1 /* 2 * INET An implementation of the TCP/IP protocol suite for the LINUX 3 * operating system. INET is implemented using the BSD Socket 4 * interface as the means of communication with the user level. 5 * 6 * Definitions for the Interfaces handler. 7 * 8 * Version: @(#)dev.h 1.0.10 08/12/93 9 * 10 * Authors: Ross Biro 11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 12 * Corey Minyard <wf-rch!minyard@relay.EU.net> 13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov> 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk> 15 * Bjorn Ekwall. <bj0rn@blox.se> 16 * Pekka Riikonen <priikone@poseidon.pspt.fi> 17 * 18 * This program is free software; you can redistribute it and/or 19 * modify it under the terms of the GNU General Public License 20 * as published by the Free Software Foundation; either version 21 * 2 of the License, or (at your option) any later version. 22 * 23 * Moved to /usr/include/linux for NET3 24 */ 25 #ifndef _LINUX_NETDEVICE_H 26 #define _LINUX_NETDEVICE_H 27 28 #include <linux/timer.h> 29 #include <linux/bug.h> 30 #include <linux/delay.h> 31 #include <linux/atomic.h> 32 #include <linux/prefetch.h> 33 #include <asm/cache.h> 34 #include <asm/byteorder.h> 35 36 #include <linux/percpu.h> 37 #include <linux/rculist.h> 38 #include <linux/workqueue.h> 39 #include <linux/dynamic_queue_limits.h> 40 41 #include <linux/ethtool.h> 42 #include <net/net_namespace.h> 43 #ifdef CONFIG_DCB 44 #include <net/dcbnl.h> 45 #endif 46 #include <net/netprio_cgroup.h> 47 #include <net/xdp.h> 48 49 #include <linux/netdev_features.h> 50 #include <linux/neighbour.h> 51 #include <uapi/linux/netdevice.h> 52 #include <uapi/linux/if_bonding.h> 53 #include <uapi/linux/pkt_cls.h> 54 #include <linux/hashtable.h> 55 56 struct netpoll_info; 57 struct device; 58 struct phy_device; 59 struct dsa_port; 60 61 struct sfp_bus; 62 /* 802.11 specific */ 63 struct wireless_dev; 64 /* 802.15.4 specific */ 65 struct wpan_dev; 66 struct mpls_dev; 67 /* UDP Tunnel offloads */ 68 struct udp_tunnel_info; 69 struct bpf_prog; 70 struct xdp_buff; 71 72 void netdev_set_default_ethtool_ops(struct net_device *dev, 73 const struct ethtool_ops *ops); 74 75 /* Backlog congestion levels */ 76 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 77 #define NET_RX_DROP 1 /* packet dropped */ 78 79 /* 80 * Transmit return codes: transmit return codes originate from three different 81 * namespaces: 82 * 83 * - qdisc return codes 84 * - driver transmit return codes 85 * - errno values 86 * 87 * Drivers are allowed to return any one of those in their hard_start_xmit() 88 * function. Real network devices commonly used with qdiscs should only return 89 * the driver transmit return codes though - when qdiscs are used, the actual 90 * transmission happens asynchronously, so the value is not propagated to 91 * higher layers. Virtual network devices transmit synchronously; in this case 92 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 93 * others are propagated to higher layers. 94 */ 95 96 /* qdisc ->enqueue() return codes. */ 97 #define NET_XMIT_SUCCESS 0x00 98 #define NET_XMIT_DROP 0x01 /* skb dropped */ 99 #define NET_XMIT_CN 0x02 /* congestion notification */ 100 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 101 102 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 103 * indicates that the device will soon be dropping packets, or already drops 104 * some packets of the same priority; prompting us to send less aggressively. */ 105 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 106 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 107 108 /* Driver transmit return codes */ 109 #define NETDEV_TX_MASK 0xf0 110 111 enum netdev_tx { 112 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 113 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 114 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 115 }; 116 typedef enum netdev_tx netdev_tx_t; 117 118 /* 119 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 120 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 121 */ 122 static inline bool dev_xmit_complete(int rc) 123 { 124 /* 125 * Positive cases with an skb consumed by a driver: 126 * - successful transmission (rc == NETDEV_TX_OK) 127 * - error while transmitting (rc < 0) 128 * - error while queueing to a different device (rc & NET_XMIT_MASK) 129 */ 130 if (likely(rc < NET_XMIT_MASK)) 131 return true; 132 133 return false; 134 } 135 136 /* 137 * Compute the worst-case header length according to the protocols 138 * used. 139 */ 140 141 #if defined(CONFIG_HYPERV_NET) 142 # define LL_MAX_HEADER 128 143 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 144 # if defined(CONFIG_MAC80211_MESH) 145 # define LL_MAX_HEADER 128 146 # else 147 # define LL_MAX_HEADER 96 148 # endif 149 #else 150 # define LL_MAX_HEADER 32 151 #endif 152 153 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 154 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 155 #define MAX_HEADER LL_MAX_HEADER 156 #else 157 #define MAX_HEADER (LL_MAX_HEADER + 48) 158 #endif 159 160 /* 161 * Old network device statistics. Fields are native words 162 * (unsigned long) so they can be read and written atomically. 163 */ 164 165 struct net_device_stats { 166 unsigned long rx_packets; 167 unsigned long tx_packets; 168 unsigned long rx_bytes; 169 unsigned long tx_bytes; 170 unsigned long rx_errors; 171 unsigned long tx_errors; 172 unsigned long rx_dropped; 173 unsigned long tx_dropped; 174 unsigned long multicast; 175 unsigned long collisions; 176 unsigned long rx_length_errors; 177 unsigned long rx_over_errors; 178 unsigned long rx_crc_errors; 179 unsigned long rx_frame_errors; 180 unsigned long rx_fifo_errors; 181 unsigned long rx_missed_errors; 182 unsigned long tx_aborted_errors; 183 unsigned long tx_carrier_errors; 184 unsigned long tx_fifo_errors; 185 unsigned long tx_heartbeat_errors; 186 unsigned long tx_window_errors; 187 unsigned long rx_compressed; 188 unsigned long tx_compressed; 189 }; 190 191 192 #include <linux/cache.h> 193 #include <linux/skbuff.h> 194 195 #ifdef CONFIG_RPS 196 #include <linux/static_key.h> 197 extern struct static_key rps_needed; 198 extern struct static_key rfs_needed; 199 #endif 200 201 struct neighbour; 202 struct neigh_parms; 203 struct sk_buff; 204 205 struct netdev_hw_addr { 206 struct list_head list; 207 unsigned char addr[MAX_ADDR_LEN]; 208 unsigned char type; 209 #define NETDEV_HW_ADDR_T_LAN 1 210 #define NETDEV_HW_ADDR_T_SAN 2 211 #define NETDEV_HW_ADDR_T_SLAVE 3 212 #define NETDEV_HW_ADDR_T_UNICAST 4 213 #define NETDEV_HW_ADDR_T_MULTICAST 5 214 bool global_use; 215 int sync_cnt; 216 int refcount; 217 int synced; 218 struct rcu_head rcu_head; 219 }; 220 221 struct netdev_hw_addr_list { 222 struct list_head list; 223 int count; 224 }; 225 226 #define netdev_hw_addr_list_count(l) ((l)->count) 227 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 228 #define netdev_hw_addr_list_for_each(ha, l) \ 229 list_for_each_entry(ha, &(l)->list, list) 230 231 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 232 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 233 #define netdev_for_each_uc_addr(ha, dev) \ 234 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 235 236 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 237 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 238 #define netdev_for_each_mc_addr(ha, dev) \ 239 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 240 241 struct hh_cache { 242 unsigned int hh_len; 243 seqlock_t hh_lock; 244 245 /* cached hardware header; allow for machine alignment needs. */ 246 #define HH_DATA_MOD 16 247 #define HH_DATA_OFF(__len) \ 248 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 249 #define HH_DATA_ALIGN(__len) \ 250 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 251 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 252 }; 253 254 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 255 * Alternative is: 256 * dev->hard_header_len ? (dev->hard_header_len + 257 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 258 * 259 * We could use other alignment values, but we must maintain the 260 * relationship HH alignment <= LL alignment. 261 */ 262 #define LL_RESERVED_SPACE(dev) \ 263 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 264 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 265 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 266 267 struct header_ops { 268 int (*create) (struct sk_buff *skb, struct net_device *dev, 269 unsigned short type, const void *daddr, 270 const void *saddr, unsigned int len); 271 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 272 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 273 void (*cache_update)(struct hh_cache *hh, 274 const struct net_device *dev, 275 const unsigned char *haddr); 276 bool (*validate)(const char *ll_header, unsigned int len); 277 }; 278 279 /* These flag bits are private to the generic network queueing 280 * layer; they may not be explicitly referenced by any other 281 * code. 282 */ 283 284 enum netdev_state_t { 285 __LINK_STATE_START, 286 __LINK_STATE_PRESENT, 287 __LINK_STATE_NOCARRIER, 288 __LINK_STATE_LINKWATCH_PENDING, 289 __LINK_STATE_DORMANT, 290 }; 291 292 293 /* 294 * This structure holds boot-time configured netdevice settings. They 295 * are then used in the device probing. 296 */ 297 struct netdev_boot_setup { 298 char name[IFNAMSIZ]; 299 struct ifmap map; 300 }; 301 #define NETDEV_BOOT_SETUP_MAX 8 302 303 int __init netdev_boot_setup(char *str); 304 305 struct gro_list { 306 struct list_head list; 307 int count; 308 }; 309 310 /* 311 * Structure for NAPI scheduling similar to tasklet but with weighting 312 */ 313 #define GRO_HASH_BUCKETS 8 314 struct napi_struct { 315 /* The poll_list must only be managed by the entity which 316 * changes the state of the NAPI_STATE_SCHED bit. This means 317 * whoever atomically sets that bit can add this napi_struct 318 * to the per-CPU poll_list, and whoever clears that bit 319 * can remove from the list right before clearing the bit. 320 */ 321 struct list_head poll_list; 322 323 unsigned long state; 324 int weight; 325 unsigned int gro_count; 326 int (*poll)(struct napi_struct *, int); 327 #ifdef CONFIG_NETPOLL 328 int poll_owner; 329 #endif 330 struct net_device *dev; 331 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 332 struct sk_buff *skb; 333 struct hrtimer timer; 334 struct list_head dev_list; 335 struct hlist_node napi_hash_node; 336 unsigned int napi_id; 337 }; 338 339 enum { 340 NAPI_STATE_SCHED, /* Poll is scheduled */ 341 NAPI_STATE_MISSED, /* reschedule a napi */ 342 NAPI_STATE_DISABLE, /* Disable pending */ 343 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 344 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */ 345 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */ 346 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */ 347 }; 348 349 enum { 350 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 351 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 352 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 353 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 354 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED), 355 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 356 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 357 }; 358 359 enum gro_result { 360 GRO_MERGED, 361 GRO_MERGED_FREE, 362 GRO_HELD, 363 GRO_NORMAL, 364 GRO_DROP, 365 GRO_CONSUMED, 366 }; 367 typedef enum gro_result gro_result_t; 368 369 /* 370 * enum rx_handler_result - Possible return values for rx_handlers. 371 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 372 * further. 373 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 374 * case skb->dev was changed by rx_handler. 375 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 376 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 377 * 378 * rx_handlers are functions called from inside __netif_receive_skb(), to do 379 * special processing of the skb, prior to delivery to protocol handlers. 380 * 381 * Currently, a net_device can only have a single rx_handler registered. Trying 382 * to register a second rx_handler will return -EBUSY. 383 * 384 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 385 * To unregister a rx_handler on a net_device, use 386 * netdev_rx_handler_unregister(). 387 * 388 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 389 * do with the skb. 390 * 391 * If the rx_handler consumed the skb in some way, it should return 392 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 393 * the skb to be delivered in some other way. 394 * 395 * If the rx_handler changed skb->dev, to divert the skb to another 396 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 397 * new device will be called if it exists. 398 * 399 * If the rx_handler decides the skb should be ignored, it should return 400 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 401 * are registered on exact device (ptype->dev == skb->dev). 402 * 403 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 404 * delivered, it should return RX_HANDLER_PASS. 405 * 406 * A device without a registered rx_handler will behave as if rx_handler 407 * returned RX_HANDLER_PASS. 408 */ 409 410 enum rx_handler_result { 411 RX_HANDLER_CONSUMED, 412 RX_HANDLER_ANOTHER, 413 RX_HANDLER_EXACT, 414 RX_HANDLER_PASS, 415 }; 416 typedef enum rx_handler_result rx_handler_result_t; 417 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 418 419 void __napi_schedule(struct napi_struct *n); 420 void __napi_schedule_irqoff(struct napi_struct *n); 421 422 static inline bool napi_disable_pending(struct napi_struct *n) 423 { 424 return test_bit(NAPI_STATE_DISABLE, &n->state); 425 } 426 427 bool napi_schedule_prep(struct napi_struct *n); 428 429 /** 430 * napi_schedule - schedule NAPI poll 431 * @n: NAPI context 432 * 433 * Schedule NAPI poll routine to be called if it is not already 434 * running. 435 */ 436 static inline void napi_schedule(struct napi_struct *n) 437 { 438 if (napi_schedule_prep(n)) 439 __napi_schedule(n); 440 } 441 442 /** 443 * napi_schedule_irqoff - schedule NAPI poll 444 * @n: NAPI context 445 * 446 * Variant of napi_schedule(), assuming hard irqs are masked. 447 */ 448 static inline void napi_schedule_irqoff(struct napi_struct *n) 449 { 450 if (napi_schedule_prep(n)) 451 __napi_schedule_irqoff(n); 452 } 453 454 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 455 static inline bool napi_reschedule(struct napi_struct *napi) 456 { 457 if (napi_schedule_prep(napi)) { 458 __napi_schedule(napi); 459 return true; 460 } 461 return false; 462 } 463 464 bool napi_complete_done(struct napi_struct *n, int work_done); 465 /** 466 * napi_complete - NAPI processing complete 467 * @n: NAPI context 468 * 469 * Mark NAPI processing as complete. 470 * Consider using napi_complete_done() instead. 471 * Return false if device should avoid rearming interrupts. 472 */ 473 static inline bool napi_complete(struct napi_struct *n) 474 { 475 return napi_complete_done(n, 0); 476 } 477 478 /** 479 * napi_hash_del - remove a NAPI from global table 480 * @napi: NAPI context 481 * 482 * Warning: caller must observe RCU grace period 483 * before freeing memory containing @napi, if 484 * this function returns true. 485 * Note: core networking stack automatically calls it 486 * from netif_napi_del(). 487 * Drivers might want to call this helper to combine all 488 * the needed RCU grace periods into a single one. 489 */ 490 bool napi_hash_del(struct napi_struct *napi); 491 492 /** 493 * napi_disable - prevent NAPI from scheduling 494 * @n: NAPI context 495 * 496 * Stop NAPI from being scheduled on this context. 497 * Waits till any outstanding processing completes. 498 */ 499 void napi_disable(struct napi_struct *n); 500 501 /** 502 * napi_enable - enable NAPI scheduling 503 * @n: NAPI context 504 * 505 * Resume NAPI from being scheduled on this context. 506 * Must be paired with napi_disable. 507 */ 508 static inline void napi_enable(struct napi_struct *n) 509 { 510 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 511 smp_mb__before_atomic(); 512 clear_bit(NAPI_STATE_SCHED, &n->state); 513 clear_bit(NAPI_STATE_NPSVC, &n->state); 514 } 515 516 /** 517 * napi_synchronize - wait until NAPI is not running 518 * @n: NAPI context 519 * 520 * Wait until NAPI is done being scheduled on this context. 521 * Waits till any outstanding processing completes but 522 * does not disable future activations. 523 */ 524 static inline void napi_synchronize(const struct napi_struct *n) 525 { 526 if (IS_ENABLED(CONFIG_SMP)) 527 while (test_bit(NAPI_STATE_SCHED, &n->state)) 528 msleep(1); 529 else 530 barrier(); 531 } 532 533 enum netdev_queue_state_t { 534 __QUEUE_STATE_DRV_XOFF, 535 __QUEUE_STATE_STACK_XOFF, 536 __QUEUE_STATE_FROZEN, 537 }; 538 539 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 540 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 541 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 542 543 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 544 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 545 QUEUE_STATE_FROZEN) 546 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 547 QUEUE_STATE_FROZEN) 548 549 /* 550 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 551 * netif_tx_* functions below are used to manipulate this flag. The 552 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 553 * queue independently. The netif_xmit_*stopped functions below are called 554 * to check if the queue has been stopped by the driver or stack (either 555 * of the XOFF bits are set in the state). Drivers should not need to call 556 * netif_xmit*stopped functions, they should only be using netif_tx_*. 557 */ 558 559 struct netdev_queue { 560 /* 561 * read-mostly part 562 */ 563 struct net_device *dev; 564 struct Qdisc __rcu *qdisc; 565 struct Qdisc *qdisc_sleeping; 566 #ifdef CONFIG_SYSFS 567 struct kobject kobj; 568 #endif 569 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 570 int numa_node; 571 #endif 572 unsigned long tx_maxrate; 573 /* 574 * Number of TX timeouts for this queue 575 * (/sys/class/net/DEV/Q/trans_timeout) 576 */ 577 unsigned long trans_timeout; 578 579 /* Subordinate device that the queue has been assigned to */ 580 struct net_device *sb_dev; 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 extern int sysctl_fb_tunnels_only_for_init_net; 599 600 static inline bool net_has_fallback_tunnels(const struct net *net) 601 { 602 return net == &init_net || 603 !IS_ENABLED(CONFIG_SYSCTL) || 604 !sysctl_fb_tunnels_only_for_init_net; 605 } 606 607 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 608 { 609 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 610 return q->numa_node; 611 #else 612 return NUMA_NO_NODE; 613 #endif 614 } 615 616 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 617 { 618 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 619 q->numa_node = node; 620 #endif 621 } 622 623 #ifdef CONFIG_RPS 624 /* 625 * This structure holds an RPS map which can be of variable length. The 626 * map is an array of CPUs. 627 */ 628 struct rps_map { 629 unsigned int len; 630 struct rcu_head rcu; 631 u16 cpus[0]; 632 }; 633 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 634 635 /* 636 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 637 * tail pointer for that CPU's input queue at the time of last enqueue, and 638 * a hardware filter index. 639 */ 640 struct rps_dev_flow { 641 u16 cpu; 642 u16 filter; 643 unsigned int last_qtail; 644 }; 645 #define RPS_NO_FILTER 0xffff 646 647 /* 648 * The rps_dev_flow_table structure contains a table of flow mappings. 649 */ 650 struct rps_dev_flow_table { 651 unsigned int mask; 652 struct rcu_head rcu; 653 struct rps_dev_flow flows[0]; 654 }; 655 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 656 ((_num) * sizeof(struct rps_dev_flow))) 657 658 /* 659 * The rps_sock_flow_table contains mappings of flows to the last CPU 660 * on which they were processed by the application (set in recvmsg). 661 * Each entry is a 32bit value. Upper part is the high-order bits 662 * of flow hash, lower part is CPU number. 663 * rps_cpu_mask is used to partition the space, depending on number of 664 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 665 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 666 * meaning we use 32-6=26 bits for the hash. 667 */ 668 struct rps_sock_flow_table { 669 u32 mask; 670 671 u32 ents[0] ____cacheline_aligned_in_smp; 672 }; 673 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 674 675 #define RPS_NO_CPU 0xffff 676 677 extern u32 rps_cpu_mask; 678 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 679 680 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 681 u32 hash) 682 { 683 if (table && hash) { 684 unsigned int index = hash & table->mask; 685 u32 val = hash & ~rps_cpu_mask; 686 687 /* We only give a hint, preemption can change CPU under us */ 688 val |= raw_smp_processor_id(); 689 690 if (table->ents[index] != val) 691 table->ents[index] = val; 692 } 693 } 694 695 #ifdef CONFIG_RFS_ACCEL 696 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 697 u16 filter_id); 698 #endif 699 #endif /* CONFIG_RPS */ 700 701 /* This structure contains an instance of an RX queue. */ 702 struct netdev_rx_queue { 703 #ifdef CONFIG_RPS 704 struct rps_map __rcu *rps_map; 705 struct rps_dev_flow_table __rcu *rps_flow_table; 706 #endif 707 struct kobject kobj; 708 struct net_device *dev; 709 struct xdp_rxq_info xdp_rxq; 710 } ____cacheline_aligned_in_smp; 711 712 /* 713 * RX queue sysfs structures and functions. 714 */ 715 struct rx_queue_attribute { 716 struct attribute attr; 717 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 718 ssize_t (*store)(struct netdev_rx_queue *queue, 719 const char *buf, size_t len); 720 }; 721 722 #ifdef CONFIG_XPS 723 /* 724 * This structure holds an XPS map which can be of variable length. The 725 * map is an array of queues. 726 */ 727 struct xps_map { 728 unsigned int len; 729 unsigned int alloc_len; 730 struct rcu_head rcu; 731 u16 queues[0]; 732 }; 733 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 734 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 735 - sizeof(struct xps_map)) / sizeof(u16)) 736 737 /* 738 * This structure holds all XPS maps for device. Maps are indexed by CPU. 739 */ 740 struct xps_dev_maps { 741 struct rcu_head rcu; 742 struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */ 743 }; 744 745 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 746 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 747 748 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 749 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 750 751 #endif /* CONFIG_XPS */ 752 753 #define TC_MAX_QUEUE 16 754 #define TC_BITMASK 15 755 /* HW offloaded queuing disciplines txq count and offset maps */ 756 struct netdev_tc_txq { 757 u16 count; 758 u16 offset; 759 }; 760 761 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 762 /* 763 * This structure is to hold information about the device 764 * configured to run FCoE protocol stack. 765 */ 766 struct netdev_fcoe_hbainfo { 767 char manufacturer[64]; 768 char serial_number[64]; 769 char hardware_version[64]; 770 char driver_version[64]; 771 char optionrom_version[64]; 772 char firmware_version[64]; 773 char model[256]; 774 char model_description[256]; 775 }; 776 #endif 777 778 #define MAX_PHYS_ITEM_ID_LEN 32 779 780 /* This structure holds a unique identifier to identify some 781 * physical item (port for example) used by a netdevice. 782 */ 783 struct netdev_phys_item_id { 784 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 785 unsigned char id_len; 786 }; 787 788 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 789 struct netdev_phys_item_id *b) 790 { 791 return a->id_len == b->id_len && 792 memcmp(a->id, b->id, a->id_len) == 0; 793 } 794 795 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 796 struct sk_buff *skb, 797 struct net_device *sb_dev); 798 799 enum tc_setup_type { 800 TC_SETUP_QDISC_MQPRIO, 801 TC_SETUP_CLSU32, 802 TC_SETUP_CLSFLOWER, 803 TC_SETUP_CLSMATCHALL, 804 TC_SETUP_CLSBPF, 805 TC_SETUP_BLOCK, 806 TC_SETUP_QDISC_CBS, 807 TC_SETUP_QDISC_RED, 808 TC_SETUP_QDISC_PRIO, 809 TC_SETUP_QDISC_MQ, 810 TC_SETUP_QDISC_ETF, 811 }; 812 813 /* These structures hold the attributes of bpf state that are being passed 814 * to the netdevice through the bpf op. 815 */ 816 enum bpf_netdev_command { 817 /* Set or clear a bpf program used in the earliest stages of packet 818 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 819 * is responsible for calling bpf_prog_put on any old progs that are 820 * stored. In case of error, the callee need not release the new prog 821 * reference, but on success it takes ownership and must bpf_prog_put 822 * when it is no longer used. 823 */ 824 XDP_SETUP_PROG, 825 XDP_SETUP_PROG_HW, 826 XDP_QUERY_PROG, 827 XDP_QUERY_PROG_HW, 828 /* BPF program for offload callbacks, invoked at program load time. */ 829 BPF_OFFLOAD_VERIFIER_PREP, 830 BPF_OFFLOAD_TRANSLATE, 831 BPF_OFFLOAD_DESTROY, 832 BPF_OFFLOAD_MAP_ALLOC, 833 BPF_OFFLOAD_MAP_FREE, 834 XDP_QUERY_XSK_UMEM, 835 XDP_SETUP_XSK_UMEM, 836 }; 837 838 struct bpf_prog_offload_ops; 839 struct netlink_ext_ack; 840 struct xdp_umem; 841 842 struct netdev_bpf { 843 enum bpf_netdev_command command; 844 union { 845 /* XDP_SETUP_PROG */ 846 struct { 847 u32 flags; 848 struct bpf_prog *prog; 849 struct netlink_ext_ack *extack; 850 }; 851 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */ 852 struct { 853 u32 prog_id; 854 /* flags with which program was installed */ 855 u32 prog_flags; 856 }; 857 /* BPF_OFFLOAD_VERIFIER_PREP */ 858 struct { 859 struct bpf_prog *prog; 860 const struct bpf_prog_offload_ops *ops; /* callee set */ 861 } verifier; 862 /* BPF_OFFLOAD_TRANSLATE, BPF_OFFLOAD_DESTROY */ 863 struct { 864 struct bpf_prog *prog; 865 } offload; 866 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 867 struct { 868 struct bpf_offloaded_map *offmap; 869 }; 870 /* XDP_SETUP_XSK_UMEM */ 871 struct { 872 struct xdp_umem *umem; 873 u16 queue_id; 874 } xsk; 875 }; 876 }; 877 878 #ifdef CONFIG_XFRM_OFFLOAD 879 struct xfrmdev_ops { 880 int (*xdo_dev_state_add) (struct xfrm_state *x); 881 void (*xdo_dev_state_delete) (struct xfrm_state *x); 882 void (*xdo_dev_state_free) (struct xfrm_state *x); 883 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 884 struct xfrm_state *x); 885 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 886 }; 887 #endif 888 889 #if IS_ENABLED(CONFIG_TLS_DEVICE) 890 enum tls_offload_ctx_dir { 891 TLS_OFFLOAD_CTX_DIR_RX, 892 TLS_OFFLOAD_CTX_DIR_TX, 893 }; 894 895 struct tls_crypto_info; 896 struct tls_context; 897 898 struct tlsdev_ops { 899 int (*tls_dev_add)(struct net_device *netdev, struct sock *sk, 900 enum tls_offload_ctx_dir direction, 901 struct tls_crypto_info *crypto_info, 902 u32 start_offload_tcp_sn); 903 void (*tls_dev_del)(struct net_device *netdev, 904 struct tls_context *ctx, 905 enum tls_offload_ctx_dir direction); 906 void (*tls_dev_resync_rx)(struct net_device *netdev, 907 struct sock *sk, u32 seq, u64 rcd_sn); 908 }; 909 #endif 910 911 struct dev_ifalias { 912 struct rcu_head rcuhead; 913 char ifalias[]; 914 }; 915 916 /* 917 * This structure defines the management hooks for network devices. 918 * The following hooks can be defined; unless noted otherwise, they are 919 * optional and can be filled with a null pointer. 920 * 921 * int (*ndo_init)(struct net_device *dev); 922 * This function is called once when a network device is registered. 923 * The network device can use this for any late stage initialization 924 * or semantic validation. It can fail with an error code which will 925 * be propagated back to register_netdev. 926 * 927 * void (*ndo_uninit)(struct net_device *dev); 928 * This function is called when device is unregistered or when registration 929 * fails. It is not called if init fails. 930 * 931 * int (*ndo_open)(struct net_device *dev); 932 * This function is called when a network device transitions to the up 933 * state. 934 * 935 * int (*ndo_stop)(struct net_device *dev); 936 * This function is called when a network device transitions to the down 937 * state. 938 * 939 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 940 * struct net_device *dev); 941 * Called when a packet needs to be transmitted. 942 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 943 * the queue before that can happen; it's for obsolete devices and weird 944 * corner cases, but the stack really does a non-trivial amount 945 * of useless work if you return NETDEV_TX_BUSY. 946 * Required; cannot be NULL. 947 * 948 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 949 * struct net_device *dev 950 * netdev_features_t features); 951 * Called by core transmit path to determine if device is capable of 952 * performing offload operations on a given packet. This is to give 953 * the device an opportunity to implement any restrictions that cannot 954 * be otherwise expressed by feature flags. The check is called with 955 * the set of features that the stack has calculated and it returns 956 * those the driver believes to be appropriate. 957 * 958 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 959 * struct net_device *sb_dev, 960 * select_queue_fallback_t fallback); 961 * Called to decide which queue to use when device supports multiple 962 * transmit queues. 963 * 964 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 965 * This function is called to allow device receiver to make 966 * changes to configuration when multicast or promiscuous is enabled. 967 * 968 * void (*ndo_set_rx_mode)(struct net_device *dev); 969 * This function is called device changes address list filtering. 970 * If driver handles unicast address filtering, it should set 971 * IFF_UNICAST_FLT in its priv_flags. 972 * 973 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 974 * This function is called when the Media Access Control address 975 * needs to be changed. If this interface is not defined, the 976 * MAC address can not be changed. 977 * 978 * int (*ndo_validate_addr)(struct net_device *dev); 979 * Test if Media Access Control address is valid for the device. 980 * 981 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 982 * Called when a user requests an ioctl which can't be handled by 983 * the generic interface code. If not defined ioctls return 984 * not supported error code. 985 * 986 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 987 * Used to set network devices bus interface parameters. This interface 988 * is retained for legacy reasons; new devices should use the bus 989 * interface (PCI) for low level management. 990 * 991 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 992 * Called when a user wants to change the Maximum Transfer Unit 993 * of a device. 994 * 995 * void (*ndo_tx_timeout)(struct net_device *dev); 996 * Callback used when the transmitter has not made any progress 997 * for dev->watchdog ticks. 998 * 999 * void (*ndo_get_stats64)(struct net_device *dev, 1000 * struct rtnl_link_stats64 *storage); 1001 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1002 * Called when a user wants to get the network device usage 1003 * statistics. Drivers must do one of the following: 1004 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1005 * rtnl_link_stats64 structure passed by the caller. 1006 * 2. Define @ndo_get_stats to update a net_device_stats structure 1007 * (which should normally be dev->stats) and return a pointer to 1008 * it. The structure may be changed asynchronously only if each 1009 * field is written atomically. 1010 * 3. Update dev->stats asynchronously and atomically, and define 1011 * neither operation. 1012 * 1013 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1014 * Return true if this device supports offload stats of this attr_id. 1015 * 1016 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1017 * void *attr_data) 1018 * Get statistics for offload operations by attr_id. Write it into the 1019 * attr_data pointer. 1020 * 1021 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1022 * If device supports VLAN filtering this function is called when a 1023 * VLAN id is registered. 1024 * 1025 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1026 * If device supports VLAN filtering this function is called when a 1027 * VLAN id is unregistered. 1028 * 1029 * void (*ndo_poll_controller)(struct net_device *dev); 1030 * 1031 * SR-IOV management functions. 1032 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1033 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1034 * u8 qos, __be16 proto); 1035 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1036 * int max_tx_rate); 1037 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1038 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1039 * int (*ndo_get_vf_config)(struct net_device *dev, 1040 * int vf, struct ifla_vf_info *ivf); 1041 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1042 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1043 * struct nlattr *port[]); 1044 * 1045 * Enable or disable the VF ability to query its RSS Redirection Table and 1046 * Hash Key. This is needed since on some devices VF share this information 1047 * with PF and querying it may introduce a theoretical security risk. 1048 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1049 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1050 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1051 * void *type_data); 1052 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1053 * This is always called from the stack with the rtnl lock held and netif 1054 * tx queues stopped. This allows the netdevice to perform queue 1055 * management safely. 1056 * 1057 * Fiber Channel over Ethernet (FCoE) offload functions. 1058 * int (*ndo_fcoe_enable)(struct net_device *dev); 1059 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1060 * so the underlying device can perform whatever needed configuration or 1061 * initialization to support acceleration of FCoE traffic. 1062 * 1063 * int (*ndo_fcoe_disable)(struct net_device *dev); 1064 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1065 * so the underlying device can perform whatever needed clean-ups to 1066 * stop supporting acceleration of FCoE traffic. 1067 * 1068 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1069 * struct scatterlist *sgl, unsigned int sgc); 1070 * Called when the FCoE Initiator wants to initialize an I/O that 1071 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1072 * perform necessary setup and returns 1 to indicate the device is set up 1073 * successfully to perform DDP on this I/O, otherwise this returns 0. 1074 * 1075 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1076 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1077 * indicated by the FC exchange id 'xid', so the underlying device can 1078 * clean up and reuse resources for later DDP requests. 1079 * 1080 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1081 * struct scatterlist *sgl, unsigned int sgc); 1082 * Called when the FCoE Target wants to initialize an I/O that 1083 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1084 * perform necessary setup and returns 1 to indicate the device is set up 1085 * successfully to perform DDP on this I/O, otherwise this returns 0. 1086 * 1087 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1088 * struct netdev_fcoe_hbainfo *hbainfo); 1089 * Called when the FCoE Protocol stack wants information on the underlying 1090 * device. This information is utilized by the FCoE protocol stack to 1091 * register attributes with Fiber Channel management service as per the 1092 * FC-GS Fabric Device Management Information(FDMI) specification. 1093 * 1094 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1095 * Called when the underlying device wants to override default World Wide 1096 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1097 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1098 * protocol stack to use. 1099 * 1100 * RFS acceleration. 1101 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1102 * u16 rxq_index, u32 flow_id); 1103 * Set hardware filter for RFS. rxq_index is the target queue index; 1104 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1105 * Return the filter ID on success, or a negative error code. 1106 * 1107 * Slave management functions (for bridge, bonding, etc). 1108 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1109 * Called to make another netdev an underling. 1110 * 1111 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1112 * Called to release previously enslaved netdev. 1113 * 1114 * Feature/offload setting functions. 1115 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1116 * netdev_features_t features); 1117 * Adjusts the requested feature flags according to device-specific 1118 * constraints, and returns the resulting flags. Must not modify 1119 * the device state. 1120 * 1121 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1122 * Called to update device configuration to new features. Passed 1123 * feature set might be less than what was returned by ndo_fix_features()). 1124 * Must return >0 or -errno if it changed dev->features itself. 1125 * 1126 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1127 * struct net_device *dev, 1128 * const unsigned char *addr, u16 vid, u16 flags) 1129 * Adds an FDB entry to dev for addr. 1130 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1131 * struct net_device *dev, 1132 * const unsigned char *addr, u16 vid) 1133 * Deletes the FDB entry from dev coresponding to addr. 1134 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1135 * struct net_device *dev, struct net_device *filter_dev, 1136 * int *idx) 1137 * Used to add FDB entries to dump requests. Implementers should add 1138 * entries to skb and update idx with the number of entries. 1139 * 1140 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1141 * u16 flags) 1142 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1143 * struct net_device *dev, u32 filter_mask, 1144 * int nlflags) 1145 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1146 * u16 flags); 1147 * 1148 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1149 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1150 * which do not represent real hardware may define this to allow their 1151 * userspace components to manage their virtual carrier state. Devices 1152 * that determine carrier state from physical hardware properties (eg 1153 * network cables) or protocol-dependent mechanisms (eg 1154 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1155 * 1156 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1157 * struct netdev_phys_item_id *ppid); 1158 * Called to get ID of physical port of this device. If driver does 1159 * not implement this, it is assumed that the hw is not able to have 1160 * multiple net devices on single physical port. 1161 * 1162 * void (*ndo_udp_tunnel_add)(struct net_device *dev, 1163 * struct udp_tunnel_info *ti); 1164 * Called by UDP tunnel to notify a driver about the UDP port and socket 1165 * address family that a UDP tunnel is listnening to. It is called only 1166 * when a new port starts listening. The operation is protected by the 1167 * RTNL. 1168 * 1169 * void (*ndo_udp_tunnel_del)(struct net_device *dev, 1170 * struct udp_tunnel_info *ti); 1171 * Called by UDP tunnel to notify the driver about a UDP port and socket 1172 * address family that the UDP tunnel is not listening to anymore. The 1173 * operation is protected by the RTNL. 1174 * 1175 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1176 * struct net_device *dev) 1177 * Called by upper layer devices to accelerate switching or other 1178 * station functionality into hardware. 'pdev is the lowerdev 1179 * to use for the offload and 'dev' is the net device that will 1180 * back the offload. Returns a pointer to the private structure 1181 * the upper layer will maintain. 1182 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1183 * Called by upper layer device to delete the station created 1184 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1185 * the station and priv is the structure returned by the add 1186 * operation. 1187 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1188 * int queue_index, u32 maxrate); 1189 * Called when a user wants to set a max-rate limitation of specific 1190 * TX queue. 1191 * int (*ndo_get_iflink)(const struct net_device *dev); 1192 * Called to get the iflink value of this device. 1193 * void (*ndo_change_proto_down)(struct net_device *dev, 1194 * bool proto_down); 1195 * This function is used to pass protocol port error state information 1196 * to the switch driver. The switch driver can react to the proto_down 1197 * by doing a phys down on the associated switch port. 1198 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1199 * This function is used to get egress tunnel information for given skb. 1200 * This is useful for retrieving outer tunnel header parameters while 1201 * sampling packet. 1202 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1203 * This function is used to specify the headroom that the skb must 1204 * consider when allocation skb during packet reception. Setting 1205 * appropriate rx headroom value allows avoiding skb head copy on 1206 * forward. Setting a negative value resets the rx headroom to the 1207 * default value. 1208 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1209 * This function is used to set or query state related to XDP on the 1210 * netdevice and manage BPF offload. See definition of 1211 * enum bpf_netdev_command for details. 1212 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1213 * u32 flags); 1214 * This function is used to submit @n XDP packets for transmit on a 1215 * netdevice. Returns number of frames successfully transmitted, frames 1216 * that got dropped are freed/returned via xdp_return_frame(). 1217 * Returns negative number, means general error invoking ndo, meaning 1218 * no frames were xmit'ed and core-caller will free all frames. 1219 */ 1220 struct net_device_ops { 1221 int (*ndo_init)(struct net_device *dev); 1222 void (*ndo_uninit)(struct net_device *dev); 1223 int (*ndo_open)(struct net_device *dev); 1224 int (*ndo_stop)(struct net_device *dev); 1225 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1226 struct net_device *dev); 1227 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1228 struct net_device *dev, 1229 netdev_features_t features); 1230 u16 (*ndo_select_queue)(struct net_device *dev, 1231 struct sk_buff *skb, 1232 struct net_device *sb_dev, 1233 select_queue_fallback_t fallback); 1234 void (*ndo_change_rx_flags)(struct net_device *dev, 1235 int flags); 1236 void (*ndo_set_rx_mode)(struct net_device *dev); 1237 int (*ndo_set_mac_address)(struct net_device *dev, 1238 void *addr); 1239 int (*ndo_validate_addr)(struct net_device *dev); 1240 int (*ndo_do_ioctl)(struct net_device *dev, 1241 struct ifreq *ifr, int cmd); 1242 int (*ndo_set_config)(struct net_device *dev, 1243 struct ifmap *map); 1244 int (*ndo_change_mtu)(struct net_device *dev, 1245 int new_mtu); 1246 int (*ndo_neigh_setup)(struct net_device *dev, 1247 struct neigh_parms *); 1248 void (*ndo_tx_timeout) (struct net_device *dev); 1249 1250 void (*ndo_get_stats64)(struct net_device *dev, 1251 struct rtnl_link_stats64 *storage); 1252 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1253 int (*ndo_get_offload_stats)(int attr_id, 1254 const struct net_device *dev, 1255 void *attr_data); 1256 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1257 1258 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1259 __be16 proto, u16 vid); 1260 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1261 __be16 proto, u16 vid); 1262 #ifdef CONFIG_NET_POLL_CONTROLLER 1263 void (*ndo_poll_controller)(struct net_device *dev); 1264 int (*ndo_netpoll_setup)(struct net_device *dev, 1265 struct netpoll_info *info); 1266 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1267 #endif 1268 int (*ndo_set_vf_mac)(struct net_device *dev, 1269 int queue, u8 *mac); 1270 int (*ndo_set_vf_vlan)(struct net_device *dev, 1271 int queue, u16 vlan, 1272 u8 qos, __be16 proto); 1273 int (*ndo_set_vf_rate)(struct net_device *dev, 1274 int vf, int min_tx_rate, 1275 int max_tx_rate); 1276 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1277 int vf, bool setting); 1278 int (*ndo_set_vf_trust)(struct net_device *dev, 1279 int vf, bool setting); 1280 int (*ndo_get_vf_config)(struct net_device *dev, 1281 int vf, 1282 struct ifla_vf_info *ivf); 1283 int (*ndo_set_vf_link_state)(struct net_device *dev, 1284 int vf, int link_state); 1285 int (*ndo_get_vf_stats)(struct net_device *dev, 1286 int vf, 1287 struct ifla_vf_stats 1288 *vf_stats); 1289 int (*ndo_set_vf_port)(struct net_device *dev, 1290 int vf, 1291 struct nlattr *port[]); 1292 int (*ndo_get_vf_port)(struct net_device *dev, 1293 int vf, struct sk_buff *skb); 1294 int (*ndo_set_vf_guid)(struct net_device *dev, 1295 int vf, u64 guid, 1296 int guid_type); 1297 int (*ndo_set_vf_rss_query_en)( 1298 struct net_device *dev, 1299 int vf, bool setting); 1300 int (*ndo_setup_tc)(struct net_device *dev, 1301 enum tc_setup_type type, 1302 void *type_data); 1303 #if IS_ENABLED(CONFIG_FCOE) 1304 int (*ndo_fcoe_enable)(struct net_device *dev); 1305 int (*ndo_fcoe_disable)(struct net_device *dev); 1306 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1307 u16 xid, 1308 struct scatterlist *sgl, 1309 unsigned int sgc); 1310 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1311 u16 xid); 1312 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1313 u16 xid, 1314 struct scatterlist *sgl, 1315 unsigned int sgc); 1316 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1317 struct netdev_fcoe_hbainfo *hbainfo); 1318 #endif 1319 1320 #if IS_ENABLED(CONFIG_LIBFCOE) 1321 #define NETDEV_FCOE_WWNN 0 1322 #define NETDEV_FCOE_WWPN 1 1323 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1324 u64 *wwn, int type); 1325 #endif 1326 1327 #ifdef CONFIG_RFS_ACCEL 1328 int (*ndo_rx_flow_steer)(struct net_device *dev, 1329 const struct sk_buff *skb, 1330 u16 rxq_index, 1331 u32 flow_id); 1332 #endif 1333 int (*ndo_add_slave)(struct net_device *dev, 1334 struct net_device *slave_dev, 1335 struct netlink_ext_ack *extack); 1336 int (*ndo_del_slave)(struct net_device *dev, 1337 struct net_device *slave_dev); 1338 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1339 netdev_features_t features); 1340 int (*ndo_set_features)(struct net_device *dev, 1341 netdev_features_t features); 1342 int (*ndo_neigh_construct)(struct net_device *dev, 1343 struct neighbour *n); 1344 void (*ndo_neigh_destroy)(struct net_device *dev, 1345 struct neighbour *n); 1346 1347 int (*ndo_fdb_add)(struct ndmsg *ndm, 1348 struct nlattr *tb[], 1349 struct net_device *dev, 1350 const unsigned char *addr, 1351 u16 vid, 1352 u16 flags); 1353 int (*ndo_fdb_del)(struct ndmsg *ndm, 1354 struct nlattr *tb[], 1355 struct net_device *dev, 1356 const unsigned char *addr, 1357 u16 vid); 1358 int (*ndo_fdb_dump)(struct sk_buff *skb, 1359 struct netlink_callback *cb, 1360 struct net_device *dev, 1361 struct net_device *filter_dev, 1362 int *idx); 1363 1364 int (*ndo_bridge_setlink)(struct net_device *dev, 1365 struct nlmsghdr *nlh, 1366 u16 flags); 1367 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1368 u32 pid, u32 seq, 1369 struct net_device *dev, 1370 u32 filter_mask, 1371 int nlflags); 1372 int (*ndo_bridge_dellink)(struct net_device *dev, 1373 struct nlmsghdr *nlh, 1374 u16 flags); 1375 int (*ndo_change_carrier)(struct net_device *dev, 1376 bool new_carrier); 1377 int (*ndo_get_phys_port_id)(struct net_device *dev, 1378 struct netdev_phys_item_id *ppid); 1379 int (*ndo_get_phys_port_name)(struct net_device *dev, 1380 char *name, size_t len); 1381 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1382 struct udp_tunnel_info *ti); 1383 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1384 struct udp_tunnel_info *ti); 1385 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1386 struct net_device *dev); 1387 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1388 void *priv); 1389 1390 int (*ndo_get_lock_subclass)(struct net_device *dev); 1391 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1392 int queue_index, 1393 u32 maxrate); 1394 int (*ndo_get_iflink)(const struct net_device *dev); 1395 int (*ndo_change_proto_down)(struct net_device *dev, 1396 bool proto_down); 1397 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1398 struct sk_buff *skb); 1399 void (*ndo_set_rx_headroom)(struct net_device *dev, 1400 int needed_headroom); 1401 int (*ndo_bpf)(struct net_device *dev, 1402 struct netdev_bpf *bpf); 1403 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1404 struct xdp_frame **xdp, 1405 u32 flags); 1406 int (*ndo_xsk_async_xmit)(struct net_device *dev, 1407 u32 queue_id); 1408 }; 1409 1410 /** 1411 * enum net_device_priv_flags - &struct net_device priv_flags 1412 * 1413 * These are the &struct net_device, they are only set internally 1414 * by drivers and used in the kernel. These flags are invisible to 1415 * userspace; this means that the order of these flags can change 1416 * during any kernel release. 1417 * 1418 * You should have a pretty good reason to be extending these flags. 1419 * 1420 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1421 * @IFF_EBRIDGE: Ethernet bridging device 1422 * @IFF_BONDING: bonding master or slave 1423 * @IFF_ISATAP: ISATAP interface (RFC4214) 1424 * @IFF_WAN_HDLC: WAN HDLC device 1425 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1426 * release skb->dst 1427 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1428 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1429 * @IFF_MACVLAN_PORT: device used as macvlan port 1430 * @IFF_BRIDGE_PORT: device used as bridge port 1431 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1432 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1433 * @IFF_UNICAST_FLT: Supports unicast filtering 1434 * @IFF_TEAM_PORT: device used as team port 1435 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1436 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1437 * change when it's running 1438 * @IFF_MACVLAN: Macvlan device 1439 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1440 * underlying stacked devices 1441 * @IFF_L3MDEV_MASTER: device is an L3 master device 1442 * @IFF_NO_QUEUE: device can run without qdisc attached 1443 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1444 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1445 * @IFF_TEAM: device is a team device 1446 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1447 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1448 * entity (i.e. the master device for bridged veth) 1449 * @IFF_MACSEC: device is a MACsec device 1450 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1451 * @IFF_FAILOVER: device is a failover master device 1452 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1453 */ 1454 enum netdev_priv_flags { 1455 IFF_802_1Q_VLAN = 1<<0, 1456 IFF_EBRIDGE = 1<<1, 1457 IFF_BONDING = 1<<2, 1458 IFF_ISATAP = 1<<3, 1459 IFF_WAN_HDLC = 1<<4, 1460 IFF_XMIT_DST_RELEASE = 1<<5, 1461 IFF_DONT_BRIDGE = 1<<6, 1462 IFF_DISABLE_NETPOLL = 1<<7, 1463 IFF_MACVLAN_PORT = 1<<8, 1464 IFF_BRIDGE_PORT = 1<<9, 1465 IFF_OVS_DATAPATH = 1<<10, 1466 IFF_TX_SKB_SHARING = 1<<11, 1467 IFF_UNICAST_FLT = 1<<12, 1468 IFF_TEAM_PORT = 1<<13, 1469 IFF_SUPP_NOFCS = 1<<14, 1470 IFF_LIVE_ADDR_CHANGE = 1<<15, 1471 IFF_MACVLAN = 1<<16, 1472 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1473 IFF_L3MDEV_MASTER = 1<<18, 1474 IFF_NO_QUEUE = 1<<19, 1475 IFF_OPENVSWITCH = 1<<20, 1476 IFF_L3MDEV_SLAVE = 1<<21, 1477 IFF_TEAM = 1<<22, 1478 IFF_RXFH_CONFIGURED = 1<<23, 1479 IFF_PHONY_HEADROOM = 1<<24, 1480 IFF_MACSEC = 1<<25, 1481 IFF_NO_RX_HANDLER = 1<<26, 1482 IFF_FAILOVER = 1<<27, 1483 IFF_FAILOVER_SLAVE = 1<<28, 1484 }; 1485 1486 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1487 #define IFF_EBRIDGE IFF_EBRIDGE 1488 #define IFF_BONDING IFF_BONDING 1489 #define IFF_ISATAP IFF_ISATAP 1490 #define IFF_WAN_HDLC IFF_WAN_HDLC 1491 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1492 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1493 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1494 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1495 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1496 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1497 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1498 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1499 #define IFF_TEAM_PORT IFF_TEAM_PORT 1500 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1501 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1502 #define IFF_MACVLAN IFF_MACVLAN 1503 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1504 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1505 #define IFF_NO_QUEUE IFF_NO_QUEUE 1506 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1507 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1508 #define IFF_TEAM IFF_TEAM 1509 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1510 #define IFF_MACSEC IFF_MACSEC 1511 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1512 #define IFF_FAILOVER IFF_FAILOVER 1513 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1514 1515 /** 1516 * struct net_device - The DEVICE structure. 1517 * 1518 * Actually, this whole structure is a big mistake. It mixes I/O 1519 * data with strictly "high-level" data, and it has to know about 1520 * almost every data structure used in the INET module. 1521 * 1522 * @name: This is the first field of the "visible" part of this structure 1523 * (i.e. as seen by users in the "Space.c" file). It is the name 1524 * of the interface. 1525 * 1526 * @name_hlist: Device name hash chain, please keep it close to name[] 1527 * @ifalias: SNMP alias 1528 * @mem_end: Shared memory end 1529 * @mem_start: Shared memory start 1530 * @base_addr: Device I/O address 1531 * @irq: Device IRQ number 1532 * 1533 * @state: Generic network queuing layer state, see netdev_state_t 1534 * @dev_list: The global list of network devices 1535 * @napi_list: List entry used for polling NAPI devices 1536 * @unreg_list: List entry when we are unregistering the 1537 * device; see the function unregister_netdev 1538 * @close_list: List entry used when we are closing the device 1539 * @ptype_all: Device-specific packet handlers for all protocols 1540 * @ptype_specific: Device-specific, protocol-specific packet handlers 1541 * 1542 * @adj_list: Directly linked devices, like slaves for bonding 1543 * @features: Currently active device features 1544 * @hw_features: User-changeable features 1545 * 1546 * @wanted_features: User-requested features 1547 * @vlan_features: Mask of features inheritable by VLAN devices 1548 * 1549 * @hw_enc_features: Mask of features inherited by encapsulating devices 1550 * This field indicates what encapsulation 1551 * offloads the hardware is capable of doing, 1552 * and drivers will need to set them appropriately. 1553 * 1554 * @mpls_features: Mask of features inheritable by MPLS 1555 * 1556 * @ifindex: interface index 1557 * @group: The group the device belongs to 1558 * 1559 * @stats: Statistics struct, which was left as a legacy, use 1560 * rtnl_link_stats64 instead 1561 * 1562 * @rx_dropped: Dropped packets by core network, 1563 * do not use this in drivers 1564 * @tx_dropped: Dropped packets by core network, 1565 * do not use this in drivers 1566 * @rx_nohandler: nohandler dropped packets by core network on 1567 * inactive devices, do not use this in drivers 1568 * @carrier_up_count: Number of times the carrier has been up 1569 * @carrier_down_count: Number of times the carrier has been down 1570 * 1571 * @wireless_handlers: List of functions to handle Wireless Extensions, 1572 * instead of ioctl, 1573 * see <net/iw_handler.h> for details. 1574 * @wireless_data: Instance data managed by the core of wireless extensions 1575 * 1576 * @netdev_ops: Includes several pointers to callbacks, 1577 * if one wants to override the ndo_*() functions 1578 * @ethtool_ops: Management operations 1579 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1580 * discovery handling. Necessary for e.g. 6LoWPAN. 1581 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1582 * of Layer 2 headers. 1583 * 1584 * @flags: Interface flags (a la BSD) 1585 * @priv_flags: Like 'flags' but invisible to userspace, 1586 * see if.h for the definitions 1587 * @gflags: Global flags ( kept as legacy ) 1588 * @padded: How much padding added by alloc_netdev() 1589 * @operstate: RFC2863 operstate 1590 * @link_mode: Mapping policy to operstate 1591 * @if_port: Selectable AUI, TP, ... 1592 * @dma: DMA channel 1593 * @mtu: Interface MTU value 1594 * @min_mtu: Interface Minimum MTU value 1595 * @max_mtu: Interface Maximum MTU value 1596 * @type: Interface hardware type 1597 * @hard_header_len: Maximum hardware header length. 1598 * @min_header_len: Minimum hardware header length 1599 * 1600 * @needed_headroom: Extra headroom the hardware may need, but not in all 1601 * cases can this be guaranteed 1602 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1603 * cases can this be guaranteed. Some cases also use 1604 * LL_MAX_HEADER instead to allocate the skb 1605 * 1606 * interface address info: 1607 * 1608 * @perm_addr: Permanent hw address 1609 * @addr_assign_type: Hw address assignment type 1610 * @addr_len: Hardware address length 1611 * @neigh_priv_len: Used in neigh_alloc() 1612 * @dev_id: Used to differentiate devices that share 1613 * the same link layer address 1614 * @dev_port: Used to differentiate devices that share 1615 * the same function 1616 * @addr_list_lock: XXX: need comments on this one 1617 * @uc_promisc: Counter that indicates promiscuous mode 1618 * has been enabled due to the need to listen to 1619 * additional unicast addresses in a device that 1620 * does not implement ndo_set_rx_mode() 1621 * @uc: unicast mac addresses 1622 * @mc: multicast mac addresses 1623 * @dev_addrs: list of device hw addresses 1624 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1625 * @promiscuity: Number of times the NIC is told to work in 1626 * promiscuous mode; if it becomes 0 the NIC will 1627 * exit promiscuous mode 1628 * @allmulti: Counter, enables or disables allmulticast mode 1629 * 1630 * @vlan_info: VLAN info 1631 * @dsa_ptr: dsa specific data 1632 * @tipc_ptr: TIPC specific data 1633 * @atalk_ptr: AppleTalk link 1634 * @ip_ptr: IPv4 specific data 1635 * @dn_ptr: DECnet specific data 1636 * @ip6_ptr: IPv6 specific data 1637 * @ax25_ptr: AX.25 specific data 1638 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1639 * 1640 * @dev_addr: Hw address (before bcast, 1641 * because most packets are unicast) 1642 * 1643 * @_rx: Array of RX queues 1644 * @num_rx_queues: Number of RX queues 1645 * allocated at register_netdev() time 1646 * @real_num_rx_queues: Number of RX queues currently active in device 1647 * 1648 * @rx_handler: handler for received packets 1649 * @rx_handler_data: XXX: need comments on this one 1650 * @miniq_ingress: ingress/clsact qdisc specific data for 1651 * ingress processing 1652 * @ingress_queue: XXX: need comments on this one 1653 * @broadcast: hw bcast address 1654 * 1655 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1656 * indexed by RX queue number. Assigned by driver. 1657 * This must only be set if the ndo_rx_flow_steer 1658 * operation is defined 1659 * @index_hlist: Device index hash chain 1660 * 1661 * @_tx: Array of TX queues 1662 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1663 * @real_num_tx_queues: Number of TX queues currently active in device 1664 * @qdisc: Root qdisc from userspace point of view 1665 * @tx_queue_len: Max frames per queue allowed 1666 * @tx_global_lock: XXX: need comments on this one 1667 * 1668 * @xps_maps: XXX: need comments on this one 1669 * @miniq_egress: clsact qdisc specific data for 1670 * egress processing 1671 * @watchdog_timeo: Represents the timeout that is used by 1672 * the watchdog (see dev_watchdog()) 1673 * @watchdog_timer: List of timers 1674 * 1675 * @pcpu_refcnt: Number of references to this device 1676 * @todo_list: Delayed register/unregister 1677 * @link_watch_list: XXX: need comments on this one 1678 * 1679 * @reg_state: Register/unregister state machine 1680 * @dismantle: Device is going to be freed 1681 * @rtnl_link_state: This enum represents the phases of creating 1682 * a new link 1683 * 1684 * @needs_free_netdev: Should unregister perform free_netdev? 1685 * @priv_destructor: Called from unregister 1686 * @npinfo: XXX: need comments on this one 1687 * @nd_net: Network namespace this network device is inside 1688 * 1689 * @ml_priv: Mid-layer private 1690 * @lstats: Loopback statistics 1691 * @tstats: Tunnel statistics 1692 * @dstats: Dummy statistics 1693 * @vstats: Virtual ethernet statistics 1694 * 1695 * @garp_port: GARP 1696 * @mrp_port: MRP 1697 * 1698 * @dev: Class/net/name entry 1699 * @sysfs_groups: Space for optional device, statistics and wireless 1700 * sysfs groups 1701 * 1702 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1703 * @rtnl_link_ops: Rtnl_link_ops 1704 * 1705 * @gso_max_size: Maximum size of generic segmentation offload 1706 * @gso_max_segs: Maximum number of segments that can be passed to the 1707 * NIC for GSO 1708 * 1709 * @dcbnl_ops: Data Center Bridging netlink ops 1710 * @num_tc: Number of traffic classes in the net device 1711 * @tc_to_txq: XXX: need comments on this one 1712 * @prio_tc_map: XXX: need comments on this one 1713 * 1714 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1715 * 1716 * @priomap: XXX: need comments on this one 1717 * @phydev: Physical device may attach itself 1718 * for hardware timestamping 1719 * @sfp_bus: attached &struct sfp_bus structure. 1720 * 1721 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1722 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1723 * 1724 * @proto_down: protocol port state information can be sent to the 1725 * switch driver and used to set the phys state of the 1726 * switch port. 1727 * 1728 * FIXME: cleanup struct net_device such that network protocol info 1729 * moves out. 1730 */ 1731 1732 struct net_device { 1733 char name[IFNAMSIZ]; 1734 struct hlist_node name_hlist; 1735 struct dev_ifalias __rcu *ifalias; 1736 /* 1737 * I/O specific fields 1738 * FIXME: Merge these and struct ifmap into one 1739 */ 1740 unsigned long mem_end; 1741 unsigned long mem_start; 1742 unsigned long base_addr; 1743 int irq; 1744 1745 /* 1746 * Some hardware also needs these fields (state,dev_list, 1747 * napi_list,unreg_list,close_list) but they are not 1748 * part of the usual set specified in Space.c. 1749 */ 1750 1751 unsigned long state; 1752 1753 struct list_head dev_list; 1754 struct list_head napi_list; 1755 struct list_head unreg_list; 1756 struct list_head close_list; 1757 struct list_head ptype_all; 1758 struct list_head ptype_specific; 1759 1760 struct { 1761 struct list_head upper; 1762 struct list_head lower; 1763 } adj_list; 1764 1765 netdev_features_t features; 1766 netdev_features_t hw_features; 1767 netdev_features_t wanted_features; 1768 netdev_features_t vlan_features; 1769 netdev_features_t hw_enc_features; 1770 netdev_features_t mpls_features; 1771 netdev_features_t gso_partial_features; 1772 1773 int ifindex; 1774 int group; 1775 1776 struct net_device_stats stats; 1777 1778 atomic_long_t rx_dropped; 1779 atomic_long_t tx_dropped; 1780 atomic_long_t rx_nohandler; 1781 1782 /* Stats to monitor link on/off, flapping */ 1783 atomic_t carrier_up_count; 1784 atomic_t carrier_down_count; 1785 1786 #ifdef CONFIG_WIRELESS_EXT 1787 const struct iw_handler_def *wireless_handlers; 1788 struct iw_public_data *wireless_data; 1789 #endif 1790 const struct net_device_ops *netdev_ops; 1791 const struct ethtool_ops *ethtool_ops; 1792 #ifdef CONFIG_NET_SWITCHDEV 1793 const struct switchdev_ops *switchdev_ops; 1794 #endif 1795 #ifdef CONFIG_NET_L3_MASTER_DEV 1796 const struct l3mdev_ops *l3mdev_ops; 1797 #endif 1798 #if IS_ENABLED(CONFIG_IPV6) 1799 const struct ndisc_ops *ndisc_ops; 1800 #endif 1801 1802 #ifdef CONFIG_XFRM_OFFLOAD 1803 const struct xfrmdev_ops *xfrmdev_ops; 1804 #endif 1805 1806 #if IS_ENABLED(CONFIG_TLS_DEVICE) 1807 const struct tlsdev_ops *tlsdev_ops; 1808 #endif 1809 1810 const struct header_ops *header_ops; 1811 1812 unsigned int flags; 1813 unsigned int priv_flags; 1814 1815 unsigned short gflags; 1816 unsigned short padded; 1817 1818 unsigned char operstate; 1819 unsigned char link_mode; 1820 1821 unsigned char if_port; 1822 unsigned char dma; 1823 1824 unsigned int mtu; 1825 unsigned int min_mtu; 1826 unsigned int max_mtu; 1827 unsigned short type; 1828 unsigned short hard_header_len; 1829 unsigned char min_header_len; 1830 1831 unsigned short needed_headroom; 1832 unsigned short needed_tailroom; 1833 1834 /* Interface address info. */ 1835 unsigned char perm_addr[MAX_ADDR_LEN]; 1836 unsigned char addr_assign_type; 1837 unsigned char addr_len; 1838 unsigned short neigh_priv_len; 1839 unsigned short dev_id; 1840 unsigned short dev_port; 1841 spinlock_t addr_list_lock; 1842 unsigned char name_assign_type; 1843 bool uc_promisc; 1844 struct netdev_hw_addr_list uc; 1845 struct netdev_hw_addr_list mc; 1846 struct netdev_hw_addr_list dev_addrs; 1847 1848 #ifdef CONFIG_SYSFS 1849 struct kset *queues_kset; 1850 #endif 1851 unsigned int promiscuity; 1852 unsigned int allmulti; 1853 1854 1855 /* Protocol-specific pointers */ 1856 1857 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1858 struct vlan_info __rcu *vlan_info; 1859 #endif 1860 #if IS_ENABLED(CONFIG_NET_DSA) 1861 struct dsa_port *dsa_ptr; 1862 #endif 1863 #if IS_ENABLED(CONFIG_TIPC) 1864 struct tipc_bearer __rcu *tipc_ptr; 1865 #endif 1866 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 1867 void *atalk_ptr; 1868 #endif 1869 struct in_device __rcu *ip_ptr; 1870 #if IS_ENABLED(CONFIG_DECNET) 1871 struct dn_dev __rcu *dn_ptr; 1872 #endif 1873 struct inet6_dev __rcu *ip6_ptr; 1874 #if IS_ENABLED(CONFIG_AX25) 1875 void *ax25_ptr; 1876 #endif 1877 struct wireless_dev *ieee80211_ptr; 1878 struct wpan_dev *ieee802154_ptr; 1879 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 1880 struct mpls_dev __rcu *mpls_ptr; 1881 #endif 1882 1883 /* 1884 * Cache lines mostly used on receive path (including eth_type_trans()) 1885 */ 1886 /* Interface address info used in eth_type_trans() */ 1887 unsigned char *dev_addr; 1888 1889 struct netdev_rx_queue *_rx; 1890 unsigned int num_rx_queues; 1891 unsigned int real_num_rx_queues; 1892 1893 struct bpf_prog __rcu *xdp_prog; 1894 unsigned long gro_flush_timeout; 1895 rx_handler_func_t __rcu *rx_handler; 1896 void __rcu *rx_handler_data; 1897 1898 #ifdef CONFIG_NET_CLS_ACT 1899 struct mini_Qdisc __rcu *miniq_ingress; 1900 #endif 1901 struct netdev_queue __rcu *ingress_queue; 1902 #ifdef CONFIG_NETFILTER_INGRESS 1903 struct nf_hook_entries __rcu *nf_hooks_ingress; 1904 #endif 1905 1906 unsigned char broadcast[MAX_ADDR_LEN]; 1907 #ifdef CONFIG_RFS_ACCEL 1908 struct cpu_rmap *rx_cpu_rmap; 1909 #endif 1910 struct hlist_node index_hlist; 1911 1912 /* 1913 * Cache lines mostly used on transmit path 1914 */ 1915 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1916 unsigned int num_tx_queues; 1917 unsigned int real_num_tx_queues; 1918 struct Qdisc *qdisc; 1919 #ifdef CONFIG_NET_SCHED 1920 DECLARE_HASHTABLE (qdisc_hash, 4); 1921 #endif 1922 unsigned int tx_queue_len; 1923 spinlock_t tx_global_lock; 1924 int watchdog_timeo; 1925 1926 #ifdef CONFIG_XPS 1927 struct xps_dev_maps __rcu *xps_cpus_map; 1928 struct xps_dev_maps __rcu *xps_rxqs_map; 1929 #endif 1930 #ifdef CONFIG_NET_CLS_ACT 1931 struct mini_Qdisc __rcu *miniq_egress; 1932 #endif 1933 1934 /* These may be needed for future network-power-down code. */ 1935 struct timer_list watchdog_timer; 1936 1937 int __percpu *pcpu_refcnt; 1938 struct list_head todo_list; 1939 1940 struct list_head link_watch_list; 1941 1942 enum { NETREG_UNINITIALIZED=0, 1943 NETREG_REGISTERED, /* completed register_netdevice */ 1944 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1945 NETREG_UNREGISTERED, /* completed unregister todo */ 1946 NETREG_RELEASED, /* called free_netdev */ 1947 NETREG_DUMMY, /* dummy device for NAPI poll */ 1948 } reg_state:8; 1949 1950 bool dismantle; 1951 1952 enum { 1953 RTNL_LINK_INITIALIZED, 1954 RTNL_LINK_INITIALIZING, 1955 } rtnl_link_state:16; 1956 1957 bool needs_free_netdev; 1958 void (*priv_destructor)(struct net_device *dev); 1959 1960 #ifdef CONFIG_NETPOLL 1961 struct netpoll_info __rcu *npinfo; 1962 #endif 1963 1964 possible_net_t nd_net; 1965 1966 /* mid-layer private */ 1967 union { 1968 void *ml_priv; 1969 struct pcpu_lstats __percpu *lstats; 1970 struct pcpu_sw_netstats __percpu *tstats; 1971 struct pcpu_dstats __percpu *dstats; 1972 struct pcpu_vstats __percpu *vstats; 1973 }; 1974 1975 #if IS_ENABLED(CONFIG_GARP) 1976 struct garp_port __rcu *garp_port; 1977 #endif 1978 #if IS_ENABLED(CONFIG_MRP) 1979 struct mrp_port __rcu *mrp_port; 1980 #endif 1981 1982 struct device dev; 1983 const struct attribute_group *sysfs_groups[4]; 1984 const struct attribute_group *sysfs_rx_queue_group; 1985 1986 const struct rtnl_link_ops *rtnl_link_ops; 1987 1988 /* for setting kernel sock attribute on TCP connection setup */ 1989 #define GSO_MAX_SIZE 65536 1990 unsigned int gso_max_size; 1991 #define GSO_MAX_SEGS 65535 1992 u16 gso_max_segs; 1993 1994 #ifdef CONFIG_DCB 1995 const struct dcbnl_rtnl_ops *dcbnl_ops; 1996 #endif 1997 s16 num_tc; 1998 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 1999 u8 prio_tc_map[TC_BITMASK + 1]; 2000 2001 #if IS_ENABLED(CONFIG_FCOE) 2002 unsigned int fcoe_ddp_xid; 2003 #endif 2004 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2005 struct netprio_map __rcu *priomap; 2006 #endif 2007 struct phy_device *phydev; 2008 struct sfp_bus *sfp_bus; 2009 struct lock_class_key *qdisc_tx_busylock; 2010 struct lock_class_key *qdisc_running_key; 2011 bool proto_down; 2012 }; 2013 #define to_net_dev(d) container_of(d, struct net_device, dev) 2014 2015 static inline bool netif_elide_gro(const struct net_device *dev) 2016 { 2017 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2018 return true; 2019 return false; 2020 } 2021 2022 #define NETDEV_ALIGN 32 2023 2024 static inline 2025 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2026 { 2027 return dev->prio_tc_map[prio & TC_BITMASK]; 2028 } 2029 2030 static inline 2031 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2032 { 2033 if (tc >= dev->num_tc) 2034 return -EINVAL; 2035 2036 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2037 return 0; 2038 } 2039 2040 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2041 void netdev_reset_tc(struct net_device *dev); 2042 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2043 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2044 2045 static inline 2046 int netdev_get_num_tc(struct net_device *dev) 2047 { 2048 return dev->num_tc; 2049 } 2050 2051 void netdev_unbind_sb_channel(struct net_device *dev, 2052 struct net_device *sb_dev); 2053 int netdev_bind_sb_channel_queue(struct net_device *dev, 2054 struct net_device *sb_dev, 2055 u8 tc, u16 count, u16 offset); 2056 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2057 static inline int netdev_get_sb_channel(struct net_device *dev) 2058 { 2059 return max_t(int, -dev->num_tc, 0); 2060 } 2061 2062 static inline 2063 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2064 unsigned int index) 2065 { 2066 return &dev->_tx[index]; 2067 } 2068 2069 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2070 const struct sk_buff *skb) 2071 { 2072 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2073 } 2074 2075 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2076 void (*f)(struct net_device *, 2077 struct netdev_queue *, 2078 void *), 2079 void *arg) 2080 { 2081 unsigned int i; 2082 2083 for (i = 0; i < dev->num_tx_queues; i++) 2084 f(dev, &dev->_tx[i], arg); 2085 } 2086 2087 #define netdev_lockdep_set_classes(dev) \ 2088 { \ 2089 static struct lock_class_key qdisc_tx_busylock_key; \ 2090 static struct lock_class_key qdisc_running_key; \ 2091 static struct lock_class_key qdisc_xmit_lock_key; \ 2092 static struct lock_class_key dev_addr_list_lock_key; \ 2093 unsigned int i; \ 2094 \ 2095 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2096 (dev)->qdisc_running_key = &qdisc_running_key; \ 2097 lockdep_set_class(&(dev)->addr_list_lock, \ 2098 &dev_addr_list_lock_key); \ 2099 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2100 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2101 &qdisc_xmit_lock_key); \ 2102 } 2103 2104 struct netdev_queue *netdev_pick_tx(struct net_device *dev, 2105 struct sk_buff *skb, 2106 struct net_device *sb_dev); 2107 2108 /* returns the headroom that the master device needs to take in account 2109 * when forwarding to this dev 2110 */ 2111 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2112 { 2113 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2114 } 2115 2116 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2117 { 2118 if (dev->netdev_ops->ndo_set_rx_headroom) 2119 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2120 } 2121 2122 /* set the device rx headroom to the dev's default */ 2123 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2124 { 2125 netdev_set_rx_headroom(dev, -1); 2126 } 2127 2128 /* 2129 * Net namespace inlines 2130 */ 2131 static inline 2132 struct net *dev_net(const struct net_device *dev) 2133 { 2134 return read_pnet(&dev->nd_net); 2135 } 2136 2137 static inline 2138 void dev_net_set(struct net_device *dev, struct net *net) 2139 { 2140 write_pnet(&dev->nd_net, net); 2141 } 2142 2143 /** 2144 * netdev_priv - access network device private data 2145 * @dev: network device 2146 * 2147 * Get network device private data 2148 */ 2149 static inline void *netdev_priv(const struct net_device *dev) 2150 { 2151 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2152 } 2153 2154 /* Set the sysfs physical device reference for the network logical device 2155 * if set prior to registration will cause a symlink during initialization. 2156 */ 2157 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2158 2159 /* Set the sysfs device type for the network logical device to allow 2160 * fine-grained identification of different network device types. For 2161 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2162 */ 2163 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2164 2165 /* Default NAPI poll() weight 2166 * Device drivers are strongly advised to not use bigger value 2167 */ 2168 #define NAPI_POLL_WEIGHT 64 2169 2170 /** 2171 * netif_napi_add - initialize a NAPI context 2172 * @dev: network device 2173 * @napi: NAPI context 2174 * @poll: polling function 2175 * @weight: default weight 2176 * 2177 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2178 * *any* of the other NAPI-related functions. 2179 */ 2180 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2181 int (*poll)(struct napi_struct *, int), int weight); 2182 2183 /** 2184 * netif_tx_napi_add - initialize a NAPI context 2185 * @dev: network device 2186 * @napi: NAPI context 2187 * @poll: polling function 2188 * @weight: default weight 2189 * 2190 * This variant of netif_napi_add() should be used from drivers using NAPI 2191 * to exclusively poll a TX queue. 2192 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2193 */ 2194 static inline void netif_tx_napi_add(struct net_device *dev, 2195 struct napi_struct *napi, 2196 int (*poll)(struct napi_struct *, int), 2197 int weight) 2198 { 2199 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2200 netif_napi_add(dev, napi, poll, weight); 2201 } 2202 2203 /** 2204 * netif_napi_del - remove a NAPI context 2205 * @napi: NAPI context 2206 * 2207 * netif_napi_del() removes a NAPI context from the network device NAPI list 2208 */ 2209 void netif_napi_del(struct napi_struct *napi); 2210 2211 struct napi_gro_cb { 2212 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2213 void *frag0; 2214 2215 /* Length of frag0. */ 2216 unsigned int frag0_len; 2217 2218 /* This indicates where we are processing relative to skb->data. */ 2219 int data_offset; 2220 2221 /* This is non-zero if the packet cannot be merged with the new skb. */ 2222 u16 flush; 2223 2224 /* Save the IP ID here and check when we get to the transport layer */ 2225 u16 flush_id; 2226 2227 /* Number of segments aggregated. */ 2228 u16 count; 2229 2230 /* Start offset for remote checksum offload */ 2231 u16 gro_remcsum_start; 2232 2233 /* jiffies when first packet was created/queued */ 2234 unsigned long age; 2235 2236 /* Used in ipv6_gro_receive() and foo-over-udp */ 2237 u16 proto; 2238 2239 /* This is non-zero if the packet may be of the same flow. */ 2240 u8 same_flow:1; 2241 2242 /* Used in tunnel GRO receive */ 2243 u8 encap_mark:1; 2244 2245 /* GRO checksum is valid */ 2246 u8 csum_valid:1; 2247 2248 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2249 u8 csum_cnt:3; 2250 2251 /* Free the skb? */ 2252 u8 free:2; 2253 #define NAPI_GRO_FREE 1 2254 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2255 2256 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2257 u8 is_ipv6:1; 2258 2259 /* Used in GRE, set in fou/gue_gro_receive */ 2260 u8 is_fou:1; 2261 2262 /* Used to determine if flush_id can be ignored */ 2263 u8 is_atomic:1; 2264 2265 /* Number of gro_receive callbacks this packet already went through */ 2266 u8 recursion_counter:4; 2267 2268 /* 1 bit hole */ 2269 2270 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2271 __wsum csum; 2272 2273 /* used in skb_gro_receive() slow path */ 2274 struct sk_buff *last; 2275 }; 2276 2277 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2278 2279 #define GRO_RECURSION_LIMIT 15 2280 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2281 { 2282 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2283 } 2284 2285 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2286 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2287 struct list_head *head, 2288 struct sk_buff *skb) 2289 { 2290 if (unlikely(gro_recursion_inc_test(skb))) { 2291 NAPI_GRO_CB(skb)->flush |= 1; 2292 return NULL; 2293 } 2294 2295 return cb(head, skb); 2296 } 2297 2298 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2299 struct sk_buff *); 2300 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2301 struct sock *sk, 2302 struct list_head *head, 2303 struct sk_buff *skb) 2304 { 2305 if (unlikely(gro_recursion_inc_test(skb))) { 2306 NAPI_GRO_CB(skb)->flush |= 1; 2307 return NULL; 2308 } 2309 2310 return cb(sk, head, skb); 2311 } 2312 2313 struct packet_type { 2314 __be16 type; /* This is really htons(ether_type). */ 2315 struct net_device *dev; /* NULL is wildcarded here */ 2316 int (*func) (struct sk_buff *, 2317 struct net_device *, 2318 struct packet_type *, 2319 struct net_device *); 2320 void (*list_func) (struct list_head *, 2321 struct packet_type *, 2322 struct net_device *); 2323 bool (*id_match)(struct packet_type *ptype, 2324 struct sock *sk); 2325 void *af_packet_priv; 2326 struct list_head list; 2327 }; 2328 2329 struct offload_callbacks { 2330 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2331 netdev_features_t features); 2332 struct sk_buff *(*gro_receive)(struct list_head *head, 2333 struct sk_buff *skb); 2334 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2335 }; 2336 2337 struct packet_offload { 2338 __be16 type; /* This is really htons(ether_type). */ 2339 u16 priority; 2340 struct offload_callbacks callbacks; 2341 struct list_head list; 2342 }; 2343 2344 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2345 struct pcpu_sw_netstats { 2346 u64 rx_packets; 2347 u64 rx_bytes; 2348 u64 tx_packets; 2349 u64 tx_bytes; 2350 struct u64_stats_sync syncp; 2351 }; 2352 2353 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2354 ({ \ 2355 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2356 if (pcpu_stats) { \ 2357 int __cpu; \ 2358 for_each_possible_cpu(__cpu) { \ 2359 typeof(type) *stat; \ 2360 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2361 u64_stats_init(&stat->syncp); \ 2362 } \ 2363 } \ 2364 pcpu_stats; \ 2365 }) 2366 2367 #define netdev_alloc_pcpu_stats(type) \ 2368 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2369 2370 enum netdev_lag_tx_type { 2371 NETDEV_LAG_TX_TYPE_UNKNOWN, 2372 NETDEV_LAG_TX_TYPE_RANDOM, 2373 NETDEV_LAG_TX_TYPE_BROADCAST, 2374 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2375 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2376 NETDEV_LAG_TX_TYPE_HASH, 2377 }; 2378 2379 enum netdev_lag_hash { 2380 NETDEV_LAG_HASH_NONE, 2381 NETDEV_LAG_HASH_L2, 2382 NETDEV_LAG_HASH_L34, 2383 NETDEV_LAG_HASH_L23, 2384 NETDEV_LAG_HASH_E23, 2385 NETDEV_LAG_HASH_E34, 2386 NETDEV_LAG_HASH_UNKNOWN, 2387 }; 2388 2389 struct netdev_lag_upper_info { 2390 enum netdev_lag_tx_type tx_type; 2391 enum netdev_lag_hash hash_type; 2392 }; 2393 2394 struct netdev_lag_lower_state_info { 2395 u8 link_up : 1, 2396 tx_enabled : 1; 2397 }; 2398 2399 #include <linux/notifier.h> 2400 2401 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2402 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2403 * adding new types. 2404 */ 2405 enum netdev_cmd { 2406 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2407 NETDEV_DOWN, 2408 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2409 detected a hardware crash and restarted 2410 - we can use this eg to kick tcp sessions 2411 once done */ 2412 NETDEV_CHANGE, /* Notify device state change */ 2413 NETDEV_REGISTER, 2414 NETDEV_UNREGISTER, 2415 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2416 NETDEV_CHANGEADDR, 2417 NETDEV_GOING_DOWN, 2418 NETDEV_CHANGENAME, 2419 NETDEV_FEAT_CHANGE, 2420 NETDEV_BONDING_FAILOVER, 2421 NETDEV_PRE_UP, 2422 NETDEV_PRE_TYPE_CHANGE, 2423 NETDEV_POST_TYPE_CHANGE, 2424 NETDEV_POST_INIT, 2425 NETDEV_RELEASE, 2426 NETDEV_NOTIFY_PEERS, 2427 NETDEV_JOIN, 2428 NETDEV_CHANGEUPPER, 2429 NETDEV_RESEND_IGMP, 2430 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2431 NETDEV_CHANGEINFODATA, 2432 NETDEV_BONDING_INFO, 2433 NETDEV_PRECHANGEUPPER, 2434 NETDEV_CHANGELOWERSTATE, 2435 NETDEV_UDP_TUNNEL_PUSH_INFO, 2436 NETDEV_UDP_TUNNEL_DROP_INFO, 2437 NETDEV_CHANGE_TX_QUEUE_LEN, 2438 NETDEV_CVLAN_FILTER_PUSH_INFO, 2439 NETDEV_CVLAN_FILTER_DROP_INFO, 2440 NETDEV_SVLAN_FILTER_PUSH_INFO, 2441 NETDEV_SVLAN_FILTER_DROP_INFO, 2442 }; 2443 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2444 2445 int register_netdevice_notifier(struct notifier_block *nb); 2446 int unregister_netdevice_notifier(struct notifier_block *nb); 2447 2448 struct netdev_notifier_info { 2449 struct net_device *dev; 2450 struct netlink_ext_ack *extack; 2451 }; 2452 2453 struct netdev_notifier_change_info { 2454 struct netdev_notifier_info info; /* must be first */ 2455 unsigned int flags_changed; 2456 }; 2457 2458 struct netdev_notifier_changeupper_info { 2459 struct netdev_notifier_info info; /* must be first */ 2460 struct net_device *upper_dev; /* new upper dev */ 2461 bool master; /* is upper dev master */ 2462 bool linking; /* is the notification for link or unlink */ 2463 void *upper_info; /* upper dev info */ 2464 }; 2465 2466 struct netdev_notifier_changelowerstate_info { 2467 struct netdev_notifier_info info; /* must be first */ 2468 void *lower_state_info; /* is lower dev state */ 2469 }; 2470 2471 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2472 struct net_device *dev) 2473 { 2474 info->dev = dev; 2475 info->extack = NULL; 2476 } 2477 2478 static inline struct net_device * 2479 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2480 { 2481 return info->dev; 2482 } 2483 2484 static inline struct netlink_ext_ack * 2485 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2486 { 2487 return info->extack; 2488 } 2489 2490 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2491 2492 2493 extern rwlock_t dev_base_lock; /* Device list lock */ 2494 2495 #define for_each_netdev(net, d) \ 2496 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2497 #define for_each_netdev_reverse(net, d) \ 2498 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2499 #define for_each_netdev_rcu(net, d) \ 2500 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2501 #define for_each_netdev_safe(net, d, n) \ 2502 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2503 #define for_each_netdev_continue(net, d) \ 2504 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2505 #define for_each_netdev_continue_rcu(net, d) \ 2506 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2507 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2508 for_each_netdev_rcu(&init_net, slave) \ 2509 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2510 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2511 2512 static inline struct net_device *next_net_device(struct net_device *dev) 2513 { 2514 struct list_head *lh; 2515 struct net *net; 2516 2517 net = dev_net(dev); 2518 lh = dev->dev_list.next; 2519 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2520 } 2521 2522 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2523 { 2524 struct list_head *lh; 2525 struct net *net; 2526 2527 net = dev_net(dev); 2528 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2529 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2530 } 2531 2532 static inline struct net_device *first_net_device(struct net *net) 2533 { 2534 return list_empty(&net->dev_base_head) ? NULL : 2535 net_device_entry(net->dev_base_head.next); 2536 } 2537 2538 static inline struct net_device *first_net_device_rcu(struct net *net) 2539 { 2540 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2541 2542 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2543 } 2544 2545 int netdev_boot_setup_check(struct net_device *dev); 2546 unsigned long netdev_boot_base(const char *prefix, int unit); 2547 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2548 const char *hwaddr); 2549 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2550 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2551 void dev_add_pack(struct packet_type *pt); 2552 void dev_remove_pack(struct packet_type *pt); 2553 void __dev_remove_pack(struct packet_type *pt); 2554 void dev_add_offload(struct packet_offload *po); 2555 void dev_remove_offload(struct packet_offload *po); 2556 2557 int dev_get_iflink(const struct net_device *dev); 2558 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2559 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2560 unsigned short mask); 2561 struct net_device *dev_get_by_name(struct net *net, const char *name); 2562 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2563 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2564 int dev_alloc_name(struct net_device *dev, const char *name); 2565 int dev_open(struct net_device *dev); 2566 void dev_close(struct net_device *dev); 2567 void dev_close_many(struct list_head *head, bool unlink); 2568 void dev_disable_lro(struct net_device *dev); 2569 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2570 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2571 struct net_device *sb_dev, 2572 select_queue_fallback_t fallback); 2573 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2574 struct net_device *sb_dev, 2575 select_queue_fallback_t fallback); 2576 int dev_queue_xmit(struct sk_buff *skb); 2577 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2578 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2579 int register_netdevice(struct net_device *dev); 2580 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2581 void unregister_netdevice_many(struct list_head *head); 2582 static inline void unregister_netdevice(struct net_device *dev) 2583 { 2584 unregister_netdevice_queue(dev, NULL); 2585 } 2586 2587 int netdev_refcnt_read(const struct net_device *dev); 2588 void free_netdev(struct net_device *dev); 2589 void netdev_freemem(struct net_device *dev); 2590 void synchronize_net(void); 2591 int init_dummy_netdev(struct net_device *dev); 2592 2593 DECLARE_PER_CPU(int, xmit_recursion); 2594 #define XMIT_RECURSION_LIMIT 10 2595 2596 static inline int dev_recursion_level(void) 2597 { 2598 return this_cpu_read(xmit_recursion); 2599 } 2600 2601 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2602 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2603 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2604 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2605 int netdev_get_name(struct net *net, char *name, int ifindex); 2606 int dev_restart(struct net_device *dev); 2607 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 2608 2609 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2610 { 2611 return NAPI_GRO_CB(skb)->data_offset; 2612 } 2613 2614 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2615 { 2616 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2617 } 2618 2619 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2620 { 2621 NAPI_GRO_CB(skb)->data_offset += len; 2622 } 2623 2624 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2625 unsigned int offset) 2626 { 2627 return NAPI_GRO_CB(skb)->frag0 + offset; 2628 } 2629 2630 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2631 { 2632 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2633 } 2634 2635 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2636 { 2637 NAPI_GRO_CB(skb)->frag0 = NULL; 2638 NAPI_GRO_CB(skb)->frag0_len = 0; 2639 } 2640 2641 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2642 unsigned int offset) 2643 { 2644 if (!pskb_may_pull(skb, hlen)) 2645 return NULL; 2646 2647 skb_gro_frag0_invalidate(skb); 2648 return skb->data + offset; 2649 } 2650 2651 static inline void *skb_gro_network_header(struct sk_buff *skb) 2652 { 2653 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2654 skb_network_offset(skb); 2655 } 2656 2657 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2658 const void *start, unsigned int len) 2659 { 2660 if (NAPI_GRO_CB(skb)->csum_valid) 2661 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2662 csum_partial(start, len, 0)); 2663 } 2664 2665 /* GRO checksum functions. These are logical equivalents of the normal 2666 * checksum functions (in skbuff.h) except that they operate on the GRO 2667 * offsets and fields in sk_buff. 2668 */ 2669 2670 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2671 2672 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2673 { 2674 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2675 } 2676 2677 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2678 bool zero_okay, 2679 __sum16 check) 2680 { 2681 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2682 skb_checksum_start_offset(skb) < 2683 skb_gro_offset(skb)) && 2684 !skb_at_gro_remcsum_start(skb) && 2685 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2686 (!zero_okay || check)); 2687 } 2688 2689 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2690 __wsum psum) 2691 { 2692 if (NAPI_GRO_CB(skb)->csum_valid && 2693 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2694 return 0; 2695 2696 NAPI_GRO_CB(skb)->csum = psum; 2697 2698 return __skb_gro_checksum_complete(skb); 2699 } 2700 2701 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2702 { 2703 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2704 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2705 NAPI_GRO_CB(skb)->csum_cnt--; 2706 } else { 2707 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2708 * verified a new top level checksum or an encapsulated one 2709 * during GRO. This saves work if we fallback to normal path. 2710 */ 2711 __skb_incr_checksum_unnecessary(skb); 2712 } 2713 } 2714 2715 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2716 compute_pseudo) \ 2717 ({ \ 2718 __sum16 __ret = 0; \ 2719 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2720 __ret = __skb_gro_checksum_validate_complete(skb, \ 2721 compute_pseudo(skb, proto)); \ 2722 if (!__ret) \ 2723 skb_gro_incr_csum_unnecessary(skb); \ 2724 __ret; \ 2725 }) 2726 2727 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2728 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2729 2730 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2731 compute_pseudo) \ 2732 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2733 2734 #define skb_gro_checksum_simple_validate(skb) \ 2735 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2736 2737 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2738 { 2739 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2740 !NAPI_GRO_CB(skb)->csum_valid); 2741 } 2742 2743 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2744 __sum16 check, __wsum pseudo) 2745 { 2746 NAPI_GRO_CB(skb)->csum = ~pseudo; 2747 NAPI_GRO_CB(skb)->csum_valid = 1; 2748 } 2749 2750 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \ 2751 do { \ 2752 if (__skb_gro_checksum_convert_check(skb)) \ 2753 __skb_gro_checksum_convert(skb, check, \ 2754 compute_pseudo(skb, proto)); \ 2755 } while (0) 2756 2757 struct gro_remcsum { 2758 int offset; 2759 __wsum delta; 2760 }; 2761 2762 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2763 { 2764 grc->offset = 0; 2765 grc->delta = 0; 2766 } 2767 2768 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2769 unsigned int off, size_t hdrlen, 2770 int start, int offset, 2771 struct gro_remcsum *grc, 2772 bool nopartial) 2773 { 2774 __wsum delta; 2775 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2776 2777 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2778 2779 if (!nopartial) { 2780 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2781 return ptr; 2782 } 2783 2784 ptr = skb_gro_header_fast(skb, off); 2785 if (skb_gro_header_hard(skb, off + plen)) { 2786 ptr = skb_gro_header_slow(skb, off + plen, off); 2787 if (!ptr) 2788 return NULL; 2789 } 2790 2791 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2792 start, offset); 2793 2794 /* Adjust skb->csum since we changed the packet */ 2795 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2796 2797 grc->offset = off + hdrlen + offset; 2798 grc->delta = delta; 2799 2800 return ptr; 2801 } 2802 2803 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2804 struct gro_remcsum *grc) 2805 { 2806 void *ptr; 2807 size_t plen = grc->offset + sizeof(u16); 2808 2809 if (!grc->delta) 2810 return; 2811 2812 ptr = skb_gro_header_fast(skb, grc->offset); 2813 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 2814 ptr = skb_gro_header_slow(skb, plen, grc->offset); 2815 if (!ptr) 2816 return; 2817 } 2818 2819 remcsum_unadjust((__sum16 *)ptr, grc->delta); 2820 } 2821 2822 #ifdef CONFIG_XFRM_OFFLOAD 2823 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2824 { 2825 if (PTR_ERR(pp) != -EINPROGRESS) 2826 NAPI_GRO_CB(skb)->flush |= flush; 2827 } 2828 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 2829 struct sk_buff *pp, 2830 int flush, 2831 struct gro_remcsum *grc) 2832 { 2833 if (PTR_ERR(pp) != -EINPROGRESS) { 2834 NAPI_GRO_CB(skb)->flush |= flush; 2835 skb_gro_remcsum_cleanup(skb, grc); 2836 skb->remcsum_offload = 0; 2837 } 2838 } 2839 #else 2840 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2841 { 2842 NAPI_GRO_CB(skb)->flush |= flush; 2843 } 2844 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 2845 struct sk_buff *pp, 2846 int flush, 2847 struct gro_remcsum *grc) 2848 { 2849 NAPI_GRO_CB(skb)->flush |= flush; 2850 skb_gro_remcsum_cleanup(skb, grc); 2851 skb->remcsum_offload = 0; 2852 } 2853 #endif 2854 2855 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2856 unsigned short type, 2857 const void *daddr, const void *saddr, 2858 unsigned int len) 2859 { 2860 if (!dev->header_ops || !dev->header_ops->create) 2861 return 0; 2862 2863 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2864 } 2865 2866 static inline int dev_parse_header(const struct sk_buff *skb, 2867 unsigned char *haddr) 2868 { 2869 const struct net_device *dev = skb->dev; 2870 2871 if (!dev->header_ops || !dev->header_ops->parse) 2872 return 0; 2873 return dev->header_ops->parse(skb, haddr); 2874 } 2875 2876 /* ll_header must have at least hard_header_len allocated */ 2877 static inline bool dev_validate_header(const struct net_device *dev, 2878 char *ll_header, int len) 2879 { 2880 if (likely(len >= dev->hard_header_len)) 2881 return true; 2882 if (len < dev->min_header_len) 2883 return false; 2884 2885 if (capable(CAP_SYS_RAWIO)) { 2886 memset(ll_header + len, 0, dev->hard_header_len - len); 2887 return true; 2888 } 2889 2890 if (dev->header_ops && dev->header_ops->validate) 2891 return dev->header_ops->validate(ll_header, len); 2892 2893 return false; 2894 } 2895 2896 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, 2897 int len, int size); 2898 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2899 static inline int unregister_gifconf(unsigned int family) 2900 { 2901 return register_gifconf(family, NULL); 2902 } 2903 2904 #ifdef CONFIG_NET_FLOW_LIMIT 2905 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2906 struct sd_flow_limit { 2907 u64 count; 2908 unsigned int num_buckets; 2909 unsigned int history_head; 2910 u16 history[FLOW_LIMIT_HISTORY]; 2911 u8 buckets[]; 2912 }; 2913 2914 extern int netdev_flow_limit_table_len; 2915 #endif /* CONFIG_NET_FLOW_LIMIT */ 2916 2917 /* 2918 * Incoming packets are placed on per-CPU queues 2919 */ 2920 struct softnet_data { 2921 struct list_head poll_list; 2922 struct sk_buff_head process_queue; 2923 2924 /* stats */ 2925 unsigned int processed; 2926 unsigned int time_squeeze; 2927 unsigned int received_rps; 2928 #ifdef CONFIG_RPS 2929 struct softnet_data *rps_ipi_list; 2930 #endif 2931 #ifdef CONFIG_NET_FLOW_LIMIT 2932 struct sd_flow_limit __rcu *flow_limit; 2933 #endif 2934 struct Qdisc *output_queue; 2935 struct Qdisc **output_queue_tailp; 2936 struct sk_buff *completion_queue; 2937 #ifdef CONFIG_XFRM_OFFLOAD 2938 struct sk_buff_head xfrm_backlog; 2939 #endif 2940 #ifdef CONFIG_RPS 2941 /* input_queue_head should be written by cpu owning this struct, 2942 * and only read by other cpus. Worth using a cache line. 2943 */ 2944 unsigned int input_queue_head ____cacheline_aligned_in_smp; 2945 2946 /* Elements below can be accessed between CPUs for RPS/RFS */ 2947 call_single_data_t csd ____cacheline_aligned_in_smp; 2948 struct softnet_data *rps_ipi_next; 2949 unsigned int cpu; 2950 unsigned int input_queue_tail; 2951 #endif 2952 unsigned int dropped; 2953 struct sk_buff_head input_pkt_queue; 2954 struct napi_struct backlog; 2955 2956 }; 2957 2958 static inline void input_queue_head_incr(struct softnet_data *sd) 2959 { 2960 #ifdef CONFIG_RPS 2961 sd->input_queue_head++; 2962 #endif 2963 } 2964 2965 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 2966 unsigned int *qtail) 2967 { 2968 #ifdef CONFIG_RPS 2969 *qtail = ++sd->input_queue_tail; 2970 #endif 2971 } 2972 2973 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 2974 2975 void __netif_schedule(struct Qdisc *q); 2976 void netif_schedule_queue(struct netdev_queue *txq); 2977 2978 static inline void netif_tx_schedule_all(struct net_device *dev) 2979 { 2980 unsigned int i; 2981 2982 for (i = 0; i < dev->num_tx_queues; i++) 2983 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 2984 } 2985 2986 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 2987 { 2988 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2989 } 2990 2991 /** 2992 * netif_start_queue - allow transmit 2993 * @dev: network device 2994 * 2995 * Allow upper layers to call the device hard_start_xmit routine. 2996 */ 2997 static inline void netif_start_queue(struct net_device *dev) 2998 { 2999 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3000 } 3001 3002 static inline void netif_tx_start_all_queues(struct net_device *dev) 3003 { 3004 unsigned int i; 3005 3006 for (i = 0; i < dev->num_tx_queues; i++) { 3007 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3008 netif_tx_start_queue(txq); 3009 } 3010 } 3011 3012 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3013 3014 /** 3015 * netif_wake_queue - restart transmit 3016 * @dev: network device 3017 * 3018 * Allow upper layers to call the device hard_start_xmit routine. 3019 * Used for flow control when transmit resources are available. 3020 */ 3021 static inline void netif_wake_queue(struct net_device *dev) 3022 { 3023 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3024 } 3025 3026 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3027 { 3028 unsigned int i; 3029 3030 for (i = 0; i < dev->num_tx_queues; i++) { 3031 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3032 netif_tx_wake_queue(txq); 3033 } 3034 } 3035 3036 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3037 { 3038 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3039 } 3040 3041 /** 3042 * netif_stop_queue - stop transmitted packets 3043 * @dev: network device 3044 * 3045 * Stop upper layers calling the device hard_start_xmit routine. 3046 * Used for flow control when transmit resources are unavailable. 3047 */ 3048 static inline void netif_stop_queue(struct net_device *dev) 3049 { 3050 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3051 } 3052 3053 void netif_tx_stop_all_queues(struct net_device *dev); 3054 3055 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3056 { 3057 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3058 } 3059 3060 /** 3061 * netif_queue_stopped - test if transmit queue is flowblocked 3062 * @dev: network device 3063 * 3064 * Test if transmit queue on device is currently unable to send. 3065 */ 3066 static inline bool netif_queue_stopped(const struct net_device *dev) 3067 { 3068 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3069 } 3070 3071 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3072 { 3073 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3074 } 3075 3076 static inline bool 3077 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3078 { 3079 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3080 } 3081 3082 static inline bool 3083 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3084 { 3085 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3086 } 3087 3088 /** 3089 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3090 * @dev_queue: pointer to transmit queue 3091 * 3092 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3093 * to give appropriate hint to the CPU. 3094 */ 3095 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3096 { 3097 #ifdef CONFIG_BQL 3098 prefetchw(&dev_queue->dql.num_queued); 3099 #endif 3100 } 3101 3102 /** 3103 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3104 * @dev_queue: pointer to transmit queue 3105 * 3106 * BQL enabled drivers might use this helper in their TX completion path, 3107 * to give appropriate hint to the CPU. 3108 */ 3109 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3110 { 3111 #ifdef CONFIG_BQL 3112 prefetchw(&dev_queue->dql.limit); 3113 #endif 3114 } 3115 3116 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3117 unsigned int bytes) 3118 { 3119 #ifdef CONFIG_BQL 3120 dql_queued(&dev_queue->dql, bytes); 3121 3122 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3123 return; 3124 3125 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3126 3127 /* 3128 * The XOFF flag must be set before checking the dql_avail below, 3129 * because in netdev_tx_completed_queue we update the dql_completed 3130 * before checking the XOFF flag. 3131 */ 3132 smp_mb(); 3133 3134 /* check again in case another CPU has just made room avail */ 3135 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3136 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3137 #endif 3138 } 3139 3140 /** 3141 * netdev_sent_queue - report the number of bytes queued to hardware 3142 * @dev: network device 3143 * @bytes: number of bytes queued to the hardware device queue 3144 * 3145 * Report the number of bytes queued for sending/completion to the network 3146 * device hardware queue. @bytes should be a good approximation and should 3147 * exactly match netdev_completed_queue() @bytes 3148 */ 3149 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3150 { 3151 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3152 } 3153 3154 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3155 unsigned int pkts, unsigned int bytes) 3156 { 3157 #ifdef CONFIG_BQL 3158 if (unlikely(!bytes)) 3159 return; 3160 3161 dql_completed(&dev_queue->dql, bytes); 3162 3163 /* 3164 * Without the memory barrier there is a small possiblity that 3165 * netdev_tx_sent_queue will miss the update and cause the queue to 3166 * be stopped forever 3167 */ 3168 smp_mb(); 3169 3170 if (dql_avail(&dev_queue->dql) < 0) 3171 return; 3172 3173 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3174 netif_schedule_queue(dev_queue); 3175 #endif 3176 } 3177 3178 /** 3179 * netdev_completed_queue - report bytes and packets completed by device 3180 * @dev: network device 3181 * @pkts: actual number of packets sent over the medium 3182 * @bytes: actual number of bytes sent over the medium 3183 * 3184 * Report the number of bytes and packets transmitted by the network device 3185 * hardware queue over the physical medium, @bytes must exactly match the 3186 * @bytes amount passed to netdev_sent_queue() 3187 */ 3188 static inline void netdev_completed_queue(struct net_device *dev, 3189 unsigned int pkts, unsigned int bytes) 3190 { 3191 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3192 } 3193 3194 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3195 { 3196 #ifdef CONFIG_BQL 3197 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3198 dql_reset(&q->dql); 3199 #endif 3200 } 3201 3202 /** 3203 * netdev_reset_queue - reset the packets and bytes count of a network device 3204 * @dev_queue: network device 3205 * 3206 * Reset the bytes and packet count of a network device and clear the 3207 * software flow control OFF bit for this network device 3208 */ 3209 static inline void netdev_reset_queue(struct net_device *dev_queue) 3210 { 3211 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3212 } 3213 3214 /** 3215 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3216 * @dev: network device 3217 * @queue_index: given tx queue index 3218 * 3219 * Returns 0 if given tx queue index >= number of device tx queues, 3220 * otherwise returns the originally passed tx queue index. 3221 */ 3222 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3223 { 3224 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3225 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3226 dev->name, queue_index, 3227 dev->real_num_tx_queues); 3228 return 0; 3229 } 3230 3231 return queue_index; 3232 } 3233 3234 /** 3235 * netif_running - test if up 3236 * @dev: network device 3237 * 3238 * Test if the device has been brought up. 3239 */ 3240 static inline bool netif_running(const struct net_device *dev) 3241 { 3242 return test_bit(__LINK_STATE_START, &dev->state); 3243 } 3244 3245 /* 3246 * Routines to manage the subqueues on a device. We only need start, 3247 * stop, and a check if it's stopped. All other device management is 3248 * done at the overall netdevice level. 3249 * Also test the device if we're multiqueue. 3250 */ 3251 3252 /** 3253 * netif_start_subqueue - allow sending packets on subqueue 3254 * @dev: network device 3255 * @queue_index: sub queue index 3256 * 3257 * Start individual transmit queue of a device with multiple transmit queues. 3258 */ 3259 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3260 { 3261 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3262 3263 netif_tx_start_queue(txq); 3264 } 3265 3266 /** 3267 * netif_stop_subqueue - stop sending packets on subqueue 3268 * @dev: network device 3269 * @queue_index: sub queue index 3270 * 3271 * Stop individual transmit queue of a device with multiple transmit queues. 3272 */ 3273 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3274 { 3275 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3276 netif_tx_stop_queue(txq); 3277 } 3278 3279 /** 3280 * netif_subqueue_stopped - test status of subqueue 3281 * @dev: network device 3282 * @queue_index: sub queue index 3283 * 3284 * Check individual transmit queue of a device with multiple transmit queues. 3285 */ 3286 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3287 u16 queue_index) 3288 { 3289 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3290 3291 return netif_tx_queue_stopped(txq); 3292 } 3293 3294 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3295 struct sk_buff *skb) 3296 { 3297 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3298 } 3299 3300 /** 3301 * netif_wake_subqueue - allow sending packets on subqueue 3302 * @dev: network device 3303 * @queue_index: sub queue index 3304 * 3305 * Resume individual transmit queue of a device with multiple transmit queues. 3306 */ 3307 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3308 { 3309 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3310 3311 netif_tx_wake_queue(txq); 3312 } 3313 3314 #ifdef CONFIG_XPS 3315 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3316 u16 index); 3317 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3318 u16 index, bool is_rxqs_map); 3319 3320 /** 3321 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3322 * @j: CPU/Rx queue index 3323 * @mask: bitmask of all cpus/rx queues 3324 * @nr_bits: number of bits in the bitmask 3325 * 3326 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3327 */ 3328 static inline bool netif_attr_test_mask(unsigned long j, 3329 const unsigned long *mask, 3330 unsigned int nr_bits) 3331 { 3332 cpu_max_bits_warn(j, nr_bits); 3333 return test_bit(j, mask); 3334 } 3335 3336 /** 3337 * netif_attr_test_online - Test for online CPU/Rx queue 3338 * @j: CPU/Rx queue index 3339 * @online_mask: bitmask for CPUs/Rx queues that are online 3340 * @nr_bits: number of bits in the bitmask 3341 * 3342 * Returns true if a CPU/Rx queue is online. 3343 */ 3344 static inline bool netif_attr_test_online(unsigned long j, 3345 const unsigned long *online_mask, 3346 unsigned int nr_bits) 3347 { 3348 cpu_max_bits_warn(j, nr_bits); 3349 3350 if (online_mask) 3351 return test_bit(j, online_mask); 3352 3353 return (j < nr_bits); 3354 } 3355 3356 /** 3357 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3358 * @n: CPU/Rx queue index 3359 * @srcp: the cpumask/Rx queue mask pointer 3360 * @nr_bits: number of bits in the bitmask 3361 * 3362 * Returns >= nr_bits if no further CPUs/Rx queues set. 3363 */ 3364 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3365 unsigned int nr_bits) 3366 { 3367 /* -1 is a legal arg here. */ 3368 if (n != -1) 3369 cpu_max_bits_warn(n, nr_bits); 3370 3371 if (srcp) 3372 return find_next_bit(srcp, nr_bits, n + 1); 3373 3374 return n + 1; 3375 } 3376 3377 /** 3378 * netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p 3379 * @n: CPU/Rx queue index 3380 * @src1p: the first CPUs/Rx queues mask pointer 3381 * @src2p: the second CPUs/Rx queues mask pointer 3382 * @nr_bits: number of bits in the bitmask 3383 * 3384 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3385 */ 3386 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3387 const unsigned long *src2p, 3388 unsigned int nr_bits) 3389 { 3390 /* -1 is a legal arg here. */ 3391 if (n != -1) 3392 cpu_max_bits_warn(n, nr_bits); 3393 3394 if (src1p && src2p) 3395 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3396 else if (src1p) 3397 return find_next_bit(src1p, nr_bits, n + 1); 3398 else if (src2p) 3399 return find_next_bit(src2p, nr_bits, n + 1); 3400 3401 return n + 1; 3402 } 3403 #else 3404 static inline int netif_set_xps_queue(struct net_device *dev, 3405 const struct cpumask *mask, 3406 u16 index) 3407 { 3408 return 0; 3409 } 3410 #endif 3411 3412 /** 3413 * netif_is_multiqueue - test if device has multiple transmit queues 3414 * @dev: network device 3415 * 3416 * Check if device has multiple transmit queues 3417 */ 3418 static inline bool netif_is_multiqueue(const struct net_device *dev) 3419 { 3420 return dev->num_tx_queues > 1; 3421 } 3422 3423 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3424 3425 #ifdef CONFIG_SYSFS 3426 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3427 #else 3428 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3429 unsigned int rxq) 3430 { 3431 return 0; 3432 } 3433 #endif 3434 3435 static inline struct netdev_rx_queue * 3436 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3437 { 3438 return dev->_rx + rxq; 3439 } 3440 3441 #ifdef CONFIG_SYSFS 3442 static inline unsigned int get_netdev_rx_queue_index( 3443 struct netdev_rx_queue *queue) 3444 { 3445 struct net_device *dev = queue->dev; 3446 int index = queue - dev->_rx; 3447 3448 BUG_ON(index >= dev->num_rx_queues); 3449 return index; 3450 } 3451 #endif 3452 3453 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3454 int netif_get_num_default_rss_queues(void); 3455 3456 enum skb_free_reason { 3457 SKB_REASON_CONSUMED, 3458 SKB_REASON_DROPPED, 3459 }; 3460 3461 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3462 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3463 3464 /* 3465 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3466 * interrupt context or with hardware interrupts being disabled. 3467 * (in_irq() || irqs_disabled()) 3468 * 3469 * We provide four helpers that can be used in following contexts : 3470 * 3471 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3472 * replacing kfree_skb(skb) 3473 * 3474 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3475 * Typically used in place of consume_skb(skb) in TX completion path 3476 * 3477 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3478 * replacing kfree_skb(skb) 3479 * 3480 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3481 * and consumed a packet. Used in place of consume_skb(skb) 3482 */ 3483 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3484 { 3485 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3486 } 3487 3488 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3489 { 3490 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3491 } 3492 3493 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3494 { 3495 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3496 } 3497 3498 static inline void dev_consume_skb_any(struct sk_buff *skb) 3499 { 3500 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3501 } 3502 3503 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3504 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3505 int netif_rx(struct sk_buff *skb); 3506 int netif_rx_ni(struct sk_buff *skb); 3507 int netif_receive_skb(struct sk_buff *skb); 3508 int netif_receive_skb_core(struct sk_buff *skb); 3509 void netif_receive_skb_list(struct list_head *head); 3510 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3511 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3512 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3513 gro_result_t napi_gro_frags(struct napi_struct *napi); 3514 struct packet_offload *gro_find_receive_by_type(__be16 type); 3515 struct packet_offload *gro_find_complete_by_type(__be16 type); 3516 3517 static inline void napi_free_frags(struct napi_struct *napi) 3518 { 3519 kfree_skb(napi->skb); 3520 napi->skb = NULL; 3521 } 3522 3523 bool netdev_is_rx_handler_busy(struct net_device *dev); 3524 int netdev_rx_handler_register(struct net_device *dev, 3525 rx_handler_func_t *rx_handler, 3526 void *rx_handler_data); 3527 void netdev_rx_handler_unregister(struct net_device *dev); 3528 3529 bool dev_valid_name(const char *name); 3530 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3531 bool *need_copyout); 3532 int dev_ifconf(struct net *net, struct ifconf *, int); 3533 int dev_ethtool(struct net *net, struct ifreq *); 3534 unsigned int dev_get_flags(const struct net_device *); 3535 int __dev_change_flags(struct net_device *, unsigned int flags); 3536 int dev_change_flags(struct net_device *, unsigned int); 3537 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3538 unsigned int gchanges); 3539 int dev_change_name(struct net_device *, const char *); 3540 int dev_set_alias(struct net_device *, const char *, size_t); 3541 int dev_get_alias(const struct net_device *, char *, size_t); 3542 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3543 int __dev_set_mtu(struct net_device *, int); 3544 int dev_set_mtu(struct net_device *, int); 3545 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3546 void dev_set_group(struct net_device *, int); 3547 int dev_set_mac_address(struct net_device *, struct sockaddr *); 3548 int dev_change_carrier(struct net_device *, bool new_carrier); 3549 int dev_get_phys_port_id(struct net_device *dev, 3550 struct netdev_phys_item_id *ppid); 3551 int dev_get_phys_port_name(struct net_device *dev, 3552 char *name, size_t len); 3553 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3554 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3555 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3556 struct netdev_queue *txq, int *ret); 3557 3558 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3559 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3560 int fd, u32 flags); 3561 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op, 3562 enum bpf_netdev_command cmd); 3563 3564 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3565 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3566 bool is_skb_forwardable(const struct net_device *dev, 3567 const struct sk_buff *skb); 3568 3569 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3570 struct sk_buff *skb) 3571 { 3572 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3573 unlikely(!is_skb_forwardable(dev, skb))) { 3574 atomic_long_inc(&dev->rx_dropped); 3575 kfree_skb(skb); 3576 return NET_RX_DROP; 3577 } 3578 3579 skb_scrub_packet(skb, true); 3580 skb->priority = 0; 3581 return 0; 3582 } 3583 3584 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3585 3586 extern int netdev_budget; 3587 extern unsigned int netdev_budget_usecs; 3588 3589 /* Called by rtnetlink.c:rtnl_unlock() */ 3590 void netdev_run_todo(void); 3591 3592 /** 3593 * dev_put - release reference to device 3594 * @dev: network device 3595 * 3596 * Release reference to device to allow it to be freed. 3597 */ 3598 static inline void dev_put(struct net_device *dev) 3599 { 3600 this_cpu_dec(*dev->pcpu_refcnt); 3601 } 3602 3603 /** 3604 * dev_hold - get reference to device 3605 * @dev: network device 3606 * 3607 * Hold reference to device to keep it from being freed. 3608 */ 3609 static inline void dev_hold(struct net_device *dev) 3610 { 3611 this_cpu_inc(*dev->pcpu_refcnt); 3612 } 3613 3614 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3615 * and _off may be called from IRQ context, but it is caller 3616 * who is responsible for serialization of these calls. 3617 * 3618 * The name carrier is inappropriate, these functions should really be 3619 * called netif_lowerlayer_*() because they represent the state of any 3620 * kind of lower layer not just hardware media. 3621 */ 3622 3623 void linkwatch_init_dev(struct net_device *dev); 3624 void linkwatch_fire_event(struct net_device *dev); 3625 void linkwatch_forget_dev(struct net_device *dev); 3626 3627 /** 3628 * netif_carrier_ok - test if carrier present 3629 * @dev: network device 3630 * 3631 * Check if carrier is present on device 3632 */ 3633 static inline bool netif_carrier_ok(const struct net_device *dev) 3634 { 3635 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3636 } 3637 3638 unsigned long dev_trans_start(struct net_device *dev); 3639 3640 void __netdev_watchdog_up(struct net_device *dev); 3641 3642 void netif_carrier_on(struct net_device *dev); 3643 3644 void netif_carrier_off(struct net_device *dev); 3645 3646 /** 3647 * netif_dormant_on - mark device as dormant. 3648 * @dev: network device 3649 * 3650 * Mark device as dormant (as per RFC2863). 3651 * 3652 * The dormant state indicates that the relevant interface is not 3653 * actually in a condition to pass packets (i.e., it is not 'up') but is 3654 * in a "pending" state, waiting for some external event. For "on- 3655 * demand" interfaces, this new state identifies the situation where the 3656 * interface is waiting for events to place it in the up state. 3657 */ 3658 static inline void netif_dormant_on(struct net_device *dev) 3659 { 3660 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3661 linkwatch_fire_event(dev); 3662 } 3663 3664 /** 3665 * netif_dormant_off - set device as not dormant. 3666 * @dev: network device 3667 * 3668 * Device is not in dormant state. 3669 */ 3670 static inline void netif_dormant_off(struct net_device *dev) 3671 { 3672 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3673 linkwatch_fire_event(dev); 3674 } 3675 3676 /** 3677 * netif_dormant - test if device is dormant 3678 * @dev: network device 3679 * 3680 * Check if device is dormant. 3681 */ 3682 static inline bool netif_dormant(const struct net_device *dev) 3683 { 3684 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3685 } 3686 3687 3688 /** 3689 * netif_oper_up - test if device is operational 3690 * @dev: network device 3691 * 3692 * Check if carrier is operational 3693 */ 3694 static inline bool netif_oper_up(const struct net_device *dev) 3695 { 3696 return (dev->operstate == IF_OPER_UP || 3697 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3698 } 3699 3700 /** 3701 * netif_device_present - is device available or removed 3702 * @dev: network device 3703 * 3704 * Check if device has not been removed from system. 3705 */ 3706 static inline bool netif_device_present(struct net_device *dev) 3707 { 3708 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3709 } 3710 3711 void netif_device_detach(struct net_device *dev); 3712 3713 void netif_device_attach(struct net_device *dev); 3714 3715 /* 3716 * Network interface message level settings 3717 */ 3718 3719 enum { 3720 NETIF_MSG_DRV = 0x0001, 3721 NETIF_MSG_PROBE = 0x0002, 3722 NETIF_MSG_LINK = 0x0004, 3723 NETIF_MSG_TIMER = 0x0008, 3724 NETIF_MSG_IFDOWN = 0x0010, 3725 NETIF_MSG_IFUP = 0x0020, 3726 NETIF_MSG_RX_ERR = 0x0040, 3727 NETIF_MSG_TX_ERR = 0x0080, 3728 NETIF_MSG_TX_QUEUED = 0x0100, 3729 NETIF_MSG_INTR = 0x0200, 3730 NETIF_MSG_TX_DONE = 0x0400, 3731 NETIF_MSG_RX_STATUS = 0x0800, 3732 NETIF_MSG_PKTDATA = 0x1000, 3733 NETIF_MSG_HW = 0x2000, 3734 NETIF_MSG_WOL = 0x4000, 3735 }; 3736 3737 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3738 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3739 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3740 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3741 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3742 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3743 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3744 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3745 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3746 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3747 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3748 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3749 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3750 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3751 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3752 3753 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3754 { 3755 /* use default */ 3756 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3757 return default_msg_enable_bits; 3758 if (debug_value == 0) /* no output */ 3759 return 0; 3760 /* set low N bits */ 3761 return (1 << debug_value) - 1; 3762 } 3763 3764 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 3765 { 3766 spin_lock(&txq->_xmit_lock); 3767 txq->xmit_lock_owner = cpu; 3768 } 3769 3770 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 3771 { 3772 __acquire(&txq->_xmit_lock); 3773 return true; 3774 } 3775 3776 static inline void __netif_tx_release(struct netdev_queue *txq) 3777 { 3778 __release(&txq->_xmit_lock); 3779 } 3780 3781 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 3782 { 3783 spin_lock_bh(&txq->_xmit_lock); 3784 txq->xmit_lock_owner = smp_processor_id(); 3785 } 3786 3787 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 3788 { 3789 bool ok = spin_trylock(&txq->_xmit_lock); 3790 if (likely(ok)) 3791 txq->xmit_lock_owner = smp_processor_id(); 3792 return ok; 3793 } 3794 3795 static inline void __netif_tx_unlock(struct netdev_queue *txq) 3796 { 3797 txq->xmit_lock_owner = -1; 3798 spin_unlock(&txq->_xmit_lock); 3799 } 3800 3801 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 3802 { 3803 txq->xmit_lock_owner = -1; 3804 spin_unlock_bh(&txq->_xmit_lock); 3805 } 3806 3807 static inline void txq_trans_update(struct netdev_queue *txq) 3808 { 3809 if (txq->xmit_lock_owner != -1) 3810 txq->trans_start = jiffies; 3811 } 3812 3813 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 3814 static inline void netif_trans_update(struct net_device *dev) 3815 { 3816 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 3817 3818 if (txq->trans_start != jiffies) 3819 txq->trans_start = jiffies; 3820 } 3821 3822 /** 3823 * netif_tx_lock - grab network device transmit lock 3824 * @dev: network device 3825 * 3826 * Get network device transmit lock 3827 */ 3828 static inline void netif_tx_lock(struct net_device *dev) 3829 { 3830 unsigned int i; 3831 int cpu; 3832 3833 spin_lock(&dev->tx_global_lock); 3834 cpu = smp_processor_id(); 3835 for (i = 0; i < dev->num_tx_queues; i++) { 3836 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3837 3838 /* We are the only thread of execution doing a 3839 * freeze, but we have to grab the _xmit_lock in 3840 * order to synchronize with threads which are in 3841 * the ->hard_start_xmit() handler and already 3842 * checked the frozen bit. 3843 */ 3844 __netif_tx_lock(txq, cpu); 3845 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 3846 __netif_tx_unlock(txq); 3847 } 3848 } 3849 3850 static inline void netif_tx_lock_bh(struct net_device *dev) 3851 { 3852 local_bh_disable(); 3853 netif_tx_lock(dev); 3854 } 3855 3856 static inline void netif_tx_unlock(struct net_device *dev) 3857 { 3858 unsigned int i; 3859 3860 for (i = 0; i < dev->num_tx_queues; i++) { 3861 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3862 3863 /* No need to grab the _xmit_lock here. If the 3864 * queue is not stopped for another reason, we 3865 * force a schedule. 3866 */ 3867 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 3868 netif_schedule_queue(txq); 3869 } 3870 spin_unlock(&dev->tx_global_lock); 3871 } 3872 3873 static inline void netif_tx_unlock_bh(struct net_device *dev) 3874 { 3875 netif_tx_unlock(dev); 3876 local_bh_enable(); 3877 } 3878 3879 #define HARD_TX_LOCK(dev, txq, cpu) { \ 3880 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3881 __netif_tx_lock(txq, cpu); \ 3882 } else { \ 3883 __netif_tx_acquire(txq); \ 3884 } \ 3885 } 3886 3887 #define HARD_TX_TRYLOCK(dev, txq) \ 3888 (((dev->features & NETIF_F_LLTX) == 0) ? \ 3889 __netif_tx_trylock(txq) : \ 3890 __netif_tx_acquire(txq)) 3891 3892 #define HARD_TX_UNLOCK(dev, txq) { \ 3893 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3894 __netif_tx_unlock(txq); \ 3895 } else { \ 3896 __netif_tx_release(txq); \ 3897 } \ 3898 } 3899 3900 static inline void netif_tx_disable(struct net_device *dev) 3901 { 3902 unsigned int i; 3903 int cpu; 3904 3905 local_bh_disable(); 3906 cpu = smp_processor_id(); 3907 for (i = 0; i < dev->num_tx_queues; i++) { 3908 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3909 3910 __netif_tx_lock(txq, cpu); 3911 netif_tx_stop_queue(txq); 3912 __netif_tx_unlock(txq); 3913 } 3914 local_bh_enable(); 3915 } 3916 3917 static inline void netif_addr_lock(struct net_device *dev) 3918 { 3919 spin_lock(&dev->addr_list_lock); 3920 } 3921 3922 static inline void netif_addr_lock_nested(struct net_device *dev) 3923 { 3924 int subclass = SINGLE_DEPTH_NESTING; 3925 3926 if (dev->netdev_ops->ndo_get_lock_subclass) 3927 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev); 3928 3929 spin_lock_nested(&dev->addr_list_lock, subclass); 3930 } 3931 3932 static inline void netif_addr_lock_bh(struct net_device *dev) 3933 { 3934 spin_lock_bh(&dev->addr_list_lock); 3935 } 3936 3937 static inline void netif_addr_unlock(struct net_device *dev) 3938 { 3939 spin_unlock(&dev->addr_list_lock); 3940 } 3941 3942 static inline void netif_addr_unlock_bh(struct net_device *dev) 3943 { 3944 spin_unlock_bh(&dev->addr_list_lock); 3945 } 3946 3947 /* 3948 * dev_addrs walker. Should be used only for read access. Call with 3949 * rcu_read_lock held. 3950 */ 3951 #define for_each_dev_addr(dev, ha) \ 3952 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 3953 3954 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 3955 3956 void ether_setup(struct net_device *dev); 3957 3958 /* Support for loadable net-drivers */ 3959 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 3960 unsigned char name_assign_type, 3961 void (*setup)(struct net_device *), 3962 unsigned int txqs, unsigned int rxqs); 3963 int dev_get_valid_name(struct net *net, struct net_device *dev, 3964 const char *name); 3965 3966 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 3967 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 3968 3969 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 3970 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 3971 count) 3972 3973 int register_netdev(struct net_device *dev); 3974 void unregister_netdev(struct net_device *dev); 3975 3976 /* General hardware address lists handling functions */ 3977 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 3978 struct netdev_hw_addr_list *from_list, int addr_len); 3979 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 3980 struct netdev_hw_addr_list *from_list, int addr_len); 3981 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 3982 struct net_device *dev, 3983 int (*sync)(struct net_device *, const unsigned char *), 3984 int (*unsync)(struct net_device *, 3985 const unsigned char *)); 3986 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 3987 struct net_device *dev, 3988 int (*unsync)(struct net_device *, 3989 const unsigned char *)); 3990 void __hw_addr_init(struct netdev_hw_addr_list *list); 3991 3992 /* Functions used for device addresses handling */ 3993 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 3994 unsigned char addr_type); 3995 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 3996 unsigned char addr_type); 3997 void dev_addr_flush(struct net_device *dev); 3998 int dev_addr_init(struct net_device *dev); 3999 4000 /* Functions used for unicast addresses handling */ 4001 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4002 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4003 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4004 int dev_uc_sync(struct net_device *to, struct net_device *from); 4005 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4006 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4007 void dev_uc_flush(struct net_device *dev); 4008 void dev_uc_init(struct net_device *dev); 4009 4010 /** 4011 * __dev_uc_sync - Synchonize device's unicast list 4012 * @dev: device to sync 4013 * @sync: function to call if address should be added 4014 * @unsync: function to call if address should be removed 4015 * 4016 * Add newly added addresses to the interface, and release 4017 * addresses that have been deleted. 4018 */ 4019 static inline int __dev_uc_sync(struct net_device *dev, 4020 int (*sync)(struct net_device *, 4021 const unsigned char *), 4022 int (*unsync)(struct net_device *, 4023 const unsigned char *)) 4024 { 4025 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4026 } 4027 4028 /** 4029 * __dev_uc_unsync - Remove synchronized addresses from device 4030 * @dev: device to sync 4031 * @unsync: function to call if address should be removed 4032 * 4033 * Remove all addresses that were added to the device by dev_uc_sync(). 4034 */ 4035 static inline void __dev_uc_unsync(struct net_device *dev, 4036 int (*unsync)(struct net_device *, 4037 const unsigned char *)) 4038 { 4039 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4040 } 4041 4042 /* Functions used for multicast addresses handling */ 4043 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4044 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4045 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4046 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4047 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4048 int dev_mc_sync(struct net_device *to, struct net_device *from); 4049 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4050 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4051 void dev_mc_flush(struct net_device *dev); 4052 void dev_mc_init(struct net_device *dev); 4053 4054 /** 4055 * __dev_mc_sync - Synchonize device's multicast list 4056 * @dev: device to sync 4057 * @sync: function to call if address should be added 4058 * @unsync: function to call if address should be removed 4059 * 4060 * Add newly added addresses to the interface, and release 4061 * addresses that have been deleted. 4062 */ 4063 static inline int __dev_mc_sync(struct net_device *dev, 4064 int (*sync)(struct net_device *, 4065 const unsigned char *), 4066 int (*unsync)(struct net_device *, 4067 const unsigned char *)) 4068 { 4069 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4070 } 4071 4072 /** 4073 * __dev_mc_unsync - Remove synchronized addresses from device 4074 * @dev: device to sync 4075 * @unsync: function to call if address should be removed 4076 * 4077 * Remove all addresses that were added to the device by dev_mc_sync(). 4078 */ 4079 static inline void __dev_mc_unsync(struct net_device *dev, 4080 int (*unsync)(struct net_device *, 4081 const unsigned char *)) 4082 { 4083 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4084 } 4085 4086 /* Functions used for secondary unicast and multicast support */ 4087 void dev_set_rx_mode(struct net_device *dev); 4088 void __dev_set_rx_mode(struct net_device *dev); 4089 int dev_set_promiscuity(struct net_device *dev, int inc); 4090 int dev_set_allmulti(struct net_device *dev, int inc); 4091 void netdev_state_change(struct net_device *dev); 4092 void netdev_notify_peers(struct net_device *dev); 4093 void netdev_features_change(struct net_device *dev); 4094 /* Load a device via the kmod */ 4095 void dev_load(struct net *net, const char *name); 4096 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4097 struct rtnl_link_stats64 *storage); 4098 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4099 const struct net_device_stats *netdev_stats); 4100 4101 extern int netdev_max_backlog; 4102 extern int netdev_tstamp_prequeue; 4103 extern int weight_p; 4104 extern int dev_weight_rx_bias; 4105 extern int dev_weight_tx_bias; 4106 extern int dev_rx_weight; 4107 extern int dev_tx_weight; 4108 4109 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4110 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4111 struct list_head **iter); 4112 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 4113 struct list_head **iter); 4114 4115 /* iterate through upper list, must be called under RCU read lock */ 4116 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4117 for (iter = &(dev)->adj_list.upper, \ 4118 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4119 updev; \ 4120 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4121 4122 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4123 int (*fn)(struct net_device *upper_dev, 4124 void *data), 4125 void *data); 4126 4127 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4128 struct net_device *upper_dev); 4129 4130 bool netdev_has_any_upper_dev(struct net_device *dev); 4131 4132 void *netdev_lower_get_next_private(struct net_device *dev, 4133 struct list_head **iter); 4134 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4135 struct list_head **iter); 4136 4137 #define netdev_for_each_lower_private(dev, priv, iter) \ 4138 for (iter = (dev)->adj_list.lower.next, \ 4139 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4140 priv; \ 4141 priv = netdev_lower_get_next_private(dev, &(iter))) 4142 4143 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4144 for (iter = &(dev)->adj_list.lower, \ 4145 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4146 priv; \ 4147 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4148 4149 void *netdev_lower_get_next(struct net_device *dev, 4150 struct list_head **iter); 4151 4152 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4153 for (iter = (dev)->adj_list.lower.next, \ 4154 ldev = netdev_lower_get_next(dev, &(iter)); \ 4155 ldev; \ 4156 ldev = netdev_lower_get_next(dev, &(iter))) 4157 4158 struct net_device *netdev_all_lower_get_next(struct net_device *dev, 4159 struct list_head **iter); 4160 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev, 4161 struct list_head **iter); 4162 4163 int netdev_walk_all_lower_dev(struct net_device *dev, 4164 int (*fn)(struct net_device *lower_dev, 4165 void *data), 4166 void *data); 4167 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4168 int (*fn)(struct net_device *lower_dev, 4169 void *data), 4170 void *data); 4171 4172 void *netdev_adjacent_get_private(struct list_head *adj_list); 4173 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4174 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4175 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4176 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4177 struct netlink_ext_ack *extack); 4178 int netdev_master_upper_dev_link(struct net_device *dev, 4179 struct net_device *upper_dev, 4180 void *upper_priv, void *upper_info, 4181 struct netlink_ext_ack *extack); 4182 void netdev_upper_dev_unlink(struct net_device *dev, 4183 struct net_device *upper_dev); 4184 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4185 void *netdev_lower_dev_get_private(struct net_device *dev, 4186 struct net_device *lower_dev); 4187 void netdev_lower_state_changed(struct net_device *lower_dev, 4188 void *lower_state_info); 4189 4190 /* RSS keys are 40 or 52 bytes long */ 4191 #define NETDEV_RSS_KEY_LEN 52 4192 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4193 void netdev_rss_key_fill(void *buffer, size_t len); 4194 4195 int dev_get_nest_level(struct net_device *dev); 4196 int skb_checksum_help(struct sk_buff *skb); 4197 int skb_crc32c_csum_help(struct sk_buff *skb); 4198 int skb_csum_hwoffload_help(struct sk_buff *skb, 4199 const netdev_features_t features); 4200 4201 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4202 netdev_features_t features, bool tx_path); 4203 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4204 netdev_features_t features); 4205 4206 struct netdev_bonding_info { 4207 ifslave slave; 4208 ifbond master; 4209 }; 4210 4211 struct netdev_notifier_bonding_info { 4212 struct netdev_notifier_info info; /* must be first */ 4213 struct netdev_bonding_info bonding_info; 4214 }; 4215 4216 void netdev_bonding_info_change(struct net_device *dev, 4217 struct netdev_bonding_info *bonding_info); 4218 4219 static inline 4220 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4221 { 4222 return __skb_gso_segment(skb, features, true); 4223 } 4224 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4225 4226 static inline bool can_checksum_protocol(netdev_features_t features, 4227 __be16 protocol) 4228 { 4229 if (protocol == htons(ETH_P_FCOE)) 4230 return !!(features & NETIF_F_FCOE_CRC); 4231 4232 /* Assume this is an IP checksum (not SCTP CRC) */ 4233 4234 if (features & NETIF_F_HW_CSUM) { 4235 /* Can checksum everything */ 4236 return true; 4237 } 4238 4239 switch (protocol) { 4240 case htons(ETH_P_IP): 4241 return !!(features & NETIF_F_IP_CSUM); 4242 case htons(ETH_P_IPV6): 4243 return !!(features & NETIF_F_IPV6_CSUM); 4244 default: 4245 return false; 4246 } 4247 } 4248 4249 #ifdef CONFIG_BUG 4250 void netdev_rx_csum_fault(struct net_device *dev); 4251 #else 4252 static inline void netdev_rx_csum_fault(struct net_device *dev) 4253 { 4254 } 4255 #endif 4256 /* rx skb timestamps */ 4257 void net_enable_timestamp(void); 4258 void net_disable_timestamp(void); 4259 4260 #ifdef CONFIG_PROC_FS 4261 int __init dev_proc_init(void); 4262 #else 4263 #define dev_proc_init() 0 4264 #endif 4265 4266 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4267 struct sk_buff *skb, struct net_device *dev, 4268 bool more) 4269 { 4270 skb->xmit_more = more ? 1 : 0; 4271 return ops->ndo_start_xmit(skb, dev); 4272 } 4273 4274 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4275 struct netdev_queue *txq, bool more) 4276 { 4277 const struct net_device_ops *ops = dev->netdev_ops; 4278 int rc; 4279 4280 rc = __netdev_start_xmit(ops, skb, dev, more); 4281 if (rc == NETDEV_TX_OK) 4282 txq_trans_update(txq); 4283 4284 return rc; 4285 } 4286 4287 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4288 const void *ns); 4289 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4290 const void *ns); 4291 4292 static inline int netdev_class_create_file(const struct class_attribute *class_attr) 4293 { 4294 return netdev_class_create_file_ns(class_attr, NULL); 4295 } 4296 4297 static inline void netdev_class_remove_file(const struct class_attribute *class_attr) 4298 { 4299 netdev_class_remove_file_ns(class_attr, NULL); 4300 } 4301 4302 extern const struct kobj_ns_type_operations net_ns_type_operations; 4303 4304 const char *netdev_drivername(const struct net_device *dev); 4305 4306 void linkwatch_run_queue(void); 4307 4308 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4309 netdev_features_t f2) 4310 { 4311 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4312 if (f1 & NETIF_F_HW_CSUM) 4313 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4314 else 4315 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4316 } 4317 4318 return f1 & f2; 4319 } 4320 4321 static inline netdev_features_t netdev_get_wanted_features( 4322 struct net_device *dev) 4323 { 4324 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4325 } 4326 netdev_features_t netdev_increment_features(netdev_features_t all, 4327 netdev_features_t one, netdev_features_t mask); 4328 4329 /* Allow TSO being used on stacked device : 4330 * Performing the GSO segmentation before last device 4331 * is a performance improvement. 4332 */ 4333 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4334 netdev_features_t mask) 4335 { 4336 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4337 } 4338 4339 int __netdev_update_features(struct net_device *dev); 4340 void netdev_update_features(struct net_device *dev); 4341 void netdev_change_features(struct net_device *dev); 4342 4343 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4344 struct net_device *dev); 4345 4346 netdev_features_t passthru_features_check(struct sk_buff *skb, 4347 struct net_device *dev, 4348 netdev_features_t features); 4349 netdev_features_t netif_skb_features(struct sk_buff *skb); 4350 4351 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4352 { 4353 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4354 4355 /* check flags correspondence */ 4356 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4357 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4358 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4359 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4360 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4361 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4362 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4363 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4364 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4365 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4366 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4367 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4368 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4369 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4370 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4371 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4372 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4373 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4374 4375 return (features & feature) == feature; 4376 } 4377 4378 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4379 { 4380 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4381 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4382 } 4383 4384 static inline bool netif_needs_gso(struct sk_buff *skb, 4385 netdev_features_t features) 4386 { 4387 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4388 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4389 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4390 } 4391 4392 static inline void netif_set_gso_max_size(struct net_device *dev, 4393 unsigned int size) 4394 { 4395 dev->gso_max_size = size; 4396 } 4397 4398 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4399 int pulled_hlen, u16 mac_offset, 4400 int mac_len) 4401 { 4402 skb->protocol = protocol; 4403 skb->encapsulation = 1; 4404 skb_push(skb, pulled_hlen); 4405 skb_reset_transport_header(skb); 4406 skb->mac_header = mac_offset; 4407 skb->network_header = skb->mac_header + mac_len; 4408 skb->mac_len = mac_len; 4409 } 4410 4411 static inline bool netif_is_macsec(const struct net_device *dev) 4412 { 4413 return dev->priv_flags & IFF_MACSEC; 4414 } 4415 4416 static inline bool netif_is_macvlan(const struct net_device *dev) 4417 { 4418 return dev->priv_flags & IFF_MACVLAN; 4419 } 4420 4421 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4422 { 4423 return dev->priv_flags & IFF_MACVLAN_PORT; 4424 } 4425 4426 static inline bool netif_is_bond_master(const struct net_device *dev) 4427 { 4428 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4429 } 4430 4431 static inline bool netif_is_bond_slave(const struct net_device *dev) 4432 { 4433 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4434 } 4435 4436 static inline bool netif_supports_nofcs(struct net_device *dev) 4437 { 4438 return dev->priv_flags & IFF_SUPP_NOFCS; 4439 } 4440 4441 static inline bool netif_is_l3_master(const struct net_device *dev) 4442 { 4443 return dev->priv_flags & IFF_L3MDEV_MASTER; 4444 } 4445 4446 static inline bool netif_is_l3_slave(const struct net_device *dev) 4447 { 4448 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4449 } 4450 4451 static inline bool netif_is_bridge_master(const struct net_device *dev) 4452 { 4453 return dev->priv_flags & IFF_EBRIDGE; 4454 } 4455 4456 static inline bool netif_is_bridge_port(const struct net_device *dev) 4457 { 4458 return dev->priv_flags & IFF_BRIDGE_PORT; 4459 } 4460 4461 static inline bool netif_is_ovs_master(const struct net_device *dev) 4462 { 4463 return dev->priv_flags & IFF_OPENVSWITCH; 4464 } 4465 4466 static inline bool netif_is_ovs_port(const struct net_device *dev) 4467 { 4468 return dev->priv_flags & IFF_OVS_DATAPATH; 4469 } 4470 4471 static inline bool netif_is_team_master(const struct net_device *dev) 4472 { 4473 return dev->priv_flags & IFF_TEAM; 4474 } 4475 4476 static inline bool netif_is_team_port(const struct net_device *dev) 4477 { 4478 return dev->priv_flags & IFF_TEAM_PORT; 4479 } 4480 4481 static inline bool netif_is_lag_master(const struct net_device *dev) 4482 { 4483 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4484 } 4485 4486 static inline bool netif_is_lag_port(const struct net_device *dev) 4487 { 4488 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4489 } 4490 4491 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4492 { 4493 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4494 } 4495 4496 static inline bool netif_is_failover(const struct net_device *dev) 4497 { 4498 return dev->priv_flags & IFF_FAILOVER; 4499 } 4500 4501 static inline bool netif_is_failover_slave(const struct net_device *dev) 4502 { 4503 return dev->priv_flags & IFF_FAILOVER_SLAVE; 4504 } 4505 4506 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4507 static inline void netif_keep_dst(struct net_device *dev) 4508 { 4509 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4510 } 4511 4512 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4513 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4514 { 4515 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4516 return dev->priv_flags & IFF_MACSEC; 4517 } 4518 4519 extern struct pernet_operations __net_initdata loopback_net_ops; 4520 4521 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4522 4523 /* netdev_printk helpers, similar to dev_printk */ 4524 4525 static inline const char *netdev_name(const struct net_device *dev) 4526 { 4527 if (!dev->name[0] || strchr(dev->name, '%')) 4528 return "(unnamed net_device)"; 4529 return dev->name; 4530 } 4531 4532 static inline bool netdev_unregistering(const struct net_device *dev) 4533 { 4534 return dev->reg_state == NETREG_UNREGISTERING; 4535 } 4536 4537 static inline const char *netdev_reg_state(const struct net_device *dev) 4538 { 4539 switch (dev->reg_state) { 4540 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4541 case NETREG_REGISTERED: return ""; 4542 case NETREG_UNREGISTERING: return " (unregistering)"; 4543 case NETREG_UNREGISTERED: return " (unregistered)"; 4544 case NETREG_RELEASED: return " (released)"; 4545 case NETREG_DUMMY: return " (dummy)"; 4546 } 4547 4548 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4549 return " (unknown)"; 4550 } 4551 4552 __printf(3, 4) 4553 void netdev_printk(const char *level, const struct net_device *dev, 4554 const char *format, ...); 4555 __printf(2, 3) 4556 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4557 __printf(2, 3) 4558 void netdev_alert(const struct net_device *dev, const char *format, ...); 4559 __printf(2, 3) 4560 void netdev_crit(const struct net_device *dev, const char *format, ...); 4561 __printf(2, 3) 4562 void netdev_err(const struct net_device *dev, const char *format, ...); 4563 __printf(2, 3) 4564 void netdev_warn(const struct net_device *dev, const char *format, ...); 4565 __printf(2, 3) 4566 void netdev_notice(const struct net_device *dev, const char *format, ...); 4567 __printf(2, 3) 4568 void netdev_info(const struct net_device *dev, const char *format, ...); 4569 4570 #define netdev_level_once(level, dev, fmt, ...) \ 4571 do { \ 4572 static bool __print_once __read_mostly; \ 4573 \ 4574 if (!__print_once) { \ 4575 __print_once = true; \ 4576 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 4577 } \ 4578 } while (0) 4579 4580 #define netdev_emerg_once(dev, fmt, ...) \ 4581 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 4582 #define netdev_alert_once(dev, fmt, ...) \ 4583 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 4584 #define netdev_crit_once(dev, fmt, ...) \ 4585 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 4586 #define netdev_err_once(dev, fmt, ...) \ 4587 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 4588 #define netdev_warn_once(dev, fmt, ...) \ 4589 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 4590 #define netdev_notice_once(dev, fmt, ...) \ 4591 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 4592 #define netdev_info_once(dev, fmt, ...) \ 4593 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 4594 4595 #define MODULE_ALIAS_NETDEV(device) \ 4596 MODULE_ALIAS("netdev-" device) 4597 4598 #if defined(CONFIG_DYNAMIC_DEBUG) 4599 #define netdev_dbg(__dev, format, args...) \ 4600 do { \ 4601 dynamic_netdev_dbg(__dev, format, ##args); \ 4602 } while (0) 4603 #elif defined(DEBUG) 4604 #define netdev_dbg(__dev, format, args...) \ 4605 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4606 #else 4607 #define netdev_dbg(__dev, format, args...) \ 4608 ({ \ 4609 if (0) \ 4610 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4611 }) 4612 #endif 4613 4614 #if defined(VERBOSE_DEBUG) 4615 #define netdev_vdbg netdev_dbg 4616 #else 4617 4618 #define netdev_vdbg(dev, format, args...) \ 4619 ({ \ 4620 if (0) \ 4621 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 4622 0; \ 4623 }) 4624 #endif 4625 4626 /* 4627 * netdev_WARN() acts like dev_printk(), but with the key difference 4628 * of using a WARN/WARN_ON to get the message out, including the 4629 * file/line information and a backtrace. 4630 */ 4631 #define netdev_WARN(dev, format, args...) \ 4632 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4633 netdev_reg_state(dev), ##args) 4634 4635 #define netdev_WARN_ONCE(dev, format, args...) \ 4636 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4637 netdev_reg_state(dev), ##args) 4638 4639 /* netif printk helpers, similar to netdev_printk */ 4640 4641 #define netif_printk(priv, type, level, dev, fmt, args...) \ 4642 do { \ 4643 if (netif_msg_##type(priv)) \ 4644 netdev_printk(level, (dev), fmt, ##args); \ 4645 } while (0) 4646 4647 #define netif_level(level, priv, type, dev, fmt, args...) \ 4648 do { \ 4649 if (netif_msg_##type(priv)) \ 4650 netdev_##level(dev, fmt, ##args); \ 4651 } while (0) 4652 4653 #define netif_emerg(priv, type, dev, fmt, args...) \ 4654 netif_level(emerg, priv, type, dev, fmt, ##args) 4655 #define netif_alert(priv, type, dev, fmt, args...) \ 4656 netif_level(alert, priv, type, dev, fmt, ##args) 4657 #define netif_crit(priv, type, dev, fmt, args...) \ 4658 netif_level(crit, priv, type, dev, fmt, ##args) 4659 #define netif_err(priv, type, dev, fmt, args...) \ 4660 netif_level(err, priv, type, dev, fmt, ##args) 4661 #define netif_warn(priv, type, dev, fmt, args...) \ 4662 netif_level(warn, priv, type, dev, fmt, ##args) 4663 #define netif_notice(priv, type, dev, fmt, args...) \ 4664 netif_level(notice, priv, type, dev, fmt, ##args) 4665 #define netif_info(priv, type, dev, fmt, args...) \ 4666 netif_level(info, priv, type, dev, fmt, ##args) 4667 4668 #if defined(CONFIG_DYNAMIC_DEBUG) 4669 #define netif_dbg(priv, type, netdev, format, args...) \ 4670 do { \ 4671 if (netif_msg_##type(priv)) \ 4672 dynamic_netdev_dbg(netdev, format, ##args); \ 4673 } while (0) 4674 #elif defined(DEBUG) 4675 #define netif_dbg(priv, type, dev, format, args...) \ 4676 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 4677 #else 4678 #define netif_dbg(priv, type, dev, format, args...) \ 4679 ({ \ 4680 if (0) \ 4681 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4682 0; \ 4683 }) 4684 #endif 4685 4686 /* if @cond then downgrade to debug, else print at @level */ 4687 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 4688 do { \ 4689 if (cond) \ 4690 netif_dbg(priv, type, netdev, fmt, ##args); \ 4691 else \ 4692 netif_ ## level(priv, type, netdev, fmt, ##args); \ 4693 } while (0) 4694 4695 #if defined(VERBOSE_DEBUG) 4696 #define netif_vdbg netif_dbg 4697 #else 4698 #define netif_vdbg(priv, type, dev, format, args...) \ 4699 ({ \ 4700 if (0) \ 4701 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4702 0; \ 4703 }) 4704 #endif 4705 4706 /* 4707 * The list of packet types we will receive (as opposed to discard) 4708 * and the routines to invoke. 4709 * 4710 * Why 16. Because with 16 the only overlap we get on a hash of the 4711 * low nibble of the protocol value is RARP/SNAP/X.25. 4712 * 4713 * 0800 IP 4714 * 0001 802.3 4715 * 0002 AX.25 4716 * 0004 802.2 4717 * 8035 RARP 4718 * 0005 SNAP 4719 * 0805 X.25 4720 * 0806 ARP 4721 * 8137 IPX 4722 * 0009 Localtalk 4723 * 86DD IPv6 4724 */ 4725 #define PTYPE_HASH_SIZE (16) 4726 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 4727 4728 #endif /* _LINUX_NETDEVICE_H */ 4729