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 int count; 256 257 /* Auxiliary tree for faster lookup on addition and deletion */ 258 struct rb_root tree; 259 }; 260 261 #define netdev_hw_addr_list_count(l) ((l)->count) 262 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 263 #define netdev_hw_addr_list_for_each(ha, l) \ 264 list_for_each_entry(ha, &(l)->list, list) 265 266 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 267 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 268 #define netdev_for_each_uc_addr(ha, dev) \ 269 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 270 #define netdev_for_each_synced_uc_addr(_ha, _dev) \ 271 netdev_for_each_uc_addr((_ha), (_dev)) \ 272 if ((_ha)->sync_cnt) 273 274 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 275 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 276 #define netdev_for_each_mc_addr(ha, dev) \ 277 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 278 #define netdev_for_each_synced_mc_addr(_ha, _dev) \ 279 netdev_for_each_mc_addr((_ha), (_dev)) \ 280 if ((_ha)->sync_cnt) 281 282 struct hh_cache { 283 unsigned int hh_len; 284 seqlock_t hh_lock; 285 286 /* cached hardware header; allow for machine alignment needs. */ 287 #define HH_DATA_MOD 16 288 #define HH_DATA_OFF(__len) \ 289 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 290 #define HH_DATA_ALIGN(__len) \ 291 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 292 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 293 }; 294 295 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 296 * Alternative is: 297 * dev->hard_header_len ? (dev->hard_header_len + 298 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 299 * 300 * We could use other alignment values, but we must maintain the 301 * relationship HH alignment <= LL alignment. 302 */ 303 #define LL_RESERVED_SPACE_EX(dev, hlen) \ 304 ((((hlen) + READ_ONCE((dev)->needed_headroom)) \ 305 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 306 #define LL_RESERVED_SPACE(dev) \ 307 LL_RESERVED_SPACE_EX(dev, (dev)->hard_header_len) 308 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 309 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \ 310 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 311 312 struct header_ops { 313 int (*create) (struct sk_buff *skb, struct net_device *dev, 314 unsigned short type, const void *daddr, 315 const void *saddr, unsigned int len); 316 int (*parse)(const struct sk_buff *skb, 317 const struct net_device *dev, 318 unsigned char *haddr); 319 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 320 void (*cache_update)(struct hh_cache *hh, 321 const struct net_device *dev, 322 const unsigned char *haddr); 323 bool (*validate)(const char *ll_header, unsigned int len); 324 __be16 (*parse_protocol)(const struct sk_buff *skb); 325 }; 326 327 /* These flag bits are private to the generic network queueing 328 * layer; they may not be explicitly referenced by any other 329 * code. 330 */ 331 332 enum netdev_state_t { 333 __LINK_STATE_START, 334 __LINK_STATE_PRESENT, 335 __LINK_STATE_NOCARRIER, 336 __LINK_STATE_LINKWATCH_PENDING, 337 __LINK_STATE_DORMANT, 338 __LINK_STATE_TESTING, 339 }; 340 341 struct gro_list { 342 struct list_head list; 343 int count; 344 }; 345 346 /* 347 * size of gro hash buckets, must be <= the number of bits in 348 * gro_node::bitmask 349 */ 350 #define GRO_HASH_BUCKETS 8 351 352 /** 353 * struct gro_node - structure to support Generic Receive Offload 354 * @bitmask: bitmask to indicate used buckets in @hash 355 * @hash: hashtable of pending aggregated skbs, separated by flows 356 * @rx_list: list of pending ``GRO_NORMAL`` skbs 357 * @rx_count: cached current length of @rx_list 358 * @cached_napi_id: napi_struct::napi_id cached for hotpath, 0 for standalone 359 */ 360 struct gro_node { 361 unsigned long bitmask; 362 struct gro_list hash[GRO_HASH_BUCKETS]; 363 struct list_head rx_list; 364 u32 rx_count; 365 u32 cached_napi_id; 366 }; 367 368 /* 369 * Structure for per-NAPI config 370 */ 371 struct napi_config { 372 u64 gro_flush_timeout; 373 u64 irq_suspend_timeout; 374 u32 defer_hard_irqs; 375 cpumask_t affinity_mask; 376 u8 threaded; 377 unsigned int napi_id; 378 }; 379 380 /* 381 * Structure for NAPI scheduling similar to tasklet but with weighting 382 */ 383 struct napi_struct { 384 /* This field should be first or softnet_data.backlog needs tweaks. */ 385 unsigned long state; 386 /* The poll_list must only be managed by the entity which 387 * changes the state of the NAPI_STATE_SCHED bit. This means 388 * whoever atomically sets that bit can add this napi_struct 389 * to the per-CPU poll_list, and whoever clears that bit 390 * can remove from the list right before clearing the bit. 391 */ 392 struct list_head poll_list; 393 394 int weight; 395 u32 defer_hard_irqs_count; 396 int (*poll)(struct napi_struct *, int); 397 #ifdef CONFIG_NETPOLL 398 /* CPU actively polling if netpoll is configured */ 399 int poll_owner; 400 #endif 401 /* CPU on which NAPI has been scheduled for processing */ 402 int list_owner; 403 struct net_device *dev; 404 struct sk_buff *skb; 405 struct gro_node gro; 406 struct hrtimer timer; 407 /* all fields past this point are write-protected by netdev_lock */ 408 struct task_struct *thread; 409 unsigned long gro_flush_timeout; 410 unsigned long irq_suspend_timeout; 411 u32 defer_hard_irqs; 412 /* control-path-only fields follow */ 413 u32 napi_id; 414 struct list_head dev_list; 415 struct hlist_node napi_hash_node; 416 int irq; 417 struct irq_affinity_notify notify; 418 int napi_rmap_idx; 419 int index; 420 struct napi_config *config; 421 }; 422 423 enum { 424 NAPI_STATE_SCHED, /* Poll is scheduled */ 425 NAPI_STATE_MISSED, /* reschedule a napi */ 426 NAPI_STATE_DISABLE, /* Disable pending */ 427 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 428 NAPI_STATE_LISTED, /* NAPI added to system lists */ 429 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 430 NAPI_STATE_IN_BUSY_POLL, /* Do not rearm NAPI interrupt */ 431 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 432 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 433 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 434 NAPI_STATE_HAS_NOTIFIER, /* Napi has an IRQ notifier */ 435 NAPI_STATE_THREADED_BUSY_POLL, /* The threaded NAPI poller will busy poll */ 436 }; 437 438 enum { 439 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 440 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 441 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 442 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 443 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 444 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 445 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 446 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 447 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 448 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 449 NAPIF_STATE_HAS_NOTIFIER = BIT(NAPI_STATE_HAS_NOTIFIER), 450 NAPIF_STATE_THREADED_BUSY_POLL = BIT(NAPI_STATE_THREADED_BUSY_POLL), 451 }; 452 453 enum gro_result { 454 GRO_MERGED, 455 GRO_MERGED_FREE, 456 GRO_HELD, 457 GRO_NORMAL, 458 GRO_CONSUMED, 459 }; 460 typedef enum gro_result gro_result_t; 461 462 /* 463 * enum rx_handler_result - Possible return values for rx_handlers. 464 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 465 * further. 466 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 467 * case skb->dev was changed by rx_handler. 468 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 469 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 470 * 471 * rx_handlers are functions called from inside __netif_receive_skb(), to do 472 * special processing of the skb, prior to delivery to protocol handlers. 473 * 474 * Currently, a net_device can only have a single rx_handler registered. Trying 475 * to register a second rx_handler will return -EBUSY. 476 * 477 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 478 * To unregister a rx_handler on a net_device, use 479 * netdev_rx_handler_unregister(). 480 * 481 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 482 * do with the skb. 483 * 484 * If the rx_handler consumed the skb in some way, it should return 485 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 486 * the skb to be delivered in some other way. 487 * 488 * If the rx_handler changed skb->dev, to divert the skb to another 489 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 490 * new device will be called if it exists. 491 * 492 * If the rx_handler decides the skb should be ignored, it should return 493 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 494 * are registered on exact device (ptype->dev == skb->dev). 495 * 496 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 497 * delivered, it should return RX_HANDLER_PASS. 498 * 499 * A device without a registered rx_handler will behave as if rx_handler 500 * returned RX_HANDLER_PASS. 501 */ 502 503 enum rx_handler_result { 504 RX_HANDLER_CONSUMED, 505 RX_HANDLER_ANOTHER, 506 RX_HANDLER_EXACT, 507 RX_HANDLER_PASS, 508 }; 509 typedef enum rx_handler_result rx_handler_result_t; 510 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 511 512 void __napi_schedule(struct napi_struct *n); 513 void __napi_schedule_irqoff(struct napi_struct *n); 514 515 static inline bool napi_disable_pending(struct napi_struct *n) 516 { 517 return test_bit(NAPI_STATE_DISABLE, &n->state); 518 } 519 520 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 521 { 522 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 523 } 524 525 /** 526 * napi_is_scheduled - test if NAPI is scheduled 527 * @n: NAPI context 528 * 529 * This check is "best-effort". With no locking implemented, 530 * a NAPI can be scheduled or terminate right after this check 531 * and produce not precise results. 532 * 533 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled 534 * should not be used normally and napi_schedule should be 535 * used instead. 536 * 537 * Use only if the driver really needs to check if a NAPI 538 * is scheduled for example in the context of delayed timer 539 * that can be skipped if a NAPI is already scheduled. 540 * 541 * Return: True if NAPI is scheduled, False otherwise. 542 */ 543 static inline bool napi_is_scheduled(struct napi_struct *n) 544 { 545 return test_bit(NAPI_STATE_SCHED, &n->state); 546 } 547 548 bool napi_schedule_prep(struct napi_struct *n); 549 550 /** 551 * napi_schedule - schedule NAPI poll 552 * @n: NAPI context 553 * 554 * Schedule NAPI poll routine to be called if it is not already 555 * running. 556 * Return: true if we schedule a NAPI or false if not. 557 * Refer to napi_schedule_prep() for additional reason on why 558 * a NAPI might not be scheduled. 559 */ 560 static inline bool napi_schedule(struct napi_struct *n) 561 { 562 if (napi_schedule_prep(n)) { 563 __napi_schedule(n); 564 return true; 565 } 566 567 return false; 568 } 569 570 /** 571 * napi_schedule_irqoff - schedule NAPI poll 572 * @n: NAPI context 573 * 574 * Variant of napi_schedule(), assuming hard irqs are masked. 575 */ 576 static inline void napi_schedule_irqoff(struct napi_struct *n) 577 { 578 if (napi_schedule_prep(n)) 579 __napi_schedule_irqoff(n); 580 } 581 582 /** 583 * napi_complete_done - NAPI processing complete 584 * @n: NAPI context 585 * @work_done: number of packets processed 586 * 587 * Mark NAPI processing as complete. Should only be called if poll budget 588 * has not been completely consumed. 589 * Prefer over napi_complete(). 590 * Return: false if device should avoid rearming interrupts. 591 */ 592 bool napi_complete_done(struct napi_struct *n, int work_done); 593 594 static inline bool napi_complete(struct napi_struct *n) 595 { 596 return napi_complete_done(n, 0); 597 } 598 599 void netif_threaded_enable(struct net_device *dev); 600 int dev_set_threaded(struct net_device *dev, 601 enum netdev_napi_threaded threaded); 602 603 void napi_disable(struct napi_struct *n); 604 void napi_disable_locked(struct napi_struct *n); 605 606 void napi_enable(struct napi_struct *n); 607 void napi_enable_locked(struct napi_struct *n); 608 609 /** 610 * napi_synchronize - wait until NAPI is not running 611 * @n: NAPI context 612 * 613 * Wait until NAPI is done being scheduled on this context. 614 * Waits till any outstanding processing completes but 615 * does not disable future activations. 616 */ 617 static inline void napi_synchronize(const struct napi_struct *n) 618 { 619 if (IS_ENABLED(CONFIG_SMP)) 620 while (test_bit(NAPI_STATE_SCHED, &n->state)) 621 msleep(1); 622 else 623 barrier(); 624 } 625 626 /** 627 * napi_if_scheduled_mark_missed - if napi is running, set the 628 * NAPIF_STATE_MISSED 629 * @n: NAPI context 630 * 631 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 632 * NAPI is scheduled. 633 **/ 634 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 635 { 636 unsigned long val, new; 637 638 val = READ_ONCE(n->state); 639 do { 640 if (val & NAPIF_STATE_DISABLE) 641 return true; 642 643 if (!(val & NAPIF_STATE_SCHED)) 644 return false; 645 646 new = val | NAPIF_STATE_MISSED; 647 } while (!try_cmpxchg(&n->state, &val, new)); 648 649 return true; 650 } 651 652 enum netdev_queue_state_t { 653 __QUEUE_STATE_DRV_XOFF, 654 __QUEUE_STATE_STACK_XOFF, 655 __QUEUE_STATE_FROZEN, 656 }; 657 658 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 659 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 660 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 661 662 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 663 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 664 QUEUE_STATE_FROZEN) 665 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 666 QUEUE_STATE_FROZEN) 667 668 /* 669 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 670 * netif_tx_* functions below are used to manipulate this flag. The 671 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 672 * queue independently. The netif_xmit_*stopped functions below are called 673 * to check if the queue has been stopped by the driver or stack (either 674 * of the XOFF bits are set in the state). Drivers should not need to call 675 * netif_xmit*stopped functions, they should only be using netif_tx_*. 676 */ 677 678 struct netdev_queue { 679 /* 680 * read-mostly part 681 */ 682 struct net_device *dev; 683 netdevice_tracker dev_tracker; 684 685 struct Qdisc __rcu *qdisc; 686 struct Qdisc __rcu *qdisc_sleeping; 687 #ifdef CONFIG_SYSFS 688 struct kobject kobj; 689 const struct attribute_group **groups; 690 #endif 691 unsigned long tx_maxrate; 692 /* 693 * Number of TX timeouts for this queue 694 * (/sys/class/net/DEV/Q/trans_timeout) 695 */ 696 atomic_long_t trans_timeout; 697 698 /* Subordinate device that the queue has been assigned to */ 699 struct net_device *sb_dev; 700 #ifdef CONFIG_XDP_SOCKETS 701 /* "ops protected", see comment about net_device::lock */ 702 struct xsk_buff_pool *pool; 703 #endif 704 705 /* 706 * write-mostly part 707 */ 708 #ifdef CONFIG_BQL 709 struct dql dql; 710 #endif 711 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 712 int xmit_lock_owner; 713 /* 714 * Time (in jiffies) of last Tx 715 */ 716 unsigned long trans_start; 717 718 unsigned long state; 719 720 /* 721 * slow- / control-path part 722 */ 723 /* NAPI instance for the queue 724 * "ops protected", see comment about net_device::lock 725 */ 726 struct napi_struct *napi; 727 728 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 729 int numa_node; 730 #endif 731 } ____cacheline_aligned_in_smp; 732 733 extern int sysctl_fb_tunnels_only_for_init_net; 734 extern int sysctl_devconf_inherit_init_net; 735 736 /* 737 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 738 * == 1 : For initns only 739 * == 2 : For none. 740 */ 741 static inline bool net_has_fallback_tunnels(const struct net *net) 742 { 743 #if IS_ENABLED(CONFIG_SYSCTL) 744 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net); 745 746 return !fb_tunnels_only_for_init_net || 747 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1); 748 #else 749 return true; 750 #endif 751 } 752 753 static inline int net_inherit_devconf(void) 754 { 755 #if IS_ENABLED(CONFIG_SYSCTL) 756 return READ_ONCE(sysctl_devconf_inherit_init_net); 757 #else 758 return 0; 759 #endif 760 } 761 762 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 763 { 764 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 765 return q->numa_node; 766 #else 767 return NUMA_NO_NODE; 768 #endif 769 } 770 771 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 772 { 773 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 774 q->numa_node = node; 775 #endif 776 } 777 778 #ifdef CONFIG_RFS_ACCEL 779 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 780 u16 filter_id); 781 #endif 782 783 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 784 enum xps_map_type { 785 XPS_CPUS = 0, 786 XPS_RXQS, 787 XPS_MAPS_MAX, 788 }; 789 790 #ifdef CONFIG_XPS 791 /* 792 * This structure holds an XPS map which can be of variable length. The 793 * map is an array of queues. 794 */ 795 struct xps_map { 796 unsigned int len; 797 unsigned int alloc_len; 798 struct rcu_head rcu; 799 u16 queues[]; 800 }; 801 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 802 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 803 - sizeof(struct xps_map)) / sizeof(u16)) 804 805 /* 806 * This structure holds all XPS maps for device. Maps are indexed by CPU. 807 * 808 * We keep track of the number of cpus/rxqs used when the struct is allocated, 809 * in nr_ids. This will help not accessing out-of-bound memory. 810 * 811 * We keep track of the number of traffic classes used when the struct is 812 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 813 * not crossing its upper bound, as the original dev->num_tc can be updated in 814 * the meantime. 815 */ 816 struct xps_dev_maps { 817 struct rcu_head rcu; 818 unsigned int nr_ids; 819 s16 num_tc; 820 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 821 }; 822 823 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 824 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 825 826 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 827 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 828 829 #endif /* CONFIG_XPS */ 830 831 #define TC_MAX_QUEUE 16 832 #define TC_BITMASK 15 833 /* HW offloaded queuing disciplines txq count and offset maps */ 834 struct netdev_tc_txq { 835 union { 836 struct { 837 u16 count; 838 u16 offset; 839 }; 840 u32 combined; 841 }; 842 }; 843 844 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 845 /* 846 * This structure is to hold information about the device 847 * configured to run FCoE protocol stack. 848 */ 849 struct netdev_fcoe_hbainfo { 850 char manufacturer[64]; 851 char serial_number[64]; 852 char hardware_version[64]; 853 char driver_version[64]; 854 char optionrom_version[64]; 855 char firmware_version[64]; 856 char model[256]; 857 char model_description[256]; 858 }; 859 #endif 860 861 #define MAX_PHYS_ITEM_ID_LEN 32 862 863 /* This structure holds a unique identifier to identify some 864 * physical item (port for example) used by a netdevice. 865 */ 866 struct netdev_phys_item_id { 867 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 868 unsigned char id_len; 869 }; 870 871 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 872 struct netdev_phys_item_id *b) 873 { 874 return a->id_len == b->id_len && 875 memcmp(a->id, b->id, a->id_len) == 0; 876 } 877 878 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 879 struct sk_buff *skb, 880 struct net_device *sb_dev); 881 882 enum net_device_path_type { 883 DEV_PATH_ETHERNET = 0, 884 DEV_PATH_VLAN, 885 DEV_PATH_BRIDGE, 886 DEV_PATH_PPPOE, 887 DEV_PATH_DSA, 888 DEV_PATH_MTK_WDMA, 889 DEV_PATH_TUN, 890 DEV_PATH_IEEE80211, 891 }; 892 893 struct net_device_path { 894 enum net_device_path_type type; 895 const struct net_device *dev; 896 union { 897 struct { 898 u16 id; 899 __be16 proto; 900 u8 h_dest[ETH_ALEN]; 901 } encap; 902 struct { 903 struct dst_entry *dst; 904 union { 905 struct in_addr src_v4; 906 struct in6_addr src_v6; 907 }; 908 union { 909 struct in_addr dst_v4; 910 struct in6_addr dst_v6; 911 }; 912 913 u8 inner_proto; 914 } tun; 915 struct { 916 enum { 917 DEV_PATH_BR_VLAN_KEEP, 918 DEV_PATH_BR_VLAN_TAG, 919 DEV_PATH_BR_VLAN_UNTAG, 920 DEV_PATH_BR_VLAN_UNTAG_HW, 921 } vlan_mode; 922 u16 vlan_id; 923 __be16 vlan_proto; 924 } bridge; 925 struct { 926 int port; 927 u16 proto; 928 } dsa; 929 struct { 930 u8 wdma_idx; 931 u8 queue; 932 u16 wcid; 933 u8 bss; 934 u8 amsdu; 935 } mtk_wdma; 936 }; 937 }; 938 939 #define NET_DEVICE_PATH_STACK_MAX 5 940 #define NET_DEVICE_PATH_VLAN_MAX 2 941 942 struct net_device_path_stack { 943 int num_paths; 944 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 945 }; 946 947 struct net_device_path_ctx { 948 const struct net_device *dev; 949 u8 daddr[ETH_ALEN]; 950 __be16 ether_type; 951 952 int num_vlans; 953 struct { 954 u16 id; 955 __be16 proto; 956 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 957 958 bool ieee80211; 959 }; 960 961 enum tc_setup_type { 962 TC_QUERY_CAPS, 963 TC_SETUP_QDISC_MQPRIO, 964 TC_SETUP_CLSU32, 965 TC_SETUP_CLSFLOWER, 966 TC_SETUP_CLSMATCHALL, 967 TC_SETUP_CLSBPF, 968 TC_SETUP_BLOCK, 969 TC_SETUP_QDISC_CBS, 970 TC_SETUP_QDISC_RED, 971 TC_SETUP_QDISC_PRIO, 972 TC_SETUP_QDISC_MQ, 973 TC_SETUP_QDISC_ETF, 974 TC_SETUP_ROOT_QDISC, 975 TC_SETUP_QDISC_GRED, 976 TC_SETUP_QDISC_TAPRIO, 977 TC_SETUP_FT, 978 TC_SETUP_QDISC_ETS, 979 TC_SETUP_QDISC_TBF, 980 TC_SETUP_QDISC_FIFO, 981 TC_SETUP_QDISC_HTB, 982 TC_SETUP_ACT, 983 }; 984 985 /* These structures hold the attributes of bpf state that are being passed 986 * to the netdevice through the bpf op. 987 */ 988 enum bpf_netdev_command { 989 /* Set or clear a bpf program used in the earliest stages of packet 990 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 991 * is responsible for calling bpf_prog_put on any old progs that are 992 * stored. In case of error, the callee need not release the new prog 993 * reference, but on success it takes ownership and must bpf_prog_put 994 * when it is no longer used. 995 */ 996 XDP_SETUP_PROG, 997 XDP_SETUP_PROG_HW, 998 /* BPF program for offload callbacks, invoked at program load time. */ 999 BPF_OFFLOAD_MAP_ALLOC, 1000 BPF_OFFLOAD_MAP_FREE, 1001 XDP_SETUP_XSK_POOL, 1002 }; 1003 1004 struct bpf_prog_offload_ops; 1005 struct netlink_ext_ack; 1006 struct xdp_umem; 1007 struct xdp_dev_bulk_queue; 1008 struct bpf_xdp_link; 1009 1010 enum bpf_xdp_mode { 1011 XDP_MODE_SKB = 0, 1012 XDP_MODE_DRV = 1, 1013 XDP_MODE_HW = 2, 1014 __MAX_XDP_MODE 1015 }; 1016 1017 struct bpf_xdp_entity { 1018 struct bpf_prog *prog; 1019 struct bpf_xdp_link *link; 1020 }; 1021 1022 struct netdev_bpf { 1023 enum bpf_netdev_command command; 1024 union { 1025 /* XDP_SETUP_PROG */ 1026 struct { 1027 u32 flags; 1028 struct bpf_prog *prog; 1029 struct netlink_ext_ack *extack; 1030 }; 1031 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 1032 struct { 1033 struct bpf_offloaded_map *offmap; 1034 }; 1035 /* XDP_SETUP_XSK_POOL */ 1036 struct { 1037 struct xsk_buff_pool *pool; 1038 u16 queue_id; 1039 } xsk; 1040 }; 1041 }; 1042 1043 /* Flags for ndo_xsk_wakeup. */ 1044 #define XDP_WAKEUP_RX (1 << 0) 1045 #define XDP_WAKEUP_TX (1 << 1) 1046 1047 #ifdef CONFIG_XFRM_OFFLOAD 1048 struct xfrmdev_ops { 1049 int (*xdo_dev_state_add)(struct net_device *dev, 1050 struct xfrm_state *x, 1051 struct netlink_ext_ack *extack); 1052 void (*xdo_dev_state_delete)(struct net_device *dev, 1053 struct xfrm_state *x); 1054 void (*xdo_dev_state_free)(struct net_device *dev, 1055 struct xfrm_state *x); 1056 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 1057 struct xfrm_state *x); 1058 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1059 void (*xdo_dev_state_update_stats) (struct xfrm_state *x); 1060 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack); 1061 void (*xdo_dev_policy_delete) (struct xfrm_policy *x); 1062 void (*xdo_dev_policy_free) (struct xfrm_policy *x); 1063 }; 1064 #endif 1065 1066 struct dev_ifalias { 1067 struct rcu_head rcuhead; 1068 char ifalias[]; 1069 }; 1070 1071 struct devlink; 1072 struct tlsdev_ops; 1073 1074 struct netdev_net_notifier { 1075 struct list_head list; 1076 struct notifier_block *nb; 1077 }; 1078 1079 /* 1080 * This structure defines the management hooks for network devices. 1081 * The following hooks can be defined; unless noted otherwise, they are 1082 * optional and can be filled with a null pointer. 1083 * 1084 * int (*ndo_init)(struct net_device *dev); 1085 * This function is called once when a network device is registered. 1086 * The network device can use this for any late stage initialization 1087 * or semantic validation. It can fail with an error code which will 1088 * be propagated back to register_netdev. 1089 * 1090 * void (*ndo_uninit)(struct net_device *dev); 1091 * This function is called when device is unregistered or when registration 1092 * fails. It is not called if init fails. 1093 * 1094 * int (*ndo_open)(struct net_device *dev); 1095 * This function is called when a network device transitions to the up 1096 * state. 1097 * 1098 * int (*ndo_stop)(struct net_device *dev); 1099 * This function is called when a network device transitions to the down 1100 * state. 1101 * 1102 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1103 * struct net_device *dev); 1104 * Called when a packet needs to be transmitted. 1105 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1106 * the queue before that can happen; it's for obsolete devices and weird 1107 * corner cases, but the stack really does a non-trivial amount 1108 * of useless work if you return NETDEV_TX_BUSY. 1109 * Required; cannot be NULL. 1110 * 1111 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1112 * struct net_device *dev 1113 * netdev_features_t features); 1114 * Called by core transmit path to determine if device is capable of 1115 * performing offload operations on a given packet. This is to give 1116 * the device an opportunity to implement any restrictions that cannot 1117 * be otherwise expressed by feature flags. The check is called with 1118 * the set of features that the stack has calculated and it returns 1119 * those the driver believes to be appropriate. 1120 * 1121 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1122 * struct net_device *sb_dev); 1123 * Called to decide which queue to use when device supports multiple 1124 * transmit queues. 1125 * 1126 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1127 * This function is called to allow device receiver to make 1128 * changes to configuration when multicast or promiscuous is enabled. 1129 * 1130 * void (*ndo_set_rx_mode)(struct net_device *dev); 1131 * This function is called device changes address list filtering. 1132 * If driver handles unicast address filtering, it should set 1133 * IFF_UNICAST_FLT in its priv_flags. 1134 * Cannot sleep, called with netif_addr_lock_bh held. 1135 * Deprecated in favor of ndo_set_rx_mode_async. 1136 * 1137 * int (*ndo_set_rx_mode_async)(struct net_device *dev, 1138 * struct netdev_hw_addr_list *uc, 1139 * struct netdev_hw_addr_list *mc); 1140 * Async version of ndo_set_rx_mode which runs in process context 1141 * with rtnl_lock and netdev_lock_ops(dev) held. The uc/mc parameters 1142 * are snapshots of the address lists - iterate with 1143 * netdev_hw_addr_list_for_each(ha, uc). Return 0 on success or a 1144 * negative errno to request a retry via the core backoff. 1145 * 1146 * void (*ndo_work)(struct net_device *dev, unsigned long events); 1147 * Run deferred work scheduled with netdev_work_sched(@events). 1148 * 1149 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1150 * This function is called when the Media Access Control address 1151 * needs to be changed. If this interface is not defined, the 1152 * MAC address can not be changed. 1153 * 1154 * int (*ndo_validate_addr)(struct net_device *dev); 1155 * Test if Media Access Control address is valid for the device. 1156 * 1157 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1158 * Old-style ioctl entry point. This is used internally by the 1159 * ieee802154 subsystem but is no longer called by the device 1160 * ioctl handler. 1161 * 1162 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1163 * Used by the bonding driver for its device specific ioctls: 1164 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1165 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1166 * 1167 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1168 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG and 1169 * SIOCSMIIREG. 1170 * 1171 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1172 * Used to set network devices bus interface parameters. This interface 1173 * is retained for legacy reasons; new devices should use the bus 1174 * interface (PCI) for low level management. 1175 * 1176 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1177 * Called when a user wants to change the Maximum Transfer Unit 1178 * of a device. 1179 * 1180 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1181 * Callback used when the transmitter has not made any progress 1182 * for dev->watchdog ticks. 1183 * 1184 * void (*ndo_get_stats64)(struct net_device *dev, 1185 * struct rtnl_link_stats64 *storage); 1186 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1187 * Called when a user wants to get the network device usage 1188 * statistics. Drivers must do one of the following: 1189 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1190 * rtnl_link_stats64 structure passed by the caller. 1191 * 2. Define @ndo_get_stats to update a net_device_stats structure 1192 * (which should normally be dev->stats) and return a pointer to 1193 * it. The structure may be changed asynchronously only if each 1194 * field is written atomically. 1195 * 3. Update dev->stats asynchronously and atomically, and define 1196 * neither operation. 1197 * 1198 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1199 * Return true if this device supports offload stats of this attr_id. 1200 * 1201 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1202 * void *attr_data) 1203 * Get statistics for offload operations by attr_id. Write it into the 1204 * attr_data pointer. 1205 * 1206 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1207 * If device supports VLAN filtering this function is called when a 1208 * VLAN id is registered. 1209 * 1210 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1211 * If device supports VLAN filtering this function is called when a 1212 * VLAN id is unregistered. 1213 * 1214 * void (*ndo_poll_controller)(struct net_device *dev); 1215 * 1216 * SR-IOV management functions. 1217 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1218 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1219 * u8 qos, __be16 proto); 1220 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1221 * int max_tx_rate); 1222 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1223 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1224 * int (*ndo_get_vf_config)(struct net_device *dev, 1225 * int vf, struct ifla_vf_info *ivf); 1226 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1227 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1228 * struct nlattr *port[]); 1229 * 1230 * Enable or disable the VF ability to query its RSS Redirection Table and 1231 * Hash Key. This is needed since on some devices VF share this information 1232 * with PF and querying it may introduce a theoretical security risk. 1233 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1234 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1235 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1236 * void *type_data); 1237 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1238 * This is always called from the stack with the rtnl lock held and netif 1239 * tx queues stopped. This allows the netdevice to perform queue 1240 * management safely. 1241 * 1242 * NB: Returning -EOPNOTSUPP for whatever commands means "this qdisc 1243 * is not offloaded (anymore, offloading may have silently stopped)", 1244 * and the offloading flag is cleared. Notably, this is also true for 1245 * dump queries (e.g. TC_*_STATS commands). If the underlying device does 1246 * not report any statistics but is still offloading, return 0 instead. 1247 * 1248 * Fiber Channel over Ethernet (FCoE) offload functions. 1249 * int (*ndo_fcoe_enable)(struct net_device *dev); 1250 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1251 * so the underlying device can perform whatever needed configuration or 1252 * initialization to support acceleration of FCoE traffic. 1253 * 1254 * int (*ndo_fcoe_disable)(struct net_device *dev); 1255 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1256 * so the underlying device can perform whatever needed clean-ups to 1257 * stop supporting acceleration of FCoE traffic. 1258 * 1259 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1260 * struct scatterlist *sgl, unsigned int sgc); 1261 * Called when the FCoE Initiator wants to initialize an I/O that 1262 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1263 * perform necessary setup and returns 1 to indicate the device is set up 1264 * successfully to perform DDP on this I/O, otherwise this returns 0. 1265 * 1266 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1267 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1268 * indicated by the FC exchange id 'xid', so the underlying device can 1269 * clean up and reuse resources for later DDP requests. 1270 * 1271 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1272 * struct scatterlist *sgl, unsigned int sgc); 1273 * Called when the FCoE Target wants to initialize an I/O that 1274 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1275 * perform necessary setup and returns 1 to indicate the device is set up 1276 * successfully to perform DDP on this I/O, otherwise this returns 0. 1277 * 1278 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1279 * struct netdev_fcoe_hbainfo *hbainfo); 1280 * Called when the FCoE Protocol stack wants information on the underlying 1281 * device. This information is utilized by the FCoE protocol stack to 1282 * register attributes with Fiber Channel management service as per the 1283 * FC-GS Fabric Device Management Information(FDMI) specification. 1284 * 1285 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1286 * Called when the underlying device wants to override default World Wide 1287 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1288 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1289 * protocol stack to use. 1290 * 1291 * RFS acceleration. 1292 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1293 * u16 rxq_index, u32 flow_id); 1294 * Set hardware filter for RFS. rxq_index is the target queue index; 1295 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1296 * Return the filter ID on success, or a negative error code. 1297 * 1298 * Slave management functions (for bridge, bonding, etc). 1299 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1300 * Called to make another netdev an underling. 1301 * 1302 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1303 * Called to release previously enslaved netdev. 1304 * 1305 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1306 * struct sk_buff *skb, 1307 * bool all_slaves); 1308 * Get the xmit slave of master device. If all_slaves is true, function 1309 * assume all the slaves can transmit. 1310 * 1311 * Feature/offload setting functions. 1312 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1313 * netdev_features_t features); 1314 * Adjusts the requested feature flags according to device-specific 1315 * constraints, and returns the resulting flags. Must not modify 1316 * the device state. 1317 * 1318 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1319 * Called to update device configuration to new features. Passed 1320 * feature set might be less than what was returned by ndo_fix_features()). 1321 * Must return >0 or -errno if it changed dev->features itself. 1322 * 1323 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1324 * struct net_device *dev, 1325 * const unsigned char *addr, u16 vid, u16 flags, 1326 * bool *notified, struct netlink_ext_ack *extack); 1327 * Adds an FDB entry to dev for addr. 1328 * Callee shall set *notified to true if it sent any appropriate 1329 * notification(s). Otherwise core will send a generic one. 1330 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1331 * struct net_device *dev, 1332 * const unsigned char *addr, u16 vid 1333 * bool *notified, struct netlink_ext_ack *extack); 1334 * Deletes the FDB entry from dev corresponding to 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_bulk)(struct nlmsghdr *nlh, struct net_device *dev, 1338 * struct netlink_ext_ack *extack); 1339 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1340 * struct net_device *dev, struct net_device *filter_dev, 1341 * int *idx) 1342 * Used to add FDB entries to dump requests. Implementers should add 1343 * entries to skb and update idx with the number of entries. 1344 * 1345 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[], 1346 * u16 nlmsg_flags, struct netlink_ext_ack *extack); 1347 * Adds an MDB entry to dev. 1348 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[], 1349 * struct netlink_ext_ack *extack); 1350 * Deletes the MDB entry from dev. 1351 * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[], 1352 * struct netlink_ext_ack *extack); 1353 * Bulk deletes MDB entries from dev. 1354 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb, 1355 * struct netlink_callback *cb); 1356 * Dumps MDB entries from dev. The first argument (marker) in the netlink 1357 * callback is used by core rtnetlink code. 1358 * 1359 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1360 * u16 flags, struct netlink_ext_ack *extack) 1361 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1362 * struct net_device *dev, u32 filter_mask, 1363 * int nlflags) 1364 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1365 * u16 flags); 1366 * 1367 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1368 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1369 * which do not represent real hardware may define this to allow their 1370 * userspace components to manage their virtual carrier state. Devices 1371 * that determine carrier state from physical hardware properties (eg 1372 * network cables) or protocol-dependent mechanisms (eg 1373 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1374 * 1375 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1376 * struct netdev_phys_item_id *ppid); 1377 * Called to get ID of physical port of this device. If driver does 1378 * not implement this, it is assumed that the hw is not able to have 1379 * multiple net devices on single physical port. 1380 * 1381 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1382 * struct netdev_phys_item_id *ppid) 1383 * Called to get the parent ID of the physical port of this device. 1384 * 1385 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1386 * struct net_device *dev) 1387 * Called by upper layer devices to accelerate switching or other 1388 * station functionality into hardware. 'pdev is the lowerdev 1389 * to use for the offload and 'dev' is the net device that will 1390 * back the offload. Returns a pointer to the private structure 1391 * the upper layer will maintain. 1392 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1393 * Called by upper layer device to delete the station created 1394 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1395 * the station and priv is the structure returned by the add 1396 * operation. 1397 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1398 * int queue_index, u32 maxrate); 1399 * Called when a user wants to set a max-rate limitation of specific 1400 * TX queue. 1401 * int (*ndo_get_iflink)(const struct net_device *dev); 1402 * Called to get the iflink value of this device. 1403 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1404 * This function is used to get egress tunnel information for given skb. 1405 * This is useful for retrieving outer tunnel header parameters while 1406 * sampling packet. 1407 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1408 * This function is used to specify the headroom that the skb must 1409 * consider when allocation skb during packet reception. Setting 1410 * appropriate rx headroom value allows avoiding skb head copy on 1411 * forward. Setting a negative value resets the rx headroom to the 1412 * default value. 1413 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1414 * This function is used to set or query state related to XDP on the 1415 * netdevice and manage BPF offload. See definition of 1416 * enum bpf_netdev_command for details. 1417 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1418 * u32 flags); 1419 * This function is used to submit @n XDP packets for transmit on a 1420 * netdevice. Returns number of frames successfully transmitted, frames 1421 * that got dropped are freed/returned via xdp_return_frame(). 1422 * Returns negative number, means general error invoking ndo, meaning 1423 * no frames were xmit'ed and core-caller will free all frames. 1424 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1425 * struct xdp_buff *xdp); 1426 * Get the xmit slave of master device based on the xdp_buff. 1427 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1428 * This function is used to wake up the softirq, ksoftirqd or kthread 1429 * responsible for sending and/or receiving packets on a specific 1430 * queue id bound to an AF_XDP socket. The flags field specifies if 1431 * only RX, only Tx, or both should be woken up using the flags 1432 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1433 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm_kern *p, 1434 * int cmd); 1435 * Add, change, delete or get information on an IPv4 tunnel. 1436 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1437 * If a device is paired with a peer device, return the peer instance. 1438 * The caller must be under RCU read context. 1439 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1440 * Get the forwarding path to reach the real device from the HW destination address 1441 * ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1442 * const struct skb_shared_hwtstamps *hwtstamps, 1443 * bool cycles); 1444 * Get hardware timestamp based on normal/adjustable time or free running 1445 * cycle counter. This function is required if physical clock supports a 1446 * free running cycle counter. 1447 * 1448 * int (*ndo_hwtstamp_get)(struct net_device *dev, 1449 * struct kernel_hwtstamp_config *kernel_config); 1450 * Get the currently configured hardware timestamping parameters for the 1451 * NIC device. 1452 * 1453 * int (*ndo_hwtstamp_set)(struct net_device *dev, 1454 * struct kernel_hwtstamp_config *kernel_config, 1455 * struct netlink_ext_ack *extack); 1456 * Change the hardware timestamping parameters for NIC device. 1457 */ 1458 struct net_device_ops { 1459 int (*ndo_init)(struct net_device *dev); 1460 void (*ndo_uninit)(struct net_device *dev); 1461 int (*ndo_open)(struct net_device *dev); 1462 int (*ndo_stop)(struct net_device *dev); 1463 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1464 struct net_device *dev); 1465 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1466 struct net_device *dev, 1467 netdev_features_t features); 1468 u16 (*ndo_select_queue)(struct net_device *dev, 1469 struct sk_buff *skb, 1470 struct net_device *sb_dev); 1471 void (*ndo_change_rx_flags)(struct net_device *dev, 1472 int flags); 1473 void (*ndo_set_rx_mode)(struct net_device *dev); 1474 int (*ndo_set_rx_mode_async)( 1475 struct net_device *dev, 1476 struct netdev_hw_addr_list *uc, 1477 struct netdev_hw_addr_list *mc); 1478 void (*ndo_work)(struct net_device *dev, 1479 unsigned long events); 1480 int (*ndo_set_mac_address)(struct net_device *dev, 1481 void *addr); 1482 int (*ndo_validate_addr)(struct net_device *dev); 1483 int (*ndo_do_ioctl)(struct net_device *dev, 1484 struct ifreq *ifr, int cmd); 1485 int (*ndo_eth_ioctl)(struct net_device *dev, 1486 struct ifreq *ifr, int cmd); 1487 int (*ndo_siocbond)(struct net_device *dev, 1488 struct ifreq *ifr, int cmd); 1489 int (*ndo_siocwandev)(struct net_device *dev, 1490 struct if_settings *ifs); 1491 int (*ndo_siocdevprivate)(struct net_device *dev, 1492 struct ifreq *ifr, 1493 void __user *data, int cmd); 1494 int (*ndo_set_config)(struct net_device *dev, 1495 struct ifmap *map); 1496 int (*ndo_change_mtu)(struct net_device *dev, 1497 int new_mtu); 1498 int (*ndo_neigh_setup)(struct net_device *dev, 1499 struct neigh_parms *); 1500 void (*ndo_tx_timeout) (struct net_device *dev, 1501 unsigned int txqueue); 1502 1503 void (*ndo_get_stats64)(struct net_device *dev, 1504 struct rtnl_link_stats64 *storage); 1505 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1506 int (*ndo_get_offload_stats)(int attr_id, 1507 const struct net_device *dev, 1508 void *attr_data); 1509 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1510 1511 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1512 __be16 proto, u16 vid); 1513 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1514 __be16 proto, u16 vid); 1515 #ifdef CONFIG_NET_POLL_CONTROLLER 1516 void (*ndo_poll_controller)(struct net_device *dev); 1517 int (*ndo_netpoll_setup)(struct net_device *dev); 1518 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1519 #endif 1520 int (*ndo_set_vf_mac)(struct net_device *dev, 1521 int queue, u8 *mac); 1522 int (*ndo_set_vf_vlan)(struct net_device *dev, 1523 int queue, u16 vlan, 1524 u8 qos, __be16 proto); 1525 int (*ndo_set_vf_rate)(struct net_device *dev, 1526 int vf, int min_tx_rate, 1527 int max_tx_rate); 1528 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1529 int vf, bool setting); 1530 int (*ndo_set_vf_trust)(struct net_device *dev, 1531 int vf, bool setting); 1532 int (*ndo_get_vf_config)(struct net_device *dev, 1533 int vf, 1534 struct ifla_vf_info *ivf); 1535 int (*ndo_set_vf_link_state)(struct net_device *dev, 1536 int vf, int link_state); 1537 int (*ndo_get_vf_stats)(struct net_device *dev, 1538 int vf, 1539 struct ifla_vf_stats 1540 *vf_stats); 1541 int (*ndo_set_vf_port)(struct net_device *dev, 1542 int vf, 1543 struct nlattr *port[]); 1544 int (*ndo_get_vf_port)(struct net_device *dev, 1545 int vf, struct sk_buff *skb); 1546 int (*ndo_get_vf_guid)(struct net_device *dev, 1547 int vf, 1548 struct ifla_vf_guid *node_guid, 1549 struct ifla_vf_guid *port_guid); 1550 int (*ndo_set_vf_guid)(struct net_device *dev, 1551 int vf, u64 guid, 1552 int guid_type); 1553 int (*ndo_set_vf_rss_query_en)( 1554 struct net_device *dev, 1555 int vf, bool setting); 1556 int (*ndo_setup_tc)(struct net_device *dev, 1557 enum tc_setup_type type, 1558 void *type_data); 1559 #if IS_ENABLED(CONFIG_FCOE) 1560 int (*ndo_fcoe_enable)(struct net_device *dev); 1561 int (*ndo_fcoe_disable)(struct net_device *dev); 1562 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1563 u16 xid, 1564 struct scatterlist *sgl, 1565 unsigned int sgc); 1566 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1567 u16 xid); 1568 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1569 u16 xid, 1570 struct scatterlist *sgl, 1571 unsigned int sgc); 1572 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1573 struct netdev_fcoe_hbainfo *hbainfo); 1574 #endif 1575 1576 #if IS_ENABLED(CONFIG_LIBFCOE) 1577 #define NETDEV_FCOE_WWNN 0 1578 #define NETDEV_FCOE_WWPN 1 1579 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1580 u64 *wwn, int type); 1581 #endif 1582 1583 #ifdef CONFIG_RFS_ACCEL 1584 int (*ndo_rx_flow_steer)(struct net_device *dev, 1585 const struct sk_buff *skb, 1586 u16 rxq_index, 1587 u32 flow_id); 1588 #endif 1589 int (*ndo_add_slave)(struct net_device *dev, 1590 struct net_device *slave_dev, 1591 struct netlink_ext_ack *extack); 1592 int (*ndo_del_slave)(struct net_device *dev, 1593 struct net_device *slave_dev); 1594 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1595 struct sk_buff *skb, 1596 bool all_slaves); 1597 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1598 struct sock *sk); 1599 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1600 netdev_features_t features); 1601 int (*ndo_set_features)(struct net_device *dev, 1602 netdev_features_t features); 1603 int (*ndo_neigh_construct)(struct net_device *dev, 1604 struct neighbour *n); 1605 void (*ndo_neigh_destroy)(struct net_device *dev, 1606 struct neighbour *n); 1607 1608 int (*ndo_fdb_add)(struct ndmsg *ndm, 1609 struct nlattr *tb[], 1610 struct net_device *dev, 1611 const unsigned char *addr, 1612 u16 vid, 1613 u16 flags, 1614 bool *notified, 1615 struct netlink_ext_ack *extack); 1616 int (*ndo_fdb_del)(struct ndmsg *ndm, 1617 struct nlattr *tb[], 1618 struct net_device *dev, 1619 const unsigned char *addr, 1620 u16 vid, 1621 bool *notified, 1622 struct netlink_ext_ack *extack); 1623 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, 1624 struct net_device *dev, 1625 struct netlink_ext_ack *extack); 1626 int (*ndo_fdb_dump)(struct sk_buff *skb, 1627 struct netlink_callback *cb, 1628 struct net_device *dev, 1629 struct net_device *filter_dev, 1630 int *idx); 1631 int (*ndo_fdb_get)(struct sk_buff *skb, 1632 struct nlattr *tb[], 1633 struct net_device *dev, 1634 const unsigned char *addr, 1635 u16 vid, u32 portid, u32 seq, 1636 struct netlink_ext_ack *extack); 1637 int (*ndo_mdb_add)(struct net_device *dev, 1638 struct nlattr *tb[], 1639 u16 nlmsg_flags, 1640 struct netlink_ext_ack *extack); 1641 int (*ndo_mdb_del)(struct net_device *dev, 1642 struct nlattr *tb[], 1643 struct netlink_ext_ack *extack); 1644 int (*ndo_mdb_del_bulk)(struct net_device *dev, 1645 struct nlattr *tb[], 1646 struct netlink_ext_ack *extack); 1647 int (*ndo_mdb_dump)(struct net_device *dev, 1648 struct sk_buff *skb, 1649 struct netlink_callback *cb); 1650 int (*ndo_mdb_get)(struct net_device *dev, 1651 struct nlattr *tb[], u32 portid, 1652 u32 seq, 1653 struct netlink_ext_ack *extack); 1654 int (*ndo_bridge_setlink)(struct net_device *dev, 1655 struct nlmsghdr *nlh, 1656 u16 flags, 1657 struct netlink_ext_ack *extack); 1658 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1659 u32 pid, u32 seq, 1660 struct net_device *dev, 1661 u32 filter_mask, 1662 int nlflags); 1663 int (*ndo_bridge_dellink)(struct net_device *dev, 1664 struct nlmsghdr *nlh, 1665 u16 flags); 1666 int (*ndo_change_carrier)(struct net_device *dev, 1667 bool new_carrier); 1668 int (*ndo_get_phys_port_id)(struct net_device *dev, 1669 struct netdev_phys_item_id *ppid); 1670 int (*ndo_get_port_parent_id)(struct net_device *dev, 1671 struct netdev_phys_item_id *ppid); 1672 int (*ndo_get_phys_port_name)(struct net_device *dev, 1673 char *name, size_t len); 1674 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1675 struct net_device *dev); 1676 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1677 void *priv); 1678 1679 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1680 int queue_index, 1681 u32 maxrate); 1682 int (*ndo_get_iflink)(const struct net_device *dev); 1683 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1684 struct sk_buff *skb); 1685 void (*ndo_set_rx_headroom)(struct net_device *dev, 1686 int needed_headroom); 1687 int (*ndo_bpf)(struct net_device *dev, 1688 struct netdev_bpf *bpf); 1689 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1690 struct xdp_frame **xdp, 1691 u32 flags); 1692 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1693 struct xdp_buff *xdp); 1694 int (*ndo_xsk_wakeup)(struct net_device *dev, 1695 u32 queue_id, u32 flags); 1696 int (*ndo_tunnel_ctl)(struct net_device *dev, 1697 struct ip_tunnel_parm_kern *p, 1698 int cmd); 1699 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1700 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1701 struct net_device_path *path); 1702 ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1703 const struct skb_shared_hwtstamps *hwtstamps, 1704 bool cycles); 1705 int (*ndo_hwtstamp_get)(struct net_device *dev, 1706 struct kernel_hwtstamp_config *kernel_config); 1707 int (*ndo_hwtstamp_set)(struct net_device *dev, 1708 struct kernel_hwtstamp_config *kernel_config, 1709 struct netlink_ext_ack *extack); 1710 1711 #if IS_ENABLED(CONFIG_NET_SHAPER) 1712 /** 1713 * @net_shaper_ops: Device shaping offload operations 1714 * see include/net/net_shapers.h 1715 */ 1716 const struct net_shaper_ops *net_shaper_ops; 1717 #endif 1718 }; 1719 1720 /** 1721 * enum netdev_priv_flags - &struct net_device priv_flags 1722 * 1723 * These are the &struct net_device, they are only set internally 1724 * by drivers and used in the kernel. These flags are invisible to 1725 * userspace; this means that the order of these flags can change 1726 * during any kernel release. 1727 * 1728 * You should add bitfield booleans after either net_device::priv_flags 1729 * (hotpath) or ::threaded (slowpath) instead of extending these flags. 1730 * 1731 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1732 * @IFF_EBRIDGE: Ethernet bridging device 1733 * @IFF_BONDING: bonding master or slave 1734 * @IFF_ISATAP: ISATAP interface (RFC4214) 1735 * @IFF_WAN_HDLC: WAN HDLC device 1736 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1737 * release skb->dst 1738 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1739 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1740 * @IFF_MACVLAN_PORT: device used as macvlan port 1741 * @IFF_BRIDGE_PORT: device used as bridge port 1742 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1743 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1744 * @IFF_UNICAST_FLT: Supports unicast filtering 1745 * @IFF_TEAM_PORT: device used as team port 1746 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1747 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1748 * change when it's running 1749 * @IFF_MACVLAN: Macvlan device 1750 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1751 * underlying stacked devices 1752 * @IFF_L3MDEV_MASTER: device is an L3 master device 1753 * @IFF_NO_QUEUE: device can run without qdisc attached 1754 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1755 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1756 * @IFF_TEAM: device is a team device 1757 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1758 * entity (i.e. the master device for bridged veth) 1759 * @IFF_MACSEC: device is a MACsec device 1760 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1761 * @IFF_FAILOVER: device is a failover master device 1762 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1763 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1764 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf 1765 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1766 * skb_headlen(skb) == 0 (data starts from frag0) 1767 */ 1768 enum netdev_priv_flags { 1769 IFF_802_1Q_VLAN = 1<<0, 1770 IFF_EBRIDGE = 1<<1, 1771 IFF_BONDING = 1<<2, 1772 IFF_ISATAP = 1<<3, 1773 IFF_WAN_HDLC = 1<<4, 1774 IFF_XMIT_DST_RELEASE = 1<<5, 1775 IFF_DONT_BRIDGE = 1<<6, 1776 IFF_DISABLE_NETPOLL = 1<<7, 1777 IFF_MACVLAN_PORT = 1<<8, 1778 IFF_BRIDGE_PORT = 1<<9, 1779 IFF_OVS_DATAPATH = 1<<10, 1780 IFF_TX_SKB_SHARING = 1<<11, 1781 IFF_UNICAST_FLT = 1<<12, 1782 IFF_TEAM_PORT = 1<<13, 1783 IFF_SUPP_NOFCS = 1<<14, 1784 IFF_LIVE_ADDR_CHANGE = 1<<15, 1785 IFF_MACVLAN = 1<<16, 1786 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1787 IFF_L3MDEV_MASTER = 1<<18, 1788 IFF_NO_QUEUE = 1<<19, 1789 IFF_OPENVSWITCH = 1<<20, 1790 IFF_L3MDEV_SLAVE = 1<<21, 1791 IFF_TEAM = 1<<22, 1792 IFF_PHONY_HEADROOM = 1<<24, 1793 IFF_MACSEC = 1<<25, 1794 IFF_NO_RX_HANDLER = 1<<26, 1795 IFF_FAILOVER = 1<<27, 1796 IFF_FAILOVER_SLAVE = 1<<28, 1797 IFF_L3MDEV_RX_HANDLER = 1<<29, 1798 IFF_NO_ADDRCONF = BIT_ULL(30), 1799 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31), 1800 }; 1801 1802 /* Specifies the type of the struct net_device::ml_priv pointer */ 1803 enum netdev_ml_priv_type { 1804 ML_PRIV_NONE, 1805 ML_PRIV_CAN, 1806 }; 1807 1808 enum netdev_stat_type { 1809 NETDEV_PCPU_STAT_NONE, 1810 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */ 1811 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */ 1812 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */ 1813 }; 1814 1815 enum netmem_tx_mode { 1816 NETMEM_TX_NONE, /* no netmem TX support */ 1817 NETMEM_TX_DMA, /* DMA-capable netmem TX (real HW) */ 1818 NETMEM_TX_NO_DMA, /* no DMA, e.g. passthrough for virtual devs */ 1819 }; 1820 1821 enum netdev_reg_state { 1822 NETREG_UNINITIALIZED = 0, 1823 NETREG_REGISTERED, /* completed register_netdevice */ 1824 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1825 NETREG_UNREGISTERED, /* completed unregister todo */ 1826 NETREG_RELEASED, /* called free_netdev */ 1827 NETREG_DUMMY, /* dummy device for NAPI poll */ 1828 }; 1829 1830 /** 1831 * struct net_device - The DEVICE structure. 1832 * 1833 * Actually, this whole structure is a big mistake. It mixes I/O 1834 * data with strictly "high-level" data, and it has to know about 1835 * almost every data structure used in the INET module. 1836 * 1837 * @priv_flags: flags invisible to userspace defined as bits, see 1838 * enum netdev_priv_flags for the definitions 1839 * @lltx: device supports lockless Tx. Deprecated for real HW 1840 * drivers. Mainly used by logical interfaces, such as 1841 * bonding and tunnels 1842 * @netmem_tx: device netmem TX mode 1843 * 1844 * @name: This is the first field of the "visible" part of this structure 1845 * (i.e. as seen by users in the "Space.c" file). It is the name 1846 * of the interface. 1847 * 1848 * @name_node: Name hashlist node 1849 * @ifalias: SNMP alias 1850 * @mem_end: Shared memory end 1851 * @mem_start: Shared memory start 1852 * @base_addr: Device I/O address 1853 * @irq: Device IRQ number 1854 * 1855 * @state: Generic network queuing layer state, see netdev_state_t 1856 * @dev_list: The global list of network devices 1857 * @napi_list: List entry used for polling NAPI devices 1858 * @unreg_list: List entry when we are unregistering the 1859 * device; see the function unregister_netdev 1860 * @unreg_list_net:List entry when we are unregistering the cross-netns 1861 * device; see the function unregister_netdevice_queue_net() 1862 * @close_list: List entry used when we are closing the device 1863 * @ptype_all: Device-specific packet handlers for all protocols 1864 * @ptype_specific: Device-specific, protocol-specific packet handlers 1865 * 1866 * @adj_list: Directly linked devices, like slaves for bonding 1867 * @features: Currently active device features 1868 * @hw_features: User-changeable features 1869 * 1870 * @wanted_features: User-requested features 1871 * @vlan_features: Mask of features inheritable by VLAN devices 1872 * 1873 * @hw_enc_features: Mask of features inherited by encapsulating devices 1874 * This field indicates what encapsulation 1875 * offloads the hardware is capable of doing, 1876 * and drivers will need to set them appropriately. 1877 * 1878 * @mpls_features: Mask of features inheritable by MPLS 1879 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1880 * @mangleid_features: Mask of features requiring MANGLEID, will be 1881 * disabled together with the latter. 1882 * 1883 * @ifindex: interface index 1884 * @group: The group the device belongs to 1885 * 1886 * @stats: Statistics struct, which was left as a legacy, use 1887 * rtnl_link_stats64 instead 1888 * 1889 * @core_stats: core networking counters, 1890 * do not use this in drivers 1891 * @carrier_up_count: Number of times the carrier has been up 1892 * @carrier_down_count: Number of times the carrier has been down 1893 * 1894 * @wireless_handlers: List of functions to handle Wireless Extensions, 1895 * instead of ioctl, 1896 * see <net/iw_handler.h> for details. 1897 * 1898 * @netdev_ops: Includes several pointers to callbacks, 1899 * if one wants to override the ndo_*() functions 1900 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks. 1901 * @xsk_tx_metadata_ops: Includes pointers to AF_XDP TX metadata callbacks. 1902 * @ethtool_ops: Management operations 1903 * @l3mdev_ops: Layer 3 master device operations 1904 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1905 * discovery handling. Necessary for e.g. 6LoWPAN. 1906 * @xfrmdev_ops: Transformation offload operations 1907 * @tlsdev_ops: Transport Layer Security offload operations 1908 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1909 * of Layer 2 headers. 1910 * 1911 * @flags: Interface flags (a la BSD) 1912 * @xdp_features: XDP capability supported by the device 1913 * @gflags: Global flags ( kept as legacy ) 1914 * @priv_len: Size of the ->priv flexible array 1915 * @priv: Flexible array containing private data 1916 * @operstate: RFC2863 operstate 1917 * @link_mode: Mapping policy to operstate 1918 * @if_port: Selectable AUI, TP, ... 1919 * @dma: DMA channel 1920 * @mtu: Interface MTU value 1921 * @min_mtu: Interface Minimum MTU value 1922 * @max_mtu: Interface Maximum MTU value 1923 * @type: Interface hardware type 1924 * @hard_header_len: Maximum hardware header length. 1925 * @min_header_len: Minimum hardware header length 1926 * 1927 * @needed_headroom: Extra headroom the hardware may need, but not in all 1928 * cases can this be guaranteed 1929 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1930 * cases can this be guaranteed. Some cases also use 1931 * LL_MAX_HEADER instead to allocate the skb 1932 * 1933 * interface address info: 1934 * 1935 * @perm_addr: Permanent hw address 1936 * @addr_assign_type: Hw address assignment type 1937 * @addr_len: Hardware address length 1938 * @upper_level: Maximum depth level of upper devices. 1939 * @lower_level: Maximum depth level of lower devices. 1940 * @threaded: napi threaded state. 1941 * @neigh_priv_len: Used in neigh_alloc() 1942 * @dev_id: Used to differentiate devices that share 1943 * the same link layer address 1944 * @dev_port: Used to differentiate devices that share 1945 * the same function 1946 * @addr_list_lock: XXX: need comments on this one 1947 * @name_assign_type: network interface name assignment type 1948 * @uc_promisc: Counter that indicates promiscuous mode 1949 * has been enabled due to the need to listen to 1950 * additional unicast addresses in a device that 1951 * does not implement ndo_set_rx_mode() 1952 * @work_node: List entry for async netdev_work processing 1953 * @work_tracker: Refcount tracker for async netdev_work 1954 * @work_pending: Driver-defined pending netdev_work, passed to 1955 * ndo_work() (see netdev_work_sched()) 1956 * @work_core_pending: Core-defined pending netdev_work (NETDEV_WORK_*) 1957 * @rx_mode_addr_cache: Recycled snapshot entries for rx_mode work 1958 * @rx_mode_retry_timer: Timer that re-queues rx_mode work after failure 1959 * @rx_mode_retry_count: Number of consecutive retries already scheduled 1960 * @uc: unicast mac addresses 1961 * @mc: multicast mac addresses 1962 * @dev_addrs: list of device hw addresses 1963 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1964 * @promiscuity: Number of times the NIC is told to work in 1965 * promiscuous mode; if it becomes 0 the NIC will 1966 * exit promiscuous mode 1967 * @allmulti: Counter, enables or disables allmulticast mode 1968 * 1969 * @vlan_info: VLAN info 1970 * @dsa_ptr: dsa specific data 1971 * @tipc_ptr: TIPC specific data 1972 * @ip_ptr: IPv4 specific data 1973 * @ip6_ptr: IPv6 specific data 1974 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1975 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1976 * device struct 1977 * @mpls_ptr: mpls_dev struct pointer 1978 * @mctp_ptr: MCTP specific data 1979 * @psp_dev: PSP crypto device registered for this netdev 1980 * 1981 * @dev_addr: Hw address (before bcast, 1982 * because most packets are unicast) 1983 * 1984 * @_rx: Array of RX queues 1985 * @num_rx_queues: Number of RX queues 1986 * allocated at register_netdev() time 1987 * @real_num_rx_queues: Number of RX queues currently active in device 1988 * @xdp_prog: XDP sockets filter program pointer 1989 * 1990 * @rx_handler: handler for received packets 1991 * @rx_handler_data: XXX: need comments on this one 1992 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing 1993 * @ingress_queue: XXX: need comments on this one 1994 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1995 * @broadcast: hw bcast address 1996 * 1997 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1998 * indexed by RX queue number. Assigned by driver. 1999 * This must only be set if the ndo_rx_flow_steer 2000 * operation is defined 2001 * @index_hlist: Device index hash chain 2002 * 2003 * @_tx: Array of TX queues 2004 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 2005 * @real_num_tx_queues: Number of TX queues currently active in device 2006 * @qdisc: Root qdisc from userspace point of view 2007 * @tx_queue_len: Max frames per queue allowed 2008 * @tx_global_lock: XXX: need comments on this one 2009 * @xdp_bulkq: XDP device bulk queue 2010 * @xps_maps: all CPUs/RXQs maps for XPS device 2011 * 2012 * @xps_maps: XXX: need comments on this one 2013 * @tcx_egress: BPF & clsact qdisc specific data for egress processing 2014 * @nf_hooks_egress: netfilter hooks executed for egress packets 2015 * @qdisc_hash: qdisc hash table 2016 * @watchdog_timeo: Represents the timeout that is used by 2017 * the watchdog (see dev_watchdog()) 2018 * @watchdog_lock: protect watchdog_ref_held 2019 * @watchdog_ref_held: True if the watchdog device ref is taken. 2020 * @watchdog_timer: List of timers 2021 * 2022 * @proto_down_reason: reason a netdev interface is held down 2023 * @pcpu_refcnt: Number of references to this device 2024 * @dev_refcnt: Number of references to this device 2025 * @refcnt_tracker: Tracker directory for tracked references to this device 2026 * @todo_list: Delayed register/unregister 2027 * @link_watch_list: XXX: need comments on this one 2028 * 2029 * @reg_state: Register/unregister state machine 2030 * @dismantle: Device is going to be freed 2031 * @needs_free_netdev: Should unregister perform free_netdev? 2032 * @priv_destructor: Called from unregister 2033 * @npinfo: XXX: need comments on this one 2034 * @nd_net: Network namespace this network device is inside 2035 * protected by @lock 2036 * 2037 * @ml_priv: Mid-layer private 2038 * @ml_priv_type: Mid-layer private type 2039 * 2040 * @pcpu_stat_type: Type of device statistics which the core should 2041 * allocate/free: none, lstats, tstats, dstats. none 2042 * means the driver is handling statistics allocation/ 2043 * freeing internally. 2044 * @lstats: Loopback statistics: packets, bytes 2045 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes 2046 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes 2047 * 2048 * @garp_port: GARP 2049 * @mrp_port: MRP 2050 * 2051 * @dm_private: Drop monitor private 2052 * 2053 * @dev: Class/net/name entry 2054 * @sysfs_groups: Space for optional device, statistics and wireless 2055 * sysfs groups 2056 * 2057 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 2058 * @rtnl_link_ops: Rtnl_link_ops 2059 * @stat_ops: Optional ops for queue-aware statistics 2060 * @queue_mgmt_ops: Optional ops for queue management 2061 * 2062 * @gso_max_size: Maximum size of generic segmentation offload 2063 * @tso_max_size: Device (as in HW) limit on the max TSO request size 2064 * @gso_max_segs: Maximum number of segments that can be passed to the 2065 * NIC for GSO 2066 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count 2067 * @gso_ipv4_max_size: Maximum size of generic segmentation offload, 2068 * for IPv4. 2069 * 2070 * @dcbnl_ops: Data Center Bridging netlink ops 2071 * @num_tc: Number of traffic classes in the net device 2072 * @tc_to_txq: XXX: need comments on this one 2073 * @prio_tc_map: XXX: need comments on this one 2074 * 2075 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 2076 * 2077 * @priomap: XXX: need comments on this one 2078 * @link_topo: Physical link topology tracking attached PHYs 2079 * @phydev: Physical device may attach itself 2080 * for hardware timestamping 2081 * @sfp_bus: attached &struct sfp_bus structure. 2082 * 2083 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 2084 * 2085 * @proto_down: protocol port state information can be sent to the 2086 * switch driver and used to set the phys state of the 2087 * switch port. 2088 * 2089 * @irq_affinity_auto: driver wants the core to store and re-assign the IRQ 2090 * affinity. Set by netif_enable_irq_affinity(), then 2091 * the driver must create a persistent napi by 2092 * netif_napi_add_config() and finally bind the napi to 2093 * IRQ (via netif_napi_set_irq()). 2094 * 2095 * @rx_cpu_rmap_auto: driver wants the core to manage the ARFS rmap. 2096 * Set by calling netif_enable_cpu_rmap(). 2097 * 2098 * @see_all_hwtstamp_requests: device wants to see calls to 2099 * ndo_hwtstamp_set() for all timestamp requests 2100 * regardless of source, even if those aren't 2101 * HWTSTAMP_SOURCE_NETDEV 2102 * @change_proto_down: device supports setting carrier via IFLA_PROTO_DOWN 2103 * @netns_immutable: interface can't change network namespaces 2104 * @fcoe_mtu: device supports maximum FCoE MTU, 2158 bytes 2105 * 2106 * @net_notifier_list: List of per-net netdev notifier block 2107 * that follow this device when it is moved 2108 * to another network namespace. 2109 * 2110 * @macsec_ops: MACsec offloading ops 2111 * 2112 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 2113 * offload capabilities of the device 2114 * @udp_tunnel_nic: UDP tunnel offload state 2115 * @ethtool: ethtool related state 2116 * @xdp_state: stores info on attached XDP BPF programs 2117 * 2118 * @nested_level: Used as a parameter of spin_lock_nested() of 2119 * dev->addr_list_lock. 2120 * @unlink_list: As netif_addr_lock() can be called recursively, 2121 * keep a list of interfaces to be deleted. 2122 * @gro_max_size: Maximum size of aggregated packet in generic 2123 * receive offload (GRO) 2124 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic 2125 * receive offload (GRO), for IPv4. 2126 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP 2127 * zero copy driver 2128 * 2129 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 2130 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 2131 * @watchdog_dev_tracker: refcount tracker used by watchdog. 2132 * @dev_registered_tracker: tracker for reference held while 2133 * registered 2134 * @offload_xstats_l3: L3 HW stats for this netdevice. 2135 * 2136 * @devlink_port: Pointer to related devlink port structure. 2137 * Assigned by a driver before netdev registration using 2138 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static 2139 * during the time netdevice is registered. 2140 * 2141 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem, 2142 * where the clock is recovered. 2143 * 2144 * @max_pacing_offload_horizon: max EDT offload horizon in nsec. 2145 * @napi_config: An array of napi_config structures containing per-NAPI 2146 * settings. 2147 * @num_napi_configs: number of allocated NAPI config structs, 2148 * always >= max(num_rx_queues, num_tx_queues). 2149 * @gro_flush_timeout: timeout for GRO layer in NAPI 2150 * @napi_defer_hard_irqs: If not zero, provides a counter that would 2151 * allow to avoid NIC hard IRQ, on busy queues. 2152 * 2153 * @neighbours: List heads pointing to this device's neighbours' 2154 * dev_list, one per address-family. 2155 * @hwprov: Tracks which PTP performs hardware packet time stamping. 2156 * 2157 * FIXME: cleanup struct net_device such that network protocol info 2158 * moves out. 2159 */ 2160 2161 struct net_device { 2162 /* Cacheline organization can be found documented in 2163 * Documentation/networking/net_cachelines/net_device.rst. 2164 * Please update the document when adding new fields. 2165 */ 2166 2167 /* TX read-mostly hotpath */ 2168 __cacheline_group_begin(net_device_read_tx); 2169 struct_group(priv_flags_fast, 2170 unsigned long priv_flags:32; 2171 unsigned long lltx:1; 2172 unsigned long netmem_tx:2; 2173 ); 2174 const struct net_device_ops *netdev_ops; 2175 const struct header_ops *header_ops; 2176 struct netdev_queue *_tx; 2177 netdev_features_t gso_partial_features; 2178 unsigned int real_num_tx_queues; 2179 unsigned int gso_max_size; 2180 unsigned int gso_ipv4_max_size; 2181 u16 gso_max_segs; 2182 s16 num_tc; 2183 /* Note : dev->mtu is often read without holding a lock. 2184 * Writers usually hold RTNL. 2185 * It is recommended to use READ_ONCE() to annotate the reads, 2186 * and to use WRITE_ONCE() to annotate the writes. 2187 */ 2188 unsigned int mtu; 2189 unsigned short needed_headroom; 2190 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2191 #ifdef CONFIG_XPS 2192 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2193 #endif 2194 #ifdef CONFIG_NETFILTER_EGRESS 2195 struct nf_hook_entries __rcu *nf_hooks_egress; 2196 #endif 2197 #ifdef CONFIG_NET_XGRESS 2198 struct bpf_mprog_entry __rcu *tcx_egress; 2199 #endif 2200 __cacheline_group_end(net_device_read_tx); 2201 2202 /* TXRX read-mostly hotpath */ 2203 __cacheline_group_begin(net_device_read_txrx); 2204 union { 2205 struct pcpu_lstats __percpu *lstats; 2206 struct pcpu_sw_netstats __percpu *tstats; 2207 struct pcpu_dstats __percpu *dstats; 2208 }; 2209 unsigned long state; 2210 unsigned int flags; 2211 unsigned short hard_header_len; 2212 enum netdev_stat_type pcpu_stat_type:8; 2213 netdev_features_t features; 2214 struct inet6_dev __rcu *ip6_ptr; 2215 __cacheline_group_end(net_device_read_txrx); 2216 2217 /* RX read-mostly hotpath */ 2218 __cacheline_group_begin(net_device_read_rx); 2219 struct bpf_prog __rcu *xdp_prog; 2220 struct list_head ptype_specific; 2221 int ifindex; 2222 unsigned int real_num_rx_queues; 2223 struct netdev_rx_queue *_rx; 2224 unsigned int gro_max_size; 2225 unsigned int gro_ipv4_max_size; 2226 rx_handler_func_t __rcu *rx_handler; 2227 void __rcu *rx_handler_data; 2228 possible_net_t nd_net; 2229 #ifdef CONFIG_NETPOLL 2230 struct netpoll_info __rcu *npinfo; 2231 #endif 2232 #ifdef CONFIG_NET_XGRESS 2233 struct bpf_mprog_entry __rcu *tcx_ingress; 2234 #endif 2235 __cacheline_group_end(net_device_read_rx); 2236 2237 char name[IFNAMSIZ]; 2238 struct netdev_name_node *name_node; 2239 struct dev_ifalias __rcu *ifalias; 2240 /* 2241 * I/O specific fields 2242 * FIXME: Merge these and struct ifmap into one 2243 */ 2244 unsigned long mem_end; 2245 unsigned long mem_start; 2246 unsigned long base_addr; 2247 2248 /* 2249 * Some hardware also needs these fields (state,dev_list, 2250 * napi_list,unreg_list,close_list) but they are not 2251 * part of the usual set specified in Space.c. 2252 */ 2253 2254 2255 struct list_head dev_list; 2256 struct list_head napi_list; 2257 struct list_head unreg_list; 2258 #ifdef CONFIG_DEBUG_NET_SMALL_RTNL 2259 struct list_head unreg_list_net; 2260 #endif 2261 struct list_head close_list; 2262 struct list_head ptype_all; 2263 2264 struct { 2265 struct list_head upper; 2266 struct list_head lower; 2267 } adj_list; 2268 2269 /* Read-mostly cache-line for fast-path access */ 2270 xdp_features_t xdp_features; 2271 const struct xdp_metadata_ops *xdp_metadata_ops; 2272 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops; 2273 unsigned short gflags; 2274 2275 unsigned short needed_tailroom; 2276 2277 netdev_features_t hw_features; 2278 netdev_features_t wanted_features; 2279 netdev_features_t vlan_features; 2280 netdev_features_t hw_enc_features; 2281 netdev_features_t mpls_features; 2282 netdev_features_t mangleid_features; 2283 2284 unsigned int min_mtu; 2285 unsigned int max_mtu; 2286 unsigned short type; 2287 unsigned char min_header_len; 2288 unsigned char name_assign_type; 2289 2290 int group; 2291 2292 struct net_device_stats stats; /* not used by modern drivers */ 2293 2294 struct net_device_core_stats __percpu *core_stats; 2295 2296 /* Stats to monitor link on/off, flapping */ 2297 atomic_t carrier_up_count; 2298 atomic_t carrier_down_count; 2299 2300 #ifdef CONFIG_WIRELESS_EXT 2301 const struct iw_handler_def *wireless_handlers; 2302 #endif 2303 const struct ethtool_ops *ethtool_ops; 2304 #ifdef CONFIG_NET_L3_MASTER_DEV 2305 const struct l3mdev_ops *l3mdev_ops; 2306 #endif 2307 #if IS_ENABLED(CONFIG_IPV6) 2308 const struct ndisc_ops *ndisc_ops; 2309 #endif 2310 2311 #ifdef CONFIG_XFRM_OFFLOAD 2312 const struct xfrmdev_ops *xfrmdev_ops; 2313 #endif 2314 2315 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2316 const struct tlsdev_ops *tlsdev_ops; 2317 #endif 2318 2319 unsigned int operstate; 2320 unsigned char link_mode; 2321 2322 unsigned char if_port; 2323 unsigned char dma; 2324 2325 /* Interface address info. */ 2326 unsigned char perm_addr[MAX_ADDR_LEN]; 2327 unsigned char addr_assign_type; 2328 unsigned char addr_len; 2329 unsigned char upper_level; 2330 unsigned char lower_level; 2331 u8 threaded; 2332 2333 unsigned short neigh_priv_len; 2334 unsigned short dev_id; 2335 unsigned short dev_port; 2336 int irq; 2337 u32 priv_len; 2338 2339 spinlock_t addr_list_lock; 2340 2341 struct netdev_hw_addr_list uc; 2342 struct netdev_hw_addr_list mc; 2343 struct netdev_hw_addr_list dev_addrs; 2344 2345 #ifdef CONFIG_SYSFS 2346 struct kset *queues_kset; 2347 #endif 2348 #ifdef CONFIG_LOCKDEP 2349 struct list_head unlink_list; 2350 #endif 2351 unsigned int promiscuity; 2352 unsigned int allmulti; 2353 bool uc_promisc; 2354 struct list_head work_node; 2355 netdevice_tracker work_tracker; 2356 unsigned long work_pending; 2357 unsigned long work_core_pending; 2358 struct netdev_hw_addr_list rx_mode_addr_cache; 2359 struct timer_list rx_mode_retry_timer; 2360 unsigned int rx_mode_retry_count; 2361 #ifdef CONFIG_LOCKDEP 2362 unsigned char nested_level; 2363 #endif 2364 2365 2366 /* Protocol-specific pointers */ 2367 struct in_device __rcu *ip_ptr; 2368 /** @fib_nh_head: nexthops associated with this netdev */ 2369 struct hlist_head fib_nh_head; 2370 2371 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2372 struct vlan_info __rcu *vlan_info; 2373 #endif 2374 #if IS_ENABLED(CONFIG_NET_DSA) 2375 struct dsa_port *dsa_ptr; 2376 #endif 2377 #if IS_ENABLED(CONFIG_TIPC) 2378 struct tipc_bearer __rcu *tipc_ptr; 2379 #endif 2380 #if IS_ENABLED(CONFIG_CFG80211) 2381 struct wireless_dev *ieee80211_ptr; 2382 #endif 2383 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2384 struct wpan_dev *ieee802154_ptr; 2385 #endif 2386 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2387 struct mpls_dev __rcu *mpls_ptr; 2388 #endif 2389 #if IS_ENABLED(CONFIG_MCTP) 2390 struct mctp_dev __rcu *mctp_ptr; 2391 #endif 2392 #if IS_ENABLED(CONFIG_INET_PSP) 2393 struct psp_dev __rcu *psp_dev; 2394 #endif 2395 2396 /* 2397 * Cache lines mostly used on receive path (including eth_type_trans()) 2398 */ 2399 /* Interface address info used in eth_type_trans() */ 2400 const unsigned char *dev_addr; 2401 2402 unsigned int num_rx_queues; 2403 #define GRO_LEGACY_MAX_SIZE 65536u 2404 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2405 * and shinfo->gso_segs is a 16bit field. 2406 */ 2407 #define GRO_MAX_SIZE (8 * 65535u) 2408 unsigned int xdp_zc_max_segs; 2409 struct netdev_queue __rcu *ingress_queue; 2410 #ifdef CONFIG_NETFILTER_INGRESS 2411 struct nf_hook_entries __rcu *nf_hooks_ingress; 2412 #endif 2413 2414 unsigned char broadcast[MAX_ADDR_LEN]; 2415 #ifdef CONFIG_RFS_ACCEL 2416 struct cpu_rmap *rx_cpu_rmap; 2417 #endif 2418 struct hlist_node index_hlist; 2419 2420 /* 2421 * Cache lines mostly used on transmit path 2422 */ 2423 unsigned int num_tx_queues; 2424 struct Qdisc __rcu *qdisc; 2425 unsigned int tx_queue_len; 2426 spinlock_t tx_global_lock; 2427 2428 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2429 2430 #ifdef CONFIG_NET_SCHED 2431 DECLARE_HASHTABLE (qdisc_hash, 4); 2432 #endif 2433 /* These may be needed for future network-power-down code. */ 2434 struct timer_list watchdog_timer; 2435 int watchdog_timeo; 2436 spinlock_t watchdog_lock; 2437 bool watchdog_ref_held; 2438 2439 u32 proto_down_reason; 2440 2441 struct list_head todo_list; 2442 2443 #ifdef CONFIG_PCPU_DEV_REFCNT 2444 int __percpu *pcpu_refcnt; 2445 #else 2446 refcount_t dev_refcnt; 2447 #endif 2448 struct ref_tracker_dir refcnt_tracker; 2449 2450 struct list_head link_watch_list; 2451 2452 u8 reg_state; 2453 2454 bool dismantle; 2455 2456 /** @moving_ns: device is changing netns, protected by @lock */ 2457 bool moving_ns; 2458 /** @rtnl_link_initializing: Device being created, suppress events */ 2459 bool rtnl_link_initializing; 2460 2461 bool needs_free_netdev; 2462 void (*priv_destructor)(struct net_device *dev); 2463 2464 /* mid-layer private */ 2465 void *ml_priv; 2466 enum netdev_ml_priv_type ml_priv_type; 2467 2468 #if IS_ENABLED(CONFIG_GARP) 2469 struct garp_port __rcu *garp_port; 2470 #endif 2471 #if IS_ENABLED(CONFIG_MRP) 2472 struct mrp_port __rcu *mrp_port; 2473 #endif 2474 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2475 struct dm_hw_stat_delta __rcu *dm_private; 2476 #endif 2477 struct device dev; 2478 const struct attribute_group *sysfs_groups[5]; 2479 const struct attribute_group *sysfs_rx_queue_group; 2480 2481 const struct rtnl_link_ops *rtnl_link_ops; 2482 2483 const struct netdev_stat_ops *stat_ops; 2484 2485 const struct netdev_queue_mgmt_ops *queue_mgmt_ops; 2486 2487 /* for setting kernel sock attribute on TCP connection setup */ 2488 #define GSO_MAX_SEGS 65535u 2489 #define GSO_LEGACY_MAX_SIZE 65536u 2490 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2491 * and shinfo->gso_segs is a 16bit field. 2492 */ 2493 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2494 2495 #define TSO_LEGACY_MAX_SIZE 65536 2496 #define TSO_MAX_SIZE UINT_MAX 2497 unsigned int tso_max_size; 2498 #define TSO_MAX_SEGS U16_MAX 2499 u16 tso_max_segs; 2500 2501 #ifdef CONFIG_DCB 2502 const struct dcbnl_rtnl_ops *dcbnl_ops; 2503 #endif 2504 u8 prio_tc_map[TC_BITMASK + 1]; 2505 2506 #if IS_ENABLED(CONFIG_FCOE) 2507 unsigned int fcoe_ddp_xid; 2508 #endif 2509 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2510 struct netprio_map __rcu *priomap; 2511 #endif 2512 struct phy_link_topology *link_topo; 2513 struct phy_device *phydev; 2514 struct sfp_bus *sfp_bus; 2515 struct lock_class_key *qdisc_tx_busylock; 2516 bool proto_down; 2517 bool irq_affinity_auto; 2518 bool rx_cpu_rmap_auto; 2519 2520 /* priv_flags_slow, ungrouped to save space */ 2521 unsigned long see_all_hwtstamp_requests:1; 2522 unsigned long change_proto_down:1; 2523 unsigned long netns_immutable:1; 2524 unsigned long fcoe_mtu:1; 2525 2526 struct list_head net_notifier_list; 2527 2528 #if IS_ENABLED(CONFIG_MACSEC) 2529 /* MACsec management functions */ 2530 const struct macsec_ops *macsec_ops; 2531 #endif 2532 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2533 struct udp_tunnel_nic *udp_tunnel_nic; 2534 2535 /** @cfg: net_device queue-related configuration */ 2536 struct netdev_config *cfg; 2537 /** 2538 * @cfg_pending: same as @cfg but when device is being actively 2539 * reconfigured includes any changes to the configuration 2540 * requested by the user, but which may or may not be rejected. 2541 */ 2542 struct netdev_config *cfg_pending; 2543 struct ethtool_netdev_state *ethtool; 2544 2545 /* protected by rtnl_lock */ 2546 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2547 2548 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2549 netdevice_tracker linkwatch_dev_tracker; 2550 netdevice_tracker watchdog_dev_tracker; 2551 netdevice_tracker dev_registered_tracker; 2552 struct rtnl_hw_stats64 *offload_xstats_l3; 2553 2554 struct devlink_port *devlink_port; 2555 2556 #if IS_ENABLED(CONFIG_DPLL) 2557 struct dpll_pin __rcu *dpll_pin; 2558 #endif 2559 #if IS_ENABLED(CONFIG_PAGE_POOL) 2560 /** @page_pools: page pools created for this netdevice */ 2561 struct hlist_head page_pools; 2562 #endif 2563 2564 /** @irq_moder: dim parameters used if IS_ENABLED(CONFIG_DIMLIB). */ 2565 struct dim_irq_moder *irq_moder; 2566 2567 u64 max_pacing_offload_horizon; 2568 struct napi_config *napi_config; 2569 u32 num_napi_configs; 2570 u32 napi_defer_hard_irqs; 2571 unsigned long gro_flush_timeout; 2572 2573 /** 2574 * @up: copy of @state's IFF_UP, but safe to read with just @lock. 2575 * May report false negatives while the device is being opened 2576 * or closed (@lock does not protect .ndo_open, or .ndo_close). 2577 */ 2578 bool up; 2579 2580 /** 2581 * @request_ops_lock: request the core to run all @netdev_ops and 2582 * @ethtool_ops under the @lock. 2583 */ 2584 bool request_ops_lock; 2585 2586 /** 2587 * @lock: netdev-scope lock, protects a small selection of fields. 2588 * Should always be taken using netdev_lock() / netdev_unlock() helpers. 2589 * Drivers are free to use it for other protection. 2590 * 2591 * For the drivers that implement shaper or queue API, the scope 2592 * of this lock is expanded to cover most ndo/queue/ethtool/sysfs 2593 * operations. Drivers may opt-in to this behavior by setting 2594 * @request_ops_lock. 2595 * 2596 * @lock protection mixes with rtnl_lock in multiple ways, fields are 2597 * either: 2598 * 2599 * - simply protected by the instance @lock; 2600 * 2601 * - double protected - writers hold both locks, readers hold either; 2602 * 2603 * - ops protected - protected by the lock held around the NDOs 2604 * and other callbacks, that is the instance lock on devices for 2605 * which netdev_need_ops_lock() returns true, otherwise by rtnl_lock; 2606 * 2607 * - double ops protected - always protected by rtnl_lock but for 2608 * devices for which netdev_need_ops_lock() returns true - also 2609 * the instance lock. 2610 * 2611 * Simply protects: 2612 * @gro_flush_timeout, @napi_defer_hard_irqs, @napi_list, 2613 * @net_shaper_hierarchy, @reg_state, @threaded 2614 * 2615 * Double protects: 2616 * @up, @moving_ns, @nd_net, @xdp_features 2617 * 2618 * Ops protects: 2619 * @cfg, @cfg_pending, @ethtool, @hwprov 2620 * 2621 * Double ops protects: 2622 * @real_num_rx_queues, @real_num_tx_queues 2623 * 2624 * Also protects some fields in: 2625 * struct napi_struct, struct netdev_queue, struct netdev_rx_queue 2626 * 2627 * Ordering: 2628 * 2629 * - take after rtnl_lock 2630 * 2631 * - for the case of netdev queue leasing, the netdev-scope lock is 2632 * taken for both the virtual and the physical device; to prevent 2633 * deadlocks, the virtual device's lock must always be acquired 2634 * before the physical device's (see netdev_nl_queue_create_doit) 2635 */ 2636 struct mutex lock; 2637 2638 #if IS_ENABLED(CONFIG_NET_SHAPER) 2639 /** 2640 * @net_shaper_hierarchy: data tracking the current shaper status 2641 * see include/net/net_shapers.h 2642 */ 2643 struct net_shaper_hierarchy *net_shaper_hierarchy; 2644 #endif 2645 2646 struct hlist_head neighbours[NEIGH_NR_TABLES]; 2647 2648 struct hwtstamp_provider __rcu *hwprov; 2649 2650 u8 priv[] ____cacheline_aligned 2651 __counted_by(priv_len); 2652 } ____cacheline_aligned; 2653 #define to_net_dev(d) container_of(d, struct net_device, dev) 2654 2655 /* 2656 * Driver should use this to assign devlink port instance to a netdevice 2657 * before it registers the netdevice. Therefore devlink_port is static 2658 * during the netdev lifetime after it is registered. 2659 */ 2660 #define SET_NETDEV_DEVLINK_PORT(dev, port) \ 2661 ({ \ 2662 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \ 2663 ((dev)->devlink_port = (port)); \ 2664 }) 2665 2666 static inline bool netif_elide_gro(const struct net_device *dev) 2667 { 2668 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2669 return true; 2670 return false; 2671 } 2672 2673 #define NETDEV_ALIGN 32 2674 2675 static inline 2676 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2677 { 2678 return READ_ONCE(dev->prio_tc_map[prio & TC_BITMASK]); 2679 } 2680 2681 static inline 2682 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2683 { 2684 if (tc >= READ_ONCE(dev->num_tc)) 2685 return -EINVAL; 2686 2687 WRITE_ONCE(dev->prio_tc_map[prio & TC_BITMASK], tc & TC_BITMASK); 2688 return 0; 2689 } 2690 2691 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2692 void netdev_reset_tc(struct net_device *dev); 2693 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2694 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2695 2696 static inline 2697 int netdev_get_num_tc(const struct net_device *dev) 2698 { 2699 return READ_ONCE(dev->num_tc); 2700 } 2701 2702 static inline void net_prefetch(void *p) 2703 { 2704 prefetch(p); 2705 #if L1_CACHE_BYTES < 128 2706 prefetch((u8 *)p + L1_CACHE_BYTES); 2707 #endif 2708 } 2709 2710 static inline void net_prefetchw(void *p) 2711 { 2712 prefetchw(p); 2713 #if L1_CACHE_BYTES < 128 2714 prefetchw((u8 *)p + L1_CACHE_BYTES); 2715 #endif 2716 } 2717 2718 void netdev_unbind_sb_channel(struct net_device *dev, 2719 struct net_device *sb_dev); 2720 int netdev_bind_sb_channel_queue(struct net_device *dev, 2721 struct net_device *sb_dev, 2722 u8 tc, u16 count, u16 offset); 2723 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2724 static inline int netdev_get_sb_channel(struct net_device *dev) 2725 { 2726 return max_t(int, -READ_ONCE(dev->num_tc), 0); 2727 } 2728 2729 static inline 2730 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2731 unsigned int index) 2732 { 2733 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues); 2734 return &dev->_tx[index]; 2735 } 2736 2737 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2738 const struct sk_buff *skb) 2739 { 2740 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2741 } 2742 2743 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2744 void (*f)(struct net_device *, 2745 struct netdev_queue *, 2746 void *), 2747 void *arg) 2748 { 2749 unsigned int i; 2750 2751 for (i = 0; i < dev->num_tx_queues; i++) 2752 f(dev, &dev->_tx[i], arg); 2753 } 2754 2755 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2756 struct net_device *sb_dev); 2757 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2758 struct sk_buff *skb, 2759 struct net_device *sb_dev); 2760 2761 /* returns the headroom that the master device needs to take in account 2762 * when forwarding to this dev 2763 */ 2764 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2765 { 2766 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2767 } 2768 2769 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2770 { 2771 if (dev->netdev_ops->ndo_set_rx_headroom) 2772 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2773 } 2774 2775 /* set the device rx headroom to the dev's default */ 2776 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2777 { 2778 netdev_set_rx_headroom(dev, -1); 2779 } 2780 2781 static inline void *netdev_get_ml_priv(struct net_device *dev, 2782 enum netdev_ml_priv_type type) 2783 { 2784 if (dev->ml_priv_type != type) 2785 return NULL; 2786 2787 return dev->ml_priv; 2788 } 2789 2790 static inline void netdev_set_ml_priv(struct net_device *dev, 2791 void *ml_priv, 2792 enum netdev_ml_priv_type type) 2793 { 2794 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2795 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2796 dev->ml_priv_type, type); 2797 WARN(!dev->ml_priv_type && dev->ml_priv, 2798 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2799 2800 dev->ml_priv = ml_priv; 2801 dev->ml_priv_type = type; 2802 } 2803 2804 /* 2805 * Net namespace inlines 2806 */ 2807 static inline 2808 struct net *dev_net(const struct net_device *dev) 2809 { 2810 return read_pnet(&dev->nd_net); 2811 } 2812 2813 static inline 2814 struct net *dev_net_rcu(const struct net_device *dev) 2815 { 2816 return read_pnet_rcu(&dev->nd_net); 2817 } 2818 2819 static inline 2820 void dev_net_set(struct net_device *dev, struct net *net) 2821 { 2822 write_pnet(&dev->nd_net, net); 2823 } 2824 2825 /** 2826 * netdev_priv - access network device private data 2827 * @dev: network device 2828 * 2829 * Get network device private data 2830 */ 2831 static inline void *netdev_priv(const struct net_device *dev) 2832 { 2833 return (void *)dev->priv; 2834 } 2835 2836 /** 2837 * netdev_from_priv() - get network device from priv 2838 * @priv: network device private data 2839 * 2840 * Returns: net_device to which @priv belongs 2841 */ 2842 static inline struct net_device *netdev_from_priv(const void *priv) 2843 { 2844 return container_of(priv, struct net_device, priv); 2845 } 2846 2847 /* Set the sysfs physical device reference for the network logical device 2848 * if set prior to registration will cause a symlink during initialization. 2849 */ 2850 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2851 2852 /* Set the sysfs device type for the network logical device to allow 2853 * fine-grained identification of different network device types. For 2854 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2855 */ 2856 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2857 2858 void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index, 2859 enum netdev_queue_type type, 2860 struct napi_struct *napi); 2861 2862 static inline void netdev_lock(struct net_device *dev) 2863 { 2864 mutex_lock(&dev->lock); 2865 } 2866 2867 static inline void netdev_unlock(struct net_device *dev) 2868 { 2869 mutex_unlock(&dev->lock); 2870 } 2871 /* Additional netdev_lock()-related helpers are in net/netdev_lock.h */ 2872 2873 void netif_napi_set_irq_locked(struct napi_struct *napi, int irq); 2874 2875 static inline void netif_napi_set_irq(struct napi_struct *napi, int irq) 2876 { 2877 netdev_lock(napi->dev); 2878 netif_napi_set_irq_locked(napi, irq); 2879 netdev_unlock(napi->dev); 2880 } 2881 2882 /* Default NAPI poll() weight 2883 * Device drivers are strongly advised to not use bigger value 2884 */ 2885 #define NAPI_POLL_WEIGHT 64 2886 2887 void netif_napi_add_weight_locked(struct net_device *dev, 2888 struct napi_struct *napi, 2889 int (*poll)(struct napi_struct *, int), 2890 int weight); 2891 2892 static inline void 2893 netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2894 int (*poll)(struct napi_struct *, int), int weight) 2895 { 2896 netdev_lock(dev); 2897 netif_napi_add_weight_locked(dev, napi, poll, weight); 2898 netdev_unlock(dev); 2899 } 2900 2901 /** 2902 * netif_napi_add() - initialize a NAPI context 2903 * @dev: network device 2904 * @napi: NAPI context 2905 * @poll: polling function 2906 * 2907 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2908 * *any* of the other NAPI-related functions. 2909 */ 2910 static inline void 2911 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2912 int (*poll)(struct napi_struct *, int)) 2913 { 2914 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2915 } 2916 2917 static inline void 2918 netif_napi_add_locked(struct net_device *dev, struct napi_struct *napi, 2919 int (*poll)(struct napi_struct *, int)) 2920 { 2921 netif_napi_add_weight_locked(dev, napi, poll, NAPI_POLL_WEIGHT); 2922 } 2923 2924 static inline void 2925 netif_napi_add_tx_weight(struct net_device *dev, 2926 struct napi_struct *napi, 2927 int (*poll)(struct napi_struct *, int), 2928 int weight) 2929 { 2930 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2931 netif_napi_add_weight(dev, napi, poll, weight); 2932 } 2933 2934 static inline void 2935 netif_napi_add_config_locked(struct net_device *dev, struct napi_struct *napi, 2936 int (*poll)(struct napi_struct *, int), int index) 2937 { 2938 napi->index = index; 2939 napi->config = &dev->napi_config[index]; 2940 netif_napi_add_weight_locked(dev, napi, poll, NAPI_POLL_WEIGHT); 2941 } 2942 2943 /** 2944 * netif_napi_add_config - initialize a NAPI context with persistent config 2945 * @dev: network device 2946 * @napi: NAPI context 2947 * @poll: polling function 2948 * @index: the NAPI index 2949 */ 2950 static inline void 2951 netif_napi_add_config(struct net_device *dev, struct napi_struct *napi, 2952 int (*poll)(struct napi_struct *, int), int index) 2953 { 2954 netdev_lock(dev); 2955 netif_napi_add_config_locked(dev, napi, poll, index); 2956 netdev_unlock(dev); 2957 } 2958 2959 /** 2960 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2961 * @dev: network device 2962 * @napi: NAPI context 2963 * @poll: polling function 2964 * 2965 * This variant of netif_napi_add() should be used from drivers using NAPI 2966 * to exclusively poll a TX queue. 2967 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2968 */ 2969 static inline void netif_napi_add_tx(struct net_device *dev, 2970 struct napi_struct *napi, 2971 int (*poll)(struct napi_struct *, int)) 2972 { 2973 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2974 } 2975 2976 void __netif_napi_del_locked(struct napi_struct *napi); 2977 2978 /** 2979 * __netif_napi_del - remove a NAPI context 2980 * @napi: NAPI context 2981 * 2982 * Warning: caller must observe RCU grace period before freeing memory 2983 * containing @napi. Drivers might want to call this helper to combine 2984 * all the needed RCU grace periods into a single one. 2985 */ 2986 static inline void __netif_napi_del(struct napi_struct *napi) 2987 { 2988 netdev_lock(napi->dev); 2989 __netif_napi_del_locked(napi); 2990 netdev_unlock(napi->dev); 2991 } 2992 2993 static inline void netif_napi_del_locked(struct napi_struct *napi) 2994 { 2995 __netif_napi_del_locked(napi); 2996 synchronize_net(); 2997 } 2998 2999 /** 3000 * netif_napi_del - remove a NAPI context 3001 * @napi: NAPI context 3002 * 3003 * netif_napi_del() removes a NAPI context from the network device NAPI list 3004 */ 3005 static inline void netif_napi_del(struct napi_struct *napi) 3006 { 3007 __netif_napi_del(napi); 3008 synchronize_net(); 3009 } 3010 3011 int netif_enable_cpu_rmap(struct net_device *dev, unsigned int num_irqs); 3012 void netif_set_affinity_auto(struct net_device *dev); 3013 3014 struct packet_type { 3015 __be16 type; /* This is really htons(ether_type). */ 3016 bool ignore_outgoing; 3017 struct net_device *dev; /* NULL is wildcarded here */ 3018 netdevice_tracker dev_tracker; 3019 int (*func) (struct sk_buff *, 3020 struct net_device *, 3021 struct packet_type *, 3022 struct net_device *); 3023 void (*list_func) (struct list_head *, 3024 struct packet_type *, 3025 struct net_device *); 3026 bool (*id_match)(struct packet_type *ptype, 3027 struct sock *sk); 3028 struct net *af_packet_net; 3029 void *af_packet_priv; 3030 struct list_head list; 3031 }; 3032 3033 struct offload_callbacks { 3034 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 3035 netdev_features_t features); 3036 struct sk_buff *(*gro_receive)(struct list_head *head, 3037 struct sk_buff *skb); 3038 int (*gro_complete)(struct sk_buff *skb, int nhoff); 3039 }; 3040 3041 struct packet_offload { 3042 __be16 type; /* This is really htons(ether_type). */ 3043 u16 priority; 3044 struct offload_callbacks callbacks; 3045 struct list_head list; 3046 }; 3047 3048 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 3049 struct pcpu_sw_netstats { 3050 u64_stats_t rx_packets; 3051 u64_stats_t rx_bytes; 3052 u64_stats_t tx_packets; 3053 u64_stats_t tx_bytes; 3054 struct u64_stats_sync syncp; 3055 } __aligned(4 * sizeof(u64)); 3056 3057 struct pcpu_dstats { 3058 u64_stats_t rx_packets; 3059 u64_stats_t rx_bytes; 3060 u64_stats_t tx_packets; 3061 u64_stats_t tx_bytes; 3062 u64_stats_t rx_drops; 3063 u64_stats_t tx_drops; 3064 struct u64_stats_sync syncp; 3065 } __aligned(8 * sizeof(u64)); 3066 3067 struct pcpu_lstats { 3068 u64_stats_t packets; 3069 u64_stats_t bytes; 3070 struct u64_stats_sync syncp; 3071 } __aligned(2 * sizeof(u64)); 3072 3073 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 3074 3075 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 3076 { 3077 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 3078 3079 u64_stats_update_begin(&tstats->syncp); 3080 u64_stats_add(&tstats->rx_bytes, len); 3081 u64_stats_inc(&tstats->rx_packets); 3082 u64_stats_update_end(&tstats->syncp); 3083 } 3084 3085 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 3086 unsigned int packets, 3087 unsigned int len) 3088 { 3089 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 3090 3091 u64_stats_update_begin(&tstats->syncp); 3092 u64_stats_add(&tstats->tx_bytes, len); 3093 u64_stats_add(&tstats->tx_packets, packets); 3094 u64_stats_update_end(&tstats->syncp); 3095 } 3096 3097 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 3098 { 3099 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 3100 3101 u64_stats_update_begin(&lstats->syncp); 3102 u64_stats_add(&lstats->bytes, len); 3103 u64_stats_inc(&lstats->packets); 3104 u64_stats_update_end(&lstats->syncp); 3105 } 3106 3107 static inline void dev_dstats_rx_add(struct net_device *dev, 3108 unsigned int len) 3109 { 3110 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3111 3112 u64_stats_update_begin(&dstats->syncp); 3113 u64_stats_inc(&dstats->rx_packets); 3114 u64_stats_add(&dstats->rx_bytes, len); 3115 u64_stats_update_end(&dstats->syncp); 3116 } 3117 3118 static inline void dev_dstats_rx_dropped(struct net_device *dev) 3119 { 3120 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3121 3122 u64_stats_update_begin(&dstats->syncp); 3123 u64_stats_inc(&dstats->rx_drops); 3124 u64_stats_update_end(&dstats->syncp); 3125 } 3126 3127 static inline void dev_dstats_rx_dropped_add(struct net_device *dev, 3128 unsigned int packets) 3129 { 3130 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3131 3132 u64_stats_update_begin(&dstats->syncp); 3133 u64_stats_add(&dstats->rx_drops, packets); 3134 u64_stats_update_end(&dstats->syncp); 3135 } 3136 3137 static inline void dev_dstats_tx_add(struct net_device *dev, 3138 unsigned int len) 3139 { 3140 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3141 3142 u64_stats_update_begin(&dstats->syncp); 3143 u64_stats_inc(&dstats->tx_packets); 3144 u64_stats_add(&dstats->tx_bytes, len); 3145 u64_stats_update_end(&dstats->syncp); 3146 } 3147 3148 static inline void dev_dstats_tx_dropped(struct net_device *dev) 3149 { 3150 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3151 3152 u64_stats_update_begin(&dstats->syncp); 3153 u64_stats_inc(&dstats->tx_drops); 3154 u64_stats_update_end(&dstats->syncp); 3155 } 3156 3157 #define __netdev_alloc_pcpu_stats(type, gfp) \ 3158 ({ \ 3159 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 3160 if (pcpu_stats) { \ 3161 int __cpu; \ 3162 for_each_possible_cpu(__cpu) { \ 3163 typeof(type) *stat; \ 3164 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 3165 u64_stats_init(&stat->syncp); \ 3166 } \ 3167 } \ 3168 pcpu_stats; \ 3169 }) 3170 3171 #define netdev_alloc_pcpu_stats(type) \ 3172 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 3173 3174 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 3175 ({ \ 3176 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 3177 if (pcpu_stats) { \ 3178 int __cpu; \ 3179 for_each_possible_cpu(__cpu) { \ 3180 typeof(type) *stat; \ 3181 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 3182 u64_stats_init(&stat->syncp); \ 3183 } \ 3184 } \ 3185 pcpu_stats; \ 3186 }) 3187 3188 enum netdev_lag_tx_type { 3189 NETDEV_LAG_TX_TYPE_UNKNOWN, 3190 NETDEV_LAG_TX_TYPE_RANDOM, 3191 NETDEV_LAG_TX_TYPE_BROADCAST, 3192 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 3193 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 3194 NETDEV_LAG_TX_TYPE_HASH, 3195 }; 3196 3197 enum netdev_lag_hash { 3198 NETDEV_LAG_HASH_NONE, 3199 NETDEV_LAG_HASH_L2, 3200 NETDEV_LAG_HASH_L34, 3201 NETDEV_LAG_HASH_L23, 3202 NETDEV_LAG_HASH_E23, 3203 NETDEV_LAG_HASH_E34, 3204 NETDEV_LAG_HASH_VLAN_SRCMAC, 3205 NETDEV_LAG_HASH_UNKNOWN, 3206 }; 3207 3208 struct netdev_lag_upper_info { 3209 enum netdev_lag_tx_type tx_type; 3210 enum netdev_lag_hash hash_type; 3211 }; 3212 3213 struct netdev_lag_lower_state_info { 3214 u8 link_up : 1, 3215 tx_enabled : 1; 3216 }; 3217 3218 #include <linux/notifier.h> 3219 3220 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 3221 * and the rtnetlink notification exclusion list in rtnetlink_event() when 3222 * adding new types. 3223 */ 3224 enum netdev_cmd { 3225 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 3226 NETDEV_DOWN, 3227 NETDEV_REBOOT, /* Tell a protocol stack a network interface 3228 detected a hardware crash and restarted 3229 - we can use this eg to kick tcp sessions 3230 once done */ 3231 NETDEV_CHANGE, /* Notify device state change */ 3232 NETDEV_REGISTER, 3233 NETDEV_UNREGISTER, 3234 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 3235 NETDEV_CHANGEADDR, /* notify after the address change */ 3236 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 3237 NETDEV_GOING_DOWN, 3238 NETDEV_CHANGENAME, 3239 NETDEV_FEAT_CHANGE, 3240 NETDEV_BONDING_FAILOVER, 3241 NETDEV_PRE_UP, 3242 NETDEV_PRE_TYPE_CHANGE, 3243 NETDEV_POST_TYPE_CHANGE, 3244 NETDEV_POST_INIT, 3245 NETDEV_PRE_UNINIT, 3246 NETDEV_RELEASE, 3247 NETDEV_NOTIFY_PEERS, 3248 NETDEV_JOIN, 3249 NETDEV_CHANGEUPPER, 3250 NETDEV_RESEND_IGMP, 3251 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 3252 NETDEV_CHANGEINFODATA, 3253 NETDEV_BONDING_INFO, 3254 NETDEV_PRECHANGEUPPER, 3255 NETDEV_CHANGELOWERSTATE, 3256 NETDEV_UDP_TUNNEL_PUSH_INFO, 3257 NETDEV_UDP_TUNNEL_DROP_INFO, 3258 NETDEV_CHANGE_TX_QUEUE_LEN, 3259 NETDEV_CVLAN_FILTER_PUSH_INFO, 3260 NETDEV_CVLAN_FILTER_DROP_INFO, 3261 NETDEV_SVLAN_FILTER_PUSH_INFO, 3262 NETDEV_SVLAN_FILTER_DROP_INFO, 3263 NETDEV_OFFLOAD_XSTATS_ENABLE, 3264 NETDEV_OFFLOAD_XSTATS_DISABLE, 3265 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 3266 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 3267 NETDEV_XDP_FEAT_CHANGE, 3268 }; 3269 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 3270 3271 int register_netdevice_notifier(struct notifier_block *nb); 3272 int unregister_netdevice_notifier(struct notifier_block *nb); 3273 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 3274 int unregister_netdevice_notifier_net(struct net *net, 3275 struct notifier_block *nb); 3276 int register_netdevice_notifier_dev_net(struct net_device *dev, 3277 struct notifier_block *nb, 3278 struct netdev_net_notifier *nn); 3279 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 3280 struct notifier_block *nb, 3281 struct netdev_net_notifier *nn); 3282 3283 struct netdev_notifier_info { 3284 struct net_device *dev; 3285 struct netlink_ext_ack *extack; 3286 }; 3287 3288 struct netdev_notifier_info_ext { 3289 struct netdev_notifier_info info; /* must be first */ 3290 union { 3291 u32 mtu; 3292 } ext; 3293 }; 3294 3295 struct netdev_notifier_change_info { 3296 struct netdev_notifier_info info; /* must be first */ 3297 unsigned int flags_changed; 3298 }; 3299 3300 struct netdev_notifier_changeupper_info { 3301 struct netdev_notifier_info info; /* must be first */ 3302 struct net_device *upper_dev; /* new upper dev */ 3303 bool master; /* is upper dev master */ 3304 bool linking; /* is the notification for link or unlink */ 3305 void *upper_info; /* upper dev info */ 3306 }; 3307 3308 struct netdev_notifier_changelowerstate_info { 3309 struct netdev_notifier_info info; /* must be first */ 3310 void *lower_state_info; /* is lower dev state */ 3311 }; 3312 3313 struct netdev_notifier_pre_changeaddr_info { 3314 struct netdev_notifier_info info; /* must be first */ 3315 const unsigned char *dev_addr; 3316 }; 3317 3318 enum netdev_offload_xstats_type { 3319 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 3320 }; 3321 3322 struct netdev_notifier_offload_xstats_info { 3323 struct netdev_notifier_info info; /* must be first */ 3324 enum netdev_offload_xstats_type type; 3325 3326 union { 3327 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 3328 struct netdev_notifier_offload_xstats_rd *report_delta; 3329 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 3330 struct netdev_notifier_offload_xstats_ru *report_used; 3331 }; 3332 }; 3333 3334 int netdev_offload_xstats_enable(struct net_device *dev, 3335 enum netdev_offload_xstats_type type, 3336 struct netlink_ext_ack *extack); 3337 int netdev_offload_xstats_disable(struct net_device *dev, 3338 enum netdev_offload_xstats_type type); 3339 bool netdev_offload_xstats_enabled(const struct net_device *dev, 3340 enum netdev_offload_xstats_type type); 3341 int netdev_offload_xstats_get(struct net_device *dev, 3342 enum netdev_offload_xstats_type type, 3343 struct rtnl_hw_stats64 *stats, bool *used, 3344 struct netlink_ext_ack *extack); 3345 void 3346 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 3347 const struct rtnl_hw_stats64 *stats); 3348 void 3349 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 3350 void netdev_offload_xstats_push_delta(struct net_device *dev, 3351 enum netdev_offload_xstats_type type, 3352 const struct rtnl_hw_stats64 *stats); 3353 3354 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 3355 struct net_device *dev) 3356 { 3357 info->dev = dev; 3358 info->extack = NULL; 3359 } 3360 3361 static inline struct net_device * 3362 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 3363 { 3364 return info->dev; 3365 } 3366 3367 static inline struct netlink_ext_ack * 3368 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 3369 { 3370 return info->extack; 3371 } 3372 3373 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 3374 int call_netdevice_notifiers_info(unsigned long val, 3375 struct netdev_notifier_info *info); 3376 3377 #define for_each_netdev(net, d) \ 3378 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 3379 #define for_each_netdev_reverse(net, d) \ 3380 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 3381 #define for_each_netdev_rcu(net, d) \ 3382 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 3383 #define for_each_netdev_safe(net, d, n) \ 3384 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 3385 #define for_each_netdev_continue(net, d) \ 3386 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 3387 #define for_each_netdev_continue_reverse(net, d) \ 3388 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 3389 dev_list) 3390 #define for_each_netdev_continue_rcu(net, d) \ 3391 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 3392 #define for_each_netdev_in_bond_rcu(bond, slave) \ 3393 for_each_netdev_rcu(dev_net_rcu(bond), slave) \ 3394 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 3395 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 3396 3397 #define for_each_netdev_dump(net, d, ifindex) \ 3398 for (; (d = xa_find(&(net)->dev_by_index, &ifindex, \ 3399 ULONG_MAX, XA_PRESENT)); ifindex++) 3400 3401 static inline struct net_device *next_net_device(struct net_device *dev) 3402 { 3403 struct list_head *lh; 3404 struct net *net; 3405 3406 net = dev_net(dev); 3407 lh = dev->dev_list.next; 3408 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3409 } 3410 3411 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 3412 { 3413 struct list_head *lh; 3414 struct net *net; 3415 3416 net = dev_net(dev); 3417 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 3418 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3419 } 3420 3421 static inline struct net_device *first_net_device(struct net *net) 3422 { 3423 return list_empty(&net->dev_base_head) ? NULL : 3424 net_device_entry(net->dev_base_head.next); 3425 } 3426 3427 struct net_device *dev_getbyhwaddr(struct net *net, unsigned short type, 3428 const char *hwaddr); 3429 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 3430 const char *hwaddr); 3431 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 3432 void dev_add_pack(struct packet_type *pt); 3433 void dev_remove_pack(struct packet_type *pt); 3434 void __dev_remove_pack(struct packet_type *pt); 3435 void dev_add_offload(struct packet_offload *po); 3436 void dev_remove_offload(struct packet_offload *po); 3437 3438 int dev_get_iflink(const struct net_device *dev); 3439 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 3440 int dev_fill_forward_path(struct net_device_path_ctx *ctx, 3441 struct net_device_path_stack *stack); 3442 void dev_fill_forward_path_release(struct net_device_path_stack *stack); 3443 struct net_device *dev_get_by_name(struct net *net, const char *name); 3444 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 3445 struct net_device *__dev_get_by_name(struct net *net, const char *name); 3446 bool netdev_name_in_use(struct net *net, const char *name); 3447 int dev_alloc_name(struct net_device *dev, const char *name); 3448 int netif_open(struct net_device *dev, struct netlink_ext_ack *extack); 3449 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 3450 void netif_close(struct net_device *dev); 3451 void dev_close(struct net_device *dev); 3452 void netif_close_many(struct list_head *head, bool unlink); 3453 void netif_disable_lro(struct net_device *dev); 3454 void dev_disable_lro(struct net_device *dev); 3455 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 3456 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 3457 struct net_device *sb_dev); 3458 3459 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3460 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3461 3462 static inline int dev_queue_xmit(struct sk_buff *skb) 3463 { 3464 return __dev_queue_xmit(skb, NULL); 3465 } 3466 3467 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3468 struct net_device *sb_dev) 3469 { 3470 return __dev_queue_xmit(skb, sb_dev); 3471 } 3472 3473 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3474 { 3475 int ret; 3476 3477 ret = __dev_direct_xmit(skb, queue_id); 3478 if (!dev_xmit_complete(ret)) 3479 kfree_skb(skb); 3480 return ret; 3481 } 3482 3483 int register_netdevice(struct net_device *dev); 3484 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3485 void unregister_netdevice_many(struct list_head *head); 3486 bool unregister_netdevice_queued(const struct net_device *dev); 3487 3488 static inline void unregister_netdevice(struct net_device *dev) 3489 { 3490 unregister_netdevice_queue(dev, NULL); 3491 } 3492 3493 #ifdef CONFIG_DEBUG_NET_SMALL_RTNL 3494 void unregister_netdevice_queue_net(struct net *net, struct net_device *dev, 3495 struct list_head *head); 3496 void unregister_netdevice_many_net(struct net *net); 3497 void unregister_netdevice_queue_many_net(struct net *net, struct list_head *head); 3498 #else 3499 static inline void unregister_netdevice_queue_net(struct net *net, 3500 struct net_device *dev, 3501 struct list_head *head) 3502 { 3503 unregister_netdevice_queue(dev, head); 3504 } 3505 3506 static inline void unregister_netdevice_queue_many_net(struct net *net, 3507 struct list_head *head) 3508 { 3509 } 3510 #endif 3511 3512 int netdev_refcnt_read(const struct net_device *dev); 3513 void free_netdev(struct net_device *dev); 3514 3515 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3516 struct sk_buff *skb, 3517 bool all_slaves); 3518 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3519 struct sock *sk); 3520 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3521 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3522 struct net_device *netdev_get_by_index(struct net *net, int ifindex, 3523 netdevice_tracker *tracker, gfp_t gfp); 3524 struct net_device *netdev_get_by_index_lock(struct net *net, int ifindex); 3525 struct net_device *netdev_get_by_name(struct net *net, const char *name, 3526 netdevice_tracker *tracker, gfp_t gfp); 3527 struct net_device *netdev_get_by_flags_rcu(struct net *net, netdevice_tracker *tracker, 3528 unsigned short flags, unsigned short mask); 3529 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3530 void netdev_copy_name(struct net_device *dev, char *name); 3531 3532 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3533 unsigned short type, 3534 const void *daddr, const void *saddr, 3535 unsigned int len) 3536 { 3537 if (!dev->header_ops || !dev->header_ops->create) 3538 return 0; 3539 3540 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3541 } 3542 3543 static inline int dev_parse_header(const struct sk_buff *skb, 3544 unsigned char *haddr) 3545 { 3546 const struct net_device *dev = skb->dev; 3547 3548 if (!dev->header_ops || !dev->header_ops->parse) 3549 return 0; 3550 return dev->header_ops->parse(skb, dev, haddr); 3551 } 3552 3553 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3554 { 3555 const struct net_device *dev = skb->dev; 3556 3557 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3558 return 0; 3559 return dev->header_ops->parse_protocol(skb); 3560 } 3561 3562 /* ll_header must have at least hard_header_len allocated */ 3563 static inline bool dev_validate_header(const struct net_device *dev, 3564 char *ll_header, int len) 3565 { 3566 if (likely(len >= dev->hard_header_len)) 3567 return true; 3568 if (len < dev->min_header_len) 3569 return false; 3570 3571 if (dev->header_ops && dev->header_ops->validate) 3572 return dev->header_ops->validate(ll_header, len); 3573 3574 return false; 3575 } 3576 3577 static inline bool dev_has_header(const struct net_device *dev) 3578 { 3579 return dev->header_ops && dev->header_ops->create; 3580 } 3581 3582 struct numa_drop_counters { 3583 atomic_t drops0 ____cacheline_aligned_in_smp; 3584 atomic_t drops1 ____cacheline_aligned_in_smp; 3585 }; 3586 3587 static inline int numa_drop_read(const struct numa_drop_counters *ndc) 3588 { 3589 return atomic_read(&ndc->drops0) + atomic_read(&ndc->drops1); 3590 } 3591 3592 static inline void numa_drop_add(struct numa_drop_counters *ndc, int val) 3593 { 3594 int n = numa_node_id() % 2; 3595 3596 if (n) 3597 atomic_add(val, &ndc->drops1); 3598 else 3599 atomic_add(val, &ndc->drops0); 3600 } 3601 3602 static inline void numa_drop_reset(struct numa_drop_counters *ndc) 3603 { 3604 atomic_set(&ndc->drops0, 0); 3605 atomic_set(&ndc->drops1, 0); 3606 } 3607 3608 /* 3609 * Incoming packets are placed on per-CPU queues 3610 */ 3611 struct softnet_data { 3612 struct list_head poll_list; 3613 struct sk_buff_head process_queue; 3614 local_lock_t process_queue_bh_lock; 3615 3616 /* stats */ 3617 unsigned int processed; 3618 unsigned int time_squeeze; 3619 #ifdef CONFIG_RPS 3620 struct softnet_data *rps_ipi_list; 3621 #endif 3622 3623 unsigned int received_rps; 3624 bool in_net_rx_action; 3625 bool in_napi_threaded_poll; 3626 3627 #ifdef CONFIG_NET_FLOW_LIMIT 3628 struct sd_flow_limit __rcu *flow_limit; 3629 #endif 3630 struct Qdisc *output_queue; 3631 struct Qdisc **output_queue_tailp; 3632 struct sk_buff *completion_queue; 3633 #ifdef CONFIG_XFRM_OFFLOAD 3634 struct sk_buff_head xfrm_backlog; 3635 #endif 3636 /* written and read only by owning cpu: */ 3637 struct netdev_xmit xmit; 3638 #ifdef CONFIG_RPS 3639 /* input_queue_head should be written by cpu owning this struct, 3640 * and only read by other cpus. Worth using a cache line. 3641 */ 3642 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3643 3644 /* Elements below can be accessed between CPUs for RPS/RFS */ 3645 call_single_data_t csd ____cacheline_aligned_in_smp; 3646 struct softnet_data *rps_ipi_next; 3647 unsigned int cpu; 3648 3649 /* We force a cacheline alignment from here, to hold together 3650 * input_queue_tail, input_pkt_queue and backlog.state. 3651 * We add holes so that backlog.state is the last field 3652 * of this cache line. 3653 */ 3654 long pad[3] ____cacheline_aligned_in_smp; 3655 unsigned int input_queue_tail; 3656 #endif 3657 struct sk_buff_head input_pkt_queue; 3658 3659 struct napi_struct backlog; 3660 3661 struct numa_drop_counters drop_counters; 3662 3663 int defer_ipi_scheduled ____cacheline_aligned_in_smp; 3664 call_single_data_t defer_csd; 3665 }; 3666 3667 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3668 3669 struct page_pool_bh { 3670 struct page_pool *pool; 3671 local_lock_t bh_lock; 3672 }; 3673 DECLARE_PER_CPU(struct page_pool_bh, system_page_pool); 3674 3675 #define XMIT_RECURSION_LIMIT 8 3676 3677 #ifndef CONFIG_PREEMPT_RT 3678 static inline int dev_recursion_level(void) 3679 { 3680 return this_cpu_read(softnet_data.xmit.recursion); 3681 } 3682 3683 static inline bool dev_xmit_recursion(void) 3684 { 3685 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3686 XMIT_RECURSION_LIMIT); 3687 } 3688 3689 static inline void dev_xmit_recursion_inc(void) 3690 { 3691 __this_cpu_inc(softnet_data.xmit.recursion); 3692 } 3693 3694 static inline void dev_xmit_recursion_dec(void) 3695 { 3696 __this_cpu_dec(softnet_data.xmit.recursion); 3697 } 3698 #else 3699 static inline int dev_recursion_level(void) 3700 { 3701 return current->net_xmit.recursion; 3702 } 3703 3704 static inline bool dev_xmit_recursion(void) 3705 { 3706 return unlikely(current->net_xmit.recursion > XMIT_RECURSION_LIMIT); 3707 } 3708 3709 static inline void dev_xmit_recursion_inc(void) 3710 { 3711 current->net_xmit.recursion++; 3712 } 3713 3714 static inline void dev_xmit_recursion_dec(void) 3715 { 3716 current->net_xmit.recursion--; 3717 } 3718 #endif 3719 3720 void __netif_schedule(struct Qdisc *q); 3721 void netif_schedule_queue(struct netdev_queue *txq); 3722 3723 static inline void netif_tx_schedule_all(struct net_device *dev) 3724 { 3725 unsigned int i; 3726 3727 for (i = 0; i < dev->num_tx_queues; i++) 3728 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3729 } 3730 3731 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3732 { 3733 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3734 } 3735 3736 /** 3737 * netif_start_queue - allow transmit 3738 * @dev: network device 3739 * 3740 * Allow upper layers to call the device hard_start_xmit routine. 3741 */ 3742 static inline void netif_start_queue(struct net_device *dev) 3743 { 3744 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3745 } 3746 3747 static inline void netif_tx_start_all_queues(struct net_device *dev) 3748 { 3749 unsigned int i; 3750 3751 for (i = 0; i < dev->num_tx_queues; i++) { 3752 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3753 netif_tx_start_queue(txq); 3754 } 3755 } 3756 3757 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3758 3759 /** 3760 * netif_wake_queue - restart transmit 3761 * @dev: network device 3762 * 3763 * Allow upper layers to call the device hard_start_xmit routine. 3764 * Used for flow control when transmit resources are available. 3765 */ 3766 static inline void netif_wake_queue(struct net_device *dev) 3767 { 3768 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3769 } 3770 3771 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3772 { 3773 unsigned int i; 3774 3775 for (i = 0; i < dev->num_tx_queues; i++) { 3776 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3777 netif_tx_wake_queue(txq); 3778 } 3779 } 3780 3781 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3782 { 3783 /* Paired with READ_ONCE() from dev_watchdog() */ 3784 WRITE_ONCE(dev_queue->trans_start, jiffies); 3785 3786 /* This barrier is paired with smp_mb() from dev_watchdog() */ 3787 smp_mb__before_atomic(); 3788 3789 /* Must be an atomic op see netif_txq_try_stop() */ 3790 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3791 } 3792 3793 /** 3794 * netif_stop_queue - stop transmitted packets 3795 * @dev: network device 3796 * 3797 * Stop upper layers calling the device hard_start_xmit routine. 3798 * Used for flow control when transmit resources are unavailable. 3799 */ 3800 static inline void netif_stop_queue(struct net_device *dev) 3801 { 3802 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3803 } 3804 3805 void netif_tx_stop_all_queues(struct net_device *dev); 3806 3807 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3808 { 3809 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3810 } 3811 3812 /** 3813 * netif_queue_stopped - test if transmit queue is flowblocked 3814 * @dev: network device 3815 * 3816 * Test if transmit queue on device is currently unable to send. 3817 */ 3818 static inline bool netif_queue_stopped(const struct net_device *dev) 3819 { 3820 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3821 } 3822 3823 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3824 { 3825 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3826 } 3827 3828 static inline bool 3829 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3830 { 3831 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3832 } 3833 3834 static inline bool 3835 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3836 { 3837 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3838 } 3839 3840 /** 3841 * netdev_queue_set_dql_min_limit - set dql minimum limit 3842 * @dev_queue: pointer to transmit queue 3843 * @min_limit: dql minimum limit 3844 * 3845 * Forces xmit_more() to return true until the minimum threshold 3846 * defined by @min_limit is reached (or until the tx queue is 3847 * empty). Warning: to be use with care, misuse will impact the 3848 * latency. 3849 */ 3850 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3851 unsigned int min_limit) 3852 { 3853 #ifdef CONFIG_BQL 3854 dev_queue->dql.min_limit = min_limit; 3855 #endif 3856 } 3857 3858 static inline int netdev_queue_dql_avail(const struct netdev_queue *txq) 3859 { 3860 #ifdef CONFIG_BQL 3861 /* Non-BQL migrated drivers will return 0, too. */ 3862 return dql_avail(&txq->dql); 3863 #else 3864 return 0; 3865 #endif 3866 } 3867 3868 /** 3869 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3870 * @dev_queue: pointer to transmit queue 3871 * 3872 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3873 * to give appropriate hint to the CPU. 3874 */ 3875 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3876 { 3877 #ifdef CONFIG_BQL 3878 prefetchw(&dev_queue->dql.num_queued); 3879 #endif 3880 } 3881 3882 /** 3883 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3884 * @dev_queue: pointer to transmit queue 3885 * 3886 * BQL enabled drivers might use this helper in their TX completion path, 3887 * to give appropriate hint to the CPU. 3888 */ 3889 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3890 { 3891 #ifdef CONFIG_BQL 3892 prefetchw(&dev_queue->dql.limit); 3893 #endif 3894 } 3895 3896 /** 3897 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3898 * @dev_queue: network device queue 3899 * @bytes: number of bytes queued to the device queue 3900 * 3901 * Report the number of bytes queued for sending/completion to the network 3902 * device hardware queue. @bytes should be a good approximation and should 3903 * exactly match netdev_completed_queue() @bytes. 3904 * This is typically called once per packet, from ndo_start_xmit(). 3905 */ 3906 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3907 unsigned int bytes) 3908 { 3909 #ifdef CONFIG_BQL 3910 dql_queued(&dev_queue->dql, bytes); 3911 3912 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3913 return; 3914 3915 /* Paired with READ_ONCE() from dev_watchdog() */ 3916 WRITE_ONCE(dev_queue->trans_start, jiffies); 3917 3918 /* This barrier is paired with smp_mb() from dev_watchdog() */ 3919 smp_mb__before_atomic(); 3920 3921 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3922 3923 /* 3924 * The XOFF flag must be set before checking the dql_avail below, 3925 * because in netdev_tx_completed_queue we update the dql_completed 3926 * before checking the XOFF flag. 3927 */ 3928 smp_mb__after_atomic(); 3929 3930 /* check again in case another CPU has just made room avail */ 3931 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3932 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3933 #endif 3934 } 3935 3936 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3937 * that they should not test BQL status themselves. 3938 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3939 * skb of a batch. 3940 * Returns true if the doorbell must be used to kick the NIC. 3941 */ 3942 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3943 unsigned int bytes, 3944 bool xmit_more) 3945 { 3946 if (xmit_more) { 3947 #ifdef CONFIG_BQL 3948 dql_queued(&dev_queue->dql, bytes); 3949 #endif 3950 return netif_tx_queue_stopped(dev_queue); 3951 } 3952 netdev_tx_sent_queue(dev_queue, bytes); 3953 return true; 3954 } 3955 3956 /** 3957 * netdev_sent_queue - report the number of bytes queued to hardware 3958 * @dev: network device 3959 * @bytes: number of bytes queued to the hardware device queue 3960 * 3961 * Report the number of bytes queued for sending/completion to the network 3962 * device hardware queue#0. @bytes should be a good approximation and should 3963 * exactly match netdev_completed_queue() @bytes. 3964 * This is typically called once per packet, from ndo_start_xmit(). 3965 */ 3966 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3967 { 3968 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3969 } 3970 3971 static inline bool __netdev_sent_queue(struct net_device *dev, 3972 unsigned int bytes, 3973 bool xmit_more) 3974 { 3975 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3976 xmit_more); 3977 } 3978 3979 /** 3980 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3981 * @dev_queue: network device queue 3982 * @pkts: number of packets (currently ignored) 3983 * @bytes: number of bytes dequeued from the device queue 3984 * 3985 * Must be called at most once per TX completion round (and not per 3986 * individual packet), so that BQL can adjust its limits appropriately. 3987 */ 3988 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3989 unsigned int pkts, unsigned int bytes) 3990 { 3991 #ifdef CONFIG_BQL 3992 if (unlikely(!bytes)) 3993 return; 3994 3995 dql_completed(&dev_queue->dql, bytes); 3996 3997 /* 3998 * Without the memory barrier there is a small possibility that 3999 * netdev_tx_sent_queue will miss the update and cause the queue to 4000 * be stopped forever 4001 */ 4002 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */ 4003 4004 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 4005 return; 4006 4007 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 4008 netif_schedule_queue(dev_queue); 4009 #endif 4010 } 4011 4012 /** 4013 * netdev_completed_queue - report bytes and packets completed by device 4014 * @dev: network device 4015 * @pkts: actual number of packets sent over the medium 4016 * @bytes: actual number of bytes sent over the medium 4017 * 4018 * Report the number of bytes and packets transmitted by the network device 4019 * hardware queue over the physical medium, @bytes must exactly match the 4020 * @bytes amount passed to netdev_sent_queue() 4021 */ 4022 static inline void netdev_completed_queue(struct net_device *dev, 4023 unsigned int pkts, unsigned int bytes) 4024 { 4025 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 4026 } 4027 4028 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 4029 { 4030 #ifdef CONFIG_BQL 4031 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 4032 dql_reset(&q->dql); 4033 #endif 4034 } 4035 4036 /** 4037 * netdev_tx_reset_subqueue - reset the BQL stats and state of a netdev queue 4038 * @dev: network device 4039 * @qid: stack index of the queue to reset 4040 */ 4041 static inline void netdev_tx_reset_subqueue(const struct net_device *dev, 4042 u32 qid) 4043 { 4044 netdev_tx_reset_queue(netdev_get_tx_queue(dev, qid)); 4045 } 4046 4047 /** 4048 * netdev_reset_queue - reset the packets and bytes count of a network device 4049 * @dev_queue: network device 4050 * 4051 * Reset the bytes and packet count of a network device and clear the 4052 * software flow control OFF bit for this network device 4053 */ 4054 static inline void netdev_reset_queue(struct net_device *dev_queue) 4055 { 4056 netdev_tx_reset_subqueue(dev_queue, 0); 4057 } 4058 4059 /** 4060 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 4061 * @dev: network device 4062 * @queue_index: given tx queue index 4063 * 4064 * Returns 0 if given tx queue index >= number of device tx queues, 4065 * otherwise returns the originally passed tx queue index. 4066 */ 4067 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 4068 { 4069 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 4070 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 4071 dev->name, queue_index, 4072 dev->real_num_tx_queues); 4073 return 0; 4074 } 4075 4076 return queue_index; 4077 } 4078 4079 /** 4080 * netif_running - test if up 4081 * @dev: network device 4082 * 4083 * Test if the device has been brought up. 4084 */ 4085 static inline bool netif_running(const struct net_device *dev) 4086 { 4087 return test_bit(__LINK_STATE_START, &dev->state); 4088 } 4089 4090 /* 4091 * Routines to manage the subqueues on a device. We only need start, 4092 * stop, and a check if it's stopped. All other device management is 4093 * done at the overall netdevice level. 4094 * Also test the device if we're multiqueue. 4095 */ 4096 4097 /** 4098 * netif_start_subqueue - allow sending packets on subqueue 4099 * @dev: network device 4100 * @queue_index: sub queue index 4101 * 4102 * Start individual transmit queue of a device with multiple transmit queues. 4103 */ 4104 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 4105 { 4106 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 4107 4108 netif_tx_start_queue(txq); 4109 } 4110 4111 /** 4112 * netif_stop_subqueue - stop sending packets on subqueue 4113 * @dev: network device 4114 * @queue_index: sub queue index 4115 * 4116 * Stop individual transmit queue of a device with multiple transmit queues. 4117 */ 4118 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 4119 { 4120 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 4121 netif_tx_stop_queue(txq); 4122 } 4123 4124 /** 4125 * __netif_subqueue_stopped - test status of subqueue 4126 * @dev: network device 4127 * @queue_index: sub queue index 4128 * 4129 * Check individual transmit queue of a device with multiple transmit queues. 4130 */ 4131 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 4132 u16 queue_index) 4133 { 4134 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 4135 4136 return netif_tx_queue_stopped(txq); 4137 } 4138 4139 /** 4140 * netif_subqueue_stopped - test status of subqueue 4141 * @dev: network device 4142 * @skb: sub queue buffer pointer 4143 * 4144 * Check individual transmit queue of a device with multiple transmit queues. 4145 */ 4146 static inline bool netif_subqueue_stopped(const struct net_device *dev, 4147 struct sk_buff *skb) 4148 { 4149 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 4150 } 4151 4152 /** 4153 * netif_wake_subqueue - allow sending packets on subqueue 4154 * @dev: network device 4155 * @queue_index: sub queue index 4156 * 4157 * Resume individual transmit queue of a device with multiple transmit queues. 4158 */ 4159 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 4160 { 4161 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 4162 4163 netif_tx_wake_queue(txq); 4164 } 4165 4166 #ifdef CONFIG_XPS 4167 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 4168 u16 index); 4169 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 4170 u16 index, enum xps_map_type type); 4171 4172 /** 4173 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 4174 * @j: CPU/Rx queue index 4175 * @mask: bitmask of all cpus/rx queues 4176 * @nr_bits: number of bits in the bitmask 4177 * 4178 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 4179 */ 4180 static inline bool netif_attr_test_mask(unsigned long j, 4181 const unsigned long *mask, 4182 unsigned int nr_bits) 4183 { 4184 cpu_max_bits_warn(j, nr_bits); 4185 return test_bit(j, mask); 4186 } 4187 4188 /** 4189 * netif_attr_test_online - Test for online CPU/Rx queue 4190 * @j: CPU/Rx queue index 4191 * @online_mask: bitmask for CPUs/Rx queues that are online 4192 * @nr_bits: number of bits in the bitmask 4193 * 4194 * Returns: true if a CPU/Rx queue is online. 4195 */ 4196 static inline bool netif_attr_test_online(unsigned long j, 4197 const unsigned long *online_mask, 4198 unsigned int nr_bits) 4199 { 4200 cpu_max_bits_warn(j, nr_bits); 4201 4202 if (online_mask) 4203 return test_bit(j, online_mask); 4204 4205 return (j < nr_bits); 4206 } 4207 4208 /** 4209 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 4210 * @n: CPU/Rx queue index 4211 * @srcp: the cpumask/Rx queue mask pointer 4212 * @nr_bits: number of bits in the bitmask 4213 * 4214 * Returns: next (after n) CPU/Rx queue index in the mask; 4215 * >= nr_bits if no further CPUs/Rx queues set. 4216 */ 4217 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 4218 unsigned int nr_bits) 4219 { 4220 /* -1 is a legal arg here. */ 4221 if (n != -1) 4222 cpu_max_bits_warn(n, nr_bits); 4223 4224 if (srcp) 4225 return find_next_bit(srcp, nr_bits, n + 1); 4226 4227 return n + 1; 4228 } 4229 4230 /** 4231 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 4232 * @n: CPU/Rx queue index 4233 * @src1p: the first CPUs/Rx queues mask pointer 4234 * @src2p: the second CPUs/Rx queues mask pointer 4235 * @nr_bits: number of bits in the bitmask 4236 * 4237 * Returns: next (after n) CPU/Rx queue index set in both masks; 4238 * >= nr_bits if no further CPUs/Rx queues set in both. 4239 */ 4240 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 4241 const unsigned long *src2p, 4242 unsigned int nr_bits) 4243 { 4244 /* -1 is a legal arg here. */ 4245 if (n != -1) 4246 cpu_max_bits_warn(n, nr_bits); 4247 4248 if (src1p && src2p) 4249 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 4250 else if (src1p) 4251 return find_next_bit(src1p, nr_bits, n + 1); 4252 else if (src2p) 4253 return find_next_bit(src2p, nr_bits, n + 1); 4254 4255 return n + 1; 4256 } 4257 #else 4258 static inline int netif_set_xps_queue(struct net_device *dev, 4259 const struct cpumask *mask, 4260 u16 index) 4261 { 4262 return 0; 4263 } 4264 4265 static inline int __netif_set_xps_queue(struct net_device *dev, 4266 const unsigned long *mask, 4267 u16 index, enum xps_map_type type) 4268 { 4269 return 0; 4270 } 4271 #endif 4272 4273 /** 4274 * netif_is_multiqueue - test if device has multiple transmit queues 4275 * @dev: network device 4276 * 4277 * Check if device has multiple transmit queues 4278 */ 4279 static inline bool netif_is_multiqueue(const struct net_device *dev) 4280 { 4281 return dev->num_tx_queues > 1; 4282 } 4283 4284 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 4285 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 4286 int netif_set_real_num_queues(struct net_device *dev, 4287 unsigned int txq, unsigned int rxq); 4288 4289 int netif_get_num_default_rss_queues(void); 4290 4291 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason); 4292 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason); 4293 4294 /* 4295 * It is not allowed to call kfree_skb() or consume_skb() from hardware 4296 * interrupt context or with hardware interrupts being disabled. 4297 * (in_hardirq() || irqs_disabled()) 4298 * 4299 * We provide four helpers that can be used in following contexts : 4300 * 4301 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 4302 * replacing kfree_skb(skb) 4303 * 4304 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 4305 * Typically used in place of consume_skb(skb) in TX completion path 4306 * 4307 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 4308 * replacing kfree_skb(skb) 4309 * 4310 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 4311 * and consumed a packet. Used in place of consume_skb(skb) 4312 */ 4313 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 4314 { 4315 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 4316 } 4317 4318 static inline void dev_consume_skb_irq(struct sk_buff *skb) 4319 { 4320 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED); 4321 } 4322 4323 static inline void dev_kfree_skb_any(struct sk_buff *skb) 4324 { 4325 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 4326 } 4327 4328 static inline void dev_consume_skb_any(struct sk_buff *skb) 4329 { 4330 dev_kfree_skb_any_reason(skb, SKB_CONSUMED); 4331 } 4332 4333 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 4334 const struct bpf_prog *xdp_prog); 4335 void generic_xdp_tx(struct sk_buff *skb, const struct bpf_prog *xdp_prog); 4336 int do_xdp_generic(const struct bpf_prog *xdp_prog, struct sk_buff **pskb); 4337 int netif_rx(struct sk_buff *skb); 4338 int __netif_rx(struct sk_buff *skb); 4339 4340 int netif_receive_skb(struct sk_buff *skb); 4341 int netif_receive_skb_core(struct sk_buff *skb); 4342 void netif_receive_skb_list_internal(struct list_head *head); 4343 void netif_receive_skb_list(struct list_head *head); 4344 gro_result_t gro_receive_skb(struct gro_node *gro, struct sk_buff *skb); 4345 4346 static inline gro_result_t napi_gro_receive(struct napi_struct *napi, 4347 struct sk_buff *skb) 4348 { 4349 return gro_receive_skb(&napi->gro, skb); 4350 } 4351 4352 struct sk_buff *napi_get_frags(struct napi_struct *napi); 4353 gro_result_t napi_gro_frags(struct napi_struct *napi); 4354 4355 static inline void napi_free_frags(struct napi_struct *napi) 4356 { 4357 kfree_skb(napi->skb); 4358 napi->skb = NULL; 4359 } 4360 4361 bool netdev_is_rx_handler_busy(struct net_device *dev); 4362 int netdev_rx_handler_register(struct net_device *dev, 4363 rx_handler_func_t *rx_handler, 4364 void *rx_handler_data); 4365 void netdev_rx_handler_unregister(struct net_device *dev); 4366 4367 bool dev_valid_name(const char *name); 4368 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 4369 { 4370 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 4371 } 4372 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 4373 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 4374 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 4375 void __user *data, bool *need_copyout); 4376 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 4377 int dev_eth_ioctl(struct net_device *dev, 4378 struct ifreq *ifr, unsigned int cmd); 4379 int generic_hwtstamp_get_lower(struct net_device *dev, 4380 struct kernel_hwtstamp_config *kernel_cfg); 4381 int generic_hwtstamp_set_lower(struct net_device *dev, 4382 struct kernel_hwtstamp_config *kernel_cfg, 4383 struct netlink_ext_ack *extack); 4384 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 4385 unsigned int netif_get_flags(const struct net_device *dev); 4386 int __dev_change_flags(struct net_device *dev, unsigned int flags, 4387 struct netlink_ext_ack *extack); 4388 int netif_change_flags(struct net_device *dev, unsigned int flags, 4389 struct netlink_ext_ack *extack); 4390 int dev_change_flags(struct net_device *dev, unsigned int flags, 4391 struct netlink_ext_ack *extack); 4392 int netif_set_alias(struct net_device *dev, const char *alias, size_t len); 4393 int dev_set_alias(struct net_device *, const char *, size_t); 4394 int dev_get_alias(const struct net_device *, char *, size_t); 4395 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 4396 const char *pat, int new_ifindex, 4397 struct netlink_ext_ack *extack); 4398 int dev_change_net_namespace(struct net_device *dev, struct net *net, 4399 const char *pat); 4400 int __netif_set_mtu(struct net_device *dev, int new_mtu); 4401 int netif_set_mtu(struct net_device *dev, int new_mtu); 4402 int dev_set_mtu(struct net_device *, int); 4403 int netif_pre_changeaddr_notify(struct net_device *dev, const char *addr, 4404 struct netlink_ext_ack *extack); 4405 int netif_set_mac_address(struct net_device *dev, struct sockaddr_storage *ss, 4406 struct netlink_ext_ack *extack); 4407 int dev_set_mac_address(struct net_device *dev, struct sockaddr_storage *ss, 4408 struct netlink_ext_ack *extack); 4409 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr_storage *ss, 4410 struct netlink_ext_ack *extack); 4411 int netif_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 4412 int netif_get_port_parent_id(struct net_device *dev, 4413 struct netdev_phys_item_id *ppid, bool recurse); 4414 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 4415 4416 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 4417 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 4418 struct netdev_queue *txq, int *ret); 4419 4420 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 4421 u8 dev_xdp_prog_count(struct net_device *dev); 4422 int netif_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf); 4423 int dev_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf); 4424 u8 dev_xdp_sb_prog_count(struct net_device *dev); 4425 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 4426 4427 u32 dev_get_min_mp_channel_count(const struct net_device *dev); 4428 4429 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4430 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4431 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 4432 bool is_skb_forwardable(const struct net_device *dev, 4433 const struct sk_buff *skb); 4434 4435 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 4436 const struct sk_buff *skb, 4437 const bool check_mtu) 4438 { 4439 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 4440 unsigned int len; 4441 4442 if (!(dev->flags & IFF_UP)) 4443 return false; 4444 4445 if (!check_mtu) 4446 return true; 4447 4448 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 4449 if (skb->len <= len) 4450 return true; 4451 4452 /* if TSO is enabled, we don't care about the length as the packet 4453 * could be forwarded without being segmented before 4454 */ 4455 if (skb_is_gso(skb)) 4456 return true; 4457 4458 return false; 4459 } 4460 4461 void netdev_core_stats_inc(struct net_device *dev, u32 offset); 4462 4463 #define DEV_CORE_STATS_INC(FIELD) \ 4464 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 4465 { \ 4466 netdev_core_stats_inc(dev, \ 4467 offsetof(struct net_device_core_stats, FIELD)); \ 4468 } 4469 DEV_CORE_STATS_INC(rx_dropped) 4470 DEV_CORE_STATS_INC(tx_dropped) 4471 DEV_CORE_STATS_INC(rx_nohandler) 4472 DEV_CORE_STATS_INC(rx_otherhost_dropped) 4473 #undef DEV_CORE_STATS_INC 4474 4475 static __always_inline int ____dev_forward_skb(struct net_device *dev, 4476 struct sk_buff *skb, 4477 const bool check_mtu) 4478 { 4479 if (skb_orphan_frags(skb, GFP_ATOMIC) || 4480 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 4481 dev_core_stats_rx_dropped_inc(dev); 4482 kfree_skb(skb); 4483 return NET_RX_DROP; 4484 } 4485 4486 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 4487 skb->priority = 0; 4488 return 0; 4489 } 4490 4491 bool dev_nit_active_rcu(const struct net_device *dev); 4492 static inline bool dev_nit_active(const struct net_device *dev) 4493 { 4494 bool ret; 4495 4496 rcu_read_lock(); 4497 ret = dev_nit_active_rcu(dev); 4498 rcu_read_unlock(); 4499 return ret; 4500 } 4501 4502 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 4503 4504 static inline void __dev_put(struct net_device *dev) 4505 { 4506 if (dev) { 4507 #ifdef CONFIG_PCPU_DEV_REFCNT 4508 this_cpu_dec(*dev->pcpu_refcnt); 4509 #else 4510 refcount_dec(&dev->dev_refcnt); 4511 #endif 4512 } 4513 } 4514 4515 static inline void __dev_hold(struct net_device *dev) 4516 { 4517 if (dev) { 4518 #ifdef CONFIG_PCPU_DEV_REFCNT 4519 this_cpu_inc(*dev->pcpu_refcnt); 4520 #else 4521 refcount_inc(&dev->dev_refcnt); 4522 #endif 4523 } 4524 } 4525 4526 static inline void __netdev_tracker_alloc(struct net_device *dev, 4527 netdevice_tracker *tracker, 4528 gfp_t gfp) 4529 { 4530 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4531 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 4532 #endif 4533 } 4534 4535 /* netdev_tracker_alloc() can upgrade a prior untracked reference 4536 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4537 */ 4538 static inline void netdev_tracker_alloc(struct net_device *dev, 4539 netdevice_tracker *tracker, gfp_t gfp) 4540 { 4541 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4542 refcount_dec(&dev->refcnt_tracker.no_tracker); 4543 __netdev_tracker_alloc(dev, tracker, gfp); 4544 #endif 4545 } 4546 4547 static inline void netdev_tracker_free(struct net_device *dev, 4548 netdevice_tracker *tracker) 4549 { 4550 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4551 ref_tracker_free(&dev->refcnt_tracker, tracker); 4552 #endif 4553 } 4554 4555 static inline void netdev_hold(struct net_device *dev, 4556 netdevice_tracker *tracker, gfp_t gfp) 4557 { 4558 if (dev) { 4559 __dev_hold(dev); 4560 __netdev_tracker_alloc(dev, tracker, gfp); 4561 } 4562 } 4563 4564 static inline void netdev_put(struct net_device *dev, 4565 netdevice_tracker *tracker) 4566 { 4567 if (dev) { 4568 netdev_tracker_free(dev, tracker); 4569 __dev_put(dev); 4570 } 4571 } 4572 4573 /** 4574 * dev_hold - get reference to device 4575 * @dev: network device 4576 * 4577 * Hold reference to device to keep it from being freed. 4578 * Try using netdev_hold() instead. 4579 */ 4580 static inline void dev_hold(struct net_device *dev) 4581 { 4582 netdev_hold(dev, NULL, GFP_ATOMIC); 4583 } 4584 4585 /** 4586 * dev_put - release reference to device 4587 * @dev: network device 4588 * 4589 * Release reference to device to allow it to be freed. 4590 * Try using netdev_put() instead. 4591 */ 4592 static inline void dev_put(struct net_device *dev) 4593 { 4594 netdev_put(dev, NULL); 4595 } 4596 4597 DEFINE_FREE(dev_put, struct net_device *, if (_T) dev_put(_T)) 4598 4599 static inline void netdev_ref_replace(struct net_device *odev, 4600 struct net_device *ndev, 4601 netdevice_tracker *tracker, 4602 gfp_t gfp) 4603 { 4604 if (odev) 4605 netdev_tracker_free(odev, tracker); 4606 4607 __dev_hold(ndev); 4608 __dev_put(odev); 4609 4610 if (ndev) 4611 __netdev_tracker_alloc(ndev, tracker, gfp); 4612 } 4613 4614 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4615 * and _off may be called from IRQ context, but it is caller 4616 * who is responsible for serialization of these calls. 4617 * 4618 * The name carrier is inappropriate, these functions should really be 4619 * called netif_lowerlayer_*() because they represent the state of any 4620 * kind of lower layer not just hardware media. 4621 */ 4622 void linkwatch_fire_event(struct net_device *dev); 4623 4624 /** 4625 * linkwatch_sync_dev - sync linkwatch for the given device 4626 * @dev: network device to sync linkwatch for 4627 * 4628 * Sync linkwatch for the given device, removing it from the 4629 * pending work list (if queued). 4630 */ 4631 void linkwatch_sync_dev(struct net_device *dev); 4632 void __linkwatch_sync_dev(struct net_device *dev); 4633 4634 /** 4635 * netif_carrier_ok - test if carrier present 4636 * @dev: network device 4637 * 4638 * Check if carrier is present on device 4639 */ 4640 static inline bool netif_carrier_ok(const struct net_device *dev) 4641 { 4642 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4643 } 4644 4645 unsigned long dev_trans_start(struct net_device *dev); 4646 4647 void netdev_watchdog_up(struct net_device *dev); 4648 4649 void netif_carrier_on(struct net_device *dev); 4650 void netif_carrier_off(struct net_device *dev); 4651 void netif_carrier_event(struct net_device *dev); 4652 4653 /** 4654 * netif_dormant_on - mark device as dormant. 4655 * @dev: network device 4656 * 4657 * Mark device as dormant (as per RFC2863). 4658 * 4659 * The dormant state indicates that the relevant interface is not 4660 * actually in a condition to pass packets (i.e., it is not 'up') but is 4661 * in a "pending" state, waiting for some external event. For "on- 4662 * demand" interfaces, this new state identifies the situation where the 4663 * interface is waiting for events to place it in the up state. 4664 */ 4665 static inline void netif_dormant_on(struct net_device *dev) 4666 { 4667 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4668 linkwatch_fire_event(dev); 4669 } 4670 4671 /** 4672 * netif_dormant_off - set device as not dormant. 4673 * @dev: network device 4674 * 4675 * Device is not in dormant state. 4676 */ 4677 static inline void netif_dormant_off(struct net_device *dev) 4678 { 4679 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4680 linkwatch_fire_event(dev); 4681 } 4682 4683 /** 4684 * netif_dormant - test if device is dormant 4685 * @dev: network device 4686 * 4687 * Check if device is dormant. 4688 */ 4689 static inline bool netif_dormant(const struct net_device *dev) 4690 { 4691 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4692 } 4693 4694 4695 /** 4696 * netif_testing_on - mark device as under test. 4697 * @dev: network device 4698 * 4699 * Mark device as under test (as per RFC2863). 4700 * 4701 * The testing state indicates that some test(s) must be performed on 4702 * the interface. After completion, of the test, the interface state 4703 * will change to up, dormant, or down, as appropriate. 4704 */ 4705 static inline void netif_testing_on(struct net_device *dev) 4706 { 4707 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4708 linkwatch_fire_event(dev); 4709 } 4710 4711 /** 4712 * netif_testing_off - set device as not under test. 4713 * @dev: network device 4714 * 4715 * Device is not in testing state. 4716 */ 4717 static inline void netif_testing_off(struct net_device *dev) 4718 { 4719 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4720 linkwatch_fire_event(dev); 4721 } 4722 4723 /** 4724 * netif_testing - test if device is under test 4725 * @dev: network device 4726 * 4727 * Check if device is under test 4728 */ 4729 static inline bool netif_testing(const struct net_device *dev) 4730 { 4731 return test_bit(__LINK_STATE_TESTING, &dev->state); 4732 } 4733 4734 4735 /** 4736 * netif_oper_up - test if device is operational 4737 * @dev: network device 4738 * 4739 * Check if carrier is operational 4740 */ 4741 static inline bool netif_oper_up(const struct net_device *dev) 4742 { 4743 unsigned int operstate = READ_ONCE(dev->operstate); 4744 4745 return operstate == IF_OPER_UP || 4746 operstate == IF_OPER_UNKNOWN /* backward compat */; 4747 } 4748 4749 /** 4750 * netif_device_present - is device available or removed 4751 * @dev: network device 4752 * 4753 * Check if device has not been removed from system. 4754 */ 4755 static inline bool netif_device_present(const struct net_device *dev) 4756 { 4757 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4758 } 4759 4760 void netif_device_detach(struct net_device *dev); 4761 4762 void netif_device_attach(struct net_device *dev); 4763 4764 /* 4765 * Network interface message level settings 4766 */ 4767 4768 enum { 4769 NETIF_MSG_DRV_BIT, 4770 NETIF_MSG_PROBE_BIT, 4771 NETIF_MSG_LINK_BIT, 4772 NETIF_MSG_TIMER_BIT, 4773 NETIF_MSG_IFDOWN_BIT, 4774 NETIF_MSG_IFUP_BIT, 4775 NETIF_MSG_RX_ERR_BIT, 4776 NETIF_MSG_TX_ERR_BIT, 4777 NETIF_MSG_TX_QUEUED_BIT, 4778 NETIF_MSG_INTR_BIT, 4779 NETIF_MSG_TX_DONE_BIT, 4780 NETIF_MSG_RX_STATUS_BIT, 4781 NETIF_MSG_PKTDATA_BIT, 4782 NETIF_MSG_HW_BIT, 4783 NETIF_MSG_WOL_BIT, 4784 4785 /* When you add a new bit above, update netif_msg_class_names array 4786 * in net/ethtool/common.c 4787 */ 4788 NETIF_MSG_CLASS_COUNT, 4789 }; 4790 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4791 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4792 4793 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4794 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4795 4796 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4797 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4798 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4799 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4800 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4801 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4802 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4803 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4804 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4805 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4806 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4807 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4808 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4809 #define NETIF_MSG_HW __NETIF_MSG(HW) 4810 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4811 4812 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4813 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4814 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4815 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4816 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4817 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4818 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4819 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4820 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4821 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4822 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4823 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4824 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4825 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4826 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4827 4828 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4829 { 4830 /* use default */ 4831 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4832 return default_msg_enable_bits; 4833 if (debug_value == 0) /* no output */ 4834 return 0; 4835 /* set low N bits */ 4836 return (1U << debug_value) - 1; 4837 } 4838 4839 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4840 { 4841 spin_lock(&txq->_xmit_lock); 4842 /* Pairs with READ_ONCE() in netif_tx_owned() */ 4843 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4844 } 4845 4846 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4847 { 4848 __acquire(&txq->_xmit_lock); 4849 return true; 4850 } 4851 4852 static inline void __netif_tx_release(struct netdev_queue *txq) 4853 { 4854 __release(&txq->_xmit_lock); 4855 } 4856 4857 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4858 { 4859 spin_lock_bh(&txq->_xmit_lock); 4860 /* Pairs with READ_ONCE() in netif_tx_owned() */ 4861 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4862 } 4863 4864 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4865 { 4866 bool ok = spin_trylock(&txq->_xmit_lock); 4867 4868 if (likely(ok)) { 4869 /* Pairs with READ_ONCE() in netif_tx_owned() */ 4870 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4871 } 4872 return ok; 4873 } 4874 4875 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4876 { 4877 /* Pairs with READ_ONCE() in netif_tx_owned() */ 4878 WRITE_ONCE(txq->xmit_lock_owner, -1); 4879 spin_unlock(&txq->_xmit_lock); 4880 } 4881 4882 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4883 { 4884 /* Pairs with READ_ONCE() in netif_tx_owned() */ 4885 WRITE_ONCE(txq->xmit_lock_owner, -1); 4886 spin_unlock_bh(&txq->_xmit_lock); 4887 } 4888 4889 /* 4890 * txq->trans_start can be read locklessly from dev_watchdog() 4891 */ 4892 static inline void txq_trans_update(const struct net_device *dev, 4893 struct netdev_queue *txq) 4894 { 4895 if (!dev->lltx) 4896 WRITE_ONCE(txq->trans_start, jiffies); 4897 } 4898 4899 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4900 { 4901 unsigned long now = jiffies; 4902 4903 if (READ_ONCE(txq->trans_start) != now) 4904 WRITE_ONCE(txq->trans_start, now); 4905 } 4906 4907 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4908 static inline void netif_trans_update(struct net_device *dev) 4909 { 4910 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4911 4912 txq_trans_cond_update(txq); 4913 } 4914 4915 /** 4916 * netif_tx_lock - grab network device transmit lock 4917 * @dev: network device 4918 * 4919 * Get network device transmit lock 4920 */ 4921 void netif_tx_lock(struct net_device *dev); 4922 4923 static inline void netif_tx_lock_bh(struct net_device *dev) 4924 { 4925 local_bh_disable(); 4926 netif_tx_lock(dev); 4927 } 4928 4929 void netif_tx_unlock(struct net_device *dev); 4930 4931 static inline void netif_tx_unlock_bh(struct net_device *dev) 4932 { 4933 netif_tx_unlock(dev); 4934 local_bh_enable(); 4935 } 4936 4937 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4938 if (!(dev)->lltx) { \ 4939 __netif_tx_lock(txq, cpu); \ 4940 } else { \ 4941 __netif_tx_acquire(txq); \ 4942 } \ 4943 } 4944 4945 #define HARD_TX_TRYLOCK(dev, txq) \ 4946 (!(dev)->lltx ? \ 4947 __netif_tx_trylock(txq) : \ 4948 __netif_tx_acquire(txq)) 4949 4950 #define HARD_TX_UNLOCK(dev, txq) { \ 4951 if (!(dev)->lltx) { \ 4952 __netif_tx_unlock(txq); \ 4953 } else { \ 4954 __netif_tx_release(txq); \ 4955 } \ 4956 } 4957 4958 static inline void netif_tx_disable(struct net_device *dev) 4959 { 4960 unsigned int i; 4961 int cpu; 4962 4963 local_bh_disable(); 4964 cpu = smp_processor_id(); 4965 spin_lock(&dev->tx_global_lock); 4966 for (i = 0; i < dev->num_tx_queues; i++) { 4967 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4968 4969 __netif_tx_lock(txq, cpu); 4970 netif_tx_stop_queue(txq); 4971 __netif_tx_unlock(txq); 4972 } 4973 spin_unlock(&dev->tx_global_lock); 4974 local_bh_enable(); 4975 } 4976 4977 #ifndef CONFIG_PREEMPT_RT 4978 static inline bool netif_tx_owned(struct netdev_queue *txq, unsigned int cpu) 4979 { 4980 /* Other cpus might concurrently change txq->xmit_lock_owner 4981 * to -1 or to their cpu id, but not to our id. 4982 */ 4983 return READ_ONCE(txq->xmit_lock_owner) == cpu; 4984 } 4985 4986 #else 4987 static inline bool netif_tx_owned(struct netdev_queue *txq, unsigned int cpu) 4988 { 4989 return rt_mutex_owner(&txq->_xmit_lock.lock) == current; 4990 } 4991 4992 #endif 4993 4994 static inline void netif_addr_lock(struct net_device *dev) 4995 { 4996 unsigned char nest_level = 0; 4997 4998 #ifdef CONFIG_LOCKDEP 4999 nest_level = dev->nested_level; 5000 #endif 5001 spin_lock_nested(&dev->addr_list_lock, nest_level); 5002 } 5003 5004 static inline void netif_addr_lock_bh(struct net_device *dev) 5005 { 5006 unsigned char nest_level = 0; 5007 5008 #ifdef CONFIG_LOCKDEP 5009 nest_level = dev->nested_level; 5010 #endif 5011 local_bh_disable(); 5012 spin_lock_nested(&dev->addr_list_lock, nest_level); 5013 } 5014 5015 static inline void netif_addr_unlock(struct net_device *dev) 5016 { 5017 spin_unlock(&dev->addr_list_lock); 5018 } 5019 5020 static inline void netif_addr_unlock_bh(struct net_device *dev) 5021 { 5022 spin_unlock_bh(&dev->addr_list_lock); 5023 } 5024 5025 /* 5026 * dev_addrs walker. Should be used only for read access. Call with 5027 * rcu_read_lock held. 5028 */ 5029 #define for_each_dev_addr(dev, ha) \ 5030 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 5031 5032 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 5033 5034 void ether_setup(struct net_device *dev); 5035 5036 /* Allocate dummy net_device */ 5037 struct net_device *alloc_netdev_dummy(int sizeof_priv); 5038 5039 /* Support for loadable net-drivers */ 5040 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 5041 unsigned char name_assign_type, 5042 void (*setup)(struct net_device *), 5043 unsigned int txqs, unsigned int rxqs); 5044 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 5045 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 5046 5047 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 5048 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 5049 count) 5050 5051 int register_netdev(struct net_device *dev); 5052 void unregister_netdev(struct net_device *dev); 5053 5054 int devm_register_netdev(struct device *dev, struct net_device *ndev); 5055 5056 /* General hardware address lists handling functions */ 5057 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 5058 struct netdev_hw_addr_list *from_list, int addr_len); 5059 int __hw_addr_sync_multiple(struct netdev_hw_addr_list *to_list, 5060 struct netdev_hw_addr_list *from_list, 5061 int addr_len); 5062 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 5063 struct netdev_hw_addr_list *from_list, int addr_len); 5064 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 5065 struct net_device *dev, 5066 int (*sync)(struct net_device *, const unsigned char *), 5067 int (*unsync)(struct net_device *, 5068 const unsigned char *)); 5069 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 5070 struct net_device *dev, 5071 int (*sync)(struct net_device *, 5072 const unsigned char *, int), 5073 int (*unsync)(struct net_device *, 5074 const unsigned char *, int)); 5075 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 5076 struct net_device *dev, 5077 int (*unsync)(struct net_device *, 5078 const unsigned char *, int)); 5079 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 5080 struct net_device *dev, 5081 int (*unsync)(struct net_device *, 5082 const unsigned char *)); 5083 void __hw_addr_init(struct netdev_hw_addr_list *list); 5084 void __hw_addr_flush(struct netdev_hw_addr_list *list); 5085 int __hw_addr_list_snapshot(struct netdev_hw_addr_list *snap, 5086 const struct netdev_hw_addr_list *list, 5087 int addr_len, struct netdev_hw_addr_list *cache); 5088 void __hw_addr_list_reconcile(struct netdev_hw_addr_list *real_list, 5089 struct netdev_hw_addr_list *work, 5090 struct netdev_hw_addr_list *ref, int addr_len, 5091 struct netdev_hw_addr_list *cache); 5092 5093 /* Functions used for device addresses handling */ 5094 void dev_addr_mod(struct net_device *dev, unsigned int offset, 5095 const void *addr, size_t len); 5096 5097 static inline void 5098 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 5099 { 5100 dev_addr_mod(dev, 0, addr, len); 5101 } 5102 5103 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 5104 { 5105 __dev_addr_set(dev, addr, dev->addr_len); 5106 } 5107 5108 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 5109 unsigned char addr_type); 5110 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 5111 unsigned char addr_type); 5112 5113 /* Functions used for unicast addresses handling */ 5114 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 5115 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 5116 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 5117 int dev_uc_sync(struct net_device *to, struct net_device *from); 5118 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 5119 void dev_uc_unsync(struct net_device *to, struct net_device *from); 5120 void dev_uc_flush(struct net_device *dev); 5121 void dev_uc_init(struct net_device *dev); 5122 5123 /** 5124 * __dev_uc_sync - Synchronize device's unicast list 5125 * @dev: device to sync 5126 * @sync: function to call if address should be added 5127 * @unsync: function to call if address should be removed 5128 * 5129 * Add newly added addresses to the interface, and release 5130 * addresses that have been deleted. 5131 */ 5132 static inline int __dev_uc_sync(struct net_device *dev, 5133 int (*sync)(struct net_device *, 5134 const unsigned char *), 5135 int (*unsync)(struct net_device *, 5136 const unsigned char *)) 5137 { 5138 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 5139 } 5140 5141 /** 5142 * __dev_uc_unsync - Remove synchronized addresses from device 5143 * @dev: device to sync 5144 * @unsync: function to call if address should be removed 5145 * 5146 * Remove all addresses that were added to the device by dev_uc_sync(). 5147 */ 5148 static inline void __dev_uc_unsync(struct net_device *dev, 5149 int (*unsync)(struct net_device *, 5150 const unsigned char *)) 5151 { 5152 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 5153 } 5154 5155 /* Functions used for multicast addresses handling */ 5156 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 5157 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 5158 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 5159 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 5160 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 5161 int dev_mc_sync(struct net_device *to, struct net_device *from); 5162 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 5163 void dev_mc_unsync(struct net_device *to, struct net_device *from); 5164 void dev_mc_flush(struct net_device *dev); 5165 void dev_mc_init(struct net_device *dev); 5166 5167 /** 5168 * __dev_mc_sync - Synchronize device's multicast list 5169 * @dev: device to sync 5170 * @sync: function to call if address should be added 5171 * @unsync: function to call if address should be removed 5172 * 5173 * Add newly added addresses to the interface, and release 5174 * addresses that have been deleted. 5175 */ 5176 static inline int __dev_mc_sync(struct net_device *dev, 5177 int (*sync)(struct net_device *, 5178 const unsigned char *), 5179 int (*unsync)(struct net_device *, 5180 const unsigned char *)) 5181 { 5182 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 5183 } 5184 5185 /** 5186 * __dev_mc_unsync - Remove synchronized addresses from device 5187 * @dev: device to sync 5188 * @unsync: function to call if address should be removed 5189 * 5190 * Remove all addresses that were added to the device by dev_mc_sync(). 5191 */ 5192 static inline void __dev_mc_unsync(struct net_device *dev, 5193 int (*unsync)(struct net_device *, 5194 const unsigned char *)) 5195 { 5196 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 5197 } 5198 5199 /* Functions used for secondary unicast and multicast support */ 5200 void dev_set_rx_mode(struct net_device *dev); 5201 void netif_rx_mode_schedule_retry(struct net_device *dev); 5202 int netif_set_promiscuity(struct net_device *dev, int inc); 5203 int dev_set_promiscuity(struct net_device *dev, int inc); 5204 int netif_set_allmulti(struct net_device *dev, int inc, bool notify); 5205 int dev_set_allmulti(struct net_device *dev, int inc); 5206 void netif_state_change(struct net_device *dev); 5207 void netdev_state_change(struct net_device *dev); 5208 void __netdev_notify_peers(struct net_device *dev); 5209 void netdev_notify_peers(struct net_device *dev); 5210 void netdev_features_change(struct net_device *dev); 5211 /* Load a device via the kmod */ 5212 void dev_load(struct net *net, const char *name); 5213 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 5214 struct rtnl_link_stats64 *storage); 5215 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 5216 const struct net_device_stats *netdev_stats); 5217 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 5218 const struct pcpu_sw_netstats __percpu *netstats); 5219 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 5220 5221 void netdev_work_sched(struct net_device *dev, unsigned long events); 5222 unsigned long netdev_work_cancel(struct net_device *dev, unsigned long mask); 5223 5224 enum { 5225 NESTED_SYNC_IMM_BIT, 5226 NESTED_SYNC_TODO_BIT, 5227 }; 5228 5229 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 5230 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 5231 5232 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 5233 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 5234 5235 struct netdev_nested_priv { 5236 unsigned char flags; 5237 void *data; 5238 }; 5239 5240 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 5241 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 5242 struct list_head **iter); 5243 5244 /* iterate through upper list, must be called under RCU read lock */ 5245 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 5246 for (iter = &(dev)->adj_list.upper, \ 5247 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 5248 updev; \ 5249 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 5250 5251 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 5252 int (*fn)(struct net_device *upper_dev, 5253 struct netdev_nested_priv *priv), 5254 struct netdev_nested_priv *priv); 5255 5256 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 5257 struct net_device *upper_dev); 5258 5259 bool netdev_has_any_upper_dev(struct net_device *dev); 5260 5261 void *netdev_lower_get_next_private(struct net_device *dev, 5262 struct list_head **iter); 5263 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 5264 struct list_head **iter); 5265 5266 #define netdev_for_each_lower_private(dev, priv, iter) \ 5267 for (iter = (dev)->adj_list.lower.next, \ 5268 priv = netdev_lower_get_next_private(dev, &(iter)); \ 5269 priv; \ 5270 priv = netdev_lower_get_next_private(dev, &(iter))) 5271 5272 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 5273 for (iter = &(dev)->adj_list.lower, \ 5274 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 5275 priv; \ 5276 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 5277 5278 void *netdev_lower_get_next(struct net_device *dev, 5279 struct list_head **iter); 5280 5281 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 5282 for (iter = (dev)->adj_list.lower.next, \ 5283 ldev = netdev_lower_get_next(dev, &(iter)); \ 5284 ldev; \ 5285 ldev = netdev_lower_get_next(dev, &(iter))) 5286 5287 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 5288 struct list_head **iter); 5289 int netdev_walk_all_lower_dev(struct net_device *dev, 5290 int (*fn)(struct net_device *lower_dev, 5291 struct netdev_nested_priv *priv), 5292 struct netdev_nested_priv *priv); 5293 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 5294 int (*fn)(struct net_device *lower_dev, 5295 struct netdev_nested_priv *priv), 5296 struct netdev_nested_priv *priv); 5297 5298 void *netdev_adjacent_get_private(struct list_head *adj_list); 5299 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 5300 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 5301 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 5302 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 5303 struct netlink_ext_ack *extack); 5304 int netdev_master_upper_dev_link(struct net_device *dev, 5305 struct net_device *upper_dev, 5306 void *upper_priv, void *upper_info, 5307 struct netlink_ext_ack *extack); 5308 void netdev_upper_dev_unlink(struct net_device *dev, 5309 struct net_device *upper_dev); 5310 int netdev_adjacent_change_prepare(struct net_device *old_dev, 5311 struct net_device *new_dev, 5312 struct net_device *dev, 5313 struct netlink_ext_ack *extack); 5314 void netdev_adjacent_change_commit(struct net_device *old_dev, 5315 struct net_device *new_dev, 5316 struct net_device *dev); 5317 void netdev_adjacent_change_abort(struct net_device *old_dev, 5318 struct net_device *new_dev, 5319 struct net_device *dev); 5320 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 5321 void *netdev_lower_dev_get_private(struct net_device *dev, 5322 struct net_device *lower_dev); 5323 void netdev_lower_state_changed(struct net_device *lower_dev, 5324 void *lower_state_info); 5325 5326 #define NETDEV_RSS_KEY_LEN 256 5327 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 5328 void netdev_rss_key_fill(void *buffer, size_t len); 5329 5330 int skb_checksum_help(struct sk_buff *skb); 5331 int skb_crc32c_csum_help(struct sk_buff *skb); 5332 int skb_csum_hwoffload_help(struct sk_buff *skb, 5333 const netdev_features_t features); 5334 5335 struct netdev_bonding_info { 5336 ifslave slave; 5337 ifbond master; 5338 }; 5339 5340 struct netdev_notifier_bonding_info { 5341 struct netdev_notifier_info info; /* must be first */ 5342 struct netdev_bonding_info bonding_info; 5343 }; 5344 5345 void netdev_bonding_info_change(struct net_device *dev, 5346 struct netdev_bonding_info *bonding_info); 5347 5348 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 5349 void ethtool_notify(struct net_device *dev, unsigned int cmd); 5350 #else 5351 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd) 5352 { 5353 } 5354 #endif 5355 5356 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 5357 5358 static inline bool can_checksum_protocol(netdev_features_t features, 5359 __be16 protocol) 5360 { 5361 if (protocol == htons(ETH_P_FCOE)) 5362 return !!(features & NETIF_F_FCOE_CRC); 5363 5364 /* Assume this is an IP checksum (not SCTP CRC) */ 5365 5366 if (features & NETIF_F_HW_CSUM) { 5367 /* Can checksum everything */ 5368 return true; 5369 } 5370 5371 switch (protocol) { 5372 case htons(ETH_P_IP): 5373 return !!(features & NETIF_F_IP_CSUM); 5374 case htons(ETH_P_IPV6): 5375 return !!(features & NETIF_F_IPV6_CSUM); 5376 default: 5377 return false; 5378 } 5379 } 5380 5381 #ifdef CONFIG_BUG 5382 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 5383 #else 5384 static inline void netdev_rx_csum_fault(struct net_device *dev, 5385 struct sk_buff *skb) 5386 { 5387 } 5388 #endif 5389 /* rx skb timestamps */ 5390 void net_enable_timestamp(void); 5391 void net_disable_timestamp(void); 5392 5393 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 5394 const struct skb_shared_hwtstamps *hwtstamps, 5395 bool cycles) 5396 { 5397 const struct net_device_ops *ops = dev->netdev_ops; 5398 5399 if (ops->ndo_get_tstamp) 5400 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 5401 5402 return hwtstamps->hwtstamp; 5403 } 5404 5405 #ifndef CONFIG_PREEMPT_RT 5406 static inline void netdev_xmit_set_more(bool more) 5407 { 5408 __this_cpu_write(softnet_data.xmit.more, more); 5409 } 5410 5411 static inline bool netdev_xmit_more(void) 5412 { 5413 return __this_cpu_read(softnet_data.xmit.more); 5414 } 5415 #else 5416 static inline void netdev_xmit_set_more(bool more) 5417 { 5418 current->net_xmit.more = more; 5419 } 5420 5421 static inline bool netdev_xmit_more(void) 5422 { 5423 return current->net_xmit.more; 5424 } 5425 #endif 5426 5427 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 5428 struct sk_buff *skb, struct net_device *dev, 5429 bool more) 5430 { 5431 netdev_xmit_set_more(more); 5432 return ops->ndo_start_xmit(skb, dev); 5433 } 5434 5435 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 5436 struct netdev_queue *txq, bool more) 5437 { 5438 const struct net_device_ops *ops = dev->netdev_ops; 5439 netdev_tx_t rc; 5440 5441 rc = __netdev_start_xmit(ops, skb, dev, more); 5442 if (rc == NETDEV_TX_OK) 5443 txq_trans_update(dev, txq); 5444 5445 return rc; 5446 } 5447 5448 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 5449 const struct ns_common *ns); 5450 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 5451 const struct ns_common *ns); 5452 5453 extern const struct kobj_ns_type_operations net_ns_type_operations; 5454 5455 const char *netdev_drivername(const struct net_device *dev); 5456 5457 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 5458 netdev_features_t f2) 5459 { 5460 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 5461 if (f1 & NETIF_F_HW_CSUM) 5462 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5463 else 5464 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5465 } 5466 5467 return f1 & f2; 5468 } 5469 5470 static inline netdev_features_t netdev_get_wanted_features( 5471 struct net_device *dev) 5472 { 5473 return (dev->features & ~dev->hw_features) | dev->wanted_features; 5474 } 5475 netdev_features_t netdev_increment_features(netdev_features_t all, 5476 netdev_features_t one, netdev_features_t mask); 5477 5478 /* Allow TSO being used on stacked device : 5479 * Performing the GSO segmentation before last device 5480 * is a performance improvement. 5481 */ 5482 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 5483 netdev_features_t mask) 5484 { 5485 return netdev_increment_features(features, NETIF_F_ALL_TSO | 5486 NETIF_F_ALL_FOR_ALL, mask); 5487 } 5488 5489 int __netdev_update_features(struct net_device *dev); 5490 void netdev_update_features(struct net_device *dev); 5491 void netdev_change_features(struct net_device *dev); 5492 void netdev_compute_master_upper_features(struct net_device *dev, bool update_header); 5493 5494 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 5495 struct net_device *dev); 5496 5497 netdev_features_t passthru_features_check(struct sk_buff *skb, 5498 struct net_device *dev, 5499 netdev_features_t features); 5500 netdev_features_t netif_skb_features(struct sk_buff *skb); 5501 void skb_warn_bad_offload(const struct sk_buff *skb); 5502 5503 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 5504 { 5505 netdev_features_t feature; 5506 5507 if (gso_type & (SKB_GSO_TCP_FIXEDID | SKB_GSO_TCP_FIXEDID_INNER)) 5508 gso_type |= __SKB_GSO_TCP_FIXEDID; 5509 5510 feature = ((netdev_features_t)gso_type << NETIF_F_GSO_SHIFT) & NETIF_F_GSO_MASK; 5511 5512 /* check flags correspondence */ 5513 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 5514 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 5515 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 5516 BUILD_BUG_ON(__SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 5517 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 5518 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 5519 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 5520 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 5521 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 5522 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 5523 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 5524 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 5525 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 5526 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 5527 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 5528 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 5529 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 5530 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 5531 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 5532 BUILD_BUG_ON(SKB_GSO_TCP_ACCECN != 5533 (NETIF_F_GSO_ACCECN >> NETIF_F_GSO_SHIFT)); 5534 5535 return (features & feature) == feature; 5536 } 5537 5538 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 5539 { 5540 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 5541 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 5542 } 5543 5544 static inline bool netif_needs_gso(struct sk_buff *skb, 5545 netdev_features_t features) 5546 { 5547 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 5548 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 5549 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 5550 } 5551 5552 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 5553 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 5554 void netif_inherit_tso_max(struct net_device *to, 5555 const struct net_device *from); 5556 5557 static inline unsigned int 5558 netif_get_gro_max_size(const struct net_device *dev, const struct sk_buff *skb) 5559 { 5560 /* pairs with WRITE_ONCE() in netif_set_gro(_ipv4)_max_size() */ 5561 return skb->protocol == htons(ETH_P_IPV6) ? 5562 READ_ONCE(dev->gro_max_size) : 5563 READ_ONCE(dev->gro_ipv4_max_size); 5564 } 5565 5566 static inline unsigned int 5567 netif_get_gso_max_size(const struct net_device *dev, const struct sk_buff *skb) 5568 { 5569 /* pairs with WRITE_ONCE() in netif_set_gso(_ipv4)_max_size() */ 5570 return skb->protocol == htons(ETH_P_IPV6) ? 5571 READ_ONCE(dev->gso_max_size) : 5572 READ_ONCE(dev->gso_ipv4_max_size); 5573 } 5574 5575 static inline bool netif_is_macsec(const struct net_device *dev) 5576 { 5577 return dev->priv_flags & IFF_MACSEC; 5578 } 5579 5580 static inline bool netif_is_macvlan(const struct net_device *dev) 5581 { 5582 return dev->priv_flags & IFF_MACVLAN; 5583 } 5584 5585 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5586 { 5587 return dev->priv_flags & IFF_MACVLAN_PORT; 5588 } 5589 5590 static inline bool netif_is_bond_master(const struct net_device *dev) 5591 { 5592 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5593 } 5594 5595 static inline bool netif_is_bond_slave(const struct net_device *dev) 5596 { 5597 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5598 } 5599 5600 static inline bool netif_supports_nofcs(struct net_device *dev) 5601 { 5602 return dev->priv_flags & IFF_SUPP_NOFCS; 5603 } 5604 5605 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5606 { 5607 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5608 } 5609 5610 static inline bool netif_is_l3_master(const struct net_device *dev) 5611 { 5612 return dev->priv_flags & IFF_L3MDEV_MASTER; 5613 } 5614 5615 static inline bool netif_is_l3_slave(const struct net_device *dev) 5616 { 5617 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5618 } 5619 5620 static inline int dev_sdif(const struct net_device *dev) 5621 { 5622 #ifdef CONFIG_NET_L3_MASTER_DEV 5623 if (netif_is_l3_slave(dev)) 5624 return dev->ifindex; 5625 #endif 5626 return 0; 5627 } 5628 5629 static inline bool netif_is_bridge_master(const struct net_device *dev) 5630 { 5631 return dev->priv_flags & IFF_EBRIDGE; 5632 } 5633 5634 static inline bool netif_is_bridge_port(const struct net_device *dev) 5635 { 5636 return dev->priv_flags & IFF_BRIDGE_PORT; 5637 } 5638 5639 static inline bool netif_is_ovs_master(const struct net_device *dev) 5640 { 5641 return dev->priv_flags & IFF_OPENVSWITCH; 5642 } 5643 5644 static inline bool netif_is_ovs_port(const struct net_device *dev) 5645 { 5646 return dev->priv_flags & IFF_OVS_DATAPATH; 5647 } 5648 5649 static inline bool netif_is_any_bridge_master(const struct net_device *dev) 5650 { 5651 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev); 5652 } 5653 5654 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5655 { 5656 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5657 } 5658 5659 static inline bool netif_is_team_master(const struct net_device *dev) 5660 { 5661 return dev->priv_flags & IFF_TEAM; 5662 } 5663 5664 static inline bool netif_is_team_port(const struct net_device *dev) 5665 { 5666 return dev->priv_flags & IFF_TEAM_PORT; 5667 } 5668 5669 static inline bool netif_is_lag_master(const struct net_device *dev) 5670 { 5671 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5672 } 5673 5674 static inline bool netif_is_lag_port(const struct net_device *dev) 5675 { 5676 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5677 } 5678 5679 bool netif_is_rxfh_configured(const struct net_device *dev); 5680 5681 static inline bool netif_is_failover(const struct net_device *dev) 5682 { 5683 return dev->priv_flags & IFF_FAILOVER; 5684 } 5685 5686 static inline bool netif_is_failover_slave(const struct net_device *dev) 5687 { 5688 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5689 } 5690 5691 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5692 static inline void netif_keep_dst(struct net_device *dev) 5693 { 5694 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5695 } 5696 5697 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5698 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5699 { 5700 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5701 return netif_is_macsec(dev); 5702 } 5703 5704 extern struct pernet_operations __net_initdata loopback_net_ops; 5705 5706 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5707 5708 /* netdev_printk helpers, similar to dev_printk */ 5709 5710 static inline const char *netdev_name(const struct net_device *dev) 5711 { 5712 if (!dev->name[0] || strchr(dev->name, '%')) 5713 return "(unnamed net_device)"; 5714 return dev->name; 5715 } 5716 5717 static inline const char *netdev_reg_state(const struct net_device *dev) 5718 { 5719 u8 reg_state = READ_ONCE(dev->reg_state); 5720 5721 switch (reg_state) { 5722 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5723 case NETREG_REGISTERED: return ""; 5724 case NETREG_UNREGISTERING: return " (unregistering)"; 5725 case NETREG_UNREGISTERED: return " (unregistered)"; 5726 case NETREG_RELEASED: return " (released)"; 5727 case NETREG_DUMMY: return " (dummy)"; 5728 } 5729 5730 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state); 5731 return " (unknown)"; 5732 } 5733 5734 #define MODULE_ALIAS_NETDEV(device) \ 5735 MODULE_ALIAS("netdev-" device) 5736 5737 /* 5738 * netdev_WARN() acts like dev_printk(), but with the key difference 5739 * of using a WARN/WARN_ON to get the message out, including the 5740 * file/line information and a backtrace. 5741 */ 5742 #define netdev_WARN(dev, format, args...) \ 5743 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5744 netdev_reg_state(dev), ##args) 5745 5746 #define netdev_WARN_ONCE(dev, format, args...) \ 5747 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5748 netdev_reg_state(dev), ##args) 5749 5750 /* 5751 * The list of packet types we will receive (as opposed to discard) 5752 * and the routines to invoke. 5753 * 5754 * Why 16. Because with 16 the only overlap we get on a hash of the 5755 * low nibble of the protocol value is RARP/SNAP/X.25. 5756 * 5757 * 0800 IP 5758 * 0001 802.3 5759 * 0002 AX.25 5760 * 0004 802.2 5761 * 8035 RARP 5762 * 0005 SNAP 5763 * 0805 X.25 5764 * 0806 ARP 5765 * 8137 IPX 5766 * 0009 Localtalk 5767 * 86DD IPv6 5768 */ 5769 #define PTYPE_HASH_SIZE (16) 5770 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5771 5772 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5773 5774 extern struct net_device *blackhole_netdev; 5775 5776 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */ 5777 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD) 5778 #define DEV_STATS_ADD(DEV, FIELD, VAL) \ 5779 atomic_long_add((VAL), &(DEV)->stats.__##FIELD) 5780 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD) 5781 5782 #endif /* _LINUX_NETDEVICE_H */ 5783