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 <linux/netdev_features.h> 44 #include <linux/neighbour.h> 45 #include <linux/netdevice_xmit.h> 46 #include <uapi/linux/netdevice.h> 47 #include <uapi/linux/if_bonding.h> 48 #include <uapi/linux/pkt_cls.h> 49 #include <uapi/linux/netdev.h> 50 #include <linux/hashtable.h> 51 #include <linux/rbtree.h> 52 #include <net/net_trackers.h> 53 #include <net/net_debug.h> 54 #include <net/dropreason-core.h> 55 #include <net/neighbour_tables.h> 56 57 struct netpoll_info; 58 struct device; 59 struct ethtool_ops; 60 struct kernel_hwtstamp_config; 61 struct phy_device; 62 struct dsa_port; 63 struct ip_tunnel_parm_kern; 64 struct macsec_context; 65 struct macsec_ops; 66 struct netdev_name_node; 67 struct sd_flow_limit; 68 struct sfp_bus; 69 /* 802.11 specific */ 70 struct wireless_dev; 71 /* 802.15.4 specific */ 72 struct wpan_dev; 73 struct mpls_dev; 74 /* UDP Tunnel offloads */ 75 struct udp_tunnel_info; 76 struct udp_tunnel_nic_info; 77 struct udp_tunnel_nic; 78 struct bpf_prog; 79 struct xdp_buff; 80 struct xdp_frame; 81 struct xdp_metadata_ops; 82 struct xdp_md; 83 struct ethtool_netdev_state; 84 struct phy_link_topology; 85 struct hwtstamp_provider; 86 87 typedef u32 xdp_features_t; 88 89 void synchronize_net(void); 90 void netdev_set_default_ethtool_ops(struct net_device *dev, 91 const struct ethtool_ops *ops); 92 void netdev_sw_irq_coalesce_default_on(struct net_device *dev); 93 94 /* Backlog congestion levels */ 95 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 96 #define NET_RX_DROP 1 /* packet dropped */ 97 98 #define MAX_NEST_DEV 8 99 100 /* 101 * Transmit return codes: transmit return codes originate from three different 102 * namespaces: 103 * 104 * - qdisc return codes 105 * - driver transmit return codes 106 * - errno values 107 * 108 * Drivers are allowed to return any one of those in their hard_start_xmit() 109 * function. Real network devices commonly used with qdiscs should only return 110 * the driver transmit return codes though - when qdiscs are used, the actual 111 * transmission happens asynchronously, so the value is not propagated to 112 * higher layers. Virtual network devices transmit synchronously; in this case 113 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 114 * others are propagated to higher layers. 115 */ 116 117 /* qdisc ->enqueue() return codes. */ 118 #define NET_XMIT_SUCCESS 0x00 119 #define NET_XMIT_DROP 0x01 /* skb dropped */ 120 #define NET_XMIT_CN 0x02 /* congestion notification */ 121 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 122 123 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 124 * indicates that the device will soon be dropping packets, or already drops 125 * some packets of the same priority; prompting us to send less aggressively. */ 126 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 127 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 128 129 /* Driver transmit return codes */ 130 #define NETDEV_TX_MASK 0xf0 131 132 enum netdev_tx { 133 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 134 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 135 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 136 }; 137 typedef enum netdev_tx netdev_tx_t; 138 139 /* 140 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 141 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 142 */ 143 static inline bool dev_xmit_complete(int rc) 144 { 145 /* 146 * Positive cases with an skb consumed by a driver: 147 * - successful transmission (rc == NETDEV_TX_OK) 148 * - error while transmitting (rc < 0) 149 * - error while queueing to a different device (rc & NET_XMIT_MASK) 150 */ 151 if (likely(rc < NET_XMIT_MASK)) 152 return true; 153 154 return false; 155 } 156 157 /* 158 * Compute the worst-case header length according to the protocols 159 * used. 160 */ 161 162 #if defined(CONFIG_HYPERV_NET) 163 # define LL_MAX_HEADER 128 164 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 165 # if defined(CONFIG_MAC80211_MESH) 166 # define LL_MAX_HEADER 128 167 # else 168 # define LL_MAX_HEADER 96 169 # endif 170 #else 171 # define LL_MAX_HEADER 32 172 #endif 173 174 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 175 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 176 #define MAX_HEADER LL_MAX_HEADER 177 #else 178 #define MAX_HEADER (LL_MAX_HEADER + 48) 179 #endif 180 181 /* 182 * Old network device statistics. Fields are native words 183 * (unsigned long) so they can be read and written atomically. 184 */ 185 186 #define NET_DEV_STAT(FIELD) \ 187 union { \ 188 unsigned long FIELD; \ 189 atomic_long_t __##FIELD; \ 190 } 191 192 struct net_device_stats { 193 NET_DEV_STAT(rx_packets); 194 NET_DEV_STAT(tx_packets); 195 NET_DEV_STAT(rx_bytes); 196 NET_DEV_STAT(tx_bytes); 197 NET_DEV_STAT(rx_errors); 198 NET_DEV_STAT(tx_errors); 199 NET_DEV_STAT(rx_dropped); 200 NET_DEV_STAT(tx_dropped); 201 NET_DEV_STAT(multicast); 202 NET_DEV_STAT(collisions); 203 NET_DEV_STAT(rx_length_errors); 204 NET_DEV_STAT(rx_over_errors); 205 NET_DEV_STAT(rx_crc_errors); 206 NET_DEV_STAT(rx_frame_errors); 207 NET_DEV_STAT(rx_fifo_errors); 208 NET_DEV_STAT(rx_missed_errors); 209 NET_DEV_STAT(tx_aborted_errors); 210 NET_DEV_STAT(tx_carrier_errors); 211 NET_DEV_STAT(tx_fifo_errors); 212 NET_DEV_STAT(tx_heartbeat_errors); 213 NET_DEV_STAT(tx_window_errors); 214 NET_DEV_STAT(rx_compressed); 215 NET_DEV_STAT(tx_compressed); 216 }; 217 #undef NET_DEV_STAT 218 219 /* per-cpu stats, allocated on demand. 220 * Try to fit them in a single cache line, for dev_get_stats() sake. 221 */ 222 struct net_device_core_stats { 223 unsigned long rx_dropped; 224 unsigned long tx_dropped; 225 unsigned long rx_nohandler; 226 unsigned long rx_otherhost_dropped; 227 } __aligned(4 * sizeof(unsigned long)); 228 229 #include <linux/cache.h> 230 #include <linux/skbuff.h> 231 232 struct neighbour; 233 struct neigh_parms; 234 struct sk_buff; 235 236 struct netdev_hw_addr { 237 struct list_head list; 238 struct rb_node node; 239 unsigned char addr[MAX_ADDR_LEN]; 240 unsigned char type; 241 #define NETDEV_HW_ADDR_T_LAN 1 242 #define NETDEV_HW_ADDR_T_SAN 2 243 #define NETDEV_HW_ADDR_T_UNICAST 3 244 #define NETDEV_HW_ADDR_T_MULTICAST 4 245 bool global_use; 246 int sync_cnt; 247 int refcount; 248 int synced; 249 struct rcu_head rcu_head; 250 }; 251 252 struct netdev_hw_addr_list { 253 struct list_head list; 254 int count; 255 256 /* Auxiliary tree for faster lookup on addition and deletion */ 257 struct rb_root tree; 258 }; 259 260 #define netdev_hw_addr_list_count(l) ((l)->count) 261 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 262 #define netdev_hw_addr_list_for_each(ha, l) \ 263 list_for_each_entry(ha, &(l)->list, list) 264 265 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 266 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 267 #define netdev_for_each_uc_addr(ha, dev) \ 268 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 269 #define netdev_for_each_synced_uc_addr(_ha, _dev) \ 270 netdev_for_each_uc_addr((_ha), (_dev)) \ 271 if ((_ha)->sync_cnt) 272 273 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 274 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 275 #define netdev_for_each_mc_addr(ha, dev) \ 276 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 277 #define netdev_for_each_synced_mc_addr(_ha, _dev) \ 278 netdev_for_each_mc_addr((_ha), (_dev)) \ 279 if ((_ha)->sync_cnt) 280 281 struct hh_cache { 282 unsigned int hh_len; 283 seqlock_t hh_lock; 284 285 /* cached hardware header; allow for machine alignment needs. */ 286 #define HH_DATA_MOD 16 287 #define HH_DATA_OFF(__len) \ 288 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 289 #define HH_DATA_ALIGN(__len) \ 290 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 291 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 292 }; 293 294 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 295 * Alternative is: 296 * dev->hard_header_len ? (dev->hard_header_len + 297 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 298 * 299 * We could use other alignment values, but we must maintain the 300 * relationship HH alignment <= LL alignment. 301 */ 302 #define LL_RESERVED_SPACE(dev) \ 303 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \ 304 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 305 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 306 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \ 307 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 308 309 struct header_ops { 310 int (*create) (struct sk_buff *skb, struct net_device *dev, 311 unsigned short type, const void *daddr, 312 const void *saddr, unsigned int len); 313 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 314 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 315 void (*cache_update)(struct hh_cache *hh, 316 const struct net_device *dev, 317 const unsigned char *haddr); 318 bool (*validate)(const char *ll_header, unsigned int len); 319 __be16 (*parse_protocol)(const struct sk_buff *skb); 320 }; 321 322 /* These flag bits are private to the generic network queueing 323 * layer; they may not be explicitly referenced by any other 324 * code. 325 */ 326 327 enum netdev_state_t { 328 __LINK_STATE_START, 329 __LINK_STATE_PRESENT, 330 __LINK_STATE_NOCARRIER, 331 __LINK_STATE_LINKWATCH_PENDING, 332 __LINK_STATE_DORMANT, 333 __LINK_STATE_TESTING, 334 }; 335 336 struct gro_list { 337 struct list_head list; 338 int count; 339 }; 340 341 /* 342 * size of gro hash buckets, must less than bit number of 343 * napi_struct::gro_bitmask 344 */ 345 #define GRO_HASH_BUCKETS 8 346 347 /* 348 * Structure for per-NAPI config 349 */ 350 struct napi_config { 351 u64 gro_flush_timeout; 352 u64 irq_suspend_timeout; 353 u32 defer_hard_irqs; 354 unsigned int napi_id; 355 }; 356 357 /* 358 * Structure for NAPI scheduling similar to tasklet but with weighting 359 */ 360 struct napi_struct { 361 /* The poll_list must only be managed by the entity which 362 * changes the state of the NAPI_STATE_SCHED bit. This means 363 * whoever atomically sets that bit can add this napi_struct 364 * to the per-CPU poll_list, and whoever clears that bit 365 * can remove from the list right before clearing the bit. 366 */ 367 struct list_head poll_list; 368 369 unsigned long state; 370 int weight; 371 u32 defer_hard_irqs_count; 372 unsigned long gro_bitmask; 373 int (*poll)(struct napi_struct *, int); 374 #ifdef CONFIG_NETPOLL 375 /* CPU actively polling if netpoll is configured */ 376 int poll_owner; 377 #endif 378 /* CPU on which NAPI has been scheduled for processing */ 379 int list_owner; 380 struct net_device *dev; 381 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 382 struct sk_buff *skb; 383 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 384 int rx_count; /* length of rx_list */ 385 unsigned int napi_id; /* protected by netdev_lock */ 386 struct hrtimer timer; 387 /* all fields past this point are write-protected by netdev_lock */ 388 struct task_struct *thread; 389 unsigned long gro_flush_timeout; 390 unsigned long irq_suspend_timeout; 391 u32 defer_hard_irqs; 392 /* control-path-only fields follow */ 393 struct list_head dev_list; 394 struct hlist_node napi_hash_node; 395 int irq; 396 int index; 397 struct napi_config *config; 398 }; 399 400 enum { 401 NAPI_STATE_SCHED, /* Poll is scheduled */ 402 NAPI_STATE_MISSED, /* reschedule a napi */ 403 NAPI_STATE_DISABLE, /* Disable pending */ 404 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 405 NAPI_STATE_LISTED, /* NAPI added to system lists */ 406 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 407 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 408 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 409 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 410 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 411 }; 412 413 enum { 414 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 415 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 416 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 417 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 418 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 419 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 420 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 421 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 422 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 423 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 424 }; 425 426 enum gro_result { 427 GRO_MERGED, 428 GRO_MERGED_FREE, 429 GRO_HELD, 430 GRO_NORMAL, 431 GRO_CONSUMED, 432 }; 433 typedef enum gro_result gro_result_t; 434 435 /* 436 * enum rx_handler_result - Possible return values for rx_handlers. 437 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 438 * further. 439 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 440 * case skb->dev was changed by rx_handler. 441 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 442 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 443 * 444 * rx_handlers are functions called from inside __netif_receive_skb(), to do 445 * special processing of the skb, prior to delivery to protocol handlers. 446 * 447 * Currently, a net_device can only have a single rx_handler registered. Trying 448 * to register a second rx_handler will return -EBUSY. 449 * 450 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 451 * To unregister a rx_handler on a net_device, use 452 * netdev_rx_handler_unregister(). 453 * 454 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 455 * do with the skb. 456 * 457 * If the rx_handler consumed the skb in some way, it should return 458 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 459 * the skb to be delivered in some other way. 460 * 461 * If the rx_handler changed skb->dev, to divert the skb to another 462 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 463 * new device will be called if it exists. 464 * 465 * If the rx_handler decides the skb should be ignored, it should return 466 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 467 * are registered on exact device (ptype->dev == skb->dev). 468 * 469 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 470 * delivered, it should return RX_HANDLER_PASS. 471 * 472 * A device without a registered rx_handler will behave as if rx_handler 473 * returned RX_HANDLER_PASS. 474 */ 475 476 enum rx_handler_result { 477 RX_HANDLER_CONSUMED, 478 RX_HANDLER_ANOTHER, 479 RX_HANDLER_EXACT, 480 RX_HANDLER_PASS, 481 }; 482 typedef enum rx_handler_result rx_handler_result_t; 483 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 484 485 void __napi_schedule(struct napi_struct *n); 486 void __napi_schedule_irqoff(struct napi_struct *n); 487 488 static inline bool napi_disable_pending(struct napi_struct *n) 489 { 490 return test_bit(NAPI_STATE_DISABLE, &n->state); 491 } 492 493 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 494 { 495 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 496 } 497 498 /** 499 * napi_is_scheduled - test if NAPI is scheduled 500 * @n: NAPI context 501 * 502 * This check is "best-effort". With no locking implemented, 503 * a NAPI can be scheduled or terminate right after this check 504 * and produce not precise results. 505 * 506 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled 507 * should not be used normally and napi_schedule should be 508 * used instead. 509 * 510 * Use only if the driver really needs to check if a NAPI 511 * is scheduled for example in the context of delayed timer 512 * that can be skipped if a NAPI is already scheduled. 513 * 514 * Return: True if NAPI is scheduled, False otherwise. 515 */ 516 static inline bool napi_is_scheduled(struct napi_struct *n) 517 { 518 return test_bit(NAPI_STATE_SCHED, &n->state); 519 } 520 521 bool napi_schedule_prep(struct napi_struct *n); 522 523 /** 524 * napi_schedule - schedule NAPI poll 525 * @n: NAPI context 526 * 527 * Schedule NAPI poll routine to be called if it is not already 528 * running. 529 * Return: true if we schedule a NAPI or false if not. 530 * Refer to napi_schedule_prep() for additional reason on why 531 * a NAPI might not be scheduled. 532 */ 533 static inline bool napi_schedule(struct napi_struct *n) 534 { 535 if (napi_schedule_prep(n)) { 536 __napi_schedule(n); 537 return true; 538 } 539 540 return false; 541 } 542 543 /** 544 * napi_schedule_irqoff - schedule NAPI poll 545 * @n: NAPI context 546 * 547 * Variant of napi_schedule(), assuming hard irqs are masked. 548 */ 549 static inline void napi_schedule_irqoff(struct napi_struct *n) 550 { 551 if (napi_schedule_prep(n)) 552 __napi_schedule_irqoff(n); 553 } 554 555 /** 556 * napi_complete_done - NAPI processing complete 557 * @n: NAPI context 558 * @work_done: number of packets processed 559 * 560 * Mark NAPI processing as complete. Should only be called if poll budget 561 * has not been completely consumed. 562 * Prefer over napi_complete(). 563 * Return: false if device should avoid rearming interrupts. 564 */ 565 bool napi_complete_done(struct napi_struct *n, int work_done); 566 567 static inline bool napi_complete(struct napi_struct *n) 568 { 569 return napi_complete_done(n, 0); 570 } 571 572 int dev_set_threaded(struct net_device *dev, bool threaded); 573 574 void napi_disable(struct napi_struct *n); 575 void napi_disable_locked(struct napi_struct *n); 576 577 void napi_enable(struct napi_struct *n); 578 void napi_enable_locked(struct napi_struct *n); 579 580 /** 581 * napi_synchronize - wait until NAPI is not running 582 * @n: NAPI context 583 * 584 * Wait until NAPI is done being scheduled on this context. 585 * Waits till any outstanding processing completes but 586 * does not disable future activations. 587 */ 588 static inline void napi_synchronize(const struct napi_struct *n) 589 { 590 if (IS_ENABLED(CONFIG_SMP)) 591 while (test_bit(NAPI_STATE_SCHED, &n->state)) 592 msleep(1); 593 else 594 barrier(); 595 } 596 597 /** 598 * napi_if_scheduled_mark_missed - if napi is running, set the 599 * NAPIF_STATE_MISSED 600 * @n: NAPI context 601 * 602 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 603 * NAPI is scheduled. 604 **/ 605 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 606 { 607 unsigned long val, new; 608 609 val = READ_ONCE(n->state); 610 do { 611 if (val & NAPIF_STATE_DISABLE) 612 return true; 613 614 if (!(val & NAPIF_STATE_SCHED)) 615 return false; 616 617 new = val | NAPIF_STATE_MISSED; 618 } while (!try_cmpxchg(&n->state, &val, new)); 619 620 return true; 621 } 622 623 enum netdev_queue_state_t { 624 __QUEUE_STATE_DRV_XOFF, 625 __QUEUE_STATE_STACK_XOFF, 626 __QUEUE_STATE_FROZEN, 627 }; 628 629 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 630 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 631 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 632 633 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 634 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 635 QUEUE_STATE_FROZEN) 636 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 637 QUEUE_STATE_FROZEN) 638 639 /* 640 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 641 * netif_tx_* functions below are used to manipulate this flag. The 642 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 643 * queue independently. The netif_xmit_*stopped functions below are called 644 * to check if the queue has been stopped by the driver or stack (either 645 * of the XOFF bits are set in the state). Drivers should not need to call 646 * netif_xmit*stopped functions, they should only be using netif_tx_*. 647 */ 648 649 struct netdev_queue { 650 /* 651 * read-mostly part 652 */ 653 struct net_device *dev; 654 netdevice_tracker dev_tracker; 655 656 struct Qdisc __rcu *qdisc; 657 struct Qdisc __rcu *qdisc_sleeping; 658 #ifdef CONFIG_SYSFS 659 struct kobject kobj; 660 #endif 661 unsigned long tx_maxrate; 662 /* 663 * Number of TX timeouts for this queue 664 * (/sys/class/net/DEV/Q/trans_timeout) 665 */ 666 atomic_long_t trans_timeout; 667 668 /* Subordinate device that the queue has been assigned to */ 669 struct net_device *sb_dev; 670 #ifdef CONFIG_XDP_SOCKETS 671 struct xsk_buff_pool *pool; 672 #endif 673 674 /* 675 * write-mostly part 676 */ 677 #ifdef CONFIG_BQL 678 struct dql dql; 679 #endif 680 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 681 int xmit_lock_owner; 682 /* 683 * Time (in jiffies) of last Tx 684 */ 685 unsigned long trans_start; 686 687 unsigned long state; 688 689 /* 690 * slow- / control-path part 691 */ 692 /* NAPI instance for the queue 693 * Readers and writers must hold RTNL 694 */ 695 struct napi_struct *napi; 696 697 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 698 int numa_node; 699 #endif 700 } ____cacheline_aligned_in_smp; 701 702 extern int sysctl_fb_tunnels_only_for_init_net; 703 extern int sysctl_devconf_inherit_init_net; 704 705 /* 706 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 707 * == 1 : For initns only 708 * == 2 : For none. 709 */ 710 static inline bool net_has_fallback_tunnels(const struct net *net) 711 { 712 #if IS_ENABLED(CONFIG_SYSCTL) 713 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net); 714 715 return !fb_tunnels_only_for_init_net || 716 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1); 717 #else 718 return true; 719 #endif 720 } 721 722 static inline int net_inherit_devconf(void) 723 { 724 #if IS_ENABLED(CONFIG_SYSCTL) 725 return READ_ONCE(sysctl_devconf_inherit_init_net); 726 #else 727 return 0; 728 #endif 729 } 730 731 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 732 { 733 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 734 return q->numa_node; 735 #else 736 return NUMA_NO_NODE; 737 #endif 738 } 739 740 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 741 { 742 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 743 q->numa_node = node; 744 #endif 745 } 746 747 #ifdef CONFIG_RFS_ACCEL 748 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 749 u16 filter_id); 750 #endif 751 752 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 753 enum xps_map_type { 754 XPS_CPUS = 0, 755 XPS_RXQS, 756 XPS_MAPS_MAX, 757 }; 758 759 #ifdef CONFIG_XPS 760 /* 761 * This structure holds an XPS map which can be of variable length. The 762 * map is an array of queues. 763 */ 764 struct xps_map { 765 unsigned int len; 766 unsigned int alloc_len; 767 struct rcu_head rcu; 768 u16 queues[]; 769 }; 770 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 771 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 772 - sizeof(struct xps_map)) / sizeof(u16)) 773 774 /* 775 * This structure holds all XPS maps for device. Maps are indexed by CPU. 776 * 777 * We keep track of the number of cpus/rxqs used when the struct is allocated, 778 * in nr_ids. This will help not accessing out-of-bound memory. 779 * 780 * We keep track of the number of traffic classes used when the struct is 781 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 782 * not crossing its upper bound, as the original dev->num_tc can be updated in 783 * the meantime. 784 */ 785 struct xps_dev_maps { 786 struct rcu_head rcu; 787 unsigned int nr_ids; 788 s16 num_tc; 789 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 790 }; 791 792 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 793 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 794 795 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 796 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 797 798 #endif /* CONFIG_XPS */ 799 800 #define TC_MAX_QUEUE 16 801 #define TC_BITMASK 15 802 /* HW offloaded queuing disciplines txq count and offset maps */ 803 struct netdev_tc_txq { 804 u16 count; 805 u16 offset; 806 }; 807 808 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 809 /* 810 * This structure is to hold information about the device 811 * configured to run FCoE protocol stack. 812 */ 813 struct netdev_fcoe_hbainfo { 814 char manufacturer[64]; 815 char serial_number[64]; 816 char hardware_version[64]; 817 char driver_version[64]; 818 char optionrom_version[64]; 819 char firmware_version[64]; 820 char model[256]; 821 char model_description[256]; 822 }; 823 #endif 824 825 #define MAX_PHYS_ITEM_ID_LEN 32 826 827 /* This structure holds a unique identifier to identify some 828 * physical item (port for example) used by a netdevice. 829 */ 830 struct netdev_phys_item_id { 831 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 832 unsigned char id_len; 833 }; 834 835 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 836 struct netdev_phys_item_id *b) 837 { 838 return a->id_len == b->id_len && 839 memcmp(a->id, b->id, a->id_len) == 0; 840 } 841 842 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 843 struct sk_buff *skb, 844 struct net_device *sb_dev); 845 846 enum net_device_path_type { 847 DEV_PATH_ETHERNET = 0, 848 DEV_PATH_VLAN, 849 DEV_PATH_BRIDGE, 850 DEV_PATH_PPPOE, 851 DEV_PATH_DSA, 852 DEV_PATH_MTK_WDMA, 853 }; 854 855 struct net_device_path { 856 enum net_device_path_type type; 857 const struct net_device *dev; 858 union { 859 struct { 860 u16 id; 861 __be16 proto; 862 u8 h_dest[ETH_ALEN]; 863 } encap; 864 struct { 865 enum { 866 DEV_PATH_BR_VLAN_KEEP, 867 DEV_PATH_BR_VLAN_TAG, 868 DEV_PATH_BR_VLAN_UNTAG, 869 DEV_PATH_BR_VLAN_UNTAG_HW, 870 } vlan_mode; 871 u16 vlan_id; 872 __be16 vlan_proto; 873 } bridge; 874 struct { 875 int port; 876 u16 proto; 877 } dsa; 878 struct { 879 u8 wdma_idx; 880 u8 queue; 881 u16 wcid; 882 u8 bss; 883 u8 amsdu; 884 } mtk_wdma; 885 }; 886 }; 887 888 #define NET_DEVICE_PATH_STACK_MAX 5 889 #define NET_DEVICE_PATH_VLAN_MAX 2 890 891 struct net_device_path_stack { 892 int num_paths; 893 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 894 }; 895 896 struct net_device_path_ctx { 897 const struct net_device *dev; 898 u8 daddr[ETH_ALEN]; 899 900 int num_vlans; 901 struct { 902 u16 id; 903 __be16 proto; 904 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 905 }; 906 907 enum tc_setup_type { 908 TC_QUERY_CAPS, 909 TC_SETUP_QDISC_MQPRIO, 910 TC_SETUP_CLSU32, 911 TC_SETUP_CLSFLOWER, 912 TC_SETUP_CLSMATCHALL, 913 TC_SETUP_CLSBPF, 914 TC_SETUP_BLOCK, 915 TC_SETUP_QDISC_CBS, 916 TC_SETUP_QDISC_RED, 917 TC_SETUP_QDISC_PRIO, 918 TC_SETUP_QDISC_MQ, 919 TC_SETUP_QDISC_ETF, 920 TC_SETUP_ROOT_QDISC, 921 TC_SETUP_QDISC_GRED, 922 TC_SETUP_QDISC_TAPRIO, 923 TC_SETUP_FT, 924 TC_SETUP_QDISC_ETS, 925 TC_SETUP_QDISC_TBF, 926 TC_SETUP_QDISC_FIFO, 927 TC_SETUP_QDISC_HTB, 928 TC_SETUP_ACT, 929 }; 930 931 /* These structures hold the attributes of bpf state that are being passed 932 * to the netdevice through the bpf op. 933 */ 934 enum bpf_netdev_command { 935 /* Set or clear a bpf program used in the earliest stages of packet 936 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 937 * is responsible for calling bpf_prog_put on any old progs that are 938 * stored. In case of error, the callee need not release the new prog 939 * reference, but on success it takes ownership and must bpf_prog_put 940 * when it is no longer used. 941 */ 942 XDP_SETUP_PROG, 943 XDP_SETUP_PROG_HW, 944 /* BPF program for offload callbacks, invoked at program load time. */ 945 BPF_OFFLOAD_MAP_ALLOC, 946 BPF_OFFLOAD_MAP_FREE, 947 XDP_SETUP_XSK_POOL, 948 }; 949 950 struct bpf_prog_offload_ops; 951 struct netlink_ext_ack; 952 struct xdp_umem; 953 struct xdp_dev_bulk_queue; 954 struct bpf_xdp_link; 955 956 enum bpf_xdp_mode { 957 XDP_MODE_SKB = 0, 958 XDP_MODE_DRV = 1, 959 XDP_MODE_HW = 2, 960 __MAX_XDP_MODE 961 }; 962 963 struct bpf_xdp_entity { 964 struct bpf_prog *prog; 965 struct bpf_xdp_link *link; 966 }; 967 968 struct netdev_bpf { 969 enum bpf_netdev_command command; 970 union { 971 /* XDP_SETUP_PROG */ 972 struct { 973 u32 flags; 974 struct bpf_prog *prog; 975 struct netlink_ext_ack *extack; 976 }; 977 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 978 struct { 979 struct bpf_offloaded_map *offmap; 980 }; 981 /* XDP_SETUP_XSK_POOL */ 982 struct { 983 struct xsk_buff_pool *pool; 984 u16 queue_id; 985 } xsk; 986 }; 987 }; 988 989 /* Flags for ndo_xsk_wakeup. */ 990 #define XDP_WAKEUP_RX (1 << 0) 991 #define XDP_WAKEUP_TX (1 << 1) 992 993 #ifdef CONFIG_XFRM_OFFLOAD 994 struct xfrmdev_ops { 995 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack); 996 void (*xdo_dev_state_delete) (struct xfrm_state *x); 997 void (*xdo_dev_state_free) (struct xfrm_state *x); 998 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 999 struct xfrm_state *x); 1000 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1001 void (*xdo_dev_state_update_stats) (struct xfrm_state *x); 1002 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack); 1003 void (*xdo_dev_policy_delete) (struct xfrm_policy *x); 1004 void (*xdo_dev_policy_free) (struct xfrm_policy *x); 1005 }; 1006 #endif 1007 1008 struct dev_ifalias { 1009 struct rcu_head rcuhead; 1010 char ifalias[]; 1011 }; 1012 1013 struct devlink; 1014 struct tlsdev_ops; 1015 1016 struct netdev_net_notifier { 1017 struct list_head list; 1018 struct notifier_block *nb; 1019 }; 1020 1021 /* 1022 * This structure defines the management hooks for network devices. 1023 * The following hooks can be defined; unless noted otherwise, they are 1024 * optional and can be filled with a null pointer. 1025 * 1026 * int (*ndo_init)(struct net_device *dev); 1027 * This function is called once when a network device is registered. 1028 * The network device can use this for any late stage initialization 1029 * or semantic validation. It can fail with an error code which will 1030 * be propagated back to register_netdev. 1031 * 1032 * void (*ndo_uninit)(struct net_device *dev); 1033 * This function is called when device is unregistered or when registration 1034 * fails. It is not called if init fails. 1035 * 1036 * int (*ndo_open)(struct net_device *dev); 1037 * This function is called when a network device transitions to the up 1038 * state. 1039 * 1040 * int (*ndo_stop)(struct net_device *dev); 1041 * This function is called when a network device transitions to the down 1042 * state. 1043 * 1044 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1045 * struct net_device *dev); 1046 * Called when a packet needs to be transmitted. 1047 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1048 * the queue before that can happen; it's for obsolete devices and weird 1049 * corner cases, but the stack really does a non-trivial amount 1050 * of useless work if you return NETDEV_TX_BUSY. 1051 * Required; cannot be NULL. 1052 * 1053 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1054 * struct net_device *dev 1055 * netdev_features_t features); 1056 * Called by core transmit path to determine if device is capable of 1057 * performing offload operations on a given packet. This is to give 1058 * the device an opportunity to implement any restrictions that cannot 1059 * be otherwise expressed by feature flags. The check is called with 1060 * the set of features that the stack has calculated and it returns 1061 * those the driver believes to be appropriate. 1062 * 1063 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1064 * struct net_device *sb_dev); 1065 * Called to decide which queue to use when device supports multiple 1066 * transmit queues. 1067 * 1068 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1069 * This function is called to allow device receiver to make 1070 * changes to configuration when multicast or promiscuous is enabled. 1071 * 1072 * void (*ndo_set_rx_mode)(struct net_device *dev); 1073 * This function is called device changes address list filtering. 1074 * If driver handles unicast address filtering, it should set 1075 * IFF_UNICAST_FLT in its priv_flags. 1076 * 1077 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1078 * This function is called when the Media Access Control address 1079 * needs to be changed. If this interface is not defined, the 1080 * MAC address can not be changed. 1081 * 1082 * int (*ndo_validate_addr)(struct net_device *dev); 1083 * Test if Media Access Control address is valid for the device. 1084 * 1085 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1086 * Old-style ioctl entry point. This is used internally by the 1087 * appletalk and ieee802154 subsystems but is no longer called by 1088 * the device ioctl handler. 1089 * 1090 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1091 * Used by the bonding driver for its device specific ioctls: 1092 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1093 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1094 * 1095 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1096 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1097 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1098 * 1099 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1100 * Used to set network devices bus interface parameters. This interface 1101 * is retained for legacy reasons; new devices should use the bus 1102 * interface (PCI) for low level management. 1103 * 1104 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1105 * Called when a user wants to change the Maximum Transfer Unit 1106 * of a device. 1107 * 1108 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1109 * Callback used when the transmitter has not made any progress 1110 * for dev->watchdog ticks. 1111 * 1112 * void (*ndo_get_stats64)(struct net_device *dev, 1113 * struct rtnl_link_stats64 *storage); 1114 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1115 * Called when a user wants to get the network device usage 1116 * statistics. Drivers must do one of the following: 1117 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1118 * rtnl_link_stats64 structure passed by the caller. 1119 * 2. Define @ndo_get_stats to update a net_device_stats structure 1120 * (which should normally be dev->stats) and return a pointer to 1121 * it. The structure may be changed asynchronously only if each 1122 * field is written atomically. 1123 * 3. Update dev->stats asynchronously and atomically, and define 1124 * neither operation. 1125 * 1126 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1127 * Return true if this device supports offload stats of this attr_id. 1128 * 1129 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1130 * void *attr_data) 1131 * Get statistics for offload operations by attr_id. Write it into the 1132 * attr_data pointer. 1133 * 1134 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1135 * If device supports VLAN filtering this function is called when a 1136 * VLAN id is registered. 1137 * 1138 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1139 * If device supports VLAN filtering this function is called when a 1140 * VLAN id is unregistered. 1141 * 1142 * void (*ndo_poll_controller)(struct net_device *dev); 1143 * 1144 * SR-IOV management functions. 1145 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1146 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1147 * u8 qos, __be16 proto); 1148 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1149 * int max_tx_rate); 1150 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1151 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1152 * int (*ndo_get_vf_config)(struct net_device *dev, 1153 * int vf, struct ifla_vf_info *ivf); 1154 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1155 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1156 * struct nlattr *port[]); 1157 * 1158 * Enable or disable the VF ability to query its RSS Redirection Table and 1159 * Hash Key. This is needed since on some devices VF share this information 1160 * with PF and querying it may introduce a theoretical security risk. 1161 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1162 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1163 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1164 * void *type_data); 1165 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1166 * This is always called from the stack with the rtnl lock held and netif 1167 * tx queues stopped. This allows the netdevice to perform queue 1168 * management safely. 1169 * 1170 * Fiber Channel over Ethernet (FCoE) offload functions. 1171 * int (*ndo_fcoe_enable)(struct net_device *dev); 1172 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1173 * so the underlying device can perform whatever needed configuration or 1174 * initialization to support acceleration of FCoE traffic. 1175 * 1176 * int (*ndo_fcoe_disable)(struct net_device *dev); 1177 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1178 * so the underlying device can perform whatever needed clean-ups to 1179 * stop supporting acceleration of FCoE traffic. 1180 * 1181 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1182 * struct scatterlist *sgl, unsigned int sgc); 1183 * Called when the FCoE Initiator wants to initialize an I/O that 1184 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1185 * perform necessary setup and returns 1 to indicate the device is set up 1186 * successfully to perform DDP on this I/O, otherwise this returns 0. 1187 * 1188 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1189 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1190 * indicated by the FC exchange id 'xid', so the underlying device can 1191 * clean up and reuse resources for later DDP requests. 1192 * 1193 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1194 * struct scatterlist *sgl, unsigned int sgc); 1195 * Called when the FCoE Target wants to initialize an I/O that 1196 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1197 * perform necessary setup and returns 1 to indicate the device is set up 1198 * successfully to perform DDP on this I/O, otherwise this returns 0. 1199 * 1200 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1201 * struct netdev_fcoe_hbainfo *hbainfo); 1202 * Called when the FCoE Protocol stack wants information on the underlying 1203 * device. This information is utilized by the FCoE protocol stack to 1204 * register attributes with Fiber Channel management service as per the 1205 * FC-GS Fabric Device Management Information(FDMI) specification. 1206 * 1207 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1208 * Called when the underlying device wants to override default World Wide 1209 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1210 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1211 * protocol stack to use. 1212 * 1213 * RFS acceleration. 1214 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1215 * u16 rxq_index, u32 flow_id); 1216 * Set hardware filter for RFS. rxq_index is the target queue index; 1217 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1218 * Return the filter ID on success, or a negative error code. 1219 * 1220 * Slave management functions (for bridge, bonding, etc). 1221 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1222 * Called to make another netdev an underling. 1223 * 1224 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1225 * Called to release previously enslaved netdev. 1226 * 1227 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1228 * struct sk_buff *skb, 1229 * bool all_slaves); 1230 * Get the xmit slave of master device. If all_slaves is true, function 1231 * assume all the slaves can transmit. 1232 * 1233 * Feature/offload setting functions. 1234 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1235 * netdev_features_t features); 1236 * Adjusts the requested feature flags according to device-specific 1237 * constraints, and returns the resulting flags. Must not modify 1238 * the device state. 1239 * 1240 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1241 * Called to update device configuration to new features. Passed 1242 * feature set might be less than what was returned by ndo_fix_features()). 1243 * Must return >0 or -errno if it changed dev->features itself. 1244 * 1245 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1246 * struct net_device *dev, 1247 * const unsigned char *addr, u16 vid, u16 flags, 1248 * bool *notified, struct netlink_ext_ack *extack); 1249 * Adds an FDB entry to dev for addr. 1250 * Callee shall set *notified to true if it sent any appropriate 1251 * notification(s). Otherwise core will send a generic one. 1252 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1253 * struct net_device *dev, 1254 * const unsigned char *addr, u16 vid 1255 * bool *notified, struct netlink_ext_ack *extack); 1256 * Deletes the FDB entry from dev corresponding to addr. 1257 * Callee shall set *notified to true if it sent any appropriate 1258 * notification(s). Otherwise core will send a generic one. 1259 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev, 1260 * struct netlink_ext_ack *extack); 1261 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1262 * struct net_device *dev, struct net_device *filter_dev, 1263 * int *idx) 1264 * Used to add FDB entries to dump requests. Implementers should add 1265 * entries to skb and update idx with the number of entries. 1266 * 1267 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[], 1268 * u16 nlmsg_flags, struct netlink_ext_ack *extack); 1269 * Adds an MDB entry to dev. 1270 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[], 1271 * struct netlink_ext_ack *extack); 1272 * Deletes the MDB entry from dev. 1273 * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[], 1274 * struct netlink_ext_ack *extack); 1275 * Bulk deletes MDB entries from dev. 1276 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb, 1277 * struct netlink_callback *cb); 1278 * Dumps MDB entries from dev. The first argument (marker) in the netlink 1279 * callback is used by core rtnetlink code. 1280 * 1281 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1282 * u16 flags, struct netlink_ext_ack *extack) 1283 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1284 * struct net_device *dev, u32 filter_mask, 1285 * int nlflags) 1286 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1287 * u16 flags); 1288 * 1289 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1290 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1291 * which do not represent real hardware may define this to allow their 1292 * userspace components to manage their virtual carrier state. Devices 1293 * that determine carrier state from physical hardware properties (eg 1294 * network cables) or protocol-dependent mechanisms (eg 1295 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1296 * 1297 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1298 * struct netdev_phys_item_id *ppid); 1299 * Called to get ID of physical port of this device. If driver does 1300 * not implement this, it is assumed that the hw is not able to have 1301 * multiple net devices on single physical port. 1302 * 1303 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1304 * struct netdev_phys_item_id *ppid) 1305 * Called to get the parent ID of the physical port of this device. 1306 * 1307 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1308 * struct net_device *dev) 1309 * Called by upper layer devices to accelerate switching or other 1310 * station functionality into hardware. 'pdev is the lowerdev 1311 * to use for the offload and 'dev' is the net device that will 1312 * back the offload. Returns a pointer to the private structure 1313 * the upper layer will maintain. 1314 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1315 * Called by upper layer device to delete the station created 1316 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1317 * the station and priv is the structure returned by the add 1318 * operation. 1319 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1320 * int queue_index, u32 maxrate); 1321 * Called when a user wants to set a max-rate limitation of specific 1322 * TX queue. 1323 * int (*ndo_get_iflink)(const struct net_device *dev); 1324 * Called to get the iflink value of this device. 1325 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1326 * This function is used to get egress tunnel information for given skb. 1327 * This is useful for retrieving outer tunnel header parameters while 1328 * sampling packet. 1329 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1330 * This function is used to specify the headroom that the skb must 1331 * consider when allocation skb during packet reception. Setting 1332 * appropriate rx headroom value allows avoiding skb head copy on 1333 * forward. Setting a negative value resets the rx headroom to the 1334 * default value. 1335 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1336 * This function is used to set or query state related to XDP on the 1337 * netdevice and manage BPF offload. See definition of 1338 * enum bpf_netdev_command for details. 1339 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1340 * u32 flags); 1341 * This function is used to submit @n XDP packets for transmit on a 1342 * netdevice. Returns number of frames successfully transmitted, frames 1343 * that got dropped are freed/returned via xdp_return_frame(). 1344 * Returns negative number, means general error invoking ndo, meaning 1345 * no frames were xmit'ed and core-caller will free all frames. 1346 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1347 * struct xdp_buff *xdp); 1348 * Get the xmit slave of master device based on the xdp_buff. 1349 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1350 * This function is used to wake up the softirq, ksoftirqd or kthread 1351 * responsible for sending and/or receiving packets on a specific 1352 * queue id bound to an AF_XDP socket. The flags field specifies if 1353 * only RX, only Tx, or both should be woken up using the flags 1354 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1355 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm_kern *p, 1356 * int cmd); 1357 * Add, change, delete or get information on an IPv4 tunnel. 1358 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1359 * If a device is paired with a peer device, return the peer instance. 1360 * The caller must be under RCU read context. 1361 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1362 * Get the forwarding path to reach the real device from the HW destination address 1363 * ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1364 * const struct skb_shared_hwtstamps *hwtstamps, 1365 * bool cycles); 1366 * Get hardware timestamp based on normal/adjustable time or free running 1367 * cycle counter. This function is required if physical clock supports a 1368 * free running cycle counter. 1369 * 1370 * int (*ndo_hwtstamp_get)(struct net_device *dev, 1371 * struct kernel_hwtstamp_config *kernel_config); 1372 * Get the currently configured hardware timestamping parameters for the 1373 * NIC device. 1374 * 1375 * int (*ndo_hwtstamp_set)(struct net_device *dev, 1376 * struct kernel_hwtstamp_config *kernel_config, 1377 * struct netlink_ext_ack *extack); 1378 * Change the hardware timestamping parameters for NIC device. 1379 */ 1380 struct net_device_ops { 1381 int (*ndo_init)(struct net_device *dev); 1382 void (*ndo_uninit)(struct net_device *dev); 1383 int (*ndo_open)(struct net_device *dev); 1384 int (*ndo_stop)(struct net_device *dev); 1385 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1386 struct net_device *dev); 1387 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1388 struct net_device *dev, 1389 netdev_features_t features); 1390 u16 (*ndo_select_queue)(struct net_device *dev, 1391 struct sk_buff *skb, 1392 struct net_device *sb_dev); 1393 void (*ndo_change_rx_flags)(struct net_device *dev, 1394 int flags); 1395 void (*ndo_set_rx_mode)(struct net_device *dev); 1396 int (*ndo_set_mac_address)(struct net_device *dev, 1397 void *addr); 1398 int (*ndo_validate_addr)(struct net_device *dev); 1399 int (*ndo_do_ioctl)(struct net_device *dev, 1400 struct ifreq *ifr, int cmd); 1401 int (*ndo_eth_ioctl)(struct net_device *dev, 1402 struct ifreq *ifr, int cmd); 1403 int (*ndo_siocbond)(struct net_device *dev, 1404 struct ifreq *ifr, int cmd); 1405 int (*ndo_siocwandev)(struct net_device *dev, 1406 struct if_settings *ifs); 1407 int (*ndo_siocdevprivate)(struct net_device *dev, 1408 struct ifreq *ifr, 1409 void __user *data, int cmd); 1410 int (*ndo_set_config)(struct net_device *dev, 1411 struct ifmap *map); 1412 int (*ndo_change_mtu)(struct net_device *dev, 1413 int new_mtu); 1414 int (*ndo_neigh_setup)(struct net_device *dev, 1415 struct neigh_parms *); 1416 void (*ndo_tx_timeout) (struct net_device *dev, 1417 unsigned int txqueue); 1418 1419 void (*ndo_get_stats64)(struct net_device *dev, 1420 struct rtnl_link_stats64 *storage); 1421 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1422 int (*ndo_get_offload_stats)(int attr_id, 1423 const struct net_device *dev, 1424 void *attr_data); 1425 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1426 1427 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1428 __be16 proto, u16 vid); 1429 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1430 __be16 proto, u16 vid); 1431 #ifdef CONFIG_NET_POLL_CONTROLLER 1432 void (*ndo_poll_controller)(struct net_device *dev); 1433 int (*ndo_netpoll_setup)(struct net_device *dev); 1434 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1435 #endif 1436 int (*ndo_set_vf_mac)(struct net_device *dev, 1437 int queue, u8 *mac); 1438 int (*ndo_set_vf_vlan)(struct net_device *dev, 1439 int queue, u16 vlan, 1440 u8 qos, __be16 proto); 1441 int (*ndo_set_vf_rate)(struct net_device *dev, 1442 int vf, int min_tx_rate, 1443 int max_tx_rate); 1444 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1445 int vf, bool setting); 1446 int (*ndo_set_vf_trust)(struct net_device *dev, 1447 int vf, bool setting); 1448 int (*ndo_get_vf_config)(struct net_device *dev, 1449 int vf, 1450 struct ifla_vf_info *ivf); 1451 int (*ndo_set_vf_link_state)(struct net_device *dev, 1452 int vf, int link_state); 1453 int (*ndo_get_vf_stats)(struct net_device *dev, 1454 int vf, 1455 struct ifla_vf_stats 1456 *vf_stats); 1457 int (*ndo_set_vf_port)(struct net_device *dev, 1458 int vf, 1459 struct nlattr *port[]); 1460 int (*ndo_get_vf_port)(struct net_device *dev, 1461 int vf, struct sk_buff *skb); 1462 int (*ndo_get_vf_guid)(struct net_device *dev, 1463 int vf, 1464 struct ifla_vf_guid *node_guid, 1465 struct ifla_vf_guid *port_guid); 1466 int (*ndo_set_vf_guid)(struct net_device *dev, 1467 int vf, u64 guid, 1468 int guid_type); 1469 int (*ndo_set_vf_rss_query_en)( 1470 struct net_device *dev, 1471 int vf, bool setting); 1472 int (*ndo_setup_tc)(struct net_device *dev, 1473 enum tc_setup_type type, 1474 void *type_data); 1475 #if IS_ENABLED(CONFIG_FCOE) 1476 int (*ndo_fcoe_enable)(struct net_device *dev); 1477 int (*ndo_fcoe_disable)(struct net_device *dev); 1478 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1479 u16 xid, 1480 struct scatterlist *sgl, 1481 unsigned int sgc); 1482 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1483 u16 xid); 1484 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1485 u16 xid, 1486 struct scatterlist *sgl, 1487 unsigned int sgc); 1488 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1489 struct netdev_fcoe_hbainfo *hbainfo); 1490 #endif 1491 1492 #if IS_ENABLED(CONFIG_LIBFCOE) 1493 #define NETDEV_FCOE_WWNN 0 1494 #define NETDEV_FCOE_WWPN 1 1495 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1496 u64 *wwn, int type); 1497 #endif 1498 1499 #ifdef CONFIG_RFS_ACCEL 1500 int (*ndo_rx_flow_steer)(struct net_device *dev, 1501 const struct sk_buff *skb, 1502 u16 rxq_index, 1503 u32 flow_id); 1504 #endif 1505 int (*ndo_add_slave)(struct net_device *dev, 1506 struct net_device *slave_dev, 1507 struct netlink_ext_ack *extack); 1508 int (*ndo_del_slave)(struct net_device *dev, 1509 struct net_device *slave_dev); 1510 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1511 struct sk_buff *skb, 1512 bool all_slaves); 1513 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1514 struct sock *sk); 1515 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1516 netdev_features_t features); 1517 int (*ndo_set_features)(struct net_device *dev, 1518 netdev_features_t features); 1519 int (*ndo_neigh_construct)(struct net_device *dev, 1520 struct neighbour *n); 1521 void (*ndo_neigh_destroy)(struct net_device *dev, 1522 struct neighbour *n); 1523 1524 int (*ndo_fdb_add)(struct ndmsg *ndm, 1525 struct nlattr *tb[], 1526 struct net_device *dev, 1527 const unsigned char *addr, 1528 u16 vid, 1529 u16 flags, 1530 bool *notified, 1531 struct netlink_ext_ack *extack); 1532 int (*ndo_fdb_del)(struct ndmsg *ndm, 1533 struct nlattr *tb[], 1534 struct net_device *dev, 1535 const unsigned char *addr, 1536 u16 vid, 1537 bool *notified, 1538 struct netlink_ext_ack *extack); 1539 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, 1540 struct net_device *dev, 1541 struct netlink_ext_ack *extack); 1542 int (*ndo_fdb_dump)(struct sk_buff *skb, 1543 struct netlink_callback *cb, 1544 struct net_device *dev, 1545 struct net_device *filter_dev, 1546 int *idx); 1547 int (*ndo_fdb_get)(struct sk_buff *skb, 1548 struct nlattr *tb[], 1549 struct net_device *dev, 1550 const unsigned char *addr, 1551 u16 vid, u32 portid, u32 seq, 1552 struct netlink_ext_ack *extack); 1553 int (*ndo_mdb_add)(struct net_device *dev, 1554 struct nlattr *tb[], 1555 u16 nlmsg_flags, 1556 struct netlink_ext_ack *extack); 1557 int (*ndo_mdb_del)(struct net_device *dev, 1558 struct nlattr *tb[], 1559 struct netlink_ext_ack *extack); 1560 int (*ndo_mdb_del_bulk)(struct net_device *dev, 1561 struct nlattr *tb[], 1562 struct netlink_ext_ack *extack); 1563 int (*ndo_mdb_dump)(struct net_device *dev, 1564 struct sk_buff *skb, 1565 struct netlink_callback *cb); 1566 int (*ndo_mdb_get)(struct net_device *dev, 1567 struct nlattr *tb[], u32 portid, 1568 u32 seq, 1569 struct netlink_ext_ack *extack); 1570 int (*ndo_bridge_setlink)(struct net_device *dev, 1571 struct nlmsghdr *nlh, 1572 u16 flags, 1573 struct netlink_ext_ack *extack); 1574 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1575 u32 pid, u32 seq, 1576 struct net_device *dev, 1577 u32 filter_mask, 1578 int nlflags); 1579 int (*ndo_bridge_dellink)(struct net_device *dev, 1580 struct nlmsghdr *nlh, 1581 u16 flags); 1582 int (*ndo_change_carrier)(struct net_device *dev, 1583 bool new_carrier); 1584 int (*ndo_get_phys_port_id)(struct net_device *dev, 1585 struct netdev_phys_item_id *ppid); 1586 int (*ndo_get_port_parent_id)(struct net_device *dev, 1587 struct netdev_phys_item_id *ppid); 1588 int (*ndo_get_phys_port_name)(struct net_device *dev, 1589 char *name, size_t len); 1590 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1591 struct net_device *dev); 1592 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1593 void *priv); 1594 1595 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1596 int queue_index, 1597 u32 maxrate); 1598 int (*ndo_get_iflink)(const struct net_device *dev); 1599 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1600 struct sk_buff *skb); 1601 void (*ndo_set_rx_headroom)(struct net_device *dev, 1602 int needed_headroom); 1603 int (*ndo_bpf)(struct net_device *dev, 1604 struct netdev_bpf *bpf); 1605 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1606 struct xdp_frame **xdp, 1607 u32 flags); 1608 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1609 struct xdp_buff *xdp); 1610 int (*ndo_xsk_wakeup)(struct net_device *dev, 1611 u32 queue_id, u32 flags); 1612 int (*ndo_tunnel_ctl)(struct net_device *dev, 1613 struct ip_tunnel_parm_kern *p, 1614 int cmd); 1615 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1616 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1617 struct net_device_path *path); 1618 ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1619 const struct skb_shared_hwtstamps *hwtstamps, 1620 bool cycles); 1621 int (*ndo_hwtstamp_get)(struct net_device *dev, 1622 struct kernel_hwtstamp_config *kernel_config); 1623 int (*ndo_hwtstamp_set)(struct net_device *dev, 1624 struct kernel_hwtstamp_config *kernel_config, 1625 struct netlink_ext_ack *extack); 1626 1627 #if IS_ENABLED(CONFIG_NET_SHAPER) 1628 /** 1629 * @net_shaper_ops: Device shaping offload operations 1630 * see include/net/net_shapers.h 1631 */ 1632 const struct net_shaper_ops *net_shaper_ops; 1633 #endif 1634 }; 1635 1636 /** 1637 * enum netdev_priv_flags - &struct net_device priv_flags 1638 * 1639 * These are the &struct net_device, they are only set internally 1640 * by drivers and used in the kernel. These flags are invisible to 1641 * userspace; this means that the order of these flags can change 1642 * during any kernel release. 1643 * 1644 * You should add bitfield booleans after either net_device::priv_flags 1645 * (hotpath) or ::threaded (slowpath) instead of extending these flags. 1646 * 1647 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1648 * @IFF_EBRIDGE: Ethernet bridging device 1649 * @IFF_BONDING: bonding master or slave 1650 * @IFF_ISATAP: ISATAP interface (RFC4214) 1651 * @IFF_WAN_HDLC: WAN HDLC device 1652 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1653 * release skb->dst 1654 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1655 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1656 * @IFF_MACVLAN_PORT: device used as macvlan port 1657 * @IFF_BRIDGE_PORT: device used as bridge port 1658 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1659 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1660 * @IFF_UNICAST_FLT: Supports unicast filtering 1661 * @IFF_TEAM_PORT: device used as team port 1662 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1663 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1664 * change when it's running 1665 * @IFF_MACVLAN: Macvlan device 1666 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1667 * underlying stacked devices 1668 * @IFF_L3MDEV_MASTER: device is an L3 master device 1669 * @IFF_NO_QUEUE: device can run without qdisc attached 1670 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1671 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1672 * @IFF_TEAM: device is a team device 1673 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1674 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1675 * entity (i.e. the master device for bridged veth) 1676 * @IFF_MACSEC: device is a MACsec device 1677 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1678 * @IFF_FAILOVER: device is a failover master device 1679 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1680 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1681 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf 1682 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1683 * skb_headlen(skb) == 0 (data starts from frag0) 1684 */ 1685 enum netdev_priv_flags { 1686 IFF_802_1Q_VLAN = 1<<0, 1687 IFF_EBRIDGE = 1<<1, 1688 IFF_BONDING = 1<<2, 1689 IFF_ISATAP = 1<<3, 1690 IFF_WAN_HDLC = 1<<4, 1691 IFF_XMIT_DST_RELEASE = 1<<5, 1692 IFF_DONT_BRIDGE = 1<<6, 1693 IFF_DISABLE_NETPOLL = 1<<7, 1694 IFF_MACVLAN_PORT = 1<<8, 1695 IFF_BRIDGE_PORT = 1<<9, 1696 IFF_OVS_DATAPATH = 1<<10, 1697 IFF_TX_SKB_SHARING = 1<<11, 1698 IFF_UNICAST_FLT = 1<<12, 1699 IFF_TEAM_PORT = 1<<13, 1700 IFF_SUPP_NOFCS = 1<<14, 1701 IFF_LIVE_ADDR_CHANGE = 1<<15, 1702 IFF_MACVLAN = 1<<16, 1703 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1704 IFF_L3MDEV_MASTER = 1<<18, 1705 IFF_NO_QUEUE = 1<<19, 1706 IFF_OPENVSWITCH = 1<<20, 1707 IFF_L3MDEV_SLAVE = 1<<21, 1708 IFF_TEAM = 1<<22, 1709 IFF_RXFH_CONFIGURED = 1<<23, 1710 IFF_PHONY_HEADROOM = 1<<24, 1711 IFF_MACSEC = 1<<25, 1712 IFF_NO_RX_HANDLER = 1<<26, 1713 IFF_FAILOVER = 1<<27, 1714 IFF_FAILOVER_SLAVE = 1<<28, 1715 IFF_L3MDEV_RX_HANDLER = 1<<29, 1716 IFF_NO_ADDRCONF = BIT_ULL(30), 1717 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31), 1718 }; 1719 1720 /* Specifies the type of the struct net_device::ml_priv pointer */ 1721 enum netdev_ml_priv_type { 1722 ML_PRIV_NONE, 1723 ML_PRIV_CAN, 1724 }; 1725 1726 enum netdev_stat_type { 1727 NETDEV_PCPU_STAT_NONE, 1728 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */ 1729 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */ 1730 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */ 1731 }; 1732 1733 enum netdev_reg_state { 1734 NETREG_UNINITIALIZED = 0, 1735 NETREG_REGISTERED, /* completed register_netdevice */ 1736 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1737 NETREG_UNREGISTERED, /* completed unregister todo */ 1738 NETREG_RELEASED, /* called free_netdev */ 1739 NETREG_DUMMY, /* dummy device for NAPI poll */ 1740 }; 1741 1742 /** 1743 * struct net_device - The DEVICE structure. 1744 * 1745 * Actually, this whole structure is a big mistake. It mixes I/O 1746 * data with strictly "high-level" data, and it has to know about 1747 * almost every data structure used in the INET module. 1748 * 1749 * @priv_flags: flags invisible to userspace defined as bits, see 1750 * enum netdev_priv_flags for the definitions 1751 * @lltx: device supports lockless Tx. Deprecated for real HW 1752 * drivers. Mainly used by logical interfaces, such as 1753 * bonding and tunnels 1754 * 1755 * @name: This is the first field of the "visible" part of this structure 1756 * (i.e. as seen by users in the "Space.c" file). It is the name 1757 * of the interface. 1758 * 1759 * @name_node: Name hashlist node 1760 * @ifalias: SNMP alias 1761 * @mem_end: Shared memory end 1762 * @mem_start: Shared memory start 1763 * @base_addr: Device I/O address 1764 * @irq: Device IRQ number 1765 * 1766 * @state: Generic network queuing layer state, see netdev_state_t 1767 * @dev_list: The global list of network devices 1768 * @napi_list: List entry used for polling NAPI devices 1769 * @unreg_list: List entry when we are unregistering the 1770 * device; see the function unregister_netdev 1771 * @close_list: List entry used when we are closing the device 1772 * @ptype_all: Device-specific packet handlers for all protocols 1773 * @ptype_specific: Device-specific, protocol-specific packet handlers 1774 * 1775 * @adj_list: Directly linked devices, like slaves for bonding 1776 * @features: Currently active device features 1777 * @hw_features: User-changeable features 1778 * 1779 * @wanted_features: User-requested features 1780 * @vlan_features: Mask of features inheritable by VLAN devices 1781 * 1782 * @hw_enc_features: Mask of features inherited by encapsulating devices 1783 * This field indicates what encapsulation 1784 * offloads the hardware is capable of doing, 1785 * and drivers will need to set them appropriately. 1786 * 1787 * @mpls_features: Mask of features inheritable by MPLS 1788 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1789 * 1790 * @ifindex: interface index 1791 * @group: The group the device belongs to 1792 * 1793 * @stats: Statistics struct, which was left as a legacy, use 1794 * rtnl_link_stats64 instead 1795 * 1796 * @core_stats: core networking counters, 1797 * do not use this in drivers 1798 * @carrier_up_count: Number of times the carrier has been up 1799 * @carrier_down_count: Number of times the carrier has been down 1800 * 1801 * @wireless_handlers: List of functions to handle Wireless Extensions, 1802 * instead of ioctl, 1803 * see <net/iw_handler.h> for details. 1804 * 1805 * @netdev_ops: Includes several pointers to callbacks, 1806 * if one wants to override the ndo_*() functions 1807 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks. 1808 * @xsk_tx_metadata_ops: Includes pointers to AF_XDP TX metadata callbacks. 1809 * @ethtool_ops: Management operations 1810 * @l3mdev_ops: Layer 3 master device operations 1811 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1812 * discovery handling. Necessary for e.g. 6LoWPAN. 1813 * @xfrmdev_ops: Transformation offload operations 1814 * @tlsdev_ops: Transport Layer Security offload operations 1815 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1816 * of Layer 2 headers. 1817 * 1818 * @flags: Interface flags (a la BSD) 1819 * @xdp_features: XDP capability supported by the device 1820 * @gflags: Global flags ( kept as legacy ) 1821 * @priv_len: Size of the ->priv flexible array 1822 * @priv: Flexible array containing private data 1823 * @operstate: RFC2863 operstate 1824 * @link_mode: Mapping policy to operstate 1825 * @if_port: Selectable AUI, TP, ... 1826 * @dma: DMA channel 1827 * @mtu: Interface MTU value 1828 * @min_mtu: Interface Minimum MTU value 1829 * @max_mtu: Interface Maximum MTU value 1830 * @type: Interface hardware type 1831 * @hard_header_len: Maximum hardware header length. 1832 * @min_header_len: Minimum hardware header length 1833 * 1834 * @needed_headroom: Extra headroom the hardware may need, but not in all 1835 * cases can this be guaranteed 1836 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1837 * cases can this be guaranteed. Some cases also use 1838 * LL_MAX_HEADER instead to allocate the skb 1839 * 1840 * interface address info: 1841 * 1842 * @perm_addr: Permanent hw address 1843 * @addr_assign_type: Hw address assignment type 1844 * @addr_len: Hardware address length 1845 * @upper_level: Maximum depth level of upper devices. 1846 * @lower_level: Maximum depth level of lower devices. 1847 * @neigh_priv_len: Used in neigh_alloc() 1848 * @dev_id: Used to differentiate devices that share 1849 * the same link layer address 1850 * @dev_port: Used to differentiate devices that share 1851 * the same function 1852 * @addr_list_lock: XXX: need comments on this one 1853 * @name_assign_type: network interface name assignment type 1854 * @uc_promisc: Counter that indicates promiscuous mode 1855 * has been enabled due to the need to listen to 1856 * additional unicast addresses in a device that 1857 * does not implement ndo_set_rx_mode() 1858 * @uc: unicast mac addresses 1859 * @mc: multicast mac addresses 1860 * @dev_addrs: list of device hw addresses 1861 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1862 * @promiscuity: Number of times the NIC is told to work in 1863 * promiscuous mode; if it becomes 0 the NIC will 1864 * exit promiscuous mode 1865 * @allmulti: Counter, enables or disables allmulticast mode 1866 * 1867 * @vlan_info: VLAN info 1868 * @dsa_ptr: dsa specific data 1869 * @tipc_ptr: TIPC specific data 1870 * @atalk_ptr: AppleTalk link 1871 * @ip_ptr: IPv4 specific data 1872 * @ip6_ptr: IPv6 specific data 1873 * @ax25_ptr: AX.25 specific data 1874 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1875 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1876 * device struct 1877 * @mpls_ptr: mpls_dev struct pointer 1878 * @mctp_ptr: MCTP specific data 1879 * 1880 * @dev_addr: Hw address (before bcast, 1881 * because most packets are unicast) 1882 * 1883 * @_rx: Array of RX queues 1884 * @num_rx_queues: Number of RX queues 1885 * allocated at register_netdev() time 1886 * @real_num_rx_queues: Number of RX queues currently active in device 1887 * @xdp_prog: XDP sockets filter program pointer 1888 * 1889 * @rx_handler: handler for received packets 1890 * @rx_handler_data: XXX: need comments on this one 1891 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing 1892 * @ingress_queue: XXX: need comments on this one 1893 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1894 * @broadcast: hw bcast address 1895 * 1896 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1897 * indexed by RX queue number. Assigned by driver. 1898 * This must only be set if the ndo_rx_flow_steer 1899 * operation is defined 1900 * @index_hlist: Device index hash chain 1901 * 1902 * @_tx: Array of TX queues 1903 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1904 * @real_num_tx_queues: Number of TX queues currently active in device 1905 * @qdisc: Root qdisc from userspace point of view 1906 * @tx_queue_len: Max frames per queue allowed 1907 * @tx_global_lock: XXX: need comments on this one 1908 * @xdp_bulkq: XDP device bulk queue 1909 * @xps_maps: all CPUs/RXQs maps for XPS device 1910 * 1911 * @xps_maps: XXX: need comments on this one 1912 * @tcx_egress: BPF & clsact qdisc specific data for 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 * 1939 * @pcpu_stat_type: Type of device statistics which the core should 1940 * allocate/free: none, lstats, tstats, dstats. none 1941 * means the driver is handling statistics allocation/ 1942 * freeing internally. 1943 * @lstats: Loopback statistics: packets, bytes 1944 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes 1945 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes 1946 * 1947 * @garp_port: GARP 1948 * @mrp_port: MRP 1949 * 1950 * @dm_private: Drop monitor private 1951 * 1952 * @dev: Class/net/name entry 1953 * @sysfs_groups: Space for optional device, statistics and wireless 1954 * sysfs groups 1955 * 1956 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1957 * @rtnl_link_ops: Rtnl_link_ops 1958 * @stat_ops: Optional ops for queue-aware statistics 1959 * @queue_mgmt_ops: Optional ops for queue management 1960 * 1961 * @gso_max_size: Maximum size of generic segmentation offload 1962 * @tso_max_size: Device (as in HW) limit on the max TSO request size 1963 * @gso_max_segs: Maximum number of segments that can be passed to the 1964 * NIC for GSO 1965 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count 1966 * @gso_ipv4_max_size: Maximum size of generic segmentation offload, 1967 * for IPv4. 1968 * 1969 * @dcbnl_ops: Data Center Bridging netlink ops 1970 * @num_tc: Number of traffic classes in the net device 1971 * @tc_to_txq: XXX: need comments on this one 1972 * @prio_tc_map: XXX: need comments on this one 1973 * 1974 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1975 * 1976 * @priomap: XXX: need comments on this one 1977 * @link_topo: Physical link topology tracking attached PHYs 1978 * @phydev: Physical device may attach itself 1979 * for hardware timestamping 1980 * @sfp_bus: attached &struct sfp_bus structure. 1981 * 1982 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1983 * 1984 * @proto_down: protocol port state information can be sent to the 1985 * switch driver and used to set the phys state of the 1986 * switch port. 1987 * 1988 * @threaded: napi threaded mode is enabled 1989 * 1990 * @see_all_hwtstamp_requests: device wants to see calls to 1991 * ndo_hwtstamp_set() for all timestamp requests 1992 * regardless of source, even if those aren't 1993 * HWTSTAMP_SOURCE_NETDEV 1994 * @change_proto_down: device supports setting carrier via IFLA_PROTO_DOWN 1995 * @netns_local: interface can't change network namespaces 1996 * @fcoe_mtu: device supports maximum FCoE MTU, 2158 bytes 1997 * 1998 * @net_notifier_list: List of per-net netdev notifier block 1999 * that follow this device when it is moved 2000 * to another network namespace. 2001 * 2002 * @macsec_ops: MACsec offloading ops 2003 * 2004 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 2005 * offload capabilities of the device 2006 * @udp_tunnel_nic: UDP tunnel offload state 2007 * @ethtool: ethtool related state 2008 * @xdp_state: stores info on attached XDP BPF programs 2009 * 2010 * @nested_level: Used as a parameter of spin_lock_nested() of 2011 * dev->addr_list_lock. 2012 * @unlink_list: As netif_addr_lock() can be called recursively, 2013 * keep a list of interfaces to be deleted. 2014 * @gro_max_size: Maximum size of aggregated packet in generic 2015 * receive offload (GRO) 2016 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic 2017 * receive offload (GRO), for IPv4. 2018 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP 2019 * zero copy driver 2020 * 2021 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 2022 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 2023 * @watchdog_dev_tracker: refcount tracker used by watchdog. 2024 * @dev_registered_tracker: tracker for reference held while 2025 * registered 2026 * @offload_xstats_l3: L3 HW stats for this netdevice. 2027 * 2028 * @devlink_port: Pointer to related devlink port structure. 2029 * Assigned by a driver before netdev registration using 2030 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static 2031 * during the time netdevice is registered. 2032 * 2033 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem, 2034 * where the clock is recovered. 2035 * 2036 * @max_pacing_offload_horizon: max EDT offload horizon in nsec. 2037 * @napi_config: An array of napi_config structures containing per-NAPI 2038 * settings. 2039 * @gro_flush_timeout: timeout for GRO layer in NAPI 2040 * @napi_defer_hard_irqs: If not zero, provides a counter that would 2041 * allow to avoid NIC hard IRQ, on busy queues. 2042 * 2043 * @neighbours: List heads pointing to this device's neighbours' 2044 * dev_list, one per address-family. 2045 * @hwprov: Tracks which PTP performs hardware packet time stamping. 2046 * 2047 * FIXME: cleanup struct net_device such that network protocol info 2048 * moves out. 2049 */ 2050 2051 struct net_device { 2052 /* Cacheline organization can be found documented in 2053 * Documentation/networking/net_cachelines/net_device.rst. 2054 * Please update the document when adding new fields. 2055 */ 2056 2057 /* TX read-mostly hotpath */ 2058 __cacheline_group_begin(net_device_read_tx); 2059 struct_group(priv_flags_fast, 2060 unsigned long priv_flags:32; 2061 unsigned long lltx:1; 2062 ); 2063 const struct net_device_ops *netdev_ops; 2064 const struct header_ops *header_ops; 2065 struct netdev_queue *_tx; 2066 netdev_features_t gso_partial_features; 2067 unsigned int real_num_tx_queues; 2068 unsigned int gso_max_size; 2069 unsigned int gso_ipv4_max_size; 2070 u16 gso_max_segs; 2071 s16 num_tc; 2072 /* Note : dev->mtu is often read without holding a lock. 2073 * Writers usually hold RTNL. 2074 * It is recommended to use READ_ONCE() to annotate the reads, 2075 * and to use WRITE_ONCE() to annotate the writes. 2076 */ 2077 unsigned int mtu; 2078 unsigned short needed_headroom; 2079 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2080 #ifdef CONFIG_XPS 2081 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2082 #endif 2083 #ifdef CONFIG_NETFILTER_EGRESS 2084 struct nf_hook_entries __rcu *nf_hooks_egress; 2085 #endif 2086 #ifdef CONFIG_NET_XGRESS 2087 struct bpf_mprog_entry __rcu *tcx_egress; 2088 #endif 2089 __cacheline_group_end(net_device_read_tx); 2090 2091 /* TXRX read-mostly hotpath */ 2092 __cacheline_group_begin(net_device_read_txrx); 2093 union { 2094 struct pcpu_lstats __percpu *lstats; 2095 struct pcpu_sw_netstats __percpu *tstats; 2096 struct pcpu_dstats __percpu *dstats; 2097 }; 2098 unsigned long state; 2099 unsigned int flags; 2100 unsigned short hard_header_len; 2101 netdev_features_t features; 2102 struct inet6_dev __rcu *ip6_ptr; 2103 __cacheline_group_end(net_device_read_txrx); 2104 2105 /* RX read-mostly hotpath */ 2106 __cacheline_group_begin(net_device_read_rx); 2107 struct bpf_prog __rcu *xdp_prog; 2108 struct list_head ptype_specific; 2109 int ifindex; 2110 unsigned int real_num_rx_queues; 2111 struct netdev_rx_queue *_rx; 2112 unsigned int gro_max_size; 2113 unsigned int gro_ipv4_max_size; 2114 rx_handler_func_t __rcu *rx_handler; 2115 void __rcu *rx_handler_data; 2116 possible_net_t nd_net; 2117 #ifdef CONFIG_NETPOLL 2118 struct netpoll_info __rcu *npinfo; 2119 #endif 2120 #ifdef CONFIG_NET_XGRESS 2121 struct bpf_mprog_entry __rcu *tcx_ingress; 2122 #endif 2123 __cacheline_group_end(net_device_read_rx); 2124 2125 char name[IFNAMSIZ]; 2126 struct netdev_name_node *name_node; 2127 struct dev_ifalias __rcu *ifalias; 2128 /* 2129 * I/O specific fields 2130 * FIXME: Merge these and struct ifmap into one 2131 */ 2132 unsigned long mem_end; 2133 unsigned long mem_start; 2134 unsigned long base_addr; 2135 2136 /* 2137 * Some hardware also needs these fields (state,dev_list, 2138 * napi_list,unreg_list,close_list) but they are not 2139 * part of the usual set specified in Space.c. 2140 */ 2141 2142 2143 struct list_head dev_list; 2144 struct list_head napi_list; 2145 struct list_head unreg_list; 2146 struct list_head close_list; 2147 struct list_head ptype_all; 2148 2149 struct { 2150 struct list_head upper; 2151 struct list_head lower; 2152 } adj_list; 2153 2154 /* Read-mostly cache-line for fast-path access */ 2155 xdp_features_t xdp_features; 2156 const struct xdp_metadata_ops *xdp_metadata_ops; 2157 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops; 2158 unsigned short gflags; 2159 2160 unsigned short needed_tailroom; 2161 2162 netdev_features_t hw_features; 2163 netdev_features_t wanted_features; 2164 netdev_features_t vlan_features; 2165 netdev_features_t hw_enc_features; 2166 netdev_features_t mpls_features; 2167 2168 unsigned int min_mtu; 2169 unsigned int max_mtu; 2170 unsigned short type; 2171 unsigned char min_header_len; 2172 unsigned char name_assign_type; 2173 2174 int group; 2175 2176 struct net_device_stats stats; /* not used by modern drivers */ 2177 2178 struct net_device_core_stats __percpu *core_stats; 2179 2180 /* Stats to monitor link on/off, flapping */ 2181 atomic_t carrier_up_count; 2182 atomic_t carrier_down_count; 2183 2184 #ifdef CONFIG_WIRELESS_EXT 2185 const struct iw_handler_def *wireless_handlers; 2186 #endif 2187 const struct ethtool_ops *ethtool_ops; 2188 #ifdef CONFIG_NET_L3_MASTER_DEV 2189 const struct l3mdev_ops *l3mdev_ops; 2190 #endif 2191 #if IS_ENABLED(CONFIG_IPV6) 2192 const struct ndisc_ops *ndisc_ops; 2193 #endif 2194 2195 #ifdef CONFIG_XFRM_OFFLOAD 2196 const struct xfrmdev_ops *xfrmdev_ops; 2197 #endif 2198 2199 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2200 const struct tlsdev_ops *tlsdev_ops; 2201 #endif 2202 2203 unsigned int operstate; 2204 unsigned char link_mode; 2205 2206 unsigned char if_port; 2207 unsigned char dma; 2208 2209 /* Interface address info. */ 2210 unsigned char perm_addr[MAX_ADDR_LEN]; 2211 unsigned char addr_assign_type; 2212 unsigned char addr_len; 2213 unsigned char upper_level; 2214 unsigned char lower_level; 2215 2216 unsigned short neigh_priv_len; 2217 unsigned short dev_id; 2218 unsigned short dev_port; 2219 int irq; 2220 u32 priv_len; 2221 2222 spinlock_t addr_list_lock; 2223 2224 struct netdev_hw_addr_list uc; 2225 struct netdev_hw_addr_list mc; 2226 struct netdev_hw_addr_list dev_addrs; 2227 2228 #ifdef CONFIG_SYSFS 2229 struct kset *queues_kset; 2230 #endif 2231 #ifdef CONFIG_LOCKDEP 2232 struct list_head unlink_list; 2233 #endif 2234 unsigned int promiscuity; 2235 unsigned int allmulti; 2236 bool uc_promisc; 2237 #ifdef CONFIG_LOCKDEP 2238 unsigned char nested_level; 2239 #endif 2240 2241 2242 /* Protocol-specific pointers */ 2243 struct in_device __rcu *ip_ptr; 2244 /** @fib_nh_head: nexthops associated with this netdev */ 2245 struct hlist_head fib_nh_head; 2246 2247 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2248 struct vlan_info __rcu *vlan_info; 2249 #endif 2250 #if IS_ENABLED(CONFIG_NET_DSA) 2251 struct dsa_port *dsa_ptr; 2252 #endif 2253 #if IS_ENABLED(CONFIG_TIPC) 2254 struct tipc_bearer __rcu *tipc_ptr; 2255 #endif 2256 #if IS_ENABLED(CONFIG_ATALK) 2257 void *atalk_ptr; 2258 #endif 2259 #if IS_ENABLED(CONFIG_AX25) 2260 struct ax25_dev __rcu *ax25_ptr; 2261 #endif 2262 #if IS_ENABLED(CONFIG_CFG80211) 2263 struct wireless_dev *ieee80211_ptr; 2264 #endif 2265 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2266 struct wpan_dev *ieee802154_ptr; 2267 #endif 2268 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2269 struct mpls_dev __rcu *mpls_ptr; 2270 #endif 2271 #if IS_ENABLED(CONFIG_MCTP) 2272 struct mctp_dev __rcu *mctp_ptr; 2273 #endif 2274 2275 /* 2276 * Cache lines mostly used on receive path (including eth_type_trans()) 2277 */ 2278 /* Interface address info used in eth_type_trans() */ 2279 const unsigned char *dev_addr; 2280 2281 unsigned int num_rx_queues; 2282 #define GRO_LEGACY_MAX_SIZE 65536u 2283 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2284 * and shinfo->gso_segs is a 16bit field. 2285 */ 2286 #define GRO_MAX_SIZE (8 * 65535u) 2287 unsigned int xdp_zc_max_segs; 2288 struct netdev_queue __rcu *ingress_queue; 2289 #ifdef CONFIG_NETFILTER_INGRESS 2290 struct nf_hook_entries __rcu *nf_hooks_ingress; 2291 #endif 2292 2293 unsigned char broadcast[MAX_ADDR_LEN]; 2294 #ifdef CONFIG_RFS_ACCEL 2295 struct cpu_rmap *rx_cpu_rmap; 2296 #endif 2297 struct hlist_node index_hlist; 2298 2299 /* 2300 * Cache lines mostly used on transmit path 2301 */ 2302 unsigned int num_tx_queues; 2303 struct Qdisc __rcu *qdisc; 2304 unsigned int tx_queue_len; 2305 spinlock_t tx_global_lock; 2306 2307 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2308 2309 #ifdef CONFIG_NET_SCHED 2310 DECLARE_HASHTABLE (qdisc_hash, 4); 2311 #endif 2312 /* These may be needed for future network-power-down code. */ 2313 struct timer_list watchdog_timer; 2314 int watchdog_timeo; 2315 2316 u32 proto_down_reason; 2317 2318 struct list_head todo_list; 2319 2320 #ifdef CONFIG_PCPU_DEV_REFCNT 2321 int __percpu *pcpu_refcnt; 2322 #else 2323 refcount_t dev_refcnt; 2324 #endif 2325 struct ref_tracker_dir refcnt_tracker; 2326 2327 struct list_head link_watch_list; 2328 2329 u8 reg_state; 2330 2331 bool dismantle; 2332 2333 enum { 2334 RTNL_LINK_INITIALIZED, 2335 RTNL_LINK_INITIALIZING, 2336 } rtnl_link_state:16; 2337 2338 bool needs_free_netdev; 2339 void (*priv_destructor)(struct net_device *dev); 2340 2341 /* mid-layer private */ 2342 void *ml_priv; 2343 enum netdev_ml_priv_type ml_priv_type; 2344 2345 enum netdev_stat_type pcpu_stat_type:8; 2346 2347 #if IS_ENABLED(CONFIG_GARP) 2348 struct garp_port __rcu *garp_port; 2349 #endif 2350 #if IS_ENABLED(CONFIG_MRP) 2351 struct mrp_port __rcu *mrp_port; 2352 #endif 2353 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2354 struct dm_hw_stat_delta __rcu *dm_private; 2355 #endif 2356 struct device dev; 2357 const struct attribute_group *sysfs_groups[4]; 2358 const struct attribute_group *sysfs_rx_queue_group; 2359 2360 const struct rtnl_link_ops *rtnl_link_ops; 2361 2362 const struct netdev_stat_ops *stat_ops; 2363 2364 const struct netdev_queue_mgmt_ops *queue_mgmt_ops; 2365 2366 /* for setting kernel sock attribute on TCP connection setup */ 2367 #define GSO_MAX_SEGS 65535u 2368 #define GSO_LEGACY_MAX_SIZE 65536u 2369 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2370 * and shinfo->gso_segs is a 16bit field. 2371 */ 2372 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2373 2374 #define TSO_LEGACY_MAX_SIZE 65536 2375 #define TSO_MAX_SIZE UINT_MAX 2376 unsigned int tso_max_size; 2377 #define TSO_MAX_SEGS U16_MAX 2378 u16 tso_max_segs; 2379 2380 #ifdef CONFIG_DCB 2381 const struct dcbnl_rtnl_ops *dcbnl_ops; 2382 #endif 2383 u8 prio_tc_map[TC_BITMASK + 1]; 2384 2385 #if IS_ENABLED(CONFIG_FCOE) 2386 unsigned int fcoe_ddp_xid; 2387 #endif 2388 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2389 struct netprio_map __rcu *priomap; 2390 #endif 2391 struct phy_link_topology *link_topo; 2392 struct phy_device *phydev; 2393 struct sfp_bus *sfp_bus; 2394 struct lock_class_key *qdisc_tx_busylock; 2395 bool proto_down; 2396 bool threaded; 2397 2398 /* priv_flags_slow, ungrouped to save space */ 2399 unsigned long see_all_hwtstamp_requests:1; 2400 unsigned long change_proto_down:1; 2401 unsigned long netns_local:1; 2402 unsigned long fcoe_mtu:1; 2403 2404 struct list_head net_notifier_list; 2405 2406 #if IS_ENABLED(CONFIG_MACSEC) 2407 /* MACsec management functions */ 2408 const struct macsec_ops *macsec_ops; 2409 #endif 2410 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2411 struct udp_tunnel_nic *udp_tunnel_nic; 2412 2413 struct ethtool_netdev_state *ethtool; 2414 2415 /* protected by rtnl_lock */ 2416 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2417 2418 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2419 netdevice_tracker linkwatch_dev_tracker; 2420 netdevice_tracker watchdog_dev_tracker; 2421 netdevice_tracker dev_registered_tracker; 2422 struct rtnl_hw_stats64 *offload_xstats_l3; 2423 2424 struct devlink_port *devlink_port; 2425 2426 #if IS_ENABLED(CONFIG_DPLL) 2427 struct dpll_pin __rcu *dpll_pin; 2428 #endif 2429 #if IS_ENABLED(CONFIG_PAGE_POOL) 2430 /** @page_pools: page pools created for this netdevice */ 2431 struct hlist_head page_pools; 2432 #endif 2433 2434 /** @irq_moder: dim parameters used if IS_ENABLED(CONFIG_DIMLIB). */ 2435 struct dim_irq_moder *irq_moder; 2436 2437 u64 max_pacing_offload_horizon; 2438 struct napi_config *napi_config; 2439 unsigned long gro_flush_timeout; 2440 u32 napi_defer_hard_irqs; 2441 2442 /** 2443 * @up: copy of @state's IFF_UP, but safe to read with just @lock. 2444 * May report false negatives while the device is being opened 2445 * or closed (@lock does not protect .ndo_open, or .ndo_close). 2446 */ 2447 bool up; 2448 2449 /** 2450 * @lock: netdev-scope lock, protects a small selection of fields. 2451 * Should always be taken using netdev_lock() / netdev_unlock() helpers. 2452 * Drivers are free to use it for other protection. 2453 * 2454 * Protects: 2455 * @gro_flush_timeout, @napi_defer_hard_irqs, @napi_list, 2456 * @net_shaper_hierarchy, @reg_state, @threaded 2457 * 2458 * Partially protects (writers must hold both @lock and rtnl_lock): 2459 * @up 2460 * 2461 * Also protects some fields in struct napi_struct. 2462 * 2463 * Ordering: take after rtnl_lock. 2464 */ 2465 struct mutex lock; 2466 2467 #if IS_ENABLED(CONFIG_NET_SHAPER) 2468 /** 2469 * @net_shaper_hierarchy: data tracking the current shaper status 2470 * see include/net/net_shapers.h 2471 */ 2472 struct net_shaper_hierarchy *net_shaper_hierarchy; 2473 #endif 2474 2475 struct hlist_head neighbours[NEIGH_NR_TABLES]; 2476 2477 struct hwtstamp_provider __rcu *hwprov; 2478 2479 u8 priv[] ____cacheline_aligned 2480 __counted_by(priv_len); 2481 } ____cacheline_aligned; 2482 #define to_net_dev(d) container_of(d, struct net_device, dev) 2483 2484 /* 2485 * Driver should use this to assign devlink port instance to a netdevice 2486 * before it registers the netdevice. Therefore devlink_port is static 2487 * during the netdev lifetime after it is registered. 2488 */ 2489 #define SET_NETDEV_DEVLINK_PORT(dev, port) \ 2490 ({ \ 2491 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \ 2492 ((dev)->devlink_port = (port)); \ 2493 }) 2494 2495 static inline bool netif_elide_gro(const struct net_device *dev) 2496 { 2497 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2498 return true; 2499 return false; 2500 } 2501 2502 #define NETDEV_ALIGN 32 2503 2504 static inline 2505 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2506 { 2507 return dev->prio_tc_map[prio & TC_BITMASK]; 2508 } 2509 2510 static inline 2511 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2512 { 2513 if (tc >= dev->num_tc) 2514 return -EINVAL; 2515 2516 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2517 return 0; 2518 } 2519 2520 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2521 void netdev_reset_tc(struct net_device *dev); 2522 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2523 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2524 2525 static inline 2526 int netdev_get_num_tc(struct net_device *dev) 2527 { 2528 return dev->num_tc; 2529 } 2530 2531 static inline void net_prefetch(void *p) 2532 { 2533 prefetch(p); 2534 #if L1_CACHE_BYTES < 128 2535 prefetch((u8 *)p + L1_CACHE_BYTES); 2536 #endif 2537 } 2538 2539 static inline void net_prefetchw(void *p) 2540 { 2541 prefetchw(p); 2542 #if L1_CACHE_BYTES < 128 2543 prefetchw((u8 *)p + L1_CACHE_BYTES); 2544 #endif 2545 } 2546 2547 void netdev_unbind_sb_channel(struct net_device *dev, 2548 struct net_device *sb_dev); 2549 int netdev_bind_sb_channel_queue(struct net_device *dev, 2550 struct net_device *sb_dev, 2551 u8 tc, u16 count, u16 offset); 2552 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2553 static inline int netdev_get_sb_channel(struct net_device *dev) 2554 { 2555 return max_t(int, -dev->num_tc, 0); 2556 } 2557 2558 static inline 2559 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2560 unsigned int index) 2561 { 2562 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues); 2563 return &dev->_tx[index]; 2564 } 2565 2566 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2567 const struct sk_buff *skb) 2568 { 2569 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2570 } 2571 2572 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2573 void (*f)(struct net_device *, 2574 struct netdev_queue *, 2575 void *), 2576 void *arg) 2577 { 2578 unsigned int i; 2579 2580 for (i = 0; i < dev->num_tx_queues; i++) 2581 f(dev, &dev->_tx[i], arg); 2582 } 2583 2584 #define netdev_lockdep_set_classes(dev) \ 2585 { \ 2586 static struct lock_class_key qdisc_tx_busylock_key; \ 2587 static struct lock_class_key qdisc_xmit_lock_key; \ 2588 static struct lock_class_key dev_addr_list_lock_key; \ 2589 unsigned int i; \ 2590 \ 2591 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2592 lockdep_set_class(&(dev)->addr_list_lock, \ 2593 &dev_addr_list_lock_key); \ 2594 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2595 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2596 &qdisc_xmit_lock_key); \ 2597 } 2598 2599 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2600 struct net_device *sb_dev); 2601 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2602 struct sk_buff *skb, 2603 struct net_device *sb_dev); 2604 2605 /* returns the headroom that the master device needs to take in account 2606 * when forwarding to this dev 2607 */ 2608 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2609 { 2610 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2611 } 2612 2613 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2614 { 2615 if (dev->netdev_ops->ndo_set_rx_headroom) 2616 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2617 } 2618 2619 /* set the device rx headroom to the dev's default */ 2620 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2621 { 2622 netdev_set_rx_headroom(dev, -1); 2623 } 2624 2625 static inline void *netdev_get_ml_priv(struct net_device *dev, 2626 enum netdev_ml_priv_type type) 2627 { 2628 if (dev->ml_priv_type != type) 2629 return NULL; 2630 2631 return dev->ml_priv; 2632 } 2633 2634 static inline void netdev_set_ml_priv(struct net_device *dev, 2635 void *ml_priv, 2636 enum netdev_ml_priv_type type) 2637 { 2638 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2639 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2640 dev->ml_priv_type, type); 2641 WARN(!dev->ml_priv_type && dev->ml_priv, 2642 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2643 2644 dev->ml_priv = ml_priv; 2645 dev->ml_priv_type = type; 2646 } 2647 2648 /* 2649 * Net namespace inlines 2650 */ 2651 static inline 2652 struct net *dev_net(const struct net_device *dev) 2653 { 2654 return read_pnet(&dev->nd_net); 2655 } 2656 2657 static inline 2658 void dev_net_set(struct net_device *dev, struct net *net) 2659 { 2660 write_pnet(&dev->nd_net, net); 2661 } 2662 2663 /** 2664 * netdev_priv - access network device private data 2665 * @dev: network device 2666 * 2667 * Get network device private data 2668 */ 2669 static inline void *netdev_priv(const struct net_device *dev) 2670 { 2671 return (void *)dev->priv; 2672 } 2673 2674 /* Set the sysfs physical device reference for the network logical device 2675 * if set prior to registration will cause a symlink during initialization. 2676 */ 2677 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2678 2679 /* Set the sysfs device type for the network logical device to allow 2680 * fine-grained identification of different network device types. For 2681 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2682 */ 2683 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2684 2685 void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index, 2686 enum netdev_queue_type type, 2687 struct napi_struct *napi); 2688 2689 static inline void netdev_lock(struct net_device *dev) 2690 { 2691 mutex_lock(&dev->lock); 2692 } 2693 2694 static inline void netdev_unlock(struct net_device *dev) 2695 { 2696 mutex_unlock(&dev->lock); 2697 } 2698 2699 static inline void netdev_assert_locked(struct net_device *dev) 2700 { 2701 lockdep_assert_held(&dev->lock); 2702 } 2703 2704 static inline void netdev_assert_locked_or_invisible(struct net_device *dev) 2705 { 2706 if (dev->reg_state == NETREG_REGISTERED || 2707 dev->reg_state == NETREG_UNREGISTERING) 2708 netdev_assert_locked(dev); 2709 } 2710 2711 static inline void netif_napi_set_irq_locked(struct napi_struct *napi, int irq) 2712 { 2713 napi->irq = irq; 2714 } 2715 2716 static inline void netif_napi_set_irq(struct napi_struct *napi, int irq) 2717 { 2718 netdev_lock(napi->dev); 2719 netif_napi_set_irq_locked(napi, irq); 2720 netdev_unlock(napi->dev); 2721 } 2722 2723 /* Default NAPI poll() weight 2724 * Device drivers are strongly advised to not use bigger value 2725 */ 2726 #define NAPI_POLL_WEIGHT 64 2727 2728 void netif_napi_add_weight_locked(struct net_device *dev, 2729 struct napi_struct *napi, 2730 int (*poll)(struct napi_struct *, int), 2731 int weight); 2732 2733 static inline void 2734 netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2735 int (*poll)(struct napi_struct *, int), int weight) 2736 { 2737 netdev_lock(dev); 2738 netif_napi_add_weight_locked(dev, napi, poll, weight); 2739 netdev_unlock(dev); 2740 } 2741 2742 /** 2743 * netif_napi_add() - initialize a NAPI context 2744 * @dev: network device 2745 * @napi: NAPI context 2746 * @poll: polling function 2747 * 2748 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2749 * *any* of the other NAPI-related functions. 2750 */ 2751 static inline void 2752 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2753 int (*poll)(struct napi_struct *, int)) 2754 { 2755 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2756 } 2757 2758 static inline void 2759 netif_napi_add_locked(struct net_device *dev, struct napi_struct *napi, 2760 int (*poll)(struct napi_struct *, int)) 2761 { 2762 netif_napi_add_weight_locked(dev, napi, poll, NAPI_POLL_WEIGHT); 2763 } 2764 2765 static inline void 2766 netif_napi_add_tx_weight(struct net_device *dev, 2767 struct napi_struct *napi, 2768 int (*poll)(struct napi_struct *, int), 2769 int weight) 2770 { 2771 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2772 netif_napi_add_weight(dev, napi, poll, weight); 2773 } 2774 2775 static inline void 2776 netif_napi_add_config_locked(struct net_device *dev, struct napi_struct *napi, 2777 int (*poll)(struct napi_struct *, int), int index) 2778 { 2779 napi->index = index; 2780 napi->config = &dev->napi_config[index]; 2781 netif_napi_add_weight_locked(dev, napi, poll, NAPI_POLL_WEIGHT); 2782 } 2783 2784 /** 2785 * netif_napi_add_config - initialize a NAPI context with persistent config 2786 * @dev: network device 2787 * @napi: NAPI context 2788 * @poll: polling function 2789 * @index: the NAPI index 2790 */ 2791 static inline void 2792 netif_napi_add_config(struct net_device *dev, struct napi_struct *napi, 2793 int (*poll)(struct napi_struct *, int), int index) 2794 { 2795 netdev_lock(dev); 2796 netif_napi_add_config_locked(dev, napi, poll, index); 2797 netdev_unlock(dev); 2798 } 2799 2800 /** 2801 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2802 * @dev: network device 2803 * @napi: NAPI context 2804 * @poll: polling function 2805 * 2806 * This variant of netif_napi_add() should be used from drivers using NAPI 2807 * to exclusively poll a TX queue. 2808 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2809 */ 2810 static inline void netif_napi_add_tx(struct net_device *dev, 2811 struct napi_struct *napi, 2812 int (*poll)(struct napi_struct *, int)) 2813 { 2814 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2815 } 2816 2817 void __netif_napi_del_locked(struct napi_struct *napi); 2818 2819 /** 2820 * __netif_napi_del - remove a NAPI context 2821 * @napi: NAPI context 2822 * 2823 * Warning: caller must observe RCU grace period before freeing memory 2824 * containing @napi. Drivers might want to call this helper to combine 2825 * all the needed RCU grace periods into a single one. 2826 */ 2827 static inline void __netif_napi_del(struct napi_struct *napi) 2828 { 2829 netdev_lock(napi->dev); 2830 __netif_napi_del_locked(napi); 2831 netdev_unlock(napi->dev); 2832 } 2833 2834 static inline void netif_napi_del_locked(struct napi_struct *napi) 2835 { 2836 __netif_napi_del_locked(napi); 2837 synchronize_net(); 2838 } 2839 2840 /** 2841 * netif_napi_del - remove a NAPI context 2842 * @napi: NAPI context 2843 * 2844 * netif_napi_del() removes a NAPI context from the network device NAPI list 2845 */ 2846 static inline void netif_napi_del(struct napi_struct *napi) 2847 { 2848 __netif_napi_del(napi); 2849 synchronize_net(); 2850 } 2851 2852 struct packet_type { 2853 __be16 type; /* This is really htons(ether_type). */ 2854 bool ignore_outgoing; 2855 struct net_device *dev; /* NULL is wildcarded here */ 2856 netdevice_tracker dev_tracker; 2857 int (*func) (struct sk_buff *, 2858 struct net_device *, 2859 struct packet_type *, 2860 struct net_device *); 2861 void (*list_func) (struct list_head *, 2862 struct packet_type *, 2863 struct net_device *); 2864 bool (*id_match)(struct packet_type *ptype, 2865 struct sock *sk); 2866 struct net *af_packet_net; 2867 void *af_packet_priv; 2868 struct list_head list; 2869 }; 2870 2871 struct offload_callbacks { 2872 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2873 netdev_features_t features); 2874 struct sk_buff *(*gro_receive)(struct list_head *head, 2875 struct sk_buff *skb); 2876 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2877 }; 2878 2879 struct packet_offload { 2880 __be16 type; /* This is really htons(ether_type). */ 2881 u16 priority; 2882 struct offload_callbacks callbacks; 2883 struct list_head list; 2884 }; 2885 2886 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2887 struct pcpu_sw_netstats { 2888 u64_stats_t rx_packets; 2889 u64_stats_t rx_bytes; 2890 u64_stats_t tx_packets; 2891 u64_stats_t tx_bytes; 2892 struct u64_stats_sync syncp; 2893 } __aligned(4 * sizeof(u64)); 2894 2895 struct pcpu_dstats { 2896 u64_stats_t rx_packets; 2897 u64_stats_t rx_bytes; 2898 u64_stats_t rx_drops; 2899 u64_stats_t tx_packets; 2900 u64_stats_t tx_bytes; 2901 u64_stats_t tx_drops; 2902 struct u64_stats_sync syncp; 2903 } __aligned(8 * sizeof(u64)); 2904 2905 struct pcpu_lstats { 2906 u64_stats_t packets; 2907 u64_stats_t bytes; 2908 struct u64_stats_sync syncp; 2909 } __aligned(2 * sizeof(u64)); 2910 2911 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2912 2913 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2914 { 2915 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2916 2917 u64_stats_update_begin(&tstats->syncp); 2918 u64_stats_add(&tstats->rx_bytes, len); 2919 u64_stats_inc(&tstats->rx_packets); 2920 u64_stats_update_end(&tstats->syncp); 2921 } 2922 2923 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2924 unsigned int packets, 2925 unsigned int len) 2926 { 2927 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2928 2929 u64_stats_update_begin(&tstats->syncp); 2930 u64_stats_add(&tstats->tx_bytes, len); 2931 u64_stats_add(&tstats->tx_packets, packets); 2932 u64_stats_update_end(&tstats->syncp); 2933 } 2934 2935 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2936 { 2937 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2938 2939 u64_stats_update_begin(&lstats->syncp); 2940 u64_stats_add(&lstats->bytes, len); 2941 u64_stats_inc(&lstats->packets); 2942 u64_stats_update_end(&lstats->syncp); 2943 } 2944 2945 static inline void dev_dstats_rx_add(struct net_device *dev, 2946 unsigned int len) 2947 { 2948 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 2949 2950 u64_stats_update_begin(&dstats->syncp); 2951 u64_stats_inc(&dstats->rx_packets); 2952 u64_stats_add(&dstats->rx_bytes, len); 2953 u64_stats_update_end(&dstats->syncp); 2954 } 2955 2956 static inline void dev_dstats_rx_dropped(struct net_device *dev) 2957 { 2958 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 2959 2960 u64_stats_update_begin(&dstats->syncp); 2961 u64_stats_inc(&dstats->rx_drops); 2962 u64_stats_update_end(&dstats->syncp); 2963 } 2964 2965 static inline void dev_dstats_tx_add(struct net_device *dev, 2966 unsigned int len) 2967 { 2968 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 2969 2970 u64_stats_update_begin(&dstats->syncp); 2971 u64_stats_inc(&dstats->tx_packets); 2972 u64_stats_add(&dstats->tx_bytes, len); 2973 u64_stats_update_end(&dstats->syncp); 2974 } 2975 2976 static inline void dev_dstats_tx_dropped(struct net_device *dev) 2977 { 2978 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 2979 2980 u64_stats_update_begin(&dstats->syncp); 2981 u64_stats_inc(&dstats->tx_drops); 2982 u64_stats_update_end(&dstats->syncp); 2983 } 2984 2985 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2986 ({ \ 2987 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2988 if (pcpu_stats) { \ 2989 int __cpu; \ 2990 for_each_possible_cpu(__cpu) { \ 2991 typeof(type) *stat; \ 2992 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2993 u64_stats_init(&stat->syncp); \ 2994 } \ 2995 } \ 2996 pcpu_stats; \ 2997 }) 2998 2999 #define netdev_alloc_pcpu_stats(type) \ 3000 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 3001 3002 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 3003 ({ \ 3004 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 3005 if (pcpu_stats) { \ 3006 int __cpu; \ 3007 for_each_possible_cpu(__cpu) { \ 3008 typeof(type) *stat; \ 3009 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 3010 u64_stats_init(&stat->syncp); \ 3011 } \ 3012 } \ 3013 pcpu_stats; \ 3014 }) 3015 3016 enum netdev_lag_tx_type { 3017 NETDEV_LAG_TX_TYPE_UNKNOWN, 3018 NETDEV_LAG_TX_TYPE_RANDOM, 3019 NETDEV_LAG_TX_TYPE_BROADCAST, 3020 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 3021 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 3022 NETDEV_LAG_TX_TYPE_HASH, 3023 }; 3024 3025 enum netdev_lag_hash { 3026 NETDEV_LAG_HASH_NONE, 3027 NETDEV_LAG_HASH_L2, 3028 NETDEV_LAG_HASH_L34, 3029 NETDEV_LAG_HASH_L23, 3030 NETDEV_LAG_HASH_E23, 3031 NETDEV_LAG_HASH_E34, 3032 NETDEV_LAG_HASH_VLAN_SRCMAC, 3033 NETDEV_LAG_HASH_UNKNOWN, 3034 }; 3035 3036 struct netdev_lag_upper_info { 3037 enum netdev_lag_tx_type tx_type; 3038 enum netdev_lag_hash hash_type; 3039 }; 3040 3041 struct netdev_lag_lower_state_info { 3042 u8 link_up : 1, 3043 tx_enabled : 1; 3044 }; 3045 3046 #include <linux/notifier.h> 3047 3048 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 3049 * and the rtnetlink notification exclusion list in rtnetlink_event() when 3050 * adding new types. 3051 */ 3052 enum netdev_cmd { 3053 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 3054 NETDEV_DOWN, 3055 NETDEV_REBOOT, /* Tell a protocol stack a network interface 3056 detected a hardware crash and restarted 3057 - we can use this eg to kick tcp sessions 3058 once done */ 3059 NETDEV_CHANGE, /* Notify device state change */ 3060 NETDEV_REGISTER, 3061 NETDEV_UNREGISTER, 3062 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 3063 NETDEV_CHANGEADDR, /* notify after the address change */ 3064 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 3065 NETDEV_GOING_DOWN, 3066 NETDEV_CHANGENAME, 3067 NETDEV_FEAT_CHANGE, 3068 NETDEV_BONDING_FAILOVER, 3069 NETDEV_PRE_UP, 3070 NETDEV_PRE_TYPE_CHANGE, 3071 NETDEV_POST_TYPE_CHANGE, 3072 NETDEV_POST_INIT, 3073 NETDEV_PRE_UNINIT, 3074 NETDEV_RELEASE, 3075 NETDEV_NOTIFY_PEERS, 3076 NETDEV_JOIN, 3077 NETDEV_CHANGEUPPER, 3078 NETDEV_RESEND_IGMP, 3079 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 3080 NETDEV_CHANGEINFODATA, 3081 NETDEV_BONDING_INFO, 3082 NETDEV_PRECHANGEUPPER, 3083 NETDEV_CHANGELOWERSTATE, 3084 NETDEV_UDP_TUNNEL_PUSH_INFO, 3085 NETDEV_UDP_TUNNEL_DROP_INFO, 3086 NETDEV_CHANGE_TX_QUEUE_LEN, 3087 NETDEV_CVLAN_FILTER_PUSH_INFO, 3088 NETDEV_CVLAN_FILTER_DROP_INFO, 3089 NETDEV_SVLAN_FILTER_PUSH_INFO, 3090 NETDEV_SVLAN_FILTER_DROP_INFO, 3091 NETDEV_OFFLOAD_XSTATS_ENABLE, 3092 NETDEV_OFFLOAD_XSTATS_DISABLE, 3093 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 3094 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 3095 NETDEV_XDP_FEAT_CHANGE, 3096 }; 3097 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 3098 3099 int register_netdevice_notifier(struct notifier_block *nb); 3100 int unregister_netdevice_notifier(struct notifier_block *nb); 3101 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 3102 int unregister_netdevice_notifier_net(struct net *net, 3103 struct notifier_block *nb); 3104 int register_netdevice_notifier_dev_net(struct net_device *dev, 3105 struct notifier_block *nb, 3106 struct netdev_net_notifier *nn); 3107 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 3108 struct notifier_block *nb, 3109 struct netdev_net_notifier *nn); 3110 3111 struct netdev_notifier_info { 3112 struct net_device *dev; 3113 struct netlink_ext_ack *extack; 3114 }; 3115 3116 struct netdev_notifier_info_ext { 3117 struct netdev_notifier_info info; /* must be first */ 3118 union { 3119 u32 mtu; 3120 } ext; 3121 }; 3122 3123 struct netdev_notifier_change_info { 3124 struct netdev_notifier_info info; /* must be first */ 3125 unsigned int flags_changed; 3126 }; 3127 3128 struct netdev_notifier_changeupper_info { 3129 struct netdev_notifier_info info; /* must be first */ 3130 struct net_device *upper_dev; /* new upper dev */ 3131 bool master; /* is upper dev master */ 3132 bool linking; /* is the notification for link or unlink */ 3133 void *upper_info; /* upper dev info */ 3134 }; 3135 3136 struct netdev_notifier_changelowerstate_info { 3137 struct netdev_notifier_info info; /* must be first */ 3138 void *lower_state_info; /* is lower dev state */ 3139 }; 3140 3141 struct netdev_notifier_pre_changeaddr_info { 3142 struct netdev_notifier_info info; /* must be first */ 3143 const unsigned char *dev_addr; 3144 }; 3145 3146 enum netdev_offload_xstats_type { 3147 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 3148 }; 3149 3150 struct netdev_notifier_offload_xstats_info { 3151 struct netdev_notifier_info info; /* must be first */ 3152 enum netdev_offload_xstats_type type; 3153 3154 union { 3155 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 3156 struct netdev_notifier_offload_xstats_rd *report_delta; 3157 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 3158 struct netdev_notifier_offload_xstats_ru *report_used; 3159 }; 3160 }; 3161 3162 int netdev_offload_xstats_enable(struct net_device *dev, 3163 enum netdev_offload_xstats_type type, 3164 struct netlink_ext_ack *extack); 3165 int netdev_offload_xstats_disable(struct net_device *dev, 3166 enum netdev_offload_xstats_type type); 3167 bool netdev_offload_xstats_enabled(const struct net_device *dev, 3168 enum netdev_offload_xstats_type type); 3169 int netdev_offload_xstats_get(struct net_device *dev, 3170 enum netdev_offload_xstats_type type, 3171 struct rtnl_hw_stats64 *stats, bool *used, 3172 struct netlink_ext_ack *extack); 3173 void 3174 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 3175 const struct rtnl_hw_stats64 *stats); 3176 void 3177 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 3178 void netdev_offload_xstats_push_delta(struct net_device *dev, 3179 enum netdev_offload_xstats_type type, 3180 const struct rtnl_hw_stats64 *stats); 3181 3182 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 3183 struct net_device *dev) 3184 { 3185 info->dev = dev; 3186 info->extack = NULL; 3187 } 3188 3189 static inline struct net_device * 3190 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 3191 { 3192 return info->dev; 3193 } 3194 3195 static inline struct netlink_ext_ack * 3196 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 3197 { 3198 return info->extack; 3199 } 3200 3201 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 3202 int call_netdevice_notifiers_info(unsigned long val, 3203 struct netdev_notifier_info *info); 3204 3205 #define for_each_netdev(net, d) \ 3206 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 3207 #define for_each_netdev_reverse(net, d) \ 3208 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 3209 #define for_each_netdev_rcu(net, d) \ 3210 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 3211 #define for_each_netdev_safe(net, d, n) \ 3212 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 3213 #define for_each_netdev_continue(net, d) \ 3214 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 3215 #define for_each_netdev_continue_reverse(net, d) \ 3216 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 3217 dev_list) 3218 #define for_each_netdev_continue_rcu(net, d) \ 3219 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 3220 #define for_each_netdev_in_bond_rcu(bond, slave) \ 3221 for_each_netdev_rcu(&init_net, slave) \ 3222 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 3223 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 3224 3225 #define for_each_netdev_dump(net, d, ifindex) \ 3226 for (; (d = xa_find(&(net)->dev_by_index, &ifindex, \ 3227 ULONG_MAX, XA_PRESENT)); ifindex++) 3228 3229 static inline struct net_device *next_net_device(struct net_device *dev) 3230 { 3231 struct list_head *lh; 3232 struct net *net; 3233 3234 net = dev_net(dev); 3235 lh = dev->dev_list.next; 3236 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3237 } 3238 3239 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 3240 { 3241 struct list_head *lh; 3242 struct net *net; 3243 3244 net = dev_net(dev); 3245 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 3246 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3247 } 3248 3249 static inline struct net_device *first_net_device(struct net *net) 3250 { 3251 return list_empty(&net->dev_base_head) ? NULL : 3252 net_device_entry(net->dev_base_head.next); 3253 } 3254 3255 static inline struct net_device *first_net_device_rcu(struct net *net) 3256 { 3257 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 3258 3259 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3260 } 3261 3262 int netdev_boot_setup_check(struct net_device *dev); 3263 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 3264 const char *hwaddr); 3265 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 3266 void dev_add_pack(struct packet_type *pt); 3267 void dev_remove_pack(struct packet_type *pt); 3268 void __dev_remove_pack(struct packet_type *pt); 3269 void dev_add_offload(struct packet_offload *po); 3270 void dev_remove_offload(struct packet_offload *po); 3271 3272 int dev_get_iflink(const struct net_device *dev); 3273 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 3274 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 3275 struct net_device_path_stack *stack); 3276 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 3277 unsigned short mask); 3278 struct net_device *dev_get_by_name(struct net *net, const char *name); 3279 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 3280 struct net_device *__dev_get_by_name(struct net *net, const char *name); 3281 bool netdev_name_in_use(struct net *net, const char *name); 3282 int dev_alloc_name(struct net_device *dev, const char *name); 3283 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 3284 void dev_close(struct net_device *dev); 3285 void dev_close_many(struct list_head *head, bool unlink); 3286 void dev_disable_lro(struct net_device *dev); 3287 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 3288 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 3289 struct net_device *sb_dev); 3290 3291 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3292 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3293 3294 static inline int dev_queue_xmit(struct sk_buff *skb) 3295 { 3296 return __dev_queue_xmit(skb, NULL); 3297 } 3298 3299 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3300 struct net_device *sb_dev) 3301 { 3302 return __dev_queue_xmit(skb, sb_dev); 3303 } 3304 3305 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3306 { 3307 int ret; 3308 3309 ret = __dev_direct_xmit(skb, queue_id); 3310 if (!dev_xmit_complete(ret)) 3311 kfree_skb(skb); 3312 return ret; 3313 } 3314 3315 int register_netdevice(struct net_device *dev); 3316 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3317 void unregister_netdevice_many(struct list_head *head); 3318 static inline void unregister_netdevice(struct net_device *dev) 3319 { 3320 unregister_netdevice_queue(dev, NULL); 3321 } 3322 3323 int netdev_refcnt_read(const struct net_device *dev); 3324 void free_netdev(struct net_device *dev); 3325 3326 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3327 struct sk_buff *skb, 3328 bool all_slaves); 3329 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3330 struct sock *sk); 3331 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3332 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3333 struct net_device *netdev_get_by_index(struct net *net, int ifindex, 3334 netdevice_tracker *tracker, gfp_t gfp); 3335 struct net_device *netdev_get_by_name(struct net *net, const char *name, 3336 netdevice_tracker *tracker, gfp_t gfp); 3337 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3338 void netdev_copy_name(struct net_device *dev, char *name); 3339 3340 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3341 unsigned short type, 3342 const void *daddr, const void *saddr, 3343 unsigned int len) 3344 { 3345 if (!dev->header_ops || !dev->header_ops->create) 3346 return 0; 3347 3348 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3349 } 3350 3351 static inline int dev_parse_header(const struct sk_buff *skb, 3352 unsigned char *haddr) 3353 { 3354 const struct net_device *dev = skb->dev; 3355 3356 if (!dev->header_ops || !dev->header_ops->parse) 3357 return 0; 3358 return dev->header_ops->parse(skb, haddr); 3359 } 3360 3361 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3362 { 3363 const struct net_device *dev = skb->dev; 3364 3365 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3366 return 0; 3367 return dev->header_ops->parse_protocol(skb); 3368 } 3369 3370 /* ll_header must have at least hard_header_len allocated */ 3371 static inline bool dev_validate_header(const struct net_device *dev, 3372 char *ll_header, int len) 3373 { 3374 if (likely(len >= dev->hard_header_len)) 3375 return true; 3376 if (len < dev->min_header_len) 3377 return false; 3378 3379 if (capable(CAP_SYS_RAWIO)) { 3380 memset(ll_header + len, 0, dev->hard_header_len - len); 3381 return true; 3382 } 3383 3384 if (dev->header_ops && dev->header_ops->validate) 3385 return dev->header_ops->validate(ll_header, len); 3386 3387 return false; 3388 } 3389 3390 static inline bool dev_has_header(const struct net_device *dev) 3391 { 3392 return dev->header_ops && dev->header_ops->create; 3393 } 3394 3395 /* 3396 * Incoming packets are placed on per-CPU queues 3397 */ 3398 struct softnet_data { 3399 struct list_head poll_list; 3400 struct sk_buff_head process_queue; 3401 local_lock_t process_queue_bh_lock; 3402 3403 /* stats */ 3404 unsigned int processed; 3405 unsigned int time_squeeze; 3406 #ifdef CONFIG_RPS 3407 struct softnet_data *rps_ipi_list; 3408 #endif 3409 3410 unsigned int received_rps; 3411 bool in_net_rx_action; 3412 bool in_napi_threaded_poll; 3413 3414 #ifdef CONFIG_NET_FLOW_LIMIT 3415 struct sd_flow_limit __rcu *flow_limit; 3416 #endif 3417 struct Qdisc *output_queue; 3418 struct Qdisc **output_queue_tailp; 3419 struct sk_buff *completion_queue; 3420 #ifdef CONFIG_XFRM_OFFLOAD 3421 struct sk_buff_head xfrm_backlog; 3422 #endif 3423 /* written and read only by owning cpu: */ 3424 struct netdev_xmit xmit; 3425 #ifdef CONFIG_RPS 3426 /* input_queue_head should be written by cpu owning this struct, 3427 * and only read by other cpus. Worth using a cache line. 3428 */ 3429 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3430 3431 /* Elements below can be accessed between CPUs for RPS/RFS */ 3432 call_single_data_t csd ____cacheline_aligned_in_smp; 3433 struct softnet_data *rps_ipi_next; 3434 unsigned int cpu; 3435 unsigned int input_queue_tail; 3436 #endif 3437 struct sk_buff_head input_pkt_queue; 3438 struct napi_struct backlog; 3439 3440 atomic_t dropped ____cacheline_aligned_in_smp; 3441 3442 /* Another possibly contended cache line */ 3443 spinlock_t defer_lock ____cacheline_aligned_in_smp; 3444 int defer_count; 3445 int defer_ipi_scheduled; 3446 struct sk_buff *defer_list; 3447 call_single_data_t defer_csd; 3448 }; 3449 3450 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3451 DECLARE_PER_CPU(struct page_pool *, system_page_pool); 3452 3453 #ifndef CONFIG_PREEMPT_RT 3454 static inline int dev_recursion_level(void) 3455 { 3456 return this_cpu_read(softnet_data.xmit.recursion); 3457 } 3458 #else 3459 static inline int dev_recursion_level(void) 3460 { 3461 return current->net_xmit.recursion; 3462 } 3463 3464 #endif 3465 3466 void __netif_schedule(struct Qdisc *q); 3467 void netif_schedule_queue(struct netdev_queue *txq); 3468 3469 static inline void netif_tx_schedule_all(struct net_device *dev) 3470 { 3471 unsigned int i; 3472 3473 for (i = 0; i < dev->num_tx_queues; i++) 3474 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3475 } 3476 3477 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3478 { 3479 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3480 } 3481 3482 /** 3483 * netif_start_queue - allow transmit 3484 * @dev: network device 3485 * 3486 * Allow upper layers to call the device hard_start_xmit routine. 3487 */ 3488 static inline void netif_start_queue(struct net_device *dev) 3489 { 3490 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3491 } 3492 3493 static inline void netif_tx_start_all_queues(struct net_device *dev) 3494 { 3495 unsigned int i; 3496 3497 for (i = 0; i < dev->num_tx_queues; i++) { 3498 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3499 netif_tx_start_queue(txq); 3500 } 3501 } 3502 3503 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3504 3505 /** 3506 * netif_wake_queue - restart transmit 3507 * @dev: network device 3508 * 3509 * Allow upper layers to call the device hard_start_xmit routine. 3510 * Used for flow control when transmit resources are available. 3511 */ 3512 static inline void netif_wake_queue(struct net_device *dev) 3513 { 3514 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3515 } 3516 3517 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3518 { 3519 unsigned int i; 3520 3521 for (i = 0; i < dev->num_tx_queues; i++) { 3522 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3523 netif_tx_wake_queue(txq); 3524 } 3525 } 3526 3527 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3528 { 3529 /* Paired with READ_ONCE() from dev_watchdog() */ 3530 WRITE_ONCE(dev_queue->trans_start, jiffies); 3531 3532 /* This barrier is paired with smp_mb() from dev_watchdog() */ 3533 smp_mb__before_atomic(); 3534 3535 /* Must be an atomic op see netif_txq_try_stop() */ 3536 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3537 } 3538 3539 /** 3540 * netif_stop_queue - stop transmitted packets 3541 * @dev: network device 3542 * 3543 * Stop upper layers calling the device hard_start_xmit routine. 3544 * Used for flow control when transmit resources are unavailable. 3545 */ 3546 static inline void netif_stop_queue(struct net_device *dev) 3547 { 3548 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3549 } 3550 3551 void netif_tx_stop_all_queues(struct net_device *dev); 3552 3553 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3554 { 3555 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3556 } 3557 3558 /** 3559 * netif_queue_stopped - test if transmit queue is flowblocked 3560 * @dev: network device 3561 * 3562 * Test if transmit queue on device is currently unable to send. 3563 */ 3564 static inline bool netif_queue_stopped(const struct net_device *dev) 3565 { 3566 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3567 } 3568 3569 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3570 { 3571 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3572 } 3573 3574 static inline bool 3575 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3576 { 3577 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3578 } 3579 3580 static inline bool 3581 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3582 { 3583 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3584 } 3585 3586 /** 3587 * netdev_queue_set_dql_min_limit - set dql minimum limit 3588 * @dev_queue: pointer to transmit queue 3589 * @min_limit: dql minimum limit 3590 * 3591 * Forces xmit_more() to return true until the minimum threshold 3592 * defined by @min_limit is reached (or until the tx queue is 3593 * empty). Warning: to be use with care, misuse will impact the 3594 * latency. 3595 */ 3596 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3597 unsigned int min_limit) 3598 { 3599 #ifdef CONFIG_BQL 3600 dev_queue->dql.min_limit = min_limit; 3601 #endif 3602 } 3603 3604 static inline int netdev_queue_dql_avail(const struct netdev_queue *txq) 3605 { 3606 #ifdef CONFIG_BQL 3607 /* Non-BQL migrated drivers will return 0, too. */ 3608 return dql_avail(&txq->dql); 3609 #else 3610 return 0; 3611 #endif 3612 } 3613 3614 /** 3615 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3616 * @dev_queue: pointer to transmit queue 3617 * 3618 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3619 * to give appropriate hint to the CPU. 3620 */ 3621 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3622 { 3623 #ifdef CONFIG_BQL 3624 prefetchw(&dev_queue->dql.num_queued); 3625 #endif 3626 } 3627 3628 /** 3629 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3630 * @dev_queue: pointer to transmit queue 3631 * 3632 * BQL enabled drivers might use this helper in their TX completion path, 3633 * to give appropriate hint to the CPU. 3634 */ 3635 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3636 { 3637 #ifdef CONFIG_BQL 3638 prefetchw(&dev_queue->dql.limit); 3639 #endif 3640 } 3641 3642 /** 3643 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3644 * @dev_queue: network device queue 3645 * @bytes: number of bytes queued to the device queue 3646 * 3647 * Report the number of bytes queued for sending/completion to the network 3648 * device hardware queue. @bytes should be a good approximation and should 3649 * exactly match netdev_completed_queue() @bytes. 3650 * This is typically called once per packet, from ndo_start_xmit(). 3651 */ 3652 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3653 unsigned int bytes) 3654 { 3655 #ifdef CONFIG_BQL 3656 dql_queued(&dev_queue->dql, bytes); 3657 3658 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3659 return; 3660 3661 /* Paired with READ_ONCE() from dev_watchdog() */ 3662 WRITE_ONCE(dev_queue->trans_start, jiffies); 3663 3664 /* This barrier is paired with smp_mb() from dev_watchdog() */ 3665 smp_mb__before_atomic(); 3666 3667 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3668 3669 /* 3670 * The XOFF flag must be set before checking the dql_avail below, 3671 * because in netdev_tx_completed_queue we update the dql_completed 3672 * before checking the XOFF flag. 3673 */ 3674 smp_mb__after_atomic(); 3675 3676 /* check again in case another CPU has just made room avail */ 3677 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3678 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3679 #endif 3680 } 3681 3682 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3683 * that they should not test BQL status themselves. 3684 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3685 * skb of a batch. 3686 * Returns true if the doorbell must be used to kick the NIC. 3687 */ 3688 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3689 unsigned int bytes, 3690 bool xmit_more) 3691 { 3692 if (xmit_more) { 3693 #ifdef CONFIG_BQL 3694 dql_queued(&dev_queue->dql, bytes); 3695 #endif 3696 return netif_tx_queue_stopped(dev_queue); 3697 } 3698 netdev_tx_sent_queue(dev_queue, bytes); 3699 return true; 3700 } 3701 3702 /** 3703 * netdev_sent_queue - report the number of bytes queued to hardware 3704 * @dev: network device 3705 * @bytes: number of bytes queued to the hardware device queue 3706 * 3707 * Report the number of bytes queued for sending/completion to the network 3708 * device hardware queue#0. @bytes should be a good approximation and should 3709 * exactly match netdev_completed_queue() @bytes. 3710 * This is typically called once per packet, from ndo_start_xmit(). 3711 */ 3712 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3713 { 3714 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3715 } 3716 3717 static inline bool __netdev_sent_queue(struct net_device *dev, 3718 unsigned int bytes, 3719 bool xmit_more) 3720 { 3721 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3722 xmit_more); 3723 } 3724 3725 /** 3726 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3727 * @dev_queue: network device queue 3728 * @pkts: number of packets (currently ignored) 3729 * @bytes: number of bytes dequeued from the device queue 3730 * 3731 * Must be called at most once per TX completion round (and not per 3732 * individual packet), so that BQL can adjust its limits appropriately. 3733 */ 3734 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3735 unsigned int pkts, unsigned int bytes) 3736 { 3737 #ifdef CONFIG_BQL 3738 if (unlikely(!bytes)) 3739 return; 3740 3741 dql_completed(&dev_queue->dql, bytes); 3742 3743 /* 3744 * Without the memory barrier there is a small possibility that 3745 * netdev_tx_sent_queue will miss the update and cause the queue to 3746 * be stopped forever 3747 */ 3748 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */ 3749 3750 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3751 return; 3752 3753 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3754 netif_schedule_queue(dev_queue); 3755 #endif 3756 } 3757 3758 /** 3759 * netdev_completed_queue - report bytes and packets completed by device 3760 * @dev: network device 3761 * @pkts: actual number of packets sent over the medium 3762 * @bytes: actual number of bytes sent over the medium 3763 * 3764 * Report the number of bytes and packets transmitted by the network device 3765 * hardware queue over the physical medium, @bytes must exactly match the 3766 * @bytes amount passed to netdev_sent_queue() 3767 */ 3768 static inline void netdev_completed_queue(struct net_device *dev, 3769 unsigned int pkts, unsigned int bytes) 3770 { 3771 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3772 } 3773 3774 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3775 { 3776 #ifdef CONFIG_BQL 3777 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3778 dql_reset(&q->dql); 3779 #endif 3780 } 3781 3782 /** 3783 * netdev_tx_reset_subqueue - reset the BQL stats and state of a netdev queue 3784 * @dev: network device 3785 * @qid: stack index of the queue to reset 3786 */ 3787 static inline void netdev_tx_reset_subqueue(const struct net_device *dev, 3788 u32 qid) 3789 { 3790 netdev_tx_reset_queue(netdev_get_tx_queue(dev, qid)); 3791 } 3792 3793 /** 3794 * netdev_reset_queue - reset the packets and bytes count of a network device 3795 * @dev_queue: network device 3796 * 3797 * Reset the bytes and packet count of a network device and clear the 3798 * software flow control OFF bit for this network device 3799 */ 3800 static inline void netdev_reset_queue(struct net_device *dev_queue) 3801 { 3802 netdev_tx_reset_subqueue(dev_queue, 0); 3803 } 3804 3805 /** 3806 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3807 * @dev: network device 3808 * @queue_index: given tx queue index 3809 * 3810 * Returns 0 if given tx queue index >= number of device tx queues, 3811 * otherwise returns the originally passed tx queue index. 3812 */ 3813 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3814 { 3815 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3816 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3817 dev->name, queue_index, 3818 dev->real_num_tx_queues); 3819 return 0; 3820 } 3821 3822 return queue_index; 3823 } 3824 3825 /** 3826 * netif_running - test if up 3827 * @dev: network device 3828 * 3829 * Test if the device has been brought up. 3830 */ 3831 static inline bool netif_running(const struct net_device *dev) 3832 { 3833 return test_bit(__LINK_STATE_START, &dev->state); 3834 } 3835 3836 /* 3837 * Routines to manage the subqueues on a device. We only need start, 3838 * stop, and a check if it's stopped. All other device management is 3839 * done at the overall netdevice level. 3840 * Also test the device if we're multiqueue. 3841 */ 3842 3843 /** 3844 * netif_start_subqueue - allow sending packets on subqueue 3845 * @dev: network device 3846 * @queue_index: sub queue index 3847 * 3848 * Start individual transmit queue of a device with multiple transmit queues. 3849 */ 3850 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3851 { 3852 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3853 3854 netif_tx_start_queue(txq); 3855 } 3856 3857 /** 3858 * netif_stop_subqueue - stop sending packets on subqueue 3859 * @dev: network device 3860 * @queue_index: sub queue index 3861 * 3862 * Stop individual transmit queue of a device with multiple transmit queues. 3863 */ 3864 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3865 { 3866 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3867 netif_tx_stop_queue(txq); 3868 } 3869 3870 /** 3871 * __netif_subqueue_stopped - test status of subqueue 3872 * @dev: network device 3873 * @queue_index: sub queue index 3874 * 3875 * Check individual transmit queue of a device with multiple transmit queues. 3876 */ 3877 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3878 u16 queue_index) 3879 { 3880 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3881 3882 return netif_tx_queue_stopped(txq); 3883 } 3884 3885 /** 3886 * netif_subqueue_stopped - test status of subqueue 3887 * @dev: network device 3888 * @skb: sub queue buffer pointer 3889 * 3890 * Check individual transmit queue of a device with multiple transmit queues. 3891 */ 3892 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3893 struct sk_buff *skb) 3894 { 3895 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3896 } 3897 3898 /** 3899 * netif_wake_subqueue - allow sending packets on subqueue 3900 * @dev: network device 3901 * @queue_index: sub queue index 3902 * 3903 * Resume individual transmit queue of a device with multiple transmit queues. 3904 */ 3905 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3906 { 3907 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3908 3909 netif_tx_wake_queue(txq); 3910 } 3911 3912 #ifdef CONFIG_XPS 3913 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3914 u16 index); 3915 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3916 u16 index, enum xps_map_type type); 3917 3918 /** 3919 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3920 * @j: CPU/Rx queue index 3921 * @mask: bitmask of all cpus/rx queues 3922 * @nr_bits: number of bits in the bitmask 3923 * 3924 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3925 */ 3926 static inline bool netif_attr_test_mask(unsigned long j, 3927 const unsigned long *mask, 3928 unsigned int nr_bits) 3929 { 3930 cpu_max_bits_warn(j, nr_bits); 3931 return test_bit(j, mask); 3932 } 3933 3934 /** 3935 * netif_attr_test_online - Test for online CPU/Rx queue 3936 * @j: CPU/Rx queue index 3937 * @online_mask: bitmask for CPUs/Rx queues that are online 3938 * @nr_bits: number of bits in the bitmask 3939 * 3940 * Returns: true if a CPU/Rx queue is online. 3941 */ 3942 static inline bool netif_attr_test_online(unsigned long j, 3943 const unsigned long *online_mask, 3944 unsigned int nr_bits) 3945 { 3946 cpu_max_bits_warn(j, nr_bits); 3947 3948 if (online_mask) 3949 return test_bit(j, online_mask); 3950 3951 return (j < nr_bits); 3952 } 3953 3954 /** 3955 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3956 * @n: CPU/Rx queue index 3957 * @srcp: the cpumask/Rx queue mask pointer 3958 * @nr_bits: number of bits in the bitmask 3959 * 3960 * Returns: next (after n) CPU/Rx queue index in the mask; 3961 * >= nr_bits if no further CPUs/Rx queues set. 3962 */ 3963 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3964 unsigned int nr_bits) 3965 { 3966 /* -1 is a legal arg here. */ 3967 if (n != -1) 3968 cpu_max_bits_warn(n, nr_bits); 3969 3970 if (srcp) 3971 return find_next_bit(srcp, nr_bits, n + 1); 3972 3973 return n + 1; 3974 } 3975 3976 /** 3977 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3978 * @n: CPU/Rx queue index 3979 * @src1p: the first CPUs/Rx queues mask pointer 3980 * @src2p: the second CPUs/Rx queues mask pointer 3981 * @nr_bits: number of bits in the bitmask 3982 * 3983 * Returns: next (after n) CPU/Rx queue index set in both masks; 3984 * >= nr_bits if no further CPUs/Rx queues set in both. 3985 */ 3986 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3987 const unsigned long *src2p, 3988 unsigned int nr_bits) 3989 { 3990 /* -1 is a legal arg here. */ 3991 if (n != -1) 3992 cpu_max_bits_warn(n, nr_bits); 3993 3994 if (src1p && src2p) 3995 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3996 else if (src1p) 3997 return find_next_bit(src1p, nr_bits, n + 1); 3998 else if (src2p) 3999 return find_next_bit(src2p, nr_bits, n + 1); 4000 4001 return n + 1; 4002 } 4003 #else 4004 static inline int netif_set_xps_queue(struct net_device *dev, 4005 const struct cpumask *mask, 4006 u16 index) 4007 { 4008 return 0; 4009 } 4010 4011 static inline int __netif_set_xps_queue(struct net_device *dev, 4012 const unsigned long *mask, 4013 u16 index, enum xps_map_type type) 4014 { 4015 return 0; 4016 } 4017 #endif 4018 4019 /** 4020 * netif_is_multiqueue - test if device has multiple transmit queues 4021 * @dev: network device 4022 * 4023 * Check if device has multiple transmit queues 4024 */ 4025 static inline bool netif_is_multiqueue(const struct net_device *dev) 4026 { 4027 return dev->num_tx_queues > 1; 4028 } 4029 4030 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 4031 4032 #ifdef CONFIG_SYSFS 4033 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 4034 #else 4035 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 4036 unsigned int rxqs) 4037 { 4038 dev->real_num_rx_queues = rxqs; 4039 return 0; 4040 } 4041 #endif 4042 int netif_set_real_num_queues(struct net_device *dev, 4043 unsigned int txq, unsigned int rxq); 4044 4045 int netif_get_num_default_rss_queues(void); 4046 4047 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason); 4048 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason); 4049 4050 /* 4051 * It is not allowed to call kfree_skb() or consume_skb() from hardware 4052 * interrupt context or with hardware interrupts being disabled. 4053 * (in_hardirq() || irqs_disabled()) 4054 * 4055 * We provide four helpers that can be used in following contexts : 4056 * 4057 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 4058 * replacing kfree_skb(skb) 4059 * 4060 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 4061 * Typically used in place of consume_skb(skb) in TX completion path 4062 * 4063 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 4064 * replacing kfree_skb(skb) 4065 * 4066 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 4067 * and consumed a packet. Used in place of consume_skb(skb) 4068 */ 4069 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 4070 { 4071 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 4072 } 4073 4074 static inline void dev_consume_skb_irq(struct sk_buff *skb) 4075 { 4076 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED); 4077 } 4078 4079 static inline void dev_kfree_skb_any(struct sk_buff *skb) 4080 { 4081 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 4082 } 4083 4084 static inline void dev_consume_skb_any(struct sk_buff *skb) 4085 { 4086 dev_kfree_skb_any_reason(skb, SKB_CONSUMED); 4087 } 4088 4089 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 4090 const struct bpf_prog *xdp_prog); 4091 void generic_xdp_tx(struct sk_buff *skb, const struct bpf_prog *xdp_prog); 4092 int do_xdp_generic(const struct bpf_prog *xdp_prog, struct sk_buff **pskb); 4093 int netif_rx(struct sk_buff *skb); 4094 int __netif_rx(struct sk_buff *skb); 4095 4096 int netif_receive_skb(struct sk_buff *skb); 4097 int netif_receive_skb_core(struct sk_buff *skb); 4098 void netif_receive_skb_list_internal(struct list_head *head); 4099 void netif_receive_skb_list(struct list_head *head); 4100 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 4101 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 4102 struct sk_buff *napi_get_frags(struct napi_struct *napi); 4103 void napi_get_frags_check(struct napi_struct *napi); 4104 gro_result_t napi_gro_frags(struct napi_struct *napi); 4105 4106 static inline void napi_free_frags(struct napi_struct *napi) 4107 { 4108 kfree_skb(napi->skb); 4109 napi->skb = NULL; 4110 } 4111 4112 bool netdev_is_rx_handler_busy(struct net_device *dev); 4113 int netdev_rx_handler_register(struct net_device *dev, 4114 rx_handler_func_t *rx_handler, 4115 void *rx_handler_data); 4116 void netdev_rx_handler_unregister(struct net_device *dev); 4117 4118 bool dev_valid_name(const char *name); 4119 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 4120 { 4121 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 4122 } 4123 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 4124 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 4125 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 4126 void __user *data, bool *need_copyout); 4127 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 4128 int generic_hwtstamp_get_lower(struct net_device *dev, 4129 struct kernel_hwtstamp_config *kernel_cfg); 4130 int generic_hwtstamp_set_lower(struct net_device *dev, 4131 struct kernel_hwtstamp_config *kernel_cfg, 4132 struct netlink_ext_ack *extack); 4133 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 4134 unsigned int dev_get_flags(const struct net_device *); 4135 int __dev_change_flags(struct net_device *dev, unsigned int flags, 4136 struct netlink_ext_ack *extack); 4137 int dev_change_flags(struct net_device *dev, unsigned int flags, 4138 struct netlink_ext_ack *extack); 4139 int dev_set_alias(struct net_device *, const char *, size_t); 4140 int dev_get_alias(const struct net_device *, char *, size_t); 4141 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 4142 const char *pat, int new_ifindex); 4143 static inline 4144 int dev_change_net_namespace(struct net_device *dev, struct net *net, 4145 const char *pat) 4146 { 4147 return __dev_change_net_namespace(dev, net, pat, 0); 4148 } 4149 int __dev_set_mtu(struct net_device *, int); 4150 int dev_set_mtu(struct net_device *, int); 4151 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 4152 struct netlink_ext_ack *extack); 4153 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 4154 struct netlink_ext_ack *extack); 4155 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 4156 struct netlink_ext_ack *extack); 4157 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 4158 int dev_get_port_parent_id(struct net_device *dev, 4159 struct netdev_phys_item_id *ppid, bool recurse); 4160 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 4161 4162 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 4163 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 4164 struct netdev_queue *txq, int *ret); 4165 4166 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 4167 u8 dev_xdp_prog_count(struct net_device *dev); 4168 int dev_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf); 4169 u8 dev_xdp_sb_prog_count(struct net_device *dev); 4170 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 4171 4172 u32 dev_get_min_mp_channel_count(const struct net_device *dev); 4173 4174 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4175 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4176 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 4177 bool is_skb_forwardable(const struct net_device *dev, 4178 const struct sk_buff *skb); 4179 4180 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 4181 const struct sk_buff *skb, 4182 const bool check_mtu) 4183 { 4184 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 4185 unsigned int len; 4186 4187 if (!(dev->flags & IFF_UP)) 4188 return false; 4189 4190 if (!check_mtu) 4191 return true; 4192 4193 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 4194 if (skb->len <= len) 4195 return true; 4196 4197 /* if TSO is enabled, we don't care about the length as the packet 4198 * could be forwarded without being segmented before 4199 */ 4200 if (skb_is_gso(skb)) 4201 return true; 4202 4203 return false; 4204 } 4205 4206 void netdev_core_stats_inc(struct net_device *dev, u32 offset); 4207 4208 #define DEV_CORE_STATS_INC(FIELD) \ 4209 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 4210 { \ 4211 netdev_core_stats_inc(dev, \ 4212 offsetof(struct net_device_core_stats, FIELD)); \ 4213 } 4214 DEV_CORE_STATS_INC(rx_dropped) 4215 DEV_CORE_STATS_INC(tx_dropped) 4216 DEV_CORE_STATS_INC(rx_nohandler) 4217 DEV_CORE_STATS_INC(rx_otherhost_dropped) 4218 #undef DEV_CORE_STATS_INC 4219 4220 static __always_inline int ____dev_forward_skb(struct net_device *dev, 4221 struct sk_buff *skb, 4222 const bool check_mtu) 4223 { 4224 if (skb_orphan_frags(skb, GFP_ATOMIC) || 4225 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 4226 dev_core_stats_rx_dropped_inc(dev); 4227 kfree_skb(skb); 4228 return NET_RX_DROP; 4229 } 4230 4231 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 4232 skb->priority = 0; 4233 return 0; 4234 } 4235 4236 bool dev_nit_active(struct net_device *dev); 4237 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 4238 4239 static inline void __dev_put(struct net_device *dev) 4240 { 4241 if (dev) { 4242 #ifdef CONFIG_PCPU_DEV_REFCNT 4243 this_cpu_dec(*dev->pcpu_refcnt); 4244 #else 4245 refcount_dec(&dev->dev_refcnt); 4246 #endif 4247 } 4248 } 4249 4250 static inline void __dev_hold(struct net_device *dev) 4251 { 4252 if (dev) { 4253 #ifdef CONFIG_PCPU_DEV_REFCNT 4254 this_cpu_inc(*dev->pcpu_refcnt); 4255 #else 4256 refcount_inc(&dev->dev_refcnt); 4257 #endif 4258 } 4259 } 4260 4261 static inline void __netdev_tracker_alloc(struct net_device *dev, 4262 netdevice_tracker *tracker, 4263 gfp_t gfp) 4264 { 4265 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4266 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 4267 #endif 4268 } 4269 4270 /* netdev_tracker_alloc() can upgrade a prior untracked reference 4271 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4272 */ 4273 static inline void netdev_tracker_alloc(struct net_device *dev, 4274 netdevice_tracker *tracker, gfp_t gfp) 4275 { 4276 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4277 refcount_dec(&dev->refcnt_tracker.no_tracker); 4278 __netdev_tracker_alloc(dev, tracker, gfp); 4279 #endif 4280 } 4281 4282 static inline void netdev_tracker_free(struct net_device *dev, 4283 netdevice_tracker *tracker) 4284 { 4285 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4286 ref_tracker_free(&dev->refcnt_tracker, tracker); 4287 #endif 4288 } 4289 4290 static inline void netdev_hold(struct net_device *dev, 4291 netdevice_tracker *tracker, gfp_t gfp) 4292 { 4293 if (dev) { 4294 __dev_hold(dev); 4295 __netdev_tracker_alloc(dev, tracker, gfp); 4296 } 4297 } 4298 4299 static inline void netdev_put(struct net_device *dev, 4300 netdevice_tracker *tracker) 4301 { 4302 if (dev) { 4303 netdev_tracker_free(dev, tracker); 4304 __dev_put(dev); 4305 } 4306 } 4307 4308 /** 4309 * dev_hold - get reference to device 4310 * @dev: network device 4311 * 4312 * Hold reference to device to keep it from being freed. 4313 * Try using netdev_hold() instead. 4314 */ 4315 static inline void dev_hold(struct net_device *dev) 4316 { 4317 netdev_hold(dev, NULL, GFP_ATOMIC); 4318 } 4319 4320 /** 4321 * dev_put - release reference to device 4322 * @dev: network device 4323 * 4324 * Release reference to device to allow it to be freed. 4325 * Try using netdev_put() instead. 4326 */ 4327 static inline void dev_put(struct net_device *dev) 4328 { 4329 netdev_put(dev, NULL); 4330 } 4331 4332 DEFINE_FREE(dev_put, struct net_device *, if (_T) dev_put(_T)) 4333 4334 static inline void netdev_ref_replace(struct net_device *odev, 4335 struct net_device *ndev, 4336 netdevice_tracker *tracker, 4337 gfp_t gfp) 4338 { 4339 if (odev) 4340 netdev_tracker_free(odev, tracker); 4341 4342 __dev_hold(ndev); 4343 __dev_put(odev); 4344 4345 if (ndev) 4346 __netdev_tracker_alloc(ndev, tracker, gfp); 4347 } 4348 4349 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4350 * and _off may be called from IRQ context, but it is caller 4351 * who is responsible for serialization of these calls. 4352 * 4353 * The name carrier is inappropriate, these functions should really be 4354 * called netif_lowerlayer_*() because they represent the state of any 4355 * kind of lower layer not just hardware media. 4356 */ 4357 void linkwatch_fire_event(struct net_device *dev); 4358 4359 /** 4360 * linkwatch_sync_dev - sync linkwatch for the given device 4361 * @dev: network device to sync linkwatch for 4362 * 4363 * Sync linkwatch for the given device, removing it from the 4364 * pending work list (if queued). 4365 */ 4366 void linkwatch_sync_dev(struct net_device *dev); 4367 4368 /** 4369 * netif_carrier_ok - test if carrier present 4370 * @dev: network device 4371 * 4372 * Check if carrier is present on device 4373 */ 4374 static inline bool netif_carrier_ok(const struct net_device *dev) 4375 { 4376 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4377 } 4378 4379 unsigned long dev_trans_start(struct net_device *dev); 4380 4381 void netdev_watchdog_up(struct net_device *dev); 4382 4383 void netif_carrier_on(struct net_device *dev); 4384 void netif_carrier_off(struct net_device *dev); 4385 void netif_carrier_event(struct net_device *dev); 4386 4387 /** 4388 * netif_dormant_on - mark device as dormant. 4389 * @dev: network device 4390 * 4391 * Mark device as dormant (as per RFC2863). 4392 * 4393 * The dormant state indicates that the relevant interface is not 4394 * actually in a condition to pass packets (i.e., it is not 'up') but is 4395 * in a "pending" state, waiting for some external event. For "on- 4396 * demand" interfaces, this new state identifies the situation where the 4397 * interface is waiting for events to place it in the up state. 4398 */ 4399 static inline void netif_dormant_on(struct net_device *dev) 4400 { 4401 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4402 linkwatch_fire_event(dev); 4403 } 4404 4405 /** 4406 * netif_dormant_off - set device as not dormant. 4407 * @dev: network device 4408 * 4409 * Device is not in dormant state. 4410 */ 4411 static inline void netif_dormant_off(struct net_device *dev) 4412 { 4413 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4414 linkwatch_fire_event(dev); 4415 } 4416 4417 /** 4418 * netif_dormant - test if device is dormant 4419 * @dev: network device 4420 * 4421 * Check if device is dormant. 4422 */ 4423 static inline bool netif_dormant(const struct net_device *dev) 4424 { 4425 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4426 } 4427 4428 4429 /** 4430 * netif_testing_on - mark device as under test. 4431 * @dev: network device 4432 * 4433 * Mark device as under test (as per RFC2863). 4434 * 4435 * The testing state indicates that some test(s) must be performed on 4436 * the interface. After completion, of the test, the interface state 4437 * will change to up, dormant, or down, as appropriate. 4438 */ 4439 static inline void netif_testing_on(struct net_device *dev) 4440 { 4441 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4442 linkwatch_fire_event(dev); 4443 } 4444 4445 /** 4446 * netif_testing_off - set device as not under test. 4447 * @dev: network device 4448 * 4449 * Device is not in testing state. 4450 */ 4451 static inline void netif_testing_off(struct net_device *dev) 4452 { 4453 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4454 linkwatch_fire_event(dev); 4455 } 4456 4457 /** 4458 * netif_testing - test if device is under test 4459 * @dev: network device 4460 * 4461 * Check if device is under test 4462 */ 4463 static inline bool netif_testing(const struct net_device *dev) 4464 { 4465 return test_bit(__LINK_STATE_TESTING, &dev->state); 4466 } 4467 4468 4469 /** 4470 * netif_oper_up - test if device is operational 4471 * @dev: network device 4472 * 4473 * Check if carrier is operational 4474 */ 4475 static inline bool netif_oper_up(const struct net_device *dev) 4476 { 4477 unsigned int operstate = READ_ONCE(dev->operstate); 4478 4479 return operstate == IF_OPER_UP || 4480 operstate == IF_OPER_UNKNOWN /* backward compat */; 4481 } 4482 4483 /** 4484 * netif_device_present - is device available or removed 4485 * @dev: network device 4486 * 4487 * Check if device has not been removed from system. 4488 */ 4489 static inline bool netif_device_present(const struct net_device *dev) 4490 { 4491 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4492 } 4493 4494 void netif_device_detach(struct net_device *dev); 4495 4496 void netif_device_attach(struct net_device *dev); 4497 4498 /* 4499 * Network interface message level settings 4500 */ 4501 4502 enum { 4503 NETIF_MSG_DRV_BIT, 4504 NETIF_MSG_PROBE_BIT, 4505 NETIF_MSG_LINK_BIT, 4506 NETIF_MSG_TIMER_BIT, 4507 NETIF_MSG_IFDOWN_BIT, 4508 NETIF_MSG_IFUP_BIT, 4509 NETIF_MSG_RX_ERR_BIT, 4510 NETIF_MSG_TX_ERR_BIT, 4511 NETIF_MSG_TX_QUEUED_BIT, 4512 NETIF_MSG_INTR_BIT, 4513 NETIF_MSG_TX_DONE_BIT, 4514 NETIF_MSG_RX_STATUS_BIT, 4515 NETIF_MSG_PKTDATA_BIT, 4516 NETIF_MSG_HW_BIT, 4517 NETIF_MSG_WOL_BIT, 4518 4519 /* When you add a new bit above, update netif_msg_class_names array 4520 * in net/ethtool/common.c 4521 */ 4522 NETIF_MSG_CLASS_COUNT, 4523 }; 4524 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4525 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4526 4527 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4528 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4529 4530 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4531 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4532 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4533 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4534 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4535 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4536 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4537 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4538 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4539 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4540 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4541 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4542 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4543 #define NETIF_MSG_HW __NETIF_MSG(HW) 4544 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4545 4546 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4547 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4548 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4549 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4550 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4551 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4552 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4553 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4554 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4555 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4556 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4557 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4558 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4559 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4560 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4561 4562 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4563 { 4564 /* use default */ 4565 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4566 return default_msg_enable_bits; 4567 if (debug_value == 0) /* no output */ 4568 return 0; 4569 /* set low N bits */ 4570 return (1U << debug_value) - 1; 4571 } 4572 4573 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4574 { 4575 spin_lock(&txq->_xmit_lock); 4576 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4577 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4578 } 4579 4580 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4581 { 4582 __acquire(&txq->_xmit_lock); 4583 return true; 4584 } 4585 4586 static inline void __netif_tx_release(struct netdev_queue *txq) 4587 { 4588 __release(&txq->_xmit_lock); 4589 } 4590 4591 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4592 { 4593 spin_lock_bh(&txq->_xmit_lock); 4594 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4595 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4596 } 4597 4598 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4599 { 4600 bool ok = spin_trylock(&txq->_xmit_lock); 4601 4602 if (likely(ok)) { 4603 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4604 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4605 } 4606 return ok; 4607 } 4608 4609 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4610 { 4611 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4612 WRITE_ONCE(txq->xmit_lock_owner, -1); 4613 spin_unlock(&txq->_xmit_lock); 4614 } 4615 4616 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4617 { 4618 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4619 WRITE_ONCE(txq->xmit_lock_owner, -1); 4620 spin_unlock_bh(&txq->_xmit_lock); 4621 } 4622 4623 /* 4624 * txq->trans_start can be read locklessly from dev_watchdog() 4625 */ 4626 static inline void txq_trans_update(struct netdev_queue *txq) 4627 { 4628 if (txq->xmit_lock_owner != -1) 4629 WRITE_ONCE(txq->trans_start, jiffies); 4630 } 4631 4632 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4633 { 4634 unsigned long now = jiffies; 4635 4636 if (READ_ONCE(txq->trans_start) != now) 4637 WRITE_ONCE(txq->trans_start, now); 4638 } 4639 4640 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4641 static inline void netif_trans_update(struct net_device *dev) 4642 { 4643 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4644 4645 txq_trans_cond_update(txq); 4646 } 4647 4648 /** 4649 * netif_tx_lock - grab network device transmit lock 4650 * @dev: network device 4651 * 4652 * Get network device transmit lock 4653 */ 4654 void netif_tx_lock(struct net_device *dev); 4655 4656 static inline void netif_tx_lock_bh(struct net_device *dev) 4657 { 4658 local_bh_disable(); 4659 netif_tx_lock(dev); 4660 } 4661 4662 void netif_tx_unlock(struct net_device *dev); 4663 4664 static inline void netif_tx_unlock_bh(struct net_device *dev) 4665 { 4666 netif_tx_unlock(dev); 4667 local_bh_enable(); 4668 } 4669 4670 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4671 if (!(dev)->lltx) { \ 4672 __netif_tx_lock(txq, cpu); \ 4673 } else { \ 4674 __netif_tx_acquire(txq); \ 4675 } \ 4676 } 4677 4678 #define HARD_TX_TRYLOCK(dev, txq) \ 4679 (!(dev)->lltx ? \ 4680 __netif_tx_trylock(txq) : \ 4681 __netif_tx_acquire(txq)) 4682 4683 #define HARD_TX_UNLOCK(dev, txq) { \ 4684 if (!(dev)->lltx) { \ 4685 __netif_tx_unlock(txq); \ 4686 } else { \ 4687 __netif_tx_release(txq); \ 4688 } \ 4689 } 4690 4691 static inline void netif_tx_disable(struct net_device *dev) 4692 { 4693 unsigned int i; 4694 int cpu; 4695 4696 local_bh_disable(); 4697 cpu = smp_processor_id(); 4698 spin_lock(&dev->tx_global_lock); 4699 for (i = 0; i < dev->num_tx_queues; i++) { 4700 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4701 4702 __netif_tx_lock(txq, cpu); 4703 netif_tx_stop_queue(txq); 4704 __netif_tx_unlock(txq); 4705 } 4706 spin_unlock(&dev->tx_global_lock); 4707 local_bh_enable(); 4708 } 4709 4710 static inline void netif_addr_lock(struct net_device *dev) 4711 { 4712 unsigned char nest_level = 0; 4713 4714 #ifdef CONFIG_LOCKDEP 4715 nest_level = dev->nested_level; 4716 #endif 4717 spin_lock_nested(&dev->addr_list_lock, nest_level); 4718 } 4719 4720 static inline void netif_addr_lock_bh(struct net_device *dev) 4721 { 4722 unsigned char nest_level = 0; 4723 4724 #ifdef CONFIG_LOCKDEP 4725 nest_level = dev->nested_level; 4726 #endif 4727 local_bh_disable(); 4728 spin_lock_nested(&dev->addr_list_lock, nest_level); 4729 } 4730 4731 static inline void netif_addr_unlock(struct net_device *dev) 4732 { 4733 spin_unlock(&dev->addr_list_lock); 4734 } 4735 4736 static inline void netif_addr_unlock_bh(struct net_device *dev) 4737 { 4738 spin_unlock_bh(&dev->addr_list_lock); 4739 } 4740 4741 /* 4742 * dev_addrs walker. Should be used only for read access. Call with 4743 * rcu_read_lock held. 4744 */ 4745 #define for_each_dev_addr(dev, ha) \ 4746 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4747 4748 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4749 4750 void ether_setup(struct net_device *dev); 4751 4752 /* Allocate dummy net_device */ 4753 struct net_device *alloc_netdev_dummy(int sizeof_priv); 4754 4755 /* Support for loadable net-drivers */ 4756 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4757 unsigned char name_assign_type, 4758 void (*setup)(struct net_device *), 4759 unsigned int txqs, unsigned int rxqs); 4760 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4761 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4762 4763 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4764 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4765 count) 4766 4767 int register_netdev(struct net_device *dev); 4768 void unregister_netdev(struct net_device *dev); 4769 4770 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4771 4772 /* General hardware address lists handling functions */ 4773 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4774 struct netdev_hw_addr_list *from_list, int addr_len); 4775 int __hw_addr_sync_multiple(struct netdev_hw_addr_list *to_list, 4776 struct netdev_hw_addr_list *from_list, 4777 int addr_len); 4778 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4779 struct netdev_hw_addr_list *from_list, int addr_len); 4780 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4781 struct net_device *dev, 4782 int (*sync)(struct net_device *, const unsigned char *), 4783 int (*unsync)(struct net_device *, 4784 const unsigned char *)); 4785 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4786 struct net_device *dev, 4787 int (*sync)(struct net_device *, 4788 const unsigned char *, int), 4789 int (*unsync)(struct net_device *, 4790 const unsigned char *, int)); 4791 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4792 struct net_device *dev, 4793 int (*unsync)(struct net_device *, 4794 const unsigned char *, int)); 4795 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4796 struct net_device *dev, 4797 int (*unsync)(struct net_device *, 4798 const unsigned char *)); 4799 void __hw_addr_init(struct netdev_hw_addr_list *list); 4800 4801 /* Functions used for device addresses handling */ 4802 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4803 const void *addr, size_t len); 4804 4805 static inline void 4806 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4807 { 4808 dev_addr_mod(dev, 0, addr, len); 4809 } 4810 4811 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4812 { 4813 __dev_addr_set(dev, addr, dev->addr_len); 4814 } 4815 4816 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4817 unsigned char addr_type); 4818 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4819 unsigned char addr_type); 4820 4821 /* Functions used for unicast addresses handling */ 4822 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4823 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4824 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4825 int dev_uc_sync(struct net_device *to, struct net_device *from); 4826 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4827 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4828 void dev_uc_flush(struct net_device *dev); 4829 void dev_uc_init(struct net_device *dev); 4830 4831 /** 4832 * __dev_uc_sync - Synchronize device's unicast list 4833 * @dev: device to sync 4834 * @sync: function to call if address should be added 4835 * @unsync: function to call if address should be removed 4836 * 4837 * Add newly added addresses to the interface, and release 4838 * addresses that have been deleted. 4839 */ 4840 static inline int __dev_uc_sync(struct net_device *dev, 4841 int (*sync)(struct net_device *, 4842 const unsigned char *), 4843 int (*unsync)(struct net_device *, 4844 const unsigned char *)) 4845 { 4846 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4847 } 4848 4849 /** 4850 * __dev_uc_unsync - Remove synchronized addresses from device 4851 * @dev: device to sync 4852 * @unsync: function to call if address should be removed 4853 * 4854 * Remove all addresses that were added to the device by dev_uc_sync(). 4855 */ 4856 static inline void __dev_uc_unsync(struct net_device *dev, 4857 int (*unsync)(struct net_device *, 4858 const unsigned char *)) 4859 { 4860 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4861 } 4862 4863 /* Functions used for multicast addresses handling */ 4864 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4865 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4866 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4867 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4868 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4869 int dev_mc_sync(struct net_device *to, struct net_device *from); 4870 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4871 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4872 void dev_mc_flush(struct net_device *dev); 4873 void dev_mc_init(struct net_device *dev); 4874 4875 /** 4876 * __dev_mc_sync - Synchronize device's multicast list 4877 * @dev: device to sync 4878 * @sync: function to call if address should be added 4879 * @unsync: function to call if address should be removed 4880 * 4881 * Add newly added addresses to the interface, and release 4882 * addresses that have been deleted. 4883 */ 4884 static inline int __dev_mc_sync(struct net_device *dev, 4885 int (*sync)(struct net_device *, 4886 const unsigned char *), 4887 int (*unsync)(struct net_device *, 4888 const unsigned char *)) 4889 { 4890 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4891 } 4892 4893 /** 4894 * __dev_mc_unsync - Remove synchronized addresses from device 4895 * @dev: device to sync 4896 * @unsync: function to call if address should be removed 4897 * 4898 * Remove all addresses that were added to the device by dev_mc_sync(). 4899 */ 4900 static inline void __dev_mc_unsync(struct net_device *dev, 4901 int (*unsync)(struct net_device *, 4902 const unsigned char *)) 4903 { 4904 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4905 } 4906 4907 /* Functions used for secondary unicast and multicast support */ 4908 void dev_set_rx_mode(struct net_device *dev); 4909 int dev_set_promiscuity(struct net_device *dev, int inc); 4910 int dev_set_allmulti(struct net_device *dev, int inc); 4911 void netdev_state_change(struct net_device *dev); 4912 void __netdev_notify_peers(struct net_device *dev); 4913 void netdev_notify_peers(struct net_device *dev); 4914 void netdev_features_change(struct net_device *dev); 4915 /* Load a device via the kmod */ 4916 void dev_load(struct net *net, const char *name); 4917 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4918 struct rtnl_link_stats64 *storage); 4919 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4920 const struct net_device_stats *netdev_stats); 4921 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4922 const struct pcpu_sw_netstats __percpu *netstats); 4923 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4924 4925 enum { 4926 NESTED_SYNC_IMM_BIT, 4927 NESTED_SYNC_TODO_BIT, 4928 }; 4929 4930 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4931 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4932 4933 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4934 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4935 4936 struct netdev_nested_priv { 4937 unsigned char flags; 4938 void *data; 4939 }; 4940 4941 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4942 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4943 struct list_head **iter); 4944 4945 /* iterate through upper list, must be called under RCU read lock */ 4946 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4947 for (iter = &(dev)->adj_list.upper, \ 4948 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4949 updev; \ 4950 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4951 4952 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4953 int (*fn)(struct net_device *upper_dev, 4954 struct netdev_nested_priv *priv), 4955 struct netdev_nested_priv *priv); 4956 4957 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4958 struct net_device *upper_dev); 4959 4960 bool netdev_has_any_upper_dev(struct net_device *dev); 4961 4962 void *netdev_lower_get_next_private(struct net_device *dev, 4963 struct list_head **iter); 4964 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4965 struct list_head **iter); 4966 4967 #define netdev_for_each_lower_private(dev, priv, iter) \ 4968 for (iter = (dev)->adj_list.lower.next, \ 4969 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4970 priv; \ 4971 priv = netdev_lower_get_next_private(dev, &(iter))) 4972 4973 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4974 for (iter = &(dev)->adj_list.lower, \ 4975 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4976 priv; \ 4977 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4978 4979 void *netdev_lower_get_next(struct net_device *dev, 4980 struct list_head **iter); 4981 4982 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4983 for (iter = (dev)->adj_list.lower.next, \ 4984 ldev = netdev_lower_get_next(dev, &(iter)); \ 4985 ldev; \ 4986 ldev = netdev_lower_get_next(dev, &(iter))) 4987 4988 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4989 struct list_head **iter); 4990 int netdev_walk_all_lower_dev(struct net_device *dev, 4991 int (*fn)(struct net_device *lower_dev, 4992 struct netdev_nested_priv *priv), 4993 struct netdev_nested_priv *priv); 4994 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4995 int (*fn)(struct net_device *lower_dev, 4996 struct netdev_nested_priv *priv), 4997 struct netdev_nested_priv *priv); 4998 4999 void *netdev_adjacent_get_private(struct list_head *adj_list); 5000 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 5001 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 5002 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 5003 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 5004 struct netlink_ext_ack *extack); 5005 int netdev_master_upper_dev_link(struct net_device *dev, 5006 struct net_device *upper_dev, 5007 void *upper_priv, void *upper_info, 5008 struct netlink_ext_ack *extack); 5009 void netdev_upper_dev_unlink(struct net_device *dev, 5010 struct net_device *upper_dev); 5011 int netdev_adjacent_change_prepare(struct net_device *old_dev, 5012 struct net_device *new_dev, 5013 struct net_device *dev, 5014 struct netlink_ext_ack *extack); 5015 void netdev_adjacent_change_commit(struct net_device *old_dev, 5016 struct net_device *new_dev, 5017 struct net_device *dev); 5018 void netdev_adjacent_change_abort(struct net_device *old_dev, 5019 struct net_device *new_dev, 5020 struct net_device *dev); 5021 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 5022 void *netdev_lower_dev_get_private(struct net_device *dev, 5023 struct net_device *lower_dev); 5024 void netdev_lower_state_changed(struct net_device *lower_dev, 5025 void *lower_state_info); 5026 5027 /* RSS keys are 40 or 52 bytes long */ 5028 #define NETDEV_RSS_KEY_LEN 52 5029 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 5030 void netdev_rss_key_fill(void *buffer, size_t len); 5031 5032 int skb_checksum_help(struct sk_buff *skb); 5033 int skb_crc32c_csum_help(struct sk_buff *skb); 5034 int skb_csum_hwoffload_help(struct sk_buff *skb, 5035 const netdev_features_t features); 5036 5037 struct netdev_bonding_info { 5038 ifslave slave; 5039 ifbond master; 5040 }; 5041 5042 struct netdev_notifier_bonding_info { 5043 struct netdev_notifier_info info; /* must be first */ 5044 struct netdev_bonding_info bonding_info; 5045 }; 5046 5047 void netdev_bonding_info_change(struct net_device *dev, 5048 struct netdev_bonding_info *bonding_info); 5049 5050 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 5051 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 5052 #else 5053 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 5054 const void *data) 5055 { 5056 } 5057 #endif 5058 5059 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 5060 5061 static inline bool can_checksum_protocol(netdev_features_t features, 5062 __be16 protocol) 5063 { 5064 if (protocol == htons(ETH_P_FCOE)) 5065 return !!(features & NETIF_F_FCOE_CRC); 5066 5067 /* Assume this is an IP checksum (not SCTP CRC) */ 5068 5069 if (features & NETIF_F_HW_CSUM) { 5070 /* Can checksum everything */ 5071 return true; 5072 } 5073 5074 switch (protocol) { 5075 case htons(ETH_P_IP): 5076 return !!(features & NETIF_F_IP_CSUM); 5077 case htons(ETH_P_IPV6): 5078 return !!(features & NETIF_F_IPV6_CSUM); 5079 default: 5080 return false; 5081 } 5082 } 5083 5084 #ifdef CONFIG_BUG 5085 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 5086 #else 5087 static inline void netdev_rx_csum_fault(struct net_device *dev, 5088 struct sk_buff *skb) 5089 { 5090 } 5091 #endif 5092 /* rx skb timestamps */ 5093 void net_enable_timestamp(void); 5094 void net_disable_timestamp(void); 5095 5096 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 5097 const struct skb_shared_hwtstamps *hwtstamps, 5098 bool cycles) 5099 { 5100 const struct net_device_ops *ops = dev->netdev_ops; 5101 5102 if (ops->ndo_get_tstamp) 5103 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 5104 5105 return hwtstamps->hwtstamp; 5106 } 5107 5108 #ifndef CONFIG_PREEMPT_RT 5109 static inline void netdev_xmit_set_more(bool more) 5110 { 5111 __this_cpu_write(softnet_data.xmit.more, more); 5112 } 5113 5114 static inline bool netdev_xmit_more(void) 5115 { 5116 return __this_cpu_read(softnet_data.xmit.more); 5117 } 5118 #else 5119 static inline void netdev_xmit_set_more(bool more) 5120 { 5121 current->net_xmit.more = more; 5122 } 5123 5124 static inline bool netdev_xmit_more(void) 5125 { 5126 return current->net_xmit.more; 5127 } 5128 #endif 5129 5130 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 5131 struct sk_buff *skb, struct net_device *dev, 5132 bool more) 5133 { 5134 netdev_xmit_set_more(more); 5135 return ops->ndo_start_xmit(skb, dev); 5136 } 5137 5138 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 5139 struct netdev_queue *txq, bool more) 5140 { 5141 const struct net_device_ops *ops = dev->netdev_ops; 5142 netdev_tx_t rc; 5143 5144 rc = __netdev_start_xmit(ops, skb, dev, more); 5145 if (rc == NETDEV_TX_OK) 5146 txq_trans_update(txq); 5147 5148 return rc; 5149 } 5150 5151 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 5152 const void *ns); 5153 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 5154 const void *ns); 5155 5156 extern const struct kobj_ns_type_operations net_ns_type_operations; 5157 5158 const char *netdev_drivername(const struct net_device *dev); 5159 5160 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 5161 netdev_features_t f2) 5162 { 5163 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 5164 if (f1 & NETIF_F_HW_CSUM) 5165 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5166 else 5167 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5168 } 5169 5170 return f1 & f2; 5171 } 5172 5173 static inline netdev_features_t netdev_get_wanted_features( 5174 struct net_device *dev) 5175 { 5176 return (dev->features & ~dev->hw_features) | dev->wanted_features; 5177 } 5178 netdev_features_t netdev_increment_features(netdev_features_t all, 5179 netdev_features_t one, netdev_features_t mask); 5180 5181 /* Allow TSO being used on stacked device : 5182 * Performing the GSO segmentation before last device 5183 * is a performance improvement. 5184 */ 5185 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 5186 netdev_features_t mask) 5187 { 5188 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 5189 } 5190 5191 int __netdev_update_features(struct net_device *dev); 5192 void netdev_update_features(struct net_device *dev); 5193 void netdev_change_features(struct net_device *dev); 5194 5195 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 5196 struct net_device *dev); 5197 5198 netdev_features_t passthru_features_check(struct sk_buff *skb, 5199 struct net_device *dev, 5200 netdev_features_t features); 5201 netdev_features_t netif_skb_features(struct sk_buff *skb); 5202 void skb_warn_bad_offload(const struct sk_buff *skb); 5203 5204 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 5205 { 5206 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 5207 5208 /* check flags correspondence */ 5209 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 5210 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 5211 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 5212 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 5213 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 5214 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 5215 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 5216 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 5217 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 5218 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 5219 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 5220 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 5221 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 5222 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 5223 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 5224 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 5225 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 5226 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 5227 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 5228 5229 return (features & feature) == feature; 5230 } 5231 5232 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 5233 { 5234 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 5235 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 5236 } 5237 5238 static inline bool netif_needs_gso(struct sk_buff *skb, 5239 netdev_features_t features) 5240 { 5241 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 5242 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 5243 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 5244 } 5245 5246 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 5247 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 5248 void netif_inherit_tso_max(struct net_device *to, 5249 const struct net_device *from); 5250 5251 static inline unsigned int 5252 netif_get_gro_max_size(const struct net_device *dev, const struct sk_buff *skb) 5253 { 5254 /* pairs with WRITE_ONCE() in netif_set_gro(_ipv4)_max_size() */ 5255 return skb->protocol == htons(ETH_P_IPV6) ? 5256 READ_ONCE(dev->gro_max_size) : 5257 READ_ONCE(dev->gro_ipv4_max_size); 5258 } 5259 5260 static inline unsigned int 5261 netif_get_gso_max_size(const struct net_device *dev, const struct sk_buff *skb) 5262 { 5263 /* pairs with WRITE_ONCE() in netif_set_gso(_ipv4)_max_size() */ 5264 return skb->protocol == htons(ETH_P_IPV6) ? 5265 READ_ONCE(dev->gso_max_size) : 5266 READ_ONCE(dev->gso_ipv4_max_size); 5267 } 5268 5269 static inline bool netif_is_macsec(const struct net_device *dev) 5270 { 5271 return dev->priv_flags & IFF_MACSEC; 5272 } 5273 5274 static inline bool netif_is_macvlan(const struct net_device *dev) 5275 { 5276 return dev->priv_flags & IFF_MACVLAN; 5277 } 5278 5279 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5280 { 5281 return dev->priv_flags & IFF_MACVLAN_PORT; 5282 } 5283 5284 static inline bool netif_is_bond_master(const struct net_device *dev) 5285 { 5286 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5287 } 5288 5289 static inline bool netif_is_bond_slave(const struct net_device *dev) 5290 { 5291 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5292 } 5293 5294 static inline bool netif_supports_nofcs(struct net_device *dev) 5295 { 5296 return dev->priv_flags & IFF_SUPP_NOFCS; 5297 } 5298 5299 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5300 { 5301 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5302 } 5303 5304 static inline bool netif_is_l3_master(const struct net_device *dev) 5305 { 5306 return dev->priv_flags & IFF_L3MDEV_MASTER; 5307 } 5308 5309 static inline bool netif_is_l3_slave(const struct net_device *dev) 5310 { 5311 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5312 } 5313 5314 static inline int dev_sdif(const struct net_device *dev) 5315 { 5316 #ifdef CONFIG_NET_L3_MASTER_DEV 5317 if (netif_is_l3_slave(dev)) 5318 return dev->ifindex; 5319 #endif 5320 return 0; 5321 } 5322 5323 static inline bool netif_is_bridge_master(const struct net_device *dev) 5324 { 5325 return dev->priv_flags & IFF_EBRIDGE; 5326 } 5327 5328 static inline bool netif_is_bridge_port(const struct net_device *dev) 5329 { 5330 return dev->priv_flags & IFF_BRIDGE_PORT; 5331 } 5332 5333 static inline bool netif_is_ovs_master(const struct net_device *dev) 5334 { 5335 return dev->priv_flags & IFF_OPENVSWITCH; 5336 } 5337 5338 static inline bool netif_is_ovs_port(const struct net_device *dev) 5339 { 5340 return dev->priv_flags & IFF_OVS_DATAPATH; 5341 } 5342 5343 static inline bool netif_is_any_bridge_master(const struct net_device *dev) 5344 { 5345 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev); 5346 } 5347 5348 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5349 { 5350 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5351 } 5352 5353 static inline bool netif_is_team_master(const struct net_device *dev) 5354 { 5355 return dev->priv_flags & IFF_TEAM; 5356 } 5357 5358 static inline bool netif_is_team_port(const struct net_device *dev) 5359 { 5360 return dev->priv_flags & IFF_TEAM_PORT; 5361 } 5362 5363 static inline bool netif_is_lag_master(const struct net_device *dev) 5364 { 5365 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5366 } 5367 5368 static inline bool netif_is_lag_port(const struct net_device *dev) 5369 { 5370 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5371 } 5372 5373 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5374 { 5375 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5376 } 5377 5378 static inline bool netif_is_failover(const struct net_device *dev) 5379 { 5380 return dev->priv_flags & IFF_FAILOVER; 5381 } 5382 5383 static inline bool netif_is_failover_slave(const struct net_device *dev) 5384 { 5385 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5386 } 5387 5388 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5389 static inline void netif_keep_dst(struct net_device *dev) 5390 { 5391 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5392 } 5393 5394 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5395 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5396 { 5397 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5398 return netif_is_macsec(dev); 5399 } 5400 5401 extern struct pernet_operations __net_initdata loopback_net_ops; 5402 5403 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5404 5405 /* netdev_printk helpers, similar to dev_printk */ 5406 5407 static inline const char *netdev_name(const struct net_device *dev) 5408 { 5409 if (!dev->name[0] || strchr(dev->name, '%')) 5410 return "(unnamed net_device)"; 5411 return dev->name; 5412 } 5413 5414 static inline const char *netdev_reg_state(const struct net_device *dev) 5415 { 5416 u8 reg_state = READ_ONCE(dev->reg_state); 5417 5418 switch (reg_state) { 5419 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5420 case NETREG_REGISTERED: return ""; 5421 case NETREG_UNREGISTERING: return " (unregistering)"; 5422 case NETREG_UNREGISTERED: return " (unregistered)"; 5423 case NETREG_RELEASED: return " (released)"; 5424 case NETREG_DUMMY: return " (dummy)"; 5425 } 5426 5427 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state); 5428 return " (unknown)"; 5429 } 5430 5431 #define MODULE_ALIAS_NETDEV(device) \ 5432 MODULE_ALIAS("netdev-" device) 5433 5434 /* 5435 * netdev_WARN() acts like dev_printk(), but with the key difference 5436 * of using a WARN/WARN_ON to get the message out, including the 5437 * file/line information and a backtrace. 5438 */ 5439 #define netdev_WARN(dev, format, args...) \ 5440 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5441 netdev_reg_state(dev), ##args) 5442 5443 #define netdev_WARN_ONCE(dev, format, args...) \ 5444 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5445 netdev_reg_state(dev), ##args) 5446 5447 /* 5448 * The list of packet types we will receive (as opposed to discard) 5449 * and the routines to invoke. 5450 * 5451 * Why 16. Because with 16 the only overlap we get on a hash of the 5452 * low nibble of the protocol value is RARP/SNAP/X.25. 5453 * 5454 * 0800 IP 5455 * 0001 802.3 5456 * 0002 AX.25 5457 * 0004 802.2 5458 * 8035 RARP 5459 * 0005 SNAP 5460 * 0805 X.25 5461 * 0806 ARP 5462 * 8137 IPX 5463 * 0009 Localtalk 5464 * 86DD IPv6 5465 */ 5466 #define PTYPE_HASH_SIZE (16) 5467 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5468 5469 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5470 5471 extern struct net_device *blackhole_netdev; 5472 5473 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */ 5474 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD) 5475 #define DEV_STATS_ADD(DEV, FIELD, VAL) \ 5476 atomic_long_add((VAL), &(DEV)->stats.__##FIELD) 5477 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD) 5478 5479 #endif /* _LINUX_NETDEVICE_H */ 5480