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 u16 count; 836 u16 offset; 837 }; 838 839 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 840 /* 841 * This structure is to hold information about the device 842 * configured to run FCoE protocol stack. 843 */ 844 struct netdev_fcoe_hbainfo { 845 char manufacturer[64]; 846 char serial_number[64]; 847 char hardware_version[64]; 848 char driver_version[64]; 849 char optionrom_version[64]; 850 char firmware_version[64]; 851 char model[256]; 852 char model_description[256]; 853 }; 854 #endif 855 856 #define MAX_PHYS_ITEM_ID_LEN 32 857 858 /* This structure holds a unique identifier to identify some 859 * physical item (port for example) used by a netdevice. 860 */ 861 struct netdev_phys_item_id { 862 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 863 unsigned char id_len; 864 }; 865 866 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 867 struct netdev_phys_item_id *b) 868 { 869 return a->id_len == b->id_len && 870 memcmp(a->id, b->id, a->id_len) == 0; 871 } 872 873 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 874 struct sk_buff *skb, 875 struct net_device *sb_dev); 876 877 enum net_device_path_type { 878 DEV_PATH_ETHERNET = 0, 879 DEV_PATH_VLAN, 880 DEV_PATH_BRIDGE, 881 DEV_PATH_PPPOE, 882 DEV_PATH_DSA, 883 DEV_PATH_MTK_WDMA, 884 DEV_PATH_TUN, 885 }; 886 887 struct net_device_path { 888 enum net_device_path_type type; 889 const struct net_device *dev; 890 union { 891 struct { 892 u16 id; 893 __be16 proto; 894 u8 h_dest[ETH_ALEN]; 895 } encap; 896 struct { 897 union { 898 struct in_addr src_v4; 899 struct in6_addr src_v6; 900 }; 901 union { 902 struct in_addr dst_v4; 903 struct in6_addr dst_v6; 904 }; 905 906 u8 l3_proto; 907 } tun; 908 struct { 909 enum { 910 DEV_PATH_BR_VLAN_KEEP, 911 DEV_PATH_BR_VLAN_TAG, 912 DEV_PATH_BR_VLAN_UNTAG, 913 DEV_PATH_BR_VLAN_UNTAG_HW, 914 } vlan_mode; 915 u16 vlan_id; 916 __be16 vlan_proto; 917 } bridge; 918 struct { 919 int port; 920 u16 proto; 921 } dsa; 922 struct { 923 u8 wdma_idx; 924 u8 queue; 925 u16 wcid; 926 u8 bss; 927 u8 amsdu; 928 } mtk_wdma; 929 }; 930 }; 931 932 #define NET_DEVICE_PATH_STACK_MAX 5 933 #define NET_DEVICE_PATH_VLAN_MAX 2 934 935 struct net_device_path_stack { 936 int num_paths; 937 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 938 }; 939 940 struct net_device_path_ctx { 941 const struct net_device *dev; 942 u8 daddr[ETH_ALEN]; 943 944 int num_vlans; 945 struct { 946 u16 id; 947 __be16 proto; 948 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 949 }; 950 951 enum tc_setup_type { 952 TC_QUERY_CAPS, 953 TC_SETUP_QDISC_MQPRIO, 954 TC_SETUP_CLSU32, 955 TC_SETUP_CLSFLOWER, 956 TC_SETUP_CLSMATCHALL, 957 TC_SETUP_CLSBPF, 958 TC_SETUP_BLOCK, 959 TC_SETUP_QDISC_CBS, 960 TC_SETUP_QDISC_RED, 961 TC_SETUP_QDISC_PRIO, 962 TC_SETUP_QDISC_MQ, 963 TC_SETUP_QDISC_ETF, 964 TC_SETUP_ROOT_QDISC, 965 TC_SETUP_QDISC_GRED, 966 TC_SETUP_QDISC_TAPRIO, 967 TC_SETUP_FT, 968 TC_SETUP_QDISC_ETS, 969 TC_SETUP_QDISC_TBF, 970 TC_SETUP_QDISC_FIFO, 971 TC_SETUP_QDISC_HTB, 972 TC_SETUP_ACT, 973 }; 974 975 /* These structures hold the attributes of bpf state that are being passed 976 * to the netdevice through the bpf op. 977 */ 978 enum bpf_netdev_command { 979 /* Set or clear a bpf program used in the earliest stages of packet 980 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 981 * is responsible for calling bpf_prog_put on any old progs that are 982 * stored. In case of error, the callee need not release the new prog 983 * reference, but on success it takes ownership and must bpf_prog_put 984 * when it is no longer used. 985 */ 986 XDP_SETUP_PROG, 987 XDP_SETUP_PROG_HW, 988 /* BPF program for offload callbacks, invoked at program load time. */ 989 BPF_OFFLOAD_MAP_ALLOC, 990 BPF_OFFLOAD_MAP_FREE, 991 XDP_SETUP_XSK_POOL, 992 }; 993 994 struct bpf_prog_offload_ops; 995 struct netlink_ext_ack; 996 struct xdp_umem; 997 struct xdp_dev_bulk_queue; 998 struct bpf_xdp_link; 999 1000 enum bpf_xdp_mode { 1001 XDP_MODE_SKB = 0, 1002 XDP_MODE_DRV = 1, 1003 XDP_MODE_HW = 2, 1004 __MAX_XDP_MODE 1005 }; 1006 1007 struct bpf_xdp_entity { 1008 struct bpf_prog *prog; 1009 struct bpf_xdp_link *link; 1010 }; 1011 1012 struct netdev_bpf { 1013 enum bpf_netdev_command command; 1014 union { 1015 /* XDP_SETUP_PROG */ 1016 struct { 1017 u32 flags; 1018 struct bpf_prog *prog; 1019 struct netlink_ext_ack *extack; 1020 }; 1021 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 1022 struct { 1023 struct bpf_offloaded_map *offmap; 1024 }; 1025 /* XDP_SETUP_XSK_POOL */ 1026 struct { 1027 struct xsk_buff_pool *pool; 1028 u16 queue_id; 1029 } xsk; 1030 }; 1031 }; 1032 1033 /* Flags for ndo_xsk_wakeup. */ 1034 #define XDP_WAKEUP_RX (1 << 0) 1035 #define XDP_WAKEUP_TX (1 << 1) 1036 1037 #ifdef CONFIG_XFRM_OFFLOAD 1038 struct xfrmdev_ops { 1039 int (*xdo_dev_state_add)(struct net_device *dev, 1040 struct xfrm_state *x, 1041 struct netlink_ext_ack *extack); 1042 void (*xdo_dev_state_delete)(struct net_device *dev, 1043 struct xfrm_state *x); 1044 void (*xdo_dev_state_free)(struct net_device *dev, 1045 struct xfrm_state *x); 1046 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 1047 struct xfrm_state *x); 1048 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1049 void (*xdo_dev_state_update_stats) (struct xfrm_state *x); 1050 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack); 1051 void (*xdo_dev_policy_delete) (struct xfrm_policy *x); 1052 void (*xdo_dev_policy_free) (struct xfrm_policy *x); 1053 }; 1054 #endif 1055 1056 struct dev_ifalias { 1057 struct rcu_head rcuhead; 1058 char ifalias[]; 1059 }; 1060 1061 struct devlink; 1062 struct tlsdev_ops; 1063 1064 struct netdev_net_notifier { 1065 struct list_head list; 1066 struct notifier_block *nb; 1067 }; 1068 1069 /* 1070 * This structure defines the management hooks for network devices. 1071 * The following hooks can be defined; unless noted otherwise, they are 1072 * optional and can be filled with a null pointer. 1073 * 1074 * int (*ndo_init)(struct net_device *dev); 1075 * This function is called once when a network device is registered. 1076 * The network device can use this for any late stage initialization 1077 * or semantic validation. It can fail with an error code which will 1078 * be propagated back to register_netdev. 1079 * 1080 * void (*ndo_uninit)(struct net_device *dev); 1081 * This function is called when device is unregistered or when registration 1082 * fails. It is not called if init fails. 1083 * 1084 * int (*ndo_open)(struct net_device *dev); 1085 * This function is called when a network device transitions to the up 1086 * state. 1087 * 1088 * int (*ndo_stop)(struct net_device *dev); 1089 * This function is called when a network device transitions to the down 1090 * state. 1091 * 1092 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1093 * struct net_device *dev); 1094 * Called when a packet needs to be transmitted. 1095 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1096 * the queue before that can happen; it's for obsolete devices and weird 1097 * corner cases, but the stack really does a non-trivial amount 1098 * of useless work if you return NETDEV_TX_BUSY. 1099 * Required; cannot be NULL. 1100 * 1101 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1102 * struct net_device *dev 1103 * netdev_features_t features); 1104 * Called by core transmit path to determine if device is capable of 1105 * performing offload operations on a given packet. This is to give 1106 * the device an opportunity to implement any restrictions that cannot 1107 * be otherwise expressed by feature flags. The check is called with 1108 * the set of features that the stack has calculated and it returns 1109 * those the driver believes to be appropriate. 1110 * 1111 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1112 * struct net_device *sb_dev); 1113 * Called to decide which queue to use when device supports multiple 1114 * transmit queues. 1115 * 1116 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1117 * This function is called to allow device receiver to make 1118 * changes to configuration when multicast or promiscuous is enabled. 1119 * 1120 * void (*ndo_set_rx_mode)(struct net_device *dev); 1121 * This function is called device changes address list filtering. 1122 * If driver handles unicast address filtering, it should set 1123 * IFF_UNICAST_FLT in its priv_flags. 1124 * Cannot sleep, called with netif_addr_lock_bh held. 1125 * Deprecated in favor of ndo_set_rx_mode_async. 1126 * 1127 * int (*ndo_set_rx_mode_async)(struct net_device *dev, 1128 * struct netdev_hw_addr_list *uc, 1129 * struct netdev_hw_addr_list *mc); 1130 * Async version of ndo_set_rx_mode which runs in process context 1131 * with rtnl_lock and netdev_lock_ops(dev) held. The uc/mc parameters 1132 * are snapshots of the address lists - iterate with 1133 * netdev_hw_addr_list_for_each(ha, uc). Return 0 on success or a 1134 * negative errno to request a retry via the core backoff. 1135 * 1136 * void (*ndo_work)(struct net_device *dev, unsigned long events); 1137 * Run deferred work scheduled with netdev_work_sched(@events). 1138 * 1139 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1140 * This function is called when the Media Access Control address 1141 * needs to be changed. If this interface is not defined, the 1142 * MAC address can not be changed. 1143 * 1144 * int (*ndo_validate_addr)(struct net_device *dev); 1145 * Test if Media Access Control address is valid for the device. 1146 * 1147 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1148 * Old-style ioctl entry point. This is used internally by the 1149 * ieee802154 subsystem but is no longer called by the device 1150 * ioctl handler. 1151 * 1152 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1153 * Used by the bonding driver for its device specific ioctls: 1154 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1155 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1156 * 1157 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1158 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG and 1159 * SIOCSMIIREG. 1160 * 1161 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1162 * Used to set network devices bus interface parameters. This interface 1163 * is retained for legacy reasons; new devices should use the bus 1164 * interface (PCI) for low level management. 1165 * 1166 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1167 * Called when a user wants to change the Maximum Transfer Unit 1168 * of a device. 1169 * 1170 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1171 * Callback used when the transmitter has not made any progress 1172 * for dev->watchdog ticks. 1173 * 1174 * void (*ndo_get_stats64)(struct net_device *dev, 1175 * struct rtnl_link_stats64 *storage); 1176 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1177 * Called when a user wants to get the network device usage 1178 * statistics. Drivers must do one of the following: 1179 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1180 * rtnl_link_stats64 structure passed by the caller. 1181 * 2. Define @ndo_get_stats to update a net_device_stats structure 1182 * (which should normally be dev->stats) and return a pointer to 1183 * it. The structure may be changed asynchronously only if each 1184 * field is written atomically. 1185 * 3. Update dev->stats asynchronously and atomically, and define 1186 * neither operation. 1187 * 1188 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1189 * Return true if this device supports offload stats of this attr_id. 1190 * 1191 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1192 * void *attr_data) 1193 * Get statistics for offload operations by attr_id. Write it into the 1194 * attr_data pointer. 1195 * 1196 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1197 * If device supports VLAN filtering this function is called when a 1198 * VLAN id is registered. 1199 * 1200 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1201 * If device supports VLAN filtering this function is called when a 1202 * VLAN id is unregistered. 1203 * 1204 * void (*ndo_poll_controller)(struct net_device *dev); 1205 * 1206 * SR-IOV management functions. 1207 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1208 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1209 * u8 qos, __be16 proto); 1210 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1211 * int max_tx_rate); 1212 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1213 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1214 * int (*ndo_get_vf_config)(struct net_device *dev, 1215 * int vf, struct ifla_vf_info *ivf); 1216 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1217 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1218 * struct nlattr *port[]); 1219 * 1220 * Enable or disable the VF ability to query its RSS Redirection Table and 1221 * Hash Key. This is needed since on some devices VF share this information 1222 * with PF and querying it may introduce a theoretical security risk. 1223 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1224 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1225 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1226 * void *type_data); 1227 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1228 * This is always called from the stack with the rtnl lock held and netif 1229 * tx queues stopped. This allows the netdevice to perform queue 1230 * management safely. 1231 * 1232 * NB: Returning -EOPNOTSUPP for whatever commands means "this qdisc 1233 * is not offloaded (anymore, offloading may have silently stopped)", 1234 * and the offloading flag is cleared. Notably, this is also true for 1235 * dump queries (e.g. TC_*_STATS commands). If the underlying device does 1236 * not report any statistics but is still offloading, return 0 instead. 1237 * 1238 * Fiber Channel over Ethernet (FCoE) offload functions. 1239 * int (*ndo_fcoe_enable)(struct net_device *dev); 1240 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1241 * so the underlying device can perform whatever needed configuration or 1242 * initialization to support acceleration of FCoE traffic. 1243 * 1244 * int (*ndo_fcoe_disable)(struct net_device *dev); 1245 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1246 * so the underlying device can perform whatever needed clean-ups to 1247 * stop supporting acceleration of FCoE traffic. 1248 * 1249 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1250 * struct scatterlist *sgl, unsigned int sgc); 1251 * Called when the FCoE Initiator wants to initialize an I/O that 1252 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1253 * perform necessary setup and returns 1 to indicate the device is set up 1254 * successfully to perform DDP on this I/O, otherwise this returns 0. 1255 * 1256 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1257 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1258 * indicated by the FC exchange id 'xid', so the underlying device can 1259 * clean up and reuse resources for later DDP requests. 1260 * 1261 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1262 * struct scatterlist *sgl, unsigned int sgc); 1263 * Called when the FCoE Target wants to initialize an I/O that 1264 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1265 * perform necessary setup and returns 1 to indicate the device is set up 1266 * successfully to perform DDP on this I/O, otherwise this returns 0. 1267 * 1268 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1269 * struct netdev_fcoe_hbainfo *hbainfo); 1270 * Called when the FCoE Protocol stack wants information on the underlying 1271 * device. This information is utilized by the FCoE protocol stack to 1272 * register attributes with Fiber Channel management service as per the 1273 * FC-GS Fabric Device Management Information(FDMI) specification. 1274 * 1275 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1276 * Called when the underlying device wants to override default World Wide 1277 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1278 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1279 * protocol stack to use. 1280 * 1281 * RFS acceleration. 1282 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1283 * u16 rxq_index, u32 flow_id); 1284 * Set hardware filter for RFS. rxq_index is the target queue index; 1285 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1286 * Return the filter ID on success, or a negative error code. 1287 * 1288 * Slave management functions (for bridge, bonding, etc). 1289 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1290 * Called to make another netdev an underling. 1291 * 1292 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1293 * Called to release previously enslaved netdev. 1294 * 1295 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1296 * struct sk_buff *skb, 1297 * bool all_slaves); 1298 * Get the xmit slave of master device. If all_slaves is true, function 1299 * assume all the slaves can transmit. 1300 * 1301 * Feature/offload setting functions. 1302 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1303 * netdev_features_t features); 1304 * Adjusts the requested feature flags according to device-specific 1305 * constraints, and returns the resulting flags. Must not modify 1306 * the device state. 1307 * 1308 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1309 * Called to update device configuration to new features. Passed 1310 * feature set might be less than what was returned by ndo_fix_features()). 1311 * Must return >0 or -errno if it changed dev->features itself. 1312 * 1313 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1314 * struct net_device *dev, 1315 * const unsigned char *addr, u16 vid, u16 flags, 1316 * bool *notified, struct netlink_ext_ack *extack); 1317 * Adds an FDB entry to dev for addr. 1318 * Callee shall set *notified to true if it sent any appropriate 1319 * notification(s). Otherwise core will send a generic one. 1320 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1321 * struct net_device *dev, 1322 * const unsigned char *addr, u16 vid 1323 * bool *notified, struct netlink_ext_ack *extack); 1324 * Deletes the FDB entry from dev corresponding to addr. 1325 * Callee shall set *notified to true if it sent any appropriate 1326 * notification(s). Otherwise core will send a generic one. 1327 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev, 1328 * struct netlink_ext_ack *extack); 1329 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1330 * struct net_device *dev, struct net_device *filter_dev, 1331 * int *idx) 1332 * Used to add FDB entries to dump requests. Implementers should add 1333 * entries to skb and update idx with the number of entries. 1334 * 1335 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[], 1336 * u16 nlmsg_flags, struct netlink_ext_ack *extack); 1337 * Adds an MDB entry to dev. 1338 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[], 1339 * struct netlink_ext_ack *extack); 1340 * Deletes the MDB entry from dev. 1341 * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[], 1342 * struct netlink_ext_ack *extack); 1343 * Bulk deletes MDB entries from dev. 1344 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb, 1345 * struct netlink_callback *cb); 1346 * Dumps MDB entries from dev. The first argument (marker) in the netlink 1347 * callback is used by core rtnetlink code. 1348 * 1349 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1350 * u16 flags, struct netlink_ext_ack *extack) 1351 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1352 * struct net_device *dev, u32 filter_mask, 1353 * int nlflags) 1354 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1355 * u16 flags); 1356 * 1357 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1358 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1359 * which do not represent real hardware may define this to allow their 1360 * userspace components to manage their virtual carrier state. Devices 1361 * that determine carrier state from physical hardware properties (eg 1362 * network cables) or protocol-dependent mechanisms (eg 1363 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1364 * 1365 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1366 * struct netdev_phys_item_id *ppid); 1367 * Called to get ID of physical port of this device. If driver does 1368 * not implement this, it is assumed that the hw is not able to have 1369 * multiple net devices on single physical port. 1370 * 1371 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1372 * struct netdev_phys_item_id *ppid) 1373 * Called to get the parent ID of the physical port of this device. 1374 * 1375 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1376 * struct net_device *dev) 1377 * Called by upper layer devices to accelerate switching or other 1378 * station functionality into hardware. 'pdev is the lowerdev 1379 * to use for the offload and 'dev' is the net device that will 1380 * back the offload. Returns a pointer to the private structure 1381 * the upper layer will maintain. 1382 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1383 * Called by upper layer device to delete the station created 1384 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1385 * the station and priv is the structure returned by the add 1386 * operation. 1387 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1388 * int queue_index, u32 maxrate); 1389 * Called when a user wants to set a max-rate limitation of specific 1390 * TX queue. 1391 * int (*ndo_get_iflink)(const struct net_device *dev); 1392 * Called to get the iflink value of this device. 1393 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1394 * This function is used to get egress tunnel information for given skb. 1395 * This is useful for retrieving outer tunnel header parameters while 1396 * sampling packet. 1397 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1398 * This function is used to specify the headroom that the skb must 1399 * consider when allocation skb during packet reception. Setting 1400 * appropriate rx headroom value allows avoiding skb head copy on 1401 * forward. Setting a negative value resets the rx headroom to the 1402 * default value. 1403 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1404 * This function is used to set or query state related to XDP on the 1405 * netdevice and manage BPF offload. See definition of 1406 * enum bpf_netdev_command for details. 1407 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1408 * u32 flags); 1409 * This function is used to submit @n XDP packets for transmit on a 1410 * netdevice. Returns number of frames successfully transmitted, frames 1411 * that got dropped are freed/returned via xdp_return_frame(). 1412 * Returns negative number, means general error invoking ndo, meaning 1413 * no frames were xmit'ed and core-caller will free all frames. 1414 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1415 * struct xdp_buff *xdp); 1416 * Get the xmit slave of master device based on the xdp_buff. 1417 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1418 * This function is used to wake up the softirq, ksoftirqd or kthread 1419 * responsible for sending and/or receiving packets on a specific 1420 * queue id bound to an AF_XDP socket. The flags field specifies if 1421 * only RX, only Tx, or both should be woken up using the flags 1422 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1423 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm_kern *p, 1424 * int cmd); 1425 * Add, change, delete or get information on an IPv4 tunnel. 1426 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1427 * If a device is paired with a peer device, return the peer instance. 1428 * The caller must be under RCU read context. 1429 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1430 * Get the forwarding path to reach the real device from the HW destination address 1431 * ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1432 * const struct skb_shared_hwtstamps *hwtstamps, 1433 * bool cycles); 1434 * Get hardware timestamp based on normal/adjustable time or free running 1435 * cycle counter. This function is required if physical clock supports a 1436 * free running cycle counter. 1437 * 1438 * int (*ndo_hwtstamp_get)(struct net_device *dev, 1439 * struct kernel_hwtstamp_config *kernel_config); 1440 * Get the currently configured hardware timestamping parameters for the 1441 * NIC device. 1442 * 1443 * int (*ndo_hwtstamp_set)(struct net_device *dev, 1444 * struct kernel_hwtstamp_config *kernel_config, 1445 * struct netlink_ext_ack *extack); 1446 * Change the hardware timestamping parameters for NIC device. 1447 */ 1448 struct net_device_ops { 1449 int (*ndo_init)(struct net_device *dev); 1450 void (*ndo_uninit)(struct net_device *dev); 1451 int (*ndo_open)(struct net_device *dev); 1452 int (*ndo_stop)(struct net_device *dev); 1453 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1454 struct net_device *dev); 1455 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1456 struct net_device *dev, 1457 netdev_features_t features); 1458 u16 (*ndo_select_queue)(struct net_device *dev, 1459 struct sk_buff *skb, 1460 struct net_device *sb_dev); 1461 void (*ndo_change_rx_flags)(struct net_device *dev, 1462 int flags); 1463 void (*ndo_set_rx_mode)(struct net_device *dev); 1464 int (*ndo_set_rx_mode_async)( 1465 struct net_device *dev, 1466 struct netdev_hw_addr_list *uc, 1467 struct netdev_hw_addr_list *mc); 1468 void (*ndo_work)(struct net_device *dev, 1469 unsigned long events); 1470 int (*ndo_set_mac_address)(struct net_device *dev, 1471 void *addr); 1472 int (*ndo_validate_addr)(struct net_device *dev); 1473 int (*ndo_do_ioctl)(struct net_device *dev, 1474 struct ifreq *ifr, int cmd); 1475 int (*ndo_eth_ioctl)(struct net_device *dev, 1476 struct ifreq *ifr, int cmd); 1477 int (*ndo_siocbond)(struct net_device *dev, 1478 struct ifreq *ifr, int cmd); 1479 int (*ndo_siocwandev)(struct net_device *dev, 1480 struct if_settings *ifs); 1481 int (*ndo_siocdevprivate)(struct net_device *dev, 1482 struct ifreq *ifr, 1483 void __user *data, int cmd); 1484 int (*ndo_set_config)(struct net_device *dev, 1485 struct ifmap *map); 1486 int (*ndo_change_mtu)(struct net_device *dev, 1487 int new_mtu); 1488 int (*ndo_neigh_setup)(struct net_device *dev, 1489 struct neigh_parms *); 1490 void (*ndo_tx_timeout) (struct net_device *dev, 1491 unsigned int txqueue); 1492 1493 void (*ndo_get_stats64)(struct net_device *dev, 1494 struct rtnl_link_stats64 *storage); 1495 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1496 int (*ndo_get_offload_stats)(int attr_id, 1497 const struct net_device *dev, 1498 void *attr_data); 1499 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1500 1501 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1502 __be16 proto, u16 vid); 1503 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1504 __be16 proto, u16 vid); 1505 #ifdef CONFIG_NET_POLL_CONTROLLER 1506 void (*ndo_poll_controller)(struct net_device *dev); 1507 int (*ndo_netpoll_setup)(struct net_device *dev); 1508 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1509 #endif 1510 int (*ndo_set_vf_mac)(struct net_device *dev, 1511 int queue, u8 *mac); 1512 int (*ndo_set_vf_vlan)(struct net_device *dev, 1513 int queue, u16 vlan, 1514 u8 qos, __be16 proto); 1515 int (*ndo_set_vf_rate)(struct net_device *dev, 1516 int vf, int min_tx_rate, 1517 int max_tx_rate); 1518 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1519 int vf, bool setting); 1520 int (*ndo_set_vf_trust)(struct net_device *dev, 1521 int vf, bool setting); 1522 int (*ndo_get_vf_config)(struct net_device *dev, 1523 int vf, 1524 struct ifla_vf_info *ivf); 1525 int (*ndo_set_vf_link_state)(struct net_device *dev, 1526 int vf, int link_state); 1527 int (*ndo_get_vf_stats)(struct net_device *dev, 1528 int vf, 1529 struct ifla_vf_stats 1530 *vf_stats); 1531 int (*ndo_set_vf_port)(struct net_device *dev, 1532 int vf, 1533 struct nlattr *port[]); 1534 int (*ndo_get_vf_port)(struct net_device *dev, 1535 int vf, struct sk_buff *skb); 1536 int (*ndo_get_vf_guid)(struct net_device *dev, 1537 int vf, 1538 struct ifla_vf_guid *node_guid, 1539 struct ifla_vf_guid *port_guid); 1540 int (*ndo_set_vf_guid)(struct net_device *dev, 1541 int vf, u64 guid, 1542 int guid_type); 1543 int (*ndo_set_vf_rss_query_en)( 1544 struct net_device *dev, 1545 int vf, bool setting); 1546 int (*ndo_setup_tc)(struct net_device *dev, 1547 enum tc_setup_type type, 1548 void *type_data); 1549 #if IS_ENABLED(CONFIG_FCOE) 1550 int (*ndo_fcoe_enable)(struct net_device *dev); 1551 int (*ndo_fcoe_disable)(struct net_device *dev); 1552 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1553 u16 xid, 1554 struct scatterlist *sgl, 1555 unsigned int sgc); 1556 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1557 u16 xid); 1558 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1559 u16 xid, 1560 struct scatterlist *sgl, 1561 unsigned int sgc); 1562 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1563 struct netdev_fcoe_hbainfo *hbainfo); 1564 #endif 1565 1566 #if IS_ENABLED(CONFIG_LIBFCOE) 1567 #define NETDEV_FCOE_WWNN 0 1568 #define NETDEV_FCOE_WWPN 1 1569 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1570 u64 *wwn, int type); 1571 #endif 1572 1573 #ifdef CONFIG_RFS_ACCEL 1574 int (*ndo_rx_flow_steer)(struct net_device *dev, 1575 const struct sk_buff *skb, 1576 u16 rxq_index, 1577 u32 flow_id); 1578 #endif 1579 int (*ndo_add_slave)(struct net_device *dev, 1580 struct net_device *slave_dev, 1581 struct netlink_ext_ack *extack); 1582 int (*ndo_del_slave)(struct net_device *dev, 1583 struct net_device *slave_dev); 1584 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1585 struct sk_buff *skb, 1586 bool all_slaves); 1587 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1588 struct sock *sk); 1589 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1590 netdev_features_t features); 1591 int (*ndo_set_features)(struct net_device *dev, 1592 netdev_features_t features); 1593 int (*ndo_neigh_construct)(struct net_device *dev, 1594 struct neighbour *n); 1595 void (*ndo_neigh_destroy)(struct net_device *dev, 1596 struct neighbour *n); 1597 1598 int (*ndo_fdb_add)(struct ndmsg *ndm, 1599 struct nlattr *tb[], 1600 struct net_device *dev, 1601 const unsigned char *addr, 1602 u16 vid, 1603 u16 flags, 1604 bool *notified, 1605 struct netlink_ext_ack *extack); 1606 int (*ndo_fdb_del)(struct ndmsg *ndm, 1607 struct nlattr *tb[], 1608 struct net_device *dev, 1609 const unsigned char *addr, 1610 u16 vid, 1611 bool *notified, 1612 struct netlink_ext_ack *extack); 1613 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, 1614 struct net_device *dev, 1615 struct netlink_ext_ack *extack); 1616 int (*ndo_fdb_dump)(struct sk_buff *skb, 1617 struct netlink_callback *cb, 1618 struct net_device *dev, 1619 struct net_device *filter_dev, 1620 int *idx); 1621 int (*ndo_fdb_get)(struct sk_buff *skb, 1622 struct nlattr *tb[], 1623 struct net_device *dev, 1624 const unsigned char *addr, 1625 u16 vid, u32 portid, u32 seq, 1626 struct netlink_ext_ack *extack); 1627 int (*ndo_mdb_add)(struct net_device *dev, 1628 struct nlattr *tb[], 1629 u16 nlmsg_flags, 1630 struct netlink_ext_ack *extack); 1631 int (*ndo_mdb_del)(struct net_device *dev, 1632 struct nlattr *tb[], 1633 struct netlink_ext_ack *extack); 1634 int (*ndo_mdb_del_bulk)(struct net_device *dev, 1635 struct nlattr *tb[], 1636 struct netlink_ext_ack *extack); 1637 int (*ndo_mdb_dump)(struct net_device *dev, 1638 struct sk_buff *skb, 1639 struct netlink_callback *cb); 1640 int (*ndo_mdb_get)(struct net_device *dev, 1641 struct nlattr *tb[], u32 portid, 1642 u32 seq, 1643 struct netlink_ext_ack *extack); 1644 int (*ndo_bridge_setlink)(struct net_device *dev, 1645 struct nlmsghdr *nlh, 1646 u16 flags, 1647 struct netlink_ext_ack *extack); 1648 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1649 u32 pid, u32 seq, 1650 struct net_device *dev, 1651 u32 filter_mask, 1652 int nlflags); 1653 int (*ndo_bridge_dellink)(struct net_device *dev, 1654 struct nlmsghdr *nlh, 1655 u16 flags); 1656 int (*ndo_change_carrier)(struct net_device *dev, 1657 bool new_carrier); 1658 int (*ndo_get_phys_port_id)(struct net_device *dev, 1659 struct netdev_phys_item_id *ppid); 1660 int (*ndo_get_port_parent_id)(struct net_device *dev, 1661 struct netdev_phys_item_id *ppid); 1662 int (*ndo_get_phys_port_name)(struct net_device *dev, 1663 char *name, size_t len); 1664 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1665 struct net_device *dev); 1666 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1667 void *priv); 1668 1669 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1670 int queue_index, 1671 u32 maxrate); 1672 int (*ndo_get_iflink)(const struct net_device *dev); 1673 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1674 struct sk_buff *skb); 1675 void (*ndo_set_rx_headroom)(struct net_device *dev, 1676 int needed_headroom); 1677 int (*ndo_bpf)(struct net_device *dev, 1678 struct netdev_bpf *bpf); 1679 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1680 struct xdp_frame **xdp, 1681 u32 flags); 1682 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1683 struct xdp_buff *xdp); 1684 int (*ndo_xsk_wakeup)(struct net_device *dev, 1685 u32 queue_id, u32 flags); 1686 int (*ndo_tunnel_ctl)(struct net_device *dev, 1687 struct ip_tunnel_parm_kern *p, 1688 int cmd); 1689 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1690 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1691 struct net_device_path *path); 1692 ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1693 const struct skb_shared_hwtstamps *hwtstamps, 1694 bool cycles); 1695 int (*ndo_hwtstamp_get)(struct net_device *dev, 1696 struct kernel_hwtstamp_config *kernel_config); 1697 int (*ndo_hwtstamp_set)(struct net_device *dev, 1698 struct kernel_hwtstamp_config *kernel_config, 1699 struct netlink_ext_ack *extack); 1700 1701 #if IS_ENABLED(CONFIG_NET_SHAPER) 1702 /** 1703 * @net_shaper_ops: Device shaping offload operations 1704 * see include/net/net_shapers.h 1705 */ 1706 const struct net_shaper_ops *net_shaper_ops; 1707 #endif 1708 }; 1709 1710 /** 1711 * enum netdev_priv_flags - &struct net_device priv_flags 1712 * 1713 * These are the &struct net_device, they are only set internally 1714 * by drivers and used in the kernel. These flags are invisible to 1715 * userspace; this means that the order of these flags can change 1716 * during any kernel release. 1717 * 1718 * You should add bitfield booleans after either net_device::priv_flags 1719 * (hotpath) or ::threaded (slowpath) instead of extending these flags. 1720 * 1721 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1722 * @IFF_EBRIDGE: Ethernet bridging device 1723 * @IFF_BONDING: bonding master or slave 1724 * @IFF_ISATAP: ISATAP interface (RFC4214) 1725 * @IFF_WAN_HDLC: WAN HDLC device 1726 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1727 * release skb->dst 1728 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1729 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1730 * @IFF_MACVLAN_PORT: device used as macvlan port 1731 * @IFF_BRIDGE_PORT: device used as bridge port 1732 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1733 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1734 * @IFF_UNICAST_FLT: Supports unicast filtering 1735 * @IFF_TEAM_PORT: device used as team port 1736 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1737 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1738 * change when it's running 1739 * @IFF_MACVLAN: Macvlan device 1740 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1741 * underlying stacked devices 1742 * @IFF_L3MDEV_MASTER: device is an L3 master device 1743 * @IFF_NO_QUEUE: device can run without qdisc attached 1744 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1745 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1746 * @IFF_TEAM: device is a team device 1747 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1748 * entity (i.e. the master device for bridged veth) 1749 * @IFF_MACSEC: device is a MACsec device 1750 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1751 * @IFF_FAILOVER: device is a failover master device 1752 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1753 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1754 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf 1755 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1756 * skb_headlen(skb) == 0 (data starts from frag0) 1757 */ 1758 enum netdev_priv_flags { 1759 IFF_802_1Q_VLAN = 1<<0, 1760 IFF_EBRIDGE = 1<<1, 1761 IFF_BONDING = 1<<2, 1762 IFF_ISATAP = 1<<3, 1763 IFF_WAN_HDLC = 1<<4, 1764 IFF_XMIT_DST_RELEASE = 1<<5, 1765 IFF_DONT_BRIDGE = 1<<6, 1766 IFF_DISABLE_NETPOLL = 1<<7, 1767 IFF_MACVLAN_PORT = 1<<8, 1768 IFF_BRIDGE_PORT = 1<<9, 1769 IFF_OVS_DATAPATH = 1<<10, 1770 IFF_TX_SKB_SHARING = 1<<11, 1771 IFF_UNICAST_FLT = 1<<12, 1772 IFF_TEAM_PORT = 1<<13, 1773 IFF_SUPP_NOFCS = 1<<14, 1774 IFF_LIVE_ADDR_CHANGE = 1<<15, 1775 IFF_MACVLAN = 1<<16, 1776 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1777 IFF_L3MDEV_MASTER = 1<<18, 1778 IFF_NO_QUEUE = 1<<19, 1779 IFF_OPENVSWITCH = 1<<20, 1780 IFF_L3MDEV_SLAVE = 1<<21, 1781 IFF_TEAM = 1<<22, 1782 IFF_PHONY_HEADROOM = 1<<24, 1783 IFF_MACSEC = 1<<25, 1784 IFF_NO_RX_HANDLER = 1<<26, 1785 IFF_FAILOVER = 1<<27, 1786 IFF_FAILOVER_SLAVE = 1<<28, 1787 IFF_L3MDEV_RX_HANDLER = 1<<29, 1788 IFF_NO_ADDRCONF = BIT_ULL(30), 1789 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31), 1790 }; 1791 1792 /* Specifies the type of the struct net_device::ml_priv pointer */ 1793 enum netdev_ml_priv_type { 1794 ML_PRIV_NONE, 1795 ML_PRIV_CAN, 1796 }; 1797 1798 enum netdev_stat_type { 1799 NETDEV_PCPU_STAT_NONE, 1800 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */ 1801 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */ 1802 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */ 1803 }; 1804 1805 enum netmem_tx_mode { 1806 NETMEM_TX_NONE, /* no netmem TX support */ 1807 NETMEM_TX_DMA, /* DMA-capable netmem TX (real HW) */ 1808 NETMEM_TX_NO_DMA, /* no DMA, e.g. passthrough for virtual devs */ 1809 }; 1810 1811 enum netdev_reg_state { 1812 NETREG_UNINITIALIZED = 0, 1813 NETREG_REGISTERED, /* completed register_netdevice */ 1814 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1815 NETREG_UNREGISTERED, /* completed unregister todo */ 1816 NETREG_RELEASED, /* called free_netdev */ 1817 NETREG_DUMMY, /* dummy device for NAPI poll */ 1818 }; 1819 1820 /** 1821 * struct net_device - The DEVICE structure. 1822 * 1823 * Actually, this whole structure is a big mistake. It mixes I/O 1824 * data with strictly "high-level" data, and it has to know about 1825 * almost every data structure used in the INET module. 1826 * 1827 * @priv_flags: flags invisible to userspace defined as bits, see 1828 * enum netdev_priv_flags for the definitions 1829 * @lltx: device supports lockless Tx. Deprecated for real HW 1830 * drivers. Mainly used by logical interfaces, such as 1831 * bonding and tunnels 1832 * @netmem_tx: device netmem TX mode 1833 * 1834 * @name: This is the first field of the "visible" part of this structure 1835 * (i.e. as seen by users in the "Space.c" file). It is the name 1836 * of the interface. 1837 * 1838 * @name_node: Name hashlist node 1839 * @ifalias: SNMP alias 1840 * @mem_end: Shared memory end 1841 * @mem_start: Shared memory start 1842 * @base_addr: Device I/O address 1843 * @irq: Device IRQ number 1844 * 1845 * @state: Generic network queuing layer state, see netdev_state_t 1846 * @dev_list: The global list of network devices 1847 * @napi_list: List entry used for polling NAPI devices 1848 * @unreg_list: List entry when we are unregistering the 1849 * device; see the function unregister_netdev 1850 * @close_list: List entry used when we are closing the device 1851 * @ptype_all: Device-specific packet handlers for all protocols 1852 * @ptype_specific: Device-specific, protocol-specific packet handlers 1853 * 1854 * @adj_list: Directly linked devices, like slaves for bonding 1855 * @features: Currently active device features 1856 * @hw_features: User-changeable features 1857 * 1858 * @wanted_features: User-requested features 1859 * @vlan_features: Mask of features inheritable by VLAN devices 1860 * 1861 * @hw_enc_features: Mask of features inherited by encapsulating devices 1862 * This field indicates what encapsulation 1863 * offloads the hardware is capable of doing, 1864 * and drivers will need to set them appropriately. 1865 * 1866 * @mpls_features: Mask of features inheritable by MPLS 1867 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1868 * @mangleid_features: Mask of features requiring MANGLEID, will be 1869 * disabled together with the latter. 1870 * 1871 * @ifindex: interface index 1872 * @group: The group the device belongs to 1873 * 1874 * @stats: Statistics struct, which was left as a legacy, use 1875 * rtnl_link_stats64 instead 1876 * 1877 * @core_stats: core networking counters, 1878 * do not use this in drivers 1879 * @carrier_up_count: Number of times the carrier has been up 1880 * @carrier_down_count: Number of times the carrier has been down 1881 * 1882 * @wireless_handlers: List of functions to handle Wireless Extensions, 1883 * instead of ioctl, 1884 * see <net/iw_handler.h> for details. 1885 * 1886 * @netdev_ops: Includes several pointers to callbacks, 1887 * if one wants to override the ndo_*() functions 1888 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks. 1889 * @xsk_tx_metadata_ops: Includes pointers to AF_XDP TX metadata callbacks. 1890 * @ethtool_ops: Management operations 1891 * @l3mdev_ops: Layer 3 master device operations 1892 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1893 * discovery handling. Necessary for e.g. 6LoWPAN. 1894 * @xfrmdev_ops: Transformation offload operations 1895 * @tlsdev_ops: Transport Layer Security offload operations 1896 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1897 * of Layer 2 headers. 1898 * 1899 * @flags: Interface flags (a la BSD) 1900 * @xdp_features: XDP capability supported by the device 1901 * @gflags: Global flags ( kept as legacy ) 1902 * @priv_len: Size of the ->priv flexible array 1903 * @priv: Flexible array containing private data 1904 * @operstate: RFC2863 operstate 1905 * @link_mode: Mapping policy to operstate 1906 * @if_port: Selectable AUI, TP, ... 1907 * @dma: DMA channel 1908 * @mtu: Interface MTU value 1909 * @min_mtu: Interface Minimum MTU value 1910 * @max_mtu: Interface Maximum MTU value 1911 * @type: Interface hardware type 1912 * @hard_header_len: Maximum hardware header length. 1913 * @min_header_len: Minimum hardware header length 1914 * 1915 * @needed_headroom: Extra headroom the hardware may need, but not in all 1916 * cases can this be guaranteed 1917 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1918 * cases can this be guaranteed. Some cases also use 1919 * LL_MAX_HEADER instead to allocate the skb 1920 * 1921 * interface address info: 1922 * 1923 * @perm_addr: Permanent hw address 1924 * @addr_assign_type: Hw address assignment type 1925 * @addr_len: Hardware address length 1926 * @upper_level: Maximum depth level of upper devices. 1927 * @lower_level: Maximum depth level of lower devices. 1928 * @threaded: napi threaded state. 1929 * @neigh_priv_len: Used in neigh_alloc() 1930 * @dev_id: Used to differentiate devices that share 1931 * the same link layer address 1932 * @dev_port: Used to differentiate devices that share 1933 * the same function 1934 * @addr_list_lock: XXX: need comments on this one 1935 * @name_assign_type: network interface name assignment type 1936 * @uc_promisc: Counter that indicates promiscuous mode 1937 * has been enabled due to the need to listen to 1938 * additional unicast addresses in a device that 1939 * does not implement ndo_set_rx_mode() 1940 * @work_node: List entry for async netdev_work processing 1941 * @work_tracker: Refcount tracker for async netdev_work 1942 * @work_pending: Driver-defined pending netdev_work, passed to 1943 * ndo_work() (see netdev_work_sched()) 1944 * @work_core_pending: Core-defined pending netdev_work (NETDEV_WORK_*) 1945 * @rx_mode_addr_cache: Recycled snapshot entries for rx_mode work 1946 * @rx_mode_retry_timer: Timer that re-queues rx_mode work after failure 1947 * @rx_mode_retry_count: Number of consecutive retries already scheduled 1948 * @uc: unicast mac addresses 1949 * @mc: multicast mac addresses 1950 * @dev_addrs: list of device hw addresses 1951 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1952 * @promiscuity: Number of times the NIC is told to work in 1953 * promiscuous mode; if it becomes 0 the NIC will 1954 * exit promiscuous mode 1955 * @allmulti: Counter, enables or disables allmulticast mode 1956 * 1957 * @vlan_info: VLAN info 1958 * @dsa_ptr: dsa specific data 1959 * @tipc_ptr: TIPC specific data 1960 * @ip_ptr: IPv4 specific data 1961 * @ip6_ptr: IPv6 specific data 1962 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1963 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1964 * device struct 1965 * @mpls_ptr: mpls_dev struct pointer 1966 * @mctp_ptr: MCTP specific data 1967 * @psp_dev: PSP crypto device registered for this netdev 1968 * 1969 * @dev_addr: Hw address (before bcast, 1970 * because most packets are unicast) 1971 * 1972 * @_rx: Array of RX queues 1973 * @num_rx_queues: Number of RX queues 1974 * allocated at register_netdev() time 1975 * @real_num_rx_queues: Number of RX queues currently active in device 1976 * @xdp_prog: XDP sockets filter program pointer 1977 * 1978 * @rx_handler: handler for received packets 1979 * @rx_handler_data: XXX: need comments on this one 1980 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing 1981 * @ingress_queue: XXX: need comments on this one 1982 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1983 * @broadcast: hw bcast address 1984 * 1985 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1986 * indexed by RX queue number. Assigned by driver. 1987 * This must only be set if the ndo_rx_flow_steer 1988 * operation is defined 1989 * @index_hlist: Device index hash chain 1990 * 1991 * @_tx: Array of TX queues 1992 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1993 * @real_num_tx_queues: Number of TX queues currently active in device 1994 * @qdisc: Root qdisc from userspace point of view 1995 * @tx_queue_len: Max frames per queue allowed 1996 * @tx_global_lock: XXX: need comments on this one 1997 * @xdp_bulkq: XDP device bulk queue 1998 * @xps_maps: all CPUs/RXQs maps for XPS device 1999 * 2000 * @xps_maps: XXX: need comments on this one 2001 * @tcx_egress: BPF & clsact qdisc specific data for egress processing 2002 * @nf_hooks_egress: netfilter hooks executed for egress packets 2003 * @qdisc_hash: qdisc hash table 2004 * @watchdog_timeo: Represents the timeout that is used by 2005 * the watchdog (see dev_watchdog()) 2006 * @watchdog_lock: protect watchdog_ref_held 2007 * @watchdog_ref_held: True if the watchdog device ref is taken. 2008 * @watchdog_timer: List of timers 2009 * 2010 * @proto_down_reason: reason a netdev interface is held down 2011 * @pcpu_refcnt: Number of references to this device 2012 * @dev_refcnt: Number of references to this device 2013 * @refcnt_tracker: Tracker directory for tracked references to this device 2014 * @todo_list: Delayed register/unregister 2015 * @link_watch_list: XXX: need comments on this one 2016 * 2017 * @reg_state: Register/unregister state machine 2018 * @dismantle: Device is going to be freed 2019 * @needs_free_netdev: Should unregister perform free_netdev? 2020 * @priv_destructor: Called from unregister 2021 * @npinfo: XXX: need comments on this one 2022 * @nd_net: Network namespace this network device is inside 2023 * protected by @lock 2024 * 2025 * @ml_priv: Mid-layer private 2026 * @ml_priv_type: Mid-layer private type 2027 * 2028 * @pcpu_stat_type: Type of device statistics which the core should 2029 * allocate/free: none, lstats, tstats, dstats. none 2030 * means the driver is handling statistics allocation/ 2031 * freeing internally. 2032 * @lstats: Loopback statistics: packets, bytes 2033 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes 2034 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes 2035 * 2036 * @garp_port: GARP 2037 * @mrp_port: MRP 2038 * 2039 * @dm_private: Drop monitor private 2040 * 2041 * @dev: Class/net/name entry 2042 * @sysfs_groups: Space for optional device, statistics and wireless 2043 * sysfs groups 2044 * 2045 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 2046 * @rtnl_link_ops: Rtnl_link_ops 2047 * @stat_ops: Optional ops for queue-aware statistics 2048 * @queue_mgmt_ops: Optional ops for queue management 2049 * 2050 * @gso_max_size: Maximum size of generic segmentation offload 2051 * @tso_max_size: Device (as in HW) limit on the max TSO request size 2052 * @gso_max_segs: Maximum number of segments that can be passed to the 2053 * NIC for GSO 2054 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count 2055 * @gso_ipv4_max_size: Maximum size of generic segmentation offload, 2056 * for IPv4. 2057 * 2058 * @dcbnl_ops: Data Center Bridging netlink ops 2059 * @num_tc: Number of traffic classes in the net device 2060 * @tc_to_txq: XXX: need comments on this one 2061 * @prio_tc_map: XXX: need comments on this one 2062 * 2063 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 2064 * 2065 * @priomap: XXX: need comments on this one 2066 * @link_topo: Physical link topology tracking attached PHYs 2067 * @phydev: Physical device may attach itself 2068 * for hardware timestamping 2069 * @sfp_bus: attached &struct sfp_bus structure. 2070 * 2071 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 2072 * 2073 * @proto_down: protocol port state information can be sent to the 2074 * switch driver and used to set the phys state of the 2075 * switch port. 2076 * 2077 * @irq_affinity_auto: driver wants the core to store and re-assign the IRQ 2078 * affinity. Set by netif_enable_irq_affinity(), then 2079 * the driver must create a persistent napi by 2080 * netif_napi_add_config() and finally bind the napi to 2081 * IRQ (via netif_napi_set_irq()). 2082 * 2083 * @rx_cpu_rmap_auto: driver wants the core to manage the ARFS rmap. 2084 * Set by calling netif_enable_cpu_rmap(). 2085 * 2086 * @see_all_hwtstamp_requests: device wants to see calls to 2087 * ndo_hwtstamp_set() for all timestamp requests 2088 * regardless of source, even if those aren't 2089 * HWTSTAMP_SOURCE_NETDEV 2090 * @change_proto_down: device supports setting carrier via IFLA_PROTO_DOWN 2091 * @netns_immutable: interface can't change network namespaces 2092 * @fcoe_mtu: device supports maximum FCoE MTU, 2158 bytes 2093 * 2094 * @net_notifier_list: List of per-net netdev notifier block 2095 * that follow this device when it is moved 2096 * to another network namespace. 2097 * 2098 * @macsec_ops: MACsec offloading ops 2099 * 2100 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 2101 * offload capabilities of the device 2102 * @udp_tunnel_nic: UDP tunnel offload state 2103 * @ethtool: ethtool related state 2104 * @xdp_state: stores info on attached XDP BPF programs 2105 * 2106 * @nested_level: Used as a parameter of spin_lock_nested() of 2107 * dev->addr_list_lock. 2108 * @unlink_list: As netif_addr_lock() can be called recursively, 2109 * keep a list of interfaces to be deleted. 2110 * @gro_max_size: Maximum size of aggregated packet in generic 2111 * receive offload (GRO) 2112 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic 2113 * receive offload (GRO), for IPv4. 2114 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP 2115 * zero copy driver 2116 * 2117 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 2118 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 2119 * @watchdog_dev_tracker: refcount tracker used by watchdog. 2120 * @dev_registered_tracker: tracker for reference held while 2121 * registered 2122 * @offload_xstats_l3: L3 HW stats for this netdevice. 2123 * 2124 * @devlink_port: Pointer to related devlink port structure. 2125 * Assigned by a driver before netdev registration using 2126 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static 2127 * during the time netdevice is registered. 2128 * 2129 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem, 2130 * where the clock is recovered. 2131 * 2132 * @max_pacing_offload_horizon: max EDT offload horizon in nsec. 2133 * @napi_config: An array of napi_config structures containing per-NAPI 2134 * settings. 2135 * @num_napi_configs: number of allocated NAPI config structs, 2136 * always >= max(num_rx_queues, num_tx_queues). 2137 * @gro_flush_timeout: timeout for GRO layer in NAPI 2138 * @napi_defer_hard_irqs: If not zero, provides a counter that would 2139 * allow to avoid NIC hard IRQ, on busy queues. 2140 * 2141 * @neighbours: List heads pointing to this device's neighbours' 2142 * dev_list, one per address-family. 2143 * @hwprov: Tracks which PTP performs hardware packet time stamping. 2144 * 2145 * FIXME: cleanup struct net_device such that network protocol info 2146 * moves out. 2147 */ 2148 2149 struct net_device { 2150 /* Cacheline organization can be found documented in 2151 * Documentation/networking/net_cachelines/net_device.rst. 2152 * Please update the document when adding new fields. 2153 */ 2154 2155 /* TX read-mostly hotpath */ 2156 __cacheline_group_begin(net_device_read_tx); 2157 struct_group(priv_flags_fast, 2158 unsigned long priv_flags:32; 2159 unsigned long lltx:1; 2160 unsigned long netmem_tx:2; 2161 ); 2162 const struct net_device_ops *netdev_ops; 2163 const struct header_ops *header_ops; 2164 struct netdev_queue *_tx; 2165 netdev_features_t gso_partial_features; 2166 unsigned int real_num_tx_queues; 2167 unsigned int gso_max_size; 2168 unsigned int gso_ipv4_max_size; 2169 u16 gso_max_segs; 2170 s16 num_tc; 2171 /* Note : dev->mtu is often read without holding a lock. 2172 * Writers usually hold RTNL. 2173 * It is recommended to use READ_ONCE() to annotate the reads, 2174 * and to use WRITE_ONCE() to annotate the writes. 2175 */ 2176 unsigned int mtu; 2177 unsigned short needed_headroom; 2178 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2179 #ifdef CONFIG_XPS 2180 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2181 #endif 2182 #ifdef CONFIG_NETFILTER_EGRESS 2183 struct nf_hook_entries __rcu *nf_hooks_egress; 2184 #endif 2185 #ifdef CONFIG_NET_XGRESS 2186 struct bpf_mprog_entry __rcu *tcx_egress; 2187 #endif 2188 __cacheline_group_end(net_device_read_tx); 2189 2190 /* TXRX read-mostly hotpath */ 2191 __cacheline_group_begin(net_device_read_txrx); 2192 union { 2193 struct pcpu_lstats __percpu *lstats; 2194 struct pcpu_sw_netstats __percpu *tstats; 2195 struct pcpu_dstats __percpu *dstats; 2196 }; 2197 unsigned long state; 2198 unsigned int flags; 2199 unsigned short hard_header_len; 2200 enum netdev_stat_type pcpu_stat_type:8; 2201 netdev_features_t features; 2202 struct inet6_dev __rcu *ip6_ptr; 2203 __cacheline_group_end(net_device_read_txrx); 2204 2205 /* RX read-mostly hotpath */ 2206 __cacheline_group_begin(net_device_read_rx); 2207 struct bpf_prog __rcu *xdp_prog; 2208 struct list_head ptype_specific; 2209 int ifindex; 2210 unsigned int real_num_rx_queues; 2211 struct netdev_rx_queue *_rx; 2212 unsigned int gro_max_size; 2213 unsigned int gro_ipv4_max_size; 2214 rx_handler_func_t __rcu *rx_handler; 2215 void __rcu *rx_handler_data; 2216 possible_net_t nd_net; 2217 #ifdef CONFIG_NETPOLL 2218 struct netpoll_info __rcu *npinfo; 2219 #endif 2220 #ifdef CONFIG_NET_XGRESS 2221 struct bpf_mprog_entry __rcu *tcx_ingress; 2222 #endif 2223 __cacheline_group_end(net_device_read_rx); 2224 2225 char name[IFNAMSIZ]; 2226 struct netdev_name_node *name_node; 2227 struct dev_ifalias __rcu *ifalias; 2228 /* 2229 * I/O specific fields 2230 * FIXME: Merge these and struct ifmap into one 2231 */ 2232 unsigned long mem_end; 2233 unsigned long mem_start; 2234 unsigned long base_addr; 2235 2236 /* 2237 * Some hardware also needs these fields (state,dev_list, 2238 * napi_list,unreg_list,close_list) but they are not 2239 * part of the usual set specified in Space.c. 2240 */ 2241 2242 2243 struct list_head dev_list; 2244 struct list_head napi_list; 2245 struct list_head unreg_list; 2246 struct list_head close_list; 2247 struct list_head ptype_all; 2248 2249 struct { 2250 struct list_head upper; 2251 struct list_head lower; 2252 } adj_list; 2253 2254 /* Read-mostly cache-line for fast-path access */ 2255 xdp_features_t xdp_features; 2256 const struct xdp_metadata_ops *xdp_metadata_ops; 2257 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops; 2258 unsigned short gflags; 2259 2260 unsigned short needed_tailroom; 2261 2262 netdev_features_t hw_features; 2263 netdev_features_t wanted_features; 2264 netdev_features_t vlan_features; 2265 netdev_features_t hw_enc_features; 2266 netdev_features_t mpls_features; 2267 netdev_features_t mangleid_features; 2268 2269 unsigned int min_mtu; 2270 unsigned int max_mtu; 2271 unsigned short type; 2272 unsigned char min_header_len; 2273 unsigned char name_assign_type; 2274 2275 int group; 2276 2277 struct net_device_stats stats; /* not used by modern drivers */ 2278 2279 struct net_device_core_stats __percpu *core_stats; 2280 2281 /* Stats to monitor link on/off, flapping */ 2282 atomic_t carrier_up_count; 2283 atomic_t carrier_down_count; 2284 2285 #ifdef CONFIG_WIRELESS_EXT 2286 const struct iw_handler_def *wireless_handlers; 2287 #endif 2288 const struct ethtool_ops *ethtool_ops; 2289 #ifdef CONFIG_NET_L3_MASTER_DEV 2290 const struct l3mdev_ops *l3mdev_ops; 2291 #endif 2292 #if IS_ENABLED(CONFIG_IPV6) 2293 const struct ndisc_ops *ndisc_ops; 2294 #endif 2295 2296 #ifdef CONFIG_XFRM_OFFLOAD 2297 const struct xfrmdev_ops *xfrmdev_ops; 2298 #endif 2299 2300 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2301 const struct tlsdev_ops *tlsdev_ops; 2302 #endif 2303 2304 unsigned int operstate; 2305 unsigned char link_mode; 2306 2307 unsigned char if_port; 2308 unsigned char dma; 2309 2310 /* Interface address info. */ 2311 unsigned char perm_addr[MAX_ADDR_LEN]; 2312 unsigned char addr_assign_type; 2313 unsigned char addr_len; 2314 unsigned char upper_level; 2315 unsigned char lower_level; 2316 u8 threaded; 2317 2318 unsigned short neigh_priv_len; 2319 unsigned short dev_id; 2320 unsigned short dev_port; 2321 int irq; 2322 u32 priv_len; 2323 2324 spinlock_t addr_list_lock; 2325 2326 struct netdev_hw_addr_list uc; 2327 struct netdev_hw_addr_list mc; 2328 struct netdev_hw_addr_list dev_addrs; 2329 2330 #ifdef CONFIG_SYSFS 2331 struct kset *queues_kset; 2332 #endif 2333 #ifdef CONFIG_LOCKDEP 2334 struct list_head unlink_list; 2335 #endif 2336 unsigned int promiscuity; 2337 unsigned int allmulti; 2338 bool uc_promisc; 2339 struct list_head work_node; 2340 netdevice_tracker work_tracker; 2341 unsigned long work_pending; 2342 unsigned long work_core_pending; 2343 struct netdev_hw_addr_list rx_mode_addr_cache; 2344 struct timer_list rx_mode_retry_timer; 2345 unsigned int rx_mode_retry_count; 2346 #ifdef CONFIG_LOCKDEP 2347 unsigned char nested_level; 2348 #endif 2349 2350 2351 /* Protocol-specific pointers */ 2352 struct in_device __rcu *ip_ptr; 2353 /** @fib_nh_head: nexthops associated with this netdev */ 2354 struct hlist_head fib_nh_head; 2355 2356 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2357 struct vlan_info __rcu *vlan_info; 2358 #endif 2359 #if IS_ENABLED(CONFIG_NET_DSA) 2360 struct dsa_port *dsa_ptr; 2361 #endif 2362 #if IS_ENABLED(CONFIG_TIPC) 2363 struct tipc_bearer __rcu *tipc_ptr; 2364 #endif 2365 #if IS_ENABLED(CONFIG_CFG80211) 2366 struct wireless_dev *ieee80211_ptr; 2367 #endif 2368 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2369 struct wpan_dev *ieee802154_ptr; 2370 #endif 2371 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2372 struct mpls_dev __rcu *mpls_ptr; 2373 #endif 2374 #if IS_ENABLED(CONFIG_MCTP) 2375 struct mctp_dev __rcu *mctp_ptr; 2376 #endif 2377 #if IS_ENABLED(CONFIG_INET_PSP) 2378 struct psp_dev __rcu *psp_dev; 2379 #endif 2380 2381 /* 2382 * Cache lines mostly used on receive path (including eth_type_trans()) 2383 */ 2384 /* Interface address info used in eth_type_trans() */ 2385 const unsigned char *dev_addr; 2386 2387 unsigned int num_rx_queues; 2388 #define GRO_LEGACY_MAX_SIZE 65536u 2389 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2390 * and shinfo->gso_segs is a 16bit field. 2391 */ 2392 #define GRO_MAX_SIZE (8 * 65535u) 2393 unsigned int xdp_zc_max_segs; 2394 struct netdev_queue __rcu *ingress_queue; 2395 #ifdef CONFIG_NETFILTER_INGRESS 2396 struct nf_hook_entries __rcu *nf_hooks_ingress; 2397 #endif 2398 2399 unsigned char broadcast[MAX_ADDR_LEN]; 2400 #ifdef CONFIG_RFS_ACCEL 2401 struct cpu_rmap *rx_cpu_rmap; 2402 #endif 2403 struct hlist_node index_hlist; 2404 2405 /* 2406 * Cache lines mostly used on transmit path 2407 */ 2408 unsigned int num_tx_queues; 2409 struct Qdisc __rcu *qdisc; 2410 unsigned int tx_queue_len; 2411 spinlock_t tx_global_lock; 2412 2413 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2414 2415 #ifdef CONFIG_NET_SCHED 2416 DECLARE_HASHTABLE (qdisc_hash, 4); 2417 #endif 2418 /* These may be needed for future network-power-down code. */ 2419 struct timer_list watchdog_timer; 2420 int watchdog_timeo; 2421 spinlock_t watchdog_lock; 2422 bool watchdog_ref_held; 2423 2424 u32 proto_down_reason; 2425 2426 struct list_head todo_list; 2427 2428 #ifdef CONFIG_PCPU_DEV_REFCNT 2429 int __percpu *pcpu_refcnt; 2430 #else 2431 refcount_t dev_refcnt; 2432 #endif 2433 struct ref_tracker_dir refcnt_tracker; 2434 2435 struct list_head link_watch_list; 2436 2437 u8 reg_state; 2438 2439 bool dismantle; 2440 2441 /** @moving_ns: device is changing netns, protected by @lock */ 2442 bool moving_ns; 2443 /** @rtnl_link_initializing: Device being created, suppress events */ 2444 bool rtnl_link_initializing; 2445 2446 bool needs_free_netdev; 2447 void (*priv_destructor)(struct net_device *dev); 2448 2449 /* mid-layer private */ 2450 void *ml_priv; 2451 enum netdev_ml_priv_type ml_priv_type; 2452 2453 #if IS_ENABLED(CONFIG_GARP) 2454 struct garp_port __rcu *garp_port; 2455 #endif 2456 #if IS_ENABLED(CONFIG_MRP) 2457 struct mrp_port __rcu *mrp_port; 2458 #endif 2459 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2460 struct dm_hw_stat_delta __rcu *dm_private; 2461 #endif 2462 struct device dev; 2463 const struct attribute_group *sysfs_groups[5]; 2464 const struct attribute_group *sysfs_rx_queue_group; 2465 2466 const struct rtnl_link_ops *rtnl_link_ops; 2467 2468 const struct netdev_stat_ops *stat_ops; 2469 2470 const struct netdev_queue_mgmt_ops *queue_mgmt_ops; 2471 2472 /* for setting kernel sock attribute on TCP connection setup */ 2473 #define GSO_MAX_SEGS 65535u 2474 #define GSO_LEGACY_MAX_SIZE 65536u 2475 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2476 * and shinfo->gso_segs is a 16bit field. 2477 */ 2478 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2479 2480 #define TSO_LEGACY_MAX_SIZE 65536 2481 #define TSO_MAX_SIZE UINT_MAX 2482 unsigned int tso_max_size; 2483 #define TSO_MAX_SEGS U16_MAX 2484 u16 tso_max_segs; 2485 2486 #ifdef CONFIG_DCB 2487 const struct dcbnl_rtnl_ops *dcbnl_ops; 2488 #endif 2489 u8 prio_tc_map[TC_BITMASK + 1]; 2490 2491 #if IS_ENABLED(CONFIG_FCOE) 2492 unsigned int fcoe_ddp_xid; 2493 #endif 2494 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2495 struct netprio_map __rcu *priomap; 2496 #endif 2497 struct phy_link_topology *link_topo; 2498 struct phy_device *phydev; 2499 struct sfp_bus *sfp_bus; 2500 struct lock_class_key *qdisc_tx_busylock; 2501 bool proto_down; 2502 bool irq_affinity_auto; 2503 bool rx_cpu_rmap_auto; 2504 2505 /* priv_flags_slow, ungrouped to save space */ 2506 unsigned long see_all_hwtstamp_requests:1; 2507 unsigned long change_proto_down:1; 2508 unsigned long netns_immutable:1; 2509 unsigned long fcoe_mtu:1; 2510 2511 struct list_head net_notifier_list; 2512 2513 #if IS_ENABLED(CONFIG_MACSEC) 2514 /* MACsec management functions */ 2515 const struct macsec_ops *macsec_ops; 2516 #endif 2517 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2518 struct udp_tunnel_nic *udp_tunnel_nic; 2519 2520 /** @cfg: net_device queue-related configuration */ 2521 struct netdev_config *cfg; 2522 /** 2523 * @cfg_pending: same as @cfg but when device is being actively 2524 * reconfigured includes any changes to the configuration 2525 * requested by the user, but which may or may not be rejected. 2526 */ 2527 struct netdev_config *cfg_pending; 2528 struct ethtool_netdev_state *ethtool; 2529 2530 /* protected by rtnl_lock */ 2531 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2532 2533 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2534 netdevice_tracker linkwatch_dev_tracker; 2535 netdevice_tracker watchdog_dev_tracker; 2536 netdevice_tracker dev_registered_tracker; 2537 struct rtnl_hw_stats64 *offload_xstats_l3; 2538 2539 struct devlink_port *devlink_port; 2540 2541 #if IS_ENABLED(CONFIG_DPLL) 2542 struct dpll_pin __rcu *dpll_pin; 2543 #endif 2544 #if IS_ENABLED(CONFIG_PAGE_POOL) 2545 /** @page_pools: page pools created for this netdevice */ 2546 struct hlist_head page_pools; 2547 #endif 2548 2549 /** @irq_moder: dim parameters used if IS_ENABLED(CONFIG_DIMLIB). */ 2550 struct dim_irq_moder *irq_moder; 2551 2552 u64 max_pacing_offload_horizon; 2553 struct napi_config *napi_config; 2554 u32 num_napi_configs; 2555 u32 napi_defer_hard_irqs; 2556 unsigned long gro_flush_timeout; 2557 2558 /** 2559 * @up: copy of @state's IFF_UP, but safe to read with just @lock. 2560 * May report false negatives while the device is being opened 2561 * or closed (@lock does not protect .ndo_open, or .ndo_close). 2562 */ 2563 bool up; 2564 2565 /** 2566 * @request_ops_lock: request the core to run all @netdev_ops and 2567 * @ethtool_ops under the @lock. 2568 */ 2569 bool request_ops_lock; 2570 2571 /** 2572 * @lock: netdev-scope lock, protects a small selection of fields. 2573 * Should always be taken using netdev_lock() / netdev_unlock() helpers. 2574 * Drivers are free to use it for other protection. 2575 * 2576 * For the drivers that implement shaper or queue API, the scope 2577 * of this lock is expanded to cover most ndo/queue/ethtool/sysfs 2578 * operations. Drivers may opt-in to this behavior by setting 2579 * @request_ops_lock. 2580 * 2581 * @lock protection mixes with rtnl_lock in multiple ways, fields are 2582 * either: 2583 * 2584 * - simply protected by the instance @lock; 2585 * 2586 * - double protected - writers hold both locks, readers hold either; 2587 * 2588 * - ops protected - protected by the lock held around the NDOs 2589 * and other callbacks, that is the instance lock on devices for 2590 * which netdev_need_ops_lock() returns true, otherwise by rtnl_lock; 2591 * 2592 * - double ops protected - always protected by rtnl_lock but for 2593 * devices for which netdev_need_ops_lock() returns true - also 2594 * the instance lock. 2595 * 2596 * Simply protects: 2597 * @gro_flush_timeout, @napi_defer_hard_irqs, @napi_list, 2598 * @net_shaper_hierarchy, @reg_state, @threaded 2599 * 2600 * Double protects: 2601 * @up, @moving_ns, @nd_net, @xdp_features 2602 * 2603 * Ops protects: 2604 * @cfg, @cfg_pending, @ethtool, @hwprov 2605 * 2606 * Double ops protects: 2607 * @real_num_rx_queues, @real_num_tx_queues 2608 * 2609 * Also protects some fields in: 2610 * struct napi_struct, struct netdev_queue, struct netdev_rx_queue 2611 * 2612 * Ordering: 2613 * 2614 * - take after rtnl_lock 2615 * 2616 * - for the case of netdev queue leasing, the netdev-scope lock is 2617 * taken for both the virtual and the physical device; to prevent 2618 * deadlocks, the virtual device's lock must always be acquired 2619 * before the physical device's (see netdev_nl_queue_create_doit) 2620 */ 2621 struct mutex lock; 2622 2623 #if IS_ENABLED(CONFIG_NET_SHAPER) 2624 /** 2625 * @net_shaper_hierarchy: data tracking the current shaper status 2626 * see include/net/net_shapers.h 2627 */ 2628 struct net_shaper_hierarchy *net_shaper_hierarchy; 2629 #endif 2630 2631 struct hlist_head neighbours[NEIGH_NR_TABLES]; 2632 2633 struct hwtstamp_provider __rcu *hwprov; 2634 2635 u8 priv[] ____cacheline_aligned 2636 __counted_by(priv_len); 2637 } ____cacheline_aligned; 2638 #define to_net_dev(d) container_of(d, struct net_device, dev) 2639 2640 /* 2641 * Driver should use this to assign devlink port instance to a netdevice 2642 * before it registers the netdevice. Therefore devlink_port is static 2643 * during the netdev lifetime after it is registered. 2644 */ 2645 #define SET_NETDEV_DEVLINK_PORT(dev, port) \ 2646 ({ \ 2647 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \ 2648 ((dev)->devlink_port = (port)); \ 2649 }) 2650 2651 static inline bool netif_elide_gro(const struct net_device *dev) 2652 { 2653 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2654 return true; 2655 return false; 2656 } 2657 2658 #define NETDEV_ALIGN 32 2659 2660 static inline 2661 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2662 { 2663 return dev->prio_tc_map[prio & TC_BITMASK]; 2664 } 2665 2666 static inline 2667 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2668 { 2669 if (tc >= dev->num_tc) 2670 return -EINVAL; 2671 2672 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2673 return 0; 2674 } 2675 2676 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2677 void netdev_reset_tc(struct net_device *dev); 2678 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2679 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2680 2681 static inline 2682 int netdev_get_num_tc(struct net_device *dev) 2683 { 2684 return dev->num_tc; 2685 } 2686 2687 static inline void net_prefetch(void *p) 2688 { 2689 prefetch(p); 2690 #if L1_CACHE_BYTES < 128 2691 prefetch((u8 *)p + L1_CACHE_BYTES); 2692 #endif 2693 } 2694 2695 static inline void net_prefetchw(void *p) 2696 { 2697 prefetchw(p); 2698 #if L1_CACHE_BYTES < 128 2699 prefetchw((u8 *)p + L1_CACHE_BYTES); 2700 #endif 2701 } 2702 2703 void netdev_unbind_sb_channel(struct net_device *dev, 2704 struct net_device *sb_dev); 2705 int netdev_bind_sb_channel_queue(struct net_device *dev, 2706 struct net_device *sb_dev, 2707 u8 tc, u16 count, u16 offset); 2708 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2709 static inline int netdev_get_sb_channel(struct net_device *dev) 2710 { 2711 return max_t(int, -dev->num_tc, 0); 2712 } 2713 2714 static inline 2715 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2716 unsigned int index) 2717 { 2718 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues); 2719 return &dev->_tx[index]; 2720 } 2721 2722 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2723 const struct sk_buff *skb) 2724 { 2725 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2726 } 2727 2728 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2729 void (*f)(struct net_device *, 2730 struct netdev_queue *, 2731 void *), 2732 void *arg) 2733 { 2734 unsigned int i; 2735 2736 for (i = 0; i < dev->num_tx_queues; i++) 2737 f(dev, &dev->_tx[i], arg); 2738 } 2739 2740 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2741 struct net_device *sb_dev); 2742 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2743 struct sk_buff *skb, 2744 struct net_device *sb_dev); 2745 2746 /* returns the headroom that the master device needs to take in account 2747 * when forwarding to this dev 2748 */ 2749 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2750 { 2751 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2752 } 2753 2754 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2755 { 2756 if (dev->netdev_ops->ndo_set_rx_headroom) 2757 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2758 } 2759 2760 /* set the device rx headroom to the dev's default */ 2761 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2762 { 2763 netdev_set_rx_headroom(dev, -1); 2764 } 2765 2766 static inline void *netdev_get_ml_priv(struct net_device *dev, 2767 enum netdev_ml_priv_type type) 2768 { 2769 if (dev->ml_priv_type != type) 2770 return NULL; 2771 2772 return dev->ml_priv; 2773 } 2774 2775 static inline void netdev_set_ml_priv(struct net_device *dev, 2776 void *ml_priv, 2777 enum netdev_ml_priv_type type) 2778 { 2779 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2780 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2781 dev->ml_priv_type, type); 2782 WARN(!dev->ml_priv_type && dev->ml_priv, 2783 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2784 2785 dev->ml_priv = ml_priv; 2786 dev->ml_priv_type = type; 2787 } 2788 2789 /* 2790 * Net namespace inlines 2791 */ 2792 static inline 2793 struct net *dev_net(const struct net_device *dev) 2794 { 2795 return read_pnet(&dev->nd_net); 2796 } 2797 2798 static inline 2799 struct net *dev_net_rcu(const struct net_device *dev) 2800 { 2801 return read_pnet_rcu(&dev->nd_net); 2802 } 2803 2804 static inline 2805 void dev_net_set(struct net_device *dev, struct net *net) 2806 { 2807 write_pnet(&dev->nd_net, net); 2808 } 2809 2810 /** 2811 * netdev_priv - access network device private data 2812 * @dev: network device 2813 * 2814 * Get network device private data 2815 */ 2816 static inline void *netdev_priv(const struct net_device *dev) 2817 { 2818 return (void *)dev->priv; 2819 } 2820 2821 /** 2822 * netdev_from_priv() - get network device from priv 2823 * @priv: network device private data 2824 * 2825 * Returns: net_device to which @priv belongs 2826 */ 2827 static inline struct net_device *netdev_from_priv(const void *priv) 2828 { 2829 return container_of(priv, struct net_device, priv); 2830 } 2831 2832 /* Set the sysfs physical device reference for the network logical device 2833 * if set prior to registration will cause a symlink during initialization. 2834 */ 2835 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2836 2837 /* Set the sysfs device type for the network logical device to allow 2838 * fine-grained identification of different network device types. For 2839 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2840 */ 2841 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2842 2843 void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index, 2844 enum netdev_queue_type type, 2845 struct napi_struct *napi); 2846 2847 static inline void netdev_lock(struct net_device *dev) 2848 { 2849 mutex_lock(&dev->lock); 2850 } 2851 2852 static inline void netdev_unlock(struct net_device *dev) 2853 { 2854 mutex_unlock(&dev->lock); 2855 } 2856 /* Additional netdev_lock()-related helpers are in net/netdev_lock.h */ 2857 2858 void netif_napi_set_irq_locked(struct napi_struct *napi, int irq); 2859 2860 static inline void netif_napi_set_irq(struct napi_struct *napi, int irq) 2861 { 2862 netdev_lock(napi->dev); 2863 netif_napi_set_irq_locked(napi, irq); 2864 netdev_unlock(napi->dev); 2865 } 2866 2867 /* Default NAPI poll() weight 2868 * Device drivers are strongly advised to not use bigger value 2869 */ 2870 #define NAPI_POLL_WEIGHT 64 2871 2872 void netif_napi_add_weight_locked(struct net_device *dev, 2873 struct napi_struct *napi, 2874 int (*poll)(struct napi_struct *, int), 2875 int weight); 2876 2877 static inline void 2878 netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2879 int (*poll)(struct napi_struct *, int), int weight) 2880 { 2881 netdev_lock(dev); 2882 netif_napi_add_weight_locked(dev, napi, poll, weight); 2883 netdev_unlock(dev); 2884 } 2885 2886 /** 2887 * netif_napi_add() - initialize a NAPI context 2888 * @dev: network device 2889 * @napi: NAPI context 2890 * @poll: polling function 2891 * 2892 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2893 * *any* of the other NAPI-related functions. 2894 */ 2895 static inline void 2896 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2897 int (*poll)(struct napi_struct *, int)) 2898 { 2899 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2900 } 2901 2902 static inline void 2903 netif_napi_add_locked(struct net_device *dev, struct napi_struct *napi, 2904 int (*poll)(struct napi_struct *, int)) 2905 { 2906 netif_napi_add_weight_locked(dev, napi, poll, NAPI_POLL_WEIGHT); 2907 } 2908 2909 static inline void 2910 netif_napi_add_tx_weight(struct net_device *dev, 2911 struct napi_struct *napi, 2912 int (*poll)(struct napi_struct *, int), 2913 int weight) 2914 { 2915 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2916 netif_napi_add_weight(dev, napi, poll, weight); 2917 } 2918 2919 static inline void 2920 netif_napi_add_config_locked(struct net_device *dev, struct napi_struct *napi, 2921 int (*poll)(struct napi_struct *, int), int index) 2922 { 2923 napi->index = index; 2924 napi->config = &dev->napi_config[index]; 2925 netif_napi_add_weight_locked(dev, napi, poll, NAPI_POLL_WEIGHT); 2926 } 2927 2928 /** 2929 * netif_napi_add_config - initialize a NAPI context with persistent config 2930 * @dev: network device 2931 * @napi: NAPI context 2932 * @poll: polling function 2933 * @index: the NAPI index 2934 */ 2935 static inline void 2936 netif_napi_add_config(struct net_device *dev, struct napi_struct *napi, 2937 int (*poll)(struct napi_struct *, int), int index) 2938 { 2939 netdev_lock(dev); 2940 netif_napi_add_config_locked(dev, napi, poll, index); 2941 netdev_unlock(dev); 2942 } 2943 2944 /** 2945 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2946 * @dev: network device 2947 * @napi: NAPI context 2948 * @poll: polling function 2949 * 2950 * This variant of netif_napi_add() should be used from drivers using NAPI 2951 * to exclusively poll a TX queue. 2952 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2953 */ 2954 static inline void netif_napi_add_tx(struct net_device *dev, 2955 struct napi_struct *napi, 2956 int (*poll)(struct napi_struct *, int)) 2957 { 2958 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2959 } 2960 2961 void __netif_napi_del_locked(struct napi_struct *napi); 2962 2963 /** 2964 * __netif_napi_del - remove a NAPI context 2965 * @napi: NAPI context 2966 * 2967 * Warning: caller must observe RCU grace period before freeing memory 2968 * containing @napi. Drivers might want to call this helper to combine 2969 * all the needed RCU grace periods into a single one. 2970 */ 2971 static inline void __netif_napi_del(struct napi_struct *napi) 2972 { 2973 netdev_lock(napi->dev); 2974 __netif_napi_del_locked(napi); 2975 netdev_unlock(napi->dev); 2976 } 2977 2978 static inline void netif_napi_del_locked(struct napi_struct *napi) 2979 { 2980 __netif_napi_del_locked(napi); 2981 synchronize_net(); 2982 } 2983 2984 /** 2985 * netif_napi_del - remove a NAPI context 2986 * @napi: NAPI context 2987 * 2988 * netif_napi_del() removes a NAPI context from the network device NAPI list 2989 */ 2990 static inline void netif_napi_del(struct napi_struct *napi) 2991 { 2992 __netif_napi_del(napi); 2993 synchronize_net(); 2994 } 2995 2996 int netif_enable_cpu_rmap(struct net_device *dev, unsigned int num_irqs); 2997 void netif_set_affinity_auto(struct net_device *dev); 2998 2999 struct packet_type { 3000 __be16 type; /* This is really htons(ether_type). */ 3001 bool ignore_outgoing; 3002 struct net_device *dev; /* NULL is wildcarded here */ 3003 netdevice_tracker dev_tracker; 3004 int (*func) (struct sk_buff *, 3005 struct net_device *, 3006 struct packet_type *, 3007 struct net_device *); 3008 void (*list_func) (struct list_head *, 3009 struct packet_type *, 3010 struct net_device *); 3011 bool (*id_match)(struct packet_type *ptype, 3012 struct sock *sk); 3013 struct net *af_packet_net; 3014 void *af_packet_priv; 3015 struct list_head list; 3016 }; 3017 3018 struct offload_callbacks { 3019 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 3020 netdev_features_t features); 3021 struct sk_buff *(*gro_receive)(struct list_head *head, 3022 struct sk_buff *skb); 3023 int (*gro_complete)(struct sk_buff *skb, int nhoff); 3024 }; 3025 3026 struct packet_offload { 3027 __be16 type; /* This is really htons(ether_type). */ 3028 u16 priority; 3029 struct offload_callbacks callbacks; 3030 struct list_head list; 3031 }; 3032 3033 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 3034 struct pcpu_sw_netstats { 3035 u64_stats_t rx_packets; 3036 u64_stats_t rx_bytes; 3037 u64_stats_t tx_packets; 3038 u64_stats_t tx_bytes; 3039 struct u64_stats_sync syncp; 3040 } __aligned(4 * sizeof(u64)); 3041 3042 struct pcpu_dstats { 3043 u64_stats_t rx_packets; 3044 u64_stats_t rx_bytes; 3045 u64_stats_t tx_packets; 3046 u64_stats_t tx_bytes; 3047 u64_stats_t rx_drops; 3048 u64_stats_t tx_drops; 3049 struct u64_stats_sync syncp; 3050 } __aligned(8 * sizeof(u64)); 3051 3052 struct pcpu_lstats { 3053 u64_stats_t packets; 3054 u64_stats_t bytes; 3055 struct u64_stats_sync syncp; 3056 } __aligned(2 * sizeof(u64)); 3057 3058 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 3059 3060 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 3061 { 3062 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 3063 3064 u64_stats_update_begin(&tstats->syncp); 3065 u64_stats_add(&tstats->rx_bytes, len); 3066 u64_stats_inc(&tstats->rx_packets); 3067 u64_stats_update_end(&tstats->syncp); 3068 } 3069 3070 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 3071 unsigned int packets, 3072 unsigned int len) 3073 { 3074 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 3075 3076 u64_stats_update_begin(&tstats->syncp); 3077 u64_stats_add(&tstats->tx_bytes, len); 3078 u64_stats_add(&tstats->tx_packets, packets); 3079 u64_stats_update_end(&tstats->syncp); 3080 } 3081 3082 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 3083 { 3084 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 3085 3086 u64_stats_update_begin(&lstats->syncp); 3087 u64_stats_add(&lstats->bytes, len); 3088 u64_stats_inc(&lstats->packets); 3089 u64_stats_update_end(&lstats->syncp); 3090 } 3091 3092 static inline void dev_dstats_rx_add(struct net_device *dev, 3093 unsigned int len) 3094 { 3095 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3096 3097 u64_stats_update_begin(&dstats->syncp); 3098 u64_stats_inc(&dstats->rx_packets); 3099 u64_stats_add(&dstats->rx_bytes, len); 3100 u64_stats_update_end(&dstats->syncp); 3101 } 3102 3103 static inline void dev_dstats_rx_dropped(struct net_device *dev) 3104 { 3105 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3106 3107 u64_stats_update_begin(&dstats->syncp); 3108 u64_stats_inc(&dstats->rx_drops); 3109 u64_stats_update_end(&dstats->syncp); 3110 } 3111 3112 static inline void dev_dstats_rx_dropped_add(struct net_device *dev, 3113 unsigned int packets) 3114 { 3115 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3116 3117 u64_stats_update_begin(&dstats->syncp); 3118 u64_stats_add(&dstats->rx_drops, packets); 3119 u64_stats_update_end(&dstats->syncp); 3120 } 3121 3122 static inline void dev_dstats_tx_add(struct net_device *dev, 3123 unsigned int len) 3124 { 3125 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3126 3127 u64_stats_update_begin(&dstats->syncp); 3128 u64_stats_inc(&dstats->tx_packets); 3129 u64_stats_add(&dstats->tx_bytes, len); 3130 u64_stats_update_end(&dstats->syncp); 3131 } 3132 3133 static inline void dev_dstats_tx_dropped(struct net_device *dev) 3134 { 3135 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3136 3137 u64_stats_update_begin(&dstats->syncp); 3138 u64_stats_inc(&dstats->tx_drops); 3139 u64_stats_update_end(&dstats->syncp); 3140 } 3141 3142 #define __netdev_alloc_pcpu_stats(type, gfp) \ 3143 ({ \ 3144 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 3145 if (pcpu_stats) { \ 3146 int __cpu; \ 3147 for_each_possible_cpu(__cpu) { \ 3148 typeof(type) *stat; \ 3149 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 3150 u64_stats_init(&stat->syncp); \ 3151 } \ 3152 } \ 3153 pcpu_stats; \ 3154 }) 3155 3156 #define netdev_alloc_pcpu_stats(type) \ 3157 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 3158 3159 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 3160 ({ \ 3161 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 3162 if (pcpu_stats) { \ 3163 int __cpu; \ 3164 for_each_possible_cpu(__cpu) { \ 3165 typeof(type) *stat; \ 3166 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 3167 u64_stats_init(&stat->syncp); \ 3168 } \ 3169 } \ 3170 pcpu_stats; \ 3171 }) 3172 3173 enum netdev_lag_tx_type { 3174 NETDEV_LAG_TX_TYPE_UNKNOWN, 3175 NETDEV_LAG_TX_TYPE_RANDOM, 3176 NETDEV_LAG_TX_TYPE_BROADCAST, 3177 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 3178 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 3179 NETDEV_LAG_TX_TYPE_HASH, 3180 }; 3181 3182 enum netdev_lag_hash { 3183 NETDEV_LAG_HASH_NONE, 3184 NETDEV_LAG_HASH_L2, 3185 NETDEV_LAG_HASH_L34, 3186 NETDEV_LAG_HASH_L23, 3187 NETDEV_LAG_HASH_E23, 3188 NETDEV_LAG_HASH_E34, 3189 NETDEV_LAG_HASH_VLAN_SRCMAC, 3190 NETDEV_LAG_HASH_UNKNOWN, 3191 }; 3192 3193 struct netdev_lag_upper_info { 3194 enum netdev_lag_tx_type tx_type; 3195 enum netdev_lag_hash hash_type; 3196 }; 3197 3198 struct netdev_lag_lower_state_info { 3199 u8 link_up : 1, 3200 tx_enabled : 1; 3201 }; 3202 3203 #include <linux/notifier.h> 3204 3205 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 3206 * and the rtnetlink notification exclusion list in rtnetlink_event() when 3207 * adding new types. 3208 */ 3209 enum netdev_cmd { 3210 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 3211 NETDEV_DOWN, 3212 NETDEV_REBOOT, /* Tell a protocol stack a network interface 3213 detected a hardware crash and restarted 3214 - we can use this eg to kick tcp sessions 3215 once done */ 3216 NETDEV_CHANGE, /* Notify device state change */ 3217 NETDEV_REGISTER, 3218 NETDEV_UNREGISTER, 3219 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 3220 NETDEV_CHANGEADDR, /* notify after the address change */ 3221 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 3222 NETDEV_GOING_DOWN, 3223 NETDEV_CHANGENAME, 3224 NETDEV_FEAT_CHANGE, 3225 NETDEV_BONDING_FAILOVER, 3226 NETDEV_PRE_UP, 3227 NETDEV_PRE_TYPE_CHANGE, 3228 NETDEV_POST_TYPE_CHANGE, 3229 NETDEV_POST_INIT, 3230 NETDEV_PRE_UNINIT, 3231 NETDEV_RELEASE, 3232 NETDEV_NOTIFY_PEERS, 3233 NETDEV_JOIN, 3234 NETDEV_CHANGEUPPER, 3235 NETDEV_RESEND_IGMP, 3236 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 3237 NETDEV_CHANGEINFODATA, 3238 NETDEV_BONDING_INFO, 3239 NETDEV_PRECHANGEUPPER, 3240 NETDEV_CHANGELOWERSTATE, 3241 NETDEV_UDP_TUNNEL_PUSH_INFO, 3242 NETDEV_UDP_TUNNEL_DROP_INFO, 3243 NETDEV_CHANGE_TX_QUEUE_LEN, 3244 NETDEV_CVLAN_FILTER_PUSH_INFO, 3245 NETDEV_CVLAN_FILTER_DROP_INFO, 3246 NETDEV_SVLAN_FILTER_PUSH_INFO, 3247 NETDEV_SVLAN_FILTER_DROP_INFO, 3248 NETDEV_OFFLOAD_XSTATS_ENABLE, 3249 NETDEV_OFFLOAD_XSTATS_DISABLE, 3250 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 3251 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 3252 NETDEV_XDP_FEAT_CHANGE, 3253 }; 3254 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 3255 3256 int register_netdevice_notifier(struct notifier_block *nb); 3257 int unregister_netdevice_notifier(struct notifier_block *nb); 3258 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 3259 int unregister_netdevice_notifier_net(struct net *net, 3260 struct notifier_block *nb); 3261 int register_netdevice_notifier_dev_net(struct net_device *dev, 3262 struct notifier_block *nb, 3263 struct netdev_net_notifier *nn); 3264 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 3265 struct notifier_block *nb, 3266 struct netdev_net_notifier *nn); 3267 3268 struct netdev_notifier_info { 3269 struct net_device *dev; 3270 struct netlink_ext_ack *extack; 3271 }; 3272 3273 struct netdev_notifier_info_ext { 3274 struct netdev_notifier_info info; /* must be first */ 3275 union { 3276 u32 mtu; 3277 } ext; 3278 }; 3279 3280 struct netdev_notifier_change_info { 3281 struct netdev_notifier_info info; /* must be first */ 3282 unsigned int flags_changed; 3283 }; 3284 3285 struct netdev_notifier_changeupper_info { 3286 struct netdev_notifier_info info; /* must be first */ 3287 struct net_device *upper_dev; /* new upper dev */ 3288 bool master; /* is upper dev master */ 3289 bool linking; /* is the notification for link or unlink */ 3290 void *upper_info; /* upper dev info */ 3291 }; 3292 3293 struct netdev_notifier_changelowerstate_info { 3294 struct netdev_notifier_info info; /* must be first */ 3295 void *lower_state_info; /* is lower dev state */ 3296 }; 3297 3298 struct netdev_notifier_pre_changeaddr_info { 3299 struct netdev_notifier_info info; /* must be first */ 3300 const unsigned char *dev_addr; 3301 }; 3302 3303 enum netdev_offload_xstats_type { 3304 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 3305 }; 3306 3307 struct netdev_notifier_offload_xstats_info { 3308 struct netdev_notifier_info info; /* must be first */ 3309 enum netdev_offload_xstats_type type; 3310 3311 union { 3312 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 3313 struct netdev_notifier_offload_xstats_rd *report_delta; 3314 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 3315 struct netdev_notifier_offload_xstats_ru *report_used; 3316 }; 3317 }; 3318 3319 int netdev_offload_xstats_enable(struct net_device *dev, 3320 enum netdev_offload_xstats_type type, 3321 struct netlink_ext_ack *extack); 3322 int netdev_offload_xstats_disable(struct net_device *dev, 3323 enum netdev_offload_xstats_type type); 3324 bool netdev_offload_xstats_enabled(const struct net_device *dev, 3325 enum netdev_offload_xstats_type type); 3326 int netdev_offload_xstats_get(struct net_device *dev, 3327 enum netdev_offload_xstats_type type, 3328 struct rtnl_hw_stats64 *stats, bool *used, 3329 struct netlink_ext_ack *extack); 3330 void 3331 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 3332 const struct rtnl_hw_stats64 *stats); 3333 void 3334 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 3335 void netdev_offload_xstats_push_delta(struct net_device *dev, 3336 enum netdev_offload_xstats_type type, 3337 const struct rtnl_hw_stats64 *stats); 3338 3339 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 3340 struct net_device *dev) 3341 { 3342 info->dev = dev; 3343 info->extack = NULL; 3344 } 3345 3346 static inline struct net_device * 3347 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 3348 { 3349 return info->dev; 3350 } 3351 3352 static inline struct netlink_ext_ack * 3353 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 3354 { 3355 return info->extack; 3356 } 3357 3358 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 3359 int call_netdevice_notifiers_info(unsigned long val, 3360 struct netdev_notifier_info *info); 3361 3362 #define for_each_netdev(net, d) \ 3363 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 3364 #define for_each_netdev_reverse(net, d) \ 3365 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 3366 #define for_each_netdev_rcu(net, d) \ 3367 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 3368 #define for_each_netdev_safe(net, d, n) \ 3369 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 3370 #define for_each_netdev_continue(net, d) \ 3371 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 3372 #define for_each_netdev_continue_reverse(net, d) \ 3373 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 3374 dev_list) 3375 #define for_each_netdev_continue_rcu(net, d) \ 3376 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 3377 #define for_each_netdev_in_bond_rcu(bond, slave) \ 3378 for_each_netdev_rcu(dev_net_rcu(bond), slave) \ 3379 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 3380 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 3381 3382 #define for_each_netdev_dump(net, d, ifindex) \ 3383 for (; (d = xa_find(&(net)->dev_by_index, &ifindex, \ 3384 ULONG_MAX, XA_PRESENT)); ifindex++) 3385 3386 static inline struct net_device *next_net_device(struct net_device *dev) 3387 { 3388 struct list_head *lh; 3389 struct net *net; 3390 3391 net = dev_net(dev); 3392 lh = dev->dev_list.next; 3393 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3394 } 3395 3396 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 3397 { 3398 struct list_head *lh; 3399 struct net *net; 3400 3401 net = dev_net(dev); 3402 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 3403 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3404 } 3405 3406 static inline struct net_device *first_net_device(struct net *net) 3407 { 3408 return list_empty(&net->dev_base_head) ? NULL : 3409 net_device_entry(net->dev_base_head.next); 3410 } 3411 3412 struct net_device *dev_getbyhwaddr(struct net *net, unsigned short type, 3413 const char *hwaddr); 3414 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 3415 const char *hwaddr); 3416 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 3417 void dev_add_pack(struct packet_type *pt); 3418 void dev_remove_pack(struct packet_type *pt); 3419 void __dev_remove_pack(struct packet_type *pt); 3420 void dev_add_offload(struct packet_offload *po); 3421 void dev_remove_offload(struct packet_offload *po); 3422 3423 int dev_get_iflink(const struct net_device *dev); 3424 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 3425 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 3426 struct net_device_path_stack *stack); 3427 struct net_device *dev_get_by_name(struct net *net, const char *name); 3428 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 3429 struct net_device *__dev_get_by_name(struct net *net, const char *name); 3430 bool netdev_name_in_use(struct net *net, const char *name); 3431 int dev_alloc_name(struct net_device *dev, const char *name); 3432 int netif_open(struct net_device *dev, struct netlink_ext_ack *extack); 3433 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 3434 void netif_close(struct net_device *dev); 3435 void dev_close(struct net_device *dev); 3436 void netif_close_many(struct list_head *head, bool unlink); 3437 void netif_disable_lro(struct net_device *dev); 3438 void dev_disable_lro(struct net_device *dev); 3439 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 3440 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 3441 struct net_device *sb_dev); 3442 3443 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3444 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3445 3446 static inline int dev_queue_xmit(struct sk_buff *skb) 3447 { 3448 return __dev_queue_xmit(skb, NULL); 3449 } 3450 3451 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3452 struct net_device *sb_dev) 3453 { 3454 return __dev_queue_xmit(skb, sb_dev); 3455 } 3456 3457 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3458 { 3459 int ret; 3460 3461 ret = __dev_direct_xmit(skb, queue_id); 3462 if (!dev_xmit_complete(ret)) 3463 kfree_skb(skb); 3464 return ret; 3465 } 3466 3467 int register_netdevice(struct net_device *dev); 3468 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3469 void unregister_netdevice_many(struct list_head *head); 3470 bool unregister_netdevice_queued(const struct net_device *dev); 3471 3472 static inline void unregister_netdevice(struct net_device *dev) 3473 { 3474 unregister_netdevice_queue(dev, NULL); 3475 } 3476 3477 int netdev_refcnt_read(const struct net_device *dev); 3478 void free_netdev(struct net_device *dev); 3479 3480 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3481 struct sk_buff *skb, 3482 bool all_slaves); 3483 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3484 struct sock *sk); 3485 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3486 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3487 struct net_device *netdev_get_by_index(struct net *net, int ifindex, 3488 netdevice_tracker *tracker, gfp_t gfp); 3489 struct net_device *netdev_get_by_index_lock(struct net *net, int ifindex); 3490 struct net_device *netdev_get_by_name(struct net *net, const char *name, 3491 netdevice_tracker *tracker, gfp_t gfp); 3492 struct net_device *netdev_get_by_flags_rcu(struct net *net, netdevice_tracker *tracker, 3493 unsigned short flags, unsigned short mask); 3494 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3495 void netdev_copy_name(struct net_device *dev, char *name); 3496 3497 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3498 unsigned short type, 3499 const void *daddr, const void *saddr, 3500 unsigned int len) 3501 { 3502 if (!dev->header_ops || !dev->header_ops->create) 3503 return 0; 3504 3505 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3506 } 3507 3508 static inline int dev_parse_header(const struct sk_buff *skb, 3509 unsigned char *haddr) 3510 { 3511 const struct net_device *dev = skb->dev; 3512 3513 if (!dev->header_ops || !dev->header_ops->parse) 3514 return 0; 3515 return dev->header_ops->parse(skb, dev, haddr); 3516 } 3517 3518 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3519 { 3520 const struct net_device *dev = skb->dev; 3521 3522 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3523 return 0; 3524 return dev->header_ops->parse_protocol(skb); 3525 } 3526 3527 /* ll_header must have at least hard_header_len allocated */ 3528 static inline bool dev_validate_header(const struct net_device *dev, 3529 char *ll_header, int len) 3530 { 3531 if (likely(len >= dev->hard_header_len)) 3532 return true; 3533 if (len < dev->min_header_len) 3534 return false; 3535 3536 if (dev->header_ops && dev->header_ops->validate) 3537 return dev->header_ops->validate(ll_header, len); 3538 3539 return false; 3540 } 3541 3542 static inline bool dev_has_header(const struct net_device *dev) 3543 { 3544 return dev->header_ops && dev->header_ops->create; 3545 } 3546 3547 struct numa_drop_counters { 3548 atomic_t drops0 ____cacheline_aligned_in_smp; 3549 atomic_t drops1 ____cacheline_aligned_in_smp; 3550 }; 3551 3552 static inline int numa_drop_read(const struct numa_drop_counters *ndc) 3553 { 3554 return atomic_read(&ndc->drops0) + atomic_read(&ndc->drops1); 3555 } 3556 3557 static inline void numa_drop_add(struct numa_drop_counters *ndc, int val) 3558 { 3559 int n = numa_node_id() % 2; 3560 3561 if (n) 3562 atomic_add(val, &ndc->drops1); 3563 else 3564 atomic_add(val, &ndc->drops0); 3565 } 3566 3567 static inline void numa_drop_reset(struct numa_drop_counters *ndc) 3568 { 3569 atomic_set(&ndc->drops0, 0); 3570 atomic_set(&ndc->drops1, 0); 3571 } 3572 3573 /* 3574 * Incoming packets are placed on per-CPU queues 3575 */ 3576 struct softnet_data { 3577 struct list_head poll_list; 3578 struct sk_buff_head process_queue; 3579 local_lock_t process_queue_bh_lock; 3580 3581 /* stats */ 3582 unsigned int processed; 3583 unsigned int time_squeeze; 3584 #ifdef CONFIG_RPS 3585 struct softnet_data *rps_ipi_list; 3586 #endif 3587 3588 unsigned int received_rps; 3589 bool in_net_rx_action; 3590 bool in_napi_threaded_poll; 3591 3592 #ifdef CONFIG_NET_FLOW_LIMIT 3593 struct sd_flow_limit __rcu *flow_limit; 3594 #endif 3595 struct Qdisc *output_queue; 3596 struct Qdisc **output_queue_tailp; 3597 struct sk_buff *completion_queue; 3598 #ifdef CONFIG_XFRM_OFFLOAD 3599 struct sk_buff_head xfrm_backlog; 3600 #endif 3601 /* written and read only by owning cpu: */ 3602 struct netdev_xmit xmit; 3603 #ifdef CONFIG_RPS 3604 /* input_queue_head should be written by cpu owning this struct, 3605 * and only read by other cpus. Worth using a cache line. 3606 */ 3607 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3608 3609 /* Elements below can be accessed between CPUs for RPS/RFS */ 3610 call_single_data_t csd ____cacheline_aligned_in_smp; 3611 struct softnet_data *rps_ipi_next; 3612 unsigned int cpu; 3613 3614 /* We force a cacheline alignment from here, to hold together 3615 * input_queue_tail, input_pkt_queue and backlog.state. 3616 * We add holes so that backlog.state is the last field 3617 * of this cache line. 3618 */ 3619 long pad[3] ____cacheline_aligned_in_smp; 3620 unsigned int input_queue_tail; 3621 #endif 3622 struct sk_buff_head input_pkt_queue; 3623 3624 struct napi_struct backlog; 3625 3626 struct numa_drop_counters drop_counters; 3627 3628 int defer_ipi_scheduled ____cacheline_aligned_in_smp; 3629 call_single_data_t defer_csd; 3630 }; 3631 3632 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3633 3634 struct page_pool_bh { 3635 struct page_pool *pool; 3636 local_lock_t bh_lock; 3637 }; 3638 DECLARE_PER_CPU(struct page_pool_bh, system_page_pool); 3639 3640 #define XMIT_RECURSION_LIMIT 8 3641 3642 #ifndef CONFIG_PREEMPT_RT 3643 static inline int dev_recursion_level(void) 3644 { 3645 return this_cpu_read(softnet_data.xmit.recursion); 3646 } 3647 3648 static inline bool dev_xmit_recursion(void) 3649 { 3650 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3651 XMIT_RECURSION_LIMIT); 3652 } 3653 3654 static inline void dev_xmit_recursion_inc(void) 3655 { 3656 __this_cpu_inc(softnet_data.xmit.recursion); 3657 } 3658 3659 static inline void dev_xmit_recursion_dec(void) 3660 { 3661 __this_cpu_dec(softnet_data.xmit.recursion); 3662 } 3663 #else 3664 static inline int dev_recursion_level(void) 3665 { 3666 return current->net_xmit.recursion; 3667 } 3668 3669 static inline bool dev_xmit_recursion(void) 3670 { 3671 return unlikely(current->net_xmit.recursion > XMIT_RECURSION_LIMIT); 3672 } 3673 3674 static inline void dev_xmit_recursion_inc(void) 3675 { 3676 current->net_xmit.recursion++; 3677 } 3678 3679 static inline void dev_xmit_recursion_dec(void) 3680 { 3681 current->net_xmit.recursion--; 3682 } 3683 #endif 3684 3685 void __netif_schedule(struct Qdisc *q); 3686 void netif_schedule_queue(struct netdev_queue *txq); 3687 3688 static inline void netif_tx_schedule_all(struct net_device *dev) 3689 { 3690 unsigned int i; 3691 3692 for (i = 0; i < dev->num_tx_queues; i++) 3693 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3694 } 3695 3696 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3697 { 3698 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3699 } 3700 3701 /** 3702 * netif_start_queue - allow transmit 3703 * @dev: network device 3704 * 3705 * Allow upper layers to call the device hard_start_xmit routine. 3706 */ 3707 static inline void netif_start_queue(struct net_device *dev) 3708 { 3709 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3710 } 3711 3712 static inline void netif_tx_start_all_queues(struct net_device *dev) 3713 { 3714 unsigned int i; 3715 3716 for (i = 0; i < dev->num_tx_queues; i++) { 3717 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3718 netif_tx_start_queue(txq); 3719 } 3720 } 3721 3722 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3723 3724 /** 3725 * netif_wake_queue - restart transmit 3726 * @dev: network device 3727 * 3728 * Allow upper layers to call the device hard_start_xmit routine. 3729 * Used for flow control when transmit resources are available. 3730 */ 3731 static inline void netif_wake_queue(struct net_device *dev) 3732 { 3733 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3734 } 3735 3736 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3737 { 3738 unsigned int i; 3739 3740 for (i = 0; i < dev->num_tx_queues; i++) { 3741 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3742 netif_tx_wake_queue(txq); 3743 } 3744 } 3745 3746 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3747 { 3748 /* Paired with READ_ONCE() from dev_watchdog() */ 3749 WRITE_ONCE(dev_queue->trans_start, jiffies); 3750 3751 /* This barrier is paired with smp_mb() from dev_watchdog() */ 3752 smp_mb__before_atomic(); 3753 3754 /* Must be an atomic op see netif_txq_try_stop() */ 3755 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3756 } 3757 3758 /** 3759 * netif_stop_queue - stop transmitted packets 3760 * @dev: network device 3761 * 3762 * Stop upper layers calling the device hard_start_xmit routine. 3763 * Used for flow control when transmit resources are unavailable. 3764 */ 3765 static inline void netif_stop_queue(struct net_device *dev) 3766 { 3767 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3768 } 3769 3770 void netif_tx_stop_all_queues(struct net_device *dev); 3771 3772 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3773 { 3774 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3775 } 3776 3777 /** 3778 * netif_queue_stopped - test if transmit queue is flowblocked 3779 * @dev: network device 3780 * 3781 * Test if transmit queue on device is currently unable to send. 3782 */ 3783 static inline bool netif_queue_stopped(const struct net_device *dev) 3784 { 3785 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3786 } 3787 3788 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3789 { 3790 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3791 } 3792 3793 static inline bool 3794 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3795 { 3796 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3797 } 3798 3799 static inline bool 3800 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3801 { 3802 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3803 } 3804 3805 /** 3806 * netdev_queue_set_dql_min_limit - set dql minimum limit 3807 * @dev_queue: pointer to transmit queue 3808 * @min_limit: dql minimum limit 3809 * 3810 * Forces xmit_more() to return true until the minimum threshold 3811 * defined by @min_limit is reached (or until the tx queue is 3812 * empty). Warning: to be use with care, misuse will impact the 3813 * latency. 3814 */ 3815 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3816 unsigned int min_limit) 3817 { 3818 #ifdef CONFIG_BQL 3819 dev_queue->dql.min_limit = min_limit; 3820 #endif 3821 } 3822 3823 static inline int netdev_queue_dql_avail(const struct netdev_queue *txq) 3824 { 3825 #ifdef CONFIG_BQL 3826 /* Non-BQL migrated drivers will return 0, too. */ 3827 return dql_avail(&txq->dql); 3828 #else 3829 return 0; 3830 #endif 3831 } 3832 3833 /** 3834 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3835 * @dev_queue: pointer to transmit queue 3836 * 3837 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3838 * to give appropriate hint to the CPU. 3839 */ 3840 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3841 { 3842 #ifdef CONFIG_BQL 3843 prefetchw(&dev_queue->dql.num_queued); 3844 #endif 3845 } 3846 3847 /** 3848 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3849 * @dev_queue: pointer to transmit queue 3850 * 3851 * BQL enabled drivers might use this helper in their TX completion path, 3852 * to give appropriate hint to the CPU. 3853 */ 3854 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3855 { 3856 #ifdef CONFIG_BQL 3857 prefetchw(&dev_queue->dql.limit); 3858 #endif 3859 } 3860 3861 /** 3862 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3863 * @dev_queue: network device queue 3864 * @bytes: number of bytes queued to the device queue 3865 * 3866 * Report the number of bytes queued for sending/completion to the network 3867 * device hardware queue. @bytes should be a good approximation and should 3868 * exactly match netdev_completed_queue() @bytes. 3869 * This is typically called once per packet, from ndo_start_xmit(). 3870 */ 3871 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3872 unsigned int bytes) 3873 { 3874 #ifdef CONFIG_BQL 3875 dql_queued(&dev_queue->dql, bytes); 3876 3877 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3878 return; 3879 3880 /* Paired with READ_ONCE() from dev_watchdog() */ 3881 WRITE_ONCE(dev_queue->trans_start, jiffies); 3882 3883 /* This barrier is paired with smp_mb() from dev_watchdog() */ 3884 smp_mb__before_atomic(); 3885 3886 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3887 3888 /* 3889 * The XOFF flag must be set before checking the dql_avail below, 3890 * because in netdev_tx_completed_queue we update the dql_completed 3891 * before checking the XOFF flag. 3892 */ 3893 smp_mb__after_atomic(); 3894 3895 /* check again in case another CPU has just made room avail */ 3896 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3897 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3898 #endif 3899 } 3900 3901 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3902 * that they should not test BQL status themselves. 3903 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3904 * skb of a batch. 3905 * Returns true if the doorbell must be used to kick the NIC. 3906 */ 3907 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3908 unsigned int bytes, 3909 bool xmit_more) 3910 { 3911 if (xmit_more) { 3912 #ifdef CONFIG_BQL 3913 dql_queued(&dev_queue->dql, bytes); 3914 #endif 3915 return netif_tx_queue_stopped(dev_queue); 3916 } 3917 netdev_tx_sent_queue(dev_queue, bytes); 3918 return true; 3919 } 3920 3921 /** 3922 * netdev_sent_queue - report the number of bytes queued to hardware 3923 * @dev: network device 3924 * @bytes: number of bytes queued to the hardware device queue 3925 * 3926 * Report the number of bytes queued for sending/completion to the network 3927 * device hardware queue#0. @bytes should be a good approximation and should 3928 * exactly match netdev_completed_queue() @bytes. 3929 * This is typically called once per packet, from ndo_start_xmit(). 3930 */ 3931 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3932 { 3933 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3934 } 3935 3936 static inline bool __netdev_sent_queue(struct net_device *dev, 3937 unsigned int bytes, 3938 bool xmit_more) 3939 { 3940 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3941 xmit_more); 3942 } 3943 3944 /** 3945 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3946 * @dev_queue: network device queue 3947 * @pkts: number of packets (currently ignored) 3948 * @bytes: number of bytes dequeued from the device queue 3949 * 3950 * Must be called at most once per TX completion round (and not per 3951 * individual packet), so that BQL can adjust its limits appropriately. 3952 */ 3953 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3954 unsigned int pkts, unsigned int bytes) 3955 { 3956 #ifdef CONFIG_BQL 3957 if (unlikely(!bytes)) 3958 return; 3959 3960 dql_completed(&dev_queue->dql, bytes); 3961 3962 /* 3963 * Without the memory barrier there is a small possibility that 3964 * netdev_tx_sent_queue will miss the update and cause the queue to 3965 * be stopped forever 3966 */ 3967 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */ 3968 3969 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3970 return; 3971 3972 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3973 netif_schedule_queue(dev_queue); 3974 #endif 3975 } 3976 3977 /** 3978 * netdev_completed_queue - report bytes and packets completed by device 3979 * @dev: network device 3980 * @pkts: actual number of packets sent over the medium 3981 * @bytes: actual number of bytes sent over the medium 3982 * 3983 * Report the number of bytes and packets transmitted by the network device 3984 * hardware queue over the physical medium, @bytes must exactly match the 3985 * @bytes amount passed to netdev_sent_queue() 3986 */ 3987 static inline void netdev_completed_queue(struct net_device *dev, 3988 unsigned int pkts, unsigned int bytes) 3989 { 3990 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3991 } 3992 3993 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3994 { 3995 #ifdef CONFIG_BQL 3996 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3997 dql_reset(&q->dql); 3998 #endif 3999 } 4000 4001 /** 4002 * netdev_tx_reset_subqueue - reset the BQL stats and state of a netdev queue 4003 * @dev: network device 4004 * @qid: stack index of the queue to reset 4005 */ 4006 static inline void netdev_tx_reset_subqueue(const struct net_device *dev, 4007 u32 qid) 4008 { 4009 netdev_tx_reset_queue(netdev_get_tx_queue(dev, qid)); 4010 } 4011 4012 /** 4013 * netdev_reset_queue - reset the packets and bytes count of a network device 4014 * @dev_queue: network device 4015 * 4016 * Reset the bytes and packet count of a network device and clear the 4017 * software flow control OFF bit for this network device 4018 */ 4019 static inline void netdev_reset_queue(struct net_device *dev_queue) 4020 { 4021 netdev_tx_reset_subqueue(dev_queue, 0); 4022 } 4023 4024 /** 4025 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 4026 * @dev: network device 4027 * @queue_index: given tx queue index 4028 * 4029 * Returns 0 if given tx queue index >= number of device tx queues, 4030 * otherwise returns the originally passed tx queue index. 4031 */ 4032 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 4033 { 4034 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 4035 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 4036 dev->name, queue_index, 4037 dev->real_num_tx_queues); 4038 return 0; 4039 } 4040 4041 return queue_index; 4042 } 4043 4044 /** 4045 * netif_running - test if up 4046 * @dev: network device 4047 * 4048 * Test if the device has been brought up. 4049 */ 4050 static inline bool netif_running(const struct net_device *dev) 4051 { 4052 return test_bit(__LINK_STATE_START, &dev->state); 4053 } 4054 4055 /* 4056 * Routines to manage the subqueues on a device. We only need start, 4057 * stop, and a check if it's stopped. All other device management is 4058 * done at the overall netdevice level. 4059 * Also test the device if we're multiqueue. 4060 */ 4061 4062 /** 4063 * netif_start_subqueue - allow sending packets on subqueue 4064 * @dev: network device 4065 * @queue_index: sub queue index 4066 * 4067 * Start individual transmit queue of a device with multiple transmit queues. 4068 */ 4069 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 4070 { 4071 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 4072 4073 netif_tx_start_queue(txq); 4074 } 4075 4076 /** 4077 * netif_stop_subqueue - stop sending packets on subqueue 4078 * @dev: network device 4079 * @queue_index: sub queue index 4080 * 4081 * Stop individual transmit queue of a device with multiple transmit queues. 4082 */ 4083 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 4084 { 4085 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 4086 netif_tx_stop_queue(txq); 4087 } 4088 4089 /** 4090 * __netif_subqueue_stopped - test status of subqueue 4091 * @dev: network device 4092 * @queue_index: sub queue index 4093 * 4094 * Check individual transmit queue of a device with multiple transmit queues. 4095 */ 4096 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 4097 u16 queue_index) 4098 { 4099 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 4100 4101 return netif_tx_queue_stopped(txq); 4102 } 4103 4104 /** 4105 * netif_subqueue_stopped - test status of subqueue 4106 * @dev: network device 4107 * @skb: sub queue buffer pointer 4108 * 4109 * Check individual transmit queue of a device with multiple transmit queues. 4110 */ 4111 static inline bool netif_subqueue_stopped(const struct net_device *dev, 4112 struct sk_buff *skb) 4113 { 4114 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 4115 } 4116 4117 /** 4118 * netif_wake_subqueue - allow sending packets on subqueue 4119 * @dev: network device 4120 * @queue_index: sub queue index 4121 * 4122 * Resume individual transmit queue of a device with multiple transmit queues. 4123 */ 4124 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 4125 { 4126 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 4127 4128 netif_tx_wake_queue(txq); 4129 } 4130 4131 #ifdef CONFIG_XPS 4132 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 4133 u16 index); 4134 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 4135 u16 index, enum xps_map_type type); 4136 4137 /** 4138 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 4139 * @j: CPU/Rx queue index 4140 * @mask: bitmask of all cpus/rx queues 4141 * @nr_bits: number of bits in the bitmask 4142 * 4143 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 4144 */ 4145 static inline bool netif_attr_test_mask(unsigned long j, 4146 const unsigned long *mask, 4147 unsigned int nr_bits) 4148 { 4149 cpu_max_bits_warn(j, nr_bits); 4150 return test_bit(j, mask); 4151 } 4152 4153 /** 4154 * netif_attr_test_online - Test for online CPU/Rx queue 4155 * @j: CPU/Rx queue index 4156 * @online_mask: bitmask for CPUs/Rx queues that are online 4157 * @nr_bits: number of bits in the bitmask 4158 * 4159 * Returns: true if a CPU/Rx queue is online. 4160 */ 4161 static inline bool netif_attr_test_online(unsigned long j, 4162 const unsigned long *online_mask, 4163 unsigned int nr_bits) 4164 { 4165 cpu_max_bits_warn(j, nr_bits); 4166 4167 if (online_mask) 4168 return test_bit(j, online_mask); 4169 4170 return (j < nr_bits); 4171 } 4172 4173 /** 4174 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 4175 * @n: CPU/Rx queue index 4176 * @srcp: the cpumask/Rx queue mask pointer 4177 * @nr_bits: number of bits in the bitmask 4178 * 4179 * Returns: next (after n) CPU/Rx queue index in the mask; 4180 * >= nr_bits if no further CPUs/Rx queues set. 4181 */ 4182 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 4183 unsigned int nr_bits) 4184 { 4185 /* -1 is a legal arg here. */ 4186 if (n != -1) 4187 cpu_max_bits_warn(n, nr_bits); 4188 4189 if (srcp) 4190 return find_next_bit(srcp, nr_bits, n + 1); 4191 4192 return n + 1; 4193 } 4194 4195 /** 4196 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 4197 * @n: CPU/Rx queue index 4198 * @src1p: the first CPUs/Rx queues mask pointer 4199 * @src2p: the second CPUs/Rx queues mask pointer 4200 * @nr_bits: number of bits in the bitmask 4201 * 4202 * Returns: next (after n) CPU/Rx queue index set in both masks; 4203 * >= nr_bits if no further CPUs/Rx queues set in both. 4204 */ 4205 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 4206 const unsigned long *src2p, 4207 unsigned int nr_bits) 4208 { 4209 /* -1 is a legal arg here. */ 4210 if (n != -1) 4211 cpu_max_bits_warn(n, nr_bits); 4212 4213 if (src1p && src2p) 4214 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 4215 else if (src1p) 4216 return find_next_bit(src1p, nr_bits, n + 1); 4217 else if (src2p) 4218 return find_next_bit(src2p, nr_bits, n + 1); 4219 4220 return n + 1; 4221 } 4222 #else 4223 static inline int netif_set_xps_queue(struct net_device *dev, 4224 const struct cpumask *mask, 4225 u16 index) 4226 { 4227 return 0; 4228 } 4229 4230 static inline int __netif_set_xps_queue(struct net_device *dev, 4231 const unsigned long *mask, 4232 u16 index, enum xps_map_type type) 4233 { 4234 return 0; 4235 } 4236 #endif 4237 4238 /** 4239 * netif_is_multiqueue - test if device has multiple transmit queues 4240 * @dev: network device 4241 * 4242 * Check if device has multiple transmit queues 4243 */ 4244 static inline bool netif_is_multiqueue(const struct net_device *dev) 4245 { 4246 return dev->num_tx_queues > 1; 4247 } 4248 4249 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 4250 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 4251 int netif_set_real_num_queues(struct net_device *dev, 4252 unsigned int txq, unsigned int rxq); 4253 4254 int netif_get_num_default_rss_queues(void); 4255 4256 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason); 4257 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason); 4258 4259 /* 4260 * It is not allowed to call kfree_skb() or consume_skb() from hardware 4261 * interrupt context or with hardware interrupts being disabled. 4262 * (in_hardirq() || irqs_disabled()) 4263 * 4264 * We provide four helpers that can be used in following contexts : 4265 * 4266 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 4267 * replacing kfree_skb(skb) 4268 * 4269 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 4270 * Typically used in place of consume_skb(skb) in TX completion path 4271 * 4272 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 4273 * replacing kfree_skb(skb) 4274 * 4275 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 4276 * and consumed a packet. Used in place of consume_skb(skb) 4277 */ 4278 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 4279 { 4280 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 4281 } 4282 4283 static inline void dev_consume_skb_irq(struct sk_buff *skb) 4284 { 4285 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED); 4286 } 4287 4288 static inline void dev_kfree_skb_any(struct sk_buff *skb) 4289 { 4290 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 4291 } 4292 4293 static inline void dev_consume_skb_any(struct sk_buff *skb) 4294 { 4295 dev_kfree_skb_any_reason(skb, SKB_CONSUMED); 4296 } 4297 4298 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 4299 const struct bpf_prog *xdp_prog); 4300 void generic_xdp_tx(struct sk_buff *skb, const struct bpf_prog *xdp_prog); 4301 int do_xdp_generic(const struct bpf_prog *xdp_prog, struct sk_buff **pskb); 4302 int netif_rx(struct sk_buff *skb); 4303 int __netif_rx(struct sk_buff *skb); 4304 4305 int netif_receive_skb(struct sk_buff *skb); 4306 int netif_receive_skb_core(struct sk_buff *skb); 4307 void netif_receive_skb_list_internal(struct list_head *head); 4308 void netif_receive_skb_list(struct list_head *head); 4309 gro_result_t gro_receive_skb(struct gro_node *gro, struct sk_buff *skb); 4310 4311 static inline gro_result_t napi_gro_receive(struct napi_struct *napi, 4312 struct sk_buff *skb) 4313 { 4314 return gro_receive_skb(&napi->gro, skb); 4315 } 4316 4317 struct sk_buff *napi_get_frags(struct napi_struct *napi); 4318 gro_result_t napi_gro_frags(struct napi_struct *napi); 4319 4320 static inline void napi_free_frags(struct napi_struct *napi) 4321 { 4322 kfree_skb(napi->skb); 4323 napi->skb = NULL; 4324 } 4325 4326 bool netdev_is_rx_handler_busy(struct net_device *dev); 4327 int netdev_rx_handler_register(struct net_device *dev, 4328 rx_handler_func_t *rx_handler, 4329 void *rx_handler_data); 4330 void netdev_rx_handler_unregister(struct net_device *dev); 4331 4332 bool dev_valid_name(const char *name); 4333 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 4334 { 4335 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 4336 } 4337 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 4338 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 4339 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 4340 void __user *data, bool *need_copyout); 4341 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 4342 int dev_eth_ioctl(struct net_device *dev, 4343 struct ifreq *ifr, unsigned int cmd); 4344 int generic_hwtstamp_get_lower(struct net_device *dev, 4345 struct kernel_hwtstamp_config *kernel_cfg); 4346 int generic_hwtstamp_set_lower(struct net_device *dev, 4347 struct kernel_hwtstamp_config *kernel_cfg, 4348 struct netlink_ext_ack *extack); 4349 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 4350 unsigned int netif_get_flags(const struct net_device *dev); 4351 int __dev_change_flags(struct net_device *dev, unsigned int flags, 4352 struct netlink_ext_ack *extack); 4353 int netif_change_flags(struct net_device *dev, unsigned int flags, 4354 struct netlink_ext_ack *extack); 4355 int dev_change_flags(struct net_device *dev, unsigned int flags, 4356 struct netlink_ext_ack *extack); 4357 int netif_set_alias(struct net_device *dev, const char *alias, size_t len); 4358 int dev_set_alias(struct net_device *, const char *, size_t); 4359 int dev_get_alias(const struct net_device *, char *, size_t); 4360 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 4361 const char *pat, int new_ifindex, 4362 struct netlink_ext_ack *extack); 4363 int dev_change_net_namespace(struct net_device *dev, struct net *net, 4364 const char *pat); 4365 int __netif_set_mtu(struct net_device *dev, int new_mtu); 4366 int netif_set_mtu(struct net_device *dev, int new_mtu); 4367 int dev_set_mtu(struct net_device *, int); 4368 int netif_pre_changeaddr_notify(struct net_device *dev, const char *addr, 4369 struct netlink_ext_ack *extack); 4370 int netif_set_mac_address(struct net_device *dev, struct sockaddr_storage *ss, 4371 struct netlink_ext_ack *extack); 4372 int dev_set_mac_address(struct net_device *dev, struct sockaddr_storage *ss, 4373 struct netlink_ext_ack *extack); 4374 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr_storage *ss, 4375 struct netlink_ext_ack *extack); 4376 int netif_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 4377 int netif_get_port_parent_id(struct net_device *dev, 4378 struct netdev_phys_item_id *ppid, bool recurse); 4379 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 4380 4381 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 4382 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 4383 struct netdev_queue *txq, int *ret); 4384 4385 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 4386 u8 dev_xdp_prog_count(struct net_device *dev); 4387 int netif_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf); 4388 int dev_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf); 4389 u8 dev_xdp_sb_prog_count(struct net_device *dev); 4390 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 4391 4392 u32 dev_get_min_mp_channel_count(const struct net_device *dev); 4393 4394 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4395 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4396 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 4397 bool is_skb_forwardable(const struct net_device *dev, 4398 const struct sk_buff *skb); 4399 4400 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 4401 const struct sk_buff *skb, 4402 const bool check_mtu) 4403 { 4404 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 4405 unsigned int len; 4406 4407 if (!(dev->flags & IFF_UP)) 4408 return false; 4409 4410 if (!check_mtu) 4411 return true; 4412 4413 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 4414 if (skb->len <= len) 4415 return true; 4416 4417 /* if TSO is enabled, we don't care about the length as the packet 4418 * could be forwarded without being segmented before 4419 */ 4420 if (skb_is_gso(skb)) 4421 return true; 4422 4423 return false; 4424 } 4425 4426 void netdev_core_stats_inc(struct net_device *dev, u32 offset); 4427 4428 #define DEV_CORE_STATS_INC(FIELD) \ 4429 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 4430 { \ 4431 netdev_core_stats_inc(dev, \ 4432 offsetof(struct net_device_core_stats, FIELD)); \ 4433 } 4434 DEV_CORE_STATS_INC(rx_dropped) 4435 DEV_CORE_STATS_INC(tx_dropped) 4436 DEV_CORE_STATS_INC(rx_nohandler) 4437 DEV_CORE_STATS_INC(rx_otherhost_dropped) 4438 #undef DEV_CORE_STATS_INC 4439 4440 static __always_inline int ____dev_forward_skb(struct net_device *dev, 4441 struct sk_buff *skb, 4442 const bool check_mtu) 4443 { 4444 if (skb_orphan_frags(skb, GFP_ATOMIC) || 4445 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 4446 dev_core_stats_rx_dropped_inc(dev); 4447 kfree_skb(skb); 4448 return NET_RX_DROP; 4449 } 4450 4451 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 4452 skb->priority = 0; 4453 return 0; 4454 } 4455 4456 bool dev_nit_active_rcu(const struct net_device *dev); 4457 static inline bool dev_nit_active(const struct net_device *dev) 4458 { 4459 bool ret; 4460 4461 rcu_read_lock(); 4462 ret = dev_nit_active_rcu(dev); 4463 rcu_read_unlock(); 4464 return ret; 4465 } 4466 4467 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 4468 4469 static inline void __dev_put(struct net_device *dev) 4470 { 4471 if (dev) { 4472 #ifdef CONFIG_PCPU_DEV_REFCNT 4473 this_cpu_dec(*dev->pcpu_refcnt); 4474 #else 4475 refcount_dec(&dev->dev_refcnt); 4476 #endif 4477 } 4478 } 4479 4480 static inline void __dev_hold(struct net_device *dev) 4481 { 4482 if (dev) { 4483 #ifdef CONFIG_PCPU_DEV_REFCNT 4484 this_cpu_inc(*dev->pcpu_refcnt); 4485 #else 4486 refcount_inc(&dev->dev_refcnt); 4487 #endif 4488 } 4489 } 4490 4491 static inline void __netdev_tracker_alloc(struct net_device *dev, 4492 netdevice_tracker *tracker, 4493 gfp_t gfp) 4494 { 4495 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4496 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 4497 #endif 4498 } 4499 4500 /* netdev_tracker_alloc() can upgrade a prior untracked reference 4501 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4502 */ 4503 static inline void netdev_tracker_alloc(struct net_device *dev, 4504 netdevice_tracker *tracker, gfp_t gfp) 4505 { 4506 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4507 refcount_dec(&dev->refcnt_tracker.no_tracker); 4508 __netdev_tracker_alloc(dev, tracker, gfp); 4509 #endif 4510 } 4511 4512 static inline void netdev_tracker_free(struct net_device *dev, 4513 netdevice_tracker *tracker) 4514 { 4515 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4516 ref_tracker_free(&dev->refcnt_tracker, tracker); 4517 #endif 4518 } 4519 4520 static inline void netdev_hold(struct net_device *dev, 4521 netdevice_tracker *tracker, gfp_t gfp) 4522 { 4523 if (dev) { 4524 __dev_hold(dev); 4525 __netdev_tracker_alloc(dev, tracker, gfp); 4526 } 4527 } 4528 4529 static inline void netdev_put(struct net_device *dev, 4530 netdevice_tracker *tracker) 4531 { 4532 if (dev) { 4533 netdev_tracker_free(dev, tracker); 4534 __dev_put(dev); 4535 } 4536 } 4537 4538 /** 4539 * dev_hold - get reference to device 4540 * @dev: network device 4541 * 4542 * Hold reference to device to keep it from being freed. 4543 * Try using netdev_hold() instead. 4544 */ 4545 static inline void dev_hold(struct net_device *dev) 4546 { 4547 netdev_hold(dev, NULL, GFP_ATOMIC); 4548 } 4549 4550 /** 4551 * dev_put - release reference to device 4552 * @dev: network device 4553 * 4554 * Release reference to device to allow it to be freed. 4555 * Try using netdev_put() instead. 4556 */ 4557 static inline void dev_put(struct net_device *dev) 4558 { 4559 netdev_put(dev, NULL); 4560 } 4561 4562 DEFINE_FREE(dev_put, struct net_device *, if (_T) dev_put(_T)) 4563 4564 static inline void netdev_ref_replace(struct net_device *odev, 4565 struct net_device *ndev, 4566 netdevice_tracker *tracker, 4567 gfp_t gfp) 4568 { 4569 if (odev) 4570 netdev_tracker_free(odev, tracker); 4571 4572 __dev_hold(ndev); 4573 __dev_put(odev); 4574 4575 if (ndev) 4576 __netdev_tracker_alloc(ndev, tracker, gfp); 4577 } 4578 4579 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4580 * and _off may be called from IRQ context, but it is caller 4581 * who is responsible for serialization of these calls. 4582 * 4583 * The name carrier is inappropriate, these functions should really be 4584 * called netif_lowerlayer_*() because they represent the state of any 4585 * kind of lower layer not just hardware media. 4586 */ 4587 void linkwatch_fire_event(struct net_device *dev); 4588 4589 /** 4590 * linkwatch_sync_dev - sync linkwatch for the given device 4591 * @dev: network device to sync linkwatch for 4592 * 4593 * Sync linkwatch for the given device, removing it from the 4594 * pending work list (if queued). 4595 */ 4596 void linkwatch_sync_dev(struct net_device *dev); 4597 void __linkwatch_sync_dev(struct net_device *dev); 4598 4599 /** 4600 * netif_carrier_ok - test if carrier present 4601 * @dev: network device 4602 * 4603 * Check if carrier is present on device 4604 */ 4605 static inline bool netif_carrier_ok(const struct net_device *dev) 4606 { 4607 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4608 } 4609 4610 unsigned long dev_trans_start(struct net_device *dev); 4611 4612 void netdev_watchdog_up(struct net_device *dev); 4613 4614 void netif_carrier_on(struct net_device *dev); 4615 void netif_carrier_off(struct net_device *dev); 4616 void netif_carrier_event(struct net_device *dev); 4617 4618 /** 4619 * netif_dormant_on - mark device as dormant. 4620 * @dev: network device 4621 * 4622 * Mark device as dormant (as per RFC2863). 4623 * 4624 * The dormant state indicates that the relevant interface is not 4625 * actually in a condition to pass packets (i.e., it is not 'up') but is 4626 * in a "pending" state, waiting for some external event. For "on- 4627 * demand" interfaces, this new state identifies the situation where the 4628 * interface is waiting for events to place it in the up state. 4629 */ 4630 static inline void netif_dormant_on(struct net_device *dev) 4631 { 4632 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4633 linkwatch_fire_event(dev); 4634 } 4635 4636 /** 4637 * netif_dormant_off - set device as not dormant. 4638 * @dev: network device 4639 * 4640 * Device is not in dormant state. 4641 */ 4642 static inline void netif_dormant_off(struct net_device *dev) 4643 { 4644 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4645 linkwatch_fire_event(dev); 4646 } 4647 4648 /** 4649 * netif_dormant - test if device is dormant 4650 * @dev: network device 4651 * 4652 * Check if device is dormant. 4653 */ 4654 static inline bool netif_dormant(const struct net_device *dev) 4655 { 4656 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4657 } 4658 4659 4660 /** 4661 * netif_testing_on - mark device as under test. 4662 * @dev: network device 4663 * 4664 * Mark device as under test (as per RFC2863). 4665 * 4666 * The testing state indicates that some test(s) must be performed on 4667 * the interface. After completion, of the test, the interface state 4668 * will change to up, dormant, or down, as appropriate. 4669 */ 4670 static inline void netif_testing_on(struct net_device *dev) 4671 { 4672 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4673 linkwatch_fire_event(dev); 4674 } 4675 4676 /** 4677 * netif_testing_off - set device as not under test. 4678 * @dev: network device 4679 * 4680 * Device is not in testing state. 4681 */ 4682 static inline void netif_testing_off(struct net_device *dev) 4683 { 4684 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4685 linkwatch_fire_event(dev); 4686 } 4687 4688 /** 4689 * netif_testing - test if device is under test 4690 * @dev: network device 4691 * 4692 * Check if device is under test 4693 */ 4694 static inline bool netif_testing(const struct net_device *dev) 4695 { 4696 return test_bit(__LINK_STATE_TESTING, &dev->state); 4697 } 4698 4699 4700 /** 4701 * netif_oper_up - test if device is operational 4702 * @dev: network device 4703 * 4704 * Check if carrier is operational 4705 */ 4706 static inline bool netif_oper_up(const struct net_device *dev) 4707 { 4708 unsigned int operstate = READ_ONCE(dev->operstate); 4709 4710 return operstate == IF_OPER_UP || 4711 operstate == IF_OPER_UNKNOWN /* backward compat */; 4712 } 4713 4714 /** 4715 * netif_device_present - is device available or removed 4716 * @dev: network device 4717 * 4718 * Check if device has not been removed from system. 4719 */ 4720 static inline bool netif_device_present(const struct net_device *dev) 4721 { 4722 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4723 } 4724 4725 void netif_device_detach(struct net_device *dev); 4726 4727 void netif_device_attach(struct net_device *dev); 4728 4729 /* 4730 * Network interface message level settings 4731 */ 4732 4733 enum { 4734 NETIF_MSG_DRV_BIT, 4735 NETIF_MSG_PROBE_BIT, 4736 NETIF_MSG_LINK_BIT, 4737 NETIF_MSG_TIMER_BIT, 4738 NETIF_MSG_IFDOWN_BIT, 4739 NETIF_MSG_IFUP_BIT, 4740 NETIF_MSG_RX_ERR_BIT, 4741 NETIF_MSG_TX_ERR_BIT, 4742 NETIF_MSG_TX_QUEUED_BIT, 4743 NETIF_MSG_INTR_BIT, 4744 NETIF_MSG_TX_DONE_BIT, 4745 NETIF_MSG_RX_STATUS_BIT, 4746 NETIF_MSG_PKTDATA_BIT, 4747 NETIF_MSG_HW_BIT, 4748 NETIF_MSG_WOL_BIT, 4749 4750 /* When you add a new bit above, update netif_msg_class_names array 4751 * in net/ethtool/common.c 4752 */ 4753 NETIF_MSG_CLASS_COUNT, 4754 }; 4755 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4756 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4757 4758 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4759 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4760 4761 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4762 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4763 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4764 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4765 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4766 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4767 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4768 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4769 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4770 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4771 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4772 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4773 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4774 #define NETIF_MSG_HW __NETIF_MSG(HW) 4775 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4776 4777 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4778 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4779 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4780 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4781 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4782 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4783 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4784 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4785 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4786 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4787 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4788 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4789 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4790 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4791 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4792 4793 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4794 { 4795 /* use default */ 4796 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4797 return default_msg_enable_bits; 4798 if (debug_value == 0) /* no output */ 4799 return 0; 4800 /* set low N bits */ 4801 return (1U << debug_value) - 1; 4802 } 4803 4804 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4805 { 4806 spin_lock(&txq->_xmit_lock); 4807 /* Pairs with READ_ONCE() in netif_tx_owned() */ 4808 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4809 } 4810 4811 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4812 { 4813 __acquire(&txq->_xmit_lock); 4814 return true; 4815 } 4816 4817 static inline void __netif_tx_release(struct netdev_queue *txq) 4818 { 4819 __release(&txq->_xmit_lock); 4820 } 4821 4822 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4823 { 4824 spin_lock_bh(&txq->_xmit_lock); 4825 /* Pairs with READ_ONCE() in netif_tx_owned() */ 4826 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4827 } 4828 4829 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4830 { 4831 bool ok = spin_trylock(&txq->_xmit_lock); 4832 4833 if (likely(ok)) { 4834 /* Pairs with READ_ONCE() in netif_tx_owned() */ 4835 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4836 } 4837 return ok; 4838 } 4839 4840 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4841 { 4842 /* Pairs with READ_ONCE() in netif_tx_owned() */ 4843 WRITE_ONCE(txq->xmit_lock_owner, -1); 4844 spin_unlock(&txq->_xmit_lock); 4845 } 4846 4847 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4848 { 4849 /* Pairs with READ_ONCE() in netif_tx_owned() */ 4850 WRITE_ONCE(txq->xmit_lock_owner, -1); 4851 spin_unlock_bh(&txq->_xmit_lock); 4852 } 4853 4854 /* 4855 * txq->trans_start can be read locklessly from dev_watchdog() 4856 */ 4857 static inline void txq_trans_update(const struct net_device *dev, 4858 struct netdev_queue *txq) 4859 { 4860 if (!dev->lltx) 4861 WRITE_ONCE(txq->trans_start, jiffies); 4862 } 4863 4864 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4865 { 4866 unsigned long now = jiffies; 4867 4868 if (READ_ONCE(txq->trans_start) != now) 4869 WRITE_ONCE(txq->trans_start, now); 4870 } 4871 4872 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4873 static inline void netif_trans_update(struct net_device *dev) 4874 { 4875 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4876 4877 txq_trans_cond_update(txq); 4878 } 4879 4880 /** 4881 * netif_tx_lock - grab network device transmit lock 4882 * @dev: network device 4883 * 4884 * Get network device transmit lock 4885 */ 4886 void netif_tx_lock(struct net_device *dev); 4887 4888 static inline void netif_tx_lock_bh(struct net_device *dev) 4889 { 4890 local_bh_disable(); 4891 netif_tx_lock(dev); 4892 } 4893 4894 void netif_tx_unlock(struct net_device *dev); 4895 4896 static inline void netif_tx_unlock_bh(struct net_device *dev) 4897 { 4898 netif_tx_unlock(dev); 4899 local_bh_enable(); 4900 } 4901 4902 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4903 if (!(dev)->lltx) { \ 4904 __netif_tx_lock(txq, cpu); \ 4905 } else { \ 4906 __netif_tx_acquire(txq); \ 4907 } \ 4908 } 4909 4910 #define HARD_TX_TRYLOCK(dev, txq) \ 4911 (!(dev)->lltx ? \ 4912 __netif_tx_trylock(txq) : \ 4913 __netif_tx_acquire(txq)) 4914 4915 #define HARD_TX_UNLOCK(dev, txq) { \ 4916 if (!(dev)->lltx) { \ 4917 __netif_tx_unlock(txq); \ 4918 } else { \ 4919 __netif_tx_release(txq); \ 4920 } \ 4921 } 4922 4923 static inline void netif_tx_disable(struct net_device *dev) 4924 { 4925 unsigned int i; 4926 int cpu; 4927 4928 local_bh_disable(); 4929 cpu = smp_processor_id(); 4930 spin_lock(&dev->tx_global_lock); 4931 for (i = 0; i < dev->num_tx_queues; i++) { 4932 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4933 4934 __netif_tx_lock(txq, cpu); 4935 netif_tx_stop_queue(txq); 4936 __netif_tx_unlock(txq); 4937 } 4938 spin_unlock(&dev->tx_global_lock); 4939 local_bh_enable(); 4940 } 4941 4942 #ifndef CONFIG_PREEMPT_RT 4943 static inline bool netif_tx_owned(struct netdev_queue *txq, unsigned int cpu) 4944 { 4945 /* Other cpus might concurrently change txq->xmit_lock_owner 4946 * to -1 or to their cpu id, but not to our id. 4947 */ 4948 return READ_ONCE(txq->xmit_lock_owner) == cpu; 4949 } 4950 4951 #else 4952 static inline bool netif_tx_owned(struct netdev_queue *txq, unsigned int cpu) 4953 { 4954 return rt_mutex_owner(&txq->_xmit_lock.lock) == current; 4955 } 4956 4957 #endif 4958 4959 static inline void netif_addr_lock(struct net_device *dev) 4960 { 4961 unsigned char nest_level = 0; 4962 4963 #ifdef CONFIG_LOCKDEP 4964 nest_level = dev->nested_level; 4965 #endif 4966 spin_lock_nested(&dev->addr_list_lock, nest_level); 4967 } 4968 4969 static inline void netif_addr_lock_bh(struct net_device *dev) 4970 { 4971 unsigned char nest_level = 0; 4972 4973 #ifdef CONFIG_LOCKDEP 4974 nest_level = dev->nested_level; 4975 #endif 4976 local_bh_disable(); 4977 spin_lock_nested(&dev->addr_list_lock, nest_level); 4978 } 4979 4980 static inline void netif_addr_unlock(struct net_device *dev) 4981 { 4982 spin_unlock(&dev->addr_list_lock); 4983 } 4984 4985 static inline void netif_addr_unlock_bh(struct net_device *dev) 4986 { 4987 spin_unlock_bh(&dev->addr_list_lock); 4988 } 4989 4990 /* 4991 * dev_addrs walker. Should be used only for read access. Call with 4992 * rcu_read_lock held. 4993 */ 4994 #define for_each_dev_addr(dev, ha) \ 4995 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4996 4997 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4998 4999 void ether_setup(struct net_device *dev); 5000 5001 /* Allocate dummy net_device */ 5002 struct net_device *alloc_netdev_dummy(int sizeof_priv); 5003 5004 /* Support for loadable net-drivers */ 5005 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 5006 unsigned char name_assign_type, 5007 void (*setup)(struct net_device *), 5008 unsigned int txqs, unsigned int rxqs); 5009 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 5010 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 5011 5012 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 5013 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 5014 count) 5015 5016 int register_netdev(struct net_device *dev); 5017 void unregister_netdev(struct net_device *dev); 5018 5019 int devm_register_netdev(struct device *dev, struct net_device *ndev); 5020 5021 /* General hardware address lists handling functions */ 5022 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 5023 struct netdev_hw_addr_list *from_list, int addr_len); 5024 int __hw_addr_sync_multiple(struct netdev_hw_addr_list *to_list, 5025 struct netdev_hw_addr_list *from_list, 5026 int addr_len); 5027 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 5028 struct netdev_hw_addr_list *from_list, int addr_len); 5029 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 5030 struct net_device *dev, 5031 int (*sync)(struct net_device *, const unsigned char *), 5032 int (*unsync)(struct net_device *, 5033 const unsigned char *)); 5034 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 5035 struct net_device *dev, 5036 int (*sync)(struct net_device *, 5037 const unsigned char *, int), 5038 int (*unsync)(struct net_device *, 5039 const unsigned char *, int)); 5040 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 5041 struct net_device *dev, 5042 int (*unsync)(struct net_device *, 5043 const unsigned char *, int)); 5044 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 5045 struct net_device *dev, 5046 int (*unsync)(struct net_device *, 5047 const unsigned char *)); 5048 void __hw_addr_init(struct netdev_hw_addr_list *list); 5049 void __hw_addr_flush(struct netdev_hw_addr_list *list); 5050 int __hw_addr_list_snapshot(struct netdev_hw_addr_list *snap, 5051 const struct netdev_hw_addr_list *list, 5052 int addr_len, struct netdev_hw_addr_list *cache); 5053 void __hw_addr_list_reconcile(struct netdev_hw_addr_list *real_list, 5054 struct netdev_hw_addr_list *work, 5055 struct netdev_hw_addr_list *ref, int addr_len, 5056 struct netdev_hw_addr_list *cache); 5057 5058 /* Functions used for device addresses handling */ 5059 void dev_addr_mod(struct net_device *dev, unsigned int offset, 5060 const void *addr, size_t len); 5061 5062 static inline void 5063 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 5064 { 5065 dev_addr_mod(dev, 0, addr, len); 5066 } 5067 5068 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 5069 { 5070 __dev_addr_set(dev, addr, dev->addr_len); 5071 } 5072 5073 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 5074 unsigned char addr_type); 5075 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 5076 unsigned char addr_type); 5077 5078 /* Functions used for unicast addresses handling */ 5079 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 5080 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 5081 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 5082 int dev_uc_sync(struct net_device *to, struct net_device *from); 5083 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 5084 void dev_uc_unsync(struct net_device *to, struct net_device *from); 5085 void dev_uc_flush(struct net_device *dev); 5086 void dev_uc_init(struct net_device *dev); 5087 5088 /** 5089 * __dev_uc_sync - Synchronize device's unicast list 5090 * @dev: device to sync 5091 * @sync: function to call if address should be added 5092 * @unsync: function to call if address should be removed 5093 * 5094 * Add newly added addresses to the interface, and release 5095 * addresses that have been deleted. 5096 */ 5097 static inline int __dev_uc_sync(struct net_device *dev, 5098 int (*sync)(struct net_device *, 5099 const unsigned char *), 5100 int (*unsync)(struct net_device *, 5101 const unsigned char *)) 5102 { 5103 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 5104 } 5105 5106 /** 5107 * __dev_uc_unsync - Remove synchronized addresses from device 5108 * @dev: device to sync 5109 * @unsync: function to call if address should be removed 5110 * 5111 * Remove all addresses that were added to the device by dev_uc_sync(). 5112 */ 5113 static inline void __dev_uc_unsync(struct net_device *dev, 5114 int (*unsync)(struct net_device *, 5115 const unsigned char *)) 5116 { 5117 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 5118 } 5119 5120 /* Functions used for multicast addresses handling */ 5121 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 5122 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 5123 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 5124 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 5125 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 5126 int dev_mc_sync(struct net_device *to, struct net_device *from); 5127 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 5128 void dev_mc_unsync(struct net_device *to, struct net_device *from); 5129 void dev_mc_flush(struct net_device *dev); 5130 void dev_mc_init(struct net_device *dev); 5131 5132 /** 5133 * __dev_mc_sync - Synchronize device's multicast list 5134 * @dev: device to sync 5135 * @sync: function to call if address should be added 5136 * @unsync: function to call if address should be removed 5137 * 5138 * Add newly added addresses to the interface, and release 5139 * addresses that have been deleted. 5140 */ 5141 static inline int __dev_mc_sync(struct net_device *dev, 5142 int (*sync)(struct net_device *, 5143 const unsigned char *), 5144 int (*unsync)(struct net_device *, 5145 const unsigned char *)) 5146 { 5147 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 5148 } 5149 5150 /** 5151 * __dev_mc_unsync - Remove synchronized addresses from device 5152 * @dev: device to sync 5153 * @unsync: function to call if address should be removed 5154 * 5155 * Remove all addresses that were added to the device by dev_mc_sync(). 5156 */ 5157 static inline void __dev_mc_unsync(struct net_device *dev, 5158 int (*unsync)(struct net_device *, 5159 const unsigned char *)) 5160 { 5161 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 5162 } 5163 5164 /* Functions used for secondary unicast and multicast support */ 5165 void dev_set_rx_mode(struct net_device *dev); 5166 void netif_rx_mode_schedule_retry(struct net_device *dev); 5167 int netif_set_promiscuity(struct net_device *dev, int inc); 5168 int dev_set_promiscuity(struct net_device *dev, int inc); 5169 int netif_set_allmulti(struct net_device *dev, int inc, bool notify); 5170 int dev_set_allmulti(struct net_device *dev, int inc); 5171 void netif_state_change(struct net_device *dev); 5172 void netdev_state_change(struct net_device *dev); 5173 void __netdev_notify_peers(struct net_device *dev); 5174 void netdev_notify_peers(struct net_device *dev); 5175 void netdev_features_change(struct net_device *dev); 5176 /* Load a device via the kmod */ 5177 void dev_load(struct net *net, const char *name); 5178 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 5179 struct rtnl_link_stats64 *storage); 5180 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 5181 const struct net_device_stats *netdev_stats); 5182 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 5183 const struct pcpu_sw_netstats __percpu *netstats); 5184 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 5185 5186 void netdev_work_sched(struct net_device *dev, unsigned long events); 5187 unsigned long netdev_work_cancel(struct net_device *dev, unsigned long mask); 5188 5189 enum { 5190 NESTED_SYNC_IMM_BIT, 5191 NESTED_SYNC_TODO_BIT, 5192 }; 5193 5194 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 5195 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 5196 5197 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 5198 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 5199 5200 struct netdev_nested_priv { 5201 unsigned char flags; 5202 void *data; 5203 }; 5204 5205 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 5206 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 5207 struct list_head **iter); 5208 5209 /* iterate through upper list, must be called under RCU read lock */ 5210 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 5211 for (iter = &(dev)->adj_list.upper, \ 5212 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 5213 updev; \ 5214 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 5215 5216 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 5217 int (*fn)(struct net_device *upper_dev, 5218 struct netdev_nested_priv *priv), 5219 struct netdev_nested_priv *priv); 5220 5221 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 5222 struct net_device *upper_dev); 5223 5224 bool netdev_has_any_upper_dev(struct net_device *dev); 5225 5226 void *netdev_lower_get_next_private(struct net_device *dev, 5227 struct list_head **iter); 5228 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 5229 struct list_head **iter); 5230 5231 #define netdev_for_each_lower_private(dev, priv, iter) \ 5232 for (iter = (dev)->adj_list.lower.next, \ 5233 priv = netdev_lower_get_next_private(dev, &(iter)); \ 5234 priv; \ 5235 priv = netdev_lower_get_next_private(dev, &(iter))) 5236 5237 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 5238 for (iter = &(dev)->adj_list.lower, \ 5239 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 5240 priv; \ 5241 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 5242 5243 void *netdev_lower_get_next(struct net_device *dev, 5244 struct list_head **iter); 5245 5246 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 5247 for (iter = (dev)->adj_list.lower.next, \ 5248 ldev = netdev_lower_get_next(dev, &(iter)); \ 5249 ldev; \ 5250 ldev = netdev_lower_get_next(dev, &(iter))) 5251 5252 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 5253 struct list_head **iter); 5254 int netdev_walk_all_lower_dev(struct net_device *dev, 5255 int (*fn)(struct net_device *lower_dev, 5256 struct netdev_nested_priv *priv), 5257 struct netdev_nested_priv *priv); 5258 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 5259 int (*fn)(struct net_device *lower_dev, 5260 struct netdev_nested_priv *priv), 5261 struct netdev_nested_priv *priv); 5262 5263 void *netdev_adjacent_get_private(struct list_head *adj_list); 5264 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 5265 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 5266 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 5267 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 5268 struct netlink_ext_ack *extack); 5269 int netdev_master_upper_dev_link(struct net_device *dev, 5270 struct net_device *upper_dev, 5271 void *upper_priv, void *upper_info, 5272 struct netlink_ext_ack *extack); 5273 void netdev_upper_dev_unlink(struct net_device *dev, 5274 struct net_device *upper_dev); 5275 int netdev_adjacent_change_prepare(struct net_device *old_dev, 5276 struct net_device *new_dev, 5277 struct net_device *dev, 5278 struct netlink_ext_ack *extack); 5279 void netdev_adjacent_change_commit(struct net_device *old_dev, 5280 struct net_device *new_dev, 5281 struct net_device *dev); 5282 void netdev_adjacent_change_abort(struct net_device *old_dev, 5283 struct net_device *new_dev, 5284 struct net_device *dev); 5285 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 5286 void *netdev_lower_dev_get_private(struct net_device *dev, 5287 struct net_device *lower_dev); 5288 void netdev_lower_state_changed(struct net_device *lower_dev, 5289 void *lower_state_info); 5290 5291 #define NETDEV_RSS_KEY_LEN 256 5292 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 5293 void netdev_rss_key_fill(void *buffer, size_t len); 5294 5295 int skb_checksum_help(struct sk_buff *skb); 5296 int skb_crc32c_csum_help(struct sk_buff *skb); 5297 int skb_csum_hwoffload_help(struct sk_buff *skb, 5298 const netdev_features_t features); 5299 5300 struct netdev_bonding_info { 5301 ifslave slave; 5302 ifbond master; 5303 }; 5304 5305 struct netdev_notifier_bonding_info { 5306 struct netdev_notifier_info info; /* must be first */ 5307 struct netdev_bonding_info bonding_info; 5308 }; 5309 5310 void netdev_bonding_info_change(struct net_device *dev, 5311 struct netdev_bonding_info *bonding_info); 5312 5313 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 5314 void ethtool_notify(struct net_device *dev, unsigned int cmd); 5315 #else 5316 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd) 5317 { 5318 } 5319 #endif 5320 5321 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 5322 5323 static inline bool can_checksum_protocol(netdev_features_t features, 5324 __be16 protocol) 5325 { 5326 if (protocol == htons(ETH_P_FCOE)) 5327 return !!(features & NETIF_F_FCOE_CRC); 5328 5329 /* Assume this is an IP checksum (not SCTP CRC) */ 5330 5331 if (features & NETIF_F_HW_CSUM) { 5332 /* Can checksum everything */ 5333 return true; 5334 } 5335 5336 switch (protocol) { 5337 case htons(ETH_P_IP): 5338 return !!(features & NETIF_F_IP_CSUM); 5339 case htons(ETH_P_IPV6): 5340 return !!(features & NETIF_F_IPV6_CSUM); 5341 default: 5342 return false; 5343 } 5344 } 5345 5346 #ifdef CONFIG_BUG 5347 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 5348 #else 5349 static inline void netdev_rx_csum_fault(struct net_device *dev, 5350 struct sk_buff *skb) 5351 { 5352 } 5353 #endif 5354 /* rx skb timestamps */ 5355 void net_enable_timestamp(void); 5356 void net_disable_timestamp(void); 5357 5358 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 5359 const struct skb_shared_hwtstamps *hwtstamps, 5360 bool cycles) 5361 { 5362 const struct net_device_ops *ops = dev->netdev_ops; 5363 5364 if (ops->ndo_get_tstamp) 5365 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 5366 5367 return hwtstamps->hwtstamp; 5368 } 5369 5370 #ifndef CONFIG_PREEMPT_RT 5371 static inline void netdev_xmit_set_more(bool more) 5372 { 5373 __this_cpu_write(softnet_data.xmit.more, more); 5374 } 5375 5376 static inline bool netdev_xmit_more(void) 5377 { 5378 return __this_cpu_read(softnet_data.xmit.more); 5379 } 5380 #else 5381 static inline void netdev_xmit_set_more(bool more) 5382 { 5383 current->net_xmit.more = more; 5384 } 5385 5386 static inline bool netdev_xmit_more(void) 5387 { 5388 return current->net_xmit.more; 5389 } 5390 #endif 5391 5392 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 5393 struct sk_buff *skb, struct net_device *dev, 5394 bool more) 5395 { 5396 netdev_xmit_set_more(more); 5397 return ops->ndo_start_xmit(skb, dev); 5398 } 5399 5400 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 5401 struct netdev_queue *txq, bool more) 5402 { 5403 const struct net_device_ops *ops = dev->netdev_ops; 5404 netdev_tx_t rc; 5405 5406 rc = __netdev_start_xmit(ops, skb, dev, more); 5407 if (rc == NETDEV_TX_OK) 5408 txq_trans_update(dev, txq); 5409 5410 return rc; 5411 } 5412 5413 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 5414 const struct ns_common *ns); 5415 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 5416 const struct ns_common *ns); 5417 5418 extern const struct kobj_ns_type_operations net_ns_type_operations; 5419 5420 const char *netdev_drivername(const struct net_device *dev); 5421 5422 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 5423 netdev_features_t f2) 5424 { 5425 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 5426 if (f1 & NETIF_F_HW_CSUM) 5427 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5428 else 5429 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5430 } 5431 5432 return f1 & f2; 5433 } 5434 5435 static inline netdev_features_t netdev_get_wanted_features( 5436 struct net_device *dev) 5437 { 5438 return (dev->features & ~dev->hw_features) | dev->wanted_features; 5439 } 5440 netdev_features_t netdev_increment_features(netdev_features_t all, 5441 netdev_features_t one, netdev_features_t mask); 5442 5443 /* Allow TSO being used on stacked device : 5444 * Performing the GSO segmentation before last device 5445 * is a performance improvement. 5446 */ 5447 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 5448 netdev_features_t mask) 5449 { 5450 return netdev_increment_features(features, NETIF_F_ALL_TSO | 5451 NETIF_F_ALL_FOR_ALL, mask); 5452 } 5453 5454 int __netdev_update_features(struct net_device *dev); 5455 void netdev_update_features(struct net_device *dev); 5456 void netdev_change_features(struct net_device *dev); 5457 void netdev_compute_master_upper_features(struct net_device *dev, bool update_header); 5458 5459 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 5460 struct net_device *dev); 5461 5462 netdev_features_t passthru_features_check(struct sk_buff *skb, 5463 struct net_device *dev, 5464 netdev_features_t features); 5465 netdev_features_t netif_skb_features(struct sk_buff *skb); 5466 void skb_warn_bad_offload(const struct sk_buff *skb); 5467 5468 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 5469 { 5470 netdev_features_t feature; 5471 5472 if (gso_type & (SKB_GSO_TCP_FIXEDID | SKB_GSO_TCP_FIXEDID_INNER)) 5473 gso_type |= __SKB_GSO_TCP_FIXEDID; 5474 5475 feature = ((netdev_features_t)gso_type << NETIF_F_GSO_SHIFT) & NETIF_F_GSO_MASK; 5476 5477 /* check flags correspondence */ 5478 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 5479 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 5480 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 5481 BUILD_BUG_ON(__SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 5482 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 5483 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 5484 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 5485 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 5486 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 5487 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 5488 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 5489 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 5490 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 5491 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 5492 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 5493 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 5494 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 5495 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 5496 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 5497 BUILD_BUG_ON(SKB_GSO_TCP_ACCECN != 5498 (NETIF_F_GSO_ACCECN >> NETIF_F_GSO_SHIFT)); 5499 5500 return (features & feature) == feature; 5501 } 5502 5503 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 5504 { 5505 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 5506 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 5507 } 5508 5509 static inline bool netif_needs_gso(struct sk_buff *skb, 5510 netdev_features_t features) 5511 { 5512 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 5513 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 5514 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 5515 } 5516 5517 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 5518 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 5519 void netif_inherit_tso_max(struct net_device *to, 5520 const struct net_device *from); 5521 5522 static inline unsigned int 5523 netif_get_gro_max_size(const struct net_device *dev, const struct sk_buff *skb) 5524 { 5525 /* pairs with WRITE_ONCE() in netif_set_gro(_ipv4)_max_size() */ 5526 return skb->protocol == htons(ETH_P_IPV6) ? 5527 READ_ONCE(dev->gro_max_size) : 5528 READ_ONCE(dev->gro_ipv4_max_size); 5529 } 5530 5531 static inline unsigned int 5532 netif_get_gso_max_size(const struct net_device *dev, const struct sk_buff *skb) 5533 { 5534 /* pairs with WRITE_ONCE() in netif_set_gso(_ipv4)_max_size() */ 5535 return skb->protocol == htons(ETH_P_IPV6) ? 5536 READ_ONCE(dev->gso_max_size) : 5537 READ_ONCE(dev->gso_ipv4_max_size); 5538 } 5539 5540 static inline bool netif_is_macsec(const struct net_device *dev) 5541 { 5542 return dev->priv_flags & IFF_MACSEC; 5543 } 5544 5545 static inline bool netif_is_macvlan(const struct net_device *dev) 5546 { 5547 return dev->priv_flags & IFF_MACVLAN; 5548 } 5549 5550 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5551 { 5552 return dev->priv_flags & IFF_MACVLAN_PORT; 5553 } 5554 5555 static inline bool netif_is_bond_master(const struct net_device *dev) 5556 { 5557 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5558 } 5559 5560 static inline bool netif_is_bond_slave(const struct net_device *dev) 5561 { 5562 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5563 } 5564 5565 static inline bool netif_supports_nofcs(struct net_device *dev) 5566 { 5567 return dev->priv_flags & IFF_SUPP_NOFCS; 5568 } 5569 5570 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5571 { 5572 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5573 } 5574 5575 static inline bool netif_is_l3_master(const struct net_device *dev) 5576 { 5577 return dev->priv_flags & IFF_L3MDEV_MASTER; 5578 } 5579 5580 static inline bool netif_is_l3_slave(const struct net_device *dev) 5581 { 5582 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5583 } 5584 5585 static inline int dev_sdif(const struct net_device *dev) 5586 { 5587 #ifdef CONFIG_NET_L3_MASTER_DEV 5588 if (netif_is_l3_slave(dev)) 5589 return dev->ifindex; 5590 #endif 5591 return 0; 5592 } 5593 5594 static inline bool netif_is_bridge_master(const struct net_device *dev) 5595 { 5596 return dev->priv_flags & IFF_EBRIDGE; 5597 } 5598 5599 static inline bool netif_is_bridge_port(const struct net_device *dev) 5600 { 5601 return dev->priv_flags & IFF_BRIDGE_PORT; 5602 } 5603 5604 static inline bool netif_is_ovs_master(const struct net_device *dev) 5605 { 5606 return dev->priv_flags & IFF_OPENVSWITCH; 5607 } 5608 5609 static inline bool netif_is_ovs_port(const struct net_device *dev) 5610 { 5611 return dev->priv_flags & IFF_OVS_DATAPATH; 5612 } 5613 5614 static inline bool netif_is_any_bridge_master(const struct net_device *dev) 5615 { 5616 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev); 5617 } 5618 5619 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5620 { 5621 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5622 } 5623 5624 static inline bool netif_is_team_master(const struct net_device *dev) 5625 { 5626 return dev->priv_flags & IFF_TEAM; 5627 } 5628 5629 static inline bool netif_is_team_port(const struct net_device *dev) 5630 { 5631 return dev->priv_flags & IFF_TEAM_PORT; 5632 } 5633 5634 static inline bool netif_is_lag_master(const struct net_device *dev) 5635 { 5636 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5637 } 5638 5639 static inline bool netif_is_lag_port(const struct net_device *dev) 5640 { 5641 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5642 } 5643 5644 bool netif_is_rxfh_configured(const struct net_device *dev); 5645 5646 static inline bool netif_is_failover(const struct net_device *dev) 5647 { 5648 return dev->priv_flags & IFF_FAILOVER; 5649 } 5650 5651 static inline bool netif_is_failover_slave(const struct net_device *dev) 5652 { 5653 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5654 } 5655 5656 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5657 static inline void netif_keep_dst(struct net_device *dev) 5658 { 5659 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5660 } 5661 5662 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5663 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5664 { 5665 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5666 return netif_is_macsec(dev); 5667 } 5668 5669 extern struct pernet_operations __net_initdata loopback_net_ops; 5670 5671 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5672 5673 /* netdev_printk helpers, similar to dev_printk */ 5674 5675 static inline const char *netdev_name(const struct net_device *dev) 5676 { 5677 if (!dev->name[0] || strchr(dev->name, '%')) 5678 return "(unnamed net_device)"; 5679 return dev->name; 5680 } 5681 5682 static inline const char *netdev_reg_state(const struct net_device *dev) 5683 { 5684 u8 reg_state = READ_ONCE(dev->reg_state); 5685 5686 switch (reg_state) { 5687 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5688 case NETREG_REGISTERED: return ""; 5689 case NETREG_UNREGISTERING: return " (unregistering)"; 5690 case NETREG_UNREGISTERED: return " (unregistered)"; 5691 case NETREG_RELEASED: return " (released)"; 5692 case NETREG_DUMMY: return " (dummy)"; 5693 } 5694 5695 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state); 5696 return " (unknown)"; 5697 } 5698 5699 #define MODULE_ALIAS_NETDEV(device) \ 5700 MODULE_ALIAS("netdev-" device) 5701 5702 /* 5703 * netdev_WARN() acts like dev_printk(), but with the key difference 5704 * of using a WARN/WARN_ON to get the message out, including the 5705 * file/line information and a backtrace. 5706 */ 5707 #define netdev_WARN(dev, format, args...) \ 5708 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5709 netdev_reg_state(dev), ##args) 5710 5711 #define netdev_WARN_ONCE(dev, format, args...) \ 5712 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5713 netdev_reg_state(dev), ##args) 5714 5715 /* 5716 * The list of packet types we will receive (as opposed to discard) 5717 * and the routines to invoke. 5718 * 5719 * Why 16. Because with 16 the only overlap we get on a hash of the 5720 * low nibble of the protocol value is RARP/SNAP/X.25. 5721 * 5722 * 0800 IP 5723 * 0001 802.3 5724 * 0002 AX.25 5725 * 0004 802.2 5726 * 8035 RARP 5727 * 0005 SNAP 5728 * 0805 X.25 5729 * 0806 ARP 5730 * 8137 IPX 5731 * 0009 Localtalk 5732 * 86DD IPv6 5733 */ 5734 #define PTYPE_HASH_SIZE (16) 5735 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5736 5737 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5738 5739 extern struct net_device *blackhole_netdev; 5740 5741 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */ 5742 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD) 5743 #define DEV_STATS_ADD(DEV, FIELD, VAL) \ 5744 atomic_long_add((VAL), &(DEV)->stats.__##FIELD) 5745 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD) 5746 5747 #endif /* _LINUX_NETDEVICE_H */ 5748