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/timer.h> 29 #include <linux/bug.h> 30 #include <linux/delay.h> 31 #include <linux/atomic.h> 32 #include <linux/prefetch.h> 33 #include <asm/cache.h> 34 #include <asm/byteorder.h> 35 36 #include <linux/percpu.h> 37 #include <linux/rculist.h> 38 #include <linux/dmaengine.h> 39 #include <linux/workqueue.h> 40 #include <linux/dynamic_queue_limits.h> 41 42 #include <linux/ethtool.h> 43 #include <net/net_namespace.h> 44 #include <net/dsa.h> 45 #ifdef CONFIG_DCB 46 #include <net/dcbnl.h> 47 #endif 48 #include <net/netprio_cgroup.h> 49 50 #include <linux/netdev_features.h> 51 #include <linux/neighbour.h> 52 #include <uapi/linux/netdevice.h> 53 #include <uapi/linux/if_bonding.h> 54 #include <uapi/linux/pkt_cls.h> 55 56 struct netpoll_info; 57 struct device; 58 struct phy_device; 59 /* 802.11 specific */ 60 struct wireless_dev; 61 /* 802.15.4 specific */ 62 struct wpan_dev; 63 struct mpls_dev; 64 /* UDP Tunnel offloads */ 65 struct udp_tunnel_info; 66 struct bpf_prog; 67 68 void netdev_set_default_ethtool_ops(struct net_device *dev, 69 const struct ethtool_ops *ops); 70 71 /* Backlog congestion levels */ 72 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 73 #define NET_RX_DROP 1 /* packet dropped */ 74 75 /* 76 * Transmit return codes: transmit return codes originate from three different 77 * namespaces: 78 * 79 * - qdisc return codes 80 * - driver transmit return codes 81 * - errno values 82 * 83 * Drivers are allowed to return any one of those in their hard_start_xmit() 84 * function. Real network devices commonly used with qdiscs should only return 85 * the driver transmit return codes though - when qdiscs are used, the actual 86 * transmission happens asynchronously, so the value is not propagated to 87 * higher layers. Virtual network devices transmit synchronously; in this case 88 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 89 * others are propagated to higher layers. 90 */ 91 92 /* qdisc ->enqueue() return codes. */ 93 #define NET_XMIT_SUCCESS 0x00 94 #define NET_XMIT_DROP 0x01 /* skb dropped */ 95 #define NET_XMIT_CN 0x02 /* congestion notification */ 96 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 97 98 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 99 * indicates that the device will soon be dropping packets, or already drops 100 * some packets of the same priority; prompting us to send less aggressively. */ 101 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 102 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 103 104 /* Driver transmit return codes */ 105 #define NETDEV_TX_MASK 0xf0 106 107 enum netdev_tx { 108 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 109 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 110 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 111 }; 112 typedef enum netdev_tx netdev_tx_t; 113 114 /* 115 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 116 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 117 */ 118 static inline bool dev_xmit_complete(int rc) 119 { 120 /* 121 * Positive cases with an skb consumed by a driver: 122 * - successful transmission (rc == NETDEV_TX_OK) 123 * - error while transmitting (rc < 0) 124 * - error while queueing to a different device (rc & NET_XMIT_MASK) 125 */ 126 if (likely(rc < NET_XMIT_MASK)) 127 return true; 128 129 return false; 130 } 131 132 /* 133 * Compute the worst-case header length according to the protocols 134 * used. 135 */ 136 137 #if defined(CONFIG_HYPERV_NET) 138 # define LL_MAX_HEADER 128 139 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 140 # if defined(CONFIG_MAC80211_MESH) 141 # define LL_MAX_HEADER 128 142 # else 143 # define LL_MAX_HEADER 96 144 # endif 145 #else 146 # define LL_MAX_HEADER 32 147 #endif 148 149 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 150 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 151 #define MAX_HEADER LL_MAX_HEADER 152 #else 153 #define MAX_HEADER (LL_MAX_HEADER + 48) 154 #endif 155 156 /* 157 * Old network device statistics. Fields are native words 158 * (unsigned long) so they can be read and written atomically. 159 */ 160 161 struct net_device_stats { 162 unsigned long rx_packets; 163 unsigned long tx_packets; 164 unsigned long rx_bytes; 165 unsigned long tx_bytes; 166 unsigned long rx_errors; 167 unsigned long tx_errors; 168 unsigned long rx_dropped; 169 unsigned long tx_dropped; 170 unsigned long multicast; 171 unsigned long collisions; 172 unsigned long rx_length_errors; 173 unsigned long rx_over_errors; 174 unsigned long rx_crc_errors; 175 unsigned long rx_frame_errors; 176 unsigned long rx_fifo_errors; 177 unsigned long rx_missed_errors; 178 unsigned long tx_aborted_errors; 179 unsigned long tx_carrier_errors; 180 unsigned long tx_fifo_errors; 181 unsigned long tx_heartbeat_errors; 182 unsigned long tx_window_errors; 183 unsigned long rx_compressed; 184 unsigned long tx_compressed; 185 }; 186 187 188 #include <linux/cache.h> 189 #include <linux/skbuff.h> 190 191 #ifdef CONFIG_RPS 192 #include <linux/static_key.h> 193 extern struct static_key rps_needed; 194 #endif 195 196 struct neighbour; 197 struct neigh_parms; 198 struct sk_buff; 199 200 struct netdev_hw_addr { 201 struct list_head list; 202 unsigned char addr[MAX_ADDR_LEN]; 203 unsigned char type; 204 #define NETDEV_HW_ADDR_T_LAN 1 205 #define NETDEV_HW_ADDR_T_SAN 2 206 #define NETDEV_HW_ADDR_T_SLAVE 3 207 #define NETDEV_HW_ADDR_T_UNICAST 4 208 #define NETDEV_HW_ADDR_T_MULTICAST 5 209 bool global_use; 210 int sync_cnt; 211 int refcount; 212 int synced; 213 struct rcu_head rcu_head; 214 }; 215 216 struct netdev_hw_addr_list { 217 struct list_head list; 218 int count; 219 }; 220 221 #define netdev_hw_addr_list_count(l) ((l)->count) 222 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 223 #define netdev_hw_addr_list_for_each(ha, l) \ 224 list_for_each_entry(ha, &(l)->list, list) 225 226 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 227 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 228 #define netdev_for_each_uc_addr(ha, dev) \ 229 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 230 231 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 232 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 233 #define netdev_for_each_mc_addr(ha, dev) \ 234 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 235 236 struct hh_cache { 237 u16 hh_len; 238 u16 __pad; 239 seqlock_t hh_lock; 240 241 /* cached hardware header; allow for machine alignment needs. */ 242 #define HH_DATA_MOD 16 243 #define HH_DATA_OFF(__len) \ 244 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 245 #define HH_DATA_ALIGN(__len) \ 246 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 247 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 248 }; 249 250 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 251 * Alternative is: 252 * dev->hard_header_len ? (dev->hard_header_len + 253 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 254 * 255 * We could use other alignment values, but we must maintain the 256 * relationship HH alignment <= LL alignment. 257 */ 258 #define LL_RESERVED_SPACE(dev) \ 259 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 260 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 261 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 262 263 struct header_ops { 264 int (*create) (struct sk_buff *skb, struct net_device *dev, 265 unsigned short type, const void *daddr, 266 const void *saddr, unsigned int len); 267 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 268 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 269 void (*cache_update)(struct hh_cache *hh, 270 const struct net_device *dev, 271 const unsigned char *haddr); 272 bool (*validate)(const char *ll_header, unsigned int len); 273 }; 274 275 /* These flag bits are private to the generic network queueing 276 * layer; they may not be explicitly referenced by any other 277 * code. 278 */ 279 280 enum netdev_state_t { 281 __LINK_STATE_START, 282 __LINK_STATE_PRESENT, 283 __LINK_STATE_NOCARRIER, 284 __LINK_STATE_LINKWATCH_PENDING, 285 __LINK_STATE_DORMANT, 286 }; 287 288 289 /* 290 * This structure holds boot-time configured netdevice settings. They 291 * are then used in the device probing. 292 */ 293 struct netdev_boot_setup { 294 char name[IFNAMSIZ]; 295 struct ifmap map; 296 }; 297 #define NETDEV_BOOT_SETUP_MAX 8 298 299 int __init netdev_boot_setup(char *str); 300 301 /* 302 * Structure for NAPI scheduling similar to tasklet but with weighting 303 */ 304 struct napi_struct { 305 /* The poll_list must only be managed by the entity which 306 * changes the state of the NAPI_STATE_SCHED bit. This means 307 * whoever atomically sets that bit can add this napi_struct 308 * to the per-CPU poll_list, and whoever clears that bit 309 * can remove from the list right before clearing the bit. 310 */ 311 struct list_head poll_list; 312 313 unsigned long state; 314 int weight; 315 unsigned int gro_count; 316 int (*poll)(struct napi_struct *, int); 317 #ifdef CONFIG_NETPOLL 318 spinlock_t poll_lock; 319 int poll_owner; 320 #endif 321 struct net_device *dev; 322 struct sk_buff *gro_list; 323 struct sk_buff *skb; 324 struct hrtimer timer; 325 struct list_head dev_list; 326 struct hlist_node napi_hash_node; 327 unsigned int napi_id; 328 }; 329 330 enum { 331 NAPI_STATE_SCHED, /* Poll is scheduled */ 332 NAPI_STATE_DISABLE, /* Disable pending */ 333 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 334 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */ 335 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */ 336 }; 337 338 enum gro_result { 339 GRO_MERGED, 340 GRO_MERGED_FREE, 341 GRO_HELD, 342 GRO_NORMAL, 343 GRO_DROP, 344 }; 345 typedef enum gro_result gro_result_t; 346 347 /* 348 * enum rx_handler_result - Possible return values for rx_handlers. 349 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 350 * further. 351 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 352 * case skb->dev was changed by rx_handler. 353 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 354 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 355 * 356 * rx_handlers are functions called from inside __netif_receive_skb(), to do 357 * special processing of the skb, prior to delivery to protocol handlers. 358 * 359 * Currently, a net_device can only have a single rx_handler registered. Trying 360 * to register a second rx_handler will return -EBUSY. 361 * 362 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 363 * To unregister a rx_handler on a net_device, use 364 * netdev_rx_handler_unregister(). 365 * 366 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 367 * do with the skb. 368 * 369 * If the rx_handler consumed the skb in some way, it should return 370 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 371 * the skb to be delivered in some other way. 372 * 373 * If the rx_handler changed skb->dev, to divert the skb to another 374 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 375 * new device will be called if it exists. 376 * 377 * If the rx_handler decides the skb should be ignored, it should return 378 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 379 * are registered on exact device (ptype->dev == skb->dev). 380 * 381 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 382 * delivered, it should return RX_HANDLER_PASS. 383 * 384 * A device without a registered rx_handler will behave as if rx_handler 385 * returned RX_HANDLER_PASS. 386 */ 387 388 enum rx_handler_result { 389 RX_HANDLER_CONSUMED, 390 RX_HANDLER_ANOTHER, 391 RX_HANDLER_EXACT, 392 RX_HANDLER_PASS, 393 }; 394 typedef enum rx_handler_result rx_handler_result_t; 395 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 396 397 void __napi_schedule(struct napi_struct *n); 398 void __napi_schedule_irqoff(struct napi_struct *n); 399 400 static inline bool napi_disable_pending(struct napi_struct *n) 401 { 402 return test_bit(NAPI_STATE_DISABLE, &n->state); 403 } 404 405 /** 406 * napi_schedule_prep - check if NAPI can be scheduled 407 * @n: NAPI context 408 * 409 * Test if NAPI routine is already running, and if not mark 410 * it as running. This is used as a condition variable to 411 * insure only one NAPI poll instance runs. We also make 412 * sure there is no pending NAPI disable. 413 */ 414 static inline bool napi_schedule_prep(struct napi_struct *n) 415 { 416 return !napi_disable_pending(n) && 417 !test_and_set_bit(NAPI_STATE_SCHED, &n->state); 418 } 419 420 /** 421 * napi_schedule - schedule NAPI poll 422 * @n: NAPI context 423 * 424 * Schedule NAPI poll routine to be called if it is not already 425 * running. 426 */ 427 static inline void napi_schedule(struct napi_struct *n) 428 { 429 if (napi_schedule_prep(n)) 430 __napi_schedule(n); 431 } 432 433 /** 434 * napi_schedule_irqoff - schedule NAPI poll 435 * @n: NAPI context 436 * 437 * Variant of napi_schedule(), assuming hard irqs are masked. 438 */ 439 static inline void napi_schedule_irqoff(struct napi_struct *n) 440 { 441 if (napi_schedule_prep(n)) 442 __napi_schedule_irqoff(n); 443 } 444 445 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 446 static inline bool napi_reschedule(struct napi_struct *napi) 447 { 448 if (napi_schedule_prep(napi)) { 449 __napi_schedule(napi); 450 return true; 451 } 452 return false; 453 } 454 455 void __napi_complete(struct napi_struct *n); 456 void napi_complete_done(struct napi_struct *n, int work_done); 457 /** 458 * napi_complete - NAPI processing complete 459 * @n: NAPI context 460 * 461 * Mark NAPI processing as complete. 462 * Consider using napi_complete_done() instead. 463 */ 464 static inline void napi_complete(struct napi_struct *n) 465 { 466 return napi_complete_done(n, 0); 467 } 468 469 /** 470 * napi_hash_add - add a NAPI to global hashtable 471 * @napi: NAPI context 472 * 473 * Generate a new napi_id and store a @napi under it in napi_hash. 474 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL). 475 * Note: This is normally automatically done from netif_napi_add(), 476 * so might disappear in a future Linux version. 477 */ 478 void napi_hash_add(struct napi_struct *napi); 479 480 /** 481 * napi_hash_del - remove a NAPI from global table 482 * @napi: NAPI context 483 * 484 * Warning: caller must observe RCU grace period 485 * before freeing memory containing @napi, if 486 * this function returns true. 487 * Note: core networking stack automatically calls it 488 * from netif_napi_del(). 489 * Drivers might want to call this helper to combine all 490 * the needed RCU grace periods into a single one. 491 */ 492 bool napi_hash_del(struct napi_struct *napi); 493 494 /** 495 * napi_disable - prevent NAPI from scheduling 496 * @n: NAPI context 497 * 498 * Stop NAPI from being scheduled on this context. 499 * Waits till any outstanding processing completes. 500 */ 501 void napi_disable(struct napi_struct *n); 502 503 /** 504 * napi_enable - enable NAPI scheduling 505 * @n: NAPI context 506 * 507 * Resume NAPI from being scheduled on this context. 508 * Must be paired with napi_disable. 509 */ 510 static inline void napi_enable(struct napi_struct *n) 511 { 512 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 513 smp_mb__before_atomic(); 514 clear_bit(NAPI_STATE_SCHED, &n->state); 515 clear_bit(NAPI_STATE_NPSVC, &n->state); 516 } 517 518 /** 519 * napi_synchronize - wait until NAPI is not running 520 * @n: NAPI context 521 * 522 * Wait until NAPI is done being scheduled on this context. 523 * Waits till any outstanding processing completes but 524 * does not disable future activations. 525 */ 526 static inline void napi_synchronize(const struct napi_struct *n) 527 { 528 if (IS_ENABLED(CONFIG_SMP)) 529 while (test_bit(NAPI_STATE_SCHED, &n->state)) 530 msleep(1); 531 else 532 barrier(); 533 } 534 535 enum netdev_queue_state_t { 536 __QUEUE_STATE_DRV_XOFF, 537 __QUEUE_STATE_STACK_XOFF, 538 __QUEUE_STATE_FROZEN, 539 }; 540 541 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 542 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 543 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 544 545 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 546 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 547 QUEUE_STATE_FROZEN) 548 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 549 QUEUE_STATE_FROZEN) 550 551 /* 552 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 553 * netif_tx_* functions below are used to manipulate this flag. The 554 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 555 * queue independently. The netif_xmit_*stopped functions below are called 556 * to check if the queue has been stopped by the driver or stack (either 557 * of the XOFF bits are set in the state). Drivers should not need to call 558 * netif_xmit*stopped functions, they should only be using netif_tx_*. 559 */ 560 561 struct netdev_queue { 562 /* 563 * read-mostly part 564 */ 565 struct net_device *dev; 566 struct Qdisc __rcu *qdisc; 567 struct Qdisc *qdisc_sleeping; 568 #ifdef CONFIG_SYSFS 569 struct kobject kobj; 570 #endif 571 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 572 int numa_node; 573 #endif 574 unsigned long tx_maxrate; 575 /* 576 * Number of TX timeouts for this queue 577 * (/sys/class/net/DEV/Q/trans_timeout) 578 */ 579 unsigned long trans_timeout; 580 /* 581 * write-mostly part 582 */ 583 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 584 int xmit_lock_owner; 585 /* 586 * Time (in jiffies) of last Tx 587 */ 588 unsigned long trans_start; 589 590 unsigned long state; 591 592 #ifdef CONFIG_BQL 593 struct dql dql; 594 #endif 595 } ____cacheline_aligned_in_smp; 596 597 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 598 { 599 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 600 return q->numa_node; 601 #else 602 return NUMA_NO_NODE; 603 #endif 604 } 605 606 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 607 { 608 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 609 q->numa_node = node; 610 #endif 611 } 612 613 #ifdef CONFIG_RPS 614 /* 615 * This structure holds an RPS map which can be of variable length. The 616 * map is an array of CPUs. 617 */ 618 struct rps_map { 619 unsigned int len; 620 struct rcu_head rcu; 621 u16 cpus[0]; 622 }; 623 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 624 625 /* 626 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 627 * tail pointer for that CPU's input queue at the time of last enqueue, and 628 * a hardware filter index. 629 */ 630 struct rps_dev_flow { 631 u16 cpu; 632 u16 filter; 633 unsigned int last_qtail; 634 }; 635 #define RPS_NO_FILTER 0xffff 636 637 /* 638 * The rps_dev_flow_table structure contains a table of flow mappings. 639 */ 640 struct rps_dev_flow_table { 641 unsigned int mask; 642 struct rcu_head rcu; 643 struct rps_dev_flow flows[0]; 644 }; 645 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 646 ((_num) * sizeof(struct rps_dev_flow))) 647 648 /* 649 * The rps_sock_flow_table contains mappings of flows to the last CPU 650 * on which they were processed by the application (set in recvmsg). 651 * Each entry is a 32bit value. Upper part is the high-order bits 652 * of flow hash, lower part is CPU number. 653 * rps_cpu_mask is used to partition the space, depending on number of 654 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 655 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 656 * meaning we use 32-6=26 bits for the hash. 657 */ 658 struct rps_sock_flow_table { 659 u32 mask; 660 661 u32 ents[0] ____cacheline_aligned_in_smp; 662 }; 663 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 664 665 #define RPS_NO_CPU 0xffff 666 667 extern u32 rps_cpu_mask; 668 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 669 670 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 671 u32 hash) 672 { 673 if (table && hash) { 674 unsigned int index = hash & table->mask; 675 u32 val = hash & ~rps_cpu_mask; 676 677 /* We only give a hint, preemption can change CPU under us */ 678 val |= raw_smp_processor_id(); 679 680 if (table->ents[index] != val) 681 table->ents[index] = val; 682 } 683 } 684 685 #ifdef CONFIG_RFS_ACCEL 686 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 687 u16 filter_id); 688 #endif 689 #endif /* CONFIG_RPS */ 690 691 /* This structure contains an instance of an RX queue. */ 692 struct netdev_rx_queue { 693 #ifdef CONFIG_RPS 694 struct rps_map __rcu *rps_map; 695 struct rps_dev_flow_table __rcu *rps_flow_table; 696 #endif 697 struct kobject kobj; 698 struct net_device *dev; 699 } ____cacheline_aligned_in_smp; 700 701 /* 702 * RX queue sysfs structures and functions. 703 */ 704 struct rx_queue_attribute { 705 struct attribute attr; 706 ssize_t (*show)(struct netdev_rx_queue *queue, 707 struct rx_queue_attribute *attr, char *buf); 708 ssize_t (*store)(struct netdev_rx_queue *queue, 709 struct rx_queue_attribute *attr, const char *buf, size_t len); 710 }; 711 712 #ifdef CONFIG_XPS 713 /* 714 * This structure holds an XPS map which can be of variable length. The 715 * map is an array of queues. 716 */ 717 struct xps_map { 718 unsigned int len; 719 unsigned int alloc_len; 720 struct rcu_head rcu; 721 u16 queues[0]; 722 }; 723 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 724 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 725 - sizeof(struct xps_map)) / sizeof(u16)) 726 727 /* 728 * This structure holds all XPS maps for device. Maps are indexed by CPU. 729 */ 730 struct xps_dev_maps { 731 struct rcu_head rcu; 732 struct xps_map __rcu *cpu_map[0]; 733 }; 734 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \ 735 (nr_cpu_ids * sizeof(struct xps_map *))) 736 #endif /* CONFIG_XPS */ 737 738 #define TC_MAX_QUEUE 16 739 #define TC_BITMASK 15 740 /* HW offloaded queuing disciplines txq count and offset maps */ 741 struct netdev_tc_txq { 742 u16 count; 743 u16 offset; 744 }; 745 746 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 747 /* 748 * This structure is to hold information about the device 749 * configured to run FCoE protocol stack. 750 */ 751 struct netdev_fcoe_hbainfo { 752 char manufacturer[64]; 753 char serial_number[64]; 754 char hardware_version[64]; 755 char driver_version[64]; 756 char optionrom_version[64]; 757 char firmware_version[64]; 758 char model[256]; 759 char model_description[256]; 760 }; 761 #endif 762 763 #define MAX_PHYS_ITEM_ID_LEN 32 764 765 /* This structure holds a unique identifier to identify some 766 * physical item (port for example) used by a netdevice. 767 */ 768 struct netdev_phys_item_id { 769 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 770 unsigned char id_len; 771 }; 772 773 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 774 struct netdev_phys_item_id *b) 775 { 776 return a->id_len == b->id_len && 777 memcmp(a->id, b->id, a->id_len) == 0; 778 } 779 780 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 781 struct sk_buff *skb); 782 783 /* These structures hold the attributes of qdisc and classifiers 784 * that are being passed to the netdevice through the setup_tc op. 785 */ 786 enum { 787 TC_SETUP_MQPRIO, 788 TC_SETUP_CLSU32, 789 TC_SETUP_CLSFLOWER, 790 TC_SETUP_MATCHALL, 791 }; 792 793 struct tc_cls_u32_offload; 794 795 struct tc_to_netdev { 796 unsigned int type; 797 union { 798 u8 tc; 799 struct tc_cls_u32_offload *cls_u32; 800 struct tc_cls_flower_offload *cls_flower; 801 struct tc_cls_matchall_offload *cls_mall; 802 }; 803 }; 804 805 /* These structures hold the attributes of xdp state that are being passed 806 * to the netdevice through the xdp op. 807 */ 808 enum xdp_netdev_command { 809 /* Set or clear a bpf program used in the earliest stages of packet 810 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 811 * is responsible for calling bpf_prog_put on any old progs that are 812 * stored. In case of error, the callee need not release the new prog 813 * reference, but on success it takes ownership and must bpf_prog_put 814 * when it is no longer used. 815 */ 816 XDP_SETUP_PROG, 817 /* Check if a bpf program is set on the device. The callee should 818 * return true if a program is currently attached and running. 819 */ 820 XDP_QUERY_PROG, 821 }; 822 823 struct netdev_xdp { 824 enum xdp_netdev_command command; 825 union { 826 /* XDP_SETUP_PROG */ 827 struct bpf_prog *prog; 828 /* XDP_QUERY_PROG */ 829 bool prog_attached; 830 }; 831 }; 832 833 /* 834 * This structure defines the management hooks for network devices. 835 * The following hooks can be defined; unless noted otherwise, they are 836 * optional and can be filled with a null pointer. 837 * 838 * int (*ndo_init)(struct net_device *dev); 839 * This function is called once when a network device is registered. 840 * The network device can use this for any late stage initialization 841 * or semantic validation. It can fail with an error code which will 842 * be propagated back to register_netdev. 843 * 844 * void (*ndo_uninit)(struct net_device *dev); 845 * This function is called when device is unregistered or when registration 846 * fails. It is not called if init fails. 847 * 848 * int (*ndo_open)(struct net_device *dev); 849 * This function is called when a network device transitions to the up 850 * state. 851 * 852 * int (*ndo_stop)(struct net_device *dev); 853 * This function is called when a network device transitions to the down 854 * state. 855 * 856 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 857 * struct net_device *dev); 858 * Called when a packet needs to be transmitted. 859 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 860 * the queue before that can happen; it's for obsolete devices and weird 861 * corner cases, but the stack really does a non-trivial amount 862 * of useless work if you return NETDEV_TX_BUSY. 863 * Required; cannot be NULL. 864 * 865 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 866 * netdev_features_t features); 867 * Adjusts the requested feature flags according to device-specific 868 * constraints, and returns the resulting flags. Must not modify 869 * the device state. 870 * 871 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 872 * void *accel_priv, select_queue_fallback_t fallback); 873 * Called to decide which queue to use when device supports multiple 874 * transmit queues. 875 * 876 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 877 * This function is called to allow device receiver to make 878 * changes to configuration when multicast or promiscuous is enabled. 879 * 880 * void (*ndo_set_rx_mode)(struct net_device *dev); 881 * This function is called device changes address list filtering. 882 * If driver handles unicast address filtering, it should set 883 * IFF_UNICAST_FLT in its priv_flags. 884 * 885 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 886 * This function is called when the Media Access Control address 887 * needs to be changed. If this interface is not defined, the 888 * MAC address can not be changed. 889 * 890 * int (*ndo_validate_addr)(struct net_device *dev); 891 * Test if Media Access Control address is valid for the device. 892 * 893 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 894 * Called when a user requests an ioctl which can't be handled by 895 * the generic interface code. If not defined ioctls return 896 * not supported error code. 897 * 898 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 899 * Used to set network devices bus interface parameters. This interface 900 * is retained for legacy reasons; new devices should use the bus 901 * interface (PCI) for low level management. 902 * 903 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 904 * Called when a user wants to change the Maximum Transfer Unit 905 * of a device. If not defined, any request to change MTU will 906 * will return an error. 907 * 908 * void (*ndo_tx_timeout)(struct net_device *dev); 909 * Callback used when the transmitter has not made any progress 910 * for dev->watchdog ticks. 911 * 912 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev, 913 * struct rtnl_link_stats64 *storage); 914 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 915 * Called when a user wants to get the network device usage 916 * statistics. Drivers must do one of the following: 917 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 918 * rtnl_link_stats64 structure passed by the caller. 919 * 2. Define @ndo_get_stats to update a net_device_stats structure 920 * (which should normally be dev->stats) and return a pointer to 921 * it. The structure may be changed asynchronously only if each 922 * field is written atomically. 923 * 3. Update dev->stats asynchronously and atomically, and define 924 * neither operation. 925 * 926 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 927 * If device supports VLAN filtering this function is called when a 928 * VLAN id is registered. 929 * 930 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 931 * If device supports VLAN filtering this function is called when a 932 * VLAN id is unregistered. 933 * 934 * void (*ndo_poll_controller)(struct net_device *dev); 935 * 936 * SR-IOV management functions. 937 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 938 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos); 939 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 940 * int max_tx_rate); 941 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 942 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 943 * int (*ndo_get_vf_config)(struct net_device *dev, 944 * int vf, struct ifla_vf_info *ivf); 945 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 946 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 947 * struct nlattr *port[]); 948 * 949 * Enable or disable the VF ability to query its RSS Redirection Table and 950 * Hash Key. This is needed since on some devices VF share this information 951 * with PF and querying it may introduce a theoretical security risk. 952 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 953 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 954 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc) 955 * Called to setup 'tc' number of traffic classes in the net device. This 956 * is always called from the stack with the rtnl lock held and netif tx 957 * queues stopped. This allows the netdevice to perform queue management 958 * safely. 959 * 960 * Fiber Channel over Ethernet (FCoE) offload functions. 961 * int (*ndo_fcoe_enable)(struct net_device *dev); 962 * Called when the FCoE protocol stack wants to start using LLD for FCoE 963 * so the underlying device can perform whatever needed configuration or 964 * initialization to support acceleration of FCoE traffic. 965 * 966 * int (*ndo_fcoe_disable)(struct net_device *dev); 967 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 968 * so the underlying device can perform whatever needed clean-ups to 969 * stop supporting acceleration of FCoE traffic. 970 * 971 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 972 * struct scatterlist *sgl, unsigned int sgc); 973 * Called when the FCoE Initiator wants to initialize an I/O that 974 * is a possible candidate for Direct Data Placement (DDP). The LLD can 975 * perform necessary setup and returns 1 to indicate the device is set up 976 * successfully to perform DDP on this I/O, otherwise this returns 0. 977 * 978 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 979 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 980 * indicated by the FC exchange id 'xid', so the underlying device can 981 * clean up and reuse resources for later DDP requests. 982 * 983 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 984 * struct scatterlist *sgl, unsigned int sgc); 985 * Called when the FCoE Target wants to initialize an I/O that 986 * is a possible candidate for Direct Data Placement (DDP). The LLD can 987 * perform necessary setup and returns 1 to indicate the device is set up 988 * successfully to perform DDP on this I/O, otherwise this returns 0. 989 * 990 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 991 * struct netdev_fcoe_hbainfo *hbainfo); 992 * Called when the FCoE Protocol stack wants information on the underlying 993 * device. This information is utilized by the FCoE protocol stack to 994 * register attributes with Fiber Channel management service as per the 995 * FC-GS Fabric Device Management Information(FDMI) specification. 996 * 997 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 998 * Called when the underlying device wants to override default World Wide 999 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1000 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1001 * protocol stack to use. 1002 * 1003 * RFS acceleration. 1004 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1005 * u16 rxq_index, u32 flow_id); 1006 * Set hardware filter for RFS. rxq_index is the target queue index; 1007 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1008 * Return the filter ID on success, or a negative error code. 1009 * 1010 * Slave management functions (for bridge, bonding, etc). 1011 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1012 * Called to make another netdev an underling. 1013 * 1014 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1015 * Called to release previously enslaved netdev. 1016 * 1017 * Feature/offload setting functions. 1018 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1019 * Called to update device configuration to new features. Passed 1020 * feature set might be less than what was returned by ndo_fix_features()). 1021 * Must return >0 or -errno if it changed dev->features itself. 1022 * 1023 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1024 * struct net_device *dev, 1025 * const unsigned char *addr, u16 vid, u16 flags) 1026 * Adds an FDB entry to dev for addr. 1027 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1028 * struct net_device *dev, 1029 * const unsigned char *addr, u16 vid) 1030 * Deletes the FDB entry from dev coresponding to addr. 1031 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1032 * struct net_device *dev, struct net_device *filter_dev, 1033 * int idx) 1034 * Used to add FDB entries to dump requests. Implementers should add 1035 * entries to skb and update idx with the number of entries. 1036 * 1037 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1038 * u16 flags) 1039 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1040 * struct net_device *dev, u32 filter_mask, 1041 * int nlflags) 1042 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1043 * u16 flags); 1044 * 1045 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1046 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1047 * which do not represent real hardware may define this to allow their 1048 * userspace components to manage their virtual carrier state. Devices 1049 * that determine carrier state from physical hardware properties (eg 1050 * network cables) or protocol-dependent mechanisms (eg 1051 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1052 * 1053 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1054 * struct netdev_phys_item_id *ppid); 1055 * Called to get ID of physical port of this device. If driver does 1056 * not implement this, it is assumed that the hw is not able to have 1057 * multiple net devices on single physical port. 1058 * 1059 * void (*ndo_udp_tunnel_add)(struct net_device *dev, 1060 * struct udp_tunnel_info *ti); 1061 * Called by UDP tunnel to notify a driver about the UDP port and socket 1062 * address family that a UDP tunnel is listnening to. It is called only 1063 * when a new port starts listening. The operation is protected by the 1064 * RTNL. 1065 * 1066 * void (*ndo_udp_tunnel_del)(struct net_device *dev, 1067 * struct udp_tunnel_info *ti); 1068 * Called by UDP tunnel to notify the driver about a UDP port and socket 1069 * address family that the UDP tunnel is not listening to anymore. The 1070 * operation is protected by the RTNL. 1071 * 1072 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1073 * struct net_device *dev) 1074 * Called by upper layer devices to accelerate switching or other 1075 * station functionality into hardware. 'pdev is the lowerdev 1076 * to use for the offload and 'dev' is the net device that will 1077 * back the offload. Returns a pointer to the private structure 1078 * the upper layer will maintain. 1079 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1080 * Called by upper layer device to delete the station created 1081 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1082 * the station and priv is the structure returned by the add 1083 * operation. 1084 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb, 1085 * struct net_device *dev, 1086 * void *priv); 1087 * Callback to use for xmit over the accelerated station. This 1088 * is used in place of ndo_start_xmit on accelerated net 1089 * devices. 1090 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1091 * struct net_device *dev 1092 * netdev_features_t features); 1093 * Called by core transmit path to determine if device is capable of 1094 * performing offload operations on a given packet. This is to give 1095 * the device an opportunity to implement any restrictions that cannot 1096 * be otherwise expressed by feature flags. The check is called with 1097 * the set of features that the stack has calculated and it returns 1098 * those the driver believes to be appropriate. 1099 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1100 * int queue_index, u32 maxrate); 1101 * Called when a user wants to set a max-rate limitation of specific 1102 * TX queue. 1103 * int (*ndo_get_iflink)(const struct net_device *dev); 1104 * Called to get the iflink value of this device. 1105 * void (*ndo_change_proto_down)(struct net_device *dev, 1106 * bool proto_down); 1107 * This function is used to pass protocol port error state information 1108 * to the switch driver. The switch driver can react to the proto_down 1109 * by doing a phys down on the associated switch port. 1110 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1111 * This function is used to get egress tunnel information for given skb. 1112 * This is useful for retrieving outer tunnel header parameters while 1113 * sampling packet. 1114 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1115 * This function is used to specify the headroom that the skb must 1116 * consider when allocation skb during packet reception. Setting 1117 * appropriate rx headroom value allows avoiding skb head copy on 1118 * forward. Setting a negative value resets the rx headroom to the 1119 * default value. 1120 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp); 1121 * This function is used to set or query state related to XDP on the 1122 * netdevice. See definition of enum xdp_netdev_command for details. 1123 * 1124 */ 1125 struct net_device_ops { 1126 int (*ndo_init)(struct net_device *dev); 1127 void (*ndo_uninit)(struct net_device *dev); 1128 int (*ndo_open)(struct net_device *dev); 1129 int (*ndo_stop)(struct net_device *dev); 1130 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1131 struct net_device *dev); 1132 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1133 struct net_device *dev, 1134 netdev_features_t features); 1135 u16 (*ndo_select_queue)(struct net_device *dev, 1136 struct sk_buff *skb, 1137 void *accel_priv, 1138 select_queue_fallback_t fallback); 1139 void (*ndo_change_rx_flags)(struct net_device *dev, 1140 int flags); 1141 void (*ndo_set_rx_mode)(struct net_device *dev); 1142 int (*ndo_set_mac_address)(struct net_device *dev, 1143 void *addr); 1144 int (*ndo_validate_addr)(struct net_device *dev); 1145 int (*ndo_do_ioctl)(struct net_device *dev, 1146 struct ifreq *ifr, int cmd); 1147 int (*ndo_set_config)(struct net_device *dev, 1148 struct ifmap *map); 1149 int (*ndo_change_mtu)(struct net_device *dev, 1150 int new_mtu); 1151 int (*ndo_neigh_setup)(struct net_device *dev, 1152 struct neigh_parms *); 1153 void (*ndo_tx_timeout) (struct net_device *dev); 1154 1155 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev, 1156 struct rtnl_link_stats64 *storage); 1157 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1158 1159 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1160 __be16 proto, u16 vid); 1161 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1162 __be16 proto, u16 vid); 1163 #ifdef CONFIG_NET_POLL_CONTROLLER 1164 void (*ndo_poll_controller)(struct net_device *dev); 1165 int (*ndo_netpoll_setup)(struct net_device *dev, 1166 struct netpoll_info *info); 1167 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1168 #endif 1169 #ifdef CONFIG_NET_RX_BUSY_POLL 1170 int (*ndo_busy_poll)(struct napi_struct *dev); 1171 #endif 1172 int (*ndo_set_vf_mac)(struct net_device *dev, 1173 int queue, u8 *mac); 1174 int (*ndo_set_vf_vlan)(struct net_device *dev, 1175 int queue, u16 vlan, u8 qos); 1176 int (*ndo_set_vf_rate)(struct net_device *dev, 1177 int vf, int min_tx_rate, 1178 int max_tx_rate); 1179 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1180 int vf, bool setting); 1181 int (*ndo_set_vf_trust)(struct net_device *dev, 1182 int vf, bool setting); 1183 int (*ndo_get_vf_config)(struct net_device *dev, 1184 int vf, 1185 struct ifla_vf_info *ivf); 1186 int (*ndo_set_vf_link_state)(struct net_device *dev, 1187 int vf, int link_state); 1188 int (*ndo_get_vf_stats)(struct net_device *dev, 1189 int vf, 1190 struct ifla_vf_stats 1191 *vf_stats); 1192 int (*ndo_set_vf_port)(struct net_device *dev, 1193 int vf, 1194 struct nlattr *port[]); 1195 int (*ndo_get_vf_port)(struct net_device *dev, 1196 int vf, struct sk_buff *skb); 1197 int (*ndo_set_vf_guid)(struct net_device *dev, 1198 int vf, u64 guid, 1199 int guid_type); 1200 int (*ndo_set_vf_rss_query_en)( 1201 struct net_device *dev, 1202 int vf, bool setting); 1203 int (*ndo_setup_tc)(struct net_device *dev, 1204 u32 handle, 1205 __be16 protocol, 1206 struct tc_to_netdev *tc); 1207 #if IS_ENABLED(CONFIG_FCOE) 1208 int (*ndo_fcoe_enable)(struct net_device *dev); 1209 int (*ndo_fcoe_disable)(struct net_device *dev); 1210 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1211 u16 xid, 1212 struct scatterlist *sgl, 1213 unsigned int sgc); 1214 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1215 u16 xid); 1216 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1217 u16 xid, 1218 struct scatterlist *sgl, 1219 unsigned int sgc); 1220 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1221 struct netdev_fcoe_hbainfo *hbainfo); 1222 #endif 1223 1224 #if IS_ENABLED(CONFIG_LIBFCOE) 1225 #define NETDEV_FCOE_WWNN 0 1226 #define NETDEV_FCOE_WWPN 1 1227 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1228 u64 *wwn, int type); 1229 #endif 1230 1231 #ifdef CONFIG_RFS_ACCEL 1232 int (*ndo_rx_flow_steer)(struct net_device *dev, 1233 const struct sk_buff *skb, 1234 u16 rxq_index, 1235 u32 flow_id); 1236 #endif 1237 int (*ndo_add_slave)(struct net_device *dev, 1238 struct net_device *slave_dev); 1239 int (*ndo_del_slave)(struct net_device *dev, 1240 struct net_device *slave_dev); 1241 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1242 netdev_features_t features); 1243 int (*ndo_set_features)(struct net_device *dev, 1244 netdev_features_t features); 1245 int (*ndo_neigh_construct)(struct net_device *dev, 1246 struct neighbour *n); 1247 void (*ndo_neigh_destroy)(struct net_device *dev, 1248 struct neighbour *n); 1249 1250 int (*ndo_fdb_add)(struct ndmsg *ndm, 1251 struct nlattr *tb[], 1252 struct net_device *dev, 1253 const unsigned char *addr, 1254 u16 vid, 1255 u16 flags); 1256 int (*ndo_fdb_del)(struct ndmsg *ndm, 1257 struct nlattr *tb[], 1258 struct net_device *dev, 1259 const unsigned char *addr, 1260 u16 vid); 1261 int (*ndo_fdb_dump)(struct sk_buff *skb, 1262 struct netlink_callback *cb, 1263 struct net_device *dev, 1264 struct net_device *filter_dev, 1265 int idx); 1266 1267 int (*ndo_bridge_setlink)(struct net_device *dev, 1268 struct nlmsghdr *nlh, 1269 u16 flags); 1270 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1271 u32 pid, u32 seq, 1272 struct net_device *dev, 1273 u32 filter_mask, 1274 int nlflags); 1275 int (*ndo_bridge_dellink)(struct net_device *dev, 1276 struct nlmsghdr *nlh, 1277 u16 flags); 1278 int (*ndo_change_carrier)(struct net_device *dev, 1279 bool new_carrier); 1280 int (*ndo_get_phys_port_id)(struct net_device *dev, 1281 struct netdev_phys_item_id *ppid); 1282 int (*ndo_get_phys_port_name)(struct net_device *dev, 1283 char *name, size_t len); 1284 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1285 struct udp_tunnel_info *ti); 1286 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1287 struct udp_tunnel_info *ti); 1288 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1289 struct net_device *dev); 1290 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1291 void *priv); 1292 1293 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb, 1294 struct net_device *dev, 1295 void *priv); 1296 int (*ndo_get_lock_subclass)(struct net_device *dev); 1297 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1298 int queue_index, 1299 u32 maxrate); 1300 int (*ndo_get_iflink)(const struct net_device *dev); 1301 int (*ndo_change_proto_down)(struct net_device *dev, 1302 bool proto_down); 1303 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1304 struct sk_buff *skb); 1305 void (*ndo_set_rx_headroom)(struct net_device *dev, 1306 int needed_headroom); 1307 int (*ndo_xdp)(struct net_device *dev, 1308 struct netdev_xdp *xdp); 1309 }; 1310 1311 /** 1312 * enum net_device_priv_flags - &struct net_device priv_flags 1313 * 1314 * These are the &struct net_device, they are only set internally 1315 * by drivers and used in the kernel. These flags are invisible to 1316 * userspace; this means that the order of these flags can change 1317 * during any kernel release. 1318 * 1319 * You should have a pretty good reason to be extending these flags. 1320 * 1321 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1322 * @IFF_EBRIDGE: Ethernet bridging device 1323 * @IFF_BONDING: bonding master or slave 1324 * @IFF_ISATAP: ISATAP interface (RFC4214) 1325 * @IFF_WAN_HDLC: WAN HDLC device 1326 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1327 * release skb->dst 1328 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1329 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1330 * @IFF_MACVLAN_PORT: device used as macvlan port 1331 * @IFF_BRIDGE_PORT: device used as bridge port 1332 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1333 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1334 * @IFF_UNICAST_FLT: Supports unicast filtering 1335 * @IFF_TEAM_PORT: device used as team port 1336 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1337 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1338 * change when it's running 1339 * @IFF_MACVLAN: Macvlan device 1340 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1341 * underlying stacked devices 1342 * @IFF_IPVLAN_MASTER: IPvlan master device 1343 * @IFF_IPVLAN_SLAVE: IPvlan slave device 1344 * @IFF_L3MDEV_MASTER: device is an L3 master device 1345 * @IFF_NO_QUEUE: device can run without qdisc attached 1346 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1347 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1348 * @IFF_TEAM: device is a team device 1349 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1350 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1351 * entity (i.e. the master device for bridged veth) 1352 * @IFF_MACSEC: device is a MACsec device 1353 */ 1354 enum netdev_priv_flags { 1355 IFF_802_1Q_VLAN = 1<<0, 1356 IFF_EBRIDGE = 1<<1, 1357 IFF_BONDING = 1<<2, 1358 IFF_ISATAP = 1<<3, 1359 IFF_WAN_HDLC = 1<<4, 1360 IFF_XMIT_DST_RELEASE = 1<<5, 1361 IFF_DONT_BRIDGE = 1<<6, 1362 IFF_DISABLE_NETPOLL = 1<<7, 1363 IFF_MACVLAN_PORT = 1<<8, 1364 IFF_BRIDGE_PORT = 1<<9, 1365 IFF_OVS_DATAPATH = 1<<10, 1366 IFF_TX_SKB_SHARING = 1<<11, 1367 IFF_UNICAST_FLT = 1<<12, 1368 IFF_TEAM_PORT = 1<<13, 1369 IFF_SUPP_NOFCS = 1<<14, 1370 IFF_LIVE_ADDR_CHANGE = 1<<15, 1371 IFF_MACVLAN = 1<<16, 1372 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1373 IFF_IPVLAN_MASTER = 1<<18, 1374 IFF_IPVLAN_SLAVE = 1<<19, 1375 IFF_L3MDEV_MASTER = 1<<20, 1376 IFF_NO_QUEUE = 1<<21, 1377 IFF_OPENVSWITCH = 1<<22, 1378 IFF_L3MDEV_SLAVE = 1<<23, 1379 IFF_TEAM = 1<<24, 1380 IFF_RXFH_CONFIGURED = 1<<25, 1381 IFF_PHONY_HEADROOM = 1<<26, 1382 IFF_MACSEC = 1<<27, 1383 }; 1384 1385 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1386 #define IFF_EBRIDGE IFF_EBRIDGE 1387 #define IFF_BONDING IFF_BONDING 1388 #define IFF_ISATAP IFF_ISATAP 1389 #define IFF_WAN_HDLC IFF_WAN_HDLC 1390 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1391 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1392 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1393 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1394 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1395 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1396 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1397 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1398 #define IFF_TEAM_PORT IFF_TEAM_PORT 1399 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1400 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1401 #define IFF_MACVLAN IFF_MACVLAN 1402 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1403 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER 1404 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE 1405 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1406 #define IFF_NO_QUEUE IFF_NO_QUEUE 1407 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1408 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1409 #define IFF_TEAM IFF_TEAM 1410 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1411 #define IFF_MACSEC IFF_MACSEC 1412 1413 /** 1414 * struct net_device - The DEVICE structure. 1415 * Actually, this whole structure is a big mistake. It mixes I/O 1416 * data with strictly "high-level" data, and it has to know about 1417 * almost every data structure used in the INET module. 1418 * 1419 * @name: This is the first field of the "visible" part of this structure 1420 * (i.e. as seen by users in the "Space.c" file). It is the name 1421 * of the interface. 1422 * 1423 * @name_hlist: Device name hash chain, please keep it close to name[] 1424 * @ifalias: SNMP alias 1425 * @mem_end: Shared memory end 1426 * @mem_start: Shared memory start 1427 * @base_addr: Device I/O address 1428 * @irq: Device IRQ number 1429 * 1430 * @carrier_changes: Stats to monitor carrier on<->off transitions 1431 * 1432 * @state: Generic network queuing layer state, see netdev_state_t 1433 * @dev_list: The global list of network devices 1434 * @napi_list: List entry used for polling NAPI devices 1435 * @unreg_list: List entry when we are unregistering the 1436 * device; see the function unregister_netdev 1437 * @close_list: List entry used when we are closing the device 1438 * @ptype_all: Device-specific packet handlers for all protocols 1439 * @ptype_specific: Device-specific, protocol-specific packet handlers 1440 * 1441 * @adj_list: Directly linked devices, like slaves for bonding 1442 * @all_adj_list: All linked devices, *including* neighbours 1443 * @features: Currently active device features 1444 * @hw_features: User-changeable features 1445 * 1446 * @wanted_features: User-requested features 1447 * @vlan_features: Mask of features inheritable by VLAN devices 1448 * 1449 * @hw_enc_features: Mask of features inherited by encapsulating devices 1450 * This field indicates what encapsulation 1451 * offloads the hardware is capable of doing, 1452 * and drivers will need to set them appropriately. 1453 * 1454 * @mpls_features: Mask of features inheritable by MPLS 1455 * 1456 * @ifindex: interface index 1457 * @group: The group the device belongs to 1458 * 1459 * @stats: Statistics struct, which was left as a legacy, use 1460 * rtnl_link_stats64 instead 1461 * 1462 * @rx_dropped: Dropped packets by core network, 1463 * do not use this in drivers 1464 * @tx_dropped: Dropped packets by core network, 1465 * do not use this in drivers 1466 * @rx_nohandler: nohandler dropped packets by core network on 1467 * inactive devices, do not use this in drivers 1468 * 1469 * @wireless_handlers: List of functions to handle Wireless Extensions, 1470 * instead of ioctl, 1471 * see <net/iw_handler.h> for details. 1472 * @wireless_data: Instance data managed by the core of wireless extensions 1473 * 1474 * @netdev_ops: Includes several pointers to callbacks, 1475 * if one wants to override the ndo_*() functions 1476 * @ethtool_ops: Management operations 1477 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1478 * discovery handling. Necessary for e.g. 6LoWPAN. 1479 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1480 * of Layer 2 headers. 1481 * 1482 * @flags: Interface flags (a la BSD) 1483 * @priv_flags: Like 'flags' but invisible to userspace, 1484 * see if.h for the definitions 1485 * @gflags: Global flags ( kept as legacy ) 1486 * @padded: How much padding added by alloc_netdev() 1487 * @operstate: RFC2863 operstate 1488 * @link_mode: Mapping policy to operstate 1489 * @if_port: Selectable AUI, TP, ... 1490 * @dma: DMA channel 1491 * @mtu: Interface MTU value 1492 * @type: Interface hardware type 1493 * @hard_header_len: Maximum hardware header length. 1494 * 1495 * @needed_headroom: Extra headroom the hardware may need, but not in all 1496 * cases can this be guaranteed 1497 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1498 * cases can this be guaranteed. Some cases also use 1499 * LL_MAX_HEADER instead to allocate the skb 1500 * 1501 * interface address info: 1502 * 1503 * @perm_addr: Permanent hw address 1504 * @addr_assign_type: Hw address assignment type 1505 * @addr_len: Hardware address length 1506 * @neigh_priv_len: Used in neigh_alloc() 1507 * @dev_id: Used to differentiate devices that share 1508 * the same link layer address 1509 * @dev_port: Used to differentiate devices that share 1510 * the same function 1511 * @addr_list_lock: XXX: need comments on this one 1512 * @uc_promisc: Counter that indicates promiscuous mode 1513 * has been enabled due to the need to listen to 1514 * additional unicast addresses in a device that 1515 * does not implement ndo_set_rx_mode() 1516 * @uc: unicast mac addresses 1517 * @mc: multicast mac addresses 1518 * @dev_addrs: list of device hw addresses 1519 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1520 * @promiscuity: Number of times the NIC is told to work in 1521 * promiscuous mode; if it becomes 0 the NIC will 1522 * exit promiscuous mode 1523 * @allmulti: Counter, enables or disables allmulticast mode 1524 * 1525 * @vlan_info: VLAN info 1526 * @dsa_ptr: dsa specific data 1527 * @tipc_ptr: TIPC specific data 1528 * @atalk_ptr: AppleTalk link 1529 * @ip_ptr: IPv4 specific data 1530 * @dn_ptr: DECnet specific data 1531 * @ip6_ptr: IPv6 specific data 1532 * @ax25_ptr: AX.25 specific data 1533 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1534 * 1535 * @last_rx: Time of last Rx 1536 * @dev_addr: Hw address (before bcast, 1537 * because most packets are unicast) 1538 * 1539 * @_rx: Array of RX queues 1540 * @num_rx_queues: Number of RX queues 1541 * allocated at register_netdev() time 1542 * @real_num_rx_queues: Number of RX queues currently active in device 1543 * 1544 * @rx_handler: handler for received packets 1545 * @rx_handler_data: XXX: need comments on this one 1546 * @ingress_queue: XXX: need comments on this one 1547 * @broadcast: hw bcast address 1548 * 1549 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1550 * indexed by RX queue number. Assigned by driver. 1551 * This must only be set if the ndo_rx_flow_steer 1552 * operation is defined 1553 * @index_hlist: Device index hash chain 1554 * 1555 * @_tx: Array of TX queues 1556 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1557 * @real_num_tx_queues: Number of TX queues currently active in device 1558 * @qdisc: Root qdisc from userspace point of view 1559 * @tx_queue_len: Max frames per queue allowed 1560 * @tx_global_lock: XXX: need comments on this one 1561 * 1562 * @xps_maps: XXX: need comments on this one 1563 * 1564 * @offload_fwd_mark: Offload device fwding mark 1565 * 1566 * @watchdog_timeo: Represents the timeout that is used by 1567 * the watchdog (see dev_watchdog()) 1568 * @watchdog_timer: List of timers 1569 * 1570 * @pcpu_refcnt: Number of references to this device 1571 * @todo_list: Delayed register/unregister 1572 * @link_watch_list: XXX: need comments on this one 1573 * 1574 * @reg_state: Register/unregister state machine 1575 * @dismantle: Device is going to be freed 1576 * @rtnl_link_state: This enum represents the phases of creating 1577 * a new link 1578 * 1579 * @destructor: Called from unregister, 1580 * can be used to call free_netdev 1581 * @npinfo: XXX: need comments on this one 1582 * @nd_net: Network namespace this network device is inside 1583 * 1584 * @ml_priv: Mid-layer private 1585 * @lstats: Loopback statistics 1586 * @tstats: Tunnel statistics 1587 * @dstats: Dummy statistics 1588 * @vstats: Virtual ethernet statistics 1589 * 1590 * @garp_port: GARP 1591 * @mrp_port: MRP 1592 * 1593 * @dev: Class/net/name entry 1594 * @sysfs_groups: Space for optional device, statistics and wireless 1595 * sysfs groups 1596 * 1597 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1598 * @rtnl_link_ops: Rtnl_link_ops 1599 * 1600 * @gso_max_size: Maximum size of generic segmentation offload 1601 * @gso_max_segs: Maximum number of segments that can be passed to the 1602 * NIC for GSO 1603 * 1604 * @dcbnl_ops: Data Center Bridging netlink ops 1605 * @num_tc: Number of traffic classes in the net device 1606 * @tc_to_txq: XXX: need comments on this one 1607 * @prio_tc_map XXX: need comments on this one 1608 * 1609 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1610 * 1611 * @priomap: XXX: need comments on this one 1612 * @phydev: Physical device may attach itself 1613 * for hardware timestamping 1614 * 1615 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1616 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1617 * 1618 * @proto_down: protocol port state information can be sent to the 1619 * switch driver and used to set the phys state of the 1620 * switch port. 1621 * 1622 * FIXME: cleanup struct net_device such that network protocol info 1623 * moves out. 1624 */ 1625 1626 struct net_device { 1627 char name[IFNAMSIZ]; 1628 struct hlist_node name_hlist; 1629 char *ifalias; 1630 /* 1631 * I/O specific fields 1632 * FIXME: Merge these and struct ifmap into one 1633 */ 1634 unsigned long mem_end; 1635 unsigned long mem_start; 1636 unsigned long base_addr; 1637 int irq; 1638 1639 atomic_t carrier_changes; 1640 1641 /* 1642 * Some hardware also needs these fields (state,dev_list, 1643 * napi_list,unreg_list,close_list) but they are not 1644 * part of the usual set specified in Space.c. 1645 */ 1646 1647 unsigned long state; 1648 1649 struct list_head dev_list; 1650 struct list_head napi_list; 1651 struct list_head unreg_list; 1652 struct list_head close_list; 1653 struct list_head ptype_all; 1654 struct list_head ptype_specific; 1655 1656 struct { 1657 struct list_head upper; 1658 struct list_head lower; 1659 } adj_list; 1660 1661 struct { 1662 struct list_head upper; 1663 struct list_head lower; 1664 } all_adj_list; 1665 1666 netdev_features_t features; 1667 netdev_features_t hw_features; 1668 netdev_features_t wanted_features; 1669 netdev_features_t vlan_features; 1670 netdev_features_t hw_enc_features; 1671 netdev_features_t mpls_features; 1672 netdev_features_t gso_partial_features; 1673 1674 int ifindex; 1675 int group; 1676 1677 struct net_device_stats stats; 1678 1679 atomic_long_t rx_dropped; 1680 atomic_long_t tx_dropped; 1681 atomic_long_t rx_nohandler; 1682 1683 #ifdef CONFIG_WIRELESS_EXT 1684 const struct iw_handler_def *wireless_handlers; 1685 struct iw_public_data *wireless_data; 1686 #endif 1687 const struct net_device_ops *netdev_ops; 1688 const struct ethtool_ops *ethtool_ops; 1689 #ifdef CONFIG_NET_SWITCHDEV 1690 const struct switchdev_ops *switchdev_ops; 1691 #endif 1692 #ifdef CONFIG_NET_L3_MASTER_DEV 1693 const struct l3mdev_ops *l3mdev_ops; 1694 #endif 1695 #if IS_ENABLED(CONFIG_IPV6) 1696 const struct ndisc_ops *ndisc_ops; 1697 #endif 1698 1699 const struct header_ops *header_ops; 1700 1701 unsigned int flags; 1702 unsigned int priv_flags; 1703 1704 unsigned short gflags; 1705 unsigned short padded; 1706 1707 unsigned char operstate; 1708 unsigned char link_mode; 1709 1710 unsigned char if_port; 1711 unsigned char dma; 1712 1713 unsigned int mtu; 1714 unsigned short type; 1715 unsigned short hard_header_len; 1716 1717 unsigned short needed_headroom; 1718 unsigned short needed_tailroom; 1719 1720 /* Interface address info. */ 1721 unsigned char perm_addr[MAX_ADDR_LEN]; 1722 unsigned char addr_assign_type; 1723 unsigned char addr_len; 1724 unsigned short neigh_priv_len; 1725 unsigned short dev_id; 1726 unsigned short dev_port; 1727 spinlock_t addr_list_lock; 1728 unsigned char name_assign_type; 1729 bool uc_promisc; 1730 struct netdev_hw_addr_list uc; 1731 struct netdev_hw_addr_list mc; 1732 struct netdev_hw_addr_list dev_addrs; 1733 1734 #ifdef CONFIG_SYSFS 1735 struct kset *queues_kset; 1736 #endif 1737 unsigned int promiscuity; 1738 unsigned int allmulti; 1739 1740 1741 /* Protocol-specific pointers */ 1742 1743 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1744 struct vlan_info __rcu *vlan_info; 1745 #endif 1746 #if IS_ENABLED(CONFIG_NET_DSA) 1747 struct dsa_switch_tree *dsa_ptr; 1748 #endif 1749 #if IS_ENABLED(CONFIG_TIPC) 1750 struct tipc_bearer __rcu *tipc_ptr; 1751 #endif 1752 void *atalk_ptr; 1753 struct in_device __rcu *ip_ptr; 1754 struct dn_dev __rcu *dn_ptr; 1755 struct inet6_dev __rcu *ip6_ptr; 1756 void *ax25_ptr; 1757 struct wireless_dev *ieee80211_ptr; 1758 struct wpan_dev *ieee802154_ptr; 1759 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 1760 struct mpls_dev __rcu *mpls_ptr; 1761 #endif 1762 1763 /* 1764 * Cache lines mostly used on receive path (including eth_type_trans()) 1765 */ 1766 unsigned long last_rx; 1767 1768 /* Interface address info used in eth_type_trans() */ 1769 unsigned char *dev_addr; 1770 1771 #ifdef CONFIG_SYSFS 1772 struct netdev_rx_queue *_rx; 1773 1774 unsigned int num_rx_queues; 1775 unsigned int real_num_rx_queues; 1776 #endif 1777 1778 unsigned long gro_flush_timeout; 1779 rx_handler_func_t __rcu *rx_handler; 1780 void __rcu *rx_handler_data; 1781 1782 #ifdef CONFIG_NET_CLS_ACT 1783 struct tcf_proto __rcu *ingress_cl_list; 1784 #endif 1785 struct netdev_queue __rcu *ingress_queue; 1786 #ifdef CONFIG_NETFILTER_INGRESS 1787 struct list_head nf_hooks_ingress; 1788 #endif 1789 1790 unsigned char broadcast[MAX_ADDR_LEN]; 1791 #ifdef CONFIG_RFS_ACCEL 1792 struct cpu_rmap *rx_cpu_rmap; 1793 #endif 1794 struct hlist_node index_hlist; 1795 1796 /* 1797 * Cache lines mostly used on transmit path 1798 */ 1799 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1800 unsigned int num_tx_queues; 1801 unsigned int real_num_tx_queues; 1802 struct Qdisc *qdisc; 1803 unsigned long tx_queue_len; 1804 spinlock_t tx_global_lock; 1805 int watchdog_timeo; 1806 1807 #ifdef CONFIG_XPS 1808 struct xps_dev_maps __rcu *xps_maps; 1809 #endif 1810 #ifdef CONFIG_NET_CLS_ACT 1811 struct tcf_proto __rcu *egress_cl_list; 1812 #endif 1813 #ifdef CONFIG_NET_SWITCHDEV 1814 u32 offload_fwd_mark; 1815 #endif 1816 1817 /* These may be needed for future network-power-down code. */ 1818 struct timer_list watchdog_timer; 1819 1820 int __percpu *pcpu_refcnt; 1821 struct list_head todo_list; 1822 1823 struct list_head link_watch_list; 1824 1825 enum { NETREG_UNINITIALIZED=0, 1826 NETREG_REGISTERED, /* completed register_netdevice */ 1827 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1828 NETREG_UNREGISTERED, /* completed unregister todo */ 1829 NETREG_RELEASED, /* called free_netdev */ 1830 NETREG_DUMMY, /* dummy device for NAPI poll */ 1831 } reg_state:8; 1832 1833 bool dismantle; 1834 1835 enum { 1836 RTNL_LINK_INITIALIZED, 1837 RTNL_LINK_INITIALIZING, 1838 } rtnl_link_state:16; 1839 1840 void (*destructor)(struct net_device *dev); 1841 1842 #ifdef CONFIG_NETPOLL 1843 struct netpoll_info __rcu *npinfo; 1844 #endif 1845 1846 possible_net_t nd_net; 1847 1848 /* mid-layer private */ 1849 union { 1850 void *ml_priv; 1851 struct pcpu_lstats __percpu *lstats; 1852 struct pcpu_sw_netstats __percpu *tstats; 1853 struct pcpu_dstats __percpu *dstats; 1854 struct pcpu_vstats __percpu *vstats; 1855 }; 1856 1857 struct garp_port __rcu *garp_port; 1858 struct mrp_port __rcu *mrp_port; 1859 1860 struct device dev; 1861 const struct attribute_group *sysfs_groups[4]; 1862 const struct attribute_group *sysfs_rx_queue_group; 1863 1864 const struct rtnl_link_ops *rtnl_link_ops; 1865 1866 /* for setting kernel sock attribute on TCP connection setup */ 1867 #define GSO_MAX_SIZE 65536 1868 unsigned int gso_max_size; 1869 #define GSO_MAX_SEGS 65535 1870 u16 gso_max_segs; 1871 1872 #ifdef CONFIG_DCB 1873 const struct dcbnl_rtnl_ops *dcbnl_ops; 1874 #endif 1875 u8 num_tc; 1876 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 1877 u8 prio_tc_map[TC_BITMASK + 1]; 1878 1879 #if IS_ENABLED(CONFIG_FCOE) 1880 unsigned int fcoe_ddp_xid; 1881 #endif 1882 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 1883 struct netprio_map __rcu *priomap; 1884 #endif 1885 struct phy_device *phydev; 1886 struct lock_class_key *qdisc_tx_busylock; 1887 struct lock_class_key *qdisc_running_key; 1888 bool proto_down; 1889 }; 1890 #define to_net_dev(d) container_of(d, struct net_device, dev) 1891 1892 #define NETDEV_ALIGN 32 1893 1894 static inline 1895 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 1896 { 1897 return dev->prio_tc_map[prio & TC_BITMASK]; 1898 } 1899 1900 static inline 1901 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 1902 { 1903 if (tc >= dev->num_tc) 1904 return -EINVAL; 1905 1906 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 1907 return 0; 1908 } 1909 1910 static inline 1911 void netdev_reset_tc(struct net_device *dev) 1912 { 1913 dev->num_tc = 0; 1914 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq)); 1915 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map)); 1916 } 1917 1918 static inline 1919 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset) 1920 { 1921 if (tc >= dev->num_tc) 1922 return -EINVAL; 1923 1924 dev->tc_to_txq[tc].count = count; 1925 dev->tc_to_txq[tc].offset = offset; 1926 return 0; 1927 } 1928 1929 static inline 1930 int netdev_set_num_tc(struct net_device *dev, u8 num_tc) 1931 { 1932 if (num_tc > TC_MAX_QUEUE) 1933 return -EINVAL; 1934 1935 dev->num_tc = num_tc; 1936 return 0; 1937 } 1938 1939 static inline 1940 int netdev_get_num_tc(struct net_device *dev) 1941 { 1942 return dev->num_tc; 1943 } 1944 1945 static inline 1946 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 1947 unsigned int index) 1948 { 1949 return &dev->_tx[index]; 1950 } 1951 1952 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 1953 const struct sk_buff *skb) 1954 { 1955 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 1956 } 1957 1958 static inline void netdev_for_each_tx_queue(struct net_device *dev, 1959 void (*f)(struct net_device *, 1960 struct netdev_queue *, 1961 void *), 1962 void *arg) 1963 { 1964 unsigned int i; 1965 1966 for (i = 0; i < dev->num_tx_queues; i++) 1967 f(dev, &dev->_tx[i], arg); 1968 } 1969 1970 #define netdev_lockdep_set_classes(dev) \ 1971 { \ 1972 static struct lock_class_key qdisc_tx_busylock_key; \ 1973 static struct lock_class_key qdisc_running_key; \ 1974 static struct lock_class_key qdisc_xmit_lock_key; \ 1975 static struct lock_class_key dev_addr_list_lock_key; \ 1976 unsigned int i; \ 1977 \ 1978 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 1979 (dev)->qdisc_running_key = &qdisc_running_key; \ 1980 lockdep_set_class(&(dev)->addr_list_lock, \ 1981 &dev_addr_list_lock_key); \ 1982 for (i = 0; i < (dev)->num_tx_queues; i++) \ 1983 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 1984 &qdisc_xmit_lock_key); \ 1985 } 1986 1987 struct netdev_queue *netdev_pick_tx(struct net_device *dev, 1988 struct sk_buff *skb, 1989 void *accel_priv); 1990 1991 /* returns the headroom that the master device needs to take in account 1992 * when forwarding to this dev 1993 */ 1994 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 1995 { 1996 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 1997 } 1998 1999 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2000 { 2001 if (dev->netdev_ops->ndo_set_rx_headroom) 2002 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2003 } 2004 2005 /* set the device rx headroom to the dev's default */ 2006 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2007 { 2008 netdev_set_rx_headroom(dev, -1); 2009 } 2010 2011 /* 2012 * Net namespace inlines 2013 */ 2014 static inline 2015 struct net *dev_net(const struct net_device *dev) 2016 { 2017 return read_pnet(&dev->nd_net); 2018 } 2019 2020 static inline 2021 void dev_net_set(struct net_device *dev, struct net *net) 2022 { 2023 write_pnet(&dev->nd_net, net); 2024 } 2025 2026 static inline bool netdev_uses_dsa(struct net_device *dev) 2027 { 2028 #if IS_ENABLED(CONFIG_NET_DSA) 2029 if (dev->dsa_ptr != NULL) 2030 return dsa_uses_tagged_protocol(dev->dsa_ptr); 2031 #endif 2032 return false; 2033 } 2034 2035 /** 2036 * netdev_priv - access network device private data 2037 * @dev: network device 2038 * 2039 * Get network device private data 2040 */ 2041 static inline void *netdev_priv(const struct net_device *dev) 2042 { 2043 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2044 } 2045 2046 /* Set the sysfs physical device reference for the network logical device 2047 * if set prior to registration will cause a symlink during initialization. 2048 */ 2049 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2050 2051 /* Set the sysfs device type for the network logical device to allow 2052 * fine-grained identification of different network device types. For 2053 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2054 */ 2055 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2056 2057 /* Default NAPI poll() weight 2058 * Device drivers are strongly advised to not use bigger value 2059 */ 2060 #define NAPI_POLL_WEIGHT 64 2061 2062 /** 2063 * netif_napi_add - initialize a NAPI context 2064 * @dev: network device 2065 * @napi: NAPI context 2066 * @poll: polling function 2067 * @weight: default weight 2068 * 2069 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2070 * *any* of the other NAPI-related functions. 2071 */ 2072 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2073 int (*poll)(struct napi_struct *, int), int weight); 2074 2075 /** 2076 * netif_tx_napi_add - initialize a NAPI context 2077 * @dev: network device 2078 * @napi: NAPI context 2079 * @poll: polling function 2080 * @weight: default weight 2081 * 2082 * This variant of netif_napi_add() should be used from drivers using NAPI 2083 * to exclusively poll a TX queue. 2084 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2085 */ 2086 static inline void netif_tx_napi_add(struct net_device *dev, 2087 struct napi_struct *napi, 2088 int (*poll)(struct napi_struct *, int), 2089 int weight) 2090 { 2091 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2092 netif_napi_add(dev, napi, poll, weight); 2093 } 2094 2095 /** 2096 * netif_napi_del - remove a NAPI context 2097 * @napi: NAPI context 2098 * 2099 * netif_napi_del() removes a NAPI context from the network device NAPI list 2100 */ 2101 void netif_napi_del(struct napi_struct *napi); 2102 2103 struct napi_gro_cb { 2104 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2105 void *frag0; 2106 2107 /* Length of frag0. */ 2108 unsigned int frag0_len; 2109 2110 /* This indicates where we are processing relative to skb->data. */ 2111 int data_offset; 2112 2113 /* This is non-zero if the packet cannot be merged with the new skb. */ 2114 u16 flush; 2115 2116 /* Save the IP ID here and check when we get to the transport layer */ 2117 u16 flush_id; 2118 2119 /* Number of segments aggregated. */ 2120 u16 count; 2121 2122 /* Start offset for remote checksum offload */ 2123 u16 gro_remcsum_start; 2124 2125 /* jiffies when first packet was created/queued */ 2126 unsigned long age; 2127 2128 /* Used in ipv6_gro_receive() and foo-over-udp */ 2129 u16 proto; 2130 2131 /* This is non-zero if the packet may be of the same flow. */ 2132 u8 same_flow:1; 2133 2134 /* Used in tunnel GRO receive */ 2135 u8 encap_mark:1; 2136 2137 /* GRO checksum is valid */ 2138 u8 csum_valid:1; 2139 2140 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2141 u8 csum_cnt:3; 2142 2143 /* Free the skb? */ 2144 u8 free:2; 2145 #define NAPI_GRO_FREE 1 2146 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2147 2148 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2149 u8 is_ipv6:1; 2150 2151 /* Used in GRE, set in fou/gue_gro_receive */ 2152 u8 is_fou:1; 2153 2154 /* Used to determine if flush_id can be ignored */ 2155 u8 is_atomic:1; 2156 2157 /* 5 bit hole */ 2158 2159 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2160 __wsum csum; 2161 2162 /* used in skb_gro_receive() slow path */ 2163 struct sk_buff *last; 2164 }; 2165 2166 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2167 2168 struct packet_type { 2169 __be16 type; /* This is really htons(ether_type). */ 2170 struct net_device *dev; /* NULL is wildcarded here */ 2171 int (*func) (struct sk_buff *, 2172 struct net_device *, 2173 struct packet_type *, 2174 struct net_device *); 2175 bool (*id_match)(struct packet_type *ptype, 2176 struct sock *sk); 2177 void *af_packet_priv; 2178 struct list_head list; 2179 }; 2180 2181 struct offload_callbacks { 2182 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2183 netdev_features_t features); 2184 struct sk_buff **(*gro_receive)(struct sk_buff **head, 2185 struct sk_buff *skb); 2186 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2187 }; 2188 2189 struct packet_offload { 2190 __be16 type; /* This is really htons(ether_type). */ 2191 u16 priority; 2192 struct offload_callbacks callbacks; 2193 struct list_head list; 2194 }; 2195 2196 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2197 struct pcpu_sw_netstats { 2198 u64 rx_packets; 2199 u64 rx_bytes; 2200 u64 tx_packets; 2201 u64 tx_bytes; 2202 struct u64_stats_sync syncp; 2203 }; 2204 2205 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2206 ({ \ 2207 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2208 if (pcpu_stats) { \ 2209 int __cpu; \ 2210 for_each_possible_cpu(__cpu) { \ 2211 typeof(type) *stat; \ 2212 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2213 u64_stats_init(&stat->syncp); \ 2214 } \ 2215 } \ 2216 pcpu_stats; \ 2217 }) 2218 2219 #define netdev_alloc_pcpu_stats(type) \ 2220 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2221 2222 enum netdev_lag_tx_type { 2223 NETDEV_LAG_TX_TYPE_UNKNOWN, 2224 NETDEV_LAG_TX_TYPE_RANDOM, 2225 NETDEV_LAG_TX_TYPE_BROADCAST, 2226 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2227 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2228 NETDEV_LAG_TX_TYPE_HASH, 2229 }; 2230 2231 struct netdev_lag_upper_info { 2232 enum netdev_lag_tx_type tx_type; 2233 }; 2234 2235 struct netdev_lag_lower_state_info { 2236 u8 link_up : 1, 2237 tx_enabled : 1; 2238 }; 2239 2240 #include <linux/notifier.h> 2241 2242 /* netdevice notifier chain. Please remember to update the rtnetlink 2243 * notification exclusion list in rtnetlink_event() when adding new 2244 * types. 2245 */ 2246 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */ 2247 #define NETDEV_DOWN 0x0002 2248 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface 2249 detected a hardware crash and restarted 2250 - we can use this eg to kick tcp sessions 2251 once done */ 2252 #define NETDEV_CHANGE 0x0004 /* Notify device state change */ 2253 #define NETDEV_REGISTER 0x0005 2254 #define NETDEV_UNREGISTER 0x0006 2255 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */ 2256 #define NETDEV_CHANGEADDR 0x0008 2257 #define NETDEV_GOING_DOWN 0x0009 2258 #define NETDEV_CHANGENAME 0x000A 2259 #define NETDEV_FEAT_CHANGE 0x000B 2260 #define NETDEV_BONDING_FAILOVER 0x000C 2261 #define NETDEV_PRE_UP 0x000D 2262 #define NETDEV_PRE_TYPE_CHANGE 0x000E 2263 #define NETDEV_POST_TYPE_CHANGE 0x000F 2264 #define NETDEV_POST_INIT 0x0010 2265 #define NETDEV_UNREGISTER_FINAL 0x0011 2266 #define NETDEV_RELEASE 0x0012 2267 #define NETDEV_NOTIFY_PEERS 0x0013 2268 #define NETDEV_JOIN 0x0014 2269 #define NETDEV_CHANGEUPPER 0x0015 2270 #define NETDEV_RESEND_IGMP 0x0016 2271 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */ 2272 #define NETDEV_CHANGEINFODATA 0x0018 2273 #define NETDEV_BONDING_INFO 0x0019 2274 #define NETDEV_PRECHANGEUPPER 0x001A 2275 #define NETDEV_CHANGELOWERSTATE 0x001B 2276 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C 2277 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E 2278 2279 int register_netdevice_notifier(struct notifier_block *nb); 2280 int unregister_netdevice_notifier(struct notifier_block *nb); 2281 2282 struct netdev_notifier_info { 2283 struct net_device *dev; 2284 }; 2285 2286 struct netdev_notifier_change_info { 2287 struct netdev_notifier_info info; /* must be first */ 2288 unsigned int flags_changed; 2289 }; 2290 2291 struct netdev_notifier_changeupper_info { 2292 struct netdev_notifier_info info; /* must be first */ 2293 struct net_device *upper_dev; /* new upper dev */ 2294 bool master; /* is upper dev master */ 2295 bool linking; /* is the notification for link or unlink */ 2296 void *upper_info; /* upper dev info */ 2297 }; 2298 2299 struct netdev_notifier_changelowerstate_info { 2300 struct netdev_notifier_info info; /* must be first */ 2301 void *lower_state_info; /* is lower dev state */ 2302 }; 2303 2304 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2305 struct net_device *dev) 2306 { 2307 info->dev = dev; 2308 } 2309 2310 static inline struct net_device * 2311 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2312 { 2313 return info->dev; 2314 } 2315 2316 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2317 2318 2319 extern rwlock_t dev_base_lock; /* Device list lock */ 2320 2321 #define for_each_netdev(net, d) \ 2322 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2323 #define for_each_netdev_reverse(net, d) \ 2324 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2325 #define for_each_netdev_rcu(net, d) \ 2326 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2327 #define for_each_netdev_safe(net, d, n) \ 2328 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2329 #define for_each_netdev_continue(net, d) \ 2330 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2331 #define for_each_netdev_continue_rcu(net, d) \ 2332 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2333 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2334 for_each_netdev_rcu(&init_net, slave) \ 2335 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2336 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2337 2338 static inline struct net_device *next_net_device(struct net_device *dev) 2339 { 2340 struct list_head *lh; 2341 struct net *net; 2342 2343 net = dev_net(dev); 2344 lh = dev->dev_list.next; 2345 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2346 } 2347 2348 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2349 { 2350 struct list_head *lh; 2351 struct net *net; 2352 2353 net = dev_net(dev); 2354 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2355 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2356 } 2357 2358 static inline struct net_device *first_net_device(struct net *net) 2359 { 2360 return list_empty(&net->dev_base_head) ? NULL : 2361 net_device_entry(net->dev_base_head.next); 2362 } 2363 2364 static inline struct net_device *first_net_device_rcu(struct net *net) 2365 { 2366 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2367 2368 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2369 } 2370 2371 int netdev_boot_setup_check(struct net_device *dev); 2372 unsigned long netdev_boot_base(const char *prefix, int unit); 2373 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2374 const char *hwaddr); 2375 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2376 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2377 void dev_add_pack(struct packet_type *pt); 2378 void dev_remove_pack(struct packet_type *pt); 2379 void __dev_remove_pack(struct packet_type *pt); 2380 void dev_add_offload(struct packet_offload *po); 2381 void dev_remove_offload(struct packet_offload *po); 2382 2383 int dev_get_iflink(const struct net_device *dev); 2384 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2385 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2386 unsigned short mask); 2387 struct net_device *dev_get_by_name(struct net *net, const char *name); 2388 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2389 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2390 int dev_alloc_name(struct net_device *dev, const char *name); 2391 int dev_open(struct net_device *dev); 2392 int dev_close(struct net_device *dev); 2393 int dev_close_many(struct list_head *head, bool unlink); 2394 void dev_disable_lro(struct net_device *dev); 2395 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2396 int dev_queue_xmit(struct sk_buff *skb); 2397 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv); 2398 int register_netdevice(struct net_device *dev); 2399 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2400 void unregister_netdevice_many(struct list_head *head); 2401 static inline void unregister_netdevice(struct net_device *dev) 2402 { 2403 unregister_netdevice_queue(dev, NULL); 2404 } 2405 2406 int netdev_refcnt_read(const struct net_device *dev); 2407 void free_netdev(struct net_device *dev); 2408 void netdev_freemem(struct net_device *dev); 2409 void synchronize_net(void); 2410 int init_dummy_netdev(struct net_device *dev); 2411 2412 DECLARE_PER_CPU(int, xmit_recursion); 2413 #define XMIT_RECURSION_LIMIT 10 2414 2415 static inline int dev_recursion_level(void) 2416 { 2417 return this_cpu_read(xmit_recursion); 2418 } 2419 2420 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2421 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2422 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2423 int netdev_get_name(struct net *net, char *name, int ifindex); 2424 int dev_restart(struct net_device *dev); 2425 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb); 2426 2427 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2428 { 2429 return NAPI_GRO_CB(skb)->data_offset; 2430 } 2431 2432 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2433 { 2434 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2435 } 2436 2437 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2438 { 2439 NAPI_GRO_CB(skb)->data_offset += len; 2440 } 2441 2442 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2443 unsigned int offset) 2444 { 2445 return NAPI_GRO_CB(skb)->frag0 + offset; 2446 } 2447 2448 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2449 { 2450 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2451 } 2452 2453 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2454 unsigned int offset) 2455 { 2456 if (!pskb_may_pull(skb, hlen)) 2457 return NULL; 2458 2459 NAPI_GRO_CB(skb)->frag0 = NULL; 2460 NAPI_GRO_CB(skb)->frag0_len = 0; 2461 return skb->data + offset; 2462 } 2463 2464 static inline void *skb_gro_network_header(struct sk_buff *skb) 2465 { 2466 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2467 skb_network_offset(skb); 2468 } 2469 2470 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2471 const void *start, unsigned int len) 2472 { 2473 if (NAPI_GRO_CB(skb)->csum_valid) 2474 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2475 csum_partial(start, len, 0)); 2476 } 2477 2478 /* GRO checksum functions. These are logical equivalents of the normal 2479 * checksum functions (in skbuff.h) except that they operate on the GRO 2480 * offsets and fields in sk_buff. 2481 */ 2482 2483 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2484 2485 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2486 { 2487 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2488 } 2489 2490 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2491 bool zero_okay, 2492 __sum16 check) 2493 { 2494 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2495 skb_checksum_start_offset(skb) < 2496 skb_gro_offset(skb)) && 2497 !skb_at_gro_remcsum_start(skb) && 2498 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2499 (!zero_okay || check)); 2500 } 2501 2502 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2503 __wsum psum) 2504 { 2505 if (NAPI_GRO_CB(skb)->csum_valid && 2506 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2507 return 0; 2508 2509 NAPI_GRO_CB(skb)->csum = psum; 2510 2511 return __skb_gro_checksum_complete(skb); 2512 } 2513 2514 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2515 { 2516 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2517 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2518 NAPI_GRO_CB(skb)->csum_cnt--; 2519 } else { 2520 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2521 * verified a new top level checksum or an encapsulated one 2522 * during GRO. This saves work if we fallback to normal path. 2523 */ 2524 __skb_incr_checksum_unnecessary(skb); 2525 } 2526 } 2527 2528 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2529 compute_pseudo) \ 2530 ({ \ 2531 __sum16 __ret = 0; \ 2532 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2533 __ret = __skb_gro_checksum_validate_complete(skb, \ 2534 compute_pseudo(skb, proto)); \ 2535 if (__ret) \ 2536 __skb_mark_checksum_bad(skb); \ 2537 else \ 2538 skb_gro_incr_csum_unnecessary(skb); \ 2539 __ret; \ 2540 }) 2541 2542 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2543 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2544 2545 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2546 compute_pseudo) \ 2547 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2548 2549 #define skb_gro_checksum_simple_validate(skb) \ 2550 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2551 2552 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2553 { 2554 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2555 !NAPI_GRO_CB(skb)->csum_valid); 2556 } 2557 2558 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2559 __sum16 check, __wsum pseudo) 2560 { 2561 NAPI_GRO_CB(skb)->csum = ~pseudo; 2562 NAPI_GRO_CB(skb)->csum_valid = 1; 2563 } 2564 2565 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \ 2566 do { \ 2567 if (__skb_gro_checksum_convert_check(skb)) \ 2568 __skb_gro_checksum_convert(skb, check, \ 2569 compute_pseudo(skb, proto)); \ 2570 } while (0) 2571 2572 struct gro_remcsum { 2573 int offset; 2574 __wsum delta; 2575 }; 2576 2577 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2578 { 2579 grc->offset = 0; 2580 grc->delta = 0; 2581 } 2582 2583 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2584 unsigned int off, size_t hdrlen, 2585 int start, int offset, 2586 struct gro_remcsum *grc, 2587 bool nopartial) 2588 { 2589 __wsum delta; 2590 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2591 2592 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2593 2594 if (!nopartial) { 2595 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2596 return ptr; 2597 } 2598 2599 ptr = skb_gro_header_fast(skb, off); 2600 if (skb_gro_header_hard(skb, off + plen)) { 2601 ptr = skb_gro_header_slow(skb, off + plen, off); 2602 if (!ptr) 2603 return NULL; 2604 } 2605 2606 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2607 start, offset); 2608 2609 /* Adjust skb->csum since we changed the packet */ 2610 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2611 2612 grc->offset = off + hdrlen + offset; 2613 grc->delta = delta; 2614 2615 return ptr; 2616 } 2617 2618 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2619 struct gro_remcsum *grc) 2620 { 2621 void *ptr; 2622 size_t plen = grc->offset + sizeof(u16); 2623 2624 if (!grc->delta) 2625 return; 2626 2627 ptr = skb_gro_header_fast(skb, grc->offset); 2628 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 2629 ptr = skb_gro_header_slow(skb, plen, grc->offset); 2630 if (!ptr) 2631 return; 2632 } 2633 2634 remcsum_unadjust((__sum16 *)ptr, grc->delta); 2635 } 2636 2637 struct skb_csum_offl_spec { 2638 __u16 ipv4_okay:1, 2639 ipv6_okay:1, 2640 encap_okay:1, 2641 ip_options_okay:1, 2642 ext_hdrs_okay:1, 2643 tcp_okay:1, 2644 udp_okay:1, 2645 sctp_okay:1, 2646 vlan_okay:1, 2647 no_encapped_ipv6:1, 2648 no_not_encapped:1; 2649 }; 2650 2651 bool __skb_csum_offload_chk(struct sk_buff *skb, 2652 const struct skb_csum_offl_spec *spec, 2653 bool *csum_encapped, 2654 bool csum_help); 2655 2656 static inline bool skb_csum_offload_chk(struct sk_buff *skb, 2657 const struct skb_csum_offl_spec *spec, 2658 bool *csum_encapped, 2659 bool csum_help) 2660 { 2661 if (skb->ip_summed != CHECKSUM_PARTIAL) 2662 return false; 2663 2664 return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help); 2665 } 2666 2667 static inline bool skb_csum_offload_chk_help(struct sk_buff *skb, 2668 const struct skb_csum_offl_spec *spec) 2669 { 2670 bool csum_encapped; 2671 2672 return skb_csum_offload_chk(skb, spec, &csum_encapped, true); 2673 } 2674 2675 static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb) 2676 { 2677 static const struct skb_csum_offl_spec csum_offl_spec = { 2678 .ipv4_okay = 1, 2679 .ip_options_okay = 1, 2680 .ipv6_okay = 1, 2681 .vlan_okay = 1, 2682 .tcp_okay = 1, 2683 .udp_okay = 1, 2684 }; 2685 2686 return skb_csum_offload_chk_help(skb, &csum_offl_spec); 2687 } 2688 2689 static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb) 2690 { 2691 static const struct skb_csum_offl_spec csum_offl_spec = { 2692 .ipv4_okay = 1, 2693 .ip_options_okay = 1, 2694 .tcp_okay = 1, 2695 .udp_okay = 1, 2696 .vlan_okay = 1, 2697 }; 2698 2699 return skb_csum_offload_chk_help(skb, &csum_offl_spec); 2700 } 2701 2702 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2703 unsigned short type, 2704 const void *daddr, const void *saddr, 2705 unsigned int len) 2706 { 2707 if (!dev->header_ops || !dev->header_ops->create) 2708 return 0; 2709 2710 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2711 } 2712 2713 static inline int dev_parse_header(const struct sk_buff *skb, 2714 unsigned char *haddr) 2715 { 2716 const struct net_device *dev = skb->dev; 2717 2718 if (!dev->header_ops || !dev->header_ops->parse) 2719 return 0; 2720 return dev->header_ops->parse(skb, haddr); 2721 } 2722 2723 /* ll_header must have at least hard_header_len allocated */ 2724 static inline bool dev_validate_header(const struct net_device *dev, 2725 char *ll_header, int len) 2726 { 2727 if (likely(len >= dev->hard_header_len)) 2728 return true; 2729 2730 if (capable(CAP_SYS_RAWIO)) { 2731 memset(ll_header + len, 0, dev->hard_header_len - len); 2732 return true; 2733 } 2734 2735 if (dev->header_ops && dev->header_ops->validate) 2736 return dev->header_ops->validate(ll_header, len); 2737 2738 return false; 2739 } 2740 2741 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len); 2742 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2743 static inline int unregister_gifconf(unsigned int family) 2744 { 2745 return register_gifconf(family, NULL); 2746 } 2747 2748 #ifdef CONFIG_NET_FLOW_LIMIT 2749 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2750 struct sd_flow_limit { 2751 u64 count; 2752 unsigned int num_buckets; 2753 unsigned int history_head; 2754 u16 history[FLOW_LIMIT_HISTORY]; 2755 u8 buckets[]; 2756 }; 2757 2758 extern int netdev_flow_limit_table_len; 2759 #endif /* CONFIG_NET_FLOW_LIMIT */ 2760 2761 /* 2762 * Incoming packets are placed on per-CPU queues 2763 */ 2764 struct softnet_data { 2765 struct list_head poll_list; 2766 struct sk_buff_head process_queue; 2767 2768 /* stats */ 2769 unsigned int processed; 2770 unsigned int time_squeeze; 2771 unsigned int received_rps; 2772 #ifdef CONFIG_RPS 2773 struct softnet_data *rps_ipi_list; 2774 #endif 2775 #ifdef CONFIG_NET_FLOW_LIMIT 2776 struct sd_flow_limit __rcu *flow_limit; 2777 #endif 2778 struct Qdisc *output_queue; 2779 struct Qdisc **output_queue_tailp; 2780 struct sk_buff *completion_queue; 2781 2782 #ifdef CONFIG_RPS 2783 /* input_queue_head should be written by cpu owning this struct, 2784 * and only read by other cpus. Worth using a cache line. 2785 */ 2786 unsigned int input_queue_head ____cacheline_aligned_in_smp; 2787 2788 /* Elements below can be accessed between CPUs for RPS/RFS */ 2789 struct call_single_data csd ____cacheline_aligned_in_smp; 2790 struct softnet_data *rps_ipi_next; 2791 unsigned int cpu; 2792 unsigned int input_queue_tail; 2793 #endif 2794 unsigned int dropped; 2795 struct sk_buff_head input_pkt_queue; 2796 struct napi_struct backlog; 2797 2798 }; 2799 2800 static inline void input_queue_head_incr(struct softnet_data *sd) 2801 { 2802 #ifdef CONFIG_RPS 2803 sd->input_queue_head++; 2804 #endif 2805 } 2806 2807 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 2808 unsigned int *qtail) 2809 { 2810 #ifdef CONFIG_RPS 2811 *qtail = ++sd->input_queue_tail; 2812 #endif 2813 } 2814 2815 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 2816 2817 void __netif_schedule(struct Qdisc *q); 2818 void netif_schedule_queue(struct netdev_queue *txq); 2819 2820 static inline void netif_tx_schedule_all(struct net_device *dev) 2821 { 2822 unsigned int i; 2823 2824 for (i = 0; i < dev->num_tx_queues; i++) 2825 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 2826 } 2827 2828 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 2829 { 2830 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2831 } 2832 2833 /** 2834 * netif_start_queue - allow transmit 2835 * @dev: network device 2836 * 2837 * Allow upper layers to call the device hard_start_xmit routine. 2838 */ 2839 static inline void netif_start_queue(struct net_device *dev) 2840 { 2841 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 2842 } 2843 2844 static inline void netif_tx_start_all_queues(struct net_device *dev) 2845 { 2846 unsigned int i; 2847 2848 for (i = 0; i < dev->num_tx_queues; i++) { 2849 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2850 netif_tx_start_queue(txq); 2851 } 2852 } 2853 2854 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 2855 2856 /** 2857 * netif_wake_queue - restart transmit 2858 * @dev: network device 2859 * 2860 * Allow upper layers to call the device hard_start_xmit routine. 2861 * Used for flow control when transmit resources are available. 2862 */ 2863 static inline void netif_wake_queue(struct net_device *dev) 2864 { 2865 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 2866 } 2867 2868 static inline void netif_tx_wake_all_queues(struct net_device *dev) 2869 { 2870 unsigned int i; 2871 2872 for (i = 0; i < dev->num_tx_queues; i++) { 2873 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2874 netif_tx_wake_queue(txq); 2875 } 2876 } 2877 2878 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 2879 { 2880 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2881 } 2882 2883 /** 2884 * netif_stop_queue - stop transmitted packets 2885 * @dev: network device 2886 * 2887 * Stop upper layers calling the device hard_start_xmit routine. 2888 * Used for flow control when transmit resources are unavailable. 2889 */ 2890 static inline void netif_stop_queue(struct net_device *dev) 2891 { 2892 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 2893 } 2894 2895 void netif_tx_stop_all_queues(struct net_device *dev); 2896 2897 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 2898 { 2899 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2900 } 2901 2902 /** 2903 * netif_queue_stopped - test if transmit queue is flowblocked 2904 * @dev: network device 2905 * 2906 * Test if transmit queue on device is currently unable to send. 2907 */ 2908 static inline bool netif_queue_stopped(const struct net_device *dev) 2909 { 2910 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 2911 } 2912 2913 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 2914 { 2915 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 2916 } 2917 2918 static inline bool 2919 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 2920 { 2921 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 2922 } 2923 2924 static inline bool 2925 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 2926 { 2927 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 2928 } 2929 2930 /** 2931 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 2932 * @dev_queue: pointer to transmit queue 2933 * 2934 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 2935 * to give appropriate hint to the CPU. 2936 */ 2937 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 2938 { 2939 #ifdef CONFIG_BQL 2940 prefetchw(&dev_queue->dql.num_queued); 2941 #endif 2942 } 2943 2944 /** 2945 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 2946 * @dev_queue: pointer to transmit queue 2947 * 2948 * BQL enabled drivers might use this helper in their TX completion path, 2949 * to give appropriate hint to the CPU. 2950 */ 2951 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 2952 { 2953 #ifdef CONFIG_BQL 2954 prefetchw(&dev_queue->dql.limit); 2955 #endif 2956 } 2957 2958 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 2959 unsigned int bytes) 2960 { 2961 #ifdef CONFIG_BQL 2962 dql_queued(&dev_queue->dql, bytes); 2963 2964 if (likely(dql_avail(&dev_queue->dql) >= 0)) 2965 return; 2966 2967 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2968 2969 /* 2970 * The XOFF flag must be set before checking the dql_avail below, 2971 * because in netdev_tx_completed_queue we update the dql_completed 2972 * before checking the XOFF flag. 2973 */ 2974 smp_mb(); 2975 2976 /* check again in case another CPU has just made room avail */ 2977 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 2978 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2979 #endif 2980 } 2981 2982 /** 2983 * netdev_sent_queue - report the number of bytes queued to hardware 2984 * @dev: network device 2985 * @bytes: number of bytes queued to the hardware device queue 2986 * 2987 * Report the number of bytes queued for sending/completion to the network 2988 * device hardware queue. @bytes should be a good approximation and should 2989 * exactly match netdev_completed_queue() @bytes 2990 */ 2991 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 2992 { 2993 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 2994 } 2995 2996 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 2997 unsigned int pkts, unsigned int bytes) 2998 { 2999 #ifdef CONFIG_BQL 3000 if (unlikely(!bytes)) 3001 return; 3002 3003 dql_completed(&dev_queue->dql, bytes); 3004 3005 /* 3006 * Without the memory barrier there is a small possiblity that 3007 * netdev_tx_sent_queue will miss the update and cause the queue to 3008 * be stopped forever 3009 */ 3010 smp_mb(); 3011 3012 if (dql_avail(&dev_queue->dql) < 0) 3013 return; 3014 3015 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3016 netif_schedule_queue(dev_queue); 3017 #endif 3018 } 3019 3020 /** 3021 * netdev_completed_queue - report bytes and packets completed by device 3022 * @dev: network device 3023 * @pkts: actual number of packets sent over the medium 3024 * @bytes: actual number of bytes sent over the medium 3025 * 3026 * Report the number of bytes and packets transmitted by the network device 3027 * hardware queue over the physical medium, @bytes must exactly match the 3028 * @bytes amount passed to netdev_sent_queue() 3029 */ 3030 static inline void netdev_completed_queue(struct net_device *dev, 3031 unsigned int pkts, unsigned int bytes) 3032 { 3033 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3034 } 3035 3036 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3037 { 3038 #ifdef CONFIG_BQL 3039 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3040 dql_reset(&q->dql); 3041 #endif 3042 } 3043 3044 /** 3045 * netdev_reset_queue - reset the packets and bytes count of a network device 3046 * @dev_queue: network device 3047 * 3048 * Reset the bytes and packet count of a network device and clear the 3049 * software flow control OFF bit for this network device 3050 */ 3051 static inline void netdev_reset_queue(struct net_device *dev_queue) 3052 { 3053 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3054 } 3055 3056 /** 3057 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3058 * @dev: network device 3059 * @queue_index: given tx queue index 3060 * 3061 * Returns 0 if given tx queue index >= number of device tx queues, 3062 * otherwise returns the originally passed tx queue index. 3063 */ 3064 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3065 { 3066 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3067 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3068 dev->name, queue_index, 3069 dev->real_num_tx_queues); 3070 return 0; 3071 } 3072 3073 return queue_index; 3074 } 3075 3076 /** 3077 * netif_running - test if up 3078 * @dev: network device 3079 * 3080 * Test if the device has been brought up. 3081 */ 3082 static inline bool netif_running(const struct net_device *dev) 3083 { 3084 return test_bit(__LINK_STATE_START, &dev->state); 3085 } 3086 3087 /* 3088 * Routines to manage the subqueues on a device. We only need start, 3089 * stop, and a check if it's stopped. All other device management is 3090 * done at the overall netdevice level. 3091 * Also test the device if we're multiqueue. 3092 */ 3093 3094 /** 3095 * netif_start_subqueue - allow sending packets on subqueue 3096 * @dev: network device 3097 * @queue_index: sub queue index 3098 * 3099 * Start individual transmit queue of a device with multiple transmit queues. 3100 */ 3101 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3102 { 3103 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3104 3105 netif_tx_start_queue(txq); 3106 } 3107 3108 /** 3109 * netif_stop_subqueue - stop sending packets on subqueue 3110 * @dev: network device 3111 * @queue_index: sub queue index 3112 * 3113 * Stop individual transmit queue of a device with multiple transmit queues. 3114 */ 3115 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3116 { 3117 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3118 netif_tx_stop_queue(txq); 3119 } 3120 3121 /** 3122 * netif_subqueue_stopped - test status of subqueue 3123 * @dev: network device 3124 * @queue_index: sub queue index 3125 * 3126 * Check individual transmit queue of a device with multiple transmit queues. 3127 */ 3128 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3129 u16 queue_index) 3130 { 3131 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3132 3133 return netif_tx_queue_stopped(txq); 3134 } 3135 3136 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3137 struct sk_buff *skb) 3138 { 3139 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3140 } 3141 3142 void netif_wake_subqueue(struct net_device *dev, u16 queue_index); 3143 3144 #ifdef CONFIG_XPS 3145 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3146 u16 index); 3147 #else 3148 static inline int netif_set_xps_queue(struct net_device *dev, 3149 const struct cpumask *mask, 3150 u16 index) 3151 { 3152 return 0; 3153 } 3154 #endif 3155 3156 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb, 3157 unsigned int num_tx_queues); 3158 3159 /* 3160 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used 3161 * as a distribution range limit for the returned value. 3162 */ 3163 static inline u16 skb_tx_hash(const struct net_device *dev, 3164 struct sk_buff *skb) 3165 { 3166 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues); 3167 } 3168 3169 /** 3170 * netif_is_multiqueue - test if device has multiple transmit queues 3171 * @dev: network device 3172 * 3173 * Check if device has multiple transmit queues 3174 */ 3175 static inline bool netif_is_multiqueue(const struct net_device *dev) 3176 { 3177 return dev->num_tx_queues > 1; 3178 } 3179 3180 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3181 3182 #ifdef CONFIG_SYSFS 3183 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3184 #else 3185 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3186 unsigned int rxq) 3187 { 3188 return 0; 3189 } 3190 #endif 3191 3192 #ifdef CONFIG_SYSFS 3193 static inline unsigned int get_netdev_rx_queue_index( 3194 struct netdev_rx_queue *queue) 3195 { 3196 struct net_device *dev = queue->dev; 3197 int index = queue - dev->_rx; 3198 3199 BUG_ON(index >= dev->num_rx_queues); 3200 return index; 3201 } 3202 #endif 3203 3204 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3205 int netif_get_num_default_rss_queues(void); 3206 3207 enum skb_free_reason { 3208 SKB_REASON_CONSUMED, 3209 SKB_REASON_DROPPED, 3210 }; 3211 3212 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3213 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3214 3215 /* 3216 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3217 * interrupt context or with hardware interrupts being disabled. 3218 * (in_irq() || irqs_disabled()) 3219 * 3220 * We provide four helpers that can be used in following contexts : 3221 * 3222 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3223 * replacing kfree_skb(skb) 3224 * 3225 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3226 * Typically used in place of consume_skb(skb) in TX completion path 3227 * 3228 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3229 * replacing kfree_skb(skb) 3230 * 3231 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3232 * and consumed a packet. Used in place of consume_skb(skb) 3233 */ 3234 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3235 { 3236 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3237 } 3238 3239 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3240 { 3241 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3242 } 3243 3244 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3245 { 3246 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3247 } 3248 3249 static inline void dev_consume_skb_any(struct sk_buff *skb) 3250 { 3251 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3252 } 3253 3254 int netif_rx(struct sk_buff *skb); 3255 int netif_rx_ni(struct sk_buff *skb); 3256 int netif_receive_skb(struct sk_buff *skb); 3257 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3258 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3259 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3260 gro_result_t napi_gro_frags(struct napi_struct *napi); 3261 struct packet_offload *gro_find_receive_by_type(__be16 type); 3262 struct packet_offload *gro_find_complete_by_type(__be16 type); 3263 3264 static inline void napi_free_frags(struct napi_struct *napi) 3265 { 3266 kfree_skb(napi->skb); 3267 napi->skb = NULL; 3268 } 3269 3270 int netdev_rx_handler_register(struct net_device *dev, 3271 rx_handler_func_t *rx_handler, 3272 void *rx_handler_data); 3273 void netdev_rx_handler_unregister(struct net_device *dev); 3274 3275 bool dev_valid_name(const char *name); 3276 int dev_ioctl(struct net *net, unsigned int cmd, void __user *); 3277 int dev_ethtool(struct net *net, struct ifreq *); 3278 unsigned int dev_get_flags(const struct net_device *); 3279 int __dev_change_flags(struct net_device *, unsigned int flags); 3280 int dev_change_flags(struct net_device *, unsigned int); 3281 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3282 unsigned int gchanges); 3283 int dev_change_name(struct net_device *, const char *); 3284 int dev_set_alias(struct net_device *, const char *, size_t); 3285 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3286 int dev_set_mtu(struct net_device *, int); 3287 void dev_set_group(struct net_device *, int); 3288 int dev_set_mac_address(struct net_device *, struct sockaddr *); 3289 int dev_change_carrier(struct net_device *, bool new_carrier); 3290 int dev_get_phys_port_id(struct net_device *dev, 3291 struct netdev_phys_item_id *ppid); 3292 int dev_get_phys_port_name(struct net_device *dev, 3293 char *name, size_t len); 3294 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3295 int dev_change_xdp_fd(struct net_device *dev, int fd); 3296 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev); 3297 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3298 struct netdev_queue *txq, int *ret); 3299 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3300 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3301 bool is_skb_forwardable(const struct net_device *dev, 3302 const struct sk_buff *skb); 3303 3304 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3305 3306 extern int netdev_budget; 3307 3308 /* Called by rtnetlink.c:rtnl_unlock() */ 3309 void netdev_run_todo(void); 3310 3311 /** 3312 * dev_put - release reference to device 3313 * @dev: network device 3314 * 3315 * Release reference to device to allow it to be freed. 3316 */ 3317 static inline void dev_put(struct net_device *dev) 3318 { 3319 this_cpu_dec(*dev->pcpu_refcnt); 3320 } 3321 3322 /** 3323 * dev_hold - get reference to device 3324 * @dev: network device 3325 * 3326 * Hold reference to device to keep it from being freed. 3327 */ 3328 static inline void dev_hold(struct net_device *dev) 3329 { 3330 this_cpu_inc(*dev->pcpu_refcnt); 3331 } 3332 3333 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3334 * and _off may be called from IRQ context, but it is caller 3335 * who is responsible for serialization of these calls. 3336 * 3337 * The name carrier is inappropriate, these functions should really be 3338 * called netif_lowerlayer_*() because they represent the state of any 3339 * kind of lower layer not just hardware media. 3340 */ 3341 3342 void linkwatch_init_dev(struct net_device *dev); 3343 void linkwatch_fire_event(struct net_device *dev); 3344 void linkwatch_forget_dev(struct net_device *dev); 3345 3346 /** 3347 * netif_carrier_ok - test if carrier present 3348 * @dev: network device 3349 * 3350 * Check if carrier is present on device 3351 */ 3352 static inline bool netif_carrier_ok(const struct net_device *dev) 3353 { 3354 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3355 } 3356 3357 unsigned long dev_trans_start(struct net_device *dev); 3358 3359 void __netdev_watchdog_up(struct net_device *dev); 3360 3361 void netif_carrier_on(struct net_device *dev); 3362 3363 void netif_carrier_off(struct net_device *dev); 3364 3365 /** 3366 * netif_dormant_on - mark device as dormant. 3367 * @dev: network device 3368 * 3369 * Mark device as dormant (as per RFC2863). 3370 * 3371 * The dormant state indicates that the relevant interface is not 3372 * actually in a condition to pass packets (i.e., it is not 'up') but is 3373 * in a "pending" state, waiting for some external event. For "on- 3374 * demand" interfaces, this new state identifies the situation where the 3375 * interface is waiting for events to place it in the up state. 3376 */ 3377 static inline void netif_dormant_on(struct net_device *dev) 3378 { 3379 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3380 linkwatch_fire_event(dev); 3381 } 3382 3383 /** 3384 * netif_dormant_off - set device as not dormant. 3385 * @dev: network device 3386 * 3387 * Device is not in dormant state. 3388 */ 3389 static inline void netif_dormant_off(struct net_device *dev) 3390 { 3391 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3392 linkwatch_fire_event(dev); 3393 } 3394 3395 /** 3396 * netif_dormant - test if carrier present 3397 * @dev: network device 3398 * 3399 * Check if carrier is present on device 3400 */ 3401 static inline bool netif_dormant(const struct net_device *dev) 3402 { 3403 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3404 } 3405 3406 3407 /** 3408 * netif_oper_up - test if device is operational 3409 * @dev: network device 3410 * 3411 * Check if carrier is operational 3412 */ 3413 static inline bool netif_oper_up(const struct net_device *dev) 3414 { 3415 return (dev->operstate == IF_OPER_UP || 3416 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3417 } 3418 3419 /** 3420 * netif_device_present - is device available or removed 3421 * @dev: network device 3422 * 3423 * Check if device has not been removed from system. 3424 */ 3425 static inline bool netif_device_present(struct net_device *dev) 3426 { 3427 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3428 } 3429 3430 void netif_device_detach(struct net_device *dev); 3431 3432 void netif_device_attach(struct net_device *dev); 3433 3434 /* 3435 * Network interface message level settings 3436 */ 3437 3438 enum { 3439 NETIF_MSG_DRV = 0x0001, 3440 NETIF_MSG_PROBE = 0x0002, 3441 NETIF_MSG_LINK = 0x0004, 3442 NETIF_MSG_TIMER = 0x0008, 3443 NETIF_MSG_IFDOWN = 0x0010, 3444 NETIF_MSG_IFUP = 0x0020, 3445 NETIF_MSG_RX_ERR = 0x0040, 3446 NETIF_MSG_TX_ERR = 0x0080, 3447 NETIF_MSG_TX_QUEUED = 0x0100, 3448 NETIF_MSG_INTR = 0x0200, 3449 NETIF_MSG_TX_DONE = 0x0400, 3450 NETIF_MSG_RX_STATUS = 0x0800, 3451 NETIF_MSG_PKTDATA = 0x1000, 3452 NETIF_MSG_HW = 0x2000, 3453 NETIF_MSG_WOL = 0x4000, 3454 }; 3455 3456 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3457 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3458 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3459 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3460 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3461 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3462 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3463 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3464 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3465 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3466 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3467 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3468 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3469 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3470 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3471 3472 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3473 { 3474 /* use default */ 3475 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3476 return default_msg_enable_bits; 3477 if (debug_value == 0) /* no output */ 3478 return 0; 3479 /* set low N bits */ 3480 return (1 << debug_value) - 1; 3481 } 3482 3483 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 3484 { 3485 spin_lock(&txq->_xmit_lock); 3486 txq->xmit_lock_owner = cpu; 3487 } 3488 3489 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 3490 { 3491 spin_lock_bh(&txq->_xmit_lock); 3492 txq->xmit_lock_owner = smp_processor_id(); 3493 } 3494 3495 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 3496 { 3497 bool ok = spin_trylock(&txq->_xmit_lock); 3498 if (likely(ok)) 3499 txq->xmit_lock_owner = smp_processor_id(); 3500 return ok; 3501 } 3502 3503 static inline void __netif_tx_unlock(struct netdev_queue *txq) 3504 { 3505 txq->xmit_lock_owner = -1; 3506 spin_unlock(&txq->_xmit_lock); 3507 } 3508 3509 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 3510 { 3511 txq->xmit_lock_owner = -1; 3512 spin_unlock_bh(&txq->_xmit_lock); 3513 } 3514 3515 static inline void txq_trans_update(struct netdev_queue *txq) 3516 { 3517 if (txq->xmit_lock_owner != -1) 3518 txq->trans_start = jiffies; 3519 } 3520 3521 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 3522 static inline void netif_trans_update(struct net_device *dev) 3523 { 3524 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 3525 3526 if (txq->trans_start != jiffies) 3527 txq->trans_start = jiffies; 3528 } 3529 3530 /** 3531 * netif_tx_lock - grab network device transmit lock 3532 * @dev: network device 3533 * 3534 * Get network device transmit lock 3535 */ 3536 static inline void netif_tx_lock(struct net_device *dev) 3537 { 3538 unsigned int i; 3539 int cpu; 3540 3541 spin_lock(&dev->tx_global_lock); 3542 cpu = smp_processor_id(); 3543 for (i = 0; i < dev->num_tx_queues; i++) { 3544 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3545 3546 /* We are the only thread of execution doing a 3547 * freeze, but we have to grab the _xmit_lock in 3548 * order to synchronize with threads which are in 3549 * the ->hard_start_xmit() handler and already 3550 * checked the frozen bit. 3551 */ 3552 __netif_tx_lock(txq, cpu); 3553 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 3554 __netif_tx_unlock(txq); 3555 } 3556 } 3557 3558 static inline void netif_tx_lock_bh(struct net_device *dev) 3559 { 3560 local_bh_disable(); 3561 netif_tx_lock(dev); 3562 } 3563 3564 static inline void netif_tx_unlock(struct net_device *dev) 3565 { 3566 unsigned int i; 3567 3568 for (i = 0; i < dev->num_tx_queues; i++) { 3569 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3570 3571 /* No need to grab the _xmit_lock here. If the 3572 * queue is not stopped for another reason, we 3573 * force a schedule. 3574 */ 3575 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 3576 netif_schedule_queue(txq); 3577 } 3578 spin_unlock(&dev->tx_global_lock); 3579 } 3580 3581 static inline void netif_tx_unlock_bh(struct net_device *dev) 3582 { 3583 netif_tx_unlock(dev); 3584 local_bh_enable(); 3585 } 3586 3587 #define HARD_TX_LOCK(dev, txq, cpu) { \ 3588 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3589 __netif_tx_lock(txq, cpu); \ 3590 } \ 3591 } 3592 3593 #define HARD_TX_TRYLOCK(dev, txq) \ 3594 (((dev->features & NETIF_F_LLTX) == 0) ? \ 3595 __netif_tx_trylock(txq) : \ 3596 true ) 3597 3598 #define HARD_TX_UNLOCK(dev, txq) { \ 3599 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3600 __netif_tx_unlock(txq); \ 3601 } \ 3602 } 3603 3604 static inline void netif_tx_disable(struct net_device *dev) 3605 { 3606 unsigned int i; 3607 int cpu; 3608 3609 local_bh_disable(); 3610 cpu = smp_processor_id(); 3611 for (i = 0; i < dev->num_tx_queues; i++) { 3612 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3613 3614 __netif_tx_lock(txq, cpu); 3615 netif_tx_stop_queue(txq); 3616 __netif_tx_unlock(txq); 3617 } 3618 local_bh_enable(); 3619 } 3620 3621 static inline void netif_addr_lock(struct net_device *dev) 3622 { 3623 spin_lock(&dev->addr_list_lock); 3624 } 3625 3626 static inline void netif_addr_lock_nested(struct net_device *dev) 3627 { 3628 int subclass = SINGLE_DEPTH_NESTING; 3629 3630 if (dev->netdev_ops->ndo_get_lock_subclass) 3631 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev); 3632 3633 spin_lock_nested(&dev->addr_list_lock, subclass); 3634 } 3635 3636 static inline void netif_addr_lock_bh(struct net_device *dev) 3637 { 3638 spin_lock_bh(&dev->addr_list_lock); 3639 } 3640 3641 static inline void netif_addr_unlock(struct net_device *dev) 3642 { 3643 spin_unlock(&dev->addr_list_lock); 3644 } 3645 3646 static inline void netif_addr_unlock_bh(struct net_device *dev) 3647 { 3648 spin_unlock_bh(&dev->addr_list_lock); 3649 } 3650 3651 /* 3652 * dev_addrs walker. Should be used only for read access. Call with 3653 * rcu_read_lock held. 3654 */ 3655 #define for_each_dev_addr(dev, ha) \ 3656 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 3657 3658 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 3659 3660 void ether_setup(struct net_device *dev); 3661 3662 /* Support for loadable net-drivers */ 3663 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 3664 unsigned char name_assign_type, 3665 void (*setup)(struct net_device *), 3666 unsigned int txqs, unsigned int rxqs); 3667 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 3668 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 3669 3670 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 3671 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 3672 count) 3673 3674 int register_netdev(struct net_device *dev); 3675 void unregister_netdev(struct net_device *dev); 3676 3677 /* General hardware address lists handling functions */ 3678 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 3679 struct netdev_hw_addr_list *from_list, int addr_len); 3680 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 3681 struct netdev_hw_addr_list *from_list, int addr_len); 3682 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 3683 struct net_device *dev, 3684 int (*sync)(struct net_device *, const unsigned char *), 3685 int (*unsync)(struct net_device *, 3686 const unsigned char *)); 3687 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 3688 struct net_device *dev, 3689 int (*unsync)(struct net_device *, 3690 const unsigned char *)); 3691 void __hw_addr_init(struct netdev_hw_addr_list *list); 3692 3693 /* Functions used for device addresses handling */ 3694 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 3695 unsigned char addr_type); 3696 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 3697 unsigned char addr_type); 3698 void dev_addr_flush(struct net_device *dev); 3699 int dev_addr_init(struct net_device *dev); 3700 3701 /* Functions used for unicast addresses handling */ 3702 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 3703 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 3704 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 3705 int dev_uc_sync(struct net_device *to, struct net_device *from); 3706 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 3707 void dev_uc_unsync(struct net_device *to, struct net_device *from); 3708 void dev_uc_flush(struct net_device *dev); 3709 void dev_uc_init(struct net_device *dev); 3710 3711 /** 3712 * __dev_uc_sync - Synchonize device's unicast list 3713 * @dev: device to sync 3714 * @sync: function to call if address should be added 3715 * @unsync: function to call if address should be removed 3716 * 3717 * Add newly added addresses to the interface, and release 3718 * addresses that have been deleted. 3719 */ 3720 static inline int __dev_uc_sync(struct net_device *dev, 3721 int (*sync)(struct net_device *, 3722 const unsigned char *), 3723 int (*unsync)(struct net_device *, 3724 const unsigned char *)) 3725 { 3726 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 3727 } 3728 3729 /** 3730 * __dev_uc_unsync - Remove synchronized addresses from device 3731 * @dev: device to sync 3732 * @unsync: function to call if address should be removed 3733 * 3734 * Remove all addresses that were added to the device by dev_uc_sync(). 3735 */ 3736 static inline void __dev_uc_unsync(struct net_device *dev, 3737 int (*unsync)(struct net_device *, 3738 const unsigned char *)) 3739 { 3740 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 3741 } 3742 3743 /* Functions used for multicast addresses handling */ 3744 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 3745 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 3746 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 3747 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 3748 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 3749 int dev_mc_sync(struct net_device *to, struct net_device *from); 3750 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 3751 void dev_mc_unsync(struct net_device *to, struct net_device *from); 3752 void dev_mc_flush(struct net_device *dev); 3753 void dev_mc_init(struct net_device *dev); 3754 3755 /** 3756 * __dev_mc_sync - Synchonize device's multicast list 3757 * @dev: device to sync 3758 * @sync: function to call if address should be added 3759 * @unsync: function to call if address should be removed 3760 * 3761 * Add newly added addresses to the interface, and release 3762 * addresses that have been deleted. 3763 */ 3764 static inline int __dev_mc_sync(struct net_device *dev, 3765 int (*sync)(struct net_device *, 3766 const unsigned char *), 3767 int (*unsync)(struct net_device *, 3768 const unsigned char *)) 3769 { 3770 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 3771 } 3772 3773 /** 3774 * __dev_mc_unsync - Remove synchronized addresses from device 3775 * @dev: device to sync 3776 * @unsync: function to call if address should be removed 3777 * 3778 * Remove all addresses that were added to the device by dev_mc_sync(). 3779 */ 3780 static inline void __dev_mc_unsync(struct net_device *dev, 3781 int (*unsync)(struct net_device *, 3782 const unsigned char *)) 3783 { 3784 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 3785 } 3786 3787 /* Functions used for secondary unicast and multicast support */ 3788 void dev_set_rx_mode(struct net_device *dev); 3789 void __dev_set_rx_mode(struct net_device *dev); 3790 int dev_set_promiscuity(struct net_device *dev, int inc); 3791 int dev_set_allmulti(struct net_device *dev, int inc); 3792 void netdev_state_change(struct net_device *dev); 3793 void netdev_notify_peers(struct net_device *dev); 3794 void netdev_features_change(struct net_device *dev); 3795 /* Load a device via the kmod */ 3796 void dev_load(struct net *net, const char *name); 3797 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 3798 struct rtnl_link_stats64 *storage); 3799 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 3800 const struct net_device_stats *netdev_stats); 3801 3802 extern int netdev_max_backlog; 3803 extern int netdev_tstamp_prequeue; 3804 extern int weight_p; 3805 3806 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 3807 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 3808 struct list_head **iter); 3809 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 3810 struct list_head **iter); 3811 3812 /* iterate through upper list, must be called under RCU read lock */ 3813 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 3814 for (iter = &(dev)->adj_list.upper, \ 3815 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 3816 updev; \ 3817 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 3818 3819 /* iterate through upper list, must be called under RCU read lock */ 3820 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \ 3821 for (iter = &(dev)->all_adj_list.upper, \ 3822 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \ 3823 updev; \ 3824 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter))) 3825 3826 void *netdev_lower_get_next_private(struct net_device *dev, 3827 struct list_head **iter); 3828 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 3829 struct list_head **iter); 3830 3831 #define netdev_for_each_lower_private(dev, priv, iter) \ 3832 for (iter = (dev)->adj_list.lower.next, \ 3833 priv = netdev_lower_get_next_private(dev, &(iter)); \ 3834 priv; \ 3835 priv = netdev_lower_get_next_private(dev, &(iter))) 3836 3837 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 3838 for (iter = &(dev)->adj_list.lower, \ 3839 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 3840 priv; \ 3841 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 3842 3843 void *netdev_lower_get_next(struct net_device *dev, 3844 struct list_head **iter); 3845 3846 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 3847 for (iter = (dev)->adj_list.lower.next, \ 3848 ldev = netdev_lower_get_next(dev, &(iter)); \ 3849 ldev; \ 3850 ldev = netdev_lower_get_next(dev, &(iter))) 3851 3852 struct net_device *netdev_all_lower_get_next(struct net_device *dev, 3853 struct list_head **iter); 3854 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev, 3855 struct list_head **iter); 3856 3857 #define netdev_for_each_all_lower_dev(dev, ldev, iter) \ 3858 for (iter = (dev)->all_adj_list.lower.next, \ 3859 ldev = netdev_all_lower_get_next(dev, &(iter)); \ 3860 ldev; \ 3861 ldev = netdev_all_lower_get_next(dev, &(iter))) 3862 3863 #define netdev_for_each_all_lower_dev_rcu(dev, ldev, iter) \ 3864 for (iter = (dev)->all_adj_list.lower.next, \ 3865 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)); \ 3866 ldev; \ 3867 ldev = netdev_all_lower_get_next_rcu(dev, &(iter))) 3868 3869 void *netdev_adjacent_get_private(struct list_head *adj_list); 3870 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 3871 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 3872 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 3873 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev); 3874 int netdev_master_upper_dev_link(struct net_device *dev, 3875 struct net_device *upper_dev, 3876 void *upper_priv, void *upper_info); 3877 void netdev_upper_dev_unlink(struct net_device *dev, 3878 struct net_device *upper_dev); 3879 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 3880 void *netdev_lower_dev_get_private(struct net_device *dev, 3881 struct net_device *lower_dev); 3882 void netdev_lower_state_changed(struct net_device *lower_dev, 3883 void *lower_state_info); 3884 int netdev_default_l2upper_neigh_construct(struct net_device *dev, 3885 struct neighbour *n); 3886 void netdev_default_l2upper_neigh_destroy(struct net_device *dev, 3887 struct neighbour *n); 3888 3889 /* RSS keys are 40 or 52 bytes long */ 3890 #define NETDEV_RSS_KEY_LEN 52 3891 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 3892 void netdev_rss_key_fill(void *buffer, size_t len); 3893 3894 int dev_get_nest_level(struct net_device *dev); 3895 int skb_checksum_help(struct sk_buff *skb); 3896 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 3897 netdev_features_t features, bool tx_path); 3898 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 3899 netdev_features_t features); 3900 3901 struct netdev_bonding_info { 3902 ifslave slave; 3903 ifbond master; 3904 }; 3905 3906 struct netdev_notifier_bonding_info { 3907 struct netdev_notifier_info info; /* must be first */ 3908 struct netdev_bonding_info bonding_info; 3909 }; 3910 3911 void netdev_bonding_info_change(struct net_device *dev, 3912 struct netdev_bonding_info *bonding_info); 3913 3914 static inline 3915 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 3916 { 3917 return __skb_gso_segment(skb, features, true); 3918 } 3919 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 3920 3921 static inline bool can_checksum_protocol(netdev_features_t features, 3922 __be16 protocol) 3923 { 3924 if (protocol == htons(ETH_P_FCOE)) 3925 return !!(features & NETIF_F_FCOE_CRC); 3926 3927 /* Assume this is an IP checksum (not SCTP CRC) */ 3928 3929 if (features & NETIF_F_HW_CSUM) { 3930 /* Can checksum everything */ 3931 return true; 3932 } 3933 3934 switch (protocol) { 3935 case htons(ETH_P_IP): 3936 return !!(features & NETIF_F_IP_CSUM); 3937 case htons(ETH_P_IPV6): 3938 return !!(features & NETIF_F_IPV6_CSUM); 3939 default: 3940 return false; 3941 } 3942 } 3943 3944 /* Map an ethertype into IP protocol if possible */ 3945 static inline int eproto_to_ipproto(int eproto) 3946 { 3947 switch (eproto) { 3948 case htons(ETH_P_IP): 3949 return IPPROTO_IP; 3950 case htons(ETH_P_IPV6): 3951 return IPPROTO_IPV6; 3952 default: 3953 return -1; 3954 } 3955 } 3956 3957 #ifdef CONFIG_BUG 3958 void netdev_rx_csum_fault(struct net_device *dev); 3959 #else 3960 static inline void netdev_rx_csum_fault(struct net_device *dev) 3961 { 3962 } 3963 #endif 3964 /* rx skb timestamps */ 3965 void net_enable_timestamp(void); 3966 void net_disable_timestamp(void); 3967 3968 #ifdef CONFIG_PROC_FS 3969 int __init dev_proc_init(void); 3970 #else 3971 #define dev_proc_init() 0 3972 #endif 3973 3974 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 3975 struct sk_buff *skb, struct net_device *dev, 3976 bool more) 3977 { 3978 skb->xmit_more = more ? 1 : 0; 3979 return ops->ndo_start_xmit(skb, dev); 3980 } 3981 3982 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 3983 struct netdev_queue *txq, bool more) 3984 { 3985 const struct net_device_ops *ops = dev->netdev_ops; 3986 int rc; 3987 3988 rc = __netdev_start_xmit(ops, skb, dev, more); 3989 if (rc == NETDEV_TX_OK) 3990 txq_trans_update(txq); 3991 3992 return rc; 3993 } 3994 3995 int netdev_class_create_file_ns(struct class_attribute *class_attr, 3996 const void *ns); 3997 void netdev_class_remove_file_ns(struct class_attribute *class_attr, 3998 const void *ns); 3999 4000 static inline int netdev_class_create_file(struct class_attribute *class_attr) 4001 { 4002 return netdev_class_create_file_ns(class_attr, NULL); 4003 } 4004 4005 static inline void netdev_class_remove_file(struct class_attribute *class_attr) 4006 { 4007 netdev_class_remove_file_ns(class_attr, NULL); 4008 } 4009 4010 extern struct kobj_ns_type_operations net_ns_type_operations; 4011 4012 const char *netdev_drivername(const struct net_device *dev); 4013 4014 void linkwatch_run_queue(void); 4015 4016 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4017 netdev_features_t f2) 4018 { 4019 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4020 if (f1 & NETIF_F_HW_CSUM) 4021 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4022 else 4023 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4024 } 4025 4026 return f1 & f2; 4027 } 4028 4029 static inline netdev_features_t netdev_get_wanted_features( 4030 struct net_device *dev) 4031 { 4032 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4033 } 4034 netdev_features_t netdev_increment_features(netdev_features_t all, 4035 netdev_features_t one, netdev_features_t mask); 4036 4037 /* Allow TSO being used on stacked device : 4038 * Performing the GSO segmentation before last device 4039 * is a performance improvement. 4040 */ 4041 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4042 netdev_features_t mask) 4043 { 4044 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4045 } 4046 4047 int __netdev_update_features(struct net_device *dev); 4048 void netdev_update_features(struct net_device *dev); 4049 void netdev_change_features(struct net_device *dev); 4050 4051 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4052 struct net_device *dev); 4053 4054 netdev_features_t passthru_features_check(struct sk_buff *skb, 4055 struct net_device *dev, 4056 netdev_features_t features); 4057 netdev_features_t netif_skb_features(struct sk_buff *skb); 4058 4059 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4060 { 4061 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4062 4063 /* check flags correspondence */ 4064 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4065 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT)); 4066 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4067 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4068 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4069 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4070 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4071 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4072 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4073 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4074 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4075 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4076 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4077 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4078 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4079 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4080 4081 return (features & feature) == feature; 4082 } 4083 4084 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4085 { 4086 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4087 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4088 } 4089 4090 static inline bool netif_needs_gso(struct sk_buff *skb, 4091 netdev_features_t features) 4092 { 4093 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4094 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4095 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4096 } 4097 4098 static inline void netif_set_gso_max_size(struct net_device *dev, 4099 unsigned int size) 4100 { 4101 dev->gso_max_size = size; 4102 } 4103 4104 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4105 int pulled_hlen, u16 mac_offset, 4106 int mac_len) 4107 { 4108 skb->protocol = protocol; 4109 skb->encapsulation = 1; 4110 skb_push(skb, pulled_hlen); 4111 skb_reset_transport_header(skb); 4112 skb->mac_header = mac_offset; 4113 skb->network_header = skb->mac_header + mac_len; 4114 skb->mac_len = mac_len; 4115 } 4116 4117 static inline bool netif_is_macsec(const struct net_device *dev) 4118 { 4119 return dev->priv_flags & IFF_MACSEC; 4120 } 4121 4122 static inline bool netif_is_macvlan(const struct net_device *dev) 4123 { 4124 return dev->priv_flags & IFF_MACVLAN; 4125 } 4126 4127 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4128 { 4129 return dev->priv_flags & IFF_MACVLAN_PORT; 4130 } 4131 4132 static inline bool netif_is_ipvlan(const struct net_device *dev) 4133 { 4134 return dev->priv_flags & IFF_IPVLAN_SLAVE; 4135 } 4136 4137 static inline bool netif_is_ipvlan_port(const struct net_device *dev) 4138 { 4139 return dev->priv_flags & IFF_IPVLAN_MASTER; 4140 } 4141 4142 static inline bool netif_is_bond_master(const struct net_device *dev) 4143 { 4144 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4145 } 4146 4147 static inline bool netif_is_bond_slave(const struct net_device *dev) 4148 { 4149 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4150 } 4151 4152 static inline bool netif_supports_nofcs(struct net_device *dev) 4153 { 4154 return dev->priv_flags & IFF_SUPP_NOFCS; 4155 } 4156 4157 static inline bool netif_is_l3_master(const struct net_device *dev) 4158 { 4159 return dev->priv_flags & IFF_L3MDEV_MASTER; 4160 } 4161 4162 static inline bool netif_is_l3_slave(const struct net_device *dev) 4163 { 4164 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4165 } 4166 4167 static inline bool netif_is_bridge_master(const struct net_device *dev) 4168 { 4169 return dev->priv_flags & IFF_EBRIDGE; 4170 } 4171 4172 static inline bool netif_is_bridge_port(const struct net_device *dev) 4173 { 4174 return dev->priv_flags & IFF_BRIDGE_PORT; 4175 } 4176 4177 static inline bool netif_is_ovs_master(const struct net_device *dev) 4178 { 4179 return dev->priv_flags & IFF_OPENVSWITCH; 4180 } 4181 4182 static inline bool netif_is_team_master(const struct net_device *dev) 4183 { 4184 return dev->priv_flags & IFF_TEAM; 4185 } 4186 4187 static inline bool netif_is_team_port(const struct net_device *dev) 4188 { 4189 return dev->priv_flags & IFF_TEAM_PORT; 4190 } 4191 4192 static inline bool netif_is_lag_master(const struct net_device *dev) 4193 { 4194 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4195 } 4196 4197 static inline bool netif_is_lag_port(const struct net_device *dev) 4198 { 4199 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4200 } 4201 4202 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4203 { 4204 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4205 } 4206 4207 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4208 static inline void netif_keep_dst(struct net_device *dev) 4209 { 4210 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4211 } 4212 4213 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4214 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4215 { 4216 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4217 return dev->priv_flags & IFF_MACSEC; 4218 } 4219 4220 extern struct pernet_operations __net_initdata loopback_net_ops; 4221 4222 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4223 4224 /* netdev_printk helpers, similar to dev_printk */ 4225 4226 static inline const char *netdev_name(const struct net_device *dev) 4227 { 4228 if (!dev->name[0] || strchr(dev->name, '%')) 4229 return "(unnamed net_device)"; 4230 return dev->name; 4231 } 4232 4233 static inline const char *netdev_reg_state(const struct net_device *dev) 4234 { 4235 switch (dev->reg_state) { 4236 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4237 case NETREG_REGISTERED: return ""; 4238 case NETREG_UNREGISTERING: return " (unregistering)"; 4239 case NETREG_UNREGISTERED: return " (unregistered)"; 4240 case NETREG_RELEASED: return " (released)"; 4241 case NETREG_DUMMY: return " (dummy)"; 4242 } 4243 4244 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4245 return " (unknown)"; 4246 } 4247 4248 __printf(3, 4) 4249 void netdev_printk(const char *level, const struct net_device *dev, 4250 const char *format, ...); 4251 __printf(2, 3) 4252 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4253 __printf(2, 3) 4254 void netdev_alert(const struct net_device *dev, const char *format, ...); 4255 __printf(2, 3) 4256 void netdev_crit(const struct net_device *dev, const char *format, ...); 4257 __printf(2, 3) 4258 void netdev_err(const struct net_device *dev, const char *format, ...); 4259 __printf(2, 3) 4260 void netdev_warn(const struct net_device *dev, const char *format, ...); 4261 __printf(2, 3) 4262 void netdev_notice(const struct net_device *dev, const char *format, ...); 4263 __printf(2, 3) 4264 void netdev_info(const struct net_device *dev, const char *format, ...); 4265 4266 #define MODULE_ALIAS_NETDEV(device) \ 4267 MODULE_ALIAS("netdev-" device) 4268 4269 #if defined(CONFIG_DYNAMIC_DEBUG) 4270 #define netdev_dbg(__dev, format, args...) \ 4271 do { \ 4272 dynamic_netdev_dbg(__dev, format, ##args); \ 4273 } while (0) 4274 #elif defined(DEBUG) 4275 #define netdev_dbg(__dev, format, args...) \ 4276 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4277 #else 4278 #define netdev_dbg(__dev, format, args...) \ 4279 ({ \ 4280 if (0) \ 4281 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4282 }) 4283 #endif 4284 4285 #if defined(VERBOSE_DEBUG) 4286 #define netdev_vdbg netdev_dbg 4287 #else 4288 4289 #define netdev_vdbg(dev, format, args...) \ 4290 ({ \ 4291 if (0) \ 4292 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 4293 0; \ 4294 }) 4295 #endif 4296 4297 /* 4298 * netdev_WARN() acts like dev_printk(), but with the key difference 4299 * of using a WARN/WARN_ON to get the message out, including the 4300 * file/line information and a backtrace. 4301 */ 4302 #define netdev_WARN(dev, format, args...) \ 4303 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \ 4304 netdev_reg_state(dev), ##args) 4305 4306 /* netif printk helpers, similar to netdev_printk */ 4307 4308 #define netif_printk(priv, type, level, dev, fmt, args...) \ 4309 do { \ 4310 if (netif_msg_##type(priv)) \ 4311 netdev_printk(level, (dev), fmt, ##args); \ 4312 } while (0) 4313 4314 #define netif_level(level, priv, type, dev, fmt, args...) \ 4315 do { \ 4316 if (netif_msg_##type(priv)) \ 4317 netdev_##level(dev, fmt, ##args); \ 4318 } while (0) 4319 4320 #define netif_emerg(priv, type, dev, fmt, args...) \ 4321 netif_level(emerg, priv, type, dev, fmt, ##args) 4322 #define netif_alert(priv, type, dev, fmt, args...) \ 4323 netif_level(alert, priv, type, dev, fmt, ##args) 4324 #define netif_crit(priv, type, dev, fmt, args...) \ 4325 netif_level(crit, priv, type, dev, fmt, ##args) 4326 #define netif_err(priv, type, dev, fmt, args...) \ 4327 netif_level(err, priv, type, dev, fmt, ##args) 4328 #define netif_warn(priv, type, dev, fmt, args...) \ 4329 netif_level(warn, priv, type, dev, fmt, ##args) 4330 #define netif_notice(priv, type, dev, fmt, args...) \ 4331 netif_level(notice, priv, type, dev, fmt, ##args) 4332 #define netif_info(priv, type, dev, fmt, args...) \ 4333 netif_level(info, priv, type, dev, fmt, ##args) 4334 4335 #if defined(CONFIG_DYNAMIC_DEBUG) 4336 #define netif_dbg(priv, type, netdev, format, args...) \ 4337 do { \ 4338 if (netif_msg_##type(priv)) \ 4339 dynamic_netdev_dbg(netdev, format, ##args); \ 4340 } while (0) 4341 #elif defined(DEBUG) 4342 #define netif_dbg(priv, type, dev, format, args...) \ 4343 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 4344 #else 4345 #define netif_dbg(priv, type, dev, format, args...) \ 4346 ({ \ 4347 if (0) \ 4348 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4349 0; \ 4350 }) 4351 #endif 4352 4353 #if defined(VERBOSE_DEBUG) 4354 #define netif_vdbg netif_dbg 4355 #else 4356 #define netif_vdbg(priv, type, dev, format, args...) \ 4357 ({ \ 4358 if (0) \ 4359 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4360 0; \ 4361 }) 4362 #endif 4363 4364 /* 4365 * The list of packet types we will receive (as opposed to discard) 4366 * and the routines to invoke. 4367 * 4368 * Why 16. Because with 16 the only overlap we get on a hash of the 4369 * low nibble of the protocol value is RARP/SNAP/X.25. 4370 * 4371 * NOTE: That is no longer true with the addition of VLAN tags. Not 4372 * sure which should go first, but I bet it won't make much 4373 * difference if we are running VLANs. The good news is that 4374 * this protocol won't be in the list unless compiled in, so 4375 * the average user (w/out VLANs) will not be adversely affected. 4376 * --BLG 4377 * 4378 * 0800 IP 4379 * 8100 802.1Q VLAN 4380 * 0001 802.3 4381 * 0002 AX.25 4382 * 0004 802.2 4383 * 8035 RARP 4384 * 0005 SNAP 4385 * 0805 X.25 4386 * 0806 ARP 4387 * 8137 IPX 4388 * 0009 Localtalk 4389 * 86DD IPv6 4390 */ 4391 #define PTYPE_HASH_SIZE (16) 4392 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 4393 4394 #endif /* _LINUX_NETDEVICE_H */ 4395