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