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 cpumask_t affinity_mask; 356 unsigned int napi_id; 357 }; 358 359 /* 360 * Structure for NAPI scheduling similar to tasklet but with weighting 361 */ 362 struct napi_struct { 363 /* The poll_list must only be managed by the entity which 364 * changes the state of the NAPI_STATE_SCHED bit. This means 365 * whoever atomically sets that bit can add this napi_struct 366 * to the per-CPU poll_list, and whoever clears that bit 367 * can remove from the list right before clearing the bit. 368 */ 369 struct list_head poll_list; 370 371 unsigned long state; 372 int weight; 373 u32 defer_hard_irqs_count; 374 unsigned long gro_bitmask; 375 int (*poll)(struct napi_struct *, int); 376 #ifdef CONFIG_NETPOLL 377 /* CPU actively polling if netpoll is configured */ 378 int poll_owner; 379 #endif 380 /* CPU on which NAPI has been scheduled for processing */ 381 int list_owner; 382 struct net_device *dev; 383 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 384 struct sk_buff *skb; 385 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 386 int rx_count; /* length of rx_list */ 387 unsigned int napi_id; /* protected by netdev_lock */ 388 struct hrtimer timer; 389 /* all fields past this point are write-protected by netdev_lock */ 390 struct task_struct *thread; 391 unsigned long gro_flush_timeout; 392 unsigned long irq_suspend_timeout; 393 u32 defer_hard_irqs; 394 /* control-path-only fields follow */ 395 struct list_head dev_list; 396 struct hlist_node napi_hash_node; 397 int irq; 398 struct irq_affinity_notify notify; 399 int napi_rmap_idx; 400 int index; 401 struct napi_config *config; 402 }; 403 404 enum { 405 NAPI_STATE_SCHED, /* Poll is scheduled */ 406 NAPI_STATE_MISSED, /* reschedule a napi */ 407 NAPI_STATE_DISABLE, /* Disable pending */ 408 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 409 NAPI_STATE_LISTED, /* NAPI added to system lists */ 410 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 411 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 412 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 413 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 414 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 415 NAPI_STATE_HAS_NOTIFIER, /* Napi has an IRQ notifier */ 416 }; 417 418 enum { 419 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 420 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 421 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 422 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 423 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 424 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 425 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 426 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 427 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 428 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 429 NAPIF_STATE_HAS_NOTIFIER = BIT(NAPI_STATE_HAS_NOTIFIER), 430 }; 431 432 enum gro_result { 433 GRO_MERGED, 434 GRO_MERGED_FREE, 435 GRO_HELD, 436 GRO_NORMAL, 437 GRO_CONSUMED, 438 }; 439 typedef enum gro_result gro_result_t; 440 441 /* 442 * enum rx_handler_result - Possible return values for rx_handlers. 443 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 444 * further. 445 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 446 * case skb->dev was changed by rx_handler. 447 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 448 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 449 * 450 * rx_handlers are functions called from inside __netif_receive_skb(), to do 451 * special processing of the skb, prior to delivery to protocol handlers. 452 * 453 * Currently, a net_device can only have a single rx_handler registered. Trying 454 * to register a second rx_handler will return -EBUSY. 455 * 456 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 457 * To unregister a rx_handler on a net_device, use 458 * netdev_rx_handler_unregister(). 459 * 460 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 461 * do with the skb. 462 * 463 * If the rx_handler consumed the skb in some way, it should return 464 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 465 * the skb to be delivered in some other way. 466 * 467 * If the rx_handler changed skb->dev, to divert the skb to another 468 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 469 * new device will be called if it exists. 470 * 471 * If the rx_handler decides the skb should be ignored, it should return 472 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 473 * are registered on exact device (ptype->dev == skb->dev). 474 * 475 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 476 * delivered, it should return RX_HANDLER_PASS. 477 * 478 * A device without a registered rx_handler will behave as if rx_handler 479 * returned RX_HANDLER_PASS. 480 */ 481 482 enum rx_handler_result { 483 RX_HANDLER_CONSUMED, 484 RX_HANDLER_ANOTHER, 485 RX_HANDLER_EXACT, 486 RX_HANDLER_PASS, 487 }; 488 typedef enum rx_handler_result rx_handler_result_t; 489 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 490 491 void __napi_schedule(struct napi_struct *n); 492 void __napi_schedule_irqoff(struct napi_struct *n); 493 494 static inline bool napi_disable_pending(struct napi_struct *n) 495 { 496 return test_bit(NAPI_STATE_DISABLE, &n->state); 497 } 498 499 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 500 { 501 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 502 } 503 504 /** 505 * napi_is_scheduled - test if NAPI is scheduled 506 * @n: NAPI context 507 * 508 * This check is "best-effort". With no locking implemented, 509 * a NAPI can be scheduled or terminate right after this check 510 * and produce not precise results. 511 * 512 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled 513 * should not be used normally and napi_schedule should be 514 * used instead. 515 * 516 * Use only if the driver really needs to check if a NAPI 517 * is scheduled for example in the context of delayed timer 518 * that can be skipped if a NAPI is already scheduled. 519 * 520 * Return: True if NAPI is scheduled, False otherwise. 521 */ 522 static inline bool napi_is_scheduled(struct napi_struct *n) 523 { 524 return test_bit(NAPI_STATE_SCHED, &n->state); 525 } 526 527 bool napi_schedule_prep(struct napi_struct *n); 528 529 /** 530 * napi_schedule - schedule NAPI poll 531 * @n: NAPI context 532 * 533 * Schedule NAPI poll routine to be called if it is not already 534 * running. 535 * Return: true if we schedule a NAPI or false if not. 536 * Refer to napi_schedule_prep() for additional reason on why 537 * a NAPI might not be scheduled. 538 */ 539 static inline bool napi_schedule(struct napi_struct *n) 540 { 541 if (napi_schedule_prep(n)) { 542 __napi_schedule(n); 543 return true; 544 } 545 546 return false; 547 } 548 549 /** 550 * napi_schedule_irqoff - schedule NAPI poll 551 * @n: NAPI context 552 * 553 * Variant of napi_schedule(), assuming hard irqs are masked. 554 */ 555 static inline void napi_schedule_irqoff(struct napi_struct *n) 556 { 557 if (napi_schedule_prep(n)) 558 __napi_schedule_irqoff(n); 559 } 560 561 /** 562 * napi_complete_done - NAPI processing complete 563 * @n: NAPI context 564 * @work_done: number of packets processed 565 * 566 * Mark NAPI processing as complete. Should only be called if poll budget 567 * has not been completely consumed. 568 * Prefer over napi_complete(). 569 * Return: false if device should avoid rearming interrupts. 570 */ 571 bool napi_complete_done(struct napi_struct *n, int work_done); 572 573 static inline bool napi_complete(struct napi_struct *n) 574 { 575 return napi_complete_done(n, 0); 576 } 577 578 int dev_set_threaded(struct net_device *dev, bool threaded); 579 580 void napi_disable(struct napi_struct *n); 581 void napi_disable_locked(struct napi_struct *n); 582 583 void napi_enable(struct napi_struct *n); 584 void napi_enable_locked(struct napi_struct *n); 585 586 /** 587 * napi_synchronize - wait until NAPI is not running 588 * @n: NAPI context 589 * 590 * Wait until NAPI is done being scheduled on this context. 591 * Waits till any outstanding processing completes but 592 * does not disable future activations. 593 */ 594 static inline void napi_synchronize(const struct napi_struct *n) 595 { 596 if (IS_ENABLED(CONFIG_SMP)) 597 while (test_bit(NAPI_STATE_SCHED, &n->state)) 598 msleep(1); 599 else 600 barrier(); 601 } 602 603 /** 604 * napi_if_scheduled_mark_missed - if napi is running, set the 605 * NAPIF_STATE_MISSED 606 * @n: NAPI context 607 * 608 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 609 * NAPI is scheduled. 610 **/ 611 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 612 { 613 unsigned long val, new; 614 615 val = READ_ONCE(n->state); 616 do { 617 if (val & NAPIF_STATE_DISABLE) 618 return true; 619 620 if (!(val & NAPIF_STATE_SCHED)) 621 return false; 622 623 new = val | NAPIF_STATE_MISSED; 624 } while (!try_cmpxchg(&n->state, &val, new)); 625 626 return true; 627 } 628 629 enum netdev_queue_state_t { 630 __QUEUE_STATE_DRV_XOFF, 631 __QUEUE_STATE_STACK_XOFF, 632 __QUEUE_STATE_FROZEN, 633 }; 634 635 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 636 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 637 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 638 639 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 640 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 641 QUEUE_STATE_FROZEN) 642 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 643 QUEUE_STATE_FROZEN) 644 645 /* 646 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 647 * netif_tx_* functions below are used to manipulate this flag. The 648 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 649 * queue independently. The netif_xmit_*stopped functions below are called 650 * to check if the queue has been stopped by the driver or stack (either 651 * of the XOFF bits are set in the state). Drivers should not need to call 652 * netif_xmit*stopped functions, they should only be using netif_tx_*. 653 */ 654 655 struct netdev_queue { 656 /* 657 * read-mostly part 658 */ 659 struct net_device *dev; 660 netdevice_tracker dev_tracker; 661 662 struct Qdisc __rcu *qdisc; 663 struct Qdisc __rcu *qdisc_sleeping; 664 #ifdef CONFIG_SYSFS 665 struct kobject kobj; 666 const struct attribute_group **groups; 667 #endif 668 unsigned long tx_maxrate; 669 /* 670 * Number of TX timeouts for this queue 671 * (/sys/class/net/DEV/Q/trans_timeout) 672 */ 673 atomic_long_t trans_timeout; 674 675 /* Subordinate device that the queue has been assigned to */ 676 struct net_device *sb_dev; 677 #ifdef CONFIG_XDP_SOCKETS 678 struct xsk_buff_pool *pool; 679 #endif 680 681 /* 682 * write-mostly part 683 */ 684 #ifdef CONFIG_BQL 685 struct dql dql; 686 #endif 687 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 688 int xmit_lock_owner; 689 /* 690 * Time (in jiffies) of last Tx 691 */ 692 unsigned long trans_start; 693 694 unsigned long state; 695 696 /* 697 * slow- / control-path part 698 */ 699 /* NAPI instance for the queue 700 * Readers and writers must hold RTNL 701 */ 702 struct napi_struct *napi; 703 704 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 705 int numa_node; 706 #endif 707 } ____cacheline_aligned_in_smp; 708 709 extern int sysctl_fb_tunnels_only_for_init_net; 710 extern int sysctl_devconf_inherit_init_net; 711 712 /* 713 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 714 * == 1 : For initns only 715 * == 2 : For none. 716 */ 717 static inline bool net_has_fallback_tunnels(const struct net *net) 718 { 719 #if IS_ENABLED(CONFIG_SYSCTL) 720 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net); 721 722 return !fb_tunnels_only_for_init_net || 723 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1); 724 #else 725 return true; 726 #endif 727 } 728 729 static inline int net_inherit_devconf(void) 730 { 731 #if IS_ENABLED(CONFIG_SYSCTL) 732 return READ_ONCE(sysctl_devconf_inherit_init_net); 733 #else 734 return 0; 735 #endif 736 } 737 738 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 739 { 740 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 741 return q->numa_node; 742 #else 743 return NUMA_NO_NODE; 744 #endif 745 } 746 747 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 748 { 749 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 750 q->numa_node = node; 751 #endif 752 } 753 754 #ifdef CONFIG_RFS_ACCEL 755 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 756 u16 filter_id); 757 #endif 758 759 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 760 enum xps_map_type { 761 XPS_CPUS = 0, 762 XPS_RXQS, 763 XPS_MAPS_MAX, 764 }; 765 766 #ifdef CONFIG_XPS 767 /* 768 * This structure holds an XPS map which can be of variable length. The 769 * map is an array of queues. 770 */ 771 struct xps_map { 772 unsigned int len; 773 unsigned int alloc_len; 774 struct rcu_head rcu; 775 u16 queues[]; 776 }; 777 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 778 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 779 - sizeof(struct xps_map)) / sizeof(u16)) 780 781 /* 782 * This structure holds all XPS maps for device. Maps are indexed by CPU. 783 * 784 * We keep track of the number of cpus/rxqs used when the struct is allocated, 785 * in nr_ids. This will help not accessing out-of-bound memory. 786 * 787 * We keep track of the number of traffic classes used when the struct is 788 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 789 * not crossing its upper bound, as the original dev->num_tc can be updated in 790 * the meantime. 791 */ 792 struct xps_dev_maps { 793 struct rcu_head rcu; 794 unsigned int nr_ids; 795 s16 num_tc; 796 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 797 }; 798 799 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 800 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 801 802 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 803 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 804 805 #endif /* CONFIG_XPS */ 806 807 #define TC_MAX_QUEUE 16 808 #define TC_BITMASK 15 809 /* HW offloaded queuing disciplines txq count and offset maps */ 810 struct netdev_tc_txq { 811 u16 count; 812 u16 offset; 813 }; 814 815 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 816 /* 817 * This structure is to hold information about the device 818 * configured to run FCoE protocol stack. 819 */ 820 struct netdev_fcoe_hbainfo { 821 char manufacturer[64]; 822 char serial_number[64]; 823 char hardware_version[64]; 824 char driver_version[64]; 825 char optionrom_version[64]; 826 char firmware_version[64]; 827 char model[256]; 828 char model_description[256]; 829 }; 830 #endif 831 832 #define MAX_PHYS_ITEM_ID_LEN 32 833 834 /* This structure holds a unique identifier to identify some 835 * physical item (port for example) used by a netdevice. 836 */ 837 struct netdev_phys_item_id { 838 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 839 unsigned char id_len; 840 }; 841 842 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 843 struct netdev_phys_item_id *b) 844 { 845 return a->id_len == b->id_len && 846 memcmp(a->id, b->id, a->id_len) == 0; 847 } 848 849 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 850 struct sk_buff *skb, 851 struct net_device *sb_dev); 852 853 enum net_device_path_type { 854 DEV_PATH_ETHERNET = 0, 855 DEV_PATH_VLAN, 856 DEV_PATH_BRIDGE, 857 DEV_PATH_PPPOE, 858 DEV_PATH_DSA, 859 DEV_PATH_MTK_WDMA, 860 }; 861 862 struct net_device_path { 863 enum net_device_path_type type; 864 const struct net_device *dev; 865 union { 866 struct { 867 u16 id; 868 __be16 proto; 869 u8 h_dest[ETH_ALEN]; 870 } encap; 871 struct { 872 enum { 873 DEV_PATH_BR_VLAN_KEEP, 874 DEV_PATH_BR_VLAN_TAG, 875 DEV_PATH_BR_VLAN_UNTAG, 876 DEV_PATH_BR_VLAN_UNTAG_HW, 877 } vlan_mode; 878 u16 vlan_id; 879 __be16 vlan_proto; 880 } bridge; 881 struct { 882 int port; 883 u16 proto; 884 } dsa; 885 struct { 886 u8 wdma_idx; 887 u8 queue; 888 u16 wcid; 889 u8 bss; 890 u8 amsdu; 891 } mtk_wdma; 892 }; 893 }; 894 895 #define NET_DEVICE_PATH_STACK_MAX 5 896 #define NET_DEVICE_PATH_VLAN_MAX 2 897 898 struct net_device_path_stack { 899 int num_paths; 900 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 901 }; 902 903 struct net_device_path_ctx { 904 const struct net_device *dev; 905 u8 daddr[ETH_ALEN]; 906 907 int num_vlans; 908 struct { 909 u16 id; 910 __be16 proto; 911 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 912 }; 913 914 enum tc_setup_type { 915 TC_QUERY_CAPS, 916 TC_SETUP_QDISC_MQPRIO, 917 TC_SETUP_CLSU32, 918 TC_SETUP_CLSFLOWER, 919 TC_SETUP_CLSMATCHALL, 920 TC_SETUP_CLSBPF, 921 TC_SETUP_BLOCK, 922 TC_SETUP_QDISC_CBS, 923 TC_SETUP_QDISC_RED, 924 TC_SETUP_QDISC_PRIO, 925 TC_SETUP_QDISC_MQ, 926 TC_SETUP_QDISC_ETF, 927 TC_SETUP_ROOT_QDISC, 928 TC_SETUP_QDISC_GRED, 929 TC_SETUP_QDISC_TAPRIO, 930 TC_SETUP_FT, 931 TC_SETUP_QDISC_ETS, 932 TC_SETUP_QDISC_TBF, 933 TC_SETUP_QDISC_FIFO, 934 TC_SETUP_QDISC_HTB, 935 TC_SETUP_ACT, 936 }; 937 938 /* These structures hold the attributes of bpf state that are being passed 939 * to the netdevice through the bpf op. 940 */ 941 enum bpf_netdev_command { 942 /* Set or clear a bpf program used in the earliest stages of packet 943 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 944 * is responsible for calling bpf_prog_put on any old progs that are 945 * stored. In case of error, the callee need not release the new prog 946 * reference, but on success it takes ownership and must bpf_prog_put 947 * when it is no longer used. 948 */ 949 XDP_SETUP_PROG, 950 XDP_SETUP_PROG_HW, 951 /* BPF program for offload callbacks, invoked at program load time. */ 952 BPF_OFFLOAD_MAP_ALLOC, 953 BPF_OFFLOAD_MAP_FREE, 954 XDP_SETUP_XSK_POOL, 955 }; 956 957 struct bpf_prog_offload_ops; 958 struct netlink_ext_ack; 959 struct xdp_umem; 960 struct xdp_dev_bulk_queue; 961 struct bpf_xdp_link; 962 963 enum bpf_xdp_mode { 964 XDP_MODE_SKB = 0, 965 XDP_MODE_DRV = 1, 966 XDP_MODE_HW = 2, 967 __MAX_XDP_MODE 968 }; 969 970 struct bpf_xdp_entity { 971 struct bpf_prog *prog; 972 struct bpf_xdp_link *link; 973 }; 974 975 struct netdev_bpf { 976 enum bpf_netdev_command command; 977 union { 978 /* XDP_SETUP_PROG */ 979 struct { 980 u32 flags; 981 struct bpf_prog *prog; 982 struct netlink_ext_ack *extack; 983 }; 984 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 985 struct { 986 struct bpf_offloaded_map *offmap; 987 }; 988 /* XDP_SETUP_XSK_POOL */ 989 struct { 990 struct xsk_buff_pool *pool; 991 u16 queue_id; 992 } xsk; 993 }; 994 }; 995 996 /* Flags for ndo_xsk_wakeup. */ 997 #define XDP_WAKEUP_RX (1 << 0) 998 #define XDP_WAKEUP_TX (1 << 1) 999 1000 #ifdef CONFIG_XFRM_OFFLOAD 1001 struct xfrmdev_ops { 1002 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack); 1003 void (*xdo_dev_state_delete) (struct xfrm_state *x); 1004 void (*xdo_dev_state_free) (struct xfrm_state *x); 1005 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 1006 struct xfrm_state *x); 1007 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1008 void (*xdo_dev_state_update_stats) (struct xfrm_state *x); 1009 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack); 1010 void (*xdo_dev_policy_delete) (struct xfrm_policy *x); 1011 void (*xdo_dev_policy_free) (struct xfrm_policy *x); 1012 }; 1013 #endif 1014 1015 struct dev_ifalias { 1016 struct rcu_head rcuhead; 1017 char ifalias[]; 1018 }; 1019 1020 struct devlink; 1021 struct tlsdev_ops; 1022 1023 struct netdev_net_notifier { 1024 struct list_head list; 1025 struct notifier_block *nb; 1026 }; 1027 1028 /* 1029 * This structure defines the management hooks for network devices. 1030 * The following hooks can be defined; unless noted otherwise, they are 1031 * optional and can be filled with a null pointer. 1032 * 1033 * int (*ndo_init)(struct net_device *dev); 1034 * This function is called once when a network device is registered. 1035 * The network device can use this for any late stage initialization 1036 * or semantic validation. It can fail with an error code which will 1037 * be propagated back to register_netdev. 1038 * 1039 * void (*ndo_uninit)(struct net_device *dev); 1040 * This function is called when device is unregistered or when registration 1041 * fails. It is not called if init fails. 1042 * 1043 * int (*ndo_open)(struct net_device *dev); 1044 * This function is called when a network device transitions to the up 1045 * state. 1046 * 1047 * int (*ndo_stop)(struct net_device *dev); 1048 * This function is called when a network device transitions to the down 1049 * state. 1050 * 1051 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1052 * struct net_device *dev); 1053 * Called when a packet needs to be transmitted. 1054 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1055 * the queue before that can happen; it's for obsolete devices and weird 1056 * corner cases, but the stack really does a non-trivial amount 1057 * of useless work if you return NETDEV_TX_BUSY. 1058 * Required; cannot be NULL. 1059 * 1060 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1061 * struct net_device *dev 1062 * netdev_features_t features); 1063 * Called by core transmit path to determine if device is capable of 1064 * performing offload operations on a given packet. This is to give 1065 * the device an opportunity to implement any restrictions that cannot 1066 * be otherwise expressed by feature flags. The check is called with 1067 * the set of features that the stack has calculated and it returns 1068 * those the driver believes to be appropriate. 1069 * 1070 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1071 * struct net_device *sb_dev); 1072 * Called to decide which queue to use when device supports multiple 1073 * transmit queues. 1074 * 1075 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1076 * This function is called to allow device receiver to make 1077 * changes to configuration when multicast or promiscuous is enabled. 1078 * 1079 * void (*ndo_set_rx_mode)(struct net_device *dev); 1080 * This function is called device changes address list filtering. 1081 * If driver handles unicast address filtering, it should set 1082 * IFF_UNICAST_FLT in its priv_flags. 1083 * 1084 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1085 * This function is called when the Media Access Control address 1086 * needs to be changed. If this interface is not defined, the 1087 * MAC address can not be changed. 1088 * 1089 * int (*ndo_validate_addr)(struct net_device *dev); 1090 * Test if Media Access Control address is valid for the device. 1091 * 1092 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1093 * Old-style ioctl entry point. This is used internally by the 1094 * ieee802154 subsystem but is no longer called by the device 1095 * ioctl handler. 1096 * 1097 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1098 * Used by the bonding driver for its device specific ioctls: 1099 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1100 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1101 * 1102 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1103 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1104 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1105 * 1106 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1107 * Used to set network devices bus interface parameters. This interface 1108 * is retained for legacy reasons; new devices should use the bus 1109 * interface (PCI) for low level management. 1110 * 1111 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1112 * Called when a user wants to change the Maximum Transfer Unit 1113 * of a device. 1114 * 1115 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1116 * Callback used when the transmitter has not made any progress 1117 * for dev->watchdog ticks. 1118 * 1119 * void (*ndo_get_stats64)(struct net_device *dev, 1120 * struct rtnl_link_stats64 *storage); 1121 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1122 * Called when a user wants to get the network device usage 1123 * statistics. Drivers must do one of the following: 1124 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1125 * rtnl_link_stats64 structure passed by the caller. 1126 * 2. Define @ndo_get_stats to update a net_device_stats structure 1127 * (which should normally be dev->stats) and return a pointer to 1128 * it. The structure may be changed asynchronously only if each 1129 * field is written atomically. 1130 * 3. Update dev->stats asynchronously and atomically, and define 1131 * neither operation. 1132 * 1133 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1134 * Return true if this device supports offload stats of this attr_id. 1135 * 1136 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1137 * void *attr_data) 1138 * Get statistics for offload operations by attr_id. Write it into the 1139 * attr_data pointer. 1140 * 1141 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1142 * If device supports VLAN filtering this function is called when a 1143 * VLAN id is registered. 1144 * 1145 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1146 * If device supports VLAN filtering this function is called when a 1147 * VLAN id is unregistered. 1148 * 1149 * void (*ndo_poll_controller)(struct net_device *dev); 1150 * 1151 * SR-IOV management functions. 1152 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1153 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1154 * u8 qos, __be16 proto); 1155 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1156 * int max_tx_rate); 1157 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1158 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1159 * int (*ndo_get_vf_config)(struct net_device *dev, 1160 * int vf, struct ifla_vf_info *ivf); 1161 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1162 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1163 * struct nlattr *port[]); 1164 * 1165 * Enable or disable the VF ability to query its RSS Redirection Table and 1166 * Hash Key. This is needed since on some devices VF share this information 1167 * with PF and querying it may introduce a theoretical security risk. 1168 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1169 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1170 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1171 * void *type_data); 1172 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1173 * This is always called from the stack with the rtnl lock held and netif 1174 * tx queues stopped. This allows the netdevice to perform queue 1175 * management safely. 1176 * 1177 * Fiber Channel over Ethernet (FCoE) offload functions. 1178 * int (*ndo_fcoe_enable)(struct net_device *dev); 1179 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1180 * so the underlying device can perform whatever needed configuration or 1181 * initialization to support acceleration of FCoE traffic. 1182 * 1183 * int (*ndo_fcoe_disable)(struct net_device *dev); 1184 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1185 * so the underlying device can perform whatever needed clean-ups to 1186 * stop supporting acceleration of FCoE traffic. 1187 * 1188 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1189 * struct scatterlist *sgl, unsigned int sgc); 1190 * Called when the FCoE Initiator wants to initialize an I/O that 1191 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1192 * perform necessary setup and returns 1 to indicate the device is set up 1193 * successfully to perform DDP on this I/O, otherwise this returns 0. 1194 * 1195 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1196 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1197 * indicated by the FC exchange id 'xid', so the underlying device can 1198 * clean up and reuse resources for later DDP requests. 1199 * 1200 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1201 * struct scatterlist *sgl, unsigned int sgc); 1202 * Called when the FCoE Target wants to initialize an I/O that 1203 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1204 * perform necessary setup and returns 1 to indicate the device is set up 1205 * successfully to perform DDP on this I/O, otherwise this returns 0. 1206 * 1207 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1208 * struct netdev_fcoe_hbainfo *hbainfo); 1209 * Called when the FCoE Protocol stack wants information on the underlying 1210 * device. This information is utilized by the FCoE protocol stack to 1211 * register attributes with Fiber Channel management service as per the 1212 * FC-GS Fabric Device Management Information(FDMI) specification. 1213 * 1214 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1215 * Called when the underlying device wants to override default World Wide 1216 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1217 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1218 * protocol stack to use. 1219 * 1220 * RFS acceleration. 1221 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1222 * u16 rxq_index, u32 flow_id); 1223 * Set hardware filter for RFS. rxq_index is the target queue index; 1224 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1225 * Return the filter ID on success, or a negative error code. 1226 * 1227 * Slave management functions (for bridge, bonding, etc). 1228 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1229 * Called to make another netdev an underling. 1230 * 1231 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1232 * Called to release previously enslaved netdev. 1233 * 1234 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1235 * struct sk_buff *skb, 1236 * bool all_slaves); 1237 * Get the xmit slave of master device. If all_slaves is true, function 1238 * assume all the slaves can transmit. 1239 * 1240 * Feature/offload setting functions. 1241 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1242 * netdev_features_t features); 1243 * Adjusts the requested feature flags according to device-specific 1244 * constraints, and returns the resulting flags. Must not modify 1245 * the device state. 1246 * 1247 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1248 * Called to update device configuration to new features. Passed 1249 * feature set might be less than what was returned by ndo_fix_features()). 1250 * Must return >0 or -errno if it changed dev->features itself. 1251 * 1252 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1253 * struct net_device *dev, 1254 * const unsigned char *addr, u16 vid, u16 flags, 1255 * bool *notified, struct netlink_ext_ack *extack); 1256 * Adds an FDB entry to dev for addr. 1257 * Callee shall set *notified to true if it sent any appropriate 1258 * notification(s). Otherwise core will send a generic one. 1259 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1260 * struct net_device *dev, 1261 * const unsigned char *addr, u16 vid 1262 * bool *notified, struct netlink_ext_ack *extack); 1263 * Deletes the FDB entry from dev corresponding to addr. 1264 * Callee shall set *notified to true if it sent any appropriate 1265 * notification(s). Otherwise core will send a generic one. 1266 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev, 1267 * struct netlink_ext_ack *extack); 1268 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1269 * struct net_device *dev, struct net_device *filter_dev, 1270 * int *idx) 1271 * Used to add FDB entries to dump requests. Implementers should add 1272 * entries to skb and update idx with the number of entries. 1273 * 1274 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[], 1275 * u16 nlmsg_flags, struct netlink_ext_ack *extack); 1276 * Adds an MDB entry to dev. 1277 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[], 1278 * struct netlink_ext_ack *extack); 1279 * Deletes the MDB entry from dev. 1280 * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[], 1281 * struct netlink_ext_ack *extack); 1282 * Bulk deletes MDB entries from dev. 1283 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb, 1284 * struct netlink_callback *cb); 1285 * Dumps MDB entries from dev. The first argument (marker) in the netlink 1286 * callback is used by core rtnetlink code. 1287 * 1288 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1289 * u16 flags, struct netlink_ext_ack *extack) 1290 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1291 * struct net_device *dev, u32 filter_mask, 1292 * int nlflags) 1293 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1294 * u16 flags); 1295 * 1296 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1297 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1298 * which do not represent real hardware may define this to allow their 1299 * userspace components to manage their virtual carrier state. Devices 1300 * that determine carrier state from physical hardware properties (eg 1301 * network cables) or protocol-dependent mechanisms (eg 1302 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1303 * 1304 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1305 * struct netdev_phys_item_id *ppid); 1306 * Called to get ID of physical port of this device. If driver does 1307 * not implement this, it is assumed that the hw is not able to have 1308 * multiple net devices on single physical port. 1309 * 1310 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1311 * struct netdev_phys_item_id *ppid) 1312 * Called to get the parent ID of the physical port of this device. 1313 * 1314 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1315 * struct net_device *dev) 1316 * Called by upper layer devices to accelerate switching or other 1317 * station functionality into hardware. 'pdev is the lowerdev 1318 * to use for the offload and 'dev' is the net device that will 1319 * back the offload. Returns a pointer to the private structure 1320 * the upper layer will maintain. 1321 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1322 * Called by upper layer device to delete the station created 1323 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1324 * the station and priv is the structure returned by the add 1325 * operation. 1326 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1327 * int queue_index, u32 maxrate); 1328 * Called when a user wants to set a max-rate limitation of specific 1329 * TX queue. 1330 * int (*ndo_get_iflink)(const struct net_device *dev); 1331 * Called to get the iflink value of this device. 1332 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1333 * This function is used to get egress tunnel information for given skb. 1334 * This is useful for retrieving outer tunnel header parameters while 1335 * sampling packet. 1336 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1337 * This function is used to specify the headroom that the skb must 1338 * consider when allocation skb during packet reception. Setting 1339 * appropriate rx headroom value allows avoiding skb head copy on 1340 * forward. Setting a negative value resets the rx headroom to the 1341 * default value. 1342 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1343 * This function is used to set or query state related to XDP on the 1344 * netdevice and manage BPF offload. See definition of 1345 * enum bpf_netdev_command for details. 1346 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1347 * u32 flags); 1348 * This function is used to submit @n XDP packets for transmit on a 1349 * netdevice. Returns number of frames successfully transmitted, frames 1350 * that got dropped are freed/returned via xdp_return_frame(). 1351 * Returns negative number, means general error invoking ndo, meaning 1352 * no frames were xmit'ed and core-caller will free all frames. 1353 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1354 * struct xdp_buff *xdp); 1355 * Get the xmit slave of master device based on the xdp_buff. 1356 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1357 * This function is used to wake up the softirq, ksoftirqd or kthread 1358 * responsible for sending and/or receiving packets on a specific 1359 * queue id bound to an AF_XDP socket. The flags field specifies if 1360 * only RX, only Tx, or both should be woken up using the flags 1361 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1362 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm_kern *p, 1363 * int cmd); 1364 * Add, change, delete or get information on an IPv4 tunnel. 1365 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1366 * If a device is paired with a peer device, return the peer instance. 1367 * The caller must be under RCU read context. 1368 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1369 * Get the forwarding path to reach the real device from the HW destination address 1370 * ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1371 * const struct skb_shared_hwtstamps *hwtstamps, 1372 * bool cycles); 1373 * Get hardware timestamp based on normal/adjustable time or free running 1374 * cycle counter. This function is required if physical clock supports a 1375 * free running cycle counter. 1376 * 1377 * int (*ndo_hwtstamp_get)(struct net_device *dev, 1378 * struct kernel_hwtstamp_config *kernel_config); 1379 * Get the currently configured hardware timestamping parameters for the 1380 * NIC device. 1381 * 1382 * int (*ndo_hwtstamp_set)(struct net_device *dev, 1383 * struct kernel_hwtstamp_config *kernel_config, 1384 * struct netlink_ext_ack *extack); 1385 * Change the hardware timestamping parameters for NIC device. 1386 */ 1387 struct net_device_ops { 1388 int (*ndo_init)(struct net_device *dev); 1389 void (*ndo_uninit)(struct net_device *dev); 1390 int (*ndo_open)(struct net_device *dev); 1391 int (*ndo_stop)(struct net_device *dev); 1392 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1393 struct net_device *dev); 1394 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1395 struct net_device *dev, 1396 netdev_features_t features); 1397 u16 (*ndo_select_queue)(struct net_device *dev, 1398 struct sk_buff *skb, 1399 struct net_device *sb_dev); 1400 void (*ndo_change_rx_flags)(struct net_device *dev, 1401 int flags); 1402 void (*ndo_set_rx_mode)(struct net_device *dev); 1403 int (*ndo_set_mac_address)(struct net_device *dev, 1404 void *addr); 1405 int (*ndo_validate_addr)(struct net_device *dev); 1406 int (*ndo_do_ioctl)(struct net_device *dev, 1407 struct ifreq *ifr, int cmd); 1408 int (*ndo_eth_ioctl)(struct net_device *dev, 1409 struct ifreq *ifr, int cmd); 1410 int (*ndo_siocbond)(struct net_device *dev, 1411 struct ifreq *ifr, int cmd); 1412 int (*ndo_siocwandev)(struct net_device *dev, 1413 struct if_settings *ifs); 1414 int (*ndo_siocdevprivate)(struct net_device *dev, 1415 struct ifreq *ifr, 1416 void __user *data, int cmd); 1417 int (*ndo_set_config)(struct net_device *dev, 1418 struct ifmap *map); 1419 int (*ndo_change_mtu)(struct net_device *dev, 1420 int new_mtu); 1421 int (*ndo_neigh_setup)(struct net_device *dev, 1422 struct neigh_parms *); 1423 void (*ndo_tx_timeout) (struct net_device *dev, 1424 unsigned int txqueue); 1425 1426 void (*ndo_get_stats64)(struct net_device *dev, 1427 struct rtnl_link_stats64 *storage); 1428 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1429 int (*ndo_get_offload_stats)(int attr_id, 1430 const struct net_device *dev, 1431 void *attr_data); 1432 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1433 1434 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1435 __be16 proto, u16 vid); 1436 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1437 __be16 proto, u16 vid); 1438 #ifdef CONFIG_NET_POLL_CONTROLLER 1439 void (*ndo_poll_controller)(struct net_device *dev); 1440 int (*ndo_netpoll_setup)(struct net_device *dev); 1441 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1442 #endif 1443 int (*ndo_set_vf_mac)(struct net_device *dev, 1444 int queue, u8 *mac); 1445 int (*ndo_set_vf_vlan)(struct net_device *dev, 1446 int queue, u16 vlan, 1447 u8 qos, __be16 proto); 1448 int (*ndo_set_vf_rate)(struct net_device *dev, 1449 int vf, int min_tx_rate, 1450 int max_tx_rate); 1451 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1452 int vf, bool setting); 1453 int (*ndo_set_vf_trust)(struct net_device *dev, 1454 int vf, bool setting); 1455 int (*ndo_get_vf_config)(struct net_device *dev, 1456 int vf, 1457 struct ifla_vf_info *ivf); 1458 int (*ndo_set_vf_link_state)(struct net_device *dev, 1459 int vf, int link_state); 1460 int (*ndo_get_vf_stats)(struct net_device *dev, 1461 int vf, 1462 struct ifla_vf_stats 1463 *vf_stats); 1464 int (*ndo_set_vf_port)(struct net_device *dev, 1465 int vf, 1466 struct nlattr *port[]); 1467 int (*ndo_get_vf_port)(struct net_device *dev, 1468 int vf, struct sk_buff *skb); 1469 int (*ndo_get_vf_guid)(struct net_device *dev, 1470 int vf, 1471 struct ifla_vf_guid *node_guid, 1472 struct ifla_vf_guid *port_guid); 1473 int (*ndo_set_vf_guid)(struct net_device *dev, 1474 int vf, u64 guid, 1475 int guid_type); 1476 int (*ndo_set_vf_rss_query_en)( 1477 struct net_device *dev, 1478 int vf, bool setting); 1479 int (*ndo_setup_tc)(struct net_device *dev, 1480 enum tc_setup_type type, 1481 void *type_data); 1482 #if IS_ENABLED(CONFIG_FCOE) 1483 int (*ndo_fcoe_enable)(struct net_device *dev); 1484 int (*ndo_fcoe_disable)(struct net_device *dev); 1485 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1486 u16 xid, 1487 struct scatterlist *sgl, 1488 unsigned int sgc); 1489 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1490 u16 xid); 1491 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1492 u16 xid, 1493 struct scatterlist *sgl, 1494 unsigned int sgc); 1495 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1496 struct netdev_fcoe_hbainfo *hbainfo); 1497 #endif 1498 1499 #if IS_ENABLED(CONFIG_LIBFCOE) 1500 #define NETDEV_FCOE_WWNN 0 1501 #define NETDEV_FCOE_WWPN 1 1502 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1503 u64 *wwn, int type); 1504 #endif 1505 1506 #ifdef CONFIG_RFS_ACCEL 1507 int (*ndo_rx_flow_steer)(struct net_device *dev, 1508 const struct sk_buff *skb, 1509 u16 rxq_index, 1510 u32 flow_id); 1511 #endif 1512 int (*ndo_add_slave)(struct net_device *dev, 1513 struct net_device *slave_dev, 1514 struct netlink_ext_ack *extack); 1515 int (*ndo_del_slave)(struct net_device *dev, 1516 struct net_device *slave_dev); 1517 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1518 struct sk_buff *skb, 1519 bool all_slaves); 1520 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1521 struct sock *sk); 1522 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1523 netdev_features_t features); 1524 int (*ndo_set_features)(struct net_device *dev, 1525 netdev_features_t features); 1526 int (*ndo_neigh_construct)(struct net_device *dev, 1527 struct neighbour *n); 1528 void (*ndo_neigh_destroy)(struct net_device *dev, 1529 struct neighbour *n); 1530 1531 int (*ndo_fdb_add)(struct ndmsg *ndm, 1532 struct nlattr *tb[], 1533 struct net_device *dev, 1534 const unsigned char *addr, 1535 u16 vid, 1536 u16 flags, 1537 bool *notified, 1538 struct netlink_ext_ack *extack); 1539 int (*ndo_fdb_del)(struct ndmsg *ndm, 1540 struct nlattr *tb[], 1541 struct net_device *dev, 1542 const unsigned char *addr, 1543 u16 vid, 1544 bool *notified, 1545 struct netlink_ext_ack *extack); 1546 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, 1547 struct net_device *dev, 1548 struct netlink_ext_ack *extack); 1549 int (*ndo_fdb_dump)(struct sk_buff *skb, 1550 struct netlink_callback *cb, 1551 struct net_device *dev, 1552 struct net_device *filter_dev, 1553 int *idx); 1554 int (*ndo_fdb_get)(struct sk_buff *skb, 1555 struct nlattr *tb[], 1556 struct net_device *dev, 1557 const unsigned char *addr, 1558 u16 vid, u32 portid, u32 seq, 1559 struct netlink_ext_ack *extack); 1560 int (*ndo_mdb_add)(struct net_device *dev, 1561 struct nlattr *tb[], 1562 u16 nlmsg_flags, 1563 struct netlink_ext_ack *extack); 1564 int (*ndo_mdb_del)(struct net_device *dev, 1565 struct nlattr *tb[], 1566 struct netlink_ext_ack *extack); 1567 int (*ndo_mdb_del_bulk)(struct net_device *dev, 1568 struct nlattr *tb[], 1569 struct netlink_ext_ack *extack); 1570 int (*ndo_mdb_dump)(struct net_device *dev, 1571 struct sk_buff *skb, 1572 struct netlink_callback *cb); 1573 int (*ndo_mdb_get)(struct net_device *dev, 1574 struct nlattr *tb[], u32 portid, 1575 u32 seq, 1576 struct netlink_ext_ack *extack); 1577 int (*ndo_bridge_setlink)(struct net_device *dev, 1578 struct nlmsghdr *nlh, 1579 u16 flags, 1580 struct netlink_ext_ack *extack); 1581 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1582 u32 pid, u32 seq, 1583 struct net_device *dev, 1584 u32 filter_mask, 1585 int nlflags); 1586 int (*ndo_bridge_dellink)(struct net_device *dev, 1587 struct nlmsghdr *nlh, 1588 u16 flags); 1589 int (*ndo_change_carrier)(struct net_device *dev, 1590 bool new_carrier); 1591 int (*ndo_get_phys_port_id)(struct net_device *dev, 1592 struct netdev_phys_item_id *ppid); 1593 int (*ndo_get_port_parent_id)(struct net_device *dev, 1594 struct netdev_phys_item_id *ppid); 1595 int (*ndo_get_phys_port_name)(struct net_device *dev, 1596 char *name, size_t len); 1597 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1598 struct net_device *dev); 1599 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1600 void *priv); 1601 1602 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1603 int queue_index, 1604 u32 maxrate); 1605 int (*ndo_get_iflink)(const struct net_device *dev); 1606 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1607 struct sk_buff *skb); 1608 void (*ndo_set_rx_headroom)(struct net_device *dev, 1609 int needed_headroom); 1610 int (*ndo_bpf)(struct net_device *dev, 1611 struct netdev_bpf *bpf); 1612 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1613 struct xdp_frame **xdp, 1614 u32 flags); 1615 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1616 struct xdp_buff *xdp); 1617 int (*ndo_xsk_wakeup)(struct net_device *dev, 1618 u32 queue_id, u32 flags); 1619 int (*ndo_tunnel_ctl)(struct net_device *dev, 1620 struct ip_tunnel_parm_kern *p, 1621 int cmd); 1622 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1623 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1624 struct net_device_path *path); 1625 ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1626 const struct skb_shared_hwtstamps *hwtstamps, 1627 bool cycles); 1628 int (*ndo_hwtstamp_get)(struct net_device *dev, 1629 struct kernel_hwtstamp_config *kernel_config); 1630 int (*ndo_hwtstamp_set)(struct net_device *dev, 1631 struct kernel_hwtstamp_config *kernel_config, 1632 struct netlink_ext_ack *extack); 1633 1634 #if IS_ENABLED(CONFIG_NET_SHAPER) 1635 /** 1636 * @net_shaper_ops: Device shaping offload operations 1637 * see include/net/net_shapers.h 1638 */ 1639 const struct net_shaper_ops *net_shaper_ops; 1640 #endif 1641 }; 1642 1643 /** 1644 * enum netdev_priv_flags - &struct net_device priv_flags 1645 * 1646 * These are the &struct net_device, they are only set internally 1647 * by drivers and used in the kernel. These flags are invisible to 1648 * userspace; this means that the order of these flags can change 1649 * during any kernel release. 1650 * 1651 * You should add bitfield booleans after either net_device::priv_flags 1652 * (hotpath) or ::threaded (slowpath) instead of extending these flags. 1653 * 1654 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1655 * @IFF_EBRIDGE: Ethernet bridging device 1656 * @IFF_BONDING: bonding master or slave 1657 * @IFF_ISATAP: ISATAP interface (RFC4214) 1658 * @IFF_WAN_HDLC: WAN HDLC device 1659 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1660 * release skb->dst 1661 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1662 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1663 * @IFF_MACVLAN_PORT: device used as macvlan port 1664 * @IFF_BRIDGE_PORT: device used as bridge port 1665 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1666 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1667 * @IFF_UNICAST_FLT: Supports unicast filtering 1668 * @IFF_TEAM_PORT: device used as team port 1669 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1670 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1671 * change when it's running 1672 * @IFF_MACVLAN: Macvlan device 1673 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1674 * underlying stacked devices 1675 * @IFF_L3MDEV_MASTER: device is an L3 master device 1676 * @IFF_NO_QUEUE: device can run without qdisc attached 1677 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1678 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1679 * @IFF_TEAM: device is a team device 1680 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1681 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1682 * entity (i.e. the master device for bridged veth) 1683 * @IFF_MACSEC: device is a MACsec device 1684 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1685 * @IFF_FAILOVER: device is a failover master device 1686 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1687 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1688 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf 1689 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1690 * skb_headlen(skb) == 0 (data starts from frag0) 1691 */ 1692 enum netdev_priv_flags { 1693 IFF_802_1Q_VLAN = 1<<0, 1694 IFF_EBRIDGE = 1<<1, 1695 IFF_BONDING = 1<<2, 1696 IFF_ISATAP = 1<<3, 1697 IFF_WAN_HDLC = 1<<4, 1698 IFF_XMIT_DST_RELEASE = 1<<5, 1699 IFF_DONT_BRIDGE = 1<<6, 1700 IFF_DISABLE_NETPOLL = 1<<7, 1701 IFF_MACVLAN_PORT = 1<<8, 1702 IFF_BRIDGE_PORT = 1<<9, 1703 IFF_OVS_DATAPATH = 1<<10, 1704 IFF_TX_SKB_SHARING = 1<<11, 1705 IFF_UNICAST_FLT = 1<<12, 1706 IFF_TEAM_PORT = 1<<13, 1707 IFF_SUPP_NOFCS = 1<<14, 1708 IFF_LIVE_ADDR_CHANGE = 1<<15, 1709 IFF_MACVLAN = 1<<16, 1710 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1711 IFF_L3MDEV_MASTER = 1<<18, 1712 IFF_NO_QUEUE = 1<<19, 1713 IFF_OPENVSWITCH = 1<<20, 1714 IFF_L3MDEV_SLAVE = 1<<21, 1715 IFF_TEAM = 1<<22, 1716 IFF_RXFH_CONFIGURED = 1<<23, 1717 IFF_PHONY_HEADROOM = 1<<24, 1718 IFF_MACSEC = 1<<25, 1719 IFF_NO_RX_HANDLER = 1<<26, 1720 IFF_FAILOVER = 1<<27, 1721 IFF_FAILOVER_SLAVE = 1<<28, 1722 IFF_L3MDEV_RX_HANDLER = 1<<29, 1723 IFF_NO_ADDRCONF = BIT_ULL(30), 1724 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31), 1725 }; 1726 1727 /* Specifies the type of the struct net_device::ml_priv pointer */ 1728 enum netdev_ml_priv_type { 1729 ML_PRIV_NONE, 1730 ML_PRIV_CAN, 1731 }; 1732 1733 enum netdev_stat_type { 1734 NETDEV_PCPU_STAT_NONE, 1735 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */ 1736 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */ 1737 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */ 1738 }; 1739 1740 enum netdev_reg_state { 1741 NETREG_UNINITIALIZED = 0, 1742 NETREG_REGISTERED, /* completed register_netdevice */ 1743 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1744 NETREG_UNREGISTERED, /* completed unregister todo */ 1745 NETREG_RELEASED, /* called free_netdev */ 1746 NETREG_DUMMY, /* dummy device for NAPI poll */ 1747 }; 1748 1749 /** 1750 * struct net_device - The DEVICE structure. 1751 * 1752 * Actually, this whole structure is a big mistake. It mixes I/O 1753 * data with strictly "high-level" data, and it has to know about 1754 * almost every data structure used in the INET module. 1755 * 1756 * @priv_flags: flags invisible to userspace defined as bits, see 1757 * enum netdev_priv_flags for the definitions 1758 * @lltx: device supports lockless Tx. Deprecated for real HW 1759 * drivers. Mainly used by logical interfaces, such as 1760 * bonding and tunnels 1761 * 1762 * @name: This is the first field of the "visible" part of this structure 1763 * (i.e. as seen by users in the "Space.c" file). It is the name 1764 * of the interface. 1765 * 1766 * @name_node: Name hashlist node 1767 * @ifalias: SNMP alias 1768 * @mem_end: Shared memory end 1769 * @mem_start: Shared memory start 1770 * @base_addr: Device I/O address 1771 * @irq: Device IRQ number 1772 * 1773 * @state: Generic network queuing layer state, see netdev_state_t 1774 * @dev_list: The global list of network devices 1775 * @napi_list: List entry used for polling NAPI devices 1776 * @unreg_list: List entry when we are unregistering the 1777 * device; see the function unregister_netdev 1778 * @close_list: List entry used when we are closing the device 1779 * @ptype_all: Device-specific packet handlers for all protocols 1780 * @ptype_specific: Device-specific, protocol-specific packet handlers 1781 * 1782 * @adj_list: Directly linked devices, like slaves for bonding 1783 * @features: Currently active device features 1784 * @hw_features: User-changeable features 1785 * 1786 * @wanted_features: User-requested features 1787 * @vlan_features: Mask of features inheritable by VLAN devices 1788 * 1789 * @hw_enc_features: Mask of features inherited by encapsulating devices 1790 * This field indicates what encapsulation 1791 * offloads the hardware is capable of doing, 1792 * and drivers will need to set them appropriately. 1793 * 1794 * @mpls_features: Mask of features inheritable by MPLS 1795 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1796 * 1797 * @ifindex: interface index 1798 * @group: The group the device belongs to 1799 * 1800 * @stats: Statistics struct, which was left as a legacy, use 1801 * rtnl_link_stats64 instead 1802 * 1803 * @core_stats: core networking counters, 1804 * do not use this in drivers 1805 * @carrier_up_count: Number of times the carrier has been up 1806 * @carrier_down_count: Number of times the carrier has been down 1807 * 1808 * @wireless_handlers: List of functions to handle Wireless Extensions, 1809 * instead of ioctl, 1810 * see <net/iw_handler.h> for details. 1811 * 1812 * @netdev_ops: Includes several pointers to callbacks, 1813 * if one wants to override the ndo_*() functions 1814 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks. 1815 * @xsk_tx_metadata_ops: Includes pointers to AF_XDP TX metadata callbacks. 1816 * @ethtool_ops: Management operations 1817 * @l3mdev_ops: Layer 3 master device operations 1818 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1819 * discovery handling. Necessary for e.g. 6LoWPAN. 1820 * @xfrmdev_ops: Transformation offload operations 1821 * @tlsdev_ops: Transport Layer Security offload operations 1822 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1823 * of Layer 2 headers. 1824 * 1825 * @flags: Interface flags (a la BSD) 1826 * @xdp_features: XDP capability supported by the device 1827 * @gflags: Global flags ( kept as legacy ) 1828 * @priv_len: Size of the ->priv flexible array 1829 * @priv: Flexible array containing private data 1830 * @operstate: RFC2863 operstate 1831 * @link_mode: Mapping policy to operstate 1832 * @if_port: Selectable AUI, TP, ... 1833 * @dma: DMA channel 1834 * @mtu: Interface MTU value 1835 * @min_mtu: Interface Minimum MTU value 1836 * @max_mtu: Interface Maximum MTU value 1837 * @type: Interface hardware type 1838 * @hard_header_len: Maximum hardware header length. 1839 * @min_header_len: Minimum hardware header length 1840 * 1841 * @needed_headroom: Extra headroom the hardware may need, but not in all 1842 * cases can this be guaranteed 1843 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1844 * cases can this be guaranteed. Some cases also use 1845 * LL_MAX_HEADER instead to allocate the skb 1846 * 1847 * interface address info: 1848 * 1849 * @perm_addr: Permanent hw address 1850 * @addr_assign_type: Hw address assignment type 1851 * @addr_len: Hardware address length 1852 * @upper_level: Maximum depth level of upper devices. 1853 * @lower_level: Maximum depth level of lower devices. 1854 * @neigh_priv_len: Used in neigh_alloc() 1855 * @dev_id: Used to differentiate devices that share 1856 * the same link layer address 1857 * @dev_port: Used to differentiate devices that share 1858 * the same function 1859 * @addr_list_lock: XXX: need comments on this one 1860 * @name_assign_type: network interface name assignment type 1861 * @uc_promisc: Counter that indicates promiscuous mode 1862 * has been enabled due to the need to listen to 1863 * additional unicast addresses in a device that 1864 * does not implement ndo_set_rx_mode() 1865 * @uc: unicast mac addresses 1866 * @mc: multicast mac addresses 1867 * @dev_addrs: list of device hw addresses 1868 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1869 * @promiscuity: Number of times the NIC is told to work in 1870 * promiscuous mode; if it becomes 0 the NIC will 1871 * exit promiscuous mode 1872 * @allmulti: Counter, enables or disables allmulticast mode 1873 * 1874 * @vlan_info: VLAN info 1875 * @dsa_ptr: dsa specific data 1876 * @tipc_ptr: TIPC specific data 1877 * @atalk_ptr: AppleTalk link 1878 * @ip_ptr: IPv4 specific data 1879 * @ip6_ptr: IPv6 specific data 1880 * @ax25_ptr: AX.25 specific data 1881 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1882 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1883 * device struct 1884 * @mpls_ptr: mpls_dev struct pointer 1885 * @mctp_ptr: MCTP specific data 1886 * 1887 * @dev_addr: Hw address (before bcast, 1888 * because most packets are unicast) 1889 * 1890 * @_rx: Array of RX queues 1891 * @num_rx_queues: Number of RX queues 1892 * allocated at register_netdev() time 1893 * @real_num_rx_queues: Number of RX queues currently active in device 1894 * @xdp_prog: XDP sockets filter program pointer 1895 * 1896 * @rx_handler: handler for received packets 1897 * @rx_handler_data: XXX: need comments on this one 1898 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing 1899 * @ingress_queue: XXX: need comments on this one 1900 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1901 * @broadcast: hw bcast address 1902 * 1903 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1904 * indexed by RX queue number. Assigned by driver. 1905 * This must only be set if the ndo_rx_flow_steer 1906 * operation is defined 1907 * @index_hlist: Device index hash chain 1908 * 1909 * @_tx: Array of TX queues 1910 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1911 * @real_num_tx_queues: Number of TX queues currently active in device 1912 * @qdisc: Root qdisc from userspace point of view 1913 * @tx_queue_len: Max frames per queue allowed 1914 * @tx_global_lock: XXX: need comments on this one 1915 * @xdp_bulkq: XDP device bulk queue 1916 * @xps_maps: all CPUs/RXQs maps for XPS device 1917 * 1918 * @xps_maps: XXX: need comments on this one 1919 * @tcx_egress: BPF & clsact qdisc specific data for egress processing 1920 * @nf_hooks_egress: netfilter hooks executed for egress packets 1921 * @qdisc_hash: qdisc hash table 1922 * @watchdog_timeo: Represents the timeout that is used by 1923 * the watchdog (see dev_watchdog()) 1924 * @watchdog_timer: List of timers 1925 * 1926 * @proto_down_reason: reason a netdev interface is held down 1927 * @pcpu_refcnt: Number of references to this device 1928 * @dev_refcnt: Number of references to this device 1929 * @refcnt_tracker: Tracker directory for tracked references to this device 1930 * @todo_list: Delayed register/unregister 1931 * @link_watch_list: XXX: need comments on this one 1932 * 1933 * @reg_state: Register/unregister state machine 1934 * @dismantle: Device is going to be freed 1935 * @rtnl_link_state: This enum represents the phases of creating 1936 * a new link 1937 * 1938 * @needs_free_netdev: Should unregister perform free_netdev? 1939 * @priv_destructor: Called from unregister 1940 * @npinfo: XXX: need comments on this one 1941 * @nd_net: Network namespace this network device is inside 1942 * 1943 * @ml_priv: Mid-layer private 1944 * @ml_priv_type: Mid-layer private type 1945 * 1946 * @pcpu_stat_type: Type of device statistics which the core should 1947 * allocate/free: none, lstats, tstats, dstats. none 1948 * means the driver is handling statistics allocation/ 1949 * freeing internally. 1950 * @lstats: Loopback statistics: packets, bytes 1951 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes 1952 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes 1953 * 1954 * @garp_port: GARP 1955 * @mrp_port: MRP 1956 * 1957 * @dm_private: Drop monitor private 1958 * 1959 * @dev: Class/net/name entry 1960 * @sysfs_groups: Space for optional device, statistics and wireless 1961 * sysfs groups 1962 * 1963 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1964 * @rtnl_link_ops: Rtnl_link_ops 1965 * @stat_ops: Optional ops for queue-aware statistics 1966 * @queue_mgmt_ops: Optional ops for queue management 1967 * 1968 * @gso_max_size: Maximum size of generic segmentation offload 1969 * @tso_max_size: Device (as in HW) limit on the max TSO request size 1970 * @gso_max_segs: Maximum number of segments that can be passed to the 1971 * NIC for GSO 1972 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count 1973 * @gso_ipv4_max_size: Maximum size of generic segmentation offload, 1974 * for IPv4. 1975 * 1976 * @dcbnl_ops: Data Center Bridging netlink ops 1977 * @num_tc: Number of traffic classes in the net device 1978 * @tc_to_txq: XXX: need comments on this one 1979 * @prio_tc_map: XXX: need comments on this one 1980 * 1981 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1982 * 1983 * @priomap: XXX: need comments on this one 1984 * @link_topo: Physical link topology tracking attached PHYs 1985 * @phydev: Physical device may attach itself 1986 * for hardware timestamping 1987 * @sfp_bus: attached &struct sfp_bus structure. 1988 * 1989 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1990 * 1991 * @proto_down: protocol port state information can be sent to the 1992 * switch driver and used to set the phys state of the 1993 * switch port. 1994 * 1995 * @threaded: napi threaded mode is enabled 1996 * 1997 * @irq_affinity_auto: driver wants the core to store and re-assign the IRQ 1998 * affinity. Set by netif_enable_irq_affinity(), then 1999 * the driver must create a persistent napi by 2000 * netif_napi_add_config() and finally bind the napi to 2001 * IRQ (via netif_napi_set_irq()). 2002 * 2003 * @rx_cpu_rmap_auto: driver wants the core to manage the ARFS rmap. 2004 * Set by calling netif_enable_cpu_rmap(). 2005 * 2006 * @see_all_hwtstamp_requests: device wants to see calls to 2007 * ndo_hwtstamp_set() for all timestamp requests 2008 * regardless of source, even if those aren't 2009 * HWTSTAMP_SOURCE_NETDEV 2010 * @change_proto_down: device supports setting carrier via IFLA_PROTO_DOWN 2011 * @netns_local: interface can't change network namespaces 2012 * @fcoe_mtu: device supports maximum FCoE MTU, 2158 bytes 2013 * 2014 * @net_notifier_list: List of per-net netdev notifier block 2015 * that follow this device when it is moved 2016 * to another network namespace. 2017 * 2018 * @macsec_ops: MACsec offloading ops 2019 * 2020 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 2021 * offload capabilities of the device 2022 * @udp_tunnel_nic: UDP tunnel offload state 2023 * @ethtool: ethtool related state 2024 * @xdp_state: stores info on attached XDP BPF programs 2025 * 2026 * @nested_level: Used as a parameter of spin_lock_nested() of 2027 * dev->addr_list_lock. 2028 * @unlink_list: As netif_addr_lock() can be called recursively, 2029 * keep a list of interfaces to be deleted. 2030 * @gro_max_size: Maximum size of aggregated packet in generic 2031 * receive offload (GRO) 2032 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic 2033 * receive offload (GRO), for IPv4. 2034 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP 2035 * zero copy driver 2036 * 2037 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 2038 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 2039 * @watchdog_dev_tracker: refcount tracker used by watchdog. 2040 * @dev_registered_tracker: tracker for reference held while 2041 * registered 2042 * @offload_xstats_l3: L3 HW stats for this netdevice. 2043 * 2044 * @devlink_port: Pointer to related devlink port structure. 2045 * Assigned by a driver before netdev registration using 2046 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static 2047 * during the time netdevice is registered. 2048 * 2049 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem, 2050 * where the clock is recovered. 2051 * 2052 * @max_pacing_offload_horizon: max EDT offload horizon in nsec. 2053 * @napi_config: An array of napi_config structures containing per-NAPI 2054 * settings. 2055 * @gro_flush_timeout: timeout for GRO layer in NAPI 2056 * @napi_defer_hard_irqs: If not zero, provides a counter that would 2057 * allow to avoid NIC hard IRQ, on busy queues. 2058 * 2059 * @neighbours: List heads pointing to this device's neighbours' 2060 * dev_list, one per address-family. 2061 * @hwprov: Tracks which PTP performs hardware packet time stamping. 2062 * 2063 * FIXME: cleanup struct net_device such that network protocol info 2064 * moves out. 2065 */ 2066 2067 struct net_device { 2068 /* Cacheline organization can be found documented in 2069 * Documentation/networking/net_cachelines/net_device.rst. 2070 * Please update the document when adding new fields. 2071 */ 2072 2073 /* TX read-mostly hotpath */ 2074 __cacheline_group_begin(net_device_read_tx); 2075 struct_group(priv_flags_fast, 2076 unsigned long priv_flags:32; 2077 unsigned long lltx:1; 2078 ); 2079 const struct net_device_ops *netdev_ops; 2080 const struct header_ops *header_ops; 2081 struct netdev_queue *_tx; 2082 netdev_features_t gso_partial_features; 2083 unsigned int real_num_tx_queues; 2084 unsigned int gso_max_size; 2085 unsigned int gso_ipv4_max_size; 2086 u16 gso_max_segs; 2087 s16 num_tc; 2088 /* Note : dev->mtu is often read without holding a lock. 2089 * Writers usually hold RTNL. 2090 * It is recommended to use READ_ONCE() to annotate the reads, 2091 * and to use WRITE_ONCE() to annotate the writes. 2092 */ 2093 unsigned int mtu; 2094 unsigned short needed_headroom; 2095 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2096 #ifdef CONFIG_XPS 2097 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2098 #endif 2099 #ifdef CONFIG_NETFILTER_EGRESS 2100 struct nf_hook_entries __rcu *nf_hooks_egress; 2101 #endif 2102 #ifdef CONFIG_NET_XGRESS 2103 struct bpf_mprog_entry __rcu *tcx_egress; 2104 #endif 2105 __cacheline_group_end(net_device_read_tx); 2106 2107 /* TXRX read-mostly hotpath */ 2108 __cacheline_group_begin(net_device_read_txrx); 2109 union { 2110 struct pcpu_lstats __percpu *lstats; 2111 struct pcpu_sw_netstats __percpu *tstats; 2112 struct pcpu_dstats __percpu *dstats; 2113 }; 2114 unsigned long state; 2115 unsigned int flags; 2116 unsigned short hard_header_len; 2117 netdev_features_t features; 2118 struct inet6_dev __rcu *ip6_ptr; 2119 __cacheline_group_end(net_device_read_txrx); 2120 2121 /* RX read-mostly hotpath */ 2122 __cacheline_group_begin(net_device_read_rx); 2123 struct bpf_prog __rcu *xdp_prog; 2124 struct list_head ptype_specific; 2125 int ifindex; 2126 unsigned int real_num_rx_queues; 2127 struct netdev_rx_queue *_rx; 2128 unsigned int gro_max_size; 2129 unsigned int gro_ipv4_max_size; 2130 rx_handler_func_t __rcu *rx_handler; 2131 void __rcu *rx_handler_data; 2132 possible_net_t nd_net; 2133 #ifdef CONFIG_NETPOLL 2134 struct netpoll_info __rcu *npinfo; 2135 #endif 2136 #ifdef CONFIG_NET_XGRESS 2137 struct bpf_mprog_entry __rcu *tcx_ingress; 2138 #endif 2139 __cacheline_group_end(net_device_read_rx); 2140 2141 char name[IFNAMSIZ]; 2142 struct netdev_name_node *name_node; 2143 struct dev_ifalias __rcu *ifalias; 2144 /* 2145 * I/O specific fields 2146 * FIXME: Merge these and struct ifmap into one 2147 */ 2148 unsigned long mem_end; 2149 unsigned long mem_start; 2150 unsigned long base_addr; 2151 2152 /* 2153 * Some hardware also needs these fields (state,dev_list, 2154 * napi_list,unreg_list,close_list) but they are not 2155 * part of the usual set specified in Space.c. 2156 */ 2157 2158 2159 struct list_head dev_list; 2160 struct list_head napi_list; 2161 struct list_head unreg_list; 2162 struct list_head close_list; 2163 struct list_head ptype_all; 2164 2165 struct { 2166 struct list_head upper; 2167 struct list_head lower; 2168 } adj_list; 2169 2170 /* Read-mostly cache-line for fast-path access */ 2171 xdp_features_t xdp_features; 2172 const struct xdp_metadata_ops *xdp_metadata_ops; 2173 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops; 2174 unsigned short gflags; 2175 2176 unsigned short needed_tailroom; 2177 2178 netdev_features_t hw_features; 2179 netdev_features_t wanted_features; 2180 netdev_features_t vlan_features; 2181 netdev_features_t hw_enc_features; 2182 netdev_features_t mpls_features; 2183 2184 unsigned int min_mtu; 2185 unsigned int max_mtu; 2186 unsigned short type; 2187 unsigned char min_header_len; 2188 unsigned char name_assign_type; 2189 2190 int group; 2191 2192 struct net_device_stats stats; /* not used by modern drivers */ 2193 2194 struct net_device_core_stats __percpu *core_stats; 2195 2196 /* Stats to monitor link on/off, flapping */ 2197 atomic_t carrier_up_count; 2198 atomic_t carrier_down_count; 2199 2200 #ifdef CONFIG_WIRELESS_EXT 2201 const struct iw_handler_def *wireless_handlers; 2202 #endif 2203 const struct ethtool_ops *ethtool_ops; 2204 #ifdef CONFIG_NET_L3_MASTER_DEV 2205 const struct l3mdev_ops *l3mdev_ops; 2206 #endif 2207 #if IS_ENABLED(CONFIG_IPV6) 2208 const struct ndisc_ops *ndisc_ops; 2209 #endif 2210 2211 #ifdef CONFIG_XFRM_OFFLOAD 2212 const struct xfrmdev_ops *xfrmdev_ops; 2213 #endif 2214 2215 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2216 const struct tlsdev_ops *tlsdev_ops; 2217 #endif 2218 2219 unsigned int operstate; 2220 unsigned char link_mode; 2221 2222 unsigned char if_port; 2223 unsigned char dma; 2224 2225 /* Interface address info. */ 2226 unsigned char perm_addr[MAX_ADDR_LEN]; 2227 unsigned char addr_assign_type; 2228 unsigned char addr_len; 2229 unsigned char upper_level; 2230 unsigned char lower_level; 2231 2232 unsigned short neigh_priv_len; 2233 unsigned short dev_id; 2234 unsigned short dev_port; 2235 int irq; 2236 u32 priv_len; 2237 2238 spinlock_t addr_list_lock; 2239 2240 struct netdev_hw_addr_list uc; 2241 struct netdev_hw_addr_list mc; 2242 struct netdev_hw_addr_list dev_addrs; 2243 2244 #ifdef CONFIG_SYSFS 2245 struct kset *queues_kset; 2246 #endif 2247 #ifdef CONFIG_LOCKDEP 2248 struct list_head unlink_list; 2249 #endif 2250 unsigned int promiscuity; 2251 unsigned int allmulti; 2252 bool uc_promisc; 2253 #ifdef CONFIG_LOCKDEP 2254 unsigned char nested_level; 2255 #endif 2256 2257 2258 /* Protocol-specific pointers */ 2259 struct in_device __rcu *ip_ptr; 2260 /** @fib_nh_head: nexthops associated with this netdev */ 2261 struct hlist_head fib_nh_head; 2262 2263 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2264 struct vlan_info __rcu *vlan_info; 2265 #endif 2266 #if IS_ENABLED(CONFIG_NET_DSA) 2267 struct dsa_port *dsa_ptr; 2268 #endif 2269 #if IS_ENABLED(CONFIG_TIPC) 2270 struct tipc_bearer __rcu *tipc_ptr; 2271 #endif 2272 #if IS_ENABLED(CONFIG_ATALK) 2273 void *atalk_ptr; 2274 #endif 2275 #if IS_ENABLED(CONFIG_AX25) 2276 struct ax25_dev __rcu *ax25_ptr; 2277 #endif 2278 #if IS_ENABLED(CONFIG_CFG80211) 2279 struct wireless_dev *ieee80211_ptr; 2280 #endif 2281 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2282 struct wpan_dev *ieee802154_ptr; 2283 #endif 2284 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2285 struct mpls_dev __rcu *mpls_ptr; 2286 #endif 2287 #if IS_ENABLED(CONFIG_MCTP) 2288 struct mctp_dev __rcu *mctp_ptr; 2289 #endif 2290 2291 /* 2292 * Cache lines mostly used on receive path (including eth_type_trans()) 2293 */ 2294 /* Interface address info used in eth_type_trans() */ 2295 const unsigned char *dev_addr; 2296 2297 unsigned int num_rx_queues; 2298 #define GRO_LEGACY_MAX_SIZE 65536u 2299 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2300 * and shinfo->gso_segs is a 16bit field. 2301 */ 2302 #define GRO_MAX_SIZE (8 * 65535u) 2303 unsigned int xdp_zc_max_segs; 2304 struct netdev_queue __rcu *ingress_queue; 2305 #ifdef CONFIG_NETFILTER_INGRESS 2306 struct nf_hook_entries __rcu *nf_hooks_ingress; 2307 #endif 2308 2309 unsigned char broadcast[MAX_ADDR_LEN]; 2310 #ifdef CONFIG_RFS_ACCEL 2311 struct cpu_rmap *rx_cpu_rmap; 2312 #endif 2313 struct hlist_node index_hlist; 2314 2315 /* 2316 * Cache lines mostly used on transmit path 2317 */ 2318 unsigned int num_tx_queues; 2319 struct Qdisc __rcu *qdisc; 2320 unsigned int tx_queue_len; 2321 spinlock_t tx_global_lock; 2322 2323 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2324 2325 #ifdef CONFIG_NET_SCHED 2326 DECLARE_HASHTABLE (qdisc_hash, 4); 2327 #endif 2328 /* These may be needed for future network-power-down code. */ 2329 struct timer_list watchdog_timer; 2330 int watchdog_timeo; 2331 2332 u32 proto_down_reason; 2333 2334 struct list_head todo_list; 2335 2336 #ifdef CONFIG_PCPU_DEV_REFCNT 2337 int __percpu *pcpu_refcnt; 2338 #else 2339 refcount_t dev_refcnt; 2340 #endif 2341 struct ref_tracker_dir refcnt_tracker; 2342 2343 struct list_head link_watch_list; 2344 2345 u8 reg_state; 2346 2347 bool dismantle; 2348 2349 enum { 2350 RTNL_LINK_INITIALIZED, 2351 RTNL_LINK_INITIALIZING, 2352 } rtnl_link_state:16; 2353 2354 bool needs_free_netdev; 2355 void (*priv_destructor)(struct net_device *dev); 2356 2357 /* mid-layer private */ 2358 void *ml_priv; 2359 enum netdev_ml_priv_type ml_priv_type; 2360 2361 enum netdev_stat_type pcpu_stat_type:8; 2362 2363 #if IS_ENABLED(CONFIG_GARP) 2364 struct garp_port __rcu *garp_port; 2365 #endif 2366 #if IS_ENABLED(CONFIG_MRP) 2367 struct mrp_port __rcu *mrp_port; 2368 #endif 2369 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2370 struct dm_hw_stat_delta __rcu *dm_private; 2371 #endif 2372 struct device dev; 2373 const struct attribute_group *sysfs_groups[4]; 2374 const struct attribute_group *sysfs_rx_queue_group; 2375 2376 const struct rtnl_link_ops *rtnl_link_ops; 2377 2378 const struct netdev_stat_ops *stat_ops; 2379 2380 const struct netdev_queue_mgmt_ops *queue_mgmt_ops; 2381 2382 /* for setting kernel sock attribute on TCP connection setup */ 2383 #define GSO_MAX_SEGS 65535u 2384 #define GSO_LEGACY_MAX_SIZE 65536u 2385 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2386 * and shinfo->gso_segs is a 16bit field. 2387 */ 2388 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2389 2390 #define TSO_LEGACY_MAX_SIZE 65536 2391 #define TSO_MAX_SIZE UINT_MAX 2392 unsigned int tso_max_size; 2393 #define TSO_MAX_SEGS U16_MAX 2394 u16 tso_max_segs; 2395 2396 #ifdef CONFIG_DCB 2397 const struct dcbnl_rtnl_ops *dcbnl_ops; 2398 #endif 2399 u8 prio_tc_map[TC_BITMASK + 1]; 2400 2401 #if IS_ENABLED(CONFIG_FCOE) 2402 unsigned int fcoe_ddp_xid; 2403 #endif 2404 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2405 struct netprio_map __rcu *priomap; 2406 #endif 2407 struct phy_link_topology *link_topo; 2408 struct phy_device *phydev; 2409 struct sfp_bus *sfp_bus; 2410 struct lock_class_key *qdisc_tx_busylock; 2411 bool proto_down; 2412 bool threaded; 2413 bool irq_affinity_auto; 2414 bool rx_cpu_rmap_auto; 2415 2416 /* priv_flags_slow, ungrouped to save space */ 2417 unsigned long see_all_hwtstamp_requests:1; 2418 unsigned long change_proto_down:1; 2419 unsigned long netns_local:1; 2420 unsigned long fcoe_mtu:1; 2421 2422 struct list_head net_notifier_list; 2423 2424 #if IS_ENABLED(CONFIG_MACSEC) 2425 /* MACsec management functions */ 2426 const struct macsec_ops *macsec_ops; 2427 #endif 2428 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2429 struct udp_tunnel_nic *udp_tunnel_nic; 2430 2431 /** @cfg: net_device queue-related configuration */ 2432 struct netdev_config *cfg; 2433 /** 2434 * @cfg_pending: same as @cfg but when device is being actively 2435 * reconfigured includes any changes to the configuration 2436 * requested by the user, but which may or may not be rejected. 2437 */ 2438 struct netdev_config *cfg_pending; 2439 struct ethtool_netdev_state *ethtool; 2440 2441 /* protected by rtnl_lock */ 2442 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2443 2444 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2445 netdevice_tracker linkwatch_dev_tracker; 2446 netdevice_tracker watchdog_dev_tracker; 2447 netdevice_tracker dev_registered_tracker; 2448 struct rtnl_hw_stats64 *offload_xstats_l3; 2449 2450 struct devlink_port *devlink_port; 2451 2452 #if IS_ENABLED(CONFIG_DPLL) 2453 struct dpll_pin __rcu *dpll_pin; 2454 #endif 2455 #if IS_ENABLED(CONFIG_PAGE_POOL) 2456 /** @page_pools: page pools created for this netdevice */ 2457 struct hlist_head page_pools; 2458 #endif 2459 2460 /** @irq_moder: dim parameters used if IS_ENABLED(CONFIG_DIMLIB). */ 2461 struct dim_irq_moder *irq_moder; 2462 2463 u64 max_pacing_offload_horizon; 2464 struct napi_config *napi_config; 2465 unsigned long gro_flush_timeout; 2466 u32 napi_defer_hard_irqs; 2467 2468 /** 2469 * @up: copy of @state's IFF_UP, but safe to read with just @lock. 2470 * May report false negatives while the device is being opened 2471 * or closed (@lock does not protect .ndo_open, or .ndo_close). 2472 */ 2473 bool up; 2474 2475 /** 2476 * @lock: netdev-scope lock, protects a small selection of fields. 2477 * Should always be taken using netdev_lock() / netdev_unlock() helpers. 2478 * Drivers are free to use it for other protection. 2479 * 2480 * Protects: 2481 * @gro_flush_timeout, @napi_defer_hard_irqs, @napi_list, 2482 * @net_shaper_hierarchy, @reg_state, @threaded 2483 * 2484 * Partially protects (writers must hold both @lock and rtnl_lock): 2485 * @up 2486 * 2487 * Also protects some fields in struct napi_struct. 2488 * 2489 * Ordering: take after rtnl_lock. 2490 */ 2491 struct mutex lock; 2492 2493 #if IS_ENABLED(CONFIG_NET_SHAPER) 2494 /** 2495 * @net_shaper_hierarchy: data tracking the current shaper status 2496 * see include/net/net_shapers.h 2497 */ 2498 struct net_shaper_hierarchy *net_shaper_hierarchy; 2499 #endif 2500 2501 struct hlist_head neighbours[NEIGH_NR_TABLES]; 2502 2503 struct hwtstamp_provider __rcu *hwprov; 2504 2505 u8 priv[] ____cacheline_aligned 2506 __counted_by(priv_len); 2507 } ____cacheline_aligned; 2508 #define to_net_dev(d) container_of(d, struct net_device, dev) 2509 2510 /* 2511 * Driver should use this to assign devlink port instance to a netdevice 2512 * before it registers the netdevice. Therefore devlink_port is static 2513 * during the netdev lifetime after it is registered. 2514 */ 2515 #define SET_NETDEV_DEVLINK_PORT(dev, port) \ 2516 ({ \ 2517 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \ 2518 ((dev)->devlink_port = (port)); \ 2519 }) 2520 2521 static inline bool netif_elide_gro(const struct net_device *dev) 2522 { 2523 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2524 return true; 2525 return false; 2526 } 2527 2528 #define NETDEV_ALIGN 32 2529 2530 static inline 2531 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2532 { 2533 return dev->prio_tc_map[prio & TC_BITMASK]; 2534 } 2535 2536 static inline 2537 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2538 { 2539 if (tc >= dev->num_tc) 2540 return -EINVAL; 2541 2542 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2543 return 0; 2544 } 2545 2546 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2547 void netdev_reset_tc(struct net_device *dev); 2548 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2549 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2550 2551 static inline 2552 int netdev_get_num_tc(struct net_device *dev) 2553 { 2554 return dev->num_tc; 2555 } 2556 2557 static inline void net_prefetch(void *p) 2558 { 2559 prefetch(p); 2560 #if L1_CACHE_BYTES < 128 2561 prefetch((u8 *)p + L1_CACHE_BYTES); 2562 #endif 2563 } 2564 2565 static inline void net_prefetchw(void *p) 2566 { 2567 prefetchw(p); 2568 #if L1_CACHE_BYTES < 128 2569 prefetchw((u8 *)p + L1_CACHE_BYTES); 2570 #endif 2571 } 2572 2573 void netdev_unbind_sb_channel(struct net_device *dev, 2574 struct net_device *sb_dev); 2575 int netdev_bind_sb_channel_queue(struct net_device *dev, 2576 struct net_device *sb_dev, 2577 u8 tc, u16 count, u16 offset); 2578 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2579 static inline int netdev_get_sb_channel(struct net_device *dev) 2580 { 2581 return max_t(int, -dev->num_tc, 0); 2582 } 2583 2584 static inline 2585 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2586 unsigned int index) 2587 { 2588 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues); 2589 return &dev->_tx[index]; 2590 } 2591 2592 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2593 const struct sk_buff *skb) 2594 { 2595 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2596 } 2597 2598 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2599 void (*f)(struct net_device *, 2600 struct netdev_queue *, 2601 void *), 2602 void *arg) 2603 { 2604 unsigned int i; 2605 2606 for (i = 0; i < dev->num_tx_queues; i++) 2607 f(dev, &dev->_tx[i], arg); 2608 } 2609 2610 #define netdev_lockdep_set_classes(dev) \ 2611 { \ 2612 static struct lock_class_key qdisc_tx_busylock_key; \ 2613 static struct lock_class_key qdisc_xmit_lock_key; \ 2614 static struct lock_class_key dev_addr_list_lock_key; \ 2615 unsigned int i; \ 2616 \ 2617 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2618 lockdep_set_class(&(dev)->addr_list_lock, \ 2619 &dev_addr_list_lock_key); \ 2620 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2621 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2622 &qdisc_xmit_lock_key); \ 2623 } 2624 2625 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2626 struct net_device *sb_dev); 2627 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2628 struct sk_buff *skb, 2629 struct net_device *sb_dev); 2630 2631 /* returns the headroom that the master device needs to take in account 2632 * when forwarding to this dev 2633 */ 2634 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2635 { 2636 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2637 } 2638 2639 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2640 { 2641 if (dev->netdev_ops->ndo_set_rx_headroom) 2642 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2643 } 2644 2645 /* set the device rx headroom to the dev's default */ 2646 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2647 { 2648 netdev_set_rx_headroom(dev, -1); 2649 } 2650 2651 static inline void *netdev_get_ml_priv(struct net_device *dev, 2652 enum netdev_ml_priv_type type) 2653 { 2654 if (dev->ml_priv_type != type) 2655 return NULL; 2656 2657 return dev->ml_priv; 2658 } 2659 2660 static inline void netdev_set_ml_priv(struct net_device *dev, 2661 void *ml_priv, 2662 enum netdev_ml_priv_type type) 2663 { 2664 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2665 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2666 dev->ml_priv_type, type); 2667 WARN(!dev->ml_priv_type && dev->ml_priv, 2668 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2669 2670 dev->ml_priv = ml_priv; 2671 dev->ml_priv_type = type; 2672 } 2673 2674 /* 2675 * Net namespace inlines 2676 */ 2677 static inline 2678 struct net *dev_net(const struct net_device *dev) 2679 { 2680 return read_pnet(&dev->nd_net); 2681 } 2682 2683 static inline 2684 struct net *dev_net_rcu(const struct net_device *dev) 2685 { 2686 return read_pnet_rcu(&dev->nd_net); 2687 } 2688 2689 static inline 2690 void dev_net_set(struct net_device *dev, struct net *net) 2691 { 2692 write_pnet(&dev->nd_net, net); 2693 } 2694 2695 /** 2696 * netdev_priv - access network device private data 2697 * @dev: network device 2698 * 2699 * Get network device private data 2700 */ 2701 static inline void *netdev_priv(const struct net_device *dev) 2702 { 2703 return (void *)dev->priv; 2704 } 2705 2706 /* Set the sysfs physical device reference for the network logical device 2707 * if set prior to registration will cause a symlink during initialization. 2708 */ 2709 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2710 2711 /* Set the sysfs device type for the network logical device to allow 2712 * fine-grained identification of different network device types. For 2713 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2714 */ 2715 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2716 2717 void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index, 2718 enum netdev_queue_type type, 2719 struct napi_struct *napi); 2720 2721 static inline void netdev_lock(struct net_device *dev) 2722 { 2723 mutex_lock(&dev->lock); 2724 } 2725 2726 static inline void netdev_unlock(struct net_device *dev) 2727 { 2728 mutex_unlock(&dev->lock); 2729 } 2730 2731 static inline void netdev_assert_locked(struct net_device *dev) 2732 { 2733 lockdep_assert_held(&dev->lock); 2734 } 2735 2736 static inline void netdev_assert_locked_or_invisible(struct net_device *dev) 2737 { 2738 if (dev->reg_state == NETREG_REGISTERED || 2739 dev->reg_state == NETREG_UNREGISTERING) 2740 netdev_assert_locked(dev); 2741 } 2742 2743 void netif_napi_set_irq_locked(struct napi_struct *napi, int irq); 2744 2745 static inline void netif_napi_set_irq(struct napi_struct *napi, int irq) 2746 { 2747 netdev_lock(napi->dev); 2748 netif_napi_set_irq_locked(napi, irq); 2749 netdev_unlock(napi->dev); 2750 } 2751 2752 /* Default NAPI poll() weight 2753 * Device drivers are strongly advised to not use bigger value 2754 */ 2755 #define NAPI_POLL_WEIGHT 64 2756 2757 void netif_napi_add_weight_locked(struct net_device *dev, 2758 struct napi_struct *napi, 2759 int (*poll)(struct napi_struct *, int), 2760 int weight); 2761 2762 static inline void 2763 netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2764 int (*poll)(struct napi_struct *, int), int weight) 2765 { 2766 netdev_lock(dev); 2767 netif_napi_add_weight_locked(dev, napi, poll, weight); 2768 netdev_unlock(dev); 2769 } 2770 2771 /** 2772 * netif_napi_add() - initialize a NAPI context 2773 * @dev: network device 2774 * @napi: NAPI context 2775 * @poll: polling function 2776 * 2777 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2778 * *any* of the other NAPI-related functions. 2779 */ 2780 static inline void 2781 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2782 int (*poll)(struct napi_struct *, int)) 2783 { 2784 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2785 } 2786 2787 static inline void 2788 netif_napi_add_locked(struct net_device *dev, struct napi_struct *napi, 2789 int (*poll)(struct napi_struct *, int)) 2790 { 2791 netif_napi_add_weight_locked(dev, napi, poll, NAPI_POLL_WEIGHT); 2792 } 2793 2794 static inline void 2795 netif_napi_add_tx_weight(struct net_device *dev, 2796 struct napi_struct *napi, 2797 int (*poll)(struct napi_struct *, int), 2798 int weight) 2799 { 2800 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2801 netif_napi_add_weight(dev, napi, poll, weight); 2802 } 2803 2804 static inline void 2805 netif_napi_add_config_locked(struct net_device *dev, struct napi_struct *napi, 2806 int (*poll)(struct napi_struct *, int), int index) 2807 { 2808 napi->index = index; 2809 napi->config = &dev->napi_config[index]; 2810 netif_napi_add_weight_locked(dev, napi, poll, NAPI_POLL_WEIGHT); 2811 } 2812 2813 /** 2814 * netif_napi_add_config - initialize a NAPI context with persistent config 2815 * @dev: network device 2816 * @napi: NAPI context 2817 * @poll: polling function 2818 * @index: the NAPI index 2819 */ 2820 static inline void 2821 netif_napi_add_config(struct net_device *dev, struct napi_struct *napi, 2822 int (*poll)(struct napi_struct *, int), int index) 2823 { 2824 netdev_lock(dev); 2825 netif_napi_add_config_locked(dev, napi, poll, index); 2826 netdev_unlock(dev); 2827 } 2828 2829 /** 2830 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2831 * @dev: network device 2832 * @napi: NAPI context 2833 * @poll: polling function 2834 * 2835 * This variant of netif_napi_add() should be used from drivers using NAPI 2836 * to exclusively poll a TX queue. 2837 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2838 */ 2839 static inline void netif_napi_add_tx(struct net_device *dev, 2840 struct napi_struct *napi, 2841 int (*poll)(struct napi_struct *, int)) 2842 { 2843 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2844 } 2845 2846 void __netif_napi_del_locked(struct napi_struct *napi); 2847 2848 /** 2849 * __netif_napi_del - remove a NAPI context 2850 * @napi: NAPI context 2851 * 2852 * Warning: caller must observe RCU grace period before freeing memory 2853 * containing @napi. Drivers might want to call this helper to combine 2854 * all the needed RCU grace periods into a single one. 2855 */ 2856 static inline void __netif_napi_del(struct napi_struct *napi) 2857 { 2858 netdev_lock(napi->dev); 2859 __netif_napi_del_locked(napi); 2860 netdev_unlock(napi->dev); 2861 } 2862 2863 static inline void netif_napi_del_locked(struct napi_struct *napi) 2864 { 2865 __netif_napi_del_locked(napi); 2866 synchronize_net(); 2867 } 2868 2869 /** 2870 * netif_napi_del - remove a NAPI context 2871 * @napi: NAPI context 2872 * 2873 * netif_napi_del() removes a NAPI context from the network device NAPI list 2874 */ 2875 static inline void netif_napi_del(struct napi_struct *napi) 2876 { 2877 __netif_napi_del(napi); 2878 synchronize_net(); 2879 } 2880 2881 int netif_enable_cpu_rmap(struct net_device *dev, unsigned int num_irqs); 2882 void netif_set_affinity_auto(struct net_device *dev); 2883 2884 struct packet_type { 2885 __be16 type; /* This is really htons(ether_type). */ 2886 bool ignore_outgoing; 2887 struct net_device *dev; /* NULL is wildcarded here */ 2888 netdevice_tracker dev_tracker; 2889 int (*func) (struct sk_buff *, 2890 struct net_device *, 2891 struct packet_type *, 2892 struct net_device *); 2893 void (*list_func) (struct list_head *, 2894 struct packet_type *, 2895 struct net_device *); 2896 bool (*id_match)(struct packet_type *ptype, 2897 struct sock *sk); 2898 struct net *af_packet_net; 2899 void *af_packet_priv; 2900 struct list_head list; 2901 }; 2902 2903 struct offload_callbacks { 2904 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2905 netdev_features_t features); 2906 struct sk_buff *(*gro_receive)(struct list_head *head, 2907 struct sk_buff *skb); 2908 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2909 }; 2910 2911 struct packet_offload { 2912 __be16 type; /* This is really htons(ether_type). */ 2913 u16 priority; 2914 struct offload_callbacks callbacks; 2915 struct list_head list; 2916 }; 2917 2918 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2919 struct pcpu_sw_netstats { 2920 u64_stats_t rx_packets; 2921 u64_stats_t rx_bytes; 2922 u64_stats_t tx_packets; 2923 u64_stats_t tx_bytes; 2924 struct u64_stats_sync syncp; 2925 } __aligned(4 * sizeof(u64)); 2926 2927 struct pcpu_dstats { 2928 u64_stats_t rx_packets; 2929 u64_stats_t rx_bytes; 2930 u64_stats_t tx_packets; 2931 u64_stats_t tx_bytes; 2932 u64_stats_t rx_drops; 2933 u64_stats_t tx_drops; 2934 struct u64_stats_sync syncp; 2935 } __aligned(8 * sizeof(u64)); 2936 2937 struct pcpu_lstats { 2938 u64_stats_t packets; 2939 u64_stats_t bytes; 2940 struct u64_stats_sync syncp; 2941 } __aligned(2 * sizeof(u64)); 2942 2943 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2944 2945 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2946 { 2947 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2948 2949 u64_stats_update_begin(&tstats->syncp); 2950 u64_stats_add(&tstats->rx_bytes, len); 2951 u64_stats_inc(&tstats->rx_packets); 2952 u64_stats_update_end(&tstats->syncp); 2953 } 2954 2955 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2956 unsigned int packets, 2957 unsigned int len) 2958 { 2959 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2960 2961 u64_stats_update_begin(&tstats->syncp); 2962 u64_stats_add(&tstats->tx_bytes, len); 2963 u64_stats_add(&tstats->tx_packets, packets); 2964 u64_stats_update_end(&tstats->syncp); 2965 } 2966 2967 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2968 { 2969 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2970 2971 u64_stats_update_begin(&lstats->syncp); 2972 u64_stats_add(&lstats->bytes, len); 2973 u64_stats_inc(&lstats->packets); 2974 u64_stats_update_end(&lstats->syncp); 2975 } 2976 2977 static inline void dev_dstats_rx_add(struct net_device *dev, 2978 unsigned int len) 2979 { 2980 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 2981 2982 u64_stats_update_begin(&dstats->syncp); 2983 u64_stats_inc(&dstats->rx_packets); 2984 u64_stats_add(&dstats->rx_bytes, len); 2985 u64_stats_update_end(&dstats->syncp); 2986 } 2987 2988 static inline void dev_dstats_rx_dropped(struct net_device *dev) 2989 { 2990 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 2991 2992 u64_stats_update_begin(&dstats->syncp); 2993 u64_stats_inc(&dstats->rx_drops); 2994 u64_stats_update_end(&dstats->syncp); 2995 } 2996 2997 static inline void dev_dstats_tx_add(struct net_device *dev, 2998 unsigned int len) 2999 { 3000 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3001 3002 u64_stats_update_begin(&dstats->syncp); 3003 u64_stats_inc(&dstats->tx_packets); 3004 u64_stats_add(&dstats->tx_bytes, len); 3005 u64_stats_update_end(&dstats->syncp); 3006 } 3007 3008 static inline void dev_dstats_tx_dropped(struct net_device *dev) 3009 { 3010 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3011 3012 u64_stats_update_begin(&dstats->syncp); 3013 u64_stats_inc(&dstats->tx_drops); 3014 u64_stats_update_end(&dstats->syncp); 3015 } 3016 3017 #define __netdev_alloc_pcpu_stats(type, gfp) \ 3018 ({ \ 3019 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 3020 if (pcpu_stats) { \ 3021 int __cpu; \ 3022 for_each_possible_cpu(__cpu) { \ 3023 typeof(type) *stat; \ 3024 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 3025 u64_stats_init(&stat->syncp); \ 3026 } \ 3027 } \ 3028 pcpu_stats; \ 3029 }) 3030 3031 #define netdev_alloc_pcpu_stats(type) \ 3032 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 3033 3034 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 3035 ({ \ 3036 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 3037 if (pcpu_stats) { \ 3038 int __cpu; \ 3039 for_each_possible_cpu(__cpu) { \ 3040 typeof(type) *stat; \ 3041 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 3042 u64_stats_init(&stat->syncp); \ 3043 } \ 3044 } \ 3045 pcpu_stats; \ 3046 }) 3047 3048 enum netdev_lag_tx_type { 3049 NETDEV_LAG_TX_TYPE_UNKNOWN, 3050 NETDEV_LAG_TX_TYPE_RANDOM, 3051 NETDEV_LAG_TX_TYPE_BROADCAST, 3052 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 3053 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 3054 NETDEV_LAG_TX_TYPE_HASH, 3055 }; 3056 3057 enum netdev_lag_hash { 3058 NETDEV_LAG_HASH_NONE, 3059 NETDEV_LAG_HASH_L2, 3060 NETDEV_LAG_HASH_L34, 3061 NETDEV_LAG_HASH_L23, 3062 NETDEV_LAG_HASH_E23, 3063 NETDEV_LAG_HASH_E34, 3064 NETDEV_LAG_HASH_VLAN_SRCMAC, 3065 NETDEV_LAG_HASH_UNKNOWN, 3066 }; 3067 3068 struct netdev_lag_upper_info { 3069 enum netdev_lag_tx_type tx_type; 3070 enum netdev_lag_hash hash_type; 3071 }; 3072 3073 struct netdev_lag_lower_state_info { 3074 u8 link_up : 1, 3075 tx_enabled : 1; 3076 }; 3077 3078 #include <linux/notifier.h> 3079 3080 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 3081 * and the rtnetlink notification exclusion list in rtnetlink_event() when 3082 * adding new types. 3083 */ 3084 enum netdev_cmd { 3085 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 3086 NETDEV_DOWN, 3087 NETDEV_REBOOT, /* Tell a protocol stack a network interface 3088 detected a hardware crash and restarted 3089 - we can use this eg to kick tcp sessions 3090 once done */ 3091 NETDEV_CHANGE, /* Notify device state change */ 3092 NETDEV_REGISTER, 3093 NETDEV_UNREGISTER, 3094 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 3095 NETDEV_CHANGEADDR, /* notify after the address change */ 3096 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 3097 NETDEV_GOING_DOWN, 3098 NETDEV_CHANGENAME, 3099 NETDEV_FEAT_CHANGE, 3100 NETDEV_BONDING_FAILOVER, 3101 NETDEV_PRE_UP, 3102 NETDEV_PRE_TYPE_CHANGE, 3103 NETDEV_POST_TYPE_CHANGE, 3104 NETDEV_POST_INIT, 3105 NETDEV_PRE_UNINIT, 3106 NETDEV_RELEASE, 3107 NETDEV_NOTIFY_PEERS, 3108 NETDEV_JOIN, 3109 NETDEV_CHANGEUPPER, 3110 NETDEV_RESEND_IGMP, 3111 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 3112 NETDEV_CHANGEINFODATA, 3113 NETDEV_BONDING_INFO, 3114 NETDEV_PRECHANGEUPPER, 3115 NETDEV_CHANGELOWERSTATE, 3116 NETDEV_UDP_TUNNEL_PUSH_INFO, 3117 NETDEV_UDP_TUNNEL_DROP_INFO, 3118 NETDEV_CHANGE_TX_QUEUE_LEN, 3119 NETDEV_CVLAN_FILTER_PUSH_INFO, 3120 NETDEV_CVLAN_FILTER_DROP_INFO, 3121 NETDEV_SVLAN_FILTER_PUSH_INFO, 3122 NETDEV_SVLAN_FILTER_DROP_INFO, 3123 NETDEV_OFFLOAD_XSTATS_ENABLE, 3124 NETDEV_OFFLOAD_XSTATS_DISABLE, 3125 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 3126 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 3127 NETDEV_XDP_FEAT_CHANGE, 3128 }; 3129 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 3130 3131 int register_netdevice_notifier(struct notifier_block *nb); 3132 int unregister_netdevice_notifier(struct notifier_block *nb); 3133 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 3134 int unregister_netdevice_notifier_net(struct net *net, 3135 struct notifier_block *nb); 3136 int register_netdevice_notifier_dev_net(struct net_device *dev, 3137 struct notifier_block *nb, 3138 struct netdev_net_notifier *nn); 3139 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 3140 struct notifier_block *nb, 3141 struct netdev_net_notifier *nn); 3142 3143 struct netdev_notifier_info { 3144 struct net_device *dev; 3145 struct netlink_ext_ack *extack; 3146 }; 3147 3148 struct netdev_notifier_info_ext { 3149 struct netdev_notifier_info info; /* must be first */ 3150 union { 3151 u32 mtu; 3152 } ext; 3153 }; 3154 3155 struct netdev_notifier_change_info { 3156 struct netdev_notifier_info info; /* must be first */ 3157 unsigned int flags_changed; 3158 }; 3159 3160 struct netdev_notifier_changeupper_info { 3161 struct netdev_notifier_info info; /* must be first */ 3162 struct net_device *upper_dev; /* new upper dev */ 3163 bool master; /* is upper dev master */ 3164 bool linking; /* is the notification for link or unlink */ 3165 void *upper_info; /* upper dev info */ 3166 }; 3167 3168 struct netdev_notifier_changelowerstate_info { 3169 struct netdev_notifier_info info; /* must be first */ 3170 void *lower_state_info; /* is lower dev state */ 3171 }; 3172 3173 struct netdev_notifier_pre_changeaddr_info { 3174 struct netdev_notifier_info info; /* must be first */ 3175 const unsigned char *dev_addr; 3176 }; 3177 3178 enum netdev_offload_xstats_type { 3179 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 3180 }; 3181 3182 struct netdev_notifier_offload_xstats_info { 3183 struct netdev_notifier_info info; /* must be first */ 3184 enum netdev_offload_xstats_type type; 3185 3186 union { 3187 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 3188 struct netdev_notifier_offload_xstats_rd *report_delta; 3189 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 3190 struct netdev_notifier_offload_xstats_ru *report_used; 3191 }; 3192 }; 3193 3194 int netdev_offload_xstats_enable(struct net_device *dev, 3195 enum netdev_offload_xstats_type type, 3196 struct netlink_ext_ack *extack); 3197 int netdev_offload_xstats_disable(struct net_device *dev, 3198 enum netdev_offload_xstats_type type); 3199 bool netdev_offload_xstats_enabled(const struct net_device *dev, 3200 enum netdev_offload_xstats_type type); 3201 int netdev_offload_xstats_get(struct net_device *dev, 3202 enum netdev_offload_xstats_type type, 3203 struct rtnl_hw_stats64 *stats, bool *used, 3204 struct netlink_ext_ack *extack); 3205 void 3206 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 3207 const struct rtnl_hw_stats64 *stats); 3208 void 3209 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 3210 void netdev_offload_xstats_push_delta(struct net_device *dev, 3211 enum netdev_offload_xstats_type type, 3212 const struct rtnl_hw_stats64 *stats); 3213 3214 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 3215 struct net_device *dev) 3216 { 3217 info->dev = dev; 3218 info->extack = NULL; 3219 } 3220 3221 static inline struct net_device * 3222 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 3223 { 3224 return info->dev; 3225 } 3226 3227 static inline struct netlink_ext_ack * 3228 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 3229 { 3230 return info->extack; 3231 } 3232 3233 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 3234 int call_netdevice_notifiers_info(unsigned long val, 3235 struct netdev_notifier_info *info); 3236 3237 #define for_each_netdev(net, d) \ 3238 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 3239 #define for_each_netdev_reverse(net, d) \ 3240 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 3241 #define for_each_netdev_rcu(net, d) \ 3242 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 3243 #define for_each_netdev_safe(net, d, n) \ 3244 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 3245 #define for_each_netdev_continue(net, d) \ 3246 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 3247 #define for_each_netdev_continue_reverse(net, d) \ 3248 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 3249 dev_list) 3250 #define for_each_netdev_continue_rcu(net, d) \ 3251 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 3252 #define for_each_netdev_in_bond_rcu(bond, slave) \ 3253 for_each_netdev_rcu(&init_net, slave) \ 3254 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 3255 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 3256 3257 #define for_each_netdev_dump(net, d, ifindex) \ 3258 for (; (d = xa_find(&(net)->dev_by_index, &ifindex, \ 3259 ULONG_MAX, XA_PRESENT)); ifindex++) 3260 3261 static inline struct net_device *next_net_device(struct net_device *dev) 3262 { 3263 struct list_head *lh; 3264 struct net *net; 3265 3266 net = dev_net(dev); 3267 lh = dev->dev_list.next; 3268 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3269 } 3270 3271 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 3272 { 3273 struct list_head *lh; 3274 struct net *net; 3275 3276 net = dev_net(dev); 3277 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 3278 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3279 } 3280 3281 static inline struct net_device *first_net_device(struct net *net) 3282 { 3283 return list_empty(&net->dev_base_head) ? NULL : 3284 net_device_entry(net->dev_base_head.next); 3285 } 3286 3287 static inline struct net_device *first_net_device_rcu(struct net *net) 3288 { 3289 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 3290 3291 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3292 } 3293 3294 int netdev_boot_setup_check(struct net_device *dev); 3295 struct net_device *dev_getbyhwaddr(struct net *net, unsigned short type, 3296 const char *hwaddr); 3297 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 3298 const char *hwaddr); 3299 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 3300 void dev_add_pack(struct packet_type *pt); 3301 void dev_remove_pack(struct packet_type *pt); 3302 void __dev_remove_pack(struct packet_type *pt); 3303 void dev_add_offload(struct packet_offload *po); 3304 void dev_remove_offload(struct packet_offload *po); 3305 3306 int dev_get_iflink(const struct net_device *dev); 3307 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 3308 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 3309 struct net_device_path_stack *stack); 3310 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 3311 unsigned short mask); 3312 struct net_device *dev_get_by_name(struct net *net, const char *name); 3313 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 3314 struct net_device *__dev_get_by_name(struct net *net, const char *name); 3315 bool netdev_name_in_use(struct net *net, const char *name); 3316 int dev_alloc_name(struct net_device *dev, const char *name); 3317 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 3318 void dev_close(struct net_device *dev); 3319 void dev_close_many(struct list_head *head, bool unlink); 3320 void dev_disable_lro(struct net_device *dev); 3321 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 3322 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 3323 struct net_device *sb_dev); 3324 3325 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3326 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3327 3328 static inline int dev_queue_xmit(struct sk_buff *skb) 3329 { 3330 return __dev_queue_xmit(skb, NULL); 3331 } 3332 3333 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3334 struct net_device *sb_dev) 3335 { 3336 return __dev_queue_xmit(skb, sb_dev); 3337 } 3338 3339 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3340 { 3341 int ret; 3342 3343 ret = __dev_direct_xmit(skb, queue_id); 3344 if (!dev_xmit_complete(ret)) 3345 kfree_skb(skb); 3346 return ret; 3347 } 3348 3349 int register_netdevice(struct net_device *dev); 3350 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3351 void unregister_netdevice_many(struct list_head *head); 3352 static inline void unregister_netdevice(struct net_device *dev) 3353 { 3354 unregister_netdevice_queue(dev, NULL); 3355 } 3356 3357 int netdev_refcnt_read(const struct net_device *dev); 3358 void free_netdev(struct net_device *dev); 3359 3360 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3361 struct sk_buff *skb, 3362 bool all_slaves); 3363 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3364 struct sock *sk); 3365 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3366 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3367 struct net_device *netdev_get_by_index(struct net *net, int ifindex, 3368 netdevice_tracker *tracker, gfp_t gfp); 3369 struct net_device *netdev_get_by_name(struct net *net, const char *name, 3370 netdevice_tracker *tracker, gfp_t gfp); 3371 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3372 void netdev_copy_name(struct net_device *dev, char *name); 3373 3374 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3375 unsigned short type, 3376 const void *daddr, const void *saddr, 3377 unsigned int len) 3378 { 3379 if (!dev->header_ops || !dev->header_ops->create) 3380 return 0; 3381 3382 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3383 } 3384 3385 static inline int dev_parse_header(const struct sk_buff *skb, 3386 unsigned char *haddr) 3387 { 3388 const struct net_device *dev = skb->dev; 3389 3390 if (!dev->header_ops || !dev->header_ops->parse) 3391 return 0; 3392 return dev->header_ops->parse(skb, haddr); 3393 } 3394 3395 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3396 { 3397 const struct net_device *dev = skb->dev; 3398 3399 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3400 return 0; 3401 return dev->header_ops->parse_protocol(skb); 3402 } 3403 3404 /* ll_header must have at least hard_header_len allocated */ 3405 static inline bool dev_validate_header(const struct net_device *dev, 3406 char *ll_header, int len) 3407 { 3408 if (likely(len >= dev->hard_header_len)) 3409 return true; 3410 if (len < dev->min_header_len) 3411 return false; 3412 3413 if (capable(CAP_SYS_RAWIO)) { 3414 memset(ll_header + len, 0, dev->hard_header_len - len); 3415 return true; 3416 } 3417 3418 if (dev->header_ops && dev->header_ops->validate) 3419 return dev->header_ops->validate(ll_header, len); 3420 3421 return false; 3422 } 3423 3424 static inline bool dev_has_header(const struct net_device *dev) 3425 { 3426 return dev->header_ops && dev->header_ops->create; 3427 } 3428 3429 /* 3430 * Incoming packets are placed on per-CPU queues 3431 */ 3432 struct softnet_data { 3433 struct list_head poll_list; 3434 struct sk_buff_head process_queue; 3435 local_lock_t process_queue_bh_lock; 3436 3437 /* stats */ 3438 unsigned int processed; 3439 unsigned int time_squeeze; 3440 #ifdef CONFIG_RPS 3441 struct softnet_data *rps_ipi_list; 3442 #endif 3443 3444 unsigned int received_rps; 3445 bool in_net_rx_action; 3446 bool in_napi_threaded_poll; 3447 3448 #ifdef CONFIG_NET_FLOW_LIMIT 3449 struct sd_flow_limit __rcu *flow_limit; 3450 #endif 3451 struct Qdisc *output_queue; 3452 struct Qdisc **output_queue_tailp; 3453 struct sk_buff *completion_queue; 3454 #ifdef CONFIG_XFRM_OFFLOAD 3455 struct sk_buff_head xfrm_backlog; 3456 #endif 3457 /* written and read only by owning cpu: */ 3458 struct netdev_xmit xmit; 3459 #ifdef CONFIG_RPS 3460 /* input_queue_head should be written by cpu owning this struct, 3461 * and only read by other cpus. Worth using a cache line. 3462 */ 3463 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3464 3465 /* Elements below can be accessed between CPUs for RPS/RFS */ 3466 call_single_data_t csd ____cacheline_aligned_in_smp; 3467 struct softnet_data *rps_ipi_next; 3468 unsigned int cpu; 3469 unsigned int input_queue_tail; 3470 #endif 3471 struct sk_buff_head input_pkt_queue; 3472 struct napi_struct backlog; 3473 3474 atomic_t dropped ____cacheline_aligned_in_smp; 3475 3476 /* Another possibly contended cache line */ 3477 spinlock_t defer_lock ____cacheline_aligned_in_smp; 3478 int defer_count; 3479 int defer_ipi_scheduled; 3480 struct sk_buff *defer_list; 3481 call_single_data_t defer_csd; 3482 }; 3483 3484 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3485 DECLARE_PER_CPU(struct page_pool *, system_page_pool); 3486 3487 #ifndef CONFIG_PREEMPT_RT 3488 static inline int dev_recursion_level(void) 3489 { 3490 return this_cpu_read(softnet_data.xmit.recursion); 3491 } 3492 #else 3493 static inline int dev_recursion_level(void) 3494 { 3495 return current->net_xmit.recursion; 3496 } 3497 3498 #endif 3499 3500 void __netif_schedule(struct Qdisc *q); 3501 void netif_schedule_queue(struct netdev_queue *txq); 3502 3503 static inline void netif_tx_schedule_all(struct net_device *dev) 3504 { 3505 unsigned int i; 3506 3507 for (i = 0; i < dev->num_tx_queues; i++) 3508 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3509 } 3510 3511 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3512 { 3513 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3514 } 3515 3516 /** 3517 * netif_start_queue - allow transmit 3518 * @dev: network device 3519 * 3520 * Allow upper layers to call the device hard_start_xmit routine. 3521 */ 3522 static inline void netif_start_queue(struct net_device *dev) 3523 { 3524 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3525 } 3526 3527 static inline void netif_tx_start_all_queues(struct net_device *dev) 3528 { 3529 unsigned int i; 3530 3531 for (i = 0; i < dev->num_tx_queues; i++) { 3532 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3533 netif_tx_start_queue(txq); 3534 } 3535 } 3536 3537 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3538 3539 /** 3540 * netif_wake_queue - restart transmit 3541 * @dev: network device 3542 * 3543 * Allow upper layers to call the device hard_start_xmit routine. 3544 * Used for flow control when transmit resources are available. 3545 */ 3546 static inline void netif_wake_queue(struct net_device *dev) 3547 { 3548 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3549 } 3550 3551 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3552 { 3553 unsigned int i; 3554 3555 for (i = 0; i < dev->num_tx_queues; i++) { 3556 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3557 netif_tx_wake_queue(txq); 3558 } 3559 } 3560 3561 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3562 { 3563 /* Paired with READ_ONCE() from dev_watchdog() */ 3564 WRITE_ONCE(dev_queue->trans_start, jiffies); 3565 3566 /* This barrier is paired with smp_mb() from dev_watchdog() */ 3567 smp_mb__before_atomic(); 3568 3569 /* Must be an atomic op see netif_txq_try_stop() */ 3570 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3571 } 3572 3573 /** 3574 * netif_stop_queue - stop transmitted packets 3575 * @dev: network device 3576 * 3577 * Stop upper layers calling the device hard_start_xmit routine. 3578 * Used for flow control when transmit resources are unavailable. 3579 */ 3580 static inline void netif_stop_queue(struct net_device *dev) 3581 { 3582 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3583 } 3584 3585 void netif_tx_stop_all_queues(struct net_device *dev); 3586 3587 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3588 { 3589 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3590 } 3591 3592 /** 3593 * netif_queue_stopped - test if transmit queue is flowblocked 3594 * @dev: network device 3595 * 3596 * Test if transmit queue on device is currently unable to send. 3597 */ 3598 static inline bool netif_queue_stopped(const struct net_device *dev) 3599 { 3600 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3601 } 3602 3603 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3604 { 3605 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3606 } 3607 3608 static inline bool 3609 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3610 { 3611 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3612 } 3613 3614 static inline bool 3615 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3616 { 3617 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3618 } 3619 3620 /** 3621 * netdev_queue_set_dql_min_limit - set dql minimum limit 3622 * @dev_queue: pointer to transmit queue 3623 * @min_limit: dql minimum limit 3624 * 3625 * Forces xmit_more() to return true until the minimum threshold 3626 * defined by @min_limit is reached (or until the tx queue is 3627 * empty). Warning: to be use with care, misuse will impact the 3628 * latency. 3629 */ 3630 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3631 unsigned int min_limit) 3632 { 3633 #ifdef CONFIG_BQL 3634 dev_queue->dql.min_limit = min_limit; 3635 #endif 3636 } 3637 3638 static inline int netdev_queue_dql_avail(const struct netdev_queue *txq) 3639 { 3640 #ifdef CONFIG_BQL 3641 /* Non-BQL migrated drivers will return 0, too. */ 3642 return dql_avail(&txq->dql); 3643 #else 3644 return 0; 3645 #endif 3646 } 3647 3648 /** 3649 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3650 * @dev_queue: pointer to transmit queue 3651 * 3652 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3653 * to give appropriate hint to the CPU. 3654 */ 3655 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3656 { 3657 #ifdef CONFIG_BQL 3658 prefetchw(&dev_queue->dql.num_queued); 3659 #endif 3660 } 3661 3662 /** 3663 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3664 * @dev_queue: pointer to transmit queue 3665 * 3666 * BQL enabled drivers might use this helper in their TX completion path, 3667 * to give appropriate hint to the CPU. 3668 */ 3669 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3670 { 3671 #ifdef CONFIG_BQL 3672 prefetchw(&dev_queue->dql.limit); 3673 #endif 3674 } 3675 3676 /** 3677 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3678 * @dev_queue: network device queue 3679 * @bytes: number of bytes queued to the device queue 3680 * 3681 * Report the number of bytes queued for sending/completion to the network 3682 * device hardware queue. @bytes should be a good approximation and should 3683 * exactly match netdev_completed_queue() @bytes. 3684 * This is typically called once per packet, from ndo_start_xmit(). 3685 */ 3686 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3687 unsigned int bytes) 3688 { 3689 #ifdef CONFIG_BQL 3690 dql_queued(&dev_queue->dql, bytes); 3691 3692 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3693 return; 3694 3695 /* Paired with READ_ONCE() from dev_watchdog() */ 3696 WRITE_ONCE(dev_queue->trans_start, jiffies); 3697 3698 /* This barrier is paired with smp_mb() from dev_watchdog() */ 3699 smp_mb__before_atomic(); 3700 3701 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3702 3703 /* 3704 * The XOFF flag must be set before checking the dql_avail below, 3705 * because in netdev_tx_completed_queue we update the dql_completed 3706 * before checking the XOFF flag. 3707 */ 3708 smp_mb__after_atomic(); 3709 3710 /* check again in case another CPU has just made room avail */ 3711 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3712 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3713 #endif 3714 } 3715 3716 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3717 * that they should not test BQL status themselves. 3718 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3719 * skb of a batch. 3720 * Returns true if the doorbell must be used to kick the NIC. 3721 */ 3722 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3723 unsigned int bytes, 3724 bool xmit_more) 3725 { 3726 if (xmit_more) { 3727 #ifdef CONFIG_BQL 3728 dql_queued(&dev_queue->dql, bytes); 3729 #endif 3730 return netif_tx_queue_stopped(dev_queue); 3731 } 3732 netdev_tx_sent_queue(dev_queue, bytes); 3733 return true; 3734 } 3735 3736 /** 3737 * netdev_sent_queue - report the number of bytes queued to hardware 3738 * @dev: network device 3739 * @bytes: number of bytes queued to the hardware device queue 3740 * 3741 * Report the number of bytes queued for sending/completion to the network 3742 * device hardware queue#0. @bytes should be a good approximation and should 3743 * exactly match netdev_completed_queue() @bytes. 3744 * This is typically called once per packet, from ndo_start_xmit(). 3745 */ 3746 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3747 { 3748 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3749 } 3750 3751 static inline bool __netdev_sent_queue(struct net_device *dev, 3752 unsigned int bytes, 3753 bool xmit_more) 3754 { 3755 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3756 xmit_more); 3757 } 3758 3759 /** 3760 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3761 * @dev_queue: network device queue 3762 * @pkts: number of packets (currently ignored) 3763 * @bytes: number of bytes dequeued from the device queue 3764 * 3765 * Must be called at most once per TX completion round (and not per 3766 * individual packet), so that BQL can adjust its limits appropriately. 3767 */ 3768 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3769 unsigned int pkts, unsigned int bytes) 3770 { 3771 #ifdef CONFIG_BQL 3772 if (unlikely(!bytes)) 3773 return; 3774 3775 dql_completed(&dev_queue->dql, bytes); 3776 3777 /* 3778 * Without the memory barrier there is a small possibility that 3779 * netdev_tx_sent_queue will miss the update and cause the queue to 3780 * be stopped forever 3781 */ 3782 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */ 3783 3784 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3785 return; 3786 3787 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3788 netif_schedule_queue(dev_queue); 3789 #endif 3790 } 3791 3792 /** 3793 * netdev_completed_queue - report bytes and packets completed by device 3794 * @dev: network device 3795 * @pkts: actual number of packets sent over the medium 3796 * @bytes: actual number of bytes sent over the medium 3797 * 3798 * Report the number of bytes and packets transmitted by the network device 3799 * hardware queue over the physical medium, @bytes must exactly match the 3800 * @bytes amount passed to netdev_sent_queue() 3801 */ 3802 static inline void netdev_completed_queue(struct net_device *dev, 3803 unsigned int pkts, unsigned int bytes) 3804 { 3805 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3806 } 3807 3808 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3809 { 3810 #ifdef CONFIG_BQL 3811 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3812 dql_reset(&q->dql); 3813 #endif 3814 } 3815 3816 /** 3817 * netdev_tx_reset_subqueue - reset the BQL stats and state of a netdev queue 3818 * @dev: network device 3819 * @qid: stack index of the queue to reset 3820 */ 3821 static inline void netdev_tx_reset_subqueue(const struct net_device *dev, 3822 u32 qid) 3823 { 3824 netdev_tx_reset_queue(netdev_get_tx_queue(dev, qid)); 3825 } 3826 3827 /** 3828 * netdev_reset_queue - reset the packets and bytes count of a network device 3829 * @dev_queue: network device 3830 * 3831 * Reset the bytes and packet count of a network device and clear the 3832 * software flow control OFF bit for this network device 3833 */ 3834 static inline void netdev_reset_queue(struct net_device *dev_queue) 3835 { 3836 netdev_tx_reset_subqueue(dev_queue, 0); 3837 } 3838 3839 /** 3840 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3841 * @dev: network device 3842 * @queue_index: given tx queue index 3843 * 3844 * Returns 0 if given tx queue index >= number of device tx queues, 3845 * otherwise returns the originally passed tx queue index. 3846 */ 3847 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3848 { 3849 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3850 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3851 dev->name, queue_index, 3852 dev->real_num_tx_queues); 3853 return 0; 3854 } 3855 3856 return queue_index; 3857 } 3858 3859 /** 3860 * netif_running - test if up 3861 * @dev: network device 3862 * 3863 * Test if the device has been brought up. 3864 */ 3865 static inline bool netif_running(const struct net_device *dev) 3866 { 3867 return test_bit(__LINK_STATE_START, &dev->state); 3868 } 3869 3870 /* 3871 * Routines to manage the subqueues on a device. We only need start, 3872 * stop, and a check if it's stopped. All other device management is 3873 * done at the overall netdevice level. 3874 * Also test the device if we're multiqueue. 3875 */ 3876 3877 /** 3878 * netif_start_subqueue - allow sending packets on subqueue 3879 * @dev: network device 3880 * @queue_index: sub queue index 3881 * 3882 * Start individual transmit queue of a device with multiple transmit queues. 3883 */ 3884 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3885 { 3886 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3887 3888 netif_tx_start_queue(txq); 3889 } 3890 3891 /** 3892 * netif_stop_subqueue - stop sending packets on subqueue 3893 * @dev: network device 3894 * @queue_index: sub queue index 3895 * 3896 * Stop individual transmit queue of a device with multiple transmit queues. 3897 */ 3898 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3899 { 3900 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3901 netif_tx_stop_queue(txq); 3902 } 3903 3904 /** 3905 * __netif_subqueue_stopped - test status of subqueue 3906 * @dev: network device 3907 * @queue_index: sub queue index 3908 * 3909 * Check individual transmit queue of a device with multiple transmit queues. 3910 */ 3911 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3912 u16 queue_index) 3913 { 3914 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3915 3916 return netif_tx_queue_stopped(txq); 3917 } 3918 3919 /** 3920 * netif_subqueue_stopped - test status of subqueue 3921 * @dev: network device 3922 * @skb: sub queue buffer pointer 3923 * 3924 * Check individual transmit queue of a device with multiple transmit queues. 3925 */ 3926 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3927 struct sk_buff *skb) 3928 { 3929 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3930 } 3931 3932 /** 3933 * netif_wake_subqueue - allow sending packets on subqueue 3934 * @dev: network device 3935 * @queue_index: sub queue index 3936 * 3937 * Resume individual transmit queue of a device with multiple transmit queues. 3938 */ 3939 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3940 { 3941 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3942 3943 netif_tx_wake_queue(txq); 3944 } 3945 3946 #ifdef CONFIG_XPS 3947 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3948 u16 index); 3949 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3950 u16 index, enum xps_map_type type); 3951 3952 /** 3953 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3954 * @j: CPU/Rx queue index 3955 * @mask: bitmask of all cpus/rx queues 3956 * @nr_bits: number of bits in the bitmask 3957 * 3958 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3959 */ 3960 static inline bool netif_attr_test_mask(unsigned long j, 3961 const unsigned long *mask, 3962 unsigned int nr_bits) 3963 { 3964 cpu_max_bits_warn(j, nr_bits); 3965 return test_bit(j, mask); 3966 } 3967 3968 /** 3969 * netif_attr_test_online - Test for online CPU/Rx queue 3970 * @j: CPU/Rx queue index 3971 * @online_mask: bitmask for CPUs/Rx queues that are online 3972 * @nr_bits: number of bits in the bitmask 3973 * 3974 * Returns: true if a CPU/Rx queue is online. 3975 */ 3976 static inline bool netif_attr_test_online(unsigned long j, 3977 const unsigned long *online_mask, 3978 unsigned int nr_bits) 3979 { 3980 cpu_max_bits_warn(j, nr_bits); 3981 3982 if (online_mask) 3983 return test_bit(j, online_mask); 3984 3985 return (j < nr_bits); 3986 } 3987 3988 /** 3989 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3990 * @n: CPU/Rx queue index 3991 * @srcp: the cpumask/Rx queue mask pointer 3992 * @nr_bits: number of bits in the bitmask 3993 * 3994 * Returns: next (after n) CPU/Rx queue index in the mask; 3995 * >= nr_bits if no further CPUs/Rx queues set. 3996 */ 3997 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3998 unsigned int nr_bits) 3999 { 4000 /* -1 is a legal arg here. */ 4001 if (n != -1) 4002 cpu_max_bits_warn(n, nr_bits); 4003 4004 if (srcp) 4005 return find_next_bit(srcp, nr_bits, n + 1); 4006 4007 return n + 1; 4008 } 4009 4010 /** 4011 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 4012 * @n: CPU/Rx queue index 4013 * @src1p: the first CPUs/Rx queues mask pointer 4014 * @src2p: the second CPUs/Rx queues mask pointer 4015 * @nr_bits: number of bits in the bitmask 4016 * 4017 * Returns: next (after n) CPU/Rx queue index set in both masks; 4018 * >= nr_bits if no further CPUs/Rx queues set in both. 4019 */ 4020 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 4021 const unsigned long *src2p, 4022 unsigned int nr_bits) 4023 { 4024 /* -1 is a legal arg here. */ 4025 if (n != -1) 4026 cpu_max_bits_warn(n, nr_bits); 4027 4028 if (src1p && src2p) 4029 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 4030 else if (src1p) 4031 return find_next_bit(src1p, nr_bits, n + 1); 4032 else if (src2p) 4033 return find_next_bit(src2p, nr_bits, n + 1); 4034 4035 return n + 1; 4036 } 4037 #else 4038 static inline int netif_set_xps_queue(struct net_device *dev, 4039 const struct cpumask *mask, 4040 u16 index) 4041 { 4042 return 0; 4043 } 4044 4045 static inline int __netif_set_xps_queue(struct net_device *dev, 4046 const unsigned long *mask, 4047 u16 index, enum xps_map_type type) 4048 { 4049 return 0; 4050 } 4051 #endif 4052 4053 /** 4054 * netif_is_multiqueue - test if device has multiple transmit queues 4055 * @dev: network device 4056 * 4057 * Check if device has multiple transmit queues 4058 */ 4059 static inline bool netif_is_multiqueue(const struct net_device *dev) 4060 { 4061 return dev->num_tx_queues > 1; 4062 } 4063 4064 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 4065 4066 #ifdef CONFIG_SYSFS 4067 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 4068 #else 4069 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 4070 unsigned int rxqs) 4071 { 4072 dev->real_num_rx_queues = rxqs; 4073 return 0; 4074 } 4075 #endif 4076 int netif_set_real_num_queues(struct net_device *dev, 4077 unsigned int txq, unsigned int rxq); 4078 4079 int netif_get_num_default_rss_queues(void); 4080 4081 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason); 4082 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason); 4083 4084 /* 4085 * It is not allowed to call kfree_skb() or consume_skb() from hardware 4086 * interrupt context or with hardware interrupts being disabled. 4087 * (in_hardirq() || irqs_disabled()) 4088 * 4089 * We provide four helpers that can be used in following contexts : 4090 * 4091 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 4092 * replacing kfree_skb(skb) 4093 * 4094 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 4095 * Typically used in place of consume_skb(skb) in TX completion path 4096 * 4097 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 4098 * replacing kfree_skb(skb) 4099 * 4100 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 4101 * and consumed a packet. Used in place of consume_skb(skb) 4102 */ 4103 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 4104 { 4105 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 4106 } 4107 4108 static inline void dev_consume_skb_irq(struct sk_buff *skb) 4109 { 4110 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED); 4111 } 4112 4113 static inline void dev_kfree_skb_any(struct sk_buff *skb) 4114 { 4115 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 4116 } 4117 4118 static inline void dev_consume_skb_any(struct sk_buff *skb) 4119 { 4120 dev_kfree_skb_any_reason(skb, SKB_CONSUMED); 4121 } 4122 4123 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 4124 const struct bpf_prog *xdp_prog); 4125 void generic_xdp_tx(struct sk_buff *skb, const struct bpf_prog *xdp_prog); 4126 int do_xdp_generic(const struct bpf_prog *xdp_prog, struct sk_buff **pskb); 4127 int netif_rx(struct sk_buff *skb); 4128 int __netif_rx(struct sk_buff *skb); 4129 4130 int netif_receive_skb(struct sk_buff *skb); 4131 int netif_receive_skb_core(struct sk_buff *skb); 4132 void netif_receive_skb_list_internal(struct list_head *head); 4133 void netif_receive_skb_list(struct list_head *head); 4134 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 4135 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 4136 struct sk_buff *napi_get_frags(struct napi_struct *napi); 4137 gro_result_t napi_gro_frags(struct napi_struct *napi); 4138 4139 static inline void napi_free_frags(struct napi_struct *napi) 4140 { 4141 kfree_skb(napi->skb); 4142 napi->skb = NULL; 4143 } 4144 4145 bool netdev_is_rx_handler_busy(struct net_device *dev); 4146 int netdev_rx_handler_register(struct net_device *dev, 4147 rx_handler_func_t *rx_handler, 4148 void *rx_handler_data); 4149 void netdev_rx_handler_unregister(struct net_device *dev); 4150 4151 bool dev_valid_name(const char *name); 4152 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 4153 { 4154 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 4155 } 4156 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 4157 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 4158 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 4159 void __user *data, bool *need_copyout); 4160 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 4161 int generic_hwtstamp_get_lower(struct net_device *dev, 4162 struct kernel_hwtstamp_config *kernel_cfg); 4163 int generic_hwtstamp_set_lower(struct net_device *dev, 4164 struct kernel_hwtstamp_config *kernel_cfg, 4165 struct netlink_ext_ack *extack); 4166 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 4167 unsigned int dev_get_flags(const struct net_device *); 4168 int __dev_change_flags(struct net_device *dev, unsigned int flags, 4169 struct netlink_ext_ack *extack); 4170 int dev_change_flags(struct net_device *dev, unsigned int flags, 4171 struct netlink_ext_ack *extack); 4172 int dev_set_alias(struct net_device *, const char *, size_t); 4173 int dev_get_alias(const struct net_device *, char *, size_t); 4174 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 4175 const char *pat, int new_ifindex); 4176 static inline 4177 int dev_change_net_namespace(struct net_device *dev, struct net *net, 4178 const char *pat) 4179 { 4180 return __dev_change_net_namespace(dev, net, pat, 0); 4181 } 4182 int __dev_set_mtu(struct net_device *, int); 4183 int dev_set_mtu(struct net_device *, int); 4184 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 4185 struct netlink_ext_ack *extack); 4186 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 4187 struct netlink_ext_ack *extack); 4188 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 4189 struct netlink_ext_ack *extack); 4190 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 4191 int dev_get_port_parent_id(struct net_device *dev, 4192 struct netdev_phys_item_id *ppid, bool recurse); 4193 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 4194 4195 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 4196 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 4197 struct netdev_queue *txq, int *ret); 4198 4199 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 4200 u8 dev_xdp_prog_count(struct net_device *dev); 4201 int dev_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf); 4202 u8 dev_xdp_sb_prog_count(struct net_device *dev); 4203 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 4204 4205 u32 dev_get_min_mp_channel_count(const struct net_device *dev); 4206 4207 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4208 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4209 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 4210 bool is_skb_forwardable(const struct net_device *dev, 4211 const struct sk_buff *skb); 4212 4213 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 4214 const struct sk_buff *skb, 4215 const bool check_mtu) 4216 { 4217 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 4218 unsigned int len; 4219 4220 if (!(dev->flags & IFF_UP)) 4221 return false; 4222 4223 if (!check_mtu) 4224 return true; 4225 4226 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 4227 if (skb->len <= len) 4228 return true; 4229 4230 /* if TSO is enabled, we don't care about the length as the packet 4231 * could be forwarded without being segmented before 4232 */ 4233 if (skb_is_gso(skb)) 4234 return true; 4235 4236 return false; 4237 } 4238 4239 void netdev_core_stats_inc(struct net_device *dev, u32 offset); 4240 4241 #define DEV_CORE_STATS_INC(FIELD) \ 4242 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 4243 { \ 4244 netdev_core_stats_inc(dev, \ 4245 offsetof(struct net_device_core_stats, FIELD)); \ 4246 } 4247 DEV_CORE_STATS_INC(rx_dropped) 4248 DEV_CORE_STATS_INC(tx_dropped) 4249 DEV_CORE_STATS_INC(rx_nohandler) 4250 DEV_CORE_STATS_INC(rx_otherhost_dropped) 4251 #undef DEV_CORE_STATS_INC 4252 4253 static __always_inline int ____dev_forward_skb(struct net_device *dev, 4254 struct sk_buff *skb, 4255 const bool check_mtu) 4256 { 4257 if (skb_orphan_frags(skb, GFP_ATOMIC) || 4258 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 4259 dev_core_stats_rx_dropped_inc(dev); 4260 kfree_skb(skb); 4261 return NET_RX_DROP; 4262 } 4263 4264 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 4265 skb->priority = 0; 4266 return 0; 4267 } 4268 4269 bool dev_nit_active(struct net_device *dev); 4270 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 4271 4272 static inline void __dev_put(struct net_device *dev) 4273 { 4274 if (dev) { 4275 #ifdef CONFIG_PCPU_DEV_REFCNT 4276 this_cpu_dec(*dev->pcpu_refcnt); 4277 #else 4278 refcount_dec(&dev->dev_refcnt); 4279 #endif 4280 } 4281 } 4282 4283 static inline void __dev_hold(struct net_device *dev) 4284 { 4285 if (dev) { 4286 #ifdef CONFIG_PCPU_DEV_REFCNT 4287 this_cpu_inc(*dev->pcpu_refcnt); 4288 #else 4289 refcount_inc(&dev->dev_refcnt); 4290 #endif 4291 } 4292 } 4293 4294 static inline void __netdev_tracker_alloc(struct net_device *dev, 4295 netdevice_tracker *tracker, 4296 gfp_t gfp) 4297 { 4298 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4299 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 4300 #endif 4301 } 4302 4303 /* netdev_tracker_alloc() can upgrade a prior untracked reference 4304 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4305 */ 4306 static inline void netdev_tracker_alloc(struct net_device *dev, 4307 netdevice_tracker *tracker, gfp_t gfp) 4308 { 4309 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4310 refcount_dec(&dev->refcnt_tracker.no_tracker); 4311 __netdev_tracker_alloc(dev, tracker, gfp); 4312 #endif 4313 } 4314 4315 static inline void netdev_tracker_free(struct net_device *dev, 4316 netdevice_tracker *tracker) 4317 { 4318 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4319 ref_tracker_free(&dev->refcnt_tracker, tracker); 4320 #endif 4321 } 4322 4323 static inline void netdev_hold(struct net_device *dev, 4324 netdevice_tracker *tracker, gfp_t gfp) 4325 { 4326 if (dev) { 4327 __dev_hold(dev); 4328 __netdev_tracker_alloc(dev, tracker, gfp); 4329 } 4330 } 4331 4332 static inline void netdev_put(struct net_device *dev, 4333 netdevice_tracker *tracker) 4334 { 4335 if (dev) { 4336 netdev_tracker_free(dev, tracker); 4337 __dev_put(dev); 4338 } 4339 } 4340 4341 /** 4342 * dev_hold - get reference to device 4343 * @dev: network device 4344 * 4345 * Hold reference to device to keep it from being freed. 4346 * Try using netdev_hold() instead. 4347 */ 4348 static inline void dev_hold(struct net_device *dev) 4349 { 4350 netdev_hold(dev, NULL, GFP_ATOMIC); 4351 } 4352 4353 /** 4354 * dev_put - release reference to device 4355 * @dev: network device 4356 * 4357 * Release reference to device to allow it to be freed. 4358 * Try using netdev_put() instead. 4359 */ 4360 static inline void dev_put(struct net_device *dev) 4361 { 4362 netdev_put(dev, NULL); 4363 } 4364 4365 DEFINE_FREE(dev_put, struct net_device *, if (_T) dev_put(_T)) 4366 4367 static inline void netdev_ref_replace(struct net_device *odev, 4368 struct net_device *ndev, 4369 netdevice_tracker *tracker, 4370 gfp_t gfp) 4371 { 4372 if (odev) 4373 netdev_tracker_free(odev, tracker); 4374 4375 __dev_hold(ndev); 4376 __dev_put(odev); 4377 4378 if (ndev) 4379 __netdev_tracker_alloc(ndev, tracker, gfp); 4380 } 4381 4382 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4383 * and _off may be called from IRQ context, but it is caller 4384 * who is responsible for serialization of these calls. 4385 * 4386 * The name carrier is inappropriate, these functions should really be 4387 * called netif_lowerlayer_*() because they represent the state of any 4388 * kind of lower layer not just hardware media. 4389 */ 4390 void linkwatch_fire_event(struct net_device *dev); 4391 4392 /** 4393 * linkwatch_sync_dev - sync linkwatch for the given device 4394 * @dev: network device to sync linkwatch for 4395 * 4396 * Sync linkwatch for the given device, removing it from the 4397 * pending work list (if queued). 4398 */ 4399 void linkwatch_sync_dev(struct net_device *dev); 4400 4401 /** 4402 * netif_carrier_ok - test if carrier present 4403 * @dev: network device 4404 * 4405 * Check if carrier is present on device 4406 */ 4407 static inline bool netif_carrier_ok(const struct net_device *dev) 4408 { 4409 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4410 } 4411 4412 unsigned long dev_trans_start(struct net_device *dev); 4413 4414 void netdev_watchdog_up(struct net_device *dev); 4415 4416 void netif_carrier_on(struct net_device *dev); 4417 void netif_carrier_off(struct net_device *dev); 4418 void netif_carrier_event(struct net_device *dev); 4419 4420 /** 4421 * netif_dormant_on - mark device as dormant. 4422 * @dev: network device 4423 * 4424 * Mark device as dormant (as per RFC2863). 4425 * 4426 * The dormant state indicates that the relevant interface is not 4427 * actually in a condition to pass packets (i.e., it is not 'up') but is 4428 * in a "pending" state, waiting for some external event. For "on- 4429 * demand" interfaces, this new state identifies the situation where the 4430 * interface is waiting for events to place it in the up state. 4431 */ 4432 static inline void netif_dormant_on(struct net_device *dev) 4433 { 4434 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4435 linkwatch_fire_event(dev); 4436 } 4437 4438 /** 4439 * netif_dormant_off - set device as not dormant. 4440 * @dev: network device 4441 * 4442 * Device is not in dormant state. 4443 */ 4444 static inline void netif_dormant_off(struct net_device *dev) 4445 { 4446 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4447 linkwatch_fire_event(dev); 4448 } 4449 4450 /** 4451 * netif_dormant - test if device is dormant 4452 * @dev: network device 4453 * 4454 * Check if device is dormant. 4455 */ 4456 static inline bool netif_dormant(const struct net_device *dev) 4457 { 4458 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4459 } 4460 4461 4462 /** 4463 * netif_testing_on - mark device as under test. 4464 * @dev: network device 4465 * 4466 * Mark device as under test (as per RFC2863). 4467 * 4468 * The testing state indicates that some test(s) must be performed on 4469 * the interface. After completion, of the test, the interface state 4470 * will change to up, dormant, or down, as appropriate. 4471 */ 4472 static inline void netif_testing_on(struct net_device *dev) 4473 { 4474 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4475 linkwatch_fire_event(dev); 4476 } 4477 4478 /** 4479 * netif_testing_off - set device as not under test. 4480 * @dev: network device 4481 * 4482 * Device is not in testing state. 4483 */ 4484 static inline void netif_testing_off(struct net_device *dev) 4485 { 4486 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4487 linkwatch_fire_event(dev); 4488 } 4489 4490 /** 4491 * netif_testing - test if device is under test 4492 * @dev: network device 4493 * 4494 * Check if device is under test 4495 */ 4496 static inline bool netif_testing(const struct net_device *dev) 4497 { 4498 return test_bit(__LINK_STATE_TESTING, &dev->state); 4499 } 4500 4501 4502 /** 4503 * netif_oper_up - test if device is operational 4504 * @dev: network device 4505 * 4506 * Check if carrier is operational 4507 */ 4508 static inline bool netif_oper_up(const struct net_device *dev) 4509 { 4510 unsigned int operstate = READ_ONCE(dev->operstate); 4511 4512 return operstate == IF_OPER_UP || 4513 operstate == IF_OPER_UNKNOWN /* backward compat */; 4514 } 4515 4516 /** 4517 * netif_device_present - is device available or removed 4518 * @dev: network device 4519 * 4520 * Check if device has not been removed from system. 4521 */ 4522 static inline bool netif_device_present(const struct net_device *dev) 4523 { 4524 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4525 } 4526 4527 void netif_device_detach(struct net_device *dev); 4528 4529 void netif_device_attach(struct net_device *dev); 4530 4531 /* 4532 * Network interface message level settings 4533 */ 4534 4535 enum { 4536 NETIF_MSG_DRV_BIT, 4537 NETIF_MSG_PROBE_BIT, 4538 NETIF_MSG_LINK_BIT, 4539 NETIF_MSG_TIMER_BIT, 4540 NETIF_MSG_IFDOWN_BIT, 4541 NETIF_MSG_IFUP_BIT, 4542 NETIF_MSG_RX_ERR_BIT, 4543 NETIF_MSG_TX_ERR_BIT, 4544 NETIF_MSG_TX_QUEUED_BIT, 4545 NETIF_MSG_INTR_BIT, 4546 NETIF_MSG_TX_DONE_BIT, 4547 NETIF_MSG_RX_STATUS_BIT, 4548 NETIF_MSG_PKTDATA_BIT, 4549 NETIF_MSG_HW_BIT, 4550 NETIF_MSG_WOL_BIT, 4551 4552 /* When you add a new bit above, update netif_msg_class_names array 4553 * in net/ethtool/common.c 4554 */ 4555 NETIF_MSG_CLASS_COUNT, 4556 }; 4557 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4558 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4559 4560 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4561 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4562 4563 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4564 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4565 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4566 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4567 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4568 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4569 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4570 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4571 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4572 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4573 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4574 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4575 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4576 #define NETIF_MSG_HW __NETIF_MSG(HW) 4577 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4578 4579 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4580 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4581 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4582 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4583 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4584 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4585 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4586 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4587 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4588 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4589 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4590 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4591 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4592 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4593 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4594 4595 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4596 { 4597 /* use default */ 4598 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4599 return default_msg_enable_bits; 4600 if (debug_value == 0) /* no output */ 4601 return 0; 4602 /* set low N bits */ 4603 return (1U << debug_value) - 1; 4604 } 4605 4606 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4607 { 4608 spin_lock(&txq->_xmit_lock); 4609 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4610 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4611 } 4612 4613 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4614 { 4615 __acquire(&txq->_xmit_lock); 4616 return true; 4617 } 4618 4619 static inline void __netif_tx_release(struct netdev_queue *txq) 4620 { 4621 __release(&txq->_xmit_lock); 4622 } 4623 4624 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4625 { 4626 spin_lock_bh(&txq->_xmit_lock); 4627 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4628 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4629 } 4630 4631 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4632 { 4633 bool ok = spin_trylock(&txq->_xmit_lock); 4634 4635 if (likely(ok)) { 4636 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4637 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4638 } 4639 return ok; 4640 } 4641 4642 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4643 { 4644 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4645 WRITE_ONCE(txq->xmit_lock_owner, -1); 4646 spin_unlock(&txq->_xmit_lock); 4647 } 4648 4649 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4650 { 4651 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4652 WRITE_ONCE(txq->xmit_lock_owner, -1); 4653 spin_unlock_bh(&txq->_xmit_lock); 4654 } 4655 4656 /* 4657 * txq->trans_start can be read locklessly from dev_watchdog() 4658 */ 4659 static inline void txq_trans_update(struct netdev_queue *txq) 4660 { 4661 if (txq->xmit_lock_owner != -1) 4662 WRITE_ONCE(txq->trans_start, jiffies); 4663 } 4664 4665 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4666 { 4667 unsigned long now = jiffies; 4668 4669 if (READ_ONCE(txq->trans_start) != now) 4670 WRITE_ONCE(txq->trans_start, now); 4671 } 4672 4673 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4674 static inline void netif_trans_update(struct net_device *dev) 4675 { 4676 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4677 4678 txq_trans_cond_update(txq); 4679 } 4680 4681 /** 4682 * netif_tx_lock - grab network device transmit lock 4683 * @dev: network device 4684 * 4685 * Get network device transmit lock 4686 */ 4687 void netif_tx_lock(struct net_device *dev); 4688 4689 static inline void netif_tx_lock_bh(struct net_device *dev) 4690 { 4691 local_bh_disable(); 4692 netif_tx_lock(dev); 4693 } 4694 4695 void netif_tx_unlock(struct net_device *dev); 4696 4697 static inline void netif_tx_unlock_bh(struct net_device *dev) 4698 { 4699 netif_tx_unlock(dev); 4700 local_bh_enable(); 4701 } 4702 4703 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4704 if (!(dev)->lltx) { \ 4705 __netif_tx_lock(txq, cpu); \ 4706 } else { \ 4707 __netif_tx_acquire(txq); \ 4708 } \ 4709 } 4710 4711 #define HARD_TX_TRYLOCK(dev, txq) \ 4712 (!(dev)->lltx ? \ 4713 __netif_tx_trylock(txq) : \ 4714 __netif_tx_acquire(txq)) 4715 4716 #define HARD_TX_UNLOCK(dev, txq) { \ 4717 if (!(dev)->lltx) { \ 4718 __netif_tx_unlock(txq); \ 4719 } else { \ 4720 __netif_tx_release(txq); \ 4721 } \ 4722 } 4723 4724 static inline void netif_tx_disable(struct net_device *dev) 4725 { 4726 unsigned int i; 4727 int cpu; 4728 4729 local_bh_disable(); 4730 cpu = smp_processor_id(); 4731 spin_lock(&dev->tx_global_lock); 4732 for (i = 0; i < dev->num_tx_queues; i++) { 4733 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4734 4735 __netif_tx_lock(txq, cpu); 4736 netif_tx_stop_queue(txq); 4737 __netif_tx_unlock(txq); 4738 } 4739 spin_unlock(&dev->tx_global_lock); 4740 local_bh_enable(); 4741 } 4742 4743 static inline void netif_addr_lock(struct net_device *dev) 4744 { 4745 unsigned char nest_level = 0; 4746 4747 #ifdef CONFIG_LOCKDEP 4748 nest_level = dev->nested_level; 4749 #endif 4750 spin_lock_nested(&dev->addr_list_lock, nest_level); 4751 } 4752 4753 static inline void netif_addr_lock_bh(struct net_device *dev) 4754 { 4755 unsigned char nest_level = 0; 4756 4757 #ifdef CONFIG_LOCKDEP 4758 nest_level = dev->nested_level; 4759 #endif 4760 local_bh_disable(); 4761 spin_lock_nested(&dev->addr_list_lock, nest_level); 4762 } 4763 4764 static inline void netif_addr_unlock(struct net_device *dev) 4765 { 4766 spin_unlock(&dev->addr_list_lock); 4767 } 4768 4769 static inline void netif_addr_unlock_bh(struct net_device *dev) 4770 { 4771 spin_unlock_bh(&dev->addr_list_lock); 4772 } 4773 4774 /* 4775 * dev_addrs walker. Should be used only for read access. Call with 4776 * rcu_read_lock held. 4777 */ 4778 #define for_each_dev_addr(dev, ha) \ 4779 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4780 4781 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4782 4783 void ether_setup(struct net_device *dev); 4784 4785 /* Allocate dummy net_device */ 4786 struct net_device *alloc_netdev_dummy(int sizeof_priv); 4787 4788 /* Support for loadable net-drivers */ 4789 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4790 unsigned char name_assign_type, 4791 void (*setup)(struct net_device *), 4792 unsigned int txqs, unsigned int rxqs); 4793 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4794 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4795 4796 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4797 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4798 count) 4799 4800 int register_netdev(struct net_device *dev); 4801 void unregister_netdev(struct net_device *dev); 4802 4803 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4804 4805 /* General hardware address lists handling functions */ 4806 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4807 struct netdev_hw_addr_list *from_list, int addr_len); 4808 int __hw_addr_sync_multiple(struct netdev_hw_addr_list *to_list, 4809 struct netdev_hw_addr_list *from_list, 4810 int addr_len); 4811 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4812 struct netdev_hw_addr_list *from_list, int addr_len); 4813 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4814 struct net_device *dev, 4815 int (*sync)(struct net_device *, const unsigned char *), 4816 int (*unsync)(struct net_device *, 4817 const unsigned char *)); 4818 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4819 struct net_device *dev, 4820 int (*sync)(struct net_device *, 4821 const unsigned char *, int), 4822 int (*unsync)(struct net_device *, 4823 const unsigned char *, int)); 4824 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4825 struct net_device *dev, 4826 int (*unsync)(struct net_device *, 4827 const unsigned char *, int)); 4828 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4829 struct net_device *dev, 4830 int (*unsync)(struct net_device *, 4831 const unsigned char *)); 4832 void __hw_addr_init(struct netdev_hw_addr_list *list); 4833 4834 /* Functions used for device addresses handling */ 4835 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4836 const void *addr, size_t len); 4837 4838 static inline void 4839 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4840 { 4841 dev_addr_mod(dev, 0, addr, len); 4842 } 4843 4844 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4845 { 4846 __dev_addr_set(dev, addr, dev->addr_len); 4847 } 4848 4849 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4850 unsigned char addr_type); 4851 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4852 unsigned char addr_type); 4853 4854 /* Functions used for unicast addresses handling */ 4855 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4856 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4857 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4858 int dev_uc_sync(struct net_device *to, struct net_device *from); 4859 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4860 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4861 void dev_uc_flush(struct net_device *dev); 4862 void dev_uc_init(struct net_device *dev); 4863 4864 /** 4865 * __dev_uc_sync - Synchronize device's unicast list 4866 * @dev: device to sync 4867 * @sync: function to call if address should be added 4868 * @unsync: function to call if address should be removed 4869 * 4870 * Add newly added addresses to the interface, and release 4871 * addresses that have been deleted. 4872 */ 4873 static inline int __dev_uc_sync(struct net_device *dev, 4874 int (*sync)(struct net_device *, 4875 const unsigned char *), 4876 int (*unsync)(struct net_device *, 4877 const unsigned char *)) 4878 { 4879 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4880 } 4881 4882 /** 4883 * __dev_uc_unsync - Remove synchronized addresses from device 4884 * @dev: device to sync 4885 * @unsync: function to call if address should be removed 4886 * 4887 * Remove all addresses that were added to the device by dev_uc_sync(). 4888 */ 4889 static inline void __dev_uc_unsync(struct net_device *dev, 4890 int (*unsync)(struct net_device *, 4891 const unsigned char *)) 4892 { 4893 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4894 } 4895 4896 /* Functions used for multicast addresses handling */ 4897 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4898 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4899 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4900 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4901 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4902 int dev_mc_sync(struct net_device *to, struct net_device *from); 4903 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4904 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4905 void dev_mc_flush(struct net_device *dev); 4906 void dev_mc_init(struct net_device *dev); 4907 4908 /** 4909 * __dev_mc_sync - Synchronize device's multicast list 4910 * @dev: device to sync 4911 * @sync: function to call if address should be added 4912 * @unsync: function to call if address should be removed 4913 * 4914 * Add newly added addresses to the interface, and release 4915 * addresses that have been deleted. 4916 */ 4917 static inline int __dev_mc_sync(struct net_device *dev, 4918 int (*sync)(struct net_device *, 4919 const unsigned char *), 4920 int (*unsync)(struct net_device *, 4921 const unsigned char *)) 4922 { 4923 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4924 } 4925 4926 /** 4927 * __dev_mc_unsync - Remove synchronized addresses from device 4928 * @dev: device to sync 4929 * @unsync: function to call if address should be removed 4930 * 4931 * Remove all addresses that were added to the device by dev_mc_sync(). 4932 */ 4933 static inline void __dev_mc_unsync(struct net_device *dev, 4934 int (*unsync)(struct net_device *, 4935 const unsigned char *)) 4936 { 4937 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4938 } 4939 4940 /* Functions used for secondary unicast and multicast support */ 4941 void dev_set_rx_mode(struct net_device *dev); 4942 int dev_set_promiscuity(struct net_device *dev, int inc); 4943 int dev_set_allmulti(struct net_device *dev, int inc); 4944 void netdev_state_change(struct net_device *dev); 4945 void __netdev_notify_peers(struct net_device *dev); 4946 void netdev_notify_peers(struct net_device *dev); 4947 void netdev_features_change(struct net_device *dev); 4948 /* Load a device via the kmod */ 4949 void dev_load(struct net *net, const char *name); 4950 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4951 struct rtnl_link_stats64 *storage); 4952 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4953 const struct net_device_stats *netdev_stats); 4954 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4955 const struct pcpu_sw_netstats __percpu *netstats); 4956 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4957 4958 enum { 4959 NESTED_SYNC_IMM_BIT, 4960 NESTED_SYNC_TODO_BIT, 4961 }; 4962 4963 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4964 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4965 4966 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4967 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4968 4969 struct netdev_nested_priv { 4970 unsigned char flags; 4971 void *data; 4972 }; 4973 4974 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4975 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4976 struct list_head **iter); 4977 4978 /* iterate through upper list, must be called under RCU read lock */ 4979 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4980 for (iter = &(dev)->adj_list.upper, \ 4981 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4982 updev; \ 4983 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4984 4985 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4986 int (*fn)(struct net_device *upper_dev, 4987 struct netdev_nested_priv *priv), 4988 struct netdev_nested_priv *priv); 4989 4990 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4991 struct net_device *upper_dev); 4992 4993 bool netdev_has_any_upper_dev(struct net_device *dev); 4994 4995 void *netdev_lower_get_next_private(struct net_device *dev, 4996 struct list_head **iter); 4997 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4998 struct list_head **iter); 4999 5000 #define netdev_for_each_lower_private(dev, priv, iter) \ 5001 for (iter = (dev)->adj_list.lower.next, \ 5002 priv = netdev_lower_get_next_private(dev, &(iter)); \ 5003 priv; \ 5004 priv = netdev_lower_get_next_private(dev, &(iter))) 5005 5006 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 5007 for (iter = &(dev)->adj_list.lower, \ 5008 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 5009 priv; \ 5010 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 5011 5012 void *netdev_lower_get_next(struct net_device *dev, 5013 struct list_head **iter); 5014 5015 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 5016 for (iter = (dev)->adj_list.lower.next, \ 5017 ldev = netdev_lower_get_next(dev, &(iter)); \ 5018 ldev; \ 5019 ldev = netdev_lower_get_next(dev, &(iter))) 5020 5021 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 5022 struct list_head **iter); 5023 int netdev_walk_all_lower_dev(struct net_device *dev, 5024 int (*fn)(struct net_device *lower_dev, 5025 struct netdev_nested_priv *priv), 5026 struct netdev_nested_priv *priv); 5027 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 5028 int (*fn)(struct net_device *lower_dev, 5029 struct netdev_nested_priv *priv), 5030 struct netdev_nested_priv *priv); 5031 5032 void *netdev_adjacent_get_private(struct list_head *adj_list); 5033 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 5034 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 5035 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 5036 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 5037 struct netlink_ext_ack *extack); 5038 int netdev_master_upper_dev_link(struct net_device *dev, 5039 struct net_device *upper_dev, 5040 void *upper_priv, void *upper_info, 5041 struct netlink_ext_ack *extack); 5042 void netdev_upper_dev_unlink(struct net_device *dev, 5043 struct net_device *upper_dev); 5044 int netdev_adjacent_change_prepare(struct net_device *old_dev, 5045 struct net_device *new_dev, 5046 struct net_device *dev, 5047 struct netlink_ext_ack *extack); 5048 void netdev_adjacent_change_commit(struct net_device *old_dev, 5049 struct net_device *new_dev, 5050 struct net_device *dev); 5051 void netdev_adjacent_change_abort(struct net_device *old_dev, 5052 struct net_device *new_dev, 5053 struct net_device *dev); 5054 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 5055 void *netdev_lower_dev_get_private(struct net_device *dev, 5056 struct net_device *lower_dev); 5057 void netdev_lower_state_changed(struct net_device *lower_dev, 5058 void *lower_state_info); 5059 5060 /* RSS keys are 40 or 52 bytes long */ 5061 #define NETDEV_RSS_KEY_LEN 52 5062 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 5063 void netdev_rss_key_fill(void *buffer, size_t len); 5064 5065 int skb_checksum_help(struct sk_buff *skb); 5066 int skb_crc32c_csum_help(struct sk_buff *skb); 5067 int skb_csum_hwoffload_help(struct sk_buff *skb, 5068 const netdev_features_t features); 5069 5070 struct netdev_bonding_info { 5071 ifslave slave; 5072 ifbond master; 5073 }; 5074 5075 struct netdev_notifier_bonding_info { 5076 struct netdev_notifier_info info; /* must be first */ 5077 struct netdev_bonding_info bonding_info; 5078 }; 5079 5080 void netdev_bonding_info_change(struct net_device *dev, 5081 struct netdev_bonding_info *bonding_info); 5082 5083 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 5084 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 5085 #else 5086 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 5087 const void *data) 5088 { 5089 } 5090 #endif 5091 5092 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 5093 5094 static inline bool can_checksum_protocol(netdev_features_t features, 5095 __be16 protocol) 5096 { 5097 if (protocol == htons(ETH_P_FCOE)) 5098 return !!(features & NETIF_F_FCOE_CRC); 5099 5100 /* Assume this is an IP checksum (not SCTP CRC) */ 5101 5102 if (features & NETIF_F_HW_CSUM) { 5103 /* Can checksum everything */ 5104 return true; 5105 } 5106 5107 switch (protocol) { 5108 case htons(ETH_P_IP): 5109 return !!(features & NETIF_F_IP_CSUM); 5110 case htons(ETH_P_IPV6): 5111 return !!(features & NETIF_F_IPV6_CSUM); 5112 default: 5113 return false; 5114 } 5115 } 5116 5117 #ifdef CONFIG_BUG 5118 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 5119 #else 5120 static inline void netdev_rx_csum_fault(struct net_device *dev, 5121 struct sk_buff *skb) 5122 { 5123 } 5124 #endif 5125 /* rx skb timestamps */ 5126 void net_enable_timestamp(void); 5127 void net_disable_timestamp(void); 5128 5129 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 5130 const struct skb_shared_hwtstamps *hwtstamps, 5131 bool cycles) 5132 { 5133 const struct net_device_ops *ops = dev->netdev_ops; 5134 5135 if (ops->ndo_get_tstamp) 5136 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 5137 5138 return hwtstamps->hwtstamp; 5139 } 5140 5141 #ifndef CONFIG_PREEMPT_RT 5142 static inline void netdev_xmit_set_more(bool more) 5143 { 5144 __this_cpu_write(softnet_data.xmit.more, more); 5145 } 5146 5147 static inline bool netdev_xmit_more(void) 5148 { 5149 return __this_cpu_read(softnet_data.xmit.more); 5150 } 5151 #else 5152 static inline void netdev_xmit_set_more(bool more) 5153 { 5154 current->net_xmit.more = more; 5155 } 5156 5157 static inline bool netdev_xmit_more(void) 5158 { 5159 return current->net_xmit.more; 5160 } 5161 #endif 5162 5163 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 5164 struct sk_buff *skb, struct net_device *dev, 5165 bool more) 5166 { 5167 netdev_xmit_set_more(more); 5168 return ops->ndo_start_xmit(skb, dev); 5169 } 5170 5171 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 5172 struct netdev_queue *txq, bool more) 5173 { 5174 const struct net_device_ops *ops = dev->netdev_ops; 5175 netdev_tx_t rc; 5176 5177 rc = __netdev_start_xmit(ops, skb, dev, more); 5178 if (rc == NETDEV_TX_OK) 5179 txq_trans_update(txq); 5180 5181 return rc; 5182 } 5183 5184 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 5185 const void *ns); 5186 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 5187 const void *ns); 5188 5189 extern const struct kobj_ns_type_operations net_ns_type_operations; 5190 5191 const char *netdev_drivername(const struct net_device *dev); 5192 5193 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 5194 netdev_features_t f2) 5195 { 5196 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 5197 if (f1 & NETIF_F_HW_CSUM) 5198 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5199 else 5200 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5201 } 5202 5203 return f1 & f2; 5204 } 5205 5206 static inline netdev_features_t netdev_get_wanted_features( 5207 struct net_device *dev) 5208 { 5209 return (dev->features & ~dev->hw_features) | dev->wanted_features; 5210 } 5211 netdev_features_t netdev_increment_features(netdev_features_t all, 5212 netdev_features_t one, netdev_features_t mask); 5213 5214 /* Allow TSO being used on stacked device : 5215 * Performing the GSO segmentation before last device 5216 * is a performance improvement. 5217 */ 5218 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 5219 netdev_features_t mask) 5220 { 5221 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 5222 } 5223 5224 int __netdev_update_features(struct net_device *dev); 5225 void netdev_update_features(struct net_device *dev); 5226 void netdev_change_features(struct net_device *dev); 5227 5228 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 5229 struct net_device *dev); 5230 5231 netdev_features_t passthru_features_check(struct sk_buff *skb, 5232 struct net_device *dev, 5233 netdev_features_t features); 5234 netdev_features_t netif_skb_features(struct sk_buff *skb); 5235 void skb_warn_bad_offload(const struct sk_buff *skb); 5236 5237 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 5238 { 5239 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 5240 5241 /* check flags correspondence */ 5242 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 5243 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 5244 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 5245 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 5246 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 5247 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 5248 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 5249 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 5250 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 5251 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 5252 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 5253 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 5254 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 5255 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 5256 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 5257 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 5258 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 5259 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 5260 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 5261 5262 return (features & feature) == feature; 5263 } 5264 5265 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 5266 { 5267 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 5268 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 5269 } 5270 5271 static inline bool netif_needs_gso(struct sk_buff *skb, 5272 netdev_features_t features) 5273 { 5274 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 5275 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 5276 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 5277 } 5278 5279 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 5280 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 5281 void netif_inherit_tso_max(struct net_device *to, 5282 const struct net_device *from); 5283 5284 static inline unsigned int 5285 netif_get_gro_max_size(const struct net_device *dev, const struct sk_buff *skb) 5286 { 5287 /* pairs with WRITE_ONCE() in netif_set_gro(_ipv4)_max_size() */ 5288 return skb->protocol == htons(ETH_P_IPV6) ? 5289 READ_ONCE(dev->gro_max_size) : 5290 READ_ONCE(dev->gro_ipv4_max_size); 5291 } 5292 5293 static inline unsigned int 5294 netif_get_gso_max_size(const struct net_device *dev, const struct sk_buff *skb) 5295 { 5296 /* pairs with WRITE_ONCE() in netif_set_gso(_ipv4)_max_size() */ 5297 return skb->protocol == htons(ETH_P_IPV6) ? 5298 READ_ONCE(dev->gso_max_size) : 5299 READ_ONCE(dev->gso_ipv4_max_size); 5300 } 5301 5302 static inline bool netif_is_macsec(const struct net_device *dev) 5303 { 5304 return dev->priv_flags & IFF_MACSEC; 5305 } 5306 5307 static inline bool netif_is_macvlan(const struct net_device *dev) 5308 { 5309 return dev->priv_flags & IFF_MACVLAN; 5310 } 5311 5312 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5313 { 5314 return dev->priv_flags & IFF_MACVLAN_PORT; 5315 } 5316 5317 static inline bool netif_is_bond_master(const struct net_device *dev) 5318 { 5319 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5320 } 5321 5322 static inline bool netif_is_bond_slave(const struct net_device *dev) 5323 { 5324 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5325 } 5326 5327 static inline bool netif_supports_nofcs(struct net_device *dev) 5328 { 5329 return dev->priv_flags & IFF_SUPP_NOFCS; 5330 } 5331 5332 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5333 { 5334 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5335 } 5336 5337 static inline bool netif_is_l3_master(const struct net_device *dev) 5338 { 5339 return dev->priv_flags & IFF_L3MDEV_MASTER; 5340 } 5341 5342 static inline bool netif_is_l3_slave(const struct net_device *dev) 5343 { 5344 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5345 } 5346 5347 static inline int dev_sdif(const struct net_device *dev) 5348 { 5349 #ifdef CONFIG_NET_L3_MASTER_DEV 5350 if (netif_is_l3_slave(dev)) 5351 return dev->ifindex; 5352 #endif 5353 return 0; 5354 } 5355 5356 static inline bool netif_is_bridge_master(const struct net_device *dev) 5357 { 5358 return dev->priv_flags & IFF_EBRIDGE; 5359 } 5360 5361 static inline bool netif_is_bridge_port(const struct net_device *dev) 5362 { 5363 return dev->priv_flags & IFF_BRIDGE_PORT; 5364 } 5365 5366 static inline bool netif_is_ovs_master(const struct net_device *dev) 5367 { 5368 return dev->priv_flags & IFF_OPENVSWITCH; 5369 } 5370 5371 static inline bool netif_is_ovs_port(const struct net_device *dev) 5372 { 5373 return dev->priv_flags & IFF_OVS_DATAPATH; 5374 } 5375 5376 static inline bool netif_is_any_bridge_master(const struct net_device *dev) 5377 { 5378 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev); 5379 } 5380 5381 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5382 { 5383 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5384 } 5385 5386 static inline bool netif_is_team_master(const struct net_device *dev) 5387 { 5388 return dev->priv_flags & IFF_TEAM; 5389 } 5390 5391 static inline bool netif_is_team_port(const struct net_device *dev) 5392 { 5393 return dev->priv_flags & IFF_TEAM_PORT; 5394 } 5395 5396 static inline bool netif_is_lag_master(const struct net_device *dev) 5397 { 5398 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5399 } 5400 5401 static inline bool netif_is_lag_port(const struct net_device *dev) 5402 { 5403 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5404 } 5405 5406 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5407 { 5408 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5409 } 5410 5411 static inline bool netif_is_failover(const struct net_device *dev) 5412 { 5413 return dev->priv_flags & IFF_FAILOVER; 5414 } 5415 5416 static inline bool netif_is_failover_slave(const struct net_device *dev) 5417 { 5418 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5419 } 5420 5421 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5422 static inline void netif_keep_dst(struct net_device *dev) 5423 { 5424 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5425 } 5426 5427 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5428 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5429 { 5430 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5431 return netif_is_macsec(dev); 5432 } 5433 5434 extern struct pernet_operations __net_initdata loopback_net_ops; 5435 5436 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5437 5438 /* netdev_printk helpers, similar to dev_printk */ 5439 5440 static inline const char *netdev_name(const struct net_device *dev) 5441 { 5442 if (!dev->name[0] || strchr(dev->name, '%')) 5443 return "(unnamed net_device)"; 5444 return dev->name; 5445 } 5446 5447 static inline const char *netdev_reg_state(const struct net_device *dev) 5448 { 5449 u8 reg_state = READ_ONCE(dev->reg_state); 5450 5451 switch (reg_state) { 5452 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5453 case NETREG_REGISTERED: return ""; 5454 case NETREG_UNREGISTERING: return " (unregistering)"; 5455 case NETREG_UNREGISTERED: return " (unregistered)"; 5456 case NETREG_RELEASED: return " (released)"; 5457 case NETREG_DUMMY: return " (dummy)"; 5458 } 5459 5460 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state); 5461 return " (unknown)"; 5462 } 5463 5464 #define MODULE_ALIAS_NETDEV(device) \ 5465 MODULE_ALIAS("netdev-" device) 5466 5467 /* 5468 * netdev_WARN() acts like dev_printk(), but with the key difference 5469 * of using a WARN/WARN_ON to get the message out, including the 5470 * file/line information and a backtrace. 5471 */ 5472 #define netdev_WARN(dev, format, args...) \ 5473 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5474 netdev_reg_state(dev), ##args) 5475 5476 #define netdev_WARN_ONCE(dev, format, args...) \ 5477 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5478 netdev_reg_state(dev), ##args) 5479 5480 /* 5481 * The list of packet types we will receive (as opposed to discard) 5482 * and the routines to invoke. 5483 * 5484 * Why 16. Because with 16 the only overlap we get on a hash of the 5485 * low nibble of the protocol value is RARP/SNAP/X.25. 5486 * 5487 * 0800 IP 5488 * 0001 802.3 5489 * 0002 AX.25 5490 * 0004 802.2 5491 * 8035 RARP 5492 * 0005 SNAP 5493 * 0805 X.25 5494 * 0806 ARP 5495 * 8137 IPX 5496 * 0009 Localtalk 5497 * 86DD IPv6 5498 */ 5499 #define PTYPE_HASH_SIZE (16) 5500 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5501 5502 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5503 5504 extern struct net_device *blackhole_netdev; 5505 5506 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */ 5507 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD) 5508 #define DEV_STATS_ADD(DEV, FIELD, VAL) \ 5509 atomic_long_add((VAL), &(DEV)->stats.__##FIELD) 5510 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD) 5511 5512 #endif /* _LINUX_NETDEVICE_H */ 5513