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 extern int sysctl_devconf_inherit_init_net; 634 635 static inline bool net_has_fallback_tunnels(const struct net *net) 636 { 637 return net == &init_net || 638 !IS_ENABLED(CONFIG_SYSCTL) || 639 !sysctl_fb_tunnels_only_for_init_net; 640 } 641 642 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 643 { 644 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 645 return q->numa_node; 646 #else 647 return NUMA_NO_NODE; 648 #endif 649 } 650 651 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 652 { 653 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 654 q->numa_node = node; 655 #endif 656 } 657 658 #ifdef CONFIG_RPS 659 /* 660 * This structure holds an RPS map which can be of variable length. The 661 * map is an array of CPUs. 662 */ 663 struct rps_map { 664 unsigned int len; 665 struct rcu_head rcu; 666 u16 cpus[0]; 667 }; 668 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 669 670 /* 671 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 672 * tail pointer for that CPU's input queue at the time of last enqueue, and 673 * a hardware filter index. 674 */ 675 struct rps_dev_flow { 676 u16 cpu; 677 u16 filter; 678 unsigned int last_qtail; 679 }; 680 #define RPS_NO_FILTER 0xffff 681 682 /* 683 * The rps_dev_flow_table structure contains a table of flow mappings. 684 */ 685 struct rps_dev_flow_table { 686 unsigned int mask; 687 struct rcu_head rcu; 688 struct rps_dev_flow flows[0]; 689 }; 690 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 691 ((_num) * sizeof(struct rps_dev_flow))) 692 693 /* 694 * The rps_sock_flow_table contains mappings of flows to the last CPU 695 * on which they were processed by the application (set in recvmsg). 696 * Each entry is a 32bit value. Upper part is the high-order bits 697 * of flow hash, lower part is CPU number. 698 * rps_cpu_mask is used to partition the space, depending on number of 699 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 700 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 701 * meaning we use 32-6=26 bits for the hash. 702 */ 703 struct rps_sock_flow_table { 704 u32 mask; 705 706 u32 ents[0] ____cacheline_aligned_in_smp; 707 }; 708 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 709 710 #define RPS_NO_CPU 0xffff 711 712 extern u32 rps_cpu_mask; 713 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 714 715 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 716 u32 hash) 717 { 718 if (table && hash) { 719 unsigned int index = hash & table->mask; 720 u32 val = hash & ~rps_cpu_mask; 721 722 /* We only give a hint, preemption can change CPU under us */ 723 val |= raw_smp_processor_id(); 724 725 if (table->ents[index] != val) 726 table->ents[index] = val; 727 } 728 } 729 730 #ifdef CONFIG_RFS_ACCEL 731 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 732 u16 filter_id); 733 #endif 734 #endif /* CONFIG_RPS */ 735 736 /* This structure contains an instance of an RX queue. */ 737 struct netdev_rx_queue { 738 #ifdef CONFIG_RPS 739 struct rps_map __rcu *rps_map; 740 struct rps_dev_flow_table __rcu *rps_flow_table; 741 #endif 742 struct kobject kobj; 743 struct net_device *dev; 744 struct xdp_rxq_info xdp_rxq; 745 #ifdef CONFIG_XDP_SOCKETS 746 struct xdp_umem *umem; 747 #endif 748 } ____cacheline_aligned_in_smp; 749 750 /* 751 * RX queue sysfs structures and functions. 752 */ 753 struct rx_queue_attribute { 754 struct attribute attr; 755 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 756 ssize_t (*store)(struct netdev_rx_queue *queue, 757 const char *buf, size_t len); 758 }; 759 760 #ifdef CONFIG_XPS 761 /* 762 * This structure holds an XPS map which can be of variable length. The 763 * map is an array of queues. 764 */ 765 struct xps_map { 766 unsigned int len; 767 unsigned int alloc_len; 768 struct rcu_head rcu; 769 u16 queues[0]; 770 }; 771 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 772 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 773 - sizeof(struct xps_map)) / sizeof(u16)) 774 775 /* 776 * This structure holds all XPS maps for device. Maps are indexed by CPU. 777 */ 778 struct xps_dev_maps { 779 struct rcu_head rcu; 780 struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */ 781 }; 782 783 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 784 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 785 786 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 787 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 788 789 #endif /* CONFIG_XPS */ 790 791 #define TC_MAX_QUEUE 16 792 #define TC_BITMASK 15 793 /* HW offloaded queuing disciplines txq count and offset maps */ 794 struct netdev_tc_txq { 795 u16 count; 796 u16 offset; 797 }; 798 799 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 800 /* 801 * This structure is to hold information about the device 802 * configured to run FCoE protocol stack. 803 */ 804 struct netdev_fcoe_hbainfo { 805 char manufacturer[64]; 806 char serial_number[64]; 807 char hardware_version[64]; 808 char driver_version[64]; 809 char optionrom_version[64]; 810 char firmware_version[64]; 811 char model[256]; 812 char model_description[256]; 813 }; 814 #endif 815 816 #define MAX_PHYS_ITEM_ID_LEN 32 817 818 /* This structure holds a unique identifier to identify some 819 * physical item (port for example) used by a netdevice. 820 */ 821 struct netdev_phys_item_id { 822 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 823 unsigned char id_len; 824 }; 825 826 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 827 struct netdev_phys_item_id *b) 828 { 829 return a->id_len == b->id_len && 830 memcmp(a->id, b->id, a->id_len) == 0; 831 } 832 833 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 834 struct sk_buff *skb, 835 struct net_device *sb_dev); 836 837 enum tc_setup_type { 838 TC_SETUP_QDISC_MQPRIO, 839 TC_SETUP_CLSU32, 840 TC_SETUP_CLSFLOWER, 841 TC_SETUP_CLSMATCHALL, 842 TC_SETUP_CLSBPF, 843 TC_SETUP_BLOCK, 844 TC_SETUP_QDISC_CBS, 845 TC_SETUP_QDISC_RED, 846 TC_SETUP_QDISC_PRIO, 847 TC_SETUP_QDISC_MQ, 848 TC_SETUP_QDISC_ETF, 849 TC_SETUP_ROOT_QDISC, 850 TC_SETUP_QDISC_GRED, 851 }; 852 853 /* These structures hold the attributes of bpf state that are being passed 854 * to the netdevice through the bpf op. 855 */ 856 enum bpf_netdev_command { 857 /* Set or clear a bpf program used in the earliest stages of packet 858 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 859 * is responsible for calling bpf_prog_put on any old progs that are 860 * stored. In case of error, the callee need not release the new prog 861 * reference, but on success it takes ownership and must bpf_prog_put 862 * when it is no longer used. 863 */ 864 XDP_SETUP_PROG, 865 XDP_SETUP_PROG_HW, 866 XDP_QUERY_PROG, 867 XDP_QUERY_PROG_HW, 868 /* BPF program for offload callbacks, invoked at program load time. */ 869 BPF_OFFLOAD_MAP_ALLOC, 870 BPF_OFFLOAD_MAP_FREE, 871 XDP_QUERY_XSK_UMEM, 872 XDP_SETUP_XSK_UMEM, 873 }; 874 875 struct bpf_prog_offload_ops; 876 struct netlink_ext_ack; 877 struct xdp_umem; 878 879 struct netdev_bpf { 880 enum bpf_netdev_command command; 881 union { 882 /* XDP_SETUP_PROG */ 883 struct { 884 u32 flags; 885 struct bpf_prog *prog; 886 struct netlink_ext_ack *extack; 887 }; 888 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */ 889 struct { 890 u32 prog_id; 891 /* flags with which program was installed */ 892 u32 prog_flags; 893 }; 894 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 895 struct { 896 struct bpf_offloaded_map *offmap; 897 }; 898 /* XDP_QUERY_XSK_UMEM, XDP_SETUP_XSK_UMEM */ 899 struct { 900 struct xdp_umem *umem; /* out for query*/ 901 u16 queue_id; /* in for query */ 902 } xsk; 903 }; 904 }; 905 906 #ifdef CONFIG_XFRM_OFFLOAD 907 struct xfrmdev_ops { 908 int (*xdo_dev_state_add) (struct xfrm_state *x); 909 void (*xdo_dev_state_delete) (struct xfrm_state *x); 910 void (*xdo_dev_state_free) (struct xfrm_state *x); 911 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 912 struct xfrm_state *x); 913 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 914 }; 915 #endif 916 917 #if IS_ENABLED(CONFIG_TLS_DEVICE) 918 enum tls_offload_ctx_dir { 919 TLS_OFFLOAD_CTX_DIR_RX, 920 TLS_OFFLOAD_CTX_DIR_TX, 921 }; 922 923 struct tls_crypto_info; 924 struct tls_context; 925 926 struct tlsdev_ops { 927 int (*tls_dev_add)(struct net_device *netdev, struct sock *sk, 928 enum tls_offload_ctx_dir direction, 929 struct tls_crypto_info *crypto_info, 930 u32 start_offload_tcp_sn); 931 void (*tls_dev_del)(struct net_device *netdev, 932 struct tls_context *ctx, 933 enum tls_offload_ctx_dir direction); 934 void (*tls_dev_resync_rx)(struct net_device *netdev, 935 struct sock *sk, u32 seq, u64 rcd_sn); 936 }; 937 #endif 938 939 struct dev_ifalias { 940 struct rcu_head rcuhead; 941 char ifalias[]; 942 }; 943 944 /* 945 * This structure defines the management hooks for network devices. 946 * The following hooks can be defined; unless noted otherwise, they are 947 * optional and can be filled with a null pointer. 948 * 949 * int (*ndo_init)(struct net_device *dev); 950 * This function is called once when a network device is registered. 951 * The network device can use this for any late stage initialization 952 * or semantic validation. It can fail with an error code which will 953 * be propagated back to register_netdev. 954 * 955 * void (*ndo_uninit)(struct net_device *dev); 956 * This function is called when device is unregistered or when registration 957 * fails. It is not called if init fails. 958 * 959 * int (*ndo_open)(struct net_device *dev); 960 * This function is called when a network device transitions to the up 961 * state. 962 * 963 * int (*ndo_stop)(struct net_device *dev); 964 * This function is called when a network device transitions to the down 965 * state. 966 * 967 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 968 * struct net_device *dev); 969 * Called when a packet needs to be transmitted. 970 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 971 * the queue before that can happen; it's for obsolete devices and weird 972 * corner cases, but the stack really does a non-trivial amount 973 * of useless work if you return NETDEV_TX_BUSY. 974 * Required; cannot be NULL. 975 * 976 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 977 * struct net_device *dev 978 * netdev_features_t features); 979 * Called by core transmit path to determine if device is capable of 980 * performing offload operations on a given packet. This is to give 981 * the device an opportunity to implement any restrictions that cannot 982 * be otherwise expressed by feature flags. The check is called with 983 * the set of features that the stack has calculated and it returns 984 * those the driver believes to be appropriate. 985 * 986 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 987 * struct net_device *sb_dev, 988 * select_queue_fallback_t fallback); 989 * Called to decide which queue to use when device supports multiple 990 * transmit queues. 991 * 992 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 993 * This function is called to allow device receiver to make 994 * changes to configuration when multicast or promiscuous is enabled. 995 * 996 * void (*ndo_set_rx_mode)(struct net_device *dev); 997 * This function is called device changes address list filtering. 998 * If driver handles unicast address filtering, it should set 999 * IFF_UNICAST_FLT in its priv_flags. 1000 * 1001 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1002 * This function is called when the Media Access Control address 1003 * needs to be changed. If this interface is not defined, the 1004 * MAC address can not be changed. 1005 * 1006 * int (*ndo_validate_addr)(struct net_device *dev); 1007 * Test if Media Access Control address is valid for the device. 1008 * 1009 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1010 * Called when a user requests an ioctl which can't be handled by 1011 * the generic interface code. If not defined ioctls return 1012 * not supported error code. 1013 * 1014 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1015 * Used to set network devices bus interface parameters. This interface 1016 * is retained for legacy reasons; new devices should use the bus 1017 * interface (PCI) for low level management. 1018 * 1019 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1020 * Called when a user wants to change the Maximum Transfer Unit 1021 * of a device. 1022 * 1023 * void (*ndo_tx_timeout)(struct net_device *dev); 1024 * Callback used when the transmitter has not made any progress 1025 * for dev->watchdog ticks. 1026 * 1027 * void (*ndo_get_stats64)(struct net_device *dev, 1028 * struct rtnl_link_stats64 *storage); 1029 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1030 * Called when a user wants to get the network device usage 1031 * statistics. Drivers must do one of the following: 1032 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1033 * rtnl_link_stats64 structure passed by the caller. 1034 * 2. Define @ndo_get_stats to update a net_device_stats structure 1035 * (which should normally be dev->stats) and return a pointer to 1036 * it. The structure may be changed asynchronously only if each 1037 * field is written atomically. 1038 * 3. Update dev->stats asynchronously and atomically, and define 1039 * neither operation. 1040 * 1041 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1042 * Return true if this device supports offload stats of this attr_id. 1043 * 1044 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1045 * void *attr_data) 1046 * Get statistics for offload operations by attr_id. Write it into the 1047 * attr_data pointer. 1048 * 1049 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1050 * If device supports VLAN filtering this function is called when a 1051 * VLAN id is registered. 1052 * 1053 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1054 * If device supports VLAN filtering this function is called when a 1055 * VLAN id is unregistered. 1056 * 1057 * void (*ndo_poll_controller)(struct net_device *dev); 1058 * 1059 * SR-IOV management functions. 1060 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1061 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1062 * u8 qos, __be16 proto); 1063 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1064 * int max_tx_rate); 1065 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1066 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1067 * int (*ndo_get_vf_config)(struct net_device *dev, 1068 * int vf, struct ifla_vf_info *ivf); 1069 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1070 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1071 * struct nlattr *port[]); 1072 * 1073 * Enable or disable the VF ability to query its RSS Redirection Table and 1074 * Hash Key. This is needed since on some devices VF share this information 1075 * with PF and querying it may introduce a theoretical security risk. 1076 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1077 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1078 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1079 * void *type_data); 1080 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1081 * This is always called from the stack with the rtnl lock held and netif 1082 * tx queues stopped. This allows the netdevice to perform queue 1083 * management safely. 1084 * 1085 * Fiber Channel over Ethernet (FCoE) offload functions. 1086 * int (*ndo_fcoe_enable)(struct net_device *dev); 1087 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1088 * so the underlying device can perform whatever needed configuration or 1089 * initialization to support acceleration of FCoE traffic. 1090 * 1091 * int (*ndo_fcoe_disable)(struct net_device *dev); 1092 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1093 * so the underlying device can perform whatever needed clean-ups to 1094 * stop supporting acceleration of FCoE traffic. 1095 * 1096 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1097 * struct scatterlist *sgl, unsigned int sgc); 1098 * Called when the FCoE Initiator wants to initialize an I/O that 1099 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1100 * perform necessary setup and returns 1 to indicate the device is set up 1101 * successfully to perform DDP on this I/O, otherwise this returns 0. 1102 * 1103 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1104 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1105 * indicated by the FC exchange id 'xid', so the underlying device can 1106 * clean up and reuse resources for later DDP requests. 1107 * 1108 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1109 * struct scatterlist *sgl, unsigned int sgc); 1110 * Called when the FCoE Target wants to initialize an I/O that 1111 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1112 * perform necessary setup and returns 1 to indicate the device is set up 1113 * successfully to perform DDP on this I/O, otherwise this returns 0. 1114 * 1115 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1116 * struct netdev_fcoe_hbainfo *hbainfo); 1117 * Called when the FCoE Protocol stack wants information on the underlying 1118 * device. This information is utilized by the FCoE protocol stack to 1119 * register attributes with Fiber Channel management service as per the 1120 * FC-GS Fabric Device Management Information(FDMI) specification. 1121 * 1122 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1123 * Called when the underlying device wants to override default World Wide 1124 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1125 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1126 * protocol stack to use. 1127 * 1128 * RFS acceleration. 1129 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1130 * u16 rxq_index, u32 flow_id); 1131 * Set hardware filter for RFS. rxq_index is the target queue index; 1132 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1133 * Return the filter ID on success, or a negative error code. 1134 * 1135 * Slave management functions (for bridge, bonding, etc). 1136 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1137 * Called to make another netdev an underling. 1138 * 1139 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1140 * Called to release previously enslaved netdev. 1141 * 1142 * Feature/offload setting functions. 1143 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1144 * netdev_features_t features); 1145 * Adjusts the requested feature flags according to device-specific 1146 * constraints, and returns the resulting flags. Must not modify 1147 * the device state. 1148 * 1149 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1150 * Called to update device configuration to new features. Passed 1151 * feature set might be less than what was returned by ndo_fix_features()). 1152 * Must return >0 or -errno if it changed dev->features itself. 1153 * 1154 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1155 * struct net_device *dev, 1156 * const unsigned char *addr, u16 vid, u16 flags, 1157 * struct netlink_ext_ack *extack); 1158 * Adds an FDB entry to dev for addr. 1159 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1160 * struct net_device *dev, 1161 * const unsigned char *addr, u16 vid) 1162 * Deletes the FDB entry from dev coresponding to addr. 1163 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1164 * struct net_device *dev, struct net_device *filter_dev, 1165 * int *idx) 1166 * Used to add FDB entries to dump requests. Implementers should add 1167 * entries to skb and update idx with the number of entries. 1168 * 1169 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1170 * u16 flags, struct netlink_ext_ack *extack) 1171 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1172 * struct net_device *dev, u32 filter_mask, 1173 * int nlflags) 1174 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1175 * u16 flags); 1176 * 1177 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1178 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1179 * which do not represent real hardware may define this to allow their 1180 * userspace components to manage their virtual carrier state. Devices 1181 * that determine carrier state from physical hardware properties (eg 1182 * network cables) or protocol-dependent mechanisms (eg 1183 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1184 * 1185 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1186 * struct netdev_phys_item_id *ppid); 1187 * Called to get ID of physical port of this device. If driver does 1188 * not implement this, it is assumed that the hw is not able to have 1189 * multiple net devices on single physical port. 1190 * 1191 * void (*ndo_udp_tunnel_add)(struct net_device *dev, 1192 * struct udp_tunnel_info *ti); 1193 * Called by UDP tunnel to notify a driver about the UDP port and socket 1194 * address family that a UDP tunnel is listnening to. It is called only 1195 * when a new port starts listening. The operation is protected by the 1196 * RTNL. 1197 * 1198 * void (*ndo_udp_tunnel_del)(struct net_device *dev, 1199 * struct udp_tunnel_info *ti); 1200 * Called by UDP tunnel to notify the driver about a UDP port and socket 1201 * address family that the UDP tunnel is not listening to anymore. The 1202 * operation is protected by the RTNL. 1203 * 1204 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1205 * struct net_device *dev) 1206 * Called by upper layer devices to accelerate switching or other 1207 * station functionality into hardware. 'pdev is the lowerdev 1208 * to use for the offload and 'dev' is the net device that will 1209 * back the offload. Returns a pointer to the private structure 1210 * the upper layer will maintain. 1211 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1212 * Called by upper layer device to delete the station created 1213 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1214 * the station and priv is the structure returned by the add 1215 * operation. 1216 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1217 * int queue_index, u32 maxrate); 1218 * Called when a user wants to set a max-rate limitation of specific 1219 * TX queue. 1220 * int (*ndo_get_iflink)(const struct net_device *dev); 1221 * Called to get the iflink value of this device. 1222 * void (*ndo_change_proto_down)(struct net_device *dev, 1223 * bool proto_down); 1224 * This function is used to pass protocol port error state information 1225 * to the switch driver. The switch driver can react to the proto_down 1226 * by doing a phys down on the associated switch port. 1227 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1228 * This function is used to get egress tunnel information for given skb. 1229 * This is useful for retrieving outer tunnel header parameters while 1230 * sampling packet. 1231 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1232 * This function is used to specify the headroom that the skb must 1233 * consider when allocation skb during packet reception. Setting 1234 * appropriate rx headroom value allows avoiding skb head copy on 1235 * forward. Setting a negative value resets the rx headroom to the 1236 * default value. 1237 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1238 * This function is used to set or query state related to XDP on the 1239 * netdevice and manage BPF offload. See definition of 1240 * enum bpf_netdev_command for details. 1241 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1242 * u32 flags); 1243 * This function is used to submit @n XDP packets for transmit on a 1244 * netdevice. Returns number of frames successfully transmitted, frames 1245 * that got dropped are freed/returned via xdp_return_frame(). 1246 * Returns negative number, means general error invoking ndo, meaning 1247 * no frames were xmit'ed and core-caller will free all frames. 1248 */ 1249 struct net_device_ops { 1250 int (*ndo_init)(struct net_device *dev); 1251 void (*ndo_uninit)(struct net_device *dev); 1252 int (*ndo_open)(struct net_device *dev); 1253 int (*ndo_stop)(struct net_device *dev); 1254 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1255 struct net_device *dev); 1256 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1257 struct net_device *dev, 1258 netdev_features_t features); 1259 u16 (*ndo_select_queue)(struct net_device *dev, 1260 struct sk_buff *skb, 1261 struct net_device *sb_dev, 1262 select_queue_fallback_t fallback); 1263 void (*ndo_change_rx_flags)(struct net_device *dev, 1264 int flags); 1265 void (*ndo_set_rx_mode)(struct net_device *dev); 1266 int (*ndo_set_mac_address)(struct net_device *dev, 1267 void *addr); 1268 int (*ndo_validate_addr)(struct net_device *dev); 1269 int (*ndo_do_ioctl)(struct net_device *dev, 1270 struct ifreq *ifr, int cmd); 1271 int (*ndo_set_config)(struct net_device *dev, 1272 struct ifmap *map); 1273 int (*ndo_change_mtu)(struct net_device *dev, 1274 int new_mtu); 1275 int (*ndo_neigh_setup)(struct net_device *dev, 1276 struct neigh_parms *); 1277 void (*ndo_tx_timeout) (struct net_device *dev); 1278 1279 void (*ndo_get_stats64)(struct net_device *dev, 1280 struct rtnl_link_stats64 *storage); 1281 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1282 int (*ndo_get_offload_stats)(int attr_id, 1283 const struct net_device *dev, 1284 void *attr_data); 1285 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1286 1287 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1288 __be16 proto, u16 vid); 1289 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1290 __be16 proto, u16 vid); 1291 #ifdef CONFIG_NET_POLL_CONTROLLER 1292 void (*ndo_poll_controller)(struct net_device *dev); 1293 int (*ndo_netpoll_setup)(struct net_device *dev, 1294 struct netpoll_info *info); 1295 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1296 #endif 1297 int (*ndo_set_vf_mac)(struct net_device *dev, 1298 int queue, u8 *mac); 1299 int (*ndo_set_vf_vlan)(struct net_device *dev, 1300 int queue, u16 vlan, 1301 u8 qos, __be16 proto); 1302 int (*ndo_set_vf_rate)(struct net_device *dev, 1303 int vf, int min_tx_rate, 1304 int max_tx_rate); 1305 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1306 int vf, bool setting); 1307 int (*ndo_set_vf_trust)(struct net_device *dev, 1308 int vf, bool setting); 1309 int (*ndo_get_vf_config)(struct net_device *dev, 1310 int vf, 1311 struct ifla_vf_info *ivf); 1312 int (*ndo_set_vf_link_state)(struct net_device *dev, 1313 int vf, int link_state); 1314 int (*ndo_get_vf_stats)(struct net_device *dev, 1315 int vf, 1316 struct ifla_vf_stats 1317 *vf_stats); 1318 int (*ndo_set_vf_port)(struct net_device *dev, 1319 int vf, 1320 struct nlattr *port[]); 1321 int (*ndo_get_vf_port)(struct net_device *dev, 1322 int vf, struct sk_buff *skb); 1323 int (*ndo_set_vf_guid)(struct net_device *dev, 1324 int vf, u64 guid, 1325 int guid_type); 1326 int (*ndo_set_vf_rss_query_en)( 1327 struct net_device *dev, 1328 int vf, bool setting); 1329 int (*ndo_setup_tc)(struct net_device *dev, 1330 enum tc_setup_type type, 1331 void *type_data); 1332 #if IS_ENABLED(CONFIG_FCOE) 1333 int (*ndo_fcoe_enable)(struct net_device *dev); 1334 int (*ndo_fcoe_disable)(struct net_device *dev); 1335 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1336 u16 xid, 1337 struct scatterlist *sgl, 1338 unsigned int sgc); 1339 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1340 u16 xid); 1341 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1342 u16 xid, 1343 struct scatterlist *sgl, 1344 unsigned int sgc); 1345 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1346 struct netdev_fcoe_hbainfo *hbainfo); 1347 #endif 1348 1349 #if IS_ENABLED(CONFIG_LIBFCOE) 1350 #define NETDEV_FCOE_WWNN 0 1351 #define NETDEV_FCOE_WWPN 1 1352 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1353 u64 *wwn, int type); 1354 #endif 1355 1356 #ifdef CONFIG_RFS_ACCEL 1357 int (*ndo_rx_flow_steer)(struct net_device *dev, 1358 const struct sk_buff *skb, 1359 u16 rxq_index, 1360 u32 flow_id); 1361 #endif 1362 int (*ndo_add_slave)(struct net_device *dev, 1363 struct net_device *slave_dev, 1364 struct netlink_ext_ack *extack); 1365 int (*ndo_del_slave)(struct net_device *dev, 1366 struct net_device *slave_dev); 1367 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1368 netdev_features_t features); 1369 int (*ndo_set_features)(struct net_device *dev, 1370 netdev_features_t features); 1371 int (*ndo_neigh_construct)(struct net_device *dev, 1372 struct neighbour *n); 1373 void (*ndo_neigh_destroy)(struct net_device *dev, 1374 struct neighbour *n); 1375 1376 int (*ndo_fdb_add)(struct ndmsg *ndm, 1377 struct nlattr *tb[], 1378 struct net_device *dev, 1379 const unsigned char *addr, 1380 u16 vid, 1381 u16 flags, 1382 struct netlink_ext_ack *extack); 1383 int (*ndo_fdb_del)(struct ndmsg *ndm, 1384 struct nlattr *tb[], 1385 struct net_device *dev, 1386 const unsigned char *addr, 1387 u16 vid); 1388 int (*ndo_fdb_dump)(struct sk_buff *skb, 1389 struct netlink_callback *cb, 1390 struct net_device *dev, 1391 struct net_device *filter_dev, 1392 int *idx); 1393 int (*ndo_fdb_get)(struct sk_buff *skb, 1394 struct nlattr *tb[], 1395 struct net_device *dev, 1396 const unsigned char *addr, 1397 u16 vid, u32 portid, u32 seq, 1398 struct netlink_ext_ack *extack); 1399 int (*ndo_bridge_setlink)(struct net_device *dev, 1400 struct nlmsghdr *nlh, 1401 u16 flags, 1402 struct netlink_ext_ack *extack); 1403 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1404 u32 pid, u32 seq, 1405 struct net_device *dev, 1406 u32 filter_mask, 1407 int nlflags); 1408 int (*ndo_bridge_dellink)(struct net_device *dev, 1409 struct nlmsghdr *nlh, 1410 u16 flags); 1411 int (*ndo_change_carrier)(struct net_device *dev, 1412 bool new_carrier); 1413 int (*ndo_get_phys_port_id)(struct net_device *dev, 1414 struct netdev_phys_item_id *ppid); 1415 int (*ndo_get_phys_port_name)(struct net_device *dev, 1416 char *name, size_t len); 1417 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1418 struct udp_tunnel_info *ti); 1419 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1420 struct udp_tunnel_info *ti); 1421 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1422 struct net_device *dev); 1423 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1424 void *priv); 1425 1426 int (*ndo_get_lock_subclass)(struct net_device *dev); 1427 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1428 int queue_index, 1429 u32 maxrate); 1430 int (*ndo_get_iflink)(const struct net_device *dev); 1431 int (*ndo_change_proto_down)(struct net_device *dev, 1432 bool proto_down); 1433 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1434 struct sk_buff *skb); 1435 void (*ndo_set_rx_headroom)(struct net_device *dev, 1436 int needed_headroom); 1437 int (*ndo_bpf)(struct net_device *dev, 1438 struct netdev_bpf *bpf); 1439 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1440 struct xdp_frame **xdp, 1441 u32 flags); 1442 int (*ndo_xsk_async_xmit)(struct net_device *dev, 1443 u32 queue_id); 1444 }; 1445 1446 /** 1447 * enum net_device_priv_flags - &struct net_device priv_flags 1448 * 1449 * These are the &struct net_device, they are only set internally 1450 * by drivers and used in the kernel. These flags are invisible to 1451 * userspace; this means that the order of these flags can change 1452 * during any kernel release. 1453 * 1454 * You should have a pretty good reason to be extending these flags. 1455 * 1456 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1457 * @IFF_EBRIDGE: Ethernet bridging device 1458 * @IFF_BONDING: bonding master or slave 1459 * @IFF_ISATAP: ISATAP interface (RFC4214) 1460 * @IFF_WAN_HDLC: WAN HDLC device 1461 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1462 * release skb->dst 1463 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1464 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1465 * @IFF_MACVLAN_PORT: device used as macvlan port 1466 * @IFF_BRIDGE_PORT: device used as bridge port 1467 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1468 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1469 * @IFF_UNICAST_FLT: Supports unicast filtering 1470 * @IFF_TEAM_PORT: device used as team port 1471 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1472 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1473 * change when it's running 1474 * @IFF_MACVLAN: Macvlan device 1475 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1476 * underlying stacked devices 1477 * @IFF_L3MDEV_MASTER: device is an L3 master device 1478 * @IFF_NO_QUEUE: device can run without qdisc attached 1479 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1480 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1481 * @IFF_TEAM: device is a team device 1482 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1483 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1484 * entity (i.e. the master device for bridged veth) 1485 * @IFF_MACSEC: device is a MACsec device 1486 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1487 * @IFF_FAILOVER: device is a failover master device 1488 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1489 */ 1490 enum netdev_priv_flags { 1491 IFF_802_1Q_VLAN = 1<<0, 1492 IFF_EBRIDGE = 1<<1, 1493 IFF_BONDING = 1<<2, 1494 IFF_ISATAP = 1<<3, 1495 IFF_WAN_HDLC = 1<<4, 1496 IFF_XMIT_DST_RELEASE = 1<<5, 1497 IFF_DONT_BRIDGE = 1<<6, 1498 IFF_DISABLE_NETPOLL = 1<<7, 1499 IFF_MACVLAN_PORT = 1<<8, 1500 IFF_BRIDGE_PORT = 1<<9, 1501 IFF_OVS_DATAPATH = 1<<10, 1502 IFF_TX_SKB_SHARING = 1<<11, 1503 IFF_UNICAST_FLT = 1<<12, 1504 IFF_TEAM_PORT = 1<<13, 1505 IFF_SUPP_NOFCS = 1<<14, 1506 IFF_LIVE_ADDR_CHANGE = 1<<15, 1507 IFF_MACVLAN = 1<<16, 1508 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1509 IFF_L3MDEV_MASTER = 1<<18, 1510 IFF_NO_QUEUE = 1<<19, 1511 IFF_OPENVSWITCH = 1<<20, 1512 IFF_L3MDEV_SLAVE = 1<<21, 1513 IFF_TEAM = 1<<22, 1514 IFF_RXFH_CONFIGURED = 1<<23, 1515 IFF_PHONY_HEADROOM = 1<<24, 1516 IFF_MACSEC = 1<<25, 1517 IFF_NO_RX_HANDLER = 1<<26, 1518 IFF_FAILOVER = 1<<27, 1519 IFF_FAILOVER_SLAVE = 1<<28, 1520 }; 1521 1522 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1523 #define IFF_EBRIDGE IFF_EBRIDGE 1524 #define IFF_BONDING IFF_BONDING 1525 #define IFF_ISATAP IFF_ISATAP 1526 #define IFF_WAN_HDLC IFF_WAN_HDLC 1527 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1528 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1529 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1530 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1531 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1532 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1533 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1534 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1535 #define IFF_TEAM_PORT IFF_TEAM_PORT 1536 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1537 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1538 #define IFF_MACVLAN IFF_MACVLAN 1539 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1540 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1541 #define IFF_NO_QUEUE IFF_NO_QUEUE 1542 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1543 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1544 #define IFF_TEAM IFF_TEAM 1545 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1546 #define IFF_MACSEC IFF_MACSEC 1547 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1548 #define IFF_FAILOVER IFF_FAILOVER 1549 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1550 1551 /** 1552 * struct net_device - The DEVICE structure. 1553 * 1554 * Actually, this whole structure is a big mistake. It mixes I/O 1555 * data with strictly "high-level" data, and it has to know about 1556 * almost every data structure used in the INET module. 1557 * 1558 * @name: This is the first field of the "visible" part of this structure 1559 * (i.e. as seen by users in the "Space.c" file). It is the name 1560 * of the interface. 1561 * 1562 * @name_hlist: Device name hash chain, please keep it close to name[] 1563 * @ifalias: SNMP alias 1564 * @mem_end: Shared memory end 1565 * @mem_start: Shared memory start 1566 * @base_addr: Device I/O address 1567 * @irq: Device IRQ number 1568 * 1569 * @state: Generic network queuing layer state, see netdev_state_t 1570 * @dev_list: The global list of network devices 1571 * @napi_list: List entry used for polling NAPI devices 1572 * @unreg_list: List entry when we are unregistering the 1573 * device; see the function unregister_netdev 1574 * @close_list: List entry used when we are closing the device 1575 * @ptype_all: Device-specific packet handlers for all protocols 1576 * @ptype_specific: Device-specific, protocol-specific packet handlers 1577 * 1578 * @adj_list: Directly linked devices, like slaves for bonding 1579 * @features: Currently active device features 1580 * @hw_features: User-changeable features 1581 * 1582 * @wanted_features: User-requested features 1583 * @vlan_features: Mask of features inheritable by VLAN devices 1584 * 1585 * @hw_enc_features: Mask of features inherited by encapsulating devices 1586 * This field indicates what encapsulation 1587 * offloads the hardware is capable of doing, 1588 * and drivers will need to set them appropriately. 1589 * 1590 * @mpls_features: Mask of features inheritable by MPLS 1591 * 1592 * @ifindex: interface index 1593 * @group: The group the device belongs to 1594 * 1595 * @stats: Statistics struct, which was left as a legacy, use 1596 * rtnl_link_stats64 instead 1597 * 1598 * @rx_dropped: Dropped packets by core network, 1599 * do not use this in drivers 1600 * @tx_dropped: Dropped packets by core network, 1601 * do not use this in drivers 1602 * @rx_nohandler: nohandler dropped packets by core network on 1603 * inactive devices, do not use this in drivers 1604 * @carrier_up_count: Number of times the carrier has been up 1605 * @carrier_down_count: Number of times the carrier has been down 1606 * 1607 * @wireless_handlers: List of functions to handle Wireless Extensions, 1608 * instead of ioctl, 1609 * see <net/iw_handler.h> for details. 1610 * @wireless_data: Instance data managed by the core of wireless extensions 1611 * 1612 * @netdev_ops: Includes several pointers to callbacks, 1613 * if one wants to override the ndo_*() functions 1614 * @ethtool_ops: Management operations 1615 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1616 * discovery handling. Necessary for e.g. 6LoWPAN. 1617 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1618 * of Layer 2 headers. 1619 * 1620 * @flags: Interface flags (a la BSD) 1621 * @priv_flags: Like 'flags' but invisible to userspace, 1622 * see if.h for the definitions 1623 * @gflags: Global flags ( kept as legacy ) 1624 * @padded: How much padding added by alloc_netdev() 1625 * @operstate: RFC2863 operstate 1626 * @link_mode: Mapping policy to operstate 1627 * @if_port: Selectable AUI, TP, ... 1628 * @dma: DMA channel 1629 * @mtu: Interface MTU value 1630 * @min_mtu: Interface Minimum MTU value 1631 * @max_mtu: Interface Maximum MTU value 1632 * @type: Interface hardware type 1633 * @hard_header_len: Maximum hardware header length. 1634 * @min_header_len: Minimum hardware header length 1635 * 1636 * @needed_headroom: Extra headroom the hardware may need, but not in all 1637 * cases can this be guaranteed 1638 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1639 * cases can this be guaranteed. Some cases also use 1640 * LL_MAX_HEADER instead to allocate the skb 1641 * 1642 * interface address info: 1643 * 1644 * @perm_addr: Permanent hw address 1645 * @addr_assign_type: Hw address assignment type 1646 * @addr_len: Hardware address length 1647 * @neigh_priv_len: Used in neigh_alloc() 1648 * @dev_id: Used to differentiate devices that share 1649 * the same link layer address 1650 * @dev_port: Used to differentiate devices that share 1651 * the same function 1652 * @addr_list_lock: XXX: need comments on this one 1653 * @uc_promisc: Counter that indicates promiscuous mode 1654 * has been enabled due to the need to listen to 1655 * additional unicast addresses in a device that 1656 * does not implement ndo_set_rx_mode() 1657 * @uc: unicast mac addresses 1658 * @mc: multicast mac addresses 1659 * @dev_addrs: list of device hw addresses 1660 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1661 * @promiscuity: Number of times the NIC is told to work in 1662 * promiscuous mode; if it becomes 0 the NIC will 1663 * exit promiscuous mode 1664 * @allmulti: Counter, enables or disables allmulticast mode 1665 * 1666 * @vlan_info: VLAN info 1667 * @dsa_ptr: dsa specific data 1668 * @tipc_ptr: TIPC specific data 1669 * @atalk_ptr: AppleTalk link 1670 * @ip_ptr: IPv4 specific data 1671 * @dn_ptr: DECnet specific data 1672 * @ip6_ptr: IPv6 specific data 1673 * @ax25_ptr: AX.25 specific data 1674 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1675 * 1676 * @dev_addr: Hw address (before bcast, 1677 * because most packets are unicast) 1678 * 1679 * @_rx: Array of RX queues 1680 * @num_rx_queues: Number of RX queues 1681 * allocated at register_netdev() time 1682 * @real_num_rx_queues: Number of RX queues currently active in device 1683 * 1684 * @rx_handler: handler for received packets 1685 * @rx_handler_data: XXX: need comments on this one 1686 * @miniq_ingress: ingress/clsact qdisc specific data for 1687 * ingress processing 1688 * @ingress_queue: XXX: need comments on this one 1689 * @broadcast: hw bcast address 1690 * 1691 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1692 * indexed by RX queue number. Assigned by driver. 1693 * This must only be set if the ndo_rx_flow_steer 1694 * operation is defined 1695 * @index_hlist: Device index hash chain 1696 * 1697 * @_tx: Array of TX queues 1698 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1699 * @real_num_tx_queues: Number of TX queues currently active in device 1700 * @qdisc: Root qdisc from userspace point of view 1701 * @tx_queue_len: Max frames per queue allowed 1702 * @tx_global_lock: XXX: need comments on this one 1703 * 1704 * @xps_maps: XXX: need comments on this one 1705 * @miniq_egress: clsact qdisc specific data for 1706 * egress processing 1707 * @watchdog_timeo: Represents the timeout that is used by 1708 * the watchdog (see dev_watchdog()) 1709 * @watchdog_timer: List of timers 1710 * 1711 * @pcpu_refcnt: Number of references to this device 1712 * @todo_list: Delayed register/unregister 1713 * @link_watch_list: XXX: need comments on this one 1714 * 1715 * @reg_state: Register/unregister state machine 1716 * @dismantle: Device is going to be freed 1717 * @rtnl_link_state: This enum represents the phases of creating 1718 * a new link 1719 * 1720 * @needs_free_netdev: Should unregister perform free_netdev? 1721 * @priv_destructor: Called from unregister 1722 * @npinfo: XXX: need comments on this one 1723 * @nd_net: Network namespace this network device is inside 1724 * 1725 * @ml_priv: Mid-layer private 1726 * @lstats: Loopback statistics 1727 * @tstats: Tunnel statistics 1728 * @dstats: Dummy statistics 1729 * @vstats: Virtual ethernet statistics 1730 * 1731 * @garp_port: GARP 1732 * @mrp_port: MRP 1733 * 1734 * @dev: Class/net/name entry 1735 * @sysfs_groups: Space for optional device, statistics and wireless 1736 * sysfs groups 1737 * 1738 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1739 * @rtnl_link_ops: Rtnl_link_ops 1740 * 1741 * @gso_max_size: Maximum size of generic segmentation offload 1742 * @gso_max_segs: Maximum number of segments that can be passed to the 1743 * NIC for GSO 1744 * 1745 * @dcbnl_ops: Data Center Bridging netlink ops 1746 * @num_tc: Number of traffic classes in the net device 1747 * @tc_to_txq: XXX: need comments on this one 1748 * @prio_tc_map: XXX: need comments on this one 1749 * 1750 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1751 * 1752 * @priomap: XXX: need comments on this one 1753 * @phydev: Physical device may attach itself 1754 * for hardware timestamping 1755 * @sfp_bus: attached &struct sfp_bus structure. 1756 * 1757 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1758 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1759 * 1760 * @proto_down: protocol port state information can be sent to the 1761 * switch driver and used to set the phys state of the 1762 * switch port. 1763 * 1764 * @wol_enabled: Wake-on-LAN is enabled 1765 * 1766 * FIXME: cleanup struct net_device such that network protocol info 1767 * moves out. 1768 */ 1769 1770 struct net_device { 1771 char name[IFNAMSIZ]; 1772 struct hlist_node name_hlist; 1773 struct dev_ifalias __rcu *ifalias; 1774 /* 1775 * I/O specific fields 1776 * FIXME: Merge these and struct ifmap into one 1777 */ 1778 unsigned long mem_end; 1779 unsigned long mem_start; 1780 unsigned long base_addr; 1781 int irq; 1782 1783 /* 1784 * Some hardware also needs these fields (state,dev_list, 1785 * napi_list,unreg_list,close_list) but they are not 1786 * part of the usual set specified in Space.c. 1787 */ 1788 1789 unsigned long state; 1790 1791 struct list_head dev_list; 1792 struct list_head napi_list; 1793 struct list_head unreg_list; 1794 struct list_head close_list; 1795 struct list_head ptype_all; 1796 struct list_head ptype_specific; 1797 1798 struct { 1799 struct list_head upper; 1800 struct list_head lower; 1801 } adj_list; 1802 1803 netdev_features_t features; 1804 netdev_features_t hw_features; 1805 netdev_features_t wanted_features; 1806 netdev_features_t vlan_features; 1807 netdev_features_t hw_enc_features; 1808 netdev_features_t mpls_features; 1809 netdev_features_t gso_partial_features; 1810 1811 int ifindex; 1812 int group; 1813 1814 struct net_device_stats stats; 1815 1816 atomic_long_t rx_dropped; 1817 atomic_long_t tx_dropped; 1818 atomic_long_t rx_nohandler; 1819 1820 /* Stats to monitor link on/off, flapping */ 1821 atomic_t carrier_up_count; 1822 atomic_t carrier_down_count; 1823 1824 #ifdef CONFIG_WIRELESS_EXT 1825 const struct iw_handler_def *wireless_handlers; 1826 struct iw_public_data *wireless_data; 1827 #endif 1828 const struct net_device_ops *netdev_ops; 1829 const struct ethtool_ops *ethtool_ops; 1830 #ifdef CONFIG_NET_SWITCHDEV 1831 const struct switchdev_ops *switchdev_ops; 1832 #endif 1833 #ifdef CONFIG_NET_L3_MASTER_DEV 1834 const struct l3mdev_ops *l3mdev_ops; 1835 #endif 1836 #if IS_ENABLED(CONFIG_IPV6) 1837 const struct ndisc_ops *ndisc_ops; 1838 #endif 1839 1840 #ifdef CONFIG_XFRM_OFFLOAD 1841 const struct xfrmdev_ops *xfrmdev_ops; 1842 #endif 1843 1844 #if IS_ENABLED(CONFIG_TLS_DEVICE) 1845 const struct tlsdev_ops *tlsdev_ops; 1846 #endif 1847 1848 const struct header_ops *header_ops; 1849 1850 unsigned int flags; 1851 unsigned int priv_flags; 1852 1853 unsigned short gflags; 1854 unsigned short padded; 1855 1856 unsigned char operstate; 1857 unsigned char link_mode; 1858 1859 unsigned char if_port; 1860 unsigned char dma; 1861 1862 unsigned int mtu; 1863 unsigned int min_mtu; 1864 unsigned int max_mtu; 1865 unsigned short type; 1866 unsigned short hard_header_len; 1867 unsigned char min_header_len; 1868 1869 unsigned short needed_headroom; 1870 unsigned short needed_tailroom; 1871 1872 /* Interface address info. */ 1873 unsigned char perm_addr[MAX_ADDR_LEN]; 1874 unsigned char addr_assign_type; 1875 unsigned char addr_len; 1876 unsigned short neigh_priv_len; 1877 unsigned short dev_id; 1878 unsigned short dev_port; 1879 spinlock_t addr_list_lock; 1880 unsigned char name_assign_type; 1881 bool uc_promisc; 1882 struct netdev_hw_addr_list uc; 1883 struct netdev_hw_addr_list mc; 1884 struct netdev_hw_addr_list dev_addrs; 1885 1886 #ifdef CONFIG_SYSFS 1887 struct kset *queues_kset; 1888 #endif 1889 unsigned int promiscuity; 1890 unsigned int allmulti; 1891 1892 1893 /* Protocol-specific pointers */ 1894 1895 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1896 struct vlan_info __rcu *vlan_info; 1897 #endif 1898 #if IS_ENABLED(CONFIG_NET_DSA) 1899 struct dsa_port *dsa_ptr; 1900 #endif 1901 #if IS_ENABLED(CONFIG_TIPC) 1902 struct tipc_bearer __rcu *tipc_ptr; 1903 #endif 1904 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 1905 void *atalk_ptr; 1906 #endif 1907 struct in_device __rcu *ip_ptr; 1908 #if IS_ENABLED(CONFIG_DECNET) 1909 struct dn_dev __rcu *dn_ptr; 1910 #endif 1911 struct inet6_dev __rcu *ip6_ptr; 1912 #if IS_ENABLED(CONFIG_AX25) 1913 void *ax25_ptr; 1914 #endif 1915 struct wireless_dev *ieee80211_ptr; 1916 struct wpan_dev *ieee802154_ptr; 1917 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 1918 struct mpls_dev __rcu *mpls_ptr; 1919 #endif 1920 1921 /* 1922 * Cache lines mostly used on receive path (including eth_type_trans()) 1923 */ 1924 /* Interface address info used in eth_type_trans() */ 1925 unsigned char *dev_addr; 1926 1927 struct netdev_rx_queue *_rx; 1928 unsigned int num_rx_queues; 1929 unsigned int real_num_rx_queues; 1930 1931 struct bpf_prog __rcu *xdp_prog; 1932 unsigned long gro_flush_timeout; 1933 rx_handler_func_t __rcu *rx_handler; 1934 void __rcu *rx_handler_data; 1935 1936 #ifdef CONFIG_NET_CLS_ACT 1937 struct mini_Qdisc __rcu *miniq_ingress; 1938 #endif 1939 struct netdev_queue __rcu *ingress_queue; 1940 #ifdef CONFIG_NETFILTER_INGRESS 1941 struct nf_hook_entries __rcu *nf_hooks_ingress; 1942 #endif 1943 1944 unsigned char broadcast[MAX_ADDR_LEN]; 1945 #ifdef CONFIG_RFS_ACCEL 1946 struct cpu_rmap *rx_cpu_rmap; 1947 #endif 1948 struct hlist_node index_hlist; 1949 1950 /* 1951 * Cache lines mostly used on transmit path 1952 */ 1953 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1954 unsigned int num_tx_queues; 1955 unsigned int real_num_tx_queues; 1956 struct Qdisc *qdisc; 1957 #ifdef CONFIG_NET_SCHED 1958 DECLARE_HASHTABLE (qdisc_hash, 4); 1959 #endif 1960 unsigned int tx_queue_len; 1961 spinlock_t tx_global_lock; 1962 int watchdog_timeo; 1963 1964 #ifdef CONFIG_XPS 1965 struct xps_dev_maps __rcu *xps_cpus_map; 1966 struct xps_dev_maps __rcu *xps_rxqs_map; 1967 #endif 1968 #ifdef CONFIG_NET_CLS_ACT 1969 struct mini_Qdisc __rcu *miniq_egress; 1970 #endif 1971 1972 /* These may be needed for future network-power-down code. */ 1973 struct timer_list watchdog_timer; 1974 1975 int __percpu *pcpu_refcnt; 1976 struct list_head todo_list; 1977 1978 struct list_head link_watch_list; 1979 1980 enum { NETREG_UNINITIALIZED=0, 1981 NETREG_REGISTERED, /* completed register_netdevice */ 1982 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1983 NETREG_UNREGISTERED, /* completed unregister todo */ 1984 NETREG_RELEASED, /* called free_netdev */ 1985 NETREG_DUMMY, /* dummy device for NAPI poll */ 1986 } reg_state:8; 1987 1988 bool dismantle; 1989 1990 enum { 1991 RTNL_LINK_INITIALIZED, 1992 RTNL_LINK_INITIALIZING, 1993 } rtnl_link_state:16; 1994 1995 bool needs_free_netdev; 1996 void (*priv_destructor)(struct net_device *dev); 1997 1998 #ifdef CONFIG_NETPOLL 1999 struct netpoll_info __rcu *npinfo; 2000 #endif 2001 2002 possible_net_t nd_net; 2003 2004 /* mid-layer private */ 2005 union { 2006 void *ml_priv; 2007 struct pcpu_lstats __percpu *lstats; 2008 struct pcpu_sw_netstats __percpu *tstats; 2009 struct pcpu_dstats __percpu *dstats; 2010 }; 2011 2012 #if IS_ENABLED(CONFIG_GARP) 2013 struct garp_port __rcu *garp_port; 2014 #endif 2015 #if IS_ENABLED(CONFIG_MRP) 2016 struct mrp_port __rcu *mrp_port; 2017 #endif 2018 2019 struct device dev; 2020 const struct attribute_group *sysfs_groups[4]; 2021 const struct attribute_group *sysfs_rx_queue_group; 2022 2023 const struct rtnl_link_ops *rtnl_link_ops; 2024 2025 /* for setting kernel sock attribute on TCP connection setup */ 2026 #define GSO_MAX_SIZE 65536 2027 unsigned int gso_max_size; 2028 #define GSO_MAX_SEGS 65535 2029 u16 gso_max_segs; 2030 2031 #ifdef CONFIG_DCB 2032 const struct dcbnl_rtnl_ops *dcbnl_ops; 2033 #endif 2034 s16 num_tc; 2035 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2036 u8 prio_tc_map[TC_BITMASK + 1]; 2037 2038 #if IS_ENABLED(CONFIG_FCOE) 2039 unsigned int fcoe_ddp_xid; 2040 #endif 2041 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2042 struct netprio_map __rcu *priomap; 2043 #endif 2044 struct phy_device *phydev; 2045 struct sfp_bus *sfp_bus; 2046 struct lock_class_key *qdisc_tx_busylock; 2047 struct lock_class_key *qdisc_running_key; 2048 bool proto_down; 2049 unsigned wol_enabled:1; 2050 }; 2051 #define to_net_dev(d) container_of(d, struct net_device, dev) 2052 2053 static inline bool netif_elide_gro(const struct net_device *dev) 2054 { 2055 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2056 return true; 2057 return false; 2058 } 2059 2060 #define NETDEV_ALIGN 32 2061 2062 static inline 2063 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2064 { 2065 return dev->prio_tc_map[prio & TC_BITMASK]; 2066 } 2067 2068 static inline 2069 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2070 { 2071 if (tc >= dev->num_tc) 2072 return -EINVAL; 2073 2074 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2075 return 0; 2076 } 2077 2078 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2079 void netdev_reset_tc(struct net_device *dev); 2080 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2081 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2082 2083 static inline 2084 int netdev_get_num_tc(struct net_device *dev) 2085 { 2086 return dev->num_tc; 2087 } 2088 2089 void netdev_unbind_sb_channel(struct net_device *dev, 2090 struct net_device *sb_dev); 2091 int netdev_bind_sb_channel_queue(struct net_device *dev, 2092 struct net_device *sb_dev, 2093 u8 tc, u16 count, u16 offset); 2094 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2095 static inline int netdev_get_sb_channel(struct net_device *dev) 2096 { 2097 return max_t(int, -dev->num_tc, 0); 2098 } 2099 2100 static inline 2101 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2102 unsigned int index) 2103 { 2104 return &dev->_tx[index]; 2105 } 2106 2107 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2108 const struct sk_buff *skb) 2109 { 2110 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2111 } 2112 2113 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2114 void (*f)(struct net_device *, 2115 struct netdev_queue *, 2116 void *), 2117 void *arg) 2118 { 2119 unsigned int i; 2120 2121 for (i = 0; i < dev->num_tx_queues; i++) 2122 f(dev, &dev->_tx[i], arg); 2123 } 2124 2125 #define netdev_lockdep_set_classes(dev) \ 2126 { \ 2127 static struct lock_class_key qdisc_tx_busylock_key; \ 2128 static struct lock_class_key qdisc_running_key; \ 2129 static struct lock_class_key qdisc_xmit_lock_key; \ 2130 static struct lock_class_key dev_addr_list_lock_key; \ 2131 unsigned int i; \ 2132 \ 2133 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2134 (dev)->qdisc_running_key = &qdisc_running_key; \ 2135 lockdep_set_class(&(dev)->addr_list_lock, \ 2136 &dev_addr_list_lock_key); \ 2137 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2138 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2139 &qdisc_xmit_lock_key); \ 2140 } 2141 2142 struct netdev_queue *netdev_pick_tx(struct net_device *dev, 2143 struct sk_buff *skb, 2144 struct net_device *sb_dev); 2145 2146 /* returns the headroom that the master device needs to take in account 2147 * when forwarding to this dev 2148 */ 2149 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2150 { 2151 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2152 } 2153 2154 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2155 { 2156 if (dev->netdev_ops->ndo_set_rx_headroom) 2157 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2158 } 2159 2160 /* set the device rx headroom to the dev's default */ 2161 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2162 { 2163 netdev_set_rx_headroom(dev, -1); 2164 } 2165 2166 /* 2167 * Net namespace inlines 2168 */ 2169 static inline 2170 struct net *dev_net(const struct net_device *dev) 2171 { 2172 return read_pnet(&dev->nd_net); 2173 } 2174 2175 static inline 2176 void dev_net_set(struct net_device *dev, struct net *net) 2177 { 2178 write_pnet(&dev->nd_net, net); 2179 } 2180 2181 /** 2182 * netdev_priv - access network device private data 2183 * @dev: network device 2184 * 2185 * Get network device private data 2186 */ 2187 static inline void *netdev_priv(const struct net_device *dev) 2188 { 2189 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2190 } 2191 2192 /* Set the sysfs physical device reference for the network logical device 2193 * if set prior to registration will cause a symlink during initialization. 2194 */ 2195 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2196 2197 /* Set the sysfs device type for the network logical device to allow 2198 * fine-grained identification of different network device types. For 2199 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2200 */ 2201 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2202 2203 /* Default NAPI poll() weight 2204 * Device drivers are strongly advised to not use bigger value 2205 */ 2206 #define NAPI_POLL_WEIGHT 64 2207 2208 /** 2209 * netif_napi_add - initialize a NAPI context 2210 * @dev: network device 2211 * @napi: NAPI context 2212 * @poll: polling function 2213 * @weight: default weight 2214 * 2215 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2216 * *any* of the other NAPI-related functions. 2217 */ 2218 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2219 int (*poll)(struct napi_struct *, int), int weight); 2220 2221 /** 2222 * netif_tx_napi_add - initialize a NAPI context 2223 * @dev: network device 2224 * @napi: NAPI context 2225 * @poll: polling function 2226 * @weight: default weight 2227 * 2228 * This variant of netif_napi_add() should be used from drivers using NAPI 2229 * to exclusively poll a TX queue. 2230 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2231 */ 2232 static inline void netif_tx_napi_add(struct net_device *dev, 2233 struct napi_struct *napi, 2234 int (*poll)(struct napi_struct *, int), 2235 int weight) 2236 { 2237 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2238 netif_napi_add(dev, napi, poll, weight); 2239 } 2240 2241 /** 2242 * netif_napi_del - remove a NAPI context 2243 * @napi: NAPI context 2244 * 2245 * netif_napi_del() removes a NAPI context from the network device NAPI list 2246 */ 2247 void netif_napi_del(struct napi_struct *napi); 2248 2249 struct napi_gro_cb { 2250 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2251 void *frag0; 2252 2253 /* Length of frag0. */ 2254 unsigned int frag0_len; 2255 2256 /* This indicates where we are processing relative to skb->data. */ 2257 int data_offset; 2258 2259 /* This is non-zero if the packet cannot be merged with the new skb. */ 2260 u16 flush; 2261 2262 /* Save the IP ID here and check when we get to the transport layer */ 2263 u16 flush_id; 2264 2265 /* Number of segments aggregated. */ 2266 u16 count; 2267 2268 /* Start offset for remote checksum offload */ 2269 u16 gro_remcsum_start; 2270 2271 /* jiffies when first packet was created/queued */ 2272 unsigned long age; 2273 2274 /* Used in ipv6_gro_receive() and foo-over-udp */ 2275 u16 proto; 2276 2277 /* This is non-zero if the packet may be of the same flow. */ 2278 u8 same_flow:1; 2279 2280 /* Used in tunnel GRO receive */ 2281 u8 encap_mark:1; 2282 2283 /* GRO checksum is valid */ 2284 u8 csum_valid:1; 2285 2286 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2287 u8 csum_cnt:3; 2288 2289 /* Free the skb? */ 2290 u8 free:2; 2291 #define NAPI_GRO_FREE 1 2292 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2293 2294 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2295 u8 is_ipv6:1; 2296 2297 /* Used in GRE, set in fou/gue_gro_receive */ 2298 u8 is_fou:1; 2299 2300 /* Used to determine if flush_id can be ignored */ 2301 u8 is_atomic:1; 2302 2303 /* Number of gro_receive callbacks this packet already went through */ 2304 u8 recursion_counter:4; 2305 2306 /* 1 bit hole */ 2307 2308 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2309 __wsum csum; 2310 2311 /* used in skb_gro_receive() slow path */ 2312 struct sk_buff *last; 2313 }; 2314 2315 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2316 2317 #define GRO_RECURSION_LIMIT 15 2318 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2319 { 2320 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2321 } 2322 2323 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2324 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2325 struct list_head *head, 2326 struct sk_buff *skb) 2327 { 2328 if (unlikely(gro_recursion_inc_test(skb))) { 2329 NAPI_GRO_CB(skb)->flush |= 1; 2330 return NULL; 2331 } 2332 2333 return cb(head, skb); 2334 } 2335 2336 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2337 struct sk_buff *); 2338 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2339 struct sock *sk, 2340 struct list_head *head, 2341 struct sk_buff *skb) 2342 { 2343 if (unlikely(gro_recursion_inc_test(skb))) { 2344 NAPI_GRO_CB(skb)->flush |= 1; 2345 return NULL; 2346 } 2347 2348 return cb(sk, head, skb); 2349 } 2350 2351 struct packet_type { 2352 __be16 type; /* This is really htons(ether_type). */ 2353 bool ignore_outgoing; 2354 struct net_device *dev; /* NULL is wildcarded here */ 2355 int (*func) (struct sk_buff *, 2356 struct net_device *, 2357 struct packet_type *, 2358 struct net_device *); 2359 void (*list_func) (struct list_head *, 2360 struct packet_type *, 2361 struct net_device *); 2362 bool (*id_match)(struct packet_type *ptype, 2363 struct sock *sk); 2364 void *af_packet_priv; 2365 struct list_head list; 2366 }; 2367 2368 struct offload_callbacks { 2369 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2370 netdev_features_t features); 2371 struct sk_buff *(*gro_receive)(struct list_head *head, 2372 struct sk_buff *skb); 2373 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2374 }; 2375 2376 struct packet_offload { 2377 __be16 type; /* This is really htons(ether_type). */ 2378 u16 priority; 2379 struct offload_callbacks callbacks; 2380 struct list_head list; 2381 }; 2382 2383 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2384 struct pcpu_sw_netstats { 2385 u64 rx_packets; 2386 u64 rx_bytes; 2387 u64 tx_packets; 2388 u64 tx_bytes; 2389 struct u64_stats_sync syncp; 2390 } __aligned(4 * sizeof(u64)); 2391 2392 struct pcpu_lstats { 2393 u64 packets; 2394 u64 bytes; 2395 struct u64_stats_sync syncp; 2396 } __aligned(2 * sizeof(u64)); 2397 2398 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2399 ({ \ 2400 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2401 if (pcpu_stats) { \ 2402 int __cpu; \ 2403 for_each_possible_cpu(__cpu) { \ 2404 typeof(type) *stat; \ 2405 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2406 u64_stats_init(&stat->syncp); \ 2407 } \ 2408 } \ 2409 pcpu_stats; \ 2410 }) 2411 2412 #define netdev_alloc_pcpu_stats(type) \ 2413 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2414 2415 enum netdev_lag_tx_type { 2416 NETDEV_LAG_TX_TYPE_UNKNOWN, 2417 NETDEV_LAG_TX_TYPE_RANDOM, 2418 NETDEV_LAG_TX_TYPE_BROADCAST, 2419 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2420 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2421 NETDEV_LAG_TX_TYPE_HASH, 2422 }; 2423 2424 enum netdev_lag_hash { 2425 NETDEV_LAG_HASH_NONE, 2426 NETDEV_LAG_HASH_L2, 2427 NETDEV_LAG_HASH_L34, 2428 NETDEV_LAG_HASH_L23, 2429 NETDEV_LAG_HASH_E23, 2430 NETDEV_LAG_HASH_E34, 2431 NETDEV_LAG_HASH_UNKNOWN, 2432 }; 2433 2434 struct netdev_lag_upper_info { 2435 enum netdev_lag_tx_type tx_type; 2436 enum netdev_lag_hash hash_type; 2437 }; 2438 2439 struct netdev_lag_lower_state_info { 2440 u8 link_up : 1, 2441 tx_enabled : 1; 2442 }; 2443 2444 #include <linux/notifier.h> 2445 2446 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2447 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2448 * adding new types. 2449 */ 2450 enum netdev_cmd { 2451 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2452 NETDEV_DOWN, 2453 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2454 detected a hardware crash and restarted 2455 - we can use this eg to kick tcp sessions 2456 once done */ 2457 NETDEV_CHANGE, /* Notify device state change */ 2458 NETDEV_REGISTER, 2459 NETDEV_UNREGISTER, 2460 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2461 NETDEV_CHANGEADDR, /* notify after the address change */ 2462 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2463 NETDEV_GOING_DOWN, 2464 NETDEV_CHANGENAME, 2465 NETDEV_FEAT_CHANGE, 2466 NETDEV_BONDING_FAILOVER, 2467 NETDEV_PRE_UP, 2468 NETDEV_PRE_TYPE_CHANGE, 2469 NETDEV_POST_TYPE_CHANGE, 2470 NETDEV_POST_INIT, 2471 NETDEV_RELEASE, 2472 NETDEV_NOTIFY_PEERS, 2473 NETDEV_JOIN, 2474 NETDEV_CHANGEUPPER, 2475 NETDEV_RESEND_IGMP, 2476 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2477 NETDEV_CHANGEINFODATA, 2478 NETDEV_BONDING_INFO, 2479 NETDEV_PRECHANGEUPPER, 2480 NETDEV_CHANGELOWERSTATE, 2481 NETDEV_UDP_TUNNEL_PUSH_INFO, 2482 NETDEV_UDP_TUNNEL_DROP_INFO, 2483 NETDEV_CHANGE_TX_QUEUE_LEN, 2484 NETDEV_CVLAN_FILTER_PUSH_INFO, 2485 NETDEV_CVLAN_FILTER_DROP_INFO, 2486 NETDEV_SVLAN_FILTER_PUSH_INFO, 2487 NETDEV_SVLAN_FILTER_DROP_INFO, 2488 }; 2489 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2490 2491 int register_netdevice_notifier(struct notifier_block *nb); 2492 int unregister_netdevice_notifier(struct notifier_block *nb); 2493 2494 struct netdev_notifier_info { 2495 struct net_device *dev; 2496 struct netlink_ext_ack *extack; 2497 }; 2498 2499 struct netdev_notifier_info_ext { 2500 struct netdev_notifier_info info; /* must be first */ 2501 union { 2502 u32 mtu; 2503 } ext; 2504 }; 2505 2506 struct netdev_notifier_change_info { 2507 struct netdev_notifier_info info; /* must be first */ 2508 unsigned int flags_changed; 2509 }; 2510 2511 struct netdev_notifier_changeupper_info { 2512 struct netdev_notifier_info info; /* must be first */ 2513 struct net_device *upper_dev; /* new upper dev */ 2514 bool master; /* is upper dev master */ 2515 bool linking; /* is the notification for link or unlink */ 2516 void *upper_info; /* upper dev info */ 2517 }; 2518 2519 struct netdev_notifier_changelowerstate_info { 2520 struct netdev_notifier_info info; /* must be first */ 2521 void *lower_state_info; /* is lower dev state */ 2522 }; 2523 2524 struct netdev_notifier_pre_changeaddr_info { 2525 struct netdev_notifier_info info; /* must be first */ 2526 const unsigned char *dev_addr; 2527 }; 2528 2529 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2530 struct net_device *dev) 2531 { 2532 info->dev = dev; 2533 info->extack = NULL; 2534 } 2535 2536 static inline struct net_device * 2537 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2538 { 2539 return info->dev; 2540 } 2541 2542 static inline struct netlink_ext_ack * 2543 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2544 { 2545 return info->extack; 2546 } 2547 2548 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2549 2550 2551 extern rwlock_t dev_base_lock; /* Device list lock */ 2552 2553 #define for_each_netdev(net, d) \ 2554 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2555 #define for_each_netdev_reverse(net, d) \ 2556 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2557 #define for_each_netdev_rcu(net, d) \ 2558 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2559 #define for_each_netdev_safe(net, d, n) \ 2560 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2561 #define for_each_netdev_continue(net, d) \ 2562 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2563 #define for_each_netdev_continue_rcu(net, d) \ 2564 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2565 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2566 for_each_netdev_rcu(&init_net, slave) \ 2567 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2568 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2569 2570 static inline struct net_device *next_net_device(struct net_device *dev) 2571 { 2572 struct list_head *lh; 2573 struct net *net; 2574 2575 net = dev_net(dev); 2576 lh = dev->dev_list.next; 2577 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2578 } 2579 2580 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2581 { 2582 struct list_head *lh; 2583 struct net *net; 2584 2585 net = dev_net(dev); 2586 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2587 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2588 } 2589 2590 static inline struct net_device *first_net_device(struct net *net) 2591 { 2592 return list_empty(&net->dev_base_head) ? NULL : 2593 net_device_entry(net->dev_base_head.next); 2594 } 2595 2596 static inline struct net_device *first_net_device_rcu(struct net *net) 2597 { 2598 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2599 2600 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2601 } 2602 2603 int netdev_boot_setup_check(struct net_device *dev); 2604 unsigned long netdev_boot_base(const char *prefix, int unit); 2605 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2606 const char *hwaddr); 2607 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2608 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2609 void dev_add_pack(struct packet_type *pt); 2610 void dev_remove_pack(struct packet_type *pt); 2611 void __dev_remove_pack(struct packet_type *pt); 2612 void dev_add_offload(struct packet_offload *po); 2613 void dev_remove_offload(struct packet_offload *po); 2614 2615 int dev_get_iflink(const struct net_device *dev); 2616 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2617 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2618 unsigned short mask); 2619 struct net_device *dev_get_by_name(struct net *net, const char *name); 2620 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2621 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2622 int dev_alloc_name(struct net_device *dev, const char *name); 2623 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2624 void dev_close(struct net_device *dev); 2625 void dev_close_many(struct list_head *head, bool unlink); 2626 void dev_disable_lro(struct net_device *dev); 2627 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2628 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2629 struct net_device *sb_dev, 2630 select_queue_fallback_t fallback); 2631 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2632 struct net_device *sb_dev, 2633 select_queue_fallback_t fallback); 2634 int dev_queue_xmit(struct sk_buff *skb); 2635 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2636 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2637 int register_netdevice(struct net_device *dev); 2638 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2639 void unregister_netdevice_many(struct list_head *head); 2640 static inline void unregister_netdevice(struct net_device *dev) 2641 { 2642 unregister_netdevice_queue(dev, NULL); 2643 } 2644 2645 int netdev_refcnt_read(const struct net_device *dev); 2646 void free_netdev(struct net_device *dev); 2647 void netdev_freemem(struct net_device *dev); 2648 void synchronize_net(void); 2649 int init_dummy_netdev(struct net_device *dev); 2650 2651 DECLARE_PER_CPU(int, xmit_recursion); 2652 #define XMIT_RECURSION_LIMIT 10 2653 2654 static inline int dev_recursion_level(void) 2655 { 2656 return this_cpu_read(xmit_recursion); 2657 } 2658 2659 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2660 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2661 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2662 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2663 int netdev_get_name(struct net *net, char *name, int ifindex); 2664 int dev_restart(struct net_device *dev); 2665 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 2666 2667 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2668 { 2669 return NAPI_GRO_CB(skb)->data_offset; 2670 } 2671 2672 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2673 { 2674 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2675 } 2676 2677 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2678 { 2679 NAPI_GRO_CB(skb)->data_offset += len; 2680 } 2681 2682 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2683 unsigned int offset) 2684 { 2685 return NAPI_GRO_CB(skb)->frag0 + offset; 2686 } 2687 2688 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2689 { 2690 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2691 } 2692 2693 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2694 { 2695 NAPI_GRO_CB(skb)->frag0 = NULL; 2696 NAPI_GRO_CB(skb)->frag0_len = 0; 2697 } 2698 2699 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2700 unsigned int offset) 2701 { 2702 if (!pskb_may_pull(skb, hlen)) 2703 return NULL; 2704 2705 skb_gro_frag0_invalidate(skb); 2706 return skb->data + offset; 2707 } 2708 2709 static inline void *skb_gro_network_header(struct sk_buff *skb) 2710 { 2711 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2712 skb_network_offset(skb); 2713 } 2714 2715 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2716 const void *start, unsigned int len) 2717 { 2718 if (NAPI_GRO_CB(skb)->csum_valid) 2719 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2720 csum_partial(start, len, 0)); 2721 } 2722 2723 /* GRO checksum functions. These are logical equivalents of the normal 2724 * checksum functions (in skbuff.h) except that they operate on the GRO 2725 * offsets and fields in sk_buff. 2726 */ 2727 2728 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2729 2730 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2731 { 2732 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2733 } 2734 2735 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2736 bool zero_okay, 2737 __sum16 check) 2738 { 2739 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2740 skb_checksum_start_offset(skb) < 2741 skb_gro_offset(skb)) && 2742 !skb_at_gro_remcsum_start(skb) && 2743 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2744 (!zero_okay || check)); 2745 } 2746 2747 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2748 __wsum psum) 2749 { 2750 if (NAPI_GRO_CB(skb)->csum_valid && 2751 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2752 return 0; 2753 2754 NAPI_GRO_CB(skb)->csum = psum; 2755 2756 return __skb_gro_checksum_complete(skb); 2757 } 2758 2759 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2760 { 2761 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2762 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2763 NAPI_GRO_CB(skb)->csum_cnt--; 2764 } else { 2765 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2766 * verified a new top level checksum or an encapsulated one 2767 * during GRO. This saves work if we fallback to normal path. 2768 */ 2769 __skb_incr_checksum_unnecessary(skb); 2770 } 2771 } 2772 2773 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2774 compute_pseudo) \ 2775 ({ \ 2776 __sum16 __ret = 0; \ 2777 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2778 __ret = __skb_gro_checksum_validate_complete(skb, \ 2779 compute_pseudo(skb, proto)); \ 2780 if (!__ret) \ 2781 skb_gro_incr_csum_unnecessary(skb); \ 2782 __ret; \ 2783 }) 2784 2785 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2786 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2787 2788 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2789 compute_pseudo) \ 2790 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2791 2792 #define skb_gro_checksum_simple_validate(skb) \ 2793 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2794 2795 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2796 { 2797 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2798 !NAPI_GRO_CB(skb)->csum_valid); 2799 } 2800 2801 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2802 __sum16 check, __wsum pseudo) 2803 { 2804 NAPI_GRO_CB(skb)->csum = ~pseudo; 2805 NAPI_GRO_CB(skb)->csum_valid = 1; 2806 } 2807 2808 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \ 2809 do { \ 2810 if (__skb_gro_checksum_convert_check(skb)) \ 2811 __skb_gro_checksum_convert(skb, check, \ 2812 compute_pseudo(skb, proto)); \ 2813 } while (0) 2814 2815 struct gro_remcsum { 2816 int offset; 2817 __wsum delta; 2818 }; 2819 2820 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2821 { 2822 grc->offset = 0; 2823 grc->delta = 0; 2824 } 2825 2826 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2827 unsigned int off, size_t hdrlen, 2828 int start, int offset, 2829 struct gro_remcsum *grc, 2830 bool nopartial) 2831 { 2832 __wsum delta; 2833 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2834 2835 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2836 2837 if (!nopartial) { 2838 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2839 return ptr; 2840 } 2841 2842 ptr = skb_gro_header_fast(skb, off); 2843 if (skb_gro_header_hard(skb, off + plen)) { 2844 ptr = skb_gro_header_slow(skb, off + plen, off); 2845 if (!ptr) 2846 return NULL; 2847 } 2848 2849 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2850 start, offset); 2851 2852 /* Adjust skb->csum since we changed the packet */ 2853 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2854 2855 grc->offset = off + hdrlen + offset; 2856 grc->delta = delta; 2857 2858 return ptr; 2859 } 2860 2861 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2862 struct gro_remcsum *grc) 2863 { 2864 void *ptr; 2865 size_t plen = grc->offset + sizeof(u16); 2866 2867 if (!grc->delta) 2868 return; 2869 2870 ptr = skb_gro_header_fast(skb, grc->offset); 2871 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 2872 ptr = skb_gro_header_slow(skb, plen, grc->offset); 2873 if (!ptr) 2874 return; 2875 } 2876 2877 remcsum_unadjust((__sum16 *)ptr, grc->delta); 2878 } 2879 2880 #ifdef CONFIG_XFRM_OFFLOAD 2881 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2882 { 2883 if (PTR_ERR(pp) != -EINPROGRESS) 2884 NAPI_GRO_CB(skb)->flush |= flush; 2885 } 2886 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 2887 struct sk_buff *pp, 2888 int flush, 2889 struct gro_remcsum *grc) 2890 { 2891 if (PTR_ERR(pp) != -EINPROGRESS) { 2892 NAPI_GRO_CB(skb)->flush |= flush; 2893 skb_gro_remcsum_cleanup(skb, grc); 2894 skb->remcsum_offload = 0; 2895 } 2896 } 2897 #else 2898 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2899 { 2900 NAPI_GRO_CB(skb)->flush |= flush; 2901 } 2902 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 2903 struct sk_buff *pp, 2904 int flush, 2905 struct gro_remcsum *grc) 2906 { 2907 NAPI_GRO_CB(skb)->flush |= flush; 2908 skb_gro_remcsum_cleanup(skb, grc); 2909 skb->remcsum_offload = 0; 2910 } 2911 #endif 2912 2913 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2914 unsigned short type, 2915 const void *daddr, const void *saddr, 2916 unsigned int len) 2917 { 2918 if (!dev->header_ops || !dev->header_ops->create) 2919 return 0; 2920 2921 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2922 } 2923 2924 static inline int dev_parse_header(const struct sk_buff *skb, 2925 unsigned char *haddr) 2926 { 2927 const struct net_device *dev = skb->dev; 2928 2929 if (!dev->header_ops || !dev->header_ops->parse) 2930 return 0; 2931 return dev->header_ops->parse(skb, haddr); 2932 } 2933 2934 /* ll_header must have at least hard_header_len allocated */ 2935 static inline bool dev_validate_header(const struct net_device *dev, 2936 char *ll_header, int len) 2937 { 2938 if (likely(len >= dev->hard_header_len)) 2939 return true; 2940 if (len < dev->min_header_len) 2941 return false; 2942 2943 if (capable(CAP_SYS_RAWIO)) { 2944 memset(ll_header + len, 0, dev->hard_header_len - len); 2945 return true; 2946 } 2947 2948 if (dev->header_ops && dev->header_ops->validate) 2949 return dev->header_ops->validate(ll_header, len); 2950 2951 return false; 2952 } 2953 2954 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, 2955 int len, int size); 2956 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2957 static inline int unregister_gifconf(unsigned int family) 2958 { 2959 return register_gifconf(family, NULL); 2960 } 2961 2962 #ifdef CONFIG_NET_FLOW_LIMIT 2963 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2964 struct sd_flow_limit { 2965 u64 count; 2966 unsigned int num_buckets; 2967 unsigned int history_head; 2968 u16 history[FLOW_LIMIT_HISTORY]; 2969 u8 buckets[]; 2970 }; 2971 2972 extern int netdev_flow_limit_table_len; 2973 #endif /* CONFIG_NET_FLOW_LIMIT */ 2974 2975 /* 2976 * Incoming packets are placed on per-CPU queues 2977 */ 2978 struct softnet_data { 2979 struct list_head poll_list; 2980 struct sk_buff_head process_queue; 2981 2982 /* stats */ 2983 unsigned int processed; 2984 unsigned int time_squeeze; 2985 unsigned int received_rps; 2986 #ifdef CONFIG_RPS 2987 struct softnet_data *rps_ipi_list; 2988 #endif 2989 #ifdef CONFIG_NET_FLOW_LIMIT 2990 struct sd_flow_limit __rcu *flow_limit; 2991 #endif 2992 struct Qdisc *output_queue; 2993 struct Qdisc **output_queue_tailp; 2994 struct sk_buff *completion_queue; 2995 #ifdef CONFIG_XFRM_OFFLOAD 2996 struct sk_buff_head xfrm_backlog; 2997 #endif 2998 #ifdef CONFIG_RPS 2999 /* input_queue_head should be written by cpu owning this struct, 3000 * and only read by other cpus. Worth using a cache line. 3001 */ 3002 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3003 3004 /* Elements below can be accessed between CPUs for RPS/RFS */ 3005 call_single_data_t csd ____cacheline_aligned_in_smp; 3006 struct softnet_data *rps_ipi_next; 3007 unsigned int cpu; 3008 unsigned int input_queue_tail; 3009 #endif 3010 unsigned int dropped; 3011 struct sk_buff_head input_pkt_queue; 3012 struct napi_struct backlog; 3013 3014 }; 3015 3016 static inline void input_queue_head_incr(struct softnet_data *sd) 3017 { 3018 #ifdef CONFIG_RPS 3019 sd->input_queue_head++; 3020 #endif 3021 } 3022 3023 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3024 unsigned int *qtail) 3025 { 3026 #ifdef CONFIG_RPS 3027 *qtail = ++sd->input_queue_tail; 3028 #endif 3029 } 3030 3031 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3032 3033 void __netif_schedule(struct Qdisc *q); 3034 void netif_schedule_queue(struct netdev_queue *txq); 3035 3036 static inline void netif_tx_schedule_all(struct net_device *dev) 3037 { 3038 unsigned int i; 3039 3040 for (i = 0; i < dev->num_tx_queues; i++) 3041 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3042 } 3043 3044 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3045 { 3046 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3047 } 3048 3049 /** 3050 * netif_start_queue - allow transmit 3051 * @dev: network device 3052 * 3053 * Allow upper layers to call the device hard_start_xmit routine. 3054 */ 3055 static inline void netif_start_queue(struct net_device *dev) 3056 { 3057 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3058 } 3059 3060 static inline void netif_tx_start_all_queues(struct net_device *dev) 3061 { 3062 unsigned int i; 3063 3064 for (i = 0; i < dev->num_tx_queues; i++) { 3065 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3066 netif_tx_start_queue(txq); 3067 } 3068 } 3069 3070 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3071 3072 /** 3073 * netif_wake_queue - restart transmit 3074 * @dev: network device 3075 * 3076 * Allow upper layers to call the device hard_start_xmit routine. 3077 * Used for flow control when transmit resources are available. 3078 */ 3079 static inline void netif_wake_queue(struct net_device *dev) 3080 { 3081 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3082 } 3083 3084 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3085 { 3086 unsigned int i; 3087 3088 for (i = 0; i < dev->num_tx_queues; i++) { 3089 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3090 netif_tx_wake_queue(txq); 3091 } 3092 } 3093 3094 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3095 { 3096 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3097 } 3098 3099 /** 3100 * netif_stop_queue - stop transmitted packets 3101 * @dev: network device 3102 * 3103 * Stop upper layers calling the device hard_start_xmit routine. 3104 * Used for flow control when transmit resources are unavailable. 3105 */ 3106 static inline void netif_stop_queue(struct net_device *dev) 3107 { 3108 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3109 } 3110 3111 void netif_tx_stop_all_queues(struct net_device *dev); 3112 3113 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3114 { 3115 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3116 } 3117 3118 /** 3119 * netif_queue_stopped - test if transmit queue is flowblocked 3120 * @dev: network device 3121 * 3122 * Test if transmit queue on device is currently unable to send. 3123 */ 3124 static inline bool netif_queue_stopped(const struct net_device *dev) 3125 { 3126 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3127 } 3128 3129 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3130 { 3131 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3132 } 3133 3134 static inline bool 3135 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3136 { 3137 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3138 } 3139 3140 static inline bool 3141 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3142 { 3143 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3144 } 3145 3146 /** 3147 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3148 * @dev_queue: pointer to transmit queue 3149 * 3150 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3151 * to give appropriate hint to the CPU. 3152 */ 3153 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3154 { 3155 #ifdef CONFIG_BQL 3156 prefetchw(&dev_queue->dql.num_queued); 3157 #endif 3158 } 3159 3160 /** 3161 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3162 * @dev_queue: pointer to transmit queue 3163 * 3164 * BQL enabled drivers might use this helper in their TX completion path, 3165 * to give appropriate hint to the CPU. 3166 */ 3167 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3168 { 3169 #ifdef CONFIG_BQL 3170 prefetchw(&dev_queue->dql.limit); 3171 #endif 3172 } 3173 3174 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3175 unsigned int bytes) 3176 { 3177 #ifdef CONFIG_BQL 3178 dql_queued(&dev_queue->dql, bytes); 3179 3180 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3181 return; 3182 3183 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3184 3185 /* 3186 * The XOFF flag must be set before checking the dql_avail below, 3187 * because in netdev_tx_completed_queue we update the dql_completed 3188 * before checking the XOFF flag. 3189 */ 3190 smp_mb(); 3191 3192 /* check again in case another CPU has just made room avail */ 3193 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3194 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3195 #endif 3196 } 3197 3198 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3199 * that they should not test BQL status themselves. 3200 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3201 * skb of a batch. 3202 * Returns true if the doorbell must be used to kick the NIC. 3203 */ 3204 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3205 unsigned int bytes, 3206 bool xmit_more) 3207 { 3208 if (xmit_more) { 3209 #ifdef CONFIG_BQL 3210 dql_queued(&dev_queue->dql, bytes); 3211 #endif 3212 return netif_tx_queue_stopped(dev_queue); 3213 } 3214 netdev_tx_sent_queue(dev_queue, bytes); 3215 return true; 3216 } 3217 3218 /** 3219 * netdev_sent_queue - report the number of bytes queued to hardware 3220 * @dev: network device 3221 * @bytes: number of bytes queued to the hardware device queue 3222 * 3223 * Report the number of bytes queued for sending/completion to the network 3224 * device hardware queue. @bytes should be a good approximation and should 3225 * exactly match netdev_completed_queue() @bytes 3226 */ 3227 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3228 { 3229 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3230 } 3231 3232 static inline bool __netdev_sent_queue(struct net_device *dev, 3233 unsigned int bytes, 3234 bool xmit_more) 3235 { 3236 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3237 xmit_more); 3238 } 3239 3240 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3241 unsigned int pkts, unsigned int bytes) 3242 { 3243 #ifdef CONFIG_BQL 3244 if (unlikely(!bytes)) 3245 return; 3246 3247 dql_completed(&dev_queue->dql, bytes); 3248 3249 /* 3250 * Without the memory barrier there is a small possiblity that 3251 * netdev_tx_sent_queue will miss the update and cause the queue to 3252 * be stopped forever 3253 */ 3254 smp_mb(); 3255 3256 if (dql_avail(&dev_queue->dql) < 0) 3257 return; 3258 3259 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3260 netif_schedule_queue(dev_queue); 3261 #endif 3262 } 3263 3264 /** 3265 * netdev_completed_queue - report bytes and packets completed by device 3266 * @dev: network device 3267 * @pkts: actual number of packets sent over the medium 3268 * @bytes: actual number of bytes sent over the medium 3269 * 3270 * Report the number of bytes and packets transmitted by the network device 3271 * hardware queue over the physical medium, @bytes must exactly match the 3272 * @bytes amount passed to netdev_sent_queue() 3273 */ 3274 static inline void netdev_completed_queue(struct net_device *dev, 3275 unsigned int pkts, unsigned int bytes) 3276 { 3277 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3278 } 3279 3280 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3281 { 3282 #ifdef CONFIG_BQL 3283 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3284 dql_reset(&q->dql); 3285 #endif 3286 } 3287 3288 /** 3289 * netdev_reset_queue - reset the packets and bytes count of a network device 3290 * @dev_queue: network device 3291 * 3292 * Reset the bytes and packet count of a network device and clear the 3293 * software flow control OFF bit for this network device 3294 */ 3295 static inline void netdev_reset_queue(struct net_device *dev_queue) 3296 { 3297 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3298 } 3299 3300 /** 3301 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3302 * @dev: network device 3303 * @queue_index: given tx queue index 3304 * 3305 * Returns 0 if given tx queue index >= number of device tx queues, 3306 * otherwise returns the originally passed tx queue index. 3307 */ 3308 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3309 { 3310 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3311 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3312 dev->name, queue_index, 3313 dev->real_num_tx_queues); 3314 return 0; 3315 } 3316 3317 return queue_index; 3318 } 3319 3320 /** 3321 * netif_running - test if up 3322 * @dev: network device 3323 * 3324 * Test if the device has been brought up. 3325 */ 3326 static inline bool netif_running(const struct net_device *dev) 3327 { 3328 return test_bit(__LINK_STATE_START, &dev->state); 3329 } 3330 3331 /* 3332 * Routines to manage the subqueues on a device. We only need start, 3333 * stop, and a check if it's stopped. All other device management is 3334 * done at the overall netdevice level. 3335 * Also test the device if we're multiqueue. 3336 */ 3337 3338 /** 3339 * netif_start_subqueue - allow sending packets on subqueue 3340 * @dev: network device 3341 * @queue_index: sub queue index 3342 * 3343 * Start individual transmit queue of a device with multiple transmit queues. 3344 */ 3345 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3346 { 3347 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3348 3349 netif_tx_start_queue(txq); 3350 } 3351 3352 /** 3353 * netif_stop_subqueue - stop sending packets on subqueue 3354 * @dev: network device 3355 * @queue_index: sub queue index 3356 * 3357 * Stop individual transmit queue of a device with multiple transmit queues. 3358 */ 3359 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3360 { 3361 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3362 netif_tx_stop_queue(txq); 3363 } 3364 3365 /** 3366 * netif_subqueue_stopped - test status of subqueue 3367 * @dev: network device 3368 * @queue_index: sub queue index 3369 * 3370 * Check individual transmit queue of a device with multiple transmit queues. 3371 */ 3372 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3373 u16 queue_index) 3374 { 3375 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3376 3377 return netif_tx_queue_stopped(txq); 3378 } 3379 3380 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3381 struct sk_buff *skb) 3382 { 3383 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3384 } 3385 3386 /** 3387 * netif_wake_subqueue - allow sending packets on subqueue 3388 * @dev: network device 3389 * @queue_index: sub queue index 3390 * 3391 * Resume individual transmit queue of a device with multiple transmit queues. 3392 */ 3393 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3394 { 3395 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3396 3397 netif_tx_wake_queue(txq); 3398 } 3399 3400 #ifdef CONFIG_XPS 3401 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3402 u16 index); 3403 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3404 u16 index, bool is_rxqs_map); 3405 3406 /** 3407 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3408 * @j: CPU/Rx queue index 3409 * @mask: bitmask of all cpus/rx queues 3410 * @nr_bits: number of bits in the bitmask 3411 * 3412 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3413 */ 3414 static inline bool netif_attr_test_mask(unsigned long j, 3415 const unsigned long *mask, 3416 unsigned int nr_bits) 3417 { 3418 cpu_max_bits_warn(j, nr_bits); 3419 return test_bit(j, mask); 3420 } 3421 3422 /** 3423 * netif_attr_test_online - Test for online CPU/Rx queue 3424 * @j: CPU/Rx queue index 3425 * @online_mask: bitmask for CPUs/Rx queues that are online 3426 * @nr_bits: number of bits in the bitmask 3427 * 3428 * Returns true if a CPU/Rx queue is online. 3429 */ 3430 static inline bool netif_attr_test_online(unsigned long j, 3431 const unsigned long *online_mask, 3432 unsigned int nr_bits) 3433 { 3434 cpu_max_bits_warn(j, nr_bits); 3435 3436 if (online_mask) 3437 return test_bit(j, online_mask); 3438 3439 return (j < nr_bits); 3440 } 3441 3442 /** 3443 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3444 * @n: CPU/Rx queue index 3445 * @srcp: the cpumask/Rx queue mask pointer 3446 * @nr_bits: number of bits in the bitmask 3447 * 3448 * Returns >= nr_bits if no further CPUs/Rx queues set. 3449 */ 3450 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3451 unsigned int nr_bits) 3452 { 3453 /* -1 is a legal arg here. */ 3454 if (n != -1) 3455 cpu_max_bits_warn(n, nr_bits); 3456 3457 if (srcp) 3458 return find_next_bit(srcp, nr_bits, n + 1); 3459 3460 return n + 1; 3461 } 3462 3463 /** 3464 * netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p 3465 * @n: CPU/Rx queue index 3466 * @src1p: the first CPUs/Rx queues mask pointer 3467 * @src2p: the second CPUs/Rx queues mask pointer 3468 * @nr_bits: number of bits in the bitmask 3469 * 3470 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3471 */ 3472 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3473 const unsigned long *src2p, 3474 unsigned int nr_bits) 3475 { 3476 /* -1 is a legal arg here. */ 3477 if (n != -1) 3478 cpu_max_bits_warn(n, nr_bits); 3479 3480 if (src1p && src2p) 3481 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3482 else if (src1p) 3483 return find_next_bit(src1p, nr_bits, n + 1); 3484 else if (src2p) 3485 return find_next_bit(src2p, nr_bits, n + 1); 3486 3487 return n + 1; 3488 } 3489 #else 3490 static inline int netif_set_xps_queue(struct net_device *dev, 3491 const struct cpumask *mask, 3492 u16 index) 3493 { 3494 return 0; 3495 } 3496 3497 static inline int __netif_set_xps_queue(struct net_device *dev, 3498 const unsigned long *mask, 3499 u16 index, bool is_rxqs_map) 3500 { 3501 return 0; 3502 } 3503 #endif 3504 3505 /** 3506 * netif_is_multiqueue - test if device has multiple transmit queues 3507 * @dev: network device 3508 * 3509 * Check if device has multiple transmit queues 3510 */ 3511 static inline bool netif_is_multiqueue(const struct net_device *dev) 3512 { 3513 return dev->num_tx_queues > 1; 3514 } 3515 3516 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3517 3518 #ifdef CONFIG_SYSFS 3519 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3520 #else 3521 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3522 unsigned int rxqs) 3523 { 3524 dev->real_num_rx_queues = rxqs; 3525 return 0; 3526 } 3527 #endif 3528 3529 static inline struct netdev_rx_queue * 3530 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3531 { 3532 return dev->_rx + rxq; 3533 } 3534 3535 #ifdef CONFIG_SYSFS 3536 static inline unsigned int get_netdev_rx_queue_index( 3537 struct netdev_rx_queue *queue) 3538 { 3539 struct net_device *dev = queue->dev; 3540 int index = queue - dev->_rx; 3541 3542 BUG_ON(index >= dev->num_rx_queues); 3543 return index; 3544 } 3545 #endif 3546 3547 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3548 int netif_get_num_default_rss_queues(void); 3549 3550 enum skb_free_reason { 3551 SKB_REASON_CONSUMED, 3552 SKB_REASON_DROPPED, 3553 }; 3554 3555 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3556 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3557 3558 /* 3559 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3560 * interrupt context or with hardware interrupts being disabled. 3561 * (in_irq() || irqs_disabled()) 3562 * 3563 * We provide four helpers that can be used in following contexts : 3564 * 3565 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3566 * replacing kfree_skb(skb) 3567 * 3568 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3569 * Typically used in place of consume_skb(skb) in TX completion path 3570 * 3571 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3572 * replacing kfree_skb(skb) 3573 * 3574 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3575 * and consumed a packet. Used in place of consume_skb(skb) 3576 */ 3577 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3578 { 3579 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3580 } 3581 3582 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3583 { 3584 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3585 } 3586 3587 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3588 { 3589 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3590 } 3591 3592 static inline void dev_consume_skb_any(struct sk_buff *skb) 3593 { 3594 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3595 } 3596 3597 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3598 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3599 int netif_rx(struct sk_buff *skb); 3600 int netif_rx_ni(struct sk_buff *skb); 3601 int netif_receive_skb(struct sk_buff *skb); 3602 int netif_receive_skb_core(struct sk_buff *skb); 3603 void netif_receive_skb_list(struct list_head *head); 3604 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3605 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3606 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3607 gro_result_t napi_gro_frags(struct napi_struct *napi); 3608 struct packet_offload *gro_find_receive_by_type(__be16 type); 3609 struct packet_offload *gro_find_complete_by_type(__be16 type); 3610 3611 static inline void napi_free_frags(struct napi_struct *napi) 3612 { 3613 kfree_skb(napi->skb); 3614 napi->skb = NULL; 3615 } 3616 3617 bool netdev_is_rx_handler_busy(struct net_device *dev); 3618 int netdev_rx_handler_register(struct net_device *dev, 3619 rx_handler_func_t *rx_handler, 3620 void *rx_handler_data); 3621 void netdev_rx_handler_unregister(struct net_device *dev); 3622 3623 bool dev_valid_name(const char *name); 3624 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3625 bool *need_copyout); 3626 int dev_ifconf(struct net *net, struct ifconf *, int); 3627 int dev_ethtool(struct net *net, struct ifreq *); 3628 unsigned int dev_get_flags(const struct net_device *); 3629 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3630 struct netlink_ext_ack *extack); 3631 int dev_change_flags(struct net_device *dev, unsigned int flags, 3632 struct netlink_ext_ack *extack); 3633 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3634 unsigned int gchanges); 3635 int dev_change_name(struct net_device *, const char *); 3636 int dev_set_alias(struct net_device *, const char *, size_t); 3637 int dev_get_alias(const struct net_device *, char *, size_t); 3638 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3639 int __dev_set_mtu(struct net_device *, int); 3640 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3641 struct netlink_ext_ack *extack); 3642 int dev_set_mtu(struct net_device *, int); 3643 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3644 void dev_set_group(struct net_device *, int); 3645 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3646 struct netlink_ext_ack *extack); 3647 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3648 struct netlink_ext_ack *extack); 3649 int dev_change_carrier(struct net_device *, bool new_carrier); 3650 int dev_get_phys_port_id(struct net_device *dev, 3651 struct netdev_phys_item_id *ppid); 3652 int dev_get_phys_port_name(struct net_device *dev, 3653 char *name, size_t len); 3654 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3655 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3656 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3657 struct netdev_queue *txq, int *ret); 3658 3659 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3660 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3661 int fd, u32 flags); 3662 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op, 3663 enum bpf_netdev_command cmd); 3664 int xdp_umem_query(struct net_device *dev, u16 queue_id); 3665 3666 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3667 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3668 bool is_skb_forwardable(const struct net_device *dev, 3669 const struct sk_buff *skb); 3670 3671 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3672 struct sk_buff *skb) 3673 { 3674 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3675 unlikely(!is_skb_forwardable(dev, skb))) { 3676 atomic_long_inc(&dev->rx_dropped); 3677 kfree_skb(skb); 3678 return NET_RX_DROP; 3679 } 3680 3681 skb_scrub_packet(skb, true); 3682 skb->priority = 0; 3683 return 0; 3684 } 3685 3686 bool dev_nit_active(struct net_device *dev); 3687 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3688 3689 extern int netdev_budget; 3690 extern unsigned int netdev_budget_usecs; 3691 3692 /* Called by rtnetlink.c:rtnl_unlock() */ 3693 void netdev_run_todo(void); 3694 3695 /** 3696 * dev_put - release reference to device 3697 * @dev: network device 3698 * 3699 * Release reference to device to allow it to be freed. 3700 */ 3701 static inline void dev_put(struct net_device *dev) 3702 { 3703 this_cpu_dec(*dev->pcpu_refcnt); 3704 } 3705 3706 /** 3707 * dev_hold - get reference to device 3708 * @dev: network device 3709 * 3710 * Hold reference to device to keep it from being freed. 3711 */ 3712 static inline void dev_hold(struct net_device *dev) 3713 { 3714 this_cpu_inc(*dev->pcpu_refcnt); 3715 } 3716 3717 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3718 * and _off may be called from IRQ context, but it is caller 3719 * who is responsible for serialization of these calls. 3720 * 3721 * The name carrier is inappropriate, these functions should really be 3722 * called netif_lowerlayer_*() because they represent the state of any 3723 * kind of lower layer not just hardware media. 3724 */ 3725 3726 void linkwatch_init_dev(struct net_device *dev); 3727 void linkwatch_fire_event(struct net_device *dev); 3728 void linkwatch_forget_dev(struct net_device *dev); 3729 3730 /** 3731 * netif_carrier_ok - test if carrier present 3732 * @dev: network device 3733 * 3734 * Check if carrier is present on device 3735 */ 3736 static inline bool netif_carrier_ok(const struct net_device *dev) 3737 { 3738 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3739 } 3740 3741 unsigned long dev_trans_start(struct net_device *dev); 3742 3743 void __netdev_watchdog_up(struct net_device *dev); 3744 3745 void netif_carrier_on(struct net_device *dev); 3746 3747 void netif_carrier_off(struct net_device *dev); 3748 3749 /** 3750 * netif_dormant_on - mark device as dormant. 3751 * @dev: network device 3752 * 3753 * Mark device as dormant (as per RFC2863). 3754 * 3755 * The dormant state indicates that the relevant interface is not 3756 * actually in a condition to pass packets (i.e., it is not 'up') but is 3757 * in a "pending" state, waiting for some external event. For "on- 3758 * demand" interfaces, this new state identifies the situation where the 3759 * interface is waiting for events to place it in the up state. 3760 */ 3761 static inline void netif_dormant_on(struct net_device *dev) 3762 { 3763 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3764 linkwatch_fire_event(dev); 3765 } 3766 3767 /** 3768 * netif_dormant_off - set device as not dormant. 3769 * @dev: network device 3770 * 3771 * Device is not in dormant state. 3772 */ 3773 static inline void netif_dormant_off(struct net_device *dev) 3774 { 3775 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3776 linkwatch_fire_event(dev); 3777 } 3778 3779 /** 3780 * netif_dormant - test if device is dormant 3781 * @dev: network device 3782 * 3783 * Check if device is dormant. 3784 */ 3785 static inline bool netif_dormant(const struct net_device *dev) 3786 { 3787 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3788 } 3789 3790 3791 /** 3792 * netif_oper_up - test if device is operational 3793 * @dev: network device 3794 * 3795 * Check if carrier is operational 3796 */ 3797 static inline bool netif_oper_up(const struct net_device *dev) 3798 { 3799 return (dev->operstate == IF_OPER_UP || 3800 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3801 } 3802 3803 /** 3804 * netif_device_present - is device available or removed 3805 * @dev: network device 3806 * 3807 * Check if device has not been removed from system. 3808 */ 3809 static inline bool netif_device_present(struct net_device *dev) 3810 { 3811 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3812 } 3813 3814 void netif_device_detach(struct net_device *dev); 3815 3816 void netif_device_attach(struct net_device *dev); 3817 3818 /* 3819 * Network interface message level settings 3820 */ 3821 3822 enum { 3823 NETIF_MSG_DRV = 0x0001, 3824 NETIF_MSG_PROBE = 0x0002, 3825 NETIF_MSG_LINK = 0x0004, 3826 NETIF_MSG_TIMER = 0x0008, 3827 NETIF_MSG_IFDOWN = 0x0010, 3828 NETIF_MSG_IFUP = 0x0020, 3829 NETIF_MSG_RX_ERR = 0x0040, 3830 NETIF_MSG_TX_ERR = 0x0080, 3831 NETIF_MSG_TX_QUEUED = 0x0100, 3832 NETIF_MSG_INTR = 0x0200, 3833 NETIF_MSG_TX_DONE = 0x0400, 3834 NETIF_MSG_RX_STATUS = 0x0800, 3835 NETIF_MSG_PKTDATA = 0x1000, 3836 NETIF_MSG_HW = 0x2000, 3837 NETIF_MSG_WOL = 0x4000, 3838 }; 3839 3840 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3841 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3842 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3843 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3844 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3845 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3846 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3847 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3848 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3849 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3850 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3851 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3852 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3853 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3854 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3855 3856 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3857 { 3858 /* use default */ 3859 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3860 return default_msg_enable_bits; 3861 if (debug_value == 0) /* no output */ 3862 return 0; 3863 /* set low N bits */ 3864 return (1 << debug_value) - 1; 3865 } 3866 3867 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 3868 { 3869 spin_lock(&txq->_xmit_lock); 3870 txq->xmit_lock_owner = cpu; 3871 } 3872 3873 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 3874 { 3875 __acquire(&txq->_xmit_lock); 3876 return true; 3877 } 3878 3879 static inline void __netif_tx_release(struct netdev_queue *txq) 3880 { 3881 __release(&txq->_xmit_lock); 3882 } 3883 3884 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 3885 { 3886 spin_lock_bh(&txq->_xmit_lock); 3887 txq->xmit_lock_owner = smp_processor_id(); 3888 } 3889 3890 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 3891 { 3892 bool ok = spin_trylock(&txq->_xmit_lock); 3893 if (likely(ok)) 3894 txq->xmit_lock_owner = smp_processor_id(); 3895 return ok; 3896 } 3897 3898 static inline void __netif_tx_unlock(struct netdev_queue *txq) 3899 { 3900 txq->xmit_lock_owner = -1; 3901 spin_unlock(&txq->_xmit_lock); 3902 } 3903 3904 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 3905 { 3906 txq->xmit_lock_owner = -1; 3907 spin_unlock_bh(&txq->_xmit_lock); 3908 } 3909 3910 static inline void txq_trans_update(struct netdev_queue *txq) 3911 { 3912 if (txq->xmit_lock_owner != -1) 3913 txq->trans_start = jiffies; 3914 } 3915 3916 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 3917 static inline void netif_trans_update(struct net_device *dev) 3918 { 3919 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 3920 3921 if (txq->trans_start != jiffies) 3922 txq->trans_start = jiffies; 3923 } 3924 3925 /** 3926 * netif_tx_lock - grab network device transmit lock 3927 * @dev: network device 3928 * 3929 * Get network device transmit lock 3930 */ 3931 static inline void netif_tx_lock(struct net_device *dev) 3932 { 3933 unsigned int i; 3934 int cpu; 3935 3936 spin_lock(&dev->tx_global_lock); 3937 cpu = smp_processor_id(); 3938 for (i = 0; i < dev->num_tx_queues; i++) { 3939 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3940 3941 /* We are the only thread of execution doing a 3942 * freeze, but we have to grab the _xmit_lock in 3943 * order to synchronize with threads which are in 3944 * the ->hard_start_xmit() handler and already 3945 * checked the frozen bit. 3946 */ 3947 __netif_tx_lock(txq, cpu); 3948 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 3949 __netif_tx_unlock(txq); 3950 } 3951 } 3952 3953 static inline void netif_tx_lock_bh(struct net_device *dev) 3954 { 3955 local_bh_disable(); 3956 netif_tx_lock(dev); 3957 } 3958 3959 static inline void netif_tx_unlock(struct net_device *dev) 3960 { 3961 unsigned int i; 3962 3963 for (i = 0; i < dev->num_tx_queues; i++) { 3964 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3965 3966 /* No need to grab the _xmit_lock here. If the 3967 * queue is not stopped for another reason, we 3968 * force a schedule. 3969 */ 3970 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 3971 netif_schedule_queue(txq); 3972 } 3973 spin_unlock(&dev->tx_global_lock); 3974 } 3975 3976 static inline void netif_tx_unlock_bh(struct net_device *dev) 3977 { 3978 netif_tx_unlock(dev); 3979 local_bh_enable(); 3980 } 3981 3982 #define HARD_TX_LOCK(dev, txq, cpu) { \ 3983 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3984 __netif_tx_lock(txq, cpu); \ 3985 } else { \ 3986 __netif_tx_acquire(txq); \ 3987 } \ 3988 } 3989 3990 #define HARD_TX_TRYLOCK(dev, txq) \ 3991 (((dev->features & NETIF_F_LLTX) == 0) ? \ 3992 __netif_tx_trylock(txq) : \ 3993 __netif_tx_acquire(txq)) 3994 3995 #define HARD_TX_UNLOCK(dev, txq) { \ 3996 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3997 __netif_tx_unlock(txq); \ 3998 } else { \ 3999 __netif_tx_release(txq); \ 4000 } \ 4001 } 4002 4003 static inline void netif_tx_disable(struct net_device *dev) 4004 { 4005 unsigned int i; 4006 int cpu; 4007 4008 local_bh_disable(); 4009 cpu = smp_processor_id(); 4010 for (i = 0; i < dev->num_tx_queues; i++) { 4011 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4012 4013 __netif_tx_lock(txq, cpu); 4014 netif_tx_stop_queue(txq); 4015 __netif_tx_unlock(txq); 4016 } 4017 local_bh_enable(); 4018 } 4019 4020 static inline void netif_addr_lock(struct net_device *dev) 4021 { 4022 spin_lock(&dev->addr_list_lock); 4023 } 4024 4025 static inline void netif_addr_lock_nested(struct net_device *dev) 4026 { 4027 int subclass = SINGLE_DEPTH_NESTING; 4028 4029 if (dev->netdev_ops->ndo_get_lock_subclass) 4030 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev); 4031 4032 spin_lock_nested(&dev->addr_list_lock, subclass); 4033 } 4034 4035 static inline void netif_addr_lock_bh(struct net_device *dev) 4036 { 4037 spin_lock_bh(&dev->addr_list_lock); 4038 } 4039 4040 static inline void netif_addr_unlock(struct net_device *dev) 4041 { 4042 spin_unlock(&dev->addr_list_lock); 4043 } 4044 4045 static inline void netif_addr_unlock_bh(struct net_device *dev) 4046 { 4047 spin_unlock_bh(&dev->addr_list_lock); 4048 } 4049 4050 /* 4051 * dev_addrs walker. Should be used only for read access. Call with 4052 * rcu_read_lock held. 4053 */ 4054 #define for_each_dev_addr(dev, ha) \ 4055 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4056 4057 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4058 4059 void ether_setup(struct net_device *dev); 4060 4061 /* Support for loadable net-drivers */ 4062 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4063 unsigned char name_assign_type, 4064 void (*setup)(struct net_device *), 4065 unsigned int txqs, unsigned int rxqs); 4066 int dev_get_valid_name(struct net *net, struct net_device *dev, 4067 const char *name); 4068 4069 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4070 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4071 4072 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4073 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4074 count) 4075 4076 int register_netdev(struct net_device *dev); 4077 void unregister_netdev(struct net_device *dev); 4078 4079 /* General hardware address lists handling functions */ 4080 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4081 struct netdev_hw_addr_list *from_list, int addr_len); 4082 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4083 struct netdev_hw_addr_list *from_list, int addr_len); 4084 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4085 struct net_device *dev, 4086 int (*sync)(struct net_device *, const unsigned char *), 4087 int (*unsync)(struct net_device *, 4088 const unsigned char *)); 4089 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4090 struct net_device *dev, 4091 int (*sync)(struct net_device *, 4092 const unsigned char *, int), 4093 int (*unsync)(struct net_device *, 4094 const unsigned char *, int)); 4095 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4096 struct net_device *dev, 4097 int (*unsync)(struct net_device *, 4098 const unsigned char *, int)); 4099 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4100 struct net_device *dev, 4101 int (*unsync)(struct net_device *, 4102 const unsigned char *)); 4103 void __hw_addr_init(struct netdev_hw_addr_list *list); 4104 4105 /* Functions used for device addresses handling */ 4106 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4107 unsigned char addr_type); 4108 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4109 unsigned char addr_type); 4110 void dev_addr_flush(struct net_device *dev); 4111 int dev_addr_init(struct net_device *dev); 4112 4113 /* Functions used for unicast addresses handling */ 4114 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4115 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4116 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4117 int dev_uc_sync(struct net_device *to, struct net_device *from); 4118 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4119 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4120 void dev_uc_flush(struct net_device *dev); 4121 void dev_uc_init(struct net_device *dev); 4122 4123 /** 4124 * __dev_uc_sync - Synchonize device's unicast list 4125 * @dev: device to sync 4126 * @sync: function to call if address should be added 4127 * @unsync: function to call if address should be removed 4128 * 4129 * Add newly added addresses to the interface, and release 4130 * addresses that have been deleted. 4131 */ 4132 static inline int __dev_uc_sync(struct net_device *dev, 4133 int (*sync)(struct net_device *, 4134 const unsigned char *), 4135 int (*unsync)(struct net_device *, 4136 const unsigned char *)) 4137 { 4138 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4139 } 4140 4141 /** 4142 * __dev_uc_unsync - Remove synchronized addresses from device 4143 * @dev: device to sync 4144 * @unsync: function to call if address should be removed 4145 * 4146 * Remove all addresses that were added to the device by dev_uc_sync(). 4147 */ 4148 static inline void __dev_uc_unsync(struct net_device *dev, 4149 int (*unsync)(struct net_device *, 4150 const unsigned char *)) 4151 { 4152 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4153 } 4154 4155 /* Functions used for multicast addresses handling */ 4156 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4157 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4158 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4159 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4160 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4161 int dev_mc_sync(struct net_device *to, struct net_device *from); 4162 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4163 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4164 void dev_mc_flush(struct net_device *dev); 4165 void dev_mc_init(struct net_device *dev); 4166 4167 /** 4168 * __dev_mc_sync - Synchonize device's multicast list 4169 * @dev: device to sync 4170 * @sync: function to call if address should be added 4171 * @unsync: function to call if address should be removed 4172 * 4173 * Add newly added addresses to the interface, and release 4174 * addresses that have been deleted. 4175 */ 4176 static inline int __dev_mc_sync(struct net_device *dev, 4177 int (*sync)(struct net_device *, 4178 const unsigned char *), 4179 int (*unsync)(struct net_device *, 4180 const unsigned char *)) 4181 { 4182 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4183 } 4184 4185 /** 4186 * __dev_mc_unsync - Remove synchronized addresses from device 4187 * @dev: device to sync 4188 * @unsync: function to call if address should be removed 4189 * 4190 * Remove all addresses that were added to the device by dev_mc_sync(). 4191 */ 4192 static inline void __dev_mc_unsync(struct net_device *dev, 4193 int (*unsync)(struct net_device *, 4194 const unsigned char *)) 4195 { 4196 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4197 } 4198 4199 /* Functions used for secondary unicast and multicast support */ 4200 void dev_set_rx_mode(struct net_device *dev); 4201 void __dev_set_rx_mode(struct net_device *dev); 4202 int dev_set_promiscuity(struct net_device *dev, int inc); 4203 int dev_set_allmulti(struct net_device *dev, int inc); 4204 void netdev_state_change(struct net_device *dev); 4205 void netdev_notify_peers(struct net_device *dev); 4206 void netdev_features_change(struct net_device *dev); 4207 /* Load a device via the kmod */ 4208 void dev_load(struct net *net, const char *name); 4209 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4210 struct rtnl_link_stats64 *storage); 4211 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4212 const struct net_device_stats *netdev_stats); 4213 4214 extern int netdev_max_backlog; 4215 extern int netdev_tstamp_prequeue; 4216 extern int weight_p; 4217 extern int dev_weight_rx_bias; 4218 extern int dev_weight_tx_bias; 4219 extern int dev_rx_weight; 4220 extern int dev_tx_weight; 4221 4222 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4223 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4224 struct list_head **iter); 4225 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 4226 struct list_head **iter); 4227 4228 /* iterate through upper list, must be called under RCU read lock */ 4229 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4230 for (iter = &(dev)->adj_list.upper, \ 4231 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4232 updev; \ 4233 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4234 4235 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4236 int (*fn)(struct net_device *upper_dev, 4237 void *data), 4238 void *data); 4239 4240 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4241 struct net_device *upper_dev); 4242 4243 bool netdev_has_any_upper_dev(struct net_device *dev); 4244 4245 void *netdev_lower_get_next_private(struct net_device *dev, 4246 struct list_head **iter); 4247 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4248 struct list_head **iter); 4249 4250 #define netdev_for_each_lower_private(dev, priv, iter) \ 4251 for (iter = (dev)->adj_list.lower.next, \ 4252 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4253 priv; \ 4254 priv = netdev_lower_get_next_private(dev, &(iter))) 4255 4256 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4257 for (iter = &(dev)->adj_list.lower, \ 4258 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4259 priv; \ 4260 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4261 4262 void *netdev_lower_get_next(struct net_device *dev, 4263 struct list_head **iter); 4264 4265 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4266 for (iter = (dev)->adj_list.lower.next, \ 4267 ldev = netdev_lower_get_next(dev, &(iter)); \ 4268 ldev; \ 4269 ldev = netdev_lower_get_next(dev, &(iter))) 4270 4271 struct net_device *netdev_all_lower_get_next(struct net_device *dev, 4272 struct list_head **iter); 4273 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev, 4274 struct list_head **iter); 4275 4276 int netdev_walk_all_lower_dev(struct net_device *dev, 4277 int (*fn)(struct net_device *lower_dev, 4278 void *data), 4279 void *data); 4280 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4281 int (*fn)(struct net_device *lower_dev, 4282 void *data), 4283 void *data); 4284 4285 void *netdev_adjacent_get_private(struct list_head *adj_list); 4286 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4287 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4288 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4289 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4290 struct netlink_ext_ack *extack); 4291 int netdev_master_upper_dev_link(struct net_device *dev, 4292 struct net_device *upper_dev, 4293 void *upper_priv, void *upper_info, 4294 struct netlink_ext_ack *extack); 4295 void netdev_upper_dev_unlink(struct net_device *dev, 4296 struct net_device *upper_dev); 4297 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4298 void *netdev_lower_dev_get_private(struct net_device *dev, 4299 struct net_device *lower_dev); 4300 void netdev_lower_state_changed(struct net_device *lower_dev, 4301 void *lower_state_info); 4302 4303 /* RSS keys are 40 or 52 bytes long */ 4304 #define NETDEV_RSS_KEY_LEN 52 4305 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4306 void netdev_rss_key_fill(void *buffer, size_t len); 4307 4308 int dev_get_nest_level(struct net_device *dev); 4309 int skb_checksum_help(struct sk_buff *skb); 4310 int skb_crc32c_csum_help(struct sk_buff *skb); 4311 int skb_csum_hwoffload_help(struct sk_buff *skb, 4312 const netdev_features_t features); 4313 4314 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4315 netdev_features_t features, bool tx_path); 4316 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4317 netdev_features_t features); 4318 4319 struct netdev_bonding_info { 4320 ifslave slave; 4321 ifbond master; 4322 }; 4323 4324 struct netdev_notifier_bonding_info { 4325 struct netdev_notifier_info info; /* must be first */ 4326 struct netdev_bonding_info bonding_info; 4327 }; 4328 4329 void netdev_bonding_info_change(struct net_device *dev, 4330 struct netdev_bonding_info *bonding_info); 4331 4332 static inline 4333 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4334 { 4335 return __skb_gso_segment(skb, features, true); 4336 } 4337 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4338 4339 static inline bool can_checksum_protocol(netdev_features_t features, 4340 __be16 protocol) 4341 { 4342 if (protocol == htons(ETH_P_FCOE)) 4343 return !!(features & NETIF_F_FCOE_CRC); 4344 4345 /* Assume this is an IP checksum (not SCTP CRC) */ 4346 4347 if (features & NETIF_F_HW_CSUM) { 4348 /* Can checksum everything */ 4349 return true; 4350 } 4351 4352 switch (protocol) { 4353 case htons(ETH_P_IP): 4354 return !!(features & NETIF_F_IP_CSUM); 4355 case htons(ETH_P_IPV6): 4356 return !!(features & NETIF_F_IPV6_CSUM); 4357 default: 4358 return false; 4359 } 4360 } 4361 4362 #ifdef CONFIG_BUG 4363 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4364 #else 4365 static inline void netdev_rx_csum_fault(struct net_device *dev, 4366 struct sk_buff *skb) 4367 { 4368 } 4369 #endif 4370 /* rx skb timestamps */ 4371 void net_enable_timestamp(void); 4372 void net_disable_timestamp(void); 4373 4374 #ifdef CONFIG_PROC_FS 4375 int __init dev_proc_init(void); 4376 #else 4377 #define dev_proc_init() 0 4378 #endif 4379 4380 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4381 struct sk_buff *skb, struct net_device *dev, 4382 bool more) 4383 { 4384 skb->xmit_more = more ? 1 : 0; 4385 return ops->ndo_start_xmit(skb, dev); 4386 } 4387 4388 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4389 struct netdev_queue *txq, bool more) 4390 { 4391 const struct net_device_ops *ops = dev->netdev_ops; 4392 netdev_tx_t rc; 4393 4394 rc = __netdev_start_xmit(ops, skb, dev, more); 4395 if (rc == NETDEV_TX_OK) 4396 txq_trans_update(txq); 4397 4398 return rc; 4399 } 4400 4401 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4402 const void *ns); 4403 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4404 const void *ns); 4405 4406 static inline int netdev_class_create_file(const struct class_attribute *class_attr) 4407 { 4408 return netdev_class_create_file_ns(class_attr, NULL); 4409 } 4410 4411 static inline void netdev_class_remove_file(const struct class_attribute *class_attr) 4412 { 4413 netdev_class_remove_file_ns(class_attr, NULL); 4414 } 4415 4416 extern const struct kobj_ns_type_operations net_ns_type_operations; 4417 4418 const char *netdev_drivername(const struct net_device *dev); 4419 4420 void linkwatch_run_queue(void); 4421 4422 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4423 netdev_features_t f2) 4424 { 4425 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4426 if (f1 & NETIF_F_HW_CSUM) 4427 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4428 else 4429 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4430 } 4431 4432 return f1 & f2; 4433 } 4434 4435 static inline netdev_features_t netdev_get_wanted_features( 4436 struct net_device *dev) 4437 { 4438 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4439 } 4440 netdev_features_t netdev_increment_features(netdev_features_t all, 4441 netdev_features_t one, netdev_features_t mask); 4442 4443 /* Allow TSO being used on stacked device : 4444 * Performing the GSO segmentation before last device 4445 * is a performance improvement. 4446 */ 4447 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4448 netdev_features_t mask) 4449 { 4450 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4451 } 4452 4453 int __netdev_update_features(struct net_device *dev); 4454 void netdev_update_features(struct net_device *dev); 4455 void netdev_change_features(struct net_device *dev); 4456 4457 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4458 struct net_device *dev); 4459 4460 netdev_features_t passthru_features_check(struct sk_buff *skb, 4461 struct net_device *dev, 4462 netdev_features_t features); 4463 netdev_features_t netif_skb_features(struct sk_buff *skb); 4464 4465 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4466 { 4467 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4468 4469 /* check flags correspondence */ 4470 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4471 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4472 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4473 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4474 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4475 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4476 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4477 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4478 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4479 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4480 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4481 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4482 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4483 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4484 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4485 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4486 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4487 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4488 4489 return (features & feature) == feature; 4490 } 4491 4492 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4493 { 4494 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4495 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4496 } 4497 4498 static inline bool netif_needs_gso(struct sk_buff *skb, 4499 netdev_features_t features) 4500 { 4501 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4502 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4503 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4504 } 4505 4506 static inline void netif_set_gso_max_size(struct net_device *dev, 4507 unsigned int size) 4508 { 4509 dev->gso_max_size = size; 4510 } 4511 4512 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4513 int pulled_hlen, u16 mac_offset, 4514 int mac_len) 4515 { 4516 skb->protocol = protocol; 4517 skb->encapsulation = 1; 4518 skb_push(skb, pulled_hlen); 4519 skb_reset_transport_header(skb); 4520 skb->mac_header = mac_offset; 4521 skb->network_header = skb->mac_header + mac_len; 4522 skb->mac_len = mac_len; 4523 } 4524 4525 static inline bool netif_is_macsec(const struct net_device *dev) 4526 { 4527 return dev->priv_flags & IFF_MACSEC; 4528 } 4529 4530 static inline bool netif_is_macvlan(const struct net_device *dev) 4531 { 4532 return dev->priv_flags & IFF_MACVLAN; 4533 } 4534 4535 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4536 { 4537 return dev->priv_flags & IFF_MACVLAN_PORT; 4538 } 4539 4540 static inline bool netif_is_bond_master(const struct net_device *dev) 4541 { 4542 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4543 } 4544 4545 static inline bool netif_is_bond_slave(const struct net_device *dev) 4546 { 4547 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4548 } 4549 4550 static inline bool netif_supports_nofcs(struct net_device *dev) 4551 { 4552 return dev->priv_flags & IFF_SUPP_NOFCS; 4553 } 4554 4555 static inline bool netif_is_l3_master(const struct net_device *dev) 4556 { 4557 return dev->priv_flags & IFF_L3MDEV_MASTER; 4558 } 4559 4560 static inline bool netif_is_l3_slave(const struct net_device *dev) 4561 { 4562 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4563 } 4564 4565 static inline bool netif_is_bridge_master(const struct net_device *dev) 4566 { 4567 return dev->priv_flags & IFF_EBRIDGE; 4568 } 4569 4570 static inline bool netif_is_bridge_port(const struct net_device *dev) 4571 { 4572 return dev->priv_flags & IFF_BRIDGE_PORT; 4573 } 4574 4575 static inline bool netif_is_ovs_master(const struct net_device *dev) 4576 { 4577 return dev->priv_flags & IFF_OPENVSWITCH; 4578 } 4579 4580 static inline bool netif_is_ovs_port(const struct net_device *dev) 4581 { 4582 return dev->priv_flags & IFF_OVS_DATAPATH; 4583 } 4584 4585 static inline bool netif_is_team_master(const struct net_device *dev) 4586 { 4587 return dev->priv_flags & IFF_TEAM; 4588 } 4589 4590 static inline bool netif_is_team_port(const struct net_device *dev) 4591 { 4592 return dev->priv_flags & IFF_TEAM_PORT; 4593 } 4594 4595 static inline bool netif_is_lag_master(const struct net_device *dev) 4596 { 4597 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4598 } 4599 4600 static inline bool netif_is_lag_port(const struct net_device *dev) 4601 { 4602 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4603 } 4604 4605 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4606 { 4607 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4608 } 4609 4610 static inline bool netif_is_failover(const struct net_device *dev) 4611 { 4612 return dev->priv_flags & IFF_FAILOVER; 4613 } 4614 4615 static inline bool netif_is_failover_slave(const struct net_device *dev) 4616 { 4617 return dev->priv_flags & IFF_FAILOVER_SLAVE; 4618 } 4619 4620 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4621 static inline void netif_keep_dst(struct net_device *dev) 4622 { 4623 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4624 } 4625 4626 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4627 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4628 { 4629 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4630 return dev->priv_flags & IFF_MACSEC; 4631 } 4632 4633 extern struct pernet_operations __net_initdata loopback_net_ops; 4634 4635 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4636 4637 /* netdev_printk helpers, similar to dev_printk */ 4638 4639 static inline const char *netdev_name(const struct net_device *dev) 4640 { 4641 if (!dev->name[0] || strchr(dev->name, '%')) 4642 return "(unnamed net_device)"; 4643 return dev->name; 4644 } 4645 4646 static inline bool netdev_unregistering(const struct net_device *dev) 4647 { 4648 return dev->reg_state == NETREG_UNREGISTERING; 4649 } 4650 4651 static inline const char *netdev_reg_state(const struct net_device *dev) 4652 { 4653 switch (dev->reg_state) { 4654 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4655 case NETREG_REGISTERED: return ""; 4656 case NETREG_UNREGISTERING: return " (unregistering)"; 4657 case NETREG_UNREGISTERED: return " (unregistered)"; 4658 case NETREG_RELEASED: return " (released)"; 4659 case NETREG_DUMMY: return " (dummy)"; 4660 } 4661 4662 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4663 return " (unknown)"; 4664 } 4665 4666 __printf(3, 4) 4667 void netdev_printk(const char *level, const struct net_device *dev, 4668 const char *format, ...); 4669 __printf(2, 3) 4670 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4671 __printf(2, 3) 4672 void netdev_alert(const struct net_device *dev, const char *format, ...); 4673 __printf(2, 3) 4674 void netdev_crit(const struct net_device *dev, const char *format, ...); 4675 __printf(2, 3) 4676 void netdev_err(const struct net_device *dev, const char *format, ...); 4677 __printf(2, 3) 4678 void netdev_warn(const struct net_device *dev, const char *format, ...); 4679 __printf(2, 3) 4680 void netdev_notice(const struct net_device *dev, const char *format, ...); 4681 __printf(2, 3) 4682 void netdev_info(const struct net_device *dev, const char *format, ...); 4683 4684 #define netdev_level_once(level, dev, fmt, ...) \ 4685 do { \ 4686 static bool __print_once __read_mostly; \ 4687 \ 4688 if (!__print_once) { \ 4689 __print_once = true; \ 4690 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 4691 } \ 4692 } while (0) 4693 4694 #define netdev_emerg_once(dev, fmt, ...) \ 4695 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 4696 #define netdev_alert_once(dev, fmt, ...) \ 4697 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 4698 #define netdev_crit_once(dev, fmt, ...) \ 4699 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 4700 #define netdev_err_once(dev, fmt, ...) \ 4701 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 4702 #define netdev_warn_once(dev, fmt, ...) \ 4703 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 4704 #define netdev_notice_once(dev, fmt, ...) \ 4705 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 4706 #define netdev_info_once(dev, fmt, ...) \ 4707 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 4708 4709 #define MODULE_ALIAS_NETDEV(device) \ 4710 MODULE_ALIAS("netdev-" device) 4711 4712 #if defined(CONFIG_DYNAMIC_DEBUG) 4713 #define netdev_dbg(__dev, format, args...) \ 4714 do { \ 4715 dynamic_netdev_dbg(__dev, format, ##args); \ 4716 } while (0) 4717 #elif defined(DEBUG) 4718 #define netdev_dbg(__dev, format, args...) \ 4719 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4720 #else 4721 #define netdev_dbg(__dev, format, args...) \ 4722 ({ \ 4723 if (0) \ 4724 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4725 }) 4726 #endif 4727 4728 #if defined(VERBOSE_DEBUG) 4729 #define netdev_vdbg netdev_dbg 4730 #else 4731 4732 #define netdev_vdbg(dev, format, args...) \ 4733 ({ \ 4734 if (0) \ 4735 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 4736 0; \ 4737 }) 4738 #endif 4739 4740 /* 4741 * netdev_WARN() acts like dev_printk(), but with the key difference 4742 * of using a WARN/WARN_ON to get the message out, including the 4743 * file/line information and a backtrace. 4744 */ 4745 #define netdev_WARN(dev, format, args...) \ 4746 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4747 netdev_reg_state(dev), ##args) 4748 4749 #define netdev_WARN_ONCE(dev, format, args...) \ 4750 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4751 netdev_reg_state(dev), ##args) 4752 4753 /* netif printk helpers, similar to netdev_printk */ 4754 4755 #define netif_printk(priv, type, level, dev, fmt, args...) \ 4756 do { \ 4757 if (netif_msg_##type(priv)) \ 4758 netdev_printk(level, (dev), fmt, ##args); \ 4759 } while (0) 4760 4761 #define netif_level(level, priv, type, dev, fmt, args...) \ 4762 do { \ 4763 if (netif_msg_##type(priv)) \ 4764 netdev_##level(dev, fmt, ##args); \ 4765 } while (0) 4766 4767 #define netif_emerg(priv, type, dev, fmt, args...) \ 4768 netif_level(emerg, priv, type, dev, fmt, ##args) 4769 #define netif_alert(priv, type, dev, fmt, args...) \ 4770 netif_level(alert, priv, type, dev, fmt, ##args) 4771 #define netif_crit(priv, type, dev, fmt, args...) \ 4772 netif_level(crit, priv, type, dev, fmt, ##args) 4773 #define netif_err(priv, type, dev, fmt, args...) \ 4774 netif_level(err, priv, type, dev, fmt, ##args) 4775 #define netif_warn(priv, type, dev, fmt, args...) \ 4776 netif_level(warn, priv, type, dev, fmt, ##args) 4777 #define netif_notice(priv, type, dev, fmt, args...) \ 4778 netif_level(notice, priv, type, dev, fmt, ##args) 4779 #define netif_info(priv, type, dev, fmt, args...) \ 4780 netif_level(info, priv, type, dev, fmt, ##args) 4781 4782 #if defined(CONFIG_DYNAMIC_DEBUG) 4783 #define netif_dbg(priv, type, netdev, format, args...) \ 4784 do { \ 4785 if (netif_msg_##type(priv)) \ 4786 dynamic_netdev_dbg(netdev, format, ##args); \ 4787 } while (0) 4788 #elif defined(DEBUG) 4789 #define netif_dbg(priv, type, dev, format, args...) \ 4790 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 4791 #else 4792 #define netif_dbg(priv, type, dev, format, args...) \ 4793 ({ \ 4794 if (0) \ 4795 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4796 0; \ 4797 }) 4798 #endif 4799 4800 /* if @cond then downgrade to debug, else print at @level */ 4801 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 4802 do { \ 4803 if (cond) \ 4804 netif_dbg(priv, type, netdev, fmt, ##args); \ 4805 else \ 4806 netif_ ## level(priv, type, netdev, fmt, ##args); \ 4807 } while (0) 4808 4809 #if defined(VERBOSE_DEBUG) 4810 #define netif_vdbg netif_dbg 4811 #else 4812 #define netif_vdbg(priv, type, dev, format, args...) \ 4813 ({ \ 4814 if (0) \ 4815 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4816 0; \ 4817 }) 4818 #endif 4819 4820 /* 4821 * The list of packet types we will receive (as opposed to discard) 4822 * and the routines to invoke. 4823 * 4824 * Why 16. Because with 16 the only overlap we get on a hash of the 4825 * low nibble of the protocol value is RARP/SNAP/X.25. 4826 * 4827 * 0800 IP 4828 * 0001 802.3 4829 * 0002 AX.25 4830 * 0004 802.2 4831 * 8035 RARP 4832 * 0005 SNAP 4833 * 0805 X.25 4834 * 0806 ARP 4835 * 8137 IPX 4836 * 0009 Localtalk 4837 * 86DD IPv6 4838 */ 4839 #define PTYPE_HASH_SIZE (16) 4840 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 4841 4842 #endif /* _LINUX_NETDEVICE_H */ 4843