1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <[email protected]> 13 * Corey Minyard <[email protected]> 14 * Donald J. Becker, <[email protected]> 15 * Alan Cox, <[email protected]> 16 * Bjorn Ekwall. <[email protected]> 17 * Pekka Riikonen <[email protected]> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 #include <asm/local.h> 32 33 #include <linux/percpu.h> 34 #include <linux/rculist.h> 35 #include <linux/workqueue.h> 36 #include <linux/dynamic_queue_limits.h> 37 38 #include <net/net_namespace.h> 39 #ifdef CONFIG_DCB 40 #include <net/dcbnl.h> 41 #endif 42 #include <net/netprio_cgroup.h> 43 #include <net/xdp.h> 44 45 #include <linux/netdev_features.h> 46 #include <linux/neighbour.h> 47 #include <uapi/linux/netdevice.h> 48 #include <uapi/linux/if_bonding.h> 49 #include <uapi/linux/pkt_cls.h> 50 #include <linux/hashtable.h> 51 #include <linux/rbtree.h> 52 #include <net/net_trackers.h> 53 #include <net/net_debug.h> 54 55 struct netpoll_info; 56 struct device; 57 struct ethtool_ops; 58 struct phy_device; 59 struct dsa_port; 60 struct ip_tunnel_parm; 61 struct macsec_context; 62 struct macsec_ops; 63 struct netdev_name_node; 64 struct sd_flow_limit; 65 struct sfp_bus; 66 /* 802.11 specific */ 67 struct wireless_dev; 68 /* 802.15.4 specific */ 69 struct wpan_dev; 70 struct mpls_dev; 71 /* UDP Tunnel offloads */ 72 struct udp_tunnel_info; 73 struct udp_tunnel_nic_info; 74 struct udp_tunnel_nic; 75 struct bpf_prog; 76 struct xdp_buff; 77 78 void synchronize_net(void); 79 void netdev_set_default_ethtool_ops(struct net_device *dev, 80 const struct ethtool_ops *ops); 81 void netdev_sw_irq_coalesce_default_on(struct net_device *dev); 82 83 /* Backlog congestion levels */ 84 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 85 #define NET_RX_DROP 1 /* packet dropped */ 86 87 #define MAX_NEST_DEV 8 88 89 /* 90 * Transmit return codes: transmit return codes originate from three different 91 * namespaces: 92 * 93 * - qdisc return codes 94 * - driver transmit return codes 95 * - errno values 96 * 97 * Drivers are allowed to return any one of those in their hard_start_xmit() 98 * function. Real network devices commonly used with qdiscs should only return 99 * the driver transmit return codes though - when qdiscs are used, the actual 100 * transmission happens asynchronously, so the value is not propagated to 101 * higher layers. Virtual network devices transmit synchronously; in this case 102 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 103 * others are propagated to higher layers. 104 */ 105 106 /* qdisc ->enqueue() return codes. */ 107 #define NET_XMIT_SUCCESS 0x00 108 #define NET_XMIT_DROP 0x01 /* skb dropped */ 109 #define NET_XMIT_CN 0x02 /* congestion notification */ 110 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 111 112 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 113 * indicates that the device will soon be dropping packets, or already drops 114 * some packets of the same priority; prompting us to send less aggressively. */ 115 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 116 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 117 118 /* Driver transmit return codes */ 119 #define NETDEV_TX_MASK 0xf0 120 121 enum netdev_tx { 122 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 123 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 124 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 125 }; 126 typedef enum netdev_tx netdev_tx_t; 127 128 /* 129 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 130 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 131 */ 132 static inline bool dev_xmit_complete(int rc) 133 { 134 /* 135 * Positive cases with an skb consumed by a driver: 136 * - successful transmission (rc == NETDEV_TX_OK) 137 * - error while transmitting (rc < 0) 138 * - error while queueing to a different device (rc & NET_XMIT_MASK) 139 */ 140 if (likely(rc < NET_XMIT_MASK)) 141 return true; 142 143 return false; 144 } 145 146 /* 147 * Compute the worst-case header length according to the protocols 148 * used. 149 */ 150 151 #if defined(CONFIG_HYPERV_NET) 152 # define LL_MAX_HEADER 128 153 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 154 # if defined(CONFIG_MAC80211_MESH) 155 # define LL_MAX_HEADER 128 156 # else 157 # define LL_MAX_HEADER 96 158 # endif 159 #else 160 # define LL_MAX_HEADER 32 161 #endif 162 163 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 164 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 165 #define MAX_HEADER LL_MAX_HEADER 166 #else 167 #define MAX_HEADER (LL_MAX_HEADER + 48) 168 #endif 169 170 /* 171 * Old network device statistics. Fields are native words 172 * (unsigned long) so they can be read and written atomically. 173 */ 174 175 #define NET_DEV_STAT(FIELD) \ 176 union { \ 177 unsigned long FIELD; \ 178 atomic_long_t __##FIELD; \ 179 } 180 181 struct net_device_stats { 182 NET_DEV_STAT(rx_packets); 183 NET_DEV_STAT(tx_packets); 184 NET_DEV_STAT(rx_bytes); 185 NET_DEV_STAT(tx_bytes); 186 NET_DEV_STAT(rx_errors); 187 NET_DEV_STAT(tx_errors); 188 NET_DEV_STAT(rx_dropped); 189 NET_DEV_STAT(tx_dropped); 190 NET_DEV_STAT(multicast); 191 NET_DEV_STAT(collisions); 192 NET_DEV_STAT(rx_length_errors); 193 NET_DEV_STAT(rx_over_errors); 194 NET_DEV_STAT(rx_crc_errors); 195 NET_DEV_STAT(rx_frame_errors); 196 NET_DEV_STAT(rx_fifo_errors); 197 NET_DEV_STAT(rx_missed_errors); 198 NET_DEV_STAT(tx_aborted_errors); 199 NET_DEV_STAT(tx_carrier_errors); 200 NET_DEV_STAT(tx_fifo_errors); 201 NET_DEV_STAT(tx_heartbeat_errors); 202 NET_DEV_STAT(tx_window_errors); 203 NET_DEV_STAT(rx_compressed); 204 NET_DEV_STAT(tx_compressed); 205 }; 206 #undef NET_DEV_STAT 207 208 /* per-cpu stats, allocated on demand. 209 * Try to fit them in a single cache line, for dev_get_stats() sake. 210 */ 211 struct net_device_core_stats { 212 unsigned long rx_dropped; 213 unsigned long tx_dropped; 214 unsigned long rx_nohandler; 215 unsigned long rx_otherhost_dropped; 216 } __aligned(4 * sizeof(unsigned long)); 217 218 #include <linux/cache.h> 219 #include <linux/skbuff.h> 220 221 #ifdef CONFIG_RPS 222 #include <linux/static_key.h> 223 extern struct static_key_false rps_needed; 224 extern struct static_key_false rfs_needed; 225 #endif 226 227 struct neighbour; 228 struct neigh_parms; 229 struct sk_buff; 230 231 struct netdev_hw_addr { 232 struct list_head list; 233 struct rb_node node; 234 unsigned char addr[MAX_ADDR_LEN]; 235 unsigned char type; 236 #define NETDEV_HW_ADDR_T_LAN 1 237 #define NETDEV_HW_ADDR_T_SAN 2 238 #define NETDEV_HW_ADDR_T_UNICAST 3 239 #define NETDEV_HW_ADDR_T_MULTICAST 4 240 bool global_use; 241 int sync_cnt; 242 int refcount; 243 int synced; 244 struct rcu_head rcu_head; 245 }; 246 247 struct netdev_hw_addr_list { 248 struct list_head list; 249 int count; 250 251 /* Auxiliary tree for faster lookup on addition and deletion */ 252 struct rb_root tree; 253 }; 254 255 #define netdev_hw_addr_list_count(l) ((l)->count) 256 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 257 #define netdev_hw_addr_list_for_each(ha, l) \ 258 list_for_each_entry(ha, &(l)->list, list) 259 260 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 261 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 262 #define netdev_for_each_uc_addr(ha, dev) \ 263 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 264 #define netdev_for_each_synced_uc_addr(_ha, _dev) \ 265 netdev_for_each_uc_addr((_ha), (_dev)) \ 266 if ((_ha)->sync_cnt) 267 268 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 269 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 270 #define netdev_for_each_mc_addr(ha, dev) \ 271 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 272 #define netdev_for_each_synced_mc_addr(_ha, _dev) \ 273 netdev_for_each_mc_addr((_ha), (_dev)) \ 274 if ((_ha)->sync_cnt) 275 276 struct hh_cache { 277 unsigned int hh_len; 278 seqlock_t hh_lock; 279 280 /* cached hardware header; allow for machine alignment needs. */ 281 #define HH_DATA_MOD 16 282 #define HH_DATA_OFF(__len) \ 283 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 284 #define HH_DATA_ALIGN(__len) \ 285 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 286 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 287 }; 288 289 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 290 * Alternative is: 291 * dev->hard_header_len ? (dev->hard_header_len + 292 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 293 * 294 * We could use other alignment values, but we must maintain the 295 * relationship HH alignment <= LL alignment. 296 */ 297 #define LL_RESERVED_SPACE(dev) \ 298 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 299 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 300 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 301 302 struct header_ops { 303 int (*create) (struct sk_buff *skb, struct net_device *dev, 304 unsigned short type, const void *daddr, 305 const void *saddr, unsigned int len); 306 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 307 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 308 void (*cache_update)(struct hh_cache *hh, 309 const struct net_device *dev, 310 const unsigned char *haddr); 311 bool (*validate)(const char *ll_header, unsigned int len); 312 __be16 (*parse_protocol)(const struct sk_buff *skb); 313 }; 314 315 /* These flag bits are private to the generic network queueing 316 * layer; they may not be explicitly referenced by any other 317 * code. 318 */ 319 320 enum netdev_state_t { 321 __LINK_STATE_START, 322 __LINK_STATE_PRESENT, 323 __LINK_STATE_NOCARRIER, 324 __LINK_STATE_LINKWATCH_PENDING, 325 __LINK_STATE_DORMANT, 326 __LINK_STATE_TESTING, 327 }; 328 329 struct gro_list { 330 struct list_head list; 331 int count; 332 }; 333 334 /* 335 * size of gro hash buckets, must less than bit number of 336 * napi_struct::gro_bitmask 337 */ 338 #define GRO_HASH_BUCKETS 8 339 340 /* 341 * Structure for NAPI scheduling similar to tasklet but with weighting 342 */ 343 struct napi_struct { 344 /* The poll_list must only be managed by the entity which 345 * changes the state of the NAPI_STATE_SCHED bit. This means 346 * whoever atomically sets that bit can add this napi_struct 347 * to the per-CPU poll_list, and whoever clears that bit 348 * can remove from the list right before clearing the bit. 349 */ 350 struct list_head poll_list; 351 352 unsigned long state; 353 int weight; 354 int defer_hard_irqs_count; 355 unsigned long gro_bitmask; 356 int (*poll)(struct napi_struct *, int); 357 #ifdef CONFIG_NETPOLL 358 int poll_owner; 359 #endif 360 struct net_device *dev; 361 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 362 struct sk_buff *skb; 363 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 364 int rx_count; /* length of rx_list */ 365 struct hrtimer timer; 366 struct list_head dev_list; 367 struct hlist_node napi_hash_node; 368 unsigned int napi_id; 369 struct task_struct *thread; 370 }; 371 372 enum { 373 NAPI_STATE_SCHED, /* Poll is scheduled */ 374 NAPI_STATE_MISSED, /* reschedule a napi */ 375 NAPI_STATE_DISABLE, /* Disable pending */ 376 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 377 NAPI_STATE_LISTED, /* NAPI added to system lists */ 378 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 379 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 380 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 381 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 382 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 383 }; 384 385 enum { 386 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 387 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 388 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 389 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 390 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 391 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 392 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 393 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 394 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 395 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 396 }; 397 398 enum gro_result { 399 GRO_MERGED, 400 GRO_MERGED_FREE, 401 GRO_HELD, 402 GRO_NORMAL, 403 GRO_CONSUMED, 404 }; 405 typedef enum gro_result gro_result_t; 406 407 /* 408 * enum rx_handler_result - Possible return values for rx_handlers. 409 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 410 * further. 411 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 412 * case skb->dev was changed by rx_handler. 413 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 414 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 415 * 416 * rx_handlers are functions called from inside __netif_receive_skb(), to do 417 * special processing of the skb, prior to delivery to protocol handlers. 418 * 419 * Currently, a net_device can only have a single rx_handler registered. Trying 420 * to register a second rx_handler will return -EBUSY. 421 * 422 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 423 * To unregister a rx_handler on a net_device, use 424 * netdev_rx_handler_unregister(). 425 * 426 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 427 * do with the skb. 428 * 429 * If the rx_handler consumed the skb in some way, it should return 430 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 431 * the skb to be delivered in some other way. 432 * 433 * If the rx_handler changed skb->dev, to divert the skb to another 434 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 435 * new device will be called if it exists. 436 * 437 * If the rx_handler decides the skb should be ignored, it should return 438 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 439 * are registered on exact device (ptype->dev == skb->dev). 440 * 441 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 442 * delivered, it should return RX_HANDLER_PASS. 443 * 444 * A device without a registered rx_handler will behave as if rx_handler 445 * returned RX_HANDLER_PASS. 446 */ 447 448 enum rx_handler_result { 449 RX_HANDLER_CONSUMED, 450 RX_HANDLER_ANOTHER, 451 RX_HANDLER_EXACT, 452 RX_HANDLER_PASS, 453 }; 454 typedef enum rx_handler_result rx_handler_result_t; 455 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 456 457 void __napi_schedule(struct napi_struct *n); 458 void __napi_schedule_irqoff(struct napi_struct *n); 459 460 static inline bool napi_disable_pending(struct napi_struct *n) 461 { 462 return test_bit(NAPI_STATE_DISABLE, &n->state); 463 } 464 465 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 466 { 467 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 468 } 469 470 bool napi_schedule_prep(struct napi_struct *n); 471 472 /** 473 * napi_schedule - schedule NAPI poll 474 * @n: NAPI context 475 * 476 * Schedule NAPI poll routine to be called if it is not already 477 * running. 478 */ 479 static inline void napi_schedule(struct napi_struct *n) 480 { 481 if (napi_schedule_prep(n)) 482 __napi_schedule(n); 483 } 484 485 /** 486 * napi_schedule_irqoff - schedule NAPI poll 487 * @n: NAPI context 488 * 489 * Variant of napi_schedule(), assuming hard irqs are masked. 490 */ 491 static inline void napi_schedule_irqoff(struct napi_struct *n) 492 { 493 if (napi_schedule_prep(n)) 494 __napi_schedule_irqoff(n); 495 } 496 497 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 498 static inline bool napi_reschedule(struct napi_struct *napi) 499 { 500 if (napi_schedule_prep(napi)) { 501 __napi_schedule(napi); 502 return true; 503 } 504 return false; 505 } 506 507 bool napi_complete_done(struct napi_struct *n, int work_done); 508 /** 509 * napi_complete - NAPI processing complete 510 * @n: NAPI context 511 * 512 * Mark NAPI processing as complete. 513 * Consider using napi_complete_done() instead. 514 * Return false if device should avoid rearming interrupts. 515 */ 516 static inline bool napi_complete(struct napi_struct *n) 517 { 518 return napi_complete_done(n, 0); 519 } 520 521 int dev_set_threaded(struct net_device *dev, bool threaded); 522 523 /** 524 * napi_disable - prevent NAPI from scheduling 525 * @n: NAPI context 526 * 527 * Stop NAPI from being scheduled on this context. 528 * Waits till any outstanding processing completes. 529 */ 530 void napi_disable(struct napi_struct *n); 531 532 void napi_enable(struct napi_struct *n); 533 534 /** 535 * napi_synchronize - wait until NAPI is not running 536 * @n: NAPI context 537 * 538 * Wait until NAPI is done being scheduled on this context. 539 * Waits till any outstanding processing completes but 540 * does not disable future activations. 541 */ 542 static inline void napi_synchronize(const struct napi_struct *n) 543 { 544 if (IS_ENABLED(CONFIG_SMP)) 545 while (test_bit(NAPI_STATE_SCHED, &n->state)) 546 msleep(1); 547 else 548 barrier(); 549 } 550 551 /** 552 * napi_if_scheduled_mark_missed - if napi is running, set the 553 * NAPIF_STATE_MISSED 554 * @n: NAPI context 555 * 556 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 557 * NAPI is scheduled. 558 **/ 559 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 560 { 561 unsigned long val, new; 562 563 val = READ_ONCE(n->state); 564 do { 565 if (val & NAPIF_STATE_DISABLE) 566 return true; 567 568 if (!(val & NAPIF_STATE_SCHED)) 569 return false; 570 571 new = val | NAPIF_STATE_MISSED; 572 } while (!try_cmpxchg(&n->state, &val, new)); 573 574 return true; 575 } 576 577 enum netdev_queue_state_t { 578 __QUEUE_STATE_DRV_XOFF, 579 __QUEUE_STATE_STACK_XOFF, 580 __QUEUE_STATE_FROZEN, 581 }; 582 583 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 584 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 585 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 586 587 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 588 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 589 QUEUE_STATE_FROZEN) 590 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 591 QUEUE_STATE_FROZEN) 592 593 /* 594 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 595 * netif_tx_* functions below are used to manipulate this flag. The 596 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 597 * queue independently. The netif_xmit_*stopped functions below are called 598 * to check if the queue has been stopped by the driver or stack (either 599 * of the XOFF bits are set in the state). Drivers should not need to call 600 * netif_xmit*stopped functions, they should only be using netif_tx_*. 601 */ 602 603 struct netdev_queue { 604 /* 605 * read-mostly part 606 */ 607 struct net_device *dev; 608 netdevice_tracker dev_tracker; 609 610 struct Qdisc __rcu *qdisc; 611 struct Qdisc *qdisc_sleeping; 612 #ifdef CONFIG_SYSFS 613 struct kobject kobj; 614 #endif 615 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 616 int numa_node; 617 #endif 618 unsigned long tx_maxrate; 619 /* 620 * Number of TX timeouts for this queue 621 * (/sys/class/net/DEV/Q/trans_timeout) 622 */ 623 atomic_long_t trans_timeout; 624 625 /* Subordinate device that the queue has been assigned to */ 626 struct net_device *sb_dev; 627 #ifdef CONFIG_XDP_SOCKETS 628 struct xsk_buff_pool *pool; 629 #endif 630 /* 631 * write-mostly part 632 */ 633 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 634 int xmit_lock_owner; 635 /* 636 * Time (in jiffies) of last Tx 637 */ 638 unsigned long trans_start; 639 640 unsigned long state; 641 642 #ifdef CONFIG_BQL 643 struct dql dql; 644 #endif 645 } ____cacheline_aligned_in_smp; 646 647 extern int sysctl_fb_tunnels_only_for_init_net; 648 extern int sysctl_devconf_inherit_init_net; 649 650 /* 651 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 652 * == 1 : For initns only 653 * == 2 : For none. 654 */ 655 static inline bool net_has_fallback_tunnels(const struct net *net) 656 { 657 #if IS_ENABLED(CONFIG_SYSCTL) 658 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net); 659 660 return !fb_tunnels_only_for_init_net || 661 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1); 662 #else 663 return true; 664 #endif 665 } 666 667 static inline int net_inherit_devconf(void) 668 { 669 #if IS_ENABLED(CONFIG_SYSCTL) 670 return READ_ONCE(sysctl_devconf_inherit_init_net); 671 #else 672 return 0; 673 #endif 674 } 675 676 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 677 { 678 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 679 return q->numa_node; 680 #else 681 return NUMA_NO_NODE; 682 #endif 683 } 684 685 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 686 { 687 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 688 q->numa_node = node; 689 #endif 690 } 691 692 #ifdef CONFIG_RPS 693 /* 694 * This structure holds an RPS map which can be of variable length. The 695 * map is an array of CPUs. 696 */ 697 struct rps_map { 698 unsigned int len; 699 struct rcu_head rcu; 700 u16 cpus[]; 701 }; 702 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 703 704 /* 705 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 706 * tail pointer for that CPU's input queue at the time of last enqueue, and 707 * a hardware filter index. 708 */ 709 struct rps_dev_flow { 710 u16 cpu; 711 u16 filter; 712 unsigned int last_qtail; 713 }; 714 #define RPS_NO_FILTER 0xffff 715 716 /* 717 * The rps_dev_flow_table structure contains a table of flow mappings. 718 */ 719 struct rps_dev_flow_table { 720 unsigned int mask; 721 struct rcu_head rcu; 722 struct rps_dev_flow flows[]; 723 }; 724 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 725 ((_num) * sizeof(struct rps_dev_flow))) 726 727 /* 728 * The rps_sock_flow_table contains mappings of flows to the last CPU 729 * on which they were processed by the application (set in recvmsg). 730 * Each entry is a 32bit value. Upper part is the high-order bits 731 * of flow hash, lower part is CPU number. 732 * rps_cpu_mask is used to partition the space, depending on number of 733 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 734 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 735 * meaning we use 32-6=26 bits for the hash. 736 */ 737 struct rps_sock_flow_table { 738 u32 mask; 739 740 u32 ents[] ____cacheline_aligned_in_smp; 741 }; 742 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 743 744 #define RPS_NO_CPU 0xffff 745 746 extern u32 rps_cpu_mask; 747 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 748 749 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 750 u32 hash) 751 { 752 if (table && hash) { 753 unsigned int index = hash & table->mask; 754 u32 val = hash & ~rps_cpu_mask; 755 756 /* We only give a hint, preemption can change CPU under us */ 757 val |= raw_smp_processor_id(); 758 759 if (table->ents[index] != val) 760 table->ents[index] = val; 761 } 762 } 763 764 #ifdef CONFIG_RFS_ACCEL 765 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 766 u16 filter_id); 767 #endif 768 #endif /* CONFIG_RPS */ 769 770 /* This structure contains an instance of an RX queue. */ 771 struct netdev_rx_queue { 772 struct xdp_rxq_info xdp_rxq; 773 #ifdef CONFIG_RPS 774 struct rps_map __rcu *rps_map; 775 struct rps_dev_flow_table __rcu *rps_flow_table; 776 #endif 777 struct kobject kobj; 778 struct net_device *dev; 779 netdevice_tracker dev_tracker; 780 781 #ifdef CONFIG_XDP_SOCKETS 782 struct xsk_buff_pool *pool; 783 #endif 784 } ____cacheline_aligned_in_smp; 785 786 /* 787 * RX queue sysfs structures and functions. 788 */ 789 struct rx_queue_attribute { 790 struct attribute attr; 791 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 792 ssize_t (*store)(struct netdev_rx_queue *queue, 793 const char *buf, size_t len); 794 }; 795 796 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 797 enum xps_map_type { 798 XPS_CPUS = 0, 799 XPS_RXQS, 800 XPS_MAPS_MAX, 801 }; 802 803 #ifdef CONFIG_XPS 804 /* 805 * This structure holds an XPS map which can be of variable length. The 806 * map is an array of queues. 807 */ 808 struct xps_map { 809 unsigned int len; 810 unsigned int alloc_len; 811 struct rcu_head rcu; 812 u16 queues[]; 813 }; 814 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 815 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 816 - sizeof(struct xps_map)) / sizeof(u16)) 817 818 /* 819 * This structure holds all XPS maps for device. Maps are indexed by CPU. 820 * 821 * We keep track of the number of cpus/rxqs used when the struct is allocated, 822 * in nr_ids. This will help not accessing out-of-bound memory. 823 * 824 * We keep track of the number of traffic classes used when the struct is 825 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 826 * not crossing its upper bound, as the original dev->num_tc can be updated in 827 * the meantime. 828 */ 829 struct xps_dev_maps { 830 struct rcu_head rcu; 831 unsigned int nr_ids; 832 s16 num_tc; 833 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 834 }; 835 836 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 837 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 838 839 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 840 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 841 842 #endif /* CONFIG_XPS */ 843 844 #define TC_MAX_QUEUE 16 845 #define TC_BITMASK 15 846 /* HW offloaded queuing disciplines txq count and offset maps */ 847 struct netdev_tc_txq { 848 u16 count; 849 u16 offset; 850 }; 851 852 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 853 /* 854 * This structure is to hold information about the device 855 * configured to run FCoE protocol stack. 856 */ 857 struct netdev_fcoe_hbainfo { 858 char manufacturer[64]; 859 char serial_number[64]; 860 char hardware_version[64]; 861 char driver_version[64]; 862 char optionrom_version[64]; 863 char firmware_version[64]; 864 char model[256]; 865 char model_description[256]; 866 }; 867 #endif 868 869 #define MAX_PHYS_ITEM_ID_LEN 32 870 871 /* This structure holds a unique identifier to identify some 872 * physical item (port for example) used by a netdevice. 873 */ 874 struct netdev_phys_item_id { 875 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 876 unsigned char id_len; 877 }; 878 879 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 880 struct netdev_phys_item_id *b) 881 { 882 return a->id_len == b->id_len && 883 memcmp(a->id, b->id, a->id_len) == 0; 884 } 885 886 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 887 struct sk_buff *skb, 888 struct net_device *sb_dev); 889 890 enum net_device_path_type { 891 DEV_PATH_ETHERNET = 0, 892 DEV_PATH_VLAN, 893 DEV_PATH_BRIDGE, 894 DEV_PATH_PPPOE, 895 DEV_PATH_DSA, 896 DEV_PATH_MTK_WDMA, 897 }; 898 899 struct net_device_path { 900 enum net_device_path_type type; 901 const struct net_device *dev; 902 union { 903 struct { 904 u16 id; 905 __be16 proto; 906 u8 h_dest[ETH_ALEN]; 907 } encap; 908 struct { 909 enum { 910 DEV_PATH_BR_VLAN_KEEP, 911 DEV_PATH_BR_VLAN_TAG, 912 DEV_PATH_BR_VLAN_UNTAG, 913 DEV_PATH_BR_VLAN_UNTAG_HW, 914 } vlan_mode; 915 u16 vlan_id; 916 __be16 vlan_proto; 917 } bridge; 918 struct { 919 int port; 920 u16 proto; 921 } dsa; 922 struct { 923 u8 wdma_idx; 924 u8 queue; 925 u16 wcid; 926 u8 bss; 927 } mtk_wdma; 928 }; 929 }; 930 931 #define NET_DEVICE_PATH_STACK_MAX 5 932 #define NET_DEVICE_PATH_VLAN_MAX 2 933 934 struct net_device_path_stack { 935 int num_paths; 936 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 937 }; 938 939 struct net_device_path_ctx { 940 const struct net_device *dev; 941 u8 daddr[ETH_ALEN]; 942 943 int num_vlans; 944 struct { 945 u16 id; 946 __be16 proto; 947 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 948 }; 949 950 enum tc_setup_type { 951 TC_QUERY_CAPS, 952 TC_SETUP_QDISC_MQPRIO, 953 TC_SETUP_CLSU32, 954 TC_SETUP_CLSFLOWER, 955 TC_SETUP_CLSMATCHALL, 956 TC_SETUP_CLSBPF, 957 TC_SETUP_BLOCK, 958 TC_SETUP_QDISC_CBS, 959 TC_SETUP_QDISC_RED, 960 TC_SETUP_QDISC_PRIO, 961 TC_SETUP_QDISC_MQ, 962 TC_SETUP_QDISC_ETF, 963 TC_SETUP_ROOT_QDISC, 964 TC_SETUP_QDISC_GRED, 965 TC_SETUP_QDISC_TAPRIO, 966 TC_SETUP_FT, 967 TC_SETUP_QDISC_ETS, 968 TC_SETUP_QDISC_TBF, 969 TC_SETUP_QDISC_FIFO, 970 TC_SETUP_QDISC_HTB, 971 TC_SETUP_ACT, 972 }; 973 974 /* These structures hold the attributes of bpf state that are being passed 975 * to the netdevice through the bpf op. 976 */ 977 enum bpf_netdev_command { 978 /* Set or clear a bpf program used in the earliest stages of packet 979 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 980 * is responsible for calling bpf_prog_put on any old progs that are 981 * stored. In case of error, the callee need not release the new prog 982 * reference, but on success it takes ownership and must bpf_prog_put 983 * when it is no longer used. 984 */ 985 XDP_SETUP_PROG, 986 XDP_SETUP_PROG_HW, 987 /* BPF program for offload callbacks, invoked at program load time. */ 988 BPF_OFFLOAD_MAP_ALLOC, 989 BPF_OFFLOAD_MAP_FREE, 990 XDP_SETUP_XSK_POOL, 991 }; 992 993 struct bpf_prog_offload_ops; 994 struct netlink_ext_ack; 995 struct xdp_umem; 996 struct xdp_dev_bulk_queue; 997 struct bpf_xdp_link; 998 999 enum bpf_xdp_mode { 1000 XDP_MODE_SKB = 0, 1001 XDP_MODE_DRV = 1, 1002 XDP_MODE_HW = 2, 1003 __MAX_XDP_MODE 1004 }; 1005 1006 struct bpf_xdp_entity { 1007 struct bpf_prog *prog; 1008 struct bpf_xdp_link *link; 1009 }; 1010 1011 struct netdev_bpf { 1012 enum bpf_netdev_command command; 1013 union { 1014 /* XDP_SETUP_PROG */ 1015 struct { 1016 u32 flags; 1017 struct bpf_prog *prog; 1018 struct netlink_ext_ack *extack; 1019 }; 1020 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 1021 struct { 1022 struct bpf_offloaded_map *offmap; 1023 }; 1024 /* XDP_SETUP_XSK_POOL */ 1025 struct { 1026 struct xsk_buff_pool *pool; 1027 u16 queue_id; 1028 } xsk; 1029 }; 1030 }; 1031 1032 /* Flags for ndo_xsk_wakeup. */ 1033 #define XDP_WAKEUP_RX (1 << 0) 1034 #define XDP_WAKEUP_TX (1 << 1) 1035 1036 #ifdef CONFIG_XFRM_OFFLOAD 1037 struct xfrmdev_ops { 1038 int (*xdo_dev_state_add) (struct xfrm_state *x); 1039 void (*xdo_dev_state_delete) (struct xfrm_state *x); 1040 void (*xdo_dev_state_free) (struct xfrm_state *x); 1041 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 1042 struct xfrm_state *x); 1043 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1044 }; 1045 #endif 1046 1047 struct dev_ifalias { 1048 struct rcu_head rcuhead; 1049 char ifalias[]; 1050 }; 1051 1052 struct devlink; 1053 struct tlsdev_ops; 1054 1055 struct netdev_net_notifier { 1056 struct list_head list; 1057 struct notifier_block *nb; 1058 }; 1059 1060 /* 1061 * This structure defines the management hooks for network devices. 1062 * The following hooks can be defined; unless noted otherwise, they are 1063 * optional and can be filled with a null pointer. 1064 * 1065 * int (*ndo_init)(struct net_device *dev); 1066 * This function is called once when a network device is registered. 1067 * The network device can use this for any late stage initialization 1068 * or semantic validation. It can fail with an error code which will 1069 * be propagated back to register_netdev. 1070 * 1071 * void (*ndo_uninit)(struct net_device *dev); 1072 * This function is called when device is unregistered or when registration 1073 * fails. It is not called if init fails. 1074 * 1075 * int (*ndo_open)(struct net_device *dev); 1076 * This function is called when a network device transitions to the up 1077 * state. 1078 * 1079 * int (*ndo_stop)(struct net_device *dev); 1080 * This function is called when a network device transitions to the down 1081 * state. 1082 * 1083 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1084 * struct net_device *dev); 1085 * Called when a packet needs to be transmitted. 1086 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1087 * the queue before that can happen; it's for obsolete devices and weird 1088 * corner cases, but the stack really does a non-trivial amount 1089 * of useless work if you return NETDEV_TX_BUSY. 1090 * Required; cannot be NULL. 1091 * 1092 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1093 * struct net_device *dev 1094 * netdev_features_t features); 1095 * Called by core transmit path to determine if device is capable of 1096 * performing offload operations on a given packet. This is to give 1097 * the device an opportunity to implement any restrictions that cannot 1098 * be otherwise expressed by feature flags. The check is called with 1099 * the set of features that the stack has calculated and it returns 1100 * those the driver believes to be appropriate. 1101 * 1102 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1103 * struct net_device *sb_dev); 1104 * Called to decide which queue to use when device supports multiple 1105 * transmit queues. 1106 * 1107 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1108 * This function is called to allow device receiver to make 1109 * changes to configuration when multicast or promiscuous is enabled. 1110 * 1111 * void (*ndo_set_rx_mode)(struct net_device *dev); 1112 * This function is called device changes address list filtering. 1113 * If driver handles unicast address filtering, it should set 1114 * IFF_UNICAST_FLT in its priv_flags. 1115 * 1116 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1117 * This function is called when the Media Access Control address 1118 * needs to be changed. If this interface is not defined, the 1119 * MAC address can not be changed. 1120 * 1121 * int (*ndo_validate_addr)(struct net_device *dev); 1122 * Test if Media Access Control address is valid for the device. 1123 * 1124 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1125 * Old-style ioctl entry point. This is used internally by the 1126 * appletalk and ieee802154 subsystems but is no longer called by 1127 * the device ioctl handler. 1128 * 1129 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1130 * Used by the bonding driver for its device specific ioctls: 1131 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1132 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1133 * 1134 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1135 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1136 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1137 * 1138 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1139 * Used to set network devices bus interface parameters. This interface 1140 * is retained for legacy reasons; new devices should use the bus 1141 * interface (PCI) for low level management. 1142 * 1143 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1144 * Called when a user wants to change the Maximum Transfer Unit 1145 * of a device. 1146 * 1147 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1148 * Callback used when the transmitter has not made any progress 1149 * for dev->watchdog ticks. 1150 * 1151 * void (*ndo_get_stats64)(struct net_device *dev, 1152 * struct rtnl_link_stats64 *storage); 1153 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1154 * Called when a user wants to get the network device usage 1155 * statistics. Drivers must do one of the following: 1156 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1157 * rtnl_link_stats64 structure passed by the caller. 1158 * 2. Define @ndo_get_stats to update a net_device_stats structure 1159 * (which should normally be dev->stats) and return a pointer to 1160 * it. The structure may be changed asynchronously only if each 1161 * field is written atomically. 1162 * 3. Update dev->stats asynchronously and atomically, and define 1163 * neither operation. 1164 * 1165 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1166 * Return true if this device supports offload stats of this attr_id. 1167 * 1168 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1169 * void *attr_data) 1170 * Get statistics for offload operations by attr_id. Write it into the 1171 * attr_data pointer. 1172 * 1173 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1174 * If device supports VLAN filtering this function is called when a 1175 * VLAN id is registered. 1176 * 1177 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1178 * If device supports VLAN filtering this function is called when a 1179 * VLAN id is unregistered. 1180 * 1181 * void (*ndo_poll_controller)(struct net_device *dev); 1182 * 1183 * SR-IOV management functions. 1184 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1185 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1186 * u8 qos, __be16 proto); 1187 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1188 * int max_tx_rate); 1189 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1190 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1191 * int (*ndo_get_vf_config)(struct net_device *dev, 1192 * int vf, struct ifla_vf_info *ivf); 1193 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1194 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1195 * struct nlattr *port[]); 1196 * 1197 * Enable or disable the VF ability to query its RSS Redirection Table and 1198 * Hash Key. This is needed since on some devices VF share this information 1199 * with PF and querying it may introduce a theoretical security risk. 1200 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1201 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1202 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1203 * void *type_data); 1204 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1205 * This is always called from the stack with the rtnl lock held and netif 1206 * tx queues stopped. This allows the netdevice to perform queue 1207 * management safely. 1208 * 1209 * Fiber Channel over Ethernet (FCoE) offload functions. 1210 * int (*ndo_fcoe_enable)(struct net_device *dev); 1211 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1212 * so the underlying device can perform whatever needed configuration or 1213 * initialization to support acceleration of FCoE traffic. 1214 * 1215 * int (*ndo_fcoe_disable)(struct net_device *dev); 1216 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1217 * so the underlying device can perform whatever needed clean-ups to 1218 * stop supporting acceleration of FCoE traffic. 1219 * 1220 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1221 * struct scatterlist *sgl, unsigned int sgc); 1222 * Called when the FCoE Initiator wants to initialize an I/O that 1223 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1224 * perform necessary setup and returns 1 to indicate the device is set up 1225 * successfully to perform DDP on this I/O, otherwise this returns 0. 1226 * 1227 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1228 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1229 * indicated by the FC exchange id 'xid', so the underlying device can 1230 * clean up and reuse resources for later DDP requests. 1231 * 1232 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1233 * struct scatterlist *sgl, unsigned int sgc); 1234 * Called when the FCoE Target wants to initialize an I/O that 1235 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1236 * perform necessary setup and returns 1 to indicate the device is set up 1237 * successfully to perform DDP on this I/O, otherwise this returns 0. 1238 * 1239 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1240 * struct netdev_fcoe_hbainfo *hbainfo); 1241 * Called when the FCoE Protocol stack wants information on the underlying 1242 * device. This information is utilized by the FCoE protocol stack to 1243 * register attributes with Fiber Channel management service as per the 1244 * FC-GS Fabric Device Management Information(FDMI) specification. 1245 * 1246 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1247 * Called when the underlying device wants to override default World Wide 1248 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1249 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1250 * protocol stack to use. 1251 * 1252 * RFS acceleration. 1253 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1254 * u16 rxq_index, u32 flow_id); 1255 * Set hardware filter for RFS. rxq_index is the target queue index; 1256 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1257 * Return the filter ID on success, or a negative error code. 1258 * 1259 * Slave management functions (for bridge, bonding, etc). 1260 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1261 * Called to make another netdev an underling. 1262 * 1263 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1264 * Called to release previously enslaved netdev. 1265 * 1266 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1267 * struct sk_buff *skb, 1268 * bool all_slaves); 1269 * Get the xmit slave of master device. If all_slaves is true, function 1270 * assume all the slaves can transmit. 1271 * 1272 * Feature/offload setting functions. 1273 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1274 * netdev_features_t features); 1275 * Adjusts the requested feature flags according to device-specific 1276 * constraints, and returns the resulting flags. Must not modify 1277 * the device state. 1278 * 1279 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1280 * Called to update device configuration to new features. Passed 1281 * feature set might be less than what was returned by ndo_fix_features()). 1282 * Must return >0 or -errno if it changed dev->features itself. 1283 * 1284 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1285 * struct net_device *dev, 1286 * const unsigned char *addr, u16 vid, u16 flags, 1287 * struct netlink_ext_ack *extack); 1288 * Adds an FDB entry to dev for addr. 1289 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1290 * struct net_device *dev, 1291 * const unsigned char *addr, u16 vid) 1292 * Deletes the FDB entry from dev coresponding to addr. 1293 * int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, struct nlattr *tb[], 1294 * struct net_device *dev, 1295 * u16 vid, 1296 * struct netlink_ext_ack *extack); 1297 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1298 * struct net_device *dev, struct net_device *filter_dev, 1299 * int *idx) 1300 * Used to add FDB entries to dump requests. Implementers should add 1301 * entries to skb and update idx with the number of entries. 1302 * 1303 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1304 * u16 flags, struct netlink_ext_ack *extack) 1305 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1306 * struct net_device *dev, u32 filter_mask, 1307 * int nlflags) 1308 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1309 * u16 flags); 1310 * 1311 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1312 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1313 * which do not represent real hardware may define this to allow their 1314 * userspace components to manage their virtual carrier state. Devices 1315 * that determine carrier state from physical hardware properties (eg 1316 * network cables) or protocol-dependent mechanisms (eg 1317 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1318 * 1319 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1320 * struct netdev_phys_item_id *ppid); 1321 * Called to get ID of physical port of this device. If driver does 1322 * not implement this, it is assumed that the hw is not able to have 1323 * multiple net devices on single physical port. 1324 * 1325 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1326 * struct netdev_phys_item_id *ppid) 1327 * Called to get the parent ID of the physical port of this device. 1328 * 1329 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1330 * struct net_device *dev) 1331 * Called by upper layer devices to accelerate switching or other 1332 * station functionality into hardware. 'pdev is the lowerdev 1333 * to use for the offload and 'dev' is the net device that will 1334 * back the offload. Returns a pointer to the private structure 1335 * the upper layer will maintain. 1336 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1337 * Called by upper layer device to delete the station created 1338 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1339 * the station and priv is the structure returned by the add 1340 * operation. 1341 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1342 * int queue_index, u32 maxrate); 1343 * Called when a user wants to set a max-rate limitation of specific 1344 * TX queue. 1345 * int (*ndo_get_iflink)(const struct net_device *dev); 1346 * Called to get the iflink value of this device. 1347 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1348 * This function is used to get egress tunnel information for given skb. 1349 * This is useful for retrieving outer tunnel header parameters while 1350 * sampling packet. 1351 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1352 * This function is used to specify the headroom that the skb must 1353 * consider when allocation skb during packet reception. Setting 1354 * appropriate rx headroom value allows avoiding skb head copy on 1355 * forward. Setting a negative value resets the rx headroom to the 1356 * default value. 1357 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1358 * This function is used to set or query state related to XDP on the 1359 * netdevice and manage BPF offload. See definition of 1360 * enum bpf_netdev_command for details. 1361 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1362 * u32 flags); 1363 * This function is used to submit @n XDP packets for transmit on a 1364 * netdevice. Returns number of frames successfully transmitted, frames 1365 * that got dropped are freed/returned via xdp_return_frame(). 1366 * Returns negative number, means general error invoking ndo, meaning 1367 * no frames were xmit'ed and core-caller will free all frames. 1368 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1369 * struct xdp_buff *xdp); 1370 * Get the xmit slave of master device based on the xdp_buff. 1371 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1372 * This function is used to wake up the softirq, ksoftirqd or kthread 1373 * responsible for sending and/or receiving packets on a specific 1374 * queue id bound to an AF_XDP socket. The flags field specifies if 1375 * only RX, only Tx, or both should be woken up using the flags 1376 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1377 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1378 * int cmd); 1379 * Add, change, delete or get information on an IPv4 tunnel. 1380 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1381 * If a device is paired with a peer device, return the peer instance. 1382 * The caller must be under RCU read context. 1383 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1384 * Get the forwarding path to reach the real device from the HW destination address 1385 * ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1386 * const struct skb_shared_hwtstamps *hwtstamps, 1387 * bool cycles); 1388 * Get hardware timestamp based on normal/adjustable time or free running 1389 * cycle counter. This function is required if physical clock supports a 1390 * free running cycle counter. 1391 */ 1392 struct net_device_ops { 1393 int (*ndo_init)(struct net_device *dev); 1394 void (*ndo_uninit)(struct net_device *dev); 1395 int (*ndo_open)(struct net_device *dev); 1396 int (*ndo_stop)(struct net_device *dev); 1397 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1398 struct net_device *dev); 1399 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1400 struct net_device *dev, 1401 netdev_features_t features); 1402 u16 (*ndo_select_queue)(struct net_device *dev, 1403 struct sk_buff *skb, 1404 struct net_device *sb_dev); 1405 void (*ndo_change_rx_flags)(struct net_device *dev, 1406 int flags); 1407 void (*ndo_set_rx_mode)(struct net_device *dev); 1408 int (*ndo_set_mac_address)(struct net_device *dev, 1409 void *addr); 1410 int (*ndo_validate_addr)(struct net_device *dev); 1411 int (*ndo_do_ioctl)(struct net_device *dev, 1412 struct ifreq *ifr, int cmd); 1413 int (*ndo_eth_ioctl)(struct net_device *dev, 1414 struct ifreq *ifr, int cmd); 1415 int (*ndo_siocbond)(struct net_device *dev, 1416 struct ifreq *ifr, int cmd); 1417 int (*ndo_siocwandev)(struct net_device *dev, 1418 struct if_settings *ifs); 1419 int (*ndo_siocdevprivate)(struct net_device *dev, 1420 struct ifreq *ifr, 1421 void __user *data, int cmd); 1422 int (*ndo_set_config)(struct net_device *dev, 1423 struct ifmap *map); 1424 int (*ndo_change_mtu)(struct net_device *dev, 1425 int new_mtu); 1426 int (*ndo_neigh_setup)(struct net_device *dev, 1427 struct neigh_parms *); 1428 void (*ndo_tx_timeout) (struct net_device *dev, 1429 unsigned int txqueue); 1430 1431 void (*ndo_get_stats64)(struct net_device *dev, 1432 struct rtnl_link_stats64 *storage); 1433 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1434 int (*ndo_get_offload_stats)(int attr_id, 1435 const struct net_device *dev, 1436 void *attr_data); 1437 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1438 1439 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1440 __be16 proto, u16 vid); 1441 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1442 __be16 proto, u16 vid); 1443 #ifdef CONFIG_NET_POLL_CONTROLLER 1444 void (*ndo_poll_controller)(struct net_device *dev); 1445 int (*ndo_netpoll_setup)(struct net_device *dev, 1446 struct netpoll_info *info); 1447 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1448 #endif 1449 int (*ndo_set_vf_mac)(struct net_device *dev, 1450 int queue, u8 *mac); 1451 int (*ndo_set_vf_vlan)(struct net_device *dev, 1452 int queue, u16 vlan, 1453 u8 qos, __be16 proto); 1454 int (*ndo_set_vf_rate)(struct net_device *dev, 1455 int vf, int min_tx_rate, 1456 int max_tx_rate); 1457 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1458 int vf, bool setting); 1459 int (*ndo_set_vf_trust)(struct net_device *dev, 1460 int vf, bool setting); 1461 int (*ndo_get_vf_config)(struct net_device *dev, 1462 int vf, 1463 struct ifla_vf_info *ivf); 1464 int (*ndo_set_vf_link_state)(struct net_device *dev, 1465 int vf, int link_state); 1466 int (*ndo_get_vf_stats)(struct net_device *dev, 1467 int vf, 1468 struct ifla_vf_stats 1469 *vf_stats); 1470 int (*ndo_set_vf_port)(struct net_device *dev, 1471 int vf, 1472 struct nlattr *port[]); 1473 int (*ndo_get_vf_port)(struct net_device *dev, 1474 int vf, struct sk_buff *skb); 1475 int (*ndo_get_vf_guid)(struct net_device *dev, 1476 int vf, 1477 struct ifla_vf_guid *node_guid, 1478 struct ifla_vf_guid *port_guid); 1479 int (*ndo_set_vf_guid)(struct net_device *dev, 1480 int vf, u64 guid, 1481 int guid_type); 1482 int (*ndo_set_vf_rss_query_en)( 1483 struct net_device *dev, 1484 int vf, bool setting); 1485 int (*ndo_setup_tc)(struct net_device *dev, 1486 enum tc_setup_type type, 1487 void *type_data); 1488 #if IS_ENABLED(CONFIG_FCOE) 1489 int (*ndo_fcoe_enable)(struct net_device *dev); 1490 int (*ndo_fcoe_disable)(struct net_device *dev); 1491 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1492 u16 xid, 1493 struct scatterlist *sgl, 1494 unsigned int sgc); 1495 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1496 u16 xid); 1497 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1498 u16 xid, 1499 struct scatterlist *sgl, 1500 unsigned int sgc); 1501 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1502 struct netdev_fcoe_hbainfo *hbainfo); 1503 #endif 1504 1505 #if IS_ENABLED(CONFIG_LIBFCOE) 1506 #define NETDEV_FCOE_WWNN 0 1507 #define NETDEV_FCOE_WWPN 1 1508 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1509 u64 *wwn, int type); 1510 #endif 1511 1512 #ifdef CONFIG_RFS_ACCEL 1513 int (*ndo_rx_flow_steer)(struct net_device *dev, 1514 const struct sk_buff *skb, 1515 u16 rxq_index, 1516 u32 flow_id); 1517 #endif 1518 int (*ndo_add_slave)(struct net_device *dev, 1519 struct net_device *slave_dev, 1520 struct netlink_ext_ack *extack); 1521 int (*ndo_del_slave)(struct net_device *dev, 1522 struct net_device *slave_dev); 1523 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1524 struct sk_buff *skb, 1525 bool all_slaves); 1526 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1527 struct sock *sk); 1528 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1529 netdev_features_t features); 1530 int (*ndo_set_features)(struct net_device *dev, 1531 netdev_features_t features); 1532 int (*ndo_neigh_construct)(struct net_device *dev, 1533 struct neighbour *n); 1534 void (*ndo_neigh_destroy)(struct net_device *dev, 1535 struct neighbour *n); 1536 1537 int (*ndo_fdb_add)(struct ndmsg *ndm, 1538 struct nlattr *tb[], 1539 struct net_device *dev, 1540 const unsigned char *addr, 1541 u16 vid, 1542 u16 flags, 1543 struct netlink_ext_ack *extack); 1544 int (*ndo_fdb_del)(struct ndmsg *ndm, 1545 struct nlattr *tb[], 1546 struct net_device *dev, 1547 const unsigned char *addr, 1548 u16 vid, struct netlink_ext_ack *extack); 1549 int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, 1550 struct nlattr *tb[], 1551 struct net_device *dev, 1552 u16 vid, 1553 struct netlink_ext_ack *extack); 1554 int (*ndo_fdb_dump)(struct sk_buff *skb, 1555 struct netlink_callback *cb, 1556 struct net_device *dev, 1557 struct net_device *filter_dev, 1558 int *idx); 1559 int (*ndo_fdb_get)(struct sk_buff *skb, 1560 struct nlattr *tb[], 1561 struct net_device *dev, 1562 const unsigned char *addr, 1563 u16 vid, u32 portid, u32 seq, 1564 struct netlink_ext_ack *extack); 1565 int (*ndo_bridge_setlink)(struct net_device *dev, 1566 struct nlmsghdr *nlh, 1567 u16 flags, 1568 struct netlink_ext_ack *extack); 1569 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1570 u32 pid, u32 seq, 1571 struct net_device *dev, 1572 u32 filter_mask, 1573 int nlflags); 1574 int (*ndo_bridge_dellink)(struct net_device *dev, 1575 struct nlmsghdr *nlh, 1576 u16 flags); 1577 int (*ndo_change_carrier)(struct net_device *dev, 1578 bool new_carrier); 1579 int (*ndo_get_phys_port_id)(struct net_device *dev, 1580 struct netdev_phys_item_id *ppid); 1581 int (*ndo_get_port_parent_id)(struct net_device *dev, 1582 struct netdev_phys_item_id *ppid); 1583 int (*ndo_get_phys_port_name)(struct net_device *dev, 1584 char *name, size_t len); 1585 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1586 struct net_device *dev); 1587 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1588 void *priv); 1589 1590 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1591 int queue_index, 1592 u32 maxrate); 1593 int (*ndo_get_iflink)(const struct net_device *dev); 1594 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1595 struct sk_buff *skb); 1596 void (*ndo_set_rx_headroom)(struct net_device *dev, 1597 int needed_headroom); 1598 int (*ndo_bpf)(struct net_device *dev, 1599 struct netdev_bpf *bpf); 1600 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1601 struct xdp_frame **xdp, 1602 u32 flags); 1603 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1604 struct xdp_buff *xdp); 1605 int (*ndo_xsk_wakeup)(struct net_device *dev, 1606 u32 queue_id, u32 flags); 1607 int (*ndo_tunnel_ctl)(struct net_device *dev, 1608 struct ip_tunnel_parm *p, int cmd); 1609 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1610 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1611 struct net_device_path *path); 1612 ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1613 const struct skb_shared_hwtstamps *hwtstamps, 1614 bool cycles); 1615 }; 1616 1617 /** 1618 * enum netdev_priv_flags - &struct net_device priv_flags 1619 * 1620 * These are the &struct net_device, they are only set internally 1621 * by drivers and used in the kernel. These flags are invisible to 1622 * userspace; this means that the order of these flags can change 1623 * during any kernel release. 1624 * 1625 * You should have a pretty good reason to be extending these flags. 1626 * 1627 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1628 * @IFF_EBRIDGE: Ethernet bridging device 1629 * @IFF_BONDING: bonding master or slave 1630 * @IFF_ISATAP: ISATAP interface (RFC4214) 1631 * @IFF_WAN_HDLC: WAN HDLC device 1632 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1633 * release skb->dst 1634 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1635 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1636 * @IFF_MACVLAN_PORT: device used as macvlan port 1637 * @IFF_BRIDGE_PORT: device used as bridge port 1638 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1639 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1640 * @IFF_UNICAST_FLT: Supports unicast filtering 1641 * @IFF_TEAM_PORT: device used as team port 1642 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1643 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1644 * change when it's running 1645 * @IFF_MACVLAN: Macvlan device 1646 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1647 * underlying stacked devices 1648 * @IFF_L3MDEV_MASTER: device is an L3 master device 1649 * @IFF_NO_QUEUE: device can run without qdisc attached 1650 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1651 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1652 * @IFF_TEAM: device is a team device 1653 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1654 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1655 * entity (i.e. the master device for bridged veth) 1656 * @IFF_MACSEC: device is a MACsec device 1657 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1658 * @IFF_FAILOVER: device is a failover master device 1659 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1660 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1661 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1662 * skb_headlen(skb) == 0 (data starts from frag0) 1663 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN 1664 */ 1665 enum netdev_priv_flags { 1666 IFF_802_1Q_VLAN = 1<<0, 1667 IFF_EBRIDGE = 1<<1, 1668 IFF_BONDING = 1<<2, 1669 IFF_ISATAP = 1<<3, 1670 IFF_WAN_HDLC = 1<<4, 1671 IFF_XMIT_DST_RELEASE = 1<<5, 1672 IFF_DONT_BRIDGE = 1<<6, 1673 IFF_DISABLE_NETPOLL = 1<<7, 1674 IFF_MACVLAN_PORT = 1<<8, 1675 IFF_BRIDGE_PORT = 1<<9, 1676 IFF_OVS_DATAPATH = 1<<10, 1677 IFF_TX_SKB_SHARING = 1<<11, 1678 IFF_UNICAST_FLT = 1<<12, 1679 IFF_TEAM_PORT = 1<<13, 1680 IFF_SUPP_NOFCS = 1<<14, 1681 IFF_LIVE_ADDR_CHANGE = 1<<15, 1682 IFF_MACVLAN = 1<<16, 1683 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1684 IFF_L3MDEV_MASTER = 1<<18, 1685 IFF_NO_QUEUE = 1<<19, 1686 IFF_OPENVSWITCH = 1<<20, 1687 IFF_L3MDEV_SLAVE = 1<<21, 1688 IFF_TEAM = 1<<22, 1689 IFF_RXFH_CONFIGURED = 1<<23, 1690 IFF_PHONY_HEADROOM = 1<<24, 1691 IFF_MACSEC = 1<<25, 1692 IFF_NO_RX_HANDLER = 1<<26, 1693 IFF_FAILOVER = 1<<27, 1694 IFF_FAILOVER_SLAVE = 1<<28, 1695 IFF_L3MDEV_RX_HANDLER = 1<<29, 1696 /* was IFF_LIVE_RENAME_OK */ 1697 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31), 1698 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32), 1699 }; 1700 1701 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1702 #define IFF_EBRIDGE IFF_EBRIDGE 1703 #define IFF_BONDING IFF_BONDING 1704 #define IFF_ISATAP IFF_ISATAP 1705 #define IFF_WAN_HDLC IFF_WAN_HDLC 1706 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1707 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1708 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1709 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1710 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1711 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1712 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1713 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1714 #define IFF_TEAM_PORT IFF_TEAM_PORT 1715 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1716 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1717 #define IFF_MACVLAN IFF_MACVLAN 1718 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1719 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1720 #define IFF_NO_QUEUE IFF_NO_QUEUE 1721 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1722 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1723 #define IFF_TEAM IFF_TEAM 1724 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1725 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM 1726 #define IFF_MACSEC IFF_MACSEC 1727 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1728 #define IFF_FAILOVER IFF_FAILOVER 1729 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1730 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1731 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR 1732 1733 /* Specifies the type of the struct net_device::ml_priv pointer */ 1734 enum netdev_ml_priv_type { 1735 ML_PRIV_NONE, 1736 ML_PRIV_CAN, 1737 }; 1738 1739 /** 1740 * struct net_device - The DEVICE structure. 1741 * 1742 * Actually, this whole structure is a big mistake. It mixes I/O 1743 * data with strictly "high-level" data, and it has to know about 1744 * almost every data structure used in the INET module. 1745 * 1746 * @name: This is the first field of the "visible" part of this structure 1747 * (i.e. as seen by users in the "Space.c" file). It is the name 1748 * of the interface. 1749 * 1750 * @name_node: Name hashlist node 1751 * @ifalias: SNMP alias 1752 * @mem_end: Shared memory end 1753 * @mem_start: Shared memory start 1754 * @base_addr: Device I/O address 1755 * @irq: Device IRQ number 1756 * 1757 * @state: Generic network queuing layer state, see netdev_state_t 1758 * @dev_list: The global list of network devices 1759 * @napi_list: List entry used for polling NAPI devices 1760 * @unreg_list: List entry when we are unregistering the 1761 * device; see the function unregister_netdev 1762 * @close_list: List entry used when we are closing the device 1763 * @ptype_all: Device-specific packet handlers for all protocols 1764 * @ptype_specific: Device-specific, protocol-specific packet handlers 1765 * 1766 * @adj_list: Directly linked devices, like slaves for bonding 1767 * @features: Currently active device features 1768 * @hw_features: User-changeable features 1769 * 1770 * @wanted_features: User-requested features 1771 * @vlan_features: Mask of features inheritable by VLAN devices 1772 * 1773 * @hw_enc_features: Mask of features inherited by encapsulating devices 1774 * This field indicates what encapsulation 1775 * offloads the hardware is capable of doing, 1776 * and drivers will need to set them appropriately. 1777 * 1778 * @mpls_features: Mask of features inheritable by MPLS 1779 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1780 * 1781 * @ifindex: interface index 1782 * @group: The group the device belongs to 1783 * 1784 * @stats: Statistics struct, which was left as a legacy, use 1785 * rtnl_link_stats64 instead 1786 * 1787 * @core_stats: core networking counters, 1788 * do not use this in drivers 1789 * @carrier_up_count: Number of times the carrier has been up 1790 * @carrier_down_count: Number of times the carrier has been down 1791 * 1792 * @wireless_handlers: List of functions to handle Wireless Extensions, 1793 * instead of ioctl, 1794 * see <net/iw_handler.h> for details. 1795 * @wireless_data: Instance data managed by the core of wireless extensions 1796 * 1797 * @netdev_ops: Includes several pointers to callbacks, 1798 * if one wants to override the ndo_*() functions 1799 * @ethtool_ops: Management operations 1800 * @l3mdev_ops: Layer 3 master device operations 1801 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1802 * discovery handling. Necessary for e.g. 6LoWPAN. 1803 * @xfrmdev_ops: Transformation offload operations 1804 * @tlsdev_ops: Transport Layer Security offload operations 1805 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1806 * of Layer 2 headers. 1807 * 1808 * @flags: Interface flags (a la BSD) 1809 * @priv_flags: Like 'flags' but invisible to userspace, 1810 * see if.h for the definitions 1811 * @gflags: Global flags ( kept as legacy ) 1812 * @padded: How much padding added by alloc_netdev() 1813 * @operstate: RFC2863 operstate 1814 * @link_mode: Mapping policy to operstate 1815 * @if_port: Selectable AUI, TP, ... 1816 * @dma: DMA channel 1817 * @mtu: Interface MTU value 1818 * @min_mtu: Interface Minimum MTU value 1819 * @max_mtu: Interface Maximum MTU value 1820 * @type: Interface hardware type 1821 * @hard_header_len: Maximum hardware header length. 1822 * @min_header_len: Minimum hardware header length 1823 * 1824 * @needed_headroom: Extra headroom the hardware may need, but not in all 1825 * cases can this be guaranteed 1826 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1827 * cases can this be guaranteed. Some cases also use 1828 * LL_MAX_HEADER instead to allocate the skb 1829 * 1830 * interface address info: 1831 * 1832 * @perm_addr: Permanent hw address 1833 * @addr_assign_type: Hw address assignment type 1834 * @addr_len: Hardware address length 1835 * @upper_level: Maximum depth level of upper devices. 1836 * @lower_level: Maximum depth level of lower devices. 1837 * @neigh_priv_len: Used in neigh_alloc() 1838 * @dev_id: Used to differentiate devices that share 1839 * the same link layer address 1840 * @dev_port: Used to differentiate devices that share 1841 * the same function 1842 * @addr_list_lock: XXX: need comments on this one 1843 * @name_assign_type: network interface name assignment type 1844 * @uc_promisc: Counter that indicates promiscuous mode 1845 * has been enabled due to the need to listen to 1846 * additional unicast addresses in a device that 1847 * does not implement ndo_set_rx_mode() 1848 * @uc: unicast mac addresses 1849 * @mc: multicast mac addresses 1850 * @dev_addrs: list of device hw addresses 1851 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1852 * @promiscuity: Number of times the NIC is told to work in 1853 * promiscuous mode; if it becomes 0 the NIC will 1854 * exit promiscuous mode 1855 * @allmulti: Counter, enables or disables allmulticast mode 1856 * 1857 * @vlan_info: VLAN info 1858 * @dsa_ptr: dsa specific data 1859 * @tipc_ptr: TIPC specific data 1860 * @atalk_ptr: AppleTalk link 1861 * @ip_ptr: IPv4 specific data 1862 * @ip6_ptr: IPv6 specific data 1863 * @ax25_ptr: AX.25 specific data 1864 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1865 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1866 * device struct 1867 * @mpls_ptr: mpls_dev struct pointer 1868 * @mctp_ptr: MCTP specific data 1869 * 1870 * @dev_addr: Hw address (before bcast, 1871 * because most packets are unicast) 1872 * 1873 * @_rx: Array of RX queues 1874 * @num_rx_queues: Number of RX queues 1875 * allocated at register_netdev() time 1876 * @real_num_rx_queues: Number of RX queues currently active in device 1877 * @xdp_prog: XDP sockets filter program pointer 1878 * @gro_flush_timeout: timeout for GRO layer in NAPI 1879 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1880 * allow to avoid NIC hard IRQ, on busy queues. 1881 * 1882 * @rx_handler: handler for received packets 1883 * @rx_handler_data: XXX: need comments on this one 1884 * @miniq_ingress: ingress/clsact qdisc specific data for 1885 * ingress processing 1886 * @ingress_queue: XXX: need comments on this one 1887 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1888 * @broadcast: hw bcast address 1889 * 1890 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1891 * indexed by RX queue number. Assigned by driver. 1892 * This must only be set if the ndo_rx_flow_steer 1893 * operation is defined 1894 * @index_hlist: Device index hash chain 1895 * 1896 * @_tx: Array of TX queues 1897 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1898 * @real_num_tx_queues: Number of TX queues currently active in device 1899 * @qdisc: Root qdisc from userspace point of view 1900 * @tx_queue_len: Max frames per queue allowed 1901 * @tx_global_lock: XXX: need comments on this one 1902 * @xdp_bulkq: XDP device bulk queue 1903 * @xps_maps: all CPUs/RXQs maps for XPS device 1904 * 1905 * @xps_maps: XXX: need comments on this one 1906 * @miniq_egress: clsact qdisc specific data for 1907 * egress processing 1908 * @nf_hooks_egress: netfilter hooks executed for egress packets 1909 * @qdisc_hash: qdisc hash table 1910 * @watchdog_timeo: Represents the timeout that is used by 1911 * the watchdog (see dev_watchdog()) 1912 * @watchdog_timer: List of timers 1913 * 1914 * @proto_down_reason: reason a netdev interface is held down 1915 * @pcpu_refcnt: Number of references to this device 1916 * @dev_refcnt: Number of references to this device 1917 * @refcnt_tracker: Tracker directory for tracked references to this device 1918 * @todo_list: Delayed register/unregister 1919 * @link_watch_list: XXX: need comments on this one 1920 * 1921 * @reg_state: Register/unregister state machine 1922 * @dismantle: Device is going to be freed 1923 * @rtnl_link_state: This enum represents the phases of creating 1924 * a new link 1925 * 1926 * @needs_free_netdev: Should unregister perform free_netdev? 1927 * @priv_destructor: Called from unregister 1928 * @npinfo: XXX: need comments on this one 1929 * @nd_net: Network namespace this network device is inside 1930 * 1931 * @ml_priv: Mid-layer private 1932 * @ml_priv_type: Mid-layer private type 1933 * @lstats: Loopback statistics 1934 * @tstats: Tunnel statistics 1935 * @dstats: Dummy statistics 1936 * @vstats: Virtual ethernet statistics 1937 * 1938 * @garp_port: GARP 1939 * @mrp_port: MRP 1940 * 1941 * @dm_private: Drop monitor private 1942 * 1943 * @dev: Class/net/name entry 1944 * @sysfs_groups: Space for optional device, statistics and wireless 1945 * sysfs groups 1946 * 1947 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1948 * @rtnl_link_ops: Rtnl_link_ops 1949 * 1950 * @gso_max_size: Maximum size of generic segmentation offload 1951 * @tso_max_size: Device (as in HW) limit on the max TSO request size 1952 * @gso_max_segs: Maximum number of segments that can be passed to the 1953 * NIC for GSO 1954 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count 1955 * 1956 * @dcbnl_ops: Data Center Bridging netlink ops 1957 * @num_tc: Number of traffic classes in the net device 1958 * @tc_to_txq: XXX: need comments on this one 1959 * @prio_tc_map: XXX: need comments on this one 1960 * 1961 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1962 * 1963 * @priomap: XXX: need comments on this one 1964 * @phydev: Physical device may attach itself 1965 * for hardware timestamping 1966 * @sfp_bus: attached &struct sfp_bus structure. 1967 * 1968 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1969 * 1970 * @proto_down: protocol port state information can be sent to the 1971 * switch driver and used to set the phys state of the 1972 * switch port. 1973 * 1974 * @wol_enabled: Wake-on-LAN is enabled 1975 * 1976 * @threaded: napi threaded mode is enabled 1977 * 1978 * @net_notifier_list: List of per-net netdev notifier block 1979 * that follow this device when it is moved 1980 * to another network namespace. 1981 * 1982 * @macsec_ops: MACsec offloading ops 1983 * 1984 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1985 * offload capabilities of the device 1986 * @udp_tunnel_nic: UDP tunnel offload state 1987 * @xdp_state: stores info on attached XDP BPF programs 1988 * 1989 * @nested_level: Used as a parameter of spin_lock_nested() of 1990 * dev->addr_list_lock. 1991 * @unlink_list: As netif_addr_lock() can be called recursively, 1992 * keep a list of interfaces to be deleted. 1993 * @gro_max_size: Maximum size of aggregated packet in generic 1994 * receive offload (GRO) 1995 * 1996 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 1997 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 1998 * @watchdog_dev_tracker: refcount tracker used by watchdog. 1999 * @dev_registered_tracker: tracker for reference held while 2000 * registered 2001 * @offload_xstats_l3: L3 HW stats for this netdevice. 2002 * 2003 * @devlink_port: Pointer to related devlink port structure. 2004 * Assigned by a driver before netdev registration using 2005 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static 2006 * during the time netdevice is registered. 2007 * 2008 * FIXME: cleanup struct net_device such that network protocol info 2009 * moves out. 2010 */ 2011 2012 struct net_device { 2013 char name[IFNAMSIZ]; 2014 struct netdev_name_node *name_node; 2015 struct dev_ifalias __rcu *ifalias; 2016 /* 2017 * I/O specific fields 2018 * FIXME: Merge these and struct ifmap into one 2019 */ 2020 unsigned long mem_end; 2021 unsigned long mem_start; 2022 unsigned long base_addr; 2023 2024 /* 2025 * Some hardware also needs these fields (state,dev_list, 2026 * napi_list,unreg_list,close_list) but they are not 2027 * part of the usual set specified in Space.c. 2028 */ 2029 2030 unsigned long state; 2031 2032 struct list_head dev_list; 2033 struct list_head napi_list; 2034 struct list_head unreg_list; 2035 struct list_head close_list; 2036 struct list_head ptype_all; 2037 struct list_head ptype_specific; 2038 2039 struct { 2040 struct list_head upper; 2041 struct list_head lower; 2042 } adj_list; 2043 2044 /* Read-mostly cache-line for fast-path access */ 2045 unsigned int flags; 2046 unsigned long long priv_flags; 2047 const struct net_device_ops *netdev_ops; 2048 int ifindex; 2049 unsigned short gflags; 2050 unsigned short hard_header_len; 2051 2052 /* Note : dev->mtu is often read without holding a lock. 2053 * Writers usually hold RTNL. 2054 * It is recommended to use READ_ONCE() to annotate the reads, 2055 * and to use WRITE_ONCE() to annotate the writes. 2056 */ 2057 unsigned int mtu; 2058 unsigned short needed_headroom; 2059 unsigned short needed_tailroom; 2060 2061 netdev_features_t features; 2062 netdev_features_t hw_features; 2063 netdev_features_t wanted_features; 2064 netdev_features_t vlan_features; 2065 netdev_features_t hw_enc_features; 2066 netdev_features_t mpls_features; 2067 netdev_features_t gso_partial_features; 2068 2069 unsigned int min_mtu; 2070 unsigned int max_mtu; 2071 unsigned short type; 2072 unsigned char min_header_len; 2073 unsigned char name_assign_type; 2074 2075 int group; 2076 2077 struct net_device_stats stats; /* not used by modern drivers */ 2078 2079 struct net_device_core_stats __percpu *core_stats; 2080 2081 /* Stats to monitor link on/off, flapping */ 2082 atomic_t carrier_up_count; 2083 atomic_t carrier_down_count; 2084 2085 #ifdef CONFIG_WIRELESS_EXT 2086 const struct iw_handler_def *wireless_handlers; 2087 struct iw_public_data *wireless_data; 2088 #endif 2089 const struct ethtool_ops *ethtool_ops; 2090 #ifdef CONFIG_NET_L3_MASTER_DEV 2091 const struct l3mdev_ops *l3mdev_ops; 2092 #endif 2093 #if IS_ENABLED(CONFIG_IPV6) 2094 const struct ndisc_ops *ndisc_ops; 2095 #endif 2096 2097 #ifdef CONFIG_XFRM_OFFLOAD 2098 const struct xfrmdev_ops *xfrmdev_ops; 2099 #endif 2100 2101 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2102 const struct tlsdev_ops *tlsdev_ops; 2103 #endif 2104 2105 const struct header_ops *header_ops; 2106 2107 unsigned char operstate; 2108 unsigned char link_mode; 2109 2110 unsigned char if_port; 2111 unsigned char dma; 2112 2113 /* Interface address info. */ 2114 unsigned char perm_addr[MAX_ADDR_LEN]; 2115 unsigned char addr_assign_type; 2116 unsigned char addr_len; 2117 unsigned char upper_level; 2118 unsigned char lower_level; 2119 2120 unsigned short neigh_priv_len; 2121 unsigned short dev_id; 2122 unsigned short dev_port; 2123 unsigned short padded; 2124 2125 spinlock_t addr_list_lock; 2126 int irq; 2127 2128 struct netdev_hw_addr_list uc; 2129 struct netdev_hw_addr_list mc; 2130 struct netdev_hw_addr_list dev_addrs; 2131 2132 #ifdef CONFIG_SYSFS 2133 struct kset *queues_kset; 2134 #endif 2135 #ifdef CONFIG_LOCKDEP 2136 struct list_head unlink_list; 2137 #endif 2138 unsigned int promiscuity; 2139 unsigned int allmulti; 2140 bool uc_promisc; 2141 #ifdef CONFIG_LOCKDEP 2142 unsigned char nested_level; 2143 #endif 2144 2145 2146 /* Protocol-specific pointers */ 2147 2148 struct in_device __rcu *ip_ptr; 2149 struct inet6_dev __rcu *ip6_ptr; 2150 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2151 struct vlan_info __rcu *vlan_info; 2152 #endif 2153 #if IS_ENABLED(CONFIG_NET_DSA) 2154 struct dsa_port *dsa_ptr; 2155 #endif 2156 #if IS_ENABLED(CONFIG_TIPC) 2157 struct tipc_bearer __rcu *tipc_ptr; 2158 #endif 2159 #if IS_ENABLED(CONFIG_ATALK) 2160 void *atalk_ptr; 2161 #endif 2162 #if IS_ENABLED(CONFIG_AX25) 2163 void *ax25_ptr; 2164 #endif 2165 #if IS_ENABLED(CONFIG_CFG80211) 2166 struct wireless_dev *ieee80211_ptr; 2167 #endif 2168 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2169 struct wpan_dev *ieee802154_ptr; 2170 #endif 2171 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2172 struct mpls_dev __rcu *mpls_ptr; 2173 #endif 2174 #if IS_ENABLED(CONFIG_MCTP) 2175 struct mctp_dev __rcu *mctp_ptr; 2176 #endif 2177 2178 /* 2179 * Cache lines mostly used on receive path (including eth_type_trans()) 2180 */ 2181 /* Interface address info used in eth_type_trans() */ 2182 const unsigned char *dev_addr; 2183 2184 struct netdev_rx_queue *_rx; 2185 unsigned int num_rx_queues; 2186 unsigned int real_num_rx_queues; 2187 2188 struct bpf_prog __rcu *xdp_prog; 2189 unsigned long gro_flush_timeout; 2190 int napi_defer_hard_irqs; 2191 #define GRO_LEGACY_MAX_SIZE 65536u 2192 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2193 * and shinfo->gso_segs is a 16bit field. 2194 */ 2195 #define GRO_MAX_SIZE (8 * 65535u) 2196 unsigned int gro_max_size; 2197 rx_handler_func_t __rcu *rx_handler; 2198 void __rcu *rx_handler_data; 2199 2200 #ifdef CONFIG_NET_CLS_ACT 2201 struct mini_Qdisc __rcu *miniq_ingress; 2202 #endif 2203 struct netdev_queue __rcu *ingress_queue; 2204 #ifdef CONFIG_NETFILTER_INGRESS 2205 struct nf_hook_entries __rcu *nf_hooks_ingress; 2206 #endif 2207 2208 unsigned char broadcast[MAX_ADDR_LEN]; 2209 #ifdef CONFIG_RFS_ACCEL 2210 struct cpu_rmap *rx_cpu_rmap; 2211 #endif 2212 struct hlist_node index_hlist; 2213 2214 /* 2215 * Cache lines mostly used on transmit path 2216 */ 2217 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2218 unsigned int num_tx_queues; 2219 unsigned int real_num_tx_queues; 2220 struct Qdisc __rcu *qdisc; 2221 unsigned int tx_queue_len; 2222 spinlock_t tx_global_lock; 2223 2224 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2225 2226 #ifdef CONFIG_XPS 2227 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2228 #endif 2229 #ifdef CONFIG_NET_CLS_ACT 2230 struct mini_Qdisc __rcu *miniq_egress; 2231 #endif 2232 #ifdef CONFIG_NETFILTER_EGRESS 2233 struct nf_hook_entries __rcu *nf_hooks_egress; 2234 #endif 2235 2236 #ifdef CONFIG_NET_SCHED 2237 DECLARE_HASHTABLE (qdisc_hash, 4); 2238 #endif 2239 /* These may be needed for future network-power-down code. */ 2240 struct timer_list watchdog_timer; 2241 int watchdog_timeo; 2242 2243 u32 proto_down_reason; 2244 2245 struct list_head todo_list; 2246 2247 #ifdef CONFIG_PCPU_DEV_REFCNT 2248 int __percpu *pcpu_refcnt; 2249 #else 2250 refcount_t dev_refcnt; 2251 #endif 2252 struct ref_tracker_dir refcnt_tracker; 2253 2254 struct list_head link_watch_list; 2255 2256 enum { NETREG_UNINITIALIZED=0, 2257 NETREG_REGISTERED, /* completed register_netdevice */ 2258 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2259 NETREG_UNREGISTERED, /* completed unregister todo */ 2260 NETREG_RELEASED, /* called free_netdev */ 2261 NETREG_DUMMY, /* dummy device for NAPI poll */ 2262 } reg_state:8; 2263 2264 bool dismantle; 2265 2266 enum { 2267 RTNL_LINK_INITIALIZED, 2268 RTNL_LINK_INITIALIZING, 2269 } rtnl_link_state:16; 2270 2271 bool needs_free_netdev; 2272 void (*priv_destructor)(struct net_device *dev); 2273 2274 #ifdef CONFIG_NETPOLL 2275 struct netpoll_info __rcu *npinfo; 2276 #endif 2277 2278 possible_net_t nd_net; 2279 2280 /* mid-layer private */ 2281 void *ml_priv; 2282 enum netdev_ml_priv_type ml_priv_type; 2283 2284 union { 2285 struct pcpu_lstats __percpu *lstats; 2286 struct pcpu_sw_netstats __percpu *tstats; 2287 struct pcpu_dstats __percpu *dstats; 2288 }; 2289 2290 #if IS_ENABLED(CONFIG_GARP) 2291 struct garp_port __rcu *garp_port; 2292 #endif 2293 #if IS_ENABLED(CONFIG_MRP) 2294 struct mrp_port __rcu *mrp_port; 2295 #endif 2296 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2297 struct dm_hw_stat_delta __rcu *dm_private; 2298 #endif 2299 struct device dev; 2300 const struct attribute_group *sysfs_groups[4]; 2301 const struct attribute_group *sysfs_rx_queue_group; 2302 2303 const struct rtnl_link_ops *rtnl_link_ops; 2304 2305 /* for setting kernel sock attribute on TCP connection setup */ 2306 #define GSO_MAX_SEGS 65535u 2307 #define GSO_LEGACY_MAX_SIZE 65536u 2308 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2309 * and shinfo->gso_segs is a 16bit field. 2310 */ 2311 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2312 2313 unsigned int gso_max_size; 2314 #define TSO_LEGACY_MAX_SIZE 65536 2315 #define TSO_MAX_SIZE UINT_MAX 2316 unsigned int tso_max_size; 2317 u16 gso_max_segs; 2318 #define TSO_MAX_SEGS U16_MAX 2319 u16 tso_max_segs; 2320 2321 #ifdef CONFIG_DCB 2322 const struct dcbnl_rtnl_ops *dcbnl_ops; 2323 #endif 2324 s16 num_tc; 2325 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2326 u8 prio_tc_map[TC_BITMASK + 1]; 2327 2328 #if IS_ENABLED(CONFIG_FCOE) 2329 unsigned int fcoe_ddp_xid; 2330 #endif 2331 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2332 struct netprio_map __rcu *priomap; 2333 #endif 2334 struct phy_device *phydev; 2335 struct sfp_bus *sfp_bus; 2336 struct lock_class_key *qdisc_tx_busylock; 2337 bool proto_down; 2338 unsigned wol_enabled:1; 2339 unsigned threaded:1; 2340 2341 struct list_head net_notifier_list; 2342 2343 #if IS_ENABLED(CONFIG_MACSEC) 2344 /* MACsec management functions */ 2345 const struct macsec_ops *macsec_ops; 2346 #endif 2347 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2348 struct udp_tunnel_nic *udp_tunnel_nic; 2349 2350 /* protected by rtnl_lock */ 2351 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2352 2353 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2354 netdevice_tracker linkwatch_dev_tracker; 2355 netdevice_tracker watchdog_dev_tracker; 2356 netdevice_tracker dev_registered_tracker; 2357 struct rtnl_hw_stats64 *offload_xstats_l3; 2358 2359 struct devlink_port *devlink_port; 2360 }; 2361 #define to_net_dev(d) container_of(d, struct net_device, dev) 2362 2363 /* 2364 * Driver should use this to assign devlink port instance to a netdevice 2365 * before it registers the netdevice. Therefore devlink_port is static 2366 * during the netdev lifetime after it is registered. 2367 */ 2368 #define SET_NETDEV_DEVLINK_PORT(dev, port) \ 2369 ({ \ 2370 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \ 2371 ((dev)->devlink_port = (port)); \ 2372 }) 2373 2374 static inline bool netif_elide_gro(const struct net_device *dev) 2375 { 2376 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2377 return true; 2378 return false; 2379 } 2380 2381 #define NETDEV_ALIGN 32 2382 2383 static inline 2384 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2385 { 2386 return dev->prio_tc_map[prio & TC_BITMASK]; 2387 } 2388 2389 static inline 2390 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2391 { 2392 if (tc >= dev->num_tc) 2393 return -EINVAL; 2394 2395 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2396 return 0; 2397 } 2398 2399 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2400 void netdev_reset_tc(struct net_device *dev); 2401 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2402 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2403 2404 static inline 2405 int netdev_get_num_tc(struct net_device *dev) 2406 { 2407 return dev->num_tc; 2408 } 2409 2410 static inline void net_prefetch(void *p) 2411 { 2412 prefetch(p); 2413 #if L1_CACHE_BYTES < 128 2414 prefetch((u8 *)p + L1_CACHE_BYTES); 2415 #endif 2416 } 2417 2418 static inline void net_prefetchw(void *p) 2419 { 2420 prefetchw(p); 2421 #if L1_CACHE_BYTES < 128 2422 prefetchw((u8 *)p + L1_CACHE_BYTES); 2423 #endif 2424 } 2425 2426 void netdev_unbind_sb_channel(struct net_device *dev, 2427 struct net_device *sb_dev); 2428 int netdev_bind_sb_channel_queue(struct net_device *dev, 2429 struct net_device *sb_dev, 2430 u8 tc, u16 count, u16 offset); 2431 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2432 static inline int netdev_get_sb_channel(struct net_device *dev) 2433 { 2434 return max_t(int, -dev->num_tc, 0); 2435 } 2436 2437 static inline 2438 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2439 unsigned int index) 2440 { 2441 return &dev->_tx[index]; 2442 } 2443 2444 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2445 const struct sk_buff *skb) 2446 { 2447 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2448 } 2449 2450 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2451 void (*f)(struct net_device *, 2452 struct netdev_queue *, 2453 void *), 2454 void *arg) 2455 { 2456 unsigned int i; 2457 2458 for (i = 0; i < dev->num_tx_queues; i++) 2459 f(dev, &dev->_tx[i], arg); 2460 } 2461 2462 #define netdev_lockdep_set_classes(dev) \ 2463 { \ 2464 static struct lock_class_key qdisc_tx_busylock_key; \ 2465 static struct lock_class_key qdisc_xmit_lock_key; \ 2466 static struct lock_class_key dev_addr_list_lock_key; \ 2467 unsigned int i; \ 2468 \ 2469 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2470 lockdep_set_class(&(dev)->addr_list_lock, \ 2471 &dev_addr_list_lock_key); \ 2472 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2473 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2474 &qdisc_xmit_lock_key); \ 2475 } 2476 2477 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2478 struct net_device *sb_dev); 2479 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2480 struct sk_buff *skb, 2481 struct net_device *sb_dev); 2482 2483 /* returns the headroom that the master device needs to take in account 2484 * when forwarding to this dev 2485 */ 2486 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2487 { 2488 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2489 } 2490 2491 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2492 { 2493 if (dev->netdev_ops->ndo_set_rx_headroom) 2494 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2495 } 2496 2497 /* set the device rx headroom to the dev's default */ 2498 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2499 { 2500 netdev_set_rx_headroom(dev, -1); 2501 } 2502 2503 static inline void *netdev_get_ml_priv(struct net_device *dev, 2504 enum netdev_ml_priv_type type) 2505 { 2506 if (dev->ml_priv_type != type) 2507 return NULL; 2508 2509 return dev->ml_priv; 2510 } 2511 2512 static inline void netdev_set_ml_priv(struct net_device *dev, 2513 void *ml_priv, 2514 enum netdev_ml_priv_type type) 2515 { 2516 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2517 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2518 dev->ml_priv_type, type); 2519 WARN(!dev->ml_priv_type && dev->ml_priv, 2520 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2521 2522 dev->ml_priv = ml_priv; 2523 dev->ml_priv_type = type; 2524 } 2525 2526 /* 2527 * Net namespace inlines 2528 */ 2529 static inline 2530 struct net *dev_net(const struct net_device *dev) 2531 { 2532 return read_pnet(&dev->nd_net); 2533 } 2534 2535 static inline 2536 void dev_net_set(struct net_device *dev, struct net *net) 2537 { 2538 write_pnet(&dev->nd_net, net); 2539 } 2540 2541 /** 2542 * netdev_priv - access network device private data 2543 * @dev: network device 2544 * 2545 * Get network device private data 2546 */ 2547 static inline void *netdev_priv(const struct net_device *dev) 2548 { 2549 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2550 } 2551 2552 /* Set the sysfs physical device reference for the network logical device 2553 * if set prior to registration will cause a symlink during initialization. 2554 */ 2555 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2556 2557 /* Set the sysfs device type for the network logical device to allow 2558 * fine-grained identification of different network device types. For 2559 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2560 */ 2561 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2562 2563 /* Default NAPI poll() weight 2564 * Device drivers are strongly advised to not use bigger value 2565 */ 2566 #define NAPI_POLL_WEIGHT 64 2567 2568 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2569 int (*poll)(struct napi_struct *, int), int weight); 2570 2571 /** 2572 * netif_napi_add() - initialize a NAPI context 2573 * @dev: network device 2574 * @napi: NAPI context 2575 * @poll: polling function 2576 * 2577 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2578 * *any* of the other NAPI-related functions. 2579 */ 2580 static inline void 2581 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2582 int (*poll)(struct napi_struct *, int)) 2583 { 2584 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2585 } 2586 2587 static inline void 2588 netif_napi_add_tx_weight(struct net_device *dev, 2589 struct napi_struct *napi, 2590 int (*poll)(struct napi_struct *, int), 2591 int weight) 2592 { 2593 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2594 netif_napi_add_weight(dev, napi, poll, weight); 2595 } 2596 2597 /** 2598 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2599 * @dev: network device 2600 * @napi: NAPI context 2601 * @poll: polling function 2602 * 2603 * This variant of netif_napi_add() should be used from drivers using NAPI 2604 * to exclusively poll a TX queue. 2605 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2606 */ 2607 static inline void netif_napi_add_tx(struct net_device *dev, 2608 struct napi_struct *napi, 2609 int (*poll)(struct napi_struct *, int)) 2610 { 2611 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2612 } 2613 2614 /** 2615 * __netif_napi_del - remove a NAPI context 2616 * @napi: NAPI context 2617 * 2618 * Warning: caller must observe RCU grace period before freeing memory 2619 * containing @napi. Drivers might want to call this helper to combine 2620 * all the needed RCU grace periods into a single one. 2621 */ 2622 void __netif_napi_del(struct napi_struct *napi); 2623 2624 /** 2625 * netif_napi_del - remove a NAPI context 2626 * @napi: NAPI context 2627 * 2628 * netif_napi_del() removes a NAPI context from the network device NAPI list 2629 */ 2630 static inline void netif_napi_del(struct napi_struct *napi) 2631 { 2632 __netif_napi_del(napi); 2633 synchronize_net(); 2634 } 2635 2636 struct packet_type { 2637 __be16 type; /* This is really htons(ether_type). */ 2638 bool ignore_outgoing; 2639 struct net_device *dev; /* NULL is wildcarded here */ 2640 netdevice_tracker dev_tracker; 2641 int (*func) (struct sk_buff *, 2642 struct net_device *, 2643 struct packet_type *, 2644 struct net_device *); 2645 void (*list_func) (struct list_head *, 2646 struct packet_type *, 2647 struct net_device *); 2648 bool (*id_match)(struct packet_type *ptype, 2649 struct sock *sk); 2650 struct net *af_packet_net; 2651 void *af_packet_priv; 2652 struct list_head list; 2653 }; 2654 2655 struct offload_callbacks { 2656 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2657 netdev_features_t features); 2658 struct sk_buff *(*gro_receive)(struct list_head *head, 2659 struct sk_buff *skb); 2660 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2661 }; 2662 2663 struct packet_offload { 2664 __be16 type; /* This is really htons(ether_type). */ 2665 u16 priority; 2666 struct offload_callbacks callbacks; 2667 struct list_head list; 2668 }; 2669 2670 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2671 struct pcpu_sw_netstats { 2672 u64_stats_t rx_packets; 2673 u64_stats_t rx_bytes; 2674 u64_stats_t tx_packets; 2675 u64_stats_t tx_bytes; 2676 struct u64_stats_sync syncp; 2677 } __aligned(4 * sizeof(u64)); 2678 2679 struct pcpu_lstats { 2680 u64_stats_t packets; 2681 u64_stats_t bytes; 2682 struct u64_stats_sync syncp; 2683 } __aligned(2 * sizeof(u64)); 2684 2685 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2686 2687 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2688 { 2689 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2690 2691 u64_stats_update_begin(&tstats->syncp); 2692 u64_stats_add(&tstats->rx_bytes, len); 2693 u64_stats_inc(&tstats->rx_packets); 2694 u64_stats_update_end(&tstats->syncp); 2695 } 2696 2697 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2698 unsigned int packets, 2699 unsigned int len) 2700 { 2701 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2702 2703 u64_stats_update_begin(&tstats->syncp); 2704 u64_stats_add(&tstats->tx_bytes, len); 2705 u64_stats_add(&tstats->tx_packets, packets); 2706 u64_stats_update_end(&tstats->syncp); 2707 } 2708 2709 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2710 { 2711 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2712 2713 u64_stats_update_begin(&lstats->syncp); 2714 u64_stats_add(&lstats->bytes, len); 2715 u64_stats_inc(&lstats->packets); 2716 u64_stats_update_end(&lstats->syncp); 2717 } 2718 2719 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2720 ({ \ 2721 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2722 if (pcpu_stats) { \ 2723 int __cpu; \ 2724 for_each_possible_cpu(__cpu) { \ 2725 typeof(type) *stat; \ 2726 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2727 u64_stats_init(&stat->syncp); \ 2728 } \ 2729 } \ 2730 pcpu_stats; \ 2731 }) 2732 2733 #define netdev_alloc_pcpu_stats(type) \ 2734 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2735 2736 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2737 ({ \ 2738 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2739 if (pcpu_stats) { \ 2740 int __cpu; \ 2741 for_each_possible_cpu(__cpu) { \ 2742 typeof(type) *stat; \ 2743 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2744 u64_stats_init(&stat->syncp); \ 2745 } \ 2746 } \ 2747 pcpu_stats; \ 2748 }) 2749 2750 enum netdev_lag_tx_type { 2751 NETDEV_LAG_TX_TYPE_UNKNOWN, 2752 NETDEV_LAG_TX_TYPE_RANDOM, 2753 NETDEV_LAG_TX_TYPE_BROADCAST, 2754 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2755 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2756 NETDEV_LAG_TX_TYPE_HASH, 2757 }; 2758 2759 enum netdev_lag_hash { 2760 NETDEV_LAG_HASH_NONE, 2761 NETDEV_LAG_HASH_L2, 2762 NETDEV_LAG_HASH_L34, 2763 NETDEV_LAG_HASH_L23, 2764 NETDEV_LAG_HASH_E23, 2765 NETDEV_LAG_HASH_E34, 2766 NETDEV_LAG_HASH_VLAN_SRCMAC, 2767 NETDEV_LAG_HASH_UNKNOWN, 2768 }; 2769 2770 struct netdev_lag_upper_info { 2771 enum netdev_lag_tx_type tx_type; 2772 enum netdev_lag_hash hash_type; 2773 }; 2774 2775 struct netdev_lag_lower_state_info { 2776 u8 link_up : 1, 2777 tx_enabled : 1; 2778 }; 2779 2780 #include <linux/notifier.h> 2781 2782 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2783 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2784 * adding new types. 2785 */ 2786 enum netdev_cmd { 2787 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2788 NETDEV_DOWN, 2789 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2790 detected a hardware crash and restarted 2791 - we can use this eg to kick tcp sessions 2792 once done */ 2793 NETDEV_CHANGE, /* Notify device state change */ 2794 NETDEV_REGISTER, 2795 NETDEV_UNREGISTER, 2796 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2797 NETDEV_CHANGEADDR, /* notify after the address change */ 2798 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2799 NETDEV_GOING_DOWN, 2800 NETDEV_CHANGENAME, 2801 NETDEV_FEAT_CHANGE, 2802 NETDEV_BONDING_FAILOVER, 2803 NETDEV_PRE_UP, 2804 NETDEV_PRE_TYPE_CHANGE, 2805 NETDEV_POST_TYPE_CHANGE, 2806 NETDEV_POST_INIT, 2807 NETDEV_PRE_UNINIT, 2808 NETDEV_RELEASE, 2809 NETDEV_NOTIFY_PEERS, 2810 NETDEV_JOIN, 2811 NETDEV_CHANGEUPPER, 2812 NETDEV_RESEND_IGMP, 2813 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2814 NETDEV_CHANGEINFODATA, 2815 NETDEV_BONDING_INFO, 2816 NETDEV_PRECHANGEUPPER, 2817 NETDEV_CHANGELOWERSTATE, 2818 NETDEV_UDP_TUNNEL_PUSH_INFO, 2819 NETDEV_UDP_TUNNEL_DROP_INFO, 2820 NETDEV_CHANGE_TX_QUEUE_LEN, 2821 NETDEV_CVLAN_FILTER_PUSH_INFO, 2822 NETDEV_CVLAN_FILTER_DROP_INFO, 2823 NETDEV_SVLAN_FILTER_PUSH_INFO, 2824 NETDEV_SVLAN_FILTER_DROP_INFO, 2825 NETDEV_OFFLOAD_XSTATS_ENABLE, 2826 NETDEV_OFFLOAD_XSTATS_DISABLE, 2827 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 2828 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 2829 }; 2830 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2831 2832 int register_netdevice_notifier(struct notifier_block *nb); 2833 int unregister_netdevice_notifier(struct notifier_block *nb); 2834 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2835 int unregister_netdevice_notifier_net(struct net *net, 2836 struct notifier_block *nb); 2837 void move_netdevice_notifier_net(struct net *src_net, struct net *dst_net, 2838 struct notifier_block *nb); 2839 int register_netdevice_notifier_dev_net(struct net_device *dev, 2840 struct notifier_block *nb, 2841 struct netdev_net_notifier *nn); 2842 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2843 struct notifier_block *nb, 2844 struct netdev_net_notifier *nn); 2845 2846 struct netdev_notifier_info { 2847 struct net_device *dev; 2848 struct netlink_ext_ack *extack; 2849 }; 2850 2851 struct netdev_notifier_info_ext { 2852 struct netdev_notifier_info info; /* must be first */ 2853 union { 2854 u32 mtu; 2855 } ext; 2856 }; 2857 2858 struct netdev_notifier_change_info { 2859 struct netdev_notifier_info info; /* must be first */ 2860 unsigned int flags_changed; 2861 }; 2862 2863 struct netdev_notifier_changeupper_info { 2864 struct netdev_notifier_info info; /* must be first */ 2865 struct net_device *upper_dev; /* new upper dev */ 2866 bool master; /* is upper dev master */ 2867 bool linking; /* is the notification for link or unlink */ 2868 void *upper_info; /* upper dev info */ 2869 }; 2870 2871 struct netdev_notifier_changelowerstate_info { 2872 struct netdev_notifier_info info; /* must be first */ 2873 void *lower_state_info; /* is lower dev state */ 2874 }; 2875 2876 struct netdev_notifier_pre_changeaddr_info { 2877 struct netdev_notifier_info info; /* must be first */ 2878 const unsigned char *dev_addr; 2879 }; 2880 2881 enum netdev_offload_xstats_type { 2882 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 2883 }; 2884 2885 struct netdev_notifier_offload_xstats_info { 2886 struct netdev_notifier_info info; /* must be first */ 2887 enum netdev_offload_xstats_type type; 2888 2889 union { 2890 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 2891 struct netdev_notifier_offload_xstats_rd *report_delta; 2892 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 2893 struct netdev_notifier_offload_xstats_ru *report_used; 2894 }; 2895 }; 2896 2897 int netdev_offload_xstats_enable(struct net_device *dev, 2898 enum netdev_offload_xstats_type type, 2899 struct netlink_ext_ack *extack); 2900 int netdev_offload_xstats_disable(struct net_device *dev, 2901 enum netdev_offload_xstats_type type); 2902 bool netdev_offload_xstats_enabled(const struct net_device *dev, 2903 enum netdev_offload_xstats_type type); 2904 int netdev_offload_xstats_get(struct net_device *dev, 2905 enum netdev_offload_xstats_type type, 2906 struct rtnl_hw_stats64 *stats, bool *used, 2907 struct netlink_ext_ack *extack); 2908 void 2909 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 2910 const struct rtnl_hw_stats64 *stats); 2911 void 2912 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 2913 void netdev_offload_xstats_push_delta(struct net_device *dev, 2914 enum netdev_offload_xstats_type type, 2915 const struct rtnl_hw_stats64 *stats); 2916 2917 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2918 struct net_device *dev) 2919 { 2920 info->dev = dev; 2921 info->extack = NULL; 2922 } 2923 2924 static inline struct net_device * 2925 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2926 { 2927 return info->dev; 2928 } 2929 2930 static inline struct netlink_ext_ack * 2931 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2932 { 2933 return info->extack; 2934 } 2935 2936 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2937 2938 2939 extern rwlock_t dev_base_lock; /* Device list lock */ 2940 2941 #define for_each_netdev(net, d) \ 2942 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2943 #define for_each_netdev_reverse(net, d) \ 2944 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2945 #define for_each_netdev_rcu(net, d) \ 2946 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2947 #define for_each_netdev_safe(net, d, n) \ 2948 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2949 #define for_each_netdev_continue(net, d) \ 2950 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2951 #define for_each_netdev_continue_reverse(net, d) \ 2952 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2953 dev_list) 2954 #define for_each_netdev_continue_rcu(net, d) \ 2955 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2956 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2957 for_each_netdev_rcu(&init_net, slave) \ 2958 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2959 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2960 2961 static inline struct net_device *next_net_device(struct net_device *dev) 2962 { 2963 struct list_head *lh; 2964 struct net *net; 2965 2966 net = dev_net(dev); 2967 lh = dev->dev_list.next; 2968 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2969 } 2970 2971 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2972 { 2973 struct list_head *lh; 2974 struct net *net; 2975 2976 net = dev_net(dev); 2977 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2978 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2979 } 2980 2981 static inline struct net_device *first_net_device(struct net *net) 2982 { 2983 return list_empty(&net->dev_base_head) ? NULL : 2984 net_device_entry(net->dev_base_head.next); 2985 } 2986 2987 static inline struct net_device *first_net_device_rcu(struct net *net) 2988 { 2989 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2990 2991 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2992 } 2993 2994 int netdev_boot_setup_check(struct net_device *dev); 2995 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2996 const char *hwaddr); 2997 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2998 void dev_add_pack(struct packet_type *pt); 2999 void dev_remove_pack(struct packet_type *pt); 3000 void __dev_remove_pack(struct packet_type *pt); 3001 void dev_add_offload(struct packet_offload *po); 3002 void dev_remove_offload(struct packet_offload *po); 3003 3004 int dev_get_iflink(const struct net_device *dev); 3005 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 3006 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 3007 struct net_device_path_stack *stack); 3008 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 3009 unsigned short mask); 3010 struct net_device *dev_get_by_name(struct net *net, const char *name); 3011 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 3012 struct net_device *__dev_get_by_name(struct net *net, const char *name); 3013 bool netdev_name_in_use(struct net *net, const char *name); 3014 int dev_alloc_name(struct net_device *dev, const char *name); 3015 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 3016 void dev_close(struct net_device *dev); 3017 void dev_close_many(struct list_head *head, bool unlink); 3018 void dev_disable_lro(struct net_device *dev); 3019 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 3020 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 3021 struct net_device *sb_dev); 3022 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 3023 struct net_device *sb_dev); 3024 3025 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3026 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3027 3028 static inline int dev_queue_xmit(struct sk_buff *skb) 3029 { 3030 return __dev_queue_xmit(skb, NULL); 3031 } 3032 3033 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3034 struct net_device *sb_dev) 3035 { 3036 return __dev_queue_xmit(skb, sb_dev); 3037 } 3038 3039 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3040 { 3041 int ret; 3042 3043 ret = __dev_direct_xmit(skb, queue_id); 3044 if (!dev_xmit_complete(ret)) 3045 kfree_skb(skb); 3046 return ret; 3047 } 3048 3049 int register_netdevice(struct net_device *dev); 3050 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3051 void unregister_netdevice_many(struct list_head *head); 3052 static inline void unregister_netdevice(struct net_device *dev) 3053 { 3054 unregister_netdevice_queue(dev, NULL); 3055 } 3056 3057 int netdev_refcnt_read(const struct net_device *dev); 3058 void free_netdev(struct net_device *dev); 3059 void netdev_freemem(struct net_device *dev); 3060 int init_dummy_netdev(struct net_device *dev); 3061 3062 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3063 struct sk_buff *skb, 3064 bool all_slaves); 3065 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3066 struct sock *sk); 3067 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3068 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3069 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3070 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 3071 int dev_restart(struct net_device *dev); 3072 3073 3074 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3075 unsigned short type, 3076 const void *daddr, const void *saddr, 3077 unsigned int len) 3078 { 3079 if (!dev->header_ops || !dev->header_ops->create) 3080 return 0; 3081 3082 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3083 } 3084 3085 static inline int dev_parse_header(const struct sk_buff *skb, 3086 unsigned char *haddr) 3087 { 3088 const struct net_device *dev = skb->dev; 3089 3090 if (!dev->header_ops || !dev->header_ops->parse) 3091 return 0; 3092 return dev->header_ops->parse(skb, haddr); 3093 } 3094 3095 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3096 { 3097 const struct net_device *dev = skb->dev; 3098 3099 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3100 return 0; 3101 return dev->header_ops->parse_protocol(skb); 3102 } 3103 3104 /* ll_header must have at least hard_header_len allocated */ 3105 static inline bool dev_validate_header(const struct net_device *dev, 3106 char *ll_header, int len) 3107 { 3108 if (likely(len >= dev->hard_header_len)) 3109 return true; 3110 if (len < dev->min_header_len) 3111 return false; 3112 3113 if (capable(CAP_SYS_RAWIO)) { 3114 memset(ll_header + len, 0, dev->hard_header_len - len); 3115 return true; 3116 } 3117 3118 if (dev->header_ops && dev->header_ops->validate) 3119 return dev->header_ops->validate(ll_header, len); 3120 3121 return false; 3122 } 3123 3124 static inline bool dev_has_header(const struct net_device *dev) 3125 { 3126 return dev->header_ops && dev->header_ops->create; 3127 } 3128 3129 /* 3130 * Incoming packets are placed on per-CPU queues 3131 */ 3132 struct softnet_data { 3133 struct list_head poll_list; 3134 struct sk_buff_head process_queue; 3135 3136 /* stats */ 3137 unsigned int processed; 3138 unsigned int time_squeeze; 3139 #ifdef CONFIG_RPS 3140 struct softnet_data *rps_ipi_list; 3141 #endif 3142 #ifdef CONFIG_NET_FLOW_LIMIT 3143 struct sd_flow_limit __rcu *flow_limit; 3144 #endif 3145 struct Qdisc *output_queue; 3146 struct Qdisc **output_queue_tailp; 3147 struct sk_buff *completion_queue; 3148 #ifdef CONFIG_XFRM_OFFLOAD 3149 struct sk_buff_head xfrm_backlog; 3150 #endif 3151 /* written and read only by owning cpu: */ 3152 struct { 3153 u16 recursion; 3154 u8 more; 3155 #ifdef CONFIG_NET_EGRESS 3156 u8 skip_txqueue; 3157 #endif 3158 } xmit; 3159 #ifdef CONFIG_RPS 3160 /* input_queue_head should be written by cpu owning this struct, 3161 * and only read by other cpus. Worth using a cache line. 3162 */ 3163 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3164 3165 /* Elements below can be accessed between CPUs for RPS/RFS */ 3166 call_single_data_t csd ____cacheline_aligned_in_smp; 3167 struct softnet_data *rps_ipi_next; 3168 unsigned int cpu; 3169 unsigned int input_queue_tail; 3170 #endif 3171 unsigned int received_rps; 3172 unsigned int dropped; 3173 struct sk_buff_head input_pkt_queue; 3174 struct napi_struct backlog; 3175 3176 /* Another possibly contended cache line */ 3177 spinlock_t defer_lock ____cacheline_aligned_in_smp; 3178 int defer_count; 3179 int defer_ipi_scheduled; 3180 struct sk_buff *defer_list; 3181 call_single_data_t defer_csd; 3182 }; 3183 3184 static inline void input_queue_head_incr(struct softnet_data *sd) 3185 { 3186 #ifdef CONFIG_RPS 3187 sd->input_queue_head++; 3188 #endif 3189 } 3190 3191 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3192 unsigned int *qtail) 3193 { 3194 #ifdef CONFIG_RPS 3195 *qtail = ++sd->input_queue_tail; 3196 #endif 3197 } 3198 3199 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3200 3201 static inline int dev_recursion_level(void) 3202 { 3203 return this_cpu_read(softnet_data.xmit.recursion); 3204 } 3205 3206 #define XMIT_RECURSION_LIMIT 8 3207 static inline bool dev_xmit_recursion(void) 3208 { 3209 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3210 XMIT_RECURSION_LIMIT); 3211 } 3212 3213 static inline void dev_xmit_recursion_inc(void) 3214 { 3215 __this_cpu_inc(softnet_data.xmit.recursion); 3216 } 3217 3218 static inline void dev_xmit_recursion_dec(void) 3219 { 3220 __this_cpu_dec(softnet_data.xmit.recursion); 3221 } 3222 3223 void __netif_schedule(struct Qdisc *q); 3224 void netif_schedule_queue(struct netdev_queue *txq); 3225 3226 static inline void netif_tx_schedule_all(struct net_device *dev) 3227 { 3228 unsigned int i; 3229 3230 for (i = 0; i < dev->num_tx_queues; i++) 3231 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3232 } 3233 3234 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3235 { 3236 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3237 } 3238 3239 /** 3240 * netif_start_queue - allow transmit 3241 * @dev: network device 3242 * 3243 * Allow upper layers to call the device hard_start_xmit routine. 3244 */ 3245 static inline void netif_start_queue(struct net_device *dev) 3246 { 3247 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3248 } 3249 3250 static inline void netif_tx_start_all_queues(struct net_device *dev) 3251 { 3252 unsigned int i; 3253 3254 for (i = 0; i < dev->num_tx_queues; i++) { 3255 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3256 netif_tx_start_queue(txq); 3257 } 3258 } 3259 3260 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3261 3262 /** 3263 * netif_wake_queue - restart transmit 3264 * @dev: network device 3265 * 3266 * Allow upper layers to call the device hard_start_xmit routine. 3267 * Used for flow control when transmit resources are available. 3268 */ 3269 static inline void netif_wake_queue(struct net_device *dev) 3270 { 3271 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3272 } 3273 3274 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3275 { 3276 unsigned int i; 3277 3278 for (i = 0; i < dev->num_tx_queues; i++) { 3279 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3280 netif_tx_wake_queue(txq); 3281 } 3282 } 3283 3284 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3285 { 3286 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3287 } 3288 3289 /** 3290 * netif_stop_queue - stop transmitted packets 3291 * @dev: network device 3292 * 3293 * Stop upper layers calling the device hard_start_xmit routine. 3294 * Used for flow control when transmit resources are unavailable. 3295 */ 3296 static inline void netif_stop_queue(struct net_device *dev) 3297 { 3298 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3299 } 3300 3301 void netif_tx_stop_all_queues(struct net_device *dev); 3302 3303 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3304 { 3305 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3306 } 3307 3308 /** 3309 * netif_queue_stopped - test if transmit queue is flowblocked 3310 * @dev: network device 3311 * 3312 * Test if transmit queue on device is currently unable to send. 3313 */ 3314 static inline bool netif_queue_stopped(const struct net_device *dev) 3315 { 3316 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3317 } 3318 3319 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3320 { 3321 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3322 } 3323 3324 static inline bool 3325 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3326 { 3327 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3328 } 3329 3330 static inline bool 3331 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3332 { 3333 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3334 } 3335 3336 /** 3337 * netdev_queue_set_dql_min_limit - set dql minimum limit 3338 * @dev_queue: pointer to transmit queue 3339 * @min_limit: dql minimum limit 3340 * 3341 * Forces xmit_more() to return true until the minimum threshold 3342 * defined by @min_limit is reached (or until the tx queue is 3343 * empty). Warning: to be use with care, misuse will impact the 3344 * latency. 3345 */ 3346 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3347 unsigned int min_limit) 3348 { 3349 #ifdef CONFIG_BQL 3350 dev_queue->dql.min_limit = min_limit; 3351 #endif 3352 } 3353 3354 /** 3355 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3356 * @dev_queue: pointer to transmit queue 3357 * 3358 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3359 * to give appropriate hint to the CPU. 3360 */ 3361 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3362 { 3363 #ifdef CONFIG_BQL 3364 prefetchw(&dev_queue->dql.num_queued); 3365 #endif 3366 } 3367 3368 /** 3369 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3370 * @dev_queue: pointer to transmit queue 3371 * 3372 * BQL enabled drivers might use this helper in their TX completion path, 3373 * to give appropriate hint to the CPU. 3374 */ 3375 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3376 { 3377 #ifdef CONFIG_BQL 3378 prefetchw(&dev_queue->dql.limit); 3379 #endif 3380 } 3381 3382 /** 3383 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3384 * @dev_queue: network device queue 3385 * @bytes: number of bytes queued to the device queue 3386 * 3387 * Report the number of bytes queued for sending/completion to the network 3388 * device hardware queue. @bytes should be a good approximation and should 3389 * exactly match netdev_completed_queue() @bytes. 3390 * This is typically called once per packet, from ndo_start_xmit(). 3391 */ 3392 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3393 unsigned int bytes) 3394 { 3395 #ifdef CONFIG_BQL 3396 dql_queued(&dev_queue->dql, bytes); 3397 3398 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3399 return; 3400 3401 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3402 3403 /* 3404 * The XOFF flag must be set before checking the dql_avail below, 3405 * because in netdev_tx_completed_queue we update the dql_completed 3406 * before checking the XOFF flag. 3407 */ 3408 smp_mb(); 3409 3410 /* check again in case another CPU has just made room avail */ 3411 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3412 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3413 #endif 3414 } 3415 3416 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3417 * that they should not test BQL status themselves. 3418 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3419 * skb of a batch. 3420 * Returns true if the doorbell must be used to kick the NIC. 3421 */ 3422 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3423 unsigned int bytes, 3424 bool xmit_more) 3425 { 3426 if (xmit_more) { 3427 #ifdef CONFIG_BQL 3428 dql_queued(&dev_queue->dql, bytes); 3429 #endif 3430 return netif_tx_queue_stopped(dev_queue); 3431 } 3432 netdev_tx_sent_queue(dev_queue, bytes); 3433 return true; 3434 } 3435 3436 /** 3437 * netdev_sent_queue - report the number of bytes queued to hardware 3438 * @dev: network device 3439 * @bytes: number of bytes queued to the hardware device queue 3440 * 3441 * Report the number of bytes queued for sending/completion to the network 3442 * device hardware queue#0. @bytes should be a good approximation and should 3443 * exactly match netdev_completed_queue() @bytes. 3444 * This is typically called once per packet, from ndo_start_xmit(). 3445 */ 3446 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3447 { 3448 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3449 } 3450 3451 static inline bool __netdev_sent_queue(struct net_device *dev, 3452 unsigned int bytes, 3453 bool xmit_more) 3454 { 3455 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3456 xmit_more); 3457 } 3458 3459 /** 3460 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3461 * @dev_queue: network device queue 3462 * @pkts: number of packets (currently ignored) 3463 * @bytes: number of bytes dequeued from the device queue 3464 * 3465 * Must be called at most once per TX completion round (and not per 3466 * individual packet), so that BQL can adjust its limits appropriately. 3467 */ 3468 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3469 unsigned int pkts, unsigned int bytes) 3470 { 3471 #ifdef CONFIG_BQL 3472 if (unlikely(!bytes)) 3473 return; 3474 3475 dql_completed(&dev_queue->dql, bytes); 3476 3477 /* 3478 * Without the memory barrier there is a small possiblity that 3479 * netdev_tx_sent_queue will miss the update and cause the queue to 3480 * be stopped forever 3481 */ 3482 smp_mb(); 3483 3484 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3485 return; 3486 3487 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3488 netif_schedule_queue(dev_queue); 3489 #endif 3490 } 3491 3492 /** 3493 * netdev_completed_queue - report bytes and packets completed by device 3494 * @dev: network device 3495 * @pkts: actual number of packets sent over the medium 3496 * @bytes: actual number of bytes sent over the medium 3497 * 3498 * Report the number of bytes and packets transmitted by the network device 3499 * hardware queue over the physical medium, @bytes must exactly match the 3500 * @bytes amount passed to netdev_sent_queue() 3501 */ 3502 static inline void netdev_completed_queue(struct net_device *dev, 3503 unsigned int pkts, unsigned int bytes) 3504 { 3505 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3506 } 3507 3508 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3509 { 3510 #ifdef CONFIG_BQL 3511 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3512 dql_reset(&q->dql); 3513 #endif 3514 } 3515 3516 /** 3517 * netdev_reset_queue - reset the packets and bytes count of a network device 3518 * @dev_queue: network device 3519 * 3520 * Reset the bytes and packet count of a network device and clear the 3521 * software flow control OFF bit for this network device 3522 */ 3523 static inline void netdev_reset_queue(struct net_device *dev_queue) 3524 { 3525 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3526 } 3527 3528 /** 3529 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3530 * @dev: network device 3531 * @queue_index: given tx queue index 3532 * 3533 * Returns 0 if given tx queue index >= number of device tx queues, 3534 * otherwise returns the originally passed tx queue index. 3535 */ 3536 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3537 { 3538 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3539 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3540 dev->name, queue_index, 3541 dev->real_num_tx_queues); 3542 return 0; 3543 } 3544 3545 return queue_index; 3546 } 3547 3548 /** 3549 * netif_running - test if up 3550 * @dev: network device 3551 * 3552 * Test if the device has been brought up. 3553 */ 3554 static inline bool netif_running(const struct net_device *dev) 3555 { 3556 return test_bit(__LINK_STATE_START, &dev->state); 3557 } 3558 3559 /* 3560 * Routines to manage the subqueues on a device. We only need start, 3561 * stop, and a check if it's stopped. All other device management is 3562 * done at the overall netdevice level. 3563 * Also test the device if we're multiqueue. 3564 */ 3565 3566 /** 3567 * netif_start_subqueue - allow sending packets on subqueue 3568 * @dev: network device 3569 * @queue_index: sub queue index 3570 * 3571 * Start individual transmit queue of a device with multiple transmit queues. 3572 */ 3573 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3574 { 3575 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3576 3577 netif_tx_start_queue(txq); 3578 } 3579 3580 /** 3581 * netif_stop_subqueue - stop sending packets on subqueue 3582 * @dev: network device 3583 * @queue_index: sub queue index 3584 * 3585 * Stop individual transmit queue of a device with multiple transmit queues. 3586 */ 3587 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3588 { 3589 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3590 netif_tx_stop_queue(txq); 3591 } 3592 3593 /** 3594 * __netif_subqueue_stopped - test status of subqueue 3595 * @dev: network device 3596 * @queue_index: sub queue index 3597 * 3598 * Check individual transmit queue of a device with multiple transmit queues. 3599 */ 3600 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3601 u16 queue_index) 3602 { 3603 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3604 3605 return netif_tx_queue_stopped(txq); 3606 } 3607 3608 /** 3609 * netif_subqueue_stopped - test status of subqueue 3610 * @dev: network device 3611 * @skb: sub queue buffer pointer 3612 * 3613 * Check individual transmit queue of a device with multiple transmit queues. 3614 */ 3615 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3616 struct sk_buff *skb) 3617 { 3618 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3619 } 3620 3621 /** 3622 * netif_wake_subqueue - allow sending packets on subqueue 3623 * @dev: network device 3624 * @queue_index: sub queue index 3625 * 3626 * Resume individual transmit queue of a device with multiple transmit queues. 3627 */ 3628 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3629 { 3630 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3631 3632 netif_tx_wake_queue(txq); 3633 } 3634 3635 #ifdef CONFIG_XPS 3636 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3637 u16 index); 3638 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3639 u16 index, enum xps_map_type type); 3640 3641 /** 3642 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3643 * @j: CPU/Rx queue index 3644 * @mask: bitmask of all cpus/rx queues 3645 * @nr_bits: number of bits in the bitmask 3646 * 3647 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3648 */ 3649 static inline bool netif_attr_test_mask(unsigned long j, 3650 const unsigned long *mask, 3651 unsigned int nr_bits) 3652 { 3653 cpu_max_bits_warn(j, nr_bits); 3654 return test_bit(j, mask); 3655 } 3656 3657 /** 3658 * netif_attr_test_online - Test for online CPU/Rx queue 3659 * @j: CPU/Rx queue index 3660 * @online_mask: bitmask for CPUs/Rx queues that are online 3661 * @nr_bits: number of bits in the bitmask 3662 * 3663 * Returns true if a CPU/Rx queue is online. 3664 */ 3665 static inline bool netif_attr_test_online(unsigned long j, 3666 const unsigned long *online_mask, 3667 unsigned int nr_bits) 3668 { 3669 cpu_max_bits_warn(j, nr_bits); 3670 3671 if (online_mask) 3672 return test_bit(j, online_mask); 3673 3674 return (j < nr_bits); 3675 } 3676 3677 /** 3678 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3679 * @n: CPU/Rx queue index 3680 * @srcp: the cpumask/Rx queue mask pointer 3681 * @nr_bits: number of bits in the bitmask 3682 * 3683 * Returns >= nr_bits if no further CPUs/Rx queues set. 3684 */ 3685 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3686 unsigned int nr_bits) 3687 { 3688 /* -1 is a legal arg here. */ 3689 if (n != -1) 3690 cpu_max_bits_warn(n, nr_bits); 3691 3692 if (srcp) 3693 return find_next_bit(srcp, nr_bits, n + 1); 3694 3695 return n + 1; 3696 } 3697 3698 /** 3699 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3700 * @n: CPU/Rx queue index 3701 * @src1p: the first CPUs/Rx queues mask pointer 3702 * @src2p: the second CPUs/Rx queues mask pointer 3703 * @nr_bits: number of bits in the bitmask 3704 * 3705 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3706 */ 3707 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3708 const unsigned long *src2p, 3709 unsigned int nr_bits) 3710 { 3711 /* -1 is a legal arg here. */ 3712 if (n != -1) 3713 cpu_max_bits_warn(n, nr_bits); 3714 3715 if (src1p && src2p) 3716 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3717 else if (src1p) 3718 return find_next_bit(src1p, nr_bits, n + 1); 3719 else if (src2p) 3720 return find_next_bit(src2p, nr_bits, n + 1); 3721 3722 return n + 1; 3723 } 3724 #else 3725 static inline int netif_set_xps_queue(struct net_device *dev, 3726 const struct cpumask *mask, 3727 u16 index) 3728 { 3729 return 0; 3730 } 3731 3732 static inline int __netif_set_xps_queue(struct net_device *dev, 3733 const unsigned long *mask, 3734 u16 index, enum xps_map_type type) 3735 { 3736 return 0; 3737 } 3738 #endif 3739 3740 /** 3741 * netif_is_multiqueue - test if device has multiple transmit queues 3742 * @dev: network device 3743 * 3744 * Check if device has multiple transmit queues 3745 */ 3746 static inline bool netif_is_multiqueue(const struct net_device *dev) 3747 { 3748 return dev->num_tx_queues > 1; 3749 } 3750 3751 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3752 3753 #ifdef CONFIG_SYSFS 3754 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3755 #else 3756 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3757 unsigned int rxqs) 3758 { 3759 dev->real_num_rx_queues = rxqs; 3760 return 0; 3761 } 3762 #endif 3763 int netif_set_real_num_queues(struct net_device *dev, 3764 unsigned int txq, unsigned int rxq); 3765 3766 static inline struct netdev_rx_queue * 3767 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3768 { 3769 return dev->_rx + rxq; 3770 } 3771 3772 #ifdef CONFIG_SYSFS 3773 static inline unsigned int get_netdev_rx_queue_index( 3774 struct netdev_rx_queue *queue) 3775 { 3776 struct net_device *dev = queue->dev; 3777 int index = queue - dev->_rx; 3778 3779 BUG_ON(index >= dev->num_rx_queues); 3780 return index; 3781 } 3782 #endif 3783 3784 int netif_get_num_default_rss_queues(void); 3785 3786 enum skb_free_reason { 3787 SKB_REASON_CONSUMED, 3788 SKB_REASON_DROPPED, 3789 }; 3790 3791 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3792 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3793 3794 /* 3795 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3796 * interrupt context or with hardware interrupts being disabled. 3797 * (in_hardirq() || irqs_disabled()) 3798 * 3799 * We provide four helpers that can be used in following contexts : 3800 * 3801 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3802 * replacing kfree_skb(skb) 3803 * 3804 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3805 * Typically used in place of consume_skb(skb) in TX completion path 3806 * 3807 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3808 * replacing kfree_skb(skb) 3809 * 3810 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3811 * and consumed a packet. Used in place of consume_skb(skb) 3812 */ 3813 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3814 { 3815 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3816 } 3817 3818 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3819 { 3820 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3821 } 3822 3823 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3824 { 3825 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3826 } 3827 3828 static inline void dev_consume_skb_any(struct sk_buff *skb) 3829 { 3830 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3831 } 3832 3833 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3834 struct bpf_prog *xdp_prog); 3835 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3836 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3837 int netif_rx(struct sk_buff *skb); 3838 int __netif_rx(struct sk_buff *skb); 3839 3840 int netif_receive_skb(struct sk_buff *skb); 3841 int netif_receive_skb_core(struct sk_buff *skb); 3842 void netif_receive_skb_list_internal(struct list_head *head); 3843 void netif_receive_skb_list(struct list_head *head); 3844 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3845 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3846 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3847 void napi_get_frags_check(struct napi_struct *napi); 3848 gro_result_t napi_gro_frags(struct napi_struct *napi); 3849 struct packet_offload *gro_find_receive_by_type(__be16 type); 3850 struct packet_offload *gro_find_complete_by_type(__be16 type); 3851 3852 static inline void napi_free_frags(struct napi_struct *napi) 3853 { 3854 kfree_skb(napi->skb); 3855 napi->skb = NULL; 3856 } 3857 3858 bool netdev_is_rx_handler_busy(struct net_device *dev); 3859 int netdev_rx_handler_register(struct net_device *dev, 3860 rx_handler_func_t *rx_handler, 3861 void *rx_handler_data); 3862 void netdev_rx_handler_unregister(struct net_device *dev); 3863 3864 bool dev_valid_name(const char *name); 3865 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 3866 { 3867 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 3868 } 3869 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 3870 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 3871 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3872 void __user *data, bool *need_copyout); 3873 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 3874 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 3875 unsigned int dev_get_flags(const struct net_device *); 3876 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3877 struct netlink_ext_ack *extack); 3878 int dev_change_flags(struct net_device *dev, unsigned int flags, 3879 struct netlink_ext_ack *extack); 3880 int dev_set_alias(struct net_device *, const char *, size_t); 3881 int dev_get_alias(const struct net_device *, char *, size_t); 3882 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 3883 const char *pat, int new_ifindex); 3884 static inline 3885 int dev_change_net_namespace(struct net_device *dev, struct net *net, 3886 const char *pat) 3887 { 3888 return __dev_change_net_namespace(dev, net, pat, 0); 3889 } 3890 int __dev_set_mtu(struct net_device *, int); 3891 int dev_set_mtu(struct net_device *, int); 3892 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3893 struct netlink_ext_ack *extack); 3894 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3895 struct netlink_ext_ack *extack); 3896 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 3897 struct netlink_ext_ack *extack); 3898 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 3899 int dev_get_port_parent_id(struct net_device *dev, 3900 struct netdev_phys_item_id *ppid, bool recurse); 3901 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3902 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3903 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3904 struct netdev_queue *txq, int *ret); 3905 3906 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3907 u8 dev_xdp_prog_count(struct net_device *dev); 3908 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3909 3910 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3911 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3912 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 3913 bool is_skb_forwardable(const struct net_device *dev, 3914 const struct sk_buff *skb); 3915 3916 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 3917 const struct sk_buff *skb, 3918 const bool check_mtu) 3919 { 3920 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 3921 unsigned int len; 3922 3923 if (!(dev->flags & IFF_UP)) 3924 return false; 3925 3926 if (!check_mtu) 3927 return true; 3928 3929 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 3930 if (skb->len <= len) 3931 return true; 3932 3933 /* if TSO is enabled, we don't care about the length as the packet 3934 * could be forwarded without being segmented before 3935 */ 3936 if (skb_is_gso(skb)) 3937 return true; 3938 3939 return false; 3940 } 3941 3942 struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev); 3943 3944 static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev) 3945 { 3946 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */ 3947 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats); 3948 3949 if (likely(p)) 3950 return p; 3951 3952 return netdev_core_stats_alloc(dev); 3953 } 3954 3955 #define DEV_CORE_STATS_INC(FIELD) \ 3956 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 3957 { \ 3958 struct net_device_core_stats __percpu *p; \ 3959 \ 3960 p = dev_core_stats(dev); \ 3961 if (p) \ 3962 this_cpu_inc(p->FIELD); \ 3963 } 3964 DEV_CORE_STATS_INC(rx_dropped) 3965 DEV_CORE_STATS_INC(tx_dropped) 3966 DEV_CORE_STATS_INC(rx_nohandler) 3967 DEV_CORE_STATS_INC(rx_otherhost_dropped) 3968 3969 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3970 struct sk_buff *skb, 3971 const bool check_mtu) 3972 { 3973 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3974 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 3975 dev_core_stats_rx_dropped_inc(dev); 3976 kfree_skb(skb); 3977 return NET_RX_DROP; 3978 } 3979 3980 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 3981 skb->priority = 0; 3982 return 0; 3983 } 3984 3985 bool dev_nit_active(struct net_device *dev); 3986 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3987 3988 static inline void __dev_put(struct net_device *dev) 3989 { 3990 if (dev) { 3991 #ifdef CONFIG_PCPU_DEV_REFCNT 3992 this_cpu_dec(*dev->pcpu_refcnt); 3993 #else 3994 refcount_dec(&dev->dev_refcnt); 3995 #endif 3996 } 3997 } 3998 3999 static inline void __dev_hold(struct net_device *dev) 4000 { 4001 if (dev) { 4002 #ifdef CONFIG_PCPU_DEV_REFCNT 4003 this_cpu_inc(*dev->pcpu_refcnt); 4004 #else 4005 refcount_inc(&dev->dev_refcnt); 4006 #endif 4007 } 4008 } 4009 4010 static inline void __netdev_tracker_alloc(struct net_device *dev, 4011 netdevice_tracker *tracker, 4012 gfp_t gfp) 4013 { 4014 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4015 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 4016 #endif 4017 } 4018 4019 /* netdev_tracker_alloc() can upgrade a prior untracked reference 4020 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4021 */ 4022 static inline void netdev_tracker_alloc(struct net_device *dev, 4023 netdevice_tracker *tracker, gfp_t gfp) 4024 { 4025 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4026 refcount_dec(&dev->refcnt_tracker.no_tracker); 4027 __netdev_tracker_alloc(dev, tracker, gfp); 4028 #endif 4029 } 4030 4031 static inline void netdev_tracker_free(struct net_device *dev, 4032 netdevice_tracker *tracker) 4033 { 4034 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4035 ref_tracker_free(&dev->refcnt_tracker, tracker); 4036 #endif 4037 } 4038 4039 static inline void netdev_hold(struct net_device *dev, 4040 netdevice_tracker *tracker, gfp_t gfp) 4041 { 4042 if (dev) { 4043 __dev_hold(dev); 4044 __netdev_tracker_alloc(dev, tracker, gfp); 4045 } 4046 } 4047 4048 static inline void netdev_put(struct net_device *dev, 4049 netdevice_tracker *tracker) 4050 { 4051 if (dev) { 4052 netdev_tracker_free(dev, tracker); 4053 __dev_put(dev); 4054 } 4055 } 4056 4057 /** 4058 * dev_hold - get reference to device 4059 * @dev: network device 4060 * 4061 * Hold reference to device to keep it from being freed. 4062 * Try using netdev_hold() instead. 4063 */ 4064 static inline void dev_hold(struct net_device *dev) 4065 { 4066 netdev_hold(dev, NULL, GFP_ATOMIC); 4067 } 4068 4069 /** 4070 * dev_put - release reference to device 4071 * @dev: network device 4072 * 4073 * Release reference to device to allow it to be freed. 4074 * Try using netdev_put() instead. 4075 */ 4076 static inline void dev_put(struct net_device *dev) 4077 { 4078 netdev_put(dev, NULL); 4079 } 4080 4081 static inline void netdev_ref_replace(struct net_device *odev, 4082 struct net_device *ndev, 4083 netdevice_tracker *tracker, 4084 gfp_t gfp) 4085 { 4086 if (odev) 4087 netdev_tracker_free(odev, tracker); 4088 4089 __dev_hold(ndev); 4090 __dev_put(odev); 4091 4092 if (ndev) 4093 __netdev_tracker_alloc(ndev, tracker, gfp); 4094 } 4095 4096 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4097 * and _off may be called from IRQ context, but it is caller 4098 * who is responsible for serialization of these calls. 4099 * 4100 * The name carrier is inappropriate, these functions should really be 4101 * called netif_lowerlayer_*() because they represent the state of any 4102 * kind of lower layer not just hardware media. 4103 */ 4104 void linkwatch_fire_event(struct net_device *dev); 4105 4106 /** 4107 * netif_carrier_ok - test if carrier present 4108 * @dev: network device 4109 * 4110 * Check if carrier is present on device 4111 */ 4112 static inline bool netif_carrier_ok(const struct net_device *dev) 4113 { 4114 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4115 } 4116 4117 unsigned long dev_trans_start(struct net_device *dev); 4118 4119 void __netdev_watchdog_up(struct net_device *dev); 4120 4121 void netif_carrier_on(struct net_device *dev); 4122 void netif_carrier_off(struct net_device *dev); 4123 void netif_carrier_event(struct net_device *dev); 4124 4125 /** 4126 * netif_dormant_on - mark device as dormant. 4127 * @dev: network device 4128 * 4129 * Mark device as dormant (as per RFC2863). 4130 * 4131 * The dormant state indicates that the relevant interface is not 4132 * actually in a condition to pass packets (i.e., it is not 'up') but is 4133 * in a "pending" state, waiting for some external event. For "on- 4134 * demand" interfaces, this new state identifies the situation where the 4135 * interface is waiting for events to place it in the up state. 4136 */ 4137 static inline void netif_dormant_on(struct net_device *dev) 4138 { 4139 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4140 linkwatch_fire_event(dev); 4141 } 4142 4143 /** 4144 * netif_dormant_off - set device as not dormant. 4145 * @dev: network device 4146 * 4147 * Device is not in dormant state. 4148 */ 4149 static inline void netif_dormant_off(struct net_device *dev) 4150 { 4151 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4152 linkwatch_fire_event(dev); 4153 } 4154 4155 /** 4156 * netif_dormant - test if device is dormant 4157 * @dev: network device 4158 * 4159 * Check if device is dormant. 4160 */ 4161 static inline bool netif_dormant(const struct net_device *dev) 4162 { 4163 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4164 } 4165 4166 4167 /** 4168 * netif_testing_on - mark device as under test. 4169 * @dev: network device 4170 * 4171 * Mark device as under test (as per RFC2863). 4172 * 4173 * The testing state indicates that some test(s) must be performed on 4174 * the interface. After completion, of the test, the interface state 4175 * will change to up, dormant, or down, as appropriate. 4176 */ 4177 static inline void netif_testing_on(struct net_device *dev) 4178 { 4179 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4180 linkwatch_fire_event(dev); 4181 } 4182 4183 /** 4184 * netif_testing_off - set device as not under test. 4185 * @dev: network device 4186 * 4187 * Device is not in testing state. 4188 */ 4189 static inline void netif_testing_off(struct net_device *dev) 4190 { 4191 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4192 linkwatch_fire_event(dev); 4193 } 4194 4195 /** 4196 * netif_testing - test if device is under test 4197 * @dev: network device 4198 * 4199 * Check if device is under test 4200 */ 4201 static inline bool netif_testing(const struct net_device *dev) 4202 { 4203 return test_bit(__LINK_STATE_TESTING, &dev->state); 4204 } 4205 4206 4207 /** 4208 * netif_oper_up - test if device is operational 4209 * @dev: network device 4210 * 4211 * Check if carrier is operational 4212 */ 4213 static inline bool netif_oper_up(const struct net_device *dev) 4214 { 4215 return (dev->operstate == IF_OPER_UP || 4216 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4217 } 4218 4219 /** 4220 * netif_device_present - is device available or removed 4221 * @dev: network device 4222 * 4223 * Check if device has not been removed from system. 4224 */ 4225 static inline bool netif_device_present(const struct net_device *dev) 4226 { 4227 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4228 } 4229 4230 void netif_device_detach(struct net_device *dev); 4231 4232 void netif_device_attach(struct net_device *dev); 4233 4234 /* 4235 * Network interface message level settings 4236 */ 4237 4238 enum { 4239 NETIF_MSG_DRV_BIT, 4240 NETIF_MSG_PROBE_BIT, 4241 NETIF_MSG_LINK_BIT, 4242 NETIF_MSG_TIMER_BIT, 4243 NETIF_MSG_IFDOWN_BIT, 4244 NETIF_MSG_IFUP_BIT, 4245 NETIF_MSG_RX_ERR_BIT, 4246 NETIF_MSG_TX_ERR_BIT, 4247 NETIF_MSG_TX_QUEUED_BIT, 4248 NETIF_MSG_INTR_BIT, 4249 NETIF_MSG_TX_DONE_BIT, 4250 NETIF_MSG_RX_STATUS_BIT, 4251 NETIF_MSG_PKTDATA_BIT, 4252 NETIF_MSG_HW_BIT, 4253 NETIF_MSG_WOL_BIT, 4254 4255 /* When you add a new bit above, update netif_msg_class_names array 4256 * in net/ethtool/common.c 4257 */ 4258 NETIF_MSG_CLASS_COUNT, 4259 }; 4260 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4261 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4262 4263 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4264 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4265 4266 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4267 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4268 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4269 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4270 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4271 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4272 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4273 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4274 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4275 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4276 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4277 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4278 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4279 #define NETIF_MSG_HW __NETIF_MSG(HW) 4280 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4281 4282 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4283 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4284 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4285 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4286 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4287 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4288 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4289 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4290 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4291 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4292 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4293 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4294 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4295 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4296 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4297 4298 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4299 { 4300 /* use default */ 4301 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4302 return default_msg_enable_bits; 4303 if (debug_value == 0) /* no output */ 4304 return 0; 4305 /* set low N bits */ 4306 return (1U << debug_value) - 1; 4307 } 4308 4309 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4310 { 4311 spin_lock(&txq->_xmit_lock); 4312 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4313 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4314 } 4315 4316 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4317 { 4318 __acquire(&txq->_xmit_lock); 4319 return true; 4320 } 4321 4322 static inline void __netif_tx_release(struct netdev_queue *txq) 4323 { 4324 __release(&txq->_xmit_lock); 4325 } 4326 4327 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4328 { 4329 spin_lock_bh(&txq->_xmit_lock); 4330 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4331 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4332 } 4333 4334 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4335 { 4336 bool ok = spin_trylock(&txq->_xmit_lock); 4337 4338 if (likely(ok)) { 4339 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4340 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4341 } 4342 return ok; 4343 } 4344 4345 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4346 { 4347 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4348 WRITE_ONCE(txq->xmit_lock_owner, -1); 4349 spin_unlock(&txq->_xmit_lock); 4350 } 4351 4352 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4353 { 4354 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4355 WRITE_ONCE(txq->xmit_lock_owner, -1); 4356 spin_unlock_bh(&txq->_xmit_lock); 4357 } 4358 4359 /* 4360 * txq->trans_start can be read locklessly from dev_watchdog() 4361 */ 4362 static inline void txq_trans_update(struct netdev_queue *txq) 4363 { 4364 if (txq->xmit_lock_owner != -1) 4365 WRITE_ONCE(txq->trans_start, jiffies); 4366 } 4367 4368 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4369 { 4370 unsigned long now = jiffies; 4371 4372 if (READ_ONCE(txq->trans_start) != now) 4373 WRITE_ONCE(txq->trans_start, now); 4374 } 4375 4376 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4377 static inline void netif_trans_update(struct net_device *dev) 4378 { 4379 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4380 4381 txq_trans_cond_update(txq); 4382 } 4383 4384 /** 4385 * netif_tx_lock - grab network device transmit lock 4386 * @dev: network device 4387 * 4388 * Get network device transmit lock 4389 */ 4390 void netif_tx_lock(struct net_device *dev); 4391 4392 static inline void netif_tx_lock_bh(struct net_device *dev) 4393 { 4394 local_bh_disable(); 4395 netif_tx_lock(dev); 4396 } 4397 4398 void netif_tx_unlock(struct net_device *dev); 4399 4400 static inline void netif_tx_unlock_bh(struct net_device *dev) 4401 { 4402 netif_tx_unlock(dev); 4403 local_bh_enable(); 4404 } 4405 4406 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4407 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4408 __netif_tx_lock(txq, cpu); \ 4409 } else { \ 4410 __netif_tx_acquire(txq); \ 4411 } \ 4412 } 4413 4414 #define HARD_TX_TRYLOCK(dev, txq) \ 4415 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4416 __netif_tx_trylock(txq) : \ 4417 __netif_tx_acquire(txq)) 4418 4419 #define HARD_TX_UNLOCK(dev, txq) { \ 4420 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4421 __netif_tx_unlock(txq); \ 4422 } else { \ 4423 __netif_tx_release(txq); \ 4424 } \ 4425 } 4426 4427 static inline void netif_tx_disable(struct net_device *dev) 4428 { 4429 unsigned int i; 4430 int cpu; 4431 4432 local_bh_disable(); 4433 cpu = smp_processor_id(); 4434 spin_lock(&dev->tx_global_lock); 4435 for (i = 0; i < dev->num_tx_queues; i++) { 4436 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4437 4438 __netif_tx_lock(txq, cpu); 4439 netif_tx_stop_queue(txq); 4440 __netif_tx_unlock(txq); 4441 } 4442 spin_unlock(&dev->tx_global_lock); 4443 local_bh_enable(); 4444 } 4445 4446 static inline void netif_addr_lock(struct net_device *dev) 4447 { 4448 unsigned char nest_level = 0; 4449 4450 #ifdef CONFIG_LOCKDEP 4451 nest_level = dev->nested_level; 4452 #endif 4453 spin_lock_nested(&dev->addr_list_lock, nest_level); 4454 } 4455 4456 static inline void netif_addr_lock_bh(struct net_device *dev) 4457 { 4458 unsigned char nest_level = 0; 4459 4460 #ifdef CONFIG_LOCKDEP 4461 nest_level = dev->nested_level; 4462 #endif 4463 local_bh_disable(); 4464 spin_lock_nested(&dev->addr_list_lock, nest_level); 4465 } 4466 4467 static inline void netif_addr_unlock(struct net_device *dev) 4468 { 4469 spin_unlock(&dev->addr_list_lock); 4470 } 4471 4472 static inline void netif_addr_unlock_bh(struct net_device *dev) 4473 { 4474 spin_unlock_bh(&dev->addr_list_lock); 4475 } 4476 4477 /* 4478 * dev_addrs walker. Should be used only for read access. Call with 4479 * rcu_read_lock held. 4480 */ 4481 #define for_each_dev_addr(dev, ha) \ 4482 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4483 4484 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4485 4486 void ether_setup(struct net_device *dev); 4487 4488 /* Support for loadable net-drivers */ 4489 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4490 unsigned char name_assign_type, 4491 void (*setup)(struct net_device *), 4492 unsigned int txqs, unsigned int rxqs); 4493 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4494 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4495 4496 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4497 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4498 count) 4499 4500 int register_netdev(struct net_device *dev); 4501 void unregister_netdev(struct net_device *dev); 4502 4503 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4504 4505 /* General hardware address lists handling functions */ 4506 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4507 struct netdev_hw_addr_list *from_list, int addr_len); 4508 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4509 struct netdev_hw_addr_list *from_list, int addr_len); 4510 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4511 struct net_device *dev, 4512 int (*sync)(struct net_device *, const unsigned char *), 4513 int (*unsync)(struct net_device *, 4514 const unsigned char *)); 4515 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4516 struct net_device *dev, 4517 int (*sync)(struct net_device *, 4518 const unsigned char *, int), 4519 int (*unsync)(struct net_device *, 4520 const unsigned char *, int)); 4521 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4522 struct net_device *dev, 4523 int (*unsync)(struct net_device *, 4524 const unsigned char *, int)); 4525 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4526 struct net_device *dev, 4527 int (*unsync)(struct net_device *, 4528 const unsigned char *)); 4529 void __hw_addr_init(struct netdev_hw_addr_list *list); 4530 4531 /* Functions used for device addresses handling */ 4532 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4533 const void *addr, size_t len); 4534 4535 static inline void 4536 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4537 { 4538 dev_addr_mod(dev, 0, addr, len); 4539 } 4540 4541 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4542 { 4543 __dev_addr_set(dev, addr, dev->addr_len); 4544 } 4545 4546 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4547 unsigned char addr_type); 4548 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4549 unsigned char addr_type); 4550 4551 /* Functions used for unicast addresses handling */ 4552 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4553 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4554 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4555 int dev_uc_sync(struct net_device *to, struct net_device *from); 4556 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4557 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4558 void dev_uc_flush(struct net_device *dev); 4559 void dev_uc_init(struct net_device *dev); 4560 4561 /** 4562 * __dev_uc_sync - Synchonize device's unicast list 4563 * @dev: device to sync 4564 * @sync: function to call if address should be added 4565 * @unsync: function to call if address should be removed 4566 * 4567 * Add newly added addresses to the interface, and release 4568 * addresses that have been deleted. 4569 */ 4570 static inline int __dev_uc_sync(struct net_device *dev, 4571 int (*sync)(struct net_device *, 4572 const unsigned char *), 4573 int (*unsync)(struct net_device *, 4574 const unsigned char *)) 4575 { 4576 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4577 } 4578 4579 /** 4580 * __dev_uc_unsync - Remove synchronized addresses from device 4581 * @dev: device to sync 4582 * @unsync: function to call if address should be removed 4583 * 4584 * Remove all addresses that were added to the device by dev_uc_sync(). 4585 */ 4586 static inline void __dev_uc_unsync(struct net_device *dev, 4587 int (*unsync)(struct net_device *, 4588 const unsigned char *)) 4589 { 4590 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4591 } 4592 4593 /* Functions used for multicast addresses handling */ 4594 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4595 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4596 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4597 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4598 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4599 int dev_mc_sync(struct net_device *to, struct net_device *from); 4600 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4601 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4602 void dev_mc_flush(struct net_device *dev); 4603 void dev_mc_init(struct net_device *dev); 4604 4605 /** 4606 * __dev_mc_sync - Synchonize device's multicast list 4607 * @dev: device to sync 4608 * @sync: function to call if address should be added 4609 * @unsync: function to call if address should be removed 4610 * 4611 * Add newly added addresses to the interface, and release 4612 * addresses that have been deleted. 4613 */ 4614 static inline int __dev_mc_sync(struct net_device *dev, 4615 int (*sync)(struct net_device *, 4616 const unsigned char *), 4617 int (*unsync)(struct net_device *, 4618 const unsigned char *)) 4619 { 4620 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4621 } 4622 4623 /** 4624 * __dev_mc_unsync - Remove synchronized addresses from device 4625 * @dev: device to sync 4626 * @unsync: function to call if address should be removed 4627 * 4628 * Remove all addresses that were added to the device by dev_mc_sync(). 4629 */ 4630 static inline void __dev_mc_unsync(struct net_device *dev, 4631 int (*unsync)(struct net_device *, 4632 const unsigned char *)) 4633 { 4634 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4635 } 4636 4637 /* Functions used for secondary unicast and multicast support */ 4638 void dev_set_rx_mode(struct net_device *dev); 4639 int dev_set_promiscuity(struct net_device *dev, int inc); 4640 int dev_set_allmulti(struct net_device *dev, int inc); 4641 void netdev_state_change(struct net_device *dev); 4642 void __netdev_notify_peers(struct net_device *dev); 4643 void netdev_notify_peers(struct net_device *dev); 4644 void netdev_features_change(struct net_device *dev); 4645 /* Load a device via the kmod */ 4646 void dev_load(struct net *net, const char *name); 4647 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4648 struct rtnl_link_stats64 *storage); 4649 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4650 const struct net_device_stats *netdev_stats); 4651 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4652 const struct pcpu_sw_netstats __percpu *netstats); 4653 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4654 4655 extern int netdev_max_backlog; 4656 extern int dev_rx_weight; 4657 extern int dev_tx_weight; 4658 extern int gro_normal_batch; 4659 4660 enum { 4661 NESTED_SYNC_IMM_BIT, 4662 NESTED_SYNC_TODO_BIT, 4663 }; 4664 4665 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4666 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4667 4668 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4669 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4670 4671 struct netdev_nested_priv { 4672 unsigned char flags; 4673 void *data; 4674 }; 4675 4676 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4677 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4678 struct list_head **iter); 4679 4680 /* iterate through upper list, must be called under RCU read lock */ 4681 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4682 for (iter = &(dev)->adj_list.upper, \ 4683 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4684 updev; \ 4685 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4686 4687 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4688 int (*fn)(struct net_device *upper_dev, 4689 struct netdev_nested_priv *priv), 4690 struct netdev_nested_priv *priv); 4691 4692 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4693 struct net_device *upper_dev); 4694 4695 bool netdev_has_any_upper_dev(struct net_device *dev); 4696 4697 void *netdev_lower_get_next_private(struct net_device *dev, 4698 struct list_head **iter); 4699 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4700 struct list_head **iter); 4701 4702 #define netdev_for_each_lower_private(dev, priv, iter) \ 4703 for (iter = (dev)->adj_list.lower.next, \ 4704 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4705 priv; \ 4706 priv = netdev_lower_get_next_private(dev, &(iter))) 4707 4708 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4709 for (iter = &(dev)->adj_list.lower, \ 4710 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4711 priv; \ 4712 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4713 4714 void *netdev_lower_get_next(struct net_device *dev, 4715 struct list_head **iter); 4716 4717 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4718 for (iter = (dev)->adj_list.lower.next, \ 4719 ldev = netdev_lower_get_next(dev, &(iter)); \ 4720 ldev; \ 4721 ldev = netdev_lower_get_next(dev, &(iter))) 4722 4723 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4724 struct list_head **iter); 4725 int netdev_walk_all_lower_dev(struct net_device *dev, 4726 int (*fn)(struct net_device *lower_dev, 4727 struct netdev_nested_priv *priv), 4728 struct netdev_nested_priv *priv); 4729 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4730 int (*fn)(struct net_device *lower_dev, 4731 struct netdev_nested_priv *priv), 4732 struct netdev_nested_priv *priv); 4733 4734 void *netdev_adjacent_get_private(struct list_head *adj_list); 4735 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4736 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4737 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4738 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4739 struct netlink_ext_ack *extack); 4740 int netdev_master_upper_dev_link(struct net_device *dev, 4741 struct net_device *upper_dev, 4742 void *upper_priv, void *upper_info, 4743 struct netlink_ext_ack *extack); 4744 void netdev_upper_dev_unlink(struct net_device *dev, 4745 struct net_device *upper_dev); 4746 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4747 struct net_device *new_dev, 4748 struct net_device *dev, 4749 struct netlink_ext_ack *extack); 4750 void netdev_adjacent_change_commit(struct net_device *old_dev, 4751 struct net_device *new_dev, 4752 struct net_device *dev); 4753 void netdev_adjacent_change_abort(struct net_device *old_dev, 4754 struct net_device *new_dev, 4755 struct net_device *dev); 4756 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4757 void *netdev_lower_dev_get_private(struct net_device *dev, 4758 struct net_device *lower_dev); 4759 void netdev_lower_state_changed(struct net_device *lower_dev, 4760 void *lower_state_info); 4761 4762 /* RSS keys are 40 or 52 bytes long */ 4763 #define NETDEV_RSS_KEY_LEN 52 4764 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4765 void netdev_rss_key_fill(void *buffer, size_t len); 4766 4767 int skb_checksum_help(struct sk_buff *skb); 4768 int skb_crc32c_csum_help(struct sk_buff *skb); 4769 int skb_csum_hwoffload_help(struct sk_buff *skb, 4770 const netdev_features_t features); 4771 4772 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4773 netdev_features_t features, bool tx_path); 4774 struct sk_buff *skb_eth_gso_segment(struct sk_buff *skb, 4775 netdev_features_t features, __be16 type); 4776 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4777 netdev_features_t features); 4778 4779 struct netdev_bonding_info { 4780 ifslave slave; 4781 ifbond master; 4782 }; 4783 4784 struct netdev_notifier_bonding_info { 4785 struct netdev_notifier_info info; /* must be first */ 4786 struct netdev_bonding_info bonding_info; 4787 }; 4788 4789 void netdev_bonding_info_change(struct net_device *dev, 4790 struct netdev_bonding_info *bonding_info); 4791 4792 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4793 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4794 #else 4795 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4796 const void *data) 4797 { 4798 } 4799 #endif 4800 4801 static inline 4802 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4803 { 4804 return __skb_gso_segment(skb, features, true); 4805 } 4806 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4807 4808 static inline bool can_checksum_protocol(netdev_features_t features, 4809 __be16 protocol) 4810 { 4811 if (protocol == htons(ETH_P_FCOE)) 4812 return !!(features & NETIF_F_FCOE_CRC); 4813 4814 /* Assume this is an IP checksum (not SCTP CRC) */ 4815 4816 if (features & NETIF_F_HW_CSUM) { 4817 /* Can checksum everything */ 4818 return true; 4819 } 4820 4821 switch (protocol) { 4822 case htons(ETH_P_IP): 4823 return !!(features & NETIF_F_IP_CSUM); 4824 case htons(ETH_P_IPV6): 4825 return !!(features & NETIF_F_IPV6_CSUM); 4826 default: 4827 return false; 4828 } 4829 } 4830 4831 #ifdef CONFIG_BUG 4832 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4833 #else 4834 static inline void netdev_rx_csum_fault(struct net_device *dev, 4835 struct sk_buff *skb) 4836 { 4837 } 4838 #endif 4839 /* rx skb timestamps */ 4840 void net_enable_timestamp(void); 4841 void net_disable_timestamp(void); 4842 4843 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 4844 const struct skb_shared_hwtstamps *hwtstamps, 4845 bool cycles) 4846 { 4847 const struct net_device_ops *ops = dev->netdev_ops; 4848 4849 if (ops->ndo_get_tstamp) 4850 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 4851 4852 return hwtstamps->hwtstamp; 4853 } 4854 4855 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4856 struct sk_buff *skb, struct net_device *dev, 4857 bool more) 4858 { 4859 __this_cpu_write(softnet_data.xmit.more, more); 4860 return ops->ndo_start_xmit(skb, dev); 4861 } 4862 4863 static inline bool netdev_xmit_more(void) 4864 { 4865 return __this_cpu_read(softnet_data.xmit.more); 4866 } 4867 4868 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4869 struct netdev_queue *txq, bool more) 4870 { 4871 const struct net_device_ops *ops = dev->netdev_ops; 4872 netdev_tx_t rc; 4873 4874 rc = __netdev_start_xmit(ops, skb, dev, more); 4875 if (rc == NETDEV_TX_OK) 4876 txq_trans_update(txq); 4877 4878 return rc; 4879 } 4880 4881 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4882 const void *ns); 4883 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4884 const void *ns); 4885 4886 extern const struct kobj_ns_type_operations net_ns_type_operations; 4887 4888 const char *netdev_drivername(const struct net_device *dev); 4889 4890 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4891 netdev_features_t f2) 4892 { 4893 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4894 if (f1 & NETIF_F_HW_CSUM) 4895 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4896 else 4897 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4898 } 4899 4900 return f1 & f2; 4901 } 4902 4903 static inline netdev_features_t netdev_get_wanted_features( 4904 struct net_device *dev) 4905 { 4906 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4907 } 4908 netdev_features_t netdev_increment_features(netdev_features_t all, 4909 netdev_features_t one, netdev_features_t mask); 4910 4911 /* Allow TSO being used on stacked device : 4912 * Performing the GSO segmentation before last device 4913 * is a performance improvement. 4914 */ 4915 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4916 netdev_features_t mask) 4917 { 4918 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4919 } 4920 4921 int __netdev_update_features(struct net_device *dev); 4922 void netdev_update_features(struct net_device *dev); 4923 void netdev_change_features(struct net_device *dev); 4924 4925 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4926 struct net_device *dev); 4927 4928 netdev_features_t passthru_features_check(struct sk_buff *skb, 4929 struct net_device *dev, 4930 netdev_features_t features); 4931 netdev_features_t netif_skb_features(struct sk_buff *skb); 4932 4933 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4934 { 4935 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4936 4937 /* check flags correspondence */ 4938 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4939 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4940 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4941 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4942 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4943 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4944 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4945 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4946 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4947 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4948 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4949 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4950 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4951 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4952 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4953 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4954 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4955 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4956 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4957 4958 return (features & feature) == feature; 4959 } 4960 4961 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4962 { 4963 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4964 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4965 } 4966 4967 static inline bool netif_needs_gso(struct sk_buff *skb, 4968 netdev_features_t features) 4969 { 4970 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4971 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4972 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4973 } 4974 4975 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 4976 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 4977 void netif_inherit_tso_max(struct net_device *to, 4978 const struct net_device *from); 4979 4980 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4981 int pulled_hlen, u16 mac_offset, 4982 int mac_len) 4983 { 4984 skb->protocol = protocol; 4985 skb->encapsulation = 1; 4986 skb_push(skb, pulled_hlen); 4987 skb_reset_transport_header(skb); 4988 skb->mac_header = mac_offset; 4989 skb->network_header = skb->mac_header + mac_len; 4990 skb->mac_len = mac_len; 4991 } 4992 4993 static inline bool netif_is_macsec(const struct net_device *dev) 4994 { 4995 return dev->priv_flags & IFF_MACSEC; 4996 } 4997 4998 static inline bool netif_is_macvlan(const struct net_device *dev) 4999 { 5000 return dev->priv_flags & IFF_MACVLAN; 5001 } 5002 5003 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5004 { 5005 return dev->priv_flags & IFF_MACVLAN_PORT; 5006 } 5007 5008 static inline bool netif_is_bond_master(const struct net_device *dev) 5009 { 5010 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5011 } 5012 5013 static inline bool netif_is_bond_slave(const struct net_device *dev) 5014 { 5015 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5016 } 5017 5018 static inline bool netif_supports_nofcs(struct net_device *dev) 5019 { 5020 return dev->priv_flags & IFF_SUPP_NOFCS; 5021 } 5022 5023 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5024 { 5025 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5026 } 5027 5028 static inline bool netif_is_l3_master(const struct net_device *dev) 5029 { 5030 return dev->priv_flags & IFF_L3MDEV_MASTER; 5031 } 5032 5033 static inline bool netif_is_l3_slave(const struct net_device *dev) 5034 { 5035 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5036 } 5037 5038 static inline bool netif_is_bridge_master(const struct net_device *dev) 5039 { 5040 return dev->priv_flags & IFF_EBRIDGE; 5041 } 5042 5043 static inline bool netif_is_bridge_port(const struct net_device *dev) 5044 { 5045 return dev->priv_flags & IFF_BRIDGE_PORT; 5046 } 5047 5048 static inline bool netif_is_ovs_master(const struct net_device *dev) 5049 { 5050 return dev->priv_flags & IFF_OPENVSWITCH; 5051 } 5052 5053 static inline bool netif_is_ovs_port(const struct net_device *dev) 5054 { 5055 return dev->priv_flags & IFF_OVS_DATAPATH; 5056 } 5057 5058 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5059 { 5060 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5061 } 5062 5063 static inline bool netif_is_team_master(const struct net_device *dev) 5064 { 5065 return dev->priv_flags & IFF_TEAM; 5066 } 5067 5068 static inline bool netif_is_team_port(const struct net_device *dev) 5069 { 5070 return dev->priv_flags & IFF_TEAM_PORT; 5071 } 5072 5073 static inline bool netif_is_lag_master(const struct net_device *dev) 5074 { 5075 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5076 } 5077 5078 static inline bool netif_is_lag_port(const struct net_device *dev) 5079 { 5080 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5081 } 5082 5083 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5084 { 5085 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5086 } 5087 5088 static inline bool netif_is_failover(const struct net_device *dev) 5089 { 5090 return dev->priv_flags & IFF_FAILOVER; 5091 } 5092 5093 static inline bool netif_is_failover_slave(const struct net_device *dev) 5094 { 5095 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5096 } 5097 5098 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5099 static inline void netif_keep_dst(struct net_device *dev) 5100 { 5101 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5102 } 5103 5104 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5105 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5106 { 5107 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5108 return netif_is_macsec(dev); 5109 } 5110 5111 extern struct pernet_operations __net_initdata loopback_net_ops; 5112 5113 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5114 5115 /* netdev_printk helpers, similar to dev_printk */ 5116 5117 static inline const char *netdev_name(const struct net_device *dev) 5118 { 5119 if (!dev->name[0] || strchr(dev->name, '%')) 5120 return "(unnamed net_device)"; 5121 return dev->name; 5122 } 5123 5124 static inline const char *netdev_reg_state(const struct net_device *dev) 5125 { 5126 switch (dev->reg_state) { 5127 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5128 case NETREG_REGISTERED: return ""; 5129 case NETREG_UNREGISTERING: return " (unregistering)"; 5130 case NETREG_UNREGISTERED: return " (unregistered)"; 5131 case NETREG_RELEASED: return " (released)"; 5132 case NETREG_DUMMY: return " (dummy)"; 5133 } 5134 5135 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5136 return " (unknown)"; 5137 } 5138 5139 #define MODULE_ALIAS_NETDEV(device) \ 5140 MODULE_ALIAS("netdev-" device) 5141 5142 /* 5143 * netdev_WARN() acts like dev_printk(), but with the key difference 5144 * of using a WARN/WARN_ON to get the message out, including the 5145 * file/line information and a backtrace. 5146 */ 5147 #define netdev_WARN(dev, format, args...) \ 5148 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5149 netdev_reg_state(dev), ##args) 5150 5151 #define netdev_WARN_ONCE(dev, format, args...) \ 5152 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5153 netdev_reg_state(dev), ##args) 5154 5155 /* 5156 * The list of packet types we will receive (as opposed to discard) 5157 * and the routines to invoke. 5158 * 5159 * Why 16. Because with 16 the only overlap we get on a hash of the 5160 * low nibble of the protocol value is RARP/SNAP/X.25. 5161 * 5162 * 0800 IP 5163 * 0001 802.3 5164 * 0002 AX.25 5165 * 0004 802.2 5166 * 8035 RARP 5167 * 0005 SNAP 5168 * 0805 X.25 5169 * 0806 ARP 5170 * 8137 IPX 5171 * 0009 Localtalk 5172 * 86DD IPv6 5173 */ 5174 #define PTYPE_HASH_SIZE (16) 5175 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5176 5177 extern struct list_head ptype_all __read_mostly; 5178 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5179 5180 extern struct net_device *blackhole_netdev; 5181 5182 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */ 5183 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD) 5184 #define DEV_STATS_ADD(DEV, FIELD, VAL) \ 5185 atomic_long_add((VAL), &(DEV)->stats.__##FIELD) 5186 5187 #endif /* _LINUX_NETDEVICE_H */ 5188