1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <[email protected]> 13 * Corey Minyard <[email protected]> 14 * Donald J. Becker, <[email protected]> 15 * Alan Cox, <[email protected]> 16 * Bjorn Ekwall. <[email protected]> 17 * Pekka Riikonen <[email protected]> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 #include <asm/local.h> 32 33 #include <linux/percpu.h> 34 #include <linux/rculist.h> 35 #include <linux/workqueue.h> 36 #include <linux/dynamic_queue_limits.h> 37 38 #include <net/net_namespace.h> 39 #ifdef CONFIG_DCB 40 #include <net/dcbnl.h> 41 #endif 42 #include <net/netprio_cgroup.h> 43 #include <net/xdp.h> 44 45 #include <linux/netdev_features.h> 46 #include <linux/neighbour.h> 47 #include <uapi/linux/netdevice.h> 48 #include <uapi/linux/if_bonding.h> 49 #include <uapi/linux/pkt_cls.h> 50 #include <linux/hashtable.h> 51 #include <linux/rbtree.h> 52 #include <net/net_trackers.h> 53 #include <net/net_debug.h> 54 55 struct netpoll_info; 56 struct device; 57 struct ethtool_ops; 58 struct phy_device; 59 struct dsa_port; 60 struct ip_tunnel_parm; 61 struct macsec_context; 62 struct macsec_ops; 63 struct netdev_name_node; 64 struct sd_flow_limit; 65 struct sfp_bus; 66 /* 802.11 specific */ 67 struct wireless_dev; 68 /* 802.15.4 specific */ 69 struct wpan_dev; 70 struct mpls_dev; 71 /* UDP Tunnel offloads */ 72 struct udp_tunnel_info; 73 struct udp_tunnel_nic_info; 74 struct udp_tunnel_nic; 75 struct bpf_prog; 76 struct xdp_buff; 77 78 void synchronize_net(void); 79 void netdev_set_default_ethtool_ops(struct net_device *dev, 80 const struct ethtool_ops *ops); 81 void netdev_sw_irq_coalesce_default_on(struct net_device *dev); 82 83 /* Backlog congestion levels */ 84 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 85 #define NET_RX_DROP 1 /* packet dropped */ 86 87 #define MAX_NEST_DEV 8 88 89 /* 90 * Transmit return codes: transmit return codes originate from three different 91 * namespaces: 92 * 93 * - qdisc return codes 94 * - driver transmit return codes 95 * - errno values 96 * 97 * Drivers are allowed to return any one of those in their hard_start_xmit() 98 * function. Real network devices commonly used with qdiscs should only return 99 * the driver transmit return codes though - when qdiscs are used, the actual 100 * transmission happens asynchronously, so the value is not propagated to 101 * higher layers. Virtual network devices transmit synchronously; in this case 102 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 103 * others are propagated to higher layers. 104 */ 105 106 /* qdisc ->enqueue() return codes. */ 107 #define NET_XMIT_SUCCESS 0x00 108 #define NET_XMIT_DROP 0x01 /* skb dropped */ 109 #define NET_XMIT_CN 0x02 /* congestion notification */ 110 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 111 112 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 113 * indicates that the device will soon be dropping packets, or already drops 114 * some packets of the same priority; prompting us to send less aggressively. */ 115 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 116 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 117 118 /* Driver transmit return codes */ 119 #define NETDEV_TX_MASK 0xf0 120 121 enum netdev_tx { 122 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 123 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 124 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 125 }; 126 typedef enum netdev_tx netdev_tx_t; 127 128 /* 129 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 130 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 131 */ 132 static inline bool dev_xmit_complete(int rc) 133 { 134 /* 135 * Positive cases with an skb consumed by a driver: 136 * - successful transmission (rc == NETDEV_TX_OK) 137 * - error while transmitting (rc < 0) 138 * - error while queueing to a different device (rc & NET_XMIT_MASK) 139 */ 140 if (likely(rc < NET_XMIT_MASK)) 141 return true; 142 143 return false; 144 } 145 146 /* 147 * Compute the worst-case header length according to the protocols 148 * used. 149 */ 150 151 #if defined(CONFIG_HYPERV_NET) 152 # define LL_MAX_HEADER 128 153 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 154 # if defined(CONFIG_MAC80211_MESH) 155 # define LL_MAX_HEADER 128 156 # else 157 # define LL_MAX_HEADER 96 158 # endif 159 #else 160 # define LL_MAX_HEADER 32 161 #endif 162 163 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 164 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 165 #define MAX_HEADER LL_MAX_HEADER 166 #else 167 #define MAX_HEADER (LL_MAX_HEADER + 48) 168 #endif 169 170 /* 171 * Old network device statistics. Fields are native words 172 * (unsigned long) so they can be read and written atomically. 173 */ 174 175 #define NET_DEV_STAT(FIELD) \ 176 union { \ 177 unsigned long FIELD; \ 178 atomic_long_t __##FIELD; \ 179 } 180 181 struct net_device_stats { 182 NET_DEV_STAT(rx_packets); 183 NET_DEV_STAT(tx_packets); 184 NET_DEV_STAT(rx_bytes); 185 NET_DEV_STAT(tx_bytes); 186 NET_DEV_STAT(rx_errors); 187 NET_DEV_STAT(tx_errors); 188 NET_DEV_STAT(rx_dropped); 189 NET_DEV_STAT(tx_dropped); 190 NET_DEV_STAT(multicast); 191 NET_DEV_STAT(collisions); 192 NET_DEV_STAT(rx_length_errors); 193 NET_DEV_STAT(rx_over_errors); 194 NET_DEV_STAT(rx_crc_errors); 195 NET_DEV_STAT(rx_frame_errors); 196 NET_DEV_STAT(rx_fifo_errors); 197 NET_DEV_STAT(rx_missed_errors); 198 NET_DEV_STAT(tx_aborted_errors); 199 NET_DEV_STAT(tx_carrier_errors); 200 NET_DEV_STAT(tx_fifo_errors); 201 NET_DEV_STAT(tx_heartbeat_errors); 202 NET_DEV_STAT(tx_window_errors); 203 NET_DEV_STAT(rx_compressed); 204 NET_DEV_STAT(tx_compressed); 205 }; 206 #undef NET_DEV_STAT 207 208 /* per-cpu stats, allocated on demand. 209 * Try to fit them in a single cache line, for dev_get_stats() sake. 210 */ 211 struct net_device_core_stats { 212 unsigned long rx_dropped; 213 unsigned long tx_dropped; 214 unsigned long rx_nohandler; 215 unsigned long rx_otherhost_dropped; 216 } __aligned(4 * sizeof(unsigned long)); 217 218 #include <linux/cache.h> 219 #include <linux/skbuff.h> 220 221 #ifdef CONFIG_RPS 222 #include <linux/static_key.h> 223 extern struct static_key_false rps_needed; 224 extern struct static_key_false rfs_needed; 225 #endif 226 227 struct neighbour; 228 struct neigh_parms; 229 struct sk_buff; 230 231 struct netdev_hw_addr { 232 struct list_head list; 233 struct rb_node node; 234 unsigned char addr[MAX_ADDR_LEN]; 235 unsigned char type; 236 #define NETDEV_HW_ADDR_T_LAN 1 237 #define NETDEV_HW_ADDR_T_SAN 2 238 #define NETDEV_HW_ADDR_T_UNICAST 3 239 #define NETDEV_HW_ADDR_T_MULTICAST 4 240 bool global_use; 241 int sync_cnt; 242 int refcount; 243 int synced; 244 struct rcu_head rcu_head; 245 }; 246 247 struct netdev_hw_addr_list { 248 struct list_head list; 249 int count; 250 251 /* Auxiliary tree for faster lookup on addition and deletion */ 252 struct rb_root tree; 253 }; 254 255 #define netdev_hw_addr_list_count(l) ((l)->count) 256 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 257 #define netdev_hw_addr_list_for_each(ha, l) \ 258 list_for_each_entry(ha, &(l)->list, list) 259 260 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 261 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 262 #define netdev_for_each_uc_addr(ha, dev) \ 263 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 264 #define netdev_for_each_synced_uc_addr(_ha, _dev) \ 265 netdev_for_each_uc_addr((_ha), (_dev)) \ 266 if ((_ha)->sync_cnt) 267 268 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 269 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 270 #define netdev_for_each_mc_addr(ha, dev) \ 271 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 272 #define netdev_for_each_synced_mc_addr(_ha, _dev) \ 273 netdev_for_each_mc_addr((_ha), (_dev)) \ 274 if ((_ha)->sync_cnt) 275 276 struct hh_cache { 277 unsigned int hh_len; 278 seqlock_t hh_lock; 279 280 /* cached hardware header; allow for machine alignment needs. */ 281 #define HH_DATA_MOD 16 282 #define HH_DATA_OFF(__len) \ 283 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 284 #define HH_DATA_ALIGN(__len) \ 285 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 286 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 287 }; 288 289 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 290 * Alternative is: 291 * dev->hard_header_len ? (dev->hard_header_len + 292 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 293 * 294 * We could use other alignment values, but we must maintain the 295 * relationship HH alignment <= LL alignment. 296 */ 297 #define LL_RESERVED_SPACE(dev) \ 298 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 299 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 300 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 301 302 struct header_ops { 303 int (*create) (struct sk_buff *skb, struct net_device *dev, 304 unsigned short type, const void *daddr, 305 const void *saddr, unsigned int len); 306 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 307 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 308 void (*cache_update)(struct hh_cache *hh, 309 const struct net_device *dev, 310 const unsigned char *haddr); 311 bool (*validate)(const char *ll_header, unsigned int len); 312 __be16 (*parse_protocol)(const struct sk_buff *skb); 313 }; 314 315 /* These flag bits are private to the generic network queueing 316 * layer; they may not be explicitly referenced by any other 317 * code. 318 */ 319 320 enum netdev_state_t { 321 __LINK_STATE_START, 322 __LINK_STATE_PRESENT, 323 __LINK_STATE_NOCARRIER, 324 __LINK_STATE_LINKWATCH_PENDING, 325 __LINK_STATE_DORMANT, 326 __LINK_STATE_TESTING, 327 }; 328 329 struct gro_list { 330 struct list_head list; 331 int count; 332 }; 333 334 /* 335 * size of gro hash buckets, must less than bit number of 336 * napi_struct::gro_bitmask 337 */ 338 #define GRO_HASH_BUCKETS 8 339 340 /* 341 * Structure for NAPI scheduling similar to tasklet but with weighting 342 */ 343 struct napi_struct { 344 /* The poll_list must only be managed by the entity which 345 * changes the state of the NAPI_STATE_SCHED bit. This means 346 * whoever atomically sets that bit can add this napi_struct 347 * to the per-CPU poll_list, and whoever clears that bit 348 * can remove from the list right before clearing the bit. 349 */ 350 struct list_head poll_list; 351 352 unsigned long state; 353 int weight; 354 int defer_hard_irqs_count; 355 unsigned long gro_bitmask; 356 int (*poll)(struct napi_struct *, int); 357 #ifdef CONFIG_NETPOLL 358 int poll_owner; 359 #endif 360 struct net_device *dev; 361 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 362 struct sk_buff *skb; 363 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 364 int rx_count; /* length of rx_list */ 365 struct hrtimer timer; 366 struct list_head dev_list; 367 struct hlist_node napi_hash_node; 368 unsigned int napi_id; 369 struct task_struct *thread; 370 }; 371 372 enum { 373 NAPI_STATE_SCHED, /* Poll is scheduled */ 374 NAPI_STATE_MISSED, /* reschedule a napi */ 375 NAPI_STATE_DISABLE, /* Disable pending */ 376 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 377 NAPI_STATE_LISTED, /* NAPI added to system lists */ 378 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 379 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 380 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 381 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 382 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 383 }; 384 385 enum { 386 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 387 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 388 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 389 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 390 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 391 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 392 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 393 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 394 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 395 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 396 }; 397 398 enum gro_result { 399 GRO_MERGED, 400 GRO_MERGED_FREE, 401 GRO_HELD, 402 GRO_NORMAL, 403 GRO_CONSUMED, 404 }; 405 typedef enum gro_result gro_result_t; 406 407 /* 408 * enum rx_handler_result - Possible return values for rx_handlers. 409 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 410 * further. 411 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 412 * case skb->dev was changed by rx_handler. 413 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 414 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 415 * 416 * rx_handlers are functions called from inside __netif_receive_skb(), to do 417 * special processing of the skb, prior to delivery to protocol handlers. 418 * 419 * Currently, a net_device can only have a single rx_handler registered. Trying 420 * to register a second rx_handler will return -EBUSY. 421 * 422 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 423 * To unregister a rx_handler on a net_device, use 424 * netdev_rx_handler_unregister(). 425 * 426 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 427 * do with the skb. 428 * 429 * If the rx_handler consumed the skb in some way, it should return 430 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 431 * the skb to be delivered in some other way. 432 * 433 * If the rx_handler changed skb->dev, to divert the skb to another 434 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 435 * new device will be called if it exists. 436 * 437 * If the rx_handler decides the skb should be ignored, it should return 438 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 439 * are registered on exact device (ptype->dev == skb->dev). 440 * 441 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 442 * delivered, it should return RX_HANDLER_PASS. 443 * 444 * A device without a registered rx_handler will behave as if rx_handler 445 * returned RX_HANDLER_PASS. 446 */ 447 448 enum rx_handler_result { 449 RX_HANDLER_CONSUMED, 450 RX_HANDLER_ANOTHER, 451 RX_HANDLER_EXACT, 452 RX_HANDLER_PASS, 453 }; 454 typedef enum rx_handler_result rx_handler_result_t; 455 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 456 457 void __napi_schedule(struct napi_struct *n); 458 void __napi_schedule_irqoff(struct napi_struct *n); 459 460 static inline bool napi_disable_pending(struct napi_struct *n) 461 { 462 return test_bit(NAPI_STATE_DISABLE, &n->state); 463 } 464 465 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 466 { 467 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 468 } 469 470 bool napi_schedule_prep(struct napi_struct *n); 471 472 /** 473 * napi_schedule - schedule NAPI poll 474 * @n: NAPI context 475 * 476 * Schedule NAPI poll routine to be called if it is not already 477 * running. 478 */ 479 static inline void napi_schedule(struct napi_struct *n) 480 { 481 if (napi_schedule_prep(n)) 482 __napi_schedule(n); 483 } 484 485 /** 486 * napi_schedule_irqoff - schedule NAPI poll 487 * @n: NAPI context 488 * 489 * Variant of napi_schedule(), assuming hard irqs are masked. 490 */ 491 static inline void napi_schedule_irqoff(struct napi_struct *n) 492 { 493 if (napi_schedule_prep(n)) 494 __napi_schedule_irqoff(n); 495 } 496 497 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 498 static inline bool napi_reschedule(struct napi_struct *napi) 499 { 500 if (napi_schedule_prep(napi)) { 501 __napi_schedule(napi); 502 return true; 503 } 504 return false; 505 } 506 507 bool napi_complete_done(struct napi_struct *n, int work_done); 508 /** 509 * napi_complete - NAPI processing complete 510 * @n: NAPI context 511 * 512 * Mark NAPI processing as complete. 513 * Consider using napi_complete_done() instead. 514 * Return false if device should avoid rearming interrupts. 515 */ 516 static inline bool napi_complete(struct napi_struct *n) 517 { 518 return napi_complete_done(n, 0); 519 } 520 521 int dev_set_threaded(struct net_device *dev, bool threaded); 522 523 /** 524 * napi_disable - prevent NAPI from scheduling 525 * @n: NAPI context 526 * 527 * Stop NAPI from being scheduled on this context. 528 * Waits till any outstanding processing completes. 529 */ 530 void napi_disable(struct napi_struct *n); 531 532 void napi_enable(struct napi_struct *n); 533 534 /** 535 * napi_synchronize - wait until NAPI is not running 536 * @n: NAPI context 537 * 538 * Wait until NAPI is done being scheduled on this context. 539 * Waits till any outstanding processing completes but 540 * does not disable future activations. 541 */ 542 static inline void napi_synchronize(const struct napi_struct *n) 543 { 544 if (IS_ENABLED(CONFIG_SMP)) 545 while (test_bit(NAPI_STATE_SCHED, &n->state)) 546 msleep(1); 547 else 548 barrier(); 549 } 550 551 /** 552 * napi_if_scheduled_mark_missed - if napi is running, set the 553 * NAPIF_STATE_MISSED 554 * @n: NAPI context 555 * 556 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 557 * NAPI is scheduled. 558 **/ 559 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 560 { 561 unsigned long val, new; 562 563 val = READ_ONCE(n->state); 564 do { 565 if (val & NAPIF_STATE_DISABLE) 566 return true; 567 568 if (!(val & NAPIF_STATE_SCHED)) 569 return false; 570 571 new = val | NAPIF_STATE_MISSED; 572 } while (!try_cmpxchg(&n->state, &val, new)); 573 574 return true; 575 } 576 577 enum netdev_queue_state_t { 578 __QUEUE_STATE_DRV_XOFF, 579 __QUEUE_STATE_STACK_XOFF, 580 __QUEUE_STATE_FROZEN, 581 }; 582 583 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 584 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 585 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 586 587 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 588 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 589 QUEUE_STATE_FROZEN) 590 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 591 QUEUE_STATE_FROZEN) 592 593 /* 594 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 595 * netif_tx_* functions below are used to manipulate this flag. The 596 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 597 * queue independently. The netif_xmit_*stopped functions below are called 598 * to check if the queue has been stopped by the driver or stack (either 599 * of the XOFF bits are set in the state). Drivers should not need to call 600 * netif_xmit*stopped functions, they should only be using netif_tx_*. 601 */ 602 603 struct netdev_queue { 604 /* 605 * read-mostly part 606 */ 607 struct net_device *dev; 608 netdevice_tracker dev_tracker; 609 610 struct Qdisc __rcu *qdisc; 611 struct Qdisc *qdisc_sleeping; 612 #ifdef CONFIG_SYSFS 613 struct kobject kobj; 614 #endif 615 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 616 int numa_node; 617 #endif 618 unsigned long tx_maxrate; 619 /* 620 * Number of TX timeouts for this queue 621 * (/sys/class/net/DEV/Q/trans_timeout) 622 */ 623 atomic_long_t trans_timeout; 624 625 /* Subordinate device that the queue has been assigned to */ 626 struct net_device *sb_dev; 627 #ifdef CONFIG_XDP_SOCKETS 628 struct xsk_buff_pool *pool; 629 #endif 630 /* 631 * write-mostly part 632 */ 633 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 634 int xmit_lock_owner; 635 /* 636 * Time (in jiffies) of last Tx 637 */ 638 unsigned long trans_start; 639 640 unsigned long state; 641 642 #ifdef CONFIG_BQL 643 struct dql dql; 644 #endif 645 } ____cacheline_aligned_in_smp; 646 647 extern int sysctl_fb_tunnels_only_for_init_net; 648 extern int sysctl_devconf_inherit_init_net; 649 650 /* 651 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 652 * == 1 : For initns only 653 * == 2 : For none. 654 */ 655 static inline bool net_has_fallback_tunnels(const struct net *net) 656 { 657 #if IS_ENABLED(CONFIG_SYSCTL) 658 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net); 659 660 return !fb_tunnels_only_for_init_net || 661 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1); 662 #else 663 return true; 664 #endif 665 } 666 667 static inline int net_inherit_devconf(void) 668 { 669 #if IS_ENABLED(CONFIG_SYSCTL) 670 return READ_ONCE(sysctl_devconf_inherit_init_net); 671 #else 672 return 0; 673 #endif 674 } 675 676 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 677 { 678 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 679 return q->numa_node; 680 #else 681 return NUMA_NO_NODE; 682 #endif 683 } 684 685 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 686 { 687 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 688 q->numa_node = node; 689 #endif 690 } 691 692 #ifdef CONFIG_RPS 693 /* 694 * This structure holds an RPS map which can be of variable length. The 695 * map is an array of CPUs. 696 */ 697 struct rps_map { 698 unsigned int len; 699 struct rcu_head rcu; 700 u16 cpus[]; 701 }; 702 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 703 704 /* 705 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 706 * tail pointer for that CPU's input queue at the time of last enqueue, and 707 * a hardware filter index. 708 */ 709 struct rps_dev_flow { 710 u16 cpu; 711 u16 filter; 712 unsigned int last_qtail; 713 }; 714 #define RPS_NO_FILTER 0xffff 715 716 /* 717 * The rps_dev_flow_table structure contains a table of flow mappings. 718 */ 719 struct rps_dev_flow_table { 720 unsigned int mask; 721 struct rcu_head rcu; 722 struct rps_dev_flow flows[]; 723 }; 724 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 725 ((_num) * sizeof(struct rps_dev_flow))) 726 727 /* 728 * The rps_sock_flow_table contains mappings of flows to the last CPU 729 * on which they were processed by the application (set in recvmsg). 730 * Each entry is a 32bit value. Upper part is the high-order bits 731 * of flow hash, lower part is CPU number. 732 * rps_cpu_mask is used to partition the space, depending on number of 733 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 734 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 735 * meaning we use 32-6=26 bits for the hash. 736 */ 737 struct rps_sock_flow_table { 738 u32 mask; 739 740 u32 ents[] ____cacheline_aligned_in_smp; 741 }; 742 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 743 744 #define RPS_NO_CPU 0xffff 745 746 extern u32 rps_cpu_mask; 747 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 748 749 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 750 u32 hash) 751 { 752 if (table && hash) { 753 unsigned int index = hash & table->mask; 754 u32 val = hash & ~rps_cpu_mask; 755 756 /* We only give a hint, preemption can change CPU under us */ 757 val |= raw_smp_processor_id(); 758 759 if (table->ents[index] != val) 760 table->ents[index] = val; 761 } 762 } 763 764 #ifdef CONFIG_RFS_ACCEL 765 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 766 u16 filter_id); 767 #endif 768 #endif /* CONFIG_RPS */ 769 770 /* This structure contains an instance of an RX queue. */ 771 struct netdev_rx_queue { 772 struct xdp_rxq_info xdp_rxq; 773 #ifdef CONFIG_RPS 774 struct rps_map __rcu *rps_map; 775 struct rps_dev_flow_table __rcu *rps_flow_table; 776 #endif 777 struct kobject kobj; 778 struct net_device *dev; 779 netdevice_tracker dev_tracker; 780 781 #ifdef CONFIG_XDP_SOCKETS 782 struct xsk_buff_pool *pool; 783 #endif 784 } ____cacheline_aligned_in_smp; 785 786 /* 787 * RX queue sysfs structures and functions. 788 */ 789 struct rx_queue_attribute { 790 struct attribute attr; 791 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 792 ssize_t (*store)(struct netdev_rx_queue *queue, 793 const char *buf, size_t len); 794 }; 795 796 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 797 enum xps_map_type { 798 XPS_CPUS = 0, 799 XPS_RXQS, 800 XPS_MAPS_MAX, 801 }; 802 803 #ifdef CONFIG_XPS 804 /* 805 * This structure holds an XPS map which can be of variable length. The 806 * map is an array of queues. 807 */ 808 struct xps_map { 809 unsigned int len; 810 unsigned int alloc_len; 811 struct rcu_head rcu; 812 u16 queues[]; 813 }; 814 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 815 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 816 - sizeof(struct xps_map)) / sizeof(u16)) 817 818 /* 819 * This structure holds all XPS maps for device. Maps are indexed by CPU. 820 * 821 * We keep track of the number of cpus/rxqs used when the struct is allocated, 822 * in nr_ids. This will help not accessing out-of-bound memory. 823 * 824 * We keep track of the number of traffic classes used when the struct is 825 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 826 * not crossing its upper bound, as the original dev->num_tc can be updated in 827 * the meantime. 828 */ 829 struct xps_dev_maps { 830 struct rcu_head rcu; 831 unsigned int nr_ids; 832 s16 num_tc; 833 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 834 }; 835 836 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 837 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 838 839 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 840 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 841 842 #endif /* CONFIG_XPS */ 843 844 #define TC_MAX_QUEUE 16 845 #define TC_BITMASK 15 846 /* HW offloaded queuing disciplines txq count and offset maps */ 847 struct netdev_tc_txq { 848 u16 count; 849 u16 offset; 850 }; 851 852 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 853 /* 854 * This structure is to hold information about the device 855 * configured to run FCoE protocol stack. 856 */ 857 struct netdev_fcoe_hbainfo { 858 char manufacturer[64]; 859 char serial_number[64]; 860 char hardware_version[64]; 861 char driver_version[64]; 862 char optionrom_version[64]; 863 char firmware_version[64]; 864 char model[256]; 865 char model_description[256]; 866 }; 867 #endif 868 869 #define MAX_PHYS_ITEM_ID_LEN 32 870 871 /* This structure holds a unique identifier to identify some 872 * physical item (port for example) used by a netdevice. 873 */ 874 struct netdev_phys_item_id { 875 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 876 unsigned char id_len; 877 }; 878 879 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 880 struct netdev_phys_item_id *b) 881 { 882 return a->id_len == b->id_len && 883 memcmp(a->id, b->id, a->id_len) == 0; 884 } 885 886 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 887 struct sk_buff *skb, 888 struct net_device *sb_dev); 889 890 enum net_device_path_type { 891 DEV_PATH_ETHERNET = 0, 892 DEV_PATH_VLAN, 893 DEV_PATH_BRIDGE, 894 DEV_PATH_PPPOE, 895 DEV_PATH_DSA, 896 DEV_PATH_MTK_WDMA, 897 }; 898 899 struct net_device_path { 900 enum net_device_path_type type; 901 const struct net_device *dev; 902 union { 903 struct { 904 u16 id; 905 __be16 proto; 906 u8 h_dest[ETH_ALEN]; 907 } encap; 908 struct { 909 enum { 910 DEV_PATH_BR_VLAN_KEEP, 911 DEV_PATH_BR_VLAN_TAG, 912 DEV_PATH_BR_VLAN_UNTAG, 913 DEV_PATH_BR_VLAN_UNTAG_HW, 914 } vlan_mode; 915 u16 vlan_id; 916 __be16 vlan_proto; 917 } bridge; 918 struct { 919 int port; 920 u16 proto; 921 } dsa; 922 struct { 923 u8 wdma_idx; 924 u8 queue; 925 u16 wcid; 926 u8 bss; 927 } mtk_wdma; 928 }; 929 }; 930 931 #define NET_DEVICE_PATH_STACK_MAX 5 932 #define NET_DEVICE_PATH_VLAN_MAX 2 933 934 struct net_device_path_stack { 935 int num_paths; 936 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 937 }; 938 939 struct net_device_path_ctx { 940 const struct net_device *dev; 941 u8 daddr[ETH_ALEN]; 942 943 int num_vlans; 944 struct { 945 u16 id; 946 __be16 proto; 947 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 948 }; 949 950 enum tc_setup_type { 951 TC_QUERY_CAPS, 952 TC_SETUP_QDISC_MQPRIO, 953 TC_SETUP_CLSU32, 954 TC_SETUP_CLSFLOWER, 955 TC_SETUP_CLSMATCHALL, 956 TC_SETUP_CLSBPF, 957 TC_SETUP_BLOCK, 958 TC_SETUP_QDISC_CBS, 959 TC_SETUP_QDISC_RED, 960 TC_SETUP_QDISC_PRIO, 961 TC_SETUP_QDISC_MQ, 962 TC_SETUP_QDISC_ETF, 963 TC_SETUP_ROOT_QDISC, 964 TC_SETUP_QDISC_GRED, 965 TC_SETUP_QDISC_TAPRIO, 966 TC_SETUP_FT, 967 TC_SETUP_QDISC_ETS, 968 TC_SETUP_QDISC_TBF, 969 TC_SETUP_QDISC_FIFO, 970 TC_SETUP_QDISC_HTB, 971 TC_SETUP_ACT, 972 }; 973 974 /* These structures hold the attributes of bpf state that are being passed 975 * to the netdevice through the bpf op. 976 */ 977 enum bpf_netdev_command { 978 /* Set or clear a bpf program used in the earliest stages of packet 979 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 980 * is responsible for calling bpf_prog_put on any old progs that are 981 * stored. In case of error, the callee need not release the new prog 982 * reference, but on success it takes ownership and must bpf_prog_put 983 * when it is no longer used. 984 */ 985 XDP_SETUP_PROG, 986 XDP_SETUP_PROG_HW, 987 /* BPF program for offload callbacks, invoked at program load time. */ 988 BPF_OFFLOAD_MAP_ALLOC, 989 BPF_OFFLOAD_MAP_FREE, 990 XDP_SETUP_XSK_POOL, 991 }; 992 993 struct bpf_prog_offload_ops; 994 struct netlink_ext_ack; 995 struct xdp_umem; 996 struct xdp_dev_bulk_queue; 997 struct bpf_xdp_link; 998 999 enum bpf_xdp_mode { 1000 XDP_MODE_SKB = 0, 1001 XDP_MODE_DRV = 1, 1002 XDP_MODE_HW = 2, 1003 __MAX_XDP_MODE 1004 }; 1005 1006 struct bpf_xdp_entity { 1007 struct bpf_prog *prog; 1008 struct bpf_xdp_link *link; 1009 }; 1010 1011 struct netdev_bpf { 1012 enum bpf_netdev_command command; 1013 union { 1014 /* XDP_SETUP_PROG */ 1015 struct { 1016 u32 flags; 1017 struct bpf_prog *prog; 1018 struct netlink_ext_ack *extack; 1019 }; 1020 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 1021 struct { 1022 struct bpf_offloaded_map *offmap; 1023 }; 1024 /* XDP_SETUP_XSK_POOL */ 1025 struct { 1026 struct xsk_buff_pool *pool; 1027 u16 queue_id; 1028 } xsk; 1029 }; 1030 }; 1031 1032 /* Flags for ndo_xsk_wakeup. */ 1033 #define XDP_WAKEUP_RX (1 << 0) 1034 #define XDP_WAKEUP_TX (1 << 1) 1035 1036 #ifdef CONFIG_XFRM_OFFLOAD 1037 struct xfrmdev_ops { 1038 int (*xdo_dev_state_add) (struct xfrm_state *x); 1039 void (*xdo_dev_state_delete) (struct xfrm_state *x); 1040 void (*xdo_dev_state_free) (struct xfrm_state *x); 1041 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 1042 struct xfrm_state *x); 1043 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1044 void (*xdo_dev_state_update_curlft) (struct xfrm_state *x); 1045 int (*xdo_dev_policy_add) (struct xfrm_policy *x); 1046 void (*xdo_dev_policy_delete) (struct xfrm_policy *x); 1047 void (*xdo_dev_policy_free) (struct xfrm_policy *x); 1048 }; 1049 #endif 1050 1051 struct dev_ifalias { 1052 struct rcu_head rcuhead; 1053 char ifalias[]; 1054 }; 1055 1056 struct devlink; 1057 struct tlsdev_ops; 1058 1059 struct netdev_net_notifier { 1060 struct list_head list; 1061 struct notifier_block *nb; 1062 }; 1063 1064 /* 1065 * This structure defines the management hooks for network devices. 1066 * The following hooks can be defined; unless noted otherwise, they are 1067 * optional and can be filled with a null pointer. 1068 * 1069 * int (*ndo_init)(struct net_device *dev); 1070 * This function is called once when a network device is registered. 1071 * The network device can use this for any late stage initialization 1072 * or semantic validation. It can fail with an error code which will 1073 * be propagated back to register_netdev. 1074 * 1075 * void (*ndo_uninit)(struct net_device *dev); 1076 * This function is called when device is unregistered or when registration 1077 * fails. It is not called if init fails. 1078 * 1079 * int (*ndo_open)(struct net_device *dev); 1080 * This function is called when a network device transitions to the up 1081 * state. 1082 * 1083 * int (*ndo_stop)(struct net_device *dev); 1084 * This function is called when a network device transitions to the down 1085 * state. 1086 * 1087 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1088 * struct net_device *dev); 1089 * Called when a packet needs to be transmitted. 1090 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1091 * the queue before that can happen; it's for obsolete devices and weird 1092 * corner cases, but the stack really does a non-trivial amount 1093 * of useless work if you return NETDEV_TX_BUSY. 1094 * Required; cannot be NULL. 1095 * 1096 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1097 * struct net_device *dev 1098 * netdev_features_t features); 1099 * Called by core transmit path to determine if device is capable of 1100 * performing offload operations on a given packet. This is to give 1101 * the device an opportunity to implement any restrictions that cannot 1102 * be otherwise expressed by feature flags. The check is called with 1103 * the set of features that the stack has calculated and it returns 1104 * those the driver believes to be appropriate. 1105 * 1106 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1107 * struct net_device *sb_dev); 1108 * Called to decide which queue to use when device supports multiple 1109 * transmit queues. 1110 * 1111 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1112 * This function is called to allow device receiver to make 1113 * changes to configuration when multicast or promiscuous is enabled. 1114 * 1115 * void (*ndo_set_rx_mode)(struct net_device *dev); 1116 * This function is called device changes address list filtering. 1117 * If driver handles unicast address filtering, it should set 1118 * IFF_UNICAST_FLT in its priv_flags. 1119 * 1120 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1121 * This function is called when the Media Access Control address 1122 * needs to be changed. If this interface is not defined, the 1123 * MAC address can not be changed. 1124 * 1125 * int (*ndo_validate_addr)(struct net_device *dev); 1126 * Test if Media Access Control address is valid for the device. 1127 * 1128 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1129 * Old-style ioctl entry point. This is used internally by the 1130 * appletalk and ieee802154 subsystems but is no longer called by 1131 * the device ioctl handler. 1132 * 1133 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1134 * Used by the bonding driver for its device specific ioctls: 1135 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1136 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1137 * 1138 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1139 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1140 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1141 * 1142 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1143 * Used to set network devices bus interface parameters. This interface 1144 * is retained for legacy reasons; new devices should use the bus 1145 * interface (PCI) for low level management. 1146 * 1147 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1148 * Called when a user wants to change the Maximum Transfer Unit 1149 * of a device. 1150 * 1151 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1152 * Callback used when the transmitter has not made any progress 1153 * for dev->watchdog ticks. 1154 * 1155 * void (*ndo_get_stats64)(struct net_device *dev, 1156 * struct rtnl_link_stats64 *storage); 1157 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1158 * Called when a user wants to get the network device usage 1159 * statistics. Drivers must do one of the following: 1160 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1161 * rtnl_link_stats64 structure passed by the caller. 1162 * 2. Define @ndo_get_stats to update a net_device_stats structure 1163 * (which should normally be dev->stats) and return a pointer to 1164 * it. The structure may be changed asynchronously only if each 1165 * field is written atomically. 1166 * 3. Update dev->stats asynchronously and atomically, and define 1167 * neither operation. 1168 * 1169 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1170 * Return true if this device supports offload stats of this attr_id. 1171 * 1172 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1173 * void *attr_data) 1174 * Get statistics for offload operations by attr_id. Write it into the 1175 * attr_data pointer. 1176 * 1177 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1178 * If device supports VLAN filtering this function is called when a 1179 * VLAN id is registered. 1180 * 1181 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1182 * If device supports VLAN filtering this function is called when a 1183 * VLAN id is unregistered. 1184 * 1185 * void (*ndo_poll_controller)(struct net_device *dev); 1186 * 1187 * SR-IOV management functions. 1188 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1189 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1190 * u8 qos, __be16 proto); 1191 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1192 * int max_tx_rate); 1193 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1194 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1195 * int (*ndo_get_vf_config)(struct net_device *dev, 1196 * int vf, struct ifla_vf_info *ivf); 1197 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1198 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1199 * struct nlattr *port[]); 1200 * 1201 * Enable or disable the VF ability to query its RSS Redirection Table and 1202 * Hash Key. This is needed since on some devices VF share this information 1203 * with PF and querying it may introduce a theoretical security risk. 1204 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1205 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1206 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1207 * void *type_data); 1208 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1209 * This is always called from the stack with the rtnl lock held and netif 1210 * tx queues stopped. This allows the netdevice to perform queue 1211 * management safely. 1212 * 1213 * Fiber Channel over Ethernet (FCoE) offload functions. 1214 * int (*ndo_fcoe_enable)(struct net_device *dev); 1215 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1216 * so the underlying device can perform whatever needed configuration or 1217 * initialization to support acceleration of FCoE traffic. 1218 * 1219 * int (*ndo_fcoe_disable)(struct net_device *dev); 1220 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1221 * so the underlying device can perform whatever needed clean-ups to 1222 * stop supporting acceleration of FCoE traffic. 1223 * 1224 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1225 * struct scatterlist *sgl, unsigned int sgc); 1226 * Called when the FCoE Initiator wants to initialize an I/O that 1227 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1228 * perform necessary setup and returns 1 to indicate the device is set up 1229 * successfully to perform DDP on this I/O, otherwise this returns 0. 1230 * 1231 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1232 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1233 * indicated by the FC exchange id 'xid', so the underlying device can 1234 * clean up and reuse resources for later DDP requests. 1235 * 1236 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1237 * struct scatterlist *sgl, unsigned int sgc); 1238 * Called when the FCoE Target wants to initialize an I/O that 1239 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1240 * perform necessary setup and returns 1 to indicate the device is set up 1241 * successfully to perform DDP on this I/O, otherwise this returns 0. 1242 * 1243 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1244 * struct netdev_fcoe_hbainfo *hbainfo); 1245 * Called when the FCoE Protocol stack wants information on the underlying 1246 * device. This information is utilized by the FCoE protocol stack to 1247 * register attributes with Fiber Channel management service as per the 1248 * FC-GS Fabric Device Management Information(FDMI) specification. 1249 * 1250 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1251 * Called when the underlying device wants to override default World Wide 1252 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1253 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1254 * protocol stack to use. 1255 * 1256 * RFS acceleration. 1257 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1258 * u16 rxq_index, u32 flow_id); 1259 * Set hardware filter for RFS. rxq_index is the target queue index; 1260 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1261 * Return the filter ID on success, or a negative error code. 1262 * 1263 * Slave management functions (for bridge, bonding, etc). 1264 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1265 * Called to make another netdev an underling. 1266 * 1267 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1268 * Called to release previously enslaved netdev. 1269 * 1270 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1271 * struct sk_buff *skb, 1272 * bool all_slaves); 1273 * Get the xmit slave of master device. If all_slaves is true, function 1274 * assume all the slaves can transmit. 1275 * 1276 * Feature/offload setting functions. 1277 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1278 * netdev_features_t features); 1279 * Adjusts the requested feature flags according to device-specific 1280 * constraints, and returns the resulting flags. Must not modify 1281 * the device state. 1282 * 1283 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1284 * Called to update device configuration to new features. Passed 1285 * feature set might be less than what was returned by ndo_fix_features()). 1286 * Must return >0 or -errno if it changed dev->features itself. 1287 * 1288 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1289 * struct net_device *dev, 1290 * const unsigned char *addr, u16 vid, u16 flags, 1291 * struct netlink_ext_ack *extack); 1292 * Adds an FDB entry to dev for addr. 1293 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1294 * struct net_device *dev, 1295 * const unsigned char *addr, u16 vid) 1296 * Deletes the FDB entry from dev coresponding to addr. 1297 * int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, struct nlattr *tb[], 1298 * struct net_device *dev, 1299 * u16 vid, 1300 * struct netlink_ext_ack *extack); 1301 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1302 * struct net_device *dev, struct net_device *filter_dev, 1303 * int *idx) 1304 * Used to add FDB entries to dump requests. Implementers should add 1305 * entries to skb and update idx with the number of entries. 1306 * 1307 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1308 * u16 flags, struct netlink_ext_ack *extack) 1309 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1310 * struct net_device *dev, u32 filter_mask, 1311 * int nlflags) 1312 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1313 * u16 flags); 1314 * 1315 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1316 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1317 * which do not represent real hardware may define this to allow their 1318 * userspace components to manage their virtual carrier state. Devices 1319 * that determine carrier state from physical hardware properties (eg 1320 * network cables) or protocol-dependent mechanisms (eg 1321 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1322 * 1323 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1324 * struct netdev_phys_item_id *ppid); 1325 * Called to get ID of physical port of this device. If driver does 1326 * not implement this, it is assumed that the hw is not able to have 1327 * multiple net devices on single physical port. 1328 * 1329 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1330 * struct netdev_phys_item_id *ppid) 1331 * Called to get the parent ID of the physical port of this device. 1332 * 1333 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1334 * struct net_device *dev) 1335 * Called by upper layer devices to accelerate switching or other 1336 * station functionality into hardware. 'pdev is the lowerdev 1337 * to use for the offload and 'dev' is the net device that will 1338 * back the offload. Returns a pointer to the private structure 1339 * the upper layer will maintain. 1340 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1341 * Called by upper layer device to delete the station created 1342 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1343 * the station and priv is the structure returned by the add 1344 * operation. 1345 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1346 * int queue_index, u32 maxrate); 1347 * Called when a user wants to set a max-rate limitation of specific 1348 * TX queue. 1349 * int (*ndo_get_iflink)(const struct net_device *dev); 1350 * Called to get the iflink value of this device. 1351 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1352 * This function is used to get egress tunnel information for given skb. 1353 * This is useful for retrieving outer tunnel header parameters while 1354 * sampling packet. 1355 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1356 * This function is used to specify the headroom that the skb must 1357 * consider when allocation skb during packet reception. Setting 1358 * appropriate rx headroom value allows avoiding skb head copy on 1359 * forward. Setting a negative value resets the rx headroom to the 1360 * default value. 1361 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1362 * This function is used to set or query state related to XDP on the 1363 * netdevice and manage BPF offload. See definition of 1364 * enum bpf_netdev_command for details. 1365 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1366 * u32 flags); 1367 * This function is used to submit @n XDP packets for transmit on a 1368 * netdevice. Returns number of frames successfully transmitted, frames 1369 * that got dropped are freed/returned via xdp_return_frame(). 1370 * Returns negative number, means general error invoking ndo, meaning 1371 * no frames were xmit'ed and core-caller will free all frames. 1372 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1373 * struct xdp_buff *xdp); 1374 * Get the xmit slave of master device based on the xdp_buff. 1375 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1376 * This function is used to wake up the softirq, ksoftirqd or kthread 1377 * responsible for sending and/or receiving packets on a specific 1378 * queue id bound to an AF_XDP socket. The flags field specifies if 1379 * only RX, only Tx, or both should be woken up using the flags 1380 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1381 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1382 * int cmd); 1383 * Add, change, delete or get information on an IPv4 tunnel. 1384 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1385 * If a device is paired with a peer device, return the peer instance. 1386 * The caller must be under RCU read context. 1387 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1388 * Get the forwarding path to reach the real device from the HW destination address 1389 * ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1390 * const struct skb_shared_hwtstamps *hwtstamps, 1391 * bool cycles); 1392 * Get hardware timestamp based on normal/adjustable time or free running 1393 * cycle counter. This function is required if physical clock supports a 1394 * free running cycle counter. 1395 */ 1396 struct net_device_ops { 1397 int (*ndo_init)(struct net_device *dev); 1398 void (*ndo_uninit)(struct net_device *dev); 1399 int (*ndo_open)(struct net_device *dev); 1400 int (*ndo_stop)(struct net_device *dev); 1401 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1402 struct net_device *dev); 1403 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1404 struct net_device *dev, 1405 netdev_features_t features); 1406 u16 (*ndo_select_queue)(struct net_device *dev, 1407 struct sk_buff *skb, 1408 struct net_device *sb_dev); 1409 void (*ndo_change_rx_flags)(struct net_device *dev, 1410 int flags); 1411 void (*ndo_set_rx_mode)(struct net_device *dev); 1412 int (*ndo_set_mac_address)(struct net_device *dev, 1413 void *addr); 1414 int (*ndo_validate_addr)(struct net_device *dev); 1415 int (*ndo_do_ioctl)(struct net_device *dev, 1416 struct ifreq *ifr, int cmd); 1417 int (*ndo_eth_ioctl)(struct net_device *dev, 1418 struct ifreq *ifr, int cmd); 1419 int (*ndo_siocbond)(struct net_device *dev, 1420 struct ifreq *ifr, int cmd); 1421 int (*ndo_siocwandev)(struct net_device *dev, 1422 struct if_settings *ifs); 1423 int (*ndo_siocdevprivate)(struct net_device *dev, 1424 struct ifreq *ifr, 1425 void __user *data, int cmd); 1426 int (*ndo_set_config)(struct net_device *dev, 1427 struct ifmap *map); 1428 int (*ndo_change_mtu)(struct net_device *dev, 1429 int new_mtu); 1430 int (*ndo_neigh_setup)(struct net_device *dev, 1431 struct neigh_parms *); 1432 void (*ndo_tx_timeout) (struct net_device *dev, 1433 unsigned int txqueue); 1434 1435 void (*ndo_get_stats64)(struct net_device *dev, 1436 struct rtnl_link_stats64 *storage); 1437 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1438 int (*ndo_get_offload_stats)(int attr_id, 1439 const struct net_device *dev, 1440 void *attr_data); 1441 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1442 1443 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1444 __be16 proto, u16 vid); 1445 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1446 __be16 proto, u16 vid); 1447 #ifdef CONFIG_NET_POLL_CONTROLLER 1448 void (*ndo_poll_controller)(struct net_device *dev); 1449 int (*ndo_netpoll_setup)(struct net_device *dev, 1450 struct netpoll_info *info); 1451 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1452 #endif 1453 int (*ndo_set_vf_mac)(struct net_device *dev, 1454 int queue, u8 *mac); 1455 int (*ndo_set_vf_vlan)(struct net_device *dev, 1456 int queue, u16 vlan, 1457 u8 qos, __be16 proto); 1458 int (*ndo_set_vf_rate)(struct net_device *dev, 1459 int vf, int min_tx_rate, 1460 int max_tx_rate); 1461 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1462 int vf, bool setting); 1463 int (*ndo_set_vf_trust)(struct net_device *dev, 1464 int vf, bool setting); 1465 int (*ndo_get_vf_config)(struct net_device *dev, 1466 int vf, 1467 struct ifla_vf_info *ivf); 1468 int (*ndo_set_vf_link_state)(struct net_device *dev, 1469 int vf, int link_state); 1470 int (*ndo_get_vf_stats)(struct net_device *dev, 1471 int vf, 1472 struct ifla_vf_stats 1473 *vf_stats); 1474 int (*ndo_set_vf_port)(struct net_device *dev, 1475 int vf, 1476 struct nlattr *port[]); 1477 int (*ndo_get_vf_port)(struct net_device *dev, 1478 int vf, struct sk_buff *skb); 1479 int (*ndo_get_vf_guid)(struct net_device *dev, 1480 int vf, 1481 struct ifla_vf_guid *node_guid, 1482 struct ifla_vf_guid *port_guid); 1483 int (*ndo_set_vf_guid)(struct net_device *dev, 1484 int vf, u64 guid, 1485 int guid_type); 1486 int (*ndo_set_vf_rss_query_en)( 1487 struct net_device *dev, 1488 int vf, bool setting); 1489 int (*ndo_setup_tc)(struct net_device *dev, 1490 enum tc_setup_type type, 1491 void *type_data); 1492 #if IS_ENABLED(CONFIG_FCOE) 1493 int (*ndo_fcoe_enable)(struct net_device *dev); 1494 int (*ndo_fcoe_disable)(struct net_device *dev); 1495 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1496 u16 xid, 1497 struct scatterlist *sgl, 1498 unsigned int sgc); 1499 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1500 u16 xid); 1501 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1502 u16 xid, 1503 struct scatterlist *sgl, 1504 unsigned int sgc); 1505 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1506 struct netdev_fcoe_hbainfo *hbainfo); 1507 #endif 1508 1509 #if IS_ENABLED(CONFIG_LIBFCOE) 1510 #define NETDEV_FCOE_WWNN 0 1511 #define NETDEV_FCOE_WWPN 1 1512 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1513 u64 *wwn, int type); 1514 #endif 1515 1516 #ifdef CONFIG_RFS_ACCEL 1517 int (*ndo_rx_flow_steer)(struct net_device *dev, 1518 const struct sk_buff *skb, 1519 u16 rxq_index, 1520 u32 flow_id); 1521 #endif 1522 int (*ndo_add_slave)(struct net_device *dev, 1523 struct net_device *slave_dev, 1524 struct netlink_ext_ack *extack); 1525 int (*ndo_del_slave)(struct net_device *dev, 1526 struct net_device *slave_dev); 1527 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1528 struct sk_buff *skb, 1529 bool all_slaves); 1530 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1531 struct sock *sk); 1532 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1533 netdev_features_t features); 1534 int (*ndo_set_features)(struct net_device *dev, 1535 netdev_features_t features); 1536 int (*ndo_neigh_construct)(struct net_device *dev, 1537 struct neighbour *n); 1538 void (*ndo_neigh_destroy)(struct net_device *dev, 1539 struct neighbour *n); 1540 1541 int (*ndo_fdb_add)(struct ndmsg *ndm, 1542 struct nlattr *tb[], 1543 struct net_device *dev, 1544 const unsigned char *addr, 1545 u16 vid, 1546 u16 flags, 1547 struct netlink_ext_ack *extack); 1548 int (*ndo_fdb_del)(struct ndmsg *ndm, 1549 struct nlattr *tb[], 1550 struct net_device *dev, 1551 const unsigned char *addr, 1552 u16 vid, struct netlink_ext_ack *extack); 1553 int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, 1554 struct nlattr *tb[], 1555 struct net_device *dev, 1556 u16 vid, 1557 struct netlink_ext_ack *extack); 1558 int (*ndo_fdb_dump)(struct sk_buff *skb, 1559 struct netlink_callback *cb, 1560 struct net_device *dev, 1561 struct net_device *filter_dev, 1562 int *idx); 1563 int (*ndo_fdb_get)(struct sk_buff *skb, 1564 struct nlattr *tb[], 1565 struct net_device *dev, 1566 const unsigned char *addr, 1567 u16 vid, u32 portid, u32 seq, 1568 struct netlink_ext_ack *extack); 1569 int (*ndo_bridge_setlink)(struct net_device *dev, 1570 struct nlmsghdr *nlh, 1571 u16 flags, 1572 struct netlink_ext_ack *extack); 1573 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1574 u32 pid, u32 seq, 1575 struct net_device *dev, 1576 u32 filter_mask, 1577 int nlflags); 1578 int (*ndo_bridge_dellink)(struct net_device *dev, 1579 struct nlmsghdr *nlh, 1580 u16 flags); 1581 int (*ndo_change_carrier)(struct net_device *dev, 1582 bool new_carrier); 1583 int (*ndo_get_phys_port_id)(struct net_device *dev, 1584 struct netdev_phys_item_id *ppid); 1585 int (*ndo_get_port_parent_id)(struct net_device *dev, 1586 struct netdev_phys_item_id *ppid); 1587 int (*ndo_get_phys_port_name)(struct net_device *dev, 1588 char *name, size_t len); 1589 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1590 struct net_device *dev); 1591 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1592 void *priv); 1593 1594 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1595 int queue_index, 1596 u32 maxrate); 1597 int (*ndo_get_iflink)(const struct net_device *dev); 1598 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1599 struct sk_buff *skb); 1600 void (*ndo_set_rx_headroom)(struct net_device *dev, 1601 int needed_headroom); 1602 int (*ndo_bpf)(struct net_device *dev, 1603 struct netdev_bpf *bpf); 1604 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1605 struct xdp_frame **xdp, 1606 u32 flags); 1607 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1608 struct xdp_buff *xdp); 1609 int (*ndo_xsk_wakeup)(struct net_device *dev, 1610 u32 queue_id, u32 flags); 1611 int (*ndo_tunnel_ctl)(struct net_device *dev, 1612 struct ip_tunnel_parm *p, int cmd); 1613 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1614 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1615 struct net_device_path *path); 1616 ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1617 const struct skb_shared_hwtstamps *hwtstamps, 1618 bool cycles); 1619 }; 1620 1621 /** 1622 * enum netdev_priv_flags - &struct net_device priv_flags 1623 * 1624 * These are the &struct net_device, they are only set internally 1625 * by drivers and used in the kernel. These flags are invisible to 1626 * userspace; this means that the order of these flags can change 1627 * during any kernel release. 1628 * 1629 * You should have a pretty good reason to be extending these flags. 1630 * 1631 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1632 * @IFF_EBRIDGE: Ethernet bridging device 1633 * @IFF_BONDING: bonding master or slave 1634 * @IFF_ISATAP: ISATAP interface (RFC4214) 1635 * @IFF_WAN_HDLC: WAN HDLC device 1636 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1637 * release skb->dst 1638 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1639 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1640 * @IFF_MACVLAN_PORT: device used as macvlan port 1641 * @IFF_BRIDGE_PORT: device used as bridge port 1642 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1643 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1644 * @IFF_UNICAST_FLT: Supports unicast filtering 1645 * @IFF_TEAM_PORT: device used as team port 1646 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1647 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1648 * change when it's running 1649 * @IFF_MACVLAN: Macvlan device 1650 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1651 * underlying stacked devices 1652 * @IFF_L3MDEV_MASTER: device is an L3 master device 1653 * @IFF_NO_QUEUE: device can run without qdisc attached 1654 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1655 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1656 * @IFF_TEAM: device is a team device 1657 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1658 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1659 * entity (i.e. the master device for bridged veth) 1660 * @IFF_MACSEC: device is a MACsec device 1661 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1662 * @IFF_FAILOVER: device is a failover master device 1663 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1664 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1665 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf 1666 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1667 * skb_headlen(skb) == 0 (data starts from frag0) 1668 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN 1669 */ 1670 enum netdev_priv_flags { 1671 IFF_802_1Q_VLAN = 1<<0, 1672 IFF_EBRIDGE = 1<<1, 1673 IFF_BONDING = 1<<2, 1674 IFF_ISATAP = 1<<3, 1675 IFF_WAN_HDLC = 1<<4, 1676 IFF_XMIT_DST_RELEASE = 1<<5, 1677 IFF_DONT_BRIDGE = 1<<6, 1678 IFF_DISABLE_NETPOLL = 1<<7, 1679 IFF_MACVLAN_PORT = 1<<8, 1680 IFF_BRIDGE_PORT = 1<<9, 1681 IFF_OVS_DATAPATH = 1<<10, 1682 IFF_TX_SKB_SHARING = 1<<11, 1683 IFF_UNICAST_FLT = 1<<12, 1684 IFF_TEAM_PORT = 1<<13, 1685 IFF_SUPP_NOFCS = 1<<14, 1686 IFF_LIVE_ADDR_CHANGE = 1<<15, 1687 IFF_MACVLAN = 1<<16, 1688 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1689 IFF_L3MDEV_MASTER = 1<<18, 1690 IFF_NO_QUEUE = 1<<19, 1691 IFF_OPENVSWITCH = 1<<20, 1692 IFF_L3MDEV_SLAVE = 1<<21, 1693 IFF_TEAM = 1<<22, 1694 IFF_RXFH_CONFIGURED = 1<<23, 1695 IFF_PHONY_HEADROOM = 1<<24, 1696 IFF_MACSEC = 1<<25, 1697 IFF_NO_RX_HANDLER = 1<<26, 1698 IFF_FAILOVER = 1<<27, 1699 IFF_FAILOVER_SLAVE = 1<<28, 1700 IFF_L3MDEV_RX_HANDLER = 1<<29, 1701 IFF_NO_ADDRCONF = BIT_ULL(30), 1702 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31), 1703 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32), 1704 }; 1705 1706 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1707 #define IFF_EBRIDGE IFF_EBRIDGE 1708 #define IFF_BONDING IFF_BONDING 1709 #define IFF_ISATAP IFF_ISATAP 1710 #define IFF_WAN_HDLC IFF_WAN_HDLC 1711 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1712 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1713 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1714 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1715 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1716 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1717 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1718 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1719 #define IFF_TEAM_PORT IFF_TEAM_PORT 1720 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1721 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1722 #define IFF_MACVLAN IFF_MACVLAN 1723 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1724 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1725 #define IFF_NO_QUEUE IFF_NO_QUEUE 1726 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1727 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1728 #define IFF_TEAM IFF_TEAM 1729 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1730 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM 1731 #define IFF_MACSEC IFF_MACSEC 1732 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1733 #define IFF_FAILOVER IFF_FAILOVER 1734 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1735 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1736 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR 1737 1738 /* Specifies the type of the struct net_device::ml_priv pointer */ 1739 enum netdev_ml_priv_type { 1740 ML_PRIV_NONE, 1741 ML_PRIV_CAN, 1742 }; 1743 1744 /** 1745 * struct net_device - The DEVICE structure. 1746 * 1747 * Actually, this whole structure is a big mistake. It mixes I/O 1748 * data with strictly "high-level" data, and it has to know about 1749 * almost every data structure used in the INET module. 1750 * 1751 * @name: This is the first field of the "visible" part of this structure 1752 * (i.e. as seen by users in the "Space.c" file). It is the name 1753 * of the interface. 1754 * 1755 * @name_node: Name hashlist node 1756 * @ifalias: SNMP alias 1757 * @mem_end: Shared memory end 1758 * @mem_start: Shared memory start 1759 * @base_addr: Device I/O address 1760 * @irq: Device IRQ number 1761 * 1762 * @state: Generic network queuing layer state, see netdev_state_t 1763 * @dev_list: The global list of network devices 1764 * @napi_list: List entry used for polling NAPI devices 1765 * @unreg_list: List entry when we are unregistering the 1766 * device; see the function unregister_netdev 1767 * @close_list: List entry used when we are closing the device 1768 * @ptype_all: Device-specific packet handlers for all protocols 1769 * @ptype_specific: Device-specific, protocol-specific packet handlers 1770 * 1771 * @adj_list: Directly linked devices, like slaves for bonding 1772 * @features: Currently active device features 1773 * @hw_features: User-changeable features 1774 * 1775 * @wanted_features: User-requested features 1776 * @vlan_features: Mask of features inheritable by VLAN devices 1777 * 1778 * @hw_enc_features: Mask of features inherited by encapsulating devices 1779 * This field indicates what encapsulation 1780 * offloads the hardware is capable of doing, 1781 * and drivers will need to set them appropriately. 1782 * 1783 * @mpls_features: Mask of features inheritable by MPLS 1784 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1785 * 1786 * @ifindex: interface index 1787 * @group: The group the device belongs to 1788 * 1789 * @stats: Statistics struct, which was left as a legacy, use 1790 * rtnl_link_stats64 instead 1791 * 1792 * @core_stats: core networking counters, 1793 * do not use this in drivers 1794 * @carrier_up_count: Number of times the carrier has been up 1795 * @carrier_down_count: Number of times the carrier has been down 1796 * 1797 * @wireless_handlers: List of functions to handle Wireless Extensions, 1798 * instead of ioctl, 1799 * see <net/iw_handler.h> for details. 1800 * @wireless_data: Instance data managed by the core of wireless extensions 1801 * 1802 * @netdev_ops: Includes several pointers to callbacks, 1803 * if one wants to override the ndo_*() functions 1804 * @ethtool_ops: Management operations 1805 * @l3mdev_ops: Layer 3 master device operations 1806 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1807 * discovery handling. Necessary for e.g. 6LoWPAN. 1808 * @xfrmdev_ops: Transformation offload operations 1809 * @tlsdev_ops: Transport Layer Security offload operations 1810 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1811 * of Layer 2 headers. 1812 * 1813 * @flags: Interface flags (a la BSD) 1814 * @priv_flags: Like 'flags' but invisible to userspace, 1815 * see if.h for the definitions 1816 * @gflags: Global flags ( kept as legacy ) 1817 * @padded: How much padding added by alloc_netdev() 1818 * @operstate: RFC2863 operstate 1819 * @link_mode: Mapping policy to operstate 1820 * @if_port: Selectable AUI, TP, ... 1821 * @dma: DMA channel 1822 * @mtu: Interface MTU value 1823 * @min_mtu: Interface Minimum MTU value 1824 * @max_mtu: Interface Maximum MTU value 1825 * @type: Interface hardware type 1826 * @hard_header_len: Maximum hardware header length. 1827 * @min_header_len: Minimum hardware header length 1828 * 1829 * @needed_headroom: Extra headroom the hardware may need, but not in all 1830 * cases can this be guaranteed 1831 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1832 * cases can this be guaranteed. Some cases also use 1833 * LL_MAX_HEADER instead to allocate the skb 1834 * 1835 * interface address info: 1836 * 1837 * @perm_addr: Permanent hw address 1838 * @addr_assign_type: Hw address assignment type 1839 * @addr_len: Hardware address length 1840 * @upper_level: Maximum depth level of upper devices. 1841 * @lower_level: Maximum depth level of lower devices. 1842 * @neigh_priv_len: Used in neigh_alloc() 1843 * @dev_id: Used to differentiate devices that share 1844 * the same link layer address 1845 * @dev_port: Used to differentiate devices that share 1846 * the same function 1847 * @addr_list_lock: XXX: need comments on this one 1848 * @name_assign_type: network interface name assignment type 1849 * @uc_promisc: Counter that indicates promiscuous mode 1850 * has been enabled due to the need to listen to 1851 * additional unicast addresses in a device that 1852 * does not implement ndo_set_rx_mode() 1853 * @uc: unicast mac addresses 1854 * @mc: multicast mac addresses 1855 * @dev_addrs: list of device hw addresses 1856 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1857 * @promiscuity: Number of times the NIC is told to work in 1858 * promiscuous mode; if it becomes 0 the NIC will 1859 * exit promiscuous mode 1860 * @allmulti: Counter, enables or disables allmulticast mode 1861 * 1862 * @vlan_info: VLAN info 1863 * @dsa_ptr: dsa specific data 1864 * @tipc_ptr: TIPC specific data 1865 * @atalk_ptr: AppleTalk link 1866 * @ip_ptr: IPv4 specific data 1867 * @ip6_ptr: IPv6 specific data 1868 * @ax25_ptr: AX.25 specific data 1869 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1870 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1871 * device struct 1872 * @mpls_ptr: mpls_dev struct pointer 1873 * @mctp_ptr: MCTP specific data 1874 * 1875 * @dev_addr: Hw address (before bcast, 1876 * because most packets are unicast) 1877 * 1878 * @_rx: Array of RX queues 1879 * @num_rx_queues: Number of RX queues 1880 * allocated at register_netdev() time 1881 * @real_num_rx_queues: Number of RX queues currently active in device 1882 * @xdp_prog: XDP sockets filter program pointer 1883 * @gro_flush_timeout: timeout for GRO layer in NAPI 1884 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1885 * allow to avoid NIC hard IRQ, on busy queues. 1886 * 1887 * @rx_handler: handler for received packets 1888 * @rx_handler_data: XXX: need comments on this one 1889 * @miniq_ingress: ingress/clsact qdisc specific data for 1890 * ingress processing 1891 * @ingress_queue: XXX: need comments on this one 1892 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1893 * @broadcast: hw bcast address 1894 * 1895 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1896 * indexed by RX queue number. Assigned by driver. 1897 * This must only be set if the ndo_rx_flow_steer 1898 * operation is defined 1899 * @index_hlist: Device index hash chain 1900 * 1901 * @_tx: Array of TX queues 1902 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1903 * @real_num_tx_queues: Number of TX queues currently active in device 1904 * @qdisc: Root qdisc from userspace point of view 1905 * @tx_queue_len: Max frames per queue allowed 1906 * @tx_global_lock: XXX: need comments on this one 1907 * @xdp_bulkq: XDP device bulk queue 1908 * @xps_maps: all CPUs/RXQs maps for XPS device 1909 * 1910 * @xps_maps: XXX: need comments on this one 1911 * @miniq_egress: clsact qdisc specific data for 1912 * egress processing 1913 * @nf_hooks_egress: netfilter hooks executed for egress packets 1914 * @qdisc_hash: qdisc hash table 1915 * @watchdog_timeo: Represents the timeout that is used by 1916 * the watchdog (see dev_watchdog()) 1917 * @watchdog_timer: List of timers 1918 * 1919 * @proto_down_reason: reason a netdev interface is held down 1920 * @pcpu_refcnt: Number of references to this device 1921 * @dev_refcnt: Number of references to this device 1922 * @refcnt_tracker: Tracker directory for tracked references to this device 1923 * @todo_list: Delayed register/unregister 1924 * @link_watch_list: XXX: need comments on this one 1925 * 1926 * @reg_state: Register/unregister state machine 1927 * @dismantle: Device is going to be freed 1928 * @rtnl_link_state: This enum represents the phases of creating 1929 * a new link 1930 * 1931 * @needs_free_netdev: Should unregister perform free_netdev? 1932 * @priv_destructor: Called from unregister 1933 * @npinfo: XXX: need comments on this one 1934 * @nd_net: Network namespace this network device is inside 1935 * 1936 * @ml_priv: Mid-layer private 1937 * @ml_priv_type: Mid-layer private type 1938 * @lstats: Loopback statistics 1939 * @tstats: Tunnel statistics 1940 * @dstats: Dummy statistics 1941 * @vstats: Virtual ethernet statistics 1942 * 1943 * @garp_port: GARP 1944 * @mrp_port: MRP 1945 * 1946 * @dm_private: Drop monitor private 1947 * 1948 * @dev: Class/net/name entry 1949 * @sysfs_groups: Space for optional device, statistics and wireless 1950 * sysfs groups 1951 * 1952 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1953 * @rtnl_link_ops: Rtnl_link_ops 1954 * 1955 * @gso_max_size: Maximum size of generic segmentation offload 1956 * @tso_max_size: Device (as in HW) limit on the max TSO request size 1957 * @gso_max_segs: Maximum number of segments that can be passed to the 1958 * NIC for GSO 1959 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count 1960 * 1961 * @dcbnl_ops: Data Center Bridging netlink ops 1962 * @num_tc: Number of traffic classes in the net device 1963 * @tc_to_txq: XXX: need comments on this one 1964 * @prio_tc_map: XXX: need comments on this one 1965 * 1966 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1967 * 1968 * @priomap: XXX: need comments on this one 1969 * @phydev: Physical device may attach itself 1970 * for hardware timestamping 1971 * @sfp_bus: attached &struct sfp_bus structure. 1972 * 1973 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1974 * 1975 * @proto_down: protocol port state information can be sent to the 1976 * switch driver and used to set the phys state of the 1977 * switch port. 1978 * 1979 * @wol_enabled: Wake-on-LAN is enabled 1980 * 1981 * @threaded: napi threaded mode is enabled 1982 * 1983 * @net_notifier_list: List of per-net netdev notifier block 1984 * that follow this device when it is moved 1985 * to another network namespace. 1986 * 1987 * @macsec_ops: MACsec offloading ops 1988 * 1989 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1990 * offload capabilities of the device 1991 * @udp_tunnel_nic: UDP tunnel offload state 1992 * @xdp_state: stores info on attached XDP BPF programs 1993 * 1994 * @nested_level: Used as a parameter of spin_lock_nested() of 1995 * dev->addr_list_lock. 1996 * @unlink_list: As netif_addr_lock() can be called recursively, 1997 * keep a list of interfaces to be deleted. 1998 * @gro_max_size: Maximum size of aggregated packet in generic 1999 * receive offload (GRO) 2000 * 2001 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 2002 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 2003 * @watchdog_dev_tracker: refcount tracker used by watchdog. 2004 * @dev_registered_tracker: tracker for reference held while 2005 * registered 2006 * @offload_xstats_l3: L3 HW stats for this netdevice. 2007 * 2008 * @devlink_port: Pointer to related devlink port structure. 2009 * Assigned by a driver before netdev registration using 2010 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static 2011 * during the time netdevice is registered. 2012 * 2013 * FIXME: cleanup struct net_device such that network protocol info 2014 * moves out. 2015 */ 2016 2017 struct net_device { 2018 char name[IFNAMSIZ]; 2019 struct netdev_name_node *name_node; 2020 struct dev_ifalias __rcu *ifalias; 2021 /* 2022 * I/O specific fields 2023 * FIXME: Merge these and struct ifmap into one 2024 */ 2025 unsigned long mem_end; 2026 unsigned long mem_start; 2027 unsigned long base_addr; 2028 2029 /* 2030 * Some hardware also needs these fields (state,dev_list, 2031 * napi_list,unreg_list,close_list) but they are not 2032 * part of the usual set specified in Space.c. 2033 */ 2034 2035 unsigned long state; 2036 2037 struct list_head dev_list; 2038 struct list_head napi_list; 2039 struct list_head unreg_list; 2040 struct list_head close_list; 2041 struct list_head ptype_all; 2042 struct list_head ptype_specific; 2043 2044 struct { 2045 struct list_head upper; 2046 struct list_head lower; 2047 } adj_list; 2048 2049 /* Read-mostly cache-line for fast-path access */ 2050 unsigned int flags; 2051 unsigned long long priv_flags; 2052 const struct net_device_ops *netdev_ops; 2053 int ifindex; 2054 unsigned short gflags; 2055 unsigned short hard_header_len; 2056 2057 /* Note : dev->mtu is often read without holding a lock. 2058 * Writers usually hold RTNL. 2059 * It is recommended to use READ_ONCE() to annotate the reads, 2060 * and to use WRITE_ONCE() to annotate the writes. 2061 */ 2062 unsigned int mtu; 2063 unsigned short needed_headroom; 2064 unsigned short needed_tailroom; 2065 2066 netdev_features_t features; 2067 netdev_features_t hw_features; 2068 netdev_features_t wanted_features; 2069 netdev_features_t vlan_features; 2070 netdev_features_t hw_enc_features; 2071 netdev_features_t mpls_features; 2072 netdev_features_t gso_partial_features; 2073 2074 unsigned int min_mtu; 2075 unsigned int max_mtu; 2076 unsigned short type; 2077 unsigned char min_header_len; 2078 unsigned char name_assign_type; 2079 2080 int group; 2081 2082 struct net_device_stats stats; /* not used by modern drivers */ 2083 2084 struct net_device_core_stats __percpu *core_stats; 2085 2086 /* Stats to monitor link on/off, flapping */ 2087 atomic_t carrier_up_count; 2088 atomic_t carrier_down_count; 2089 2090 #ifdef CONFIG_WIRELESS_EXT 2091 const struct iw_handler_def *wireless_handlers; 2092 struct iw_public_data *wireless_data; 2093 #endif 2094 const struct ethtool_ops *ethtool_ops; 2095 #ifdef CONFIG_NET_L3_MASTER_DEV 2096 const struct l3mdev_ops *l3mdev_ops; 2097 #endif 2098 #if IS_ENABLED(CONFIG_IPV6) 2099 const struct ndisc_ops *ndisc_ops; 2100 #endif 2101 2102 #ifdef CONFIG_XFRM_OFFLOAD 2103 const struct xfrmdev_ops *xfrmdev_ops; 2104 #endif 2105 2106 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2107 const struct tlsdev_ops *tlsdev_ops; 2108 #endif 2109 2110 const struct header_ops *header_ops; 2111 2112 unsigned char operstate; 2113 unsigned char link_mode; 2114 2115 unsigned char if_port; 2116 unsigned char dma; 2117 2118 /* Interface address info. */ 2119 unsigned char perm_addr[MAX_ADDR_LEN]; 2120 unsigned char addr_assign_type; 2121 unsigned char addr_len; 2122 unsigned char upper_level; 2123 unsigned char lower_level; 2124 2125 unsigned short neigh_priv_len; 2126 unsigned short dev_id; 2127 unsigned short dev_port; 2128 unsigned short padded; 2129 2130 spinlock_t addr_list_lock; 2131 int irq; 2132 2133 struct netdev_hw_addr_list uc; 2134 struct netdev_hw_addr_list mc; 2135 struct netdev_hw_addr_list dev_addrs; 2136 2137 #ifdef CONFIG_SYSFS 2138 struct kset *queues_kset; 2139 #endif 2140 #ifdef CONFIG_LOCKDEP 2141 struct list_head unlink_list; 2142 #endif 2143 unsigned int promiscuity; 2144 unsigned int allmulti; 2145 bool uc_promisc; 2146 #ifdef CONFIG_LOCKDEP 2147 unsigned char nested_level; 2148 #endif 2149 2150 2151 /* Protocol-specific pointers */ 2152 2153 struct in_device __rcu *ip_ptr; 2154 struct inet6_dev __rcu *ip6_ptr; 2155 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2156 struct vlan_info __rcu *vlan_info; 2157 #endif 2158 #if IS_ENABLED(CONFIG_NET_DSA) 2159 struct dsa_port *dsa_ptr; 2160 #endif 2161 #if IS_ENABLED(CONFIG_TIPC) 2162 struct tipc_bearer __rcu *tipc_ptr; 2163 #endif 2164 #if IS_ENABLED(CONFIG_ATALK) 2165 void *atalk_ptr; 2166 #endif 2167 #if IS_ENABLED(CONFIG_AX25) 2168 void *ax25_ptr; 2169 #endif 2170 #if IS_ENABLED(CONFIG_CFG80211) 2171 struct wireless_dev *ieee80211_ptr; 2172 #endif 2173 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2174 struct wpan_dev *ieee802154_ptr; 2175 #endif 2176 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2177 struct mpls_dev __rcu *mpls_ptr; 2178 #endif 2179 #if IS_ENABLED(CONFIG_MCTP) 2180 struct mctp_dev __rcu *mctp_ptr; 2181 #endif 2182 2183 /* 2184 * Cache lines mostly used on receive path (including eth_type_trans()) 2185 */ 2186 /* Interface address info used in eth_type_trans() */ 2187 const unsigned char *dev_addr; 2188 2189 struct netdev_rx_queue *_rx; 2190 unsigned int num_rx_queues; 2191 unsigned int real_num_rx_queues; 2192 2193 struct bpf_prog __rcu *xdp_prog; 2194 unsigned long gro_flush_timeout; 2195 int napi_defer_hard_irqs; 2196 #define GRO_LEGACY_MAX_SIZE 65536u 2197 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2198 * and shinfo->gso_segs is a 16bit field. 2199 */ 2200 #define GRO_MAX_SIZE (8 * 65535u) 2201 unsigned int gro_max_size; 2202 rx_handler_func_t __rcu *rx_handler; 2203 void __rcu *rx_handler_data; 2204 2205 #ifdef CONFIG_NET_CLS_ACT 2206 struct mini_Qdisc __rcu *miniq_ingress; 2207 #endif 2208 struct netdev_queue __rcu *ingress_queue; 2209 #ifdef CONFIG_NETFILTER_INGRESS 2210 struct nf_hook_entries __rcu *nf_hooks_ingress; 2211 #endif 2212 2213 unsigned char broadcast[MAX_ADDR_LEN]; 2214 #ifdef CONFIG_RFS_ACCEL 2215 struct cpu_rmap *rx_cpu_rmap; 2216 #endif 2217 struct hlist_node index_hlist; 2218 2219 /* 2220 * Cache lines mostly used on transmit path 2221 */ 2222 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2223 unsigned int num_tx_queues; 2224 unsigned int real_num_tx_queues; 2225 struct Qdisc __rcu *qdisc; 2226 unsigned int tx_queue_len; 2227 spinlock_t tx_global_lock; 2228 2229 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2230 2231 #ifdef CONFIG_XPS 2232 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2233 #endif 2234 #ifdef CONFIG_NET_CLS_ACT 2235 struct mini_Qdisc __rcu *miniq_egress; 2236 #endif 2237 #ifdef CONFIG_NETFILTER_EGRESS 2238 struct nf_hook_entries __rcu *nf_hooks_egress; 2239 #endif 2240 2241 #ifdef CONFIG_NET_SCHED 2242 DECLARE_HASHTABLE (qdisc_hash, 4); 2243 #endif 2244 /* These may be needed for future network-power-down code. */ 2245 struct timer_list watchdog_timer; 2246 int watchdog_timeo; 2247 2248 u32 proto_down_reason; 2249 2250 struct list_head todo_list; 2251 2252 #ifdef CONFIG_PCPU_DEV_REFCNT 2253 int __percpu *pcpu_refcnt; 2254 #else 2255 refcount_t dev_refcnt; 2256 #endif 2257 struct ref_tracker_dir refcnt_tracker; 2258 2259 struct list_head link_watch_list; 2260 2261 enum { NETREG_UNINITIALIZED=0, 2262 NETREG_REGISTERED, /* completed register_netdevice */ 2263 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2264 NETREG_UNREGISTERED, /* completed unregister todo */ 2265 NETREG_RELEASED, /* called free_netdev */ 2266 NETREG_DUMMY, /* dummy device for NAPI poll */ 2267 } reg_state:8; 2268 2269 bool dismantle; 2270 2271 enum { 2272 RTNL_LINK_INITIALIZED, 2273 RTNL_LINK_INITIALIZING, 2274 } rtnl_link_state:16; 2275 2276 bool needs_free_netdev; 2277 void (*priv_destructor)(struct net_device *dev); 2278 2279 #ifdef CONFIG_NETPOLL 2280 struct netpoll_info __rcu *npinfo; 2281 #endif 2282 2283 possible_net_t nd_net; 2284 2285 /* mid-layer private */ 2286 void *ml_priv; 2287 enum netdev_ml_priv_type ml_priv_type; 2288 2289 union { 2290 struct pcpu_lstats __percpu *lstats; 2291 struct pcpu_sw_netstats __percpu *tstats; 2292 struct pcpu_dstats __percpu *dstats; 2293 }; 2294 2295 #if IS_ENABLED(CONFIG_GARP) 2296 struct garp_port __rcu *garp_port; 2297 #endif 2298 #if IS_ENABLED(CONFIG_MRP) 2299 struct mrp_port __rcu *mrp_port; 2300 #endif 2301 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2302 struct dm_hw_stat_delta __rcu *dm_private; 2303 #endif 2304 struct device dev; 2305 const struct attribute_group *sysfs_groups[4]; 2306 const struct attribute_group *sysfs_rx_queue_group; 2307 2308 const struct rtnl_link_ops *rtnl_link_ops; 2309 2310 /* for setting kernel sock attribute on TCP connection setup */ 2311 #define GSO_MAX_SEGS 65535u 2312 #define GSO_LEGACY_MAX_SIZE 65536u 2313 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2314 * and shinfo->gso_segs is a 16bit field. 2315 */ 2316 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2317 2318 unsigned int gso_max_size; 2319 #define TSO_LEGACY_MAX_SIZE 65536 2320 #define TSO_MAX_SIZE UINT_MAX 2321 unsigned int tso_max_size; 2322 u16 gso_max_segs; 2323 #define TSO_MAX_SEGS U16_MAX 2324 u16 tso_max_segs; 2325 2326 #ifdef CONFIG_DCB 2327 const struct dcbnl_rtnl_ops *dcbnl_ops; 2328 #endif 2329 s16 num_tc; 2330 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2331 u8 prio_tc_map[TC_BITMASK + 1]; 2332 2333 #if IS_ENABLED(CONFIG_FCOE) 2334 unsigned int fcoe_ddp_xid; 2335 #endif 2336 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2337 struct netprio_map __rcu *priomap; 2338 #endif 2339 struct phy_device *phydev; 2340 struct sfp_bus *sfp_bus; 2341 struct lock_class_key *qdisc_tx_busylock; 2342 bool proto_down; 2343 unsigned wol_enabled:1; 2344 unsigned threaded:1; 2345 2346 struct list_head net_notifier_list; 2347 2348 #if IS_ENABLED(CONFIG_MACSEC) 2349 /* MACsec management functions */ 2350 const struct macsec_ops *macsec_ops; 2351 #endif 2352 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2353 struct udp_tunnel_nic *udp_tunnel_nic; 2354 2355 /* protected by rtnl_lock */ 2356 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2357 2358 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2359 netdevice_tracker linkwatch_dev_tracker; 2360 netdevice_tracker watchdog_dev_tracker; 2361 netdevice_tracker dev_registered_tracker; 2362 struct rtnl_hw_stats64 *offload_xstats_l3; 2363 2364 struct devlink_port *devlink_port; 2365 }; 2366 #define to_net_dev(d) container_of(d, struct net_device, dev) 2367 2368 /* 2369 * Driver should use this to assign devlink port instance to a netdevice 2370 * before it registers the netdevice. Therefore devlink_port is static 2371 * during the netdev lifetime after it is registered. 2372 */ 2373 #define SET_NETDEV_DEVLINK_PORT(dev, port) \ 2374 ({ \ 2375 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \ 2376 ((dev)->devlink_port = (port)); \ 2377 }) 2378 2379 static inline bool netif_elide_gro(const struct net_device *dev) 2380 { 2381 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2382 return true; 2383 return false; 2384 } 2385 2386 #define NETDEV_ALIGN 32 2387 2388 static inline 2389 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2390 { 2391 return dev->prio_tc_map[prio & TC_BITMASK]; 2392 } 2393 2394 static inline 2395 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2396 { 2397 if (tc >= dev->num_tc) 2398 return -EINVAL; 2399 2400 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2401 return 0; 2402 } 2403 2404 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2405 void netdev_reset_tc(struct net_device *dev); 2406 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2407 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2408 2409 static inline 2410 int netdev_get_num_tc(struct net_device *dev) 2411 { 2412 return dev->num_tc; 2413 } 2414 2415 static inline void net_prefetch(void *p) 2416 { 2417 prefetch(p); 2418 #if L1_CACHE_BYTES < 128 2419 prefetch((u8 *)p + L1_CACHE_BYTES); 2420 #endif 2421 } 2422 2423 static inline void net_prefetchw(void *p) 2424 { 2425 prefetchw(p); 2426 #if L1_CACHE_BYTES < 128 2427 prefetchw((u8 *)p + L1_CACHE_BYTES); 2428 #endif 2429 } 2430 2431 void netdev_unbind_sb_channel(struct net_device *dev, 2432 struct net_device *sb_dev); 2433 int netdev_bind_sb_channel_queue(struct net_device *dev, 2434 struct net_device *sb_dev, 2435 u8 tc, u16 count, u16 offset); 2436 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2437 static inline int netdev_get_sb_channel(struct net_device *dev) 2438 { 2439 return max_t(int, -dev->num_tc, 0); 2440 } 2441 2442 static inline 2443 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2444 unsigned int index) 2445 { 2446 return &dev->_tx[index]; 2447 } 2448 2449 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2450 const struct sk_buff *skb) 2451 { 2452 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2453 } 2454 2455 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2456 void (*f)(struct net_device *, 2457 struct netdev_queue *, 2458 void *), 2459 void *arg) 2460 { 2461 unsigned int i; 2462 2463 for (i = 0; i < dev->num_tx_queues; i++) 2464 f(dev, &dev->_tx[i], arg); 2465 } 2466 2467 #define netdev_lockdep_set_classes(dev) \ 2468 { \ 2469 static struct lock_class_key qdisc_tx_busylock_key; \ 2470 static struct lock_class_key qdisc_xmit_lock_key; \ 2471 static struct lock_class_key dev_addr_list_lock_key; \ 2472 unsigned int i; \ 2473 \ 2474 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2475 lockdep_set_class(&(dev)->addr_list_lock, \ 2476 &dev_addr_list_lock_key); \ 2477 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2478 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2479 &qdisc_xmit_lock_key); \ 2480 } 2481 2482 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2483 struct net_device *sb_dev); 2484 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2485 struct sk_buff *skb, 2486 struct net_device *sb_dev); 2487 2488 /* returns the headroom that the master device needs to take in account 2489 * when forwarding to this dev 2490 */ 2491 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2492 { 2493 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2494 } 2495 2496 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2497 { 2498 if (dev->netdev_ops->ndo_set_rx_headroom) 2499 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2500 } 2501 2502 /* set the device rx headroom to the dev's default */ 2503 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2504 { 2505 netdev_set_rx_headroom(dev, -1); 2506 } 2507 2508 static inline void *netdev_get_ml_priv(struct net_device *dev, 2509 enum netdev_ml_priv_type type) 2510 { 2511 if (dev->ml_priv_type != type) 2512 return NULL; 2513 2514 return dev->ml_priv; 2515 } 2516 2517 static inline void netdev_set_ml_priv(struct net_device *dev, 2518 void *ml_priv, 2519 enum netdev_ml_priv_type type) 2520 { 2521 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2522 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2523 dev->ml_priv_type, type); 2524 WARN(!dev->ml_priv_type && dev->ml_priv, 2525 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2526 2527 dev->ml_priv = ml_priv; 2528 dev->ml_priv_type = type; 2529 } 2530 2531 /* 2532 * Net namespace inlines 2533 */ 2534 static inline 2535 struct net *dev_net(const struct net_device *dev) 2536 { 2537 return read_pnet(&dev->nd_net); 2538 } 2539 2540 static inline 2541 void dev_net_set(struct net_device *dev, struct net *net) 2542 { 2543 write_pnet(&dev->nd_net, net); 2544 } 2545 2546 /** 2547 * netdev_priv - access network device private data 2548 * @dev: network device 2549 * 2550 * Get network device private data 2551 */ 2552 static inline void *netdev_priv(const struct net_device *dev) 2553 { 2554 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2555 } 2556 2557 /* Set the sysfs physical device reference for the network logical device 2558 * if set prior to registration will cause a symlink during initialization. 2559 */ 2560 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2561 2562 /* Set the sysfs device type for the network logical device to allow 2563 * fine-grained identification of different network device types. For 2564 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2565 */ 2566 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2567 2568 /* Default NAPI poll() weight 2569 * Device drivers are strongly advised to not use bigger value 2570 */ 2571 #define NAPI_POLL_WEIGHT 64 2572 2573 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2574 int (*poll)(struct napi_struct *, int), int weight); 2575 2576 /** 2577 * netif_napi_add() - initialize a NAPI context 2578 * @dev: network device 2579 * @napi: NAPI context 2580 * @poll: polling function 2581 * 2582 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2583 * *any* of the other NAPI-related functions. 2584 */ 2585 static inline void 2586 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2587 int (*poll)(struct napi_struct *, int)) 2588 { 2589 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2590 } 2591 2592 static inline void 2593 netif_napi_add_tx_weight(struct net_device *dev, 2594 struct napi_struct *napi, 2595 int (*poll)(struct napi_struct *, int), 2596 int weight) 2597 { 2598 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2599 netif_napi_add_weight(dev, napi, poll, weight); 2600 } 2601 2602 /** 2603 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2604 * @dev: network device 2605 * @napi: NAPI context 2606 * @poll: polling function 2607 * 2608 * This variant of netif_napi_add() should be used from drivers using NAPI 2609 * to exclusively poll a TX queue. 2610 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2611 */ 2612 static inline void netif_napi_add_tx(struct net_device *dev, 2613 struct napi_struct *napi, 2614 int (*poll)(struct napi_struct *, int)) 2615 { 2616 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2617 } 2618 2619 /** 2620 * __netif_napi_del - remove a NAPI context 2621 * @napi: NAPI context 2622 * 2623 * Warning: caller must observe RCU grace period before freeing memory 2624 * containing @napi. Drivers might want to call this helper to combine 2625 * all the needed RCU grace periods into a single one. 2626 */ 2627 void __netif_napi_del(struct napi_struct *napi); 2628 2629 /** 2630 * netif_napi_del - remove a NAPI context 2631 * @napi: NAPI context 2632 * 2633 * netif_napi_del() removes a NAPI context from the network device NAPI list 2634 */ 2635 static inline void netif_napi_del(struct napi_struct *napi) 2636 { 2637 __netif_napi_del(napi); 2638 synchronize_net(); 2639 } 2640 2641 struct packet_type { 2642 __be16 type; /* This is really htons(ether_type). */ 2643 bool ignore_outgoing; 2644 struct net_device *dev; /* NULL is wildcarded here */ 2645 netdevice_tracker dev_tracker; 2646 int (*func) (struct sk_buff *, 2647 struct net_device *, 2648 struct packet_type *, 2649 struct net_device *); 2650 void (*list_func) (struct list_head *, 2651 struct packet_type *, 2652 struct net_device *); 2653 bool (*id_match)(struct packet_type *ptype, 2654 struct sock *sk); 2655 struct net *af_packet_net; 2656 void *af_packet_priv; 2657 struct list_head list; 2658 }; 2659 2660 struct offload_callbacks { 2661 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2662 netdev_features_t features); 2663 struct sk_buff *(*gro_receive)(struct list_head *head, 2664 struct sk_buff *skb); 2665 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2666 }; 2667 2668 struct packet_offload { 2669 __be16 type; /* This is really htons(ether_type). */ 2670 u16 priority; 2671 struct offload_callbacks callbacks; 2672 struct list_head list; 2673 }; 2674 2675 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2676 struct pcpu_sw_netstats { 2677 u64_stats_t rx_packets; 2678 u64_stats_t rx_bytes; 2679 u64_stats_t tx_packets; 2680 u64_stats_t tx_bytes; 2681 struct u64_stats_sync syncp; 2682 } __aligned(4 * sizeof(u64)); 2683 2684 struct pcpu_lstats { 2685 u64_stats_t packets; 2686 u64_stats_t bytes; 2687 struct u64_stats_sync syncp; 2688 } __aligned(2 * sizeof(u64)); 2689 2690 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2691 2692 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2693 { 2694 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2695 2696 u64_stats_update_begin(&tstats->syncp); 2697 u64_stats_add(&tstats->rx_bytes, len); 2698 u64_stats_inc(&tstats->rx_packets); 2699 u64_stats_update_end(&tstats->syncp); 2700 } 2701 2702 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2703 unsigned int packets, 2704 unsigned int len) 2705 { 2706 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2707 2708 u64_stats_update_begin(&tstats->syncp); 2709 u64_stats_add(&tstats->tx_bytes, len); 2710 u64_stats_add(&tstats->tx_packets, packets); 2711 u64_stats_update_end(&tstats->syncp); 2712 } 2713 2714 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2715 { 2716 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2717 2718 u64_stats_update_begin(&lstats->syncp); 2719 u64_stats_add(&lstats->bytes, len); 2720 u64_stats_inc(&lstats->packets); 2721 u64_stats_update_end(&lstats->syncp); 2722 } 2723 2724 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2725 ({ \ 2726 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2727 if (pcpu_stats) { \ 2728 int __cpu; \ 2729 for_each_possible_cpu(__cpu) { \ 2730 typeof(type) *stat; \ 2731 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2732 u64_stats_init(&stat->syncp); \ 2733 } \ 2734 } \ 2735 pcpu_stats; \ 2736 }) 2737 2738 #define netdev_alloc_pcpu_stats(type) \ 2739 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2740 2741 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2742 ({ \ 2743 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2744 if (pcpu_stats) { \ 2745 int __cpu; \ 2746 for_each_possible_cpu(__cpu) { \ 2747 typeof(type) *stat; \ 2748 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2749 u64_stats_init(&stat->syncp); \ 2750 } \ 2751 } \ 2752 pcpu_stats; \ 2753 }) 2754 2755 enum netdev_lag_tx_type { 2756 NETDEV_LAG_TX_TYPE_UNKNOWN, 2757 NETDEV_LAG_TX_TYPE_RANDOM, 2758 NETDEV_LAG_TX_TYPE_BROADCAST, 2759 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2760 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2761 NETDEV_LAG_TX_TYPE_HASH, 2762 }; 2763 2764 enum netdev_lag_hash { 2765 NETDEV_LAG_HASH_NONE, 2766 NETDEV_LAG_HASH_L2, 2767 NETDEV_LAG_HASH_L34, 2768 NETDEV_LAG_HASH_L23, 2769 NETDEV_LAG_HASH_E23, 2770 NETDEV_LAG_HASH_E34, 2771 NETDEV_LAG_HASH_VLAN_SRCMAC, 2772 NETDEV_LAG_HASH_UNKNOWN, 2773 }; 2774 2775 struct netdev_lag_upper_info { 2776 enum netdev_lag_tx_type tx_type; 2777 enum netdev_lag_hash hash_type; 2778 }; 2779 2780 struct netdev_lag_lower_state_info { 2781 u8 link_up : 1, 2782 tx_enabled : 1; 2783 }; 2784 2785 #include <linux/notifier.h> 2786 2787 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2788 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2789 * adding new types. 2790 */ 2791 enum netdev_cmd { 2792 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2793 NETDEV_DOWN, 2794 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2795 detected a hardware crash and restarted 2796 - we can use this eg to kick tcp sessions 2797 once done */ 2798 NETDEV_CHANGE, /* Notify device state change */ 2799 NETDEV_REGISTER, 2800 NETDEV_UNREGISTER, 2801 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2802 NETDEV_CHANGEADDR, /* notify after the address change */ 2803 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2804 NETDEV_GOING_DOWN, 2805 NETDEV_CHANGENAME, 2806 NETDEV_FEAT_CHANGE, 2807 NETDEV_BONDING_FAILOVER, 2808 NETDEV_PRE_UP, 2809 NETDEV_PRE_TYPE_CHANGE, 2810 NETDEV_POST_TYPE_CHANGE, 2811 NETDEV_POST_INIT, 2812 NETDEV_PRE_UNINIT, 2813 NETDEV_RELEASE, 2814 NETDEV_NOTIFY_PEERS, 2815 NETDEV_JOIN, 2816 NETDEV_CHANGEUPPER, 2817 NETDEV_RESEND_IGMP, 2818 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2819 NETDEV_CHANGEINFODATA, 2820 NETDEV_BONDING_INFO, 2821 NETDEV_PRECHANGEUPPER, 2822 NETDEV_CHANGELOWERSTATE, 2823 NETDEV_UDP_TUNNEL_PUSH_INFO, 2824 NETDEV_UDP_TUNNEL_DROP_INFO, 2825 NETDEV_CHANGE_TX_QUEUE_LEN, 2826 NETDEV_CVLAN_FILTER_PUSH_INFO, 2827 NETDEV_CVLAN_FILTER_DROP_INFO, 2828 NETDEV_SVLAN_FILTER_PUSH_INFO, 2829 NETDEV_SVLAN_FILTER_DROP_INFO, 2830 NETDEV_OFFLOAD_XSTATS_ENABLE, 2831 NETDEV_OFFLOAD_XSTATS_DISABLE, 2832 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 2833 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 2834 }; 2835 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2836 2837 int register_netdevice_notifier(struct notifier_block *nb); 2838 int unregister_netdevice_notifier(struct notifier_block *nb); 2839 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2840 int unregister_netdevice_notifier_net(struct net *net, 2841 struct notifier_block *nb); 2842 int register_netdevice_notifier_dev_net(struct net_device *dev, 2843 struct notifier_block *nb, 2844 struct netdev_net_notifier *nn); 2845 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2846 struct notifier_block *nb, 2847 struct netdev_net_notifier *nn); 2848 2849 struct netdev_notifier_info { 2850 struct net_device *dev; 2851 struct netlink_ext_ack *extack; 2852 }; 2853 2854 struct netdev_notifier_info_ext { 2855 struct netdev_notifier_info info; /* must be first */ 2856 union { 2857 u32 mtu; 2858 } ext; 2859 }; 2860 2861 struct netdev_notifier_change_info { 2862 struct netdev_notifier_info info; /* must be first */ 2863 unsigned int flags_changed; 2864 }; 2865 2866 struct netdev_notifier_changeupper_info { 2867 struct netdev_notifier_info info; /* must be first */ 2868 struct net_device *upper_dev; /* new upper dev */ 2869 bool master; /* is upper dev master */ 2870 bool linking; /* is the notification for link or unlink */ 2871 void *upper_info; /* upper dev info */ 2872 }; 2873 2874 struct netdev_notifier_changelowerstate_info { 2875 struct netdev_notifier_info info; /* must be first */ 2876 void *lower_state_info; /* is lower dev state */ 2877 }; 2878 2879 struct netdev_notifier_pre_changeaddr_info { 2880 struct netdev_notifier_info info; /* must be first */ 2881 const unsigned char *dev_addr; 2882 }; 2883 2884 enum netdev_offload_xstats_type { 2885 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 2886 }; 2887 2888 struct netdev_notifier_offload_xstats_info { 2889 struct netdev_notifier_info info; /* must be first */ 2890 enum netdev_offload_xstats_type type; 2891 2892 union { 2893 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 2894 struct netdev_notifier_offload_xstats_rd *report_delta; 2895 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 2896 struct netdev_notifier_offload_xstats_ru *report_used; 2897 }; 2898 }; 2899 2900 int netdev_offload_xstats_enable(struct net_device *dev, 2901 enum netdev_offload_xstats_type type, 2902 struct netlink_ext_ack *extack); 2903 int netdev_offload_xstats_disable(struct net_device *dev, 2904 enum netdev_offload_xstats_type type); 2905 bool netdev_offload_xstats_enabled(const struct net_device *dev, 2906 enum netdev_offload_xstats_type type); 2907 int netdev_offload_xstats_get(struct net_device *dev, 2908 enum netdev_offload_xstats_type type, 2909 struct rtnl_hw_stats64 *stats, bool *used, 2910 struct netlink_ext_ack *extack); 2911 void 2912 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 2913 const struct rtnl_hw_stats64 *stats); 2914 void 2915 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 2916 void netdev_offload_xstats_push_delta(struct net_device *dev, 2917 enum netdev_offload_xstats_type type, 2918 const struct rtnl_hw_stats64 *stats); 2919 2920 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2921 struct net_device *dev) 2922 { 2923 info->dev = dev; 2924 info->extack = NULL; 2925 } 2926 2927 static inline struct net_device * 2928 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2929 { 2930 return info->dev; 2931 } 2932 2933 static inline struct netlink_ext_ack * 2934 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2935 { 2936 return info->extack; 2937 } 2938 2939 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2940 2941 2942 extern rwlock_t dev_base_lock; /* Device list lock */ 2943 2944 #define for_each_netdev(net, d) \ 2945 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2946 #define for_each_netdev_reverse(net, d) \ 2947 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2948 #define for_each_netdev_rcu(net, d) \ 2949 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2950 #define for_each_netdev_safe(net, d, n) \ 2951 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2952 #define for_each_netdev_continue(net, d) \ 2953 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2954 #define for_each_netdev_continue_reverse(net, d) \ 2955 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2956 dev_list) 2957 #define for_each_netdev_continue_rcu(net, d) \ 2958 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2959 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2960 for_each_netdev_rcu(&init_net, slave) \ 2961 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2962 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2963 2964 static inline struct net_device *next_net_device(struct net_device *dev) 2965 { 2966 struct list_head *lh; 2967 struct net *net; 2968 2969 net = dev_net(dev); 2970 lh = dev->dev_list.next; 2971 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2972 } 2973 2974 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2975 { 2976 struct list_head *lh; 2977 struct net *net; 2978 2979 net = dev_net(dev); 2980 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2981 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2982 } 2983 2984 static inline struct net_device *first_net_device(struct net *net) 2985 { 2986 return list_empty(&net->dev_base_head) ? NULL : 2987 net_device_entry(net->dev_base_head.next); 2988 } 2989 2990 static inline struct net_device *first_net_device_rcu(struct net *net) 2991 { 2992 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2993 2994 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2995 } 2996 2997 int netdev_boot_setup_check(struct net_device *dev); 2998 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2999 const char *hwaddr); 3000 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 3001 void dev_add_pack(struct packet_type *pt); 3002 void dev_remove_pack(struct packet_type *pt); 3003 void __dev_remove_pack(struct packet_type *pt); 3004 void dev_add_offload(struct packet_offload *po); 3005 void dev_remove_offload(struct packet_offload *po); 3006 3007 int dev_get_iflink(const struct net_device *dev); 3008 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 3009 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 3010 struct net_device_path_stack *stack); 3011 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 3012 unsigned short mask); 3013 struct net_device *dev_get_by_name(struct net *net, const char *name); 3014 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 3015 struct net_device *__dev_get_by_name(struct net *net, const char *name); 3016 bool netdev_name_in_use(struct net *net, const char *name); 3017 int dev_alloc_name(struct net_device *dev, const char *name); 3018 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 3019 void dev_close(struct net_device *dev); 3020 void dev_close_many(struct list_head *head, bool unlink); 3021 void dev_disable_lro(struct net_device *dev); 3022 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 3023 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 3024 struct net_device *sb_dev); 3025 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 3026 struct net_device *sb_dev); 3027 3028 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3029 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3030 3031 static inline int dev_queue_xmit(struct sk_buff *skb) 3032 { 3033 return __dev_queue_xmit(skb, NULL); 3034 } 3035 3036 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3037 struct net_device *sb_dev) 3038 { 3039 return __dev_queue_xmit(skb, sb_dev); 3040 } 3041 3042 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3043 { 3044 int ret; 3045 3046 ret = __dev_direct_xmit(skb, queue_id); 3047 if (!dev_xmit_complete(ret)) 3048 kfree_skb(skb); 3049 return ret; 3050 } 3051 3052 int register_netdevice(struct net_device *dev); 3053 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3054 void unregister_netdevice_many(struct list_head *head); 3055 static inline void unregister_netdevice(struct net_device *dev) 3056 { 3057 unregister_netdevice_queue(dev, NULL); 3058 } 3059 3060 int netdev_refcnt_read(const struct net_device *dev); 3061 void free_netdev(struct net_device *dev); 3062 void netdev_freemem(struct net_device *dev); 3063 int init_dummy_netdev(struct net_device *dev); 3064 3065 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3066 struct sk_buff *skb, 3067 bool all_slaves); 3068 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3069 struct sock *sk); 3070 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3071 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3072 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3073 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 3074 int dev_restart(struct net_device *dev); 3075 3076 3077 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3078 unsigned short type, 3079 const void *daddr, const void *saddr, 3080 unsigned int len) 3081 { 3082 if (!dev->header_ops || !dev->header_ops->create) 3083 return 0; 3084 3085 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3086 } 3087 3088 static inline int dev_parse_header(const struct sk_buff *skb, 3089 unsigned char *haddr) 3090 { 3091 const struct net_device *dev = skb->dev; 3092 3093 if (!dev->header_ops || !dev->header_ops->parse) 3094 return 0; 3095 return dev->header_ops->parse(skb, haddr); 3096 } 3097 3098 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3099 { 3100 const struct net_device *dev = skb->dev; 3101 3102 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3103 return 0; 3104 return dev->header_ops->parse_protocol(skb); 3105 } 3106 3107 /* ll_header must have at least hard_header_len allocated */ 3108 static inline bool dev_validate_header(const struct net_device *dev, 3109 char *ll_header, int len) 3110 { 3111 if (likely(len >= dev->hard_header_len)) 3112 return true; 3113 if (len < dev->min_header_len) 3114 return false; 3115 3116 if (capable(CAP_SYS_RAWIO)) { 3117 memset(ll_header + len, 0, dev->hard_header_len - len); 3118 return true; 3119 } 3120 3121 if (dev->header_ops && dev->header_ops->validate) 3122 return dev->header_ops->validate(ll_header, len); 3123 3124 return false; 3125 } 3126 3127 static inline bool dev_has_header(const struct net_device *dev) 3128 { 3129 return dev->header_ops && dev->header_ops->create; 3130 } 3131 3132 /* 3133 * Incoming packets are placed on per-CPU queues 3134 */ 3135 struct softnet_data { 3136 struct list_head poll_list; 3137 struct sk_buff_head process_queue; 3138 3139 /* stats */ 3140 unsigned int processed; 3141 unsigned int time_squeeze; 3142 #ifdef CONFIG_RPS 3143 struct softnet_data *rps_ipi_list; 3144 #endif 3145 #ifdef CONFIG_NET_FLOW_LIMIT 3146 struct sd_flow_limit __rcu *flow_limit; 3147 #endif 3148 struct Qdisc *output_queue; 3149 struct Qdisc **output_queue_tailp; 3150 struct sk_buff *completion_queue; 3151 #ifdef CONFIG_XFRM_OFFLOAD 3152 struct sk_buff_head xfrm_backlog; 3153 #endif 3154 /* written and read only by owning cpu: */ 3155 struct { 3156 u16 recursion; 3157 u8 more; 3158 #ifdef CONFIG_NET_EGRESS 3159 u8 skip_txqueue; 3160 #endif 3161 } xmit; 3162 #ifdef CONFIG_RPS 3163 /* input_queue_head should be written by cpu owning this struct, 3164 * and only read by other cpus. Worth using a cache line. 3165 */ 3166 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3167 3168 /* Elements below can be accessed between CPUs for RPS/RFS */ 3169 call_single_data_t csd ____cacheline_aligned_in_smp; 3170 struct softnet_data *rps_ipi_next; 3171 unsigned int cpu; 3172 unsigned int input_queue_tail; 3173 #endif 3174 unsigned int received_rps; 3175 unsigned int dropped; 3176 struct sk_buff_head input_pkt_queue; 3177 struct napi_struct backlog; 3178 3179 /* Another possibly contended cache line */ 3180 spinlock_t defer_lock ____cacheline_aligned_in_smp; 3181 int defer_count; 3182 int defer_ipi_scheduled; 3183 struct sk_buff *defer_list; 3184 call_single_data_t defer_csd; 3185 }; 3186 3187 static inline void input_queue_head_incr(struct softnet_data *sd) 3188 { 3189 #ifdef CONFIG_RPS 3190 sd->input_queue_head++; 3191 #endif 3192 } 3193 3194 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3195 unsigned int *qtail) 3196 { 3197 #ifdef CONFIG_RPS 3198 *qtail = ++sd->input_queue_tail; 3199 #endif 3200 } 3201 3202 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3203 3204 static inline int dev_recursion_level(void) 3205 { 3206 return this_cpu_read(softnet_data.xmit.recursion); 3207 } 3208 3209 #define XMIT_RECURSION_LIMIT 8 3210 static inline bool dev_xmit_recursion(void) 3211 { 3212 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3213 XMIT_RECURSION_LIMIT); 3214 } 3215 3216 static inline void dev_xmit_recursion_inc(void) 3217 { 3218 __this_cpu_inc(softnet_data.xmit.recursion); 3219 } 3220 3221 static inline void dev_xmit_recursion_dec(void) 3222 { 3223 __this_cpu_dec(softnet_data.xmit.recursion); 3224 } 3225 3226 void __netif_schedule(struct Qdisc *q); 3227 void netif_schedule_queue(struct netdev_queue *txq); 3228 3229 static inline void netif_tx_schedule_all(struct net_device *dev) 3230 { 3231 unsigned int i; 3232 3233 for (i = 0; i < dev->num_tx_queues; i++) 3234 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3235 } 3236 3237 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3238 { 3239 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3240 } 3241 3242 /** 3243 * netif_start_queue - allow transmit 3244 * @dev: network device 3245 * 3246 * Allow upper layers to call the device hard_start_xmit routine. 3247 */ 3248 static inline void netif_start_queue(struct net_device *dev) 3249 { 3250 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3251 } 3252 3253 static inline void netif_tx_start_all_queues(struct net_device *dev) 3254 { 3255 unsigned int i; 3256 3257 for (i = 0; i < dev->num_tx_queues; i++) { 3258 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3259 netif_tx_start_queue(txq); 3260 } 3261 } 3262 3263 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3264 3265 /** 3266 * netif_wake_queue - restart transmit 3267 * @dev: network device 3268 * 3269 * Allow upper layers to call the device hard_start_xmit routine. 3270 * Used for flow control when transmit resources are available. 3271 */ 3272 static inline void netif_wake_queue(struct net_device *dev) 3273 { 3274 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3275 } 3276 3277 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3278 { 3279 unsigned int i; 3280 3281 for (i = 0; i < dev->num_tx_queues; i++) { 3282 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3283 netif_tx_wake_queue(txq); 3284 } 3285 } 3286 3287 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3288 { 3289 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3290 } 3291 3292 /** 3293 * netif_stop_queue - stop transmitted packets 3294 * @dev: network device 3295 * 3296 * Stop upper layers calling the device hard_start_xmit routine. 3297 * Used for flow control when transmit resources are unavailable. 3298 */ 3299 static inline void netif_stop_queue(struct net_device *dev) 3300 { 3301 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3302 } 3303 3304 void netif_tx_stop_all_queues(struct net_device *dev); 3305 3306 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3307 { 3308 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3309 } 3310 3311 /** 3312 * netif_queue_stopped - test if transmit queue is flowblocked 3313 * @dev: network device 3314 * 3315 * Test if transmit queue on device is currently unable to send. 3316 */ 3317 static inline bool netif_queue_stopped(const struct net_device *dev) 3318 { 3319 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3320 } 3321 3322 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3323 { 3324 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3325 } 3326 3327 static inline bool 3328 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3329 { 3330 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3331 } 3332 3333 static inline bool 3334 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3335 { 3336 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3337 } 3338 3339 /** 3340 * netdev_queue_set_dql_min_limit - set dql minimum limit 3341 * @dev_queue: pointer to transmit queue 3342 * @min_limit: dql minimum limit 3343 * 3344 * Forces xmit_more() to return true until the minimum threshold 3345 * defined by @min_limit is reached (or until the tx queue is 3346 * empty). Warning: to be use with care, misuse will impact the 3347 * latency. 3348 */ 3349 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3350 unsigned int min_limit) 3351 { 3352 #ifdef CONFIG_BQL 3353 dev_queue->dql.min_limit = min_limit; 3354 #endif 3355 } 3356 3357 /** 3358 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3359 * @dev_queue: pointer to transmit queue 3360 * 3361 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3362 * to give appropriate hint to the CPU. 3363 */ 3364 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3365 { 3366 #ifdef CONFIG_BQL 3367 prefetchw(&dev_queue->dql.num_queued); 3368 #endif 3369 } 3370 3371 /** 3372 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3373 * @dev_queue: pointer to transmit queue 3374 * 3375 * BQL enabled drivers might use this helper in their TX completion path, 3376 * to give appropriate hint to the CPU. 3377 */ 3378 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3379 { 3380 #ifdef CONFIG_BQL 3381 prefetchw(&dev_queue->dql.limit); 3382 #endif 3383 } 3384 3385 /** 3386 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3387 * @dev_queue: network device queue 3388 * @bytes: number of bytes queued to the device queue 3389 * 3390 * Report the number of bytes queued for sending/completion to the network 3391 * device hardware queue. @bytes should be a good approximation and should 3392 * exactly match netdev_completed_queue() @bytes. 3393 * This is typically called once per packet, from ndo_start_xmit(). 3394 */ 3395 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3396 unsigned int bytes) 3397 { 3398 #ifdef CONFIG_BQL 3399 dql_queued(&dev_queue->dql, bytes); 3400 3401 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3402 return; 3403 3404 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3405 3406 /* 3407 * The XOFF flag must be set before checking the dql_avail below, 3408 * because in netdev_tx_completed_queue we update the dql_completed 3409 * before checking the XOFF flag. 3410 */ 3411 smp_mb(); 3412 3413 /* check again in case another CPU has just made room avail */ 3414 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3415 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3416 #endif 3417 } 3418 3419 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3420 * that they should not test BQL status themselves. 3421 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3422 * skb of a batch. 3423 * Returns true if the doorbell must be used to kick the NIC. 3424 */ 3425 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3426 unsigned int bytes, 3427 bool xmit_more) 3428 { 3429 if (xmit_more) { 3430 #ifdef CONFIG_BQL 3431 dql_queued(&dev_queue->dql, bytes); 3432 #endif 3433 return netif_tx_queue_stopped(dev_queue); 3434 } 3435 netdev_tx_sent_queue(dev_queue, bytes); 3436 return true; 3437 } 3438 3439 /** 3440 * netdev_sent_queue - report the number of bytes queued to hardware 3441 * @dev: network device 3442 * @bytes: number of bytes queued to the hardware device queue 3443 * 3444 * Report the number of bytes queued for sending/completion to the network 3445 * device hardware queue#0. @bytes should be a good approximation and should 3446 * exactly match netdev_completed_queue() @bytes. 3447 * This is typically called once per packet, from ndo_start_xmit(). 3448 */ 3449 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3450 { 3451 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3452 } 3453 3454 static inline bool __netdev_sent_queue(struct net_device *dev, 3455 unsigned int bytes, 3456 bool xmit_more) 3457 { 3458 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3459 xmit_more); 3460 } 3461 3462 /** 3463 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3464 * @dev_queue: network device queue 3465 * @pkts: number of packets (currently ignored) 3466 * @bytes: number of bytes dequeued from the device queue 3467 * 3468 * Must be called at most once per TX completion round (and not per 3469 * individual packet), so that BQL can adjust its limits appropriately. 3470 */ 3471 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3472 unsigned int pkts, unsigned int bytes) 3473 { 3474 #ifdef CONFIG_BQL 3475 if (unlikely(!bytes)) 3476 return; 3477 3478 dql_completed(&dev_queue->dql, bytes); 3479 3480 /* 3481 * Without the memory barrier there is a small possiblity that 3482 * netdev_tx_sent_queue will miss the update and cause the queue to 3483 * be stopped forever 3484 */ 3485 smp_mb(); 3486 3487 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3488 return; 3489 3490 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3491 netif_schedule_queue(dev_queue); 3492 #endif 3493 } 3494 3495 /** 3496 * netdev_completed_queue - report bytes and packets completed by device 3497 * @dev: network device 3498 * @pkts: actual number of packets sent over the medium 3499 * @bytes: actual number of bytes sent over the medium 3500 * 3501 * Report the number of bytes and packets transmitted by the network device 3502 * hardware queue over the physical medium, @bytes must exactly match the 3503 * @bytes amount passed to netdev_sent_queue() 3504 */ 3505 static inline void netdev_completed_queue(struct net_device *dev, 3506 unsigned int pkts, unsigned int bytes) 3507 { 3508 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3509 } 3510 3511 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3512 { 3513 #ifdef CONFIG_BQL 3514 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3515 dql_reset(&q->dql); 3516 #endif 3517 } 3518 3519 /** 3520 * netdev_reset_queue - reset the packets and bytes count of a network device 3521 * @dev_queue: network device 3522 * 3523 * Reset the bytes and packet count of a network device and clear the 3524 * software flow control OFF bit for this network device 3525 */ 3526 static inline void netdev_reset_queue(struct net_device *dev_queue) 3527 { 3528 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3529 } 3530 3531 /** 3532 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3533 * @dev: network device 3534 * @queue_index: given tx queue index 3535 * 3536 * Returns 0 if given tx queue index >= number of device tx queues, 3537 * otherwise returns the originally passed tx queue index. 3538 */ 3539 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3540 { 3541 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3542 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3543 dev->name, queue_index, 3544 dev->real_num_tx_queues); 3545 return 0; 3546 } 3547 3548 return queue_index; 3549 } 3550 3551 /** 3552 * netif_running - test if up 3553 * @dev: network device 3554 * 3555 * Test if the device has been brought up. 3556 */ 3557 static inline bool netif_running(const struct net_device *dev) 3558 { 3559 return test_bit(__LINK_STATE_START, &dev->state); 3560 } 3561 3562 /* 3563 * Routines to manage the subqueues on a device. We only need start, 3564 * stop, and a check if it's stopped. All other device management is 3565 * done at the overall netdevice level. 3566 * Also test the device if we're multiqueue. 3567 */ 3568 3569 /** 3570 * netif_start_subqueue - allow sending packets on subqueue 3571 * @dev: network device 3572 * @queue_index: sub queue index 3573 * 3574 * Start individual transmit queue of a device with multiple transmit queues. 3575 */ 3576 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3577 { 3578 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3579 3580 netif_tx_start_queue(txq); 3581 } 3582 3583 /** 3584 * netif_stop_subqueue - stop sending packets on subqueue 3585 * @dev: network device 3586 * @queue_index: sub queue index 3587 * 3588 * Stop individual transmit queue of a device with multiple transmit queues. 3589 */ 3590 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3591 { 3592 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3593 netif_tx_stop_queue(txq); 3594 } 3595 3596 /** 3597 * __netif_subqueue_stopped - test status of subqueue 3598 * @dev: network device 3599 * @queue_index: sub queue index 3600 * 3601 * Check individual transmit queue of a device with multiple transmit queues. 3602 */ 3603 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3604 u16 queue_index) 3605 { 3606 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3607 3608 return netif_tx_queue_stopped(txq); 3609 } 3610 3611 /** 3612 * netif_subqueue_stopped - test status of subqueue 3613 * @dev: network device 3614 * @skb: sub queue buffer pointer 3615 * 3616 * Check individual transmit queue of a device with multiple transmit queues. 3617 */ 3618 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3619 struct sk_buff *skb) 3620 { 3621 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3622 } 3623 3624 /** 3625 * netif_wake_subqueue - allow sending packets on subqueue 3626 * @dev: network device 3627 * @queue_index: sub queue index 3628 * 3629 * Resume individual transmit queue of a device with multiple transmit queues. 3630 */ 3631 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3632 { 3633 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3634 3635 netif_tx_wake_queue(txq); 3636 } 3637 3638 #ifdef CONFIG_XPS 3639 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3640 u16 index); 3641 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3642 u16 index, enum xps_map_type type); 3643 3644 /** 3645 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3646 * @j: CPU/Rx queue index 3647 * @mask: bitmask of all cpus/rx queues 3648 * @nr_bits: number of bits in the bitmask 3649 * 3650 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3651 */ 3652 static inline bool netif_attr_test_mask(unsigned long j, 3653 const unsigned long *mask, 3654 unsigned int nr_bits) 3655 { 3656 cpu_max_bits_warn(j, nr_bits); 3657 return test_bit(j, mask); 3658 } 3659 3660 /** 3661 * netif_attr_test_online - Test for online CPU/Rx queue 3662 * @j: CPU/Rx queue index 3663 * @online_mask: bitmask for CPUs/Rx queues that are online 3664 * @nr_bits: number of bits in the bitmask 3665 * 3666 * Returns true if a CPU/Rx queue is online. 3667 */ 3668 static inline bool netif_attr_test_online(unsigned long j, 3669 const unsigned long *online_mask, 3670 unsigned int nr_bits) 3671 { 3672 cpu_max_bits_warn(j, nr_bits); 3673 3674 if (online_mask) 3675 return test_bit(j, online_mask); 3676 3677 return (j < nr_bits); 3678 } 3679 3680 /** 3681 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3682 * @n: CPU/Rx queue index 3683 * @srcp: the cpumask/Rx queue mask pointer 3684 * @nr_bits: number of bits in the bitmask 3685 * 3686 * Returns >= nr_bits if no further CPUs/Rx queues set. 3687 */ 3688 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3689 unsigned int nr_bits) 3690 { 3691 /* -1 is a legal arg here. */ 3692 if (n != -1) 3693 cpu_max_bits_warn(n, nr_bits); 3694 3695 if (srcp) 3696 return find_next_bit(srcp, nr_bits, n + 1); 3697 3698 return n + 1; 3699 } 3700 3701 /** 3702 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3703 * @n: CPU/Rx queue index 3704 * @src1p: the first CPUs/Rx queues mask pointer 3705 * @src2p: the second CPUs/Rx queues mask pointer 3706 * @nr_bits: number of bits in the bitmask 3707 * 3708 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3709 */ 3710 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3711 const unsigned long *src2p, 3712 unsigned int nr_bits) 3713 { 3714 /* -1 is a legal arg here. */ 3715 if (n != -1) 3716 cpu_max_bits_warn(n, nr_bits); 3717 3718 if (src1p && src2p) 3719 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3720 else if (src1p) 3721 return find_next_bit(src1p, nr_bits, n + 1); 3722 else if (src2p) 3723 return find_next_bit(src2p, nr_bits, n + 1); 3724 3725 return n + 1; 3726 } 3727 #else 3728 static inline int netif_set_xps_queue(struct net_device *dev, 3729 const struct cpumask *mask, 3730 u16 index) 3731 { 3732 return 0; 3733 } 3734 3735 static inline int __netif_set_xps_queue(struct net_device *dev, 3736 const unsigned long *mask, 3737 u16 index, enum xps_map_type type) 3738 { 3739 return 0; 3740 } 3741 #endif 3742 3743 /** 3744 * netif_is_multiqueue - test if device has multiple transmit queues 3745 * @dev: network device 3746 * 3747 * Check if device has multiple transmit queues 3748 */ 3749 static inline bool netif_is_multiqueue(const struct net_device *dev) 3750 { 3751 return dev->num_tx_queues > 1; 3752 } 3753 3754 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3755 3756 #ifdef CONFIG_SYSFS 3757 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3758 #else 3759 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3760 unsigned int rxqs) 3761 { 3762 dev->real_num_rx_queues = rxqs; 3763 return 0; 3764 } 3765 #endif 3766 int netif_set_real_num_queues(struct net_device *dev, 3767 unsigned int txq, unsigned int rxq); 3768 3769 static inline struct netdev_rx_queue * 3770 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3771 { 3772 return dev->_rx + rxq; 3773 } 3774 3775 #ifdef CONFIG_SYSFS 3776 static inline unsigned int get_netdev_rx_queue_index( 3777 struct netdev_rx_queue *queue) 3778 { 3779 struct net_device *dev = queue->dev; 3780 int index = queue - dev->_rx; 3781 3782 BUG_ON(index >= dev->num_rx_queues); 3783 return index; 3784 } 3785 #endif 3786 3787 int netif_get_num_default_rss_queues(void); 3788 3789 enum skb_free_reason { 3790 SKB_REASON_CONSUMED, 3791 SKB_REASON_DROPPED, 3792 }; 3793 3794 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3795 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3796 3797 /* 3798 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3799 * interrupt context or with hardware interrupts being disabled. 3800 * (in_hardirq() || irqs_disabled()) 3801 * 3802 * We provide four helpers that can be used in following contexts : 3803 * 3804 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3805 * replacing kfree_skb(skb) 3806 * 3807 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3808 * Typically used in place of consume_skb(skb) in TX completion path 3809 * 3810 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3811 * replacing kfree_skb(skb) 3812 * 3813 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3814 * and consumed a packet. Used in place of consume_skb(skb) 3815 */ 3816 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3817 { 3818 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3819 } 3820 3821 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3822 { 3823 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3824 } 3825 3826 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3827 { 3828 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3829 } 3830 3831 static inline void dev_consume_skb_any(struct sk_buff *skb) 3832 { 3833 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3834 } 3835 3836 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3837 struct bpf_prog *xdp_prog); 3838 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3839 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3840 int netif_rx(struct sk_buff *skb); 3841 int __netif_rx(struct sk_buff *skb); 3842 3843 int netif_receive_skb(struct sk_buff *skb); 3844 int netif_receive_skb_core(struct sk_buff *skb); 3845 void netif_receive_skb_list_internal(struct list_head *head); 3846 void netif_receive_skb_list(struct list_head *head); 3847 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3848 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3849 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3850 void napi_get_frags_check(struct napi_struct *napi); 3851 gro_result_t napi_gro_frags(struct napi_struct *napi); 3852 struct packet_offload *gro_find_receive_by_type(__be16 type); 3853 struct packet_offload *gro_find_complete_by_type(__be16 type); 3854 3855 static inline void napi_free_frags(struct napi_struct *napi) 3856 { 3857 kfree_skb(napi->skb); 3858 napi->skb = NULL; 3859 } 3860 3861 bool netdev_is_rx_handler_busy(struct net_device *dev); 3862 int netdev_rx_handler_register(struct net_device *dev, 3863 rx_handler_func_t *rx_handler, 3864 void *rx_handler_data); 3865 void netdev_rx_handler_unregister(struct net_device *dev); 3866 3867 bool dev_valid_name(const char *name); 3868 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 3869 { 3870 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 3871 } 3872 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 3873 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 3874 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3875 void __user *data, bool *need_copyout); 3876 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 3877 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 3878 unsigned int dev_get_flags(const struct net_device *); 3879 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3880 struct netlink_ext_ack *extack); 3881 int dev_change_flags(struct net_device *dev, unsigned int flags, 3882 struct netlink_ext_ack *extack); 3883 int dev_set_alias(struct net_device *, const char *, size_t); 3884 int dev_get_alias(const struct net_device *, char *, size_t); 3885 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 3886 const char *pat, int new_ifindex); 3887 static inline 3888 int dev_change_net_namespace(struct net_device *dev, struct net *net, 3889 const char *pat) 3890 { 3891 return __dev_change_net_namespace(dev, net, pat, 0); 3892 } 3893 int __dev_set_mtu(struct net_device *, int); 3894 int dev_set_mtu(struct net_device *, int); 3895 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3896 struct netlink_ext_ack *extack); 3897 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3898 struct netlink_ext_ack *extack); 3899 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 3900 struct netlink_ext_ack *extack); 3901 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 3902 int dev_get_port_parent_id(struct net_device *dev, 3903 struct netdev_phys_item_id *ppid, bool recurse); 3904 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3905 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3906 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3907 struct netdev_queue *txq, int *ret); 3908 3909 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3910 u8 dev_xdp_prog_count(struct net_device *dev); 3911 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3912 3913 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3914 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3915 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 3916 bool is_skb_forwardable(const struct net_device *dev, 3917 const struct sk_buff *skb); 3918 3919 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 3920 const struct sk_buff *skb, 3921 const bool check_mtu) 3922 { 3923 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 3924 unsigned int len; 3925 3926 if (!(dev->flags & IFF_UP)) 3927 return false; 3928 3929 if (!check_mtu) 3930 return true; 3931 3932 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 3933 if (skb->len <= len) 3934 return true; 3935 3936 /* if TSO is enabled, we don't care about the length as the packet 3937 * could be forwarded without being segmented before 3938 */ 3939 if (skb_is_gso(skb)) 3940 return true; 3941 3942 return false; 3943 } 3944 3945 struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev); 3946 3947 static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev) 3948 { 3949 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */ 3950 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats); 3951 3952 if (likely(p)) 3953 return p; 3954 3955 return netdev_core_stats_alloc(dev); 3956 } 3957 3958 #define DEV_CORE_STATS_INC(FIELD) \ 3959 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 3960 { \ 3961 struct net_device_core_stats __percpu *p; \ 3962 \ 3963 p = dev_core_stats(dev); \ 3964 if (p) \ 3965 this_cpu_inc(p->FIELD); \ 3966 } 3967 DEV_CORE_STATS_INC(rx_dropped) 3968 DEV_CORE_STATS_INC(tx_dropped) 3969 DEV_CORE_STATS_INC(rx_nohandler) 3970 DEV_CORE_STATS_INC(rx_otherhost_dropped) 3971 3972 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3973 struct sk_buff *skb, 3974 const bool check_mtu) 3975 { 3976 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3977 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 3978 dev_core_stats_rx_dropped_inc(dev); 3979 kfree_skb(skb); 3980 return NET_RX_DROP; 3981 } 3982 3983 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 3984 skb->priority = 0; 3985 return 0; 3986 } 3987 3988 bool dev_nit_active(struct net_device *dev); 3989 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3990 3991 static inline void __dev_put(struct net_device *dev) 3992 { 3993 if (dev) { 3994 #ifdef CONFIG_PCPU_DEV_REFCNT 3995 this_cpu_dec(*dev->pcpu_refcnt); 3996 #else 3997 refcount_dec(&dev->dev_refcnt); 3998 #endif 3999 } 4000 } 4001 4002 static inline void __dev_hold(struct net_device *dev) 4003 { 4004 if (dev) { 4005 #ifdef CONFIG_PCPU_DEV_REFCNT 4006 this_cpu_inc(*dev->pcpu_refcnt); 4007 #else 4008 refcount_inc(&dev->dev_refcnt); 4009 #endif 4010 } 4011 } 4012 4013 static inline void __netdev_tracker_alloc(struct net_device *dev, 4014 netdevice_tracker *tracker, 4015 gfp_t gfp) 4016 { 4017 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4018 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 4019 #endif 4020 } 4021 4022 /* netdev_tracker_alloc() can upgrade a prior untracked reference 4023 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4024 */ 4025 static inline void netdev_tracker_alloc(struct net_device *dev, 4026 netdevice_tracker *tracker, gfp_t gfp) 4027 { 4028 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4029 refcount_dec(&dev->refcnt_tracker.no_tracker); 4030 __netdev_tracker_alloc(dev, tracker, gfp); 4031 #endif 4032 } 4033 4034 static inline void netdev_tracker_free(struct net_device *dev, 4035 netdevice_tracker *tracker) 4036 { 4037 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4038 ref_tracker_free(&dev->refcnt_tracker, tracker); 4039 #endif 4040 } 4041 4042 static inline void netdev_hold(struct net_device *dev, 4043 netdevice_tracker *tracker, gfp_t gfp) 4044 { 4045 if (dev) { 4046 __dev_hold(dev); 4047 __netdev_tracker_alloc(dev, tracker, gfp); 4048 } 4049 } 4050 4051 static inline void netdev_put(struct net_device *dev, 4052 netdevice_tracker *tracker) 4053 { 4054 if (dev) { 4055 netdev_tracker_free(dev, tracker); 4056 __dev_put(dev); 4057 } 4058 } 4059 4060 /** 4061 * dev_hold - get reference to device 4062 * @dev: network device 4063 * 4064 * Hold reference to device to keep it from being freed. 4065 * Try using netdev_hold() instead. 4066 */ 4067 static inline void dev_hold(struct net_device *dev) 4068 { 4069 netdev_hold(dev, NULL, GFP_ATOMIC); 4070 } 4071 4072 /** 4073 * dev_put - release reference to device 4074 * @dev: network device 4075 * 4076 * Release reference to device to allow it to be freed. 4077 * Try using netdev_put() instead. 4078 */ 4079 static inline void dev_put(struct net_device *dev) 4080 { 4081 netdev_put(dev, NULL); 4082 } 4083 4084 static inline void netdev_ref_replace(struct net_device *odev, 4085 struct net_device *ndev, 4086 netdevice_tracker *tracker, 4087 gfp_t gfp) 4088 { 4089 if (odev) 4090 netdev_tracker_free(odev, tracker); 4091 4092 __dev_hold(ndev); 4093 __dev_put(odev); 4094 4095 if (ndev) 4096 __netdev_tracker_alloc(ndev, tracker, gfp); 4097 } 4098 4099 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4100 * and _off may be called from IRQ context, but it is caller 4101 * who is responsible for serialization of these calls. 4102 * 4103 * The name carrier is inappropriate, these functions should really be 4104 * called netif_lowerlayer_*() because they represent the state of any 4105 * kind of lower layer not just hardware media. 4106 */ 4107 void linkwatch_fire_event(struct net_device *dev); 4108 4109 /** 4110 * netif_carrier_ok - test if carrier present 4111 * @dev: network device 4112 * 4113 * Check if carrier is present on device 4114 */ 4115 static inline bool netif_carrier_ok(const struct net_device *dev) 4116 { 4117 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4118 } 4119 4120 unsigned long dev_trans_start(struct net_device *dev); 4121 4122 void __netdev_watchdog_up(struct net_device *dev); 4123 4124 void netif_carrier_on(struct net_device *dev); 4125 void netif_carrier_off(struct net_device *dev); 4126 void netif_carrier_event(struct net_device *dev); 4127 4128 /** 4129 * netif_dormant_on - mark device as dormant. 4130 * @dev: network device 4131 * 4132 * Mark device as dormant (as per RFC2863). 4133 * 4134 * The dormant state indicates that the relevant interface is not 4135 * actually in a condition to pass packets (i.e., it is not 'up') but is 4136 * in a "pending" state, waiting for some external event. For "on- 4137 * demand" interfaces, this new state identifies the situation where the 4138 * interface is waiting for events to place it in the up state. 4139 */ 4140 static inline void netif_dormant_on(struct net_device *dev) 4141 { 4142 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4143 linkwatch_fire_event(dev); 4144 } 4145 4146 /** 4147 * netif_dormant_off - set device as not dormant. 4148 * @dev: network device 4149 * 4150 * Device is not in dormant state. 4151 */ 4152 static inline void netif_dormant_off(struct net_device *dev) 4153 { 4154 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4155 linkwatch_fire_event(dev); 4156 } 4157 4158 /** 4159 * netif_dormant - test if device is dormant 4160 * @dev: network device 4161 * 4162 * Check if device is dormant. 4163 */ 4164 static inline bool netif_dormant(const struct net_device *dev) 4165 { 4166 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4167 } 4168 4169 4170 /** 4171 * netif_testing_on - mark device as under test. 4172 * @dev: network device 4173 * 4174 * Mark device as under test (as per RFC2863). 4175 * 4176 * The testing state indicates that some test(s) must be performed on 4177 * the interface. After completion, of the test, the interface state 4178 * will change to up, dormant, or down, as appropriate. 4179 */ 4180 static inline void netif_testing_on(struct net_device *dev) 4181 { 4182 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4183 linkwatch_fire_event(dev); 4184 } 4185 4186 /** 4187 * netif_testing_off - set device as not under test. 4188 * @dev: network device 4189 * 4190 * Device is not in testing state. 4191 */ 4192 static inline void netif_testing_off(struct net_device *dev) 4193 { 4194 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4195 linkwatch_fire_event(dev); 4196 } 4197 4198 /** 4199 * netif_testing - test if device is under test 4200 * @dev: network device 4201 * 4202 * Check if device is under test 4203 */ 4204 static inline bool netif_testing(const struct net_device *dev) 4205 { 4206 return test_bit(__LINK_STATE_TESTING, &dev->state); 4207 } 4208 4209 4210 /** 4211 * netif_oper_up - test if device is operational 4212 * @dev: network device 4213 * 4214 * Check if carrier is operational 4215 */ 4216 static inline bool netif_oper_up(const struct net_device *dev) 4217 { 4218 return (dev->operstate == IF_OPER_UP || 4219 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4220 } 4221 4222 /** 4223 * netif_device_present - is device available or removed 4224 * @dev: network device 4225 * 4226 * Check if device has not been removed from system. 4227 */ 4228 static inline bool netif_device_present(const struct net_device *dev) 4229 { 4230 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4231 } 4232 4233 void netif_device_detach(struct net_device *dev); 4234 4235 void netif_device_attach(struct net_device *dev); 4236 4237 /* 4238 * Network interface message level settings 4239 */ 4240 4241 enum { 4242 NETIF_MSG_DRV_BIT, 4243 NETIF_MSG_PROBE_BIT, 4244 NETIF_MSG_LINK_BIT, 4245 NETIF_MSG_TIMER_BIT, 4246 NETIF_MSG_IFDOWN_BIT, 4247 NETIF_MSG_IFUP_BIT, 4248 NETIF_MSG_RX_ERR_BIT, 4249 NETIF_MSG_TX_ERR_BIT, 4250 NETIF_MSG_TX_QUEUED_BIT, 4251 NETIF_MSG_INTR_BIT, 4252 NETIF_MSG_TX_DONE_BIT, 4253 NETIF_MSG_RX_STATUS_BIT, 4254 NETIF_MSG_PKTDATA_BIT, 4255 NETIF_MSG_HW_BIT, 4256 NETIF_MSG_WOL_BIT, 4257 4258 /* When you add a new bit above, update netif_msg_class_names array 4259 * in net/ethtool/common.c 4260 */ 4261 NETIF_MSG_CLASS_COUNT, 4262 }; 4263 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4264 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4265 4266 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4267 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4268 4269 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4270 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4271 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4272 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4273 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4274 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4275 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4276 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4277 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4278 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4279 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4280 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4281 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4282 #define NETIF_MSG_HW __NETIF_MSG(HW) 4283 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4284 4285 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4286 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4287 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4288 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4289 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4290 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4291 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4292 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4293 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4294 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4295 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4296 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4297 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4298 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4299 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4300 4301 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4302 { 4303 /* use default */ 4304 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4305 return default_msg_enable_bits; 4306 if (debug_value == 0) /* no output */ 4307 return 0; 4308 /* set low N bits */ 4309 return (1U << debug_value) - 1; 4310 } 4311 4312 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4313 { 4314 spin_lock(&txq->_xmit_lock); 4315 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4316 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4317 } 4318 4319 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4320 { 4321 __acquire(&txq->_xmit_lock); 4322 return true; 4323 } 4324 4325 static inline void __netif_tx_release(struct netdev_queue *txq) 4326 { 4327 __release(&txq->_xmit_lock); 4328 } 4329 4330 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4331 { 4332 spin_lock_bh(&txq->_xmit_lock); 4333 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4334 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4335 } 4336 4337 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4338 { 4339 bool ok = spin_trylock(&txq->_xmit_lock); 4340 4341 if (likely(ok)) { 4342 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4343 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4344 } 4345 return ok; 4346 } 4347 4348 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4349 { 4350 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4351 WRITE_ONCE(txq->xmit_lock_owner, -1); 4352 spin_unlock(&txq->_xmit_lock); 4353 } 4354 4355 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4356 { 4357 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4358 WRITE_ONCE(txq->xmit_lock_owner, -1); 4359 spin_unlock_bh(&txq->_xmit_lock); 4360 } 4361 4362 /* 4363 * txq->trans_start can be read locklessly from dev_watchdog() 4364 */ 4365 static inline void txq_trans_update(struct netdev_queue *txq) 4366 { 4367 if (txq->xmit_lock_owner != -1) 4368 WRITE_ONCE(txq->trans_start, jiffies); 4369 } 4370 4371 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4372 { 4373 unsigned long now = jiffies; 4374 4375 if (READ_ONCE(txq->trans_start) != now) 4376 WRITE_ONCE(txq->trans_start, now); 4377 } 4378 4379 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4380 static inline void netif_trans_update(struct net_device *dev) 4381 { 4382 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4383 4384 txq_trans_cond_update(txq); 4385 } 4386 4387 /** 4388 * netif_tx_lock - grab network device transmit lock 4389 * @dev: network device 4390 * 4391 * Get network device transmit lock 4392 */ 4393 void netif_tx_lock(struct net_device *dev); 4394 4395 static inline void netif_tx_lock_bh(struct net_device *dev) 4396 { 4397 local_bh_disable(); 4398 netif_tx_lock(dev); 4399 } 4400 4401 void netif_tx_unlock(struct net_device *dev); 4402 4403 static inline void netif_tx_unlock_bh(struct net_device *dev) 4404 { 4405 netif_tx_unlock(dev); 4406 local_bh_enable(); 4407 } 4408 4409 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4410 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4411 __netif_tx_lock(txq, cpu); \ 4412 } else { \ 4413 __netif_tx_acquire(txq); \ 4414 } \ 4415 } 4416 4417 #define HARD_TX_TRYLOCK(dev, txq) \ 4418 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4419 __netif_tx_trylock(txq) : \ 4420 __netif_tx_acquire(txq)) 4421 4422 #define HARD_TX_UNLOCK(dev, txq) { \ 4423 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4424 __netif_tx_unlock(txq); \ 4425 } else { \ 4426 __netif_tx_release(txq); \ 4427 } \ 4428 } 4429 4430 static inline void netif_tx_disable(struct net_device *dev) 4431 { 4432 unsigned int i; 4433 int cpu; 4434 4435 local_bh_disable(); 4436 cpu = smp_processor_id(); 4437 spin_lock(&dev->tx_global_lock); 4438 for (i = 0; i < dev->num_tx_queues; i++) { 4439 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4440 4441 __netif_tx_lock(txq, cpu); 4442 netif_tx_stop_queue(txq); 4443 __netif_tx_unlock(txq); 4444 } 4445 spin_unlock(&dev->tx_global_lock); 4446 local_bh_enable(); 4447 } 4448 4449 static inline void netif_addr_lock(struct net_device *dev) 4450 { 4451 unsigned char nest_level = 0; 4452 4453 #ifdef CONFIG_LOCKDEP 4454 nest_level = dev->nested_level; 4455 #endif 4456 spin_lock_nested(&dev->addr_list_lock, nest_level); 4457 } 4458 4459 static inline void netif_addr_lock_bh(struct net_device *dev) 4460 { 4461 unsigned char nest_level = 0; 4462 4463 #ifdef CONFIG_LOCKDEP 4464 nest_level = dev->nested_level; 4465 #endif 4466 local_bh_disable(); 4467 spin_lock_nested(&dev->addr_list_lock, nest_level); 4468 } 4469 4470 static inline void netif_addr_unlock(struct net_device *dev) 4471 { 4472 spin_unlock(&dev->addr_list_lock); 4473 } 4474 4475 static inline void netif_addr_unlock_bh(struct net_device *dev) 4476 { 4477 spin_unlock_bh(&dev->addr_list_lock); 4478 } 4479 4480 /* 4481 * dev_addrs walker. Should be used only for read access. Call with 4482 * rcu_read_lock held. 4483 */ 4484 #define for_each_dev_addr(dev, ha) \ 4485 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4486 4487 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4488 4489 void ether_setup(struct net_device *dev); 4490 4491 /* Support for loadable net-drivers */ 4492 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4493 unsigned char name_assign_type, 4494 void (*setup)(struct net_device *), 4495 unsigned int txqs, unsigned int rxqs); 4496 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4497 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4498 4499 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4500 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4501 count) 4502 4503 int register_netdev(struct net_device *dev); 4504 void unregister_netdev(struct net_device *dev); 4505 4506 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4507 4508 /* General hardware address lists handling functions */ 4509 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4510 struct netdev_hw_addr_list *from_list, int addr_len); 4511 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4512 struct netdev_hw_addr_list *from_list, int addr_len); 4513 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4514 struct net_device *dev, 4515 int (*sync)(struct net_device *, const unsigned char *), 4516 int (*unsync)(struct net_device *, 4517 const unsigned char *)); 4518 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4519 struct net_device *dev, 4520 int (*sync)(struct net_device *, 4521 const unsigned char *, int), 4522 int (*unsync)(struct net_device *, 4523 const unsigned char *, int)); 4524 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4525 struct net_device *dev, 4526 int (*unsync)(struct net_device *, 4527 const unsigned char *, int)); 4528 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4529 struct net_device *dev, 4530 int (*unsync)(struct net_device *, 4531 const unsigned char *)); 4532 void __hw_addr_init(struct netdev_hw_addr_list *list); 4533 4534 /* Functions used for device addresses handling */ 4535 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4536 const void *addr, size_t len); 4537 4538 static inline void 4539 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4540 { 4541 dev_addr_mod(dev, 0, addr, len); 4542 } 4543 4544 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4545 { 4546 __dev_addr_set(dev, addr, dev->addr_len); 4547 } 4548 4549 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4550 unsigned char addr_type); 4551 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4552 unsigned char addr_type); 4553 4554 /* Functions used for unicast addresses handling */ 4555 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4556 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4557 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4558 int dev_uc_sync(struct net_device *to, struct net_device *from); 4559 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4560 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4561 void dev_uc_flush(struct net_device *dev); 4562 void dev_uc_init(struct net_device *dev); 4563 4564 /** 4565 * __dev_uc_sync - Synchonize device's unicast list 4566 * @dev: device to sync 4567 * @sync: function to call if address should be added 4568 * @unsync: function to call if address should be removed 4569 * 4570 * Add newly added addresses to the interface, and release 4571 * addresses that have been deleted. 4572 */ 4573 static inline int __dev_uc_sync(struct net_device *dev, 4574 int (*sync)(struct net_device *, 4575 const unsigned char *), 4576 int (*unsync)(struct net_device *, 4577 const unsigned char *)) 4578 { 4579 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4580 } 4581 4582 /** 4583 * __dev_uc_unsync - Remove synchronized addresses from device 4584 * @dev: device to sync 4585 * @unsync: function to call if address should be removed 4586 * 4587 * Remove all addresses that were added to the device by dev_uc_sync(). 4588 */ 4589 static inline void __dev_uc_unsync(struct net_device *dev, 4590 int (*unsync)(struct net_device *, 4591 const unsigned char *)) 4592 { 4593 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4594 } 4595 4596 /* Functions used for multicast addresses handling */ 4597 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4598 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4599 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4600 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4601 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4602 int dev_mc_sync(struct net_device *to, struct net_device *from); 4603 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4604 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4605 void dev_mc_flush(struct net_device *dev); 4606 void dev_mc_init(struct net_device *dev); 4607 4608 /** 4609 * __dev_mc_sync - Synchonize device's multicast list 4610 * @dev: device to sync 4611 * @sync: function to call if address should be added 4612 * @unsync: function to call if address should be removed 4613 * 4614 * Add newly added addresses to the interface, and release 4615 * addresses that have been deleted. 4616 */ 4617 static inline int __dev_mc_sync(struct net_device *dev, 4618 int (*sync)(struct net_device *, 4619 const unsigned char *), 4620 int (*unsync)(struct net_device *, 4621 const unsigned char *)) 4622 { 4623 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4624 } 4625 4626 /** 4627 * __dev_mc_unsync - Remove synchronized addresses from device 4628 * @dev: device to sync 4629 * @unsync: function to call if address should be removed 4630 * 4631 * Remove all addresses that were added to the device by dev_mc_sync(). 4632 */ 4633 static inline void __dev_mc_unsync(struct net_device *dev, 4634 int (*unsync)(struct net_device *, 4635 const unsigned char *)) 4636 { 4637 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4638 } 4639 4640 /* Functions used for secondary unicast and multicast support */ 4641 void dev_set_rx_mode(struct net_device *dev); 4642 int dev_set_promiscuity(struct net_device *dev, int inc); 4643 int dev_set_allmulti(struct net_device *dev, int inc); 4644 void netdev_state_change(struct net_device *dev); 4645 void __netdev_notify_peers(struct net_device *dev); 4646 void netdev_notify_peers(struct net_device *dev); 4647 void netdev_features_change(struct net_device *dev); 4648 /* Load a device via the kmod */ 4649 void dev_load(struct net *net, const char *name); 4650 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4651 struct rtnl_link_stats64 *storage); 4652 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4653 const struct net_device_stats *netdev_stats); 4654 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4655 const struct pcpu_sw_netstats __percpu *netstats); 4656 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4657 4658 extern int netdev_max_backlog; 4659 extern int dev_rx_weight; 4660 extern int dev_tx_weight; 4661 extern int gro_normal_batch; 4662 4663 enum { 4664 NESTED_SYNC_IMM_BIT, 4665 NESTED_SYNC_TODO_BIT, 4666 }; 4667 4668 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4669 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4670 4671 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4672 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4673 4674 struct netdev_nested_priv { 4675 unsigned char flags; 4676 void *data; 4677 }; 4678 4679 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4680 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4681 struct list_head **iter); 4682 4683 /* iterate through upper list, must be called under RCU read lock */ 4684 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4685 for (iter = &(dev)->adj_list.upper, \ 4686 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4687 updev; \ 4688 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4689 4690 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4691 int (*fn)(struct net_device *upper_dev, 4692 struct netdev_nested_priv *priv), 4693 struct netdev_nested_priv *priv); 4694 4695 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4696 struct net_device *upper_dev); 4697 4698 bool netdev_has_any_upper_dev(struct net_device *dev); 4699 4700 void *netdev_lower_get_next_private(struct net_device *dev, 4701 struct list_head **iter); 4702 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4703 struct list_head **iter); 4704 4705 #define netdev_for_each_lower_private(dev, priv, iter) \ 4706 for (iter = (dev)->adj_list.lower.next, \ 4707 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4708 priv; \ 4709 priv = netdev_lower_get_next_private(dev, &(iter))) 4710 4711 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4712 for (iter = &(dev)->adj_list.lower, \ 4713 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4714 priv; \ 4715 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4716 4717 void *netdev_lower_get_next(struct net_device *dev, 4718 struct list_head **iter); 4719 4720 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4721 for (iter = (dev)->adj_list.lower.next, \ 4722 ldev = netdev_lower_get_next(dev, &(iter)); \ 4723 ldev; \ 4724 ldev = netdev_lower_get_next(dev, &(iter))) 4725 4726 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4727 struct list_head **iter); 4728 int netdev_walk_all_lower_dev(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 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4733 int (*fn)(struct net_device *lower_dev, 4734 struct netdev_nested_priv *priv), 4735 struct netdev_nested_priv *priv); 4736 4737 void *netdev_adjacent_get_private(struct list_head *adj_list); 4738 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4739 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4740 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4741 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4742 struct netlink_ext_ack *extack); 4743 int netdev_master_upper_dev_link(struct net_device *dev, 4744 struct net_device *upper_dev, 4745 void *upper_priv, void *upper_info, 4746 struct netlink_ext_ack *extack); 4747 void netdev_upper_dev_unlink(struct net_device *dev, 4748 struct net_device *upper_dev); 4749 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4750 struct net_device *new_dev, 4751 struct net_device *dev, 4752 struct netlink_ext_ack *extack); 4753 void netdev_adjacent_change_commit(struct net_device *old_dev, 4754 struct net_device *new_dev, 4755 struct net_device *dev); 4756 void netdev_adjacent_change_abort(struct net_device *old_dev, 4757 struct net_device *new_dev, 4758 struct net_device *dev); 4759 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4760 void *netdev_lower_dev_get_private(struct net_device *dev, 4761 struct net_device *lower_dev); 4762 void netdev_lower_state_changed(struct net_device *lower_dev, 4763 void *lower_state_info); 4764 4765 /* RSS keys are 40 or 52 bytes long */ 4766 #define NETDEV_RSS_KEY_LEN 52 4767 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4768 void netdev_rss_key_fill(void *buffer, size_t len); 4769 4770 int skb_checksum_help(struct sk_buff *skb); 4771 int skb_crc32c_csum_help(struct sk_buff *skb); 4772 int skb_csum_hwoffload_help(struct sk_buff *skb, 4773 const netdev_features_t features); 4774 4775 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4776 netdev_features_t features, bool tx_path); 4777 struct sk_buff *skb_eth_gso_segment(struct sk_buff *skb, 4778 netdev_features_t features, __be16 type); 4779 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4780 netdev_features_t features); 4781 4782 struct netdev_bonding_info { 4783 ifslave slave; 4784 ifbond master; 4785 }; 4786 4787 struct netdev_notifier_bonding_info { 4788 struct netdev_notifier_info info; /* must be first */ 4789 struct netdev_bonding_info bonding_info; 4790 }; 4791 4792 void netdev_bonding_info_change(struct net_device *dev, 4793 struct netdev_bonding_info *bonding_info); 4794 4795 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4796 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4797 #else 4798 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4799 const void *data) 4800 { 4801 } 4802 #endif 4803 4804 static inline 4805 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4806 { 4807 return __skb_gso_segment(skb, features, true); 4808 } 4809 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4810 4811 static inline bool can_checksum_protocol(netdev_features_t features, 4812 __be16 protocol) 4813 { 4814 if (protocol == htons(ETH_P_FCOE)) 4815 return !!(features & NETIF_F_FCOE_CRC); 4816 4817 /* Assume this is an IP checksum (not SCTP CRC) */ 4818 4819 if (features & NETIF_F_HW_CSUM) { 4820 /* Can checksum everything */ 4821 return true; 4822 } 4823 4824 switch (protocol) { 4825 case htons(ETH_P_IP): 4826 return !!(features & NETIF_F_IP_CSUM); 4827 case htons(ETH_P_IPV6): 4828 return !!(features & NETIF_F_IPV6_CSUM); 4829 default: 4830 return false; 4831 } 4832 } 4833 4834 #ifdef CONFIG_BUG 4835 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4836 #else 4837 static inline void netdev_rx_csum_fault(struct net_device *dev, 4838 struct sk_buff *skb) 4839 { 4840 } 4841 #endif 4842 /* rx skb timestamps */ 4843 void net_enable_timestamp(void); 4844 void net_disable_timestamp(void); 4845 4846 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 4847 const struct skb_shared_hwtstamps *hwtstamps, 4848 bool cycles) 4849 { 4850 const struct net_device_ops *ops = dev->netdev_ops; 4851 4852 if (ops->ndo_get_tstamp) 4853 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 4854 4855 return hwtstamps->hwtstamp; 4856 } 4857 4858 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4859 struct sk_buff *skb, struct net_device *dev, 4860 bool more) 4861 { 4862 __this_cpu_write(softnet_data.xmit.more, more); 4863 return ops->ndo_start_xmit(skb, dev); 4864 } 4865 4866 static inline bool netdev_xmit_more(void) 4867 { 4868 return __this_cpu_read(softnet_data.xmit.more); 4869 } 4870 4871 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4872 struct netdev_queue *txq, bool more) 4873 { 4874 const struct net_device_ops *ops = dev->netdev_ops; 4875 netdev_tx_t rc; 4876 4877 rc = __netdev_start_xmit(ops, skb, dev, more); 4878 if (rc == NETDEV_TX_OK) 4879 txq_trans_update(txq); 4880 4881 return rc; 4882 } 4883 4884 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4885 const void *ns); 4886 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4887 const void *ns); 4888 4889 extern const struct kobj_ns_type_operations net_ns_type_operations; 4890 4891 const char *netdev_drivername(const struct net_device *dev); 4892 4893 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4894 netdev_features_t f2) 4895 { 4896 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4897 if (f1 & NETIF_F_HW_CSUM) 4898 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4899 else 4900 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4901 } 4902 4903 return f1 & f2; 4904 } 4905 4906 static inline netdev_features_t netdev_get_wanted_features( 4907 struct net_device *dev) 4908 { 4909 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4910 } 4911 netdev_features_t netdev_increment_features(netdev_features_t all, 4912 netdev_features_t one, netdev_features_t mask); 4913 4914 /* Allow TSO being used on stacked device : 4915 * Performing the GSO segmentation before last device 4916 * is a performance improvement. 4917 */ 4918 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4919 netdev_features_t mask) 4920 { 4921 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4922 } 4923 4924 int __netdev_update_features(struct net_device *dev); 4925 void netdev_update_features(struct net_device *dev); 4926 void netdev_change_features(struct net_device *dev); 4927 4928 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4929 struct net_device *dev); 4930 4931 netdev_features_t passthru_features_check(struct sk_buff *skb, 4932 struct net_device *dev, 4933 netdev_features_t features); 4934 netdev_features_t netif_skb_features(struct sk_buff *skb); 4935 4936 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4937 { 4938 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4939 4940 /* check flags correspondence */ 4941 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4942 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4943 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4944 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4945 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4946 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4947 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4948 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4949 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4950 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4951 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4952 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4953 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4954 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4955 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4956 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4957 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4958 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4959 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4960 4961 return (features & feature) == feature; 4962 } 4963 4964 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4965 { 4966 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4967 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4968 } 4969 4970 static inline bool netif_needs_gso(struct sk_buff *skb, 4971 netdev_features_t features) 4972 { 4973 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4974 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4975 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4976 } 4977 4978 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 4979 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 4980 void netif_inherit_tso_max(struct net_device *to, 4981 const struct net_device *from); 4982 4983 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4984 int pulled_hlen, u16 mac_offset, 4985 int mac_len) 4986 { 4987 skb->protocol = protocol; 4988 skb->encapsulation = 1; 4989 skb_push(skb, pulled_hlen); 4990 skb_reset_transport_header(skb); 4991 skb->mac_header = mac_offset; 4992 skb->network_header = skb->mac_header + mac_len; 4993 skb->mac_len = mac_len; 4994 } 4995 4996 static inline bool netif_is_macsec(const struct net_device *dev) 4997 { 4998 return dev->priv_flags & IFF_MACSEC; 4999 } 5000 5001 static inline bool netif_is_macvlan(const struct net_device *dev) 5002 { 5003 return dev->priv_flags & IFF_MACVLAN; 5004 } 5005 5006 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5007 { 5008 return dev->priv_flags & IFF_MACVLAN_PORT; 5009 } 5010 5011 static inline bool netif_is_bond_master(const struct net_device *dev) 5012 { 5013 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5014 } 5015 5016 static inline bool netif_is_bond_slave(const struct net_device *dev) 5017 { 5018 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5019 } 5020 5021 static inline bool netif_supports_nofcs(struct net_device *dev) 5022 { 5023 return dev->priv_flags & IFF_SUPP_NOFCS; 5024 } 5025 5026 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5027 { 5028 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5029 } 5030 5031 static inline bool netif_is_l3_master(const struct net_device *dev) 5032 { 5033 return dev->priv_flags & IFF_L3MDEV_MASTER; 5034 } 5035 5036 static inline bool netif_is_l3_slave(const struct net_device *dev) 5037 { 5038 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5039 } 5040 5041 static inline bool netif_is_bridge_master(const struct net_device *dev) 5042 { 5043 return dev->priv_flags & IFF_EBRIDGE; 5044 } 5045 5046 static inline bool netif_is_bridge_port(const struct net_device *dev) 5047 { 5048 return dev->priv_flags & IFF_BRIDGE_PORT; 5049 } 5050 5051 static inline bool netif_is_ovs_master(const struct net_device *dev) 5052 { 5053 return dev->priv_flags & IFF_OPENVSWITCH; 5054 } 5055 5056 static inline bool netif_is_ovs_port(const struct net_device *dev) 5057 { 5058 return dev->priv_flags & IFF_OVS_DATAPATH; 5059 } 5060 5061 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5062 { 5063 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5064 } 5065 5066 static inline bool netif_is_team_master(const struct net_device *dev) 5067 { 5068 return dev->priv_flags & IFF_TEAM; 5069 } 5070 5071 static inline bool netif_is_team_port(const struct net_device *dev) 5072 { 5073 return dev->priv_flags & IFF_TEAM_PORT; 5074 } 5075 5076 static inline bool netif_is_lag_master(const struct net_device *dev) 5077 { 5078 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5079 } 5080 5081 static inline bool netif_is_lag_port(const struct net_device *dev) 5082 { 5083 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5084 } 5085 5086 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5087 { 5088 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5089 } 5090 5091 static inline bool netif_is_failover(const struct net_device *dev) 5092 { 5093 return dev->priv_flags & IFF_FAILOVER; 5094 } 5095 5096 static inline bool netif_is_failover_slave(const struct net_device *dev) 5097 { 5098 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5099 } 5100 5101 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5102 static inline void netif_keep_dst(struct net_device *dev) 5103 { 5104 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5105 } 5106 5107 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5108 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5109 { 5110 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5111 return netif_is_macsec(dev); 5112 } 5113 5114 extern struct pernet_operations __net_initdata loopback_net_ops; 5115 5116 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5117 5118 /* netdev_printk helpers, similar to dev_printk */ 5119 5120 static inline const char *netdev_name(const struct net_device *dev) 5121 { 5122 if (!dev->name[0] || strchr(dev->name, '%')) 5123 return "(unnamed net_device)"; 5124 return dev->name; 5125 } 5126 5127 static inline const char *netdev_reg_state(const struct net_device *dev) 5128 { 5129 switch (dev->reg_state) { 5130 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5131 case NETREG_REGISTERED: return ""; 5132 case NETREG_UNREGISTERING: return " (unregistering)"; 5133 case NETREG_UNREGISTERED: return " (unregistered)"; 5134 case NETREG_RELEASED: return " (released)"; 5135 case NETREG_DUMMY: return " (dummy)"; 5136 } 5137 5138 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5139 return " (unknown)"; 5140 } 5141 5142 #define MODULE_ALIAS_NETDEV(device) \ 5143 MODULE_ALIAS("netdev-" device) 5144 5145 /* 5146 * netdev_WARN() acts like dev_printk(), but with the key difference 5147 * of using a WARN/WARN_ON to get the message out, including the 5148 * file/line information and a backtrace. 5149 */ 5150 #define netdev_WARN(dev, format, args...) \ 5151 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5152 netdev_reg_state(dev), ##args) 5153 5154 #define netdev_WARN_ONCE(dev, format, args...) \ 5155 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5156 netdev_reg_state(dev), ##args) 5157 5158 /* 5159 * The list of packet types we will receive (as opposed to discard) 5160 * and the routines to invoke. 5161 * 5162 * Why 16. Because with 16 the only overlap we get on a hash of the 5163 * low nibble of the protocol value is RARP/SNAP/X.25. 5164 * 5165 * 0800 IP 5166 * 0001 802.3 5167 * 0002 AX.25 5168 * 0004 802.2 5169 * 8035 RARP 5170 * 0005 SNAP 5171 * 0805 X.25 5172 * 0806 ARP 5173 * 8137 IPX 5174 * 0009 Localtalk 5175 * 86DD IPv6 5176 */ 5177 #define PTYPE_HASH_SIZE (16) 5178 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5179 5180 extern struct list_head ptype_all __read_mostly; 5181 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5182 5183 extern struct net_device *blackhole_netdev; 5184 5185 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */ 5186 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD) 5187 #define DEV_STATS_ADD(DEV, FIELD, VAL) \ 5188 atomic_long_add((VAL), &(DEV)->stats.__##FIELD) 5189 5190 #endif /* _LINUX_NETDEVICE_H */ 5191