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 GRO_CONSUMED, 356 }; 357 typedef enum gro_result gro_result_t; 358 359 /* 360 * enum rx_handler_result - Possible return values for rx_handlers. 361 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 362 * further. 363 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 364 * case skb->dev was changed by rx_handler. 365 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 366 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 367 * 368 * rx_handlers are functions called from inside __netif_receive_skb(), to do 369 * special processing of the skb, prior to delivery to protocol handlers. 370 * 371 * Currently, a net_device can only have a single rx_handler registered. Trying 372 * to register a second rx_handler will return -EBUSY. 373 * 374 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 375 * To unregister a rx_handler on a net_device, use 376 * netdev_rx_handler_unregister(). 377 * 378 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 379 * do with the skb. 380 * 381 * If the rx_handler consumed the skb in some way, it should return 382 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 383 * the skb to be delivered in some other way. 384 * 385 * If the rx_handler changed skb->dev, to divert the skb to another 386 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 387 * new device will be called if it exists. 388 * 389 * If the rx_handler decides the skb should be ignored, it should return 390 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 391 * are registered on exact device (ptype->dev == skb->dev). 392 * 393 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 394 * delivered, it should return RX_HANDLER_PASS. 395 * 396 * A device without a registered rx_handler will behave as if rx_handler 397 * returned RX_HANDLER_PASS. 398 */ 399 400 enum rx_handler_result { 401 RX_HANDLER_CONSUMED, 402 RX_HANDLER_ANOTHER, 403 RX_HANDLER_EXACT, 404 RX_HANDLER_PASS, 405 }; 406 typedef enum rx_handler_result rx_handler_result_t; 407 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 408 409 void __napi_schedule(struct napi_struct *n); 410 void __napi_schedule_irqoff(struct napi_struct *n); 411 412 static inline bool napi_disable_pending(struct napi_struct *n) 413 { 414 return test_bit(NAPI_STATE_DISABLE, &n->state); 415 } 416 417 /** 418 * napi_schedule_prep - check if NAPI can be scheduled 419 * @n: NAPI context 420 * 421 * Test if NAPI routine is already running, and if not mark 422 * it as running. This is used as a condition variable to 423 * insure only one NAPI poll instance runs. We also make 424 * sure there is no pending NAPI disable. 425 */ 426 static inline bool napi_schedule_prep(struct napi_struct *n) 427 { 428 return !napi_disable_pending(n) && 429 !test_and_set_bit(NAPI_STATE_SCHED, &n->state); 430 } 431 432 /** 433 * napi_schedule - schedule NAPI poll 434 * @n: NAPI context 435 * 436 * Schedule NAPI poll routine to be called if it is not already 437 * running. 438 */ 439 static inline void napi_schedule(struct napi_struct *n) 440 { 441 if (napi_schedule_prep(n)) 442 __napi_schedule(n); 443 } 444 445 /** 446 * napi_schedule_irqoff - schedule NAPI poll 447 * @n: NAPI context 448 * 449 * Variant of napi_schedule(), assuming hard irqs are masked. 450 */ 451 static inline void napi_schedule_irqoff(struct napi_struct *n) 452 { 453 if (napi_schedule_prep(n)) 454 __napi_schedule_irqoff(n); 455 } 456 457 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 458 static inline bool napi_reschedule(struct napi_struct *napi) 459 { 460 if (napi_schedule_prep(napi)) { 461 __napi_schedule(napi); 462 return true; 463 } 464 return false; 465 } 466 467 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_features_check)(struct sk_buff *skb, 870 * struct net_device *dev 871 * netdev_features_t features); 872 * Called by core transmit path to determine if device is capable of 873 * performing offload operations on a given packet. This is to give 874 * the device an opportunity to implement any restrictions that cannot 875 * be otherwise expressed by feature flags. The check is called with 876 * the set of features that the stack has calculated and it returns 877 * those the driver believes to be appropriate. 878 * 879 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 880 * void *accel_priv, select_queue_fallback_t fallback); 881 * Called to decide which queue to use when device supports multiple 882 * transmit queues. 883 * 884 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 885 * This function is called to allow device receiver to make 886 * changes to configuration when multicast or promiscuous is enabled. 887 * 888 * void (*ndo_set_rx_mode)(struct net_device *dev); 889 * This function is called device changes address list filtering. 890 * If driver handles unicast address filtering, it should set 891 * IFF_UNICAST_FLT in its priv_flags. 892 * 893 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 894 * This function is called when the Media Access Control address 895 * needs to be changed. If this interface is not defined, the 896 * MAC address can not be changed. 897 * 898 * int (*ndo_validate_addr)(struct net_device *dev); 899 * Test if Media Access Control address is valid for the device. 900 * 901 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 902 * Called when a user requests an ioctl which can't be handled by 903 * the generic interface code. If not defined ioctls return 904 * not supported error code. 905 * 906 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 907 * Used to set network devices bus interface parameters. This interface 908 * is retained for legacy reasons; new devices should use the bus 909 * interface (PCI) for low level management. 910 * 911 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 912 * Called when a user wants to change the Maximum Transfer Unit 913 * of a device. If not defined, any request to change MTU will 914 * will return an error. 915 * 916 * void (*ndo_tx_timeout)(struct net_device *dev); 917 * Callback used when the transmitter has not made any progress 918 * for dev->watchdog ticks. 919 * 920 * void (*ndo_get_stats64)(struct net_device *dev, 921 * struct rtnl_link_stats64 *storage); 922 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 923 * Called when a user wants to get the network device usage 924 * statistics. Drivers must do one of the following: 925 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 926 * rtnl_link_stats64 structure passed by the caller. 927 * 2. Define @ndo_get_stats to update a net_device_stats structure 928 * (which should normally be dev->stats) and return a pointer to 929 * it. The structure may be changed asynchronously only if each 930 * field is written atomically. 931 * 3. Update dev->stats asynchronously and atomically, and define 932 * neither operation. 933 * 934 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 935 * Return true if this device supports offload stats of this attr_id. 936 * 937 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 938 * void *attr_data) 939 * Get statistics for offload operations by attr_id. Write it into the 940 * attr_data pointer. 941 * 942 * int (*ndo_vlan_rx_add_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 registered. 945 * 946 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 947 * If device supports VLAN filtering this function is called when a 948 * VLAN id is unregistered. 949 * 950 * void (*ndo_poll_controller)(struct net_device *dev); 951 * 952 * SR-IOV management functions. 953 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 954 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 955 * u8 qos, __be16 proto); 956 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 957 * int max_tx_rate); 958 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 959 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 960 * int (*ndo_get_vf_config)(struct net_device *dev, 961 * int vf, struct ifla_vf_info *ivf); 962 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 963 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 964 * struct nlattr *port[]); 965 * 966 * Enable or disable the VF ability to query its RSS Redirection Table and 967 * Hash Key. This is needed since on some devices VF share this information 968 * with PF and querying it may introduce a theoretical security risk. 969 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 970 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 971 * int (*ndo_setup_tc)(struct net_device *dev, u32 handle, 972 * __be16 protocol, struct tc_to_netdev *tc); 973 * Called to setup any 'tc' scheduler, classifier or action on @dev. 974 * This is always called from the stack with the rtnl lock held and netif 975 * tx queues stopped. This allows the netdevice to perform queue 976 * management safely. 977 * 978 * Fiber Channel over Ethernet (FCoE) offload functions. 979 * int (*ndo_fcoe_enable)(struct net_device *dev); 980 * Called when the FCoE protocol stack wants to start using LLD for FCoE 981 * so the underlying device can perform whatever needed configuration or 982 * initialization to support acceleration of FCoE traffic. 983 * 984 * int (*ndo_fcoe_disable)(struct net_device *dev); 985 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 986 * so the underlying device can perform whatever needed clean-ups to 987 * stop supporting acceleration of FCoE traffic. 988 * 989 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 990 * struct scatterlist *sgl, unsigned int sgc); 991 * Called when the FCoE Initiator wants to initialize an I/O that 992 * is a possible candidate for Direct Data Placement (DDP). The LLD can 993 * perform necessary setup and returns 1 to indicate the device is set up 994 * successfully to perform DDP on this I/O, otherwise this returns 0. 995 * 996 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 997 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 998 * indicated by the FC exchange id 'xid', so the underlying device can 999 * clean up and reuse resources for later DDP requests. 1000 * 1001 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1002 * struct scatterlist *sgl, unsigned int sgc); 1003 * Called when the FCoE Target wants to initialize an I/O that 1004 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1005 * perform necessary setup and returns 1 to indicate the device is set up 1006 * successfully to perform DDP on this I/O, otherwise this returns 0. 1007 * 1008 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1009 * struct netdev_fcoe_hbainfo *hbainfo); 1010 * Called when the FCoE Protocol stack wants information on the underlying 1011 * device. This information is utilized by the FCoE protocol stack to 1012 * register attributes with Fiber Channel management service as per the 1013 * FC-GS Fabric Device Management Information(FDMI) specification. 1014 * 1015 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1016 * Called when the underlying device wants to override default World Wide 1017 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1018 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1019 * protocol stack to use. 1020 * 1021 * RFS acceleration. 1022 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1023 * u16 rxq_index, u32 flow_id); 1024 * Set hardware filter for RFS. rxq_index is the target queue index; 1025 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1026 * Return the filter ID on success, or a negative error code. 1027 * 1028 * Slave management functions (for bridge, bonding, etc). 1029 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1030 * Called to make another netdev an underling. 1031 * 1032 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1033 * Called to release previously enslaved netdev. 1034 * 1035 * Feature/offload setting functions. 1036 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1037 * netdev_features_t features); 1038 * Adjusts the requested feature flags according to device-specific 1039 * constraints, and returns the resulting flags. Must not modify 1040 * the device state. 1041 * 1042 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1043 * Called to update device configuration to new features. Passed 1044 * feature set might be less than what was returned by ndo_fix_features()). 1045 * Must return >0 or -errno if it changed dev->features itself. 1046 * 1047 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1048 * struct net_device *dev, 1049 * const unsigned char *addr, u16 vid, u16 flags) 1050 * Adds an FDB entry to dev for addr. 1051 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1052 * struct net_device *dev, 1053 * const unsigned char *addr, u16 vid) 1054 * Deletes the FDB entry from dev coresponding to addr. 1055 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1056 * struct net_device *dev, struct net_device *filter_dev, 1057 * int *idx) 1058 * Used to add FDB entries to dump requests. Implementers should add 1059 * entries to skb and update idx with the number of entries. 1060 * 1061 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1062 * u16 flags) 1063 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1064 * struct net_device *dev, u32 filter_mask, 1065 * int nlflags) 1066 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1067 * u16 flags); 1068 * 1069 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1070 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1071 * which do not represent real hardware may define this to allow their 1072 * userspace components to manage their virtual carrier state. Devices 1073 * that determine carrier state from physical hardware properties (eg 1074 * network cables) or protocol-dependent mechanisms (eg 1075 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1076 * 1077 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1078 * struct netdev_phys_item_id *ppid); 1079 * Called to get ID of physical port of this device. If driver does 1080 * not implement this, it is assumed that the hw is not able to have 1081 * multiple net devices on single physical port. 1082 * 1083 * void (*ndo_udp_tunnel_add)(struct net_device *dev, 1084 * struct udp_tunnel_info *ti); 1085 * Called by UDP tunnel to notify a driver about the UDP port and socket 1086 * address family that a UDP tunnel is listnening to. It is called only 1087 * when a new port starts listening. The operation is protected by the 1088 * RTNL. 1089 * 1090 * void (*ndo_udp_tunnel_del)(struct net_device *dev, 1091 * struct udp_tunnel_info *ti); 1092 * Called by UDP tunnel to notify the driver about a UDP port and socket 1093 * address family that the UDP tunnel is not listening to anymore. The 1094 * operation is protected by the RTNL. 1095 * 1096 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1097 * struct net_device *dev) 1098 * Called by upper layer devices to accelerate switching or other 1099 * station functionality into hardware. 'pdev is the lowerdev 1100 * to use for the offload and 'dev' is the net device that will 1101 * back the offload. Returns a pointer to the private structure 1102 * the upper layer will maintain. 1103 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1104 * Called by upper layer device to delete the station created 1105 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1106 * the station and priv is the structure returned by the add 1107 * operation. 1108 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb, 1109 * struct net_device *dev, 1110 * void *priv); 1111 * Callback to use for xmit over the accelerated station. This 1112 * is used in place of ndo_start_xmit on accelerated net 1113 * devices. 1114 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1115 * int queue_index, u32 maxrate); 1116 * Called when a user wants to set a max-rate limitation of specific 1117 * TX queue. 1118 * int (*ndo_get_iflink)(const struct net_device *dev); 1119 * Called to get the iflink value of this device. 1120 * void (*ndo_change_proto_down)(struct net_device *dev, 1121 * bool proto_down); 1122 * This function is used to pass protocol port error state information 1123 * to the switch driver. The switch driver can react to the proto_down 1124 * by doing a phys down on the associated switch port. 1125 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1126 * This function is used to get egress tunnel information for given skb. 1127 * This is useful for retrieving outer tunnel header parameters while 1128 * sampling packet. 1129 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1130 * This function is used to specify the headroom that the skb must 1131 * consider when allocation skb during packet reception. Setting 1132 * appropriate rx headroom value allows avoiding skb head copy on 1133 * forward. Setting a negative value resets the rx headroom to the 1134 * default value. 1135 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp); 1136 * This function is used to set or query state related to XDP on the 1137 * netdevice. See definition of enum xdp_netdev_command for details. 1138 * 1139 */ 1140 struct net_device_ops { 1141 int (*ndo_init)(struct net_device *dev); 1142 void (*ndo_uninit)(struct net_device *dev); 1143 int (*ndo_open)(struct net_device *dev); 1144 int (*ndo_stop)(struct net_device *dev); 1145 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1146 struct net_device *dev); 1147 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1148 struct net_device *dev, 1149 netdev_features_t features); 1150 u16 (*ndo_select_queue)(struct net_device *dev, 1151 struct sk_buff *skb, 1152 void *accel_priv, 1153 select_queue_fallback_t fallback); 1154 void (*ndo_change_rx_flags)(struct net_device *dev, 1155 int flags); 1156 void (*ndo_set_rx_mode)(struct net_device *dev); 1157 int (*ndo_set_mac_address)(struct net_device *dev, 1158 void *addr); 1159 int (*ndo_validate_addr)(struct net_device *dev); 1160 int (*ndo_do_ioctl)(struct net_device *dev, 1161 struct ifreq *ifr, int cmd); 1162 int (*ndo_set_config)(struct net_device *dev, 1163 struct ifmap *map); 1164 int (*ndo_change_mtu)(struct net_device *dev, 1165 int new_mtu); 1166 int (*ndo_neigh_setup)(struct net_device *dev, 1167 struct neigh_parms *); 1168 void (*ndo_tx_timeout) (struct net_device *dev); 1169 1170 void (*ndo_get_stats64)(struct net_device *dev, 1171 struct rtnl_link_stats64 *storage); 1172 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1173 int (*ndo_get_offload_stats)(int attr_id, 1174 const struct net_device *dev, 1175 void *attr_data); 1176 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1177 1178 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1179 __be16 proto, u16 vid); 1180 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1181 __be16 proto, u16 vid); 1182 #ifdef CONFIG_NET_POLL_CONTROLLER 1183 void (*ndo_poll_controller)(struct net_device *dev); 1184 int (*ndo_netpoll_setup)(struct net_device *dev, 1185 struct netpoll_info *info); 1186 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1187 #endif 1188 int (*ndo_set_vf_mac)(struct net_device *dev, 1189 int queue, u8 *mac); 1190 int (*ndo_set_vf_vlan)(struct net_device *dev, 1191 int queue, u16 vlan, 1192 u8 qos, __be16 proto); 1193 int (*ndo_set_vf_rate)(struct net_device *dev, 1194 int vf, int min_tx_rate, 1195 int max_tx_rate); 1196 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1197 int vf, bool setting); 1198 int (*ndo_set_vf_trust)(struct net_device *dev, 1199 int vf, bool setting); 1200 int (*ndo_get_vf_config)(struct net_device *dev, 1201 int vf, 1202 struct ifla_vf_info *ivf); 1203 int (*ndo_set_vf_link_state)(struct net_device *dev, 1204 int vf, int link_state); 1205 int (*ndo_get_vf_stats)(struct net_device *dev, 1206 int vf, 1207 struct ifla_vf_stats 1208 *vf_stats); 1209 int (*ndo_set_vf_port)(struct net_device *dev, 1210 int vf, 1211 struct nlattr *port[]); 1212 int (*ndo_get_vf_port)(struct net_device *dev, 1213 int vf, struct sk_buff *skb); 1214 int (*ndo_set_vf_guid)(struct net_device *dev, 1215 int vf, u64 guid, 1216 int guid_type); 1217 int (*ndo_set_vf_rss_query_en)( 1218 struct net_device *dev, 1219 int vf, bool setting); 1220 int (*ndo_setup_tc)(struct net_device *dev, 1221 u32 handle, 1222 __be16 protocol, 1223 struct tc_to_netdev *tc); 1224 #if IS_ENABLED(CONFIG_FCOE) 1225 int (*ndo_fcoe_enable)(struct net_device *dev); 1226 int (*ndo_fcoe_disable)(struct net_device *dev); 1227 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1228 u16 xid, 1229 struct scatterlist *sgl, 1230 unsigned int sgc); 1231 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1232 u16 xid); 1233 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1234 u16 xid, 1235 struct scatterlist *sgl, 1236 unsigned int sgc); 1237 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1238 struct netdev_fcoe_hbainfo *hbainfo); 1239 #endif 1240 1241 #if IS_ENABLED(CONFIG_LIBFCOE) 1242 #define NETDEV_FCOE_WWNN 0 1243 #define NETDEV_FCOE_WWPN 1 1244 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1245 u64 *wwn, int type); 1246 #endif 1247 1248 #ifdef CONFIG_RFS_ACCEL 1249 int (*ndo_rx_flow_steer)(struct net_device *dev, 1250 const struct sk_buff *skb, 1251 u16 rxq_index, 1252 u32 flow_id); 1253 #endif 1254 int (*ndo_add_slave)(struct net_device *dev, 1255 struct net_device *slave_dev); 1256 int (*ndo_del_slave)(struct net_device *dev, 1257 struct net_device *slave_dev); 1258 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1259 netdev_features_t features); 1260 int (*ndo_set_features)(struct net_device *dev, 1261 netdev_features_t features); 1262 int (*ndo_neigh_construct)(struct net_device *dev, 1263 struct neighbour *n); 1264 void (*ndo_neigh_destroy)(struct net_device *dev, 1265 struct neighbour *n); 1266 1267 int (*ndo_fdb_add)(struct ndmsg *ndm, 1268 struct nlattr *tb[], 1269 struct net_device *dev, 1270 const unsigned char *addr, 1271 u16 vid, 1272 u16 flags); 1273 int (*ndo_fdb_del)(struct ndmsg *ndm, 1274 struct nlattr *tb[], 1275 struct net_device *dev, 1276 const unsigned char *addr, 1277 u16 vid); 1278 int (*ndo_fdb_dump)(struct sk_buff *skb, 1279 struct netlink_callback *cb, 1280 struct net_device *dev, 1281 struct net_device *filter_dev, 1282 int *idx); 1283 1284 int (*ndo_bridge_setlink)(struct net_device *dev, 1285 struct nlmsghdr *nlh, 1286 u16 flags); 1287 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1288 u32 pid, u32 seq, 1289 struct net_device *dev, 1290 u32 filter_mask, 1291 int nlflags); 1292 int (*ndo_bridge_dellink)(struct net_device *dev, 1293 struct nlmsghdr *nlh, 1294 u16 flags); 1295 int (*ndo_change_carrier)(struct net_device *dev, 1296 bool new_carrier); 1297 int (*ndo_get_phys_port_id)(struct net_device *dev, 1298 struct netdev_phys_item_id *ppid); 1299 int (*ndo_get_phys_port_name)(struct net_device *dev, 1300 char *name, size_t len); 1301 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1302 struct udp_tunnel_info *ti); 1303 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1304 struct udp_tunnel_info *ti); 1305 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1306 struct net_device *dev); 1307 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1308 void *priv); 1309 1310 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb, 1311 struct net_device *dev, 1312 void *priv); 1313 int (*ndo_get_lock_subclass)(struct net_device *dev); 1314 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1315 int queue_index, 1316 u32 maxrate); 1317 int (*ndo_get_iflink)(const struct net_device *dev); 1318 int (*ndo_change_proto_down)(struct net_device *dev, 1319 bool proto_down); 1320 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1321 struct sk_buff *skb); 1322 void (*ndo_set_rx_headroom)(struct net_device *dev, 1323 int needed_headroom); 1324 int (*ndo_xdp)(struct net_device *dev, 1325 struct netdev_xdp *xdp); 1326 }; 1327 1328 /** 1329 * enum net_device_priv_flags - &struct net_device priv_flags 1330 * 1331 * These are the &struct net_device, they are only set internally 1332 * by drivers and used in the kernel. These flags are invisible to 1333 * userspace; this means that the order of these flags can change 1334 * during any kernel release. 1335 * 1336 * You should have a pretty good reason to be extending these flags. 1337 * 1338 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1339 * @IFF_EBRIDGE: Ethernet bridging device 1340 * @IFF_BONDING: bonding master or slave 1341 * @IFF_ISATAP: ISATAP interface (RFC4214) 1342 * @IFF_WAN_HDLC: WAN HDLC device 1343 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1344 * release skb->dst 1345 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1346 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1347 * @IFF_MACVLAN_PORT: device used as macvlan port 1348 * @IFF_BRIDGE_PORT: device used as bridge port 1349 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1350 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1351 * @IFF_UNICAST_FLT: Supports unicast filtering 1352 * @IFF_TEAM_PORT: device used as team port 1353 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1354 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1355 * change when it's running 1356 * @IFF_MACVLAN: Macvlan device 1357 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1358 * underlying stacked devices 1359 * @IFF_IPVLAN_MASTER: IPvlan master device 1360 * @IFF_IPVLAN_SLAVE: IPvlan slave device 1361 * @IFF_L3MDEV_MASTER: device is an L3 master device 1362 * @IFF_NO_QUEUE: device can run without qdisc attached 1363 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1364 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1365 * @IFF_TEAM: device is a team device 1366 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1367 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1368 * entity (i.e. the master device for bridged veth) 1369 * @IFF_MACSEC: device is a MACsec device 1370 */ 1371 enum netdev_priv_flags { 1372 IFF_802_1Q_VLAN = 1<<0, 1373 IFF_EBRIDGE = 1<<1, 1374 IFF_BONDING = 1<<2, 1375 IFF_ISATAP = 1<<3, 1376 IFF_WAN_HDLC = 1<<4, 1377 IFF_XMIT_DST_RELEASE = 1<<5, 1378 IFF_DONT_BRIDGE = 1<<6, 1379 IFF_DISABLE_NETPOLL = 1<<7, 1380 IFF_MACVLAN_PORT = 1<<8, 1381 IFF_BRIDGE_PORT = 1<<9, 1382 IFF_OVS_DATAPATH = 1<<10, 1383 IFF_TX_SKB_SHARING = 1<<11, 1384 IFF_UNICAST_FLT = 1<<12, 1385 IFF_TEAM_PORT = 1<<13, 1386 IFF_SUPP_NOFCS = 1<<14, 1387 IFF_LIVE_ADDR_CHANGE = 1<<15, 1388 IFF_MACVLAN = 1<<16, 1389 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1390 IFF_IPVLAN_MASTER = 1<<18, 1391 IFF_IPVLAN_SLAVE = 1<<19, 1392 IFF_L3MDEV_MASTER = 1<<20, 1393 IFF_NO_QUEUE = 1<<21, 1394 IFF_OPENVSWITCH = 1<<22, 1395 IFF_L3MDEV_SLAVE = 1<<23, 1396 IFF_TEAM = 1<<24, 1397 IFF_RXFH_CONFIGURED = 1<<25, 1398 IFF_PHONY_HEADROOM = 1<<26, 1399 IFF_MACSEC = 1<<27, 1400 }; 1401 1402 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1403 #define IFF_EBRIDGE IFF_EBRIDGE 1404 #define IFF_BONDING IFF_BONDING 1405 #define IFF_ISATAP IFF_ISATAP 1406 #define IFF_WAN_HDLC IFF_WAN_HDLC 1407 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1408 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1409 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1410 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1411 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1412 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1413 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1414 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1415 #define IFF_TEAM_PORT IFF_TEAM_PORT 1416 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1417 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1418 #define IFF_MACVLAN IFF_MACVLAN 1419 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1420 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER 1421 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE 1422 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1423 #define IFF_NO_QUEUE IFF_NO_QUEUE 1424 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1425 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1426 #define IFF_TEAM IFF_TEAM 1427 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1428 #define IFF_MACSEC IFF_MACSEC 1429 1430 /** 1431 * struct net_device - The DEVICE structure. 1432 * Actually, this whole structure is a big mistake. It mixes I/O 1433 * data with strictly "high-level" data, and it has to know about 1434 * almost every data structure used in the INET module. 1435 * 1436 * @name: This is the first field of the "visible" part of this structure 1437 * (i.e. as seen by users in the "Space.c" file). It is the name 1438 * of the interface. 1439 * 1440 * @name_hlist: Device name hash chain, please keep it close to name[] 1441 * @ifalias: SNMP alias 1442 * @mem_end: Shared memory end 1443 * @mem_start: Shared memory start 1444 * @base_addr: Device I/O address 1445 * @irq: Device IRQ number 1446 * 1447 * @carrier_changes: Stats to monitor carrier on<->off transitions 1448 * 1449 * @state: Generic network queuing layer state, see netdev_state_t 1450 * @dev_list: The global list of network devices 1451 * @napi_list: List entry used for polling NAPI devices 1452 * @unreg_list: List entry when we are unregistering the 1453 * device; see the function unregister_netdev 1454 * @close_list: List entry used when we are closing the device 1455 * @ptype_all: Device-specific packet handlers for all protocols 1456 * @ptype_specific: Device-specific, protocol-specific packet handlers 1457 * 1458 * @adj_list: Directly linked devices, like slaves for bonding 1459 * @features: Currently active device features 1460 * @hw_features: User-changeable features 1461 * 1462 * @wanted_features: User-requested features 1463 * @vlan_features: Mask of features inheritable by VLAN devices 1464 * 1465 * @hw_enc_features: Mask of features inherited by encapsulating devices 1466 * This field indicates what encapsulation 1467 * offloads the hardware is capable of doing, 1468 * and drivers will need to set them appropriately. 1469 * 1470 * @mpls_features: Mask of features inheritable by MPLS 1471 * 1472 * @ifindex: interface index 1473 * @group: The group the device belongs to 1474 * 1475 * @stats: Statistics struct, which was left as a legacy, use 1476 * rtnl_link_stats64 instead 1477 * 1478 * @rx_dropped: Dropped packets by core network, 1479 * do not use this in drivers 1480 * @tx_dropped: Dropped packets by core network, 1481 * do not use this in drivers 1482 * @rx_nohandler: nohandler dropped packets by core network on 1483 * inactive devices, do not use this in drivers 1484 * 1485 * @wireless_handlers: List of functions to handle Wireless Extensions, 1486 * instead of ioctl, 1487 * see <net/iw_handler.h> for details. 1488 * @wireless_data: Instance data managed by the core of wireless extensions 1489 * 1490 * @netdev_ops: Includes several pointers to callbacks, 1491 * if one wants to override the ndo_*() functions 1492 * @ethtool_ops: Management operations 1493 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1494 * discovery handling. Necessary for e.g. 6LoWPAN. 1495 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1496 * of Layer 2 headers. 1497 * 1498 * @flags: Interface flags (a la BSD) 1499 * @priv_flags: Like 'flags' but invisible to userspace, 1500 * see if.h for the definitions 1501 * @gflags: Global flags ( kept as legacy ) 1502 * @padded: How much padding added by alloc_netdev() 1503 * @operstate: RFC2863 operstate 1504 * @link_mode: Mapping policy to operstate 1505 * @if_port: Selectable AUI, TP, ... 1506 * @dma: DMA channel 1507 * @mtu: Interface MTU value 1508 * @min_mtu: Interface Minimum MTU value 1509 * @max_mtu: Interface Maximum MTU value 1510 * @type: Interface hardware type 1511 * @hard_header_len: Maximum hardware header length. 1512 * @min_header_len: Minimum 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 unsigned short min_header_len; 1730 1731 unsigned short needed_headroom; 1732 unsigned short needed_tailroom; 1733 1734 /* Interface address info. */ 1735 unsigned char perm_addr[MAX_ADDR_LEN]; 1736 unsigned char addr_assign_type; 1737 unsigned char addr_len; 1738 unsigned short neigh_priv_len; 1739 unsigned short dev_id; 1740 unsigned short dev_port; 1741 spinlock_t addr_list_lock; 1742 unsigned char name_assign_type; 1743 bool uc_promisc; 1744 struct netdev_hw_addr_list uc; 1745 struct netdev_hw_addr_list mc; 1746 struct netdev_hw_addr_list dev_addrs; 1747 1748 #ifdef CONFIG_SYSFS 1749 struct kset *queues_kset; 1750 #endif 1751 unsigned int promiscuity; 1752 unsigned int allmulti; 1753 1754 1755 /* Protocol-specific pointers */ 1756 1757 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1758 struct vlan_info __rcu *vlan_info; 1759 #endif 1760 #if IS_ENABLED(CONFIG_NET_DSA) 1761 struct dsa_switch_tree *dsa_ptr; 1762 #endif 1763 #if IS_ENABLED(CONFIG_TIPC) 1764 struct tipc_bearer __rcu *tipc_ptr; 1765 #endif 1766 void *atalk_ptr; 1767 struct in_device __rcu *ip_ptr; 1768 struct dn_dev __rcu *dn_ptr; 1769 struct inet6_dev __rcu *ip6_ptr; 1770 void *ax25_ptr; 1771 struct wireless_dev *ieee80211_ptr; 1772 struct wpan_dev *ieee802154_ptr; 1773 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 1774 struct mpls_dev __rcu *mpls_ptr; 1775 #endif 1776 1777 /* 1778 * Cache lines mostly used on receive path (including eth_type_trans()) 1779 */ 1780 /* Interface address info used in eth_type_trans() */ 1781 unsigned char *dev_addr; 1782 1783 #ifdef CONFIG_SYSFS 1784 struct netdev_rx_queue *_rx; 1785 1786 unsigned int num_rx_queues; 1787 unsigned int real_num_rx_queues; 1788 #endif 1789 1790 unsigned long gro_flush_timeout; 1791 rx_handler_func_t __rcu *rx_handler; 1792 void __rcu *rx_handler_data; 1793 1794 #ifdef CONFIG_NET_CLS_ACT 1795 struct tcf_proto __rcu *ingress_cl_list; 1796 #endif 1797 struct netdev_queue __rcu *ingress_queue; 1798 #ifdef CONFIG_NETFILTER_INGRESS 1799 struct nf_hook_entry __rcu *nf_hooks_ingress; 1800 #endif 1801 1802 unsigned char broadcast[MAX_ADDR_LEN]; 1803 #ifdef CONFIG_RFS_ACCEL 1804 struct cpu_rmap *rx_cpu_rmap; 1805 #endif 1806 struct hlist_node index_hlist; 1807 1808 /* 1809 * Cache lines mostly used on transmit path 1810 */ 1811 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1812 unsigned int num_tx_queues; 1813 unsigned int real_num_tx_queues; 1814 struct Qdisc *qdisc; 1815 #ifdef CONFIG_NET_SCHED 1816 DECLARE_HASHTABLE (qdisc_hash, 4); 1817 #endif 1818 unsigned long tx_queue_len; 1819 spinlock_t tx_global_lock; 1820 int watchdog_timeo; 1821 1822 #ifdef CONFIG_XPS 1823 struct xps_dev_maps __rcu *xps_maps; 1824 #endif 1825 #ifdef CONFIG_NET_CLS_ACT 1826 struct tcf_proto __rcu *egress_cl_list; 1827 #endif 1828 1829 /* These may be needed for future network-power-down code. */ 1830 struct timer_list watchdog_timer; 1831 1832 int __percpu *pcpu_refcnt; 1833 struct list_head todo_list; 1834 1835 struct list_head link_watch_list; 1836 1837 enum { NETREG_UNINITIALIZED=0, 1838 NETREG_REGISTERED, /* completed register_netdevice */ 1839 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1840 NETREG_UNREGISTERED, /* completed unregister todo */ 1841 NETREG_RELEASED, /* called free_netdev */ 1842 NETREG_DUMMY, /* dummy device for NAPI poll */ 1843 } reg_state:8; 1844 1845 bool dismantle; 1846 1847 enum { 1848 RTNL_LINK_INITIALIZED, 1849 RTNL_LINK_INITIALIZING, 1850 } rtnl_link_state:16; 1851 1852 void (*destructor)(struct net_device *dev); 1853 1854 #ifdef CONFIG_NETPOLL 1855 struct netpoll_info __rcu *npinfo; 1856 #endif 1857 1858 possible_net_t nd_net; 1859 1860 /* mid-layer private */ 1861 union { 1862 void *ml_priv; 1863 struct pcpu_lstats __percpu *lstats; 1864 struct pcpu_sw_netstats __percpu *tstats; 1865 struct pcpu_dstats __percpu *dstats; 1866 struct pcpu_vstats __percpu *vstats; 1867 }; 1868 1869 #if IS_ENABLED(CONFIG_GARP) 1870 struct garp_port __rcu *garp_port; 1871 #endif 1872 #if IS_ENABLED(CONFIG_MRP) 1873 struct mrp_port __rcu *mrp_port; 1874 #endif 1875 1876 struct device dev; 1877 const struct attribute_group *sysfs_groups[4]; 1878 const struct attribute_group *sysfs_rx_queue_group; 1879 1880 const struct rtnl_link_ops *rtnl_link_ops; 1881 1882 /* for setting kernel sock attribute on TCP connection setup */ 1883 #define GSO_MAX_SIZE 65536 1884 unsigned int gso_max_size; 1885 #define GSO_MAX_SEGS 65535 1886 u16 gso_max_segs; 1887 1888 #ifdef CONFIG_DCB 1889 const struct dcbnl_rtnl_ops *dcbnl_ops; 1890 #endif 1891 u8 num_tc; 1892 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 1893 u8 prio_tc_map[TC_BITMASK + 1]; 1894 1895 #if IS_ENABLED(CONFIG_FCOE) 1896 unsigned int fcoe_ddp_xid; 1897 #endif 1898 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 1899 struct netprio_map __rcu *priomap; 1900 #endif 1901 struct phy_device *phydev; 1902 struct lock_class_key *qdisc_tx_busylock; 1903 struct lock_class_key *qdisc_running_key; 1904 bool proto_down; 1905 }; 1906 #define to_net_dev(d) container_of(d, struct net_device, dev) 1907 1908 #define NETDEV_ALIGN 32 1909 1910 static inline 1911 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 1912 { 1913 return dev->prio_tc_map[prio & TC_BITMASK]; 1914 } 1915 1916 static inline 1917 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 1918 { 1919 if (tc >= dev->num_tc) 1920 return -EINVAL; 1921 1922 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 1923 return 0; 1924 } 1925 1926 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 1927 void netdev_reset_tc(struct net_device *dev); 1928 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 1929 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 1930 1931 static inline 1932 int netdev_get_num_tc(struct net_device *dev) 1933 { 1934 return dev->num_tc; 1935 } 1936 1937 static inline 1938 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 1939 unsigned int index) 1940 { 1941 return &dev->_tx[index]; 1942 } 1943 1944 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 1945 const struct sk_buff *skb) 1946 { 1947 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 1948 } 1949 1950 static inline void netdev_for_each_tx_queue(struct net_device *dev, 1951 void (*f)(struct net_device *, 1952 struct netdev_queue *, 1953 void *), 1954 void *arg) 1955 { 1956 unsigned int i; 1957 1958 for (i = 0; i < dev->num_tx_queues; i++) 1959 f(dev, &dev->_tx[i], arg); 1960 } 1961 1962 #define netdev_lockdep_set_classes(dev) \ 1963 { \ 1964 static struct lock_class_key qdisc_tx_busylock_key; \ 1965 static struct lock_class_key qdisc_running_key; \ 1966 static struct lock_class_key qdisc_xmit_lock_key; \ 1967 static struct lock_class_key dev_addr_list_lock_key; \ 1968 unsigned int i; \ 1969 \ 1970 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 1971 (dev)->qdisc_running_key = &qdisc_running_key; \ 1972 lockdep_set_class(&(dev)->addr_list_lock, \ 1973 &dev_addr_list_lock_key); \ 1974 for (i = 0; i < (dev)->num_tx_queues; i++) \ 1975 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 1976 &qdisc_xmit_lock_key); \ 1977 } 1978 1979 struct netdev_queue *netdev_pick_tx(struct net_device *dev, 1980 struct sk_buff *skb, 1981 void *accel_priv); 1982 1983 /* returns the headroom that the master device needs to take in account 1984 * when forwarding to this dev 1985 */ 1986 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 1987 { 1988 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 1989 } 1990 1991 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 1992 { 1993 if (dev->netdev_ops->ndo_set_rx_headroom) 1994 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 1995 } 1996 1997 /* set the device rx headroom to the dev's default */ 1998 static inline void netdev_reset_rx_headroom(struct net_device *dev) 1999 { 2000 netdev_set_rx_headroom(dev, -1); 2001 } 2002 2003 /* 2004 * Net namespace inlines 2005 */ 2006 static inline 2007 struct net *dev_net(const struct net_device *dev) 2008 { 2009 return read_pnet(&dev->nd_net); 2010 } 2011 2012 static inline 2013 void dev_net_set(struct net_device *dev, struct net *net) 2014 { 2015 write_pnet(&dev->nd_net, net); 2016 } 2017 2018 static inline bool netdev_uses_dsa(struct net_device *dev) 2019 { 2020 #if IS_ENABLED(CONFIG_NET_DSA) 2021 if (dev->dsa_ptr != NULL) 2022 return dsa_uses_tagged_protocol(dev->dsa_ptr); 2023 #endif 2024 return false; 2025 } 2026 2027 /** 2028 * netdev_priv - access network device private data 2029 * @dev: network device 2030 * 2031 * Get network device private data 2032 */ 2033 static inline void *netdev_priv(const struct net_device *dev) 2034 { 2035 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2036 } 2037 2038 /* Set the sysfs physical device reference for the network logical device 2039 * if set prior to registration will cause a symlink during initialization. 2040 */ 2041 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2042 2043 /* Set the sysfs device type for the network logical device to allow 2044 * fine-grained identification of different network device types. For 2045 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2046 */ 2047 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2048 2049 /* Default NAPI poll() weight 2050 * Device drivers are strongly advised to not use bigger value 2051 */ 2052 #define NAPI_POLL_WEIGHT 64 2053 2054 /** 2055 * netif_napi_add - initialize a NAPI context 2056 * @dev: network device 2057 * @napi: NAPI context 2058 * @poll: polling function 2059 * @weight: default weight 2060 * 2061 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2062 * *any* of the other NAPI-related functions. 2063 */ 2064 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2065 int (*poll)(struct napi_struct *, int), int weight); 2066 2067 /** 2068 * netif_tx_napi_add - initialize a NAPI context 2069 * @dev: network device 2070 * @napi: NAPI context 2071 * @poll: polling function 2072 * @weight: default weight 2073 * 2074 * This variant of netif_napi_add() should be used from drivers using NAPI 2075 * to exclusively poll a TX queue. 2076 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2077 */ 2078 static inline void netif_tx_napi_add(struct net_device *dev, 2079 struct napi_struct *napi, 2080 int (*poll)(struct napi_struct *, int), 2081 int weight) 2082 { 2083 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2084 netif_napi_add(dev, napi, poll, weight); 2085 } 2086 2087 /** 2088 * netif_napi_del - remove a NAPI context 2089 * @napi: NAPI context 2090 * 2091 * netif_napi_del() removes a NAPI context from the network device NAPI list 2092 */ 2093 void netif_napi_del(struct napi_struct *napi); 2094 2095 struct napi_gro_cb { 2096 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2097 void *frag0; 2098 2099 /* Length of frag0. */ 2100 unsigned int frag0_len; 2101 2102 /* This indicates where we are processing relative to skb->data. */ 2103 int data_offset; 2104 2105 /* This is non-zero if the packet cannot be merged with the new skb. */ 2106 u16 flush; 2107 2108 /* Save the IP ID here and check when we get to the transport layer */ 2109 u16 flush_id; 2110 2111 /* Number of segments aggregated. */ 2112 u16 count; 2113 2114 /* Start offset for remote checksum offload */ 2115 u16 gro_remcsum_start; 2116 2117 /* jiffies when first packet was created/queued */ 2118 unsigned long age; 2119 2120 /* Used in ipv6_gro_receive() and foo-over-udp */ 2121 u16 proto; 2122 2123 /* This is non-zero if the packet may be of the same flow. */ 2124 u8 same_flow:1; 2125 2126 /* Used in tunnel GRO receive */ 2127 u8 encap_mark:1; 2128 2129 /* GRO checksum is valid */ 2130 u8 csum_valid:1; 2131 2132 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2133 u8 csum_cnt:3; 2134 2135 /* Free the skb? */ 2136 u8 free:2; 2137 #define NAPI_GRO_FREE 1 2138 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2139 2140 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2141 u8 is_ipv6:1; 2142 2143 /* Used in GRE, set in fou/gue_gro_receive */ 2144 u8 is_fou:1; 2145 2146 /* Used to determine if flush_id can be ignored */ 2147 u8 is_atomic:1; 2148 2149 /* Number of gro_receive callbacks this packet already went through */ 2150 u8 recursion_counter:4; 2151 2152 /* 1 bit hole */ 2153 2154 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2155 __wsum csum; 2156 2157 /* used in skb_gro_receive() slow path */ 2158 struct sk_buff *last; 2159 }; 2160 2161 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2162 2163 #define GRO_RECURSION_LIMIT 15 2164 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2165 { 2166 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2167 } 2168 2169 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *); 2170 static inline struct sk_buff **call_gro_receive(gro_receive_t cb, 2171 struct sk_buff **head, 2172 struct sk_buff *skb) 2173 { 2174 if (unlikely(gro_recursion_inc_test(skb))) { 2175 NAPI_GRO_CB(skb)->flush |= 1; 2176 return NULL; 2177 } 2178 2179 return cb(head, skb); 2180 } 2181 2182 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **, 2183 struct sk_buff *); 2184 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb, 2185 struct sock *sk, 2186 struct sk_buff **head, 2187 struct sk_buff *skb) 2188 { 2189 if (unlikely(gro_recursion_inc_test(skb))) { 2190 NAPI_GRO_CB(skb)->flush |= 1; 2191 return NULL; 2192 } 2193 2194 return cb(sk, head, skb); 2195 } 2196 2197 struct packet_type { 2198 __be16 type; /* This is really htons(ether_type). */ 2199 struct net_device *dev; /* NULL is wildcarded here */ 2200 int (*func) (struct sk_buff *, 2201 struct net_device *, 2202 struct packet_type *, 2203 struct net_device *); 2204 bool (*id_match)(struct packet_type *ptype, 2205 struct sock *sk); 2206 void *af_packet_priv; 2207 struct list_head list; 2208 }; 2209 2210 struct offload_callbacks { 2211 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2212 netdev_features_t features); 2213 struct sk_buff **(*gro_receive)(struct sk_buff **head, 2214 struct sk_buff *skb); 2215 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2216 }; 2217 2218 struct packet_offload { 2219 __be16 type; /* This is really htons(ether_type). */ 2220 u16 priority; 2221 struct offload_callbacks callbacks; 2222 struct list_head list; 2223 }; 2224 2225 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2226 struct pcpu_sw_netstats { 2227 u64 rx_packets; 2228 u64 rx_bytes; 2229 u64 tx_packets; 2230 u64 tx_bytes; 2231 struct u64_stats_sync syncp; 2232 }; 2233 2234 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2235 ({ \ 2236 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2237 if (pcpu_stats) { \ 2238 int __cpu; \ 2239 for_each_possible_cpu(__cpu) { \ 2240 typeof(type) *stat; \ 2241 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2242 u64_stats_init(&stat->syncp); \ 2243 } \ 2244 } \ 2245 pcpu_stats; \ 2246 }) 2247 2248 #define netdev_alloc_pcpu_stats(type) \ 2249 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2250 2251 enum netdev_lag_tx_type { 2252 NETDEV_LAG_TX_TYPE_UNKNOWN, 2253 NETDEV_LAG_TX_TYPE_RANDOM, 2254 NETDEV_LAG_TX_TYPE_BROADCAST, 2255 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2256 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2257 NETDEV_LAG_TX_TYPE_HASH, 2258 }; 2259 2260 struct netdev_lag_upper_info { 2261 enum netdev_lag_tx_type tx_type; 2262 }; 2263 2264 struct netdev_lag_lower_state_info { 2265 u8 link_up : 1, 2266 tx_enabled : 1; 2267 }; 2268 2269 #include <linux/notifier.h> 2270 2271 /* netdevice notifier chain. Please remember to update the rtnetlink 2272 * notification exclusion list in rtnetlink_event() when adding new 2273 * types. 2274 */ 2275 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */ 2276 #define NETDEV_DOWN 0x0002 2277 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface 2278 detected a hardware crash and restarted 2279 - we can use this eg to kick tcp sessions 2280 once done */ 2281 #define NETDEV_CHANGE 0x0004 /* Notify device state change */ 2282 #define NETDEV_REGISTER 0x0005 2283 #define NETDEV_UNREGISTER 0x0006 2284 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */ 2285 #define NETDEV_CHANGEADDR 0x0008 2286 #define NETDEV_GOING_DOWN 0x0009 2287 #define NETDEV_CHANGENAME 0x000A 2288 #define NETDEV_FEAT_CHANGE 0x000B 2289 #define NETDEV_BONDING_FAILOVER 0x000C 2290 #define NETDEV_PRE_UP 0x000D 2291 #define NETDEV_PRE_TYPE_CHANGE 0x000E 2292 #define NETDEV_POST_TYPE_CHANGE 0x000F 2293 #define NETDEV_POST_INIT 0x0010 2294 #define NETDEV_UNREGISTER_FINAL 0x0011 2295 #define NETDEV_RELEASE 0x0012 2296 #define NETDEV_NOTIFY_PEERS 0x0013 2297 #define NETDEV_JOIN 0x0014 2298 #define NETDEV_CHANGEUPPER 0x0015 2299 #define NETDEV_RESEND_IGMP 0x0016 2300 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */ 2301 #define NETDEV_CHANGEINFODATA 0x0018 2302 #define NETDEV_BONDING_INFO 0x0019 2303 #define NETDEV_PRECHANGEUPPER 0x001A 2304 #define NETDEV_CHANGELOWERSTATE 0x001B 2305 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C 2306 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E 2307 2308 int register_netdevice_notifier(struct notifier_block *nb); 2309 int unregister_netdevice_notifier(struct notifier_block *nb); 2310 2311 struct netdev_notifier_info { 2312 struct net_device *dev; 2313 }; 2314 2315 struct netdev_notifier_change_info { 2316 struct netdev_notifier_info info; /* must be first */ 2317 unsigned int flags_changed; 2318 }; 2319 2320 struct netdev_notifier_changeupper_info { 2321 struct netdev_notifier_info info; /* must be first */ 2322 struct net_device *upper_dev; /* new upper dev */ 2323 bool master; /* is upper dev master */ 2324 bool linking; /* is the notification for link or unlink */ 2325 void *upper_info; /* upper dev info */ 2326 }; 2327 2328 struct netdev_notifier_changelowerstate_info { 2329 struct netdev_notifier_info info; /* must be first */ 2330 void *lower_state_info; /* is lower dev state */ 2331 }; 2332 2333 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2334 struct net_device *dev) 2335 { 2336 info->dev = dev; 2337 } 2338 2339 static inline struct net_device * 2340 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2341 { 2342 return info->dev; 2343 } 2344 2345 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2346 2347 2348 extern rwlock_t dev_base_lock; /* Device list lock */ 2349 2350 #define for_each_netdev(net, d) \ 2351 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2352 #define for_each_netdev_reverse(net, d) \ 2353 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2354 #define for_each_netdev_rcu(net, d) \ 2355 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2356 #define for_each_netdev_safe(net, d, n) \ 2357 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2358 #define for_each_netdev_continue(net, d) \ 2359 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2360 #define for_each_netdev_continue_rcu(net, d) \ 2361 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2362 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2363 for_each_netdev_rcu(&init_net, slave) \ 2364 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2365 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2366 2367 static inline struct net_device *next_net_device(struct net_device *dev) 2368 { 2369 struct list_head *lh; 2370 struct net *net; 2371 2372 net = dev_net(dev); 2373 lh = dev->dev_list.next; 2374 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2375 } 2376 2377 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2378 { 2379 struct list_head *lh; 2380 struct net *net; 2381 2382 net = dev_net(dev); 2383 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2384 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2385 } 2386 2387 static inline struct net_device *first_net_device(struct net *net) 2388 { 2389 return list_empty(&net->dev_base_head) ? NULL : 2390 net_device_entry(net->dev_base_head.next); 2391 } 2392 2393 static inline struct net_device *first_net_device_rcu(struct net *net) 2394 { 2395 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2396 2397 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2398 } 2399 2400 int netdev_boot_setup_check(struct net_device *dev); 2401 unsigned long netdev_boot_base(const char *prefix, int unit); 2402 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2403 const char *hwaddr); 2404 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2405 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2406 void dev_add_pack(struct packet_type *pt); 2407 void dev_remove_pack(struct packet_type *pt); 2408 void __dev_remove_pack(struct packet_type *pt); 2409 void dev_add_offload(struct packet_offload *po); 2410 void dev_remove_offload(struct packet_offload *po); 2411 2412 int dev_get_iflink(const struct net_device *dev); 2413 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2414 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2415 unsigned short mask); 2416 struct net_device *dev_get_by_name(struct net *net, const char *name); 2417 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2418 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2419 int dev_alloc_name(struct net_device *dev, const char *name); 2420 int dev_open(struct net_device *dev); 2421 int dev_close(struct net_device *dev); 2422 int dev_close_many(struct list_head *head, bool unlink); 2423 void dev_disable_lro(struct net_device *dev); 2424 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2425 int dev_queue_xmit(struct sk_buff *skb); 2426 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv); 2427 int register_netdevice(struct net_device *dev); 2428 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2429 void unregister_netdevice_many(struct list_head *head); 2430 static inline void unregister_netdevice(struct net_device *dev) 2431 { 2432 unregister_netdevice_queue(dev, NULL); 2433 } 2434 2435 int netdev_refcnt_read(const struct net_device *dev); 2436 void free_netdev(struct net_device *dev); 2437 void netdev_freemem(struct net_device *dev); 2438 void synchronize_net(void); 2439 int init_dummy_netdev(struct net_device *dev); 2440 2441 DECLARE_PER_CPU(int, xmit_recursion); 2442 #define XMIT_RECURSION_LIMIT 10 2443 2444 static inline int dev_recursion_level(void) 2445 { 2446 return this_cpu_read(xmit_recursion); 2447 } 2448 2449 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2450 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2451 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2452 int netdev_get_name(struct net *net, char *name, int ifindex); 2453 int dev_restart(struct net_device *dev); 2454 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb); 2455 2456 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2457 { 2458 return NAPI_GRO_CB(skb)->data_offset; 2459 } 2460 2461 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2462 { 2463 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2464 } 2465 2466 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2467 { 2468 NAPI_GRO_CB(skb)->data_offset += len; 2469 } 2470 2471 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2472 unsigned int offset) 2473 { 2474 return NAPI_GRO_CB(skb)->frag0 + offset; 2475 } 2476 2477 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2478 { 2479 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2480 } 2481 2482 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2483 { 2484 NAPI_GRO_CB(skb)->frag0 = NULL; 2485 NAPI_GRO_CB(skb)->frag0_len = 0; 2486 } 2487 2488 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2489 unsigned int offset) 2490 { 2491 if (!pskb_may_pull(skb, hlen)) 2492 return NULL; 2493 2494 skb_gro_frag0_invalidate(skb); 2495 return skb->data + offset; 2496 } 2497 2498 static inline void *skb_gro_network_header(struct sk_buff *skb) 2499 { 2500 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2501 skb_network_offset(skb); 2502 } 2503 2504 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2505 const void *start, unsigned int len) 2506 { 2507 if (NAPI_GRO_CB(skb)->csum_valid) 2508 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2509 csum_partial(start, len, 0)); 2510 } 2511 2512 /* GRO checksum functions. These are logical equivalents of the normal 2513 * checksum functions (in skbuff.h) except that they operate on the GRO 2514 * offsets and fields in sk_buff. 2515 */ 2516 2517 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2518 2519 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2520 { 2521 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2522 } 2523 2524 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2525 bool zero_okay, 2526 __sum16 check) 2527 { 2528 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2529 skb_checksum_start_offset(skb) < 2530 skb_gro_offset(skb)) && 2531 !skb_at_gro_remcsum_start(skb) && 2532 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2533 (!zero_okay || check)); 2534 } 2535 2536 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2537 __wsum psum) 2538 { 2539 if (NAPI_GRO_CB(skb)->csum_valid && 2540 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2541 return 0; 2542 2543 NAPI_GRO_CB(skb)->csum = psum; 2544 2545 return __skb_gro_checksum_complete(skb); 2546 } 2547 2548 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2549 { 2550 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2551 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2552 NAPI_GRO_CB(skb)->csum_cnt--; 2553 } else { 2554 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2555 * verified a new top level checksum or an encapsulated one 2556 * during GRO. This saves work if we fallback to normal path. 2557 */ 2558 __skb_incr_checksum_unnecessary(skb); 2559 } 2560 } 2561 2562 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2563 compute_pseudo) \ 2564 ({ \ 2565 __sum16 __ret = 0; \ 2566 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2567 __ret = __skb_gro_checksum_validate_complete(skb, \ 2568 compute_pseudo(skb, proto)); \ 2569 if (__ret) \ 2570 __skb_mark_checksum_bad(skb); \ 2571 else \ 2572 skb_gro_incr_csum_unnecessary(skb); \ 2573 __ret; \ 2574 }) 2575 2576 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2577 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2578 2579 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2580 compute_pseudo) \ 2581 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2582 2583 #define skb_gro_checksum_simple_validate(skb) \ 2584 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2585 2586 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2587 { 2588 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2589 !NAPI_GRO_CB(skb)->csum_valid); 2590 } 2591 2592 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2593 __sum16 check, __wsum pseudo) 2594 { 2595 NAPI_GRO_CB(skb)->csum = ~pseudo; 2596 NAPI_GRO_CB(skb)->csum_valid = 1; 2597 } 2598 2599 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \ 2600 do { \ 2601 if (__skb_gro_checksum_convert_check(skb)) \ 2602 __skb_gro_checksum_convert(skb, check, \ 2603 compute_pseudo(skb, proto)); \ 2604 } while (0) 2605 2606 struct gro_remcsum { 2607 int offset; 2608 __wsum delta; 2609 }; 2610 2611 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2612 { 2613 grc->offset = 0; 2614 grc->delta = 0; 2615 } 2616 2617 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2618 unsigned int off, size_t hdrlen, 2619 int start, int offset, 2620 struct gro_remcsum *grc, 2621 bool nopartial) 2622 { 2623 __wsum delta; 2624 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2625 2626 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2627 2628 if (!nopartial) { 2629 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2630 return ptr; 2631 } 2632 2633 ptr = skb_gro_header_fast(skb, off); 2634 if (skb_gro_header_hard(skb, off + plen)) { 2635 ptr = skb_gro_header_slow(skb, off + plen, off); 2636 if (!ptr) 2637 return NULL; 2638 } 2639 2640 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2641 start, offset); 2642 2643 /* Adjust skb->csum since we changed the packet */ 2644 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2645 2646 grc->offset = off + hdrlen + offset; 2647 grc->delta = delta; 2648 2649 return ptr; 2650 } 2651 2652 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2653 struct gro_remcsum *grc) 2654 { 2655 void *ptr; 2656 size_t plen = grc->offset + sizeof(u16); 2657 2658 if (!grc->delta) 2659 return; 2660 2661 ptr = skb_gro_header_fast(skb, grc->offset); 2662 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 2663 ptr = skb_gro_header_slow(skb, plen, grc->offset); 2664 if (!ptr) 2665 return; 2666 } 2667 2668 remcsum_unadjust((__sum16 *)ptr, grc->delta); 2669 } 2670 2671 #ifdef CONFIG_XFRM_OFFLOAD 2672 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush) 2673 { 2674 if (PTR_ERR(pp) != -EINPROGRESS) 2675 NAPI_GRO_CB(skb)->flush |= flush; 2676 } 2677 #else 2678 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush) 2679 { 2680 NAPI_GRO_CB(skb)->flush |= flush; 2681 } 2682 #endif 2683 2684 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2685 unsigned short type, 2686 const void *daddr, const void *saddr, 2687 unsigned int len) 2688 { 2689 if (!dev->header_ops || !dev->header_ops->create) 2690 return 0; 2691 2692 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2693 } 2694 2695 static inline int dev_parse_header(const struct sk_buff *skb, 2696 unsigned char *haddr) 2697 { 2698 const struct net_device *dev = skb->dev; 2699 2700 if (!dev->header_ops || !dev->header_ops->parse) 2701 return 0; 2702 return dev->header_ops->parse(skb, haddr); 2703 } 2704 2705 /* ll_header must have at least hard_header_len allocated */ 2706 static inline bool dev_validate_header(const struct net_device *dev, 2707 char *ll_header, int len) 2708 { 2709 if (likely(len >= dev->hard_header_len)) 2710 return true; 2711 if (len < dev->min_header_len) 2712 return false; 2713 2714 if (capable(CAP_SYS_RAWIO)) { 2715 memset(ll_header + len, 0, dev->hard_header_len - len); 2716 return true; 2717 } 2718 2719 if (dev->header_ops && dev->header_ops->validate) 2720 return dev->header_ops->validate(ll_header, len); 2721 2722 return false; 2723 } 2724 2725 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len); 2726 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2727 static inline int unregister_gifconf(unsigned int family) 2728 { 2729 return register_gifconf(family, NULL); 2730 } 2731 2732 #ifdef CONFIG_NET_FLOW_LIMIT 2733 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2734 struct sd_flow_limit { 2735 u64 count; 2736 unsigned int num_buckets; 2737 unsigned int history_head; 2738 u16 history[FLOW_LIMIT_HISTORY]; 2739 u8 buckets[]; 2740 }; 2741 2742 extern int netdev_flow_limit_table_len; 2743 #endif /* CONFIG_NET_FLOW_LIMIT */ 2744 2745 /* 2746 * Incoming packets are placed on per-CPU queues 2747 */ 2748 struct softnet_data { 2749 struct list_head poll_list; 2750 struct sk_buff_head process_queue; 2751 2752 /* stats */ 2753 unsigned int processed; 2754 unsigned int time_squeeze; 2755 unsigned int received_rps; 2756 #ifdef CONFIG_RPS 2757 struct softnet_data *rps_ipi_list; 2758 #endif 2759 #ifdef CONFIG_NET_FLOW_LIMIT 2760 struct sd_flow_limit __rcu *flow_limit; 2761 #endif 2762 struct Qdisc *output_queue; 2763 struct Qdisc **output_queue_tailp; 2764 struct sk_buff *completion_queue; 2765 2766 #ifdef CONFIG_RPS 2767 /* input_queue_head should be written by cpu owning this struct, 2768 * and only read by other cpus. Worth using a cache line. 2769 */ 2770 unsigned int input_queue_head ____cacheline_aligned_in_smp; 2771 2772 /* Elements below can be accessed between CPUs for RPS/RFS */ 2773 struct call_single_data csd ____cacheline_aligned_in_smp; 2774 struct softnet_data *rps_ipi_next; 2775 unsigned int cpu; 2776 unsigned int input_queue_tail; 2777 #endif 2778 unsigned int dropped; 2779 struct sk_buff_head input_pkt_queue; 2780 struct napi_struct backlog; 2781 2782 }; 2783 2784 static inline void input_queue_head_incr(struct softnet_data *sd) 2785 { 2786 #ifdef CONFIG_RPS 2787 sd->input_queue_head++; 2788 #endif 2789 } 2790 2791 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 2792 unsigned int *qtail) 2793 { 2794 #ifdef CONFIG_RPS 2795 *qtail = ++sd->input_queue_tail; 2796 #endif 2797 } 2798 2799 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 2800 2801 void __netif_schedule(struct Qdisc *q); 2802 void netif_schedule_queue(struct netdev_queue *txq); 2803 2804 static inline void netif_tx_schedule_all(struct net_device *dev) 2805 { 2806 unsigned int i; 2807 2808 for (i = 0; i < dev->num_tx_queues; i++) 2809 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 2810 } 2811 2812 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 2813 { 2814 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2815 } 2816 2817 /** 2818 * netif_start_queue - allow transmit 2819 * @dev: network device 2820 * 2821 * Allow upper layers to call the device hard_start_xmit routine. 2822 */ 2823 static inline void netif_start_queue(struct net_device *dev) 2824 { 2825 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 2826 } 2827 2828 static inline void netif_tx_start_all_queues(struct net_device *dev) 2829 { 2830 unsigned int i; 2831 2832 for (i = 0; i < dev->num_tx_queues; i++) { 2833 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2834 netif_tx_start_queue(txq); 2835 } 2836 } 2837 2838 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 2839 2840 /** 2841 * netif_wake_queue - restart transmit 2842 * @dev: network device 2843 * 2844 * Allow upper layers to call the device hard_start_xmit routine. 2845 * Used for flow control when transmit resources are available. 2846 */ 2847 static inline void netif_wake_queue(struct net_device *dev) 2848 { 2849 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 2850 } 2851 2852 static inline void netif_tx_wake_all_queues(struct net_device *dev) 2853 { 2854 unsigned int i; 2855 2856 for (i = 0; i < dev->num_tx_queues; i++) { 2857 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2858 netif_tx_wake_queue(txq); 2859 } 2860 } 2861 2862 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 2863 { 2864 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2865 } 2866 2867 /** 2868 * netif_stop_queue - stop transmitted packets 2869 * @dev: network device 2870 * 2871 * Stop upper layers calling the device hard_start_xmit routine. 2872 * Used for flow control when transmit resources are unavailable. 2873 */ 2874 static inline void netif_stop_queue(struct net_device *dev) 2875 { 2876 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 2877 } 2878 2879 void netif_tx_stop_all_queues(struct net_device *dev); 2880 2881 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 2882 { 2883 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2884 } 2885 2886 /** 2887 * netif_queue_stopped - test if transmit queue is flowblocked 2888 * @dev: network device 2889 * 2890 * Test if transmit queue on device is currently unable to send. 2891 */ 2892 static inline bool netif_queue_stopped(const struct net_device *dev) 2893 { 2894 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 2895 } 2896 2897 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 2898 { 2899 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 2900 } 2901 2902 static inline bool 2903 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 2904 { 2905 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 2906 } 2907 2908 static inline bool 2909 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 2910 { 2911 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 2912 } 2913 2914 /** 2915 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 2916 * @dev_queue: pointer to transmit queue 2917 * 2918 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 2919 * to give appropriate hint to the CPU. 2920 */ 2921 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 2922 { 2923 #ifdef CONFIG_BQL 2924 prefetchw(&dev_queue->dql.num_queued); 2925 #endif 2926 } 2927 2928 /** 2929 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 2930 * @dev_queue: pointer to transmit queue 2931 * 2932 * BQL enabled drivers might use this helper in their TX completion path, 2933 * to give appropriate hint to the CPU. 2934 */ 2935 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 2936 { 2937 #ifdef CONFIG_BQL 2938 prefetchw(&dev_queue->dql.limit); 2939 #endif 2940 } 2941 2942 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 2943 unsigned int bytes) 2944 { 2945 #ifdef CONFIG_BQL 2946 dql_queued(&dev_queue->dql, bytes); 2947 2948 if (likely(dql_avail(&dev_queue->dql) >= 0)) 2949 return; 2950 2951 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2952 2953 /* 2954 * The XOFF flag must be set before checking the dql_avail below, 2955 * because in netdev_tx_completed_queue we update the dql_completed 2956 * before checking the XOFF flag. 2957 */ 2958 smp_mb(); 2959 2960 /* check again in case another CPU has just made room avail */ 2961 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 2962 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2963 #endif 2964 } 2965 2966 /** 2967 * netdev_sent_queue - report the number of bytes queued to hardware 2968 * @dev: network device 2969 * @bytes: number of bytes queued to the hardware device queue 2970 * 2971 * Report the number of bytes queued for sending/completion to the network 2972 * device hardware queue. @bytes should be a good approximation and should 2973 * exactly match netdev_completed_queue() @bytes 2974 */ 2975 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 2976 { 2977 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 2978 } 2979 2980 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 2981 unsigned int pkts, unsigned int bytes) 2982 { 2983 #ifdef CONFIG_BQL 2984 if (unlikely(!bytes)) 2985 return; 2986 2987 dql_completed(&dev_queue->dql, bytes); 2988 2989 /* 2990 * Without the memory barrier there is a small possiblity that 2991 * netdev_tx_sent_queue will miss the update and cause the queue to 2992 * be stopped forever 2993 */ 2994 smp_mb(); 2995 2996 if (dql_avail(&dev_queue->dql) < 0) 2997 return; 2998 2999 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3000 netif_schedule_queue(dev_queue); 3001 #endif 3002 } 3003 3004 /** 3005 * netdev_completed_queue - report bytes and packets completed by device 3006 * @dev: network device 3007 * @pkts: actual number of packets sent over the medium 3008 * @bytes: actual number of bytes sent over the medium 3009 * 3010 * Report the number of bytes and packets transmitted by the network device 3011 * hardware queue over the physical medium, @bytes must exactly match the 3012 * @bytes amount passed to netdev_sent_queue() 3013 */ 3014 static inline void netdev_completed_queue(struct net_device *dev, 3015 unsigned int pkts, unsigned int bytes) 3016 { 3017 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3018 } 3019 3020 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3021 { 3022 #ifdef CONFIG_BQL 3023 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3024 dql_reset(&q->dql); 3025 #endif 3026 } 3027 3028 /** 3029 * netdev_reset_queue - reset the packets and bytes count of a network device 3030 * @dev_queue: network device 3031 * 3032 * Reset the bytes and packet count of a network device and clear the 3033 * software flow control OFF bit for this network device 3034 */ 3035 static inline void netdev_reset_queue(struct net_device *dev_queue) 3036 { 3037 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3038 } 3039 3040 /** 3041 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3042 * @dev: network device 3043 * @queue_index: given tx queue index 3044 * 3045 * Returns 0 if given tx queue index >= number of device tx queues, 3046 * otherwise returns the originally passed tx queue index. 3047 */ 3048 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3049 { 3050 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3051 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3052 dev->name, queue_index, 3053 dev->real_num_tx_queues); 3054 return 0; 3055 } 3056 3057 return queue_index; 3058 } 3059 3060 /** 3061 * netif_running - test if up 3062 * @dev: network device 3063 * 3064 * Test if the device has been brought up. 3065 */ 3066 static inline bool netif_running(const struct net_device *dev) 3067 { 3068 return test_bit(__LINK_STATE_START, &dev->state); 3069 } 3070 3071 /* 3072 * Routines to manage the subqueues on a device. We only need start, 3073 * stop, and a check if it's stopped. All other device management is 3074 * done at the overall netdevice level. 3075 * Also test the device if we're multiqueue. 3076 */ 3077 3078 /** 3079 * netif_start_subqueue - allow sending packets on subqueue 3080 * @dev: network device 3081 * @queue_index: sub queue index 3082 * 3083 * Start individual transmit queue of a device with multiple transmit queues. 3084 */ 3085 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3086 { 3087 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3088 3089 netif_tx_start_queue(txq); 3090 } 3091 3092 /** 3093 * netif_stop_subqueue - stop sending packets on subqueue 3094 * @dev: network device 3095 * @queue_index: sub queue index 3096 * 3097 * Stop individual transmit queue of a device with multiple transmit queues. 3098 */ 3099 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3100 { 3101 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3102 netif_tx_stop_queue(txq); 3103 } 3104 3105 /** 3106 * netif_subqueue_stopped - test status of subqueue 3107 * @dev: network device 3108 * @queue_index: sub queue index 3109 * 3110 * Check individual transmit queue of a device with multiple transmit queues. 3111 */ 3112 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3113 u16 queue_index) 3114 { 3115 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3116 3117 return netif_tx_queue_stopped(txq); 3118 } 3119 3120 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3121 struct sk_buff *skb) 3122 { 3123 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3124 } 3125 3126 /** 3127 * netif_wake_subqueue - allow sending packets on subqueue 3128 * @dev: network device 3129 * @queue_index: sub queue index 3130 * 3131 * Resume individual transmit queue of a device with multiple transmit queues. 3132 */ 3133 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3134 { 3135 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3136 3137 netif_tx_wake_queue(txq); 3138 } 3139 3140 #ifdef CONFIG_XPS 3141 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3142 u16 index); 3143 #else 3144 static inline int netif_set_xps_queue(struct net_device *dev, 3145 const struct cpumask *mask, 3146 u16 index) 3147 { 3148 return 0; 3149 } 3150 #endif 3151 3152 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb, 3153 unsigned int num_tx_queues); 3154 3155 /* 3156 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used 3157 * as a distribution range limit for the returned value. 3158 */ 3159 static inline u16 skb_tx_hash(const struct net_device *dev, 3160 struct sk_buff *skb) 3161 { 3162 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues); 3163 } 3164 3165 /** 3166 * netif_is_multiqueue - test if device has multiple transmit queues 3167 * @dev: network device 3168 * 3169 * Check if device has multiple transmit queues 3170 */ 3171 static inline bool netif_is_multiqueue(const struct net_device *dev) 3172 { 3173 return dev->num_tx_queues > 1; 3174 } 3175 3176 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3177 3178 #ifdef CONFIG_SYSFS 3179 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3180 #else 3181 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3182 unsigned int rxq) 3183 { 3184 return 0; 3185 } 3186 #endif 3187 3188 #ifdef CONFIG_SYSFS 3189 static inline unsigned int get_netdev_rx_queue_index( 3190 struct netdev_rx_queue *queue) 3191 { 3192 struct net_device *dev = queue->dev; 3193 int index = queue - dev->_rx; 3194 3195 BUG_ON(index >= dev->num_rx_queues); 3196 return index; 3197 } 3198 #endif 3199 3200 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3201 int netif_get_num_default_rss_queues(void); 3202 3203 enum skb_free_reason { 3204 SKB_REASON_CONSUMED, 3205 SKB_REASON_DROPPED, 3206 }; 3207 3208 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3209 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3210 3211 /* 3212 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3213 * interrupt context or with hardware interrupts being disabled. 3214 * (in_irq() || irqs_disabled()) 3215 * 3216 * We provide four helpers that can be used in following contexts : 3217 * 3218 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3219 * replacing kfree_skb(skb) 3220 * 3221 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3222 * Typically used in place of consume_skb(skb) in TX completion path 3223 * 3224 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3225 * replacing kfree_skb(skb) 3226 * 3227 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3228 * and consumed a packet. Used in place of consume_skb(skb) 3229 */ 3230 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3231 { 3232 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3233 } 3234 3235 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3236 { 3237 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3238 } 3239 3240 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3241 { 3242 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3243 } 3244 3245 static inline void dev_consume_skb_any(struct sk_buff *skb) 3246 { 3247 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3248 } 3249 3250 int netif_rx(struct sk_buff *skb); 3251 int netif_rx_ni(struct sk_buff *skb); 3252 int netif_receive_skb(struct sk_buff *skb); 3253 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3254 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3255 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3256 gro_result_t napi_gro_frags(struct napi_struct *napi); 3257 struct packet_offload *gro_find_receive_by_type(__be16 type); 3258 struct packet_offload *gro_find_complete_by_type(__be16 type); 3259 3260 static inline void napi_free_frags(struct napi_struct *napi) 3261 { 3262 kfree_skb(napi->skb); 3263 napi->skb = NULL; 3264 } 3265 3266 bool netdev_is_rx_handler_busy(struct net_device *dev); 3267 int netdev_rx_handler_register(struct net_device *dev, 3268 rx_handler_func_t *rx_handler, 3269 void *rx_handler_data); 3270 void netdev_rx_handler_unregister(struct net_device *dev); 3271 3272 bool dev_valid_name(const char *name); 3273 int dev_ioctl(struct net *net, unsigned int cmd, void __user *); 3274 int dev_ethtool(struct net *net, struct ifreq *); 3275 unsigned int dev_get_flags(const struct net_device *); 3276 int __dev_change_flags(struct net_device *, unsigned int flags); 3277 int dev_change_flags(struct net_device *, unsigned int); 3278 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3279 unsigned int gchanges); 3280 int dev_change_name(struct net_device *, const char *); 3281 int dev_set_alias(struct net_device *, const char *, size_t); 3282 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3283 int dev_set_mtu(struct net_device *, int); 3284 void dev_set_group(struct net_device *, int); 3285 int dev_set_mac_address(struct net_device *, struct sockaddr *); 3286 int dev_change_carrier(struct net_device *, bool new_carrier); 3287 int dev_get_phys_port_id(struct net_device *dev, 3288 struct netdev_phys_item_id *ppid); 3289 int dev_get_phys_port_name(struct net_device *dev, 3290 char *name, size_t len); 3291 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3292 int dev_change_xdp_fd(struct net_device *dev, int fd, u32 flags); 3293 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev); 3294 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3295 struct netdev_queue *txq, int *ret); 3296 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3297 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3298 bool is_skb_forwardable(const struct net_device *dev, 3299 const struct sk_buff *skb); 3300 3301 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3302 struct sk_buff *skb) 3303 { 3304 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3305 unlikely(!is_skb_forwardable(dev, skb))) { 3306 atomic_long_inc(&dev->rx_dropped); 3307 kfree_skb(skb); 3308 return NET_RX_DROP; 3309 } 3310 3311 skb_scrub_packet(skb, true); 3312 skb->priority = 0; 3313 return 0; 3314 } 3315 3316 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3317 3318 extern int netdev_budget; 3319 3320 /* Called by rtnetlink.c:rtnl_unlock() */ 3321 void netdev_run_todo(void); 3322 3323 /** 3324 * dev_put - release reference to device 3325 * @dev: network device 3326 * 3327 * Release reference to device to allow it to be freed. 3328 */ 3329 static inline void dev_put(struct net_device *dev) 3330 { 3331 this_cpu_dec(*dev->pcpu_refcnt); 3332 } 3333 3334 /** 3335 * dev_hold - get reference to device 3336 * @dev: network device 3337 * 3338 * Hold reference to device to keep it from being freed. 3339 */ 3340 static inline void dev_hold(struct net_device *dev) 3341 { 3342 this_cpu_inc(*dev->pcpu_refcnt); 3343 } 3344 3345 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3346 * and _off may be called from IRQ context, but it is caller 3347 * who is responsible for serialization of these calls. 3348 * 3349 * The name carrier is inappropriate, these functions should really be 3350 * called netif_lowerlayer_*() because they represent the state of any 3351 * kind of lower layer not just hardware media. 3352 */ 3353 3354 void linkwatch_init_dev(struct net_device *dev); 3355 void linkwatch_fire_event(struct net_device *dev); 3356 void linkwatch_forget_dev(struct net_device *dev); 3357 3358 /** 3359 * netif_carrier_ok - test if carrier present 3360 * @dev: network device 3361 * 3362 * Check if carrier is present on device 3363 */ 3364 static inline bool netif_carrier_ok(const struct net_device *dev) 3365 { 3366 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3367 } 3368 3369 unsigned long dev_trans_start(struct net_device *dev); 3370 3371 void __netdev_watchdog_up(struct net_device *dev); 3372 3373 void netif_carrier_on(struct net_device *dev); 3374 3375 void netif_carrier_off(struct net_device *dev); 3376 3377 /** 3378 * netif_dormant_on - mark device as dormant. 3379 * @dev: network device 3380 * 3381 * Mark device as dormant (as per RFC2863). 3382 * 3383 * The dormant state indicates that the relevant interface is not 3384 * actually in a condition to pass packets (i.e., it is not 'up') but is 3385 * in a "pending" state, waiting for some external event. For "on- 3386 * demand" interfaces, this new state identifies the situation where the 3387 * interface is waiting for events to place it in the up state. 3388 */ 3389 static inline void netif_dormant_on(struct net_device *dev) 3390 { 3391 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3392 linkwatch_fire_event(dev); 3393 } 3394 3395 /** 3396 * netif_dormant_off - set device as not dormant. 3397 * @dev: network device 3398 * 3399 * Device is not in dormant state. 3400 */ 3401 static inline void netif_dormant_off(struct net_device *dev) 3402 { 3403 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3404 linkwatch_fire_event(dev); 3405 } 3406 3407 /** 3408 * netif_dormant - test if carrier present 3409 * @dev: network device 3410 * 3411 * Check if carrier is present on device 3412 */ 3413 static inline bool netif_dormant(const struct net_device *dev) 3414 { 3415 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3416 } 3417 3418 3419 /** 3420 * netif_oper_up - test if device is operational 3421 * @dev: network device 3422 * 3423 * Check if carrier is operational 3424 */ 3425 static inline bool netif_oper_up(const struct net_device *dev) 3426 { 3427 return (dev->operstate == IF_OPER_UP || 3428 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3429 } 3430 3431 /** 3432 * netif_device_present - is device available or removed 3433 * @dev: network device 3434 * 3435 * Check if device has not been removed from system. 3436 */ 3437 static inline bool netif_device_present(struct net_device *dev) 3438 { 3439 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3440 } 3441 3442 void netif_device_detach(struct net_device *dev); 3443 3444 void netif_device_attach(struct net_device *dev); 3445 3446 /* 3447 * Network interface message level settings 3448 */ 3449 3450 enum { 3451 NETIF_MSG_DRV = 0x0001, 3452 NETIF_MSG_PROBE = 0x0002, 3453 NETIF_MSG_LINK = 0x0004, 3454 NETIF_MSG_TIMER = 0x0008, 3455 NETIF_MSG_IFDOWN = 0x0010, 3456 NETIF_MSG_IFUP = 0x0020, 3457 NETIF_MSG_RX_ERR = 0x0040, 3458 NETIF_MSG_TX_ERR = 0x0080, 3459 NETIF_MSG_TX_QUEUED = 0x0100, 3460 NETIF_MSG_INTR = 0x0200, 3461 NETIF_MSG_TX_DONE = 0x0400, 3462 NETIF_MSG_RX_STATUS = 0x0800, 3463 NETIF_MSG_PKTDATA = 0x1000, 3464 NETIF_MSG_HW = 0x2000, 3465 NETIF_MSG_WOL = 0x4000, 3466 }; 3467 3468 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3469 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3470 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3471 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3472 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3473 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3474 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3475 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3476 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3477 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3478 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3479 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3480 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3481 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3482 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3483 3484 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3485 { 3486 /* use default */ 3487 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3488 return default_msg_enable_bits; 3489 if (debug_value == 0) /* no output */ 3490 return 0; 3491 /* set low N bits */ 3492 return (1 << debug_value) - 1; 3493 } 3494 3495 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 3496 { 3497 spin_lock(&txq->_xmit_lock); 3498 txq->xmit_lock_owner = cpu; 3499 } 3500 3501 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 3502 { 3503 __acquire(&txq->_xmit_lock); 3504 return true; 3505 } 3506 3507 static inline void __netif_tx_release(struct netdev_queue *txq) 3508 { 3509 __release(&txq->_xmit_lock); 3510 } 3511 3512 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 3513 { 3514 spin_lock_bh(&txq->_xmit_lock); 3515 txq->xmit_lock_owner = smp_processor_id(); 3516 } 3517 3518 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 3519 { 3520 bool ok = spin_trylock(&txq->_xmit_lock); 3521 if (likely(ok)) 3522 txq->xmit_lock_owner = smp_processor_id(); 3523 return ok; 3524 } 3525 3526 static inline void __netif_tx_unlock(struct netdev_queue *txq) 3527 { 3528 txq->xmit_lock_owner = -1; 3529 spin_unlock(&txq->_xmit_lock); 3530 } 3531 3532 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 3533 { 3534 txq->xmit_lock_owner = -1; 3535 spin_unlock_bh(&txq->_xmit_lock); 3536 } 3537 3538 static inline void txq_trans_update(struct netdev_queue *txq) 3539 { 3540 if (txq->xmit_lock_owner != -1) 3541 txq->trans_start = jiffies; 3542 } 3543 3544 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 3545 static inline void netif_trans_update(struct net_device *dev) 3546 { 3547 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 3548 3549 if (txq->trans_start != jiffies) 3550 txq->trans_start = jiffies; 3551 } 3552 3553 /** 3554 * netif_tx_lock - grab network device transmit lock 3555 * @dev: network device 3556 * 3557 * Get network device transmit lock 3558 */ 3559 static inline void netif_tx_lock(struct net_device *dev) 3560 { 3561 unsigned int i; 3562 int cpu; 3563 3564 spin_lock(&dev->tx_global_lock); 3565 cpu = smp_processor_id(); 3566 for (i = 0; i < dev->num_tx_queues; i++) { 3567 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3568 3569 /* We are the only thread of execution doing a 3570 * freeze, but we have to grab the _xmit_lock in 3571 * order to synchronize with threads which are in 3572 * the ->hard_start_xmit() handler and already 3573 * checked the frozen bit. 3574 */ 3575 __netif_tx_lock(txq, cpu); 3576 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 3577 __netif_tx_unlock(txq); 3578 } 3579 } 3580 3581 static inline void netif_tx_lock_bh(struct net_device *dev) 3582 { 3583 local_bh_disable(); 3584 netif_tx_lock(dev); 3585 } 3586 3587 static inline void netif_tx_unlock(struct net_device *dev) 3588 { 3589 unsigned int i; 3590 3591 for (i = 0; i < dev->num_tx_queues; i++) { 3592 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3593 3594 /* No need to grab the _xmit_lock here. If the 3595 * queue is not stopped for another reason, we 3596 * force a schedule. 3597 */ 3598 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 3599 netif_schedule_queue(txq); 3600 } 3601 spin_unlock(&dev->tx_global_lock); 3602 } 3603 3604 static inline void netif_tx_unlock_bh(struct net_device *dev) 3605 { 3606 netif_tx_unlock(dev); 3607 local_bh_enable(); 3608 } 3609 3610 #define HARD_TX_LOCK(dev, txq, cpu) { \ 3611 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3612 __netif_tx_lock(txq, cpu); \ 3613 } else { \ 3614 __netif_tx_acquire(txq); \ 3615 } \ 3616 } 3617 3618 #define HARD_TX_TRYLOCK(dev, txq) \ 3619 (((dev->features & NETIF_F_LLTX) == 0) ? \ 3620 __netif_tx_trylock(txq) : \ 3621 __netif_tx_acquire(txq)) 3622 3623 #define HARD_TX_UNLOCK(dev, txq) { \ 3624 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3625 __netif_tx_unlock(txq); \ 3626 } else { \ 3627 __netif_tx_release(txq); \ 3628 } \ 3629 } 3630 3631 static inline void netif_tx_disable(struct net_device *dev) 3632 { 3633 unsigned int i; 3634 int cpu; 3635 3636 local_bh_disable(); 3637 cpu = smp_processor_id(); 3638 for (i = 0; i < dev->num_tx_queues; i++) { 3639 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3640 3641 __netif_tx_lock(txq, cpu); 3642 netif_tx_stop_queue(txq); 3643 __netif_tx_unlock(txq); 3644 } 3645 local_bh_enable(); 3646 } 3647 3648 static inline void netif_addr_lock(struct net_device *dev) 3649 { 3650 spin_lock(&dev->addr_list_lock); 3651 } 3652 3653 static inline void netif_addr_lock_nested(struct net_device *dev) 3654 { 3655 int subclass = SINGLE_DEPTH_NESTING; 3656 3657 if (dev->netdev_ops->ndo_get_lock_subclass) 3658 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev); 3659 3660 spin_lock_nested(&dev->addr_list_lock, subclass); 3661 } 3662 3663 static inline void netif_addr_lock_bh(struct net_device *dev) 3664 { 3665 spin_lock_bh(&dev->addr_list_lock); 3666 } 3667 3668 static inline void netif_addr_unlock(struct net_device *dev) 3669 { 3670 spin_unlock(&dev->addr_list_lock); 3671 } 3672 3673 static inline void netif_addr_unlock_bh(struct net_device *dev) 3674 { 3675 spin_unlock_bh(&dev->addr_list_lock); 3676 } 3677 3678 /* 3679 * dev_addrs walker. Should be used only for read access. Call with 3680 * rcu_read_lock held. 3681 */ 3682 #define for_each_dev_addr(dev, ha) \ 3683 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 3684 3685 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 3686 3687 void ether_setup(struct net_device *dev); 3688 3689 /* Support for loadable net-drivers */ 3690 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 3691 unsigned char name_assign_type, 3692 void (*setup)(struct net_device *), 3693 unsigned int txqs, unsigned int rxqs); 3694 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 3695 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 3696 3697 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 3698 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 3699 count) 3700 3701 int register_netdev(struct net_device *dev); 3702 void unregister_netdev(struct net_device *dev); 3703 3704 /* General hardware address lists handling functions */ 3705 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 3706 struct netdev_hw_addr_list *from_list, int addr_len); 3707 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 3708 struct netdev_hw_addr_list *from_list, int addr_len); 3709 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 3710 struct net_device *dev, 3711 int (*sync)(struct net_device *, const unsigned char *), 3712 int (*unsync)(struct net_device *, 3713 const unsigned char *)); 3714 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 3715 struct net_device *dev, 3716 int (*unsync)(struct net_device *, 3717 const unsigned char *)); 3718 void __hw_addr_init(struct netdev_hw_addr_list *list); 3719 3720 /* Functions used for device addresses handling */ 3721 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 3722 unsigned char addr_type); 3723 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 3724 unsigned char addr_type); 3725 void dev_addr_flush(struct net_device *dev); 3726 int dev_addr_init(struct net_device *dev); 3727 3728 /* Functions used for unicast addresses handling */ 3729 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 3730 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 3731 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 3732 int dev_uc_sync(struct net_device *to, struct net_device *from); 3733 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 3734 void dev_uc_unsync(struct net_device *to, struct net_device *from); 3735 void dev_uc_flush(struct net_device *dev); 3736 void dev_uc_init(struct net_device *dev); 3737 3738 /** 3739 * __dev_uc_sync - Synchonize device's unicast list 3740 * @dev: device to sync 3741 * @sync: function to call if address should be added 3742 * @unsync: function to call if address should be removed 3743 * 3744 * Add newly added addresses to the interface, and release 3745 * addresses that have been deleted. 3746 */ 3747 static inline int __dev_uc_sync(struct net_device *dev, 3748 int (*sync)(struct net_device *, 3749 const unsigned char *), 3750 int (*unsync)(struct net_device *, 3751 const unsigned char *)) 3752 { 3753 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 3754 } 3755 3756 /** 3757 * __dev_uc_unsync - Remove synchronized addresses from device 3758 * @dev: device to sync 3759 * @unsync: function to call if address should be removed 3760 * 3761 * Remove all addresses that were added to the device by dev_uc_sync(). 3762 */ 3763 static inline void __dev_uc_unsync(struct net_device *dev, 3764 int (*unsync)(struct net_device *, 3765 const unsigned char *)) 3766 { 3767 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 3768 } 3769 3770 /* Functions used for multicast addresses handling */ 3771 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 3772 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 3773 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 3774 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 3775 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 3776 int dev_mc_sync(struct net_device *to, struct net_device *from); 3777 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 3778 void dev_mc_unsync(struct net_device *to, struct net_device *from); 3779 void dev_mc_flush(struct net_device *dev); 3780 void dev_mc_init(struct net_device *dev); 3781 3782 /** 3783 * __dev_mc_sync - Synchonize device's multicast list 3784 * @dev: device to sync 3785 * @sync: function to call if address should be added 3786 * @unsync: function to call if address should be removed 3787 * 3788 * Add newly added addresses to the interface, and release 3789 * addresses that have been deleted. 3790 */ 3791 static inline int __dev_mc_sync(struct net_device *dev, 3792 int (*sync)(struct net_device *, 3793 const unsigned char *), 3794 int (*unsync)(struct net_device *, 3795 const unsigned char *)) 3796 { 3797 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 3798 } 3799 3800 /** 3801 * __dev_mc_unsync - Remove synchronized addresses from device 3802 * @dev: device to sync 3803 * @unsync: function to call if address should be removed 3804 * 3805 * Remove all addresses that were added to the device by dev_mc_sync(). 3806 */ 3807 static inline void __dev_mc_unsync(struct net_device *dev, 3808 int (*unsync)(struct net_device *, 3809 const unsigned char *)) 3810 { 3811 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 3812 } 3813 3814 /* Functions used for secondary unicast and multicast support */ 3815 void dev_set_rx_mode(struct net_device *dev); 3816 void __dev_set_rx_mode(struct net_device *dev); 3817 int dev_set_promiscuity(struct net_device *dev, int inc); 3818 int dev_set_allmulti(struct net_device *dev, int inc); 3819 void netdev_state_change(struct net_device *dev); 3820 void netdev_notify_peers(struct net_device *dev); 3821 void netdev_features_change(struct net_device *dev); 3822 /* Load a device via the kmod */ 3823 void dev_load(struct net *net, const char *name); 3824 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 3825 struct rtnl_link_stats64 *storage); 3826 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 3827 const struct net_device_stats *netdev_stats); 3828 3829 extern int netdev_max_backlog; 3830 extern int netdev_tstamp_prequeue; 3831 extern int weight_p; 3832 extern int dev_weight_rx_bias; 3833 extern int dev_weight_tx_bias; 3834 extern int dev_rx_weight; 3835 extern int dev_tx_weight; 3836 3837 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 3838 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 3839 struct list_head **iter); 3840 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 3841 struct list_head **iter); 3842 3843 /* iterate through upper list, must be called under RCU read lock */ 3844 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 3845 for (iter = &(dev)->adj_list.upper, \ 3846 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 3847 updev; \ 3848 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 3849 3850 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 3851 int (*fn)(struct net_device *upper_dev, 3852 void *data), 3853 void *data); 3854 3855 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 3856 struct net_device *upper_dev); 3857 3858 void *netdev_lower_get_next_private(struct net_device *dev, 3859 struct list_head **iter); 3860 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 3861 struct list_head **iter); 3862 3863 #define netdev_for_each_lower_private(dev, priv, iter) \ 3864 for (iter = (dev)->adj_list.lower.next, \ 3865 priv = netdev_lower_get_next_private(dev, &(iter)); \ 3866 priv; \ 3867 priv = netdev_lower_get_next_private(dev, &(iter))) 3868 3869 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 3870 for (iter = &(dev)->adj_list.lower, \ 3871 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 3872 priv; \ 3873 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 3874 3875 void *netdev_lower_get_next(struct net_device *dev, 3876 struct list_head **iter); 3877 3878 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 3879 for (iter = (dev)->adj_list.lower.next, \ 3880 ldev = netdev_lower_get_next(dev, &(iter)); \ 3881 ldev; \ 3882 ldev = netdev_lower_get_next(dev, &(iter))) 3883 3884 struct net_device *netdev_all_lower_get_next(struct net_device *dev, 3885 struct list_head **iter); 3886 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev, 3887 struct list_head **iter); 3888 3889 int netdev_walk_all_lower_dev(struct net_device *dev, 3890 int (*fn)(struct net_device *lower_dev, 3891 void *data), 3892 void *data); 3893 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 3894 int (*fn)(struct net_device *lower_dev, 3895 void *data), 3896 void *data); 3897 3898 void *netdev_adjacent_get_private(struct list_head *adj_list); 3899 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 3900 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 3901 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 3902 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev); 3903 int netdev_master_upper_dev_link(struct net_device *dev, 3904 struct net_device *upper_dev, 3905 void *upper_priv, void *upper_info); 3906 void netdev_upper_dev_unlink(struct net_device *dev, 3907 struct net_device *upper_dev); 3908 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 3909 void *netdev_lower_dev_get_private(struct net_device *dev, 3910 struct net_device *lower_dev); 3911 void netdev_lower_state_changed(struct net_device *lower_dev, 3912 void *lower_state_info); 3913 3914 /* RSS keys are 40 or 52 bytes long */ 3915 #define NETDEV_RSS_KEY_LEN 52 3916 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 3917 void netdev_rss_key_fill(void *buffer, size_t len); 3918 3919 int dev_get_nest_level(struct net_device *dev); 3920 int skb_checksum_help(struct sk_buff *skb); 3921 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 3922 netdev_features_t features, bool tx_path); 3923 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 3924 netdev_features_t features); 3925 3926 struct netdev_bonding_info { 3927 ifslave slave; 3928 ifbond master; 3929 }; 3930 3931 struct netdev_notifier_bonding_info { 3932 struct netdev_notifier_info info; /* must be first */ 3933 struct netdev_bonding_info bonding_info; 3934 }; 3935 3936 void netdev_bonding_info_change(struct net_device *dev, 3937 struct netdev_bonding_info *bonding_info); 3938 3939 static inline 3940 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 3941 { 3942 return __skb_gso_segment(skb, features, true); 3943 } 3944 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 3945 3946 static inline bool can_checksum_protocol(netdev_features_t features, 3947 __be16 protocol) 3948 { 3949 if (protocol == htons(ETH_P_FCOE)) 3950 return !!(features & NETIF_F_FCOE_CRC); 3951 3952 /* Assume this is an IP checksum (not SCTP CRC) */ 3953 3954 if (features & NETIF_F_HW_CSUM) { 3955 /* Can checksum everything */ 3956 return true; 3957 } 3958 3959 switch (protocol) { 3960 case htons(ETH_P_IP): 3961 return !!(features & NETIF_F_IP_CSUM); 3962 case htons(ETH_P_IPV6): 3963 return !!(features & NETIF_F_IPV6_CSUM); 3964 default: 3965 return false; 3966 } 3967 } 3968 3969 #ifdef CONFIG_BUG 3970 void netdev_rx_csum_fault(struct net_device *dev); 3971 #else 3972 static inline void netdev_rx_csum_fault(struct net_device *dev) 3973 { 3974 } 3975 #endif 3976 /* rx skb timestamps */ 3977 void net_enable_timestamp(void); 3978 void net_disable_timestamp(void); 3979 3980 #ifdef CONFIG_PROC_FS 3981 int __init dev_proc_init(void); 3982 #else 3983 #define dev_proc_init() 0 3984 #endif 3985 3986 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 3987 struct sk_buff *skb, struct net_device *dev, 3988 bool more) 3989 { 3990 skb->xmit_more = more ? 1 : 0; 3991 return ops->ndo_start_xmit(skb, dev); 3992 } 3993 3994 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 3995 struct netdev_queue *txq, bool more) 3996 { 3997 const struct net_device_ops *ops = dev->netdev_ops; 3998 int rc; 3999 4000 rc = __netdev_start_xmit(ops, skb, dev, more); 4001 if (rc == NETDEV_TX_OK) 4002 txq_trans_update(txq); 4003 4004 return rc; 4005 } 4006 4007 int netdev_class_create_file_ns(struct class_attribute *class_attr, 4008 const void *ns); 4009 void netdev_class_remove_file_ns(struct class_attribute *class_attr, 4010 const void *ns); 4011 4012 static inline int netdev_class_create_file(struct class_attribute *class_attr) 4013 { 4014 return netdev_class_create_file_ns(class_attr, NULL); 4015 } 4016 4017 static inline void netdev_class_remove_file(struct class_attribute *class_attr) 4018 { 4019 netdev_class_remove_file_ns(class_attr, NULL); 4020 } 4021 4022 extern struct kobj_ns_type_operations net_ns_type_operations; 4023 4024 const char *netdev_drivername(const struct net_device *dev); 4025 4026 void linkwatch_run_queue(void); 4027 4028 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4029 netdev_features_t f2) 4030 { 4031 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4032 if (f1 & NETIF_F_HW_CSUM) 4033 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4034 else 4035 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4036 } 4037 4038 return f1 & f2; 4039 } 4040 4041 static inline netdev_features_t netdev_get_wanted_features( 4042 struct net_device *dev) 4043 { 4044 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4045 } 4046 netdev_features_t netdev_increment_features(netdev_features_t all, 4047 netdev_features_t one, netdev_features_t mask); 4048 4049 /* Allow TSO being used on stacked device : 4050 * Performing the GSO segmentation before last device 4051 * is a performance improvement. 4052 */ 4053 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4054 netdev_features_t mask) 4055 { 4056 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4057 } 4058 4059 int __netdev_update_features(struct net_device *dev); 4060 void netdev_update_features(struct net_device *dev); 4061 void netdev_change_features(struct net_device *dev); 4062 4063 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4064 struct net_device *dev); 4065 4066 netdev_features_t passthru_features_check(struct sk_buff *skb, 4067 struct net_device *dev, 4068 netdev_features_t features); 4069 netdev_features_t netif_skb_features(struct sk_buff *skb); 4070 4071 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4072 { 4073 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4074 4075 /* check flags correspondence */ 4076 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4077 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT)); 4078 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4079 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4080 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4081 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4082 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4083 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4084 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4085 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4086 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4087 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4088 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4089 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4090 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4091 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4092 4093 return (features & feature) == feature; 4094 } 4095 4096 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4097 { 4098 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4099 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4100 } 4101 4102 static inline bool netif_needs_gso(struct sk_buff *skb, 4103 netdev_features_t features) 4104 { 4105 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4106 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4107 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4108 } 4109 4110 static inline void netif_set_gso_max_size(struct net_device *dev, 4111 unsigned int size) 4112 { 4113 dev->gso_max_size = size; 4114 } 4115 4116 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4117 int pulled_hlen, u16 mac_offset, 4118 int mac_len) 4119 { 4120 skb->protocol = protocol; 4121 skb->encapsulation = 1; 4122 skb_push(skb, pulled_hlen); 4123 skb_reset_transport_header(skb); 4124 skb->mac_header = mac_offset; 4125 skb->network_header = skb->mac_header + mac_len; 4126 skb->mac_len = mac_len; 4127 } 4128 4129 static inline bool netif_is_macsec(const struct net_device *dev) 4130 { 4131 return dev->priv_flags & IFF_MACSEC; 4132 } 4133 4134 static inline bool netif_is_macvlan(const struct net_device *dev) 4135 { 4136 return dev->priv_flags & IFF_MACVLAN; 4137 } 4138 4139 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4140 { 4141 return dev->priv_flags & IFF_MACVLAN_PORT; 4142 } 4143 4144 static inline bool netif_is_ipvlan(const struct net_device *dev) 4145 { 4146 return dev->priv_flags & IFF_IPVLAN_SLAVE; 4147 } 4148 4149 static inline bool netif_is_ipvlan_port(const struct net_device *dev) 4150 { 4151 return dev->priv_flags & IFF_IPVLAN_MASTER; 4152 } 4153 4154 static inline bool netif_is_bond_master(const struct net_device *dev) 4155 { 4156 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4157 } 4158 4159 static inline bool netif_is_bond_slave(const struct net_device *dev) 4160 { 4161 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4162 } 4163 4164 static inline bool netif_supports_nofcs(struct net_device *dev) 4165 { 4166 return dev->priv_flags & IFF_SUPP_NOFCS; 4167 } 4168 4169 static inline bool netif_is_l3_master(const struct net_device *dev) 4170 { 4171 return dev->priv_flags & IFF_L3MDEV_MASTER; 4172 } 4173 4174 static inline bool netif_is_l3_slave(const struct net_device *dev) 4175 { 4176 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4177 } 4178 4179 static inline bool netif_is_bridge_master(const struct net_device *dev) 4180 { 4181 return dev->priv_flags & IFF_EBRIDGE; 4182 } 4183 4184 static inline bool netif_is_bridge_port(const struct net_device *dev) 4185 { 4186 return dev->priv_flags & IFF_BRIDGE_PORT; 4187 } 4188 4189 static inline bool netif_is_ovs_master(const struct net_device *dev) 4190 { 4191 return dev->priv_flags & IFF_OPENVSWITCH; 4192 } 4193 4194 static inline bool netif_is_team_master(const struct net_device *dev) 4195 { 4196 return dev->priv_flags & IFF_TEAM; 4197 } 4198 4199 static inline bool netif_is_team_port(const struct net_device *dev) 4200 { 4201 return dev->priv_flags & IFF_TEAM_PORT; 4202 } 4203 4204 static inline bool netif_is_lag_master(const struct net_device *dev) 4205 { 4206 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4207 } 4208 4209 static inline bool netif_is_lag_port(const struct net_device *dev) 4210 { 4211 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4212 } 4213 4214 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4215 { 4216 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4217 } 4218 4219 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4220 static inline void netif_keep_dst(struct net_device *dev) 4221 { 4222 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4223 } 4224 4225 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4226 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4227 { 4228 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4229 return dev->priv_flags & IFF_MACSEC; 4230 } 4231 4232 extern struct pernet_operations __net_initdata loopback_net_ops; 4233 4234 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4235 4236 /* netdev_printk helpers, similar to dev_printk */ 4237 4238 static inline const char *netdev_name(const struct net_device *dev) 4239 { 4240 if (!dev->name[0] || strchr(dev->name, '%')) 4241 return "(unnamed net_device)"; 4242 return dev->name; 4243 } 4244 4245 static inline const char *netdev_reg_state(const struct net_device *dev) 4246 { 4247 switch (dev->reg_state) { 4248 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4249 case NETREG_REGISTERED: return ""; 4250 case NETREG_UNREGISTERING: return " (unregistering)"; 4251 case NETREG_UNREGISTERED: return " (unregistered)"; 4252 case NETREG_RELEASED: return " (released)"; 4253 case NETREG_DUMMY: return " (dummy)"; 4254 } 4255 4256 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4257 return " (unknown)"; 4258 } 4259 4260 __printf(3, 4) 4261 void netdev_printk(const char *level, const struct net_device *dev, 4262 const char *format, ...); 4263 __printf(2, 3) 4264 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4265 __printf(2, 3) 4266 void netdev_alert(const struct net_device *dev, const char *format, ...); 4267 __printf(2, 3) 4268 void netdev_crit(const struct net_device *dev, const char *format, ...); 4269 __printf(2, 3) 4270 void netdev_err(const struct net_device *dev, const char *format, ...); 4271 __printf(2, 3) 4272 void netdev_warn(const struct net_device *dev, const char *format, ...); 4273 __printf(2, 3) 4274 void netdev_notice(const struct net_device *dev, const char *format, ...); 4275 __printf(2, 3) 4276 void netdev_info(const struct net_device *dev, const char *format, ...); 4277 4278 #define MODULE_ALIAS_NETDEV(device) \ 4279 MODULE_ALIAS("netdev-" device) 4280 4281 #if defined(CONFIG_DYNAMIC_DEBUG) 4282 #define netdev_dbg(__dev, format, args...) \ 4283 do { \ 4284 dynamic_netdev_dbg(__dev, format, ##args); \ 4285 } while (0) 4286 #elif defined(DEBUG) 4287 #define netdev_dbg(__dev, format, args...) \ 4288 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4289 #else 4290 #define netdev_dbg(__dev, format, args...) \ 4291 ({ \ 4292 if (0) \ 4293 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4294 }) 4295 #endif 4296 4297 #if defined(VERBOSE_DEBUG) 4298 #define netdev_vdbg netdev_dbg 4299 #else 4300 4301 #define netdev_vdbg(dev, format, args...) \ 4302 ({ \ 4303 if (0) \ 4304 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 4305 0; \ 4306 }) 4307 #endif 4308 4309 /* 4310 * netdev_WARN() acts like dev_printk(), but with the key difference 4311 * of using a WARN/WARN_ON to get the message out, including the 4312 * file/line information and a backtrace. 4313 */ 4314 #define netdev_WARN(dev, format, args...) \ 4315 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \ 4316 netdev_reg_state(dev), ##args) 4317 4318 /* netif printk helpers, similar to netdev_printk */ 4319 4320 #define netif_printk(priv, type, level, dev, fmt, args...) \ 4321 do { \ 4322 if (netif_msg_##type(priv)) \ 4323 netdev_printk(level, (dev), fmt, ##args); \ 4324 } while (0) 4325 4326 #define netif_level(level, priv, type, dev, fmt, args...) \ 4327 do { \ 4328 if (netif_msg_##type(priv)) \ 4329 netdev_##level(dev, fmt, ##args); \ 4330 } while (0) 4331 4332 #define netif_emerg(priv, type, dev, fmt, args...) \ 4333 netif_level(emerg, priv, type, dev, fmt, ##args) 4334 #define netif_alert(priv, type, dev, fmt, args...) \ 4335 netif_level(alert, priv, type, dev, fmt, ##args) 4336 #define netif_crit(priv, type, dev, fmt, args...) \ 4337 netif_level(crit, priv, type, dev, fmt, ##args) 4338 #define netif_err(priv, type, dev, fmt, args...) \ 4339 netif_level(err, priv, type, dev, fmt, ##args) 4340 #define netif_warn(priv, type, dev, fmt, args...) \ 4341 netif_level(warn, priv, type, dev, fmt, ##args) 4342 #define netif_notice(priv, type, dev, fmt, args...) \ 4343 netif_level(notice, priv, type, dev, fmt, ##args) 4344 #define netif_info(priv, type, dev, fmt, args...) \ 4345 netif_level(info, priv, type, dev, fmt, ##args) 4346 4347 #if defined(CONFIG_DYNAMIC_DEBUG) 4348 #define netif_dbg(priv, type, netdev, format, args...) \ 4349 do { \ 4350 if (netif_msg_##type(priv)) \ 4351 dynamic_netdev_dbg(netdev, format, ##args); \ 4352 } while (0) 4353 #elif defined(DEBUG) 4354 #define netif_dbg(priv, type, dev, format, args...) \ 4355 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 4356 #else 4357 #define netif_dbg(priv, type, dev, format, args...) \ 4358 ({ \ 4359 if (0) \ 4360 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4361 0; \ 4362 }) 4363 #endif 4364 4365 /* if @cond then downgrade to debug, else print at @level */ 4366 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 4367 do { \ 4368 if (cond) \ 4369 netif_dbg(priv, type, netdev, fmt, ##args); \ 4370 else \ 4371 netif_ ## level(priv, type, netdev, fmt, ##args); \ 4372 } while (0) 4373 4374 #if defined(VERBOSE_DEBUG) 4375 #define netif_vdbg netif_dbg 4376 #else 4377 #define netif_vdbg(priv, type, dev, format, args...) \ 4378 ({ \ 4379 if (0) \ 4380 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4381 0; \ 4382 }) 4383 #endif 4384 4385 /* 4386 * The list of packet types we will receive (as opposed to discard) 4387 * and the routines to invoke. 4388 * 4389 * Why 16. Because with 16 the only overlap we get on a hash of the 4390 * low nibble of the protocol value is RARP/SNAP/X.25. 4391 * 4392 * NOTE: That is no longer true with the addition of VLAN tags. Not 4393 * sure which should go first, but I bet it won't make much 4394 * difference if we are running VLANs. The good news is that 4395 * this protocol won't be in the list unless compiled in, so 4396 * the average user (w/out VLANs) will not be adversely affected. 4397 * --BLG 4398 * 4399 * 0800 IP 4400 * 8100 802.1Q VLAN 4401 * 0001 802.3 4402 * 0002 AX.25 4403 * 0004 802.2 4404 * 8035 RARP 4405 * 0005 SNAP 4406 * 0805 X.25 4407 * 0806 ARP 4408 * 8137 IPX 4409 * 0009 Localtalk 4410 * 86DD IPv6 4411 */ 4412 #define PTYPE_HASH_SIZE (16) 4413 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 4414 4415 #endif /* _LINUX_NETDEVICE_H */ 4416