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