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