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, <waltje@uWalt.NL.Mugnet.ORG> 13 * Corey Minyard <wf-rch!minyard@relay.EU.net> 14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov> 15 * Alan Cox, <alan@lxorguk.ukuu.org.uk> 16 * Bjorn Ekwall. <bj0rn@blox.se> 17 * Pekka Riikonen <priikone@poseidon.pspt.fi> 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) 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 /* Only changed through the __hw_addr_count_* helpers after 256 * __hw_addr_init() 257 */ 258 int _count; 259 260 /* Auxiliary tree for faster lookup on addition and deletion */ 261 struct rb_root tree; 262 263 /* Set for dev->mc only, the device the list belongs to */ 264 struct net_device *mc_dev; 265 }; 266 267 #define netdev_hw_addr_list_count(l) ((l)->_count) 268 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 269 #define netdev_hw_addr_list_for_each(ha, l) \ 270 list_for_each_entry(ha, &(l)->list, list) 271 272 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 273 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 274 #define netdev_for_each_uc_addr(ha, dev) \ 275 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 276 #define netdev_for_each_synced_uc_addr(_ha, _dev) \ 277 netdev_for_each_uc_addr((_ha), (_dev)) \ 278 if ((_ha)->sync_cnt) 279 280 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 281 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 282 #define netdev_for_each_mc_addr(ha, dev) \ 283 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 284 #define netdev_for_each_synced_mc_addr(_ha, _dev) \ 285 netdev_for_each_mc_addr((_ha), (_dev)) \ 286 if ((_ha)->sync_cnt) 287 288 struct hh_cache { 289 unsigned int hh_len; 290 seqlock_t hh_lock; 291 292 /* cached hardware header; allow for machine alignment needs. */ 293 #define HH_DATA_MOD 16 294 #define HH_DATA_OFF(__len) \ 295 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 296 #define HH_DATA_ALIGN(__len) \ 297 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 298 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 299 }; 300 301 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 302 * Alternative is: 303 * dev->hard_header_len ? (dev->hard_header_len + 304 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 305 * 306 * We could use other alignment values, but we must maintain the 307 * relationship HH alignment <= LL alignment. 308 */ 309 #define LL_RESERVED_SPACE_EX(dev, hlen) \ 310 ((((hlen) + READ_ONCE((dev)->needed_headroom)) \ 311 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 312 #define LL_RESERVED_SPACE(dev) \ 313 LL_RESERVED_SPACE_EX(dev, (dev)->hard_header_len) 314 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 315 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \ 316 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 317 318 struct header_ops { 319 int (*create) (struct sk_buff *skb, struct net_device *dev, 320 unsigned short type, const void *daddr, 321 const void *saddr, unsigned int len); 322 int (*parse)(const struct sk_buff *skb, 323 const struct net_device *dev, 324 unsigned char *haddr); 325 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 326 void (*cache_update)(struct hh_cache *hh, 327 const struct net_device *dev, 328 const unsigned char *haddr); 329 bool (*validate)(const char *ll_header, unsigned int len); 330 __be16 (*parse_protocol)(const struct sk_buff *skb); 331 }; 332 333 /* These flag bits are private to the generic network queueing 334 * layer; they may not be explicitly referenced by any other 335 * code. 336 */ 337 338 enum netdev_state_t { 339 __LINK_STATE_START, 340 __LINK_STATE_PRESENT, 341 __LINK_STATE_NOCARRIER, 342 __LINK_STATE_LINKWATCH_PENDING, 343 __LINK_STATE_DORMANT, 344 __LINK_STATE_TESTING, 345 }; 346 347 struct gro_list { 348 struct list_head list; 349 int count; 350 }; 351 352 /* 353 * size of gro hash buckets, must be <= the number of bits in 354 * gro_node::bitmask 355 */ 356 #define GRO_HASH_BUCKETS 8 357 358 /** 359 * struct gro_node - structure to support Generic Receive Offload 360 * @bitmask: bitmask to indicate used buckets in @hash 361 * @hash: hashtable of pending aggregated skbs, separated by flows 362 * @rx_list: list of pending ``GRO_NORMAL`` skbs 363 * @rx_count: cached current length of @rx_list 364 * @cached_napi_id: napi_struct::napi_id cached for hotpath, 0 for standalone 365 */ 366 struct gro_node { 367 unsigned long bitmask; 368 struct gro_list hash[GRO_HASH_BUCKETS]; 369 struct list_head rx_list; 370 u32 rx_count; 371 u32 cached_napi_id; 372 }; 373 374 /* 375 * Structure for per-NAPI config 376 */ 377 struct napi_config { 378 u64 gro_flush_timeout; 379 u64 irq_suspend_timeout; 380 u32 defer_hard_irqs; 381 cpumask_t affinity_mask; 382 u8 threaded; 383 unsigned int napi_id; 384 }; 385 386 /* 387 * Structure for NAPI scheduling similar to tasklet but with weighting 388 */ 389 struct napi_struct { 390 /* This field should be first or softnet_data.backlog needs tweaks. */ 391 unsigned long state; 392 /* The poll_list must only be managed by the entity which 393 * changes the state of the NAPI_STATE_SCHED bit. This means 394 * whoever atomically sets that bit can add this napi_struct 395 * to the per-CPU poll_list, and whoever clears that bit 396 * can remove from the list right before clearing the bit. 397 */ 398 struct list_head poll_list; 399 400 int weight; 401 u32 defer_hard_irqs_count; 402 int (*poll)(struct napi_struct *, int); 403 #ifdef CONFIG_NETPOLL 404 /* CPU actively polling if netpoll is configured */ 405 int poll_owner; 406 #endif 407 /* CPU on which NAPI has been scheduled for processing */ 408 int list_owner; 409 struct net_device *dev; 410 struct sk_buff *skb; 411 struct gro_node gro; 412 struct hrtimer timer; 413 /* all fields past this point are write-protected by netdev_lock */ 414 struct task_struct *thread; 415 unsigned long gro_flush_timeout; 416 unsigned long irq_suspend_timeout; 417 u32 defer_hard_irqs; 418 /* control-path-only fields follow */ 419 u32 napi_id; 420 struct list_head dev_list; 421 struct hlist_node napi_hash_node; 422 int irq; 423 struct irq_affinity_notify notify; 424 int napi_rmap_idx; 425 int index; 426 struct napi_config *config; 427 }; 428 429 enum { 430 NAPI_STATE_SCHED, /* Poll is scheduled */ 431 NAPI_STATE_MISSED, /* reschedule a napi */ 432 NAPI_STATE_DISABLE, /* Disable pending */ 433 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 434 NAPI_STATE_LISTED, /* NAPI added to system lists */ 435 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 436 NAPI_STATE_IN_BUSY_POLL, /* Do not rearm NAPI interrupt */ 437 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 438 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 439 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 440 NAPI_STATE_HAS_NOTIFIER, /* Napi has an IRQ notifier */ 441 NAPI_STATE_THREADED_BUSY_POLL, /* The threaded NAPI poller will busy poll */ 442 NAPI_STATE_PERCPU, /* NAPI handler is run per-CPU - incompatible with threaded mode */ 443 }; 444 445 enum { 446 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 447 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 448 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 449 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 450 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 451 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 452 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 453 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 454 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 455 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 456 NAPIF_STATE_HAS_NOTIFIER = BIT(NAPI_STATE_HAS_NOTIFIER), 457 NAPIF_STATE_THREADED_BUSY_POLL = BIT(NAPI_STATE_THREADED_BUSY_POLL), 458 NAPIF_STATE_PERCPU = BIT(NAPI_STATE_PERCPU), 459 }; 460 461 enum gro_result { 462 GRO_MERGED, 463 GRO_MERGED_FREE, 464 GRO_HELD, 465 GRO_NORMAL, 466 GRO_CONSUMED, 467 }; 468 typedef enum gro_result gro_result_t; 469 470 /* 471 * enum rx_handler_result - Possible return values for rx_handlers. 472 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 473 * further. 474 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 475 * case skb->dev was changed by rx_handler. 476 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 477 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 478 * 479 * rx_handlers are functions called from inside __netif_receive_skb(), to do 480 * special processing of the skb, prior to delivery to protocol handlers. 481 * 482 * Currently, a net_device can only have a single rx_handler registered. Trying 483 * to register a second rx_handler will return -EBUSY. 484 * 485 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 486 * To unregister a rx_handler on a net_device, use 487 * netdev_rx_handler_unregister(). 488 * 489 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 490 * do with the skb. 491 * 492 * If the rx_handler consumed the skb in some way, it should return 493 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 494 * the skb to be delivered in some other way. 495 * 496 * If the rx_handler changed skb->dev, to divert the skb to another 497 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 498 * new device will be called if it exists. 499 * 500 * If the rx_handler decides the skb should be ignored, it should return 501 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 502 * are registered on exact device (ptype->dev == skb->dev). 503 * 504 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 505 * delivered, it should return RX_HANDLER_PASS. 506 * 507 * A device without a registered rx_handler will behave as if rx_handler 508 * returned RX_HANDLER_PASS. 509 */ 510 511 enum rx_handler_result { 512 RX_HANDLER_CONSUMED, 513 RX_HANDLER_ANOTHER, 514 RX_HANDLER_EXACT, 515 RX_HANDLER_PASS, 516 }; 517 typedef enum rx_handler_result rx_handler_result_t; 518 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 519 520 void __napi_schedule(struct napi_struct *n); 521 void __napi_schedule_irqoff(struct napi_struct *n); 522 523 static inline bool napi_disable_pending(struct napi_struct *n) 524 { 525 return test_bit(NAPI_STATE_DISABLE, &n->state); 526 } 527 528 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 529 { 530 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 531 } 532 533 /** 534 * napi_is_scheduled - test if NAPI is scheduled 535 * @n: NAPI context 536 * 537 * This check is "best-effort". With no locking implemented, 538 * a NAPI can be scheduled or terminate right after this check 539 * and produce not precise results. 540 * 541 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled 542 * should not be used normally and napi_schedule should be 543 * used instead. 544 * 545 * Use only if the driver really needs to check if a NAPI 546 * is scheduled for example in the context of delayed timer 547 * that can be skipped if a NAPI is already scheduled. 548 * 549 * Return: True if NAPI is scheduled, False otherwise. 550 */ 551 static inline bool napi_is_scheduled(struct napi_struct *n) 552 { 553 return test_bit(NAPI_STATE_SCHED, &n->state); 554 } 555 556 bool napi_schedule_prep(struct napi_struct *n); 557 558 /** 559 * napi_schedule - schedule NAPI poll 560 * @n: NAPI context 561 * 562 * Schedule NAPI poll routine to be called if it is not already 563 * running. 564 * Return: true if we schedule a NAPI or false if not. 565 * Refer to napi_schedule_prep() for additional reason on why 566 * a NAPI might not be scheduled. 567 */ 568 static inline bool napi_schedule(struct napi_struct *n) 569 { 570 if (napi_schedule_prep(n)) { 571 __napi_schedule(n); 572 return true; 573 } 574 575 return false; 576 } 577 578 /** 579 * napi_schedule_irqoff - schedule NAPI poll 580 * @n: NAPI context 581 * 582 * Variant of napi_schedule(), assuming hard irqs are masked. 583 */ 584 static inline void napi_schedule_irqoff(struct napi_struct *n) 585 { 586 if (napi_schedule_prep(n)) 587 __napi_schedule_irqoff(n); 588 } 589 590 /** 591 * napi_complete_done - NAPI processing complete 592 * @n: NAPI context 593 * @work_done: number of packets processed 594 * 595 * Mark NAPI processing as complete. Should only be called if poll budget 596 * has not been completely consumed. 597 * Prefer over napi_complete(). 598 * Return: false if device should avoid rearming interrupts. 599 */ 600 bool napi_complete_done(struct napi_struct *n, int work_done); 601 602 static inline bool napi_complete(struct napi_struct *n) 603 { 604 return napi_complete_done(n, 0); 605 } 606 607 void netif_threaded_enable(struct net_device *dev); 608 int dev_set_threaded(struct net_device *dev, 609 enum netdev_napi_threaded threaded); 610 611 void napi_disable(struct napi_struct *n); 612 void napi_disable_locked(struct napi_struct *n); 613 614 void napi_enable(struct napi_struct *n); 615 void napi_enable_locked(struct napi_struct *n); 616 617 /** 618 * napi_synchronize - wait until NAPI is not running 619 * @n: NAPI context 620 * 621 * Wait until NAPI is done being scheduled on this context. 622 * Waits till any outstanding processing completes but 623 * does not disable future activations. 624 */ 625 static inline void napi_synchronize(const struct napi_struct *n) 626 { 627 if (IS_ENABLED(CONFIG_SMP)) 628 while (test_bit(NAPI_STATE_SCHED, &n->state)) 629 msleep(1); 630 else 631 barrier(); 632 } 633 634 /** 635 * napi_if_scheduled_mark_missed - if napi is running, set the 636 * NAPIF_STATE_MISSED 637 * @n: NAPI context 638 * 639 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 640 * NAPI is scheduled. 641 **/ 642 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 643 { 644 unsigned long val, new; 645 646 val = READ_ONCE(n->state); 647 do { 648 if (val & NAPIF_STATE_DISABLE) 649 return true; 650 651 if (!(val & NAPIF_STATE_SCHED)) 652 return false; 653 654 new = val | NAPIF_STATE_MISSED; 655 } while (!try_cmpxchg(&n->state, &val, new)); 656 657 return true; 658 } 659 660 enum netdev_queue_state_t { 661 __QUEUE_STATE_DRV_XOFF, 662 __QUEUE_STATE_STACK_XOFF, 663 __QUEUE_STATE_FROZEN, 664 }; 665 666 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 667 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 668 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 669 670 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 671 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 672 QUEUE_STATE_FROZEN) 673 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 674 QUEUE_STATE_FROZEN) 675 676 /* 677 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 678 * netif_tx_* functions below are used to manipulate this flag. The 679 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 680 * queue independently. The netif_xmit_*stopped functions below are called 681 * to check if the queue has been stopped by the driver or stack (either 682 * of the XOFF bits are set in the state). Drivers should not need to call 683 * netif_xmit*stopped functions, they should only be using netif_tx_*. 684 */ 685 686 struct netdev_queue { 687 /* 688 * read-mostly part 689 */ 690 struct net_device *dev; 691 netdevice_tracker dev_tracker; 692 693 struct Qdisc __rcu *qdisc; 694 struct Qdisc __rcu *qdisc_sleeping; 695 #ifdef CONFIG_SYSFS 696 struct kobject kobj; 697 const struct attribute_group **groups; 698 #endif 699 unsigned long tx_maxrate; 700 /* 701 * Number of TX timeouts for this queue 702 * (/sys/class/net/DEV/Q/trans_timeout) 703 */ 704 atomic_long_t trans_timeout; 705 706 /* Subordinate device that the queue has been assigned to */ 707 struct net_device *sb_dev; 708 #ifdef CONFIG_XDP_SOCKETS 709 /* "ops protected", see comment about net_device::lock */ 710 struct xsk_buff_pool *pool; 711 #endif 712 713 /* 714 * write-mostly part 715 */ 716 #ifdef CONFIG_BQL 717 struct dql dql; 718 #endif 719 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 720 int xmit_lock_owner; 721 /* 722 * Time (in jiffies) of last Tx 723 */ 724 unsigned long trans_start; 725 726 unsigned long state; 727 728 /* 729 * slow- / control-path part 730 */ 731 /* NAPI instance for the queue 732 * "ops protected", see comment about net_device::lock 733 */ 734 struct napi_struct *napi; 735 736 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 737 int numa_node; 738 #endif 739 } ____cacheline_aligned_in_smp; 740 741 extern int sysctl_fb_tunnels_only_for_init_net; 742 extern int sysctl_devconf_inherit_init_net; 743 744 /* 745 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 746 * == 1 : For initns only 747 * == 2 : For none. 748 */ 749 static inline bool net_has_fallback_tunnels(const struct net *net) 750 { 751 #if IS_ENABLED(CONFIG_SYSCTL) 752 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net); 753 754 return !fb_tunnels_only_for_init_net || 755 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1); 756 #else 757 return true; 758 #endif 759 } 760 761 static inline int net_inherit_devconf(void) 762 { 763 #if IS_ENABLED(CONFIG_SYSCTL) 764 return READ_ONCE(sysctl_devconf_inherit_init_net); 765 #else 766 return 0; 767 #endif 768 } 769 770 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 771 { 772 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 773 return q->numa_node; 774 #else 775 return NUMA_NO_NODE; 776 #endif 777 } 778 779 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 780 { 781 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 782 q->numa_node = node; 783 #endif 784 } 785 786 #ifdef CONFIG_RFS_ACCEL 787 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 788 u16 filter_id); 789 #endif 790 791 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 792 enum xps_map_type { 793 XPS_CPUS = 0, 794 XPS_RXQS, 795 XPS_MAPS_MAX, 796 }; 797 798 #ifdef CONFIG_XPS 799 /* 800 * This structure holds an XPS map which can be of variable length. The 801 * map is an array of queues. 802 */ 803 struct xps_map { 804 unsigned int len; 805 unsigned int alloc_len; 806 struct rcu_head rcu; 807 u16 queues[]; 808 }; 809 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 810 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 811 - sizeof(struct xps_map)) / sizeof(u16)) 812 813 /* 814 * This structure holds all XPS maps for device. Maps are indexed by CPU. 815 * 816 * We keep track of the number of cpus/rxqs used when the struct is allocated, 817 * in nr_ids. This will help not accessing out-of-bound memory. 818 * 819 * We keep track of the number of traffic classes used when the struct is 820 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 821 * not crossing its upper bound, as the original dev->num_tc can be updated in 822 * the meantime. 823 */ 824 struct xps_dev_maps { 825 struct rcu_head rcu; 826 unsigned int nr_ids; 827 s16 num_tc; 828 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 829 }; 830 831 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 832 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 833 834 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 835 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 836 837 #endif /* CONFIG_XPS */ 838 839 #define TC_MAX_QUEUE 16 840 #define TC_BITMASK 15 841 /* HW offloaded queuing disciplines txq count and offset maps */ 842 struct netdev_tc_txq { 843 union { 844 struct { 845 u16 count; 846 u16 offset; 847 }; 848 u32 combined; 849 }; 850 }; 851 852 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 853 /* 854 * This structure is to hold information about the device 855 * configured to run FCoE protocol stack. 856 */ 857 struct netdev_fcoe_hbainfo { 858 char manufacturer[64]; 859 char serial_number[64]; 860 char hardware_version[64]; 861 char driver_version[64]; 862 char optionrom_version[64]; 863 char firmware_version[64]; 864 char model[256]; 865 char model_description[256]; 866 }; 867 #endif 868 869 #define MAX_PHYS_ITEM_ID_LEN 32 870 871 /* This structure holds a unique identifier to identify some 872 * physical item (port for example) used by a netdevice. 873 */ 874 struct netdev_phys_item_id { 875 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 876 unsigned char id_len; 877 }; 878 879 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 880 struct netdev_phys_item_id *b) 881 { 882 return a->id_len == b->id_len && 883 memcmp(a->id, b->id, a->id_len) == 0; 884 } 885 886 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 887 struct sk_buff *skb, 888 struct net_device *sb_dev); 889 890 enum net_device_path_type { 891 DEV_PATH_ETHERNET = 0, 892 DEV_PATH_VLAN, 893 DEV_PATH_BRIDGE, 894 DEV_PATH_PPPOE, 895 DEV_PATH_DSA, 896 DEV_PATH_MTK_WDMA, 897 DEV_PATH_TUN, 898 DEV_PATH_IEEE80211, 899 }; 900 901 struct net_device_path { 902 enum net_device_path_type type; 903 const struct net_device *dev; 904 union { 905 struct { 906 u16 id; 907 __be16 proto; 908 u8 h_dest[ETH_ALEN]; 909 } encap; 910 struct { 911 struct dst_entry *dst; 912 union { 913 struct in_addr src_v4; 914 struct in6_addr src_v6; 915 }; 916 union { 917 struct in_addr dst_v4; 918 struct in6_addr dst_v6; 919 }; 920 921 u8 inner_proto; 922 u8 encap_proto; 923 } tun; 924 struct { 925 enum { 926 DEV_PATH_BR_VLAN_KEEP, 927 DEV_PATH_BR_VLAN_TAG, 928 DEV_PATH_BR_VLAN_UNTAG, 929 DEV_PATH_BR_VLAN_UNTAG_HW, 930 } vlan_mode; 931 u16 vlan_id; 932 __be16 vlan_proto; 933 } bridge; 934 struct { 935 int port; 936 u16 proto; 937 } dsa; 938 struct { 939 u8 wdma_idx; 940 u8 queue; 941 u16 wcid; 942 u8 bss; 943 u8 amsdu; 944 } mtk_wdma; 945 }; 946 }; 947 948 #define NET_DEVICE_PATH_STACK_MAX 5 949 #define NET_DEVICE_PATH_VLAN_MAX 2 950 951 struct net_device_path_stack { 952 int num_paths; 953 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 954 }; 955 956 struct net_device_path_ctx { 957 const struct net_device *dev; 958 u8 daddr[ETH_ALEN]; 959 __be16 ether_type; 960 961 int num_vlans; 962 struct { 963 u16 id; 964 __be16 proto; 965 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 966 967 bool ieee80211; 968 }; 969 970 enum tc_setup_type { 971 TC_QUERY_CAPS, 972 TC_SETUP_QDISC_MQPRIO, 973 TC_SETUP_CLSU32, 974 TC_SETUP_CLSFLOWER, 975 TC_SETUP_CLSMATCHALL, 976 TC_SETUP_CLSBPF, 977 TC_SETUP_BLOCK, 978 TC_SETUP_QDISC_CBS, 979 TC_SETUP_QDISC_RED, 980 TC_SETUP_QDISC_PRIO, 981 TC_SETUP_QDISC_MQ, 982 TC_SETUP_QDISC_ETF, 983 TC_SETUP_ROOT_QDISC, 984 TC_SETUP_QDISC_GRED, 985 TC_SETUP_QDISC_TAPRIO, 986 TC_SETUP_FT, 987 TC_SETUP_QDISC_ETS, 988 TC_SETUP_QDISC_TBF, 989 TC_SETUP_QDISC_FIFO, 990 TC_SETUP_QDISC_HTB, 991 TC_SETUP_ACT, 992 }; 993 994 /* These structures hold the attributes of bpf state that are being passed 995 * to the netdevice through the bpf op. 996 */ 997 enum bpf_netdev_command { 998 /* Set or clear a bpf program used in the earliest stages of packet 999 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 1000 * is responsible for calling bpf_prog_put on any old progs that are 1001 * stored. In case of error, the callee need not release the new prog 1002 * reference, but on success it takes ownership and must bpf_prog_put 1003 * when it is no longer used. 1004 */ 1005 XDP_SETUP_PROG, 1006 XDP_SETUP_PROG_HW, 1007 /* BPF program for offload callbacks, invoked at program load time. */ 1008 BPF_OFFLOAD_MAP_ALLOC, 1009 BPF_OFFLOAD_MAP_FREE, 1010 XDP_SETUP_XSK_POOL, 1011 }; 1012 1013 struct bpf_prog_offload_ops; 1014 struct netlink_ext_ack; 1015 struct xdp_umem; 1016 struct xdp_dev_bulk_queue; 1017 struct bpf_xdp_link; 1018 1019 enum bpf_xdp_mode { 1020 XDP_MODE_SKB = 0, 1021 XDP_MODE_DRV = 1, 1022 XDP_MODE_HW = 2, 1023 __MAX_XDP_MODE 1024 }; 1025 1026 struct bpf_xdp_entity { 1027 struct bpf_prog *prog; 1028 struct bpf_xdp_link *link; 1029 }; 1030 1031 struct netdev_bpf { 1032 enum bpf_netdev_command command; 1033 union { 1034 /* XDP_SETUP_PROG */ 1035 struct { 1036 u32 flags; 1037 struct bpf_prog *prog; 1038 struct netlink_ext_ack *extack; 1039 }; 1040 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 1041 struct { 1042 struct bpf_offloaded_map *offmap; 1043 }; 1044 /* XDP_SETUP_XSK_POOL */ 1045 struct { 1046 struct xsk_buff_pool *pool; 1047 u16 queue_id; 1048 } xsk; 1049 }; 1050 }; 1051 1052 /* Flags for ndo_xsk_wakeup. */ 1053 #define XDP_WAKEUP_RX (1 << 0) 1054 #define XDP_WAKEUP_TX (1 << 1) 1055 1056 #ifdef CONFIG_XFRM_OFFLOAD 1057 struct xfrmdev_ops { 1058 int (*xdo_dev_state_add)(struct net_device *dev, 1059 struct xfrm_state *x, 1060 struct netlink_ext_ack *extack); 1061 void (*xdo_dev_state_delete)(struct net_device *dev, 1062 struct xfrm_state *x); 1063 void (*xdo_dev_state_free)(struct net_device *dev, 1064 struct xfrm_state *x); 1065 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 1066 struct xfrm_state *x); 1067 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1068 void (*xdo_dev_state_update_stats) (struct xfrm_state *x); 1069 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack); 1070 void (*xdo_dev_policy_delete) (struct xfrm_policy *x); 1071 void (*xdo_dev_policy_free) (struct xfrm_policy *x); 1072 }; 1073 #endif 1074 1075 struct dev_ifalias { 1076 struct rcu_head rcuhead; 1077 char ifalias[]; 1078 }; 1079 1080 struct devlink; 1081 struct tlsdev_ops; 1082 1083 struct netdev_net_notifier { 1084 struct list_head list; 1085 struct notifier_block *nb; 1086 }; 1087 1088 /* 1089 * This structure defines the management hooks for network devices. 1090 * The following hooks can be defined; unless noted otherwise, they are 1091 * optional and can be filled with a null pointer. 1092 * 1093 * int (*ndo_init)(struct net_device *dev); 1094 * This function is called once when a network device is registered. 1095 * The network device can use this for any late stage initialization 1096 * or semantic validation. It can fail with an error code which will 1097 * be propagated back to register_netdev. 1098 * 1099 * void (*ndo_uninit)(struct net_device *dev); 1100 * This function is called when device is unregistered or when registration 1101 * fails. It is not called if init fails. 1102 * 1103 * int (*ndo_open)(struct net_device *dev); 1104 * This function is called when a network device transitions to the up 1105 * state. 1106 * 1107 * int (*ndo_stop)(struct net_device *dev); 1108 * This function is called when a network device transitions to the down 1109 * state. 1110 * 1111 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1112 * struct net_device *dev); 1113 * Called when a packet needs to be transmitted. 1114 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1115 * the queue before that can happen; it's for obsolete devices and weird 1116 * corner cases, but the stack really does a non-trivial amount 1117 * of useless work if you return NETDEV_TX_BUSY. 1118 * Required; cannot be NULL. 1119 * 1120 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1121 * struct net_device *dev 1122 * netdev_features_t features); 1123 * Called by core transmit path to determine if device is capable of 1124 * performing offload operations on a given packet. This is to give 1125 * the device an opportunity to implement any restrictions that cannot 1126 * be otherwise expressed by feature flags. The check is called with 1127 * the set of features that the stack has calculated and it returns 1128 * those the driver believes to be appropriate. 1129 * 1130 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1131 * struct net_device *sb_dev); 1132 * Called to decide which queue to use when device supports multiple 1133 * transmit queues. 1134 * 1135 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1136 * This function is called to allow device receiver to make 1137 * changes to configuration when multicast or promiscuous is enabled. 1138 * 1139 * void (*ndo_set_rx_mode)(struct net_device *dev); 1140 * This function is called device changes address list filtering. 1141 * If driver handles unicast address filtering, it should set 1142 * IFF_UNICAST_FLT in its priv_flags. 1143 * Cannot sleep, called with netif_addr_lock_bh held. 1144 * Deprecated in favor of ndo_set_rx_mode_async. 1145 * 1146 * int (*ndo_set_rx_mode_async)(struct net_device *dev, 1147 * struct netdev_hw_addr_list *uc, 1148 * struct netdev_hw_addr_list *mc); 1149 * Async version of ndo_set_rx_mode which runs in process context 1150 * under the netdev instance lock for "ops locked" drivers, or 1151 * rtnl_lock for all other drivers. The uc/mc parameters 1152 * are snapshots of the address lists - iterate with 1153 * netdev_hw_addr_list_for_each(ha, uc). Return 0 on success or a 1154 * negative errno to request a retry via the core backoff. 1155 * 1156 * void (*ndo_work)(struct net_device *dev, unsigned long events); 1157 * Run deferred work scheduled with netdev_work_sched(@events). 1158 * Runs in process context under the netdev instance lock for "ops 1159 * locked" drivers, or rtnl_lock for all other drivers. 1160 * 1161 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1162 * This function is called when the Media Access Control address 1163 * needs to be changed. If this interface is not defined, the 1164 * MAC address can not be changed. 1165 * 1166 * int (*ndo_validate_addr)(struct net_device *dev); 1167 * Test if Media Access Control address is valid for the device. 1168 * 1169 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1170 * Used by the bonding driver for its device specific ioctls: 1171 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1172 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1173 * 1174 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1175 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG and 1176 * SIOCSMIIREG. 1177 * 1178 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1179 * Used to set network devices bus interface parameters. This interface 1180 * is retained for legacy reasons; new devices should use the bus 1181 * interface (PCI) for low level management. 1182 * 1183 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1184 * Called when a user wants to change the Maximum Transfer Unit 1185 * of a device. 1186 * 1187 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1188 * Callback used when the transmitter has not made any progress 1189 * for dev->watchdog ticks. 1190 * 1191 * void (*ndo_get_stats64)(struct net_device *dev, 1192 * struct rtnl_link_stats64 *storage); 1193 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1194 * Called when a user wants to get the network device usage 1195 * statistics. Drivers must do one of the following: 1196 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1197 * rtnl_link_stats64 structure passed by the caller. 1198 * 2. Define @ndo_get_stats to update a net_device_stats structure 1199 * (which should normally be dev->stats) and return a pointer to 1200 * it. The structure may be changed asynchronously only if each 1201 * field is written atomically. 1202 * 3. Update dev->stats asynchronously and atomically, and define 1203 * neither operation. 1204 * 1205 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1206 * Return true if this device supports offload stats of this attr_id. 1207 * 1208 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1209 * void *attr_data) 1210 * Get statistics for offload operations by attr_id. Write it into the 1211 * attr_data pointer. 1212 * 1213 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1214 * If device supports VLAN filtering this function is called when a 1215 * VLAN id is registered. 1216 * 1217 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1218 * If device supports VLAN filtering this function is called when a 1219 * VLAN id is unregistered. 1220 * 1221 * void (*ndo_poll_controller)(struct net_device *dev); 1222 * 1223 * SR-IOV management functions. 1224 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1225 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1226 * u8 qos, __be16 proto); 1227 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1228 * int max_tx_rate); 1229 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1230 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1231 * int (*ndo_get_vf_config)(struct net_device *dev, 1232 * int vf, struct ifla_vf_info *ivf); 1233 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1234 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1235 * struct nlattr *port[]); 1236 * 1237 * Enable or disable the VF ability to query its RSS Redirection Table and 1238 * Hash Key. This is needed since on some devices VF share this information 1239 * with PF and querying it may introduce a theoretical security risk. 1240 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1241 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1242 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1243 * void *type_data); 1244 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1245 * This is always called from the stack with the rtnl lock held and netif 1246 * tx queues stopped. This allows the netdevice to perform queue 1247 * management safely. 1248 * 1249 * NB: Returning -EOPNOTSUPP for whatever commands means "this qdisc 1250 * is not offloaded (anymore, offloading may have silently stopped)", 1251 * and the offloading flag is cleared. Notably, this is also true for 1252 * dump queries (e.g. TC_*_STATS commands). If the underlying device does 1253 * not report any statistics but is still offloading, return 0 instead. 1254 * 1255 * Fiber Channel over Ethernet (FCoE) offload functions. 1256 * int (*ndo_fcoe_enable)(struct net_device *dev); 1257 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1258 * so the underlying device can perform whatever needed configuration or 1259 * initialization to support acceleration of FCoE traffic. 1260 * 1261 * int (*ndo_fcoe_disable)(struct net_device *dev); 1262 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1263 * so the underlying device can perform whatever needed clean-ups to 1264 * stop supporting acceleration of FCoE traffic. 1265 * 1266 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1267 * struct scatterlist *sgl, unsigned int sgc); 1268 * Called when the FCoE Initiator wants to initialize an I/O that 1269 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1270 * perform necessary setup and returns 1 to indicate the device is set up 1271 * successfully to perform DDP on this I/O, otherwise this returns 0. 1272 * 1273 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1274 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1275 * indicated by the FC exchange id 'xid', so the underlying device can 1276 * clean up and reuse resources for later DDP requests. 1277 * 1278 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1279 * struct scatterlist *sgl, unsigned int sgc); 1280 * Called when the FCoE Target wants to initialize an I/O that 1281 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1282 * perform necessary setup and returns 1 to indicate the device is set up 1283 * successfully to perform DDP on this I/O, otherwise this returns 0. 1284 * 1285 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1286 * struct netdev_fcoe_hbainfo *hbainfo); 1287 * Called when the FCoE Protocol stack wants information on the underlying 1288 * device. This information is utilized by the FCoE protocol stack to 1289 * register attributes with Fiber Channel management service as per the 1290 * FC-GS Fabric Device Management Information(FDMI) specification. 1291 * 1292 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1293 * Called when the underlying device wants to override default World Wide 1294 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1295 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1296 * protocol stack to use. 1297 * 1298 * RFS acceleration. 1299 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1300 * u16 rxq_index, u32 flow_id); 1301 * Set hardware filter for RFS. rxq_index is the target queue index; 1302 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1303 * Return the filter ID on success, or a negative error code. 1304 * 1305 * Slave management functions (for bridge, bonding, etc). 1306 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1307 * Called to make another netdev an underling. 1308 * 1309 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1310 * Called to release previously enslaved netdev. 1311 * 1312 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1313 * struct sk_buff *skb, 1314 * bool all_slaves); 1315 * Get the xmit slave of master device. If all_slaves is true, function 1316 * assume all the slaves can transmit. 1317 * 1318 * Feature/offload setting functions. 1319 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1320 * netdev_features_t features); 1321 * Adjusts the requested feature flags according to device-specific 1322 * constraints, and returns the resulting flags. Must not modify 1323 * the device state. 1324 * 1325 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1326 * Called to update device configuration to new features. Passed 1327 * feature set might be less than what was returned by ndo_fix_features()). 1328 * Must return >0 or -errno if it changed dev->features itself. 1329 * 1330 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1331 * struct net_device *dev, 1332 * const unsigned char *addr, u16 vid, u16 flags, 1333 * bool *notified, struct netlink_ext_ack *extack); 1334 * Adds an FDB entry to dev for addr. 1335 * Callee shall set *notified to true if it sent any appropriate 1336 * notification(s). Otherwise core will send a generic one. 1337 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1338 * struct net_device *dev, 1339 * const unsigned char *addr, u16 vid 1340 * bool *notified, struct netlink_ext_ack *extack); 1341 * Deletes the FDB entry from dev corresponding to addr. 1342 * Callee shall set *notified to true if it sent any appropriate 1343 * notification(s). Otherwise core will send a generic one. 1344 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev, 1345 * struct netlink_ext_ack *extack); 1346 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1347 * struct net_device *dev, struct net_device *filter_dev, 1348 * int *idx) 1349 * Used to add FDB entries to dump requests. Implementers should add 1350 * entries to skb and update idx with the number of entries. 1351 * 1352 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[], 1353 * u16 nlmsg_flags, struct netlink_ext_ack *extack); 1354 * Adds an MDB entry to dev. 1355 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[], 1356 * struct netlink_ext_ack *extack); 1357 * Deletes the MDB entry from dev. 1358 * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[], 1359 * struct netlink_ext_ack *extack); 1360 * Bulk deletes MDB entries from dev. 1361 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb, 1362 * struct netlink_callback *cb); 1363 * Dumps MDB entries from dev. The first argument (marker) in the netlink 1364 * callback is used by core rtnetlink code. 1365 * 1366 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1367 * u16 flags, struct netlink_ext_ack *extack) 1368 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1369 * struct net_device *dev, u32 filter_mask, 1370 * int nlflags) 1371 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1372 * u16 flags); 1373 * 1374 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1375 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1376 * which do not represent real hardware may define this to allow their 1377 * userspace components to manage their virtual carrier state. Devices 1378 * that determine carrier state from physical hardware properties (eg 1379 * network cables) or protocol-dependent mechanisms (eg 1380 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1381 * 1382 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1383 * struct netdev_phys_item_id *ppid); 1384 * Called to get ID of physical port of this device. If driver does 1385 * not implement this, it is assumed that the hw is not able to have 1386 * multiple net devices on single physical port. 1387 * 1388 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1389 * struct netdev_phys_item_id *ppid) 1390 * Called to get the parent ID of the physical port of this device. 1391 * 1392 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1393 * struct net_device *dev) 1394 * Called by upper layer devices to accelerate switching or other 1395 * station functionality into hardware. 'pdev is the lowerdev 1396 * to use for the offload and 'dev' is the net device that will 1397 * back the offload. Returns a pointer to the private structure 1398 * the upper layer will maintain. 1399 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1400 * Called by upper layer device to delete the station created 1401 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1402 * the station and priv is the structure returned by the add 1403 * operation. 1404 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1405 * int queue_index, u32 maxrate); 1406 * Called when a user wants to set a max-rate limitation of specific 1407 * TX queue. 1408 * int (*ndo_get_iflink)(const struct net_device *dev); 1409 * Called to get the iflink value of this device. 1410 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1411 * This function is used to get egress tunnel information for given skb. 1412 * This is useful for retrieving outer tunnel header parameters while 1413 * sampling packet. 1414 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1415 * This function is used to specify the headroom that the skb must 1416 * consider when allocation skb during packet reception. Setting 1417 * appropriate rx headroom value allows avoiding skb head copy on 1418 * forward. Setting a negative value resets the rx headroom to the 1419 * default value. 1420 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1421 * This function is used to set or query state related to XDP on the 1422 * netdevice and manage BPF offload. See definition of 1423 * enum bpf_netdev_command for details. 1424 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1425 * u32 flags); 1426 * This function is used to submit @n XDP packets for transmit on a 1427 * netdevice. Returns number of frames successfully transmitted, frames 1428 * that got dropped are freed/returned via xdp_return_frame(). 1429 * Returns negative number, means general error invoking ndo, meaning 1430 * no frames were xmit'ed and core-caller will free all frames. 1431 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1432 * struct xdp_buff *xdp); 1433 * Get the xmit slave of master device based on the xdp_buff. 1434 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1435 * This function is used to wake up the softirq, ksoftirqd or kthread 1436 * responsible for sending and/or receiving packets on a specific 1437 * queue id bound to an AF_XDP socket. The flags field specifies if 1438 * only RX, only Tx, or both should be woken up using the flags 1439 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1440 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm_kern *p, 1441 * int cmd); 1442 * Add, change, delete or get information on an IPv4 tunnel. 1443 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1444 * If a device is paired with a peer device, return the peer instance. 1445 * The caller must be under RCU read context. 1446 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1447 * Get the forwarding path to reach the real device from the HW destination address 1448 * ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1449 * const struct skb_shared_hwtstamps *hwtstamps, 1450 * bool cycles); 1451 * Get hardware timestamp based on normal/adjustable time or free running 1452 * cycle counter. This function is required if physical clock supports a 1453 * free running cycle counter. 1454 * 1455 * int (*ndo_hwtstamp_get)(struct net_device *dev, 1456 * struct kernel_hwtstamp_config *kernel_config); 1457 * Get the currently configured hardware timestamping parameters for the 1458 * NIC device. 1459 * 1460 * int (*ndo_hwtstamp_set)(struct net_device *dev, 1461 * struct kernel_hwtstamp_config *kernel_config, 1462 * struct netlink_ext_ack *extack); 1463 * Change the hardware timestamping parameters for NIC device. 1464 */ 1465 struct net_device_ops { 1466 int (*ndo_init)(struct net_device *dev); 1467 void (*ndo_uninit)(struct net_device *dev); 1468 int (*ndo_open)(struct net_device *dev); 1469 int (*ndo_stop)(struct net_device *dev); 1470 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1471 struct net_device *dev); 1472 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1473 struct net_device *dev, 1474 netdev_features_t features); 1475 u16 (*ndo_select_queue)(struct net_device *dev, 1476 struct sk_buff *skb, 1477 struct net_device *sb_dev); 1478 void (*ndo_change_rx_flags)(struct net_device *dev, 1479 int flags); 1480 void (*ndo_set_rx_mode)(struct net_device *dev); 1481 int (*ndo_set_rx_mode_async)( 1482 struct net_device *dev, 1483 struct netdev_hw_addr_list *uc, 1484 struct netdev_hw_addr_list *mc); 1485 void (*ndo_work)(struct net_device *dev, 1486 unsigned long events); 1487 int (*ndo_set_mac_address)(struct net_device *dev, 1488 void *addr); 1489 int (*ndo_validate_addr)(struct net_device *dev); 1490 int (*ndo_eth_ioctl)(struct net_device *dev, 1491 struct ifreq *ifr, int cmd); 1492 int (*ndo_siocbond)(struct net_device *dev, 1493 struct ifreq *ifr, int cmd); 1494 int (*ndo_siocwandev)(struct net_device *dev, 1495 struct if_settings *ifs); 1496 int (*ndo_siocdevprivate)(struct net_device *dev, 1497 struct ifreq *ifr, 1498 void __user *data, int cmd); 1499 int (*ndo_set_config)(struct net_device *dev, 1500 struct ifmap *map); 1501 int (*ndo_change_mtu)(struct net_device *dev, 1502 int new_mtu); 1503 int (*ndo_neigh_setup)(struct net_device *dev, 1504 struct neigh_parms *); 1505 void (*ndo_tx_timeout) (struct net_device *dev, 1506 unsigned int txqueue); 1507 1508 void (*ndo_get_stats64)(struct net_device *dev, 1509 struct rtnl_link_stats64 *storage); 1510 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1511 int (*ndo_get_offload_stats)(int attr_id, 1512 const struct net_device *dev, 1513 void *attr_data); 1514 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1515 1516 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1517 __be16 proto, u16 vid); 1518 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1519 __be16 proto, u16 vid); 1520 #ifdef CONFIG_NET_POLL_CONTROLLER 1521 void (*ndo_poll_controller)(struct net_device *dev); 1522 int (*ndo_netpoll_setup)(struct net_device *dev); 1523 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1524 #endif 1525 int (*ndo_set_vf_mac)(struct net_device *dev, 1526 int queue, u8 *mac); 1527 int (*ndo_set_vf_vlan)(struct net_device *dev, 1528 int queue, u16 vlan, 1529 u8 qos, __be16 proto); 1530 int (*ndo_set_vf_rate)(struct net_device *dev, 1531 int vf, int min_tx_rate, 1532 int max_tx_rate); 1533 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1534 int vf, bool setting); 1535 int (*ndo_set_vf_trust)(struct net_device *dev, 1536 int vf, bool setting); 1537 int (*ndo_get_vf_config)(struct net_device *dev, 1538 int vf, 1539 struct ifla_vf_info *ivf); 1540 int (*ndo_set_vf_link_state)(struct net_device *dev, 1541 int vf, int link_state); 1542 int (*ndo_get_vf_stats)(struct net_device *dev, 1543 int vf, 1544 struct ifla_vf_stats 1545 *vf_stats); 1546 int (*ndo_set_vf_port)(struct net_device *dev, 1547 int vf, 1548 struct nlattr *port[]); 1549 int (*ndo_get_vf_port)(struct net_device *dev, 1550 int vf, struct sk_buff *skb); 1551 int (*ndo_get_vf_guid)(struct net_device *dev, 1552 int vf, 1553 struct ifla_vf_guid *node_guid, 1554 struct ifla_vf_guid *port_guid); 1555 int (*ndo_set_vf_guid)(struct net_device *dev, 1556 int vf, u64 guid, 1557 int guid_type); 1558 int (*ndo_set_vf_rss_query_en)( 1559 struct net_device *dev, 1560 int vf, bool setting); 1561 int (*ndo_setup_tc)(struct net_device *dev, 1562 enum tc_setup_type type, 1563 void *type_data); 1564 #if IS_ENABLED(CONFIG_FCOE) 1565 int (*ndo_fcoe_enable)(struct net_device *dev); 1566 int (*ndo_fcoe_disable)(struct net_device *dev); 1567 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1568 u16 xid, 1569 struct scatterlist *sgl, 1570 unsigned int sgc); 1571 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1572 u16 xid); 1573 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1574 u16 xid, 1575 struct scatterlist *sgl, 1576 unsigned int sgc); 1577 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1578 struct netdev_fcoe_hbainfo *hbainfo); 1579 #endif 1580 1581 #if IS_ENABLED(CONFIG_LIBFCOE) 1582 #define NETDEV_FCOE_WWNN 0 1583 #define NETDEV_FCOE_WWPN 1 1584 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1585 u64 *wwn, int type); 1586 #endif 1587 1588 #ifdef CONFIG_RFS_ACCEL 1589 int (*ndo_rx_flow_steer)(struct net_device *dev, 1590 const struct sk_buff *skb, 1591 u16 rxq_index, 1592 u32 flow_id); 1593 #endif 1594 int (*ndo_add_slave)(struct net_device *dev, 1595 struct net_device *slave_dev, 1596 struct netlink_ext_ack *extack); 1597 int (*ndo_del_slave)(struct net_device *dev, 1598 struct net_device *slave_dev); 1599 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1600 struct sk_buff *skb, 1601 bool all_slaves); 1602 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1603 struct sock *sk); 1604 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1605 netdev_features_t features); 1606 int (*ndo_set_features)(struct net_device *dev, 1607 netdev_features_t features); 1608 int (*ndo_neigh_construct)(struct net_device *dev, 1609 struct neighbour *n); 1610 void (*ndo_neigh_destroy)(struct net_device *dev, 1611 struct neighbour *n); 1612 1613 int (*ndo_fdb_add)(struct ndmsg *ndm, 1614 struct nlattr *tb[], 1615 struct net_device *dev, 1616 const unsigned char *addr, 1617 u16 vid, 1618 u16 flags, 1619 bool *notified, 1620 struct netlink_ext_ack *extack); 1621 int (*ndo_fdb_del)(struct ndmsg *ndm, 1622 struct nlattr *tb[], 1623 struct net_device *dev, 1624 const unsigned char *addr, 1625 u16 vid, 1626 bool *notified, 1627 struct netlink_ext_ack *extack); 1628 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, 1629 struct net_device *dev, 1630 struct netlink_ext_ack *extack); 1631 int (*ndo_fdb_dump)(struct sk_buff *skb, 1632 struct netlink_callback *cb, 1633 struct net_device *dev, 1634 struct net_device *filter_dev, 1635 int *idx); 1636 int (*ndo_fdb_get)(struct sk_buff *skb, 1637 struct nlattr *tb[], 1638 struct net_device *dev, 1639 const unsigned char *addr, 1640 u16 vid, u32 portid, u32 seq, 1641 struct netlink_ext_ack *extack); 1642 int (*ndo_mdb_add)(struct net_device *dev, 1643 struct nlattr *tb[], 1644 u16 nlmsg_flags, 1645 struct netlink_ext_ack *extack); 1646 int (*ndo_mdb_del)(struct net_device *dev, 1647 struct nlattr *tb[], 1648 struct netlink_ext_ack *extack); 1649 int (*ndo_mdb_del_bulk)(struct net_device *dev, 1650 struct nlattr *tb[], 1651 struct netlink_ext_ack *extack); 1652 int (*ndo_mdb_dump)(struct net_device *dev, 1653 struct sk_buff *skb, 1654 struct netlink_callback *cb); 1655 int (*ndo_mdb_get)(struct net_device *dev, 1656 struct nlattr *tb[], u32 portid, 1657 u32 seq, 1658 struct netlink_ext_ack *extack); 1659 int (*ndo_bridge_setlink)(struct net_device *dev, 1660 struct nlmsghdr *nlh, 1661 u16 flags, 1662 struct netlink_ext_ack *extack); 1663 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1664 u32 pid, u32 seq, 1665 struct net_device *dev, 1666 u32 filter_mask, 1667 int nlflags); 1668 int (*ndo_bridge_dellink)(struct net_device *dev, 1669 struct nlmsghdr *nlh, 1670 u16 flags); 1671 int (*ndo_change_carrier)(struct net_device *dev, 1672 bool new_carrier); 1673 int (*ndo_get_phys_port_id)(struct net_device *dev, 1674 struct netdev_phys_item_id *ppid); 1675 int (*ndo_get_port_parent_id)(struct net_device *dev, 1676 struct netdev_phys_item_id *ppid); 1677 int (*ndo_get_phys_port_name)(struct net_device *dev, 1678 char *name, size_t len); 1679 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1680 struct net_device *dev); 1681 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1682 void *priv); 1683 1684 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1685 int queue_index, 1686 u32 maxrate); 1687 int (*ndo_get_iflink)(const struct net_device *dev); 1688 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1689 struct sk_buff *skb); 1690 void (*ndo_set_rx_headroom)(struct net_device *dev, 1691 int needed_headroom); 1692 int (*ndo_bpf)(struct net_device *dev, 1693 struct netdev_bpf *bpf); 1694 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1695 struct xdp_frame **xdp, 1696 u32 flags); 1697 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1698 struct xdp_buff *xdp); 1699 int (*ndo_xsk_wakeup)(struct net_device *dev, 1700 u32 queue_id, u32 flags); 1701 int (*ndo_tunnel_ctl)(struct net_device *dev, 1702 struct ip_tunnel_parm_kern *p, 1703 int cmd); 1704 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1705 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1706 struct net_device_path *path); 1707 ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1708 const struct skb_shared_hwtstamps *hwtstamps, 1709 bool cycles); 1710 int (*ndo_hwtstamp_get)(struct net_device *dev, 1711 struct kernel_hwtstamp_config *kernel_config); 1712 int (*ndo_hwtstamp_set)(struct net_device *dev, 1713 struct kernel_hwtstamp_config *kernel_config, 1714 struct netlink_ext_ack *extack); 1715 1716 #if IS_ENABLED(CONFIG_NET_SHAPER) 1717 /** 1718 * @net_shaper_ops: Device shaping offload operations 1719 * see include/net/net_shapers.h 1720 */ 1721 const struct net_shaper_ops *net_shaper_ops; 1722 #endif 1723 }; 1724 1725 /** 1726 * enum netdev_priv_flags - &struct net_device priv_flags 1727 * 1728 * These are the &struct net_device, they are only set internally 1729 * by drivers and used in the kernel. These flags are invisible to 1730 * userspace; this means that the order of these flags can change 1731 * during any kernel release. 1732 * 1733 * You should add bitfield booleans after either net_device::priv_flags 1734 * (hotpath) or ::threaded (slowpath) instead of extending these flags. 1735 * 1736 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1737 * @IFF_EBRIDGE: Ethernet bridging device 1738 * @IFF_BONDING: bonding master or slave 1739 * @IFF_ISATAP: ISATAP interface (RFC4214) 1740 * @IFF_WAN_HDLC: WAN HDLC device 1741 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1742 * release skb->dst 1743 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1744 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1745 * @IFF_MACVLAN_PORT: device used as macvlan port 1746 * @IFF_BRIDGE_PORT: device used as bridge port 1747 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1748 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1749 * @IFF_UNICAST_FLT: Supports unicast filtering 1750 * @IFF_TEAM_PORT: device used as team port 1751 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1752 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1753 * change when it's running 1754 * @IFF_MACVLAN: Macvlan device 1755 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1756 * underlying stacked devices 1757 * @IFF_L3MDEV_MASTER: device is an L3 master device 1758 * @IFF_NO_QUEUE: device can run without qdisc attached 1759 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1760 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1761 * @IFF_TEAM: device is a team device 1762 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1763 * entity (i.e. the master device for bridged veth) 1764 * @IFF_MACSEC: device is a MACsec device 1765 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1766 * @IFF_FAILOVER: device is a failover master device 1767 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1768 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1769 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf 1770 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1771 * skb_headlen(skb) == 0 (data starts from frag0) 1772 */ 1773 enum netdev_priv_flags { 1774 IFF_802_1Q_VLAN = 1<<0, 1775 IFF_EBRIDGE = 1<<1, 1776 IFF_BONDING = 1<<2, 1777 IFF_ISATAP = 1<<3, 1778 IFF_WAN_HDLC = 1<<4, 1779 IFF_XMIT_DST_RELEASE = 1<<5, 1780 IFF_DONT_BRIDGE = 1<<6, 1781 IFF_DISABLE_NETPOLL = 1<<7, 1782 IFF_MACVLAN_PORT = 1<<8, 1783 IFF_BRIDGE_PORT = 1<<9, 1784 IFF_OVS_DATAPATH = 1<<10, 1785 IFF_TX_SKB_SHARING = 1<<11, 1786 IFF_UNICAST_FLT = 1<<12, 1787 IFF_TEAM_PORT = 1<<13, 1788 IFF_SUPP_NOFCS = 1<<14, 1789 IFF_LIVE_ADDR_CHANGE = 1<<15, 1790 IFF_MACVLAN = 1<<16, 1791 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1792 IFF_L3MDEV_MASTER = 1<<18, 1793 IFF_NO_QUEUE = 1<<19, 1794 IFF_OPENVSWITCH = 1<<20, 1795 IFF_L3MDEV_SLAVE = 1<<21, 1796 IFF_TEAM = 1<<22, 1797 IFF_PHONY_HEADROOM = 1<<24, 1798 IFF_MACSEC = 1<<25, 1799 IFF_NO_RX_HANDLER = 1<<26, 1800 IFF_FAILOVER = 1<<27, 1801 IFF_FAILOVER_SLAVE = 1<<28, 1802 IFF_L3MDEV_RX_HANDLER = 1<<29, 1803 IFF_NO_ADDRCONF = BIT_ULL(30), 1804 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31), 1805 }; 1806 1807 /* Specifies the type of the struct net_device::ml_priv pointer */ 1808 enum netdev_ml_priv_type { 1809 ML_PRIV_NONE, 1810 ML_PRIV_CAN, 1811 }; 1812 1813 enum netdev_stat_type { 1814 NETDEV_PCPU_STAT_NONE, 1815 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */ 1816 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */ 1817 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */ 1818 }; 1819 1820 enum netmem_tx_mode { 1821 NETMEM_TX_NONE, /* no netmem TX support */ 1822 NETMEM_TX_DMA, /* DMA-capable netmem TX (real HW) */ 1823 NETMEM_TX_NO_DMA, /* no DMA, e.g. passthrough for virtual devs */ 1824 }; 1825 1826 enum netdev_reg_state { 1827 NETREG_UNINITIALIZED = 0, 1828 NETREG_REGISTERED, /* completed register_netdevice */ 1829 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1830 NETREG_UNREGISTERED, /* completed unregister todo */ 1831 NETREG_RELEASED, /* called free_netdev */ 1832 NETREG_DUMMY, /* dummy device for NAPI poll */ 1833 }; 1834 1835 /** 1836 * struct net_device - The DEVICE structure. 1837 * 1838 * Actually, this whole structure is a big mistake. It mixes I/O 1839 * data with strictly "high-level" data, and it has to know about 1840 * almost every data structure used in the INET module. 1841 * 1842 * @priv_flags: flags invisible to userspace defined as bits, see 1843 * enum netdev_priv_flags for the definitions 1844 * @lltx: device supports lockless Tx. Deprecated for real HW 1845 * drivers. Mainly used by logical interfaces, such as 1846 * bonding and tunnels 1847 * @netmem_tx: device netmem TX mode 1848 * @pacing_offload: enable EDT pacing offload. 1849 * 1850 * @name: This is the first field of the "visible" part of this structure 1851 * (i.e. as seen by users in the "Space.c" file). It is the name 1852 * of the interface. 1853 * 1854 * @name_node: Name hashlist node 1855 * @ifalias: SNMP alias 1856 * @mem_end: Shared memory end 1857 * @mem_start: Shared memory start 1858 * @base_addr: Device I/O address 1859 * @irq: Device IRQ number 1860 * 1861 * @state: Generic network queuing layer state, see netdev_state_t 1862 * @dev_list: The global list of network devices 1863 * @napi_list: List entry used for polling NAPI devices 1864 * @unreg_list: List entry when we are unregistering the 1865 * device; see the function unregister_netdev 1866 * @unreg_list_net:List entry when we are unregistering the cross-netns 1867 * device; see the function unregister_netdevice_queue_net() 1868 * @close_list: List entry used when we are closing the device 1869 * @ptype_all: Device-specific packet handlers for all protocols 1870 * @ptype_specific: Device-specific, protocol-specific packet handlers 1871 * 1872 * @adj_list: Directly linked devices, like slaves for bonding 1873 * @features: Currently active device features 1874 * @hw_features: User-changeable features 1875 * 1876 * @wanted_features: User-requested features 1877 * @vlan_features: Mask of features inheritable by VLAN devices 1878 * 1879 * @hw_enc_features: Mask of features inherited by encapsulating devices 1880 * This field indicates what encapsulation 1881 * offloads the hardware is capable of doing, 1882 * and drivers will need to set them appropriately. 1883 * 1884 * @mpls_features: Mask of features inheritable by MPLS 1885 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1886 * @mangleid_features: Mask of features requiring MANGLEID, will be 1887 * disabled together with the latter. 1888 * 1889 * @ifindex: interface index 1890 * @group: The group the device belongs to 1891 * 1892 * @stats: Statistics struct, which was left as a legacy, use 1893 * rtnl_link_stats64 instead 1894 * 1895 * @core_stats: core networking counters, 1896 * do not use this in drivers 1897 * @carrier_up_count: Number of times the carrier has been up 1898 * @carrier_down_count: Number of times the carrier has been down 1899 * 1900 * @wireless_handlers: List of functions to handle Wireless Extensions, 1901 * instead of ioctl, 1902 * see <net/iw_handler.h> for details. 1903 * 1904 * @netdev_ops: Includes several pointers to callbacks, 1905 * if one wants to override the ndo_*() functions 1906 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks. 1907 * @xsk_tx_metadata_ops: Includes pointers to AF_XDP TX metadata callbacks. 1908 * @ethtool_ops: Management operations 1909 * @l3mdev_ops: Layer 3 master device operations 1910 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1911 * discovery handling. Necessary for e.g. 6LoWPAN. 1912 * @xfrmdev_ops: Transformation offload operations 1913 * @tlsdev_ops: Transport Layer Security offload operations 1914 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1915 * of Layer 2 headers. 1916 * 1917 * @flags: Interface flags (a la BSD) 1918 * @xdp_features: XDP capability supported by the device 1919 * @gflags: Global flags ( kept as legacy ) 1920 * @priv_len: Size of the ->priv flexible array 1921 * @priv: Flexible array containing private data 1922 * @operstate: RFC2863 operstate 1923 * @link_mode: Mapping policy to operstate 1924 * @if_port: Selectable AUI, TP, ... 1925 * @dma: DMA channel 1926 * @mtu: Interface MTU value 1927 * @min_mtu: Interface Minimum MTU value 1928 * @max_mtu: Interface Maximum MTU value 1929 * @type: Interface hardware type 1930 * @hard_header_len: Maximum hardware header length. 1931 * @min_header_len: Minimum hardware header length 1932 * 1933 * @needed_headroom: Extra headroom the hardware may need, but not in all 1934 * cases can this be guaranteed 1935 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1936 * cases can this be guaranteed. Some cases also use 1937 * LL_MAX_HEADER instead to allocate the skb 1938 * 1939 * interface address info: 1940 * 1941 * @perm_addr: Permanent hw address 1942 * @addr_assign_type: Hw address assignment type 1943 * @addr_len: Hardware address length 1944 * @upper_level: Maximum depth level of upper devices. 1945 * @lower_level: Maximum depth level of lower devices. 1946 * @threaded: napi threaded state. 1947 * @neigh_priv_len: Used in neigh_alloc() 1948 * @dev_id: Used to differentiate devices that share 1949 * the same link layer address 1950 * @dev_port: Used to differentiate devices that share 1951 * the same function 1952 * @addr_list_lock: XXX: need comments on this one 1953 * @name_assign_type: network interface name assignment type 1954 * @uc_promisc: Counter that indicates promiscuous mode 1955 * has been enabled due to the need to listen to 1956 * additional unicast addresses in a device that 1957 * does not implement ndo_set_rx_mode() 1958 * @work_node: List entry for async netdev_work processing 1959 * @work_tracker: Refcount tracker for async netdev_work 1960 * @work_pending: Driver-defined pending netdev_work, passed to 1961 * ndo_work() (see netdev_work_sched()) 1962 * @work_core_pending: Core-defined pending netdev_work (NETDEV_WORK_*) 1963 * @rx_mode_addr_cache: Recycled snapshot entries for rx_mode work 1964 * @rx_mode_retry_timer: Timer that re-queues rx_mode work after failure 1965 * @rx_mode_retry_count: Number of consecutive retries already scheduled 1966 * @uc: unicast mac addresses 1967 * @mc: multicast mac addresses 1968 * @dev_addrs: list of device hw addresses 1969 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1970 * @promiscuity: Number of times the NIC is told to work in 1971 * promiscuous mode; if it becomes 0 the NIC will 1972 * exit promiscuous mode 1973 * @allmulti: Counter, enables or disables allmulticast mode 1974 * 1975 * @vlan_info: VLAN info 1976 * @dsa_ptr: dsa specific data 1977 * @tipc_ptr: TIPC specific data 1978 * @ip_ptr: IPv4 specific data 1979 * @ip6_ptr: IPv6 specific data 1980 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1981 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1982 * device struct 1983 * @mpls_ptr: mpls_dev struct pointer 1984 * @mctp_ptr: MCTP specific data 1985 * @psp_dev: PSP crypto device registered for this netdev 1986 * 1987 * @dev_addr: Hw address (before bcast, 1988 * because most packets are unicast) 1989 * 1990 * @_rx: Array of RX queues 1991 * @num_rx_queues: Number of RX queues 1992 * allocated at register_netdev() time 1993 * @real_num_rx_queues: Number of RX queues currently active in device 1994 * @xdp_prog: XDP sockets filter program pointer 1995 * 1996 * @rx_handler: handler for received packets 1997 * @rx_handler_data: XXX: need comments on this one 1998 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing 1999 * @ingress_queue: XXX: need comments on this one 2000 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 2001 * @broadcast: hw bcast address 2002 * 2003 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 2004 * indexed by RX queue number. Assigned by driver. 2005 * This must only be set if the ndo_rx_flow_steer 2006 * operation is defined 2007 * @index_hlist: Device index hash chain 2008 * 2009 * @_tx: Array of TX queues 2010 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 2011 * @real_num_tx_queues: Number of TX queues currently active in device 2012 * @qdisc: Root qdisc from userspace point of view 2013 * @tx_queue_len: Max frames per queue allowed 2014 * @tx_global_lock: XXX: need comments on this one 2015 * @xdp_bulkq: XDP device bulk queue 2016 * @xps_maps: all CPUs/RXQs maps for XPS device 2017 * 2018 * @xps_maps: XXX: need comments on this one 2019 * @tcx_egress: BPF & clsact qdisc specific data for egress processing 2020 * @nf_hooks_egress: netfilter hooks executed for egress packets 2021 * @qdisc_hash: qdisc hash table 2022 * @watchdog_timeo: Represents the timeout that is used by 2023 * the watchdog (see dev_watchdog()) 2024 * @watchdog_lock: protect watchdog_ref_held 2025 * @watchdog_ref_held: True if the watchdog device ref is taken. 2026 * @watchdog_timer: List of timers 2027 * 2028 * @proto_down_reason: reason a netdev interface is held down 2029 * @pcpu_refcnt: Number of references to this device 2030 * @dev_refcnt: Number of references to this device 2031 * @refcnt_tracker: Tracker directory for tracked references to this device 2032 * @todo_list: Delayed register/unregister 2033 * @link_watch_list: XXX: need comments on this one 2034 * 2035 * @reg_state: Register/unregister state machine 2036 * @dismantle: Device is going to be freed 2037 * @needs_free_netdev: Should unregister perform free_netdev? 2038 * @priv_destructor: Called from unregister 2039 * @npinfo: XXX: need comments on this one 2040 * @nd_net: Network namespace this network device is inside 2041 * protected by @lock 2042 * 2043 * @ml_priv: Mid-layer private 2044 * @ml_priv_type: Mid-layer private type 2045 * 2046 * @pcpu_stat_type: Type of device statistics which the core should 2047 * allocate/free: none, lstats, tstats, dstats. none 2048 * means the driver is handling statistics allocation/ 2049 * freeing internally. 2050 * @lstats: Loopback statistics: packets, bytes 2051 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes 2052 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes 2053 * 2054 * @garp_port: GARP 2055 * @mrp_port: MRP 2056 * 2057 * @dm_private: Drop monitor private 2058 * 2059 * @dev: Class/net/name entry 2060 * @sysfs_groups: Space for optional device, statistics and wireless 2061 * sysfs groups 2062 * 2063 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 2064 * @rtnl_link_ops: Rtnl_link_ops 2065 * @stat_ops: Optional ops for queue-aware statistics 2066 * @queue_mgmt_ops: Optional ops for queue management 2067 * 2068 * @gso_max_size: Maximum size of generic segmentation offload 2069 * @tso_max_size: Device (as in HW) limit on the max TSO request size 2070 * @gso_max_segs: Maximum number of segments that can be passed to the 2071 * NIC for GSO 2072 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count 2073 * @gso_ipv4_max_size: Maximum size of generic segmentation offload, 2074 * for IPv4. 2075 * 2076 * @dcbnl_ops: Data Center Bridging netlink ops 2077 * @num_tc: Number of traffic classes in the net device 2078 * @tc_to_txq: XXX: need comments on this one 2079 * @prio_tc_map: XXX: need comments on this one 2080 * 2081 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 2082 * 2083 * @priomap: XXX: need comments on this one 2084 * @link_topo: Physical link topology tracking attached PHYs 2085 * @phydev: Physical device may attach itself 2086 * for hardware timestamping 2087 * @sfp_bus: attached &struct sfp_bus structure. 2088 * 2089 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 2090 * 2091 * @proto_down: protocol port state information can be sent to the 2092 * switch driver and used to set the phys state of the 2093 * switch port. 2094 * 2095 * @irq_affinity_auto: driver wants the core to store and re-assign the IRQ 2096 * affinity. Set by netif_enable_irq_affinity(), then 2097 * the driver must create a persistent napi by 2098 * netif_napi_add_config() and finally bind the napi to 2099 * IRQ (via netif_napi_set_irq()). 2100 * 2101 * @rx_cpu_rmap_auto: driver wants the core to manage the ARFS rmap. 2102 * Set by calling netif_enable_cpu_rmap(). 2103 * 2104 * @see_all_hwtstamp_requests: device wants to see calls to 2105 * ndo_hwtstamp_set() for all timestamp requests 2106 * regardless of source, even if those aren't 2107 * HWTSTAMP_SOURCE_NETDEV 2108 * @change_proto_down: device supports setting carrier via IFLA_PROTO_DOWN 2109 * @netns_immutable: interface can't change network namespaces 2110 * @fcoe_mtu: device supports maximum FCoE MTU, 2158 bytes 2111 * 2112 * @net_notifier_list: List of per-net netdev notifier block 2113 * that follow this device when it is moved 2114 * to another network namespace. 2115 * 2116 * @macsec_ops: MACsec offloading ops 2117 * 2118 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 2119 * offload capabilities of the device 2120 * @udp_tunnel_nic: UDP tunnel offload state 2121 * @ethtool: ethtool related state 2122 * @xdp_state: stores info on attached XDP BPF programs 2123 * 2124 * @nested_level: Used as a parameter of spin_lock_nested() of 2125 * dev->addr_list_lock. 2126 * @unlink_list: As netif_addr_lock() can be called recursively, 2127 * keep a list of interfaces to be deleted. 2128 * @gro_max_size: Maximum size of aggregated packet in generic 2129 * receive offload (GRO) 2130 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic 2131 * receive offload (GRO), for IPv4. 2132 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP 2133 * zero copy driver 2134 * 2135 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 2136 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 2137 * @watchdog_dev_tracker: refcount tracker used by watchdog. 2138 * @dev_registered_tracker: tracker for reference held while 2139 * registered 2140 * @offload_xstats_l3: L3 HW stats for this netdevice. 2141 * 2142 * @devlink_port: Pointer to related devlink port structure. 2143 * Assigned by a driver before netdev registration using 2144 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static 2145 * during the time netdevice is registered. 2146 * 2147 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem, 2148 * where the clock is recovered. 2149 * 2150 * @max_pacing_offload_horizon: max EDT offload horizon in nsec. 2151 * @napi_config: An array of napi_config structures containing per-NAPI 2152 * settings. 2153 * @num_napi_configs: number of allocated NAPI config structs, 2154 * always >= max(num_rx_queues, num_tx_queues). 2155 * @gro_flush_timeout: timeout for GRO layer in NAPI 2156 * @napi_defer_hard_irqs: If not zero, provides a counter that would 2157 * allow to avoid NIC hard IRQ, on busy queues. 2158 * 2159 * @neighbours: List heads pointing to this device's neighbours' 2160 * dev_list, one per address-family. 2161 * @hwprov: Tracks which PTP performs hardware packet time stamping. 2162 * 2163 * FIXME: cleanup struct net_device such that network protocol info 2164 * moves out. 2165 */ 2166 2167 struct net_device { 2168 /* Cacheline organization can be found documented in 2169 * Documentation/networking/net_cachelines/net_device.rst. 2170 * Please update the document when adding new fields. 2171 */ 2172 2173 /* TX read-mostly hotpath */ 2174 __cacheline_group_begin(net_device_read_tx); 2175 struct_group(priv_flags_fast, 2176 unsigned long priv_flags:32; 2177 unsigned long lltx:1; 2178 unsigned long netmem_tx:2; 2179 unsigned long pacing_offload:1; 2180 ); 2181 const struct net_device_ops *netdev_ops; 2182 const struct header_ops *header_ops; 2183 struct netdev_queue *_tx; 2184 netdev_features_t gso_partial_features; 2185 unsigned int real_num_tx_queues; 2186 unsigned int gso_max_size; 2187 unsigned int gso_ipv4_max_size; 2188 u16 gso_max_segs; 2189 s16 num_tc; 2190 /* Note : dev->mtu is often read without holding a lock. 2191 * Writers usually hold RTNL. 2192 * It is recommended to use READ_ONCE() to annotate the reads, 2193 * and to use WRITE_ONCE() to annotate the writes. 2194 */ 2195 unsigned int mtu; 2196 unsigned short needed_headroom; 2197 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2198 #ifdef CONFIG_XPS 2199 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2200 #endif 2201 #ifdef CONFIG_NETFILTER_EGRESS 2202 struct nf_hook_entries __rcu *nf_hooks_egress; 2203 #endif 2204 #ifdef CONFIG_NET_XGRESS 2205 struct bpf_mprog_entry __rcu *tcx_egress; 2206 #endif 2207 __cacheline_group_end(net_device_read_tx); 2208 2209 /* TXRX read-mostly hotpath */ 2210 __cacheline_group_begin(net_device_read_txrx); 2211 union { 2212 struct pcpu_lstats __percpu *lstats; 2213 struct pcpu_sw_netstats __percpu *tstats; 2214 struct pcpu_dstats __percpu *dstats; 2215 }; 2216 unsigned long state; 2217 unsigned int flags; 2218 unsigned short hard_header_len; 2219 enum netdev_stat_type pcpu_stat_type:8; 2220 netdev_features_t features; 2221 struct inet6_dev __rcu *ip6_ptr; 2222 __cacheline_group_end(net_device_read_txrx); 2223 2224 /* RX read-mostly hotpath */ 2225 __cacheline_group_begin(net_device_read_rx); 2226 struct bpf_prog __rcu *xdp_prog; 2227 struct list_head ptype_specific; 2228 int ifindex; 2229 unsigned int real_num_rx_queues; 2230 struct netdev_rx_queue *_rx; 2231 unsigned int gro_max_size; 2232 unsigned int gro_ipv4_max_size; 2233 rx_handler_func_t __rcu *rx_handler; 2234 void __rcu *rx_handler_data; 2235 possible_net_t nd_net; 2236 #ifdef CONFIG_NETPOLL 2237 struct netpoll_info __rcu *npinfo; 2238 #endif 2239 #ifdef CONFIG_NET_XGRESS 2240 struct bpf_mprog_entry __rcu *tcx_ingress; 2241 #endif 2242 __cacheline_group_end(net_device_read_rx); 2243 2244 char name[IFNAMSIZ]; 2245 struct netdev_name_node *name_node; 2246 struct dev_ifalias __rcu *ifalias; 2247 /* 2248 * I/O specific fields 2249 * FIXME: Merge these and struct ifmap into one 2250 */ 2251 unsigned long mem_end; 2252 unsigned long mem_start; 2253 unsigned long base_addr; 2254 2255 /* 2256 * Some hardware also needs these fields (state,dev_list, 2257 * napi_list,unreg_list,close_list) but they are not 2258 * part of the usual set specified in Space.c. 2259 */ 2260 2261 2262 struct list_head dev_list; 2263 struct list_head napi_list; 2264 struct list_head unreg_list; 2265 #ifdef CONFIG_DEBUG_NET_SMALL_RTNL 2266 struct list_head unreg_list_net; 2267 #endif 2268 struct list_head close_list; 2269 struct list_head ptype_all; 2270 2271 struct { 2272 struct list_head upper; 2273 struct list_head lower; 2274 } adj_list; 2275 2276 /* Read-mostly cache-line for fast-path access */ 2277 xdp_features_t xdp_features; 2278 const struct xdp_metadata_ops *xdp_metadata_ops; 2279 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops; 2280 unsigned short gflags; 2281 2282 unsigned short needed_tailroom; 2283 2284 netdev_features_t hw_features; 2285 netdev_features_t wanted_features; 2286 netdev_features_t vlan_features; 2287 netdev_features_t hw_enc_features; 2288 netdev_features_t mpls_features; 2289 netdev_features_t mangleid_features; 2290 2291 unsigned int min_mtu; 2292 unsigned int max_mtu; 2293 unsigned short type; 2294 unsigned char min_header_len; 2295 unsigned char name_assign_type; 2296 2297 int group; 2298 2299 struct net_device_stats stats; /* not used by modern drivers */ 2300 2301 struct net_device_core_stats __percpu *core_stats; 2302 2303 /* Stats to monitor link on/off, flapping */ 2304 atomic_t carrier_up_count; 2305 atomic_t carrier_down_count; 2306 2307 #ifdef CONFIG_WIRELESS_EXT 2308 const struct iw_handler_def *wireless_handlers; 2309 #endif 2310 const struct ethtool_ops *ethtool_ops; 2311 #ifdef CONFIG_NET_L3_MASTER_DEV 2312 const struct l3mdev_ops *l3mdev_ops; 2313 #endif 2314 #if IS_ENABLED(CONFIG_IPV6) 2315 const struct ndisc_ops *ndisc_ops; 2316 #endif 2317 2318 #ifdef CONFIG_XFRM_OFFLOAD 2319 const struct xfrmdev_ops *xfrmdev_ops; 2320 #endif 2321 2322 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2323 const struct tlsdev_ops *tlsdev_ops; 2324 #endif 2325 2326 unsigned int operstate; 2327 unsigned char link_mode; 2328 2329 unsigned char if_port; 2330 unsigned char dma; 2331 2332 /* Interface address info. */ 2333 unsigned char perm_addr[MAX_ADDR_LEN]; 2334 unsigned char addr_assign_type; 2335 unsigned char addr_len; 2336 unsigned char upper_level; 2337 unsigned char lower_level; 2338 u8 threaded; 2339 2340 unsigned short neigh_priv_len; 2341 unsigned short dev_id; 2342 unsigned short dev_port; 2343 int irq; 2344 u32 priv_len; 2345 2346 spinlock_t addr_list_lock; 2347 2348 struct netdev_hw_addr_list uc; 2349 struct netdev_hw_addr_list mc; 2350 struct netdev_hw_addr_list dev_addrs; 2351 2352 #ifdef CONFIG_SYSFS 2353 struct kset *queues_kset; 2354 #endif 2355 #ifdef CONFIG_LOCKDEP 2356 struct list_head unlink_list; 2357 #endif 2358 unsigned int promiscuity; 2359 unsigned int allmulti; 2360 bool uc_promisc; 2361 struct list_head work_node; 2362 netdevice_tracker work_tracker; 2363 unsigned long work_pending; 2364 unsigned long work_core_pending; 2365 struct netdev_hw_addr_list rx_mode_addr_cache; 2366 struct timer_list rx_mode_retry_timer; 2367 unsigned int rx_mode_retry_count; 2368 #ifdef CONFIG_LOCKDEP 2369 unsigned char nested_level; 2370 #endif 2371 2372 2373 /* Protocol-specific pointers */ 2374 struct in_device __rcu *ip_ptr; 2375 /** @fib_nh_head: nexthops associated with this netdev */ 2376 struct hlist_head fib_nh_head; 2377 2378 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2379 struct vlan_info __rcu *vlan_info; 2380 #endif 2381 #if IS_ENABLED(CONFIG_NET_DSA) 2382 struct dsa_port *dsa_ptr; 2383 #endif 2384 #if IS_ENABLED(CONFIG_TIPC) 2385 struct tipc_bearer __rcu *tipc_ptr; 2386 #endif 2387 #if IS_ENABLED(CONFIG_CFG80211) 2388 struct wireless_dev *ieee80211_ptr; 2389 #endif 2390 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2391 struct wpan_dev *ieee802154_ptr; 2392 #endif 2393 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2394 struct mpls_dev __rcu *mpls_ptr; 2395 #endif 2396 #if IS_ENABLED(CONFIG_MCTP) 2397 struct mctp_dev __rcu *mctp_ptr; 2398 #endif 2399 #if IS_ENABLED(CONFIG_INET_PSP) 2400 struct psp_dev __rcu *psp_dev; 2401 #endif 2402 2403 /* 2404 * Cache lines mostly used on receive path (including eth_type_trans()) 2405 */ 2406 /* Interface address info used in eth_type_trans() */ 2407 const unsigned char *dev_addr; 2408 2409 unsigned int num_rx_queues; 2410 #define GRO_LEGACY_MAX_SIZE 65536u 2411 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2412 * and shinfo->gso_segs is a 16bit field. 2413 */ 2414 #define GRO_MAX_SIZE (8 * 65535u) 2415 unsigned int xdp_zc_max_segs; 2416 struct netdev_queue __rcu *ingress_queue; 2417 #ifdef CONFIG_NETFILTER_INGRESS 2418 struct nf_hook_entries __rcu *nf_hooks_ingress; 2419 #endif 2420 2421 unsigned char broadcast[MAX_ADDR_LEN]; 2422 #ifdef CONFIG_RFS_ACCEL 2423 struct cpu_rmap *rx_cpu_rmap; 2424 #endif 2425 struct hlist_node index_hlist; 2426 2427 /* 2428 * Cache lines mostly used on transmit path 2429 */ 2430 unsigned int num_tx_queues; 2431 struct Qdisc __rcu *qdisc; 2432 unsigned int tx_queue_len; 2433 spinlock_t tx_global_lock; 2434 2435 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2436 2437 #ifdef CONFIG_NET_SCHED 2438 DECLARE_HASHTABLE (qdisc_hash, 4); 2439 #endif 2440 /* These may be needed for future network-power-down code. */ 2441 struct timer_list watchdog_timer; 2442 int watchdog_timeo; 2443 spinlock_t watchdog_lock; 2444 bool watchdog_ref_held; 2445 2446 u32 proto_down_reason; 2447 2448 struct list_head todo_list; 2449 2450 #ifdef CONFIG_PCPU_DEV_REFCNT 2451 int __percpu *pcpu_refcnt; 2452 #else 2453 refcount_t dev_refcnt; 2454 #endif 2455 struct ref_tracker_dir refcnt_tracker; 2456 2457 struct list_head link_watch_list; 2458 2459 u8 reg_state; 2460 2461 bool dismantle; 2462 2463 /** @moving_ns: device is changing netns, protected by @lock */ 2464 bool moving_ns; 2465 /** @rtnl_link_initializing: Device being created, suppress events */ 2466 bool rtnl_link_initializing; 2467 2468 bool needs_free_netdev; 2469 void (*priv_destructor)(struct net_device *dev); 2470 2471 /* mid-layer private */ 2472 void *ml_priv; 2473 enum netdev_ml_priv_type ml_priv_type; 2474 2475 #if IS_ENABLED(CONFIG_GARP) 2476 struct garp_port __rcu *garp_port; 2477 #endif 2478 #if IS_ENABLED(CONFIG_MRP) 2479 struct mrp_port __rcu *mrp_port; 2480 #endif 2481 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2482 struct dm_hw_stat_delta __rcu *dm_private; 2483 #endif 2484 struct device dev; 2485 const struct attribute_group *sysfs_groups[5]; 2486 const struct attribute_group *sysfs_rx_queue_group; 2487 2488 const struct rtnl_link_ops *rtnl_link_ops; 2489 2490 const struct netdev_stat_ops *stat_ops; 2491 2492 const struct netdev_queue_mgmt_ops *queue_mgmt_ops; 2493 2494 /* for setting kernel sock attribute on TCP connection setup */ 2495 #define GSO_MAX_SEGS 65535u 2496 #define GSO_LEGACY_MAX_SIZE 65536u 2497 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2498 * and shinfo->gso_segs is a 16bit field. 2499 */ 2500 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2501 2502 #define TSO_LEGACY_MAX_SIZE 65536 2503 #define TSO_MAX_SIZE UINT_MAX 2504 unsigned int tso_max_size; 2505 #define TSO_MAX_SEGS U16_MAX 2506 u16 tso_max_segs; 2507 2508 #ifdef CONFIG_DCB 2509 const struct dcbnl_rtnl_ops *dcbnl_ops; 2510 #endif 2511 u8 prio_tc_map[TC_BITMASK + 1]; 2512 2513 #if IS_ENABLED(CONFIG_FCOE) 2514 unsigned int fcoe_ddp_xid; 2515 #endif 2516 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2517 struct netprio_map __rcu *priomap; 2518 #endif 2519 struct phy_link_topology *link_topo; 2520 struct phy_device *phydev; 2521 struct sfp_bus *sfp_bus; 2522 struct lock_class_key *qdisc_tx_busylock; 2523 bool proto_down; 2524 bool irq_affinity_auto; 2525 bool rx_cpu_rmap_auto; 2526 2527 /* priv_flags_slow, ungrouped to save space */ 2528 unsigned long see_all_hwtstamp_requests:1; 2529 unsigned long change_proto_down:1; 2530 unsigned long netns_immutable:1; 2531 unsigned long fcoe_mtu:1; 2532 2533 struct list_head net_notifier_list; 2534 2535 #if IS_ENABLED(CONFIG_MACSEC) 2536 /* MACsec management functions */ 2537 const struct macsec_ops *macsec_ops; 2538 #endif 2539 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2540 struct udp_tunnel_nic *udp_tunnel_nic; 2541 2542 /** @cfg: net_device queue-related configuration */ 2543 struct netdev_config *cfg; 2544 /** 2545 * @cfg_pending: same as @cfg but when device is being actively 2546 * reconfigured includes any changes to the configuration 2547 * requested by the user, but which may or may not be rejected. 2548 */ 2549 struct netdev_config *cfg_pending; 2550 struct ethtool_netdev_state *ethtool; 2551 2552 /* protected by rtnl_lock */ 2553 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2554 2555 #ifdef CONFIG_XDP_SOCKETS 2556 /** @xsk_pools: assigned AF_XDP pools, ops protected */ 2557 struct hlist_head xsk_pools; 2558 #endif 2559 2560 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2561 netdevice_tracker linkwatch_dev_tracker; 2562 netdevice_tracker watchdog_dev_tracker; 2563 netdevice_tracker dev_registered_tracker; 2564 struct rtnl_hw_stats64 *offload_xstats_l3; 2565 2566 struct devlink_port *devlink_port; 2567 2568 #if IS_ENABLED(CONFIG_DPLL) 2569 struct dpll_pin __rcu *dpll_pin; 2570 #endif 2571 #if IS_ENABLED(CONFIG_PAGE_POOL) 2572 /** @page_pools: page pools created for this netdevice */ 2573 struct hlist_head page_pools; 2574 #endif 2575 2576 /** @irq_moder: dim parameters used if IS_ENABLED(CONFIG_DIMLIB). */ 2577 struct dim_irq_moder *irq_moder; 2578 2579 u64 max_pacing_offload_horizon; 2580 struct napi_config *napi_config; 2581 u32 num_napi_configs; 2582 u32 napi_defer_hard_irqs; 2583 unsigned long gro_flush_timeout; 2584 2585 /** 2586 * @up: copy of @state's IFF_UP, but safe to read with just @lock. 2587 * May report false negatives while the device is being opened 2588 * or closed (@lock does not protect .ndo_open, or .ndo_close). 2589 */ 2590 bool up; 2591 2592 /** 2593 * @request_ops_lock: request the core to run all @netdev_ops and 2594 * @ethtool_ops under the @lock. 2595 */ 2596 bool request_ops_lock; 2597 2598 /** 2599 * @lock: netdev-scope lock, protects a small selection of fields. 2600 * Should always be taken using netdev_lock() / netdev_unlock() helpers. 2601 * Drivers are free to use it for other protection. 2602 * 2603 * For the drivers that implement shaper or queue API, the scope 2604 * of this lock is expanded to cover most ndo/queue/ethtool/sysfs 2605 * operations. Drivers may opt-in to this behavior by setting 2606 * @request_ops_lock. 2607 * 2608 * @lock protection mixes with rtnl_lock in multiple ways, fields are 2609 * either: 2610 * 2611 * - simply protected by the instance @lock; 2612 * 2613 * - double protected - writers hold both locks, readers hold either; 2614 * 2615 * - ops protected - protected by the lock held around the NDOs 2616 * and other callbacks, that is the instance lock on devices for 2617 * which netdev_need_ops_lock() returns true, otherwise by rtnl_lock; 2618 * 2619 * - double ops protected - always protected by rtnl_lock but for 2620 * devices for which netdev_need_ops_lock() returns true - also 2621 * the instance lock. 2622 * 2623 * Simply protects: 2624 * @gro_flush_timeout, @napi_defer_hard_irqs, @napi_list, 2625 * @net_shaper_hierarchy, @reg_state, @threaded 2626 * 2627 * Double protects: 2628 * @up, @moving_ns, @nd_net, @xdp_features 2629 * 2630 * Ops protects: 2631 * @cfg, @cfg_pending, @ethtool, @hwprov 2632 * 2633 * Double ops protects: 2634 * @real_num_rx_queues, @real_num_tx_queues 2635 * 2636 * Also protects some fields in: 2637 * struct napi_struct, struct netdev_queue, struct netdev_rx_queue 2638 * 2639 * Ordering: 2640 * 2641 * - take after rtnl_lock 2642 * 2643 * - for the case of netdev queue leasing, the netdev-scope lock is 2644 * taken for both the virtual and the physical device; to prevent 2645 * deadlocks, the virtual device's lock must always be acquired 2646 * before the physical device's (see netdev_nl_queue_create_doit) 2647 */ 2648 struct mutex lock; 2649 2650 #if IS_ENABLED(CONFIG_NET_SHAPER) 2651 /** 2652 * @net_shaper_hierarchy: data tracking the current shaper status 2653 * see include/net/net_shapers.h 2654 */ 2655 struct net_shaper_hierarchy *net_shaper_hierarchy; 2656 #endif 2657 2658 struct hlist_head neighbours[NEIGH_NR_TABLES]; 2659 2660 struct hwtstamp_provider __rcu *hwprov; 2661 2662 u8 priv[] ____cacheline_aligned 2663 __counted_by(priv_len); 2664 } ____cacheline_aligned; 2665 #define to_net_dev(d) container_of(d, struct net_device, dev) 2666 2667 /* 2668 * Driver should use this to assign devlink port instance to a netdevice 2669 * before it registers the netdevice. Therefore devlink_port is static 2670 * during the netdev lifetime after it is registered. 2671 */ 2672 #define SET_NETDEV_DEVLINK_PORT(dev, port) \ 2673 ({ \ 2674 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \ 2675 ((dev)->devlink_port = (port)); \ 2676 }) 2677 2678 static inline bool netif_elide_gro(const struct net_device *dev) 2679 { 2680 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2681 return true; 2682 return false; 2683 } 2684 2685 #define NETDEV_ALIGN 32 2686 2687 static inline 2688 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2689 { 2690 return READ_ONCE(dev->prio_tc_map[prio & TC_BITMASK]); 2691 } 2692 2693 static inline 2694 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2695 { 2696 if (tc >= READ_ONCE(dev->num_tc)) 2697 return -EINVAL; 2698 2699 WRITE_ONCE(dev->prio_tc_map[prio & TC_BITMASK], tc & TC_BITMASK); 2700 return 0; 2701 } 2702 2703 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2704 void netdev_reset_tc(struct net_device *dev); 2705 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2706 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2707 2708 static inline 2709 int netdev_get_num_tc(const struct net_device *dev) 2710 { 2711 return READ_ONCE(dev->num_tc); 2712 } 2713 2714 static inline void net_prefetch(void *p) 2715 { 2716 prefetch(p); 2717 #if L1_CACHE_BYTES < 128 2718 prefetch((u8 *)p + L1_CACHE_BYTES); 2719 #endif 2720 } 2721 2722 static inline void net_prefetchw(void *p) 2723 { 2724 prefetchw(p); 2725 #if L1_CACHE_BYTES < 128 2726 prefetchw((u8 *)p + L1_CACHE_BYTES); 2727 #endif 2728 } 2729 2730 void netdev_unbind_sb_channel(struct net_device *dev, 2731 struct net_device *sb_dev); 2732 int netdev_bind_sb_channel_queue(struct net_device *dev, 2733 struct net_device *sb_dev, 2734 u8 tc, u16 count, u16 offset); 2735 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2736 static inline int netdev_get_sb_channel(struct net_device *dev) 2737 { 2738 return max_t(int, -READ_ONCE(dev->num_tc), 0); 2739 } 2740 2741 static inline 2742 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2743 unsigned int index) 2744 { 2745 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues); 2746 return &dev->_tx[index]; 2747 } 2748 2749 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2750 const struct sk_buff *skb) 2751 { 2752 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2753 } 2754 2755 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2756 void (*f)(struct net_device *, 2757 struct netdev_queue *, 2758 void *), 2759 void *arg) 2760 { 2761 unsigned int i; 2762 2763 for (i = 0; i < dev->num_tx_queues; i++) 2764 f(dev, &dev->_tx[i], arg); 2765 } 2766 2767 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2768 struct net_device *sb_dev); 2769 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2770 struct sk_buff *skb, 2771 struct net_device *sb_dev); 2772 2773 /* returns the headroom that the master device needs to take in account 2774 * when forwarding to this dev 2775 */ 2776 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2777 { 2778 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2779 } 2780 2781 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2782 { 2783 if (dev->netdev_ops->ndo_set_rx_headroom) 2784 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2785 } 2786 2787 /* set the device rx headroom to the dev's default */ 2788 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2789 { 2790 netdev_set_rx_headroom(dev, -1); 2791 } 2792 2793 static inline void *netdev_get_ml_priv(struct net_device *dev, 2794 enum netdev_ml_priv_type type) 2795 { 2796 if (dev->ml_priv_type != type) 2797 return NULL; 2798 2799 return dev->ml_priv; 2800 } 2801 2802 static inline void netdev_set_ml_priv(struct net_device *dev, 2803 void *ml_priv, 2804 enum netdev_ml_priv_type type) 2805 { 2806 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2807 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2808 dev->ml_priv_type, type); 2809 WARN(!dev->ml_priv_type && dev->ml_priv, 2810 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2811 2812 dev->ml_priv = ml_priv; 2813 dev->ml_priv_type = type; 2814 } 2815 2816 /* 2817 * Net namespace inlines 2818 */ 2819 static inline 2820 struct net *dev_net(const struct net_device *dev) 2821 { 2822 return read_pnet(&dev->nd_net); 2823 } 2824 2825 static inline 2826 struct net *dev_net_rcu(const struct net_device *dev) 2827 { 2828 return read_pnet_rcu(&dev->nd_net); 2829 } 2830 2831 static inline 2832 void dev_net_set(struct net_device *dev, struct net *net) 2833 { 2834 write_pnet(&dev->nd_net, net); 2835 } 2836 2837 /** 2838 * netdev_priv - access network device private data 2839 * @dev: network device 2840 * 2841 * Get network device private data 2842 */ 2843 static inline void *netdev_priv(const struct net_device *dev) 2844 { 2845 return (void *)dev->priv; 2846 } 2847 2848 /** 2849 * netdev_from_priv() - get network device from priv 2850 * @priv: network device private data 2851 * 2852 * Returns: net_device to which @priv belongs 2853 */ 2854 static inline struct net_device *netdev_from_priv(const void *priv) 2855 { 2856 /* 2857 * When GCC can see that @priv came from netdev_priv(), it can 2858 * compute an object size of 0 for the net_device and its members, 2859 * which makes FORTIFY_SOURCE fail valid accesses through the result: 2860 * https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127875 2861 */ 2862 OPTIMIZER_HIDE_VAR(priv); 2863 return container_of(priv, struct net_device, priv); 2864 } 2865 2866 /* Set the sysfs physical device reference for the network logical device 2867 * if set prior to registration will cause a symlink during initialization. 2868 */ 2869 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2870 2871 /* Set the sysfs device type for the network logical device to allow 2872 * fine-grained identification of different network device types. For 2873 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2874 */ 2875 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2876 2877 void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index, 2878 enum netdev_queue_type type, 2879 struct napi_struct *napi); 2880 2881 static inline void netdev_lock(struct net_device *dev) 2882 { 2883 mutex_lock(&dev->lock); 2884 } 2885 2886 static inline void netdev_unlock(struct net_device *dev) 2887 { 2888 mutex_unlock(&dev->lock); 2889 } 2890 /* Additional netdev_lock()-related helpers are in net/netdev_lock.h */ 2891 2892 void netif_napi_set_irq_locked(struct napi_struct *napi, int irq); 2893 2894 static inline void netif_napi_set_irq(struct napi_struct *napi, int irq) 2895 { 2896 netdev_lock(napi->dev); 2897 netif_napi_set_irq_locked(napi, irq); 2898 netdev_unlock(napi->dev); 2899 } 2900 2901 /* Default NAPI poll() weight 2902 * Device drivers are strongly advised to not use bigger value 2903 */ 2904 #define NAPI_POLL_WEIGHT 64 2905 2906 void netif_napi_add_weight_locked(struct net_device *dev, 2907 struct napi_struct *napi, 2908 int (*poll)(struct napi_struct *, int), 2909 int weight); 2910 2911 static inline void 2912 netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2913 int (*poll)(struct napi_struct *, int), int weight) 2914 { 2915 netdev_lock(dev); 2916 netif_napi_add_weight_locked(dev, napi, poll, weight); 2917 netdev_unlock(dev); 2918 } 2919 2920 /** 2921 * netif_napi_add() - initialize a NAPI context 2922 * @dev: network device 2923 * @napi: NAPI context 2924 * @poll: polling function 2925 * 2926 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2927 * *any* of the other NAPI-related functions. 2928 */ 2929 static inline void 2930 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2931 int (*poll)(struct napi_struct *, int)) 2932 { 2933 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2934 } 2935 2936 static inline void 2937 netif_napi_add_locked(struct net_device *dev, struct napi_struct *napi, 2938 int (*poll)(struct napi_struct *, int)) 2939 { 2940 netif_napi_add_weight_locked(dev, napi, poll, NAPI_POLL_WEIGHT); 2941 } 2942 2943 static inline void 2944 netif_napi_add_tx_weight(struct net_device *dev, 2945 struct napi_struct *napi, 2946 int (*poll)(struct napi_struct *, int), 2947 int weight) 2948 { 2949 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2950 netif_napi_add_weight(dev, napi, poll, weight); 2951 } 2952 2953 static inline void 2954 netif_napi_add_config_locked(struct net_device *dev, struct napi_struct *napi, 2955 int (*poll)(struct napi_struct *, int), int index) 2956 { 2957 napi->index = index; 2958 napi->config = &dev->napi_config[index]; 2959 netif_napi_add_weight_locked(dev, napi, poll, NAPI_POLL_WEIGHT); 2960 } 2961 2962 /** 2963 * netif_napi_add_config - initialize a NAPI context with persistent config 2964 * @dev: network device 2965 * @napi: NAPI context 2966 * @poll: polling function 2967 * @index: the NAPI index 2968 */ 2969 static inline void 2970 netif_napi_add_config(struct net_device *dev, struct napi_struct *napi, 2971 int (*poll)(struct napi_struct *, int), int index) 2972 { 2973 netdev_lock(dev); 2974 netif_napi_add_config_locked(dev, napi, poll, index); 2975 netdev_unlock(dev); 2976 } 2977 2978 /** 2979 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2980 * @dev: network device 2981 * @napi: NAPI context 2982 * @poll: polling function 2983 * 2984 * This variant of netif_napi_add() should be used from drivers using NAPI 2985 * to exclusively poll a TX queue. 2986 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2987 */ 2988 static inline void netif_napi_add_tx(struct net_device *dev, 2989 struct napi_struct *napi, 2990 int (*poll)(struct napi_struct *, int)) 2991 { 2992 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2993 } 2994 2995 void __netif_napi_del_locked(struct napi_struct *napi); 2996 2997 /** 2998 * __netif_napi_del - remove a NAPI context 2999 * @napi: NAPI context 3000 * 3001 * Warning: caller must observe RCU grace period before freeing memory 3002 * containing @napi. Drivers might want to call this helper to combine 3003 * all the needed RCU grace periods into a single one. 3004 */ 3005 static inline void __netif_napi_del(struct napi_struct *napi) 3006 { 3007 netdev_lock(napi->dev); 3008 __netif_napi_del_locked(napi); 3009 netdev_unlock(napi->dev); 3010 } 3011 3012 static inline void netif_napi_del_locked(struct napi_struct *napi) 3013 { 3014 __netif_napi_del_locked(napi); 3015 synchronize_net(); 3016 } 3017 3018 /** 3019 * netif_napi_del - remove a NAPI context 3020 * @napi: NAPI context 3021 * 3022 * netif_napi_del() removes a NAPI context from the network device NAPI list 3023 */ 3024 static inline void netif_napi_del(struct napi_struct *napi) 3025 { 3026 __netif_napi_del(napi); 3027 synchronize_net(); 3028 } 3029 3030 int netif_enable_cpu_rmap(struct net_device *dev, unsigned int num_irqs); 3031 void netif_set_affinity_auto(struct net_device *dev); 3032 3033 struct packet_type { 3034 __be16 type; /* This is really htons(ether_type). */ 3035 bool ignore_outgoing; 3036 struct net_device *dev; /* NULL is wildcarded here */ 3037 netdevice_tracker dev_tracker; 3038 int (*func) (struct sk_buff *, 3039 struct net_device *, 3040 struct packet_type *, 3041 struct net_device *); 3042 void (*list_func) (struct list_head *, 3043 struct packet_type *, 3044 struct net_device *); 3045 bool (*id_match)(struct packet_type *ptype, 3046 struct sock *sk); 3047 struct net *af_packet_net; 3048 void *af_packet_priv; 3049 struct list_head list; 3050 }; 3051 3052 struct offload_callbacks { 3053 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 3054 netdev_features_t features); 3055 struct sk_buff *(*gro_receive)(struct list_head *head, 3056 struct sk_buff *skb); 3057 int (*gro_complete)(struct sk_buff *skb, int nhoff); 3058 }; 3059 3060 struct packet_offload { 3061 __be16 type; /* This is really htons(ether_type). */ 3062 u16 priority; 3063 struct offload_callbacks callbacks; 3064 struct list_head list; 3065 }; 3066 3067 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 3068 struct pcpu_sw_netstats { 3069 u64_stats_t rx_packets; 3070 u64_stats_t rx_bytes; 3071 u64_stats_t tx_packets; 3072 u64_stats_t tx_bytes; 3073 struct u64_stats_sync syncp; 3074 } __aligned(4 * sizeof(u64)); 3075 3076 struct pcpu_dstats { 3077 u64_stats_t rx_packets; 3078 u64_stats_t rx_bytes; 3079 u64_stats_t tx_packets; 3080 u64_stats_t tx_bytes; 3081 u64_stats_t rx_drops; 3082 u64_stats_t tx_drops; 3083 struct u64_stats_sync syncp; 3084 } __aligned(8 * sizeof(u64)); 3085 3086 struct pcpu_lstats { 3087 u64_stats_t packets; 3088 u64_stats_t bytes; 3089 struct u64_stats_sync syncp; 3090 } __aligned(2 * sizeof(u64)); 3091 3092 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 3093 3094 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 3095 { 3096 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 3097 3098 u64_stats_update_begin(&tstats->syncp); 3099 u64_stats_add(&tstats->rx_bytes, len); 3100 u64_stats_inc(&tstats->rx_packets); 3101 u64_stats_update_end(&tstats->syncp); 3102 } 3103 3104 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 3105 unsigned int packets, 3106 unsigned int len) 3107 { 3108 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 3109 3110 u64_stats_update_begin(&tstats->syncp); 3111 u64_stats_add(&tstats->tx_bytes, len); 3112 u64_stats_add(&tstats->tx_packets, packets); 3113 u64_stats_update_end(&tstats->syncp); 3114 } 3115 3116 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 3117 { 3118 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 3119 3120 u64_stats_update_begin(&lstats->syncp); 3121 u64_stats_add(&lstats->bytes, len); 3122 u64_stats_inc(&lstats->packets); 3123 u64_stats_update_end(&lstats->syncp); 3124 } 3125 3126 static inline void dev_dstats_rx_add(struct net_device *dev, 3127 unsigned int len) 3128 { 3129 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3130 3131 u64_stats_update_begin(&dstats->syncp); 3132 u64_stats_inc(&dstats->rx_packets); 3133 u64_stats_add(&dstats->rx_bytes, len); 3134 u64_stats_update_end(&dstats->syncp); 3135 } 3136 3137 static inline void dev_dstats_rx_dropped(struct net_device *dev) 3138 { 3139 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3140 3141 u64_stats_update_begin(&dstats->syncp); 3142 u64_stats_inc(&dstats->rx_drops); 3143 u64_stats_update_end(&dstats->syncp); 3144 } 3145 3146 static inline void dev_dstats_rx_dropped_add(struct net_device *dev, 3147 unsigned int packets) 3148 { 3149 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3150 3151 u64_stats_update_begin(&dstats->syncp); 3152 u64_stats_add(&dstats->rx_drops, packets); 3153 u64_stats_update_end(&dstats->syncp); 3154 } 3155 3156 static inline void dev_dstats_tx_add(struct net_device *dev, 3157 unsigned int len) 3158 { 3159 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3160 3161 u64_stats_update_begin(&dstats->syncp); 3162 u64_stats_inc(&dstats->tx_packets); 3163 u64_stats_add(&dstats->tx_bytes, len); 3164 u64_stats_update_end(&dstats->syncp); 3165 } 3166 3167 static inline void dev_dstats_tx_dropped(struct net_device *dev) 3168 { 3169 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3170 3171 u64_stats_update_begin(&dstats->syncp); 3172 u64_stats_inc(&dstats->tx_drops); 3173 u64_stats_update_end(&dstats->syncp); 3174 } 3175 3176 #define __netdev_alloc_pcpu_stats(type, gfp) \ 3177 ({ \ 3178 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 3179 if (pcpu_stats) { \ 3180 int __cpu; \ 3181 for_each_possible_cpu(__cpu) { \ 3182 typeof(type) *stat; \ 3183 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 3184 u64_stats_init(&stat->syncp); \ 3185 } \ 3186 } \ 3187 pcpu_stats; \ 3188 }) 3189 3190 #define netdev_alloc_pcpu_stats(type) \ 3191 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 3192 3193 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 3194 ({ \ 3195 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 3196 if (pcpu_stats) { \ 3197 int __cpu; \ 3198 for_each_possible_cpu(__cpu) { \ 3199 typeof(type) *stat; \ 3200 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 3201 u64_stats_init(&stat->syncp); \ 3202 } \ 3203 } \ 3204 pcpu_stats; \ 3205 }) 3206 3207 enum netdev_lag_tx_type { 3208 NETDEV_LAG_TX_TYPE_UNKNOWN, 3209 NETDEV_LAG_TX_TYPE_RANDOM, 3210 NETDEV_LAG_TX_TYPE_BROADCAST, 3211 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 3212 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 3213 NETDEV_LAG_TX_TYPE_HASH, 3214 }; 3215 3216 enum netdev_lag_hash { 3217 NETDEV_LAG_HASH_NONE, 3218 NETDEV_LAG_HASH_L2, 3219 NETDEV_LAG_HASH_L34, 3220 NETDEV_LAG_HASH_L23, 3221 NETDEV_LAG_HASH_E23, 3222 NETDEV_LAG_HASH_E34, 3223 NETDEV_LAG_HASH_VLAN_SRCMAC, 3224 NETDEV_LAG_HASH_UNKNOWN, 3225 }; 3226 3227 struct netdev_lag_upper_info { 3228 enum netdev_lag_tx_type tx_type; 3229 enum netdev_lag_hash hash_type; 3230 }; 3231 3232 struct netdev_lag_lower_state_info { 3233 u8 link_up : 1, 3234 tx_enabled : 1; 3235 }; 3236 3237 #include <linux/notifier.h> 3238 3239 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 3240 * and the rtnetlink notification exclusion list in rtnetlink_event() when 3241 * adding new types. 3242 */ 3243 enum netdev_cmd { 3244 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 3245 NETDEV_DOWN, 3246 NETDEV_REBOOT, /* Tell a protocol stack a network interface 3247 detected a hardware crash and restarted 3248 - we can use this eg to kick tcp sessions 3249 once done */ 3250 NETDEV_CHANGE, /* Notify device state change */ 3251 NETDEV_REGISTER, 3252 NETDEV_UNREGISTER, 3253 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 3254 NETDEV_CHANGEADDR, /* notify after the address change */ 3255 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 3256 NETDEV_GOING_DOWN, 3257 NETDEV_CHANGENAME, 3258 NETDEV_FEAT_CHANGE, 3259 NETDEV_BONDING_FAILOVER, 3260 NETDEV_PRE_UP, 3261 NETDEV_PRE_TYPE_CHANGE, 3262 NETDEV_POST_TYPE_CHANGE, 3263 NETDEV_POST_INIT, 3264 NETDEV_PRE_UNINIT, 3265 NETDEV_RELEASE, 3266 NETDEV_NOTIFY_PEERS, 3267 NETDEV_JOIN, 3268 NETDEV_CHANGEUPPER, 3269 NETDEV_RESEND_IGMP, 3270 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 3271 NETDEV_CHANGEINFODATA, 3272 NETDEV_BONDING_INFO, 3273 NETDEV_PRECHANGEUPPER, 3274 NETDEV_CHANGELOWERSTATE, 3275 NETDEV_UDP_TUNNEL_PUSH_INFO, 3276 NETDEV_UDP_TUNNEL_DROP_INFO, 3277 NETDEV_CHANGE_TX_QUEUE_LEN, 3278 NETDEV_CVLAN_FILTER_PUSH_INFO, 3279 NETDEV_CVLAN_FILTER_DROP_INFO, 3280 NETDEV_SVLAN_FILTER_PUSH_INFO, 3281 NETDEV_SVLAN_FILTER_DROP_INFO, 3282 NETDEV_OFFLOAD_XSTATS_ENABLE, 3283 NETDEV_OFFLOAD_XSTATS_DISABLE, 3284 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 3285 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 3286 NETDEV_XDP_FEAT_CHANGE, 3287 }; 3288 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 3289 3290 int register_netdevice_notifier(struct notifier_block *nb); 3291 int unregister_netdevice_notifier(struct notifier_block *nb); 3292 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 3293 int unregister_netdevice_notifier_net(struct net *net, 3294 struct notifier_block *nb); 3295 int register_netdevice_notifier_dev_net(struct net_device *dev, 3296 struct notifier_block *nb, 3297 struct netdev_net_notifier *nn); 3298 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 3299 struct notifier_block *nb, 3300 struct netdev_net_notifier *nn); 3301 3302 struct netdev_notifier_info { 3303 struct net_device *dev; 3304 struct netlink_ext_ack *extack; 3305 }; 3306 3307 struct netdev_notifier_info_ext { 3308 struct netdev_notifier_info info; /* must be first */ 3309 union { 3310 u32 mtu; 3311 } ext; 3312 }; 3313 3314 struct netdev_notifier_change_info { 3315 struct netdev_notifier_info info; /* must be first */ 3316 unsigned int flags_changed; 3317 }; 3318 3319 struct netdev_notifier_changeupper_info { 3320 struct netdev_notifier_info info; /* must be first */ 3321 struct net_device *upper_dev; /* new upper dev */ 3322 bool master; /* is upper dev master */ 3323 bool linking; /* is the notification for link or unlink */ 3324 void *upper_info; /* upper dev info */ 3325 }; 3326 3327 struct netdev_notifier_changelowerstate_info { 3328 struct netdev_notifier_info info; /* must be first */ 3329 void *lower_state_info; /* is lower dev state */ 3330 }; 3331 3332 struct netdev_notifier_pre_changeaddr_info { 3333 struct netdev_notifier_info info; /* must be first */ 3334 const unsigned char *dev_addr; 3335 }; 3336 3337 enum netdev_offload_xstats_type { 3338 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 3339 }; 3340 3341 struct netdev_notifier_offload_xstats_info { 3342 struct netdev_notifier_info info; /* must be first */ 3343 enum netdev_offload_xstats_type type; 3344 3345 union { 3346 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 3347 struct netdev_notifier_offload_xstats_rd *report_delta; 3348 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 3349 struct netdev_notifier_offload_xstats_ru *report_used; 3350 }; 3351 }; 3352 3353 int netdev_offload_xstats_enable(struct net_device *dev, 3354 enum netdev_offload_xstats_type type, 3355 struct netlink_ext_ack *extack); 3356 int netdev_offload_xstats_disable(struct net_device *dev, 3357 enum netdev_offload_xstats_type type); 3358 bool netdev_offload_xstats_enabled(const struct net_device *dev, 3359 enum netdev_offload_xstats_type type); 3360 int netdev_offload_xstats_get(struct net_device *dev, 3361 enum netdev_offload_xstats_type type, 3362 struct rtnl_hw_stats64 *stats, bool *used, 3363 struct netlink_ext_ack *extack); 3364 void 3365 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 3366 const struct rtnl_hw_stats64 *stats); 3367 void 3368 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 3369 void netdev_offload_xstats_push_delta(struct net_device *dev, 3370 enum netdev_offload_xstats_type type, 3371 const struct rtnl_hw_stats64 *stats); 3372 3373 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 3374 struct net_device *dev) 3375 { 3376 info->dev = dev; 3377 info->extack = NULL; 3378 } 3379 3380 static inline struct net_device * 3381 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 3382 { 3383 return info->dev; 3384 } 3385 3386 static inline struct netlink_ext_ack * 3387 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 3388 { 3389 return info->extack; 3390 } 3391 3392 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 3393 int call_netdevice_notifiers_info(unsigned long val, 3394 struct netdev_notifier_info *info); 3395 3396 #define for_each_netdev(net, d) \ 3397 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 3398 #define for_each_netdev_reverse(net, d) \ 3399 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 3400 #define for_each_netdev_rcu(net, d) \ 3401 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 3402 #define for_each_netdev_safe(net, d, n) \ 3403 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 3404 #define for_each_netdev_continue(net, d) \ 3405 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 3406 #define for_each_netdev_continue_reverse(net, d) \ 3407 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 3408 dev_list) 3409 #define for_each_netdev_continue_rcu(net, d) \ 3410 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 3411 #define for_each_netdev_in_bond_rcu(bond, slave) \ 3412 for_each_netdev_rcu(dev_net_rcu(bond), slave) \ 3413 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 3414 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 3415 3416 #define for_each_netdev_dump(net, d, ifindex) \ 3417 for (; (d = xa_find(&(net)->dev_by_index, &ifindex, \ 3418 ULONG_MAX, XA_PRESENT)); ifindex++) 3419 3420 static inline struct net_device *next_net_device(struct net_device *dev) 3421 { 3422 struct list_head *lh; 3423 struct net *net; 3424 3425 net = dev_net(dev); 3426 lh = dev->dev_list.next; 3427 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3428 } 3429 3430 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 3431 { 3432 struct list_head *lh; 3433 struct net *net; 3434 3435 net = dev_net(dev); 3436 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 3437 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3438 } 3439 3440 static inline struct net_device *first_net_device(struct net *net) 3441 { 3442 return list_empty(&net->dev_base_head) ? NULL : 3443 net_device_entry(net->dev_base_head.next); 3444 } 3445 3446 struct net_device *dev_getbyhwaddr(struct net *net, unsigned short type, 3447 const char *hwaddr); 3448 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 3449 const char *hwaddr); 3450 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 3451 void dev_add_pack(struct packet_type *pt); 3452 void dev_remove_pack(struct packet_type *pt); 3453 void __dev_remove_pack(struct packet_type *pt); 3454 void dev_add_offload(struct packet_offload *po); 3455 void dev_remove_offload(struct packet_offload *po); 3456 3457 int dev_get_iflink(const struct net_device *dev); 3458 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 3459 int dev_fill_forward_path(struct net_device_path_ctx *ctx, 3460 struct net_device_path_stack *stack); 3461 void dev_fill_forward_path_release(struct net_device_path_stack *stack); 3462 struct net_device *dev_get_by_name(struct net *net, const char *name); 3463 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 3464 struct net_device *__dev_get_by_name(struct net *net, const char *name); 3465 bool netdev_name_in_use(struct net *net, const char *name); 3466 int dev_alloc_name(struct net_device *dev, const char *name); 3467 int netif_open(struct net_device *dev, struct netlink_ext_ack *extack); 3468 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 3469 void netif_close(struct net_device *dev); 3470 void dev_close(struct net_device *dev); 3471 void netif_close_many(struct list_head *head, bool unlink); 3472 void netif_disable_lro(struct net_device *dev); 3473 void dev_disable_lro(struct net_device *dev); 3474 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 3475 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 3476 struct net_device *sb_dev); 3477 3478 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3479 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3480 3481 static inline int dev_queue_xmit(struct sk_buff *skb) 3482 { 3483 return __dev_queue_xmit(skb, NULL); 3484 } 3485 3486 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3487 struct net_device *sb_dev) 3488 { 3489 return __dev_queue_xmit(skb, sb_dev); 3490 } 3491 3492 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3493 { 3494 int ret; 3495 3496 ret = __dev_direct_xmit(skb, queue_id); 3497 if (!dev_xmit_complete(ret)) 3498 kfree_skb(skb); 3499 return ret; 3500 } 3501 3502 int register_netdevice(struct net_device *dev); 3503 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3504 void unregister_netdevice_many(struct list_head *head); 3505 bool unregister_netdevice_queued(const struct net_device *dev); 3506 3507 static inline void unregister_netdevice(struct net_device *dev) 3508 { 3509 unregister_netdevice_queue(dev, NULL); 3510 } 3511 3512 #ifdef CONFIG_DEBUG_NET_SMALL_RTNL 3513 void unregister_netdevice_queue_net(struct net *net, struct net_device *dev, 3514 struct list_head *head); 3515 void unregister_netdevice_many_net(struct net *net); 3516 void unregister_netdevice_queue_many_net(struct net *net, struct list_head *head); 3517 #else 3518 static inline void unregister_netdevice_queue_net(struct net *net, 3519 struct net_device *dev, 3520 struct list_head *head) 3521 { 3522 unregister_netdevice_queue(dev, head); 3523 } 3524 3525 static inline void unregister_netdevice_queue_many_net(struct net *net, 3526 struct list_head *head) 3527 { 3528 } 3529 #endif 3530 3531 int netdev_refcnt_read(const struct net_device *dev); 3532 void free_netdev(struct net_device *dev); 3533 3534 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3535 struct sk_buff *skb, 3536 bool all_slaves); 3537 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3538 struct sock *sk); 3539 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3540 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3541 struct net_device *netdev_get_by_index(struct net *net, int ifindex, 3542 netdevice_tracker *tracker, gfp_t gfp); 3543 struct net_device *netdev_get_by_index_lock(struct net *net, int ifindex); 3544 struct net_device *netdev_get_by_name(struct net *net, const char *name, 3545 netdevice_tracker *tracker, gfp_t gfp); 3546 struct net_device *netdev_get_by_flags_rcu(struct net *net, netdevice_tracker *tracker, 3547 unsigned short flags, unsigned short mask); 3548 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3549 void netdev_copy_name(struct net_device *dev, char *name); 3550 3551 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3552 unsigned short type, 3553 const void *daddr, const void *saddr, 3554 unsigned int len) 3555 { 3556 if (!dev->header_ops || !dev->header_ops->create) 3557 return 0; 3558 3559 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3560 } 3561 3562 static inline int dev_parse_header(const struct sk_buff *skb, 3563 unsigned char *haddr) 3564 { 3565 const struct net_device *dev = skb->dev; 3566 3567 if (!dev->header_ops || !dev->header_ops->parse) 3568 return 0; 3569 return dev->header_ops->parse(skb, dev, haddr); 3570 } 3571 3572 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3573 { 3574 const struct net_device *dev = skb->dev; 3575 3576 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3577 return 0; 3578 return dev->header_ops->parse_protocol(skb); 3579 } 3580 3581 /* ll_header must have at least hard_header_len allocated */ 3582 static inline bool dev_validate_header(const struct net_device *dev, 3583 char *ll_header, int len) 3584 { 3585 if (likely(len >= dev->hard_header_len)) 3586 return true; 3587 if (len < dev->min_header_len) 3588 return false; 3589 3590 if (dev->header_ops && dev->header_ops->validate) 3591 return dev->header_ops->validate(ll_header, len); 3592 3593 return false; 3594 } 3595 3596 static inline bool dev_has_header(const struct net_device *dev) 3597 { 3598 return dev->header_ops && dev->header_ops->create; 3599 } 3600 3601 struct numa_drop_counters { 3602 atomic_t drops0 ____cacheline_aligned_in_smp; 3603 atomic_t drops1 ____cacheline_aligned_in_smp; 3604 }; 3605 3606 static inline int numa_drop_read(const struct numa_drop_counters *ndc) 3607 { 3608 return atomic_read(&ndc->drops0) + atomic_read(&ndc->drops1); 3609 } 3610 3611 static inline void numa_drop_add(struct numa_drop_counters *ndc, int val) 3612 { 3613 int n = numa_node_id() % 2; 3614 3615 if (n) 3616 atomic_add(val, &ndc->drops1); 3617 else 3618 atomic_add(val, &ndc->drops0); 3619 } 3620 3621 static inline void numa_drop_reset(struct numa_drop_counters *ndc) 3622 { 3623 atomic_set(&ndc->drops0, 0); 3624 atomic_set(&ndc->drops1, 0); 3625 } 3626 3627 /* 3628 * Incoming packets are placed on per-CPU queues 3629 */ 3630 struct softnet_data { 3631 struct list_head poll_list; 3632 struct sk_buff_head process_queue; 3633 local_lock_t process_queue_bh_lock; 3634 3635 /* stats */ 3636 unsigned int processed; 3637 unsigned int time_squeeze; 3638 #ifdef CONFIG_RPS 3639 struct softnet_data *rps_ipi_list; 3640 #endif 3641 3642 unsigned int received_rps; 3643 bool in_net_rx_action; 3644 bool in_napi_threaded_poll; 3645 3646 #ifdef CONFIG_NET_FLOW_LIMIT 3647 struct sd_flow_limit __rcu *flow_limit; 3648 #endif 3649 struct Qdisc *output_queue; 3650 struct Qdisc **output_queue_tailp; 3651 struct sk_buff *completion_queue; 3652 #ifdef CONFIG_XFRM_OFFLOAD 3653 struct sk_buff_head xfrm_backlog; 3654 #endif 3655 /* written and read only by owning cpu: */ 3656 struct netdev_xmit xmit; 3657 #ifdef CONFIG_RPS 3658 /* input_queue_head should be written by cpu owning this struct, 3659 * and only read by other cpus. Worth using a cache line. 3660 */ 3661 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3662 3663 /* Elements below can be accessed between CPUs for RPS/RFS */ 3664 call_single_data_t csd ____cacheline_aligned_in_smp; 3665 struct softnet_data *rps_ipi_next; 3666 unsigned int cpu; 3667 3668 /* We force a cacheline alignment from here, to hold together 3669 * input_queue_tail, input_pkt_queue and backlog.state. 3670 * We add holes so that backlog.state is the last field 3671 * of this cache line. 3672 */ 3673 long pad[3] ____cacheline_aligned_in_smp; 3674 unsigned int input_queue_tail; 3675 #endif 3676 struct sk_buff_head input_pkt_queue; 3677 3678 struct napi_struct backlog; 3679 3680 struct numa_drop_counters drop_counters; 3681 3682 int defer_ipi_scheduled ____cacheline_aligned_in_smp; 3683 call_single_data_t defer_csd; 3684 }; 3685 3686 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3687 3688 struct page_pool_bh { 3689 struct page_pool *pool; 3690 local_lock_t bh_lock; 3691 }; 3692 DECLARE_PER_CPU(struct page_pool_bh, system_page_pool); 3693 3694 #ifdef CONFIG_KASAN 3695 #define XMIT_RECURSION_LIMIT 4 3696 #else 3697 #define XMIT_RECURSION_LIMIT 8 3698 #endif 3699 3700 #ifndef CONFIG_PREEMPT_RT 3701 static inline int dev_recursion_level(void) 3702 { 3703 return this_cpu_read(softnet_data.xmit.recursion); 3704 } 3705 3706 static inline bool dev_xmit_recursion(void) 3707 { 3708 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3709 XMIT_RECURSION_LIMIT); 3710 } 3711 3712 static inline void dev_xmit_recursion_inc(void) 3713 { 3714 __this_cpu_inc(softnet_data.xmit.recursion); 3715 } 3716 3717 static inline void dev_xmit_recursion_dec(void) 3718 { 3719 __this_cpu_dec(softnet_data.xmit.recursion); 3720 } 3721 #else 3722 static inline int dev_recursion_level(void) 3723 { 3724 return current->net_xmit.recursion; 3725 } 3726 3727 static inline bool dev_xmit_recursion(void) 3728 { 3729 return unlikely(current->net_xmit.recursion > XMIT_RECURSION_LIMIT); 3730 } 3731 3732 static inline void dev_xmit_recursion_inc(void) 3733 { 3734 current->net_xmit.recursion++; 3735 } 3736 3737 static inline void dev_xmit_recursion_dec(void) 3738 { 3739 current->net_xmit.recursion--; 3740 } 3741 #endif 3742 3743 void __netif_schedule(struct Qdisc *q); 3744 void netif_schedule_queue(struct netdev_queue *txq); 3745 3746 static inline void netif_tx_schedule_all(struct net_device *dev) 3747 { 3748 unsigned int i; 3749 3750 for (i = 0; i < dev->num_tx_queues; i++) 3751 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3752 } 3753 3754 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3755 { 3756 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3757 } 3758 3759 /** 3760 * netif_start_queue - allow transmit 3761 * @dev: network device 3762 * 3763 * Allow upper layers to call the device hard_start_xmit routine. 3764 */ 3765 static inline void netif_start_queue(struct net_device *dev) 3766 { 3767 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3768 } 3769 3770 static inline void netif_tx_start_all_queues(struct net_device *dev) 3771 { 3772 unsigned int i; 3773 3774 for (i = 0; i < dev->num_tx_queues; i++) { 3775 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3776 netif_tx_start_queue(txq); 3777 } 3778 } 3779 3780 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3781 3782 /** 3783 * netif_wake_queue - restart transmit 3784 * @dev: network device 3785 * 3786 * Allow upper layers to call the device hard_start_xmit routine. 3787 * Used for flow control when transmit resources are available. 3788 */ 3789 static inline void netif_wake_queue(struct net_device *dev) 3790 { 3791 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3792 } 3793 3794 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3795 { 3796 unsigned int i; 3797 3798 for (i = 0; i < dev->num_tx_queues; i++) { 3799 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3800 netif_tx_wake_queue(txq); 3801 } 3802 } 3803 3804 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3805 { 3806 /* Paired with READ_ONCE() from dev_watchdog() */ 3807 WRITE_ONCE(dev_queue->trans_start, jiffies); 3808 3809 /* This barrier is paired with smp_mb() from dev_watchdog() */ 3810 smp_mb__before_atomic(); 3811 3812 /* Must be an atomic op see netif_txq_try_stop() */ 3813 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3814 } 3815 3816 /** 3817 * netif_stop_queue - stop transmitted packets 3818 * @dev: network device 3819 * 3820 * Stop upper layers calling the device hard_start_xmit routine. 3821 * Used for flow control when transmit resources are unavailable. 3822 */ 3823 static inline void netif_stop_queue(struct net_device *dev) 3824 { 3825 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3826 } 3827 3828 void netif_tx_stop_all_queues(struct net_device *dev); 3829 3830 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3831 { 3832 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3833 } 3834 3835 /** 3836 * netif_queue_stopped - test if transmit queue is flowblocked 3837 * @dev: network device 3838 * 3839 * Test if transmit queue on device is currently unable to send. 3840 */ 3841 static inline bool netif_queue_stopped(const struct net_device *dev) 3842 { 3843 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3844 } 3845 3846 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3847 { 3848 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3849 } 3850 3851 static inline bool 3852 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3853 { 3854 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3855 } 3856 3857 static inline bool 3858 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3859 { 3860 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3861 } 3862 3863 /** 3864 * netdev_queue_set_dql_min_limit - set dql minimum limit 3865 * @dev_queue: pointer to transmit queue 3866 * @min_limit: dql minimum limit 3867 * 3868 * Forces xmit_more() to return true until the minimum threshold 3869 * defined by @min_limit is reached (or until the tx queue is 3870 * empty). Warning: to be use with care, misuse will impact the 3871 * latency. 3872 */ 3873 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3874 unsigned int min_limit) 3875 { 3876 #ifdef CONFIG_BQL 3877 dev_queue->dql.min_limit = min_limit; 3878 #endif 3879 } 3880 3881 static inline int netdev_queue_dql_avail(const struct netdev_queue *txq) 3882 { 3883 #ifdef CONFIG_BQL 3884 /* Non-BQL migrated drivers will return 0, too. */ 3885 return dql_avail(&txq->dql); 3886 #else 3887 return 0; 3888 #endif 3889 } 3890 3891 /** 3892 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3893 * @dev_queue: pointer to transmit queue 3894 * 3895 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3896 * to give appropriate hint to the CPU. 3897 */ 3898 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3899 { 3900 #ifdef CONFIG_BQL 3901 prefetchw(&dev_queue->dql.num_queued); 3902 #endif 3903 } 3904 3905 /** 3906 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3907 * @dev_queue: pointer to transmit queue 3908 * 3909 * BQL enabled drivers might use this helper in their TX completion path, 3910 * to give appropriate hint to the CPU. 3911 */ 3912 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3913 { 3914 #ifdef CONFIG_BQL 3915 prefetchw(&dev_queue->dql.limit); 3916 #endif 3917 } 3918 3919 /** 3920 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3921 * @dev_queue: network device queue 3922 * @bytes: number of bytes queued to the device queue 3923 * 3924 * Report the number of bytes queued for sending/completion to the network 3925 * device hardware queue. @bytes should be a good approximation and should 3926 * exactly match netdev_completed_queue() @bytes. 3927 * This is typically called once per packet, from ndo_start_xmit(). 3928 */ 3929 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3930 unsigned int bytes) 3931 { 3932 #ifdef CONFIG_BQL 3933 dql_queued(&dev_queue->dql, bytes); 3934 3935 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3936 return; 3937 3938 /* Paired with READ_ONCE() from dev_watchdog() */ 3939 WRITE_ONCE(dev_queue->trans_start, jiffies); 3940 3941 /* This barrier is paired with smp_mb() from dev_watchdog() */ 3942 smp_mb__before_atomic(); 3943 3944 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3945 3946 /* 3947 * The XOFF flag must be set before checking the dql_avail below, 3948 * because in netdev_tx_completed_queue we update the dql_completed 3949 * before checking the XOFF flag. 3950 */ 3951 smp_mb__after_atomic(); 3952 3953 /* check again in case another CPU has just made room avail */ 3954 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3955 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3956 #endif 3957 } 3958 3959 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3960 * that they should not test BQL status themselves. 3961 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3962 * skb of a batch. 3963 * Returns true if the doorbell must be used to kick the NIC. 3964 */ 3965 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3966 unsigned int bytes, 3967 bool xmit_more) 3968 { 3969 if (xmit_more) { 3970 #ifdef CONFIG_BQL 3971 dql_queued(&dev_queue->dql, bytes); 3972 #endif 3973 return netif_tx_queue_stopped(dev_queue); 3974 } 3975 netdev_tx_sent_queue(dev_queue, bytes); 3976 return true; 3977 } 3978 3979 /** 3980 * netdev_sent_queue - report the number of bytes queued to hardware 3981 * @dev: network device 3982 * @bytes: number of bytes queued to the hardware device queue 3983 * 3984 * Report the number of bytes queued for sending/completion to the network 3985 * device hardware queue#0. @bytes should be a good approximation and should 3986 * exactly match netdev_completed_queue() @bytes. 3987 * This is typically called once per packet, from ndo_start_xmit(). 3988 */ 3989 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3990 { 3991 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3992 } 3993 3994 static inline bool __netdev_sent_queue(struct net_device *dev, 3995 unsigned int bytes, 3996 bool xmit_more) 3997 { 3998 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3999 xmit_more); 4000 } 4001 4002 /** 4003 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 4004 * @dev_queue: network device queue 4005 * @pkts: number of packets (currently ignored) 4006 * @bytes: number of bytes dequeued from the device queue 4007 * 4008 * Must be called at most once per TX completion round (and not per 4009 * individual packet), so that BQL can adjust its limits appropriately. 4010 */ 4011 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 4012 unsigned int pkts, unsigned int bytes) 4013 { 4014 #ifdef CONFIG_BQL 4015 if (unlikely(!bytes)) 4016 return; 4017 4018 dql_completed(&dev_queue->dql, bytes); 4019 4020 /* 4021 * Without the memory barrier there is a small possibility that 4022 * netdev_tx_sent_queue will miss the update and cause the queue to 4023 * be stopped forever 4024 */ 4025 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */ 4026 4027 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 4028 return; 4029 4030 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 4031 netif_schedule_queue(dev_queue); 4032 #endif 4033 } 4034 4035 /** 4036 * netdev_completed_queue - report bytes and packets completed by device 4037 * @dev: network device 4038 * @pkts: actual number of packets sent over the medium 4039 * @bytes: actual number of bytes sent over the medium 4040 * 4041 * Report the number of bytes and packets transmitted by the network device 4042 * hardware queue over the physical medium, @bytes must exactly match the 4043 * @bytes amount passed to netdev_sent_queue() 4044 */ 4045 static inline void netdev_completed_queue(struct net_device *dev, 4046 unsigned int pkts, unsigned int bytes) 4047 { 4048 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 4049 } 4050 4051 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 4052 { 4053 #ifdef CONFIG_BQL 4054 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 4055 dql_reset(&q->dql); 4056 #endif 4057 } 4058 4059 /** 4060 * netdev_tx_reset_subqueue - reset the BQL stats and state of a netdev queue 4061 * @dev: network device 4062 * @qid: stack index of the queue to reset 4063 */ 4064 static inline void netdev_tx_reset_subqueue(const struct net_device *dev, 4065 u32 qid) 4066 { 4067 netdev_tx_reset_queue(netdev_get_tx_queue(dev, qid)); 4068 } 4069 4070 /** 4071 * netdev_reset_queue - reset the packets and bytes count of a network device 4072 * @dev_queue: network device 4073 * 4074 * Reset the bytes and packet count of a network device and clear the 4075 * software flow control OFF bit for this network device 4076 */ 4077 static inline void netdev_reset_queue(struct net_device *dev_queue) 4078 { 4079 netdev_tx_reset_subqueue(dev_queue, 0); 4080 } 4081 4082 /** 4083 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 4084 * @dev: network device 4085 * @queue_index: given tx queue index 4086 * 4087 * Returns 0 if given tx queue index >= number of device tx queues, 4088 * otherwise returns the originally passed tx queue index. 4089 */ 4090 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 4091 { 4092 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 4093 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 4094 dev->name, queue_index, 4095 dev->real_num_tx_queues); 4096 return 0; 4097 } 4098 4099 return queue_index; 4100 } 4101 4102 /** 4103 * netif_running - test if up 4104 * @dev: network device 4105 * 4106 * Test if the device has been brought up. 4107 */ 4108 static inline bool netif_running(const struct net_device *dev) 4109 { 4110 return test_bit(__LINK_STATE_START, &dev->state); 4111 } 4112 4113 /* 4114 * Routines to manage the subqueues on a device. We only need start, 4115 * stop, and a check if it's stopped. All other device management is 4116 * done at the overall netdevice level. 4117 * Also test the device if we're multiqueue. 4118 */ 4119 4120 /** 4121 * netif_start_subqueue - allow sending packets on subqueue 4122 * @dev: network device 4123 * @queue_index: sub queue index 4124 * 4125 * Start individual transmit queue of a device with multiple transmit queues. 4126 */ 4127 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 4128 { 4129 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 4130 4131 netif_tx_start_queue(txq); 4132 } 4133 4134 /** 4135 * netif_stop_subqueue - stop sending packets on subqueue 4136 * @dev: network device 4137 * @queue_index: sub queue index 4138 * 4139 * Stop individual transmit queue of a device with multiple transmit queues. 4140 */ 4141 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 4142 { 4143 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 4144 netif_tx_stop_queue(txq); 4145 } 4146 4147 /** 4148 * __netif_subqueue_stopped - test status of subqueue 4149 * @dev: network device 4150 * @queue_index: sub queue index 4151 * 4152 * Check individual transmit queue of a device with multiple transmit queues. 4153 */ 4154 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 4155 u16 queue_index) 4156 { 4157 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 4158 4159 return netif_tx_queue_stopped(txq); 4160 } 4161 4162 /** 4163 * netif_subqueue_stopped - test status of subqueue 4164 * @dev: network device 4165 * @skb: sub queue buffer pointer 4166 * 4167 * Check individual transmit queue of a device with multiple transmit queues. 4168 */ 4169 static inline bool netif_subqueue_stopped(const struct net_device *dev, 4170 struct sk_buff *skb) 4171 { 4172 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 4173 } 4174 4175 /** 4176 * netif_wake_subqueue - allow sending packets on subqueue 4177 * @dev: network device 4178 * @queue_index: sub queue index 4179 * 4180 * Resume individual transmit queue of a device with multiple transmit queues. 4181 */ 4182 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 4183 { 4184 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 4185 4186 netif_tx_wake_queue(txq); 4187 } 4188 4189 #ifdef CONFIG_XPS 4190 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 4191 u16 index); 4192 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 4193 u16 index, enum xps_map_type type); 4194 4195 /** 4196 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 4197 * @j: CPU/Rx queue index 4198 * @mask: bitmask of all cpus/rx queues 4199 * @nr_bits: number of bits in the bitmask 4200 * 4201 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 4202 */ 4203 static inline bool netif_attr_test_mask(unsigned long j, 4204 const unsigned long *mask, 4205 unsigned int nr_bits) 4206 { 4207 cpu_max_bits_warn(j, nr_bits); 4208 return test_bit(j, mask); 4209 } 4210 4211 /** 4212 * netif_attr_test_online - Test for online CPU/Rx queue 4213 * @j: CPU/Rx queue index 4214 * @online_mask: bitmask for CPUs/Rx queues that are online 4215 * @nr_bits: number of bits in the bitmask 4216 * 4217 * Returns: true if a CPU/Rx queue is online. 4218 */ 4219 static inline bool netif_attr_test_online(unsigned long j, 4220 const unsigned long *online_mask, 4221 unsigned int nr_bits) 4222 { 4223 cpu_max_bits_warn(j, nr_bits); 4224 4225 if (online_mask) 4226 return test_bit(j, online_mask); 4227 4228 return (j < nr_bits); 4229 } 4230 4231 /** 4232 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 4233 * @n: CPU/Rx queue index 4234 * @srcp: the cpumask/Rx queue mask pointer 4235 * @nr_bits: number of bits in the bitmask 4236 * 4237 * Returns: next (after n) CPU/Rx queue index in the mask; 4238 * >= nr_bits if no further CPUs/Rx queues set. 4239 */ 4240 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 4241 unsigned int nr_bits) 4242 { 4243 /* -1 is a legal arg here. */ 4244 if (n != -1) 4245 cpu_max_bits_warn(n, nr_bits); 4246 4247 if (srcp) 4248 return find_next_bit(srcp, nr_bits, n + 1); 4249 4250 return n + 1; 4251 } 4252 4253 /** 4254 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 4255 * @n: CPU/Rx queue index 4256 * @src1p: the first CPUs/Rx queues mask pointer 4257 * @src2p: the second CPUs/Rx queues mask pointer 4258 * @nr_bits: number of bits in the bitmask 4259 * 4260 * Returns: next (after n) CPU/Rx queue index set in both masks; 4261 * >= nr_bits if no further CPUs/Rx queues set in both. 4262 */ 4263 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 4264 const unsigned long *src2p, 4265 unsigned int nr_bits) 4266 { 4267 /* -1 is a legal arg here. */ 4268 if (n != -1) 4269 cpu_max_bits_warn(n, nr_bits); 4270 4271 if (src1p && src2p) 4272 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 4273 else if (src1p) 4274 return find_next_bit(src1p, nr_bits, n + 1); 4275 else if (src2p) 4276 return find_next_bit(src2p, nr_bits, n + 1); 4277 4278 return n + 1; 4279 } 4280 #else 4281 static inline int netif_set_xps_queue(struct net_device *dev, 4282 const struct cpumask *mask, 4283 u16 index) 4284 { 4285 return 0; 4286 } 4287 4288 static inline int __netif_set_xps_queue(struct net_device *dev, 4289 const unsigned long *mask, 4290 u16 index, enum xps_map_type type) 4291 { 4292 return 0; 4293 } 4294 #endif 4295 4296 /** 4297 * netif_is_multiqueue - test if device has multiple transmit queues 4298 * @dev: network device 4299 * 4300 * Check if device has multiple transmit queues 4301 */ 4302 static inline bool netif_is_multiqueue(const struct net_device *dev) 4303 { 4304 return dev->num_tx_queues > 1; 4305 } 4306 4307 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 4308 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 4309 int netif_set_real_num_queues(struct net_device *dev, 4310 unsigned int txq, unsigned int rxq); 4311 4312 int netif_get_num_default_rss_queues(void); 4313 4314 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason); 4315 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason); 4316 4317 /* 4318 * It is not allowed to call kfree_skb() or consume_skb() from hardware 4319 * interrupt context or with hardware interrupts being disabled. 4320 * (in_hardirq() || irqs_disabled()) 4321 * 4322 * We provide four helpers that can be used in following contexts : 4323 * 4324 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 4325 * replacing kfree_skb(skb) 4326 * 4327 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 4328 * Typically used in place of consume_skb(skb) in TX completion path 4329 * 4330 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 4331 * replacing kfree_skb(skb) 4332 * 4333 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 4334 * and consumed a packet. Used in place of consume_skb(skb) 4335 */ 4336 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 4337 { 4338 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 4339 } 4340 4341 static inline void dev_consume_skb_irq(struct sk_buff *skb) 4342 { 4343 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED); 4344 } 4345 4346 static inline void dev_kfree_skb_any(struct sk_buff *skb) 4347 { 4348 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 4349 } 4350 4351 static inline void dev_consume_skb_any(struct sk_buff *skb) 4352 { 4353 dev_kfree_skb_any_reason(skb, SKB_CONSUMED); 4354 } 4355 4356 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 4357 const struct bpf_prog *xdp_prog); 4358 void generic_xdp_tx(struct sk_buff *skb, const struct bpf_prog *xdp_prog); 4359 int do_xdp_generic(const struct bpf_prog *xdp_prog, struct sk_buff **pskb); 4360 int netif_rx(struct sk_buff *skb); 4361 int __netif_rx(struct sk_buff *skb); 4362 4363 int netif_receive_skb(struct sk_buff *skb); 4364 int netif_receive_skb_core(struct sk_buff *skb); 4365 void netif_receive_skb_list_internal(struct list_head *head); 4366 void netif_receive_skb_list(struct list_head *head); 4367 gro_result_t gro_receive_skb(struct gro_node *gro, struct sk_buff *skb); 4368 4369 static inline gro_result_t napi_gro_receive(struct napi_struct *napi, 4370 struct sk_buff *skb) 4371 { 4372 return gro_receive_skb(&napi->gro, skb); 4373 } 4374 4375 struct sk_buff *napi_get_frags(struct napi_struct *napi); 4376 gro_result_t napi_gro_frags(struct napi_struct *napi); 4377 4378 static inline void napi_free_frags(struct napi_struct *napi) 4379 { 4380 kfree_skb(napi->skb); 4381 napi->skb = NULL; 4382 } 4383 4384 bool netdev_is_rx_handler_busy(struct net_device *dev); 4385 int netdev_rx_handler_register(struct net_device *dev, 4386 rx_handler_func_t *rx_handler, 4387 void *rx_handler_data); 4388 void netdev_rx_handler_unregister(struct net_device *dev); 4389 4390 bool dev_valid_name(const char *name); 4391 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 4392 { 4393 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 4394 } 4395 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 4396 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 4397 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 4398 void __user *data, bool *need_copyout); 4399 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 4400 int dev_eth_ioctl(struct net_device *dev, 4401 struct ifreq *ifr, unsigned int cmd); 4402 int generic_hwtstamp_get_lower(struct net_device *dev, 4403 struct kernel_hwtstamp_config *kernel_cfg); 4404 int generic_hwtstamp_set_lower(struct net_device *dev, 4405 struct kernel_hwtstamp_config *kernel_cfg, 4406 struct netlink_ext_ack *extack); 4407 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 4408 unsigned int netif_get_flags(const struct net_device *dev); 4409 int __dev_change_flags(struct net_device *dev, unsigned int flags, 4410 struct netlink_ext_ack *extack); 4411 int netif_change_flags(struct net_device *dev, unsigned int flags, 4412 struct netlink_ext_ack *extack); 4413 int dev_change_flags(struct net_device *dev, unsigned int flags, 4414 struct netlink_ext_ack *extack); 4415 int netif_set_alias(struct net_device *dev, const char *alias, size_t len); 4416 int dev_set_alias(struct net_device *, const char *, size_t); 4417 int dev_get_alias(const struct net_device *, char *, size_t); 4418 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 4419 const char *pat, int new_ifindex, 4420 struct netlink_ext_ack *extack); 4421 int dev_change_net_namespace(struct net_device *dev, struct net *net, 4422 const char *pat); 4423 int __netif_set_mtu(struct net_device *dev, int new_mtu); 4424 int netif_set_mtu(struct net_device *dev, int new_mtu); 4425 int dev_set_mtu(struct net_device *, int); 4426 int netif_pre_changeaddr_notify(struct net_device *dev, const char *addr, 4427 struct netlink_ext_ack *extack); 4428 int netif_set_mac_address(struct net_device *dev, struct sockaddr_storage *ss, 4429 struct netlink_ext_ack *extack); 4430 int dev_set_mac_address(struct net_device *dev, struct sockaddr_storage *ss, 4431 struct netlink_ext_ack *extack); 4432 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr_storage *ss, 4433 struct netlink_ext_ack *extack); 4434 int netif_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 4435 int netif_get_port_parent_id(struct net_device *dev, 4436 struct netdev_phys_item_id *ppid, bool recurse); 4437 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 4438 4439 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 4440 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 4441 struct netdev_queue *txq, int *ret); 4442 4443 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 4444 u8 dev_xdp_prog_count(struct net_device *dev); 4445 int netif_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf); 4446 int dev_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf); 4447 u8 dev_xdp_sb_prog_count(struct net_device *dev); 4448 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 4449 4450 u32 dev_get_min_mp_channel_count(const struct net_device *dev); 4451 4452 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4453 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4454 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 4455 bool is_skb_forwardable(const struct net_device *dev, 4456 const struct sk_buff *skb); 4457 4458 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 4459 const struct sk_buff *skb, 4460 const bool check_mtu) 4461 { 4462 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 4463 unsigned int len; 4464 4465 if (!(dev->flags & IFF_UP)) 4466 return false; 4467 4468 if (!check_mtu) 4469 return true; 4470 4471 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 4472 if (skb->len <= len) 4473 return true; 4474 4475 /* if TSO is enabled, we don't care about the length as the packet 4476 * could be forwarded without being segmented before 4477 */ 4478 if (skb_is_gso(skb)) 4479 return true; 4480 4481 return false; 4482 } 4483 4484 void netdev_core_stats_inc(struct net_device *dev, u32 offset); 4485 4486 #define DEV_CORE_STATS_INC(FIELD) \ 4487 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 4488 { \ 4489 netdev_core_stats_inc(dev, \ 4490 offsetof(struct net_device_core_stats, FIELD)); \ 4491 } 4492 DEV_CORE_STATS_INC(rx_dropped) 4493 DEV_CORE_STATS_INC(tx_dropped) 4494 DEV_CORE_STATS_INC(rx_nohandler) 4495 DEV_CORE_STATS_INC(rx_otherhost_dropped) 4496 #undef DEV_CORE_STATS_INC 4497 4498 static __always_inline int ____dev_forward_skb(struct net_device *dev, 4499 struct sk_buff *skb, 4500 const bool check_mtu) 4501 { 4502 if (skb_orphan_frags(skb, GFP_ATOMIC) || 4503 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 4504 dev_core_stats_rx_dropped_inc(dev); 4505 kfree_skb(skb); 4506 return NET_RX_DROP; 4507 } 4508 4509 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 4510 skb->priority = 0; 4511 return 0; 4512 } 4513 4514 bool dev_nit_active_rcu(const struct net_device *dev); 4515 static inline bool dev_nit_active(const struct net_device *dev) 4516 { 4517 bool ret; 4518 4519 rcu_read_lock(); 4520 ret = dev_nit_active_rcu(dev); 4521 rcu_read_unlock(); 4522 return ret; 4523 } 4524 4525 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 4526 4527 static inline void __dev_put(struct net_device *dev) 4528 { 4529 if (dev) { 4530 #ifdef CONFIG_PCPU_DEV_REFCNT 4531 this_cpu_dec(*dev->pcpu_refcnt); 4532 #else 4533 refcount_dec(&dev->dev_refcnt); 4534 #endif 4535 } 4536 } 4537 4538 static inline void __dev_hold(struct net_device *dev) 4539 { 4540 if (dev) { 4541 #ifdef CONFIG_PCPU_DEV_REFCNT 4542 this_cpu_inc(*dev->pcpu_refcnt); 4543 #else 4544 refcount_inc(&dev->dev_refcnt); 4545 #endif 4546 } 4547 } 4548 4549 static inline void __netdev_tracker_alloc(struct net_device *dev, 4550 netdevice_tracker *tracker, 4551 gfp_t gfp) 4552 { 4553 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4554 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 4555 #endif 4556 } 4557 4558 /* netdev_tracker_alloc() can upgrade a prior untracked reference 4559 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4560 */ 4561 static inline void netdev_tracker_alloc(struct net_device *dev, 4562 netdevice_tracker *tracker, gfp_t gfp) 4563 { 4564 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4565 refcount_dec(&dev->refcnt_tracker.no_tracker); 4566 __netdev_tracker_alloc(dev, tracker, gfp); 4567 #endif 4568 } 4569 4570 static inline void netdev_tracker_free(struct net_device *dev, 4571 netdevice_tracker *tracker) 4572 { 4573 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4574 ref_tracker_free(&dev->refcnt_tracker, tracker); 4575 #endif 4576 } 4577 4578 static inline void netdev_hold(struct net_device *dev, 4579 netdevice_tracker *tracker, gfp_t gfp) 4580 { 4581 if (dev) { 4582 __dev_hold(dev); 4583 __netdev_tracker_alloc(dev, tracker, gfp); 4584 } 4585 } 4586 4587 static inline void netdev_put(struct net_device *dev, 4588 netdevice_tracker *tracker) 4589 { 4590 if (dev) { 4591 netdev_tracker_free(dev, tracker); 4592 __dev_put(dev); 4593 } 4594 } 4595 4596 /** 4597 * dev_hold - get reference to device 4598 * @dev: network device 4599 * 4600 * Hold reference to device to keep it from being freed. 4601 * Try using netdev_hold() instead. 4602 */ 4603 static inline void dev_hold(struct net_device *dev) 4604 { 4605 netdev_hold(dev, NULL, GFP_ATOMIC); 4606 } 4607 4608 /** 4609 * dev_put - release reference to device 4610 * @dev: network device 4611 * 4612 * Release reference to device to allow it to be freed. 4613 * Try using netdev_put() instead. 4614 */ 4615 static inline void dev_put(struct net_device *dev) 4616 { 4617 netdev_put(dev, NULL); 4618 } 4619 4620 DEFINE_FREE(dev_put, struct net_device *, if (_T) dev_put(_T)) 4621 4622 static inline void netdev_ref_replace(struct net_device *odev, 4623 struct net_device *ndev, 4624 netdevice_tracker *tracker, 4625 gfp_t gfp) 4626 { 4627 if (odev) 4628 netdev_tracker_free(odev, tracker); 4629 4630 __dev_hold(ndev); 4631 __dev_put(odev); 4632 4633 if (ndev) 4634 __netdev_tracker_alloc(ndev, tracker, gfp); 4635 } 4636 4637 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4638 * and _off may be called from IRQ context, but it is caller 4639 * who is responsible for serialization of these calls. 4640 * 4641 * The name carrier is inappropriate, these functions should really be 4642 * called netif_lowerlayer_*() because they represent the state of any 4643 * kind of lower layer not just hardware media. 4644 */ 4645 void linkwatch_fire_event(struct net_device *dev); 4646 4647 /** 4648 * linkwatch_sync_dev - sync linkwatch for the given device 4649 * @dev: network device to sync linkwatch for 4650 * 4651 * Sync linkwatch for the given device, removing it from the 4652 * pending work list (if queued). 4653 */ 4654 void linkwatch_sync_dev(struct net_device *dev); 4655 void __linkwatch_sync_dev(struct net_device *dev); 4656 4657 /** 4658 * netif_carrier_ok - test if carrier present 4659 * @dev: network device 4660 * 4661 * Check if carrier is present on device 4662 */ 4663 static inline bool netif_carrier_ok(const struct net_device *dev) 4664 { 4665 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4666 } 4667 4668 unsigned long dev_trans_start(struct net_device *dev); 4669 4670 void netdev_watchdog_up(struct net_device *dev); 4671 4672 void netif_carrier_on(struct net_device *dev); 4673 void netif_carrier_off(struct net_device *dev); 4674 void netif_carrier_event(struct net_device *dev); 4675 4676 /** 4677 * netif_dormant_on - mark device as dormant. 4678 * @dev: network device 4679 * 4680 * Mark device as dormant (as per RFC2863). 4681 * 4682 * The dormant state indicates that the relevant interface is not 4683 * actually in a condition to pass packets (i.e., it is not 'up') but is 4684 * in a "pending" state, waiting for some external event. For "on- 4685 * demand" interfaces, this new state identifies the situation where the 4686 * interface is waiting for events to place it in the up state. 4687 */ 4688 static inline void netif_dormant_on(struct net_device *dev) 4689 { 4690 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4691 linkwatch_fire_event(dev); 4692 } 4693 4694 /** 4695 * netif_dormant_off - set device as not dormant. 4696 * @dev: network device 4697 * 4698 * Device is not in dormant state. 4699 */ 4700 static inline void netif_dormant_off(struct net_device *dev) 4701 { 4702 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4703 linkwatch_fire_event(dev); 4704 } 4705 4706 /** 4707 * netif_dormant - test if device is dormant 4708 * @dev: network device 4709 * 4710 * Check if device is dormant. 4711 */ 4712 static inline bool netif_dormant(const struct net_device *dev) 4713 { 4714 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4715 } 4716 4717 4718 /** 4719 * netif_testing_on - mark device as under test. 4720 * @dev: network device 4721 * 4722 * Mark device as under test (as per RFC2863). 4723 * 4724 * The testing state indicates that some test(s) must be performed on 4725 * the interface. After completion, of the test, the interface state 4726 * will change to up, dormant, or down, as appropriate. 4727 */ 4728 static inline void netif_testing_on(struct net_device *dev) 4729 { 4730 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4731 linkwatch_fire_event(dev); 4732 } 4733 4734 /** 4735 * netif_testing_off - set device as not under test. 4736 * @dev: network device 4737 * 4738 * Device is not in testing state. 4739 */ 4740 static inline void netif_testing_off(struct net_device *dev) 4741 { 4742 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4743 linkwatch_fire_event(dev); 4744 } 4745 4746 /** 4747 * netif_testing - test if device is under test 4748 * @dev: network device 4749 * 4750 * Check if device is under test 4751 */ 4752 static inline bool netif_testing(const struct net_device *dev) 4753 { 4754 return test_bit(__LINK_STATE_TESTING, &dev->state); 4755 } 4756 4757 4758 /** 4759 * netif_oper_up - test if device is operational 4760 * @dev: network device 4761 * 4762 * Check if carrier is operational 4763 */ 4764 static inline bool netif_oper_up(const struct net_device *dev) 4765 { 4766 unsigned int operstate = READ_ONCE(dev->operstate); 4767 4768 return operstate == IF_OPER_UP || 4769 operstate == IF_OPER_UNKNOWN /* backward compat */; 4770 } 4771 4772 /** 4773 * netif_device_present - is device available or removed 4774 * @dev: network device 4775 * 4776 * Check if device has not been removed from system. 4777 */ 4778 static inline bool netif_device_present(const struct net_device *dev) 4779 { 4780 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4781 } 4782 4783 void netif_device_detach(struct net_device *dev); 4784 4785 void netif_device_attach(struct net_device *dev); 4786 4787 /* 4788 * Network interface message level settings 4789 */ 4790 4791 enum { 4792 NETIF_MSG_DRV_BIT, 4793 NETIF_MSG_PROBE_BIT, 4794 NETIF_MSG_LINK_BIT, 4795 NETIF_MSG_TIMER_BIT, 4796 NETIF_MSG_IFDOWN_BIT, 4797 NETIF_MSG_IFUP_BIT, 4798 NETIF_MSG_RX_ERR_BIT, 4799 NETIF_MSG_TX_ERR_BIT, 4800 NETIF_MSG_TX_QUEUED_BIT, 4801 NETIF_MSG_INTR_BIT, 4802 NETIF_MSG_TX_DONE_BIT, 4803 NETIF_MSG_RX_STATUS_BIT, 4804 NETIF_MSG_PKTDATA_BIT, 4805 NETIF_MSG_HW_BIT, 4806 NETIF_MSG_WOL_BIT, 4807 4808 /* When you add a new bit above, update netif_msg_class_names array 4809 * in net/ethtool/common.c 4810 */ 4811 NETIF_MSG_CLASS_COUNT, 4812 }; 4813 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4814 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4815 4816 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4817 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4818 4819 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4820 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4821 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4822 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4823 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4824 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4825 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4826 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4827 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4828 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4829 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4830 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4831 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4832 #define NETIF_MSG_HW __NETIF_MSG(HW) 4833 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4834 4835 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4836 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4837 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4838 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4839 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4840 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4841 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4842 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4843 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4844 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4845 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4846 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4847 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4848 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4849 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4850 4851 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4852 { 4853 /* use default */ 4854 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4855 return default_msg_enable_bits; 4856 if (debug_value == 0) /* no output */ 4857 return 0; 4858 /* set low N bits */ 4859 return (1U << debug_value) - 1; 4860 } 4861 4862 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4863 { 4864 spin_lock(&txq->_xmit_lock); 4865 /* Pairs with READ_ONCE() in netif_tx_owned() */ 4866 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4867 } 4868 4869 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4870 { 4871 __acquire(&txq->_xmit_lock); 4872 return true; 4873 } 4874 4875 static inline void __netif_tx_release(struct netdev_queue *txq) 4876 { 4877 __release(&txq->_xmit_lock); 4878 } 4879 4880 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4881 { 4882 spin_lock_bh(&txq->_xmit_lock); 4883 /* Pairs with READ_ONCE() in netif_tx_owned() */ 4884 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4885 } 4886 4887 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4888 { 4889 bool ok = spin_trylock(&txq->_xmit_lock); 4890 4891 if (likely(ok)) { 4892 /* Pairs with READ_ONCE() in netif_tx_owned() */ 4893 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4894 } 4895 return ok; 4896 } 4897 4898 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4899 { 4900 /* Pairs with READ_ONCE() in netif_tx_owned() */ 4901 WRITE_ONCE(txq->xmit_lock_owner, -1); 4902 spin_unlock(&txq->_xmit_lock); 4903 } 4904 4905 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4906 { 4907 /* Pairs with READ_ONCE() in netif_tx_owned() */ 4908 WRITE_ONCE(txq->xmit_lock_owner, -1); 4909 spin_unlock_bh(&txq->_xmit_lock); 4910 } 4911 4912 /* 4913 * txq->trans_start can be read locklessly from dev_watchdog() 4914 */ 4915 static inline void txq_trans_update(const struct net_device *dev, 4916 struct netdev_queue *txq) 4917 { 4918 if (!dev->lltx) 4919 WRITE_ONCE(txq->trans_start, jiffies); 4920 } 4921 4922 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4923 { 4924 unsigned long now = jiffies; 4925 4926 if (READ_ONCE(txq->trans_start) != now) 4927 WRITE_ONCE(txq->trans_start, now); 4928 } 4929 4930 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4931 static inline void netif_trans_update(struct net_device *dev) 4932 { 4933 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4934 4935 txq_trans_cond_update(txq); 4936 } 4937 4938 /** 4939 * netif_tx_lock - grab network device transmit lock 4940 * @dev: network device 4941 * 4942 * Get network device transmit lock 4943 */ 4944 void netif_tx_lock(struct net_device *dev); 4945 4946 static inline void netif_tx_lock_bh(struct net_device *dev) 4947 { 4948 local_bh_disable(); 4949 netif_tx_lock(dev); 4950 } 4951 4952 void netif_tx_unlock(struct net_device *dev); 4953 4954 static inline void netif_tx_unlock_bh(struct net_device *dev) 4955 { 4956 netif_tx_unlock(dev); 4957 local_bh_enable(); 4958 } 4959 4960 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4961 if (!(dev)->lltx) { \ 4962 __netif_tx_lock(txq, cpu); \ 4963 } else { \ 4964 __netif_tx_acquire(txq); \ 4965 } \ 4966 } 4967 4968 #define HARD_TX_TRYLOCK(dev, txq) \ 4969 (!(dev)->lltx ? \ 4970 __netif_tx_trylock(txq) : \ 4971 __netif_tx_acquire(txq)) 4972 4973 #define HARD_TX_UNLOCK(dev, txq) { \ 4974 if (!(dev)->lltx) { \ 4975 __netif_tx_unlock(txq); \ 4976 } else { \ 4977 __netif_tx_release(txq); \ 4978 } \ 4979 } 4980 4981 static inline void netif_tx_disable(struct net_device *dev) 4982 { 4983 unsigned int i; 4984 int cpu; 4985 4986 local_bh_disable(); 4987 cpu = smp_processor_id(); 4988 spin_lock(&dev->tx_global_lock); 4989 for (i = 0; i < dev->num_tx_queues; i++) { 4990 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4991 4992 __netif_tx_lock(txq, cpu); 4993 netif_tx_stop_queue(txq); 4994 __netif_tx_unlock(txq); 4995 } 4996 spin_unlock(&dev->tx_global_lock); 4997 local_bh_enable(); 4998 } 4999 5000 #ifndef CONFIG_PREEMPT_RT 5001 static inline bool netif_tx_owned(struct netdev_queue *txq, unsigned int cpu) 5002 { 5003 /* Other cpus might concurrently change txq->xmit_lock_owner 5004 * to -1 or to their cpu id, but not to our id. 5005 */ 5006 return READ_ONCE(txq->xmit_lock_owner) == cpu; 5007 } 5008 5009 #else 5010 static inline bool netif_tx_owned(struct netdev_queue *txq, unsigned int cpu) 5011 { 5012 return rt_mutex_owner(&txq->_xmit_lock.lock) == current; 5013 } 5014 5015 #endif 5016 5017 static inline void netif_addr_lock(struct net_device *dev) 5018 { 5019 unsigned char nest_level = 0; 5020 5021 #ifdef CONFIG_LOCKDEP 5022 nest_level = dev->nested_level; 5023 #endif 5024 spin_lock_nested(&dev->addr_list_lock, nest_level); 5025 } 5026 5027 static inline void netif_addr_lock_bh(struct net_device *dev) 5028 { 5029 unsigned char nest_level = 0; 5030 5031 #ifdef CONFIG_LOCKDEP 5032 nest_level = dev->nested_level; 5033 #endif 5034 local_bh_disable(); 5035 spin_lock_nested(&dev->addr_list_lock, nest_level); 5036 } 5037 5038 static inline void netif_addr_unlock(struct net_device *dev) 5039 { 5040 spin_unlock(&dev->addr_list_lock); 5041 } 5042 5043 static inline void netif_addr_unlock_bh(struct net_device *dev) 5044 { 5045 spin_unlock_bh(&dev->addr_list_lock); 5046 } 5047 5048 /* 5049 * dev_addrs walker. Should be used only for read access. Call with 5050 * rcu_read_lock held. 5051 */ 5052 #define for_each_dev_addr(dev, ha) \ 5053 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 5054 5055 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 5056 5057 void ether_setup(struct net_device *dev); 5058 5059 /* Allocate dummy net_device */ 5060 struct net_device *alloc_netdev_dummy(int sizeof_priv); 5061 5062 /* Support for loadable net-drivers */ 5063 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 5064 unsigned char name_assign_type, 5065 void (*setup)(struct net_device *), 5066 unsigned int txqs, unsigned int rxqs); 5067 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 5068 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 5069 5070 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 5071 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 5072 count) 5073 5074 int register_netdev(struct net_device *dev); 5075 void unregister_netdev(struct net_device *dev); 5076 5077 int devm_register_netdev(struct device *dev, struct net_device *ndev); 5078 5079 /* General hardware address lists handling functions */ 5080 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 5081 struct netdev_hw_addr_list *from_list, int addr_len); 5082 int __hw_addr_sync_multiple(struct netdev_hw_addr_list *to_list, 5083 struct netdev_hw_addr_list *from_list, 5084 int addr_len); 5085 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 5086 struct netdev_hw_addr_list *from_list, int addr_len); 5087 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 5088 struct net_device *dev, 5089 int (*sync)(struct net_device *, const unsigned char *), 5090 int (*unsync)(struct net_device *, 5091 const unsigned char *)); 5092 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 5093 struct net_device *dev, 5094 int (*sync)(struct net_device *, 5095 const unsigned char *, int), 5096 int (*unsync)(struct net_device *, 5097 const unsigned char *, int)); 5098 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 5099 struct net_device *dev, 5100 int (*unsync)(struct net_device *, 5101 const unsigned char *, int)); 5102 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 5103 struct net_device *dev, 5104 int (*unsync)(struct net_device *, 5105 const unsigned char *)); 5106 void __hw_addr_init(struct netdev_hw_addr_list *list); 5107 void __hw_addr_flush(struct netdev_hw_addr_list *list); 5108 int __hw_addr_list_snapshot(struct netdev_hw_addr_list *snap, 5109 const struct netdev_hw_addr_list *list, 5110 int addr_len, struct netdev_hw_addr_list *cache); 5111 void __hw_addr_list_reconcile(struct netdev_hw_addr_list *real_list, 5112 struct netdev_hw_addr_list *work, 5113 struct netdev_hw_addr_list *ref, int addr_len, 5114 struct netdev_hw_addr_list *cache); 5115 5116 /* Functions used for device addresses handling */ 5117 void dev_addr_mod(struct net_device *dev, unsigned int offset, 5118 const void *addr, size_t len); 5119 5120 static inline void 5121 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 5122 { 5123 dev_addr_mod(dev, 0, addr, len); 5124 } 5125 5126 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 5127 { 5128 __dev_addr_set(dev, addr, dev->addr_len); 5129 } 5130 5131 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 5132 unsigned char addr_type); 5133 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 5134 unsigned char addr_type); 5135 5136 /* Functions used for unicast addresses handling */ 5137 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 5138 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 5139 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 5140 int dev_uc_sync(struct net_device *to, struct net_device *from); 5141 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 5142 void dev_uc_unsync(struct net_device *to, struct net_device *from); 5143 void dev_uc_flush(struct net_device *dev); 5144 void dev_uc_init(struct net_device *dev); 5145 5146 /** 5147 * __dev_uc_sync - Synchronize device's unicast list 5148 * @dev: device to sync 5149 * @sync: function to call if address should be added 5150 * @unsync: function to call if address should be removed 5151 * 5152 * Add newly added addresses to the interface, and release 5153 * addresses that have been deleted. 5154 */ 5155 static inline int __dev_uc_sync(struct net_device *dev, 5156 int (*sync)(struct net_device *, 5157 const unsigned char *), 5158 int (*unsync)(struct net_device *, 5159 const unsigned char *)) 5160 { 5161 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 5162 } 5163 5164 /** 5165 * __dev_uc_unsync - Remove synchronized addresses from device 5166 * @dev: device to sync 5167 * @unsync: function to call if address should be removed 5168 * 5169 * Remove all addresses that were added to the device by dev_uc_sync(). 5170 */ 5171 static inline void __dev_uc_unsync(struct net_device *dev, 5172 int (*unsync)(struct net_device *, 5173 const unsigned char *)) 5174 { 5175 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 5176 } 5177 5178 /* Functions used for multicast addresses handling */ 5179 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 5180 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 5181 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 5182 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 5183 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 5184 int dev_mc_sync(struct net_device *to, struct net_device *from); 5185 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 5186 void dev_mc_unsync(struct net_device *to, struct net_device *from); 5187 void dev_mc_flush(struct net_device *dev); 5188 void dev_mc_init(struct net_device *dev); 5189 int dev_mc_dump(struct sk_buff *skb, struct netlink_callback *cb); 5190 5191 /** 5192 * __dev_mc_sync - Synchronize device's multicast list 5193 * @dev: device to sync 5194 * @sync: function to call if address should be added 5195 * @unsync: function to call if address should be removed 5196 * 5197 * Add newly added addresses to the interface, and release 5198 * addresses that have been deleted. 5199 */ 5200 static inline int __dev_mc_sync(struct net_device *dev, 5201 int (*sync)(struct net_device *, 5202 const unsigned char *), 5203 int (*unsync)(struct net_device *, 5204 const unsigned char *)) 5205 { 5206 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 5207 } 5208 5209 /** 5210 * __dev_mc_unsync - Remove synchronized addresses from device 5211 * @dev: device to sync 5212 * @unsync: function to call if address should be removed 5213 * 5214 * Remove all addresses that were added to the device by dev_mc_sync(). 5215 */ 5216 static inline void __dev_mc_unsync(struct net_device *dev, 5217 int (*unsync)(struct net_device *, 5218 const unsigned char *)) 5219 { 5220 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 5221 } 5222 5223 /* Functions used for secondary unicast and multicast support */ 5224 void dev_set_rx_mode(struct net_device *dev); 5225 void netif_rx_mode_schedule_retry(struct net_device *dev); 5226 int netif_set_promiscuity(struct net_device *dev, int inc); 5227 int dev_set_promiscuity(struct net_device *dev, int inc); 5228 int netif_set_allmulti(struct net_device *dev, int inc, bool notify); 5229 int dev_set_allmulti(struct net_device *dev, int inc); 5230 void netif_state_change(struct net_device *dev); 5231 void netdev_state_change(struct net_device *dev); 5232 void __netdev_notify_peers(struct net_device *dev); 5233 void netdev_notify_peers(struct net_device *dev); 5234 void netdev_features_change(struct net_device *dev); 5235 /* Load a device via the kmod */ 5236 void dev_load(struct net *net, const char *name); 5237 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 5238 struct rtnl_link_stats64 *storage); 5239 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 5240 const struct net_device_stats *netdev_stats); 5241 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 5242 const struct pcpu_sw_netstats __percpu *netstats); 5243 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 5244 5245 void netdev_work_sched(struct net_device *dev, unsigned long events); 5246 unsigned long netdev_work_cancel(struct net_device *dev, unsigned long mask); 5247 5248 enum { 5249 NESTED_SYNC_IMM_BIT, 5250 NESTED_SYNC_TODO_BIT, 5251 }; 5252 5253 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 5254 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 5255 5256 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 5257 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 5258 5259 struct netdev_nested_priv { 5260 unsigned char flags; 5261 void *data; 5262 }; 5263 5264 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 5265 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 5266 struct list_head **iter); 5267 5268 /* iterate through upper list, must be called under RCU read lock */ 5269 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 5270 for (iter = &(dev)->adj_list.upper, \ 5271 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 5272 updev; \ 5273 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 5274 5275 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 5276 int (*fn)(struct net_device *upper_dev, 5277 struct netdev_nested_priv *priv), 5278 struct netdev_nested_priv *priv); 5279 5280 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 5281 struct net_device *upper_dev); 5282 5283 bool netdev_has_any_upper_dev(struct net_device *dev); 5284 5285 void *netdev_lower_get_next_private(struct net_device *dev, 5286 struct list_head **iter); 5287 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 5288 struct list_head **iter); 5289 5290 #define netdev_for_each_lower_private(dev, priv, iter) \ 5291 for (iter = (dev)->adj_list.lower.next, \ 5292 priv = netdev_lower_get_next_private(dev, &(iter)); \ 5293 priv; \ 5294 priv = netdev_lower_get_next_private(dev, &(iter))) 5295 5296 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 5297 for (iter = &(dev)->adj_list.lower, \ 5298 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 5299 priv; \ 5300 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 5301 5302 void *netdev_lower_get_next(struct net_device *dev, 5303 struct list_head **iter); 5304 5305 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 5306 for (iter = (dev)->adj_list.lower.next, \ 5307 ldev = netdev_lower_get_next(dev, &(iter)); \ 5308 ldev; \ 5309 ldev = netdev_lower_get_next(dev, &(iter))) 5310 5311 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 5312 struct list_head **iter); 5313 int netdev_walk_all_lower_dev(struct net_device *dev, 5314 int (*fn)(struct net_device *lower_dev, 5315 struct netdev_nested_priv *priv), 5316 struct netdev_nested_priv *priv); 5317 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 5318 int (*fn)(struct net_device *lower_dev, 5319 struct netdev_nested_priv *priv), 5320 struct netdev_nested_priv *priv); 5321 5322 void *netdev_adjacent_get_private(struct list_head *adj_list); 5323 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 5324 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 5325 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 5326 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 5327 struct netlink_ext_ack *extack); 5328 int netdev_master_upper_dev_link(struct net_device *dev, 5329 struct net_device *upper_dev, 5330 void *upper_priv, void *upper_info, 5331 struct netlink_ext_ack *extack); 5332 void netdev_upper_dev_unlink(struct net_device *dev, 5333 struct net_device *upper_dev); 5334 int netdev_adjacent_change_prepare(struct net_device *old_dev, 5335 struct net_device *new_dev, 5336 struct net_device *dev, 5337 struct netlink_ext_ack *extack); 5338 void netdev_adjacent_change_commit(struct net_device *old_dev, 5339 struct net_device *new_dev, 5340 struct net_device *dev); 5341 void netdev_adjacent_change_abort(struct net_device *old_dev, 5342 struct net_device *new_dev, 5343 struct net_device *dev); 5344 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 5345 void *netdev_lower_dev_get_private(struct net_device *dev, 5346 struct net_device *lower_dev); 5347 void netdev_lower_state_changed(struct net_device *lower_dev, 5348 void *lower_state_info); 5349 5350 #define NETDEV_RSS_KEY_LEN 256 5351 void netdev_rss_key_fill(void *buffer, size_t len); 5352 5353 int skb_checksum_help(struct sk_buff *skb); 5354 int skb_crc32c_csum_help(struct sk_buff *skb); 5355 int skb_csum_hwoffload_help(struct sk_buff *skb, 5356 const netdev_features_t features); 5357 5358 struct netdev_bonding_info { 5359 ifslave slave; 5360 ifbond master; 5361 }; 5362 5363 struct netdev_notifier_bonding_info { 5364 struct netdev_notifier_info info; /* must be first */ 5365 struct netdev_bonding_info bonding_info; 5366 }; 5367 5368 void netdev_bonding_info_change(struct net_device *dev, 5369 struct netdev_bonding_info *bonding_info); 5370 5371 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 5372 void ethtool_notify(struct net_device *dev, unsigned int cmd); 5373 #else 5374 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd) 5375 { 5376 } 5377 #endif 5378 5379 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 5380 5381 static inline bool can_checksum_protocol(netdev_features_t features, 5382 __be16 protocol) 5383 { 5384 if (protocol == htons(ETH_P_FCOE)) 5385 return !!(features & NETIF_F_FCOE_CRC); 5386 5387 /* Assume this is an IP checksum (not SCTP CRC) */ 5388 5389 if (features & NETIF_F_HW_CSUM) { 5390 /* Can checksum everything */ 5391 return true; 5392 } 5393 5394 switch (protocol) { 5395 case htons(ETH_P_IP): 5396 return !!(features & NETIF_F_IP_CSUM); 5397 case htons(ETH_P_IPV6): 5398 return !!(features & NETIF_F_IPV6_CSUM); 5399 default: 5400 return false; 5401 } 5402 } 5403 5404 #ifdef CONFIG_BUG 5405 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 5406 #else 5407 static inline void netdev_rx_csum_fault(struct net_device *dev, 5408 struct sk_buff *skb) 5409 { 5410 } 5411 #endif 5412 /* rx skb timestamps */ 5413 void net_enable_timestamp(void); 5414 void net_disable_timestamp(void); 5415 5416 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 5417 const struct skb_shared_hwtstamps *hwtstamps, 5418 bool cycles) 5419 { 5420 const struct net_device_ops *ops = dev->netdev_ops; 5421 5422 if (ops->ndo_get_tstamp) 5423 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 5424 5425 return hwtstamps->hwtstamp; 5426 } 5427 5428 #ifndef CONFIG_PREEMPT_RT 5429 static inline void netdev_xmit_set_more(bool more) 5430 { 5431 __this_cpu_write(softnet_data.xmit.more, more); 5432 } 5433 5434 static inline bool netdev_xmit_more(void) 5435 { 5436 return __this_cpu_read(softnet_data.xmit.more); 5437 } 5438 #else 5439 static inline void netdev_xmit_set_more(bool more) 5440 { 5441 current->net_xmit.more = more; 5442 } 5443 5444 static inline bool netdev_xmit_more(void) 5445 { 5446 return current->net_xmit.more; 5447 } 5448 #endif 5449 5450 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 5451 struct sk_buff *skb, struct net_device *dev, 5452 bool more) 5453 { 5454 netdev_xmit_set_more(more); 5455 return ops->ndo_start_xmit(skb, dev); 5456 } 5457 5458 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 5459 struct netdev_queue *txq, bool more) 5460 { 5461 const struct net_device_ops *ops = dev->netdev_ops; 5462 netdev_tx_t rc; 5463 5464 rc = __netdev_start_xmit(ops, skb, dev, more); 5465 if (rc == NETDEV_TX_OK) 5466 txq_trans_update(dev, txq); 5467 5468 return rc; 5469 } 5470 5471 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 5472 const struct ns_common *ns); 5473 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 5474 const struct ns_common *ns); 5475 5476 extern const struct kobj_ns_type_operations net_ns_type_operations; 5477 5478 const char *netdev_drivername(const struct net_device *dev); 5479 5480 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 5481 netdev_features_t f2) 5482 { 5483 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 5484 if (f1 & NETIF_F_HW_CSUM) 5485 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5486 else 5487 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5488 } 5489 5490 return f1 & f2; 5491 } 5492 5493 static inline netdev_features_t netdev_get_wanted_features( 5494 struct net_device *dev) 5495 { 5496 return (dev->features & ~dev->hw_features) | dev->wanted_features; 5497 } 5498 netdev_features_t netdev_increment_features(netdev_features_t all, 5499 netdev_features_t one, netdev_features_t mask); 5500 5501 /* Allow TSO being used on stacked device : 5502 * Performing the GSO segmentation before last device 5503 * is a performance improvement. 5504 */ 5505 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 5506 netdev_features_t mask) 5507 { 5508 return netdev_increment_features(features, NETIF_F_ALL_TSO | 5509 NETIF_F_ALL_FOR_ALL, mask); 5510 } 5511 5512 int __netdev_update_features(struct net_device *dev); 5513 void netdev_update_features(struct net_device *dev); 5514 void netdev_change_features(struct net_device *dev); 5515 void netdev_compute_master_upper_features(struct net_device *dev, bool update_header); 5516 5517 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 5518 struct net_device *dev); 5519 5520 netdev_features_t passthru_features_check(struct sk_buff *skb, 5521 struct net_device *dev, 5522 netdev_features_t features); 5523 netdev_features_t __netif_skb_features(struct sk_buff *skb, 5524 bool check_gso_limits); 5525 5526 static inline netdev_features_t netif_skb_features(struct sk_buff *skb) 5527 { 5528 return __netif_skb_features(skb, true); 5529 } 5530 5531 void skb_warn_bad_offload(const struct sk_buff *skb); 5532 5533 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 5534 { 5535 netdev_features_t feature; 5536 5537 if (gso_type & (SKB_GSO_TCP_FIXEDID | SKB_GSO_TCP_FIXEDID_INNER)) 5538 gso_type |= __SKB_GSO_TCP_FIXEDID; 5539 5540 feature = ((netdev_features_t)gso_type << NETIF_F_GSO_SHIFT) & NETIF_F_GSO_MASK; 5541 5542 /* check flags correspondence */ 5543 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 5544 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 5545 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 5546 BUILD_BUG_ON(__SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 5547 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 5548 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 5549 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 5550 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 5551 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 5552 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 5553 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 5554 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 5555 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 5556 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 5557 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 5558 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 5559 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 5560 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 5561 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 5562 BUILD_BUG_ON(SKB_GSO_TCP_ACCECN != 5563 (NETIF_F_GSO_ACCECN >> NETIF_F_GSO_SHIFT)); 5564 5565 return (features & feature) == feature; 5566 } 5567 5568 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 5569 { 5570 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 5571 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 5572 } 5573 5574 static inline bool netif_needs_gso(struct sk_buff *skb, 5575 netdev_features_t features) 5576 { 5577 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 5578 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 5579 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 5580 } 5581 5582 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 5583 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 5584 void netif_inherit_tso_max(struct net_device *to, 5585 const struct net_device *from); 5586 5587 static inline unsigned int 5588 netif_get_gro_max_size(const struct net_device *dev, const struct sk_buff *skb) 5589 { 5590 /* pairs with WRITE_ONCE() in netif_set_gro(_ipv4)_max_size() */ 5591 return skb->protocol == htons(ETH_P_IPV6) ? 5592 READ_ONCE(dev->gro_max_size) : 5593 READ_ONCE(dev->gro_ipv4_max_size); 5594 } 5595 5596 static inline bool netif_is_macsec(const struct net_device *dev) 5597 { 5598 return dev->priv_flags & IFF_MACSEC; 5599 } 5600 5601 static inline bool netif_is_macvlan(const struct net_device *dev) 5602 { 5603 return dev->priv_flags & IFF_MACVLAN; 5604 } 5605 5606 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5607 { 5608 return dev->priv_flags & IFF_MACVLAN_PORT; 5609 } 5610 5611 static inline bool netif_is_bond_master(const struct net_device *dev) 5612 { 5613 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5614 } 5615 5616 static inline bool netif_is_bond_slave(const struct net_device *dev) 5617 { 5618 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5619 } 5620 5621 static inline bool netif_supports_nofcs(struct net_device *dev) 5622 { 5623 return dev->priv_flags & IFF_SUPP_NOFCS; 5624 } 5625 5626 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5627 { 5628 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5629 } 5630 5631 static inline bool netif_is_l3_master(const struct net_device *dev) 5632 { 5633 return IS_ENABLED(CONFIG_NET_VRF) && (dev->priv_flags & IFF_L3MDEV_MASTER); 5634 } 5635 5636 static inline bool netif_is_l3_slave(const struct net_device *dev) 5637 { 5638 return IS_ENABLED(CONFIG_NET_VRF) && (dev->priv_flags & IFF_L3MDEV_SLAVE); 5639 } 5640 5641 static inline int dev_sdif(const struct net_device *dev) 5642 { 5643 #ifdef CONFIG_NET_L3_MASTER_DEV 5644 if (netif_is_l3_slave(dev)) 5645 return dev->ifindex; 5646 #endif 5647 return 0; 5648 } 5649 5650 static inline bool netif_is_bridge_master(const struct net_device *dev) 5651 { 5652 return dev->priv_flags & IFF_EBRIDGE; 5653 } 5654 5655 static inline bool netif_is_bridge_port(const struct net_device *dev) 5656 { 5657 return dev->priv_flags & IFF_BRIDGE_PORT; 5658 } 5659 5660 static inline bool netif_is_ovs_master(const struct net_device *dev) 5661 { 5662 return dev->priv_flags & IFF_OPENVSWITCH; 5663 } 5664 5665 static inline bool netif_is_ovs_port(const struct net_device *dev) 5666 { 5667 return dev->priv_flags & IFF_OVS_DATAPATH; 5668 } 5669 5670 static inline bool netif_is_any_bridge_master(const struct net_device *dev) 5671 { 5672 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev); 5673 } 5674 5675 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5676 { 5677 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5678 } 5679 5680 static inline bool netif_is_team_master(const struct net_device *dev) 5681 { 5682 return dev->priv_flags & IFF_TEAM; 5683 } 5684 5685 static inline bool netif_is_team_port(const struct net_device *dev) 5686 { 5687 return dev->priv_flags & IFF_TEAM_PORT; 5688 } 5689 5690 static inline bool netif_is_lag_master(const struct net_device *dev) 5691 { 5692 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5693 } 5694 5695 static inline bool netif_is_lag_port(const struct net_device *dev) 5696 { 5697 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5698 } 5699 5700 bool netif_is_rxfh_configured(const struct net_device *dev); 5701 5702 static inline bool netif_is_failover(const struct net_device *dev) 5703 { 5704 return dev->priv_flags & IFF_FAILOVER; 5705 } 5706 5707 static inline bool netif_is_failover_slave(const struct net_device *dev) 5708 { 5709 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5710 } 5711 5712 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5713 static inline void netif_keep_dst(struct net_device *dev) 5714 { 5715 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5716 } 5717 5718 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5719 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5720 { 5721 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5722 return netif_is_macsec(dev); 5723 } 5724 5725 extern struct pernet_operations __net_initdata loopback_net_ops; 5726 5727 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5728 5729 /* netdev_printk helpers, similar to dev_printk */ 5730 5731 static inline const char *netdev_name(const struct net_device *dev) 5732 { 5733 if (!dev->name[0] || strchr(dev->name, '%')) 5734 return "(unnamed net_device)"; 5735 return dev->name; 5736 } 5737 5738 /* Caller holds RTNL, netdev->lock or RCU */ 5739 static inline bool netif_is_alive(const struct net_device *dev) 5740 { 5741 return READ_ONCE(dev->reg_state) <= NETREG_REGISTERED; 5742 } 5743 5744 static inline const char *netdev_reg_state(const struct net_device *dev) 5745 { 5746 u8 reg_state = READ_ONCE(dev->reg_state); 5747 5748 switch (reg_state) { 5749 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5750 case NETREG_REGISTERED: return ""; 5751 case NETREG_UNREGISTERING: return " (unregistering)"; 5752 case NETREG_UNREGISTERED: return " (unregistered)"; 5753 case NETREG_RELEASED: return " (released)"; 5754 case NETREG_DUMMY: return " (dummy)"; 5755 } 5756 5757 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state); 5758 return " (unknown)"; 5759 } 5760 5761 #define MODULE_ALIAS_NETDEV(device) \ 5762 MODULE_ALIAS("netdev-" device) 5763 5764 /* 5765 * netdev_WARN() acts like dev_printk(), but with the key difference 5766 * of using a WARN/WARN_ON to get the message out, including the 5767 * file/line information and a backtrace. 5768 */ 5769 #define netdev_WARN(dev, format, args...) \ 5770 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5771 netdev_reg_state(dev), ##args) 5772 5773 #define netdev_WARN_ONCE(dev, format, args...) \ 5774 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5775 netdev_reg_state(dev), ##args) 5776 5777 /* 5778 * The list of packet types we will receive (as opposed to discard) 5779 * and the routines to invoke. 5780 * 5781 * Why 16. Because with 16 the only overlap we get on a hash of the 5782 * low nibble of the protocol value is RARP/SNAP/X.25. 5783 * 5784 * 0800 IP 5785 * 0001 802.3 5786 * 0002 AX.25 5787 * 0004 802.2 5788 * 8035 RARP 5789 * 0005 SNAP 5790 * 0805 X.25 5791 * 0806 ARP 5792 * 8137 IPX 5793 * 0009 Localtalk 5794 * 86DD IPv6 5795 */ 5796 #define PTYPE_HASH_SIZE (16) 5797 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5798 5799 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5800 5801 extern struct net_device *blackhole_netdev; 5802 5803 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */ 5804 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD) 5805 #define DEV_STATS_ADD(DEV, FIELD, VAL) \ 5806 atomic_long_add((VAL), &(DEV)->stats.__##FIELD) 5807 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD) 5808 5809 #endif /* _LINUX_NETDEVICE_H */ 5810