1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <[email protected]> 13 * Corey Minyard <[email protected]> 14 * Donald J. Becker, <[email protected]> 15 * Alan Cox, <[email protected]> 16 * Bjorn Ekwall. <[email protected]> 17 * Pekka Riikonen <[email protected]> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 32 #include <linux/percpu.h> 33 #include <linux/rculist.h> 34 #include <linux/workqueue.h> 35 #include <linux/dynamic_queue_limits.h> 36 37 #include <linux/ethtool.h> 38 #include <net/net_namespace.h> 39 #ifdef CONFIG_DCB 40 #include <net/dcbnl.h> 41 #endif 42 #include <net/netprio_cgroup.h> 43 #include <net/xdp.h> 44 45 #include <linux/netdev_features.h> 46 #include <linux/neighbour.h> 47 #include <uapi/linux/netdevice.h> 48 #include <uapi/linux/if_bonding.h> 49 #include <uapi/linux/pkt_cls.h> 50 #include <linux/hashtable.h> 51 52 struct netpoll_info; 53 struct device; 54 struct phy_device; 55 struct dsa_port; 56 57 struct sfp_bus; 58 /* 802.11 specific */ 59 struct wireless_dev; 60 /* 802.15.4 specific */ 61 struct wpan_dev; 62 struct mpls_dev; 63 /* UDP Tunnel offloads */ 64 struct udp_tunnel_info; 65 struct bpf_prog; 66 struct xdp_buff; 67 68 void netdev_set_default_ethtool_ops(struct net_device *dev, 69 const struct ethtool_ops *ops); 70 71 /* Backlog congestion levels */ 72 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 73 #define NET_RX_DROP 1 /* packet dropped */ 74 75 /* 76 * Transmit return codes: transmit return codes originate from three different 77 * namespaces: 78 * 79 * - qdisc return codes 80 * - driver transmit return codes 81 * - errno values 82 * 83 * Drivers are allowed to return any one of those in their hard_start_xmit() 84 * function. Real network devices commonly used with qdiscs should only return 85 * the driver transmit return codes though - when qdiscs are used, the actual 86 * transmission happens asynchronously, so the value is not propagated to 87 * higher layers. Virtual network devices transmit synchronously; in this case 88 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 89 * others are propagated to higher layers. 90 */ 91 92 /* qdisc ->enqueue() return codes. */ 93 #define NET_XMIT_SUCCESS 0x00 94 #define NET_XMIT_DROP 0x01 /* skb dropped */ 95 #define NET_XMIT_CN 0x02 /* congestion notification */ 96 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 97 98 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 99 * indicates that the device will soon be dropping packets, or already drops 100 * some packets of the same priority; prompting us to send less aggressively. */ 101 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 102 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 103 104 /* Driver transmit return codes */ 105 #define NETDEV_TX_MASK 0xf0 106 107 enum netdev_tx { 108 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 109 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 110 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 111 }; 112 typedef enum netdev_tx netdev_tx_t; 113 114 /* 115 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 116 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 117 */ 118 static inline bool dev_xmit_complete(int rc) 119 { 120 /* 121 * Positive cases with an skb consumed by a driver: 122 * - successful transmission (rc == NETDEV_TX_OK) 123 * - error while transmitting (rc < 0) 124 * - error while queueing to a different device (rc & NET_XMIT_MASK) 125 */ 126 if (likely(rc < NET_XMIT_MASK)) 127 return true; 128 129 return false; 130 } 131 132 /* 133 * Compute the worst-case header length according to the protocols 134 * used. 135 */ 136 137 #if defined(CONFIG_HYPERV_NET) 138 # define LL_MAX_HEADER 128 139 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 140 # if defined(CONFIG_MAC80211_MESH) 141 # define LL_MAX_HEADER 128 142 # else 143 # define LL_MAX_HEADER 96 144 # endif 145 #else 146 # define LL_MAX_HEADER 32 147 #endif 148 149 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 150 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 151 #define MAX_HEADER LL_MAX_HEADER 152 #else 153 #define MAX_HEADER (LL_MAX_HEADER + 48) 154 #endif 155 156 /* 157 * Old network device statistics. Fields are native words 158 * (unsigned long) so they can be read and written atomically. 159 */ 160 161 struct net_device_stats { 162 unsigned long rx_packets; 163 unsigned long tx_packets; 164 unsigned long rx_bytes; 165 unsigned long tx_bytes; 166 unsigned long rx_errors; 167 unsigned long tx_errors; 168 unsigned long rx_dropped; 169 unsigned long tx_dropped; 170 unsigned long multicast; 171 unsigned long collisions; 172 unsigned long rx_length_errors; 173 unsigned long rx_over_errors; 174 unsigned long rx_crc_errors; 175 unsigned long rx_frame_errors; 176 unsigned long rx_fifo_errors; 177 unsigned long rx_missed_errors; 178 unsigned long tx_aborted_errors; 179 unsigned long tx_carrier_errors; 180 unsigned long tx_fifo_errors; 181 unsigned long tx_heartbeat_errors; 182 unsigned long tx_window_errors; 183 unsigned long rx_compressed; 184 unsigned long tx_compressed; 185 }; 186 187 188 #include <linux/cache.h> 189 #include <linux/skbuff.h> 190 191 #ifdef CONFIG_RPS 192 #include <linux/static_key.h> 193 extern struct static_key_false rps_needed; 194 extern struct static_key_false rfs_needed; 195 #endif 196 197 struct neighbour; 198 struct neigh_parms; 199 struct sk_buff; 200 201 struct netdev_hw_addr { 202 struct list_head list; 203 unsigned char addr[MAX_ADDR_LEN]; 204 unsigned char type; 205 #define NETDEV_HW_ADDR_T_LAN 1 206 #define NETDEV_HW_ADDR_T_SAN 2 207 #define NETDEV_HW_ADDR_T_SLAVE 3 208 #define NETDEV_HW_ADDR_T_UNICAST 4 209 #define NETDEV_HW_ADDR_T_MULTICAST 5 210 bool global_use; 211 int sync_cnt; 212 int refcount; 213 int synced; 214 struct rcu_head rcu_head; 215 }; 216 217 struct netdev_hw_addr_list { 218 struct list_head list; 219 int count; 220 }; 221 222 #define netdev_hw_addr_list_count(l) ((l)->count) 223 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 224 #define netdev_hw_addr_list_for_each(ha, l) \ 225 list_for_each_entry(ha, &(l)->list, list) 226 227 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 228 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 229 #define netdev_for_each_uc_addr(ha, dev) \ 230 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 231 232 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 233 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 234 #define netdev_for_each_mc_addr(ha, dev) \ 235 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 236 237 struct hh_cache { 238 unsigned int hh_len; 239 seqlock_t hh_lock; 240 241 /* cached hardware header; allow for machine alignment needs. */ 242 #define HH_DATA_MOD 16 243 #define HH_DATA_OFF(__len) \ 244 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 245 #define HH_DATA_ALIGN(__len) \ 246 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 247 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 248 }; 249 250 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 251 * Alternative is: 252 * dev->hard_header_len ? (dev->hard_header_len + 253 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 254 * 255 * We could use other alignment values, but we must maintain the 256 * relationship HH alignment <= LL alignment. 257 */ 258 #define LL_RESERVED_SPACE(dev) \ 259 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 260 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 261 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 262 263 struct header_ops { 264 int (*create) (struct sk_buff *skb, struct net_device *dev, 265 unsigned short type, const void *daddr, 266 const void *saddr, unsigned int len); 267 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 268 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 269 void (*cache_update)(struct hh_cache *hh, 270 const struct net_device *dev, 271 const unsigned char *haddr); 272 bool (*validate)(const char *ll_header, unsigned int len); 273 __be16 (*parse_protocol)(const struct sk_buff *skb); 274 }; 275 276 /* These flag bits are private to the generic network queueing 277 * layer; they may not be explicitly referenced by any other 278 * code. 279 */ 280 281 enum netdev_state_t { 282 __LINK_STATE_START, 283 __LINK_STATE_PRESENT, 284 __LINK_STATE_NOCARRIER, 285 __LINK_STATE_LINKWATCH_PENDING, 286 __LINK_STATE_DORMANT, 287 }; 288 289 290 /* 291 * This structure holds boot-time configured netdevice settings. They 292 * are then used in the device probing. 293 */ 294 struct netdev_boot_setup { 295 char name[IFNAMSIZ]; 296 struct ifmap map; 297 }; 298 #define NETDEV_BOOT_SETUP_MAX 8 299 300 int __init netdev_boot_setup(char *str); 301 302 struct gro_list { 303 struct list_head list; 304 int count; 305 }; 306 307 /* 308 * size of gro hash buckets, must less than bit number of 309 * napi_struct::gro_bitmask 310 */ 311 #define GRO_HASH_BUCKETS 8 312 313 /* 314 * Structure for NAPI scheduling similar to tasklet but with weighting 315 */ 316 struct napi_struct { 317 /* The poll_list must only be managed by the entity which 318 * changes the state of the NAPI_STATE_SCHED bit. This means 319 * whoever atomically sets that bit can add this napi_struct 320 * to the per-CPU poll_list, and whoever clears that bit 321 * can remove from the list right before clearing the bit. 322 */ 323 struct list_head poll_list; 324 325 unsigned long state; 326 int weight; 327 unsigned long gro_bitmask; 328 int (*poll)(struct napi_struct *, int); 329 #ifdef CONFIG_NETPOLL 330 int poll_owner; 331 #endif 332 struct net_device *dev; 333 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 334 struct sk_buff *skb; 335 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 336 int rx_count; /* length of rx_list */ 337 struct hrtimer timer; 338 struct list_head dev_list; 339 struct hlist_node napi_hash_node; 340 unsigned int napi_id; 341 }; 342 343 enum { 344 NAPI_STATE_SCHED, /* Poll is scheduled */ 345 NAPI_STATE_MISSED, /* reschedule a napi */ 346 NAPI_STATE_DISABLE, /* Disable pending */ 347 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 348 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */ 349 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */ 350 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */ 351 }; 352 353 enum { 354 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 355 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 356 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 357 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 358 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED), 359 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 360 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 361 }; 362 363 enum gro_result { 364 GRO_MERGED, 365 GRO_MERGED_FREE, 366 GRO_HELD, 367 GRO_NORMAL, 368 GRO_DROP, 369 GRO_CONSUMED, 370 }; 371 typedef enum gro_result gro_result_t; 372 373 /* 374 * enum rx_handler_result - Possible return values for rx_handlers. 375 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 376 * further. 377 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 378 * case skb->dev was changed by rx_handler. 379 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 380 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 381 * 382 * rx_handlers are functions called from inside __netif_receive_skb(), to do 383 * special processing of the skb, prior to delivery to protocol handlers. 384 * 385 * Currently, a net_device can only have a single rx_handler registered. Trying 386 * to register a second rx_handler will return -EBUSY. 387 * 388 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 389 * To unregister a rx_handler on a net_device, use 390 * netdev_rx_handler_unregister(). 391 * 392 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 393 * do with the skb. 394 * 395 * If the rx_handler consumed the skb in some way, it should return 396 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 397 * the skb to be delivered in some other way. 398 * 399 * If the rx_handler changed skb->dev, to divert the skb to another 400 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 401 * new device will be called if it exists. 402 * 403 * If the rx_handler decides the skb should be ignored, it should return 404 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 405 * are registered on exact device (ptype->dev == skb->dev). 406 * 407 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 408 * delivered, it should return RX_HANDLER_PASS. 409 * 410 * A device without a registered rx_handler will behave as if rx_handler 411 * returned RX_HANDLER_PASS. 412 */ 413 414 enum rx_handler_result { 415 RX_HANDLER_CONSUMED, 416 RX_HANDLER_ANOTHER, 417 RX_HANDLER_EXACT, 418 RX_HANDLER_PASS, 419 }; 420 typedef enum rx_handler_result rx_handler_result_t; 421 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 422 423 void __napi_schedule(struct napi_struct *n); 424 void __napi_schedule_irqoff(struct napi_struct *n); 425 426 static inline bool napi_disable_pending(struct napi_struct *n) 427 { 428 return test_bit(NAPI_STATE_DISABLE, &n->state); 429 } 430 431 bool napi_schedule_prep(struct napi_struct *n); 432 433 /** 434 * napi_schedule - schedule NAPI poll 435 * @n: NAPI context 436 * 437 * Schedule NAPI poll routine to be called if it is not already 438 * running. 439 */ 440 static inline void napi_schedule(struct napi_struct *n) 441 { 442 if (napi_schedule_prep(n)) 443 __napi_schedule(n); 444 } 445 446 /** 447 * napi_schedule_irqoff - schedule NAPI poll 448 * @n: NAPI context 449 * 450 * Variant of napi_schedule(), assuming hard irqs are masked. 451 */ 452 static inline void napi_schedule_irqoff(struct napi_struct *n) 453 { 454 if (napi_schedule_prep(n)) 455 __napi_schedule_irqoff(n); 456 } 457 458 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 459 static inline bool napi_reschedule(struct napi_struct *napi) 460 { 461 if (napi_schedule_prep(napi)) { 462 __napi_schedule(napi); 463 return true; 464 } 465 return false; 466 } 467 468 bool napi_complete_done(struct napi_struct *n, int work_done); 469 /** 470 * napi_complete - NAPI processing complete 471 * @n: NAPI context 472 * 473 * Mark NAPI processing as complete. 474 * Consider using napi_complete_done() instead. 475 * Return false if device should avoid rearming interrupts. 476 */ 477 static inline bool napi_complete(struct napi_struct *n) 478 { 479 return napi_complete_done(n, 0); 480 } 481 482 /** 483 * napi_hash_del - remove a NAPI from global table 484 * @napi: NAPI context 485 * 486 * Warning: caller must observe RCU grace period 487 * before freeing memory containing @napi, if 488 * this function returns true. 489 * Note: core networking stack automatically calls it 490 * from netif_napi_del(). 491 * Drivers might want to call this helper to combine all 492 * the needed RCU grace periods into a single one. 493 */ 494 bool napi_hash_del(struct napi_struct *napi); 495 496 /** 497 * napi_disable - prevent NAPI from scheduling 498 * @n: NAPI context 499 * 500 * Stop NAPI from being scheduled on this context. 501 * Waits till any outstanding processing completes. 502 */ 503 void napi_disable(struct napi_struct *n); 504 505 /** 506 * napi_enable - enable NAPI scheduling 507 * @n: NAPI context 508 * 509 * Resume NAPI from being scheduled on this context. 510 * Must be paired with napi_disable. 511 */ 512 static inline void napi_enable(struct napi_struct *n) 513 { 514 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 515 smp_mb__before_atomic(); 516 clear_bit(NAPI_STATE_SCHED, &n->state); 517 clear_bit(NAPI_STATE_NPSVC, &n->state); 518 } 519 520 /** 521 * napi_synchronize - wait until NAPI is not running 522 * @n: NAPI context 523 * 524 * Wait until NAPI is done being scheduled on this context. 525 * Waits till any outstanding processing completes but 526 * does not disable future activations. 527 */ 528 static inline void napi_synchronize(const struct napi_struct *n) 529 { 530 if (IS_ENABLED(CONFIG_SMP)) 531 while (test_bit(NAPI_STATE_SCHED, &n->state)) 532 msleep(1); 533 else 534 barrier(); 535 } 536 537 /** 538 * napi_if_scheduled_mark_missed - if napi is running, set the 539 * NAPIF_STATE_MISSED 540 * @n: NAPI context 541 * 542 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 543 * NAPI is scheduled. 544 **/ 545 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 546 { 547 unsigned long val, new; 548 549 do { 550 val = READ_ONCE(n->state); 551 if (val & NAPIF_STATE_DISABLE) 552 return true; 553 554 if (!(val & NAPIF_STATE_SCHED)) 555 return false; 556 557 new = val | NAPIF_STATE_MISSED; 558 } while (cmpxchg(&n->state, val, new) != val); 559 560 return true; 561 } 562 563 enum netdev_queue_state_t { 564 __QUEUE_STATE_DRV_XOFF, 565 __QUEUE_STATE_STACK_XOFF, 566 __QUEUE_STATE_FROZEN, 567 }; 568 569 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 570 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 571 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 572 573 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 574 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 575 QUEUE_STATE_FROZEN) 576 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 577 QUEUE_STATE_FROZEN) 578 579 /* 580 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 581 * netif_tx_* functions below are used to manipulate this flag. The 582 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 583 * queue independently. The netif_xmit_*stopped functions below are called 584 * to check if the queue has been stopped by the driver or stack (either 585 * of the XOFF bits are set in the state). Drivers should not need to call 586 * netif_xmit*stopped functions, they should only be using netif_tx_*. 587 */ 588 589 struct netdev_queue { 590 /* 591 * read-mostly part 592 */ 593 struct net_device *dev; 594 struct Qdisc __rcu *qdisc; 595 struct Qdisc *qdisc_sleeping; 596 #ifdef CONFIG_SYSFS 597 struct kobject kobj; 598 #endif 599 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 600 int numa_node; 601 #endif 602 unsigned long tx_maxrate; 603 /* 604 * Number of TX timeouts for this queue 605 * (/sys/class/net/DEV/Q/trans_timeout) 606 */ 607 unsigned long trans_timeout; 608 609 /* Subordinate device that the queue has been assigned to */ 610 struct net_device *sb_dev; 611 #ifdef CONFIG_XDP_SOCKETS 612 struct xdp_umem *umem; 613 #endif 614 /* 615 * write-mostly part 616 */ 617 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 618 int xmit_lock_owner; 619 /* 620 * Time (in jiffies) of last Tx 621 */ 622 unsigned long trans_start; 623 624 unsigned long state; 625 626 #ifdef CONFIG_BQL 627 struct dql dql; 628 #endif 629 } ____cacheline_aligned_in_smp; 630 631 extern int sysctl_fb_tunnels_only_for_init_net; 632 extern int sysctl_devconf_inherit_init_net; 633 634 static inline bool net_has_fallback_tunnels(const struct net *net) 635 { 636 return net == &init_net || 637 !IS_ENABLED(CONFIG_SYSCTL) || 638 !sysctl_fb_tunnels_only_for_init_net; 639 } 640 641 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 642 { 643 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 644 return q->numa_node; 645 #else 646 return NUMA_NO_NODE; 647 #endif 648 } 649 650 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 651 { 652 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 653 q->numa_node = node; 654 #endif 655 } 656 657 #ifdef CONFIG_RPS 658 /* 659 * This structure holds an RPS map which can be of variable length. The 660 * map is an array of CPUs. 661 */ 662 struct rps_map { 663 unsigned int len; 664 struct rcu_head rcu; 665 u16 cpus[0]; 666 }; 667 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 668 669 /* 670 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 671 * tail pointer for that CPU's input queue at the time of last enqueue, and 672 * a hardware filter index. 673 */ 674 struct rps_dev_flow { 675 u16 cpu; 676 u16 filter; 677 unsigned int last_qtail; 678 }; 679 #define RPS_NO_FILTER 0xffff 680 681 /* 682 * The rps_dev_flow_table structure contains a table of flow mappings. 683 */ 684 struct rps_dev_flow_table { 685 unsigned int mask; 686 struct rcu_head rcu; 687 struct rps_dev_flow flows[0]; 688 }; 689 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 690 ((_num) * sizeof(struct rps_dev_flow))) 691 692 /* 693 * The rps_sock_flow_table contains mappings of flows to the last CPU 694 * on which they were processed by the application (set in recvmsg). 695 * Each entry is a 32bit value. Upper part is the high-order bits 696 * of flow hash, lower part is CPU number. 697 * rps_cpu_mask is used to partition the space, depending on number of 698 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 699 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 700 * meaning we use 32-6=26 bits for the hash. 701 */ 702 struct rps_sock_flow_table { 703 u32 mask; 704 705 u32 ents[0] ____cacheline_aligned_in_smp; 706 }; 707 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 708 709 #define RPS_NO_CPU 0xffff 710 711 extern u32 rps_cpu_mask; 712 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 713 714 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 715 u32 hash) 716 { 717 if (table && hash) { 718 unsigned int index = hash & table->mask; 719 u32 val = hash & ~rps_cpu_mask; 720 721 /* We only give a hint, preemption can change CPU under us */ 722 val |= raw_smp_processor_id(); 723 724 if (table->ents[index] != val) 725 table->ents[index] = val; 726 } 727 } 728 729 #ifdef CONFIG_RFS_ACCEL 730 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 731 u16 filter_id); 732 #endif 733 #endif /* CONFIG_RPS */ 734 735 /* This structure contains an instance of an RX queue. */ 736 struct netdev_rx_queue { 737 #ifdef CONFIG_RPS 738 struct rps_map __rcu *rps_map; 739 struct rps_dev_flow_table __rcu *rps_flow_table; 740 #endif 741 struct kobject kobj; 742 struct net_device *dev; 743 struct xdp_rxq_info xdp_rxq; 744 #ifdef CONFIG_XDP_SOCKETS 745 struct xdp_umem *umem; 746 #endif 747 } ____cacheline_aligned_in_smp; 748 749 /* 750 * RX queue sysfs structures and functions. 751 */ 752 struct rx_queue_attribute { 753 struct attribute attr; 754 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 755 ssize_t (*store)(struct netdev_rx_queue *queue, 756 const char *buf, size_t len); 757 }; 758 759 #ifdef CONFIG_XPS 760 /* 761 * This structure holds an XPS map which can be of variable length. The 762 * map is an array of queues. 763 */ 764 struct xps_map { 765 unsigned int len; 766 unsigned int alloc_len; 767 struct rcu_head rcu; 768 u16 queues[0]; 769 }; 770 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 771 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 772 - sizeof(struct xps_map)) / sizeof(u16)) 773 774 /* 775 * This structure holds all XPS maps for device. Maps are indexed by CPU. 776 */ 777 struct xps_dev_maps { 778 struct rcu_head rcu; 779 struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */ 780 }; 781 782 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 783 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 784 785 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 786 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 787 788 #endif /* CONFIG_XPS */ 789 790 #define TC_MAX_QUEUE 16 791 #define TC_BITMASK 15 792 /* HW offloaded queuing disciplines txq count and offset maps */ 793 struct netdev_tc_txq { 794 u16 count; 795 u16 offset; 796 }; 797 798 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 799 /* 800 * This structure is to hold information about the device 801 * configured to run FCoE protocol stack. 802 */ 803 struct netdev_fcoe_hbainfo { 804 char manufacturer[64]; 805 char serial_number[64]; 806 char hardware_version[64]; 807 char driver_version[64]; 808 char optionrom_version[64]; 809 char firmware_version[64]; 810 char model[256]; 811 char model_description[256]; 812 }; 813 #endif 814 815 #define MAX_PHYS_ITEM_ID_LEN 32 816 817 /* This structure holds a unique identifier to identify some 818 * physical item (port for example) used by a netdevice. 819 */ 820 struct netdev_phys_item_id { 821 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 822 unsigned char id_len; 823 }; 824 825 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 826 struct netdev_phys_item_id *b) 827 { 828 return a->id_len == b->id_len && 829 memcmp(a->id, b->id, a->id_len) == 0; 830 } 831 832 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 833 struct sk_buff *skb, 834 struct net_device *sb_dev); 835 836 enum tc_setup_type { 837 TC_SETUP_QDISC_MQPRIO, 838 TC_SETUP_CLSU32, 839 TC_SETUP_CLSFLOWER, 840 TC_SETUP_CLSMATCHALL, 841 TC_SETUP_CLSBPF, 842 TC_SETUP_BLOCK, 843 TC_SETUP_QDISC_CBS, 844 TC_SETUP_QDISC_RED, 845 TC_SETUP_QDISC_PRIO, 846 TC_SETUP_QDISC_MQ, 847 TC_SETUP_QDISC_ETF, 848 TC_SETUP_ROOT_QDISC, 849 TC_SETUP_QDISC_GRED, 850 TC_SETUP_QDISC_TAPRIO, 851 TC_SETUP_FT, 852 TC_SETUP_QDISC_ETS, 853 }; 854 855 /* These structures hold the attributes of bpf state that are being passed 856 * to the netdevice through the bpf op. 857 */ 858 enum bpf_netdev_command { 859 /* Set or clear a bpf program used in the earliest stages of packet 860 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 861 * is responsible for calling bpf_prog_put on any old progs that are 862 * stored. In case of error, the callee need not release the new prog 863 * reference, but on success it takes ownership and must bpf_prog_put 864 * when it is no longer used. 865 */ 866 XDP_SETUP_PROG, 867 XDP_SETUP_PROG_HW, 868 XDP_QUERY_PROG, 869 XDP_QUERY_PROG_HW, 870 /* BPF program for offload callbacks, invoked at program load time. */ 871 BPF_OFFLOAD_MAP_ALLOC, 872 BPF_OFFLOAD_MAP_FREE, 873 XDP_SETUP_XSK_UMEM, 874 }; 875 876 struct bpf_prog_offload_ops; 877 struct netlink_ext_ack; 878 struct xdp_umem; 879 struct xdp_dev_bulk_queue; 880 881 struct netdev_bpf { 882 enum bpf_netdev_command command; 883 union { 884 /* XDP_SETUP_PROG */ 885 struct { 886 u32 flags; 887 struct bpf_prog *prog; 888 struct netlink_ext_ack *extack; 889 }; 890 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */ 891 struct { 892 u32 prog_id; 893 /* flags with which program was installed */ 894 u32 prog_flags; 895 }; 896 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 897 struct { 898 struct bpf_offloaded_map *offmap; 899 }; 900 /* XDP_SETUP_XSK_UMEM */ 901 struct { 902 struct xdp_umem *umem; 903 u16 queue_id; 904 } xsk; 905 }; 906 }; 907 908 /* Flags for ndo_xsk_wakeup. */ 909 #define XDP_WAKEUP_RX (1 << 0) 910 #define XDP_WAKEUP_TX (1 << 1) 911 912 #ifdef CONFIG_XFRM_OFFLOAD 913 struct xfrmdev_ops { 914 int (*xdo_dev_state_add) (struct xfrm_state *x); 915 void (*xdo_dev_state_delete) (struct xfrm_state *x); 916 void (*xdo_dev_state_free) (struct xfrm_state *x); 917 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 918 struct xfrm_state *x); 919 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 920 }; 921 #endif 922 923 struct dev_ifalias { 924 struct rcu_head rcuhead; 925 char ifalias[]; 926 }; 927 928 struct devlink; 929 struct tlsdev_ops; 930 931 struct netdev_name_node { 932 struct hlist_node hlist; 933 struct list_head list; 934 struct net_device *dev; 935 const char *name; 936 }; 937 938 int netdev_name_node_alt_create(struct net_device *dev, const char *name); 939 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name); 940 941 /* 942 * This structure defines the management hooks for network devices. 943 * The following hooks can be defined; unless noted otherwise, they are 944 * optional and can be filled with a null pointer. 945 * 946 * int (*ndo_init)(struct net_device *dev); 947 * This function is called once when a network device is registered. 948 * The network device can use this for any late stage initialization 949 * or semantic validation. It can fail with an error code which will 950 * be propagated back to register_netdev. 951 * 952 * void (*ndo_uninit)(struct net_device *dev); 953 * This function is called when device is unregistered or when registration 954 * fails. It is not called if init fails. 955 * 956 * int (*ndo_open)(struct net_device *dev); 957 * This function is called when a network device transitions to the up 958 * state. 959 * 960 * int (*ndo_stop)(struct net_device *dev); 961 * This function is called when a network device transitions to the down 962 * state. 963 * 964 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 965 * struct net_device *dev); 966 * Called when a packet needs to be transmitted. 967 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 968 * the queue before that can happen; it's for obsolete devices and weird 969 * corner cases, but the stack really does a non-trivial amount 970 * of useless work if you return NETDEV_TX_BUSY. 971 * Required; cannot be NULL. 972 * 973 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 974 * struct net_device *dev 975 * netdev_features_t features); 976 * Called by core transmit path to determine if device is capable of 977 * performing offload operations on a given packet. This is to give 978 * the device an opportunity to implement any restrictions that cannot 979 * be otherwise expressed by feature flags. The check is called with 980 * the set of features that the stack has calculated and it returns 981 * those the driver believes to be appropriate. 982 * 983 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 984 * struct net_device *sb_dev); 985 * Called to decide which queue to use when device supports multiple 986 * transmit queues. 987 * 988 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 989 * This function is called to allow device receiver to make 990 * changes to configuration when multicast or promiscuous is enabled. 991 * 992 * void (*ndo_set_rx_mode)(struct net_device *dev); 993 * This function is called device changes address list filtering. 994 * If driver handles unicast address filtering, it should set 995 * IFF_UNICAST_FLT in its priv_flags. 996 * 997 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 998 * This function is called when the Media Access Control address 999 * needs to be changed. If this interface is not defined, the 1000 * MAC address can not be changed. 1001 * 1002 * int (*ndo_validate_addr)(struct net_device *dev); 1003 * Test if Media Access Control address is valid for the device. 1004 * 1005 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1006 * Called when a user requests an ioctl which can't be handled by 1007 * the generic interface code. If not defined ioctls return 1008 * not supported error code. 1009 * 1010 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1011 * Used to set network devices bus interface parameters. This interface 1012 * is retained for legacy reasons; new devices should use the bus 1013 * interface (PCI) for low level management. 1014 * 1015 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1016 * Called when a user wants to change the Maximum Transfer Unit 1017 * of a device. 1018 * 1019 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1020 * Callback used when the transmitter has not made any progress 1021 * for dev->watchdog ticks. 1022 * 1023 * void (*ndo_get_stats64)(struct net_device *dev, 1024 * struct rtnl_link_stats64 *storage); 1025 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1026 * Called when a user wants to get the network device usage 1027 * statistics. Drivers must do one of the following: 1028 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1029 * rtnl_link_stats64 structure passed by the caller. 1030 * 2. Define @ndo_get_stats to update a net_device_stats structure 1031 * (which should normally be dev->stats) and return a pointer to 1032 * it. The structure may be changed asynchronously only if each 1033 * field is written atomically. 1034 * 3. Update dev->stats asynchronously and atomically, and define 1035 * neither operation. 1036 * 1037 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1038 * Return true if this device supports offload stats of this attr_id. 1039 * 1040 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1041 * void *attr_data) 1042 * Get statistics for offload operations by attr_id. Write it into the 1043 * attr_data pointer. 1044 * 1045 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1046 * If device supports VLAN filtering this function is called when a 1047 * VLAN id is registered. 1048 * 1049 * int (*ndo_vlan_rx_kill_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 unregistered. 1052 * 1053 * void (*ndo_poll_controller)(struct net_device *dev); 1054 * 1055 * SR-IOV management functions. 1056 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1057 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1058 * u8 qos, __be16 proto); 1059 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1060 * int max_tx_rate); 1061 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1062 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1063 * int (*ndo_get_vf_config)(struct net_device *dev, 1064 * int vf, struct ifla_vf_info *ivf); 1065 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1066 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1067 * struct nlattr *port[]); 1068 * 1069 * Enable or disable the VF ability to query its RSS Redirection Table and 1070 * Hash Key. This is needed since on some devices VF share this information 1071 * with PF and querying it may introduce a theoretical security risk. 1072 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1073 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1074 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1075 * void *type_data); 1076 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1077 * This is always called from the stack with the rtnl lock held and netif 1078 * tx queues stopped. This allows the netdevice to perform queue 1079 * management safely. 1080 * 1081 * Fiber Channel over Ethernet (FCoE) offload functions. 1082 * int (*ndo_fcoe_enable)(struct net_device *dev); 1083 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1084 * so the underlying device can perform whatever needed configuration or 1085 * initialization to support acceleration of FCoE traffic. 1086 * 1087 * int (*ndo_fcoe_disable)(struct net_device *dev); 1088 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1089 * so the underlying device can perform whatever needed clean-ups to 1090 * stop supporting acceleration of FCoE traffic. 1091 * 1092 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1093 * struct scatterlist *sgl, unsigned int sgc); 1094 * Called when the FCoE Initiator wants to initialize an I/O that 1095 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1096 * perform necessary setup and returns 1 to indicate the device is set up 1097 * successfully to perform DDP on this I/O, otherwise this returns 0. 1098 * 1099 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1100 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1101 * indicated by the FC exchange id 'xid', so the underlying device can 1102 * clean up and reuse resources for later DDP requests. 1103 * 1104 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1105 * struct scatterlist *sgl, unsigned int sgc); 1106 * Called when the FCoE Target wants to initialize an I/O that 1107 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1108 * perform necessary setup and returns 1 to indicate the device is set up 1109 * successfully to perform DDP on this I/O, otherwise this returns 0. 1110 * 1111 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1112 * struct netdev_fcoe_hbainfo *hbainfo); 1113 * Called when the FCoE Protocol stack wants information on the underlying 1114 * device. This information is utilized by the FCoE protocol stack to 1115 * register attributes with Fiber Channel management service as per the 1116 * FC-GS Fabric Device Management Information(FDMI) specification. 1117 * 1118 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1119 * Called when the underlying device wants to override default World Wide 1120 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1121 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1122 * protocol stack to use. 1123 * 1124 * RFS acceleration. 1125 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1126 * u16 rxq_index, u32 flow_id); 1127 * Set hardware filter for RFS. rxq_index is the target queue index; 1128 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1129 * Return the filter ID on success, or a negative error code. 1130 * 1131 * Slave management functions (for bridge, bonding, etc). 1132 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1133 * Called to make another netdev an underling. 1134 * 1135 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1136 * Called to release previously enslaved netdev. 1137 * 1138 * Feature/offload setting functions. 1139 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1140 * netdev_features_t features); 1141 * Adjusts the requested feature flags according to device-specific 1142 * constraints, and returns the resulting flags. Must not modify 1143 * the device state. 1144 * 1145 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1146 * Called to update device configuration to new features. Passed 1147 * feature set might be less than what was returned by ndo_fix_features()). 1148 * Must return >0 or -errno if it changed dev->features itself. 1149 * 1150 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1151 * struct net_device *dev, 1152 * const unsigned char *addr, u16 vid, u16 flags, 1153 * struct netlink_ext_ack *extack); 1154 * Adds an FDB entry to dev for addr. 1155 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1156 * struct net_device *dev, 1157 * const unsigned char *addr, u16 vid) 1158 * Deletes the FDB entry from dev coresponding to addr. 1159 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1160 * struct net_device *dev, struct net_device *filter_dev, 1161 * int *idx) 1162 * Used to add FDB entries to dump requests. Implementers should add 1163 * entries to skb and update idx with the number of entries. 1164 * 1165 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1166 * u16 flags, struct netlink_ext_ack *extack) 1167 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1168 * struct net_device *dev, u32 filter_mask, 1169 * int nlflags) 1170 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1171 * u16 flags); 1172 * 1173 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1174 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1175 * which do not represent real hardware may define this to allow their 1176 * userspace components to manage their virtual carrier state. Devices 1177 * that determine carrier state from physical hardware properties (eg 1178 * network cables) or protocol-dependent mechanisms (eg 1179 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1180 * 1181 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1182 * struct netdev_phys_item_id *ppid); 1183 * Called to get ID of physical port of this device. If driver does 1184 * not implement this, it is assumed that the hw is not able to have 1185 * multiple net devices on single physical port. 1186 * 1187 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1188 * struct netdev_phys_item_id *ppid) 1189 * Called to get the parent ID of the physical port of this device. 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 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1249 * This function is used to wake up the softirq, ksoftirqd or kthread 1250 * responsible for sending and/or receiving packets on a specific 1251 * queue id bound to an AF_XDP socket. The flags field specifies if 1252 * only RX, only Tx, or both should be woken up using the flags 1253 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1254 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1255 * Get devlink port instance associated with a given netdev. 1256 * Called with a reference on the netdevice and devlink locks only, 1257 * rtnl_lock is not held. 1258 */ 1259 struct net_device_ops { 1260 int (*ndo_init)(struct net_device *dev); 1261 void (*ndo_uninit)(struct net_device *dev); 1262 int (*ndo_open)(struct net_device *dev); 1263 int (*ndo_stop)(struct net_device *dev); 1264 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1265 struct net_device *dev); 1266 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1267 struct net_device *dev, 1268 netdev_features_t features); 1269 u16 (*ndo_select_queue)(struct net_device *dev, 1270 struct sk_buff *skb, 1271 struct net_device *sb_dev); 1272 void (*ndo_change_rx_flags)(struct net_device *dev, 1273 int flags); 1274 void (*ndo_set_rx_mode)(struct net_device *dev); 1275 int (*ndo_set_mac_address)(struct net_device *dev, 1276 void *addr); 1277 int (*ndo_validate_addr)(struct net_device *dev); 1278 int (*ndo_do_ioctl)(struct net_device *dev, 1279 struct ifreq *ifr, int cmd); 1280 int (*ndo_set_config)(struct net_device *dev, 1281 struct ifmap *map); 1282 int (*ndo_change_mtu)(struct net_device *dev, 1283 int new_mtu); 1284 int (*ndo_neigh_setup)(struct net_device *dev, 1285 struct neigh_parms *); 1286 void (*ndo_tx_timeout) (struct net_device *dev, 1287 unsigned int txqueue); 1288 1289 void (*ndo_get_stats64)(struct net_device *dev, 1290 struct rtnl_link_stats64 *storage); 1291 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1292 int (*ndo_get_offload_stats)(int attr_id, 1293 const struct net_device *dev, 1294 void *attr_data); 1295 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1296 1297 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1298 __be16 proto, u16 vid); 1299 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1300 __be16 proto, u16 vid); 1301 #ifdef CONFIG_NET_POLL_CONTROLLER 1302 void (*ndo_poll_controller)(struct net_device *dev); 1303 int (*ndo_netpoll_setup)(struct net_device *dev, 1304 struct netpoll_info *info); 1305 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1306 #endif 1307 int (*ndo_set_vf_mac)(struct net_device *dev, 1308 int queue, u8 *mac); 1309 int (*ndo_set_vf_vlan)(struct net_device *dev, 1310 int queue, u16 vlan, 1311 u8 qos, __be16 proto); 1312 int (*ndo_set_vf_rate)(struct net_device *dev, 1313 int vf, int min_tx_rate, 1314 int max_tx_rate); 1315 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1316 int vf, bool setting); 1317 int (*ndo_set_vf_trust)(struct net_device *dev, 1318 int vf, bool setting); 1319 int (*ndo_get_vf_config)(struct net_device *dev, 1320 int vf, 1321 struct ifla_vf_info *ivf); 1322 int (*ndo_set_vf_link_state)(struct net_device *dev, 1323 int vf, int link_state); 1324 int (*ndo_get_vf_stats)(struct net_device *dev, 1325 int vf, 1326 struct ifla_vf_stats 1327 *vf_stats); 1328 int (*ndo_set_vf_port)(struct net_device *dev, 1329 int vf, 1330 struct nlattr *port[]); 1331 int (*ndo_get_vf_port)(struct net_device *dev, 1332 int vf, struct sk_buff *skb); 1333 int (*ndo_get_vf_guid)(struct net_device *dev, 1334 int vf, 1335 struct ifla_vf_guid *node_guid, 1336 struct ifla_vf_guid *port_guid); 1337 int (*ndo_set_vf_guid)(struct net_device *dev, 1338 int vf, u64 guid, 1339 int guid_type); 1340 int (*ndo_set_vf_rss_query_en)( 1341 struct net_device *dev, 1342 int vf, bool setting); 1343 int (*ndo_setup_tc)(struct net_device *dev, 1344 enum tc_setup_type type, 1345 void *type_data); 1346 #if IS_ENABLED(CONFIG_FCOE) 1347 int (*ndo_fcoe_enable)(struct net_device *dev); 1348 int (*ndo_fcoe_disable)(struct net_device *dev); 1349 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1350 u16 xid, 1351 struct scatterlist *sgl, 1352 unsigned int sgc); 1353 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1354 u16 xid); 1355 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1356 u16 xid, 1357 struct scatterlist *sgl, 1358 unsigned int sgc); 1359 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1360 struct netdev_fcoe_hbainfo *hbainfo); 1361 #endif 1362 1363 #if IS_ENABLED(CONFIG_LIBFCOE) 1364 #define NETDEV_FCOE_WWNN 0 1365 #define NETDEV_FCOE_WWPN 1 1366 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1367 u64 *wwn, int type); 1368 #endif 1369 1370 #ifdef CONFIG_RFS_ACCEL 1371 int (*ndo_rx_flow_steer)(struct net_device *dev, 1372 const struct sk_buff *skb, 1373 u16 rxq_index, 1374 u32 flow_id); 1375 #endif 1376 int (*ndo_add_slave)(struct net_device *dev, 1377 struct net_device *slave_dev, 1378 struct netlink_ext_ack *extack); 1379 int (*ndo_del_slave)(struct net_device *dev, 1380 struct net_device *slave_dev); 1381 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1382 netdev_features_t features); 1383 int (*ndo_set_features)(struct net_device *dev, 1384 netdev_features_t features); 1385 int (*ndo_neigh_construct)(struct net_device *dev, 1386 struct neighbour *n); 1387 void (*ndo_neigh_destroy)(struct net_device *dev, 1388 struct neighbour *n); 1389 1390 int (*ndo_fdb_add)(struct ndmsg *ndm, 1391 struct nlattr *tb[], 1392 struct net_device *dev, 1393 const unsigned char *addr, 1394 u16 vid, 1395 u16 flags, 1396 struct netlink_ext_ack *extack); 1397 int (*ndo_fdb_del)(struct ndmsg *ndm, 1398 struct nlattr *tb[], 1399 struct net_device *dev, 1400 const unsigned char *addr, 1401 u16 vid); 1402 int (*ndo_fdb_dump)(struct sk_buff *skb, 1403 struct netlink_callback *cb, 1404 struct net_device *dev, 1405 struct net_device *filter_dev, 1406 int *idx); 1407 int (*ndo_fdb_get)(struct sk_buff *skb, 1408 struct nlattr *tb[], 1409 struct net_device *dev, 1410 const unsigned char *addr, 1411 u16 vid, u32 portid, u32 seq, 1412 struct netlink_ext_ack *extack); 1413 int (*ndo_bridge_setlink)(struct net_device *dev, 1414 struct nlmsghdr *nlh, 1415 u16 flags, 1416 struct netlink_ext_ack *extack); 1417 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1418 u32 pid, u32 seq, 1419 struct net_device *dev, 1420 u32 filter_mask, 1421 int nlflags); 1422 int (*ndo_bridge_dellink)(struct net_device *dev, 1423 struct nlmsghdr *nlh, 1424 u16 flags); 1425 int (*ndo_change_carrier)(struct net_device *dev, 1426 bool new_carrier); 1427 int (*ndo_get_phys_port_id)(struct net_device *dev, 1428 struct netdev_phys_item_id *ppid); 1429 int (*ndo_get_port_parent_id)(struct net_device *dev, 1430 struct netdev_phys_item_id *ppid); 1431 int (*ndo_get_phys_port_name)(struct net_device *dev, 1432 char *name, size_t len); 1433 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1434 struct udp_tunnel_info *ti); 1435 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1436 struct udp_tunnel_info *ti); 1437 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1438 struct net_device *dev); 1439 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1440 void *priv); 1441 1442 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1443 int queue_index, 1444 u32 maxrate); 1445 int (*ndo_get_iflink)(const struct net_device *dev); 1446 int (*ndo_change_proto_down)(struct net_device *dev, 1447 bool proto_down); 1448 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1449 struct sk_buff *skb); 1450 void (*ndo_set_rx_headroom)(struct net_device *dev, 1451 int needed_headroom); 1452 int (*ndo_bpf)(struct net_device *dev, 1453 struct netdev_bpf *bpf); 1454 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1455 struct xdp_frame **xdp, 1456 u32 flags); 1457 int (*ndo_xsk_wakeup)(struct net_device *dev, 1458 u32 queue_id, u32 flags); 1459 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1460 }; 1461 1462 /** 1463 * enum net_device_priv_flags - &struct net_device priv_flags 1464 * 1465 * These are the &struct net_device, they are only set internally 1466 * by drivers and used in the kernel. These flags are invisible to 1467 * userspace; this means that the order of these flags can change 1468 * during any kernel release. 1469 * 1470 * You should have a pretty good reason to be extending these flags. 1471 * 1472 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1473 * @IFF_EBRIDGE: Ethernet bridging device 1474 * @IFF_BONDING: bonding master or slave 1475 * @IFF_ISATAP: ISATAP interface (RFC4214) 1476 * @IFF_WAN_HDLC: WAN HDLC device 1477 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1478 * release skb->dst 1479 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1480 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1481 * @IFF_MACVLAN_PORT: device used as macvlan port 1482 * @IFF_BRIDGE_PORT: device used as bridge port 1483 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1484 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1485 * @IFF_UNICAST_FLT: Supports unicast filtering 1486 * @IFF_TEAM_PORT: device used as team port 1487 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1488 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1489 * change when it's running 1490 * @IFF_MACVLAN: Macvlan device 1491 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1492 * underlying stacked devices 1493 * @IFF_L3MDEV_MASTER: device is an L3 master device 1494 * @IFF_NO_QUEUE: device can run without qdisc attached 1495 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1496 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1497 * @IFF_TEAM: device is a team device 1498 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1499 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1500 * entity (i.e. the master device for bridged veth) 1501 * @IFF_MACSEC: device is a MACsec device 1502 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1503 * @IFF_FAILOVER: device is a failover master device 1504 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1505 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1506 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1507 */ 1508 enum netdev_priv_flags { 1509 IFF_802_1Q_VLAN = 1<<0, 1510 IFF_EBRIDGE = 1<<1, 1511 IFF_BONDING = 1<<2, 1512 IFF_ISATAP = 1<<3, 1513 IFF_WAN_HDLC = 1<<4, 1514 IFF_XMIT_DST_RELEASE = 1<<5, 1515 IFF_DONT_BRIDGE = 1<<6, 1516 IFF_DISABLE_NETPOLL = 1<<7, 1517 IFF_MACVLAN_PORT = 1<<8, 1518 IFF_BRIDGE_PORT = 1<<9, 1519 IFF_OVS_DATAPATH = 1<<10, 1520 IFF_TX_SKB_SHARING = 1<<11, 1521 IFF_UNICAST_FLT = 1<<12, 1522 IFF_TEAM_PORT = 1<<13, 1523 IFF_SUPP_NOFCS = 1<<14, 1524 IFF_LIVE_ADDR_CHANGE = 1<<15, 1525 IFF_MACVLAN = 1<<16, 1526 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1527 IFF_L3MDEV_MASTER = 1<<18, 1528 IFF_NO_QUEUE = 1<<19, 1529 IFF_OPENVSWITCH = 1<<20, 1530 IFF_L3MDEV_SLAVE = 1<<21, 1531 IFF_TEAM = 1<<22, 1532 IFF_RXFH_CONFIGURED = 1<<23, 1533 IFF_PHONY_HEADROOM = 1<<24, 1534 IFF_MACSEC = 1<<25, 1535 IFF_NO_RX_HANDLER = 1<<26, 1536 IFF_FAILOVER = 1<<27, 1537 IFF_FAILOVER_SLAVE = 1<<28, 1538 IFF_L3MDEV_RX_HANDLER = 1<<29, 1539 IFF_LIVE_RENAME_OK = 1<<30, 1540 }; 1541 1542 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1543 #define IFF_EBRIDGE IFF_EBRIDGE 1544 #define IFF_BONDING IFF_BONDING 1545 #define IFF_ISATAP IFF_ISATAP 1546 #define IFF_WAN_HDLC IFF_WAN_HDLC 1547 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1548 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1549 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1550 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1551 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1552 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1553 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1554 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1555 #define IFF_TEAM_PORT IFF_TEAM_PORT 1556 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1557 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1558 #define IFF_MACVLAN IFF_MACVLAN 1559 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1560 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1561 #define IFF_NO_QUEUE IFF_NO_QUEUE 1562 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1563 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1564 #define IFF_TEAM IFF_TEAM 1565 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1566 #define IFF_MACSEC IFF_MACSEC 1567 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1568 #define IFF_FAILOVER IFF_FAILOVER 1569 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1570 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1571 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1572 1573 /** 1574 * struct net_device - The DEVICE structure. 1575 * 1576 * Actually, this whole structure is a big mistake. It mixes I/O 1577 * data with strictly "high-level" data, and it has to know about 1578 * almost every data structure used in the INET module. 1579 * 1580 * @name: This is the first field of the "visible" part of this structure 1581 * (i.e. as seen by users in the "Space.c" file). It is the name 1582 * of the interface. 1583 * 1584 * @name_node: Name hashlist node 1585 * @ifalias: SNMP alias 1586 * @mem_end: Shared memory end 1587 * @mem_start: Shared memory start 1588 * @base_addr: Device I/O address 1589 * @irq: Device IRQ number 1590 * 1591 * @state: Generic network queuing layer state, see netdev_state_t 1592 * @dev_list: The global list of network devices 1593 * @napi_list: List entry used for polling NAPI devices 1594 * @unreg_list: List entry when we are unregistering the 1595 * device; see the function unregister_netdev 1596 * @close_list: List entry used when we are closing the device 1597 * @ptype_all: Device-specific packet handlers for all protocols 1598 * @ptype_specific: Device-specific, protocol-specific packet handlers 1599 * 1600 * @adj_list: Directly linked devices, like slaves for bonding 1601 * @features: Currently active device features 1602 * @hw_features: User-changeable features 1603 * 1604 * @wanted_features: User-requested features 1605 * @vlan_features: Mask of features inheritable by VLAN devices 1606 * 1607 * @hw_enc_features: Mask of features inherited by encapsulating devices 1608 * This field indicates what encapsulation 1609 * offloads the hardware is capable of doing, 1610 * and drivers will need to set them appropriately. 1611 * 1612 * @mpls_features: Mask of features inheritable by MPLS 1613 * 1614 * @ifindex: interface index 1615 * @group: The group the device belongs to 1616 * 1617 * @stats: Statistics struct, which was left as a legacy, use 1618 * rtnl_link_stats64 instead 1619 * 1620 * @rx_dropped: Dropped packets by core network, 1621 * do not use this in drivers 1622 * @tx_dropped: Dropped packets by core network, 1623 * do not use this in drivers 1624 * @rx_nohandler: nohandler dropped packets by core network on 1625 * inactive devices, do not use this in drivers 1626 * @carrier_up_count: Number of times the carrier has been up 1627 * @carrier_down_count: Number of times the carrier has been down 1628 * 1629 * @wireless_handlers: List of functions to handle Wireless Extensions, 1630 * instead of ioctl, 1631 * see <net/iw_handler.h> for details. 1632 * @wireless_data: Instance data managed by the core of wireless extensions 1633 * 1634 * @netdev_ops: Includes several pointers to callbacks, 1635 * if one wants to override the ndo_*() functions 1636 * @ethtool_ops: Management operations 1637 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1638 * discovery handling. Necessary for e.g. 6LoWPAN. 1639 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1640 * of Layer 2 headers. 1641 * 1642 * @flags: Interface flags (a la BSD) 1643 * @priv_flags: Like 'flags' but invisible to userspace, 1644 * see if.h for the definitions 1645 * @gflags: Global flags ( kept as legacy ) 1646 * @padded: How much padding added by alloc_netdev() 1647 * @operstate: RFC2863 operstate 1648 * @link_mode: Mapping policy to operstate 1649 * @if_port: Selectable AUI, TP, ... 1650 * @dma: DMA channel 1651 * @mtu: Interface MTU value 1652 * @min_mtu: Interface Minimum MTU value 1653 * @max_mtu: Interface Maximum MTU value 1654 * @type: Interface hardware type 1655 * @hard_header_len: Maximum hardware header length. 1656 * @min_header_len: Minimum hardware header length 1657 * 1658 * @needed_headroom: Extra headroom the hardware may need, but not in all 1659 * cases can this be guaranteed 1660 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1661 * cases can this be guaranteed. Some cases also use 1662 * LL_MAX_HEADER instead to allocate the skb 1663 * 1664 * interface address info: 1665 * 1666 * @perm_addr: Permanent hw address 1667 * @addr_assign_type: Hw address assignment type 1668 * @addr_len: Hardware address length 1669 * @upper_level: Maximum depth level of upper devices. 1670 * @lower_level: Maximum depth level of lower devices. 1671 * @neigh_priv_len: Used in neigh_alloc() 1672 * @dev_id: Used to differentiate devices that share 1673 * the same link layer address 1674 * @dev_port: Used to differentiate devices that share 1675 * the same function 1676 * @addr_list_lock: XXX: need comments on this one 1677 * @uc_promisc: Counter that indicates promiscuous mode 1678 * has been enabled due to the need to listen to 1679 * additional unicast addresses in a device that 1680 * does not implement ndo_set_rx_mode() 1681 * @uc: unicast mac addresses 1682 * @mc: multicast mac addresses 1683 * @dev_addrs: list of device hw addresses 1684 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1685 * @promiscuity: Number of times the NIC is told to work in 1686 * promiscuous mode; if it becomes 0 the NIC will 1687 * exit promiscuous mode 1688 * @allmulti: Counter, enables or disables allmulticast mode 1689 * 1690 * @vlan_info: VLAN info 1691 * @dsa_ptr: dsa specific data 1692 * @tipc_ptr: TIPC specific data 1693 * @atalk_ptr: AppleTalk link 1694 * @ip_ptr: IPv4 specific data 1695 * @dn_ptr: DECnet specific data 1696 * @ip6_ptr: IPv6 specific data 1697 * @ax25_ptr: AX.25 specific data 1698 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1699 * 1700 * @dev_addr: Hw address (before bcast, 1701 * because most packets are unicast) 1702 * 1703 * @_rx: Array of RX queues 1704 * @num_rx_queues: Number of RX queues 1705 * allocated at register_netdev() time 1706 * @real_num_rx_queues: Number of RX queues currently active in device 1707 * 1708 * @rx_handler: handler for received packets 1709 * @rx_handler_data: XXX: need comments on this one 1710 * @miniq_ingress: ingress/clsact qdisc specific data for 1711 * ingress processing 1712 * @ingress_queue: XXX: need comments on this one 1713 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1714 * @broadcast: hw bcast address 1715 * 1716 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1717 * indexed by RX queue number. Assigned by driver. 1718 * This must only be set if the ndo_rx_flow_steer 1719 * operation is defined 1720 * @index_hlist: Device index hash chain 1721 * 1722 * @_tx: Array of TX queues 1723 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1724 * @real_num_tx_queues: Number of TX queues currently active in device 1725 * @qdisc: Root qdisc from userspace point of view 1726 * @tx_queue_len: Max frames per queue allowed 1727 * @tx_global_lock: XXX: need comments on this one 1728 * 1729 * @xps_maps: XXX: need comments on this one 1730 * @miniq_egress: clsact qdisc specific data for 1731 * egress processing 1732 * @watchdog_timeo: Represents the timeout that is used by 1733 * the watchdog (see dev_watchdog()) 1734 * @watchdog_timer: List of timers 1735 * 1736 * @pcpu_refcnt: Number of references to this device 1737 * @todo_list: Delayed register/unregister 1738 * @link_watch_list: XXX: need comments on this one 1739 * 1740 * @reg_state: Register/unregister state machine 1741 * @dismantle: Device is going to be freed 1742 * @rtnl_link_state: This enum represents the phases of creating 1743 * a new link 1744 * 1745 * @needs_free_netdev: Should unregister perform free_netdev? 1746 * @priv_destructor: Called from unregister 1747 * @npinfo: XXX: need comments on this one 1748 * @nd_net: Network namespace this network device is inside 1749 * 1750 * @ml_priv: Mid-layer private 1751 * @lstats: Loopback statistics 1752 * @tstats: Tunnel statistics 1753 * @dstats: Dummy statistics 1754 * @vstats: Virtual ethernet statistics 1755 * 1756 * @garp_port: GARP 1757 * @mrp_port: MRP 1758 * 1759 * @dev: Class/net/name entry 1760 * @sysfs_groups: Space for optional device, statistics and wireless 1761 * sysfs groups 1762 * 1763 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1764 * @rtnl_link_ops: Rtnl_link_ops 1765 * 1766 * @gso_max_size: Maximum size of generic segmentation offload 1767 * @gso_max_segs: Maximum number of segments that can be passed to the 1768 * NIC for GSO 1769 * 1770 * @dcbnl_ops: Data Center Bridging netlink ops 1771 * @num_tc: Number of traffic classes in the net device 1772 * @tc_to_txq: XXX: need comments on this one 1773 * @prio_tc_map: XXX: need comments on this one 1774 * 1775 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1776 * 1777 * @priomap: XXX: need comments on this one 1778 * @phydev: Physical device may attach itself 1779 * for hardware timestamping 1780 * @sfp_bus: attached &struct sfp_bus structure. 1781 * @qdisc_tx_busylock_key: lockdep class annotating Qdisc->busylock 1782 * spinlock 1783 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1784 * @qdisc_xmit_lock_key: lockdep class annotating 1785 * netdev_queue->_xmit_lock spinlock 1786 * @addr_list_lock_key: lockdep class annotating 1787 * net_device->addr_list_lock spinlock 1788 * 1789 * @proto_down: protocol port state information can be sent to the 1790 * switch driver and used to set the phys state of the 1791 * switch port. 1792 * 1793 * @wol_enabled: Wake-on-LAN is enabled 1794 * 1795 * FIXME: cleanup struct net_device such that network protocol info 1796 * moves out. 1797 */ 1798 1799 struct net_device { 1800 char name[IFNAMSIZ]; 1801 struct netdev_name_node *name_node; 1802 struct dev_ifalias __rcu *ifalias; 1803 /* 1804 * I/O specific fields 1805 * FIXME: Merge these and struct ifmap into one 1806 */ 1807 unsigned long mem_end; 1808 unsigned long mem_start; 1809 unsigned long base_addr; 1810 int irq; 1811 1812 /* 1813 * Some hardware also needs these fields (state,dev_list, 1814 * napi_list,unreg_list,close_list) but they are not 1815 * part of the usual set specified in Space.c. 1816 */ 1817 1818 unsigned long state; 1819 1820 struct list_head dev_list; 1821 struct list_head napi_list; 1822 struct list_head unreg_list; 1823 struct list_head close_list; 1824 struct list_head ptype_all; 1825 struct list_head ptype_specific; 1826 1827 struct { 1828 struct list_head upper; 1829 struct list_head lower; 1830 } adj_list; 1831 1832 netdev_features_t features; 1833 netdev_features_t hw_features; 1834 netdev_features_t wanted_features; 1835 netdev_features_t vlan_features; 1836 netdev_features_t hw_enc_features; 1837 netdev_features_t mpls_features; 1838 netdev_features_t gso_partial_features; 1839 1840 int ifindex; 1841 int group; 1842 1843 struct net_device_stats stats; 1844 1845 atomic_long_t rx_dropped; 1846 atomic_long_t tx_dropped; 1847 atomic_long_t rx_nohandler; 1848 1849 /* Stats to monitor link on/off, flapping */ 1850 atomic_t carrier_up_count; 1851 atomic_t carrier_down_count; 1852 1853 #ifdef CONFIG_WIRELESS_EXT 1854 const struct iw_handler_def *wireless_handlers; 1855 struct iw_public_data *wireless_data; 1856 #endif 1857 const struct net_device_ops *netdev_ops; 1858 const struct ethtool_ops *ethtool_ops; 1859 #ifdef CONFIG_NET_L3_MASTER_DEV 1860 const struct l3mdev_ops *l3mdev_ops; 1861 #endif 1862 #if IS_ENABLED(CONFIG_IPV6) 1863 const struct ndisc_ops *ndisc_ops; 1864 #endif 1865 1866 #ifdef CONFIG_XFRM_OFFLOAD 1867 const struct xfrmdev_ops *xfrmdev_ops; 1868 #endif 1869 1870 #if IS_ENABLED(CONFIG_TLS_DEVICE) 1871 const struct tlsdev_ops *tlsdev_ops; 1872 #endif 1873 1874 const struct header_ops *header_ops; 1875 1876 unsigned int flags; 1877 unsigned int priv_flags; 1878 1879 unsigned short gflags; 1880 unsigned short padded; 1881 1882 unsigned char operstate; 1883 unsigned char link_mode; 1884 1885 unsigned char if_port; 1886 unsigned char dma; 1887 1888 /* Note : dev->mtu is often read without holding a lock. 1889 * Writers usually hold RTNL. 1890 * It is recommended to use READ_ONCE() to annotate the reads, 1891 * and to use WRITE_ONCE() to annotate the writes. 1892 */ 1893 unsigned int mtu; 1894 unsigned int min_mtu; 1895 unsigned int max_mtu; 1896 unsigned short type; 1897 unsigned short hard_header_len; 1898 unsigned char min_header_len; 1899 1900 unsigned short needed_headroom; 1901 unsigned short needed_tailroom; 1902 1903 /* Interface address info. */ 1904 unsigned char perm_addr[MAX_ADDR_LEN]; 1905 unsigned char addr_assign_type; 1906 unsigned char addr_len; 1907 unsigned char upper_level; 1908 unsigned char lower_level; 1909 unsigned short neigh_priv_len; 1910 unsigned short dev_id; 1911 unsigned short dev_port; 1912 spinlock_t addr_list_lock; 1913 unsigned char name_assign_type; 1914 bool uc_promisc; 1915 struct netdev_hw_addr_list uc; 1916 struct netdev_hw_addr_list mc; 1917 struct netdev_hw_addr_list dev_addrs; 1918 1919 #ifdef CONFIG_SYSFS 1920 struct kset *queues_kset; 1921 #endif 1922 unsigned int promiscuity; 1923 unsigned int allmulti; 1924 1925 1926 /* Protocol-specific pointers */ 1927 1928 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1929 struct vlan_info __rcu *vlan_info; 1930 #endif 1931 #if IS_ENABLED(CONFIG_NET_DSA) 1932 struct dsa_port *dsa_ptr; 1933 #endif 1934 #if IS_ENABLED(CONFIG_TIPC) 1935 struct tipc_bearer __rcu *tipc_ptr; 1936 #endif 1937 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 1938 void *atalk_ptr; 1939 #endif 1940 struct in_device __rcu *ip_ptr; 1941 #if IS_ENABLED(CONFIG_DECNET) 1942 struct dn_dev __rcu *dn_ptr; 1943 #endif 1944 struct inet6_dev __rcu *ip6_ptr; 1945 #if IS_ENABLED(CONFIG_AX25) 1946 void *ax25_ptr; 1947 #endif 1948 struct wireless_dev *ieee80211_ptr; 1949 struct wpan_dev *ieee802154_ptr; 1950 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 1951 struct mpls_dev __rcu *mpls_ptr; 1952 #endif 1953 1954 /* 1955 * Cache lines mostly used on receive path (including eth_type_trans()) 1956 */ 1957 /* Interface address info used in eth_type_trans() */ 1958 unsigned char *dev_addr; 1959 1960 struct netdev_rx_queue *_rx; 1961 unsigned int num_rx_queues; 1962 unsigned int real_num_rx_queues; 1963 1964 struct bpf_prog __rcu *xdp_prog; 1965 unsigned long gro_flush_timeout; 1966 rx_handler_func_t __rcu *rx_handler; 1967 void __rcu *rx_handler_data; 1968 1969 #ifdef CONFIG_NET_CLS_ACT 1970 struct mini_Qdisc __rcu *miniq_ingress; 1971 #endif 1972 struct netdev_queue __rcu *ingress_queue; 1973 #ifdef CONFIG_NETFILTER_INGRESS 1974 struct nf_hook_entries __rcu *nf_hooks_ingress; 1975 #endif 1976 1977 unsigned char broadcast[MAX_ADDR_LEN]; 1978 #ifdef CONFIG_RFS_ACCEL 1979 struct cpu_rmap *rx_cpu_rmap; 1980 #endif 1981 struct hlist_node index_hlist; 1982 1983 /* 1984 * Cache lines mostly used on transmit path 1985 */ 1986 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1987 unsigned int num_tx_queues; 1988 unsigned int real_num_tx_queues; 1989 struct Qdisc *qdisc; 1990 unsigned int tx_queue_len; 1991 spinlock_t tx_global_lock; 1992 1993 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 1994 1995 #ifdef CONFIG_XPS 1996 struct xps_dev_maps __rcu *xps_cpus_map; 1997 struct xps_dev_maps __rcu *xps_rxqs_map; 1998 #endif 1999 #ifdef CONFIG_NET_CLS_ACT 2000 struct mini_Qdisc __rcu *miniq_egress; 2001 #endif 2002 2003 #ifdef CONFIG_NET_SCHED 2004 DECLARE_HASHTABLE (qdisc_hash, 4); 2005 #endif 2006 /* These may be needed for future network-power-down code. */ 2007 struct timer_list watchdog_timer; 2008 int watchdog_timeo; 2009 2010 struct list_head todo_list; 2011 int __percpu *pcpu_refcnt; 2012 2013 struct list_head link_watch_list; 2014 2015 enum { NETREG_UNINITIALIZED=0, 2016 NETREG_REGISTERED, /* completed register_netdevice */ 2017 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2018 NETREG_UNREGISTERED, /* completed unregister todo */ 2019 NETREG_RELEASED, /* called free_netdev */ 2020 NETREG_DUMMY, /* dummy device for NAPI poll */ 2021 } reg_state:8; 2022 2023 bool dismantle; 2024 2025 enum { 2026 RTNL_LINK_INITIALIZED, 2027 RTNL_LINK_INITIALIZING, 2028 } rtnl_link_state:16; 2029 2030 bool needs_free_netdev; 2031 void (*priv_destructor)(struct net_device *dev); 2032 2033 #ifdef CONFIG_NETPOLL 2034 struct netpoll_info __rcu *npinfo; 2035 #endif 2036 2037 possible_net_t nd_net; 2038 2039 /* mid-layer private */ 2040 union { 2041 void *ml_priv; 2042 struct pcpu_lstats __percpu *lstats; 2043 struct pcpu_sw_netstats __percpu *tstats; 2044 struct pcpu_dstats __percpu *dstats; 2045 }; 2046 2047 #if IS_ENABLED(CONFIG_GARP) 2048 struct garp_port __rcu *garp_port; 2049 #endif 2050 #if IS_ENABLED(CONFIG_MRP) 2051 struct mrp_port __rcu *mrp_port; 2052 #endif 2053 2054 struct device dev; 2055 const struct attribute_group *sysfs_groups[4]; 2056 const struct attribute_group *sysfs_rx_queue_group; 2057 2058 const struct rtnl_link_ops *rtnl_link_ops; 2059 2060 /* for setting kernel sock attribute on TCP connection setup */ 2061 #define GSO_MAX_SIZE 65536 2062 unsigned int gso_max_size; 2063 #define GSO_MAX_SEGS 65535 2064 u16 gso_max_segs; 2065 2066 #ifdef CONFIG_DCB 2067 const struct dcbnl_rtnl_ops *dcbnl_ops; 2068 #endif 2069 s16 num_tc; 2070 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2071 u8 prio_tc_map[TC_BITMASK + 1]; 2072 2073 #if IS_ENABLED(CONFIG_FCOE) 2074 unsigned int fcoe_ddp_xid; 2075 #endif 2076 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2077 struct netprio_map __rcu *priomap; 2078 #endif 2079 struct phy_device *phydev; 2080 struct sfp_bus *sfp_bus; 2081 struct lock_class_key qdisc_tx_busylock_key; 2082 struct lock_class_key qdisc_running_key; 2083 struct lock_class_key qdisc_xmit_lock_key; 2084 struct lock_class_key addr_list_lock_key; 2085 bool proto_down; 2086 unsigned wol_enabled:1; 2087 }; 2088 #define to_net_dev(d) container_of(d, struct net_device, dev) 2089 2090 static inline bool netif_elide_gro(const struct net_device *dev) 2091 { 2092 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2093 return true; 2094 return false; 2095 } 2096 2097 #define NETDEV_ALIGN 32 2098 2099 static inline 2100 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2101 { 2102 return dev->prio_tc_map[prio & TC_BITMASK]; 2103 } 2104 2105 static inline 2106 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2107 { 2108 if (tc >= dev->num_tc) 2109 return -EINVAL; 2110 2111 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2112 return 0; 2113 } 2114 2115 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2116 void netdev_reset_tc(struct net_device *dev); 2117 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2118 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2119 2120 static inline 2121 int netdev_get_num_tc(struct net_device *dev) 2122 { 2123 return dev->num_tc; 2124 } 2125 2126 void netdev_unbind_sb_channel(struct net_device *dev, 2127 struct net_device *sb_dev); 2128 int netdev_bind_sb_channel_queue(struct net_device *dev, 2129 struct net_device *sb_dev, 2130 u8 tc, u16 count, u16 offset); 2131 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2132 static inline int netdev_get_sb_channel(struct net_device *dev) 2133 { 2134 return max_t(int, -dev->num_tc, 0); 2135 } 2136 2137 static inline 2138 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2139 unsigned int index) 2140 { 2141 return &dev->_tx[index]; 2142 } 2143 2144 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2145 const struct sk_buff *skb) 2146 { 2147 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2148 } 2149 2150 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2151 void (*f)(struct net_device *, 2152 struct netdev_queue *, 2153 void *), 2154 void *arg) 2155 { 2156 unsigned int i; 2157 2158 for (i = 0; i < dev->num_tx_queues; i++) 2159 f(dev, &dev->_tx[i], arg); 2160 } 2161 2162 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2163 struct net_device *sb_dev); 2164 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2165 struct sk_buff *skb, 2166 struct net_device *sb_dev); 2167 2168 /* returns the headroom that the master device needs to take in account 2169 * when forwarding to this dev 2170 */ 2171 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2172 { 2173 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2174 } 2175 2176 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2177 { 2178 if (dev->netdev_ops->ndo_set_rx_headroom) 2179 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2180 } 2181 2182 /* set the device rx headroom to the dev's default */ 2183 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2184 { 2185 netdev_set_rx_headroom(dev, -1); 2186 } 2187 2188 /* 2189 * Net namespace inlines 2190 */ 2191 static inline 2192 struct net *dev_net(const struct net_device *dev) 2193 { 2194 return read_pnet(&dev->nd_net); 2195 } 2196 2197 static inline 2198 void dev_net_set(struct net_device *dev, struct net *net) 2199 { 2200 write_pnet(&dev->nd_net, net); 2201 } 2202 2203 /** 2204 * netdev_priv - access network device private data 2205 * @dev: network device 2206 * 2207 * Get network device private data 2208 */ 2209 static inline void *netdev_priv(const struct net_device *dev) 2210 { 2211 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2212 } 2213 2214 /* Set the sysfs physical device reference for the network logical device 2215 * if set prior to registration will cause a symlink during initialization. 2216 */ 2217 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2218 2219 /* Set the sysfs device type for the network logical device to allow 2220 * fine-grained identification of different network device types. For 2221 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2222 */ 2223 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2224 2225 /* Default NAPI poll() weight 2226 * Device drivers are strongly advised to not use bigger value 2227 */ 2228 #define NAPI_POLL_WEIGHT 64 2229 2230 /** 2231 * netif_napi_add - initialize a NAPI context 2232 * @dev: network device 2233 * @napi: NAPI context 2234 * @poll: polling function 2235 * @weight: default weight 2236 * 2237 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2238 * *any* of the other NAPI-related functions. 2239 */ 2240 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2241 int (*poll)(struct napi_struct *, int), int weight); 2242 2243 /** 2244 * netif_tx_napi_add - initialize a NAPI context 2245 * @dev: network device 2246 * @napi: NAPI context 2247 * @poll: polling function 2248 * @weight: default weight 2249 * 2250 * This variant of netif_napi_add() should be used from drivers using NAPI 2251 * to exclusively poll a TX queue. 2252 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2253 */ 2254 static inline void netif_tx_napi_add(struct net_device *dev, 2255 struct napi_struct *napi, 2256 int (*poll)(struct napi_struct *, int), 2257 int weight) 2258 { 2259 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2260 netif_napi_add(dev, napi, poll, weight); 2261 } 2262 2263 /** 2264 * netif_napi_del - remove a NAPI context 2265 * @napi: NAPI context 2266 * 2267 * netif_napi_del() removes a NAPI context from the network device NAPI list 2268 */ 2269 void netif_napi_del(struct napi_struct *napi); 2270 2271 struct napi_gro_cb { 2272 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2273 void *frag0; 2274 2275 /* Length of frag0. */ 2276 unsigned int frag0_len; 2277 2278 /* This indicates where we are processing relative to skb->data. */ 2279 int data_offset; 2280 2281 /* This is non-zero if the packet cannot be merged with the new skb. */ 2282 u16 flush; 2283 2284 /* Save the IP ID here and check when we get to the transport layer */ 2285 u16 flush_id; 2286 2287 /* Number of segments aggregated. */ 2288 u16 count; 2289 2290 /* Start offset for remote checksum offload */ 2291 u16 gro_remcsum_start; 2292 2293 /* jiffies when first packet was created/queued */ 2294 unsigned long age; 2295 2296 /* Used in ipv6_gro_receive() and foo-over-udp */ 2297 u16 proto; 2298 2299 /* This is non-zero if the packet may be of the same flow. */ 2300 u8 same_flow:1; 2301 2302 /* Used in tunnel GRO receive */ 2303 u8 encap_mark:1; 2304 2305 /* GRO checksum is valid */ 2306 u8 csum_valid:1; 2307 2308 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2309 u8 csum_cnt:3; 2310 2311 /* Free the skb? */ 2312 u8 free:2; 2313 #define NAPI_GRO_FREE 1 2314 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2315 2316 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2317 u8 is_ipv6:1; 2318 2319 /* Used in GRE, set in fou/gue_gro_receive */ 2320 u8 is_fou:1; 2321 2322 /* Used to determine if flush_id can be ignored */ 2323 u8 is_atomic:1; 2324 2325 /* Number of gro_receive callbacks this packet already went through */ 2326 u8 recursion_counter:4; 2327 2328 /* 1 bit hole */ 2329 2330 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2331 __wsum csum; 2332 2333 /* used in skb_gro_receive() slow path */ 2334 struct sk_buff *last; 2335 }; 2336 2337 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2338 2339 #define GRO_RECURSION_LIMIT 15 2340 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2341 { 2342 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2343 } 2344 2345 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2346 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2347 struct list_head *head, 2348 struct sk_buff *skb) 2349 { 2350 if (unlikely(gro_recursion_inc_test(skb))) { 2351 NAPI_GRO_CB(skb)->flush |= 1; 2352 return NULL; 2353 } 2354 2355 return cb(head, skb); 2356 } 2357 2358 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2359 struct sk_buff *); 2360 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2361 struct sock *sk, 2362 struct list_head *head, 2363 struct sk_buff *skb) 2364 { 2365 if (unlikely(gro_recursion_inc_test(skb))) { 2366 NAPI_GRO_CB(skb)->flush |= 1; 2367 return NULL; 2368 } 2369 2370 return cb(sk, head, skb); 2371 } 2372 2373 struct packet_type { 2374 __be16 type; /* This is really htons(ether_type). */ 2375 bool ignore_outgoing; 2376 struct net_device *dev; /* NULL is wildcarded here */ 2377 int (*func) (struct sk_buff *, 2378 struct net_device *, 2379 struct packet_type *, 2380 struct net_device *); 2381 void (*list_func) (struct list_head *, 2382 struct packet_type *, 2383 struct net_device *); 2384 bool (*id_match)(struct packet_type *ptype, 2385 struct sock *sk); 2386 void *af_packet_priv; 2387 struct list_head list; 2388 }; 2389 2390 struct offload_callbacks { 2391 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2392 netdev_features_t features); 2393 struct sk_buff *(*gro_receive)(struct list_head *head, 2394 struct sk_buff *skb); 2395 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2396 }; 2397 2398 struct packet_offload { 2399 __be16 type; /* This is really htons(ether_type). */ 2400 u16 priority; 2401 struct offload_callbacks callbacks; 2402 struct list_head list; 2403 }; 2404 2405 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2406 struct pcpu_sw_netstats { 2407 u64 rx_packets; 2408 u64 rx_bytes; 2409 u64 tx_packets; 2410 u64 tx_bytes; 2411 struct u64_stats_sync syncp; 2412 } __aligned(4 * sizeof(u64)); 2413 2414 struct pcpu_lstats { 2415 u64_stats_t packets; 2416 u64_stats_t bytes; 2417 struct u64_stats_sync syncp; 2418 } __aligned(2 * sizeof(u64)); 2419 2420 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2421 2422 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2423 { 2424 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2425 2426 u64_stats_update_begin(&lstats->syncp); 2427 u64_stats_add(&lstats->bytes, len); 2428 u64_stats_inc(&lstats->packets); 2429 u64_stats_update_end(&lstats->syncp); 2430 } 2431 2432 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2433 ({ \ 2434 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2435 if (pcpu_stats) { \ 2436 int __cpu; \ 2437 for_each_possible_cpu(__cpu) { \ 2438 typeof(type) *stat; \ 2439 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2440 u64_stats_init(&stat->syncp); \ 2441 } \ 2442 } \ 2443 pcpu_stats; \ 2444 }) 2445 2446 #define netdev_alloc_pcpu_stats(type) \ 2447 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2448 2449 enum netdev_lag_tx_type { 2450 NETDEV_LAG_TX_TYPE_UNKNOWN, 2451 NETDEV_LAG_TX_TYPE_RANDOM, 2452 NETDEV_LAG_TX_TYPE_BROADCAST, 2453 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2454 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2455 NETDEV_LAG_TX_TYPE_HASH, 2456 }; 2457 2458 enum netdev_lag_hash { 2459 NETDEV_LAG_HASH_NONE, 2460 NETDEV_LAG_HASH_L2, 2461 NETDEV_LAG_HASH_L34, 2462 NETDEV_LAG_HASH_L23, 2463 NETDEV_LAG_HASH_E23, 2464 NETDEV_LAG_HASH_E34, 2465 NETDEV_LAG_HASH_UNKNOWN, 2466 }; 2467 2468 struct netdev_lag_upper_info { 2469 enum netdev_lag_tx_type tx_type; 2470 enum netdev_lag_hash hash_type; 2471 }; 2472 2473 struct netdev_lag_lower_state_info { 2474 u8 link_up : 1, 2475 tx_enabled : 1; 2476 }; 2477 2478 #include <linux/notifier.h> 2479 2480 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2481 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2482 * adding new types. 2483 */ 2484 enum netdev_cmd { 2485 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2486 NETDEV_DOWN, 2487 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2488 detected a hardware crash and restarted 2489 - we can use this eg to kick tcp sessions 2490 once done */ 2491 NETDEV_CHANGE, /* Notify device state change */ 2492 NETDEV_REGISTER, 2493 NETDEV_UNREGISTER, 2494 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2495 NETDEV_CHANGEADDR, /* notify after the address change */ 2496 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2497 NETDEV_GOING_DOWN, 2498 NETDEV_CHANGENAME, 2499 NETDEV_FEAT_CHANGE, 2500 NETDEV_BONDING_FAILOVER, 2501 NETDEV_PRE_UP, 2502 NETDEV_PRE_TYPE_CHANGE, 2503 NETDEV_POST_TYPE_CHANGE, 2504 NETDEV_POST_INIT, 2505 NETDEV_RELEASE, 2506 NETDEV_NOTIFY_PEERS, 2507 NETDEV_JOIN, 2508 NETDEV_CHANGEUPPER, 2509 NETDEV_RESEND_IGMP, 2510 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2511 NETDEV_CHANGEINFODATA, 2512 NETDEV_BONDING_INFO, 2513 NETDEV_PRECHANGEUPPER, 2514 NETDEV_CHANGELOWERSTATE, 2515 NETDEV_UDP_TUNNEL_PUSH_INFO, 2516 NETDEV_UDP_TUNNEL_DROP_INFO, 2517 NETDEV_CHANGE_TX_QUEUE_LEN, 2518 NETDEV_CVLAN_FILTER_PUSH_INFO, 2519 NETDEV_CVLAN_FILTER_DROP_INFO, 2520 NETDEV_SVLAN_FILTER_PUSH_INFO, 2521 NETDEV_SVLAN_FILTER_DROP_INFO, 2522 }; 2523 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2524 2525 int register_netdevice_notifier(struct notifier_block *nb); 2526 int unregister_netdevice_notifier(struct notifier_block *nb); 2527 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2528 int unregister_netdevice_notifier_net(struct net *net, 2529 struct notifier_block *nb); 2530 2531 struct netdev_notifier_info { 2532 struct net_device *dev; 2533 struct netlink_ext_ack *extack; 2534 }; 2535 2536 struct netdev_notifier_info_ext { 2537 struct netdev_notifier_info info; /* must be first */ 2538 union { 2539 u32 mtu; 2540 } ext; 2541 }; 2542 2543 struct netdev_notifier_change_info { 2544 struct netdev_notifier_info info; /* must be first */ 2545 unsigned int flags_changed; 2546 }; 2547 2548 struct netdev_notifier_changeupper_info { 2549 struct netdev_notifier_info info; /* must be first */ 2550 struct net_device *upper_dev; /* new upper dev */ 2551 bool master; /* is upper dev master */ 2552 bool linking; /* is the notification for link or unlink */ 2553 void *upper_info; /* upper dev info */ 2554 }; 2555 2556 struct netdev_notifier_changelowerstate_info { 2557 struct netdev_notifier_info info; /* must be first */ 2558 void *lower_state_info; /* is lower dev state */ 2559 }; 2560 2561 struct netdev_notifier_pre_changeaddr_info { 2562 struct netdev_notifier_info info; /* must be first */ 2563 const unsigned char *dev_addr; 2564 }; 2565 2566 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2567 struct net_device *dev) 2568 { 2569 info->dev = dev; 2570 info->extack = NULL; 2571 } 2572 2573 static inline struct net_device * 2574 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2575 { 2576 return info->dev; 2577 } 2578 2579 static inline struct netlink_ext_ack * 2580 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2581 { 2582 return info->extack; 2583 } 2584 2585 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2586 2587 2588 extern rwlock_t dev_base_lock; /* Device list lock */ 2589 2590 #define for_each_netdev(net, d) \ 2591 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2592 #define for_each_netdev_reverse(net, d) \ 2593 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2594 #define for_each_netdev_rcu(net, d) \ 2595 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2596 #define for_each_netdev_safe(net, d, n) \ 2597 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2598 #define for_each_netdev_continue(net, d) \ 2599 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2600 #define for_each_netdev_continue_reverse(net, d) \ 2601 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2602 dev_list) 2603 #define for_each_netdev_continue_rcu(net, d) \ 2604 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2605 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2606 for_each_netdev_rcu(&init_net, slave) \ 2607 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2608 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2609 2610 static inline struct net_device *next_net_device(struct net_device *dev) 2611 { 2612 struct list_head *lh; 2613 struct net *net; 2614 2615 net = dev_net(dev); 2616 lh = dev->dev_list.next; 2617 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2618 } 2619 2620 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2621 { 2622 struct list_head *lh; 2623 struct net *net; 2624 2625 net = dev_net(dev); 2626 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2627 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2628 } 2629 2630 static inline struct net_device *first_net_device(struct net *net) 2631 { 2632 return list_empty(&net->dev_base_head) ? NULL : 2633 net_device_entry(net->dev_base_head.next); 2634 } 2635 2636 static inline struct net_device *first_net_device_rcu(struct net *net) 2637 { 2638 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2639 2640 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2641 } 2642 2643 int netdev_boot_setup_check(struct net_device *dev); 2644 unsigned long netdev_boot_base(const char *prefix, int unit); 2645 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2646 const char *hwaddr); 2647 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2648 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2649 void dev_add_pack(struct packet_type *pt); 2650 void dev_remove_pack(struct packet_type *pt); 2651 void __dev_remove_pack(struct packet_type *pt); 2652 void dev_add_offload(struct packet_offload *po); 2653 void dev_remove_offload(struct packet_offload *po); 2654 2655 int dev_get_iflink(const struct net_device *dev); 2656 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2657 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2658 unsigned short mask); 2659 struct net_device *dev_get_by_name(struct net *net, const char *name); 2660 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2661 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2662 int dev_alloc_name(struct net_device *dev, const char *name); 2663 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2664 void dev_close(struct net_device *dev); 2665 void dev_close_many(struct list_head *head, bool unlink); 2666 void dev_disable_lro(struct net_device *dev); 2667 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2668 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2669 struct net_device *sb_dev); 2670 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2671 struct net_device *sb_dev); 2672 int dev_queue_xmit(struct sk_buff *skb); 2673 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2674 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2675 int register_netdevice(struct net_device *dev); 2676 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2677 void unregister_netdevice_many(struct list_head *head); 2678 static inline void unregister_netdevice(struct net_device *dev) 2679 { 2680 unregister_netdevice_queue(dev, NULL); 2681 } 2682 2683 int netdev_refcnt_read(const struct net_device *dev); 2684 void free_netdev(struct net_device *dev); 2685 void netdev_freemem(struct net_device *dev); 2686 void synchronize_net(void); 2687 int init_dummy_netdev(struct net_device *dev); 2688 2689 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2690 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2691 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2692 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2693 int netdev_get_name(struct net *net, char *name, int ifindex); 2694 int dev_restart(struct net_device *dev); 2695 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 2696 2697 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2698 { 2699 return NAPI_GRO_CB(skb)->data_offset; 2700 } 2701 2702 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2703 { 2704 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2705 } 2706 2707 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2708 { 2709 NAPI_GRO_CB(skb)->data_offset += len; 2710 } 2711 2712 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2713 unsigned int offset) 2714 { 2715 return NAPI_GRO_CB(skb)->frag0 + offset; 2716 } 2717 2718 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2719 { 2720 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2721 } 2722 2723 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2724 { 2725 NAPI_GRO_CB(skb)->frag0 = NULL; 2726 NAPI_GRO_CB(skb)->frag0_len = 0; 2727 } 2728 2729 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2730 unsigned int offset) 2731 { 2732 if (!pskb_may_pull(skb, hlen)) 2733 return NULL; 2734 2735 skb_gro_frag0_invalidate(skb); 2736 return skb->data + offset; 2737 } 2738 2739 static inline void *skb_gro_network_header(struct sk_buff *skb) 2740 { 2741 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2742 skb_network_offset(skb); 2743 } 2744 2745 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2746 const void *start, unsigned int len) 2747 { 2748 if (NAPI_GRO_CB(skb)->csum_valid) 2749 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2750 csum_partial(start, len, 0)); 2751 } 2752 2753 /* GRO checksum functions. These are logical equivalents of the normal 2754 * checksum functions (in skbuff.h) except that they operate on the GRO 2755 * offsets and fields in sk_buff. 2756 */ 2757 2758 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2759 2760 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2761 { 2762 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2763 } 2764 2765 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2766 bool zero_okay, 2767 __sum16 check) 2768 { 2769 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2770 skb_checksum_start_offset(skb) < 2771 skb_gro_offset(skb)) && 2772 !skb_at_gro_remcsum_start(skb) && 2773 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2774 (!zero_okay || check)); 2775 } 2776 2777 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2778 __wsum psum) 2779 { 2780 if (NAPI_GRO_CB(skb)->csum_valid && 2781 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2782 return 0; 2783 2784 NAPI_GRO_CB(skb)->csum = psum; 2785 2786 return __skb_gro_checksum_complete(skb); 2787 } 2788 2789 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2790 { 2791 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2792 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2793 NAPI_GRO_CB(skb)->csum_cnt--; 2794 } else { 2795 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2796 * verified a new top level checksum or an encapsulated one 2797 * during GRO. This saves work if we fallback to normal path. 2798 */ 2799 __skb_incr_checksum_unnecessary(skb); 2800 } 2801 } 2802 2803 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2804 compute_pseudo) \ 2805 ({ \ 2806 __sum16 __ret = 0; \ 2807 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2808 __ret = __skb_gro_checksum_validate_complete(skb, \ 2809 compute_pseudo(skb, proto)); \ 2810 if (!__ret) \ 2811 skb_gro_incr_csum_unnecessary(skb); \ 2812 __ret; \ 2813 }) 2814 2815 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2816 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2817 2818 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2819 compute_pseudo) \ 2820 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2821 2822 #define skb_gro_checksum_simple_validate(skb) \ 2823 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2824 2825 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2826 { 2827 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2828 !NAPI_GRO_CB(skb)->csum_valid); 2829 } 2830 2831 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2832 __wsum pseudo) 2833 { 2834 NAPI_GRO_CB(skb)->csum = ~pseudo; 2835 NAPI_GRO_CB(skb)->csum_valid = 1; 2836 } 2837 2838 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \ 2839 do { \ 2840 if (__skb_gro_checksum_convert_check(skb)) \ 2841 __skb_gro_checksum_convert(skb, \ 2842 compute_pseudo(skb, proto)); \ 2843 } while (0) 2844 2845 struct gro_remcsum { 2846 int offset; 2847 __wsum delta; 2848 }; 2849 2850 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2851 { 2852 grc->offset = 0; 2853 grc->delta = 0; 2854 } 2855 2856 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2857 unsigned int off, size_t hdrlen, 2858 int start, int offset, 2859 struct gro_remcsum *grc, 2860 bool nopartial) 2861 { 2862 __wsum delta; 2863 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2864 2865 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2866 2867 if (!nopartial) { 2868 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2869 return ptr; 2870 } 2871 2872 ptr = skb_gro_header_fast(skb, off); 2873 if (skb_gro_header_hard(skb, off + plen)) { 2874 ptr = skb_gro_header_slow(skb, off + plen, off); 2875 if (!ptr) 2876 return NULL; 2877 } 2878 2879 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2880 start, offset); 2881 2882 /* Adjust skb->csum since we changed the packet */ 2883 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2884 2885 grc->offset = off + hdrlen + offset; 2886 grc->delta = delta; 2887 2888 return ptr; 2889 } 2890 2891 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2892 struct gro_remcsum *grc) 2893 { 2894 void *ptr; 2895 size_t plen = grc->offset + sizeof(u16); 2896 2897 if (!grc->delta) 2898 return; 2899 2900 ptr = skb_gro_header_fast(skb, grc->offset); 2901 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 2902 ptr = skb_gro_header_slow(skb, plen, grc->offset); 2903 if (!ptr) 2904 return; 2905 } 2906 2907 remcsum_unadjust((__sum16 *)ptr, grc->delta); 2908 } 2909 2910 #ifdef CONFIG_XFRM_OFFLOAD 2911 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2912 { 2913 if (PTR_ERR(pp) != -EINPROGRESS) 2914 NAPI_GRO_CB(skb)->flush |= flush; 2915 } 2916 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 2917 struct sk_buff *pp, 2918 int flush, 2919 struct gro_remcsum *grc) 2920 { 2921 if (PTR_ERR(pp) != -EINPROGRESS) { 2922 NAPI_GRO_CB(skb)->flush |= flush; 2923 skb_gro_remcsum_cleanup(skb, grc); 2924 skb->remcsum_offload = 0; 2925 } 2926 } 2927 #else 2928 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2929 { 2930 NAPI_GRO_CB(skb)->flush |= flush; 2931 } 2932 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 2933 struct sk_buff *pp, 2934 int flush, 2935 struct gro_remcsum *grc) 2936 { 2937 NAPI_GRO_CB(skb)->flush |= flush; 2938 skb_gro_remcsum_cleanup(skb, grc); 2939 skb->remcsum_offload = 0; 2940 } 2941 #endif 2942 2943 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2944 unsigned short type, 2945 const void *daddr, const void *saddr, 2946 unsigned int len) 2947 { 2948 if (!dev->header_ops || !dev->header_ops->create) 2949 return 0; 2950 2951 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2952 } 2953 2954 static inline int dev_parse_header(const struct sk_buff *skb, 2955 unsigned char *haddr) 2956 { 2957 const struct net_device *dev = skb->dev; 2958 2959 if (!dev->header_ops || !dev->header_ops->parse) 2960 return 0; 2961 return dev->header_ops->parse(skb, haddr); 2962 } 2963 2964 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 2965 { 2966 const struct net_device *dev = skb->dev; 2967 2968 if (!dev->header_ops || !dev->header_ops->parse_protocol) 2969 return 0; 2970 return dev->header_ops->parse_protocol(skb); 2971 } 2972 2973 /* ll_header must have at least hard_header_len allocated */ 2974 static inline bool dev_validate_header(const struct net_device *dev, 2975 char *ll_header, int len) 2976 { 2977 if (likely(len >= dev->hard_header_len)) 2978 return true; 2979 if (len < dev->min_header_len) 2980 return false; 2981 2982 if (capable(CAP_SYS_RAWIO)) { 2983 memset(ll_header + len, 0, dev->hard_header_len - len); 2984 return true; 2985 } 2986 2987 if (dev->header_ops && dev->header_ops->validate) 2988 return dev->header_ops->validate(ll_header, len); 2989 2990 return false; 2991 } 2992 2993 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, 2994 int len, int size); 2995 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2996 static inline int unregister_gifconf(unsigned int family) 2997 { 2998 return register_gifconf(family, NULL); 2999 } 3000 3001 #ifdef CONFIG_NET_FLOW_LIMIT 3002 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 3003 struct sd_flow_limit { 3004 u64 count; 3005 unsigned int num_buckets; 3006 unsigned int history_head; 3007 u16 history[FLOW_LIMIT_HISTORY]; 3008 u8 buckets[]; 3009 }; 3010 3011 extern int netdev_flow_limit_table_len; 3012 #endif /* CONFIG_NET_FLOW_LIMIT */ 3013 3014 /* 3015 * Incoming packets are placed on per-CPU queues 3016 */ 3017 struct softnet_data { 3018 struct list_head poll_list; 3019 struct sk_buff_head process_queue; 3020 3021 /* stats */ 3022 unsigned int processed; 3023 unsigned int time_squeeze; 3024 unsigned int received_rps; 3025 #ifdef CONFIG_RPS 3026 struct softnet_data *rps_ipi_list; 3027 #endif 3028 #ifdef CONFIG_NET_FLOW_LIMIT 3029 struct sd_flow_limit __rcu *flow_limit; 3030 #endif 3031 struct Qdisc *output_queue; 3032 struct Qdisc **output_queue_tailp; 3033 struct sk_buff *completion_queue; 3034 #ifdef CONFIG_XFRM_OFFLOAD 3035 struct sk_buff_head xfrm_backlog; 3036 #endif 3037 /* written and read only by owning cpu: */ 3038 struct { 3039 u16 recursion; 3040 u8 more; 3041 } xmit; 3042 #ifdef CONFIG_RPS 3043 /* input_queue_head should be written by cpu owning this struct, 3044 * and only read by other cpus. Worth using a cache line. 3045 */ 3046 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3047 3048 /* Elements below can be accessed between CPUs for RPS/RFS */ 3049 call_single_data_t csd ____cacheline_aligned_in_smp; 3050 struct softnet_data *rps_ipi_next; 3051 unsigned int cpu; 3052 unsigned int input_queue_tail; 3053 #endif 3054 unsigned int dropped; 3055 struct sk_buff_head input_pkt_queue; 3056 struct napi_struct backlog; 3057 3058 }; 3059 3060 static inline void input_queue_head_incr(struct softnet_data *sd) 3061 { 3062 #ifdef CONFIG_RPS 3063 sd->input_queue_head++; 3064 #endif 3065 } 3066 3067 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3068 unsigned int *qtail) 3069 { 3070 #ifdef CONFIG_RPS 3071 *qtail = ++sd->input_queue_tail; 3072 #endif 3073 } 3074 3075 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3076 3077 static inline int dev_recursion_level(void) 3078 { 3079 return this_cpu_read(softnet_data.xmit.recursion); 3080 } 3081 3082 #define XMIT_RECURSION_LIMIT 10 3083 static inline bool dev_xmit_recursion(void) 3084 { 3085 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3086 XMIT_RECURSION_LIMIT); 3087 } 3088 3089 static inline void dev_xmit_recursion_inc(void) 3090 { 3091 __this_cpu_inc(softnet_data.xmit.recursion); 3092 } 3093 3094 static inline void dev_xmit_recursion_dec(void) 3095 { 3096 __this_cpu_dec(softnet_data.xmit.recursion); 3097 } 3098 3099 void __netif_schedule(struct Qdisc *q); 3100 void netif_schedule_queue(struct netdev_queue *txq); 3101 3102 static inline void netif_tx_schedule_all(struct net_device *dev) 3103 { 3104 unsigned int i; 3105 3106 for (i = 0; i < dev->num_tx_queues; i++) 3107 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3108 } 3109 3110 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3111 { 3112 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3113 } 3114 3115 /** 3116 * netif_start_queue - allow transmit 3117 * @dev: network device 3118 * 3119 * Allow upper layers to call the device hard_start_xmit routine. 3120 */ 3121 static inline void netif_start_queue(struct net_device *dev) 3122 { 3123 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3124 } 3125 3126 static inline void netif_tx_start_all_queues(struct net_device *dev) 3127 { 3128 unsigned int i; 3129 3130 for (i = 0; i < dev->num_tx_queues; i++) { 3131 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3132 netif_tx_start_queue(txq); 3133 } 3134 } 3135 3136 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3137 3138 /** 3139 * netif_wake_queue - restart transmit 3140 * @dev: network device 3141 * 3142 * Allow upper layers to call the device hard_start_xmit routine. 3143 * Used for flow control when transmit resources are available. 3144 */ 3145 static inline void netif_wake_queue(struct net_device *dev) 3146 { 3147 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3148 } 3149 3150 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3151 { 3152 unsigned int i; 3153 3154 for (i = 0; i < dev->num_tx_queues; i++) { 3155 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3156 netif_tx_wake_queue(txq); 3157 } 3158 } 3159 3160 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3161 { 3162 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3163 } 3164 3165 /** 3166 * netif_stop_queue - stop transmitted packets 3167 * @dev: network device 3168 * 3169 * Stop upper layers calling the device hard_start_xmit routine. 3170 * Used for flow control when transmit resources are unavailable. 3171 */ 3172 static inline void netif_stop_queue(struct net_device *dev) 3173 { 3174 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3175 } 3176 3177 void netif_tx_stop_all_queues(struct net_device *dev); 3178 void netdev_update_lockdep_key(struct net_device *dev); 3179 3180 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3181 { 3182 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3183 } 3184 3185 /** 3186 * netif_queue_stopped - test if transmit queue is flowblocked 3187 * @dev: network device 3188 * 3189 * Test if transmit queue on device is currently unable to send. 3190 */ 3191 static inline bool netif_queue_stopped(const struct net_device *dev) 3192 { 3193 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3194 } 3195 3196 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3197 { 3198 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3199 } 3200 3201 static inline bool 3202 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3203 { 3204 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3205 } 3206 3207 static inline bool 3208 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3209 { 3210 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3211 } 3212 3213 /** 3214 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3215 * @dev_queue: pointer to transmit queue 3216 * 3217 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3218 * to give appropriate hint to the CPU. 3219 */ 3220 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3221 { 3222 #ifdef CONFIG_BQL 3223 prefetchw(&dev_queue->dql.num_queued); 3224 #endif 3225 } 3226 3227 /** 3228 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3229 * @dev_queue: pointer to transmit queue 3230 * 3231 * BQL enabled drivers might use this helper in their TX completion path, 3232 * to give appropriate hint to the CPU. 3233 */ 3234 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3235 { 3236 #ifdef CONFIG_BQL 3237 prefetchw(&dev_queue->dql.limit); 3238 #endif 3239 } 3240 3241 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3242 unsigned int bytes) 3243 { 3244 #ifdef CONFIG_BQL 3245 dql_queued(&dev_queue->dql, bytes); 3246 3247 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3248 return; 3249 3250 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3251 3252 /* 3253 * The XOFF flag must be set before checking the dql_avail below, 3254 * because in netdev_tx_completed_queue we update the dql_completed 3255 * before checking the XOFF flag. 3256 */ 3257 smp_mb(); 3258 3259 /* check again in case another CPU has just made room avail */ 3260 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3261 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3262 #endif 3263 } 3264 3265 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3266 * that they should not test BQL status themselves. 3267 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3268 * skb of a batch. 3269 * Returns true if the doorbell must be used to kick the NIC. 3270 */ 3271 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3272 unsigned int bytes, 3273 bool xmit_more) 3274 { 3275 if (xmit_more) { 3276 #ifdef CONFIG_BQL 3277 dql_queued(&dev_queue->dql, bytes); 3278 #endif 3279 return netif_tx_queue_stopped(dev_queue); 3280 } 3281 netdev_tx_sent_queue(dev_queue, bytes); 3282 return true; 3283 } 3284 3285 /** 3286 * netdev_sent_queue - report the number of bytes queued to hardware 3287 * @dev: network device 3288 * @bytes: number of bytes queued to the hardware device queue 3289 * 3290 * Report the number of bytes queued for sending/completion to the network 3291 * device hardware queue. @bytes should be a good approximation and should 3292 * exactly match netdev_completed_queue() @bytes 3293 */ 3294 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3295 { 3296 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3297 } 3298 3299 static inline bool __netdev_sent_queue(struct net_device *dev, 3300 unsigned int bytes, 3301 bool xmit_more) 3302 { 3303 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3304 xmit_more); 3305 } 3306 3307 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3308 unsigned int pkts, unsigned int bytes) 3309 { 3310 #ifdef CONFIG_BQL 3311 if (unlikely(!bytes)) 3312 return; 3313 3314 dql_completed(&dev_queue->dql, bytes); 3315 3316 /* 3317 * Without the memory barrier there is a small possiblity that 3318 * netdev_tx_sent_queue will miss the update and cause the queue to 3319 * be stopped forever 3320 */ 3321 smp_mb(); 3322 3323 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3324 return; 3325 3326 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3327 netif_schedule_queue(dev_queue); 3328 #endif 3329 } 3330 3331 /** 3332 * netdev_completed_queue - report bytes and packets completed by device 3333 * @dev: network device 3334 * @pkts: actual number of packets sent over the medium 3335 * @bytes: actual number of bytes sent over the medium 3336 * 3337 * Report the number of bytes and packets transmitted by the network device 3338 * hardware queue over the physical medium, @bytes must exactly match the 3339 * @bytes amount passed to netdev_sent_queue() 3340 */ 3341 static inline void netdev_completed_queue(struct net_device *dev, 3342 unsigned int pkts, unsigned int bytes) 3343 { 3344 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3345 } 3346 3347 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3348 { 3349 #ifdef CONFIG_BQL 3350 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3351 dql_reset(&q->dql); 3352 #endif 3353 } 3354 3355 /** 3356 * netdev_reset_queue - reset the packets and bytes count of a network device 3357 * @dev_queue: network device 3358 * 3359 * Reset the bytes and packet count of a network device and clear the 3360 * software flow control OFF bit for this network device 3361 */ 3362 static inline void netdev_reset_queue(struct net_device *dev_queue) 3363 { 3364 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3365 } 3366 3367 /** 3368 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3369 * @dev: network device 3370 * @queue_index: given tx queue index 3371 * 3372 * Returns 0 if given tx queue index >= number of device tx queues, 3373 * otherwise returns the originally passed tx queue index. 3374 */ 3375 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3376 { 3377 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3378 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3379 dev->name, queue_index, 3380 dev->real_num_tx_queues); 3381 return 0; 3382 } 3383 3384 return queue_index; 3385 } 3386 3387 /** 3388 * netif_running - test if up 3389 * @dev: network device 3390 * 3391 * Test if the device has been brought up. 3392 */ 3393 static inline bool netif_running(const struct net_device *dev) 3394 { 3395 return test_bit(__LINK_STATE_START, &dev->state); 3396 } 3397 3398 /* 3399 * Routines to manage the subqueues on a device. We only need start, 3400 * stop, and a check if it's stopped. All other device management is 3401 * done at the overall netdevice level. 3402 * Also test the device if we're multiqueue. 3403 */ 3404 3405 /** 3406 * netif_start_subqueue - allow sending packets on subqueue 3407 * @dev: network device 3408 * @queue_index: sub queue index 3409 * 3410 * Start individual transmit queue of a device with multiple transmit queues. 3411 */ 3412 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3413 { 3414 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3415 3416 netif_tx_start_queue(txq); 3417 } 3418 3419 /** 3420 * netif_stop_subqueue - stop sending packets on subqueue 3421 * @dev: network device 3422 * @queue_index: sub queue index 3423 * 3424 * Stop individual transmit queue of a device with multiple transmit queues. 3425 */ 3426 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3427 { 3428 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3429 netif_tx_stop_queue(txq); 3430 } 3431 3432 /** 3433 * netif_subqueue_stopped - test status of subqueue 3434 * @dev: network device 3435 * @queue_index: sub queue index 3436 * 3437 * Check individual transmit queue of a device with multiple transmit queues. 3438 */ 3439 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3440 u16 queue_index) 3441 { 3442 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3443 3444 return netif_tx_queue_stopped(txq); 3445 } 3446 3447 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3448 struct sk_buff *skb) 3449 { 3450 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3451 } 3452 3453 /** 3454 * netif_wake_subqueue - allow sending packets on subqueue 3455 * @dev: network device 3456 * @queue_index: sub queue index 3457 * 3458 * Resume individual transmit queue of a device with multiple transmit queues. 3459 */ 3460 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3461 { 3462 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3463 3464 netif_tx_wake_queue(txq); 3465 } 3466 3467 #ifdef CONFIG_XPS 3468 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3469 u16 index); 3470 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3471 u16 index, bool is_rxqs_map); 3472 3473 /** 3474 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3475 * @j: CPU/Rx queue index 3476 * @mask: bitmask of all cpus/rx queues 3477 * @nr_bits: number of bits in the bitmask 3478 * 3479 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3480 */ 3481 static inline bool netif_attr_test_mask(unsigned long j, 3482 const unsigned long *mask, 3483 unsigned int nr_bits) 3484 { 3485 cpu_max_bits_warn(j, nr_bits); 3486 return test_bit(j, mask); 3487 } 3488 3489 /** 3490 * netif_attr_test_online - Test for online CPU/Rx queue 3491 * @j: CPU/Rx queue index 3492 * @online_mask: bitmask for CPUs/Rx queues that are online 3493 * @nr_bits: number of bits in the bitmask 3494 * 3495 * Returns true if a CPU/Rx queue is online. 3496 */ 3497 static inline bool netif_attr_test_online(unsigned long j, 3498 const unsigned long *online_mask, 3499 unsigned int nr_bits) 3500 { 3501 cpu_max_bits_warn(j, nr_bits); 3502 3503 if (online_mask) 3504 return test_bit(j, online_mask); 3505 3506 return (j < nr_bits); 3507 } 3508 3509 /** 3510 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3511 * @n: CPU/Rx queue index 3512 * @srcp: the cpumask/Rx queue mask pointer 3513 * @nr_bits: number of bits in the bitmask 3514 * 3515 * Returns >= nr_bits if no further CPUs/Rx queues set. 3516 */ 3517 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3518 unsigned int nr_bits) 3519 { 3520 /* -1 is a legal arg here. */ 3521 if (n != -1) 3522 cpu_max_bits_warn(n, nr_bits); 3523 3524 if (srcp) 3525 return find_next_bit(srcp, nr_bits, n + 1); 3526 3527 return n + 1; 3528 } 3529 3530 /** 3531 * netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p 3532 * @n: CPU/Rx queue index 3533 * @src1p: the first CPUs/Rx queues mask pointer 3534 * @src2p: the second CPUs/Rx queues mask pointer 3535 * @nr_bits: number of bits in the bitmask 3536 * 3537 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3538 */ 3539 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3540 const unsigned long *src2p, 3541 unsigned int nr_bits) 3542 { 3543 /* -1 is a legal arg here. */ 3544 if (n != -1) 3545 cpu_max_bits_warn(n, nr_bits); 3546 3547 if (src1p && src2p) 3548 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3549 else if (src1p) 3550 return find_next_bit(src1p, nr_bits, n + 1); 3551 else if (src2p) 3552 return find_next_bit(src2p, nr_bits, n + 1); 3553 3554 return n + 1; 3555 } 3556 #else 3557 static inline int netif_set_xps_queue(struct net_device *dev, 3558 const struct cpumask *mask, 3559 u16 index) 3560 { 3561 return 0; 3562 } 3563 3564 static inline int __netif_set_xps_queue(struct net_device *dev, 3565 const unsigned long *mask, 3566 u16 index, bool is_rxqs_map) 3567 { 3568 return 0; 3569 } 3570 #endif 3571 3572 /** 3573 * netif_is_multiqueue - test if device has multiple transmit queues 3574 * @dev: network device 3575 * 3576 * Check if device has multiple transmit queues 3577 */ 3578 static inline bool netif_is_multiqueue(const struct net_device *dev) 3579 { 3580 return dev->num_tx_queues > 1; 3581 } 3582 3583 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3584 3585 #ifdef CONFIG_SYSFS 3586 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3587 #else 3588 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3589 unsigned int rxqs) 3590 { 3591 dev->real_num_rx_queues = rxqs; 3592 return 0; 3593 } 3594 #endif 3595 3596 static inline struct netdev_rx_queue * 3597 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3598 { 3599 return dev->_rx + rxq; 3600 } 3601 3602 #ifdef CONFIG_SYSFS 3603 static inline unsigned int get_netdev_rx_queue_index( 3604 struct netdev_rx_queue *queue) 3605 { 3606 struct net_device *dev = queue->dev; 3607 int index = queue - dev->_rx; 3608 3609 BUG_ON(index >= dev->num_rx_queues); 3610 return index; 3611 } 3612 #endif 3613 3614 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3615 int netif_get_num_default_rss_queues(void); 3616 3617 enum skb_free_reason { 3618 SKB_REASON_CONSUMED, 3619 SKB_REASON_DROPPED, 3620 }; 3621 3622 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3623 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3624 3625 /* 3626 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3627 * interrupt context or with hardware interrupts being disabled. 3628 * (in_irq() || irqs_disabled()) 3629 * 3630 * We provide four helpers that can be used in following contexts : 3631 * 3632 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3633 * replacing kfree_skb(skb) 3634 * 3635 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3636 * Typically used in place of consume_skb(skb) in TX completion path 3637 * 3638 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3639 * replacing kfree_skb(skb) 3640 * 3641 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3642 * and consumed a packet. Used in place of consume_skb(skb) 3643 */ 3644 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3645 { 3646 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3647 } 3648 3649 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3650 { 3651 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3652 } 3653 3654 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3655 { 3656 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3657 } 3658 3659 static inline void dev_consume_skb_any(struct sk_buff *skb) 3660 { 3661 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3662 } 3663 3664 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3665 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3666 int netif_rx(struct sk_buff *skb); 3667 int netif_rx_ni(struct sk_buff *skb); 3668 int netif_receive_skb(struct sk_buff *skb); 3669 int netif_receive_skb_core(struct sk_buff *skb); 3670 void netif_receive_skb_list(struct list_head *head); 3671 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3672 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3673 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3674 gro_result_t napi_gro_frags(struct napi_struct *napi); 3675 struct packet_offload *gro_find_receive_by_type(__be16 type); 3676 struct packet_offload *gro_find_complete_by_type(__be16 type); 3677 3678 static inline void napi_free_frags(struct napi_struct *napi) 3679 { 3680 kfree_skb(napi->skb); 3681 napi->skb = NULL; 3682 } 3683 3684 bool netdev_is_rx_handler_busy(struct net_device *dev); 3685 int netdev_rx_handler_register(struct net_device *dev, 3686 rx_handler_func_t *rx_handler, 3687 void *rx_handler_data); 3688 void netdev_rx_handler_unregister(struct net_device *dev); 3689 3690 bool dev_valid_name(const char *name); 3691 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3692 bool *need_copyout); 3693 int dev_ifconf(struct net *net, struct ifconf *, int); 3694 int dev_ethtool(struct net *net, struct ifreq *); 3695 unsigned int dev_get_flags(const struct net_device *); 3696 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3697 struct netlink_ext_ack *extack); 3698 int dev_change_flags(struct net_device *dev, unsigned int flags, 3699 struct netlink_ext_ack *extack); 3700 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3701 unsigned int gchanges); 3702 int dev_change_name(struct net_device *, const char *); 3703 int dev_set_alias(struct net_device *, const char *, size_t); 3704 int dev_get_alias(const struct net_device *, char *, size_t); 3705 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3706 int __dev_set_mtu(struct net_device *, int); 3707 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3708 struct netlink_ext_ack *extack); 3709 int dev_set_mtu(struct net_device *, int); 3710 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3711 void dev_set_group(struct net_device *, int); 3712 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3713 struct netlink_ext_ack *extack); 3714 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3715 struct netlink_ext_ack *extack); 3716 int dev_change_carrier(struct net_device *, bool new_carrier); 3717 int dev_get_phys_port_id(struct net_device *dev, 3718 struct netdev_phys_item_id *ppid); 3719 int dev_get_phys_port_name(struct net_device *dev, 3720 char *name, size_t len); 3721 int dev_get_port_parent_id(struct net_device *dev, 3722 struct netdev_phys_item_id *ppid, bool recurse); 3723 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3724 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3725 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down); 3726 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3727 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3728 struct netdev_queue *txq, int *ret); 3729 3730 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3731 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3732 int fd, u32 flags); 3733 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op, 3734 enum bpf_netdev_command cmd); 3735 int xdp_umem_query(struct net_device *dev, u16 queue_id); 3736 3737 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3738 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3739 bool is_skb_forwardable(const struct net_device *dev, 3740 const struct sk_buff *skb); 3741 3742 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3743 struct sk_buff *skb) 3744 { 3745 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3746 unlikely(!is_skb_forwardable(dev, skb))) { 3747 atomic_long_inc(&dev->rx_dropped); 3748 kfree_skb(skb); 3749 return NET_RX_DROP; 3750 } 3751 3752 skb_scrub_packet(skb, true); 3753 skb->priority = 0; 3754 return 0; 3755 } 3756 3757 bool dev_nit_active(struct net_device *dev); 3758 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3759 3760 extern int netdev_budget; 3761 extern unsigned int netdev_budget_usecs; 3762 3763 /* Called by rtnetlink.c:rtnl_unlock() */ 3764 void netdev_run_todo(void); 3765 3766 /** 3767 * dev_put - release reference to device 3768 * @dev: network device 3769 * 3770 * Release reference to device to allow it to be freed. 3771 */ 3772 static inline void dev_put(struct net_device *dev) 3773 { 3774 this_cpu_dec(*dev->pcpu_refcnt); 3775 } 3776 3777 /** 3778 * dev_hold - get reference to device 3779 * @dev: network device 3780 * 3781 * Hold reference to device to keep it from being freed. 3782 */ 3783 static inline void dev_hold(struct net_device *dev) 3784 { 3785 this_cpu_inc(*dev->pcpu_refcnt); 3786 } 3787 3788 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3789 * and _off may be called from IRQ context, but it is caller 3790 * who is responsible for serialization of these calls. 3791 * 3792 * The name carrier is inappropriate, these functions should really be 3793 * called netif_lowerlayer_*() because they represent the state of any 3794 * kind of lower layer not just hardware media. 3795 */ 3796 3797 void linkwatch_init_dev(struct net_device *dev); 3798 void linkwatch_fire_event(struct net_device *dev); 3799 void linkwatch_forget_dev(struct net_device *dev); 3800 3801 /** 3802 * netif_carrier_ok - test if carrier present 3803 * @dev: network device 3804 * 3805 * Check if carrier is present on device 3806 */ 3807 static inline bool netif_carrier_ok(const struct net_device *dev) 3808 { 3809 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3810 } 3811 3812 unsigned long dev_trans_start(struct net_device *dev); 3813 3814 void __netdev_watchdog_up(struct net_device *dev); 3815 3816 void netif_carrier_on(struct net_device *dev); 3817 3818 void netif_carrier_off(struct net_device *dev); 3819 3820 /** 3821 * netif_dormant_on - mark device as dormant. 3822 * @dev: network device 3823 * 3824 * Mark device as dormant (as per RFC2863). 3825 * 3826 * The dormant state indicates that the relevant interface is not 3827 * actually in a condition to pass packets (i.e., it is not 'up') but is 3828 * in a "pending" state, waiting for some external event. For "on- 3829 * demand" interfaces, this new state identifies the situation where the 3830 * interface is waiting for events to place it in the up state. 3831 */ 3832 static inline void netif_dormant_on(struct net_device *dev) 3833 { 3834 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3835 linkwatch_fire_event(dev); 3836 } 3837 3838 /** 3839 * netif_dormant_off - set device as not dormant. 3840 * @dev: network device 3841 * 3842 * Device is not in dormant state. 3843 */ 3844 static inline void netif_dormant_off(struct net_device *dev) 3845 { 3846 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3847 linkwatch_fire_event(dev); 3848 } 3849 3850 /** 3851 * netif_dormant - test if device is dormant 3852 * @dev: network device 3853 * 3854 * Check if device is dormant. 3855 */ 3856 static inline bool netif_dormant(const struct net_device *dev) 3857 { 3858 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3859 } 3860 3861 3862 /** 3863 * netif_oper_up - test if device is operational 3864 * @dev: network device 3865 * 3866 * Check if carrier is operational 3867 */ 3868 static inline bool netif_oper_up(const struct net_device *dev) 3869 { 3870 return (dev->operstate == IF_OPER_UP || 3871 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3872 } 3873 3874 /** 3875 * netif_device_present - is device available or removed 3876 * @dev: network device 3877 * 3878 * Check if device has not been removed from system. 3879 */ 3880 static inline bool netif_device_present(struct net_device *dev) 3881 { 3882 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3883 } 3884 3885 void netif_device_detach(struct net_device *dev); 3886 3887 void netif_device_attach(struct net_device *dev); 3888 3889 /* 3890 * Network interface message level settings 3891 */ 3892 3893 enum { 3894 NETIF_MSG_DRV = 0x0001, 3895 NETIF_MSG_PROBE = 0x0002, 3896 NETIF_MSG_LINK = 0x0004, 3897 NETIF_MSG_TIMER = 0x0008, 3898 NETIF_MSG_IFDOWN = 0x0010, 3899 NETIF_MSG_IFUP = 0x0020, 3900 NETIF_MSG_RX_ERR = 0x0040, 3901 NETIF_MSG_TX_ERR = 0x0080, 3902 NETIF_MSG_TX_QUEUED = 0x0100, 3903 NETIF_MSG_INTR = 0x0200, 3904 NETIF_MSG_TX_DONE = 0x0400, 3905 NETIF_MSG_RX_STATUS = 0x0800, 3906 NETIF_MSG_PKTDATA = 0x1000, 3907 NETIF_MSG_HW = 0x2000, 3908 NETIF_MSG_WOL = 0x4000, 3909 }; 3910 3911 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3912 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3913 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3914 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3915 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3916 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3917 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3918 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3919 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3920 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3921 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3922 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3923 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3924 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3925 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3926 3927 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3928 { 3929 /* use default */ 3930 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3931 return default_msg_enable_bits; 3932 if (debug_value == 0) /* no output */ 3933 return 0; 3934 /* set low N bits */ 3935 return (1U << debug_value) - 1; 3936 } 3937 3938 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 3939 { 3940 spin_lock(&txq->_xmit_lock); 3941 txq->xmit_lock_owner = cpu; 3942 } 3943 3944 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 3945 { 3946 __acquire(&txq->_xmit_lock); 3947 return true; 3948 } 3949 3950 static inline void __netif_tx_release(struct netdev_queue *txq) 3951 { 3952 __release(&txq->_xmit_lock); 3953 } 3954 3955 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 3956 { 3957 spin_lock_bh(&txq->_xmit_lock); 3958 txq->xmit_lock_owner = smp_processor_id(); 3959 } 3960 3961 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 3962 { 3963 bool ok = spin_trylock(&txq->_xmit_lock); 3964 if (likely(ok)) 3965 txq->xmit_lock_owner = smp_processor_id(); 3966 return ok; 3967 } 3968 3969 static inline void __netif_tx_unlock(struct netdev_queue *txq) 3970 { 3971 txq->xmit_lock_owner = -1; 3972 spin_unlock(&txq->_xmit_lock); 3973 } 3974 3975 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 3976 { 3977 txq->xmit_lock_owner = -1; 3978 spin_unlock_bh(&txq->_xmit_lock); 3979 } 3980 3981 static inline void txq_trans_update(struct netdev_queue *txq) 3982 { 3983 if (txq->xmit_lock_owner != -1) 3984 txq->trans_start = jiffies; 3985 } 3986 3987 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 3988 static inline void netif_trans_update(struct net_device *dev) 3989 { 3990 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 3991 3992 if (txq->trans_start != jiffies) 3993 txq->trans_start = jiffies; 3994 } 3995 3996 /** 3997 * netif_tx_lock - grab network device transmit lock 3998 * @dev: network device 3999 * 4000 * Get network device transmit lock 4001 */ 4002 static inline void netif_tx_lock(struct net_device *dev) 4003 { 4004 unsigned int i; 4005 int cpu; 4006 4007 spin_lock(&dev->tx_global_lock); 4008 cpu = smp_processor_id(); 4009 for (i = 0; i < dev->num_tx_queues; i++) { 4010 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4011 4012 /* We are the only thread of execution doing a 4013 * freeze, but we have to grab the _xmit_lock in 4014 * order to synchronize with threads which are in 4015 * the ->hard_start_xmit() handler and already 4016 * checked the frozen bit. 4017 */ 4018 __netif_tx_lock(txq, cpu); 4019 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 4020 __netif_tx_unlock(txq); 4021 } 4022 } 4023 4024 static inline void netif_tx_lock_bh(struct net_device *dev) 4025 { 4026 local_bh_disable(); 4027 netif_tx_lock(dev); 4028 } 4029 4030 static inline void netif_tx_unlock(struct net_device *dev) 4031 { 4032 unsigned int i; 4033 4034 for (i = 0; i < dev->num_tx_queues; i++) { 4035 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4036 4037 /* No need to grab the _xmit_lock here. If the 4038 * queue is not stopped for another reason, we 4039 * force a schedule. 4040 */ 4041 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 4042 netif_schedule_queue(txq); 4043 } 4044 spin_unlock(&dev->tx_global_lock); 4045 } 4046 4047 static inline void netif_tx_unlock_bh(struct net_device *dev) 4048 { 4049 netif_tx_unlock(dev); 4050 local_bh_enable(); 4051 } 4052 4053 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4054 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4055 __netif_tx_lock(txq, cpu); \ 4056 } else { \ 4057 __netif_tx_acquire(txq); \ 4058 } \ 4059 } 4060 4061 #define HARD_TX_TRYLOCK(dev, txq) \ 4062 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4063 __netif_tx_trylock(txq) : \ 4064 __netif_tx_acquire(txq)) 4065 4066 #define HARD_TX_UNLOCK(dev, txq) { \ 4067 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4068 __netif_tx_unlock(txq); \ 4069 } else { \ 4070 __netif_tx_release(txq); \ 4071 } \ 4072 } 4073 4074 static inline void netif_tx_disable(struct net_device *dev) 4075 { 4076 unsigned int i; 4077 int cpu; 4078 4079 local_bh_disable(); 4080 cpu = smp_processor_id(); 4081 for (i = 0; i < dev->num_tx_queues; i++) { 4082 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4083 4084 __netif_tx_lock(txq, cpu); 4085 netif_tx_stop_queue(txq); 4086 __netif_tx_unlock(txq); 4087 } 4088 local_bh_enable(); 4089 } 4090 4091 static inline void netif_addr_lock(struct net_device *dev) 4092 { 4093 spin_lock(&dev->addr_list_lock); 4094 } 4095 4096 static inline void netif_addr_lock_bh(struct net_device *dev) 4097 { 4098 spin_lock_bh(&dev->addr_list_lock); 4099 } 4100 4101 static inline void netif_addr_unlock(struct net_device *dev) 4102 { 4103 spin_unlock(&dev->addr_list_lock); 4104 } 4105 4106 static inline void netif_addr_unlock_bh(struct net_device *dev) 4107 { 4108 spin_unlock_bh(&dev->addr_list_lock); 4109 } 4110 4111 /* 4112 * dev_addrs walker. Should be used only for read access. Call with 4113 * rcu_read_lock held. 4114 */ 4115 #define for_each_dev_addr(dev, ha) \ 4116 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4117 4118 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4119 4120 void ether_setup(struct net_device *dev); 4121 4122 /* Support for loadable net-drivers */ 4123 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4124 unsigned char name_assign_type, 4125 void (*setup)(struct net_device *), 4126 unsigned int txqs, unsigned int rxqs); 4127 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4128 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4129 4130 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4131 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4132 count) 4133 4134 int register_netdev(struct net_device *dev); 4135 void unregister_netdev(struct net_device *dev); 4136 4137 /* General hardware address lists handling functions */ 4138 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4139 struct netdev_hw_addr_list *from_list, int addr_len); 4140 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4141 struct netdev_hw_addr_list *from_list, int addr_len); 4142 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4143 struct net_device *dev, 4144 int (*sync)(struct net_device *, const unsigned char *), 4145 int (*unsync)(struct net_device *, 4146 const unsigned char *)); 4147 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4148 struct net_device *dev, 4149 int (*sync)(struct net_device *, 4150 const unsigned char *, int), 4151 int (*unsync)(struct net_device *, 4152 const unsigned char *, int)); 4153 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4154 struct net_device *dev, 4155 int (*unsync)(struct net_device *, 4156 const unsigned char *, int)); 4157 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4158 struct net_device *dev, 4159 int (*unsync)(struct net_device *, 4160 const unsigned char *)); 4161 void __hw_addr_init(struct netdev_hw_addr_list *list); 4162 4163 /* Functions used for device addresses handling */ 4164 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4165 unsigned char addr_type); 4166 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4167 unsigned char addr_type); 4168 void dev_addr_flush(struct net_device *dev); 4169 int dev_addr_init(struct net_device *dev); 4170 4171 /* Functions used for unicast addresses handling */ 4172 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4173 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4174 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4175 int dev_uc_sync(struct net_device *to, struct net_device *from); 4176 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4177 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4178 void dev_uc_flush(struct net_device *dev); 4179 void dev_uc_init(struct net_device *dev); 4180 4181 /** 4182 * __dev_uc_sync - Synchonize device's unicast list 4183 * @dev: device to sync 4184 * @sync: function to call if address should be added 4185 * @unsync: function to call if address should be removed 4186 * 4187 * Add newly added addresses to the interface, and release 4188 * addresses that have been deleted. 4189 */ 4190 static inline int __dev_uc_sync(struct net_device *dev, 4191 int (*sync)(struct net_device *, 4192 const unsigned char *), 4193 int (*unsync)(struct net_device *, 4194 const unsigned char *)) 4195 { 4196 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4197 } 4198 4199 /** 4200 * __dev_uc_unsync - Remove synchronized addresses from device 4201 * @dev: device to sync 4202 * @unsync: function to call if address should be removed 4203 * 4204 * Remove all addresses that were added to the device by dev_uc_sync(). 4205 */ 4206 static inline void __dev_uc_unsync(struct net_device *dev, 4207 int (*unsync)(struct net_device *, 4208 const unsigned char *)) 4209 { 4210 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4211 } 4212 4213 /* Functions used for multicast addresses handling */ 4214 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4215 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4216 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4217 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4218 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4219 int dev_mc_sync(struct net_device *to, struct net_device *from); 4220 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4221 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4222 void dev_mc_flush(struct net_device *dev); 4223 void dev_mc_init(struct net_device *dev); 4224 4225 /** 4226 * __dev_mc_sync - Synchonize device's multicast list 4227 * @dev: device to sync 4228 * @sync: function to call if address should be added 4229 * @unsync: function to call if address should be removed 4230 * 4231 * Add newly added addresses to the interface, and release 4232 * addresses that have been deleted. 4233 */ 4234 static inline int __dev_mc_sync(struct net_device *dev, 4235 int (*sync)(struct net_device *, 4236 const unsigned char *), 4237 int (*unsync)(struct net_device *, 4238 const unsigned char *)) 4239 { 4240 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4241 } 4242 4243 /** 4244 * __dev_mc_unsync - Remove synchronized addresses from device 4245 * @dev: device to sync 4246 * @unsync: function to call if address should be removed 4247 * 4248 * Remove all addresses that were added to the device by dev_mc_sync(). 4249 */ 4250 static inline void __dev_mc_unsync(struct net_device *dev, 4251 int (*unsync)(struct net_device *, 4252 const unsigned char *)) 4253 { 4254 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4255 } 4256 4257 /* Functions used for secondary unicast and multicast support */ 4258 void dev_set_rx_mode(struct net_device *dev); 4259 void __dev_set_rx_mode(struct net_device *dev); 4260 int dev_set_promiscuity(struct net_device *dev, int inc); 4261 int dev_set_allmulti(struct net_device *dev, int inc); 4262 void netdev_state_change(struct net_device *dev); 4263 void netdev_notify_peers(struct net_device *dev); 4264 void netdev_features_change(struct net_device *dev); 4265 /* Load a device via the kmod */ 4266 void dev_load(struct net *net, const char *name); 4267 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4268 struct rtnl_link_stats64 *storage); 4269 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4270 const struct net_device_stats *netdev_stats); 4271 4272 extern int netdev_max_backlog; 4273 extern int netdev_tstamp_prequeue; 4274 extern int weight_p; 4275 extern int dev_weight_rx_bias; 4276 extern int dev_weight_tx_bias; 4277 extern int dev_rx_weight; 4278 extern int dev_tx_weight; 4279 extern int gro_normal_batch; 4280 4281 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4282 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4283 struct list_head **iter); 4284 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 4285 struct list_head **iter); 4286 4287 /* iterate through upper list, must be called under RCU read lock */ 4288 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4289 for (iter = &(dev)->adj_list.upper, \ 4290 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4291 updev; \ 4292 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4293 4294 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4295 int (*fn)(struct net_device *upper_dev, 4296 void *data), 4297 void *data); 4298 4299 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4300 struct net_device *upper_dev); 4301 4302 bool netdev_has_any_upper_dev(struct net_device *dev); 4303 4304 void *netdev_lower_get_next_private(struct net_device *dev, 4305 struct list_head **iter); 4306 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4307 struct list_head **iter); 4308 4309 #define netdev_for_each_lower_private(dev, priv, iter) \ 4310 for (iter = (dev)->adj_list.lower.next, \ 4311 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4312 priv; \ 4313 priv = netdev_lower_get_next_private(dev, &(iter))) 4314 4315 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4316 for (iter = &(dev)->adj_list.lower, \ 4317 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4318 priv; \ 4319 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4320 4321 void *netdev_lower_get_next(struct net_device *dev, 4322 struct list_head **iter); 4323 4324 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4325 for (iter = (dev)->adj_list.lower.next, \ 4326 ldev = netdev_lower_get_next(dev, &(iter)); \ 4327 ldev; \ 4328 ldev = netdev_lower_get_next(dev, &(iter))) 4329 4330 struct net_device *netdev_all_lower_get_next(struct net_device *dev, 4331 struct list_head **iter); 4332 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev, 4333 struct list_head **iter); 4334 4335 int netdev_walk_all_lower_dev(struct net_device *dev, 4336 int (*fn)(struct net_device *lower_dev, 4337 void *data), 4338 void *data); 4339 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4340 int (*fn)(struct net_device *lower_dev, 4341 void *data), 4342 void *data); 4343 4344 void *netdev_adjacent_get_private(struct list_head *adj_list); 4345 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4346 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4347 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4348 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4349 struct netlink_ext_ack *extack); 4350 int netdev_master_upper_dev_link(struct net_device *dev, 4351 struct net_device *upper_dev, 4352 void *upper_priv, void *upper_info, 4353 struct netlink_ext_ack *extack); 4354 void netdev_upper_dev_unlink(struct net_device *dev, 4355 struct net_device *upper_dev); 4356 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4357 struct net_device *new_dev, 4358 struct net_device *dev, 4359 struct netlink_ext_ack *extack); 4360 void netdev_adjacent_change_commit(struct net_device *old_dev, 4361 struct net_device *new_dev, 4362 struct net_device *dev); 4363 void netdev_adjacent_change_abort(struct net_device *old_dev, 4364 struct net_device *new_dev, 4365 struct net_device *dev); 4366 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4367 void *netdev_lower_dev_get_private(struct net_device *dev, 4368 struct net_device *lower_dev); 4369 void netdev_lower_state_changed(struct net_device *lower_dev, 4370 void *lower_state_info); 4371 4372 /* RSS keys are 40 or 52 bytes long */ 4373 #define NETDEV_RSS_KEY_LEN 52 4374 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4375 void netdev_rss_key_fill(void *buffer, size_t len); 4376 4377 int skb_checksum_help(struct sk_buff *skb); 4378 int skb_crc32c_csum_help(struct sk_buff *skb); 4379 int skb_csum_hwoffload_help(struct sk_buff *skb, 4380 const netdev_features_t features); 4381 4382 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4383 netdev_features_t features, bool tx_path); 4384 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4385 netdev_features_t features); 4386 4387 struct netdev_bonding_info { 4388 ifslave slave; 4389 ifbond master; 4390 }; 4391 4392 struct netdev_notifier_bonding_info { 4393 struct netdev_notifier_info info; /* must be first */ 4394 struct netdev_bonding_info bonding_info; 4395 }; 4396 4397 void netdev_bonding_info_change(struct net_device *dev, 4398 struct netdev_bonding_info *bonding_info); 4399 4400 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4401 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4402 #else 4403 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4404 const void *data) 4405 { 4406 } 4407 #endif 4408 4409 static inline 4410 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4411 { 4412 return __skb_gso_segment(skb, features, true); 4413 } 4414 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4415 4416 static inline bool can_checksum_protocol(netdev_features_t features, 4417 __be16 protocol) 4418 { 4419 if (protocol == htons(ETH_P_FCOE)) 4420 return !!(features & NETIF_F_FCOE_CRC); 4421 4422 /* Assume this is an IP checksum (not SCTP CRC) */ 4423 4424 if (features & NETIF_F_HW_CSUM) { 4425 /* Can checksum everything */ 4426 return true; 4427 } 4428 4429 switch (protocol) { 4430 case htons(ETH_P_IP): 4431 return !!(features & NETIF_F_IP_CSUM); 4432 case htons(ETH_P_IPV6): 4433 return !!(features & NETIF_F_IPV6_CSUM); 4434 default: 4435 return false; 4436 } 4437 } 4438 4439 #ifdef CONFIG_BUG 4440 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4441 #else 4442 static inline void netdev_rx_csum_fault(struct net_device *dev, 4443 struct sk_buff *skb) 4444 { 4445 } 4446 #endif 4447 /* rx skb timestamps */ 4448 void net_enable_timestamp(void); 4449 void net_disable_timestamp(void); 4450 4451 #ifdef CONFIG_PROC_FS 4452 int __init dev_proc_init(void); 4453 #else 4454 #define dev_proc_init() 0 4455 #endif 4456 4457 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4458 struct sk_buff *skb, struct net_device *dev, 4459 bool more) 4460 { 4461 __this_cpu_write(softnet_data.xmit.more, more); 4462 return ops->ndo_start_xmit(skb, dev); 4463 } 4464 4465 static inline bool netdev_xmit_more(void) 4466 { 4467 return __this_cpu_read(softnet_data.xmit.more); 4468 } 4469 4470 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4471 struct netdev_queue *txq, bool more) 4472 { 4473 const struct net_device_ops *ops = dev->netdev_ops; 4474 netdev_tx_t rc; 4475 4476 rc = __netdev_start_xmit(ops, skb, dev, more); 4477 if (rc == NETDEV_TX_OK) 4478 txq_trans_update(txq); 4479 4480 return rc; 4481 } 4482 4483 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4484 const void *ns); 4485 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4486 const void *ns); 4487 4488 static inline int netdev_class_create_file(const struct class_attribute *class_attr) 4489 { 4490 return netdev_class_create_file_ns(class_attr, NULL); 4491 } 4492 4493 static inline void netdev_class_remove_file(const struct class_attribute *class_attr) 4494 { 4495 netdev_class_remove_file_ns(class_attr, NULL); 4496 } 4497 4498 extern const struct kobj_ns_type_operations net_ns_type_operations; 4499 4500 const char *netdev_drivername(const struct net_device *dev); 4501 4502 void linkwatch_run_queue(void); 4503 4504 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4505 netdev_features_t f2) 4506 { 4507 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4508 if (f1 & NETIF_F_HW_CSUM) 4509 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4510 else 4511 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4512 } 4513 4514 return f1 & f2; 4515 } 4516 4517 static inline netdev_features_t netdev_get_wanted_features( 4518 struct net_device *dev) 4519 { 4520 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4521 } 4522 netdev_features_t netdev_increment_features(netdev_features_t all, 4523 netdev_features_t one, netdev_features_t mask); 4524 4525 /* Allow TSO being used on stacked device : 4526 * Performing the GSO segmentation before last device 4527 * is a performance improvement. 4528 */ 4529 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4530 netdev_features_t mask) 4531 { 4532 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4533 } 4534 4535 int __netdev_update_features(struct net_device *dev); 4536 void netdev_update_features(struct net_device *dev); 4537 void netdev_change_features(struct net_device *dev); 4538 4539 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4540 struct net_device *dev); 4541 4542 netdev_features_t passthru_features_check(struct sk_buff *skb, 4543 struct net_device *dev, 4544 netdev_features_t features); 4545 netdev_features_t netif_skb_features(struct sk_buff *skb); 4546 4547 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4548 { 4549 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4550 4551 /* check flags correspondence */ 4552 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4553 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4554 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4555 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4556 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4557 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4558 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4559 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4560 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4561 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4562 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4563 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4564 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4565 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4566 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4567 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4568 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4569 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4570 4571 return (features & feature) == feature; 4572 } 4573 4574 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4575 { 4576 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4577 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4578 } 4579 4580 static inline bool netif_needs_gso(struct sk_buff *skb, 4581 netdev_features_t features) 4582 { 4583 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4584 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4585 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4586 } 4587 4588 static inline void netif_set_gso_max_size(struct net_device *dev, 4589 unsigned int size) 4590 { 4591 dev->gso_max_size = size; 4592 } 4593 4594 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4595 int pulled_hlen, u16 mac_offset, 4596 int mac_len) 4597 { 4598 skb->protocol = protocol; 4599 skb->encapsulation = 1; 4600 skb_push(skb, pulled_hlen); 4601 skb_reset_transport_header(skb); 4602 skb->mac_header = mac_offset; 4603 skb->network_header = skb->mac_header + mac_len; 4604 skb->mac_len = mac_len; 4605 } 4606 4607 static inline bool netif_is_macsec(const struct net_device *dev) 4608 { 4609 return dev->priv_flags & IFF_MACSEC; 4610 } 4611 4612 static inline bool netif_is_macvlan(const struct net_device *dev) 4613 { 4614 return dev->priv_flags & IFF_MACVLAN; 4615 } 4616 4617 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4618 { 4619 return dev->priv_flags & IFF_MACVLAN_PORT; 4620 } 4621 4622 static inline bool netif_is_bond_master(const struct net_device *dev) 4623 { 4624 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4625 } 4626 4627 static inline bool netif_is_bond_slave(const struct net_device *dev) 4628 { 4629 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4630 } 4631 4632 static inline bool netif_supports_nofcs(struct net_device *dev) 4633 { 4634 return dev->priv_flags & IFF_SUPP_NOFCS; 4635 } 4636 4637 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4638 { 4639 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4640 } 4641 4642 static inline bool netif_is_l3_master(const struct net_device *dev) 4643 { 4644 return dev->priv_flags & IFF_L3MDEV_MASTER; 4645 } 4646 4647 static inline bool netif_is_l3_slave(const struct net_device *dev) 4648 { 4649 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4650 } 4651 4652 static inline bool netif_is_bridge_master(const struct net_device *dev) 4653 { 4654 return dev->priv_flags & IFF_EBRIDGE; 4655 } 4656 4657 static inline bool netif_is_bridge_port(const struct net_device *dev) 4658 { 4659 return dev->priv_flags & IFF_BRIDGE_PORT; 4660 } 4661 4662 static inline bool netif_is_ovs_master(const struct net_device *dev) 4663 { 4664 return dev->priv_flags & IFF_OPENVSWITCH; 4665 } 4666 4667 static inline bool netif_is_ovs_port(const struct net_device *dev) 4668 { 4669 return dev->priv_flags & IFF_OVS_DATAPATH; 4670 } 4671 4672 static inline bool netif_is_team_master(const struct net_device *dev) 4673 { 4674 return dev->priv_flags & IFF_TEAM; 4675 } 4676 4677 static inline bool netif_is_team_port(const struct net_device *dev) 4678 { 4679 return dev->priv_flags & IFF_TEAM_PORT; 4680 } 4681 4682 static inline bool netif_is_lag_master(const struct net_device *dev) 4683 { 4684 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4685 } 4686 4687 static inline bool netif_is_lag_port(const struct net_device *dev) 4688 { 4689 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4690 } 4691 4692 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4693 { 4694 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4695 } 4696 4697 static inline bool netif_is_failover(const struct net_device *dev) 4698 { 4699 return dev->priv_flags & IFF_FAILOVER; 4700 } 4701 4702 static inline bool netif_is_failover_slave(const struct net_device *dev) 4703 { 4704 return dev->priv_flags & IFF_FAILOVER_SLAVE; 4705 } 4706 4707 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4708 static inline void netif_keep_dst(struct net_device *dev) 4709 { 4710 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4711 } 4712 4713 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4714 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4715 { 4716 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4717 return dev->priv_flags & IFF_MACSEC; 4718 } 4719 4720 extern struct pernet_operations __net_initdata loopback_net_ops; 4721 4722 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4723 4724 /* netdev_printk helpers, similar to dev_printk */ 4725 4726 static inline const char *netdev_name(const struct net_device *dev) 4727 { 4728 if (!dev->name[0] || strchr(dev->name, '%')) 4729 return "(unnamed net_device)"; 4730 return dev->name; 4731 } 4732 4733 static inline bool netdev_unregistering(const struct net_device *dev) 4734 { 4735 return dev->reg_state == NETREG_UNREGISTERING; 4736 } 4737 4738 static inline const char *netdev_reg_state(const struct net_device *dev) 4739 { 4740 switch (dev->reg_state) { 4741 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4742 case NETREG_REGISTERED: return ""; 4743 case NETREG_UNREGISTERING: return " (unregistering)"; 4744 case NETREG_UNREGISTERED: return " (unregistered)"; 4745 case NETREG_RELEASED: return " (released)"; 4746 case NETREG_DUMMY: return " (dummy)"; 4747 } 4748 4749 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4750 return " (unknown)"; 4751 } 4752 4753 __printf(3, 4) __cold 4754 void netdev_printk(const char *level, const struct net_device *dev, 4755 const char *format, ...); 4756 __printf(2, 3) __cold 4757 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4758 __printf(2, 3) __cold 4759 void netdev_alert(const struct net_device *dev, const char *format, ...); 4760 __printf(2, 3) __cold 4761 void netdev_crit(const struct net_device *dev, const char *format, ...); 4762 __printf(2, 3) __cold 4763 void netdev_err(const struct net_device *dev, const char *format, ...); 4764 __printf(2, 3) __cold 4765 void netdev_warn(const struct net_device *dev, const char *format, ...); 4766 __printf(2, 3) __cold 4767 void netdev_notice(const struct net_device *dev, const char *format, ...); 4768 __printf(2, 3) __cold 4769 void netdev_info(const struct net_device *dev, const char *format, ...); 4770 4771 #define netdev_level_once(level, dev, fmt, ...) \ 4772 do { \ 4773 static bool __print_once __read_mostly; \ 4774 \ 4775 if (!__print_once) { \ 4776 __print_once = true; \ 4777 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 4778 } \ 4779 } while (0) 4780 4781 #define netdev_emerg_once(dev, fmt, ...) \ 4782 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 4783 #define netdev_alert_once(dev, fmt, ...) \ 4784 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 4785 #define netdev_crit_once(dev, fmt, ...) \ 4786 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 4787 #define netdev_err_once(dev, fmt, ...) \ 4788 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 4789 #define netdev_warn_once(dev, fmt, ...) \ 4790 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 4791 #define netdev_notice_once(dev, fmt, ...) \ 4792 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 4793 #define netdev_info_once(dev, fmt, ...) \ 4794 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 4795 4796 #define MODULE_ALIAS_NETDEV(device) \ 4797 MODULE_ALIAS("netdev-" device) 4798 4799 #if defined(CONFIG_DYNAMIC_DEBUG) 4800 #define netdev_dbg(__dev, format, args...) \ 4801 do { \ 4802 dynamic_netdev_dbg(__dev, format, ##args); \ 4803 } while (0) 4804 #elif defined(DEBUG) 4805 #define netdev_dbg(__dev, format, args...) \ 4806 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4807 #else 4808 #define netdev_dbg(__dev, format, args...) \ 4809 ({ \ 4810 if (0) \ 4811 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4812 }) 4813 #endif 4814 4815 #if defined(VERBOSE_DEBUG) 4816 #define netdev_vdbg netdev_dbg 4817 #else 4818 4819 #define netdev_vdbg(dev, format, args...) \ 4820 ({ \ 4821 if (0) \ 4822 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 4823 0; \ 4824 }) 4825 #endif 4826 4827 /* 4828 * netdev_WARN() acts like dev_printk(), but with the key difference 4829 * of using a WARN/WARN_ON to get the message out, including the 4830 * file/line information and a backtrace. 4831 */ 4832 #define netdev_WARN(dev, format, args...) \ 4833 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4834 netdev_reg_state(dev), ##args) 4835 4836 #define netdev_WARN_ONCE(dev, format, args...) \ 4837 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4838 netdev_reg_state(dev), ##args) 4839 4840 /* netif printk helpers, similar to netdev_printk */ 4841 4842 #define netif_printk(priv, type, level, dev, fmt, args...) \ 4843 do { \ 4844 if (netif_msg_##type(priv)) \ 4845 netdev_printk(level, (dev), fmt, ##args); \ 4846 } while (0) 4847 4848 #define netif_level(level, priv, type, dev, fmt, args...) \ 4849 do { \ 4850 if (netif_msg_##type(priv)) \ 4851 netdev_##level(dev, fmt, ##args); \ 4852 } while (0) 4853 4854 #define netif_emerg(priv, type, dev, fmt, args...) \ 4855 netif_level(emerg, priv, type, dev, fmt, ##args) 4856 #define netif_alert(priv, type, dev, fmt, args...) \ 4857 netif_level(alert, priv, type, dev, fmt, ##args) 4858 #define netif_crit(priv, type, dev, fmt, args...) \ 4859 netif_level(crit, priv, type, dev, fmt, ##args) 4860 #define netif_err(priv, type, dev, fmt, args...) \ 4861 netif_level(err, priv, type, dev, fmt, ##args) 4862 #define netif_warn(priv, type, dev, fmt, args...) \ 4863 netif_level(warn, priv, type, dev, fmt, ##args) 4864 #define netif_notice(priv, type, dev, fmt, args...) \ 4865 netif_level(notice, priv, type, dev, fmt, ##args) 4866 #define netif_info(priv, type, dev, fmt, args...) \ 4867 netif_level(info, priv, type, dev, fmt, ##args) 4868 4869 #if defined(CONFIG_DYNAMIC_DEBUG) 4870 #define netif_dbg(priv, type, netdev, format, args...) \ 4871 do { \ 4872 if (netif_msg_##type(priv)) \ 4873 dynamic_netdev_dbg(netdev, format, ##args); \ 4874 } while (0) 4875 #elif defined(DEBUG) 4876 #define netif_dbg(priv, type, dev, format, args...) \ 4877 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 4878 #else 4879 #define netif_dbg(priv, type, dev, format, args...) \ 4880 ({ \ 4881 if (0) \ 4882 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4883 0; \ 4884 }) 4885 #endif 4886 4887 /* if @cond then downgrade to debug, else print at @level */ 4888 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 4889 do { \ 4890 if (cond) \ 4891 netif_dbg(priv, type, netdev, fmt, ##args); \ 4892 else \ 4893 netif_ ## level(priv, type, netdev, fmt, ##args); \ 4894 } while (0) 4895 4896 #if defined(VERBOSE_DEBUG) 4897 #define netif_vdbg netif_dbg 4898 #else 4899 #define netif_vdbg(priv, type, dev, format, args...) \ 4900 ({ \ 4901 if (0) \ 4902 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4903 0; \ 4904 }) 4905 #endif 4906 4907 /* 4908 * The list of packet types we will receive (as opposed to discard) 4909 * and the routines to invoke. 4910 * 4911 * Why 16. Because with 16 the only overlap we get on a hash of the 4912 * low nibble of the protocol value is RARP/SNAP/X.25. 4913 * 4914 * 0800 IP 4915 * 0001 802.3 4916 * 0002 AX.25 4917 * 0004 802.2 4918 * 8035 RARP 4919 * 0005 SNAP 4920 * 0805 X.25 4921 * 0806 ARP 4922 * 8137 IPX 4923 * 0009 Localtalk 4924 * 86DD IPv6 4925 */ 4926 #define PTYPE_HASH_SIZE (16) 4927 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 4928 4929 extern struct net_device *blackhole_netdev; 4930 4931 #endif /* _LINUX_NETDEVICE_H */ 4932