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 #include <asm/local.h> 32 33 #include <linux/percpu.h> 34 #include <linux/rculist.h> 35 #include <linux/workqueue.h> 36 #include <linux/dynamic_queue_limits.h> 37 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 #include <linux/rbtree.h> 52 #include <net/net_trackers.h> 53 #include <net/net_debug.h> 54 55 struct netpoll_info; 56 struct device; 57 struct ethtool_ops; 58 struct phy_device; 59 struct dsa_port; 60 struct ip_tunnel_parm; 61 struct macsec_context; 62 struct macsec_ops; 63 struct netdev_name_node; 64 struct sd_flow_limit; 65 struct sfp_bus; 66 /* 802.11 specific */ 67 struct wireless_dev; 68 /* 802.15.4 specific */ 69 struct wpan_dev; 70 struct mpls_dev; 71 /* UDP Tunnel offloads */ 72 struct udp_tunnel_info; 73 struct udp_tunnel_nic_info; 74 struct udp_tunnel_nic; 75 struct bpf_prog; 76 struct xdp_buff; 77 78 void synchronize_net(void); 79 void netdev_set_default_ethtool_ops(struct net_device *dev, 80 const struct ethtool_ops *ops); 81 82 /* Backlog congestion levels */ 83 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 84 #define NET_RX_DROP 1 /* packet dropped */ 85 86 #define MAX_NEST_DEV 8 87 88 /* 89 * Transmit return codes: transmit return codes originate from three different 90 * namespaces: 91 * 92 * - qdisc return codes 93 * - driver transmit return codes 94 * - errno values 95 * 96 * Drivers are allowed to return any one of those in their hard_start_xmit() 97 * function. Real network devices commonly used with qdiscs should only return 98 * the driver transmit return codes though - when qdiscs are used, the actual 99 * transmission happens asynchronously, so the value is not propagated to 100 * higher layers. Virtual network devices transmit synchronously; in this case 101 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 102 * others are propagated to higher layers. 103 */ 104 105 /* qdisc ->enqueue() return codes. */ 106 #define NET_XMIT_SUCCESS 0x00 107 #define NET_XMIT_DROP 0x01 /* skb dropped */ 108 #define NET_XMIT_CN 0x02 /* congestion notification */ 109 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 110 111 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 112 * indicates that the device will soon be dropping packets, or already drops 113 * some packets of the same priority; prompting us to send less aggressively. */ 114 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 115 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 116 117 /* Driver transmit return codes */ 118 #define NETDEV_TX_MASK 0xf0 119 120 enum netdev_tx { 121 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 122 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 123 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 124 }; 125 typedef enum netdev_tx netdev_tx_t; 126 127 /* 128 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 129 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 130 */ 131 static inline bool dev_xmit_complete(int rc) 132 { 133 /* 134 * Positive cases with an skb consumed by a driver: 135 * - successful transmission (rc == NETDEV_TX_OK) 136 * - error while transmitting (rc < 0) 137 * - error while queueing to a different device (rc & NET_XMIT_MASK) 138 */ 139 if (likely(rc < NET_XMIT_MASK)) 140 return true; 141 142 return false; 143 } 144 145 /* 146 * Compute the worst-case header length according to the protocols 147 * used. 148 */ 149 150 #if defined(CONFIG_HYPERV_NET) 151 # define LL_MAX_HEADER 128 152 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 153 # if defined(CONFIG_MAC80211_MESH) 154 # define LL_MAX_HEADER 128 155 # else 156 # define LL_MAX_HEADER 96 157 # endif 158 #else 159 # define LL_MAX_HEADER 32 160 #endif 161 162 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 163 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 164 #define MAX_HEADER LL_MAX_HEADER 165 #else 166 #define MAX_HEADER (LL_MAX_HEADER + 48) 167 #endif 168 169 /* 170 * Old network device statistics. Fields are native words 171 * (unsigned long) so they can be read and written atomically. 172 */ 173 174 struct net_device_stats { 175 unsigned long rx_packets; 176 unsigned long tx_packets; 177 unsigned long rx_bytes; 178 unsigned long tx_bytes; 179 unsigned long rx_errors; 180 unsigned long tx_errors; 181 unsigned long rx_dropped; 182 unsigned long tx_dropped; 183 unsigned long multicast; 184 unsigned long collisions; 185 unsigned long rx_length_errors; 186 unsigned long rx_over_errors; 187 unsigned long rx_crc_errors; 188 unsigned long rx_frame_errors; 189 unsigned long rx_fifo_errors; 190 unsigned long rx_missed_errors; 191 unsigned long tx_aborted_errors; 192 unsigned long tx_carrier_errors; 193 unsigned long tx_fifo_errors; 194 unsigned long tx_heartbeat_errors; 195 unsigned long tx_window_errors; 196 unsigned long rx_compressed; 197 unsigned long tx_compressed; 198 }; 199 200 /* per-cpu stats, allocated on demand. 201 * Try to fit them in a single cache line, for dev_get_stats() sake. 202 */ 203 struct net_device_core_stats { 204 unsigned long rx_dropped; 205 unsigned long tx_dropped; 206 unsigned long rx_nohandler; 207 unsigned long rx_otherhost_dropped; 208 } __aligned(4 * sizeof(unsigned long)); 209 210 #include <linux/cache.h> 211 #include <linux/skbuff.h> 212 213 #ifdef CONFIG_RPS 214 #include <linux/static_key.h> 215 extern struct static_key_false rps_needed; 216 extern struct static_key_false rfs_needed; 217 #endif 218 219 struct neighbour; 220 struct neigh_parms; 221 struct sk_buff; 222 223 struct netdev_hw_addr { 224 struct list_head list; 225 struct rb_node node; 226 unsigned char addr[MAX_ADDR_LEN]; 227 unsigned char type; 228 #define NETDEV_HW_ADDR_T_LAN 1 229 #define NETDEV_HW_ADDR_T_SAN 2 230 #define NETDEV_HW_ADDR_T_UNICAST 3 231 #define NETDEV_HW_ADDR_T_MULTICAST 4 232 bool global_use; 233 int sync_cnt; 234 int refcount; 235 int synced; 236 struct rcu_head rcu_head; 237 }; 238 239 struct netdev_hw_addr_list { 240 struct list_head list; 241 int count; 242 243 /* Auxiliary tree for faster lookup on addition and deletion */ 244 struct rb_root tree; 245 }; 246 247 #define netdev_hw_addr_list_count(l) ((l)->count) 248 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 249 #define netdev_hw_addr_list_for_each(ha, l) \ 250 list_for_each_entry(ha, &(l)->list, list) 251 252 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 253 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 254 #define netdev_for_each_uc_addr(ha, dev) \ 255 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 256 257 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 258 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 259 #define netdev_for_each_mc_addr(ha, dev) \ 260 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 261 262 struct hh_cache { 263 unsigned int hh_len; 264 seqlock_t hh_lock; 265 266 /* cached hardware header; allow for machine alignment needs. */ 267 #define HH_DATA_MOD 16 268 #define HH_DATA_OFF(__len) \ 269 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 270 #define HH_DATA_ALIGN(__len) \ 271 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 272 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 273 }; 274 275 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 276 * Alternative is: 277 * dev->hard_header_len ? (dev->hard_header_len + 278 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 279 * 280 * We could use other alignment values, but we must maintain the 281 * relationship HH alignment <= LL alignment. 282 */ 283 #define LL_RESERVED_SPACE(dev) \ 284 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 285 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 286 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 287 288 struct header_ops { 289 int (*create) (struct sk_buff *skb, struct net_device *dev, 290 unsigned short type, const void *daddr, 291 const void *saddr, unsigned int len); 292 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 293 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 294 void (*cache_update)(struct hh_cache *hh, 295 const struct net_device *dev, 296 const unsigned char *haddr); 297 bool (*validate)(const char *ll_header, unsigned int len); 298 __be16 (*parse_protocol)(const struct sk_buff *skb); 299 }; 300 301 /* These flag bits are private to the generic network queueing 302 * layer; they may not be explicitly referenced by any other 303 * code. 304 */ 305 306 enum netdev_state_t { 307 __LINK_STATE_START, 308 __LINK_STATE_PRESENT, 309 __LINK_STATE_NOCARRIER, 310 __LINK_STATE_LINKWATCH_PENDING, 311 __LINK_STATE_DORMANT, 312 __LINK_STATE_TESTING, 313 }; 314 315 struct gro_list { 316 struct list_head list; 317 int count; 318 }; 319 320 /* 321 * size of gro hash buckets, must less than bit number of 322 * napi_struct::gro_bitmask 323 */ 324 #define GRO_HASH_BUCKETS 8 325 326 /* 327 * Structure for NAPI scheduling similar to tasklet but with weighting 328 */ 329 struct napi_struct { 330 /* The poll_list must only be managed by the entity which 331 * changes the state of the NAPI_STATE_SCHED bit. This means 332 * whoever atomically sets that bit can add this napi_struct 333 * to the per-CPU poll_list, and whoever clears that bit 334 * can remove from the list right before clearing the bit. 335 */ 336 struct list_head poll_list; 337 338 unsigned long state; 339 int weight; 340 int defer_hard_irqs_count; 341 unsigned long gro_bitmask; 342 int (*poll)(struct napi_struct *, int); 343 #ifdef CONFIG_NETPOLL 344 int poll_owner; 345 #endif 346 struct net_device *dev; 347 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 348 struct sk_buff *skb; 349 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 350 int rx_count; /* length of rx_list */ 351 struct hrtimer timer; 352 struct list_head dev_list; 353 struct hlist_node napi_hash_node; 354 unsigned int napi_id; 355 struct task_struct *thread; 356 }; 357 358 enum { 359 NAPI_STATE_SCHED, /* Poll is scheduled */ 360 NAPI_STATE_MISSED, /* reschedule a napi */ 361 NAPI_STATE_DISABLE, /* Disable pending */ 362 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 363 NAPI_STATE_LISTED, /* NAPI added to system lists */ 364 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 365 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 366 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 367 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 368 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 369 }; 370 371 enum { 372 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 373 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 374 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 375 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 376 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 377 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 378 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 379 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 380 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 381 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 382 }; 383 384 enum gro_result { 385 GRO_MERGED, 386 GRO_MERGED_FREE, 387 GRO_HELD, 388 GRO_NORMAL, 389 GRO_CONSUMED, 390 }; 391 typedef enum gro_result gro_result_t; 392 393 /* 394 * enum rx_handler_result - Possible return values for rx_handlers. 395 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 396 * further. 397 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 398 * case skb->dev was changed by rx_handler. 399 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 400 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 401 * 402 * rx_handlers are functions called from inside __netif_receive_skb(), to do 403 * special processing of the skb, prior to delivery to protocol handlers. 404 * 405 * Currently, a net_device can only have a single rx_handler registered. Trying 406 * to register a second rx_handler will return -EBUSY. 407 * 408 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 409 * To unregister a rx_handler on a net_device, use 410 * netdev_rx_handler_unregister(). 411 * 412 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 413 * do with the skb. 414 * 415 * If the rx_handler consumed the skb in some way, it should return 416 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 417 * the skb to be delivered in some other way. 418 * 419 * If the rx_handler changed skb->dev, to divert the skb to another 420 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 421 * new device will be called if it exists. 422 * 423 * If the rx_handler decides the skb should be ignored, it should return 424 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 425 * are registered on exact device (ptype->dev == skb->dev). 426 * 427 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 428 * delivered, it should return RX_HANDLER_PASS. 429 * 430 * A device without a registered rx_handler will behave as if rx_handler 431 * returned RX_HANDLER_PASS. 432 */ 433 434 enum rx_handler_result { 435 RX_HANDLER_CONSUMED, 436 RX_HANDLER_ANOTHER, 437 RX_HANDLER_EXACT, 438 RX_HANDLER_PASS, 439 }; 440 typedef enum rx_handler_result rx_handler_result_t; 441 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 442 443 void __napi_schedule(struct napi_struct *n); 444 void __napi_schedule_irqoff(struct napi_struct *n); 445 446 static inline bool napi_disable_pending(struct napi_struct *n) 447 { 448 return test_bit(NAPI_STATE_DISABLE, &n->state); 449 } 450 451 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 452 { 453 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 454 } 455 456 bool napi_schedule_prep(struct napi_struct *n); 457 458 /** 459 * napi_schedule - schedule NAPI poll 460 * @n: NAPI context 461 * 462 * Schedule NAPI poll routine to be called if it is not already 463 * running. 464 */ 465 static inline void napi_schedule(struct napi_struct *n) 466 { 467 if (napi_schedule_prep(n)) 468 __napi_schedule(n); 469 } 470 471 /** 472 * napi_schedule_irqoff - schedule NAPI poll 473 * @n: NAPI context 474 * 475 * Variant of napi_schedule(), assuming hard irqs are masked. 476 */ 477 static inline void napi_schedule_irqoff(struct napi_struct *n) 478 { 479 if (napi_schedule_prep(n)) 480 __napi_schedule_irqoff(n); 481 } 482 483 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 484 static inline bool napi_reschedule(struct napi_struct *napi) 485 { 486 if (napi_schedule_prep(napi)) { 487 __napi_schedule(napi); 488 return true; 489 } 490 return false; 491 } 492 493 bool napi_complete_done(struct napi_struct *n, int work_done); 494 /** 495 * napi_complete - NAPI processing complete 496 * @n: NAPI context 497 * 498 * Mark NAPI processing as complete. 499 * Consider using napi_complete_done() instead. 500 * Return false if device should avoid rearming interrupts. 501 */ 502 static inline bool napi_complete(struct napi_struct *n) 503 { 504 return napi_complete_done(n, 0); 505 } 506 507 int dev_set_threaded(struct net_device *dev, bool threaded); 508 509 /** 510 * napi_disable - prevent NAPI from scheduling 511 * @n: NAPI context 512 * 513 * Stop NAPI from being scheduled on this context. 514 * Waits till any outstanding processing completes. 515 */ 516 void napi_disable(struct napi_struct *n); 517 518 void napi_enable(struct napi_struct *n); 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 val = READ_ONCE(n->state); 550 do { 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 (!try_cmpxchg(&n->state, &val, new)); 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 netdevice_tracker dev_tracker; 595 596 struct Qdisc __rcu *qdisc; 597 struct Qdisc *qdisc_sleeping; 598 #ifdef CONFIG_SYSFS 599 struct kobject kobj; 600 #endif 601 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 602 int numa_node; 603 #endif 604 unsigned long tx_maxrate; 605 /* 606 * Number of TX timeouts for this queue 607 * (/sys/class/net/DEV/Q/trans_timeout) 608 */ 609 atomic_long_t trans_timeout; 610 611 /* Subordinate device that the queue has been assigned to */ 612 struct net_device *sb_dev; 613 #ifdef CONFIG_XDP_SOCKETS 614 struct xsk_buff_pool *pool; 615 #endif 616 /* 617 * write-mostly part 618 */ 619 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 620 int xmit_lock_owner; 621 /* 622 * Time (in jiffies) of last Tx 623 */ 624 unsigned long trans_start; 625 626 unsigned long state; 627 628 #ifdef CONFIG_BQL 629 struct dql dql; 630 #endif 631 } ____cacheline_aligned_in_smp; 632 633 extern int sysctl_fb_tunnels_only_for_init_net; 634 extern int sysctl_devconf_inherit_init_net; 635 636 /* 637 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 638 * == 1 : For initns only 639 * == 2 : For none. 640 */ 641 static inline bool net_has_fallback_tunnels(const struct net *net) 642 { 643 #if IS_ENABLED(CONFIG_SYSCTL) 644 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net); 645 646 return !fb_tunnels_only_for_init_net || 647 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1); 648 #else 649 return true; 650 #endif 651 } 652 653 static inline int net_inherit_devconf(void) 654 { 655 #if IS_ENABLED(CONFIG_SYSCTL) 656 return READ_ONCE(sysctl_devconf_inherit_init_net); 657 #else 658 return 0; 659 #endif 660 } 661 662 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 663 { 664 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 665 return q->numa_node; 666 #else 667 return NUMA_NO_NODE; 668 #endif 669 } 670 671 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 672 { 673 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 674 q->numa_node = node; 675 #endif 676 } 677 678 #ifdef CONFIG_RPS 679 /* 680 * This structure holds an RPS map which can be of variable length. The 681 * map is an array of CPUs. 682 */ 683 struct rps_map { 684 unsigned int len; 685 struct rcu_head rcu; 686 u16 cpus[]; 687 }; 688 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 689 690 /* 691 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 692 * tail pointer for that CPU's input queue at the time of last enqueue, and 693 * a hardware filter index. 694 */ 695 struct rps_dev_flow { 696 u16 cpu; 697 u16 filter; 698 unsigned int last_qtail; 699 }; 700 #define RPS_NO_FILTER 0xffff 701 702 /* 703 * The rps_dev_flow_table structure contains a table of flow mappings. 704 */ 705 struct rps_dev_flow_table { 706 unsigned int mask; 707 struct rcu_head rcu; 708 struct rps_dev_flow flows[]; 709 }; 710 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 711 ((_num) * sizeof(struct rps_dev_flow))) 712 713 /* 714 * The rps_sock_flow_table contains mappings of flows to the last CPU 715 * on which they were processed by the application (set in recvmsg). 716 * Each entry is a 32bit value. Upper part is the high-order bits 717 * of flow hash, lower part is CPU number. 718 * rps_cpu_mask is used to partition the space, depending on number of 719 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 720 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 721 * meaning we use 32-6=26 bits for the hash. 722 */ 723 struct rps_sock_flow_table { 724 u32 mask; 725 726 u32 ents[] ____cacheline_aligned_in_smp; 727 }; 728 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 729 730 #define RPS_NO_CPU 0xffff 731 732 extern u32 rps_cpu_mask; 733 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 734 735 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 736 u32 hash) 737 { 738 if (table && hash) { 739 unsigned int index = hash & table->mask; 740 u32 val = hash & ~rps_cpu_mask; 741 742 /* We only give a hint, preemption can change CPU under us */ 743 val |= raw_smp_processor_id(); 744 745 if (table->ents[index] != val) 746 table->ents[index] = val; 747 } 748 } 749 750 #ifdef CONFIG_RFS_ACCEL 751 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 752 u16 filter_id); 753 #endif 754 #endif /* CONFIG_RPS */ 755 756 /* This structure contains an instance of an RX queue. */ 757 struct netdev_rx_queue { 758 struct xdp_rxq_info xdp_rxq; 759 #ifdef CONFIG_RPS 760 struct rps_map __rcu *rps_map; 761 struct rps_dev_flow_table __rcu *rps_flow_table; 762 #endif 763 struct kobject kobj; 764 struct net_device *dev; 765 netdevice_tracker dev_tracker; 766 767 #ifdef CONFIG_XDP_SOCKETS 768 struct xsk_buff_pool *pool; 769 #endif 770 } ____cacheline_aligned_in_smp; 771 772 /* 773 * RX queue sysfs structures and functions. 774 */ 775 struct rx_queue_attribute { 776 struct attribute attr; 777 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 778 ssize_t (*store)(struct netdev_rx_queue *queue, 779 const char *buf, size_t len); 780 }; 781 782 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 783 enum xps_map_type { 784 XPS_CPUS = 0, 785 XPS_RXQS, 786 XPS_MAPS_MAX, 787 }; 788 789 #ifdef CONFIG_XPS 790 /* 791 * This structure holds an XPS map which can be of variable length. The 792 * map is an array of queues. 793 */ 794 struct xps_map { 795 unsigned int len; 796 unsigned int alloc_len; 797 struct rcu_head rcu; 798 u16 queues[]; 799 }; 800 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 801 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 802 - sizeof(struct xps_map)) / sizeof(u16)) 803 804 /* 805 * This structure holds all XPS maps for device. Maps are indexed by CPU. 806 * 807 * We keep track of the number of cpus/rxqs used when the struct is allocated, 808 * in nr_ids. This will help not accessing out-of-bound memory. 809 * 810 * We keep track of the number of traffic classes used when the struct is 811 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 812 * not crossing its upper bound, as the original dev->num_tc can be updated in 813 * the meantime. 814 */ 815 struct xps_dev_maps { 816 struct rcu_head rcu; 817 unsigned int nr_ids; 818 s16 num_tc; 819 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 820 }; 821 822 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 823 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 824 825 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 826 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 827 828 #endif /* CONFIG_XPS */ 829 830 #define TC_MAX_QUEUE 16 831 #define TC_BITMASK 15 832 /* HW offloaded queuing disciplines txq count and offset maps */ 833 struct netdev_tc_txq { 834 u16 count; 835 u16 offset; 836 }; 837 838 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 839 /* 840 * This structure is to hold information about the device 841 * configured to run FCoE protocol stack. 842 */ 843 struct netdev_fcoe_hbainfo { 844 char manufacturer[64]; 845 char serial_number[64]; 846 char hardware_version[64]; 847 char driver_version[64]; 848 char optionrom_version[64]; 849 char firmware_version[64]; 850 char model[256]; 851 char model_description[256]; 852 }; 853 #endif 854 855 #define MAX_PHYS_ITEM_ID_LEN 32 856 857 /* This structure holds a unique identifier to identify some 858 * physical item (port for example) used by a netdevice. 859 */ 860 struct netdev_phys_item_id { 861 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 862 unsigned char id_len; 863 }; 864 865 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 866 struct netdev_phys_item_id *b) 867 { 868 return a->id_len == b->id_len && 869 memcmp(a->id, b->id, a->id_len) == 0; 870 } 871 872 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 873 struct sk_buff *skb, 874 struct net_device *sb_dev); 875 876 enum net_device_path_type { 877 DEV_PATH_ETHERNET = 0, 878 DEV_PATH_VLAN, 879 DEV_PATH_BRIDGE, 880 DEV_PATH_PPPOE, 881 DEV_PATH_DSA, 882 DEV_PATH_MTK_WDMA, 883 }; 884 885 struct net_device_path { 886 enum net_device_path_type type; 887 const struct net_device *dev; 888 union { 889 struct { 890 u16 id; 891 __be16 proto; 892 u8 h_dest[ETH_ALEN]; 893 } encap; 894 struct { 895 enum { 896 DEV_PATH_BR_VLAN_KEEP, 897 DEV_PATH_BR_VLAN_TAG, 898 DEV_PATH_BR_VLAN_UNTAG, 899 DEV_PATH_BR_VLAN_UNTAG_HW, 900 } vlan_mode; 901 u16 vlan_id; 902 __be16 vlan_proto; 903 } bridge; 904 struct { 905 int port; 906 u16 proto; 907 } dsa; 908 struct { 909 u8 wdma_idx; 910 u8 queue; 911 u16 wcid; 912 u8 bss; 913 } mtk_wdma; 914 }; 915 }; 916 917 #define NET_DEVICE_PATH_STACK_MAX 5 918 #define NET_DEVICE_PATH_VLAN_MAX 2 919 920 struct net_device_path_stack { 921 int num_paths; 922 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 923 }; 924 925 struct net_device_path_ctx { 926 const struct net_device *dev; 927 u8 daddr[ETH_ALEN]; 928 929 int num_vlans; 930 struct { 931 u16 id; 932 __be16 proto; 933 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 934 }; 935 936 enum tc_setup_type { 937 TC_SETUP_QDISC_MQPRIO, 938 TC_SETUP_CLSU32, 939 TC_SETUP_CLSFLOWER, 940 TC_SETUP_CLSMATCHALL, 941 TC_SETUP_CLSBPF, 942 TC_SETUP_BLOCK, 943 TC_SETUP_QDISC_CBS, 944 TC_SETUP_QDISC_RED, 945 TC_SETUP_QDISC_PRIO, 946 TC_SETUP_QDISC_MQ, 947 TC_SETUP_QDISC_ETF, 948 TC_SETUP_ROOT_QDISC, 949 TC_SETUP_QDISC_GRED, 950 TC_SETUP_QDISC_TAPRIO, 951 TC_SETUP_FT, 952 TC_SETUP_QDISC_ETS, 953 TC_SETUP_QDISC_TBF, 954 TC_SETUP_QDISC_FIFO, 955 TC_SETUP_QDISC_HTB, 956 TC_SETUP_ACT, 957 }; 958 959 /* These structures hold the attributes of bpf state that are being passed 960 * to the netdevice through the bpf op. 961 */ 962 enum bpf_netdev_command { 963 /* Set or clear a bpf program used in the earliest stages of packet 964 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 965 * is responsible for calling bpf_prog_put on any old progs that are 966 * stored. In case of error, the callee need not release the new prog 967 * reference, but on success it takes ownership and must bpf_prog_put 968 * when it is no longer used. 969 */ 970 XDP_SETUP_PROG, 971 XDP_SETUP_PROG_HW, 972 /* BPF program for offload callbacks, invoked at program load time. */ 973 BPF_OFFLOAD_MAP_ALLOC, 974 BPF_OFFLOAD_MAP_FREE, 975 XDP_SETUP_XSK_POOL, 976 }; 977 978 struct bpf_prog_offload_ops; 979 struct netlink_ext_ack; 980 struct xdp_umem; 981 struct xdp_dev_bulk_queue; 982 struct bpf_xdp_link; 983 984 enum bpf_xdp_mode { 985 XDP_MODE_SKB = 0, 986 XDP_MODE_DRV = 1, 987 XDP_MODE_HW = 2, 988 __MAX_XDP_MODE 989 }; 990 991 struct bpf_xdp_entity { 992 struct bpf_prog *prog; 993 struct bpf_xdp_link *link; 994 }; 995 996 struct netdev_bpf { 997 enum bpf_netdev_command command; 998 union { 999 /* XDP_SETUP_PROG */ 1000 struct { 1001 u32 flags; 1002 struct bpf_prog *prog; 1003 struct netlink_ext_ack *extack; 1004 }; 1005 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 1006 struct { 1007 struct bpf_offloaded_map *offmap; 1008 }; 1009 /* XDP_SETUP_XSK_POOL */ 1010 struct { 1011 struct xsk_buff_pool *pool; 1012 u16 queue_id; 1013 } xsk; 1014 }; 1015 }; 1016 1017 /* Flags for ndo_xsk_wakeup. */ 1018 #define XDP_WAKEUP_RX (1 << 0) 1019 #define XDP_WAKEUP_TX (1 << 1) 1020 1021 #ifdef CONFIG_XFRM_OFFLOAD 1022 struct xfrmdev_ops { 1023 int (*xdo_dev_state_add) (struct xfrm_state *x); 1024 void (*xdo_dev_state_delete) (struct xfrm_state *x); 1025 void (*xdo_dev_state_free) (struct xfrm_state *x); 1026 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 1027 struct xfrm_state *x); 1028 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1029 }; 1030 #endif 1031 1032 struct dev_ifalias { 1033 struct rcu_head rcuhead; 1034 char ifalias[]; 1035 }; 1036 1037 struct devlink; 1038 struct tlsdev_ops; 1039 1040 struct netdev_net_notifier { 1041 struct list_head list; 1042 struct notifier_block *nb; 1043 }; 1044 1045 /* 1046 * This structure defines the management hooks for network devices. 1047 * The following hooks can be defined; unless noted otherwise, they are 1048 * optional and can be filled with a null pointer. 1049 * 1050 * int (*ndo_init)(struct net_device *dev); 1051 * This function is called once when a network device is registered. 1052 * The network device can use this for any late stage initialization 1053 * or semantic validation. It can fail with an error code which will 1054 * be propagated back to register_netdev. 1055 * 1056 * void (*ndo_uninit)(struct net_device *dev); 1057 * This function is called when device is unregistered or when registration 1058 * fails. It is not called if init fails. 1059 * 1060 * int (*ndo_open)(struct net_device *dev); 1061 * This function is called when a network device transitions to the up 1062 * state. 1063 * 1064 * int (*ndo_stop)(struct net_device *dev); 1065 * This function is called when a network device transitions to the down 1066 * state. 1067 * 1068 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1069 * struct net_device *dev); 1070 * Called when a packet needs to be transmitted. 1071 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1072 * the queue before that can happen; it's for obsolete devices and weird 1073 * corner cases, but the stack really does a non-trivial amount 1074 * of useless work if you return NETDEV_TX_BUSY. 1075 * Required; cannot be NULL. 1076 * 1077 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1078 * struct net_device *dev 1079 * netdev_features_t features); 1080 * Called by core transmit path to determine if device is capable of 1081 * performing offload operations on a given packet. This is to give 1082 * the device an opportunity to implement any restrictions that cannot 1083 * be otherwise expressed by feature flags. The check is called with 1084 * the set of features that the stack has calculated and it returns 1085 * those the driver believes to be appropriate. 1086 * 1087 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1088 * struct net_device *sb_dev); 1089 * Called to decide which queue to use when device supports multiple 1090 * transmit queues. 1091 * 1092 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1093 * This function is called to allow device receiver to make 1094 * changes to configuration when multicast or promiscuous is enabled. 1095 * 1096 * void (*ndo_set_rx_mode)(struct net_device *dev); 1097 * This function is called device changes address list filtering. 1098 * If driver handles unicast address filtering, it should set 1099 * IFF_UNICAST_FLT in its priv_flags. 1100 * 1101 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1102 * This function is called when the Media Access Control address 1103 * needs to be changed. If this interface is not defined, the 1104 * MAC address can not be changed. 1105 * 1106 * int (*ndo_validate_addr)(struct net_device *dev); 1107 * Test if Media Access Control address is valid for the device. 1108 * 1109 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1110 * Old-style ioctl entry point. This is used internally by the 1111 * appletalk and ieee802154 subsystems but is no longer called by 1112 * the device ioctl handler. 1113 * 1114 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1115 * Used by the bonding driver for its device specific ioctls: 1116 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1117 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1118 * 1119 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1120 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1121 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1122 * 1123 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1124 * Used to set network devices bus interface parameters. This interface 1125 * is retained for legacy reasons; new devices should use the bus 1126 * interface (PCI) for low level management. 1127 * 1128 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1129 * Called when a user wants to change the Maximum Transfer Unit 1130 * of a device. 1131 * 1132 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1133 * Callback used when the transmitter has not made any progress 1134 * for dev->watchdog ticks. 1135 * 1136 * void (*ndo_get_stats64)(struct net_device *dev, 1137 * struct rtnl_link_stats64 *storage); 1138 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1139 * Called when a user wants to get the network device usage 1140 * statistics. Drivers must do one of the following: 1141 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1142 * rtnl_link_stats64 structure passed by the caller. 1143 * 2. Define @ndo_get_stats to update a net_device_stats structure 1144 * (which should normally be dev->stats) and return a pointer to 1145 * it. The structure may be changed asynchronously only if each 1146 * field is written atomically. 1147 * 3. Update dev->stats asynchronously and atomically, and define 1148 * neither operation. 1149 * 1150 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1151 * Return true if this device supports offload stats of this attr_id. 1152 * 1153 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1154 * void *attr_data) 1155 * Get statistics for offload operations by attr_id. Write it into the 1156 * attr_data pointer. 1157 * 1158 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1159 * If device supports VLAN filtering this function is called when a 1160 * VLAN id is registered. 1161 * 1162 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1163 * If device supports VLAN filtering this function is called when a 1164 * VLAN id is unregistered. 1165 * 1166 * void (*ndo_poll_controller)(struct net_device *dev); 1167 * 1168 * SR-IOV management functions. 1169 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1170 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1171 * u8 qos, __be16 proto); 1172 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1173 * int max_tx_rate); 1174 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1175 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1176 * int (*ndo_get_vf_config)(struct net_device *dev, 1177 * int vf, struct ifla_vf_info *ivf); 1178 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1179 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1180 * struct nlattr *port[]); 1181 * 1182 * Enable or disable the VF ability to query its RSS Redirection Table and 1183 * Hash Key. This is needed since on some devices VF share this information 1184 * with PF and querying it may introduce a theoretical security risk. 1185 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1186 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1187 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1188 * void *type_data); 1189 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1190 * This is always called from the stack with the rtnl lock held and netif 1191 * tx queues stopped. This allows the netdevice to perform queue 1192 * management safely. 1193 * 1194 * Fiber Channel over Ethernet (FCoE) offload functions. 1195 * int (*ndo_fcoe_enable)(struct net_device *dev); 1196 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1197 * so the underlying device can perform whatever needed configuration or 1198 * initialization to support acceleration of FCoE traffic. 1199 * 1200 * int (*ndo_fcoe_disable)(struct net_device *dev); 1201 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1202 * so the underlying device can perform whatever needed clean-ups to 1203 * stop supporting acceleration of FCoE traffic. 1204 * 1205 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1206 * struct scatterlist *sgl, unsigned int sgc); 1207 * Called when the FCoE Initiator wants to initialize an I/O that 1208 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1209 * perform necessary setup and returns 1 to indicate the device is set up 1210 * successfully to perform DDP on this I/O, otherwise this returns 0. 1211 * 1212 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1213 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1214 * indicated by the FC exchange id 'xid', so the underlying device can 1215 * clean up and reuse resources for later DDP requests. 1216 * 1217 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1218 * struct scatterlist *sgl, unsigned int sgc); 1219 * Called when the FCoE Target wants to initialize an I/O that 1220 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1221 * perform necessary setup and returns 1 to indicate the device is set up 1222 * successfully to perform DDP on this I/O, otherwise this returns 0. 1223 * 1224 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1225 * struct netdev_fcoe_hbainfo *hbainfo); 1226 * Called when the FCoE Protocol stack wants information on the underlying 1227 * device. This information is utilized by the FCoE protocol stack to 1228 * register attributes with Fiber Channel management service as per the 1229 * FC-GS Fabric Device Management Information(FDMI) specification. 1230 * 1231 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1232 * Called when the underlying device wants to override default World Wide 1233 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1234 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1235 * protocol stack to use. 1236 * 1237 * RFS acceleration. 1238 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1239 * u16 rxq_index, u32 flow_id); 1240 * Set hardware filter for RFS. rxq_index is the target queue index; 1241 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1242 * Return the filter ID on success, or a negative error code. 1243 * 1244 * Slave management functions (for bridge, bonding, etc). 1245 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1246 * Called to make another netdev an underling. 1247 * 1248 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1249 * Called to release previously enslaved netdev. 1250 * 1251 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1252 * struct sk_buff *skb, 1253 * bool all_slaves); 1254 * Get the xmit slave of master device. If all_slaves is true, function 1255 * assume all the slaves can transmit. 1256 * 1257 * Feature/offload setting functions. 1258 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1259 * netdev_features_t features); 1260 * Adjusts the requested feature flags according to device-specific 1261 * constraints, and returns the resulting flags. Must not modify 1262 * the device state. 1263 * 1264 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1265 * Called to update device configuration to new features. Passed 1266 * feature set might be less than what was returned by ndo_fix_features()). 1267 * Must return >0 or -errno if it changed dev->features itself. 1268 * 1269 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1270 * struct net_device *dev, 1271 * const unsigned char *addr, u16 vid, u16 flags, 1272 * struct netlink_ext_ack *extack); 1273 * Adds an FDB entry to dev for addr. 1274 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1275 * struct net_device *dev, 1276 * const unsigned char *addr, u16 vid) 1277 * Deletes the FDB entry from dev coresponding to addr. 1278 * int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, struct nlattr *tb[], 1279 * struct net_device *dev, 1280 * u16 vid, 1281 * struct netlink_ext_ack *extack); 1282 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1283 * struct net_device *dev, struct net_device *filter_dev, 1284 * int *idx) 1285 * Used to add FDB entries to dump requests. Implementers should add 1286 * entries to skb and update idx with the number of entries. 1287 * 1288 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1289 * u16 flags, struct netlink_ext_ack *extack) 1290 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1291 * struct net_device *dev, u32 filter_mask, 1292 * int nlflags) 1293 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1294 * u16 flags); 1295 * 1296 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1297 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1298 * which do not represent real hardware may define this to allow their 1299 * userspace components to manage their virtual carrier state. Devices 1300 * that determine carrier state from physical hardware properties (eg 1301 * network cables) or protocol-dependent mechanisms (eg 1302 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1303 * 1304 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1305 * struct netdev_phys_item_id *ppid); 1306 * Called to get ID of physical port of this device. If driver does 1307 * not implement this, it is assumed that the hw is not able to have 1308 * multiple net devices on single physical port. 1309 * 1310 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1311 * struct netdev_phys_item_id *ppid) 1312 * Called to get the parent ID of the physical port of this device. 1313 * 1314 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1315 * struct net_device *dev) 1316 * Called by upper layer devices to accelerate switching or other 1317 * station functionality into hardware. 'pdev is the lowerdev 1318 * to use for the offload and 'dev' is the net device that will 1319 * back the offload. Returns a pointer to the private structure 1320 * the upper layer will maintain. 1321 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1322 * Called by upper layer device to delete the station created 1323 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1324 * the station and priv is the structure returned by the add 1325 * operation. 1326 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1327 * int queue_index, u32 maxrate); 1328 * Called when a user wants to set a max-rate limitation of specific 1329 * TX queue. 1330 * int (*ndo_get_iflink)(const struct net_device *dev); 1331 * Called to get the iflink value of this device. 1332 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1333 * This function is used to get egress tunnel information for given skb. 1334 * This is useful for retrieving outer tunnel header parameters while 1335 * sampling packet. 1336 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1337 * This function is used to specify the headroom that the skb must 1338 * consider when allocation skb during packet reception. Setting 1339 * appropriate rx headroom value allows avoiding skb head copy on 1340 * forward. Setting a negative value resets the rx headroom to the 1341 * default value. 1342 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1343 * This function is used to set or query state related to XDP on the 1344 * netdevice and manage BPF offload. See definition of 1345 * enum bpf_netdev_command for details. 1346 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1347 * u32 flags); 1348 * This function is used to submit @n XDP packets for transmit on a 1349 * netdevice. Returns number of frames successfully transmitted, frames 1350 * that got dropped are freed/returned via xdp_return_frame(). 1351 * Returns negative number, means general error invoking ndo, meaning 1352 * no frames were xmit'ed and core-caller will free all frames. 1353 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1354 * struct xdp_buff *xdp); 1355 * Get the xmit slave of master device based on the xdp_buff. 1356 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1357 * This function is used to wake up the softirq, ksoftirqd or kthread 1358 * responsible for sending and/or receiving packets on a specific 1359 * queue id bound to an AF_XDP socket. The flags field specifies if 1360 * only RX, only Tx, or both should be woken up using the flags 1361 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1362 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1363 * Get devlink port instance associated with a given netdev. 1364 * Called with a reference on the netdevice and devlink locks only, 1365 * rtnl_lock is not held. 1366 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1367 * int cmd); 1368 * Add, change, delete or get information on an IPv4 tunnel. 1369 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1370 * If a device is paired with a peer device, return the peer instance. 1371 * The caller must be under RCU read context. 1372 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1373 * Get the forwarding path to reach the real device from the HW destination address 1374 * ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1375 * const struct skb_shared_hwtstamps *hwtstamps, 1376 * bool cycles); 1377 * Get hardware timestamp based on normal/adjustable time or free running 1378 * cycle counter. This function is required if physical clock supports a 1379 * free running cycle counter. 1380 */ 1381 struct net_device_ops { 1382 int (*ndo_init)(struct net_device *dev); 1383 void (*ndo_uninit)(struct net_device *dev); 1384 int (*ndo_open)(struct net_device *dev); 1385 int (*ndo_stop)(struct net_device *dev); 1386 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1387 struct net_device *dev); 1388 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1389 struct net_device *dev, 1390 netdev_features_t features); 1391 u16 (*ndo_select_queue)(struct net_device *dev, 1392 struct sk_buff *skb, 1393 struct net_device *sb_dev); 1394 void (*ndo_change_rx_flags)(struct net_device *dev, 1395 int flags); 1396 void (*ndo_set_rx_mode)(struct net_device *dev); 1397 int (*ndo_set_mac_address)(struct net_device *dev, 1398 void *addr); 1399 int (*ndo_validate_addr)(struct net_device *dev); 1400 int (*ndo_do_ioctl)(struct net_device *dev, 1401 struct ifreq *ifr, int cmd); 1402 int (*ndo_eth_ioctl)(struct net_device *dev, 1403 struct ifreq *ifr, int cmd); 1404 int (*ndo_siocbond)(struct net_device *dev, 1405 struct ifreq *ifr, int cmd); 1406 int (*ndo_siocwandev)(struct net_device *dev, 1407 struct if_settings *ifs); 1408 int (*ndo_siocdevprivate)(struct net_device *dev, 1409 struct ifreq *ifr, 1410 void __user *data, int cmd); 1411 int (*ndo_set_config)(struct net_device *dev, 1412 struct ifmap *map); 1413 int (*ndo_change_mtu)(struct net_device *dev, 1414 int new_mtu); 1415 int (*ndo_neigh_setup)(struct net_device *dev, 1416 struct neigh_parms *); 1417 void (*ndo_tx_timeout) (struct net_device *dev, 1418 unsigned int txqueue); 1419 1420 void (*ndo_get_stats64)(struct net_device *dev, 1421 struct rtnl_link_stats64 *storage); 1422 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1423 int (*ndo_get_offload_stats)(int attr_id, 1424 const struct net_device *dev, 1425 void *attr_data); 1426 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1427 1428 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1429 __be16 proto, u16 vid); 1430 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1431 __be16 proto, u16 vid); 1432 #ifdef CONFIG_NET_POLL_CONTROLLER 1433 void (*ndo_poll_controller)(struct net_device *dev); 1434 int (*ndo_netpoll_setup)(struct net_device *dev, 1435 struct netpoll_info *info); 1436 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1437 #endif 1438 int (*ndo_set_vf_mac)(struct net_device *dev, 1439 int queue, u8 *mac); 1440 int (*ndo_set_vf_vlan)(struct net_device *dev, 1441 int queue, u16 vlan, 1442 u8 qos, __be16 proto); 1443 int (*ndo_set_vf_rate)(struct net_device *dev, 1444 int vf, int min_tx_rate, 1445 int max_tx_rate); 1446 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1447 int vf, bool setting); 1448 int (*ndo_set_vf_trust)(struct net_device *dev, 1449 int vf, bool setting); 1450 int (*ndo_get_vf_config)(struct net_device *dev, 1451 int vf, 1452 struct ifla_vf_info *ivf); 1453 int (*ndo_set_vf_link_state)(struct net_device *dev, 1454 int vf, int link_state); 1455 int (*ndo_get_vf_stats)(struct net_device *dev, 1456 int vf, 1457 struct ifla_vf_stats 1458 *vf_stats); 1459 int (*ndo_set_vf_port)(struct net_device *dev, 1460 int vf, 1461 struct nlattr *port[]); 1462 int (*ndo_get_vf_port)(struct net_device *dev, 1463 int vf, struct sk_buff *skb); 1464 int (*ndo_get_vf_guid)(struct net_device *dev, 1465 int vf, 1466 struct ifla_vf_guid *node_guid, 1467 struct ifla_vf_guid *port_guid); 1468 int (*ndo_set_vf_guid)(struct net_device *dev, 1469 int vf, u64 guid, 1470 int guid_type); 1471 int (*ndo_set_vf_rss_query_en)( 1472 struct net_device *dev, 1473 int vf, bool setting); 1474 int (*ndo_setup_tc)(struct net_device *dev, 1475 enum tc_setup_type type, 1476 void *type_data); 1477 #if IS_ENABLED(CONFIG_FCOE) 1478 int (*ndo_fcoe_enable)(struct net_device *dev); 1479 int (*ndo_fcoe_disable)(struct net_device *dev); 1480 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1481 u16 xid, 1482 struct scatterlist *sgl, 1483 unsigned int sgc); 1484 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1485 u16 xid); 1486 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1487 u16 xid, 1488 struct scatterlist *sgl, 1489 unsigned int sgc); 1490 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1491 struct netdev_fcoe_hbainfo *hbainfo); 1492 #endif 1493 1494 #if IS_ENABLED(CONFIG_LIBFCOE) 1495 #define NETDEV_FCOE_WWNN 0 1496 #define NETDEV_FCOE_WWPN 1 1497 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1498 u64 *wwn, int type); 1499 #endif 1500 1501 #ifdef CONFIG_RFS_ACCEL 1502 int (*ndo_rx_flow_steer)(struct net_device *dev, 1503 const struct sk_buff *skb, 1504 u16 rxq_index, 1505 u32 flow_id); 1506 #endif 1507 int (*ndo_add_slave)(struct net_device *dev, 1508 struct net_device *slave_dev, 1509 struct netlink_ext_ack *extack); 1510 int (*ndo_del_slave)(struct net_device *dev, 1511 struct net_device *slave_dev); 1512 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1513 struct sk_buff *skb, 1514 bool all_slaves); 1515 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1516 struct sock *sk); 1517 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1518 netdev_features_t features); 1519 int (*ndo_set_features)(struct net_device *dev, 1520 netdev_features_t features); 1521 int (*ndo_neigh_construct)(struct net_device *dev, 1522 struct neighbour *n); 1523 void (*ndo_neigh_destroy)(struct net_device *dev, 1524 struct neighbour *n); 1525 1526 int (*ndo_fdb_add)(struct ndmsg *ndm, 1527 struct nlattr *tb[], 1528 struct net_device *dev, 1529 const unsigned char *addr, 1530 u16 vid, 1531 u16 flags, 1532 struct netlink_ext_ack *extack); 1533 int (*ndo_fdb_del)(struct ndmsg *ndm, 1534 struct nlattr *tb[], 1535 struct net_device *dev, 1536 const unsigned char *addr, 1537 u16 vid, struct netlink_ext_ack *extack); 1538 int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, 1539 struct nlattr *tb[], 1540 struct net_device *dev, 1541 u16 vid, 1542 struct netlink_ext_ack *extack); 1543 int (*ndo_fdb_dump)(struct sk_buff *skb, 1544 struct netlink_callback *cb, 1545 struct net_device *dev, 1546 struct net_device *filter_dev, 1547 int *idx); 1548 int (*ndo_fdb_get)(struct sk_buff *skb, 1549 struct nlattr *tb[], 1550 struct net_device *dev, 1551 const unsigned char *addr, 1552 u16 vid, u32 portid, u32 seq, 1553 struct netlink_ext_ack *extack); 1554 int (*ndo_bridge_setlink)(struct net_device *dev, 1555 struct nlmsghdr *nlh, 1556 u16 flags, 1557 struct netlink_ext_ack *extack); 1558 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1559 u32 pid, u32 seq, 1560 struct net_device *dev, 1561 u32 filter_mask, 1562 int nlflags); 1563 int (*ndo_bridge_dellink)(struct net_device *dev, 1564 struct nlmsghdr *nlh, 1565 u16 flags); 1566 int (*ndo_change_carrier)(struct net_device *dev, 1567 bool new_carrier); 1568 int (*ndo_get_phys_port_id)(struct net_device *dev, 1569 struct netdev_phys_item_id *ppid); 1570 int (*ndo_get_port_parent_id)(struct net_device *dev, 1571 struct netdev_phys_item_id *ppid); 1572 int (*ndo_get_phys_port_name)(struct net_device *dev, 1573 char *name, size_t len); 1574 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1575 struct net_device *dev); 1576 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1577 void *priv); 1578 1579 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1580 int queue_index, 1581 u32 maxrate); 1582 int (*ndo_get_iflink)(const struct net_device *dev); 1583 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1584 struct sk_buff *skb); 1585 void (*ndo_set_rx_headroom)(struct net_device *dev, 1586 int needed_headroom); 1587 int (*ndo_bpf)(struct net_device *dev, 1588 struct netdev_bpf *bpf); 1589 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1590 struct xdp_frame **xdp, 1591 u32 flags); 1592 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1593 struct xdp_buff *xdp); 1594 int (*ndo_xsk_wakeup)(struct net_device *dev, 1595 u32 queue_id, u32 flags); 1596 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1597 int (*ndo_tunnel_ctl)(struct net_device *dev, 1598 struct ip_tunnel_parm *p, int cmd); 1599 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1600 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1601 struct net_device_path *path); 1602 ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1603 const struct skb_shared_hwtstamps *hwtstamps, 1604 bool cycles); 1605 }; 1606 1607 /** 1608 * enum netdev_priv_flags - &struct net_device priv_flags 1609 * 1610 * These are the &struct net_device, they are only set internally 1611 * by drivers and used in the kernel. These flags are invisible to 1612 * userspace; this means that the order of these flags can change 1613 * during any kernel release. 1614 * 1615 * You should have a pretty good reason to be extending these flags. 1616 * 1617 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1618 * @IFF_EBRIDGE: Ethernet bridging device 1619 * @IFF_BONDING: bonding master or slave 1620 * @IFF_ISATAP: ISATAP interface (RFC4214) 1621 * @IFF_WAN_HDLC: WAN HDLC device 1622 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1623 * release skb->dst 1624 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1625 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1626 * @IFF_MACVLAN_PORT: device used as macvlan port 1627 * @IFF_BRIDGE_PORT: device used as bridge port 1628 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1629 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1630 * @IFF_UNICAST_FLT: Supports unicast filtering 1631 * @IFF_TEAM_PORT: device used as team port 1632 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1633 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1634 * change when it's running 1635 * @IFF_MACVLAN: Macvlan device 1636 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1637 * underlying stacked devices 1638 * @IFF_L3MDEV_MASTER: device is an L3 master device 1639 * @IFF_NO_QUEUE: device can run without qdisc attached 1640 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1641 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1642 * @IFF_TEAM: device is a team device 1643 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1644 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1645 * entity (i.e. the master device for bridged veth) 1646 * @IFF_MACSEC: device is a MACsec device 1647 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1648 * @IFF_FAILOVER: device is a failover master device 1649 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1650 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1651 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1652 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1653 * skb_headlen(skb) == 0 (data starts from frag0) 1654 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN 1655 */ 1656 enum netdev_priv_flags { 1657 IFF_802_1Q_VLAN = 1<<0, 1658 IFF_EBRIDGE = 1<<1, 1659 IFF_BONDING = 1<<2, 1660 IFF_ISATAP = 1<<3, 1661 IFF_WAN_HDLC = 1<<4, 1662 IFF_XMIT_DST_RELEASE = 1<<5, 1663 IFF_DONT_BRIDGE = 1<<6, 1664 IFF_DISABLE_NETPOLL = 1<<7, 1665 IFF_MACVLAN_PORT = 1<<8, 1666 IFF_BRIDGE_PORT = 1<<9, 1667 IFF_OVS_DATAPATH = 1<<10, 1668 IFF_TX_SKB_SHARING = 1<<11, 1669 IFF_UNICAST_FLT = 1<<12, 1670 IFF_TEAM_PORT = 1<<13, 1671 IFF_SUPP_NOFCS = 1<<14, 1672 IFF_LIVE_ADDR_CHANGE = 1<<15, 1673 IFF_MACVLAN = 1<<16, 1674 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1675 IFF_L3MDEV_MASTER = 1<<18, 1676 IFF_NO_QUEUE = 1<<19, 1677 IFF_OPENVSWITCH = 1<<20, 1678 IFF_L3MDEV_SLAVE = 1<<21, 1679 IFF_TEAM = 1<<22, 1680 IFF_RXFH_CONFIGURED = 1<<23, 1681 IFF_PHONY_HEADROOM = 1<<24, 1682 IFF_MACSEC = 1<<25, 1683 IFF_NO_RX_HANDLER = 1<<26, 1684 IFF_FAILOVER = 1<<27, 1685 IFF_FAILOVER_SLAVE = 1<<28, 1686 IFF_L3MDEV_RX_HANDLER = 1<<29, 1687 IFF_LIVE_RENAME_OK = 1<<30, 1688 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31), 1689 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32), 1690 }; 1691 1692 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1693 #define IFF_EBRIDGE IFF_EBRIDGE 1694 #define IFF_BONDING IFF_BONDING 1695 #define IFF_ISATAP IFF_ISATAP 1696 #define IFF_WAN_HDLC IFF_WAN_HDLC 1697 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1698 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1699 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1700 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1701 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1702 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1703 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1704 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1705 #define IFF_TEAM_PORT IFF_TEAM_PORT 1706 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1707 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1708 #define IFF_MACVLAN IFF_MACVLAN 1709 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1710 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1711 #define IFF_NO_QUEUE IFF_NO_QUEUE 1712 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1713 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1714 #define IFF_TEAM IFF_TEAM 1715 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1716 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM 1717 #define IFF_MACSEC IFF_MACSEC 1718 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1719 #define IFF_FAILOVER IFF_FAILOVER 1720 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1721 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1722 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1723 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR 1724 1725 /* Specifies the type of the struct net_device::ml_priv pointer */ 1726 enum netdev_ml_priv_type { 1727 ML_PRIV_NONE, 1728 ML_PRIV_CAN, 1729 }; 1730 1731 /** 1732 * struct net_device - The DEVICE structure. 1733 * 1734 * Actually, this whole structure is a big mistake. It mixes I/O 1735 * data with strictly "high-level" data, and it has to know about 1736 * almost every data structure used in the INET module. 1737 * 1738 * @name: This is the first field of the "visible" part of this structure 1739 * (i.e. as seen by users in the "Space.c" file). It is the name 1740 * of the interface. 1741 * 1742 * @name_node: Name hashlist node 1743 * @ifalias: SNMP alias 1744 * @mem_end: Shared memory end 1745 * @mem_start: Shared memory start 1746 * @base_addr: Device I/O address 1747 * @irq: Device IRQ number 1748 * 1749 * @state: Generic network queuing layer state, see netdev_state_t 1750 * @dev_list: The global list of network devices 1751 * @napi_list: List entry used for polling NAPI devices 1752 * @unreg_list: List entry when we are unregistering the 1753 * device; see the function unregister_netdev 1754 * @close_list: List entry used when we are closing the device 1755 * @ptype_all: Device-specific packet handlers for all protocols 1756 * @ptype_specific: Device-specific, protocol-specific packet handlers 1757 * 1758 * @adj_list: Directly linked devices, like slaves for bonding 1759 * @features: Currently active device features 1760 * @hw_features: User-changeable features 1761 * 1762 * @wanted_features: User-requested features 1763 * @vlan_features: Mask of features inheritable by VLAN devices 1764 * 1765 * @hw_enc_features: Mask of features inherited by encapsulating devices 1766 * This field indicates what encapsulation 1767 * offloads the hardware is capable of doing, 1768 * and drivers will need to set them appropriately. 1769 * 1770 * @mpls_features: Mask of features inheritable by MPLS 1771 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1772 * 1773 * @ifindex: interface index 1774 * @group: The group the device belongs to 1775 * 1776 * @stats: Statistics struct, which was left as a legacy, use 1777 * rtnl_link_stats64 instead 1778 * 1779 * @core_stats: core networking counters, 1780 * do not use this in drivers 1781 * @carrier_up_count: Number of times the carrier has been up 1782 * @carrier_down_count: Number of times the carrier has been down 1783 * 1784 * @wireless_handlers: List of functions to handle Wireless Extensions, 1785 * instead of ioctl, 1786 * see <net/iw_handler.h> for details. 1787 * @wireless_data: Instance data managed by the core of wireless extensions 1788 * 1789 * @netdev_ops: Includes several pointers to callbacks, 1790 * if one wants to override the ndo_*() functions 1791 * @ethtool_ops: Management operations 1792 * @l3mdev_ops: Layer 3 master device operations 1793 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1794 * discovery handling. Necessary for e.g. 6LoWPAN. 1795 * @xfrmdev_ops: Transformation offload operations 1796 * @tlsdev_ops: Transport Layer Security offload operations 1797 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1798 * of Layer 2 headers. 1799 * 1800 * @flags: Interface flags (a la BSD) 1801 * @priv_flags: Like 'flags' but invisible to userspace, 1802 * see if.h for the definitions 1803 * @gflags: Global flags ( kept as legacy ) 1804 * @padded: How much padding added by alloc_netdev() 1805 * @operstate: RFC2863 operstate 1806 * @link_mode: Mapping policy to operstate 1807 * @if_port: Selectable AUI, TP, ... 1808 * @dma: DMA channel 1809 * @mtu: Interface MTU value 1810 * @min_mtu: Interface Minimum MTU value 1811 * @max_mtu: Interface Maximum MTU value 1812 * @type: Interface hardware type 1813 * @hard_header_len: Maximum hardware header length. 1814 * @min_header_len: Minimum hardware header length 1815 * 1816 * @needed_headroom: Extra headroom the hardware may need, but not in all 1817 * cases can this be guaranteed 1818 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1819 * cases can this be guaranteed. Some cases also use 1820 * LL_MAX_HEADER instead to allocate the skb 1821 * 1822 * interface address info: 1823 * 1824 * @perm_addr: Permanent hw address 1825 * @addr_assign_type: Hw address assignment type 1826 * @addr_len: Hardware address length 1827 * @upper_level: Maximum depth level of upper devices. 1828 * @lower_level: Maximum depth level of lower devices. 1829 * @neigh_priv_len: Used in neigh_alloc() 1830 * @dev_id: Used to differentiate devices that share 1831 * the same link layer address 1832 * @dev_port: Used to differentiate devices that share 1833 * the same function 1834 * @addr_list_lock: XXX: need comments on this one 1835 * @name_assign_type: network interface name assignment type 1836 * @uc_promisc: Counter that indicates promiscuous mode 1837 * has been enabled due to the need to listen to 1838 * additional unicast addresses in a device that 1839 * does not implement ndo_set_rx_mode() 1840 * @uc: unicast mac addresses 1841 * @mc: multicast mac addresses 1842 * @dev_addrs: list of device hw addresses 1843 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1844 * @promiscuity: Number of times the NIC is told to work in 1845 * promiscuous mode; if it becomes 0 the NIC will 1846 * exit promiscuous mode 1847 * @allmulti: Counter, enables or disables allmulticast mode 1848 * 1849 * @vlan_info: VLAN info 1850 * @dsa_ptr: dsa specific data 1851 * @tipc_ptr: TIPC specific data 1852 * @atalk_ptr: AppleTalk link 1853 * @ip_ptr: IPv4 specific data 1854 * @ip6_ptr: IPv6 specific data 1855 * @ax25_ptr: AX.25 specific data 1856 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1857 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1858 * device struct 1859 * @mpls_ptr: mpls_dev struct pointer 1860 * @mctp_ptr: MCTP specific data 1861 * 1862 * @dev_addr: Hw address (before bcast, 1863 * because most packets are unicast) 1864 * 1865 * @_rx: Array of RX queues 1866 * @num_rx_queues: Number of RX queues 1867 * allocated at register_netdev() time 1868 * @real_num_rx_queues: Number of RX queues currently active in device 1869 * @xdp_prog: XDP sockets filter program pointer 1870 * @gro_flush_timeout: timeout for GRO layer in NAPI 1871 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1872 * allow to avoid NIC hard IRQ, on busy queues. 1873 * 1874 * @rx_handler: handler for received packets 1875 * @rx_handler_data: XXX: need comments on this one 1876 * @miniq_ingress: ingress/clsact qdisc specific data for 1877 * ingress processing 1878 * @ingress_queue: XXX: need comments on this one 1879 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1880 * @broadcast: hw bcast address 1881 * 1882 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1883 * indexed by RX queue number. Assigned by driver. 1884 * This must only be set if the ndo_rx_flow_steer 1885 * operation is defined 1886 * @index_hlist: Device index hash chain 1887 * 1888 * @_tx: Array of TX queues 1889 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1890 * @real_num_tx_queues: Number of TX queues currently active in device 1891 * @qdisc: Root qdisc from userspace point of view 1892 * @tx_queue_len: Max frames per queue allowed 1893 * @tx_global_lock: XXX: need comments on this one 1894 * @xdp_bulkq: XDP device bulk queue 1895 * @xps_maps: all CPUs/RXQs maps for XPS device 1896 * 1897 * @xps_maps: XXX: need comments on this one 1898 * @miniq_egress: clsact qdisc specific data for 1899 * egress processing 1900 * @nf_hooks_egress: netfilter hooks executed for egress packets 1901 * @qdisc_hash: qdisc hash table 1902 * @watchdog_timeo: Represents the timeout that is used by 1903 * the watchdog (see dev_watchdog()) 1904 * @watchdog_timer: List of timers 1905 * 1906 * @proto_down_reason: reason a netdev interface is held down 1907 * @pcpu_refcnt: Number of references to this device 1908 * @dev_refcnt: Number of references to this device 1909 * @refcnt_tracker: Tracker directory for tracked references to this device 1910 * @todo_list: Delayed register/unregister 1911 * @link_watch_list: XXX: need comments on this one 1912 * 1913 * @reg_state: Register/unregister state machine 1914 * @dismantle: Device is going to be freed 1915 * @rtnl_link_state: This enum represents the phases of creating 1916 * a new link 1917 * 1918 * @needs_free_netdev: Should unregister perform free_netdev? 1919 * @priv_destructor: Called from unregister 1920 * @npinfo: XXX: need comments on this one 1921 * @nd_net: Network namespace this network device is inside 1922 * 1923 * @ml_priv: Mid-layer private 1924 * @ml_priv_type: Mid-layer private type 1925 * @lstats: Loopback statistics 1926 * @tstats: Tunnel statistics 1927 * @dstats: Dummy statistics 1928 * @vstats: Virtual ethernet statistics 1929 * 1930 * @garp_port: GARP 1931 * @mrp_port: MRP 1932 * 1933 * @dm_private: Drop monitor private 1934 * 1935 * @dev: Class/net/name entry 1936 * @sysfs_groups: Space for optional device, statistics and wireless 1937 * sysfs groups 1938 * 1939 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1940 * @rtnl_link_ops: Rtnl_link_ops 1941 * 1942 * @gso_max_size: Maximum size of generic segmentation offload 1943 * @tso_max_size: Device (as in HW) limit on the max TSO request size 1944 * @gso_max_segs: Maximum number of segments that can be passed to the 1945 * NIC for GSO 1946 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count 1947 * 1948 * @dcbnl_ops: Data Center Bridging netlink ops 1949 * @num_tc: Number of traffic classes in the net device 1950 * @tc_to_txq: XXX: need comments on this one 1951 * @prio_tc_map: XXX: need comments on this one 1952 * 1953 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1954 * 1955 * @priomap: XXX: need comments on this one 1956 * @phydev: Physical device may attach itself 1957 * for hardware timestamping 1958 * @sfp_bus: attached &struct sfp_bus structure. 1959 * 1960 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1961 * 1962 * @proto_down: protocol port state information can be sent to the 1963 * switch driver and used to set the phys state of the 1964 * switch port. 1965 * 1966 * @wol_enabled: Wake-on-LAN is enabled 1967 * 1968 * @threaded: napi threaded mode is enabled 1969 * 1970 * @net_notifier_list: List of per-net netdev notifier block 1971 * that follow this device when it is moved 1972 * to another network namespace. 1973 * 1974 * @macsec_ops: MACsec offloading ops 1975 * 1976 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1977 * offload capabilities of the device 1978 * @udp_tunnel_nic: UDP tunnel offload state 1979 * @xdp_state: stores info on attached XDP BPF programs 1980 * 1981 * @nested_level: Used as a parameter of spin_lock_nested() of 1982 * dev->addr_list_lock. 1983 * @unlink_list: As netif_addr_lock() can be called recursively, 1984 * keep a list of interfaces to be deleted. 1985 * @gro_max_size: Maximum size of aggregated packet in generic 1986 * receive offload (GRO) 1987 * 1988 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 1989 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 1990 * @watchdog_dev_tracker: refcount tracker used by watchdog. 1991 * @dev_registered_tracker: tracker for reference held while 1992 * registered 1993 * @offload_xstats_l3: L3 HW stats for this netdevice. 1994 * 1995 * FIXME: cleanup struct net_device such that network protocol info 1996 * moves out. 1997 */ 1998 1999 struct net_device { 2000 char name[IFNAMSIZ]; 2001 struct netdev_name_node *name_node; 2002 struct dev_ifalias __rcu *ifalias; 2003 /* 2004 * I/O specific fields 2005 * FIXME: Merge these and struct ifmap into one 2006 */ 2007 unsigned long mem_end; 2008 unsigned long mem_start; 2009 unsigned long base_addr; 2010 2011 /* 2012 * Some hardware also needs these fields (state,dev_list, 2013 * napi_list,unreg_list,close_list) but they are not 2014 * part of the usual set specified in Space.c. 2015 */ 2016 2017 unsigned long state; 2018 2019 struct list_head dev_list; 2020 struct list_head napi_list; 2021 struct list_head unreg_list; 2022 struct list_head close_list; 2023 struct list_head ptype_all; 2024 struct list_head ptype_specific; 2025 2026 struct { 2027 struct list_head upper; 2028 struct list_head lower; 2029 } adj_list; 2030 2031 /* Read-mostly cache-line for fast-path access */ 2032 unsigned int flags; 2033 unsigned long long priv_flags; 2034 const struct net_device_ops *netdev_ops; 2035 int ifindex; 2036 unsigned short gflags; 2037 unsigned short hard_header_len; 2038 2039 /* Note : dev->mtu is often read without holding a lock. 2040 * Writers usually hold RTNL. 2041 * It is recommended to use READ_ONCE() to annotate the reads, 2042 * and to use WRITE_ONCE() to annotate the writes. 2043 */ 2044 unsigned int mtu; 2045 unsigned short needed_headroom; 2046 unsigned short needed_tailroom; 2047 2048 netdev_features_t features; 2049 netdev_features_t hw_features; 2050 netdev_features_t wanted_features; 2051 netdev_features_t vlan_features; 2052 netdev_features_t hw_enc_features; 2053 netdev_features_t mpls_features; 2054 netdev_features_t gso_partial_features; 2055 2056 unsigned int min_mtu; 2057 unsigned int max_mtu; 2058 unsigned short type; 2059 unsigned char min_header_len; 2060 unsigned char name_assign_type; 2061 2062 int group; 2063 2064 struct net_device_stats stats; /* not used by modern drivers */ 2065 2066 struct net_device_core_stats __percpu *core_stats; 2067 2068 /* Stats to monitor link on/off, flapping */ 2069 atomic_t carrier_up_count; 2070 atomic_t carrier_down_count; 2071 2072 #ifdef CONFIG_WIRELESS_EXT 2073 const struct iw_handler_def *wireless_handlers; 2074 struct iw_public_data *wireless_data; 2075 #endif 2076 const struct ethtool_ops *ethtool_ops; 2077 #ifdef CONFIG_NET_L3_MASTER_DEV 2078 const struct l3mdev_ops *l3mdev_ops; 2079 #endif 2080 #if IS_ENABLED(CONFIG_IPV6) 2081 const struct ndisc_ops *ndisc_ops; 2082 #endif 2083 2084 #ifdef CONFIG_XFRM_OFFLOAD 2085 const struct xfrmdev_ops *xfrmdev_ops; 2086 #endif 2087 2088 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2089 const struct tlsdev_ops *tlsdev_ops; 2090 #endif 2091 2092 const struct header_ops *header_ops; 2093 2094 unsigned char operstate; 2095 unsigned char link_mode; 2096 2097 unsigned char if_port; 2098 unsigned char dma; 2099 2100 /* Interface address info. */ 2101 unsigned char perm_addr[MAX_ADDR_LEN]; 2102 unsigned char addr_assign_type; 2103 unsigned char addr_len; 2104 unsigned char upper_level; 2105 unsigned char lower_level; 2106 2107 unsigned short neigh_priv_len; 2108 unsigned short dev_id; 2109 unsigned short dev_port; 2110 unsigned short padded; 2111 2112 spinlock_t addr_list_lock; 2113 int irq; 2114 2115 struct netdev_hw_addr_list uc; 2116 struct netdev_hw_addr_list mc; 2117 struct netdev_hw_addr_list dev_addrs; 2118 2119 #ifdef CONFIG_SYSFS 2120 struct kset *queues_kset; 2121 #endif 2122 #ifdef CONFIG_LOCKDEP 2123 struct list_head unlink_list; 2124 #endif 2125 unsigned int promiscuity; 2126 unsigned int allmulti; 2127 bool uc_promisc; 2128 #ifdef CONFIG_LOCKDEP 2129 unsigned char nested_level; 2130 #endif 2131 2132 2133 /* Protocol-specific pointers */ 2134 2135 struct in_device __rcu *ip_ptr; 2136 struct inet6_dev __rcu *ip6_ptr; 2137 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2138 struct vlan_info __rcu *vlan_info; 2139 #endif 2140 #if IS_ENABLED(CONFIG_NET_DSA) 2141 struct dsa_port *dsa_ptr; 2142 #endif 2143 #if IS_ENABLED(CONFIG_TIPC) 2144 struct tipc_bearer __rcu *tipc_ptr; 2145 #endif 2146 #if IS_ENABLED(CONFIG_ATALK) 2147 void *atalk_ptr; 2148 #endif 2149 #if IS_ENABLED(CONFIG_AX25) 2150 void *ax25_ptr; 2151 #endif 2152 #if IS_ENABLED(CONFIG_CFG80211) 2153 struct wireless_dev *ieee80211_ptr; 2154 #endif 2155 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2156 struct wpan_dev *ieee802154_ptr; 2157 #endif 2158 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2159 struct mpls_dev __rcu *mpls_ptr; 2160 #endif 2161 #if IS_ENABLED(CONFIG_MCTP) 2162 struct mctp_dev __rcu *mctp_ptr; 2163 #endif 2164 2165 /* 2166 * Cache lines mostly used on receive path (including eth_type_trans()) 2167 */ 2168 /* Interface address info used in eth_type_trans() */ 2169 const unsigned char *dev_addr; 2170 2171 struct netdev_rx_queue *_rx; 2172 unsigned int num_rx_queues; 2173 unsigned int real_num_rx_queues; 2174 2175 struct bpf_prog __rcu *xdp_prog; 2176 unsigned long gro_flush_timeout; 2177 int napi_defer_hard_irqs; 2178 #define GRO_LEGACY_MAX_SIZE 65536u 2179 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2180 * and shinfo->gso_segs is a 16bit field. 2181 */ 2182 #define GRO_MAX_SIZE (8 * 65535u) 2183 unsigned int gro_max_size; 2184 rx_handler_func_t __rcu *rx_handler; 2185 void __rcu *rx_handler_data; 2186 2187 #ifdef CONFIG_NET_CLS_ACT 2188 struct mini_Qdisc __rcu *miniq_ingress; 2189 #endif 2190 struct netdev_queue __rcu *ingress_queue; 2191 #ifdef CONFIG_NETFILTER_INGRESS 2192 struct nf_hook_entries __rcu *nf_hooks_ingress; 2193 #endif 2194 2195 unsigned char broadcast[MAX_ADDR_LEN]; 2196 #ifdef CONFIG_RFS_ACCEL 2197 struct cpu_rmap *rx_cpu_rmap; 2198 #endif 2199 struct hlist_node index_hlist; 2200 2201 /* 2202 * Cache lines mostly used on transmit path 2203 */ 2204 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2205 unsigned int num_tx_queues; 2206 unsigned int real_num_tx_queues; 2207 struct Qdisc __rcu *qdisc; 2208 unsigned int tx_queue_len; 2209 spinlock_t tx_global_lock; 2210 2211 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2212 2213 #ifdef CONFIG_XPS 2214 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2215 #endif 2216 #ifdef CONFIG_NET_CLS_ACT 2217 struct mini_Qdisc __rcu *miniq_egress; 2218 #endif 2219 #ifdef CONFIG_NETFILTER_EGRESS 2220 struct nf_hook_entries __rcu *nf_hooks_egress; 2221 #endif 2222 2223 #ifdef CONFIG_NET_SCHED 2224 DECLARE_HASHTABLE (qdisc_hash, 4); 2225 #endif 2226 /* These may be needed for future network-power-down code. */ 2227 struct timer_list watchdog_timer; 2228 int watchdog_timeo; 2229 2230 u32 proto_down_reason; 2231 2232 struct list_head todo_list; 2233 2234 #ifdef CONFIG_PCPU_DEV_REFCNT 2235 int __percpu *pcpu_refcnt; 2236 #else 2237 refcount_t dev_refcnt; 2238 #endif 2239 struct ref_tracker_dir refcnt_tracker; 2240 2241 struct list_head link_watch_list; 2242 2243 enum { NETREG_UNINITIALIZED=0, 2244 NETREG_REGISTERED, /* completed register_netdevice */ 2245 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2246 NETREG_UNREGISTERED, /* completed unregister todo */ 2247 NETREG_RELEASED, /* called free_netdev */ 2248 NETREG_DUMMY, /* dummy device for NAPI poll */ 2249 } reg_state:8; 2250 2251 bool dismantle; 2252 2253 enum { 2254 RTNL_LINK_INITIALIZED, 2255 RTNL_LINK_INITIALIZING, 2256 } rtnl_link_state:16; 2257 2258 bool needs_free_netdev; 2259 void (*priv_destructor)(struct net_device *dev); 2260 2261 #ifdef CONFIG_NETPOLL 2262 struct netpoll_info __rcu *npinfo; 2263 #endif 2264 2265 possible_net_t nd_net; 2266 2267 /* mid-layer private */ 2268 void *ml_priv; 2269 enum netdev_ml_priv_type ml_priv_type; 2270 2271 union { 2272 struct pcpu_lstats __percpu *lstats; 2273 struct pcpu_sw_netstats __percpu *tstats; 2274 struct pcpu_dstats __percpu *dstats; 2275 }; 2276 2277 #if IS_ENABLED(CONFIG_GARP) 2278 struct garp_port __rcu *garp_port; 2279 #endif 2280 #if IS_ENABLED(CONFIG_MRP) 2281 struct mrp_port __rcu *mrp_port; 2282 #endif 2283 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2284 struct dm_hw_stat_delta __rcu *dm_private; 2285 #endif 2286 struct device dev; 2287 const struct attribute_group *sysfs_groups[4]; 2288 const struct attribute_group *sysfs_rx_queue_group; 2289 2290 const struct rtnl_link_ops *rtnl_link_ops; 2291 2292 /* for setting kernel sock attribute on TCP connection setup */ 2293 #define GSO_MAX_SEGS 65535u 2294 #define GSO_LEGACY_MAX_SIZE 65536u 2295 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2296 * and shinfo->gso_segs is a 16bit field. 2297 */ 2298 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2299 2300 unsigned int gso_max_size; 2301 #define TSO_LEGACY_MAX_SIZE 65536 2302 #define TSO_MAX_SIZE UINT_MAX 2303 unsigned int tso_max_size; 2304 u16 gso_max_segs; 2305 #define TSO_MAX_SEGS U16_MAX 2306 u16 tso_max_segs; 2307 2308 #ifdef CONFIG_DCB 2309 const struct dcbnl_rtnl_ops *dcbnl_ops; 2310 #endif 2311 s16 num_tc; 2312 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2313 u8 prio_tc_map[TC_BITMASK + 1]; 2314 2315 #if IS_ENABLED(CONFIG_FCOE) 2316 unsigned int fcoe_ddp_xid; 2317 #endif 2318 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2319 struct netprio_map __rcu *priomap; 2320 #endif 2321 struct phy_device *phydev; 2322 struct sfp_bus *sfp_bus; 2323 struct lock_class_key *qdisc_tx_busylock; 2324 bool proto_down; 2325 unsigned wol_enabled:1; 2326 unsigned threaded:1; 2327 2328 struct list_head net_notifier_list; 2329 2330 #if IS_ENABLED(CONFIG_MACSEC) 2331 /* MACsec management functions */ 2332 const struct macsec_ops *macsec_ops; 2333 #endif 2334 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2335 struct udp_tunnel_nic *udp_tunnel_nic; 2336 2337 /* protected by rtnl_lock */ 2338 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2339 2340 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2341 netdevice_tracker linkwatch_dev_tracker; 2342 netdevice_tracker watchdog_dev_tracker; 2343 netdevice_tracker dev_registered_tracker; 2344 struct rtnl_hw_stats64 *offload_xstats_l3; 2345 }; 2346 #define to_net_dev(d) container_of(d, struct net_device, dev) 2347 2348 static inline bool netif_elide_gro(const struct net_device *dev) 2349 { 2350 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2351 return true; 2352 return false; 2353 } 2354 2355 #define NETDEV_ALIGN 32 2356 2357 static inline 2358 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2359 { 2360 return dev->prio_tc_map[prio & TC_BITMASK]; 2361 } 2362 2363 static inline 2364 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2365 { 2366 if (tc >= dev->num_tc) 2367 return -EINVAL; 2368 2369 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2370 return 0; 2371 } 2372 2373 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2374 void netdev_reset_tc(struct net_device *dev); 2375 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2376 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2377 2378 static inline 2379 int netdev_get_num_tc(struct net_device *dev) 2380 { 2381 return dev->num_tc; 2382 } 2383 2384 static inline void net_prefetch(void *p) 2385 { 2386 prefetch(p); 2387 #if L1_CACHE_BYTES < 128 2388 prefetch((u8 *)p + L1_CACHE_BYTES); 2389 #endif 2390 } 2391 2392 static inline void net_prefetchw(void *p) 2393 { 2394 prefetchw(p); 2395 #if L1_CACHE_BYTES < 128 2396 prefetchw((u8 *)p + L1_CACHE_BYTES); 2397 #endif 2398 } 2399 2400 void netdev_unbind_sb_channel(struct net_device *dev, 2401 struct net_device *sb_dev); 2402 int netdev_bind_sb_channel_queue(struct net_device *dev, 2403 struct net_device *sb_dev, 2404 u8 tc, u16 count, u16 offset); 2405 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2406 static inline int netdev_get_sb_channel(struct net_device *dev) 2407 { 2408 return max_t(int, -dev->num_tc, 0); 2409 } 2410 2411 static inline 2412 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2413 unsigned int index) 2414 { 2415 return &dev->_tx[index]; 2416 } 2417 2418 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2419 const struct sk_buff *skb) 2420 { 2421 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2422 } 2423 2424 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2425 void (*f)(struct net_device *, 2426 struct netdev_queue *, 2427 void *), 2428 void *arg) 2429 { 2430 unsigned int i; 2431 2432 for (i = 0; i < dev->num_tx_queues; i++) 2433 f(dev, &dev->_tx[i], arg); 2434 } 2435 2436 #define netdev_lockdep_set_classes(dev) \ 2437 { \ 2438 static struct lock_class_key qdisc_tx_busylock_key; \ 2439 static struct lock_class_key qdisc_xmit_lock_key; \ 2440 static struct lock_class_key dev_addr_list_lock_key; \ 2441 unsigned int i; \ 2442 \ 2443 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2444 lockdep_set_class(&(dev)->addr_list_lock, \ 2445 &dev_addr_list_lock_key); \ 2446 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2447 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2448 &qdisc_xmit_lock_key); \ 2449 } 2450 2451 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2452 struct net_device *sb_dev); 2453 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2454 struct sk_buff *skb, 2455 struct net_device *sb_dev); 2456 2457 /* returns the headroom that the master device needs to take in account 2458 * when forwarding to this dev 2459 */ 2460 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2461 { 2462 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2463 } 2464 2465 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2466 { 2467 if (dev->netdev_ops->ndo_set_rx_headroom) 2468 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2469 } 2470 2471 /* set the device rx headroom to the dev's default */ 2472 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2473 { 2474 netdev_set_rx_headroom(dev, -1); 2475 } 2476 2477 static inline void *netdev_get_ml_priv(struct net_device *dev, 2478 enum netdev_ml_priv_type type) 2479 { 2480 if (dev->ml_priv_type != type) 2481 return NULL; 2482 2483 return dev->ml_priv; 2484 } 2485 2486 static inline void netdev_set_ml_priv(struct net_device *dev, 2487 void *ml_priv, 2488 enum netdev_ml_priv_type type) 2489 { 2490 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2491 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2492 dev->ml_priv_type, type); 2493 WARN(!dev->ml_priv_type && dev->ml_priv, 2494 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2495 2496 dev->ml_priv = ml_priv; 2497 dev->ml_priv_type = type; 2498 } 2499 2500 /* 2501 * Net namespace inlines 2502 */ 2503 static inline 2504 struct net *dev_net(const struct net_device *dev) 2505 { 2506 return read_pnet(&dev->nd_net); 2507 } 2508 2509 static inline 2510 void dev_net_set(struct net_device *dev, struct net *net) 2511 { 2512 write_pnet(&dev->nd_net, net); 2513 } 2514 2515 /** 2516 * netdev_priv - access network device private data 2517 * @dev: network device 2518 * 2519 * Get network device private data 2520 */ 2521 static inline void *netdev_priv(const struct net_device *dev) 2522 { 2523 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2524 } 2525 2526 /* Set the sysfs physical device reference for the network logical device 2527 * if set prior to registration will cause a symlink during initialization. 2528 */ 2529 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2530 2531 /* Set the sysfs device type for the network logical device to allow 2532 * fine-grained identification of different network device types. For 2533 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2534 */ 2535 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2536 2537 /* Default NAPI poll() weight 2538 * Device drivers are strongly advised to not use bigger value 2539 */ 2540 #define NAPI_POLL_WEIGHT 64 2541 2542 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2543 int (*poll)(struct napi_struct *, int), int weight); 2544 2545 /** 2546 * netif_napi_add() - initialize a NAPI context 2547 * @dev: network device 2548 * @napi: NAPI context 2549 * @poll: polling function 2550 * @weight: default weight 2551 * 2552 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2553 * *any* of the other NAPI-related functions. 2554 */ 2555 static inline void 2556 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2557 int (*poll)(struct napi_struct *, int), int weight) 2558 { 2559 netif_napi_add_weight(dev, napi, poll, weight); 2560 } 2561 2562 static inline void 2563 netif_napi_add_tx_weight(struct net_device *dev, 2564 struct napi_struct *napi, 2565 int (*poll)(struct napi_struct *, int), 2566 int weight) 2567 { 2568 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2569 netif_napi_add_weight(dev, napi, poll, weight); 2570 } 2571 2572 /** 2573 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2574 * @dev: network device 2575 * @napi: NAPI context 2576 * @poll: polling function 2577 * 2578 * This variant of netif_napi_add() should be used from drivers using NAPI 2579 * to exclusively poll a TX queue. 2580 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2581 */ 2582 static inline void netif_napi_add_tx(struct net_device *dev, 2583 struct napi_struct *napi, 2584 int (*poll)(struct napi_struct *, int)) 2585 { 2586 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2587 } 2588 2589 /** 2590 * __netif_napi_del - remove a NAPI context 2591 * @napi: NAPI context 2592 * 2593 * Warning: caller must observe RCU grace period before freeing memory 2594 * containing @napi. Drivers might want to call this helper to combine 2595 * all the needed RCU grace periods into a single one. 2596 */ 2597 void __netif_napi_del(struct napi_struct *napi); 2598 2599 /** 2600 * netif_napi_del - remove a NAPI context 2601 * @napi: NAPI context 2602 * 2603 * netif_napi_del() removes a NAPI context from the network device NAPI list 2604 */ 2605 static inline void netif_napi_del(struct napi_struct *napi) 2606 { 2607 __netif_napi_del(napi); 2608 synchronize_net(); 2609 } 2610 2611 struct packet_type { 2612 __be16 type; /* This is really htons(ether_type). */ 2613 bool ignore_outgoing; 2614 struct net_device *dev; /* NULL is wildcarded here */ 2615 netdevice_tracker dev_tracker; 2616 int (*func) (struct sk_buff *, 2617 struct net_device *, 2618 struct packet_type *, 2619 struct net_device *); 2620 void (*list_func) (struct list_head *, 2621 struct packet_type *, 2622 struct net_device *); 2623 bool (*id_match)(struct packet_type *ptype, 2624 struct sock *sk); 2625 struct net *af_packet_net; 2626 void *af_packet_priv; 2627 struct list_head list; 2628 }; 2629 2630 struct offload_callbacks { 2631 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2632 netdev_features_t features); 2633 struct sk_buff *(*gro_receive)(struct list_head *head, 2634 struct sk_buff *skb); 2635 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2636 }; 2637 2638 struct packet_offload { 2639 __be16 type; /* This is really htons(ether_type). */ 2640 u16 priority; 2641 struct offload_callbacks callbacks; 2642 struct list_head list; 2643 }; 2644 2645 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2646 struct pcpu_sw_netstats { 2647 u64_stats_t rx_packets; 2648 u64_stats_t rx_bytes; 2649 u64_stats_t tx_packets; 2650 u64_stats_t tx_bytes; 2651 struct u64_stats_sync syncp; 2652 } __aligned(4 * sizeof(u64)); 2653 2654 struct pcpu_lstats { 2655 u64_stats_t packets; 2656 u64_stats_t bytes; 2657 struct u64_stats_sync syncp; 2658 } __aligned(2 * sizeof(u64)); 2659 2660 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2661 2662 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2663 { 2664 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2665 2666 u64_stats_update_begin(&tstats->syncp); 2667 u64_stats_add(&tstats->rx_bytes, len); 2668 u64_stats_inc(&tstats->rx_packets); 2669 u64_stats_update_end(&tstats->syncp); 2670 } 2671 2672 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2673 unsigned int packets, 2674 unsigned int len) 2675 { 2676 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2677 2678 u64_stats_update_begin(&tstats->syncp); 2679 u64_stats_add(&tstats->tx_bytes, len); 2680 u64_stats_add(&tstats->tx_packets, packets); 2681 u64_stats_update_end(&tstats->syncp); 2682 } 2683 2684 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2685 { 2686 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2687 2688 u64_stats_update_begin(&lstats->syncp); 2689 u64_stats_add(&lstats->bytes, len); 2690 u64_stats_inc(&lstats->packets); 2691 u64_stats_update_end(&lstats->syncp); 2692 } 2693 2694 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2695 ({ \ 2696 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2697 if (pcpu_stats) { \ 2698 int __cpu; \ 2699 for_each_possible_cpu(__cpu) { \ 2700 typeof(type) *stat; \ 2701 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2702 u64_stats_init(&stat->syncp); \ 2703 } \ 2704 } \ 2705 pcpu_stats; \ 2706 }) 2707 2708 #define netdev_alloc_pcpu_stats(type) \ 2709 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2710 2711 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2712 ({ \ 2713 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2714 if (pcpu_stats) { \ 2715 int __cpu; \ 2716 for_each_possible_cpu(__cpu) { \ 2717 typeof(type) *stat; \ 2718 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2719 u64_stats_init(&stat->syncp); \ 2720 } \ 2721 } \ 2722 pcpu_stats; \ 2723 }) 2724 2725 enum netdev_lag_tx_type { 2726 NETDEV_LAG_TX_TYPE_UNKNOWN, 2727 NETDEV_LAG_TX_TYPE_RANDOM, 2728 NETDEV_LAG_TX_TYPE_BROADCAST, 2729 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2730 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2731 NETDEV_LAG_TX_TYPE_HASH, 2732 }; 2733 2734 enum netdev_lag_hash { 2735 NETDEV_LAG_HASH_NONE, 2736 NETDEV_LAG_HASH_L2, 2737 NETDEV_LAG_HASH_L34, 2738 NETDEV_LAG_HASH_L23, 2739 NETDEV_LAG_HASH_E23, 2740 NETDEV_LAG_HASH_E34, 2741 NETDEV_LAG_HASH_VLAN_SRCMAC, 2742 NETDEV_LAG_HASH_UNKNOWN, 2743 }; 2744 2745 struct netdev_lag_upper_info { 2746 enum netdev_lag_tx_type tx_type; 2747 enum netdev_lag_hash hash_type; 2748 }; 2749 2750 struct netdev_lag_lower_state_info { 2751 u8 link_up : 1, 2752 tx_enabled : 1; 2753 }; 2754 2755 #include <linux/notifier.h> 2756 2757 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2758 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2759 * adding new types. 2760 */ 2761 enum netdev_cmd { 2762 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2763 NETDEV_DOWN, 2764 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2765 detected a hardware crash and restarted 2766 - we can use this eg to kick tcp sessions 2767 once done */ 2768 NETDEV_CHANGE, /* Notify device state change */ 2769 NETDEV_REGISTER, 2770 NETDEV_UNREGISTER, 2771 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2772 NETDEV_CHANGEADDR, /* notify after the address change */ 2773 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2774 NETDEV_GOING_DOWN, 2775 NETDEV_CHANGENAME, 2776 NETDEV_FEAT_CHANGE, 2777 NETDEV_BONDING_FAILOVER, 2778 NETDEV_PRE_UP, 2779 NETDEV_PRE_TYPE_CHANGE, 2780 NETDEV_POST_TYPE_CHANGE, 2781 NETDEV_POST_INIT, 2782 NETDEV_RELEASE, 2783 NETDEV_NOTIFY_PEERS, 2784 NETDEV_JOIN, 2785 NETDEV_CHANGEUPPER, 2786 NETDEV_RESEND_IGMP, 2787 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2788 NETDEV_CHANGEINFODATA, 2789 NETDEV_BONDING_INFO, 2790 NETDEV_PRECHANGEUPPER, 2791 NETDEV_CHANGELOWERSTATE, 2792 NETDEV_UDP_TUNNEL_PUSH_INFO, 2793 NETDEV_UDP_TUNNEL_DROP_INFO, 2794 NETDEV_CHANGE_TX_QUEUE_LEN, 2795 NETDEV_CVLAN_FILTER_PUSH_INFO, 2796 NETDEV_CVLAN_FILTER_DROP_INFO, 2797 NETDEV_SVLAN_FILTER_PUSH_INFO, 2798 NETDEV_SVLAN_FILTER_DROP_INFO, 2799 NETDEV_OFFLOAD_XSTATS_ENABLE, 2800 NETDEV_OFFLOAD_XSTATS_DISABLE, 2801 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 2802 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 2803 }; 2804 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2805 2806 int register_netdevice_notifier(struct notifier_block *nb); 2807 int unregister_netdevice_notifier(struct notifier_block *nb); 2808 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2809 int unregister_netdevice_notifier_net(struct net *net, 2810 struct notifier_block *nb); 2811 int register_netdevice_notifier_dev_net(struct net_device *dev, 2812 struct notifier_block *nb, 2813 struct netdev_net_notifier *nn); 2814 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2815 struct notifier_block *nb, 2816 struct netdev_net_notifier *nn); 2817 2818 struct netdev_notifier_info { 2819 struct net_device *dev; 2820 struct netlink_ext_ack *extack; 2821 }; 2822 2823 struct netdev_notifier_info_ext { 2824 struct netdev_notifier_info info; /* must be first */ 2825 union { 2826 u32 mtu; 2827 } ext; 2828 }; 2829 2830 struct netdev_notifier_change_info { 2831 struct netdev_notifier_info info; /* must be first */ 2832 unsigned int flags_changed; 2833 }; 2834 2835 struct netdev_notifier_changeupper_info { 2836 struct netdev_notifier_info info; /* must be first */ 2837 struct net_device *upper_dev; /* new upper dev */ 2838 bool master; /* is upper dev master */ 2839 bool linking; /* is the notification for link or unlink */ 2840 void *upper_info; /* upper dev info */ 2841 }; 2842 2843 struct netdev_notifier_changelowerstate_info { 2844 struct netdev_notifier_info info; /* must be first */ 2845 void *lower_state_info; /* is lower dev state */ 2846 }; 2847 2848 struct netdev_notifier_pre_changeaddr_info { 2849 struct netdev_notifier_info info; /* must be first */ 2850 const unsigned char *dev_addr; 2851 }; 2852 2853 enum netdev_offload_xstats_type { 2854 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 2855 }; 2856 2857 struct netdev_notifier_offload_xstats_info { 2858 struct netdev_notifier_info info; /* must be first */ 2859 enum netdev_offload_xstats_type type; 2860 2861 union { 2862 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 2863 struct netdev_notifier_offload_xstats_rd *report_delta; 2864 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 2865 struct netdev_notifier_offload_xstats_ru *report_used; 2866 }; 2867 }; 2868 2869 int netdev_offload_xstats_enable(struct net_device *dev, 2870 enum netdev_offload_xstats_type type, 2871 struct netlink_ext_ack *extack); 2872 int netdev_offload_xstats_disable(struct net_device *dev, 2873 enum netdev_offload_xstats_type type); 2874 bool netdev_offload_xstats_enabled(const struct net_device *dev, 2875 enum netdev_offload_xstats_type type); 2876 int netdev_offload_xstats_get(struct net_device *dev, 2877 enum netdev_offload_xstats_type type, 2878 struct rtnl_hw_stats64 *stats, bool *used, 2879 struct netlink_ext_ack *extack); 2880 void 2881 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 2882 const struct rtnl_hw_stats64 *stats); 2883 void 2884 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 2885 void netdev_offload_xstats_push_delta(struct net_device *dev, 2886 enum netdev_offload_xstats_type type, 2887 const struct rtnl_hw_stats64 *stats); 2888 2889 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2890 struct net_device *dev) 2891 { 2892 info->dev = dev; 2893 info->extack = NULL; 2894 } 2895 2896 static inline struct net_device * 2897 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2898 { 2899 return info->dev; 2900 } 2901 2902 static inline struct netlink_ext_ack * 2903 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2904 { 2905 return info->extack; 2906 } 2907 2908 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2909 2910 2911 extern rwlock_t dev_base_lock; /* Device list lock */ 2912 2913 #define for_each_netdev(net, d) \ 2914 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2915 #define for_each_netdev_reverse(net, d) \ 2916 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2917 #define for_each_netdev_rcu(net, d) \ 2918 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2919 #define for_each_netdev_safe(net, d, n) \ 2920 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2921 #define for_each_netdev_continue(net, d) \ 2922 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2923 #define for_each_netdev_continue_reverse(net, d) \ 2924 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2925 dev_list) 2926 #define for_each_netdev_continue_rcu(net, d) \ 2927 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2928 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2929 for_each_netdev_rcu(&init_net, slave) \ 2930 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2931 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2932 2933 static inline struct net_device *next_net_device(struct net_device *dev) 2934 { 2935 struct list_head *lh; 2936 struct net *net; 2937 2938 net = dev_net(dev); 2939 lh = dev->dev_list.next; 2940 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2941 } 2942 2943 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2944 { 2945 struct list_head *lh; 2946 struct net *net; 2947 2948 net = dev_net(dev); 2949 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2950 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2951 } 2952 2953 static inline struct net_device *first_net_device(struct net *net) 2954 { 2955 return list_empty(&net->dev_base_head) ? NULL : 2956 net_device_entry(net->dev_base_head.next); 2957 } 2958 2959 static inline struct net_device *first_net_device_rcu(struct net *net) 2960 { 2961 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2962 2963 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2964 } 2965 2966 int netdev_boot_setup_check(struct net_device *dev); 2967 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2968 const char *hwaddr); 2969 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2970 void dev_add_pack(struct packet_type *pt); 2971 void dev_remove_pack(struct packet_type *pt); 2972 void __dev_remove_pack(struct packet_type *pt); 2973 void dev_add_offload(struct packet_offload *po); 2974 void dev_remove_offload(struct packet_offload *po); 2975 2976 int dev_get_iflink(const struct net_device *dev); 2977 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2978 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 2979 struct net_device_path_stack *stack); 2980 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2981 unsigned short mask); 2982 struct net_device *dev_get_by_name(struct net *net, const char *name); 2983 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2984 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2985 bool netdev_name_in_use(struct net *net, const char *name); 2986 int dev_alloc_name(struct net_device *dev, const char *name); 2987 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2988 void dev_close(struct net_device *dev); 2989 void dev_close_many(struct list_head *head, bool unlink); 2990 void dev_disable_lro(struct net_device *dev); 2991 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2992 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2993 struct net_device *sb_dev); 2994 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2995 struct net_device *sb_dev); 2996 2997 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 2998 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2999 3000 static inline int dev_queue_xmit(struct sk_buff *skb) 3001 { 3002 return __dev_queue_xmit(skb, NULL); 3003 } 3004 3005 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3006 struct net_device *sb_dev) 3007 { 3008 return __dev_queue_xmit(skb, sb_dev); 3009 } 3010 3011 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3012 { 3013 int ret; 3014 3015 ret = __dev_direct_xmit(skb, queue_id); 3016 if (!dev_xmit_complete(ret)) 3017 kfree_skb(skb); 3018 return ret; 3019 } 3020 3021 int register_netdevice(struct net_device *dev); 3022 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3023 void unregister_netdevice_many(struct list_head *head); 3024 static inline void unregister_netdevice(struct net_device *dev) 3025 { 3026 unregister_netdevice_queue(dev, NULL); 3027 } 3028 3029 int netdev_refcnt_read(const struct net_device *dev); 3030 void free_netdev(struct net_device *dev); 3031 void netdev_freemem(struct net_device *dev); 3032 int init_dummy_netdev(struct net_device *dev); 3033 3034 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3035 struct sk_buff *skb, 3036 bool all_slaves); 3037 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3038 struct sock *sk); 3039 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3040 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3041 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3042 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 3043 int dev_restart(struct net_device *dev); 3044 3045 3046 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3047 unsigned short type, 3048 const void *daddr, const void *saddr, 3049 unsigned int len) 3050 { 3051 if (!dev->header_ops || !dev->header_ops->create) 3052 return 0; 3053 3054 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3055 } 3056 3057 static inline int dev_parse_header(const struct sk_buff *skb, 3058 unsigned char *haddr) 3059 { 3060 const struct net_device *dev = skb->dev; 3061 3062 if (!dev->header_ops || !dev->header_ops->parse) 3063 return 0; 3064 return dev->header_ops->parse(skb, haddr); 3065 } 3066 3067 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3068 { 3069 const struct net_device *dev = skb->dev; 3070 3071 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3072 return 0; 3073 return dev->header_ops->parse_protocol(skb); 3074 } 3075 3076 /* ll_header must have at least hard_header_len allocated */ 3077 static inline bool dev_validate_header(const struct net_device *dev, 3078 char *ll_header, int len) 3079 { 3080 if (likely(len >= dev->hard_header_len)) 3081 return true; 3082 if (len < dev->min_header_len) 3083 return false; 3084 3085 if (capable(CAP_SYS_RAWIO)) { 3086 memset(ll_header + len, 0, dev->hard_header_len - len); 3087 return true; 3088 } 3089 3090 if (dev->header_ops && dev->header_ops->validate) 3091 return dev->header_ops->validate(ll_header, len); 3092 3093 return false; 3094 } 3095 3096 static inline bool dev_has_header(const struct net_device *dev) 3097 { 3098 return dev->header_ops && dev->header_ops->create; 3099 } 3100 3101 /* 3102 * Incoming packets are placed on per-CPU queues 3103 */ 3104 struct softnet_data { 3105 struct list_head poll_list; 3106 struct sk_buff_head process_queue; 3107 3108 /* stats */ 3109 unsigned int processed; 3110 unsigned int time_squeeze; 3111 unsigned int received_rps; 3112 #ifdef CONFIG_RPS 3113 struct softnet_data *rps_ipi_list; 3114 #endif 3115 #ifdef CONFIG_NET_FLOW_LIMIT 3116 struct sd_flow_limit __rcu *flow_limit; 3117 #endif 3118 struct Qdisc *output_queue; 3119 struct Qdisc **output_queue_tailp; 3120 struct sk_buff *completion_queue; 3121 #ifdef CONFIG_XFRM_OFFLOAD 3122 struct sk_buff_head xfrm_backlog; 3123 #endif 3124 /* written and read only by owning cpu: */ 3125 struct { 3126 u16 recursion; 3127 u8 more; 3128 #ifdef CONFIG_NET_EGRESS 3129 u8 skip_txqueue; 3130 #endif 3131 } xmit; 3132 #ifdef CONFIG_RPS 3133 /* input_queue_head should be written by cpu owning this struct, 3134 * and only read by other cpus. Worth using a cache line. 3135 */ 3136 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3137 3138 /* Elements below can be accessed between CPUs for RPS/RFS */ 3139 call_single_data_t csd ____cacheline_aligned_in_smp; 3140 struct softnet_data *rps_ipi_next; 3141 unsigned int cpu; 3142 unsigned int input_queue_tail; 3143 #endif 3144 unsigned int dropped; 3145 struct sk_buff_head input_pkt_queue; 3146 struct napi_struct backlog; 3147 3148 /* Another possibly contended cache line */ 3149 spinlock_t defer_lock ____cacheline_aligned_in_smp; 3150 int defer_count; 3151 int defer_ipi_scheduled; 3152 struct sk_buff *defer_list; 3153 call_single_data_t defer_csd; 3154 }; 3155 3156 static inline void input_queue_head_incr(struct softnet_data *sd) 3157 { 3158 #ifdef CONFIG_RPS 3159 sd->input_queue_head++; 3160 #endif 3161 } 3162 3163 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3164 unsigned int *qtail) 3165 { 3166 #ifdef CONFIG_RPS 3167 *qtail = ++sd->input_queue_tail; 3168 #endif 3169 } 3170 3171 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3172 3173 static inline int dev_recursion_level(void) 3174 { 3175 return this_cpu_read(softnet_data.xmit.recursion); 3176 } 3177 3178 #define XMIT_RECURSION_LIMIT 8 3179 static inline bool dev_xmit_recursion(void) 3180 { 3181 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3182 XMIT_RECURSION_LIMIT); 3183 } 3184 3185 static inline void dev_xmit_recursion_inc(void) 3186 { 3187 __this_cpu_inc(softnet_data.xmit.recursion); 3188 } 3189 3190 static inline void dev_xmit_recursion_dec(void) 3191 { 3192 __this_cpu_dec(softnet_data.xmit.recursion); 3193 } 3194 3195 void __netif_schedule(struct Qdisc *q); 3196 void netif_schedule_queue(struct netdev_queue *txq); 3197 3198 static inline void netif_tx_schedule_all(struct net_device *dev) 3199 { 3200 unsigned int i; 3201 3202 for (i = 0; i < dev->num_tx_queues; i++) 3203 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3204 } 3205 3206 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3207 { 3208 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3209 } 3210 3211 /** 3212 * netif_start_queue - allow transmit 3213 * @dev: network device 3214 * 3215 * Allow upper layers to call the device hard_start_xmit routine. 3216 */ 3217 static inline void netif_start_queue(struct net_device *dev) 3218 { 3219 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3220 } 3221 3222 static inline void netif_tx_start_all_queues(struct net_device *dev) 3223 { 3224 unsigned int i; 3225 3226 for (i = 0; i < dev->num_tx_queues; i++) { 3227 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3228 netif_tx_start_queue(txq); 3229 } 3230 } 3231 3232 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3233 3234 /** 3235 * netif_wake_queue - restart transmit 3236 * @dev: network device 3237 * 3238 * Allow upper layers to call the device hard_start_xmit routine. 3239 * Used for flow control when transmit resources are available. 3240 */ 3241 static inline void netif_wake_queue(struct net_device *dev) 3242 { 3243 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3244 } 3245 3246 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3247 { 3248 unsigned int i; 3249 3250 for (i = 0; i < dev->num_tx_queues; i++) { 3251 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3252 netif_tx_wake_queue(txq); 3253 } 3254 } 3255 3256 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3257 { 3258 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3259 } 3260 3261 /** 3262 * netif_stop_queue - stop transmitted packets 3263 * @dev: network device 3264 * 3265 * Stop upper layers calling the device hard_start_xmit routine. 3266 * Used for flow control when transmit resources are unavailable. 3267 */ 3268 static inline void netif_stop_queue(struct net_device *dev) 3269 { 3270 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3271 } 3272 3273 void netif_tx_stop_all_queues(struct net_device *dev); 3274 3275 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3276 { 3277 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3278 } 3279 3280 /** 3281 * netif_queue_stopped - test if transmit queue is flowblocked 3282 * @dev: network device 3283 * 3284 * Test if transmit queue on device is currently unable to send. 3285 */ 3286 static inline bool netif_queue_stopped(const struct net_device *dev) 3287 { 3288 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3289 } 3290 3291 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3292 { 3293 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3294 } 3295 3296 static inline bool 3297 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3298 { 3299 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3300 } 3301 3302 static inline bool 3303 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3304 { 3305 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3306 } 3307 3308 /** 3309 * netdev_queue_set_dql_min_limit - set dql minimum limit 3310 * @dev_queue: pointer to transmit queue 3311 * @min_limit: dql minimum limit 3312 * 3313 * Forces xmit_more() to return true until the minimum threshold 3314 * defined by @min_limit is reached (or until the tx queue is 3315 * empty). Warning: to be use with care, misuse will impact the 3316 * latency. 3317 */ 3318 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3319 unsigned int min_limit) 3320 { 3321 #ifdef CONFIG_BQL 3322 dev_queue->dql.min_limit = min_limit; 3323 #endif 3324 } 3325 3326 /** 3327 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3328 * @dev_queue: pointer to transmit queue 3329 * 3330 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3331 * to give appropriate hint to the CPU. 3332 */ 3333 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3334 { 3335 #ifdef CONFIG_BQL 3336 prefetchw(&dev_queue->dql.num_queued); 3337 #endif 3338 } 3339 3340 /** 3341 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3342 * @dev_queue: pointer to transmit queue 3343 * 3344 * BQL enabled drivers might use this helper in their TX completion path, 3345 * to give appropriate hint to the CPU. 3346 */ 3347 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3348 { 3349 #ifdef CONFIG_BQL 3350 prefetchw(&dev_queue->dql.limit); 3351 #endif 3352 } 3353 3354 /** 3355 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3356 * @dev_queue: network device queue 3357 * @bytes: number of bytes queued to the device queue 3358 * 3359 * Report the number of bytes queued for sending/completion to the network 3360 * device hardware queue. @bytes should be a good approximation and should 3361 * exactly match netdev_completed_queue() @bytes. 3362 * This is typically called once per packet, from ndo_start_xmit(). 3363 */ 3364 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3365 unsigned int bytes) 3366 { 3367 #ifdef CONFIG_BQL 3368 dql_queued(&dev_queue->dql, bytes); 3369 3370 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3371 return; 3372 3373 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3374 3375 /* 3376 * The XOFF flag must be set before checking the dql_avail below, 3377 * because in netdev_tx_completed_queue we update the dql_completed 3378 * before checking the XOFF flag. 3379 */ 3380 smp_mb(); 3381 3382 /* check again in case another CPU has just made room avail */ 3383 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3384 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3385 #endif 3386 } 3387 3388 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3389 * that they should not test BQL status themselves. 3390 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3391 * skb of a batch. 3392 * Returns true if the doorbell must be used to kick the NIC. 3393 */ 3394 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3395 unsigned int bytes, 3396 bool xmit_more) 3397 { 3398 if (xmit_more) { 3399 #ifdef CONFIG_BQL 3400 dql_queued(&dev_queue->dql, bytes); 3401 #endif 3402 return netif_tx_queue_stopped(dev_queue); 3403 } 3404 netdev_tx_sent_queue(dev_queue, bytes); 3405 return true; 3406 } 3407 3408 /** 3409 * netdev_sent_queue - report the number of bytes queued to hardware 3410 * @dev: network device 3411 * @bytes: number of bytes queued to the hardware device queue 3412 * 3413 * Report the number of bytes queued for sending/completion to the network 3414 * device hardware queue#0. @bytes should be a good approximation and should 3415 * exactly match netdev_completed_queue() @bytes. 3416 * This is typically called once per packet, from ndo_start_xmit(). 3417 */ 3418 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3419 { 3420 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3421 } 3422 3423 static inline bool __netdev_sent_queue(struct net_device *dev, 3424 unsigned int bytes, 3425 bool xmit_more) 3426 { 3427 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3428 xmit_more); 3429 } 3430 3431 /** 3432 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3433 * @dev_queue: network device queue 3434 * @pkts: number of packets (currently ignored) 3435 * @bytes: number of bytes dequeued from the device queue 3436 * 3437 * Must be called at most once per TX completion round (and not per 3438 * individual packet), so that BQL can adjust its limits appropriately. 3439 */ 3440 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3441 unsigned int pkts, unsigned int bytes) 3442 { 3443 #ifdef CONFIG_BQL 3444 if (unlikely(!bytes)) 3445 return; 3446 3447 dql_completed(&dev_queue->dql, bytes); 3448 3449 /* 3450 * Without the memory barrier there is a small possiblity that 3451 * netdev_tx_sent_queue will miss the update and cause the queue to 3452 * be stopped forever 3453 */ 3454 smp_mb(); 3455 3456 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3457 return; 3458 3459 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3460 netif_schedule_queue(dev_queue); 3461 #endif 3462 } 3463 3464 /** 3465 * netdev_completed_queue - report bytes and packets completed by device 3466 * @dev: network device 3467 * @pkts: actual number of packets sent over the medium 3468 * @bytes: actual number of bytes sent over the medium 3469 * 3470 * Report the number of bytes and packets transmitted by the network device 3471 * hardware queue over the physical medium, @bytes must exactly match the 3472 * @bytes amount passed to netdev_sent_queue() 3473 */ 3474 static inline void netdev_completed_queue(struct net_device *dev, 3475 unsigned int pkts, unsigned int bytes) 3476 { 3477 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3478 } 3479 3480 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3481 { 3482 #ifdef CONFIG_BQL 3483 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3484 dql_reset(&q->dql); 3485 #endif 3486 } 3487 3488 /** 3489 * netdev_reset_queue - reset the packets and bytes count of a network device 3490 * @dev_queue: network device 3491 * 3492 * Reset the bytes and packet count of a network device and clear the 3493 * software flow control OFF bit for this network device 3494 */ 3495 static inline void netdev_reset_queue(struct net_device *dev_queue) 3496 { 3497 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3498 } 3499 3500 /** 3501 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3502 * @dev: network device 3503 * @queue_index: given tx queue index 3504 * 3505 * Returns 0 if given tx queue index >= number of device tx queues, 3506 * otherwise returns the originally passed tx queue index. 3507 */ 3508 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3509 { 3510 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3511 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3512 dev->name, queue_index, 3513 dev->real_num_tx_queues); 3514 return 0; 3515 } 3516 3517 return queue_index; 3518 } 3519 3520 /** 3521 * netif_running - test if up 3522 * @dev: network device 3523 * 3524 * Test if the device has been brought up. 3525 */ 3526 static inline bool netif_running(const struct net_device *dev) 3527 { 3528 return test_bit(__LINK_STATE_START, &dev->state); 3529 } 3530 3531 /* 3532 * Routines to manage the subqueues on a device. We only need start, 3533 * stop, and a check if it's stopped. All other device management is 3534 * done at the overall netdevice level. 3535 * Also test the device if we're multiqueue. 3536 */ 3537 3538 /** 3539 * netif_start_subqueue - allow sending packets on subqueue 3540 * @dev: network device 3541 * @queue_index: sub queue index 3542 * 3543 * Start individual transmit queue of a device with multiple transmit queues. 3544 */ 3545 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3546 { 3547 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3548 3549 netif_tx_start_queue(txq); 3550 } 3551 3552 /** 3553 * netif_stop_subqueue - stop sending packets on subqueue 3554 * @dev: network device 3555 * @queue_index: sub queue index 3556 * 3557 * Stop individual transmit queue of a device with multiple transmit queues. 3558 */ 3559 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3560 { 3561 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3562 netif_tx_stop_queue(txq); 3563 } 3564 3565 /** 3566 * __netif_subqueue_stopped - test status of subqueue 3567 * @dev: network device 3568 * @queue_index: sub queue index 3569 * 3570 * Check individual transmit queue of a device with multiple transmit queues. 3571 */ 3572 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3573 u16 queue_index) 3574 { 3575 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3576 3577 return netif_tx_queue_stopped(txq); 3578 } 3579 3580 /** 3581 * netif_subqueue_stopped - test status of subqueue 3582 * @dev: network device 3583 * @skb: sub queue buffer pointer 3584 * 3585 * Check individual transmit queue of a device with multiple transmit queues. 3586 */ 3587 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3588 struct sk_buff *skb) 3589 { 3590 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3591 } 3592 3593 /** 3594 * netif_wake_subqueue - allow sending packets on subqueue 3595 * @dev: network device 3596 * @queue_index: sub queue index 3597 * 3598 * Resume individual transmit queue of a device with multiple transmit queues. 3599 */ 3600 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3601 { 3602 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3603 3604 netif_tx_wake_queue(txq); 3605 } 3606 3607 #ifdef CONFIG_XPS 3608 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3609 u16 index); 3610 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3611 u16 index, enum xps_map_type type); 3612 3613 /** 3614 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3615 * @j: CPU/Rx queue index 3616 * @mask: bitmask of all cpus/rx queues 3617 * @nr_bits: number of bits in the bitmask 3618 * 3619 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3620 */ 3621 static inline bool netif_attr_test_mask(unsigned long j, 3622 const unsigned long *mask, 3623 unsigned int nr_bits) 3624 { 3625 cpu_max_bits_warn(j, nr_bits); 3626 return test_bit(j, mask); 3627 } 3628 3629 /** 3630 * netif_attr_test_online - Test for online CPU/Rx queue 3631 * @j: CPU/Rx queue index 3632 * @online_mask: bitmask for CPUs/Rx queues that are online 3633 * @nr_bits: number of bits in the bitmask 3634 * 3635 * Returns true if a CPU/Rx queue is online. 3636 */ 3637 static inline bool netif_attr_test_online(unsigned long j, 3638 const unsigned long *online_mask, 3639 unsigned int nr_bits) 3640 { 3641 cpu_max_bits_warn(j, nr_bits); 3642 3643 if (online_mask) 3644 return test_bit(j, online_mask); 3645 3646 return (j < nr_bits); 3647 } 3648 3649 /** 3650 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3651 * @n: CPU/Rx queue index 3652 * @srcp: the cpumask/Rx queue mask pointer 3653 * @nr_bits: number of bits in the bitmask 3654 * 3655 * Returns >= nr_bits if no further CPUs/Rx queues set. 3656 */ 3657 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3658 unsigned int nr_bits) 3659 { 3660 /* -1 is a legal arg here. */ 3661 if (n != -1) 3662 cpu_max_bits_warn(n, nr_bits); 3663 3664 if (srcp) 3665 return find_next_bit(srcp, nr_bits, n + 1); 3666 3667 return n + 1; 3668 } 3669 3670 /** 3671 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3672 * @n: CPU/Rx queue index 3673 * @src1p: the first CPUs/Rx queues mask pointer 3674 * @src2p: the second CPUs/Rx queues mask pointer 3675 * @nr_bits: number of bits in the bitmask 3676 * 3677 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3678 */ 3679 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3680 const unsigned long *src2p, 3681 unsigned int nr_bits) 3682 { 3683 /* -1 is a legal arg here. */ 3684 if (n != -1) 3685 cpu_max_bits_warn(n, nr_bits); 3686 3687 if (src1p && src2p) 3688 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3689 else if (src1p) 3690 return find_next_bit(src1p, nr_bits, n + 1); 3691 else if (src2p) 3692 return find_next_bit(src2p, nr_bits, n + 1); 3693 3694 return n + 1; 3695 } 3696 #else 3697 static inline int netif_set_xps_queue(struct net_device *dev, 3698 const struct cpumask *mask, 3699 u16 index) 3700 { 3701 return 0; 3702 } 3703 3704 static inline int __netif_set_xps_queue(struct net_device *dev, 3705 const unsigned long *mask, 3706 u16 index, enum xps_map_type type) 3707 { 3708 return 0; 3709 } 3710 #endif 3711 3712 /** 3713 * netif_is_multiqueue - test if device has multiple transmit queues 3714 * @dev: network device 3715 * 3716 * Check if device has multiple transmit queues 3717 */ 3718 static inline bool netif_is_multiqueue(const struct net_device *dev) 3719 { 3720 return dev->num_tx_queues > 1; 3721 } 3722 3723 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3724 3725 #ifdef CONFIG_SYSFS 3726 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3727 #else 3728 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3729 unsigned int rxqs) 3730 { 3731 dev->real_num_rx_queues = rxqs; 3732 return 0; 3733 } 3734 #endif 3735 int netif_set_real_num_queues(struct net_device *dev, 3736 unsigned int txq, unsigned int rxq); 3737 3738 static inline struct netdev_rx_queue * 3739 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3740 { 3741 return dev->_rx + rxq; 3742 } 3743 3744 #ifdef CONFIG_SYSFS 3745 static inline unsigned int get_netdev_rx_queue_index( 3746 struct netdev_rx_queue *queue) 3747 { 3748 struct net_device *dev = queue->dev; 3749 int index = queue - dev->_rx; 3750 3751 BUG_ON(index >= dev->num_rx_queues); 3752 return index; 3753 } 3754 #endif 3755 3756 int netif_get_num_default_rss_queues(void); 3757 3758 enum skb_free_reason { 3759 SKB_REASON_CONSUMED, 3760 SKB_REASON_DROPPED, 3761 }; 3762 3763 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3764 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3765 3766 /* 3767 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3768 * interrupt context or with hardware interrupts being disabled. 3769 * (in_hardirq() || irqs_disabled()) 3770 * 3771 * We provide four helpers that can be used in following contexts : 3772 * 3773 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3774 * replacing kfree_skb(skb) 3775 * 3776 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3777 * Typically used in place of consume_skb(skb) in TX completion path 3778 * 3779 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3780 * replacing kfree_skb(skb) 3781 * 3782 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3783 * and consumed a packet. Used in place of consume_skb(skb) 3784 */ 3785 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3786 { 3787 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3788 } 3789 3790 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3791 { 3792 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3793 } 3794 3795 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3796 { 3797 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3798 } 3799 3800 static inline void dev_consume_skb_any(struct sk_buff *skb) 3801 { 3802 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3803 } 3804 3805 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3806 struct bpf_prog *xdp_prog); 3807 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3808 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3809 int netif_rx(struct sk_buff *skb); 3810 int __netif_rx(struct sk_buff *skb); 3811 3812 int netif_receive_skb(struct sk_buff *skb); 3813 int netif_receive_skb_core(struct sk_buff *skb); 3814 void netif_receive_skb_list_internal(struct list_head *head); 3815 void netif_receive_skb_list(struct list_head *head); 3816 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3817 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3818 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3819 gro_result_t napi_gro_frags(struct napi_struct *napi); 3820 struct packet_offload *gro_find_receive_by_type(__be16 type); 3821 struct packet_offload *gro_find_complete_by_type(__be16 type); 3822 3823 static inline void napi_free_frags(struct napi_struct *napi) 3824 { 3825 kfree_skb(napi->skb); 3826 napi->skb = NULL; 3827 } 3828 3829 bool netdev_is_rx_handler_busy(struct net_device *dev); 3830 int netdev_rx_handler_register(struct net_device *dev, 3831 rx_handler_func_t *rx_handler, 3832 void *rx_handler_data); 3833 void netdev_rx_handler_unregister(struct net_device *dev); 3834 3835 bool dev_valid_name(const char *name); 3836 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 3837 { 3838 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 3839 } 3840 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 3841 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 3842 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3843 void __user *data, bool *need_copyout); 3844 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 3845 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 3846 unsigned int dev_get_flags(const struct net_device *); 3847 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3848 struct netlink_ext_ack *extack); 3849 int dev_change_flags(struct net_device *dev, unsigned int flags, 3850 struct netlink_ext_ack *extack); 3851 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3852 unsigned int gchanges); 3853 int dev_set_alias(struct net_device *, const char *, size_t); 3854 int dev_get_alias(const struct net_device *, char *, size_t); 3855 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 3856 const char *pat, int new_ifindex); 3857 static inline 3858 int dev_change_net_namespace(struct net_device *dev, struct net *net, 3859 const char *pat) 3860 { 3861 return __dev_change_net_namespace(dev, net, pat, 0); 3862 } 3863 int __dev_set_mtu(struct net_device *, int); 3864 int dev_set_mtu(struct net_device *, int); 3865 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3866 struct netlink_ext_ack *extack); 3867 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3868 struct netlink_ext_ack *extack); 3869 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 3870 struct netlink_ext_ack *extack); 3871 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 3872 int dev_get_port_parent_id(struct net_device *dev, 3873 struct netdev_phys_item_id *ppid, bool recurse); 3874 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3875 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3876 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3877 struct netdev_queue *txq, int *ret); 3878 3879 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3880 u8 dev_xdp_prog_count(struct net_device *dev); 3881 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3882 3883 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3884 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3885 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 3886 bool is_skb_forwardable(const struct net_device *dev, 3887 const struct sk_buff *skb); 3888 3889 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 3890 const struct sk_buff *skb, 3891 const bool check_mtu) 3892 { 3893 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 3894 unsigned int len; 3895 3896 if (!(dev->flags & IFF_UP)) 3897 return false; 3898 3899 if (!check_mtu) 3900 return true; 3901 3902 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 3903 if (skb->len <= len) 3904 return true; 3905 3906 /* if TSO is enabled, we don't care about the length as the packet 3907 * could be forwarded without being segmented before 3908 */ 3909 if (skb_is_gso(skb)) 3910 return true; 3911 3912 return false; 3913 } 3914 3915 struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev); 3916 3917 static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev) 3918 { 3919 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */ 3920 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats); 3921 3922 if (likely(p)) 3923 return p; 3924 3925 return netdev_core_stats_alloc(dev); 3926 } 3927 3928 #define DEV_CORE_STATS_INC(FIELD) \ 3929 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 3930 { \ 3931 struct net_device_core_stats __percpu *p; \ 3932 \ 3933 p = dev_core_stats(dev); \ 3934 if (p) \ 3935 this_cpu_inc(p->FIELD); \ 3936 } 3937 DEV_CORE_STATS_INC(rx_dropped) 3938 DEV_CORE_STATS_INC(tx_dropped) 3939 DEV_CORE_STATS_INC(rx_nohandler) 3940 DEV_CORE_STATS_INC(rx_otherhost_dropped) 3941 3942 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3943 struct sk_buff *skb, 3944 const bool check_mtu) 3945 { 3946 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3947 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 3948 dev_core_stats_rx_dropped_inc(dev); 3949 kfree_skb(skb); 3950 return NET_RX_DROP; 3951 } 3952 3953 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 3954 skb->priority = 0; 3955 return 0; 3956 } 3957 3958 bool dev_nit_active(struct net_device *dev); 3959 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3960 3961 static inline void __dev_put(struct net_device *dev) 3962 { 3963 if (dev) { 3964 #ifdef CONFIG_PCPU_DEV_REFCNT 3965 this_cpu_dec(*dev->pcpu_refcnt); 3966 #else 3967 refcount_dec(&dev->dev_refcnt); 3968 #endif 3969 } 3970 } 3971 3972 static inline void __dev_hold(struct net_device *dev) 3973 { 3974 if (dev) { 3975 #ifdef CONFIG_PCPU_DEV_REFCNT 3976 this_cpu_inc(*dev->pcpu_refcnt); 3977 #else 3978 refcount_inc(&dev->dev_refcnt); 3979 #endif 3980 } 3981 } 3982 3983 static inline void __netdev_tracker_alloc(struct net_device *dev, 3984 netdevice_tracker *tracker, 3985 gfp_t gfp) 3986 { 3987 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 3988 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 3989 #endif 3990 } 3991 3992 /* netdev_tracker_alloc() can upgrade a prior untracked reference 3993 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 3994 */ 3995 static inline void netdev_tracker_alloc(struct net_device *dev, 3996 netdevice_tracker *tracker, gfp_t gfp) 3997 { 3998 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 3999 refcount_dec(&dev->refcnt_tracker.no_tracker); 4000 __netdev_tracker_alloc(dev, tracker, gfp); 4001 #endif 4002 } 4003 4004 static inline void netdev_tracker_free(struct net_device *dev, 4005 netdevice_tracker *tracker) 4006 { 4007 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4008 ref_tracker_free(&dev->refcnt_tracker, tracker); 4009 #endif 4010 } 4011 4012 static inline void netdev_hold(struct net_device *dev, 4013 netdevice_tracker *tracker, gfp_t gfp) 4014 { 4015 if (dev) { 4016 __dev_hold(dev); 4017 __netdev_tracker_alloc(dev, tracker, gfp); 4018 } 4019 } 4020 4021 static inline void netdev_put(struct net_device *dev, 4022 netdevice_tracker *tracker) 4023 { 4024 if (dev) { 4025 netdev_tracker_free(dev, tracker); 4026 __dev_put(dev); 4027 } 4028 } 4029 4030 /** 4031 * dev_hold - get reference to device 4032 * @dev: network device 4033 * 4034 * Hold reference to device to keep it from being freed. 4035 * Try using netdev_hold() instead. 4036 */ 4037 static inline void dev_hold(struct net_device *dev) 4038 { 4039 netdev_hold(dev, NULL, GFP_ATOMIC); 4040 } 4041 4042 /** 4043 * dev_put - release reference to device 4044 * @dev: network device 4045 * 4046 * Release reference to device to allow it to be freed. 4047 * Try using netdev_put() instead. 4048 */ 4049 static inline void dev_put(struct net_device *dev) 4050 { 4051 netdev_put(dev, NULL); 4052 } 4053 4054 static inline void netdev_ref_replace(struct net_device *odev, 4055 struct net_device *ndev, 4056 netdevice_tracker *tracker, 4057 gfp_t gfp) 4058 { 4059 if (odev) 4060 netdev_tracker_free(odev, tracker); 4061 4062 __dev_hold(ndev); 4063 __dev_put(odev); 4064 4065 if (ndev) 4066 __netdev_tracker_alloc(ndev, tracker, gfp); 4067 } 4068 4069 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4070 * and _off may be called from IRQ context, but it is caller 4071 * who is responsible for serialization of these calls. 4072 * 4073 * The name carrier is inappropriate, these functions should really be 4074 * called netif_lowerlayer_*() because they represent the state of any 4075 * kind of lower layer not just hardware media. 4076 */ 4077 void linkwatch_fire_event(struct net_device *dev); 4078 4079 /** 4080 * netif_carrier_ok - test if carrier present 4081 * @dev: network device 4082 * 4083 * Check if carrier is present on device 4084 */ 4085 static inline bool netif_carrier_ok(const struct net_device *dev) 4086 { 4087 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4088 } 4089 4090 unsigned long dev_trans_start(struct net_device *dev); 4091 4092 void __netdev_watchdog_up(struct net_device *dev); 4093 4094 void netif_carrier_on(struct net_device *dev); 4095 void netif_carrier_off(struct net_device *dev); 4096 void netif_carrier_event(struct net_device *dev); 4097 4098 /** 4099 * netif_dormant_on - mark device as dormant. 4100 * @dev: network device 4101 * 4102 * Mark device as dormant (as per RFC2863). 4103 * 4104 * The dormant state indicates that the relevant interface is not 4105 * actually in a condition to pass packets (i.e., it is not 'up') but is 4106 * in a "pending" state, waiting for some external event. For "on- 4107 * demand" interfaces, this new state identifies the situation where the 4108 * interface is waiting for events to place it in the up state. 4109 */ 4110 static inline void netif_dormant_on(struct net_device *dev) 4111 { 4112 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4113 linkwatch_fire_event(dev); 4114 } 4115 4116 /** 4117 * netif_dormant_off - set device as not dormant. 4118 * @dev: network device 4119 * 4120 * Device is not in dormant state. 4121 */ 4122 static inline void netif_dormant_off(struct net_device *dev) 4123 { 4124 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4125 linkwatch_fire_event(dev); 4126 } 4127 4128 /** 4129 * netif_dormant - test if device is dormant 4130 * @dev: network device 4131 * 4132 * Check if device is dormant. 4133 */ 4134 static inline bool netif_dormant(const struct net_device *dev) 4135 { 4136 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4137 } 4138 4139 4140 /** 4141 * netif_testing_on - mark device as under test. 4142 * @dev: network device 4143 * 4144 * Mark device as under test (as per RFC2863). 4145 * 4146 * The testing state indicates that some test(s) must be performed on 4147 * the interface. After completion, of the test, the interface state 4148 * will change to up, dormant, or down, as appropriate. 4149 */ 4150 static inline void netif_testing_on(struct net_device *dev) 4151 { 4152 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4153 linkwatch_fire_event(dev); 4154 } 4155 4156 /** 4157 * netif_testing_off - set device as not under test. 4158 * @dev: network device 4159 * 4160 * Device is not in testing state. 4161 */ 4162 static inline void netif_testing_off(struct net_device *dev) 4163 { 4164 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4165 linkwatch_fire_event(dev); 4166 } 4167 4168 /** 4169 * netif_testing - test if device is under test 4170 * @dev: network device 4171 * 4172 * Check if device is under test 4173 */ 4174 static inline bool netif_testing(const struct net_device *dev) 4175 { 4176 return test_bit(__LINK_STATE_TESTING, &dev->state); 4177 } 4178 4179 4180 /** 4181 * netif_oper_up - test if device is operational 4182 * @dev: network device 4183 * 4184 * Check if carrier is operational 4185 */ 4186 static inline bool netif_oper_up(const struct net_device *dev) 4187 { 4188 return (dev->operstate == IF_OPER_UP || 4189 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4190 } 4191 4192 /** 4193 * netif_device_present - is device available or removed 4194 * @dev: network device 4195 * 4196 * Check if device has not been removed from system. 4197 */ 4198 static inline bool netif_device_present(const struct net_device *dev) 4199 { 4200 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4201 } 4202 4203 void netif_device_detach(struct net_device *dev); 4204 4205 void netif_device_attach(struct net_device *dev); 4206 4207 /* 4208 * Network interface message level settings 4209 */ 4210 4211 enum { 4212 NETIF_MSG_DRV_BIT, 4213 NETIF_MSG_PROBE_BIT, 4214 NETIF_MSG_LINK_BIT, 4215 NETIF_MSG_TIMER_BIT, 4216 NETIF_MSG_IFDOWN_BIT, 4217 NETIF_MSG_IFUP_BIT, 4218 NETIF_MSG_RX_ERR_BIT, 4219 NETIF_MSG_TX_ERR_BIT, 4220 NETIF_MSG_TX_QUEUED_BIT, 4221 NETIF_MSG_INTR_BIT, 4222 NETIF_MSG_TX_DONE_BIT, 4223 NETIF_MSG_RX_STATUS_BIT, 4224 NETIF_MSG_PKTDATA_BIT, 4225 NETIF_MSG_HW_BIT, 4226 NETIF_MSG_WOL_BIT, 4227 4228 /* When you add a new bit above, update netif_msg_class_names array 4229 * in net/ethtool/common.c 4230 */ 4231 NETIF_MSG_CLASS_COUNT, 4232 }; 4233 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4234 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4235 4236 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4237 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4238 4239 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4240 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4241 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4242 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4243 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4244 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4245 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4246 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4247 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4248 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4249 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4250 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4251 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4252 #define NETIF_MSG_HW __NETIF_MSG(HW) 4253 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4254 4255 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4256 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4257 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4258 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4259 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4260 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4261 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4262 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4263 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4264 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4265 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4266 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4267 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4268 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4269 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4270 4271 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4272 { 4273 /* use default */ 4274 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4275 return default_msg_enable_bits; 4276 if (debug_value == 0) /* no output */ 4277 return 0; 4278 /* set low N bits */ 4279 return (1U << debug_value) - 1; 4280 } 4281 4282 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4283 { 4284 spin_lock(&txq->_xmit_lock); 4285 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4286 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4287 } 4288 4289 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4290 { 4291 __acquire(&txq->_xmit_lock); 4292 return true; 4293 } 4294 4295 static inline void __netif_tx_release(struct netdev_queue *txq) 4296 { 4297 __release(&txq->_xmit_lock); 4298 } 4299 4300 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4301 { 4302 spin_lock_bh(&txq->_xmit_lock); 4303 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4304 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4305 } 4306 4307 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4308 { 4309 bool ok = spin_trylock(&txq->_xmit_lock); 4310 4311 if (likely(ok)) { 4312 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4313 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4314 } 4315 return ok; 4316 } 4317 4318 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4319 { 4320 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4321 WRITE_ONCE(txq->xmit_lock_owner, -1); 4322 spin_unlock(&txq->_xmit_lock); 4323 } 4324 4325 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4326 { 4327 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4328 WRITE_ONCE(txq->xmit_lock_owner, -1); 4329 spin_unlock_bh(&txq->_xmit_lock); 4330 } 4331 4332 /* 4333 * txq->trans_start can be read locklessly from dev_watchdog() 4334 */ 4335 static inline void txq_trans_update(struct netdev_queue *txq) 4336 { 4337 if (txq->xmit_lock_owner != -1) 4338 WRITE_ONCE(txq->trans_start, jiffies); 4339 } 4340 4341 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4342 { 4343 unsigned long now = jiffies; 4344 4345 if (READ_ONCE(txq->trans_start) != now) 4346 WRITE_ONCE(txq->trans_start, now); 4347 } 4348 4349 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4350 static inline void netif_trans_update(struct net_device *dev) 4351 { 4352 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4353 4354 txq_trans_cond_update(txq); 4355 } 4356 4357 /** 4358 * netif_tx_lock - grab network device transmit lock 4359 * @dev: network device 4360 * 4361 * Get network device transmit lock 4362 */ 4363 void netif_tx_lock(struct net_device *dev); 4364 4365 static inline void netif_tx_lock_bh(struct net_device *dev) 4366 { 4367 local_bh_disable(); 4368 netif_tx_lock(dev); 4369 } 4370 4371 void netif_tx_unlock(struct net_device *dev); 4372 4373 static inline void netif_tx_unlock_bh(struct net_device *dev) 4374 { 4375 netif_tx_unlock(dev); 4376 local_bh_enable(); 4377 } 4378 4379 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4380 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4381 __netif_tx_lock(txq, cpu); \ 4382 } else { \ 4383 __netif_tx_acquire(txq); \ 4384 } \ 4385 } 4386 4387 #define HARD_TX_TRYLOCK(dev, txq) \ 4388 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4389 __netif_tx_trylock(txq) : \ 4390 __netif_tx_acquire(txq)) 4391 4392 #define HARD_TX_UNLOCK(dev, txq) { \ 4393 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4394 __netif_tx_unlock(txq); \ 4395 } else { \ 4396 __netif_tx_release(txq); \ 4397 } \ 4398 } 4399 4400 static inline void netif_tx_disable(struct net_device *dev) 4401 { 4402 unsigned int i; 4403 int cpu; 4404 4405 local_bh_disable(); 4406 cpu = smp_processor_id(); 4407 spin_lock(&dev->tx_global_lock); 4408 for (i = 0; i < dev->num_tx_queues; i++) { 4409 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4410 4411 __netif_tx_lock(txq, cpu); 4412 netif_tx_stop_queue(txq); 4413 __netif_tx_unlock(txq); 4414 } 4415 spin_unlock(&dev->tx_global_lock); 4416 local_bh_enable(); 4417 } 4418 4419 static inline void netif_addr_lock(struct net_device *dev) 4420 { 4421 unsigned char nest_level = 0; 4422 4423 #ifdef CONFIG_LOCKDEP 4424 nest_level = dev->nested_level; 4425 #endif 4426 spin_lock_nested(&dev->addr_list_lock, nest_level); 4427 } 4428 4429 static inline void netif_addr_lock_bh(struct net_device *dev) 4430 { 4431 unsigned char nest_level = 0; 4432 4433 #ifdef CONFIG_LOCKDEP 4434 nest_level = dev->nested_level; 4435 #endif 4436 local_bh_disable(); 4437 spin_lock_nested(&dev->addr_list_lock, nest_level); 4438 } 4439 4440 static inline void netif_addr_unlock(struct net_device *dev) 4441 { 4442 spin_unlock(&dev->addr_list_lock); 4443 } 4444 4445 static inline void netif_addr_unlock_bh(struct net_device *dev) 4446 { 4447 spin_unlock_bh(&dev->addr_list_lock); 4448 } 4449 4450 /* 4451 * dev_addrs walker. Should be used only for read access. Call with 4452 * rcu_read_lock held. 4453 */ 4454 #define for_each_dev_addr(dev, ha) \ 4455 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4456 4457 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4458 4459 void ether_setup(struct net_device *dev); 4460 4461 /* Support for loadable net-drivers */ 4462 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4463 unsigned char name_assign_type, 4464 void (*setup)(struct net_device *), 4465 unsigned int txqs, unsigned int rxqs); 4466 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4467 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4468 4469 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4470 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4471 count) 4472 4473 int register_netdev(struct net_device *dev); 4474 void unregister_netdev(struct net_device *dev); 4475 4476 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4477 4478 /* General hardware address lists handling functions */ 4479 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4480 struct netdev_hw_addr_list *from_list, int addr_len); 4481 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4482 struct netdev_hw_addr_list *from_list, int addr_len); 4483 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4484 struct net_device *dev, 4485 int (*sync)(struct net_device *, const unsigned char *), 4486 int (*unsync)(struct net_device *, 4487 const unsigned char *)); 4488 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4489 struct net_device *dev, 4490 int (*sync)(struct net_device *, 4491 const unsigned char *, int), 4492 int (*unsync)(struct net_device *, 4493 const unsigned char *, int)); 4494 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4495 struct net_device *dev, 4496 int (*unsync)(struct net_device *, 4497 const unsigned char *, int)); 4498 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4499 struct net_device *dev, 4500 int (*unsync)(struct net_device *, 4501 const unsigned char *)); 4502 void __hw_addr_init(struct netdev_hw_addr_list *list); 4503 4504 /* Functions used for device addresses handling */ 4505 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4506 const void *addr, size_t len); 4507 4508 static inline void 4509 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4510 { 4511 dev_addr_mod(dev, 0, addr, len); 4512 } 4513 4514 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4515 { 4516 __dev_addr_set(dev, addr, dev->addr_len); 4517 } 4518 4519 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4520 unsigned char addr_type); 4521 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4522 unsigned char addr_type); 4523 4524 /* Functions used for unicast addresses handling */ 4525 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4526 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4527 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4528 int dev_uc_sync(struct net_device *to, struct net_device *from); 4529 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4530 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4531 void dev_uc_flush(struct net_device *dev); 4532 void dev_uc_init(struct net_device *dev); 4533 4534 /** 4535 * __dev_uc_sync - Synchonize device's unicast list 4536 * @dev: device to sync 4537 * @sync: function to call if address should be added 4538 * @unsync: function to call if address should be removed 4539 * 4540 * Add newly added addresses to the interface, and release 4541 * addresses that have been deleted. 4542 */ 4543 static inline int __dev_uc_sync(struct net_device *dev, 4544 int (*sync)(struct net_device *, 4545 const unsigned char *), 4546 int (*unsync)(struct net_device *, 4547 const unsigned char *)) 4548 { 4549 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4550 } 4551 4552 /** 4553 * __dev_uc_unsync - Remove synchronized addresses from device 4554 * @dev: device to sync 4555 * @unsync: function to call if address should be removed 4556 * 4557 * Remove all addresses that were added to the device by dev_uc_sync(). 4558 */ 4559 static inline void __dev_uc_unsync(struct net_device *dev, 4560 int (*unsync)(struct net_device *, 4561 const unsigned char *)) 4562 { 4563 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4564 } 4565 4566 /* Functions used for multicast addresses handling */ 4567 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4568 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4569 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4570 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4571 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4572 int dev_mc_sync(struct net_device *to, struct net_device *from); 4573 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4574 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4575 void dev_mc_flush(struct net_device *dev); 4576 void dev_mc_init(struct net_device *dev); 4577 4578 /** 4579 * __dev_mc_sync - Synchonize device's multicast list 4580 * @dev: device to sync 4581 * @sync: function to call if address should be added 4582 * @unsync: function to call if address should be removed 4583 * 4584 * Add newly added addresses to the interface, and release 4585 * addresses that have been deleted. 4586 */ 4587 static inline int __dev_mc_sync(struct net_device *dev, 4588 int (*sync)(struct net_device *, 4589 const unsigned char *), 4590 int (*unsync)(struct net_device *, 4591 const unsigned char *)) 4592 { 4593 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4594 } 4595 4596 /** 4597 * __dev_mc_unsync - Remove synchronized addresses from device 4598 * @dev: device to sync 4599 * @unsync: function to call if address should be removed 4600 * 4601 * Remove all addresses that were added to the device by dev_mc_sync(). 4602 */ 4603 static inline void __dev_mc_unsync(struct net_device *dev, 4604 int (*unsync)(struct net_device *, 4605 const unsigned char *)) 4606 { 4607 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4608 } 4609 4610 /* Functions used for secondary unicast and multicast support */ 4611 void dev_set_rx_mode(struct net_device *dev); 4612 int dev_set_promiscuity(struct net_device *dev, int inc); 4613 int dev_set_allmulti(struct net_device *dev, int inc); 4614 void netdev_state_change(struct net_device *dev); 4615 void __netdev_notify_peers(struct net_device *dev); 4616 void netdev_notify_peers(struct net_device *dev); 4617 void netdev_features_change(struct net_device *dev); 4618 /* Load a device via the kmod */ 4619 void dev_load(struct net *net, const char *name); 4620 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4621 struct rtnl_link_stats64 *storage); 4622 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4623 const struct net_device_stats *netdev_stats); 4624 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4625 const struct pcpu_sw_netstats __percpu *netstats); 4626 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4627 4628 extern int netdev_max_backlog; 4629 extern int dev_rx_weight; 4630 extern int dev_tx_weight; 4631 extern int gro_normal_batch; 4632 4633 enum { 4634 NESTED_SYNC_IMM_BIT, 4635 NESTED_SYNC_TODO_BIT, 4636 }; 4637 4638 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4639 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4640 4641 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4642 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4643 4644 struct netdev_nested_priv { 4645 unsigned char flags; 4646 void *data; 4647 }; 4648 4649 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4650 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4651 struct list_head **iter); 4652 4653 /* iterate through upper list, must be called under RCU read lock */ 4654 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4655 for (iter = &(dev)->adj_list.upper, \ 4656 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4657 updev; \ 4658 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4659 4660 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4661 int (*fn)(struct net_device *upper_dev, 4662 struct netdev_nested_priv *priv), 4663 struct netdev_nested_priv *priv); 4664 4665 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4666 struct net_device *upper_dev); 4667 4668 bool netdev_has_any_upper_dev(struct net_device *dev); 4669 4670 void *netdev_lower_get_next_private(struct net_device *dev, 4671 struct list_head **iter); 4672 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4673 struct list_head **iter); 4674 4675 #define netdev_for_each_lower_private(dev, priv, iter) \ 4676 for (iter = (dev)->adj_list.lower.next, \ 4677 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4678 priv; \ 4679 priv = netdev_lower_get_next_private(dev, &(iter))) 4680 4681 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4682 for (iter = &(dev)->adj_list.lower, \ 4683 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4684 priv; \ 4685 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4686 4687 void *netdev_lower_get_next(struct net_device *dev, 4688 struct list_head **iter); 4689 4690 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4691 for (iter = (dev)->adj_list.lower.next, \ 4692 ldev = netdev_lower_get_next(dev, &(iter)); \ 4693 ldev; \ 4694 ldev = netdev_lower_get_next(dev, &(iter))) 4695 4696 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4697 struct list_head **iter); 4698 int netdev_walk_all_lower_dev(struct net_device *dev, 4699 int (*fn)(struct net_device *lower_dev, 4700 struct netdev_nested_priv *priv), 4701 struct netdev_nested_priv *priv); 4702 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4703 int (*fn)(struct net_device *lower_dev, 4704 struct netdev_nested_priv *priv), 4705 struct netdev_nested_priv *priv); 4706 4707 void *netdev_adjacent_get_private(struct list_head *adj_list); 4708 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4709 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4710 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4711 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4712 struct netlink_ext_ack *extack); 4713 int netdev_master_upper_dev_link(struct net_device *dev, 4714 struct net_device *upper_dev, 4715 void *upper_priv, void *upper_info, 4716 struct netlink_ext_ack *extack); 4717 void netdev_upper_dev_unlink(struct net_device *dev, 4718 struct net_device *upper_dev); 4719 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4720 struct net_device *new_dev, 4721 struct net_device *dev, 4722 struct netlink_ext_ack *extack); 4723 void netdev_adjacent_change_commit(struct net_device *old_dev, 4724 struct net_device *new_dev, 4725 struct net_device *dev); 4726 void netdev_adjacent_change_abort(struct net_device *old_dev, 4727 struct net_device *new_dev, 4728 struct net_device *dev); 4729 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4730 void *netdev_lower_dev_get_private(struct net_device *dev, 4731 struct net_device *lower_dev); 4732 void netdev_lower_state_changed(struct net_device *lower_dev, 4733 void *lower_state_info); 4734 4735 /* RSS keys are 40 or 52 bytes long */ 4736 #define NETDEV_RSS_KEY_LEN 52 4737 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4738 void netdev_rss_key_fill(void *buffer, size_t len); 4739 4740 int skb_checksum_help(struct sk_buff *skb); 4741 int skb_crc32c_csum_help(struct sk_buff *skb); 4742 int skb_csum_hwoffload_help(struct sk_buff *skb, 4743 const netdev_features_t features); 4744 4745 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4746 netdev_features_t features, bool tx_path); 4747 struct sk_buff *skb_eth_gso_segment(struct sk_buff *skb, 4748 netdev_features_t features, __be16 type); 4749 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4750 netdev_features_t features); 4751 4752 struct netdev_bonding_info { 4753 ifslave slave; 4754 ifbond master; 4755 }; 4756 4757 struct netdev_notifier_bonding_info { 4758 struct netdev_notifier_info info; /* must be first */ 4759 struct netdev_bonding_info bonding_info; 4760 }; 4761 4762 void netdev_bonding_info_change(struct net_device *dev, 4763 struct netdev_bonding_info *bonding_info); 4764 4765 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4766 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4767 #else 4768 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4769 const void *data) 4770 { 4771 } 4772 #endif 4773 4774 static inline 4775 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4776 { 4777 return __skb_gso_segment(skb, features, true); 4778 } 4779 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4780 4781 static inline bool can_checksum_protocol(netdev_features_t features, 4782 __be16 protocol) 4783 { 4784 if (protocol == htons(ETH_P_FCOE)) 4785 return !!(features & NETIF_F_FCOE_CRC); 4786 4787 /* Assume this is an IP checksum (not SCTP CRC) */ 4788 4789 if (features & NETIF_F_HW_CSUM) { 4790 /* Can checksum everything */ 4791 return true; 4792 } 4793 4794 switch (protocol) { 4795 case htons(ETH_P_IP): 4796 return !!(features & NETIF_F_IP_CSUM); 4797 case htons(ETH_P_IPV6): 4798 return !!(features & NETIF_F_IPV6_CSUM); 4799 default: 4800 return false; 4801 } 4802 } 4803 4804 #ifdef CONFIG_BUG 4805 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4806 #else 4807 static inline void netdev_rx_csum_fault(struct net_device *dev, 4808 struct sk_buff *skb) 4809 { 4810 } 4811 #endif 4812 /* rx skb timestamps */ 4813 void net_enable_timestamp(void); 4814 void net_disable_timestamp(void); 4815 4816 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 4817 const struct skb_shared_hwtstamps *hwtstamps, 4818 bool cycles) 4819 { 4820 const struct net_device_ops *ops = dev->netdev_ops; 4821 4822 if (ops->ndo_get_tstamp) 4823 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 4824 4825 return hwtstamps->hwtstamp; 4826 } 4827 4828 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4829 struct sk_buff *skb, struct net_device *dev, 4830 bool more) 4831 { 4832 __this_cpu_write(softnet_data.xmit.more, more); 4833 return ops->ndo_start_xmit(skb, dev); 4834 } 4835 4836 static inline bool netdev_xmit_more(void) 4837 { 4838 return __this_cpu_read(softnet_data.xmit.more); 4839 } 4840 4841 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4842 struct netdev_queue *txq, bool more) 4843 { 4844 const struct net_device_ops *ops = dev->netdev_ops; 4845 netdev_tx_t rc; 4846 4847 rc = __netdev_start_xmit(ops, skb, dev, more); 4848 if (rc == NETDEV_TX_OK) 4849 txq_trans_update(txq); 4850 4851 return rc; 4852 } 4853 4854 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4855 const void *ns); 4856 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4857 const void *ns); 4858 4859 extern const struct kobj_ns_type_operations net_ns_type_operations; 4860 4861 const char *netdev_drivername(const struct net_device *dev); 4862 4863 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4864 netdev_features_t f2) 4865 { 4866 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4867 if (f1 & NETIF_F_HW_CSUM) 4868 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4869 else 4870 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4871 } 4872 4873 return f1 & f2; 4874 } 4875 4876 static inline netdev_features_t netdev_get_wanted_features( 4877 struct net_device *dev) 4878 { 4879 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4880 } 4881 netdev_features_t netdev_increment_features(netdev_features_t all, 4882 netdev_features_t one, netdev_features_t mask); 4883 4884 /* Allow TSO being used on stacked device : 4885 * Performing the GSO segmentation before last device 4886 * is a performance improvement. 4887 */ 4888 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4889 netdev_features_t mask) 4890 { 4891 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4892 } 4893 4894 int __netdev_update_features(struct net_device *dev); 4895 void netdev_update_features(struct net_device *dev); 4896 void netdev_change_features(struct net_device *dev); 4897 4898 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4899 struct net_device *dev); 4900 4901 netdev_features_t passthru_features_check(struct sk_buff *skb, 4902 struct net_device *dev, 4903 netdev_features_t features); 4904 netdev_features_t netif_skb_features(struct sk_buff *skb); 4905 4906 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4907 { 4908 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4909 4910 /* check flags correspondence */ 4911 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4912 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4913 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4914 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4915 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4916 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4917 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4918 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4919 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4920 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4921 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4922 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4923 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4924 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4925 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4926 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4927 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4928 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4929 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4930 4931 return (features & feature) == feature; 4932 } 4933 4934 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4935 { 4936 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4937 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4938 } 4939 4940 static inline bool netif_needs_gso(struct sk_buff *skb, 4941 netdev_features_t features) 4942 { 4943 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4944 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4945 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4946 } 4947 4948 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 4949 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 4950 void netif_inherit_tso_max(struct net_device *to, 4951 const struct net_device *from); 4952 4953 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4954 int pulled_hlen, u16 mac_offset, 4955 int mac_len) 4956 { 4957 skb->protocol = protocol; 4958 skb->encapsulation = 1; 4959 skb_push(skb, pulled_hlen); 4960 skb_reset_transport_header(skb); 4961 skb->mac_header = mac_offset; 4962 skb->network_header = skb->mac_header + mac_len; 4963 skb->mac_len = mac_len; 4964 } 4965 4966 static inline bool netif_is_macsec(const struct net_device *dev) 4967 { 4968 return dev->priv_flags & IFF_MACSEC; 4969 } 4970 4971 static inline bool netif_is_macvlan(const struct net_device *dev) 4972 { 4973 return dev->priv_flags & IFF_MACVLAN; 4974 } 4975 4976 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4977 { 4978 return dev->priv_flags & IFF_MACVLAN_PORT; 4979 } 4980 4981 static inline bool netif_is_bond_master(const struct net_device *dev) 4982 { 4983 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4984 } 4985 4986 static inline bool netif_is_bond_slave(const struct net_device *dev) 4987 { 4988 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4989 } 4990 4991 static inline bool netif_supports_nofcs(struct net_device *dev) 4992 { 4993 return dev->priv_flags & IFF_SUPP_NOFCS; 4994 } 4995 4996 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4997 { 4998 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4999 } 5000 5001 static inline bool netif_is_l3_master(const struct net_device *dev) 5002 { 5003 return dev->priv_flags & IFF_L3MDEV_MASTER; 5004 } 5005 5006 static inline bool netif_is_l3_slave(const struct net_device *dev) 5007 { 5008 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5009 } 5010 5011 static inline bool netif_is_bridge_master(const struct net_device *dev) 5012 { 5013 return dev->priv_flags & IFF_EBRIDGE; 5014 } 5015 5016 static inline bool netif_is_bridge_port(const struct net_device *dev) 5017 { 5018 return dev->priv_flags & IFF_BRIDGE_PORT; 5019 } 5020 5021 static inline bool netif_is_ovs_master(const struct net_device *dev) 5022 { 5023 return dev->priv_flags & IFF_OPENVSWITCH; 5024 } 5025 5026 static inline bool netif_is_ovs_port(const struct net_device *dev) 5027 { 5028 return dev->priv_flags & IFF_OVS_DATAPATH; 5029 } 5030 5031 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5032 { 5033 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5034 } 5035 5036 static inline bool netif_is_team_master(const struct net_device *dev) 5037 { 5038 return dev->priv_flags & IFF_TEAM; 5039 } 5040 5041 static inline bool netif_is_team_port(const struct net_device *dev) 5042 { 5043 return dev->priv_flags & IFF_TEAM_PORT; 5044 } 5045 5046 static inline bool netif_is_lag_master(const struct net_device *dev) 5047 { 5048 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5049 } 5050 5051 static inline bool netif_is_lag_port(const struct net_device *dev) 5052 { 5053 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5054 } 5055 5056 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5057 { 5058 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5059 } 5060 5061 static inline bool netif_is_failover(const struct net_device *dev) 5062 { 5063 return dev->priv_flags & IFF_FAILOVER; 5064 } 5065 5066 static inline bool netif_is_failover_slave(const struct net_device *dev) 5067 { 5068 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5069 } 5070 5071 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5072 static inline void netif_keep_dst(struct net_device *dev) 5073 { 5074 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5075 } 5076 5077 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5078 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5079 { 5080 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5081 return netif_is_macsec(dev); 5082 } 5083 5084 extern struct pernet_operations __net_initdata loopback_net_ops; 5085 5086 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5087 5088 /* netdev_printk helpers, similar to dev_printk */ 5089 5090 static inline const char *netdev_name(const struct net_device *dev) 5091 { 5092 if (!dev->name[0] || strchr(dev->name, '%')) 5093 return "(unnamed net_device)"; 5094 return dev->name; 5095 } 5096 5097 static inline bool netdev_unregistering(const struct net_device *dev) 5098 { 5099 return dev->reg_state == NETREG_UNREGISTERING; 5100 } 5101 5102 static inline const char *netdev_reg_state(const struct net_device *dev) 5103 { 5104 switch (dev->reg_state) { 5105 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5106 case NETREG_REGISTERED: return ""; 5107 case NETREG_UNREGISTERING: return " (unregistering)"; 5108 case NETREG_UNREGISTERED: return " (unregistered)"; 5109 case NETREG_RELEASED: return " (released)"; 5110 case NETREG_DUMMY: return " (dummy)"; 5111 } 5112 5113 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5114 return " (unknown)"; 5115 } 5116 5117 #define MODULE_ALIAS_NETDEV(device) \ 5118 MODULE_ALIAS("netdev-" device) 5119 5120 /* 5121 * netdev_WARN() acts like dev_printk(), but with the key difference 5122 * of using a WARN/WARN_ON to get the message out, including the 5123 * file/line information and a backtrace. 5124 */ 5125 #define netdev_WARN(dev, format, args...) \ 5126 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5127 netdev_reg_state(dev), ##args) 5128 5129 #define netdev_WARN_ONCE(dev, format, args...) \ 5130 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5131 netdev_reg_state(dev), ##args) 5132 5133 /* 5134 * The list of packet types we will receive (as opposed to discard) 5135 * and the routines to invoke. 5136 * 5137 * Why 16. Because with 16 the only overlap we get on a hash of the 5138 * low nibble of the protocol value is RARP/SNAP/X.25. 5139 * 5140 * 0800 IP 5141 * 0001 802.3 5142 * 0002 AX.25 5143 * 0004 802.2 5144 * 8035 RARP 5145 * 0005 SNAP 5146 * 0805 X.25 5147 * 0806 ARP 5148 * 8137 IPX 5149 * 0009 Localtalk 5150 * 86DD IPv6 5151 */ 5152 #define PTYPE_HASH_SIZE (16) 5153 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5154 5155 extern struct list_head ptype_all __read_mostly; 5156 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5157 5158 extern struct net_device *blackhole_netdev; 5159 5160 #endif /* _LINUX_NETDEVICE_H */ 5161