1 /* 2 * INET An implementation of the TCP/IP protocol suite for the LINUX 3 * operating system. INET is implemented using the BSD Socket 4 * interface as the means of communication with the user level. 5 * 6 * Definitions for the Interfaces handler. 7 * 8 * Version: @(#)dev.h 1.0.10 08/12/93 9 * 10 * Authors: Ross Biro 11 * Fred N. van Kempen, <[email protected]> 12 * Corey Minyard <[email protected]> 13 * Donald J. Becker, <[email protected]> 14 * Alan Cox, <[email protected]> 15 * Bjorn Ekwall. <[email protected]> 16 * Pekka Riikonen <[email protected]> 17 * 18 * This program is free software; you can redistribute it and/or 19 * modify it under the terms of the GNU General Public License 20 * as published by the Free Software Foundation; either version 21 * 2 of the License, or (at your option) any later version. 22 * 23 * Moved to /usr/include/linux for NET3 24 */ 25 #ifndef _LINUX_NETDEVICE_H 26 #define _LINUX_NETDEVICE_H 27 28 #include <linux/timer.h> 29 #include <linux/bug.h> 30 #include <linux/delay.h> 31 #include <linux/atomic.h> 32 #include <linux/prefetch.h> 33 #include <asm/cache.h> 34 #include <asm/byteorder.h> 35 36 #include <linux/percpu.h> 37 #include <linux/rculist.h> 38 #include <linux/dmaengine.h> 39 #include <linux/workqueue.h> 40 #include <linux/dynamic_queue_limits.h> 41 42 #include <linux/ethtool.h> 43 #include <net/net_namespace.h> 44 #include <net/dsa.h> 45 #ifdef CONFIG_DCB 46 #include <net/dcbnl.h> 47 #endif 48 #include <net/netprio_cgroup.h> 49 50 #include <linux/netdev_features.h> 51 #include <linux/neighbour.h> 52 #include <uapi/linux/netdevice.h> 53 #include <uapi/linux/if_bonding.h> 54 #include <uapi/linux/pkt_cls.h> 55 #include <linux/hashtable.h> 56 57 struct netpoll_info; 58 struct device; 59 struct phy_device; 60 /* 802.11 specific */ 61 struct wireless_dev; 62 /* 802.15.4 specific */ 63 struct wpan_dev; 64 struct mpls_dev; 65 /* UDP Tunnel offloads */ 66 struct udp_tunnel_info; 67 struct bpf_prog; 68 69 void netdev_set_default_ethtool_ops(struct net_device *dev, 70 const struct ethtool_ops *ops); 71 72 /* Backlog congestion levels */ 73 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 74 #define NET_RX_DROP 1 /* packet dropped */ 75 76 /* 77 * Transmit return codes: transmit return codes originate from three different 78 * namespaces: 79 * 80 * - qdisc return codes 81 * - driver transmit return codes 82 * - errno values 83 * 84 * Drivers are allowed to return any one of those in their hard_start_xmit() 85 * function. Real network devices commonly used with qdiscs should only return 86 * the driver transmit return codes though - when qdiscs are used, the actual 87 * transmission happens asynchronously, so the value is not propagated to 88 * higher layers. Virtual network devices transmit synchronously; in this case 89 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 90 * others are propagated to higher layers. 91 */ 92 93 /* qdisc ->enqueue() return codes. */ 94 #define NET_XMIT_SUCCESS 0x00 95 #define NET_XMIT_DROP 0x01 /* skb dropped */ 96 #define NET_XMIT_CN 0x02 /* congestion notification */ 97 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 98 99 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 100 * indicates that the device will soon be dropping packets, or already drops 101 * some packets of the same priority; prompting us to send less aggressively. */ 102 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 103 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 104 105 /* Driver transmit return codes */ 106 #define NETDEV_TX_MASK 0xf0 107 108 enum netdev_tx { 109 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 110 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 111 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 112 }; 113 typedef enum netdev_tx netdev_tx_t; 114 115 /* 116 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 117 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 118 */ 119 static inline bool dev_xmit_complete(int rc) 120 { 121 /* 122 * Positive cases with an skb consumed by a driver: 123 * - successful transmission (rc == NETDEV_TX_OK) 124 * - error while transmitting (rc < 0) 125 * - error while queueing to a different device (rc & NET_XMIT_MASK) 126 */ 127 if (likely(rc < NET_XMIT_MASK)) 128 return true; 129 130 return false; 131 } 132 133 /* 134 * Compute the worst-case header length according to the protocols 135 * used. 136 */ 137 138 #if defined(CONFIG_HYPERV_NET) 139 # define LL_MAX_HEADER 128 140 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 141 # if defined(CONFIG_MAC80211_MESH) 142 # define LL_MAX_HEADER 128 143 # else 144 # define LL_MAX_HEADER 96 145 # endif 146 #else 147 # define LL_MAX_HEADER 32 148 #endif 149 150 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 151 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 152 #define MAX_HEADER LL_MAX_HEADER 153 #else 154 #define MAX_HEADER (LL_MAX_HEADER + 48) 155 #endif 156 157 /* 158 * Old network device statistics. Fields are native words 159 * (unsigned long) so they can be read and written atomically. 160 */ 161 162 struct net_device_stats { 163 unsigned long rx_packets; 164 unsigned long tx_packets; 165 unsigned long rx_bytes; 166 unsigned long tx_bytes; 167 unsigned long rx_errors; 168 unsigned long tx_errors; 169 unsigned long rx_dropped; 170 unsigned long tx_dropped; 171 unsigned long multicast; 172 unsigned long collisions; 173 unsigned long rx_length_errors; 174 unsigned long rx_over_errors; 175 unsigned long rx_crc_errors; 176 unsigned long rx_frame_errors; 177 unsigned long rx_fifo_errors; 178 unsigned long rx_missed_errors; 179 unsigned long tx_aborted_errors; 180 unsigned long tx_carrier_errors; 181 unsigned long tx_fifo_errors; 182 unsigned long tx_heartbeat_errors; 183 unsigned long tx_window_errors; 184 unsigned long rx_compressed; 185 unsigned long tx_compressed; 186 }; 187 188 189 #include <linux/cache.h> 190 #include <linux/skbuff.h> 191 192 #ifdef CONFIG_RPS 193 #include <linux/static_key.h> 194 extern struct static_key rps_needed; 195 extern struct static_key rfs_needed; 196 #endif 197 198 struct neighbour; 199 struct neigh_parms; 200 struct sk_buff; 201 202 struct netdev_hw_addr { 203 struct list_head list; 204 unsigned char addr[MAX_ADDR_LEN]; 205 unsigned char type; 206 #define NETDEV_HW_ADDR_T_LAN 1 207 #define NETDEV_HW_ADDR_T_SAN 2 208 #define NETDEV_HW_ADDR_T_SLAVE 3 209 #define NETDEV_HW_ADDR_T_UNICAST 4 210 #define NETDEV_HW_ADDR_T_MULTICAST 5 211 bool global_use; 212 int sync_cnt; 213 int refcount; 214 int synced; 215 struct rcu_head rcu_head; 216 }; 217 218 struct netdev_hw_addr_list { 219 struct list_head list; 220 int count; 221 }; 222 223 #define netdev_hw_addr_list_count(l) ((l)->count) 224 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 225 #define netdev_hw_addr_list_for_each(ha, l) \ 226 list_for_each_entry(ha, &(l)->list, list) 227 228 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 229 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 230 #define netdev_for_each_uc_addr(ha, dev) \ 231 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 232 233 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 234 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 235 #define netdev_for_each_mc_addr(ha, dev) \ 236 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 237 238 struct hh_cache { 239 u16 hh_len; 240 u16 __pad; 241 seqlock_t hh_lock; 242 243 /* cached hardware header; allow for machine alignment needs. */ 244 #define HH_DATA_MOD 16 245 #define HH_DATA_OFF(__len) \ 246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 247 #define HH_DATA_ALIGN(__len) \ 248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 249 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 250 }; 251 252 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 253 * Alternative is: 254 * dev->hard_header_len ? (dev->hard_header_len + 255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 256 * 257 * We could use other alignment values, but we must maintain the 258 * relationship HH alignment <= LL alignment. 259 */ 260 #define LL_RESERVED_SPACE(dev) \ 261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 262 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 264 265 struct header_ops { 266 int (*create) (struct sk_buff *skb, struct net_device *dev, 267 unsigned short type, const void *daddr, 268 const void *saddr, unsigned int len); 269 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 270 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 271 void (*cache_update)(struct hh_cache *hh, 272 const struct net_device *dev, 273 const unsigned char *haddr); 274 bool (*validate)(const char *ll_header, unsigned int len); 275 }; 276 277 /* These flag bits are private to the generic network queueing 278 * layer; they may not be explicitly referenced by any other 279 * code. 280 */ 281 282 enum netdev_state_t { 283 __LINK_STATE_START, 284 __LINK_STATE_PRESENT, 285 __LINK_STATE_NOCARRIER, 286 __LINK_STATE_LINKWATCH_PENDING, 287 __LINK_STATE_DORMANT, 288 }; 289 290 291 /* 292 * This structure holds boot-time configured netdevice settings. They 293 * are then used in the device probing. 294 */ 295 struct netdev_boot_setup { 296 char name[IFNAMSIZ]; 297 struct ifmap map; 298 }; 299 #define NETDEV_BOOT_SETUP_MAX 8 300 301 int __init netdev_boot_setup(char *str); 302 303 /* 304 * Structure for NAPI scheduling similar to tasklet but with weighting 305 */ 306 struct napi_struct { 307 /* The poll_list must only be managed by the entity which 308 * changes the state of the NAPI_STATE_SCHED bit. This means 309 * whoever atomically sets that bit can add this napi_struct 310 * to the per-CPU poll_list, and whoever clears that bit 311 * can remove from the list right before clearing the bit. 312 */ 313 struct list_head poll_list; 314 315 unsigned long state; 316 int weight; 317 unsigned int gro_count; 318 int (*poll)(struct napi_struct *, int); 319 #ifdef CONFIG_NETPOLL 320 int poll_owner; 321 #endif 322 struct net_device *dev; 323 struct sk_buff *gro_list; 324 struct sk_buff *skb; 325 struct hrtimer timer; 326 struct list_head dev_list; 327 struct hlist_node napi_hash_node; 328 unsigned int napi_id; 329 }; 330 331 enum { 332 NAPI_STATE_SCHED, /* Poll is scheduled */ 333 NAPI_STATE_DISABLE, /* Disable pending */ 334 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 335 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */ 336 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */ 337 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */ 338 }; 339 340 enum { 341 NAPIF_STATE_SCHED = (1UL << NAPI_STATE_SCHED), 342 NAPIF_STATE_DISABLE = (1UL << NAPI_STATE_DISABLE), 343 NAPIF_STATE_NPSVC = (1UL << NAPI_STATE_NPSVC), 344 NAPIF_STATE_HASHED = (1UL << NAPI_STATE_HASHED), 345 NAPIF_STATE_NO_BUSY_POLL = (1UL << NAPI_STATE_NO_BUSY_POLL), 346 NAPIF_STATE_IN_BUSY_POLL = (1UL << NAPI_STATE_IN_BUSY_POLL), 347 }; 348 349 enum gro_result { 350 GRO_MERGED, 351 GRO_MERGED_FREE, 352 GRO_HELD, 353 GRO_NORMAL, 354 GRO_DROP, 355 }; 356 typedef enum gro_result gro_result_t; 357 358 /* 359 * enum rx_handler_result - Possible return values for rx_handlers. 360 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 361 * further. 362 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 363 * case skb->dev was changed by rx_handler. 364 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 365 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 366 * 367 * rx_handlers are functions called from inside __netif_receive_skb(), to do 368 * special processing of the skb, prior to delivery to protocol handlers. 369 * 370 * Currently, a net_device can only have a single rx_handler registered. Trying 371 * to register a second rx_handler will return -EBUSY. 372 * 373 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 374 * To unregister a rx_handler on a net_device, use 375 * netdev_rx_handler_unregister(). 376 * 377 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 378 * do with the skb. 379 * 380 * If the rx_handler consumed the skb in some way, it should return 381 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 382 * the skb to be delivered in some other way. 383 * 384 * If the rx_handler changed skb->dev, to divert the skb to another 385 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 386 * new device will be called if it exists. 387 * 388 * If the rx_handler decides the skb should be ignored, it should return 389 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 390 * are registered on exact device (ptype->dev == skb->dev). 391 * 392 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 393 * delivered, it should return RX_HANDLER_PASS. 394 * 395 * A device without a registered rx_handler will behave as if rx_handler 396 * returned RX_HANDLER_PASS. 397 */ 398 399 enum rx_handler_result { 400 RX_HANDLER_CONSUMED, 401 RX_HANDLER_ANOTHER, 402 RX_HANDLER_EXACT, 403 RX_HANDLER_PASS, 404 }; 405 typedef enum rx_handler_result rx_handler_result_t; 406 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 407 408 void __napi_schedule(struct napi_struct *n); 409 void __napi_schedule_irqoff(struct napi_struct *n); 410 411 static inline bool napi_disable_pending(struct napi_struct *n) 412 { 413 return test_bit(NAPI_STATE_DISABLE, &n->state); 414 } 415 416 /** 417 * napi_schedule_prep - check if NAPI can be scheduled 418 * @n: NAPI context 419 * 420 * Test if NAPI routine is already running, and if not mark 421 * it as running. This is used as a condition variable to 422 * insure only one NAPI poll instance runs. We also make 423 * sure there is no pending NAPI disable. 424 */ 425 static inline bool napi_schedule_prep(struct napi_struct *n) 426 { 427 return !napi_disable_pending(n) && 428 !test_and_set_bit(NAPI_STATE_SCHED, &n->state); 429 } 430 431 /** 432 * napi_schedule - schedule NAPI poll 433 * @n: NAPI context 434 * 435 * Schedule NAPI poll routine to be called if it is not already 436 * running. 437 */ 438 static inline void napi_schedule(struct napi_struct *n) 439 { 440 if (napi_schedule_prep(n)) 441 __napi_schedule(n); 442 } 443 444 /** 445 * napi_schedule_irqoff - schedule NAPI poll 446 * @n: NAPI context 447 * 448 * Variant of napi_schedule(), assuming hard irqs are masked. 449 */ 450 static inline void napi_schedule_irqoff(struct napi_struct *n) 451 { 452 if (napi_schedule_prep(n)) 453 __napi_schedule_irqoff(n); 454 } 455 456 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 457 static inline bool napi_reschedule(struct napi_struct *napi) 458 { 459 if (napi_schedule_prep(napi)) { 460 __napi_schedule(napi); 461 return true; 462 } 463 return false; 464 } 465 466 bool __napi_complete(struct napi_struct *n); 467 bool napi_complete_done(struct napi_struct *n, int work_done); 468 /** 469 * napi_complete - NAPI processing complete 470 * @n: NAPI context 471 * 472 * Mark NAPI processing as complete. 473 * Consider using napi_complete_done() instead. 474 * Return false if device should avoid rearming interrupts. 475 */ 476 static inline bool napi_complete(struct napi_struct *n) 477 { 478 return napi_complete_done(n, 0); 479 } 480 481 /** 482 * napi_hash_del - remove a NAPI from global table 483 * @napi: NAPI context 484 * 485 * Warning: caller must observe RCU grace period 486 * before freeing memory containing @napi, if 487 * this function returns true. 488 * Note: core networking stack automatically calls it 489 * from netif_napi_del(). 490 * Drivers might want to call this helper to combine all 491 * the needed RCU grace periods into a single one. 492 */ 493 bool napi_hash_del(struct napi_struct *napi); 494 495 /** 496 * napi_disable - prevent NAPI from scheduling 497 * @n: NAPI context 498 * 499 * Stop NAPI from being scheduled on this context. 500 * Waits till any outstanding processing completes. 501 */ 502 void napi_disable(struct napi_struct *n); 503 504 /** 505 * napi_enable - enable NAPI scheduling 506 * @n: NAPI context 507 * 508 * Resume NAPI from being scheduled on this context. 509 * Must be paired with napi_disable. 510 */ 511 static inline void napi_enable(struct napi_struct *n) 512 { 513 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 514 smp_mb__before_atomic(); 515 clear_bit(NAPI_STATE_SCHED, &n->state); 516 clear_bit(NAPI_STATE_NPSVC, &n->state); 517 } 518 519 /** 520 * napi_synchronize - wait until NAPI is not running 521 * @n: NAPI context 522 * 523 * Wait until NAPI is done being scheduled on this context. 524 * Waits till any outstanding processing completes but 525 * does not disable future activations. 526 */ 527 static inline void napi_synchronize(const struct napi_struct *n) 528 { 529 if (IS_ENABLED(CONFIG_SMP)) 530 while (test_bit(NAPI_STATE_SCHED, &n->state)) 531 msleep(1); 532 else 533 barrier(); 534 } 535 536 enum netdev_queue_state_t { 537 __QUEUE_STATE_DRV_XOFF, 538 __QUEUE_STATE_STACK_XOFF, 539 __QUEUE_STATE_FROZEN, 540 }; 541 542 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 543 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 544 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 545 546 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 547 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 548 QUEUE_STATE_FROZEN) 549 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 550 QUEUE_STATE_FROZEN) 551 552 /* 553 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 554 * netif_tx_* functions below are used to manipulate this flag. The 555 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 556 * queue independently. The netif_xmit_*stopped functions below are called 557 * to check if the queue has been stopped by the driver or stack (either 558 * of the XOFF bits are set in the state). Drivers should not need to call 559 * netif_xmit*stopped functions, they should only be using netif_tx_*. 560 */ 561 562 struct netdev_queue { 563 /* 564 * read-mostly part 565 */ 566 struct net_device *dev; 567 struct Qdisc __rcu *qdisc; 568 struct Qdisc *qdisc_sleeping; 569 #ifdef CONFIG_SYSFS 570 struct kobject kobj; 571 #endif 572 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 573 int numa_node; 574 #endif 575 unsigned long tx_maxrate; 576 /* 577 * Number of TX timeouts for this queue 578 * (/sys/class/net/DEV/Q/trans_timeout) 579 */ 580 unsigned long trans_timeout; 581 /* 582 * write-mostly part 583 */ 584 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 585 int xmit_lock_owner; 586 /* 587 * Time (in jiffies) of last Tx 588 */ 589 unsigned long trans_start; 590 591 unsigned long state; 592 593 #ifdef CONFIG_BQL 594 struct dql dql; 595 #endif 596 } ____cacheline_aligned_in_smp; 597 598 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 599 { 600 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 601 return q->numa_node; 602 #else 603 return NUMA_NO_NODE; 604 #endif 605 } 606 607 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 608 { 609 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 610 q->numa_node = node; 611 #endif 612 } 613 614 #ifdef CONFIG_RPS 615 /* 616 * This structure holds an RPS map which can be of variable length. The 617 * map is an array of CPUs. 618 */ 619 struct rps_map { 620 unsigned int len; 621 struct rcu_head rcu; 622 u16 cpus[0]; 623 }; 624 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 625 626 /* 627 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 628 * tail pointer for that CPU's input queue at the time of last enqueue, and 629 * a hardware filter index. 630 */ 631 struct rps_dev_flow { 632 u16 cpu; 633 u16 filter; 634 unsigned int last_qtail; 635 }; 636 #define RPS_NO_FILTER 0xffff 637 638 /* 639 * The rps_dev_flow_table structure contains a table of flow mappings. 640 */ 641 struct rps_dev_flow_table { 642 unsigned int mask; 643 struct rcu_head rcu; 644 struct rps_dev_flow flows[0]; 645 }; 646 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 647 ((_num) * sizeof(struct rps_dev_flow))) 648 649 /* 650 * The rps_sock_flow_table contains mappings of flows to the last CPU 651 * on which they were processed by the application (set in recvmsg). 652 * Each entry is a 32bit value. Upper part is the high-order bits 653 * of flow hash, lower part is CPU number. 654 * rps_cpu_mask is used to partition the space, depending on number of 655 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 656 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 657 * meaning we use 32-6=26 bits for the hash. 658 */ 659 struct rps_sock_flow_table { 660 u32 mask; 661 662 u32 ents[0] ____cacheline_aligned_in_smp; 663 }; 664 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 665 666 #define RPS_NO_CPU 0xffff 667 668 extern u32 rps_cpu_mask; 669 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 670 671 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 672 u32 hash) 673 { 674 if (table && hash) { 675 unsigned int index = hash & table->mask; 676 u32 val = hash & ~rps_cpu_mask; 677 678 /* We only give a hint, preemption can change CPU under us */ 679 val |= raw_smp_processor_id(); 680 681 if (table->ents[index] != val) 682 table->ents[index] = val; 683 } 684 } 685 686 #ifdef CONFIG_RFS_ACCEL 687 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 688 u16 filter_id); 689 #endif 690 #endif /* CONFIG_RPS */ 691 692 /* This structure contains an instance of an RX queue. */ 693 struct netdev_rx_queue { 694 #ifdef CONFIG_RPS 695 struct rps_map __rcu *rps_map; 696 struct rps_dev_flow_table __rcu *rps_flow_table; 697 #endif 698 struct kobject kobj; 699 struct net_device *dev; 700 } ____cacheline_aligned_in_smp; 701 702 /* 703 * RX queue sysfs structures and functions. 704 */ 705 struct rx_queue_attribute { 706 struct attribute attr; 707 ssize_t (*show)(struct netdev_rx_queue *queue, 708 struct rx_queue_attribute *attr, char *buf); 709 ssize_t (*store)(struct netdev_rx_queue *queue, 710 struct rx_queue_attribute *attr, const char *buf, size_t len); 711 }; 712 713 #ifdef CONFIG_XPS 714 /* 715 * This structure holds an XPS map which can be of variable length. The 716 * map is an array of queues. 717 */ 718 struct xps_map { 719 unsigned int len; 720 unsigned int alloc_len; 721 struct rcu_head rcu; 722 u16 queues[0]; 723 }; 724 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 725 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 726 - sizeof(struct xps_map)) / sizeof(u16)) 727 728 /* 729 * This structure holds all XPS maps for device. Maps are indexed by CPU. 730 */ 731 struct xps_dev_maps { 732 struct rcu_head rcu; 733 struct xps_map __rcu *cpu_map[0]; 734 }; 735 #define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 736 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 737 #endif /* CONFIG_XPS */ 738 739 #define TC_MAX_QUEUE 16 740 #define TC_BITMASK 15 741 /* HW offloaded queuing disciplines txq count and offset maps */ 742 struct netdev_tc_txq { 743 u16 count; 744 u16 offset; 745 }; 746 747 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 748 /* 749 * This structure is to hold information about the device 750 * configured to run FCoE protocol stack. 751 */ 752 struct netdev_fcoe_hbainfo { 753 char manufacturer[64]; 754 char serial_number[64]; 755 char hardware_version[64]; 756 char driver_version[64]; 757 char optionrom_version[64]; 758 char firmware_version[64]; 759 char model[256]; 760 char model_description[256]; 761 }; 762 #endif 763 764 #define MAX_PHYS_ITEM_ID_LEN 32 765 766 /* This structure holds a unique identifier to identify some 767 * physical item (port for example) used by a netdevice. 768 */ 769 struct netdev_phys_item_id { 770 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 771 unsigned char id_len; 772 }; 773 774 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 775 struct netdev_phys_item_id *b) 776 { 777 return a->id_len == b->id_len && 778 memcmp(a->id, b->id, a->id_len) == 0; 779 } 780 781 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 782 struct sk_buff *skb); 783 784 /* These structures hold the attributes of qdisc and classifiers 785 * that are being passed to the netdevice through the setup_tc op. 786 */ 787 enum { 788 TC_SETUP_MQPRIO, 789 TC_SETUP_CLSU32, 790 TC_SETUP_CLSFLOWER, 791 TC_SETUP_MATCHALL, 792 TC_SETUP_CLSBPF, 793 }; 794 795 struct tc_cls_u32_offload; 796 797 struct tc_to_netdev { 798 unsigned int type; 799 union { 800 u8 tc; 801 struct tc_cls_u32_offload *cls_u32; 802 struct tc_cls_flower_offload *cls_flower; 803 struct tc_cls_matchall_offload *cls_mall; 804 struct tc_cls_bpf_offload *cls_bpf; 805 }; 806 bool egress_dev; 807 }; 808 809 /* These structures hold the attributes of xdp state that are being passed 810 * to the netdevice through the xdp op. 811 */ 812 enum xdp_netdev_command { 813 /* Set or clear a bpf program used in the earliest stages of packet 814 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 815 * is responsible for calling bpf_prog_put on any old progs that are 816 * stored. In case of error, the callee need not release the new prog 817 * reference, but on success it takes ownership and must bpf_prog_put 818 * when it is no longer used. 819 */ 820 XDP_SETUP_PROG, 821 /* Check if a bpf program is set on the device. The callee should 822 * return true if a program is currently attached and running. 823 */ 824 XDP_QUERY_PROG, 825 }; 826 827 struct netdev_xdp { 828 enum xdp_netdev_command command; 829 union { 830 /* XDP_SETUP_PROG */ 831 struct bpf_prog *prog; 832 /* XDP_QUERY_PROG */ 833 bool prog_attached; 834 }; 835 }; 836 837 /* 838 * This structure defines the management hooks for network devices. 839 * The following hooks can be defined; unless noted otherwise, they are 840 * optional and can be filled with a null pointer. 841 * 842 * int (*ndo_init)(struct net_device *dev); 843 * This function is called once when a network device is registered. 844 * The network device can use this for any late stage initialization 845 * or semantic validation. It can fail with an error code which will 846 * be propagated back to register_netdev. 847 * 848 * void (*ndo_uninit)(struct net_device *dev); 849 * This function is called when device is unregistered or when registration 850 * fails. It is not called if init fails. 851 * 852 * int (*ndo_open)(struct net_device *dev); 853 * This function is called when a network device transitions to the up 854 * state. 855 * 856 * int (*ndo_stop)(struct net_device *dev); 857 * This function is called when a network device transitions to the down 858 * state. 859 * 860 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 861 * struct net_device *dev); 862 * Called when a packet needs to be transmitted. 863 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 864 * the queue before that can happen; it's for obsolete devices and weird 865 * corner cases, but the stack really does a non-trivial amount 866 * of useless work if you return NETDEV_TX_BUSY. 867 * Required; cannot be NULL. 868 * 869 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 870 * netdev_features_t features); 871 * Adjusts the requested feature flags according to device-specific 872 * constraints, and returns the resulting flags. Must not modify 873 * the device state. 874 * 875 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 876 * void *accel_priv, select_queue_fallback_t fallback); 877 * Called to decide which queue to use when device supports multiple 878 * transmit queues. 879 * 880 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 881 * This function is called to allow device receiver to make 882 * changes to configuration when multicast or promiscuous is enabled. 883 * 884 * void (*ndo_set_rx_mode)(struct net_device *dev); 885 * This function is called device changes address list filtering. 886 * If driver handles unicast address filtering, it should set 887 * IFF_UNICAST_FLT in its priv_flags. 888 * 889 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 890 * This function is called when the Media Access Control address 891 * needs to be changed. If this interface is not defined, the 892 * MAC address can not be changed. 893 * 894 * int (*ndo_validate_addr)(struct net_device *dev); 895 * Test if Media Access Control address is valid for the device. 896 * 897 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 898 * Called when a user requests an ioctl which can't be handled by 899 * the generic interface code. If not defined ioctls return 900 * not supported error code. 901 * 902 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 903 * Used to set network devices bus interface parameters. This interface 904 * is retained for legacy reasons; new devices should use the bus 905 * interface (PCI) for low level management. 906 * 907 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 908 * Called when a user wants to change the Maximum Transfer Unit 909 * of a device. If not defined, any request to change MTU will 910 * will return an error. 911 * 912 * void (*ndo_tx_timeout)(struct net_device *dev); 913 * Callback used when the transmitter has not made any progress 914 * for dev->watchdog ticks. 915 * 916 * void (*ndo_get_stats64)(struct net_device *dev, 917 * struct rtnl_link_stats64 *storage); 918 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 919 * Called when a user wants to get the network device usage 920 * statistics. Drivers must do one of the following: 921 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 922 * rtnl_link_stats64 structure passed by the caller. 923 * 2. Define @ndo_get_stats to update a net_device_stats structure 924 * (which should normally be dev->stats) and return a pointer to 925 * it. The structure may be changed asynchronously only if each 926 * field is written atomically. 927 * 3. Update dev->stats asynchronously and atomically, and define 928 * neither operation. 929 * 930 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 931 * Return true if this device supports offload stats of this attr_id. 932 * 933 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 934 * void *attr_data) 935 * Get statistics for offload operations by attr_id. Write it into the 936 * attr_data pointer. 937 * 938 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 939 * If device supports VLAN filtering this function is called when a 940 * VLAN id is registered. 941 * 942 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 943 * If device supports VLAN filtering this function is called when a 944 * VLAN id is unregistered. 945 * 946 * void (*ndo_poll_controller)(struct net_device *dev); 947 * 948 * SR-IOV management functions. 949 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 950 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 951 * u8 qos, __be16 proto); 952 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 953 * int max_tx_rate); 954 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 955 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 956 * int (*ndo_get_vf_config)(struct net_device *dev, 957 * int vf, struct ifla_vf_info *ivf); 958 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 959 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 960 * struct nlattr *port[]); 961 * 962 * Enable or disable the VF ability to query its RSS Redirection Table and 963 * Hash Key. This is needed since on some devices VF share this information 964 * with PF and querying it may introduce a theoretical security risk. 965 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 966 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 967 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc) 968 * Called to setup 'tc' number of traffic classes in the net device. This 969 * is always called from the stack with the rtnl lock held and netif tx 970 * queues stopped. This allows the netdevice to perform queue management 971 * safely. 972 * 973 * Fiber Channel over Ethernet (FCoE) offload functions. 974 * int (*ndo_fcoe_enable)(struct net_device *dev); 975 * Called when the FCoE protocol stack wants to start using LLD for FCoE 976 * so the underlying device can perform whatever needed configuration or 977 * initialization to support acceleration of FCoE traffic. 978 * 979 * int (*ndo_fcoe_disable)(struct net_device *dev); 980 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 981 * so the underlying device can perform whatever needed clean-ups to 982 * stop supporting acceleration of FCoE traffic. 983 * 984 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 985 * struct scatterlist *sgl, unsigned int sgc); 986 * Called when the FCoE Initiator wants to initialize an I/O that 987 * is a possible candidate for Direct Data Placement (DDP). The LLD can 988 * perform necessary setup and returns 1 to indicate the device is set up 989 * successfully to perform DDP on this I/O, otherwise this returns 0. 990 * 991 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 992 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 993 * indicated by the FC exchange id 'xid', so the underlying device can 994 * clean up and reuse resources for later DDP requests. 995 * 996 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 997 * struct scatterlist *sgl, unsigned int sgc); 998 * Called when the FCoE Target wants to initialize an I/O that 999 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1000 * perform necessary setup and returns 1 to indicate the device is set up 1001 * successfully to perform DDP on this I/O, otherwise this returns 0. 1002 * 1003 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1004 * struct netdev_fcoe_hbainfo *hbainfo); 1005 * Called when the FCoE Protocol stack wants information on the underlying 1006 * device. This information is utilized by the FCoE protocol stack to 1007 * register attributes with Fiber Channel management service as per the 1008 * FC-GS Fabric Device Management Information(FDMI) specification. 1009 * 1010 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1011 * Called when the underlying device wants to override default World Wide 1012 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1013 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1014 * protocol stack to use. 1015 * 1016 * RFS acceleration. 1017 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1018 * u16 rxq_index, u32 flow_id); 1019 * Set hardware filter for RFS. rxq_index is the target queue index; 1020 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1021 * Return the filter ID on success, or a negative error code. 1022 * 1023 * Slave management functions (for bridge, bonding, etc). 1024 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1025 * Called to make another netdev an underling. 1026 * 1027 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1028 * Called to release previously enslaved netdev. 1029 * 1030 * Feature/offload setting functions. 1031 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1032 * Called to update device configuration to new features. Passed 1033 * feature set might be less than what was returned by ndo_fix_features()). 1034 * Must return >0 or -errno if it changed dev->features itself. 1035 * 1036 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1037 * struct net_device *dev, 1038 * const unsigned char *addr, u16 vid, u16 flags) 1039 * Adds an FDB entry to dev for addr. 1040 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1041 * struct net_device *dev, 1042 * const unsigned char *addr, u16 vid) 1043 * Deletes the FDB entry from dev coresponding to addr. 1044 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1045 * struct net_device *dev, struct net_device *filter_dev, 1046 * int *idx) 1047 * Used to add FDB entries to dump requests. Implementers should add 1048 * entries to skb and update idx with the number of entries. 1049 * 1050 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1051 * u16 flags) 1052 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1053 * struct net_device *dev, u32 filter_mask, 1054 * int nlflags) 1055 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1056 * u16 flags); 1057 * 1058 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1059 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1060 * which do not represent real hardware may define this to allow their 1061 * userspace components to manage their virtual carrier state. Devices 1062 * that determine carrier state from physical hardware properties (eg 1063 * network cables) or protocol-dependent mechanisms (eg 1064 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1065 * 1066 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1067 * struct netdev_phys_item_id *ppid); 1068 * Called to get ID of physical port of this device. If driver does 1069 * not implement this, it is assumed that the hw is not able to have 1070 * multiple net devices on single physical port. 1071 * 1072 * void (*ndo_udp_tunnel_add)(struct net_device *dev, 1073 * struct udp_tunnel_info *ti); 1074 * Called by UDP tunnel to notify a driver about the UDP port and socket 1075 * address family that a UDP tunnel is listnening to. It is called only 1076 * when a new port starts listening. The operation is protected by the 1077 * RTNL. 1078 * 1079 * void (*ndo_udp_tunnel_del)(struct net_device *dev, 1080 * struct udp_tunnel_info *ti); 1081 * Called by UDP tunnel to notify the driver about a UDP port and socket 1082 * address family that the UDP tunnel is not listening to anymore. The 1083 * operation is protected by the RTNL. 1084 * 1085 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1086 * struct net_device *dev) 1087 * Called by upper layer devices to accelerate switching or other 1088 * station functionality into hardware. 'pdev is the lowerdev 1089 * to use for the offload and 'dev' is the net device that will 1090 * back the offload. Returns a pointer to the private structure 1091 * the upper layer will maintain. 1092 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1093 * Called by upper layer device to delete the station created 1094 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1095 * the station and priv is the structure returned by the add 1096 * operation. 1097 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb, 1098 * struct net_device *dev, 1099 * void *priv); 1100 * Callback to use for xmit over the accelerated station. This 1101 * is used in place of ndo_start_xmit on accelerated net 1102 * devices. 1103 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1104 * struct net_device *dev 1105 * netdev_features_t features); 1106 * Called by core transmit path to determine if device is capable of 1107 * performing offload operations on a given packet. This is to give 1108 * the device an opportunity to implement any restrictions that cannot 1109 * be otherwise expressed by feature flags. The check is called with 1110 * the set of features that the stack has calculated and it returns 1111 * those the driver believes to be appropriate. 1112 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1113 * int queue_index, u32 maxrate); 1114 * Called when a user wants to set a max-rate limitation of specific 1115 * TX queue. 1116 * int (*ndo_get_iflink)(const struct net_device *dev); 1117 * Called to get the iflink value of this device. 1118 * void (*ndo_change_proto_down)(struct net_device *dev, 1119 * bool proto_down); 1120 * This function is used to pass protocol port error state information 1121 * to the switch driver. The switch driver can react to the proto_down 1122 * by doing a phys down on the associated switch port. 1123 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1124 * This function is used to get egress tunnel information for given skb. 1125 * This is useful for retrieving outer tunnel header parameters while 1126 * sampling packet. 1127 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1128 * This function is used to specify the headroom that the skb must 1129 * consider when allocation skb during packet reception. Setting 1130 * appropriate rx headroom value allows avoiding skb head copy on 1131 * forward. Setting a negative value resets the rx headroom to the 1132 * default value. 1133 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp); 1134 * This function is used to set or query state related to XDP on the 1135 * netdevice. See definition of enum xdp_netdev_command for details. 1136 * 1137 */ 1138 struct net_device_ops { 1139 int (*ndo_init)(struct net_device *dev); 1140 void (*ndo_uninit)(struct net_device *dev); 1141 int (*ndo_open)(struct net_device *dev); 1142 int (*ndo_stop)(struct net_device *dev); 1143 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1144 struct net_device *dev); 1145 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1146 struct net_device *dev, 1147 netdev_features_t features); 1148 u16 (*ndo_select_queue)(struct net_device *dev, 1149 struct sk_buff *skb, 1150 void *accel_priv, 1151 select_queue_fallback_t fallback); 1152 void (*ndo_change_rx_flags)(struct net_device *dev, 1153 int flags); 1154 void (*ndo_set_rx_mode)(struct net_device *dev); 1155 int (*ndo_set_mac_address)(struct net_device *dev, 1156 void *addr); 1157 int (*ndo_validate_addr)(struct net_device *dev); 1158 int (*ndo_do_ioctl)(struct net_device *dev, 1159 struct ifreq *ifr, int cmd); 1160 int (*ndo_set_config)(struct net_device *dev, 1161 struct ifmap *map); 1162 int (*ndo_change_mtu)(struct net_device *dev, 1163 int new_mtu); 1164 int (*ndo_neigh_setup)(struct net_device *dev, 1165 struct neigh_parms *); 1166 void (*ndo_tx_timeout) (struct net_device *dev); 1167 1168 void (*ndo_get_stats64)(struct net_device *dev, 1169 struct rtnl_link_stats64 *storage); 1170 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1171 int (*ndo_get_offload_stats)(int attr_id, 1172 const struct net_device *dev, 1173 void *attr_data); 1174 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1175 1176 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1177 __be16 proto, u16 vid); 1178 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1179 __be16 proto, u16 vid); 1180 #ifdef CONFIG_NET_POLL_CONTROLLER 1181 void (*ndo_poll_controller)(struct net_device *dev); 1182 int (*ndo_netpoll_setup)(struct net_device *dev, 1183 struct netpoll_info *info); 1184 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1185 #endif 1186 #ifdef CONFIG_NET_RX_BUSY_POLL 1187 int (*ndo_busy_poll)(struct napi_struct *dev); 1188 #endif 1189 int (*ndo_set_vf_mac)(struct net_device *dev, 1190 int queue, u8 *mac); 1191 int (*ndo_set_vf_vlan)(struct net_device *dev, 1192 int queue, u16 vlan, 1193 u8 qos, __be16 proto); 1194 int (*ndo_set_vf_rate)(struct net_device *dev, 1195 int vf, int min_tx_rate, 1196 int max_tx_rate); 1197 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1198 int vf, bool setting); 1199 int (*ndo_set_vf_trust)(struct net_device *dev, 1200 int vf, bool setting); 1201 int (*ndo_get_vf_config)(struct net_device *dev, 1202 int vf, 1203 struct ifla_vf_info *ivf); 1204 int (*ndo_set_vf_link_state)(struct net_device *dev, 1205 int vf, int link_state); 1206 int (*ndo_get_vf_stats)(struct net_device *dev, 1207 int vf, 1208 struct ifla_vf_stats 1209 *vf_stats); 1210 int (*ndo_set_vf_port)(struct net_device *dev, 1211 int vf, 1212 struct nlattr *port[]); 1213 int (*ndo_get_vf_port)(struct net_device *dev, 1214 int vf, struct sk_buff *skb); 1215 int (*ndo_set_vf_guid)(struct net_device *dev, 1216 int vf, u64 guid, 1217 int guid_type); 1218 int (*ndo_set_vf_rss_query_en)( 1219 struct net_device *dev, 1220 int vf, bool setting); 1221 int (*ndo_setup_tc)(struct net_device *dev, 1222 u32 handle, 1223 __be16 protocol, 1224 struct tc_to_netdev *tc); 1225 #if IS_ENABLED(CONFIG_FCOE) 1226 int (*ndo_fcoe_enable)(struct net_device *dev); 1227 int (*ndo_fcoe_disable)(struct net_device *dev); 1228 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1229 u16 xid, 1230 struct scatterlist *sgl, 1231 unsigned int sgc); 1232 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1233 u16 xid); 1234 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1235 u16 xid, 1236 struct scatterlist *sgl, 1237 unsigned int sgc); 1238 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1239 struct netdev_fcoe_hbainfo *hbainfo); 1240 #endif 1241 1242 #if IS_ENABLED(CONFIG_LIBFCOE) 1243 #define NETDEV_FCOE_WWNN 0 1244 #define NETDEV_FCOE_WWPN 1 1245 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1246 u64 *wwn, int type); 1247 #endif 1248 1249 #ifdef CONFIG_RFS_ACCEL 1250 int (*ndo_rx_flow_steer)(struct net_device *dev, 1251 const struct sk_buff *skb, 1252 u16 rxq_index, 1253 u32 flow_id); 1254 #endif 1255 int (*ndo_add_slave)(struct net_device *dev, 1256 struct net_device *slave_dev); 1257 int (*ndo_del_slave)(struct net_device *dev, 1258 struct net_device *slave_dev); 1259 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1260 netdev_features_t features); 1261 int (*ndo_set_features)(struct net_device *dev, 1262 netdev_features_t features); 1263 int (*ndo_neigh_construct)(struct net_device *dev, 1264 struct neighbour *n); 1265 void (*ndo_neigh_destroy)(struct net_device *dev, 1266 struct neighbour *n); 1267 1268 int (*ndo_fdb_add)(struct ndmsg *ndm, 1269 struct nlattr *tb[], 1270 struct net_device *dev, 1271 const unsigned char *addr, 1272 u16 vid, 1273 u16 flags); 1274 int (*ndo_fdb_del)(struct ndmsg *ndm, 1275 struct nlattr *tb[], 1276 struct net_device *dev, 1277 const unsigned char *addr, 1278 u16 vid); 1279 int (*ndo_fdb_dump)(struct sk_buff *skb, 1280 struct netlink_callback *cb, 1281 struct net_device *dev, 1282 struct net_device *filter_dev, 1283 int *idx); 1284 1285 int (*ndo_bridge_setlink)(struct net_device *dev, 1286 struct nlmsghdr *nlh, 1287 u16 flags); 1288 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1289 u32 pid, u32 seq, 1290 struct net_device *dev, 1291 u32 filter_mask, 1292 int nlflags); 1293 int (*ndo_bridge_dellink)(struct net_device *dev, 1294 struct nlmsghdr *nlh, 1295 u16 flags); 1296 int (*ndo_change_carrier)(struct net_device *dev, 1297 bool new_carrier); 1298 int (*ndo_get_phys_port_id)(struct net_device *dev, 1299 struct netdev_phys_item_id *ppid); 1300 int (*ndo_get_phys_port_name)(struct net_device *dev, 1301 char *name, size_t len); 1302 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1303 struct udp_tunnel_info *ti); 1304 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1305 struct udp_tunnel_info *ti); 1306 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1307 struct net_device *dev); 1308 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1309 void *priv); 1310 1311 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb, 1312 struct net_device *dev, 1313 void *priv); 1314 int (*ndo_get_lock_subclass)(struct net_device *dev); 1315 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1316 int queue_index, 1317 u32 maxrate); 1318 int (*ndo_get_iflink)(const struct net_device *dev); 1319 int (*ndo_change_proto_down)(struct net_device *dev, 1320 bool proto_down); 1321 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1322 struct sk_buff *skb); 1323 void (*ndo_set_rx_headroom)(struct net_device *dev, 1324 int needed_headroom); 1325 int (*ndo_xdp)(struct net_device *dev, 1326 struct netdev_xdp *xdp); 1327 }; 1328 1329 /** 1330 * enum net_device_priv_flags - &struct net_device priv_flags 1331 * 1332 * These are the &struct net_device, they are only set internally 1333 * by drivers and used in the kernel. These flags are invisible to 1334 * userspace; this means that the order of these flags can change 1335 * during any kernel release. 1336 * 1337 * You should have a pretty good reason to be extending these flags. 1338 * 1339 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1340 * @IFF_EBRIDGE: Ethernet bridging device 1341 * @IFF_BONDING: bonding master or slave 1342 * @IFF_ISATAP: ISATAP interface (RFC4214) 1343 * @IFF_WAN_HDLC: WAN HDLC device 1344 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1345 * release skb->dst 1346 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1347 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1348 * @IFF_MACVLAN_PORT: device used as macvlan port 1349 * @IFF_BRIDGE_PORT: device used as bridge port 1350 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1351 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1352 * @IFF_UNICAST_FLT: Supports unicast filtering 1353 * @IFF_TEAM_PORT: device used as team port 1354 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1355 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1356 * change when it's running 1357 * @IFF_MACVLAN: Macvlan device 1358 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1359 * underlying stacked devices 1360 * @IFF_IPVLAN_MASTER: IPvlan master device 1361 * @IFF_IPVLAN_SLAVE: IPvlan slave device 1362 * @IFF_L3MDEV_MASTER: device is an L3 master device 1363 * @IFF_NO_QUEUE: device can run without qdisc attached 1364 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1365 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1366 * @IFF_TEAM: device is a team device 1367 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1368 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1369 * entity (i.e. the master device for bridged veth) 1370 * @IFF_MACSEC: device is a MACsec device 1371 */ 1372 enum netdev_priv_flags { 1373 IFF_802_1Q_VLAN = 1<<0, 1374 IFF_EBRIDGE = 1<<1, 1375 IFF_BONDING = 1<<2, 1376 IFF_ISATAP = 1<<3, 1377 IFF_WAN_HDLC = 1<<4, 1378 IFF_XMIT_DST_RELEASE = 1<<5, 1379 IFF_DONT_BRIDGE = 1<<6, 1380 IFF_DISABLE_NETPOLL = 1<<7, 1381 IFF_MACVLAN_PORT = 1<<8, 1382 IFF_BRIDGE_PORT = 1<<9, 1383 IFF_OVS_DATAPATH = 1<<10, 1384 IFF_TX_SKB_SHARING = 1<<11, 1385 IFF_UNICAST_FLT = 1<<12, 1386 IFF_TEAM_PORT = 1<<13, 1387 IFF_SUPP_NOFCS = 1<<14, 1388 IFF_LIVE_ADDR_CHANGE = 1<<15, 1389 IFF_MACVLAN = 1<<16, 1390 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1391 IFF_IPVLAN_MASTER = 1<<18, 1392 IFF_IPVLAN_SLAVE = 1<<19, 1393 IFF_L3MDEV_MASTER = 1<<20, 1394 IFF_NO_QUEUE = 1<<21, 1395 IFF_OPENVSWITCH = 1<<22, 1396 IFF_L3MDEV_SLAVE = 1<<23, 1397 IFF_TEAM = 1<<24, 1398 IFF_RXFH_CONFIGURED = 1<<25, 1399 IFF_PHONY_HEADROOM = 1<<26, 1400 IFF_MACSEC = 1<<27, 1401 }; 1402 1403 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1404 #define IFF_EBRIDGE IFF_EBRIDGE 1405 #define IFF_BONDING IFF_BONDING 1406 #define IFF_ISATAP IFF_ISATAP 1407 #define IFF_WAN_HDLC IFF_WAN_HDLC 1408 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1409 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1410 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1411 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1412 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1413 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1414 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1415 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1416 #define IFF_TEAM_PORT IFF_TEAM_PORT 1417 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1418 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1419 #define IFF_MACVLAN IFF_MACVLAN 1420 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1421 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER 1422 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE 1423 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1424 #define IFF_NO_QUEUE IFF_NO_QUEUE 1425 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1426 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1427 #define IFF_TEAM IFF_TEAM 1428 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1429 #define IFF_MACSEC IFF_MACSEC 1430 1431 /** 1432 * struct net_device - The DEVICE structure. 1433 * Actually, this whole structure is a big mistake. It mixes I/O 1434 * data with strictly "high-level" data, and it has to know about 1435 * almost every data structure used in the INET module. 1436 * 1437 * @name: This is the first field of the "visible" part of this structure 1438 * (i.e. as seen by users in the "Space.c" file). It is the name 1439 * of the interface. 1440 * 1441 * @name_hlist: Device name hash chain, please keep it close to name[] 1442 * @ifalias: SNMP alias 1443 * @mem_end: Shared memory end 1444 * @mem_start: Shared memory start 1445 * @base_addr: Device I/O address 1446 * @irq: Device IRQ number 1447 * 1448 * @carrier_changes: Stats to monitor carrier on<->off transitions 1449 * 1450 * @state: Generic network queuing layer state, see netdev_state_t 1451 * @dev_list: The global list of network devices 1452 * @napi_list: List entry used for polling NAPI devices 1453 * @unreg_list: List entry when we are unregistering the 1454 * device; see the function unregister_netdev 1455 * @close_list: List entry used when we are closing the device 1456 * @ptype_all: Device-specific packet handlers for all protocols 1457 * @ptype_specific: Device-specific, protocol-specific packet handlers 1458 * 1459 * @adj_list: Directly linked devices, like slaves for bonding 1460 * @features: Currently active device features 1461 * @hw_features: User-changeable features 1462 * 1463 * @wanted_features: User-requested features 1464 * @vlan_features: Mask of features inheritable by VLAN devices 1465 * 1466 * @hw_enc_features: Mask of features inherited by encapsulating devices 1467 * This field indicates what encapsulation 1468 * offloads the hardware is capable of doing, 1469 * and drivers will need to set them appropriately. 1470 * 1471 * @mpls_features: Mask of features inheritable by MPLS 1472 * 1473 * @ifindex: interface index 1474 * @group: The group the device belongs to 1475 * 1476 * @stats: Statistics struct, which was left as a legacy, use 1477 * rtnl_link_stats64 instead 1478 * 1479 * @rx_dropped: Dropped packets by core network, 1480 * do not use this in drivers 1481 * @tx_dropped: Dropped packets by core network, 1482 * do not use this in drivers 1483 * @rx_nohandler: nohandler dropped packets by core network on 1484 * inactive devices, do not use this in drivers 1485 * 1486 * @wireless_handlers: List of functions to handle Wireless Extensions, 1487 * instead of ioctl, 1488 * see <net/iw_handler.h> for details. 1489 * @wireless_data: Instance data managed by the core of wireless extensions 1490 * 1491 * @netdev_ops: Includes several pointers to callbacks, 1492 * if one wants to override the ndo_*() functions 1493 * @ethtool_ops: Management operations 1494 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1495 * discovery handling. Necessary for e.g. 6LoWPAN. 1496 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1497 * of Layer 2 headers. 1498 * 1499 * @flags: Interface flags (a la BSD) 1500 * @priv_flags: Like 'flags' but invisible to userspace, 1501 * see if.h for the definitions 1502 * @gflags: Global flags ( kept as legacy ) 1503 * @padded: How much padding added by alloc_netdev() 1504 * @operstate: RFC2863 operstate 1505 * @link_mode: Mapping policy to operstate 1506 * @if_port: Selectable AUI, TP, ... 1507 * @dma: DMA channel 1508 * @mtu: Interface MTU value 1509 * @min_mtu: Interface Minimum MTU value 1510 * @max_mtu: Interface Maximum MTU value 1511 * @type: Interface hardware type 1512 * @hard_header_len: Maximum hardware header length. 1513 * 1514 * @needed_headroom: Extra headroom the hardware may need, but not in all 1515 * cases can this be guaranteed 1516 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1517 * cases can this be guaranteed. Some cases also use 1518 * LL_MAX_HEADER instead to allocate the skb 1519 * 1520 * interface address info: 1521 * 1522 * @perm_addr: Permanent hw address 1523 * @addr_assign_type: Hw address assignment type 1524 * @addr_len: Hardware address length 1525 * @neigh_priv_len: Used in neigh_alloc() 1526 * @dev_id: Used to differentiate devices that share 1527 * the same link layer address 1528 * @dev_port: Used to differentiate devices that share 1529 * the same function 1530 * @addr_list_lock: XXX: need comments on this one 1531 * @uc_promisc: Counter that indicates promiscuous mode 1532 * has been enabled due to the need to listen to 1533 * additional unicast addresses in a device that 1534 * does not implement ndo_set_rx_mode() 1535 * @uc: unicast mac addresses 1536 * @mc: multicast mac addresses 1537 * @dev_addrs: list of device hw addresses 1538 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1539 * @promiscuity: Number of times the NIC is told to work in 1540 * promiscuous mode; if it becomes 0 the NIC will 1541 * exit promiscuous mode 1542 * @allmulti: Counter, enables or disables allmulticast mode 1543 * 1544 * @vlan_info: VLAN info 1545 * @dsa_ptr: dsa specific data 1546 * @tipc_ptr: TIPC specific data 1547 * @atalk_ptr: AppleTalk link 1548 * @ip_ptr: IPv4 specific data 1549 * @dn_ptr: DECnet specific data 1550 * @ip6_ptr: IPv6 specific data 1551 * @ax25_ptr: AX.25 specific data 1552 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1553 * 1554 * @dev_addr: Hw address (before bcast, 1555 * because most packets are unicast) 1556 * 1557 * @_rx: Array of RX queues 1558 * @num_rx_queues: Number of RX queues 1559 * allocated at register_netdev() time 1560 * @real_num_rx_queues: Number of RX queues currently active in device 1561 * 1562 * @rx_handler: handler for received packets 1563 * @rx_handler_data: XXX: need comments on this one 1564 * @ingress_queue: XXX: need comments on this one 1565 * @broadcast: hw bcast address 1566 * 1567 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1568 * indexed by RX queue number. Assigned by driver. 1569 * This must only be set if the ndo_rx_flow_steer 1570 * operation is defined 1571 * @index_hlist: Device index hash chain 1572 * 1573 * @_tx: Array of TX queues 1574 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1575 * @real_num_tx_queues: Number of TX queues currently active in device 1576 * @qdisc: Root qdisc from userspace point of view 1577 * @tx_queue_len: Max frames per queue allowed 1578 * @tx_global_lock: XXX: need comments on this one 1579 * 1580 * @xps_maps: XXX: need comments on this one 1581 * 1582 * @watchdog_timeo: Represents the timeout that is used by 1583 * the watchdog (see dev_watchdog()) 1584 * @watchdog_timer: List of timers 1585 * 1586 * @pcpu_refcnt: Number of references to this device 1587 * @todo_list: Delayed register/unregister 1588 * @link_watch_list: XXX: need comments on this one 1589 * 1590 * @reg_state: Register/unregister state machine 1591 * @dismantle: Device is going to be freed 1592 * @rtnl_link_state: This enum represents the phases of creating 1593 * a new link 1594 * 1595 * @destructor: Called from unregister, 1596 * can be used to call free_netdev 1597 * @npinfo: XXX: need comments on this one 1598 * @nd_net: Network namespace this network device is inside 1599 * 1600 * @ml_priv: Mid-layer private 1601 * @lstats: Loopback statistics 1602 * @tstats: Tunnel statistics 1603 * @dstats: Dummy statistics 1604 * @vstats: Virtual ethernet statistics 1605 * 1606 * @garp_port: GARP 1607 * @mrp_port: MRP 1608 * 1609 * @dev: Class/net/name entry 1610 * @sysfs_groups: Space for optional device, statistics and wireless 1611 * sysfs groups 1612 * 1613 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1614 * @rtnl_link_ops: Rtnl_link_ops 1615 * 1616 * @gso_max_size: Maximum size of generic segmentation offload 1617 * @gso_max_segs: Maximum number of segments that can be passed to the 1618 * NIC for GSO 1619 * 1620 * @dcbnl_ops: Data Center Bridging netlink ops 1621 * @num_tc: Number of traffic classes in the net device 1622 * @tc_to_txq: XXX: need comments on this one 1623 * @prio_tc_map: XXX: need comments on this one 1624 * 1625 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1626 * 1627 * @priomap: XXX: need comments on this one 1628 * @phydev: Physical device may attach itself 1629 * for hardware timestamping 1630 * 1631 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1632 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1633 * 1634 * @proto_down: protocol port state information can be sent to the 1635 * switch driver and used to set the phys state of the 1636 * switch port. 1637 * 1638 * FIXME: cleanup struct net_device such that network protocol info 1639 * moves out. 1640 */ 1641 1642 struct net_device { 1643 char name[IFNAMSIZ]; 1644 struct hlist_node name_hlist; 1645 char *ifalias; 1646 /* 1647 * I/O specific fields 1648 * FIXME: Merge these and struct ifmap into one 1649 */ 1650 unsigned long mem_end; 1651 unsigned long mem_start; 1652 unsigned long base_addr; 1653 int irq; 1654 1655 atomic_t carrier_changes; 1656 1657 /* 1658 * Some hardware also needs these fields (state,dev_list, 1659 * napi_list,unreg_list,close_list) but they are not 1660 * part of the usual set specified in Space.c. 1661 */ 1662 1663 unsigned long state; 1664 1665 struct list_head dev_list; 1666 struct list_head napi_list; 1667 struct list_head unreg_list; 1668 struct list_head close_list; 1669 struct list_head ptype_all; 1670 struct list_head ptype_specific; 1671 1672 struct { 1673 struct list_head upper; 1674 struct list_head lower; 1675 } adj_list; 1676 1677 netdev_features_t features; 1678 netdev_features_t hw_features; 1679 netdev_features_t wanted_features; 1680 netdev_features_t vlan_features; 1681 netdev_features_t hw_enc_features; 1682 netdev_features_t mpls_features; 1683 netdev_features_t gso_partial_features; 1684 1685 int ifindex; 1686 int group; 1687 1688 struct net_device_stats stats; 1689 1690 atomic_long_t rx_dropped; 1691 atomic_long_t tx_dropped; 1692 atomic_long_t rx_nohandler; 1693 1694 #ifdef CONFIG_WIRELESS_EXT 1695 const struct iw_handler_def *wireless_handlers; 1696 struct iw_public_data *wireless_data; 1697 #endif 1698 const struct net_device_ops *netdev_ops; 1699 const struct ethtool_ops *ethtool_ops; 1700 #ifdef CONFIG_NET_SWITCHDEV 1701 const struct switchdev_ops *switchdev_ops; 1702 #endif 1703 #ifdef CONFIG_NET_L3_MASTER_DEV 1704 const struct l3mdev_ops *l3mdev_ops; 1705 #endif 1706 #if IS_ENABLED(CONFIG_IPV6) 1707 const struct ndisc_ops *ndisc_ops; 1708 #endif 1709 1710 const struct header_ops *header_ops; 1711 1712 unsigned int flags; 1713 unsigned int priv_flags; 1714 1715 unsigned short gflags; 1716 unsigned short padded; 1717 1718 unsigned char operstate; 1719 unsigned char link_mode; 1720 1721 unsigned char if_port; 1722 unsigned char dma; 1723 1724 unsigned int mtu; 1725 unsigned int min_mtu; 1726 unsigned int max_mtu; 1727 unsigned short type; 1728 unsigned short hard_header_len; 1729 1730 unsigned short needed_headroom; 1731 unsigned short needed_tailroom; 1732 1733 /* Interface address info. */ 1734 unsigned char perm_addr[MAX_ADDR_LEN]; 1735 unsigned char addr_assign_type; 1736 unsigned char addr_len; 1737 unsigned short neigh_priv_len; 1738 unsigned short dev_id; 1739 unsigned short dev_port; 1740 spinlock_t addr_list_lock; 1741 unsigned char name_assign_type; 1742 bool uc_promisc; 1743 struct netdev_hw_addr_list uc; 1744 struct netdev_hw_addr_list mc; 1745 struct netdev_hw_addr_list dev_addrs; 1746 1747 #ifdef CONFIG_SYSFS 1748 struct kset *queues_kset; 1749 #endif 1750 unsigned int promiscuity; 1751 unsigned int allmulti; 1752 1753 1754 /* Protocol-specific pointers */ 1755 1756 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1757 struct vlan_info __rcu *vlan_info; 1758 #endif 1759 #if IS_ENABLED(CONFIG_NET_DSA) 1760 struct dsa_switch_tree *dsa_ptr; 1761 #endif 1762 #if IS_ENABLED(CONFIG_TIPC) 1763 struct tipc_bearer __rcu *tipc_ptr; 1764 #endif 1765 void *atalk_ptr; 1766 struct in_device __rcu *ip_ptr; 1767 struct dn_dev __rcu *dn_ptr; 1768 struct inet6_dev __rcu *ip6_ptr; 1769 void *ax25_ptr; 1770 struct wireless_dev *ieee80211_ptr; 1771 struct wpan_dev *ieee802154_ptr; 1772 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 1773 struct mpls_dev __rcu *mpls_ptr; 1774 #endif 1775 1776 /* 1777 * Cache lines mostly used on receive path (including eth_type_trans()) 1778 */ 1779 /* Interface address info used in eth_type_trans() */ 1780 unsigned char *dev_addr; 1781 1782 #ifdef CONFIG_SYSFS 1783 struct netdev_rx_queue *_rx; 1784 1785 unsigned int num_rx_queues; 1786 unsigned int real_num_rx_queues; 1787 #endif 1788 1789 unsigned long gro_flush_timeout; 1790 rx_handler_func_t __rcu *rx_handler; 1791 void __rcu *rx_handler_data; 1792 1793 #ifdef CONFIG_NET_CLS_ACT 1794 struct tcf_proto __rcu *ingress_cl_list; 1795 #endif 1796 struct netdev_queue __rcu *ingress_queue; 1797 #ifdef CONFIG_NETFILTER_INGRESS 1798 struct nf_hook_entry __rcu *nf_hooks_ingress; 1799 #endif 1800 1801 unsigned char broadcast[MAX_ADDR_LEN]; 1802 #ifdef CONFIG_RFS_ACCEL 1803 struct cpu_rmap *rx_cpu_rmap; 1804 #endif 1805 struct hlist_node index_hlist; 1806 1807 /* 1808 * Cache lines mostly used on transmit path 1809 */ 1810 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1811 unsigned int num_tx_queues; 1812 unsigned int real_num_tx_queues; 1813 struct Qdisc *qdisc; 1814 #ifdef CONFIG_NET_SCHED 1815 DECLARE_HASHTABLE (qdisc_hash, 4); 1816 #endif 1817 unsigned long tx_queue_len; 1818 spinlock_t tx_global_lock; 1819 int watchdog_timeo; 1820 1821 #ifdef CONFIG_XPS 1822 struct xps_dev_maps __rcu *xps_maps; 1823 #endif 1824 #ifdef CONFIG_NET_CLS_ACT 1825 struct tcf_proto __rcu *egress_cl_list; 1826 #endif 1827 1828 /* These may be needed for future network-power-down code. */ 1829 struct timer_list watchdog_timer; 1830 1831 int __percpu *pcpu_refcnt; 1832 struct list_head todo_list; 1833 1834 struct list_head link_watch_list; 1835 1836 enum { NETREG_UNINITIALIZED=0, 1837 NETREG_REGISTERED, /* completed register_netdevice */ 1838 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1839 NETREG_UNREGISTERED, /* completed unregister todo */ 1840 NETREG_RELEASED, /* called free_netdev */ 1841 NETREG_DUMMY, /* dummy device for NAPI poll */ 1842 } reg_state:8; 1843 1844 bool dismantle; 1845 1846 enum { 1847 RTNL_LINK_INITIALIZED, 1848 RTNL_LINK_INITIALIZING, 1849 } rtnl_link_state:16; 1850 1851 void (*destructor)(struct net_device *dev); 1852 1853 #ifdef CONFIG_NETPOLL 1854 struct netpoll_info __rcu *npinfo; 1855 #endif 1856 1857 possible_net_t nd_net; 1858 1859 /* mid-layer private */ 1860 union { 1861 void *ml_priv; 1862 struct pcpu_lstats __percpu *lstats; 1863 struct pcpu_sw_netstats __percpu *tstats; 1864 struct pcpu_dstats __percpu *dstats; 1865 struct pcpu_vstats __percpu *vstats; 1866 }; 1867 1868 struct garp_port __rcu *garp_port; 1869 struct mrp_port __rcu *mrp_port; 1870 1871 struct device dev; 1872 const struct attribute_group *sysfs_groups[4]; 1873 const struct attribute_group *sysfs_rx_queue_group; 1874 1875 const struct rtnl_link_ops *rtnl_link_ops; 1876 1877 /* for setting kernel sock attribute on TCP connection setup */ 1878 #define GSO_MAX_SIZE 65536 1879 unsigned int gso_max_size; 1880 #define GSO_MAX_SEGS 65535 1881 u16 gso_max_segs; 1882 1883 #ifdef CONFIG_DCB 1884 const struct dcbnl_rtnl_ops *dcbnl_ops; 1885 #endif 1886 u8 num_tc; 1887 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 1888 u8 prio_tc_map[TC_BITMASK + 1]; 1889 1890 #if IS_ENABLED(CONFIG_FCOE) 1891 unsigned int fcoe_ddp_xid; 1892 #endif 1893 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 1894 struct netprio_map __rcu *priomap; 1895 #endif 1896 struct phy_device *phydev; 1897 struct lock_class_key *qdisc_tx_busylock; 1898 struct lock_class_key *qdisc_running_key; 1899 bool proto_down; 1900 }; 1901 #define to_net_dev(d) container_of(d, struct net_device, dev) 1902 1903 #define NETDEV_ALIGN 32 1904 1905 static inline 1906 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 1907 { 1908 return dev->prio_tc_map[prio & TC_BITMASK]; 1909 } 1910 1911 static inline 1912 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 1913 { 1914 if (tc >= dev->num_tc) 1915 return -EINVAL; 1916 1917 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 1918 return 0; 1919 } 1920 1921 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 1922 void netdev_reset_tc(struct net_device *dev); 1923 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 1924 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 1925 1926 static inline 1927 int netdev_get_num_tc(struct net_device *dev) 1928 { 1929 return dev->num_tc; 1930 } 1931 1932 static inline 1933 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 1934 unsigned int index) 1935 { 1936 return &dev->_tx[index]; 1937 } 1938 1939 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 1940 const struct sk_buff *skb) 1941 { 1942 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 1943 } 1944 1945 static inline void netdev_for_each_tx_queue(struct net_device *dev, 1946 void (*f)(struct net_device *, 1947 struct netdev_queue *, 1948 void *), 1949 void *arg) 1950 { 1951 unsigned int i; 1952 1953 for (i = 0; i < dev->num_tx_queues; i++) 1954 f(dev, &dev->_tx[i], arg); 1955 } 1956 1957 #define netdev_lockdep_set_classes(dev) \ 1958 { \ 1959 static struct lock_class_key qdisc_tx_busylock_key; \ 1960 static struct lock_class_key qdisc_running_key; \ 1961 static struct lock_class_key qdisc_xmit_lock_key; \ 1962 static struct lock_class_key dev_addr_list_lock_key; \ 1963 unsigned int i; \ 1964 \ 1965 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 1966 (dev)->qdisc_running_key = &qdisc_running_key; \ 1967 lockdep_set_class(&(dev)->addr_list_lock, \ 1968 &dev_addr_list_lock_key); \ 1969 for (i = 0; i < (dev)->num_tx_queues; i++) \ 1970 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 1971 &qdisc_xmit_lock_key); \ 1972 } 1973 1974 struct netdev_queue *netdev_pick_tx(struct net_device *dev, 1975 struct sk_buff *skb, 1976 void *accel_priv); 1977 1978 /* returns the headroom that the master device needs to take in account 1979 * when forwarding to this dev 1980 */ 1981 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 1982 { 1983 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 1984 } 1985 1986 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 1987 { 1988 if (dev->netdev_ops->ndo_set_rx_headroom) 1989 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 1990 } 1991 1992 /* set the device rx headroom to the dev's default */ 1993 static inline void netdev_reset_rx_headroom(struct net_device *dev) 1994 { 1995 netdev_set_rx_headroom(dev, -1); 1996 } 1997 1998 /* 1999 * Net namespace inlines 2000 */ 2001 static inline 2002 struct net *dev_net(const struct net_device *dev) 2003 { 2004 return read_pnet(&dev->nd_net); 2005 } 2006 2007 static inline 2008 void dev_net_set(struct net_device *dev, struct net *net) 2009 { 2010 write_pnet(&dev->nd_net, net); 2011 } 2012 2013 static inline bool netdev_uses_dsa(struct net_device *dev) 2014 { 2015 #if IS_ENABLED(CONFIG_NET_DSA) 2016 if (dev->dsa_ptr != NULL) 2017 return dsa_uses_tagged_protocol(dev->dsa_ptr); 2018 #endif 2019 return false; 2020 } 2021 2022 /** 2023 * netdev_priv - access network device private data 2024 * @dev: network device 2025 * 2026 * Get network device private data 2027 */ 2028 static inline void *netdev_priv(const struct net_device *dev) 2029 { 2030 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2031 } 2032 2033 /* Set the sysfs physical device reference for the network logical device 2034 * if set prior to registration will cause a symlink during initialization. 2035 */ 2036 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2037 2038 /* Set the sysfs device type for the network logical device to allow 2039 * fine-grained identification of different network device types. For 2040 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2041 */ 2042 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2043 2044 /* Default NAPI poll() weight 2045 * Device drivers are strongly advised to not use bigger value 2046 */ 2047 #define NAPI_POLL_WEIGHT 64 2048 2049 /** 2050 * netif_napi_add - initialize a NAPI context 2051 * @dev: network device 2052 * @napi: NAPI context 2053 * @poll: polling function 2054 * @weight: default weight 2055 * 2056 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2057 * *any* of the other NAPI-related functions. 2058 */ 2059 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2060 int (*poll)(struct napi_struct *, int), int weight); 2061 2062 /** 2063 * netif_tx_napi_add - initialize a NAPI context 2064 * @dev: network device 2065 * @napi: NAPI context 2066 * @poll: polling function 2067 * @weight: default weight 2068 * 2069 * This variant of netif_napi_add() should be used from drivers using NAPI 2070 * to exclusively poll a TX queue. 2071 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2072 */ 2073 static inline void netif_tx_napi_add(struct net_device *dev, 2074 struct napi_struct *napi, 2075 int (*poll)(struct napi_struct *, int), 2076 int weight) 2077 { 2078 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2079 netif_napi_add(dev, napi, poll, weight); 2080 } 2081 2082 /** 2083 * netif_napi_del - remove a NAPI context 2084 * @napi: NAPI context 2085 * 2086 * netif_napi_del() removes a NAPI context from the network device NAPI list 2087 */ 2088 void netif_napi_del(struct napi_struct *napi); 2089 2090 struct napi_gro_cb { 2091 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2092 void *frag0; 2093 2094 /* Length of frag0. */ 2095 unsigned int frag0_len; 2096 2097 /* This indicates where we are processing relative to skb->data. */ 2098 int data_offset; 2099 2100 /* This is non-zero if the packet cannot be merged with the new skb. */ 2101 u16 flush; 2102 2103 /* Save the IP ID here and check when we get to the transport layer */ 2104 u16 flush_id; 2105 2106 /* Number of segments aggregated. */ 2107 u16 count; 2108 2109 /* Start offset for remote checksum offload */ 2110 u16 gro_remcsum_start; 2111 2112 /* jiffies when first packet was created/queued */ 2113 unsigned long age; 2114 2115 /* Used in ipv6_gro_receive() and foo-over-udp */ 2116 u16 proto; 2117 2118 /* This is non-zero if the packet may be of the same flow. */ 2119 u8 same_flow:1; 2120 2121 /* Used in tunnel GRO receive */ 2122 u8 encap_mark:1; 2123 2124 /* GRO checksum is valid */ 2125 u8 csum_valid:1; 2126 2127 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2128 u8 csum_cnt:3; 2129 2130 /* Free the skb? */ 2131 u8 free:2; 2132 #define NAPI_GRO_FREE 1 2133 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2134 2135 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2136 u8 is_ipv6:1; 2137 2138 /* Used in GRE, set in fou/gue_gro_receive */ 2139 u8 is_fou:1; 2140 2141 /* Used to determine if flush_id can be ignored */ 2142 u8 is_atomic:1; 2143 2144 /* Number of gro_receive callbacks this packet already went through */ 2145 u8 recursion_counter:4; 2146 2147 /* 1 bit hole */ 2148 2149 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2150 __wsum csum; 2151 2152 /* used in skb_gro_receive() slow path */ 2153 struct sk_buff *last; 2154 }; 2155 2156 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2157 2158 #define GRO_RECURSION_LIMIT 15 2159 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2160 { 2161 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2162 } 2163 2164 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *); 2165 static inline struct sk_buff **call_gro_receive(gro_receive_t cb, 2166 struct sk_buff **head, 2167 struct sk_buff *skb) 2168 { 2169 if (unlikely(gro_recursion_inc_test(skb))) { 2170 NAPI_GRO_CB(skb)->flush |= 1; 2171 return NULL; 2172 } 2173 2174 return cb(head, skb); 2175 } 2176 2177 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **, 2178 struct sk_buff *); 2179 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb, 2180 struct sock *sk, 2181 struct sk_buff **head, 2182 struct sk_buff *skb) 2183 { 2184 if (unlikely(gro_recursion_inc_test(skb))) { 2185 NAPI_GRO_CB(skb)->flush |= 1; 2186 return NULL; 2187 } 2188 2189 return cb(sk, head, skb); 2190 } 2191 2192 struct packet_type { 2193 __be16 type; /* This is really htons(ether_type). */ 2194 struct net_device *dev; /* NULL is wildcarded here */ 2195 int (*func) (struct sk_buff *, 2196 struct net_device *, 2197 struct packet_type *, 2198 struct net_device *); 2199 bool (*id_match)(struct packet_type *ptype, 2200 struct sock *sk); 2201 void *af_packet_priv; 2202 struct list_head list; 2203 }; 2204 2205 struct offload_callbacks { 2206 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2207 netdev_features_t features); 2208 struct sk_buff **(*gro_receive)(struct sk_buff **head, 2209 struct sk_buff *skb); 2210 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2211 }; 2212 2213 struct packet_offload { 2214 __be16 type; /* This is really htons(ether_type). */ 2215 u16 priority; 2216 struct offload_callbacks callbacks; 2217 struct list_head list; 2218 }; 2219 2220 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2221 struct pcpu_sw_netstats { 2222 u64 rx_packets; 2223 u64 rx_bytes; 2224 u64 tx_packets; 2225 u64 tx_bytes; 2226 struct u64_stats_sync syncp; 2227 }; 2228 2229 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2230 ({ \ 2231 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2232 if (pcpu_stats) { \ 2233 int __cpu; \ 2234 for_each_possible_cpu(__cpu) { \ 2235 typeof(type) *stat; \ 2236 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2237 u64_stats_init(&stat->syncp); \ 2238 } \ 2239 } \ 2240 pcpu_stats; \ 2241 }) 2242 2243 #define netdev_alloc_pcpu_stats(type) \ 2244 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2245 2246 enum netdev_lag_tx_type { 2247 NETDEV_LAG_TX_TYPE_UNKNOWN, 2248 NETDEV_LAG_TX_TYPE_RANDOM, 2249 NETDEV_LAG_TX_TYPE_BROADCAST, 2250 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2251 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2252 NETDEV_LAG_TX_TYPE_HASH, 2253 }; 2254 2255 struct netdev_lag_upper_info { 2256 enum netdev_lag_tx_type tx_type; 2257 }; 2258 2259 struct netdev_lag_lower_state_info { 2260 u8 link_up : 1, 2261 tx_enabled : 1; 2262 }; 2263 2264 #include <linux/notifier.h> 2265 2266 /* netdevice notifier chain. Please remember to update the rtnetlink 2267 * notification exclusion list in rtnetlink_event() when adding new 2268 * types. 2269 */ 2270 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */ 2271 #define NETDEV_DOWN 0x0002 2272 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface 2273 detected a hardware crash and restarted 2274 - we can use this eg to kick tcp sessions 2275 once done */ 2276 #define NETDEV_CHANGE 0x0004 /* Notify device state change */ 2277 #define NETDEV_REGISTER 0x0005 2278 #define NETDEV_UNREGISTER 0x0006 2279 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */ 2280 #define NETDEV_CHANGEADDR 0x0008 2281 #define NETDEV_GOING_DOWN 0x0009 2282 #define NETDEV_CHANGENAME 0x000A 2283 #define NETDEV_FEAT_CHANGE 0x000B 2284 #define NETDEV_BONDING_FAILOVER 0x000C 2285 #define NETDEV_PRE_UP 0x000D 2286 #define NETDEV_PRE_TYPE_CHANGE 0x000E 2287 #define NETDEV_POST_TYPE_CHANGE 0x000F 2288 #define NETDEV_POST_INIT 0x0010 2289 #define NETDEV_UNREGISTER_FINAL 0x0011 2290 #define NETDEV_RELEASE 0x0012 2291 #define NETDEV_NOTIFY_PEERS 0x0013 2292 #define NETDEV_JOIN 0x0014 2293 #define NETDEV_CHANGEUPPER 0x0015 2294 #define NETDEV_RESEND_IGMP 0x0016 2295 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */ 2296 #define NETDEV_CHANGEINFODATA 0x0018 2297 #define NETDEV_BONDING_INFO 0x0019 2298 #define NETDEV_PRECHANGEUPPER 0x001A 2299 #define NETDEV_CHANGELOWERSTATE 0x001B 2300 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C 2301 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E 2302 2303 int register_netdevice_notifier(struct notifier_block *nb); 2304 int unregister_netdevice_notifier(struct notifier_block *nb); 2305 2306 struct netdev_notifier_info { 2307 struct net_device *dev; 2308 }; 2309 2310 struct netdev_notifier_change_info { 2311 struct netdev_notifier_info info; /* must be first */ 2312 unsigned int flags_changed; 2313 }; 2314 2315 struct netdev_notifier_changeupper_info { 2316 struct netdev_notifier_info info; /* must be first */ 2317 struct net_device *upper_dev; /* new upper dev */ 2318 bool master; /* is upper dev master */ 2319 bool linking; /* is the notification for link or unlink */ 2320 void *upper_info; /* upper dev info */ 2321 }; 2322 2323 struct netdev_notifier_changelowerstate_info { 2324 struct netdev_notifier_info info; /* must be first */ 2325 void *lower_state_info; /* is lower dev state */ 2326 }; 2327 2328 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2329 struct net_device *dev) 2330 { 2331 info->dev = dev; 2332 } 2333 2334 static inline struct net_device * 2335 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2336 { 2337 return info->dev; 2338 } 2339 2340 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2341 2342 2343 extern rwlock_t dev_base_lock; /* Device list lock */ 2344 2345 #define for_each_netdev(net, d) \ 2346 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2347 #define for_each_netdev_reverse(net, d) \ 2348 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2349 #define for_each_netdev_rcu(net, d) \ 2350 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2351 #define for_each_netdev_safe(net, d, n) \ 2352 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2353 #define for_each_netdev_continue(net, d) \ 2354 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2355 #define for_each_netdev_continue_rcu(net, d) \ 2356 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2357 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2358 for_each_netdev_rcu(&init_net, slave) \ 2359 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2360 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2361 2362 static inline struct net_device *next_net_device(struct net_device *dev) 2363 { 2364 struct list_head *lh; 2365 struct net *net; 2366 2367 net = dev_net(dev); 2368 lh = dev->dev_list.next; 2369 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2370 } 2371 2372 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2373 { 2374 struct list_head *lh; 2375 struct net *net; 2376 2377 net = dev_net(dev); 2378 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2379 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2380 } 2381 2382 static inline struct net_device *first_net_device(struct net *net) 2383 { 2384 return list_empty(&net->dev_base_head) ? NULL : 2385 net_device_entry(net->dev_base_head.next); 2386 } 2387 2388 static inline struct net_device *first_net_device_rcu(struct net *net) 2389 { 2390 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2391 2392 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2393 } 2394 2395 int netdev_boot_setup_check(struct net_device *dev); 2396 unsigned long netdev_boot_base(const char *prefix, int unit); 2397 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2398 const char *hwaddr); 2399 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2400 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2401 void dev_add_pack(struct packet_type *pt); 2402 void dev_remove_pack(struct packet_type *pt); 2403 void __dev_remove_pack(struct packet_type *pt); 2404 void dev_add_offload(struct packet_offload *po); 2405 void dev_remove_offload(struct packet_offload *po); 2406 2407 int dev_get_iflink(const struct net_device *dev); 2408 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2409 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2410 unsigned short mask); 2411 struct net_device *dev_get_by_name(struct net *net, const char *name); 2412 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2413 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2414 int dev_alloc_name(struct net_device *dev, const char *name); 2415 int dev_open(struct net_device *dev); 2416 int dev_close(struct net_device *dev); 2417 int dev_close_many(struct list_head *head, bool unlink); 2418 void dev_disable_lro(struct net_device *dev); 2419 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2420 int dev_queue_xmit(struct sk_buff *skb); 2421 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv); 2422 int register_netdevice(struct net_device *dev); 2423 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2424 void unregister_netdevice_many(struct list_head *head); 2425 static inline void unregister_netdevice(struct net_device *dev) 2426 { 2427 unregister_netdevice_queue(dev, NULL); 2428 } 2429 2430 int netdev_refcnt_read(const struct net_device *dev); 2431 void free_netdev(struct net_device *dev); 2432 void netdev_freemem(struct net_device *dev); 2433 void synchronize_net(void); 2434 int init_dummy_netdev(struct net_device *dev); 2435 2436 DECLARE_PER_CPU(int, xmit_recursion); 2437 #define XMIT_RECURSION_LIMIT 10 2438 2439 static inline int dev_recursion_level(void) 2440 { 2441 return this_cpu_read(xmit_recursion); 2442 } 2443 2444 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2445 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2446 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2447 int netdev_get_name(struct net *net, char *name, int ifindex); 2448 int dev_restart(struct net_device *dev); 2449 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb); 2450 2451 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2452 { 2453 return NAPI_GRO_CB(skb)->data_offset; 2454 } 2455 2456 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2457 { 2458 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2459 } 2460 2461 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2462 { 2463 NAPI_GRO_CB(skb)->data_offset += len; 2464 } 2465 2466 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2467 unsigned int offset) 2468 { 2469 return NAPI_GRO_CB(skb)->frag0 + offset; 2470 } 2471 2472 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2473 { 2474 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2475 } 2476 2477 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2478 { 2479 NAPI_GRO_CB(skb)->frag0 = NULL; 2480 NAPI_GRO_CB(skb)->frag0_len = 0; 2481 } 2482 2483 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2484 unsigned int offset) 2485 { 2486 if (!pskb_may_pull(skb, hlen)) 2487 return NULL; 2488 2489 skb_gro_frag0_invalidate(skb); 2490 return skb->data + offset; 2491 } 2492 2493 static inline void *skb_gro_network_header(struct sk_buff *skb) 2494 { 2495 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2496 skb_network_offset(skb); 2497 } 2498 2499 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2500 const void *start, unsigned int len) 2501 { 2502 if (NAPI_GRO_CB(skb)->csum_valid) 2503 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2504 csum_partial(start, len, 0)); 2505 } 2506 2507 /* GRO checksum functions. These are logical equivalents of the normal 2508 * checksum functions (in skbuff.h) except that they operate on the GRO 2509 * offsets and fields in sk_buff. 2510 */ 2511 2512 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2513 2514 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2515 { 2516 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2517 } 2518 2519 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2520 bool zero_okay, 2521 __sum16 check) 2522 { 2523 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2524 skb_checksum_start_offset(skb) < 2525 skb_gro_offset(skb)) && 2526 !skb_at_gro_remcsum_start(skb) && 2527 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2528 (!zero_okay || check)); 2529 } 2530 2531 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2532 __wsum psum) 2533 { 2534 if (NAPI_GRO_CB(skb)->csum_valid && 2535 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2536 return 0; 2537 2538 NAPI_GRO_CB(skb)->csum = psum; 2539 2540 return __skb_gro_checksum_complete(skb); 2541 } 2542 2543 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2544 { 2545 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2546 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2547 NAPI_GRO_CB(skb)->csum_cnt--; 2548 } else { 2549 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2550 * verified a new top level checksum or an encapsulated one 2551 * during GRO. This saves work if we fallback to normal path. 2552 */ 2553 __skb_incr_checksum_unnecessary(skb); 2554 } 2555 } 2556 2557 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2558 compute_pseudo) \ 2559 ({ \ 2560 __sum16 __ret = 0; \ 2561 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2562 __ret = __skb_gro_checksum_validate_complete(skb, \ 2563 compute_pseudo(skb, proto)); \ 2564 if (__ret) \ 2565 __skb_mark_checksum_bad(skb); \ 2566 else \ 2567 skb_gro_incr_csum_unnecessary(skb); \ 2568 __ret; \ 2569 }) 2570 2571 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2572 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2573 2574 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2575 compute_pseudo) \ 2576 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2577 2578 #define skb_gro_checksum_simple_validate(skb) \ 2579 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2580 2581 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2582 { 2583 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2584 !NAPI_GRO_CB(skb)->csum_valid); 2585 } 2586 2587 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2588 __sum16 check, __wsum pseudo) 2589 { 2590 NAPI_GRO_CB(skb)->csum = ~pseudo; 2591 NAPI_GRO_CB(skb)->csum_valid = 1; 2592 } 2593 2594 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \ 2595 do { \ 2596 if (__skb_gro_checksum_convert_check(skb)) \ 2597 __skb_gro_checksum_convert(skb, check, \ 2598 compute_pseudo(skb, proto)); \ 2599 } while (0) 2600 2601 struct gro_remcsum { 2602 int offset; 2603 __wsum delta; 2604 }; 2605 2606 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2607 { 2608 grc->offset = 0; 2609 grc->delta = 0; 2610 } 2611 2612 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2613 unsigned int off, size_t hdrlen, 2614 int start, int offset, 2615 struct gro_remcsum *grc, 2616 bool nopartial) 2617 { 2618 __wsum delta; 2619 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2620 2621 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2622 2623 if (!nopartial) { 2624 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2625 return ptr; 2626 } 2627 2628 ptr = skb_gro_header_fast(skb, off); 2629 if (skb_gro_header_hard(skb, off + plen)) { 2630 ptr = skb_gro_header_slow(skb, off + plen, off); 2631 if (!ptr) 2632 return NULL; 2633 } 2634 2635 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2636 start, offset); 2637 2638 /* Adjust skb->csum since we changed the packet */ 2639 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2640 2641 grc->offset = off + hdrlen + offset; 2642 grc->delta = delta; 2643 2644 return ptr; 2645 } 2646 2647 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2648 struct gro_remcsum *grc) 2649 { 2650 void *ptr; 2651 size_t plen = grc->offset + sizeof(u16); 2652 2653 if (!grc->delta) 2654 return; 2655 2656 ptr = skb_gro_header_fast(skb, grc->offset); 2657 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 2658 ptr = skb_gro_header_slow(skb, plen, grc->offset); 2659 if (!ptr) 2660 return; 2661 } 2662 2663 remcsum_unadjust((__sum16 *)ptr, grc->delta); 2664 } 2665 2666 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2667 unsigned short type, 2668 const void *daddr, const void *saddr, 2669 unsigned int len) 2670 { 2671 if (!dev->header_ops || !dev->header_ops->create) 2672 return 0; 2673 2674 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2675 } 2676 2677 static inline int dev_parse_header(const struct sk_buff *skb, 2678 unsigned char *haddr) 2679 { 2680 const struct net_device *dev = skb->dev; 2681 2682 if (!dev->header_ops || !dev->header_ops->parse) 2683 return 0; 2684 return dev->header_ops->parse(skb, haddr); 2685 } 2686 2687 /* ll_header must have at least hard_header_len allocated */ 2688 static inline bool dev_validate_header(const struct net_device *dev, 2689 char *ll_header, int len) 2690 { 2691 if (likely(len >= dev->hard_header_len)) 2692 return true; 2693 2694 if (capable(CAP_SYS_RAWIO)) { 2695 memset(ll_header + len, 0, dev->hard_header_len - len); 2696 return true; 2697 } 2698 2699 if (dev->header_ops && dev->header_ops->validate) 2700 return dev->header_ops->validate(ll_header, len); 2701 2702 return false; 2703 } 2704 2705 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len); 2706 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2707 static inline int unregister_gifconf(unsigned int family) 2708 { 2709 return register_gifconf(family, NULL); 2710 } 2711 2712 #ifdef CONFIG_NET_FLOW_LIMIT 2713 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2714 struct sd_flow_limit { 2715 u64 count; 2716 unsigned int num_buckets; 2717 unsigned int history_head; 2718 u16 history[FLOW_LIMIT_HISTORY]; 2719 u8 buckets[]; 2720 }; 2721 2722 extern int netdev_flow_limit_table_len; 2723 #endif /* CONFIG_NET_FLOW_LIMIT */ 2724 2725 /* 2726 * Incoming packets are placed on per-CPU queues 2727 */ 2728 struct softnet_data { 2729 struct list_head poll_list; 2730 struct sk_buff_head process_queue; 2731 2732 /* stats */ 2733 unsigned int processed; 2734 unsigned int time_squeeze; 2735 unsigned int received_rps; 2736 #ifdef CONFIG_RPS 2737 struct softnet_data *rps_ipi_list; 2738 #endif 2739 #ifdef CONFIG_NET_FLOW_LIMIT 2740 struct sd_flow_limit __rcu *flow_limit; 2741 #endif 2742 struct Qdisc *output_queue; 2743 struct Qdisc **output_queue_tailp; 2744 struct sk_buff *completion_queue; 2745 2746 #ifdef CONFIG_RPS 2747 /* input_queue_head should be written by cpu owning this struct, 2748 * and only read by other cpus. Worth using a cache line. 2749 */ 2750 unsigned int input_queue_head ____cacheline_aligned_in_smp; 2751 2752 /* Elements below can be accessed between CPUs for RPS/RFS */ 2753 struct call_single_data csd ____cacheline_aligned_in_smp; 2754 struct softnet_data *rps_ipi_next; 2755 unsigned int cpu; 2756 unsigned int input_queue_tail; 2757 #endif 2758 unsigned int dropped; 2759 struct sk_buff_head input_pkt_queue; 2760 struct napi_struct backlog; 2761 2762 }; 2763 2764 static inline void input_queue_head_incr(struct softnet_data *sd) 2765 { 2766 #ifdef CONFIG_RPS 2767 sd->input_queue_head++; 2768 #endif 2769 } 2770 2771 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 2772 unsigned int *qtail) 2773 { 2774 #ifdef CONFIG_RPS 2775 *qtail = ++sd->input_queue_tail; 2776 #endif 2777 } 2778 2779 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 2780 2781 void __netif_schedule(struct Qdisc *q); 2782 void netif_schedule_queue(struct netdev_queue *txq); 2783 2784 static inline void netif_tx_schedule_all(struct net_device *dev) 2785 { 2786 unsigned int i; 2787 2788 for (i = 0; i < dev->num_tx_queues; i++) 2789 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 2790 } 2791 2792 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 2793 { 2794 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2795 } 2796 2797 /** 2798 * netif_start_queue - allow transmit 2799 * @dev: network device 2800 * 2801 * Allow upper layers to call the device hard_start_xmit routine. 2802 */ 2803 static inline void netif_start_queue(struct net_device *dev) 2804 { 2805 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 2806 } 2807 2808 static inline void netif_tx_start_all_queues(struct net_device *dev) 2809 { 2810 unsigned int i; 2811 2812 for (i = 0; i < dev->num_tx_queues; i++) { 2813 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2814 netif_tx_start_queue(txq); 2815 } 2816 } 2817 2818 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 2819 2820 /** 2821 * netif_wake_queue - restart transmit 2822 * @dev: network device 2823 * 2824 * Allow upper layers to call the device hard_start_xmit routine. 2825 * Used for flow control when transmit resources are available. 2826 */ 2827 static inline void netif_wake_queue(struct net_device *dev) 2828 { 2829 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 2830 } 2831 2832 static inline void netif_tx_wake_all_queues(struct net_device *dev) 2833 { 2834 unsigned int i; 2835 2836 for (i = 0; i < dev->num_tx_queues; i++) { 2837 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2838 netif_tx_wake_queue(txq); 2839 } 2840 } 2841 2842 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 2843 { 2844 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2845 } 2846 2847 /** 2848 * netif_stop_queue - stop transmitted packets 2849 * @dev: network device 2850 * 2851 * Stop upper layers calling the device hard_start_xmit routine. 2852 * Used for flow control when transmit resources are unavailable. 2853 */ 2854 static inline void netif_stop_queue(struct net_device *dev) 2855 { 2856 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 2857 } 2858 2859 void netif_tx_stop_all_queues(struct net_device *dev); 2860 2861 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 2862 { 2863 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2864 } 2865 2866 /** 2867 * netif_queue_stopped - test if transmit queue is flowblocked 2868 * @dev: network device 2869 * 2870 * Test if transmit queue on device is currently unable to send. 2871 */ 2872 static inline bool netif_queue_stopped(const struct net_device *dev) 2873 { 2874 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 2875 } 2876 2877 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 2878 { 2879 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 2880 } 2881 2882 static inline bool 2883 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 2884 { 2885 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 2886 } 2887 2888 static inline bool 2889 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 2890 { 2891 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 2892 } 2893 2894 /** 2895 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 2896 * @dev_queue: pointer to transmit queue 2897 * 2898 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 2899 * to give appropriate hint to the CPU. 2900 */ 2901 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 2902 { 2903 #ifdef CONFIG_BQL 2904 prefetchw(&dev_queue->dql.num_queued); 2905 #endif 2906 } 2907 2908 /** 2909 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 2910 * @dev_queue: pointer to transmit queue 2911 * 2912 * BQL enabled drivers might use this helper in their TX completion path, 2913 * to give appropriate hint to the CPU. 2914 */ 2915 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 2916 { 2917 #ifdef CONFIG_BQL 2918 prefetchw(&dev_queue->dql.limit); 2919 #endif 2920 } 2921 2922 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 2923 unsigned int bytes) 2924 { 2925 #ifdef CONFIG_BQL 2926 dql_queued(&dev_queue->dql, bytes); 2927 2928 if (likely(dql_avail(&dev_queue->dql) >= 0)) 2929 return; 2930 2931 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2932 2933 /* 2934 * The XOFF flag must be set before checking the dql_avail below, 2935 * because in netdev_tx_completed_queue we update the dql_completed 2936 * before checking the XOFF flag. 2937 */ 2938 smp_mb(); 2939 2940 /* check again in case another CPU has just made room avail */ 2941 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 2942 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2943 #endif 2944 } 2945 2946 /** 2947 * netdev_sent_queue - report the number of bytes queued to hardware 2948 * @dev: network device 2949 * @bytes: number of bytes queued to the hardware device queue 2950 * 2951 * Report the number of bytes queued for sending/completion to the network 2952 * device hardware queue. @bytes should be a good approximation and should 2953 * exactly match netdev_completed_queue() @bytes 2954 */ 2955 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 2956 { 2957 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 2958 } 2959 2960 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 2961 unsigned int pkts, unsigned int bytes) 2962 { 2963 #ifdef CONFIG_BQL 2964 if (unlikely(!bytes)) 2965 return; 2966 2967 dql_completed(&dev_queue->dql, bytes); 2968 2969 /* 2970 * Without the memory barrier there is a small possiblity that 2971 * netdev_tx_sent_queue will miss the update and cause the queue to 2972 * be stopped forever 2973 */ 2974 smp_mb(); 2975 2976 if (dql_avail(&dev_queue->dql) < 0) 2977 return; 2978 2979 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 2980 netif_schedule_queue(dev_queue); 2981 #endif 2982 } 2983 2984 /** 2985 * netdev_completed_queue - report bytes and packets completed by device 2986 * @dev: network device 2987 * @pkts: actual number of packets sent over the medium 2988 * @bytes: actual number of bytes sent over the medium 2989 * 2990 * Report the number of bytes and packets transmitted by the network device 2991 * hardware queue over the physical medium, @bytes must exactly match the 2992 * @bytes amount passed to netdev_sent_queue() 2993 */ 2994 static inline void netdev_completed_queue(struct net_device *dev, 2995 unsigned int pkts, unsigned int bytes) 2996 { 2997 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 2998 } 2999 3000 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3001 { 3002 #ifdef CONFIG_BQL 3003 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3004 dql_reset(&q->dql); 3005 #endif 3006 } 3007 3008 /** 3009 * netdev_reset_queue - reset the packets and bytes count of a network device 3010 * @dev_queue: network device 3011 * 3012 * Reset the bytes and packet count of a network device and clear the 3013 * software flow control OFF bit for this network device 3014 */ 3015 static inline void netdev_reset_queue(struct net_device *dev_queue) 3016 { 3017 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3018 } 3019 3020 /** 3021 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3022 * @dev: network device 3023 * @queue_index: given tx queue index 3024 * 3025 * Returns 0 if given tx queue index >= number of device tx queues, 3026 * otherwise returns the originally passed tx queue index. 3027 */ 3028 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3029 { 3030 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3031 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3032 dev->name, queue_index, 3033 dev->real_num_tx_queues); 3034 return 0; 3035 } 3036 3037 return queue_index; 3038 } 3039 3040 /** 3041 * netif_running - test if up 3042 * @dev: network device 3043 * 3044 * Test if the device has been brought up. 3045 */ 3046 static inline bool netif_running(const struct net_device *dev) 3047 { 3048 return test_bit(__LINK_STATE_START, &dev->state); 3049 } 3050 3051 /* 3052 * Routines to manage the subqueues on a device. We only need start, 3053 * stop, and a check if it's stopped. All other device management is 3054 * done at the overall netdevice level. 3055 * Also test the device if we're multiqueue. 3056 */ 3057 3058 /** 3059 * netif_start_subqueue - allow sending packets on subqueue 3060 * @dev: network device 3061 * @queue_index: sub queue index 3062 * 3063 * Start individual transmit queue of a device with multiple transmit queues. 3064 */ 3065 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3066 { 3067 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3068 3069 netif_tx_start_queue(txq); 3070 } 3071 3072 /** 3073 * netif_stop_subqueue - stop sending packets on subqueue 3074 * @dev: network device 3075 * @queue_index: sub queue index 3076 * 3077 * Stop individual transmit queue of a device with multiple transmit queues. 3078 */ 3079 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3080 { 3081 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3082 netif_tx_stop_queue(txq); 3083 } 3084 3085 /** 3086 * netif_subqueue_stopped - test status of subqueue 3087 * @dev: network device 3088 * @queue_index: sub queue index 3089 * 3090 * Check individual transmit queue of a device with multiple transmit queues. 3091 */ 3092 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3093 u16 queue_index) 3094 { 3095 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3096 3097 return netif_tx_queue_stopped(txq); 3098 } 3099 3100 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3101 struct sk_buff *skb) 3102 { 3103 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3104 } 3105 3106 /** 3107 * netif_wake_subqueue - allow sending packets on subqueue 3108 * @dev: network device 3109 * @queue_index: sub queue index 3110 * 3111 * Resume individual transmit queue of a device with multiple transmit queues. 3112 */ 3113 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3114 { 3115 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3116 3117 netif_tx_wake_queue(txq); 3118 } 3119 3120 #ifdef CONFIG_XPS 3121 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3122 u16 index); 3123 #else 3124 static inline int netif_set_xps_queue(struct net_device *dev, 3125 const struct cpumask *mask, 3126 u16 index) 3127 { 3128 return 0; 3129 } 3130 #endif 3131 3132 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb, 3133 unsigned int num_tx_queues); 3134 3135 /* 3136 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used 3137 * as a distribution range limit for the returned value. 3138 */ 3139 static inline u16 skb_tx_hash(const struct net_device *dev, 3140 struct sk_buff *skb) 3141 { 3142 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues); 3143 } 3144 3145 /** 3146 * netif_is_multiqueue - test if device has multiple transmit queues 3147 * @dev: network device 3148 * 3149 * Check if device has multiple transmit queues 3150 */ 3151 static inline bool netif_is_multiqueue(const struct net_device *dev) 3152 { 3153 return dev->num_tx_queues > 1; 3154 } 3155 3156 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3157 3158 #ifdef CONFIG_SYSFS 3159 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3160 #else 3161 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3162 unsigned int rxq) 3163 { 3164 return 0; 3165 } 3166 #endif 3167 3168 #ifdef CONFIG_SYSFS 3169 static inline unsigned int get_netdev_rx_queue_index( 3170 struct netdev_rx_queue *queue) 3171 { 3172 struct net_device *dev = queue->dev; 3173 int index = queue - dev->_rx; 3174 3175 BUG_ON(index >= dev->num_rx_queues); 3176 return index; 3177 } 3178 #endif 3179 3180 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3181 int netif_get_num_default_rss_queues(void); 3182 3183 enum skb_free_reason { 3184 SKB_REASON_CONSUMED, 3185 SKB_REASON_DROPPED, 3186 }; 3187 3188 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3189 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3190 3191 /* 3192 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3193 * interrupt context or with hardware interrupts being disabled. 3194 * (in_irq() || irqs_disabled()) 3195 * 3196 * We provide four helpers that can be used in following contexts : 3197 * 3198 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3199 * replacing kfree_skb(skb) 3200 * 3201 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3202 * Typically used in place of consume_skb(skb) in TX completion path 3203 * 3204 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3205 * replacing kfree_skb(skb) 3206 * 3207 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3208 * and consumed a packet. Used in place of consume_skb(skb) 3209 */ 3210 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3211 { 3212 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3213 } 3214 3215 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3216 { 3217 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3218 } 3219 3220 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3221 { 3222 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3223 } 3224 3225 static inline void dev_consume_skb_any(struct sk_buff *skb) 3226 { 3227 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3228 } 3229 3230 int netif_rx(struct sk_buff *skb); 3231 int netif_rx_ni(struct sk_buff *skb); 3232 int netif_receive_skb(struct sk_buff *skb); 3233 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3234 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3235 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3236 gro_result_t napi_gro_frags(struct napi_struct *napi); 3237 struct packet_offload *gro_find_receive_by_type(__be16 type); 3238 struct packet_offload *gro_find_complete_by_type(__be16 type); 3239 3240 static inline void napi_free_frags(struct napi_struct *napi) 3241 { 3242 kfree_skb(napi->skb); 3243 napi->skb = NULL; 3244 } 3245 3246 bool netdev_is_rx_handler_busy(struct net_device *dev); 3247 int netdev_rx_handler_register(struct net_device *dev, 3248 rx_handler_func_t *rx_handler, 3249 void *rx_handler_data); 3250 void netdev_rx_handler_unregister(struct net_device *dev); 3251 3252 bool dev_valid_name(const char *name); 3253 int dev_ioctl(struct net *net, unsigned int cmd, void __user *); 3254 int dev_ethtool(struct net *net, struct ifreq *); 3255 unsigned int dev_get_flags(const struct net_device *); 3256 int __dev_change_flags(struct net_device *, unsigned int flags); 3257 int dev_change_flags(struct net_device *, unsigned int); 3258 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3259 unsigned int gchanges); 3260 int dev_change_name(struct net_device *, const char *); 3261 int dev_set_alias(struct net_device *, const char *, size_t); 3262 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3263 int dev_set_mtu(struct net_device *, int); 3264 void dev_set_group(struct net_device *, int); 3265 int dev_set_mac_address(struct net_device *, struct sockaddr *); 3266 int dev_change_carrier(struct net_device *, bool new_carrier); 3267 int dev_get_phys_port_id(struct net_device *dev, 3268 struct netdev_phys_item_id *ppid); 3269 int dev_get_phys_port_name(struct net_device *dev, 3270 char *name, size_t len); 3271 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3272 int dev_change_xdp_fd(struct net_device *dev, int fd, u32 flags); 3273 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev); 3274 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3275 struct netdev_queue *txq, int *ret); 3276 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3277 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3278 bool is_skb_forwardable(const struct net_device *dev, 3279 const struct sk_buff *skb); 3280 3281 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3282 struct sk_buff *skb) 3283 { 3284 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3285 unlikely(!is_skb_forwardable(dev, skb))) { 3286 atomic_long_inc(&dev->rx_dropped); 3287 kfree_skb(skb); 3288 return NET_RX_DROP; 3289 } 3290 3291 skb_scrub_packet(skb, true); 3292 skb->priority = 0; 3293 return 0; 3294 } 3295 3296 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3297 3298 extern int netdev_budget; 3299 3300 /* Called by rtnetlink.c:rtnl_unlock() */ 3301 void netdev_run_todo(void); 3302 3303 /** 3304 * dev_put - release reference to device 3305 * @dev: network device 3306 * 3307 * Release reference to device to allow it to be freed. 3308 */ 3309 static inline void dev_put(struct net_device *dev) 3310 { 3311 this_cpu_dec(*dev->pcpu_refcnt); 3312 } 3313 3314 /** 3315 * dev_hold - get reference to device 3316 * @dev: network device 3317 * 3318 * Hold reference to device to keep it from being freed. 3319 */ 3320 static inline void dev_hold(struct net_device *dev) 3321 { 3322 this_cpu_inc(*dev->pcpu_refcnt); 3323 } 3324 3325 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3326 * and _off may be called from IRQ context, but it is caller 3327 * who is responsible for serialization of these calls. 3328 * 3329 * The name carrier is inappropriate, these functions should really be 3330 * called netif_lowerlayer_*() because they represent the state of any 3331 * kind of lower layer not just hardware media. 3332 */ 3333 3334 void linkwatch_init_dev(struct net_device *dev); 3335 void linkwatch_fire_event(struct net_device *dev); 3336 void linkwatch_forget_dev(struct net_device *dev); 3337 3338 /** 3339 * netif_carrier_ok - test if carrier present 3340 * @dev: network device 3341 * 3342 * Check if carrier is present on device 3343 */ 3344 static inline bool netif_carrier_ok(const struct net_device *dev) 3345 { 3346 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3347 } 3348 3349 unsigned long dev_trans_start(struct net_device *dev); 3350 3351 void __netdev_watchdog_up(struct net_device *dev); 3352 3353 void netif_carrier_on(struct net_device *dev); 3354 3355 void netif_carrier_off(struct net_device *dev); 3356 3357 /** 3358 * netif_dormant_on - mark device as dormant. 3359 * @dev: network device 3360 * 3361 * Mark device as dormant (as per RFC2863). 3362 * 3363 * The dormant state indicates that the relevant interface is not 3364 * actually in a condition to pass packets (i.e., it is not 'up') but is 3365 * in a "pending" state, waiting for some external event. For "on- 3366 * demand" interfaces, this new state identifies the situation where the 3367 * interface is waiting for events to place it in the up state. 3368 */ 3369 static inline void netif_dormant_on(struct net_device *dev) 3370 { 3371 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3372 linkwatch_fire_event(dev); 3373 } 3374 3375 /** 3376 * netif_dormant_off - set device as not dormant. 3377 * @dev: network device 3378 * 3379 * Device is not in dormant state. 3380 */ 3381 static inline void netif_dormant_off(struct net_device *dev) 3382 { 3383 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3384 linkwatch_fire_event(dev); 3385 } 3386 3387 /** 3388 * netif_dormant - test if carrier present 3389 * @dev: network device 3390 * 3391 * Check if carrier is present on device 3392 */ 3393 static inline bool netif_dormant(const struct net_device *dev) 3394 { 3395 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3396 } 3397 3398 3399 /** 3400 * netif_oper_up - test if device is operational 3401 * @dev: network device 3402 * 3403 * Check if carrier is operational 3404 */ 3405 static inline bool netif_oper_up(const struct net_device *dev) 3406 { 3407 return (dev->operstate == IF_OPER_UP || 3408 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3409 } 3410 3411 /** 3412 * netif_device_present - is device available or removed 3413 * @dev: network device 3414 * 3415 * Check if device has not been removed from system. 3416 */ 3417 static inline bool netif_device_present(struct net_device *dev) 3418 { 3419 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3420 } 3421 3422 void netif_device_detach(struct net_device *dev); 3423 3424 void netif_device_attach(struct net_device *dev); 3425 3426 /* 3427 * Network interface message level settings 3428 */ 3429 3430 enum { 3431 NETIF_MSG_DRV = 0x0001, 3432 NETIF_MSG_PROBE = 0x0002, 3433 NETIF_MSG_LINK = 0x0004, 3434 NETIF_MSG_TIMER = 0x0008, 3435 NETIF_MSG_IFDOWN = 0x0010, 3436 NETIF_MSG_IFUP = 0x0020, 3437 NETIF_MSG_RX_ERR = 0x0040, 3438 NETIF_MSG_TX_ERR = 0x0080, 3439 NETIF_MSG_TX_QUEUED = 0x0100, 3440 NETIF_MSG_INTR = 0x0200, 3441 NETIF_MSG_TX_DONE = 0x0400, 3442 NETIF_MSG_RX_STATUS = 0x0800, 3443 NETIF_MSG_PKTDATA = 0x1000, 3444 NETIF_MSG_HW = 0x2000, 3445 NETIF_MSG_WOL = 0x4000, 3446 }; 3447 3448 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3449 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3450 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3451 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3452 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3453 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3454 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3455 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3456 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3457 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3458 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3459 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3460 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3461 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3462 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3463 3464 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3465 { 3466 /* use default */ 3467 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3468 return default_msg_enable_bits; 3469 if (debug_value == 0) /* no output */ 3470 return 0; 3471 /* set low N bits */ 3472 return (1 << debug_value) - 1; 3473 } 3474 3475 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 3476 { 3477 spin_lock(&txq->_xmit_lock); 3478 txq->xmit_lock_owner = cpu; 3479 } 3480 3481 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 3482 { 3483 __acquire(&txq->_xmit_lock); 3484 return true; 3485 } 3486 3487 static inline void __netif_tx_release(struct netdev_queue *txq) 3488 { 3489 __release(&txq->_xmit_lock); 3490 } 3491 3492 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 3493 { 3494 spin_lock_bh(&txq->_xmit_lock); 3495 txq->xmit_lock_owner = smp_processor_id(); 3496 } 3497 3498 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 3499 { 3500 bool ok = spin_trylock(&txq->_xmit_lock); 3501 if (likely(ok)) 3502 txq->xmit_lock_owner = smp_processor_id(); 3503 return ok; 3504 } 3505 3506 static inline void __netif_tx_unlock(struct netdev_queue *txq) 3507 { 3508 txq->xmit_lock_owner = -1; 3509 spin_unlock(&txq->_xmit_lock); 3510 } 3511 3512 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 3513 { 3514 txq->xmit_lock_owner = -1; 3515 spin_unlock_bh(&txq->_xmit_lock); 3516 } 3517 3518 static inline void txq_trans_update(struct netdev_queue *txq) 3519 { 3520 if (txq->xmit_lock_owner != -1) 3521 txq->trans_start = jiffies; 3522 } 3523 3524 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 3525 static inline void netif_trans_update(struct net_device *dev) 3526 { 3527 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 3528 3529 if (txq->trans_start != jiffies) 3530 txq->trans_start = jiffies; 3531 } 3532 3533 /** 3534 * netif_tx_lock - grab network device transmit lock 3535 * @dev: network device 3536 * 3537 * Get network device transmit lock 3538 */ 3539 static inline void netif_tx_lock(struct net_device *dev) 3540 { 3541 unsigned int i; 3542 int cpu; 3543 3544 spin_lock(&dev->tx_global_lock); 3545 cpu = smp_processor_id(); 3546 for (i = 0; i < dev->num_tx_queues; i++) { 3547 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3548 3549 /* We are the only thread of execution doing a 3550 * freeze, but we have to grab the _xmit_lock in 3551 * order to synchronize with threads which are in 3552 * the ->hard_start_xmit() handler and already 3553 * checked the frozen bit. 3554 */ 3555 __netif_tx_lock(txq, cpu); 3556 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 3557 __netif_tx_unlock(txq); 3558 } 3559 } 3560 3561 static inline void netif_tx_lock_bh(struct net_device *dev) 3562 { 3563 local_bh_disable(); 3564 netif_tx_lock(dev); 3565 } 3566 3567 static inline void netif_tx_unlock(struct net_device *dev) 3568 { 3569 unsigned int i; 3570 3571 for (i = 0; i < dev->num_tx_queues; i++) { 3572 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3573 3574 /* No need to grab the _xmit_lock here. If the 3575 * queue is not stopped for another reason, we 3576 * force a schedule. 3577 */ 3578 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 3579 netif_schedule_queue(txq); 3580 } 3581 spin_unlock(&dev->tx_global_lock); 3582 } 3583 3584 static inline void netif_tx_unlock_bh(struct net_device *dev) 3585 { 3586 netif_tx_unlock(dev); 3587 local_bh_enable(); 3588 } 3589 3590 #define HARD_TX_LOCK(dev, txq, cpu) { \ 3591 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3592 __netif_tx_lock(txq, cpu); \ 3593 } else { \ 3594 __netif_tx_acquire(txq); \ 3595 } \ 3596 } 3597 3598 #define HARD_TX_TRYLOCK(dev, txq) \ 3599 (((dev->features & NETIF_F_LLTX) == 0) ? \ 3600 __netif_tx_trylock(txq) : \ 3601 __netif_tx_acquire(txq)) 3602 3603 #define HARD_TX_UNLOCK(dev, txq) { \ 3604 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3605 __netif_tx_unlock(txq); \ 3606 } else { \ 3607 __netif_tx_release(txq); \ 3608 } \ 3609 } 3610 3611 static inline void netif_tx_disable(struct net_device *dev) 3612 { 3613 unsigned int i; 3614 int cpu; 3615 3616 local_bh_disable(); 3617 cpu = smp_processor_id(); 3618 for (i = 0; i < dev->num_tx_queues; i++) { 3619 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3620 3621 __netif_tx_lock(txq, cpu); 3622 netif_tx_stop_queue(txq); 3623 __netif_tx_unlock(txq); 3624 } 3625 local_bh_enable(); 3626 } 3627 3628 static inline void netif_addr_lock(struct net_device *dev) 3629 { 3630 spin_lock(&dev->addr_list_lock); 3631 } 3632 3633 static inline void netif_addr_lock_nested(struct net_device *dev) 3634 { 3635 int subclass = SINGLE_DEPTH_NESTING; 3636 3637 if (dev->netdev_ops->ndo_get_lock_subclass) 3638 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev); 3639 3640 spin_lock_nested(&dev->addr_list_lock, subclass); 3641 } 3642 3643 static inline void netif_addr_lock_bh(struct net_device *dev) 3644 { 3645 spin_lock_bh(&dev->addr_list_lock); 3646 } 3647 3648 static inline void netif_addr_unlock(struct net_device *dev) 3649 { 3650 spin_unlock(&dev->addr_list_lock); 3651 } 3652 3653 static inline void netif_addr_unlock_bh(struct net_device *dev) 3654 { 3655 spin_unlock_bh(&dev->addr_list_lock); 3656 } 3657 3658 /* 3659 * dev_addrs walker. Should be used only for read access. Call with 3660 * rcu_read_lock held. 3661 */ 3662 #define for_each_dev_addr(dev, ha) \ 3663 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 3664 3665 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 3666 3667 void ether_setup(struct net_device *dev); 3668 3669 /* Support for loadable net-drivers */ 3670 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 3671 unsigned char name_assign_type, 3672 void (*setup)(struct net_device *), 3673 unsigned int txqs, unsigned int rxqs); 3674 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 3675 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 3676 3677 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 3678 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 3679 count) 3680 3681 int register_netdev(struct net_device *dev); 3682 void unregister_netdev(struct net_device *dev); 3683 3684 /* General hardware address lists handling functions */ 3685 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 3686 struct netdev_hw_addr_list *from_list, int addr_len); 3687 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 3688 struct netdev_hw_addr_list *from_list, int addr_len); 3689 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 3690 struct net_device *dev, 3691 int (*sync)(struct net_device *, const unsigned char *), 3692 int (*unsync)(struct net_device *, 3693 const unsigned char *)); 3694 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 3695 struct net_device *dev, 3696 int (*unsync)(struct net_device *, 3697 const unsigned char *)); 3698 void __hw_addr_init(struct netdev_hw_addr_list *list); 3699 3700 /* Functions used for device addresses handling */ 3701 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 3702 unsigned char addr_type); 3703 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 3704 unsigned char addr_type); 3705 void dev_addr_flush(struct net_device *dev); 3706 int dev_addr_init(struct net_device *dev); 3707 3708 /* Functions used for unicast addresses handling */ 3709 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 3710 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 3711 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 3712 int dev_uc_sync(struct net_device *to, struct net_device *from); 3713 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 3714 void dev_uc_unsync(struct net_device *to, struct net_device *from); 3715 void dev_uc_flush(struct net_device *dev); 3716 void dev_uc_init(struct net_device *dev); 3717 3718 /** 3719 * __dev_uc_sync - Synchonize device's unicast list 3720 * @dev: device to sync 3721 * @sync: function to call if address should be added 3722 * @unsync: function to call if address should be removed 3723 * 3724 * Add newly added addresses to the interface, and release 3725 * addresses that have been deleted. 3726 */ 3727 static inline int __dev_uc_sync(struct net_device *dev, 3728 int (*sync)(struct net_device *, 3729 const unsigned char *), 3730 int (*unsync)(struct net_device *, 3731 const unsigned char *)) 3732 { 3733 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 3734 } 3735 3736 /** 3737 * __dev_uc_unsync - Remove synchronized addresses from device 3738 * @dev: device to sync 3739 * @unsync: function to call if address should be removed 3740 * 3741 * Remove all addresses that were added to the device by dev_uc_sync(). 3742 */ 3743 static inline void __dev_uc_unsync(struct net_device *dev, 3744 int (*unsync)(struct net_device *, 3745 const unsigned char *)) 3746 { 3747 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 3748 } 3749 3750 /* Functions used for multicast addresses handling */ 3751 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 3752 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 3753 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 3754 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 3755 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 3756 int dev_mc_sync(struct net_device *to, struct net_device *from); 3757 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 3758 void dev_mc_unsync(struct net_device *to, struct net_device *from); 3759 void dev_mc_flush(struct net_device *dev); 3760 void dev_mc_init(struct net_device *dev); 3761 3762 /** 3763 * __dev_mc_sync - Synchonize device's multicast list 3764 * @dev: device to sync 3765 * @sync: function to call if address should be added 3766 * @unsync: function to call if address should be removed 3767 * 3768 * Add newly added addresses to the interface, and release 3769 * addresses that have been deleted. 3770 */ 3771 static inline int __dev_mc_sync(struct net_device *dev, 3772 int (*sync)(struct net_device *, 3773 const unsigned char *), 3774 int (*unsync)(struct net_device *, 3775 const unsigned char *)) 3776 { 3777 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 3778 } 3779 3780 /** 3781 * __dev_mc_unsync - Remove synchronized addresses from device 3782 * @dev: device to sync 3783 * @unsync: function to call if address should be removed 3784 * 3785 * Remove all addresses that were added to the device by dev_mc_sync(). 3786 */ 3787 static inline void __dev_mc_unsync(struct net_device *dev, 3788 int (*unsync)(struct net_device *, 3789 const unsigned char *)) 3790 { 3791 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 3792 } 3793 3794 /* Functions used for secondary unicast and multicast support */ 3795 void dev_set_rx_mode(struct net_device *dev); 3796 void __dev_set_rx_mode(struct net_device *dev); 3797 int dev_set_promiscuity(struct net_device *dev, int inc); 3798 int dev_set_allmulti(struct net_device *dev, int inc); 3799 void netdev_state_change(struct net_device *dev); 3800 void netdev_notify_peers(struct net_device *dev); 3801 void netdev_features_change(struct net_device *dev); 3802 /* Load a device via the kmod */ 3803 void dev_load(struct net *net, const char *name); 3804 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 3805 struct rtnl_link_stats64 *storage); 3806 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 3807 const struct net_device_stats *netdev_stats); 3808 3809 extern int netdev_max_backlog; 3810 extern int netdev_tstamp_prequeue; 3811 extern int weight_p; 3812 extern int dev_weight_rx_bias; 3813 extern int dev_weight_tx_bias; 3814 extern int dev_rx_weight; 3815 extern int dev_tx_weight; 3816 3817 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 3818 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 3819 struct list_head **iter); 3820 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 3821 struct list_head **iter); 3822 3823 /* iterate through upper list, must be called under RCU read lock */ 3824 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 3825 for (iter = &(dev)->adj_list.upper, \ 3826 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 3827 updev; \ 3828 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 3829 3830 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 3831 int (*fn)(struct net_device *upper_dev, 3832 void *data), 3833 void *data); 3834 3835 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 3836 struct net_device *upper_dev); 3837 3838 void *netdev_lower_get_next_private(struct net_device *dev, 3839 struct list_head **iter); 3840 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 3841 struct list_head **iter); 3842 3843 #define netdev_for_each_lower_private(dev, priv, iter) \ 3844 for (iter = (dev)->adj_list.lower.next, \ 3845 priv = netdev_lower_get_next_private(dev, &(iter)); \ 3846 priv; \ 3847 priv = netdev_lower_get_next_private(dev, &(iter))) 3848 3849 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 3850 for (iter = &(dev)->adj_list.lower, \ 3851 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 3852 priv; \ 3853 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 3854 3855 void *netdev_lower_get_next(struct net_device *dev, 3856 struct list_head **iter); 3857 3858 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 3859 for (iter = (dev)->adj_list.lower.next, \ 3860 ldev = netdev_lower_get_next(dev, &(iter)); \ 3861 ldev; \ 3862 ldev = netdev_lower_get_next(dev, &(iter))) 3863 3864 struct net_device *netdev_all_lower_get_next(struct net_device *dev, 3865 struct list_head **iter); 3866 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev, 3867 struct list_head **iter); 3868 3869 int netdev_walk_all_lower_dev(struct net_device *dev, 3870 int (*fn)(struct net_device *lower_dev, 3871 void *data), 3872 void *data); 3873 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 3874 int (*fn)(struct net_device *lower_dev, 3875 void *data), 3876 void *data); 3877 3878 void *netdev_adjacent_get_private(struct list_head *adj_list); 3879 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 3880 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 3881 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 3882 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev); 3883 int netdev_master_upper_dev_link(struct net_device *dev, 3884 struct net_device *upper_dev, 3885 void *upper_priv, void *upper_info); 3886 void netdev_upper_dev_unlink(struct net_device *dev, 3887 struct net_device *upper_dev); 3888 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 3889 void *netdev_lower_dev_get_private(struct net_device *dev, 3890 struct net_device *lower_dev); 3891 void netdev_lower_state_changed(struct net_device *lower_dev, 3892 void *lower_state_info); 3893 int netdev_default_l2upper_neigh_construct(struct net_device *dev, 3894 struct neighbour *n); 3895 void netdev_default_l2upper_neigh_destroy(struct net_device *dev, 3896 struct neighbour *n); 3897 3898 /* RSS keys are 40 or 52 bytes long */ 3899 #define NETDEV_RSS_KEY_LEN 52 3900 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 3901 void netdev_rss_key_fill(void *buffer, size_t len); 3902 3903 int dev_get_nest_level(struct net_device *dev); 3904 int skb_checksum_help(struct sk_buff *skb); 3905 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 3906 netdev_features_t features, bool tx_path); 3907 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 3908 netdev_features_t features); 3909 3910 struct netdev_bonding_info { 3911 ifslave slave; 3912 ifbond master; 3913 }; 3914 3915 struct netdev_notifier_bonding_info { 3916 struct netdev_notifier_info info; /* must be first */ 3917 struct netdev_bonding_info bonding_info; 3918 }; 3919 3920 void netdev_bonding_info_change(struct net_device *dev, 3921 struct netdev_bonding_info *bonding_info); 3922 3923 static inline 3924 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 3925 { 3926 return __skb_gso_segment(skb, features, true); 3927 } 3928 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 3929 3930 static inline bool can_checksum_protocol(netdev_features_t features, 3931 __be16 protocol) 3932 { 3933 if (protocol == htons(ETH_P_FCOE)) 3934 return !!(features & NETIF_F_FCOE_CRC); 3935 3936 /* Assume this is an IP checksum (not SCTP CRC) */ 3937 3938 if (features & NETIF_F_HW_CSUM) { 3939 /* Can checksum everything */ 3940 return true; 3941 } 3942 3943 switch (protocol) { 3944 case htons(ETH_P_IP): 3945 return !!(features & NETIF_F_IP_CSUM); 3946 case htons(ETH_P_IPV6): 3947 return !!(features & NETIF_F_IPV6_CSUM); 3948 default: 3949 return false; 3950 } 3951 } 3952 3953 #ifdef CONFIG_BUG 3954 void netdev_rx_csum_fault(struct net_device *dev); 3955 #else 3956 static inline void netdev_rx_csum_fault(struct net_device *dev) 3957 { 3958 } 3959 #endif 3960 /* rx skb timestamps */ 3961 void net_enable_timestamp(void); 3962 void net_disable_timestamp(void); 3963 3964 #ifdef CONFIG_PROC_FS 3965 int __init dev_proc_init(void); 3966 #else 3967 #define dev_proc_init() 0 3968 #endif 3969 3970 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 3971 struct sk_buff *skb, struct net_device *dev, 3972 bool more) 3973 { 3974 skb->xmit_more = more ? 1 : 0; 3975 return ops->ndo_start_xmit(skb, dev); 3976 } 3977 3978 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 3979 struct netdev_queue *txq, bool more) 3980 { 3981 const struct net_device_ops *ops = dev->netdev_ops; 3982 int rc; 3983 3984 rc = __netdev_start_xmit(ops, skb, dev, more); 3985 if (rc == NETDEV_TX_OK) 3986 txq_trans_update(txq); 3987 3988 return rc; 3989 } 3990 3991 int netdev_class_create_file_ns(struct class_attribute *class_attr, 3992 const void *ns); 3993 void netdev_class_remove_file_ns(struct class_attribute *class_attr, 3994 const void *ns); 3995 3996 static inline int netdev_class_create_file(struct class_attribute *class_attr) 3997 { 3998 return netdev_class_create_file_ns(class_attr, NULL); 3999 } 4000 4001 static inline void netdev_class_remove_file(struct class_attribute *class_attr) 4002 { 4003 netdev_class_remove_file_ns(class_attr, NULL); 4004 } 4005 4006 extern struct kobj_ns_type_operations net_ns_type_operations; 4007 4008 const char *netdev_drivername(const struct net_device *dev); 4009 4010 void linkwatch_run_queue(void); 4011 4012 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4013 netdev_features_t f2) 4014 { 4015 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4016 if (f1 & NETIF_F_HW_CSUM) 4017 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4018 else 4019 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4020 } 4021 4022 return f1 & f2; 4023 } 4024 4025 static inline netdev_features_t netdev_get_wanted_features( 4026 struct net_device *dev) 4027 { 4028 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4029 } 4030 netdev_features_t netdev_increment_features(netdev_features_t all, 4031 netdev_features_t one, netdev_features_t mask); 4032 4033 /* Allow TSO being used on stacked device : 4034 * Performing the GSO segmentation before last device 4035 * is a performance improvement. 4036 */ 4037 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4038 netdev_features_t mask) 4039 { 4040 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4041 } 4042 4043 int __netdev_update_features(struct net_device *dev); 4044 void netdev_update_features(struct net_device *dev); 4045 void netdev_change_features(struct net_device *dev); 4046 4047 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4048 struct net_device *dev); 4049 4050 netdev_features_t passthru_features_check(struct sk_buff *skb, 4051 struct net_device *dev, 4052 netdev_features_t features); 4053 netdev_features_t netif_skb_features(struct sk_buff *skb); 4054 4055 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4056 { 4057 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4058 4059 /* check flags correspondence */ 4060 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4061 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT)); 4062 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4063 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4064 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4065 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4066 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4067 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4068 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4069 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4070 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4071 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4072 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4073 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4074 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4075 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4076 4077 return (features & feature) == feature; 4078 } 4079 4080 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4081 { 4082 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4083 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4084 } 4085 4086 static inline bool netif_needs_gso(struct sk_buff *skb, 4087 netdev_features_t features) 4088 { 4089 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4090 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4091 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4092 } 4093 4094 static inline void netif_set_gso_max_size(struct net_device *dev, 4095 unsigned int size) 4096 { 4097 dev->gso_max_size = size; 4098 } 4099 4100 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4101 int pulled_hlen, u16 mac_offset, 4102 int mac_len) 4103 { 4104 skb->protocol = protocol; 4105 skb->encapsulation = 1; 4106 skb_push(skb, pulled_hlen); 4107 skb_reset_transport_header(skb); 4108 skb->mac_header = mac_offset; 4109 skb->network_header = skb->mac_header + mac_len; 4110 skb->mac_len = mac_len; 4111 } 4112 4113 static inline bool netif_is_macsec(const struct net_device *dev) 4114 { 4115 return dev->priv_flags & IFF_MACSEC; 4116 } 4117 4118 static inline bool netif_is_macvlan(const struct net_device *dev) 4119 { 4120 return dev->priv_flags & IFF_MACVLAN; 4121 } 4122 4123 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4124 { 4125 return dev->priv_flags & IFF_MACVLAN_PORT; 4126 } 4127 4128 static inline bool netif_is_ipvlan(const struct net_device *dev) 4129 { 4130 return dev->priv_flags & IFF_IPVLAN_SLAVE; 4131 } 4132 4133 static inline bool netif_is_ipvlan_port(const struct net_device *dev) 4134 { 4135 return dev->priv_flags & IFF_IPVLAN_MASTER; 4136 } 4137 4138 static inline bool netif_is_bond_master(const struct net_device *dev) 4139 { 4140 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4141 } 4142 4143 static inline bool netif_is_bond_slave(const struct net_device *dev) 4144 { 4145 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4146 } 4147 4148 static inline bool netif_supports_nofcs(struct net_device *dev) 4149 { 4150 return dev->priv_flags & IFF_SUPP_NOFCS; 4151 } 4152 4153 static inline bool netif_is_l3_master(const struct net_device *dev) 4154 { 4155 return dev->priv_flags & IFF_L3MDEV_MASTER; 4156 } 4157 4158 static inline bool netif_is_l3_slave(const struct net_device *dev) 4159 { 4160 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4161 } 4162 4163 static inline bool netif_is_bridge_master(const struct net_device *dev) 4164 { 4165 return dev->priv_flags & IFF_EBRIDGE; 4166 } 4167 4168 static inline bool netif_is_bridge_port(const struct net_device *dev) 4169 { 4170 return dev->priv_flags & IFF_BRIDGE_PORT; 4171 } 4172 4173 static inline bool netif_is_ovs_master(const struct net_device *dev) 4174 { 4175 return dev->priv_flags & IFF_OPENVSWITCH; 4176 } 4177 4178 static inline bool netif_is_team_master(const struct net_device *dev) 4179 { 4180 return dev->priv_flags & IFF_TEAM; 4181 } 4182 4183 static inline bool netif_is_team_port(const struct net_device *dev) 4184 { 4185 return dev->priv_flags & IFF_TEAM_PORT; 4186 } 4187 4188 static inline bool netif_is_lag_master(const struct net_device *dev) 4189 { 4190 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4191 } 4192 4193 static inline bool netif_is_lag_port(const struct net_device *dev) 4194 { 4195 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4196 } 4197 4198 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4199 { 4200 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4201 } 4202 4203 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4204 static inline void netif_keep_dst(struct net_device *dev) 4205 { 4206 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4207 } 4208 4209 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4210 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4211 { 4212 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4213 return dev->priv_flags & IFF_MACSEC; 4214 } 4215 4216 extern struct pernet_operations __net_initdata loopback_net_ops; 4217 4218 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4219 4220 /* netdev_printk helpers, similar to dev_printk */ 4221 4222 static inline const char *netdev_name(const struct net_device *dev) 4223 { 4224 if (!dev->name[0] || strchr(dev->name, '%')) 4225 return "(unnamed net_device)"; 4226 return dev->name; 4227 } 4228 4229 static inline const char *netdev_reg_state(const struct net_device *dev) 4230 { 4231 switch (dev->reg_state) { 4232 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4233 case NETREG_REGISTERED: return ""; 4234 case NETREG_UNREGISTERING: return " (unregistering)"; 4235 case NETREG_UNREGISTERED: return " (unregistered)"; 4236 case NETREG_RELEASED: return " (released)"; 4237 case NETREG_DUMMY: return " (dummy)"; 4238 } 4239 4240 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4241 return " (unknown)"; 4242 } 4243 4244 __printf(3, 4) 4245 void netdev_printk(const char *level, const struct net_device *dev, 4246 const char *format, ...); 4247 __printf(2, 3) 4248 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4249 __printf(2, 3) 4250 void netdev_alert(const struct net_device *dev, const char *format, ...); 4251 __printf(2, 3) 4252 void netdev_crit(const struct net_device *dev, const char *format, ...); 4253 __printf(2, 3) 4254 void netdev_err(const struct net_device *dev, const char *format, ...); 4255 __printf(2, 3) 4256 void netdev_warn(const struct net_device *dev, const char *format, ...); 4257 __printf(2, 3) 4258 void netdev_notice(const struct net_device *dev, const char *format, ...); 4259 __printf(2, 3) 4260 void netdev_info(const struct net_device *dev, const char *format, ...); 4261 4262 #define MODULE_ALIAS_NETDEV(device) \ 4263 MODULE_ALIAS("netdev-" device) 4264 4265 #if defined(CONFIG_DYNAMIC_DEBUG) 4266 #define netdev_dbg(__dev, format, args...) \ 4267 do { \ 4268 dynamic_netdev_dbg(__dev, format, ##args); \ 4269 } while (0) 4270 #elif defined(DEBUG) 4271 #define netdev_dbg(__dev, format, args...) \ 4272 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4273 #else 4274 #define netdev_dbg(__dev, format, args...) \ 4275 ({ \ 4276 if (0) \ 4277 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4278 }) 4279 #endif 4280 4281 #if defined(VERBOSE_DEBUG) 4282 #define netdev_vdbg netdev_dbg 4283 #else 4284 4285 #define netdev_vdbg(dev, format, args...) \ 4286 ({ \ 4287 if (0) \ 4288 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 4289 0; \ 4290 }) 4291 #endif 4292 4293 /* 4294 * netdev_WARN() acts like dev_printk(), but with the key difference 4295 * of using a WARN/WARN_ON to get the message out, including the 4296 * file/line information and a backtrace. 4297 */ 4298 #define netdev_WARN(dev, format, args...) \ 4299 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \ 4300 netdev_reg_state(dev), ##args) 4301 4302 /* netif printk helpers, similar to netdev_printk */ 4303 4304 #define netif_printk(priv, type, level, dev, fmt, args...) \ 4305 do { \ 4306 if (netif_msg_##type(priv)) \ 4307 netdev_printk(level, (dev), fmt, ##args); \ 4308 } while (0) 4309 4310 #define netif_level(level, priv, type, dev, fmt, args...) \ 4311 do { \ 4312 if (netif_msg_##type(priv)) \ 4313 netdev_##level(dev, fmt, ##args); \ 4314 } while (0) 4315 4316 #define netif_emerg(priv, type, dev, fmt, args...) \ 4317 netif_level(emerg, priv, type, dev, fmt, ##args) 4318 #define netif_alert(priv, type, dev, fmt, args...) \ 4319 netif_level(alert, priv, type, dev, fmt, ##args) 4320 #define netif_crit(priv, type, dev, fmt, args...) \ 4321 netif_level(crit, priv, type, dev, fmt, ##args) 4322 #define netif_err(priv, type, dev, fmt, args...) \ 4323 netif_level(err, priv, type, dev, fmt, ##args) 4324 #define netif_warn(priv, type, dev, fmt, args...) \ 4325 netif_level(warn, priv, type, dev, fmt, ##args) 4326 #define netif_notice(priv, type, dev, fmt, args...) \ 4327 netif_level(notice, priv, type, dev, fmt, ##args) 4328 #define netif_info(priv, type, dev, fmt, args...) \ 4329 netif_level(info, priv, type, dev, fmt, ##args) 4330 4331 #if defined(CONFIG_DYNAMIC_DEBUG) 4332 #define netif_dbg(priv, type, netdev, format, args...) \ 4333 do { \ 4334 if (netif_msg_##type(priv)) \ 4335 dynamic_netdev_dbg(netdev, format, ##args); \ 4336 } while (0) 4337 #elif defined(DEBUG) 4338 #define netif_dbg(priv, type, dev, format, args...) \ 4339 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 4340 #else 4341 #define netif_dbg(priv, type, dev, format, args...) \ 4342 ({ \ 4343 if (0) \ 4344 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4345 0; \ 4346 }) 4347 #endif 4348 4349 #if defined(VERBOSE_DEBUG) 4350 #define netif_vdbg netif_dbg 4351 #else 4352 #define netif_vdbg(priv, type, dev, format, args...) \ 4353 ({ \ 4354 if (0) \ 4355 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4356 0; \ 4357 }) 4358 #endif 4359 4360 /* 4361 * The list of packet types we will receive (as opposed to discard) 4362 * and the routines to invoke. 4363 * 4364 * Why 16. Because with 16 the only overlap we get on a hash of the 4365 * low nibble of the protocol value is RARP/SNAP/X.25. 4366 * 4367 * NOTE: That is no longer true with the addition of VLAN tags. Not 4368 * sure which should go first, but I bet it won't make much 4369 * difference if we are running VLANs. The good news is that 4370 * this protocol won't be in the list unless compiled in, so 4371 * the average user (w/out VLANs) will not be adversely affected. 4372 * --BLG 4373 * 4374 * 0800 IP 4375 * 8100 802.1Q VLAN 4376 * 0001 802.3 4377 * 0002 AX.25 4378 * 0004 802.2 4379 * 8035 RARP 4380 * 0005 SNAP 4381 * 0805 X.25 4382 * 0806 ARP 4383 * 8137 IPX 4384 * 0009 Localtalk 4385 * 86DD IPv6 4386 */ 4387 #define PTYPE_HASH_SIZE (16) 4388 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 4389 4390 #endif /* _LINUX_NETDEVICE_H */ 4391