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 <uapi/linux/netdevice.h> 46 #include <uapi/linux/if_bonding.h> 47 #include <uapi/linux/pkt_cls.h> 48 #include <uapi/linux/netdev.h> 49 #include <linux/hashtable.h> 50 #include <linux/rbtree.h> 51 #include <net/net_trackers.h> 52 #include <net/net_debug.h> 53 #include <net/dropreason-core.h> 54 #include <linux/phy_link_topology_core.h> 55 56 struct netpoll_info; 57 struct device; 58 struct ethtool_ops; 59 struct kernel_hwtstamp_config; 60 struct phy_device; 61 struct dsa_port; 62 struct ip_tunnel_parm; 63 struct macsec_context; 64 struct macsec_ops; 65 struct netdev_name_node; 66 struct sd_flow_limit; 67 struct sfp_bus; 68 /* 802.11 specific */ 69 struct wireless_dev; 70 /* 802.15.4 specific */ 71 struct wpan_dev; 72 struct mpls_dev; 73 /* UDP Tunnel offloads */ 74 struct udp_tunnel_info; 75 struct udp_tunnel_nic_info; 76 struct udp_tunnel_nic; 77 struct bpf_prog; 78 struct xdp_buff; 79 struct xdp_frame; 80 struct xdp_metadata_ops; 81 struct xdp_md; 82 /* DPLL specific */ 83 struct dpll_pin; 84 85 typedef u32 xdp_features_t; 86 87 void synchronize_net(void); 88 void netdev_set_default_ethtool_ops(struct net_device *dev, 89 const struct ethtool_ops *ops); 90 void netdev_sw_irq_coalesce_default_on(struct net_device *dev); 91 92 /* Backlog congestion levels */ 93 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 94 #define NET_RX_DROP 1 /* packet dropped */ 95 96 #define MAX_NEST_DEV 8 97 98 /* 99 * Transmit return codes: transmit return codes originate from three different 100 * namespaces: 101 * 102 * - qdisc return codes 103 * - driver transmit return codes 104 * - errno values 105 * 106 * Drivers are allowed to return any one of those in their hard_start_xmit() 107 * function. Real network devices commonly used with qdiscs should only return 108 * the driver transmit return codes though - when qdiscs are used, the actual 109 * transmission happens asynchronously, so the value is not propagated to 110 * higher layers. Virtual network devices transmit synchronously; in this case 111 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 112 * others are propagated to higher layers. 113 */ 114 115 /* qdisc ->enqueue() return codes. */ 116 #define NET_XMIT_SUCCESS 0x00 117 #define NET_XMIT_DROP 0x01 /* skb dropped */ 118 #define NET_XMIT_CN 0x02 /* congestion notification */ 119 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 120 121 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 122 * indicates that the device will soon be dropping packets, or already drops 123 * some packets of the same priority; prompting us to send less aggressively. */ 124 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 125 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 126 127 /* Driver transmit return codes */ 128 #define NETDEV_TX_MASK 0xf0 129 130 enum netdev_tx { 131 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 132 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 133 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 134 }; 135 typedef enum netdev_tx netdev_tx_t; 136 137 /* 138 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 139 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 140 */ 141 static inline bool dev_xmit_complete(int rc) 142 { 143 /* 144 * Positive cases with an skb consumed by a driver: 145 * - successful transmission (rc == NETDEV_TX_OK) 146 * - error while transmitting (rc < 0) 147 * - error while queueing to a different device (rc & NET_XMIT_MASK) 148 */ 149 if (likely(rc < NET_XMIT_MASK)) 150 return true; 151 152 return false; 153 } 154 155 /* 156 * Compute the worst-case header length according to the protocols 157 * used. 158 */ 159 160 #if defined(CONFIG_HYPERV_NET) 161 # define LL_MAX_HEADER 128 162 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 163 # if defined(CONFIG_MAC80211_MESH) 164 # define LL_MAX_HEADER 128 165 # else 166 # define LL_MAX_HEADER 96 167 # endif 168 #else 169 # define LL_MAX_HEADER 32 170 #endif 171 172 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 173 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 174 #define MAX_HEADER LL_MAX_HEADER 175 #else 176 #define MAX_HEADER (LL_MAX_HEADER + 48) 177 #endif 178 179 /* 180 * Old network device statistics. Fields are native words 181 * (unsigned long) so they can be read and written atomically. 182 */ 183 184 #define NET_DEV_STAT(FIELD) \ 185 union { \ 186 unsigned long FIELD; \ 187 atomic_long_t __##FIELD; \ 188 } 189 190 struct net_device_stats { 191 NET_DEV_STAT(rx_packets); 192 NET_DEV_STAT(tx_packets); 193 NET_DEV_STAT(rx_bytes); 194 NET_DEV_STAT(tx_bytes); 195 NET_DEV_STAT(rx_errors); 196 NET_DEV_STAT(tx_errors); 197 NET_DEV_STAT(rx_dropped); 198 NET_DEV_STAT(tx_dropped); 199 NET_DEV_STAT(multicast); 200 NET_DEV_STAT(collisions); 201 NET_DEV_STAT(rx_length_errors); 202 NET_DEV_STAT(rx_over_errors); 203 NET_DEV_STAT(rx_crc_errors); 204 NET_DEV_STAT(rx_frame_errors); 205 NET_DEV_STAT(rx_fifo_errors); 206 NET_DEV_STAT(rx_missed_errors); 207 NET_DEV_STAT(tx_aborted_errors); 208 NET_DEV_STAT(tx_carrier_errors); 209 NET_DEV_STAT(tx_fifo_errors); 210 NET_DEV_STAT(tx_heartbeat_errors); 211 NET_DEV_STAT(tx_window_errors); 212 NET_DEV_STAT(rx_compressed); 213 NET_DEV_STAT(tx_compressed); 214 }; 215 #undef NET_DEV_STAT 216 217 /* per-cpu stats, allocated on demand. 218 * Try to fit them in a single cache line, for dev_get_stats() sake. 219 */ 220 struct net_device_core_stats { 221 unsigned long rx_dropped; 222 unsigned long tx_dropped; 223 unsigned long rx_nohandler; 224 unsigned long rx_otherhost_dropped; 225 } __aligned(4 * sizeof(unsigned long)); 226 227 #include <linux/cache.h> 228 #include <linux/skbuff.h> 229 230 #ifdef CONFIG_RPS 231 #include <linux/static_key.h> 232 extern struct static_key_false rps_needed; 233 extern struct static_key_false rfs_needed; 234 #endif 235 236 struct neighbour; 237 struct neigh_parms; 238 struct sk_buff; 239 240 struct netdev_hw_addr { 241 struct list_head list; 242 struct rb_node node; 243 unsigned char addr[MAX_ADDR_LEN]; 244 unsigned char type; 245 #define NETDEV_HW_ADDR_T_LAN 1 246 #define NETDEV_HW_ADDR_T_SAN 2 247 #define NETDEV_HW_ADDR_T_UNICAST 3 248 #define NETDEV_HW_ADDR_T_MULTICAST 4 249 bool global_use; 250 int sync_cnt; 251 int refcount; 252 int synced; 253 struct rcu_head rcu_head; 254 }; 255 256 struct netdev_hw_addr_list { 257 struct list_head list; 258 int count; 259 260 /* Auxiliary tree for faster lookup on addition and deletion */ 261 struct rb_root tree; 262 }; 263 264 #define netdev_hw_addr_list_count(l) ((l)->count) 265 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 266 #define netdev_hw_addr_list_for_each(ha, l) \ 267 list_for_each_entry(ha, &(l)->list, list) 268 269 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 270 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 271 #define netdev_for_each_uc_addr(ha, dev) \ 272 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 273 #define netdev_for_each_synced_uc_addr(_ha, _dev) \ 274 netdev_for_each_uc_addr((_ha), (_dev)) \ 275 if ((_ha)->sync_cnt) 276 277 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 278 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 279 #define netdev_for_each_mc_addr(ha, dev) \ 280 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 281 #define netdev_for_each_synced_mc_addr(_ha, _dev) \ 282 netdev_for_each_mc_addr((_ha), (_dev)) \ 283 if ((_ha)->sync_cnt) 284 285 struct hh_cache { 286 unsigned int hh_len; 287 seqlock_t hh_lock; 288 289 /* cached hardware header; allow for machine alignment needs. */ 290 #define HH_DATA_MOD 16 291 #define HH_DATA_OFF(__len) \ 292 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 293 #define HH_DATA_ALIGN(__len) \ 294 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 295 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 296 }; 297 298 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 299 * Alternative is: 300 * dev->hard_header_len ? (dev->hard_header_len + 301 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 302 * 303 * We could use other alignment values, but we must maintain the 304 * relationship HH alignment <= LL alignment. 305 */ 306 #define LL_RESERVED_SPACE(dev) \ 307 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \ 308 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 309 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 310 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \ 311 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 312 313 struct header_ops { 314 int (*create) (struct sk_buff *skb, struct net_device *dev, 315 unsigned short type, const void *daddr, 316 const void *saddr, unsigned int len); 317 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 318 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 319 void (*cache_update)(struct hh_cache *hh, 320 const struct net_device *dev, 321 const unsigned char *haddr); 322 bool (*validate)(const char *ll_header, unsigned int len); 323 __be16 (*parse_protocol)(const struct sk_buff *skb); 324 }; 325 326 /* These flag bits are private to the generic network queueing 327 * layer; they may not be explicitly referenced by any other 328 * code. 329 */ 330 331 enum netdev_state_t { 332 __LINK_STATE_START, 333 __LINK_STATE_PRESENT, 334 __LINK_STATE_NOCARRIER, 335 __LINK_STATE_LINKWATCH_PENDING, 336 __LINK_STATE_DORMANT, 337 __LINK_STATE_TESTING, 338 }; 339 340 struct gro_list { 341 struct list_head list; 342 int count; 343 }; 344 345 /* 346 * size of gro hash buckets, must less than bit number of 347 * napi_struct::gro_bitmask 348 */ 349 #define GRO_HASH_BUCKETS 8 350 351 /* 352 * Structure for NAPI scheduling similar to tasklet but with weighting 353 */ 354 struct napi_struct { 355 /* The poll_list must only be managed by the entity which 356 * changes the state of the NAPI_STATE_SCHED bit. This means 357 * whoever atomically sets that bit can add this napi_struct 358 * to the per-CPU poll_list, and whoever clears that bit 359 * can remove from the list right before clearing the bit. 360 */ 361 struct list_head poll_list; 362 363 unsigned long state; 364 int weight; 365 int defer_hard_irqs_count; 366 unsigned long gro_bitmask; 367 int (*poll)(struct napi_struct *, int); 368 #ifdef CONFIG_NETPOLL 369 /* CPU actively polling if netpoll is configured */ 370 int poll_owner; 371 #endif 372 /* CPU on which NAPI has been scheduled for processing */ 373 int list_owner; 374 struct net_device *dev; 375 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 376 struct sk_buff *skb; 377 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 378 int rx_count; /* length of rx_list */ 379 unsigned int napi_id; 380 struct hrtimer timer; 381 struct task_struct *thread; 382 /* control-path-only fields follow */ 383 struct list_head dev_list; 384 struct hlist_node napi_hash_node; 385 int irq; 386 }; 387 388 enum { 389 NAPI_STATE_SCHED, /* Poll is scheduled */ 390 NAPI_STATE_MISSED, /* reschedule a napi */ 391 NAPI_STATE_DISABLE, /* Disable pending */ 392 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 393 NAPI_STATE_LISTED, /* NAPI added to system lists */ 394 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 395 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 396 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 397 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 398 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 399 }; 400 401 enum { 402 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 403 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 404 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 405 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 406 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 407 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 408 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 409 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 410 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 411 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 412 }; 413 414 enum gro_result { 415 GRO_MERGED, 416 GRO_MERGED_FREE, 417 GRO_HELD, 418 GRO_NORMAL, 419 GRO_CONSUMED, 420 }; 421 typedef enum gro_result gro_result_t; 422 423 /* 424 * enum rx_handler_result - Possible return values for rx_handlers. 425 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 426 * further. 427 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 428 * case skb->dev was changed by rx_handler. 429 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 430 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 431 * 432 * rx_handlers are functions called from inside __netif_receive_skb(), to do 433 * special processing of the skb, prior to delivery to protocol handlers. 434 * 435 * Currently, a net_device can only have a single rx_handler registered. Trying 436 * to register a second rx_handler will return -EBUSY. 437 * 438 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 439 * To unregister a rx_handler on a net_device, use 440 * netdev_rx_handler_unregister(). 441 * 442 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 443 * do with the skb. 444 * 445 * If the rx_handler consumed the skb in some way, it should return 446 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 447 * the skb to be delivered in some other way. 448 * 449 * If the rx_handler changed skb->dev, to divert the skb to another 450 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 451 * new device will be called if it exists. 452 * 453 * If the rx_handler decides the skb should be ignored, it should return 454 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 455 * are registered on exact device (ptype->dev == skb->dev). 456 * 457 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 458 * delivered, it should return RX_HANDLER_PASS. 459 * 460 * A device without a registered rx_handler will behave as if rx_handler 461 * returned RX_HANDLER_PASS. 462 */ 463 464 enum rx_handler_result { 465 RX_HANDLER_CONSUMED, 466 RX_HANDLER_ANOTHER, 467 RX_HANDLER_EXACT, 468 RX_HANDLER_PASS, 469 }; 470 typedef enum rx_handler_result rx_handler_result_t; 471 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 472 473 void __napi_schedule(struct napi_struct *n); 474 void __napi_schedule_irqoff(struct napi_struct *n); 475 476 static inline bool napi_disable_pending(struct napi_struct *n) 477 { 478 return test_bit(NAPI_STATE_DISABLE, &n->state); 479 } 480 481 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 482 { 483 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 484 } 485 486 /** 487 * napi_is_scheduled - test if NAPI is scheduled 488 * @n: NAPI context 489 * 490 * This check is "best-effort". With no locking implemented, 491 * a NAPI can be scheduled or terminate right after this check 492 * and produce not precise results. 493 * 494 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled 495 * should not be used normally and napi_schedule should be 496 * used instead. 497 * 498 * Use only if the driver really needs to check if a NAPI 499 * is scheduled for example in the context of delayed timer 500 * that can be skipped if a NAPI is already scheduled. 501 * 502 * Return True if NAPI is scheduled, False otherwise. 503 */ 504 static inline bool napi_is_scheduled(struct napi_struct *n) 505 { 506 return test_bit(NAPI_STATE_SCHED, &n->state); 507 } 508 509 bool napi_schedule_prep(struct napi_struct *n); 510 511 /** 512 * napi_schedule - schedule NAPI poll 513 * @n: NAPI context 514 * 515 * Schedule NAPI poll routine to be called if it is not already 516 * running. 517 * Return true if we schedule a NAPI or false if not. 518 * Refer to napi_schedule_prep() for additional reason on why 519 * a NAPI might not be scheduled. 520 */ 521 static inline bool napi_schedule(struct napi_struct *n) 522 { 523 if (napi_schedule_prep(n)) { 524 __napi_schedule(n); 525 return true; 526 } 527 528 return false; 529 } 530 531 /** 532 * napi_schedule_irqoff - schedule NAPI poll 533 * @n: NAPI context 534 * 535 * Variant of napi_schedule(), assuming hard irqs are masked. 536 */ 537 static inline void napi_schedule_irqoff(struct napi_struct *n) 538 { 539 if (napi_schedule_prep(n)) 540 __napi_schedule_irqoff(n); 541 } 542 543 /** 544 * napi_complete_done - NAPI processing complete 545 * @n: NAPI context 546 * @work_done: number of packets processed 547 * 548 * Mark NAPI processing as complete. Should only be called if poll budget 549 * has not been completely consumed. 550 * Prefer over napi_complete(). 551 * Return false if device should avoid rearming interrupts. 552 */ 553 bool napi_complete_done(struct napi_struct *n, int work_done); 554 555 static inline bool napi_complete(struct napi_struct *n) 556 { 557 return napi_complete_done(n, 0); 558 } 559 560 int dev_set_threaded(struct net_device *dev, bool threaded); 561 562 /** 563 * napi_disable - prevent NAPI from scheduling 564 * @n: NAPI context 565 * 566 * Stop NAPI from being scheduled on this context. 567 * Waits till any outstanding processing completes. 568 */ 569 void napi_disable(struct napi_struct *n); 570 571 void napi_enable(struct napi_struct *n); 572 573 /** 574 * napi_synchronize - wait until NAPI is not running 575 * @n: NAPI context 576 * 577 * Wait until NAPI is done being scheduled on this context. 578 * Waits till any outstanding processing completes but 579 * does not disable future activations. 580 */ 581 static inline void napi_synchronize(const struct napi_struct *n) 582 { 583 if (IS_ENABLED(CONFIG_SMP)) 584 while (test_bit(NAPI_STATE_SCHED, &n->state)) 585 msleep(1); 586 else 587 barrier(); 588 } 589 590 /** 591 * napi_if_scheduled_mark_missed - if napi is running, set the 592 * NAPIF_STATE_MISSED 593 * @n: NAPI context 594 * 595 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 596 * NAPI is scheduled. 597 **/ 598 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 599 { 600 unsigned long val, new; 601 602 val = READ_ONCE(n->state); 603 do { 604 if (val & NAPIF_STATE_DISABLE) 605 return true; 606 607 if (!(val & NAPIF_STATE_SCHED)) 608 return false; 609 610 new = val | NAPIF_STATE_MISSED; 611 } while (!try_cmpxchg(&n->state, &val, new)); 612 613 return true; 614 } 615 616 enum netdev_queue_state_t { 617 __QUEUE_STATE_DRV_XOFF, 618 __QUEUE_STATE_STACK_XOFF, 619 __QUEUE_STATE_FROZEN, 620 }; 621 622 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 623 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 624 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 625 626 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 627 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 628 QUEUE_STATE_FROZEN) 629 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 630 QUEUE_STATE_FROZEN) 631 632 /* 633 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 634 * netif_tx_* functions below are used to manipulate this flag. The 635 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 636 * queue independently. The netif_xmit_*stopped functions below are called 637 * to check if the queue has been stopped by the driver or stack (either 638 * of the XOFF bits are set in the state). Drivers should not need to call 639 * netif_xmit*stopped functions, they should only be using netif_tx_*. 640 */ 641 642 struct netdev_queue { 643 /* 644 * read-mostly part 645 */ 646 struct net_device *dev; 647 netdevice_tracker dev_tracker; 648 649 struct Qdisc __rcu *qdisc; 650 struct Qdisc __rcu *qdisc_sleeping; 651 #ifdef CONFIG_SYSFS 652 struct kobject kobj; 653 #endif 654 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 655 int numa_node; 656 #endif 657 unsigned long tx_maxrate; 658 /* 659 * Number of TX timeouts for this queue 660 * (/sys/class/net/DEV/Q/trans_timeout) 661 */ 662 atomic_long_t trans_timeout; 663 664 /* Subordinate device that the queue has been assigned to */ 665 struct net_device *sb_dev; 666 #ifdef CONFIG_XDP_SOCKETS 667 struct xsk_buff_pool *pool; 668 #endif 669 /* NAPI instance for the queue 670 * Readers and writers must hold RTNL 671 */ 672 struct napi_struct *napi; 673 /* 674 * write-mostly part 675 */ 676 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 677 int xmit_lock_owner; 678 /* 679 * Time (in jiffies) of last Tx 680 */ 681 unsigned long trans_start; 682 683 unsigned long state; 684 685 #ifdef CONFIG_BQL 686 struct dql dql; 687 #endif 688 } ____cacheline_aligned_in_smp; 689 690 extern int sysctl_fb_tunnels_only_for_init_net; 691 extern int sysctl_devconf_inherit_init_net; 692 693 /* 694 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 695 * == 1 : For initns only 696 * == 2 : For none. 697 */ 698 static inline bool net_has_fallback_tunnels(const struct net *net) 699 { 700 #if IS_ENABLED(CONFIG_SYSCTL) 701 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net); 702 703 return !fb_tunnels_only_for_init_net || 704 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1); 705 #else 706 return true; 707 #endif 708 } 709 710 static inline int net_inherit_devconf(void) 711 { 712 #if IS_ENABLED(CONFIG_SYSCTL) 713 return READ_ONCE(sysctl_devconf_inherit_init_net); 714 #else 715 return 0; 716 #endif 717 } 718 719 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 720 { 721 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 722 return q->numa_node; 723 #else 724 return NUMA_NO_NODE; 725 #endif 726 } 727 728 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 729 { 730 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 731 q->numa_node = node; 732 #endif 733 } 734 735 #ifdef CONFIG_RPS 736 /* 737 * This structure holds an RPS map which can be of variable length. The 738 * map is an array of CPUs. 739 */ 740 struct rps_map { 741 unsigned int len; 742 struct rcu_head rcu; 743 u16 cpus[]; 744 }; 745 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 746 747 /* 748 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 749 * tail pointer for that CPU's input queue at the time of last enqueue, and 750 * a hardware filter index. 751 */ 752 struct rps_dev_flow { 753 u16 cpu; 754 u16 filter; 755 unsigned int last_qtail; 756 }; 757 #define RPS_NO_FILTER 0xffff 758 759 /* 760 * The rps_dev_flow_table structure contains a table of flow mappings. 761 */ 762 struct rps_dev_flow_table { 763 unsigned int mask; 764 struct rcu_head rcu; 765 struct rps_dev_flow flows[]; 766 }; 767 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 768 ((_num) * sizeof(struct rps_dev_flow))) 769 770 /* 771 * The rps_sock_flow_table contains mappings of flows to the last CPU 772 * on which they were processed by the application (set in recvmsg). 773 * Each entry is a 32bit value. Upper part is the high-order bits 774 * of flow hash, lower part is CPU number. 775 * rps_cpu_mask is used to partition the space, depending on number of 776 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 777 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 778 * meaning we use 32-6=26 bits for the hash. 779 */ 780 struct rps_sock_flow_table { 781 u32 mask; 782 783 u32 ents[] ____cacheline_aligned_in_smp; 784 }; 785 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 786 787 #define RPS_NO_CPU 0xffff 788 789 extern u32 rps_cpu_mask; 790 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 791 792 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 793 u32 hash) 794 { 795 if (table && hash) { 796 unsigned int index = hash & table->mask; 797 u32 val = hash & ~rps_cpu_mask; 798 799 /* We only give a hint, preemption can change CPU under us */ 800 val |= raw_smp_processor_id(); 801 802 /* The following WRITE_ONCE() is paired with the READ_ONCE() 803 * here, and another one in get_rps_cpu(). 804 */ 805 if (READ_ONCE(table->ents[index]) != val) 806 WRITE_ONCE(table->ents[index], val); 807 } 808 } 809 810 #ifdef CONFIG_RFS_ACCEL 811 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 812 u16 filter_id); 813 #endif 814 #endif /* CONFIG_RPS */ 815 816 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 817 enum xps_map_type { 818 XPS_CPUS = 0, 819 XPS_RXQS, 820 XPS_MAPS_MAX, 821 }; 822 823 #ifdef CONFIG_XPS 824 /* 825 * This structure holds an XPS map which can be of variable length. The 826 * map is an array of queues. 827 */ 828 struct xps_map { 829 unsigned int len; 830 unsigned int alloc_len; 831 struct rcu_head rcu; 832 u16 queues[]; 833 }; 834 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 835 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 836 - sizeof(struct xps_map)) / sizeof(u16)) 837 838 /* 839 * This structure holds all XPS maps for device. Maps are indexed by CPU. 840 * 841 * We keep track of the number of cpus/rxqs used when the struct is allocated, 842 * in nr_ids. This will help not accessing out-of-bound memory. 843 * 844 * We keep track of the number of traffic classes used when the struct is 845 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 846 * not crossing its upper bound, as the original dev->num_tc can be updated in 847 * the meantime. 848 */ 849 struct xps_dev_maps { 850 struct rcu_head rcu; 851 unsigned int nr_ids; 852 s16 num_tc; 853 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 854 }; 855 856 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 857 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 858 859 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 860 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 861 862 #endif /* CONFIG_XPS */ 863 864 #define TC_MAX_QUEUE 16 865 #define TC_BITMASK 15 866 /* HW offloaded queuing disciplines txq count and offset maps */ 867 struct netdev_tc_txq { 868 u16 count; 869 u16 offset; 870 }; 871 872 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 873 /* 874 * This structure is to hold information about the device 875 * configured to run FCoE protocol stack. 876 */ 877 struct netdev_fcoe_hbainfo { 878 char manufacturer[64]; 879 char serial_number[64]; 880 char hardware_version[64]; 881 char driver_version[64]; 882 char optionrom_version[64]; 883 char firmware_version[64]; 884 char model[256]; 885 char model_description[256]; 886 }; 887 #endif 888 889 #define MAX_PHYS_ITEM_ID_LEN 32 890 891 /* This structure holds a unique identifier to identify some 892 * physical item (port for example) used by a netdevice. 893 */ 894 struct netdev_phys_item_id { 895 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 896 unsigned char id_len; 897 }; 898 899 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 900 struct netdev_phys_item_id *b) 901 { 902 return a->id_len == b->id_len && 903 memcmp(a->id, b->id, a->id_len) == 0; 904 } 905 906 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 907 struct sk_buff *skb, 908 struct net_device *sb_dev); 909 910 enum net_device_path_type { 911 DEV_PATH_ETHERNET = 0, 912 DEV_PATH_VLAN, 913 DEV_PATH_BRIDGE, 914 DEV_PATH_PPPOE, 915 DEV_PATH_DSA, 916 DEV_PATH_MTK_WDMA, 917 }; 918 919 struct net_device_path { 920 enum net_device_path_type type; 921 const struct net_device *dev; 922 union { 923 struct { 924 u16 id; 925 __be16 proto; 926 u8 h_dest[ETH_ALEN]; 927 } encap; 928 struct { 929 enum { 930 DEV_PATH_BR_VLAN_KEEP, 931 DEV_PATH_BR_VLAN_TAG, 932 DEV_PATH_BR_VLAN_UNTAG, 933 DEV_PATH_BR_VLAN_UNTAG_HW, 934 } vlan_mode; 935 u16 vlan_id; 936 __be16 vlan_proto; 937 } bridge; 938 struct { 939 int port; 940 u16 proto; 941 } dsa; 942 struct { 943 u8 wdma_idx; 944 u8 queue; 945 u16 wcid; 946 u8 bss; 947 u8 amsdu; 948 } mtk_wdma; 949 }; 950 }; 951 952 #define NET_DEVICE_PATH_STACK_MAX 5 953 #define NET_DEVICE_PATH_VLAN_MAX 2 954 955 struct net_device_path_stack { 956 int num_paths; 957 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 958 }; 959 960 struct net_device_path_ctx { 961 const struct net_device *dev; 962 u8 daddr[ETH_ALEN]; 963 964 int num_vlans; 965 struct { 966 u16 id; 967 __be16 proto; 968 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 969 }; 970 971 enum tc_setup_type { 972 TC_QUERY_CAPS, 973 TC_SETUP_QDISC_MQPRIO, 974 TC_SETUP_CLSU32, 975 TC_SETUP_CLSFLOWER, 976 TC_SETUP_CLSMATCHALL, 977 TC_SETUP_CLSBPF, 978 TC_SETUP_BLOCK, 979 TC_SETUP_QDISC_CBS, 980 TC_SETUP_QDISC_RED, 981 TC_SETUP_QDISC_PRIO, 982 TC_SETUP_QDISC_MQ, 983 TC_SETUP_QDISC_ETF, 984 TC_SETUP_ROOT_QDISC, 985 TC_SETUP_QDISC_GRED, 986 TC_SETUP_QDISC_TAPRIO, 987 TC_SETUP_FT, 988 TC_SETUP_QDISC_ETS, 989 TC_SETUP_QDISC_TBF, 990 TC_SETUP_QDISC_FIFO, 991 TC_SETUP_QDISC_HTB, 992 TC_SETUP_ACT, 993 }; 994 995 /* These structures hold the attributes of bpf state that are being passed 996 * to the netdevice through the bpf op. 997 */ 998 enum bpf_netdev_command { 999 /* Set or clear a bpf program used in the earliest stages of packet 1000 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 1001 * is responsible for calling bpf_prog_put on any old progs that are 1002 * stored. In case of error, the callee need not release the new prog 1003 * reference, but on success it takes ownership and must bpf_prog_put 1004 * when it is no longer used. 1005 */ 1006 XDP_SETUP_PROG, 1007 XDP_SETUP_PROG_HW, 1008 /* BPF program for offload callbacks, invoked at program load time. */ 1009 BPF_OFFLOAD_MAP_ALLOC, 1010 BPF_OFFLOAD_MAP_FREE, 1011 XDP_SETUP_XSK_POOL, 1012 }; 1013 1014 struct bpf_prog_offload_ops; 1015 struct netlink_ext_ack; 1016 struct xdp_umem; 1017 struct xdp_dev_bulk_queue; 1018 struct bpf_xdp_link; 1019 1020 enum bpf_xdp_mode { 1021 XDP_MODE_SKB = 0, 1022 XDP_MODE_DRV = 1, 1023 XDP_MODE_HW = 2, 1024 __MAX_XDP_MODE 1025 }; 1026 1027 struct bpf_xdp_entity { 1028 struct bpf_prog *prog; 1029 struct bpf_xdp_link *link; 1030 }; 1031 1032 struct netdev_bpf { 1033 enum bpf_netdev_command command; 1034 union { 1035 /* XDP_SETUP_PROG */ 1036 struct { 1037 u32 flags; 1038 struct bpf_prog *prog; 1039 struct netlink_ext_ack *extack; 1040 }; 1041 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 1042 struct { 1043 struct bpf_offloaded_map *offmap; 1044 }; 1045 /* XDP_SETUP_XSK_POOL */ 1046 struct { 1047 struct xsk_buff_pool *pool; 1048 u16 queue_id; 1049 } xsk; 1050 }; 1051 }; 1052 1053 /* Flags for ndo_xsk_wakeup. */ 1054 #define XDP_WAKEUP_RX (1 << 0) 1055 #define XDP_WAKEUP_TX (1 << 1) 1056 1057 #ifdef CONFIG_XFRM_OFFLOAD 1058 struct xfrmdev_ops { 1059 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack); 1060 void (*xdo_dev_state_delete) (struct xfrm_state *x); 1061 void (*xdo_dev_state_free) (struct xfrm_state *x); 1062 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 1063 struct xfrm_state *x); 1064 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1065 void (*xdo_dev_state_update_curlft) (struct xfrm_state *x); 1066 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack); 1067 void (*xdo_dev_policy_delete) (struct xfrm_policy *x); 1068 void (*xdo_dev_policy_free) (struct xfrm_policy *x); 1069 }; 1070 #endif 1071 1072 struct dev_ifalias { 1073 struct rcu_head rcuhead; 1074 char ifalias[]; 1075 }; 1076 1077 struct devlink; 1078 struct tlsdev_ops; 1079 1080 struct netdev_net_notifier { 1081 struct list_head list; 1082 struct notifier_block *nb; 1083 }; 1084 1085 /* 1086 * This structure defines the management hooks for network devices. 1087 * The following hooks can be defined; unless noted otherwise, they are 1088 * optional and can be filled with a null pointer. 1089 * 1090 * int (*ndo_init)(struct net_device *dev); 1091 * This function is called once when a network device is registered. 1092 * The network device can use this for any late stage initialization 1093 * or semantic validation. It can fail with an error code which will 1094 * be propagated back to register_netdev. 1095 * 1096 * void (*ndo_uninit)(struct net_device *dev); 1097 * This function is called when device is unregistered or when registration 1098 * fails. It is not called if init fails. 1099 * 1100 * int (*ndo_open)(struct net_device *dev); 1101 * This function is called when a network device transitions to the up 1102 * state. 1103 * 1104 * int (*ndo_stop)(struct net_device *dev); 1105 * This function is called when a network device transitions to the down 1106 * state. 1107 * 1108 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1109 * struct net_device *dev); 1110 * Called when a packet needs to be transmitted. 1111 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1112 * the queue before that can happen; it's for obsolete devices and weird 1113 * corner cases, but the stack really does a non-trivial amount 1114 * of useless work if you return NETDEV_TX_BUSY. 1115 * Required; cannot be NULL. 1116 * 1117 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1118 * struct net_device *dev 1119 * netdev_features_t features); 1120 * Called by core transmit path to determine if device is capable of 1121 * performing offload operations on a given packet. This is to give 1122 * the device an opportunity to implement any restrictions that cannot 1123 * be otherwise expressed by feature flags. The check is called with 1124 * the set of features that the stack has calculated and it returns 1125 * those the driver believes to be appropriate. 1126 * 1127 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1128 * struct net_device *sb_dev); 1129 * Called to decide which queue to use when device supports multiple 1130 * transmit queues. 1131 * 1132 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1133 * This function is called to allow device receiver to make 1134 * changes to configuration when multicast or promiscuous is enabled. 1135 * 1136 * void (*ndo_set_rx_mode)(struct net_device *dev); 1137 * This function is called device changes address list filtering. 1138 * If driver handles unicast address filtering, it should set 1139 * IFF_UNICAST_FLT in its priv_flags. 1140 * 1141 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1142 * This function is called when the Media Access Control address 1143 * needs to be changed. If this interface is not defined, the 1144 * MAC address can not be changed. 1145 * 1146 * int (*ndo_validate_addr)(struct net_device *dev); 1147 * Test if Media Access Control address is valid for the device. 1148 * 1149 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1150 * Old-style ioctl entry point. This is used internally by the 1151 * appletalk and ieee802154 subsystems but is no longer called by 1152 * the device ioctl handler. 1153 * 1154 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1155 * Used by the bonding driver for its device specific ioctls: 1156 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1157 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1158 * 1159 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1160 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1161 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1162 * 1163 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1164 * Used to set network devices bus interface parameters. This interface 1165 * is retained for legacy reasons; new devices should use the bus 1166 * interface (PCI) for low level management. 1167 * 1168 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1169 * Called when a user wants to change the Maximum Transfer Unit 1170 * of a device. 1171 * 1172 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1173 * Callback used when the transmitter has not made any progress 1174 * for dev->watchdog ticks. 1175 * 1176 * void (*ndo_get_stats64)(struct net_device *dev, 1177 * struct rtnl_link_stats64 *storage); 1178 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1179 * Called when a user wants to get the network device usage 1180 * statistics. Drivers must do one of the following: 1181 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1182 * rtnl_link_stats64 structure passed by the caller. 1183 * 2. Define @ndo_get_stats to update a net_device_stats structure 1184 * (which should normally be dev->stats) and return a pointer to 1185 * it. The structure may be changed asynchronously only if each 1186 * field is written atomically. 1187 * 3. Update dev->stats asynchronously and atomically, and define 1188 * neither operation. 1189 * 1190 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1191 * Return true if this device supports offload stats of this attr_id. 1192 * 1193 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1194 * void *attr_data) 1195 * Get statistics for offload operations by attr_id. Write it into the 1196 * attr_data pointer. 1197 * 1198 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1199 * If device supports VLAN filtering this function is called when a 1200 * VLAN id is registered. 1201 * 1202 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1203 * If device supports VLAN filtering this function is called when a 1204 * VLAN id is unregistered. 1205 * 1206 * void (*ndo_poll_controller)(struct net_device *dev); 1207 * 1208 * SR-IOV management functions. 1209 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1210 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1211 * u8 qos, __be16 proto); 1212 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1213 * int max_tx_rate); 1214 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1215 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1216 * int (*ndo_get_vf_config)(struct net_device *dev, 1217 * int vf, struct ifla_vf_info *ivf); 1218 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1219 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1220 * struct nlattr *port[]); 1221 * 1222 * Enable or disable the VF ability to query its RSS Redirection Table and 1223 * Hash Key. This is needed since on some devices VF share this information 1224 * with PF and querying it may introduce a theoretical security risk. 1225 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1226 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1227 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1228 * void *type_data); 1229 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1230 * This is always called from the stack with the rtnl lock held and netif 1231 * tx queues stopped. This allows the netdevice to perform queue 1232 * management safely. 1233 * 1234 * Fiber Channel over Ethernet (FCoE) offload functions. 1235 * int (*ndo_fcoe_enable)(struct net_device *dev); 1236 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1237 * so the underlying device can perform whatever needed configuration or 1238 * initialization to support acceleration of FCoE traffic. 1239 * 1240 * int (*ndo_fcoe_disable)(struct net_device *dev); 1241 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1242 * so the underlying device can perform whatever needed clean-ups to 1243 * stop supporting acceleration of FCoE traffic. 1244 * 1245 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1246 * struct scatterlist *sgl, unsigned int sgc); 1247 * Called when the FCoE Initiator wants to initialize an I/O that 1248 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1249 * perform necessary setup and returns 1 to indicate the device is set up 1250 * successfully to perform DDP on this I/O, otherwise this returns 0. 1251 * 1252 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1253 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1254 * indicated by the FC exchange id 'xid', so the underlying device can 1255 * clean up and reuse resources for later DDP requests. 1256 * 1257 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1258 * struct scatterlist *sgl, unsigned int sgc); 1259 * Called when the FCoE Target wants to initialize an I/O that 1260 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1261 * perform necessary setup and returns 1 to indicate the device is set up 1262 * successfully to perform DDP on this I/O, otherwise this returns 0. 1263 * 1264 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1265 * struct netdev_fcoe_hbainfo *hbainfo); 1266 * Called when the FCoE Protocol stack wants information on the underlying 1267 * device. This information is utilized by the FCoE protocol stack to 1268 * register attributes with Fiber Channel management service as per the 1269 * FC-GS Fabric Device Management Information(FDMI) specification. 1270 * 1271 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1272 * Called when the underlying device wants to override default World Wide 1273 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1274 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1275 * protocol stack to use. 1276 * 1277 * RFS acceleration. 1278 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1279 * u16 rxq_index, u32 flow_id); 1280 * Set hardware filter for RFS. rxq_index is the target queue index; 1281 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1282 * Return the filter ID on success, or a negative error code. 1283 * 1284 * Slave management functions (for bridge, bonding, etc). 1285 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1286 * Called to make another netdev an underling. 1287 * 1288 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1289 * Called to release previously enslaved netdev. 1290 * 1291 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1292 * struct sk_buff *skb, 1293 * bool all_slaves); 1294 * Get the xmit slave of master device. If all_slaves is true, function 1295 * assume all the slaves can transmit. 1296 * 1297 * Feature/offload setting functions. 1298 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1299 * netdev_features_t features); 1300 * Adjusts the requested feature flags according to device-specific 1301 * constraints, and returns the resulting flags. Must not modify 1302 * the device state. 1303 * 1304 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1305 * Called to update device configuration to new features. Passed 1306 * feature set might be less than what was returned by ndo_fix_features()). 1307 * Must return >0 or -errno if it changed dev->features itself. 1308 * 1309 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1310 * struct net_device *dev, 1311 * const unsigned char *addr, u16 vid, u16 flags, 1312 * struct netlink_ext_ack *extack); 1313 * Adds an FDB entry to dev for addr. 1314 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1315 * struct net_device *dev, 1316 * const unsigned char *addr, u16 vid) 1317 * Deletes the FDB entry from dev coresponding to addr. 1318 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev, 1319 * struct netlink_ext_ack *extack); 1320 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1321 * struct net_device *dev, struct net_device *filter_dev, 1322 * int *idx) 1323 * Used to add FDB entries to dump requests. Implementers should add 1324 * entries to skb and update idx with the number of entries. 1325 * 1326 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[], 1327 * u16 nlmsg_flags, struct netlink_ext_ack *extack); 1328 * Adds an MDB entry to dev. 1329 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[], 1330 * struct netlink_ext_ack *extack); 1331 * Deletes the MDB entry from dev. 1332 * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[], 1333 * struct netlink_ext_ack *extack); 1334 * Bulk deletes MDB entries from dev. 1335 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb, 1336 * struct netlink_callback *cb); 1337 * Dumps MDB entries from dev. The first argument (marker) in the netlink 1338 * callback is used by core rtnetlink code. 1339 * 1340 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1341 * u16 flags, struct netlink_ext_ack *extack) 1342 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1343 * struct net_device *dev, u32 filter_mask, 1344 * int nlflags) 1345 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1346 * u16 flags); 1347 * 1348 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1349 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1350 * which do not represent real hardware may define this to allow their 1351 * userspace components to manage their virtual carrier state. Devices 1352 * that determine carrier state from physical hardware properties (eg 1353 * network cables) or protocol-dependent mechanisms (eg 1354 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1355 * 1356 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1357 * struct netdev_phys_item_id *ppid); 1358 * Called to get ID of physical port of this device. If driver does 1359 * not implement this, it is assumed that the hw is not able to have 1360 * multiple net devices on single physical port. 1361 * 1362 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1363 * struct netdev_phys_item_id *ppid) 1364 * Called to get the parent ID of the physical port of this device. 1365 * 1366 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1367 * struct net_device *dev) 1368 * Called by upper layer devices to accelerate switching or other 1369 * station functionality into hardware. 'pdev is the lowerdev 1370 * to use for the offload and 'dev' is the net device that will 1371 * back the offload. Returns a pointer to the private structure 1372 * the upper layer will maintain. 1373 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1374 * Called by upper layer device to delete the station created 1375 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1376 * the station and priv is the structure returned by the add 1377 * operation. 1378 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1379 * int queue_index, u32 maxrate); 1380 * Called when a user wants to set a max-rate limitation of specific 1381 * TX queue. 1382 * int (*ndo_get_iflink)(const struct net_device *dev); 1383 * Called to get the iflink value of this device. 1384 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1385 * This function is used to get egress tunnel information for given skb. 1386 * This is useful for retrieving outer tunnel header parameters while 1387 * sampling packet. 1388 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1389 * This function is used to specify the headroom that the skb must 1390 * consider when allocation skb during packet reception. Setting 1391 * appropriate rx headroom value allows avoiding skb head copy on 1392 * forward. Setting a negative value resets the rx headroom to the 1393 * default value. 1394 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1395 * This function is used to set or query state related to XDP on the 1396 * netdevice and manage BPF offload. See definition of 1397 * enum bpf_netdev_command for details. 1398 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1399 * u32 flags); 1400 * This function is used to submit @n XDP packets for transmit on a 1401 * netdevice. Returns number of frames successfully transmitted, frames 1402 * that got dropped are freed/returned via xdp_return_frame(). 1403 * Returns negative number, means general error invoking ndo, meaning 1404 * no frames were xmit'ed and core-caller will free all frames. 1405 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1406 * struct xdp_buff *xdp); 1407 * Get the xmit slave of master device based on the xdp_buff. 1408 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1409 * This function is used to wake up the softirq, ksoftirqd or kthread 1410 * responsible for sending and/or receiving packets on a specific 1411 * queue id bound to an AF_XDP socket. The flags field specifies if 1412 * only RX, only Tx, or both should be woken up using the flags 1413 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1414 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1415 * int cmd); 1416 * Add, change, delete or get information on an IPv4 tunnel. 1417 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1418 * If a device is paired with a peer device, return the peer instance. 1419 * The caller must be under RCU read context. 1420 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1421 * Get the forwarding path to reach the real device from the HW destination address 1422 * ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1423 * const struct skb_shared_hwtstamps *hwtstamps, 1424 * bool cycles); 1425 * Get hardware timestamp based on normal/adjustable time or free running 1426 * cycle counter. This function is required if physical clock supports a 1427 * free running cycle counter. 1428 * 1429 * int (*ndo_hwtstamp_get)(struct net_device *dev, 1430 * struct kernel_hwtstamp_config *kernel_config); 1431 * Get the currently configured hardware timestamping parameters for the 1432 * NIC device. 1433 * 1434 * int (*ndo_hwtstamp_set)(struct net_device *dev, 1435 * struct kernel_hwtstamp_config *kernel_config, 1436 * struct netlink_ext_ack *extack); 1437 * Change the hardware timestamping parameters for NIC device. 1438 */ 1439 struct net_device_ops { 1440 int (*ndo_init)(struct net_device *dev); 1441 void (*ndo_uninit)(struct net_device *dev); 1442 int (*ndo_open)(struct net_device *dev); 1443 int (*ndo_stop)(struct net_device *dev); 1444 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1445 struct net_device *dev); 1446 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1447 struct net_device *dev, 1448 netdev_features_t features); 1449 u16 (*ndo_select_queue)(struct net_device *dev, 1450 struct sk_buff *skb, 1451 struct net_device *sb_dev); 1452 void (*ndo_change_rx_flags)(struct net_device *dev, 1453 int flags); 1454 void (*ndo_set_rx_mode)(struct net_device *dev); 1455 int (*ndo_set_mac_address)(struct net_device *dev, 1456 void *addr); 1457 int (*ndo_validate_addr)(struct net_device *dev); 1458 int (*ndo_do_ioctl)(struct net_device *dev, 1459 struct ifreq *ifr, int cmd); 1460 int (*ndo_eth_ioctl)(struct net_device *dev, 1461 struct ifreq *ifr, int cmd); 1462 int (*ndo_siocbond)(struct net_device *dev, 1463 struct ifreq *ifr, int cmd); 1464 int (*ndo_siocwandev)(struct net_device *dev, 1465 struct if_settings *ifs); 1466 int (*ndo_siocdevprivate)(struct net_device *dev, 1467 struct ifreq *ifr, 1468 void __user *data, int cmd); 1469 int (*ndo_set_config)(struct net_device *dev, 1470 struct ifmap *map); 1471 int (*ndo_change_mtu)(struct net_device *dev, 1472 int new_mtu); 1473 int (*ndo_neigh_setup)(struct net_device *dev, 1474 struct neigh_parms *); 1475 void (*ndo_tx_timeout) (struct net_device *dev, 1476 unsigned int txqueue); 1477 1478 void (*ndo_get_stats64)(struct net_device *dev, 1479 struct rtnl_link_stats64 *storage); 1480 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1481 int (*ndo_get_offload_stats)(int attr_id, 1482 const struct net_device *dev, 1483 void *attr_data); 1484 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1485 1486 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1487 __be16 proto, u16 vid); 1488 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1489 __be16 proto, u16 vid); 1490 #ifdef CONFIG_NET_POLL_CONTROLLER 1491 void (*ndo_poll_controller)(struct net_device *dev); 1492 int (*ndo_netpoll_setup)(struct net_device *dev, 1493 struct netpoll_info *info); 1494 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1495 #endif 1496 int (*ndo_set_vf_mac)(struct net_device *dev, 1497 int queue, u8 *mac); 1498 int (*ndo_set_vf_vlan)(struct net_device *dev, 1499 int queue, u16 vlan, 1500 u8 qos, __be16 proto); 1501 int (*ndo_set_vf_rate)(struct net_device *dev, 1502 int vf, int min_tx_rate, 1503 int max_tx_rate); 1504 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1505 int vf, bool setting); 1506 int (*ndo_set_vf_trust)(struct net_device *dev, 1507 int vf, bool setting); 1508 int (*ndo_get_vf_config)(struct net_device *dev, 1509 int vf, 1510 struct ifla_vf_info *ivf); 1511 int (*ndo_set_vf_link_state)(struct net_device *dev, 1512 int vf, int link_state); 1513 int (*ndo_get_vf_stats)(struct net_device *dev, 1514 int vf, 1515 struct ifla_vf_stats 1516 *vf_stats); 1517 int (*ndo_set_vf_port)(struct net_device *dev, 1518 int vf, 1519 struct nlattr *port[]); 1520 int (*ndo_get_vf_port)(struct net_device *dev, 1521 int vf, struct sk_buff *skb); 1522 int (*ndo_get_vf_guid)(struct net_device *dev, 1523 int vf, 1524 struct ifla_vf_guid *node_guid, 1525 struct ifla_vf_guid *port_guid); 1526 int (*ndo_set_vf_guid)(struct net_device *dev, 1527 int vf, u64 guid, 1528 int guid_type); 1529 int (*ndo_set_vf_rss_query_en)( 1530 struct net_device *dev, 1531 int vf, bool setting); 1532 int (*ndo_setup_tc)(struct net_device *dev, 1533 enum tc_setup_type type, 1534 void *type_data); 1535 #if IS_ENABLED(CONFIG_FCOE) 1536 int (*ndo_fcoe_enable)(struct net_device *dev); 1537 int (*ndo_fcoe_disable)(struct net_device *dev); 1538 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1539 u16 xid, 1540 struct scatterlist *sgl, 1541 unsigned int sgc); 1542 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1543 u16 xid); 1544 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1545 u16 xid, 1546 struct scatterlist *sgl, 1547 unsigned int sgc); 1548 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1549 struct netdev_fcoe_hbainfo *hbainfo); 1550 #endif 1551 1552 #if IS_ENABLED(CONFIG_LIBFCOE) 1553 #define NETDEV_FCOE_WWNN 0 1554 #define NETDEV_FCOE_WWPN 1 1555 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1556 u64 *wwn, int type); 1557 #endif 1558 1559 #ifdef CONFIG_RFS_ACCEL 1560 int (*ndo_rx_flow_steer)(struct net_device *dev, 1561 const struct sk_buff *skb, 1562 u16 rxq_index, 1563 u32 flow_id); 1564 #endif 1565 int (*ndo_add_slave)(struct net_device *dev, 1566 struct net_device *slave_dev, 1567 struct netlink_ext_ack *extack); 1568 int (*ndo_del_slave)(struct net_device *dev, 1569 struct net_device *slave_dev); 1570 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1571 struct sk_buff *skb, 1572 bool all_slaves); 1573 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1574 struct sock *sk); 1575 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1576 netdev_features_t features); 1577 int (*ndo_set_features)(struct net_device *dev, 1578 netdev_features_t features); 1579 int (*ndo_neigh_construct)(struct net_device *dev, 1580 struct neighbour *n); 1581 void (*ndo_neigh_destroy)(struct net_device *dev, 1582 struct neighbour *n); 1583 1584 int (*ndo_fdb_add)(struct ndmsg *ndm, 1585 struct nlattr *tb[], 1586 struct net_device *dev, 1587 const unsigned char *addr, 1588 u16 vid, 1589 u16 flags, 1590 struct netlink_ext_ack *extack); 1591 int (*ndo_fdb_del)(struct ndmsg *ndm, 1592 struct nlattr *tb[], 1593 struct net_device *dev, 1594 const unsigned char *addr, 1595 u16 vid, struct netlink_ext_ack *extack); 1596 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, 1597 struct net_device *dev, 1598 struct netlink_ext_ack *extack); 1599 int (*ndo_fdb_dump)(struct sk_buff *skb, 1600 struct netlink_callback *cb, 1601 struct net_device *dev, 1602 struct net_device *filter_dev, 1603 int *idx); 1604 int (*ndo_fdb_get)(struct sk_buff *skb, 1605 struct nlattr *tb[], 1606 struct net_device *dev, 1607 const unsigned char *addr, 1608 u16 vid, u32 portid, u32 seq, 1609 struct netlink_ext_ack *extack); 1610 int (*ndo_mdb_add)(struct net_device *dev, 1611 struct nlattr *tb[], 1612 u16 nlmsg_flags, 1613 struct netlink_ext_ack *extack); 1614 int (*ndo_mdb_del)(struct net_device *dev, 1615 struct nlattr *tb[], 1616 struct netlink_ext_ack *extack); 1617 int (*ndo_mdb_del_bulk)(struct net_device *dev, 1618 struct nlattr *tb[], 1619 struct netlink_ext_ack *extack); 1620 int (*ndo_mdb_dump)(struct net_device *dev, 1621 struct sk_buff *skb, 1622 struct netlink_callback *cb); 1623 int (*ndo_mdb_get)(struct net_device *dev, 1624 struct nlattr *tb[], u32 portid, 1625 u32 seq, 1626 struct netlink_ext_ack *extack); 1627 int (*ndo_bridge_setlink)(struct net_device *dev, 1628 struct nlmsghdr *nlh, 1629 u16 flags, 1630 struct netlink_ext_ack *extack); 1631 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1632 u32 pid, u32 seq, 1633 struct net_device *dev, 1634 u32 filter_mask, 1635 int nlflags); 1636 int (*ndo_bridge_dellink)(struct net_device *dev, 1637 struct nlmsghdr *nlh, 1638 u16 flags); 1639 int (*ndo_change_carrier)(struct net_device *dev, 1640 bool new_carrier); 1641 int (*ndo_get_phys_port_id)(struct net_device *dev, 1642 struct netdev_phys_item_id *ppid); 1643 int (*ndo_get_port_parent_id)(struct net_device *dev, 1644 struct netdev_phys_item_id *ppid); 1645 int (*ndo_get_phys_port_name)(struct net_device *dev, 1646 char *name, size_t len); 1647 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1648 struct net_device *dev); 1649 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1650 void *priv); 1651 1652 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1653 int queue_index, 1654 u32 maxrate); 1655 int (*ndo_get_iflink)(const struct net_device *dev); 1656 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1657 struct sk_buff *skb); 1658 void (*ndo_set_rx_headroom)(struct net_device *dev, 1659 int needed_headroom); 1660 int (*ndo_bpf)(struct net_device *dev, 1661 struct netdev_bpf *bpf); 1662 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1663 struct xdp_frame **xdp, 1664 u32 flags); 1665 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1666 struct xdp_buff *xdp); 1667 int (*ndo_xsk_wakeup)(struct net_device *dev, 1668 u32 queue_id, u32 flags); 1669 int (*ndo_tunnel_ctl)(struct net_device *dev, 1670 struct ip_tunnel_parm *p, int cmd); 1671 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1672 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1673 struct net_device_path *path); 1674 ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1675 const struct skb_shared_hwtstamps *hwtstamps, 1676 bool cycles); 1677 int (*ndo_hwtstamp_get)(struct net_device *dev, 1678 struct kernel_hwtstamp_config *kernel_config); 1679 int (*ndo_hwtstamp_set)(struct net_device *dev, 1680 struct kernel_hwtstamp_config *kernel_config, 1681 struct netlink_ext_ack *extack); 1682 }; 1683 1684 /** 1685 * enum netdev_priv_flags - &struct net_device priv_flags 1686 * 1687 * These are the &struct net_device, they are only set internally 1688 * by drivers and used in the kernel. These flags are invisible to 1689 * userspace; this means that the order of these flags can change 1690 * during any kernel release. 1691 * 1692 * You should have a pretty good reason to be extending these flags. 1693 * 1694 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1695 * @IFF_EBRIDGE: Ethernet bridging device 1696 * @IFF_BONDING: bonding master or slave 1697 * @IFF_ISATAP: ISATAP interface (RFC4214) 1698 * @IFF_WAN_HDLC: WAN HDLC device 1699 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1700 * release skb->dst 1701 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1702 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1703 * @IFF_MACVLAN_PORT: device used as macvlan port 1704 * @IFF_BRIDGE_PORT: device used as bridge port 1705 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1706 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1707 * @IFF_UNICAST_FLT: Supports unicast filtering 1708 * @IFF_TEAM_PORT: device used as team port 1709 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1710 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1711 * change when it's running 1712 * @IFF_MACVLAN: Macvlan device 1713 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1714 * underlying stacked devices 1715 * @IFF_L3MDEV_MASTER: device is an L3 master device 1716 * @IFF_NO_QUEUE: device can run without qdisc attached 1717 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1718 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1719 * @IFF_TEAM: device is a team device 1720 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1721 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1722 * entity (i.e. the master device for bridged veth) 1723 * @IFF_MACSEC: device is a MACsec device 1724 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1725 * @IFF_FAILOVER: device is a failover master device 1726 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1727 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1728 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf 1729 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1730 * skb_headlen(skb) == 0 (data starts from frag0) 1731 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN 1732 * @IFF_SEE_ALL_HWTSTAMP_REQUESTS: device wants to see calls to 1733 * ndo_hwtstamp_set() for all timestamp requests regardless of source, 1734 * even if those aren't HWTSTAMP_SOURCE_NETDEV. 1735 */ 1736 enum netdev_priv_flags { 1737 IFF_802_1Q_VLAN = 1<<0, 1738 IFF_EBRIDGE = 1<<1, 1739 IFF_BONDING = 1<<2, 1740 IFF_ISATAP = 1<<3, 1741 IFF_WAN_HDLC = 1<<4, 1742 IFF_XMIT_DST_RELEASE = 1<<5, 1743 IFF_DONT_BRIDGE = 1<<6, 1744 IFF_DISABLE_NETPOLL = 1<<7, 1745 IFF_MACVLAN_PORT = 1<<8, 1746 IFF_BRIDGE_PORT = 1<<9, 1747 IFF_OVS_DATAPATH = 1<<10, 1748 IFF_TX_SKB_SHARING = 1<<11, 1749 IFF_UNICAST_FLT = 1<<12, 1750 IFF_TEAM_PORT = 1<<13, 1751 IFF_SUPP_NOFCS = 1<<14, 1752 IFF_LIVE_ADDR_CHANGE = 1<<15, 1753 IFF_MACVLAN = 1<<16, 1754 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1755 IFF_L3MDEV_MASTER = 1<<18, 1756 IFF_NO_QUEUE = 1<<19, 1757 IFF_OPENVSWITCH = 1<<20, 1758 IFF_L3MDEV_SLAVE = 1<<21, 1759 IFF_TEAM = 1<<22, 1760 IFF_RXFH_CONFIGURED = 1<<23, 1761 IFF_PHONY_HEADROOM = 1<<24, 1762 IFF_MACSEC = 1<<25, 1763 IFF_NO_RX_HANDLER = 1<<26, 1764 IFF_FAILOVER = 1<<27, 1765 IFF_FAILOVER_SLAVE = 1<<28, 1766 IFF_L3MDEV_RX_HANDLER = 1<<29, 1767 IFF_NO_ADDRCONF = BIT_ULL(30), 1768 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31), 1769 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32), 1770 IFF_SEE_ALL_HWTSTAMP_REQUESTS = BIT_ULL(33), 1771 }; 1772 1773 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1774 #define IFF_EBRIDGE IFF_EBRIDGE 1775 #define IFF_BONDING IFF_BONDING 1776 #define IFF_ISATAP IFF_ISATAP 1777 #define IFF_WAN_HDLC IFF_WAN_HDLC 1778 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1779 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1780 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1781 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1782 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1783 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1784 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1785 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1786 #define IFF_TEAM_PORT IFF_TEAM_PORT 1787 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1788 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1789 #define IFF_MACVLAN IFF_MACVLAN 1790 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1791 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1792 #define IFF_NO_QUEUE IFF_NO_QUEUE 1793 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1794 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1795 #define IFF_TEAM IFF_TEAM 1796 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1797 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM 1798 #define IFF_MACSEC IFF_MACSEC 1799 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1800 #define IFF_FAILOVER IFF_FAILOVER 1801 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1802 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1803 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR 1804 1805 /* Specifies the type of the struct net_device::ml_priv pointer */ 1806 enum netdev_ml_priv_type { 1807 ML_PRIV_NONE, 1808 ML_PRIV_CAN, 1809 }; 1810 1811 enum netdev_stat_type { 1812 NETDEV_PCPU_STAT_NONE, 1813 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */ 1814 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */ 1815 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */ 1816 }; 1817 1818 /** 1819 * struct net_device - The DEVICE structure. 1820 * 1821 * Actually, this whole structure is a big mistake. It mixes I/O 1822 * data with strictly "high-level" data, and it has to know about 1823 * almost every data structure used in the INET module. 1824 * 1825 * @name: This is the first field of the "visible" part of this structure 1826 * (i.e. as seen by users in the "Space.c" file). It is the name 1827 * of the interface. 1828 * 1829 * @name_node: Name hashlist node 1830 * @ifalias: SNMP alias 1831 * @mem_end: Shared memory end 1832 * @mem_start: Shared memory start 1833 * @base_addr: Device I/O address 1834 * @irq: Device IRQ number 1835 * 1836 * @state: Generic network queuing layer state, see netdev_state_t 1837 * @dev_list: The global list of network devices 1838 * @napi_list: List entry used for polling NAPI devices 1839 * @unreg_list: List entry when we are unregistering the 1840 * device; see the function unregister_netdev 1841 * @close_list: List entry used when we are closing the device 1842 * @ptype_all: Device-specific packet handlers for all protocols 1843 * @ptype_specific: Device-specific, protocol-specific packet handlers 1844 * 1845 * @adj_list: Directly linked devices, like slaves for bonding 1846 * @features: Currently active device features 1847 * @hw_features: User-changeable features 1848 * 1849 * @wanted_features: User-requested features 1850 * @vlan_features: Mask of features inheritable by VLAN devices 1851 * 1852 * @hw_enc_features: Mask of features inherited by encapsulating devices 1853 * This field indicates what encapsulation 1854 * offloads the hardware is capable of doing, 1855 * and drivers will need to set them appropriately. 1856 * 1857 * @mpls_features: Mask of features inheritable by MPLS 1858 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1859 * 1860 * @ifindex: interface index 1861 * @group: The group the device belongs to 1862 * 1863 * @stats: Statistics struct, which was left as a legacy, use 1864 * rtnl_link_stats64 instead 1865 * 1866 * @core_stats: core networking counters, 1867 * do not use this in drivers 1868 * @carrier_up_count: Number of times the carrier has been up 1869 * @carrier_down_count: Number of times the carrier has been down 1870 * 1871 * @wireless_handlers: List of functions to handle Wireless Extensions, 1872 * instead of ioctl, 1873 * see <net/iw_handler.h> for details. 1874 * @wireless_data: Instance data managed by the core of wireless extensions 1875 * 1876 * @netdev_ops: Includes several pointers to callbacks, 1877 * if one wants to override the ndo_*() functions 1878 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks. 1879 * @xsk_tx_metadata_ops: Includes pointers to AF_XDP TX metadata callbacks. 1880 * @ethtool_ops: Management operations 1881 * @l3mdev_ops: Layer 3 master device operations 1882 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1883 * discovery handling. Necessary for e.g. 6LoWPAN. 1884 * @xfrmdev_ops: Transformation offload operations 1885 * @tlsdev_ops: Transport Layer Security offload operations 1886 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1887 * of Layer 2 headers. 1888 * 1889 * @flags: Interface flags (a la BSD) 1890 * @xdp_features: XDP capability supported by the device 1891 * @priv_flags: Like 'flags' but invisible to userspace, 1892 * see if.h for the definitions 1893 * @gflags: Global flags ( kept as legacy ) 1894 * @padded: How much padding added by alloc_netdev() 1895 * @operstate: RFC2863 operstate 1896 * @link_mode: Mapping policy to operstate 1897 * @if_port: Selectable AUI, TP, ... 1898 * @dma: DMA channel 1899 * @mtu: Interface MTU value 1900 * @min_mtu: Interface Minimum MTU value 1901 * @max_mtu: Interface Maximum MTU value 1902 * @type: Interface hardware type 1903 * @hard_header_len: Maximum hardware header length. 1904 * @min_header_len: Minimum hardware header length 1905 * 1906 * @needed_headroom: Extra headroom the hardware may need, but not in all 1907 * cases can this be guaranteed 1908 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1909 * cases can this be guaranteed. Some cases also use 1910 * LL_MAX_HEADER instead to allocate the skb 1911 * 1912 * interface address info: 1913 * 1914 * @perm_addr: Permanent hw address 1915 * @addr_assign_type: Hw address assignment type 1916 * @addr_len: Hardware address length 1917 * @upper_level: Maximum depth level of upper devices. 1918 * @lower_level: Maximum depth level of lower devices. 1919 * @neigh_priv_len: Used in neigh_alloc() 1920 * @dev_id: Used to differentiate devices that share 1921 * the same link layer address 1922 * @dev_port: Used to differentiate devices that share 1923 * the same function 1924 * @addr_list_lock: XXX: need comments on this one 1925 * @name_assign_type: network interface name assignment type 1926 * @uc_promisc: Counter that indicates promiscuous mode 1927 * has been enabled due to the need to listen to 1928 * additional unicast addresses in a device that 1929 * does not implement ndo_set_rx_mode() 1930 * @uc: unicast mac addresses 1931 * @mc: multicast mac addresses 1932 * @dev_addrs: list of device hw addresses 1933 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1934 * @promiscuity: Number of times the NIC is told to work in 1935 * promiscuous mode; if it becomes 0 the NIC will 1936 * exit promiscuous mode 1937 * @allmulti: Counter, enables or disables allmulticast mode 1938 * 1939 * @vlan_info: VLAN info 1940 * @dsa_ptr: dsa specific data 1941 * @tipc_ptr: TIPC specific data 1942 * @atalk_ptr: AppleTalk link 1943 * @ip_ptr: IPv4 specific data 1944 * @ip6_ptr: IPv6 specific data 1945 * @ax25_ptr: AX.25 specific data 1946 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1947 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1948 * device struct 1949 * @mpls_ptr: mpls_dev struct pointer 1950 * @mctp_ptr: MCTP specific data 1951 * 1952 * @dev_addr: Hw address (before bcast, 1953 * because most packets are unicast) 1954 * 1955 * @_rx: Array of RX queues 1956 * @num_rx_queues: Number of RX queues 1957 * allocated at register_netdev() time 1958 * @real_num_rx_queues: Number of RX queues currently active in device 1959 * @xdp_prog: XDP sockets filter program pointer 1960 * @gro_flush_timeout: timeout for GRO layer in NAPI 1961 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1962 * allow to avoid NIC hard IRQ, on busy queues. 1963 * 1964 * @rx_handler: handler for received packets 1965 * @rx_handler_data: XXX: need comments on this one 1966 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing 1967 * @ingress_queue: XXX: need comments on this one 1968 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1969 * @broadcast: hw bcast address 1970 * 1971 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1972 * indexed by RX queue number. Assigned by driver. 1973 * This must only be set if the ndo_rx_flow_steer 1974 * operation is defined 1975 * @index_hlist: Device index hash chain 1976 * 1977 * @_tx: Array of TX queues 1978 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1979 * @real_num_tx_queues: Number of TX queues currently active in device 1980 * @qdisc: Root qdisc from userspace point of view 1981 * @tx_queue_len: Max frames per queue allowed 1982 * @tx_global_lock: XXX: need comments on this one 1983 * @xdp_bulkq: XDP device bulk queue 1984 * @xps_maps: all CPUs/RXQs maps for XPS device 1985 * 1986 * @xps_maps: XXX: need comments on this one 1987 * @tcx_egress: BPF & clsact qdisc specific data for egress processing 1988 * @nf_hooks_egress: netfilter hooks executed for egress packets 1989 * @qdisc_hash: qdisc hash table 1990 * @watchdog_timeo: Represents the timeout that is used by 1991 * the watchdog (see dev_watchdog()) 1992 * @watchdog_timer: List of timers 1993 * 1994 * @proto_down_reason: reason a netdev interface is held down 1995 * @pcpu_refcnt: Number of references to this device 1996 * @dev_refcnt: Number of references to this device 1997 * @refcnt_tracker: Tracker directory for tracked references to this device 1998 * @todo_list: Delayed register/unregister 1999 * @link_watch_list: XXX: need comments on this one 2000 * 2001 * @reg_state: Register/unregister state machine 2002 * @dismantle: Device is going to be freed 2003 * @rtnl_link_state: This enum represents the phases of creating 2004 * a new link 2005 * 2006 * @needs_free_netdev: Should unregister perform free_netdev? 2007 * @priv_destructor: Called from unregister 2008 * @npinfo: XXX: need comments on this one 2009 * @nd_net: Network namespace this network device is inside 2010 * 2011 * @ml_priv: Mid-layer private 2012 * @ml_priv_type: Mid-layer private type 2013 * 2014 * @pcpu_stat_type: Type of device statistics which the core should 2015 * allocate/free: none, lstats, tstats, dstats. none 2016 * means the driver is handling statistics allocation/ 2017 * freeing internally. 2018 * @lstats: Loopback statistics: packets, bytes 2019 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes 2020 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes 2021 * 2022 * @garp_port: GARP 2023 * @mrp_port: MRP 2024 * 2025 * @dm_private: Drop monitor private 2026 * 2027 * @dev: Class/net/name entry 2028 * @sysfs_groups: Space for optional device, statistics and wireless 2029 * sysfs groups 2030 * 2031 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 2032 * @rtnl_link_ops: Rtnl_link_ops 2033 * 2034 * @gso_max_size: Maximum size of generic segmentation offload 2035 * @tso_max_size: Device (as in HW) limit on the max TSO request size 2036 * @gso_max_segs: Maximum number of segments that can be passed to the 2037 * NIC for GSO 2038 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count 2039 * @gso_ipv4_max_size: Maximum size of generic segmentation offload, 2040 * for IPv4. 2041 * 2042 * @dcbnl_ops: Data Center Bridging netlink ops 2043 * @num_tc: Number of traffic classes in the net device 2044 * @tc_to_txq: XXX: need comments on this one 2045 * @prio_tc_map: XXX: need comments on this one 2046 * 2047 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 2048 * 2049 * @priomap: XXX: need comments on this one 2050 * @link_topo: Physical link topology tracking attached PHYs 2051 * @phydev: Physical device may attach itself 2052 * for hardware timestamping 2053 * @sfp_bus: attached &struct sfp_bus structure. 2054 * 2055 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 2056 * 2057 * @proto_down: protocol port state information can be sent to the 2058 * switch driver and used to set the phys state of the 2059 * switch port. 2060 * 2061 * @wol_enabled: Wake-on-LAN is enabled 2062 * 2063 * @threaded: napi threaded mode is enabled 2064 * 2065 * @net_notifier_list: List of per-net netdev notifier block 2066 * that follow this device when it is moved 2067 * to another network namespace. 2068 * 2069 * @macsec_ops: MACsec offloading ops 2070 * 2071 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 2072 * offload capabilities of the device 2073 * @udp_tunnel_nic: UDP tunnel offload state 2074 * @xdp_state: stores info on attached XDP BPF programs 2075 * 2076 * @nested_level: Used as a parameter of spin_lock_nested() of 2077 * dev->addr_list_lock. 2078 * @unlink_list: As netif_addr_lock() can be called recursively, 2079 * keep a list of interfaces to be deleted. 2080 * @gro_max_size: Maximum size of aggregated packet in generic 2081 * receive offload (GRO) 2082 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic 2083 * receive offload (GRO), for IPv4. 2084 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP 2085 * zero copy driver 2086 * 2087 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 2088 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 2089 * @watchdog_dev_tracker: refcount tracker used by watchdog. 2090 * @dev_registered_tracker: tracker for reference held while 2091 * registered 2092 * @offload_xstats_l3: L3 HW stats for this netdevice. 2093 * 2094 * @devlink_port: Pointer to related devlink port structure. 2095 * Assigned by a driver before netdev registration using 2096 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static 2097 * during the time netdevice is registered. 2098 * 2099 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem, 2100 * where the clock is recovered. 2101 * 2102 * FIXME: cleanup struct net_device such that network protocol info 2103 * moves out. 2104 */ 2105 2106 struct net_device { 2107 /* Cacheline organization can be found documented in 2108 * Documentation/networking/net_cachelines/net_device.rst. 2109 * Please update the document when adding new fields. 2110 */ 2111 2112 /* TX read-mostly hotpath */ 2113 __cacheline_group_begin(net_device_read_tx); 2114 unsigned long long priv_flags; 2115 const struct net_device_ops *netdev_ops; 2116 const struct header_ops *header_ops; 2117 struct netdev_queue *_tx; 2118 netdev_features_t gso_partial_features; 2119 unsigned int real_num_tx_queues; 2120 unsigned int gso_max_size; 2121 unsigned int gso_ipv4_max_size; 2122 u16 gso_max_segs; 2123 s16 num_tc; 2124 /* Note : dev->mtu is often read without holding a lock. 2125 * Writers usually hold RTNL. 2126 * It is recommended to use READ_ONCE() to annotate the reads, 2127 * and to use WRITE_ONCE() to annotate the writes. 2128 */ 2129 unsigned int mtu; 2130 unsigned short needed_headroom; 2131 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2132 #ifdef CONFIG_XPS 2133 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2134 #endif 2135 #ifdef CONFIG_NETFILTER_EGRESS 2136 struct nf_hook_entries __rcu *nf_hooks_egress; 2137 #endif 2138 #ifdef CONFIG_NET_XGRESS 2139 struct bpf_mprog_entry __rcu *tcx_egress; 2140 #endif 2141 __cacheline_group_end(net_device_read_tx); 2142 2143 /* TXRX read-mostly hotpath */ 2144 __cacheline_group_begin(net_device_read_txrx); 2145 unsigned int flags; 2146 unsigned short hard_header_len; 2147 netdev_features_t features; 2148 struct inet6_dev __rcu *ip6_ptr; 2149 __cacheline_group_end(net_device_read_txrx); 2150 2151 /* RX read-mostly hotpath */ 2152 __cacheline_group_begin(net_device_read_rx); 2153 struct list_head ptype_specific; 2154 int ifindex; 2155 unsigned int real_num_rx_queues; 2156 struct netdev_rx_queue *_rx; 2157 unsigned long gro_flush_timeout; 2158 int napi_defer_hard_irqs; 2159 unsigned int gro_max_size; 2160 unsigned int gro_ipv4_max_size; 2161 rx_handler_func_t __rcu *rx_handler; 2162 void __rcu *rx_handler_data; 2163 possible_net_t nd_net; 2164 #ifdef CONFIG_NETPOLL 2165 struct netpoll_info __rcu *npinfo; 2166 #endif 2167 #ifdef CONFIG_NET_XGRESS 2168 struct bpf_mprog_entry __rcu *tcx_ingress; 2169 #endif 2170 __cacheline_group_end(net_device_read_rx); 2171 2172 char name[IFNAMSIZ]; 2173 struct netdev_name_node *name_node; 2174 struct dev_ifalias __rcu *ifalias; 2175 /* 2176 * I/O specific fields 2177 * FIXME: Merge these and struct ifmap into one 2178 */ 2179 unsigned long mem_end; 2180 unsigned long mem_start; 2181 unsigned long base_addr; 2182 2183 /* 2184 * Some hardware also needs these fields (state,dev_list, 2185 * napi_list,unreg_list,close_list) but they are not 2186 * part of the usual set specified in Space.c. 2187 */ 2188 2189 unsigned long state; 2190 2191 struct list_head dev_list; 2192 struct list_head napi_list; 2193 struct list_head unreg_list; 2194 struct list_head close_list; 2195 struct list_head ptype_all; 2196 2197 struct { 2198 struct list_head upper; 2199 struct list_head lower; 2200 } adj_list; 2201 2202 /* Read-mostly cache-line for fast-path access */ 2203 xdp_features_t xdp_features; 2204 const struct xdp_metadata_ops *xdp_metadata_ops; 2205 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops; 2206 unsigned short gflags; 2207 2208 unsigned short needed_tailroom; 2209 2210 netdev_features_t hw_features; 2211 netdev_features_t wanted_features; 2212 netdev_features_t vlan_features; 2213 netdev_features_t hw_enc_features; 2214 netdev_features_t mpls_features; 2215 2216 unsigned int min_mtu; 2217 unsigned int max_mtu; 2218 unsigned short type; 2219 unsigned char min_header_len; 2220 unsigned char name_assign_type; 2221 2222 int group; 2223 2224 struct net_device_stats stats; /* not used by modern drivers */ 2225 2226 struct net_device_core_stats __percpu *core_stats; 2227 2228 /* Stats to monitor link on/off, flapping */ 2229 atomic_t carrier_up_count; 2230 atomic_t carrier_down_count; 2231 2232 #ifdef CONFIG_WIRELESS_EXT 2233 const struct iw_handler_def *wireless_handlers; 2234 struct iw_public_data *wireless_data; 2235 #endif 2236 const struct ethtool_ops *ethtool_ops; 2237 #ifdef CONFIG_NET_L3_MASTER_DEV 2238 const struct l3mdev_ops *l3mdev_ops; 2239 #endif 2240 #if IS_ENABLED(CONFIG_IPV6) 2241 const struct ndisc_ops *ndisc_ops; 2242 #endif 2243 2244 #ifdef CONFIG_XFRM_OFFLOAD 2245 const struct xfrmdev_ops *xfrmdev_ops; 2246 #endif 2247 2248 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2249 const struct tlsdev_ops *tlsdev_ops; 2250 #endif 2251 2252 unsigned char operstate; 2253 unsigned char link_mode; 2254 2255 unsigned char if_port; 2256 unsigned char dma; 2257 2258 /* Interface address info. */ 2259 unsigned char perm_addr[MAX_ADDR_LEN]; 2260 unsigned char addr_assign_type; 2261 unsigned char addr_len; 2262 unsigned char upper_level; 2263 unsigned char lower_level; 2264 2265 unsigned short neigh_priv_len; 2266 unsigned short dev_id; 2267 unsigned short dev_port; 2268 unsigned short padded; 2269 2270 spinlock_t addr_list_lock; 2271 int irq; 2272 2273 struct netdev_hw_addr_list uc; 2274 struct netdev_hw_addr_list mc; 2275 struct netdev_hw_addr_list dev_addrs; 2276 2277 #ifdef CONFIG_SYSFS 2278 struct kset *queues_kset; 2279 #endif 2280 #ifdef CONFIG_LOCKDEP 2281 struct list_head unlink_list; 2282 #endif 2283 unsigned int promiscuity; 2284 unsigned int allmulti; 2285 bool uc_promisc; 2286 #ifdef CONFIG_LOCKDEP 2287 unsigned char nested_level; 2288 #endif 2289 2290 2291 /* Protocol-specific pointers */ 2292 struct in_device __rcu *ip_ptr; 2293 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2294 struct vlan_info __rcu *vlan_info; 2295 #endif 2296 #if IS_ENABLED(CONFIG_NET_DSA) 2297 struct dsa_port *dsa_ptr; 2298 #endif 2299 #if IS_ENABLED(CONFIG_TIPC) 2300 struct tipc_bearer __rcu *tipc_ptr; 2301 #endif 2302 #if IS_ENABLED(CONFIG_ATALK) 2303 void *atalk_ptr; 2304 #endif 2305 #if IS_ENABLED(CONFIG_AX25) 2306 void *ax25_ptr; 2307 #endif 2308 #if IS_ENABLED(CONFIG_CFG80211) 2309 struct wireless_dev *ieee80211_ptr; 2310 #endif 2311 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2312 struct wpan_dev *ieee802154_ptr; 2313 #endif 2314 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2315 struct mpls_dev __rcu *mpls_ptr; 2316 #endif 2317 #if IS_ENABLED(CONFIG_MCTP) 2318 struct mctp_dev __rcu *mctp_ptr; 2319 #endif 2320 2321 /* 2322 * Cache lines mostly used on receive path (including eth_type_trans()) 2323 */ 2324 /* Interface address info used in eth_type_trans() */ 2325 const unsigned char *dev_addr; 2326 2327 unsigned int num_rx_queues; 2328 struct bpf_prog __rcu *xdp_prog; 2329 #define GRO_LEGACY_MAX_SIZE 65536u 2330 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2331 * and shinfo->gso_segs is a 16bit field. 2332 */ 2333 #define GRO_MAX_SIZE (8 * 65535u) 2334 unsigned int xdp_zc_max_segs; 2335 struct netdev_queue __rcu *ingress_queue; 2336 #ifdef CONFIG_NETFILTER_INGRESS 2337 struct nf_hook_entries __rcu *nf_hooks_ingress; 2338 #endif 2339 2340 unsigned char broadcast[MAX_ADDR_LEN]; 2341 #ifdef CONFIG_RFS_ACCEL 2342 struct cpu_rmap *rx_cpu_rmap; 2343 #endif 2344 struct hlist_node index_hlist; 2345 2346 /* 2347 * Cache lines mostly used on transmit path 2348 */ 2349 unsigned int num_tx_queues; 2350 struct Qdisc __rcu *qdisc; 2351 unsigned int tx_queue_len; 2352 spinlock_t tx_global_lock; 2353 2354 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2355 2356 #ifdef CONFIG_NET_SCHED 2357 DECLARE_HASHTABLE (qdisc_hash, 4); 2358 #endif 2359 /* These may be needed for future network-power-down code. */ 2360 struct timer_list watchdog_timer; 2361 int watchdog_timeo; 2362 2363 u32 proto_down_reason; 2364 2365 struct list_head todo_list; 2366 2367 #ifdef CONFIG_PCPU_DEV_REFCNT 2368 int __percpu *pcpu_refcnt; 2369 #else 2370 refcount_t dev_refcnt; 2371 #endif 2372 struct ref_tracker_dir refcnt_tracker; 2373 2374 struct list_head link_watch_list; 2375 2376 enum { NETREG_UNINITIALIZED=0, 2377 NETREG_REGISTERED, /* completed register_netdevice */ 2378 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2379 NETREG_UNREGISTERED, /* completed unregister todo */ 2380 NETREG_RELEASED, /* called free_netdev */ 2381 NETREG_DUMMY, /* dummy device for NAPI poll */ 2382 } reg_state:8; 2383 2384 bool dismantle; 2385 2386 enum { 2387 RTNL_LINK_INITIALIZED, 2388 RTNL_LINK_INITIALIZING, 2389 } rtnl_link_state:16; 2390 2391 bool needs_free_netdev; 2392 void (*priv_destructor)(struct net_device *dev); 2393 2394 /* mid-layer private */ 2395 void *ml_priv; 2396 enum netdev_ml_priv_type ml_priv_type; 2397 2398 enum netdev_stat_type pcpu_stat_type:8; 2399 union { 2400 struct pcpu_lstats __percpu *lstats; 2401 struct pcpu_sw_netstats __percpu *tstats; 2402 struct pcpu_dstats __percpu *dstats; 2403 }; 2404 2405 #if IS_ENABLED(CONFIG_GARP) 2406 struct garp_port __rcu *garp_port; 2407 #endif 2408 #if IS_ENABLED(CONFIG_MRP) 2409 struct mrp_port __rcu *mrp_port; 2410 #endif 2411 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2412 struct dm_hw_stat_delta __rcu *dm_private; 2413 #endif 2414 struct device dev; 2415 const struct attribute_group *sysfs_groups[4]; 2416 const struct attribute_group *sysfs_rx_queue_group; 2417 2418 const struct rtnl_link_ops *rtnl_link_ops; 2419 2420 /* for setting kernel sock attribute on TCP connection setup */ 2421 #define GSO_MAX_SEGS 65535u 2422 #define GSO_LEGACY_MAX_SIZE 65536u 2423 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2424 * and shinfo->gso_segs is a 16bit field. 2425 */ 2426 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2427 2428 #define TSO_LEGACY_MAX_SIZE 65536 2429 #define TSO_MAX_SIZE UINT_MAX 2430 unsigned int tso_max_size; 2431 #define TSO_MAX_SEGS U16_MAX 2432 u16 tso_max_segs; 2433 2434 #ifdef CONFIG_DCB 2435 const struct dcbnl_rtnl_ops *dcbnl_ops; 2436 #endif 2437 u8 prio_tc_map[TC_BITMASK + 1]; 2438 2439 #if IS_ENABLED(CONFIG_FCOE) 2440 unsigned int fcoe_ddp_xid; 2441 #endif 2442 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2443 struct netprio_map __rcu *priomap; 2444 #endif 2445 struct phy_link_topology link_topo; 2446 struct phy_device *phydev; 2447 struct sfp_bus *sfp_bus; 2448 struct lock_class_key *qdisc_tx_busylock; 2449 bool proto_down; 2450 unsigned wol_enabled:1; 2451 unsigned threaded:1; 2452 2453 struct list_head net_notifier_list; 2454 2455 #if IS_ENABLED(CONFIG_MACSEC) 2456 /* MACsec management functions */ 2457 const struct macsec_ops *macsec_ops; 2458 #endif 2459 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2460 struct udp_tunnel_nic *udp_tunnel_nic; 2461 2462 /* protected by rtnl_lock */ 2463 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2464 2465 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2466 netdevice_tracker linkwatch_dev_tracker; 2467 netdevice_tracker watchdog_dev_tracker; 2468 netdevice_tracker dev_registered_tracker; 2469 struct rtnl_hw_stats64 *offload_xstats_l3; 2470 2471 struct devlink_port *devlink_port; 2472 2473 #if IS_ENABLED(CONFIG_DPLL) 2474 struct dpll_pin *dpll_pin; 2475 #endif 2476 #if IS_ENABLED(CONFIG_PAGE_POOL) 2477 /** @page_pools: page pools created for this netdevice */ 2478 struct hlist_head page_pools; 2479 #endif 2480 }; 2481 #define to_net_dev(d) container_of(d, struct net_device, dev) 2482 2483 /* 2484 * Driver should use this to assign devlink port instance to a netdevice 2485 * before it registers the netdevice. Therefore devlink_port is static 2486 * during the netdev lifetime after it is registered. 2487 */ 2488 #define SET_NETDEV_DEVLINK_PORT(dev, port) \ 2489 ({ \ 2490 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \ 2491 ((dev)->devlink_port = (port)); \ 2492 }) 2493 2494 static inline bool netif_elide_gro(const struct net_device *dev) 2495 { 2496 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2497 return true; 2498 return false; 2499 } 2500 2501 #define NETDEV_ALIGN 32 2502 2503 static inline 2504 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2505 { 2506 return dev->prio_tc_map[prio & TC_BITMASK]; 2507 } 2508 2509 static inline 2510 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2511 { 2512 if (tc >= dev->num_tc) 2513 return -EINVAL; 2514 2515 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2516 return 0; 2517 } 2518 2519 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2520 void netdev_reset_tc(struct net_device *dev); 2521 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2522 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2523 2524 static inline 2525 int netdev_get_num_tc(struct net_device *dev) 2526 { 2527 return dev->num_tc; 2528 } 2529 2530 static inline void net_prefetch(void *p) 2531 { 2532 prefetch(p); 2533 #if L1_CACHE_BYTES < 128 2534 prefetch((u8 *)p + L1_CACHE_BYTES); 2535 #endif 2536 } 2537 2538 static inline void net_prefetchw(void *p) 2539 { 2540 prefetchw(p); 2541 #if L1_CACHE_BYTES < 128 2542 prefetchw((u8 *)p + L1_CACHE_BYTES); 2543 #endif 2544 } 2545 2546 void netdev_unbind_sb_channel(struct net_device *dev, 2547 struct net_device *sb_dev); 2548 int netdev_bind_sb_channel_queue(struct net_device *dev, 2549 struct net_device *sb_dev, 2550 u8 tc, u16 count, u16 offset); 2551 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2552 static inline int netdev_get_sb_channel(struct net_device *dev) 2553 { 2554 return max_t(int, -dev->num_tc, 0); 2555 } 2556 2557 static inline 2558 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2559 unsigned int index) 2560 { 2561 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues); 2562 return &dev->_tx[index]; 2563 } 2564 2565 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2566 const struct sk_buff *skb) 2567 { 2568 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2569 } 2570 2571 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2572 void (*f)(struct net_device *, 2573 struct netdev_queue *, 2574 void *), 2575 void *arg) 2576 { 2577 unsigned int i; 2578 2579 for (i = 0; i < dev->num_tx_queues; i++) 2580 f(dev, &dev->_tx[i], arg); 2581 } 2582 2583 #define netdev_lockdep_set_classes(dev) \ 2584 { \ 2585 static struct lock_class_key qdisc_tx_busylock_key; \ 2586 static struct lock_class_key qdisc_xmit_lock_key; \ 2587 static struct lock_class_key dev_addr_list_lock_key; \ 2588 unsigned int i; \ 2589 \ 2590 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2591 lockdep_set_class(&(dev)->addr_list_lock, \ 2592 &dev_addr_list_lock_key); \ 2593 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2594 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2595 &qdisc_xmit_lock_key); \ 2596 } 2597 2598 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2599 struct net_device *sb_dev); 2600 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2601 struct sk_buff *skb, 2602 struct net_device *sb_dev); 2603 2604 /* returns the headroom that the master device needs to take in account 2605 * when forwarding to this dev 2606 */ 2607 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2608 { 2609 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2610 } 2611 2612 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2613 { 2614 if (dev->netdev_ops->ndo_set_rx_headroom) 2615 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2616 } 2617 2618 /* set the device rx headroom to the dev's default */ 2619 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2620 { 2621 netdev_set_rx_headroom(dev, -1); 2622 } 2623 2624 static inline void *netdev_get_ml_priv(struct net_device *dev, 2625 enum netdev_ml_priv_type type) 2626 { 2627 if (dev->ml_priv_type != type) 2628 return NULL; 2629 2630 return dev->ml_priv; 2631 } 2632 2633 static inline void netdev_set_ml_priv(struct net_device *dev, 2634 void *ml_priv, 2635 enum netdev_ml_priv_type type) 2636 { 2637 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2638 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2639 dev->ml_priv_type, type); 2640 WARN(!dev->ml_priv_type && dev->ml_priv, 2641 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2642 2643 dev->ml_priv = ml_priv; 2644 dev->ml_priv_type = type; 2645 } 2646 2647 /* 2648 * Net namespace inlines 2649 */ 2650 static inline 2651 struct net *dev_net(const struct net_device *dev) 2652 { 2653 return read_pnet(&dev->nd_net); 2654 } 2655 2656 static inline 2657 void dev_net_set(struct net_device *dev, struct net *net) 2658 { 2659 write_pnet(&dev->nd_net, net); 2660 } 2661 2662 /** 2663 * netdev_priv - access network device private data 2664 * @dev: network device 2665 * 2666 * Get network device private data 2667 */ 2668 static inline void *netdev_priv(const struct net_device *dev) 2669 { 2670 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2671 } 2672 2673 /* Set the sysfs physical device reference for the network logical device 2674 * if set prior to registration will cause a symlink during initialization. 2675 */ 2676 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2677 2678 /* Set the sysfs device type for the network logical device to allow 2679 * fine-grained identification of different network device types. For 2680 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2681 */ 2682 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2683 2684 void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index, 2685 enum netdev_queue_type type, 2686 struct napi_struct *napi); 2687 2688 static inline void netif_napi_set_irq(struct napi_struct *napi, int irq) 2689 { 2690 napi->irq = irq; 2691 } 2692 2693 /* Default NAPI poll() weight 2694 * Device drivers are strongly advised to not use bigger value 2695 */ 2696 #define NAPI_POLL_WEIGHT 64 2697 2698 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2699 int (*poll)(struct napi_struct *, int), int weight); 2700 2701 /** 2702 * netif_napi_add() - initialize a NAPI context 2703 * @dev: network device 2704 * @napi: NAPI context 2705 * @poll: polling function 2706 * 2707 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2708 * *any* of the other NAPI-related functions. 2709 */ 2710 static inline void 2711 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2712 int (*poll)(struct napi_struct *, int)) 2713 { 2714 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2715 } 2716 2717 static inline void 2718 netif_napi_add_tx_weight(struct net_device *dev, 2719 struct napi_struct *napi, 2720 int (*poll)(struct napi_struct *, int), 2721 int weight) 2722 { 2723 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2724 netif_napi_add_weight(dev, napi, poll, weight); 2725 } 2726 2727 /** 2728 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2729 * @dev: network device 2730 * @napi: NAPI context 2731 * @poll: polling function 2732 * 2733 * This variant of netif_napi_add() should be used from drivers using NAPI 2734 * to exclusively poll a TX queue. 2735 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2736 */ 2737 static inline void netif_napi_add_tx(struct net_device *dev, 2738 struct napi_struct *napi, 2739 int (*poll)(struct napi_struct *, int)) 2740 { 2741 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2742 } 2743 2744 /** 2745 * __netif_napi_del - remove a NAPI context 2746 * @napi: NAPI context 2747 * 2748 * Warning: caller must observe RCU grace period before freeing memory 2749 * containing @napi. Drivers might want to call this helper to combine 2750 * all the needed RCU grace periods into a single one. 2751 */ 2752 void __netif_napi_del(struct napi_struct *napi); 2753 2754 /** 2755 * netif_napi_del - remove a NAPI context 2756 * @napi: NAPI context 2757 * 2758 * netif_napi_del() removes a NAPI context from the network device NAPI list 2759 */ 2760 static inline void netif_napi_del(struct napi_struct *napi) 2761 { 2762 __netif_napi_del(napi); 2763 synchronize_net(); 2764 } 2765 2766 struct packet_type { 2767 __be16 type; /* This is really htons(ether_type). */ 2768 bool ignore_outgoing; 2769 struct net_device *dev; /* NULL is wildcarded here */ 2770 netdevice_tracker dev_tracker; 2771 int (*func) (struct sk_buff *, 2772 struct net_device *, 2773 struct packet_type *, 2774 struct net_device *); 2775 void (*list_func) (struct list_head *, 2776 struct packet_type *, 2777 struct net_device *); 2778 bool (*id_match)(struct packet_type *ptype, 2779 struct sock *sk); 2780 struct net *af_packet_net; 2781 void *af_packet_priv; 2782 struct list_head list; 2783 }; 2784 2785 struct offload_callbacks { 2786 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2787 netdev_features_t features); 2788 struct sk_buff *(*gro_receive)(struct list_head *head, 2789 struct sk_buff *skb); 2790 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2791 }; 2792 2793 struct packet_offload { 2794 __be16 type; /* This is really htons(ether_type). */ 2795 u16 priority; 2796 struct offload_callbacks callbacks; 2797 struct list_head list; 2798 }; 2799 2800 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2801 struct pcpu_sw_netstats { 2802 u64_stats_t rx_packets; 2803 u64_stats_t rx_bytes; 2804 u64_stats_t tx_packets; 2805 u64_stats_t tx_bytes; 2806 struct u64_stats_sync syncp; 2807 } __aligned(4 * sizeof(u64)); 2808 2809 struct pcpu_dstats { 2810 u64 rx_packets; 2811 u64 rx_bytes; 2812 u64 rx_drops; 2813 u64 tx_packets; 2814 u64 tx_bytes; 2815 u64 tx_drops; 2816 struct u64_stats_sync syncp; 2817 } __aligned(8 * sizeof(u64)); 2818 2819 struct pcpu_lstats { 2820 u64_stats_t packets; 2821 u64_stats_t bytes; 2822 struct u64_stats_sync syncp; 2823 } __aligned(2 * sizeof(u64)); 2824 2825 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2826 2827 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2828 { 2829 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2830 2831 u64_stats_update_begin(&tstats->syncp); 2832 u64_stats_add(&tstats->rx_bytes, len); 2833 u64_stats_inc(&tstats->rx_packets); 2834 u64_stats_update_end(&tstats->syncp); 2835 } 2836 2837 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2838 unsigned int packets, 2839 unsigned int len) 2840 { 2841 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2842 2843 u64_stats_update_begin(&tstats->syncp); 2844 u64_stats_add(&tstats->tx_bytes, len); 2845 u64_stats_add(&tstats->tx_packets, packets); 2846 u64_stats_update_end(&tstats->syncp); 2847 } 2848 2849 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2850 { 2851 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2852 2853 u64_stats_update_begin(&lstats->syncp); 2854 u64_stats_add(&lstats->bytes, len); 2855 u64_stats_inc(&lstats->packets); 2856 u64_stats_update_end(&lstats->syncp); 2857 } 2858 2859 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2860 ({ \ 2861 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2862 if (pcpu_stats) { \ 2863 int __cpu; \ 2864 for_each_possible_cpu(__cpu) { \ 2865 typeof(type) *stat; \ 2866 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2867 u64_stats_init(&stat->syncp); \ 2868 } \ 2869 } \ 2870 pcpu_stats; \ 2871 }) 2872 2873 #define netdev_alloc_pcpu_stats(type) \ 2874 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2875 2876 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2877 ({ \ 2878 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2879 if (pcpu_stats) { \ 2880 int __cpu; \ 2881 for_each_possible_cpu(__cpu) { \ 2882 typeof(type) *stat; \ 2883 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2884 u64_stats_init(&stat->syncp); \ 2885 } \ 2886 } \ 2887 pcpu_stats; \ 2888 }) 2889 2890 enum netdev_lag_tx_type { 2891 NETDEV_LAG_TX_TYPE_UNKNOWN, 2892 NETDEV_LAG_TX_TYPE_RANDOM, 2893 NETDEV_LAG_TX_TYPE_BROADCAST, 2894 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2895 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2896 NETDEV_LAG_TX_TYPE_HASH, 2897 }; 2898 2899 enum netdev_lag_hash { 2900 NETDEV_LAG_HASH_NONE, 2901 NETDEV_LAG_HASH_L2, 2902 NETDEV_LAG_HASH_L34, 2903 NETDEV_LAG_HASH_L23, 2904 NETDEV_LAG_HASH_E23, 2905 NETDEV_LAG_HASH_E34, 2906 NETDEV_LAG_HASH_VLAN_SRCMAC, 2907 NETDEV_LAG_HASH_UNKNOWN, 2908 }; 2909 2910 struct netdev_lag_upper_info { 2911 enum netdev_lag_tx_type tx_type; 2912 enum netdev_lag_hash hash_type; 2913 }; 2914 2915 struct netdev_lag_lower_state_info { 2916 u8 link_up : 1, 2917 tx_enabled : 1; 2918 }; 2919 2920 #include <linux/notifier.h> 2921 2922 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2923 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2924 * adding new types. 2925 */ 2926 enum netdev_cmd { 2927 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2928 NETDEV_DOWN, 2929 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2930 detected a hardware crash and restarted 2931 - we can use this eg to kick tcp sessions 2932 once done */ 2933 NETDEV_CHANGE, /* Notify device state change */ 2934 NETDEV_REGISTER, 2935 NETDEV_UNREGISTER, 2936 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2937 NETDEV_CHANGEADDR, /* notify after the address change */ 2938 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2939 NETDEV_GOING_DOWN, 2940 NETDEV_CHANGENAME, 2941 NETDEV_FEAT_CHANGE, 2942 NETDEV_BONDING_FAILOVER, 2943 NETDEV_PRE_UP, 2944 NETDEV_PRE_TYPE_CHANGE, 2945 NETDEV_POST_TYPE_CHANGE, 2946 NETDEV_POST_INIT, 2947 NETDEV_PRE_UNINIT, 2948 NETDEV_RELEASE, 2949 NETDEV_NOTIFY_PEERS, 2950 NETDEV_JOIN, 2951 NETDEV_CHANGEUPPER, 2952 NETDEV_RESEND_IGMP, 2953 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2954 NETDEV_CHANGEINFODATA, 2955 NETDEV_BONDING_INFO, 2956 NETDEV_PRECHANGEUPPER, 2957 NETDEV_CHANGELOWERSTATE, 2958 NETDEV_UDP_TUNNEL_PUSH_INFO, 2959 NETDEV_UDP_TUNNEL_DROP_INFO, 2960 NETDEV_CHANGE_TX_QUEUE_LEN, 2961 NETDEV_CVLAN_FILTER_PUSH_INFO, 2962 NETDEV_CVLAN_FILTER_DROP_INFO, 2963 NETDEV_SVLAN_FILTER_PUSH_INFO, 2964 NETDEV_SVLAN_FILTER_DROP_INFO, 2965 NETDEV_OFFLOAD_XSTATS_ENABLE, 2966 NETDEV_OFFLOAD_XSTATS_DISABLE, 2967 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 2968 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 2969 NETDEV_XDP_FEAT_CHANGE, 2970 }; 2971 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2972 2973 int register_netdevice_notifier(struct notifier_block *nb); 2974 int unregister_netdevice_notifier(struct notifier_block *nb); 2975 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2976 int unregister_netdevice_notifier_net(struct net *net, 2977 struct notifier_block *nb); 2978 int register_netdevice_notifier_dev_net(struct net_device *dev, 2979 struct notifier_block *nb, 2980 struct netdev_net_notifier *nn); 2981 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2982 struct notifier_block *nb, 2983 struct netdev_net_notifier *nn); 2984 2985 struct netdev_notifier_info { 2986 struct net_device *dev; 2987 struct netlink_ext_ack *extack; 2988 }; 2989 2990 struct netdev_notifier_info_ext { 2991 struct netdev_notifier_info info; /* must be first */ 2992 union { 2993 u32 mtu; 2994 } ext; 2995 }; 2996 2997 struct netdev_notifier_change_info { 2998 struct netdev_notifier_info info; /* must be first */ 2999 unsigned int flags_changed; 3000 }; 3001 3002 struct netdev_notifier_changeupper_info { 3003 struct netdev_notifier_info info; /* must be first */ 3004 struct net_device *upper_dev; /* new upper dev */ 3005 bool master; /* is upper dev master */ 3006 bool linking; /* is the notification for link or unlink */ 3007 void *upper_info; /* upper dev info */ 3008 }; 3009 3010 struct netdev_notifier_changelowerstate_info { 3011 struct netdev_notifier_info info; /* must be first */ 3012 void *lower_state_info; /* is lower dev state */ 3013 }; 3014 3015 struct netdev_notifier_pre_changeaddr_info { 3016 struct netdev_notifier_info info; /* must be first */ 3017 const unsigned char *dev_addr; 3018 }; 3019 3020 enum netdev_offload_xstats_type { 3021 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 3022 }; 3023 3024 struct netdev_notifier_offload_xstats_info { 3025 struct netdev_notifier_info info; /* must be first */ 3026 enum netdev_offload_xstats_type type; 3027 3028 union { 3029 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 3030 struct netdev_notifier_offload_xstats_rd *report_delta; 3031 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 3032 struct netdev_notifier_offload_xstats_ru *report_used; 3033 }; 3034 }; 3035 3036 int netdev_offload_xstats_enable(struct net_device *dev, 3037 enum netdev_offload_xstats_type type, 3038 struct netlink_ext_ack *extack); 3039 int netdev_offload_xstats_disable(struct net_device *dev, 3040 enum netdev_offload_xstats_type type); 3041 bool netdev_offload_xstats_enabled(const struct net_device *dev, 3042 enum netdev_offload_xstats_type type); 3043 int netdev_offload_xstats_get(struct net_device *dev, 3044 enum netdev_offload_xstats_type type, 3045 struct rtnl_hw_stats64 *stats, bool *used, 3046 struct netlink_ext_ack *extack); 3047 void 3048 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 3049 const struct rtnl_hw_stats64 *stats); 3050 void 3051 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 3052 void netdev_offload_xstats_push_delta(struct net_device *dev, 3053 enum netdev_offload_xstats_type type, 3054 const struct rtnl_hw_stats64 *stats); 3055 3056 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 3057 struct net_device *dev) 3058 { 3059 info->dev = dev; 3060 info->extack = NULL; 3061 } 3062 3063 static inline struct net_device * 3064 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 3065 { 3066 return info->dev; 3067 } 3068 3069 static inline struct netlink_ext_ack * 3070 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 3071 { 3072 return info->extack; 3073 } 3074 3075 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 3076 int call_netdevice_notifiers_info(unsigned long val, 3077 struct netdev_notifier_info *info); 3078 3079 extern rwlock_t dev_base_lock; /* Device list lock */ 3080 3081 #define for_each_netdev(net, d) \ 3082 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 3083 #define for_each_netdev_reverse(net, d) \ 3084 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 3085 #define for_each_netdev_rcu(net, d) \ 3086 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 3087 #define for_each_netdev_safe(net, d, n) \ 3088 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 3089 #define for_each_netdev_continue(net, d) \ 3090 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 3091 #define for_each_netdev_continue_reverse(net, d) \ 3092 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 3093 dev_list) 3094 #define for_each_netdev_continue_rcu(net, d) \ 3095 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 3096 #define for_each_netdev_in_bond_rcu(bond, slave) \ 3097 for_each_netdev_rcu(&init_net, slave) \ 3098 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 3099 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 3100 3101 #define for_each_netdev_dump(net, d, ifindex) \ 3102 xa_for_each_start(&(net)->dev_by_index, (ifindex), (d), (ifindex)) 3103 3104 static inline struct net_device *next_net_device(struct net_device *dev) 3105 { 3106 struct list_head *lh; 3107 struct net *net; 3108 3109 net = dev_net(dev); 3110 lh = dev->dev_list.next; 3111 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3112 } 3113 3114 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 3115 { 3116 struct list_head *lh; 3117 struct net *net; 3118 3119 net = dev_net(dev); 3120 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 3121 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3122 } 3123 3124 static inline struct net_device *first_net_device(struct net *net) 3125 { 3126 return list_empty(&net->dev_base_head) ? NULL : 3127 net_device_entry(net->dev_base_head.next); 3128 } 3129 3130 static inline struct net_device *first_net_device_rcu(struct net *net) 3131 { 3132 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 3133 3134 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3135 } 3136 3137 int netdev_boot_setup_check(struct net_device *dev); 3138 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 3139 const char *hwaddr); 3140 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 3141 void dev_add_pack(struct packet_type *pt); 3142 void dev_remove_pack(struct packet_type *pt); 3143 void __dev_remove_pack(struct packet_type *pt); 3144 void dev_add_offload(struct packet_offload *po); 3145 void dev_remove_offload(struct packet_offload *po); 3146 3147 int dev_get_iflink(const struct net_device *dev); 3148 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 3149 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 3150 struct net_device_path_stack *stack); 3151 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 3152 unsigned short mask); 3153 struct net_device *dev_get_by_name(struct net *net, const char *name); 3154 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 3155 struct net_device *__dev_get_by_name(struct net *net, const char *name); 3156 bool netdev_name_in_use(struct net *net, const char *name); 3157 int dev_alloc_name(struct net_device *dev, const char *name); 3158 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 3159 void dev_close(struct net_device *dev); 3160 void dev_close_many(struct list_head *head, bool unlink); 3161 void dev_disable_lro(struct net_device *dev); 3162 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 3163 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 3164 struct net_device *sb_dev); 3165 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 3166 struct net_device *sb_dev); 3167 3168 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3169 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3170 3171 static inline int dev_queue_xmit(struct sk_buff *skb) 3172 { 3173 return __dev_queue_xmit(skb, NULL); 3174 } 3175 3176 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3177 struct net_device *sb_dev) 3178 { 3179 return __dev_queue_xmit(skb, sb_dev); 3180 } 3181 3182 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3183 { 3184 int ret; 3185 3186 ret = __dev_direct_xmit(skb, queue_id); 3187 if (!dev_xmit_complete(ret)) 3188 kfree_skb(skb); 3189 return ret; 3190 } 3191 3192 int register_netdevice(struct net_device *dev); 3193 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3194 void unregister_netdevice_many(struct list_head *head); 3195 static inline void unregister_netdevice(struct net_device *dev) 3196 { 3197 unregister_netdevice_queue(dev, NULL); 3198 } 3199 3200 int netdev_refcnt_read(const struct net_device *dev); 3201 void free_netdev(struct net_device *dev); 3202 void netdev_freemem(struct net_device *dev); 3203 int init_dummy_netdev(struct net_device *dev); 3204 3205 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3206 struct sk_buff *skb, 3207 bool all_slaves); 3208 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3209 struct sock *sk); 3210 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3211 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3212 struct net_device *netdev_get_by_index(struct net *net, int ifindex, 3213 netdevice_tracker *tracker, gfp_t gfp); 3214 struct net_device *netdev_get_by_name(struct net *net, const char *name, 3215 netdevice_tracker *tracker, gfp_t gfp); 3216 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3217 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 3218 3219 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3220 unsigned short type, 3221 const void *daddr, const void *saddr, 3222 unsigned int len) 3223 { 3224 if (!dev->header_ops || !dev->header_ops->create) 3225 return 0; 3226 3227 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3228 } 3229 3230 static inline int dev_parse_header(const struct sk_buff *skb, 3231 unsigned char *haddr) 3232 { 3233 const struct net_device *dev = skb->dev; 3234 3235 if (!dev->header_ops || !dev->header_ops->parse) 3236 return 0; 3237 return dev->header_ops->parse(skb, haddr); 3238 } 3239 3240 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3241 { 3242 const struct net_device *dev = skb->dev; 3243 3244 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3245 return 0; 3246 return dev->header_ops->parse_protocol(skb); 3247 } 3248 3249 /* ll_header must have at least hard_header_len allocated */ 3250 static inline bool dev_validate_header(const struct net_device *dev, 3251 char *ll_header, int len) 3252 { 3253 if (likely(len >= dev->hard_header_len)) 3254 return true; 3255 if (len < dev->min_header_len) 3256 return false; 3257 3258 if (capable(CAP_SYS_RAWIO)) { 3259 memset(ll_header + len, 0, dev->hard_header_len - len); 3260 return true; 3261 } 3262 3263 if (dev->header_ops && dev->header_ops->validate) 3264 return dev->header_ops->validate(ll_header, len); 3265 3266 return false; 3267 } 3268 3269 static inline bool dev_has_header(const struct net_device *dev) 3270 { 3271 return dev->header_ops && dev->header_ops->create; 3272 } 3273 3274 /* 3275 * Incoming packets are placed on per-CPU queues 3276 */ 3277 struct softnet_data { 3278 struct list_head poll_list; 3279 struct sk_buff_head process_queue; 3280 3281 /* stats */ 3282 unsigned int processed; 3283 unsigned int time_squeeze; 3284 #ifdef CONFIG_RPS 3285 struct softnet_data *rps_ipi_list; 3286 #endif 3287 3288 bool in_net_rx_action; 3289 bool in_napi_threaded_poll; 3290 3291 #ifdef CONFIG_NET_FLOW_LIMIT 3292 struct sd_flow_limit __rcu *flow_limit; 3293 #endif 3294 struct Qdisc *output_queue; 3295 struct Qdisc **output_queue_tailp; 3296 struct sk_buff *completion_queue; 3297 #ifdef CONFIG_XFRM_OFFLOAD 3298 struct sk_buff_head xfrm_backlog; 3299 #endif 3300 /* written and read only by owning cpu: */ 3301 struct { 3302 u16 recursion; 3303 u8 more; 3304 #ifdef CONFIG_NET_EGRESS 3305 u8 skip_txqueue; 3306 #endif 3307 } xmit; 3308 #ifdef CONFIG_RPS 3309 /* input_queue_head should be written by cpu owning this struct, 3310 * and only read by other cpus. Worth using a cache line. 3311 */ 3312 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3313 3314 /* Elements below can be accessed between CPUs for RPS/RFS */ 3315 call_single_data_t csd ____cacheline_aligned_in_smp; 3316 struct softnet_data *rps_ipi_next; 3317 unsigned int cpu; 3318 unsigned int input_queue_tail; 3319 #endif 3320 unsigned int received_rps; 3321 unsigned int dropped; 3322 struct sk_buff_head input_pkt_queue; 3323 struct napi_struct backlog; 3324 3325 /* Another possibly contended cache line */ 3326 spinlock_t defer_lock ____cacheline_aligned_in_smp; 3327 int defer_count; 3328 int defer_ipi_scheduled; 3329 struct sk_buff *defer_list; 3330 call_single_data_t defer_csd; 3331 }; 3332 3333 static inline void input_queue_head_incr(struct softnet_data *sd) 3334 { 3335 #ifdef CONFIG_RPS 3336 sd->input_queue_head++; 3337 #endif 3338 } 3339 3340 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3341 unsigned int *qtail) 3342 { 3343 #ifdef CONFIG_RPS 3344 *qtail = ++sd->input_queue_tail; 3345 #endif 3346 } 3347 3348 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3349 3350 static inline int dev_recursion_level(void) 3351 { 3352 return this_cpu_read(softnet_data.xmit.recursion); 3353 } 3354 3355 #define XMIT_RECURSION_LIMIT 8 3356 static inline bool dev_xmit_recursion(void) 3357 { 3358 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3359 XMIT_RECURSION_LIMIT); 3360 } 3361 3362 static inline void dev_xmit_recursion_inc(void) 3363 { 3364 __this_cpu_inc(softnet_data.xmit.recursion); 3365 } 3366 3367 static inline void dev_xmit_recursion_dec(void) 3368 { 3369 __this_cpu_dec(softnet_data.xmit.recursion); 3370 } 3371 3372 void __netif_schedule(struct Qdisc *q); 3373 void netif_schedule_queue(struct netdev_queue *txq); 3374 3375 static inline void netif_tx_schedule_all(struct net_device *dev) 3376 { 3377 unsigned int i; 3378 3379 for (i = 0; i < dev->num_tx_queues; i++) 3380 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3381 } 3382 3383 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3384 { 3385 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3386 } 3387 3388 /** 3389 * netif_start_queue - allow transmit 3390 * @dev: network device 3391 * 3392 * Allow upper layers to call the device hard_start_xmit routine. 3393 */ 3394 static inline void netif_start_queue(struct net_device *dev) 3395 { 3396 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3397 } 3398 3399 static inline void netif_tx_start_all_queues(struct net_device *dev) 3400 { 3401 unsigned int i; 3402 3403 for (i = 0; i < dev->num_tx_queues; i++) { 3404 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3405 netif_tx_start_queue(txq); 3406 } 3407 } 3408 3409 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3410 3411 /** 3412 * netif_wake_queue - restart transmit 3413 * @dev: network device 3414 * 3415 * Allow upper layers to call the device hard_start_xmit routine. 3416 * Used for flow control when transmit resources are available. 3417 */ 3418 static inline void netif_wake_queue(struct net_device *dev) 3419 { 3420 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3421 } 3422 3423 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3424 { 3425 unsigned int i; 3426 3427 for (i = 0; i < dev->num_tx_queues; i++) { 3428 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3429 netif_tx_wake_queue(txq); 3430 } 3431 } 3432 3433 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3434 { 3435 /* Must be an atomic op see netif_txq_try_stop() */ 3436 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3437 } 3438 3439 /** 3440 * netif_stop_queue - stop transmitted packets 3441 * @dev: network device 3442 * 3443 * Stop upper layers calling the device hard_start_xmit routine. 3444 * Used for flow control when transmit resources are unavailable. 3445 */ 3446 static inline void netif_stop_queue(struct net_device *dev) 3447 { 3448 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3449 } 3450 3451 void netif_tx_stop_all_queues(struct net_device *dev); 3452 3453 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3454 { 3455 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3456 } 3457 3458 /** 3459 * netif_queue_stopped - test if transmit queue is flowblocked 3460 * @dev: network device 3461 * 3462 * Test if transmit queue on device is currently unable to send. 3463 */ 3464 static inline bool netif_queue_stopped(const struct net_device *dev) 3465 { 3466 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3467 } 3468 3469 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3470 { 3471 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3472 } 3473 3474 static inline bool 3475 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3476 { 3477 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3478 } 3479 3480 static inline bool 3481 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3482 { 3483 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3484 } 3485 3486 /** 3487 * netdev_queue_set_dql_min_limit - set dql minimum limit 3488 * @dev_queue: pointer to transmit queue 3489 * @min_limit: dql minimum limit 3490 * 3491 * Forces xmit_more() to return true until the minimum threshold 3492 * defined by @min_limit is reached (or until the tx queue is 3493 * empty). Warning: to be use with care, misuse will impact the 3494 * latency. 3495 */ 3496 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3497 unsigned int min_limit) 3498 { 3499 #ifdef CONFIG_BQL 3500 dev_queue->dql.min_limit = min_limit; 3501 #endif 3502 } 3503 3504 /** 3505 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3506 * @dev_queue: pointer to transmit queue 3507 * 3508 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3509 * to give appropriate hint to the CPU. 3510 */ 3511 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3512 { 3513 #ifdef CONFIG_BQL 3514 prefetchw(&dev_queue->dql.num_queued); 3515 #endif 3516 } 3517 3518 /** 3519 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3520 * @dev_queue: pointer to transmit queue 3521 * 3522 * BQL enabled drivers might use this helper in their TX completion path, 3523 * to give appropriate hint to the CPU. 3524 */ 3525 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3526 { 3527 #ifdef CONFIG_BQL 3528 prefetchw(&dev_queue->dql.limit); 3529 #endif 3530 } 3531 3532 /** 3533 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3534 * @dev_queue: network device queue 3535 * @bytes: number of bytes queued to the device queue 3536 * 3537 * Report the number of bytes queued for sending/completion to the network 3538 * device hardware queue. @bytes should be a good approximation and should 3539 * exactly match netdev_completed_queue() @bytes. 3540 * This is typically called once per packet, from ndo_start_xmit(). 3541 */ 3542 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3543 unsigned int bytes) 3544 { 3545 #ifdef CONFIG_BQL 3546 dql_queued(&dev_queue->dql, bytes); 3547 3548 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3549 return; 3550 3551 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3552 3553 /* 3554 * The XOFF flag must be set before checking the dql_avail below, 3555 * because in netdev_tx_completed_queue we update the dql_completed 3556 * before checking the XOFF flag. 3557 */ 3558 smp_mb(); 3559 3560 /* check again in case another CPU has just made room avail */ 3561 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3562 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3563 #endif 3564 } 3565 3566 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3567 * that they should not test BQL status themselves. 3568 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3569 * skb of a batch. 3570 * Returns true if the doorbell must be used to kick the NIC. 3571 */ 3572 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3573 unsigned int bytes, 3574 bool xmit_more) 3575 { 3576 if (xmit_more) { 3577 #ifdef CONFIG_BQL 3578 dql_queued(&dev_queue->dql, bytes); 3579 #endif 3580 return netif_tx_queue_stopped(dev_queue); 3581 } 3582 netdev_tx_sent_queue(dev_queue, bytes); 3583 return true; 3584 } 3585 3586 /** 3587 * netdev_sent_queue - report the number of bytes queued to hardware 3588 * @dev: network device 3589 * @bytes: number of bytes queued to the hardware device queue 3590 * 3591 * Report the number of bytes queued for sending/completion to the network 3592 * device hardware queue#0. @bytes should be a good approximation and should 3593 * exactly match netdev_completed_queue() @bytes. 3594 * This is typically called once per packet, from ndo_start_xmit(). 3595 */ 3596 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3597 { 3598 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3599 } 3600 3601 static inline bool __netdev_sent_queue(struct net_device *dev, 3602 unsigned int bytes, 3603 bool xmit_more) 3604 { 3605 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3606 xmit_more); 3607 } 3608 3609 /** 3610 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3611 * @dev_queue: network device queue 3612 * @pkts: number of packets (currently ignored) 3613 * @bytes: number of bytes dequeued from the device queue 3614 * 3615 * Must be called at most once per TX completion round (and not per 3616 * individual packet), so that BQL can adjust its limits appropriately. 3617 */ 3618 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3619 unsigned int pkts, unsigned int bytes) 3620 { 3621 #ifdef CONFIG_BQL 3622 if (unlikely(!bytes)) 3623 return; 3624 3625 dql_completed(&dev_queue->dql, bytes); 3626 3627 /* 3628 * Without the memory barrier there is a small possiblity that 3629 * netdev_tx_sent_queue will miss the update and cause the queue to 3630 * be stopped forever 3631 */ 3632 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */ 3633 3634 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3635 return; 3636 3637 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3638 netif_schedule_queue(dev_queue); 3639 #endif 3640 } 3641 3642 /** 3643 * netdev_completed_queue - report bytes and packets completed by device 3644 * @dev: network device 3645 * @pkts: actual number of packets sent over the medium 3646 * @bytes: actual number of bytes sent over the medium 3647 * 3648 * Report the number of bytes and packets transmitted by the network device 3649 * hardware queue over the physical medium, @bytes must exactly match the 3650 * @bytes amount passed to netdev_sent_queue() 3651 */ 3652 static inline void netdev_completed_queue(struct net_device *dev, 3653 unsigned int pkts, unsigned int bytes) 3654 { 3655 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3656 } 3657 3658 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3659 { 3660 #ifdef CONFIG_BQL 3661 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3662 dql_reset(&q->dql); 3663 #endif 3664 } 3665 3666 /** 3667 * netdev_reset_queue - reset the packets and bytes count of a network device 3668 * @dev_queue: network device 3669 * 3670 * Reset the bytes and packet count of a network device and clear the 3671 * software flow control OFF bit for this network device 3672 */ 3673 static inline void netdev_reset_queue(struct net_device *dev_queue) 3674 { 3675 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3676 } 3677 3678 /** 3679 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3680 * @dev: network device 3681 * @queue_index: given tx queue index 3682 * 3683 * Returns 0 if given tx queue index >= number of device tx queues, 3684 * otherwise returns the originally passed tx queue index. 3685 */ 3686 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3687 { 3688 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3689 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3690 dev->name, queue_index, 3691 dev->real_num_tx_queues); 3692 return 0; 3693 } 3694 3695 return queue_index; 3696 } 3697 3698 /** 3699 * netif_running - test if up 3700 * @dev: network device 3701 * 3702 * Test if the device has been brought up. 3703 */ 3704 static inline bool netif_running(const struct net_device *dev) 3705 { 3706 return test_bit(__LINK_STATE_START, &dev->state); 3707 } 3708 3709 /* 3710 * Routines to manage the subqueues on a device. We only need start, 3711 * stop, and a check if it's stopped. All other device management is 3712 * done at the overall netdevice level. 3713 * Also test the device if we're multiqueue. 3714 */ 3715 3716 /** 3717 * netif_start_subqueue - allow sending packets on subqueue 3718 * @dev: network device 3719 * @queue_index: sub queue index 3720 * 3721 * Start individual transmit queue of a device with multiple transmit queues. 3722 */ 3723 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3724 { 3725 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3726 3727 netif_tx_start_queue(txq); 3728 } 3729 3730 /** 3731 * netif_stop_subqueue - stop sending packets on subqueue 3732 * @dev: network device 3733 * @queue_index: sub queue index 3734 * 3735 * Stop individual transmit queue of a device with multiple transmit queues. 3736 */ 3737 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3738 { 3739 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3740 netif_tx_stop_queue(txq); 3741 } 3742 3743 /** 3744 * __netif_subqueue_stopped - test status of subqueue 3745 * @dev: network device 3746 * @queue_index: sub queue index 3747 * 3748 * Check individual transmit queue of a device with multiple transmit queues. 3749 */ 3750 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3751 u16 queue_index) 3752 { 3753 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3754 3755 return netif_tx_queue_stopped(txq); 3756 } 3757 3758 /** 3759 * netif_subqueue_stopped - test status of subqueue 3760 * @dev: network device 3761 * @skb: sub queue buffer pointer 3762 * 3763 * Check individual transmit queue of a device with multiple transmit queues. 3764 */ 3765 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3766 struct sk_buff *skb) 3767 { 3768 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3769 } 3770 3771 /** 3772 * netif_wake_subqueue - allow sending packets on subqueue 3773 * @dev: network device 3774 * @queue_index: sub queue index 3775 * 3776 * Resume individual transmit queue of a device with multiple transmit queues. 3777 */ 3778 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3779 { 3780 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3781 3782 netif_tx_wake_queue(txq); 3783 } 3784 3785 #ifdef CONFIG_XPS 3786 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3787 u16 index); 3788 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3789 u16 index, enum xps_map_type type); 3790 3791 /** 3792 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3793 * @j: CPU/Rx queue index 3794 * @mask: bitmask of all cpus/rx queues 3795 * @nr_bits: number of bits in the bitmask 3796 * 3797 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3798 */ 3799 static inline bool netif_attr_test_mask(unsigned long j, 3800 const unsigned long *mask, 3801 unsigned int nr_bits) 3802 { 3803 cpu_max_bits_warn(j, nr_bits); 3804 return test_bit(j, mask); 3805 } 3806 3807 /** 3808 * netif_attr_test_online - Test for online CPU/Rx queue 3809 * @j: CPU/Rx queue index 3810 * @online_mask: bitmask for CPUs/Rx queues that are online 3811 * @nr_bits: number of bits in the bitmask 3812 * 3813 * Returns true if a CPU/Rx queue is online. 3814 */ 3815 static inline bool netif_attr_test_online(unsigned long j, 3816 const unsigned long *online_mask, 3817 unsigned int nr_bits) 3818 { 3819 cpu_max_bits_warn(j, nr_bits); 3820 3821 if (online_mask) 3822 return test_bit(j, online_mask); 3823 3824 return (j < nr_bits); 3825 } 3826 3827 /** 3828 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3829 * @n: CPU/Rx queue index 3830 * @srcp: the cpumask/Rx queue mask pointer 3831 * @nr_bits: number of bits in the bitmask 3832 * 3833 * Returns >= nr_bits if no further CPUs/Rx queues set. 3834 */ 3835 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3836 unsigned int nr_bits) 3837 { 3838 /* -1 is a legal arg here. */ 3839 if (n != -1) 3840 cpu_max_bits_warn(n, nr_bits); 3841 3842 if (srcp) 3843 return find_next_bit(srcp, nr_bits, n + 1); 3844 3845 return n + 1; 3846 } 3847 3848 /** 3849 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3850 * @n: CPU/Rx queue index 3851 * @src1p: the first CPUs/Rx queues mask pointer 3852 * @src2p: the second CPUs/Rx queues mask pointer 3853 * @nr_bits: number of bits in the bitmask 3854 * 3855 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3856 */ 3857 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3858 const unsigned long *src2p, 3859 unsigned int nr_bits) 3860 { 3861 /* -1 is a legal arg here. */ 3862 if (n != -1) 3863 cpu_max_bits_warn(n, nr_bits); 3864 3865 if (src1p && src2p) 3866 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3867 else if (src1p) 3868 return find_next_bit(src1p, nr_bits, n + 1); 3869 else if (src2p) 3870 return find_next_bit(src2p, nr_bits, n + 1); 3871 3872 return n + 1; 3873 } 3874 #else 3875 static inline int netif_set_xps_queue(struct net_device *dev, 3876 const struct cpumask *mask, 3877 u16 index) 3878 { 3879 return 0; 3880 } 3881 3882 static inline int __netif_set_xps_queue(struct net_device *dev, 3883 const unsigned long *mask, 3884 u16 index, enum xps_map_type type) 3885 { 3886 return 0; 3887 } 3888 #endif 3889 3890 /** 3891 * netif_is_multiqueue - test if device has multiple transmit queues 3892 * @dev: network device 3893 * 3894 * Check if device has multiple transmit queues 3895 */ 3896 static inline bool netif_is_multiqueue(const struct net_device *dev) 3897 { 3898 return dev->num_tx_queues > 1; 3899 } 3900 3901 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3902 3903 #ifdef CONFIG_SYSFS 3904 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3905 #else 3906 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3907 unsigned int rxqs) 3908 { 3909 dev->real_num_rx_queues = rxqs; 3910 return 0; 3911 } 3912 #endif 3913 int netif_set_real_num_queues(struct net_device *dev, 3914 unsigned int txq, unsigned int rxq); 3915 3916 int netif_get_num_default_rss_queues(void); 3917 3918 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason); 3919 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason); 3920 3921 /* 3922 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3923 * interrupt context or with hardware interrupts being disabled. 3924 * (in_hardirq() || irqs_disabled()) 3925 * 3926 * We provide four helpers that can be used in following contexts : 3927 * 3928 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3929 * replacing kfree_skb(skb) 3930 * 3931 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3932 * Typically used in place of consume_skb(skb) in TX completion path 3933 * 3934 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3935 * replacing kfree_skb(skb) 3936 * 3937 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3938 * and consumed a packet. Used in place of consume_skb(skb) 3939 */ 3940 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3941 { 3942 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 3943 } 3944 3945 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3946 { 3947 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED); 3948 } 3949 3950 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3951 { 3952 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 3953 } 3954 3955 static inline void dev_consume_skb_any(struct sk_buff *skb) 3956 { 3957 dev_kfree_skb_any_reason(skb, SKB_CONSUMED); 3958 } 3959 3960 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3961 struct bpf_prog *xdp_prog); 3962 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3963 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3964 int netif_rx(struct sk_buff *skb); 3965 int __netif_rx(struct sk_buff *skb); 3966 3967 int netif_receive_skb(struct sk_buff *skb); 3968 int netif_receive_skb_core(struct sk_buff *skb); 3969 void netif_receive_skb_list_internal(struct list_head *head); 3970 void netif_receive_skb_list(struct list_head *head); 3971 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3972 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3973 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3974 void napi_get_frags_check(struct napi_struct *napi); 3975 gro_result_t napi_gro_frags(struct napi_struct *napi); 3976 struct packet_offload *gro_find_receive_by_type(__be16 type); 3977 struct packet_offload *gro_find_complete_by_type(__be16 type); 3978 3979 static inline void napi_free_frags(struct napi_struct *napi) 3980 { 3981 kfree_skb(napi->skb); 3982 napi->skb = NULL; 3983 } 3984 3985 bool netdev_is_rx_handler_busy(struct net_device *dev); 3986 int netdev_rx_handler_register(struct net_device *dev, 3987 rx_handler_func_t *rx_handler, 3988 void *rx_handler_data); 3989 void netdev_rx_handler_unregister(struct net_device *dev); 3990 3991 bool dev_valid_name(const char *name); 3992 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 3993 { 3994 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 3995 } 3996 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 3997 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 3998 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3999 void __user *data, bool *need_copyout); 4000 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 4001 int generic_hwtstamp_get_lower(struct net_device *dev, 4002 struct kernel_hwtstamp_config *kernel_cfg); 4003 int generic_hwtstamp_set_lower(struct net_device *dev, 4004 struct kernel_hwtstamp_config *kernel_cfg, 4005 struct netlink_ext_ack *extack); 4006 int dev_set_hwtstamp_phylib(struct net_device *dev, 4007 struct kernel_hwtstamp_config *cfg, 4008 struct netlink_ext_ack *extack); 4009 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 4010 unsigned int dev_get_flags(const struct net_device *); 4011 int __dev_change_flags(struct net_device *dev, unsigned int flags, 4012 struct netlink_ext_ack *extack); 4013 int dev_change_flags(struct net_device *dev, unsigned int flags, 4014 struct netlink_ext_ack *extack); 4015 int dev_set_alias(struct net_device *, const char *, size_t); 4016 int dev_get_alias(const struct net_device *, char *, size_t); 4017 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 4018 const char *pat, int new_ifindex); 4019 static inline 4020 int dev_change_net_namespace(struct net_device *dev, struct net *net, 4021 const char *pat) 4022 { 4023 return __dev_change_net_namespace(dev, net, pat, 0); 4024 } 4025 int __dev_set_mtu(struct net_device *, int); 4026 int dev_set_mtu(struct net_device *, int); 4027 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 4028 struct netlink_ext_ack *extack); 4029 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 4030 struct netlink_ext_ack *extack); 4031 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 4032 struct netlink_ext_ack *extack); 4033 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 4034 int dev_get_port_parent_id(struct net_device *dev, 4035 struct netdev_phys_item_id *ppid, bool recurse); 4036 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 4037 void netdev_dpll_pin_set(struct net_device *dev, struct dpll_pin *dpll_pin); 4038 void netdev_dpll_pin_clear(struct net_device *dev); 4039 4040 static inline struct dpll_pin *netdev_dpll_pin(const struct net_device *dev) 4041 { 4042 #if IS_ENABLED(CONFIG_DPLL) 4043 return dev->dpll_pin; 4044 #else 4045 return NULL; 4046 #endif 4047 } 4048 4049 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 4050 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 4051 struct netdev_queue *txq, int *ret); 4052 4053 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 4054 u8 dev_xdp_prog_count(struct net_device *dev); 4055 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 4056 4057 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4058 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4059 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 4060 bool is_skb_forwardable(const struct net_device *dev, 4061 const struct sk_buff *skb); 4062 4063 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 4064 const struct sk_buff *skb, 4065 const bool check_mtu) 4066 { 4067 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 4068 unsigned int len; 4069 4070 if (!(dev->flags & IFF_UP)) 4071 return false; 4072 4073 if (!check_mtu) 4074 return true; 4075 4076 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 4077 if (skb->len <= len) 4078 return true; 4079 4080 /* if TSO is enabled, we don't care about the length as the packet 4081 * could be forwarded without being segmented before 4082 */ 4083 if (skb_is_gso(skb)) 4084 return true; 4085 4086 return false; 4087 } 4088 4089 void netdev_core_stats_inc(struct net_device *dev, u32 offset); 4090 4091 #define DEV_CORE_STATS_INC(FIELD) \ 4092 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 4093 { \ 4094 netdev_core_stats_inc(dev, \ 4095 offsetof(struct net_device_core_stats, FIELD)); \ 4096 } 4097 DEV_CORE_STATS_INC(rx_dropped) 4098 DEV_CORE_STATS_INC(tx_dropped) 4099 DEV_CORE_STATS_INC(rx_nohandler) 4100 DEV_CORE_STATS_INC(rx_otherhost_dropped) 4101 #undef DEV_CORE_STATS_INC 4102 4103 static __always_inline int ____dev_forward_skb(struct net_device *dev, 4104 struct sk_buff *skb, 4105 const bool check_mtu) 4106 { 4107 if (skb_orphan_frags(skb, GFP_ATOMIC) || 4108 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 4109 dev_core_stats_rx_dropped_inc(dev); 4110 kfree_skb(skb); 4111 return NET_RX_DROP; 4112 } 4113 4114 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 4115 skb->priority = 0; 4116 return 0; 4117 } 4118 4119 bool dev_nit_active(struct net_device *dev); 4120 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 4121 4122 static inline void __dev_put(struct net_device *dev) 4123 { 4124 if (dev) { 4125 #ifdef CONFIG_PCPU_DEV_REFCNT 4126 this_cpu_dec(*dev->pcpu_refcnt); 4127 #else 4128 refcount_dec(&dev->dev_refcnt); 4129 #endif 4130 } 4131 } 4132 4133 static inline void __dev_hold(struct net_device *dev) 4134 { 4135 if (dev) { 4136 #ifdef CONFIG_PCPU_DEV_REFCNT 4137 this_cpu_inc(*dev->pcpu_refcnt); 4138 #else 4139 refcount_inc(&dev->dev_refcnt); 4140 #endif 4141 } 4142 } 4143 4144 static inline void __netdev_tracker_alloc(struct net_device *dev, 4145 netdevice_tracker *tracker, 4146 gfp_t gfp) 4147 { 4148 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4149 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 4150 #endif 4151 } 4152 4153 /* netdev_tracker_alloc() can upgrade a prior untracked reference 4154 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4155 */ 4156 static inline void netdev_tracker_alloc(struct net_device *dev, 4157 netdevice_tracker *tracker, gfp_t gfp) 4158 { 4159 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4160 refcount_dec(&dev->refcnt_tracker.no_tracker); 4161 __netdev_tracker_alloc(dev, tracker, gfp); 4162 #endif 4163 } 4164 4165 static inline void netdev_tracker_free(struct net_device *dev, 4166 netdevice_tracker *tracker) 4167 { 4168 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4169 ref_tracker_free(&dev->refcnt_tracker, tracker); 4170 #endif 4171 } 4172 4173 static inline void netdev_hold(struct net_device *dev, 4174 netdevice_tracker *tracker, gfp_t gfp) 4175 { 4176 if (dev) { 4177 __dev_hold(dev); 4178 __netdev_tracker_alloc(dev, tracker, gfp); 4179 } 4180 } 4181 4182 static inline void netdev_put(struct net_device *dev, 4183 netdevice_tracker *tracker) 4184 { 4185 if (dev) { 4186 netdev_tracker_free(dev, tracker); 4187 __dev_put(dev); 4188 } 4189 } 4190 4191 /** 4192 * dev_hold - get reference to device 4193 * @dev: network device 4194 * 4195 * Hold reference to device to keep it from being freed. 4196 * Try using netdev_hold() instead. 4197 */ 4198 static inline void dev_hold(struct net_device *dev) 4199 { 4200 netdev_hold(dev, NULL, GFP_ATOMIC); 4201 } 4202 4203 /** 4204 * dev_put - release reference to device 4205 * @dev: network device 4206 * 4207 * Release reference to device to allow it to be freed. 4208 * Try using netdev_put() instead. 4209 */ 4210 static inline void dev_put(struct net_device *dev) 4211 { 4212 netdev_put(dev, NULL); 4213 } 4214 4215 static inline void netdev_ref_replace(struct net_device *odev, 4216 struct net_device *ndev, 4217 netdevice_tracker *tracker, 4218 gfp_t gfp) 4219 { 4220 if (odev) 4221 netdev_tracker_free(odev, tracker); 4222 4223 __dev_hold(ndev); 4224 __dev_put(odev); 4225 4226 if (ndev) 4227 __netdev_tracker_alloc(ndev, tracker, gfp); 4228 } 4229 4230 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4231 * and _off may be called from IRQ context, but it is caller 4232 * who is responsible for serialization of these calls. 4233 * 4234 * The name carrier is inappropriate, these functions should really be 4235 * called netif_lowerlayer_*() because they represent the state of any 4236 * kind of lower layer not just hardware media. 4237 */ 4238 void linkwatch_fire_event(struct net_device *dev); 4239 4240 /** 4241 * linkwatch_sync_dev - sync linkwatch for the given device 4242 * @dev: network device to sync linkwatch for 4243 * 4244 * Sync linkwatch for the given device, removing it from the 4245 * pending work list (if queued). 4246 */ 4247 void linkwatch_sync_dev(struct net_device *dev); 4248 4249 /** 4250 * netif_carrier_ok - test if carrier present 4251 * @dev: network device 4252 * 4253 * Check if carrier is present on device 4254 */ 4255 static inline bool netif_carrier_ok(const struct net_device *dev) 4256 { 4257 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4258 } 4259 4260 unsigned long dev_trans_start(struct net_device *dev); 4261 4262 void __netdev_watchdog_up(struct net_device *dev); 4263 4264 void netif_carrier_on(struct net_device *dev); 4265 void netif_carrier_off(struct net_device *dev); 4266 void netif_carrier_event(struct net_device *dev); 4267 4268 /** 4269 * netif_dormant_on - mark device as dormant. 4270 * @dev: network device 4271 * 4272 * Mark device as dormant (as per RFC2863). 4273 * 4274 * The dormant state indicates that the relevant interface is not 4275 * actually in a condition to pass packets (i.e., it is not 'up') but is 4276 * in a "pending" state, waiting for some external event. For "on- 4277 * demand" interfaces, this new state identifies the situation where the 4278 * interface is waiting for events to place it in the up state. 4279 */ 4280 static inline void netif_dormant_on(struct net_device *dev) 4281 { 4282 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4283 linkwatch_fire_event(dev); 4284 } 4285 4286 /** 4287 * netif_dormant_off - set device as not dormant. 4288 * @dev: network device 4289 * 4290 * Device is not in dormant state. 4291 */ 4292 static inline void netif_dormant_off(struct net_device *dev) 4293 { 4294 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4295 linkwatch_fire_event(dev); 4296 } 4297 4298 /** 4299 * netif_dormant - test if device is dormant 4300 * @dev: network device 4301 * 4302 * Check if device is dormant. 4303 */ 4304 static inline bool netif_dormant(const struct net_device *dev) 4305 { 4306 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4307 } 4308 4309 4310 /** 4311 * netif_testing_on - mark device as under test. 4312 * @dev: network device 4313 * 4314 * Mark device as under test (as per RFC2863). 4315 * 4316 * The testing state indicates that some test(s) must be performed on 4317 * the interface. After completion, of the test, the interface state 4318 * will change to up, dormant, or down, as appropriate. 4319 */ 4320 static inline void netif_testing_on(struct net_device *dev) 4321 { 4322 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4323 linkwatch_fire_event(dev); 4324 } 4325 4326 /** 4327 * netif_testing_off - set device as not under test. 4328 * @dev: network device 4329 * 4330 * Device is not in testing state. 4331 */ 4332 static inline void netif_testing_off(struct net_device *dev) 4333 { 4334 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4335 linkwatch_fire_event(dev); 4336 } 4337 4338 /** 4339 * netif_testing - test if device is under test 4340 * @dev: network device 4341 * 4342 * Check if device is under test 4343 */ 4344 static inline bool netif_testing(const struct net_device *dev) 4345 { 4346 return test_bit(__LINK_STATE_TESTING, &dev->state); 4347 } 4348 4349 4350 /** 4351 * netif_oper_up - test if device is operational 4352 * @dev: network device 4353 * 4354 * Check if carrier is operational 4355 */ 4356 static inline bool netif_oper_up(const struct net_device *dev) 4357 { 4358 return (dev->operstate == IF_OPER_UP || 4359 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4360 } 4361 4362 /** 4363 * netif_device_present - is device available or removed 4364 * @dev: network device 4365 * 4366 * Check if device has not been removed from system. 4367 */ 4368 static inline bool netif_device_present(const struct net_device *dev) 4369 { 4370 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4371 } 4372 4373 void netif_device_detach(struct net_device *dev); 4374 4375 void netif_device_attach(struct net_device *dev); 4376 4377 /* 4378 * Network interface message level settings 4379 */ 4380 4381 enum { 4382 NETIF_MSG_DRV_BIT, 4383 NETIF_MSG_PROBE_BIT, 4384 NETIF_MSG_LINK_BIT, 4385 NETIF_MSG_TIMER_BIT, 4386 NETIF_MSG_IFDOWN_BIT, 4387 NETIF_MSG_IFUP_BIT, 4388 NETIF_MSG_RX_ERR_BIT, 4389 NETIF_MSG_TX_ERR_BIT, 4390 NETIF_MSG_TX_QUEUED_BIT, 4391 NETIF_MSG_INTR_BIT, 4392 NETIF_MSG_TX_DONE_BIT, 4393 NETIF_MSG_RX_STATUS_BIT, 4394 NETIF_MSG_PKTDATA_BIT, 4395 NETIF_MSG_HW_BIT, 4396 NETIF_MSG_WOL_BIT, 4397 4398 /* When you add a new bit above, update netif_msg_class_names array 4399 * in net/ethtool/common.c 4400 */ 4401 NETIF_MSG_CLASS_COUNT, 4402 }; 4403 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4404 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4405 4406 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4407 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4408 4409 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4410 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4411 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4412 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4413 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4414 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4415 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4416 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4417 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4418 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4419 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4420 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4421 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4422 #define NETIF_MSG_HW __NETIF_MSG(HW) 4423 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4424 4425 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4426 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4427 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4428 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4429 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4430 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4431 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4432 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4433 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4434 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4435 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4436 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4437 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4438 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4439 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4440 4441 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4442 { 4443 /* use default */ 4444 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4445 return default_msg_enable_bits; 4446 if (debug_value == 0) /* no output */ 4447 return 0; 4448 /* set low N bits */ 4449 return (1U << debug_value) - 1; 4450 } 4451 4452 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4453 { 4454 spin_lock(&txq->_xmit_lock); 4455 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4456 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4457 } 4458 4459 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4460 { 4461 __acquire(&txq->_xmit_lock); 4462 return true; 4463 } 4464 4465 static inline void __netif_tx_release(struct netdev_queue *txq) 4466 { 4467 __release(&txq->_xmit_lock); 4468 } 4469 4470 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4471 { 4472 spin_lock_bh(&txq->_xmit_lock); 4473 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4474 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4475 } 4476 4477 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4478 { 4479 bool ok = spin_trylock(&txq->_xmit_lock); 4480 4481 if (likely(ok)) { 4482 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4483 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4484 } 4485 return ok; 4486 } 4487 4488 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4489 { 4490 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4491 WRITE_ONCE(txq->xmit_lock_owner, -1); 4492 spin_unlock(&txq->_xmit_lock); 4493 } 4494 4495 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4496 { 4497 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4498 WRITE_ONCE(txq->xmit_lock_owner, -1); 4499 spin_unlock_bh(&txq->_xmit_lock); 4500 } 4501 4502 /* 4503 * txq->trans_start can be read locklessly from dev_watchdog() 4504 */ 4505 static inline void txq_trans_update(struct netdev_queue *txq) 4506 { 4507 if (txq->xmit_lock_owner != -1) 4508 WRITE_ONCE(txq->trans_start, jiffies); 4509 } 4510 4511 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4512 { 4513 unsigned long now = jiffies; 4514 4515 if (READ_ONCE(txq->trans_start) != now) 4516 WRITE_ONCE(txq->trans_start, now); 4517 } 4518 4519 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4520 static inline void netif_trans_update(struct net_device *dev) 4521 { 4522 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4523 4524 txq_trans_cond_update(txq); 4525 } 4526 4527 /** 4528 * netif_tx_lock - grab network device transmit lock 4529 * @dev: network device 4530 * 4531 * Get network device transmit lock 4532 */ 4533 void netif_tx_lock(struct net_device *dev); 4534 4535 static inline void netif_tx_lock_bh(struct net_device *dev) 4536 { 4537 local_bh_disable(); 4538 netif_tx_lock(dev); 4539 } 4540 4541 void netif_tx_unlock(struct net_device *dev); 4542 4543 static inline void netif_tx_unlock_bh(struct net_device *dev) 4544 { 4545 netif_tx_unlock(dev); 4546 local_bh_enable(); 4547 } 4548 4549 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4550 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4551 __netif_tx_lock(txq, cpu); \ 4552 } else { \ 4553 __netif_tx_acquire(txq); \ 4554 } \ 4555 } 4556 4557 #define HARD_TX_TRYLOCK(dev, txq) \ 4558 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4559 __netif_tx_trylock(txq) : \ 4560 __netif_tx_acquire(txq)) 4561 4562 #define HARD_TX_UNLOCK(dev, txq) { \ 4563 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4564 __netif_tx_unlock(txq); \ 4565 } else { \ 4566 __netif_tx_release(txq); \ 4567 } \ 4568 } 4569 4570 static inline void netif_tx_disable(struct net_device *dev) 4571 { 4572 unsigned int i; 4573 int cpu; 4574 4575 local_bh_disable(); 4576 cpu = smp_processor_id(); 4577 spin_lock(&dev->tx_global_lock); 4578 for (i = 0; i < dev->num_tx_queues; i++) { 4579 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4580 4581 __netif_tx_lock(txq, cpu); 4582 netif_tx_stop_queue(txq); 4583 __netif_tx_unlock(txq); 4584 } 4585 spin_unlock(&dev->tx_global_lock); 4586 local_bh_enable(); 4587 } 4588 4589 static inline void netif_addr_lock(struct net_device *dev) 4590 { 4591 unsigned char nest_level = 0; 4592 4593 #ifdef CONFIG_LOCKDEP 4594 nest_level = dev->nested_level; 4595 #endif 4596 spin_lock_nested(&dev->addr_list_lock, nest_level); 4597 } 4598 4599 static inline void netif_addr_lock_bh(struct net_device *dev) 4600 { 4601 unsigned char nest_level = 0; 4602 4603 #ifdef CONFIG_LOCKDEP 4604 nest_level = dev->nested_level; 4605 #endif 4606 local_bh_disable(); 4607 spin_lock_nested(&dev->addr_list_lock, nest_level); 4608 } 4609 4610 static inline void netif_addr_unlock(struct net_device *dev) 4611 { 4612 spin_unlock(&dev->addr_list_lock); 4613 } 4614 4615 static inline void netif_addr_unlock_bh(struct net_device *dev) 4616 { 4617 spin_unlock_bh(&dev->addr_list_lock); 4618 } 4619 4620 /* 4621 * dev_addrs walker. Should be used only for read access. Call with 4622 * rcu_read_lock held. 4623 */ 4624 #define for_each_dev_addr(dev, ha) \ 4625 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4626 4627 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4628 4629 void ether_setup(struct net_device *dev); 4630 4631 /* Support for loadable net-drivers */ 4632 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4633 unsigned char name_assign_type, 4634 void (*setup)(struct net_device *), 4635 unsigned int txqs, unsigned int rxqs); 4636 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4637 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4638 4639 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4640 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4641 count) 4642 4643 int register_netdev(struct net_device *dev); 4644 void unregister_netdev(struct net_device *dev); 4645 4646 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4647 4648 /* General hardware address lists handling functions */ 4649 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4650 struct netdev_hw_addr_list *from_list, int addr_len); 4651 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4652 struct netdev_hw_addr_list *from_list, int addr_len); 4653 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4654 struct net_device *dev, 4655 int (*sync)(struct net_device *, const unsigned char *), 4656 int (*unsync)(struct net_device *, 4657 const unsigned char *)); 4658 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4659 struct net_device *dev, 4660 int (*sync)(struct net_device *, 4661 const unsigned char *, int), 4662 int (*unsync)(struct net_device *, 4663 const unsigned char *, int)); 4664 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4665 struct net_device *dev, 4666 int (*unsync)(struct net_device *, 4667 const unsigned char *, int)); 4668 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4669 struct net_device *dev, 4670 int (*unsync)(struct net_device *, 4671 const unsigned char *)); 4672 void __hw_addr_init(struct netdev_hw_addr_list *list); 4673 4674 /* Functions used for device addresses handling */ 4675 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4676 const void *addr, size_t len); 4677 4678 static inline void 4679 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4680 { 4681 dev_addr_mod(dev, 0, addr, len); 4682 } 4683 4684 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4685 { 4686 __dev_addr_set(dev, addr, dev->addr_len); 4687 } 4688 4689 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4690 unsigned char addr_type); 4691 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4692 unsigned char addr_type); 4693 4694 /* Functions used for unicast addresses handling */ 4695 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4696 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4697 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4698 int dev_uc_sync(struct net_device *to, struct net_device *from); 4699 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4700 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4701 void dev_uc_flush(struct net_device *dev); 4702 void dev_uc_init(struct net_device *dev); 4703 4704 /** 4705 * __dev_uc_sync - Synchonize device's unicast list 4706 * @dev: device to sync 4707 * @sync: function to call if address should be added 4708 * @unsync: function to call if address should be removed 4709 * 4710 * Add newly added addresses to the interface, and release 4711 * addresses that have been deleted. 4712 */ 4713 static inline int __dev_uc_sync(struct net_device *dev, 4714 int (*sync)(struct net_device *, 4715 const unsigned char *), 4716 int (*unsync)(struct net_device *, 4717 const unsigned char *)) 4718 { 4719 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4720 } 4721 4722 /** 4723 * __dev_uc_unsync - Remove synchronized addresses from device 4724 * @dev: device to sync 4725 * @unsync: function to call if address should be removed 4726 * 4727 * Remove all addresses that were added to the device by dev_uc_sync(). 4728 */ 4729 static inline void __dev_uc_unsync(struct net_device *dev, 4730 int (*unsync)(struct net_device *, 4731 const unsigned char *)) 4732 { 4733 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4734 } 4735 4736 /* Functions used for multicast addresses handling */ 4737 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4738 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4739 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4740 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4741 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4742 int dev_mc_sync(struct net_device *to, struct net_device *from); 4743 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4744 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4745 void dev_mc_flush(struct net_device *dev); 4746 void dev_mc_init(struct net_device *dev); 4747 4748 /** 4749 * __dev_mc_sync - Synchonize device's multicast list 4750 * @dev: device to sync 4751 * @sync: function to call if address should be added 4752 * @unsync: function to call if address should be removed 4753 * 4754 * Add newly added addresses to the interface, and release 4755 * addresses that have been deleted. 4756 */ 4757 static inline int __dev_mc_sync(struct net_device *dev, 4758 int (*sync)(struct net_device *, 4759 const unsigned char *), 4760 int (*unsync)(struct net_device *, 4761 const unsigned char *)) 4762 { 4763 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4764 } 4765 4766 /** 4767 * __dev_mc_unsync - Remove synchronized addresses from device 4768 * @dev: device to sync 4769 * @unsync: function to call if address should be removed 4770 * 4771 * Remove all addresses that were added to the device by dev_mc_sync(). 4772 */ 4773 static inline void __dev_mc_unsync(struct net_device *dev, 4774 int (*unsync)(struct net_device *, 4775 const unsigned char *)) 4776 { 4777 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4778 } 4779 4780 /* Functions used for secondary unicast and multicast support */ 4781 void dev_set_rx_mode(struct net_device *dev); 4782 int dev_set_promiscuity(struct net_device *dev, int inc); 4783 int dev_set_allmulti(struct net_device *dev, int inc); 4784 void netdev_state_change(struct net_device *dev); 4785 void __netdev_notify_peers(struct net_device *dev); 4786 void netdev_notify_peers(struct net_device *dev); 4787 void netdev_features_change(struct net_device *dev); 4788 /* Load a device via the kmod */ 4789 void dev_load(struct net *net, const char *name); 4790 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4791 struct rtnl_link_stats64 *storage); 4792 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4793 const struct net_device_stats *netdev_stats); 4794 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4795 const struct pcpu_sw_netstats __percpu *netstats); 4796 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4797 4798 extern int netdev_max_backlog; 4799 extern int dev_rx_weight; 4800 extern int dev_tx_weight; 4801 extern int gro_normal_batch; 4802 4803 enum { 4804 NESTED_SYNC_IMM_BIT, 4805 NESTED_SYNC_TODO_BIT, 4806 }; 4807 4808 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4809 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4810 4811 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4812 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4813 4814 struct netdev_nested_priv { 4815 unsigned char flags; 4816 void *data; 4817 }; 4818 4819 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4820 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4821 struct list_head **iter); 4822 4823 /* iterate through upper list, must be called under RCU read lock */ 4824 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4825 for (iter = &(dev)->adj_list.upper, \ 4826 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4827 updev; \ 4828 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4829 4830 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4831 int (*fn)(struct net_device *upper_dev, 4832 struct netdev_nested_priv *priv), 4833 struct netdev_nested_priv *priv); 4834 4835 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4836 struct net_device *upper_dev); 4837 4838 bool netdev_has_any_upper_dev(struct net_device *dev); 4839 4840 void *netdev_lower_get_next_private(struct net_device *dev, 4841 struct list_head **iter); 4842 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4843 struct list_head **iter); 4844 4845 #define netdev_for_each_lower_private(dev, priv, iter) \ 4846 for (iter = (dev)->adj_list.lower.next, \ 4847 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4848 priv; \ 4849 priv = netdev_lower_get_next_private(dev, &(iter))) 4850 4851 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4852 for (iter = &(dev)->adj_list.lower, \ 4853 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4854 priv; \ 4855 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4856 4857 void *netdev_lower_get_next(struct net_device *dev, 4858 struct list_head **iter); 4859 4860 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4861 for (iter = (dev)->adj_list.lower.next, \ 4862 ldev = netdev_lower_get_next(dev, &(iter)); \ 4863 ldev; \ 4864 ldev = netdev_lower_get_next(dev, &(iter))) 4865 4866 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4867 struct list_head **iter); 4868 int netdev_walk_all_lower_dev(struct net_device *dev, 4869 int (*fn)(struct net_device *lower_dev, 4870 struct netdev_nested_priv *priv), 4871 struct netdev_nested_priv *priv); 4872 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4873 int (*fn)(struct net_device *lower_dev, 4874 struct netdev_nested_priv *priv), 4875 struct netdev_nested_priv *priv); 4876 4877 void *netdev_adjacent_get_private(struct list_head *adj_list); 4878 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4879 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4880 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4881 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4882 struct netlink_ext_ack *extack); 4883 int netdev_master_upper_dev_link(struct net_device *dev, 4884 struct net_device *upper_dev, 4885 void *upper_priv, void *upper_info, 4886 struct netlink_ext_ack *extack); 4887 void netdev_upper_dev_unlink(struct net_device *dev, 4888 struct net_device *upper_dev); 4889 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4890 struct net_device *new_dev, 4891 struct net_device *dev, 4892 struct netlink_ext_ack *extack); 4893 void netdev_adjacent_change_commit(struct net_device *old_dev, 4894 struct net_device *new_dev, 4895 struct net_device *dev); 4896 void netdev_adjacent_change_abort(struct net_device *old_dev, 4897 struct net_device *new_dev, 4898 struct net_device *dev); 4899 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4900 void *netdev_lower_dev_get_private(struct net_device *dev, 4901 struct net_device *lower_dev); 4902 void netdev_lower_state_changed(struct net_device *lower_dev, 4903 void *lower_state_info); 4904 4905 /* RSS keys are 40 or 52 bytes long */ 4906 #define NETDEV_RSS_KEY_LEN 52 4907 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4908 void netdev_rss_key_fill(void *buffer, size_t len); 4909 4910 int skb_checksum_help(struct sk_buff *skb); 4911 int skb_crc32c_csum_help(struct sk_buff *skb); 4912 int skb_csum_hwoffload_help(struct sk_buff *skb, 4913 const netdev_features_t features); 4914 4915 struct netdev_bonding_info { 4916 ifslave slave; 4917 ifbond master; 4918 }; 4919 4920 struct netdev_notifier_bonding_info { 4921 struct netdev_notifier_info info; /* must be first */ 4922 struct netdev_bonding_info bonding_info; 4923 }; 4924 4925 void netdev_bonding_info_change(struct net_device *dev, 4926 struct netdev_bonding_info *bonding_info); 4927 4928 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4929 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4930 #else 4931 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4932 const void *data) 4933 { 4934 } 4935 #endif 4936 4937 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4938 4939 static inline bool can_checksum_protocol(netdev_features_t features, 4940 __be16 protocol) 4941 { 4942 if (protocol == htons(ETH_P_FCOE)) 4943 return !!(features & NETIF_F_FCOE_CRC); 4944 4945 /* Assume this is an IP checksum (not SCTP CRC) */ 4946 4947 if (features & NETIF_F_HW_CSUM) { 4948 /* Can checksum everything */ 4949 return true; 4950 } 4951 4952 switch (protocol) { 4953 case htons(ETH_P_IP): 4954 return !!(features & NETIF_F_IP_CSUM); 4955 case htons(ETH_P_IPV6): 4956 return !!(features & NETIF_F_IPV6_CSUM); 4957 default: 4958 return false; 4959 } 4960 } 4961 4962 #ifdef CONFIG_BUG 4963 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4964 #else 4965 static inline void netdev_rx_csum_fault(struct net_device *dev, 4966 struct sk_buff *skb) 4967 { 4968 } 4969 #endif 4970 /* rx skb timestamps */ 4971 void net_enable_timestamp(void); 4972 void net_disable_timestamp(void); 4973 4974 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 4975 const struct skb_shared_hwtstamps *hwtstamps, 4976 bool cycles) 4977 { 4978 const struct net_device_ops *ops = dev->netdev_ops; 4979 4980 if (ops->ndo_get_tstamp) 4981 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 4982 4983 return hwtstamps->hwtstamp; 4984 } 4985 4986 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4987 struct sk_buff *skb, struct net_device *dev, 4988 bool more) 4989 { 4990 __this_cpu_write(softnet_data.xmit.more, more); 4991 return ops->ndo_start_xmit(skb, dev); 4992 } 4993 4994 static inline bool netdev_xmit_more(void) 4995 { 4996 return __this_cpu_read(softnet_data.xmit.more); 4997 } 4998 4999 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 5000 struct netdev_queue *txq, bool more) 5001 { 5002 const struct net_device_ops *ops = dev->netdev_ops; 5003 netdev_tx_t rc; 5004 5005 rc = __netdev_start_xmit(ops, skb, dev, more); 5006 if (rc == NETDEV_TX_OK) 5007 txq_trans_update(txq); 5008 5009 return rc; 5010 } 5011 5012 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 5013 const void *ns); 5014 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 5015 const void *ns); 5016 5017 extern const struct kobj_ns_type_operations net_ns_type_operations; 5018 5019 const char *netdev_drivername(const struct net_device *dev); 5020 5021 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 5022 netdev_features_t f2) 5023 { 5024 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 5025 if (f1 & NETIF_F_HW_CSUM) 5026 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5027 else 5028 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5029 } 5030 5031 return f1 & f2; 5032 } 5033 5034 static inline netdev_features_t netdev_get_wanted_features( 5035 struct net_device *dev) 5036 { 5037 return (dev->features & ~dev->hw_features) | dev->wanted_features; 5038 } 5039 netdev_features_t netdev_increment_features(netdev_features_t all, 5040 netdev_features_t one, netdev_features_t mask); 5041 5042 /* Allow TSO being used on stacked device : 5043 * Performing the GSO segmentation before last device 5044 * is a performance improvement. 5045 */ 5046 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 5047 netdev_features_t mask) 5048 { 5049 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 5050 } 5051 5052 int __netdev_update_features(struct net_device *dev); 5053 void netdev_update_features(struct net_device *dev); 5054 void netdev_change_features(struct net_device *dev); 5055 5056 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 5057 struct net_device *dev); 5058 5059 netdev_features_t passthru_features_check(struct sk_buff *skb, 5060 struct net_device *dev, 5061 netdev_features_t features); 5062 netdev_features_t netif_skb_features(struct sk_buff *skb); 5063 void skb_warn_bad_offload(const struct sk_buff *skb); 5064 5065 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 5066 { 5067 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 5068 5069 /* check flags correspondence */ 5070 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 5071 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 5072 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 5073 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 5074 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 5075 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 5076 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 5077 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 5078 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 5079 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 5080 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 5081 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 5082 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 5083 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 5084 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 5085 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 5086 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 5087 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 5088 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 5089 5090 return (features & feature) == feature; 5091 } 5092 5093 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 5094 { 5095 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 5096 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 5097 } 5098 5099 static inline bool netif_needs_gso(struct sk_buff *skb, 5100 netdev_features_t features) 5101 { 5102 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 5103 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 5104 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 5105 } 5106 5107 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 5108 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 5109 void netif_inherit_tso_max(struct net_device *to, 5110 const struct net_device *from); 5111 5112 static inline bool netif_is_macsec(const struct net_device *dev) 5113 { 5114 return dev->priv_flags & IFF_MACSEC; 5115 } 5116 5117 static inline bool netif_is_macvlan(const struct net_device *dev) 5118 { 5119 return dev->priv_flags & IFF_MACVLAN; 5120 } 5121 5122 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5123 { 5124 return dev->priv_flags & IFF_MACVLAN_PORT; 5125 } 5126 5127 static inline bool netif_is_bond_master(const struct net_device *dev) 5128 { 5129 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5130 } 5131 5132 static inline bool netif_is_bond_slave(const struct net_device *dev) 5133 { 5134 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5135 } 5136 5137 static inline bool netif_supports_nofcs(struct net_device *dev) 5138 { 5139 return dev->priv_flags & IFF_SUPP_NOFCS; 5140 } 5141 5142 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5143 { 5144 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5145 } 5146 5147 static inline bool netif_is_l3_master(const struct net_device *dev) 5148 { 5149 return dev->priv_flags & IFF_L3MDEV_MASTER; 5150 } 5151 5152 static inline bool netif_is_l3_slave(const struct net_device *dev) 5153 { 5154 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5155 } 5156 5157 static inline int dev_sdif(const struct net_device *dev) 5158 { 5159 #ifdef CONFIG_NET_L3_MASTER_DEV 5160 if (netif_is_l3_slave(dev)) 5161 return dev->ifindex; 5162 #endif 5163 return 0; 5164 } 5165 5166 static inline bool netif_is_bridge_master(const struct net_device *dev) 5167 { 5168 return dev->priv_flags & IFF_EBRIDGE; 5169 } 5170 5171 static inline bool netif_is_bridge_port(const struct net_device *dev) 5172 { 5173 return dev->priv_flags & IFF_BRIDGE_PORT; 5174 } 5175 5176 static inline bool netif_is_ovs_master(const struct net_device *dev) 5177 { 5178 return dev->priv_flags & IFF_OPENVSWITCH; 5179 } 5180 5181 static inline bool netif_is_ovs_port(const struct net_device *dev) 5182 { 5183 return dev->priv_flags & IFF_OVS_DATAPATH; 5184 } 5185 5186 static inline bool netif_is_any_bridge_master(const struct net_device *dev) 5187 { 5188 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev); 5189 } 5190 5191 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5192 { 5193 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5194 } 5195 5196 static inline bool netif_is_team_master(const struct net_device *dev) 5197 { 5198 return dev->priv_flags & IFF_TEAM; 5199 } 5200 5201 static inline bool netif_is_team_port(const struct net_device *dev) 5202 { 5203 return dev->priv_flags & IFF_TEAM_PORT; 5204 } 5205 5206 static inline bool netif_is_lag_master(const struct net_device *dev) 5207 { 5208 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5209 } 5210 5211 static inline bool netif_is_lag_port(const struct net_device *dev) 5212 { 5213 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5214 } 5215 5216 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5217 { 5218 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5219 } 5220 5221 static inline bool netif_is_failover(const struct net_device *dev) 5222 { 5223 return dev->priv_flags & IFF_FAILOVER; 5224 } 5225 5226 static inline bool netif_is_failover_slave(const struct net_device *dev) 5227 { 5228 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5229 } 5230 5231 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5232 static inline void netif_keep_dst(struct net_device *dev) 5233 { 5234 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5235 } 5236 5237 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5238 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5239 { 5240 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5241 return netif_is_macsec(dev); 5242 } 5243 5244 extern struct pernet_operations __net_initdata loopback_net_ops; 5245 5246 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5247 5248 /* netdev_printk helpers, similar to dev_printk */ 5249 5250 static inline const char *netdev_name(const struct net_device *dev) 5251 { 5252 if (!dev->name[0] || strchr(dev->name, '%')) 5253 return "(unnamed net_device)"; 5254 return dev->name; 5255 } 5256 5257 static inline const char *netdev_reg_state(const struct net_device *dev) 5258 { 5259 switch (dev->reg_state) { 5260 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5261 case NETREG_REGISTERED: return ""; 5262 case NETREG_UNREGISTERING: return " (unregistering)"; 5263 case NETREG_UNREGISTERED: return " (unregistered)"; 5264 case NETREG_RELEASED: return " (released)"; 5265 case NETREG_DUMMY: return " (dummy)"; 5266 } 5267 5268 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5269 return " (unknown)"; 5270 } 5271 5272 #define MODULE_ALIAS_NETDEV(device) \ 5273 MODULE_ALIAS("netdev-" device) 5274 5275 /* 5276 * netdev_WARN() acts like dev_printk(), but with the key difference 5277 * of using a WARN/WARN_ON to get the message out, including the 5278 * file/line information and a backtrace. 5279 */ 5280 #define netdev_WARN(dev, format, args...) \ 5281 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5282 netdev_reg_state(dev), ##args) 5283 5284 #define netdev_WARN_ONCE(dev, format, args...) \ 5285 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5286 netdev_reg_state(dev), ##args) 5287 5288 /* 5289 * The list of packet types we will receive (as opposed to discard) 5290 * and the routines to invoke. 5291 * 5292 * Why 16. Because with 16 the only overlap we get on a hash of the 5293 * low nibble of the protocol value is RARP/SNAP/X.25. 5294 * 5295 * 0800 IP 5296 * 0001 802.3 5297 * 0002 AX.25 5298 * 0004 802.2 5299 * 8035 RARP 5300 * 0005 SNAP 5301 * 0805 X.25 5302 * 0806 ARP 5303 * 8137 IPX 5304 * 0009 Localtalk 5305 * 86DD IPv6 5306 */ 5307 #define PTYPE_HASH_SIZE (16) 5308 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5309 5310 extern struct list_head ptype_all __read_mostly; 5311 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5312 5313 extern struct net_device *blackhole_netdev; 5314 5315 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */ 5316 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD) 5317 #define DEV_STATS_ADD(DEV, FIELD, VAL) \ 5318 atomic_long_add((VAL), &(DEV)->stats.__##FIELD) 5319 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD) 5320 5321 #endif /* _LINUX_NETDEVICE_H */ 5322