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 * size of gro hash buckets, must less than bit number of 312 * napi_struct::gro_bitmask 313 */ 314 #define GRO_HASH_BUCKETS 8 315 316 /* 317 * Structure for NAPI scheduling similar to tasklet but with weighting 318 */ 319 struct napi_struct { 320 /* The poll_list must only be managed by the entity which 321 * changes the state of the NAPI_STATE_SCHED bit. This means 322 * whoever atomically sets that bit can add this napi_struct 323 * to the per-CPU poll_list, and whoever clears that bit 324 * can remove from the list right before clearing the bit. 325 */ 326 struct list_head poll_list; 327 328 unsigned long state; 329 int weight; 330 unsigned long gro_bitmask; 331 int (*poll)(struct napi_struct *, int); 332 #ifdef CONFIG_NETPOLL 333 int poll_owner; 334 #endif 335 struct net_device *dev; 336 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 337 struct sk_buff *skb; 338 struct hrtimer timer; 339 struct list_head dev_list; 340 struct hlist_node napi_hash_node; 341 unsigned int napi_id; 342 }; 343 344 enum { 345 NAPI_STATE_SCHED, /* Poll is scheduled */ 346 NAPI_STATE_MISSED, /* reschedule a napi */ 347 NAPI_STATE_DISABLE, /* Disable pending */ 348 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 349 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */ 350 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */ 351 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */ 352 }; 353 354 enum { 355 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 356 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 357 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 358 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 359 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED), 360 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 361 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 362 }; 363 364 enum gro_result { 365 GRO_MERGED, 366 GRO_MERGED_FREE, 367 GRO_HELD, 368 GRO_NORMAL, 369 GRO_DROP, 370 GRO_CONSUMED, 371 }; 372 typedef enum gro_result gro_result_t; 373 374 /* 375 * enum rx_handler_result - Possible return values for rx_handlers. 376 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 377 * further. 378 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 379 * case skb->dev was changed by rx_handler. 380 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 381 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 382 * 383 * rx_handlers are functions called from inside __netif_receive_skb(), to do 384 * special processing of the skb, prior to delivery to protocol handlers. 385 * 386 * Currently, a net_device can only have a single rx_handler registered. Trying 387 * to register a second rx_handler will return -EBUSY. 388 * 389 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 390 * To unregister a rx_handler on a net_device, use 391 * netdev_rx_handler_unregister(). 392 * 393 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 394 * do with the skb. 395 * 396 * If the rx_handler consumed the skb in some way, it should return 397 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 398 * the skb to be delivered in some other way. 399 * 400 * If the rx_handler changed skb->dev, to divert the skb to another 401 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 402 * new device will be called if it exists. 403 * 404 * If the rx_handler decides the skb should be ignored, it should return 405 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 406 * are registered on exact device (ptype->dev == skb->dev). 407 * 408 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 409 * delivered, it should return RX_HANDLER_PASS. 410 * 411 * A device without a registered rx_handler will behave as if rx_handler 412 * returned RX_HANDLER_PASS. 413 */ 414 415 enum rx_handler_result { 416 RX_HANDLER_CONSUMED, 417 RX_HANDLER_ANOTHER, 418 RX_HANDLER_EXACT, 419 RX_HANDLER_PASS, 420 }; 421 typedef enum rx_handler_result rx_handler_result_t; 422 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 423 424 void __napi_schedule(struct napi_struct *n); 425 void __napi_schedule_irqoff(struct napi_struct *n); 426 427 static inline bool napi_disable_pending(struct napi_struct *n) 428 { 429 return test_bit(NAPI_STATE_DISABLE, &n->state); 430 } 431 432 bool napi_schedule_prep(struct napi_struct *n); 433 434 /** 435 * napi_schedule - schedule NAPI poll 436 * @n: NAPI context 437 * 438 * Schedule NAPI poll routine to be called if it is not already 439 * running. 440 */ 441 static inline void napi_schedule(struct napi_struct *n) 442 { 443 if (napi_schedule_prep(n)) 444 __napi_schedule(n); 445 } 446 447 /** 448 * napi_schedule_irqoff - schedule NAPI poll 449 * @n: NAPI context 450 * 451 * Variant of napi_schedule(), assuming hard irqs are masked. 452 */ 453 static inline void napi_schedule_irqoff(struct napi_struct *n) 454 { 455 if (napi_schedule_prep(n)) 456 __napi_schedule_irqoff(n); 457 } 458 459 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 460 static inline bool napi_reschedule(struct napi_struct *napi) 461 { 462 if (napi_schedule_prep(napi)) { 463 __napi_schedule(napi); 464 return true; 465 } 466 return false; 467 } 468 469 bool napi_complete_done(struct napi_struct *n, int work_done); 470 /** 471 * napi_complete - NAPI processing complete 472 * @n: NAPI context 473 * 474 * Mark NAPI processing as complete. 475 * Consider using napi_complete_done() instead. 476 * Return false if device should avoid rearming interrupts. 477 */ 478 static inline bool napi_complete(struct napi_struct *n) 479 { 480 return napi_complete_done(n, 0); 481 } 482 483 /** 484 * napi_hash_del - remove a NAPI from global table 485 * @napi: NAPI context 486 * 487 * Warning: caller must observe RCU grace period 488 * before freeing memory containing @napi, if 489 * this function returns true. 490 * Note: core networking stack automatically calls it 491 * from netif_napi_del(). 492 * Drivers might want to call this helper to combine all 493 * the needed RCU grace periods into a single one. 494 */ 495 bool napi_hash_del(struct napi_struct *napi); 496 497 /** 498 * napi_disable - prevent NAPI from scheduling 499 * @n: NAPI context 500 * 501 * Stop NAPI from being scheduled on this context. 502 * Waits till any outstanding processing completes. 503 */ 504 void napi_disable(struct napi_struct *n); 505 506 /** 507 * napi_enable - enable NAPI scheduling 508 * @n: NAPI context 509 * 510 * Resume NAPI from being scheduled on this context. 511 * Must be paired with napi_disable. 512 */ 513 static inline void napi_enable(struct napi_struct *n) 514 { 515 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 516 smp_mb__before_atomic(); 517 clear_bit(NAPI_STATE_SCHED, &n->state); 518 clear_bit(NAPI_STATE_NPSVC, &n->state); 519 } 520 521 /** 522 * napi_synchronize - wait until NAPI is not running 523 * @n: NAPI context 524 * 525 * Wait until NAPI is done being scheduled on this context. 526 * Waits till any outstanding processing completes but 527 * does not disable future activations. 528 */ 529 static inline void napi_synchronize(const struct napi_struct *n) 530 { 531 if (IS_ENABLED(CONFIG_SMP)) 532 while (test_bit(NAPI_STATE_SCHED, &n->state)) 533 msleep(1); 534 else 535 barrier(); 536 } 537 538 /** 539 * napi_if_scheduled_mark_missed - if napi is running, set the 540 * NAPIF_STATE_MISSED 541 * @n: NAPI context 542 * 543 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 544 * NAPI is scheduled. 545 **/ 546 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 547 { 548 unsigned long val, new; 549 550 do { 551 val = READ_ONCE(n->state); 552 if (val & NAPIF_STATE_DISABLE) 553 return true; 554 555 if (!(val & NAPIF_STATE_SCHED)) 556 return false; 557 558 new = val | NAPIF_STATE_MISSED; 559 } while (cmpxchg(&n->state, val, new) != val); 560 561 return true; 562 } 563 564 enum netdev_queue_state_t { 565 __QUEUE_STATE_DRV_XOFF, 566 __QUEUE_STATE_STACK_XOFF, 567 __QUEUE_STATE_FROZEN, 568 }; 569 570 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 571 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 572 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 573 574 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 575 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 576 QUEUE_STATE_FROZEN) 577 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 578 QUEUE_STATE_FROZEN) 579 580 /* 581 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 582 * netif_tx_* functions below are used to manipulate this flag. The 583 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 584 * queue independently. The netif_xmit_*stopped functions below are called 585 * to check if the queue has been stopped by the driver or stack (either 586 * of the XOFF bits are set in the state). Drivers should not need to call 587 * netif_xmit*stopped functions, they should only be using netif_tx_*. 588 */ 589 590 struct netdev_queue { 591 /* 592 * read-mostly part 593 */ 594 struct net_device *dev; 595 struct Qdisc __rcu *qdisc; 596 struct Qdisc *qdisc_sleeping; 597 #ifdef CONFIG_SYSFS 598 struct kobject kobj; 599 #endif 600 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 601 int numa_node; 602 #endif 603 unsigned long tx_maxrate; 604 /* 605 * Number of TX timeouts for this queue 606 * (/sys/class/net/DEV/Q/trans_timeout) 607 */ 608 unsigned long trans_timeout; 609 610 /* Subordinate device that the queue has been assigned to */ 611 struct net_device *sb_dev; 612 #ifdef CONFIG_XDP_SOCKETS 613 struct xdp_umem *umem; 614 #endif 615 /* 616 * write-mostly part 617 */ 618 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 619 int xmit_lock_owner; 620 /* 621 * Time (in jiffies) of last Tx 622 */ 623 unsigned long trans_start; 624 625 unsigned long state; 626 627 #ifdef CONFIG_BQL 628 struct dql dql; 629 #endif 630 } ____cacheline_aligned_in_smp; 631 632 extern int sysctl_fb_tunnels_only_for_init_net; 633 634 static inline bool net_has_fallback_tunnels(const struct net *net) 635 { 636 return net == &init_net || 637 !IS_ENABLED(CONFIG_SYSCTL) || 638 !sysctl_fb_tunnels_only_for_init_net; 639 } 640 641 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 642 { 643 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 644 return q->numa_node; 645 #else 646 return NUMA_NO_NODE; 647 #endif 648 } 649 650 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 651 { 652 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 653 q->numa_node = node; 654 #endif 655 } 656 657 #ifdef CONFIG_RPS 658 /* 659 * This structure holds an RPS map which can be of variable length. The 660 * map is an array of CPUs. 661 */ 662 struct rps_map { 663 unsigned int len; 664 struct rcu_head rcu; 665 u16 cpus[0]; 666 }; 667 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 668 669 /* 670 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 671 * tail pointer for that CPU's input queue at the time of last enqueue, and 672 * a hardware filter index. 673 */ 674 struct rps_dev_flow { 675 u16 cpu; 676 u16 filter; 677 unsigned int last_qtail; 678 }; 679 #define RPS_NO_FILTER 0xffff 680 681 /* 682 * The rps_dev_flow_table structure contains a table of flow mappings. 683 */ 684 struct rps_dev_flow_table { 685 unsigned int mask; 686 struct rcu_head rcu; 687 struct rps_dev_flow flows[0]; 688 }; 689 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 690 ((_num) * sizeof(struct rps_dev_flow))) 691 692 /* 693 * The rps_sock_flow_table contains mappings of flows to the last CPU 694 * on which they were processed by the application (set in recvmsg). 695 * Each entry is a 32bit value. Upper part is the high-order bits 696 * of flow hash, lower part is CPU number. 697 * rps_cpu_mask is used to partition the space, depending on number of 698 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 699 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 700 * meaning we use 32-6=26 bits for the hash. 701 */ 702 struct rps_sock_flow_table { 703 u32 mask; 704 705 u32 ents[0] ____cacheline_aligned_in_smp; 706 }; 707 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 708 709 #define RPS_NO_CPU 0xffff 710 711 extern u32 rps_cpu_mask; 712 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 713 714 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 715 u32 hash) 716 { 717 if (table && hash) { 718 unsigned int index = hash & table->mask; 719 u32 val = hash & ~rps_cpu_mask; 720 721 /* We only give a hint, preemption can change CPU under us */ 722 val |= raw_smp_processor_id(); 723 724 if (table->ents[index] != val) 725 table->ents[index] = val; 726 } 727 } 728 729 #ifdef CONFIG_RFS_ACCEL 730 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 731 u16 filter_id); 732 #endif 733 #endif /* CONFIG_RPS */ 734 735 /* This structure contains an instance of an RX queue. */ 736 struct netdev_rx_queue { 737 #ifdef CONFIG_RPS 738 struct rps_map __rcu *rps_map; 739 struct rps_dev_flow_table __rcu *rps_flow_table; 740 #endif 741 struct kobject kobj; 742 struct net_device *dev; 743 struct xdp_rxq_info xdp_rxq; 744 #ifdef CONFIG_XDP_SOCKETS 745 struct xdp_umem *umem; 746 #endif 747 } ____cacheline_aligned_in_smp; 748 749 /* 750 * RX queue sysfs structures and functions. 751 */ 752 struct rx_queue_attribute { 753 struct attribute attr; 754 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 755 ssize_t (*store)(struct netdev_rx_queue *queue, 756 const char *buf, size_t len); 757 }; 758 759 #ifdef CONFIG_XPS 760 /* 761 * This structure holds an XPS map which can be of variable length. The 762 * map is an array of queues. 763 */ 764 struct xps_map { 765 unsigned int len; 766 unsigned int alloc_len; 767 struct rcu_head rcu; 768 u16 queues[0]; 769 }; 770 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 771 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 772 - sizeof(struct xps_map)) / sizeof(u16)) 773 774 /* 775 * This structure holds all XPS maps for device. Maps are indexed by CPU. 776 */ 777 struct xps_dev_maps { 778 struct rcu_head rcu; 779 struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */ 780 }; 781 782 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 783 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 784 785 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 786 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 787 788 #endif /* CONFIG_XPS */ 789 790 #define TC_MAX_QUEUE 16 791 #define TC_BITMASK 15 792 /* HW offloaded queuing disciplines txq count and offset maps */ 793 struct netdev_tc_txq { 794 u16 count; 795 u16 offset; 796 }; 797 798 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 799 /* 800 * This structure is to hold information about the device 801 * configured to run FCoE protocol stack. 802 */ 803 struct netdev_fcoe_hbainfo { 804 char manufacturer[64]; 805 char serial_number[64]; 806 char hardware_version[64]; 807 char driver_version[64]; 808 char optionrom_version[64]; 809 char firmware_version[64]; 810 char model[256]; 811 char model_description[256]; 812 }; 813 #endif 814 815 #define MAX_PHYS_ITEM_ID_LEN 32 816 817 /* This structure holds a unique identifier to identify some 818 * physical item (port for example) used by a netdevice. 819 */ 820 struct netdev_phys_item_id { 821 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 822 unsigned char id_len; 823 }; 824 825 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 826 struct netdev_phys_item_id *b) 827 { 828 return a->id_len == b->id_len && 829 memcmp(a->id, b->id, a->id_len) == 0; 830 } 831 832 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 833 struct sk_buff *skb, 834 struct net_device *sb_dev); 835 836 enum tc_setup_type { 837 TC_SETUP_QDISC_MQPRIO, 838 TC_SETUP_CLSU32, 839 TC_SETUP_CLSFLOWER, 840 TC_SETUP_CLSMATCHALL, 841 TC_SETUP_CLSBPF, 842 TC_SETUP_BLOCK, 843 TC_SETUP_QDISC_CBS, 844 TC_SETUP_QDISC_RED, 845 TC_SETUP_QDISC_PRIO, 846 TC_SETUP_QDISC_MQ, 847 TC_SETUP_QDISC_ETF, 848 TC_SETUP_ROOT_QDISC, 849 TC_SETUP_QDISC_GRED, 850 }; 851 852 /* These structures hold the attributes of bpf state that are being passed 853 * to the netdevice through the bpf op. 854 */ 855 enum bpf_netdev_command { 856 /* Set or clear a bpf program used in the earliest stages of packet 857 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 858 * is responsible for calling bpf_prog_put on any old progs that are 859 * stored. In case of error, the callee need not release the new prog 860 * reference, but on success it takes ownership and must bpf_prog_put 861 * when it is no longer used. 862 */ 863 XDP_SETUP_PROG, 864 XDP_SETUP_PROG_HW, 865 XDP_QUERY_PROG, 866 XDP_QUERY_PROG_HW, 867 /* BPF program for offload callbacks, invoked at program load time. */ 868 BPF_OFFLOAD_VERIFIER_PREP, 869 BPF_OFFLOAD_TRANSLATE, 870 BPF_OFFLOAD_DESTROY, 871 BPF_OFFLOAD_MAP_ALLOC, 872 BPF_OFFLOAD_MAP_FREE, 873 XDP_QUERY_XSK_UMEM, 874 XDP_SETUP_XSK_UMEM, 875 }; 876 877 struct bpf_prog_offload_ops; 878 struct netlink_ext_ack; 879 struct xdp_umem; 880 881 struct netdev_bpf { 882 enum bpf_netdev_command command; 883 union { 884 /* XDP_SETUP_PROG */ 885 struct { 886 u32 flags; 887 struct bpf_prog *prog; 888 struct netlink_ext_ack *extack; 889 }; 890 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */ 891 struct { 892 u32 prog_id; 893 /* flags with which program was installed */ 894 u32 prog_flags; 895 }; 896 /* BPF_OFFLOAD_VERIFIER_PREP */ 897 struct { 898 struct bpf_prog *prog; 899 const struct bpf_prog_offload_ops *ops; /* callee set */ 900 } verifier; 901 /* BPF_OFFLOAD_TRANSLATE, BPF_OFFLOAD_DESTROY */ 902 struct { 903 struct bpf_prog *prog; 904 } offload; 905 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 906 struct { 907 struct bpf_offloaded_map *offmap; 908 }; 909 /* XDP_QUERY_XSK_UMEM, XDP_SETUP_XSK_UMEM */ 910 struct { 911 struct xdp_umem *umem; /* out for query*/ 912 u16 queue_id; /* in for query */ 913 } xsk; 914 }; 915 }; 916 917 #ifdef CONFIG_XFRM_OFFLOAD 918 struct xfrmdev_ops { 919 int (*xdo_dev_state_add) (struct xfrm_state *x); 920 void (*xdo_dev_state_delete) (struct xfrm_state *x); 921 void (*xdo_dev_state_free) (struct xfrm_state *x); 922 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 923 struct xfrm_state *x); 924 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 925 }; 926 #endif 927 928 #if IS_ENABLED(CONFIG_TLS_DEVICE) 929 enum tls_offload_ctx_dir { 930 TLS_OFFLOAD_CTX_DIR_RX, 931 TLS_OFFLOAD_CTX_DIR_TX, 932 }; 933 934 struct tls_crypto_info; 935 struct tls_context; 936 937 struct tlsdev_ops { 938 int (*tls_dev_add)(struct net_device *netdev, struct sock *sk, 939 enum tls_offload_ctx_dir direction, 940 struct tls_crypto_info *crypto_info, 941 u32 start_offload_tcp_sn); 942 void (*tls_dev_del)(struct net_device *netdev, 943 struct tls_context *ctx, 944 enum tls_offload_ctx_dir direction); 945 void (*tls_dev_resync_rx)(struct net_device *netdev, 946 struct sock *sk, u32 seq, u64 rcd_sn); 947 }; 948 #endif 949 950 struct dev_ifalias { 951 struct rcu_head rcuhead; 952 char ifalias[]; 953 }; 954 955 /* 956 * This structure defines the management hooks for network devices. 957 * The following hooks can be defined; unless noted otherwise, they are 958 * optional and can be filled with a null pointer. 959 * 960 * int (*ndo_init)(struct net_device *dev); 961 * This function is called once when a network device is registered. 962 * The network device can use this for any late stage initialization 963 * or semantic validation. It can fail with an error code which will 964 * be propagated back to register_netdev. 965 * 966 * void (*ndo_uninit)(struct net_device *dev); 967 * This function is called when device is unregistered or when registration 968 * fails. It is not called if init fails. 969 * 970 * int (*ndo_open)(struct net_device *dev); 971 * This function is called when a network device transitions to the up 972 * state. 973 * 974 * int (*ndo_stop)(struct net_device *dev); 975 * This function is called when a network device transitions to the down 976 * state. 977 * 978 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 979 * struct net_device *dev); 980 * Called when a packet needs to be transmitted. 981 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 982 * the queue before that can happen; it's for obsolete devices and weird 983 * corner cases, but the stack really does a non-trivial amount 984 * of useless work if you return NETDEV_TX_BUSY. 985 * Required; cannot be NULL. 986 * 987 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 988 * struct net_device *dev 989 * netdev_features_t features); 990 * Called by core transmit path to determine if device is capable of 991 * performing offload operations on a given packet. This is to give 992 * the device an opportunity to implement any restrictions that cannot 993 * be otherwise expressed by feature flags. The check is called with 994 * the set of features that the stack has calculated and it returns 995 * those the driver believes to be appropriate. 996 * 997 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 998 * struct net_device *sb_dev, 999 * select_queue_fallback_t fallback); 1000 * Called to decide which queue to use when device supports multiple 1001 * transmit queues. 1002 * 1003 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1004 * This function is called to allow device receiver to make 1005 * changes to configuration when multicast or promiscuous is enabled. 1006 * 1007 * void (*ndo_set_rx_mode)(struct net_device *dev); 1008 * This function is called device changes address list filtering. 1009 * If driver handles unicast address filtering, it should set 1010 * IFF_UNICAST_FLT in its priv_flags. 1011 * 1012 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1013 * This function is called when the Media Access Control address 1014 * needs to be changed. If this interface is not defined, the 1015 * MAC address can not be changed. 1016 * 1017 * int (*ndo_validate_addr)(struct net_device *dev); 1018 * Test if Media Access Control address is valid for the device. 1019 * 1020 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1021 * Called when a user requests an ioctl which can't be handled by 1022 * the generic interface code. If not defined ioctls return 1023 * not supported error code. 1024 * 1025 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1026 * Used to set network devices bus interface parameters. This interface 1027 * is retained for legacy reasons; new devices should use the bus 1028 * interface (PCI) for low level management. 1029 * 1030 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1031 * Called when a user wants to change the Maximum Transfer Unit 1032 * of a device. 1033 * 1034 * void (*ndo_tx_timeout)(struct net_device *dev); 1035 * Callback used when the transmitter has not made any progress 1036 * for dev->watchdog ticks. 1037 * 1038 * void (*ndo_get_stats64)(struct net_device *dev, 1039 * struct rtnl_link_stats64 *storage); 1040 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1041 * Called when a user wants to get the network device usage 1042 * statistics. Drivers must do one of the following: 1043 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1044 * rtnl_link_stats64 structure passed by the caller. 1045 * 2. Define @ndo_get_stats to update a net_device_stats structure 1046 * (which should normally be dev->stats) and return a pointer to 1047 * it. The structure may be changed asynchronously only if each 1048 * field is written atomically. 1049 * 3. Update dev->stats asynchronously and atomically, and define 1050 * neither operation. 1051 * 1052 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1053 * Return true if this device supports offload stats of this attr_id. 1054 * 1055 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1056 * void *attr_data) 1057 * Get statistics for offload operations by attr_id. Write it into the 1058 * attr_data pointer. 1059 * 1060 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1061 * If device supports VLAN filtering this function is called when a 1062 * VLAN id is registered. 1063 * 1064 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1065 * If device supports VLAN filtering this function is called when a 1066 * VLAN id is unregistered. 1067 * 1068 * void (*ndo_poll_controller)(struct net_device *dev); 1069 * 1070 * SR-IOV management functions. 1071 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1072 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1073 * u8 qos, __be16 proto); 1074 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1075 * int max_tx_rate); 1076 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1077 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1078 * int (*ndo_get_vf_config)(struct net_device *dev, 1079 * int vf, struct ifla_vf_info *ivf); 1080 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1081 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1082 * struct nlattr *port[]); 1083 * 1084 * Enable or disable the VF ability to query its RSS Redirection Table and 1085 * Hash Key. This is needed since on some devices VF share this information 1086 * with PF and querying it may introduce a theoretical security risk. 1087 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1088 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1089 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1090 * void *type_data); 1091 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1092 * This is always called from the stack with the rtnl lock held and netif 1093 * tx queues stopped. This allows the netdevice to perform queue 1094 * management safely. 1095 * 1096 * Fiber Channel over Ethernet (FCoE) offload functions. 1097 * int (*ndo_fcoe_enable)(struct net_device *dev); 1098 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1099 * so the underlying device can perform whatever needed configuration or 1100 * initialization to support acceleration of FCoE traffic. 1101 * 1102 * int (*ndo_fcoe_disable)(struct net_device *dev); 1103 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1104 * so the underlying device can perform whatever needed clean-ups to 1105 * stop supporting acceleration of FCoE traffic. 1106 * 1107 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1108 * struct scatterlist *sgl, unsigned int sgc); 1109 * Called when the FCoE Initiator wants to initialize an I/O that 1110 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1111 * perform necessary setup and returns 1 to indicate the device is set up 1112 * successfully to perform DDP on this I/O, otherwise this returns 0. 1113 * 1114 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1115 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1116 * indicated by the FC exchange id 'xid', so the underlying device can 1117 * clean up and reuse resources for later DDP requests. 1118 * 1119 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1120 * struct scatterlist *sgl, unsigned int sgc); 1121 * Called when the FCoE Target wants to initialize an I/O that 1122 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1123 * perform necessary setup and returns 1 to indicate the device is set up 1124 * successfully to perform DDP on this I/O, otherwise this returns 0. 1125 * 1126 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1127 * struct netdev_fcoe_hbainfo *hbainfo); 1128 * Called when the FCoE Protocol stack wants information on the underlying 1129 * device. This information is utilized by the FCoE protocol stack to 1130 * register attributes with Fiber Channel management service as per the 1131 * FC-GS Fabric Device Management Information(FDMI) specification. 1132 * 1133 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1134 * Called when the underlying device wants to override default World Wide 1135 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1136 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1137 * protocol stack to use. 1138 * 1139 * RFS acceleration. 1140 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1141 * u16 rxq_index, u32 flow_id); 1142 * Set hardware filter for RFS. rxq_index is the target queue index; 1143 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1144 * Return the filter ID on success, or a negative error code. 1145 * 1146 * Slave management functions (for bridge, bonding, etc). 1147 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1148 * Called to make another netdev an underling. 1149 * 1150 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1151 * Called to release previously enslaved netdev. 1152 * 1153 * Feature/offload setting functions. 1154 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1155 * netdev_features_t features); 1156 * Adjusts the requested feature flags according to device-specific 1157 * constraints, and returns the resulting flags. Must not modify 1158 * the device state. 1159 * 1160 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1161 * Called to update device configuration to new features. Passed 1162 * feature set might be less than what was returned by ndo_fix_features()). 1163 * Must return >0 or -errno if it changed dev->features itself. 1164 * 1165 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1166 * struct net_device *dev, 1167 * const unsigned char *addr, u16 vid, u16 flags) 1168 * Adds an FDB entry to dev for addr. 1169 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1170 * struct net_device *dev, 1171 * const unsigned char *addr, u16 vid) 1172 * Deletes the FDB entry from dev coresponding to addr. 1173 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1174 * struct net_device *dev, struct net_device *filter_dev, 1175 * int *idx) 1176 * Used to add FDB entries to dump requests. Implementers should add 1177 * entries to skb and update idx with the number of entries. 1178 * 1179 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1180 * u16 flags) 1181 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1182 * struct net_device *dev, u32 filter_mask, 1183 * int nlflags) 1184 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1185 * u16 flags); 1186 * 1187 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1188 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1189 * which do not represent real hardware may define this to allow their 1190 * userspace components to manage their virtual carrier state. Devices 1191 * that determine carrier state from physical hardware properties (eg 1192 * network cables) or protocol-dependent mechanisms (eg 1193 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1194 * 1195 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1196 * struct netdev_phys_item_id *ppid); 1197 * Called to get ID of physical port of this device. If driver does 1198 * not implement this, it is assumed that the hw is not able to have 1199 * multiple net devices on single physical port. 1200 * 1201 * void (*ndo_udp_tunnel_add)(struct net_device *dev, 1202 * struct udp_tunnel_info *ti); 1203 * Called by UDP tunnel to notify a driver about the UDP port and socket 1204 * address family that a UDP tunnel is listnening to. It is called only 1205 * when a new port starts listening. The operation is protected by the 1206 * RTNL. 1207 * 1208 * void (*ndo_udp_tunnel_del)(struct net_device *dev, 1209 * struct udp_tunnel_info *ti); 1210 * Called by UDP tunnel to notify the driver about a UDP port and socket 1211 * address family that the UDP tunnel is not listening to anymore. The 1212 * operation is protected by the RTNL. 1213 * 1214 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1215 * struct net_device *dev) 1216 * Called by upper layer devices to accelerate switching or other 1217 * station functionality into hardware. 'pdev is the lowerdev 1218 * to use for the offload and 'dev' is the net device that will 1219 * back the offload. Returns a pointer to the private structure 1220 * the upper layer will maintain. 1221 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1222 * Called by upper layer device to delete the station created 1223 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1224 * the station and priv is the structure returned by the add 1225 * operation. 1226 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1227 * int queue_index, u32 maxrate); 1228 * Called when a user wants to set a max-rate limitation of specific 1229 * TX queue. 1230 * int (*ndo_get_iflink)(const struct net_device *dev); 1231 * Called to get the iflink value of this device. 1232 * void (*ndo_change_proto_down)(struct net_device *dev, 1233 * bool proto_down); 1234 * This function is used to pass protocol port error state information 1235 * to the switch driver. The switch driver can react to the proto_down 1236 * by doing a phys down on the associated switch port. 1237 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1238 * This function is used to get egress tunnel information for given skb. 1239 * This is useful for retrieving outer tunnel header parameters while 1240 * sampling packet. 1241 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1242 * This function is used to specify the headroom that the skb must 1243 * consider when allocation skb during packet reception. Setting 1244 * appropriate rx headroom value allows avoiding skb head copy on 1245 * forward. Setting a negative value resets the rx headroom to the 1246 * default value. 1247 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1248 * This function is used to set or query state related to XDP on the 1249 * netdevice and manage BPF offload. See definition of 1250 * enum bpf_netdev_command for details. 1251 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1252 * u32 flags); 1253 * This function is used to submit @n XDP packets for transmit on a 1254 * netdevice. Returns number of frames successfully transmitted, frames 1255 * that got dropped are freed/returned via xdp_return_frame(). 1256 * Returns negative number, means general error invoking ndo, meaning 1257 * no frames were xmit'ed and core-caller will free all frames. 1258 */ 1259 struct net_device_ops { 1260 int (*ndo_init)(struct net_device *dev); 1261 void (*ndo_uninit)(struct net_device *dev); 1262 int (*ndo_open)(struct net_device *dev); 1263 int (*ndo_stop)(struct net_device *dev); 1264 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1265 struct net_device *dev); 1266 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1267 struct net_device *dev, 1268 netdev_features_t features); 1269 u16 (*ndo_select_queue)(struct net_device *dev, 1270 struct sk_buff *skb, 1271 struct net_device *sb_dev, 1272 select_queue_fallback_t fallback); 1273 void (*ndo_change_rx_flags)(struct net_device *dev, 1274 int flags); 1275 void (*ndo_set_rx_mode)(struct net_device *dev); 1276 int (*ndo_set_mac_address)(struct net_device *dev, 1277 void *addr); 1278 int (*ndo_validate_addr)(struct net_device *dev); 1279 int (*ndo_do_ioctl)(struct net_device *dev, 1280 struct ifreq *ifr, int cmd); 1281 int (*ndo_set_config)(struct net_device *dev, 1282 struct ifmap *map); 1283 int (*ndo_change_mtu)(struct net_device *dev, 1284 int new_mtu); 1285 int (*ndo_neigh_setup)(struct net_device *dev, 1286 struct neigh_parms *); 1287 void (*ndo_tx_timeout) (struct net_device *dev); 1288 1289 void (*ndo_get_stats64)(struct net_device *dev, 1290 struct rtnl_link_stats64 *storage); 1291 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1292 int (*ndo_get_offload_stats)(int attr_id, 1293 const struct net_device *dev, 1294 void *attr_data); 1295 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1296 1297 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1298 __be16 proto, u16 vid); 1299 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1300 __be16 proto, u16 vid); 1301 #ifdef CONFIG_NET_POLL_CONTROLLER 1302 void (*ndo_poll_controller)(struct net_device *dev); 1303 int (*ndo_netpoll_setup)(struct net_device *dev, 1304 struct netpoll_info *info); 1305 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1306 #endif 1307 int (*ndo_set_vf_mac)(struct net_device *dev, 1308 int queue, u8 *mac); 1309 int (*ndo_set_vf_vlan)(struct net_device *dev, 1310 int queue, u16 vlan, 1311 u8 qos, __be16 proto); 1312 int (*ndo_set_vf_rate)(struct net_device *dev, 1313 int vf, int min_tx_rate, 1314 int max_tx_rate); 1315 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1316 int vf, bool setting); 1317 int (*ndo_set_vf_trust)(struct net_device *dev, 1318 int vf, bool setting); 1319 int (*ndo_get_vf_config)(struct net_device *dev, 1320 int vf, 1321 struct ifla_vf_info *ivf); 1322 int (*ndo_set_vf_link_state)(struct net_device *dev, 1323 int vf, int link_state); 1324 int (*ndo_get_vf_stats)(struct net_device *dev, 1325 int vf, 1326 struct ifla_vf_stats 1327 *vf_stats); 1328 int (*ndo_set_vf_port)(struct net_device *dev, 1329 int vf, 1330 struct nlattr *port[]); 1331 int (*ndo_get_vf_port)(struct net_device *dev, 1332 int vf, struct sk_buff *skb); 1333 int (*ndo_set_vf_guid)(struct net_device *dev, 1334 int vf, u64 guid, 1335 int guid_type); 1336 int (*ndo_set_vf_rss_query_en)( 1337 struct net_device *dev, 1338 int vf, bool setting); 1339 int (*ndo_setup_tc)(struct net_device *dev, 1340 enum tc_setup_type type, 1341 void *type_data); 1342 #if IS_ENABLED(CONFIG_FCOE) 1343 int (*ndo_fcoe_enable)(struct net_device *dev); 1344 int (*ndo_fcoe_disable)(struct net_device *dev); 1345 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1346 u16 xid, 1347 struct scatterlist *sgl, 1348 unsigned int sgc); 1349 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1350 u16 xid); 1351 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1352 u16 xid, 1353 struct scatterlist *sgl, 1354 unsigned int sgc); 1355 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1356 struct netdev_fcoe_hbainfo *hbainfo); 1357 #endif 1358 1359 #if IS_ENABLED(CONFIG_LIBFCOE) 1360 #define NETDEV_FCOE_WWNN 0 1361 #define NETDEV_FCOE_WWPN 1 1362 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1363 u64 *wwn, int type); 1364 #endif 1365 1366 #ifdef CONFIG_RFS_ACCEL 1367 int (*ndo_rx_flow_steer)(struct net_device *dev, 1368 const struct sk_buff *skb, 1369 u16 rxq_index, 1370 u32 flow_id); 1371 #endif 1372 int (*ndo_add_slave)(struct net_device *dev, 1373 struct net_device *slave_dev, 1374 struct netlink_ext_ack *extack); 1375 int (*ndo_del_slave)(struct net_device *dev, 1376 struct net_device *slave_dev); 1377 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1378 netdev_features_t features); 1379 int (*ndo_set_features)(struct net_device *dev, 1380 netdev_features_t features); 1381 int (*ndo_neigh_construct)(struct net_device *dev, 1382 struct neighbour *n); 1383 void (*ndo_neigh_destroy)(struct net_device *dev, 1384 struct neighbour *n); 1385 1386 int (*ndo_fdb_add)(struct ndmsg *ndm, 1387 struct nlattr *tb[], 1388 struct net_device *dev, 1389 const unsigned char *addr, 1390 u16 vid, 1391 u16 flags); 1392 int (*ndo_fdb_del)(struct ndmsg *ndm, 1393 struct nlattr *tb[], 1394 struct net_device *dev, 1395 const unsigned char *addr, 1396 u16 vid); 1397 int (*ndo_fdb_dump)(struct sk_buff *skb, 1398 struct netlink_callback *cb, 1399 struct net_device *dev, 1400 struct net_device *filter_dev, 1401 int *idx); 1402 1403 int (*ndo_bridge_setlink)(struct net_device *dev, 1404 struct nlmsghdr *nlh, 1405 u16 flags); 1406 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1407 u32 pid, u32 seq, 1408 struct net_device *dev, 1409 u32 filter_mask, 1410 int nlflags); 1411 int (*ndo_bridge_dellink)(struct net_device *dev, 1412 struct nlmsghdr *nlh, 1413 u16 flags); 1414 int (*ndo_change_carrier)(struct net_device *dev, 1415 bool new_carrier); 1416 int (*ndo_get_phys_port_id)(struct net_device *dev, 1417 struct netdev_phys_item_id *ppid); 1418 int (*ndo_get_phys_port_name)(struct net_device *dev, 1419 char *name, size_t len); 1420 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1421 struct udp_tunnel_info *ti); 1422 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1423 struct udp_tunnel_info *ti); 1424 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1425 struct net_device *dev); 1426 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1427 void *priv); 1428 1429 int (*ndo_get_lock_subclass)(struct net_device *dev); 1430 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1431 int queue_index, 1432 u32 maxrate); 1433 int (*ndo_get_iflink)(const struct net_device *dev); 1434 int (*ndo_change_proto_down)(struct net_device *dev, 1435 bool proto_down); 1436 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1437 struct sk_buff *skb); 1438 void (*ndo_set_rx_headroom)(struct net_device *dev, 1439 int needed_headroom); 1440 int (*ndo_bpf)(struct net_device *dev, 1441 struct netdev_bpf *bpf); 1442 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1443 struct xdp_frame **xdp, 1444 u32 flags); 1445 int (*ndo_xsk_async_xmit)(struct net_device *dev, 1446 u32 queue_id); 1447 }; 1448 1449 /** 1450 * enum net_device_priv_flags - &struct net_device priv_flags 1451 * 1452 * These are the &struct net_device, they are only set internally 1453 * by drivers and used in the kernel. These flags are invisible to 1454 * userspace; this means that the order of these flags can change 1455 * during any kernel release. 1456 * 1457 * You should have a pretty good reason to be extending these flags. 1458 * 1459 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1460 * @IFF_EBRIDGE: Ethernet bridging device 1461 * @IFF_BONDING: bonding master or slave 1462 * @IFF_ISATAP: ISATAP interface (RFC4214) 1463 * @IFF_WAN_HDLC: WAN HDLC device 1464 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1465 * release skb->dst 1466 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1467 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1468 * @IFF_MACVLAN_PORT: device used as macvlan port 1469 * @IFF_BRIDGE_PORT: device used as bridge port 1470 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1471 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1472 * @IFF_UNICAST_FLT: Supports unicast filtering 1473 * @IFF_TEAM_PORT: device used as team port 1474 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1475 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1476 * change when it's running 1477 * @IFF_MACVLAN: Macvlan device 1478 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1479 * underlying stacked devices 1480 * @IFF_L3MDEV_MASTER: device is an L3 master device 1481 * @IFF_NO_QUEUE: device can run without qdisc attached 1482 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1483 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1484 * @IFF_TEAM: device is a team device 1485 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1486 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1487 * entity (i.e. the master device for bridged veth) 1488 * @IFF_MACSEC: device is a MACsec device 1489 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1490 * @IFF_FAILOVER: device is a failover master device 1491 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1492 */ 1493 enum netdev_priv_flags { 1494 IFF_802_1Q_VLAN = 1<<0, 1495 IFF_EBRIDGE = 1<<1, 1496 IFF_BONDING = 1<<2, 1497 IFF_ISATAP = 1<<3, 1498 IFF_WAN_HDLC = 1<<4, 1499 IFF_XMIT_DST_RELEASE = 1<<5, 1500 IFF_DONT_BRIDGE = 1<<6, 1501 IFF_DISABLE_NETPOLL = 1<<7, 1502 IFF_MACVLAN_PORT = 1<<8, 1503 IFF_BRIDGE_PORT = 1<<9, 1504 IFF_OVS_DATAPATH = 1<<10, 1505 IFF_TX_SKB_SHARING = 1<<11, 1506 IFF_UNICAST_FLT = 1<<12, 1507 IFF_TEAM_PORT = 1<<13, 1508 IFF_SUPP_NOFCS = 1<<14, 1509 IFF_LIVE_ADDR_CHANGE = 1<<15, 1510 IFF_MACVLAN = 1<<16, 1511 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1512 IFF_L3MDEV_MASTER = 1<<18, 1513 IFF_NO_QUEUE = 1<<19, 1514 IFF_OPENVSWITCH = 1<<20, 1515 IFF_L3MDEV_SLAVE = 1<<21, 1516 IFF_TEAM = 1<<22, 1517 IFF_RXFH_CONFIGURED = 1<<23, 1518 IFF_PHONY_HEADROOM = 1<<24, 1519 IFF_MACSEC = 1<<25, 1520 IFF_NO_RX_HANDLER = 1<<26, 1521 IFF_FAILOVER = 1<<27, 1522 IFF_FAILOVER_SLAVE = 1<<28, 1523 }; 1524 1525 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1526 #define IFF_EBRIDGE IFF_EBRIDGE 1527 #define IFF_BONDING IFF_BONDING 1528 #define IFF_ISATAP IFF_ISATAP 1529 #define IFF_WAN_HDLC IFF_WAN_HDLC 1530 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1531 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1532 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1533 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1534 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1535 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1536 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1537 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1538 #define IFF_TEAM_PORT IFF_TEAM_PORT 1539 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1540 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1541 #define IFF_MACVLAN IFF_MACVLAN 1542 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1543 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1544 #define IFF_NO_QUEUE IFF_NO_QUEUE 1545 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1546 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1547 #define IFF_TEAM IFF_TEAM 1548 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1549 #define IFF_MACSEC IFF_MACSEC 1550 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1551 #define IFF_FAILOVER IFF_FAILOVER 1552 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1553 1554 /** 1555 * struct net_device - The DEVICE structure. 1556 * 1557 * Actually, this whole structure is a big mistake. It mixes I/O 1558 * data with strictly "high-level" data, and it has to know about 1559 * almost every data structure used in the INET module. 1560 * 1561 * @name: This is the first field of the "visible" part of this structure 1562 * (i.e. as seen by users in the "Space.c" file). It is the name 1563 * of the interface. 1564 * 1565 * @name_hlist: Device name hash chain, please keep it close to name[] 1566 * @ifalias: SNMP alias 1567 * @mem_end: Shared memory end 1568 * @mem_start: Shared memory start 1569 * @base_addr: Device I/O address 1570 * @irq: Device IRQ number 1571 * 1572 * @state: Generic network queuing layer state, see netdev_state_t 1573 * @dev_list: The global list of network devices 1574 * @napi_list: List entry used for polling NAPI devices 1575 * @unreg_list: List entry when we are unregistering the 1576 * device; see the function unregister_netdev 1577 * @close_list: List entry used when we are closing the device 1578 * @ptype_all: Device-specific packet handlers for all protocols 1579 * @ptype_specific: Device-specific, protocol-specific packet handlers 1580 * 1581 * @adj_list: Directly linked devices, like slaves for bonding 1582 * @features: Currently active device features 1583 * @hw_features: User-changeable features 1584 * 1585 * @wanted_features: User-requested features 1586 * @vlan_features: Mask of features inheritable by VLAN devices 1587 * 1588 * @hw_enc_features: Mask of features inherited by encapsulating devices 1589 * This field indicates what encapsulation 1590 * offloads the hardware is capable of doing, 1591 * and drivers will need to set them appropriately. 1592 * 1593 * @mpls_features: Mask of features inheritable by MPLS 1594 * 1595 * @ifindex: interface index 1596 * @group: The group the device belongs to 1597 * 1598 * @stats: Statistics struct, which was left as a legacy, use 1599 * rtnl_link_stats64 instead 1600 * 1601 * @rx_dropped: Dropped packets by core network, 1602 * do not use this in drivers 1603 * @tx_dropped: Dropped packets by core network, 1604 * do not use this in drivers 1605 * @rx_nohandler: nohandler dropped packets by core network on 1606 * inactive devices, do not use this in drivers 1607 * @carrier_up_count: Number of times the carrier has been up 1608 * @carrier_down_count: Number of times the carrier has been down 1609 * 1610 * @wireless_handlers: List of functions to handle Wireless Extensions, 1611 * instead of ioctl, 1612 * see <net/iw_handler.h> for details. 1613 * @wireless_data: Instance data managed by the core of wireless extensions 1614 * 1615 * @netdev_ops: Includes several pointers to callbacks, 1616 * if one wants to override the ndo_*() functions 1617 * @ethtool_ops: Management operations 1618 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1619 * discovery handling. Necessary for e.g. 6LoWPAN. 1620 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1621 * of Layer 2 headers. 1622 * 1623 * @flags: Interface flags (a la BSD) 1624 * @priv_flags: Like 'flags' but invisible to userspace, 1625 * see if.h for the definitions 1626 * @gflags: Global flags ( kept as legacy ) 1627 * @padded: How much padding added by alloc_netdev() 1628 * @operstate: RFC2863 operstate 1629 * @link_mode: Mapping policy to operstate 1630 * @if_port: Selectable AUI, TP, ... 1631 * @dma: DMA channel 1632 * @mtu: Interface MTU value 1633 * @min_mtu: Interface Minimum MTU value 1634 * @max_mtu: Interface Maximum MTU value 1635 * @type: Interface hardware type 1636 * @hard_header_len: Maximum hardware header length. 1637 * @min_header_len: Minimum hardware header length 1638 * 1639 * @needed_headroom: Extra headroom the hardware may need, but not in all 1640 * cases can this be guaranteed 1641 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1642 * cases can this be guaranteed. Some cases also use 1643 * LL_MAX_HEADER instead to allocate the skb 1644 * 1645 * interface address info: 1646 * 1647 * @perm_addr: Permanent hw address 1648 * @addr_assign_type: Hw address assignment type 1649 * @addr_len: Hardware address length 1650 * @neigh_priv_len: Used in neigh_alloc() 1651 * @dev_id: Used to differentiate devices that share 1652 * the same link layer address 1653 * @dev_port: Used to differentiate devices that share 1654 * the same function 1655 * @addr_list_lock: XXX: need comments on this one 1656 * @uc_promisc: Counter that indicates promiscuous mode 1657 * has been enabled due to the need to listen to 1658 * additional unicast addresses in a device that 1659 * does not implement ndo_set_rx_mode() 1660 * @uc: unicast mac addresses 1661 * @mc: multicast mac addresses 1662 * @dev_addrs: list of device hw addresses 1663 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1664 * @promiscuity: Number of times the NIC is told to work in 1665 * promiscuous mode; if it becomes 0 the NIC will 1666 * exit promiscuous mode 1667 * @allmulti: Counter, enables or disables allmulticast mode 1668 * 1669 * @vlan_info: VLAN info 1670 * @dsa_ptr: dsa specific data 1671 * @tipc_ptr: TIPC specific data 1672 * @atalk_ptr: AppleTalk link 1673 * @ip_ptr: IPv4 specific data 1674 * @dn_ptr: DECnet specific data 1675 * @ip6_ptr: IPv6 specific data 1676 * @ax25_ptr: AX.25 specific data 1677 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1678 * 1679 * @dev_addr: Hw address (before bcast, 1680 * because most packets are unicast) 1681 * 1682 * @_rx: Array of RX queues 1683 * @num_rx_queues: Number of RX queues 1684 * allocated at register_netdev() time 1685 * @real_num_rx_queues: Number of RX queues currently active in device 1686 * 1687 * @rx_handler: handler for received packets 1688 * @rx_handler_data: XXX: need comments on this one 1689 * @miniq_ingress: ingress/clsact qdisc specific data for 1690 * ingress processing 1691 * @ingress_queue: XXX: need comments on this one 1692 * @broadcast: hw bcast address 1693 * 1694 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1695 * indexed by RX queue number. Assigned by driver. 1696 * This must only be set if the ndo_rx_flow_steer 1697 * operation is defined 1698 * @index_hlist: Device index hash chain 1699 * 1700 * @_tx: Array of TX queues 1701 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1702 * @real_num_tx_queues: Number of TX queues currently active in device 1703 * @qdisc: Root qdisc from userspace point of view 1704 * @tx_queue_len: Max frames per queue allowed 1705 * @tx_global_lock: XXX: need comments on this one 1706 * 1707 * @xps_maps: XXX: need comments on this one 1708 * @miniq_egress: clsact qdisc specific data for 1709 * egress processing 1710 * @watchdog_timeo: Represents the timeout that is used by 1711 * the watchdog (see dev_watchdog()) 1712 * @watchdog_timer: List of timers 1713 * 1714 * @pcpu_refcnt: Number of references to this device 1715 * @todo_list: Delayed register/unregister 1716 * @link_watch_list: XXX: need comments on this one 1717 * 1718 * @reg_state: Register/unregister state machine 1719 * @dismantle: Device is going to be freed 1720 * @rtnl_link_state: This enum represents the phases of creating 1721 * a new link 1722 * 1723 * @needs_free_netdev: Should unregister perform free_netdev? 1724 * @priv_destructor: Called from unregister 1725 * @npinfo: XXX: need comments on this one 1726 * @nd_net: Network namespace this network device is inside 1727 * 1728 * @ml_priv: Mid-layer private 1729 * @lstats: Loopback statistics 1730 * @tstats: Tunnel statistics 1731 * @dstats: Dummy statistics 1732 * @vstats: Virtual ethernet statistics 1733 * 1734 * @garp_port: GARP 1735 * @mrp_port: MRP 1736 * 1737 * @dev: Class/net/name entry 1738 * @sysfs_groups: Space for optional device, statistics and wireless 1739 * sysfs groups 1740 * 1741 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1742 * @rtnl_link_ops: Rtnl_link_ops 1743 * 1744 * @gso_max_size: Maximum size of generic segmentation offload 1745 * @gso_max_segs: Maximum number of segments that can be passed to the 1746 * NIC for GSO 1747 * 1748 * @dcbnl_ops: Data Center Bridging netlink ops 1749 * @num_tc: Number of traffic classes in the net device 1750 * @tc_to_txq: XXX: need comments on this one 1751 * @prio_tc_map: XXX: need comments on this one 1752 * 1753 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1754 * 1755 * @priomap: XXX: need comments on this one 1756 * @phydev: Physical device may attach itself 1757 * for hardware timestamping 1758 * @sfp_bus: attached &struct sfp_bus structure. 1759 * 1760 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1761 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1762 * 1763 * @proto_down: protocol port state information can be sent to the 1764 * switch driver and used to set the phys state of the 1765 * switch port. 1766 * 1767 * @wol_enabled: Wake-on-LAN is enabled 1768 * 1769 * FIXME: cleanup struct net_device such that network protocol info 1770 * moves out. 1771 */ 1772 1773 struct net_device { 1774 char name[IFNAMSIZ]; 1775 struct hlist_node name_hlist; 1776 struct dev_ifalias __rcu *ifalias; 1777 /* 1778 * I/O specific fields 1779 * FIXME: Merge these and struct ifmap into one 1780 */ 1781 unsigned long mem_end; 1782 unsigned long mem_start; 1783 unsigned long base_addr; 1784 int irq; 1785 1786 /* 1787 * Some hardware also needs these fields (state,dev_list, 1788 * napi_list,unreg_list,close_list) but they are not 1789 * part of the usual set specified in Space.c. 1790 */ 1791 1792 unsigned long state; 1793 1794 struct list_head dev_list; 1795 struct list_head napi_list; 1796 struct list_head unreg_list; 1797 struct list_head close_list; 1798 struct list_head ptype_all; 1799 struct list_head ptype_specific; 1800 1801 struct { 1802 struct list_head upper; 1803 struct list_head lower; 1804 } adj_list; 1805 1806 netdev_features_t features; 1807 netdev_features_t hw_features; 1808 netdev_features_t wanted_features; 1809 netdev_features_t vlan_features; 1810 netdev_features_t hw_enc_features; 1811 netdev_features_t mpls_features; 1812 netdev_features_t gso_partial_features; 1813 1814 int ifindex; 1815 int group; 1816 1817 struct net_device_stats stats; 1818 1819 atomic_long_t rx_dropped; 1820 atomic_long_t tx_dropped; 1821 atomic_long_t rx_nohandler; 1822 1823 /* Stats to monitor link on/off, flapping */ 1824 atomic_t carrier_up_count; 1825 atomic_t carrier_down_count; 1826 1827 #ifdef CONFIG_WIRELESS_EXT 1828 const struct iw_handler_def *wireless_handlers; 1829 struct iw_public_data *wireless_data; 1830 #endif 1831 const struct net_device_ops *netdev_ops; 1832 const struct ethtool_ops *ethtool_ops; 1833 #ifdef CONFIG_NET_SWITCHDEV 1834 const struct switchdev_ops *switchdev_ops; 1835 #endif 1836 #ifdef CONFIG_NET_L3_MASTER_DEV 1837 const struct l3mdev_ops *l3mdev_ops; 1838 #endif 1839 #if IS_ENABLED(CONFIG_IPV6) 1840 const struct ndisc_ops *ndisc_ops; 1841 #endif 1842 1843 #ifdef CONFIG_XFRM_OFFLOAD 1844 const struct xfrmdev_ops *xfrmdev_ops; 1845 #endif 1846 1847 #if IS_ENABLED(CONFIG_TLS_DEVICE) 1848 const struct tlsdev_ops *tlsdev_ops; 1849 #endif 1850 1851 const struct header_ops *header_ops; 1852 1853 unsigned int flags; 1854 unsigned int priv_flags; 1855 1856 unsigned short gflags; 1857 unsigned short padded; 1858 1859 unsigned char operstate; 1860 unsigned char link_mode; 1861 1862 unsigned char if_port; 1863 unsigned char dma; 1864 1865 unsigned int mtu; 1866 unsigned int min_mtu; 1867 unsigned int max_mtu; 1868 unsigned short type; 1869 unsigned short hard_header_len; 1870 unsigned char min_header_len; 1871 1872 unsigned short needed_headroom; 1873 unsigned short needed_tailroom; 1874 1875 /* Interface address info. */ 1876 unsigned char perm_addr[MAX_ADDR_LEN]; 1877 unsigned char addr_assign_type; 1878 unsigned char addr_len; 1879 unsigned short neigh_priv_len; 1880 unsigned short dev_id; 1881 unsigned short dev_port; 1882 spinlock_t addr_list_lock; 1883 unsigned char name_assign_type; 1884 bool uc_promisc; 1885 struct netdev_hw_addr_list uc; 1886 struct netdev_hw_addr_list mc; 1887 struct netdev_hw_addr_list dev_addrs; 1888 1889 #ifdef CONFIG_SYSFS 1890 struct kset *queues_kset; 1891 #endif 1892 unsigned int promiscuity; 1893 unsigned int allmulti; 1894 1895 1896 /* Protocol-specific pointers */ 1897 1898 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1899 struct vlan_info __rcu *vlan_info; 1900 #endif 1901 #if IS_ENABLED(CONFIG_NET_DSA) 1902 struct dsa_port *dsa_ptr; 1903 #endif 1904 #if IS_ENABLED(CONFIG_TIPC) 1905 struct tipc_bearer __rcu *tipc_ptr; 1906 #endif 1907 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 1908 void *atalk_ptr; 1909 #endif 1910 struct in_device __rcu *ip_ptr; 1911 #if IS_ENABLED(CONFIG_DECNET) 1912 struct dn_dev __rcu *dn_ptr; 1913 #endif 1914 struct inet6_dev __rcu *ip6_ptr; 1915 #if IS_ENABLED(CONFIG_AX25) 1916 void *ax25_ptr; 1917 #endif 1918 struct wireless_dev *ieee80211_ptr; 1919 struct wpan_dev *ieee802154_ptr; 1920 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 1921 struct mpls_dev __rcu *mpls_ptr; 1922 #endif 1923 1924 /* 1925 * Cache lines mostly used on receive path (including eth_type_trans()) 1926 */ 1927 /* Interface address info used in eth_type_trans() */ 1928 unsigned char *dev_addr; 1929 1930 struct netdev_rx_queue *_rx; 1931 unsigned int num_rx_queues; 1932 unsigned int real_num_rx_queues; 1933 1934 struct bpf_prog __rcu *xdp_prog; 1935 unsigned long gro_flush_timeout; 1936 rx_handler_func_t __rcu *rx_handler; 1937 void __rcu *rx_handler_data; 1938 1939 #ifdef CONFIG_NET_CLS_ACT 1940 struct mini_Qdisc __rcu *miniq_ingress; 1941 #endif 1942 struct netdev_queue __rcu *ingress_queue; 1943 #ifdef CONFIG_NETFILTER_INGRESS 1944 struct nf_hook_entries __rcu *nf_hooks_ingress; 1945 #endif 1946 1947 unsigned char broadcast[MAX_ADDR_LEN]; 1948 #ifdef CONFIG_RFS_ACCEL 1949 struct cpu_rmap *rx_cpu_rmap; 1950 #endif 1951 struct hlist_node index_hlist; 1952 1953 /* 1954 * Cache lines mostly used on transmit path 1955 */ 1956 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1957 unsigned int num_tx_queues; 1958 unsigned int real_num_tx_queues; 1959 struct Qdisc *qdisc; 1960 #ifdef CONFIG_NET_SCHED 1961 DECLARE_HASHTABLE (qdisc_hash, 4); 1962 #endif 1963 unsigned int tx_queue_len; 1964 spinlock_t tx_global_lock; 1965 int watchdog_timeo; 1966 1967 #ifdef CONFIG_XPS 1968 struct xps_dev_maps __rcu *xps_cpus_map; 1969 struct xps_dev_maps __rcu *xps_rxqs_map; 1970 #endif 1971 #ifdef CONFIG_NET_CLS_ACT 1972 struct mini_Qdisc __rcu *miniq_egress; 1973 #endif 1974 1975 /* These may be needed for future network-power-down code. */ 1976 struct timer_list watchdog_timer; 1977 1978 int __percpu *pcpu_refcnt; 1979 struct list_head todo_list; 1980 1981 struct list_head link_watch_list; 1982 1983 enum { NETREG_UNINITIALIZED=0, 1984 NETREG_REGISTERED, /* completed register_netdevice */ 1985 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1986 NETREG_UNREGISTERED, /* completed unregister todo */ 1987 NETREG_RELEASED, /* called free_netdev */ 1988 NETREG_DUMMY, /* dummy device for NAPI poll */ 1989 } reg_state:8; 1990 1991 bool dismantle; 1992 1993 enum { 1994 RTNL_LINK_INITIALIZED, 1995 RTNL_LINK_INITIALIZING, 1996 } rtnl_link_state:16; 1997 1998 bool needs_free_netdev; 1999 void (*priv_destructor)(struct net_device *dev); 2000 2001 #ifdef CONFIG_NETPOLL 2002 struct netpoll_info __rcu *npinfo; 2003 #endif 2004 2005 possible_net_t nd_net; 2006 2007 /* mid-layer private */ 2008 union { 2009 void *ml_priv; 2010 struct pcpu_lstats __percpu *lstats; 2011 struct pcpu_sw_netstats __percpu *tstats; 2012 struct pcpu_dstats __percpu *dstats; 2013 }; 2014 2015 #if IS_ENABLED(CONFIG_GARP) 2016 struct garp_port __rcu *garp_port; 2017 #endif 2018 #if IS_ENABLED(CONFIG_MRP) 2019 struct mrp_port __rcu *mrp_port; 2020 #endif 2021 2022 struct device dev; 2023 const struct attribute_group *sysfs_groups[4]; 2024 const struct attribute_group *sysfs_rx_queue_group; 2025 2026 const struct rtnl_link_ops *rtnl_link_ops; 2027 2028 /* for setting kernel sock attribute on TCP connection setup */ 2029 #define GSO_MAX_SIZE 65536 2030 unsigned int gso_max_size; 2031 #define GSO_MAX_SEGS 65535 2032 u16 gso_max_segs; 2033 2034 #ifdef CONFIG_DCB 2035 const struct dcbnl_rtnl_ops *dcbnl_ops; 2036 #endif 2037 s16 num_tc; 2038 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2039 u8 prio_tc_map[TC_BITMASK + 1]; 2040 2041 #if IS_ENABLED(CONFIG_FCOE) 2042 unsigned int fcoe_ddp_xid; 2043 #endif 2044 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2045 struct netprio_map __rcu *priomap; 2046 #endif 2047 struct phy_device *phydev; 2048 struct sfp_bus *sfp_bus; 2049 struct lock_class_key *qdisc_tx_busylock; 2050 struct lock_class_key *qdisc_running_key; 2051 bool proto_down; 2052 unsigned wol_enabled:1; 2053 }; 2054 #define to_net_dev(d) container_of(d, struct net_device, dev) 2055 2056 static inline bool netif_elide_gro(const struct net_device *dev) 2057 { 2058 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2059 return true; 2060 return false; 2061 } 2062 2063 #define NETDEV_ALIGN 32 2064 2065 static inline 2066 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2067 { 2068 return dev->prio_tc_map[prio & TC_BITMASK]; 2069 } 2070 2071 static inline 2072 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2073 { 2074 if (tc >= dev->num_tc) 2075 return -EINVAL; 2076 2077 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2078 return 0; 2079 } 2080 2081 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2082 void netdev_reset_tc(struct net_device *dev); 2083 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2084 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2085 2086 static inline 2087 int netdev_get_num_tc(struct net_device *dev) 2088 { 2089 return dev->num_tc; 2090 } 2091 2092 void netdev_unbind_sb_channel(struct net_device *dev, 2093 struct net_device *sb_dev); 2094 int netdev_bind_sb_channel_queue(struct net_device *dev, 2095 struct net_device *sb_dev, 2096 u8 tc, u16 count, u16 offset); 2097 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2098 static inline int netdev_get_sb_channel(struct net_device *dev) 2099 { 2100 return max_t(int, -dev->num_tc, 0); 2101 } 2102 2103 static inline 2104 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2105 unsigned int index) 2106 { 2107 return &dev->_tx[index]; 2108 } 2109 2110 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2111 const struct sk_buff *skb) 2112 { 2113 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2114 } 2115 2116 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2117 void (*f)(struct net_device *, 2118 struct netdev_queue *, 2119 void *), 2120 void *arg) 2121 { 2122 unsigned int i; 2123 2124 for (i = 0; i < dev->num_tx_queues; i++) 2125 f(dev, &dev->_tx[i], arg); 2126 } 2127 2128 #define netdev_lockdep_set_classes(dev) \ 2129 { \ 2130 static struct lock_class_key qdisc_tx_busylock_key; \ 2131 static struct lock_class_key qdisc_running_key; \ 2132 static struct lock_class_key qdisc_xmit_lock_key; \ 2133 static struct lock_class_key dev_addr_list_lock_key; \ 2134 unsigned int i; \ 2135 \ 2136 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2137 (dev)->qdisc_running_key = &qdisc_running_key; \ 2138 lockdep_set_class(&(dev)->addr_list_lock, \ 2139 &dev_addr_list_lock_key); \ 2140 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2141 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2142 &qdisc_xmit_lock_key); \ 2143 } 2144 2145 struct netdev_queue *netdev_pick_tx(struct net_device *dev, 2146 struct sk_buff *skb, 2147 struct net_device *sb_dev); 2148 2149 /* returns the headroom that the master device needs to take in account 2150 * when forwarding to this dev 2151 */ 2152 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2153 { 2154 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2155 } 2156 2157 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2158 { 2159 if (dev->netdev_ops->ndo_set_rx_headroom) 2160 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2161 } 2162 2163 /* set the device rx headroom to the dev's default */ 2164 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2165 { 2166 netdev_set_rx_headroom(dev, -1); 2167 } 2168 2169 /* 2170 * Net namespace inlines 2171 */ 2172 static inline 2173 struct net *dev_net(const struct net_device *dev) 2174 { 2175 return read_pnet(&dev->nd_net); 2176 } 2177 2178 static inline 2179 void dev_net_set(struct net_device *dev, struct net *net) 2180 { 2181 write_pnet(&dev->nd_net, net); 2182 } 2183 2184 /** 2185 * netdev_priv - access network device private data 2186 * @dev: network device 2187 * 2188 * Get network device private data 2189 */ 2190 static inline void *netdev_priv(const struct net_device *dev) 2191 { 2192 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2193 } 2194 2195 /* Set the sysfs physical device reference for the network logical device 2196 * if set prior to registration will cause a symlink during initialization. 2197 */ 2198 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2199 2200 /* Set the sysfs device type for the network logical device to allow 2201 * fine-grained identification of different network device types. For 2202 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2203 */ 2204 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2205 2206 /* Default NAPI poll() weight 2207 * Device drivers are strongly advised to not use bigger value 2208 */ 2209 #define NAPI_POLL_WEIGHT 64 2210 2211 /** 2212 * netif_napi_add - initialize a NAPI context 2213 * @dev: network device 2214 * @napi: NAPI context 2215 * @poll: polling function 2216 * @weight: default weight 2217 * 2218 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2219 * *any* of the other NAPI-related functions. 2220 */ 2221 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2222 int (*poll)(struct napi_struct *, int), int weight); 2223 2224 /** 2225 * netif_tx_napi_add - initialize a NAPI context 2226 * @dev: network device 2227 * @napi: NAPI context 2228 * @poll: polling function 2229 * @weight: default weight 2230 * 2231 * This variant of netif_napi_add() should be used from drivers using NAPI 2232 * to exclusively poll a TX queue. 2233 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2234 */ 2235 static inline void netif_tx_napi_add(struct net_device *dev, 2236 struct napi_struct *napi, 2237 int (*poll)(struct napi_struct *, int), 2238 int weight) 2239 { 2240 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2241 netif_napi_add(dev, napi, poll, weight); 2242 } 2243 2244 /** 2245 * netif_napi_del - remove a NAPI context 2246 * @napi: NAPI context 2247 * 2248 * netif_napi_del() removes a NAPI context from the network device NAPI list 2249 */ 2250 void netif_napi_del(struct napi_struct *napi); 2251 2252 struct napi_gro_cb { 2253 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2254 void *frag0; 2255 2256 /* Length of frag0. */ 2257 unsigned int frag0_len; 2258 2259 /* This indicates where we are processing relative to skb->data. */ 2260 int data_offset; 2261 2262 /* This is non-zero if the packet cannot be merged with the new skb. */ 2263 u16 flush; 2264 2265 /* Save the IP ID here and check when we get to the transport layer */ 2266 u16 flush_id; 2267 2268 /* Number of segments aggregated. */ 2269 u16 count; 2270 2271 /* Start offset for remote checksum offload */ 2272 u16 gro_remcsum_start; 2273 2274 /* jiffies when first packet was created/queued */ 2275 unsigned long age; 2276 2277 /* Used in ipv6_gro_receive() and foo-over-udp */ 2278 u16 proto; 2279 2280 /* This is non-zero if the packet may be of the same flow. */ 2281 u8 same_flow:1; 2282 2283 /* Used in tunnel GRO receive */ 2284 u8 encap_mark:1; 2285 2286 /* GRO checksum is valid */ 2287 u8 csum_valid:1; 2288 2289 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2290 u8 csum_cnt:3; 2291 2292 /* Free the skb? */ 2293 u8 free:2; 2294 #define NAPI_GRO_FREE 1 2295 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2296 2297 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2298 u8 is_ipv6:1; 2299 2300 /* Used in GRE, set in fou/gue_gro_receive */ 2301 u8 is_fou:1; 2302 2303 /* Used to determine if flush_id can be ignored */ 2304 u8 is_atomic:1; 2305 2306 /* Number of gro_receive callbacks this packet already went through */ 2307 u8 recursion_counter:4; 2308 2309 /* 1 bit hole */ 2310 2311 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2312 __wsum csum; 2313 2314 /* used in skb_gro_receive() slow path */ 2315 struct sk_buff *last; 2316 }; 2317 2318 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2319 2320 #define GRO_RECURSION_LIMIT 15 2321 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2322 { 2323 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2324 } 2325 2326 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2327 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2328 struct list_head *head, 2329 struct sk_buff *skb) 2330 { 2331 if (unlikely(gro_recursion_inc_test(skb))) { 2332 NAPI_GRO_CB(skb)->flush |= 1; 2333 return NULL; 2334 } 2335 2336 return cb(head, skb); 2337 } 2338 2339 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2340 struct sk_buff *); 2341 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2342 struct sock *sk, 2343 struct list_head *head, 2344 struct sk_buff *skb) 2345 { 2346 if (unlikely(gro_recursion_inc_test(skb))) { 2347 NAPI_GRO_CB(skb)->flush |= 1; 2348 return NULL; 2349 } 2350 2351 return cb(sk, head, skb); 2352 } 2353 2354 struct packet_type { 2355 __be16 type; /* This is really htons(ether_type). */ 2356 bool ignore_outgoing; 2357 struct net_device *dev; /* NULL is wildcarded here */ 2358 int (*func) (struct sk_buff *, 2359 struct net_device *, 2360 struct packet_type *, 2361 struct net_device *); 2362 void (*list_func) (struct list_head *, 2363 struct packet_type *, 2364 struct net_device *); 2365 bool (*id_match)(struct packet_type *ptype, 2366 struct sock *sk); 2367 void *af_packet_priv; 2368 struct list_head list; 2369 }; 2370 2371 struct offload_callbacks { 2372 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2373 netdev_features_t features); 2374 struct sk_buff *(*gro_receive)(struct list_head *head, 2375 struct sk_buff *skb); 2376 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2377 }; 2378 2379 struct packet_offload { 2380 __be16 type; /* This is really htons(ether_type). */ 2381 u16 priority; 2382 struct offload_callbacks callbacks; 2383 struct list_head list; 2384 }; 2385 2386 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2387 struct pcpu_sw_netstats { 2388 u64 rx_packets; 2389 u64 rx_bytes; 2390 u64 tx_packets; 2391 u64 tx_bytes; 2392 struct u64_stats_sync syncp; 2393 } __aligned(4 * sizeof(u64)); 2394 2395 struct pcpu_lstats { 2396 u64 packets; 2397 u64 bytes; 2398 struct u64_stats_sync syncp; 2399 } __aligned(2 * sizeof(u64)); 2400 2401 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2402 ({ \ 2403 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2404 if (pcpu_stats) { \ 2405 int __cpu; \ 2406 for_each_possible_cpu(__cpu) { \ 2407 typeof(type) *stat; \ 2408 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2409 u64_stats_init(&stat->syncp); \ 2410 } \ 2411 } \ 2412 pcpu_stats; \ 2413 }) 2414 2415 #define netdev_alloc_pcpu_stats(type) \ 2416 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2417 2418 enum netdev_lag_tx_type { 2419 NETDEV_LAG_TX_TYPE_UNKNOWN, 2420 NETDEV_LAG_TX_TYPE_RANDOM, 2421 NETDEV_LAG_TX_TYPE_BROADCAST, 2422 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2423 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2424 NETDEV_LAG_TX_TYPE_HASH, 2425 }; 2426 2427 enum netdev_lag_hash { 2428 NETDEV_LAG_HASH_NONE, 2429 NETDEV_LAG_HASH_L2, 2430 NETDEV_LAG_HASH_L34, 2431 NETDEV_LAG_HASH_L23, 2432 NETDEV_LAG_HASH_E23, 2433 NETDEV_LAG_HASH_E34, 2434 NETDEV_LAG_HASH_UNKNOWN, 2435 }; 2436 2437 struct netdev_lag_upper_info { 2438 enum netdev_lag_tx_type tx_type; 2439 enum netdev_lag_hash hash_type; 2440 }; 2441 2442 struct netdev_lag_lower_state_info { 2443 u8 link_up : 1, 2444 tx_enabled : 1; 2445 }; 2446 2447 #include <linux/notifier.h> 2448 2449 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2450 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2451 * adding new types. 2452 */ 2453 enum netdev_cmd { 2454 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2455 NETDEV_DOWN, 2456 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2457 detected a hardware crash and restarted 2458 - we can use this eg to kick tcp sessions 2459 once done */ 2460 NETDEV_CHANGE, /* Notify device state change */ 2461 NETDEV_REGISTER, 2462 NETDEV_UNREGISTER, 2463 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2464 NETDEV_CHANGEADDR, 2465 NETDEV_GOING_DOWN, 2466 NETDEV_CHANGENAME, 2467 NETDEV_FEAT_CHANGE, 2468 NETDEV_BONDING_FAILOVER, 2469 NETDEV_PRE_UP, 2470 NETDEV_PRE_TYPE_CHANGE, 2471 NETDEV_POST_TYPE_CHANGE, 2472 NETDEV_POST_INIT, 2473 NETDEV_RELEASE, 2474 NETDEV_NOTIFY_PEERS, 2475 NETDEV_JOIN, 2476 NETDEV_CHANGEUPPER, 2477 NETDEV_RESEND_IGMP, 2478 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2479 NETDEV_CHANGEINFODATA, 2480 NETDEV_BONDING_INFO, 2481 NETDEV_PRECHANGEUPPER, 2482 NETDEV_CHANGELOWERSTATE, 2483 NETDEV_UDP_TUNNEL_PUSH_INFO, 2484 NETDEV_UDP_TUNNEL_DROP_INFO, 2485 NETDEV_CHANGE_TX_QUEUE_LEN, 2486 NETDEV_CVLAN_FILTER_PUSH_INFO, 2487 NETDEV_CVLAN_FILTER_DROP_INFO, 2488 NETDEV_SVLAN_FILTER_PUSH_INFO, 2489 NETDEV_SVLAN_FILTER_DROP_INFO, 2490 }; 2491 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2492 2493 int register_netdevice_notifier(struct notifier_block *nb); 2494 int unregister_netdevice_notifier(struct notifier_block *nb); 2495 2496 struct netdev_notifier_info { 2497 struct net_device *dev; 2498 struct netlink_ext_ack *extack; 2499 }; 2500 2501 struct netdev_notifier_info_ext { 2502 struct netdev_notifier_info info; /* must be first */ 2503 union { 2504 u32 mtu; 2505 } ext; 2506 }; 2507 2508 struct netdev_notifier_change_info { 2509 struct netdev_notifier_info info; /* must be first */ 2510 unsigned int flags_changed; 2511 }; 2512 2513 struct netdev_notifier_changeupper_info { 2514 struct netdev_notifier_info info; /* must be first */ 2515 struct net_device *upper_dev; /* new upper dev */ 2516 bool master; /* is upper dev master */ 2517 bool linking; /* is the notification for link or unlink */ 2518 void *upper_info; /* upper dev info */ 2519 }; 2520 2521 struct netdev_notifier_changelowerstate_info { 2522 struct netdev_notifier_info info; /* must be first */ 2523 void *lower_state_info; /* is lower dev state */ 2524 }; 2525 2526 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2527 struct net_device *dev) 2528 { 2529 info->dev = dev; 2530 info->extack = NULL; 2531 } 2532 2533 static inline struct net_device * 2534 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2535 { 2536 return info->dev; 2537 } 2538 2539 static inline struct netlink_ext_ack * 2540 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2541 { 2542 return info->extack; 2543 } 2544 2545 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2546 2547 2548 extern rwlock_t dev_base_lock; /* Device list lock */ 2549 2550 #define for_each_netdev(net, d) \ 2551 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2552 #define for_each_netdev_reverse(net, d) \ 2553 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2554 #define for_each_netdev_rcu(net, d) \ 2555 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2556 #define for_each_netdev_safe(net, d, n) \ 2557 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2558 #define for_each_netdev_continue(net, d) \ 2559 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2560 #define for_each_netdev_continue_rcu(net, d) \ 2561 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2562 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2563 for_each_netdev_rcu(&init_net, slave) \ 2564 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2565 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2566 2567 static inline struct net_device *next_net_device(struct net_device *dev) 2568 { 2569 struct list_head *lh; 2570 struct net *net; 2571 2572 net = dev_net(dev); 2573 lh = dev->dev_list.next; 2574 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2575 } 2576 2577 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2578 { 2579 struct list_head *lh; 2580 struct net *net; 2581 2582 net = dev_net(dev); 2583 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2584 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2585 } 2586 2587 static inline struct net_device *first_net_device(struct net *net) 2588 { 2589 return list_empty(&net->dev_base_head) ? NULL : 2590 net_device_entry(net->dev_base_head.next); 2591 } 2592 2593 static inline struct net_device *first_net_device_rcu(struct net *net) 2594 { 2595 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2596 2597 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2598 } 2599 2600 int netdev_boot_setup_check(struct net_device *dev); 2601 unsigned long netdev_boot_base(const char *prefix, int unit); 2602 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2603 const char *hwaddr); 2604 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2605 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2606 void dev_add_pack(struct packet_type *pt); 2607 void dev_remove_pack(struct packet_type *pt); 2608 void __dev_remove_pack(struct packet_type *pt); 2609 void dev_add_offload(struct packet_offload *po); 2610 void dev_remove_offload(struct packet_offload *po); 2611 2612 int dev_get_iflink(const struct net_device *dev); 2613 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2614 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2615 unsigned short mask); 2616 struct net_device *dev_get_by_name(struct net *net, const char *name); 2617 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2618 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2619 int dev_alloc_name(struct net_device *dev, const char *name); 2620 int dev_open(struct net_device *dev); 2621 void dev_close(struct net_device *dev); 2622 void dev_close_many(struct list_head *head, bool unlink); 2623 void dev_disable_lro(struct net_device *dev); 2624 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2625 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2626 struct net_device *sb_dev, 2627 select_queue_fallback_t fallback); 2628 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2629 struct net_device *sb_dev, 2630 select_queue_fallback_t fallback); 2631 int dev_queue_xmit(struct sk_buff *skb); 2632 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2633 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2634 int register_netdevice(struct net_device *dev); 2635 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2636 void unregister_netdevice_many(struct list_head *head); 2637 static inline void unregister_netdevice(struct net_device *dev) 2638 { 2639 unregister_netdevice_queue(dev, NULL); 2640 } 2641 2642 int netdev_refcnt_read(const struct net_device *dev); 2643 void free_netdev(struct net_device *dev); 2644 void netdev_freemem(struct net_device *dev); 2645 void synchronize_net(void); 2646 int init_dummy_netdev(struct net_device *dev); 2647 2648 DECLARE_PER_CPU(int, xmit_recursion); 2649 #define XMIT_RECURSION_LIMIT 10 2650 2651 static inline int dev_recursion_level(void) 2652 { 2653 return this_cpu_read(xmit_recursion); 2654 } 2655 2656 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2657 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2658 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2659 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2660 int netdev_get_name(struct net *net, char *name, int ifindex); 2661 int dev_restart(struct net_device *dev); 2662 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 2663 2664 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2665 { 2666 return NAPI_GRO_CB(skb)->data_offset; 2667 } 2668 2669 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2670 { 2671 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2672 } 2673 2674 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2675 { 2676 NAPI_GRO_CB(skb)->data_offset += len; 2677 } 2678 2679 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2680 unsigned int offset) 2681 { 2682 return NAPI_GRO_CB(skb)->frag0 + offset; 2683 } 2684 2685 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2686 { 2687 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2688 } 2689 2690 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2691 { 2692 NAPI_GRO_CB(skb)->frag0 = NULL; 2693 NAPI_GRO_CB(skb)->frag0_len = 0; 2694 } 2695 2696 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2697 unsigned int offset) 2698 { 2699 if (!pskb_may_pull(skb, hlen)) 2700 return NULL; 2701 2702 skb_gro_frag0_invalidate(skb); 2703 return skb->data + offset; 2704 } 2705 2706 static inline void *skb_gro_network_header(struct sk_buff *skb) 2707 { 2708 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2709 skb_network_offset(skb); 2710 } 2711 2712 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2713 const void *start, unsigned int len) 2714 { 2715 if (NAPI_GRO_CB(skb)->csum_valid) 2716 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2717 csum_partial(start, len, 0)); 2718 } 2719 2720 /* GRO checksum functions. These are logical equivalents of the normal 2721 * checksum functions (in skbuff.h) except that they operate on the GRO 2722 * offsets and fields in sk_buff. 2723 */ 2724 2725 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2726 2727 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2728 { 2729 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2730 } 2731 2732 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2733 bool zero_okay, 2734 __sum16 check) 2735 { 2736 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2737 skb_checksum_start_offset(skb) < 2738 skb_gro_offset(skb)) && 2739 !skb_at_gro_remcsum_start(skb) && 2740 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2741 (!zero_okay || check)); 2742 } 2743 2744 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2745 __wsum psum) 2746 { 2747 if (NAPI_GRO_CB(skb)->csum_valid && 2748 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2749 return 0; 2750 2751 NAPI_GRO_CB(skb)->csum = psum; 2752 2753 return __skb_gro_checksum_complete(skb); 2754 } 2755 2756 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2757 { 2758 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2759 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2760 NAPI_GRO_CB(skb)->csum_cnt--; 2761 } else { 2762 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2763 * verified a new top level checksum or an encapsulated one 2764 * during GRO. This saves work if we fallback to normal path. 2765 */ 2766 __skb_incr_checksum_unnecessary(skb); 2767 } 2768 } 2769 2770 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2771 compute_pseudo) \ 2772 ({ \ 2773 __sum16 __ret = 0; \ 2774 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2775 __ret = __skb_gro_checksum_validate_complete(skb, \ 2776 compute_pseudo(skb, proto)); \ 2777 if (!__ret) \ 2778 skb_gro_incr_csum_unnecessary(skb); \ 2779 __ret; \ 2780 }) 2781 2782 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2783 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2784 2785 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2786 compute_pseudo) \ 2787 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2788 2789 #define skb_gro_checksum_simple_validate(skb) \ 2790 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2791 2792 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2793 { 2794 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2795 !NAPI_GRO_CB(skb)->csum_valid); 2796 } 2797 2798 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2799 __sum16 check, __wsum pseudo) 2800 { 2801 NAPI_GRO_CB(skb)->csum = ~pseudo; 2802 NAPI_GRO_CB(skb)->csum_valid = 1; 2803 } 2804 2805 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \ 2806 do { \ 2807 if (__skb_gro_checksum_convert_check(skb)) \ 2808 __skb_gro_checksum_convert(skb, check, \ 2809 compute_pseudo(skb, proto)); \ 2810 } while (0) 2811 2812 struct gro_remcsum { 2813 int offset; 2814 __wsum delta; 2815 }; 2816 2817 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2818 { 2819 grc->offset = 0; 2820 grc->delta = 0; 2821 } 2822 2823 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2824 unsigned int off, size_t hdrlen, 2825 int start, int offset, 2826 struct gro_remcsum *grc, 2827 bool nopartial) 2828 { 2829 __wsum delta; 2830 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2831 2832 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2833 2834 if (!nopartial) { 2835 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2836 return ptr; 2837 } 2838 2839 ptr = skb_gro_header_fast(skb, off); 2840 if (skb_gro_header_hard(skb, off + plen)) { 2841 ptr = skb_gro_header_slow(skb, off + plen, off); 2842 if (!ptr) 2843 return NULL; 2844 } 2845 2846 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2847 start, offset); 2848 2849 /* Adjust skb->csum since we changed the packet */ 2850 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2851 2852 grc->offset = off + hdrlen + offset; 2853 grc->delta = delta; 2854 2855 return ptr; 2856 } 2857 2858 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2859 struct gro_remcsum *grc) 2860 { 2861 void *ptr; 2862 size_t plen = grc->offset + sizeof(u16); 2863 2864 if (!grc->delta) 2865 return; 2866 2867 ptr = skb_gro_header_fast(skb, grc->offset); 2868 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 2869 ptr = skb_gro_header_slow(skb, plen, grc->offset); 2870 if (!ptr) 2871 return; 2872 } 2873 2874 remcsum_unadjust((__sum16 *)ptr, grc->delta); 2875 } 2876 2877 #ifdef CONFIG_XFRM_OFFLOAD 2878 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2879 { 2880 if (PTR_ERR(pp) != -EINPROGRESS) 2881 NAPI_GRO_CB(skb)->flush |= flush; 2882 } 2883 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 2884 struct sk_buff *pp, 2885 int flush, 2886 struct gro_remcsum *grc) 2887 { 2888 if (PTR_ERR(pp) != -EINPROGRESS) { 2889 NAPI_GRO_CB(skb)->flush |= flush; 2890 skb_gro_remcsum_cleanup(skb, grc); 2891 skb->remcsum_offload = 0; 2892 } 2893 } 2894 #else 2895 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2896 { 2897 NAPI_GRO_CB(skb)->flush |= flush; 2898 } 2899 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 2900 struct sk_buff *pp, 2901 int flush, 2902 struct gro_remcsum *grc) 2903 { 2904 NAPI_GRO_CB(skb)->flush |= flush; 2905 skb_gro_remcsum_cleanup(skb, grc); 2906 skb->remcsum_offload = 0; 2907 } 2908 #endif 2909 2910 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2911 unsigned short type, 2912 const void *daddr, const void *saddr, 2913 unsigned int len) 2914 { 2915 if (!dev->header_ops || !dev->header_ops->create) 2916 return 0; 2917 2918 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2919 } 2920 2921 static inline int dev_parse_header(const struct sk_buff *skb, 2922 unsigned char *haddr) 2923 { 2924 const struct net_device *dev = skb->dev; 2925 2926 if (!dev->header_ops || !dev->header_ops->parse) 2927 return 0; 2928 return dev->header_ops->parse(skb, haddr); 2929 } 2930 2931 /* ll_header must have at least hard_header_len allocated */ 2932 static inline bool dev_validate_header(const struct net_device *dev, 2933 char *ll_header, int len) 2934 { 2935 if (likely(len >= dev->hard_header_len)) 2936 return true; 2937 if (len < dev->min_header_len) 2938 return false; 2939 2940 if (capable(CAP_SYS_RAWIO)) { 2941 memset(ll_header + len, 0, dev->hard_header_len - len); 2942 return true; 2943 } 2944 2945 if (dev->header_ops && dev->header_ops->validate) 2946 return dev->header_ops->validate(ll_header, len); 2947 2948 return false; 2949 } 2950 2951 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, 2952 int len, int size); 2953 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2954 static inline int unregister_gifconf(unsigned int family) 2955 { 2956 return register_gifconf(family, NULL); 2957 } 2958 2959 #ifdef CONFIG_NET_FLOW_LIMIT 2960 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2961 struct sd_flow_limit { 2962 u64 count; 2963 unsigned int num_buckets; 2964 unsigned int history_head; 2965 u16 history[FLOW_LIMIT_HISTORY]; 2966 u8 buckets[]; 2967 }; 2968 2969 extern int netdev_flow_limit_table_len; 2970 #endif /* CONFIG_NET_FLOW_LIMIT */ 2971 2972 /* 2973 * Incoming packets are placed on per-CPU queues 2974 */ 2975 struct softnet_data { 2976 struct list_head poll_list; 2977 struct sk_buff_head process_queue; 2978 2979 /* stats */ 2980 unsigned int processed; 2981 unsigned int time_squeeze; 2982 unsigned int received_rps; 2983 #ifdef CONFIG_RPS 2984 struct softnet_data *rps_ipi_list; 2985 #endif 2986 #ifdef CONFIG_NET_FLOW_LIMIT 2987 struct sd_flow_limit __rcu *flow_limit; 2988 #endif 2989 struct Qdisc *output_queue; 2990 struct Qdisc **output_queue_tailp; 2991 struct sk_buff *completion_queue; 2992 #ifdef CONFIG_XFRM_OFFLOAD 2993 struct sk_buff_head xfrm_backlog; 2994 #endif 2995 #ifdef CONFIG_RPS 2996 /* input_queue_head should be written by cpu owning this struct, 2997 * and only read by other cpus. Worth using a cache line. 2998 */ 2999 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3000 3001 /* Elements below can be accessed between CPUs for RPS/RFS */ 3002 call_single_data_t csd ____cacheline_aligned_in_smp; 3003 struct softnet_data *rps_ipi_next; 3004 unsigned int cpu; 3005 unsigned int input_queue_tail; 3006 #endif 3007 unsigned int dropped; 3008 struct sk_buff_head input_pkt_queue; 3009 struct napi_struct backlog; 3010 3011 }; 3012 3013 static inline void input_queue_head_incr(struct softnet_data *sd) 3014 { 3015 #ifdef CONFIG_RPS 3016 sd->input_queue_head++; 3017 #endif 3018 } 3019 3020 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3021 unsigned int *qtail) 3022 { 3023 #ifdef CONFIG_RPS 3024 *qtail = ++sd->input_queue_tail; 3025 #endif 3026 } 3027 3028 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3029 3030 void __netif_schedule(struct Qdisc *q); 3031 void netif_schedule_queue(struct netdev_queue *txq); 3032 3033 static inline void netif_tx_schedule_all(struct net_device *dev) 3034 { 3035 unsigned int i; 3036 3037 for (i = 0; i < dev->num_tx_queues; i++) 3038 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3039 } 3040 3041 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3042 { 3043 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3044 } 3045 3046 /** 3047 * netif_start_queue - allow transmit 3048 * @dev: network device 3049 * 3050 * Allow upper layers to call the device hard_start_xmit routine. 3051 */ 3052 static inline void netif_start_queue(struct net_device *dev) 3053 { 3054 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3055 } 3056 3057 static inline void netif_tx_start_all_queues(struct net_device *dev) 3058 { 3059 unsigned int i; 3060 3061 for (i = 0; i < dev->num_tx_queues; i++) { 3062 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3063 netif_tx_start_queue(txq); 3064 } 3065 } 3066 3067 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3068 3069 /** 3070 * netif_wake_queue - restart transmit 3071 * @dev: network device 3072 * 3073 * Allow upper layers to call the device hard_start_xmit routine. 3074 * Used for flow control when transmit resources are available. 3075 */ 3076 static inline void netif_wake_queue(struct net_device *dev) 3077 { 3078 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3079 } 3080 3081 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3082 { 3083 unsigned int i; 3084 3085 for (i = 0; i < dev->num_tx_queues; i++) { 3086 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3087 netif_tx_wake_queue(txq); 3088 } 3089 } 3090 3091 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3092 { 3093 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3094 } 3095 3096 /** 3097 * netif_stop_queue - stop transmitted packets 3098 * @dev: network device 3099 * 3100 * Stop upper layers calling the device hard_start_xmit routine. 3101 * Used for flow control when transmit resources are unavailable. 3102 */ 3103 static inline void netif_stop_queue(struct net_device *dev) 3104 { 3105 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3106 } 3107 3108 void netif_tx_stop_all_queues(struct net_device *dev); 3109 3110 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3111 { 3112 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3113 } 3114 3115 /** 3116 * netif_queue_stopped - test if transmit queue is flowblocked 3117 * @dev: network device 3118 * 3119 * Test if transmit queue on device is currently unable to send. 3120 */ 3121 static inline bool netif_queue_stopped(const struct net_device *dev) 3122 { 3123 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3124 } 3125 3126 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3127 { 3128 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3129 } 3130 3131 static inline bool 3132 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3133 { 3134 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3135 } 3136 3137 static inline bool 3138 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3139 { 3140 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3141 } 3142 3143 /** 3144 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3145 * @dev_queue: pointer to transmit queue 3146 * 3147 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3148 * to give appropriate hint to the CPU. 3149 */ 3150 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3151 { 3152 #ifdef CONFIG_BQL 3153 prefetchw(&dev_queue->dql.num_queued); 3154 #endif 3155 } 3156 3157 /** 3158 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3159 * @dev_queue: pointer to transmit queue 3160 * 3161 * BQL enabled drivers might use this helper in their TX completion path, 3162 * to give appropriate hint to the CPU. 3163 */ 3164 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3165 { 3166 #ifdef CONFIG_BQL 3167 prefetchw(&dev_queue->dql.limit); 3168 #endif 3169 } 3170 3171 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3172 unsigned int bytes) 3173 { 3174 #ifdef CONFIG_BQL 3175 dql_queued(&dev_queue->dql, bytes); 3176 3177 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3178 return; 3179 3180 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3181 3182 /* 3183 * The XOFF flag must be set before checking the dql_avail below, 3184 * because in netdev_tx_completed_queue we update the dql_completed 3185 * before checking the XOFF flag. 3186 */ 3187 smp_mb(); 3188 3189 /* check again in case another CPU has just made room avail */ 3190 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3191 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3192 #endif 3193 } 3194 3195 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3196 * that they should not test BQL status themselves. 3197 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3198 * skb of a batch. 3199 * Returns true if the doorbell must be used to kick the NIC. 3200 */ 3201 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3202 unsigned int bytes, 3203 bool xmit_more) 3204 { 3205 if (xmit_more) { 3206 #ifdef CONFIG_BQL 3207 dql_queued(&dev_queue->dql, bytes); 3208 #endif 3209 return netif_tx_queue_stopped(dev_queue); 3210 } 3211 netdev_tx_sent_queue(dev_queue, bytes); 3212 return true; 3213 } 3214 3215 /** 3216 * netdev_sent_queue - report the number of bytes queued to hardware 3217 * @dev: network device 3218 * @bytes: number of bytes queued to the hardware device queue 3219 * 3220 * Report the number of bytes queued for sending/completion to the network 3221 * device hardware queue. @bytes should be a good approximation and should 3222 * exactly match netdev_completed_queue() @bytes 3223 */ 3224 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3225 { 3226 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3227 } 3228 3229 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3230 unsigned int pkts, unsigned int bytes) 3231 { 3232 #ifdef CONFIG_BQL 3233 if (unlikely(!bytes)) 3234 return; 3235 3236 dql_completed(&dev_queue->dql, bytes); 3237 3238 /* 3239 * Without the memory barrier there is a small possiblity that 3240 * netdev_tx_sent_queue will miss the update and cause the queue to 3241 * be stopped forever 3242 */ 3243 smp_mb(); 3244 3245 if (dql_avail(&dev_queue->dql) < 0) 3246 return; 3247 3248 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3249 netif_schedule_queue(dev_queue); 3250 #endif 3251 } 3252 3253 /** 3254 * netdev_completed_queue - report bytes and packets completed by device 3255 * @dev: network device 3256 * @pkts: actual number of packets sent over the medium 3257 * @bytes: actual number of bytes sent over the medium 3258 * 3259 * Report the number of bytes and packets transmitted by the network device 3260 * hardware queue over the physical medium, @bytes must exactly match the 3261 * @bytes amount passed to netdev_sent_queue() 3262 */ 3263 static inline void netdev_completed_queue(struct net_device *dev, 3264 unsigned int pkts, unsigned int bytes) 3265 { 3266 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3267 } 3268 3269 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3270 { 3271 #ifdef CONFIG_BQL 3272 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3273 dql_reset(&q->dql); 3274 #endif 3275 } 3276 3277 /** 3278 * netdev_reset_queue - reset the packets and bytes count of a network device 3279 * @dev_queue: network device 3280 * 3281 * Reset the bytes and packet count of a network device and clear the 3282 * software flow control OFF bit for this network device 3283 */ 3284 static inline void netdev_reset_queue(struct net_device *dev_queue) 3285 { 3286 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3287 } 3288 3289 /** 3290 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3291 * @dev: network device 3292 * @queue_index: given tx queue index 3293 * 3294 * Returns 0 if given tx queue index >= number of device tx queues, 3295 * otherwise returns the originally passed tx queue index. 3296 */ 3297 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3298 { 3299 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3300 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3301 dev->name, queue_index, 3302 dev->real_num_tx_queues); 3303 return 0; 3304 } 3305 3306 return queue_index; 3307 } 3308 3309 /** 3310 * netif_running - test if up 3311 * @dev: network device 3312 * 3313 * Test if the device has been brought up. 3314 */ 3315 static inline bool netif_running(const struct net_device *dev) 3316 { 3317 return test_bit(__LINK_STATE_START, &dev->state); 3318 } 3319 3320 /* 3321 * Routines to manage the subqueues on a device. We only need start, 3322 * stop, and a check if it's stopped. All other device management is 3323 * done at the overall netdevice level. 3324 * Also test the device if we're multiqueue. 3325 */ 3326 3327 /** 3328 * netif_start_subqueue - allow sending packets on subqueue 3329 * @dev: network device 3330 * @queue_index: sub queue index 3331 * 3332 * Start individual transmit queue of a device with multiple transmit queues. 3333 */ 3334 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3335 { 3336 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3337 3338 netif_tx_start_queue(txq); 3339 } 3340 3341 /** 3342 * netif_stop_subqueue - stop sending packets on subqueue 3343 * @dev: network device 3344 * @queue_index: sub queue index 3345 * 3346 * Stop individual transmit queue of a device with multiple transmit queues. 3347 */ 3348 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3349 { 3350 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3351 netif_tx_stop_queue(txq); 3352 } 3353 3354 /** 3355 * netif_subqueue_stopped - test status of subqueue 3356 * @dev: network device 3357 * @queue_index: sub queue index 3358 * 3359 * Check individual transmit queue of a device with multiple transmit queues. 3360 */ 3361 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3362 u16 queue_index) 3363 { 3364 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3365 3366 return netif_tx_queue_stopped(txq); 3367 } 3368 3369 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3370 struct sk_buff *skb) 3371 { 3372 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3373 } 3374 3375 /** 3376 * netif_wake_subqueue - allow sending packets on subqueue 3377 * @dev: network device 3378 * @queue_index: sub queue index 3379 * 3380 * Resume individual transmit queue of a device with multiple transmit queues. 3381 */ 3382 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3383 { 3384 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3385 3386 netif_tx_wake_queue(txq); 3387 } 3388 3389 #ifdef CONFIG_XPS 3390 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3391 u16 index); 3392 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3393 u16 index, bool is_rxqs_map); 3394 3395 /** 3396 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3397 * @j: CPU/Rx queue index 3398 * @mask: bitmask of all cpus/rx queues 3399 * @nr_bits: number of bits in the bitmask 3400 * 3401 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3402 */ 3403 static inline bool netif_attr_test_mask(unsigned long j, 3404 const unsigned long *mask, 3405 unsigned int nr_bits) 3406 { 3407 cpu_max_bits_warn(j, nr_bits); 3408 return test_bit(j, mask); 3409 } 3410 3411 /** 3412 * netif_attr_test_online - Test for online CPU/Rx queue 3413 * @j: CPU/Rx queue index 3414 * @online_mask: bitmask for CPUs/Rx queues that are online 3415 * @nr_bits: number of bits in the bitmask 3416 * 3417 * Returns true if a CPU/Rx queue is online. 3418 */ 3419 static inline bool netif_attr_test_online(unsigned long j, 3420 const unsigned long *online_mask, 3421 unsigned int nr_bits) 3422 { 3423 cpu_max_bits_warn(j, nr_bits); 3424 3425 if (online_mask) 3426 return test_bit(j, online_mask); 3427 3428 return (j < nr_bits); 3429 } 3430 3431 /** 3432 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3433 * @n: CPU/Rx queue index 3434 * @srcp: the cpumask/Rx queue mask pointer 3435 * @nr_bits: number of bits in the bitmask 3436 * 3437 * Returns >= nr_bits if no further CPUs/Rx queues set. 3438 */ 3439 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3440 unsigned int nr_bits) 3441 { 3442 /* -1 is a legal arg here. */ 3443 if (n != -1) 3444 cpu_max_bits_warn(n, nr_bits); 3445 3446 if (srcp) 3447 return find_next_bit(srcp, nr_bits, n + 1); 3448 3449 return n + 1; 3450 } 3451 3452 /** 3453 * netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p 3454 * @n: CPU/Rx queue index 3455 * @src1p: the first CPUs/Rx queues mask pointer 3456 * @src2p: the second CPUs/Rx queues mask pointer 3457 * @nr_bits: number of bits in the bitmask 3458 * 3459 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3460 */ 3461 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3462 const unsigned long *src2p, 3463 unsigned int nr_bits) 3464 { 3465 /* -1 is a legal arg here. */ 3466 if (n != -1) 3467 cpu_max_bits_warn(n, nr_bits); 3468 3469 if (src1p && src2p) 3470 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3471 else if (src1p) 3472 return find_next_bit(src1p, nr_bits, n + 1); 3473 else if (src2p) 3474 return find_next_bit(src2p, nr_bits, n + 1); 3475 3476 return n + 1; 3477 } 3478 #else 3479 static inline int netif_set_xps_queue(struct net_device *dev, 3480 const struct cpumask *mask, 3481 u16 index) 3482 { 3483 return 0; 3484 } 3485 3486 static inline int __netif_set_xps_queue(struct net_device *dev, 3487 const unsigned long *mask, 3488 u16 index, bool is_rxqs_map) 3489 { 3490 return 0; 3491 } 3492 #endif 3493 3494 /** 3495 * netif_is_multiqueue - test if device has multiple transmit queues 3496 * @dev: network device 3497 * 3498 * Check if device has multiple transmit queues 3499 */ 3500 static inline bool netif_is_multiqueue(const struct net_device *dev) 3501 { 3502 return dev->num_tx_queues > 1; 3503 } 3504 3505 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3506 3507 #ifdef CONFIG_SYSFS 3508 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3509 #else 3510 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3511 unsigned int rxqs) 3512 { 3513 dev->real_num_rx_queues = rxqs; 3514 return 0; 3515 } 3516 #endif 3517 3518 static inline struct netdev_rx_queue * 3519 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3520 { 3521 return dev->_rx + rxq; 3522 } 3523 3524 #ifdef CONFIG_SYSFS 3525 static inline unsigned int get_netdev_rx_queue_index( 3526 struct netdev_rx_queue *queue) 3527 { 3528 struct net_device *dev = queue->dev; 3529 int index = queue - dev->_rx; 3530 3531 BUG_ON(index >= dev->num_rx_queues); 3532 return index; 3533 } 3534 #endif 3535 3536 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3537 int netif_get_num_default_rss_queues(void); 3538 3539 enum skb_free_reason { 3540 SKB_REASON_CONSUMED, 3541 SKB_REASON_DROPPED, 3542 }; 3543 3544 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3545 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3546 3547 /* 3548 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3549 * interrupt context or with hardware interrupts being disabled. 3550 * (in_irq() || irqs_disabled()) 3551 * 3552 * We provide four helpers that can be used in following contexts : 3553 * 3554 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3555 * replacing kfree_skb(skb) 3556 * 3557 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3558 * Typically used in place of consume_skb(skb) in TX completion path 3559 * 3560 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3561 * replacing kfree_skb(skb) 3562 * 3563 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3564 * and consumed a packet. Used in place of consume_skb(skb) 3565 */ 3566 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3567 { 3568 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3569 } 3570 3571 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3572 { 3573 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3574 } 3575 3576 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3577 { 3578 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3579 } 3580 3581 static inline void dev_consume_skb_any(struct sk_buff *skb) 3582 { 3583 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3584 } 3585 3586 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3587 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3588 int netif_rx(struct sk_buff *skb); 3589 int netif_rx_ni(struct sk_buff *skb); 3590 int netif_receive_skb(struct sk_buff *skb); 3591 int netif_receive_skb_core(struct sk_buff *skb); 3592 void netif_receive_skb_list(struct list_head *head); 3593 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3594 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3595 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3596 gro_result_t napi_gro_frags(struct napi_struct *napi); 3597 struct packet_offload *gro_find_receive_by_type(__be16 type); 3598 struct packet_offload *gro_find_complete_by_type(__be16 type); 3599 3600 static inline void napi_free_frags(struct napi_struct *napi) 3601 { 3602 kfree_skb(napi->skb); 3603 napi->skb = NULL; 3604 } 3605 3606 bool netdev_is_rx_handler_busy(struct net_device *dev); 3607 int netdev_rx_handler_register(struct net_device *dev, 3608 rx_handler_func_t *rx_handler, 3609 void *rx_handler_data); 3610 void netdev_rx_handler_unregister(struct net_device *dev); 3611 3612 bool dev_valid_name(const char *name); 3613 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3614 bool *need_copyout); 3615 int dev_ifconf(struct net *net, struct ifconf *, int); 3616 int dev_ethtool(struct net *net, struct ifreq *); 3617 unsigned int dev_get_flags(const struct net_device *); 3618 int __dev_change_flags(struct net_device *, unsigned int flags); 3619 int dev_change_flags(struct net_device *, unsigned int); 3620 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3621 unsigned int gchanges); 3622 int dev_change_name(struct net_device *, const char *); 3623 int dev_set_alias(struct net_device *, const char *, size_t); 3624 int dev_get_alias(const struct net_device *, char *, size_t); 3625 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3626 int __dev_set_mtu(struct net_device *, int); 3627 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3628 struct netlink_ext_ack *extack); 3629 int dev_set_mtu(struct net_device *, int); 3630 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3631 void dev_set_group(struct net_device *, int); 3632 int dev_set_mac_address(struct net_device *, struct sockaddr *); 3633 int dev_change_carrier(struct net_device *, bool new_carrier); 3634 int dev_get_phys_port_id(struct net_device *dev, 3635 struct netdev_phys_item_id *ppid); 3636 int dev_get_phys_port_name(struct net_device *dev, 3637 char *name, size_t len); 3638 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3639 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3640 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3641 struct netdev_queue *txq, int *ret); 3642 3643 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3644 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3645 int fd, u32 flags); 3646 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op, 3647 enum bpf_netdev_command cmd); 3648 int xdp_umem_query(struct net_device *dev, u16 queue_id); 3649 3650 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3651 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3652 bool is_skb_forwardable(const struct net_device *dev, 3653 const struct sk_buff *skb); 3654 3655 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3656 struct sk_buff *skb) 3657 { 3658 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3659 unlikely(!is_skb_forwardable(dev, skb))) { 3660 atomic_long_inc(&dev->rx_dropped); 3661 kfree_skb(skb); 3662 return NET_RX_DROP; 3663 } 3664 3665 skb_scrub_packet(skb, true); 3666 skb->priority = 0; 3667 return 0; 3668 } 3669 3670 bool dev_nit_active(struct net_device *dev); 3671 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3672 3673 extern int netdev_budget; 3674 extern unsigned int netdev_budget_usecs; 3675 3676 /* Called by rtnetlink.c:rtnl_unlock() */ 3677 void netdev_run_todo(void); 3678 3679 /** 3680 * dev_put - release reference to device 3681 * @dev: network device 3682 * 3683 * Release reference to device to allow it to be freed. 3684 */ 3685 static inline void dev_put(struct net_device *dev) 3686 { 3687 this_cpu_dec(*dev->pcpu_refcnt); 3688 } 3689 3690 /** 3691 * dev_hold - get reference to device 3692 * @dev: network device 3693 * 3694 * Hold reference to device to keep it from being freed. 3695 */ 3696 static inline void dev_hold(struct net_device *dev) 3697 { 3698 this_cpu_inc(*dev->pcpu_refcnt); 3699 } 3700 3701 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3702 * and _off may be called from IRQ context, but it is caller 3703 * who is responsible for serialization of these calls. 3704 * 3705 * The name carrier is inappropriate, these functions should really be 3706 * called netif_lowerlayer_*() because they represent the state of any 3707 * kind of lower layer not just hardware media. 3708 */ 3709 3710 void linkwatch_init_dev(struct net_device *dev); 3711 void linkwatch_fire_event(struct net_device *dev); 3712 void linkwatch_forget_dev(struct net_device *dev); 3713 3714 /** 3715 * netif_carrier_ok - test if carrier present 3716 * @dev: network device 3717 * 3718 * Check if carrier is present on device 3719 */ 3720 static inline bool netif_carrier_ok(const struct net_device *dev) 3721 { 3722 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3723 } 3724 3725 unsigned long dev_trans_start(struct net_device *dev); 3726 3727 void __netdev_watchdog_up(struct net_device *dev); 3728 3729 void netif_carrier_on(struct net_device *dev); 3730 3731 void netif_carrier_off(struct net_device *dev); 3732 3733 /** 3734 * netif_dormant_on - mark device as dormant. 3735 * @dev: network device 3736 * 3737 * Mark device as dormant (as per RFC2863). 3738 * 3739 * The dormant state indicates that the relevant interface is not 3740 * actually in a condition to pass packets (i.e., it is not 'up') but is 3741 * in a "pending" state, waiting for some external event. For "on- 3742 * demand" interfaces, this new state identifies the situation where the 3743 * interface is waiting for events to place it in the up state. 3744 */ 3745 static inline void netif_dormant_on(struct net_device *dev) 3746 { 3747 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3748 linkwatch_fire_event(dev); 3749 } 3750 3751 /** 3752 * netif_dormant_off - set device as not dormant. 3753 * @dev: network device 3754 * 3755 * Device is not in dormant state. 3756 */ 3757 static inline void netif_dormant_off(struct net_device *dev) 3758 { 3759 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3760 linkwatch_fire_event(dev); 3761 } 3762 3763 /** 3764 * netif_dormant - test if device is dormant 3765 * @dev: network device 3766 * 3767 * Check if device is dormant. 3768 */ 3769 static inline bool netif_dormant(const struct net_device *dev) 3770 { 3771 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3772 } 3773 3774 3775 /** 3776 * netif_oper_up - test if device is operational 3777 * @dev: network device 3778 * 3779 * Check if carrier is operational 3780 */ 3781 static inline bool netif_oper_up(const struct net_device *dev) 3782 { 3783 return (dev->operstate == IF_OPER_UP || 3784 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3785 } 3786 3787 /** 3788 * netif_device_present - is device available or removed 3789 * @dev: network device 3790 * 3791 * Check if device has not been removed from system. 3792 */ 3793 static inline bool netif_device_present(struct net_device *dev) 3794 { 3795 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3796 } 3797 3798 void netif_device_detach(struct net_device *dev); 3799 3800 void netif_device_attach(struct net_device *dev); 3801 3802 /* 3803 * Network interface message level settings 3804 */ 3805 3806 enum { 3807 NETIF_MSG_DRV = 0x0001, 3808 NETIF_MSG_PROBE = 0x0002, 3809 NETIF_MSG_LINK = 0x0004, 3810 NETIF_MSG_TIMER = 0x0008, 3811 NETIF_MSG_IFDOWN = 0x0010, 3812 NETIF_MSG_IFUP = 0x0020, 3813 NETIF_MSG_RX_ERR = 0x0040, 3814 NETIF_MSG_TX_ERR = 0x0080, 3815 NETIF_MSG_TX_QUEUED = 0x0100, 3816 NETIF_MSG_INTR = 0x0200, 3817 NETIF_MSG_TX_DONE = 0x0400, 3818 NETIF_MSG_RX_STATUS = 0x0800, 3819 NETIF_MSG_PKTDATA = 0x1000, 3820 NETIF_MSG_HW = 0x2000, 3821 NETIF_MSG_WOL = 0x4000, 3822 }; 3823 3824 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3825 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3826 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3827 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3828 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3829 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3830 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3831 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3832 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3833 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3834 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3835 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3836 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3837 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3838 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3839 3840 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3841 { 3842 /* use default */ 3843 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3844 return default_msg_enable_bits; 3845 if (debug_value == 0) /* no output */ 3846 return 0; 3847 /* set low N bits */ 3848 return (1 << debug_value) - 1; 3849 } 3850 3851 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 3852 { 3853 spin_lock(&txq->_xmit_lock); 3854 txq->xmit_lock_owner = cpu; 3855 } 3856 3857 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 3858 { 3859 __acquire(&txq->_xmit_lock); 3860 return true; 3861 } 3862 3863 static inline void __netif_tx_release(struct netdev_queue *txq) 3864 { 3865 __release(&txq->_xmit_lock); 3866 } 3867 3868 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 3869 { 3870 spin_lock_bh(&txq->_xmit_lock); 3871 txq->xmit_lock_owner = smp_processor_id(); 3872 } 3873 3874 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 3875 { 3876 bool ok = spin_trylock(&txq->_xmit_lock); 3877 if (likely(ok)) 3878 txq->xmit_lock_owner = smp_processor_id(); 3879 return ok; 3880 } 3881 3882 static inline void __netif_tx_unlock(struct netdev_queue *txq) 3883 { 3884 txq->xmit_lock_owner = -1; 3885 spin_unlock(&txq->_xmit_lock); 3886 } 3887 3888 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 3889 { 3890 txq->xmit_lock_owner = -1; 3891 spin_unlock_bh(&txq->_xmit_lock); 3892 } 3893 3894 static inline void txq_trans_update(struct netdev_queue *txq) 3895 { 3896 if (txq->xmit_lock_owner != -1) 3897 txq->trans_start = jiffies; 3898 } 3899 3900 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 3901 static inline void netif_trans_update(struct net_device *dev) 3902 { 3903 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 3904 3905 if (txq->trans_start != jiffies) 3906 txq->trans_start = jiffies; 3907 } 3908 3909 /** 3910 * netif_tx_lock - grab network device transmit lock 3911 * @dev: network device 3912 * 3913 * Get network device transmit lock 3914 */ 3915 static inline void netif_tx_lock(struct net_device *dev) 3916 { 3917 unsigned int i; 3918 int cpu; 3919 3920 spin_lock(&dev->tx_global_lock); 3921 cpu = smp_processor_id(); 3922 for (i = 0; i < dev->num_tx_queues; i++) { 3923 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3924 3925 /* We are the only thread of execution doing a 3926 * freeze, but we have to grab the _xmit_lock in 3927 * order to synchronize with threads which are in 3928 * the ->hard_start_xmit() handler and already 3929 * checked the frozen bit. 3930 */ 3931 __netif_tx_lock(txq, cpu); 3932 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 3933 __netif_tx_unlock(txq); 3934 } 3935 } 3936 3937 static inline void netif_tx_lock_bh(struct net_device *dev) 3938 { 3939 local_bh_disable(); 3940 netif_tx_lock(dev); 3941 } 3942 3943 static inline void netif_tx_unlock(struct net_device *dev) 3944 { 3945 unsigned int i; 3946 3947 for (i = 0; i < dev->num_tx_queues; i++) { 3948 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3949 3950 /* No need to grab the _xmit_lock here. If the 3951 * queue is not stopped for another reason, we 3952 * force a schedule. 3953 */ 3954 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 3955 netif_schedule_queue(txq); 3956 } 3957 spin_unlock(&dev->tx_global_lock); 3958 } 3959 3960 static inline void netif_tx_unlock_bh(struct net_device *dev) 3961 { 3962 netif_tx_unlock(dev); 3963 local_bh_enable(); 3964 } 3965 3966 #define HARD_TX_LOCK(dev, txq, cpu) { \ 3967 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3968 __netif_tx_lock(txq, cpu); \ 3969 } else { \ 3970 __netif_tx_acquire(txq); \ 3971 } \ 3972 } 3973 3974 #define HARD_TX_TRYLOCK(dev, txq) \ 3975 (((dev->features & NETIF_F_LLTX) == 0) ? \ 3976 __netif_tx_trylock(txq) : \ 3977 __netif_tx_acquire(txq)) 3978 3979 #define HARD_TX_UNLOCK(dev, txq) { \ 3980 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3981 __netif_tx_unlock(txq); \ 3982 } else { \ 3983 __netif_tx_release(txq); \ 3984 } \ 3985 } 3986 3987 static inline void netif_tx_disable(struct net_device *dev) 3988 { 3989 unsigned int i; 3990 int cpu; 3991 3992 local_bh_disable(); 3993 cpu = smp_processor_id(); 3994 for (i = 0; i < dev->num_tx_queues; i++) { 3995 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3996 3997 __netif_tx_lock(txq, cpu); 3998 netif_tx_stop_queue(txq); 3999 __netif_tx_unlock(txq); 4000 } 4001 local_bh_enable(); 4002 } 4003 4004 static inline void netif_addr_lock(struct net_device *dev) 4005 { 4006 spin_lock(&dev->addr_list_lock); 4007 } 4008 4009 static inline void netif_addr_lock_nested(struct net_device *dev) 4010 { 4011 int subclass = SINGLE_DEPTH_NESTING; 4012 4013 if (dev->netdev_ops->ndo_get_lock_subclass) 4014 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev); 4015 4016 spin_lock_nested(&dev->addr_list_lock, subclass); 4017 } 4018 4019 static inline void netif_addr_lock_bh(struct net_device *dev) 4020 { 4021 spin_lock_bh(&dev->addr_list_lock); 4022 } 4023 4024 static inline void netif_addr_unlock(struct net_device *dev) 4025 { 4026 spin_unlock(&dev->addr_list_lock); 4027 } 4028 4029 static inline void netif_addr_unlock_bh(struct net_device *dev) 4030 { 4031 spin_unlock_bh(&dev->addr_list_lock); 4032 } 4033 4034 /* 4035 * dev_addrs walker. Should be used only for read access. Call with 4036 * rcu_read_lock held. 4037 */ 4038 #define for_each_dev_addr(dev, ha) \ 4039 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4040 4041 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4042 4043 void ether_setup(struct net_device *dev); 4044 4045 /* Support for loadable net-drivers */ 4046 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4047 unsigned char name_assign_type, 4048 void (*setup)(struct net_device *), 4049 unsigned int txqs, unsigned int rxqs); 4050 int dev_get_valid_name(struct net *net, struct net_device *dev, 4051 const char *name); 4052 4053 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4054 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4055 4056 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4057 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4058 count) 4059 4060 int register_netdev(struct net_device *dev); 4061 void unregister_netdev(struct net_device *dev); 4062 4063 /* General hardware address lists handling functions */ 4064 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4065 struct netdev_hw_addr_list *from_list, int addr_len); 4066 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4067 struct netdev_hw_addr_list *from_list, int addr_len); 4068 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4069 struct net_device *dev, 4070 int (*sync)(struct net_device *, const unsigned char *), 4071 int (*unsync)(struct net_device *, 4072 const unsigned char *)); 4073 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4074 struct net_device *dev, 4075 int (*sync)(struct net_device *, 4076 const unsigned char *, int), 4077 int (*unsync)(struct net_device *, 4078 const unsigned char *, int)); 4079 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4080 struct net_device *dev, 4081 int (*unsync)(struct net_device *, 4082 const unsigned char *, int)); 4083 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4084 struct net_device *dev, 4085 int (*unsync)(struct net_device *, 4086 const unsigned char *)); 4087 void __hw_addr_init(struct netdev_hw_addr_list *list); 4088 4089 /* Functions used for device addresses handling */ 4090 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4091 unsigned char addr_type); 4092 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4093 unsigned char addr_type); 4094 void dev_addr_flush(struct net_device *dev); 4095 int dev_addr_init(struct net_device *dev); 4096 4097 /* Functions used for unicast addresses handling */ 4098 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4099 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4100 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4101 int dev_uc_sync(struct net_device *to, struct net_device *from); 4102 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4103 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4104 void dev_uc_flush(struct net_device *dev); 4105 void dev_uc_init(struct net_device *dev); 4106 4107 /** 4108 * __dev_uc_sync - Synchonize device's unicast list 4109 * @dev: device to sync 4110 * @sync: function to call if address should be added 4111 * @unsync: function to call if address should be removed 4112 * 4113 * Add newly added addresses to the interface, and release 4114 * addresses that have been deleted. 4115 */ 4116 static inline int __dev_uc_sync(struct net_device *dev, 4117 int (*sync)(struct net_device *, 4118 const unsigned char *), 4119 int (*unsync)(struct net_device *, 4120 const unsigned char *)) 4121 { 4122 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4123 } 4124 4125 /** 4126 * __dev_uc_unsync - Remove synchronized addresses from device 4127 * @dev: device to sync 4128 * @unsync: function to call if address should be removed 4129 * 4130 * Remove all addresses that were added to the device by dev_uc_sync(). 4131 */ 4132 static inline void __dev_uc_unsync(struct net_device *dev, 4133 int (*unsync)(struct net_device *, 4134 const unsigned char *)) 4135 { 4136 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4137 } 4138 4139 /* Functions used for multicast addresses handling */ 4140 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4141 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4142 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4143 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4144 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4145 int dev_mc_sync(struct net_device *to, struct net_device *from); 4146 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4147 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4148 void dev_mc_flush(struct net_device *dev); 4149 void dev_mc_init(struct net_device *dev); 4150 4151 /** 4152 * __dev_mc_sync - Synchonize device's multicast list 4153 * @dev: device to sync 4154 * @sync: function to call if address should be added 4155 * @unsync: function to call if address should be removed 4156 * 4157 * Add newly added addresses to the interface, and release 4158 * addresses that have been deleted. 4159 */ 4160 static inline int __dev_mc_sync(struct net_device *dev, 4161 int (*sync)(struct net_device *, 4162 const unsigned char *), 4163 int (*unsync)(struct net_device *, 4164 const unsigned char *)) 4165 { 4166 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4167 } 4168 4169 /** 4170 * __dev_mc_unsync - Remove synchronized addresses from device 4171 * @dev: device to sync 4172 * @unsync: function to call if address should be removed 4173 * 4174 * Remove all addresses that were added to the device by dev_mc_sync(). 4175 */ 4176 static inline void __dev_mc_unsync(struct net_device *dev, 4177 int (*unsync)(struct net_device *, 4178 const unsigned char *)) 4179 { 4180 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4181 } 4182 4183 /* Functions used for secondary unicast and multicast support */ 4184 void dev_set_rx_mode(struct net_device *dev); 4185 void __dev_set_rx_mode(struct net_device *dev); 4186 int dev_set_promiscuity(struct net_device *dev, int inc); 4187 int dev_set_allmulti(struct net_device *dev, int inc); 4188 void netdev_state_change(struct net_device *dev); 4189 void netdev_notify_peers(struct net_device *dev); 4190 void netdev_features_change(struct net_device *dev); 4191 /* Load a device via the kmod */ 4192 void dev_load(struct net *net, const char *name); 4193 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4194 struct rtnl_link_stats64 *storage); 4195 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4196 const struct net_device_stats *netdev_stats); 4197 4198 extern int netdev_max_backlog; 4199 extern int netdev_tstamp_prequeue; 4200 extern int weight_p; 4201 extern int dev_weight_rx_bias; 4202 extern int dev_weight_tx_bias; 4203 extern int dev_rx_weight; 4204 extern int dev_tx_weight; 4205 4206 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4207 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4208 struct list_head **iter); 4209 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 4210 struct list_head **iter); 4211 4212 /* iterate through upper list, must be called under RCU read lock */ 4213 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4214 for (iter = &(dev)->adj_list.upper, \ 4215 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4216 updev; \ 4217 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4218 4219 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4220 int (*fn)(struct net_device *upper_dev, 4221 void *data), 4222 void *data); 4223 4224 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4225 struct net_device *upper_dev); 4226 4227 bool netdev_has_any_upper_dev(struct net_device *dev); 4228 4229 void *netdev_lower_get_next_private(struct net_device *dev, 4230 struct list_head **iter); 4231 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4232 struct list_head **iter); 4233 4234 #define netdev_for_each_lower_private(dev, priv, iter) \ 4235 for (iter = (dev)->adj_list.lower.next, \ 4236 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4237 priv; \ 4238 priv = netdev_lower_get_next_private(dev, &(iter))) 4239 4240 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4241 for (iter = &(dev)->adj_list.lower, \ 4242 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4243 priv; \ 4244 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4245 4246 void *netdev_lower_get_next(struct net_device *dev, 4247 struct list_head **iter); 4248 4249 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4250 for (iter = (dev)->adj_list.lower.next, \ 4251 ldev = netdev_lower_get_next(dev, &(iter)); \ 4252 ldev; \ 4253 ldev = netdev_lower_get_next(dev, &(iter))) 4254 4255 struct net_device *netdev_all_lower_get_next(struct net_device *dev, 4256 struct list_head **iter); 4257 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev, 4258 struct list_head **iter); 4259 4260 int netdev_walk_all_lower_dev(struct net_device *dev, 4261 int (*fn)(struct net_device *lower_dev, 4262 void *data), 4263 void *data); 4264 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4265 int (*fn)(struct net_device *lower_dev, 4266 void *data), 4267 void *data); 4268 4269 void *netdev_adjacent_get_private(struct list_head *adj_list); 4270 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4271 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4272 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4273 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4274 struct netlink_ext_ack *extack); 4275 int netdev_master_upper_dev_link(struct net_device *dev, 4276 struct net_device *upper_dev, 4277 void *upper_priv, void *upper_info, 4278 struct netlink_ext_ack *extack); 4279 void netdev_upper_dev_unlink(struct net_device *dev, 4280 struct net_device *upper_dev); 4281 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4282 void *netdev_lower_dev_get_private(struct net_device *dev, 4283 struct net_device *lower_dev); 4284 void netdev_lower_state_changed(struct net_device *lower_dev, 4285 void *lower_state_info); 4286 4287 /* RSS keys are 40 or 52 bytes long */ 4288 #define NETDEV_RSS_KEY_LEN 52 4289 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4290 void netdev_rss_key_fill(void *buffer, size_t len); 4291 4292 int dev_get_nest_level(struct net_device *dev); 4293 int skb_checksum_help(struct sk_buff *skb); 4294 int skb_crc32c_csum_help(struct sk_buff *skb); 4295 int skb_csum_hwoffload_help(struct sk_buff *skb, 4296 const netdev_features_t features); 4297 4298 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4299 netdev_features_t features, bool tx_path); 4300 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4301 netdev_features_t features); 4302 4303 struct netdev_bonding_info { 4304 ifslave slave; 4305 ifbond master; 4306 }; 4307 4308 struct netdev_notifier_bonding_info { 4309 struct netdev_notifier_info info; /* must be first */ 4310 struct netdev_bonding_info bonding_info; 4311 }; 4312 4313 void netdev_bonding_info_change(struct net_device *dev, 4314 struct netdev_bonding_info *bonding_info); 4315 4316 static inline 4317 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4318 { 4319 return __skb_gso_segment(skb, features, true); 4320 } 4321 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4322 4323 static inline bool can_checksum_protocol(netdev_features_t features, 4324 __be16 protocol) 4325 { 4326 if (protocol == htons(ETH_P_FCOE)) 4327 return !!(features & NETIF_F_FCOE_CRC); 4328 4329 /* Assume this is an IP checksum (not SCTP CRC) */ 4330 4331 if (features & NETIF_F_HW_CSUM) { 4332 /* Can checksum everything */ 4333 return true; 4334 } 4335 4336 switch (protocol) { 4337 case htons(ETH_P_IP): 4338 return !!(features & NETIF_F_IP_CSUM); 4339 case htons(ETH_P_IPV6): 4340 return !!(features & NETIF_F_IPV6_CSUM); 4341 default: 4342 return false; 4343 } 4344 } 4345 4346 #ifdef CONFIG_BUG 4347 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4348 #else 4349 static inline void netdev_rx_csum_fault(struct net_device *dev, 4350 struct sk_buff *skb) 4351 { 4352 } 4353 #endif 4354 /* rx skb timestamps */ 4355 void net_enable_timestamp(void); 4356 void net_disable_timestamp(void); 4357 4358 #ifdef CONFIG_PROC_FS 4359 int __init dev_proc_init(void); 4360 #else 4361 #define dev_proc_init() 0 4362 #endif 4363 4364 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4365 struct sk_buff *skb, struct net_device *dev, 4366 bool more) 4367 { 4368 skb->xmit_more = more ? 1 : 0; 4369 return ops->ndo_start_xmit(skb, dev); 4370 } 4371 4372 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4373 struct netdev_queue *txq, bool more) 4374 { 4375 const struct net_device_ops *ops = dev->netdev_ops; 4376 int rc; 4377 4378 rc = __netdev_start_xmit(ops, skb, dev, more); 4379 if (rc == NETDEV_TX_OK) 4380 txq_trans_update(txq); 4381 4382 return rc; 4383 } 4384 4385 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4386 const void *ns); 4387 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4388 const void *ns); 4389 4390 static inline int netdev_class_create_file(const struct class_attribute *class_attr) 4391 { 4392 return netdev_class_create_file_ns(class_attr, NULL); 4393 } 4394 4395 static inline void netdev_class_remove_file(const struct class_attribute *class_attr) 4396 { 4397 netdev_class_remove_file_ns(class_attr, NULL); 4398 } 4399 4400 extern const struct kobj_ns_type_operations net_ns_type_operations; 4401 4402 const char *netdev_drivername(const struct net_device *dev); 4403 4404 void linkwatch_run_queue(void); 4405 4406 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4407 netdev_features_t f2) 4408 { 4409 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4410 if (f1 & NETIF_F_HW_CSUM) 4411 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4412 else 4413 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4414 } 4415 4416 return f1 & f2; 4417 } 4418 4419 static inline netdev_features_t netdev_get_wanted_features( 4420 struct net_device *dev) 4421 { 4422 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4423 } 4424 netdev_features_t netdev_increment_features(netdev_features_t all, 4425 netdev_features_t one, netdev_features_t mask); 4426 4427 /* Allow TSO being used on stacked device : 4428 * Performing the GSO segmentation before last device 4429 * is a performance improvement. 4430 */ 4431 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4432 netdev_features_t mask) 4433 { 4434 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4435 } 4436 4437 int __netdev_update_features(struct net_device *dev); 4438 void netdev_update_features(struct net_device *dev); 4439 void netdev_change_features(struct net_device *dev); 4440 4441 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4442 struct net_device *dev); 4443 4444 netdev_features_t passthru_features_check(struct sk_buff *skb, 4445 struct net_device *dev, 4446 netdev_features_t features); 4447 netdev_features_t netif_skb_features(struct sk_buff *skb); 4448 4449 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4450 { 4451 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4452 4453 /* check flags correspondence */ 4454 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4455 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4456 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4457 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4458 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4459 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4460 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4461 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4462 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4463 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4464 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4465 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4466 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4467 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4468 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4469 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4470 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4471 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4472 4473 return (features & feature) == feature; 4474 } 4475 4476 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4477 { 4478 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4479 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4480 } 4481 4482 static inline bool netif_needs_gso(struct sk_buff *skb, 4483 netdev_features_t features) 4484 { 4485 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4486 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4487 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4488 } 4489 4490 static inline void netif_set_gso_max_size(struct net_device *dev, 4491 unsigned int size) 4492 { 4493 dev->gso_max_size = size; 4494 } 4495 4496 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4497 int pulled_hlen, u16 mac_offset, 4498 int mac_len) 4499 { 4500 skb->protocol = protocol; 4501 skb->encapsulation = 1; 4502 skb_push(skb, pulled_hlen); 4503 skb_reset_transport_header(skb); 4504 skb->mac_header = mac_offset; 4505 skb->network_header = skb->mac_header + mac_len; 4506 skb->mac_len = mac_len; 4507 } 4508 4509 static inline bool netif_is_macsec(const struct net_device *dev) 4510 { 4511 return dev->priv_flags & IFF_MACSEC; 4512 } 4513 4514 static inline bool netif_is_macvlan(const struct net_device *dev) 4515 { 4516 return dev->priv_flags & IFF_MACVLAN; 4517 } 4518 4519 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4520 { 4521 return dev->priv_flags & IFF_MACVLAN_PORT; 4522 } 4523 4524 static inline bool netif_is_bond_master(const struct net_device *dev) 4525 { 4526 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4527 } 4528 4529 static inline bool netif_is_bond_slave(const struct net_device *dev) 4530 { 4531 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4532 } 4533 4534 static inline bool netif_supports_nofcs(struct net_device *dev) 4535 { 4536 return dev->priv_flags & IFF_SUPP_NOFCS; 4537 } 4538 4539 static inline bool netif_is_l3_master(const struct net_device *dev) 4540 { 4541 return dev->priv_flags & IFF_L3MDEV_MASTER; 4542 } 4543 4544 static inline bool netif_is_l3_slave(const struct net_device *dev) 4545 { 4546 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4547 } 4548 4549 static inline bool netif_is_bridge_master(const struct net_device *dev) 4550 { 4551 return dev->priv_flags & IFF_EBRIDGE; 4552 } 4553 4554 static inline bool netif_is_bridge_port(const struct net_device *dev) 4555 { 4556 return dev->priv_flags & IFF_BRIDGE_PORT; 4557 } 4558 4559 static inline bool netif_is_ovs_master(const struct net_device *dev) 4560 { 4561 return dev->priv_flags & IFF_OPENVSWITCH; 4562 } 4563 4564 static inline bool netif_is_ovs_port(const struct net_device *dev) 4565 { 4566 return dev->priv_flags & IFF_OVS_DATAPATH; 4567 } 4568 4569 static inline bool netif_is_team_master(const struct net_device *dev) 4570 { 4571 return dev->priv_flags & IFF_TEAM; 4572 } 4573 4574 static inline bool netif_is_team_port(const struct net_device *dev) 4575 { 4576 return dev->priv_flags & IFF_TEAM_PORT; 4577 } 4578 4579 static inline bool netif_is_lag_master(const struct net_device *dev) 4580 { 4581 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4582 } 4583 4584 static inline bool netif_is_lag_port(const struct net_device *dev) 4585 { 4586 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4587 } 4588 4589 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4590 { 4591 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4592 } 4593 4594 static inline bool netif_is_failover(const struct net_device *dev) 4595 { 4596 return dev->priv_flags & IFF_FAILOVER; 4597 } 4598 4599 static inline bool netif_is_failover_slave(const struct net_device *dev) 4600 { 4601 return dev->priv_flags & IFF_FAILOVER_SLAVE; 4602 } 4603 4604 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4605 static inline void netif_keep_dst(struct net_device *dev) 4606 { 4607 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4608 } 4609 4610 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4611 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4612 { 4613 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4614 return dev->priv_flags & IFF_MACSEC; 4615 } 4616 4617 extern struct pernet_operations __net_initdata loopback_net_ops; 4618 4619 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4620 4621 /* netdev_printk helpers, similar to dev_printk */ 4622 4623 static inline const char *netdev_name(const struct net_device *dev) 4624 { 4625 if (!dev->name[0] || strchr(dev->name, '%')) 4626 return "(unnamed net_device)"; 4627 return dev->name; 4628 } 4629 4630 static inline bool netdev_unregistering(const struct net_device *dev) 4631 { 4632 return dev->reg_state == NETREG_UNREGISTERING; 4633 } 4634 4635 static inline const char *netdev_reg_state(const struct net_device *dev) 4636 { 4637 switch (dev->reg_state) { 4638 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4639 case NETREG_REGISTERED: return ""; 4640 case NETREG_UNREGISTERING: return " (unregistering)"; 4641 case NETREG_UNREGISTERED: return " (unregistered)"; 4642 case NETREG_RELEASED: return " (released)"; 4643 case NETREG_DUMMY: return " (dummy)"; 4644 } 4645 4646 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4647 return " (unknown)"; 4648 } 4649 4650 __printf(3, 4) 4651 void netdev_printk(const char *level, const struct net_device *dev, 4652 const char *format, ...); 4653 __printf(2, 3) 4654 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4655 __printf(2, 3) 4656 void netdev_alert(const struct net_device *dev, const char *format, ...); 4657 __printf(2, 3) 4658 void netdev_crit(const struct net_device *dev, const char *format, ...); 4659 __printf(2, 3) 4660 void netdev_err(const struct net_device *dev, const char *format, ...); 4661 __printf(2, 3) 4662 void netdev_warn(const struct net_device *dev, const char *format, ...); 4663 __printf(2, 3) 4664 void netdev_notice(const struct net_device *dev, const char *format, ...); 4665 __printf(2, 3) 4666 void netdev_info(const struct net_device *dev, const char *format, ...); 4667 4668 #define netdev_level_once(level, dev, fmt, ...) \ 4669 do { \ 4670 static bool __print_once __read_mostly; \ 4671 \ 4672 if (!__print_once) { \ 4673 __print_once = true; \ 4674 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 4675 } \ 4676 } while (0) 4677 4678 #define netdev_emerg_once(dev, fmt, ...) \ 4679 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 4680 #define netdev_alert_once(dev, fmt, ...) \ 4681 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 4682 #define netdev_crit_once(dev, fmt, ...) \ 4683 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 4684 #define netdev_err_once(dev, fmt, ...) \ 4685 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 4686 #define netdev_warn_once(dev, fmt, ...) \ 4687 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 4688 #define netdev_notice_once(dev, fmt, ...) \ 4689 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 4690 #define netdev_info_once(dev, fmt, ...) \ 4691 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 4692 4693 #define MODULE_ALIAS_NETDEV(device) \ 4694 MODULE_ALIAS("netdev-" device) 4695 4696 #if defined(CONFIG_DYNAMIC_DEBUG) 4697 #define netdev_dbg(__dev, format, args...) \ 4698 do { \ 4699 dynamic_netdev_dbg(__dev, format, ##args); \ 4700 } while (0) 4701 #elif defined(DEBUG) 4702 #define netdev_dbg(__dev, format, args...) \ 4703 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4704 #else 4705 #define netdev_dbg(__dev, format, args...) \ 4706 ({ \ 4707 if (0) \ 4708 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4709 }) 4710 #endif 4711 4712 #if defined(VERBOSE_DEBUG) 4713 #define netdev_vdbg netdev_dbg 4714 #else 4715 4716 #define netdev_vdbg(dev, format, args...) \ 4717 ({ \ 4718 if (0) \ 4719 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 4720 0; \ 4721 }) 4722 #endif 4723 4724 /* 4725 * netdev_WARN() acts like dev_printk(), but with the key difference 4726 * of using a WARN/WARN_ON to get the message out, including the 4727 * file/line information and a backtrace. 4728 */ 4729 #define netdev_WARN(dev, format, args...) \ 4730 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4731 netdev_reg_state(dev), ##args) 4732 4733 #define netdev_WARN_ONCE(dev, format, args...) \ 4734 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4735 netdev_reg_state(dev), ##args) 4736 4737 /* netif printk helpers, similar to netdev_printk */ 4738 4739 #define netif_printk(priv, type, level, dev, fmt, args...) \ 4740 do { \ 4741 if (netif_msg_##type(priv)) \ 4742 netdev_printk(level, (dev), fmt, ##args); \ 4743 } while (0) 4744 4745 #define netif_level(level, priv, type, dev, fmt, args...) \ 4746 do { \ 4747 if (netif_msg_##type(priv)) \ 4748 netdev_##level(dev, fmt, ##args); \ 4749 } while (0) 4750 4751 #define netif_emerg(priv, type, dev, fmt, args...) \ 4752 netif_level(emerg, priv, type, dev, fmt, ##args) 4753 #define netif_alert(priv, type, dev, fmt, args...) \ 4754 netif_level(alert, priv, type, dev, fmt, ##args) 4755 #define netif_crit(priv, type, dev, fmt, args...) \ 4756 netif_level(crit, priv, type, dev, fmt, ##args) 4757 #define netif_err(priv, type, dev, fmt, args...) \ 4758 netif_level(err, priv, type, dev, fmt, ##args) 4759 #define netif_warn(priv, type, dev, fmt, args...) \ 4760 netif_level(warn, priv, type, dev, fmt, ##args) 4761 #define netif_notice(priv, type, dev, fmt, args...) \ 4762 netif_level(notice, priv, type, dev, fmt, ##args) 4763 #define netif_info(priv, type, dev, fmt, args...) \ 4764 netif_level(info, priv, type, dev, fmt, ##args) 4765 4766 #if defined(CONFIG_DYNAMIC_DEBUG) 4767 #define netif_dbg(priv, type, netdev, format, args...) \ 4768 do { \ 4769 if (netif_msg_##type(priv)) \ 4770 dynamic_netdev_dbg(netdev, format, ##args); \ 4771 } while (0) 4772 #elif defined(DEBUG) 4773 #define netif_dbg(priv, type, dev, format, args...) \ 4774 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 4775 #else 4776 #define netif_dbg(priv, type, dev, format, args...) \ 4777 ({ \ 4778 if (0) \ 4779 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4780 0; \ 4781 }) 4782 #endif 4783 4784 /* if @cond then downgrade to debug, else print at @level */ 4785 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 4786 do { \ 4787 if (cond) \ 4788 netif_dbg(priv, type, netdev, fmt, ##args); \ 4789 else \ 4790 netif_ ## level(priv, type, netdev, fmt, ##args); \ 4791 } while (0) 4792 4793 #if defined(VERBOSE_DEBUG) 4794 #define netif_vdbg netif_dbg 4795 #else 4796 #define netif_vdbg(priv, type, dev, format, args...) \ 4797 ({ \ 4798 if (0) \ 4799 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4800 0; \ 4801 }) 4802 #endif 4803 4804 /* 4805 * The list of packet types we will receive (as opposed to discard) 4806 * and the routines to invoke. 4807 * 4808 * Why 16. Because with 16 the only overlap we get on a hash of the 4809 * low nibble of the protocol value is RARP/SNAP/X.25. 4810 * 4811 * 0800 IP 4812 * 0001 802.3 4813 * 0002 AX.25 4814 * 0004 802.2 4815 * 8035 RARP 4816 * 0005 SNAP 4817 * 0805 X.25 4818 * 0806 ARP 4819 * 8137 IPX 4820 * 0009 Localtalk 4821 * 86DD IPv6 4822 */ 4823 #define PTYPE_HASH_SIZE (16) 4824 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 4825 4826 #endif /* _LINUX_NETDEVICE_H */ 4827