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, <waltje@uWalt.NL.Mugnet.ORG> 13 * Corey Minyard <wf-rch!minyard@relay.EU.net> 14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov> 15 * Alan Cox, <alan@lxorguk.ukuu.org.uk> 16 * Bjorn Ekwall. <bj0rn@blox.se> 17 * Pekka Riikonen <priikone@poseidon.pspt.fi> 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 do { 550 val = READ_ONCE(n->state); 551 if (val & NAPIF_STATE_DISABLE) 552 return true; 553 554 if (!(val & NAPIF_STATE_SCHED)) 555 return false; 556 557 new = val | NAPIF_STATE_MISSED; 558 } while (cmpxchg(&n->state, val, new) != val); 559 560 return true; 561 } 562 563 enum netdev_queue_state_t { 564 __QUEUE_STATE_DRV_XOFF, 565 __QUEUE_STATE_STACK_XOFF, 566 __QUEUE_STATE_FROZEN, 567 }; 568 569 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 570 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 571 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 572 573 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 574 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 575 QUEUE_STATE_FROZEN) 576 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 577 QUEUE_STATE_FROZEN) 578 579 /* 580 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 581 * netif_tx_* functions below are used to manipulate this flag. The 582 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 583 * queue independently. The netif_xmit_*stopped functions below are called 584 * to check if the queue has been stopped by the driver or stack (either 585 * of the XOFF bits are set in the state). Drivers should not need to call 586 * netif_xmit*stopped functions, they should only be using netif_tx_*. 587 */ 588 589 struct netdev_queue { 590 /* 591 * read-mostly part 592 */ 593 struct net_device *dev; 594 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 * @dn_ptr: DECnet specific data 1855 * @ip6_ptr: IPv6 specific data 1856 * @ax25_ptr: AX.25 specific data 1857 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1858 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1859 * device struct 1860 * @mpls_ptr: mpls_dev struct pointer 1861 * @mctp_ptr: MCTP specific data 1862 * 1863 * @dev_addr: Hw address (before bcast, 1864 * because most packets are unicast) 1865 * 1866 * @_rx: Array of RX queues 1867 * @num_rx_queues: Number of RX queues 1868 * allocated at register_netdev() time 1869 * @real_num_rx_queues: Number of RX queues currently active in device 1870 * @xdp_prog: XDP sockets filter program pointer 1871 * @gro_flush_timeout: timeout for GRO layer in NAPI 1872 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1873 * allow to avoid NIC hard IRQ, on busy queues. 1874 * 1875 * @rx_handler: handler for received packets 1876 * @rx_handler_data: XXX: need comments on this one 1877 * @miniq_ingress: ingress/clsact qdisc specific data for 1878 * ingress processing 1879 * @ingress_queue: XXX: need comments on this one 1880 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1881 * @broadcast: hw bcast address 1882 * 1883 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1884 * indexed by RX queue number. Assigned by driver. 1885 * This must only be set if the ndo_rx_flow_steer 1886 * operation is defined 1887 * @index_hlist: Device index hash chain 1888 * 1889 * @_tx: Array of TX queues 1890 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1891 * @real_num_tx_queues: Number of TX queues currently active in device 1892 * @qdisc: Root qdisc from userspace point of view 1893 * @tx_queue_len: Max frames per queue allowed 1894 * @tx_global_lock: XXX: need comments on this one 1895 * @xdp_bulkq: XDP device bulk queue 1896 * @xps_maps: all CPUs/RXQs maps for XPS device 1897 * 1898 * @xps_maps: XXX: need comments on this one 1899 * @miniq_egress: clsact qdisc specific data for 1900 * egress processing 1901 * @nf_hooks_egress: netfilter hooks executed for egress packets 1902 * @qdisc_hash: qdisc hash table 1903 * @watchdog_timeo: Represents the timeout that is used by 1904 * the watchdog (see dev_watchdog()) 1905 * @watchdog_timer: List of timers 1906 * 1907 * @proto_down_reason: reason a netdev interface is held down 1908 * @pcpu_refcnt: Number of references to this device 1909 * @dev_refcnt: Number of references to this device 1910 * @refcnt_tracker: Tracker directory for tracked references to this device 1911 * @todo_list: Delayed register/unregister 1912 * @link_watch_list: XXX: need comments on this one 1913 * 1914 * @reg_state: Register/unregister state machine 1915 * @dismantle: Device is going to be freed 1916 * @rtnl_link_state: This enum represents the phases of creating 1917 * a new link 1918 * 1919 * @needs_free_netdev: Should unregister perform free_netdev? 1920 * @priv_destructor: Called from unregister 1921 * @npinfo: XXX: need comments on this one 1922 * @nd_net: Network namespace this network device is inside 1923 * 1924 * @ml_priv: Mid-layer private 1925 * @ml_priv_type: Mid-layer private type 1926 * @lstats: Loopback statistics 1927 * @tstats: Tunnel statistics 1928 * @dstats: Dummy statistics 1929 * @vstats: Virtual ethernet statistics 1930 * 1931 * @garp_port: GARP 1932 * @mrp_port: MRP 1933 * 1934 * @dm_private: Drop monitor private 1935 * 1936 * @dev: Class/net/name entry 1937 * @sysfs_groups: Space for optional device, statistics and wireless 1938 * sysfs groups 1939 * 1940 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1941 * @rtnl_link_ops: Rtnl_link_ops 1942 * 1943 * @gso_max_size: Maximum size of generic segmentation offload 1944 * @tso_max_size: Device (as in HW) limit on the max TSO request size 1945 * @gso_max_segs: Maximum number of segments that can be passed to the 1946 * NIC for GSO 1947 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count 1948 * 1949 * @dcbnl_ops: Data Center Bridging netlink ops 1950 * @num_tc: Number of traffic classes in the net device 1951 * @tc_to_txq: XXX: need comments on this one 1952 * @prio_tc_map: XXX: need comments on this one 1953 * 1954 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1955 * 1956 * @priomap: XXX: need comments on this one 1957 * @phydev: Physical device may attach itself 1958 * for hardware timestamping 1959 * @sfp_bus: attached &struct sfp_bus structure. 1960 * 1961 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1962 * 1963 * @proto_down: protocol port state information can be sent to the 1964 * switch driver and used to set the phys state of the 1965 * switch port. 1966 * 1967 * @wol_enabled: Wake-on-LAN is enabled 1968 * 1969 * @threaded: napi threaded mode is enabled 1970 * 1971 * @net_notifier_list: List of per-net netdev notifier block 1972 * that follow this device when it is moved 1973 * to another network namespace. 1974 * 1975 * @macsec_ops: MACsec offloading ops 1976 * 1977 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1978 * offload capabilities of the device 1979 * @udp_tunnel_nic: UDP tunnel offload state 1980 * @xdp_state: stores info on attached XDP BPF programs 1981 * 1982 * @nested_level: Used as a parameter of spin_lock_nested() of 1983 * dev->addr_list_lock. 1984 * @unlink_list: As netif_addr_lock() can be called recursively, 1985 * keep a list of interfaces to be deleted. 1986 * @gro_max_size: Maximum size of aggregated packet in generic 1987 * receive offload (GRO) 1988 * 1989 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 1990 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 1991 * @watchdog_dev_tracker: refcount tracker used by watchdog. 1992 * @dev_registered_tracker: tracker for reference held while 1993 * registered 1994 * @offload_xstats_l3: L3 HW stats for this netdevice. 1995 * 1996 * FIXME: cleanup struct net_device such that network protocol info 1997 * moves out. 1998 */ 1999 2000 struct net_device { 2001 char name[IFNAMSIZ]; 2002 struct netdev_name_node *name_node; 2003 struct dev_ifalias __rcu *ifalias; 2004 /* 2005 * I/O specific fields 2006 * FIXME: Merge these and struct ifmap into one 2007 */ 2008 unsigned long mem_end; 2009 unsigned long mem_start; 2010 unsigned long base_addr; 2011 2012 /* 2013 * Some hardware also needs these fields (state,dev_list, 2014 * napi_list,unreg_list,close_list) but they are not 2015 * part of the usual set specified in Space.c. 2016 */ 2017 2018 unsigned long state; 2019 2020 struct list_head dev_list; 2021 struct list_head napi_list; 2022 struct list_head unreg_list; 2023 struct list_head close_list; 2024 struct list_head ptype_all; 2025 struct list_head ptype_specific; 2026 2027 struct { 2028 struct list_head upper; 2029 struct list_head lower; 2030 } adj_list; 2031 2032 /* Read-mostly cache-line for fast-path access */ 2033 unsigned int flags; 2034 unsigned long long priv_flags; 2035 const struct net_device_ops *netdev_ops; 2036 int ifindex; 2037 unsigned short gflags; 2038 unsigned short hard_header_len; 2039 2040 /* Note : dev->mtu is often read without holding a lock. 2041 * Writers usually hold RTNL. 2042 * It is recommended to use READ_ONCE() to annotate the reads, 2043 * and to use WRITE_ONCE() to annotate the writes. 2044 */ 2045 unsigned int mtu; 2046 unsigned short needed_headroom; 2047 unsigned short needed_tailroom; 2048 2049 netdev_features_t features; 2050 netdev_features_t hw_features; 2051 netdev_features_t wanted_features; 2052 netdev_features_t vlan_features; 2053 netdev_features_t hw_enc_features; 2054 netdev_features_t mpls_features; 2055 netdev_features_t gso_partial_features; 2056 2057 unsigned int min_mtu; 2058 unsigned int max_mtu; 2059 unsigned short type; 2060 unsigned char min_header_len; 2061 unsigned char name_assign_type; 2062 2063 int group; 2064 2065 struct net_device_stats stats; /* not used by modern drivers */ 2066 2067 struct net_device_core_stats __percpu *core_stats; 2068 2069 /* Stats to monitor link on/off, flapping */ 2070 atomic_t carrier_up_count; 2071 atomic_t carrier_down_count; 2072 2073 #ifdef CONFIG_WIRELESS_EXT 2074 const struct iw_handler_def *wireless_handlers; 2075 struct iw_public_data *wireless_data; 2076 #endif 2077 const struct ethtool_ops *ethtool_ops; 2078 #ifdef CONFIG_NET_L3_MASTER_DEV 2079 const struct l3mdev_ops *l3mdev_ops; 2080 #endif 2081 #if IS_ENABLED(CONFIG_IPV6) 2082 const struct ndisc_ops *ndisc_ops; 2083 #endif 2084 2085 #ifdef CONFIG_XFRM_OFFLOAD 2086 const struct xfrmdev_ops *xfrmdev_ops; 2087 #endif 2088 2089 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2090 const struct tlsdev_ops *tlsdev_ops; 2091 #endif 2092 2093 const struct header_ops *header_ops; 2094 2095 unsigned char operstate; 2096 unsigned char link_mode; 2097 2098 unsigned char if_port; 2099 unsigned char dma; 2100 2101 /* Interface address info. */ 2102 unsigned char perm_addr[MAX_ADDR_LEN]; 2103 unsigned char addr_assign_type; 2104 unsigned char addr_len; 2105 unsigned char upper_level; 2106 unsigned char lower_level; 2107 2108 unsigned short neigh_priv_len; 2109 unsigned short dev_id; 2110 unsigned short dev_port; 2111 unsigned short padded; 2112 2113 spinlock_t addr_list_lock; 2114 int irq; 2115 2116 struct netdev_hw_addr_list uc; 2117 struct netdev_hw_addr_list mc; 2118 struct netdev_hw_addr_list dev_addrs; 2119 2120 #ifdef CONFIG_SYSFS 2121 struct kset *queues_kset; 2122 #endif 2123 #ifdef CONFIG_LOCKDEP 2124 struct list_head unlink_list; 2125 #endif 2126 unsigned int promiscuity; 2127 unsigned int allmulti; 2128 bool uc_promisc; 2129 #ifdef CONFIG_LOCKDEP 2130 unsigned char nested_level; 2131 #endif 2132 2133 2134 /* Protocol-specific pointers */ 2135 2136 struct in_device __rcu *ip_ptr; 2137 struct inet6_dev __rcu *ip6_ptr; 2138 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2139 struct vlan_info __rcu *vlan_info; 2140 #endif 2141 #if IS_ENABLED(CONFIG_NET_DSA) 2142 struct dsa_port *dsa_ptr; 2143 #endif 2144 #if IS_ENABLED(CONFIG_TIPC) 2145 struct tipc_bearer __rcu *tipc_ptr; 2146 #endif 2147 #if IS_ENABLED(CONFIG_ATALK) 2148 void *atalk_ptr; 2149 #endif 2150 #if IS_ENABLED(CONFIG_DECNET) 2151 struct dn_dev __rcu *dn_ptr; 2152 #endif 2153 #if IS_ENABLED(CONFIG_AX25) 2154 void *ax25_ptr; 2155 #endif 2156 #if IS_ENABLED(CONFIG_CFG80211) 2157 struct wireless_dev *ieee80211_ptr; 2158 #endif 2159 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2160 struct wpan_dev *ieee802154_ptr; 2161 #endif 2162 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2163 struct mpls_dev __rcu *mpls_ptr; 2164 #endif 2165 #if IS_ENABLED(CONFIG_MCTP) 2166 struct mctp_dev __rcu *mctp_ptr; 2167 #endif 2168 2169 /* 2170 * Cache lines mostly used on receive path (including eth_type_trans()) 2171 */ 2172 /* Interface address info used in eth_type_trans() */ 2173 const unsigned char *dev_addr; 2174 2175 struct netdev_rx_queue *_rx; 2176 unsigned int num_rx_queues; 2177 unsigned int real_num_rx_queues; 2178 2179 struct bpf_prog __rcu *xdp_prog; 2180 unsigned long gro_flush_timeout; 2181 int napi_defer_hard_irqs; 2182 #define GRO_LEGACY_MAX_SIZE 65536u 2183 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2184 * and shinfo->gso_segs is a 16bit field. 2185 */ 2186 #define GRO_MAX_SIZE (8 * 65535u) 2187 unsigned int gro_max_size; 2188 rx_handler_func_t __rcu *rx_handler; 2189 void __rcu *rx_handler_data; 2190 2191 #ifdef CONFIG_NET_CLS_ACT 2192 struct mini_Qdisc __rcu *miniq_ingress; 2193 #endif 2194 struct netdev_queue __rcu *ingress_queue; 2195 #ifdef CONFIG_NETFILTER_INGRESS 2196 struct nf_hook_entries __rcu *nf_hooks_ingress; 2197 #endif 2198 2199 unsigned char broadcast[MAX_ADDR_LEN]; 2200 #ifdef CONFIG_RFS_ACCEL 2201 struct cpu_rmap *rx_cpu_rmap; 2202 #endif 2203 struct hlist_node index_hlist; 2204 2205 /* 2206 * Cache lines mostly used on transmit path 2207 */ 2208 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2209 unsigned int num_tx_queues; 2210 unsigned int real_num_tx_queues; 2211 struct Qdisc __rcu *qdisc; 2212 unsigned int tx_queue_len; 2213 spinlock_t tx_global_lock; 2214 2215 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2216 2217 #ifdef CONFIG_XPS 2218 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2219 #endif 2220 #ifdef CONFIG_NET_CLS_ACT 2221 struct mini_Qdisc __rcu *miniq_egress; 2222 #endif 2223 #ifdef CONFIG_NETFILTER_EGRESS 2224 struct nf_hook_entries __rcu *nf_hooks_egress; 2225 #endif 2226 2227 #ifdef CONFIG_NET_SCHED 2228 DECLARE_HASHTABLE (qdisc_hash, 4); 2229 #endif 2230 /* These may be needed for future network-power-down code. */ 2231 struct timer_list watchdog_timer; 2232 int watchdog_timeo; 2233 2234 u32 proto_down_reason; 2235 2236 struct list_head todo_list; 2237 2238 #ifdef CONFIG_PCPU_DEV_REFCNT 2239 int __percpu *pcpu_refcnt; 2240 #else 2241 refcount_t dev_refcnt; 2242 #endif 2243 struct ref_tracker_dir refcnt_tracker; 2244 2245 struct list_head link_watch_list; 2246 2247 enum { NETREG_UNINITIALIZED=0, 2248 NETREG_REGISTERED, /* completed register_netdevice */ 2249 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2250 NETREG_UNREGISTERED, /* completed unregister todo */ 2251 NETREG_RELEASED, /* called free_netdev */ 2252 NETREG_DUMMY, /* dummy device for NAPI poll */ 2253 } reg_state:8; 2254 2255 bool dismantle; 2256 2257 enum { 2258 RTNL_LINK_INITIALIZED, 2259 RTNL_LINK_INITIALIZING, 2260 } rtnl_link_state:16; 2261 2262 bool needs_free_netdev; 2263 void (*priv_destructor)(struct net_device *dev); 2264 2265 #ifdef CONFIG_NETPOLL 2266 struct netpoll_info __rcu *npinfo; 2267 #endif 2268 2269 possible_net_t nd_net; 2270 2271 /* mid-layer private */ 2272 void *ml_priv; 2273 enum netdev_ml_priv_type ml_priv_type; 2274 2275 union { 2276 struct pcpu_lstats __percpu *lstats; 2277 struct pcpu_sw_netstats __percpu *tstats; 2278 struct pcpu_dstats __percpu *dstats; 2279 }; 2280 2281 #if IS_ENABLED(CONFIG_GARP) 2282 struct garp_port __rcu *garp_port; 2283 #endif 2284 #if IS_ENABLED(CONFIG_MRP) 2285 struct mrp_port __rcu *mrp_port; 2286 #endif 2287 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2288 struct dm_hw_stat_delta __rcu *dm_private; 2289 #endif 2290 struct device dev; 2291 const struct attribute_group *sysfs_groups[4]; 2292 const struct attribute_group *sysfs_rx_queue_group; 2293 2294 const struct rtnl_link_ops *rtnl_link_ops; 2295 2296 /* for setting kernel sock attribute on TCP connection setup */ 2297 #define GSO_MAX_SEGS 65535u 2298 #define GSO_LEGACY_MAX_SIZE 65536u 2299 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2300 * and shinfo->gso_segs is a 16bit field. 2301 */ 2302 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2303 2304 unsigned int gso_max_size; 2305 #define TSO_LEGACY_MAX_SIZE 65536 2306 #define TSO_MAX_SIZE UINT_MAX 2307 unsigned int tso_max_size; 2308 u16 gso_max_segs; 2309 #define TSO_MAX_SEGS U16_MAX 2310 u16 tso_max_segs; 2311 2312 #ifdef CONFIG_DCB 2313 const struct dcbnl_rtnl_ops *dcbnl_ops; 2314 #endif 2315 s16 num_tc; 2316 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2317 u8 prio_tc_map[TC_BITMASK + 1]; 2318 2319 #if IS_ENABLED(CONFIG_FCOE) 2320 unsigned int fcoe_ddp_xid; 2321 #endif 2322 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2323 struct netprio_map __rcu *priomap; 2324 #endif 2325 struct phy_device *phydev; 2326 struct sfp_bus *sfp_bus; 2327 struct lock_class_key *qdisc_tx_busylock; 2328 bool proto_down; 2329 unsigned wol_enabled:1; 2330 unsigned threaded:1; 2331 2332 struct list_head net_notifier_list; 2333 2334 #if IS_ENABLED(CONFIG_MACSEC) 2335 /* MACsec management functions */ 2336 const struct macsec_ops *macsec_ops; 2337 #endif 2338 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2339 struct udp_tunnel_nic *udp_tunnel_nic; 2340 2341 /* protected by rtnl_lock */ 2342 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2343 2344 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2345 netdevice_tracker linkwatch_dev_tracker; 2346 netdevice_tracker watchdog_dev_tracker; 2347 netdevice_tracker dev_registered_tracker; 2348 struct rtnl_hw_stats64 *offload_xstats_l3; 2349 }; 2350 #define to_net_dev(d) container_of(d, struct net_device, dev) 2351 2352 static inline bool netif_elide_gro(const struct net_device *dev) 2353 { 2354 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2355 return true; 2356 return false; 2357 } 2358 2359 #define NETDEV_ALIGN 32 2360 2361 static inline 2362 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2363 { 2364 return dev->prio_tc_map[prio & TC_BITMASK]; 2365 } 2366 2367 static inline 2368 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2369 { 2370 if (tc >= dev->num_tc) 2371 return -EINVAL; 2372 2373 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2374 return 0; 2375 } 2376 2377 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2378 void netdev_reset_tc(struct net_device *dev); 2379 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2380 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2381 2382 static inline 2383 int netdev_get_num_tc(struct net_device *dev) 2384 { 2385 return dev->num_tc; 2386 } 2387 2388 static inline void net_prefetch(void *p) 2389 { 2390 prefetch(p); 2391 #if L1_CACHE_BYTES < 128 2392 prefetch((u8 *)p + L1_CACHE_BYTES); 2393 #endif 2394 } 2395 2396 static inline void net_prefetchw(void *p) 2397 { 2398 prefetchw(p); 2399 #if L1_CACHE_BYTES < 128 2400 prefetchw((u8 *)p + L1_CACHE_BYTES); 2401 #endif 2402 } 2403 2404 void netdev_unbind_sb_channel(struct net_device *dev, 2405 struct net_device *sb_dev); 2406 int netdev_bind_sb_channel_queue(struct net_device *dev, 2407 struct net_device *sb_dev, 2408 u8 tc, u16 count, u16 offset); 2409 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2410 static inline int netdev_get_sb_channel(struct net_device *dev) 2411 { 2412 return max_t(int, -dev->num_tc, 0); 2413 } 2414 2415 static inline 2416 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2417 unsigned int index) 2418 { 2419 return &dev->_tx[index]; 2420 } 2421 2422 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2423 const struct sk_buff *skb) 2424 { 2425 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2426 } 2427 2428 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2429 void (*f)(struct net_device *, 2430 struct netdev_queue *, 2431 void *), 2432 void *arg) 2433 { 2434 unsigned int i; 2435 2436 for (i = 0; i < dev->num_tx_queues; i++) 2437 f(dev, &dev->_tx[i], arg); 2438 } 2439 2440 #define netdev_lockdep_set_classes(dev) \ 2441 { \ 2442 static struct lock_class_key qdisc_tx_busylock_key; \ 2443 static struct lock_class_key qdisc_xmit_lock_key; \ 2444 static struct lock_class_key dev_addr_list_lock_key; \ 2445 unsigned int i; \ 2446 \ 2447 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2448 lockdep_set_class(&(dev)->addr_list_lock, \ 2449 &dev_addr_list_lock_key); \ 2450 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2451 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2452 &qdisc_xmit_lock_key); \ 2453 } 2454 2455 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2456 struct net_device *sb_dev); 2457 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2458 struct sk_buff *skb, 2459 struct net_device *sb_dev); 2460 2461 /* returns the headroom that the master device needs to take in account 2462 * when forwarding to this dev 2463 */ 2464 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2465 { 2466 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2467 } 2468 2469 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2470 { 2471 if (dev->netdev_ops->ndo_set_rx_headroom) 2472 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2473 } 2474 2475 /* set the device rx headroom to the dev's default */ 2476 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2477 { 2478 netdev_set_rx_headroom(dev, -1); 2479 } 2480 2481 static inline void *netdev_get_ml_priv(struct net_device *dev, 2482 enum netdev_ml_priv_type type) 2483 { 2484 if (dev->ml_priv_type != type) 2485 return NULL; 2486 2487 return dev->ml_priv; 2488 } 2489 2490 static inline void netdev_set_ml_priv(struct net_device *dev, 2491 void *ml_priv, 2492 enum netdev_ml_priv_type type) 2493 { 2494 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2495 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2496 dev->ml_priv_type, type); 2497 WARN(!dev->ml_priv_type && dev->ml_priv, 2498 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2499 2500 dev->ml_priv = ml_priv; 2501 dev->ml_priv_type = type; 2502 } 2503 2504 /* 2505 * Net namespace inlines 2506 */ 2507 static inline 2508 struct net *dev_net(const struct net_device *dev) 2509 { 2510 return read_pnet(&dev->nd_net); 2511 } 2512 2513 static inline 2514 void dev_net_set(struct net_device *dev, struct net *net) 2515 { 2516 write_pnet(&dev->nd_net, net); 2517 } 2518 2519 /** 2520 * netdev_priv - access network device private data 2521 * @dev: network device 2522 * 2523 * Get network device private data 2524 */ 2525 static inline void *netdev_priv(const struct net_device *dev) 2526 { 2527 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2528 } 2529 2530 /* Set the sysfs physical device reference for the network logical device 2531 * if set prior to registration will cause a symlink during initialization. 2532 */ 2533 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2534 2535 /* Set the sysfs device type for the network logical device to allow 2536 * fine-grained identification of different network device types. For 2537 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2538 */ 2539 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2540 2541 /* Default NAPI poll() weight 2542 * Device drivers are strongly advised to not use bigger value 2543 */ 2544 #define NAPI_POLL_WEIGHT 64 2545 2546 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2547 int (*poll)(struct napi_struct *, int), int weight); 2548 2549 /** 2550 * netif_napi_add() - initialize a NAPI context 2551 * @dev: network device 2552 * @napi: NAPI context 2553 * @poll: polling function 2554 * @weight: default weight 2555 * 2556 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2557 * *any* of the other NAPI-related functions. 2558 */ 2559 static inline void 2560 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2561 int (*poll)(struct napi_struct *, int), int weight) 2562 { 2563 netif_napi_add_weight(dev, napi, poll, weight); 2564 } 2565 2566 static inline void 2567 netif_napi_add_tx_weight(struct net_device *dev, 2568 struct napi_struct *napi, 2569 int (*poll)(struct napi_struct *, int), 2570 int weight) 2571 { 2572 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2573 netif_napi_add_weight(dev, napi, poll, weight); 2574 } 2575 2576 #define netif_tx_napi_add netif_napi_add_tx_weight 2577 2578 /** 2579 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2580 * @dev: network device 2581 * @napi: NAPI context 2582 * @poll: polling function 2583 * 2584 * This variant of netif_napi_add() should be used from drivers using NAPI 2585 * to exclusively poll a TX queue. 2586 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2587 */ 2588 static inline void netif_napi_add_tx(struct net_device *dev, 2589 struct napi_struct *napi, 2590 int (*poll)(struct napi_struct *, int)) 2591 { 2592 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2593 } 2594 2595 /** 2596 * __netif_napi_del - remove a NAPI context 2597 * @napi: NAPI context 2598 * 2599 * Warning: caller must observe RCU grace period before freeing memory 2600 * containing @napi. Drivers might want to call this helper to combine 2601 * all the needed RCU grace periods into a single one. 2602 */ 2603 void __netif_napi_del(struct napi_struct *napi); 2604 2605 /** 2606 * netif_napi_del - remove a NAPI context 2607 * @napi: NAPI context 2608 * 2609 * netif_napi_del() removes a NAPI context from the network device NAPI list 2610 */ 2611 static inline void netif_napi_del(struct napi_struct *napi) 2612 { 2613 __netif_napi_del(napi); 2614 synchronize_net(); 2615 } 2616 2617 struct packet_type { 2618 __be16 type; /* This is really htons(ether_type). */ 2619 bool ignore_outgoing; 2620 struct net_device *dev; /* NULL is wildcarded here */ 2621 netdevice_tracker dev_tracker; 2622 int (*func) (struct sk_buff *, 2623 struct net_device *, 2624 struct packet_type *, 2625 struct net_device *); 2626 void (*list_func) (struct list_head *, 2627 struct packet_type *, 2628 struct net_device *); 2629 bool (*id_match)(struct packet_type *ptype, 2630 struct sock *sk); 2631 struct net *af_packet_net; 2632 void *af_packet_priv; 2633 struct list_head list; 2634 }; 2635 2636 struct offload_callbacks { 2637 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2638 netdev_features_t features); 2639 struct sk_buff *(*gro_receive)(struct list_head *head, 2640 struct sk_buff *skb); 2641 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2642 }; 2643 2644 struct packet_offload { 2645 __be16 type; /* This is really htons(ether_type). */ 2646 u16 priority; 2647 struct offload_callbacks callbacks; 2648 struct list_head list; 2649 }; 2650 2651 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2652 struct pcpu_sw_netstats { 2653 u64_stats_t rx_packets; 2654 u64_stats_t rx_bytes; 2655 u64_stats_t tx_packets; 2656 u64_stats_t tx_bytes; 2657 struct u64_stats_sync syncp; 2658 } __aligned(4 * sizeof(u64)); 2659 2660 struct pcpu_lstats { 2661 u64_stats_t packets; 2662 u64_stats_t bytes; 2663 struct u64_stats_sync syncp; 2664 } __aligned(2 * sizeof(u64)); 2665 2666 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2667 2668 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2669 { 2670 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2671 2672 u64_stats_update_begin(&tstats->syncp); 2673 u64_stats_add(&tstats->rx_bytes, len); 2674 u64_stats_inc(&tstats->rx_packets); 2675 u64_stats_update_end(&tstats->syncp); 2676 } 2677 2678 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2679 unsigned int packets, 2680 unsigned int len) 2681 { 2682 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2683 2684 u64_stats_update_begin(&tstats->syncp); 2685 u64_stats_add(&tstats->tx_bytes, len); 2686 u64_stats_add(&tstats->tx_packets, packets); 2687 u64_stats_update_end(&tstats->syncp); 2688 } 2689 2690 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2691 { 2692 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2693 2694 u64_stats_update_begin(&lstats->syncp); 2695 u64_stats_add(&lstats->bytes, len); 2696 u64_stats_inc(&lstats->packets); 2697 u64_stats_update_end(&lstats->syncp); 2698 } 2699 2700 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2701 ({ \ 2702 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2703 if (pcpu_stats) { \ 2704 int __cpu; \ 2705 for_each_possible_cpu(__cpu) { \ 2706 typeof(type) *stat; \ 2707 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2708 u64_stats_init(&stat->syncp); \ 2709 } \ 2710 } \ 2711 pcpu_stats; \ 2712 }) 2713 2714 #define netdev_alloc_pcpu_stats(type) \ 2715 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2716 2717 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2718 ({ \ 2719 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2720 if (pcpu_stats) { \ 2721 int __cpu; \ 2722 for_each_possible_cpu(__cpu) { \ 2723 typeof(type) *stat; \ 2724 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2725 u64_stats_init(&stat->syncp); \ 2726 } \ 2727 } \ 2728 pcpu_stats; \ 2729 }) 2730 2731 enum netdev_lag_tx_type { 2732 NETDEV_LAG_TX_TYPE_UNKNOWN, 2733 NETDEV_LAG_TX_TYPE_RANDOM, 2734 NETDEV_LAG_TX_TYPE_BROADCAST, 2735 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2736 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2737 NETDEV_LAG_TX_TYPE_HASH, 2738 }; 2739 2740 enum netdev_lag_hash { 2741 NETDEV_LAG_HASH_NONE, 2742 NETDEV_LAG_HASH_L2, 2743 NETDEV_LAG_HASH_L34, 2744 NETDEV_LAG_HASH_L23, 2745 NETDEV_LAG_HASH_E23, 2746 NETDEV_LAG_HASH_E34, 2747 NETDEV_LAG_HASH_VLAN_SRCMAC, 2748 NETDEV_LAG_HASH_UNKNOWN, 2749 }; 2750 2751 struct netdev_lag_upper_info { 2752 enum netdev_lag_tx_type tx_type; 2753 enum netdev_lag_hash hash_type; 2754 }; 2755 2756 struct netdev_lag_lower_state_info { 2757 u8 link_up : 1, 2758 tx_enabled : 1; 2759 }; 2760 2761 #include <linux/notifier.h> 2762 2763 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2764 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2765 * adding new types. 2766 */ 2767 enum netdev_cmd { 2768 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2769 NETDEV_DOWN, 2770 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2771 detected a hardware crash and restarted 2772 - we can use this eg to kick tcp sessions 2773 once done */ 2774 NETDEV_CHANGE, /* Notify device state change */ 2775 NETDEV_REGISTER, 2776 NETDEV_UNREGISTER, 2777 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2778 NETDEV_CHANGEADDR, /* notify after the address change */ 2779 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2780 NETDEV_GOING_DOWN, 2781 NETDEV_CHANGENAME, 2782 NETDEV_FEAT_CHANGE, 2783 NETDEV_BONDING_FAILOVER, 2784 NETDEV_PRE_UP, 2785 NETDEV_PRE_TYPE_CHANGE, 2786 NETDEV_POST_TYPE_CHANGE, 2787 NETDEV_POST_INIT, 2788 NETDEV_RELEASE, 2789 NETDEV_NOTIFY_PEERS, 2790 NETDEV_JOIN, 2791 NETDEV_CHANGEUPPER, 2792 NETDEV_RESEND_IGMP, 2793 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2794 NETDEV_CHANGEINFODATA, 2795 NETDEV_BONDING_INFO, 2796 NETDEV_PRECHANGEUPPER, 2797 NETDEV_CHANGELOWERSTATE, 2798 NETDEV_UDP_TUNNEL_PUSH_INFO, 2799 NETDEV_UDP_TUNNEL_DROP_INFO, 2800 NETDEV_CHANGE_TX_QUEUE_LEN, 2801 NETDEV_CVLAN_FILTER_PUSH_INFO, 2802 NETDEV_CVLAN_FILTER_DROP_INFO, 2803 NETDEV_SVLAN_FILTER_PUSH_INFO, 2804 NETDEV_SVLAN_FILTER_DROP_INFO, 2805 NETDEV_OFFLOAD_XSTATS_ENABLE, 2806 NETDEV_OFFLOAD_XSTATS_DISABLE, 2807 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 2808 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 2809 }; 2810 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2811 2812 int register_netdevice_notifier(struct notifier_block *nb); 2813 int unregister_netdevice_notifier(struct notifier_block *nb); 2814 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2815 int unregister_netdevice_notifier_net(struct net *net, 2816 struct notifier_block *nb); 2817 int register_netdevice_notifier_dev_net(struct net_device *dev, 2818 struct notifier_block *nb, 2819 struct netdev_net_notifier *nn); 2820 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2821 struct notifier_block *nb, 2822 struct netdev_net_notifier *nn); 2823 2824 struct netdev_notifier_info { 2825 struct net_device *dev; 2826 struct netlink_ext_ack *extack; 2827 }; 2828 2829 struct netdev_notifier_info_ext { 2830 struct netdev_notifier_info info; /* must be first */ 2831 union { 2832 u32 mtu; 2833 } ext; 2834 }; 2835 2836 struct netdev_notifier_change_info { 2837 struct netdev_notifier_info info; /* must be first */ 2838 unsigned int flags_changed; 2839 }; 2840 2841 struct netdev_notifier_changeupper_info { 2842 struct netdev_notifier_info info; /* must be first */ 2843 struct net_device *upper_dev; /* new upper dev */ 2844 bool master; /* is upper dev master */ 2845 bool linking; /* is the notification for link or unlink */ 2846 void *upper_info; /* upper dev info */ 2847 }; 2848 2849 struct netdev_notifier_changelowerstate_info { 2850 struct netdev_notifier_info info; /* must be first */ 2851 void *lower_state_info; /* is lower dev state */ 2852 }; 2853 2854 struct netdev_notifier_pre_changeaddr_info { 2855 struct netdev_notifier_info info; /* must be first */ 2856 const unsigned char *dev_addr; 2857 }; 2858 2859 enum netdev_offload_xstats_type { 2860 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 2861 }; 2862 2863 struct netdev_notifier_offload_xstats_info { 2864 struct netdev_notifier_info info; /* must be first */ 2865 enum netdev_offload_xstats_type type; 2866 2867 union { 2868 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 2869 struct netdev_notifier_offload_xstats_rd *report_delta; 2870 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 2871 struct netdev_notifier_offload_xstats_ru *report_used; 2872 }; 2873 }; 2874 2875 int netdev_offload_xstats_enable(struct net_device *dev, 2876 enum netdev_offload_xstats_type type, 2877 struct netlink_ext_ack *extack); 2878 int netdev_offload_xstats_disable(struct net_device *dev, 2879 enum netdev_offload_xstats_type type); 2880 bool netdev_offload_xstats_enabled(const struct net_device *dev, 2881 enum netdev_offload_xstats_type type); 2882 int netdev_offload_xstats_get(struct net_device *dev, 2883 enum netdev_offload_xstats_type type, 2884 struct rtnl_hw_stats64 *stats, bool *used, 2885 struct netlink_ext_ack *extack); 2886 void 2887 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 2888 const struct rtnl_hw_stats64 *stats); 2889 void 2890 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 2891 void netdev_offload_xstats_push_delta(struct net_device *dev, 2892 enum netdev_offload_xstats_type type, 2893 const struct rtnl_hw_stats64 *stats); 2894 2895 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2896 struct net_device *dev) 2897 { 2898 info->dev = dev; 2899 info->extack = NULL; 2900 } 2901 2902 static inline struct net_device * 2903 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2904 { 2905 return info->dev; 2906 } 2907 2908 static inline struct netlink_ext_ack * 2909 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2910 { 2911 return info->extack; 2912 } 2913 2914 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2915 2916 2917 extern rwlock_t dev_base_lock; /* Device list lock */ 2918 2919 #define for_each_netdev(net, d) \ 2920 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2921 #define for_each_netdev_reverse(net, d) \ 2922 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2923 #define for_each_netdev_rcu(net, d) \ 2924 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2925 #define for_each_netdev_safe(net, d, n) \ 2926 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2927 #define for_each_netdev_continue(net, d) \ 2928 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2929 #define for_each_netdev_continue_reverse(net, d) \ 2930 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2931 dev_list) 2932 #define for_each_netdev_continue_rcu(net, d) \ 2933 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2934 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2935 for_each_netdev_rcu(&init_net, slave) \ 2936 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2937 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2938 2939 static inline struct net_device *next_net_device(struct net_device *dev) 2940 { 2941 struct list_head *lh; 2942 struct net *net; 2943 2944 net = dev_net(dev); 2945 lh = dev->dev_list.next; 2946 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2947 } 2948 2949 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2950 { 2951 struct list_head *lh; 2952 struct net *net; 2953 2954 net = dev_net(dev); 2955 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2956 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2957 } 2958 2959 static inline struct net_device *first_net_device(struct net *net) 2960 { 2961 return list_empty(&net->dev_base_head) ? NULL : 2962 net_device_entry(net->dev_base_head.next); 2963 } 2964 2965 static inline struct net_device *first_net_device_rcu(struct net *net) 2966 { 2967 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2968 2969 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2970 } 2971 2972 int netdev_boot_setup_check(struct net_device *dev); 2973 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2974 const char *hwaddr); 2975 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2976 void dev_add_pack(struct packet_type *pt); 2977 void dev_remove_pack(struct packet_type *pt); 2978 void __dev_remove_pack(struct packet_type *pt); 2979 void dev_add_offload(struct packet_offload *po); 2980 void dev_remove_offload(struct packet_offload *po); 2981 2982 int dev_get_iflink(const struct net_device *dev); 2983 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2984 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 2985 struct net_device_path_stack *stack); 2986 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2987 unsigned short mask); 2988 struct net_device *dev_get_by_name(struct net *net, const char *name); 2989 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2990 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2991 bool netdev_name_in_use(struct net *net, const char *name); 2992 int dev_alloc_name(struct net_device *dev, const char *name); 2993 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2994 void dev_close(struct net_device *dev); 2995 void dev_close_many(struct list_head *head, bool unlink); 2996 void dev_disable_lro(struct net_device *dev); 2997 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2998 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2999 struct net_device *sb_dev); 3000 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 3001 struct net_device *sb_dev); 3002 3003 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3004 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3005 3006 static inline int dev_queue_xmit(struct sk_buff *skb) 3007 { 3008 return __dev_queue_xmit(skb, NULL); 3009 } 3010 3011 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3012 struct net_device *sb_dev) 3013 { 3014 return __dev_queue_xmit(skb, sb_dev); 3015 } 3016 3017 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3018 { 3019 int ret; 3020 3021 ret = __dev_direct_xmit(skb, queue_id); 3022 if (!dev_xmit_complete(ret)) 3023 kfree_skb(skb); 3024 return ret; 3025 } 3026 3027 int register_netdevice(struct net_device *dev); 3028 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3029 void unregister_netdevice_many(struct list_head *head); 3030 static inline void unregister_netdevice(struct net_device *dev) 3031 { 3032 unregister_netdevice_queue(dev, NULL); 3033 } 3034 3035 int netdev_refcnt_read(const struct net_device *dev); 3036 void free_netdev(struct net_device *dev); 3037 void netdev_freemem(struct net_device *dev); 3038 int init_dummy_netdev(struct net_device *dev); 3039 3040 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3041 struct sk_buff *skb, 3042 bool all_slaves); 3043 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3044 struct sock *sk); 3045 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3046 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3047 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3048 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 3049 int dev_restart(struct net_device *dev); 3050 3051 3052 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3053 unsigned short type, 3054 const void *daddr, const void *saddr, 3055 unsigned int len) 3056 { 3057 if (!dev->header_ops || !dev->header_ops->create) 3058 return 0; 3059 3060 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3061 } 3062 3063 static inline int dev_parse_header(const struct sk_buff *skb, 3064 unsigned char *haddr) 3065 { 3066 const struct net_device *dev = skb->dev; 3067 3068 if (!dev->header_ops || !dev->header_ops->parse) 3069 return 0; 3070 return dev->header_ops->parse(skb, haddr); 3071 } 3072 3073 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3074 { 3075 const struct net_device *dev = skb->dev; 3076 3077 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3078 return 0; 3079 return dev->header_ops->parse_protocol(skb); 3080 } 3081 3082 /* ll_header must have at least hard_header_len allocated */ 3083 static inline bool dev_validate_header(const struct net_device *dev, 3084 char *ll_header, int len) 3085 { 3086 if (likely(len >= dev->hard_header_len)) 3087 return true; 3088 if (len < dev->min_header_len) 3089 return false; 3090 3091 if (capable(CAP_SYS_RAWIO)) { 3092 memset(ll_header + len, 0, dev->hard_header_len - len); 3093 return true; 3094 } 3095 3096 if (dev->header_ops && dev->header_ops->validate) 3097 return dev->header_ops->validate(ll_header, len); 3098 3099 return false; 3100 } 3101 3102 static inline bool dev_has_header(const struct net_device *dev) 3103 { 3104 return dev->header_ops && dev->header_ops->create; 3105 } 3106 3107 /* 3108 * Incoming packets are placed on per-CPU queues 3109 */ 3110 struct softnet_data { 3111 struct list_head poll_list; 3112 struct sk_buff_head process_queue; 3113 3114 /* stats */ 3115 unsigned int processed; 3116 unsigned int time_squeeze; 3117 unsigned int received_rps; 3118 #ifdef CONFIG_RPS 3119 struct softnet_data *rps_ipi_list; 3120 #endif 3121 #ifdef CONFIG_NET_FLOW_LIMIT 3122 struct sd_flow_limit __rcu *flow_limit; 3123 #endif 3124 struct Qdisc *output_queue; 3125 struct Qdisc **output_queue_tailp; 3126 struct sk_buff *completion_queue; 3127 #ifdef CONFIG_XFRM_OFFLOAD 3128 struct sk_buff_head xfrm_backlog; 3129 #endif 3130 /* written and read only by owning cpu: */ 3131 struct { 3132 u16 recursion; 3133 u8 more; 3134 #ifdef CONFIG_NET_EGRESS 3135 u8 skip_txqueue; 3136 #endif 3137 } xmit; 3138 #ifdef CONFIG_RPS 3139 /* input_queue_head should be written by cpu owning this struct, 3140 * and only read by other cpus. Worth using a cache line. 3141 */ 3142 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3143 3144 /* Elements below can be accessed between CPUs for RPS/RFS */ 3145 call_single_data_t csd ____cacheline_aligned_in_smp; 3146 struct softnet_data *rps_ipi_next; 3147 unsigned int cpu; 3148 unsigned int input_queue_tail; 3149 #endif 3150 unsigned int dropped; 3151 struct sk_buff_head input_pkt_queue; 3152 struct napi_struct backlog; 3153 3154 /* Another possibly contended cache line */ 3155 spinlock_t defer_lock ____cacheline_aligned_in_smp; 3156 int defer_count; 3157 int defer_ipi_scheduled; 3158 struct sk_buff *defer_list; 3159 call_single_data_t defer_csd; 3160 }; 3161 3162 static inline void input_queue_head_incr(struct softnet_data *sd) 3163 { 3164 #ifdef CONFIG_RPS 3165 sd->input_queue_head++; 3166 #endif 3167 } 3168 3169 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3170 unsigned int *qtail) 3171 { 3172 #ifdef CONFIG_RPS 3173 *qtail = ++sd->input_queue_tail; 3174 #endif 3175 } 3176 3177 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3178 3179 static inline int dev_recursion_level(void) 3180 { 3181 return this_cpu_read(softnet_data.xmit.recursion); 3182 } 3183 3184 #define XMIT_RECURSION_LIMIT 8 3185 static inline bool dev_xmit_recursion(void) 3186 { 3187 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3188 XMIT_RECURSION_LIMIT); 3189 } 3190 3191 static inline void dev_xmit_recursion_inc(void) 3192 { 3193 __this_cpu_inc(softnet_data.xmit.recursion); 3194 } 3195 3196 static inline void dev_xmit_recursion_dec(void) 3197 { 3198 __this_cpu_dec(softnet_data.xmit.recursion); 3199 } 3200 3201 void __netif_schedule(struct Qdisc *q); 3202 void netif_schedule_queue(struct netdev_queue *txq); 3203 3204 static inline void netif_tx_schedule_all(struct net_device *dev) 3205 { 3206 unsigned int i; 3207 3208 for (i = 0; i < dev->num_tx_queues; i++) 3209 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3210 } 3211 3212 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3213 { 3214 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3215 } 3216 3217 /** 3218 * netif_start_queue - allow transmit 3219 * @dev: network device 3220 * 3221 * Allow upper layers to call the device hard_start_xmit routine. 3222 */ 3223 static inline void netif_start_queue(struct net_device *dev) 3224 { 3225 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3226 } 3227 3228 static inline void netif_tx_start_all_queues(struct net_device *dev) 3229 { 3230 unsigned int i; 3231 3232 for (i = 0; i < dev->num_tx_queues; i++) { 3233 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3234 netif_tx_start_queue(txq); 3235 } 3236 } 3237 3238 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3239 3240 /** 3241 * netif_wake_queue - restart transmit 3242 * @dev: network device 3243 * 3244 * Allow upper layers to call the device hard_start_xmit routine. 3245 * Used for flow control when transmit resources are available. 3246 */ 3247 static inline void netif_wake_queue(struct net_device *dev) 3248 { 3249 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3250 } 3251 3252 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3253 { 3254 unsigned int i; 3255 3256 for (i = 0; i < dev->num_tx_queues; i++) { 3257 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3258 netif_tx_wake_queue(txq); 3259 } 3260 } 3261 3262 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3263 { 3264 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3265 } 3266 3267 /** 3268 * netif_stop_queue - stop transmitted packets 3269 * @dev: network device 3270 * 3271 * Stop upper layers calling the device hard_start_xmit routine. 3272 * Used for flow control when transmit resources are unavailable. 3273 */ 3274 static inline void netif_stop_queue(struct net_device *dev) 3275 { 3276 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3277 } 3278 3279 void netif_tx_stop_all_queues(struct net_device *dev); 3280 3281 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3282 { 3283 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3284 } 3285 3286 /** 3287 * netif_queue_stopped - test if transmit queue is flowblocked 3288 * @dev: network device 3289 * 3290 * Test if transmit queue on device is currently unable to send. 3291 */ 3292 static inline bool netif_queue_stopped(const struct net_device *dev) 3293 { 3294 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3295 } 3296 3297 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3298 { 3299 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3300 } 3301 3302 static inline bool 3303 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3304 { 3305 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3306 } 3307 3308 static inline bool 3309 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3310 { 3311 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3312 } 3313 3314 /** 3315 * netdev_queue_set_dql_min_limit - set dql minimum limit 3316 * @dev_queue: pointer to transmit queue 3317 * @min_limit: dql minimum limit 3318 * 3319 * Forces xmit_more() to return true until the minimum threshold 3320 * defined by @min_limit is reached (or until the tx queue is 3321 * empty). Warning: to be use with care, misuse will impact the 3322 * latency. 3323 */ 3324 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3325 unsigned int min_limit) 3326 { 3327 #ifdef CONFIG_BQL 3328 dev_queue->dql.min_limit = min_limit; 3329 #endif 3330 } 3331 3332 /** 3333 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3334 * @dev_queue: pointer to transmit queue 3335 * 3336 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3337 * to give appropriate hint to the CPU. 3338 */ 3339 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3340 { 3341 #ifdef CONFIG_BQL 3342 prefetchw(&dev_queue->dql.num_queued); 3343 #endif 3344 } 3345 3346 /** 3347 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3348 * @dev_queue: pointer to transmit queue 3349 * 3350 * BQL enabled drivers might use this helper in their TX completion path, 3351 * to give appropriate hint to the CPU. 3352 */ 3353 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3354 { 3355 #ifdef CONFIG_BQL 3356 prefetchw(&dev_queue->dql.limit); 3357 #endif 3358 } 3359 3360 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3361 unsigned int bytes) 3362 { 3363 #ifdef CONFIG_BQL 3364 dql_queued(&dev_queue->dql, bytes); 3365 3366 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3367 return; 3368 3369 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3370 3371 /* 3372 * The XOFF flag must be set before checking the dql_avail below, 3373 * because in netdev_tx_completed_queue we update the dql_completed 3374 * before checking the XOFF flag. 3375 */ 3376 smp_mb(); 3377 3378 /* check again in case another CPU has just made room avail */ 3379 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3380 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3381 #endif 3382 } 3383 3384 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3385 * that they should not test BQL status themselves. 3386 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3387 * skb of a batch. 3388 * Returns true if the doorbell must be used to kick the NIC. 3389 */ 3390 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3391 unsigned int bytes, 3392 bool xmit_more) 3393 { 3394 if (xmit_more) { 3395 #ifdef CONFIG_BQL 3396 dql_queued(&dev_queue->dql, bytes); 3397 #endif 3398 return netif_tx_queue_stopped(dev_queue); 3399 } 3400 netdev_tx_sent_queue(dev_queue, bytes); 3401 return true; 3402 } 3403 3404 /** 3405 * netdev_sent_queue - report the number of bytes queued to hardware 3406 * @dev: network device 3407 * @bytes: number of bytes queued to the hardware device queue 3408 * 3409 * Report the number of bytes queued for sending/completion to the network 3410 * device hardware queue. @bytes should be a good approximation and should 3411 * exactly match netdev_completed_queue() @bytes 3412 */ 3413 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3414 { 3415 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3416 } 3417 3418 static inline bool __netdev_sent_queue(struct net_device *dev, 3419 unsigned int bytes, 3420 bool xmit_more) 3421 { 3422 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3423 xmit_more); 3424 } 3425 3426 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3427 unsigned int pkts, unsigned int bytes) 3428 { 3429 #ifdef CONFIG_BQL 3430 if (unlikely(!bytes)) 3431 return; 3432 3433 dql_completed(&dev_queue->dql, bytes); 3434 3435 /* 3436 * Without the memory barrier there is a small possiblity that 3437 * netdev_tx_sent_queue will miss the update and cause the queue to 3438 * be stopped forever 3439 */ 3440 smp_mb(); 3441 3442 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3443 return; 3444 3445 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3446 netif_schedule_queue(dev_queue); 3447 #endif 3448 } 3449 3450 /** 3451 * netdev_completed_queue - report bytes and packets completed by device 3452 * @dev: network device 3453 * @pkts: actual number of packets sent over the medium 3454 * @bytes: actual number of bytes sent over the medium 3455 * 3456 * Report the number of bytes and packets transmitted by the network device 3457 * hardware queue over the physical medium, @bytes must exactly match the 3458 * @bytes amount passed to netdev_sent_queue() 3459 */ 3460 static inline void netdev_completed_queue(struct net_device *dev, 3461 unsigned int pkts, unsigned int bytes) 3462 { 3463 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3464 } 3465 3466 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3467 { 3468 #ifdef CONFIG_BQL 3469 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3470 dql_reset(&q->dql); 3471 #endif 3472 } 3473 3474 /** 3475 * netdev_reset_queue - reset the packets and bytes count of a network device 3476 * @dev_queue: network device 3477 * 3478 * Reset the bytes and packet count of a network device and clear the 3479 * software flow control OFF bit for this network device 3480 */ 3481 static inline void netdev_reset_queue(struct net_device *dev_queue) 3482 { 3483 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3484 } 3485 3486 /** 3487 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3488 * @dev: network device 3489 * @queue_index: given tx queue index 3490 * 3491 * Returns 0 if given tx queue index >= number of device tx queues, 3492 * otherwise returns the originally passed tx queue index. 3493 */ 3494 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3495 { 3496 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3497 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3498 dev->name, queue_index, 3499 dev->real_num_tx_queues); 3500 return 0; 3501 } 3502 3503 return queue_index; 3504 } 3505 3506 /** 3507 * netif_running - test if up 3508 * @dev: network device 3509 * 3510 * Test if the device has been brought up. 3511 */ 3512 static inline bool netif_running(const struct net_device *dev) 3513 { 3514 return test_bit(__LINK_STATE_START, &dev->state); 3515 } 3516 3517 /* 3518 * Routines to manage the subqueues on a device. We only need start, 3519 * stop, and a check if it's stopped. All other device management is 3520 * done at the overall netdevice level. 3521 * Also test the device if we're multiqueue. 3522 */ 3523 3524 /** 3525 * netif_start_subqueue - allow sending packets on subqueue 3526 * @dev: network device 3527 * @queue_index: sub queue index 3528 * 3529 * Start individual transmit queue of a device with multiple transmit queues. 3530 */ 3531 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3532 { 3533 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3534 3535 netif_tx_start_queue(txq); 3536 } 3537 3538 /** 3539 * netif_stop_subqueue - stop sending packets on subqueue 3540 * @dev: network device 3541 * @queue_index: sub queue index 3542 * 3543 * Stop individual transmit queue of a device with multiple transmit queues. 3544 */ 3545 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3546 { 3547 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3548 netif_tx_stop_queue(txq); 3549 } 3550 3551 /** 3552 * __netif_subqueue_stopped - test status of subqueue 3553 * @dev: network device 3554 * @queue_index: sub queue index 3555 * 3556 * Check individual transmit queue of a device with multiple transmit queues. 3557 */ 3558 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3559 u16 queue_index) 3560 { 3561 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3562 3563 return netif_tx_queue_stopped(txq); 3564 } 3565 3566 /** 3567 * netif_subqueue_stopped - test status of subqueue 3568 * @dev: network device 3569 * @skb: sub queue buffer pointer 3570 * 3571 * Check individual transmit queue of a device with multiple transmit queues. 3572 */ 3573 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3574 struct sk_buff *skb) 3575 { 3576 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3577 } 3578 3579 /** 3580 * netif_wake_subqueue - allow sending packets on subqueue 3581 * @dev: network device 3582 * @queue_index: sub queue index 3583 * 3584 * Resume individual transmit queue of a device with multiple transmit queues. 3585 */ 3586 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3587 { 3588 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3589 3590 netif_tx_wake_queue(txq); 3591 } 3592 3593 #ifdef CONFIG_XPS 3594 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3595 u16 index); 3596 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3597 u16 index, enum xps_map_type type); 3598 3599 /** 3600 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3601 * @j: CPU/Rx queue index 3602 * @mask: bitmask of all cpus/rx queues 3603 * @nr_bits: number of bits in the bitmask 3604 * 3605 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3606 */ 3607 static inline bool netif_attr_test_mask(unsigned long j, 3608 const unsigned long *mask, 3609 unsigned int nr_bits) 3610 { 3611 cpu_max_bits_warn(j, nr_bits); 3612 return test_bit(j, mask); 3613 } 3614 3615 /** 3616 * netif_attr_test_online - Test for online CPU/Rx queue 3617 * @j: CPU/Rx queue index 3618 * @online_mask: bitmask for CPUs/Rx queues that are online 3619 * @nr_bits: number of bits in the bitmask 3620 * 3621 * Returns true if a CPU/Rx queue is online. 3622 */ 3623 static inline bool netif_attr_test_online(unsigned long j, 3624 const unsigned long *online_mask, 3625 unsigned int nr_bits) 3626 { 3627 cpu_max_bits_warn(j, nr_bits); 3628 3629 if (online_mask) 3630 return test_bit(j, online_mask); 3631 3632 return (j < nr_bits); 3633 } 3634 3635 /** 3636 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3637 * @n: CPU/Rx queue index 3638 * @srcp: the cpumask/Rx queue mask pointer 3639 * @nr_bits: number of bits in the bitmask 3640 * 3641 * Returns >= nr_bits if no further CPUs/Rx queues set. 3642 */ 3643 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3644 unsigned int nr_bits) 3645 { 3646 /* -1 is a legal arg here. */ 3647 if (n != -1) 3648 cpu_max_bits_warn(n, nr_bits); 3649 3650 if (srcp) 3651 return find_next_bit(srcp, nr_bits, n + 1); 3652 3653 return n + 1; 3654 } 3655 3656 /** 3657 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3658 * @n: CPU/Rx queue index 3659 * @src1p: the first CPUs/Rx queues mask pointer 3660 * @src2p: the second CPUs/Rx queues mask pointer 3661 * @nr_bits: number of bits in the bitmask 3662 * 3663 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3664 */ 3665 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3666 const unsigned long *src2p, 3667 unsigned int nr_bits) 3668 { 3669 /* -1 is a legal arg here. */ 3670 if (n != -1) 3671 cpu_max_bits_warn(n, nr_bits); 3672 3673 if (src1p && src2p) 3674 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3675 else if (src1p) 3676 return find_next_bit(src1p, nr_bits, n + 1); 3677 else if (src2p) 3678 return find_next_bit(src2p, nr_bits, n + 1); 3679 3680 return n + 1; 3681 } 3682 #else 3683 static inline int netif_set_xps_queue(struct net_device *dev, 3684 const struct cpumask *mask, 3685 u16 index) 3686 { 3687 return 0; 3688 } 3689 3690 static inline int __netif_set_xps_queue(struct net_device *dev, 3691 const unsigned long *mask, 3692 u16 index, enum xps_map_type type) 3693 { 3694 return 0; 3695 } 3696 #endif 3697 3698 /** 3699 * netif_is_multiqueue - test if device has multiple transmit queues 3700 * @dev: network device 3701 * 3702 * Check if device has multiple transmit queues 3703 */ 3704 static inline bool netif_is_multiqueue(const struct net_device *dev) 3705 { 3706 return dev->num_tx_queues > 1; 3707 } 3708 3709 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3710 3711 #ifdef CONFIG_SYSFS 3712 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3713 #else 3714 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3715 unsigned int rxqs) 3716 { 3717 dev->real_num_rx_queues = rxqs; 3718 return 0; 3719 } 3720 #endif 3721 int netif_set_real_num_queues(struct net_device *dev, 3722 unsigned int txq, unsigned int rxq); 3723 3724 static inline struct netdev_rx_queue * 3725 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3726 { 3727 return dev->_rx + rxq; 3728 } 3729 3730 #ifdef CONFIG_SYSFS 3731 static inline unsigned int get_netdev_rx_queue_index( 3732 struct netdev_rx_queue *queue) 3733 { 3734 struct net_device *dev = queue->dev; 3735 int index = queue - dev->_rx; 3736 3737 BUG_ON(index >= dev->num_rx_queues); 3738 return index; 3739 } 3740 #endif 3741 3742 int netif_get_num_default_rss_queues(void); 3743 3744 enum skb_free_reason { 3745 SKB_REASON_CONSUMED, 3746 SKB_REASON_DROPPED, 3747 }; 3748 3749 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3750 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3751 3752 /* 3753 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3754 * interrupt context or with hardware interrupts being disabled. 3755 * (in_hardirq() || irqs_disabled()) 3756 * 3757 * We provide four helpers that can be used in following contexts : 3758 * 3759 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3760 * replacing kfree_skb(skb) 3761 * 3762 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3763 * Typically used in place of consume_skb(skb) in TX completion path 3764 * 3765 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3766 * replacing kfree_skb(skb) 3767 * 3768 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3769 * and consumed a packet. Used in place of consume_skb(skb) 3770 */ 3771 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3772 { 3773 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3774 } 3775 3776 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3777 { 3778 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3779 } 3780 3781 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3782 { 3783 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3784 } 3785 3786 static inline void dev_consume_skb_any(struct sk_buff *skb) 3787 { 3788 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3789 } 3790 3791 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3792 struct bpf_prog *xdp_prog); 3793 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3794 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3795 int netif_rx(struct sk_buff *skb); 3796 int __netif_rx(struct sk_buff *skb); 3797 3798 int netif_receive_skb(struct sk_buff *skb); 3799 int netif_receive_skb_core(struct sk_buff *skb); 3800 void netif_receive_skb_list_internal(struct list_head *head); 3801 void netif_receive_skb_list(struct list_head *head); 3802 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3803 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3804 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3805 gro_result_t napi_gro_frags(struct napi_struct *napi); 3806 struct packet_offload *gro_find_receive_by_type(__be16 type); 3807 struct packet_offload *gro_find_complete_by_type(__be16 type); 3808 3809 static inline void napi_free_frags(struct napi_struct *napi) 3810 { 3811 kfree_skb(napi->skb); 3812 napi->skb = NULL; 3813 } 3814 3815 bool netdev_is_rx_handler_busy(struct net_device *dev); 3816 int netdev_rx_handler_register(struct net_device *dev, 3817 rx_handler_func_t *rx_handler, 3818 void *rx_handler_data); 3819 void netdev_rx_handler_unregister(struct net_device *dev); 3820 3821 bool dev_valid_name(const char *name); 3822 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 3823 { 3824 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 3825 } 3826 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 3827 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 3828 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3829 void __user *data, bool *need_copyout); 3830 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 3831 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 3832 unsigned int dev_get_flags(const struct net_device *); 3833 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3834 struct netlink_ext_ack *extack); 3835 int dev_change_flags(struct net_device *dev, unsigned int flags, 3836 struct netlink_ext_ack *extack); 3837 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3838 unsigned int gchanges); 3839 int dev_set_alias(struct net_device *, const char *, size_t); 3840 int dev_get_alias(const struct net_device *, char *, size_t); 3841 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 3842 const char *pat, int new_ifindex); 3843 static inline 3844 int dev_change_net_namespace(struct net_device *dev, struct net *net, 3845 const char *pat) 3846 { 3847 return __dev_change_net_namespace(dev, net, pat, 0); 3848 } 3849 int __dev_set_mtu(struct net_device *, int); 3850 int dev_set_mtu(struct net_device *, int); 3851 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3852 struct netlink_ext_ack *extack); 3853 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3854 struct netlink_ext_ack *extack); 3855 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 3856 struct netlink_ext_ack *extack); 3857 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 3858 int dev_get_port_parent_id(struct net_device *dev, 3859 struct netdev_phys_item_id *ppid, bool recurse); 3860 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3861 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3862 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3863 struct netdev_queue *txq, int *ret); 3864 3865 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3866 u8 dev_xdp_prog_count(struct net_device *dev); 3867 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3868 3869 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3870 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3871 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 3872 bool is_skb_forwardable(const struct net_device *dev, 3873 const struct sk_buff *skb); 3874 3875 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 3876 const struct sk_buff *skb, 3877 const bool check_mtu) 3878 { 3879 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 3880 unsigned int len; 3881 3882 if (!(dev->flags & IFF_UP)) 3883 return false; 3884 3885 if (!check_mtu) 3886 return true; 3887 3888 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 3889 if (skb->len <= len) 3890 return true; 3891 3892 /* if TSO is enabled, we don't care about the length as the packet 3893 * could be forwarded without being segmented before 3894 */ 3895 if (skb_is_gso(skb)) 3896 return true; 3897 3898 return false; 3899 } 3900 3901 struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev); 3902 3903 static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev) 3904 { 3905 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */ 3906 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats); 3907 3908 if (likely(p)) 3909 return p; 3910 3911 return netdev_core_stats_alloc(dev); 3912 } 3913 3914 #define DEV_CORE_STATS_INC(FIELD) \ 3915 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 3916 { \ 3917 struct net_device_core_stats __percpu *p; \ 3918 \ 3919 p = dev_core_stats(dev); \ 3920 if (p) \ 3921 this_cpu_inc(p->FIELD); \ 3922 } 3923 DEV_CORE_STATS_INC(rx_dropped) 3924 DEV_CORE_STATS_INC(tx_dropped) 3925 DEV_CORE_STATS_INC(rx_nohandler) 3926 DEV_CORE_STATS_INC(rx_otherhost_dropped) 3927 3928 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3929 struct sk_buff *skb, 3930 const bool check_mtu) 3931 { 3932 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3933 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 3934 dev_core_stats_rx_dropped_inc(dev); 3935 kfree_skb(skb); 3936 return NET_RX_DROP; 3937 } 3938 3939 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 3940 skb->priority = 0; 3941 return 0; 3942 } 3943 3944 bool dev_nit_active(struct net_device *dev); 3945 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3946 3947 static inline void __dev_put(struct net_device *dev) 3948 { 3949 if (dev) { 3950 #ifdef CONFIG_PCPU_DEV_REFCNT 3951 this_cpu_dec(*dev->pcpu_refcnt); 3952 #else 3953 refcount_dec(&dev->dev_refcnt); 3954 #endif 3955 } 3956 } 3957 3958 static inline void __dev_hold(struct net_device *dev) 3959 { 3960 if (dev) { 3961 #ifdef CONFIG_PCPU_DEV_REFCNT 3962 this_cpu_inc(*dev->pcpu_refcnt); 3963 #else 3964 refcount_inc(&dev->dev_refcnt); 3965 #endif 3966 } 3967 } 3968 3969 static inline void __netdev_tracker_alloc(struct net_device *dev, 3970 netdevice_tracker *tracker, 3971 gfp_t gfp) 3972 { 3973 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 3974 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 3975 #endif 3976 } 3977 3978 /* netdev_tracker_alloc() can upgrade a prior untracked reference 3979 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 3980 */ 3981 static inline void netdev_tracker_alloc(struct net_device *dev, 3982 netdevice_tracker *tracker, gfp_t gfp) 3983 { 3984 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 3985 refcount_dec(&dev->refcnt_tracker.no_tracker); 3986 __netdev_tracker_alloc(dev, tracker, gfp); 3987 #endif 3988 } 3989 3990 static inline void netdev_tracker_free(struct net_device *dev, 3991 netdevice_tracker *tracker) 3992 { 3993 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 3994 ref_tracker_free(&dev->refcnt_tracker, tracker); 3995 #endif 3996 } 3997 3998 static inline void netdev_hold(struct net_device *dev, 3999 netdevice_tracker *tracker, gfp_t gfp) 4000 { 4001 if (dev) { 4002 __dev_hold(dev); 4003 __netdev_tracker_alloc(dev, tracker, gfp); 4004 } 4005 } 4006 4007 static inline void netdev_put(struct net_device *dev, 4008 netdevice_tracker *tracker) 4009 { 4010 if (dev) { 4011 netdev_tracker_free(dev, tracker); 4012 __dev_put(dev); 4013 } 4014 } 4015 4016 /** 4017 * dev_hold - get reference to device 4018 * @dev: network device 4019 * 4020 * Hold reference to device to keep it from being freed. 4021 * Try using netdev_hold() instead. 4022 */ 4023 static inline void dev_hold(struct net_device *dev) 4024 { 4025 netdev_hold(dev, NULL, GFP_ATOMIC); 4026 } 4027 4028 /** 4029 * dev_put - release reference to device 4030 * @dev: network device 4031 * 4032 * Release reference to device to allow it to be freed. 4033 * Try using netdev_put() instead. 4034 */ 4035 static inline void dev_put(struct net_device *dev) 4036 { 4037 netdev_put(dev, NULL); 4038 } 4039 4040 static inline void netdev_ref_replace(struct net_device *odev, 4041 struct net_device *ndev, 4042 netdevice_tracker *tracker, 4043 gfp_t gfp) 4044 { 4045 if (odev) 4046 netdev_tracker_free(odev, tracker); 4047 4048 __dev_hold(ndev); 4049 __dev_put(odev); 4050 4051 if (ndev) 4052 __netdev_tracker_alloc(ndev, tracker, gfp); 4053 } 4054 4055 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4056 * and _off may be called from IRQ context, but it is caller 4057 * who is responsible for serialization of these calls. 4058 * 4059 * The name carrier is inappropriate, these functions should really be 4060 * called netif_lowerlayer_*() because they represent the state of any 4061 * kind of lower layer not just hardware media. 4062 */ 4063 void linkwatch_fire_event(struct net_device *dev); 4064 4065 /** 4066 * netif_carrier_ok - test if carrier present 4067 * @dev: network device 4068 * 4069 * Check if carrier is present on device 4070 */ 4071 static inline bool netif_carrier_ok(const struct net_device *dev) 4072 { 4073 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4074 } 4075 4076 unsigned long dev_trans_start(struct net_device *dev); 4077 4078 void __netdev_watchdog_up(struct net_device *dev); 4079 4080 void netif_carrier_on(struct net_device *dev); 4081 void netif_carrier_off(struct net_device *dev); 4082 void netif_carrier_event(struct net_device *dev); 4083 4084 /** 4085 * netif_dormant_on - mark device as dormant. 4086 * @dev: network device 4087 * 4088 * Mark device as dormant (as per RFC2863). 4089 * 4090 * The dormant state indicates that the relevant interface is not 4091 * actually in a condition to pass packets (i.e., it is not 'up') but is 4092 * in a "pending" state, waiting for some external event. For "on- 4093 * demand" interfaces, this new state identifies the situation where the 4094 * interface is waiting for events to place it in the up state. 4095 */ 4096 static inline void netif_dormant_on(struct net_device *dev) 4097 { 4098 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4099 linkwatch_fire_event(dev); 4100 } 4101 4102 /** 4103 * netif_dormant_off - set device as not dormant. 4104 * @dev: network device 4105 * 4106 * Device is not in dormant state. 4107 */ 4108 static inline void netif_dormant_off(struct net_device *dev) 4109 { 4110 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4111 linkwatch_fire_event(dev); 4112 } 4113 4114 /** 4115 * netif_dormant - test if device is dormant 4116 * @dev: network device 4117 * 4118 * Check if device is dormant. 4119 */ 4120 static inline bool netif_dormant(const struct net_device *dev) 4121 { 4122 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4123 } 4124 4125 4126 /** 4127 * netif_testing_on - mark device as under test. 4128 * @dev: network device 4129 * 4130 * Mark device as under test (as per RFC2863). 4131 * 4132 * The testing state indicates that some test(s) must be performed on 4133 * the interface. After completion, of the test, the interface state 4134 * will change to up, dormant, or down, as appropriate. 4135 */ 4136 static inline void netif_testing_on(struct net_device *dev) 4137 { 4138 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4139 linkwatch_fire_event(dev); 4140 } 4141 4142 /** 4143 * netif_testing_off - set device as not under test. 4144 * @dev: network device 4145 * 4146 * Device is not in testing state. 4147 */ 4148 static inline void netif_testing_off(struct net_device *dev) 4149 { 4150 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4151 linkwatch_fire_event(dev); 4152 } 4153 4154 /** 4155 * netif_testing - test if device is under test 4156 * @dev: network device 4157 * 4158 * Check if device is under test 4159 */ 4160 static inline bool netif_testing(const struct net_device *dev) 4161 { 4162 return test_bit(__LINK_STATE_TESTING, &dev->state); 4163 } 4164 4165 4166 /** 4167 * netif_oper_up - test if device is operational 4168 * @dev: network device 4169 * 4170 * Check if carrier is operational 4171 */ 4172 static inline bool netif_oper_up(const struct net_device *dev) 4173 { 4174 return (dev->operstate == IF_OPER_UP || 4175 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4176 } 4177 4178 /** 4179 * netif_device_present - is device available or removed 4180 * @dev: network device 4181 * 4182 * Check if device has not been removed from system. 4183 */ 4184 static inline bool netif_device_present(const struct net_device *dev) 4185 { 4186 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4187 } 4188 4189 void netif_device_detach(struct net_device *dev); 4190 4191 void netif_device_attach(struct net_device *dev); 4192 4193 /* 4194 * Network interface message level settings 4195 */ 4196 4197 enum { 4198 NETIF_MSG_DRV_BIT, 4199 NETIF_MSG_PROBE_BIT, 4200 NETIF_MSG_LINK_BIT, 4201 NETIF_MSG_TIMER_BIT, 4202 NETIF_MSG_IFDOWN_BIT, 4203 NETIF_MSG_IFUP_BIT, 4204 NETIF_MSG_RX_ERR_BIT, 4205 NETIF_MSG_TX_ERR_BIT, 4206 NETIF_MSG_TX_QUEUED_BIT, 4207 NETIF_MSG_INTR_BIT, 4208 NETIF_MSG_TX_DONE_BIT, 4209 NETIF_MSG_RX_STATUS_BIT, 4210 NETIF_MSG_PKTDATA_BIT, 4211 NETIF_MSG_HW_BIT, 4212 NETIF_MSG_WOL_BIT, 4213 4214 /* When you add a new bit above, update netif_msg_class_names array 4215 * in net/ethtool/common.c 4216 */ 4217 NETIF_MSG_CLASS_COUNT, 4218 }; 4219 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4220 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4221 4222 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4223 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4224 4225 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4226 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4227 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4228 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4229 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4230 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4231 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4232 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4233 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4234 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4235 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4236 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4237 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4238 #define NETIF_MSG_HW __NETIF_MSG(HW) 4239 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4240 4241 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4242 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4243 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4244 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4245 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4246 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4247 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4248 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4249 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4250 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4251 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4252 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4253 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4254 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4255 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4256 4257 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4258 { 4259 /* use default */ 4260 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4261 return default_msg_enable_bits; 4262 if (debug_value == 0) /* no output */ 4263 return 0; 4264 /* set low N bits */ 4265 return (1U << debug_value) - 1; 4266 } 4267 4268 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4269 { 4270 spin_lock(&txq->_xmit_lock); 4271 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4272 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4273 } 4274 4275 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4276 { 4277 __acquire(&txq->_xmit_lock); 4278 return true; 4279 } 4280 4281 static inline void __netif_tx_release(struct netdev_queue *txq) 4282 { 4283 __release(&txq->_xmit_lock); 4284 } 4285 4286 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4287 { 4288 spin_lock_bh(&txq->_xmit_lock); 4289 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4290 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4291 } 4292 4293 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4294 { 4295 bool ok = spin_trylock(&txq->_xmit_lock); 4296 4297 if (likely(ok)) { 4298 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4299 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4300 } 4301 return ok; 4302 } 4303 4304 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4305 { 4306 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4307 WRITE_ONCE(txq->xmit_lock_owner, -1); 4308 spin_unlock(&txq->_xmit_lock); 4309 } 4310 4311 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4312 { 4313 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4314 WRITE_ONCE(txq->xmit_lock_owner, -1); 4315 spin_unlock_bh(&txq->_xmit_lock); 4316 } 4317 4318 /* 4319 * txq->trans_start can be read locklessly from dev_watchdog() 4320 */ 4321 static inline void txq_trans_update(struct netdev_queue *txq) 4322 { 4323 if (txq->xmit_lock_owner != -1) 4324 WRITE_ONCE(txq->trans_start, jiffies); 4325 } 4326 4327 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4328 { 4329 unsigned long now = jiffies; 4330 4331 if (READ_ONCE(txq->trans_start) != now) 4332 WRITE_ONCE(txq->trans_start, now); 4333 } 4334 4335 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4336 static inline void netif_trans_update(struct net_device *dev) 4337 { 4338 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4339 4340 txq_trans_cond_update(txq); 4341 } 4342 4343 /** 4344 * netif_tx_lock - grab network device transmit lock 4345 * @dev: network device 4346 * 4347 * Get network device transmit lock 4348 */ 4349 void netif_tx_lock(struct net_device *dev); 4350 4351 static inline void netif_tx_lock_bh(struct net_device *dev) 4352 { 4353 local_bh_disable(); 4354 netif_tx_lock(dev); 4355 } 4356 4357 void netif_tx_unlock(struct net_device *dev); 4358 4359 static inline void netif_tx_unlock_bh(struct net_device *dev) 4360 { 4361 netif_tx_unlock(dev); 4362 local_bh_enable(); 4363 } 4364 4365 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4366 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4367 __netif_tx_lock(txq, cpu); \ 4368 } else { \ 4369 __netif_tx_acquire(txq); \ 4370 } \ 4371 } 4372 4373 #define HARD_TX_TRYLOCK(dev, txq) \ 4374 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4375 __netif_tx_trylock(txq) : \ 4376 __netif_tx_acquire(txq)) 4377 4378 #define HARD_TX_UNLOCK(dev, txq) { \ 4379 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4380 __netif_tx_unlock(txq); \ 4381 } else { \ 4382 __netif_tx_release(txq); \ 4383 } \ 4384 } 4385 4386 static inline void netif_tx_disable(struct net_device *dev) 4387 { 4388 unsigned int i; 4389 int cpu; 4390 4391 local_bh_disable(); 4392 cpu = smp_processor_id(); 4393 spin_lock(&dev->tx_global_lock); 4394 for (i = 0; i < dev->num_tx_queues; i++) { 4395 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4396 4397 __netif_tx_lock(txq, cpu); 4398 netif_tx_stop_queue(txq); 4399 __netif_tx_unlock(txq); 4400 } 4401 spin_unlock(&dev->tx_global_lock); 4402 local_bh_enable(); 4403 } 4404 4405 static inline void netif_addr_lock(struct net_device *dev) 4406 { 4407 unsigned char nest_level = 0; 4408 4409 #ifdef CONFIG_LOCKDEP 4410 nest_level = dev->nested_level; 4411 #endif 4412 spin_lock_nested(&dev->addr_list_lock, nest_level); 4413 } 4414 4415 static inline void netif_addr_lock_bh(struct net_device *dev) 4416 { 4417 unsigned char nest_level = 0; 4418 4419 #ifdef CONFIG_LOCKDEP 4420 nest_level = dev->nested_level; 4421 #endif 4422 local_bh_disable(); 4423 spin_lock_nested(&dev->addr_list_lock, nest_level); 4424 } 4425 4426 static inline void netif_addr_unlock(struct net_device *dev) 4427 { 4428 spin_unlock(&dev->addr_list_lock); 4429 } 4430 4431 static inline void netif_addr_unlock_bh(struct net_device *dev) 4432 { 4433 spin_unlock_bh(&dev->addr_list_lock); 4434 } 4435 4436 /* 4437 * dev_addrs walker. Should be used only for read access. Call with 4438 * rcu_read_lock held. 4439 */ 4440 #define for_each_dev_addr(dev, ha) \ 4441 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4442 4443 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4444 4445 void ether_setup(struct net_device *dev); 4446 4447 /* Support for loadable net-drivers */ 4448 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4449 unsigned char name_assign_type, 4450 void (*setup)(struct net_device *), 4451 unsigned int txqs, unsigned int rxqs); 4452 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4453 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4454 4455 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4456 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4457 count) 4458 4459 int register_netdev(struct net_device *dev); 4460 void unregister_netdev(struct net_device *dev); 4461 4462 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4463 4464 /* General hardware address lists handling functions */ 4465 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4466 struct netdev_hw_addr_list *from_list, int addr_len); 4467 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4468 struct netdev_hw_addr_list *from_list, int addr_len); 4469 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4470 struct net_device *dev, 4471 int (*sync)(struct net_device *, const unsigned char *), 4472 int (*unsync)(struct net_device *, 4473 const unsigned char *)); 4474 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4475 struct net_device *dev, 4476 int (*sync)(struct net_device *, 4477 const unsigned char *, int), 4478 int (*unsync)(struct net_device *, 4479 const unsigned char *, int)); 4480 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4481 struct net_device *dev, 4482 int (*unsync)(struct net_device *, 4483 const unsigned char *, int)); 4484 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4485 struct net_device *dev, 4486 int (*unsync)(struct net_device *, 4487 const unsigned char *)); 4488 void __hw_addr_init(struct netdev_hw_addr_list *list); 4489 4490 /* Functions used for device addresses handling */ 4491 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4492 const void *addr, size_t len); 4493 4494 static inline void 4495 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4496 { 4497 dev_addr_mod(dev, 0, addr, len); 4498 } 4499 4500 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4501 { 4502 __dev_addr_set(dev, addr, dev->addr_len); 4503 } 4504 4505 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4506 unsigned char addr_type); 4507 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4508 unsigned char addr_type); 4509 4510 /* Functions used for unicast addresses handling */ 4511 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4512 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4513 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4514 int dev_uc_sync(struct net_device *to, struct net_device *from); 4515 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4516 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4517 void dev_uc_flush(struct net_device *dev); 4518 void dev_uc_init(struct net_device *dev); 4519 4520 /** 4521 * __dev_uc_sync - Synchonize device's unicast list 4522 * @dev: device to sync 4523 * @sync: function to call if address should be added 4524 * @unsync: function to call if address should be removed 4525 * 4526 * Add newly added addresses to the interface, and release 4527 * addresses that have been deleted. 4528 */ 4529 static inline int __dev_uc_sync(struct net_device *dev, 4530 int (*sync)(struct net_device *, 4531 const unsigned char *), 4532 int (*unsync)(struct net_device *, 4533 const unsigned char *)) 4534 { 4535 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4536 } 4537 4538 /** 4539 * __dev_uc_unsync - Remove synchronized addresses from device 4540 * @dev: device to sync 4541 * @unsync: function to call if address should be removed 4542 * 4543 * Remove all addresses that were added to the device by dev_uc_sync(). 4544 */ 4545 static inline void __dev_uc_unsync(struct net_device *dev, 4546 int (*unsync)(struct net_device *, 4547 const unsigned char *)) 4548 { 4549 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4550 } 4551 4552 /* Functions used for multicast addresses handling */ 4553 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4554 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4555 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4556 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4557 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4558 int dev_mc_sync(struct net_device *to, struct net_device *from); 4559 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4560 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4561 void dev_mc_flush(struct net_device *dev); 4562 void dev_mc_init(struct net_device *dev); 4563 4564 /** 4565 * __dev_mc_sync - Synchonize device's multicast list 4566 * @dev: device to sync 4567 * @sync: function to call if address should be added 4568 * @unsync: function to call if address should be removed 4569 * 4570 * Add newly added addresses to the interface, and release 4571 * addresses that have been deleted. 4572 */ 4573 static inline int __dev_mc_sync(struct net_device *dev, 4574 int (*sync)(struct net_device *, 4575 const unsigned char *), 4576 int (*unsync)(struct net_device *, 4577 const unsigned char *)) 4578 { 4579 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4580 } 4581 4582 /** 4583 * __dev_mc_unsync - Remove synchronized addresses from device 4584 * @dev: device to sync 4585 * @unsync: function to call if address should be removed 4586 * 4587 * Remove all addresses that were added to the device by dev_mc_sync(). 4588 */ 4589 static inline void __dev_mc_unsync(struct net_device *dev, 4590 int (*unsync)(struct net_device *, 4591 const unsigned char *)) 4592 { 4593 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4594 } 4595 4596 /* Functions used for secondary unicast and multicast support */ 4597 void dev_set_rx_mode(struct net_device *dev); 4598 int dev_set_promiscuity(struct net_device *dev, int inc); 4599 int dev_set_allmulti(struct net_device *dev, int inc); 4600 void netdev_state_change(struct net_device *dev); 4601 void __netdev_notify_peers(struct net_device *dev); 4602 void netdev_notify_peers(struct net_device *dev); 4603 void netdev_features_change(struct net_device *dev); 4604 /* Load a device via the kmod */ 4605 void dev_load(struct net *net, const char *name); 4606 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4607 struct rtnl_link_stats64 *storage); 4608 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4609 const struct net_device_stats *netdev_stats); 4610 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4611 const struct pcpu_sw_netstats __percpu *netstats); 4612 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4613 4614 extern int netdev_max_backlog; 4615 extern int dev_rx_weight; 4616 extern int dev_tx_weight; 4617 extern int gro_normal_batch; 4618 4619 enum { 4620 NESTED_SYNC_IMM_BIT, 4621 NESTED_SYNC_TODO_BIT, 4622 }; 4623 4624 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4625 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4626 4627 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4628 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4629 4630 struct netdev_nested_priv { 4631 unsigned char flags; 4632 void *data; 4633 }; 4634 4635 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4636 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4637 struct list_head **iter); 4638 4639 /* iterate through upper list, must be called under RCU read lock */ 4640 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4641 for (iter = &(dev)->adj_list.upper, \ 4642 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4643 updev; \ 4644 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4645 4646 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4647 int (*fn)(struct net_device *upper_dev, 4648 struct netdev_nested_priv *priv), 4649 struct netdev_nested_priv *priv); 4650 4651 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4652 struct net_device *upper_dev); 4653 4654 bool netdev_has_any_upper_dev(struct net_device *dev); 4655 4656 void *netdev_lower_get_next_private(struct net_device *dev, 4657 struct list_head **iter); 4658 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4659 struct list_head **iter); 4660 4661 #define netdev_for_each_lower_private(dev, priv, iter) \ 4662 for (iter = (dev)->adj_list.lower.next, \ 4663 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4664 priv; \ 4665 priv = netdev_lower_get_next_private(dev, &(iter))) 4666 4667 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4668 for (iter = &(dev)->adj_list.lower, \ 4669 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4670 priv; \ 4671 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4672 4673 void *netdev_lower_get_next(struct net_device *dev, 4674 struct list_head **iter); 4675 4676 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4677 for (iter = (dev)->adj_list.lower.next, \ 4678 ldev = netdev_lower_get_next(dev, &(iter)); \ 4679 ldev; \ 4680 ldev = netdev_lower_get_next(dev, &(iter))) 4681 4682 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4683 struct list_head **iter); 4684 int netdev_walk_all_lower_dev(struct net_device *dev, 4685 int (*fn)(struct net_device *lower_dev, 4686 struct netdev_nested_priv *priv), 4687 struct netdev_nested_priv *priv); 4688 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4689 int (*fn)(struct net_device *lower_dev, 4690 struct netdev_nested_priv *priv), 4691 struct netdev_nested_priv *priv); 4692 4693 void *netdev_adjacent_get_private(struct list_head *adj_list); 4694 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4695 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4696 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4697 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4698 struct netlink_ext_ack *extack); 4699 int netdev_master_upper_dev_link(struct net_device *dev, 4700 struct net_device *upper_dev, 4701 void *upper_priv, void *upper_info, 4702 struct netlink_ext_ack *extack); 4703 void netdev_upper_dev_unlink(struct net_device *dev, 4704 struct net_device *upper_dev); 4705 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4706 struct net_device *new_dev, 4707 struct net_device *dev, 4708 struct netlink_ext_ack *extack); 4709 void netdev_adjacent_change_commit(struct net_device *old_dev, 4710 struct net_device *new_dev, 4711 struct net_device *dev); 4712 void netdev_adjacent_change_abort(struct net_device *old_dev, 4713 struct net_device *new_dev, 4714 struct net_device *dev); 4715 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4716 void *netdev_lower_dev_get_private(struct net_device *dev, 4717 struct net_device *lower_dev); 4718 void netdev_lower_state_changed(struct net_device *lower_dev, 4719 void *lower_state_info); 4720 4721 /* RSS keys are 40 or 52 bytes long */ 4722 #define NETDEV_RSS_KEY_LEN 52 4723 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4724 void netdev_rss_key_fill(void *buffer, size_t len); 4725 4726 int skb_checksum_help(struct sk_buff *skb); 4727 int skb_crc32c_csum_help(struct sk_buff *skb); 4728 int skb_csum_hwoffload_help(struct sk_buff *skb, 4729 const netdev_features_t features); 4730 4731 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4732 netdev_features_t features, bool tx_path); 4733 struct sk_buff *skb_eth_gso_segment(struct sk_buff *skb, 4734 netdev_features_t features, __be16 type); 4735 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4736 netdev_features_t features); 4737 4738 struct netdev_bonding_info { 4739 ifslave slave; 4740 ifbond master; 4741 }; 4742 4743 struct netdev_notifier_bonding_info { 4744 struct netdev_notifier_info info; /* must be first */ 4745 struct netdev_bonding_info bonding_info; 4746 }; 4747 4748 void netdev_bonding_info_change(struct net_device *dev, 4749 struct netdev_bonding_info *bonding_info); 4750 4751 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4752 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4753 #else 4754 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4755 const void *data) 4756 { 4757 } 4758 #endif 4759 4760 static inline 4761 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4762 { 4763 return __skb_gso_segment(skb, features, true); 4764 } 4765 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4766 4767 static inline bool can_checksum_protocol(netdev_features_t features, 4768 __be16 protocol) 4769 { 4770 if (protocol == htons(ETH_P_FCOE)) 4771 return !!(features & NETIF_F_FCOE_CRC); 4772 4773 /* Assume this is an IP checksum (not SCTP CRC) */ 4774 4775 if (features & NETIF_F_HW_CSUM) { 4776 /* Can checksum everything */ 4777 return true; 4778 } 4779 4780 switch (protocol) { 4781 case htons(ETH_P_IP): 4782 return !!(features & NETIF_F_IP_CSUM); 4783 case htons(ETH_P_IPV6): 4784 return !!(features & NETIF_F_IPV6_CSUM); 4785 default: 4786 return false; 4787 } 4788 } 4789 4790 #ifdef CONFIG_BUG 4791 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4792 #else 4793 static inline void netdev_rx_csum_fault(struct net_device *dev, 4794 struct sk_buff *skb) 4795 { 4796 } 4797 #endif 4798 /* rx skb timestamps */ 4799 void net_enable_timestamp(void); 4800 void net_disable_timestamp(void); 4801 4802 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 4803 const struct skb_shared_hwtstamps *hwtstamps, 4804 bool cycles) 4805 { 4806 const struct net_device_ops *ops = dev->netdev_ops; 4807 4808 if (ops->ndo_get_tstamp) 4809 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 4810 4811 return hwtstamps->hwtstamp; 4812 } 4813 4814 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4815 struct sk_buff *skb, struct net_device *dev, 4816 bool more) 4817 { 4818 __this_cpu_write(softnet_data.xmit.more, more); 4819 return ops->ndo_start_xmit(skb, dev); 4820 } 4821 4822 static inline bool netdev_xmit_more(void) 4823 { 4824 return __this_cpu_read(softnet_data.xmit.more); 4825 } 4826 4827 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4828 struct netdev_queue *txq, bool more) 4829 { 4830 const struct net_device_ops *ops = dev->netdev_ops; 4831 netdev_tx_t rc; 4832 4833 rc = __netdev_start_xmit(ops, skb, dev, more); 4834 if (rc == NETDEV_TX_OK) 4835 txq_trans_update(txq); 4836 4837 return rc; 4838 } 4839 4840 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4841 const void *ns); 4842 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4843 const void *ns); 4844 4845 extern const struct kobj_ns_type_operations net_ns_type_operations; 4846 4847 const char *netdev_drivername(const struct net_device *dev); 4848 4849 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4850 netdev_features_t f2) 4851 { 4852 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4853 if (f1 & NETIF_F_HW_CSUM) 4854 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4855 else 4856 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4857 } 4858 4859 return f1 & f2; 4860 } 4861 4862 static inline netdev_features_t netdev_get_wanted_features( 4863 struct net_device *dev) 4864 { 4865 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4866 } 4867 netdev_features_t netdev_increment_features(netdev_features_t all, 4868 netdev_features_t one, netdev_features_t mask); 4869 4870 /* Allow TSO being used on stacked device : 4871 * Performing the GSO segmentation before last device 4872 * is a performance improvement. 4873 */ 4874 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4875 netdev_features_t mask) 4876 { 4877 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4878 } 4879 4880 int __netdev_update_features(struct net_device *dev); 4881 void netdev_update_features(struct net_device *dev); 4882 void netdev_change_features(struct net_device *dev); 4883 4884 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4885 struct net_device *dev); 4886 4887 netdev_features_t passthru_features_check(struct sk_buff *skb, 4888 struct net_device *dev, 4889 netdev_features_t features); 4890 netdev_features_t netif_skb_features(struct sk_buff *skb); 4891 4892 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4893 { 4894 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4895 4896 /* check flags correspondence */ 4897 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4898 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4899 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4900 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4901 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4902 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4903 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4904 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4905 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4906 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4907 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4908 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4909 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4910 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4911 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4912 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4913 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4914 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4915 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4916 4917 return (features & feature) == feature; 4918 } 4919 4920 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4921 { 4922 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4923 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4924 } 4925 4926 static inline bool netif_needs_gso(struct sk_buff *skb, 4927 netdev_features_t features) 4928 { 4929 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4930 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4931 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4932 } 4933 4934 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 4935 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 4936 void netif_inherit_tso_max(struct net_device *to, 4937 const struct net_device *from); 4938 4939 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4940 int pulled_hlen, u16 mac_offset, 4941 int mac_len) 4942 { 4943 skb->protocol = protocol; 4944 skb->encapsulation = 1; 4945 skb_push(skb, pulled_hlen); 4946 skb_reset_transport_header(skb); 4947 skb->mac_header = mac_offset; 4948 skb->network_header = skb->mac_header + mac_len; 4949 skb->mac_len = mac_len; 4950 } 4951 4952 static inline bool netif_is_macsec(const struct net_device *dev) 4953 { 4954 return dev->priv_flags & IFF_MACSEC; 4955 } 4956 4957 static inline bool netif_is_macvlan(const struct net_device *dev) 4958 { 4959 return dev->priv_flags & IFF_MACVLAN; 4960 } 4961 4962 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4963 { 4964 return dev->priv_flags & IFF_MACVLAN_PORT; 4965 } 4966 4967 static inline bool netif_is_bond_master(const struct net_device *dev) 4968 { 4969 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4970 } 4971 4972 static inline bool netif_is_bond_slave(const struct net_device *dev) 4973 { 4974 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4975 } 4976 4977 static inline bool netif_supports_nofcs(struct net_device *dev) 4978 { 4979 return dev->priv_flags & IFF_SUPP_NOFCS; 4980 } 4981 4982 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4983 { 4984 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4985 } 4986 4987 static inline bool netif_is_l3_master(const struct net_device *dev) 4988 { 4989 return dev->priv_flags & IFF_L3MDEV_MASTER; 4990 } 4991 4992 static inline bool netif_is_l3_slave(const struct net_device *dev) 4993 { 4994 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4995 } 4996 4997 static inline bool netif_is_bridge_master(const struct net_device *dev) 4998 { 4999 return dev->priv_flags & IFF_EBRIDGE; 5000 } 5001 5002 static inline bool netif_is_bridge_port(const struct net_device *dev) 5003 { 5004 return dev->priv_flags & IFF_BRIDGE_PORT; 5005 } 5006 5007 static inline bool netif_is_ovs_master(const struct net_device *dev) 5008 { 5009 return dev->priv_flags & IFF_OPENVSWITCH; 5010 } 5011 5012 static inline bool netif_is_ovs_port(const struct net_device *dev) 5013 { 5014 return dev->priv_flags & IFF_OVS_DATAPATH; 5015 } 5016 5017 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5018 { 5019 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5020 } 5021 5022 static inline bool netif_is_team_master(const struct net_device *dev) 5023 { 5024 return dev->priv_flags & IFF_TEAM; 5025 } 5026 5027 static inline bool netif_is_team_port(const struct net_device *dev) 5028 { 5029 return dev->priv_flags & IFF_TEAM_PORT; 5030 } 5031 5032 static inline bool netif_is_lag_master(const struct net_device *dev) 5033 { 5034 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5035 } 5036 5037 static inline bool netif_is_lag_port(const struct net_device *dev) 5038 { 5039 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5040 } 5041 5042 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5043 { 5044 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5045 } 5046 5047 static inline bool netif_is_failover(const struct net_device *dev) 5048 { 5049 return dev->priv_flags & IFF_FAILOVER; 5050 } 5051 5052 static inline bool netif_is_failover_slave(const struct net_device *dev) 5053 { 5054 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5055 } 5056 5057 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5058 static inline void netif_keep_dst(struct net_device *dev) 5059 { 5060 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5061 } 5062 5063 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5064 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5065 { 5066 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5067 return netif_is_macsec(dev); 5068 } 5069 5070 extern struct pernet_operations __net_initdata loopback_net_ops; 5071 5072 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5073 5074 /* netdev_printk helpers, similar to dev_printk */ 5075 5076 static inline const char *netdev_name(const struct net_device *dev) 5077 { 5078 if (!dev->name[0] || strchr(dev->name, '%')) 5079 return "(unnamed net_device)"; 5080 return dev->name; 5081 } 5082 5083 static inline bool netdev_unregistering(const struct net_device *dev) 5084 { 5085 return dev->reg_state == NETREG_UNREGISTERING; 5086 } 5087 5088 static inline const char *netdev_reg_state(const struct net_device *dev) 5089 { 5090 switch (dev->reg_state) { 5091 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5092 case NETREG_REGISTERED: return ""; 5093 case NETREG_UNREGISTERING: return " (unregistering)"; 5094 case NETREG_UNREGISTERED: return " (unregistered)"; 5095 case NETREG_RELEASED: return " (released)"; 5096 case NETREG_DUMMY: return " (dummy)"; 5097 } 5098 5099 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5100 return " (unknown)"; 5101 } 5102 5103 #define MODULE_ALIAS_NETDEV(device) \ 5104 MODULE_ALIAS("netdev-" device) 5105 5106 /* 5107 * netdev_WARN() acts like dev_printk(), but with the key difference 5108 * of using a WARN/WARN_ON to get the message out, including the 5109 * file/line information and a backtrace. 5110 */ 5111 #define netdev_WARN(dev, format, args...) \ 5112 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5113 netdev_reg_state(dev), ##args) 5114 5115 #define netdev_WARN_ONCE(dev, format, args...) \ 5116 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5117 netdev_reg_state(dev), ##args) 5118 5119 /* 5120 * The list of packet types we will receive (as opposed to discard) 5121 * and the routines to invoke. 5122 * 5123 * Why 16. Because with 16 the only overlap we get on a hash of the 5124 * low nibble of the protocol value is RARP/SNAP/X.25. 5125 * 5126 * 0800 IP 5127 * 0001 802.3 5128 * 0002 AX.25 5129 * 0004 802.2 5130 * 8035 RARP 5131 * 0005 SNAP 5132 * 0805 X.25 5133 * 0806 ARP 5134 * 8137 IPX 5135 * 0009 Localtalk 5136 * 86DD IPv6 5137 */ 5138 #define PTYPE_HASH_SIZE (16) 5139 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5140 5141 extern struct list_head ptype_all __read_mostly; 5142 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5143 5144 extern struct net_device *blackhole_netdev; 5145 5146 #endif /* _LINUX_NETDEVICE_H */ 5147