1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 13 * Corey Minyard <wf-rch!minyard@relay.EU.net> 14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov> 15 * Alan Cox, <alan@lxorguk.ukuu.org.uk> 16 * Bjorn Ekwall. <bj0rn@blox.se> 17 * Pekka Riikonen <priikone@poseidon.pspt.fi> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 #include <asm/local.h> 32 33 #include <linux/percpu.h> 34 #include <linux/rculist.h> 35 #include <linux/workqueue.h> 36 #include <linux/dynamic_queue_limits.h> 37 38 #include <net/net_namespace.h> 39 #ifdef CONFIG_DCB 40 #include <net/dcbnl.h> 41 #endif 42 #include <net/netprio_cgroup.h> 43 #include <net/xdp.h> 44 45 #include <linux/netdev_features.h> 46 #include <linux/neighbour.h> 47 #include <uapi/linux/netdevice.h> 48 #include <uapi/linux/if_bonding.h> 49 #include <uapi/linux/pkt_cls.h> 50 #include <linux/hashtable.h> 51 #include <linux/rbtree.h> 52 #include <net/net_trackers.h> 53 #include <net/net_debug.h> 54 55 struct netpoll_info; 56 struct device; 57 struct ethtool_ops; 58 struct phy_device; 59 struct dsa_port; 60 struct ip_tunnel_parm; 61 struct macsec_context; 62 struct macsec_ops; 63 struct netdev_name_node; 64 struct sd_flow_limit; 65 struct sfp_bus; 66 /* 802.11 specific */ 67 struct wireless_dev; 68 /* 802.15.4 specific */ 69 struct wpan_dev; 70 struct mpls_dev; 71 /* UDP Tunnel offloads */ 72 struct udp_tunnel_info; 73 struct udp_tunnel_nic_info; 74 struct udp_tunnel_nic; 75 struct bpf_prog; 76 struct xdp_buff; 77 78 void synchronize_net(void); 79 void netdev_set_default_ethtool_ops(struct net_device *dev, 80 const struct ethtool_ops *ops); 81 82 /* Backlog congestion levels */ 83 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 84 #define NET_RX_DROP 1 /* packet dropped */ 85 86 #define MAX_NEST_DEV 8 87 88 /* 89 * Transmit return codes: transmit return codes originate from three different 90 * namespaces: 91 * 92 * - qdisc return codes 93 * - driver transmit return codes 94 * - errno values 95 * 96 * Drivers are allowed to return any one of those in their hard_start_xmit() 97 * function. Real network devices commonly used with qdiscs should only return 98 * the driver transmit return codes though - when qdiscs are used, the actual 99 * transmission happens asynchronously, so the value is not propagated to 100 * higher layers. Virtual network devices transmit synchronously; in this case 101 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 102 * others are propagated to higher layers. 103 */ 104 105 /* qdisc ->enqueue() return codes. */ 106 #define NET_XMIT_SUCCESS 0x00 107 #define NET_XMIT_DROP 0x01 /* skb dropped */ 108 #define NET_XMIT_CN 0x02 /* congestion notification */ 109 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 110 111 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 112 * indicates that the device will soon be dropping packets, or already drops 113 * some packets of the same priority; prompting us to send less aggressively. */ 114 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 115 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 116 117 /* Driver transmit return codes */ 118 #define NETDEV_TX_MASK 0xf0 119 120 enum netdev_tx { 121 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 122 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 123 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 124 }; 125 typedef enum netdev_tx netdev_tx_t; 126 127 /* 128 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 129 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 130 */ 131 static inline bool dev_xmit_complete(int rc) 132 { 133 /* 134 * Positive cases with an skb consumed by a driver: 135 * - successful transmission (rc == NETDEV_TX_OK) 136 * - error while transmitting (rc < 0) 137 * - error while queueing to a different device (rc & NET_XMIT_MASK) 138 */ 139 if (likely(rc < NET_XMIT_MASK)) 140 return true; 141 142 return false; 143 } 144 145 /* 146 * Compute the worst-case header length according to the protocols 147 * used. 148 */ 149 150 #if defined(CONFIG_HYPERV_NET) 151 # define LL_MAX_HEADER 128 152 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 153 # if defined(CONFIG_MAC80211_MESH) 154 # define LL_MAX_HEADER 128 155 # else 156 # define LL_MAX_HEADER 96 157 # endif 158 #else 159 # define LL_MAX_HEADER 32 160 #endif 161 162 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 163 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 164 #define MAX_HEADER LL_MAX_HEADER 165 #else 166 #define MAX_HEADER (LL_MAX_HEADER + 48) 167 #endif 168 169 /* 170 * Old network device statistics. Fields are native words 171 * (unsigned long) so they can be read and written atomically. 172 */ 173 174 #define NET_DEV_STAT(FIELD) \ 175 union { \ 176 unsigned long FIELD; \ 177 atomic_long_t __##FIELD; \ 178 } 179 180 struct net_device_stats { 181 NET_DEV_STAT(rx_packets); 182 NET_DEV_STAT(tx_packets); 183 NET_DEV_STAT(rx_bytes); 184 NET_DEV_STAT(tx_bytes); 185 NET_DEV_STAT(rx_errors); 186 NET_DEV_STAT(tx_errors); 187 NET_DEV_STAT(rx_dropped); 188 NET_DEV_STAT(tx_dropped); 189 NET_DEV_STAT(multicast); 190 NET_DEV_STAT(collisions); 191 NET_DEV_STAT(rx_length_errors); 192 NET_DEV_STAT(rx_over_errors); 193 NET_DEV_STAT(rx_crc_errors); 194 NET_DEV_STAT(rx_frame_errors); 195 NET_DEV_STAT(rx_fifo_errors); 196 NET_DEV_STAT(rx_missed_errors); 197 NET_DEV_STAT(tx_aborted_errors); 198 NET_DEV_STAT(tx_carrier_errors); 199 NET_DEV_STAT(tx_fifo_errors); 200 NET_DEV_STAT(tx_heartbeat_errors); 201 NET_DEV_STAT(tx_window_errors); 202 NET_DEV_STAT(rx_compressed); 203 NET_DEV_STAT(tx_compressed); 204 }; 205 #undef NET_DEV_STAT 206 207 /* per-cpu stats, allocated on demand. 208 * Try to fit them in a single cache line, for dev_get_stats() sake. 209 */ 210 struct net_device_core_stats { 211 unsigned long rx_dropped; 212 unsigned long tx_dropped; 213 unsigned long rx_nohandler; 214 unsigned long rx_otherhost_dropped; 215 } __aligned(4 * sizeof(unsigned long)); 216 217 #include <linux/cache.h> 218 #include <linux/skbuff.h> 219 220 #ifdef CONFIG_RPS 221 #include <linux/static_key.h> 222 extern struct static_key_false rps_needed; 223 extern struct static_key_false rfs_needed; 224 #endif 225 226 struct neighbour; 227 struct neigh_parms; 228 struct sk_buff; 229 230 struct netdev_hw_addr { 231 struct list_head list; 232 struct rb_node node; 233 unsigned char addr[MAX_ADDR_LEN]; 234 unsigned char type; 235 #define NETDEV_HW_ADDR_T_LAN 1 236 #define NETDEV_HW_ADDR_T_SAN 2 237 #define NETDEV_HW_ADDR_T_UNICAST 3 238 #define NETDEV_HW_ADDR_T_MULTICAST 4 239 bool global_use; 240 int sync_cnt; 241 int refcount; 242 int synced; 243 struct rcu_head rcu_head; 244 }; 245 246 struct netdev_hw_addr_list { 247 struct list_head list; 248 int count; 249 250 /* Auxiliary tree for faster lookup on addition and deletion */ 251 struct rb_root tree; 252 }; 253 254 #define netdev_hw_addr_list_count(l) ((l)->count) 255 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 256 #define netdev_hw_addr_list_for_each(ha, l) \ 257 list_for_each_entry(ha, &(l)->list, list) 258 259 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 260 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 261 #define netdev_for_each_uc_addr(ha, dev) \ 262 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 263 #define netdev_for_each_synced_uc_addr(_ha, _dev) \ 264 netdev_for_each_uc_addr((_ha), (_dev)) \ 265 if ((_ha)->sync_cnt) 266 267 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 268 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 269 #define netdev_for_each_mc_addr(ha, dev) \ 270 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 271 #define netdev_for_each_synced_mc_addr(_ha, _dev) \ 272 netdev_for_each_mc_addr((_ha), (_dev)) \ 273 if ((_ha)->sync_cnt) 274 275 struct hh_cache { 276 unsigned int hh_len; 277 seqlock_t hh_lock; 278 279 /* cached hardware header; allow for machine alignment needs. */ 280 #define HH_DATA_MOD 16 281 #define HH_DATA_OFF(__len) \ 282 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 283 #define HH_DATA_ALIGN(__len) \ 284 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 285 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 286 }; 287 288 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 289 * Alternative is: 290 * dev->hard_header_len ? (dev->hard_header_len + 291 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 292 * 293 * We could use other alignment values, but we must maintain the 294 * relationship HH alignment <= LL alignment. 295 */ 296 #define LL_RESERVED_SPACE(dev) \ 297 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 298 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 299 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 300 301 struct header_ops { 302 int (*create) (struct sk_buff *skb, struct net_device *dev, 303 unsigned short type, const void *daddr, 304 const void *saddr, unsigned int len); 305 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 306 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 307 void (*cache_update)(struct hh_cache *hh, 308 const struct net_device *dev, 309 const unsigned char *haddr); 310 bool (*validate)(const char *ll_header, unsigned int len); 311 __be16 (*parse_protocol)(const struct sk_buff *skb); 312 }; 313 314 /* These flag bits are private to the generic network queueing 315 * layer; they may not be explicitly referenced by any other 316 * code. 317 */ 318 319 enum netdev_state_t { 320 __LINK_STATE_START, 321 __LINK_STATE_PRESENT, 322 __LINK_STATE_NOCARRIER, 323 __LINK_STATE_LINKWATCH_PENDING, 324 __LINK_STATE_DORMANT, 325 __LINK_STATE_TESTING, 326 }; 327 328 struct gro_list { 329 struct list_head list; 330 int count; 331 }; 332 333 /* 334 * size of gro hash buckets, must less than bit number of 335 * napi_struct::gro_bitmask 336 */ 337 #define GRO_HASH_BUCKETS 8 338 339 /* 340 * Structure for NAPI scheduling similar to tasklet but with weighting 341 */ 342 struct napi_struct { 343 /* The poll_list must only be managed by the entity which 344 * changes the state of the NAPI_STATE_SCHED bit. This means 345 * whoever atomically sets that bit can add this napi_struct 346 * to the per-CPU poll_list, and whoever clears that bit 347 * can remove from the list right before clearing the bit. 348 */ 349 struct list_head poll_list; 350 351 unsigned long state; 352 int weight; 353 int defer_hard_irqs_count; 354 unsigned long gro_bitmask; 355 int (*poll)(struct napi_struct *, int); 356 #ifdef CONFIG_NETPOLL 357 int poll_owner; 358 #endif 359 struct net_device *dev; 360 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 361 struct sk_buff *skb; 362 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 363 int rx_count; /* length of rx_list */ 364 struct hrtimer timer; 365 struct list_head dev_list; 366 struct hlist_node napi_hash_node; 367 unsigned int napi_id; 368 struct task_struct *thread; 369 }; 370 371 enum { 372 NAPI_STATE_SCHED, /* Poll is scheduled */ 373 NAPI_STATE_MISSED, /* reschedule a napi */ 374 NAPI_STATE_DISABLE, /* Disable pending */ 375 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 376 NAPI_STATE_LISTED, /* NAPI added to system lists */ 377 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 378 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 379 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 380 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 381 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 382 }; 383 384 enum { 385 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 386 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 387 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 388 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 389 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 390 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 391 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 392 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 393 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 394 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 395 }; 396 397 enum gro_result { 398 GRO_MERGED, 399 GRO_MERGED_FREE, 400 GRO_HELD, 401 GRO_NORMAL, 402 GRO_CONSUMED, 403 }; 404 typedef enum gro_result gro_result_t; 405 406 /* 407 * enum rx_handler_result - Possible return values for rx_handlers. 408 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 409 * further. 410 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 411 * case skb->dev was changed by rx_handler. 412 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 413 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 414 * 415 * rx_handlers are functions called from inside __netif_receive_skb(), to do 416 * special processing of the skb, prior to delivery to protocol handlers. 417 * 418 * Currently, a net_device can only have a single rx_handler registered. Trying 419 * to register a second rx_handler will return -EBUSY. 420 * 421 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 422 * To unregister a rx_handler on a net_device, use 423 * netdev_rx_handler_unregister(). 424 * 425 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 426 * do with the skb. 427 * 428 * If the rx_handler consumed the skb in some way, it should return 429 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 430 * the skb to be delivered in some other way. 431 * 432 * If the rx_handler changed skb->dev, to divert the skb to another 433 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 434 * new device will be called if it exists. 435 * 436 * If the rx_handler decides the skb should be ignored, it should return 437 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 438 * are registered on exact device (ptype->dev == skb->dev). 439 * 440 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 441 * delivered, it should return RX_HANDLER_PASS. 442 * 443 * A device without a registered rx_handler will behave as if rx_handler 444 * returned RX_HANDLER_PASS. 445 */ 446 447 enum rx_handler_result { 448 RX_HANDLER_CONSUMED, 449 RX_HANDLER_ANOTHER, 450 RX_HANDLER_EXACT, 451 RX_HANDLER_PASS, 452 }; 453 typedef enum rx_handler_result rx_handler_result_t; 454 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 455 456 void __napi_schedule(struct napi_struct *n); 457 void __napi_schedule_irqoff(struct napi_struct *n); 458 459 static inline bool napi_disable_pending(struct napi_struct *n) 460 { 461 return test_bit(NAPI_STATE_DISABLE, &n->state); 462 } 463 464 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 465 { 466 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 467 } 468 469 bool napi_schedule_prep(struct napi_struct *n); 470 471 /** 472 * napi_schedule - schedule NAPI poll 473 * @n: NAPI context 474 * 475 * Schedule NAPI poll routine to be called if it is not already 476 * running. 477 */ 478 static inline void napi_schedule(struct napi_struct *n) 479 { 480 if (napi_schedule_prep(n)) 481 __napi_schedule(n); 482 } 483 484 /** 485 * napi_schedule_irqoff - schedule NAPI poll 486 * @n: NAPI context 487 * 488 * Variant of napi_schedule(), assuming hard irqs are masked. 489 */ 490 static inline void napi_schedule_irqoff(struct napi_struct *n) 491 { 492 if (napi_schedule_prep(n)) 493 __napi_schedule_irqoff(n); 494 } 495 496 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 497 static inline bool napi_reschedule(struct napi_struct *napi) 498 { 499 if (napi_schedule_prep(napi)) { 500 __napi_schedule(napi); 501 return true; 502 } 503 return false; 504 } 505 506 bool napi_complete_done(struct napi_struct *n, int work_done); 507 /** 508 * napi_complete - NAPI processing complete 509 * @n: NAPI context 510 * 511 * Mark NAPI processing as complete. 512 * Consider using napi_complete_done() instead. 513 * Return false if device should avoid rearming interrupts. 514 */ 515 static inline bool napi_complete(struct napi_struct *n) 516 { 517 return napi_complete_done(n, 0); 518 } 519 520 int dev_set_threaded(struct net_device *dev, bool threaded); 521 522 /** 523 * napi_disable - prevent NAPI from scheduling 524 * @n: NAPI context 525 * 526 * Stop NAPI from being scheduled on this context. 527 * Waits till any outstanding processing completes. 528 */ 529 void napi_disable(struct napi_struct *n); 530 531 void napi_enable(struct napi_struct *n); 532 533 /** 534 * napi_synchronize - wait until NAPI is not running 535 * @n: NAPI context 536 * 537 * Wait until NAPI is done being scheduled on this context. 538 * Waits till any outstanding processing completes but 539 * does not disable future activations. 540 */ 541 static inline void napi_synchronize(const struct napi_struct *n) 542 { 543 if (IS_ENABLED(CONFIG_SMP)) 544 while (test_bit(NAPI_STATE_SCHED, &n->state)) 545 msleep(1); 546 else 547 barrier(); 548 } 549 550 /** 551 * napi_if_scheduled_mark_missed - if napi is running, set the 552 * NAPIF_STATE_MISSED 553 * @n: NAPI context 554 * 555 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 556 * NAPI is scheduled. 557 **/ 558 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 559 { 560 unsigned long val, new; 561 562 val = READ_ONCE(n->state); 563 do { 564 if (val & NAPIF_STATE_DISABLE) 565 return true; 566 567 if (!(val & NAPIF_STATE_SCHED)) 568 return false; 569 570 new = val | NAPIF_STATE_MISSED; 571 } while (!try_cmpxchg(&n->state, &val, new)); 572 573 return true; 574 } 575 576 enum netdev_queue_state_t { 577 __QUEUE_STATE_DRV_XOFF, 578 __QUEUE_STATE_STACK_XOFF, 579 __QUEUE_STATE_FROZEN, 580 }; 581 582 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 583 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 584 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 585 586 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 587 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 588 QUEUE_STATE_FROZEN) 589 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 590 QUEUE_STATE_FROZEN) 591 592 /* 593 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 594 * netif_tx_* functions below are used to manipulate this flag. The 595 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 596 * queue independently. The netif_xmit_*stopped functions below are called 597 * to check if the queue has been stopped by the driver or stack (either 598 * of the XOFF bits are set in the state). Drivers should not need to call 599 * netif_xmit*stopped functions, they should only be using netif_tx_*. 600 */ 601 602 struct netdev_queue { 603 /* 604 * read-mostly part 605 */ 606 struct net_device *dev; 607 netdevice_tracker dev_tracker; 608 609 struct Qdisc __rcu *qdisc; 610 struct Qdisc *qdisc_sleeping; 611 #ifdef CONFIG_SYSFS 612 struct kobject kobj; 613 #endif 614 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 615 int numa_node; 616 #endif 617 unsigned long tx_maxrate; 618 /* 619 * Number of TX timeouts for this queue 620 * (/sys/class/net/DEV/Q/trans_timeout) 621 */ 622 atomic_long_t trans_timeout; 623 624 /* Subordinate device that the queue has been assigned to */ 625 struct net_device *sb_dev; 626 #ifdef CONFIG_XDP_SOCKETS 627 struct xsk_buff_pool *pool; 628 #endif 629 /* 630 * write-mostly part 631 */ 632 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 633 int xmit_lock_owner; 634 /* 635 * Time (in jiffies) of last Tx 636 */ 637 unsigned long trans_start; 638 639 unsigned long state; 640 641 #ifdef CONFIG_BQL 642 struct dql dql; 643 #endif 644 } ____cacheline_aligned_in_smp; 645 646 extern int sysctl_fb_tunnels_only_for_init_net; 647 extern int sysctl_devconf_inherit_init_net; 648 649 /* 650 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 651 * == 1 : For initns only 652 * == 2 : For none. 653 */ 654 static inline bool net_has_fallback_tunnels(const struct net *net) 655 { 656 #if IS_ENABLED(CONFIG_SYSCTL) 657 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net); 658 659 return !fb_tunnels_only_for_init_net || 660 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1); 661 #else 662 return true; 663 #endif 664 } 665 666 static inline int net_inherit_devconf(void) 667 { 668 #if IS_ENABLED(CONFIG_SYSCTL) 669 return READ_ONCE(sysctl_devconf_inherit_init_net); 670 #else 671 return 0; 672 #endif 673 } 674 675 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 676 { 677 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 678 return q->numa_node; 679 #else 680 return NUMA_NO_NODE; 681 #endif 682 } 683 684 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 685 { 686 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 687 q->numa_node = node; 688 #endif 689 } 690 691 #ifdef CONFIG_RPS 692 /* 693 * This structure holds an RPS map which can be of variable length. The 694 * map is an array of CPUs. 695 */ 696 struct rps_map { 697 unsigned int len; 698 struct rcu_head rcu; 699 u16 cpus[]; 700 }; 701 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 702 703 /* 704 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 705 * tail pointer for that CPU's input queue at the time of last enqueue, and 706 * a hardware filter index. 707 */ 708 struct rps_dev_flow { 709 u16 cpu; 710 u16 filter; 711 unsigned int last_qtail; 712 }; 713 #define RPS_NO_FILTER 0xffff 714 715 /* 716 * The rps_dev_flow_table structure contains a table of flow mappings. 717 */ 718 struct rps_dev_flow_table { 719 unsigned int mask; 720 struct rcu_head rcu; 721 struct rps_dev_flow flows[]; 722 }; 723 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 724 ((_num) * sizeof(struct rps_dev_flow))) 725 726 /* 727 * The rps_sock_flow_table contains mappings of flows to the last CPU 728 * on which they were processed by the application (set in recvmsg). 729 * Each entry is a 32bit value. Upper part is the high-order bits 730 * of flow hash, lower part is CPU number. 731 * rps_cpu_mask is used to partition the space, depending on number of 732 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 733 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 734 * meaning we use 32-6=26 bits for the hash. 735 */ 736 struct rps_sock_flow_table { 737 u32 mask; 738 739 u32 ents[] ____cacheline_aligned_in_smp; 740 }; 741 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 742 743 #define RPS_NO_CPU 0xffff 744 745 extern u32 rps_cpu_mask; 746 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 747 748 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 749 u32 hash) 750 { 751 if (table && hash) { 752 unsigned int index = hash & table->mask; 753 u32 val = hash & ~rps_cpu_mask; 754 755 /* We only give a hint, preemption can change CPU under us */ 756 val |= raw_smp_processor_id(); 757 758 if (table->ents[index] != val) 759 table->ents[index] = val; 760 } 761 } 762 763 #ifdef CONFIG_RFS_ACCEL 764 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 765 u16 filter_id); 766 #endif 767 #endif /* CONFIG_RPS */ 768 769 /* This structure contains an instance of an RX queue. */ 770 struct netdev_rx_queue { 771 struct xdp_rxq_info xdp_rxq; 772 #ifdef CONFIG_RPS 773 struct rps_map __rcu *rps_map; 774 struct rps_dev_flow_table __rcu *rps_flow_table; 775 #endif 776 struct kobject kobj; 777 struct net_device *dev; 778 netdevice_tracker dev_tracker; 779 780 #ifdef CONFIG_XDP_SOCKETS 781 struct xsk_buff_pool *pool; 782 #endif 783 } ____cacheline_aligned_in_smp; 784 785 /* 786 * RX queue sysfs structures and functions. 787 */ 788 struct rx_queue_attribute { 789 struct attribute attr; 790 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 791 ssize_t (*store)(struct netdev_rx_queue *queue, 792 const char *buf, size_t len); 793 }; 794 795 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 796 enum xps_map_type { 797 XPS_CPUS = 0, 798 XPS_RXQS, 799 XPS_MAPS_MAX, 800 }; 801 802 #ifdef CONFIG_XPS 803 /* 804 * This structure holds an XPS map which can be of variable length. The 805 * map is an array of queues. 806 */ 807 struct xps_map { 808 unsigned int len; 809 unsigned int alloc_len; 810 struct rcu_head rcu; 811 u16 queues[]; 812 }; 813 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 814 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 815 - sizeof(struct xps_map)) / sizeof(u16)) 816 817 /* 818 * This structure holds all XPS maps for device. Maps are indexed by CPU. 819 * 820 * We keep track of the number of cpus/rxqs used when the struct is allocated, 821 * in nr_ids. This will help not accessing out-of-bound memory. 822 * 823 * We keep track of the number of traffic classes used when the struct is 824 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 825 * not crossing its upper bound, as the original dev->num_tc can be updated in 826 * the meantime. 827 */ 828 struct xps_dev_maps { 829 struct rcu_head rcu; 830 unsigned int nr_ids; 831 s16 num_tc; 832 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 833 }; 834 835 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 836 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 837 838 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 839 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 840 841 #endif /* CONFIG_XPS */ 842 843 #define TC_MAX_QUEUE 16 844 #define TC_BITMASK 15 845 /* HW offloaded queuing disciplines txq count and offset maps */ 846 struct netdev_tc_txq { 847 u16 count; 848 u16 offset; 849 }; 850 851 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 852 /* 853 * This structure is to hold information about the device 854 * configured to run FCoE protocol stack. 855 */ 856 struct netdev_fcoe_hbainfo { 857 char manufacturer[64]; 858 char serial_number[64]; 859 char hardware_version[64]; 860 char driver_version[64]; 861 char optionrom_version[64]; 862 char firmware_version[64]; 863 char model[256]; 864 char model_description[256]; 865 }; 866 #endif 867 868 #define MAX_PHYS_ITEM_ID_LEN 32 869 870 /* This structure holds a unique identifier to identify some 871 * physical item (port for example) used by a netdevice. 872 */ 873 struct netdev_phys_item_id { 874 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 875 unsigned char id_len; 876 }; 877 878 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 879 struct netdev_phys_item_id *b) 880 { 881 return a->id_len == b->id_len && 882 memcmp(a->id, b->id, a->id_len) == 0; 883 } 884 885 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 886 struct sk_buff *skb, 887 struct net_device *sb_dev); 888 889 enum net_device_path_type { 890 DEV_PATH_ETHERNET = 0, 891 DEV_PATH_VLAN, 892 DEV_PATH_BRIDGE, 893 DEV_PATH_PPPOE, 894 DEV_PATH_DSA, 895 DEV_PATH_MTK_WDMA, 896 }; 897 898 struct net_device_path { 899 enum net_device_path_type type; 900 const struct net_device *dev; 901 union { 902 struct { 903 u16 id; 904 __be16 proto; 905 u8 h_dest[ETH_ALEN]; 906 } encap; 907 struct { 908 enum { 909 DEV_PATH_BR_VLAN_KEEP, 910 DEV_PATH_BR_VLAN_TAG, 911 DEV_PATH_BR_VLAN_UNTAG, 912 DEV_PATH_BR_VLAN_UNTAG_HW, 913 } vlan_mode; 914 u16 vlan_id; 915 __be16 vlan_proto; 916 } bridge; 917 struct { 918 int port; 919 u16 proto; 920 } dsa; 921 struct { 922 u8 wdma_idx; 923 u8 queue; 924 u16 wcid; 925 u8 bss; 926 } mtk_wdma; 927 }; 928 }; 929 930 #define NET_DEVICE_PATH_STACK_MAX 5 931 #define NET_DEVICE_PATH_VLAN_MAX 2 932 933 struct net_device_path_stack { 934 int num_paths; 935 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 936 }; 937 938 struct net_device_path_ctx { 939 const struct net_device *dev; 940 u8 daddr[ETH_ALEN]; 941 942 int num_vlans; 943 struct { 944 u16 id; 945 __be16 proto; 946 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 947 }; 948 949 enum tc_setup_type { 950 TC_QUERY_CAPS, 951 TC_SETUP_QDISC_MQPRIO, 952 TC_SETUP_CLSU32, 953 TC_SETUP_CLSFLOWER, 954 TC_SETUP_CLSMATCHALL, 955 TC_SETUP_CLSBPF, 956 TC_SETUP_BLOCK, 957 TC_SETUP_QDISC_CBS, 958 TC_SETUP_QDISC_RED, 959 TC_SETUP_QDISC_PRIO, 960 TC_SETUP_QDISC_MQ, 961 TC_SETUP_QDISC_ETF, 962 TC_SETUP_ROOT_QDISC, 963 TC_SETUP_QDISC_GRED, 964 TC_SETUP_QDISC_TAPRIO, 965 TC_SETUP_FT, 966 TC_SETUP_QDISC_ETS, 967 TC_SETUP_QDISC_TBF, 968 TC_SETUP_QDISC_FIFO, 969 TC_SETUP_QDISC_HTB, 970 TC_SETUP_ACT, 971 }; 972 973 /* These structures hold the attributes of bpf state that are being passed 974 * to the netdevice through the bpf op. 975 */ 976 enum bpf_netdev_command { 977 /* Set or clear a bpf program used in the earliest stages of packet 978 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 979 * is responsible for calling bpf_prog_put on any old progs that are 980 * stored. In case of error, the callee need not release the new prog 981 * reference, but on success it takes ownership and must bpf_prog_put 982 * when it is no longer used. 983 */ 984 XDP_SETUP_PROG, 985 XDP_SETUP_PROG_HW, 986 /* BPF program for offload callbacks, invoked at program load time. */ 987 BPF_OFFLOAD_MAP_ALLOC, 988 BPF_OFFLOAD_MAP_FREE, 989 XDP_SETUP_XSK_POOL, 990 }; 991 992 struct bpf_prog_offload_ops; 993 struct netlink_ext_ack; 994 struct xdp_umem; 995 struct xdp_dev_bulk_queue; 996 struct bpf_xdp_link; 997 998 enum bpf_xdp_mode { 999 XDP_MODE_SKB = 0, 1000 XDP_MODE_DRV = 1, 1001 XDP_MODE_HW = 2, 1002 __MAX_XDP_MODE 1003 }; 1004 1005 struct bpf_xdp_entity { 1006 struct bpf_prog *prog; 1007 struct bpf_xdp_link *link; 1008 }; 1009 1010 struct netdev_bpf { 1011 enum bpf_netdev_command command; 1012 union { 1013 /* XDP_SETUP_PROG */ 1014 struct { 1015 u32 flags; 1016 struct bpf_prog *prog; 1017 struct netlink_ext_ack *extack; 1018 }; 1019 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 1020 struct { 1021 struct bpf_offloaded_map *offmap; 1022 }; 1023 /* XDP_SETUP_XSK_POOL */ 1024 struct { 1025 struct xsk_buff_pool *pool; 1026 u16 queue_id; 1027 } xsk; 1028 }; 1029 }; 1030 1031 /* Flags for ndo_xsk_wakeup. */ 1032 #define XDP_WAKEUP_RX (1 << 0) 1033 #define XDP_WAKEUP_TX (1 << 1) 1034 1035 #ifdef CONFIG_XFRM_OFFLOAD 1036 struct xfrmdev_ops { 1037 int (*xdo_dev_state_add) (struct xfrm_state *x); 1038 void (*xdo_dev_state_delete) (struct xfrm_state *x); 1039 void (*xdo_dev_state_free) (struct xfrm_state *x); 1040 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 1041 struct xfrm_state *x); 1042 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1043 }; 1044 #endif 1045 1046 struct dev_ifalias { 1047 struct rcu_head rcuhead; 1048 char ifalias[]; 1049 }; 1050 1051 struct devlink; 1052 struct tlsdev_ops; 1053 1054 struct netdev_net_notifier { 1055 struct list_head list; 1056 struct notifier_block *nb; 1057 }; 1058 1059 /* 1060 * This structure defines the management hooks for network devices. 1061 * The following hooks can be defined; unless noted otherwise, they are 1062 * optional and can be filled with a null pointer. 1063 * 1064 * int (*ndo_init)(struct net_device *dev); 1065 * This function is called once when a network device is registered. 1066 * The network device can use this for any late stage initialization 1067 * or semantic validation. It can fail with an error code which will 1068 * be propagated back to register_netdev. 1069 * 1070 * void (*ndo_uninit)(struct net_device *dev); 1071 * This function is called when device is unregistered or when registration 1072 * fails. It is not called if init fails. 1073 * 1074 * int (*ndo_open)(struct net_device *dev); 1075 * This function is called when a network device transitions to the up 1076 * state. 1077 * 1078 * int (*ndo_stop)(struct net_device *dev); 1079 * This function is called when a network device transitions to the down 1080 * state. 1081 * 1082 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1083 * struct net_device *dev); 1084 * Called when a packet needs to be transmitted. 1085 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1086 * the queue before that can happen; it's for obsolete devices and weird 1087 * corner cases, but the stack really does a non-trivial amount 1088 * of useless work if you return NETDEV_TX_BUSY. 1089 * Required; cannot be NULL. 1090 * 1091 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1092 * struct net_device *dev 1093 * netdev_features_t features); 1094 * Called by core transmit path to determine if device is capable of 1095 * performing offload operations on a given packet. This is to give 1096 * the device an opportunity to implement any restrictions that cannot 1097 * be otherwise expressed by feature flags. The check is called with 1098 * the set of features that the stack has calculated and it returns 1099 * those the driver believes to be appropriate. 1100 * 1101 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1102 * struct net_device *sb_dev); 1103 * Called to decide which queue to use when device supports multiple 1104 * transmit queues. 1105 * 1106 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1107 * This function is called to allow device receiver to make 1108 * changes to configuration when multicast or promiscuous is enabled. 1109 * 1110 * void (*ndo_set_rx_mode)(struct net_device *dev); 1111 * This function is called device changes address list filtering. 1112 * If driver handles unicast address filtering, it should set 1113 * IFF_UNICAST_FLT in its priv_flags. 1114 * 1115 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1116 * This function is called when the Media Access Control address 1117 * needs to be changed. If this interface is not defined, the 1118 * MAC address can not be changed. 1119 * 1120 * int (*ndo_validate_addr)(struct net_device *dev); 1121 * Test if Media Access Control address is valid for the device. 1122 * 1123 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1124 * Old-style ioctl entry point. This is used internally by the 1125 * appletalk and ieee802154 subsystems but is no longer called by 1126 * the device ioctl handler. 1127 * 1128 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1129 * Used by the bonding driver for its device specific ioctls: 1130 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1131 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1132 * 1133 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1134 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1135 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1136 * 1137 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1138 * Used to set network devices bus interface parameters. This interface 1139 * is retained for legacy reasons; new devices should use the bus 1140 * interface (PCI) for low level management. 1141 * 1142 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1143 * Called when a user wants to change the Maximum Transfer Unit 1144 * of a device. 1145 * 1146 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1147 * Callback used when the transmitter has not made any progress 1148 * for dev->watchdog ticks. 1149 * 1150 * void (*ndo_get_stats64)(struct net_device *dev, 1151 * struct rtnl_link_stats64 *storage); 1152 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1153 * Called when a user wants to get the network device usage 1154 * statistics. Drivers must do one of the following: 1155 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1156 * rtnl_link_stats64 structure passed by the caller. 1157 * 2. Define @ndo_get_stats to update a net_device_stats structure 1158 * (which should normally be dev->stats) and return a pointer to 1159 * it. The structure may be changed asynchronously only if each 1160 * field is written atomically. 1161 * 3. Update dev->stats asynchronously and atomically, and define 1162 * neither operation. 1163 * 1164 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1165 * Return true if this device supports offload stats of this attr_id. 1166 * 1167 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1168 * void *attr_data) 1169 * Get statistics for offload operations by attr_id. Write it into the 1170 * attr_data pointer. 1171 * 1172 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1173 * If device supports VLAN filtering this function is called when a 1174 * VLAN id is registered. 1175 * 1176 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1177 * If device supports VLAN filtering this function is called when a 1178 * VLAN id is unregistered. 1179 * 1180 * void (*ndo_poll_controller)(struct net_device *dev); 1181 * 1182 * SR-IOV management functions. 1183 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1184 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1185 * u8 qos, __be16 proto); 1186 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1187 * int max_tx_rate); 1188 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1189 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1190 * int (*ndo_get_vf_config)(struct net_device *dev, 1191 * int vf, struct ifla_vf_info *ivf); 1192 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1193 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1194 * struct nlattr *port[]); 1195 * 1196 * Enable or disable the VF ability to query its RSS Redirection Table and 1197 * Hash Key. This is needed since on some devices VF share this information 1198 * with PF and querying it may introduce a theoretical security risk. 1199 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1200 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1201 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1202 * void *type_data); 1203 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1204 * This is always called from the stack with the rtnl lock held and netif 1205 * tx queues stopped. This allows the netdevice to perform queue 1206 * management safely. 1207 * 1208 * Fiber Channel over Ethernet (FCoE) offload functions. 1209 * int (*ndo_fcoe_enable)(struct net_device *dev); 1210 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1211 * so the underlying device can perform whatever needed configuration or 1212 * initialization to support acceleration of FCoE traffic. 1213 * 1214 * int (*ndo_fcoe_disable)(struct net_device *dev); 1215 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1216 * so the underlying device can perform whatever needed clean-ups to 1217 * stop supporting acceleration of FCoE traffic. 1218 * 1219 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1220 * struct scatterlist *sgl, unsigned int sgc); 1221 * Called when the FCoE Initiator wants to initialize an I/O that 1222 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1223 * perform necessary setup and returns 1 to indicate the device is set up 1224 * successfully to perform DDP on this I/O, otherwise this returns 0. 1225 * 1226 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1227 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1228 * indicated by the FC exchange id 'xid', so the underlying device can 1229 * clean up and reuse resources for later DDP requests. 1230 * 1231 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1232 * struct scatterlist *sgl, unsigned int sgc); 1233 * Called when the FCoE Target wants to initialize an I/O that 1234 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1235 * perform necessary setup and returns 1 to indicate the device is set up 1236 * successfully to perform DDP on this I/O, otherwise this returns 0. 1237 * 1238 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1239 * struct netdev_fcoe_hbainfo *hbainfo); 1240 * Called when the FCoE Protocol stack wants information on the underlying 1241 * device. This information is utilized by the FCoE protocol stack to 1242 * register attributes with Fiber Channel management service as per the 1243 * FC-GS Fabric Device Management Information(FDMI) specification. 1244 * 1245 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1246 * Called when the underlying device wants to override default World Wide 1247 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1248 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1249 * protocol stack to use. 1250 * 1251 * RFS acceleration. 1252 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1253 * u16 rxq_index, u32 flow_id); 1254 * Set hardware filter for RFS. rxq_index is the target queue index; 1255 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1256 * Return the filter ID on success, or a negative error code. 1257 * 1258 * Slave management functions (for bridge, bonding, etc). 1259 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1260 * Called to make another netdev an underling. 1261 * 1262 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1263 * Called to release previously enslaved netdev. 1264 * 1265 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1266 * struct sk_buff *skb, 1267 * bool all_slaves); 1268 * Get the xmit slave of master device. If all_slaves is true, function 1269 * assume all the slaves can transmit. 1270 * 1271 * Feature/offload setting functions. 1272 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1273 * netdev_features_t features); 1274 * Adjusts the requested feature flags according to device-specific 1275 * constraints, and returns the resulting flags. Must not modify 1276 * the device state. 1277 * 1278 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1279 * Called to update device configuration to new features. Passed 1280 * feature set might be less than what was returned by ndo_fix_features()). 1281 * Must return >0 or -errno if it changed dev->features itself. 1282 * 1283 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1284 * struct net_device *dev, 1285 * const unsigned char *addr, u16 vid, u16 flags, 1286 * struct netlink_ext_ack *extack); 1287 * Adds an FDB entry to dev for addr. 1288 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1289 * struct net_device *dev, 1290 * const unsigned char *addr, u16 vid) 1291 * Deletes the FDB entry from dev coresponding to addr. 1292 * int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, struct nlattr *tb[], 1293 * struct net_device *dev, 1294 * u16 vid, 1295 * struct netlink_ext_ack *extack); 1296 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1297 * struct net_device *dev, struct net_device *filter_dev, 1298 * int *idx) 1299 * Used to add FDB entries to dump requests. Implementers should add 1300 * entries to skb and update idx with the number of entries. 1301 * 1302 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1303 * u16 flags, struct netlink_ext_ack *extack) 1304 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1305 * struct net_device *dev, u32 filter_mask, 1306 * int nlflags) 1307 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1308 * u16 flags); 1309 * 1310 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1311 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1312 * which do not represent real hardware may define this to allow their 1313 * userspace components to manage their virtual carrier state. Devices 1314 * that determine carrier state from physical hardware properties (eg 1315 * network cables) or protocol-dependent mechanisms (eg 1316 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1317 * 1318 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1319 * struct netdev_phys_item_id *ppid); 1320 * Called to get ID of physical port of this device. If driver does 1321 * not implement this, it is assumed that the hw is not able to have 1322 * multiple net devices on single physical port. 1323 * 1324 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1325 * struct netdev_phys_item_id *ppid) 1326 * Called to get the parent ID of the physical port of this device. 1327 * 1328 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1329 * struct net_device *dev) 1330 * Called by upper layer devices to accelerate switching or other 1331 * station functionality into hardware. 'pdev is the lowerdev 1332 * to use for the offload and 'dev' is the net device that will 1333 * back the offload. Returns a pointer to the private structure 1334 * the upper layer will maintain. 1335 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1336 * Called by upper layer device to delete the station created 1337 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1338 * the station and priv is the structure returned by the add 1339 * operation. 1340 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1341 * int queue_index, u32 maxrate); 1342 * Called when a user wants to set a max-rate limitation of specific 1343 * TX queue. 1344 * int (*ndo_get_iflink)(const struct net_device *dev); 1345 * Called to get the iflink value of this device. 1346 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1347 * This function is used to get egress tunnel information for given skb. 1348 * This is useful for retrieving outer tunnel header parameters while 1349 * sampling packet. 1350 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1351 * This function is used to specify the headroom that the skb must 1352 * consider when allocation skb during packet reception. Setting 1353 * appropriate rx headroom value allows avoiding skb head copy on 1354 * forward. Setting a negative value resets the rx headroom to the 1355 * default value. 1356 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1357 * This function is used to set or query state related to XDP on the 1358 * netdevice and manage BPF offload. See definition of 1359 * enum bpf_netdev_command for details. 1360 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1361 * u32 flags); 1362 * This function is used to submit @n XDP packets for transmit on a 1363 * netdevice. Returns number of frames successfully transmitted, frames 1364 * that got dropped are freed/returned via xdp_return_frame(). 1365 * Returns negative number, means general error invoking ndo, meaning 1366 * no frames were xmit'ed and core-caller will free all frames. 1367 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1368 * struct xdp_buff *xdp); 1369 * Get the xmit slave of master device based on the xdp_buff. 1370 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1371 * This function is used to wake up the softirq, ksoftirqd or kthread 1372 * responsible for sending and/or receiving packets on a specific 1373 * queue id bound to an AF_XDP socket. The flags field specifies if 1374 * only RX, only Tx, or both should be woken up using the flags 1375 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1376 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1377 * int cmd); 1378 * Add, change, delete or get information on an IPv4 tunnel. 1379 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1380 * If a device is paired with a peer device, return the peer instance. 1381 * The caller must be under RCU read context. 1382 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1383 * Get the forwarding path to reach the real device from the HW destination address 1384 * ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1385 * const struct skb_shared_hwtstamps *hwtstamps, 1386 * bool cycles); 1387 * Get hardware timestamp based on normal/adjustable time or free running 1388 * cycle counter. This function is required if physical clock supports a 1389 * free running cycle counter. 1390 */ 1391 struct net_device_ops { 1392 int (*ndo_init)(struct net_device *dev); 1393 void (*ndo_uninit)(struct net_device *dev); 1394 int (*ndo_open)(struct net_device *dev); 1395 int (*ndo_stop)(struct net_device *dev); 1396 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1397 struct net_device *dev); 1398 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1399 struct net_device *dev, 1400 netdev_features_t features); 1401 u16 (*ndo_select_queue)(struct net_device *dev, 1402 struct sk_buff *skb, 1403 struct net_device *sb_dev); 1404 void (*ndo_change_rx_flags)(struct net_device *dev, 1405 int flags); 1406 void (*ndo_set_rx_mode)(struct net_device *dev); 1407 int (*ndo_set_mac_address)(struct net_device *dev, 1408 void *addr); 1409 int (*ndo_validate_addr)(struct net_device *dev); 1410 int (*ndo_do_ioctl)(struct net_device *dev, 1411 struct ifreq *ifr, int cmd); 1412 int (*ndo_eth_ioctl)(struct net_device *dev, 1413 struct ifreq *ifr, int cmd); 1414 int (*ndo_siocbond)(struct net_device *dev, 1415 struct ifreq *ifr, int cmd); 1416 int (*ndo_siocwandev)(struct net_device *dev, 1417 struct if_settings *ifs); 1418 int (*ndo_siocdevprivate)(struct net_device *dev, 1419 struct ifreq *ifr, 1420 void __user *data, int cmd); 1421 int (*ndo_set_config)(struct net_device *dev, 1422 struct ifmap *map); 1423 int (*ndo_change_mtu)(struct net_device *dev, 1424 int new_mtu); 1425 int (*ndo_neigh_setup)(struct net_device *dev, 1426 struct neigh_parms *); 1427 void (*ndo_tx_timeout) (struct net_device *dev, 1428 unsigned int txqueue); 1429 1430 void (*ndo_get_stats64)(struct net_device *dev, 1431 struct rtnl_link_stats64 *storage); 1432 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1433 int (*ndo_get_offload_stats)(int attr_id, 1434 const struct net_device *dev, 1435 void *attr_data); 1436 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1437 1438 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1439 __be16 proto, u16 vid); 1440 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1441 __be16 proto, u16 vid); 1442 #ifdef CONFIG_NET_POLL_CONTROLLER 1443 void (*ndo_poll_controller)(struct net_device *dev); 1444 int (*ndo_netpoll_setup)(struct net_device *dev, 1445 struct netpoll_info *info); 1446 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1447 #endif 1448 int (*ndo_set_vf_mac)(struct net_device *dev, 1449 int queue, u8 *mac); 1450 int (*ndo_set_vf_vlan)(struct net_device *dev, 1451 int queue, u16 vlan, 1452 u8 qos, __be16 proto); 1453 int (*ndo_set_vf_rate)(struct net_device *dev, 1454 int vf, int min_tx_rate, 1455 int max_tx_rate); 1456 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1457 int vf, bool setting); 1458 int (*ndo_set_vf_trust)(struct net_device *dev, 1459 int vf, bool setting); 1460 int (*ndo_get_vf_config)(struct net_device *dev, 1461 int vf, 1462 struct ifla_vf_info *ivf); 1463 int (*ndo_set_vf_link_state)(struct net_device *dev, 1464 int vf, int link_state); 1465 int (*ndo_get_vf_stats)(struct net_device *dev, 1466 int vf, 1467 struct ifla_vf_stats 1468 *vf_stats); 1469 int (*ndo_set_vf_port)(struct net_device *dev, 1470 int vf, 1471 struct nlattr *port[]); 1472 int (*ndo_get_vf_port)(struct net_device *dev, 1473 int vf, struct sk_buff *skb); 1474 int (*ndo_get_vf_guid)(struct net_device *dev, 1475 int vf, 1476 struct ifla_vf_guid *node_guid, 1477 struct ifla_vf_guid *port_guid); 1478 int (*ndo_set_vf_guid)(struct net_device *dev, 1479 int vf, u64 guid, 1480 int guid_type); 1481 int (*ndo_set_vf_rss_query_en)( 1482 struct net_device *dev, 1483 int vf, bool setting); 1484 int (*ndo_setup_tc)(struct net_device *dev, 1485 enum tc_setup_type type, 1486 void *type_data); 1487 #if IS_ENABLED(CONFIG_FCOE) 1488 int (*ndo_fcoe_enable)(struct net_device *dev); 1489 int (*ndo_fcoe_disable)(struct net_device *dev); 1490 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1491 u16 xid, 1492 struct scatterlist *sgl, 1493 unsigned int sgc); 1494 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1495 u16 xid); 1496 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1497 u16 xid, 1498 struct scatterlist *sgl, 1499 unsigned int sgc); 1500 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1501 struct netdev_fcoe_hbainfo *hbainfo); 1502 #endif 1503 1504 #if IS_ENABLED(CONFIG_LIBFCOE) 1505 #define NETDEV_FCOE_WWNN 0 1506 #define NETDEV_FCOE_WWPN 1 1507 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1508 u64 *wwn, int type); 1509 #endif 1510 1511 #ifdef CONFIG_RFS_ACCEL 1512 int (*ndo_rx_flow_steer)(struct net_device *dev, 1513 const struct sk_buff *skb, 1514 u16 rxq_index, 1515 u32 flow_id); 1516 #endif 1517 int (*ndo_add_slave)(struct net_device *dev, 1518 struct net_device *slave_dev, 1519 struct netlink_ext_ack *extack); 1520 int (*ndo_del_slave)(struct net_device *dev, 1521 struct net_device *slave_dev); 1522 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1523 struct sk_buff *skb, 1524 bool all_slaves); 1525 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1526 struct sock *sk); 1527 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1528 netdev_features_t features); 1529 int (*ndo_set_features)(struct net_device *dev, 1530 netdev_features_t features); 1531 int (*ndo_neigh_construct)(struct net_device *dev, 1532 struct neighbour *n); 1533 void (*ndo_neigh_destroy)(struct net_device *dev, 1534 struct neighbour *n); 1535 1536 int (*ndo_fdb_add)(struct ndmsg *ndm, 1537 struct nlattr *tb[], 1538 struct net_device *dev, 1539 const unsigned char *addr, 1540 u16 vid, 1541 u16 flags, 1542 struct netlink_ext_ack *extack); 1543 int (*ndo_fdb_del)(struct ndmsg *ndm, 1544 struct nlattr *tb[], 1545 struct net_device *dev, 1546 const unsigned char *addr, 1547 u16 vid, struct netlink_ext_ack *extack); 1548 int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, 1549 struct nlattr *tb[], 1550 struct net_device *dev, 1551 u16 vid, 1552 struct netlink_ext_ack *extack); 1553 int (*ndo_fdb_dump)(struct sk_buff *skb, 1554 struct netlink_callback *cb, 1555 struct net_device *dev, 1556 struct net_device *filter_dev, 1557 int *idx); 1558 int (*ndo_fdb_get)(struct sk_buff *skb, 1559 struct nlattr *tb[], 1560 struct net_device *dev, 1561 const unsigned char *addr, 1562 u16 vid, u32 portid, u32 seq, 1563 struct netlink_ext_ack *extack); 1564 int (*ndo_bridge_setlink)(struct net_device *dev, 1565 struct nlmsghdr *nlh, 1566 u16 flags, 1567 struct netlink_ext_ack *extack); 1568 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1569 u32 pid, u32 seq, 1570 struct net_device *dev, 1571 u32 filter_mask, 1572 int nlflags); 1573 int (*ndo_bridge_dellink)(struct net_device *dev, 1574 struct nlmsghdr *nlh, 1575 u16 flags); 1576 int (*ndo_change_carrier)(struct net_device *dev, 1577 bool new_carrier); 1578 int (*ndo_get_phys_port_id)(struct net_device *dev, 1579 struct netdev_phys_item_id *ppid); 1580 int (*ndo_get_port_parent_id)(struct net_device *dev, 1581 struct netdev_phys_item_id *ppid); 1582 int (*ndo_get_phys_port_name)(struct net_device *dev, 1583 char *name, size_t len); 1584 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1585 struct net_device *dev); 1586 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1587 void *priv); 1588 1589 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1590 int queue_index, 1591 u32 maxrate); 1592 int (*ndo_get_iflink)(const struct net_device *dev); 1593 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1594 struct sk_buff *skb); 1595 void (*ndo_set_rx_headroom)(struct net_device *dev, 1596 int needed_headroom); 1597 int (*ndo_bpf)(struct net_device *dev, 1598 struct netdev_bpf *bpf); 1599 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1600 struct xdp_frame **xdp, 1601 u32 flags); 1602 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1603 struct xdp_buff *xdp); 1604 int (*ndo_xsk_wakeup)(struct net_device *dev, 1605 u32 queue_id, u32 flags); 1606 int (*ndo_tunnel_ctl)(struct net_device *dev, 1607 struct ip_tunnel_parm *p, int cmd); 1608 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1609 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1610 struct net_device_path *path); 1611 ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1612 const struct skb_shared_hwtstamps *hwtstamps, 1613 bool cycles); 1614 }; 1615 1616 /** 1617 * enum netdev_priv_flags - &struct net_device priv_flags 1618 * 1619 * These are the &struct net_device, they are only set internally 1620 * by drivers and used in the kernel. These flags are invisible to 1621 * userspace; this means that the order of these flags can change 1622 * during any kernel release. 1623 * 1624 * You should have a pretty good reason to be extending these flags. 1625 * 1626 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1627 * @IFF_EBRIDGE: Ethernet bridging device 1628 * @IFF_BONDING: bonding master or slave 1629 * @IFF_ISATAP: ISATAP interface (RFC4214) 1630 * @IFF_WAN_HDLC: WAN HDLC device 1631 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1632 * release skb->dst 1633 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1634 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1635 * @IFF_MACVLAN_PORT: device used as macvlan port 1636 * @IFF_BRIDGE_PORT: device used as bridge port 1637 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1638 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1639 * @IFF_UNICAST_FLT: Supports unicast filtering 1640 * @IFF_TEAM_PORT: device used as team port 1641 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1642 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1643 * change when it's running 1644 * @IFF_MACVLAN: Macvlan device 1645 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1646 * underlying stacked devices 1647 * @IFF_L3MDEV_MASTER: device is an L3 master device 1648 * @IFF_NO_QUEUE: device can run without qdisc attached 1649 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1650 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1651 * @IFF_TEAM: device is a team device 1652 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1653 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1654 * entity (i.e. the master device for bridged veth) 1655 * @IFF_MACSEC: device is a MACsec device 1656 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1657 * @IFF_FAILOVER: device is a failover master device 1658 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1659 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1660 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1661 * skb_headlen(skb) == 0 (data starts from frag0) 1662 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN 1663 */ 1664 enum netdev_priv_flags { 1665 IFF_802_1Q_VLAN = 1<<0, 1666 IFF_EBRIDGE = 1<<1, 1667 IFF_BONDING = 1<<2, 1668 IFF_ISATAP = 1<<3, 1669 IFF_WAN_HDLC = 1<<4, 1670 IFF_XMIT_DST_RELEASE = 1<<5, 1671 IFF_DONT_BRIDGE = 1<<6, 1672 IFF_DISABLE_NETPOLL = 1<<7, 1673 IFF_MACVLAN_PORT = 1<<8, 1674 IFF_BRIDGE_PORT = 1<<9, 1675 IFF_OVS_DATAPATH = 1<<10, 1676 IFF_TX_SKB_SHARING = 1<<11, 1677 IFF_UNICAST_FLT = 1<<12, 1678 IFF_TEAM_PORT = 1<<13, 1679 IFF_SUPP_NOFCS = 1<<14, 1680 IFF_LIVE_ADDR_CHANGE = 1<<15, 1681 IFF_MACVLAN = 1<<16, 1682 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1683 IFF_L3MDEV_MASTER = 1<<18, 1684 IFF_NO_QUEUE = 1<<19, 1685 IFF_OPENVSWITCH = 1<<20, 1686 IFF_L3MDEV_SLAVE = 1<<21, 1687 IFF_TEAM = 1<<22, 1688 IFF_RXFH_CONFIGURED = 1<<23, 1689 IFF_PHONY_HEADROOM = 1<<24, 1690 IFF_MACSEC = 1<<25, 1691 IFF_NO_RX_HANDLER = 1<<26, 1692 IFF_FAILOVER = 1<<27, 1693 IFF_FAILOVER_SLAVE = 1<<28, 1694 IFF_L3MDEV_RX_HANDLER = 1<<29, 1695 /* was IFF_LIVE_RENAME_OK */ 1696 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31), 1697 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32), 1698 }; 1699 1700 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1701 #define IFF_EBRIDGE IFF_EBRIDGE 1702 #define IFF_BONDING IFF_BONDING 1703 #define IFF_ISATAP IFF_ISATAP 1704 #define IFF_WAN_HDLC IFF_WAN_HDLC 1705 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1706 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1707 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1708 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1709 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1710 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1711 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1712 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1713 #define IFF_TEAM_PORT IFF_TEAM_PORT 1714 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1715 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1716 #define IFF_MACVLAN IFF_MACVLAN 1717 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1718 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1719 #define IFF_NO_QUEUE IFF_NO_QUEUE 1720 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1721 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1722 #define IFF_TEAM IFF_TEAM 1723 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1724 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM 1725 #define IFF_MACSEC IFF_MACSEC 1726 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1727 #define IFF_FAILOVER IFF_FAILOVER 1728 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1729 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1730 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR 1731 1732 /* Specifies the type of the struct net_device::ml_priv pointer */ 1733 enum netdev_ml_priv_type { 1734 ML_PRIV_NONE, 1735 ML_PRIV_CAN, 1736 }; 1737 1738 /** 1739 * struct net_device - The DEVICE structure. 1740 * 1741 * Actually, this whole structure is a big mistake. It mixes I/O 1742 * data with strictly "high-level" data, and it has to know about 1743 * almost every data structure used in the INET module. 1744 * 1745 * @name: This is the first field of the "visible" part of this structure 1746 * (i.e. as seen by users in the "Space.c" file). It is the name 1747 * of the interface. 1748 * 1749 * @name_node: Name hashlist node 1750 * @ifalias: SNMP alias 1751 * @mem_end: Shared memory end 1752 * @mem_start: Shared memory start 1753 * @base_addr: Device I/O address 1754 * @irq: Device IRQ number 1755 * 1756 * @state: Generic network queuing layer state, see netdev_state_t 1757 * @dev_list: The global list of network devices 1758 * @napi_list: List entry used for polling NAPI devices 1759 * @unreg_list: List entry when we are unregistering the 1760 * device; see the function unregister_netdev 1761 * @close_list: List entry used when we are closing the device 1762 * @ptype_all: Device-specific packet handlers for all protocols 1763 * @ptype_specific: Device-specific, protocol-specific packet handlers 1764 * 1765 * @adj_list: Directly linked devices, like slaves for bonding 1766 * @features: Currently active device features 1767 * @hw_features: User-changeable features 1768 * 1769 * @wanted_features: User-requested features 1770 * @vlan_features: Mask of features inheritable by VLAN devices 1771 * 1772 * @hw_enc_features: Mask of features inherited by encapsulating devices 1773 * This field indicates what encapsulation 1774 * offloads the hardware is capable of doing, 1775 * and drivers will need to set them appropriately. 1776 * 1777 * @mpls_features: Mask of features inheritable by MPLS 1778 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1779 * 1780 * @ifindex: interface index 1781 * @group: The group the device belongs to 1782 * 1783 * @stats: Statistics struct, which was left as a legacy, use 1784 * rtnl_link_stats64 instead 1785 * 1786 * @core_stats: core networking counters, 1787 * do not use this in drivers 1788 * @carrier_up_count: Number of times the carrier has been up 1789 * @carrier_down_count: Number of times the carrier has been down 1790 * 1791 * @wireless_handlers: List of functions to handle Wireless Extensions, 1792 * instead of ioctl, 1793 * see <net/iw_handler.h> for details. 1794 * @wireless_data: Instance data managed by the core of wireless extensions 1795 * 1796 * @netdev_ops: Includes several pointers to callbacks, 1797 * if one wants to override the ndo_*() functions 1798 * @ethtool_ops: Management operations 1799 * @l3mdev_ops: Layer 3 master device operations 1800 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1801 * discovery handling. Necessary for e.g. 6LoWPAN. 1802 * @xfrmdev_ops: Transformation offload operations 1803 * @tlsdev_ops: Transport Layer Security offload operations 1804 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1805 * of Layer 2 headers. 1806 * 1807 * @flags: Interface flags (a la BSD) 1808 * @priv_flags: Like 'flags' but invisible to userspace, 1809 * see if.h for the definitions 1810 * @gflags: Global flags ( kept as legacy ) 1811 * @padded: How much padding added by alloc_netdev() 1812 * @operstate: RFC2863 operstate 1813 * @link_mode: Mapping policy to operstate 1814 * @if_port: Selectable AUI, TP, ... 1815 * @dma: DMA channel 1816 * @mtu: Interface MTU value 1817 * @min_mtu: Interface Minimum MTU value 1818 * @max_mtu: Interface Maximum MTU value 1819 * @type: Interface hardware type 1820 * @hard_header_len: Maximum hardware header length. 1821 * @min_header_len: Minimum hardware header length 1822 * 1823 * @needed_headroom: Extra headroom the hardware may need, but not in all 1824 * cases can this be guaranteed 1825 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1826 * cases can this be guaranteed. Some cases also use 1827 * LL_MAX_HEADER instead to allocate the skb 1828 * 1829 * interface address info: 1830 * 1831 * @perm_addr: Permanent hw address 1832 * @addr_assign_type: Hw address assignment type 1833 * @addr_len: Hardware address length 1834 * @upper_level: Maximum depth level of upper devices. 1835 * @lower_level: Maximum depth level of lower devices. 1836 * @neigh_priv_len: Used in neigh_alloc() 1837 * @dev_id: Used to differentiate devices that share 1838 * the same link layer address 1839 * @dev_port: Used to differentiate devices that share 1840 * the same function 1841 * @addr_list_lock: XXX: need comments on this one 1842 * @name_assign_type: network interface name assignment type 1843 * @uc_promisc: Counter that indicates promiscuous mode 1844 * has been enabled due to the need to listen to 1845 * additional unicast addresses in a device that 1846 * does not implement ndo_set_rx_mode() 1847 * @uc: unicast mac addresses 1848 * @mc: multicast mac addresses 1849 * @dev_addrs: list of device hw addresses 1850 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1851 * @promiscuity: Number of times the NIC is told to work in 1852 * promiscuous mode; if it becomes 0 the NIC will 1853 * exit promiscuous mode 1854 * @allmulti: Counter, enables or disables allmulticast mode 1855 * 1856 * @vlan_info: VLAN info 1857 * @dsa_ptr: dsa specific data 1858 * @tipc_ptr: TIPC specific data 1859 * @atalk_ptr: AppleTalk link 1860 * @ip_ptr: IPv4 specific data 1861 * @ip6_ptr: IPv6 specific data 1862 * @ax25_ptr: AX.25 specific data 1863 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1864 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1865 * device struct 1866 * @mpls_ptr: mpls_dev struct pointer 1867 * @mctp_ptr: MCTP specific data 1868 * 1869 * @dev_addr: Hw address (before bcast, 1870 * because most packets are unicast) 1871 * 1872 * @_rx: Array of RX queues 1873 * @num_rx_queues: Number of RX queues 1874 * allocated at register_netdev() time 1875 * @real_num_rx_queues: Number of RX queues currently active in device 1876 * @xdp_prog: XDP sockets filter program pointer 1877 * @gro_flush_timeout: timeout for GRO layer in NAPI 1878 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1879 * allow to avoid NIC hard IRQ, on busy queues. 1880 * 1881 * @rx_handler: handler for received packets 1882 * @rx_handler_data: XXX: need comments on this one 1883 * @miniq_ingress: ingress/clsact qdisc specific data for 1884 * ingress processing 1885 * @ingress_queue: XXX: need comments on this one 1886 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1887 * @broadcast: hw bcast address 1888 * 1889 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1890 * indexed by RX queue number. Assigned by driver. 1891 * This must only be set if the ndo_rx_flow_steer 1892 * operation is defined 1893 * @index_hlist: Device index hash chain 1894 * 1895 * @_tx: Array of TX queues 1896 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1897 * @real_num_tx_queues: Number of TX queues currently active in device 1898 * @qdisc: Root qdisc from userspace point of view 1899 * @tx_queue_len: Max frames per queue allowed 1900 * @tx_global_lock: XXX: need comments on this one 1901 * @xdp_bulkq: XDP device bulk queue 1902 * @xps_maps: all CPUs/RXQs maps for XPS device 1903 * 1904 * @xps_maps: XXX: need comments on this one 1905 * @miniq_egress: clsact qdisc specific data for 1906 * egress processing 1907 * @nf_hooks_egress: netfilter hooks executed for egress packets 1908 * @qdisc_hash: qdisc hash table 1909 * @watchdog_timeo: Represents the timeout that is used by 1910 * the watchdog (see dev_watchdog()) 1911 * @watchdog_timer: List of timers 1912 * 1913 * @proto_down_reason: reason a netdev interface is held down 1914 * @pcpu_refcnt: Number of references to this device 1915 * @dev_refcnt: Number of references to this device 1916 * @refcnt_tracker: Tracker directory for tracked references to this device 1917 * @todo_list: Delayed register/unregister 1918 * @link_watch_list: XXX: need comments on this one 1919 * 1920 * @reg_state: Register/unregister state machine 1921 * @dismantle: Device is going to be freed 1922 * @rtnl_link_state: This enum represents the phases of creating 1923 * a new link 1924 * 1925 * @needs_free_netdev: Should unregister perform free_netdev? 1926 * @priv_destructor: Called from unregister 1927 * @npinfo: XXX: need comments on this one 1928 * @nd_net: Network namespace this network device is inside 1929 * 1930 * @ml_priv: Mid-layer private 1931 * @ml_priv_type: Mid-layer private type 1932 * @lstats: Loopback statistics 1933 * @tstats: Tunnel statistics 1934 * @dstats: Dummy statistics 1935 * @vstats: Virtual ethernet statistics 1936 * 1937 * @garp_port: GARP 1938 * @mrp_port: MRP 1939 * 1940 * @dm_private: Drop monitor private 1941 * 1942 * @dev: Class/net/name entry 1943 * @sysfs_groups: Space for optional device, statistics and wireless 1944 * sysfs groups 1945 * 1946 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1947 * @rtnl_link_ops: Rtnl_link_ops 1948 * 1949 * @gso_max_size: Maximum size of generic segmentation offload 1950 * @tso_max_size: Device (as in HW) limit on the max TSO request size 1951 * @gso_max_segs: Maximum number of segments that can be passed to the 1952 * NIC for GSO 1953 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count 1954 * 1955 * @dcbnl_ops: Data Center Bridging netlink ops 1956 * @num_tc: Number of traffic classes in the net device 1957 * @tc_to_txq: XXX: need comments on this one 1958 * @prio_tc_map: XXX: need comments on this one 1959 * 1960 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1961 * 1962 * @priomap: XXX: need comments on this one 1963 * @phydev: Physical device may attach itself 1964 * for hardware timestamping 1965 * @sfp_bus: attached &struct sfp_bus structure. 1966 * 1967 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1968 * 1969 * @proto_down: protocol port state information can be sent to the 1970 * switch driver and used to set the phys state of the 1971 * switch port. 1972 * 1973 * @wol_enabled: Wake-on-LAN is enabled 1974 * 1975 * @threaded: napi threaded mode is enabled 1976 * 1977 * @net_notifier_list: List of per-net netdev notifier block 1978 * that follow this device when it is moved 1979 * to another network namespace. 1980 * 1981 * @macsec_ops: MACsec offloading ops 1982 * 1983 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1984 * offload capabilities of the device 1985 * @udp_tunnel_nic: UDP tunnel offload state 1986 * @xdp_state: stores info on attached XDP BPF programs 1987 * 1988 * @nested_level: Used as a parameter of spin_lock_nested() of 1989 * dev->addr_list_lock. 1990 * @unlink_list: As netif_addr_lock() can be called recursively, 1991 * keep a list of interfaces to be deleted. 1992 * @gro_max_size: Maximum size of aggregated packet in generic 1993 * receive offload (GRO) 1994 * 1995 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 1996 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 1997 * @watchdog_dev_tracker: refcount tracker used by watchdog. 1998 * @dev_registered_tracker: tracker for reference held while 1999 * registered 2000 * @offload_xstats_l3: L3 HW stats for this netdevice. 2001 * 2002 * @devlink_port: Pointer to related devlink port structure. 2003 * Assigned by a driver before netdev registration using 2004 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static 2005 * during the time netdevice is registered. 2006 * 2007 * FIXME: cleanup struct net_device such that network protocol info 2008 * moves out. 2009 */ 2010 2011 struct net_device { 2012 char name[IFNAMSIZ]; 2013 struct netdev_name_node *name_node; 2014 struct dev_ifalias __rcu *ifalias; 2015 /* 2016 * I/O specific fields 2017 * FIXME: Merge these and struct ifmap into one 2018 */ 2019 unsigned long mem_end; 2020 unsigned long mem_start; 2021 unsigned long base_addr; 2022 2023 /* 2024 * Some hardware also needs these fields (state,dev_list, 2025 * napi_list,unreg_list,close_list) but they are not 2026 * part of the usual set specified in Space.c. 2027 */ 2028 2029 unsigned long state; 2030 2031 struct list_head dev_list; 2032 struct list_head napi_list; 2033 struct list_head unreg_list; 2034 struct list_head close_list; 2035 struct list_head ptype_all; 2036 struct list_head ptype_specific; 2037 2038 struct { 2039 struct list_head upper; 2040 struct list_head lower; 2041 } adj_list; 2042 2043 /* Read-mostly cache-line for fast-path access */ 2044 unsigned int flags; 2045 unsigned long long priv_flags; 2046 const struct net_device_ops *netdev_ops; 2047 int ifindex; 2048 unsigned short gflags; 2049 unsigned short hard_header_len; 2050 2051 /* Note : dev->mtu is often read without holding a lock. 2052 * Writers usually hold RTNL. 2053 * It is recommended to use READ_ONCE() to annotate the reads, 2054 * and to use WRITE_ONCE() to annotate the writes. 2055 */ 2056 unsigned int mtu; 2057 unsigned short needed_headroom; 2058 unsigned short needed_tailroom; 2059 2060 netdev_features_t features; 2061 netdev_features_t hw_features; 2062 netdev_features_t wanted_features; 2063 netdev_features_t vlan_features; 2064 netdev_features_t hw_enc_features; 2065 netdev_features_t mpls_features; 2066 netdev_features_t gso_partial_features; 2067 2068 unsigned int min_mtu; 2069 unsigned int max_mtu; 2070 unsigned short type; 2071 unsigned char min_header_len; 2072 unsigned char name_assign_type; 2073 2074 int group; 2075 2076 struct net_device_stats stats; /* not used by modern drivers */ 2077 2078 struct net_device_core_stats __percpu *core_stats; 2079 2080 /* Stats to monitor link on/off, flapping */ 2081 atomic_t carrier_up_count; 2082 atomic_t carrier_down_count; 2083 2084 #ifdef CONFIG_WIRELESS_EXT 2085 const struct iw_handler_def *wireless_handlers; 2086 struct iw_public_data *wireless_data; 2087 #endif 2088 const struct ethtool_ops *ethtool_ops; 2089 #ifdef CONFIG_NET_L3_MASTER_DEV 2090 const struct l3mdev_ops *l3mdev_ops; 2091 #endif 2092 #if IS_ENABLED(CONFIG_IPV6) 2093 const struct ndisc_ops *ndisc_ops; 2094 #endif 2095 2096 #ifdef CONFIG_XFRM_OFFLOAD 2097 const struct xfrmdev_ops *xfrmdev_ops; 2098 #endif 2099 2100 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2101 const struct tlsdev_ops *tlsdev_ops; 2102 #endif 2103 2104 const struct header_ops *header_ops; 2105 2106 unsigned char operstate; 2107 unsigned char link_mode; 2108 2109 unsigned char if_port; 2110 unsigned char dma; 2111 2112 /* Interface address info. */ 2113 unsigned char perm_addr[MAX_ADDR_LEN]; 2114 unsigned char addr_assign_type; 2115 unsigned char addr_len; 2116 unsigned char upper_level; 2117 unsigned char lower_level; 2118 2119 unsigned short neigh_priv_len; 2120 unsigned short dev_id; 2121 unsigned short dev_port; 2122 unsigned short padded; 2123 2124 spinlock_t addr_list_lock; 2125 int irq; 2126 2127 struct netdev_hw_addr_list uc; 2128 struct netdev_hw_addr_list mc; 2129 struct netdev_hw_addr_list dev_addrs; 2130 2131 #ifdef CONFIG_SYSFS 2132 struct kset *queues_kset; 2133 #endif 2134 #ifdef CONFIG_LOCKDEP 2135 struct list_head unlink_list; 2136 #endif 2137 unsigned int promiscuity; 2138 unsigned int allmulti; 2139 bool uc_promisc; 2140 #ifdef CONFIG_LOCKDEP 2141 unsigned char nested_level; 2142 #endif 2143 2144 2145 /* Protocol-specific pointers */ 2146 2147 struct in_device __rcu *ip_ptr; 2148 struct inet6_dev __rcu *ip6_ptr; 2149 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2150 struct vlan_info __rcu *vlan_info; 2151 #endif 2152 #if IS_ENABLED(CONFIG_NET_DSA) 2153 struct dsa_port *dsa_ptr; 2154 #endif 2155 #if IS_ENABLED(CONFIG_TIPC) 2156 struct tipc_bearer __rcu *tipc_ptr; 2157 #endif 2158 #if IS_ENABLED(CONFIG_ATALK) 2159 void *atalk_ptr; 2160 #endif 2161 #if IS_ENABLED(CONFIG_AX25) 2162 void *ax25_ptr; 2163 #endif 2164 #if IS_ENABLED(CONFIG_CFG80211) 2165 struct wireless_dev *ieee80211_ptr; 2166 #endif 2167 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2168 struct wpan_dev *ieee802154_ptr; 2169 #endif 2170 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2171 struct mpls_dev __rcu *mpls_ptr; 2172 #endif 2173 #if IS_ENABLED(CONFIG_MCTP) 2174 struct mctp_dev __rcu *mctp_ptr; 2175 #endif 2176 2177 /* 2178 * Cache lines mostly used on receive path (including eth_type_trans()) 2179 */ 2180 /* Interface address info used in eth_type_trans() */ 2181 const unsigned char *dev_addr; 2182 2183 struct netdev_rx_queue *_rx; 2184 unsigned int num_rx_queues; 2185 unsigned int real_num_rx_queues; 2186 2187 struct bpf_prog __rcu *xdp_prog; 2188 unsigned long gro_flush_timeout; 2189 int napi_defer_hard_irqs; 2190 #define GRO_LEGACY_MAX_SIZE 65536u 2191 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2192 * and shinfo->gso_segs is a 16bit field. 2193 */ 2194 #define GRO_MAX_SIZE (8 * 65535u) 2195 unsigned int gro_max_size; 2196 rx_handler_func_t __rcu *rx_handler; 2197 void __rcu *rx_handler_data; 2198 2199 #ifdef CONFIG_NET_CLS_ACT 2200 struct mini_Qdisc __rcu *miniq_ingress; 2201 #endif 2202 struct netdev_queue __rcu *ingress_queue; 2203 #ifdef CONFIG_NETFILTER_INGRESS 2204 struct nf_hook_entries __rcu *nf_hooks_ingress; 2205 #endif 2206 2207 unsigned char broadcast[MAX_ADDR_LEN]; 2208 #ifdef CONFIG_RFS_ACCEL 2209 struct cpu_rmap *rx_cpu_rmap; 2210 #endif 2211 struct hlist_node index_hlist; 2212 2213 /* 2214 * Cache lines mostly used on transmit path 2215 */ 2216 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2217 unsigned int num_tx_queues; 2218 unsigned int real_num_tx_queues; 2219 struct Qdisc __rcu *qdisc; 2220 unsigned int tx_queue_len; 2221 spinlock_t tx_global_lock; 2222 2223 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2224 2225 #ifdef CONFIG_XPS 2226 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2227 #endif 2228 #ifdef CONFIG_NET_CLS_ACT 2229 struct mini_Qdisc __rcu *miniq_egress; 2230 #endif 2231 #ifdef CONFIG_NETFILTER_EGRESS 2232 struct nf_hook_entries __rcu *nf_hooks_egress; 2233 #endif 2234 2235 #ifdef CONFIG_NET_SCHED 2236 DECLARE_HASHTABLE (qdisc_hash, 4); 2237 #endif 2238 /* These may be needed for future network-power-down code. */ 2239 struct timer_list watchdog_timer; 2240 int watchdog_timeo; 2241 2242 u32 proto_down_reason; 2243 2244 struct list_head todo_list; 2245 2246 #ifdef CONFIG_PCPU_DEV_REFCNT 2247 int __percpu *pcpu_refcnt; 2248 #else 2249 refcount_t dev_refcnt; 2250 #endif 2251 struct ref_tracker_dir refcnt_tracker; 2252 2253 struct list_head link_watch_list; 2254 2255 enum { NETREG_UNINITIALIZED=0, 2256 NETREG_REGISTERED, /* completed register_netdevice */ 2257 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2258 NETREG_UNREGISTERED, /* completed unregister todo */ 2259 NETREG_RELEASED, /* called free_netdev */ 2260 NETREG_DUMMY, /* dummy device for NAPI poll */ 2261 } reg_state:8; 2262 2263 bool dismantle; 2264 2265 enum { 2266 RTNL_LINK_INITIALIZED, 2267 RTNL_LINK_INITIALIZING, 2268 } rtnl_link_state:16; 2269 2270 bool needs_free_netdev; 2271 void (*priv_destructor)(struct net_device *dev); 2272 2273 #ifdef CONFIG_NETPOLL 2274 struct netpoll_info __rcu *npinfo; 2275 #endif 2276 2277 possible_net_t nd_net; 2278 2279 /* mid-layer private */ 2280 void *ml_priv; 2281 enum netdev_ml_priv_type ml_priv_type; 2282 2283 union { 2284 struct pcpu_lstats __percpu *lstats; 2285 struct pcpu_sw_netstats __percpu *tstats; 2286 struct pcpu_dstats __percpu *dstats; 2287 }; 2288 2289 #if IS_ENABLED(CONFIG_GARP) 2290 struct garp_port __rcu *garp_port; 2291 #endif 2292 #if IS_ENABLED(CONFIG_MRP) 2293 struct mrp_port __rcu *mrp_port; 2294 #endif 2295 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2296 struct dm_hw_stat_delta __rcu *dm_private; 2297 #endif 2298 struct device dev; 2299 const struct attribute_group *sysfs_groups[4]; 2300 const struct attribute_group *sysfs_rx_queue_group; 2301 2302 const struct rtnl_link_ops *rtnl_link_ops; 2303 2304 /* for setting kernel sock attribute on TCP connection setup */ 2305 #define GSO_MAX_SEGS 65535u 2306 #define GSO_LEGACY_MAX_SIZE 65536u 2307 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2308 * and shinfo->gso_segs is a 16bit field. 2309 */ 2310 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2311 2312 unsigned int gso_max_size; 2313 #define TSO_LEGACY_MAX_SIZE 65536 2314 #define TSO_MAX_SIZE UINT_MAX 2315 unsigned int tso_max_size; 2316 u16 gso_max_segs; 2317 #define TSO_MAX_SEGS U16_MAX 2318 u16 tso_max_segs; 2319 2320 #ifdef CONFIG_DCB 2321 const struct dcbnl_rtnl_ops *dcbnl_ops; 2322 #endif 2323 s16 num_tc; 2324 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2325 u8 prio_tc_map[TC_BITMASK + 1]; 2326 2327 #if IS_ENABLED(CONFIG_FCOE) 2328 unsigned int fcoe_ddp_xid; 2329 #endif 2330 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2331 struct netprio_map __rcu *priomap; 2332 #endif 2333 struct phy_device *phydev; 2334 struct sfp_bus *sfp_bus; 2335 struct lock_class_key *qdisc_tx_busylock; 2336 bool proto_down; 2337 unsigned wol_enabled:1; 2338 unsigned threaded:1; 2339 2340 struct list_head net_notifier_list; 2341 2342 #if IS_ENABLED(CONFIG_MACSEC) 2343 /* MACsec management functions */ 2344 const struct macsec_ops *macsec_ops; 2345 #endif 2346 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2347 struct udp_tunnel_nic *udp_tunnel_nic; 2348 2349 /* protected by rtnl_lock */ 2350 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2351 2352 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2353 netdevice_tracker linkwatch_dev_tracker; 2354 netdevice_tracker watchdog_dev_tracker; 2355 netdevice_tracker dev_registered_tracker; 2356 struct rtnl_hw_stats64 *offload_xstats_l3; 2357 2358 struct devlink_port *devlink_port; 2359 }; 2360 #define to_net_dev(d) container_of(d, struct net_device, dev) 2361 2362 /* 2363 * Driver should use this to assign devlink port instance to a netdevice 2364 * before it registers the netdevice. Therefore devlink_port is static 2365 * during the netdev lifetime after it is registered. 2366 */ 2367 #define SET_NETDEV_DEVLINK_PORT(dev, port) \ 2368 ({ \ 2369 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \ 2370 ((dev)->devlink_port = (port)); \ 2371 }) 2372 2373 static inline bool netif_elide_gro(const struct net_device *dev) 2374 { 2375 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2376 return true; 2377 return false; 2378 } 2379 2380 #define NETDEV_ALIGN 32 2381 2382 static inline 2383 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2384 { 2385 return dev->prio_tc_map[prio & TC_BITMASK]; 2386 } 2387 2388 static inline 2389 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2390 { 2391 if (tc >= dev->num_tc) 2392 return -EINVAL; 2393 2394 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2395 return 0; 2396 } 2397 2398 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2399 void netdev_reset_tc(struct net_device *dev); 2400 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2401 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2402 2403 static inline 2404 int netdev_get_num_tc(struct net_device *dev) 2405 { 2406 return dev->num_tc; 2407 } 2408 2409 static inline void net_prefetch(void *p) 2410 { 2411 prefetch(p); 2412 #if L1_CACHE_BYTES < 128 2413 prefetch((u8 *)p + L1_CACHE_BYTES); 2414 #endif 2415 } 2416 2417 static inline void net_prefetchw(void *p) 2418 { 2419 prefetchw(p); 2420 #if L1_CACHE_BYTES < 128 2421 prefetchw((u8 *)p + L1_CACHE_BYTES); 2422 #endif 2423 } 2424 2425 void netdev_unbind_sb_channel(struct net_device *dev, 2426 struct net_device *sb_dev); 2427 int netdev_bind_sb_channel_queue(struct net_device *dev, 2428 struct net_device *sb_dev, 2429 u8 tc, u16 count, u16 offset); 2430 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2431 static inline int netdev_get_sb_channel(struct net_device *dev) 2432 { 2433 return max_t(int, -dev->num_tc, 0); 2434 } 2435 2436 static inline 2437 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2438 unsigned int index) 2439 { 2440 return &dev->_tx[index]; 2441 } 2442 2443 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2444 const struct sk_buff *skb) 2445 { 2446 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2447 } 2448 2449 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2450 void (*f)(struct net_device *, 2451 struct netdev_queue *, 2452 void *), 2453 void *arg) 2454 { 2455 unsigned int i; 2456 2457 for (i = 0; i < dev->num_tx_queues; i++) 2458 f(dev, &dev->_tx[i], arg); 2459 } 2460 2461 #define netdev_lockdep_set_classes(dev) \ 2462 { \ 2463 static struct lock_class_key qdisc_tx_busylock_key; \ 2464 static struct lock_class_key qdisc_xmit_lock_key; \ 2465 static struct lock_class_key dev_addr_list_lock_key; \ 2466 unsigned int i; \ 2467 \ 2468 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2469 lockdep_set_class(&(dev)->addr_list_lock, \ 2470 &dev_addr_list_lock_key); \ 2471 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2472 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2473 &qdisc_xmit_lock_key); \ 2474 } 2475 2476 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2477 struct net_device *sb_dev); 2478 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2479 struct sk_buff *skb, 2480 struct net_device *sb_dev); 2481 2482 /* returns the headroom that the master device needs to take in account 2483 * when forwarding to this dev 2484 */ 2485 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2486 { 2487 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2488 } 2489 2490 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2491 { 2492 if (dev->netdev_ops->ndo_set_rx_headroom) 2493 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2494 } 2495 2496 /* set the device rx headroom to the dev's default */ 2497 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2498 { 2499 netdev_set_rx_headroom(dev, -1); 2500 } 2501 2502 static inline void *netdev_get_ml_priv(struct net_device *dev, 2503 enum netdev_ml_priv_type type) 2504 { 2505 if (dev->ml_priv_type != type) 2506 return NULL; 2507 2508 return dev->ml_priv; 2509 } 2510 2511 static inline void netdev_set_ml_priv(struct net_device *dev, 2512 void *ml_priv, 2513 enum netdev_ml_priv_type type) 2514 { 2515 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2516 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2517 dev->ml_priv_type, type); 2518 WARN(!dev->ml_priv_type && dev->ml_priv, 2519 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2520 2521 dev->ml_priv = ml_priv; 2522 dev->ml_priv_type = type; 2523 } 2524 2525 /* 2526 * Net namespace inlines 2527 */ 2528 static inline 2529 struct net *dev_net(const struct net_device *dev) 2530 { 2531 return read_pnet(&dev->nd_net); 2532 } 2533 2534 static inline 2535 void dev_net_set(struct net_device *dev, struct net *net) 2536 { 2537 write_pnet(&dev->nd_net, net); 2538 } 2539 2540 /** 2541 * netdev_priv - access network device private data 2542 * @dev: network device 2543 * 2544 * Get network device private data 2545 */ 2546 static inline void *netdev_priv(const struct net_device *dev) 2547 { 2548 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2549 } 2550 2551 /* Set the sysfs physical device reference for the network logical device 2552 * if set prior to registration will cause a symlink during initialization. 2553 */ 2554 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2555 2556 /* Set the sysfs device type for the network logical device to allow 2557 * fine-grained identification of different network device types. For 2558 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2559 */ 2560 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2561 2562 /* Default NAPI poll() weight 2563 * Device drivers are strongly advised to not use bigger value 2564 */ 2565 #define NAPI_POLL_WEIGHT 64 2566 2567 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2568 int (*poll)(struct napi_struct *, int), int weight); 2569 2570 /** 2571 * netif_napi_add() - initialize a NAPI context 2572 * @dev: network device 2573 * @napi: NAPI context 2574 * @poll: polling function 2575 * 2576 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2577 * *any* of the other NAPI-related functions. 2578 */ 2579 static inline void 2580 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2581 int (*poll)(struct napi_struct *, int)) 2582 { 2583 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2584 } 2585 2586 static inline void 2587 netif_napi_add_tx_weight(struct net_device *dev, 2588 struct napi_struct *napi, 2589 int (*poll)(struct napi_struct *, int), 2590 int weight) 2591 { 2592 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2593 netif_napi_add_weight(dev, napi, poll, weight); 2594 } 2595 2596 /** 2597 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2598 * @dev: network device 2599 * @napi: NAPI context 2600 * @poll: polling function 2601 * 2602 * This variant of netif_napi_add() should be used from drivers using NAPI 2603 * to exclusively poll a TX queue. 2604 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2605 */ 2606 static inline void netif_napi_add_tx(struct net_device *dev, 2607 struct napi_struct *napi, 2608 int (*poll)(struct napi_struct *, int)) 2609 { 2610 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2611 } 2612 2613 /** 2614 * __netif_napi_del - remove a NAPI context 2615 * @napi: NAPI context 2616 * 2617 * Warning: caller must observe RCU grace period before freeing memory 2618 * containing @napi. Drivers might want to call this helper to combine 2619 * all the needed RCU grace periods into a single one. 2620 */ 2621 void __netif_napi_del(struct napi_struct *napi); 2622 2623 /** 2624 * netif_napi_del - remove a NAPI context 2625 * @napi: NAPI context 2626 * 2627 * netif_napi_del() removes a NAPI context from the network device NAPI list 2628 */ 2629 static inline void netif_napi_del(struct napi_struct *napi) 2630 { 2631 __netif_napi_del(napi); 2632 synchronize_net(); 2633 } 2634 2635 struct packet_type { 2636 __be16 type; /* This is really htons(ether_type). */ 2637 bool ignore_outgoing; 2638 struct net_device *dev; /* NULL is wildcarded here */ 2639 netdevice_tracker dev_tracker; 2640 int (*func) (struct sk_buff *, 2641 struct net_device *, 2642 struct packet_type *, 2643 struct net_device *); 2644 void (*list_func) (struct list_head *, 2645 struct packet_type *, 2646 struct net_device *); 2647 bool (*id_match)(struct packet_type *ptype, 2648 struct sock *sk); 2649 struct net *af_packet_net; 2650 void *af_packet_priv; 2651 struct list_head list; 2652 }; 2653 2654 struct offload_callbacks { 2655 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2656 netdev_features_t features); 2657 struct sk_buff *(*gro_receive)(struct list_head *head, 2658 struct sk_buff *skb); 2659 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2660 }; 2661 2662 struct packet_offload { 2663 __be16 type; /* This is really htons(ether_type). */ 2664 u16 priority; 2665 struct offload_callbacks callbacks; 2666 struct list_head list; 2667 }; 2668 2669 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2670 struct pcpu_sw_netstats { 2671 u64_stats_t rx_packets; 2672 u64_stats_t rx_bytes; 2673 u64_stats_t tx_packets; 2674 u64_stats_t tx_bytes; 2675 struct u64_stats_sync syncp; 2676 } __aligned(4 * sizeof(u64)); 2677 2678 struct pcpu_lstats { 2679 u64_stats_t packets; 2680 u64_stats_t bytes; 2681 struct u64_stats_sync syncp; 2682 } __aligned(2 * sizeof(u64)); 2683 2684 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2685 2686 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2687 { 2688 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2689 2690 u64_stats_update_begin(&tstats->syncp); 2691 u64_stats_add(&tstats->rx_bytes, len); 2692 u64_stats_inc(&tstats->rx_packets); 2693 u64_stats_update_end(&tstats->syncp); 2694 } 2695 2696 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2697 unsigned int packets, 2698 unsigned int len) 2699 { 2700 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2701 2702 u64_stats_update_begin(&tstats->syncp); 2703 u64_stats_add(&tstats->tx_bytes, len); 2704 u64_stats_add(&tstats->tx_packets, packets); 2705 u64_stats_update_end(&tstats->syncp); 2706 } 2707 2708 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2709 { 2710 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2711 2712 u64_stats_update_begin(&lstats->syncp); 2713 u64_stats_add(&lstats->bytes, len); 2714 u64_stats_inc(&lstats->packets); 2715 u64_stats_update_end(&lstats->syncp); 2716 } 2717 2718 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2719 ({ \ 2720 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2721 if (pcpu_stats) { \ 2722 int __cpu; \ 2723 for_each_possible_cpu(__cpu) { \ 2724 typeof(type) *stat; \ 2725 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2726 u64_stats_init(&stat->syncp); \ 2727 } \ 2728 } \ 2729 pcpu_stats; \ 2730 }) 2731 2732 #define netdev_alloc_pcpu_stats(type) \ 2733 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2734 2735 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2736 ({ \ 2737 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2738 if (pcpu_stats) { \ 2739 int __cpu; \ 2740 for_each_possible_cpu(__cpu) { \ 2741 typeof(type) *stat; \ 2742 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2743 u64_stats_init(&stat->syncp); \ 2744 } \ 2745 } \ 2746 pcpu_stats; \ 2747 }) 2748 2749 enum netdev_lag_tx_type { 2750 NETDEV_LAG_TX_TYPE_UNKNOWN, 2751 NETDEV_LAG_TX_TYPE_RANDOM, 2752 NETDEV_LAG_TX_TYPE_BROADCAST, 2753 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2754 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2755 NETDEV_LAG_TX_TYPE_HASH, 2756 }; 2757 2758 enum netdev_lag_hash { 2759 NETDEV_LAG_HASH_NONE, 2760 NETDEV_LAG_HASH_L2, 2761 NETDEV_LAG_HASH_L34, 2762 NETDEV_LAG_HASH_L23, 2763 NETDEV_LAG_HASH_E23, 2764 NETDEV_LAG_HASH_E34, 2765 NETDEV_LAG_HASH_VLAN_SRCMAC, 2766 NETDEV_LAG_HASH_UNKNOWN, 2767 }; 2768 2769 struct netdev_lag_upper_info { 2770 enum netdev_lag_tx_type tx_type; 2771 enum netdev_lag_hash hash_type; 2772 }; 2773 2774 struct netdev_lag_lower_state_info { 2775 u8 link_up : 1, 2776 tx_enabled : 1; 2777 }; 2778 2779 #include <linux/notifier.h> 2780 2781 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2782 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2783 * adding new types. 2784 */ 2785 enum netdev_cmd { 2786 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2787 NETDEV_DOWN, 2788 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2789 detected a hardware crash and restarted 2790 - we can use this eg to kick tcp sessions 2791 once done */ 2792 NETDEV_CHANGE, /* Notify device state change */ 2793 NETDEV_REGISTER, 2794 NETDEV_UNREGISTER, 2795 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2796 NETDEV_CHANGEADDR, /* notify after the address change */ 2797 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2798 NETDEV_GOING_DOWN, 2799 NETDEV_CHANGENAME, 2800 NETDEV_FEAT_CHANGE, 2801 NETDEV_BONDING_FAILOVER, 2802 NETDEV_PRE_UP, 2803 NETDEV_PRE_TYPE_CHANGE, 2804 NETDEV_POST_TYPE_CHANGE, 2805 NETDEV_POST_INIT, 2806 NETDEV_PRE_UNINIT, 2807 NETDEV_RELEASE, 2808 NETDEV_NOTIFY_PEERS, 2809 NETDEV_JOIN, 2810 NETDEV_CHANGEUPPER, 2811 NETDEV_RESEND_IGMP, 2812 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2813 NETDEV_CHANGEINFODATA, 2814 NETDEV_BONDING_INFO, 2815 NETDEV_PRECHANGEUPPER, 2816 NETDEV_CHANGELOWERSTATE, 2817 NETDEV_UDP_TUNNEL_PUSH_INFO, 2818 NETDEV_UDP_TUNNEL_DROP_INFO, 2819 NETDEV_CHANGE_TX_QUEUE_LEN, 2820 NETDEV_CVLAN_FILTER_PUSH_INFO, 2821 NETDEV_CVLAN_FILTER_DROP_INFO, 2822 NETDEV_SVLAN_FILTER_PUSH_INFO, 2823 NETDEV_SVLAN_FILTER_DROP_INFO, 2824 NETDEV_OFFLOAD_XSTATS_ENABLE, 2825 NETDEV_OFFLOAD_XSTATS_DISABLE, 2826 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 2827 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 2828 }; 2829 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2830 2831 int register_netdevice_notifier(struct notifier_block *nb); 2832 int unregister_netdevice_notifier(struct notifier_block *nb); 2833 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2834 int unregister_netdevice_notifier_net(struct net *net, 2835 struct notifier_block *nb); 2836 void move_netdevice_notifier_net(struct net *src_net, struct net *dst_net, 2837 struct notifier_block *nb); 2838 int register_netdevice_notifier_dev_net(struct net_device *dev, 2839 struct notifier_block *nb, 2840 struct netdev_net_notifier *nn); 2841 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2842 struct notifier_block *nb, 2843 struct netdev_net_notifier *nn); 2844 2845 struct netdev_notifier_info { 2846 struct net_device *dev; 2847 struct netlink_ext_ack *extack; 2848 }; 2849 2850 struct netdev_notifier_info_ext { 2851 struct netdev_notifier_info info; /* must be first */ 2852 union { 2853 u32 mtu; 2854 } ext; 2855 }; 2856 2857 struct netdev_notifier_change_info { 2858 struct netdev_notifier_info info; /* must be first */ 2859 unsigned int flags_changed; 2860 }; 2861 2862 struct netdev_notifier_changeupper_info { 2863 struct netdev_notifier_info info; /* must be first */ 2864 struct net_device *upper_dev; /* new upper dev */ 2865 bool master; /* is upper dev master */ 2866 bool linking; /* is the notification for link or unlink */ 2867 void *upper_info; /* upper dev info */ 2868 }; 2869 2870 struct netdev_notifier_changelowerstate_info { 2871 struct netdev_notifier_info info; /* must be first */ 2872 void *lower_state_info; /* is lower dev state */ 2873 }; 2874 2875 struct netdev_notifier_pre_changeaddr_info { 2876 struct netdev_notifier_info info; /* must be first */ 2877 const unsigned char *dev_addr; 2878 }; 2879 2880 enum netdev_offload_xstats_type { 2881 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 2882 }; 2883 2884 struct netdev_notifier_offload_xstats_info { 2885 struct netdev_notifier_info info; /* must be first */ 2886 enum netdev_offload_xstats_type type; 2887 2888 union { 2889 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 2890 struct netdev_notifier_offload_xstats_rd *report_delta; 2891 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 2892 struct netdev_notifier_offload_xstats_ru *report_used; 2893 }; 2894 }; 2895 2896 int netdev_offload_xstats_enable(struct net_device *dev, 2897 enum netdev_offload_xstats_type type, 2898 struct netlink_ext_ack *extack); 2899 int netdev_offload_xstats_disable(struct net_device *dev, 2900 enum netdev_offload_xstats_type type); 2901 bool netdev_offload_xstats_enabled(const struct net_device *dev, 2902 enum netdev_offload_xstats_type type); 2903 int netdev_offload_xstats_get(struct net_device *dev, 2904 enum netdev_offload_xstats_type type, 2905 struct rtnl_hw_stats64 *stats, bool *used, 2906 struct netlink_ext_ack *extack); 2907 void 2908 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 2909 const struct rtnl_hw_stats64 *stats); 2910 void 2911 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 2912 void netdev_offload_xstats_push_delta(struct net_device *dev, 2913 enum netdev_offload_xstats_type type, 2914 const struct rtnl_hw_stats64 *stats); 2915 2916 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2917 struct net_device *dev) 2918 { 2919 info->dev = dev; 2920 info->extack = NULL; 2921 } 2922 2923 static inline struct net_device * 2924 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2925 { 2926 return info->dev; 2927 } 2928 2929 static inline struct netlink_ext_ack * 2930 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2931 { 2932 return info->extack; 2933 } 2934 2935 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2936 2937 2938 extern rwlock_t dev_base_lock; /* Device list lock */ 2939 2940 #define for_each_netdev(net, d) \ 2941 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2942 #define for_each_netdev_reverse(net, d) \ 2943 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2944 #define for_each_netdev_rcu(net, d) \ 2945 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2946 #define for_each_netdev_safe(net, d, n) \ 2947 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2948 #define for_each_netdev_continue(net, d) \ 2949 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2950 #define for_each_netdev_continue_reverse(net, d) \ 2951 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2952 dev_list) 2953 #define for_each_netdev_continue_rcu(net, d) \ 2954 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2955 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2956 for_each_netdev_rcu(&init_net, slave) \ 2957 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2958 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2959 2960 static inline struct net_device *next_net_device(struct net_device *dev) 2961 { 2962 struct list_head *lh; 2963 struct net *net; 2964 2965 net = dev_net(dev); 2966 lh = dev->dev_list.next; 2967 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2968 } 2969 2970 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2971 { 2972 struct list_head *lh; 2973 struct net *net; 2974 2975 net = dev_net(dev); 2976 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2977 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2978 } 2979 2980 static inline struct net_device *first_net_device(struct net *net) 2981 { 2982 return list_empty(&net->dev_base_head) ? NULL : 2983 net_device_entry(net->dev_base_head.next); 2984 } 2985 2986 static inline struct net_device *first_net_device_rcu(struct net *net) 2987 { 2988 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2989 2990 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2991 } 2992 2993 int netdev_boot_setup_check(struct net_device *dev); 2994 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2995 const char *hwaddr); 2996 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2997 void dev_add_pack(struct packet_type *pt); 2998 void dev_remove_pack(struct packet_type *pt); 2999 void __dev_remove_pack(struct packet_type *pt); 3000 void dev_add_offload(struct packet_offload *po); 3001 void dev_remove_offload(struct packet_offload *po); 3002 3003 int dev_get_iflink(const struct net_device *dev); 3004 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 3005 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 3006 struct net_device_path_stack *stack); 3007 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 3008 unsigned short mask); 3009 struct net_device *dev_get_by_name(struct net *net, const char *name); 3010 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 3011 struct net_device *__dev_get_by_name(struct net *net, const char *name); 3012 bool netdev_name_in_use(struct net *net, const char *name); 3013 int dev_alloc_name(struct net_device *dev, const char *name); 3014 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 3015 void dev_close(struct net_device *dev); 3016 void dev_close_many(struct list_head *head, bool unlink); 3017 void dev_disable_lro(struct net_device *dev); 3018 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 3019 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 3020 struct net_device *sb_dev); 3021 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 3022 struct net_device *sb_dev); 3023 3024 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3025 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3026 3027 static inline int dev_queue_xmit(struct sk_buff *skb) 3028 { 3029 return __dev_queue_xmit(skb, NULL); 3030 } 3031 3032 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3033 struct net_device *sb_dev) 3034 { 3035 return __dev_queue_xmit(skb, sb_dev); 3036 } 3037 3038 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3039 { 3040 int ret; 3041 3042 ret = __dev_direct_xmit(skb, queue_id); 3043 if (!dev_xmit_complete(ret)) 3044 kfree_skb(skb); 3045 return ret; 3046 } 3047 3048 int register_netdevice(struct net_device *dev); 3049 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3050 void unregister_netdevice_many(struct list_head *head); 3051 static inline void unregister_netdevice(struct net_device *dev) 3052 { 3053 unregister_netdevice_queue(dev, NULL); 3054 } 3055 3056 int netdev_refcnt_read(const struct net_device *dev); 3057 void free_netdev(struct net_device *dev); 3058 void netdev_freemem(struct net_device *dev); 3059 int init_dummy_netdev(struct net_device *dev); 3060 3061 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3062 struct sk_buff *skb, 3063 bool all_slaves); 3064 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3065 struct sock *sk); 3066 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3067 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3068 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3069 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 3070 int dev_restart(struct net_device *dev); 3071 3072 3073 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3074 unsigned short type, 3075 const void *daddr, const void *saddr, 3076 unsigned int len) 3077 { 3078 if (!dev->header_ops || !dev->header_ops->create) 3079 return 0; 3080 3081 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3082 } 3083 3084 static inline int dev_parse_header(const struct sk_buff *skb, 3085 unsigned char *haddr) 3086 { 3087 const struct net_device *dev = skb->dev; 3088 3089 if (!dev->header_ops || !dev->header_ops->parse) 3090 return 0; 3091 return dev->header_ops->parse(skb, haddr); 3092 } 3093 3094 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3095 { 3096 const struct net_device *dev = skb->dev; 3097 3098 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3099 return 0; 3100 return dev->header_ops->parse_protocol(skb); 3101 } 3102 3103 /* ll_header must have at least hard_header_len allocated */ 3104 static inline bool dev_validate_header(const struct net_device *dev, 3105 char *ll_header, int len) 3106 { 3107 if (likely(len >= dev->hard_header_len)) 3108 return true; 3109 if (len < dev->min_header_len) 3110 return false; 3111 3112 if (capable(CAP_SYS_RAWIO)) { 3113 memset(ll_header + len, 0, dev->hard_header_len - len); 3114 return true; 3115 } 3116 3117 if (dev->header_ops && dev->header_ops->validate) 3118 return dev->header_ops->validate(ll_header, len); 3119 3120 return false; 3121 } 3122 3123 static inline bool dev_has_header(const struct net_device *dev) 3124 { 3125 return dev->header_ops && dev->header_ops->create; 3126 } 3127 3128 /* 3129 * Incoming packets are placed on per-CPU queues 3130 */ 3131 struct softnet_data { 3132 struct list_head poll_list; 3133 struct sk_buff_head process_queue; 3134 3135 /* stats */ 3136 unsigned int processed; 3137 unsigned int time_squeeze; 3138 #ifdef CONFIG_RPS 3139 struct softnet_data *rps_ipi_list; 3140 #endif 3141 #ifdef CONFIG_NET_FLOW_LIMIT 3142 struct sd_flow_limit __rcu *flow_limit; 3143 #endif 3144 struct Qdisc *output_queue; 3145 struct Qdisc **output_queue_tailp; 3146 struct sk_buff *completion_queue; 3147 #ifdef CONFIG_XFRM_OFFLOAD 3148 struct sk_buff_head xfrm_backlog; 3149 #endif 3150 /* written and read only by owning cpu: */ 3151 struct { 3152 u16 recursion; 3153 u8 more; 3154 #ifdef CONFIG_NET_EGRESS 3155 u8 skip_txqueue; 3156 #endif 3157 } xmit; 3158 #ifdef CONFIG_RPS 3159 /* input_queue_head should be written by cpu owning this struct, 3160 * and only read by other cpus. Worth using a cache line. 3161 */ 3162 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3163 3164 /* Elements below can be accessed between CPUs for RPS/RFS */ 3165 call_single_data_t csd ____cacheline_aligned_in_smp; 3166 struct softnet_data *rps_ipi_next; 3167 unsigned int cpu; 3168 unsigned int input_queue_tail; 3169 #endif 3170 unsigned int received_rps; 3171 unsigned int dropped; 3172 struct sk_buff_head input_pkt_queue; 3173 struct napi_struct backlog; 3174 3175 /* Another possibly contended cache line */ 3176 spinlock_t defer_lock ____cacheline_aligned_in_smp; 3177 int defer_count; 3178 int defer_ipi_scheduled; 3179 struct sk_buff *defer_list; 3180 call_single_data_t defer_csd; 3181 }; 3182 3183 static inline void input_queue_head_incr(struct softnet_data *sd) 3184 { 3185 #ifdef CONFIG_RPS 3186 sd->input_queue_head++; 3187 #endif 3188 } 3189 3190 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3191 unsigned int *qtail) 3192 { 3193 #ifdef CONFIG_RPS 3194 *qtail = ++sd->input_queue_tail; 3195 #endif 3196 } 3197 3198 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3199 3200 static inline int dev_recursion_level(void) 3201 { 3202 return this_cpu_read(softnet_data.xmit.recursion); 3203 } 3204 3205 #define XMIT_RECURSION_LIMIT 8 3206 static inline bool dev_xmit_recursion(void) 3207 { 3208 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3209 XMIT_RECURSION_LIMIT); 3210 } 3211 3212 static inline void dev_xmit_recursion_inc(void) 3213 { 3214 __this_cpu_inc(softnet_data.xmit.recursion); 3215 } 3216 3217 static inline void dev_xmit_recursion_dec(void) 3218 { 3219 __this_cpu_dec(softnet_data.xmit.recursion); 3220 } 3221 3222 void __netif_schedule(struct Qdisc *q); 3223 void netif_schedule_queue(struct netdev_queue *txq); 3224 3225 static inline void netif_tx_schedule_all(struct net_device *dev) 3226 { 3227 unsigned int i; 3228 3229 for (i = 0; i < dev->num_tx_queues; i++) 3230 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3231 } 3232 3233 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3234 { 3235 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3236 } 3237 3238 /** 3239 * netif_start_queue - allow transmit 3240 * @dev: network device 3241 * 3242 * Allow upper layers to call the device hard_start_xmit routine. 3243 */ 3244 static inline void netif_start_queue(struct net_device *dev) 3245 { 3246 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3247 } 3248 3249 static inline void netif_tx_start_all_queues(struct net_device *dev) 3250 { 3251 unsigned int i; 3252 3253 for (i = 0; i < dev->num_tx_queues; i++) { 3254 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3255 netif_tx_start_queue(txq); 3256 } 3257 } 3258 3259 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3260 3261 /** 3262 * netif_wake_queue - restart transmit 3263 * @dev: network device 3264 * 3265 * Allow upper layers to call the device hard_start_xmit routine. 3266 * Used for flow control when transmit resources are available. 3267 */ 3268 static inline void netif_wake_queue(struct net_device *dev) 3269 { 3270 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3271 } 3272 3273 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3274 { 3275 unsigned int i; 3276 3277 for (i = 0; i < dev->num_tx_queues; i++) { 3278 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3279 netif_tx_wake_queue(txq); 3280 } 3281 } 3282 3283 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3284 { 3285 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3286 } 3287 3288 /** 3289 * netif_stop_queue - stop transmitted packets 3290 * @dev: network device 3291 * 3292 * Stop upper layers calling the device hard_start_xmit routine. 3293 * Used for flow control when transmit resources are unavailable. 3294 */ 3295 static inline void netif_stop_queue(struct net_device *dev) 3296 { 3297 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3298 } 3299 3300 void netif_tx_stop_all_queues(struct net_device *dev); 3301 3302 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3303 { 3304 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3305 } 3306 3307 /** 3308 * netif_queue_stopped - test if transmit queue is flowblocked 3309 * @dev: network device 3310 * 3311 * Test if transmit queue on device is currently unable to send. 3312 */ 3313 static inline bool netif_queue_stopped(const struct net_device *dev) 3314 { 3315 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3316 } 3317 3318 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3319 { 3320 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3321 } 3322 3323 static inline bool 3324 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3325 { 3326 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3327 } 3328 3329 static inline bool 3330 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3331 { 3332 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3333 } 3334 3335 /** 3336 * netdev_queue_set_dql_min_limit - set dql minimum limit 3337 * @dev_queue: pointer to transmit queue 3338 * @min_limit: dql minimum limit 3339 * 3340 * Forces xmit_more() to return true until the minimum threshold 3341 * defined by @min_limit is reached (or until the tx queue is 3342 * empty). Warning: to be use with care, misuse will impact the 3343 * latency. 3344 */ 3345 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3346 unsigned int min_limit) 3347 { 3348 #ifdef CONFIG_BQL 3349 dev_queue->dql.min_limit = min_limit; 3350 #endif 3351 } 3352 3353 /** 3354 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3355 * @dev_queue: pointer to transmit queue 3356 * 3357 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3358 * to give appropriate hint to the CPU. 3359 */ 3360 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3361 { 3362 #ifdef CONFIG_BQL 3363 prefetchw(&dev_queue->dql.num_queued); 3364 #endif 3365 } 3366 3367 /** 3368 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3369 * @dev_queue: pointer to transmit queue 3370 * 3371 * BQL enabled drivers might use this helper in their TX completion path, 3372 * to give appropriate hint to the CPU. 3373 */ 3374 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3375 { 3376 #ifdef CONFIG_BQL 3377 prefetchw(&dev_queue->dql.limit); 3378 #endif 3379 } 3380 3381 /** 3382 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3383 * @dev_queue: network device queue 3384 * @bytes: number of bytes queued to the device queue 3385 * 3386 * Report the number of bytes queued for sending/completion to the network 3387 * device hardware queue. @bytes should be a good approximation and should 3388 * exactly match netdev_completed_queue() @bytes. 3389 * This is typically called once per packet, from ndo_start_xmit(). 3390 */ 3391 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3392 unsigned int bytes) 3393 { 3394 #ifdef CONFIG_BQL 3395 dql_queued(&dev_queue->dql, bytes); 3396 3397 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3398 return; 3399 3400 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3401 3402 /* 3403 * The XOFF flag must be set before checking the dql_avail below, 3404 * because in netdev_tx_completed_queue we update the dql_completed 3405 * before checking the XOFF flag. 3406 */ 3407 smp_mb(); 3408 3409 /* check again in case another CPU has just made room avail */ 3410 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3411 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3412 #endif 3413 } 3414 3415 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3416 * that they should not test BQL status themselves. 3417 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3418 * skb of a batch. 3419 * Returns true if the doorbell must be used to kick the NIC. 3420 */ 3421 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3422 unsigned int bytes, 3423 bool xmit_more) 3424 { 3425 if (xmit_more) { 3426 #ifdef CONFIG_BQL 3427 dql_queued(&dev_queue->dql, bytes); 3428 #endif 3429 return netif_tx_queue_stopped(dev_queue); 3430 } 3431 netdev_tx_sent_queue(dev_queue, bytes); 3432 return true; 3433 } 3434 3435 /** 3436 * netdev_sent_queue - report the number of bytes queued to hardware 3437 * @dev: network device 3438 * @bytes: number of bytes queued to the hardware device queue 3439 * 3440 * Report the number of bytes queued for sending/completion to the network 3441 * device hardware queue#0. @bytes should be a good approximation and should 3442 * exactly match netdev_completed_queue() @bytes. 3443 * This is typically called once per packet, from ndo_start_xmit(). 3444 */ 3445 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3446 { 3447 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3448 } 3449 3450 static inline bool __netdev_sent_queue(struct net_device *dev, 3451 unsigned int bytes, 3452 bool xmit_more) 3453 { 3454 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3455 xmit_more); 3456 } 3457 3458 /** 3459 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3460 * @dev_queue: network device queue 3461 * @pkts: number of packets (currently ignored) 3462 * @bytes: number of bytes dequeued from the device queue 3463 * 3464 * Must be called at most once per TX completion round (and not per 3465 * individual packet), so that BQL can adjust its limits appropriately. 3466 */ 3467 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3468 unsigned int pkts, unsigned int bytes) 3469 { 3470 #ifdef CONFIG_BQL 3471 if (unlikely(!bytes)) 3472 return; 3473 3474 dql_completed(&dev_queue->dql, bytes); 3475 3476 /* 3477 * Without the memory barrier there is a small possiblity that 3478 * netdev_tx_sent_queue will miss the update and cause the queue to 3479 * be stopped forever 3480 */ 3481 smp_mb(); 3482 3483 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3484 return; 3485 3486 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3487 netif_schedule_queue(dev_queue); 3488 #endif 3489 } 3490 3491 /** 3492 * netdev_completed_queue - report bytes and packets completed by device 3493 * @dev: network device 3494 * @pkts: actual number of packets sent over the medium 3495 * @bytes: actual number of bytes sent over the medium 3496 * 3497 * Report the number of bytes and packets transmitted by the network device 3498 * hardware queue over the physical medium, @bytes must exactly match the 3499 * @bytes amount passed to netdev_sent_queue() 3500 */ 3501 static inline void netdev_completed_queue(struct net_device *dev, 3502 unsigned int pkts, unsigned int bytes) 3503 { 3504 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3505 } 3506 3507 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3508 { 3509 #ifdef CONFIG_BQL 3510 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3511 dql_reset(&q->dql); 3512 #endif 3513 } 3514 3515 /** 3516 * netdev_reset_queue - reset the packets and bytes count of a network device 3517 * @dev_queue: network device 3518 * 3519 * Reset the bytes and packet count of a network device and clear the 3520 * software flow control OFF bit for this network device 3521 */ 3522 static inline void netdev_reset_queue(struct net_device *dev_queue) 3523 { 3524 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3525 } 3526 3527 /** 3528 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3529 * @dev: network device 3530 * @queue_index: given tx queue index 3531 * 3532 * Returns 0 if given tx queue index >= number of device tx queues, 3533 * otherwise returns the originally passed tx queue index. 3534 */ 3535 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3536 { 3537 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3538 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3539 dev->name, queue_index, 3540 dev->real_num_tx_queues); 3541 return 0; 3542 } 3543 3544 return queue_index; 3545 } 3546 3547 /** 3548 * netif_running - test if up 3549 * @dev: network device 3550 * 3551 * Test if the device has been brought up. 3552 */ 3553 static inline bool netif_running(const struct net_device *dev) 3554 { 3555 return test_bit(__LINK_STATE_START, &dev->state); 3556 } 3557 3558 /* 3559 * Routines to manage the subqueues on a device. We only need start, 3560 * stop, and a check if it's stopped. All other device management is 3561 * done at the overall netdevice level. 3562 * Also test the device if we're multiqueue. 3563 */ 3564 3565 /** 3566 * netif_start_subqueue - allow sending packets on subqueue 3567 * @dev: network device 3568 * @queue_index: sub queue index 3569 * 3570 * Start individual transmit queue of a device with multiple transmit queues. 3571 */ 3572 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3573 { 3574 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3575 3576 netif_tx_start_queue(txq); 3577 } 3578 3579 /** 3580 * netif_stop_subqueue - stop sending packets on subqueue 3581 * @dev: network device 3582 * @queue_index: sub queue index 3583 * 3584 * Stop individual transmit queue of a device with multiple transmit queues. 3585 */ 3586 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3587 { 3588 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3589 netif_tx_stop_queue(txq); 3590 } 3591 3592 /** 3593 * __netif_subqueue_stopped - test status of subqueue 3594 * @dev: network device 3595 * @queue_index: sub queue index 3596 * 3597 * Check individual transmit queue of a device with multiple transmit queues. 3598 */ 3599 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3600 u16 queue_index) 3601 { 3602 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3603 3604 return netif_tx_queue_stopped(txq); 3605 } 3606 3607 /** 3608 * netif_subqueue_stopped - test status of subqueue 3609 * @dev: network device 3610 * @skb: sub queue buffer pointer 3611 * 3612 * Check individual transmit queue of a device with multiple transmit queues. 3613 */ 3614 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3615 struct sk_buff *skb) 3616 { 3617 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3618 } 3619 3620 /** 3621 * netif_wake_subqueue - allow sending packets on subqueue 3622 * @dev: network device 3623 * @queue_index: sub queue index 3624 * 3625 * Resume individual transmit queue of a device with multiple transmit queues. 3626 */ 3627 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3628 { 3629 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3630 3631 netif_tx_wake_queue(txq); 3632 } 3633 3634 #ifdef CONFIG_XPS 3635 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3636 u16 index); 3637 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3638 u16 index, enum xps_map_type type); 3639 3640 /** 3641 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3642 * @j: CPU/Rx queue index 3643 * @mask: bitmask of all cpus/rx queues 3644 * @nr_bits: number of bits in the bitmask 3645 * 3646 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3647 */ 3648 static inline bool netif_attr_test_mask(unsigned long j, 3649 const unsigned long *mask, 3650 unsigned int nr_bits) 3651 { 3652 cpu_max_bits_warn(j, nr_bits); 3653 return test_bit(j, mask); 3654 } 3655 3656 /** 3657 * netif_attr_test_online - Test for online CPU/Rx queue 3658 * @j: CPU/Rx queue index 3659 * @online_mask: bitmask for CPUs/Rx queues that are online 3660 * @nr_bits: number of bits in the bitmask 3661 * 3662 * Returns true if a CPU/Rx queue is online. 3663 */ 3664 static inline bool netif_attr_test_online(unsigned long j, 3665 const unsigned long *online_mask, 3666 unsigned int nr_bits) 3667 { 3668 cpu_max_bits_warn(j, nr_bits); 3669 3670 if (online_mask) 3671 return test_bit(j, online_mask); 3672 3673 return (j < nr_bits); 3674 } 3675 3676 /** 3677 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3678 * @n: CPU/Rx queue index 3679 * @srcp: the cpumask/Rx queue mask pointer 3680 * @nr_bits: number of bits in the bitmask 3681 * 3682 * Returns >= nr_bits if no further CPUs/Rx queues set. 3683 */ 3684 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3685 unsigned int nr_bits) 3686 { 3687 /* -1 is a legal arg here. */ 3688 if (n != -1) 3689 cpu_max_bits_warn(n, nr_bits); 3690 3691 if (srcp) 3692 return find_next_bit(srcp, nr_bits, n + 1); 3693 3694 return n + 1; 3695 } 3696 3697 /** 3698 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3699 * @n: CPU/Rx queue index 3700 * @src1p: the first CPUs/Rx queues mask pointer 3701 * @src2p: the second CPUs/Rx queues mask pointer 3702 * @nr_bits: number of bits in the bitmask 3703 * 3704 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3705 */ 3706 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3707 const unsigned long *src2p, 3708 unsigned int nr_bits) 3709 { 3710 /* -1 is a legal arg here. */ 3711 if (n != -1) 3712 cpu_max_bits_warn(n, nr_bits); 3713 3714 if (src1p && src2p) 3715 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3716 else if (src1p) 3717 return find_next_bit(src1p, nr_bits, n + 1); 3718 else if (src2p) 3719 return find_next_bit(src2p, nr_bits, n + 1); 3720 3721 return n + 1; 3722 } 3723 #else 3724 static inline int netif_set_xps_queue(struct net_device *dev, 3725 const struct cpumask *mask, 3726 u16 index) 3727 { 3728 return 0; 3729 } 3730 3731 static inline int __netif_set_xps_queue(struct net_device *dev, 3732 const unsigned long *mask, 3733 u16 index, enum xps_map_type type) 3734 { 3735 return 0; 3736 } 3737 #endif 3738 3739 /** 3740 * netif_is_multiqueue - test if device has multiple transmit queues 3741 * @dev: network device 3742 * 3743 * Check if device has multiple transmit queues 3744 */ 3745 static inline bool netif_is_multiqueue(const struct net_device *dev) 3746 { 3747 return dev->num_tx_queues > 1; 3748 } 3749 3750 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3751 3752 #ifdef CONFIG_SYSFS 3753 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3754 #else 3755 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3756 unsigned int rxqs) 3757 { 3758 dev->real_num_rx_queues = rxqs; 3759 return 0; 3760 } 3761 #endif 3762 int netif_set_real_num_queues(struct net_device *dev, 3763 unsigned int txq, unsigned int rxq); 3764 3765 static inline struct netdev_rx_queue * 3766 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3767 { 3768 return dev->_rx + rxq; 3769 } 3770 3771 #ifdef CONFIG_SYSFS 3772 static inline unsigned int get_netdev_rx_queue_index( 3773 struct netdev_rx_queue *queue) 3774 { 3775 struct net_device *dev = queue->dev; 3776 int index = queue - dev->_rx; 3777 3778 BUG_ON(index >= dev->num_rx_queues); 3779 return index; 3780 } 3781 #endif 3782 3783 int netif_get_num_default_rss_queues(void); 3784 3785 enum skb_free_reason { 3786 SKB_REASON_CONSUMED, 3787 SKB_REASON_DROPPED, 3788 }; 3789 3790 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3791 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3792 3793 /* 3794 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3795 * interrupt context or with hardware interrupts being disabled. 3796 * (in_hardirq() || irqs_disabled()) 3797 * 3798 * We provide four helpers that can be used in following contexts : 3799 * 3800 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3801 * replacing kfree_skb(skb) 3802 * 3803 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3804 * Typically used in place of consume_skb(skb) in TX completion path 3805 * 3806 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3807 * replacing kfree_skb(skb) 3808 * 3809 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3810 * and consumed a packet. Used in place of consume_skb(skb) 3811 */ 3812 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3813 { 3814 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3815 } 3816 3817 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3818 { 3819 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3820 } 3821 3822 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3823 { 3824 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3825 } 3826 3827 static inline void dev_consume_skb_any(struct sk_buff *skb) 3828 { 3829 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3830 } 3831 3832 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3833 struct bpf_prog *xdp_prog); 3834 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3835 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3836 int netif_rx(struct sk_buff *skb); 3837 int __netif_rx(struct sk_buff *skb); 3838 3839 int netif_receive_skb(struct sk_buff *skb); 3840 int netif_receive_skb_core(struct sk_buff *skb); 3841 void netif_receive_skb_list_internal(struct list_head *head); 3842 void netif_receive_skb_list(struct list_head *head); 3843 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3844 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3845 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3846 void napi_get_frags_check(struct napi_struct *napi); 3847 gro_result_t napi_gro_frags(struct napi_struct *napi); 3848 struct packet_offload *gro_find_receive_by_type(__be16 type); 3849 struct packet_offload *gro_find_complete_by_type(__be16 type); 3850 3851 static inline void napi_free_frags(struct napi_struct *napi) 3852 { 3853 kfree_skb(napi->skb); 3854 napi->skb = NULL; 3855 } 3856 3857 bool netdev_is_rx_handler_busy(struct net_device *dev); 3858 int netdev_rx_handler_register(struct net_device *dev, 3859 rx_handler_func_t *rx_handler, 3860 void *rx_handler_data); 3861 void netdev_rx_handler_unregister(struct net_device *dev); 3862 3863 bool dev_valid_name(const char *name); 3864 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 3865 { 3866 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 3867 } 3868 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 3869 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 3870 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3871 void __user *data, bool *need_copyout); 3872 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 3873 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 3874 unsigned int dev_get_flags(const struct net_device *); 3875 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3876 struct netlink_ext_ack *extack); 3877 int dev_change_flags(struct net_device *dev, unsigned int flags, 3878 struct netlink_ext_ack *extack); 3879 int dev_set_alias(struct net_device *, const char *, size_t); 3880 int dev_get_alias(const struct net_device *, char *, size_t); 3881 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 3882 const char *pat, int new_ifindex); 3883 static inline 3884 int dev_change_net_namespace(struct net_device *dev, struct net *net, 3885 const char *pat) 3886 { 3887 return __dev_change_net_namespace(dev, net, pat, 0); 3888 } 3889 int __dev_set_mtu(struct net_device *, int); 3890 int dev_set_mtu(struct net_device *, int); 3891 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3892 struct netlink_ext_ack *extack); 3893 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3894 struct netlink_ext_ack *extack); 3895 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 3896 struct netlink_ext_ack *extack); 3897 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 3898 int dev_get_port_parent_id(struct net_device *dev, 3899 struct netdev_phys_item_id *ppid, bool recurse); 3900 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3901 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3902 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3903 struct netdev_queue *txq, int *ret); 3904 3905 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3906 u8 dev_xdp_prog_count(struct net_device *dev); 3907 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3908 3909 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3910 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3911 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 3912 bool is_skb_forwardable(const struct net_device *dev, 3913 const struct sk_buff *skb); 3914 3915 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 3916 const struct sk_buff *skb, 3917 const bool check_mtu) 3918 { 3919 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 3920 unsigned int len; 3921 3922 if (!(dev->flags & IFF_UP)) 3923 return false; 3924 3925 if (!check_mtu) 3926 return true; 3927 3928 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 3929 if (skb->len <= len) 3930 return true; 3931 3932 /* if TSO is enabled, we don't care about the length as the packet 3933 * could be forwarded without being segmented before 3934 */ 3935 if (skb_is_gso(skb)) 3936 return true; 3937 3938 return false; 3939 } 3940 3941 struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev); 3942 3943 static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev) 3944 { 3945 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */ 3946 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats); 3947 3948 if (likely(p)) 3949 return p; 3950 3951 return netdev_core_stats_alloc(dev); 3952 } 3953 3954 #define DEV_CORE_STATS_INC(FIELD) \ 3955 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 3956 { \ 3957 struct net_device_core_stats __percpu *p; \ 3958 \ 3959 p = dev_core_stats(dev); \ 3960 if (p) \ 3961 this_cpu_inc(p->FIELD); \ 3962 } 3963 DEV_CORE_STATS_INC(rx_dropped) 3964 DEV_CORE_STATS_INC(tx_dropped) 3965 DEV_CORE_STATS_INC(rx_nohandler) 3966 DEV_CORE_STATS_INC(rx_otherhost_dropped) 3967 3968 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3969 struct sk_buff *skb, 3970 const bool check_mtu) 3971 { 3972 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3973 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 3974 dev_core_stats_rx_dropped_inc(dev); 3975 kfree_skb(skb); 3976 return NET_RX_DROP; 3977 } 3978 3979 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 3980 skb->priority = 0; 3981 return 0; 3982 } 3983 3984 bool dev_nit_active(struct net_device *dev); 3985 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3986 3987 static inline void __dev_put(struct net_device *dev) 3988 { 3989 if (dev) { 3990 #ifdef CONFIG_PCPU_DEV_REFCNT 3991 this_cpu_dec(*dev->pcpu_refcnt); 3992 #else 3993 refcount_dec(&dev->dev_refcnt); 3994 #endif 3995 } 3996 } 3997 3998 static inline void __dev_hold(struct net_device *dev) 3999 { 4000 if (dev) { 4001 #ifdef CONFIG_PCPU_DEV_REFCNT 4002 this_cpu_inc(*dev->pcpu_refcnt); 4003 #else 4004 refcount_inc(&dev->dev_refcnt); 4005 #endif 4006 } 4007 } 4008 4009 static inline void __netdev_tracker_alloc(struct net_device *dev, 4010 netdevice_tracker *tracker, 4011 gfp_t gfp) 4012 { 4013 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4014 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 4015 #endif 4016 } 4017 4018 /* netdev_tracker_alloc() can upgrade a prior untracked reference 4019 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4020 */ 4021 static inline void netdev_tracker_alloc(struct net_device *dev, 4022 netdevice_tracker *tracker, gfp_t gfp) 4023 { 4024 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4025 refcount_dec(&dev->refcnt_tracker.no_tracker); 4026 __netdev_tracker_alloc(dev, tracker, gfp); 4027 #endif 4028 } 4029 4030 static inline void netdev_tracker_free(struct net_device *dev, 4031 netdevice_tracker *tracker) 4032 { 4033 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4034 ref_tracker_free(&dev->refcnt_tracker, tracker); 4035 #endif 4036 } 4037 4038 static inline void netdev_hold(struct net_device *dev, 4039 netdevice_tracker *tracker, gfp_t gfp) 4040 { 4041 if (dev) { 4042 __dev_hold(dev); 4043 __netdev_tracker_alloc(dev, tracker, gfp); 4044 } 4045 } 4046 4047 static inline void netdev_put(struct net_device *dev, 4048 netdevice_tracker *tracker) 4049 { 4050 if (dev) { 4051 netdev_tracker_free(dev, tracker); 4052 __dev_put(dev); 4053 } 4054 } 4055 4056 /** 4057 * dev_hold - get reference to device 4058 * @dev: network device 4059 * 4060 * Hold reference to device to keep it from being freed. 4061 * Try using netdev_hold() instead. 4062 */ 4063 static inline void dev_hold(struct net_device *dev) 4064 { 4065 netdev_hold(dev, NULL, GFP_ATOMIC); 4066 } 4067 4068 /** 4069 * dev_put - release reference to device 4070 * @dev: network device 4071 * 4072 * Release reference to device to allow it to be freed. 4073 * Try using netdev_put() instead. 4074 */ 4075 static inline void dev_put(struct net_device *dev) 4076 { 4077 netdev_put(dev, NULL); 4078 } 4079 4080 static inline void netdev_ref_replace(struct net_device *odev, 4081 struct net_device *ndev, 4082 netdevice_tracker *tracker, 4083 gfp_t gfp) 4084 { 4085 if (odev) 4086 netdev_tracker_free(odev, tracker); 4087 4088 __dev_hold(ndev); 4089 __dev_put(odev); 4090 4091 if (ndev) 4092 __netdev_tracker_alloc(ndev, tracker, gfp); 4093 } 4094 4095 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4096 * and _off may be called from IRQ context, but it is caller 4097 * who is responsible for serialization of these calls. 4098 * 4099 * The name carrier is inappropriate, these functions should really be 4100 * called netif_lowerlayer_*() because they represent the state of any 4101 * kind of lower layer not just hardware media. 4102 */ 4103 void linkwatch_fire_event(struct net_device *dev); 4104 4105 /** 4106 * netif_carrier_ok - test if carrier present 4107 * @dev: network device 4108 * 4109 * Check if carrier is present on device 4110 */ 4111 static inline bool netif_carrier_ok(const struct net_device *dev) 4112 { 4113 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4114 } 4115 4116 unsigned long dev_trans_start(struct net_device *dev); 4117 4118 void __netdev_watchdog_up(struct net_device *dev); 4119 4120 void netif_carrier_on(struct net_device *dev); 4121 void netif_carrier_off(struct net_device *dev); 4122 void netif_carrier_event(struct net_device *dev); 4123 4124 /** 4125 * netif_dormant_on - mark device as dormant. 4126 * @dev: network device 4127 * 4128 * Mark device as dormant (as per RFC2863). 4129 * 4130 * The dormant state indicates that the relevant interface is not 4131 * actually in a condition to pass packets (i.e., it is not 'up') but is 4132 * in a "pending" state, waiting for some external event. For "on- 4133 * demand" interfaces, this new state identifies the situation where the 4134 * interface is waiting for events to place it in the up state. 4135 */ 4136 static inline void netif_dormant_on(struct net_device *dev) 4137 { 4138 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4139 linkwatch_fire_event(dev); 4140 } 4141 4142 /** 4143 * netif_dormant_off - set device as not dormant. 4144 * @dev: network device 4145 * 4146 * Device is not in dormant state. 4147 */ 4148 static inline void netif_dormant_off(struct net_device *dev) 4149 { 4150 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4151 linkwatch_fire_event(dev); 4152 } 4153 4154 /** 4155 * netif_dormant - test if device is dormant 4156 * @dev: network device 4157 * 4158 * Check if device is dormant. 4159 */ 4160 static inline bool netif_dormant(const struct net_device *dev) 4161 { 4162 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4163 } 4164 4165 4166 /** 4167 * netif_testing_on - mark device as under test. 4168 * @dev: network device 4169 * 4170 * Mark device as under test (as per RFC2863). 4171 * 4172 * The testing state indicates that some test(s) must be performed on 4173 * the interface. After completion, of the test, the interface state 4174 * will change to up, dormant, or down, as appropriate. 4175 */ 4176 static inline void netif_testing_on(struct net_device *dev) 4177 { 4178 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4179 linkwatch_fire_event(dev); 4180 } 4181 4182 /** 4183 * netif_testing_off - set device as not under test. 4184 * @dev: network device 4185 * 4186 * Device is not in testing state. 4187 */ 4188 static inline void netif_testing_off(struct net_device *dev) 4189 { 4190 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4191 linkwatch_fire_event(dev); 4192 } 4193 4194 /** 4195 * netif_testing - test if device is under test 4196 * @dev: network device 4197 * 4198 * Check if device is under test 4199 */ 4200 static inline bool netif_testing(const struct net_device *dev) 4201 { 4202 return test_bit(__LINK_STATE_TESTING, &dev->state); 4203 } 4204 4205 4206 /** 4207 * netif_oper_up - test if device is operational 4208 * @dev: network device 4209 * 4210 * Check if carrier is operational 4211 */ 4212 static inline bool netif_oper_up(const struct net_device *dev) 4213 { 4214 return (dev->operstate == IF_OPER_UP || 4215 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4216 } 4217 4218 /** 4219 * netif_device_present - is device available or removed 4220 * @dev: network device 4221 * 4222 * Check if device has not been removed from system. 4223 */ 4224 static inline bool netif_device_present(const struct net_device *dev) 4225 { 4226 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4227 } 4228 4229 void netif_device_detach(struct net_device *dev); 4230 4231 void netif_device_attach(struct net_device *dev); 4232 4233 /* 4234 * Network interface message level settings 4235 */ 4236 4237 enum { 4238 NETIF_MSG_DRV_BIT, 4239 NETIF_MSG_PROBE_BIT, 4240 NETIF_MSG_LINK_BIT, 4241 NETIF_MSG_TIMER_BIT, 4242 NETIF_MSG_IFDOWN_BIT, 4243 NETIF_MSG_IFUP_BIT, 4244 NETIF_MSG_RX_ERR_BIT, 4245 NETIF_MSG_TX_ERR_BIT, 4246 NETIF_MSG_TX_QUEUED_BIT, 4247 NETIF_MSG_INTR_BIT, 4248 NETIF_MSG_TX_DONE_BIT, 4249 NETIF_MSG_RX_STATUS_BIT, 4250 NETIF_MSG_PKTDATA_BIT, 4251 NETIF_MSG_HW_BIT, 4252 NETIF_MSG_WOL_BIT, 4253 4254 /* When you add a new bit above, update netif_msg_class_names array 4255 * in net/ethtool/common.c 4256 */ 4257 NETIF_MSG_CLASS_COUNT, 4258 }; 4259 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4260 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4261 4262 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4263 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4264 4265 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4266 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4267 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4268 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4269 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4270 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4271 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4272 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4273 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4274 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4275 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4276 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4277 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4278 #define NETIF_MSG_HW __NETIF_MSG(HW) 4279 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4280 4281 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4282 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4283 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4284 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4285 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4286 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4287 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4288 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4289 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4290 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4291 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4292 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4293 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4294 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4295 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4296 4297 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4298 { 4299 /* use default */ 4300 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4301 return default_msg_enable_bits; 4302 if (debug_value == 0) /* no output */ 4303 return 0; 4304 /* set low N bits */ 4305 return (1U << debug_value) - 1; 4306 } 4307 4308 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4309 { 4310 spin_lock(&txq->_xmit_lock); 4311 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4312 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4313 } 4314 4315 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4316 { 4317 __acquire(&txq->_xmit_lock); 4318 return true; 4319 } 4320 4321 static inline void __netif_tx_release(struct netdev_queue *txq) 4322 { 4323 __release(&txq->_xmit_lock); 4324 } 4325 4326 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4327 { 4328 spin_lock_bh(&txq->_xmit_lock); 4329 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4330 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4331 } 4332 4333 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4334 { 4335 bool ok = spin_trylock(&txq->_xmit_lock); 4336 4337 if (likely(ok)) { 4338 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4339 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4340 } 4341 return ok; 4342 } 4343 4344 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4345 { 4346 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4347 WRITE_ONCE(txq->xmit_lock_owner, -1); 4348 spin_unlock(&txq->_xmit_lock); 4349 } 4350 4351 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4352 { 4353 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4354 WRITE_ONCE(txq->xmit_lock_owner, -1); 4355 spin_unlock_bh(&txq->_xmit_lock); 4356 } 4357 4358 /* 4359 * txq->trans_start can be read locklessly from dev_watchdog() 4360 */ 4361 static inline void txq_trans_update(struct netdev_queue *txq) 4362 { 4363 if (txq->xmit_lock_owner != -1) 4364 WRITE_ONCE(txq->trans_start, jiffies); 4365 } 4366 4367 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4368 { 4369 unsigned long now = jiffies; 4370 4371 if (READ_ONCE(txq->trans_start) != now) 4372 WRITE_ONCE(txq->trans_start, now); 4373 } 4374 4375 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4376 static inline void netif_trans_update(struct net_device *dev) 4377 { 4378 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4379 4380 txq_trans_cond_update(txq); 4381 } 4382 4383 /** 4384 * netif_tx_lock - grab network device transmit lock 4385 * @dev: network device 4386 * 4387 * Get network device transmit lock 4388 */ 4389 void netif_tx_lock(struct net_device *dev); 4390 4391 static inline void netif_tx_lock_bh(struct net_device *dev) 4392 { 4393 local_bh_disable(); 4394 netif_tx_lock(dev); 4395 } 4396 4397 void netif_tx_unlock(struct net_device *dev); 4398 4399 static inline void netif_tx_unlock_bh(struct net_device *dev) 4400 { 4401 netif_tx_unlock(dev); 4402 local_bh_enable(); 4403 } 4404 4405 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4406 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4407 __netif_tx_lock(txq, cpu); \ 4408 } else { \ 4409 __netif_tx_acquire(txq); \ 4410 } \ 4411 } 4412 4413 #define HARD_TX_TRYLOCK(dev, txq) \ 4414 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4415 __netif_tx_trylock(txq) : \ 4416 __netif_tx_acquire(txq)) 4417 4418 #define HARD_TX_UNLOCK(dev, txq) { \ 4419 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4420 __netif_tx_unlock(txq); \ 4421 } else { \ 4422 __netif_tx_release(txq); \ 4423 } \ 4424 } 4425 4426 static inline void netif_tx_disable(struct net_device *dev) 4427 { 4428 unsigned int i; 4429 int cpu; 4430 4431 local_bh_disable(); 4432 cpu = smp_processor_id(); 4433 spin_lock(&dev->tx_global_lock); 4434 for (i = 0; i < dev->num_tx_queues; i++) { 4435 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4436 4437 __netif_tx_lock(txq, cpu); 4438 netif_tx_stop_queue(txq); 4439 __netif_tx_unlock(txq); 4440 } 4441 spin_unlock(&dev->tx_global_lock); 4442 local_bh_enable(); 4443 } 4444 4445 static inline void netif_addr_lock(struct net_device *dev) 4446 { 4447 unsigned char nest_level = 0; 4448 4449 #ifdef CONFIG_LOCKDEP 4450 nest_level = dev->nested_level; 4451 #endif 4452 spin_lock_nested(&dev->addr_list_lock, nest_level); 4453 } 4454 4455 static inline void netif_addr_lock_bh(struct net_device *dev) 4456 { 4457 unsigned char nest_level = 0; 4458 4459 #ifdef CONFIG_LOCKDEP 4460 nest_level = dev->nested_level; 4461 #endif 4462 local_bh_disable(); 4463 spin_lock_nested(&dev->addr_list_lock, nest_level); 4464 } 4465 4466 static inline void netif_addr_unlock(struct net_device *dev) 4467 { 4468 spin_unlock(&dev->addr_list_lock); 4469 } 4470 4471 static inline void netif_addr_unlock_bh(struct net_device *dev) 4472 { 4473 spin_unlock_bh(&dev->addr_list_lock); 4474 } 4475 4476 /* 4477 * dev_addrs walker. Should be used only for read access. Call with 4478 * rcu_read_lock held. 4479 */ 4480 #define for_each_dev_addr(dev, ha) \ 4481 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4482 4483 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4484 4485 void ether_setup(struct net_device *dev); 4486 4487 /* Support for loadable net-drivers */ 4488 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4489 unsigned char name_assign_type, 4490 void (*setup)(struct net_device *), 4491 unsigned int txqs, unsigned int rxqs); 4492 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4493 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4494 4495 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4496 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4497 count) 4498 4499 int register_netdev(struct net_device *dev); 4500 void unregister_netdev(struct net_device *dev); 4501 4502 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4503 4504 /* General hardware address lists handling functions */ 4505 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4506 struct netdev_hw_addr_list *from_list, int addr_len); 4507 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4508 struct netdev_hw_addr_list *from_list, int addr_len); 4509 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4510 struct net_device *dev, 4511 int (*sync)(struct net_device *, const unsigned char *), 4512 int (*unsync)(struct net_device *, 4513 const unsigned char *)); 4514 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4515 struct net_device *dev, 4516 int (*sync)(struct net_device *, 4517 const unsigned char *, int), 4518 int (*unsync)(struct net_device *, 4519 const unsigned char *, int)); 4520 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4521 struct net_device *dev, 4522 int (*unsync)(struct net_device *, 4523 const unsigned char *, int)); 4524 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4525 struct net_device *dev, 4526 int (*unsync)(struct net_device *, 4527 const unsigned char *)); 4528 void __hw_addr_init(struct netdev_hw_addr_list *list); 4529 4530 /* Functions used for device addresses handling */ 4531 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4532 const void *addr, size_t len); 4533 4534 static inline void 4535 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4536 { 4537 dev_addr_mod(dev, 0, addr, len); 4538 } 4539 4540 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4541 { 4542 __dev_addr_set(dev, addr, dev->addr_len); 4543 } 4544 4545 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4546 unsigned char addr_type); 4547 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4548 unsigned char addr_type); 4549 4550 /* Functions used for unicast addresses handling */ 4551 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4552 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4553 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4554 int dev_uc_sync(struct net_device *to, struct net_device *from); 4555 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4556 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4557 void dev_uc_flush(struct net_device *dev); 4558 void dev_uc_init(struct net_device *dev); 4559 4560 /** 4561 * __dev_uc_sync - Synchonize device's unicast list 4562 * @dev: device to sync 4563 * @sync: function to call if address should be added 4564 * @unsync: function to call if address should be removed 4565 * 4566 * Add newly added addresses to the interface, and release 4567 * addresses that have been deleted. 4568 */ 4569 static inline int __dev_uc_sync(struct net_device *dev, 4570 int (*sync)(struct net_device *, 4571 const unsigned char *), 4572 int (*unsync)(struct net_device *, 4573 const unsigned char *)) 4574 { 4575 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4576 } 4577 4578 /** 4579 * __dev_uc_unsync - Remove synchronized addresses from device 4580 * @dev: device to sync 4581 * @unsync: function to call if address should be removed 4582 * 4583 * Remove all addresses that were added to the device by dev_uc_sync(). 4584 */ 4585 static inline void __dev_uc_unsync(struct net_device *dev, 4586 int (*unsync)(struct net_device *, 4587 const unsigned char *)) 4588 { 4589 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4590 } 4591 4592 /* Functions used for multicast addresses handling */ 4593 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4594 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4595 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4596 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4597 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4598 int dev_mc_sync(struct net_device *to, struct net_device *from); 4599 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4600 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4601 void dev_mc_flush(struct net_device *dev); 4602 void dev_mc_init(struct net_device *dev); 4603 4604 /** 4605 * __dev_mc_sync - Synchonize device's multicast list 4606 * @dev: device to sync 4607 * @sync: function to call if address should be added 4608 * @unsync: function to call if address should be removed 4609 * 4610 * Add newly added addresses to the interface, and release 4611 * addresses that have been deleted. 4612 */ 4613 static inline int __dev_mc_sync(struct net_device *dev, 4614 int (*sync)(struct net_device *, 4615 const unsigned char *), 4616 int (*unsync)(struct net_device *, 4617 const unsigned char *)) 4618 { 4619 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4620 } 4621 4622 /** 4623 * __dev_mc_unsync - Remove synchronized addresses from device 4624 * @dev: device to sync 4625 * @unsync: function to call if address should be removed 4626 * 4627 * Remove all addresses that were added to the device by dev_mc_sync(). 4628 */ 4629 static inline void __dev_mc_unsync(struct net_device *dev, 4630 int (*unsync)(struct net_device *, 4631 const unsigned char *)) 4632 { 4633 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4634 } 4635 4636 /* Functions used for secondary unicast and multicast support */ 4637 void dev_set_rx_mode(struct net_device *dev); 4638 int dev_set_promiscuity(struct net_device *dev, int inc); 4639 int dev_set_allmulti(struct net_device *dev, int inc); 4640 void netdev_state_change(struct net_device *dev); 4641 void __netdev_notify_peers(struct net_device *dev); 4642 void netdev_notify_peers(struct net_device *dev); 4643 void netdev_features_change(struct net_device *dev); 4644 /* Load a device via the kmod */ 4645 void dev_load(struct net *net, const char *name); 4646 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4647 struct rtnl_link_stats64 *storage); 4648 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4649 const struct net_device_stats *netdev_stats); 4650 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4651 const struct pcpu_sw_netstats __percpu *netstats); 4652 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4653 4654 extern int netdev_max_backlog; 4655 extern int dev_rx_weight; 4656 extern int dev_tx_weight; 4657 extern int gro_normal_batch; 4658 4659 enum { 4660 NESTED_SYNC_IMM_BIT, 4661 NESTED_SYNC_TODO_BIT, 4662 }; 4663 4664 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4665 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4666 4667 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4668 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4669 4670 struct netdev_nested_priv { 4671 unsigned char flags; 4672 void *data; 4673 }; 4674 4675 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4676 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4677 struct list_head **iter); 4678 4679 /* iterate through upper list, must be called under RCU read lock */ 4680 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4681 for (iter = &(dev)->adj_list.upper, \ 4682 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4683 updev; \ 4684 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4685 4686 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4687 int (*fn)(struct net_device *upper_dev, 4688 struct netdev_nested_priv *priv), 4689 struct netdev_nested_priv *priv); 4690 4691 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4692 struct net_device *upper_dev); 4693 4694 bool netdev_has_any_upper_dev(struct net_device *dev); 4695 4696 void *netdev_lower_get_next_private(struct net_device *dev, 4697 struct list_head **iter); 4698 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4699 struct list_head **iter); 4700 4701 #define netdev_for_each_lower_private(dev, priv, iter) \ 4702 for (iter = (dev)->adj_list.lower.next, \ 4703 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4704 priv; \ 4705 priv = netdev_lower_get_next_private(dev, &(iter))) 4706 4707 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4708 for (iter = &(dev)->adj_list.lower, \ 4709 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4710 priv; \ 4711 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4712 4713 void *netdev_lower_get_next(struct net_device *dev, 4714 struct list_head **iter); 4715 4716 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4717 for (iter = (dev)->adj_list.lower.next, \ 4718 ldev = netdev_lower_get_next(dev, &(iter)); \ 4719 ldev; \ 4720 ldev = netdev_lower_get_next(dev, &(iter))) 4721 4722 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4723 struct list_head **iter); 4724 int netdev_walk_all_lower_dev(struct net_device *dev, 4725 int (*fn)(struct net_device *lower_dev, 4726 struct netdev_nested_priv *priv), 4727 struct netdev_nested_priv *priv); 4728 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4729 int (*fn)(struct net_device *lower_dev, 4730 struct netdev_nested_priv *priv), 4731 struct netdev_nested_priv *priv); 4732 4733 void *netdev_adjacent_get_private(struct list_head *adj_list); 4734 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4735 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4736 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4737 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4738 struct netlink_ext_ack *extack); 4739 int netdev_master_upper_dev_link(struct net_device *dev, 4740 struct net_device *upper_dev, 4741 void *upper_priv, void *upper_info, 4742 struct netlink_ext_ack *extack); 4743 void netdev_upper_dev_unlink(struct net_device *dev, 4744 struct net_device *upper_dev); 4745 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4746 struct net_device *new_dev, 4747 struct net_device *dev, 4748 struct netlink_ext_ack *extack); 4749 void netdev_adjacent_change_commit(struct net_device *old_dev, 4750 struct net_device *new_dev, 4751 struct net_device *dev); 4752 void netdev_adjacent_change_abort(struct net_device *old_dev, 4753 struct net_device *new_dev, 4754 struct net_device *dev); 4755 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4756 void *netdev_lower_dev_get_private(struct net_device *dev, 4757 struct net_device *lower_dev); 4758 void netdev_lower_state_changed(struct net_device *lower_dev, 4759 void *lower_state_info); 4760 4761 /* RSS keys are 40 or 52 bytes long */ 4762 #define NETDEV_RSS_KEY_LEN 52 4763 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4764 void netdev_rss_key_fill(void *buffer, size_t len); 4765 4766 int skb_checksum_help(struct sk_buff *skb); 4767 int skb_crc32c_csum_help(struct sk_buff *skb); 4768 int skb_csum_hwoffload_help(struct sk_buff *skb, 4769 const netdev_features_t features); 4770 4771 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4772 netdev_features_t features, bool tx_path); 4773 struct sk_buff *skb_eth_gso_segment(struct sk_buff *skb, 4774 netdev_features_t features, __be16 type); 4775 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4776 netdev_features_t features); 4777 4778 struct netdev_bonding_info { 4779 ifslave slave; 4780 ifbond master; 4781 }; 4782 4783 struct netdev_notifier_bonding_info { 4784 struct netdev_notifier_info info; /* must be first */ 4785 struct netdev_bonding_info bonding_info; 4786 }; 4787 4788 void netdev_bonding_info_change(struct net_device *dev, 4789 struct netdev_bonding_info *bonding_info); 4790 4791 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4792 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4793 #else 4794 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4795 const void *data) 4796 { 4797 } 4798 #endif 4799 4800 static inline 4801 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4802 { 4803 return __skb_gso_segment(skb, features, true); 4804 } 4805 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4806 4807 static inline bool can_checksum_protocol(netdev_features_t features, 4808 __be16 protocol) 4809 { 4810 if (protocol == htons(ETH_P_FCOE)) 4811 return !!(features & NETIF_F_FCOE_CRC); 4812 4813 /* Assume this is an IP checksum (not SCTP CRC) */ 4814 4815 if (features & NETIF_F_HW_CSUM) { 4816 /* Can checksum everything */ 4817 return true; 4818 } 4819 4820 switch (protocol) { 4821 case htons(ETH_P_IP): 4822 return !!(features & NETIF_F_IP_CSUM); 4823 case htons(ETH_P_IPV6): 4824 return !!(features & NETIF_F_IPV6_CSUM); 4825 default: 4826 return false; 4827 } 4828 } 4829 4830 #ifdef CONFIG_BUG 4831 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4832 #else 4833 static inline void netdev_rx_csum_fault(struct net_device *dev, 4834 struct sk_buff *skb) 4835 { 4836 } 4837 #endif 4838 /* rx skb timestamps */ 4839 void net_enable_timestamp(void); 4840 void net_disable_timestamp(void); 4841 4842 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 4843 const struct skb_shared_hwtstamps *hwtstamps, 4844 bool cycles) 4845 { 4846 const struct net_device_ops *ops = dev->netdev_ops; 4847 4848 if (ops->ndo_get_tstamp) 4849 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 4850 4851 return hwtstamps->hwtstamp; 4852 } 4853 4854 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4855 struct sk_buff *skb, struct net_device *dev, 4856 bool more) 4857 { 4858 __this_cpu_write(softnet_data.xmit.more, more); 4859 return ops->ndo_start_xmit(skb, dev); 4860 } 4861 4862 static inline bool netdev_xmit_more(void) 4863 { 4864 return __this_cpu_read(softnet_data.xmit.more); 4865 } 4866 4867 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4868 struct netdev_queue *txq, bool more) 4869 { 4870 const struct net_device_ops *ops = dev->netdev_ops; 4871 netdev_tx_t rc; 4872 4873 rc = __netdev_start_xmit(ops, skb, dev, more); 4874 if (rc == NETDEV_TX_OK) 4875 txq_trans_update(txq); 4876 4877 return rc; 4878 } 4879 4880 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4881 const void *ns); 4882 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4883 const void *ns); 4884 4885 extern const struct kobj_ns_type_operations net_ns_type_operations; 4886 4887 const char *netdev_drivername(const struct net_device *dev); 4888 4889 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4890 netdev_features_t f2) 4891 { 4892 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4893 if (f1 & NETIF_F_HW_CSUM) 4894 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4895 else 4896 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4897 } 4898 4899 return f1 & f2; 4900 } 4901 4902 static inline netdev_features_t netdev_get_wanted_features( 4903 struct net_device *dev) 4904 { 4905 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4906 } 4907 netdev_features_t netdev_increment_features(netdev_features_t all, 4908 netdev_features_t one, netdev_features_t mask); 4909 4910 /* Allow TSO being used on stacked device : 4911 * Performing the GSO segmentation before last device 4912 * is a performance improvement. 4913 */ 4914 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4915 netdev_features_t mask) 4916 { 4917 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4918 } 4919 4920 int __netdev_update_features(struct net_device *dev); 4921 void netdev_update_features(struct net_device *dev); 4922 void netdev_change_features(struct net_device *dev); 4923 4924 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4925 struct net_device *dev); 4926 4927 netdev_features_t passthru_features_check(struct sk_buff *skb, 4928 struct net_device *dev, 4929 netdev_features_t features); 4930 netdev_features_t netif_skb_features(struct sk_buff *skb); 4931 4932 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4933 { 4934 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4935 4936 /* check flags correspondence */ 4937 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4938 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4939 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4940 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4941 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4942 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4943 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4944 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4945 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4946 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4947 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4948 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4949 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4950 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4951 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4952 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4953 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4954 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4955 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4956 4957 return (features & feature) == feature; 4958 } 4959 4960 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4961 { 4962 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4963 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4964 } 4965 4966 static inline bool netif_needs_gso(struct sk_buff *skb, 4967 netdev_features_t features) 4968 { 4969 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4970 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4971 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4972 } 4973 4974 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 4975 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 4976 void netif_inherit_tso_max(struct net_device *to, 4977 const struct net_device *from); 4978 4979 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4980 int pulled_hlen, u16 mac_offset, 4981 int mac_len) 4982 { 4983 skb->protocol = protocol; 4984 skb->encapsulation = 1; 4985 skb_push(skb, pulled_hlen); 4986 skb_reset_transport_header(skb); 4987 skb->mac_header = mac_offset; 4988 skb->network_header = skb->mac_header + mac_len; 4989 skb->mac_len = mac_len; 4990 } 4991 4992 static inline bool netif_is_macsec(const struct net_device *dev) 4993 { 4994 return dev->priv_flags & IFF_MACSEC; 4995 } 4996 4997 static inline bool netif_is_macvlan(const struct net_device *dev) 4998 { 4999 return dev->priv_flags & IFF_MACVLAN; 5000 } 5001 5002 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5003 { 5004 return dev->priv_flags & IFF_MACVLAN_PORT; 5005 } 5006 5007 static inline bool netif_is_bond_master(const struct net_device *dev) 5008 { 5009 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5010 } 5011 5012 static inline bool netif_is_bond_slave(const struct net_device *dev) 5013 { 5014 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5015 } 5016 5017 static inline bool netif_supports_nofcs(struct net_device *dev) 5018 { 5019 return dev->priv_flags & IFF_SUPP_NOFCS; 5020 } 5021 5022 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5023 { 5024 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5025 } 5026 5027 static inline bool netif_is_l3_master(const struct net_device *dev) 5028 { 5029 return dev->priv_flags & IFF_L3MDEV_MASTER; 5030 } 5031 5032 static inline bool netif_is_l3_slave(const struct net_device *dev) 5033 { 5034 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5035 } 5036 5037 static inline bool netif_is_bridge_master(const struct net_device *dev) 5038 { 5039 return dev->priv_flags & IFF_EBRIDGE; 5040 } 5041 5042 static inline bool netif_is_bridge_port(const struct net_device *dev) 5043 { 5044 return dev->priv_flags & IFF_BRIDGE_PORT; 5045 } 5046 5047 static inline bool netif_is_ovs_master(const struct net_device *dev) 5048 { 5049 return dev->priv_flags & IFF_OPENVSWITCH; 5050 } 5051 5052 static inline bool netif_is_ovs_port(const struct net_device *dev) 5053 { 5054 return dev->priv_flags & IFF_OVS_DATAPATH; 5055 } 5056 5057 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5058 { 5059 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5060 } 5061 5062 static inline bool netif_is_team_master(const struct net_device *dev) 5063 { 5064 return dev->priv_flags & IFF_TEAM; 5065 } 5066 5067 static inline bool netif_is_team_port(const struct net_device *dev) 5068 { 5069 return dev->priv_flags & IFF_TEAM_PORT; 5070 } 5071 5072 static inline bool netif_is_lag_master(const struct net_device *dev) 5073 { 5074 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5075 } 5076 5077 static inline bool netif_is_lag_port(const struct net_device *dev) 5078 { 5079 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5080 } 5081 5082 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5083 { 5084 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5085 } 5086 5087 static inline bool netif_is_failover(const struct net_device *dev) 5088 { 5089 return dev->priv_flags & IFF_FAILOVER; 5090 } 5091 5092 static inline bool netif_is_failover_slave(const struct net_device *dev) 5093 { 5094 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5095 } 5096 5097 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5098 static inline void netif_keep_dst(struct net_device *dev) 5099 { 5100 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5101 } 5102 5103 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5104 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5105 { 5106 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5107 return netif_is_macsec(dev); 5108 } 5109 5110 extern struct pernet_operations __net_initdata loopback_net_ops; 5111 5112 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5113 5114 /* netdev_printk helpers, similar to dev_printk */ 5115 5116 static inline const char *netdev_name(const struct net_device *dev) 5117 { 5118 if (!dev->name[0] || strchr(dev->name, '%')) 5119 return "(unnamed net_device)"; 5120 return dev->name; 5121 } 5122 5123 static inline const char *netdev_reg_state(const struct net_device *dev) 5124 { 5125 switch (dev->reg_state) { 5126 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5127 case NETREG_REGISTERED: return ""; 5128 case NETREG_UNREGISTERING: return " (unregistering)"; 5129 case NETREG_UNREGISTERED: return " (unregistered)"; 5130 case NETREG_RELEASED: return " (released)"; 5131 case NETREG_DUMMY: return " (dummy)"; 5132 } 5133 5134 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5135 return " (unknown)"; 5136 } 5137 5138 #define MODULE_ALIAS_NETDEV(device) \ 5139 MODULE_ALIAS("netdev-" device) 5140 5141 /* 5142 * netdev_WARN() acts like dev_printk(), but with the key difference 5143 * of using a WARN/WARN_ON to get the message out, including the 5144 * file/line information and a backtrace. 5145 */ 5146 #define netdev_WARN(dev, format, args...) \ 5147 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5148 netdev_reg_state(dev), ##args) 5149 5150 #define netdev_WARN_ONCE(dev, format, args...) \ 5151 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5152 netdev_reg_state(dev), ##args) 5153 5154 /* 5155 * The list of packet types we will receive (as opposed to discard) 5156 * and the routines to invoke. 5157 * 5158 * Why 16. Because with 16 the only overlap we get on a hash of the 5159 * low nibble of the protocol value is RARP/SNAP/X.25. 5160 * 5161 * 0800 IP 5162 * 0001 802.3 5163 * 0002 AX.25 5164 * 0004 802.2 5165 * 8035 RARP 5166 * 0005 SNAP 5167 * 0805 X.25 5168 * 0806 ARP 5169 * 8137 IPX 5170 * 0009 Localtalk 5171 * 86DD IPv6 5172 */ 5173 #define PTYPE_HASH_SIZE (16) 5174 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5175 5176 extern struct list_head ptype_all __read_mostly; 5177 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5178 5179 extern struct net_device *blackhole_netdev; 5180 5181 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */ 5182 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD) 5183 #define DEV_STATS_ADD(DEV, FIELD, VAL) \ 5184 atomic_long_add((VAL), &(DEV)->stats.__##FIELD) 5185 5186 #endif /* _LINUX_NETDEVICE_H */ 5187