1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <[email protected]> 13 * Corey Minyard <[email protected]> 14 * Donald J. Becker, <[email protected]> 15 * Alan Cox, <[email protected]> 16 * Bjorn Ekwall. <[email protected]> 17 * Pekka Riikonen <[email protected]> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 #include <asm/local.h> 32 33 #include <linux/percpu.h> 34 #include <linux/rculist.h> 35 #include <linux/workqueue.h> 36 #include <linux/dynamic_queue_limits.h> 37 38 #include <net/net_namespace.h> 39 #ifdef CONFIG_DCB 40 #include <net/dcbnl.h> 41 #endif 42 #include <net/netprio_cgroup.h> 43 #include <linux/netdev_features.h> 44 #include <linux/neighbour.h> 45 #include <linux/netdevice_xmit.h> 46 #include <uapi/linux/netdevice.h> 47 #include <uapi/linux/if_bonding.h> 48 #include <uapi/linux/pkt_cls.h> 49 #include <uapi/linux/netdev.h> 50 #include <linux/hashtable.h> 51 #include <linux/rbtree.h> 52 #include <net/net_trackers.h> 53 #include <net/net_debug.h> 54 #include <net/dropreason-core.h> 55 #include <net/neighbour_tables.h> 56 57 struct netpoll_info; 58 struct device; 59 struct ethtool_ops; 60 struct kernel_hwtstamp_config; 61 struct phy_device; 62 struct dsa_port; 63 struct ip_tunnel_parm_kern; 64 struct macsec_context; 65 struct macsec_ops; 66 struct netdev_config; 67 struct netdev_name_node; 68 struct sd_flow_limit; 69 struct sfp_bus; 70 /* 802.11 specific */ 71 struct wireless_dev; 72 /* 802.15.4 specific */ 73 struct wpan_dev; 74 struct mpls_dev; 75 /* UDP Tunnel offloads */ 76 struct udp_tunnel_info; 77 struct udp_tunnel_nic_info; 78 struct udp_tunnel_nic; 79 struct bpf_prog; 80 struct xdp_buff; 81 struct xdp_frame; 82 struct xdp_metadata_ops; 83 struct xdp_md; 84 struct ethtool_netdev_state; 85 struct phy_link_topology; 86 struct hwtstamp_provider; 87 88 typedef u32 xdp_features_t; 89 90 void synchronize_net(void); 91 void netdev_set_default_ethtool_ops(struct net_device *dev, 92 const struct ethtool_ops *ops); 93 void netdev_sw_irq_coalesce_default_on(struct net_device *dev); 94 95 /* Backlog congestion levels */ 96 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 97 #define NET_RX_DROP 1 /* packet dropped */ 98 99 #define MAX_NEST_DEV 8 100 101 /* 102 * Transmit return codes: transmit return codes originate from three different 103 * namespaces: 104 * 105 * - qdisc return codes 106 * - driver transmit return codes 107 * - errno values 108 * 109 * Drivers are allowed to return any one of those in their hard_start_xmit() 110 * function. Real network devices commonly used with qdiscs should only return 111 * the driver transmit return codes though - when qdiscs are used, the actual 112 * transmission happens asynchronously, so the value is not propagated to 113 * higher layers. Virtual network devices transmit synchronously; in this case 114 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 115 * others are propagated to higher layers. 116 */ 117 118 /* qdisc ->enqueue() return codes. */ 119 #define NET_XMIT_SUCCESS 0x00 120 #define NET_XMIT_DROP 0x01 /* skb dropped */ 121 #define NET_XMIT_CN 0x02 /* congestion notification */ 122 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 123 124 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 125 * indicates that the device will soon be dropping packets, or already drops 126 * some packets of the same priority; prompting us to send less aggressively. */ 127 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 128 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 129 130 /* Driver transmit return codes */ 131 #define NETDEV_TX_MASK 0xf0 132 133 enum netdev_tx { 134 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 135 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 136 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 137 }; 138 typedef enum netdev_tx netdev_tx_t; 139 140 /* 141 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 142 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 143 */ 144 static inline bool dev_xmit_complete(int rc) 145 { 146 /* 147 * Positive cases with an skb consumed by a driver: 148 * - successful transmission (rc == NETDEV_TX_OK) 149 * - error while transmitting (rc < 0) 150 * - error while queueing to a different device (rc & NET_XMIT_MASK) 151 */ 152 if (likely(rc < NET_XMIT_MASK)) 153 return true; 154 155 return false; 156 } 157 158 /* 159 * Compute the worst-case header length according to the protocols 160 * used. 161 */ 162 163 #if defined(CONFIG_HYPERV_NET) 164 # define LL_MAX_HEADER 128 165 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 166 # if defined(CONFIG_MAC80211_MESH) 167 # define LL_MAX_HEADER 128 168 # else 169 # define LL_MAX_HEADER 96 170 # endif 171 #else 172 # define LL_MAX_HEADER 32 173 #endif 174 175 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 176 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 177 #define MAX_HEADER LL_MAX_HEADER 178 #else 179 #define MAX_HEADER (LL_MAX_HEADER + 48) 180 #endif 181 182 /* 183 * Old network device statistics. Fields are native words 184 * (unsigned long) so they can be read and written atomically. 185 */ 186 187 #define NET_DEV_STAT(FIELD) \ 188 union { \ 189 unsigned long FIELD; \ 190 atomic_long_t __##FIELD; \ 191 } 192 193 struct net_device_stats { 194 NET_DEV_STAT(rx_packets); 195 NET_DEV_STAT(tx_packets); 196 NET_DEV_STAT(rx_bytes); 197 NET_DEV_STAT(tx_bytes); 198 NET_DEV_STAT(rx_errors); 199 NET_DEV_STAT(tx_errors); 200 NET_DEV_STAT(rx_dropped); 201 NET_DEV_STAT(tx_dropped); 202 NET_DEV_STAT(multicast); 203 NET_DEV_STAT(collisions); 204 NET_DEV_STAT(rx_length_errors); 205 NET_DEV_STAT(rx_over_errors); 206 NET_DEV_STAT(rx_crc_errors); 207 NET_DEV_STAT(rx_frame_errors); 208 NET_DEV_STAT(rx_fifo_errors); 209 NET_DEV_STAT(rx_missed_errors); 210 NET_DEV_STAT(tx_aborted_errors); 211 NET_DEV_STAT(tx_carrier_errors); 212 NET_DEV_STAT(tx_fifo_errors); 213 NET_DEV_STAT(tx_heartbeat_errors); 214 NET_DEV_STAT(tx_window_errors); 215 NET_DEV_STAT(rx_compressed); 216 NET_DEV_STAT(tx_compressed); 217 }; 218 #undef NET_DEV_STAT 219 220 /* per-cpu stats, allocated on demand. 221 * Try to fit them in a single cache line, for dev_get_stats() sake. 222 */ 223 struct net_device_core_stats { 224 unsigned long rx_dropped; 225 unsigned long tx_dropped; 226 unsigned long rx_nohandler; 227 unsigned long rx_otherhost_dropped; 228 } __aligned(4 * sizeof(unsigned long)); 229 230 #include <linux/cache.h> 231 #include <linux/skbuff.h> 232 233 struct neighbour; 234 struct neigh_parms; 235 struct sk_buff; 236 237 struct netdev_hw_addr { 238 struct list_head list; 239 struct rb_node node; 240 unsigned char addr[MAX_ADDR_LEN]; 241 unsigned char type; 242 #define NETDEV_HW_ADDR_T_LAN 1 243 #define NETDEV_HW_ADDR_T_SAN 2 244 #define NETDEV_HW_ADDR_T_UNICAST 3 245 #define NETDEV_HW_ADDR_T_MULTICAST 4 246 bool global_use; 247 int sync_cnt; 248 int refcount; 249 int synced; 250 struct rcu_head rcu_head; 251 }; 252 253 struct netdev_hw_addr_list { 254 struct list_head list; 255 int count; 256 257 /* Auxiliary tree for faster lookup on addition and deletion */ 258 struct rb_root tree; 259 }; 260 261 #define netdev_hw_addr_list_count(l) ((l)->count) 262 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 263 #define netdev_hw_addr_list_for_each(ha, l) \ 264 list_for_each_entry(ha, &(l)->list, list) 265 266 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 267 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 268 #define netdev_for_each_uc_addr(ha, dev) \ 269 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 270 #define netdev_for_each_synced_uc_addr(_ha, _dev) \ 271 netdev_for_each_uc_addr((_ha), (_dev)) \ 272 if ((_ha)->sync_cnt) 273 274 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 275 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 276 #define netdev_for_each_mc_addr(ha, dev) \ 277 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 278 #define netdev_for_each_synced_mc_addr(_ha, _dev) \ 279 netdev_for_each_mc_addr((_ha), (_dev)) \ 280 if ((_ha)->sync_cnt) 281 282 struct hh_cache { 283 unsigned int hh_len; 284 seqlock_t hh_lock; 285 286 /* cached hardware header; allow for machine alignment needs. */ 287 #define HH_DATA_MOD 16 288 #define HH_DATA_OFF(__len) \ 289 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 290 #define HH_DATA_ALIGN(__len) \ 291 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 292 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 293 }; 294 295 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 296 * Alternative is: 297 * dev->hard_header_len ? (dev->hard_header_len + 298 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 299 * 300 * We could use other alignment values, but we must maintain the 301 * relationship HH alignment <= LL alignment. 302 */ 303 #define LL_RESERVED_SPACE(dev) \ 304 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \ 305 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 306 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 307 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \ 308 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 309 310 struct header_ops { 311 int (*create) (struct sk_buff *skb, struct net_device *dev, 312 unsigned short type, const void *daddr, 313 const void *saddr, unsigned int len); 314 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 315 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 316 void (*cache_update)(struct hh_cache *hh, 317 const struct net_device *dev, 318 const unsigned char *haddr); 319 bool (*validate)(const char *ll_header, unsigned int len); 320 __be16 (*parse_protocol)(const struct sk_buff *skb); 321 }; 322 323 /* These flag bits are private to the generic network queueing 324 * layer; they may not be explicitly referenced by any other 325 * code. 326 */ 327 328 enum netdev_state_t { 329 __LINK_STATE_START, 330 __LINK_STATE_PRESENT, 331 __LINK_STATE_NOCARRIER, 332 __LINK_STATE_LINKWATCH_PENDING, 333 __LINK_STATE_DORMANT, 334 __LINK_STATE_TESTING, 335 }; 336 337 struct gro_list { 338 struct list_head list; 339 int count; 340 }; 341 342 /* 343 * size of gro hash buckets, must be <= the number of bits in 344 * gro_node::bitmask 345 */ 346 #define GRO_HASH_BUCKETS 8 347 348 /** 349 * struct gro_node - structure to support Generic Receive Offload 350 * @bitmask: bitmask to indicate used buckets in @hash 351 * @hash: hashtable of pending aggregated skbs, separated by flows 352 * @rx_list: list of pending ``GRO_NORMAL`` skbs 353 * @rx_count: cached current length of @rx_list 354 * @cached_napi_id: napi_struct::napi_id cached for hotpath, 0 for standalone 355 */ 356 struct gro_node { 357 unsigned long bitmask; 358 struct gro_list hash[GRO_HASH_BUCKETS]; 359 struct list_head rx_list; 360 u32 rx_count; 361 u32 cached_napi_id; 362 }; 363 364 /* 365 * Structure for per-NAPI config 366 */ 367 struct napi_config { 368 u64 gro_flush_timeout; 369 u64 irq_suspend_timeout; 370 u32 defer_hard_irqs; 371 cpumask_t affinity_mask; 372 unsigned int napi_id; 373 }; 374 375 /* 376 * Structure for NAPI scheduling similar to tasklet but with weighting 377 */ 378 struct napi_struct { 379 /* The poll_list must only be managed by the entity which 380 * changes the state of the NAPI_STATE_SCHED bit. This means 381 * whoever atomically sets that bit can add this napi_struct 382 * to the per-CPU poll_list, and whoever clears that bit 383 * can remove from the list right before clearing the bit. 384 */ 385 struct list_head poll_list; 386 387 unsigned long state; 388 int weight; 389 u32 defer_hard_irqs_count; 390 int (*poll)(struct napi_struct *, int); 391 #ifdef CONFIG_NETPOLL 392 /* CPU actively polling if netpoll is configured */ 393 int poll_owner; 394 #endif 395 /* CPU on which NAPI has been scheduled for processing */ 396 int list_owner; 397 struct net_device *dev; 398 struct sk_buff *skb; 399 struct gro_node gro; 400 struct hrtimer timer; 401 /* all fields past this point are write-protected by netdev_lock */ 402 struct task_struct *thread; 403 unsigned long gro_flush_timeout; 404 unsigned long irq_suspend_timeout; 405 u32 defer_hard_irqs; 406 /* control-path-only fields follow */ 407 u32 napi_id; 408 struct list_head dev_list; 409 struct hlist_node napi_hash_node; 410 int irq; 411 struct irq_affinity_notify notify; 412 int napi_rmap_idx; 413 int index; 414 struct napi_config *config; 415 }; 416 417 enum { 418 NAPI_STATE_SCHED, /* Poll is scheduled */ 419 NAPI_STATE_MISSED, /* reschedule a napi */ 420 NAPI_STATE_DISABLE, /* Disable pending */ 421 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 422 NAPI_STATE_LISTED, /* NAPI added to system lists */ 423 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 424 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 425 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 426 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 427 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 428 NAPI_STATE_HAS_NOTIFIER, /* Napi has an IRQ notifier */ 429 }; 430 431 enum { 432 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 433 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 434 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 435 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 436 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 437 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 438 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 439 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 440 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 441 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 442 NAPIF_STATE_HAS_NOTIFIER = BIT(NAPI_STATE_HAS_NOTIFIER), 443 }; 444 445 enum gro_result { 446 GRO_MERGED, 447 GRO_MERGED_FREE, 448 GRO_HELD, 449 GRO_NORMAL, 450 GRO_CONSUMED, 451 }; 452 typedef enum gro_result gro_result_t; 453 454 /* 455 * enum rx_handler_result - Possible return values for rx_handlers. 456 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 457 * further. 458 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 459 * case skb->dev was changed by rx_handler. 460 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 461 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 462 * 463 * rx_handlers are functions called from inside __netif_receive_skb(), to do 464 * special processing of the skb, prior to delivery to protocol handlers. 465 * 466 * Currently, a net_device can only have a single rx_handler registered. Trying 467 * to register a second rx_handler will return -EBUSY. 468 * 469 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 470 * To unregister a rx_handler on a net_device, use 471 * netdev_rx_handler_unregister(). 472 * 473 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 474 * do with the skb. 475 * 476 * If the rx_handler consumed the skb in some way, it should return 477 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 478 * the skb to be delivered in some other way. 479 * 480 * If the rx_handler changed skb->dev, to divert the skb to another 481 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 482 * new device will be called if it exists. 483 * 484 * If the rx_handler decides the skb should be ignored, it should return 485 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 486 * are registered on exact device (ptype->dev == skb->dev). 487 * 488 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 489 * delivered, it should return RX_HANDLER_PASS. 490 * 491 * A device without a registered rx_handler will behave as if rx_handler 492 * returned RX_HANDLER_PASS. 493 */ 494 495 enum rx_handler_result { 496 RX_HANDLER_CONSUMED, 497 RX_HANDLER_ANOTHER, 498 RX_HANDLER_EXACT, 499 RX_HANDLER_PASS, 500 }; 501 typedef enum rx_handler_result rx_handler_result_t; 502 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 503 504 void __napi_schedule(struct napi_struct *n); 505 void __napi_schedule_irqoff(struct napi_struct *n); 506 507 static inline bool napi_disable_pending(struct napi_struct *n) 508 { 509 return test_bit(NAPI_STATE_DISABLE, &n->state); 510 } 511 512 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 513 { 514 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 515 } 516 517 /** 518 * napi_is_scheduled - test if NAPI is scheduled 519 * @n: NAPI context 520 * 521 * This check is "best-effort". With no locking implemented, 522 * a NAPI can be scheduled or terminate right after this check 523 * and produce not precise results. 524 * 525 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled 526 * should not be used normally and napi_schedule should be 527 * used instead. 528 * 529 * Use only if the driver really needs to check if a NAPI 530 * is scheduled for example in the context of delayed timer 531 * that can be skipped if a NAPI is already scheduled. 532 * 533 * Return: True if NAPI is scheduled, False otherwise. 534 */ 535 static inline bool napi_is_scheduled(struct napi_struct *n) 536 { 537 return test_bit(NAPI_STATE_SCHED, &n->state); 538 } 539 540 bool napi_schedule_prep(struct napi_struct *n); 541 542 /** 543 * napi_schedule - schedule NAPI poll 544 * @n: NAPI context 545 * 546 * Schedule NAPI poll routine to be called if it is not already 547 * running. 548 * Return: true if we schedule a NAPI or false if not. 549 * Refer to napi_schedule_prep() for additional reason on why 550 * a NAPI might not be scheduled. 551 */ 552 static inline bool napi_schedule(struct napi_struct *n) 553 { 554 if (napi_schedule_prep(n)) { 555 __napi_schedule(n); 556 return true; 557 } 558 559 return false; 560 } 561 562 /** 563 * napi_schedule_irqoff - schedule NAPI poll 564 * @n: NAPI context 565 * 566 * Variant of napi_schedule(), assuming hard irqs are masked. 567 */ 568 static inline void napi_schedule_irqoff(struct napi_struct *n) 569 { 570 if (napi_schedule_prep(n)) 571 __napi_schedule_irqoff(n); 572 } 573 574 /** 575 * napi_complete_done - NAPI processing complete 576 * @n: NAPI context 577 * @work_done: number of packets processed 578 * 579 * Mark NAPI processing as complete. Should only be called if poll budget 580 * has not been completely consumed. 581 * Prefer over napi_complete(). 582 * Return: false if device should avoid rearming interrupts. 583 */ 584 bool napi_complete_done(struct napi_struct *n, int work_done); 585 586 static inline bool napi_complete(struct napi_struct *n) 587 { 588 return napi_complete_done(n, 0); 589 } 590 591 int dev_set_threaded(struct net_device *dev, bool threaded); 592 593 void napi_disable(struct napi_struct *n); 594 void napi_disable_locked(struct napi_struct *n); 595 596 void napi_enable(struct napi_struct *n); 597 void napi_enable_locked(struct napi_struct *n); 598 599 /** 600 * napi_synchronize - wait until NAPI is not running 601 * @n: NAPI context 602 * 603 * Wait until NAPI is done being scheduled on this context. 604 * Waits till any outstanding processing completes but 605 * does not disable future activations. 606 */ 607 static inline void napi_synchronize(const struct napi_struct *n) 608 { 609 if (IS_ENABLED(CONFIG_SMP)) 610 while (test_bit(NAPI_STATE_SCHED, &n->state)) 611 msleep(1); 612 else 613 barrier(); 614 } 615 616 /** 617 * napi_if_scheduled_mark_missed - if napi is running, set the 618 * NAPIF_STATE_MISSED 619 * @n: NAPI context 620 * 621 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 622 * NAPI is scheduled. 623 **/ 624 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 625 { 626 unsigned long val, new; 627 628 val = READ_ONCE(n->state); 629 do { 630 if (val & NAPIF_STATE_DISABLE) 631 return true; 632 633 if (!(val & NAPIF_STATE_SCHED)) 634 return false; 635 636 new = val | NAPIF_STATE_MISSED; 637 } while (!try_cmpxchg(&n->state, &val, new)); 638 639 return true; 640 } 641 642 enum netdev_queue_state_t { 643 __QUEUE_STATE_DRV_XOFF, 644 __QUEUE_STATE_STACK_XOFF, 645 __QUEUE_STATE_FROZEN, 646 }; 647 648 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 649 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 650 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 651 652 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 653 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 654 QUEUE_STATE_FROZEN) 655 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 656 QUEUE_STATE_FROZEN) 657 658 /* 659 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 660 * netif_tx_* functions below are used to manipulate this flag. The 661 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 662 * queue independently. The netif_xmit_*stopped functions below are called 663 * to check if the queue has been stopped by the driver or stack (either 664 * of the XOFF bits are set in the state). Drivers should not need to call 665 * netif_xmit*stopped functions, they should only be using netif_tx_*. 666 */ 667 668 struct netdev_queue { 669 /* 670 * read-mostly part 671 */ 672 struct net_device *dev; 673 netdevice_tracker dev_tracker; 674 675 struct Qdisc __rcu *qdisc; 676 struct Qdisc __rcu *qdisc_sleeping; 677 #ifdef CONFIG_SYSFS 678 struct kobject kobj; 679 const struct attribute_group **groups; 680 #endif 681 unsigned long tx_maxrate; 682 /* 683 * Number of TX timeouts for this queue 684 * (/sys/class/net/DEV/Q/trans_timeout) 685 */ 686 atomic_long_t trans_timeout; 687 688 /* Subordinate device that the queue has been assigned to */ 689 struct net_device *sb_dev; 690 #ifdef CONFIG_XDP_SOCKETS 691 struct xsk_buff_pool *pool; 692 #endif 693 694 /* 695 * write-mostly part 696 */ 697 #ifdef CONFIG_BQL 698 struct dql dql; 699 #endif 700 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 701 int xmit_lock_owner; 702 /* 703 * Time (in jiffies) of last Tx 704 */ 705 unsigned long trans_start; 706 707 unsigned long state; 708 709 /* 710 * slow- / control-path part 711 */ 712 /* NAPI instance for the queue 713 * Readers and writers must hold RTNL 714 */ 715 struct napi_struct *napi; 716 717 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 718 int numa_node; 719 #endif 720 } ____cacheline_aligned_in_smp; 721 722 extern int sysctl_fb_tunnels_only_for_init_net; 723 extern int sysctl_devconf_inherit_init_net; 724 725 /* 726 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 727 * == 1 : For initns only 728 * == 2 : For none. 729 */ 730 static inline bool net_has_fallback_tunnels(const struct net *net) 731 { 732 #if IS_ENABLED(CONFIG_SYSCTL) 733 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net); 734 735 return !fb_tunnels_only_for_init_net || 736 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1); 737 #else 738 return true; 739 #endif 740 } 741 742 static inline int net_inherit_devconf(void) 743 { 744 #if IS_ENABLED(CONFIG_SYSCTL) 745 return READ_ONCE(sysctl_devconf_inherit_init_net); 746 #else 747 return 0; 748 #endif 749 } 750 751 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 752 { 753 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 754 return q->numa_node; 755 #else 756 return NUMA_NO_NODE; 757 #endif 758 } 759 760 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 761 { 762 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 763 q->numa_node = node; 764 #endif 765 } 766 767 #ifdef CONFIG_RFS_ACCEL 768 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 769 u16 filter_id); 770 #endif 771 772 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 773 enum xps_map_type { 774 XPS_CPUS = 0, 775 XPS_RXQS, 776 XPS_MAPS_MAX, 777 }; 778 779 #ifdef CONFIG_XPS 780 /* 781 * This structure holds an XPS map which can be of variable length. The 782 * map is an array of queues. 783 */ 784 struct xps_map { 785 unsigned int len; 786 unsigned int alloc_len; 787 struct rcu_head rcu; 788 u16 queues[]; 789 }; 790 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 791 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 792 - sizeof(struct xps_map)) / sizeof(u16)) 793 794 /* 795 * This structure holds all XPS maps for device. Maps are indexed by CPU. 796 * 797 * We keep track of the number of cpus/rxqs used when the struct is allocated, 798 * in nr_ids. This will help not accessing out-of-bound memory. 799 * 800 * We keep track of the number of traffic classes used when the struct is 801 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 802 * not crossing its upper bound, as the original dev->num_tc can be updated in 803 * the meantime. 804 */ 805 struct xps_dev_maps { 806 struct rcu_head rcu; 807 unsigned int nr_ids; 808 s16 num_tc; 809 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 810 }; 811 812 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 813 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 814 815 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 816 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 817 818 #endif /* CONFIG_XPS */ 819 820 #define TC_MAX_QUEUE 16 821 #define TC_BITMASK 15 822 /* HW offloaded queuing disciplines txq count and offset maps */ 823 struct netdev_tc_txq { 824 u16 count; 825 u16 offset; 826 }; 827 828 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 829 /* 830 * This structure is to hold information about the device 831 * configured to run FCoE protocol stack. 832 */ 833 struct netdev_fcoe_hbainfo { 834 char manufacturer[64]; 835 char serial_number[64]; 836 char hardware_version[64]; 837 char driver_version[64]; 838 char optionrom_version[64]; 839 char firmware_version[64]; 840 char model[256]; 841 char model_description[256]; 842 }; 843 #endif 844 845 #define MAX_PHYS_ITEM_ID_LEN 32 846 847 /* This structure holds a unique identifier to identify some 848 * physical item (port for example) used by a netdevice. 849 */ 850 struct netdev_phys_item_id { 851 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 852 unsigned char id_len; 853 }; 854 855 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 856 struct netdev_phys_item_id *b) 857 { 858 return a->id_len == b->id_len && 859 memcmp(a->id, b->id, a->id_len) == 0; 860 } 861 862 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 863 struct sk_buff *skb, 864 struct net_device *sb_dev); 865 866 enum net_device_path_type { 867 DEV_PATH_ETHERNET = 0, 868 DEV_PATH_VLAN, 869 DEV_PATH_BRIDGE, 870 DEV_PATH_PPPOE, 871 DEV_PATH_DSA, 872 DEV_PATH_MTK_WDMA, 873 }; 874 875 struct net_device_path { 876 enum net_device_path_type type; 877 const struct net_device *dev; 878 union { 879 struct { 880 u16 id; 881 __be16 proto; 882 u8 h_dest[ETH_ALEN]; 883 } encap; 884 struct { 885 enum { 886 DEV_PATH_BR_VLAN_KEEP, 887 DEV_PATH_BR_VLAN_TAG, 888 DEV_PATH_BR_VLAN_UNTAG, 889 DEV_PATH_BR_VLAN_UNTAG_HW, 890 } vlan_mode; 891 u16 vlan_id; 892 __be16 vlan_proto; 893 } bridge; 894 struct { 895 int port; 896 u16 proto; 897 } dsa; 898 struct { 899 u8 wdma_idx; 900 u8 queue; 901 u16 wcid; 902 u8 bss; 903 u8 amsdu; 904 } mtk_wdma; 905 }; 906 }; 907 908 #define NET_DEVICE_PATH_STACK_MAX 5 909 #define NET_DEVICE_PATH_VLAN_MAX 2 910 911 struct net_device_path_stack { 912 int num_paths; 913 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 914 }; 915 916 struct net_device_path_ctx { 917 const struct net_device *dev; 918 u8 daddr[ETH_ALEN]; 919 920 int num_vlans; 921 struct { 922 u16 id; 923 __be16 proto; 924 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 925 }; 926 927 enum tc_setup_type { 928 TC_QUERY_CAPS, 929 TC_SETUP_QDISC_MQPRIO, 930 TC_SETUP_CLSU32, 931 TC_SETUP_CLSFLOWER, 932 TC_SETUP_CLSMATCHALL, 933 TC_SETUP_CLSBPF, 934 TC_SETUP_BLOCK, 935 TC_SETUP_QDISC_CBS, 936 TC_SETUP_QDISC_RED, 937 TC_SETUP_QDISC_PRIO, 938 TC_SETUP_QDISC_MQ, 939 TC_SETUP_QDISC_ETF, 940 TC_SETUP_ROOT_QDISC, 941 TC_SETUP_QDISC_GRED, 942 TC_SETUP_QDISC_TAPRIO, 943 TC_SETUP_FT, 944 TC_SETUP_QDISC_ETS, 945 TC_SETUP_QDISC_TBF, 946 TC_SETUP_QDISC_FIFO, 947 TC_SETUP_QDISC_HTB, 948 TC_SETUP_ACT, 949 }; 950 951 /* These structures hold the attributes of bpf state that are being passed 952 * to the netdevice through the bpf op. 953 */ 954 enum bpf_netdev_command { 955 /* Set or clear a bpf program used in the earliest stages of packet 956 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 957 * is responsible for calling bpf_prog_put on any old progs that are 958 * stored. In case of error, the callee need not release the new prog 959 * reference, but on success it takes ownership and must bpf_prog_put 960 * when it is no longer used. 961 */ 962 XDP_SETUP_PROG, 963 XDP_SETUP_PROG_HW, 964 /* BPF program for offload callbacks, invoked at program load time. */ 965 BPF_OFFLOAD_MAP_ALLOC, 966 BPF_OFFLOAD_MAP_FREE, 967 XDP_SETUP_XSK_POOL, 968 }; 969 970 struct bpf_prog_offload_ops; 971 struct netlink_ext_ack; 972 struct xdp_umem; 973 struct xdp_dev_bulk_queue; 974 struct bpf_xdp_link; 975 976 enum bpf_xdp_mode { 977 XDP_MODE_SKB = 0, 978 XDP_MODE_DRV = 1, 979 XDP_MODE_HW = 2, 980 __MAX_XDP_MODE 981 }; 982 983 struct bpf_xdp_entity { 984 struct bpf_prog *prog; 985 struct bpf_xdp_link *link; 986 }; 987 988 struct netdev_bpf { 989 enum bpf_netdev_command command; 990 union { 991 /* XDP_SETUP_PROG */ 992 struct { 993 u32 flags; 994 struct bpf_prog *prog; 995 struct netlink_ext_ack *extack; 996 }; 997 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 998 struct { 999 struct bpf_offloaded_map *offmap; 1000 }; 1001 /* XDP_SETUP_XSK_POOL */ 1002 struct { 1003 struct xsk_buff_pool *pool; 1004 u16 queue_id; 1005 } xsk; 1006 }; 1007 }; 1008 1009 /* Flags for ndo_xsk_wakeup. */ 1010 #define XDP_WAKEUP_RX (1 << 0) 1011 #define XDP_WAKEUP_TX (1 << 1) 1012 1013 #ifdef CONFIG_XFRM_OFFLOAD 1014 struct xfrmdev_ops { 1015 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack); 1016 void (*xdo_dev_state_delete) (struct xfrm_state *x); 1017 void (*xdo_dev_state_free) (struct xfrm_state *x); 1018 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 1019 struct xfrm_state *x); 1020 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1021 void (*xdo_dev_state_update_stats) (struct xfrm_state *x); 1022 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack); 1023 void (*xdo_dev_policy_delete) (struct xfrm_policy *x); 1024 void (*xdo_dev_policy_free) (struct xfrm_policy *x); 1025 }; 1026 #endif 1027 1028 struct dev_ifalias { 1029 struct rcu_head rcuhead; 1030 char ifalias[]; 1031 }; 1032 1033 struct devlink; 1034 struct tlsdev_ops; 1035 1036 struct netdev_net_notifier { 1037 struct list_head list; 1038 struct notifier_block *nb; 1039 }; 1040 1041 /* 1042 * This structure defines the management hooks for network devices. 1043 * The following hooks can be defined; unless noted otherwise, they are 1044 * optional and can be filled with a null pointer. 1045 * 1046 * int (*ndo_init)(struct net_device *dev); 1047 * This function is called once when a network device is registered. 1048 * The network device can use this for any late stage initialization 1049 * or semantic validation. It can fail with an error code which will 1050 * be propagated back to register_netdev. 1051 * 1052 * void (*ndo_uninit)(struct net_device *dev); 1053 * This function is called when device is unregistered or when registration 1054 * fails. It is not called if init fails. 1055 * 1056 * int (*ndo_open)(struct net_device *dev); 1057 * This function is called when a network device transitions to the up 1058 * state. 1059 * 1060 * int (*ndo_stop)(struct net_device *dev); 1061 * This function is called when a network device transitions to the down 1062 * state. 1063 * 1064 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1065 * struct net_device *dev); 1066 * Called when a packet needs to be transmitted. 1067 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1068 * the queue before that can happen; it's for obsolete devices and weird 1069 * corner cases, but the stack really does a non-trivial amount 1070 * of useless work if you return NETDEV_TX_BUSY. 1071 * Required; cannot be NULL. 1072 * 1073 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1074 * struct net_device *dev 1075 * netdev_features_t features); 1076 * Called by core transmit path to determine if device is capable of 1077 * performing offload operations on a given packet. This is to give 1078 * the device an opportunity to implement any restrictions that cannot 1079 * be otherwise expressed by feature flags. The check is called with 1080 * the set of features that the stack has calculated and it returns 1081 * those the driver believes to be appropriate. 1082 * 1083 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1084 * struct net_device *sb_dev); 1085 * Called to decide which queue to use when device supports multiple 1086 * transmit queues. 1087 * 1088 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1089 * This function is called to allow device receiver to make 1090 * changes to configuration when multicast or promiscuous is enabled. 1091 * 1092 * void (*ndo_set_rx_mode)(struct net_device *dev); 1093 * This function is called device changes address list filtering. 1094 * If driver handles unicast address filtering, it should set 1095 * IFF_UNICAST_FLT in its priv_flags. 1096 * 1097 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1098 * This function is called when the Media Access Control address 1099 * needs to be changed. If this interface is not defined, the 1100 * MAC address can not be changed. 1101 * 1102 * int (*ndo_validate_addr)(struct net_device *dev); 1103 * Test if Media Access Control address is valid for the device. 1104 * 1105 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1106 * Old-style ioctl entry point. This is used internally by the 1107 * ieee802154 subsystem but is no longer called by the device 1108 * ioctl handler. 1109 * 1110 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1111 * Used by the bonding driver for its device specific ioctls: 1112 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1113 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1114 * 1115 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1116 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1117 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1118 * 1119 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1120 * Used to set network devices bus interface parameters. This interface 1121 * is retained for legacy reasons; new devices should use the bus 1122 * interface (PCI) for low level management. 1123 * 1124 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1125 * Called when a user wants to change the Maximum Transfer Unit 1126 * of a device. 1127 * 1128 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1129 * Callback used when the transmitter has not made any progress 1130 * for dev->watchdog ticks. 1131 * 1132 * void (*ndo_get_stats64)(struct net_device *dev, 1133 * struct rtnl_link_stats64 *storage); 1134 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1135 * Called when a user wants to get the network device usage 1136 * statistics. Drivers must do one of the following: 1137 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1138 * rtnl_link_stats64 structure passed by the caller. 1139 * 2. Define @ndo_get_stats to update a net_device_stats structure 1140 * (which should normally be dev->stats) and return a pointer to 1141 * it. The structure may be changed asynchronously only if each 1142 * field is written atomically. 1143 * 3. Update dev->stats asynchronously and atomically, and define 1144 * neither operation. 1145 * 1146 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1147 * Return true if this device supports offload stats of this attr_id. 1148 * 1149 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1150 * void *attr_data) 1151 * Get statistics for offload operations by attr_id. Write it into the 1152 * attr_data pointer. 1153 * 1154 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1155 * If device supports VLAN filtering this function is called when a 1156 * VLAN id is registered. 1157 * 1158 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1159 * If device supports VLAN filtering this function is called when a 1160 * VLAN id is unregistered. 1161 * 1162 * void (*ndo_poll_controller)(struct net_device *dev); 1163 * 1164 * SR-IOV management functions. 1165 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1166 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1167 * u8 qos, __be16 proto); 1168 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1169 * int max_tx_rate); 1170 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1171 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1172 * int (*ndo_get_vf_config)(struct net_device *dev, 1173 * int vf, struct ifla_vf_info *ivf); 1174 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1175 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1176 * struct nlattr *port[]); 1177 * 1178 * Enable or disable the VF ability to query its RSS Redirection Table and 1179 * Hash Key. This is needed since on some devices VF share this information 1180 * with PF and querying it may introduce a theoretical security risk. 1181 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1182 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1183 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1184 * void *type_data); 1185 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1186 * This is always called from the stack with the rtnl lock held and netif 1187 * tx queues stopped. This allows the netdevice to perform queue 1188 * management safely. 1189 * 1190 * Fiber Channel over Ethernet (FCoE) offload functions. 1191 * int (*ndo_fcoe_enable)(struct net_device *dev); 1192 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1193 * so the underlying device can perform whatever needed configuration or 1194 * initialization to support acceleration of FCoE traffic. 1195 * 1196 * int (*ndo_fcoe_disable)(struct net_device *dev); 1197 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1198 * so the underlying device can perform whatever needed clean-ups to 1199 * stop supporting acceleration of FCoE traffic. 1200 * 1201 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1202 * struct scatterlist *sgl, unsigned int sgc); 1203 * Called when the FCoE Initiator wants to initialize an I/O that 1204 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1205 * perform necessary setup and returns 1 to indicate the device is set up 1206 * successfully to perform DDP on this I/O, otherwise this returns 0. 1207 * 1208 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1209 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1210 * indicated by the FC exchange id 'xid', so the underlying device can 1211 * clean up and reuse resources for later DDP requests. 1212 * 1213 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1214 * struct scatterlist *sgl, unsigned int sgc); 1215 * Called when the FCoE Target wants to initialize an I/O that 1216 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1217 * perform necessary setup and returns 1 to indicate the device is set up 1218 * successfully to perform DDP on this I/O, otherwise this returns 0. 1219 * 1220 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1221 * struct netdev_fcoe_hbainfo *hbainfo); 1222 * Called when the FCoE Protocol stack wants information on the underlying 1223 * device. This information is utilized by the FCoE protocol stack to 1224 * register attributes with Fiber Channel management service as per the 1225 * FC-GS Fabric Device Management Information(FDMI) specification. 1226 * 1227 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1228 * Called when the underlying device wants to override default World Wide 1229 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1230 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1231 * protocol stack to use. 1232 * 1233 * RFS acceleration. 1234 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1235 * u16 rxq_index, u32 flow_id); 1236 * Set hardware filter for RFS. rxq_index is the target queue index; 1237 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1238 * Return the filter ID on success, or a negative error code. 1239 * 1240 * Slave management functions (for bridge, bonding, etc). 1241 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1242 * Called to make another netdev an underling. 1243 * 1244 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1245 * Called to release previously enslaved netdev. 1246 * 1247 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1248 * struct sk_buff *skb, 1249 * bool all_slaves); 1250 * Get the xmit slave of master device. If all_slaves is true, function 1251 * assume all the slaves can transmit. 1252 * 1253 * Feature/offload setting functions. 1254 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1255 * netdev_features_t features); 1256 * Adjusts the requested feature flags according to device-specific 1257 * constraints, and returns the resulting flags. Must not modify 1258 * the device state. 1259 * 1260 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1261 * Called to update device configuration to new features. Passed 1262 * feature set might be less than what was returned by ndo_fix_features()). 1263 * Must return >0 or -errno if it changed dev->features itself. 1264 * 1265 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1266 * struct net_device *dev, 1267 * const unsigned char *addr, u16 vid, u16 flags, 1268 * bool *notified, struct netlink_ext_ack *extack); 1269 * Adds an FDB entry to dev for addr. 1270 * Callee shall set *notified to true if it sent any appropriate 1271 * notification(s). Otherwise core will send a generic one. 1272 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1273 * struct net_device *dev, 1274 * const unsigned char *addr, u16 vid 1275 * bool *notified, struct netlink_ext_ack *extack); 1276 * Deletes the FDB entry from dev corresponding to addr. 1277 * Callee shall set *notified to true if it sent any appropriate 1278 * notification(s). Otherwise core will send a generic one. 1279 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev, 1280 * struct netlink_ext_ack *extack); 1281 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1282 * struct net_device *dev, struct net_device *filter_dev, 1283 * int *idx) 1284 * Used to add FDB entries to dump requests. Implementers should add 1285 * entries to skb and update idx with the number of entries. 1286 * 1287 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[], 1288 * u16 nlmsg_flags, struct netlink_ext_ack *extack); 1289 * Adds an MDB entry to dev. 1290 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[], 1291 * struct netlink_ext_ack *extack); 1292 * Deletes the MDB entry from dev. 1293 * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[], 1294 * struct netlink_ext_ack *extack); 1295 * Bulk deletes MDB entries from dev. 1296 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb, 1297 * struct netlink_callback *cb); 1298 * Dumps MDB entries from dev. The first argument (marker) in the netlink 1299 * callback is used by core rtnetlink code. 1300 * 1301 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1302 * u16 flags, struct netlink_ext_ack *extack) 1303 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1304 * struct net_device *dev, u32 filter_mask, 1305 * int nlflags) 1306 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1307 * u16 flags); 1308 * 1309 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1310 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1311 * which do not represent real hardware may define this to allow their 1312 * userspace components to manage their virtual carrier state. Devices 1313 * that determine carrier state from physical hardware properties (eg 1314 * network cables) or protocol-dependent mechanisms (eg 1315 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1316 * 1317 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1318 * struct netdev_phys_item_id *ppid); 1319 * Called to get ID of physical port of this device. If driver does 1320 * not implement this, it is assumed that the hw is not able to have 1321 * multiple net devices on single physical port. 1322 * 1323 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1324 * struct netdev_phys_item_id *ppid) 1325 * Called to get the parent ID of the physical port of this device. 1326 * 1327 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1328 * struct net_device *dev) 1329 * Called by upper layer devices to accelerate switching or other 1330 * station functionality into hardware. 'pdev is the lowerdev 1331 * to use for the offload and 'dev' is the net device that will 1332 * back the offload. Returns a pointer to the private structure 1333 * the upper layer will maintain. 1334 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1335 * Called by upper layer device to delete the station created 1336 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1337 * the station and priv is the structure returned by the add 1338 * operation. 1339 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1340 * int queue_index, u32 maxrate); 1341 * Called when a user wants to set a max-rate limitation of specific 1342 * TX queue. 1343 * int (*ndo_get_iflink)(const struct net_device *dev); 1344 * Called to get the iflink value of this device. 1345 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1346 * This function is used to get egress tunnel information for given skb. 1347 * This is useful for retrieving outer tunnel header parameters while 1348 * sampling packet. 1349 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1350 * This function is used to specify the headroom that the skb must 1351 * consider when allocation skb during packet reception. Setting 1352 * appropriate rx headroom value allows avoiding skb head copy on 1353 * forward. Setting a negative value resets the rx headroom to the 1354 * default value. 1355 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1356 * This function is used to set or query state related to XDP on the 1357 * netdevice and manage BPF offload. See definition of 1358 * enum bpf_netdev_command for details. 1359 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1360 * u32 flags); 1361 * This function is used to submit @n XDP packets for transmit on a 1362 * netdevice. Returns number of frames successfully transmitted, frames 1363 * that got dropped are freed/returned via xdp_return_frame(). 1364 * Returns negative number, means general error invoking ndo, meaning 1365 * no frames were xmit'ed and core-caller will free all frames. 1366 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1367 * struct xdp_buff *xdp); 1368 * Get the xmit slave of master device based on the xdp_buff. 1369 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1370 * This function is used to wake up the softirq, ksoftirqd or kthread 1371 * responsible for sending and/or receiving packets on a specific 1372 * queue id bound to an AF_XDP socket. The flags field specifies if 1373 * only RX, only Tx, or both should be woken up using the flags 1374 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1375 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm_kern *p, 1376 * int cmd); 1377 * Add, change, delete or get information on an IPv4 tunnel. 1378 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1379 * If a device is paired with a peer device, return the peer instance. 1380 * The caller must be under RCU read context. 1381 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1382 * Get the forwarding path to reach the real device from the HW destination address 1383 * ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1384 * const struct skb_shared_hwtstamps *hwtstamps, 1385 * bool cycles); 1386 * Get hardware timestamp based on normal/adjustable time or free running 1387 * cycle counter. This function is required if physical clock supports a 1388 * free running cycle counter. 1389 * 1390 * int (*ndo_hwtstamp_get)(struct net_device *dev, 1391 * struct kernel_hwtstamp_config *kernel_config); 1392 * Get the currently configured hardware timestamping parameters for the 1393 * NIC device. 1394 * 1395 * int (*ndo_hwtstamp_set)(struct net_device *dev, 1396 * struct kernel_hwtstamp_config *kernel_config, 1397 * struct netlink_ext_ack *extack); 1398 * Change the hardware timestamping parameters for NIC device. 1399 */ 1400 struct net_device_ops { 1401 int (*ndo_init)(struct net_device *dev); 1402 void (*ndo_uninit)(struct net_device *dev); 1403 int (*ndo_open)(struct net_device *dev); 1404 int (*ndo_stop)(struct net_device *dev); 1405 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1406 struct net_device *dev); 1407 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1408 struct net_device *dev, 1409 netdev_features_t features); 1410 u16 (*ndo_select_queue)(struct net_device *dev, 1411 struct sk_buff *skb, 1412 struct net_device *sb_dev); 1413 void (*ndo_change_rx_flags)(struct net_device *dev, 1414 int flags); 1415 void (*ndo_set_rx_mode)(struct net_device *dev); 1416 int (*ndo_set_mac_address)(struct net_device *dev, 1417 void *addr); 1418 int (*ndo_validate_addr)(struct net_device *dev); 1419 int (*ndo_do_ioctl)(struct net_device *dev, 1420 struct ifreq *ifr, int cmd); 1421 int (*ndo_eth_ioctl)(struct net_device *dev, 1422 struct ifreq *ifr, int cmd); 1423 int (*ndo_siocbond)(struct net_device *dev, 1424 struct ifreq *ifr, int cmd); 1425 int (*ndo_siocwandev)(struct net_device *dev, 1426 struct if_settings *ifs); 1427 int (*ndo_siocdevprivate)(struct net_device *dev, 1428 struct ifreq *ifr, 1429 void __user *data, int cmd); 1430 int (*ndo_set_config)(struct net_device *dev, 1431 struct ifmap *map); 1432 int (*ndo_change_mtu)(struct net_device *dev, 1433 int new_mtu); 1434 int (*ndo_neigh_setup)(struct net_device *dev, 1435 struct neigh_parms *); 1436 void (*ndo_tx_timeout) (struct net_device *dev, 1437 unsigned int txqueue); 1438 1439 void (*ndo_get_stats64)(struct net_device *dev, 1440 struct rtnl_link_stats64 *storage); 1441 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1442 int (*ndo_get_offload_stats)(int attr_id, 1443 const struct net_device *dev, 1444 void *attr_data); 1445 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1446 1447 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1448 __be16 proto, u16 vid); 1449 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1450 __be16 proto, u16 vid); 1451 #ifdef CONFIG_NET_POLL_CONTROLLER 1452 void (*ndo_poll_controller)(struct net_device *dev); 1453 int (*ndo_netpoll_setup)(struct net_device *dev); 1454 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1455 #endif 1456 int (*ndo_set_vf_mac)(struct net_device *dev, 1457 int queue, u8 *mac); 1458 int (*ndo_set_vf_vlan)(struct net_device *dev, 1459 int queue, u16 vlan, 1460 u8 qos, __be16 proto); 1461 int (*ndo_set_vf_rate)(struct net_device *dev, 1462 int vf, int min_tx_rate, 1463 int max_tx_rate); 1464 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1465 int vf, bool setting); 1466 int (*ndo_set_vf_trust)(struct net_device *dev, 1467 int vf, bool setting); 1468 int (*ndo_get_vf_config)(struct net_device *dev, 1469 int vf, 1470 struct ifla_vf_info *ivf); 1471 int (*ndo_set_vf_link_state)(struct net_device *dev, 1472 int vf, int link_state); 1473 int (*ndo_get_vf_stats)(struct net_device *dev, 1474 int vf, 1475 struct ifla_vf_stats 1476 *vf_stats); 1477 int (*ndo_set_vf_port)(struct net_device *dev, 1478 int vf, 1479 struct nlattr *port[]); 1480 int (*ndo_get_vf_port)(struct net_device *dev, 1481 int vf, struct sk_buff *skb); 1482 int (*ndo_get_vf_guid)(struct net_device *dev, 1483 int vf, 1484 struct ifla_vf_guid *node_guid, 1485 struct ifla_vf_guid *port_guid); 1486 int (*ndo_set_vf_guid)(struct net_device *dev, 1487 int vf, u64 guid, 1488 int guid_type); 1489 int (*ndo_set_vf_rss_query_en)( 1490 struct net_device *dev, 1491 int vf, bool setting); 1492 int (*ndo_setup_tc)(struct net_device *dev, 1493 enum tc_setup_type type, 1494 void *type_data); 1495 #if IS_ENABLED(CONFIG_FCOE) 1496 int (*ndo_fcoe_enable)(struct net_device *dev); 1497 int (*ndo_fcoe_disable)(struct net_device *dev); 1498 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1499 u16 xid, 1500 struct scatterlist *sgl, 1501 unsigned int sgc); 1502 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1503 u16 xid); 1504 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1505 u16 xid, 1506 struct scatterlist *sgl, 1507 unsigned int sgc); 1508 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1509 struct netdev_fcoe_hbainfo *hbainfo); 1510 #endif 1511 1512 #if IS_ENABLED(CONFIG_LIBFCOE) 1513 #define NETDEV_FCOE_WWNN 0 1514 #define NETDEV_FCOE_WWPN 1 1515 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1516 u64 *wwn, int type); 1517 #endif 1518 1519 #ifdef CONFIG_RFS_ACCEL 1520 int (*ndo_rx_flow_steer)(struct net_device *dev, 1521 const struct sk_buff *skb, 1522 u16 rxq_index, 1523 u32 flow_id); 1524 #endif 1525 int (*ndo_add_slave)(struct net_device *dev, 1526 struct net_device *slave_dev, 1527 struct netlink_ext_ack *extack); 1528 int (*ndo_del_slave)(struct net_device *dev, 1529 struct net_device *slave_dev); 1530 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1531 struct sk_buff *skb, 1532 bool all_slaves); 1533 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1534 struct sock *sk); 1535 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1536 netdev_features_t features); 1537 int (*ndo_set_features)(struct net_device *dev, 1538 netdev_features_t features); 1539 int (*ndo_neigh_construct)(struct net_device *dev, 1540 struct neighbour *n); 1541 void (*ndo_neigh_destroy)(struct net_device *dev, 1542 struct neighbour *n); 1543 1544 int (*ndo_fdb_add)(struct ndmsg *ndm, 1545 struct nlattr *tb[], 1546 struct net_device *dev, 1547 const unsigned char *addr, 1548 u16 vid, 1549 u16 flags, 1550 bool *notified, 1551 struct netlink_ext_ack *extack); 1552 int (*ndo_fdb_del)(struct ndmsg *ndm, 1553 struct nlattr *tb[], 1554 struct net_device *dev, 1555 const unsigned char *addr, 1556 u16 vid, 1557 bool *notified, 1558 struct netlink_ext_ack *extack); 1559 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, 1560 struct net_device *dev, 1561 struct netlink_ext_ack *extack); 1562 int (*ndo_fdb_dump)(struct sk_buff *skb, 1563 struct netlink_callback *cb, 1564 struct net_device *dev, 1565 struct net_device *filter_dev, 1566 int *idx); 1567 int (*ndo_fdb_get)(struct sk_buff *skb, 1568 struct nlattr *tb[], 1569 struct net_device *dev, 1570 const unsigned char *addr, 1571 u16 vid, u32 portid, u32 seq, 1572 struct netlink_ext_ack *extack); 1573 int (*ndo_mdb_add)(struct net_device *dev, 1574 struct nlattr *tb[], 1575 u16 nlmsg_flags, 1576 struct netlink_ext_ack *extack); 1577 int (*ndo_mdb_del)(struct net_device *dev, 1578 struct nlattr *tb[], 1579 struct netlink_ext_ack *extack); 1580 int (*ndo_mdb_del_bulk)(struct net_device *dev, 1581 struct nlattr *tb[], 1582 struct netlink_ext_ack *extack); 1583 int (*ndo_mdb_dump)(struct net_device *dev, 1584 struct sk_buff *skb, 1585 struct netlink_callback *cb); 1586 int (*ndo_mdb_get)(struct net_device *dev, 1587 struct nlattr *tb[], u32 portid, 1588 u32 seq, 1589 struct netlink_ext_ack *extack); 1590 int (*ndo_bridge_setlink)(struct net_device *dev, 1591 struct nlmsghdr *nlh, 1592 u16 flags, 1593 struct netlink_ext_ack *extack); 1594 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1595 u32 pid, u32 seq, 1596 struct net_device *dev, 1597 u32 filter_mask, 1598 int nlflags); 1599 int (*ndo_bridge_dellink)(struct net_device *dev, 1600 struct nlmsghdr *nlh, 1601 u16 flags); 1602 int (*ndo_change_carrier)(struct net_device *dev, 1603 bool new_carrier); 1604 int (*ndo_get_phys_port_id)(struct net_device *dev, 1605 struct netdev_phys_item_id *ppid); 1606 int (*ndo_get_port_parent_id)(struct net_device *dev, 1607 struct netdev_phys_item_id *ppid); 1608 int (*ndo_get_phys_port_name)(struct net_device *dev, 1609 char *name, size_t len); 1610 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1611 struct net_device *dev); 1612 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1613 void *priv); 1614 1615 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1616 int queue_index, 1617 u32 maxrate); 1618 int (*ndo_get_iflink)(const struct net_device *dev); 1619 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1620 struct sk_buff *skb); 1621 void (*ndo_set_rx_headroom)(struct net_device *dev, 1622 int needed_headroom); 1623 int (*ndo_bpf)(struct net_device *dev, 1624 struct netdev_bpf *bpf); 1625 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1626 struct xdp_frame **xdp, 1627 u32 flags); 1628 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1629 struct xdp_buff *xdp); 1630 int (*ndo_xsk_wakeup)(struct net_device *dev, 1631 u32 queue_id, u32 flags); 1632 int (*ndo_tunnel_ctl)(struct net_device *dev, 1633 struct ip_tunnel_parm_kern *p, 1634 int cmd); 1635 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1636 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1637 struct net_device_path *path); 1638 ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1639 const struct skb_shared_hwtstamps *hwtstamps, 1640 bool cycles); 1641 int (*ndo_hwtstamp_get)(struct net_device *dev, 1642 struct kernel_hwtstamp_config *kernel_config); 1643 int (*ndo_hwtstamp_set)(struct net_device *dev, 1644 struct kernel_hwtstamp_config *kernel_config, 1645 struct netlink_ext_ack *extack); 1646 1647 #if IS_ENABLED(CONFIG_NET_SHAPER) 1648 /** 1649 * @net_shaper_ops: Device shaping offload operations 1650 * see include/net/net_shapers.h 1651 */ 1652 const struct net_shaper_ops *net_shaper_ops; 1653 #endif 1654 }; 1655 1656 /** 1657 * enum netdev_priv_flags - &struct net_device priv_flags 1658 * 1659 * These are the &struct net_device, they are only set internally 1660 * by drivers and used in the kernel. These flags are invisible to 1661 * userspace; this means that the order of these flags can change 1662 * during any kernel release. 1663 * 1664 * You should add bitfield booleans after either net_device::priv_flags 1665 * (hotpath) or ::threaded (slowpath) instead of extending these flags. 1666 * 1667 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1668 * @IFF_EBRIDGE: Ethernet bridging device 1669 * @IFF_BONDING: bonding master or slave 1670 * @IFF_ISATAP: ISATAP interface (RFC4214) 1671 * @IFF_WAN_HDLC: WAN HDLC device 1672 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1673 * release skb->dst 1674 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1675 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1676 * @IFF_MACVLAN_PORT: device used as macvlan port 1677 * @IFF_BRIDGE_PORT: device used as bridge port 1678 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1679 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1680 * @IFF_UNICAST_FLT: Supports unicast filtering 1681 * @IFF_TEAM_PORT: device used as team port 1682 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1683 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1684 * change when it's running 1685 * @IFF_MACVLAN: Macvlan device 1686 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1687 * underlying stacked devices 1688 * @IFF_L3MDEV_MASTER: device is an L3 master device 1689 * @IFF_NO_QUEUE: device can run without qdisc attached 1690 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1691 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1692 * @IFF_TEAM: device is a team device 1693 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1694 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1695 * entity (i.e. the master device for bridged veth) 1696 * @IFF_MACSEC: device is a MACsec device 1697 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1698 * @IFF_FAILOVER: device is a failover master device 1699 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1700 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1701 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf 1702 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1703 * skb_headlen(skb) == 0 (data starts from frag0) 1704 */ 1705 enum netdev_priv_flags { 1706 IFF_802_1Q_VLAN = 1<<0, 1707 IFF_EBRIDGE = 1<<1, 1708 IFF_BONDING = 1<<2, 1709 IFF_ISATAP = 1<<3, 1710 IFF_WAN_HDLC = 1<<4, 1711 IFF_XMIT_DST_RELEASE = 1<<5, 1712 IFF_DONT_BRIDGE = 1<<6, 1713 IFF_DISABLE_NETPOLL = 1<<7, 1714 IFF_MACVLAN_PORT = 1<<8, 1715 IFF_BRIDGE_PORT = 1<<9, 1716 IFF_OVS_DATAPATH = 1<<10, 1717 IFF_TX_SKB_SHARING = 1<<11, 1718 IFF_UNICAST_FLT = 1<<12, 1719 IFF_TEAM_PORT = 1<<13, 1720 IFF_SUPP_NOFCS = 1<<14, 1721 IFF_LIVE_ADDR_CHANGE = 1<<15, 1722 IFF_MACVLAN = 1<<16, 1723 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1724 IFF_L3MDEV_MASTER = 1<<18, 1725 IFF_NO_QUEUE = 1<<19, 1726 IFF_OPENVSWITCH = 1<<20, 1727 IFF_L3MDEV_SLAVE = 1<<21, 1728 IFF_TEAM = 1<<22, 1729 IFF_RXFH_CONFIGURED = 1<<23, 1730 IFF_PHONY_HEADROOM = 1<<24, 1731 IFF_MACSEC = 1<<25, 1732 IFF_NO_RX_HANDLER = 1<<26, 1733 IFF_FAILOVER = 1<<27, 1734 IFF_FAILOVER_SLAVE = 1<<28, 1735 IFF_L3MDEV_RX_HANDLER = 1<<29, 1736 IFF_NO_ADDRCONF = BIT_ULL(30), 1737 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31), 1738 }; 1739 1740 /* Specifies the type of the struct net_device::ml_priv pointer */ 1741 enum netdev_ml_priv_type { 1742 ML_PRIV_NONE, 1743 ML_PRIV_CAN, 1744 }; 1745 1746 enum netdev_stat_type { 1747 NETDEV_PCPU_STAT_NONE, 1748 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */ 1749 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */ 1750 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */ 1751 }; 1752 1753 enum netdev_reg_state { 1754 NETREG_UNINITIALIZED = 0, 1755 NETREG_REGISTERED, /* completed register_netdevice */ 1756 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1757 NETREG_UNREGISTERED, /* completed unregister todo */ 1758 NETREG_RELEASED, /* called free_netdev */ 1759 NETREG_DUMMY, /* dummy device for NAPI poll */ 1760 }; 1761 1762 /** 1763 * struct net_device - The DEVICE structure. 1764 * 1765 * Actually, this whole structure is a big mistake. It mixes I/O 1766 * data with strictly "high-level" data, and it has to know about 1767 * almost every data structure used in the INET module. 1768 * 1769 * @priv_flags: flags invisible to userspace defined as bits, see 1770 * enum netdev_priv_flags for the definitions 1771 * @lltx: device supports lockless Tx. Deprecated for real HW 1772 * drivers. Mainly used by logical interfaces, such as 1773 * bonding and tunnels 1774 * 1775 * @name: This is the first field of the "visible" part of this structure 1776 * (i.e. as seen by users in the "Space.c" file). It is the name 1777 * of the interface. 1778 * 1779 * @name_node: Name hashlist node 1780 * @ifalias: SNMP alias 1781 * @mem_end: Shared memory end 1782 * @mem_start: Shared memory start 1783 * @base_addr: Device I/O address 1784 * @irq: Device IRQ number 1785 * 1786 * @state: Generic network queuing layer state, see netdev_state_t 1787 * @dev_list: The global list of network devices 1788 * @napi_list: List entry used for polling NAPI devices 1789 * @unreg_list: List entry when we are unregistering the 1790 * device; see the function unregister_netdev 1791 * @close_list: List entry used when we are closing the device 1792 * @ptype_all: Device-specific packet handlers for all protocols 1793 * @ptype_specific: Device-specific, protocol-specific packet handlers 1794 * 1795 * @adj_list: Directly linked devices, like slaves for bonding 1796 * @features: Currently active device features 1797 * @hw_features: User-changeable features 1798 * 1799 * @wanted_features: User-requested features 1800 * @vlan_features: Mask of features inheritable by VLAN devices 1801 * 1802 * @hw_enc_features: Mask of features inherited by encapsulating devices 1803 * This field indicates what encapsulation 1804 * offloads the hardware is capable of doing, 1805 * and drivers will need to set them appropriately. 1806 * 1807 * @mpls_features: Mask of features inheritable by MPLS 1808 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1809 * 1810 * @ifindex: interface index 1811 * @group: The group the device belongs to 1812 * 1813 * @stats: Statistics struct, which was left as a legacy, use 1814 * rtnl_link_stats64 instead 1815 * 1816 * @core_stats: core networking counters, 1817 * do not use this in drivers 1818 * @carrier_up_count: Number of times the carrier has been up 1819 * @carrier_down_count: Number of times the carrier has been down 1820 * 1821 * @wireless_handlers: List of functions to handle Wireless Extensions, 1822 * instead of ioctl, 1823 * see <net/iw_handler.h> for details. 1824 * 1825 * @netdev_ops: Includes several pointers to callbacks, 1826 * if one wants to override the ndo_*() functions 1827 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks. 1828 * @xsk_tx_metadata_ops: Includes pointers to AF_XDP TX metadata callbacks. 1829 * @ethtool_ops: Management operations 1830 * @l3mdev_ops: Layer 3 master device operations 1831 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1832 * discovery handling. Necessary for e.g. 6LoWPAN. 1833 * @xfrmdev_ops: Transformation offload operations 1834 * @tlsdev_ops: Transport Layer Security offload operations 1835 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1836 * of Layer 2 headers. 1837 * 1838 * @flags: Interface flags (a la BSD) 1839 * @xdp_features: XDP capability supported by the device 1840 * @gflags: Global flags ( kept as legacy ) 1841 * @priv_len: Size of the ->priv flexible array 1842 * @priv: Flexible array containing private data 1843 * @operstate: RFC2863 operstate 1844 * @link_mode: Mapping policy to operstate 1845 * @if_port: Selectable AUI, TP, ... 1846 * @dma: DMA channel 1847 * @mtu: Interface MTU value 1848 * @min_mtu: Interface Minimum MTU value 1849 * @max_mtu: Interface Maximum MTU value 1850 * @type: Interface hardware type 1851 * @hard_header_len: Maximum hardware header length. 1852 * @min_header_len: Minimum hardware header length 1853 * 1854 * @needed_headroom: Extra headroom the hardware may need, but not in all 1855 * cases can this be guaranteed 1856 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1857 * cases can this be guaranteed. Some cases also use 1858 * LL_MAX_HEADER instead to allocate the skb 1859 * 1860 * interface address info: 1861 * 1862 * @perm_addr: Permanent hw address 1863 * @addr_assign_type: Hw address assignment type 1864 * @addr_len: Hardware address length 1865 * @upper_level: Maximum depth level of upper devices. 1866 * @lower_level: Maximum depth level of lower devices. 1867 * @neigh_priv_len: Used in neigh_alloc() 1868 * @dev_id: Used to differentiate devices that share 1869 * the same link layer address 1870 * @dev_port: Used to differentiate devices that share 1871 * the same function 1872 * @addr_list_lock: XXX: need comments on this one 1873 * @name_assign_type: network interface name assignment type 1874 * @uc_promisc: Counter that indicates promiscuous mode 1875 * has been enabled due to the need to listen to 1876 * additional unicast addresses in a device that 1877 * does not implement ndo_set_rx_mode() 1878 * @uc: unicast mac addresses 1879 * @mc: multicast mac addresses 1880 * @dev_addrs: list of device hw addresses 1881 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1882 * @promiscuity: Number of times the NIC is told to work in 1883 * promiscuous mode; if it becomes 0 the NIC will 1884 * exit promiscuous mode 1885 * @allmulti: Counter, enables or disables allmulticast mode 1886 * 1887 * @vlan_info: VLAN info 1888 * @dsa_ptr: dsa specific data 1889 * @tipc_ptr: TIPC specific data 1890 * @atalk_ptr: AppleTalk link 1891 * @ip_ptr: IPv4 specific data 1892 * @ip6_ptr: IPv6 specific data 1893 * @ax25_ptr: AX.25 specific data 1894 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1895 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1896 * device struct 1897 * @mpls_ptr: mpls_dev struct pointer 1898 * @mctp_ptr: MCTP specific data 1899 * 1900 * @dev_addr: Hw address (before bcast, 1901 * because most packets are unicast) 1902 * 1903 * @_rx: Array of RX queues 1904 * @num_rx_queues: Number of RX queues 1905 * allocated at register_netdev() time 1906 * @real_num_rx_queues: Number of RX queues currently active in device 1907 * @xdp_prog: XDP sockets filter program pointer 1908 * 1909 * @rx_handler: handler for received packets 1910 * @rx_handler_data: XXX: need comments on this one 1911 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing 1912 * @ingress_queue: XXX: need comments on this one 1913 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1914 * @broadcast: hw bcast address 1915 * 1916 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1917 * indexed by RX queue number. Assigned by driver. 1918 * This must only be set if the ndo_rx_flow_steer 1919 * operation is defined 1920 * @index_hlist: Device index hash chain 1921 * 1922 * @_tx: Array of TX queues 1923 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1924 * @real_num_tx_queues: Number of TX queues currently active in device 1925 * @qdisc: Root qdisc from userspace point of view 1926 * @tx_queue_len: Max frames per queue allowed 1927 * @tx_global_lock: XXX: need comments on this one 1928 * @xdp_bulkq: XDP device bulk queue 1929 * @xps_maps: all CPUs/RXQs maps for XPS device 1930 * 1931 * @xps_maps: XXX: need comments on this one 1932 * @tcx_egress: BPF & clsact qdisc specific data for egress processing 1933 * @nf_hooks_egress: netfilter hooks executed for egress packets 1934 * @qdisc_hash: qdisc hash table 1935 * @watchdog_timeo: Represents the timeout that is used by 1936 * the watchdog (see dev_watchdog()) 1937 * @watchdog_timer: List of timers 1938 * 1939 * @proto_down_reason: reason a netdev interface is held down 1940 * @pcpu_refcnt: Number of references to this device 1941 * @dev_refcnt: Number of references to this device 1942 * @refcnt_tracker: Tracker directory for tracked references to this device 1943 * @todo_list: Delayed register/unregister 1944 * @link_watch_list: XXX: need comments on this one 1945 * 1946 * @reg_state: Register/unregister state machine 1947 * @dismantle: Device is going to be freed 1948 * @rtnl_link_state: This enum represents the phases of creating 1949 * a new link 1950 * 1951 * @needs_free_netdev: Should unregister perform free_netdev? 1952 * @priv_destructor: Called from unregister 1953 * @npinfo: XXX: need comments on this one 1954 * @nd_net: Network namespace this network device is inside 1955 * 1956 * @ml_priv: Mid-layer private 1957 * @ml_priv_type: Mid-layer private type 1958 * 1959 * @pcpu_stat_type: Type of device statistics which the core should 1960 * allocate/free: none, lstats, tstats, dstats. none 1961 * means the driver is handling statistics allocation/ 1962 * freeing internally. 1963 * @lstats: Loopback statistics: packets, bytes 1964 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes 1965 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes 1966 * 1967 * @garp_port: GARP 1968 * @mrp_port: MRP 1969 * 1970 * @dm_private: Drop monitor private 1971 * 1972 * @dev: Class/net/name entry 1973 * @sysfs_groups: Space for optional device, statistics and wireless 1974 * sysfs groups 1975 * 1976 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1977 * @rtnl_link_ops: Rtnl_link_ops 1978 * @stat_ops: Optional ops for queue-aware statistics 1979 * @queue_mgmt_ops: Optional ops for queue management 1980 * 1981 * @gso_max_size: Maximum size of generic segmentation offload 1982 * @tso_max_size: Device (as in HW) limit on the max TSO request size 1983 * @gso_max_segs: Maximum number of segments that can be passed to the 1984 * NIC for GSO 1985 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count 1986 * @gso_ipv4_max_size: Maximum size of generic segmentation offload, 1987 * for IPv4. 1988 * 1989 * @dcbnl_ops: Data Center Bridging netlink ops 1990 * @num_tc: Number of traffic classes in the net device 1991 * @tc_to_txq: XXX: need comments on this one 1992 * @prio_tc_map: XXX: need comments on this one 1993 * 1994 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1995 * 1996 * @priomap: XXX: need comments on this one 1997 * @link_topo: Physical link topology tracking attached PHYs 1998 * @phydev: Physical device may attach itself 1999 * for hardware timestamping 2000 * @sfp_bus: attached &struct sfp_bus structure. 2001 * 2002 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 2003 * 2004 * @proto_down: protocol port state information can be sent to the 2005 * switch driver and used to set the phys state of the 2006 * switch port. 2007 * 2008 * @threaded: napi threaded mode is enabled 2009 * 2010 * @irq_affinity_auto: driver wants the core to store and re-assign the IRQ 2011 * affinity. Set by netif_enable_irq_affinity(), then 2012 * the driver must create a persistent napi by 2013 * netif_napi_add_config() and finally bind the napi to 2014 * IRQ (via netif_napi_set_irq()). 2015 * 2016 * @rx_cpu_rmap_auto: driver wants the core to manage the ARFS rmap. 2017 * Set by calling netif_enable_cpu_rmap(). 2018 * 2019 * @see_all_hwtstamp_requests: device wants to see calls to 2020 * ndo_hwtstamp_set() for all timestamp requests 2021 * regardless of source, even if those aren't 2022 * HWTSTAMP_SOURCE_NETDEV 2023 * @change_proto_down: device supports setting carrier via IFLA_PROTO_DOWN 2024 * @netns_immutable: interface can't change network namespaces 2025 * @fcoe_mtu: device supports maximum FCoE MTU, 2158 bytes 2026 * 2027 * @net_notifier_list: List of per-net netdev notifier block 2028 * that follow this device when it is moved 2029 * to another network namespace. 2030 * 2031 * @macsec_ops: MACsec offloading ops 2032 * 2033 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 2034 * offload capabilities of the device 2035 * @udp_tunnel_nic: UDP tunnel offload state 2036 * @ethtool: ethtool related state 2037 * @xdp_state: stores info on attached XDP BPF programs 2038 * 2039 * @nested_level: Used as a parameter of spin_lock_nested() of 2040 * dev->addr_list_lock. 2041 * @unlink_list: As netif_addr_lock() can be called recursively, 2042 * keep a list of interfaces to be deleted. 2043 * @gro_max_size: Maximum size of aggregated packet in generic 2044 * receive offload (GRO) 2045 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic 2046 * receive offload (GRO), for IPv4. 2047 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP 2048 * zero copy driver 2049 * 2050 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 2051 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 2052 * @watchdog_dev_tracker: refcount tracker used by watchdog. 2053 * @dev_registered_tracker: tracker for reference held while 2054 * registered 2055 * @offload_xstats_l3: L3 HW stats for this netdevice. 2056 * 2057 * @devlink_port: Pointer to related devlink port structure. 2058 * Assigned by a driver before netdev registration using 2059 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static 2060 * during the time netdevice is registered. 2061 * 2062 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem, 2063 * where the clock is recovered. 2064 * 2065 * @max_pacing_offload_horizon: max EDT offload horizon in nsec. 2066 * @napi_config: An array of napi_config structures containing per-NAPI 2067 * settings. 2068 * @gro_flush_timeout: timeout for GRO layer in NAPI 2069 * @napi_defer_hard_irqs: If not zero, provides a counter that would 2070 * allow to avoid NIC hard IRQ, on busy queues. 2071 * 2072 * @neighbours: List heads pointing to this device's neighbours' 2073 * dev_list, one per address-family. 2074 * @hwprov: Tracks which PTP performs hardware packet time stamping. 2075 * 2076 * FIXME: cleanup struct net_device such that network protocol info 2077 * moves out. 2078 */ 2079 2080 struct net_device { 2081 /* Cacheline organization can be found documented in 2082 * Documentation/networking/net_cachelines/net_device.rst. 2083 * Please update the document when adding new fields. 2084 */ 2085 2086 /* TX read-mostly hotpath */ 2087 __cacheline_group_begin(net_device_read_tx); 2088 struct_group(priv_flags_fast, 2089 unsigned long priv_flags:32; 2090 unsigned long lltx:1; 2091 ); 2092 const struct net_device_ops *netdev_ops; 2093 const struct header_ops *header_ops; 2094 struct netdev_queue *_tx; 2095 netdev_features_t gso_partial_features; 2096 unsigned int real_num_tx_queues; 2097 unsigned int gso_max_size; 2098 unsigned int gso_ipv4_max_size; 2099 u16 gso_max_segs; 2100 s16 num_tc; 2101 /* Note : dev->mtu is often read without holding a lock. 2102 * Writers usually hold RTNL. 2103 * It is recommended to use READ_ONCE() to annotate the reads, 2104 * and to use WRITE_ONCE() to annotate the writes. 2105 */ 2106 unsigned int mtu; 2107 unsigned short needed_headroom; 2108 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2109 #ifdef CONFIG_XPS 2110 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2111 #endif 2112 #ifdef CONFIG_NETFILTER_EGRESS 2113 struct nf_hook_entries __rcu *nf_hooks_egress; 2114 #endif 2115 #ifdef CONFIG_NET_XGRESS 2116 struct bpf_mprog_entry __rcu *tcx_egress; 2117 #endif 2118 __cacheline_group_end(net_device_read_tx); 2119 2120 /* TXRX read-mostly hotpath */ 2121 __cacheline_group_begin(net_device_read_txrx); 2122 union { 2123 struct pcpu_lstats __percpu *lstats; 2124 struct pcpu_sw_netstats __percpu *tstats; 2125 struct pcpu_dstats __percpu *dstats; 2126 }; 2127 unsigned long state; 2128 unsigned int flags; 2129 unsigned short hard_header_len; 2130 netdev_features_t features; 2131 struct inet6_dev __rcu *ip6_ptr; 2132 __cacheline_group_end(net_device_read_txrx); 2133 2134 /* RX read-mostly hotpath */ 2135 __cacheline_group_begin(net_device_read_rx); 2136 struct bpf_prog __rcu *xdp_prog; 2137 struct list_head ptype_specific; 2138 int ifindex; 2139 unsigned int real_num_rx_queues; 2140 struct netdev_rx_queue *_rx; 2141 unsigned int gro_max_size; 2142 unsigned int gro_ipv4_max_size; 2143 rx_handler_func_t __rcu *rx_handler; 2144 void __rcu *rx_handler_data; 2145 possible_net_t nd_net; 2146 #ifdef CONFIG_NETPOLL 2147 struct netpoll_info __rcu *npinfo; 2148 #endif 2149 #ifdef CONFIG_NET_XGRESS 2150 struct bpf_mprog_entry __rcu *tcx_ingress; 2151 #endif 2152 __cacheline_group_end(net_device_read_rx); 2153 2154 char name[IFNAMSIZ]; 2155 struct netdev_name_node *name_node; 2156 struct dev_ifalias __rcu *ifalias; 2157 /* 2158 * I/O specific fields 2159 * FIXME: Merge these and struct ifmap into one 2160 */ 2161 unsigned long mem_end; 2162 unsigned long mem_start; 2163 unsigned long base_addr; 2164 2165 /* 2166 * Some hardware also needs these fields (state,dev_list, 2167 * napi_list,unreg_list,close_list) but they are not 2168 * part of the usual set specified in Space.c. 2169 */ 2170 2171 2172 struct list_head dev_list; 2173 struct list_head napi_list; 2174 struct list_head unreg_list; 2175 struct list_head close_list; 2176 struct list_head ptype_all; 2177 2178 struct { 2179 struct list_head upper; 2180 struct list_head lower; 2181 } adj_list; 2182 2183 /* Read-mostly cache-line for fast-path access */ 2184 xdp_features_t xdp_features; 2185 const struct xdp_metadata_ops *xdp_metadata_ops; 2186 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops; 2187 unsigned short gflags; 2188 2189 unsigned short needed_tailroom; 2190 2191 netdev_features_t hw_features; 2192 netdev_features_t wanted_features; 2193 netdev_features_t vlan_features; 2194 netdev_features_t hw_enc_features; 2195 netdev_features_t mpls_features; 2196 2197 unsigned int min_mtu; 2198 unsigned int max_mtu; 2199 unsigned short type; 2200 unsigned char min_header_len; 2201 unsigned char name_assign_type; 2202 2203 int group; 2204 2205 struct net_device_stats stats; /* not used by modern drivers */ 2206 2207 struct net_device_core_stats __percpu *core_stats; 2208 2209 /* Stats to monitor link on/off, flapping */ 2210 atomic_t carrier_up_count; 2211 atomic_t carrier_down_count; 2212 2213 #ifdef CONFIG_WIRELESS_EXT 2214 const struct iw_handler_def *wireless_handlers; 2215 #endif 2216 const struct ethtool_ops *ethtool_ops; 2217 #ifdef CONFIG_NET_L3_MASTER_DEV 2218 const struct l3mdev_ops *l3mdev_ops; 2219 #endif 2220 #if IS_ENABLED(CONFIG_IPV6) 2221 const struct ndisc_ops *ndisc_ops; 2222 #endif 2223 2224 #ifdef CONFIG_XFRM_OFFLOAD 2225 const struct xfrmdev_ops *xfrmdev_ops; 2226 #endif 2227 2228 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2229 const struct tlsdev_ops *tlsdev_ops; 2230 #endif 2231 2232 unsigned int operstate; 2233 unsigned char link_mode; 2234 2235 unsigned char if_port; 2236 unsigned char dma; 2237 2238 /* Interface address info. */ 2239 unsigned char perm_addr[MAX_ADDR_LEN]; 2240 unsigned char addr_assign_type; 2241 unsigned char addr_len; 2242 unsigned char upper_level; 2243 unsigned char lower_level; 2244 2245 unsigned short neigh_priv_len; 2246 unsigned short dev_id; 2247 unsigned short dev_port; 2248 int irq; 2249 u32 priv_len; 2250 2251 spinlock_t addr_list_lock; 2252 2253 struct netdev_hw_addr_list uc; 2254 struct netdev_hw_addr_list mc; 2255 struct netdev_hw_addr_list dev_addrs; 2256 2257 #ifdef CONFIG_SYSFS 2258 struct kset *queues_kset; 2259 #endif 2260 #ifdef CONFIG_LOCKDEP 2261 struct list_head unlink_list; 2262 #endif 2263 unsigned int promiscuity; 2264 unsigned int allmulti; 2265 bool uc_promisc; 2266 #ifdef CONFIG_LOCKDEP 2267 unsigned char nested_level; 2268 #endif 2269 2270 2271 /* Protocol-specific pointers */ 2272 struct in_device __rcu *ip_ptr; 2273 /** @fib_nh_head: nexthops associated with this netdev */ 2274 struct hlist_head fib_nh_head; 2275 2276 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2277 struct vlan_info __rcu *vlan_info; 2278 #endif 2279 #if IS_ENABLED(CONFIG_NET_DSA) 2280 struct dsa_port *dsa_ptr; 2281 #endif 2282 #if IS_ENABLED(CONFIG_TIPC) 2283 struct tipc_bearer __rcu *tipc_ptr; 2284 #endif 2285 #if IS_ENABLED(CONFIG_ATALK) 2286 void *atalk_ptr; 2287 #endif 2288 #if IS_ENABLED(CONFIG_AX25) 2289 struct ax25_dev __rcu *ax25_ptr; 2290 #endif 2291 #if IS_ENABLED(CONFIG_CFG80211) 2292 struct wireless_dev *ieee80211_ptr; 2293 #endif 2294 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2295 struct wpan_dev *ieee802154_ptr; 2296 #endif 2297 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2298 struct mpls_dev __rcu *mpls_ptr; 2299 #endif 2300 #if IS_ENABLED(CONFIG_MCTP) 2301 struct mctp_dev __rcu *mctp_ptr; 2302 #endif 2303 2304 /* 2305 * Cache lines mostly used on receive path (including eth_type_trans()) 2306 */ 2307 /* Interface address info used in eth_type_trans() */ 2308 const unsigned char *dev_addr; 2309 2310 unsigned int num_rx_queues; 2311 #define GRO_LEGACY_MAX_SIZE 65536u 2312 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2313 * and shinfo->gso_segs is a 16bit field. 2314 */ 2315 #define GRO_MAX_SIZE (8 * 65535u) 2316 unsigned int xdp_zc_max_segs; 2317 struct netdev_queue __rcu *ingress_queue; 2318 #ifdef CONFIG_NETFILTER_INGRESS 2319 struct nf_hook_entries __rcu *nf_hooks_ingress; 2320 #endif 2321 2322 unsigned char broadcast[MAX_ADDR_LEN]; 2323 #ifdef CONFIG_RFS_ACCEL 2324 struct cpu_rmap *rx_cpu_rmap; 2325 #endif 2326 struct hlist_node index_hlist; 2327 2328 /* 2329 * Cache lines mostly used on transmit path 2330 */ 2331 unsigned int num_tx_queues; 2332 struct Qdisc __rcu *qdisc; 2333 unsigned int tx_queue_len; 2334 spinlock_t tx_global_lock; 2335 2336 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2337 2338 #ifdef CONFIG_NET_SCHED 2339 DECLARE_HASHTABLE (qdisc_hash, 4); 2340 #endif 2341 /* These may be needed for future network-power-down code. */ 2342 struct timer_list watchdog_timer; 2343 int watchdog_timeo; 2344 2345 u32 proto_down_reason; 2346 2347 struct list_head todo_list; 2348 2349 #ifdef CONFIG_PCPU_DEV_REFCNT 2350 int __percpu *pcpu_refcnt; 2351 #else 2352 refcount_t dev_refcnt; 2353 #endif 2354 struct ref_tracker_dir refcnt_tracker; 2355 2356 struct list_head link_watch_list; 2357 2358 u8 reg_state; 2359 2360 bool dismantle; 2361 2362 enum { 2363 RTNL_LINK_INITIALIZED, 2364 RTNL_LINK_INITIALIZING, 2365 } rtnl_link_state:16; 2366 2367 bool needs_free_netdev; 2368 void (*priv_destructor)(struct net_device *dev); 2369 2370 /* mid-layer private */ 2371 void *ml_priv; 2372 enum netdev_ml_priv_type ml_priv_type; 2373 2374 enum netdev_stat_type pcpu_stat_type:8; 2375 2376 #if IS_ENABLED(CONFIG_GARP) 2377 struct garp_port __rcu *garp_port; 2378 #endif 2379 #if IS_ENABLED(CONFIG_MRP) 2380 struct mrp_port __rcu *mrp_port; 2381 #endif 2382 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2383 struct dm_hw_stat_delta __rcu *dm_private; 2384 #endif 2385 struct device dev; 2386 const struct attribute_group *sysfs_groups[4]; 2387 const struct attribute_group *sysfs_rx_queue_group; 2388 2389 const struct rtnl_link_ops *rtnl_link_ops; 2390 2391 const struct netdev_stat_ops *stat_ops; 2392 2393 const struct netdev_queue_mgmt_ops *queue_mgmt_ops; 2394 2395 /* for setting kernel sock attribute on TCP connection setup */ 2396 #define GSO_MAX_SEGS 65535u 2397 #define GSO_LEGACY_MAX_SIZE 65536u 2398 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2399 * and shinfo->gso_segs is a 16bit field. 2400 */ 2401 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2402 2403 #define TSO_LEGACY_MAX_SIZE 65536 2404 #define TSO_MAX_SIZE UINT_MAX 2405 unsigned int tso_max_size; 2406 #define TSO_MAX_SEGS U16_MAX 2407 u16 tso_max_segs; 2408 2409 #ifdef CONFIG_DCB 2410 const struct dcbnl_rtnl_ops *dcbnl_ops; 2411 #endif 2412 u8 prio_tc_map[TC_BITMASK + 1]; 2413 2414 #if IS_ENABLED(CONFIG_FCOE) 2415 unsigned int fcoe_ddp_xid; 2416 #endif 2417 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2418 struct netprio_map __rcu *priomap; 2419 #endif 2420 struct phy_link_topology *link_topo; 2421 struct phy_device *phydev; 2422 struct sfp_bus *sfp_bus; 2423 struct lock_class_key *qdisc_tx_busylock; 2424 bool proto_down; 2425 bool threaded; 2426 bool irq_affinity_auto; 2427 bool rx_cpu_rmap_auto; 2428 2429 /* priv_flags_slow, ungrouped to save space */ 2430 unsigned long see_all_hwtstamp_requests:1; 2431 unsigned long change_proto_down:1; 2432 unsigned long netns_immutable:1; 2433 unsigned long fcoe_mtu:1; 2434 2435 struct list_head net_notifier_list; 2436 2437 #if IS_ENABLED(CONFIG_MACSEC) 2438 /* MACsec management functions */ 2439 const struct macsec_ops *macsec_ops; 2440 #endif 2441 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2442 struct udp_tunnel_nic *udp_tunnel_nic; 2443 2444 /** @cfg: net_device queue-related configuration */ 2445 struct netdev_config *cfg; 2446 /** 2447 * @cfg_pending: same as @cfg but when device is being actively 2448 * reconfigured includes any changes to the configuration 2449 * requested by the user, but which may or may not be rejected. 2450 */ 2451 struct netdev_config *cfg_pending; 2452 struct ethtool_netdev_state *ethtool; 2453 2454 /* protected by rtnl_lock */ 2455 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2456 2457 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2458 netdevice_tracker linkwatch_dev_tracker; 2459 netdevice_tracker watchdog_dev_tracker; 2460 netdevice_tracker dev_registered_tracker; 2461 struct rtnl_hw_stats64 *offload_xstats_l3; 2462 2463 struct devlink_port *devlink_port; 2464 2465 #if IS_ENABLED(CONFIG_DPLL) 2466 struct dpll_pin __rcu *dpll_pin; 2467 #endif 2468 #if IS_ENABLED(CONFIG_PAGE_POOL) 2469 /** @page_pools: page pools created for this netdevice */ 2470 struct hlist_head page_pools; 2471 #endif 2472 2473 /** @irq_moder: dim parameters used if IS_ENABLED(CONFIG_DIMLIB). */ 2474 struct dim_irq_moder *irq_moder; 2475 2476 u64 max_pacing_offload_horizon; 2477 struct napi_config *napi_config; 2478 unsigned long gro_flush_timeout; 2479 u32 napi_defer_hard_irqs; 2480 2481 /** 2482 * @up: copy of @state's IFF_UP, but safe to read with just @lock. 2483 * May report false negatives while the device is being opened 2484 * or closed (@lock does not protect .ndo_open, or .ndo_close). 2485 */ 2486 bool up; 2487 2488 /** 2489 * @request_ops_lock: request the core to run all @netdev_ops and 2490 * @ethtool_ops under the @lock. 2491 */ 2492 bool request_ops_lock; 2493 2494 /** 2495 * @lock: netdev-scope lock, protects a small selection of fields. 2496 * Should always be taken using netdev_lock() / netdev_unlock() helpers. 2497 * Drivers are free to use it for other protection. 2498 * 2499 * Protects: 2500 * @gro_flush_timeout, @napi_defer_hard_irqs, @napi_list, 2501 * @net_shaper_hierarchy, @reg_state, @threaded, @dev_addr 2502 * 2503 * Partially protects (writers must hold both @lock and rtnl_lock): 2504 * @up 2505 * 2506 * Also protects some fields in struct napi_struct. 2507 * 2508 * Ordering: take after rtnl_lock. 2509 */ 2510 struct mutex lock; 2511 2512 #if IS_ENABLED(CONFIG_NET_SHAPER) 2513 /** 2514 * @net_shaper_hierarchy: data tracking the current shaper status 2515 * see include/net/net_shapers.h 2516 */ 2517 struct net_shaper_hierarchy *net_shaper_hierarchy; 2518 #endif 2519 2520 struct hlist_head neighbours[NEIGH_NR_TABLES]; 2521 2522 struct hwtstamp_provider __rcu *hwprov; 2523 2524 u8 priv[] ____cacheline_aligned 2525 __counted_by(priv_len); 2526 } ____cacheline_aligned; 2527 #define to_net_dev(d) container_of(d, struct net_device, dev) 2528 2529 /* 2530 * Driver should use this to assign devlink port instance to a netdevice 2531 * before it registers the netdevice. Therefore devlink_port is static 2532 * during the netdev lifetime after it is registered. 2533 */ 2534 #define SET_NETDEV_DEVLINK_PORT(dev, port) \ 2535 ({ \ 2536 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \ 2537 ((dev)->devlink_port = (port)); \ 2538 }) 2539 2540 static inline bool netif_elide_gro(const struct net_device *dev) 2541 { 2542 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2543 return true; 2544 return false; 2545 } 2546 2547 #define NETDEV_ALIGN 32 2548 2549 static inline 2550 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2551 { 2552 return dev->prio_tc_map[prio & TC_BITMASK]; 2553 } 2554 2555 static inline 2556 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2557 { 2558 if (tc >= dev->num_tc) 2559 return -EINVAL; 2560 2561 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2562 return 0; 2563 } 2564 2565 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2566 void netdev_reset_tc(struct net_device *dev); 2567 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2568 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2569 2570 static inline 2571 int netdev_get_num_tc(struct net_device *dev) 2572 { 2573 return dev->num_tc; 2574 } 2575 2576 static inline void net_prefetch(void *p) 2577 { 2578 prefetch(p); 2579 #if L1_CACHE_BYTES < 128 2580 prefetch((u8 *)p + L1_CACHE_BYTES); 2581 #endif 2582 } 2583 2584 static inline void net_prefetchw(void *p) 2585 { 2586 prefetchw(p); 2587 #if L1_CACHE_BYTES < 128 2588 prefetchw((u8 *)p + L1_CACHE_BYTES); 2589 #endif 2590 } 2591 2592 void netdev_unbind_sb_channel(struct net_device *dev, 2593 struct net_device *sb_dev); 2594 int netdev_bind_sb_channel_queue(struct net_device *dev, 2595 struct net_device *sb_dev, 2596 u8 tc, u16 count, u16 offset); 2597 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2598 static inline int netdev_get_sb_channel(struct net_device *dev) 2599 { 2600 return max_t(int, -dev->num_tc, 0); 2601 } 2602 2603 static inline 2604 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2605 unsigned int index) 2606 { 2607 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues); 2608 return &dev->_tx[index]; 2609 } 2610 2611 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2612 const struct sk_buff *skb) 2613 { 2614 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2615 } 2616 2617 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2618 void (*f)(struct net_device *, 2619 struct netdev_queue *, 2620 void *), 2621 void *arg) 2622 { 2623 unsigned int i; 2624 2625 for (i = 0; i < dev->num_tx_queues; i++) 2626 f(dev, &dev->_tx[i], arg); 2627 } 2628 2629 static inline int netdev_lock_cmp_fn(const struct lockdep_map *a, 2630 const struct lockdep_map *b) 2631 { 2632 /* Only lower devices currently grab the instance lock, so no 2633 * real ordering issues can occur. In the near future, only 2634 * hardware devices will grab instance lock which also does not 2635 * involve any ordering. Suppress lockdep ordering warnings 2636 * until (if) we start grabbing instance lock on pure SW 2637 * devices (bond/team/veth/etc). 2638 */ 2639 if (a == b) 2640 return 0; 2641 return -1; 2642 } 2643 2644 #define netdev_lockdep_set_classes(dev) \ 2645 { \ 2646 static struct lock_class_key qdisc_tx_busylock_key; \ 2647 static struct lock_class_key qdisc_xmit_lock_key; \ 2648 static struct lock_class_key dev_addr_list_lock_key; \ 2649 static struct lock_class_key dev_instance_lock_key; \ 2650 unsigned int i; \ 2651 \ 2652 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2653 lockdep_set_class(&(dev)->addr_list_lock, \ 2654 &dev_addr_list_lock_key); \ 2655 lockdep_set_class(&(dev)->lock, \ 2656 &dev_instance_lock_key); \ 2657 lock_set_cmp_fn(&dev->lock, netdev_lock_cmp_fn, NULL); \ 2658 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2659 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2660 &qdisc_xmit_lock_key); \ 2661 } 2662 2663 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2664 struct net_device *sb_dev); 2665 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2666 struct sk_buff *skb, 2667 struct net_device *sb_dev); 2668 2669 /* returns the headroom that the master device needs to take in account 2670 * when forwarding to this dev 2671 */ 2672 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2673 { 2674 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2675 } 2676 2677 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2678 { 2679 if (dev->netdev_ops->ndo_set_rx_headroom) 2680 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2681 } 2682 2683 /* set the device rx headroom to the dev's default */ 2684 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2685 { 2686 netdev_set_rx_headroom(dev, -1); 2687 } 2688 2689 static inline void *netdev_get_ml_priv(struct net_device *dev, 2690 enum netdev_ml_priv_type type) 2691 { 2692 if (dev->ml_priv_type != type) 2693 return NULL; 2694 2695 return dev->ml_priv; 2696 } 2697 2698 static inline void netdev_set_ml_priv(struct net_device *dev, 2699 void *ml_priv, 2700 enum netdev_ml_priv_type type) 2701 { 2702 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2703 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2704 dev->ml_priv_type, type); 2705 WARN(!dev->ml_priv_type && dev->ml_priv, 2706 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2707 2708 dev->ml_priv = ml_priv; 2709 dev->ml_priv_type = type; 2710 } 2711 2712 /* 2713 * Net namespace inlines 2714 */ 2715 static inline 2716 struct net *dev_net(const struct net_device *dev) 2717 { 2718 return read_pnet(&dev->nd_net); 2719 } 2720 2721 static inline 2722 struct net *dev_net_rcu(const struct net_device *dev) 2723 { 2724 return read_pnet_rcu(&dev->nd_net); 2725 } 2726 2727 static inline 2728 void dev_net_set(struct net_device *dev, struct net *net) 2729 { 2730 write_pnet(&dev->nd_net, net); 2731 } 2732 2733 /** 2734 * netdev_priv - access network device private data 2735 * @dev: network device 2736 * 2737 * Get network device private data 2738 */ 2739 static inline void *netdev_priv(const struct net_device *dev) 2740 { 2741 return (void *)dev->priv; 2742 } 2743 2744 /* Set the sysfs physical device reference for the network logical device 2745 * if set prior to registration will cause a symlink during initialization. 2746 */ 2747 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2748 2749 /* Set the sysfs device type for the network logical device to allow 2750 * fine-grained identification of different network device types. For 2751 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2752 */ 2753 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2754 2755 void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index, 2756 enum netdev_queue_type type, 2757 struct napi_struct *napi); 2758 2759 static inline void netdev_lock(struct net_device *dev) 2760 { 2761 mutex_lock(&dev->lock); 2762 } 2763 2764 static inline void netdev_unlock(struct net_device *dev) 2765 { 2766 mutex_unlock(&dev->lock); 2767 } 2768 2769 static inline void netdev_assert_locked(struct net_device *dev) 2770 { 2771 lockdep_assert_held(&dev->lock); 2772 } 2773 2774 static inline void netdev_assert_locked_or_invisible(struct net_device *dev) 2775 { 2776 if (dev->reg_state == NETREG_REGISTERED || 2777 dev->reg_state == NETREG_UNREGISTERING) 2778 netdev_assert_locked(dev); 2779 } 2780 2781 static inline bool netdev_need_ops_lock(struct net_device *dev) 2782 { 2783 bool ret = dev->request_ops_lock || !!dev->queue_mgmt_ops; 2784 2785 #if IS_ENABLED(CONFIG_NET_SHAPER) 2786 ret |= !!dev->netdev_ops->net_shaper_ops; 2787 #endif 2788 2789 return ret; 2790 } 2791 2792 static inline void netdev_lock_ops(struct net_device *dev) 2793 { 2794 if (netdev_need_ops_lock(dev)) 2795 netdev_lock(dev); 2796 } 2797 2798 static inline void netdev_unlock_ops(struct net_device *dev) 2799 { 2800 if (netdev_need_ops_lock(dev)) 2801 netdev_unlock(dev); 2802 } 2803 2804 static inline void netdev_ops_assert_locked(struct net_device *dev) 2805 { 2806 if (netdev_need_ops_lock(dev)) 2807 lockdep_assert_held(&dev->lock); 2808 } 2809 2810 void netif_napi_set_irq_locked(struct napi_struct *napi, int irq); 2811 2812 static inline void netif_napi_set_irq(struct napi_struct *napi, int irq) 2813 { 2814 netdev_lock(napi->dev); 2815 netif_napi_set_irq_locked(napi, irq); 2816 netdev_unlock(napi->dev); 2817 } 2818 2819 /* Default NAPI poll() weight 2820 * Device drivers are strongly advised to not use bigger value 2821 */ 2822 #define NAPI_POLL_WEIGHT 64 2823 2824 void netif_napi_add_weight_locked(struct net_device *dev, 2825 struct napi_struct *napi, 2826 int (*poll)(struct napi_struct *, int), 2827 int weight); 2828 2829 static inline void 2830 netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2831 int (*poll)(struct napi_struct *, int), int weight) 2832 { 2833 netdev_lock(dev); 2834 netif_napi_add_weight_locked(dev, napi, poll, weight); 2835 netdev_unlock(dev); 2836 } 2837 2838 /** 2839 * netif_napi_add() - initialize a NAPI context 2840 * @dev: network device 2841 * @napi: NAPI context 2842 * @poll: polling function 2843 * 2844 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2845 * *any* of the other NAPI-related functions. 2846 */ 2847 static inline void 2848 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2849 int (*poll)(struct napi_struct *, int)) 2850 { 2851 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2852 } 2853 2854 static inline void 2855 netif_napi_add_locked(struct net_device *dev, struct napi_struct *napi, 2856 int (*poll)(struct napi_struct *, int)) 2857 { 2858 netif_napi_add_weight_locked(dev, napi, poll, NAPI_POLL_WEIGHT); 2859 } 2860 2861 static inline void 2862 netif_napi_add_tx_weight(struct net_device *dev, 2863 struct napi_struct *napi, 2864 int (*poll)(struct napi_struct *, int), 2865 int weight) 2866 { 2867 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2868 netif_napi_add_weight(dev, napi, poll, weight); 2869 } 2870 2871 static inline void 2872 netif_napi_add_config_locked(struct net_device *dev, struct napi_struct *napi, 2873 int (*poll)(struct napi_struct *, int), int index) 2874 { 2875 napi->index = index; 2876 napi->config = &dev->napi_config[index]; 2877 netif_napi_add_weight_locked(dev, napi, poll, NAPI_POLL_WEIGHT); 2878 } 2879 2880 /** 2881 * netif_napi_add_config - initialize a NAPI context with persistent config 2882 * @dev: network device 2883 * @napi: NAPI context 2884 * @poll: polling function 2885 * @index: the NAPI index 2886 */ 2887 static inline void 2888 netif_napi_add_config(struct net_device *dev, struct napi_struct *napi, 2889 int (*poll)(struct napi_struct *, int), int index) 2890 { 2891 netdev_lock(dev); 2892 netif_napi_add_config_locked(dev, napi, poll, index); 2893 netdev_unlock(dev); 2894 } 2895 2896 /** 2897 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2898 * @dev: network device 2899 * @napi: NAPI context 2900 * @poll: polling function 2901 * 2902 * This variant of netif_napi_add() should be used from drivers using NAPI 2903 * to exclusively poll a TX queue. 2904 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2905 */ 2906 static inline void netif_napi_add_tx(struct net_device *dev, 2907 struct napi_struct *napi, 2908 int (*poll)(struct napi_struct *, int)) 2909 { 2910 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2911 } 2912 2913 void __netif_napi_del_locked(struct napi_struct *napi); 2914 2915 /** 2916 * __netif_napi_del - remove a NAPI context 2917 * @napi: NAPI context 2918 * 2919 * Warning: caller must observe RCU grace period before freeing memory 2920 * containing @napi. Drivers might want to call this helper to combine 2921 * all the needed RCU grace periods into a single one. 2922 */ 2923 static inline void __netif_napi_del(struct napi_struct *napi) 2924 { 2925 netdev_lock(napi->dev); 2926 __netif_napi_del_locked(napi); 2927 netdev_unlock(napi->dev); 2928 } 2929 2930 static inline void netif_napi_del_locked(struct napi_struct *napi) 2931 { 2932 __netif_napi_del_locked(napi); 2933 synchronize_net(); 2934 } 2935 2936 /** 2937 * netif_napi_del - remove a NAPI context 2938 * @napi: NAPI context 2939 * 2940 * netif_napi_del() removes a NAPI context from the network device NAPI list 2941 */ 2942 static inline void netif_napi_del(struct napi_struct *napi) 2943 { 2944 __netif_napi_del(napi); 2945 synchronize_net(); 2946 } 2947 2948 int netif_enable_cpu_rmap(struct net_device *dev, unsigned int num_irqs); 2949 void netif_set_affinity_auto(struct net_device *dev); 2950 2951 struct packet_type { 2952 __be16 type; /* This is really htons(ether_type). */ 2953 bool ignore_outgoing; 2954 struct net_device *dev; /* NULL is wildcarded here */ 2955 netdevice_tracker dev_tracker; 2956 int (*func) (struct sk_buff *, 2957 struct net_device *, 2958 struct packet_type *, 2959 struct net_device *); 2960 void (*list_func) (struct list_head *, 2961 struct packet_type *, 2962 struct net_device *); 2963 bool (*id_match)(struct packet_type *ptype, 2964 struct sock *sk); 2965 struct net *af_packet_net; 2966 void *af_packet_priv; 2967 struct list_head list; 2968 }; 2969 2970 struct offload_callbacks { 2971 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2972 netdev_features_t features); 2973 struct sk_buff *(*gro_receive)(struct list_head *head, 2974 struct sk_buff *skb); 2975 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2976 }; 2977 2978 struct packet_offload { 2979 __be16 type; /* This is really htons(ether_type). */ 2980 u16 priority; 2981 struct offload_callbacks callbacks; 2982 struct list_head list; 2983 }; 2984 2985 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2986 struct pcpu_sw_netstats { 2987 u64_stats_t rx_packets; 2988 u64_stats_t rx_bytes; 2989 u64_stats_t tx_packets; 2990 u64_stats_t tx_bytes; 2991 struct u64_stats_sync syncp; 2992 } __aligned(4 * sizeof(u64)); 2993 2994 struct pcpu_dstats { 2995 u64_stats_t rx_packets; 2996 u64_stats_t rx_bytes; 2997 u64_stats_t tx_packets; 2998 u64_stats_t tx_bytes; 2999 u64_stats_t rx_drops; 3000 u64_stats_t tx_drops; 3001 struct u64_stats_sync syncp; 3002 } __aligned(8 * sizeof(u64)); 3003 3004 struct pcpu_lstats { 3005 u64_stats_t packets; 3006 u64_stats_t bytes; 3007 struct u64_stats_sync syncp; 3008 } __aligned(2 * sizeof(u64)); 3009 3010 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 3011 3012 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 3013 { 3014 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 3015 3016 u64_stats_update_begin(&tstats->syncp); 3017 u64_stats_add(&tstats->rx_bytes, len); 3018 u64_stats_inc(&tstats->rx_packets); 3019 u64_stats_update_end(&tstats->syncp); 3020 } 3021 3022 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 3023 unsigned int packets, 3024 unsigned int len) 3025 { 3026 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 3027 3028 u64_stats_update_begin(&tstats->syncp); 3029 u64_stats_add(&tstats->tx_bytes, len); 3030 u64_stats_add(&tstats->tx_packets, packets); 3031 u64_stats_update_end(&tstats->syncp); 3032 } 3033 3034 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 3035 { 3036 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 3037 3038 u64_stats_update_begin(&lstats->syncp); 3039 u64_stats_add(&lstats->bytes, len); 3040 u64_stats_inc(&lstats->packets); 3041 u64_stats_update_end(&lstats->syncp); 3042 } 3043 3044 static inline void dev_dstats_rx_add(struct net_device *dev, 3045 unsigned int len) 3046 { 3047 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3048 3049 u64_stats_update_begin(&dstats->syncp); 3050 u64_stats_inc(&dstats->rx_packets); 3051 u64_stats_add(&dstats->rx_bytes, len); 3052 u64_stats_update_end(&dstats->syncp); 3053 } 3054 3055 static inline void dev_dstats_rx_dropped(struct net_device *dev) 3056 { 3057 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3058 3059 u64_stats_update_begin(&dstats->syncp); 3060 u64_stats_inc(&dstats->rx_drops); 3061 u64_stats_update_end(&dstats->syncp); 3062 } 3063 3064 static inline void dev_dstats_tx_add(struct net_device *dev, 3065 unsigned int len) 3066 { 3067 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3068 3069 u64_stats_update_begin(&dstats->syncp); 3070 u64_stats_inc(&dstats->tx_packets); 3071 u64_stats_add(&dstats->tx_bytes, len); 3072 u64_stats_update_end(&dstats->syncp); 3073 } 3074 3075 static inline void dev_dstats_tx_dropped(struct net_device *dev) 3076 { 3077 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3078 3079 u64_stats_update_begin(&dstats->syncp); 3080 u64_stats_inc(&dstats->tx_drops); 3081 u64_stats_update_end(&dstats->syncp); 3082 } 3083 3084 #define __netdev_alloc_pcpu_stats(type, gfp) \ 3085 ({ \ 3086 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 3087 if (pcpu_stats) { \ 3088 int __cpu; \ 3089 for_each_possible_cpu(__cpu) { \ 3090 typeof(type) *stat; \ 3091 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 3092 u64_stats_init(&stat->syncp); \ 3093 } \ 3094 } \ 3095 pcpu_stats; \ 3096 }) 3097 3098 #define netdev_alloc_pcpu_stats(type) \ 3099 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 3100 3101 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 3102 ({ \ 3103 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 3104 if (pcpu_stats) { \ 3105 int __cpu; \ 3106 for_each_possible_cpu(__cpu) { \ 3107 typeof(type) *stat; \ 3108 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 3109 u64_stats_init(&stat->syncp); \ 3110 } \ 3111 } \ 3112 pcpu_stats; \ 3113 }) 3114 3115 enum netdev_lag_tx_type { 3116 NETDEV_LAG_TX_TYPE_UNKNOWN, 3117 NETDEV_LAG_TX_TYPE_RANDOM, 3118 NETDEV_LAG_TX_TYPE_BROADCAST, 3119 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 3120 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 3121 NETDEV_LAG_TX_TYPE_HASH, 3122 }; 3123 3124 enum netdev_lag_hash { 3125 NETDEV_LAG_HASH_NONE, 3126 NETDEV_LAG_HASH_L2, 3127 NETDEV_LAG_HASH_L34, 3128 NETDEV_LAG_HASH_L23, 3129 NETDEV_LAG_HASH_E23, 3130 NETDEV_LAG_HASH_E34, 3131 NETDEV_LAG_HASH_VLAN_SRCMAC, 3132 NETDEV_LAG_HASH_UNKNOWN, 3133 }; 3134 3135 struct netdev_lag_upper_info { 3136 enum netdev_lag_tx_type tx_type; 3137 enum netdev_lag_hash hash_type; 3138 }; 3139 3140 struct netdev_lag_lower_state_info { 3141 u8 link_up : 1, 3142 tx_enabled : 1; 3143 }; 3144 3145 #include <linux/notifier.h> 3146 3147 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 3148 * and the rtnetlink notification exclusion list in rtnetlink_event() when 3149 * adding new types. 3150 */ 3151 enum netdev_cmd { 3152 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 3153 NETDEV_DOWN, 3154 NETDEV_REBOOT, /* Tell a protocol stack a network interface 3155 detected a hardware crash and restarted 3156 - we can use this eg to kick tcp sessions 3157 once done */ 3158 NETDEV_CHANGE, /* Notify device state change */ 3159 NETDEV_REGISTER, 3160 NETDEV_UNREGISTER, 3161 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 3162 NETDEV_CHANGEADDR, /* notify after the address change */ 3163 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 3164 NETDEV_GOING_DOWN, 3165 NETDEV_CHANGENAME, 3166 NETDEV_FEAT_CHANGE, 3167 NETDEV_BONDING_FAILOVER, 3168 NETDEV_PRE_UP, 3169 NETDEV_PRE_TYPE_CHANGE, 3170 NETDEV_POST_TYPE_CHANGE, 3171 NETDEV_POST_INIT, 3172 NETDEV_PRE_UNINIT, 3173 NETDEV_RELEASE, 3174 NETDEV_NOTIFY_PEERS, 3175 NETDEV_JOIN, 3176 NETDEV_CHANGEUPPER, 3177 NETDEV_RESEND_IGMP, 3178 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 3179 NETDEV_CHANGEINFODATA, 3180 NETDEV_BONDING_INFO, 3181 NETDEV_PRECHANGEUPPER, 3182 NETDEV_CHANGELOWERSTATE, 3183 NETDEV_UDP_TUNNEL_PUSH_INFO, 3184 NETDEV_UDP_TUNNEL_DROP_INFO, 3185 NETDEV_CHANGE_TX_QUEUE_LEN, 3186 NETDEV_CVLAN_FILTER_PUSH_INFO, 3187 NETDEV_CVLAN_FILTER_DROP_INFO, 3188 NETDEV_SVLAN_FILTER_PUSH_INFO, 3189 NETDEV_SVLAN_FILTER_DROP_INFO, 3190 NETDEV_OFFLOAD_XSTATS_ENABLE, 3191 NETDEV_OFFLOAD_XSTATS_DISABLE, 3192 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 3193 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 3194 NETDEV_XDP_FEAT_CHANGE, 3195 }; 3196 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 3197 3198 int register_netdevice_notifier(struct notifier_block *nb); 3199 int unregister_netdevice_notifier(struct notifier_block *nb); 3200 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 3201 int unregister_netdevice_notifier_net(struct net *net, 3202 struct notifier_block *nb); 3203 int register_netdevice_notifier_dev_net(struct net_device *dev, 3204 struct notifier_block *nb, 3205 struct netdev_net_notifier *nn); 3206 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 3207 struct notifier_block *nb, 3208 struct netdev_net_notifier *nn); 3209 3210 struct netdev_notifier_info { 3211 struct net_device *dev; 3212 struct netlink_ext_ack *extack; 3213 }; 3214 3215 struct netdev_notifier_info_ext { 3216 struct netdev_notifier_info info; /* must be first */ 3217 union { 3218 u32 mtu; 3219 } ext; 3220 }; 3221 3222 struct netdev_notifier_change_info { 3223 struct netdev_notifier_info info; /* must be first */ 3224 unsigned int flags_changed; 3225 }; 3226 3227 struct netdev_notifier_changeupper_info { 3228 struct netdev_notifier_info info; /* must be first */ 3229 struct net_device *upper_dev; /* new upper dev */ 3230 bool master; /* is upper dev master */ 3231 bool linking; /* is the notification for link or unlink */ 3232 void *upper_info; /* upper dev info */ 3233 }; 3234 3235 struct netdev_notifier_changelowerstate_info { 3236 struct netdev_notifier_info info; /* must be first */ 3237 void *lower_state_info; /* is lower dev state */ 3238 }; 3239 3240 struct netdev_notifier_pre_changeaddr_info { 3241 struct netdev_notifier_info info; /* must be first */ 3242 const unsigned char *dev_addr; 3243 }; 3244 3245 enum netdev_offload_xstats_type { 3246 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 3247 }; 3248 3249 struct netdev_notifier_offload_xstats_info { 3250 struct netdev_notifier_info info; /* must be first */ 3251 enum netdev_offload_xstats_type type; 3252 3253 union { 3254 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 3255 struct netdev_notifier_offload_xstats_rd *report_delta; 3256 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 3257 struct netdev_notifier_offload_xstats_ru *report_used; 3258 }; 3259 }; 3260 3261 int netdev_offload_xstats_enable(struct net_device *dev, 3262 enum netdev_offload_xstats_type type, 3263 struct netlink_ext_ack *extack); 3264 int netdev_offload_xstats_disable(struct net_device *dev, 3265 enum netdev_offload_xstats_type type); 3266 bool netdev_offload_xstats_enabled(const struct net_device *dev, 3267 enum netdev_offload_xstats_type type); 3268 int netdev_offload_xstats_get(struct net_device *dev, 3269 enum netdev_offload_xstats_type type, 3270 struct rtnl_hw_stats64 *stats, bool *used, 3271 struct netlink_ext_ack *extack); 3272 void 3273 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 3274 const struct rtnl_hw_stats64 *stats); 3275 void 3276 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 3277 void netdev_offload_xstats_push_delta(struct net_device *dev, 3278 enum netdev_offload_xstats_type type, 3279 const struct rtnl_hw_stats64 *stats); 3280 3281 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 3282 struct net_device *dev) 3283 { 3284 info->dev = dev; 3285 info->extack = NULL; 3286 } 3287 3288 static inline struct net_device * 3289 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 3290 { 3291 return info->dev; 3292 } 3293 3294 static inline struct netlink_ext_ack * 3295 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 3296 { 3297 return info->extack; 3298 } 3299 3300 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 3301 int call_netdevice_notifiers_info(unsigned long val, 3302 struct netdev_notifier_info *info); 3303 3304 #define for_each_netdev(net, d) \ 3305 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 3306 #define for_each_netdev_reverse(net, d) \ 3307 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 3308 #define for_each_netdev_rcu(net, d) \ 3309 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 3310 #define for_each_netdev_safe(net, d, n) \ 3311 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 3312 #define for_each_netdev_continue(net, d) \ 3313 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 3314 #define for_each_netdev_continue_reverse(net, d) \ 3315 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 3316 dev_list) 3317 #define for_each_netdev_continue_rcu(net, d) \ 3318 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 3319 #define for_each_netdev_in_bond_rcu(bond, slave) \ 3320 for_each_netdev_rcu(&init_net, slave) \ 3321 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 3322 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 3323 3324 #define for_each_netdev_dump(net, d, ifindex) \ 3325 for (; (d = xa_find(&(net)->dev_by_index, &ifindex, \ 3326 ULONG_MAX, XA_PRESENT)); ifindex++) 3327 3328 static inline struct net_device *next_net_device(struct net_device *dev) 3329 { 3330 struct list_head *lh; 3331 struct net *net; 3332 3333 net = dev_net(dev); 3334 lh = dev->dev_list.next; 3335 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3336 } 3337 3338 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 3339 { 3340 struct list_head *lh; 3341 struct net *net; 3342 3343 net = dev_net(dev); 3344 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 3345 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3346 } 3347 3348 static inline struct net_device *first_net_device(struct net *net) 3349 { 3350 return list_empty(&net->dev_base_head) ? NULL : 3351 net_device_entry(net->dev_base_head.next); 3352 } 3353 3354 static inline struct net_device *first_net_device_rcu(struct net *net) 3355 { 3356 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 3357 3358 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3359 } 3360 3361 int netdev_boot_setup_check(struct net_device *dev); 3362 struct net_device *dev_getbyhwaddr(struct net *net, unsigned short type, 3363 const char *hwaddr); 3364 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 3365 const char *hwaddr); 3366 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 3367 void dev_add_pack(struct packet_type *pt); 3368 void dev_remove_pack(struct packet_type *pt); 3369 void __dev_remove_pack(struct packet_type *pt); 3370 void dev_add_offload(struct packet_offload *po); 3371 void dev_remove_offload(struct packet_offload *po); 3372 3373 int dev_get_iflink(const struct net_device *dev); 3374 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 3375 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 3376 struct net_device_path_stack *stack); 3377 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 3378 unsigned short mask); 3379 struct net_device *dev_get_by_name(struct net *net, const char *name); 3380 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 3381 struct net_device *__dev_get_by_name(struct net *net, const char *name); 3382 bool netdev_name_in_use(struct net *net, const char *name); 3383 int dev_alloc_name(struct net_device *dev, const char *name); 3384 int netif_open(struct net_device *dev, struct netlink_ext_ack *extack); 3385 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 3386 void netif_close(struct net_device *dev); 3387 void dev_close(struct net_device *dev); 3388 void dev_close_many(struct list_head *head, bool unlink); 3389 int dev_setup_tc(struct net_device *dev, enum tc_setup_type type, 3390 void *type_data); 3391 void netif_disable_lro(struct net_device *dev); 3392 void dev_disable_lro(struct net_device *dev); 3393 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 3394 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 3395 struct net_device *sb_dev); 3396 3397 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3398 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3399 3400 static inline int dev_queue_xmit(struct sk_buff *skb) 3401 { 3402 return __dev_queue_xmit(skb, NULL); 3403 } 3404 3405 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3406 struct net_device *sb_dev) 3407 { 3408 return __dev_queue_xmit(skb, sb_dev); 3409 } 3410 3411 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3412 { 3413 int ret; 3414 3415 ret = __dev_direct_xmit(skb, queue_id); 3416 if (!dev_xmit_complete(ret)) 3417 kfree_skb(skb); 3418 return ret; 3419 } 3420 3421 int register_netdevice(struct net_device *dev); 3422 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3423 void unregister_netdevice_many(struct list_head *head); 3424 static inline void unregister_netdevice(struct net_device *dev) 3425 { 3426 unregister_netdevice_queue(dev, NULL); 3427 } 3428 3429 int netdev_refcnt_read(const struct net_device *dev); 3430 void free_netdev(struct net_device *dev); 3431 3432 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3433 struct sk_buff *skb, 3434 bool all_slaves); 3435 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3436 struct sock *sk); 3437 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3438 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3439 struct net_device *netdev_get_by_index(struct net *net, int ifindex, 3440 netdevice_tracker *tracker, gfp_t gfp); 3441 struct net_device *netdev_get_by_name(struct net *net, const char *name, 3442 netdevice_tracker *tracker, gfp_t gfp); 3443 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3444 void netdev_copy_name(struct net_device *dev, char *name); 3445 3446 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3447 unsigned short type, 3448 const void *daddr, const void *saddr, 3449 unsigned int len) 3450 { 3451 if (!dev->header_ops || !dev->header_ops->create) 3452 return 0; 3453 3454 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3455 } 3456 3457 static inline int dev_parse_header(const struct sk_buff *skb, 3458 unsigned char *haddr) 3459 { 3460 const struct net_device *dev = skb->dev; 3461 3462 if (!dev->header_ops || !dev->header_ops->parse) 3463 return 0; 3464 return dev->header_ops->parse(skb, haddr); 3465 } 3466 3467 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3468 { 3469 const struct net_device *dev = skb->dev; 3470 3471 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3472 return 0; 3473 return dev->header_ops->parse_protocol(skb); 3474 } 3475 3476 /* ll_header must have at least hard_header_len allocated */ 3477 static inline bool dev_validate_header(const struct net_device *dev, 3478 char *ll_header, int len) 3479 { 3480 if (likely(len >= dev->hard_header_len)) 3481 return true; 3482 if (len < dev->min_header_len) 3483 return false; 3484 3485 if (capable(CAP_SYS_RAWIO)) { 3486 memset(ll_header + len, 0, dev->hard_header_len - len); 3487 return true; 3488 } 3489 3490 if (dev->header_ops && dev->header_ops->validate) 3491 return dev->header_ops->validate(ll_header, len); 3492 3493 return false; 3494 } 3495 3496 static inline bool dev_has_header(const struct net_device *dev) 3497 { 3498 return dev->header_ops && dev->header_ops->create; 3499 } 3500 3501 /* 3502 * Incoming packets are placed on per-CPU queues 3503 */ 3504 struct softnet_data { 3505 struct list_head poll_list; 3506 struct sk_buff_head process_queue; 3507 local_lock_t process_queue_bh_lock; 3508 3509 /* stats */ 3510 unsigned int processed; 3511 unsigned int time_squeeze; 3512 #ifdef CONFIG_RPS 3513 struct softnet_data *rps_ipi_list; 3514 #endif 3515 3516 unsigned int received_rps; 3517 bool in_net_rx_action; 3518 bool in_napi_threaded_poll; 3519 3520 #ifdef CONFIG_NET_FLOW_LIMIT 3521 struct sd_flow_limit __rcu *flow_limit; 3522 #endif 3523 struct Qdisc *output_queue; 3524 struct Qdisc **output_queue_tailp; 3525 struct sk_buff *completion_queue; 3526 #ifdef CONFIG_XFRM_OFFLOAD 3527 struct sk_buff_head xfrm_backlog; 3528 #endif 3529 /* written and read only by owning cpu: */ 3530 struct netdev_xmit xmit; 3531 #ifdef CONFIG_RPS 3532 /* input_queue_head should be written by cpu owning this struct, 3533 * and only read by other cpus. Worth using a cache line. 3534 */ 3535 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3536 3537 /* Elements below can be accessed between CPUs for RPS/RFS */ 3538 call_single_data_t csd ____cacheline_aligned_in_smp; 3539 struct softnet_data *rps_ipi_next; 3540 unsigned int cpu; 3541 unsigned int input_queue_tail; 3542 #endif 3543 struct sk_buff_head input_pkt_queue; 3544 struct napi_struct backlog; 3545 3546 atomic_t dropped ____cacheline_aligned_in_smp; 3547 3548 /* Another possibly contended cache line */ 3549 spinlock_t defer_lock ____cacheline_aligned_in_smp; 3550 int defer_count; 3551 int defer_ipi_scheduled; 3552 struct sk_buff *defer_list; 3553 call_single_data_t defer_csd; 3554 }; 3555 3556 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3557 DECLARE_PER_CPU(struct page_pool *, system_page_pool); 3558 3559 #ifndef CONFIG_PREEMPT_RT 3560 static inline int dev_recursion_level(void) 3561 { 3562 return this_cpu_read(softnet_data.xmit.recursion); 3563 } 3564 #else 3565 static inline int dev_recursion_level(void) 3566 { 3567 return current->net_xmit.recursion; 3568 } 3569 3570 #endif 3571 3572 void __netif_schedule(struct Qdisc *q); 3573 void netif_schedule_queue(struct netdev_queue *txq); 3574 3575 static inline void netif_tx_schedule_all(struct net_device *dev) 3576 { 3577 unsigned int i; 3578 3579 for (i = 0; i < dev->num_tx_queues; i++) 3580 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3581 } 3582 3583 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3584 { 3585 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3586 } 3587 3588 /** 3589 * netif_start_queue - allow transmit 3590 * @dev: network device 3591 * 3592 * Allow upper layers to call the device hard_start_xmit routine. 3593 */ 3594 static inline void netif_start_queue(struct net_device *dev) 3595 { 3596 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3597 } 3598 3599 static inline void netif_tx_start_all_queues(struct net_device *dev) 3600 { 3601 unsigned int i; 3602 3603 for (i = 0; i < dev->num_tx_queues; i++) { 3604 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3605 netif_tx_start_queue(txq); 3606 } 3607 } 3608 3609 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3610 3611 /** 3612 * netif_wake_queue - restart transmit 3613 * @dev: network device 3614 * 3615 * Allow upper layers to call the device hard_start_xmit routine. 3616 * Used for flow control when transmit resources are available. 3617 */ 3618 static inline void netif_wake_queue(struct net_device *dev) 3619 { 3620 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3621 } 3622 3623 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3624 { 3625 unsigned int i; 3626 3627 for (i = 0; i < dev->num_tx_queues; i++) { 3628 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3629 netif_tx_wake_queue(txq); 3630 } 3631 } 3632 3633 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3634 { 3635 /* Paired with READ_ONCE() from dev_watchdog() */ 3636 WRITE_ONCE(dev_queue->trans_start, jiffies); 3637 3638 /* This barrier is paired with smp_mb() from dev_watchdog() */ 3639 smp_mb__before_atomic(); 3640 3641 /* Must be an atomic op see netif_txq_try_stop() */ 3642 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3643 } 3644 3645 /** 3646 * netif_stop_queue - stop transmitted packets 3647 * @dev: network device 3648 * 3649 * Stop upper layers calling the device hard_start_xmit routine. 3650 * Used for flow control when transmit resources are unavailable. 3651 */ 3652 static inline void netif_stop_queue(struct net_device *dev) 3653 { 3654 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3655 } 3656 3657 void netif_tx_stop_all_queues(struct net_device *dev); 3658 3659 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3660 { 3661 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3662 } 3663 3664 /** 3665 * netif_queue_stopped - test if transmit queue is flowblocked 3666 * @dev: network device 3667 * 3668 * Test if transmit queue on device is currently unable to send. 3669 */ 3670 static inline bool netif_queue_stopped(const struct net_device *dev) 3671 { 3672 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3673 } 3674 3675 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3676 { 3677 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3678 } 3679 3680 static inline bool 3681 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3682 { 3683 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3684 } 3685 3686 static inline bool 3687 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3688 { 3689 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3690 } 3691 3692 /** 3693 * netdev_queue_set_dql_min_limit - set dql minimum limit 3694 * @dev_queue: pointer to transmit queue 3695 * @min_limit: dql minimum limit 3696 * 3697 * Forces xmit_more() to return true until the minimum threshold 3698 * defined by @min_limit is reached (or until the tx queue is 3699 * empty). Warning: to be use with care, misuse will impact the 3700 * latency. 3701 */ 3702 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3703 unsigned int min_limit) 3704 { 3705 #ifdef CONFIG_BQL 3706 dev_queue->dql.min_limit = min_limit; 3707 #endif 3708 } 3709 3710 static inline int netdev_queue_dql_avail(const struct netdev_queue *txq) 3711 { 3712 #ifdef CONFIG_BQL 3713 /* Non-BQL migrated drivers will return 0, too. */ 3714 return dql_avail(&txq->dql); 3715 #else 3716 return 0; 3717 #endif 3718 } 3719 3720 /** 3721 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3722 * @dev_queue: pointer to transmit queue 3723 * 3724 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3725 * to give appropriate hint to the CPU. 3726 */ 3727 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3728 { 3729 #ifdef CONFIG_BQL 3730 prefetchw(&dev_queue->dql.num_queued); 3731 #endif 3732 } 3733 3734 /** 3735 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3736 * @dev_queue: pointer to transmit queue 3737 * 3738 * BQL enabled drivers might use this helper in their TX completion path, 3739 * to give appropriate hint to the CPU. 3740 */ 3741 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3742 { 3743 #ifdef CONFIG_BQL 3744 prefetchw(&dev_queue->dql.limit); 3745 #endif 3746 } 3747 3748 /** 3749 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3750 * @dev_queue: network device queue 3751 * @bytes: number of bytes queued to the device queue 3752 * 3753 * Report the number of bytes queued for sending/completion to the network 3754 * device hardware queue. @bytes should be a good approximation and should 3755 * exactly match netdev_completed_queue() @bytes. 3756 * This is typically called once per packet, from ndo_start_xmit(). 3757 */ 3758 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3759 unsigned int bytes) 3760 { 3761 #ifdef CONFIG_BQL 3762 dql_queued(&dev_queue->dql, bytes); 3763 3764 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3765 return; 3766 3767 /* Paired with READ_ONCE() from dev_watchdog() */ 3768 WRITE_ONCE(dev_queue->trans_start, jiffies); 3769 3770 /* This barrier is paired with smp_mb() from dev_watchdog() */ 3771 smp_mb__before_atomic(); 3772 3773 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3774 3775 /* 3776 * The XOFF flag must be set before checking the dql_avail below, 3777 * because in netdev_tx_completed_queue we update the dql_completed 3778 * before checking the XOFF flag. 3779 */ 3780 smp_mb__after_atomic(); 3781 3782 /* check again in case another CPU has just made room avail */ 3783 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3784 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3785 #endif 3786 } 3787 3788 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3789 * that they should not test BQL status themselves. 3790 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3791 * skb of a batch. 3792 * Returns true if the doorbell must be used to kick the NIC. 3793 */ 3794 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3795 unsigned int bytes, 3796 bool xmit_more) 3797 { 3798 if (xmit_more) { 3799 #ifdef CONFIG_BQL 3800 dql_queued(&dev_queue->dql, bytes); 3801 #endif 3802 return netif_tx_queue_stopped(dev_queue); 3803 } 3804 netdev_tx_sent_queue(dev_queue, bytes); 3805 return true; 3806 } 3807 3808 /** 3809 * netdev_sent_queue - report the number of bytes queued to hardware 3810 * @dev: network device 3811 * @bytes: number of bytes queued to the hardware device queue 3812 * 3813 * Report the number of bytes queued for sending/completion to the network 3814 * device hardware queue#0. @bytes should be a good approximation and should 3815 * exactly match netdev_completed_queue() @bytes. 3816 * This is typically called once per packet, from ndo_start_xmit(). 3817 */ 3818 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3819 { 3820 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3821 } 3822 3823 static inline bool __netdev_sent_queue(struct net_device *dev, 3824 unsigned int bytes, 3825 bool xmit_more) 3826 { 3827 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3828 xmit_more); 3829 } 3830 3831 /** 3832 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3833 * @dev_queue: network device queue 3834 * @pkts: number of packets (currently ignored) 3835 * @bytes: number of bytes dequeued from the device queue 3836 * 3837 * Must be called at most once per TX completion round (and not per 3838 * individual packet), so that BQL can adjust its limits appropriately. 3839 */ 3840 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3841 unsigned int pkts, unsigned int bytes) 3842 { 3843 #ifdef CONFIG_BQL 3844 if (unlikely(!bytes)) 3845 return; 3846 3847 dql_completed(&dev_queue->dql, bytes); 3848 3849 /* 3850 * Without the memory barrier there is a small possibility that 3851 * netdev_tx_sent_queue will miss the update and cause the queue to 3852 * be stopped forever 3853 */ 3854 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */ 3855 3856 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3857 return; 3858 3859 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3860 netif_schedule_queue(dev_queue); 3861 #endif 3862 } 3863 3864 /** 3865 * netdev_completed_queue - report bytes and packets completed by device 3866 * @dev: network device 3867 * @pkts: actual number of packets sent over the medium 3868 * @bytes: actual number of bytes sent over the medium 3869 * 3870 * Report the number of bytes and packets transmitted by the network device 3871 * hardware queue over the physical medium, @bytes must exactly match the 3872 * @bytes amount passed to netdev_sent_queue() 3873 */ 3874 static inline void netdev_completed_queue(struct net_device *dev, 3875 unsigned int pkts, unsigned int bytes) 3876 { 3877 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3878 } 3879 3880 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3881 { 3882 #ifdef CONFIG_BQL 3883 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3884 dql_reset(&q->dql); 3885 #endif 3886 } 3887 3888 /** 3889 * netdev_tx_reset_subqueue - reset the BQL stats and state of a netdev queue 3890 * @dev: network device 3891 * @qid: stack index of the queue to reset 3892 */ 3893 static inline void netdev_tx_reset_subqueue(const struct net_device *dev, 3894 u32 qid) 3895 { 3896 netdev_tx_reset_queue(netdev_get_tx_queue(dev, qid)); 3897 } 3898 3899 /** 3900 * netdev_reset_queue - reset the packets and bytes count of a network device 3901 * @dev_queue: network device 3902 * 3903 * Reset the bytes and packet count of a network device and clear the 3904 * software flow control OFF bit for this network device 3905 */ 3906 static inline void netdev_reset_queue(struct net_device *dev_queue) 3907 { 3908 netdev_tx_reset_subqueue(dev_queue, 0); 3909 } 3910 3911 /** 3912 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3913 * @dev: network device 3914 * @queue_index: given tx queue index 3915 * 3916 * Returns 0 if given tx queue index >= number of device tx queues, 3917 * otherwise returns the originally passed tx queue index. 3918 */ 3919 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3920 { 3921 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3922 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3923 dev->name, queue_index, 3924 dev->real_num_tx_queues); 3925 return 0; 3926 } 3927 3928 return queue_index; 3929 } 3930 3931 /** 3932 * netif_running - test if up 3933 * @dev: network device 3934 * 3935 * Test if the device has been brought up. 3936 */ 3937 static inline bool netif_running(const struct net_device *dev) 3938 { 3939 return test_bit(__LINK_STATE_START, &dev->state); 3940 } 3941 3942 /* 3943 * Routines to manage the subqueues on a device. We only need start, 3944 * stop, and a check if it's stopped. All other device management is 3945 * done at the overall netdevice level. 3946 * Also test the device if we're multiqueue. 3947 */ 3948 3949 /** 3950 * netif_start_subqueue - allow sending packets on subqueue 3951 * @dev: network device 3952 * @queue_index: sub queue index 3953 * 3954 * Start individual transmit queue of a device with multiple transmit queues. 3955 */ 3956 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3957 { 3958 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3959 3960 netif_tx_start_queue(txq); 3961 } 3962 3963 /** 3964 * netif_stop_subqueue - stop sending packets on subqueue 3965 * @dev: network device 3966 * @queue_index: sub queue index 3967 * 3968 * Stop individual transmit queue of a device with multiple transmit queues. 3969 */ 3970 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3971 { 3972 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3973 netif_tx_stop_queue(txq); 3974 } 3975 3976 /** 3977 * __netif_subqueue_stopped - test status of subqueue 3978 * @dev: network device 3979 * @queue_index: sub queue index 3980 * 3981 * Check individual transmit queue of a device with multiple transmit queues. 3982 */ 3983 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3984 u16 queue_index) 3985 { 3986 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3987 3988 return netif_tx_queue_stopped(txq); 3989 } 3990 3991 /** 3992 * netif_subqueue_stopped - test status of subqueue 3993 * @dev: network device 3994 * @skb: sub queue buffer pointer 3995 * 3996 * Check individual transmit queue of a device with multiple transmit queues. 3997 */ 3998 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3999 struct sk_buff *skb) 4000 { 4001 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 4002 } 4003 4004 /** 4005 * netif_wake_subqueue - allow sending packets on subqueue 4006 * @dev: network device 4007 * @queue_index: sub queue index 4008 * 4009 * Resume individual transmit queue of a device with multiple transmit queues. 4010 */ 4011 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 4012 { 4013 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 4014 4015 netif_tx_wake_queue(txq); 4016 } 4017 4018 #ifdef CONFIG_XPS 4019 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 4020 u16 index); 4021 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 4022 u16 index, enum xps_map_type type); 4023 4024 /** 4025 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 4026 * @j: CPU/Rx queue index 4027 * @mask: bitmask of all cpus/rx queues 4028 * @nr_bits: number of bits in the bitmask 4029 * 4030 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 4031 */ 4032 static inline bool netif_attr_test_mask(unsigned long j, 4033 const unsigned long *mask, 4034 unsigned int nr_bits) 4035 { 4036 cpu_max_bits_warn(j, nr_bits); 4037 return test_bit(j, mask); 4038 } 4039 4040 /** 4041 * netif_attr_test_online - Test for online CPU/Rx queue 4042 * @j: CPU/Rx queue index 4043 * @online_mask: bitmask for CPUs/Rx queues that are online 4044 * @nr_bits: number of bits in the bitmask 4045 * 4046 * Returns: true if a CPU/Rx queue is online. 4047 */ 4048 static inline bool netif_attr_test_online(unsigned long j, 4049 const unsigned long *online_mask, 4050 unsigned int nr_bits) 4051 { 4052 cpu_max_bits_warn(j, nr_bits); 4053 4054 if (online_mask) 4055 return test_bit(j, online_mask); 4056 4057 return (j < nr_bits); 4058 } 4059 4060 /** 4061 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 4062 * @n: CPU/Rx queue index 4063 * @srcp: the cpumask/Rx queue mask pointer 4064 * @nr_bits: number of bits in the bitmask 4065 * 4066 * Returns: next (after n) CPU/Rx queue index in the mask; 4067 * >= nr_bits if no further CPUs/Rx queues set. 4068 */ 4069 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 4070 unsigned int nr_bits) 4071 { 4072 /* -1 is a legal arg here. */ 4073 if (n != -1) 4074 cpu_max_bits_warn(n, nr_bits); 4075 4076 if (srcp) 4077 return find_next_bit(srcp, nr_bits, n + 1); 4078 4079 return n + 1; 4080 } 4081 4082 /** 4083 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 4084 * @n: CPU/Rx queue index 4085 * @src1p: the first CPUs/Rx queues mask pointer 4086 * @src2p: the second CPUs/Rx queues mask pointer 4087 * @nr_bits: number of bits in the bitmask 4088 * 4089 * Returns: next (after n) CPU/Rx queue index set in both masks; 4090 * >= nr_bits if no further CPUs/Rx queues set in both. 4091 */ 4092 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 4093 const unsigned long *src2p, 4094 unsigned int nr_bits) 4095 { 4096 /* -1 is a legal arg here. */ 4097 if (n != -1) 4098 cpu_max_bits_warn(n, nr_bits); 4099 4100 if (src1p && src2p) 4101 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 4102 else if (src1p) 4103 return find_next_bit(src1p, nr_bits, n + 1); 4104 else if (src2p) 4105 return find_next_bit(src2p, nr_bits, n + 1); 4106 4107 return n + 1; 4108 } 4109 #else 4110 static inline int netif_set_xps_queue(struct net_device *dev, 4111 const struct cpumask *mask, 4112 u16 index) 4113 { 4114 return 0; 4115 } 4116 4117 static inline int __netif_set_xps_queue(struct net_device *dev, 4118 const unsigned long *mask, 4119 u16 index, enum xps_map_type type) 4120 { 4121 return 0; 4122 } 4123 #endif 4124 4125 /** 4126 * netif_is_multiqueue - test if device has multiple transmit queues 4127 * @dev: network device 4128 * 4129 * Check if device has multiple transmit queues 4130 */ 4131 static inline bool netif_is_multiqueue(const struct net_device *dev) 4132 { 4133 return dev->num_tx_queues > 1; 4134 } 4135 4136 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 4137 4138 #ifdef CONFIG_SYSFS 4139 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 4140 #else 4141 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 4142 unsigned int rxqs) 4143 { 4144 dev->real_num_rx_queues = rxqs; 4145 return 0; 4146 } 4147 #endif 4148 int netif_set_real_num_queues(struct net_device *dev, 4149 unsigned int txq, unsigned int rxq); 4150 4151 int netif_get_num_default_rss_queues(void); 4152 4153 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason); 4154 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason); 4155 4156 /* 4157 * It is not allowed to call kfree_skb() or consume_skb() from hardware 4158 * interrupt context or with hardware interrupts being disabled. 4159 * (in_hardirq() || irqs_disabled()) 4160 * 4161 * We provide four helpers that can be used in following contexts : 4162 * 4163 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 4164 * replacing kfree_skb(skb) 4165 * 4166 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 4167 * Typically used in place of consume_skb(skb) in TX completion path 4168 * 4169 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 4170 * replacing kfree_skb(skb) 4171 * 4172 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 4173 * and consumed a packet. Used in place of consume_skb(skb) 4174 */ 4175 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 4176 { 4177 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 4178 } 4179 4180 static inline void dev_consume_skb_irq(struct sk_buff *skb) 4181 { 4182 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED); 4183 } 4184 4185 static inline void dev_kfree_skb_any(struct sk_buff *skb) 4186 { 4187 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 4188 } 4189 4190 static inline void dev_consume_skb_any(struct sk_buff *skb) 4191 { 4192 dev_kfree_skb_any_reason(skb, SKB_CONSUMED); 4193 } 4194 4195 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 4196 const struct bpf_prog *xdp_prog); 4197 void generic_xdp_tx(struct sk_buff *skb, const struct bpf_prog *xdp_prog); 4198 int do_xdp_generic(const struct bpf_prog *xdp_prog, struct sk_buff **pskb); 4199 int netif_rx(struct sk_buff *skb); 4200 int __netif_rx(struct sk_buff *skb); 4201 4202 int netif_receive_skb(struct sk_buff *skb); 4203 int netif_receive_skb_core(struct sk_buff *skb); 4204 void netif_receive_skb_list_internal(struct list_head *head); 4205 void netif_receive_skb_list(struct list_head *head); 4206 gro_result_t gro_receive_skb(struct gro_node *gro, struct sk_buff *skb); 4207 4208 static inline gro_result_t napi_gro_receive(struct napi_struct *napi, 4209 struct sk_buff *skb) 4210 { 4211 return gro_receive_skb(&napi->gro, skb); 4212 } 4213 4214 struct sk_buff *napi_get_frags(struct napi_struct *napi); 4215 gro_result_t napi_gro_frags(struct napi_struct *napi); 4216 4217 static inline void napi_free_frags(struct napi_struct *napi) 4218 { 4219 kfree_skb(napi->skb); 4220 napi->skb = NULL; 4221 } 4222 4223 bool netdev_is_rx_handler_busy(struct net_device *dev); 4224 int netdev_rx_handler_register(struct net_device *dev, 4225 rx_handler_func_t *rx_handler, 4226 void *rx_handler_data); 4227 void netdev_rx_handler_unregister(struct net_device *dev); 4228 4229 bool dev_valid_name(const char *name); 4230 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 4231 { 4232 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 4233 } 4234 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 4235 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 4236 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 4237 void __user *data, bool *need_copyout); 4238 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 4239 int dev_eth_ioctl(struct net_device *dev, 4240 struct ifreq *ifr, unsigned int cmd); 4241 int generic_hwtstamp_get_lower(struct net_device *dev, 4242 struct kernel_hwtstamp_config *kernel_cfg); 4243 int generic_hwtstamp_set_lower(struct net_device *dev, 4244 struct kernel_hwtstamp_config *kernel_cfg, 4245 struct netlink_ext_ack *extack); 4246 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 4247 unsigned int dev_get_flags(const struct net_device *); 4248 int __dev_change_flags(struct net_device *dev, unsigned int flags, 4249 struct netlink_ext_ack *extack); 4250 int netif_change_flags(struct net_device *dev, unsigned int flags, 4251 struct netlink_ext_ack *extack); 4252 int dev_change_flags(struct net_device *dev, unsigned int flags, 4253 struct netlink_ext_ack *extack); 4254 int netif_set_alias(struct net_device *dev, const char *alias, size_t len); 4255 int dev_set_alias(struct net_device *, const char *, size_t); 4256 int dev_get_alias(const struct net_device *, char *, size_t); 4257 int netif_change_net_namespace(struct net_device *dev, struct net *net, 4258 const char *pat, int new_ifindex, 4259 struct netlink_ext_ack *extack); 4260 int dev_change_net_namespace(struct net_device *dev, struct net *net, 4261 const char *pat); 4262 int __dev_set_mtu(struct net_device *, int); 4263 int netif_set_mtu(struct net_device *dev, int new_mtu); 4264 int dev_set_mtu(struct net_device *, int); 4265 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 4266 struct netlink_ext_ack *extack); 4267 int netif_set_mac_address(struct net_device *dev, struct sockaddr *sa, 4268 struct netlink_ext_ack *extack); 4269 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 4270 struct netlink_ext_ack *extack); 4271 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 4272 int dev_get_port_parent_id(struct net_device *dev, 4273 struct netdev_phys_item_id *ppid, bool recurse); 4274 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 4275 4276 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 4277 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 4278 struct netdev_queue *txq, int *ret); 4279 4280 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 4281 u8 dev_xdp_prog_count(struct net_device *dev); 4282 int netif_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf); 4283 int dev_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf); 4284 u8 dev_xdp_sb_prog_count(struct net_device *dev); 4285 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 4286 4287 u32 dev_get_min_mp_channel_count(const struct net_device *dev); 4288 4289 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4290 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4291 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 4292 bool is_skb_forwardable(const struct net_device *dev, 4293 const struct sk_buff *skb); 4294 4295 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 4296 const struct sk_buff *skb, 4297 const bool check_mtu) 4298 { 4299 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 4300 unsigned int len; 4301 4302 if (!(dev->flags & IFF_UP)) 4303 return false; 4304 4305 if (!check_mtu) 4306 return true; 4307 4308 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 4309 if (skb->len <= len) 4310 return true; 4311 4312 /* if TSO is enabled, we don't care about the length as the packet 4313 * could be forwarded without being segmented before 4314 */ 4315 if (skb_is_gso(skb)) 4316 return true; 4317 4318 return false; 4319 } 4320 4321 void netdev_core_stats_inc(struct net_device *dev, u32 offset); 4322 4323 #define DEV_CORE_STATS_INC(FIELD) \ 4324 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 4325 { \ 4326 netdev_core_stats_inc(dev, \ 4327 offsetof(struct net_device_core_stats, FIELD)); \ 4328 } 4329 DEV_CORE_STATS_INC(rx_dropped) 4330 DEV_CORE_STATS_INC(tx_dropped) 4331 DEV_CORE_STATS_INC(rx_nohandler) 4332 DEV_CORE_STATS_INC(rx_otherhost_dropped) 4333 #undef DEV_CORE_STATS_INC 4334 4335 static __always_inline int ____dev_forward_skb(struct net_device *dev, 4336 struct sk_buff *skb, 4337 const bool check_mtu) 4338 { 4339 if (skb_orphan_frags(skb, GFP_ATOMIC) || 4340 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 4341 dev_core_stats_rx_dropped_inc(dev); 4342 kfree_skb(skb); 4343 return NET_RX_DROP; 4344 } 4345 4346 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 4347 skb->priority = 0; 4348 return 0; 4349 } 4350 4351 bool dev_nit_active(struct net_device *dev); 4352 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 4353 4354 static inline void __dev_put(struct net_device *dev) 4355 { 4356 if (dev) { 4357 #ifdef CONFIG_PCPU_DEV_REFCNT 4358 this_cpu_dec(*dev->pcpu_refcnt); 4359 #else 4360 refcount_dec(&dev->dev_refcnt); 4361 #endif 4362 } 4363 } 4364 4365 static inline void __dev_hold(struct net_device *dev) 4366 { 4367 if (dev) { 4368 #ifdef CONFIG_PCPU_DEV_REFCNT 4369 this_cpu_inc(*dev->pcpu_refcnt); 4370 #else 4371 refcount_inc(&dev->dev_refcnt); 4372 #endif 4373 } 4374 } 4375 4376 static inline void __netdev_tracker_alloc(struct net_device *dev, 4377 netdevice_tracker *tracker, 4378 gfp_t gfp) 4379 { 4380 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4381 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 4382 #endif 4383 } 4384 4385 /* netdev_tracker_alloc() can upgrade a prior untracked reference 4386 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4387 */ 4388 static inline void netdev_tracker_alloc(struct net_device *dev, 4389 netdevice_tracker *tracker, gfp_t gfp) 4390 { 4391 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4392 refcount_dec(&dev->refcnt_tracker.no_tracker); 4393 __netdev_tracker_alloc(dev, tracker, gfp); 4394 #endif 4395 } 4396 4397 static inline void netdev_tracker_free(struct net_device *dev, 4398 netdevice_tracker *tracker) 4399 { 4400 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4401 ref_tracker_free(&dev->refcnt_tracker, tracker); 4402 #endif 4403 } 4404 4405 static inline void netdev_hold(struct net_device *dev, 4406 netdevice_tracker *tracker, gfp_t gfp) 4407 { 4408 if (dev) { 4409 __dev_hold(dev); 4410 __netdev_tracker_alloc(dev, tracker, gfp); 4411 } 4412 } 4413 4414 static inline void netdev_put(struct net_device *dev, 4415 netdevice_tracker *tracker) 4416 { 4417 if (dev) { 4418 netdev_tracker_free(dev, tracker); 4419 __dev_put(dev); 4420 } 4421 } 4422 4423 /** 4424 * dev_hold - get reference to device 4425 * @dev: network device 4426 * 4427 * Hold reference to device to keep it from being freed. 4428 * Try using netdev_hold() instead. 4429 */ 4430 static inline void dev_hold(struct net_device *dev) 4431 { 4432 netdev_hold(dev, NULL, GFP_ATOMIC); 4433 } 4434 4435 /** 4436 * dev_put - release reference to device 4437 * @dev: network device 4438 * 4439 * Release reference to device to allow it to be freed. 4440 * Try using netdev_put() instead. 4441 */ 4442 static inline void dev_put(struct net_device *dev) 4443 { 4444 netdev_put(dev, NULL); 4445 } 4446 4447 DEFINE_FREE(dev_put, struct net_device *, if (_T) dev_put(_T)) 4448 4449 static inline void netdev_ref_replace(struct net_device *odev, 4450 struct net_device *ndev, 4451 netdevice_tracker *tracker, 4452 gfp_t gfp) 4453 { 4454 if (odev) 4455 netdev_tracker_free(odev, tracker); 4456 4457 __dev_hold(ndev); 4458 __dev_put(odev); 4459 4460 if (ndev) 4461 __netdev_tracker_alloc(ndev, tracker, gfp); 4462 } 4463 4464 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4465 * and _off may be called from IRQ context, but it is caller 4466 * who is responsible for serialization of these calls. 4467 * 4468 * The name carrier is inappropriate, these functions should really be 4469 * called netif_lowerlayer_*() because they represent the state of any 4470 * kind of lower layer not just hardware media. 4471 */ 4472 void linkwatch_fire_event(struct net_device *dev); 4473 4474 /** 4475 * linkwatch_sync_dev - sync linkwatch for the given device 4476 * @dev: network device to sync linkwatch for 4477 * 4478 * Sync linkwatch for the given device, removing it from the 4479 * pending work list (if queued). 4480 */ 4481 void linkwatch_sync_dev(struct net_device *dev); 4482 4483 /** 4484 * netif_carrier_ok - test if carrier present 4485 * @dev: network device 4486 * 4487 * Check if carrier is present on device 4488 */ 4489 static inline bool netif_carrier_ok(const struct net_device *dev) 4490 { 4491 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4492 } 4493 4494 unsigned long dev_trans_start(struct net_device *dev); 4495 4496 void netdev_watchdog_up(struct net_device *dev); 4497 4498 void netif_carrier_on(struct net_device *dev); 4499 void netif_carrier_off(struct net_device *dev); 4500 void netif_carrier_event(struct net_device *dev); 4501 4502 /** 4503 * netif_dormant_on - mark device as dormant. 4504 * @dev: network device 4505 * 4506 * Mark device as dormant (as per RFC2863). 4507 * 4508 * The dormant state indicates that the relevant interface is not 4509 * actually in a condition to pass packets (i.e., it is not 'up') but is 4510 * in a "pending" state, waiting for some external event. For "on- 4511 * demand" interfaces, this new state identifies the situation where the 4512 * interface is waiting for events to place it in the up state. 4513 */ 4514 static inline void netif_dormant_on(struct net_device *dev) 4515 { 4516 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4517 linkwatch_fire_event(dev); 4518 } 4519 4520 /** 4521 * netif_dormant_off - set device as not dormant. 4522 * @dev: network device 4523 * 4524 * Device is not in dormant state. 4525 */ 4526 static inline void netif_dormant_off(struct net_device *dev) 4527 { 4528 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4529 linkwatch_fire_event(dev); 4530 } 4531 4532 /** 4533 * netif_dormant - test if device is dormant 4534 * @dev: network device 4535 * 4536 * Check if device is dormant. 4537 */ 4538 static inline bool netif_dormant(const struct net_device *dev) 4539 { 4540 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4541 } 4542 4543 4544 /** 4545 * netif_testing_on - mark device as under test. 4546 * @dev: network device 4547 * 4548 * Mark device as under test (as per RFC2863). 4549 * 4550 * The testing state indicates that some test(s) must be performed on 4551 * the interface. After completion, of the test, the interface state 4552 * will change to up, dormant, or down, as appropriate. 4553 */ 4554 static inline void netif_testing_on(struct net_device *dev) 4555 { 4556 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4557 linkwatch_fire_event(dev); 4558 } 4559 4560 /** 4561 * netif_testing_off - set device as not under test. 4562 * @dev: network device 4563 * 4564 * Device is not in testing state. 4565 */ 4566 static inline void netif_testing_off(struct net_device *dev) 4567 { 4568 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4569 linkwatch_fire_event(dev); 4570 } 4571 4572 /** 4573 * netif_testing - test if device is under test 4574 * @dev: network device 4575 * 4576 * Check if device is under test 4577 */ 4578 static inline bool netif_testing(const struct net_device *dev) 4579 { 4580 return test_bit(__LINK_STATE_TESTING, &dev->state); 4581 } 4582 4583 4584 /** 4585 * netif_oper_up - test if device is operational 4586 * @dev: network device 4587 * 4588 * Check if carrier is operational 4589 */ 4590 static inline bool netif_oper_up(const struct net_device *dev) 4591 { 4592 unsigned int operstate = READ_ONCE(dev->operstate); 4593 4594 return operstate == IF_OPER_UP || 4595 operstate == IF_OPER_UNKNOWN /* backward compat */; 4596 } 4597 4598 /** 4599 * netif_device_present - is device available or removed 4600 * @dev: network device 4601 * 4602 * Check if device has not been removed from system. 4603 */ 4604 static inline bool netif_device_present(const struct net_device *dev) 4605 { 4606 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4607 } 4608 4609 void netif_device_detach(struct net_device *dev); 4610 4611 void netif_device_attach(struct net_device *dev); 4612 4613 /* 4614 * Network interface message level settings 4615 */ 4616 4617 enum { 4618 NETIF_MSG_DRV_BIT, 4619 NETIF_MSG_PROBE_BIT, 4620 NETIF_MSG_LINK_BIT, 4621 NETIF_MSG_TIMER_BIT, 4622 NETIF_MSG_IFDOWN_BIT, 4623 NETIF_MSG_IFUP_BIT, 4624 NETIF_MSG_RX_ERR_BIT, 4625 NETIF_MSG_TX_ERR_BIT, 4626 NETIF_MSG_TX_QUEUED_BIT, 4627 NETIF_MSG_INTR_BIT, 4628 NETIF_MSG_TX_DONE_BIT, 4629 NETIF_MSG_RX_STATUS_BIT, 4630 NETIF_MSG_PKTDATA_BIT, 4631 NETIF_MSG_HW_BIT, 4632 NETIF_MSG_WOL_BIT, 4633 4634 /* When you add a new bit above, update netif_msg_class_names array 4635 * in net/ethtool/common.c 4636 */ 4637 NETIF_MSG_CLASS_COUNT, 4638 }; 4639 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4640 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4641 4642 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4643 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4644 4645 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4646 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4647 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4648 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4649 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4650 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4651 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4652 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4653 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4654 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4655 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4656 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4657 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4658 #define NETIF_MSG_HW __NETIF_MSG(HW) 4659 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4660 4661 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4662 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4663 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4664 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4665 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4666 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4667 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4668 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4669 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4670 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4671 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4672 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4673 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4674 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4675 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4676 4677 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4678 { 4679 /* use default */ 4680 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4681 return default_msg_enable_bits; 4682 if (debug_value == 0) /* no output */ 4683 return 0; 4684 /* set low N bits */ 4685 return (1U << debug_value) - 1; 4686 } 4687 4688 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4689 { 4690 spin_lock(&txq->_xmit_lock); 4691 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4692 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4693 } 4694 4695 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4696 { 4697 __acquire(&txq->_xmit_lock); 4698 return true; 4699 } 4700 4701 static inline void __netif_tx_release(struct netdev_queue *txq) 4702 { 4703 __release(&txq->_xmit_lock); 4704 } 4705 4706 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4707 { 4708 spin_lock_bh(&txq->_xmit_lock); 4709 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4710 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4711 } 4712 4713 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4714 { 4715 bool ok = spin_trylock(&txq->_xmit_lock); 4716 4717 if (likely(ok)) { 4718 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4719 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4720 } 4721 return ok; 4722 } 4723 4724 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4725 { 4726 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4727 WRITE_ONCE(txq->xmit_lock_owner, -1); 4728 spin_unlock(&txq->_xmit_lock); 4729 } 4730 4731 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4732 { 4733 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4734 WRITE_ONCE(txq->xmit_lock_owner, -1); 4735 spin_unlock_bh(&txq->_xmit_lock); 4736 } 4737 4738 /* 4739 * txq->trans_start can be read locklessly from dev_watchdog() 4740 */ 4741 static inline void txq_trans_update(struct netdev_queue *txq) 4742 { 4743 if (txq->xmit_lock_owner != -1) 4744 WRITE_ONCE(txq->trans_start, jiffies); 4745 } 4746 4747 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4748 { 4749 unsigned long now = jiffies; 4750 4751 if (READ_ONCE(txq->trans_start) != now) 4752 WRITE_ONCE(txq->trans_start, now); 4753 } 4754 4755 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4756 static inline void netif_trans_update(struct net_device *dev) 4757 { 4758 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4759 4760 txq_trans_cond_update(txq); 4761 } 4762 4763 /** 4764 * netif_tx_lock - grab network device transmit lock 4765 * @dev: network device 4766 * 4767 * Get network device transmit lock 4768 */ 4769 void netif_tx_lock(struct net_device *dev); 4770 4771 static inline void netif_tx_lock_bh(struct net_device *dev) 4772 { 4773 local_bh_disable(); 4774 netif_tx_lock(dev); 4775 } 4776 4777 void netif_tx_unlock(struct net_device *dev); 4778 4779 static inline void netif_tx_unlock_bh(struct net_device *dev) 4780 { 4781 netif_tx_unlock(dev); 4782 local_bh_enable(); 4783 } 4784 4785 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4786 if (!(dev)->lltx) { \ 4787 __netif_tx_lock(txq, cpu); \ 4788 } else { \ 4789 __netif_tx_acquire(txq); \ 4790 } \ 4791 } 4792 4793 #define HARD_TX_TRYLOCK(dev, txq) \ 4794 (!(dev)->lltx ? \ 4795 __netif_tx_trylock(txq) : \ 4796 __netif_tx_acquire(txq)) 4797 4798 #define HARD_TX_UNLOCK(dev, txq) { \ 4799 if (!(dev)->lltx) { \ 4800 __netif_tx_unlock(txq); \ 4801 } else { \ 4802 __netif_tx_release(txq); \ 4803 } \ 4804 } 4805 4806 static inline void netif_tx_disable(struct net_device *dev) 4807 { 4808 unsigned int i; 4809 int cpu; 4810 4811 local_bh_disable(); 4812 cpu = smp_processor_id(); 4813 spin_lock(&dev->tx_global_lock); 4814 for (i = 0; i < dev->num_tx_queues; i++) { 4815 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4816 4817 __netif_tx_lock(txq, cpu); 4818 netif_tx_stop_queue(txq); 4819 __netif_tx_unlock(txq); 4820 } 4821 spin_unlock(&dev->tx_global_lock); 4822 local_bh_enable(); 4823 } 4824 4825 static inline void netif_addr_lock(struct net_device *dev) 4826 { 4827 unsigned char nest_level = 0; 4828 4829 #ifdef CONFIG_LOCKDEP 4830 nest_level = dev->nested_level; 4831 #endif 4832 spin_lock_nested(&dev->addr_list_lock, nest_level); 4833 } 4834 4835 static inline void netif_addr_lock_bh(struct net_device *dev) 4836 { 4837 unsigned char nest_level = 0; 4838 4839 #ifdef CONFIG_LOCKDEP 4840 nest_level = dev->nested_level; 4841 #endif 4842 local_bh_disable(); 4843 spin_lock_nested(&dev->addr_list_lock, nest_level); 4844 } 4845 4846 static inline void netif_addr_unlock(struct net_device *dev) 4847 { 4848 spin_unlock(&dev->addr_list_lock); 4849 } 4850 4851 static inline void netif_addr_unlock_bh(struct net_device *dev) 4852 { 4853 spin_unlock_bh(&dev->addr_list_lock); 4854 } 4855 4856 /* 4857 * dev_addrs walker. Should be used only for read access. Call with 4858 * rcu_read_lock held. 4859 */ 4860 #define for_each_dev_addr(dev, ha) \ 4861 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4862 4863 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4864 4865 void ether_setup(struct net_device *dev); 4866 4867 /* Allocate dummy net_device */ 4868 struct net_device *alloc_netdev_dummy(int sizeof_priv); 4869 4870 /* Support for loadable net-drivers */ 4871 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4872 unsigned char name_assign_type, 4873 void (*setup)(struct net_device *), 4874 unsigned int txqs, unsigned int rxqs); 4875 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4876 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4877 4878 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4879 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4880 count) 4881 4882 int register_netdev(struct net_device *dev); 4883 void unregister_netdev(struct net_device *dev); 4884 4885 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4886 4887 /* General hardware address lists handling functions */ 4888 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4889 struct netdev_hw_addr_list *from_list, int addr_len); 4890 int __hw_addr_sync_multiple(struct netdev_hw_addr_list *to_list, 4891 struct netdev_hw_addr_list *from_list, 4892 int addr_len); 4893 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4894 struct netdev_hw_addr_list *from_list, int addr_len); 4895 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4896 struct net_device *dev, 4897 int (*sync)(struct net_device *, const unsigned char *), 4898 int (*unsync)(struct net_device *, 4899 const unsigned char *)); 4900 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4901 struct net_device *dev, 4902 int (*sync)(struct net_device *, 4903 const unsigned char *, int), 4904 int (*unsync)(struct net_device *, 4905 const unsigned char *, int)); 4906 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4907 struct net_device *dev, 4908 int (*unsync)(struct net_device *, 4909 const unsigned char *, int)); 4910 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4911 struct net_device *dev, 4912 int (*unsync)(struct net_device *, 4913 const unsigned char *)); 4914 void __hw_addr_init(struct netdev_hw_addr_list *list); 4915 4916 /* Functions used for device addresses handling */ 4917 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4918 const void *addr, size_t len); 4919 4920 static inline void 4921 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4922 { 4923 dev_addr_mod(dev, 0, addr, len); 4924 } 4925 4926 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4927 { 4928 __dev_addr_set(dev, addr, dev->addr_len); 4929 } 4930 4931 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4932 unsigned char addr_type); 4933 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4934 unsigned char addr_type); 4935 4936 /* Functions used for unicast addresses handling */ 4937 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4938 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4939 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4940 int dev_uc_sync(struct net_device *to, struct net_device *from); 4941 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4942 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4943 void dev_uc_flush(struct net_device *dev); 4944 void dev_uc_init(struct net_device *dev); 4945 4946 /** 4947 * __dev_uc_sync - Synchronize device's unicast list 4948 * @dev: device to sync 4949 * @sync: function to call if address should be added 4950 * @unsync: function to call if address should be removed 4951 * 4952 * Add newly added addresses to the interface, and release 4953 * addresses that have been deleted. 4954 */ 4955 static inline int __dev_uc_sync(struct net_device *dev, 4956 int (*sync)(struct net_device *, 4957 const unsigned char *), 4958 int (*unsync)(struct net_device *, 4959 const unsigned char *)) 4960 { 4961 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4962 } 4963 4964 /** 4965 * __dev_uc_unsync - Remove synchronized addresses from device 4966 * @dev: device to sync 4967 * @unsync: function to call if address should be removed 4968 * 4969 * Remove all addresses that were added to the device by dev_uc_sync(). 4970 */ 4971 static inline void __dev_uc_unsync(struct net_device *dev, 4972 int (*unsync)(struct net_device *, 4973 const unsigned char *)) 4974 { 4975 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4976 } 4977 4978 /* Functions used for multicast addresses handling */ 4979 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4980 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4981 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4982 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4983 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4984 int dev_mc_sync(struct net_device *to, struct net_device *from); 4985 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4986 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4987 void dev_mc_flush(struct net_device *dev); 4988 void dev_mc_init(struct net_device *dev); 4989 4990 /** 4991 * __dev_mc_sync - Synchronize device's multicast list 4992 * @dev: device to sync 4993 * @sync: function to call if address should be added 4994 * @unsync: function to call if address should be removed 4995 * 4996 * Add newly added addresses to the interface, and release 4997 * addresses that have been deleted. 4998 */ 4999 static inline int __dev_mc_sync(struct net_device *dev, 5000 int (*sync)(struct net_device *, 5001 const unsigned char *), 5002 int (*unsync)(struct net_device *, 5003 const unsigned char *)) 5004 { 5005 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 5006 } 5007 5008 /** 5009 * __dev_mc_unsync - Remove synchronized addresses from device 5010 * @dev: device to sync 5011 * @unsync: function to call if address should be removed 5012 * 5013 * Remove all addresses that were added to the device by dev_mc_sync(). 5014 */ 5015 static inline void __dev_mc_unsync(struct net_device *dev, 5016 int (*unsync)(struct net_device *, 5017 const unsigned char *)) 5018 { 5019 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 5020 } 5021 5022 /* Functions used for secondary unicast and multicast support */ 5023 void dev_set_rx_mode(struct net_device *dev); 5024 int dev_set_promiscuity(struct net_device *dev, int inc); 5025 int netif_set_allmulti(struct net_device *dev, int inc, bool notify); 5026 int dev_set_allmulti(struct net_device *dev, int inc); 5027 void netdev_state_change(struct net_device *dev); 5028 void __netdev_notify_peers(struct net_device *dev); 5029 void netdev_notify_peers(struct net_device *dev); 5030 void netdev_features_change(struct net_device *dev); 5031 /* Load a device via the kmod */ 5032 void dev_load(struct net *net, const char *name); 5033 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 5034 struct rtnl_link_stats64 *storage); 5035 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 5036 const struct net_device_stats *netdev_stats); 5037 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 5038 const struct pcpu_sw_netstats __percpu *netstats); 5039 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 5040 5041 enum { 5042 NESTED_SYNC_IMM_BIT, 5043 NESTED_SYNC_TODO_BIT, 5044 }; 5045 5046 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 5047 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 5048 5049 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 5050 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 5051 5052 struct netdev_nested_priv { 5053 unsigned char flags; 5054 void *data; 5055 }; 5056 5057 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 5058 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 5059 struct list_head **iter); 5060 5061 /* iterate through upper list, must be called under RCU read lock */ 5062 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 5063 for (iter = &(dev)->adj_list.upper, \ 5064 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 5065 updev; \ 5066 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 5067 5068 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 5069 int (*fn)(struct net_device *upper_dev, 5070 struct netdev_nested_priv *priv), 5071 struct netdev_nested_priv *priv); 5072 5073 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 5074 struct net_device *upper_dev); 5075 5076 bool netdev_has_any_upper_dev(struct net_device *dev); 5077 5078 void *netdev_lower_get_next_private(struct net_device *dev, 5079 struct list_head **iter); 5080 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 5081 struct list_head **iter); 5082 5083 #define netdev_for_each_lower_private(dev, priv, iter) \ 5084 for (iter = (dev)->adj_list.lower.next, \ 5085 priv = netdev_lower_get_next_private(dev, &(iter)); \ 5086 priv; \ 5087 priv = netdev_lower_get_next_private(dev, &(iter))) 5088 5089 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 5090 for (iter = &(dev)->adj_list.lower, \ 5091 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 5092 priv; \ 5093 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 5094 5095 void *netdev_lower_get_next(struct net_device *dev, 5096 struct list_head **iter); 5097 5098 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 5099 for (iter = (dev)->adj_list.lower.next, \ 5100 ldev = netdev_lower_get_next(dev, &(iter)); \ 5101 ldev; \ 5102 ldev = netdev_lower_get_next(dev, &(iter))) 5103 5104 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 5105 struct list_head **iter); 5106 int netdev_walk_all_lower_dev(struct net_device *dev, 5107 int (*fn)(struct net_device *lower_dev, 5108 struct netdev_nested_priv *priv), 5109 struct netdev_nested_priv *priv); 5110 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 5111 int (*fn)(struct net_device *lower_dev, 5112 struct netdev_nested_priv *priv), 5113 struct netdev_nested_priv *priv); 5114 5115 void *netdev_adjacent_get_private(struct list_head *adj_list); 5116 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 5117 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 5118 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 5119 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 5120 struct netlink_ext_ack *extack); 5121 int netdev_master_upper_dev_link(struct net_device *dev, 5122 struct net_device *upper_dev, 5123 void *upper_priv, void *upper_info, 5124 struct netlink_ext_ack *extack); 5125 void netdev_upper_dev_unlink(struct net_device *dev, 5126 struct net_device *upper_dev); 5127 int netdev_adjacent_change_prepare(struct net_device *old_dev, 5128 struct net_device *new_dev, 5129 struct net_device *dev, 5130 struct netlink_ext_ack *extack); 5131 void netdev_adjacent_change_commit(struct net_device *old_dev, 5132 struct net_device *new_dev, 5133 struct net_device *dev); 5134 void netdev_adjacent_change_abort(struct net_device *old_dev, 5135 struct net_device *new_dev, 5136 struct net_device *dev); 5137 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 5138 void *netdev_lower_dev_get_private(struct net_device *dev, 5139 struct net_device *lower_dev); 5140 void netdev_lower_state_changed(struct net_device *lower_dev, 5141 void *lower_state_info); 5142 5143 /* RSS keys are 40 or 52 bytes long */ 5144 #define NETDEV_RSS_KEY_LEN 52 5145 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 5146 void netdev_rss_key_fill(void *buffer, size_t len); 5147 5148 int skb_checksum_help(struct sk_buff *skb); 5149 int skb_crc32c_csum_help(struct sk_buff *skb); 5150 int skb_csum_hwoffload_help(struct sk_buff *skb, 5151 const netdev_features_t features); 5152 5153 struct netdev_bonding_info { 5154 ifslave slave; 5155 ifbond master; 5156 }; 5157 5158 struct netdev_notifier_bonding_info { 5159 struct netdev_notifier_info info; /* must be first */ 5160 struct netdev_bonding_info bonding_info; 5161 }; 5162 5163 void netdev_bonding_info_change(struct net_device *dev, 5164 struct netdev_bonding_info *bonding_info); 5165 5166 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 5167 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 5168 #else 5169 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 5170 const void *data) 5171 { 5172 } 5173 #endif 5174 5175 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 5176 5177 static inline bool can_checksum_protocol(netdev_features_t features, 5178 __be16 protocol) 5179 { 5180 if (protocol == htons(ETH_P_FCOE)) 5181 return !!(features & NETIF_F_FCOE_CRC); 5182 5183 /* Assume this is an IP checksum (not SCTP CRC) */ 5184 5185 if (features & NETIF_F_HW_CSUM) { 5186 /* Can checksum everything */ 5187 return true; 5188 } 5189 5190 switch (protocol) { 5191 case htons(ETH_P_IP): 5192 return !!(features & NETIF_F_IP_CSUM); 5193 case htons(ETH_P_IPV6): 5194 return !!(features & NETIF_F_IPV6_CSUM); 5195 default: 5196 return false; 5197 } 5198 } 5199 5200 #ifdef CONFIG_BUG 5201 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 5202 #else 5203 static inline void netdev_rx_csum_fault(struct net_device *dev, 5204 struct sk_buff *skb) 5205 { 5206 } 5207 #endif 5208 /* rx skb timestamps */ 5209 void net_enable_timestamp(void); 5210 void net_disable_timestamp(void); 5211 5212 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 5213 const struct skb_shared_hwtstamps *hwtstamps, 5214 bool cycles) 5215 { 5216 const struct net_device_ops *ops = dev->netdev_ops; 5217 5218 if (ops->ndo_get_tstamp) 5219 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 5220 5221 return hwtstamps->hwtstamp; 5222 } 5223 5224 #ifndef CONFIG_PREEMPT_RT 5225 static inline void netdev_xmit_set_more(bool more) 5226 { 5227 __this_cpu_write(softnet_data.xmit.more, more); 5228 } 5229 5230 static inline bool netdev_xmit_more(void) 5231 { 5232 return __this_cpu_read(softnet_data.xmit.more); 5233 } 5234 #else 5235 static inline void netdev_xmit_set_more(bool more) 5236 { 5237 current->net_xmit.more = more; 5238 } 5239 5240 static inline bool netdev_xmit_more(void) 5241 { 5242 return current->net_xmit.more; 5243 } 5244 #endif 5245 5246 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 5247 struct sk_buff *skb, struct net_device *dev, 5248 bool more) 5249 { 5250 netdev_xmit_set_more(more); 5251 return ops->ndo_start_xmit(skb, dev); 5252 } 5253 5254 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 5255 struct netdev_queue *txq, bool more) 5256 { 5257 const struct net_device_ops *ops = dev->netdev_ops; 5258 netdev_tx_t rc; 5259 5260 rc = __netdev_start_xmit(ops, skb, dev, more); 5261 if (rc == NETDEV_TX_OK) 5262 txq_trans_update(txq); 5263 5264 return rc; 5265 } 5266 5267 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 5268 const void *ns); 5269 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 5270 const void *ns); 5271 5272 extern const struct kobj_ns_type_operations net_ns_type_operations; 5273 5274 const char *netdev_drivername(const struct net_device *dev); 5275 5276 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 5277 netdev_features_t f2) 5278 { 5279 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 5280 if (f1 & NETIF_F_HW_CSUM) 5281 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5282 else 5283 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5284 } 5285 5286 return f1 & f2; 5287 } 5288 5289 static inline netdev_features_t netdev_get_wanted_features( 5290 struct net_device *dev) 5291 { 5292 return (dev->features & ~dev->hw_features) | dev->wanted_features; 5293 } 5294 netdev_features_t netdev_increment_features(netdev_features_t all, 5295 netdev_features_t one, netdev_features_t mask); 5296 5297 /* Allow TSO being used on stacked device : 5298 * Performing the GSO segmentation before last device 5299 * is a performance improvement. 5300 */ 5301 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 5302 netdev_features_t mask) 5303 { 5304 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 5305 } 5306 5307 int __netdev_update_features(struct net_device *dev); 5308 void netdev_update_features(struct net_device *dev); 5309 void netdev_change_features(struct net_device *dev); 5310 5311 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 5312 struct net_device *dev); 5313 5314 netdev_features_t passthru_features_check(struct sk_buff *skb, 5315 struct net_device *dev, 5316 netdev_features_t features); 5317 netdev_features_t netif_skb_features(struct sk_buff *skb); 5318 void skb_warn_bad_offload(const struct sk_buff *skb); 5319 5320 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 5321 { 5322 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 5323 5324 /* check flags correspondence */ 5325 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 5326 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 5327 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 5328 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 5329 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 5330 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 5331 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 5332 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 5333 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 5334 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 5335 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 5336 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 5337 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 5338 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 5339 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 5340 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 5341 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 5342 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 5343 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 5344 5345 return (features & feature) == feature; 5346 } 5347 5348 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 5349 { 5350 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 5351 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 5352 } 5353 5354 static inline bool netif_needs_gso(struct sk_buff *skb, 5355 netdev_features_t features) 5356 { 5357 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 5358 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 5359 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 5360 } 5361 5362 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 5363 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 5364 void netif_inherit_tso_max(struct net_device *to, 5365 const struct net_device *from); 5366 5367 static inline unsigned int 5368 netif_get_gro_max_size(const struct net_device *dev, const struct sk_buff *skb) 5369 { 5370 /* pairs with WRITE_ONCE() in netif_set_gro(_ipv4)_max_size() */ 5371 return skb->protocol == htons(ETH_P_IPV6) ? 5372 READ_ONCE(dev->gro_max_size) : 5373 READ_ONCE(dev->gro_ipv4_max_size); 5374 } 5375 5376 static inline unsigned int 5377 netif_get_gso_max_size(const struct net_device *dev, const struct sk_buff *skb) 5378 { 5379 /* pairs with WRITE_ONCE() in netif_set_gso(_ipv4)_max_size() */ 5380 return skb->protocol == htons(ETH_P_IPV6) ? 5381 READ_ONCE(dev->gso_max_size) : 5382 READ_ONCE(dev->gso_ipv4_max_size); 5383 } 5384 5385 static inline bool netif_is_macsec(const struct net_device *dev) 5386 { 5387 return dev->priv_flags & IFF_MACSEC; 5388 } 5389 5390 static inline bool netif_is_macvlan(const struct net_device *dev) 5391 { 5392 return dev->priv_flags & IFF_MACVLAN; 5393 } 5394 5395 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5396 { 5397 return dev->priv_flags & IFF_MACVLAN_PORT; 5398 } 5399 5400 static inline bool netif_is_bond_master(const struct net_device *dev) 5401 { 5402 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5403 } 5404 5405 static inline bool netif_is_bond_slave(const struct net_device *dev) 5406 { 5407 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5408 } 5409 5410 static inline bool netif_supports_nofcs(struct net_device *dev) 5411 { 5412 return dev->priv_flags & IFF_SUPP_NOFCS; 5413 } 5414 5415 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5416 { 5417 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5418 } 5419 5420 static inline bool netif_is_l3_master(const struct net_device *dev) 5421 { 5422 return dev->priv_flags & IFF_L3MDEV_MASTER; 5423 } 5424 5425 static inline bool netif_is_l3_slave(const struct net_device *dev) 5426 { 5427 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5428 } 5429 5430 static inline int dev_sdif(const struct net_device *dev) 5431 { 5432 #ifdef CONFIG_NET_L3_MASTER_DEV 5433 if (netif_is_l3_slave(dev)) 5434 return dev->ifindex; 5435 #endif 5436 return 0; 5437 } 5438 5439 static inline bool netif_is_bridge_master(const struct net_device *dev) 5440 { 5441 return dev->priv_flags & IFF_EBRIDGE; 5442 } 5443 5444 static inline bool netif_is_bridge_port(const struct net_device *dev) 5445 { 5446 return dev->priv_flags & IFF_BRIDGE_PORT; 5447 } 5448 5449 static inline bool netif_is_ovs_master(const struct net_device *dev) 5450 { 5451 return dev->priv_flags & IFF_OPENVSWITCH; 5452 } 5453 5454 static inline bool netif_is_ovs_port(const struct net_device *dev) 5455 { 5456 return dev->priv_flags & IFF_OVS_DATAPATH; 5457 } 5458 5459 static inline bool netif_is_any_bridge_master(const struct net_device *dev) 5460 { 5461 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev); 5462 } 5463 5464 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5465 { 5466 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5467 } 5468 5469 static inline bool netif_is_team_master(const struct net_device *dev) 5470 { 5471 return dev->priv_flags & IFF_TEAM; 5472 } 5473 5474 static inline bool netif_is_team_port(const struct net_device *dev) 5475 { 5476 return dev->priv_flags & IFF_TEAM_PORT; 5477 } 5478 5479 static inline bool netif_is_lag_master(const struct net_device *dev) 5480 { 5481 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5482 } 5483 5484 static inline bool netif_is_lag_port(const struct net_device *dev) 5485 { 5486 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5487 } 5488 5489 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5490 { 5491 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5492 } 5493 5494 static inline bool netif_is_failover(const struct net_device *dev) 5495 { 5496 return dev->priv_flags & IFF_FAILOVER; 5497 } 5498 5499 static inline bool netif_is_failover_slave(const struct net_device *dev) 5500 { 5501 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5502 } 5503 5504 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5505 static inline void netif_keep_dst(struct net_device *dev) 5506 { 5507 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5508 } 5509 5510 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5511 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5512 { 5513 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5514 return netif_is_macsec(dev); 5515 } 5516 5517 extern struct pernet_operations __net_initdata loopback_net_ops; 5518 5519 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5520 5521 /* netdev_printk helpers, similar to dev_printk */ 5522 5523 static inline const char *netdev_name(const struct net_device *dev) 5524 { 5525 if (!dev->name[0] || strchr(dev->name, '%')) 5526 return "(unnamed net_device)"; 5527 return dev->name; 5528 } 5529 5530 static inline const char *netdev_reg_state(const struct net_device *dev) 5531 { 5532 u8 reg_state = READ_ONCE(dev->reg_state); 5533 5534 switch (reg_state) { 5535 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5536 case NETREG_REGISTERED: return ""; 5537 case NETREG_UNREGISTERING: return " (unregistering)"; 5538 case NETREG_UNREGISTERED: return " (unregistered)"; 5539 case NETREG_RELEASED: return " (released)"; 5540 case NETREG_DUMMY: return " (dummy)"; 5541 } 5542 5543 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state); 5544 return " (unknown)"; 5545 } 5546 5547 #define MODULE_ALIAS_NETDEV(device) \ 5548 MODULE_ALIAS("netdev-" device) 5549 5550 /* 5551 * netdev_WARN() acts like dev_printk(), but with the key difference 5552 * of using a WARN/WARN_ON to get the message out, including the 5553 * file/line information and a backtrace. 5554 */ 5555 #define netdev_WARN(dev, format, args...) \ 5556 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5557 netdev_reg_state(dev), ##args) 5558 5559 #define netdev_WARN_ONCE(dev, format, args...) \ 5560 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5561 netdev_reg_state(dev), ##args) 5562 5563 /* 5564 * The list of packet types we will receive (as opposed to discard) 5565 * and the routines to invoke. 5566 * 5567 * Why 16. Because with 16 the only overlap we get on a hash of the 5568 * low nibble of the protocol value is RARP/SNAP/X.25. 5569 * 5570 * 0800 IP 5571 * 0001 802.3 5572 * 0002 AX.25 5573 * 0004 802.2 5574 * 8035 RARP 5575 * 0005 SNAP 5576 * 0805 X.25 5577 * 0806 ARP 5578 * 8137 IPX 5579 * 0009 Localtalk 5580 * 86DD IPv6 5581 */ 5582 #define PTYPE_HASH_SIZE (16) 5583 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5584 5585 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5586 5587 extern struct net_device *blackhole_netdev; 5588 5589 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */ 5590 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD) 5591 #define DEV_STATS_ADD(DEV, FIELD, VAL) \ 5592 atomic_long_add((VAL), &(DEV)->stats.__##FIELD) 5593 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD) 5594 5595 #endif /* _LINUX_NETDEVICE_H */ 5596