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_config; 67 struct netdev_name_node; 68 struct sd_flow_limit; 69 struct sfp_bus; 70 /* 802.11 specific */ 71 struct wireless_dev; 72 /* 802.15.4 specific */ 73 struct wpan_dev; 74 struct mpls_dev; 75 /* UDP Tunnel offloads */ 76 struct udp_tunnel_info; 77 struct udp_tunnel_nic_info; 78 struct udp_tunnel_nic; 79 struct bpf_prog; 80 struct xdp_buff; 81 struct xdp_frame; 82 struct xdp_metadata_ops; 83 struct xdp_md; 84 struct ethtool_netdev_state; 85 struct phy_link_topology; 86 struct hwtstamp_provider; 87 88 typedef u32 xdp_features_t; 89 90 void synchronize_net(void); 91 void netdev_set_default_ethtool_ops(struct net_device *dev, 92 const struct ethtool_ops *ops); 93 void netdev_sw_irq_coalesce_default_on(struct net_device *dev); 94 95 /* Backlog congestion levels */ 96 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 97 #define NET_RX_DROP 1 /* packet dropped */ 98 99 #define MAX_NEST_DEV 8 100 101 /* 102 * Transmit return codes: transmit return codes originate from three different 103 * namespaces: 104 * 105 * - qdisc return codes 106 * - driver transmit return codes 107 * - errno values 108 * 109 * Drivers are allowed to return any one of those in their hard_start_xmit() 110 * function. Real network devices commonly used with qdiscs should only return 111 * the driver transmit return codes though - when qdiscs are used, the actual 112 * transmission happens asynchronously, so the value is not propagated to 113 * higher layers. Virtual network devices transmit synchronously; in this case 114 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 115 * others are propagated to higher layers. 116 */ 117 118 /* qdisc ->enqueue() return codes. */ 119 #define NET_XMIT_SUCCESS 0x00 120 #define NET_XMIT_DROP 0x01 /* skb dropped */ 121 #define NET_XMIT_CN 0x02 /* congestion notification */ 122 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 123 124 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 125 * indicates that the device will soon be dropping packets, or already drops 126 * some packets of the same priority; prompting us to send less aggressively. */ 127 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 128 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 129 130 /* Driver transmit return codes */ 131 #define NETDEV_TX_MASK 0xf0 132 133 enum netdev_tx { 134 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 135 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 136 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 137 }; 138 typedef enum netdev_tx netdev_tx_t; 139 140 /* 141 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 142 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 143 */ 144 static inline bool dev_xmit_complete(int rc) 145 { 146 /* 147 * Positive cases with an skb consumed by a driver: 148 * - successful transmission (rc == NETDEV_TX_OK) 149 * - error while transmitting (rc < 0) 150 * - error while queueing to a different device (rc & NET_XMIT_MASK) 151 */ 152 if (likely(rc < NET_XMIT_MASK)) 153 return true; 154 155 return false; 156 } 157 158 /* 159 * Compute the worst-case header length according to the protocols 160 * used. 161 */ 162 163 #if defined(CONFIG_HYPERV_NET) 164 # define LL_MAX_HEADER 128 165 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 166 # if defined(CONFIG_MAC80211_MESH) 167 # define LL_MAX_HEADER 128 168 # else 169 # define LL_MAX_HEADER 96 170 # endif 171 #else 172 # define LL_MAX_HEADER 32 173 #endif 174 175 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 176 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 177 #define MAX_HEADER LL_MAX_HEADER 178 #else 179 #define MAX_HEADER (LL_MAX_HEADER + 48) 180 #endif 181 182 /* 183 * Old network device statistics. Fields are native words 184 * (unsigned long) so they can be read and written atomically. 185 */ 186 187 #define NET_DEV_STAT(FIELD) \ 188 union { \ 189 unsigned long FIELD; \ 190 atomic_long_t __##FIELD; \ 191 } 192 193 struct net_device_stats { 194 NET_DEV_STAT(rx_packets); 195 NET_DEV_STAT(tx_packets); 196 NET_DEV_STAT(rx_bytes); 197 NET_DEV_STAT(tx_bytes); 198 NET_DEV_STAT(rx_errors); 199 NET_DEV_STAT(tx_errors); 200 NET_DEV_STAT(rx_dropped); 201 NET_DEV_STAT(tx_dropped); 202 NET_DEV_STAT(multicast); 203 NET_DEV_STAT(collisions); 204 NET_DEV_STAT(rx_length_errors); 205 NET_DEV_STAT(rx_over_errors); 206 NET_DEV_STAT(rx_crc_errors); 207 NET_DEV_STAT(rx_frame_errors); 208 NET_DEV_STAT(rx_fifo_errors); 209 NET_DEV_STAT(rx_missed_errors); 210 NET_DEV_STAT(tx_aborted_errors); 211 NET_DEV_STAT(tx_carrier_errors); 212 NET_DEV_STAT(tx_fifo_errors); 213 NET_DEV_STAT(tx_heartbeat_errors); 214 NET_DEV_STAT(tx_window_errors); 215 NET_DEV_STAT(rx_compressed); 216 NET_DEV_STAT(tx_compressed); 217 }; 218 #undef NET_DEV_STAT 219 220 /* per-cpu stats, allocated on demand. 221 * Try to fit them in a single cache line, for dev_get_stats() sake. 222 */ 223 struct net_device_core_stats { 224 unsigned long rx_dropped; 225 unsigned long tx_dropped; 226 unsigned long rx_nohandler; 227 unsigned long rx_otherhost_dropped; 228 } __aligned(4 * sizeof(unsigned long)); 229 230 #include <linux/cache.h> 231 #include <linux/skbuff.h> 232 233 struct neighbour; 234 struct neigh_parms; 235 struct sk_buff; 236 237 struct netdev_hw_addr { 238 struct list_head list; 239 struct rb_node node; 240 unsigned char addr[MAX_ADDR_LEN]; 241 unsigned char type; 242 #define NETDEV_HW_ADDR_T_LAN 1 243 #define NETDEV_HW_ADDR_T_SAN 2 244 #define NETDEV_HW_ADDR_T_UNICAST 3 245 #define NETDEV_HW_ADDR_T_MULTICAST 4 246 bool global_use; 247 int sync_cnt; 248 int refcount; 249 int synced; 250 struct rcu_head rcu_head; 251 }; 252 253 struct netdev_hw_addr_list { 254 struct list_head list; 255 int count; 256 257 /* Auxiliary tree for faster lookup on addition and deletion */ 258 struct rb_root tree; 259 }; 260 261 #define netdev_hw_addr_list_count(l) ((l)->count) 262 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 263 #define netdev_hw_addr_list_for_each(ha, l) \ 264 list_for_each_entry(ha, &(l)->list, list) 265 266 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 267 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 268 #define netdev_for_each_uc_addr(ha, dev) \ 269 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 270 #define netdev_for_each_synced_uc_addr(_ha, _dev) \ 271 netdev_for_each_uc_addr((_ha), (_dev)) \ 272 if ((_ha)->sync_cnt) 273 274 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 275 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 276 #define netdev_for_each_mc_addr(ha, dev) \ 277 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 278 #define netdev_for_each_synced_mc_addr(_ha, _dev) \ 279 netdev_for_each_mc_addr((_ha), (_dev)) \ 280 if ((_ha)->sync_cnt) 281 282 struct hh_cache { 283 unsigned int hh_len; 284 seqlock_t hh_lock; 285 286 /* cached hardware header; allow for machine alignment needs. */ 287 #define HH_DATA_MOD 16 288 #define HH_DATA_OFF(__len) \ 289 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 290 #define HH_DATA_ALIGN(__len) \ 291 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 292 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 293 }; 294 295 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 296 * Alternative is: 297 * dev->hard_header_len ? (dev->hard_header_len + 298 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 299 * 300 * We could use other alignment values, but we must maintain the 301 * relationship HH alignment <= LL alignment. 302 */ 303 #define LL_RESERVED_SPACE(dev) \ 304 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \ 305 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 306 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 307 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \ 308 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 309 310 struct header_ops { 311 int (*create) (struct sk_buff *skb, struct net_device *dev, 312 unsigned short type, const void *daddr, 313 const void *saddr, unsigned int len); 314 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 315 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 316 void (*cache_update)(struct hh_cache *hh, 317 const struct net_device *dev, 318 const unsigned char *haddr); 319 bool (*validate)(const char *ll_header, unsigned int len); 320 __be16 (*parse_protocol)(const struct sk_buff *skb); 321 }; 322 323 /* These flag bits are private to the generic network queueing 324 * layer; they may not be explicitly referenced by any other 325 * code. 326 */ 327 328 enum netdev_state_t { 329 __LINK_STATE_START, 330 __LINK_STATE_PRESENT, 331 __LINK_STATE_NOCARRIER, 332 __LINK_STATE_LINKWATCH_PENDING, 333 __LINK_STATE_DORMANT, 334 __LINK_STATE_TESTING, 335 }; 336 337 struct gro_list { 338 struct list_head list; 339 int count; 340 }; 341 342 /* 343 * size of gro hash buckets, must less than bit number of 344 * napi_struct::gro_bitmask 345 */ 346 #define GRO_HASH_BUCKETS 8 347 348 /* 349 * Structure for per-NAPI config 350 */ 351 struct napi_config { 352 u64 gro_flush_timeout; 353 u64 irq_suspend_timeout; 354 u32 defer_hard_irqs; 355 unsigned int napi_id; 356 }; 357 358 /* 359 * Structure for NAPI scheduling similar to tasklet but with weighting 360 */ 361 struct napi_struct { 362 /* The poll_list must only be managed by the entity which 363 * changes the state of the NAPI_STATE_SCHED bit. This means 364 * whoever atomically sets that bit can add this napi_struct 365 * to the per-CPU poll_list, and whoever clears that bit 366 * can remove from the list right before clearing the bit. 367 */ 368 struct list_head poll_list; 369 370 unsigned long state; 371 int weight; 372 u32 defer_hard_irqs_count; 373 unsigned long gro_bitmask; 374 int (*poll)(struct napi_struct *, int); 375 #ifdef CONFIG_NETPOLL 376 /* CPU actively polling if netpoll is configured */ 377 int poll_owner; 378 #endif 379 /* CPU on which NAPI has been scheduled for processing */ 380 int list_owner; 381 struct net_device *dev; 382 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 383 struct sk_buff *skb; 384 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 385 int rx_count; /* length of rx_list */ 386 unsigned int napi_id; /* protected by netdev_lock */ 387 struct hrtimer timer; 388 /* all fields past this point are write-protected by netdev_lock */ 389 struct task_struct *thread; 390 unsigned long gro_flush_timeout; 391 unsigned long irq_suspend_timeout; 392 u32 defer_hard_irqs; 393 /* control-path-only fields follow */ 394 struct list_head dev_list; 395 struct hlist_node napi_hash_node; 396 int irq; 397 int index; 398 struct napi_config *config; 399 }; 400 401 enum { 402 NAPI_STATE_SCHED, /* Poll is scheduled */ 403 NAPI_STATE_MISSED, /* reschedule a napi */ 404 NAPI_STATE_DISABLE, /* Disable pending */ 405 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 406 NAPI_STATE_LISTED, /* NAPI added to system lists */ 407 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 408 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 409 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 410 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 411 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 412 }; 413 414 enum { 415 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 416 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 417 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 418 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 419 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 420 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 421 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 422 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 423 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 424 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 425 }; 426 427 enum gro_result { 428 GRO_MERGED, 429 GRO_MERGED_FREE, 430 GRO_HELD, 431 GRO_NORMAL, 432 GRO_CONSUMED, 433 }; 434 typedef enum gro_result gro_result_t; 435 436 /* 437 * enum rx_handler_result - Possible return values for rx_handlers. 438 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 439 * further. 440 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 441 * case skb->dev was changed by rx_handler. 442 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 443 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 444 * 445 * rx_handlers are functions called from inside __netif_receive_skb(), to do 446 * special processing of the skb, prior to delivery to protocol handlers. 447 * 448 * Currently, a net_device can only have a single rx_handler registered. Trying 449 * to register a second rx_handler will return -EBUSY. 450 * 451 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 452 * To unregister a rx_handler on a net_device, use 453 * netdev_rx_handler_unregister(). 454 * 455 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 456 * do with the skb. 457 * 458 * If the rx_handler consumed the skb in some way, it should return 459 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 460 * the skb to be delivered in some other way. 461 * 462 * If the rx_handler changed skb->dev, to divert the skb to another 463 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 464 * new device will be called if it exists. 465 * 466 * If the rx_handler decides the skb should be ignored, it should return 467 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 468 * are registered on exact device (ptype->dev == skb->dev). 469 * 470 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 471 * delivered, it should return RX_HANDLER_PASS. 472 * 473 * A device without a registered rx_handler will behave as if rx_handler 474 * returned RX_HANDLER_PASS. 475 */ 476 477 enum rx_handler_result { 478 RX_HANDLER_CONSUMED, 479 RX_HANDLER_ANOTHER, 480 RX_HANDLER_EXACT, 481 RX_HANDLER_PASS, 482 }; 483 typedef enum rx_handler_result rx_handler_result_t; 484 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 485 486 void __napi_schedule(struct napi_struct *n); 487 void __napi_schedule_irqoff(struct napi_struct *n); 488 489 static inline bool napi_disable_pending(struct napi_struct *n) 490 { 491 return test_bit(NAPI_STATE_DISABLE, &n->state); 492 } 493 494 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 495 { 496 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 497 } 498 499 /** 500 * napi_is_scheduled - test if NAPI is scheduled 501 * @n: NAPI context 502 * 503 * This check is "best-effort". With no locking implemented, 504 * a NAPI can be scheduled or terminate right after this check 505 * and produce not precise results. 506 * 507 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled 508 * should not be used normally and napi_schedule should be 509 * used instead. 510 * 511 * Use only if the driver really needs to check if a NAPI 512 * is scheduled for example in the context of delayed timer 513 * that can be skipped if a NAPI is already scheduled. 514 * 515 * Return: True if NAPI is scheduled, False otherwise. 516 */ 517 static inline bool napi_is_scheduled(struct napi_struct *n) 518 { 519 return test_bit(NAPI_STATE_SCHED, &n->state); 520 } 521 522 bool napi_schedule_prep(struct napi_struct *n); 523 524 /** 525 * napi_schedule - schedule NAPI poll 526 * @n: NAPI context 527 * 528 * Schedule NAPI poll routine to be called if it is not already 529 * running. 530 * Return: true if we schedule a NAPI or false if not. 531 * Refer to napi_schedule_prep() for additional reason on why 532 * a NAPI might not be scheduled. 533 */ 534 static inline bool napi_schedule(struct napi_struct *n) 535 { 536 if (napi_schedule_prep(n)) { 537 __napi_schedule(n); 538 return true; 539 } 540 541 return false; 542 } 543 544 /** 545 * napi_schedule_irqoff - schedule NAPI poll 546 * @n: NAPI context 547 * 548 * Variant of napi_schedule(), assuming hard irqs are masked. 549 */ 550 static inline void napi_schedule_irqoff(struct napi_struct *n) 551 { 552 if (napi_schedule_prep(n)) 553 __napi_schedule_irqoff(n); 554 } 555 556 /** 557 * napi_complete_done - NAPI processing complete 558 * @n: NAPI context 559 * @work_done: number of packets processed 560 * 561 * Mark NAPI processing as complete. Should only be called if poll budget 562 * has not been completely consumed. 563 * Prefer over napi_complete(). 564 * Return: false if device should avoid rearming interrupts. 565 */ 566 bool napi_complete_done(struct napi_struct *n, int work_done); 567 568 static inline bool napi_complete(struct napi_struct *n) 569 { 570 return napi_complete_done(n, 0); 571 } 572 573 int dev_set_threaded(struct net_device *dev, bool threaded); 574 575 void napi_disable(struct napi_struct *n); 576 void napi_disable_locked(struct napi_struct *n); 577 578 void napi_enable(struct napi_struct *n); 579 void napi_enable_locked(struct napi_struct *n); 580 581 /** 582 * napi_synchronize - wait until NAPI is not running 583 * @n: NAPI context 584 * 585 * Wait until NAPI is done being scheduled on this context. 586 * Waits till any outstanding processing completes but 587 * does not disable future activations. 588 */ 589 static inline void napi_synchronize(const struct napi_struct *n) 590 { 591 if (IS_ENABLED(CONFIG_SMP)) 592 while (test_bit(NAPI_STATE_SCHED, &n->state)) 593 msleep(1); 594 else 595 barrier(); 596 } 597 598 /** 599 * napi_if_scheduled_mark_missed - if napi is running, set the 600 * NAPIF_STATE_MISSED 601 * @n: NAPI context 602 * 603 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 604 * NAPI is scheduled. 605 **/ 606 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 607 { 608 unsigned long val, new; 609 610 val = READ_ONCE(n->state); 611 do { 612 if (val & NAPIF_STATE_DISABLE) 613 return true; 614 615 if (!(val & NAPIF_STATE_SCHED)) 616 return false; 617 618 new = val | NAPIF_STATE_MISSED; 619 } while (!try_cmpxchg(&n->state, &val, new)); 620 621 return true; 622 } 623 624 enum netdev_queue_state_t { 625 __QUEUE_STATE_DRV_XOFF, 626 __QUEUE_STATE_STACK_XOFF, 627 __QUEUE_STATE_FROZEN, 628 }; 629 630 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 631 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 632 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 633 634 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 635 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 636 QUEUE_STATE_FROZEN) 637 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 638 QUEUE_STATE_FROZEN) 639 640 /* 641 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 642 * netif_tx_* functions below are used to manipulate this flag. The 643 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 644 * queue independently. The netif_xmit_*stopped functions below are called 645 * to check if the queue has been stopped by the driver or stack (either 646 * of the XOFF bits are set in the state). Drivers should not need to call 647 * netif_xmit*stopped functions, they should only be using netif_tx_*. 648 */ 649 650 struct netdev_queue { 651 /* 652 * read-mostly part 653 */ 654 struct net_device *dev; 655 netdevice_tracker dev_tracker; 656 657 struct Qdisc __rcu *qdisc; 658 struct Qdisc __rcu *qdisc_sleeping; 659 #ifdef CONFIG_SYSFS 660 struct kobject kobj; 661 const struct attribute_group **groups; 662 #endif 663 unsigned long tx_maxrate; 664 /* 665 * Number of TX timeouts for this queue 666 * (/sys/class/net/DEV/Q/trans_timeout) 667 */ 668 atomic_long_t trans_timeout; 669 670 /* Subordinate device that the queue has been assigned to */ 671 struct net_device *sb_dev; 672 #ifdef CONFIG_XDP_SOCKETS 673 struct xsk_buff_pool *pool; 674 #endif 675 676 /* 677 * write-mostly part 678 */ 679 #ifdef CONFIG_BQL 680 struct dql dql; 681 #endif 682 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 683 int xmit_lock_owner; 684 /* 685 * Time (in jiffies) of last Tx 686 */ 687 unsigned long trans_start; 688 689 unsigned long state; 690 691 /* 692 * slow- / control-path part 693 */ 694 /* NAPI instance for the queue 695 * Readers and writers must hold RTNL 696 */ 697 struct napi_struct *napi; 698 699 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 700 int numa_node; 701 #endif 702 } ____cacheline_aligned_in_smp; 703 704 extern int sysctl_fb_tunnels_only_for_init_net; 705 extern int sysctl_devconf_inherit_init_net; 706 707 /* 708 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 709 * == 1 : For initns only 710 * == 2 : For none. 711 */ 712 static inline bool net_has_fallback_tunnels(const struct net *net) 713 { 714 #if IS_ENABLED(CONFIG_SYSCTL) 715 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net); 716 717 return !fb_tunnels_only_for_init_net || 718 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1); 719 #else 720 return true; 721 #endif 722 } 723 724 static inline int net_inherit_devconf(void) 725 { 726 #if IS_ENABLED(CONFIG_SYSCTL) 727 return READ_ONCE(sysctl_devconf_inherit_init_net); 728 #else 729 return 0; 730 #endif 731 } 732 733 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 734 { 735 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 736 return q->numa_node; 737 #else 738 return NUMA_NO_NODE; 739 #endif 740 } 741 742 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 743 { 744 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 745 q->numa_node = node; 746 #endif 747 } 748 749 #ifdef CONFIG_RFS_ACCEL 750 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 751 u16 filter_id); 752 #endif 753 754 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 755 enum xps_map_type { 756 XPS_CPUS = 0, 757 XPS_RXQS, 758 XPS_MAPS_MAX, 759 }; 760 761 #ifdef CONFIG_XPS 762 /* 763 * This structure holds an XPS map which can be of variable length. The 764 * map is an array of queues. 765 */ 766 struct xps_map { 767 unsigned int len; 768 unsigned int alloc_len; 769 struct rcu_head rcu; 770 u16 queues[]; 771 }; 772 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 773 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 774 - sizeof(struct xps_map)) / sizeof(u16)) 775 776 /* 777 * This structure holds all XPS maps for device. Maps are indexed by CPU. 778 * 779 * We keep track of the number of cpus/rxqs used when the struct is allocated, 780 * in nr_ids. This will help not accessing out-of-bound memory. 781 * 782 * We keep track of the number of traffic classes used when the struct is 783 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 784 * not crossing its upper bound, as the original dev->num_tc can be updated in 785 * the meantime. 786 */ 787 struct xps_dev_maps { 788 struct rcu_head rcu; 789 unsigned int nr_ids; 790 s16 num_tc; 791 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 792 }; 793 794 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 795 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 796 797 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 798 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 799 800 #endif /* CONFIG_XPS */ 801 802 #define TC_MAX_QUEUE 16 803 #define TC_BITMASK 15 804 /* HW offloaded queuing disciplines txq count and offset maps */ 805 struct netdev_tc_txq { 806 u16 count; 807 u16 offset; 808 }; 809 810 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 811 /* 812 * This structure is to hold information about the device 813 * configured to run FCoE protocol stack. 814 */ 815 struct netdev_fcoe_hbainfo { 816 char manufacturer[64]; 817 char serial_number[64]; 818 char hardware_version[64]; 819 char driver_version[64]; 820 char optionrom_version[64]; 821 char firmware_version[64]; 822 char model[256]; 823 char model_description[256]; 824 }; 825 #endif 826 827 #define MAX_PHYS_ITEM_ID_LEN 32 828 829 /* This structure holds a unique identifier to identify some 830 * physical item (port for example) used by a netdevice. 831 */ 832 struct netdev_phys_item_id { 833 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 834 unsigned char id_len; 835 }; 836 837 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 838 struct netdev_phys_item_id *b) 839 { 840 return a->id_len == b->id_len && 841 memcmp(a->id, b->id, a->id_len) == 0; 842 } 843 844 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 845 struct sk_buff *skb, 846 struct net_device *sb_dev); 847 848 enum net_device_path_type { 849 DEV_PATH_ETHERNET = 0, 850 DEV_PATH_VLAN, 851 DEV_PATH_BRIDGE, 852 DEV_PATH_PPPOE, 853 DEV_PATH_DSA, 854 DEV_PATH_MTK_WDMA, 855 }; 856 857 struct net_device_path { 858 enum net_device_path_type type; 859 const struct net_device *dev; 860 union { 861 struct { 862 u16 id; 863 __be16 proto; 864 u8 h_dest[ETH_ALEN]; 865 } encap; 866 struct { 867 enum { 868 DEV_PATH_BR_VLAN_KEEP, 869 DEV_PATH_BR_VLAN_TAG, 870 DEV_PATH_BR_VLAN_UNTAG, 871 DEV_PATH_BR_VLAN_UNTAG_HW, 872 } vlan_mode; 873 u16 vlan_id; 874 __be16 vlan_proto; 875 } bridge; 876 struct { 877 int port; 878 u16 proto; 879 } dsa; 880 struct { 881 u8 wdma_idx; 882 u8 queue; 883 u16 wcid; 884 u8 bss; 885 u8 amsdu; 886 } mtk_wdma; 887 }; 888 }; 889 890 #define NET_DEVICE_PATH_STACK_MAX 5 891 #define NET_DEVICE_PATH_VLAN_MAX 2 892 893 struct net_device_path_stack { 894 int num_paths; 895 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 896 }; 897 898 struct net_device_path_ctx { 899 const struct net_device *dev; 900 u8 daddr[ETH_ALEN]; 901 902 int num_vlans; 903 struct { 904 u16 id; 905 __be16 proto; 906 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 907 }; 908 909 enum tc_setup_type { 910 TC_QUERY_CAPS, 911 TC_SETUP_QDISC_MQPRIO, 912 TC_SETUP_CLSU32, 913 TC_SETUP_CLSFLOWER, 914 TC_SETUP_CLSMATCHALL, 915 TC_SETUP_CLSBPF, 916 TC_SETUP_BLOCK, 917 TC_SETUP_QDISC_CBS, 918 TC_SETUP_QDISC_RED, 919 TC_SETUP_QDISC_PRIO, 920 TC_SETUP_QDISC_MQ, 921 TC_SETUP_QDISC_ETF, 922 TC_SETUP_ROOT_QDISC, 923 TC_SETUP_QDISC_GRED, 924 TC_SETUP_QDISC_TAPRIO, 925 TC_SETUP_FT, 926 TC_SETUP_QDISC_ETS, 927 TC_SETUP_QDISC_TBF, 928 TC_SETUP_QDISC_FIFO, 929 TC_SETUP_QDISC_HTB, 930 TC_SETUP_ACT, 931 }; 932 933 /* These structures hold the attributes of bpf state that are being passed 934 * to the netdevice through the bpf op. 935 */ 936 enum bpf_netdev_command { 937 /* Set or clear a bpf program used in the earliest stages of packet 938 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 939 * is responsible for calling bpf_prog_put on any old progs that are 940 * stored. In case of error, the callee need not release the new prog 941 * reference, but on success it takes ownership and must bpf_prog_put 942 * when it is no longer used. 943 */ 944 XDP_SETUP_PROG, 945 XDP_SETUP_PROG_HW, 946 /* BPF program for offload callbacks, invoked at program load time. */ 947 BPF_OFFLOAD_MAP_ALLOC, 948 BPF_OFFLOAD_MAP_FREE, 949 XDP_SETUP_XSK_POOL, 950 }; 951 952 struct bpf_prog_offload_ops; 953 struct netlink_ext_ack; 954 struct xdp_umem; 955 struct xdp_dev_bulk_queue; 956 struct bpf_xdp_link; 957 958 enum bpf_xdp_mode { 959 XDP_MODE_SKB = 0, 960 XDP_MODE_DRV = 1, 961 XDP_MODE_HW = 2, 962 __MAX_XDP_MODE 963 }; 964 965 struct bpf_xdp_entity { 966 struct bpf_prog *prog; 967 struct bpf_xdp_link *link; 968 }; 969 970 struct netdev_bpf { 971 enum bpf_netdev_command command; 972 union { 973 /* XDP_SETUP_PROG */ 974 struct { 975 u32 flags; 976 struct bpf_prog *prog; 977 struct netlink_ext_ack *extack; 978 }; 979 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 980 struct { 981 struct bpf_offloaded_map *offmap; 982 }; 983 /* XDP_SETUP_XSK_POOL */ 984 struct { 985 struct xsk_buff_pool *pool; 986 u16 queue_id; 987 } xsk; 988 }; 989 }; 990 991 /* Flags for ndo_xsk_wakeup. */ 992 #define XDP_WAKEUP_RX (1 << 0) 993 #define XDP_WAKEUP_TX (1 << 1) 994 995 #ifdef CONFIG_XFRM_OFFLOAD 996 struct xfrmdev_ops { 997 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack); 998 void (*xdo_dev_state_delete) (struct xfrm_state *x); 999 void (*xdo_dev_state_free) (struct xfrm_state *x); 1000 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 1001 struct xfrm_state *x); 1002 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1003 void (*xdo_dev_state_update_stats) (struct xfrm_state *x); 1004 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack); 1005 void (*xdo_dev_policy_delete) (struct xfrm_policy *x); 1006 void (*xdo_dev_policy_free) (struct xfrm_policy *x); 1007 }; 1008 #endif 1009 1010 struct dev_ifalias { 1011 struct rcu_head rcuhead; 1012 char ifalias[]; 1013 }; 1014 1015 struct devlink; 1016 struct tlsdev_ops; 1017 1018 struct netdev_net_notifier { 1019 struct list_head list; 1020 struct notifier_block *nb; 1021 }; 1022 1023 /* 1024 * This structure defines the management hooks for network devices. 1025 * The following hooks can be defined; unless noted otherwise, they are 1026 * optional and can be filled with a null pointer. 1027 * 1028 * int (*ndo_init)(struct net_device *dev); 1029 * This function is called once when a network device is registered. 1030 * The network device can use this for any late stage initialization 1031 * or semantic validation. It can fail with an error code which will 1032 * be propagated back to register_netdev. 1033 * 1034 * void (*ndo_uninit)(struct net_device *dev); 1035 * This function is called when device is unregistered or when registration 1036 * fails. It is not called if init fails. 1037 * 1038 * int (*ndo_open)(struct net_device *dev); 1039 * This function is called when a network device transitions to the up 1040 * state. 1041 * 1042 * int (*ndo_stop)(struct net_device *dev); 1043 * This function is called when a network device transitions to the down 1044 * state. 1045 * 1046 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1047 * struct net_device *dev); 1048 * Called when a packet needs to be transmitted. 1049 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1050 * the queue before that can happen; it's for obsolete devices and weird 1051 * corner cases, but the stack really does a non-trivial amount 1052 * of useless work if you return NETDEV_TX_BUSY. 1053 * Required; cannot be NULL. 1054 * 1055 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1056 * struct net_device *dev 1057 * netdev_features_t features); 1058 * Called by core transmit path to determine if device is capable of 1059 * performing offload operations on a given packet. This is to give 1060 * the device an opportunity to implement any restrictions that cannot 1061 * be otherwise expressed by feature flags. The check is called with 1062 * the set of features that the stack has calculated and it returns 1063 * those the driver believes to be appropriate. 1064 * 1065 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1066 * struct net_device *sb_dev); 1067 * Called to decide which queue to use when device supports multiple 1068 * transmit queues. 1069 * 1070 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1071 * This function is called to allow device receiver to make 1072 * changes to configuration when multicast or promiscuous is enabled. 1073 * 1074 * void (*ndo_set_rx_mode)(struct net_device *dev); 1075 * This function is called device changes address list filtering. 1076 * If driver handles unicast address filtering, it should set 1077 * IFF_UNICAST_FLT in its priv_flags. 1078 * 1079 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1080 * This function is called when the Media Access Control address 1081 * needs to be changed. If this interface is not defined, the 1082 * MAC address can not be changed. 1083 * 1084 * int (*ndo_validate_addr)(struct net_device *dev); 1085 * Test if Media Access Control address is valid for the device. 1086 * 1087 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1088 * Old-style ioctl entry point. This is used internally by the 1089 * ieee802154 subsystem but is no longer called by the device 1090 * ioctl handler. 1091 * 1092 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1093 * Used by the bonding driver for its device specific ioctls: 1094 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1095 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1096 * 1097 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1098 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1099 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1100 * 1101 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1102 * Used to set network devices bus interface parameters. This interface 1103 * is retained for legacy reasons; new devices should use the bus 1104 * interface (PCI) for low level management. 1105 * 1106 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1107 * Called when a user wants to change the Maximum Transfer Unit 1108 * of a device. 1109 * 1110 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1111 * Callback used when the transmitter has not made any progress 1112 * for dev->watchdog ticks. 1113 * 1114 * void (*ndo_get_stats64)(struct net_device *dev, 1115 * struct rtnl_link_stats64 *storage); 1116 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1117 * Called when a user wants to get the network device usage 1118 * statistics. Drivers must do one of the following: 1119 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1120 * rtnl_link_stats64 structure passed by the caller. 1121 * 2. Define @ndo_get_stats to update a net_device_stats structure 1122 * (which should normally be dev->stats) and return a pointer to 1123 * it. The structure may be changed asynchronously only if each 1124 * field is written atomically. 1125 * 3. Update dev->stats asynchronously and atomically, and define 1126 * neither operation. 1127 * 1128 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1129 * Return true if this device supports offload stats of this attr_id. 1130 * 1131 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1132 * void *attr_data) 1133 * Get statistics for offload operations by attr_id. Write it into the 1134 * attr_data pointer. 1135 * 1136 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1137 * If device supports VLAN filtering this function is called when a 1138 * VLAN id is registered. 1139 * 1140 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1141 * If device supports VLAN filtering this function is called when a 1142 * VLAN id is unregistered. 1143 * 1144 * void (*ndo_poll_controller)(struct net_device *dev); 1145 * 1146 * SR-IOV management functions. 1147 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1148 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1149 * u8 qos, __be16 proto); 1150 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1151 * int max_tx_rate); 1152 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1153 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1154 * int (*ndo_get_vf_config)(struct net_device *dev, 1155 * int vf, struct ifla_vf_info *ivf); 1156 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1157 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1158 * struct nlattr *port[]); 1159 * 1160 * Enable or disable the VF ability to query its RSS Redirection Table and 1161 * Hash Key. This is needed since on some devices VF share this information 1162 * with PF and querying it may introduce a theoretical security risk. 1163 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1164 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1165 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1166 * void *type_data); 1167 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1168 * This is always called from the stack with the rtnl lock held and netif 1169 * tx queues stopped. This allows the netdevice to perform queue 1170 * management safely. 1171 * 1172 * Fiber Channel over Ethernet (FCoE) offload functions. 1173 * int (*ndo_fcoe_enable)(struct net_device *dev); 1174 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1175 * so the underlying device can perform whatever needed configuration or 1176 * initialization to support acceleration of FCoE traffic. 1177 * 1178 * int (*ndo_fcoe_disable)(struct net_device *dev); 1179 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1180 * so the underlying device can perform whatever needed clean-ups to 1181 * stop supporting acceleration of FCoE traffic. 1182 * 1183 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1184 * struct scatterlist *sgl, unsigned int sgc); 1185 * Called when the FCoE Initiator wants to initialize an I/O that 1186 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1187 * perform necessary setup and returns 1 to indicate the device is set up 1188 * successfully to perform DDP on this I/O, otherwise this returns 0. 1189 * 1190 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1191 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1192 * indicated by the FC exchange id 'xid', so the underlying device can 1193 * clean up and reuse resources for later DDP requests. 1194 * 1195 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1196 * struct scatterlist *sgl, unsigned int sgc); 1197 * Called when the FCoE Target wants to initialize an I/O that 1198 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1199 * perform necessary setup and returns 1 to indicate the device is set up 1200 * successfully to perform DDP on this I/O, otherwise this returns 0. 1201 * 1202 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1203 * struct netdev_fcoe_hbainfo *hbainfo); 1204 * Called when the FCoE Protocol stack wants information on the underlying 1205 * device. This information is utilized by the FCoE protocol stack to 1206 * register attributes with Fiber Channel management service as per the 1207 * FC-GS Fabric Device Management Information(FDMI) specification. 1208 * 1209 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1210 * Called when the underlying device wants to override default World Wide 1211 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1212 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1213 * protocol stack to use. 1214 * 1215 * RFS acceleration. 1216 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1217 * u16 rxq_index, u32 flow_id); 1218 * Set hardware filter for RFS. rxq_index is the target queue index; 1219 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1220 * Return the filter ID on success, or a negative error code. 1221 * 1222 * Slave management functions (for bridge, bonding, etc). 1223 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1224 * Called to make another netdev an underling. 1225 * 1226 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1227 * Called to release previously enslaved netdev. 1228 * 1229 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1230 * struct sk_buff *skb, 1231 * bool all_slaves); 1232 * Get the xmit slave of master device. If all_slaves is true, function 1233 * assume all the slaves can transmit. 1234 * 1235 * Feature/offload setting functions. 1236 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1237 * netdev_features_t features); 1238 * Adjusts the requested feature flags according to device-specific 1239 * constraints, and returns the resulting flags. Must not modify 1240 * the device state. 1241 * 1242 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1243 * Called to update device configuration to new features. Passed 1244 * feature set might be less than what was returned by ndo_fix_features()). 1245 * Must return >0 or -errno if it changed dev->features itself. 1246 * 1247 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1248 * struct net_device *dev, 1249 * const unsigned char *addr, u16 vid, u16 flags, 1250 * bool *notified, struct netlink_ext_ack *extack); 1251 * Adds an FDB entry to dev for addr. 1252 * Callee shall set *notified to true if it sent any appropriate 1253 * notification(s). Otherwise core will send a generic one. 1254 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1255 * struct net_device *dev, 1256 * const unsigned char *addr, u16 vid 1257 * bool *notified, struct netlink_ext_ack *extack); 1258 * Deletes the FDB entry from dev corresponding to addr. 1259 * Callee shall set *notified to true if it sent any appropriate 1260 * notification(s). Otherwise core will send a generic one. 1261 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev, 1262 * struct netlink_ext_ack *extack); 1263 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1264 * struct net_device *dev, struct net_device *filter_dev, 1265 * int *idx) 1266 * Used to add FDB entries to dump requests. Implementers should add 1267 * entries to skb and update idx with the number of entries. 1268 * 1269 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[], 1270 * u16 nlmsg_flags, struct netlink_ext_ack *extack); 1271 * Adds an MDB entry to dev. 1272 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[], 1273 * struct netlink_ext_ack *extack); 1274 * Deletes the MDB entry from dev. 1275 * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[], 1276 * struct netlink_ext_ack *extack); 1277 * Bulk deletes MDB entries from dev. 1278 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb, 1279 * struct netlink_callback *cb); 1280 * Dumps MDB entries from dev. The first argument (marker) in the netlink 1281 * callback is used by core rtnetlink code. 1282 * 1283 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1284 * u16 flags, struct netlink_ext_ack *extack) 1285 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1286 * struct net_device *dev, u32 filter_mask, 1287 * int nlflags) 1288 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1289 * u16 flags); 1290 * 1291 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1292 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1293 * which do not represent real hardware may define this to allow their 1294 * userspace components to manage their virtual carrier state. Devices 1295 * that determine carrier state from physical hardware properties (eg 1296 * network cables) or protocol-dependent mechanisms (eg 1297 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1298 * 1299 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1300 * struct netdev_phys_item_id *ppid); 1301 * Called to get ID of physical port of this device. If driver does 1302 * not implement this, it is assumed that the hw is not able to have 1303 * multiple net devices on single physical port. 1304 * 1305 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1306 * struct netdev_phys_item_id *ppid) 1307 * Called to get the parent ID of the physical port of this device. 1308 * 1309 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1310 * struct net_device *dev) 1311 * Called by upper layer devices to accelerate switching or other 1312 * station functionality into hardware. 'pdev is the lowerdev 1313 * to use for the offload and 'dev' is the net device that will 1314 * back the offload. Returns a pointer to the private structure 1315 * the upper layer will maintain. 1316 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1317 * Called by upper layer device to delete the station created 1318 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1319 * the station and priv is the structure returned by the add 1320 * operation. 1321 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1322 * int queue_index, u32 maxrate); 1323 * Called when a user wants to set a max-rate limitation of specific 1324 * TX queue. 1325 * int (*ndo_get_iflink)(const struct net_device *dev); 1326 * Called to get the iflink value of this device. 1327 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1328 * This function is used to get egress tunnel information for given skb. 1329 * This is useful for retrieving outer tunnel header parameters while 1330 * sampling packet. 1331 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1332 * This function is used to specify the headroom that the skb must 1333 * consider when allocation skb during packet reception. Setting 1334 * appropriate rx headroom value allows avoiding skb head copy on 1335 * forward. Setting a negative value resets the rx headroom to the 1336 * default value. 1337 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1338 * This function is used to set or query state related to XDP on the 1339 * netdevice and manage BPF offload. See definition of 1340 * enum bpf_netdev_command for details. 1341 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1342 * u32 flags); 1343 * This function is used to submit @n XDP packets for transmit on a 1344 * netdevice. Returns number of frames successfully transmitted, frames 1345 * that got dropped are freed/returned via xdp_return_frame(). 1346 * Returns negative number, means general error invoking ndo, meaning 1347 * no frames were xmit'ed and core-caller will free all frames. 1348 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1349 * struct xdp_buff *xdp); 1350 * Get the xmit slave of master device based on the xdp_buff. 1351 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1352 * This function is used to wake up the softirq, ksoftirqd or kthread 1353 * responsible for sending and/or receiving packets on a specific 1354 * queue id bound to an AF_XDP socket. The flags field specifies if 1355 * only RX, only Tx, or both should be woken up using the flags 1356 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1357 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm_kern *p, 1358 * int cmd); 1359 * Add, change, delete or get information on an IPv4 tunnel. 1360 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1361 * If a device is paired with a peer device, return the peer instance. 1362 * The caller must be under RCU read context. 1363 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1364 * Get the forwarding path to reach the real device from the HW destination address 1365 * ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1366 * const struct skb_shared_hwtstamps *hwtstamps, 1367 * bool cycles); 1368 * Get hardware timestamp based on normal/adjustable time or free running 1369 * cycle counter. This function is required if physical clock supports a 1370 * free running cycle counter. 1371 * 1372 * int (*ndo_hwtstamp_get)(struct net_device *dev, 1373 * struct kernel_hwtstamp_config *kernel_config); 1374 * Get the currently configured hardware timestamping parameters for the 1375 * NIC device. 1376 * 1377 * int (*ndo_hwtstamp_set)(struct net_device *dev, 1378 * struct kernel_hwtstamp_config *kernel_config, 1379 * struct netlink_ext_ack *extack); 1380 * Change the hardware timestamping parameters for NIC device. 1381 */ 1382 struct net_device_ops { 1383 int (*ndo_init)(struct net_device *dev); 1384 void (*ndo_uninit)(struct net_device *dev); 1385 int (*ndo_open)(struct net_device *dev); 1386 int (*ndo_stop)(struct net_device *dev); 1387 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1388 struct net_device *dev); 1389 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1390 struct net_device *dev, 1391 netdev_features_t features); 1392 u16 (*ndo_select_queue)(struct net_device *dev, 1393 struct sk_buff *skb, 1394 struct net_device *sb_dev); 1395 void (*ndo_change_rx_flags)(struct net_device *dev, 1396 int flags); 1397 void (*ndo_set_rx_mode)(struct net_device *dev); 1398 int (*ndo_set_mac_address)(struct net_device *dev, 1399 void *addr); 1400 int (*ndo_validate_addr)(struct net_device *dev); 1401 int (*ndo_do_ioctl)(struct net_device *dev, 1402 struct ifreq *ifr, int cmd); 1403 int (*ndo_eth_ioctl)(struct net_device *dev, 1404 struct ifreq *ifr, int cmd); 1405 int (*ndo_siocbond)(struct net_device *dev, 1406 struct ifreq *ifr, int cmd); 1407 int (*ndo_siocwandev)(struct net_device *dev, 1408 struct if_settings *ifs); 1409 int (*ndo_siocdevprivate)(struct net_device *dev, 1410 struct ifreq *ifr, 1411 void __user *data, int cmd); 1412 int (*ndo_set_config)(struct net_device *dev, 1413 struct ifmap *map); 1414 int (*ndo_change_mtu)(struct net_device *dev, 1415 int new_mtu); 1416 int (*ndo_neigh_setup)(struct net_device *dev, 1417 struct neigh_parms *); 1418 void (*ndo_tx_timeout) (struct net_device *dev, 1419 unsigned int txqueue); 1420 1421 void (*ndo_get_stats64)(struct net_device *dev, 1422 struct rtnl_link_stats64 *storage); 1423 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1424 int (*ndo_get_offload_stats)(int attr_id, 1425 const struct net_device *dev, 1426 void *attr_data); 1427 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1428 1429 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1430 __be16 proto, u16 vid); 1431 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1432 __be16 proto, u16 vid); 1433 #ifdef CONFIG_NET_POLL_CONTROLLER 1434 void (*ndo_poll_controller)(struct net_device *dev); 1435 int (*ndo_netpoll_setup)(struct net_device *dev); 1436 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1437 #endif 1438 int (*ndo_set_vf_mac)(struct net_device *dev, 1439 int queue, u8 *mac); 1440 int (*ndo_set_vf_vlan)(struct net_device *dev, 1441 int queue, u16 vlan, 1442 u8 qos, __be16 proto); 1443 int (*ndo_set_vf_rate)(struct net_device *dev, 1444 int vf, int min_tx_rate, 1445 int max_tx_rate); 1446 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1447 int vf, bool setting); 1448 int (*ndo_set_vf_trust)(struct net_device *dev, 1449 int vf, bool setting); 1450 int (*ndo_get_vf_config)(struct net_device *dev, 1451 int vf, 1452 struct ifla_vf_info *ivf); 1453 int (*ndo_set_vf_link_state)(struct net_device *dev, 1454 int vf, int link_state); 1455 int (*ndo_get_vf_stats)(struct net_device *dev, 1456 int vf, 1457 struct ifla_vf_stats 1458 *vf_stats); 1459 int (*ndo_set_vf_port)(struct net_device *dev, 1460 int vf, 1461 struct nlattr *port[]); 1462 int (*ndo_get_vf_port)(struct net_device *dev, 1463 int vf, struct sk_buff *skb); 1464 int (*ndo_get_vf_guid)(struct net_device *dev, 1465 int vf, 1466 struct ifla_vf_guid *node_guid, 1467 struct ifla_vf_guid *port_guid); 1468 int (*ndo_set_vf_guid)(struct net_device *dev, 1469 int vf, u64 guid, 1470 int guid_type); 1471 int (*ndo_set_vf_rss_query_en)( 1472 struct net_device *dev, 1473 int vf, bool setting); 1474 int (*ndo_setup_tc)(struct net_device *dev, 1475 enum tc_setup_type type, 1476 void *type_data); 1477 #if IS_ENABLED(CONFIG_FCOE) 1478 int (*ndo_fcoe_enable)(struct net_device *dev); 1479 int (*ndo_fcoe_disable)(struct net_device *dev); 1480 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1481 u16 xid, 1482 struct scatterlist *sgl, 1483 unsigned int sgc); 1484 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1485 u16 xid); 1486 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1487 u16 xid, 1488 struct scatterlist *sgl, 1489 unsigned int sgc); 1490 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1491 struct netdev_fcoe_hbainfo *hbainfo); 1492 #endif 1493 1494 #if IS_ENABLED(CONFIG_LIBFCOE) 1495 #define NETDEV_FCOE_WWNN 0 1496 #define NETDEV_FCOE_WWPN 1 1497 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1498 u64 *wwn, int type); 1499 #endif 1500 1501 #ifdef CONFIG_RFS_ACCEL 1502 int (*ndo_rx_flow_steer)(struct net_device *dev, 1503 const struct sk_buff *skb, 1504 u16 rxq_index, 1505 u32 flow_id); 1506 #endif 1507 int (*ndo_add_slave)(struct net_device *dev, 1508 struct net_device *slave_dev, 1509 struct netlink_ext_ack *extack); 1510 int (*ndo_del_slave)(struct net_device *dev, 1511 struct net_device *slave_dev); 1512 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1513 struct sk_buff *skb, 1514 bool all_slaves); 1515 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1516 struct sock *sk); 1517 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1518 netdev_features_t features); 1519 int (*ndo_set_features)(struct net_device *dev, 1520 netdev_features_t features); 1521 int (*ndo_neigh_construct)(struct net_device *dev, 1522 struct neighbour *n); 1523 void (*ndo_neigh_destroy)(struct net_device *dev, 1524 struct neighbour *n); 1525 1526 int (*ndo_fdb_add)(struct ndmsg *ndm, 1527 struct nlattr *tb[], 1528 struct net_device *dev, 1529 const unsigned char *addr, 1530 u16 vid, 1531 u16 flags, 1532 bool *notified, 1533 struct netlink_ext_ack *extack); 1534 int (*ndo_fdb_del)(struct ndmsg *ndm, 1535 struct nlattr *tb[], 1536 struct net_device *dev, 1537 const unsigned char *addr, 1538 u16 vid, 1539 bool *notified, 1540 struct netlink_ext_ack *extack); 1541 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, 1542 struct net_device *dev, 1543 struct netlink_ext_ack *extack); 1544 int (*ndo_fdb_dump)(struct sk_buff *skb, 1545 struct netlink_callback *cb, 1546 struct net_device *dev, 1547 struct net_device *filter_dev, 1548 int *idx); 1549 int (*ndo_fdb_get)(struct sk_buff *skb, 1550 struct nlattr *tb[], 1551 struct net_device *dev, 1552 const unsigned char *addr, 1553 u16 vid, u32 portid, u32 seq, 1554 struct netlink_ext_ack *extack); 1555 int (*ndo_mdb_add)(struct net_device *dev, 1556 struct nlattr *tb[], 1557 u16 nlmsg_flags, 1558 struct netlink_ext_ack *extack); 1559 int (*ndo_mdb_del)(struct net_device *dev, 1560 struct nlattr *tb[], 1561 struct netlink_ext_ack *extack); 1562 int (*ndo_mdb_del_bulk)(struct net_device *dev, 1563 struct nlattr *tb[], 1564 struct netlink_ext_ack *extack); 1565 int (*ndo_mdb_dump)(struct net_device *dev, 1566 struct sk_buff *skb, 1567 struct netlink_callback *cb); 1568 int (*ndo_mdb_get)(struct net_device *dev, 1569 struct nlattr *tb[], u32 portid, 1570 u32 seq, 1571 struct netlink_ext_ack *extack); 1572 int (*ndo_bridge_setlink)(struct net_device *dev, 1573 struct nlmsghdr *nlh, 1574 u16 flags, 1575 struct netlink_ext_ack *extack); 1576 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1577 u32 pid, u32 seq, 1578 struct net_device *dev, 1579 u32 filter_mask, 1580 int nlflags); 1581 int (*ndo_bridge_dellink)(struct net_device *dev, 1582 struct nlmsghdr *nlh, 1583 u16 flags); 1584 int (*ndo_change_carrier)(struct net_device *dev, 1585 bool new_carrier); 1586 int (*ndo_get_phys_port_id)(struct net_device *dev, 1587 struct netdev_phys_item_id *ppid); 1588 int (*ndo_get_port_parent_id)(struct net_device *dev, 1589 struct netdev_phys_item_id *ppid); 1590 int (*ndo_get_phys_port_name)(struct net_device *dev, 1591 char *name, size_t len); 1592 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1593 struct net_device *dev); 1594 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1595 void *priv); 1596 1597 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1598 int queue_index, 1599 u32 maxrate); 1600 int (*ndo_get_iflink)(const struct net_device *dev); 1601 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1602 struct sk_buff *skb); 1603 void (*ndo_set_rx_headroom)(struct net_device *dev, 1604 int needed_headroom); 1605 int (*ndo_bpf)(struct net_device *dev, 1606 struct netdev_bpf *bpf); 1607 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1608 struct xdp_frame **xdp, 1609 u32 flags); 1610 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1611 struct xdp_buff *xdp); 1612 int (*ndo_xsk_wakeup)(struct net_device *dev, 1613 u32 queue_id, u32 flags); 1614 int (*ndo_tunnel_ctl)(struct net_device *dev, 1615 struct ip_tunnel_parm_kern *p, 1616 int cmd); 1617 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1618 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1619 struct net_device_path *path); 1620 ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1621 const struct skb_shared_hwtstamps *hwtstamps, 1622 bool cycles); 1623 int (*ndo_hwtstamp_get)(struct net_device *dev, 1624 struct kernel_hwtstamp_config *kernel_config); 1625 int (*ndo_hwtstamp_set)(struct net_device *dev, 1626 struct kernel_hwtstamp_config *kernel_config, 1627 struct netlink_ext_ack *extack); 1628 1629 #if IS_ENABLED(CONFIG_NET_SHAPER) 1630 /** 1631 * @net_shaper_ops: Device shaping offload operations 1632 * see include/net/net_shapers.h 1633 */ 1634 const struct net_shaper_ops *net_shaper_ops; 1635 #endif 1636 }; 1637 1638 /** 1639 * enum netdev_priv_flags - &struct net_device priv_flags 1640 * 1641 * These are the &struct net_device, they are only set internally 1642 * by drivers and used in the kernel. These flags are invisible to 1643 * userspace; this means that the order of these flags can change 1644 * during any kernel release. 1645 * 1646 * You should add bitfield booleans after either net_device::priv_flags 1647 * (hotpath) or ::threaded (slowpath) instead of extending these flags. 1648 * 1649 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1650 * @IFF_EBRIDGE: Ethernet bridging device 1651 * @IFF_BONDING: bonding master or slave 1652 * @IFF_ISATAP: ISATAP interface (RFC4214) 1653 * @IFF_WAN_HDLC: WAN HDLC device 1654 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1655 * release skb->dst 1656 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1657 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1658 * @IFF_MACVLAN_PORT: device used as macvlan port 1659 * @IFF_BRIDGE_PORT: device used as bridge port 1660 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1661 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1662 * @IFF_UNICAST_FLT: Supports unicast filtering 1663 * @IFF_TEAM_PORT: device used as team port 1664 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1665 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1666 * change when it's running 1667 * @IFF_MACVLAN: Macvlan device 1668 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1669 * underlying stacked devices 1670 * @IFF_L3MDEV_MASTER: device is an L3 master device 1671 * @IFF_NO_QUEUE: device can run without qdisc attached 1672 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1673 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1674 * @IFF_TEAM: device is a team device 1675 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1676 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1677 * entity (i.e. the master device for bridged veth) 1678 * @IFF_MACSEC: device is a MACsec device 1679 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1680 * @IFF_FAILOVER: device is a failover master device 1681 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1682 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1683 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf 1684 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1685 * skb_headlen(skb) == 0 (data starts from frag0) 1686 */ 1687 enum netdev_priv_flags { 1688 IFF_802_1Q_VLAN = 1<<0, 1689 IFF_EBRIDGE = 1<<1, 1690 IFF_BONDING = 1<<2, 1691 IFF_ISATAP = 1<<3, 1692 IFF_WAN_HDLC = 1<<4, 1693 IFF_XMIT_DST_RELEASE = 1<<5, 1694 IFF_DONT_BRIDGE = 1<<6, 1695 IFF_DISABLE_NETPOLL = 1<<7, 1696 IFF_MACVLAN_PORT = 1<<8, 1697 IFF_BRIDGE_PORT = 1<<9, 1698 IFF_OVS_DATAPATH = 1<<10, 1699 IFF_TX_SKB_SHARING = 1<<11, 1700 IFF_UNICAST_FLT = 1<<12, 1701 IFF_TEAM_PORT = 1<<13, 1702 IFF_SUPP_NOFCS = 1<<14, 1703 IFF_LIVE_ADDR_CHANGE = 1<<15, 1704 IFF_MACVLAN = 1<<16, 1705 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1706 IFF_L3MDEV_MASTER = 1<<18, 1707 IFF_NO_QUEUE = 1<<19, 1708 IFF_OPENVSWITCH = 1<<20, 1709 IFF_L3MDEV_SLAVE = 1<<21, 1710 IFF_TEAM = 1<<22, 1711 IFF_RXFH_CONFIGURED = 1<<23, 1712 IFF_PHONY_HEADROOM = 1<<24, 1713 IFF_MACSEC = 1<<25, 1714 IFF_NO_RX_HANDLER = 1<<26, 1715 IFF_FAILOVER = 1<<27, 1716 IFF_FAILOVER_SLAVE = 1<<28, 1717 IFF_L3MDEV_RX_HANDLER = 1<<29, 1718 IFF_NO_ADDRCONF = BIT_ULL(30), 1719 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31), 1720 }; 1721 1722 /* Specifies the type of the struct net_device::ml_priv pointer */ 1723 enum netdev_ml_priv_type { 1724 ML_PRIV_NONE, 1725 ML_PRIV_CAN, 1726 }; 1727 1728 enum netdev_stat_type { 1729 NETDEV_PCPU_STAT_NONE, 1730 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */ 1731 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */ 1732 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */ 1733 }; 1734 1735 enum netdev_reg_state { 1736 NETREG_UNINITIALIZED = 0, 1737 NETREG_REGISTERED, /* completed register_netdevice */ 1738 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1739 NETREG_UNREGISTERED, /* completed unregister todo */ 1740 NETREG_RELEASED, /* called free_netdev */ 1741 NETREG_DUMMY, /* dummy device for NAPI poll */ 1742 }; 1743 1744 /** 1745 * struct net_device - The DEVICE structure. 1746 * 1747 * Actually, this whole structure is a big mistake. It mixes I/O 1748 * data with strictly "high-level" data, and it has to know about 1749 * almost every data structure used in the INET module. 1750 * 1751 * @priv_flags: flags invisible to userspace defined as bits, see 1752 * enum netdev_priv_flags for the definitions 1753 * @lltx: device supports lockless Tx. Deprecated for real HW 1754 * drivers. Mainly used by logical interfaces, such as 1755 * bonding and tunnels 1756 * 1757 * @name: This is the first field of the "visible" part of this structure 1758 * (i.e. as seen by users in the "Space.c" file). It is the name 1759 * of the interface. 1760 * 1761 * @name_node: Name hashlist node 1762 * @ifalias: SNMP alias 1763 * @mem_end: Shared memory end 1764 * @mem_start: Shared memory start 1765 * @base_addr: Device I/O address 1766 * @irq: Device IRQ number 1767 * 1768 * @state: Generic network queuing layer state, see netdev_state_t 1769 * @dev_list: The global list of network devices 1770 * @napi_list: List entry used for polling NAPI devices 1771 * @unreg_list: List entry when we are unregistering the 1772 * device; see the function unregister_netdev 1773 * @close_list: List entry used when we are closing the device 1774 * @ptype_all: Device-specific packet handlers for all protocols 1775 * @ptype_specific: Device-specific, protocol-specific packet handlers 1776 * 1777 * @adj_list: Directly linked devices, like slaves for bonding 1778 * @features: Currently active device features 1779 * @hw_features: User-changeable features 1780 * 1781 * @wanted_features: User-requested features 1782 * @vlan_features: Mask of features inheritable by VLAN devices 1783 * 1784 * @hw_enc_features: Mask of features inherited by encapsulating devices 1785 * This field indicates what encapsulation 1786 * offloads the hardware is capable of doing, 1787 * and drivers will need to set them appropriately. 1788 * 1789 * @mpls_features: Mask of features inheritable by MPLS 1790 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1791 * 1792 * @ifindex: interface index 1793 * @group: The group the device belongs to 1794 * 1795 * @stats: Statistics struct, which was left as a legacy, use 1796 * rtnl_link_stats64 instead 1797 * 1798 * @core_stats: core networking counters, 1799 * do not use this in drivers 1800 * @carrier_up_count: Number of times the carrier has been up 1801 * @carrier_down_count: Number of times the carrier has been down 1802 * 1803 * @wireless_handlers: List of functions to handle Wireless Extensions, 1804 * instead of ioctl, 1805 * see <net/iw_handler.h> for details. 1806 * 1807 * @netdev_ops: Includes several pointers to callbacks, 1808 * if one wants to override the ndo_*() functions 1809 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks. 1810 * @xsk_tx_metadata_ops: Includes pointers to AF_XDP TX metadata callbacks. 1811 * @ethtool_ops: Management operations 1812 * @l3mdev_ops: Layer 3 master device operations 1813 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1814 * discovery handling. Necessary for e.g. 6LoWPAN. 1815 * @xfrmdev_ops: Transformation offload operations 1816 * @tlsdev_ops: Transport Layer Security offload operations 1817 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1818 * of Layer 2 headers. 1819 * 1820 * @flags: Interface flags (a la BSD) 1821 * @xdp_features: XDP capability supported by the device 1822 * @gflags: Global flags ( kept as legacy ) 1823 * @priv_len: Size of the ->priv flexible array 1824 * @priv: Flexible array containing private data 1825 * @operstate: RFC2863 operstate 1826 * @link_mode: Mapping policy to operstate 1827 * @if_port: Selectable AUI, TP, ... 1828 * @dma: DMA channel 1829 * @mtu: Interface MTU value 1830 * @min_mtu: Interface Minimum MTU value 1831 * @max_mtu: Interface Maximum MTU value 1832 * @type: Interface hardware type 1833 * @hard_header_len: Maximum hardware header length. 1834 * @min_header_len: Minimum hardware header length 1835 * 1836 * @needed_headroom: Extra headroom the hardware may need, but not in all 1837 * cases can this be guaranteed 1838 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1839 * cases can this be guaranteed. Some cases also use 1840 * LL_MAX_HEADER instead to allocate the skb 1841 * 1842 * interface address info: 1843 * 1844 * @perm_addr: Permanent hw address 1845 * @addr_assign_type: Hw address assignment type 1846 * @addr_len: Hardware address length 1847 * @upper_level: Maximum depth level of upper devices. 1848 * @lower_level: Maximum depth level of lower devices. 1849 * @neigh_priv_len: Used in neigh_alloc() 1850 * @dev_id: Used to differentiate devices that share 1851 * the same link layer address 1852 * @dev_port: Used to differentiate devices that share 1853 * the same function 1854 * @addr_list_lock: XXX: need comments on this one 1855 * @name_assign_type: network interface name assignment type 1856 * @uc_promisc: Counter that indicates promiscuous mode 1857 * has been enabled due to the need to listen to 1858 * additional unicast addresses in a device that 1859 * does not implement ndo_set_rx_mode() 1860 * @uc: unicast mac addresses 1861 * @mc: multicast mac addresses 1862 * @dev_addrs: list of device hw addresses 1863 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1864 * @promiscuity: Number of times the NIC is told to work in 1865 * promiscuous mode; if it becomes 0 the NIC will 1866 * exit promiscuous mode 1867 * @allmulti: Counter, enables or disables allmulticast mode 1868 * 1869 * @vlan_info: VLAN info 1870 * @dsa_ptr: dsa specific data 1871 * @tipc_ptr: TIPC specific data 1872 * @atalk_ptr: AppleTalk link 1873 * @ip_ptr: IPv4 specific data 1874 * @ip6_ptr: IPv6 specific data 1875 * @ax25_ptr: AX.25 specific data 1876 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1877 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1878 * device struct 1879 * @mpls_ptr: mpls_dev struct pointer 1880 * @mctp_ptr: MCTP specific data 1881 * 1882 * @dev_addr: Hw address (before bcast, 1883 * because most packets are unicast) 1884 * 1885 * @_rx: Array of RX queues 1886 * @num_rx_queues: Number of RX queues 1887 * allocated at register_netdev() time 1888 * @real_num_rx_queues: Number of RX queues currently active in device 1889 * @xdp_prog: XDP sockets filter program pointer 1890 * 1891 * @rx_handler: handler for received packets 1892 * @rx_handler_data: XXX: need comments on this one 1893 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing 1894 * @ingress_queue: XXX: need comments on this one 1895 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1896 * @broadcast: hw bcast address 1897 * 1898 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1899 * indexed by RX queue number. Assigned by driver. 1900 * This must only be set if the ndo_rx_flow_steer 1901 * operation is defined 1902 * @index_hlist: Device index hash chain 1903 * 1904 * @_tx: Array of TX queues 1905 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1906 * @real_num_tx_queues: Number of TX queues currently active in device 1907 * @qdisc: Root qdisc from userspace point of view 1908 * @tx_queue_len: Max frames per queue allowed 1909 * @tx_global_lock: XXX: need comments on this one 1910 * @xdp_bulkq: XDP device bulk queue 1911 * @xps_maps: all CPUs/RXQs maps for XPS device 1912 * 1913 * @xps_maps: XXX: need comments on this one 1914 * @tcx_egress: BPF & clsact qdisc specific data for egress processing 1915 * @nf_hooks_egress: netfilter hooks executed for egress packets 1916 * @qdisc_hash: qdisc hash table 1917 * @watchdog_timeo: Represents the timeout that is used by 1918 * the watchdog (see dev_watchdog()) 1919 * @watchdog_timer: List of timers 1920 * 1921 * @proto_down_reason: reason a netdev interface is held down 1922 * @pcpu_refcnt: Number of references to this device 1923 * @dev_refcnt: Number of references to this device 1924 * @refcnt_tracker: Tracker directory for tracked references to this device 1925 * @todo_list: Delayed register/unregister 1926 * @link_watch_list: XXX: need comments on this one 1927 * 1928 * @reg_state: Register/unregister state machine 1929 * @dismantle: Device is going to be freed 1930 * @rtnl_link_state: This enum represents the phases of creating 1931 * a new link 1932 * 1933 * @needs_free_netdev: Should unregister perform free_netdev? 1934 * @priv_destructor: Called from unregister 1935 * @npinfo: XXX: need comments on this one 1936 * @nd_net: Network namespace this network device is inside 1937 * 1938 * @ml_priv: Mid-layer private 1939 * @ml_priv_type: Mid-layer private type 1940 * 1941 * @pcpu_stat_type: Type of device statistics which the core should 1942 * allocate/free: none, lstats, tstats, dstats. none 1943 * means the driver is handling statistics allocation/ 1944 * freeing internally. 1945 * @lstats: Loopback statistics: packets, bytes 1946 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes 1947 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes 1948 * 1949 * @garp_port: GARP 1950 * @mrp_port: MRP 1951 * 1952 * @dm_private: Drop monitor private 1953 * 1954 * @dev: Class/net/name entry 1955 * @sysfs_groups: Space for optional device, statistics and wireless 1956 * sysfs groups 1957 * 1958 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1959 * @rtnl_link_ops: Rtnl_link_ops 1960 * @stat_ops: Optional ops for queue-aware statistics 1961 * @queue_mgmt_ops: Optional ops for queue management 1962 * 1963 * @gso_max_size: Maximum size of generic segmentation offload 1964 * @tso_max_size: Device (as in HW) limit on the max TSO request size 1965 * @gso_max_segs: Maximum number of segments that can be passed to the 1966 * NIC for GSO 1967 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count 1968 * @gso_ipv4_max_size: Maximum size of generic segmentation offload, 1969 * for IPv4. 1970 * 1971 * @dcbnl_ops: Data Center Bridging netlink ops 1972 * @num_tc: Number of traffic classes in the net device 1973 * @tc_to_txq: XXX: need comments on this one 1974 * @prio_tc_map: XXX: need comments on this one 1975 * 1976 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1977 * 1978 * @priomap: XXX: need comments on this one 1979 * @link_topo: Physical link topology tracking attached PHYs 1980 * @phydev: Physical device may attach itself 1981 * for hardware timestamping 1982 * @sfp_bus: attached &struct sfp_bus structure. 1983 * 1984 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1985 * 1986 * @proto_down: protocol port state information can be sent to the 1987 * switch driver and used to set the phys state of the 1988 * switch port. 1989 * 1990 * @threaded: napi threaded mode is enabled 1991 * 1992 * @see_all_hwtstamp_requests: device wants to see calls to 1993 * ndo_hwtstamp_set() for all timestamp requests 1994 * regardless of source, even if those aren't 1995 * HWTSTAMP_SOURCE_NETDEV 1996 * @change_proto_down: device supports setting carrier via IFLA_PROTO_DOWN 1997 * @netns_local: interface can't change network namespaces 1998 * @fcoe_mtu: device supports maximum FCoE MTU, 2158 bytes 1999 * 2000 * @net_notifier_list: List of per-net netdev notifier block 2001 * that follow this device when it is moved 2002 * to another network namespace. 2003 * 2004 * @macsec_ops: MACsec offloading ops 2005 * 2006 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 2007 * offload capabilities of the device 2008 * @udp_tunnel_nic: UDP tunnel offload state 2009 * @ethtool: ethtool related state 2010 * @xdp_state: stores info on attached XDP BPF programs 2011 * 2012 * @nested_level: Used as a parameter of spin_lock_nested() of 2013 * dev->addr_list_lock. 2014 * @unlink_list: As netif_addr_lock() can be called recursively, 2015 * keep a list of interfaces to be deleted. 2016 * @gro_max_size: Maximum size of aggregated packet in generic 2017 * receive offload (GRO) 2018 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic 2019 * receive offload (GRO), for IPv4. 2020 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP 2021 * zero copy driver 2022 * 2023 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 2024 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 2025 * @watchdog_dev_tracker: refcount tracker used by watchdog. 2026 * @dev_registered_tracker: tracker for reference held while 2027 * registered 2028 * @offload_xstats_l3: L3 HW stats for this netdevice. 2029 * 2030 * @devlink_port: Pointer to related devlink port structure. 2031 * Assigned by a driver before netdev registration using 2032 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static 2033 * during the time netdevice is registered. 2034 * 2035 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem, 2036 * where the clock is recovered. 2037 * 2038 * @max_pacing_offload_horizon: max EDT offload horizon in nsec. 2039 * @napi_config: An array of napi_config structures containing per-NAPI 2040 * settings. 2041 * @gro_flush_timeout: timeout for GRO layer in NAPI 2042 * @napi_defer_hard_irqs: If not zero, provides a counter that would 2043 * allow to avoid NIC hard IRQ, on busy queues. 2044 * 2045 * @neighbours: List heads pointing to this device's neighbours' 2046 * dev_list, one per address-family. 2047 * @hwprov: Tracks which PTP performs hardware packet time stamping. 2048 * 2049 * FIXME: cleanup struct net_device such that network protocol info 2050 * moves out. 2051 */ 2052 2053 struct net_device { 2054 /* Cacheline organization can be found documented in 2055 * Documentation/networking/net_cachelines/net_device.rst. 2056 * Please update the document when adding new fields. 2057 */ 2058 2059 /* TX read-mostly hotpath */ 2060 __cacheline_group_begin(net_device_read_tx); 2061 struct_group(priv_flags_fast, 2062 unsigned long priv_flags:32; 2063 unsigned long lltx:1; 2064 ); 2065 const struct net_device_ops *netdev_ops; 2066 const struct header_ops *header_ops; 2067 struct netdev_queue *_tx; 2068 netdev_features_t gso_partial_features; 2069 unsigned int real_num_tx_queues; 2070 unsigned int gso_max_size; 2071 unsigned int gso_ipv4_max_size; 2072 u16 gso_max_segs; 2073 s16 num_tc; 2074 /* Note : dev->mtu is often read without holding a lock. 2075 * Writers usually hold RTNL. 2076 * It is recommended to use READ_ONCE() to annotate the reads, 2077 * and to use WRITE_ONCE() to annotate the writes. 2078 */ 2079 unsigned int mtu; 2080 unsigned short needed_headroom; 2081 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2082 #ifdef CONFIG_XPS 2083 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2084 #endif 2085 #ifdef CONFIG_NETFILTER_EGRESS 2086 struct nf_hook_entries __rcu *nf_hooks_egress; 2087 #endif 2088 #ifdef CONFIG_NET_XGRESS 2089 struct bpf_mprog_entry __rcu *tcx_egress; 2090 #endif 2091 __cacheline_group_end(net_device_read_tx); 2092 2093 /* TXRX read-mostly hotpath */ 2094 __cacheline_group_begin(net_device_read_txrx); 2095 union { 2096 struct pcpu_lstats __percpu *lstats; 2097 struct pcpu_sw_netstats __percpu *tstats; 2098 struct pcpu_dstats __percpu *dstats; 2099 }; 2100 unsigned long state; 2101 unsigned int flags; 2102 unsigned short hard_header_len; 2103 netdev_features_t features; 2104 struct inet6_dev __rcu *ip6_ptr; 2105 __cacheline_group_end(net_device_read_txrx); 2106 2107 /* RX read-mostly hotpath */ 2108 __cacheline_group_begin(net_device_read_rx); 2109 struct bpf_prog __rcu *xdp_prog; 2110 struct list_head ptype_specific; 2111 int ifindex; 2112 unsigned int real_num_rx_queues; 2113 struct netdev_rx_queue *_rx; 2114 unsigned int gro_max_size; 2115 unsigned int gro_ipv4_max_size; 2116 rx_handler_func_t __rcu *rx_handler; 2117 void __rcu *rx_handler_data; 2118 possible_net_t nd_net; 2119 #ifdef CONFIG_NETPOLL 2120 struct netpoll_info __rcu *npinfo; 2121 #endif 2122 #ifdef CONFIG_NET_XGRESS 2123 struct bpf_mprog_entry __rcu *tcx_ingress; 2124 #endif 2125 __cacheline_group_end(net_device_read_rx); 2126 2127 char name[IFNAMSIZ]; 2128 struct netdev_name_node *name_node; 2129 struct dev_ifalias __rcu *ifalias; 2130 /* 2131 * I/O specific fields 2132 * FIXME: Merge these and struct ifmap into one 2133 */ 2134 unsigned long mem_end; 2135 unsigned long mem_start; 2136 unsigned long base_addr; 2137 2138 /* 2139 * Some hardware also needs these fields (state,dev_list, 2140 * napi_list,unreg_list,close_list) but they are not 2141 * part of the usual set specified in Space.c. 2142 */ 2143 2144 2145 struct list_head dev_list; 2146 struct list_head napi_list; 2147 struct list_head unreg_list; 2148 struct list_head close_list; 2149 struct list_head ptype_all; 2150 2151 struct { 2152 struct list_head upper; 2153 struct list_head lower; 2154 } adj_list; 2155 2156 /* Read-mostly cache-line for fast-path access */ 2157 xdp_features_t xdp_features; 2158 const struct xdp_metadata_ops *xdp_metadata_ops; 2159 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops; 2160 unsigned short gflags; 2161 2162 unsigned short needed_tailroom; 2163 2164 netdev_features_t hw_features; 2165 netdev_features_t wanted_features; 2166 netdev_features_t vlan_features; 2167 netdev_features_t hw_enc_features; 2168 netdev_features_t mpls_features; 2169 2170 unsigned int min_mtu; 2171 unsigned int max_mtu; 2172 unsigned short type; 2173 unsigned char min_header_len; 2174 unsigned char name_assign_type; 2175 2176 int group; 2177 2178 struct net_device_stats stats; /* not used by modern drivers */ 2179 2180 struct net_device_core_stats __percpu *core_stats; 2181 2182 /* Stats to monitor link on/off, flapping */ 2183 atomic_t carrier_up_count; 2184 atomic_t carrier_down_count; 2185 2186 #ifdef CONFIG_WIRELESS_EXT 2187 const struct iw_handler_def *wireless_handlers; 2188 #endif 2189 const struct ethtool_ops *ethtool_ops; 2190 #ifdef CONFIG_NET_L3_MASTER_DEV 2191 const struct l3mdev_ops *l3mdev_ops; 2192 #endif 2193 #if IS_ENABLED(CONFIG_IPV6) 2194 const struct ndisc_ops *ndisc_ops; 2195 #endif 2196 2197 #ifdef CONFIG_XFRM_OFFLOAD 2198 const struct xfrmdev_ops *xfrmdev_ops; 2199 #endif 2200 2201 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2202 const struct tlsdev_ops *tlsdev_ops; 2203 #endif 2204 2205 unsigned int operstate; 2206 unsigned char link_mode; 2207 2208 unsigned char if_port; 2209 unsigned char dma; 2210 2211 /* Interface address info. */ 2212 unsigned char perm_addr[MAX_ADDR_LEN]; 2213 unsigned char addr_assign_type; 2214 unsigned char addr_len; 2215 unsigned char upper_level; 2216 unsigned char lower_level; 2217 2218 unsigned short neigh_priv_len; 2219 unsigned short dev_id; 2220 unsigned short dev_port; 2221 int irq; 2222 u32 priv_len; 2223 2224 spinlock_t addr_list_lock; 2225 2226 struct netdev_hw_addr_list uc; 2227 struct netdev_hw_addr_list mc; 2228 struct netdev_hw_addr_list dev_addrs; 2229 2230 #ifdef CONFIG_SYSFS 2231 struct kset *queues_kset; 2232 #endif 2233 #ifdef CONFIG_LOCKDEP 2234 struct list_head unlink_list; 2235 #endif 2236 unsigned int promiscuity; 2237 unsigned int allmulti; 2238 bool uc_promisc; 2239 #ifdef CONFIG_LOCKDEP 2240 unsigned char nested_level; 2241 #endif 2242 2243 2244 /* Protocol-specific pointers */ 2245 struct in_device __rcu *ip_ptr; 2246 /** @fib_nh_head: nexthops associated with this netdev */ 2247 struct hlist_head fib_nh_head; 2248 2249 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2250 struct vlan_info __rcu *vlan_info; 2251 #endif 2252 #if IS_ENABLED(CONFIG_NET_DSA) 2253 struct dsa_port *dsa_ptr; 2254 #endif 2255 #if IS_ENABLED(CONFIG_TIPC) 2256 struct tipc_bearer __rcu *tipc_ptr; 2257 #endif 2258 #if IS_ENABLED(CONFIG_ATALK) 2259 void *atalk_ptr; 2260 #endif 2261 #if IS_ENABLED(CONFIG_AX25) 2262 struct ax25_dev __rcu *ax25_ptr; 2263 #endif 2264 #if IS_ENABLED(CONFIG_CFG80211) 2265 struct wireless_dev *ieee80211_ptr; 2266 #endif 2267 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2268 struct wpan_dev *ieee802154_ptr; 2269 #endif 2270 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2271 struct mpls_dev __rcu *mpls_ptr; 2272 #endif 2273 #if IS_ENABLED(CONFIG_MCTP) 2274 struct mctp_dev __rcu *mctp_ptr; 2275 #endif 2276 2277 /* 2278 * Cache lines mostly used on receive path (including eth_type_trans()) 2279 */ 2280 /* Interface address info used in eth_type_trans() */ 2281 const unsigned char *dev_addr; 2282 2283 unsigned int num_rx_queues; 2284 #define GRO_LEGACY_MAX_SIZE 65536u 2285 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2286 * and shinfo->gso_segs is a 16bit field. 2287 */ 2288 #define GRO_MAX_SIZE (8 * 65535u) 2289 unsigned int xdp_zc_max_segs; 2290 struct netdev_queue __rcu *ingress_queue; 2291 #ifdef CONFIG_NETFILTER_INGRESS 2292 struct nf_hook_entries __rcu *nf_hooks_ingress; 2293 #endif 2294 2295 unsigned char broadcast[MAX_ADDR_LEN]; 2296 #ifdef CONFIG_RFS_ACCEL 2297 struct cpu_rmap *rx_cpu_rmap; 2298 #endif 2299 struct hlist_node index_hlist; 2300 2301 /* 2302 * Cache lines mostly used on transmit path 2303 */ 2304 unsigned int num_tx_queues; 2305 struct Qdisc __rcu *qdisc; 2306 unsigned int tx_queue_len; 2307 spinlock_t tx_global_lock; 2308 2309 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2310 2311 #ifdef CONFIG_NET_SCHED 2312 DECLARE_HASHTABLE (qdisc_hash, 4); 2313 #endif 2314 /* These may be needed for future network-power-down code. */ 2315 struct timer_list watchdog_timer; 2316 int watchdog_timeo; 2317 2318 u32 proto_down_reason; 2319 2320 struct list_head todo_list; 2321 2322 #ifdef CONFIG_PCPU_DEV_REFCNT 2323 int __percpu *pcpu_refcnt; 2324 #else 2325 refcount_t dev_refcnt; 2326 #endif 2327 struct ref_tracker_dir refcnt_tracker; 2328 2329 struct list_head link_watch_list; 2330 2331 u8 reg_state; 2332 2333 bool dismantle; 2334 2335 enum { 2336 RTNL_LINK_INITIALIZED, 2337 RTNL_LINK_INITIALIZING, 2338 } rtnl_link_state:16; 2339 2340 bool needs_free_netdev; 2341 void (*priv_destructor)(struct net_device *dev); 2342 2343 /* mid-layer private */ 2344 void *ml_priv; 2345 enum netdev_ml_priv_type ml_priv_type; 2346 2347 enum netdev_stat_type pcpu_stat_type:8; 2348 2349 #if IS_ENABLED(CONFIG_GARP) 2350 struct garp_port __rcu *garp_port; 2351 #endif 2352 #if IS_ENABLED(CONFIG_MRP) 2353 struct mrp_port __rcu *mrp_port; 2354 #endif 2355 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2356 struct dm_hw_stat_delta __rcu *dm_private; 2357 #endif 2358 struct device dev; 2359 const struct attribute_group *sysfs_groups[4]; 2360 const struct attribute_group *sysfs_rx_queue_group; 2361 2362 const struct rtnl_link_ops *rtnl_link_ops; 2363 2364 const struct netdev_stat_ops *stat_ops; 2365 2366 const struct netdev_queue_mgmt_ops *queue_mgmt_ops; 2367 2368 /* for setting kernel sock attribute on TCP connection setup */ 2369 #define GSO_MAX_SEGS 65535u 2370 #define GSO_LEGACY_MAX_SIZE 65536u 2371 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2372 * and shinfo->gso_segs is a 16bit field. 2373 */ 2374 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2375 2376 #define TSO_LEGACY_MAX_SIZE 65536 2377 #define TSO_MAX_SIZE UINT_MAX 2378 unsigned int tso_max_size; 2379 #define TSO_MAX_SEGS U16_MAX 2380 u16 tso_max_segs; 2381 2382 #ifdef CONFIG_DCB 2383 const struct dcbnl_rtnl_ops *dcbnl_ops; 2384 #endif 2385 u8 prio_tc_map[TC_BITMASK + 1]; 2386 2387 #if IS_ENABLED(CONFIG_FCOE) 2388 unsigned int fcoe_ddp_xid; 2389 #endif 2390 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2391 struct netprio_map __rcu *priomap; 2392 #endif 2393 struct phy_link_topology *link_topo; 2394 struct phy_device *phydev; 2395 struct sfp_bus *sfp_bus; 2396 struct lock_class_key *qdisc_tx_busylock; 2397 bool proto_down; 2398 bool threaded; 2399 2400 /* priv_flags_slow, ungrouped to save space */ 2401 unsigned long see_all_hwtstamp_requests:1; 2402 unsigned long change_proto_down:1; 2403 unsigned long netns_local:1; 2404 unsigned long fcoe_mtu:1; 2405 2406 struct list_head net_notifier_list; 2407 2408 #if IS_ENABLED(CONFIG_MACSEC) 2409 /* MACsec management functions */ 2410 const struct macsec_ops *macsec_ops; 2411 #endif 2412 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2413 struct udp_tunnel_nic *udp_tunnel_nic; 2414 2415 /** @cfg: net_device queue-related configuration */ 2416 struct netdev_config *cfg; 2417 /** 2418 * @cfg_pending: same as @cfg but when device is being actively 2419 * reconfigured includes any changes to the configuration 2420 * requested by the user, but which may or may not be rejected. 2421 */ 2422 struct netdev_config *cfg_pending; 2423 struct ethtool_netdev_state *ethtool; 2424 2425 /* protected by rtnl_lock */ 2426 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2427 2428 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2429 netdevice_tracker linkwatch_dev_tracker; 2430 netdevice_tracker watchdog_dev_tracker; 2431 netdevice_tracker dev_registered_tracker; 2432 struct rtnl_hw_stats64 *offload_xstats_l3; 2433 2434 struct devlink_port *devlink_port; 2435 2436 #if IS_ENABLED(CONFIG_DPLL) 2437 struct dpll_pin __rcu *dpll_pin; 2438 #endif 2439 #if IS_ENABLED(CONFIG_PAGE_POOL) 2440 /** @page_pools: page pools created for this netdevice */ 2441 struct hlist_head page_pools; 2442 #endif 2443 2444 /** @irq_moder: dim parameters used if IS_ENABLED(CONFIG_DIMLIB). */ 2445 struct dim_irq_moder *irq_moder; 2446 2447 u64 max_pacing_offload_horizon; 2448 struct napi_config *napi_config; 2449 unsigned long gro_flush_timeout; 2450 u32 napi_defer_hard_irqs; 2451 2452 /** 2453 * @up: copy of @state's IFF_UP, but safe to read with just @lock. 2454 * May report false negatives while the device is being opened 2455 * or closed (@lock does not protect .ndo_open, or .ndo_close). 2456 */ 2457 bool up; 2458 2459 /** 2460 * @lock: netdev-scope lock, protects a small selection of fields. 2461 * Should always be taken using netdev_lock() / netdev_unlock() helpers. 2462 * Drivers are free to use it for other protection. 2463 * 2464 * Protects: 2465 * @gro_flush_timeout, @napi_defer_hard_irqs, @napi_list, 2466 * @net_shaper_hierarchy, @reg_state, @threaded 2467 * 2468 * Partially protects (writers must hold both @lock and rtnl_lock): 2469 * @up 2470 * 2471 * Also protects some fields in struct napi_struct. 2472 * 2473 * Ordering: take after rtnl_lock. 2474 */ 2475 struct mutex lock; 2476 2477 #if IS_ENABLED(CONFIG_NET_SHAPER) 2478 /** 2479 * @net_shaper_hierarchy: data tracking the current shaper status 2480 * see include/net/net_shapers.h 2481 */ 2482 struct net_shaper_hierarchy *net_shaper_hierarchy; 2483 #endif 2484 2485 struct hlist_head neighbours[NEIGH_NR_TABLES]; 2486 2487 struct hwtstamp_provider __rcu *hwprov; 2488 2489 u8 priv[] ____cacheline_aligned 2490 __counted_by(priv_len); 2491 } ____cacheline_aligned; 2492 #define to_net_dev(d) container_of(d, struct net_device, dev) 2493 2494 /* 2495 * Driver should use this to assign devlink port instance to a netdevice 2496 * before it registers the netdevice. Therefore devlink_port is static 2497 * during the netdev lifetime after it is registered. 2498 */ 2499 #define SET_NETDEV_DEVLINK_PORT(dev, port) \ 2500 ({ \ 2501 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \ 2502 ((dev)->devlink_port = (port)); \ 2503 }) 2504 2505 static inline bool netif_elide_gro(const struct net_device *dev) 2506 { 2507 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2508 return true; 2509 return false; 2510 } 2511 2512 #define NETDEV_ALIGN 32 2513 2514 static inline 2515 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2516 { 2517 return dev->prio_tc_map[prio & TC_BITMASK]; 2518 } 2519 2520 static inline 2521 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2522 { 2523 if (tc >= dev->num_tc) 2524 return -EINVAL; 2525 2526 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2527 return 0; 2528 } 2529 2530 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2531 void netdev_reset_tc(struct net_device *dev); 2532 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2533 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2534 2535 static inline 2536 int netdev_get_num_tc(struct net_device *dev) 2537 { 2538 return dev->num_tc; 2539 } 2540 2541 static inline void net_prefetch(void *p) 2542 { 2543 prefetch(p); 2544 #if L1_CACHE_BYTES < 128 2545 prefetch((u8 *)p + L1_CACHE_BYTES); 2546 #endif 2547 } 2548 2549 static inline void net_prefetchw(void *p) 2550 { 2551 prefetchw(p); 2552 #if L1_CACHE_BYTES < 128 2553 prefetchw((u8 *)p + L1_CACHE_BYTES); 2554 #endif 2555 } 2556 2557 void netdev_unbind_sb_channel(struct net_device *dev, 2558 struct net_device *sb_dev); 2559 int netdev_bind_sb_channel_queue(struct net_device *dev, 2560 struct net_device *sb_dev, 2561 u8 tc, u16 count, u16 offset); 2562 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2563 static inline int netdev_get_sb_channel(struct net_device *dev) 2564 { 2565 return max_t(int, -dev->num_tc, 0); 2566 } 2567 2568 static inline 2569 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2570 unsigned int index) 2571 { 2572 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues); 2573 return &dev->_tx[index]; 2574 } 2575 2576 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2577 const struct sk_buff *skb) 2578 { 2579 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2580 } 2581 2582 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2583 void (*f)(struct net_device *, 2584 struct netdev_queue *, 2585 void *), 2586 void *arg) 2587 { 2588 unsigned int i; 2589 2590 for (i = 0; i < dev->num_tx_queues; i++) 2591 f(dev, &dev->_tx[i], arg); 2592 } 2593 2594 #define netdev_lockdep_set_classes(dev) \ 2595 { \ 2596 static struct lock_class_key qdisc_tx_busylock_key; \ 2597 static struct lock_class_key qdisc_xmit_lock_key; \ 2598 static struct lock_class_key dev_addr_list_lock_key; \ 2599 unsigned int i; \ 2600 \ 2601 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2602 lockdep_set_class(&(dev)->addr_list_lock, \ 2603 &dev_addr_list_lock_key); \ 2604 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2605 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2606 &qdisc_xmit_lock_key); \ 2607 } 2608 2609 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2610 struct net_device *sb_dev); 2611 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2612 struct sk_buff *skb, 2613 struct net_device *sb_dev); 2614 2615 /* returns the headroom that the master device needs to take in account 2616 * when forwarding to this dev 2617 */ 2618 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2619 { 2620 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2621 } 2622 2623 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2624 { 2625 if (dev->netdev_ops->ndo_set_rx_headroom) 2626 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2627 } 2628 2629 /* set the device rx headroom to the dev's default */ 2630 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2631 { 2632 netdev_set_rx_headroom(dev, -1); 2633 } 2634 2635 static inline void *netdev_get_ml_priv(struct net_device *dev, 2636 enum netdev_ml_priv_type type) 2637 { 2638 if (dev->ml_priv_type != type) 2639 return NULL; 2640 2641 return dev->ml_priv; 2642 } 2643 2644 static inline void netdev_set_ml_priv(struct net_device *dev, 2645 void *ml_priv, 2646 enum netdev_ml_priv_type type) 2647 { 2648 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2649 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2650 dev->ml_priv_type, type); 2651 WARN(!dev->ml_priv_type && dev->ml_priv, 2652 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2653 2654 dev->ml_priv = ml_priv; 2655 dev->ml_priv_type = type; 2656 } 2657 2658 /* 2659 * Net namespace inlines 2660 */ 2661 static inline 2662 struct net *dev_net(const struct net_device *dev) 2663 { 2664 return read_pnet(&dev->nd_net); 2665 } 2666 2667 static inline 2668 struct net *dev_net_rcu(const struct net_device *dev) 2669 { 2670 return read_pnet_rcu(&dev->nd_net); 2671 } 2672 2673 static inline 2674 void dev_net_set(struct net_device *dev, struct net *net) 2675 { 2676 write_pnet(&dev->nd_net, net); 2677 } 2678 2679 /** 2680 * netdev_priv - access network device private data 2681 * @dev: network device 2682 * 2683 * Get network device private data 2684 */ 2685 static inline void *netdev_priv(const struct net_device *dev) 2686 { 2687 return (void *)dev->priv; 2688 } 2689 2690 /* Set the sysfs physical device reference for the network logical device 2691 * if set prior to registration will cause a symlink during initialization. 2692 */ 2693 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2694 2695 /* Set the sysfs device type for the network logical device to allow 2696 * fine-grained identification of different network device types. For 2697 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2698 */ 2699 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2700 2701 void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index, 2702 enum netdev_queue_type type, 2703 struct napi_struct *napi); 2704 2705 static inline void netdev_lock(struct net_device *dev) 2706 { 2707 mutex_lock(&dev->lock); 2708 } 2709 2710 static inline void netdev_unlock(struct net_device *dev) 2711 { 2712 mutex_unlock(&dev->lock); 2713 } 2714 2715 static inline void netdev_assert_locked(struct net_device *dev) 2716 { 2717 lockdep_assert_held(&dev->lock); 2718 } 2719 2720 static inline void netdev_assert_locked_or_invisible(struct net_device *dev) 2721 { 2722 if (dev->reg_state == NETREG_REGISTERED || 2723 dev->reg_state == NETREG_UNREGISTERING) 2724 netdev_assert_locked(dev); 2725 } 2726 2727 static inline void netif_napi_set_irq_locked(struct napi_struct *napi, int irq) 2728 { 2729 napi->irq = irq; 2730 } 2731 2732 static inline void netif_napi_set_irq(struct napi_struct *napi, int irq) 2733 { 2734 netdev_lock(napi->dev); 2735 netif_napi_set_irq_locked(napi, irq); 2736 netdev_unlock(napi->dev); 2737 } 2738 2739 /* Default NAPI poll() weight 2740 * Device drivers are strongly advised to not use bigger value 2741 */ 2742 #define NAPI_POLL_WEIGHT 64 2743 2744 void netif_napi_add_weight_locked(struct net_device *dev, 2745 struct napi_struct *napi, 2746 int (*poll)(struct napi_struct *, int), 2747 int weight); 2748 2749 static inline void 2750 netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2751 int (*poll)(struct napi_struct *, int), int weight) 2752 { 2753 netdev_lock(dev); 2754 netif_napi_add_weight_locked(dev, napi, poll, weight); 2755 netdev_unlock(dev); 2756 } 2757 2758 /** 2759 * netif_napi_add() - initialize a NAPI context 2760 * @dev: network device 2761 * @napi: NAPI context 2762 * @poll: polling function 2763 * 2764 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2765 * *any* of the other NAPI-related functions. 2766 */ 2767 static inline void 2768 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2769 int (*poll)(struct napi_struct *, int)) 2770 { 2771 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2772 } 2773 2774 static inline void 2775 netif_napi_add_locked(struct net_device *dev, struct napi_struct *napi, 2776 int (*poll)(struct napi_struct *, int)) 2777 { 2778 netif_napi_add_weight_locked(dev, napi, poll, NAPI_POLL_WEIGHT); 2779 } 2780 2781 static inline void 2782 netif_napi_add_tx_weight(struct net_device *dev, 2783 struct napi_struct *napi, 2784 int (*poll)(struct napi_struct *, int), 2785 int weight) 2786 { 2787 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2788 netif_napi_add_weight(dev, napi, poll, weight); 2789 } 2790 2791 static inline void 2792 netif_napi_add_config_locked(struct net_device *dev, struct napi_struct *napi, 2793 int (*poll)(struct napi_struct *, int), int index) 2794 { 2795 napi->index = index; 2796 napi->config = &dev->napi_config[index]; 2797 netif_napi_add_weight_locked(dev, napi, poll, NAPI_POLL_WEIGHT); 2798 } 2799 2800 /** 2801 * netif_napi_add_config - initialize a NAPI context with persistent config 2802 * @dev: network device 2803 * @napi: NAPI context 2804 * @poll: polling function 2805 * @index: the NAPI index 2806 */ 2807 static inline void 2808 netif_napi_add_config(struct net_device *dev, struct napi_struct *napi, 2809 int (*poll)(struct napi_struct *, int), int index) 2810 { 2811 netdev_lock(dev); 2812 netif_napi_add_config_locked(dev, napi, poll, index); 2813 netdev_unlock(dev); 2814 } 2815 2816 /** 2817 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2818 * @dev: network device 2819 * @napi: NAPI context 2820 * @poll: polling function 2821 * 2822 * This variant of netif_napi_add() should be used from drivers using NAPI 2823 * to exclusively poll a TX queue. 2824 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2825 */ 2826 static inline void netif_napi_add_tx(struct net_device *dev, 2827 struct napi_struct *napi, 2828 int (*poll)(struct napi_struct *, int)) 2829 { 2830 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2831 } 2832 2833 void __netif_napi_del_locked(struct napi_struct *napi); 2834 2835 /** 2836 * __netif_napi_del - remove a NAPI context 2837 * @napi: NAPI context 2838 * 2839 * Warning: caller must observe RCU grace period before freeing memory 2840 * containing @napi. Drivers might want to call this helper to combine 2841 * all the needed RCU grace periods into a single one. 2842 */ 2843 static inline void __netif_napi_del(struct napi_struct *napi) 2844 { 2845 netdev_lock(napi->dev); 2846 __netif_napi_del_locked(napi); 2847 netdev_unlock(napi->dev); 2848 } 2849 2850 static inline void netif_napi_del_locked(struct napi_struct *napi) 2851 { 2852 __netif_napi_del_locked(napi); 2853 synchronize_net(); 2854 } 2855 2856 /** 2857 * netif_napi_del - remove a NAPI context 2858 * @napi: NAPI context 2859 * 2860 * netif_napi_del() removes a NAPI context from the network device NAPI list 2861 */ 2862 static inline void netif_napi_del(struct napi_struct *napi) 2863 { 2864 __netif_napi_del(napi); 2865 synchronize_net(); 2866 } 2867 2868 struct packet_type { 2869 __be16 type; /* This is really htons(ether_type). */ 2870 bool ignore_outgoing; 2871 struct net_device *dev; /* NULL is wildcarded here */ 2872 netdevice_tracker dev_tracker; 2873 int (*func) (struct sk_buff *, 2874 struct net_device *, 2875 struct packet_type *, 2876 struct net_device *); 2877 void (*list_func) (struct list_head *, 2878 struct packet_type *, 2879 struct net_device *); 2880 bool (*id_match)(struct packet_type *ptype, 2881 struct sock *sk); 2882 struct net *af_packet_net; 2883 void *af_packet_priv; 2884 struct list_head list; 2885 }; 2886 2887 struct offload_callbacks { 2888 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2889 netdev_features_t features); 2890 struct sk_buff *(*gro_receive)(struct list_head *head, 2891 struct sk_buff *skb); 2892 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2893 }; 2894 2895 struct packet_offload { 2896 __be16 type; /* This is really htons(ether_type). */ 2897 u16 priority; 2898 struct offload_callbacks callbacks; 2899 struct list_head list; 2900 }; 2901 2902 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2903 struct pcpu_sw_netstats { 2904 u64_stats_t rx_packets; 2905 u64_stats_t rx_bytes; 2906 u64_stats_t tx_packets; 2907 u64_stats_t tx_bytes; 2908 struct u64_stats_sync syncp; 2909 } __aligned(4 * sizeof(u64)); 2910 2911 struct pcpu_dstats { 2912 u64_stats_t rx_packets; 2913 u64_stats_t rx_bytes; 2914 u64_stats_t tx_packets; 2915 u64_stats_t tx_bytes; 2916 u64_stats_t rx_drops; 2917 u64_stats_t tx_drops; 2918 struct u64_stats_sync syncp; 2919 } __aligned(8 * sizeof(u64)); 2920 2921 struct pcpu_lstats { 2922 u64_stats_t packets; 2923 u64_stats_t bytes; 2924 struct u64_stats_sync syncp; 2925 } __aligned(2 * sizeof(u64)); 2926 2927 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2928 2929 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2930 { 2931 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2932 2933 u64_stats_update_begin(&tstats->syncp); 2934 u64_stats_add(&tstats->rx_bytes, len); 2935 u64_stats_inc(&tstats->rx_packets); 2936 u64_stats_update_end(&tstats->syncp); 2937 } 2938 2939 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2940 unsigned int packets, 2941 unsigned int len) 2942 { 2943 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2944 2945 u64_stats_update_begin(&tstats->syncp); 2946 u64_stats_add(&tstats->tx_bytes, len); 2947 u64_stats_add(&tstats->tx_packets, packets); 2948 u64_stats_update_end(&tstats->syncp); 2949 } 2950 2951 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2952 { 2953 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2954 2955 u64_stats_update_begin(&lstats->syncp); 2956 u64_stats_add(&lstats->bytes, len); 2957 u64_stats_inc(&lstats->packets); 2958 u64_stats_update_end(&lstats->syncp); 2959 } 2960 2961 static inline void dev_dstats_rx_add(struct net_device *dev, 2962 unsigned int len) 2963 { 2964 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 2965 2966 u64_stats_update_begin(&dstats->syncp); 2967 u64_stats_inc(&dstats->rx_packets); 2968 u64_stats_add(&dstats->rx_bytes, len); 2969 u64_stats_update_end(&dstats->syncp); 2970 } 2971 2972 static inline void dev_dstats_rx_dropped(struct net_device *dev) 2973 { 2974 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 2975 2976 u64_stats_update_begin(&dstats->syncp); 2977 u64_stats_inc(&dstats->rx_drops); 2978 u64_stats_update_end(&dstats->syncp); 2979 } 2980 2981 static inline void dev_dstats_tx_add(struct net_device *dev, 2982 unsigned int len) 2983 { 2984 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 2985 2986 u64_stats_update_begin(&dstats->syncp); 2987 u64_stats_inc(&dstats->tx_packets); 2988 u64_stats_add(&dstats->tx_bytes, len); 2989 u64_stats_update_end(&dstats->syncp); 2990 } 2991 2992 static inline void dev_dstats_tx_dropped(struct net_device *dev) 2993 { 2994 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 2995 2996 u64_stats_update_begin(&dstats->syncp); 2997 u64_stats_inc(&dstats->tx_drops); 2998 u64_stats_update_end(&dstats->syncp); 2999 } 3000 3001 #define __netdev_alloc_pcpu_stats(type, gfp) \ 3002 ({ \ 3003 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 3004 if (pcpu_stats) { \ 3005 int __cpu; \ 3006 for_each_possible_cpu(__cpu) { \ 3007 typeof(type) *stat; \ 3008 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 3009 u64_stats_init(&stat->syncp); \ 3010 } \ 3011 } \ 3012 pcpu_stats; \ 3013 }) 3014 3015 #define netdev_alloc_pcpu_stats(type) \ 3016 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 3017 3018 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 3019 ({ \ 3020 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 3021 if (pcpu_stats) { \ 3022 int __cpu; \ 3023 for_each_possible_cpu(__cpu) { \ 3024 typeof(type) *stat; \ 3025 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 3026 u64_stats_init(&stat->syncp); \ 3027 } \ 3028 } \ 3029 pcpu_stats; \ 3030 }) 3031 3032 enum netdev_lag_tx_type { 3033 NETDEV_LAG_TX_TYPE_UNKNOWN, 3034 NETDEV_LAG_TX_TYPE_RANDOM, 3035 NETDEV_LAG_TX_TYPE_BROADCAST, 3036 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 3037 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 3038 NETDEV_LAG_TX_TYPE_HASH, 3039 }; 3040 3041 enum netdev_lag_hash { 3042 NETDEV_LAG_HASH_NONE, 3043 NETDEV_LAG_HASH_L2, 3044 NETDEV_LAG_HASH_L34, 3045 NETDEV_LAG_HASH_L23, 3046 NETDEV_LAG_HASH_E23, 3047 NETDEV_LAG_HASH_E34, 3048 NETDEV_LAG_HASH_VLAN_SRCMAC, 3049 NETDEV_LAG_HASH_UNKNOWN, 3050 }; 3051 3052 struct netdev_lag_upper_info { 3053 enum netdev_lag_tx_type tx_type; 3054 enum netdev_lag_hash hash_type; 3055 }; 3056 3057 struct netdev_lag_lower_state_info { 3058 u8 link_up : 1, 3059 tx_enabled : 1; 3060 }; 3061 3062 #include <linux/notifier.h> 3063 3064 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 3065 * and the rtnetlink notification exclusion list in rtnetlink_event() when 3066 * adding new types. 3067 */ 3068 enum netdev_cmd { 3069 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 3070 NETDEV_DOWN, 3071 NETDEV_REBOOT, /* Tell a protocol stack a network interface 3072 detected a hardware crash and restarted 3073 - we can use this eg to kick tcp sessions 3074 once done */ 3075 NETDEV_CHANGE, /* Notify device state change */ 3076 NETDEV_REGISTER, 3077 NETDEV_UNREGISTER, 3078 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 3079 NETDEV_CHANGEADDR, /* notify after the address change */ 3080 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 3081 NETDEV_GOING_DOWN, 3082 NETDEV_CHANGENAME, 3083 NETDEV_FEAT_CHANGE, 3084 NETDEV_BONDING_FAILOVER, 3085 NETDEV_PRE_UP, 3086 NETDEV_PRE_TYPE_CHANGE, 3087 NETDEV_POST_TYPE_CHANGE, 3088 NETDEV_POST_INIT, 3089 NETDEV_PRE_UNINIT, 3090 NETDEV_RELEASE, 3091 NETDEV_NOTIFY_PEERS, 3092 NETDEV_JOIN, 3093 NETDEV_CHANGEUPPER, 3094 NETDEV_RESEND_IGMP, 3095 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 3096 NETDEV_CHANGEINFODATA, 3097 NETDEV_BONDING_INFO, 3098 NETDEV_PRECHANGEUPPER, 3099 NETDEV_CHANGELOWERSTATE, 3100 NETDEV_UDP_TUNNEL_PUSH_INFO, 3101 NETDEV_UDP_TUNNEL_DROP_INFO, 3102 NETDEV_CHANGE_TX_QUEUE_LEN, 3103 NETDEV_CVLAN_FILTER_PUSH_INFO, 3104 NETDEV_CVLAN_FILTER_DROP_INFO, 3105 NETDEV_SVLAN_FILTER_PUSH_INFO, 3106 NETDEV_SVLAN_FILTER_DROP_INFO, 3107 NETDEV_OFFLOAD_XSTATS_ENABLE, 3108 NETDEV_OFFLOAD_XSTATS_DISABLE, 3109 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 3110 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 3111 NETDEV_XDP_FEAT_CHANGE, 3112 }; 3113 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 3114 3115 int register_netdevice_notifier(struct notifier_block *nb); 3116 int unregister_netdevice_notifier(struct notifier_block *nb); 3117 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 3118 int unregister_netdevice_notifier_net(struct net *net, 3119 struct notifier_block *nb); 3120 int register_netdevice_notifier_dev_net(struct net_device *dev, 3121 struct notifier_block *nb, 3122 struct netdev_net_notifier *nn); 3123 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 3124 struct notifier_block *nb, 3125 struct netdev_net_notifier *nn); 3126 3127 struct netdev_notifier_info { 3128 struct net_device *dev; 3129 struct netlink_ext_ack *extack; 3130 }; 3131 3132 struct netdev_notifier_info_ext { 3133 struct netdev_notifier_info info; /* must be first */ 3134 union { 3135 u32 mtu; 3136 } ext; 3137 }; 3138 3139 struct netdev_notifier_change_info { 3140 struct netdev_notifier_info info; /* must be first */ 3141 unsigned int flags_changed; 3142 }; 3143 3144 struct netdev_notifier_changeupper_info { 3145 struct netdev_notifier_info info; /* must be first */ 3146 struct net_device *upper_dev; /* new upper dev */ 3147 bool master; /* is upper dev master */ 3148 bool linking; /* is the notification for link or unlink */ 3149 void *upper_info; /* upper dev info */ 3150 }; 3151 3152 struct netdev_notifier_changelowerstate_info { 3153 struct netdev_notifier_info info; /* must be first */ 3154 void *lower_state_info; /* is lower dev state */ 3155 }; 3156 3157 struct netdev_notifier_pre_changeaddr_info { 3158 struct netdev_notifier_info info; /* must be first */ 3159 const unsigned char *dev_addr; 3160 }; 3161 3162 enum netdev_offload_xstats_type { 3163 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 3164 }; 3165 3166 struct netdev_notifier_offload_xstats_info { 3167 struct netdev_notifier_info info; /* must be first */ 3168 enum netdev_offload_xstats_type type; 3169 3170 union { 3171 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 3172 struct netdev_notifier_offload_xstats_rd *report_delta; 3173 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 3174 struct netdev_notifier_offload_xstats_ru *report_used; 3175 }; 3176 }; 3177 3178 int netdev_offload_xstats_enable(struct net_device *dev, 3179 enum netdev_offload_xstats_type type, 3180 struct netlink_ext_ack *extack); 3181 int netdev_offload_xstats_disable(struct net_device *dev, 3182 enum netdev_offload_xstats_type type); 3183 bool netdev_offload_xstats_enabled(const struct net_device *dev, 3184 enum netdev_offload_xstats_type type); 3185 int netdev_offload_xstats_get(struct net_device *dev, 3186 enum netdev_offload_xstats_type type, 3187 struct rtnl_hw_stats64 *stats, bool *used, 3188 struct netlink_ext_ack *extack); 3189 void 3190 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 3191 const struct rtnl_hw_stats64 *stats); 3192 void 3193 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 3194 void netdev_offload_xstats_push_delta(struct net_device *dev, 3195 enum netdev_offload_xstats_type type, 3196 const struct rtnl_hw_stats64 *stats); 3197 3198 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 3199 struct net_device *dev) 3200 { 3201 info->dev = dev; 3202 info->extack = NULL; 3203 } 3204 3205 static inline struct net_device * 3206 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 3207 { 3208 return info->dev; 3209 } 3210 3211 static inline struct netlink_ext_ack * 3212 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 3213 { 3214 return info->extack; 3215 } 3216 3217 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 3218 int call_netdevice_notifiers_info(unsigned long val, 3219 struct netdev_notifier_info *info); 3220 3221 #define for_each_netdev(net, d) \ 3222 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 3223 #define for_each_netdev_reverse(net, d) \ 3224 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 3225 #define for_each_netdev_rcu(net, d) \ 3226 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 3227 #define for_each_netdev_safe(net, d, n) \ 3228 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 3229 #define for_each_netdev_continue(net, d) \ 3230 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 3231 #define for_each_netdev_continue_reverse(net, d) \ 3232 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 3233 dev_list) 3234 #define for_each_netdev_continue_rcu(net, d) \ 3235 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 3236 #define for_each_netdev_in_bond_rcu(bond, slave) \ 3237 for_each_netdev_rcu(&init_net, slave) \ 3238 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 3239 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 3240 3241 #define for_each_netdev_dump(net, d, ifindex) \ 3242 for (; (d = xa_find(&(net)->dev_by_index, &ifindex, \ 3243 ULONG_MAX, XA_PRESENT)); ifindex++) 3244 3245 static inline struct net_device *next_net_device(struct net_device *dev) 3246 { 3247 struct list_head *lh; 3248 struct net *net; 3249 3250 net = dev_net(dev); 3251 lh = dev->dev_list.next; 3252 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3253 } 3254 3255 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 3256 { 3257 struct list_head *lh; 3258 struct net *net; 3259 3260 net = dev_net(dev); 3261 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 3262 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3263 } 3264 3265 static inline struct net_device *first_net_device(struct net *net) 3266 { 3267 return list_empty(&net->dev_base_head) ? NULL : 3268 net_device_entry(net->dev_base_head.next); 3269 } 3270 3271 static inline struct net_device *first_net_device_rcu(struct net *net) 3272 { 3273 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 3274 3275 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3276 } 3277 3278 int netdev_boot_setup_check(struct net_device *dev); 3279 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 3280 const char *hwaddr); 3281 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 3282 void dev_add_pack(struct packet_type *pt); 3283 void dev_remove_pack(struct packet_type *pt); 3284 void __dev_remove_pack(struct packet_type *pt); 3285 void dev_add_offload(struct packet_offload *po); 3286 void dev_remove_offload(struct packet_offload *po); 3287 3288 int dev_get_iflink(const struct net_device *dev); 3289 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 3290 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 3291 struct net_device_path_stack *stack); 3292 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 3293 unsigned short mask); 3294 struct net_device *dev_get_by_name(struct net *net, const char *name); 3295 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 3296 struct net_device *__dev_get_by_name(struct net *net, const char *name); 3297 bool netdev_name_in_use(struct net *net, const char *name); 3298 int dev_alloc_name(struct net_device *dev, const char *name); 3299 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 3300 void dev_close(struct net_device *dev); 3301 void dev_close_many(struct list_head *head, bool unlink); 3302 void dev_disable_lro(struct net_device *dev); 3303 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 3304 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 3305 struct net_device *sb_dev); 3306 3307 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3308 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3309 3310 static inline int dev_queue_xmit(struct sk_buff *skb) 3311 { 3312 return __dev_queue_xmit(skb, NULL); 3313 } 3314 3315 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3316 struct net_device *sb_dev) 3317 { 3318 return __dev_queue_xmit(skb, sb_dev); 3319 } 3320 3321 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3322 { 3323 int ret; 3324 3325 ret = __dev_direct_xmit(skb, queue_id); 3326 if (!dev_xmit_complete(ret)) 3327 kfree_skb(skb); 3328 return ret; 3329 } 3330 3331 int register_netdevice(struct net_device *dev); 3332 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3333 void unregister_netdevice_many(struct list_head *head); 3334 static inline void unregister_netdevice(struct net_device *dev) 3335 { 3336 unregister_netdevice_queue(dev, NULL); 3337 } 3338 3339 int netdev_refcnt_read(const struct net_device *dev); 3340 void free_netdev(struct net_device *dev); 3341 3342 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3343 struct sk_buff *skb, 3344 bool all_slaves); 3345 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3346 struct sock *sk); 3347 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3348 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3349 struct net_device *netdev_get_by_index(struct net *net, int ifindex, 3350 netdevice_tracker *tracker, gfp_t gfp); 3351 struct net_device *netdev_get_by_name(struct net *net, const char *name, 3352 netdevice_tracker *tracker, gfp_t gfp); 3353 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3354 void netdev_copy_name(struct net_device *dev, char *name); 3355 3356 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3357 unsigned short type, 3358 const void *daddr, const void *saddr, 3359 unsigned int len) 3360 { 3361 if (!dev->header_ops || !dev->header_ops->create) 3362 return 0; 3363 3364 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3365 } 3366 3367 static inline int dev_parse_header(const struct sk_buff *skb, 3368 unsigned char *haddr) 3369 { 3370 const struct net_device *dev = skb->dev; 3371 3372 if (!dev->header_ops || !dev->header_ops->parse) 3373 return 0; 3374 return dev->header_ops->parse(skb, haddr); 3375 } 3376 3377 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3378 { 3379 const struct net_device *dev = skb->dev; 3380 3381 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3382 return 0; 3383 return dev->header_ops->parse_protocol(skb); 3384 } 3385 3386 /* ll_header must have at least hard_header_len allocated */ 3387 static inline bool dev_validate_header(const struct net_device *dev, 3388 char *ll_header, int len) 3389 { 3390 if (likely(len >= dev->hard_header_len)) 3391 return true; 3392 if (len < dev->min_header_len) 3393 return false; 3394 3395 if (capable(CAP_SYS_RAWIO)) { 3396 memset(ll_header + len, 0, dev->hard_header_len - len); 3397 return true; 3398 } 3399 3400 if (dev->header_ops && dev->header_ops->validate) 3401 return dev->header_ops->validate(ll_header, len); 3402 3403 return false; 3404 } 3405 3406 static inline bool dev_has_header(const struct net_device *dev) 3407 { 3408 return dev->header_ops && dev->header_ops->create; 3409 } 3410 3411 /* 3412 * Incoming packets are placed on per-CPU queues 3413 */ 3414 struct softnet_data { 3415 struct list_head poll_list; 3416 struct sk_buff_head process_queue; 3417 local_lock_t process_queue_bh_lock; 3418 3419 /* stats */ 3420 unsigned int processed; 3421 unsigned int time_squeeze; 3422 #ifdef CONFIG_RPS 3423 struct softnet_data *rps_ipi_list; 3424 #endif 3425 3426 unsigned int received_rps; 3427 bool in_net_rx_action; 3428 bool in_napi_threaded_poll; 3429 3430 #ifdef CONFIG_NET_FLOW_LIMIT 3431 struct sd_flow_limit __rcu *flow_limit; 3432 #endif 3433 struct Qdisc *output_queue; 3434 struct Qdisc **output_queue_tailp; 3435 struct sk_buff *completion_queue; 3436 #ifdef CONFIG_XFRM_OFFLOAD 3437 struct sk_buff_head xfrm_backlog; 3438 #endif 3439 /* written and read only by owning cpu: */ 3440 struct netdev_xmit xmit; 3441 #ifdef CONFIG_RPS 3442 /* input_queue_head should be written by cpu owning this struct, 3443 * and only read by other cpus. Worth using a cache line. 3444 */ 3445 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3446 3447 /* Elements below can be accessed between CPUs for RPS/RFS */ 3448 call_single_data_t csd ____cacheline_aligned_in_smp; 3449 struct softnet_data *rps_ipi_next; 3450 unsigned int cpu; 3451 unsigned int input_queue_tail; 3452 #endif 3453 struct sk_buff_head input_pkt_queue; 3454 struct napi_struct backlog; 3455 3456 atomic_t dropped ____cacheline_aligned_in_smp; 3457 3458 /* Another possibly contended cache line */ 3459 spinlock_t defer_lock ____cacheline_aligned_in_smp; 3460 int defer_count; 3461 int defer_ipi_scheduled; 3462 struct sk_buff *defer_list; 3463 call_single_data_t defer_csd; 3464 }; 3465 3466 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3467 DECLARE_PER_CPU(struct page_pool *, system_page_pool); 3468 3469 #ifndef CONFIG_PREEMPT_RT 3470 static inline int dev_recursion_level(void) 3471 { 3472 return this_cpu_read(softnet_data.xmit.recursion); 3473 } 3474 #else 3475 static inline int dev_recursion_level(void) 3476 { 3477 return current->net_xmit.recursion; 3478 } 3479 3480 #endif 3481 3482 void __netif_schedule(struct Qdisc *q); 3483 void netif_schedule_queue(struct netdev_queue *txq); 3484 3485 static inline void netif_tx_schedule_all(struct net_device *dev) 3486 { 3487 unsigned int i; 3488 3489 for (i = 0; i < dev->num_tx_queues; i++) 3490 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3491 } 3492 3493 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3494 { 3495 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3496 } 3497 3498 /** 3499 * netif_start_queue - allow transmit 3500 * @dev: network device 3501 * 3502 * Allow upper layers to call the device hard_start_xmit routine. 3503 */ 3504 static inline void netif_start_queue(struct net_device *dev) 3505 { 3506 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3507 } 3508 3509 static inline void netif_tx_start_all_queues(struct net_device *dev) 3510 { 3511 unsigned int i; 3512 3513 for (i = 0; i < dev->num_tx_queues; i++) { 3514 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3515 netif_tx_start_queue(txq); 3516 } 3517 } 3518 3519 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3520 3521 /** 3522 * netif_wake_queue - restart transmit 3523 * @dev: network device 3524 * 3525 * Allow upper layers to call the device hard_start_xmit routine. 3526 * Used for flow control when transmit resources are available. 3527 */ 3528 static inline void netif_wake_queue(struct net_device *dev) 3529 { 3530 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3531 } 3532 3533 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3534 { 3535 unsigned int i; 3536 3537 for (i = 0; i < dev->num_tx_queues; i++) { 3538 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3539 netif_tx_wake_queue(txq); 3540 } 3541 } 3542 3543 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3544 { 3545 /* Paired with READ_ONCE() from dev_watchdog() */ 3546 WRITE_ONCE(dev_queue->trans_start, jiffies); 3547 3548 /* This barrier is paired with smp_mb() from dev_watchdog() */ 3549 smp_mb__before_atomic(); 3550 3551 /* Must be an atomic op see netif_txq_try_stop() */ 3552 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3553 } 3554 3555 /** 3556 * netif_stop_queue - stop transmitted packets 3557 * @dev: network device 3558 * 3559 * Stop upper layers calling the device hard_start_xmit routine. 3560 * Used for flow control when transmit resources are unavailable. 3561 */ 3562 static inline void netif_stop_queue(struct net_device *dev) 3563 { 3564 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3565 } 3566 3567 void netif_tx_stop_all_queues(struct net_device *dev); 3568 3569 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3570 { 3571 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3572 } 3573 3574 /** 3575 * netif_queue_stopped - test if transmit queue is flowblocked 3576 * @dev: network device 3577 * 3578 * Test if transmit queue on device is currently unable to send. 3579 */ 3580 static inline bool netif_queue_stopped(const struct net_device *dev) 3581 { 3582 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3583 } 3584 3585 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3586 { 3587 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3588 } 3589 3590 static inline bool 3591 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3592 { 3593 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3594 } 3595 3596 static inline bool 3597 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3598 { 3599 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3600 } 3601 3602 /** 3603 * netdev_queue_set_dql_min_limit - set dql minimum limit 3604 * @dev_queue: pointer to transmit queue 3605 * @min_limit: dql minimum limit 3606 * 3607 * Forces xmit_more() to return true until the minimum threshold 3608 * defined by @min_limit is reached (or until the tx queue is 3609 * empty). Warning: to be use with care, misuse will impact the 3610 * latency. 3611 */ 3612 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3613 unsigned int min_limit) 3614 { 3615 #ifdef CONFIG_BQL 3616 dev_queue->dql.min_limit = min_limit; 3617 #endif 3618 } 3619 3620 static inline int netdev_queue_dql_avail(const struct netdev_queue *txq) 3621 { 3622 #ifdef CONFIG_BQL 3623 /* Non-BQL migrated drivers will return 0, too. */ 3624 return dql_avail(&txq->dql); 3625 #else 3626 return 0; 3627 #endif 3628 } 3629 3630 /** 3631 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3632 * @dev_queue: pointer to transmit queue 3633 * 3634 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3635 * to give appropriate hint to the CPU. 3636 */ 3637 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3638 { 3639 #ifdef CONFIG_BQL 3640 prefetchw(&dev_queue->dql.num_queued); 3641 #endif 3642 } 3643 3644 /** 3645 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3646 * @dev_queue: pointer to transmit queue 3647 * 3648 * BQL enabled drivers might use this helper in their TX completion path, 3649 * to give appropriate hint to the CPU. 3650 */ 3651 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3652 { 3653 #ifdef CONFIG_BQL 3654 prefetchw(&dev_queue->dql.limit); 3655 #endif 3656 } 3657 3658 /** 3659 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3660 * @dev_queue: network device queue 3661 * @bytes: number of bytes queued to the device queue 3662 * 3663 * Report the number of bytes queued for sending/completion to the network 3664 * device hardware queue. @bytes should be a good approximation and should 3665 * exactly match netdev_completed_queue() @bytes. 3666 * This is typically called once per packet, from ndo_start_xmit(). 3667 */ 3668 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3669 unsigned int bytes) 3670 { 3671 #ifdef CONFIG_BQL 3672 dql_queued(&dev_queue->dql, bytes); 3673 3674 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3675 return; 3676 3677 /* Paired with READ_ONCE() from dev_watchdog() */ 3678 WRITE_ONCE(dev_queue->trans_start, jiffies); 3679 3680 /* This barrier is paired with smp_mb() from dev_watchdog() */ 3681 smp_mb__before_atomic(); 3682 3683 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3684 3685 /* 3686 * The XOFF flag must be set before checking the dql_avail below, 3687 * because in netdev_tx_completed_queue we update the dql_completed 3688 * before checking the XOFF flag. 3689 */ 3690 smp_mb__after_atomic(); 3691 3692 /* check again in case another CPU has just made room avail */ 3693 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3694 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3695 #endif 3696 } 3697 3698 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3699 * that they should not test BQL status themselves. 3700 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3701 * skb of a batch. 3702 * Returns true if the doorbell must be used to kick the NIC. 3703 */ 3704 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3705 unsigned int bytes, 3706 bool xmit_more) 3707 { 3708 if (xmit_more) { 3709 #ifdef CONFIG_BQL 3710 dql_queued(&dev_queue->dql, bytes); 3711 #endif 3712 return netif_tx_queue_stopped(dev_queue); 3713 } 3714 netdev_tx_sent_queue(dev_queue, bytes); 3715 return true; 3716 } 3717 3718 /** 3719 * netdev_sent_queue - report the number of bytes queued to hardware 3720 * @dev: network device 3721 * @bytes: number of bytes queued to the hardware device queue 3722 * 3723 * Report the number of bytes queued for sending/completion to the network 3724 * device hardware queue#0. @bytes should be a good approximation and should 3725 * exactly match netdev_completed_queue() @bytes. 3726 * This is typically called once per packet, from ndo_start_xmit(). 3727 */ 3728 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3729 { 3730 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3731 } 3732 3733 static inline bool __netdev_sent_queue(struct net_device *dev, 3734 unsigned int bytes, 3735 bool xmit_more) 3736 { 3737 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3738 xmit_more); 3739 } 3740 3741 /** 3742 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3743 * @dev_queue: network device queue 3744 * @pkts: number of packets (currently ignored) 3745 * @bytes: number of bytes dequeued from the device queue 3746 * 3747 * Must be called at most once per TX completion round (and not per 3748 * individual packet), so that BQL can adjust its limits appropriately. 3749 */ 3750 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3751 unsigned int pkts, unsigned int bytes) 3752 { 3753 #ifdef CONFIG_BQL 3754 if (unlikely(!bytes)) 3755 return; 3756 3757 dql_completed(&dev_queue->dql, bytes); 3758 3759 /* 3760 * Without the memory barrier there is a small possibility that 3761 * netdev_tx_sent_queue will miss the update and cause the queue to 3762 * be stopped forever 3763 */ 3764 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */ 3765 3766 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3767 return; 3768 3769 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3770 netif_schedule_queue(dev_queue); 3771 #endif 3772 } 3773 3774 /** 3775 * netdev_completed_queue - report bytes and packets completed by device 3776 * @dev: network device 3777 * @pkts: actual number of packets sent over the medium 3778 * @bytes: actual number of bytes sent over the medium 3779 * 3780 * Report the number of bytes and packets transmitted by the network device 3781 * hardware queue over the physical medium, @bytes must exactly match the 3782 * @bytes amount passed to netdev_sent_queue() 3783 */ 3784 static inline void netdev_completed_queue(struct net_device *dev, 3785 unsigned int pkts, unsigned int bytes) 3786 { 3787 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3788 } 3789 3790 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3791 { 3792 #ifdef CONFIG_BQL 3793 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3794 dql_reset(&q->dql); 3795 #endif 3796 } 3797 3798 /** 3799 * netdev_tx_reset_subqueue - reset the BQL stats and state of a netdev queue 3800 * @dev: network device 3801 * @qid: stack index of the queue to reset 3802 */ 3803 static inline void netdev_tx_reset_subqueue(const struct net_device *dev, 3804 u32 qid) 3805 { 3806 netdev_tx_reset_queue(netdev_get_tx_queue(dev, qid)); 3807 } 3808 3809 /** 3810 * netdev_reset_queue - reset the packets and bytes count of a network device 3811 * @dev_queue: network device 3812 * 3813 * Reset the bytes and packet count of a network device and clear the 3814 * software flow control OFF bit for this network device 3815 */ 3816 static inline void netdev_reset_queue(struct net_device *dev_queue) 3817 { 3818 netdev_tx_reset_subqueue(dev_queue, 0); 3819 } 3820 3821 /** 3822 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3823 * @dev: network device 3824 * @queue_index: given tx queue index 3825 * 3826 * Returns 0 if given tx queue index >= number of device tx queues, 3827 * otherwise returns the originally passed tx queue index. 3828 */ 3829 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3830 { 3831 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3832 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3833 dev->name, queue_index, 3834 dev->real_num_tx_queues); 3835 return 0; 3836 } 3837 3838 return queue_index; 3839 } 3840 3841 /** 3842 * netif_running - test if up 3843 * @dev: network device 3844 * 3845 * Test if the device has been brought up. 3846 */ 3847 static inline bool netif_running(const struct net_device *dev) 3848 { 3849 return test_bit(__LINK_STATE_START, &dev->state); 3850 } 3851 3852 /* 3853 * Routines to manage the subqueues on a device. We only need start, 3854 * stop, and a check if it's stopped. All other device management is 3855 * done at the overall netdevice level. 3856 * Also test the device if we're multiqueue. 3857 */ 3858 3859 /** 3860 * netif_start_subqueue - allow sending packets on subqueue 3861 * @dev: network device 3862 * @queue_index: sub queue index 3863 * 3864 * Start individual transmit queue of a device with multiple transmit queues. 3865 */ 3866 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3867 { 3868 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3869 3870 netif_tx_start_queue(txq); 3871 } 3872 3873 /** 3874 * netif_stop_subqueue - stop sending packets on subqueue 3875 * @dev: network device 3876 * @queue_index: sub queue index 3877 * 3878 * Stop individual transmit queue of a device with multiple transmit queues. 3879 */ 3880 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3881 { 3882 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3883 netif_tx_stop_queue(txq); 3884 } 3885 3886 /** 3887 * __netif_subqueue_stopped - test status of subqueue 3888 * @dev: network device 3889 * @queue_index: sub queue index 3890 * 3891 * Check individual transmit queue of a device with multiple transmit queues. 3892 */ 3893 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3894 u16 queue_index) 3895 { 3896 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3897 3898 return netif_tx_queue_stopped(txq); 3899 } 3900 3901 /** 3902 * netif_subqueue_stopped - test status of subqueue 3903 * @dev: network device 3904 * @skb: sub queue buffer pointer 3905 * 3906 * Check individual transmit queue of a device with multiple transmit queues. 3907 */ 3908 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3909 struct sk_buff *skb) 3910 { 3911 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3912 } 3913 3914 /** 3915 * netif_wake_subqueue - allow sending packets on subqueue 3916 * @dev: network device 3917 * @queue_index: sub queue index 3918 * 3919 * Resume individual transmit queue of a device with multiple transmit queues. 3920 */ 3921 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3922 { 3923 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3924 3925 netif_tx_wake_queue(txq); 3926 } 3927 3928 #ifdef CONFIG_XPS 3929 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3930 u16 index); 3931 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3932 u16 index, enum xps_map_type type); 3933 3934 /** 3935 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3936 * @j: CPU/Rx queue index 3937 * @mask: bitmask of all cpus/rx queues 3938 * @nr_bits: number of bits in the bitmask 3939 * 3940 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3941 */ 3942 static inline bool netif_attr_test_mask(unsigned long j, 3943 const unsigned long *mask, 3944 unsigned int nr_bits) 3945 { 3946 cpu_max_bits_warn(j, nr_bits); 3947 return test_bit(j, mask); 3948 } 3949 3950 /** 3951 * netif_attr_test_online - Test for online CPU/Rx queue 3952 * @j: CPU/Rx queue index 3953 * @online_mask: bitmask for CPUs/Rx queues that are online 3954 * @nr_bits: number of bits in the bitmask 3955 * 3956 * Returns: true if a CPU/Rx queue is online. 3957 */ 3958 static inline bool netif_attr_test_online(unsigned long j, 3959 const unsigned long *online_mask, 3960 unsigned int nr_bits) 3961 { 3962 cpu_max_bits_warn(j, nr_bits); 3963 3964 if (online_mask) 3965 return test_bit(j, online_mask); 3966 3967 return (j < nr_bits); 3968 } 3969 3970 /** 3971 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3972 * @n: CPU/Rx queue index 3973 * @srcp: the cpumask/Rx queue mask pointer 3974 * @nr_bits: number of bits in the bitmask 3975 * 3976 * Returns: next (after n) CPU/Rx queue index in the mask; 3977 * >= nr_bits if no further CPUs/Rx queues set. 3978 */ 3979 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3980 unsigned int nr_bits) 3981 { 3982 /* -1 is a legal arg here. */ 3983 if (n != -1) 3984 cpu_max_bits_warn(n, nr_bits); 3985 3986 if (srcp) 3987 return find_next_bit(srcp, nr_bits, n + 1); 3988 3989 return n + 1; 3990 } 3991 3992 /** 3993 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3994 * @n: CPU/Rx queue index 3995 * @src1p: the first CPUs/Rx queues mask pointer 3996 * @src2p: the second CPUs/Rx queues mask pointer 3997 * @nr_bits: number of bits in the bitmask 3998 * 3999 * Returns: next (after n) CPU/Rx queue index set in both masks; 4000 * >= nr_bits if no further CPUs/Rx queues set in both. 4001 */ 4002 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 4003 const unsigned long *src2p, 4004 unsigned int nr_bits) 4005 { 4006 /* -1 is a legal arg here. */ 4007 if (n != -1) 4008 cpu_max_bits_warn(n, nr_bits); 4009 4010 if (src1p && src2p) 4011 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 4012 else if (src1p) 4013 return find_next_bit(src1p, nr_bits, n + 1); 4014 else if (src2p) 4015 return find_next_bit(src2p, nr_bits, n + 1); 4016 4017 return n + 1; 4018 } 4019 #else 4020 static inline int netif_set_xps_queue(struct net_device *dev, 4021 const struct cpumask *mask, 4022 u16 index) 4023 { 4024 return 0; 4025 } 4026 4027 static inline int __netif_set_xps_queue(struct net_device *dev, 4028 const unsigned long *mask, 4029 u16 index, enum xps_map_type type) 4030 { 4031 return 0; 4032 } 4033 #endif 4034 4035 /** 4036 * netif_is_multiqueue - test if device has multiple transmit queues 4037 * @dev: network device 4038 * 4039 * Check if device has multiple transmit queues 4040 */ 4041 static inline bool netif_is_multiqueue(const struct net_device *dev) 4042 { 4043 return dev->num_tx_queues > 1; 4044 } 4045 4046 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 4047 4048 #ifdef CONFIG_SYSFS 4049 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 4050 #else 4051 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 4052 unsigned int rxqs) 4053 { 4054 dev->real_num_rx_queues = rxqs; 4055 return 0; 4056 } 4057 #endif 4058 int netif_set_real_num_queues(struct net_device *dev, 4059 unsigned int txq, unsigned int rxq); 4060 4061 int netif_get_num_default_rss_queues(void); 4062 4063 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason); 4064 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason); 4065 4066 /* 4067 * It is not allowed to call kfree_skb() or consume_skb() from hardware 4068 * interrupt context or with hardware interrupts being disabled. 4069 * (in_hardirq() || irqs_disabled()) 4070 * 4071 * We provide four helpers that can be used in following contexts : 4072 * 4073 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 4074 * replacing kfree_skb(skb) 4075 * 4076 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 4077 * Typically used in place of consume_skb(skb) in TX completion path 4078 * 4079 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 4080 * replacing kfree_skb(skb) 4081 * 4082 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 4083 * and consumed a packet. Used in place of consume_skb(skb) 4084 */ 4085 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 4086 { 4087 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 4088 } 4089 4090 static inline void dev_consume_skb_irq(struct sk_buff *skb) 4091 { 4092 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED); 4093 } 4094 4095 static inline void dev_kfree_skb_any(struct sk_buff *skb) 4096 { 4097 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 4098 } 4099 4100 static inline void dev_consume_skb_any(struct sk_buff *skb) 4101 { 4102 dev_kfree_skb_any_reason(skb, SKB_CONSUMED); 4103 } 4104 4105 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 4106 const struct bpf_prog *xdp_prog); 4107 void generic_xdp_tx(struct sk_buff *skb, const struct bpf_prog *xdp_prog); 4108 int do_xdp_generic(const struct bpf_prog *xdp_prog, struct sk_buff **pskb); 4109 int netif_rx(struct sk_buff *skb); 4110 int __netif_rx(struct sk_buff *skb); 4111 4112 int netif_receive_skb(struct sk_buff *skb); 4113 int netif_receive_skb_core(struct sk_buff *skb); 4114 void netif_receive_skb_list_internal(struct list_head *head); 4115 void netif_receive_skb_list(struct list_head *head); 4116 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 4117 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 4118 struct sk_buff *napi_get_frags(struct napi_struct *napi); 4119 void napi_get_frags_check(struct napi_struct *napi); 4120 gro_result_t napi_gro_frags(struct napi_struct *napi); 4121 4122 static inline void napi_free_frags(struct napi_struct *napi) 4123 { 4124 kfree_skb(napi->skb); 4125 napi->skb = NULL; 4126 } 4127 4128 bool netdev_is_rx_handler_busy(struct net_device *dev); 4129 int netdev_rx_handler_register(struct net_device *dev, 4130 rx_handler_func_t *rx_handler, 4131 void *rx_handler_data); 4132 void netdev_rx_handler_unregister(struct net_device *dev); 4133 4134 bool dev_valid_name(const char *name); 4135 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 4136 { 4137 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 4138 } 4139 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 4140 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 4141 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 4142 void __user *data, bool *need_copyout); 4143 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 4144 int generic_hwtstamp_get_lower(struct net_device *dev, 4145 struct kernel_hwtstamp_config *kernel_cfg); 4146 int generic_hwtstamp_set_lower(struct net_device *dev, 4147 struct kernel_hwtstamp_config *kernel_cfg, 4148 struct netlink_ext_ack *extack); 4149 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 4150 unsigned int dev_get_flags(const struct net_device *); 4151 int __dev_change_flags(struct net_device *dev, unsigned int flags, 4152 struct netlink_ext_ack *extack); 4153 int dev_change_flags(struct net_device *dev, unsigned int flags, 4154 struct netlink_ext_ack *extack); 4155 int dev_set_alias(struct net_device *, const char *, size_t); 4156 int dev_get_alias(const struct net_device *, char *, size_t); 4157 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 4158 const char *pat, int new_ifindex); 4159 static inline 4160 int dev_change_net_namespace(struct net_device *dev, struct net *net, 4161 const char *pat) 4162 { 4163 return __dev_change_net_namespace(dev, net, pat, 0); 4164 } 4165 int __dev_set_mtu(struct net_device *, int); 4166 int dev_set_mtu(struct net_device *, int); 4167 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 4168 struct netlink_ext_ack *extack); 4169 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 4170 struct netlink_ext_ack *extack); 4171 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 4172 struct netlink_ext_ack *extack); 4173 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 4174 int dev_get_port_parent_id(struct net_device *dev, 4175 struct netdev_phys_item_id *ppid, bool recurse); 4176 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 4177 4178 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 4179 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 4180 struct netdev_queue *txq, int *ret); 4181 4182 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 4183 u8 dev_xdp_prog_count(struct net_device *dev); 4184 int dev_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf); 4185 u8 dev_xdp_sb_prog_count(struct net_device *dev); 4186 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 4187 4188 u32 dev_get_min_mp_channel_count(const struct net_device *dev); 4189 4190 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4191 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4192 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 4193 bool is_skb_forwardable(const struct net_device *dev, 4194 const struct sk_buff *skb); 4195 4196 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 4197 const struct sk_buff *skb, 4198 const bool check_mtu) 4199 { 4200 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 4201 unsigned int len; 4202 4203 if (!(dev->flags & IFF_UP)) 4204 return false; 4205 4206 if (!check_mtu) 4207 return true; 4208 4209 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 4210 if (skb->len <= len) 4211 return true; 4212 4213 /* if TSO is enabled, we don't care about the length as the packet 4214 * could be forwarded without being segmented before 4215 */ 4216 if (skb_is_gso(skb)) 4217 return true; 4218 4219 return false; 4220 } 4221 4222 void netdev_core_stats_inc(struct net_device *dev, u32 offset); 4223 4224 #define DEV_CORE_STATS_INC(FIELD) \ 4225 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 4226 { \ 4227 netdev_core_stats_inc(dev, \ 4228 offsetof(struct net_device_core_stats, FIELD)); \ 4229 } 4230 DEV_CORE_STATS_INC(rx_dropped) 4231 DEV_CORE_STATS_INC(tx_dropped) 4232 DEV_CORE_STATS_INC(rx_nohandler) 4233 DEV_CORE_STATS_INC(rx_otherhost_dropped) 4234 #undef DEV_CORE_STATS_INC 4235 4236 static __always_inline int ____dev_forward_skb(struct net_device *dev, 4237 struct sk_buff *skb, 4238 const bool check_mtu) 4239 { 4240 if (skb_orphan_frags(skb, GFP_ATOMIC) || 4241 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 4242 dev_core_stats_rx_dropped_inc(dev); 4243 kfree_skb(skb); 4244 return NET_RX_DROP; 4245 } 4246 4247 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 4248 skb->priority = 0; 4249 return 0; 4250 } 4251 4252 bool dev_nit_active(struct net_device *dev); 4253 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 4254 4255 static inline void __dev_put(struct net_device *dev) 4256 { 4257 if (dev) { 4258 #ifdef CONFIG_PCPU_DEV_REFCNT 4259 this_cpu_dec(*dev->pcpu_refcnt); 4260 #else 4261 refcount_dec(&dev->dev_refcnt); 4262 #endif 4263 } 4264 } 4265 4266 static inline void __dev_hold(struct net_device *dev) 4267 { 4268 if (dev) { 4269 #ifdef CONFIG_PCPU_DEV_REFCNT 4270 this_cpu_inc(*dev->pcpu_refcnt); 4271 #else 4272 refcount_inc(&dev->dev_refcnt); 4273 #endif 4274 } 4275 } 4276 4277 static inline void __netdev_tracker_alloc(struct net_device *dev, 4278 netdevice_tracker *tracker, 4279 gfp_t gfp) 4280 { 4281 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4282 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 4283 #endif 4284 } 4285 4286 /* netdev_tracker_alloc() can upgrade a prior untracked reference 4287 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4288 */ 4289 static inline void netdev_tracker_alloc(struct net_device *dev, 4290 netdevice_tracker *tracker, gfp_t gfp) 4291 { 4292 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4293 refcount_dec(&dev->refcnt_tracker.no_tracker); 4294 __netdev_tracker_alloc(dev, tracker, gfp); 4295 #endif 4296 } 4297 4298 static inline void netdev_tracker_free(struct net_device *dev, 4299 netdevice_tracker *tracker) 4300 { 4301 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4302 ref_tracker_free(&dev->refcnt_tracker, tracker); 4303 #endif 4304 } 4305 4306 static inline void netdev_hold(struct net_device *dev, 4307 netdevice_tracker *tracker, gfp_t gfp) 4308 { 4309 if (dev) { 4310 __dev_hold(dev); 4311 __netdev_tracker_alloc(dev, tracker, gfp); 4312 } 4313 } 4314 4315 static inline void netdev_put(struct net_device *dev, 4316 netdevice_tracker *tracker) 4317 { 4318 if (dev) { 4319 netdev_tracker_free(dev, tracker); 4320 __dev_put(dev); 4321 } 4322 } 4323 4324 /** 4325 * dev_hold - get reference to device 4326 * @dev: network device 4327 * 4328 * Hold reference to device to keep it from being freed. 4329 * Try using netdev_hold() instead. 4330 */ 4331 static inline void dev_hold(struct net_device *dev) 4332 { 4333 netdev_hold(dev, NULL, GFP_ATOMIC); 4334 } 4335 4336 /** 4337 * dev_put - release reference to device 4338 * @dev: network device 4339 * 4340 * Release reference to device to allow it to be freed. 4341 * Try using netdev_put() instead. 4342 */ 4343 static inline void dev_put(struct net_device *dev) 4344 { 4345 netdev_put(dev, NULL); 4346 } 4347 4348 DEFINE_FREE(dev_put, struct net_device *, if (_T) dev_put(_T)) 4349 4350 static inline void netdev_ref_replace(struct net_device *odev, 4351 struct net_device *ndev, 4352 netdevice_tracker *tracker, 4353 gfp_t gfp) 4354 { 4355 if (odev) 4356 netdev_tracker_free(odev, tracker); 4357 4358 __dev_hold(ndev); 4359 __dev_put(odev); 4360 4361 if (ndev) 4362 __netdev_tracker_alloc(ndev, tracker, gfp); 4363 } 4364 4365 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4366 * and _off may be called from IRQ context, but it is caller 4367 * who is responsible for serialization of these calls. 4368 * 4369 * The name carrier is inappropriate, these functions should really be 4370 * called netif_lowerlayer_*() because they represent the state of any 4371 * kind of lower layer not just hardware media. 4372 */ 4373 void linkwatch_fire_event(struct net_device *dev); 4374 4375 /** 4376 * linkwatch_sync_dev - sync linkwatch for the given device 4377 * @dev: network device to sync linkwatch for 4378 * 4379 * Sync linkwatch for the given device, removing it from the 4380 * pending work list (if queued). 4381 */ 4382 void linkwatch_sync_dev(struct net_device *dev); 4383 4384 /** 4385 * netif_carrier_ok - test if carrier present 4386 * @dev: network device 4387 * 4388 * Check if carrier is present on device 4389 */ 4390 static inline bool netif_carrier_ok(const struct net_device *dev) 4391 { 4392 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4393 } 4394 4395 unsigned long dev_trans_start(struct net_device *dev); 4396 4397 void netdev_watchdog_up(struct net_device *dev); 4398 4399 void netif_carrier_on(struct net_device *dev); 4400 void netif_carrier_off(struct net_device *dev); 4401 void netif_carrier_event(struct net_device *dev); 4402 4403 /** 4404 * netif_dormant_on - mark device as dormant. 4405 * @dev: network device 4406 * 4407 * Mark device as dormant (as per RFC2863). 4408 * 4409 * The dormant state indicates that the relevant interface is not 4410 * actually in a condition to pass packets (i.e., it is not 'up') but is 4411 * in a "pending" state, waiting for some external event. For "on- 4412 * demand" interfaces, this new state identifies the situation where the 4413 * interface is waiting for events to place it in the up state. 4414 */ 4415 static inline void netif_dormant_on(struct net_device *dev) 4416 { 4417 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4418 linkwatch_fire_event(dev); 4419 } 4420 4421 /** 4422 * netif_dormant_off - set device as not dormant. 4423 * @dev: network device 4424 * 4425 * Device is not in dormant state. 4426 */ 4427 static inline void netif_dormant_off(struct net_device *dev) 4428 { 4429 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4430 linkwatch_fire_event(dev); 4431 } 4432 4433 /** 4434 * netif_dormant - test if device is dormant 4435 * @dev: network device 4436 * 4437 * Check if device is dormant. 4438 */ 4439 static inline bool netif_dormant(const struct net_device *dev) 4440 { 4441 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4442 } 4443 4444 4445 /** 4446 * netif_testing_on - mark device as under test. 4447 * @dev: network device 4448 * 4449 * Mark device as under test (as per RFC2863). 4450 * 4451 * The testing state indicates that some test(s) must be performed on 4452 * the interface. After completion, of the test, the interface state 4453 * will change to up, dormant, or down, as appropriate. 4454 */ 4455 static inline void netif_testing_on(struct net_device *dev) 4456 { 4457 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4458 linkwatch_fire_event(dev); 4459 } 4460 4461 /** 4462 * netif_testing_off - set device as not under test. 4463 * @dev: network device 4464 * 4465 * Device is not in testing state. 4466 */ 4467 static inline void netif_testing_off(struct net_device *dev) 4468 { 4469 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4470 linkwatch_fire_event(dev); 4471 } 4472 4473 /** 4474 * netif_testing - test if device is under test 4475 * @dev: network device 4476 * 4477 * Check if device is under test 4478 */ 4479 static inline bool netif_testing(const struct net_device *dev) 4480 { 4481 return test_bit(__LINK_STATE_TESTING, &dev->state); 4482 } 4483 4484 4485 /** 4486 * netif_oper_up - test if device is operational 4487 * @dev: network device 4488 * 4489 * Check if carrier is operational 4490 */ 4491 static inline bool netif_oper_up(const struct net_device *dev) 4492 { 4493 unsigned int operstate = READ_ONCE(dev->operstate); 4494 4495 return operstate == IF_OPER_UP || 4496 operstate == IF_OPER_UNKNOWN /* backward compat */; 4497 } 4498 4499 /** 4500 * netif_device_present - is device available or removed 4501 * @dev: network device 4502 * 4503 * Check if device has not been removed from system. 4504 */ 4505 static inline bool netif_device_present(const struct net_device *dev) 4506 { 4507 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4508 } 4509 4510 void netif_device_detach(struct net_device *dev); 4511 4512 void netif_device_attach(struct net_device *dev); 4513 4514 /* 4515 * Network interface message level settings 4516 */ 4517 4518 enum { 4519 NETIF_MSG_DRV_BIT, 4520 NETIF_MSG_PROBE_BIT, 4521 NETIF_MSG_LINK_BIT, 4522 NETIF_MSG_TIMER_BIT, 4523 NETIF_MSG_IFDOWN_BIT, 4524 NETIF_MSG_IFUP_BIT, 4525 NETIF_MSG_RX_ERR_BIT, 4526 NETIF_MSG_TX_ERR_BIT, 4527 NETIF_MSG_TX_QUEUED_BIT, 4528 NETIF_MSG_INTR_BIT, 4529 NETIF_MSG_TX_DONE_BIT, 4530 NETIF_MSG_RX_STATUS_BIT, 4531 NETIF_MSG_PKTDATA_BIT, 4532 NETIF_MSG_HW_BIT, 4533 NETIF_MSG_WOL_BIT, 4534 4535 /* When you add a new bit above, update netif_msg_class_names array 4536 * in net/ethtool/common.c 4537 */ 4538 NETIF_MSG_CLASS_COUNT, 4539 }; 4540 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4541 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4542 4543 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4544 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4545 4546 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4547 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4548 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4549 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4550 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4551 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4552 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4553 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4554 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4555 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4556 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4557 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4558 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4559 #define NETIF_MSG_HW __NETIF_MSG(HW) 4560 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4561 4562 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4563 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4564 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4565 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4566 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4567 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4568 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4569 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4570 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4571 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4572 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4573 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4574 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4575 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4576 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4577 4578 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4579 { 4580 /* use default */ 4581 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4582 return default_msg_enable_bits; 4583 if (debug_value == 0) /* no output */ 4584 return 0; 4585 /* set low N bits */ 4586 return (1U << debug_value) - 1; 4587 } 4588 4589 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4590 { 4591 spin_lock(&txq->_xmit_lock); 4592 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4593 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4594 } 4595 4596 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4597 { 4598 __acquire(&txq->_xmit_lock); 4599 return true; 4600 } 4601 4602 static inline void __netif_tx_release(struct netdev_queue *txq) 4603 { 4604 __release(&txq->_xmit_lock); 4605 } 4606 4607 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4608 { 4609 spin_lock_bh(&txq->_xmit_lock); 4610 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4611 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4612 } 4613 4614 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4615 { 4616 bool ok = spin_trylock(&txq->_xmit_lock); 4617 4618 if (likely(ok)) { 4619 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4620 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4621 } 4622 return ok; 4623 } 4624 4625 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4626 { 4627 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4628 WRITE_ONCE(txq->xmit_lock_owner, -1); 4629 spin_unlock(&txq->_xmit_lock); 4630 } 4631 4632 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4633 { 4634 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4635 WRITE_ONCE(txq->xmit_lock_owner, -1); 4636 spin_unlock_bh(&txq->_xmit_lock); 4637 } 4638 4639 /* 4640 * txq->trans_start can be read locklessly from dev_watchdog() 4641 */ 4642 static inline void txq_trans_update(struct netdev_queue *txq) 4643 { 4644 if (txq->xmit_lock_owner != -1) 4645 WRITE_ONCE(txq->trans_start, jiffies); 4646 } 4647 4648 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4649 { 4650 unsigned long now = jiffies; 4651 4652 if (READ_ONCE(txq->trans_start) != now) 4653 WRITE_ONCE(txq->trans_start, now); 4654 } 4655 4656 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4657 static inline void netif_trans_update(struct net_device *dev) 4658 { 4659 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4660 4661 txq_trans_cond_update(txq); 4662 } 4663 4664 /** 4665 * netif_tx_lock - grab network device transmit lock 4666 * @dev: network device 4667 * 4668 * Get network device transmit lock 4669 */ 4670 void netif_tx_lock(struct net_device *dev); 4671 4672 static inline void netif_tx_lock_bh(struct net_device *dev) 4673 { 4674 local_bh_disable(); 4675 netif_tx_lock(dev); 4676 } 4677 4678 void netif_tx_unlock(struct net_device *dev); 4679 4680 static inline void netif_tx_unlock_bh(struct net_device *dev) 4681 { 4682 netif_tx_unlock(dev); 4683 local_bh_enable(); 4684 } 4685 4686 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4687 if (!(dev)->lltx) { \ 4688 __netif_tx_lock(txq, cpu); \ 4689 } else { \ 4690 __netif_tx_acquire(txq); \ 4691 } \ 4692 } 4693 4694 #define HARD_TX_TRYLOCK(dev, txq) \ 4695 (!(dev)->lltx ? \ 4696 __netif_tx_trylock(txq) : \ 4697 __netif_tx_acquire(txq)) 4698 4699 #define HARD_TX_UNLOCK(dev, txq) { \ 4700 if (!(dev)->lltx) { \ 4701 __netif_tx_unlock(txq); \ 4702 } else { \ 4703 __netif_tx_release(txq); \ 4704 } \ 4705 } 4706 4707 static inline void netif_tx_disable(struct net_device *dev) 4708 { 4709 unsigned int i; 4710 int cpu; 4711 4712 local_bh_disable(); 4713 cpu = smp_processor_id(); 4714 spin_lock(&dev->tx_global_lock); 4715 for (i = 0; i < dev->num_tx_queues; i++) { 4716 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4717 4718 __netif_tx_lock(txq, cpu); 4719 netif_tx_stop_queue(txq); 4720 __netif_tx_unlock(txq); 4721 } 4722 spin_unlock(&dev->tx_global_lock); 4723 local_bh_enable(); 4724 } 4725 4726 static inline void netif_addr_lock(struct net_device *dev) 4727 { 4728 unsigned char nest_level = 0; 4729 4730 #ifdef CONFIG_LOCKDEP 4731 nest_level = dev->nested_level; 4732 #endif 4733 spin_lock_nested(&dev->addr_list_lock, nest_level); 4734 } 4735 4736 static inline void netif_addr_lock_bh(struct net_device *dev) 4737 { 4738 unsigned char nest_level = 0; 4739 4740 #ifdef CONFIG_LOCKDEP 4741 nest_level = dev->nested_level; 4742 #endif 4743 local_bh_disable(); 4744 spin_lock_nested(&dev->addr_list_lock, nest_level); 4745 } 4746 4747 static inline void netif_addr_unlock(struct net_device *dev) 4748 { 4749 spin_unlock(&dev->addr_list_lock); 4750 } 4751 4752 static inline void netif_addr_unlock_bh(struct net_device *dev) 4753 { 4754 spin_unlock_bh(&dev->addr_list_lock); 4755 } 4756 4757 /* 4758 * dev_addrs walker. Should be used only for read access. Call with 4759 * rcu_read_lock held. 4760 */ 4761 #define for_each_dev_addr(dev, ha) \ 4762 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4763 4764 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4765 4766 void ether_setup(struct net_device *dev); 4767 4768 /* Allocate dummy net_device */ 4769 struct net_device *alloc_netdev_dummy(int sizeof_priv); 4770 4771 /* Support for loadable net-drivers */ 4772 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4773 unsigned char name_assign_type, 4774 void (*setup)(struct net_device *), 4775 unsigned int txqs, unsigned int rxqs); 4776 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4777 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4778 4779 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4780 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4781 count) 4782 4783 int register_netdev(struct net_device *dev); 4784 void unregister_netdev(struct net_device *dev); 4785 4786 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4787 4788 /* General hardware address lists handling functions */ 4789 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4790 struct netdev_hw_addr_list *from_list, int addr_len); 4791 int __hw_addr_sync_multiple(struct netdev_hw_addr_list *to_list, 4792 struct netdev_hw_addr_list *from_list, 4793 int addr_len); 4794 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4795 struct netdev_hw_addr_list *from_list, int addr_len); 4796 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4797 struct net_device *dev, 4798 int (*sync)(struct net_device *, const unsigned char *), 4799 int (*unsync)(struct net_device *, 4800 const unsigned char *)); 4801 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4802 struct net_device *dev, 4803 int (*sync)(struct net_device *, 4804 const unsigned char *, int), 4805 int (*unsync)(struct net_device *, 4806 const unsigned char *, int)); 4807 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4808 struct net_device *dev, 4809 int (*unsync)(struct net_device *, 4810 const unsigned char *, int)); 4811 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4812 struct net_device *dev, 4813 int (*unsync)(struct net_device *, 4814 const unsigned char *)); 4815 void __hw_addr_init(struct netdev_hw_addr_list *list); 4816 4817 /* Functions used for device addresses handling */ 4818 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4819 const void *addr, size_t len); 4820 4821 static inline void 4822 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4823 { 4824 dev_addr_mod(dev, 0, addr, len); 4825 } 4826 4827 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4828 { 4829 __dev_addr_set(dev, addr, dev->addr_len); 4830 } 4831 4832 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4833 unsigned char addr_type); 4834 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4835 unsigned char addr_type); 4836 4837 /* Functions used for unicast addresses handling */ 4838 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4839 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4840 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4841 int dev_uc_sync(struct net_device *to, struct net_device *from); 4842 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4843 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4844 void dev_uc_flush(struct net_device *dev); 4845 void dev_uc_init(struct net_device *dev); 4846 4847 /** 4848 * __dev_uc_sync - Synchronize device's unicast list 4849 * @dev: device to sync 4850 * @sync: function to call if address should be added 4851 * @unsync: function to call if address should be removed 4852 * 4853 * Add newly added addresses to the interface, and release 4854 * addresses that have been deleted. 4855 */ 4856 static inline int __dev_uc_sync(struct net_device *dev, 4857 int (*sync)(struct net_device *, 4858 const unsigned char *), 4859 int (*unsync)(struct net_device *, 4860 const unsigned char *)) 4861 { 4862 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4863 } 4864 4865 /** 4866 * __dev_uc_unsync - Remove synchronized addresses from device 4867 * @dev: device to sync 4868 * @unsync: function to call if address should be removed 4869 * 4870 * Remove all addresses that were added to the device by dev_uc_sync(). 4871 */ 4872 static inline void __dev_uc_unsync(struct net_device *dev, 4873 int (*unsync)(struct net_device *, 4874 const unsigned char *)) 4875 { 4876 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4877 } 4878 4879 /* Functions used for multicast addresses handling */ 4880 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4881 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4882 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4883 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4884 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4885 int dev_mc_sync(struct net_device *to, struct net_device *from); 4886 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4887 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4888 void dev_mc_flush(struct net_device *dev); 4889 void dev_mc_init(struct net_device *dev); 4890 4891 /** 4892 * __dev_mc_sync - Synchronize device's multicast list 4893 * @dev: device to sync 4894 * @sync: function to call if address should be added 4895 * @unsync: function to call if address should be removed 4896 * 4897 * Add newly added addresses to the interface, and release 4898 * addresses that have been deleted. 4899 */ 4900 static inline int __dev_mc_sync(struct net_device *dev, 4901 int (*sync)(struct net_device *, 4902 const unsigned char *), 4903 int (*unsync)(struct net_device *, 4904 const unsigned char *)) 4905 { 4906 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4907 } 4908 4909 /** 4910 * __dev_mc_unsync - Remove synchronized addresses from device 4911 * @dev: device to sync 4912 * @unsync: function to call if address should be removed 4913 * 4914 * Remove all addresses that were added to the device by dev_mc_sync(). 4915 */ 4916 static inline void __dev_mc_unsync(struct net_device *dev, 4917 int (*unsync)(struct net_device *, 4918 const unsigned char *)) 4919 { 4920 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4921 } 4922 4923 /* Functions used for secondary unicast and multicast support */ 4924 void dev_set_rx_mode(struct net_device *dev); 4925 int dev_set_promiscuity(struct net_device *dev, int inc); 4926 int dev_set_allmulti(struct net_device *dev, int inc); 4927 void netdev_state_change(struct net_device *dev); 4928 void __netdev_notify_peers(struct net_device *dev); 4929 void netdev_notify_peers(struct net_device *dev); 4930 void netdev_features_change(struct net_device *dev); 4931 /* Load a device via the kmod */ 4932 void dev_load(struct net *net, const char *name); 4933 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4934 struct rtnl_link_stats64 *storage); 4935 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4936 const struct net_device_stats *netdev_stats); 4937 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4938 const struct pcpu_sw_netstats __percpu *netstats); 4939 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4940 4941 enum { 4942 NESTED_SYNC_IMM_BIT, 4943 NESTED_SYNC_TODO_BIT, 4944 }; 4945 4946 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4947 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4948 4949 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4950 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4951 4952 struct netdev_nested_priv { 4953 unsigned char flags; 4954 void *data; 4955 }; 4956 4957 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4958 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4959 struct list_head **iter); 4960 4961 /* iterate through upper list, must be called under RCU read lock */ 4962 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4963 for (iter = &(dev)->adj_list.upper, \ 4964 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4965 updev; \ 4966 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4967 4968 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4969 int (*fn)(struct net_device *upper_dev, 4970 struct netdev_nested_priv *priv), 4971 struct netdev_nested_priv *priv); 4972 4973 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4974 struct net_device *upper_dev); 4975 4976 bool netdev_has_any_upper_dev(struct net_device *dev); 4977 4978 void *netdev_lower_get_next_private(struct net_device *dev, 4979 struct list_head **iter); 4980 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4981 struct list_head **iter); 4982 4983 #define netdev_for_each_lower_private(dev, priv, iter) \ 4984 for (iter = (dev)->adj_list.lower.next, \ 4985 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4986 priv; \ 4987 priv = netdev_lower_get_next_private(dev, &(iter))) 4988 4989 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4990 for (iter = &(dev)->adj_list.lower, \ 4991 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4992 priv; \ 4993 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4994 4995 void *netdev_lower_get_next(struct net_device *dev, 4996 struct list_head **iter); 4997 4998 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4999 for (iter = (dev)->adj_list.lower.next, \ 5000 ldev = netdev_lower_get_next(dev, &(iter)); \ 5001 ldev; \ 5002 ldev = netdev_lower_get_next(dev, &(iter))) 5003 5004 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 5005 struct list_head **iter); 5006 int netdev_walk_all_lower_dev(struct net_device *dev, 5007 int (*fn)(struct net_device *lower_dev, 5008 struct netdev_nested_priv *priv), 5009 struct netdev_nested_priv *priv); 5010 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 5011 int (*fn)(struct net_device *lower_dev, 5012 struct netdev_nested_priv *priv), 5013 struct netdev_nested_priv *priv); 5014 5015 void *netdev_adjacent_get_private(struct list_head *adj_list); 5016 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 5017 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 5018 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 5019 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 5020 struct netlink_ext_ack *extack); 5021 int netdev_master_upper_dev_link(struct net_device *dev, 5022 struct net_device *upper_dev, 5023 void *upper_priv, void *upper_info, 5024 struct netlink_ext_ack *extack); 5025 void netdev_upper_dev_unlink(struct net_device *dev, 5026 struct net_device *upper_dev); 5027 int netdev_adjacent_change_prepare(struct net_device *old_dev, 5028 struct net_device *new_dev, 5029 struct net_device *dev, 5030 struct netlink_ext_ack *extack); 5031 void netdev_adjacent_change_commit(struct net_device *old_dev, 5032 struct net_device *new_dev, 5033 struct net_device *dev); 5034 void netdev_adjacent_change_abort(struct net_device *old_dev, 5035 struct net_device *new_dev, 5036 struct net_device *dev); 5037 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 5038 void *netdev_lower_dev_get_private(struct net_device *dev, 5039 struct net_device *lower_dev); 5040 void netdev_lower_state_changed(struct net_device *lower_dev, 5041 void *lower_state_info); 5042 5043 /* RSS keys are 40 or 52 bytes long */ 5044 #define NETDEV_RSS_KEY_LEN 52 5045 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 5046 void netdev_rss_key_fill(void *buffer, size_t len); 5047 5048 int skb_checksum_help(struct sk_buff *skb); 5049 int skb_crc32c_csum_help(struct sk_buff *skb); 5050 int skb_csum_hwoffload_help(struct sk_buff *skb, 5051 const netdev_features_t features); 5052 5053 struct netdev_bonding_info { 5054 ifslave slave; 5055 ifbond master; 5056 }; 5057 5058 struct netdev_notifier_bonding_info { 5059 struct netdev_notifier_info info; /* must be first */ 5060 struct netdev_bonding_info bonding_info; 5061 }; 5062 5063 void netdev_bonding_info_change(struct net_device *dev, 5064 struct netdev_bonding_info *bonding_info); 5065 5066 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 5067 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 5068 #else 5069 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 5070 const void *data) 5071 { 5072 } 5073 #endif 5074 5075 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 5076 5077 static inline bool can_checksum_protocol(netdev_features_t features, 5078 __be16 protocol) 5079 { 5080 if (protocol == htons(ETH_P_FCOE)) 5081 return !!(features & NETIF_F_FCOE_CRC); 5082 5083 /* Assume this is an IP checksum (not SCTP CRC) */ 5084 5085 if (features & NETIF_F_HW_CSUM) { 5086 /* Can checksum everything */ 5087 return true; 5088 } 5089 5090 switch (protocol) { 5091 case htons(ETH_P_IP): 5092 return !!(features & NETIF_F_IP_CSUM); 5093 case htons(ETH_P_IPV6): 5094 return !!(features & NETIF_F_IPV6_CSUM); 5095 default: 5096 return false; 5097 } 5098 } 5099 5100 #ifdef CONFIG_BUG 5101 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 5102 #else 5103 static inline void netdev_rx_csum_fault(struct net_device *dev, 5104 struct sk_buff *skb) 5105 { 5106 } 5107 #endif 5108 /* rx skb timestamps */ 5109 void net_enable_timestamp(void); 5110 void net_disable_timestamp(void); 5111 5112 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 5113 const struct skb_shared_hwtstamps *hwtstamps, 5114 bool cycles) 5115 { 5116 const struct net_device_ops *ops = dev->netdev_ops; 5117 5118 if (ops->ndo_get_tstamp) 5119 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 5120 5121 return hwtstamps->hwtstamp; 5122 } 5123 5124 #ifndef CONFIG_PREEMPT_RT 5125 static inline void netdev_xmit_set_more(bool more) 5126 { 5127 __this_cpu_write(softnet_data.xmit.more, more); 5128 } 5129 5130 static inline bool netdev_xmit_more(void) 5131 { 5132 return __this_cpu_read(softnet_data.xmit.more); 5133 } 5134 #else 5135 static inline void netdev_xmit_set_more(bool more) 5136 { 5137 current->net_xmit.more = more; 5138 } 5139 5140 static inline bool netdev_xmit_more(void) 5141 { 5142 return current->net_xmit.more; 5143 } 5144 #endif 5145 5146 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 5147 struct sk_buff *skb, struct net_device *dev, 5148 bool more) 5149 { 5150 netdev_xmit_set_more(more); 5151 return ops->ndo_start_xmit(skb, dev); 5152 } 5153 5154 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 5155 struct netdev_queue *txq, bool more) 5156 { 5157 const struct net_device_ops *ops = dev->netdev_ops; 5158 netdev_tx_t rc; 5159 5160 rc = __netdev_start_xmit(ops, skb, dev, more); 5161 if (rc == NETDEV_TX_OK) 5162 txq_trans_update(txq); 5163 5164 return rc; 5165 } 5166 5167 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 5168 const void *ns); 5169 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 5170 const void *ns); 5171 5172 extern const struct kobj_ns_type_operations net_ns_type_operations; 5173 5174 const char *netdev_drivername(const struct net_device *dev); 5175 5176 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 5177 netdev_features_t f2) 5178 { 5179 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 5180 if (f1 & NETIF_F_HW_CSUM) 5181 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5182 else 5183 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5184 } 5185 5186 return f1 & f2; 5187 } 5188 5189 static inline netdev_features_t netdev_get_wanted_features( 5190 struct net_device *dev) 5191 { 5192 return (dev->features & ~dev->hw_features) | dev->wanted_features; 5193 } 5194 netdev_features_t netdev_increment_features(netdev_features_t all, 5195 netdev_features_t one, netdev_features_t mask); 5196 5197 /* Allow TSO being used on stacked device : 5198 * Performing the GSO segmentation before last device 5199 * is a performance improvement. 5200 */ 5201 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 5202 netdev_features_t mask) 5203 { 5204 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 5205 } 5206 5207 int __netdev_update_features(struct net_device *dev); 5208 void netdev_update_features(struct net_device *dev); 5209 void netdev_change_features(struct net_device *dev); 5210 5211 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 5212 struct net_device *dev); 5213 5214 netdev_features_t passthru_features_check(struct sk_buff *skb, 5215 struct net_device *dev, 5216 netdev_features_t features); 5217 netdev_features_t netif_skb_features(struct sk_buff *skb); 5218 void skb_warn_bad_offload(const struct sk_buff *skb); 5219 5220 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 5221 { 5222 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 5223 5224 /* check flags correspondence */ 5225 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 5226 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 5227 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 5228 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 5229 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 5230 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 5231 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 5232 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 5233 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 5234 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 5235 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 5236 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 5237 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 5238 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 5239 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 5240 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 5241 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 5242 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 5243 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 5244 5245 return (features & feature) == feature; 5246 } 5247 5248 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 5249 { 5250 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 5251 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 5252 } 5253 5254 static inline bool netif_needs_gso(struct sk_buff *skb, 5255 netdev_features_t features) 5256 { 5257 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 5258 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 5259 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 5260 } 5261 5262 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 5263 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 5264 void netif_inherit_tso_max(struct net_device *to, 5265 const struct net_device *from); 5266 5267 static inline unsigned int 5268 netif_get_gro_max_size(const struct net_device *dev, const struct sk_buff *skb) 5269 { 5270 /* pairs with WRITE_ONCE() in netif_set_gro(_ipv4)_max_size() */ 5271 return skb->protocol == htons(ETH_P_IPV6) ? 5272 READ_ONCE(dev->gro_max_size) : 5273 READ_ONCE(dev->gro_ipv4_max_size); 5274 } 5275 5276 static inline unsigned int 5277 netif_get_gso_max_size(const struct net_device *dev, const struct sk_buff *skb) 5278 { 5279 /* pairs with WRITE_ONCE() in netif_set_gso(_ipv4)_max_size() */ 5280 return skb->protocol == htons(ETH_P_IPV6) ? 5281 READ_ONCE(dev->gso_max_size) : 5282 READ_ONCE(dev->gso_ipv4_max_size); 5283 } 5284 5285 static inline bool netif_is_macsec(const struct net_device *dev) 5286 { 5287 return dev->priv_flags & IFF_MACSEC; 5288 } 5289 5290 static inline bool netif_is_macvlan(const struct net_device *dev) 5291 { 5292 return dev->priv_flags & IFF_MACVLAN; 5293 } 5294 5295 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5296 { 5297 return dev->priv_flags & IFF_MACVLAN_PORT; 5298 } 5299 5300 static inline bool netif_is_bond_master(const struct net_device *dev) 5301 { 5302 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5303 } 5304 5305 static inline bool netif_is_bond_slave(const struct net_device *dev) 5306 { 5307 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5308 } 5309 5310 static inline bool netif_supports_nofcs(struct net_device *dev) 5311 { 5312 return dev->priv_flags & IFF_SUPP_NOFCS; 5313 } 5314 5315 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5316 { 5317 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5318 } 5319 5320 static inline bool netif_is_l3_master(const struct net_device *dev) 5321 { 5322 return dev->priv_flags & IFF_L3MDEV_MASTER; 5323 } 5324 5325 static inline bool netif_is_l3_slave(const struct net_device *dev) 5326 { 5327 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5328 } 5329 5330 static inline int dev_sdif(const struct net_device *dev) 5331 { 5332 #ifdef CONFIG_NET_L3_MASTER_DEV 5333 if (netif_is_l3_slave(dev)) 5334 return dev->ifindex; 5335 #endif 5336 return 0; 5337 } 5338 5339 static inline bool netif_is_bridge_master(const struct net_device *dev) 5340 { 5341 return dev->priv_flags & IFF_EBRIDGE; 5342 } 5343 5344 static inline bool netif_is_bridge_port(const struct net_device *dev) 5345 { 5346 return dev->priv_flags & IFF_BRIDGE_PORT; 5347 } 5348 5349 static inline bool netif_is_ovs_master(const struct net_device *dev) 5350 { 5351 return dev->priv_flags & IFF_OPENVSWITCH; 5352 } 5353 5354 static inline bool netif_is_ovs_port(const struct net_device *dev) 5355 { 5356 return dev->priv_flags & IFF_OVS_DATAPATH; 5357 } 5358 5359 static inline bool netif_is_any_bridge_master(const struct net_device *dev) 5360 { 5361 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev); 5362 } 5363 5364 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5365 { 5366 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5367 } 5368 5369 static inline bool netif_is_team_master(const struct net_device *dev) 5370 { 5371 return dev->priv_flags & IFF_TEAM; 5372 } 5373 5374 static inline bool netif_is_team_port(const struct net_device *dev) 5375 { 5376 return dev->priv_flags & IFF_TEAM_PORT; 5377 } 5378 5379 static inline bool netif_is_lag_master(const struct net_device *dev) 5380 { 5381 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5382 } 5383 5384 static inline bool netif_is_lag_port(const struct net_device *dev) 5385 { 5386 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5387 } 5388 5389 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5390 { 5391 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5392 } 5393 5394 static inline bool netif_is_failover(const struct net_device *dev) 5395 { 5396 return dev->priv_flags & IFF_FAILOVER; 5397 } 5398 5399 static inline bool netif_is_failover_slave(const struct net_device *dev) 5400 { 5401 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5402 } 5403 5404 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5405 static inline void netif_keep_dst(struct net_device *dev) 5406 { 5407 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5408 } 5409 5410 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5411 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5412 { 5413 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5414 return netif_is_macsec(dev); 5415 } 5416 5417 extern struct pernet_operations __net_initdata loopback_net_ops; 5418 5419 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5420 5421 /* netdev_printk helpers, similar to dev_printk */ 5422 5423 static inline const char *netdev_name(const struct net_device *dev) 5424 { 5425 if (!dev->name[0] || strchr(dev->name, '%')) 5426 return "(unnamed net_device)"; 5427 return dev->name; 5428 } 5429 5430 static inline const char *netdev_reg_state(const struct net_device *dev) 5431 { 5432 u8 reg_state = READ_ONCE(dev->reg_state); 5433 5434 switch (reg_state) { 5435 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5436 case NETREG_REGISTERED: return ""; 5437 case NETREG_UNREGISTERING: return " (unregistering)"; 5438 case NETREG_UNREGISTERED: return " (unregistered)"; 5439 case NETREG_RELEASED: return " (released)"; 5440 case NETREG_DUMMY: return " (dummy)"; 5441 } 5442 5443 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state); 5444 return " (unknown)"; 5445 } 5446 5447 #define MODULE_ALIAS_NETDEV(device) \ 5448 MODULE_ALIAS("netdev-" device) 5449 5450 /* 5451 * netdev_WARN() acts like dev_printk(), but with the key difference 5452 * of using a WARN/WARN_ON to get the message out, including the 5453 * file/line information and a backtrace. 5454 */ 5455 #define netdev_WARN(dev, format, args...) \ 5456 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5457 netdev_reg_state(dev), ##args) 5458 5459 #define netdev_WARN_ONCE(dev, format, args...) \ 5460 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5461 netdev_reg_state(dev), ##args) 5462 5463 /* 5464 * The list of packet types we will receive (as opposed to discard) 5465 * and the routines to invoke. 5466 * 5467 * Why 16. Because with 16 the only overlap we get on a hash of the 5468 * low nibble of the protocol value is RARP/SNAP/X.25. 5469 * 5470 * 0800 IP 5471 * 0001 802.3 5472 * 0002 AX.25 5473 * 0004 802.2 5474 * 8035 RARP 5475 * 0005 SNAP 5476 * 0805 X.25 5477 * 0806 ARP 5478 * 8137 IPX 5479 * 0009 Localtalk 5480 * 86DD IPv6 5481 */ 5482 #define PTYPE_HASH_SIZE (16) 5483 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5484 5485 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5486 5487 extern struct net_device *blackhole_netdev; 5488 5489 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */ 5490 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD) 5491 #define DEV_STATS_ADD(DEV, FIELD, VAL) \ 5492 atomic_long_add((VAL), &(DEV)->stats.__##FIELD) 5493 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD) 5494 5495 #endif /* _LINUX_NETDEVICE_H */ 5496