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 #define netdev_for_each_synced_uc_addr(_ha, _dev) \ 257 netdev_for_each_uc_addr((_ha), (_dev)) \ 258 if ((_ha)->sync_cnt) 259 260 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 261 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 262 #define netdev_for_each_mc_addr(ha, dev) \ 263 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 264 #define netdev_for_each_synced_mc_addr(_ha, _dev) \ 265 netdev_for_each_mc_addr((_ha), (_dev)) \ 266 if ((_ha)->sync_cnt) 267 268 struct hh_cache { 269 unsigned int hh_len; 270 seqlock_t hh_lock; 271 272 /* cached hardware header; allow for machine alignment needs. */ 273 #define HH_DATA_MOD 16 274 #define HH_DATA_OFF(__len) \ 275 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 276 #define HH_DATA_ALIGN(__len) \ 277 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 278 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 279 }; 280 281 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 282 * Alternative is: 283 * dev->hard_header_len ? (dev->hard_header_len + 284 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 285 * 286 * We could use other alignment values, but we must maintain the 287 * relationship HH alignment <= LL alignment. 288 */ 289 #define LL_RESERVED_SPACE(dev) \ 290 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 291 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 292 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 293 294 struct header_ops { 295 int (*create) (struct sk_buff *skb, struct net_device *dev, 296 unsigned short type, const void *daddr, 297 const void *saddr, unsigned int len); 298 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 299 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 300 void (*cache_update)(struct hh_cache *hh, 301 const struct net_device *dev, 302 const unsigned char *haddr); 303 bool (*validate)(const char *ll_header, unsigned int len); 304 __be16 (*parse_protocol)(const struct sk_buff *skb); 305 }; 306 307 /* These flag bits are private to the generic network queueing 308 * layer; they may not be explicitly referenced by any other 309 * code. 310 */ 311 312 enum netdev_state_t { 313 __LINK_STATE_START, 314 __LINK_STATE_PRESENT, 315 __LINK_STATE_NOCARRIER, 316 __LINK_STATE_LINKWATCH_PENDING, 317 __LINK_STATE_DORMANT, 318 __LINK_STATE_TESTING, 319 }; 320 321 struct gro_list { 322 struct list_head list; 323 int count; 324 }; 325 326 /* 327 * size of gro hash buckets, must less than bit number of 328 * napi_struct::gro_bitmask 329 */ 330 #define GRO_HASH_BUCKETS 8 331 332 /* 333 * Structure for NAPI scheduling similar to tasklet but with weighting 334 */ 335 struct napi_struct { 336 /* The poll_list must only be managed by the entity which 337 * changes the state of the NAPI_STATE_SCHED bit. This means 338 * whoever atomically sets that bit can add this napi_struct 339 * to the per-CPU poll_list, and whoever clears that bit 340 * can remove from the list right before clearing the bit. 341 */ 342 struct list_head poll_list; 343 344 unsigned long state; 345 int weight; 346 int defer_hard_irqs_count; 347 unsigned long gro_bitmask; 348 int (*poll)(struct napi_struct *, int); 349 #ifdef CONFIG_NETPOLL 350 int poll_owner; 351 #endif 352 struct net_device *dev; 353 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 354 struct sk_buff *skb; 355 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 356 int rx_count; /* length of rx_list */ 357 struct hrtimer timer; 358 struct list_head dev_list; 359 struct hlist_node napi_hash_node; 360 unsigned int napi_id; 361 struct task_struct *thread; 362 }; 363 364 enum { 365 NAPI_STATE_SCHED, /* Poll is scheduled */ 366 NAPI_STATE_MISSED, /* reschedule a napi */ 367 NAPI_STATE_DISABLE, /* Disable pending */ 368 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 369 NAPI_STATE_LISTED, /* NAPI added to system lists */ 370 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 371 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 372 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 373 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 374 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 375 }; 376 377 enum { 378 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 379 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 380 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 381 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 382 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 383 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 384 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 385 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 386 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 387 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 388 }; 389 390 enum gro_result { 391 GRO_MERGED, 392 GRO_MERGED_FREE, 393 GRO_HELD, 394 GRO_NORMAL, 395 GRO_CONSUMED, 396 }; 397 typedef enum gro_result gro_result_t; 398 399 /* 400 * enum rx_handler_result - Possible return values for rx_handlers. 401 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 402 * further. 403 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 404 * case skb->dev was changed by rx_handler. 405 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 406 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 407 * 408 * rx_handlers are functions called from inside __netif_receive_skb(), to do 409 * special processing of the skb, prior to delivery to protocol handlers. 410 * 411 * Currently, a net_device can only have a single rx_handler registered. Trying 412 * to register a second rx_handler will return -EBUSY. 413 * 414 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 415 * To unregister a rx_handler on a net_device, use 416 * netdev_rx_handler_unregister(). 417 * 418 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 419 * do with the skb. 420 * 421 * If the rx_handler consumed the skb in some way, it should return 422 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 423 * the skb to be delivered in some other way. 424 * 425 * If the rx_handler changed skb->dev, to divert the skb to another 426 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 427 * new device will be called if it exists. 428 * 429 * If the rx_handler decides the skb should be ignored, it should return 430 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 431 * are registered on exact device (ptype->dev == skb->dev). 432 * 433 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 434 * delivered, it should return RX_HANDLER_PASS. 435 * 436 * A device without a registered rx_handler will behave as if rx_handler 437 * returned RX_HANDLER_PASS. 438 */ 439 440 enum rx_handler_result { 441 RX_HANDLER_CONSUMED, 442 RX_HANDLER_ANOTHER, 443 RX_HANDLER_EXACT, 444 RX_HANDLER_PASS, 445 }; 446 typedef enum rx_handler_result rx_handler_result_t; 447 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 448 449 void __napi_schedule(struct napi_struct *n); 450 void __napi_schedule_irqoff(struct napi_struct *n); 451 452 static inline bool napi_disable_pending(struct napi_struct *n) 453 { 454 return test_bit(NAPI_STATE_DISABLE, &n->state); 455 } 456 457 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 458 { 459 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 460 } 461 462 bool napi_schedule_prep(struct napi_struct *n); 463 464 /** 465 * napi_schedule - schedule NAPI poll 466 * @n: NAPI context 467 * 468 * Schedule NAPI poll routine to be called if it is not already 469 * running. 470 */ 471 static inline void napi_schedule(struct napi_struct *n) 472 { 473 if (napi_schedule_prep(n)) 474 __napi_schedule(n); 475 } 476 477 /** 478 * napi_schedule_irqoff - schedule NAPI poll 479 * @n: NAPI context 480 * 481 * Variant of napi_schedule(), assuming hard irqs are masked. 482 */ 483 static inline void napi_schedule_irqoff(struct napi_struct *n) 484 { 485 if (napi_schedule_prep(n)) 486 __napi_schedule_irqoff(n); 487 } 488 489 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 490 static inline bool napi_reschedule(struct napi_struct *napi) 491 { 492 if (napi_schedule_prep(napi)) { 493 __napi_schedule(napi); 494 return true; 495 } 496 return false; 497 } 498 499 bool napi_complete_done(struct napi_struct *n, int work_done); 500 /** 501 * napi_complete - NAPI processing complete 502 * @n: NAPI context 503 * 504 * Mark NAPI processing as complete. 505 * Consider using napi_complete_done() instead. 506 * Return false if device should avoid rearming interrupts. 507 */ 508 static inline bool napi_complete(struct napi_struct *n) 509 { 510 return napi_complete_done(n, 0); 511 } 512 513 int dev_set_threaded(struct net_device *dev, bool threaded); 514 515 /** 516 * napi_disable - prevent NAPI from scheduling 517 * @n: NAPI context 518 * 519 * Stop NAPI from being scheduled on this context. 520 * Waits till any outstanding processing completes. 521 */ 522 void napi_disable(struct napi_struct *n); 523 524 void napi_enable(struct napi_struct *n); 525 526 /** 527 * napi_synchronize - wait until NAPI is not running 528 * @n: NAPI context 529 * 530 * Wait until NAPI is done being scheduled on this context. 531 * Waits till any outstanding processing completes but 532 * does not disable future activations. 533 */ 534 static inline void napi_synchronize(const struct napi_struct *n) 535 { 536 if (IS_ENABLED(CONFIG_SMP)) 537 while (test_bit(NAPI_STATE_SCHED, &n->state)) 538 msleep(1); 539 else 540 barrier(); 541 } 542 543 /** 544 * napi_if_scheduled_mark_missed - if napi is running, set the 545 * NAPIF_STATE_MISSED 546 * @n: NAPI context 547 * 548 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 549 * NAPI is scheduled. 550 **/ 551 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 552 { 553 unsigned long val, new; 554 555 val = READ_ONCE(n->state); 556 do { 557 if (val & NAPIF_STATE_DISABLE) 558 return true; 559 560 if (!(val & NAPIF_STATE_SCHED)) 561 return false; 562 563 new = val | NAPIF_STATE_MISSED; 564 } while (!try_cmpxchg(&n->state, &val, new)); 565 566 return true; 567 } 568 569 enum netdev_queue_state_t { 570 __QUEUE_STATE_DRV_XOFF, 571 __QUEUE_STATE_STACK_XOFF, 572 __QUEUE_STATE_FROZEN, 573 }; 574 575 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 576 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 577 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 578 579 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 580 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 581 QUEUE_STATE_FROZEN) 582 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 583 QUEUE_STATE_FROZEN) 584 585 /* 586 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 587 * netif_tx_* functions below are used to manipulate this flag. The 588 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 589 * queue independently. The netif_xmit_*stopped functions below are called 590 * to check if the queue has been stopped by the driver or stack (either 591 * of the XOFF bits are set in the state). Drivers should not need to call 592 * netif_xmit*stopped functions, they should only be using netif_tx_*. 593 */ 594 595 struct netdev_queue { 596 /* 597 * read-mostly part 598 */ 599 struct net_device *dev; 600 netdevice_tracker dev_tracker; 601 602 struct Qdisc __rcu *qdisc; 603 struct Qdisc *qdisc_sleeping; 604 #ifdef CONFIG_SYSFS 605 struct kobject kobj; 606 #endif 607 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 608 int numa_node; 609 #endif 610 unsigned long tx_maxrate; 611 /* 612 * Number of TX timeouts for this queue 613 * (/sys/class/net/DEV/Q/trans_timeout) 614 */ 615 atomic_long_t trans_timeout; 616 617 /* Subordinate device that the queue has been assigned to */ 618 struct net_device *sb_dev; 619 #ifdef CONFIG_XDP_SOCKETS 620 struct xsk_buff_pool *pool; 621 #endif 622 /* 623 * write-mostly part 624 */ 625 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 626 int xmit_lock_owner; 627 /* 628 * Time (in jiffies) of last Tx 629 */ 630 unsigned long trans_start; 631 632 unsigned long state; 633 634 #ifdef CONFIG_BQL 635 struct dql dql; 636 #endif 637 } ____cacheline_aligned_in_smp; 638 639 extern int sysctl_fb_tunnels_only_for_init_net; 640 extern int sysctl_devconf_inherit_init_net; 641 642 /* 643 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 644 * == 1 : For initns only 645 * == 2 : For none. 646 */ 647 static inline bool net_has_fallback_tunnels(const struct net *net) 648 { 649 #if IS_ENABLED(CONFIG_SYSCTL) 650 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net); 651 652 return !fb_tunnels_only_for_init_net || 653 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1); 654 #else 655 return true; 656 #endif 657 } 658 659 static inline int net_inherit_devconf(void) 660 { 661 #if IS_ENABLED(CONFIG_SYSCTL) 662 return READ_ONCE(sysctl_devconf_inherit_init_net); 663 #else 664 return 0; 665 #endif 666 } 667 668 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 669 { 670 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 671 return q->numa_node; 672 #else 673 return NUMA_NO_NODE; 674 #endif 675 } 676 677 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 678 { 679 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 680 q->numa_node = node; 681 #endif 682 } 683 684 #ifdef CONFIG_RPS 685 /* 686 * This structure holds an RPS map which can be of variable length. The 687 * map is an array of CPUs. 688 */ 689 struct rps_map { 690 unsigned int len; 691 struct rcu_head rcu; 692 u16 cpus[]; 693 }; 694 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 695 696 /* 697 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 698 * tail pointer for that CPU's input queue at the time of last enqueue, and 699 * a hardware filter index. 700 */ 701 struct rps_dev_flow { 702 u16 cpu; 703 u16 filter; 704 unsigned int last_qtail; 705 }; 706 #define RPS_NO_FILTER 0xffff 707 708 /* 709 * The rps_dev_flow_table structure contains a table of flow mappings. 710 */ 711 struct rps_dev_flow_table { 712 unsigned int mask; 713 struct rcu_head rcu; 714 struct rps_dev_flow flows[]; 715 }; 716 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 717 ((_num) * sizeof(struct rps_dev_flow))) 718 719 /* 720 * The rps_sock_flow_table contains mappings of flows to the last CPU 721 * on which they were processed by the application (set in recvmsg). 722 * Each entry is a 32bit value. Upper part is the high-order bits 723 * of flow hash, lower part is CPU number. 724 * rps_cpu_mask is used to partition the space, depending on number of 725 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 726 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 727 * meaning we use 32-6=26 bits for the hash. 728 */ 729 struct rps_sock_flow_table { 730 u32 mask; 731 732 u32 ents[] ____cacheline_aligned_in_smp; 733 }; 734 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 735 736 #define RPS_NO_CPU 0xffff 737 738 extern u32 rps_cpu_mask; 739 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 740 741 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 742 u32 hash) 743 { 744 if (table && hash) { 745 unsigned int index = hash & table->mask; 746 u32 val = hash & ~rps_cpu_mask; 747 748 /* We only give a hint, preemption can change CPU under us */ 749 val |= raw_smp_processor_id(); 750 751 if (table->ents[index] != val) 752 table->ents[index] = val; 753 } 754 } 755 756 #ifdef CONFIG_RFS_ACCEL 757 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 758 u16 filter_id); 759 #endif 760 #endif /* CONFIG_RPS */ 761 762 /* This structure contains an instance of an RX queue. */ 763 struct netdev_rx_queue { 764 struct xdp_rxq_info xdp_rxq; 765 #ifdef CONFIG_RPS 766 struct rps_map __rcu *rps_map; 767 struct rps_dev_flow_table __rcu *rps_flow_table; 768 #endif 769 struct kobject kobj; 770 struct net_device *dev; 771 netdevice_tracker dev_tracker; 772 773 #ifdef CONFIG_XDP_SOCKETS 774 struct xsk_buff_pool *pool; 775 #endif 776 } ____cacheline_aligned_in_smp; 777 778 /* 779 * RX queue sysfs structures and functions. 780 */ 781 struct rx_queue_attribute { 782 struct attribute attr; 783 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 784 ssize_t (*store)(struct netdev_rx_queue *queue, 785 const char *buf, size_t len); 786 }; 787 788 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 789 enum xps_map_type { 790 XPS_CPUS = 0, 791 XPS_RXQS, 792 XPS_MAPS_MAX, 793 }; 794 795 #ifdef CONFIG_XPS 796 /* 797 * This structure holds an XPS map which can be of variable length. The 798 * map is an array of queues. 799 */ 800 struct xps_map { 801 unsigned int len; 802 unsigned int alloc_len; 803 struct rcu_head rcu; 804 u16 queues[]; 805 }; 806 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 807 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 808 - sizeof(struct xps_map)) / sizeof(u16)) 809 810 /* 811 * This structure holds all XPS maps for device. Maps are indexed by CPU. 812 * 813 * We keep track of the number of cpus/rxqs used when the struct is allocated, 814 * in nr_ids. This will help not accessing out-of-bound memory. 815 * 816 * We keep track of the number of traffic classes used when the struct is 817 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 818 * not crossing its upper bound, as the original dev->num_tc can be updated in 819 * the meantime. 820 */ 821 struct xps_dev_maps { 822 struct rcu_head rcu; 823 unsigned int nr_ids; 824 s16 num_tc; 825 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 826 }; 827 828 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 829 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 830 831 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 832 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 833 834 #endif /* CONFIG_XPS */ 835 836 #define TC_MAX_QUEUE 16 837 #define TC_BITMASK 15 838 /* HW offloaded queuing disciplines txq count and offset maps */ 839 struct netdev_tc_txq { 840 u16 count; 841 u16 offset; 842 }; 843 844 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 845 /* 846 * This structure is to hold information about the device 847 * configured to run FCoE protocol stack. 848 */ 849 struct netdev_fcoe_hbainfo { 850 char manufacturer[64]; 851 char serial_number[64]; 852 char hardware_version[64]; 853 char driver_version[64]; 854 char optionrom_version[64]; 855 char firmware_version[64]; 856 char model[256]; 857 char model_description[256]; 858 }; 859 #endif 860 861 #define MAX_PHYS_ITEM_ID_LEN 32 862 863 /* This structure holds a unique identifier to identify some 864 * physical item (port for example) used by a netdevice. 865 */ 866 struct netdev_phys_item_id { 867 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 868 unsigned char id_len; 869 }; 870 871 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 872 struct netdev_phys_item_id *b) 873 { 874 return a->id_len == b->id_len && 875 memcmp(a->id, b->id, a->id_len) == 0; 876 } 877 878 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 879 struct sk_buff *skb, 880 struct net_device *sb_dev); 881 882 enum net_device_path_type { 883 DEV_PATH_ETHERNET = 0, 884 DEV_PATH_VLAN, 885 DEV_PATH_BRIDGE, 886 DEV_PATH_PPPOE, 887 DEV_PATH_DSA, 888 DEV_PATH_MTK_WDMA, 889 }; 890 891 struct net_device_path { 892 enum net_device_path_type type; 893 const struct net_device *dev; 894 union { 895 struct { 896 u16 id; 897 __be16 proto; 898 u8 h_dest[ETH_ALEN]; 899 } encap; 900 struct { 901 enum { 902 DEV_PATH_BR_VLAN_KEEP, 903 DEV_PATH_BR_VLAN_TAG, 904 DEV_PATH_BR_VLAN_UNTAG, 905 DEV_PATH_BR_VLAN_UNTAG_HW, 906 } vlan_mode; 907 u16 vlan_id; 908 __be16 vlan_proto; 909 } bridge; 910 struct { 911 int port; 912 u16 proto; 913 } dsa; 914 struct { 915 u8 wdma_idx; 916 u8 queue; 917 u16 wcid; 918 u8 bss; 919 } mtk_wdma; 920 }; 921 }; 922 923 #define NET_DEVICE_PATH_STACK_MAX 5 924 #define NET_DEVICE_PATH_VLAN_MAX 2 925 926 struct net_device_path_stack { 927 int num_paths; 928 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 929 }; 930 931 struct net_device_path_ctx { 932 const struct net_device *dev; 933 u8 daddr[ETH_ALEN]; 934 935 int num_vlans; 936 struct { 937 u16 id; 938 __be16 proto; 939 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 940 }; 941 942 enum tc_setup_type { 943 TC_SETUP_QDISC_MQPRIO, 944 TC_SETUP_CLSU32, 945 TC_SETUP_CLSFLOWER, 946 TC_SETUP_CLSMATCHALL, 947 TC_SETUP_CLSBPF, 948 TC_SETUP_BLOCK, 949 TC_SETUP_QDISC_CBS, 950 TC_SETUP_QDISC_RED, 951 TC_SETUP_QDISC_PRIO, 952 TC_SETUP_QDISC_MQ, 953 TC_SETUP_QDISC_ETF, 954 TC_SETUP_ROOT_QDISC, 955 TC_SETUP_QDISC_GRED, 956 TC_SETUP_QDISC_TAPRIO, 957 TC_SETUP_FT, 958 TC_SETUP_QDISC_ETS, 959 TC_SETUP_QDISC_TBF, 960 TC_SETUP_QDISC_FIFO, 961 TC_SETUP_QDISC_HTB, 962 TC_SETUP_ACT, 963 }; 964 965 /* These structures hold the attributes of bpf state that are being passed 966 * to the netdevice through the bpf op. 967 */ 968 enum bpf_netdev_command { 969 /* Set or clear a bpf program used in the earliest stages of packet 970 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 971 * is responsible for calling bpf_prog_put on any old progs that are 972 * stored. In case of error, the callee need not release the new prog 973 * reference, but on success it takes ownership and must bpf_prog_put 974 * when it is no longer used. 975 */ 976 XDP_SETUP_PROG, 977 XDP_SETUP_PROG_HW, 978 /* BPF program for offload callbacks, invoked at program load time. */ 979 BPF_OFFLOAD_MAP_ALLOC, 980 BPF_OFFLOAD_MAP_FREE, 981 XDP_SETUP_XSK_POOL, 982 }; 983 984 struct bpf_prog_offload_ops; 985 struct netlink_ext_ack; 986 struct xdp_umem; 987 struct xdp_dev_bulk_queue; 988 struct bpf_xdp_link; 989 990 enum bpf_xdp_mode { 991 XDP_MODE_SKB = 0, 992 XDP_MODE_DRV = 1, 993 XDP_MODE_HW = 2, 994 __MAX_XDP_MODE 995 }; 996 997 struct bpf_xdp_entity { 998 struct bpf_prog *prog; 999 struct bpf_xdp_link *link; 1000 }; 1001 1002 struct netdev_bpf { 1003 enum bpf_netdev_command command; 1004 union { 1005 /* XDP_SETUP_PROG */ 1006 struct { 1007 u32 flags; 1008 struct bpf_prog *prog; 1009 struct netlink_ext_ack *extack; 1010 }; 1011 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 1012 struct { 1013 struct bpf_offloaded_map *offmap; 1014 }; 1015 /* XDP_SETUP_XSK_POOL */ 1016 struct { 1017 struct xsk_buff_pool *pool; 1018 u16 queue_id; 1019 } xsk; 1020 }; 1021 }; 1022 1023 /* Flags for ndo_xsk_wakeup. */ 1024 #define XDP_WAKEUP_RX (1 << 0) 1025 #define XDP_WAKEUP_TX (1 << 1) 1026 1027 #ifdef CONFIG_XFRM_OFFLOAD 1028 struct xfrmdev_ops { 1029 int (*xdo_dev_state_add) (struct xfrm_state *x); 1030 void (*xdo_dev_state_delete) (struct xfrm_state *x); 1031 void (*xdo_dev_state_free) (struct xfrm_state *x); 1032 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 1033 struct xfrm_state *x); 1034 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1035 }; 1036 #endif 1037 1038 struct dev_ifalias { 1039 struct rcu_head rcuhead; 1040 char ifalias[]; 1041 }; 1042 1043 struct devlink; 1044 struct tlsdev_ops; 1045 1046 struct netdev_net_notifier { 1047 struct list_head list; 1048 struct notifier_block *nb; 1049 }; 1050 1051 /* 1052 * This structure defines the management hooks for network devices. 1053 * The following hooks can be defined; unless noted otherwise, they are 1054 * optional and can be filled with a null pointer. 1055 * 1056 * int (*ndo_init)(struct net_device *dev); 1057 * This function is called once when a network device is registered. 1058 * The network device can use this for any late stage initialization 1059 * or semantic validation. It can fail with an error code which will 1060 * be propagated back to register_netdev. 1061 * 1062 * void (*ndo_uninit)(struct net_device *dev); 1063 * This function is called when device is unregistered or when registration 1064 * fails. It is not called if init fails. 1065 * 1066 * int (*ndo_open)(struct net_device *dev); 1067 * This function is called when a network device transitions to the up 1068 * state. 1069 * 1070 * int (*ndo_stop)(struct net_device *dev); 1071 * This function is called when a network device transitions to the down 1072 * state. 1073 * 1074 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1075 * struct net_device *dev); 1076 * Called when a packet needs to be transmitted. 1077 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1078 * the queue before that can happen; it's for obsolete devices and weird 1079 * corner cases, but the stack really does a non-trivial amount 1080 * of useless work if you return NETDEV_TX_BUSY. 1081 * Required; cannot be NULL. 1082 * 1083 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1084 * struct net_device *dev 1085 * netdev_features_t features); 1086 * Called by core transmit path to determine if device is capable of 1087 * performing offload operations on a given packet. This is to give 1088 * the device an opportunity to implement any restrictions that cannot 1089 * be otherwise expressed by feature flags. The check is called with 1090 * the set of features that the stack has calculated and it returns 1091 * those the driver believes to be appropriate. 1092 * 1093 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1094 * struct net_device *sb_dev); 1095 * Called to decide which queue to use when device supports multiple 1096 * transmit queues. 1097 * 1098 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1099 * This function is called to allow device receiver to make 1100 * changes to configuration when multicast or promiscuous is enabled. 1101 * 1102 * void (*ndo_set_rx_mode)(struct net_device *dev); 1103 * This function is called device changes address list filtering. 1104 * If driver handles unicast address filtering, it should set 1105 * IFF_UNICAST_FLT in its priv_flags. 1106 * 1107 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1108 * This function is called when the Media Access Control address 1109 * needs to be changed. If this interface is not defined, the 1110 * MAC address can not be changed. 1111 * 1112 * int (*ndo_validate_addr)(struct net_device *dev); 1113 * Test if Media Access Control address is valid for the device. 1114 * 1115 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1116 * Old-style ioctl entry point. This is used internally by the 1117 * appletalk and ieee802154 subsystems but is no longer called by 1118 * the device ioctl handler. 1119 * 1120 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1121 * Used by the bonding driver for its device specific ioctls: 1122 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1123 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1124 * 1125 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1126 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1127 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1128 * 1129 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1130 * Used to set network devices bus interface parameters. This interface 1131 * is retained for legacy reasons; new devices should use the bus 1132 * interface (PCI) for low level management. 1133 * 1134 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1135 * Called when a user wants to change the Maximum Transfer Unit 1136 * of a device. 1137 * 1138 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1139 * Callback used when the transmitter has not made any progress 1140 * for dev->watchdog ticks. 1141 * 1142 * void (*ndo_get_stats64)(struct net_device *dev, 1143 * struct rtnl_link_stats64 *storage); 1144 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1145 * Called when a user wants to get the network device usage 1146 * statistics. Drivers must do one of the following: 1147 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1148 * rtnl_link_stats64 structure passed by the caller. 1149 * 2. Define @ndo_get_stats to update a net_device_stats structure 1150 * (which should normally be dev->stats) and return a pointer to 1151 * it. The structure may be changed asynchronously only if each 1152 * field is written atomically. 1153 * 3. Update dev->stats asynchronously and atomically, and define 1154 * neither operation. 1155 * 1156 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1157 * Return true if this device supports offload stats of this attr_id. 1158 * 1159 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1160 * void *attr_data) 1161 * Get statistics for offload operations by attr_id. Write it into the 1162 * attr_data pointer. 1163 * 1164 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1165 * If device supports VLAN filtering this function is called when a 1166 * VLAN id is registered. 1167 * 1168 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1169 * If device supports VLAN filtering this function is called when a 1170 * VLAN id is unregistered. 1171 * 1172 * void (*ndo_poll_controller)(struct net_device *dev); 1173 * 1174 * SR-IOV management functions. 1175 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1176 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1177 * u8 qos, __be16 proto); 1178 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1179 * int max_tx_rate); 1180 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1181 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1182 * int (*ndo_get_vf_config)(struct net_device *dev, 1183 * int vf, struct ifla_vf_info *ivf); 1184 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1185 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1186 * struct nlattr *port[]); 1187 * 1188 * Enable or disable the VF ability to query its RSS Redirection Table and 1189 * Hash Key. This is needed since on some devices VF share this information 1190 * with PF and querying it may introduce a theoretical security risk. 1191 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1192 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1193 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1194 * void *type_data); 1195 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1196 * This is always called from the stack with the rtnl lock held and netif 1197 * tx queues stopped. This allows the netdevice to perform queue 1198 * management safely. 1199 * 1200 * Fiber Channel over Ethernet (FCoE) offload functions. 1201 * int (*ndo_fcoe_enable)(struct net_device *dev); 1202 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1203 * so the underlying device can perform whatever needed configuration or 1204 * initialization to support acceleration of FCoE traffic. 1205 * 1206 * int (*ndo_fcoe_disable)(struct net_device *dev); 1207 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1208 * so the underlying device can perform whatever needed clean-ups to 1209 * stop supporting acceleration of FCoE traffic. 1210 * 1211 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1212 * struct scatterlist *sgl, unsigned int sgc); 1213 * Called when the FCoE Initiator wants to initialize an I/O that 1214 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1215 * perform necessary setup and returns 1 to indicate the device is set up 1216 * successfully to perform DDP on this I/O, otherwise this returns 0. 1217 * 1218 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1219 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1220 * indicated by the FC exchange id 'xid', so the underlying device can 1221 * clean up and reuse resources for later DDP requests. 1222 * 1223 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1224 * struct scatterlist *sgl, unsigned int sgc); 1225 * Called when the FCoE Target wants to initialize an I/O that 1226 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1227 * perform necessary setup and returns 1 to indicate the device is set up 1228 * successfully to perform DDP on this I/O, otherwise this returns 0. 1229 * 1230 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1231 * struct netdev_fcoe_hbainfo *hbainfo); 1232 * Called when the FCoE Protocol stack wants information on the underlying 1233 * device. This information is utilized by the FCoE protocol stack to 1234 * register attributes with Fiber Channel management service as per the 1235 * FC-GS Fabric Device Management Information(FDMI) specification. 1236 * 1237 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1238 * Called when the underlying device wants to override default World Wide 1239 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1240 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1241 * protocol stack to use. 1242 * 1243 * RFS acceleration. 1244 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1245 * u16 rxq_index, u32 flow_id); 1246 * Set hardware filter for RFS. rxq_index is the target queue index; 1247 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1248 * Return the filter ID on success, or a negative error code. 1249 * 1250 * Slave management functions (for bridge, bonding, etc). 1251 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1252 * Called to make another netdev an underling. 1253 * 1254 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1255 * Called to release previously enslaved netdev. 1256 * 1257 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1258 * struct sk_buff *skb, 1259 * bool all_slaves); 1260 * Get the xmit slave of master device. If all_slaves is true, function 1261 * assume all the slaves can transmit. 1262 * 1263 * Feature/offload setting functions. 1264 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1265 * netdev_features_t features); 1266 * Adjusts the requested feature flags according to device-specific 1267 * constraints, and returns the resulting flags. Must not modify 1268 * the device state. 1269 * 1270 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1271 * Called to update device configuration to new features. Passed 1272 * feature set might be less than what was returned by ndo_fix_features()). 1273 * Must return >0 or -errno if it changed dev->features itself. 1274 * 1275 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1276 * struct net_device *dev, 1277 * const unsigned char *addr, u16 vid, u16 flags, 1278 * struct netlink_ext_ack *extack); 1279 * Adds an FDB entry to dev for addr. 1280 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1281 * struct net_device *dev, 1282 * const unsigned char *addr, u16 vid) 1283 * Deletes the FDB entry from dev coresponding to addr. 1284 * int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, struct nlattr *tb[], 1285 * struct net_device *dev, 1286 * u16 vid, 1287 * struct netlink_ext_ack *extack); 1288 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1289 * struct net_device *dev, struct net_device *filter_dev, 1290 * int *idx) 1291 * Used to add FDB entries to dump requests. Implementers should add 1292 * entries to skb and update idx with the number of entries. 1293 * 1294 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1295 * u16 flags, struct netlink_ext_ack *extack) 1296 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1297 * struct net_device *dev, u32 filter_mask, 1298 * int nlflags) 1299 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1300 * u16 flags); 1301 * 1302 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1303 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1304 * which do not represent real hardware may define this to allow their 1305 * userspace components to manage their virtual carrier state. Devices 1306 * that determine carrier state from physical hardware properties (eg 1307 * network cables) or protocol-dependent mechanisms (eg 1308 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1309 * 1310 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1311 * struct netdev_phys_item_id *ppid); 1312 * Called to get ID of physical port of this device. If driver does 1313 * not implement this, it is assumed that the hw is not able to have 1314 * multiple net devices on single physical port. 1315 * 1316 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1317 * struct netdev_phys_item_id *ppid) 1318 * Called to get the parent ID of the physical port of this device. 1319 * 1320 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1321 * struct net_device *dev) 1322 * Called by upper layer devices to accelerate switching or other 1323 * station functionality into hardware. 'pdev is the lowerdev 1324 * to use for the offload and 'dev' is the net device that will 1325 * back the offload. Returns a pointer to the private structure 1326 * the upper layer will maintain. 1327 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1328 * Called by upper layer device to delete the station created 1329 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1330 * the station and priv is the structure returned by the add 1331 * operation. 1332 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1333 * int queue_index, u32 maxrate); 1334 * Called when a user wants to set a max-rate limitation of specific 1335 * TX queue. 1336 * int (*ndo_get_iflink)(const struct net_device *dev); 1337 * Called to get the iflink value of this device. 1338 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1339 * This function is used to get egress tunnel information for given skb. 1340 * This is useful for retrieving outer tunnel header parameters while 1341 * sampling packet. 1342 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1343 * This function is used to specify the headroom that the skb must 1344 * consider when allocation skb during packet reception. Setting 1345 * appropriate rx headroom value allows avoiding skb head copy on 1346 * forward. Setting a negative value resets the rx headroom to the 1347 * default value. 1348 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1349 * This function is used to set or query state related to XDP on the 1350 * netdevice and manage BPF offload. See definition of 1351 * enum bpf_netdev_command for details. 1352 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1353 * u32 flags); 1354 * This function is used to submit @n XDP packets for transmit on a 1355 * netdevice. Returns number of frames successfully transmitted, frames 1356 * that got dropped are freed/returned via xdp_return_frame(). 1357 * Returns negative number, means general error invoking ndo, meaning 1358 * no frames were xmit'ed and core-caller will free all frames. 1359 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1360 * struct xdp_buff *xdp); 1361 * Get the xmit slave of master device based on the xdp_buff. 1362 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1363 * This function is used to wake up the softirq, ksoftirqd or kthread 1364 * responsible for sending and/or receiving packets on a specific 1365 * queue id bound to an AF_XDP socket. The flags field specifies if 1366 * only RX, only Tx, or both should be woken up using the flags 1367 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1368 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1369 * Get devlink port instance associated with a given netdev. 1370 * Called with a reference on the netdevice and devlink locks only, 1371 * rtnl_lock is not held. 1372 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1373 * int cmd); 1374 * Add, change, delete or get information on an IPv4 tunnel. 1375 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1376 * If a device is paired with a peer device, return the peer instance. 1377 * The caller must be under RCU read context. 1378 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1379 * Get the forwarding path to reach the real device from the HW destination address 1380 * ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1381 * const struct skb_shared_hwtstamps *hwtstamps, 1382 * bool cycles); 1383 * Get hardware timestamp based on normal/adjustable time or free running 1384 * cycle counter. This function is required if physical clock supports a 1385 * free running cycle counter. 1386 */ 1387 struct net_device_ops { 1388 int (*ndo_init)(struct net_device *dev); 1389 void (*ndo_uninit)(struct net_device *dev); 1390 int (*ndo_open)(struct net_device *dev); 1391 int (*ndo_stop)(struct net_device *dev); 1392 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1393 struct net_device *dev); 1394 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1395 struct net_device *dev, 1396 netdev_features_t features); 1397 u16 (*ndo_select_queue)(struct net_device *dev, 1398 struct sk_buff *skb, 1399 struct net_device *sb_dev); 1400 void (*ndo_change_rx_flags)(struct net_device *dev, 1401 int flags); 1402 void (*ndo_set_rx_mode)(struct net_device *dev); 1403 int (*ndo_set_mac_address)(struct net_device *dev, 1404 void *addr); 1405 int (*ndo_validate_addr)(struct net_device *dev); 1406 int (*ndo_do_ioctl)(struct net_device *dev, 1407 struct ifreq *ifr, int cmd); 1408 int (*ndo_eth_ioctl)(struct net_device *dev, 1409 struct ifreq *ifr, int cmd); 1410 int (*ndo_siocbond)(struct net_device *dev, 1411 struct ifreq *ifr, int cmd); 1412 int (*ndo_siocwandev)(struct net_device *dev, 1413 struct if_settings *ifs); 1414 int (*ndo_siocdevprivate)(struct net_device *dev, 1415 struct ifreq *ifr, 1416 void __user *data, int cmd); 1417 int (*ndo_set_config)(struct net_device *dev, 1418 struct ifmap *map); 1419 int (*ndo_change_mtu)(struct net_device *dev, 1420 int new_mtu); 1421 int (*ndo_neigh_setup)(struct net_device *dev, 1422 struct neigh_parms *); 1423 void (*ndo_tx_timeout) (struct net_device *dev, 1424 unsigned int txqueue); 1425 1426 void (*ndo_get_stats64)(struct net_device *dev, 1427 struct rtnl_link_stats64 *storage); 1428 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1429 int (*ndo_get_offload_stats)(int attr_id, 1430 const struct net_device *dev, 1431 void *attr_data); 1432 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1433 1434 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1435 __be16 proto, u16 vid); 1436 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1437 __be16 proto, u16 vid); 1438 #ifdef CONFIG_NET_POLL_CONTROLLER 1439 void (*ndo_poll_controller)(struct net_device *dev); 1440 int (*ndo_netpoll_setup)(struct net_device *dev, 1441 struct netpoll_info *info); 1442 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1443 #endif 1444 int (*ndo_set_vf_mac)(struct net_device *dev, 1445 int queue, u8 *mac); 1446 int (*ndo_set_vf_vlan)(struct net_device *dev, 1447 int queue, u16 vlan, 1448 u8 qos, __be16 proto); 1449 int (*ndo_set_vf_rate)(struct net_device *dev, 1450 int vf, int min_tx_rate, 1451 int max_tx_rate); 1452 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1453 int vf, bool setting); 1454 int (*ndo_set_vf_trust)(struct net_device *dev, 1455 int vf, bool setting); 1456 int (*ndo_get_vf_config)(struct net_device *dev, 1457 int vf, 1458 struct ifla_vf_info *ivf); 1459 int (*ndo_set_vf_link_state)(struct net_device *dev, 1460 int vf, int link_state); 1461 int (*ndo_get_vf_stats)(struct net_device *dev, 1462 int vf, 1463 struct ifla_vf_stats 1464 *vf_stats); 1465 int (*ndo_set_vf_port)(struct net_device *dev, 1466 int vf, 1467 struct nlattr *port[]); 1468 int (*ndo_get_vf_port)(struct net_device *dev, 1469 int vf, struct sk_buff *skb); 1470 int (*ndo_get_vf_guid)(struct net_device *dev, 1471 int vf, 1472 struct ifla_vf_guid *node_guid, 1473 struct ifla_vf_guid *port_guid); 1474 int (*ndo_set_vf_guid)(struct net_device *dev, 1475 int vf, u64 guid, 1476 int guid_type); 1477 int (*ndo_set_vf_rss_query_en)( 1478 struct net_device *dev, 1479 int vf, bool setting); 1480 int (*ndo_setup_tc)(struct net_device *dev, 1481 enum tc_setup_type type, 1482 void *type_data); 1483 #if IS_ENABLED(CONFIG_FCOE) 1484 int (*ndo_fcoe_enable)(struct net_device *dev); 1485 int (*ndo_fcoe_disable)(struct net_device *dev); 1486 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1487 u16 xid, 1488 struct scatterlist *sgl, 1489 unsigned int sgc); 1490 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1491 u16 xid); 1492 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1493 u16 xid, 1494 struct scatterlist *sgl, 1495 unsigned int sgc); 1496 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1497 struct netdev_fcoe_hbainfo *hbainfo); 1498 #endif 1499 1500 #if IS_ENABLED(CONFIG_LIBFCOE) 1501 #define NETDEV_FCOE_WWNN 0 1502 #define NETDEV_FCOE_WWPN 1 1503 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1504 u64 *wwn, int type); 1505 #endif 1506 1507 #ifdef CONFIG_RFS_ACCEL 1508 int (*ndo_rx_flow_steer)(struct net_device *dev, 1509 const struct sk_buff *skb, 1510 u16 rxq_index, 1511 u32 flow_id); 1512 #endif 1513 int (*ndo_add_slave)(struct net_device *dev, 1514 struct net_device *slave_dev, 1515 struct netlink_ext_ack *extack); 1516 int (*ndo_del_slave)(struct net_device *dev, 1517 struct net_device *slave_dev); 1518 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1519 struct sk_buff *skb, 1520 bool all_slaves); 1521 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1522 struct sock *sk); 1523 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1524 netdev_features_t features); 1525 int (*ndo_set_features)(struct net_device *dev, 1526 netdev_features_t features); 1527 int (*ndo_neigh_construct)(struct net_device *dev, 1528 struct neighbour *n); 1529 void (*ndo_neigh_destroy)(struct net_device *dev, 1530 struct neighbour *n); 1531 1532 int (*ndo_fdb_add)(struct ndmsg *ndm, 1533 struct nlattr *tb[], 1534 struct net_device *dev, 1535 const unsigned char *addr, 1536 u16 vid, 1537 u16 flags, 1538 struct netlink_ext_ack *extack); 1539 int (*ndo_fdb_del)(struct ndmsg *ndm, 1540 struct nlattr *tb[], 1541 struct net_device *dev, 1542 const unsigned char *addr, 1543 u16 vid, struct netlink_ext_ack *extack); 1544 int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, 1545 struct nlattr *tb[], 1546 struct net_device *dev, 1547 u16 vid, 1548 struct netlink_ext_ack *extack); 1549 int (*ndo_fdb_dump)(struct sk_buff *skb, 1550 struct netlink_callback *cb, 1551 struct net_device *dev, 1552 struct net_device *filter_dev, 1553 int *idx); 1554 int (*ndo_fdb_get)(struct sk_buff *skb, 1555 struct nlattr *tb[], 1556 struct net_device *dev, 1557 const unsigned char *addr, 1558 u16 vid, u32 portid, u32 seq, 1559 struct netlink_ext_ack *extack); 1560 int (*ndo_bridge_setlink)(struct net_device *dev, 1561 struct nlmsghdr *nlh, 1562 u16 flags, 1563 struct netlink_ext_ack *extack); 1564 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1565 u32 pid, u32 seq, 1566 struct net_device *dev, 1567 u32 filter_mask, 1568 int nlflags); 1569 int (*ndo_bridge_dellink)(struct net_device *dev, 1570 struct nlmsghdr *nlh, 1571 u16 flags); 1572 int (*ndo_change_carrier)(struct net_device *dev, 1573 bool new_carrier); 1574 int (*ndo_get_phys_port_id)(struct net_device *dev, 1575 struct netdev_phys_item_id *ppid); 1576 int (*ndo_get_port_parent_id)(struct net_device *dev, 1577 struct netdev_phys_item_id *ppid); 1578 int (*ndo_get_phys_port_name)(struct net_device *dev, 1579 char *name, size_t len); 1580 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1581 struct net_device *dev); 1582 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1583 void *priv); 1584 1585 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1586 int queue_index, 1587 u32 maxrate); 1588 int (*ndo_get_iflink)(const struct net_device *dev); 1589 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1590 struct sk_buff *skb); 1591 void (*ndo_set_rx_headroom)(struct net_device *dev, 1592 int needed_headroom); 1593 int (*ndo_bpf)(struct net_device *dev, 1594 struct netdev_bpf *bpf); 1595 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1596 struct xdp_frame **xdp, 1597 u32 flags); 1598 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1599 struct xdp_buff *xdp); 1600 int (*ndo_xsk_wakeup)(struct net_device *dev, 1601 u32 queue_id, u32 flags); 1602 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1603 int (*ndo_tunnel_ctl)(struct net_device *dev, 1604 struct ip_tunnel_parm *p, int cmd); 1605 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1606 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1607 struct net_device_path *path); 1608 ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1609 const struct skb_shared_hwtstamps *hwtstamps, 1610 bool cycles); 1611 }; 1612 1613 /** 1614 * enum netdev_priv_flags - &struct net_device priv_flags 1615 * 1616 * These are the &struct net_device, they are only set internally 1617 * by drivers and used in the kernel. These flags are invisible to 1618 * userspace; this means that the order of these flags can change 1619 * during any kernel release. 1620 * 1621 * You should have a pretty good reason to be extending these flags. 1622 * 1623 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1624 * @IFF_EBRIDGE: Ethernet bridging device 1625 * @IFF_BONDING: bonding master or slave 1626 * @IFF_ISATAP: ISATAP interface (RFC4214) 1627 * @IFF_WAN_HDLC: WAN HDLC device 1628 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1629 * release skb->dst 1630 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1631 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1632 * @IFF_MACVLAN_PORT: device used as macvlan port 1633 * @IFF_BRIDGE_PORT: device used as bridge port 1634 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1635 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1636 * @IFF_UNICAST_FLT: Supports unicast filtering 1637 * @IFF_TEAM_PORT: device used as team port 1638 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1639 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1640 * change when it's running 1641 * @IFF_MACVLAN: Macvlan device 1642 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1643 * underlying stacked devices 1644 * @IFF_L3MDEV_MASTER: device is an L3 master device 1645 * @IFF_NO_QUEUE: device can run without qdisc attached 1646 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1647 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1648 * @IFF_TEAM: device is a team device 1649 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1650 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1651 * entity (i.e. the master device for bridged veth) 1652 * @IFF_MACSEC: device is a MACsec device 1653 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1654 * @IFF_FAILOVER: device is a failover master device 1655 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1656 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1657 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1658 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1659 * skb_headlen(skb) == 0 (data starts from frag0) 1660 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN 1661 */ 1662 enum netdev_priv_flags { 1663 IFF_802_1Q_VLAN = 1<<0, 1664 IFF_EBRIDGE = 1<<1, 1665 IFF_BONDING = 1<<2, 1666 IFF_ISATAP = 1<<3, 1667 IFF_WAN_HDLC = 1<<4, 1668 IFF_XMIT_DST_RELEASE = 1<<5, 1669 IFF_DONT_BRIDGE = 1<<6, 1670 IFF_DISABLE_NETPOLL = 1<<7, 1671 IFF_MACVLAN_PORT = 1<<8, 1672 IFF_BRIDGE_PORT = 1<<9, 1673 IFF_OVS_DATAPATH = 1<<10, 1674 IFF_TX_SKB_SHARING = 1<<11, 1675 IFF_UNICAST_FLT = 1<<12, 1676 IFF_TEAM_PORT = 1<<13, 1677 IFF_SUPP_NOFCS = 1<<14, 1678 IFF_LIVE_ADDR_CHANGE = 1<<15, 1679 IFF_MACVLAN = 1<<16, 1680 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1681 IFF_L3MDEV_MASTER = 1<<18, 1682 IFF_NO_QUEUE = 1<<19, 1683 IFF_OPENVSWITCH = 1<<20, 1684 IFF_L3MDEV_SLAVE = 1<<21, 1685 IFF_TEAM = 1<<22, 1686 IFF_RXFH_CONFIGURED = 1<<23, 1687 IFF_PHONY_HEADROOM = 1<<24, 1688 IFF_MACSEC = 1<<25, 1689 IFF_NO_RX_HANDLER = 1<<26, 1690 IFF_FAILOVER = 1<<27, 1691 IFF_FAILOVER_SLAVE = 1<<28, 1692 IFF_L3MDEV_RX_HANDLER = 1<<29, 1693 IFF_LIVE_RENAME_OK = 1<<30, 1694 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31), 1695 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32), 1696 }; 1697 1698 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1699 #define IFF_EBRIDGE IFF_EBRIDGE 1700 #define IFF_BONDING IFF_BONDING 1701 #define IFF_ISATAP IFF_ISATAP 1702 #define IFF_WAN_HDLC IFF_WAN_HDLC 1703 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1704 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1705 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1706 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1707 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1708 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1709 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1710 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1711 #define IFF_TEAM_PORT IFF_TEAM_PORT 1712 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1713 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1714 #define IFF_MACVLAN IFF_MACVLAN 1715 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1716 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1717 #define IFF_NO_QUEUE IFF_NO_QUEUE 1718 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1719 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1720 #define IFF_TEAM IFF_TEAM 1721 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1722 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM 1723 #define IFF_MACSEC IFF_MACSEC 1724 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1725 #define IFF_FAILOVER IFF_FAILOVER 1726 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1727 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1728 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1729 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR 1730 1731 /* Specifies the type of the struct net_device::ml_priv pointer */ 1732 enum netdev_ml_priv_type { 1733 ML_PRIV_NONE, 1734 ML_PRIV_CAN, 1735 }; 1736 1737 /** 1738 * struct net_device - The DEVICE structure. 1739 * 1740 * Actually, this whole structure is a big mistake. It mixes I/O 1741 * data with strictly "high-level" data, and it has to know about 1742 * almost every data structure used in the INET module. 1743 * 1744 * @name: This is the first field of the "visible" part of this structure 1745 * (i.e. as seen by users in the "Space.c" file). It is the name 1746 * of the interface. 1747 * 1748 * @name_node: Name hashlist node 1749 * @ifalias: SNMP alias 1750 * @mem_end: Shared memory end 1751 * @mem_start: Shared memory start 1752 * @base_addr: Device I/O address 1753 * @irq: Device IRQ number 1754 * 1755 * @state: Generic network queuing layer state, see netdev_state_t 1756 * @dev_list: The global list of network devices 1757 * @napi_list: List entry used for polling NAPI devices 1758 * @unreg_list: List entry when we are unregistering the 1759 * device; see the function unregister_netdev 1760 * @close_list: List entry used when we are closing the device 1761 * @ptype_all: Device-specific packet handlers for all protocols 1762 * @ptype_specific: Device-specific, protocol-specific packet handlers 1763 * 1764 * @adj_list: Directly linked devices, like slaves for bonding 1765 * @features: Currently active device features 1766 * @hw_features: User-changeable features 1767 * 1768 * @wanted_features: User-requested features 1769 * @vlan_features: Mask of features inheritable by VLAN devices 1770 * 1771 * @hw_enc_features: Mask of features inherited by encapsulating devices 1772 * This field indicates what encapsulation 1773 * offloads the hardware is capable of doing, 1774 * and drivers will need to set them appropriately. 1775 * 1776 * @mpls_features: Mask of features inheritable by MPLS 1777 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1778 * 1779 * @ifindex: interface index 1780 * @group: The group the device belongs to 1781 * 1782 * @stats: Statistics struct, which was left as a legacy, use 1783 * rtnl_link_stats64 instead 1784 * 1785 * @core_stats: core networking counters, 1786 * do not use this in drivers 1787 * @carrier_up_count: Number of times the carrier has been up 1788 * @carrier_down_count: Number of times the carrier has been down 1789 * 1790 * @wireless_handlers: List of functions to handle Wireless Extensions, 1791 * instead of ioctl, 1792 * see <net/iw_handler.h> for details. 1793 * @wireless_data: Instance data managed by the core of wireless extensions 1794 * 1795 * @netdev_ops: Includes several pointers to callbacks, 1796 * if one wants to override the ndo_*() functions 1797 * @ethtool_ops: Management operations 1798 * @l3mdev_ops: Layer 3 master device operations 1799 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1800 * discovery handling. Necessary for e.g. 6LoWPAN. 1801 * @xfrmdev_ops: Transformation offload operations 1802 * @tlsdev_ops: Transport Layer Security offload operations 1803 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1804 * of Layer 2 headers. 1805 * 1806 * @flags: Interface flags (a la BSD) 1807 * @priv_flags: Like 'flags' but invisible to userspace, 1808 * see if.h for the definitions 1809 * @gflags: Global flags ( kept as legacy ) 1810 * @padded: How much padding added by alloc_netdev() 1811 * @operstate: RFC2863 operstate 1812 * @link_mode: Mapping policy to operstate 1813 * @if_port: Selectable AUI, TP, ... 1814 * @dma: DMA channel 1815 * @mtu: Interface MTU value 1816 * @min_mtu: Interface Minimum MTU value 1817 * @max_mtu: Interface Maximum MTU value 1818 * @type: Interface hardware type 1819 * @hard_header_len: Maximum hardware header length. 1820 * @min_header_len: Minimum hardware header length 1821 * 1822 * @needed_headroom: Extra headroom the hardware may need, but not in all 1823 * cases can this be guaranteed 1824 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1825 * cases can this be guaranteed. Some cases also use 1826 * LL_MAX_HEADER instead to allocate the skb 1827 * 1828 * interface address info: 1829 * 1830 * @perm_addr: Permanent hw address 1831 * @addr_assign_type: Hw address assignment type 1832 * @addr_len: Hardware address length 1833 * @upper_level: Maximum depth level of upper devices. 1834 * @lower_level: Maximum depth level of lower devices. 1835 * @neigh_priv_len: Used in neigh_alloc() 1836 * @dev_id: Used to differentiate devices that share 1837 * the same link layer address 1838 * @dev_port: Used to differentiate devices that share 1839 * the same function 1840 * @addr_list_lock: XXX: need comments on this one 1841 * @name_assign_type: network interface name assignment type 1842 * @uc_promisc: Counter that indicates promiscuous mode 1843 * has been enabled due to the need to listen to 1844 * additional unicast addresses in a device that 1845 * does not implement ndo_set_rx_mode() 1846 * @uc: unicast mac addresses 1847 * @mc: multicast mac addresses 1848 * @dev_addrs: list of device hw addresses 1849 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1850 * @promiscuity: Number of times the NIC is told to work in 1851 * promiscuous mode; if it becomes 0 the NIC will 1852 * exit promiscuous mode 1853 * @allmulti: Counter, enables or disables allmulticast mode 1854 * 1855 * @vlan_info: VLAN info 1856 * @dsa_ptr: dsa specific data 1857 * @tipc_ptr: TIPC specific data 1858 * @atalk_ptr: AppleTalk link 1859 * @ip_ptr: IPv4 specific data 1860 * @ip6_ptr: IPv6 specific data 1861 * @ax25_ptr: AX.25 specific data 1862 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1863 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1864 * device struct 1865 * @mpls_ptr: mpls_dev struct pointer 1866 * @mctp_ptr: MCTP specific data 1867 * 1868 * @dev_addr: Hw address (before bcast, 1869 * because most packets are unicast) 1870 * 1871 * @_rx: Array of RX queues 1872 * @num_rx_queues: Number of RX queues 1873 * allocated at register_netdev() time 1874 * @real_num_rx_queues: Number of RX queues currently active in device 1875 * @xdp_prog: XDP sockets filter program pointer 1876 * @gro_flush_timeout: timeout for GRO layer in NAPI 1877 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1878 * allow to avoid NIC hard IRQ, on busy queues. 1879 * 1880 * @rx_handler: handler for received packets 1881 * @rx_handler_data: XXX: need comments on this one 1882 * @miniq_ingress: ingress/clsact qdisc specific data for 1883 * ingress processing 1884 * @ingress_queue: XXX: need comments on this one 1885 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1886 * @broadcast: hw bcast address 1887 * 1888 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1889 * indexed by RX queue number. Assigned by driver. 1890 * This must only be set if the ndo_rx_flow_steer 1891 * operation is defined 1892 * @index_hlist: Device index hash chain 1893 * 1894 * @_tx: Array of TX queues 1895 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1896 * @real_num_tx_queues: Number of TX queues currently active in device 1897 * @qdisc: Root qdisc from userspace point of view 1898 * @tx_queue_len: Max frames per queue allowed 1899 * @tx_global_lock: XXX: need comments on this one 1900 * @xdp_bulkq: XDP device bulk queue 1901 * @xps_maps: all CPUs/RXQs maps for XPS device 1902 * 1903 * @xps_maps: XXX: need comments on this one 1904 * @miniq_egress: clsact qdisc specific data for 1905 * egress processing 1906 * @nf_hooks_egress: netfilter hooks executed for egress packets 1907 * @qdisc_hash: qdisc hash table 1908 * @watchdog_timeo: Represents the timeout that is used by 1909 * the watchdog (see dev_watchdog()) 1910 * @watchdog_timer: List of timers 1911 * 1912 * @proto_down_reason: reason a netdev interface is held down 1913 * @pcpu_refcnt: Number of references to this device 1914 * @dev_refcnt: Number of references to this device 1915 * @refcnt_tracker: Tracker directory for tracked references to this device 1916 * @todo_list: Delayed register/unregister 1917 * @link_watch_list: XXX: need comments on this one 1918 * 1919 * @reg_state: Register/unregister state machine 1920 * @dismantle: Device is going to be freed 1921 * @rtnl_link_state: This enum represents the phases of creating 1922 * a new link 1923 * 1924 * @needs_free_netdev: Should unregister perform free_netdev? 1925 * @priv_destructor: Called from unregister 1926 * @npinfo: XXX: need comments on this one 1927 * @nd_net: Network namespace this network device is inside 1928 * 1929 * @ml_priv: Mid-layer private 1930 * @ml_priv_type: Mid-layer private type 1931 * @lstats: Loopback statistics 1932 * @tstats: Tunnel statistics 1933 * @dstats: Dummy statistics 1934 * @vstats: Virtual ethernet statistics 1935 * 1936 * @garp_port: GARP 1937 * @mrp_port: MRP 1938 * 1939 * @dm_private: Drop monitor private 1940 * 1941 * @dev: Class/net/name entry 1942 * @sysfs_groups: Space for optional device, statistics and wireless 1943 * sysfs groups 1944 * 1945 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1946 * @rtnl_link_ops: Rtnl_link_ops 1947 * 1948 * @gso_max_size: Maximum size of generic segmentation offload 1949 * @tso_max_size: Device (as in HW) limit on the max TSO request size 1950 * @gso_max_segs: Maximum number of segments that can be passed to the 1951 * NIC for GSO 1952 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count 1953 * 1954 * @dcbnl_ops: Data Center Bridging netlink ops 1955 * @num_tc: Number of traffic classes in the net device 1956 * @tc_to_txq: XXX: need comments on this one 1957 * @prio_tc_map: XXX: need comments on this one 1958 * 1959 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1960 * 1961 * @priomap: XXX: need comments on this one 1962 * @phydev: Physical device may attach itself 1963 * for hardware timestamping 1964 * @sfp_bus: attached &struct sfp_bus structure. 1965 * 1966 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1967 * 1968 * @proto_down: protocol port state information can be sent to the 1969 * switch driver and used to set the phys state of the 1970 * switch port. 1971 * 1972 * @wol_enabled: Wake-on-LAN is enabled 1973 * 1974 * @threaded: napi threaded mode is enabled 1975 * 1976 * @net_notifier_list: List of per-net netdev notifier block 1977 * that follow this device when it is moved 1978 * to another network namespace. 1979 * 1980 * @macsec_ops: MACsec offloading ops 1981 * 1982 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1983 * offload capabilities of the device 1984 * @udp_tunnel_nic: UDP tunnel offload state 1985 * @xdp_state: stores info on attached XDP BPF programs 1986 * 1987 * @nested_level: Used as a parameter of spin_lock_nested() of 1988 * dev->addr_list_lock. 1989 * @unlink_list: As netif_addr_lock() can be called recursively, 1990 * keep a list of interfaces to be deleted. 1991 * @gro_max_size: Maximum size of aggregated packet in generic 1992 * receive offload (GRO) 1993 * 1994 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 1995 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 1996 * @watchdog_dev_tracker: refcount tracker used by watchdog. 1997 * @dev_registered_tracker: tracker for reference held while 1998 * registered 1999 * @offload_xstats_l3: L3 HW stats for this netdevice. 2000 * 2001 * FIXME: cleanup struct net_device such that network protocol info 2002 * moves out. 2003 */ 2004 2005 struct net_device { 2006 char name[IFNAMSIZ]; 2007 struct netdev_name_node *name_node; 2008 struct dev_ifalias __rcu *ifalias; 2009 /* 2010 * I/O specific fields 2011 * FIXME: Merge these and struct ifmap into one 2012 */ 2013 unsigned long mem_end; 2014 unsigned long mem_start; 2015 unsigned long base_addr; 2016 2017 /* 2018 * Some hardware also needs these fields (state,dev_list, 2019 * napi_list,unreg_list,close_list) but they are not 2020 * part of the usual set specified in Space.c. 2021 */ 2022 2023 unsigned long state; 2024 2025 struct list_head dev_list; 2026 struct list_head napi_list; 2027 struct list_head unreg_list; 2028 struct list_head close_list; 2029 struct list_head ptype_all; 2030 struct list_head ptype_specific; 2031 2032 struct { 2033 struct list_head upper; 2034 struct list_head lower; 2035 } adj_list; 2036 2037 /* Read-mostly cache-line for fast-path access */ 2038 unsigned int flags; 2039 unsigned long long priv_flags; 2040 const struct net_device_ops *netdev_ops; 2041 int ifindex; 2042 unsigned short gflags; 2043 unsigned short hard_header_len; 2044 2045 /* Note : dev->mtu is often read without holding a lock. 2046 * Writers usually hold RTNL. 2047 * It is recommended to use READ_ONCE() to annotate the reads, 2048 * and to use WRITE_ONCE() to annotate the writes. 2049 */ 2050 unsigned int mtu; 2051 unsigned short needed_headroom; 2052 unsigned short needed_tailroom; 2053 2054 netdev_features_t features; 2055 netdev_features_t hw_features; 2056 netdev_features_t wanted_features; 2057 netdev_features_t vlan_features; 2058 netdev_features_t hw_enc_features; 2059 netdev_features_t mpls_features; 2060 netdev_features_t gso_partial_features; 2061 2062 unsigned int min_mtu; 2063 unsigned int max_mtu; 2064 unsigned short type; 2065 unsigned char min_header_len; 2066 unsigned char name_assign_type; 2067 2068 int group; 2069 2070 struct net_device_stats stats; /* not used by modern drivers */ 2071 2072 struct net_device_core_stats __percpu *core_stats; 2073 2074 /* Stats to monitor link on/off, flapping */ 2075 atomic_t carrier_up_count; 2076 atomic_t carrier_down_count; 2077 2078 #ifdef CONFIG_WIRELESS_EXT 2079 const struct iw_handler_def *wireless_handlers; 2080 struct iw_public_data *wireless_data; 2081 #endif 2082 const struct ethtool_ops *ethtool_ops; 2083 #ifdef CONFIG_NET_L3_MASTER_DEV 2084 const struct l3mdev_ops *l3mdev_ops; 2085 #endif 2086 #if IS_ENABLED(CONFIG_IPV6) 2087 const struct ndisc_ops *ndisc_ops; 2088 #endif 2089 2090 #ifdef CONFIG_XFRM_OFFLOAD 2091 const struct xfrmdev_ops *xfrmdev_ops; 2092 #endif 2093 2094 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2095 const struct tlsdev_ops *tlsdev_ops; 2096 #endif 2097 2098 const struct header_ops *header_ops; 2099 2100 unsigned char operstate; 2101 unsigned char link_mode; 2102 2103 unsigned char if_port; 2104 unsigned char dma; 2105 2106 /* Interface address info. */ 2107 unsigned char perm_addr[MAX_ADDR_LEN]; 2108 unsigned char addr_assign_type; 2109 unsigned char addr_len; 2110 unsigned char upper_level; 2111 unsigned char lower_level; 2112 2113 unsigned short neigh_priv_len; 2114 unsigned short dev_id; 2115 unsigned short dev_port; 2116 unsigned short padded; 2117 2118 spinlock_t addr_list_lock; 2119 int irq; 2120 2121 struct netdev_hw_addr_list uc; 2122 struct netdev_hw_addr_list mc; 2123 struct netdev_hw_addr_list dev_addrs; 2124 2125 #ifdef CONFIG_SYSFS 2126 struct kset *queues_kset; 2127 #endif 2128 #ifdef CONFIG_LOCKDEP 2129 struct list_head unlink_list; 2130 #endif 2131 unsigned int promiscuity; 2132 unsigned int allmulti; 2133 bool uc_promisc; 2134 #ifdef CONFIG_LOCKDEP 2135 unsigned char nested_level; 2136 #endif 2137 2138 2139 /* Protocol-specific pointers */ 2140 2141 struct in_device __rcu *ip_ptr; 2142 struct inet6_dev __rcu *ip6_ptr; 2143 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2144 struct vlan_info __rcu *vlan_info; 2145 #endif 2146 #if IS_ENABLED(CONFIG_NET_DSA) 2147 struct dsa_port *dsa_ptr; 2148 #endif 2149 #if IS_ENABLED(CONFIG_TIPC) 2150 struct tipc_bearer __rcu *tipc_ptr; 2151 #endif 2152 #if IS_ENABLED(CONFIG_ATALK) 2153 void *atalk_ptr; 2154 #endif 2155 #if IS_ENABLED(CONFIG_AX25) 2156 void *ax25_ptr; 2157 #endif 2158 #if IS_ENABLED(CONFIG_CFG80211) 2159 struct wireless_dev *ieee80211_ptr; 2160 #endif 2161 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2162 struct wpan_dev *ieee802154_ptr; 2163 #endif 2164 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2165 struct mpls_dev __rcu *mpls_ptr; 2166 #endif 2167 #if IS_ENABLED(CONFIG_MCTP) 2168 struct mctp_dev __rcu *mctp_ptr; 2169 #endif 2170 2171 /* 2172 * Cache lines mostly used on receive path (including eth_type_trans()) 2173 */ 2174 /* Interface address info used in eth_type_trans() */ 2175 const unsigned char *dev_addr; 2176 2177 struct netdev_rx_queue *_rx; 2178 unsigned int num_rx_queues; 2179 unsigned int real_num_rx_queues; 2180 2181 struct bpf_prog __rcu *xdp_prog; 2182 unsigned long gro_flush_timeout; 2183 int napi_defer_hard_irqs; 2184 #define GRO_LEGACY_MAX_SIZE 65536u 2185 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2186 * and shinfo->gso_segs is a 16bit field. 2187 */ 2188 #define GRO_MAX_SIZE (8 * 65535u) 2189 unsigned int gro_max_size; 2190 rx_handler_func_t __rcu *rx_handler; 2191 void __rcu *rx_handler_data; 2192 2193 #ifdef CONFIG_NET_CLS_ACT 2194 struct mini_Qdisc __rcu *miniq_ingress; 2195 #endif 2196 struct netdev_queue __rcu *ingress_queue; 2197 #ifdef CONFIG_NETFILTER_INGRESS 2198 struct nf_hook_entries __rcu *nf_hooks_ingress; 2199 #endif 2200 2201 unsigned char broadcast[MAX_ADDR_LEN]; 2202 #ifdef CONFIG_RFS_ACCEL 2203 struct cpu_rmap *rx_cpu_rmap; 2204 #endif 2205 struct hlist_node index_hlist; 2206 2207 /* 2208 * Cache lines mostly used on transmit path 2209 */ 2210 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2211 unsigned int num_tx_queues; 2212 unsigned int real_num_tx_queues; 2213 struct Qdisc __rcu *qdisc; 2214 unsigned int tx_queue_len; 2215 spinlock_t tx_global_lock; 2216 2217 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2218 2219 #ifdef CONFIG_XPS 2220 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2221 #endif 2222 #ifdef CONFIG_NET_CLS_ACT 2223 struct mini_Qdisc __rcu *miniq_egress; 2224 #endif 2225 #ifdef CONFIG_NETFILTER_EGRESS 2226 struct nf_hook_entries __rcu *nf_hooks_egress; 2227 #endif 2228 2229 #ifdef CONFIG_NET_SCHED 2230 DECLARE_HASHTABLE (qdisc_hash, 4); 2231 #endif 2232 /* These may be needed for future network-power-down code. */ 2233 struct timer_list watchdog_timer; 2234 int watchdog_timeo; 2235 2236 u32 proto_down_reason; 2237 2238 struct list_head todo_list; 2239 2240 #ifdef CONFIG_PCPU_DEV_REFCNT 2241 int __percpu *pcpu_refcnt; 2242 #else 2243 refcount_t dev_refcnt; 2244 #endif 2245 struct ref_tracker_dir refcnt_tracker; 2246 2247 struct list_head link_watch_list; 2248 2249 enum { NETREG_UNINITIALIZED=0, 2250 NETREG_REGISTERED, /* completed register_netdevice */ 2251 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2252 NETREG_UNREGISTERED, /* completed unregister todo */ 2253 NETREG_RELEASED, /* called free_netdev */ 2254 NETREG_DUMMY, /* dummy device for NAPI poll */ 2255 } reg_state:8; 2256 2257 bool dismantle; 2258 2259 enum { 2260 RTNL_LINK_INITIALIZED, 2261 RTNL_LINK_INITIALIZING, 2262 } rtnl_link_state:16; 2263 2264 bool needs_free_netdev; 2265 void (*priv_destructor)(struct net_device *dev); 2266 2267 #ifdef CONFIG_NETPOLL 2268 struct netpoll_info __rcu *npinfo; 2269 #endif 2270 2271 possible_net_t nd_net; 2272 2273 /* mid-layer private */ 2274 void *ml_priv; 2275 enum netdev_ml_priv_type ml_priv_type; 2276 2277 union { 2278 struct pcpu_lstats __percpu *lstats; 2279 struct pcpu_sw_netstats __percpu *tstats; 2280 struct pcpu_dstats __percpu *dstats; 2281 }; 2282 2283 #if IS_ENABLED(CONFIG_GARP) 2284 struct garp_port __rcu *garp_port; 2285 #endif 2286 #if IS_ENABLED(CONFIG_MRP) 2287 struct mrp_port __rcu *mrp_port; 2288 #endif 2289 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2290 struct dm_hw_stat_delta __rcu *dm_private; 2291 #endif 2292 struct device dev; 2293 const struct attribute_group *sysfs_groups[4]; 2294 const struct attribute_group *sysfs_rx_queue_group; 2295 2296 const struct rtnl_link_ops *rtnl_link_ops; 2297 2298 /* for setting kernel sock attribute on TCP connection setup */ 2299 #define GSO_MAX_SEGS 65535u 2300 #define GSO_LEGACY_MAX_SIZE 65536u 2301 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2302 * and shinfo->gso_segs is a 16bit field. 2303 */ 2304 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2305 2306 unsigned int gso_max_size; 2307 #define TSO_LEGACY_MAX_SIZE 65536 2308 #define TSO_MAX_SIZE UINT_MAX 2309 unsigned int tso_max_size; 2310 u16 gso_max_segs; 2311 #define TSO_MAX_SEGS U16_MAX 2312 u16 tso_max_segs; 2313 2314 #ifdef CONFIG_DCB 2315 const struct dcbnl_rtnl_ops *dcbnl_ops; 2316 #endif 2317 s16 num_tc; 2318 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2319 u8 prio_tc_map[TC_BITMASK + 1]; 2320 2321 #if IS_ENABLED(CONFIG_FCOE) 2322 unsigned int fcoe_ddp_xid; 2323 #endif 2324 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2325 struct netprio_map __rcu *priomap; 2326 #endif 2327 struct phy_device *phydev; 2328 struct sfp_bus *sfp_bus; 2329 struct lock_class_key *qdisc_tx_busylock; 2330 bool proto_down; 2331 unsigned wol_enabled:1; 2332 unsigned threaded:1; 2333 2334 struct list_head net_notifier_list; 2335 2336 #if IS_ENABLED(CONFIG_MACSEC) 2337 /* MACsec management functions */ 2338 const struct macsec_ops *macsec_ops; 2339 #endif 2340 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2341 struct udp_tunnel_nic *udp_tunnel_nic; 2342 2343 /* protected by rtnl_lock */ 2344 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2345 2346 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2347 netdevice_tracker linkwatch_dev_tracker; 2348 netdevice_tracker watchdog_dev_tracker; 2349 netdevice_tracker dev_registered_tracker; 2350 struct rtnl_hw_stats64 *offload_xstats_l3; 2351 }; 2352 #define to_net_dev(d) container_of(d, struct net_device, dev) 2353 2354 static inline bool netif_elide_gro(const struct net_device *dev) 2355 { 2356 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2357 return true; 2358 return false; 2359 } 2360 2361 #define NETDEV_ALIGN 32 2362 2363 static inline 2364 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2365 { 2366 return dev->prio_tc_map[prio & TC_BITMASK]; 2367 } 2368 2369 static inline 2370 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2371 { 2372 if (tc >= dev->num_tc) 2373 return -EINVAL; 2374 2375 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2376 return 0; 2377 } 2378 2379 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2380 void netdev_reset_tc(struct net_device *dev); 2381 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2382 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2383 2384 static inline 2385 int netdev_get_num_tc(struct net_device *dev) 2386 { 2387 return dev->num_tc; 2388 } 2389 2390 static inline void net_prefetch(void *p) 2391 { 2392 prefetch(p); 2393 #if L1_CACHE_BYTES < 128 2394 prefetch((u8 *)p + L1_CACHE_BYTES); 2395 #endif 2396 } 2397 2398 static inline void net_prefetchw(void *p) 2399 { 2400 prefetchw(p); 2401 #if L1_CACHE_BYTES < 128 2402 prefetchw((u8 *)p + L1_CACHE_BYTES); 2403 #endif 2404 } 2405 2406 void netdev_unbind_sb_channel(struct net_device *dev, 2407 struct net_device *sb_dev); 2408 int netdev_bind_sb_channel_queue(struct net_device *dev, 2409 struct net_device *sb_dev, 2410 u8 tc, u16 count, u16 offset); 2411 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2412 static inline int netdev_get_sb_channel(struct net_device *dev) 2413 { 2414 return max_t(int, -dev->num_tc, 0); 2415 } 2416 2417 static inline 2418 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2419 unsigned int index) 2420 { 2421 return &dev->_tx[index]; 2422 } 2423 2424 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2425 const struct sk_buff *skb) 2426 { 2427 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2428 } 2429 2430 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2431 void (*f)(struct net_device *, 2432 struct netdev_queue *, 2433 void *), 2434 void *arg) 2435 { 2436 unsigned int i; 2437 2438 for (i = 0; i < dev->num_tx_queues; i++) 2439 f(dev, &dev->_tx[i], arg); 2440 } 2441 2442 #define netdev_lockdep_set_classes(dev) \ 2443 { \ 2444 static struct lock_class_key qdisc_tx_busylock_key; \ 2445 static struct lock_class_key qdisc_xmit_lock_key; \ 2446 static struct lock_class_key dev_addr_list_lock_key; \ 2447 unsigned int i; \ 2448 \ 2449 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2450 lockdep_set_class(&(dev)->addr_list_lock, \ 2451 &dev_addr_list_lock_key); \ 2452 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2453 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2454 &qdisc_xmit_lock_key); \ 2455 } 2456 2457 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2458 struct net_device *sb_dev); 2459 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2460 struct sk_buff *skb, 2461 struct net_device *sb_dev); 2462 2463 /* returns the headroom that the master device needs to take in account 2464 * when forwarding to this dev 2465 */ 2466 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2467 { 2468 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2469 } 2470 2471 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2472 { 2473 if (dev->netdev_ops->ndo_set_rx_headroom) 2474 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2475 } 2476 2477 /* set the device rx headroom to the dev's default */ 2478 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2479 { 2480 netdev_set_rx_headroom(dev, -1); 2481 } 2482 2483 static inline void *netdev_get_ml_priv(struct net_device *dev, 2484 enum netdev_ml_priv_type type) 2485 { 2486 if (dev->ml_priv_type != type) 2487 return NULL; 2488 2489 return dev->ml_priv; 2490 } 2491 2492 static inline void netdev_set_ml_priv(struct net_device *dev, 2493 void *ml_priv, 2494 enum netdev_ml_priv_type type) 2495 { 2496 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2497 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2498 dev->ml_priv_type, type); 2499 WARN(!dev->ml_priv_type && dev->ml_priv, 2500 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2501 2502 dev->ml_priv = ml_priv; 2503 dev->ml_priv_type = type; 2504 } 2505 2506 /* 2507 * Net namespace inlines 2508 */ 2509 static inline 2510 struct net *dev_net(const struct net_device *dev) 2511 { 2512 return read_pnet(&dev->nd_net); 2513 } 2514 2515 static inline 2516 void dev_net_set(struct net_device *dev, struct net *net) 2517 { 2518 write_pnet(&dev->nd_net, net); 2519 } 2520 2521 /** 2522 * netdev_priv - access network device private data 2523 * @dev: network device 2524 * 2525 * Get network device private data 2526 */ 2527 static inline void *netdev_priv(const struct net_device *dev) 2528 { 2529 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2530 } 2531 2532 /* Set the sysfs physical device reference for the network logical device 2533 * if set prior to registration will cause a symlink during initialization. 2534 */ 2535 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2536 2537 /* Set the sysfs device type for the network logical device to allow 2538 * fine-grained identification of different network device types. For 2539 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2540 */ 2541 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2542 2543 /* Default NAPI poll() weight 2544 * Device drivers are strongly advised to not use bigger value 2545 */ 2546 #define NAPI_POLL_WEIGHT 64 2547 2548 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2549 int (*poll)(struct napi_struct *, int), int weight); 2550 2551 /** 2552 * netif_napi_add() - initialize a NAPI context 2553 * @dev: network device 2554 * @napi: NAPI context 2555 * @poll: polling function 2556 * @weight: default weight 2557 * 2558 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2559 * *any* of the other NAPI-related functions. 2560 */ 2561 static inline void 2562 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2563 int (*poll)(struct napi_struct *, int), int weight) 2564 { 2565 netif_napi_add_weight(dev, napi, poll, weight); 2566 } 2567 2568 static inline void 2569 netif_napi_add_tx_weight(struct net_device *dev, 2570 struct napi_struct *napi, 2571 int (*poll)(struct napi_struct *, int), 2572 int weight) 2573 { 2574 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2575 netif_napi_add_weight(dev, napi, poll, weight); 2576 } 2577 2578 /** 2579 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2580 * @dev: network device 2581 * @napi: NAPI context 2582 * @poll: polling function 2583 * 2584 * This variant of netif_napi_add() should be used from drivers using NAPI 2585 * to exclusively poll a TX queue. 2586 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2587 */ 2588 static inline void netif_napi_add_tx(struct net_device *dev, 2589 struct napi_struct *napi, 2590 int (*poll)(struct napi_struct *, int)) 2591 { 2592 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2593 } 2594 2595 /** 2596 * __netif_napi_del - remove a NAPI context 2597 * @napi: NAPI context 2598 * 2599 * Warning: caller must observe RCU grace period before freeing memory 2600 * containing @napi. Drivers might want to call this helper to combine 2601 * all the needed RCU grace periods into a single one. 2602 */ 2603 void __netif_napi_del(struct napi_struct *napi); 2604 2605 /** 2606 * netif_napi_del - remove a NAPI context 2607 * @napi: NAPI context 2608 * 2609 * netif_napi_del() removes a NAPI context from the network device NAPI list 2610 */ 2611 static inline void netif_napi_del(struct napi_struct *napi) 2612 { 2613 __netif_napi_del(napi); 2614 synchronize_net(); 2615 } 2616 2617 struct packet_type { 2618 __be16 type; /* This is really htons(ether_type). */ 2619 bool ignore_outgoing; 2620 struct net_device *dev; /* NULL is wildcarded here */ 2621 netdevice_tracker dev_tracker; 2622 int (*func) (struct sk_buff *, 2623 struct net_device *, 2624 struct packet_type *, 2625 struct net_device *); 2626 void (*list_func) (struct list_head *, 2627 struct packet_type *, 2628 struct net_device *); 2629 bool (*id_match)(struct packet_type *ptype, 2630 struct sock *sk); 2631 struct net *af_packet_net; 2632 void *af_packet_priv; 2633 struct list_head list; 2634 }; 2635 2636 struct offload_callbacks { 2637 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2638 netdev_features_t features); 2639 struct sk_buff *(*gro_receive)(struct list_head *head, 2640 struct sk_buff *skb); 2641 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2642 }; 2643 2644 struct packet_offload { 2645 __be16 type; /* This is really htons(ether_type). */ 2646 u16 priority; 2647 struct offload_callbacks callbacks; 2648 struct list_head list; 2649 }; 2650 2651 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2652 struct pcpu_sw_netstats { 2653 u64_stats_t rx_packets; 2654 u64_stats_t rx_bytes; 2655 u64_stats_t tx_packets; 2656 u64_stats_t tx_bytes; 2657 struct u64_stats_sync syncp; 2658 } __aligned(4 * sizeof(u64)); 2659 2660 struct pcpu_lstats { 2661 u64_stats_t packets; 2662 u64_stats_t bytes; 2663 struct u64_stats_sync syncp; 2664 } __aligned(2 * sizeof(u64)); 2665 2666 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2667 2668 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2669 { 2670 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2671 2672 u64_stats_update_begin(&tstats->syncp); 2673 u64_stats_add(&tstats->rx_bytes, len); 2674 u64_stats_inc(&tstats->rx_packets); 2675 u64_stats_update_end(&tstats->syncp); 2676 } 2677 2678 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2679 unsigned int packets, 2680 unsigned int len) 2681 { 2682 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2683 2684 u64_stats_update_begin(&tstats->syncp); 2685 u64_stats_add(&tstats->tx_bytes, len); 2686 u64_stats_add(&tstats->tx_packets, packets); 2687 u64_stats_update_end(&tstats->syncp); 2688 } 2689 2690 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2691 { 2692 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2693 2694 u64_stats_update_begin(&lstats->syncp); 2695 u64_stats_add(&lstats->bytes, len); 2696 u64_stats_inc(&lstats->packets); 2697 u64_stats_update_end(&lstats->syncp); 2698 } 2699 2700 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2701 ({ \ 2702 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2703 if (pcpu_stats) { \ 2704 int __cpu; \ 2705 for_each_possible_cpu(__cpu) { \ 2706 typeof(type) *stat; \ 2707 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2708 u64_stats_init(&stat->syncp); \ 2709 } \ 2710 } \ 2711 pcpu_stats; \ 2712 }) 2713 2714 #define netdev_alloc_pcpu_stats(type) \ 2715 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2716 2717 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2718 ({ \ 2719 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2720 if (pcpu_stats) { \ 2721 int __cpu; \ 2722 for_each_possible_cpu(__cpu) { \ 2723 typeof(type) *stat; \ 2724 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2725 u64_stats_init(&stat->syncp); \ 2726 } \ 2727 } \ 2728 pcpu_stats; \ 2729 }) 2730 2731 enum netdev_lag_tx_type { 2732 NETDEV_LAG_TX_TYPE_UNKNOWN, 2733 NETDEV_LAG_TX_TYPE_RANDOM, 2734 NETDEV_LAG_TX_TYPE_BROADCAST, 2735 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2736 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2737 NETDEV_LAG_TX_TYPE_HASH, 2738 }; 2739 2740 enum netdev_lag_hash { 2741 NETDEV_LAG_HASH_NONE, 2742 NETDEV_LAG_HASH_L2, 2743 NETDEV_LAG_HASH_L34, 2744 NETDEV_LAG_HASH_L23, 2745 NETDEV_LAG_HASH_E23, 2746 NETDEV_LAG_HASH_E34, 2747 NETDEV_LAG_HASH_VLAN_SRCMAC, 2748 NETDEV_LAG_HASH_UNKNOWN, 2749 }; 2750 2751 struct netdev_lag_upper_info { 2752 enum netdev_lag_tx_type tx_type; 2753 enum netdev_lag_hash hash_type; 2754 }; 2755 2756 struct netdev_lag_lower_state_info { 2757 u8 link_up : 1, 2758 tx_enabled : 1; 2759 }; 2760 2761 #include <linux/notifier.h> 2762 2763 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2764 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2765 * adding new types. 2766 */ 2767 enum netdev_cmd { 2768 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2769 NETDEV_DOWN, 2770 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2771 detected a hardware crash and restarted 2772 - we can use this eg to kick tcp sessions 2773 once done */ 2774 NETDEV_CHANGE, /* Notify device state change */ 2775 NETDEV_REGISTER, 2776 NETDEV_UNREGISTER, 2777 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2778 NETDEV_CHANGEADDR, /* notify after the address change */ 2779 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2780 NETDEV_GOING_DOWN, 2781 NETDEV_CHANGENAME, 2782 NETDEV_FEAT_CHANGE, 2783 NETDEV_BONDING_FAILOVER, 2784 NETDEV_PRE_UP, 2785 NETDEV_PRE_TYPE_CHANGE, 2786 NETDEV_POST_TYPE_CHANGE, 2787 NETDEV_POST_INIT, 2788 NETDEV_RELEASE, 2789 NETDEV_NOTIFY_PEERS, 2790 NETDEV_JOIN, 2791 NETDEV_CHANGEUPPER, 2792 NETDEV_RESEND_IGMP, 2793 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2794 NETDEV_CHANGEINFODATA, 2795 NETDEV_BONDING_INFO, 2796 NETDEV_PRECHANGEUPPER, 2797 NETDEV_CHANGELOWERSTATE, 2798 NETDEV_UDP_TUNNEL_PUSH_INFO, 2799 NETDEV_UDP_TUNNEL_DROP_INFO, 2800 NETDEV_CHANGE_TX_QUEUE_LEN, 2801 NETDEV_CVLAN_FILTER_PUSH_INFO, 2802 NETDEV_CVLAN_FILTER_DROP_INFO, 2803 NETDEV_SVLAN_FILTER_PUSH_INFO, 2804 NETDEV_SVLAN_FILTER_DROP_INFO, 2805 NETDEV_OFFLOAD_XSTATS_ENABLE, 2806 NETDEV_OFFLOAD_XSTATS_DISABLE, 2807 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 2808 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 2809 }; 2810 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2811 2812 int register_netdevice_notifier(struct notifier_block *nb); 2813 int unregister_netdevice_notifier(struct notifier_block *nb); 2814 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2815 int unregister_netdevice_notifier_net(struct net *net, 2816 struct notifier_block *nb); 2817 int register_netdevice_notifier_dev_net(struct net_device *dev, 2818 struct notifier_block *nb, 2819 struct netdev_net_notifier *nn); 2820 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2821 struct notifier_block *nb, 2822 struct netdev_net_notifier *nn); 2823 2824 struct netdev_notifier_info { 2825 struct net_device *dev; 2826 struct netlink_ext_ack *extack; 2827 }; 2828 2829 struct netdev_notifier_info_ext { 2830 struct netdev_notifier_info info; /* must be first */ 2831 union { 2832 u32 mtu; 2833 } ext; 2834 }; 2835 2836 struct netdev_notifier_change_info { 2837 struct netdev_notifier_info info; /* must be first */ 2838 unsigned int flags_changed; 2839 }; 2840 2841 struct netdev_notifier_changeupper_info { 2842 struct netdev_notifier_info info; /* must be first */ 2843 struct net_device *upper_dev; /* new upper dev */ 2844 bool master; /* is upper dev master */ 2845 bool linking; /* is the notification for link or unlink */ 2846 void *upper_info; /* upper dev info */ 2847 }; 2848 2849 struct netdev_notifier_changelowerstate_info { 2850 struct netdev_notifier_info info; /* must be first */ 2851 void *lower_state_info; /* is lower dev state */ 2852 }; 2853 2854 struct netdev_notifier_pre_changeaddr_info { 2855 struct netdev_notifier_info info; /* must be first */ 2856 const unsigned char *dev_addr; 2857 }; 2858 2859 enum netdev_offload_xstats_type { 2860 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 2861 }; 2862 2863 struct netdev_notifier_offload_xstats_info { 2864 struct netdev_notifier_info info; /* must be first */ 2865 enum netdev_offload_xstats_type type; 2866 2867 union { 2868 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 2869 struct netdev_notifier_offload_xstats_rd *report_delta; 2870 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 2871 struct netdev_notifier_offload_xstats_ru *report_used; 2872 }; 2873 }; 2874 2875 int netdev_offload_xstats_enable(struct net_device *dev, 2876 enum netdev_offload_xstats_type type, 2877 struct netlink_ext_ack *extack); 2878 int netdev_offload_xstats_disable(struct net_device *dev, 2879 enum netdev_offload_xstats_type type); 2880 bool netdev_offload_xstats_enabled(const struct net_device *dev, 2881 enum netdev_offload_xstats_type type); 2882 int netdev_offload_xstats_get(struct net_device *dev, 2883 enum netdev_offload_xstats_type type, 2884 struct rtnl_hw_stats64 *stats, bool *used, 2885 struct netlink_ext_ack *extack); 2886 void 2887 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 2888 const struct rtnl_hw_stats64 *stats); 2889 void 2890 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 2891 void netdev_offload_xstats_push_delta(struct net_device *dev, 2892 enum netdev_offload_xstats_type type, 2893 const struct rtnl_hw_stats64 *stats); 2894 2895 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2896 struct net_device *dev) 2897 { 2898 info->dev = dev; 2899 info->extack = NULL; 2900 } 2901 2902 static inline struct net_device * 2903 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2904 { 2905 return info->dev; 2906 } 2907 2908 static inline struct netlink_ext_ack * 2909 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2910 { 2911 return info->extack; 2912 } 2913 2914 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2915 2916 2917 extern rwlock_t dev_base_lock; /* Device list lock */ 2918 2919 #define for_each_netdev(net, d) \ 2920 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2921 #define for_each_netdev_reverse(net, d) \ 2922 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2923 #define for_each_netdev_rcu(net, d) \ 2924 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2925 #define for_each_netdev_safe(net, d, n) \ 2926 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2927 #define for_each_netdev_continue(net, d) \ 2928 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2929 #define for_each_netdev_continue_reverse(net, d) \ 2930 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2931 dev_list) 2932 #define for_each_netdev_continue_rcu(net, d) \ 2933 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2934 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2935 for_each_netdev_rcu(&init_net, slave) \ 2936 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2937 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2938 2939 static inline struct net_device *next_net_device(struct net_device *dev) 2940 { 2941 struct list_head *lh; 2942 struct net *net; 2943 2944 net = dev_net(dev); 2945 lh = dev->dev_list.next; 2946 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2947 } 2948 2949 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2950 { 2951 struct list_head *lh; 2952 struct net *net; 2953 2954 net = dev_net(dev); 2955 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2956 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2957 } 2958 2959 static inline struct net_device *first_net_device(struct net *net) 2960 { 2961 return list_empty(&net->dev_base_head) ? NULL : 2962 net_device_entry(net->dev_base_head.next); 2963 } 2964 2965 static inline struct net_device *first_net_device_rcu(struct net *net) 2966 { 2967 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2968 2969 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2970 } 2971 2972 int netdev_boot_setup_check(struct net_device *dev); 2973 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2974 const char *hwaddr); 2975 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2976 void dev_add_pack(struct packet_type *pt); 2977 void dev_remove_pack(struct packet_type *pt); 2978 void __dev_remove_pack(struct packet_type *pt); 2979 void dev_add_offload(struct packet_offload *po); 2980 void dev_remove_offload(struct packet_offload *po); 2981 2982 int dev_get_iflink(const struct net_device *dev); 2983 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2984 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 2985 struct net_device_path_stack *stack); 2986 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2987 unsigned short mask); 2988 struct net_device *dev_get_by_name(struct net *net, const char *name); 2989 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2990 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2991 bool netdev_name_in_use(struct net *net, const char *name); 2992 int dev_alloc_name(struct net_device *dev, const char *name); 2993 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2994 void dev_close(struct net_device *dev); 2995 void dev_close_many(struct list_head *head, bool unlink); 2996 void dev_disable_lro(struct net_device *dev); 2997 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2998 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2999 struct net_device *sb_dev); 3000 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 3001 struct net_device *sb_dev); 3002 3003 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3004 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3005 3006 static inline int dev_queue_xmit(struct sk_buff *skb) 3007 { 3008 return __dev_queue_xmit(skb, NULL); 3009 } 3010 3011 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3012 struct net_device *sb_dev) 3013 { 3014 return __dev_queue_xmit(skb, sb_dev); 3015 } 3016 3017 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3018 { 3019 int ret; 3020 3021 ret = __dev_direct_xmit(skb, queue_id); 3022 if (!dev_xmit_complete(ret)) 3023 kfree_skb(skb); 3024 return ret; 3025 } 3026 3027 int register_netdevice(struct net_device *dev); 3028 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3029 void unregister_netdevice_many(struct list_head *head); 3030 static inline void unregister_netdevice(struct net_device *dev) 3031 { 3032 unregister_netdevice_queue(dev, NULL); 3033 } 3034 3035 int netdev_refcnt_read(const struct net_device *dev); 3036 void free_netdev(struct net_device *dev); 3037 void netdev_freemem(struct net_device *dev); 3038 int init_dummy_netdev(struct net_device *dev); 3039 3040 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3041 struct sk_buff *skb, 3042 bool all_slaves); 3043 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3044 struct sock *sk); 3045 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3046 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3047 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3048 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 3049 int dev_restart(struct net_device *dev); 3050 3051 3052 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3053 unsigned short type, 3054 const void *daddr, const void *saddr, 3055 unsigned int len) 3056 { 3057 if (!dev->header_ops || !dev->header_ops->create) 3058 return 0; 3059 3060 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3061 } 3062 3063 static inline int dev_parse_header(const struct sk_buff *skb, 3064 unsigned char *haddr) 3065 { 3066 const struct net_device *dev = skb->dev; 3067 3068 if (!dev->header_ops || !dev->header_ops->parse) 3069 return 0; 3070 return dev->header_ops->parse(skb, haddr); 3071 } 3072 3073 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3074 { 3075 const struct net_device *dev = skb->dev; 3076 3077 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3078 return 0; 3079 return dev->header_ops->parse_protocol(skb); 3080 } 3081 3082 /* ll_header must have at least hard_header_len allocated */ 3083 static inline bool dev_validate_header(const struct net_device *dev, 3084 char *ll_header, int len) 3085 { 3086 if (likely(len >= dev->hard_header_len)) 3087 return true; 3088 if (len < dev->min_header_len) 3089 return false; 3090 3091 if (capable(CAP_SYS_RAWIO)) { 3092 memset(ll_header + len, 0, dev->hard_header_len - len); 3093 return true; 3094 } 3095 3096 if (dev->header_ops && dev->header_ops->validate) 3097 return dev->header_ops->validate(ll_header, len); 3098 3099 return false; 3100 } 3101 3102 static inline bool dev_has_header(const struct net_device *dev) 3103 { 3104 return dev->header_ops && dev->header_ops->create; 3105 } 3106 3107 /* 3108 * Incoming packets are placed on per-CPU queues 3109 */ 3110 struct softnet_data { 3111 struct list_head poll_list; 3112 struct sk_buff_head process_queue; 3113 3114 /* stats */ 3115 unsigned int processed; 3116 unsigned int time_squeeze; 3117 unsigned int received_rps; 3118 #ifdef CONFIG_RPS 3119 struct softnet_data *rps_ipi_list; 3120 #endif 3121 #ifdef CONFIG_NET_FLOW_LIMIT 3122 struct sd_flow_limit __rcu *flow_limit; 3123 #endif 3124 struct Qdisc *output_queue; 3125 struct Qdisc **output_queue_tailp; 3126 struct sk_buff *completion_queue; 3127 #ifdef CONFIG_XFRM_OFFLOAD 3128 struct sk_buff_head xfrm_backlog; 3129 #endif 3130 /* written and read only by owning cpu: */ 3131 struct { 3132 u16 recursion; 3133 u8 more; 3134 #ifdef CONFIG_NET_EGRESS 3135 u8 skip_txqueue; 3136 #endif 3137 } xmit; 3138 #ifdef CONFIG_RPS 3139 /* input_queue_head should be written by cpu owning this struct, 3140 * and only read by other cpus. Worth using a cache line. 3141 */ 3142 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3143 3144 /* Elements below can be accessed between CPUs for RPS/RFS */ 3145 call_single_data_t csd ____cacheline_aligned_in_smp; 3146 struct softnet_data *rps_ipi_next; 3147 unsigned int cpu; 3148 unsigned int input_queue_tail; 3149 #endif 3150 unsigned int dropped; 3151 struct sk_buff_head input_pkt_queue; 3152 struct napi_struct backlog; 3153 3154 /* Another possibly contended cache line */ 3155 spinlock_t defer_lock ____cacheline_aligned_in_smp; 3156 int defer_count; 3157 int defer_ipi_scheduled; 3158 struct sk_buff *defer_list; 3159 call_single_data_t defer_csd; 3160 }; 3161 3162 static inline void input_queue_head_incr(struct softnet_data *sd) 3163 { 3164 #ifdef CONFIG_RPS 3165 sd->input_queue_head++; 3166 #endif 3167 } 3168 3169 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3170 unsigned int *qtail) 3171 { 3172 #ifdef CONFIG_RPS 3173 *qtail = ++sd->input_queue_tail; 3174 #endif 3175 } 3176 3177 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3178 3179 static inline int dev_recursion_level(void) 3180 { 3181 return this_cpu_read(softnet_data.xmit.recursion); 3182 } 3183 3184 #define XMIT_RECURSION_LIMIT 8 3185 static inline bool dev_xmit_recursion(void) 3186 { 3187 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3188 XMIT_RECURSION_LIMIT); 3189 } 3190 3191 static inline void dev_xmit_recursion_inc(void) 3192 { 3193 __this_cpu_inc(softnet_data.xmit.recursion); 3194 } 3195 3196 static inline void dev_xmit_recursion_dec(void) 3197 { 3198 __this_cpu_dec(softnet_data.xmit.recursion); 3199 } 3200 3201 void __netif_schedule(struct Qdisc *q); 3202 void netif_schedule_queue(struct netdev_queue *txq); 3203 3204 static inline void netif_tx_schedule_all(struct net_device *dev) 3205 { 3206 unsigned int i; 3207 3208 for (i = 0; i < dev->num_tx_queues; i++) 3209 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3210 } 3211 3212 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3213 { 3214 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3215 } 3216 3217 /** 3218 * netif_start_queue - allow transmit 3219 * @dev: network device 3220 * 3221 * Allow upper layers to call the device hard_start_xmit routine. 3222 */ 3223 static inline void netif_start_queue(struct net_device *dev) 3224 { 3225 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3226 } 3227 3228 static inline void netif_tx_start_all_queues(struct net_device *dev) 3229 { 3230 unsigned int i; 3231 3232 for (i = 0; i < dev->num_tx_queues; i++) { 3233 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3234 netif_tx_start_queue(txq); 3235 } 3236 } 3237 3238 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3239 3240 /** 3241 * netif_wake_queue - restart transmit 3242 * @dev: network device 3243 * 3244 * Allow upper layers to call the device hard_start_xmit routine. 3245 * Used for flow control when transmit resources are available. 3246 */ 3247 static inline void netif_wake_queue(struct net_device *dev) 3248 { 3249 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3250 } 3251 3252 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3253 { 3254 unsigned int i; 3255 3256 for (i = 0; i < dev->num_tx_queues; i++) { 3257 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3258 netif_tx_wake_queue(txq); 3259 } 3260 } 3261 3262 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3263 { 3264 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3265 } 3266 3267 /** 3268 * netif_stop_queue - stop transmitted packets 3269 * @dev: network device 3270 * 3271 * Stop upper layers calling the device hard_start_xmit routine. 3272 * Used for flow control when transmit resources are unavailable. 3273 */ 3274 static inline void netif_stop_queue(struct net_device *dev) 3275 { 3276 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3277 } 3278 3279 void netif_tx_stop_all_queues(struct net_device *dev); 3280 3281 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3282 { 3283 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3284 } 3285 3286 /** 3287 * netif_queue_stopped - test if transmit queue is flowblocked 3288 * @dev: network device 3289 * 3290 * Test if transmit queue on device is currently unable to send. 3291 */ 3292 static inline bool netif_queue_stopped(const struct net_device *dev) 3293 { 3294 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3295 } 3296 3297 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3298 { 3299 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3300 } 3301 3302 static inline bool 3303 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3304 { 3305 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3306 } 3307 3308 static inline bool 3309 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3310 { 3311 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3312 } 3313 3314 /** 3315 * netdev_queue_set_dql_min_limit - set dql minimum limit 3316 * @dev_queue: pointer to transmit queue 3317 * @min_limit: dql minimum limit 3318 * 3319 * Forces xmit_more() to return true until the minimum threshold 3320 * defined by @min_limit is reached (or until the tx queue is 3321 * empty). Warning: to be use with care, misuse will impact the 3322 * latency. 3323 */ 3324 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3325 unsigned int min_limit) 3326 { 3327 #ifdef CONFIG_BQL 3328 dev_queue->dql.min_limit = min_limit; 3329 #endif 3330 } 3331 3332 /** 3333 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3334 * @dev_queue: pointer to transmit queue 3335 * 3336 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3337 * to give appropriate hint to the CPU. 3338 */ 3339 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3340 { 3341 #ifdef CONFIG_BQL 3342 prefetchw(&dev_queue->dql.num_queued); 3343 #endif 3344 } 3345 3346 /** 3347 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3348 * @dev_queue: pointer to transmit queue 3349 * 3350 * BQL enabled drivers might use this helper in their TX completion path, 3351 * to give appropriate hint to the CPU. 3352 */ 3353 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3354 { 3355 #ifdef CONFIG_BQL 3356 prefetchw(&dev_queue->dql.limit); 3357 #endif 3358 } 3359 3360 /** 3361 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3362 * @dev_queue: network device queue 3363 * @bytes: number of bytes queued to the device queue 3364 * 3365 * Report the number of bytes queued for sending/completion to the network 3366 * device hardware queue. @bytes should be a good approximation and should 3367 * exactly match netdev_completed_queue() @bytes. 3368 * This is typically called once per packet, from ndo_start_xmit(). 3369 */ 3370 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3371 unsigned int bytes) 3372 { 3373 #ifdef CONFIG_BQL 3374 dql_queued(&dev_queue->dql, bytes); 3375 3376 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3377 return; 3378 3379 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3380 3381 /* 3382 * The XOFF flag must be set before checking the dql_avail below, 3383 * because in netdev_tx_completed_queue we update the dql_completed 3384 * before checking the XOFF flag. 3385 */ 3386 smp_mb(); 3387 3388 /* check again in case another CPU has just made room avail */ 3389 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3390 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3391 #endif 3392 } 3393 3394 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3395 * that they should not test BQL status themselves. 3396 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3397 * skb of a batch. 3398 * Returns true if the doorbell must be used to kick the NIC. 3399 */ 3400 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3401 unsigned int bytes, 3402 bool xmit_more) 3403 { 3404 if (xmit_more) { 3405 #ifdef CONFIG_BQL 3406 dql_queued(&dev_queue->dql, bytes); 3407 #endif 3408 return netif_tx_queue_stopped(dev_queue); 3409 } 3410 netdev_tx_sent_queue(dev_queue, bytes); 3411 return true; 3412 } 3413 3414 /** 3415 * netdev_sent_queue - report the number of bytes queued to hardware 3416 * @dev: network device 3417 * @bytes: number of bytes queued to the hardware device queue 3418 * 3419 * Report the number of bytes queued for sending/completion to the network 3420 * device hardware queue#0. @bytes should be a good approximation and should 3421 * exactly match netdev_completed_queue() @bytes. 3422 * This is typically called once per packet, from ndo_start_xmit(). 3423 */ 3424 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3425 { 3426 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3427 } 3428 3429 static inline bool __netdev_sent_queue(struct net_device *dev, 3430 unsigned int bytes, 3431 bool xmit_more) 3432 { 3433 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3434 xmit_more); 3435 } 3436 3437 /** 3438 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3439 * @dev_queue: network device queue 3440 * @pkts: number of packets (currently ignored) 3441 * @bytes: number of bytes dequeued from the device queue 3442 * 3443 * Must be called at most once per TX completion round (and not per 3444 * individual packet), so that BQL can adjust its limits appropriately. 3445 */ 3446 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3447 unsigned int pkts, unsigned int bytes) 3448 { 3449 #ifdef CONFIG_BQL 3450 if (unlikely(!bytes)) 3451 return; 3452 3453 dql_completed(&dev_queue->dql, bytes); 3454 3455 /* 3456 * Without the memory barrier there is a small possiblity that 3457 * netdev_tx_sent_queue will miss the update and cause the queue to 3458 * be stopped forever 3459 */ 3460 smp_mb(); 3461 3462 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3463 return; 3464 3465 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3466 netif_schedule_queue(dev_queue); 3467 #endif 3468 } 3469 3470 /** 3471 * netdev_completed_queue - report bytes and packets completed by device 3472 * @dev: network device 3473 * @pkts: actual number of packets sent over the medium 3474 * @bytes: actual number of bytes sent over the medium 3475 * 3476 * Report the number of bytes and packets transmitted by the network device 3477 * hardware queue over the physical medium, @bytes must exactly match the 3478 * @bytes amount passed to netdev_sent_queue() 3479 */ 3480 static inline void netdev_completed_queue(struct net_device *dev, 3481 unsigned int pkts, unsigned int bytes) 3482 { 3483 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3484 } 3485 3486 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3487 { 3488 #ifdef CONFIG_BQL 3489 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3490 dql_reset(&q->dql); 3491 #endif 3492 } 3493 3494 /** 3495 * netdev_reset_queue - reset the packets and bytes count of a network device 3496 * @dev_queue: network device 3497 * 3498 * Reset the bytes and packet count of a network device and clear the 3499 * software flow control OFF bit for this network device 3500 */ 3501 static inline void netdev_reset_queue(struct net_device *dev_queue) 3502 { 3503 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3504 } 3505 3506 /** 3507 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3508 * @dev: network device 3509 * @queue_index: given tx queue index 3510 * 3511 * Returns 0 if given tx queue index >= number of device tx queues, 3512 * otherwise returns the originally passed tx queue index. 3513 */ 3514 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3515 { 3516 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3517 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3518 dev->name, queue_index, 3519 dev->real_num_tx_queues); 3520 return 0; 3521 } 3522 3523 return queue_index; 3524 } 3525 3526 /** 3527 * netif_running - test if up 3528 * @dev: network device 3529 * 3530 * Test if the device has been brought up. 3531 */ 3532 static inline bool netif_running(const struct net_device *dev) 3533 { 3534 return test_bit(__LINK_STATE_START, &dev->state); 3535 } 3536 3537 /* 3538 * Routines to manage the subqueues on a device. We only need start, 3539 * stop, and a check if it's stopped. All other device management is 3540 * done at the overall netdevice level. 3541 * Also test the device if we're multiqueue. 3542 */ 3543 3544 /** 3545 * netif_start_subqueue - allow sending packets on subqueue 3546 * @dev: network device 3547 * @queue_index: sub queue index 3548 * 3549 * Start individual transmit queue of a device with multiple transmit queues. 3550 */ 3551 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3552 { 3553 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3554 3555 netif_tx_start_queue(txq); 3556 } 3557 3558 /** 3559 * netif_stop_subqueue - stop sending packets on subqueue 3560 * @dev: network device 3561 * @queue_index: sub queue index 3562 * 3563 * Stop individual transmit queue of a device with multiple transmit queues. 3564 */ 3565 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3566 { 3567 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3568 netif_tx_stop_queue(txq); 3569 } 3570 3571 /** 3572 * __netif_subqueue_stopped - test status of subqueue 3573 * @dev: network device 3574 * @queue_index: sub queue index 3575 * 3576 * Check individual transmit queue of a device with multiple transmit queues. 3577 */ 3578 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3579 u16 queue_index) 3580 { 3581 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3582 3583 return netif_tx_queue_stopped(txq); 3584 } 3585 3586 /** 3587 * netif_subqueue_stopped - test status of subqueue 3588 * @dev: network device 3589 * @skb: sub queue buffer pointer 3590 * 3591 * Check individual transmit queue of a device with multiple transmit queues. 3592 */ 3593 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3594 struct sk_buff *skb) 3595 { 3596 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3597 } 3598 3599 /** 3600 * netif_wake_subqueue - allow sending packets on subqueue 3601 * @dev: network device 3602 * @queue_index: sub queue index 3603 * 3604 * Resume individual transmit queue of a device with multiple transmit queues. 3605 */ 3606 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3607 { 3608 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3609 3610 netif_tx_wake_queue(txq); 3611 } 3612 3613 #ifdef CONFIG_XPS 3614 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3615 u16 index); 3616 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3617 u16 index, enum xps_map_type type); 3618 3619 /** 3620 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3621 * @j: CPU/Rx queue index 3622 * @mask: bitmask of all cpus/rx queues 3623 * @nr_bits: number of bits in the bitmask 3624 * 3625 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3626 */ 3627 static inline bool netif_attr_test_mask(unsigned long j, 3628 const unsigned long *mask, 3629 unsigned int nr_bits) 3630 { 3631 cpu_max_bits_warn(j, nr_bits); 3632 return test_bit(j, mask); 3633 } 3634 3635 /** 3636 * netif_attr_test_online - Test for online CPU/Rx queue 3637 * @j: CPU/Rx queue index 3638 * @online_mask: bitmask for CPUs/Rx queues that are online 3639 * @nr_bits: number of bits in the bitmask 3640 * 3641 * Returns true if a CPU/Rx queue is online. 3642 */ 3643 static inline bool netif_attr_test_online(unsigned long j, 3644 const unsigned long *online_mask, 3645 unsigned int nr_bits) 3646 { 3647 cpu_max_bits_warn(j, nr_bits); 3648 3649 if (online_mask) 3650 return test_bit(j, online_mask); 3651 3652 return (j < nr_bits); 3653 } 3654 3655 /** 3656 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3657 * @n: CPU/Rx queue index 3658 * @srcp: the cpumask/Rx queue mask pointer 3659 * @nr_bits: number of bits in the bitmask 3660 * 3661 * Returns >= nr_bits if no further CPUs/Rx queues set. 3662 */ 3663 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3664 unsigned int nr_bits) 3665 { 3666 /* -1 is a legal arg here. */ 3667 if (n != -1) 3668 cpu_max_bits_warn(n, nr_bits); 3669 3670 if (srcp) 3671 return find_next_bit(srcp, nr_bits, n + 1); 3672 3673 return n + 1; 3674 } 3675 3676 /** 3677 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3678 * @n: CPU/Rx queue index 3679 * @src1p: the first CPUs/Rx queues mask pointer 3680 * @src2p: the second CPUs/Rx queues mask pointer 3681 * @nr_bits: number of bits in the bitmask 3682 * 3683 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3684 */ 3685 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3686 const unsigned long *src2p, 3687 unsigned int nr_bits) 3688 { 3689 /* -1 is a legal arg here. */ 3690 if (n != -1) 3691 cpu_max_bits_warn(n, nr_bits); 3692 3693 if (src1p && src2p) 3694 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3695 else if (src1p) 3696 return find_next_bit(src1p, nr_bits, n + 1); 3697 else if (src2p) 3698 return find_next_bit(src2p, nr_bits, n + 1); 3699 3700 return n + 1; 3701 } 3702 #else 3703 static inline int netif_set_xps_queue(struct net_device *dev, 3704 const struct cpumask *mask, 3705 u16 index) 3706 { 3707 return 0; 3708 } 3709 3710 static inline int __netif_set_xps_queue(struct net_device *dev, 3711 const unsigned long *mask, 3712 u16 index, enum xps_map_type type) 3713 { 3714 return 0; 3715 } 3716 #endif 3717 3718 /** 3719 * netif_is_multiqueue - test if device has multiple transmit queues 3720 * @dev: network device 3721 * 3722 * Check if device has multiple transmit queues 3723 */ 3724 static inline bool netif_is_multiqueue(const struct net_device *dev) 3725 { 3726 return dev->num_tx_queues > 1; 3727 } 3728 3729 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3730 3731 #ifdef CONFIG_SYSFS 3732 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3733 #else 3734 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3735 unsigned int rxqs) 3736 { 3737 dev->real_num_rx_queues = rxqs; 3738 return 0; 3739 } 3740 #endif 3741 int netif_set_real_num_queues(struct net_device *dev, 3742 unsigned int txq, unsigned int rxq); 3743 3744 static inline struct netdev_rx_queue * 3745 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3746 { 3747 return dev->_rx + rxq; 3748 } 3749 3750 #ifdef CONFIG_SYSFS 3751 static inline unsigned int get_netdev_rx_queue_index( 3752 struct netdev_rx_queue *queue) 3753 { 3754 struct net_device *dev = queue->dev; 3755 int index = queue - dev->_rx; 3756 3757 BUG_ON(index >= dev->num_rx_queues); 3758 return index; 3759 } 3760 #endif 3761 3762 int netif_get_num_default_rss_queues(void); 3763 3764 enum skb_free_reason { 3765 SKB_REASON_CONSUMED, 3766 SKB_REASON_DROPPED, 3767 }; 3768 3769 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3770 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3771 3772 /* 3773 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3774 * interrupt context or with hardware interrupts being disabled. 3775 * (in_hardirq() || irqs_disabled()) 3776 * 3777 * We provide four helpers that can be used in following contexts : 3778 * 3779 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3780 * replacing kfree_skb(skb) 3781 * 3782 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3783 * Typically used in place of consume_skb(skb) in TX completion path 3784 * 3785 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3786 * replacing kfree_skb(skb) 3787 * 3788 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3789 * and consumed a packet. Used in place of consume_skb(skb) 3790 */ 3791 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3792 { 3793 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3794 } 3795 3796 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3797 { 3798 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3799 } 3800 3801 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3802 { 3803 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3804 } 3805 3806 static inline void dev_consume_skb_any(struct sk_buff *skb) 3807 { 3808 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3809 } 3810 3811 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3812 struct bpf_prog *xdp_prog); 3813 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3814 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3815 int netif_rx(struct sk_buff *skb); 3816 int __netif_rx(struct sk_buff *skb); 3817 3818 int netif_receive_skb(struct sk_buff *skb); 3819 int netif_receive_skb_core(struct sk_buff *skb); 3820 void netif_receive_skb_list_internal(struct list_head *head); 3821 void netif_receive_skb_list(struct list_head *head); 3822 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3823 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3824 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3825 gro_result_t napi_gro_frags(struct napi_struct *napi); 3826 struct packet_offload *gro_find_receive_by_type(__be16 type); 3827 struct packet_offload *gro_find_complete_by_type(__be16 type); 3828 3829 static inline void napi_free_frags(struct napi_struct *napi) 3830 { 3831 kfree_skb(napi->skb); 3832 napi->skb = NULL; 3833 } 3834 3835 bool netdev_is_rx_handler_busy(struct net_device *dev); 3836 int netdev_rx_handler_register(struct net_device *dev, 3837 rx_handler_func_t *rx_handler, 3838 void *rx_handler_data); 3839 void netdev_rx_handler_unregister(struct net_device *dev); 3840 3841 bool dev_valid_name(const char *name); 3842 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 3843 { 3844 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 3845 } 3846 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 3847 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 3848 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3849 void __user *data, bool *need_copyout); 3850 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 3851 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 3852 unsigned int dev_get_flags(const struct net_device *); 3853 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3854 struct netlink_ext_ack *extack); 3855 int dev_change_flags(struct net_device *dev, unsigned int flags, 3856 struct netlink_ext_ack *extack); 3857 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3858 unsigned int gchanges); 3859 int dev_set_alias(struct net_device *, const char *, size_t); 3860 int dev_get_alias(const struct net_device *, char *, size_t); 3861 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 3862 const char *pat, int new_ifindex); 3863 static inline 3864 int dev_change_net_namespace(struct net_device *dev, struct net *net, 3865 const char *pat) 3866 { 3867 return __dev_change_net_namespace(dev, net, pat, 0); 3868 } 3869 int __dev_set_mtu(struct net_device *, int); 3870 int dev_set_mtu(struct net_device *, int); 3871 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3872 struct netlink_ext_ack *extack); 3873 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3874 struct netlink_ext_ack *extack); 3875 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 3876 struct netlink_ext_ack *extack); 3877 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 3878 int dev_get_port_parent_id(struct net_device *dev, 3879 struct netdev_phys_item_id *ppid, bool recurse); 3880 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3881 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3882 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3883 struct netdev_queue *txq, int *ret); 3884 3885 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3886 u8 dev_xdp_prog_count(struct net_device *dev); 3887 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3888 3889 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3890 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3891 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 3892 bool is_skb_forwardable(const struct net_device *dev, 3893 const struct sk_buff *skb); 3894 3895 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 3896 const struct sk_buff *skb, 3897 const bool check_mtu) 3898 { 3899 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 3900 unsigned int len; 3901 3902 if (!(dev->flags & IFF_UP)) 3903 return false; 3904 3905 if (!check_mtu) 3906 return true; 3907 3908 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 3909 if (skb->len <= len) 3910 return true; 3911 3912 /* if TSO is enabled, we don't care about the length as the packet 3913 * could be forwarded without being segmented before 3914 */ 3915 if (skb_is_gso(skb)) 3916 return true; 3917 3918 return false; 3919 } 3920 3921 struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev); 3922 3923 static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev) 3924 { 3925 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */ 3926 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats); 3927 3928 if (likely(p)) 3929 return p; 3930 3931 return netdev_core_stats_alloc(dev); 3932 } 3933 3934 #define DEV_CORE_STATS_INC(FIELD) \ 3935 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 3936 { \ 3937 struct net_device_core_stats __percpu *p; \ 3938 \ 3939 p = dev_core_stats(dev); \ 3940 if (p) \ 3941 this_cpu_inc(p->FIELD); \ 3942 } 3943 DEV_CORE_STATS_INC(rx_dropped) 3944 DEV_CORE_STATS_INC(tx_dropped) 3945 DEV_CORE_STATS_INC(rx_nohandler) 3946 DEV_CORE_STATS_INC(rx_otherhost_dropped) 3947 3948 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3949 struct sk_buff *skb, 3950 const bool check_mtu) 3951 { 3952 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3953 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 3954 dev_core_stats_rx_dropped_inc(dev); 3955 kfree_skb(skb); 3956 return NET_RX_DROP; 3957 } 3958 3959 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 3960 skb->priority = 0; 3961 return 0; 3962 } 3963 3964 bool dev_nit_active(struct net_device *dev); 3965 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3966 3967 static inline void __dev_put(struct net_device *dev) 3968 { 3969 if (dev) { 3970 #ifdef CONFIG_PCPU_DEV_REFCNT 3971 this_cpu_dec(*dev->pcpu_refcnt); 3972 #else 3973 refcount_dec(&dev->dev_refcnt); 3974 #endif 3975 } 3976 } 3977 3978 static inline void __dev_hold(struct net_device *dev) 3979 { 3980 if (dev) { 3981 #ifdef CONFIG_PCPU_DEV_REFCNT 3982 this_cpu_inc(*dev->pcpu_refcnt); 3983 #else 3984 refcount_inc(&dev->dev_refcnt); 3985 #endif 3986 } 3987 } 3988 3989 static inline void __netdev_tracker_alloc(struct net_device *dev, 3990 netdevice_tracker *tracker, 3991 gfp_t gfp) 3992 { 3993 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 3994 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 3995 #endif 3996 } 3997 3998 /* netdev_tracker_alloc() can upgrade a prior untracked reference 3999 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4000 */ 4001 static inline void netdev_tracker_alloc(struct net_device *dev, 4002 netdevice_tracker *tracker, gfp_t gfp) 4003 { 4004 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4005 refcount_dec(&dev->refcnt_tracker.no_tracker); 4006 __netdev_tracker_alloc(dev, tracker, gfp); 4007 #endif 4008 } 4009 4010 static inline void netdev_tracker_free(struct net_device *dev, 4011 netdevice_tracker *tracker) 4012 { 4013 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4014 ref_tracker_free(&dev->refcnt_tracker, tracker); 4015 #endif 4016 } 4017 4018 static inline void netdev_hold(struct net_device *dev, 4019 netdevice_tracker *tracker, gfp_t gfp) 4020 { 4021 if (dev) { 4022 __dev_hold(dev); 4023 __netdev_tracker_alloc(dev, tracker, gfp); 4024 } 4025 } 4026 4027 static inline void netdev_put(struct net_device *dev, 4028 netdevice_tracker *tracker) 4029 { 4030 if (dev) { 4031 netdev_tracker_free(dev, tracker); 4032 __dev_put(dev); 4033 } 4034 } 4035 4036 /** 4037 * dev_hold - get reference to device 4038 * @dev: network device 4039 * 4040 * Hold reference to device to keep it from being freed. 4041 * Try using netdev_hold() instead. 4042 */ 4043 static inline void dev_hold(struct net_device *dev) 4044 { 4045 netdev_hold(dev, NULL, GFP_ATOMIC); 4046 } 4047 4048 /** 4049 * dev_put - release reference to device 4050 * @dev: network device 4051 * 4052 * Release reference to device to allow it to be freed. 4053 * Try using netdev_put() instead. 4054 */ 4055 static inline void dev_put(struct net_device *dev) 4056 { 4057 netdev_put(dev, NULL); 4058 } 4059 4060 static inline void netdev_ref_replace(struct net_device *odev, 4061 struct net_device *ndev, 4062 netdevice_tracker *tracker, 4063 gfp_t gfp) 4064 { 4065 if (odev) 4066 netdev_tracker_free(odev, tracker); 4067 4068 __dev_hold(ndev); 4069 __dev_put(odev); 4070 4071 if (ndev) 4072 __netdev_tracker_alloc(ndev, tracker, gfp); 4073 } 4074 4075 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4076 * and _off may be called from IRQ context, but it is caller 4077 * who is responsible for serialization of these calls. 4078 * 4079 * The name carrier is inappropriate, these functions should really be 4080 * called netif_lowerlayer_*() because they represent the state of any 4081 * kind of lower layer not just hardware media. 4082 */ 4083 void linkwatch_fire_event(struct net_device *dev); 4084 4085 /** 4086 * netif_carrier_ok - test if carrier present 4087 * @dev: network device 4088 * 4089 * Check if carrier is present on device 4090 */ 4091 static inline bool netif_carrier_ok(const struct net_device *dev) 4092 { 4093 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4094 } 4095 4096 unsigned long dev_trans_start(struct net_device *dev); 4097 4098 void __netdev_watchdog_up(struct net_device *dev); 4099 4100 void netif_carrier_on(struct net_device *dev); 4101 void netif_carrier_off(struct net_device *dev); 4102 void netif_carrier_event(struct net_device *dev); 4103 4104 /** 4105 * netif_dormant_on - mark device as dormant. 4106 * @dev: network device 4107 * 4108 * Mark device as dormant (as per RFC2863). 4109 * 4110 * The dormant state indicates that the relevant interface is not 4111 * actually in a condition to pass packets (i.e., it is not 'up') but is 4112 * in a "pending" state, waiting for some external event. For "on- 4113 * demand" interfaces, this new state identifies the situation where the 4114 * interface is waiting for events to place it in the up state. 4115 */ 4116 static inline void netif_dormant_on(struct net_device *dev) 4117 { 4118 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4119 linkwatch_fire_event(dev); 4120 } 4121 4122 /** 4123 * netif_dormant_off - set device as not dormant. 4124 * @dev: network device 4125 * 4126 * Device is not in dormant state. 4127 */ 4128 static inline void netif_dormant_off(struct net_device *dev) 4129 { 4130 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4131 linkwatch_fire_event(dev); 4132 } 4133 4134 /** 4135 * netif_dormant - test if device is dormant 4136 * @dev: network device 4137 * 4138 * Check if device is dormant. 4139 */ 4140 static inline bool netif_dormant(const struct net_device *dev) 4141 { 4142 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4143 } 4144 4145 4146 /** 4147 * netif_testing_on - mark device as under test. 4148 * @dev: network device 4149 * 4150 * Mark device as under test (as per RFC2863). 4151 * 4152 * The testing state indicates that some test(s) must be performed on 4153 * the interface. After completion, of the test, the interface state 4154 * will change to up, dormant, or down, as appropriate. 4155 */ 4156 static inline void netif_testing_on(struct net_device *dev) 4157 { 4158 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4159 linkwatch_fire_event(dev); 4160 } 4161 4162 /** 4163 * netif_testing_off - set device as not under test. 4164 * @dev: network device 4165 * 4166 * Device is not in testing state. 4167 */ 4168 static inline void netif_testing_off(struct net_device *dev) 4169 { 4170 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4171 linkwatch_fire_event(dev); 4172 } 4173 4174 /** 4175 * netif_testing - test if device is under test 4176 * @dev: network device 4177 * 4178 * Check if device is under test 4179 */ 4180 static inline bool netif_testing(const struct net_device *dev) 4181 { 4182 return test_bit(__LINK_STATE_TESTING, &dev->state); 4183 } 4184 4185 4186 /** 4187 * netif_oper_up - test if device is operational 4188 * @dev: network device 4189 * 4190 * Check if carrier is operational 4191 */ 4192 static inline bool netif_oper_up(const struct net_device *dev) 4193 { 4194 return (dev->operstate == IF_OPER_UP || 4195 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4196 } 4197 4198 /** 4199 * netif_device_present - is device available or removed 4200 * @dev: network device 4201 * 4202 * Check if device has not been removed from system. 4203 */ 4204 static inline bool netif_device_present(const struct net_device *dev) 4205 { 4206 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4207 } 4208 4209 void netif_device_detach(struct net_device *dev); 4210 4211 void netif_device_attach(struct net_device *dev); 4212 4213 /* 4214 * Network interface message level settings 4215 */ 4216 4217 enum { 4218 NETIF_MSG_DRV_BIT, 4219 NETIF_MSG_PROBE_BIT, 4220 NETIF_MSG_LINK_BIT, 4221 NETIF_MSG_TIMER_BIT, 4222 NETIF_MSG_IFDOWN_BIT, 4223 NETIF_MSG_IFUP_BIT, 4224 NETIF_MSG_RX_ERR_BIT, 4225 NETIF_MSG_TX_ERR_BIT, 4226 NETIF_MSG_TX_QUEUED_BIT, 4227 NETIF_MSG_INTR_BIT, 4228 NETIF_MSG_TX_DONE_BIT, 4229 NETIF_MSG_RX_STATUS_BIT, 4230 NETIF_MSG_PKTDATA_BIT, 4231 NETIF_MSG_HW_BIT, 4232 NETIF_MSG_WOL_BIT, 4233 4234 /* When you add a new bit above, update netif_msg_class_names array 4235 * in net/ethtool/common.c 4236 */ 4237 NETIF_MSG_CLASS_COUNT, 4238 }; 4239 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4240 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4241 4242 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4243 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4244 4245 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4246 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4247 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4248 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4249 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4250 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4251 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4252 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4253 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4254 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4255 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4256 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4257 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4258 #define NETIF_MSG_HW __NETIF_MSG(HW) 4259 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4260 4261 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4262 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4263 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4264 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4265 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4266 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4267 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4268 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4269 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4270 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4271 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4272 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4273 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4274 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4275 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4276 4277 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4278 { 4279 /* use default */ 4280 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4281 return default_msg_enable_bits; 4282 if (debug_value == 0) /* no output */ 4283 return 0; 4284 /* set low N bits */ 4285 return (1U << debug_value) - 1; 4286 } 4287 4288 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4289 { 4290 spin_lock(&txq->_xmit_lock); 4291 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4292 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4293 } 4294 4295 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4296 { 4297 __acquire(&txq->_xmit_lock); 4298 return true; 4299 } 4300 4301 static inline void __netif_tx_release(struct netdev_queue *txq) 4302 { 4303 __release(&txq->_xmit_lock); 4304 } 4305 4306 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4307 { 4308 spin_lock_bh(&txq->_xmit_lock); 4309 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4310 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4311 } 4312 4313 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4314 { 4315 bool ok = spin_trylock(&txq->_xmit_lock); 4316 4317 if (likely(ok)) { 4318 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4319 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4320 } 4321 return ok; 4322 } 4323 4324 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4325 { 4326 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4327 WRITE_ONCE(txq->xmit_lock_owner, -1); 4328 spin_unlock(&txq->_xmit_lock); 4329 } 4330 4331 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4332 { 4333 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4334 WRITE_ONCE(txq->xmit_lock_owner, -1); 4335 spin_unlock_bh(&txq->_xmit_lock); 4336 } 4337 4338 /* 4339 * txq->trans_start can be read locklessly from dev_watchdog() 4340 */ 4341 static inline void txq_trans_update(struct netdev_queue *txq) 4342 { 4343 if (txq->xmit_lock_owner != -1) 4344 WRITE_ONCE(txq->trans_start, jiffies); 4345 } 4346 4347 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4348 { 4349 unsigned long now = jiffies; 4350 4351 if (READ_ONCE(txq->trans_start) != now) 4352 WRITE_ONCE(txq->trans_start, now); 4353 } 4354 4355 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4356 static inline void netif_trans_update(struct net_device *dev) 4357 { 4358 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4359 4360 txq_trans_cond_update(txq); 4361 } 4362 4363 /** 4364 * netif_tx_lock - grab network device transmit lock 4365 * @dev: network device 4366 * 4367 * Get network device transmit lock 4368 */ 4369 void netif_tx_lock(struct net_device *dev); 4370 4371 static inline void netif_tx_lock_bh(struct net_device *dev) 4372 { 4373 local_bh_disable(); 4374 netif_tx_lock(dev); 4375 } 4376 4377 void netif_tx_unlock(struct net_device *dev); 4378 4379 static inline void netif_tx_unlock_bh(struct net_device *dev) 4380 { 4381 netif_tx_unlock(dev); 4382 local_bh_enable(); 4383 } 4384 4385 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4386 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4387 __netif_tx_lock(txq, cpu); \ 4388 } else { \ 4389 __netif_tx_acquire(txq); \ 4390 } \ 4391 } 4392 4393 #define HARD_TX_TRYLOCK(dev, txq) \ 4394 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4395 __netif_tx_trylock(txq) : \ 4396 __netif_tx_acquire(txq)) 4397 4398 #define HARD_TX_UNLOCK(dev, txq) { \ 4399 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4400 __netif_tx_unlock(txq); \ 4401 } else { \ 4402 __netif_tx_release(txq); \ 4403 } \ 4404 } 4405 4406 static inline void netif_tx_disable(struct net_device *dev) 4407 { 4408 unsigned int i; 4409 int cpu; 4410 4411 local_bh_disable(); 4412 cpu = smp_processor_id(); 4413 spin_lock(&dev->tx_global_lock); 4414 for (i = 0; i < dev->num_tx_queues; i++) { 4415 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4416 4417 __netif_tx_lock(txq, cpu); 4418 netif_tx_stop_queue(txq); 4419 __netif_tx_unlock(txq); 4420 } 4421 spin_unlock(&dev->tx_global_lock); 4422 local_bh_enable(); 4423 } 4424 4425 static inline void netif_addr_lock(struct net_device *dev) 4426 { 4427 unsigned char nest_level = 0; 4428 4429 #ifdef CONFIG_LOCKDEP 4430 nest_level = dev->nested_level; 4431 #endif 4432 spin_lock_nested(&dev->addr_list_lock, nest_level); 4433 } 4434 4435 static inline void netif_addr_lock_bh(struct net_device *dev) 4436 { 4437 unsigned char nest_level = 0; 4438 4439 #ifdef CONFIG_LOCKDEP 4440 nest_level = dev->nested_level; 4441 #endif 4442 local_bh_disable(); 4443 spin_lock_nested(&dev->addr_list_lock, nest_level); 4444 } 4445 4446 static inline void netif_addr_unlock(struct net_device *dev) 4447 { 4448 spin_unlock(&dev->addr_list_lock); 4449 } 4450 4451 static inline void netif_addr_unlock_bh(struct net_device *dev) 4452 { 4453 spin_unlock_bh(&dev->addr_list_lock); 4454 } 4455 4456 /* 4457 * dev_addrs walker. Should be used only for read access. Call with 4458 * rcu_read_lock held. 4459 */ 4460 #define for_each_dev_addr(dev, ha) \ 4461 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4462 4463 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4464 4465 void ether_setup(struct net_device *dev); 4466 4467 /* Support for loadable net-drivers */ 4468 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4469 unsigned char name_assign_type, 4470 void (*setup)(struct net_device *), 4471 unsigned int txqs, unsigned int rxqs); 4472 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4473 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4474 4475 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4476 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4477 count) 4478 4479 int register_netdev(struct net_device *dev); 4480 void unregister_netdev(struct net_device *dev); 4481 4482 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4483 4484 /* General hardware address lists handling functions */ 4485 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4486 struct netdev_hw_addr_list *from_list, int addr_len); 4487 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4488 struct netdev_hw_addr_list *from_list, int addr_len); 4489 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4490 struct net_device *dev, 4491 int (*sync)(struct net_device *, const unsigned char *), 4492 int (*unsync)(struct net_device *, 4493 const unsigned char *)); 4494 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4495 struct net_device *dev, 4496 int (*sync)(struct net_device *, 4497 const unsigned char *, int), 4498 int (*unsync)(struct net_device *, 4499 const unsigned char *, int)); 4500 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4501 struct net_device *dev, 4502 int (*unsync)(struct net_device *, 4503 const unsigned char *, int)); 4504 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4505 struct net_device *dev, 4506 int (*unsync)(struct net_device *, 4507 const unsigned char *)); 4508 void __hw_addr_init(struct netdev_hw_addr_list *list); 4509 4510 /* Functions used for device addresses handling */ 4511 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4512 const void *addr, size_t len); 4513 4514 static inline void 4515 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4516 { 4517 dev_addr_mod(dev, 0, addr, len); 4518 } 4519 4520 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4521 { 4522 __dev_addr_set(dev, addr, dev->addr_len); 4523 } 4524 4525 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4526 unsigned char addr_type); 4527 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4528 unsigned char addr_type); 4529 4530 /* Functions used for unicast addresses handling */ 4531 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4532 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4533 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4534 int dev_uc_sync(struct net_device *to, struct net_device *from); 4535 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4536 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4537 void dev_uc_flush(struct net_device *dev); 4538 void dev_uc_init(struct net_device *dev); 4539 4540 /** 4541 * __dev_uc_sync - Synchonize device's unicast list 4542 * @dev: device to sync 4543 * @sync: function to call if address should be added 4544 * @unsync: function to call if address should be removed 4545 * 4546 * Add newly added addresses to the interface, and release 4547 * addresses that have been deleted. 4548 */ 4549 static inline int __dev_uc_sync(struct net_device *dev, 4550 int (*sync)(struct net_device *, 4551 const unsigned char *), 4552 int (*unsync)(struct net_device *, 4553 const unsigned char *)) 4554 { 4555 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4556 } 4557 4558 /** 4559 * __dev_uc_unsync - Remove synchronized addresses from device 4560 * @dev: device to sync 4561 * @unsync: function to call if address should be removed 4562 * 4563 * Remove all addresses that were added to the device by dev_uc_sync(). 4564 */ 4565 static inline void __dev_uc_unsync(struct net_device *dev, 4566 int (*unsync)(struct net_device *, 4567 const unsigned char *)) 4568 { 4569 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4570 } 4571 4572 /* Functions used for multicast addresses handling */ 4573 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4574 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4575 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4576 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4577 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4578 int dev_mc_sync(struct net_device *to, struct net_device *from); 4579 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4580 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4581 void dev_mc_flush(struct net_device *dev); 4582 void dev_mc_init(struct net_device *dev); 4583 4584 /** 4585 * __dev_mc_sync - Synchonize device's multicast list 4586 * @dev: device to sync 4587 * @sync: function to call if address should be added 4588 * @unsync: function to call if address should be removed 4589 * 4590 * Add newly added addresses to the interface, and release 4591 * addresses that have been deleted. 4592 */ 4593 static inline int __dev_mc_sync(struct net_device *dev, 4594 int (*sync)(struct net_device *, 4595 const unsigned char *), 4596 int (*unsync)(struct net_device *, 4597 const unsigned char *)) 4598 { 4599 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4600 } 4601 4602 /** 4603 * __dev_mc_unsync - Remove synchronized addresses from device 4604 * @dev: device to sync 4605 * @unsync: function to call if address should be removed 4606 * 4607 * Remove all addresses that were added to the device by dev_mc_sync(). 4608 */ 4609 static inline void __dev_mc_unsync(struct net_device *dev, 4610 int (*unsync)(struct net_device *, 4611 const unsigned char *)) 4612 { 4613 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4614 } 4615 4616 /* Functions used for secondary unicast and multicast support */ 4617 void dev_set_rx_mode(struct net_device *dev); 4618 int dev_set_promiscuity(struct net_device *dev, int inc); 4619 int dev_set_allmulti(struct net_device *dev, int inc); 4620 void netdev_state_change(struct net_device *dev); 4621 void __netdev_notify_peers(struct net_device *dev); 4622 void netdev_notify_peers(struct net_device *dev); 4623 void netdev_features_change(struct net_device *dev); 4624 /* Load a device via the kmod */ 4625 void dev_load(struct net *net, const char *name); 4626 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4627 struct rtnl_link_stats64 *storage); 4628 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4629 const struct net_device_stats *netdev_stats); 4630 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4631 const struct pcpu_sw_netstats __percpu *netstats); 4632 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4633 4634 extern int netdev_max_backlog; 4635 extern int dev_rx_weight; 4636 extern int dev_tx_weight; 4637 extern int gro_normal_batch; 4638 4639 enum { 4640 NESTED_SYNC_IMM_BIT, 4641 NESTED_SYNC_TODO_BIT, 4642 }; 4643 4644 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4645 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4646 4647 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4648 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4649 4650 struct netdev_nested_priv { 4651 unsigned char flags; 4652 void *data; 4653 }; 4654 4655 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4656 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4657 struct list_head **iter); 4658 4659 /* iterate through upper list, must be called under RCU read lock */ 4660 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4661 for (iter = &(dev)->adj_list.upper, \ 4662 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4663 updev; \ 4664 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4665 4666 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4667 int (*fn)(struct net_device *upper_dev, 4668 struct netdev_nested_priv *priv), 4669 struct netdev_nested_priv *priv); 4670 4671 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4672 struct net_device *upper_dev); 4673 4674 bool netdev_has_any_upper_dev(struct net_device *dev); 4675 4676 void *netdev_lower_get_next_private(struct net_device *dev, 4677 struct list_head **iter); 4678 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4679 struct list_head **iter); 4680 4681 #define netdev_for_each_lower_private(dev, priv, iter) \ 4682 for (iter = (dev)->adj_list.lower.next, \ 4683 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4684 priv; \ 4685 priv = netdev_lower_get_next_private(dev, &(iter))) 4686 4687 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4688 for (iter = &(dev)->adj_list.lower, \ 4689 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4690 priv; \ 4691 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4692 4693 void *netdev_lower_get_next(struct net_device *dev, 4694 struct list_head **iter); 4695 4696 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4697 for (iter = (dev)->adj_list.lower.next, \ 4698 ldev = netdev_lower_get_next(dev, &(iter)); \ 4699 ldev; \ 4700 ldev = netdev_lower_get_next(dev, &(iter))) 4701 4702 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4703 struct list_head **iter); 4704 int netdev_walk_all_lower_dev(struct net_device *dev, 4705 int (*fn)(struct net_device *lower_dev, 4706 struct netdev_nested_priv *priv), 4707 struct netdev_nested_priv *priv); 4708 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4709 int (*fn)(struct net_device *lower_dev, 4710 struct netdev_nested_priv *priv), 4711 struct netdev_nested_priv *priv); 4712 4713 void *netdev_adjacent_get_private(struct list_head *adj_list); 4714 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4715 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4716 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4717 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4718 struct netlink_ext_ack *extack); 4719 int netdev_master_upper_dev_link(struct net_device *dev, 4720 struct net_device *upper_dev, 4721 void *upper_priv, void *upper_info, 4722 struct netlink_ext_ack *extack); 4723 void netdev_upper_dev_unlink(struct net_device *dev, 4724 struct net_device *upper_dev); 4725 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4726 struct net_device *new_dev, 4727 struct net_device *dev, 4728 struct netlink_ext_ack *extack); 4729 void netdev_adjacent_change_commit(struct net_device *old_dev, 4730 struct net_device *new_dev, 4731 struct net_device *dev); 4732 void netdev_adjacent_change_abort(struct net_device *old_dev, 4733 struct net_device *new_dev, 4734 struct net_device *dev); 4735 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4736 void *netdev_lower_dev_get_private(struct net_device *dev, 4737 struct net_device *lower_dev); 4738 void netdev_lower_state_changed(struct net_device *lower_dev, 4739 void *lower_state_info); 4740 4741 /* RSS keys are 40 or 52 bytes long */ 4742 #define NETDEV_RSS_KEY_LEN 52 4743 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4744 void netdev_rss_key_fill(void *buffer, size_t len); 4745 4746 int skb_checksum_help(struct sk_buff *skb); 4747 int skb_crc32c_csum_help(struct sk_buff *skb); 4748 int skb_csum_hwoffload_help(struct sk_buff *skb, 4749 const netdev_features_t features); 4750 4751 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4752 netdev_features_t features, bool tx_path); 4753 struct sk_buff *skb_eth_gso_segment(struct sk_buff *skb, 4754 netdev_features_t features, __be16 type); 4755 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4756 netdev_features_t features); 4757 4758 struct netdev_bonding_info { 4759 ifslave slave; 4760 ifbond master; 4761 }; 4762 4763 struct netdev_notifier_bonding_info { 4764 struct netdev_notifier_info info; /* must be first */ 4765 struct netdev_bonding_info bonding_info; 4766 }; 4767 4768 void netdev_bonding_info_change(struct net_device *dev, 4769 struct netdev_bonding_info *bonding_info); 4770 4771 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4772 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4773 #else 4774 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4775 const void *data) 4776 { 4777 } 4778 #endif 4779 4780 static inline 4781 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4782 { 4783 return __skb_gso_segment(skb, features, true); 4784 } 4785 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4786 4787 static inline bool can_checksum_protocol(netdev_features_t features, 4788 __be16 protocol) 4789 { 4790 if (protocol == htons(ETH_P_FCOE)) 4791 return !!(features & NETIF_F_FCOE_CRC); 4792 4793 /* Assume this is an IP checksum (not SCTP CRC) */ 4794 4795 if (features & NETIF_F_HW_CSUM) { 4796 /* Can checksum everything */ 4797 return true; 4798 } 4799 4800 switch (protocol) { 4801 case htons(ETH_P_IP): 4802 return !!(features & NETIF_F_IP_CSUM); 4803 case htons(ETH_P_IPV6): 4804 return !!(features & NETIF_F_IPV6_CSUM); 4805 default: 4806 return false; 4807 } 4808 } 4809 4810 #ifdef CONFIG_BUG 4811 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4812 #else 4813 static inline void netdev_rx_csum_fault(struct net_device *dev, 4814 struct sk_buff *skb) 4815 { 4816 } 4817 #endif 4818 /* rx skb timestamps */ 4819 void net_enable_timestamp(void); 4820 void net_disable_timestamp(void); 4821 4822 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 4823 const struct skb_shared_hwtstamps *hwtstamps, 4824 bool cycles) 4825 { 4826 const struct net_device_ops *ops = dev->netdev_ops; 4827 4828 if (ops->ndo_get_tstamp) 4829 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 4830 4831 return hwtstamps->hwtstamp; 4832 } 4833 4834 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4835 struct sk_buff *skb, struct net_device *dev, 4836 bool more) 4837 { 4838 __this_cpu_write(softnet_data.xmit.more, more); 4839 return ops->ndo_start_xmit(skb, dev); 4840 } 4841 4842 static inline bool netdev_xmit_more(void) 4843 { 4844 return __this_cpu_read(softnet_data.xmit.more); 4845 } 4846 4847 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4848 struct netdev_queue *txq, bool more) 4849 { 4850 const struct net_device_ops *ops = dev->netdev_ops; 4851 netdev_tx_t rc; 4852 4853 rc = __netdev_start_xmit(ops, skb, dev, more); 4854 if (rc == NETDEV_TX_OK) 4855 txq_trans_update(txq); 4856 4857 return rc; 4858 } 4859 4860 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4861 const void *ns); 4862 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4863 const void *ns); 4864 4865 extern const struct kobj_ns_type_operations net_ns_type_operations; 4866 4867 const char *netdev_drivername(const struct net_device *dev); 4868 4869 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4870 netdev_features_t f2) 4871 { 4872 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4873 if (f1 & NETIF_F_HW_CSUM) 4874 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4875 else 4876 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4877 } 4878 4879 return f1 & f2; 4880 } 4881 4882 static inline netdev_features_t netdev_get_wanted_features( 4883 struct net_device *dev) 4884 { 4885 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4886 } 4887 netdev_features_t netdev_increment_features(netdev_features_t all, 4888 netdev_features_t one, netdev_features_t mask); 4889 4890 /* Allow TSO being used on stacked device : 4891 * Performing the GSO segmentation before last device 4892 * is a performance improvement. 4893 */ 4894 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4895 netdev_features_t mask) 4896 { 4897 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4898 } 4899 4900 int __netdev_update_features(struct net_device *dev); 4901 void netdev_update_features(struct net_device *dev); 4902 void netdev_change_features(struct net_device *dev); 4903 4904 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4905 struct net_device *dev); 4906 4907 netdev_features_t passthru_features_check(struct sk_buff *skb, 4908 struct net_device *dev, 4909 netdev_features_t features); 4910 netdev_features_t netif_skb_features(struct sk_buff *skb); 4911 4912 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4913 { 4914 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4915 4916 /* check flags correspondence */ 4917 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4918 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4919 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4920 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4921 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4922 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4923 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4924 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4925 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4926 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4927 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4928 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4929 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4930 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4931 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4932 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4933 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4934 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4935 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4936 4937 return (features & feature) == feature; 4938 } 4939 4940 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4941 { 4942 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4943 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4944 } 4945 4946 static inline bool netif_needs_gso(struct sk_buff *skb, 4947 netdev_features_t features) 4948 { 4949 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4950 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4951 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4952 } 4953 4954 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 4955 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 4956 void netif_inherit_tso_max(struct net_device *to, 4957 const struct net_device *from); 4958 4959 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4960 int pulled_hlen, u16 mac_offset, 4961 int mac_len) 4962 { 4963 skb->protocol = protocol; 4964 skb->encapsulation = 1; 4965 skb_push(skb, pulled_hlen); 4966 skb_reset_transport_header(skb); 4967 skb->mac_header = mac_offset; 4968 skb->network_header = skb->mac_header + mac_len; 4969 skb->mac_len = mac_len; 4970 } 4971 4972 static inline bool netif_is_macsec(const struct net_device *dev) 4973 { 4974 return dev->priv_flags & IFF_MACSEC; 4975 } 4976 4977 static inline bool netif_is_macvlan(const struct net_device *dev) 4978 { 4979 return dev->priv_flags & IFF_MACVLAN; 4980 } 4981 4982 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4983 { 4984 return dev->priv_flags & IFF_MACVLAN_PORT; 4985 } 4986 4987 static inline bool netif_is_bond_master(const struct net_device *dev) 4988 { 4989 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4990 } 4991 4992 static inline bool netif_is_bond_slave(const struct net_device *dev) 4993 { 4994 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4995 } 4996 4997 static inline bool netif_supports_nofcs(struct net_device *dev) 4998 { 4999 return dev->priv_flags & IFF_SUPP_NOFCS; 5000 } 5001 5002 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5003 { 5004 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5005 } 5006 5007 static inline bool netif_is_l3_master(const struct net_device *dev) 5008 { 5009 return dev->priv_flags & IFF_L3MDEV_MASTER; 5010 } 5011 5012 static inline bool netif_is_l3_slave(const struct net_device *dev) 5013 { 5014 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5015 } 5016 5017 static inline bool netif_is_bridge_master(const struct net_device *dev) 5018 { 5019 return dev->priv_flags & IFF_EBRIDGE; 5020 } 5021 5022 static inline bool netif_is_bridge_port(const struct net_device *dev) 5023 { 5024 return dev->priv_flags & IFF_BRIDGE_PORT; 5025 } 5026 5027 static inline bool netif_is_ovs_master(const struct net_device *dev) 5028 { 5029 return dev->priv_flags & IFF_OPENVSWITCH; 5030 } 5031 5032 static inline bool netif_is_ovs_port(const struct net_device *dev) 5033 { 5034 return dev->priv_flags & IFF_OVS_DATAPATH; 5035 } 5036 5037 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5038 { 5039 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5040 } 5041 5042 static inline bool netif_is_team_master(const struct net_device *dev) 5043 { 5044 return dev->priv_flags & IFF_TEAM; 5045 } 5046 5047 static inline bool netif_is_team_port(const struct net_device *dev) 5048 { 5049 return dev->priv_flags & IFF_TEAM_PORT; 5050 } 5051 5052 static inline bool netif_is_lag_master(const struct net_device *dev) 5053 { 5054 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5055 } 5056 5057 static inline bool netif_is_lag_port(const struct net_device *dev) 5058 { 5059 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5060 } 5061 5062 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5063 { 5064 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5065 } 5066 5067 static inline bool netif_is_failover(const struct net_device *dev) 5068 { 5069 return dev->priv_flags & IFF_FAILOVER; 5070 } 5071 5072 static inline bool netif_is_failover_slave(const struct net_device *dev) 5073 { 5074 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5075 } 5076 5077 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5078 static inline void netif_keep_dst(struct net_device *dev) 5079 { 5080 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5081 } 5082 5083 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5084 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5085 { 5086 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5087 return netif_is_macsec(dev); 5088 } 5089 5090 extern struct pernet_operations __net_initdata loopback_net_ops; 5091 5092 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5093 5094 /* netdev_printk helpers, similar to dev_printk */ 5095 5096 static inline const char *netdev_name(const struct net_device *dev) 5097 { 5098 if (!dev->name[0] || strchr(dev->name, '%')) 5099 return "(unnamed net_device)"; 5100 return dev->name; 5101 } 5102 5103 static inline bool netdev_unregistering(const struct net_device *dev) 5104 { 5105 return dev->reg_state == NETREG_UNREGISTERING; 5106 } 5107 5108 static inline const char *netdev_reg_state(const struct net_device *dev) 5109 { 5110 switch (dev->reg_state) { 5111 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5112 case NETREG_REGISTERED: return ""; 5113 case NETREG_UNREGISTERING: return " (unregistering)"; 5114 case NETREG_UNREGISTERED: return " (unregistered)"; 5115 case NETREG_RELEASED: return " (released)"; 5116 case NETREG_DUMMY: return " (dummy)"; 5117 } 5118 5119 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5120 return " (unknown)"; 5121 } 5122 5123 #define MODULE_ALIAS_NETDEV(device) \ 5124 MODULE_ALIAS("netdev-" device) 5125 5126 /* 5127 * netdev_WARN() acts like dev_printk(), but with the key difference 5128 * of using a WARN/WARN_ON to get the message out, including the 5129 * file/line information and a backtrace. 5130 */ 5131 #define netdev_WARN(dev, format, args...) \ 5132 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5133 netdev_reg_state(dev), ##args) 5134 5135 #define netdev_WARN_ONCE(dev, format, args...) \ 5136 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5137 netdev_reg_state(dev), ##args) 5138 5139 /* 5140 * The list of packet types we will receive (as opposed to discard) 5141 * and the routines to invoke. 5142 * 5143 * Why 16. Because with 16 the only overlap we get on a hash of the 5144 * low nibble of the protocol value is RARP/SNAP/X.25. 5145 * 5146 * 0800 IP 5147 * 0001 802.3 5148 * 0002 AX.25 5149 * 0004 802.2 5150 * 8035 RARP 5151 * 0005 SNAP 5152 * 0805 X.25 5153 * 0806 ARP 5154 * 8137 IPX 5155 * 0009 Localtalk 5156 * 86DD IPv6 5157 */ 5158 #define PTYPE_HASH_SIZE (16) 5159 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5160 5161 extern struct list_head ptype_all __read_mostly; 5162 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5163 5164 extern struct net_device *blackhole_netdev; 5165 5166 #endif /* _LINUX_NETDEVICE_H */ 5167