * Waits till any outstanding processing completes but 520 * does not disable future activations. 521 */ 522 static inline void napi_synchronize(const struct napi_struct *n) 523 { 524 while (test_bit(NAPI_STATE_SCHED, &n->state)) 525 msleep(1); 526 } 527 #else 528 # define napi_synchronize(n) barrier() 529 #endif 530 531 enum netdev_queue_state_t { 532 __QUEUE_STATE_DRV_XOFF, 533 __QUEUE_STATE_STACK_XOFF, 534 __QUEUE_STATE_FROZEN, 535 }; 536 537 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 538 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 539 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 540 541 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 542 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 543 QUEUE_STATE_FROZEN) 544 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 545 QUEUE_STATE_FROZEN) 546 547 /* 548 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 549 * netif_tx_* functions below are used to manipulate this flag. The 550 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 551 * queue independently. The netif_xmit_*stopped functions below are called 552 * to check if the queue has been stopped by the driver or stack (either 553 * of the XOFF bits are set in the state). Drivers should not need to call 554 * netif_xmit*stopped functions, they should only be using netif_tx_*. 555 */ 556 557 struct netdev_queue { 558 /* 559 * read mostly part 560 */ 561 struct net_device *dev; 562 struct Qdisc __rcu *qdisc; 563 struct Qdisc *qdisc_sleeping; 564 #ifdef CONFIG_SYSFS 565 struct kobject kobj; 566 #endif 567 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 568 int numa_node; 569 #endif 570 /* 571 * write mostly part 572 */ 573 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 574 int xmit_lock_owner; 575 /* 576 * please use this field instead of dev->trans_start 577 */ 578 unsigned long trans_start; 579 580 /* 581 * Number of TX timeouts for this queue 582 * (/sys/class/net/DEV/Q/trans_timeout) 583 */ 584 unsigned long trans_timeout; 585 586 unsigned long state; 587 588 #ifdef CONFIG_BQL 589 struct dql dql; 590 #endif 591 } ____cacheline_aligned_in_smp; 592 593 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 594 { 595 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 596 return q->numa_node; 597 #else 598 return NUMA_NO_NODE; 599 #endif 600 } 601 602 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 603 { 604 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 605 q->numa_node = node; 606 #endif 607 } 608 609 #ifdef CONFIG_RPS 610 /* 611 * This structure holds an RPS map which can be of variable length. The 612 * map is an array of CPUs. 613 */ 614 struct rps_map { 615 unsigned int len; 616 struct rcu_head rcu; 617 u16 cpus[0]; 618 }; 619 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 620 621 /* 622 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 623 * tail pointer for that CPU's input queue at the time of last enqueue, and 624 * a hardware filter index. 625 */ 626 struct rps_dev_flow { 627 u16 cpu; 628 u16 filter; 629 unsigned int last_qtail; 630 }; 631 #define RPS_NO_FILTER 0xffff 632 633 /* 634 * The rps_dev_flow_table structure contains a table of flow mappings. 635 */ 636 struct rps_dev_flow_table { 637 unsigned int mask; 638 struct rcu_head rcu; 639 struct rps_dev_flow flows[0]; 640 }; 641 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 642 ((_num) * sizeof(struct rps_dev_flow))) 643 644 /* 645 * The rps_sock_flow_table contains mappings of flows to the last CPU 646 * on which they were processed by the application (set in recvmsg). 647 * Each entry is a 32bit value. Upper part is the high order bits 648 * of flow hash, lower part is cpu number. 649 * rps_cpu_mask is used to partition the space, depending on number of 650 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 651 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f, 652 * meaning we use 32-6=26 bits for the hash. 653 */ 654 struct rps_sock_flow_table { 655 u32 mask; 656 657 u32 ents[0] ____cacheline_aligned_in_smp; 658 }; 659 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 660 661 #define RPS_NO_CPU 0xffff 662 663 extern u32 rps_cpu_mask; 664 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 665 666 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 667 u32 hash) 668 { 669 if (table && hash) { 670 unsigned int index = hash & table->mask; 671 u32 val = hash & ~rps_cpu_mask; 672 673 /* We only give a hint, preemption can change cpu under us */ 674 val |= raw_smp_processor_id(); 675 676 if (table->ents[index] != val) 677 table->ents[index] = val; 678 } 679 } 680 681 #ifdef CONFIG_RFS_ACCEL 682 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 683 u16 filter_id); 684 #endif 685 #endif /* CONFIG_RPS */ 686 687 /* This structure contains an instance of an RX queue. */ 688 struct netdev_rx_queue { 689 #ifdef CONFIG_RPS 690 struct rps_map __rcu *rps_map; 691 struct rps_dev_flow_table __rcu *rps_flow_table; 692 #endif 693 struct kobject kobj; 694 struct net_device *dev; 695 } ____cacheline_aligned_in_smp; 696 697 /* 698 * RX queue sysfs structures and functions. 699 */ 700 struct rx_queue_attribute { 701 struct attribute attr; 702 ssize_t (*show)(struct netdev_rx_queue *queue, 703 struct rx_queue_attribute *attr, char *buf); 704 ssize_t (*store)(struct netdev_rx_queue *queue, 705 struct rx_queue_attribute *attr, const char *buf, size_t len); 706 }; 707 708 #ifdef CONFIG_XPS 709 /* 710 * This structure holds an XPS map which can be of variable length. The 711 * map is an array of queues. 712 */ 713 struct xps_map { 714 unsigned int len; 715 unsigned int alloc_len; 716 struct rcu_head rcu; 717 u16 queues[0]; 718 }; 719 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 720 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \ 721 / sizeof(u16)) 722 723 /* 724 * This structure holds all XPS maps for device. Maps are indexed by CPU. 725 */ 726 struct xps_dev_maps { 727 struct rcu_head rcu; 728 struct xps_map __rcu *cpu_map[0]; 729 }; 730 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \ 731 (nr_cpu_ids * sizeof(struct xps_map *))) 732 #endif /* CONFIG_XPS */ 733 734 #define TC_MAX_QUEUE 16 735 #define TC_BITMASK 15 736 /* HW offloaded queuing disciplines txq count and offset maps */ 737 struct netdev_tc_txq { 738 u16 count; 739 u16 offset; 740 }; 741 742 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 743 /* 744 * This structure is to hold information about the device 745 * configured to run FCoE protocol stack. 746 */ 747 struct netdev_fcoe_hbainfo { 748 char manufacturer[64]; 749 char serial_number[64]; 750 char hardware_version[64]; 751 char driver_version[64]; 752 char optionrom_version[64]; 753 char firmware_version[64]; 754 char model[256]; 755 char model_description[256]; 756 }; 757 #endif 758 759 #define MAX_PHYS_ITEM_ID_LEN 32 760 761 /* This structure holds a unique identifier to identify some 762 * physical item (port for example) used by a netdevice. 763 */ 764 struct netdev_phys_item_id { 765 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 766 unsigned char id_len; 767 }; 768 769 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 770 struct sk_buff *skb); 771 772 /* 773 * This structure defines the management hooks for network devices. 774 * The following hooks can be defined; unless noted otherwise, they are 775 * optional and can be filled with a null pointer. 776 * 777 * int (*ndo_init)(struct net_device *dev); 778 * This function is called once when network device is registered. 779 * The network device can use this to any late stage initializaton 780 * or semantic validattion. It can fail with an error code which will 781 * be propogated back to register_netdev 782 * 783 * void (*ndo_uninit)(struct net_device *dev); 784 * This function is called when device is unregistered or when registration 785 * fails. It is not called if init fails. 786 * 787 * int (*ndo_open)(struct net_device *dev); 788 * This function is called when network device transistions to the up 789 * state. 790 * 791 * int (*ndo_stop)(struct net_device *dev); 792 * This function is called when network device transistions to the down 793 * state. 794 * 795 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 796 * struct net_device *dev); 797 * Called when a packet needs to be transmitted. 798 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY. 799 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX) 800 * Required can not be NULL. 801 * 802 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 803 * void *accel_priv, select_queue_fallback_t fallback); 804 * Called to decide which queue to when device supports multiple 805 * transmit queues. 806 * 807 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 808 * This function is called to allow device receiver to make 809 * changes to configuration when multicast or promiscious is enabled. 810 * 811 * void (*ndo_set_rx_mode)(struct net_device *dev); 812 * This function is called device changes address list filtering. 813 * If driver handles unicast address filtering, it should set 814 * IFF_UNICAST_FLT to its priv_flags. 815 * 816 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 817 * This function is called when the Media Access Control address 818 * needs to be changed. If this interface is not defined, the 819 * mac address can not be changed. 820 * 821 * int (*ndo_validate_addr)(struct net_device *dev); 822 * Test if Media Access Control address is valid for the device. 823 * 824 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 825 * Called when a user request an ioctl which can't be handled by 826 * the generic interface code. If not defined ioctl's return 827 * not supported error code. 828 * 829 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 830 * Used to set network devices bus interface parameters. This interface 831 * is retained for legacy reason, new devices should use the bus 832 * interface (PCI) for low level management. 833 * 834 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 835 * Called when a user wants to change the Maximum Transfer Unit 836 * of a device. If not defined, any request to change MTU will 837 * will return an error. 838 * 839 * void (*ndo_tx_timeout)(struct net_device *dev); 840 * Callback uses when the transmitter has not made any progress 841 * for dev->watchdog ticks. 842 * 843 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev, 844 * struct rtnl_link_stats64 *storage); 845 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 846 * Called when a user wants to get the network device usage 847 * statistics. Drivers must do one of the following: 848 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 849 * rtnl_link_stats64 structure passed by the caller. 850 * 2. Define @ndo_get_stats to update a net_device_stats structure 851 * (which should normally be dev->stats) and return a pointer to 852 * it. The structure may be changed asynchronously only if each 853 * field is written atomically. 854 * 3. Update dev->stats asynchronously and atomically, and define 855 * neither operation. 856 * 857 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 858 * If device support VLAN filtering this function is called when a 859 * VLAN id is registered. 860 * 861 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 862 * If device support VLAN filtering this function is called when a 863 * VLAN id is unregistered. 864 * 865 * void (*ndo_poll_controller)(struct net_device *dev); 866 * 867 * SR-IOV management functions. 868 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 869 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos); 870 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 871 * int max_tx_rate); 872 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 873 * int (*ndo_get_vf_config)(struct net_device *dev, 874 * int vf, struct ifla_vf_info *ivf); 875 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 876 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 877 * struct nlattr *port[]); 878 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 879 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc) 880 * Called to setup 'tc' number of traffic classes in the net device. This 881 * is always called from the stack with the rtnl lock held and netif tx 882 * queues stopped. This allows the netdevice to perform queue management 883 * safely. 884 * 885 * Fiber Channel over Ethernet (FCoE) offload functions. 886 * int (*ndo_fcoe_enable)(struct net_device *dev); 887 * Called when the FCoE protocol stack wants to start using LLD for FCoE 888 * so the underlying device can perform whatever needed configuration or 889 * initialization to support acceleration of FCoE traffic. 890 * 891 * int (*ndo_fcoe_disable)(struct net_device *dev); 892 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 893 * so the underlying device can perform whatever needed clean-ups to 894 * stop supporting acceleration of FCoE traffic. 895 * 896 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 897 * struct scatterlist *sgl, unsigned int sgc); 898 * Called when the FCoE Initiator wants to initialize an I/O that 899 * is a possible candidate for Direct Data Placement (DDP). The LLD can 900 * perform necessary setup and returns 1 to indicate the device is set up 901 * successfully to perform DDP on this I/O, otherwise this returns 0. 902 * 903 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 904 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 905 * indicated by the FC exchange id 'xid', so the underlying device can 906 * clean up and reuse resources for later DDP requests. 907 * 908 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 909 * struct scatterlist *sgl, unsigned int sgc); 910 * Called when the FCoE Target wants to initialize an I/O that 911 * is a possible candidate for Direct Data Placement (DDP). The LLD can 912 * perform necessary setup and returns 1 to indicate the device is set up 913 * successfully to perform DDP on this I/O, otherwise this returns 0. 914 * 915 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 916 * struct netdev_fcoe_hbainfo *hbainfo); 917 * Called when the FCoE Protocol stack wants information on the underlying 918 * device. This information is utilized by the FCoE protocol stack to 919 * register attributes with Fiber Channel management service as per the 920 * FC-GS Fabric Device Management Information(FDMI) specification. 921 * 922 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 923 * Called when the underlying device wants to override default World Wide 924 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 925 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 926 * protocol stack to use. 927 * 928 * RFS acceleration. 929 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 930 * u16 rxq_index, u32 flow_id); 931 * Set hardware filter for RFS. rxq_index is the target queue index; 932 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 933 * Return the filter ID on success, or a negative error code. 934 * 935 * Slave management functions (for bridge, bonding, etc). 936 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 937 * Called to make another netdev an underling. 938 * 939 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 940 * Called to release previously enslaved netdev. 941 * 942 * Feature/offload setting functions. 943 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 944 * netdev_features_t features); 945 * Adjusts the requested feature flags according to device-specific 946 * constraints, and returns the resulting flags. Must not modify 947 * the device state. 948 * 949 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 950 * Called to update device configuration to new features. Passed 951 * feature set might be less than what was returned by ndo_fix_features()). 952 * Must return >0 or -errno if it changed dev->features itself. 953 * 954 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 955 * struct net_device *dev, 956 * const unsigned char *addr, u16 vid, u16 flags) 957 * Adds an FDB entry to dev for addr. 958 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 959 * struct net_device *dev, 960 * const unsigned char *addr, u16 vid) 961 * Deletes the FDB entry from dev coresponding to addr. 962 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 963 * struct net_device *dev, struct net_device *filter_dev, 964 * int idx) 965 * Used to add FDB entries to dump requests. Implementers should add 966 * entries to skb and update idx with the number of entries. 967 * 968 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh) 969 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 970 * struct net_device *dev, u32 filter_mask) 971 * 972 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 973 * Called to change device carrier. Soft-devices (like dummy, team, etc) 974 * which do not represent real hardware may define this to allow their 975 * userspace components to manage their virtual carrier state. Devices 976 * that determine carrier state from physical hardware properties (eg 977 * network cables) or protocol-dependent mechanisms (eg 978 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 979 * 980 * int (*ndo_get_phys_port_id)(struct net_device *dev, 981 * struct netdev_phys_item_id *ppid); 982 * Called to get ID of physical port of this device. If driver does 983 * not implement this, it is assumed that the hw is not able to have 984 * multiple net devices on single physical port. 985 * 986 * void (*ndo_add_vxlan_port)(struct net_device *dev, 987 * sa_family_t sa_family, __be16 port); 988 * Called by vxlan to notiy a driver about the UDP port and socket 989 * address family that vxlan is listnening to. It is called only when 990 * a new port starts listening. The operation is protected by the 991 * vxlan_net->sock_lock. 992 * 993 * void (*ndo_del_vxlan_port)(struct net_device *dev, 994 * sa_family_t sa_family, __be16 port); 995 * Called by vxlan to notify the driver about a UDP port and socket 996 * address family that vxlan is not listening to anymore. The operation 997 * is protected by the vxlan_net->sock_lock. 998 * 999 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1000 * struct net_device *dev) 1001 * Called by upper layer devices to accelerate switching or other 1002 * station functionality into hardware. 'pdev is the lowerdev 1003 * to use for the offload and 'dev' is the net device that will 1004 * back the offload. Returns a pointer to the private structure 1005 * the upper layer will maintain. 1006 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1007 * Called by upper layer device to delete the station created 1008 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1009 * the station and priv is the structure returned by the add 1010 * operation. 1011 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb, 1012 * struct net_device *dev, 1013 * void *priv); 1014 * Callback to use for xmit over the accelerated station. This 1015 * is used in place of ndo_start_xmit on accelerated net 1016 * devices. 1017 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb, 1018 * struct net_device *dev 1019 * netdev_features_t features); 1020 * Called by core transmit path to determine if device is capable of 1021 * performing offload operations on a given packet. This is to give 1022 * the device an opportunity to implement any restrictions that cannot 1023 * be otherwise expressed by feature flags. The check is called with 1024 * the set of features that the stack has calculated and it returns 1025 * those the driver believes to be appropriate. 1026 * 1027 * int (*ndo_switch_parent_id_get)(struct net_device *dev, 1028 * struct netdev_phys_item_id *psid); 1029 * Called to get an ID of the switch chip this port is part of. 1030 * If driver implements this, it indicates that it represents a port 1031 * of a switch chip. 1032 * int (*ndo_switch_port_stp_update)(struct net_device *dev, u8 state); 1033 * Called to notify switch device port of bridge port STP 1034 * state change. 1035 */ 1036 struct net_device_ops { 1037 int (*ndo_init)(struct net_device *dev); 1038 void (*ndo_uninit)(struct net_device *dev); 1039 int (*ndo_open)(struct net_device *dev); 1040 int (*ndo_stop)(struct net_device *dev); 1041 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb, 1042 struct net_device *dev); 1043 u16 (*ndo_select_queue)(struct net_device *dev, 1044 struct sk_buff *skb, 1045 void *accel_priv, 1046 select_queue_fallback_t fallback); 1047 void (*ndo_change_rx_flags)(struct net_device *dev, 1048 int flags); 1049 void (*ndo_set_rx_mode)(struct net_device *dev); 1050 int (*ndo_set_mac_address)(struct net_device *dev, 1051 void *addr); 1052 int (*ndo_validate_addr)(struct net_device *dev); 1053 int (*ndo_do_ioctl)(struct net_device *dev, 1054 struct ifreq *ifr, int cmd); 1055 int (*ndo_set_config)(struct net_device *dev, 1056 struct ifmap *map); 1057 int (*ndo_change_mtu)(struct net_device *dev, 1058 int new_mtu); 1059 int (*ndo_neigh_setup)(struct net_device *dev, 1060 struct neigh_parms *); 1061 void (*ndo_tx_timeout) (struct net_device *dev); 1062 1063 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev, 1064 struct rtnl_link_stats64 *storage); 1065 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1066 1067 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1068 __be16 proto, u16 vid); 1069 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1070 __be16 proto, u16 vid); 1071 #ifdef CONFIG_NET_POLL_CONTROLLER 1072 void (*ndo_poll_controller)(struct net_device *dev); 1073 int (*ndo_netpoll_setup)(struct net_device *dev, 1074 struct netpoll_info *info); 1075 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1076 #endif 1077 #ifdef CONFIG_NET_RX_BUSY_POLL 1078 int (*ndo_busy_poll)(struct napi_struct *dev); 1079 #endif 1080 int (*ndo_set_vf_mac)(struct net_device *dev, 1081 int queue, u8 *mac); 1082 int (*ndo_set_vf_vlan)(struct net_device *dev, 1083 int queue, u16 vlan, u8 qos); 1084 int (*ndo_set_vf_rate)(struct net_device *dev, 1085 int vf, int min_tx_rate, 1086 int max_tx_rate); 1087 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1088 int vf, bool setting); 1089 int (*ndo_get_vf_config)(struct net_device *dev, 1090 int vf, 1091 struct ifla_vf_info *ivf); 1092 int (*ndo_set_vf_link_state)(struct net_device *dev, 1093 int vf, int link_state); 1094 int (*ndo_set_vf_port)(struct net_device *dev, 1095 int vf, 1096 struct nlattr *port[]); 1097 int (*ndo_get_vf_port)(struct net_device *dev, 1098 int vf, struct sk_buff *skb); 1099 int (*ndo_setup_tc)(struct net_device *dev, u8 tc); 1100 #if IS_ENABLED(CONFIG_FCOE) 1101 int (*ndo_fcoe_enable)(struct net_device *dev); 1102 int (*ndo_fcoe_disable)(struct net_device *dev); 1103 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1104 u16 xid, 1105 struct scatterlist *sgl, 1106 unsigned int sgc); 1107 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1108 u16 xid); 1109 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1110 u16 xid, 1111 struct scatterlist *sgl, 1112 unsigned int sgc); 1113 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1114 struct netdev_fcoe_hbainfo *hbainfo); 1115 #endif 1116 1117 #if IS_ENABLED(CONFIG_LIBFCOE) 1118 #define NETDEV_FCOE_WWNN 0 1119 #define NETDEV_FCOE_WWPN 1 1120 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1121 u64 *wwn, int type); 1122 #endif 1123 1124 #ifdef CONFIG_RFS_ACCEL 1125 int (*ndo_rx_flow_steer)(struct net_device *dev, 1126 const struct sk_buff *skb, 1127 u16 rxq_index, 1128 u32 flow_id); 1129 #endif 1130 int (*ndo_add_slave)(struct net_device *dev, 1131 struct net_device *slave_dev); 1132 int (*ndo_del_slave)(struct net_device *dev, 1133 struct net_device *slave_dev); 1134 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1135 netdev_features_t features); 1136 int (*ndo_set_features)(struct net_device *dev, 1137 netdev_features_t features); 1138 int (*ndo_neigh_construct)(struct neighbour *n); 1139 void (*ndo_neigh_destroy)(struct neighbour *n); 1140 1141 int (*ndo_fdb_add)(struct ndmsg *ndm, 1142 struct nlattr *tb[], 1143 struct net_device *dev, 1144 const unsigned char *addr, 1145 u16 vid, 1146 u16 flags); 1147 int (*ndo_fdb_del)(struct ndmsg *ndm, 1148 struct nlattr *tb[], 1149 struct net_device *dev, 1150 const unsigned char *addr, 1151 u16 vid); 1152 int (*ndo_fdb_dump)(struct sk_buff *skb, 1153 struct netlink_callback *cb, 1154 struct net_device *dev, 1155 struct net_device *filter_dev, 1156 int idx); 1157 1158 int (*ndo_bridge_setlink)(struct net_device *dev, 1159 struct nlmsghdr *nlh, 1160 u16 flags); 1161 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1162 u32 pid, u32 seq, 1163 struct net_device *dev, 1164 u32 filter_mask); 1165 int (*ndo_bridge_dellink)(struct net_device *dev, 1166 struct nlmsghdr *nlh, 1167 u16 flags); 1168 int (*ndo_change_carrier)(struct net_device *dev, 1169 bool new_carrier); 1170 int (*ndo_get_phys_port_id)(struct net_device *dev, 1171 struct netdev_phys_item_id *ppid); 1172 void (*ndo_add_vxlan_port)(struct net_device *dev, 1173 sa_family_t sa_family, 1174 __be16 port); 1175 void (*ndo_del_vxlan_port)(struct net_device *dev, 1176 sa_family_t sa_family, 1177 __be16 port); 1178 1179 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1180 struct net_device *dev); 1181 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1182 void *priv); 1183 1184 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb, 1185 struct net_device *dev, 1186 void *priv); 1187 int (*ndo_get_lock_subclass)(struct net_device *dev); 1188 netdev_features_t (*ndo_features_check) (struct sk_buff *skb, 1189 struct net_device *dev, 1190 netdev_features_t features); 1191 #ifdef CONFIG_NET_SWITCHDEV 1192 int (*ndo_switch_parent_id_get)(struct net_device *dev, 1193 struct netdev_phys_item_id *psid); 1194 int (*ndo_switch_port_stp_update)(struct net_device *dev, 1195 u8 state); 1196 #endif 1197 }; 1198 1199 /** 1200 * enum net_device_priv_flags - &struct net_device priv_flags 1201 * 1202 * These are the &struct net_device, they are only set internally 1203 * by drivers and used in the kernel. These flags are invisible to 1204 * userspace, this means that the order of these flags can change 1205 * during any kernel release. 1206 * 1207 * You should have a pretty good reason to be extending these flags. 1208 * 1209 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1210 * @IFF_EBRIDGE: Ethernet bridging device 1211 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active 1212 * @IFF_MASTER_8023AD: bonding master, 802.3ad 1213 * @IFF_MASTER_ALB: bonding master, balance-alb 1214 * @IFF_BONDING: bonding master or slave 1215 * @IFF_SLAVE_NEEDARP: need ARPs for validation 1216 * @IFF_ISATAP: ISATAP interface (RFC4214) 1217 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use 1218 * @IFF_WAN_HDLC: WAN HDLC device 1219 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1220 * release skb->dst 1221 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1222 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1223 * @IFF_MACVLAN_PORT: device used as macvlan port 1224 * @IFF_BRIDGE_PORT: device used as bridge port 1225 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1226 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1227 * @IFF_UNICAST_FLT: Supports unicast filtering 1228 * @IFF_TEAM_PORT: device used as team port 1229 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1230 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1231 * change when it's running 1232 * @IFF_MACVLAN: Macvlan device 1233 */ 1234 enum netdev_priv_flags { 1235 IFF_802_1Q_VLAN = 1<<0, 1236 IFF_EBRIDGE = 1<<1, 1237 IFF_SLAVE_INACTIVE = 1<<2, 1238 IFF_MASTER_8023AD = 1<<3, 1239 IFF_MASTER_ALB = 1<<4, 1240 IFF_BONDING = 1<<5, 1241 IFF_SLAVE_NEEDARP = 1<<6, 1242 IFF_ISATAP = 1<<7, 1243 IFF_MASTER_ARPMON = 1<<8, 1244 IFF_WAN_HDLC = 1<<9, 1245 IFF_XMIT_DST_RELEASE = 1<<10, 1246 IFF_DONT_BRIDGE = 1<<11, 1247 IFF_DISABLE_NETPOLL = 1<<12, 1248 IFF_MACVLAN_PORT = 1<<13, 1249 IFF_BRIDGE_PORT = 1<<14, 1250 IFF_OVS_DATAPATH = 1<<15, 1251 IFF_TX_SKB_SHARING = 1<<16, 1252 IFF_UNICAST_FLT = 1<<17, 1253 IFF_TEAM_PORT = 1<<18, 1254 IFF_SUPP_NOFCS = 1<<19, 1255 IFF_LIVE_ADDR_CHANGE = 1<<20, 1256 IFF_MACVLAN = 1<<21, 1257 IFF_XMIT_DST_RELEASE_PERM = 1<<22, 1258 IFF_IPVLAN_MASTER = 1<<23, 1259 IFF_IPVLAN_SLAVE = 1<<24, 1260 }; 1261 1262 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1263 #define IFF_EBRIDGE IFF_EBRIDGE 1264 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE 1265 #define IFF_MASTER_8023AD IFF_MASTER_8023AD 1266 #define IFF_MASTER_ALB IFF_MASTER_ALB 1267 #define IFF_BONDING IFF_BONDING 1268 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP 1269 #define IFF_ISATAP IFF_ISATAP 1270 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON 1271 #define IFF_WAN_HDLC IFF_WAN_HDLC 1272 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1273 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1274 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1275 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1276 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1277 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1278 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1279 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1280 #define IFF_TEAM_PORT IFF_TEAM_PORT 1281 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1282 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1283 #define IFF_MACVLAN IFF_MACVLAN 1284 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1285 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER 1286 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE 1287 1288 /** 1289 * struct net_device - The DEVICE structure. 1290 * Actually, this whole structure is a big mistake. It mixes I/O 1291 * data with strictly "high-level" data, and it has to know about 1292 * almost every data structure used in the INET module. 1293 * 1294 * @name: This is the first field of the "visible" part of this structure 1295 * (i.e. as seen by users in the "Space.c" file). It is the name 1296 * of the interface. 1297 * 1298 * @name_hlist: Device name hash chain, please keep it close to name[] 1299 * @ifalias: SNMP alias 1300 * @mem_end: Shared memory end 1301 * @mem_start: Shared memory start 1302 * @base_addr: Device I/O address 1303 * @irq: Device IRQ number 1304 * 1305 * @state: Generic network queuing layer state, see netdev_state_t 1306 * @dev_list: The global list of network devices 1307 * @napi_list: List entry, that is used for polling napi devices 1308 * @unreg_list: List entry, that is used, when we are unregistering the 1309 * device, see the function unregister_netdev 1310 * @close_list: List entry, that is used, when we are closing the device 1311 * 1312 * @adj_list: Directly linked devices, like slaves for bonding 1313 * @all_adj_list: All linked devices, *including* neighbours 1314 * @features: Currently active device features 1315 * @hw_features: User-changeable features 1316 * 1317 * @wanted_features: User-requested features 1318 * @vlan_features: Mask of features inheritable by VLAN devices 1319 * 1320 * @hw_enc_features: Mask of features inherited by encapsulating devices 1321 * This field indicates what encapsulation 1322 * offloads the hardware is capable of doing, 1323 * and drivers will need to set them appropriately. 1324 * 1325 * @mpls_features: Mask of features inheritable by MPLS 1326 * 1327 * @ifindex: interface index 1328 * @iflink: unique device identifier 1329 * 1330 * @stats: Statistics struct, which was left as a legacy, use 1331 * rtnl_link_stats64 instead 1332 * 1333 * @rx_dropped: Dropped packets by core network, 1334 * do not use this in drivers 1335 * @tx_dropped: Dropped packets by core network, 1336 * do not use this in drivers 1337 * 1338 * @carrier_changes: Stats to monitor carrier on<->off transitions 1339 * 1340 * @wireless_handlers: List of functions to handle Wireless Extensions, 1341 * instead of ioctl, 1342 * see <net/iw_handler.h> for details. 1343 * @wireless_data: Instance data managed by the core of wireless extensions 1344 * 1345 * @netdev_ops: Includes several pointers to callbacks, 1346 * if one wants to override the ndo_*() functions 1347 * @ethtool_ops: Management operations 1348 * @fwd_ops: Management operations 1349 * @header_ops: Includes callbacks for creating,parsing,rebuilding,etc 1350 * of Layer 2 headers. 1351 * 1352 * @flags: Interface flags (a la BSD) 1353 * @priv_flags: Like 'flags' but invisible to userspace, 1354 * see if.h for the definitions 1355 * @gflags: Global flags ( kept as legacy ) 1356 * @padded: How much padding added by alloc_netdev() 1357 * @operstate: RFC2863 operstate 1358 * @link_mode: Mapping policy to operstate 1359 * @if_port: Selectable AUI, TP, ... 1360 * @dma: DMA channel 1361 * @mtu: Interface MTU value 1362 * @type: Interface hardware type 1363 * @hard_header_len: Hardware header length 1364 * 1365 * @needed_headroom: Extra headroom the hardware may need, but not in all 1366 * cases can this be guaranteed 1367 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1368 * cases can this be guaranteed. Some cases also use 1369 * LL_MAX_HEADER instead to allocate the skb 1370 * 1371 * interface address info: 1372 * 1373 * @perm_addr: Permanent hw address 1374 * @addr_assign_type: Hw address assignment type 1375 * @addr_len: Hardware address length 1376 * @neigh_priv_len; Used in neigh_alloc(), 1377 * initialized only in atm/clip.c 1378 * @dev_id: Used to differentiate devices that share 1379 * the same link layer address 1380 * @dev_port: Used to differentiate devices that share 1381 * the same function 1382 * @addr_list_lock: XXX: need comments on this one 1383 * @uc: unicast mac addresses 1384 * @mc: multicast mac addresses 1385 * @dev_addrs: list of device hw addresses 1386 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1387 * @uc_promisc: Counter, that indicates, that promiscuous mode 1388 * has been enabled due to the need to listen to 1389 * additional unicast addresses in a device that 1390 * does not implement ndo_set_rx_mode() 1391 * @promiscuity: Number of times, the NIC is told to work in 1392 * Promiscuous mode, if it becomes 0 the NIC will 1393 * exit from working in Promiscuous mode 1394 * @allmulti: Counter, enables or disables allmulticast mode 1395 * 1396 * @vlan_info: VLAN info 1397 * @dsa_ptr: dsa specific data 1398 * @tipc_ptr: TIPC specific data 1399 * @atalk_ptr: AppleTalk link 1400 * @ip_ptr: IPv4 specific data 1401 * @dn_ptr: DECnet specific data 1402 * @ip6_ptr: IPv6 specific data 1403 * @ax25_ptr: AX.25 specific data 1404 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1405 * 1406 * @last_rx: Time of last Rx 1407 * @dev_addr: Hw address (before bcast, 1408 * because most packets are unicast) 1409 * 1410 * @_rx: Array of RX queues 1411 * @num_rx_queues: Number of RX queues 1412 * allocated at register_netdev() time 1413 * @real_num_rx_queues: Number of RX queues currently active in device 1414 * 1415 * @rx_handler: handler for received packets 1416 * @rx_handler_data: XXX: need comments on this one 1417 * @ingress_queue: XXX: need comments on this one 1418 * @broadcast: hw bcast address 1419 * 1420 * @_tx: Array of TX queues 1421 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1422 * @real_num_tx_queues: Number of TX queues currently active in device 1423 * @qdisc: Root qdisc from userspace point of view 1424 * @tx_queue_len: Max frames per queue allowed 1425 * @tx_global_lock: XXX: need comments on this one 1426 * 1427 * @xps_maps: XXX: need comments on this one 1428 * 1429 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1430 * indexed by RX queue number. Assigned by driver. 1431 * This must only be set if the ndo_rx_flow_steer 1432 * operation is defined 1433 * 1434 * @trans_start: Time (in jiffies) of last Tx 1435 * @watchdog_timeo: Represents the timeout that is used by 1436 * the watchdog ( see dev_watchdog() ) 1437 * @watchdog_timer: List of timers 1438 * 1439 * @pcpu_refcnt: Number of references to this device 1440 * @todo_list: Delayed register/unregister 1441 * @index_hlist: Device index hash chain 1442 * @link_watch_list: XXX: need comments on this one 1443 * 1444 * @reg_state: Register/unregister state machine 1445 * @dismantle: Device is going to be freed 1446 * @rtnl_link_state: This enum represents the phases of creating 1447 * a new link 1448 * 1449 * @destructor: Called from unregister, 1450 * can be used to call free_netdev 1451 * @npinfo: XXX: need comments on this one 1452 * @nd_net: Network namespace this network device is inside 1453 * 1454 * @ml_priv: Mid-layer private 1455 * @lstats: Loopback statistics 1456 * @tstats: Tunnel statistics 1457 * @dstats: Dummy statistics 1458 * @vstats: Virtual ethernet statistics 1459 * 1460 * @garp_port: GARP 1461 * @mrp_port: MRP 1462 * 1463 * @dev: Class/net/name entry 1464 * @sysfs_groups: Space for optional device, statistics and wireless 1465 * sysfs groups 1466 * 1467 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1468 * @rtnl_link_ops: Rtnl_link_ops 1469 * 1470 * @gso_max_size: Maximum size of generic segmentation offload 1471 * @gso_max_segs: Maximum number of segments that can be passed to the 1472 * NIC for GSO 1473 * @gso_min_segs: Minimum number of segments that can be passed to the 1474 * NIC for GSO 1475 * 1476 * @dcbnl_ops: Data Center Bridging netlink ops 1477 * @num_tc: Number of traffic classes in the net device 1478 * @tc_to_txq: XXX: need comments on this one 1479 * @prio_tc_map XXX: need comments on this one 1480 * 1481 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1482 * 1483 * @priomap: XXX: need comments on this one 1484 * @phydev: Physical device may attach itself 1485 * for hardware timestamping 1486 * 1487 * @qdisc_tx_busylock: XXX: need comments on this one 1488 * 1489 * @group: The group, that the device belongs to 1490 * @pm_qos_req: Power Management QoS object 1491 * 1492 * FIXME: cleanup struct net_device such that network protocol info 1493 * moves out. 1494 */ 1495 1496 struct net_device { 1497 char name[IFNAMSIZ]; 1498 struct hlist_node name_hlist; 1499 char *ifalias; 1500 /* 1501 * I/O specific fields 1502 * FIXME: Merge these and struct ifmap into one 1503 */ 1504 unsigned long mem_end; 1505 unsigned long mem_start; 1506 unsigned long base_addr; 1507 int irq; 1508 1509 /* 1510 * Some hardware also needs these fields (state,dev_list, 1511 * napi_list,unreg_list,close_list) but they are not 1512 * part of the usual set specified in Space.c. 1513 */ 1514 1515 unsigned long state; 1516 1517 struct list_head dev_list; 1518 struct list_head napi_list; 1519 struct list_head unreg_list; 1520 struct list_head close_list; 1521 struct list_head ptype_all; 1522 struct list_head ptype_specific; 1523 1524 struct { 1525 struct list_head upper; 1526 struct list_head lower; 1527 } adj_list; 1528 1529 struct { 1530 struct list_head upper; 1531 struct list_head lower; 1532 } all_adj_list; 1533 1534 netdev_features_t features; 1535 netdev_features_t hw_features; 1536 netdev_features_t wanted_features; 1537 netdev_features_t vlan_features; 1538 netdev_features_t hw_enc_features; 1539 netdev_features_t mpls_features; 1540 1541 int ifindex; 1542 int iflink; 1543 1544 struct net_device_stats stats; 1545 1546 atomic_long_t rx_dropped; 1547 atomic_long_t tx_dropped; 1548 1549 atomic_t carrier_changes; 1550 1551 #ifdef CONFIG_WIRELESS_EXT 1552 const struct iw_handler_def * wireless_handlers; 1553 struct iw_public_data * wireless_data; 1554 #endif 1555 const struct net_device_ops *netdev_ops; 1556 const struct ethtool_ops *ethtool_ops; 1557 const struct forwarding_accel_ops *fwd_ops; 1558 1559 const struct header_ops *header_ops; 1560 1561 unsigned int flags; 1562 unsigned int priv_flags; 1563 1564 unsigned short gflags; 1565 unsigned short padded; 1566 1567 unsigned char operstate; 1568 unsigned char link_mode; 1569 1570 unsigned char if_port; 1571 unsigned char dma; 1572 1573 unsigned int mtu; 1574 unsigned short type; 1575 unsigned short hard_header_len; 1576 1577 unsigned short needed_headroom; 1578 unsigned short needed_tailroom; 1579 1580 /* Interface address info. */ 1581 unsigned char perm_addr[MAX_ADDR_LEN]; 1582 unsigned char addr_assign_type; 1583 unsigned char addr_len; 1584 unsigned short neigh_priv_len; 1585 unsigned short dev_id; 1586 unsigned short dev_port; 1587 spinlock_t addr_list_lock; 1588 struct netdev_hw_addr_list uc; 1589 struct netdev_hw_addr_list mc; 1590 struct netdev_hw_addr_list dev_addrs; 1591 1592 #ifdef CONFIG_SYSFS 1593 struct kset *queues_kset; 1594 #endif 1595 1596 unsigned char name_assign_type; 1597 1598 bool uc_promisc; 1599 unsigned int promiscuity; 1600 unsigned int allmulti; 1601 1602 1603 /* Protocol specific pointers */ 1604 1605 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1606 struct vlan_info __rcu *vlan_info; 1607 #endif 1608 #if IS_ENABLED(CONFIG_NET_DSA) 1609 struct dsa_switch_tree *dsa_ptr; 1610 #endif 1611 #if IS_ENABLED(CONFIG_TIPC) 1612 struct tipc_bearer __rcu *tipc_ptr; 1613 #endif 1614 void *atalk_ptr; 1615 struct in_device __rcu *ip_ptr; 1616 struct dn_dev __rcu *dn_ptr; 1617 struct inet6_dev __rcu *ip6_ptr; 1618 void *ax25_ptr; 1619 struct wireless_dev *ieee80211_ptr; 1620 struct wpan_dev *ieee802154_ptr; 1621 1622 /* 1623 * Cache lines mostly used on receive path (including eth_type_trans()) 1624 */ 1625 unsigned long last_rx; 1626 1627 /* Interface address info used in eth_type_trans() */ 1628 unsigned char *dev_addr; 1629 1630 1631 #ifdef CONFIG_SYSFS 1632 struct netdev_rx_queue *_rx; 1633 1634 unsigned int num_rx_queues; 1635 unsigned int real_num_rx_queues; 1636 1637 #endif 1638 1639 unsigned long gro_flush_timeout; 1640 rx_handler_func_t __rcu *rx_handler; 1641 void __rcu *rx_handler_data; 1642 1643 struct netdev_queue __rcu *ingress_queue; 1644 unsigned char broadcast[MAX_ADDR_LEN]; 1645 1646 1647 /* 1648 * Cache lines mostly used on transmit path 1649 */ 1650 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1651 unsigned int num_tx_queues; 1652 unsigned int real_num_tx_queues; 1653 struct Qdisc *qdisc; 1654 unsigned long tx_queue_len; 1655 spinlock_t tx_global_lock; 1656 1657 #ifdef CONFIG_XPS 1658 struct xps_dev_maps __rcu *xps_maps; 1659 #endif 1660 #ifdef CONFIG_RFS_ACCEL 1661 struct cpu_rmap *rx_cpu_rmap; 1662 #endif 1663 1664 /* These may be needed for future network-power-down code. */ 1665 1666 /* 1667 * trans_start here is expensive for high speed devices on SMP, 1668 * please use netdev_queue->trans_start instead. 1669 */ 1670 unsigned long trans_start; 1671 1672 int watchdog_timeo; 1673 struct timer_list watchdog_timer; 1674 1675 int __percpu *pcpu_refcnt; 1676 struct list_head todo_list; 1677 1678 struct hlist_node index_hlist; 1679 struct list_head link_watch_list; 1680 1681 enum { NETREG_UNINITIALIZED=0, 1682 NETREG_REGISTERED, /* completed register_netdevice */ 1683 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1684 NETREG_UNREGISTERED, /* completed unregister todo */ 1685 NETREG_RELEASED, /* called free_netdev */ 1686 NETREG_DUMMY, /* dummy device for NAPI poll */ 1687 } reg_state:8; 1688 1689 bool dismantle; 1690 1691 enum { 1692 RTNL_LINK_INITIALIZED, 1693 RTNL_LINK_INITIALIZING, 1694 } rtnl_link_state:16; 1695 1696 void (*destructor)(struct net_device *dev); 1697 1698 #ifdef CONFIG_NETPOLL 1699 struct netpoll_info __rcu *npinfo; 1700 #endif 1701 1702 #ifdef CONFIG_NET_NS 1703 struct net *nd_net; 1704 #endif 1705 1706 /* mid-layer private */ 1707 union { 1708 void *ml_priv; 1709 struct pcpu_lstats __percpu *lstats; 1710 struct pcpu_sw_netstats __percpu *tstats; 1711 struct pcpu_dstats __percpu *dstats; 1712 struct pcpu_vstats __percpu *vstats; 1713 }; 1714 1715 struct garp_port __rcu *garp_port; 1716 struct mrp_port __rcu *mrp_port; 1717 1718 struct device dev; 1719 const struct attribute_group *sysfs_groups[4]; 1720 const struct attribute_group *sysfs_rx_queue_group; 1721 1722 const struct rtnl_link_ops *rtnl_link_ops; 1723 1724 /* for setting kernel sock attribute on TCP connection setup */ 1725 #define GSO_MAX_SIZE 65536 1726 unsigned int gso_max_size; 1727 #define GSO_MAX_SEGS 65535 1728 u16 gso_max_segs; 1729 u16 gso_min_segs; 1730 #ifdef CONFIG_DCB 1731 const struct dcbnl_rtnl_ops *dcbnl_ops; 1732 #endif 1733 u8 num_tc; 1734 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 1735 u8 prio_tc_map[TC_BITMASK + 1]; 1736 1737 #if IS_ENABLED(CONFIG_FCOE) 1738 unsigned int fcoe_ddp_xid; 1739 #endif 1740 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 1741 struct netprio_map __rcu *priomap; 1742 #endif 1743 struct phy_device *phydev; 1744 struct lock_class_key *qdisc_tx_busylock; 1745 int group; 1746 struct pm_qos_request pm_qos_req; 1747 }; 1748 #define to_net_dev(d) container_of(d, struct net_device, dev) 1749 1750 #define NETDEV_ALIGN 32 1751 1752 static inline 1753 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 1754 { 1755 return dev->prio_tc_map[prio & TC_BITMASK]; 1756 } 1757 1758 static inline 1759 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 1760 { 1761 if (tc >= dev->num_tc) 1762 return -EINVAL; 1763 1764 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 1765 return 0; 1766 } 1767 1768 static inline 1769 void netdev_reset_tc(struct net_device *dev) 1770 { 1771 dev->num_tc = 0; 1772 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq)); 1773 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map)); 1774 } 1775 1776 static inline 1777 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset) 1778 { 1779 if (tc >= dev->num_tc) 1780 return -EINVAL; 1781 1782 dev->tc_to_txq[tc].count = count; 1783 dev->tc_to_txq[tc].offset = offset; 1784 return 0; 1785 } 1786 1787 static inline 1788 int netdev_set_num_tc(struct net_device *dev, u8 num_tc) 1789 { 1790 if (num_tc > TC_MAX_QUEUE) 1791 return -EINVAL; 1792 1793 dev->num_tc = num_tc; 1794 return 0; 1795 } 1796 1797 static inline 1798 int netdev_get_num_tc(struct net_device *dev) 1799 { 1800 return dev->num_tc; 1801 } 1802 1803 static inline 1804 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 1805 unsigned int index) 1806 { 1807 return &dev->_tx[index]; 1808 } 1809 1810 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 1811 const struct sk_buff *skb) 1812 { 1813 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 1814 } 1815 1816 static inline void netdev_for_each_tx_queue(struct net_device *dev, 1817 void (*f)(struct net_device *, 1818 struct netdev_queue *, 1819 void *), 1820 void *arg) 1821 { 1822 unsigned int i; 1823 1824 for (i = 0; i < dev->num_tx_queues; i++) 1825 f(dev, &dev->_tx[i], arg); 1826 } 1827 1828 struct netdev_queue *netdev_pick_tx(struct net_device *dev, 1829 struct sk_buff *skb, 1830 void *accel_priv); 1831 1832 /* 1833 * Net namespace inlines 1834 */ 1835 static inline 1836 struct net *dev_net(const struct net_device *dev) 1837 { 1838 return read_pnet(&dev->nd_net); 1839 } 1840 1841 static inline 1842 void dev_net_set(struct net_device *dev, struct net *net) 1843 { 1844 #ifdef CONFIG_NET_NS 1845 release_net(dev->nd_net); 1846 dev->nd_net = hold_net(net); 1847 #endif 1848 } 1849 1850 static inline bool netdev_uses_dsa(struct net_device *dev) 1851 { 1852 #if IS_ENABLED(CONFIG_NET_DSA) 1853 if (dev->dsa_ptr != NULL) 1854 return dsa_uses_tagged_protocol(dev->dsa_ptr); 1855 #endif 1856 return false; 1857 } 1858 1859 /** 1860 * netdev_priv - access network device private data 1861 * @dev: network device 1862 * 1863 * Get network device private data 1864 */ 1865 static inline void *netdev_priv(const struct net_device *dev) 1866 { 1867 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 1868 } 1869 1870 /* Set the sysfs physical device reference for the network logical device 1871 * if set prior to registration will cause a symlink during initialization. 1872 */ 1873 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 1874 1875 /* Set the sysfs device type for the network logical device to allow 1876 * fine-grained identification of different network device types. For 1877 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc. 1878 */ 1879 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 1880 1881 /* Default NAPI poll() weight 1882 * Device drivers are strongly advised to not use bigger value 1883 */ 1884 #define NAPI_POLL_WEIGHT 64 1885 1886 /** 1887 * netif_napi_add - initialize a napi context 1888 * @dev: network device 1889 * @napi: napi context 1890 * @poll: polling function 1891 * @weight: default weight 1892 * 1893 * netif_napi_add() must be used to initialize a napi context prior to calling 1894 * *any* of the other napi related functions. 1895 */ 1896 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 1897 int (*poll)(struct napi_struct *, int), int weight); 1898 1899 /** 1900 * netif_napi_del - remove a napi context 1901 * @napi: napi context 1902 * 1903 * netif_napi_del() removes a napi context from the network device napi list 1904 */ 1905 void netif_napi_del(struct napi_struct *napi); 1906 1907 struct napi_gro_cb { 1908 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 1909 void *frag0; 1910 1911 /* Length of frag0. */ 1912 unsigned int frag0_len; 1913 1914 /* This indicates where we are processing relative to skb->data. */ 1915 int data_offset; 1916 1917 /* This is non-zero if the packet cannot be merged with the new skb. */ 1918 u16 flush; 1919 1920 /* Save the IP ID here and check when we get to the transport layer */ 1921 u16 flush_id; 1922 1923 /* Number of segments aggregated. */ 1924 u16 count; 1925 1926 /* Start offset for remote checksum offload */ 1927 u16 gro_remcsum_start; 1928 1929 /* jiffies when first packet was created/queued */ 1930 unsigned long age; 1931 1932 /* Used in ipv6_gro_receive() and foo-over-udp */ 1933 u16 proto; 1934 1935 /* This is non-zero if the packet may be of the same flow. */ 1936 u8 same_flow:1; 1937 1938 /* Used in udp_gro_receive */ 1939 u8 udp_mark:1; 1940 1941 /* GRO checksum is valid */ 1942 u8 csum_valid:1; 1943 1944 /* Number of checksums via CHECKSUM_UNNECESSARY */ 1945 u8 csum_cnt:3; 1946 1947 /* Free the skb? */ 1948 u8 free:2; 1949 #define NAPI_GRO_FREE 1 1950 #define NAPI_GRO_FREE_STOLEN_HEAD 2 1951 1952 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 1953 u8 is_ipv6:1; 1954 1955 /* 7 bit hole */ 1956 1957 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 1958 __wsum csum; 1959 1960 /* used in skb_gro_receive() slow path */ 1961 struct sk_buff *last; 1962 }; 1963 1964 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 1965 1966 struct packet_type { 1967 __be16 type; /* This is really htons(ether_type). */ 1968 struct net_device *dev; /* NULL is wildcarded here */ 1969 int (*func) (struct sk_buff *, 1970 struct net_device *, 1971 struct packet_type *, 1972 struct net_device *); 1973 bool (*id_match)(struct packet_type *ptype, 1974 struct sock *sk); 1975 void *af_packet_priv; 1976 struct list_head list; 1977 }; 1978 1979 struct offload_callbacks { 1980 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 1981 netdev_features_t features); 1982 struct sk_buff **(*gro_receive)(struct sk_buff **head, 1983 struct sk_buff *skb); 1984 int (*gro_complete)(struct sk_buff *skb, int nhoff); 1985 }; 1986 1987 struct packet_offload { 1988 __be16 type; /* This is really htons(ether_type). */ 1989 struct offload_callbacks callbacks; 1990 struct list_head list; 1991 }; 1992 1993 struct udp_offload; 1994 1995 struct udp_offload_callbacks { 1996 struct sk_buff **(*gro_receive)(struct sk_buff **head, 1997 struct sk_buff *skb, 1998 struct udp_offload *uoff); 1999 int (*gro_complete)(struct sk_buff *skb, 2000 int nhoff, 2001 struct udp_offload *uoff); 2002 }; 2003 2004 struct udp_offload { 2005 __be16 port; 2006 u8 ipproto; 2007 struct udp_offload_callbacks callbacks; 2008 }; 2009 2010 /* often modified stats are per cpu, other are shared (netdev->stats) */ 2011 struct pcpu_sw_netstats { 2012 u64 rx_packets; 2013 u64 rx_bytes; 2014 u64 tx_packets; 2015 u64 tx_bytes; 2016 struct u64_stats_sync syncp; 2017 }; 2018 2019 #define netdev_alloc_pcpu_stats(type) \ 2020 ({ \ 2021 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \ 2022 if (pcpu_stats) { \ 2023 int i; \ 2024 for_each_possible_cpu(i) { \ 2025 typeof(type) *stat; \ 2026 stat = per_cpu_ptr(pcpu_stats, i); \ 2027 u64_stats_init(&stat->syncp); \ 2028 } \ 2029 } \ 2030 pcpu_stats; \ 2031 }) 2032 2033 #include <linux/notifier.h> 2034 2035 /* netdevice notifier chain. Please remember to update the rtnetlink 2036 * notification exclusion list in rtnetlink_event() when adding new 2037 * types. 2038 */ 2039 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */ 2040 #define NETDEV_DOWN 0x0002 2041 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface 2042 detected a hardware crash and restarted 2043 - we can use this eg to kick tcp sessions 2044 once done */ 2045 #define NETDEV_CHANGE 0x0004 /* Notify device state change */ 2046 #define NETDEV_REGISTER 0x0005 2047 #define NETDEV_UNREGISTER 0x0006 2048 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */ 2049 #define NETDEV_CHANGEADDR 0x0008 2050 #define NETDEV_GOING_DOWN 0x0009 2051 #define NETDEV_CHANGENAME 0x000A 2052 #define NETDEV_FEAT_CHANGE 0x000B 2053 #define NETDEV_BONDING_FAILOVER 0x000C 2054 #define NETDEV_PRE_UP 0x000D 2055 #define NETDEV_PRE_TYPE_CHANGE 0x000E 2056 #define NETDEV_POST_TYPE_CHANGE 0x000F 2057 #define NETDEV_POST_INIT 0x0010 2058 #define NETDEV_UNREGISTER_FINAL 0x0011 2059 #define NETDEV_RELEASE 0x0012 2060 #define NETDEV_NOTIFY_PEERS 0x0013 2061 #define NETDEV_JOIN 0x0014 2062 #define NETDEV_CHANGEUPPER 0x0015 2063 #define NETDEV_RESEND_IGMP 0x0016 2064 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */ 2065 #define NETDEV_CHANGEINFODATA 0x0018 2066 #define NETDEV_BONDING_INFO 0x0019 2067 2068 int register_netdevice_notifier(struct notifier_block *nb); 2069 int unregister_netdevice_notifier(struct notifier_block *nb); 2070 2071 struct netdev_notifier_info { 2072 struct net_device *dev; 2073 }; 2074 2075 struct netdev_notifier_change_info { 2076 struct netdev_notifier_info info; /* must be first */ 2077 unsigned int flags_changed; 2078 }; 2079 2080 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2081 struct net_device *dev) 2082 { 2083 info->dev = dev; 2084 } 2085 2086 static inline struct net_device * 2087 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2088 { 2089 return info->dev; 2090 } 2091 2092 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2093 2094 2095 extern rwlock_t dev_base_lock; /* Device list lock */ 2096 2097 #define for_each_netdev(net, d) \ 2098 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2099 #define for_each_netdev_reverse(net, d) \ 2100 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2101 #define for_each_netdev_rcu(net, d) \ 2102 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2103 #define for_each_netdev_safe(net, d, n) \ 2104 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2105 #define for_each_netdev_continue(net, d) \ 2106 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2107 #define for_each_netdev_continue_rcu(net, d) \ 2108 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2109 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2110 for_each_netdev_rcu(&init_net, slave) \ 2111 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2112 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2113 2114 static inline struct net_device *next_net_device(struct net_device *dev) 2115 { 2116 struct list_head *lh; 2117 struct net *net; 2118 2119 net = dev_net(dev); 2120 lh = dev->dev_list.next; 2121 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2122 } 2123 2124 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2125 { 2126 struct list_head *lh; 2127 struct net *net; 2128 2129 net = dev_net(dev); 2130 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2131 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2132 } 2133 2134 static inline struct net_device *first_net_device(struct net *net) 2135 { 2136 return list_empty(&net->dev_base_head) ? NULL : 2137 net_device_entry(net->dev_base_head.next); 2138 } 2139 2140 static inline struct net_device *first_net_device_rcu(struct net *net) 2141 { 2142 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2143 2144 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2145 } 2146 2147 int netdev_boot_setup_check(struct net_device *dev); 2148 unsigned long netdev_boot_base(const char *prefix, int unit); 2149 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2150 const char *hwaddr); 2151 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2152 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2153 void dev_add_pack(struct packet_type *pt); 2154 void dev_remove_pack(struct packet_type *pt); 2155 void __dev_remove_pack(struct packet_type *pt); 2156 void dev_add_offload(struct packet_offload *po); 2157 void dev_remove_offload(struct packet_offload *po); 2158 2159 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2160 unsigned short mask); 2161 struct net_device *dev_get_by_name(struct net *net, const char *name); 2162 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2163 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2164 int dev_alloc_name(struct net_device *dev, const char *name); 2165 int dev_open(struct net_device *dev); 2166 int dev_close(struct net_device *dev); 2167 void dev_disable_lro(struct net_device *dev); 2168 int dev_loopback_xmit(struct sk_buff *newskb); 2169 int dev_queue_xmit(struct sk_buff *skb); 2170 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv); 2171 int register_netdevice(struct net_device *dev); 2172 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2173 void unregister_netdevice_many(struct list_head *head); 2174 static inline void unregister_netdevice(struct net_device *dev) 2175 { 2176 unregister_netdevice_queue(dev, NULL); 2177 } 2178 2179 int netdev_refcnt_read(const struct net_device *dev); 2180 void free_netdev(struct net_device *dev); 2181 void netdev_freemem(struct net_device *dev); 2182 void synchronize_net(void); 2183 int init_dummy_netdev(struct net_device *dev); 2184 2185 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2186 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2187 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2188 int netdev_get_name(struct net *net, char *name, int ifindex); 2189 int dev_restart(struct net_device *dev); 2190 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb); 2191 2192 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2193 { 2194 return NAPI_GRO_CB(skb)->data_offset; 2195 } 2196 2197 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2198 { 2199 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2200 } 2201 2202 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2203 { 2204 NAPI_GRO_CB(skb)->data_offset += len; 2205 } 2206 2207 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2208 unsigned int offset) 2209 { 2210 return NAPI_GRO_CB(skb)->frag0 + offset; 2211 } 2212 2213 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2214 { 2215 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2216 } 2217 2218 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2219 unsigned int offset) 2220 { 2221 if (!pskb_may_pull(skb, hlen)) 2222 return NULL; 2223 2224 NAPI_GRO_CB(skb)->frag0 = NULL; 2225 NAPI_GRO_CB(skb)->frag0_len = 0; 2226 return skb->data + offset; 2227 } 2228 2229 static inline void *skb_gro_network_header(struct sk_buff *skb) 2230 { 2231 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2232 skb_network_offset(skb); 2233 } 2234 2235 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2236 const void *start, unsigned int len) 2237 { 2238 if (NAPI_GRO_CB(skb)->csum_valid) 2239 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2240 csum_partial(start, len, 0)); 2241 } 2242 2243 /* GRO checksum functions. These are logical equivalents of the normal 2244 * checksum functions (in skbuff.h) except that they operate on the GRO 2245 * offsets and fields in sk_buff. 2246 */ 2247 2248 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2249 2250 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2251 { 2252 return (NAPI_GRO_CB(skb)->gro_remcsum_start - skb_headroom(skb) == 2253 skb_gro_offset(skb)); 2254 } 2255 2256 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2257 bool zero_okay, 2258 __sum16 check) 2259 { 2260 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2261 skb_checksum_start_offset(skb) < 2262 skb_gro_offset(skb)) && 2263 !skb_at_gro_remcsum_start(skb) && 2264 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2265 (!zero_okay || check)); 2266 } 2267 2268 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2269 __wsum psum) 2270 { 2271 if (NAPI_GRO_CB(skb)->csum_valid && 2272 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2273 return 0; 2274 2275 NAPI_GRO_CB(skb)->csum = psum; 2276 2277 return __skb_gro_checksum_complete(skb); 2278 } 2279 2280 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2281 { 2282 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2283 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2284 NAPI_GRO_CB(skb)->csum_cnt--; 2285 } else { 2286 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2287 * verified a new top level checksum or an encapsulated one 2288 * during GRO. This saves work if we fallback to normal path. 2289 */ 2290 __skb_incr_checksum_unnecessary(skb); 2291 } 2292 } 2293 2294 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2295 compute_pseudo) \ 2296 ({ \ 2297 __sum16 __ret = 0; \ 2298 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2299 __ret = __skb_gro_checksum_validate_complete(skb, \ 2300 compute_pseudo(skb, proto)); \ 2301 if (__ret) \ 2302 __skb_mark_checksum_bad(skb); \ 2303 else \ 2304 skb_gro_incr_csum_unnecessary(skb); \ 2305 __ret; \ 2306 }) 2307 2308 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2309 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2310 2311 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2312 compute_pseudo) \ 2313 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2314 2315 #define skb_gro_checksum_simple_validate(skb) \ 2316 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2317 2318 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2319 { 2320 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2321 !NAPI_GRO_CB(skb)->csum_valid); 2322 } 2323 2324 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2325 __sum16 check, __wsum pseudo) 2326 { 2327 NAPI_GRO_CB(skb)->csum = ~pseudo; 2328 NAPI_GRO_CB(skb)->csum_valid = 1; 2329 } 2330 2331 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \ 2332 do { \ 2333 if (__skb_gro_checksum_convert_check(skb)) \ 2334 __skb_gro_checksum_convert(skb, check, \ 2335 compute_pseudo(skb, proto)); \ 2336 } while (0) 2337 2338 struct gro_remcsum { 2339 int offset; 2340 __wsum delta; 2341 }; 2342 2343 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2344 { 2345 grc->delta = 0; 2346 } 2347 2348 static inline void skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2349 int start, int offset, 2350 struct gro_remcsum *grc, 2351 bool nopartial) 2352 { 2353 __wsum delta; 2354 2355 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2356 2357 if (!nopartial) { 2358 NAPI_GRO_CB(skb)->gro_remcsum_start = 2359 ((unsigned char *)ptr + start) - skb->head; 2360 return; 2361 } 2362 2363 delta = remcsum_adjust(ptr, NAPI_GRO_CB(skb)->csum, start, offset); 2364 2365 /* Adjust skb->csum since we changed the packet */ 2366 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2367 2368 grc->offset = (ptr + offset) - (void *)skb->head; 2369 grc->delta = delta; 2370 } 2371 2372 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2373 struct gro_remcsum *grc) 2374 { 2375 if (!grc->delta) 2376 return; 2377 2378 remcsum_unadjust((__sum16 *)(skb->head + grc->offset), grc->delta); 2379 } 2380 2381 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2382 unsigned short type, 2383 const void *daddr, const void *saddr, 2384 unsigned int len) 2385 { 2386 if (!dev->header_ops || !dev->header_ops->create) 2387 return 0; 2388 2389 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2390 } 2391 2392 static inline int dev_parse_header(const struct sk_buff *skb, 2393 unsigned char *haddr) 2394 { 2395 const struct net_device *dev = skb->dev; 2396 2397 if (!dev->header_ops || !dev->header_ops->parse) 2398 return 0; 2399 return dev->header_ops->parse(skb, haddr); 2400 } 2401 2402 static inline int dev_rebuild_header(struct sk_buff *skb) 2403 { 2404 const struct net_device *dev = skb->dev; 2405 2406 if (!dev->header_ops || !dev->header_ops->rebuild) 2407 return 0; 2408 return dev->header_ops->rebuild(skb); 2409 } 2410 2411 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len); 2412 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2413 static inline int unregister_gifconf(unsigned int family) 2414 { 2415 return register_gifconf(family, NULL); 2416 } 2417 2418 #ifdef CONFIG_NET_FLOW_LIMIT 2419 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2420 struct sd_flow_limit { 2421 u64 count; 2422 unsigned int num_buckets; 2423 unsigned int history_head; 2424 u16 history[FLOW_LIMIT_HISTORY]; 2425 u8 buckets[]; 2426 }; 2427 2428 extern int netdev_flow_limit_table_len; 2429 #endif /* CONFIG_NET_FLOW_LIMIT */ 2430 2431 /* 2432 * Incoming packets are placed on per-cpu queues 2433 */ 2434 struct softnet_data { 2435 struct list_head poll_list; 2436 struct sk_buff_head process_queue; 2437 2438 /* stats */ 2439 unsigned int processed; 2440 unsigned int time_squeeze; 2441 unsigned int cpu_collision; 2442 unsigned int received_rps; 2443 #ifdef CONFIG_RPS 2444 struct softnet_data *rps_ipi_list; 2445 #endif 2446 #ifdef CONFIG_NET_FLOW_LIMIT 2447 struct sd_flow_limit __rcu *flow_limit; 2448 #endif 2449 struct Qdisc *output_queue; 2450 struct Qdisc **output_queue_tailp; 2451 struct sk_buff *completion_queue; 2452 2453 #ifdef CONFIG_RPS 2454 /* Elements below can be accessed between CPUs for RPS */ 2455 struct call_single_data csd ____cacheline_aligned_in_smp; 2456 struct softnet_data *rps_ipi_next; 2457 unsigned int cpu; 2458 unsigned int input_queue_head; 2459 unsigned int input_queue_tail; 2460 #endif 2461 unsigned int dropped; 2462 struct sk_buff_head input_pkt_queue; 2463 struct napi_struct backlog; 2464 2465 }; 2466 2467 static inline void input_queue_head_incr(struct softnet_data *sd) 2468 { 2469 #ifdef CONFIG_RPS 2470 sd->input_queue_head++; 2471 #endif 2472 } 2473 2474 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 2475 unsigned int *qtail) 2476 { 2477 #ifdef CONFIG_RPS 2478 *qtail = ++sd->input_queue_tail; 2479 #endif 2480 } 2481 2482 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 2483 2484 void __netif_schedule(struct Qdisc *q); 2485 void netif_schedule_queue(struct netdev_queue *txq); 2486 2487 static inline void netif_tx_schedule_all(struct net_device *dev) 2488 { 2489 unsigned int i; 2490 2491 for (i = 0; i < dev->num_tx_queues; i++) 2492 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 2493 } 2494 2495 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 2496 { 2497 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2498 } 2499 2500 /** 2501 * netif_start_queue - allow transmit 2502 * @dev: network device 2503 * 2504 * Allow upper layers to call the device hard_start_xmit routine. 2505 */ 2506 static inline void netif_start_queue(struct net_device *dev) 2507 { 2508 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 2509 } 2510 2511 static inline void netif_tx_start_all_queues(struct net_device *dev) 2512 { 2513 unsigned int i; 2514 2515 for (i = 0; i < dev->num_tx_queues; i++) { 2516 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2517 netif_tx_start_queue(txq); 2518 } 2519 } 2520 2521 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 2522 2523 /** 2524 * netif_wake_queue - restart transmit 2525 * @dev: network device 2526 * 2527 * Allow upper layers to call the device hard_start_xmit routine. 2528 * Used for flow control when transmit resources are available. 2529 */ 2530 static inline void netif_wake_queue(struct net_device *dev) 2531 { 2532 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 2533 } 2534 2535 static inline void netif_tx_wake_all_queues(struct net_device *dev) 2536 { 2537 unsigned int i; 2538 2539 for (i = 0; i < dev->num_tx_queues; i++) { 2540 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2541 netif_tx_wake_queue(txq); 2542 } 2543 } 2544 2545 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 2546 { 2547 if (WARN_ON(!dev_queue)) { 2548 pr_info("netif_stop_queue() cannot be called before register_netdev()\n"); 2549 return; 2550 } 2551 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2552 } 2553 2554 /** 2555 * netif_stop_queue - stop transmitted packets 2556 * @dev: network device 2557 * 2558 * Stop upper layers calling the device hard_start_xmit routine. 2559 * Used for flow control when transmit resources are unavailable. 2560 */ 2561 static inline void netif_stop_queue(struct net_device *dev) 2562 { 2563 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 2564 } 2565 2566 static inline void netif_tx_stop_all_queues(struct net_device *dev) 2567 { 2568 unsigned int i; 2569 2570 for (i = 0; i < dev->num_tx_queues; i++) { 2571 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2572 netif_tx_stop_queue(txq); 2573 } 2574 } 2575 2576 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 2577 { 2578 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2579 } 2580 2581 /** 2582 * netif_queue_stopped - test if transmit queue is flowblocked 2583 * @dev: network device 2584 * 2585 * Test if transmit queue on device is currently unable to send. 2586 */ 2587 static inline bool netif_queue_stopped(const struct net_device *dev) 2588 { 2589 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 2590 } 2591 2592 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 2593 { 2594 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 2595 } 2596 2597 static inline bool 2598 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 2599 { 2600 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 2601 } 2602 2603 static inline bool 2604 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 2605 { 2606 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 2607 } 2608 2609 /** 2610 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 2611 * @dev_queue: pointer to transmit queue 2612 * 2613 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 2614 * to give appropriate hint to the cpu. 2615 */ 2616 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 2617 { 2618 #ifdef CONFIG_BQL 2619 prefetchw(&dev_queue->dql.num_queued); 2620 #endif 2621 } 2622 2623 /** 2624 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 2625 * @dev_queue: pointer to transmit queue 2626 * 2627 * BQL enabled drivers might use this helper in their TX completion path, 2628 * to give appropriate hint to the cpu. 2629 */ 2630 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 2631 { 2632 #ifdef CONFIG_BQL 2633 prefetchw(&dev_queue->dql.limit); 2634 #endif 2635 } 2636 2637 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 2638 unsigned int bytes) 2639 { 2640 #ifdef CONFIG_BQL 2641 dql_queued(&dev_queue->dql, bytes); 2642 2643 if (likely(dql_avail(&dev_queue->dql) >= 0)) 2644 return; 2645 2646 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2647 2648 /* 2649 * The XOFF flag must be set before checking the dql_avail below, 2650 * because in netdev_tx_completed_queue we update the dql_completed 2651 * before checking the XOFF flag. 2652 */ 2653 smp_mb(); 2654 2655 /* check again in case another CPU has just made room avail */ 2656 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 2657 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2658 #endif 2659 } 2660 2661 /** 2662 * netdev_sent_queue - report the number of bytes queued to hardware 2663 * @dev: network device 2664 * @bytes: number of bytes queued to the hardware device queue 2665 * 2666 * Report the number of bytes queued for sending/completion to the network 2667 * device hardware queue. @bytes should be a good approximation and should 2668 * exactly match netdev_completed_queue() @bytes 2669 */ 2670 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 2671 { 2672 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 2673 } 2674 2675 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 2676 unsigned int pkts, unsigned int bytes) 2677 { 2678 #ifdef CONFIG_BQL 2679 if (unlikely(!bytes)) 2680 return; 2681 2682 dql_completed(&dev_queue->dql, bytes); 2683 2684 /* 2685 * Without the memory barrier there is a small possiblity that 2686 * netdev_tx_sent_queue will miss the update and cause the queue to 2687 * be stopped forever 2688 */ 2689 smp_mb(); 2690 2691 if (dql_avail(&dev_queue->dql) < 0) 2692 return; 2693 2694 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 2695 netif_schedule_queue(dev_queue); 2696 #endif 2697 } 2698 2699 /** 2700 * netdev_completed_queue - report bytes and packets completed by device 2701 * @dev: network device 2702 * @pkts: actual number of packets sent over the medium 2703 * @bytes: actual number of bytes sent over the medium 2704 * 2705 * Report the number of bytes and packets transmitted by the network device 2706 * hardware queue over the physical medium, @bytes must exactly match the 2707 * @bytes amount passed to netdev_sent_queue() 2708 */ 2709 static inline void netdev_completed_queue(struct net_device *dev, 2710 unsigned int pkts, unsigned int bytes) 2711 { 2712 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 2713 } 2714 2715 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 2716 { 2717 #ifdef CONFIG_BQL 2718 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 2719 dql_reset(&q->dql); 2720 #endif 2721 } 2722 2723 /** 2724 * netdev_reset_queue - reset the packets and bytes count of a network device 2725 * @dev_queue: network device 2726 * 2727 * Reset the bytes and packet count of a network device and clear the 2728 * software flow control OFF bit for this network device 2729 */ 2730 static inline void netdev_reset_queue(struct net_device *dev_queue) 2731 { 2732 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 2733 } 2734 2735 /** 2736 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 2737 * @dev: network device 2738 * @queue_index: given tx queue index 2739 * 2740 * Returns 0 if given tx queue index >= number of device tx queues, 2741 * otherwise returns the originally passed tx queue index. 2742 */ 2743 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 2744 { 2745 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 2746 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 2747 dev->name, queue_index, 2748 dev->real_num_tx_queues); 2749 return 0; 2750 } 2751 2752 return queue_index; 2753 } 2754 2755 /** 2756 * netif_running - test if up 2757 * @dev: network device 2758 * 2759 * Test if the device has been brought up. 2760 */ 2761 static inline bool netif_running(const struct net_device *dev) 2762 { 2763 return test_bit(__LINK_STATE_START, &dev->state); 2764 } 2765 2766 /* 2767 * Routines to manage the subqueues on a device. We only need start 2768 * stop, and a check if it's stopped. All other device management is 2769 * done at the overall netdevice level. 2770 * Also test the device if we're multiqueue. 2771 */ 2772 2773 /** 2774 * netif_start_subqueue - allow sending packets on subqueue 2775 * @dev: network device 2776 * @queue_index: sub queue index 2777 * 2778 * Start individual transmit queue of a device with multiple transmit queues. 2779 */ 2780 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 2781 { 2782 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2783 2784 netif_tx_start_queue(txq); 2785 } 2786 2787 /** 2788 * netif_stop_subqueue - stop sending packets on subqueue 2789 * @dev: network device 2790 * @queue_index: sub queue index 2791 * 2792 * Stop individual transmit queue of a device with multiple transmit queues. 2793 */ 2794 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 2795 { 2796 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2797 netif_tx_stop_queue(txq); 2798 } 2799 2800 /** 2801 * netif_subqueue_stopped - test status of subqueue 2802 * @dev: network device 2803 * @queue_index: sub queue index 2804 * 2805 * Check individual transmit queue of a device with multiple transmit queues. 2806 */ 2807 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 2808 u16 queue_index) 2809 { 2810 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2811 2812 return netif_tx_queue_stopped(txq); 2813 } 2814 2815 static inline bool netif_subqueue_stopped(const struct net_device *dev, 2816 struct sk_buff *skb) 2817 { 2818 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 2819 } 2820 2821 void netif_wake_subqueue(struct net_device *dev, u16 queue_index); 2822 2823 #ifdef CONFIG_XPS 2824 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 2825 u16 index); 2826 #else 2827 static inline int netif_set_xps_queue(struct net_device *dev, 2828 const struct cpumask *mask, 2829 u16 index) 2830 { 2831 return 0; 2832 } 2833 #endif 2834 2835 /* 2836 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used 2837 * as a distribution range limit for the returned value. 2838 */ 2839 static inline u16 skb_tx_hash(const struct net_device *dev, 2840 struct sk_buff *skb) 2841 { 2842 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues); 2843 } 2844 2845 /** 2846 * netif_is_multiqueue - test if device has multiple transmit queues 2847 * @dev: network device 2848 * 2849 * Check if device has multiple transmit queues 2850 */ 2851 static inline bool netif_is_multiqueue(const struct net_device *dev) 2852 { 2853 return dev->num_tx_queues > 1; 2854 } 2855 2856 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 2857 2858 #ifdef CONFIG_SYSFS 2859 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 2860 #else 2861 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 2862 unsigned int rxq) 2863 { 2864 return 0; 2865 } 2866 #endif 2867 2868 #ifdef CONFIG_SYSFS 2869 static inline unsigned int get_netdev_rx_queue_index( 2870 struct netdev_rx_queue *queue) 2871 { 2872 struct net_device *dev = queue->dev; 2873 int index = queue - dev->_rx; 2874 2875 BUG_ON(index >= dev->num_rx_queues); 2876 return index; 2877 } 2878 #endif 2879 2880 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 2881 int netif_get_num_default_rss_queues(void); 2882 2883 enum skb_free_reason { 2884 SKB_REASON_CONSUMED, 2885 SKB_REASON_DROPPED, 2886 }; 2887 2888 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 2889 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 2890 2891 /* 2892 * It is not allowed to call kfree_skb() or consume_skb() from hardware 2893 * interrupt context or with hardware interrupts being disabled. 2894 * (in_irq() || irqs_disabled()) 2895 * 2896 * We provide four helpers that can be used in following contexts : 2897 * 2898 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 2899 * replacing kfree_skb(skb) 2900 * 2901 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 2902 * Typically used in place of consume_skb(skb) in TX completion path 2903 * 2904 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 2905 * replacing kfree_skb(skb) 2906 * 2907 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 2908 * and consumed a packet. Used in place of consume_skb(skb) 2909 */ 2910 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 2911 { 2912 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 2913 } 2914 2915 static inline void dev_consume_skb_irq(struct sk_buff *skb) 2916 { 2917 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 2918 } 2919 2920 static inline void dev_kfree_skb_any(struct sk_buff *skb) 2921 { 2922 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 2923 } 2924 2925 static inline void dev_consume_skb_any(struct sk_buff *skb) 2926 { 2927 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 2928 } 2929 2930 int netif_rx(struct sk_buff *skb); 2931 int netif_rx_ni(struct sk_buff *skb); 2932 int netif_receive_skb(struct sk_buff *skb); 2933 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 2934 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 2935 struct sk_buff *napi_get_frags(struct napi_struct *napi); 2936 gro_result_t napi_gro_frags(struct napi_struct *napi); 2937 struct packet_offload *gro_find_receive_by_type(__be16 type); 2938 struct packet_offload *gro_find_complete_by_type(__be16 type); 2939 2940 static inline void napi_free_frags(struct napi_struct *napi) 2941 { 2942 kfree_skb(napi->skb); 2943 napi->skb = NULL; 2944 } 2945 2946 int netdev_rx_handler_register(struct net_device *dev, 2947 rx_handler_func_t *rx_handler, 2948 void *rx_handler_data); 2949 void netdev_rx_handler_unregister(struct net_device *dev); 2950 2951 bool dev_valid_name(const char *name); 2952 int dev_ioctl(struct net *net, unsigned int cmd, void __user *); 2953 int dev_ethtool(struct net *net, struct ifreq *); 2954 unsigned int dev_get_flags(const struct net_device *); 2955 int __dev_change_flags(struct net_device *, unsigned int flags); 2956 int dev_change_flags(struct net_device *, unsigned int); 2957 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 2958 unsigned int gchanges); 2959 int dev_change_name(struct net_device *, const char *); 2960 int dev_set_alias(struct net_device *, const char *, size_t); 2961 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 2962 int dev_set_mtu(struct net_device *, int); 2963 void dev_set_group(struct net_device *, int); 2964 int dev_set_mac_address(struct net_device *, struct sockaddr *); 2965 int dev_change_carrier(struct net_device *, bool new_carrier); 2966 int dev_get_phys_port_id(struct net_device *dev, 2967 struct netdev_phys_item_id *ppid); 2968 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev); 2969 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 2970 struct netdev_queue *txq, int *ret); 2971 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 2972 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 2973 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb); 2974 2975 extern int netdev_budget; 2976 2977 /* Called by rtnetlink.c:rtnl_unlock() */ 2978 void netdev_run_todo(void); 2979 2980 /** 2981 * dev_put - release reference to device 2982 * @dev: network device 2983 * 2984 * Release reference to device to allow it to be freed. 2985 */ 2986 static inline void dev_put(struct net_device *dev) 2987 { 2988 this_cpu_dec(*dev->pcpu_refcnt); 2989 } 2990 2991 /** 2992 * dev_hold - get reference to device 2993 * @dev: network device 2994 * 2995 * Hold reference to device to keep it from being freed. 2996 */ 2997 static inline void dev_hold(struct net_device *dev) 2998 { 2999 this_cpu_inc(*dev->pcpu_refcnt); 3000 } 3001 3002 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3003 * and _off may be called from IRQ context, but it is caller 3004 * who is responsible for serialization of these calls. 3005 * 3006 * The name carrier is inappropriate, these functions should really be 3007 * called netif_lowerlayer_*() because they represent the state of any 3008 * kind of lower layer not just hardware media. 3009 */ 3010 3011 void linkwatch_init_dev(struct net_device *dev); 3012 void linkwatch_fire_event(struct net_device *dev); 3013 void linkwatch_forget_dev(struct net_device *dev); 3014 3015 /** 3016 * netif_carrier_ok - test if carrier present 3017 * @dev: network device 3018 * 3019 * Check if carrier is present on device 3020 */ 3021 static inline bool netif_carrier_ok(const struct net_device *dev) 3022 { 3023 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3024 } 3025 3026 unsigned long dev_trans_start(struct net_device *dev); 3027 3028 void __netdev_watchdog_up(struct net_device *dev); 3029 3030 void netif_carrier_on(struct net_device *dev); 3031 3032 void netif_carrier_off(struct net_device *dev); 3033 3034 /** 3035 * netif_dormant_on - mark device as dormant. 3036 * @dev: network device 3037 * 3038 * Mark device as dormant (as per RFC2863). 3039 * 3040 * The dormant state indicates that the relevant interface is not 3041 * actually in a condition to pass packets (i.e., it is not 'up') but is 3042 * in a "pending" state, waiting for some external event. For "on- 3043 * demand" interfaces, this new state identifies the situation where the 3044 * interface is waiting for events to place it in the up state. 3045 * 3046 */ 3047 static inline void netif_dormant_on(struct net_device *dev) 3048 { 3049 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3050 linkwatch_fire_event(dev); 3051 } 3052 3053 /** 3054 * netif_dormant_off - set device as not dormant. 3055 * @dev: network device 3056 * 3057 * Device is not in dormant state. 3058 */ 3059 static inline void netif_dormant_off(struct net_device *dev) 3060 { 3061 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3062 linkwatch_fire_event(dev); 3063 } 3064 3065 /** 3066 * netif_dormant - test if carrier present 3067 * @dev: network device 3068 * 3069 * Check if carrier is present on device 3070 */ 3071 static inline bool netif_dormant(const struct net_device *dev) 3072 { 3073 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3074 } 3075 3076 3077 /** 3078 * netif_oper_up - test if device is operational 3079 * @dev: network device 3080 * 3081 * Check if carrier is operational 3082 */ 3083 static inline bool netif_oper_up(const struct net_device *dev) 3084 { 3085 return (dev->operstate == IF_OPER_UP || 3086 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3087 } 3088 3089 /** 3090 * netif_device_present - is device available or removed 3091 * @dev: network device 3092 * 3093 * Check if device has not been removed from system. 3094 */ 3095 static inline bool netif_device_present(struct net_device *dev) 3096 { 3097 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3098 } 3099 3100 void netif_device_detach(struct net_device *dev); 3101 3102 void netif_device_attach(struct net_device *dev); 3103 3104 /* 3105 * Network interface message level settings 3106 */ 3107 3108 enum { 3109 NETIF_MSG_DRV = 0x0001, 3110 NETIF_MSG_PROBE = 0x0002, 3111 NETIF_MSG_LINK = 0x0004, 3112 NETIF_MSG_TIMER = 0x0008, 3113 NETIF_MSG_IFDOWN = 0x0010, 3114 NETIF_MSG_IFUP = 0x0020, 3115 NETIF_MSG_RX_ERR = 0x0040, 3116 NETIF_MSG_TX_ERR = 0x0080, 3117 NETIF_MSG_TX_QUEUED = 0x0100, 3118 NETIF_MSG_INTR = 0x0200, 3119 NETIF_MSG_TX_DONE = 0x0400, 3120 NETIF_MSG_RX_STATUS = 0x0800, 3121 NETIF_MSG_PKTDATA = 0x1000, 3122 NETIF_MSG_HW = 0x2000, 3123 NETIF_MSG_WOL = 0x4000, 3124 }; 3125 3126 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3127 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3128 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3129 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3130 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3131 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3132 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3133 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3134 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3135 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3136 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3137 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3138 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3139 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3140 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3141 3142 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3143 { 3144 /* use default */ 3145 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3146 return default_msg_enable_bits; 3147 if (debug_value == 0) /* no output */ 3148 return 0; 3149 /* set low N bits */ 3150 return (1 << debug_value) - 1; 3151 } 3152 3153 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 3154 { 3155 spin_lock(&txq->_xmit_lock); 3156 txq->xmit_lock_owner = cpu; 3157 } 3158 3159 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 3160 { 3161 spin_lock_bh(&txq->_xmit_lock); 3162 txq->xmit_lock_owner = smp_processor_id(); 3163 } 3164 3165 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 3166 { 3167 bool ok = spin_trylock(&txq->_xmit_lock); 3168 if (likely(ok)) 3169 txq->xmit_lock_owner = smp_processor_id(); 3170 return ok; 3171 } 3172 3173 static inline void __netif_tx_unlock(struct netdev_queue *txq) 3174 { 3175 txq->xmit_lock_owner = -1; 3176 spin_unlock(&txq->_xmit_lock); 3177 } 3178 3179 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 3180 { 3181 txq->xmit_lock_owner = -1; 3182 spin_unlock_bh(&txq->_xmit_lock); 3183 } 3184 3185 static inline void txq_trans_update(struct netdev_queue *txq) 3186 { 3187 if (txq->xmit_lock_owner != -1) 3188 txq->trans_start = jiffies; 3189 } 3190 3191 /** 3192 * netif_tx_lock - grab network device transmit lock 3193 * @dev: network device 3194 * 3195 * Get network device transmit lock 3196 */ 3197 static inline void netif_tx_lock(struct net_device *dev) 3198 { 3199 unsigned int i; 3200 int cpu; 3201 3202 spin_lock(&dev->tx_global_lock); 3203 cpu = smp_processor_id(); 3204 for (i = 0; i < dev->num_tx_queues; i++) { 3205 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3206 3207 /* We are the only thread of execution doing a 3208 * freeze, but we have to grab the _xmit_lock in 3209 * order to synchronize with threads which are in 3210 * the ->hard_start_xmit() handler and already 3211 * checked the frozen bit. 3212 */ 3213 __netif_tx_lock(txq, cpu); 3214 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 3215 __netif_tx_unlock(txq); 3216 } 3217 } 3218 3219 static inline void netif_tx_lock_bh(struct net_device *dev) 3220 { 3221 local_bh_disable(); 3222 netif_tx_lock(dev); 3223 } 3224 3225 static inline void netif_tx_unlock(struct net_device *dev) 3226 { 3227 unsigned int i; 3228 3229 for (i = 0; i < dev->num_tx_queues; i++) { 3230 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3231 3232 /* No need to grab the _xmit_lock here. If the 3233 * queue is not stopped for another reason, we 3234 * force a schedule. 3235 */ 3236 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 3237 netif_schedule_queue(txq); 3238 } 3239 spin_unlock(&dev->tx_global_lock); 3240 } 3241 3242 static inline void netif_tx_unlock_bh(struct net_device *dev) 3243 { 3244 netif_tx_unlock(dev); 3245 local_bh_enable(); 3246 } 3247 3248 #define HARD_TX_LOCK(dev, txq, cpu) { \ 3249 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3250 __netif_tx_lock(txq, cpu); \ 3251 } \ 3252 } 3253 3254 #define HARD_TX_TRYLOCK(dev, txq) \ 3255 (((dev->features & NETIF_F_LLTX) == 0) ? \ 3256 __netif_tx_trylock(txq) : \ 3257 true ) 3258 3259 #define HARD_TX_UNLOCK(dev, txq) { \ 3260 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3261 __netif_tx_unlock(txq); \ 3262 } \ 3263 } 3264 3265 static inline void netif_tx_disable(struct net_device *dev) 3266 { 3267 unsigned int i; 3268 int cpu; 3269 3270 local_bh_disable(); 3271 cpu = smp_processor_id(); 3272 for (i = 0; i < dev->num_tx_queues; i++) { 3273 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3274 3275 __netif_tx_lock(txq, cpu); 3276 netif_tx_stop_queue(txq); 3277 __netif_tx_unlock(txq); 3278 } 3279 local_bh_enable(); 3280 } 3281 3282 static inline void netif_addr_lock(struct net_device *dev) 3283 { 3284 spin_lock(&dev->addr_list_lock); 3285 } 3286 3287 static inline void netif_addr_lock_nested(struct net_device *dev) 3288 { 3289 int subclass = SINGLE_DEPTH_NESTING; 3290 3291 if (dev->netdev_ops->ndo_get_lock_subclass) 3292 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev); 3293 3294 spin_lock_nested(&dev->addr_list_lock, subclass); 3295 } 3296 3297 static inline void netif_addr_lock_bh(struct net_device *dev) 3298 { 3299 spin_lock_bh(&dev->addr_list_lock); 3300 } 3301 3302 static inline void netif_addr_unlock(struct net_device *dev) 3303 { 3304 spin_unlock(&dev->addr_list_lock); 3305 } 3306 3307 static inline void netif_addr_unlock_bh(struct net_device *dev) 3308 { 3309 spin_unlock_bh(&dev->addr_list_lock); 3310 } 3311 3312 /* 3313 * dev_addrs walker. Should be used only for read access. Call with 3314 * rcu_read_lock held. 3315 */ 3316 #define for_each_dev_addr(dev, ha) \ 3317 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 3318 3319 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 3320 3321 void ether_setup(struct net_device *dev); 3322 3323 /* Support for loadable net-drivers */ 3324 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 3325 unsigned char name_assign_type, 3326 void (*setup)(struct net_device *), 3327 unsigned int txqs, unsigned int rxqs); 3328 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 3329 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 3330 3331 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 3332 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 3333 count) 3334 3335 int register_netdev(struct net_device *dev); 3336 void unregister_netdev(struct net_device *dev); 3337 3338 /* General hardware address lists handling functions */ 3339 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 3340 struct netdev_hw_addr_list *from_list, int addr_len); 3341 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 3342 struct netdev_hw_addr_list *from_list, int addr_len); 3343 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 3344 struct net_device *dev, 3345 int (*sync)(struct net_device *, const unsigned char *), 3346 int (*unsync)(struct net_device *, 3347 const unsigned char *)); 3348 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 3349 struct net_device *dev, 3350 int (*unsync)(struct net_device *, 3351 const unsigned char *)); 3352 void __hw_addr_init(struct netdev_hw_addr_list *list); 3353 3354 /* Functions used for device addresses handling */ 3355 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 3356 unsigned char addr_type); 3357 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 3358 unsigned char addr_type); 3359 void dev_addr_flush(struct net_device *dev); 3360 int dev_addr_init(struct net_device *dev); 3361 3362 /* Functions used for unicast addresses handling */ 3363 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 3364 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 3365 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 3366 int dev_uc_sync(struct net_device *to, struct net_device *from); 3367 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 3368 void dev_uc_unsync(struct net_device *to, struct net_device *from); 3369 void dev_uc_flush(struct net_device *dev); 3370 void dev_uc_init(struct net_device *dev); 3371 3372 /** 3373 * __dev_uc_sync - Synchonize device's unicast list 3374 * @dev: device to sync 3375 * @sync: function to call if address should be added 3376 * @unsync: function to call if address should be removed 3377 * 3378 * Add newly added addresses to the interface, and release 3379 * addresses that have been deleted. 3380 **/ 3381 static inline int __dev_uc_sync(struct net_device *dev, 3382 int (*sync)(struct net_device *, 3383 const unsigned char *), 3384 int (*unsync)(struct net_device *, 3385 const unsigned char *)) 3386 { 3387 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 3388 } 3389 3390 /** 3391 * __dev_uc_unsync - Remove synchronized addresses from device 3392 * @dev: device to sync 3393 * @unsync: function to call if address should be removed 3394 * 3395 * Remove all addresses that were added to the device by dev_uc_sync(). 3396 **/ 3397 static inline void __dev_uc_unsync(struct net_device *dev, 3398 int (*unsync)(struct net_device *, 3399 const unsigned char *)) 3400 { 3401 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 3402 } 3403 3404 /* Functions used for multicast addresses handling */ 3405 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 3406 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 3407 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 3408 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 3409 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 3410 int dev_mc_sync(struct net_device *to, struct net_device *from); 3411 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 3412 void dev_mc_unsync(struct net_device *to, struct net_device *from); 3413 void dev_mc_flush(struct net_device *dev); 3414 void dev_mc_init(struct net_device *dev); 3415 3416 /** 3417 * __dev_mc_sync - Synchonize device's multicast list 3418 * @dev: device to sync 3419 * @sync: function to call if address should be added 3420 * @unsync: function to call if address should be removed 3421 * 3422 * Add newly added addresses to the interface, and release 3423 * addresses that have been deleted. 3424 **/ 3425 static inline int __dev_mc_sync(struct net_device *dev, 3426 int (*sync)(struct net_device *, 3427 const unsigned char *), 3428 int (*unsync)(struct net_device *, 3429 const unsigned char *)) 3430 { 3431 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 3432 } 3433 3434 /** 3435 * __dev_mc_unsync - Remove synchronized addresses from device 3436 * @dev: device to sync 3437 * @unsync: function to call if address should be removed 3438 * 3439 * Remove all addresses that were added to the device by dev_mc_sync(). 3440 **/ 3441 static inline void __dev_mc_unsync(struct net_device *dev, 3442 int (*unsync)(struct net_device *, 3443 const unsigned char *)) 3444 { 3445 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 3446 } 3447 3448 /* Functions used for secondary unicast and multicast support */ 3449 void dev_set_rx_mode(struct net_device *dev); 3450 void __dev_set_rx_mode(struct net_device *dev); 3451 int dev_set_promiscuity(struct net_device *dev, int inc); 3452 int dev_set_allmulti(struct net_device *dev, int inc); 3453 void netdev_state_change(struct net_device *dev); 3454 void netdev_notify_peers(struct net_device *dev); 3455 void netdev_features_change(struct net_device *dev); 3456 /* Load a device via the kmod */ 3457 void dev_load(struct net *net, const char *name); 3458 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 3459 struct rtnl_link_stats64 *storage); 3460 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 3461 const struct net_device_stats *netdev_stats); 3462 3463 extern int netdev_max_backlog; 3464 extern int netdev_tstamp_prequeue; 3465 extern int weight_p; 3466 extern int bpf_jit_enable; 3467 3468 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 3469 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 3470 struct list_head **iter); 3471 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 3472 struct list_head **iter); 3473 3474 /* iterate through upper list, must be called under RCU read lock */ 3475 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 3476 for (iter = &(dev)->adj_list.upper, \ 3477 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 3478 updev; \ 3479 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 3480 3481 /* iterate through upper list, must be called under RCU read lock */ 3482 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \ 3483 for (iter = &(dev)->all_adj_list.upper, \ 3484 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \ 3485 updev; \ 3486 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter))) 3487 3488 void *netdev_lower_get_next_private(struct net_device *dev, 3489 struct list_head **iter); 3490 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 3491 struct list_head **iter); 3492 3493 #define netdev_for_each_lower_private(dev, priv, iter) \ 3494 for (iter = (dev)->adj_list.lower.next, \ 3495 priv = netdev_lower_get_next_private(dev, &(iter)); \ 3496 priv; \ 3497 priv = netdev_lower_get_next_private(dev, &(iter))) 3498 3499 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 3500 for (iter = &(dev)->adj_list.lower, \ 3501 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 3502 priv; \ 3503 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 3504 3505 void *netdev_lower_get_next(struct net_device *dev, 3506 struct list_head **iter); 3507 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 3508 for (iter = &(dev)->adj_list.lower, \ 3509 ldev = netdev_lower_get_next(dev, &(iter)); \ 3510 ldev; \ 3511 ldev = netdev_lower_get_next(dev, &(iter))) 3512 3513 void *netdev_adjacent_get_private(struct list_head *adj_list); 3514 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 3515 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 3516 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 3517 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev); 3518 int netdev_master_upper_dev_link(struct net_device *dev, 3519 struct net_device *upper_dev); 3520 int netdev_master_upper_dev_link_private(struct net_device *dev, 3521 struct net_device *upper_dev, 3522 void *private); 3523 void netdev_upper_dev_unlink(struct net_device *dev, 3524 struct net_device *upper_dev); 3525 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 3526 void *netdev_lower_dev_get_private(struct net_device *dev, 3527 struct net_device *lower_dev); 3528 3529 /* RSS keys are 40 or 52 bytes long */ 3530 #define NETDEV_RSS_KEY_LEN 52 3531 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN]; 3532 void netdev_rss_key_fill(void *buffer, size_t len); 3533 3534 int dev_get_nest_level(struct net_device *dev, 3535 bool (*type_check)(struct net_device *dev)); 3536 int skb_checksum_help(struct sk_buff *skb); 3537 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 3538 netdev_features_t features, bool tx_path); 3539 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 3540 netdev_features_t features); 3541 3542 struct netdev_bonding_info { 3543 ifslave slave; 3544 ifbond master; 3545 }; 3546 3547 struct netdev_notifier_bonding_info { 3548 struct netdev_notifier_info info; /* must be first */ 3549 struct netdev_bonding_info bonding_info; 3550 }; 3551 3552 void netdev_bonding_info_change(struct net_device *dev, 3553 struct netdev_bonding_info *bonding_info); 3554 3555 static inline 3556 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 3557 { 3558 return __skb_gso_segment(skb, features, true); 3559 } 3560 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 3561 3562 static inline bool can_checksum_protocol(netdev_features_t features, 3563 __be16 protocol) 3564 { 3565 return ((features & NETIF_F_GEN_CSUM) || 3566 ((features & NETIF_F_V4_CSUM) && 3567 protocol == htons(ETH_P_IP)) || 3568 ((features & NETIF_F_V6_CSUM) && 3569 protocol == htons(ETH_P_IPV6)) || 3570 ((features & NETIF_F_FCOE_CRC) && 3571 protocol == htons(ETH_P_FCOE))); 3572 } 3573 3574 #ifdef CONFIG_BUG 3575 void netdev_rx_csum_fault(struct net_device *dev); 3576 #else 3577 static inline void netdev_rx_csum_fault(struct net_device *dev) 3578 { 3579 } 3580 #endif 3581 /* rx skb timestamps */ 3582 void net_enable_timestamp(void); 3583 void net_disable_timestamp(void); 3584 3585 #ifdef CONFIG_PROC_FS 3586 int __init dev_proc_init(void); 3587 #else 3588 #define dev_proc_init() 0 3589 #endif 3590 3591 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 3592 struct sk_buff *skb, struct net_device *dev, 3593 bool more) 3594 { 3595 skb->xmit_more = more ? 1 : 0; 3596 return ops->ndo_start_xmit(skb, dev); 3597 } 3598 3599 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 3600 struct netdev_queue *txq, bool more) 3601 { 3602 const struct net_device_ops *ops = dev->netdev_ops; 3603 int rc; 3604 3605 rc = __netdev_start_xmit(ops, skb, dev, more); 3606 if (rc == NETDEV_TX_OK) 3607 txq_trans_update(txq); 3608 3609 return rc; 3610 } 3611 3612 int netdev_class_create_file_ns(struct class_attribute *class_attr, 3613 const void *ns); 3614 void netdev_class_remove_file_ns(struct class_attribute *class_attr, 3615 const void *ns); 3616 3617 static inline int netdev_class_create_file(struct class_attribute *class_attr) 3618 { 3619 return netdev_class_create_file_ns(class_attr, NULL); 3620 } 3621 3622 static inline void netdev_class_remove_file(struct class_attribute *class_attr) 3623 { 3624 netdev_class_remove_file_ns(class_attr, NULL); 3625 } 3626 3627 extern struct kobj_ns_type_operations net_ns_type_operations; 3628 3629 const char *netdev_drivername(const struct net_device *dev); 3630 3631 void linkwatch_run_queue(void); 3632 3633 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 3634 netdev_features_t f2) 3635 { 3636 if (f1 & NETIF_F_GEN_CSUM) 3637 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM); 3638 if (f2 & NETIF_F_GEN_CSUM) 3639 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM); 3640 f1 &= f2; 3641 if (f1 & NETIF_F_GEN_CSUM) 3642 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM); 3643 3644 return f1; 3645 } 3646 3647 static inline netdev_features_t netdev_get_wanted_features( 3648 struct net_device *dev) 3649 { 3650 return (dev->features & ~dev->hw_features) | dev->wanted_features; 3651 } 3652 netdev_features_t netdev_increment_features(netdev_features_t all, 3653 netdev_features_t one, netdev_features_t mask); 3654 3655 /* Allow TSO being used on stacked device : 3656 * Performing the GSO segmentation before last device 3657 * is a performance improvement. 3658 */ 3659 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 3660 netdev_features_t mask) 3661 { 3662 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 3663 } 3664 3665 int __netdev_update_features(struct net_device *dev); 3666 void netdev_update_features(struct net_device *dev); 3667 void netdev_change_features(struct net_device *dev); 3668 3669 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 3670 struct net_device *dev); 3671 3672 netdev_features_t netif_skb_features(struct sk_buff *skb); 3673 3674 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 3675 { 3676 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT; 3677 3678 /* check flags correspondence */ 3679 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 3680 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT)); 3681 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 3682 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 3683 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 3684 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 3685 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 3686 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 3687 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT)); 3688 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT)); 3689 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 3690 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 3691 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 3692 3693 return (features & feature) == feature; 3694 } 3695 3696 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 3697 { 3698 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 3699 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 3700 } 3701 3702 static inline bool netif_needs_gso(struct net_device *dev, struct sk_buff *skb, 3703 netdev_features_t features) 3704 { 3705 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 3706 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 3707 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 3708 } 3709 3710 static inline void netif_set_gso_max_size(struct net_device *dev, 3711 unsigned int size) 3712 { 3713 dev->gso_max_size = size; 3714 } 3715 3716 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 3717 int pulled_hlen, u16 mac_offset, 3718 int mac_len) 3719 { 3720 skb->protocol = protocol; 3721 skb->encapsulation = 1; 3722 skb_push(skb, pulled_hlen); 3723 skb_reset_transport_header(skb); 3724 skb->mac_header = mac_offset; 3725 skb->network_header = skb->mac_header + mac_len; 3726 skb->mac_len = mac_len; 3727 } 3728 3729 static inline bool netif_is_macvlan(struct net_device *dev) 3730 { 3731 return dev->priv_flags & IFF_MACVLAN; 3732 } 3733 3734 static inline bool netif_is_macvlan_port(struct net_device *dev) 3735 { 3736 return dev->priv_flags & IFF_MACVLAN_PORT; 3737 } 3738 3739 static inline bool netif_is_ipvlan(struct net_device *dev) 3740 { 3741 return dev->priv_flags & IFF_IPVLAN_SLAVE; 3742 } 3743 3744 static inline bool netif_is_ipvlan_port(struct net_device *dev) 3745 { 3746 return dev->priv_flags & IFF_IPVLAN_MASTER; 3747 } 3748 3749 static inline bool netif_is_bond_master(struct net_device *dev) 3750 { 3751 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 3752 } 3753 3754 static inline bool netif_is_bond_slave(struct net_device *dev) 3755 { 3756 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 3757 } 3758 3759 static inline bool netif_supports_nofcs(struct net_device *dev) 3760 { 3761 return dev->priv_flags & IFF_SUPP_NOFCS; 3762 } 3763 3764 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 3765 static inline void netif_keep_dst(struct net_device *dev) 3766 { 3767 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 3768 } 3769 3770 extern struct pernet_operations __net_initdata loopback_net_ops; 3771 3772 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 3773 3774 /* netdev_printk helpers, similar to dev_printk */ 3775 3776 static inline const char *netdev_name(const struct net_device *dev) 3777 { 3778 if (!dev->name[0] || strchr(dev->name, '%')) 3779 return "(unnamed net_device)"; 3780 return dev->name; 3781 } 3782 3783 static inline const char *netdev_reg_state(const struct net_device *dev) 3784 { 3785 switch (dev->reg_state) { 3786 case NETREG_UNINITIALIZED: return " (uninitialized)"; 3787 case NETREG_REGISTERED: return ""; 3788 case NETREG_UNREGISTERING: return " (unregistering)"; 3789 case NETREG_UNREGISTERED: return " (unregistered)"; 3790 case NETREG_RELEASED: return " (released)"; 3791 case NETREG_DUMMY: return " (dummy)"; 3792 } 3793 3794 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 3795 return " (unknown)"; 3796 } 3797 3798 __printf(3, 4) 3799 void netdev_printk(const char *level, const struct net_device *dev, 3800 const char *format, ...); 3801 __printf(2, 3) 3802 void netdev_emerg(const struct net_device *dev, const char *format, ...); 3803 __printf(2, 3) 3804 void netdev_alert(const struct net_device *dev, const char *format, ...); 3805 __printf(2, 3) 3806 void netdev_crit(const struct net_device *dev, const char *format, ...); 3807 __printf(2, 3) 3808 void netdev_err(const struct net_device *dev, const char *format, ...); 3809 __printf(2, 3) 3810 void netdev_warn(const struct net_device *dev, const char *format, ...); 3811 __printf(2, 3) 3812 void netdev_notice(const struct net_device *dev, const char *format, ...); 3813 __printf(2, 3) 3814 void netdev_info(const struct net_device *dev, const char *format, ...); 3815 3816 #define MODULE_ALIAS_NETDEV(device) \ 3817 MODULE_ALIAS("netdev-" device) 3818 3819 #if defined(CONFIG_DYNAMIC_DEBUG) 3820 #define netdev_dbg(__dev, format, args...) \ 3821 do { \ 3822 dynamic_netdev_dbg(__dev, format, ##args); \ 3823 } while (0) 3824 #elif defined(DEBUG) 3825 #define netdev_dbg(__dev, format, args...) \ 3826 netdev_printk(KERN_DEBUG, __dev, format, ##args) 3827 #else 3828 #define netdev_dbg(__dev, format, args...) \ 3829 ({ \ 3830 if (0) \ 3831 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 3832 }) 3833 #endif 3834 3835 #if defined(VERBOSE_DEBUG) 3836 #define netdev_vdbg netdev_dbg 3837 #else 3838 3839 #define netdev_vdbg(dev, format, args...) \ 3840 ({ \ 3841 if (0) \ 3842 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 3843 0; \ 3844 }) 3845 #endif 3846 3847 /* 3848 * netdev_WARN() acts like dev_printk(), but with the key difference 3849 * of using a WARN/WARN_ON to get the message out, including the 3850 * file/line information and a backtrace. 3851 */ 3852 #define netdev_WARN(dev, format, args...) \ 3853 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \ 3854 netdev_reg_state(dev), ##args) 3855 3856 /* netif printk helpers, similar to netdev_printk */ 3857 3858 #define netif_printk(priv, type, level, dev, fmt, args...) \ 3859 do { \ 3860 if (netif_msg_##type(priv)) \ 3861 netdev_printk(level, (dev), fmt, ##args); \ 3862 } while (0) 3863 3864 #define netif_level(level, priv, type, dev, fmt, args...) \ 3865 do { \ 3866 if (netif_msg_##type(priv)) \ 3867 netdev_##level(dev, fmt, ##args); \ 3868 } while (0) 3869 3870 #define netif_emerg(priv, type, dev, fmt, args...) \ 3871 netif_level(emerg, priv, type, dev, fmt, ##args) 3872 #define netif_alert(priv, type, dev, fmt, args...) \ 3873 netif_level(alert, priv, type, dev, fmt, ##args) 3874 #define netif_crit(priv, type, dev, fmt, args...) \ 3875 netif_level(crit, priv, type, dev, fmt, ##args) 3876 #define netif_err(priv, type, dev, fmt, args...) \ 3877 netif_level(err, priv, type, dev, fmt, ##args) 3878 #define netif_warn(priv, type, dev, fmt, args...) \ 3879 netif_level(warn, priv, type, dev, fmt, ##args) 3880 #define netif_notice(priv, type, dev, fmt, args...) \ 3881 netif_level(notice, priv, type, dev, fmt, ##args) 3882 #define netif_info(priv, type, dev, fmt, args...) \ 3883 netif_level(info, priv, type, dev, fmt, ##args) 3884 3885 #if defined(CONFIG_DYNAMIC_DEBUG) 3886 #define netif_dbg(priv, type, netdev, format, args...) \ 3887 do { \ 3888 if (netif_msg_##type(priv)) \ 3889 dynamic_netdev_dbg(netdev, format, ##args); \ 3890 } while (0) 3891 #elif defined(DEBUG) 3892 #define netif_dbg(priv, type, dev, format, args...) \ 3893 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 3894 #else 3895 #define netif_dbg(priv, type, dev, format, args...) \ 3896 ({ \ 3897 if (0) \ 3898 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 3899 0; \ 3900 }) 3901 #endif 3902 3903 #if defined(VERBOSE_DEBUG) 3904 #define netif_vdbg netif_dbg 3905 #else 3906 #define netif_vdbg(priv, type, dev, format, args...) \ 3907 ({ \ 3908 if (0) \ 3909 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 3910 0; \ 3911 }) 3912 #endif 3913 3914 /* 3915 * The list of packet types we will receive (as opposed to discard) 3916 * and the routines to invoke. 3917 * 3918 * Why 16. Because with 16 the only overlap we get on a hash of the 3919 * low nibble of the protocol value is RARP/SNAP/X.25. 3920 * 3921 * NOTE: That is no longer true with the addition of VLAN tags. Not 3922 * sure which should go first, but I bet it won't make much 3923 * difference if we are running VLANs. The good news is that 3924 * this protocol won't be in the list unless compiled in, so 3925 * the average user (w/out VLANs) will not be adversely affected. 3926 * --BLG 3927 * 3928 * 0800 IP 3929 * 8100 802.1Q VLAN 3930 * 0001 802.3 3931 * 0002 AX.25 3932 * 0004 802.2 3933 * 8035 RARP 3934 * 0005 SNAP 3935 * 0805 X.25 3936 * 0806 ARP 3937 * 8137 IPX 3938 * 0009 Localtalk 3939 * 86DD IPv6 3940 */ 3941 #define PTYPE_HASH_SIZE (16) 3942 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 3943 3944 #endif /* _LINUX_NETDEVICE_H */ 3945